2 * iterator/iterator.c - iterative resolver DNS query response module
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 a module that performs recursive iterative DNS query
44 #include "iterator/iterator.h"
45 #include "iterator/iter_utils.h"
46 #include "iterator/iter_hints.h"
47 #include "iterator/iter_fwd.h"
48 #include "iterator/iter_donotq.h"
49 #include "iterator/iter_delegpt.h"
50 #include "iterator/iter_resptype.h"
51 #include "iterator/iter_scrub.h"
52 #include "iterator/iter_priv.h"
53 #include "validator/val_neg.h"
54 #include "services/cache/dns.h"
55 #include "services/cache/infra.h"
56 #include "services/authzone.h"
57 #include "util/module.h"
58 #include "util/netevent.h"
59 #include "util/net_help.h"
60 #include "util/regional.h"
61 #include "util/data/dname.h"
62 #include "util/data/msgencode.h"
63 #include "util/fptr_wlist.h"
64 #include "util/config_file.h"
65 #include "util/random.h"
66 #include "sldns/rrdef.h"
67 #include "sldns/wire2str.h"
68 #include "sldns/str2wire.h"
69 #include "sldns/parseutil.h"
70 #include "sldns/sbuffer.h"
73 int UNKNOWN_SERVER_NICENESS = 376;
75 int USEFUL_SERVER_TOP_TIMEOUT = 120000;
76 /* Equals USEFUL_SERVER_TOP_TIMEOUT*4 */
77 int BLACKLIST_PENALTY = (120000*4);
79 static void target_count_increase_nx(struct iter_qstate* iq, int num);
82 iter_init(struct module_env* env, int id)
84 struct iter_env* iter_env = (struct iter_env*)calloc(1,
85 sizeof(struct iter_env));
87 log_err("malloc failure");
90 env->modinfo[id] = (void*)iter_env;
92 lock_basic_init(&iter_env->queries_ratelimit_lock);
93 lock_protect(&iter_env->queries_ratelimit_lock,
94 &iter_env->num_queries_ratelimited,
95 sizeof(iter_env->num_queries_ratelimited));
97 if(!iter_apply_cfg(iter_env, env->cfg)) {
98 log_err("iterator: could not apply configuration settings.");
105 /** delete caps_whitelist element */
107 caps_free(struct rbnode_type* n, void* ATTR_UNUSED(d))
110 free(((struct name_tree_node*)n)->name);
116 iter_deinit(struct module_env* env, int id)
118 struct iter_env* iter_env;
119 if(!env || !env->modinfo[id])
121 iter_env = (struct iter_env*)env->modinfo[id];
122 lock_basic_destroy(&iter_env->queries_ratelimit_lock);
123 free(iter_env->target_fetch_policy);
124 priv_delete(iter_env->priv);
125 donotq_delete(iter_env->donotq);
126 if(iter_env->caps_white) {
127 traverse_postorder(iter_env->caps_white, caps_free, NULL);
128 free(iter_env->caps_white);
131 env->modinfo[id] = NULL;
134 /** new query for iterator */
136 iter_new(struct module_qstate* qstate, int id)
138 struct iter_qstate* iq = (struct iter_qstate*)regional_alloc(
139 qstate->region, sizeof(struct iter_qstate));
140 qstate->minfo[id] = iq;
143 memset(iq, 0, sizeof(*iq));
144 iq->state = INIT_REQUEST_STATE;
145 iq->final_state = FINISHED_STATE;
146 iq->an_prepend_list = NULL;
147 iq->an_prepend_last = NULL;
148 iq->ns_prepend_list = NULL;
149 iq->ns_prepend_last = NULL;
152 iq->num_target_queries = 0;
153 iq->num_current_queries = 0;
154 iq->query_restart_count = 0;
155 iq->referral_count = 0;
157 iq->ratelimit_ok = 0;
158 iq->target_count = NULL;
159 iq->dp_target_count = 0;
160 iq->wait_priming_stub = 0;
161 iq->refetch_glue = 0;
162 iq->dnssec_expected = 0;
163 iq->dnssec_lame_query = 0;
164 iq->chase_flags = qstate->query_flags;
165 /* Start with the (current) qname. */
166 iq->qchase = qstate->qinfo;
167 outbound_list_init(&iq->outlist);
168 iq->minimise_count = 0;
169 iq->timeout_count = 0;
170 if (qstate->env->cfg->qname_minimisation)
171 iq->minimisation_state = INIT_MINIMISE_STATE;
173 iq->minimisation_state = DONOT_MINIMISE_STATE;
175 memset(&iq->qinfo_out, 0, sizeof(struct query_info));
180 * Transition to the next state. This can be used to advance a currently
181 * processing event. It cannot be used to reactivate a forEvent.
183 * @param iq: iterator query state
184 * @param nextstate The state to transition to.
185 * @return true. This is so this can be called as the return value for the
186 * actual process*State() methods. (Transitioning to the next state
187 * implies further processing).
190 next_state(struct iter_qstate* iq, enum iter_state nextstate)
192 /* If transitioning to a "response" state, make sure that there is a
194 if(iter_state_is_responsestate(nextstate)) {
195 if(iq->response == NULL) {
196 log_err("transitioning to response state sans "
200 iq->state = nextstate;
205 * Transition an event to its final state. Final states always either return
206 * a result up the module chain, or reactivate a dependent event. Which
207 * final state to transition to is set in the module state for the event when
208 * it was created, and depends on the original purpose of the event.
210 * The response is stored in the qstate->buf buffer.
212 * @param iq: iterator query state
213 * @return false. This is so this method can be used as the return value for
214 * the processState methods. (Transitioning to the final state
217 final_state(struct iter_qstate* iq)
219 return next_state(iq, iq->final_state);
223 * Callback routine to handle errors in parent query states
224 * @param qstate: query state that failed.
225 * @param id: module id.
226 * @param super: super state.
229 error_supers(struct module_qstate* qstate, int id, struct module_qstate* super)
231 struct iter_env* ie = (struct iter_env*)qstate->env->modinfo[id];
232 struct iter_qstate* super_iq = (struct iter_qstate*)super->minfo[id];
234 if(qstate->qinfo.qtype == LDNS_RR_TYPE_A ||
235 qstate->qinfo.qtype == LDNS_RR_TYPE_AAAA) {
236 /* mark address as failed. */
237 struct delegpt_ns* dpns = NULL;
238 super_iq->num_target_queries--;
240 dpns = delegpt_find_ns(super_iq->dp,
241 qstate->qinfo.qname, qstate->qinfo.qname_len);
244 /* this can happen, for eg. qname minimisation asked
245 * for an NXDOMAIN to be validated, and used qtype
246 * A for that, and the error of that, the name, is
247 * not listed in super_iq->dp */
248 verbose(VERB_ALGO, "subq error, but not interested");
249 log_query_info(VERB_ALGO, "superq", &super->qinfo);
252 /* see if the failure did get (parent-lame) info */
253 if(!cache_fill_missing(super->env, super_iq->qchase.qclass,
254 super->region, super_iq->dp))
255 log_err("out of memory adding missing");
257 delegpt_mark_neg(dpns, qstate->qinfo.qtype);
258 if((dpns->got4 == 2 || (!ie->supports_ipv4 && !ie->use_nat64)) &&
259 (dpns->got6 == 2 || !ie->supports_ipv6)) {
260 dpns->resolved = 1; /* mark as failed */
261 target_count_increase_nx(super_iq, 1);
264 if(qstate->qinfo.qtype == LDNS_RR_TYPE_NS) {
265 /* prime failed to get delegation */
268 /* evaluate targets again */
269 super_iq->state = QUERYTARGETS_STATE;
270 /* super becomes runnable, and will process this change */
274 * Return an error to the client
275 * @param qstate: our query state
276 * @param id: module id
277 * @param rcode: error code (DNS errcode).
278 * @return: 0 for use by caller, to make notation easy, like:
279 * return error_response(..).
282 error_response(struct module_qstate* qstate, int id, int rcode)
284 verbose(VERB_QUERY, "return error response %s",
285 sldns_lookup_by_id(sldns_rcodes, rcode)?
286 sldns_lookup_by_id(sldns_rcodes, rcode)->name:"??");
287 qstate->return_rcode = rcode;
288 qstate->return_msg = NULL;
289 qstate->ext_state[id] = module_finished;
294 * Return an error to the client and cache the error code in the
295 * message cache (so per qname, qtype, qclass).
296 * @param qstate: our query state
297 * @param id: module id
298 * @param rcode: error code (DNS errcode).
299 * @return: 0 for use by caller, to make notation easy, like:
300 * return error_response(..).
303 error_response_cache(struct module_qstate* qstate, int id, int rcode)
305 struct reply_info err;
306 struct msgreply_entry* msg;
307 if(qstate->no_cache_store) {
308 return error_response(qstate, id, rcode);
310 if(qstate->prefetch_leeway > NORR_TTL) {
311 verbose(VERB_ALGO, "error response for prefetch in cache");
312 /* attempt to adjust the cache entry prefetch */
313 if(dns_cache_prefetch_adjust(qstate->env, &qstate->qinfo,
314 NORR_TTL, qstate->query_flags))
315 return error_response(qstate, id, rcode);
316 /* if that fails (not in cache), fall through to store err */
318 if((msg=msg_cache_lookup(qstate->env,
319 qstate->qinfo.qname, qstate->qinfo.qname_len,
320 qstate->qinfo.qtype, qstate->qinfo.qclass,
321 qstate->query_flags, 0,
322 qstate->env->cfg->serve_expired_ttl_reset)) != NULL) {
323 struct reply_info* rep = (struct reply_info*)msg->entry.data;
324 if(qstate->env->cfg->serve_expired &&
325 qstate->env->cfg->serve_expired_ttl_reset && rep &&
326 *qstate->env->now + qstate->env->cfg->serve_expired_ttl
327 > rep->serve_expired_ttl) {
328 verbose(VERB_ALGO, "reset serve-expired-ttl for "
329 "response in cache");
330 rep->serve_expired_ttl = *qstate->env->now +
331 qstate->env->cfg->serve_expired_ttl;
333 if(rep && (FLAGS_GET_RCODE(rep->flags) ==
334 LDNS_RCODE_NOERROR ||
335 FLAGS_GET_RCODE(rep->flags) ==
336 LDNS_RCODE_NXDOMAIN ||
337 FLAGS_GET_RCODE(rep->flags) ==
338 LDNS_RCODE_YXDOMAIN) &&
339 (qstate->env->cfg->serve_expired ||
340 *qstate->env->now <= rep->ttl)) {
341 /* we have a good entry, don't overwrite */
342 lock_rw_unlock(&msg->entry.lock);
343 return error_response(qstate, id, rcode);
345 lock_rw_unlock(&msg->entry.lock);
346 /* nothing interesting is cached (already error response or
347 * expired good record when we don't serve expired), so this
348 * servfail cache entry is useful (stops waste of time on this
349 * servfail NORR_TTL) */
352 memset(&err, 0, sizeof(err));
353 err.flags = (uint16_t)(BIT_QR | BIT_RA);
354 FLAGS_SET_RCODE(err.flags, rcode);
357 err.prefetch_ttl = PREFETCH_TTL_CALC(err.ttl);
358 err.serve_expired_ttl = NORR_TTL;
359 /* do not waste time trying to validate this servfail */
360 err.security = sec_status_indeterminate;
361 verbose(VERB_ALGO, "store error response in message cache");
362 iter_dns_store(qstate->env, &qstate->qinfo, &err, 0, 0, 0, NULL,
363 qstate->query_flags, qstate->qstarttime);
364 return error_response(qstate, id, rcode);
367 /** check if prepend item is duplicate item */
369 prepend_is_duplicate(struct ub_packed_rrset_key** sets, size_t to,
370 struct ub_packed_rrset_key* dup)
373 for(i=0; i<to; i++) {
374 if(sets[i]->rk.type == dup->rk.type &&
375 sets[i]->rk.rrset_class == dup->rk.rrset_class &&
376 sets[i]->rk.dname_len == dup->rk.dname_len &&
377 query_dname_compare(sets[i]->rk.dname, dup->rk.dname)
384 /** prepend the prepend list in the answer and authority section of dns_msg */
386 iter_prepend(struct iter_qstate* iq, struct dns_msg* msg,
387 struct regional* region)
389 struct iter_prep_list* p;
390 struct ub_packed_rrset_key** sets;
391 size_t num_an = 0, num_ns = 0;;
392 for(p = iq->an_prepend_list; p; p = p->next)
394 for(p = iq->ns_prepend_list; p; p = p->next)
396 if(num_an + num_ns == 0)
398 verbose(VERB_ALGO, "prepending %d rrsets", (int)num_an + (int)num_ns);
399 if(num_an > RR_COUNT_MAX || num_ns > RR_COUNT_MAX ||
400 msg->rep->rrset_count > RR_COUNT_MAX) return 0; /* overflow */
401 sets = regional_alloc(region, (num_an+num_ns+msg->rep->rrset_count) *
402 sizeof(struct ub_packed_rrset_key*));
407 for(p = iq->an_prepend_list; p; p = p->next) {
408 sets[num_an++] = p->rrset;
409 if(ub_packed_rrset_ttl(p->rrset) < msg->rep->ttl)
410 msg->rep->ttl = ub_packed_rrset_ttl(p->rrset);
412 memcpy(sets+num_an, msg->rep->rrsets, msg->rep->an_numrrsets *
413 sizeof(struct ub_packed_rrset_key*));
416 for(p = iq->ns_prepend_list; p; p = p->next) {
417 if(prepend_is_duplicate(sets+msg->rep->an_numrrsets+num_an,
418 num_ns, p->rrset) || prepend_is_duplicate(
419 msg->rep->rrsets+msg->rep->an_numrrsets,
420 msg->rep->ns_numrrsets, p->rrset))
422 sets[msg->rep->an_numrrsets + num_an + num_ns++] = p->rrset;
423 if(ub_packed_rrset_ttl(p->rrset) < msg->rep->ttl)
424 msg->rep->ttl = ub_packed_rrset_ttl(p->rrset);
426 memcpy(sets + num_an + msg->rep->an_numrrsets + num_ns,
427 msg->rep->rrsets + msg->rep->an_numrrsets,
428 (msg->rep->ns_numrrsets + msg->rep->ar_numrrsets) *
429 sizeof(struct ub_packed_rrset_key*));
431 /* NXDOMAIN rcode can stay if we prepended DNAME/CNAMEs, because
432 * this is what recursors should give. */
433 msg->rep->rrset_count += num_an + num_ns;
434 msg->rep->an_numrrsets += num_an;
435 msg->rep->ns_numrrsets += num_ns;
436 msg->rep->rrsets = sets;
441 * Find rrset in ANSWER prepend list.
442 * to avoid duplicate DNAMEs when a DNAME is traversed twice.
443 * @param iq: iterator query state.
444 * @param rrset: rrset to add.
445 * @return false if not found
448 iter_find_rrset_in_prepend_answer(struct iter_qstate* iq,
449 struct ub_packed_rrset_key* rrset)
451 struct iter_prep_list* p = iq->an_prepend_list;
453 if(ub_rrset_compare(p->rrset, rrset) == 0 &&
454 rrsetdata_equal((struct packed_rrset_data*)p->rrset
455 ->entry.data, (struct packed_rrset_data*)rrset
464 * Add rrset to ANSWER prepend list
465 * @param qstate: query state.
466 * @param iq: iterator query state.
467 * @param rrset: rrset to add.
468 * @return false on failure (malloc).
471 iter_add_prepend_answer(struct module_qstate* qstate, struct iter_qstate* iq,
472 struct ub_packed_rrset_key* rrset)
474 struct iter_prep_list* p = (struct iter_prep_list*)regional_alloc(
475 qstate->region, sizeof(struct iter_prep_list));
481 if(iq->an_prepend_last)
482 iq->an_prepend_last->next = p;
483 else iq->an_prepend_list = p;
484 iq->an_prepend_last = p;
489 * Add rrset to AUTHORITY prepend list
490 * @param qstate: query state.
491 * @param iq: iterator query state.
492 * @param rrset: rrset to add.
493 * @return false on failure (malloc).
496 iter_add_prepend_auth(struct module_qstate* qstate, struct iter_qstate* iq,
497 struct ub_packed_rrset_key* rrset)
499 struct iter_prep_list* p = (struct iter_prep_list*)regional_alloc(
500 qstate->region, sizeof(struct iter_prep_list));
506 if(iq->ns_prepend_last)
507 iq->ns_prepend_last->next = p;
508 else iq->ns_prepend_list = p;
509 iq->ns_prepend_last = p;
514 * Given a CNAME response (defined as a response containing a CNAME or DNAME
515 * that does not answer the request), process the response, modifying the
516 * state as necessary. This follows the CNAME/DNAME chain and returns the
519 * sets the new query name, after following the CNAME/DNAME chain.
520 * @param qstate: query state.
521 * @param iq: iterator query state.
522 * @param msg: the response.
523 * @param mname: returned target new query name.
524 * @param mname_len: length of mname.
525 * @return false on (malloc) error.
528 handle_cname_response(struct module_qstate* qstate, struct iter_qstate* iq,
529 struct dns_msg* msg, uint8_t** mname, size_t* mname_len)
532 /* Start with the (current) qname. */
533 *mname = iq->qchase.qname;
534 *mname_len = iq->qchase.qname_len;
536 /* Iterate over the ANSWER rrsets in order, looking for CNAMEs and
538 for(i=0; i<msg->rep->an_numrrsets; i++) {
539 struct ub_packed_rrset_key* r = msg->rep->rrsets[i];
540 /* If there is a (relevant) DNAME, add it to the list.
541 * We always expect there to be CNAME that was generated
542 * by this DNAME following, so we don't process the DNAME
544 if(ntohs(r->rk.type) == LDNS_RR_TYPE_DNAME &&
545 dname_strict_subdomain_c(*mname, r->rk.dname) &&
546 !iter_find_rrset_in_prepend_answer(iq, r)) {
547 if(!iter_add_prepend_answer(qstate, iq, r))
552 if(ntohs(r->rk.type) == LDNS_RR_TYPE_CNAME &&
553 query_dname_compare(*mname, r->rk.dname) == 0 &&
554 !iter_find_rrset_in_prepend_answer(iq, r)) {
555 /* Add this relevant CNAME rrset to the prepend list.*/
556 if(!iter_add_prepend_answer(qstate, iq, r))
558 get_cname_target(r, mname, mname_len);
561 /* Other rrsets in the section are ignored. */
563 /* add authority rrsets to authority prepend, for wildcarded CNAMEs */
564 for(i=msg->rep->an_numrrsets; i<msg->rep->an_numrrsets +
565 msg->rep->ns_numrrsets; i++) {
566 struct ub_packed_rrset_key* r = msg->rep->rrsets[i];
567 /* only add NSEC/NSEC3, as they may be needed for validation */
568 if(ntohs(r->rk.type) == LDNS_RR_TYPE_NSEC ||
569 ntohs(r->rk.type) == LDNS_RR_TYPE_NSEC3) {
570 if(!iter_add_prepend_auth(qstate, iq, r))
577 /** fill fail address for later recovery */
579 fill_fail_addr(struct iter_qstate* iq, struct sockaddr_storage* addr,
583 iq->fail_addr_type = 0;
586 if(((struct sockaddr_in*)addr)->sin_family == AF_INET) {
587 iq->fail_addr_type = 4;
588 memcpy(&iq->fail_addr.in,
589 &((struct sockaddr_in*)addr)->sin_addr,
590 sizeof(iq->fail_addr.in));
593 else if(((struct sockaddr_in*)addr)->sin_family == AF_INET6) {
594 iq->fail_addr_type = 6;
595 memcpy(&iq->fail_addr.in6,
596 &((struct sockaddr_in6*)addr)->sin6_addr,
597 sizeof(iq->fail_addr.in6));
601 iq->fail_addr_type = 0;
605 /** print fail addr to string */
607 print_fail_addr(struct iter_qstate* iq, char* buf, size_t len)
609 if(iq->fail_addr_type == 4) {
610 if(inet_ntop(AF_INET, &iq->fail_addr.in, buf,
611 (socklen_t)len) == 0)
612 (void)strlcpy(buf, "(inet_ntop error)", len);
615 else if(iq->fail_addr_type == 6) {
616 if(inet_ntop(AF_INET6, &iq->fail_addr.in6, buf,
617 (socklen_t)len) == 0)
618 (void)strlcpy(buf, "(inet_ntop error)", len);
622 (void)strlcpy(buf, "", len);
625 /** add response specific error information for log servfail */
627 errinf_reply(struct module_qstate* qstate, struct iter_qstate* iq)
629 if(qstate->env->cfg->val_log_level < 2 && !qstate->env->cfg->log_servfail)
631 if((qstate->reply && qstate->reply->remote_addrlen != 0) ||
632 (iq->fail_addr_type != 0)) {
633 char from[256], frm[512];
634 if(qstate->reply && qstate->reply->remote_addrlen != 0)
635 addr_to_str(&qstate->reply->remote_addr,
636 qstate->reply->remote_addrlen, from,
639 print_fail_addr(iq, from, sizeof(from));
640 snprintf(frm, sizeof(frm), "from %s", from);
643 if(iq->scrub_failures || iq->parse_failures) {
644 if(iq->scrub_failures)
645 errinf(qstate, "upstream response failed scrub");
646 if(iq->parse_failures)
647 errinf(qstate, "could not parse upstream response");
648 } else if(iq->response == NULL && iq->timeout_count != 0) {
649 errinf(qstate, "upstream server timeout");
650 } else if(iq->response == NULL) {
651 errinf(qstate, "no server to query");
653 if(iq->dp->target_list == NULL)
654 errinf(qstate, "no addresses for nameservers");
655 else errinf(qstate, "nameserver addresses not usable");
656 if(iq->dp->nslist == NULL)
657 errinf(qstate, "have no nameserver names");
659 errinf(qstate, "NS record was dnssec bogus");
662 if(iq->response && iq->response->rep) {
663 if(FLAGS_GET_RCODE(iq->response->rep->flags) != 0) {
664 char rcode[256], rc[32];
665 (void)sldns_wire2str_rcode_buf(
666 FLAGS_GET_RCODE(iq->response->rep->flags),
668 snprintf(rcode, sizeof(rcode), "got %s", rc);
669 errinf(qstate, rcode);
672 if(iq->response->rep->an_numrrsets == 0) {
673 errinf(qstate, "nodata answer");
679 /** see if last resort is possible - does config allow queries to parent */
681 can_have_last_resort(struct module_env* env, uint8_t* nm, size_t nmlen,
682 uint16_t qclass, struct delegpt** retdp)
684 struct delegpt* fwddp;
685 struct iter_hints_stub* stub;
686 int labs = dname_count_labels(nm);
687 /* do not process a last resort (the parent side) if a stub
688 * or forward is configured, because we do not want to go 'above'
689 * the configured servers */
690 if(!dname_is_root(nm) && (stub = (struct iter_hints_stub*)
691 name_tree_find(&env->hints->tree, nm, nmlen, labs, qclass)) &&
692 /* has_parent side is turned off for stub_first, where we
693 * are allowed to go to the parent */
694 stub->dp->has_parent_side_NS) {
695 if(retdp) *retdp = stub->dp;
698 if((fwddp = forwards_find(env->fwds, nm, qclass)) &&
699 /* has_parent_side is turned off for forward_first, where
700 * we are allowed to go to the parent */
701 fwddp->has_parent_side_NS) {
702 if(retdp) *retdp = fwddp;
708 /** see if target name is caps-for-id whitelisted */
710 is_caps_whitelisted(struct iter_env* ie, struct iter_qstate* iq)
712 if(!ie->caps_white) return 0; /* no whitelist, or no capsforid */
713 return name_tree_lookup(ie->caps_white, iq->qchase.qname,
714 iq->qchase.qname_len, dname_count_labels(iq->qchase.qname),
715 iq->qchase.qclass) != NULL;
719 * Create target count structure for this query. This is always explicitly
720 * created for the parent query.
723 target_count_create(struct iter_qstate* iq)
725 if(!iq->target_count) {
726 iq->target_count = (int*)calloc(TARGET_COUNT_MAX, sizeof(int));
727 /* if calloc fails we simply do not track this number */
728 if(iq->target_count) {
729 iq->target_count[TARGET_COUNT_REF] = 1;
730 iq->nxns_dp = (uint8_t**)calloc(1, sizeof(uint8_t*));
736 target_count_increase(struct iter_qstate* iq, int num)
738 target_count_create(iq);
740 iq->target_count[TARGET_COUNT_QUERIES] += num;
741 iq->dp_target_count++;
745 target_count_increase_nx(struct iter_qstate* iq, int num)
747 target_count_create(iq);
749 iq->target_count[TARGET_COUNT_NX] += num;
753 * Generate a subrequest.
754 * Generate a local request event. Local events are tied to this module, and
755 * have a corresponding (first tier) event that is waiting for this event to
756 * resolve to continue.
758 * @param qname The query name for this request.
759 * @param qnamelen length of qname
760 * @param qtype The query type for this request.
761 * @param qclass The query class for this request.
762 * @param qstate The event that is generating this event.
763 * @param id: module id.
764 * @param iq: The iterator state that is generating this event.
765 * @param initial_state The initial response state (normally this
766 * is QUERY_RESP_STATE, unless it is known that the request won't
767 * need iterative processing
768 * @param finalstate The final state for the response to this request.
769 * @param subq_ret: if newly allocated, the subquerystate, or NULL if it does
770 * not need initialisation.
771 * @param v: if true, validation is done on the subquery.
772 * @param detached: true if this qstate should not attach to the subquery
773 * @return false on error (malloc).
776 generate_sub_request(uint8_t* qname, size_t qnamelen, uint16_t qtype,
777 uint16_t qclass, struct module_qstate* qstate, int id,
778 struct iter_qstate* iq, enum iter_state initial_state,
779 enum iter_state finalstate, struct module_qstate** subq_ret, int v,
782 struct module_qstate* subq = NULL;
783 struct iter_qstate* subiq = NULL;
784 uint16_t qflags = 0; /* OPCODE QUERY, no flags */
785 struct query_info qinf;
786 int prime = (finalstate == PRIME_RESP_STATE)?1:0;
789 qinf.qname_len = qnamelen;
791 qinf.qclass = qclass;
792 qinf.local_alias = NULL;
794 /* RD should be set only when sending the query back through the INIT
796 if(initial_state == INIT_REQUEST_STATE)
798 /* We set the CD flag so we can send this through the "head" of
799 * the resolution chain, which might have a validator. We are
800 * uninterested in validating things not on the direct resolution
808 struct mesh_state* sub = NULL;
809 fptr_ok(fptr_whitelist_modenv_add_sub(
810 qstate->env->add_sub));
811 if(!(*qstate->env->add_sub)(qstate, &qinf,
812 qflags, prime, valrec, &subq, &sub)){
817 /* attach subquery, lookup existing or make a new one */
818 fptr_ok(fptr_whitelist_modenv_attach_sub(
819 qstate->env->attach_sub));
820 if(!(*qstate->env->attach_sub)(qstate, &qinf, qflags, prime,
827 /* initialise the new subquery */
829 subq->ext_state[id] = module_state_initial;
830 subq->minfo[id] = regional_alloc(subq->region,
831 sizeof(struct iter_qstate));
832 if(!subq->minfo[id]) {
833 log_err("init subq: out of memory");
834 fptr_ok(fptr_whitelist_modenv_kill_sub(
835 qstate->env->kill_sub));
836 (*qstate->env->kill_sub)(subq);
839 subiq = (struct iter_qstate*)subq->minfo[id];
840 memset(subiq, 0, sizeof(*subiq));
841 subiq->num_target_queries = 0;
842 target_count_create(iq);
843 subiq->target_count = iq->target_count;
844 if(iq->target_count) {
845 iq->target_count[TARGET_COUNT_REF] ++; /* extra reference */
846 subiq->nxns_dp = iq->nxns_dp;
848 subiq->dp_target_count = 0;
849 subiq->num_current_queries = 0;
850 subiq->depth = iq->depth+1;
851 outbound_list_init(&subiq->outlist);
852 subiq->state = initial_state;
853 subiq->final_state = finalstate;
854 subiq->qchase = subq->qinfo;
855 subiq->chase_flags = subq->query_flags;
856 subiq->refetch_glue = 0;
857 if(qstate->env->cfg->qname_minimisation)
858 subiq->minimisation_state = INIT_MINIMISE_STATE;
860 subiq->minimisation_state = DONOT_MINIMISE_STATE;
861 memset(&subiq->qinfo_out, 0, sizeof(struct query_info));
867 * Generate and send a root priming request.
868 * @param qstate: the qtstate that triggered the need to prime.
869 * @param iq: iterator query state.
870 * @param id: module id.
871 * @param qclass: the class to prime.
872 * @return 0 on failure
875 prime_root(struct module_qstate* qstate, struct iter_qstate* iq, int id,
879 struct module_qstate* subq;
880 verbose(VERB_DETAIL, "priming . %s NS",
881 sldns_lookup_by_id(sldns_rr_classes, (int)qclass)?
882 sldns_lookup_by_id(sldns_rr_classes, (int)qclass)->name:"??");
883 dp = hints_lookup_root(qstate->env->hints, qclass);
885 verbose(VERB_ALGO, "Cannot prime due to lack of hints");
888 /* Priming requests start at the QUERYTARGETS state, skipping
889 * the normal INIT state logic (which would cause an infloop). */
890 if(!generate_sub_request((uint8_t*)"\000", 1, LDNS_RR_TYPE_NS,
891 qclass, qstate, id, iq, QUERYTARGETS_STATE, PRIME_RESP_STATE,
893 verbose(VERB_ALGO, "could not prime root");
897 struct iter_qstate* subiq =
898 (struct iter_qstate*)subq->minfo[id];
899 /* Set the initial delegation point to the hint.
900 * copy dp, it is now part of the root prime query.
901 * dp was part of in the fixed hints structure. */
902 subiq->dp = delegpt_copy(dp, subq->region);
904 log_err("out of memory priming root, copydp");
905 fptr_ok(fptr_whitelist_modenv_kill_sub(
906 qstate->env->kill_sub));
907 (*qstate->env->kill_sub)(subq);
910 /* there should not be any target queries. */
911 subiq->num_target_queries = 0;
912 subiq->dnssec_expected = iter_indicates_dnssec(
913 qstate->env, subiq->dp, NULL, subq->qinfo.qclass);
916 /* this module stops, our submodule starts, and does the query. */
917 qstate->ext_state[id] = module_wait_subquery;
922 * Generate and process a stub priming request. This method tests for the
923 * need to prime a stub zone, so it is safe to call for every request.
925 * @param qstate: the qtstate that triggered the need to prime.
926 * @param iq: iterator query state.
927 * @param id: module id.
928 * @param qname: request name.
929 * @param qclass: request class.
930 * @return true if a priming subrequest was made, false if not. The will only
931 * issue a priming request if it detects an unprimed stub.
932 * Uses value of 2 to signal during stub-prime in root-prime situation
933 * that a noprime-stub is available and resolution can continue.
936 prime_stub(struct module_qstate* qstate, struct iter_qstate* iq, int id,
937 uint8_t* qname, uint16_t qclass)
939 /* Lookup the stub hint. This will return null if the stub doesn't
940 * need to be re-primed. */
941 struct iter_hints_stub* stub;
942 struct delegpt* stub_dp;
943 struct module_qstate* subq;
946 stub = hints_lookup_stub(qstate->env->hints, qname, qclass, iq->dp);
947 /* The stub (if there is one) does not need priming. */
951 /* if we have an auth_zone dp, and stub is equal, don't prime stub
952 * yet, unless we want to fallback and avoid the auth_zone */
953 if(!iq->auth_zone_avoid && iq->dp && iq->dp->auth_dp &&
954 query_dname_compare(iq->dp->name, stub_dp->name) == 0)
957 /* is it a noprime stub (always use) */
960 if(iq->dp == NULL) r = 2;
961 /* copy the dp out of the fixed hints structure, so that
962 * it can be changed when servicing this query */
963 iq->dp = delegpt_copy(stub_dp, qstate->region);
965 log_err("out of memory priming stub");
966 errinf(qstate, "malloc failure, priming stub");
967 (void)error_response(qstate, id, LDNS_RCODE_SERVFAIL);
968 return 1; /* return 1 to make module stop, with error */
970 log_nametypeclass(VERB_DETAIL, "use stub", stub_dp->name,
971 LDNS_RR_TYPE_NS, qclass);
975 /* Otherwise, we need to (re)prime the stub. */
976 log_nametypeclass(VERB_DETAIL, "priming stub", stub_dp->name,
977 LDNS_RR_TYPE_NS, qclass);
979 /* Stub priming events start at the QUERYTARGETS state to avoid the
980 * redundant INIT state processing. */
981 if(!generate_sub_request(stub_dp->name, stub_dp->namelen,
982 LDNS_RR_TYPE_NS, qclass, qstate, id, iq,
983 QUERYTARGETS_STATE, PRIME_RESP_STATE, &subq, 0, 0)) {
984 verbose(VERB_ALGO, "could not prime stub");
985 errinf(qstate, "could not generate lookup for stub prime");
986 (void)error_response(qstate, id, LDNS_RCODE_SERVFAIL);
987 return 1; /* return 1 to make module stop, with error */
990 struct iter_qstate* subiq =
991 (struct iter_qstate*)subq->minfo[id];
993 /* Set the initial delegation point to the hint. */
994 /* make copy to avoid use of stub dp by different qs/threads */
995 subiq->dp = delegpt_copy(stub_dp, subq->region);
997 log_err("out of memory priming stub, copydp");
998 fptr_ok(fptr_whitelist_modenv_kill_sub(
999 qstate->env->kill_sub));
1000 (*qstate->env->kill_sub)(subq);
1001 errinf(qstate, "malloc failure, in stub prime");
1002 (void)error_response(qstate, id, LDNS_RCODE_SERVFAIL);
1003 return 1; /* return 1 to make module stop, with error */
1005 /* there should not be any target queries -- although there
1006 * wouldn't be anyway, since stub hints never have
1007 * missing targets. */
1008 subiq->num_target_queries = 0;
1009 subiq->wait_priming_stub = 1;
1010 subiq->dnssec_expected = iter_indicates_dnssec(
1011 qstate->env, subiq->dp, NULL, subq->qinfo.qclass);
1014 /* this module stops, our submodule starts, and does the query. */
1015 qstate->ext_state[id] = module_wait_subquery;
1020 * Generate a delegation point for an auth zone (unless cached dp is better)
1021 * false on alloc failure.
1024 auth_zone_delegpt(struct module_qstate* qstate, struct iter_qstate* iq,
1025 uint8_t* delname, size_t delnamelen)
1027 struct auth_zone* z;
1028 if(iq->auth_zone_avoid)
1031 delname = iq->qchase.qname;
1032 delnamelen = iq->qchase.qname_len;
1034 lock_rw_rdlock(&qstate->env->auth_zones->lock);
1035 z = auth_zones_find_zone(qstate->env->auth_zones, delname, delnamelen,
1036 qstate->qinfo.qclass);
1038 lock_rw_unlock(&qstate->env->auth_zones->lock);
1041 lock_rw_rdlock(&z->lock);
1042 lock_rw_unlock(&qstate->env->auth_zones->lock);
1043 if(z->for_upstream) {
1044 if(iq->dp && query_dname_compare(z->name, iq->dp->name) == 0
1045 && iq->dp->auth_dp && qstate->blacklist &&
1046 z->fallback_enabled) {
1047 /* cache is blacklisted and fallback, and we
1048 * already have an auth_zone dp */
1049 if(verbosity>=VERB_ALGO) {
1051 dname_str(z->name, buf);
1052 verbose(VERB_ALGO, "auth_zone %s "
1053 "fallback because cache blacklisted",
1056 lock_rw_unlock(&z->lock);
1060 if(iq->dp==NULL || dname_subdomain_c(z->name, iq->dp->name)) {
1062 if(qstate->blacklist && z->fallback_enabled) {
1063 /* cache is blacklisted because of a DNSSEC
1064 * validation failure, and the zone allows
1065 * fallback to the internet, query there. */
1066 if(verbosity>=VERB_ALGO) {
1068 dname_str(z->name, buf);
1069 verbose(VERB_ALGO, "auth_zone %s "
1070 "fallback because cache blacklisted",
1073 lock_rw_unlock(&z->lock);
1076 dp = (struct delegpt*)regional_alloc_zero(
1077 qstate->region, sizeof(*dp));
1079 log_err("alloc failure");
1080 if(z->fallback_enabled) {
1081 lock_rw_unlock(&z->lock);
1082 return 1; /* just fallback */
1084 lock_rw_unlock(&z->lock);
1085 errinf(qstate, "malloc failure");
1088 dp->name = regional_alloc_init(qstate->region,
1089 z->name, z->namelen);
1091 log_err("alloc failure");
1092 if(z->fallback_enabled) {
1093 lock_rw_unlock(&z->lock);
1094 return 1; /* just fallback */
1096 lock_rw_unlock(&z->lock);
1097 errinf(qstate, "malloc failure");
1100 dp->namelen = z->namelen;
1101 dp->namelabs = z->namelabs;
1107 lock_rw_unlock(&z->lock);
1112 * Generate A and AAAA checks for glue that is in-zone for the referral
1113 * we just got to obtain authoritative information on the addresses.
1115 * @param qstate: the qtstate that triggered the need to prime.
1116 * @param iq: iterator query state.
1117 * @param id: module id.
1120 generate_a_aaaa_check(struct module_qstate* qstate, struct iter_qstate* iq,
1123 struct iter_env* ie = (struct iter_env*)qstate->env->modinfo[id];
1124 struct module_qstate* subq;
1126 struct reply_info* rep = iq->response->rep;
1127 struct ub_packed_rrset_key* s;
1130 if(iq->depth == ie->max_dependency_depth)
1132 /* walk through additional, and check if in-zone,
1133 * only relevant A, AAAA are left after scrub anyway */
1134 for(i=rep->an_numrrsets+rep->ns_numrrsets; i<rep->rrset_count; i++) {
1136 /* check *ALL* addresses that are transmitted in additional*/
1137 /* is it an address ? */
1138 if( !(ntohs(s->rk.type)==LDNS_RR_TYPE_A ||
1139 ntohs(s->rk.type)==LDNS_RR_TYPE_AAAA)) {
1142 /* is this query the same as the A/AAAA check for it */
1143 if(qstate->qinfo.qtype == ntohs(s->rk.type) &&
1144 qstate->qinfo.qclass == ntohs(s->rk.rrset_class) &&
1145 query_dname_compare(qstate->qinfo.qname,
1147 (qstate->query_flags&BIT_RD) &&
1148 !(qstate->query_flags&BIT_CD))
1151 /* generate subrequest for it */
1152 log_nametypeclass(VERB_ALGO, "schedule addr fetch",
1153 s->rk.dname, ntohs(s->rk.type),
1154 ntohs(s->rk.rrset_class));
1155 if(!generate_sub_request(s->rk.dname, s->rk.dname_len,
1156 ntohs(s->rk.type), ntohs(s->rk.rrset_class),
1158 INIT_REQUEST_STATE, FINISHED_STATE, &subq, 1, 0)) {
1159 verbose(VERB_ALGO, "could not generate addr check");
1162 /* ignore subq - not need for more init */
1167 * Generate a NS check request to obtain authoritative information
1170 * @param qstate: the qstate that triggered the need to prime.
1171 * @param iq: iterator query state.
1172 * @param id: module id.
1175 generate_ns_check(struct module_qstate* qstate, struct iter_qstate* iq, int id)
1177 struct iter_env* ie = (struct iter_env*)qstate->env->modinfo[id];
1178 struct module_qstate* subq;
1181 if(iq->depth == ie->max_dependency_depth)
1183 if(!can_have_last_resort(qstate->env, iq->dp->name, iq->dp->namelen,
1184 iq->qchase.qclass, NULL))
1186 /* is this query the same as the nscheck? */
1187 if(qstate->qinfo.qtype == LDNS_RR_TYPE_NS &&
1188 query_dname_compare(iq->dp->name, qstate->qinfo.qname)==0 &&
1189 (qstate->query_flags&BIT_RD) && !(qstate->query_flags&BIT_CD)){
1190 /* spawn off A, AAAA queries for in-zone glue to check */
1191 generate_a_aaaa_check(qstate, iq, id);
1194 /* no need to get the NS record for DS, it is above the zonecut */
1195 if(qstate->qinfo.qtype == LDNS_RR_TYPE_DS)
1198 log_nametypeclass(VERB_ALGO, "schedule ns fetch",
1199 iq->dp->name, LDNS_RR_TYPE_NS, iq->qchase.qclass);
1200 if(!generate_sub_request(iq->dp->name, iq->dp->namelen,
1201 LDNS_RR_TYPE_NS, iq->qchase.qclass, qstate, id, iq,
1202 INIT_REQUEST_STATE, FINISHED_STATE, &subq, 1, 0)) {
1203 verbose(VERB_ALGO, "could not generate ns check");
1207 struct iter_qstate* subiq =
1208 (struct iter_qstate*)subq->minfo[id];
1210 /* make copy to avoid use of stub dp by different qs/threads */
1211 /* refetch glue to start higher up the tree */
1212 subiq->refetch_glue = 1;
1213 subiq->dp = delegpt_copy(iq->dp, subq->region);
1215 log_err("out of memory generating ns check, copydp");
1216 fptr_ok(fptr_whitelist_modenv_kill_sub(
1217 qstate->env->kill_sub));
1218 (*qstate->env->kill_sub)(subq);
1225 * Generate a DNSKEY prefetch query to get the DNSKEY for the DS record we
1226 * just got in a referral (where we have dnssec_expected, thus have trust
1227 * anchors above it). Note that right after calling this routine the
1228 * iterator detached subqueries (because of following the referral), and thus
1229 * the DNSKEY query becomes detached, its return stored in the cache for
1230 * later lookup by the validator. This cache lookup by the validator avoids
1231 * the roundtrip incurred by the DNSKEY query. The DNSKEY query is now
1232 * performed at about the same time the original query is sent to the domain,
1233 * thus the two answers are likely to be returned at about the same time,
1234 * saving a roundtrip from the validated lookup.
1236 * @param qstate: the qtstate that triggered the need to prime.
1237 * @param iq: iterator query state.
1238 * @param id: module id.
1241 generate_dnskey_prefetch(struct module_qstate* qstate,
1242 struct iter_qstate* iq, int id)
1244 struct module_qstate* subq;
1247 /* is this query the same as the prefetch? */
1248 if(qstate->qinfo.qtype == LDNS_RR_TYPE_DNSKEY &&
1249 query_dname_compare(iq->dp->name, qstate->qinfo.qname)==0 &&
1250 (qstate->query_flags&BIT_RD) && !(qstate->query_flags&BIT_CD)){
1253 /* we do not generate this prefetch when the query list is full,
1254 * the query is fetched, if needed, when the validator wants it.
1255 * At that time the validator waits for it, after spawning it.
1256 * This means there is one state that uses cpu and a socket, the
1257 * spawned while this one waits, and not several at the same time,
1258 * if we had created the lookup here. And this helps to keep
1259 * the total load down, but the query still succeeds to resolve. */
1260 if(mesh_jostle_exceeded(qstate->env->mesh))
1263 /* if the DNSKEY is in the cache this lookup will stop quickly */
1264 log_nametypeclass(VERB_ALGO, "schedule dnskey prefetch",
1265 iq->dp->name, LDNS_RR_TYPE_DNSKEY, iq->qchase.qclass);
1266 if(!generate_sub_request(iq->dp->name, iq->dp->namelen,
1267 LDNS_RR_TYPE_DNSKEY, iq->qchase.qclass, qstate, id, iq,
1268 INIT_REQUEST_STATE, FINISHED_STATE, &subq, 0, 0)) {
1269 /* we'll be slower, but it'll work */
1270 verbose(VERB_ALGO, "could not generate dnskey prefetch");
1274 struct iter_qstate* subiq =
1275 (struct iter_qstate*)subq->minfo[id];
1276 /* this qstate has the right delegation for the dnskey lookup*/
1277 /* make copy to avoid use of stub dp by different qs/threads */
1278 subiq->dp = delegpt_copy(iq->dp, subq->region);
1279 /* if !subiq->dp, it'll start from the cache, no problem */
1284 * See if the query needs forwarding.
1286 * @param qstate: query state.
1287 * @param iq: iterator query state.
1288 * @return true if the request is forwarded, false if not.
1289 * If returns true but, iq->dp is NULL then a malloc failure occurred.
1292 forward_request(struct module_qstate* qstate, struct iter_qstate* iq)
1295 uint8_t* delname = iq->qchase.qname;
1296 size_t delnamelen = iq->qchase.qname_len;
1297 if(iq->refetch_glue && iq->dp) {
1298 delname = iq->dp->name;
1299 delnamelen = iq->dp->namelen;
1301 /* strip one label off of DS query to lookup higher for it */
1302 if( (iq->qchase.qtype == LDNS_RR_TYPE_DS || iq->refetch_glue)
1303 && !dname_is_root(iq->qchase.qname))
1304 dname_remove_label(&delname, &delnamelen);
1305 dp = forwards_lookup(qstate->env->fwds, delname, iq->qchase.qclass);
1308 /* send recursion desired to forward addr */
1309 iq->chase_flags |= BIT_RD;
1310 iq->dp = delegpt_copy(dp, qstate->region);
1311 /* iq->dp checked by caller */
1312 verbose(VERB_ALGO, "forwarding request");
1317 * Process the initial part of the request handling. This state roughly
1318 * corresponds to resolver algorithms steps 1 (find answer in cache) and 2
1319 * (find the best servers to ask).
1321 * Note that all requests start here, and query restarts revisit this state.
1323 * This state either generates: 1) a response, from cache or error, 2) a
1324 * priming event, or 3) forwards the request to the next state (init2,
1327 * @param qstate: query state.
1328 * @param iq: iterator query state.
1329 * @param ie: iterator shared global environment.
1330 * @param id: module id.
1331 * @return true if the event needs more request processing immediately,
1335 processInitRequest(struct module_qstate* qstate, struct iter_qstate* iq,
1336 struct iter_env* ie, int id)
1338 uint8_t* delname, *dpname=NULL;
1339 size_t delnamelen, dpnamelen=0;
1340 struct dns_msg* msg = NULL;
1342 log_query_info(VERB_DETAIL, "resolving", &qstate->qinfo);
1345 /* We enforce a maximum number of query restarts. This is primarily a
1346 * cheap way to prevent CNAME loops. */
1347 if(iq->query_restart_count > ie->max_query_restarts) {
1348 verbose(VERB_QUERY, "request has exceeded the maximum number"
1349 " of query restarts with %d", iq->query_restart_count);
1350 errinf(qstate, "request has exceeded the maximum number "
1351 "restarts (eg. indirections)");
1352 if(iq->qchase.qname)
1353 errinf_dname(qstate, "stop at", iq->qchase.qname);
1354 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
1357 /* We enforce a maximum recursion/dependency depth -- in general,
1358 * this is unnecessary for dependency loops (although it will
1359 * catch those), but it provides a sensible limit to the amount
1360 * of work required to answer a given query. */
1361 verbose(VERB_ALGO, "request has dependency depth of %d", iq->depth);
1362 if(iq->depth > ie->max_dependency_depth) {
1363 verbose(VERB_QUERY, "request has exceeded the maximum "
1364 "dependency depth with depth of %d", iq->depth);
1365 errinf(qstate, "request has exceeded the maximum dependency "
1366 "depth (eg. nameserver lookup recursion)");
1367 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
1370 /* If the request is qclass=ANY, setup to generate each class */
1371 if(qstate->qinfo.qclass == LDNS_RR_CLASS_ANY) {
1372 iq->qchase.qclass = 0;
1373 return next_state(iq, COLLECT_CLASS_STATE);
1377 * If we are restricted by a forward-zone or a stub-zone, we
1378 * can't re-fetch glue for this delegation point.
1379 * we won’t try to re-fetch glue if the iq->dp is null.
1381 if (iq->refetch_glue &&
1383 !can_have_last_resort(qstate->env, iq->dp->name,
1384 iq->dp->namelen, iq->qchase.qclass, NULL)) {
1385 iq->refetch_glue = 0;
1388 /* Resolver Algorithm Step 1 -- Look for the answer in local data. */
1390 /* This either results in a query restart (CNAME cache response), a
1391 * terminating response (ANSWER), or a cache miss (null). */
1393 if (iter_stub_fwd_no_cache(qstate, &iq->qchase, &dpname, &dpnamelen)) {
1394 /* Asked to not query cache. */
1395 verbose(VERB_ALGO, "no-cache set, going to the network");
1396 qstate->no_cache_lookup = 1;
1397 qstate->no_cache_store = 1;
1399 } else if(qstate->blacklist) {
1400 /* if cache, or anything else, was blacklisted then
1401 * getting older results from cache is a bad idea, no cache */
1402 verbose(VERB_ALGO, "cache blacklisted, going to the network");
1404 } else if(!qstate->no_cache_lookup) {
1405 msg = dns_cache_lookup(qstate->env, iq->qchase.qname,
1406 iq->qchase.qname_len, iq->qchase.qtype,
1407 iq->qchase.qclass, qstate->query_flags,
1408 qstate->region, qstate->env->scratch, 0, dpname,
1410 if(!msg && qstate->env->neg_cache &&
1411 iter_qname_indicates_dnssec(qstate->env, &iq->qchase)) {
1412 /* lookup in negative cache; may result in
1413 * NOERROR/NODATA or NXDOMAIN answers that need validation */
1414 msg = val_neg_getmsg(qstate->env->neg_cache, &iq->qchase,
1415 qstate->region, qstate->env->rrset_cache,
1416 qstate->env->scratch_buffer,
1417 *qstate->env->now, 1/*add SOA*/, NULL,
1420 /* item taken from cache does not match our query name, thus
1421 * security needs to be re-examined later */
1422 if(msg && query_dname_compare(qstate->qinfo.qname,
1423 iq->qchase.qname) != 0)
1424 msg->rep->security = sec_status_unchecked;
1427 /* handle positive cache response */
1428 enum response_type type = response_type_from_cache(msg,
1430 if(verbosity >= VERB_ALGO) {
1431 log_dns_msg("msg from cache lookup", &msg->qinfo,
1433 verbose(VERB_ALGO, "msg ttl is %d, prefetch ttl %d",
1435 (int)msg->rep->prefetch_ttl);
1438 if(type == RESPONSE_TYPE_CNAME) {
1441 verbose(VERB_ALGO, "returning CNAME response from "
1443 if(!handle_cname_response(qstate, iq, msg,
1445 errinf(qstate, "failed to prepend CNAME "
1446 "components, malloc failure");
1447 return error_response(qstate, id,
1448 LDNS_RCODE_SERVFAIL);
1450 iq->qchase.qname = sname;
1451 iq->qchase.qname_len = slen;
1452 /* This *is* a query restart, even if it is a cheap
1455 iq->refetch_glue = 0;
1456 iq->query_restart_count++;
1458 iq->dp_target_count = 0;
1459 sock_list_insert(&qstate->reply_origin, NULL, 0, qstate->region);
1460 if(qstate->env->cfg->qname_minimisation)
1461 iq->minimisation_state = INIT_MINIMISE_STATE;
1462 return next_state(iq, INIT_REQUEST_STATE);
1465 /* if from cache, NULL, else insert 'cache IP' len=0 */
1466 if(qstate->reply_origin)
1467 sock_list_insert(&qstate->reply_origin, NULL, 0, qstate->region);
1468 if(FLAGS_GET_RCODE(msg->rep->flags) == LDNS_RCODE_SERVFAIL)
1469 errinf(qstate, "SERVFAIL in cache");
1470 /* it is an answer, response, to final state */
1471 verbose(VERB_ALGO, "returning answer from cache.");
1473 return final_state(iq);
1476 /* attempt to forward the request */
1477 if(forward_request(qstate, iq))
1480 log_err("alloc failure for forward dp");
1481 errinf(qstate, "malloc failure for forward zone");
1482 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
1484 if((qstate->query_flags&BIT_RD)==0) {
1485 /* If the server accepts RD=0 queries and forwards
1486 * with RD=1, then if the server is listed as an NS
1487 * entry, it starts query loops. Stop that loop by
1488 * disallowing the query. The RD=0 was previously used
1489 * to check the cache with allow_snoop. For stubs,
1490 * the iterator pass would have primed the stub and
1491 * then cached information can be used for further
1493 verbose(VERB_ALGO, "cannot forward RD=0 query, to stop query loops");
1494 errinf(qstate, "cannot forward RD=0 query");
1495 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
1497 iq->refetch_glue = 0;
1498 iq->minimisation_state = DONOT_MINIMISE_STATE;
1499 /* the request has been forwarded.
1500 * forwarded requests need to be immediately sent to the
1501 * next state, QUERYTARGETS. */
1502 return next_state(iq, QUERYTARGETS_STATE);
1505 /* Resolver Algorithm Step 2 -- find the "best" servers. */
1507 /* first, adjust for DS queries. To avoid the grandparent problem,
1508 * we just look for the closest set of server to the parent of qname.
1509 * When re-fetching glue we also need to ask the parent.
1511 if(iq->refetch_glue) {
1513 log_err("internal or malloc fail: no dp for refetch");
1514 errinf(qstate, "malloc failure, for delegation info");
1515 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
1517 delname = iq->dp->name;
1518 delnamelen = iq->dp->namelen;
1520 delname = iq->qchase.qname;
1521 delnamelen = iq->qchase.qname_len;
1523 if(iq->qchase.qtype == LDNS_RR_TYPE_DS || iq->refetch_glue ||
1524 (iq->qchase.qtype == LDNS_RR_TYPE_NS && qstate->prefetch_leeway
1525 && can_have_last_resort(qstate->env, delname, delnamelen, iq->qchase.qclass, NULL))) {
1526 /* remove first label from delname, root goes to hints,
1527 * but only to fetch glue, not for qtype=DS. */
1528 /* also when prefetching an NS record, fetch it again from
1529 * its parent, just as if it expired, so that you do not
1530 * get stuck on an older nameserver that gives old NSrecords */
1531 if(dname_is_root(delname) && (iq->refetch_glue ||
1532 (iq->qchase.qtype == LDNS_RR_TYPE_NS &&
1533 qstate->prefetch_leeway)))
1534 delname = NULL; /* go to root priming */
1535 else dname_remove_label(&delname, &delnamelen);
1537 /* delname is the name to lookup a delegation for. If NULL rootprime */
1540 /* Lookup the delegation in the cache. If null, then the
1541 * cache needs to be primed for the qclass. */
1543 iq->dp = dns_cache_find_delegation(qstate->env, delname,
1544 delnamelen, iq->qchase.qtype, iq->qchase.qclass,
1545 qstate->region, &iq->deleg_msg,
1546 *qstate->env->now+qstate->prefetch_leeway, 1,
1550 /* If the cache has returned nothing, then we have a
1551 * root priming situation. */
1552 if(iq->dp == NULL) {
1554 /* if under auth zone, no prime needed */
1555 if(!auth_zone_delegpt(qstate, iq, delname, delnamelen))
1556 return error_response(qstate, id,
1557 LDNS_RCODE_SERVFAIL);
1558 if(iq->dp) /* use auth zone dp */
1559 return next_state(iq, INIT_REQUEST_2_STATE);
1560 /* if there is a stub, then no root prime needed */
1561 r = prime_stub(qstate, iq, id, delname,
1564 break; /* got noprime-stub-zone, continue */
1566 return 0; /* stub prime request made */
1567 if(forwards_lookup_root(qstate->env->fwds,
1568 iq->qchase.qclass)) {
1569 /* forward zone root, no root prime needed */
1570 /* fill in some dp - safety belt */
1571 iq->dp = hints_lookup_root(qstate->env->hints,
1574 log_err("internal error: no hints dp");
1575 errinf(qstate, "no hints for this class");
1576 return error_response(qstate, id,
1577 LDNS_RCODE_SERVFAIL);
1579 iq->dp = delegpt_copy(iq->dp, qstate->region);
1581 log_err("out of memory in safety belt");
1582 errinf(qstate, "malloc failure, in safety belt");
1583 return error_response(qstate, id,
1584 LDNS_RCODE_SERVFAIL);
1586 return next_state(iq, INIT_REQUEST_2_STATE);
1588 /* Note that the result of this will set a new
1589 * DelegationPoint based on the result of priming. */
1590 if(!prime_root(qstate, iq, id, iq->qchase.qclass))
1591 return error_response(qstate, id,
1592 LDNS_RCODE_REFUSED);
1594 /* priming creates and sends a subordinate query, with
1595 * this query as the parent. So further processing for
1596 * this event will stop until reactivated by the
1597 * results of priming. */
1600 if(!iq->ratelimit_ok && qstate->prefetch_leeway)
1601 iq->ratelimit_ok = 1; /* allow prefetches, this keeps
1602 otherwise valid data in the cache */
1604 /* see if this dp not useless.
1606 * o all NS items are required glue.
1607 * or the query is for NS item that is required glue.
1608 * o no addresses are provided.
1610 * Instead, go up one level, and try to get even further
1611 * If the root was useless, use safety belt information.
1612 * Only check cache returns, because replies for servers
1613 * could be useless but lead to loops (bumping into the
1614 * same server reply) if useless-checked.
1616 if(iter_dp_is_useless(&qstate->qinfo, qstate->query_flags,
1617 iq->dp, ie->supports_ipv4, ie->supports_ipv6,
1619 struct delegpt* retdp = NULL;
1620 if(!can_have_last_resort(qstate->env, iq->dp->name, iq->dp->namelen, iq->qchase.qclass, &retdp)) {
1622 verbose(VERB_QUERY, "cache has stub "
1623 "or fwd but no addresses, "
1624 "fallback to config");
1625 iq->dp = delegpt_copy(retdp,
1628 log_err("out of memory in "
1629 "stub/fwd fallback");
1630 errinf(qstate, "malloc failure, for fallback to config");
1631 return error_response(qstate,
1632 id, LDNS_RCODE_SERVFAIL);
1636 verbose(VERB_ALGO, "useless dp "
1637 "but cannot go up, servfail");
1638 delegpt_log(VERB_ALGO, iq->dp);
1639 errinf(qstate, "no useful nameservers, "
1640 "and cannot go up");
1641 errinf_dname(qstate, "for zone", iq->dp->name);
1642 return error_response(qstate, id,
1643 LDNS_RCODE_SERVFAIL);
1645 if(dname_is_root(iq->dp->name)) {
1646 /* use safety belt */
1647 verbose(VERB_QUERY, "Cache has root NS but "
1648 "no addresses. Fallback to the safety belt.");
1649 iq->dp = hints_lookup_root(qstate->env->hints,
1651 /* note deleg_msg is from previous lookup,
1652 * but RD is on, so it is not used */
1654 log_err("internal error: no hints dp");
1655 return error_response(qstate, id,
1656 LDNS_RCODE_REFUSED);
1658 iq->dp = delegpt_copy(iq->dp, qstate->region);
1660 log_err("out of memory in safety belt");
1661 errinf(qstate, "malloc failure, in safety belt, for root");
1662 return error_response(qstate, id,
1663 LDNS_RCODE_SERVFAIL);
1668 "cache delegation was useless:");
1669 delegpt_log(VERB_ALGO, iq->dp);
1671 delname = iq->dp->name;
1672 delnamelen = iq->dp->namelen;
1673 dname_remove_label(&delname, &delnamelen);
1678 verbose(VERB_ALGO, "cache delegation returns delegpt");
1679 delegpt_log(VERB_ALGO, iq->dp);
1681 /* Otherwise, set the current delegation point and move on to the
1683 return next_state(iq, INIT_REQUEST_2_STATE);
1687 * Process the second part of the initial request handling. This state
1688 * basically exists so that queries that generate root priming events have
1689 * the same init processing as ones that do not. Request events that reach
1690 * this state must have a valid currentDelegationPoint set.
1692 * This part is primarily handling stub zone priming. Events that reach this
1693 * state must have a current delegation point.
1695 * @param qstate: query state.
1696 * @param iq: iterator query state.
1697 * @param id: module id.
1698 * @return true if the event needs more request processing immediately,
1702 processInitRequest2(struct module_qstate* qstate, struct iter_qstate* iq,
1707 log_query_info(VERB_QUERY, "resolving (init part 2): ",
1710 delname = iq->qchase.qname;
1711 delnamelen = iq->qchase.qname_len;
1712 if(iq->refetch_glue) {
1713 struct iter_hints_stub* stub;
1715 log_err("internal or malloc fail: no dp for refetch");
1716 errinf(qstate, "malloc failure, no delegation info");
1717 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
1719 /* Do not send queries above stub, do not set delname to dp if
1720 * this is above stub without stub-first. */
1721 stub = hints_lookup_stub(
1722 qstate->env->hints, iq->qchase.qname, iq->qchase.qclass,
1724 if(!stub || !stub->dp->has_parent_side_NS ||
1725 dname_subdomain_c(iq->dp->name, stub->dp->name)) {
1726 delname = iq->dp->name;
1727 delnamelen = iq->dp->namelen;
1730 if(iq->qchase.qtype == LDNS_RR_TYPE_DS || iq->refetch_glue) {
1731 if(!dname_is_root(delname))
1732 dname_remove_label(&delname, &delnamelen);
1733 iq->refetch_glue = 0; /* if CNAME causes restart, no refetch */
1736 /* see if we have an auth zone to answer from, improves dp from cache
1737 * (if any dp from cache) with auth zone dp, if that is lower */
1738 if(!auth_zone_delegpt(qstate, iq, delname, delnamelen))
1739 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
1741 /* Check to see if we need to prime a stub zone. */
1742 if(prime_stub(qstate, iq, id, delname, iq->qchase.qclass)) {
1743 /* A priming sub request was made */
1747 /* most events just get forwarded to the next state. */
1748 return next_state(iq, INIT_REQUEST_3_STATE);
1752 * Process the third part of the initial request handling. This state exists
1753 * as a separate state so that queries that generate stub priming events
1754 * will get the tail end of the init process but not repeat the stub priming
1757 * @param qstate: query state.
1758 * @param iq: iterator query state.
1759 * @param id: module id.
1760 * @return true, advancing the event to the QUERYTARGETS_STATE.
1763 processInitRequest3(struct module_qstate* qstate, struct iter_qstate* iq,
1766 log_query_info(VERB_QUERY, "resolving (init part 3): ",
1768 /* if the cache reply dp equals a validation anchor or msg has DS,
1769 * then DNSSEC RRSIGs are expected in the reply */
1770 iq->dnssec_expected = iter_indicates_dnssec(qstate->env, iq->dp,
1771 iq->deleg_msg, iq->qchase.qclass);
1773 /* If the RD flag wasn't set, then we just finish with the
1774 * cached referral as the response. */
1775 if(!(qstate->query_flags & BIT_RD) && iq->deleg_msg) {
1776 iq->response = iq->deleg_msg;
1777 if(verbosity >= VERB_ALGO && iq->response)
1778 log_dns_msg("no RD requested, using delegation msg",
1779 &iq->response->qinfo, iq->response->rep);
1780 if(qstate->reply_origin)
1781 sock_list_insert(&qstate->reply_origin, NULL, 0, qstate->region);
1782 return final_state(iq);
1784 /* After this point, unset the RD flag -- this query is going to
1785 * be sent to an auth. server. */
1786 iq->chase_flags &= ~BIT_RD;
1788 /* if dnssec expected, fetch key for the trust-anchor or cached-DS */
1789 if(iq->dnssec_expected && qstate->env->cfg->prefetch_key &&
1790 !(qstate->query_flags&BIT_CD)) {
1791 generate_dnskey_prefetch(qstate, iq, id);
1792 fptr_ok(fptr_whitelist_modenv_detach_subs(
1793 qstate->env->detach_subs));
1794 (*qstate->env->detach_subs)(qstate);
1797 /* Jump to the next state. */
1798 return next_state(iq, QUERYTARGETS_STATE);
1802 * Given a basic query, generate a parent-side "target" query.
1803 * These are subordinate queries for missing delegation point target addresses,
1804 * for which only the parent of the delegation provides correct IP addresses.
1806 * @param qstate: query state.
1807 * @param iq: iterator query state.
1808 * @param id: module id.
1809 * @param name: target qname.
1810 * @param namelen: target qname length.
1811 * @param qtype: target qtype (either A or AAAA).
1812 * @param qclass: target qclass.
1813 * @return true on success, false on failure.
1816 generate_parentside_target_query(struct module_qstate* qstate,
1817 struct iter_qstate* iq, int id, uint8_t* name, size_t namelen,
1818 uint16_t qtype, uint16_t qclass)
1820 struct module_qstate* subq;
1821 if(!generate_sub_request(name, namelen, qtype, qclass, qstate,
1822 id, iq, INIT_REQUEST_STATE, FINISHED_STATE, &subq, 0, 0))
1825 struct iter_qstate* subiq =
1826 (struct iter_qstate*)subq->minfo[id];
1827 /* blacklist the cache - we want to fetch parent stuff */
1828 sock_list_insert(&subq->blacklist, NULL, 0, subq->region);
1829 subiq->query_for_pside_glue = 1;
1830 if(dname_subdomain_c(name, iq->dp->name)) {
1831 subiq->dp = delegpt_copy(iq->dp, subq->region);
1832 subiq->dnssec_expected = iter_indicates_dnssec(
1833 qstate->env, subiq->dp, NULL,
1834 subq->qinfo.qclass);
1835 subiq->refetch_glue = 1;
1837 subiq->dp = dns_cache_find_delegation(qstate->env,
1838 name, namelen, qtype, qclass, subq->region,
1840 *qstate->env->now+subq->prefetch_leeway,
1842 /* if no dp, then it's from root, refetch unneeded */
1844 subiq->dnssec_expected = iter_indicates_dnssec(
1845 qstate->env, subiq->dp, NULL,
1846 subq->qinfo.qclass);
1847 subiq->refetch_glue = 1;
1851 log_nametypeclass(VERB_QUERY, "new pside target", name, qtype, qclass);
1856 * Given a basic query, generate a "target" query. These are subordinate
1857 * queries for missing delegation point target addresses.
1859 * @param qstate: query state.
1860 * @param iq: iterator query state.
1861 * @param id: module id.
1862 * @param name: target qname.
1863 * @param namelen: target qname length.
1864 * @param qtype: target qtype (either A or AAAA).
1865 * @param qclass: target qclass.
1866 * @return true on success, false on failure.
1869 generate_target_query(struct module_qstate* qstate, struct iter_qstate* iq,
1870 int id, uint8_t* name, size_t namelen, uint16_t qtype, uint16_t qclass)
1872 struct module_qstate* subq;
1873 if(!generate_sub_request(name, namelen, qtype, qclass, qstate,
1874 id, iq, INIT_REQUEST_STATE, FINISHED_STATE, &subq, 0, 0))
1876 log_nametypeclass(VERB_QUERY, "new target", name, qtype, qclass);
1881 * Given an event at a certain state, generate zero or more target queries
1882 * for it's current delegation point.
1884 * @param qstate: query state.
1885 * @param iq: iterator query state.
1886 * @param ie: iterator shared global environment.
1887 * @param id: module id.
1888 * @param maxtargets: The maximum number of targets to query for.
1889 * if it is negative, there is no maximum number of targets.
1890 * @param num: returns the number of queries generated and processed,
1891 * which may be zero if there were no missing targets.
1892 * @return false on error.
1895 query_for_targets(struct module_qstate* qstate, struct iter_qstate* iq,
1896 struct iter_env* ie, int id, int maxtargets, int* num)
1898 int query_count = 0;
1899 struct delegpt_ns* ns;
1903 iter_mark_cycle_targets(qstate, iq->dp);
1904 missing = (int)delegpt_count_missing_targets(iq->dp, NULL);
1905 log_assert(maxtargets != 0); /* that would not be useful */
1907 /* Generate target requests. Basically, any missing targets
1908 * are queried for here, regardless if it is necessary to do
1909 * so to continue processing. */
1910 if(maxtargets < 0 || maxtargets > missing)
1912 else toget = maxtargets;
1918 /* now that we are sure that a target query is going to be made,
1919 * check the limits. */
1920 if(iq->depth == ie->max_dependency_depth)
1922 if(iq->depth > 0 && iq->target_count &&
1923 iq->target_count[TARGET_COUNT_QUERIES] > MAX_TARGET_COUNT) {
1924 char s[LDNS_MAX_DOMAINLEN+1];
1925 dname_str(qstate->qinfo.qname, s);
1926 verbose(VERB_QUERY, "request %s has exceeded the maximum "
1927 "number of glue fetches %d", s,
1928 iq->target_count[TARGET_COUNT_QUERIES]);
1931 if(iq->dp_target_count > MAX_DP_TARGET_COUNT) {
1932 char s[LDNS_MAX_DOMAINLEN+1];
1933 dname_str(qstate->qinfo.qname, s);
1934 verbose(VERB_QUERY, "request %s has exceeded the maximum "
1935 "number of glue fetches %d to a single delegation point",
1936 s, iq->dp_target_count);
1940 /* select 'toget' items from the total of 'missing' items */
1941 log_assert(toget <= missing);
1943 /* loop over missing targets */
1944 for(ns = iq->dp->nslist; ns; ns = ns->next) {
1948 /* randomly select this item with probability toget/missing */
1949 if(!iter_ns_probability(qstate->env->rnd, toget, missing)) {
1950 /* do not select this one, next; select toget number
1951 * of items from a list one less in size */
1956 if(ie->supports_ipv6 &&
1957 ((ns->lame && !ns->done_pside6) ||
1958 (!ns->lame && !ns->got6))) {
1959 /* Send the AAAA request. */
1960 if(!generate_target_query(qstate, iq, id,
1961 ns->name, ns->namelen,
1962 LDNS_RR_TYPE_AAAA, iq->qchase.qclass)) {
1965 qstate->ext_state[id] = module_wait_subquery;
1969 /* If the mesh query list is full, exit the loop here.
1970 * This makes the routine spawn one query at a time,
1971 * and this means there is no query state load
1972 * increase, because the spawned state uses cpu and a
1973 * socket while this state waits for that spawned
1974 * state. Next time we can look up further targets */
1975 if(mesh_jostle_exceeded(qstate->env->mesh))
1978 /* Send the A request. */
1979 if((ie->supports_ipv4 || ie->use_nat64) &&
1980 ((ns->lame && !ns->done_pside4) ||
1981 (!ns->lame && !ns->got4))) {
1982 if(!generate_target_query(qstate, iq, id,
1983 ns->name, ns->namelen,
1984 LDNS_RR_TYPE_A, iq->qchase.qclass)) {
1987 qstate->ext_state[id] = module_wait_subquery;
1991 /* If the mesh query list is full, exit the loop. */
1992 if(mesh_jostle_exceeded(qstate->env->mesh))
1996 /* mark this target as in progress. */
2005 qstate->ext_state[id] = module_wait_subquery;
2011 * Called by processQueryTargets when it would like extra targets to query
2012 * but it seems to be out of options. At last resort some less appealing
2013 * options are explored. If there are no more options, the result is SERVFAIL
2015 * @param qstate: query state.
2016 * @param iq: iterator query state.
2017 * @param ie: iterator shared global environment.
2018 * @param id: module id.
2019 * @return true if the event requires more request processing immediately,
2023 processLastResort(struct module_qstate* qstate, struct iter_qstate* iq,
2024 struct iter_env* ie, int id)
2026 struct delegpt_ns* ns;
2027 int query_count = 0;
2028 verbose(VERB_ALGO, "No more query targets, attempting last resort");
2031 if(!can_have_last_resort(qstate->env, iq->dp->name, iq->dp->namelen,
2032 iq->qchase.qclass, NULL)) {
2033 /* fail -- no more targets, no more hope of targets, no hope
2035 errinf(qstate, "all the configured stub or forward servers failed,");
2036 errinf_dname(qstate, "at zone", iq->dp->name);
2037 errinf_reply(qstate, iq);
2038 verbose(VERB_QUERY, "configured stub or forward servers failed -- returning SERVFAIL");
2039 return error_response_cache(qstate, id, LDNS_RCODE_SERVFAIL);
2041 if(!iq->dp->has_parent_side_NS && dname_is_root(iq->dp->name)) {
2042 struct delegpt* p = hints_lookup_root(qstate->env->hints,
2045 struct delegpt_addr* a;
2046 iq->chase_flags &= ~BIT_RD; /* go to authorities */
2047 for(ns = p->nslist; ns; ns=ns->next) {
2048 (void)delegpt_add_ns(iq->dp, qstate->region,
2049 ns->name, ns->lame, ns->tls_auth_name,
2052 for(a = p->target_list; a; a=a->next_target) {
2053 (void)delegpt_add_addr(iq->dp, qstate->region,
2054 &a->addr, a->addrlen, a->bogus,
2055 a->lame, a->tls_auth_name, -1, NULL);
2058 iq->dp->has_parent_side_NS = 1;
2059 } else if(!iq->dp->has_parent_side_NS) {
2060 if(!iter_lookup_parent_NS_from_cache(qstate->env, iq->dp,
2061 qstate->region, &qstate->qinfo)
2062 || !iq->dp->has_parent_side_NS) {
2063 /* if: malloc failure in lookup go up to try */
2064 /* if: no parent NS in cache - go up one level */
2065 verbose(VERB_ALGO, "try to grab parent NS");
2066 iq->store_parent_NS = iq->dp;
2067 iq->chase_flags &= ~BIT_RD; /* go to authorities */
2068 iq->deleg_msg = NULL;
2069 iq->refetch_glue = 1;
2070 iq->query_restart_count++;
2072 iq->dp_target_count = 0;
2073 if(qstate->env->cfg->qname_minimisation)
2074 iq->minimisation_state = INIT_MINIMISE_STATE;
2075 return next_state(iq, INIT_REQUEST_STATE);
2078 /* see if that makes new names available */
2079 if(!cache_fill_missing(qstate->env, iq->qchase.qclass,
2080 qstate->region, iq->dp))
2081 log_err("out of memory in cache_fill_missing");
2082 if(iq->dp->usable_list) {
2083 verbose(VERB_ALGO, "try parent-side-name, w. glue from cache");
2084 return next_state(iq, QUERYTARGETS_STATE);
2086 /* try to fill out parent glue from cache */
2087 if(iter_lookup_parent_glue_from_cache(qstate->env, iq->dp,
2088 qstate->region, &qstate->qinfo)) {
2089 /* got parent stuff from cache, see if we can continue */
2090 verbose(VERB_ALGO, "try parent-side glue from cache");
2091 return next_state(iq, QUERYTARGETS_STATE);
2093 /* query for an extra name added by the parent-NS record */
2094 if(delegpt_count_missing_targets(iq->dp, NULL) > 0) {
2096 verbose(VERB_ALGO, "try parent-side target name");
2097 if(!query_for_targets(qstate, iq, ie, id, 1, &qs)) {
2098 errinf(qstate, "could not fetch nameserver");
2099 errinf_dname(qstate, "at zone", iq->dp->name);
2100 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
2102 iq->num_target_queries += qs;
2103 target_count_increase(iq, qs);
2105 qstate->ext_state[id] = module_wait_subquery;
2106 return 0; /* and wait for them */
2109 if(iq->depth == ie->max_dependency_depth) {
2110 verbose(VERB_QUERY, "maxdepth and need more nameservers, fail");
2111 errinf(qstate, "cannot fetch more nameservers because at max dependency depth");
2112 return error_response_cache(qstate, id, LDNS_RCODE_SERVFAIL);
2114 if(iq->depth > 0 && iq->target_count &&
2115 iq->target_count[TARGET_COUNT_QUERIES] > MAX_TARGET_COUNT) {
2116 char s[LDNS_MAX_DOMAINLEN+1];
2117 dname_str(qstate->qinfo.qname, s);
2118 verbose(VERB_QUERY, "request %s has exceeded the maximum "
2119 "number of glue fetches %d", s,
2120 iq->target_count[TARGET_COUNT_QUERIES]);
2121 errinf(qstate, "exceeded the maximum number of glue fetches");
2122 return error_response_cache(qstate, id, LDNS_RCODE_SERVFAIL);
2124 /* mark cycle targets for parent-side lookups */
2125 iter_mark_pside_cycle_targets(qstate, iq->dp);
2126 /* see if we can issue queries to get nameserver addresses */
2127 /* this lookup is not randomized, but sequential. */
2128 for(ns = iq->dp->nslist; ns; ns = ns->next) {
2129 /* if this nameserver is at a delegation point, but that
2130 * delegation point is a stub and we cannot go higher, skip*/
2131 if( ((ie->supports_ipv6 && !ns->done_pside6) ||
2132 ((ie->supports_ipv4 || ie->use_nat64) && !ns->done_pside4)) &&
2133 !can_have_last_resort(qstate->env, ns->name, ns->namelen,
2134 iq->qchase.qclass, NULL)) {
2135 log_nametypeclass(VERB_ALGO, "cannot pside lookup ns "
2136 "because it is also a stub/forward,",
2137 ns->name, LDNS_RR_TYPE_NS, iq->qchase.qclass);
2138 if(ie->supports_ipv6) ns->done_pside6 = 1;
2139 if(ie->supports_ipv4 || ie->use_nat64) ns->done_pside4 = 1;
2142 /* query for parent-side A and AAAA for nameservers */
2143 if(ie->supports_ipv6 && !ns->done_pside6) {
2144 /* Send the AAAA request. */
2145 if(!generate_parentside_target_query(qstate, iq, id,
2146 ns->name, ns->namelen,
2147 LDNS_RR_TYPE_AAAA, iq->qchase.qclass)) {
2148 errinf_dname(qstate, "could not generate nameserver AAAA lookup for", ns->name);
2149 return error_response(qstate, id,
2150 LDNS_RCODE_SERVFAIL);
2152 ns->done_pside6 = 1;
2154 if(mesh_jostle_exceeded(qstate->env->mesh)) {
2155 /* Wait for the lookup; do not spawn multiple
2156 * lookups at a time. */
2157 verbose(VERB_ALGO, "try parent-side glue lookup");
2158 iq->num_target_queries += query_count;
2159 target_count_increase(iq, query_count);
2160 qstate->ext_state[id] = module_wait_subquery;
2164 if((ie->supports_ipv4 || ie->use_nat64) && !ns->done_pside4) {
2165 /* Send the A request. */
2166 if(!generate_parentside_target_query(qstate, iq, id,
2167 ns->name, ns->namelen,
2168 LDNS_RR_TYPE_A, iq->qchase.qclass)) {
2169 errinf_dname(qstate, "could not generate nameserver A lookup for", ns->name);
2170 return error_response(qstate, id,
2171 LDNS_RCODE_SERVFAIL);
2173 ns->done_pside4 = 1;
2176 if(query_count != 0) { /* suspend to await results */
2177 verbose(VERB_ALGO, "try parent-side glue lookup");
2178 iq->num_target_queries += query_count;
2179 target_count_increase(iq, query_count);
2180 qstate->ext_state[id] = module_wait_subquery;
2185 /* if this was a parent-side glue query itself, then store that
2186 * failure in cache. */
2187 if(!qstate->no_cache_store && iq->query_for_pside_glue
2189 iter_store_parentside_neg(qstate->env, &qstate->qinfo,
2190 iq->deleg_msg?iq->deleg_msg->rep:
2191 (iq->response?iq->response->rep:NULL));
2193 errinf(qstate, "all servers for this domain failed,");
2194 errinf_dname(qstate, "at zone", iq->dp->name);
2195 errinf_reply(qstate, iq);
2196 verbose(VERB_QUERY, "out of query targets -- returning SERVFAIL");
2197 /* fail -- no more targets, no more hope of targets, no hope
2199 return error_response_cache(qstate, id, LDNS_RCODE_SERVFAIL);
2203 * Try to find the NS record set that will resolve a qtype DS query. Due
2204 * to grandparent/grandchild reasons we did not get a proper lookup right
2205 * away. We need to create type NS queries until we get the right parent
2206 * for this lookup. We remove labels from the query to find the right point.
2207 * If we end up at the old dp name, then there is no solution.
2209 * @param qstate: query state.
2210 * @param iq: iterator query state.
2211 * @param id: module id.
2212 * @return true if the event requires more immediate processing, false if
2213 * not. This is generally only true when forwarding the request to
2214 * the final state (i.e., on answer).
2217 processDSNSFind(struct module_qstate* qstate, struct iter_qstate* iq, int id)
2219 struct module_qstate* subq = NULL;
2220 verbose(VERB_ALGO, "processDSNSFind");
2222 if(!iq->dsns_point) {
2224 iq->dsns_point = iq->qchase.qname;
2225 iq->dsns_point_len = iq->qchase.qname_len;
2227 /* robustcheck for internal error: we are not underneath the dp */
2228 if(!dname_subdomain_c(iq->dsns_point, iq->dp->name)) {
2229 errinf_dname(qstate, "for DS query parent-child nameserver search the query is not under the zone", iq->dp->name);
2230 return error_response_cache(qstate, id, LDNS_RCODE_SERVFAIL);
2233 /* go up one (more) step, until we hit the dp, if so, end */
2234 dname_remove_label(&iq->dsns_point, &iq->dsns_point_len);
2235 if(query_dname_compare(iq->dsns_point, iq->dp->name) == 0) {
2236 /* there was no inbetween nameserver, use the old delegation
2237 * point again. And this time, because dsns_point is nonNULL
2238 * we are going to accept the (bad) result */
2239 iq->state = QUERYTARGETS_STATE;
2242 iq->state = DSNS_FIND_STATE;
2244 /* spawn NS lookup (validation not needed, this is for DS lookup) */
2245 log_nametypeclass(VERB_ALGO, "fetch nameservers",
2246 iq->dsns_point, LDNS_RR_TYPE_NS, iq->qchase.qclass);
2247 if(!generate_sub_request(iq->dsns_point, iq->dsns_point_len,
2248 LDNS_RR_TYPE_NS, iq->qchase.qclass, qstate, id, iq,
2249 INIT_REQUEST_STATE, FINISHED_STATE, &subq, 0, 0)) {
2250 errinf_dname(qstate, "for DS query parent-child nameserver search, could not generate NS lookup for", iq->dsns_point);
2251 return error_response_cache(qstate, id, LDNS_RCODE_SERVFAIL);
2258 * Check if we wait responses for sent queries and update the iterator's
2262 check_waiting_queries(struct iter_qstate* iq, struct module_qstate* qstate,
2265 if(iq->num_target_queries>0 && iq->num_current_queries>0) {
2266 verbose(VERB_ALGO, "waiting for %d targets to "
2267 "resolve or %d outstanding queries to "
2268 "respond", iq->num_target_queries,
2269 iq->num_current_queries);
2270 qstate->ext_state[id] = module_wait_reply;
2271 } else if(iq->num_target_queries>0) {
2272 verbose(VERB_ALGO, "waiting for %d targets to "
2273 "resolve", iq->num_target_queries);
2274 qstate->ext_state[id] = module_wait_subquery;
2276 verbose(VERB_ALGO, "waiting for %d "
2277 "outstanding queries to respond",
2278 iq->num_current_queries);
2279 qstate->ext_state[id] = module_wait_reply;
2284 * This is the request event state where the request will be sent to one of
2285 * its current query targets. This state also handles issuing target lookup
2286 * queries for missing target IP addresses. Queries typically iterate on
2287 * this state, both when they are just trying different targets for a given
2288 * delegation point, and when they change delegation points. This state
2289 * roughly corresponds to RFC 1034 algorithm steps 3 and 4.
2291 * @param qstate: query state.
2292 * @param iq: iterator query state.
2293 * @param ie: iterator shared global environment.
2294 * @param id: module id.
2295 * @return true if the event requires more request processing immediately,
2296 * false if not. This state only returns true when it is generating
2297 * a SERVFAIL response because the query has hit a dead end.
2300 processQueryTargets(struct module_qstate* qstate, struct iter_qstate* iq,
2301 struct iter_env* ie, int id)
2304 struct delegpt_addr* target;
2305 struct outbound_entry* outq;
2306 struct sockaddr_storage real_addr;
2307 socklen_t real_addrlen;
2308 int auth_fallback = 0;
2309 uint8_t* qout_orig = NULL;
2310 size_t qout_orig_len = 0;
2311 int sq_check_ratelimit = 1;
2312 int sq_was_ratelimited = 0;
2313 int can_do_promisc = 0;
2315 /* NOTE: a request will encounter this state for each target it
2316 * needs to send a query to. That is, at least one per referral,
2317 * more if some targets timeout or return throwaway answers. */
2319 log_query_info(VERB_QUERY, "processQueryTargets:", &qstate->qinfo);
2320 verbose(VERB_ALGO, "processQueryTargets: targetqueries %d, "
2321 "currentqueries %d sentcount %d", iq->num_target_queries,
2322 iq->num_current_queries, iq->sent_count);
2324 /* Make sure that we haven't run away */
2325 if(iq->referral_count > MAX_REFERRAL_COUNT) {
2326 verbose(VERB_QUERY, "request has exceeded the maximum "
2327 "number of referrrals with %d", iq->referral_count);
2328 errinf(qstate, "exceeded the maximum of referrals");
2329 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
2331 if(iq->sent_count > ie->max_sent_count) {
2332 verbose(VERB_QUERY, "request has exceeded the maximum "
2333 "number of sends with %d", iq->sent_count);
2334 errinf(qstate, "exceeded the maximum number of sends");
2335 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
2338 /* Check if we reached MAX_TARGET_NX limit without a fallback activation. */
2339 if(iq->target_count && !*iq->nxns_dp &&
2340 iq->target_count[TARGET_COUNT_NX] > MAX_TARGET_NX) {
2341 struct delegpt_ns* ns;
2342 /* If we can wait for resolution, do so. */
2343 if(iq->num_target_queries>0 || iq->num_current_queries>0) {
2344 check_waiting_queries(iq, qstate, id);
2347 verbose(VERB_ALGO, "request has exceeded the maximum "
2348 "number of nxdomain nameserver lookups (%d) with %d",
2349 MAX_TARGET_NX, iq->target_count[TARGET_COUNT_NX]);
2350 /* Check for dp because we require one below */
2352 verbose(VERB_QUERY, "Failed to get a delegation, "
2354 errinf(qstate, "failed to get a delegation (eg. prime "
2356 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
2358 /* We reached the limit but we already have parent side
2359 * information; stop resolution */
2360 if(iq->dp->has_parent_side_NS) {
2361 verbose(VERB_ALGO, "parent-side information is "
2362 "already present for the delegation point, no "
2363 "fallback possible");
2364 errinf(qstate, "exceeded the maximum nameserver nxdomains");
2365 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
2367 verbose(VERB_ALGO, "initiating parent-side fallback for "
2368 "nxdomain nameserver lookups");
2369 /* Mark all the current NSes as resolved to allow for parent
2371 for(ns=iq->dp->nslist; ns; ns=ns->next) {
2374 /* Note the delegation point that triggered the NXNS fallback;
2375 * no reason for shared queries to keep trying there.
2376 * This also marks the fallback activation. */
2377 *iq->nxns_dp = malloc(iq->dp->namelen);
2379 verbose(VERB_ALGO, "out of memory while initiating "
2381 errinf(qstate, "exceeded the maximum nameserver "
2382 "nxdomains (malloc)");
2383 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
2385 memcpy(*iq->nxns_dp, iq->dp->name, iq->dp->namelen);
2386 } else if(iq->target_count && *iq->nxns_dp) {
2387 /* Handle the NXNS fallback case. */
2388 /* If we can wait for resolution, do so. */
2389 if(iq->num_target_queries>0 || iq->num_current_queries>0) {
2390 check_waiting_queries(iq, qstate, id);
2393 /* Check for dp because we require one below */
2395 verbose(VERB_QUERY, "Failed to get a delegation, "
2397 errinf(qstate, "failed to get a delegation (eg. prime "
2399 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
2402 if(iq->target_count[TARGET_COUNT_NX] > MAX_TARGET_NX_FALLBACK) {
2403 verbose(VERB_ALGO, "request has exceeded the maximum "
2404 "number of fallback nxdomain nameserver "
2405 "lookups (%d) with %d", MAX_TARGET_NX_FALLBACK,
2406 iq->target_count[TARGET_COUNT_NX]);
2407 errinf(qstate, "exceeded the maximum nameserver nxdomains");
2408 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
2411 if(!iq->dp->has_parent_side_NS) {
2412 struct delegpt_ns* ns;
2413 if(!dname_canonical_compare(*iq->nxns_dp, iq->dp->name)) {
2414 verbose(VERB_ALGO, "this delegation point "
2415 "initiated the fallback, marking the "
2416 "nslist as resolved");
2417 for(ns=iq->dp->nslist; ns; ns=ns->next) {
2424 /* Make sure we have a delegation point, otherwise priming failed
2425 * or another failure occurred */
2427 verbose(VERB_QUERY, "Failed to get a delegation, giving up");
2428 errinf(qstate, "failed to get a delegation (eg. prime failure)");
2429 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
2431 if(!ie->supports_ipv6)
2432 delegpt_no_ipv6(iq->dp);
2433 if(!ie->supports_ipv4 && !ie->use_nat64)
2434 delegpt_no_ipv4(iq->dp);
2435 delegpt_log(VERB_ALGO, iq->dp);
2437 if(iq->num_current_queries>0) {
2438 /* already busy answering a query, this restart is because
2439 * more delegpt addrs became available, wait for existing
2441 verbose(VERB_ALGO, "woke up, but wait for outstanding query");
2442 qstate->ext_state[id] = module_wait_reply;
2446 if(iq->minimisation_state == INIT_MINIMISE_STATE
2447 && !(iq->chase_flags & BIT_RD)) {
2448 /* (Re)set qinfo_out to (new) delegation point, except when
2449 * qinfo_out is already a subdomain of dp. This happens when
2450 * increasing by more than one label at once (QNAMEs with more
2451 * than MAX_MINIMISE_COUNT labels). */
2452 if(!(iq->qinfo_out.qname_len
2453 && dname_subdomain_c(iq->qchase.qname,
2454 iq->qinfo_out.qname)
2455 && dname_subdomain_c(iq->qinfo_out.qname,
2457 iq->qinfo_out.qname = iq->dp->name;
2458 iq->qinfo_out.qname_len = iq->dp->namelen;
2459 iq->qinfo_out.qtype = LDNS_RR_TYPE_A;
2460 iq->qinfo_out.qclass = iq->qchase.qclass;
2461 iq->qinfo_out.local_alias = NULL;
2462 iq->minimise_count = 0;
2465 iq->minimisation_state = MINIMISE_STATE;
2467 if(iq->minimisation_state == MINIMISE_STATE) {
2468 int qchaselabs = dname_count_labels(iq->qchase.qname);
2469 int labdiff = qchaselabs -
2470 dname_count_labels(iq->qinfo_out.qname);
2472 qout_orig = iq->qinfo_out.qname;
2473 qout_orig_len = iq->qinfo_out.qname_len;
2474 iq->qinfo_out.qname = iq->qchase.qname;
2475 iq->qinfo_out.qname_len = iq->qchase.qname_len;
2476 iq->minimise_count++;
2477 iq->timeout_count = 0;
2479 iter_dec_attempts(iq->dp, 1, ie->outbound_msg_retry);
2481 /* Limit number of iterations for QNAMEs with more
2482 * than MAX_MINIMISE_COUNT labels. Send first MINIMISE_ONE_LAB
2483 * labels of QNAME always individually.
2485 if(qchaselabs > MAX_MINIMISE_COUNT && labdiff > 1 &&
2486 iq->minimise_count > MINIMISE_ONE_LAB) {
2487 if(iq->minimise_count < MAX_MINIMISE_COUNT) {
2488 int multilabs = qchaselabs - 1 -
2490 int extralabs = multilabs /
2491 MINIMISE_MULTIPLE_LABS;
2493 if (MAX_MINIMISE_COUNT - iq->minimise_count >=
2494 multilabs % MINIMISE_MULTIPLE_LABS)
2495 /* Default behaviour is to add 1 label
2496 * every iteration. Therefore, decrement
2497 * the extralabs by 1 */
2499 if (extralabs < labdiff)
2500 labdiff -= extralabs;
2504 /* Last minimised iteration, send all labels with
2511 verbose(VERB_QUERY, "removing %d labels", labdiff-1);
2512 dname_remove_labels(&iq->qinfo_out.qname,
2513 &iq->qinfo_out.qname_len,
2516 if(labdiff < 1 || (labdiff < 2
2517 && (iq->qchase.qtype == LDNS_RR_TYPE_DS
2518 || iq->qchase.qtype == LDNS_RR_TYPE_A)))
2519 /* Stop minimising this query, resolve "as usual" */
2520 iq->minimisation_state = DONOT_MINIMISE_STATE;
2521 else if(!qstate->no_cache_lookup) {
2522 struct dns_msg* msg = dns_cache_lookup(qstate->env,
2523 iq->qinfo_out.qname, iq->qinfo_out.qname_len,
2524 iq->qinfo_out.qtype, iq->qinfo_out.qclass,
2525 qstate->query_flags, qstate->region,
2526 qstate->env->scratch, 0, iq->dp->name,
2528 if(msg && FLAGS_GET_RCODE(msg->rep->flags) ==
2530 /* no need to send query if it is already
2531 * cached as NOERROR */
2533 if(msg && FLAGS_GET_RCODE(msg->rep->flags) ==
2534 LDNS_RCODE_NXDOMAIN &&
2535 qstate->env->need_to_validate &&
2536 qstate->env->cfg->harden_below_nxdomain) {
2537 if(msg->rep->security == sec_status_secure) {
2539 return final_state(iq);
2541 if(msg->rep->security == sec_status_unchecked) {
2542 struct module_qstate* subq = NULL;
2543 if(!generate_sub_request(
2544 iq->qinfo_out.qname,
2545 iq->qinfo_out.qname_len,
2546 iq->qinfo_out.qtype,
2547 iq->qinfo_out.qclass,
2550 FINISHED_STATE, &subq, 1, 1))
2552 "could not validate NXDOMAIN "
2556 if(msg && FLAGS_GET_RCODE(msg->rep->flags) ==
2557 LDNS_RCODE_NXDOMAIN) {
2558 /* return and add a label in the next
2559 * minimisation iteration.
2565 if(iq->minimisation_state == SKIP_MINIMISE_STATE) {
2566 if(iq->timeout_count < MAX_MINIMISE_TIMEOUT_COUNT)
2567 /* Do not increment qname, continue incrementing next
2569 iq->minimisation_state = MINIMISE_STATE;
2570 else if(!qstate->env->cfg->qname_minimisation_strict)
2571 /* Too many time-outs detected for this QNAME and QTYPE.
2572 * We give up, disable QNAME minimisation. */
2573 iq->minimisation_state = DONOT_MINIMISE_STATE;
2575 if(iq->minimisation_state == DONOT_MINIMISE_STATE)
2576 iq->qinfo_out = iq->qchase;
2578 /* now find an answer to this query */
2579 /* see if authority zones have an answer */
2580 /* now we know the dp, we can check the auth zone for locally hosted
2582 if(!iq->auth_zone_avoid && qstate->blacklist) {
2583 if(auth_zones_can_fallback(qstate->env->auth_zones,
2584 iq->dp->name, iq->dp->namelen, iq->qinfo_out.qclass)) {
2585 /* if cache is blacklisted and this zone allows us
2586 * to fallback to the internet, then do so, and
2587 * fetch results from the internet servers */
2588 iq->auth_zone_avoid = 1;
2591 if(iq->auth_zone_avoid) {
2592 iq->auth_zone_avoid = 0;
2594 } else if(auth_zones_lookup(qstate->env->auth_zones, &iq->qinfo_out,
2595 qstate->region, &iq->response, &auth_fallback, iq->dp->name,
2597 /* use this as a response to be processed by the iterator */
2598 if(verbosity >= VERB_ALGO) {
2599 log_dns_msg("msg from auth zone",
2600 &iq->response->qinfo, iq->response->rep);
2602 if((iq->chase_flags&BIT_RD) && !(iq->response->rep->flags&BIT_AA)) {
2603 verbose(VERB_ALGO, "forwarder, ignoring referral from auth zone");
2605 lock_rw_wrlock(&qstate->env->auth_zones->lock);
2606 qstate->env->auth_zones->num_query_up++;
2607 lock_rw_unlock(&qstate->env->auth_zones->lock);
2608 iq->num_current_queries++;
2609 iq->chase_to_rd = 0;
2610 iq->dnssec_lame_query = 0;
2611 iq->auth_zone_response = 1;
2612 return next_state(iq, QUERY_RESP_STATE);
2615 iq->auth_zone_response = 0;
2616 if(auth_fallback == 0) {
2617 /* like we got servfail from the auth zone lookup, and
2618 * no internet fallback */
2619 verbose(VERB_ALGO, "auth zone lookup failed, no fallback,"
2621 errinf(qstate, "auth zone lookup failed, fallback is off");
2622 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
2624 if(iq->dp->auth_dp) {
2625 /* we wanted to fallback, but had no delegpt, only the
2626 * auth zone generated delegpt, create an actual one */
2627 iq->auth_zone_avoid = 1;
2628 return next_state(iq, INIT_REQUEST_STATE);
2630 /* but mostly, fallback==1 (like, when no such auth zone exists)
2631 * and we continue with lookups */
2634 /* < not <=, because although the array is large enough for <=, the
2635 * generated query will immediately be discarded due to depth and
2636 * that servfail is cached, which is not good as opportunism goes. */
2637 if(iq->depth < ie->max_dependency_depth
2638 && iq->num_target_queries == 0
2639 && (!iq->target_count || iq->target_count[TARGET_COUNT_NX]==0)
2640 && iq->sent_count < TARGET_FETCH_STOP) {
2643 /* if the mesh query list is full, then do not waste cpu and sockets to
2644 * fetch promiscuous targets. They can be looked up when needed. */
2645 if(can_do_promisc && !mesh_jostle_exceeded(qstate->env->mesh)) {
2646 tf_policy = ie->target_fetch_policy[iq->depth];
2649 /* if in 0x20 fallback get as many targets as possible */
2650 if(iq->caps_fallback) {
2652 size_t naddr, nres, navail;
2653 if(!query_for_targets(qstate, iq, ie, id, -1, &extra)) {
2654 errinf(qstate, "could not fetch nameservers for 0x20 fallback");
2655 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
2657 iq->num_target_queries += extra;
2658 target_count_increase(iq, extra);
2659 if(iq->num_target_queries > 0) {
2660 /* wait to get all targets, we want to try em */
2661 verbose(VERB_ALGO, "wait for all targets for fallback");
2662 qstate->ext_state[id] = module_wait_reply;
2663 /* undo qname minimise step because we'll get back here
2665 if(qout_orig && iq->minimise_count > 0) {
2666 iq->minimise_count--;
2667 iq->qinfo_out.qname = qout_orig;
2668 iq->qinfo_out.qname_len = qout_orig_len;
2672 /* did we do enough fallback queries already? */
2673 delegpt_count_addr(iq->dp, &naddr, &nres, &navail);
2674 /* the current caps_server is the number of fallbacks sent.
2675 * the original query is one that matched too, so we have
2676 * caps_server+1 number of matching queries now */
2677 if(iq->caps_server+1 >= naddr*3 ||
2678 iq->caps_server*2+2 >= (size_t)ie->max_sent_count) {
2679 /* *2 on sentcount check because ipv6 may fail */
2680 /* we're done, process the response */
2681 verbose(VERB_ALGO, "0x20 fallback had %d responses "
2682 "match for %d wanted, done.",
2683 (int)iq->caps_server+1, (int)naddr*3);
2684 iq->response = iq->caps_response;
2685 iq->caps_fallback = 0;
2686 iter_dec_attempts(iq->dp, 3, ie->outbound_msg_retry); /* space for fallback */
2687 iq->num_current_queries++; /* RespState decrements it*/
2688 iq->referral_count++; /* make sure we don't loop */
2690 iq->dp_target_count = 0;
2691 iq->state = QUERY_RESP_STATE;
2694 verbose(VERB_ALGO, "0x20 fallback number %d",
2695 (int)iq->caps_server);
2697 /* if there is a policy to fetch missing targets
2698 * opportunistically, do it. we rely on the fact that once a
2699 * query (or queries) for a missing name have been issued,
2700 * they will not show up again. */
2701 } else if(tf_policy != 0) {
2703 verbose(VERB_ALGO, "attempt to get extra %d targets",
2705 (void)query_for_targets(qstate, iq, ie, id, tf_policy, &extra);
2706 /* errors ignored, these targets are not strictly necessary for
2707 * this result, we do not have to reply with SERVFAIL */
2708 iq->num_target_queries += extra;
2709 target_count_increase(iq, extra);
2712 /* Add the current set of unused targets to our queue. */
2713 delegpt_add_unused_targets(iq->dp);
2715 if(qstate->env->auth_zones) {
2716 /* apply rpz triggers at query time */
2717 struct dns_msg* forged_response = rpz_callback_from_iterator_module(qstate, iq);
2718 if(forged_response != NULL) {
2719 qstate->ext_state[id] = module_finished;
2720 qstate->return_rcode = LDNS_RCODE_NOERROR;
2721 qstate->return_msg = forged_response;
2722 iq->response = forged_response;
2723 next_state(iq, FINISHED_STATE);
2724 if(!iter_prepend(iq, qstate->return_msg, qstate->region)) {
2725 log_err("rpz: prepend rrsets: out of memory");
2726 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
2732 /* Select the next usable target, filtering out unsuitable targets. */
2733 target = iter_server_selection(ie, qstate->env, iq->dp,
2734 iq->dp->name, iq->dp->namelen, iq->qchase.qtype,
2735 &iq->dnssec_lame_query, &iq->chase_to_rd,
2736 iq->num_target_queries, qstate->blacklist,
2737 qstate->prefetch_leeway);
2739 /* If no usable target was selected... */
2741 /* Here we distinguish between three states: generate a new
2742 * target query, just wait, or quit (with a SERVFAIL).
2743 * We have the following information: number of active
2744 * target queries, number of active current queries,
2745 * the presence of missing targets at this delegation
2746 * point, and the given query target policy. */
2748 /* Check for the wait condition. If this is true, then
2749 * an action must be taken. */
2750 if(iq->num_target_queries==0 && iq->num_current_queries==0) {
2751 /* If there is nothing to wait for, then we need
2752 * to distinguish between generating (a) new target
2753 * query, or failing. */
2754 if(delegpt_count_missing_targets(iq->dp, NULL) > 0) {
2756 verbose(VERB_ALGO, "querying for next "
2758 if(!query_for_targets(qstate, iq, ie, id,
2760 errinf(qstate, "could not fetch nameserver");
2761 errinf_dname(qstate, "at zone", iq->dp->name);
2762 return error_response(qstate, id,
2763 LDNS_RCODE_SERVFAIL);
2766 delegpt_count_missing_targets(iq->dp, NULL) == 0){
2767 /* it looked like there were missing
2768 * targets, but they did not turn up.
2769 * Try the bad choices again (if any),
2770 * when we get back here missing==0,
2771 * so this is not a loop. */
2774 iq->num_target_queries += qs;
2775 target_count_increase(iq, qs);
2777 /* Since a target query might have been made, we
2778 * need to check again. */
2779 if(iq->num_target_queries == 0) {
2780 /* if in capsforid fallback, instead of last
2781 * resort, we agree with the current reply
2782 * we have (if any) (our count of addrs bad)*/
2783 if(iq->caps_fallback && iq->caps_reply) {
2784 /* we're done, process the response */
2785 verbose(VERB_ALGO, "0x20 fallback had %d responses, "
2786 "but no more servers except "
2787 "last resort, done.",
2788 (int)iq->caps_server+1);
2789 iq->response = iq->caps_response;
2790 iq->caps_fallback = 0;
2791 iter_dec_attempts(iq->dp, 3, ie->outbound_msg_retry); /* space for fallback */
2792 iq->num_current_queries++; /* RespState decrements it*/
2793 iq->referral_count++; /* make sure we don't loop */
2795 iq->dp_target_count = 0;
2796 iq->state = QUERY_RESP_STATE;
2799 return processLastResort(qstate, iq, ie, id);
2803 /* otherwise, we have no current targets, so submerge
2804 * until one of the target or direct queries return. */
2805 verbose(VERB_ALGO, "no current targets");
2806 check_waiting_queries(iq, qstate, id);
2807 /* undo qname minimise step because we'll get back here
2809 if(qout_orig && iq->minimise_count > 0) {
2810 iq->minimise_count--;
2811 iq->qinfo_out.qname = qout_orig;
2812 iq->qinfo_out.qname_len = qout_orig_len;
2817 /* We have a target. We could have created promiscuous target
2818 * queries but we are currently under pressure (mesh_jostle_exceeded).
2819 * If we are configured to allow promiscuous target queries and haven't
2820 * gone out to the network for a target query for this delegation, then
2821 * it is possible to slip in a promiscuous one with a 1/10 chance. */
2822 if(can_do_promisc && tf_policy == 0 && iq->depth == 0
2823 && iq->depth < ie->max_dependency_depth
2824 && ie->target_fetch_policy[iq->depth] != 0
2825 && iq->dp_target_count == 0
2826 && !ub_random_max(qstate->env->rnd, 10)) {
2828 verbose(VERB_ALGO, "available target exists in cache but "
2829 "attempt to get extra 1 target");
2830 (void)query_for_targets(qstate, iq, ie, id, 1, &extra);
2831 /* errors ignored, these targets are not strictly necessary for
2832 * this result, we do not have to reply with SERVFAIL */
2834 iq->num_target_queries += extra;
2835 target_count_increase(iq, extra);
2836 check_waiting_queries(iq, qstate, id);
2837 /* undo qname minimise step because we'll get back here
2839 if(qout_orig && iq->minimise_count > 0) {
2840 iq->minimise_count--;
2841 iq->qinfo_out.qname = qout_orig;
2842 iq->qinfo_out.qname_len = qout_orig_len;
2848 /* Do not check ratelimit for forwarding queries or if we already got a
2850 sq_check_ratelimit = (!(iq->chase_flags & BIT_RD) && !iq->ratelimit_ok);
2851 /* We have a valid target. */
2852 if(verbosity >= VERB_QUERY) {
2853 log_query_info(VERB_QUERY, "sending query:", &iq->qinfo_out);
2854 log_name_addr(VERB_QUERY, "sending to target:", iq->dp->name,
2855 &target->addr, target->addrlen);
2856 verbose(VERB_ALGO, "dnssec status: %s%s",
2857 iq->dnssec_expected?"expected": "not expected",
2858 iq->dnssec_lame_query?" but lame_query anyway": "");
2861 real_addr = target->addr;
2862 real_addrlen = target->addrlen;
2864 if(ie->use_nat64 && target->addr.ss_family == AF_INET) {
2865 addr_to_nat64(&target->addr, &ie->nat64_prefix_addr,
2866 ie->nat64_prefix_addrlen, ie->nat64_prefix_net,
2867 &real_addr, &real_addrlen);
2868 log_name_addr(VERB_QUERY, "applied NAT64:",
2869 iq->dp->name, &real_addr, real_addrlen);
2872 fptr_ok(fptr_whitelist_modenv_send_query(qstate->env->send_query));
2873 outq = (*qstate->env->send_query)(&iq->qinfo_out,
2874 iq->chase_flags | (iq->chase_to_rd?BIT_RD:0),
2875 /* unset CD if to forwarder(RD set) and not dnssec retry
2876 * (blacklist nonempty) and no trust-anchors are configured
2877 * above the qname or on the first attempt when dnssec is on */
2878 EDNS_DO| ((iq->chase_to_rd||(iq->chase_flags&BIT_RD)!=0)&&
2879 !qstate->blacklist&&(!iter_qname_indicates_dnssec(qstate->env,
2880 &iq->qinfo_out)||target->attempts==1)?0:BIT_CD),
2881 iq->dnssec_expected, iq->caps_fallback || is_caps_whitelisted(
2882 ie, iq), sq_check_ratelimit, &real_addr, real_addrlen,
2883 iq->dp->name, iq->dp->namelen,
2884 (iq->dp->tcp_upstream || qstate->env->cfg->tcp_upstream),
2885 (iq->dp->ssl_upstream || qstate->env->cfg->ssl_upstream),
2886 target->tls_auth_name, qstate, &sq_was_ratelimited);
2888 if(sq_was_ratelimited) {
2889 lock_basic_lock(&ie->queries_ratelimit_lock);
2890 ie->num_queries_ratelimited++;
2891 lock_basic_unlock(&ie->queries_ratelimit_lock);
2892 verbose(VERB_ALGO, "query exceeded ratelimits");
2893 qstate->was_ratelimited = 1;
2894 errinf_dname(qstate, "exceeded ratelimit for zone",
2896 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
2898 log_addr(VERB_QUERY, "error sending query to auth server",
2899 &real_addr, real_addrlen);
2900 if(qstate->env->cfg->qname_minimisation)
2901 iq->minimisation_state = SKIP_MINIMISE_STATE;
2902 return next_state(iq, QUERYTARGETS_STATE);
2904 outbound_list_insert(&iq->outlist, outq);
2905 iq->num_current_queries++;
2907 qstate->ext_state[id] = module_wait_reply;
2912 /** find NS rrset in given list */
2913 static struct ub_packed_rrset_key*
2914 find_NS(struct reply_info* rep, size_t from, size_t to)
2917 for(i=from; i<to; i++) {
2918 if(ntohs(rep->rrsets[i]->rk.type) == LDNS_RR_TYPE_NS)
2919 return rep->rrsets[i];
2926 * Process the query response. All queries end up at this state first. This
2927 * process generally consists of analyzing the response and routing the
2928 * event to the next state (either bouncing it back to a request state, or
2929 * terminating the processing for this event).
2931 * @param qstate: query state.
2932 * @param iq: iterator query state.
2933 * @param ie: iterator shared global environment.
2934 * @param id: module id.
2935 * @return true if the event requires more immediate processing, false if
2936 * not. This is generally only true when forwarding the request to
2937 * the final state (i.e., on answer).
2940 processQueryResponse(struct module_qstate* qstate, struct iter_qstate* iq,
2941 struct iter_env* ie, int id)
2943 int dnsseclame = 0, origtypecname = 0;
2944 enum response_type type;
2946 iq->num_current_queries--;
2948 if(!inplace_cb_query_response_call(qstate->env, qstate, iq->response))
2949 log_err("unable to call query_response callback");
2951 if(iq->response == NULL) {
2952 /* Don't increment qname when QNAME minimisation is enabled */
2953 if(qstate->env->cfg->qname_minimisation) {
2954 iq->minimisation_state = SKIP_MINIMISE_STATE;
2956 iq->timeout_count++;
2957 iq->chase_to_rd = 0;
2958 iq->dnssec_lame_query = 0;
2959 verbose(VERB_ALGO, "query response was timeout");
2960 return next_state(iq, QUERYTARGETS_STATE);
2962 iq->timeout_count = 0;
2963 type = response_type_from_server(
2964 (int)((iq->chase_flags&BIT_RD) || iq->chase_to_rd),
2965 iq->response, &iq->qinfo_out, iq->dp);
2966 iq->chase_to_rd = 0;
2967 /* remove TC flag, if this is erroneously set by TCP upstream */
2968 iq->response->rep->flags &= ~BIT_TC;
2969 if(type == RESPONSE_TYPE_REFERRAL && (iq->chase_flags&BIT_RD) &&
2970 !iq->auth_zone_response) {
2971 /* When forwarding (RD bit is set), we handle referrals
2972 * differently. No queries should be sent elsewhere */
2973 type = RESPONSE_TYPE_ANSWER;
2975 if(!qstate->env->cfg->disable_dnssec_lame_check && iq->dnssec_expected
2976 && !iq->dnssec_lame_query &&
2977 !(iq->chase_flags&BIT_RD)
2978 && iq->sent_count < DNSSEC_LAME_DETECT_COUNT
2979 && type != RESPONSE_TYPE_LAME
2980 && type != RESPONSE_TYPE_REC_LAME
2981 && type != RESPONSE_TYPE_THROWAWAY
2982 && type != RESPONSE_TYPE_UNTYPED) {
2983 /* a possible answer, see if it is missing DNSSEC */
2984 /* but not when forwarding, so we dont mark fwder lame */
2985 if(!iter_msg_has_dnssec(iq->response)) {
2986 /* Mark this address as dnsseclame in this dp,
2987 * because that will make serverselection disprefer
2988 * it, but also, once it is the only final option,
2989 * use dnssec-lame-bypass if it needs to query there.*/
2991 struct delegpt_addr* a = delegpt_find_addr(
2992 iq->dp, &qstate->reply->remote_addr,
2993 qstate->reply->remote_addrlen);
2994 if(a) a->dnsseclame = 1;
2996 /* test the answer is from the zone we expected,
2997 * otherwise, (due to parent,child on same server), we
2998 * might mark the server,zone lame inappropriately */
2999 if(!iter_msg_from_zone(iq->response, iq->dp, type,
3001 qstate->reply = NULL;
3002 type = RESPONSE_TYPE_LAME;
3005 } else iq->dnssec_lame_query = 0;
3006 /* see if referral brings us close to the target */
3007 if(type == RESPONSE_TYPE_REFERRAL) {
3008 struct ub_packed_rrset_key* ns = find_NS(
3009 iq->response->rep, iq->response->rep->an_numrrsets,
3010 iq->response->rep->an_numrrsets
3011 + iq->response->rep->ns_numrrsets);
3012 if(!ns) ns = find_NS(iq->response->rep, 0,
3013 iq->response->rep->an_numrrsets);
3014 if(!ns || !dname_strict_subdomain_c(ns->rk.dname, iq->dp->name)
3015 || !dname_subdomain_c(iq->qchase.qname, ns->rk.dname)){
3016 verbose(VERB_ALGO, "bad referral, throwaway");
3017 type = RESPONSE_TYPE_THROWAWAY;
3019 iter_scrub_ds(iq->response, ns, iq->dp->name);
3020 } else iter_scrub_ds(iq->response, NULL, NULL);
3021 if(type == RESPONSE_TYPE_THROWAWAY &&
3022 FLAGS_GET_RCODE(iq->response->rep->flags) == LDNS_RCODE_YXDOMAIN) {
3023 /* YXDOMAIN is a permanent error, no need to retry */
3024 type = RESPONSE_TYPE_ANSWER;
3026 if(type == RESPONSE_TYPE_CNAME)
3028 if(type == RESPONSE_TYPE_CNAME && iq->response->rep->an_numrrsets >= 1
3029 && ntohs(iq->response->rep->rrsets[0]->rk.type) == LDNS_RR_TYPE_DNAME) {
3030 uint8_t* sname = NULL;
3031 size_t snamelen = 0;
3032 get_cname_target(iq->response->rep->rrsets[0], &sname,
3034 if(snamelen && dname_subdomain_c(sname, iq->response->rep->rrsets[0]->rk.dname)) {
3035 /* DNAME to a subdomain loop; do not recurse */
3036 type = RESPONSE_TYPE_ANSWER;
3038 } else if(type == RESPONSE_TYPE_CNAME &&
3039 iq->qchase.qtype == LDNS_RR_TYPE_CNAME &&
3040 iq->minimisation_state == MINIMISE_STATE &&
3041 query_dname_compare(iq->qchase.qname, iq->qinfo_out.qname) == 0) {
3042 /* The minimised query for full QTYPE and hidden QTYPE can be
3043 * classified as CNAME response type, even when the original
3044 * QTYPE=CNAME. This should be treated as answer response type.
3046 type = RESPONSE_TYPE_ANSWER;
3049 /* handle each of the type cases */
3050 if(type == RESPONSE_TYPE_ANSWER) {
3051 /* ANSWER type responses terminate the query algorithm,
3052 * so they sent on their */
3053 if(verbosity >= VERB_DETAIL) {
3054 verbose(VERB_DETAIL, "query response was %s",
3055 FLAGS_GET_RCODE(iq->response->rep->flags)
3056 ==LDNS_RCODE_NXDOMAIN?"NXDOMAIN ANSWER":
3057 (iq->response->rep->an_numrrsets?"ANSWER":
3060 /* if qtype is DS, check we have the right level of answer,
3061 * like grandchild answer but we need the middle, reject it */
3062 if(iq->qchase.qtype == LDNS_RR_TYPE_DS && !iq->dsns_point
3063 && !(iq->chase_flags&BIT_RD)
3064 && iter_ds_toolow(iq->response, iq->dp)
3065 && iter_dp_cangodown(&iq->qchase, iq->dp)) {
3066 /* close down outstanding requests to be discarded */
3067 outbound_list_clear(&iq->outlist);
3068 iq->num_current_queries = 0;
3069 fptr_ok(fptr_whitelist_modenv_detach_subs(
3070 qstate->env->detach_subs));
3071 (*qstate->env->detach_subs)(qstate);
3072 iq->num_target_queries = 0;
3073 return processDSNSFind(qstate, iq, id);
3075 if(!qstate->no_cache_store)
3076 iter_dns_store(qstate->env, &iq->response->qinfo,
3078 iq->qchase.qtype != iq->response->qinfo.qtype,
3079 qstate->prefetch_leeway,
3080 iq->dp&&iq->dp->has_parent_side_NS,
3081 qstate->region, qstate->query_flags,
3082 qstate->qstarttime);
3083 /* close down outstanding requests to be discarded */
3084 outbound_list_clear(&iq->outlist);
3085 iq->num_current_queries = 0;
3086 fptr_ok(fptr_whitelist_modenv_detach_subs(
3087 qstate->env->detach_subs));
3088 (*qstate->env->detach_subs)(qstate);
3089 iq->num_target_queries = 0;
3091 sock_list_insert(&qstate->reply_origin,
3092 &qstate->reply->remote_addr,
3093 qstate->reply->remote_addrlen, qstate->region);
3094 if(iq->minimisation_state != DONOT_MINIMISE_STATE
3095 && !(iq->chase_flags & BIT_RD)) {
3096 if(FLAGS_GET_RCODE(iq->response->rep->flags) !=
3097 LDNS_RCODE_NOERROR) {
3098 if(qstate->env->cfg->qname_minimisation_strict) {
3099 if(FLAGS_GET_RCODE(iq->response->rep->flags) ==
3100 LDNS_RCODE_NXDOMAIN) {
3101 iter_scrub_nxdomain(iq->response);
3102 return final_state(iq);
3104 return error_response(qstate, id,
3105 LDNS_RCODE_SERVFAIL);
3107 /* Best effort qname-minimisation.
3108 * Stop minimising and send full query when
3109 * RCODE is not NOERROR. */
3110 iq->minimisation_state = DONOT_MINIMISE_STATE;
3112 if(FLAGS_GET_RCODE(iq->response->rep->flags) ==
3113 LDNS_RCODE_NXDOMAIN && !origtypecname) {
3114 /* Stop resolving when NXDOMAIN is DNSSEC
3115 * signed. Based on assumption that nameservers
3116 * serving signed zones do not return NXDOMAIN
3117 * for empty-non-terminals. */
3118 /* If this response is actually a CNAME type,
3119 * the nxdomain rcode may not be for the qname,
3120 * and so it is not the final response. */
3121 if(iq->dnssec_expected)
3122 return final_state(iq);
3123 /* Make subrequest to validate intermediate
3124 * NXDOMAIN if harden-below-nxdomain is
3126 if(qstate->env->cfg->harden_below_nxdomain &&
3127 qstate->env->need_to_validate) {
3128 struct module_qstate* subq = NULL;
3129 log_query_info(VERB_QUERY,
3130 "schedule NXDOMAIN validation:",
3131 &iq->response->qinfo);
3132 if(!generate_sub_request(
3133 iq->response->qinfo.qname,
3134 iq->response->qinfo.qname_len,
3135 iq->response->qinfo.qtype,
3136 iq->response->qinfo.qclass,
3139 FINISHED_STATE, &subq, 1, 1))
3141 "could not validate NXDOMAIN "
3145 return next_state(iq, QUERYTARGETS_STATE);
3147 return final_state(iq);
3148 } else if(type == RESPONSE_TYPE_REFERRAL) {
3149 /* REFERRAL type responses get a reset of the
3150 * delegation point, and back to the QUERYTARGETS_STATE. */
3151 verbose(VERB_DETAIL, "query response was REFERRAL");
3153 /* if hardened, only store referral if we asked for it */
3154 if(!qstate->no_cache_store &&
3155 (!qstate->env->cfg->harden_referral_path ||
3156 ( qstate->qinfo.qtype == LDNS_RR_TYPE_NS
3157 && (qstate->query_flags&BIT_RD)
3158 && !(qstate->query_flags&BIT_CD)
3159 /* we know that all other NS rrsets are scrubbed
3160 * away, thus on referral only one is left.
3161 * see if that equals the query name... */
3162 && ( /* auth section, but sometimes in answer section*/
3163 reply_find_rrset_section_ns(iq->response->rep,
3164 iq->qchase.qname, iq->qchase.qname_len,
3165 LDNS_RR_TYPE_NS, iq->qchase.qclass)
3166 || reply_find_rrset_section_an(iq->response->rep,
3167 iq->qchase.qname, iq->qchase.qname_len,
3168 LDNS_RR_TYPE_NS, iq->qchase.qclass)
3171 /* Store the referral under the current query */
3172 /* no prefetch-leeway, since its not the answer */
3173 iter_dns_store(qstate->env, &iq->response->qinfo,
3174 iq->response->rep, 1, 0, 0, NULL, 0,
3175 qstate->qstarttime);
3176 if(iq->store_parent_NS)
3177 iter_store_parentside_NS(qstate->env,
3179 if(qstate->env->neg_cache)
3180 val_neg_addreferral(qstate->env->neg_cache,
3181 iq->response->rep, iq->dp->name);
3183 /* store parent-side-in-zone-glue, if directly queried for */
3184 if(!qstate->no_cache_store && iq->query_for_pside_glue
3185 && !iq->pside_glue) {
3186 iq->pside_glue = reply_find_rrset(iq->response->rep,
3187 iq->qchase.qname, iq->qchase.qname_len,
3188 iq->qchase.qtype, iq->qchase.qclass);
3189 if(iq->pside_glue) {
3190 log_rrset_key(VERB_ALGO, "found parent-side "
3191 "glue", iq->pside_glue);
3192 iter_store_parentside_rrset(qstate->env,
3197 /* Reset the event state, setting the current delegation
3198 * point to the referral. */
3199 iq->deleg_msg = iq->response;
3200 iq->dp = delegpt_from_message(iq->response, qstate->region);
3201 if (qstate->env->cfg->qname_minimisation)
3202 iq->minimisation_state = INIT_MINIMISE_STATE;
3204 errinf(qstate, "malloc failure, for delegation point");
3205 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
3207 if(!cache_fill_missing(qstate->env, iq->qchase.qclass,
3208 qstate->region, iq->dp)) {
3209 errinf(qstate, "malloc failure, copy extra info into delegation point");
3210 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
3212 if(iq->store_parent_NS && query_dname_compare(iq->dp->name,
3213 iq->store_parent_NS->name) == 0)
3214 iter_merge_retry_counts(iq->dp, iq->store_parent_NS,
3215 ie->outbound_msg_retry);
3216 delegpt_log(VERB_ALGO, iq->dp);
3217 /* Count this as a referral. */
3218 iq->referral_count++;
3220 iq->dp_target_count = 0;
3221 /* see if the next dp is a trust anchor, or a DS was sent
3222 * along, indicating dnssec is expected for next zone */
3223 iq->dnssec_expected = iter_indicates_dnssec(qstate->env,
3224 iq->dp, iq->response, iq->qchase.qclass);
3225 /* if dnssec, validating then also fetch the key for the DS */
3226 if(iq->dnssec_expected && qstate->env->cfg->prefetch_key &&
3227 !(qstate->query_flags&BIT_CD))
3228 generate_dnskey_prefetch(qstate, iq, id);
3230 /* spawn off NS and addr to auth servers for the NS we just
3231 * got in the referral. This gets authoritative answer
3232 * (answer section trust level) rrset.
3233 * right after, we detach the subs, answer goes to cache. */
3234 if(qstate->env->cfg->harden_referral_path)
3235 generate_ns_check(qstate, iq, id);
3237 /* stop current outstanding queries.
3238 * FIXME: should the outstanding queries be waited for and
3239 * handled? Say by a subquery that inherits the outbound_entry.
3241 outbound_list_clear(&iq->outlist);
3242 iq->num_current_queries = 0;
3243 fptr_ok(fptr_whitelist_modenv_detach_subs(
3244 qstate->env->detach_subs));
3245 (*qstate->env->detach_subs)(qstate);
3246 iq->num_target_queries = 0;
3247 iq->response = NULL;
3248 iq->fail_addr_type = 0;
3249 verbose(VERB_ALGO, "cleared outbound list for next round");
3250 return next_state(iq, QUERYTARGETS_STATE);
3251 } else if(type == RESPONSE_TYPE_CNAME) {
3252 uint8_t* sname = NULL;
3253 size_t snamelen = 0;
3254 /* CNAME type responses get a query restart (i.e., get a
3255 * reset of the query state and go back to INIT_REQUEST_STATE).
3257 verbose(VERB_DETAIL, "query response was CNAME");
3258 if(verbosity >= VERB_ALGO)
3259 log_dns_msg("cname msg", &iq->response->qinfo,
3261 /* if qtype is DS, check we have the right level of answer,
3262 * like grandchild answer but we need the middle, reject it */
3263 if(iq->qchase.qtype == LDNS_RR_TYPE_DS && !iq->dsns_point
3264 && !(iq->chase_flags&BIT_RD)
3265 && iter_ds_toolow(iq->response, iq->dp)
3266 && iter_dp_cangodown(&iq->qchase, iq->dp)) {
3267 outbound_list_clear(&iq->outlist);
3268 iq->num_current_queries = 0;
3269 fptr_ok(fptr_whitelist_modenv_detach_subs(
3270 qstate->env->detach_subs));
3271 (*qstate->env->detach_subs)(qstate);
3272 iq->num_target_queries = 0;
3273 return processDSNSFind(qstate, iq, id);
3275 /* Process the CNAME response. */
3276 if(!handle_cname_response(qstate, iq, iq->response,
3277 &sname, &snamelen)) {
3278 errinf(qstate, "malloc failure, CNAME info");
3279 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
3281 /* cache the CNAME response under the current query */
3282 /* NOTE : set referral=1, so that rrsets get stored but not
3283 * the partial query answer (CNAME only). */
3284 /* prefetchleeway applied because this updates answer parts */
3285 if(!qstate->no_cache_store)
3286 iter_dns_store(qstate->env, &iq->response->qinfo,
3287 iq->response->rep, 1, qstate->prefetch_leeway,
3288 iq->dp&&iq->dp->has_parent_side_NS, NULL,
3289 qstate->query_flags, qstate->qstarttime);
3290 /* set the current request's qname to the new value. */
3291 iq->qchase.qname = sname;
3292 iq->qchase.qname_len = snamelen;
3293 if(qstate->env->auth_zones) {
3294 /* apply rpz qname triggers after cname */
3295 struct dns_msg* forged_response =
3296 rpz_callback_from_iterator_cname(qstate, iq);
3297 while(forged_response && reply_find_rrset_section_an(
3298 forged_response->rep, iq->qchase.qname,
3299 iq->qchase.qname_len, LDNS_RR_TYPE_CNAME,
3300 iq->qchase.qclass)) {
3301 /* another cname to follow */
3302 if(!handle_cname_response(qstate, iq, forged_response,
3303 &sname, &snamelen)) {
3304 errinf(qstate, "malloc failure, CNAME info");
3305 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
3307 iq->qchase.qname = sname;
3308 iq->qchase.qname_len = snamelen;
3310 rpz_callback_from_iterator_cname(qstate, iq);
3312 if(forged_response != NULL) {
3313 qstate->ext_state[id] = module_finished;
3314 qstate->return_rcode = LDNS_RCODE_NOERROR;
3315 qstate->return_msg = forged_response;
3316 iq->response = forged_response;
3317 next_state(iq, FINISHED_STATE);
3318 if(!iter_prepend(iq, qstate->return_msg, qstate->region)) {
3319 log_err("rpz: after cname, prepend rrsets: out of memory");
3320 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
3322 qstate->return_msg->qinfo = qstate->qinfo;
3326 /* Clear the query state, since this is a query restart. */
3327 iq->deleg_msg = NULL;
3329 iq->dsns_point = NULL;
3330 iq->auth_zone_response = 0;
3332 iq->dp_target_count = 0;
3333 if(iq->minimisation_state != MINIMISE_STATE)
3334 /* Only count as query restart when it is not an extra
3335 * query as result of qname minimisation. */
3336 iq->query_restart_count++;
3337 if(qstate->env->cfg->qname_minimisation)
3338 iq->minimisation_state = INIT_MINIMISE_STATE;
3340 /* stop current outstanding queries.
3341 * FIXME: should the outstanding queries be waited for and
3342 * handled? Say by a subquery that inherits the outbound_entry.
3344 outbound_list_clear(&iq->outlist);
3345 iq->num_current_queries = 0;
3346 fptr_ok(fptr_whitelist_modenv_detach_subs(
3347 qstate->env->detach_subs));
3348 (*qstate->env->detach_subs)(qstate);
3349 iq->num_target_queries = 0;
3351 sock_list_insert(&qstate->reply_origin,
3352 &qstate->reply->remote_addr,
3353 qstate->reply->remote_addrlen, qstate->region);
3354 verbose(VERB_ALGO, "cleared outbound list for query restart");
3355 /* go to INIT_REQUEST_STATE for new qname. */
3356 return next_state(iq, INIT_REQUEST_STATE);
3357 } else if(type == RESPONSE_TYPE_LAME) {
3358 /* Cache the LAMEness. */
3359 verbose(VERB_DETAIL, "query response was %sLAME",
3360 dnsseclame?"DNSSEC ":"");
3361 if(!dname_subdomain_c(iq->qchase.qname, iq->dp->name)) {
3362 log_err("mark lame: mismatch in qname and dpname");
3363 /* throwaway this reply below */
3364 } else if(qstate->reply) {
3365 /* need addr for lameness cache, but we may have
3366 * gotten this from cache, so test to be sure */
3367 if(!infra_set_lame(qstate->env->infra_cache,
3368 &qstate->reply->remote_addr,
3369 qstate->reply->remote_addrlen,
3370 iq->dp->name, iq->dp->namelen,
3371 *qstate->env->now, dnsseclame, 0,
3373 log_err("mark host lame: out of memory");
3375 } else if(type == RESPONSE_TYPE_REC_LAME) {
3376 /* Cache the LAMEness. */
3377 verbose(VERB_DETAIL, "query response REC_LAME: "
3378 "recursive but not authoritative server");
3379 if(!dname_subdomain_c(iq->qchase.qname, iq->dp->name)) {
3380 log_err("mark rec_lame: mismatch in qname and dpname");
3381 /* throwaway this reply below */
3382 } else if(qstate->reply) {
3383 /* need addr for lameness cache, but we may have
3384 * gotten this from cache, so test to be sure */
3385 verbose(VERB_DETAIL, "mark as REC_LAME");
3386 if(!infra_set_lame(qstate->env->infra_cache,
3387 &qstate->reply->remote_addr,
3388 qstate->reply->remote_addrlen,
3389 iq->dp->name, iq->dp->namelen,
3390 *qstate->env->now, 0, 1, iq->qchase.qtype))
3391 log_err("mark host lame: out of memory");
3393 } else if(type == RESPONSE_TYPE_THROWAWAY) {
3394 /* LAME and THROWAWAY responses are handled the same way.
3395 * In this case, the event is just sent directly back to
3396 * the QUERYTARGETS_STATE without resetting anything,
3397 * because, clearly, the next target must be tried. */
3398 verbose(VERB_DETAIL, "query response was THROWAWAY");
3400 log_warn("A query response came back with an unknown type: %d",
3404 /* LAME, THROWAWAY and "unknown" all end up here.
3405 * Recycle to the QUERYTARGETS state to hopefully try a
3406 * different target. */
3407 if (qstate->env->cfg->qname_minimisation &&
3408 !qstate->env->cfg->qname_minimisation_strict)
3409 iq->minimisation_state = DONOT_MINIMISE_STATE;
3410 if(iq->auth_zone_response) {
3411 /* can we fallback? */
3412 iq->auth_zone_response = 0;
3413 if(!auth_zones_can_fallback(qstate->env->auth_zones,
3414 iq->dp->name, iq->dp->namelen, qstate->qinfo.qclass)) {
3415 verbose(VERB_ALGO, "auth zone response bad, and no"
3416 " fallback possible, servfail");
3417 errinf_dname(qstate, "response is bad, no fallback, "
3418 "for auth zone", iq->dp->name);
3419 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
3421 verbose(VERB_ALGO, "auth zone response was bad, "
3422 "fallback enabled");
3423 iq->auth_zone_avoid = 1;
3424 if(iq->dp->auth_dp) {
3425 /* we are using a dp for the auth zone, with no
3426 * nameservers, get one first */
3428 return next_state(iq, INIT_REQUEST_STATE);
3431 return next_state(iq, QUERYTARGETS_STATE);
3435 * Return priming query results to interested super querystates.
3437 * Sets the delegation point and delegation message (not nonRD queries).
3438 * This is a callback from walk_supers.
3440 * @param qstate: priming query state that finished.
3441 * @param id: module id.
3442 * @param forq: the qstate for which priming has been done.
3445 prime_supers(struct module_qstate* qstate, int id, struct module_qstate* forq)
3447 struct iter_qstate* foriq = (struct iter_qstate*)forq->minfo[id];
3448 struct delegpt* dp = NULL;
3450 log_assert(qstate->is_priming || foriq->wait_priming_stub);
3451 log_assert(qstate->return_rcode == LDNS_RCODE_NOERROR);
3452 /* Convert our response to a delegation point */
3453 dp = delegpt_from_message(qstate->return_msg, forq->region);
3455 /* if there is no convertible delegation point, then
3456 * the ANSWER type was (presumably) a negative answer. */
3457 verbose(VERB_ALGO, "prime response was not a positive "
3460 foriq->state = QUERYTARGETS_STATE;
3464 log_query_info(VERB_DETAIL, "priming successful for", &qstate->qinfo);
3465 delegpt_log(VERB_ALGO, dp);
3467 foriq->deleg_msg = dns_copy_msg(qstate->return_msg, forq->region);
3468 if(!foriq->deleg_msg) {
3469 log_err("copy prime response: out of memory");
3471 foriq->state = QUERYTARGETS_STATE;
3475 /* root priming responses go to init stage 2, priming stub
3476 * responses to to stage 3. */
3477 if(foriq->wait_priming_stub) {
3478 foriq->state = INIT_REQUEST_3_STATE;
3479 foriq->wait_priming_stub = 0;
3480 } else foriq->state = INIT_REQUEST_2_STATE;
3481 /* because we are finished, the parent will be reactivated */
3485 * This handles the response to a priming query. This is used to handle both
3486 * root and stub priming responses. This is basically the equivalent of the
3487 * QUERY_RESP_STATE, but will not handle CNAME responses and will treat
3488 * REFERRALs as ANSWERS. It will also update and reactivate the originating
3491 * @param qstate: query state.
3492 * @param id: module id.
3493 * @return true if the event needs more immediate processing, false if not.
3494 * This state always returns false.
3497 processPrimeResponse(struct module_qstate* qstate, int id)
3499 struct iter_qstate* iq = (struct iter_qstate*)qstate->minfo[id];
3500 enum response_type type;
3501 iq->response->rep->flags &= ~(BIT_RD|BIT_RA); /* ignore rec-lame */
3502 type = response_type_from_server(
3503 (int)((iq->chase_flags&BIT_RD) || iq->chase_to_rd),
3504 iq->response, &iq->qchase, iq->dp);
3505 if(type == RESPONSE_TYPE_ANSWER) {
3506 qstate->return_rcode = LDNS_RCODE_NOERROR;
3507 qstate->return_msg = iq->response;
3509 errinf(qstate, "prime response did not get an answer");
3510 errinf_dname(qstate, "for", qstate->qinfo.qname);
3511 qstate->return_rcode = LDNS_RCODE_SERVFAIL;
3512 qstate->return_msg = NULL;
3515 /* validate the root or stub after priming (if enabled).
3516 * This is the same query as the prime query, but with validation.
3517 * Now that we are primed, the additional queries that validation
3518 * may need can be resolved. */
3519 if(qstate->env->cfg->harden_referral_path) {
3520 struct module_qstate* subq = NULL;
3521 log_nametypeclass(VERB_ALGO, "schedule prime validation",
3522 qstate->qinfo.qname, qstate->qinfo.qtype,
3523 qstate->qinfo.qclass);
3524 if(!generate_sub_request(qstate->qinfo.qname,
3525 qstate->qinfo.qname_len, qstate->qinfo.qtype,
3526 qstate->qinfo.qclass, qstate, id, iq,
3527 INIT_REQUEST_STATE, FINISHED_STATE, &subq, 1, 0)) {
3528 verbose(VERB_ALGO, "could not generate prime check");
3530 generate_a_aaaa_check(qstate, iq, id);
3533 /* This event is finished. */
3534 qstate->ext_state[id] = module_finished;
3539 * Do final processing on responses to target queries. Events reach this
3540 * state after the iterative resolution algorithm terminates. This state is
3541 * responsible for reactivating the original event, and housekeeping related
3542 * to received target responses (caching, updating the current delegation
3544 * Callback from walk_supers for every super state that is interested in
3545 * the results from this query.
3547 * @param qstate: query state.
3548 * @param id: module id.
3549 * @param forq: super query state.
3552 processTargetResponse(struct module_qstate* qstate, int id,
3553 struct module_qstate* forq)
3555 struct iter_env* ie = (struct iter_env*)qstate->env->modinfo[id];
3556 struct iter_qstate* iq = (struct iter_qstate*)qstate->minfo[id];
3557 struct iter_qstate* foriq = (struct iter_qstate*)forq->minfo[id];
3558 struct ub_packed_rrset_key* rrset;
3559 struct delegpt_ns* dpns;
3560 log_assert(qstate->return_rcode == LDNS_RCODE_NOERROR);
3562 foriq->state = QUERYTARGETS_STATE;
3563 log_query_info(VERB_ALGO, "processTargetResponse", &qstate->qinfo);
3564 log_query_info(VERB_ALGO, "processTargetResponse super", &forq->qinfo);
3566 /* Tell the originating event that this target query has finished
3567 * (regardless if it succeeded or not). */
3568 foriq->num_target_queries--;
3570 /* check to see if parent event is still interested (in orig name). */
3572 verbose(VERB_ALGO, "subq: parent not interested, was reset");
3573 return; /* not interested anymore */
3575 dpns = delegpt_find_ns(foriq->dp, qstate->qinfo.qname,
3576 qstate->qinfo.qname_len);
3578 /* If not interested, just stop processing this event */
3579 verbose(VERB_ALGO, "subq: parent not interested anymore");
3580 /* could be because parent was jostled out of the cache,
3581 and a new identical query arrived, that does not want it*/
3585 /* if iq->query_for_pside_glue then add the pside_glue (marked lame) */
3586 if(iq->pside_glue) {
3587 /* if the pside_glue is NULL, then it could not be found,
3588 * the done_pside is already set when created and a cache
3589 * entry created in processFinished so nothing to do here */
3590 log_rrset_key(VERB_ALGO, "add parentside glue to dp",
3592 if(!delegpt_add_rrset(foriq->dp, forq->region,
3593 iq->pside_glue, 1, NULL))
3594 log_err("out of memory adding pside glue");
3597 /* This response is relevant to the current query, so we
3598 * add (attempt to add, anyway) this target(s) and reactivate
3599 * the original event.
3600 * NOTE: we could only look for the AnswerRRset if the
3601 * response type was ANSWER. */
3602 rrset = reply_find_answer_rrset(&iq->qchase, qstate->return_msg->rep);
3605 /* if CNAMEs have been followed - add new NS to delegpt. */
3606 /* BTW. RFC 1918 says NS should not have got CNAMEs. Robust. */
3607 if(!delegpt_find_ns(foriq->dp, rrset->rk.dname,
3608 rrset->rk.dname_len)) {
3609 /* if dpns->lame then set newcname ns lame too */
3610 if(!delegpt_add_ns(foriq->dp, forq->region,
3611 rrset->rk.dname, dpns->lame, dpns->tls_auth_name,
3613 log_err("out of memory adding cnamed-ns");
3615 /* if dpns->lame then set the address(es) lame too */
3616 if(!delegpt_add_rrset(foriq->dp, forq->region, rrset,
3617 dpns->lame, &additions))
3618 log_err("out of memory adding targets");
3620 /* no new addresses, increase the nxns counter, like
3621 * this could be a list of wildcards with no new
3623 target_count_increase_nx(foriq, 1);
3625 verbose(VERB_ALGO, "added target response");
3626 delegpt_log(VERB_ALGO, foriq->dp);
3628 verbose(VERB_ALGO, "iterator TargetResponse failed");
3629 delegpt_mark_neg(dpns, qstate->qinfo.qtype);
3630 if((dpns->got4 == 2 || (!ie->supports_ipv4 && !ie->use_nat64)) &&
3631 (dpns->got6 == 2 || !ie->supports_ipv6)) {
3632 dpns->resolved = 1; /* fail the target */
3633 /* do not count cached answers */
3634 if(qstate->reply_origin && qstate->reply_origin->len != 0) {
3635 target_count_increase_nx(foriq, 1);
3642 * Process response for DS NS Find queries, that attempt to find the delegation
3643 * point where we ask the DS query from.
3645 * @param qstate: query state.
3646 * @param id: module id.
3647 * @param forq: super query state.
3650 processDSNSResponse(struct module_qstate* qstate, int id,
3651 struct module_qstate* forq)
3653 struct iter_qstate* foriq = (struct iter_qstate*)forq->minfo[id];
3655 /* if the finished (iq->response) query has no NS set: continue
3656 * up to look for the right dp; nothing to change, do DPNSstate */
3657 if(qstate->return_rcode != LDNS_RCODE_NOERROR)
3658 return; /* seek further */
3659 /* find the NS RRset (without allowing CNAMEs) */
3660 if(!reply_find_rrset(qstate->return_msg->rep, qstate->qinfo.qname,
3661 qstate->qinfo.qname_len, LDNS_RR_TYPE_NS,
3662 qstate->qinfo.qclass)){
3663 return; /* seek further */
3666 /* else, store as DP and continue at querytargets */
3667 foriq->state = QUERYTARGETS_STATE;
3668 foriq->dp = delegpt_from_message(qstate->return_msg, forq->region);
3670 log_err("out of memory in dsns dp alloc");
3671 errinf(qstate, "malloc failure, in DS search");
3672 return; /* dp==NULL in QUERYTARGETS makes SERVFAIL */
3674 /* success, go query the querytargets in the new dp (and go down) */
3678 * Process response for qclass=ANY queries for a particular class.
3679 * Append to result or error-exit.
3681 * @param qstate: query state.
3682 * @param id: module id.
3683 * @param forq: super query state.
3686 processClassResponse(struct module_qstate* qstate, int id,
3687 struct module_qstate* forq)
3689 struct iter_qstate* foriq = (struct iter_qstate*)forq->minfo[id];
3690 struct dns_msg* from = qstate->return_msg;
3691 log_query_info(VERB_ALGO, "processClassResponse", &qstate->qinfo);
3692 log_query_info(VERB_ALGO, "processClassResponse super", &forq->qinfo);
3693 if(qstate->return_rcode != LDNS_RCODE_NOERROR) {
3694 /* cause servfail for qclass ANY query */
3695 foriq->response = NULL;
3696 foriq->state = FINISHED_STATE;
3700 if(!foriq->response) {
3701 /* allocate the response: copy RCODE, sec_state */
3702 foriq->response = dns_copy_msg(from, forq->region);
3703 if(!foriq->response) {
3704 log_err("malloc failed for qclass ANY response");
3705 foriq->state = FINISHED_STATE;
3708 foriq->response->qinfo.qclass = forq->qinfo.qclass;
3709 /* qclass ANY does not receive the AA flag on replies */
3710 foriq->response->rep->authoritative = 0;
3712 struct dns_msg* to = foriq->response;
3713 /* add _from_ this response _to_ existing collection */
3714 /* if there are records, copy RCODE */
3715 /* lower sec_state if this message is lower */
3716 if(from->rep->rrset_count != 0) {
3717 size_t n = from->rep->rrset_count+to->rep->rrset_count;
3718 struct ub_packed_rrset_key** dest, **d;
3719 /* copy appropriate rcode */
3720 to->rep->flags = from->rep->flags;
3722 if(from->rep->rrset_count > RR_COUNT_MAX ||
3723 to->rep->rrset_count > RR_COUNT_MAX) {
3724 log_err("malloc failed (too many rrsets) in collect ANY");
3725 foriq->state = FINISHED_STATE;
3726 return; /* integer overflow protection */
3728 dest = regional_alloc(forq->region, sizeof(dest[0])*n);
3730 log_err("malloc failed in collect ANY");
3731 foriq->state = FINISHED_STATE;
3736 memcpy(dest, to->rep->rrsets, to->rep->an_numrrsets
3738 dest += to->rep->an_numrrsets;
3739 memcpy(dest, from->rep->rrsets, from->rep->an_numrrsets
3741 dest += from->rep->an_numrrsets;
3743 memcpy(dest, to->rep->rrsets+to->rep->an_numrrsets,
3744 to->rep->ns_numrrsets * sizeof(dest[0]));
3745 dest += to->rep->ns_numrrsets;
3746 memcpy(dest, from->rep->rrsets+from->rep->an_numrrsets,
3747 from->rep->ns_numrrsets * sizeof(dest[0]));
3748 dest += from->rep->ns_numrrsets;
3750 memcpy(dest, to->rep->rrsets+to->rep->an_numrrsets+
3751 to->rep->ns_numrrsets,
3752 to->rep->ar_numrrsets * sizeof(dest[0]));
3753 dest += to->rep->ar_numrrsets;
3754 memcpy(dest, from->rep->rrsets+from->rep->an_numrrsets+
3755 from->rep->ns_numrrsets,
3756 from->rep->ar_numrrsets * sizeof(dest[0]));
3758 to->rep->rrsets = d;
3759 to->rep->an_numrrsets += from->rep->an_numrrsets;
3760 to->rep->ns_numrrsets += from->rep->ns_numrrsets;
3761 to->rep->ar_numrrsets += from->rep->ar_numrrsets;
3762 to->rep->rrset_count = n;
3764 if(from->rep->security < to->rep->security) /* lowest sec */
3765 to->rep->security = from->rep->security;
3766 if(from->rep->qdcount != 0) /* insert qd if appropriate */
3767 to->rep->qdcount = from->rep->qdcount;
3768 if(from->rep->ttl < to->rep->ttl) /* use smallest TTL */
3769 to->rep->ttl = from->rep->ttl;
3770 if(from->rep->prefetch_ttl < to->rep->prefetch_ttl)
3771 to->rep->prefetch_ttl = from->rep->prefetch_ttl;
3772 if(from->rep->serve_expired_ttl < to->rep->serve_expired_ttl)
3773 to->rep->serve_expired_ttl = from->rep->serve_expired_ttl;
3776 foriq->num_current_queries --;
3777 if(foriq->num_current_queries == 0)
3778 foriq->state = FINISHED_STATE;
3782 * Collect class ANY responses and make them into one response. This
3783 * state is started and it creates queries for all classes (that have
3784 * root hints). The answers are then collected.
3786 * @param qstate: query state.
3787 * @param id: module id.
3788 * @return true if the event needs more immediate processing, false if not.
3791 processCollectClass(struct module_qstate* qstate, int id)
3793 struct iter_qstate* iq = (struct iter_qstate*)qstate->minfo[id];
3794 struct module_qstate* subq;
3795 /* If qchase.qclass == 0 then send out queries for all classes.
3796 * Otherwise, do nothing (wait for all answers to arrive and the
3797 * processClassResponse to put them together, and that moves us
3798 * towards the Finished state when done. */
3799 if(iq->qchase.qclass == 0) {
3801 iq->qchase.qclass = LDNS_RR_CLASS_ANY;
3802 while(iter_get_next_root(qstate->env->hints,
3803 qstate->env->fwds, &c)) {
3804 /* generate query for this class */
3805 log_nametypeclass(VERB_ALGO, "spawn collect query",
3806 qstate->qinfo.qname, qstate->qinfo.qtype, c);
3807 if(!generate_sub_request(qstate->qinfo.qname,
3808 qstate->qinfo.qname_len, qstate->qinfo.qtype,
3809 c, qstate, id, iq, INIT_REQUEST_STATE,
3810 FINISHED_STATE, &subq,
3811 (int)!(qstate->query_flags&BIT_CD), 0)) {
3812 errinf(qstate, "could not generate class ANY"
3814 return error_response(qstate, id,
3815 LDNS_RCODE_SERVFAIL);
3817 /* ignore subq, no special init required */
3818 iq->num_current_queries ++;
3823 /* if no roots are configured at all, return */
3824 if(iq->num_current_queries == 0) {
3825 verbose(VERB_ALGO, "No root hints or fwds, giving up "
3827 return error_response(qstate, id, LDNS_RCODE_REFUSED);
3829 /* return false, wait for queries to return */
3831 /* if woke up here because of an answer, wait for more answers */
3836 * This handles the final state for first-tier responses (i.e., responses to
3837 * externally generated queries).
3839 * @param qstate: query state.
3840 * @param iq: iterator query state.
3841 * @param id: module id.
3842 * @return true if the event needs more processing, false if not. Since this
3843 * is the final state for an event, it always returns false.
3846 processFinished(struct module_qstate* qstate, struct iter_qstate* iq,
3849 log_query_info(VERB_QUERY, "finishing processing for",
3852 /* store negative cache element for parent side glue. */
3853 if(!qstate->no_cache_store && iq->query_for_pside_glue
3855 iter_store_parentside_neg(qstate->env, &qstate->qinfo,
3856 iq->deleg_msg?iq->deleg_msg->rep:
3857 (iq->response?iq->response->rep:NULL));
3859 verbose(VERB_ALGO, "No response is set, servfail");
3860 errinf(qstate, "(no response found at query finish)");
3861 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
3864 /* Make sure that the RA flag is set (since the presence of
3865 * this module means that recursion is available) */
3866 iq->response->rep->flags |= BIT_RA;
3868 /* Clear the AA flag */
3869 /* FIXME: does this action go here or in some other module? */
3870 iq->response->rep->flags &= ~BIT_AA;
3872 /* make sure QR flag is on */
3873 iq->response->rep->flags |= BIT_QR;
3875 /* explicitly set the EDE string to NULL */
3876 iq->response->rep->reason_bogus_str = NULL;
3878 /* we have finished processing this query */
3879 qstate->ext_state[id] = module_finished;
3881 /* TODO: we are using a private TTL, trim the response. */
3882 /* if (mPrivateTTL > 0){IterUtils.setPrivateTTL(resp, mPrivateTTL); } */
3884 /* prepend any items we have accumulated */
3885 if(iq->an_prepend_list || iq->ns_prepend_list) {
3886 if(!iter_prepend(iq, iq->response, qstate->region)) {
3887 log_err("prepend rrsets: out of memory");
3888 return error_response(qstate, id, LDNS_RCODE_SERVFAIL);
3890 /* reset the query name back */
3891 iq->response->qinfo = qstate->qinfo;
3892 /* the security state depends on the combination */
3893 iq->response->rep->security = sec_status_unchecked;
3894 /* store message with the finished prepended items,
3895 * but only if we did recursion. The nonrecursion referral
3896 * from cache does not need to be stored in the msg cache. */
3897 if(!qstate->no_cache_store && qstate->query_flags&BIT_RD) {
3898 iter_dns_store(qstate->env, &qstate->qinfo,
3899 iq->response->rep, 0, qstate->prefetch_leeway,
3900 iq->dp&&iq->dp->has_parent_side_NS,
3901 qstate->region, qstate->query_flags,
3902 qstate->qstarttime);
3905 qstate->return_rcode = LDNS_RCODE_NOERROR;
3906 qstate->return_msg = iq->response;
3911 * Return priming query results to interested super querystates.
3913 * Sets the delegation point and delegation message (not nonRD queries).
3914 * This is a callback from walk_supers.
3916 * @param qstate: query state that finished.
3917 * @param id: module id.
3918 * @param super: the qstate to inform.
3921 iter_inform_super(struct module_qstate* qstate, int id,
3922 struct module_qstate* super)
3924 if(!qstate->is_priming && super->qinfo.qclass == LDNS_RR_CLASS_ANY)
3925 processClassResponse(qstate, id, super);
3926 else if(super->qinfo.qtype == LDNS_RR_TYPE_DS && ((struct iter_qstate*)
3927 super->minfo[id])->state == DSNS_FIND_STATE)
3928 processDSNSResponse(qstate, id, super);
3929 else if(qstate->return_rcode != LDNS_RCODE_NOERROR)
3930 error_supers(qstate, id, super);
3931 else if(qstate->is_priming)
3932 prime_supers(qstate, id, super);
3933 else processTargetResponse(qstate, id, super);
3937 * Handle iterator state.
3938 * Handle events. This is the real processing loop for events, responsible
3939 * for moving events through the various states. If a processing method
3940 * returns true, then it will be advanced to the next state. If false, then
3941 * processing will stop.
3943 * @param qstate: query state.
3944 * @param ie: iterator shared global environment.
3945 * @param iq: iterator query state.
3946 * @param id: module id.
3949 iter_handle(struct module_qstate* qstate, struct iter_qstate* iq,
3950 struct iter_env* ie, int id)
3954 verbose(VERB_ALGO, "iter_handle processing q with state %s",
3955 iter_state_to_string(iq->state));
3957 case INIT_REQUEST_STATE:
3958 cont = processInitRequest(qstate, iq, ie, id);
3960 case INIT_REQUEST_2_STATE:
3961 cont = processInitRequest2(qstate, iq, id);
3963 case INIT_REQUEST_3_STATE:
3964 cont = processInitRequest3(qstate, iq, id);
3966 case QUERYTARGETS_STATE:
3967 cont = processQueryTargets(qstate, iq, ie, id);
3969 case QUERY_RESP_STATE:
3970 cont = processQueryResponse(qstate, iq, ie, id);
3972 case PRIME_RESP_STATE:
3973 cont = processPrimeResponse(qstate, id);
3975 case COLLECT_CLASS_STATE:
3976 cont = processCollectClass(qstate, id);
3978 case DSNS_FIND_STATE:
3979 cont = processDSNSFind(qstate, iq, id);
3981 case FINISHED_STATE:
3982 cont = processFinished(qstate, iq, id);
3985 log_warn("iterator: invalid state: %d",
3994 * This is the primary entry point for processing request events. Note that
3995 * this method should only be used by external modules.
3996 * @param qstate: query state.
3997 * @param ie: iterator shared global environment.
3998 * @param iq: iterator query state.
3999 * @param id: module id.
4002 process_request(struct module_qstate* qstate, struct iter_qstate* iq,
4003 struct iter_env* ie, int id)
4005 /* external requests start in the INIT state, and finish using the
4006 * FINISHED state. */
4007 iq->state = INIT_REQUEST_STATE;
4008 iq->final_state = FINISHED_STATE;
4009 verbose(VERB_ALGO, "process_request: new external request event");
4010 iter_handle(qstate, iq, ie, id);
4013 /** process authoritative server reply */
4015 process_response(struct module_qstate* qstate, struct iter_qstate* iq,
4016 struct iter_env* ie, int id, struct outbound_entry* outbound,
4017 enum module_ev event)
4019 struct msg_parse* prs;
4020 struct edns_data edns;
4023 verbose(VERB_ALGO, "process_response: new external response event");
4024 iq->response = NULL;
4025 iq->state = QUERY_RESP_STATE;
4026 if(event == module_event_noreply || event == module_event_error) {
4027 if(event == module_event_noreply && iq->timeout_count >= 3 &&
4028 qstate->env->cfg->use_caps_bits_for_id &&
4029 !iq->caps_fallback && !is_caps_whitelisted(ie, iq)) {
4030 /* start fallback */
4031 iq->caps_fallback = 1;
4032 iq->caps_server = 0;
4033 iq->caps_reply = NULL;
4034 iq->caps_response = NULL;
4035 iq->caps_minimisation_state = DONOT_MINIMISE_STATE;
4036 iq->state = QUERYTARGETS_STATE;
4037 iq->num_current_queries--;
4038 /* need fresh attempts for the 0x20 fallback, if
4039 * that was the cause for the failure */
4040 iter_dec_attempts(iq->dp, 3, ie->outbound_msg_retry);
4041 verbose(VERB_DETAIL, "Capsforid: timeouts, starting fallback");
4046 if( (event != module_event_reply && event != module_event_capsfail)
4047 || !qstate->reply) {
4048 log_err("Bad event combined with response");
4049 outbound_list_remove(&iq->outlist, outbound);
4050 errinf(qstate, "module iterator received wrong internal event with a response message");
4051 (void)error_response(qstate, id, LDNS_RCODE_SERVFAIL);
4056 fill_fail_addr(iq, &qstate->reply->remote_addr,
4057 qstate->reply->remote_addrlen);
4058 prs = (struct msg_parse*)regional_alloc(qstate->env->scratch,
4059 sizeof(struct msg_parse));
4061 log_err("out of memory on incoming message");
4062 /* like packet got dropped */
4065 memset(prs, 0, sizeof(*prs));
4066 memset(&edns, 0, sizeof(edns));
4067 pkt = qstate->reply->c->buffer;
4068 sldns_buffer_set_position(pkt, 0);
4069 if(parse_packet(pkt, prs, qstate->env->scratch) != LDNS_RCODE_NOERROR) {
4070 verbose(VERB_ALGO, "parse error on reply packet");
4071 iq->parse_failures++;
4074 /* edns is not examined, but removed from message to help cache */
4075 if(parse_extract_edns_from_response_msg(prs, &edns, qstate->env->scratch) !=
4076 LDNS_RCODE_NOERROR) {
4077 iq->parse_failures++;
4081 /* Copy the edns options we may got from the back end */
4082 if(edns.opt_list_in) {
4083 qstate->edns_opts_back_in = edns_opt_copy_region(edns.opt_list_in,
4085 if(!qstate->edns_opts_back_in) {
4086 log_err("out of memory on incoming message");
4087 /* like packet got dropped */
4090 if(!inplace_cb_edns_back_parsed_call(qstate->env, qstate)) {
4091 log_err("unable to call edns_back_parsed callback");
4096 /* remove CD-bit, we asked for in case we handle validation ourself */
4097 prs->flags &= ~BIT_CD;
4099 /* normalize and sanitize: easy to delete items from linked lists */
4100 if(!scrub_message(pkt, prs, &iq->qinfo_out, iq->dp->name,
4101 qstate->env->scratch, qstate->env, ie)) {
4102 /* if 0x20 enabled, start fallback, but we have no message */
4103 if(event == module_event_capsfail && !iq->caps_fallback) {
4104 iq->caps_fallback = 1;
4105 iq->caps_server = 0;
4106 iq->caps_reply = NULL;
4107 iq->caps_response = NULL;
4108 iq->caps_minimisation_state = DONOT_MINIMISE_STATE;
4109 iq->state = QUERYTARGETS_STATE;
4110 iq->num_current_queries--;
4111 verbose(VERB_DETAIL, "Capsforid: scrub failed, starting fallback with no response");
4113 iq->scrub_failures++;
4117 /* allocate response dns_msg in region */
4118 iq->response = dns_alloc_msg(pkt, prs, qstate->region);
4121 log_query_info(VERB_DETAIL, "response for", &qstate->qinfo);
4122 log_name_addr(VERB_DETAIL, "reply from", iq->dp->name,
4123 &qstate->reply->remote_addr, qstate->reply->remote_addrlen);
4124 if(verbosity >= VERB_ALGO)
4125 log_dns_msg("incoming scrubbed packet:", &iq->response->qinfo,
4128 if(event == module_event_capsfail || iq->caps_fallback) {
4129 if(qstate->env->cfg->qname_minimisation &&
4130 iq->minimisation_state != DONOT_MINIMISE_STATE) {
4131 /* Skip QNAME minimisation for next query, since that
4132 * one has to match the current query. */
4133 iq->minimisation_state = SKIP_MINIMISE_STATE;
4135 /* for fallback we care about main answer, not additionals */
4136 /* removing that makes comparison more likely to succeed */
4137 caps_strip_reply(iq->response->rep);
4139 if(iq->caps_fallback &&
4140 iq->caps_minimisation_state != iq->minimisation_state) {
4141 /* QNAME minimisation state has changed, restart caps
4143 iq->caps_fallback = 0;
4146 if(!iq->caps_fallback) {
4147 /* start fallback */
4148 iq->caps_fallback = 1;
4149 iq->caps_server = 0;
4150 iq->caps_reply = iq->response->rep;
4151 iq->caps_response = iq->response;
4152 iq->caps_minimisation_state = iq->minimisation_state;
4153 iq->state = QUERYTARGETS_STATE;
4154 iq->num_current_queries--;
4155 verbose(VERB_DETAIL, "Capsforid: starting fallback");
4158 /* check if reply is the same, otherwise, fail */
4159 if(!iq->caps_reply) {
4160 iq->caps_reply = iq->response->rep;
4161 iq->caps_response = iq->response;
4162 iq->caps_server = -1; /*become zero at ++,
4163 so that we start the full set of trials */
4164 } else if(caps_failed_rcode(iq->caps_reply) &&
4165 !caps_failed_rcode(iq->response->rep)) {
4166 /* prefer to upgrade to non-SERVFAIL */
4167 iq->caps_reply = iq->response->rep;
4168 iq->caps_response = iq->response;
4169 } else if(!caps_failed_rcode(iq->caps_reply) &&
4170 caps_failed_rcode(iq->response->rep)) {
4171 /* if we have non-SERVFAIL as answer then
4172 * we can ignore SERVFAILs for the equality
4174 /* no instructions here, skip other else */
4175 } else if(caps_failed_rcode(iq->caps_reply) &&
4176 caps_failed_rcode(iq->response->rep)) {
4177 /* failure is same as other failure in fallbk*/
4178 /* no instructions here, skip other else */
4179 } else if(!reply_equal(iq->response->rep, iq->caps_reply,
4180 qstate->env->scratch)) {
4181 verbose(VERB_DETAIL, "Capsforid fallback: "
4182 "getting different replies, failed");
4183 outbound_list_remove(&iq->outlist, outbound);
4184 errinf(qstate, "0x20 failed, then got different replies in fallback");
4185 (void)error_response(qstate, id,
4186 LDNS_RCODE_SERVFAIL);
4189 /* continue the fallback procedure at next server */
4191 iq->state = QUERYTARGETS_STATE;
4192 iq->num_current_queries--;
4193 verbose(VERB_DETAIL, "Capsforid: reply is equal. "
4194 "go to next fallback");
4198 iq->caps_fallback = 0; /* if we were in fallback, 0x20 is OK now */
4201 outbound_list_remove(&iq->outlist, outbound);
4202 iter_handle(qstate, iq, ie, id);
4206 iter_operate(struct module_qstate* qstate, enum module_ev event, int id,
4207 struct outbound_entry* outbound)
4209 struct iter_env* ie = (struct iter_env*)qstate->env->modinfo[id];
4210 struct iter_qstate* iq = (struct iter_qstate*)qstate->minfo[id];
4211 verbose(VERB_QUERY, "iterator[module %d] operate: extstate:%s event:%s",
4212 id, strextstate(qstate->ext_state[id]), strmodulevent(event));
4213 if(iq) log_query_info(VERB_QUERY, "iterator operate: query",
4215 if(iq && qstate->qinfo.qname != iq->qchase.qname)
4216 log_query_info(VERB_QUERY, "iterator operate: chased to",
4219 /* perform iterator state machine */
4220 if((event == module_event_new || event == module_event_pass) &&
4222 if(!iter_new(qstate, id)) {
4223 errinf(qstate, "malloc failure, new iterator module allocation");
4224 (void)error_response(qstate, id, LDNS_RCODE_SERVFAIL);
4227 iq = (struct iter_qstate*)qstate->minfo[id];
4228 process_request(qstate, iq, ie, id);
4231 if(iq && event == module_event_pass) {
4232 iter_handle(qstate, iq, ie, id);
4235 if(iq && outbound) {
4236 process_response(qstate, iq, ie, id, outbound, event);
4239 if(event == module_event_error) {
4240 verbose(VERB_ALGO, "got called with event error, giving up");
4241 errinf(qstate, "iterator module got the error event");
4242 (void)error_response(qstate, id, LDNS_RCODE_SERVFAIL);
4246 log_err("bad event for iterator");
4247 errinf(qstate, "iterator module received wrong event");
4248 (void)error_response(qstate, id, LDNS_RCODE_SERVFAIL);
4252 iter_clear(struct module_qstate* qstate, int id)
4254 struct iter_qstate* iq;
4257 iq = (struct iter_qstate*)qstate->minfo[id];
4259 outbound_list_clear(&iq->outlist);
4260 if(iq->target_count && --iq->target_count[TARGET_COUNT_REF] == 0) {
4261 free(iq->target_count);
4262 if(*iq->nxns_dp) free(*iq->nxns_dp);
4265 iq->num_current_queries = 0;
4267 qstate->minfo[id] = NULL;
4271 iter_get_mem(struct module_env* env, int id)
4273 struct iter_env* ie = (struct iter_env*)env->modinfo[id];
4276 return sizeof(*ie) + sizeof(int)*((size_t)ie->max_dependency_depth+1)
4277 + donotq_get_mem(ie->donotq) + priv_get_mem(ie->priv);
4281 * The iterator function block
4283 static struct module_func_block iter_block = {
4285 &iter_init, &iter_deinit, &iter_operate, &iter_inform_super,
4286 &iter_clear, &iter_get_mem
4289 struct module_func_block*
4290 iter_get_funcblock(void)
4296 iter_state_to_string(enum iter_state state)
4300 case INIT_REQUEST_STATE :
4301 return "INIT REQUEST STATE";
4302 case INIT_REQUEST_2_STATE :
4303 return "INIT REQUEST STATE (stage 2)";
4304 case INIT_REQUEST_3_STATE:
4305 return "INIT REQUEST STATE (stage 3)";
4306 case QUERYTARGETS_STATE :
4307 return "QUERY TARGETS STATE";
4308 case PRIME_RESP_STATE :
4309 return "PRIME RESPONSE STATE";
4310 case COLLECT_CLASS_STATE :
4311 return "COLLECT CLASS STATE";
4312 case DSNS_FIND_STATE :
4313 return "DSNS FIND STATE";
4314 case QUERY_RESP_STATE :
4315 return "QUERY RESPONSE STATE";
4316 case FINISHED_STATE :
4317 return "FINISHED RESPONSE STATE";
4319 return "UNKNOWN ITER STATE";
4324 iter_state_is_responsestate(enum iter_state s)
4327 case INIT_REQUEST_STATE :
4328 case INIT_REQUEST_2_STATE :
4329 case INIT_REQUEST_3_STATE :
4330 case QUERYTARGETS_STATE :
4331 case COLLECT_CLASS_STATE :