iter_utils.c revision 307729
1/*
2 * iterator/iter_utils.c - iterative resolver module utility functions.
3 *
4 * Copyright (c) 2007, NLnet Labs. All rights reserved.
5 *
6 * This software is open source.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 *
12 * Redistributions of source code must retain the above copyright notice,
13 * this list of conditions and the following disclaimer.
14 *
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.
18 *
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.
22 *
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.
34 */
35
36/**
37 * \file
38 *
39 * This file contains functions to assist the iterator module.
40 * Configuration options. Forward zones.
41 */
42#include "config.h"
43#include "iterator/iter_utils.h"
44#include "iterator/iterator.h"
45#include "iterator/iter_hints.h"
46#include "iterator/iter_fwd.h"
47#include "iterator/iter_donotq.h"
48#include "iterator/iter_delegpt.h"
49#include "iterator/iter_priv.h"
50#include "services/cache/infra.h"
51#include "services/cache/dns.h"
52#include "services/cache/rrset.h"
53#include "util/net_help.h"
54#include "util/module.h"
55#include "util/log.h"
56#include "util/config_file.h"
57#include "util/regional.h"
58#include "util/data/msgparse.h"
59#include "util/data/dname.h"
60#include "util/random.h"
61#include "util/fptr_wlist.h"
62#include "validator/val_anchor.h"
63#include "validator/val_kcache.h"
64#include "validator/val_kentry.h"
65#include "validator/val_utils.h"
66#include "validator/val_sigcrypt.h"
67#include "sldns/sbuffer.h"
68#include "sldns/str2wire.h"
69
70/** time when nameserver glue is said to be 'recent' */
71#define SUSPICION_RECENT_EXPIRY 86400
72/** penalty to validation failed blacklisted IPs */
73#define BLACKLIST_PENALTY (USEFUL_SERVER_TOP_TIMEOUT*4)
74
75/** fillup fetch policy array */
76static void
77fetch_fill(struct iter_env* ie, const char* str)
78{
79	char* s = (char*)str, *e;
80	int i;
81	for(i=0; i<ie->max_dependency_depth+1; i++) {
82		ie->target_fetch_policy[i] = strtol(s, &e, 10);
83		if(s == e)
84			fatal_exit("cannot parse fetch policy number %s", s);
85		s = e;
86	}
87}
88
89/** Read config string that represents the target fetch policy */
90static int
91read_fetch_policy(struct iter_env* ie, const char* str)
92{
93	int count = cfg_count_numbers(str);
94	if(count < 1) {
95		log_err("Cannot parse target fetch policy: \"%s\"", str);
96		return 0;
97	}
98	ie->max_dependency_depth = count - 1;
99	ie->target_fetch_policy = (int*)calloc(
100		(size_t)ie->max_dependency_depth+1, sizeof(int));
101	if(!ie->target_fetch_policy) {
102		log_err("alloc fetch policy: out of memory");
103		return 0;
104	}
105	fetch_fill(ie, str);
106	return 1;
107}
108
109/** apply config caps whitelist items to name tree */
110static int
111caps_white_apply_cfg(rbtree_t* ntree, struct config_file* cfg)
112{
113	struct config_strlist* p;
114	for(p=cfg->caps_whitelist; p; p=p->next) {
115		struct name_tree_node* n;
116		size_t len;
117		uint8_t* nm = sldns_str2wire_dname(p->str, &len);
118		if(!nm) {
119			log_err("could not parse %s", p->str);
120			return 0;
121		}
122		n = (struct name_tree_node*)calloc(1, sizeof(*n));
123		if(!n) {
124			log_err("out of memory");
125			free(nm);
126			return 0;
127		}
128		n->node.key = n;
129		n->name = nm;
130		n->len = len;
131		n->labs = dname_count_labels(nm);
132		n->dclass = LDNS_RR_CLASS_IN;
133		if(!name_tree_insert(ntree, n, nm, len, n->labs, n->dclass)) {
134			/* duplicate element ignored, idempotent */
135			free(n->name);
136			free(n);
137		}
138	}
139	name_tree_init_parents(ntree);
140	return 1;
141}
142
143int
144iter_apply_cfg(struct iter_env* iter_env, struct config_file* cfg)
145{
146	int i;
147	/* target fetch policy */
148	if(!read_fetch_policy(iter_env, cfg->target_fetch_policy))
149		return 0;
150	for(i=0; i<iter_env->max_dependency_depth+1; i++)
151		verbose(VERB_QUERY, "target fetch policy for level %d is %d",
152			i, iter_env->target_fetch_policy[i]);
153
154	if(!iter_env->donotq)
155		iter_env->donotq = donotq_create();
156	if(!iter_env->donotq || !donotq_apply_cfg(iter_env->donotq, cfg)) {
157		log_err("Could not set donotqueryaddresses");
158		return 0;
159	}
160	if(!iter_env->priv)
161		iter_env->priv = priv_create();
162	if(!iter_env->priv || !priv_apply_cfg(iter_env->priv, cfg)) {
163		log_err("Could not set private addresses");
164		return 0;
165	}
166	if(cfg->caps_whitelist) {
167		if(!iter_env->caps_white)
168			iter_env->caps_white = rbtree_create(name_tree_compare);
169		if(!iter_env->caps_white || !caps_white_apply_cfg(
170			iter_env->caps_white, cfg)) {
171			log_err("Could not set capsforid whitelist");
172			return 0;
173		}
174
175	}
176	iter_env->supports_ipv6 = cfg->do_ip6;
177	iter_env->supports_ipv4 = cfg->do_ip4;
178	return 1;
179}
180
181/** filter out unsuitable targets
182 * @param iter_env: iterator environment with ipv6-support flag.
183 * @param env: module environment with infra cache.
184 * @param name: zone name
185 * @param namelen: length of name
186 * @param qtype: query type (host order).
187 * @param now: current time
188 * @param a: address in delegation point we are examining.
189 * @return an integer that signals the target suitability.
190 *	as follows:
191 *	-1: The address should be omitted from the list.
192 *	    Because:
193 *		o The address is bogus (DNSSEC validation failure).
194 *		o Listed as donotquery
195 *		o is ipv6 but no ipv6 support (in operating system).
196 *		o is ipv4 but no ipv4 support (in operating system).
197 *		o is lame
198 *	Otherwise, an rtt in milliseconds.
199 *	0 .. USEFUL_SERVER_TOP_TIMEOUT-1
200 *		The roundtrip time timeout estimate. less than 2 minutes.
201 *		Note that util/rtt.c has a MIN_TIMEOUT of 50 msec, thus
202 *		values 0 .. 49 are not used, unless that is changed.
203 *	USEFUL_SERVER_TOP_TIMEOUT
204 *		This value exactly is given for unresponsive blacklisted.
205 *	USEFUL_SERVER_TOP_TIMEOUT+1
206 *		For non-blacklisted servers: huge timeout, but has traffic.
207 *	USEFUL_SERVER_TOP_TIMEOUT*1 ..
208 *		parent-side lame servers get this penalty. A dispreferential
209 *		server. (lame in delegpt).
210 *	USEFUL_SERVER_TOP_TIMEOUT*2 ..
211 *		dnsseclame servers get penalty
212 *	USEFUL_SERVER_TOP_TIMEOUT*3 ..
213 *		recursion lame servers get penalty
214 *	UNKNOWN_SERVER_NICENESS
215 *		If no information is known about the server, this is
216 *		returned. 376 msec or so.
217 *	+BLACKLIST_PENALTY (of USEFUL_TOP_TIMEOUT*4) for dnssec failed IPs.
218 *
219 * When a final value is chosen that is dnsseclame ; dnsseclameness checking
220 * is turned off (so we do not discard the reply).
221 * When a final value is chosen that is recursionlame; RD bit is set on query.
222 * Because of the numbers this means recursionlame also have dnssec lameness
223 * checking turned off.
224 */
225static int
226iter_filter_unsuitable(struct iter_env* iter_env, struct module_env* env,
227	uint8_t* name, size_t namelen, uint16_t qtype, time_t now,
228	struct delegpt_addr* a)
229{
230	int rtt, lame, reclame, dnsseclame;
231	if(a->bogus)
232		return -1; /* address of server is bogus */
233	if(donotq_lookup(iter_env->donotq, &a->addr, a->addrlen)) {
234		log_addr(VERB_ALGO, "skip addr on the donotquery list",
235			&a->addr, a->addrlen);
236		return -1; /* server is on the donotquery list */
237	}
238	if(!iter_env->supports_ipv6 && addr_is_ip6(&a->addr, a->addrlen)) {
239		return -1; /* there is no ip6 available */
240	}
241	if(!iter_env->supports_ipv4 && !addr_is_ip6(&a->addr, a->addrlen)) {
242		return -1; /* there is no ip4 available */
243	}
244	/* check lameness - need zone , class info */
245	if(infra_get_lame_rtt(env->infra_cache, &a->addr, a->addrlen,
246		name, namelen, qtype, &lame, &dnsseclame, &reclame,
247		&rtt, now)) {
248		log_addr(VERB_ALGO, "servselect", &a->addr, a->addrlen);
249		verbose(VERB_ALGO, "   rtt=%d%s%s%s%s", rtt,
250			lame?" LAME":"",
251			dnsseclame?" DNSSEC_LAME":"",
252			reclame?" REC_LAME":"",
253			a->lame?" ADDR_LAME":"");
254		if(lame)
255			return -1; /* server is lame */
256		else if(rtt >= USEFUL_SERVER_TOP_TIMEOUT)
257			/* server is unresponsive,
258			 * we used to return TOP_TIMEOUT, but fairly useless,
259			 * because if == TOP_TIMEOUT is dropped because
260			 * blacklisted later, instead, remove it here, so
261			 * other choices (that are not blacklisted) can be
262			 * tried */
263			return -1;
264		/* select remainder from worst to best */
265		else if(reclame)
266			return rtt+USEFUL_SERVER_TOP_TIMEOUT*3; /* nonpref */
267		else if(dnsseclame || a->dnsseclame)
268			return rtt+USEFUL_SERVER_TOP_TIMEOUT*2; /* nonpref */
269		else if(a->lame)
270			return rtt+USEFUL_SERVER_TOP_TIMEOUT+1; /* nonpref */
271		else	return rtt;
272	}
273	/* no server information present */
274	if(a->dnsseclame)
275		return UNKNOWN_SERVER_NICENESS+USEFUL_SERVER_TOP_TIMEOUT*2; /* nonpref */
276	else if(a->lame)
277		return USEFUL_SERVER_TOP_TIMEOUT+1+UNKNOWN_SERVER_NICENESS; /* nonpref */
278	return UNKNOWN_SERVER_NICENESS;
279}
280
281/** lookup RTT information, and also store fastest rtt (if any) */
282static int
283iter_fill_rtt(struct iter_env* iter_env, struct module_env* env,
284	uint8_t* name, size_t namelen, uint16_t qtype, time_t now,
285	struct delegpt* dp, int* best_rtt, struct sock_list* blacklist)
286{
287	int got_it = 0;
288	struct delegpt_addr* a;
289	if(dp->bogus)
290		return 0; /* NS bogus, all bogus, nothing found */
291	for(a=dp->result_list; a; a = a->next_result) {
292		a->sel_rtt = iter_filter_unsuitable(iter_env, env,
293			name, namelen, qtype, now, a);
294		if(a->sel_rtt != -1) {
295			if(sock_list_find(blacklist, &a->addr, a->addrlen))
296				a->sel_rtt += BLACKLIST_PENALTY;
297
298			if(!got_it) {
299				*best_rtt = a->sel_rtt;
300				got_it = 1;
301			} else if(a->sel_rtt < *best_rtt) {
302				*best_rtt = a->sel_rtt;
303			}
304		}
305	}
306	return got_it;
307}
308
309/** filter the address list, putting best targets at front,
310 * returns number of best targets (or 0, no suitable targets) */
311static int
312iter_filter_order(struct iter_env* iter_env, struct module_env* env,
313	uint8_t* name, size_t namelen, uint16_t qtype, time_t now,
314	struct delegpt* dp, int* selected_rtt, int open_target,
315	struct sock_list* blacklist)
316{
317	int got_num = 0, low_rtt = 0, swap_to_front;
318	struct delegpt_addr* a, *n, *prev=NULL;
319
320	/* fillup sel_rtt and find best rtt in the bunch */
321	got_num = iter_fill_rtt(iter_env, env, name, namelen, qtype, now, dp,
322		&low_rtt, blacklist);
323	if(got_num == 0)
324		return 0;
325	if(low_rtt >= USEFUL_SERVER_TOP_TIMEOUT &&
326		(delegpt_count_missing_targets(dp) > 0 || open_target > 0)) {
327		verbose(VERB_ALGO, "Bad choices, trying to get more choice");
328		return 0; /* we want more choice. The best choice is a bad one.
329			     return 0 to force the caller to fetch more */
330	}
331
332	got_num = 0;
333	a = dp->result_list;
334	while(a) {
335		/* skip unsuitable targets */
336		if(a->sel_rtt == -1) {
337			prev = a;
338			a = a->next_result;
339			continue;
340		}
341		/* classify the server address and determine what to do */
342		swap_to_front = 0;
343		if(a->sel_rtt >= low_rtt && a->sel_rtt - low_rtt <= RTT_BAND) {
344			got_num++;
345			swap_to_front = 1;
346		} else if(a->sel_rtt<low_rtt && low_rtt-a->sel_rtt<=RTT_BAND) {
347			got_num++;
348			swap_to_front = 1;
349		}
350		/* swap to front if necessary, or move to next result */
351		if(swap_to_front && prev) {
352			n = a->next_result;
353			prev->next_result = n;
354			a->next_result = dp->result_list;
355			dp->result_list = a;
356			a = n;
357		} else {
358			prev = a;
359			a = a->next_result;
360		}
361	}
362	*selected_rtt = low_rtt;
363
364	if (env->cfg->prefer_ip6) {
365		int got_num6 = 0;
366		int low_rtt6 = 0;
367		int i;
368		prev = NULL;
369		a = dp->result_list;
370		for(i = 0; i < got_num; i++) {
371			swap_to_front = 0;
372			if(a->addr.ss_family == AF_INET6) {
373				got_num6++;
374				swap_to_front = 1;
375				if(low_rtt6 == 0 || a->sel_rtt < low_rtt6) {
376					low_rtt6 = a->sel_rtt;
377				}
378			}
379			/* swap to front if IPv6, or move to next result */
380			if(swap_to_front && prev) {
381				n = a->next_result;
382				prev->next_result = n;
383				a->next_result = dp->result_list;
384				dp->result_list = a;
385				a = n;
386			} else {
387				prev = a;
388				a = a->next_result;
389			}
390		}
391		if(got_num6 > 0) {
392			got_num = got_num6;
393			*selected_rtt = low_rtt6;
394		}
395	}
396	return got_num;
397}
398
399struct delegpt_addr*
400iter_server_selection(struct iter_env* iter_env,
401	struct module_env* env, struct delegpt* dp,
402	uint8_t* name, size_t namelen, uint16_t qtype, int* dnssec_lame,
403	int* chase_to_rd, int open_target, struct sock_list* blacklist)
404{
405	int sel;
406	int selrtt;
407	struct delegpt_addr* a, *prev;
408	int num = iter_filter_order(iter_env, env, name, namelen, qtype,
409		*env->now, dp, &selrtt, open_target, blacklist);
410
411	if(num == 0)
412		return NULL;
413	verbose(VERB_ALGO, "selrtt %d", selrtt);
414	if(selrtt > BLACKLIST_PENALTY) {
415		if(selrtt-BLACKLIST_PENALTY > USEFUL_SERVER_TOP_TIMEOUT*3) {
416			verbose(VERB_ALGO, "chase to "
417				"blacklisted recursion lame server");
418			*chase_to_rd = 1;
419		}
420		if(selrtt-BLACKLIST_PENALTY > USEFUL_SERVER_TOP_TIMEOUT*2) {
421			verbose(VERB_ALGO, "chase to "
422				"blacklisted dnssec lame server");
423			*dnssec_lame = 1;
424		}
425	} else {
426		if(selrtt > USEFUL_SERVER_TOP_TIMEOUT*3) {
427			verbose(VERB_ALGO, "chase to recursion lame server");
428			*chase_to_rd = 1;
429		}
430		if(selrtt > USEFUL_SERVER_TOP_TIMEOUT*2) {
431			verbose(VERB_ALGO, "chase to dnssec lame server");
432			*dnssec_lame = 1;
433		}
434		if(selrtt == USEFUL_SERVER_TOP_TIMEOUT) {
435			verbose(VERB_ALGO, "chase to blacklisted lame server");
436			return NULL;
437		}
438	}
439
440	if(num == 1) {
441		a = dp->result_list;
442		if(++a->attempts < OUTBOUND_MSG_RETRY)
443			return a;
444		dp->result_list = a->next_result;
445		return a;
446	}
447
448	/* randomly select a target from the list */
449	log_assert(num > 1);
450	/* grab secure random number, to pick unexpected server.
451	 * also we need it to be threadsafe. */
452	sel = ub_random_max(env->rnd, num);
453	a = dp->result_list;
454	prev = NULL;
455	while(sel > 0 && a) {
456		prev = a;
457		a = a->next_result;
458		sel--;
459	}
460	if(!a)  /* robustness */
461		return NULL;
462	if(++a->attempts < OUTBOUND_MSG_RETRY)
463		return a;
464	/* remove it from the delegation point result list */
465	if(prev)
466		prev->next_result = a->next_result;
467	else	dp->result_list = a->next_result;
468	return a;
469}
470
471struct dns_msg*
472dns_alloc_msg(sldns_buffer* pkt, struct msg_parse* msg,
473	struct regional* region)
474{
475	struct dns_msg* m = (struct dns_msg*)regional_alloc(region,
476		sizeof(struct dns_msg));
477	if(!m)
478		return NULL;
479	memset(m, 0, sizeof(*m));
480	if(!parse_create_msg(pkt, msg, NULL, &m->qinfo, &m->rep, region)) {
481		log_err("malloc failure: allocating incoming dns_msg");
482		return NULL;
483	}
484	return m;
485}
486
487struct dns_msg*
488dns_copy_msg(struct dns_msg* from, struct regional* region)
489{
490	struct dns_msg* m = (struct dns_msg*)regional_alloc(region,
491		sizeof(struct dns_msg));
492	if(!m)
493		return NULL;
494	m->qinfo = from->qinfo;
495	if(!(m->qinfo.qname = regional_alloc_init(region, from->qinfo.qname,
496		from->qinfo.qname_len)))
497		return NULL;
498	if(!(m->rep = reply_info_copy(from->rep, NULL, region)))
499		return NULL;
500	return m;
501}
502
503void
504iter_dns_store(struct module_env* env, struct query_info* msgqinf,
505	struct reply_info* msgrep, int is_referral, time_t leeway, int pside,
506	struct regional* region, uint16_t flags)
507{
508	if(!dns_cache_store(env, msgqinf, msgrep, is_referral, leeway,
509		pside, region, flags))
510		log_err("out of memory: cannot store data in cache");
511}
512
513int
514iter_ns_probability(struct ub_randstate* rnd, int n, int m)
515{
516	int sel;
517	if(n == m) /* 100% chance */
518		return 1;
519	/* we do not need secure random numbers here, but
520	 * we do need it to be threadsafe, so we use this */
521	sel = ub_random_max(rnd, m);
522	return (sel < n);
523}
524
525/** detect dependency cycle for query and target */
526static int
527causes_cycle(struct module_qstate* qstate, uint8_t* name, size_t namelen,
528	uint16_t t, uint16_t c)
529{
530	struct query_info qinf;
531	qinf.qname = name;
532	qinf.qname_len = namelen;
533	qinf.qtype = t;
534	qinf.qclass = c;
535	fptr_ok(fptr_whitelist_modenv_detect_cycle(
536		qstate->env->detect_cycle));
537	return (*qstate->env->detect_cycle)(qstate, &qinf,
538		(uint16_t)(BIT_RD|BIT_CD), qstate->is_priming,
539		qstate->is_valrec);
540}
541
542void
543iter_mark_cycle_targets(struct module_qstate* qstate, struct delegpt* dp)
544{
545	struct delegpt_ns* ns;
546	for(ns = dp->nslist; ns; ns = ns->next) {
547		if(ns->resolved)
548			continue;
549		/* see if this ns as target causes dependency cycle */
550		if(causes_cycle(qstate, ns->name, ns->namelen,
551			LDNS_RR_TYPE_AAAA, qstate->qinfo.qclass) ||
552		   causes_cycle(qstate, ns->name, ns->namelen,
553			LDNS_RR_TYPE_A, qstate->qinfo.qclass)) {
554			log_nametypeclass(VERB_QUERY, "skipping target due "
555			 	"to dependency cycle (harden-glue: no may "
556				"fix some of the cycles)",
557				ns->name, LDNS_RR_TYPE_A,
558				qstate->qinfo.qclass);
559			ns->resolved = 1;
560		}
561	}
562}
563
564void
565iter_mark_pside_cycle_targets(struct module_qstate* qstate, struct delegpt* dp)
566{
567	struct delegpt_ns* ns;
568	for(ns = dp->nslist; ns; ns = ns->next) {
569		if(ns->done_pside4 && ns->done_pside6)
570			continue;
571		/* see if this ns as target causes dependency cycle */
572		if(causes_cycle(qstate, ns->name, ns->namelen,
573			LDNS_RR_TYPE_A, qstate->qinfo.qclass)) {
574			log_nametypeclass(VERB_QUERY, "skipping target due "
575			 	"to dependency cycle", ns->name,
576				LDNS_RR_TYPE_A, qstate->qinfo.qclass);
577			ns->done_pside4 = 1;
578		}
579		if(causes_cycle(qstate, ns->name, ns->namelen,
580			LDNS_RR_TYPE_AAAA, qstate->qinfo.qclass)) {
581			log_nametypeclass(VERB_QUERY, "skipping target due "
582			 	"to dependency cycle", ns->name,
583				LDNS_RR_TYPE_AAAA, qstate->qinfo.qclass);
584			ns->done_pside6 = 1;
585		}
586	}
587}
588
589int
590iter_dp_is_useless(struct query_info* qinfo, uint16_t qflags,
591	struct delegpt* dp)
592{
593	struct delegpt_ns* ns;
594	/* check:
595	 *      o RD qflag is on.
596	 *      o no addresses are provided.
597	 *      o all NS items are required glue.
598	 * OR
599	 *      o RD qflag is on.
600	 *      o no addresses are provided.
601	 *      o the query is for one of the nameservers in dp,
602	 *        and that nameserver is a glue-name for this dp.
603	 */
604	if(!(qflags&BIT_RD))
605		return 0;
606	/* either available or unused targets */
607	if(dp->usable_list || dp->result_list)
608		return 0;
609
610	/* see if query is for one of the nameservers, which is glue */
611	if( (qinfo->qtype == LDNS_RR_TYPE_A ||
612		qinfo->qtype == LDNS_RR_TYPE_AAAA) &&
613		dname_subdomain_c(qinfo->qname, dp->name) &&
614		delegpt_find_ns(dp, qinfo->qname, qinfo->qname_len))
615		return 1;
616
617	for(ns = dp->nslist; ns; ns = ns->next) {
618		if(ns->resolved) /* skip failed targets */
619			continue;
620		if(!dname_subdomain_c(ns->name, dp->name))
621			return 0; /* one address is not required glue */
622	}
623	return 1;
624}
625
626int
627iter_indicates_dnssec_fwd(struct module_env* env, struct query_info *qinfo)
628{
629	struct trust_anchor* a;
630	if(!env || !env->anchors || !qinfo || !qinfo->qname)
631		return 0;
632	/* a trust anchor exists above the name? */
633	if((a=anchors_lookup(env->anchors, qinfo->qname, qinfo->qname_len,
634		qinfo->qclass))) {
635		if(a->numDS == 0 && a->numDNSKEY == 0) {
636			/* insecure trust point */
637			lock_basic_unlock(&a->lock);
638			return 0;
639		}
640		lock_basic_unlock(&a->lock);
641		return 1;
642	}
643	/* no trust anchor above it. */
644	return 0;
645}
646
647int
648iter_indicates_dnssec(struct module_env* env, struct delegpt* dp,
649        struct dns_msg* msg, uint16_t dclass)
650{
651	struct trust_anchor* a;
652	/* information not available, !env->anchors can be common */
653	if(!env || !env->anchors || !dp || !dp->name)
654		return 0;
655	/* a trust anchor exists with this name, RRSIGs expected */
656	if((a=anchor_find(env->anchors, dp->name, dp->namelabs, dp->namelen,
657		dclass))) {
658		lock_basic_unlock(&a->lock);
659		return 1;
660	}
661	/* see if DS rrset was given, in AUTH section */
662	if(msg && msg->rep &&
663		reply_find_rrset_section_ns(msg->rep, dp->name, dp->namelen,
664		LDNS_RR_TYPE_DS, dclass))
665		return 1;
666	/* look in key cache */
667	if(env->key_cache) {
668		struct key_entry_key* kk = key_cache_obtain(env->key_cache,
669			dp->name, dp->namelen, dclass, env->scratch, *env->now);
670		if(kk) {
671			if(query_dname_compare(kk->name, dp->name) == 0) {
672			  if(key_entry_isgood(kk) || key_entry_isbad(kk)) {
673				regional_free_all(env->scratch);
674				return 1;
675			  } else if(key_entry_isnull(kk)) {
676				regional_free_all(env->scratch);
677				return 0;
678			  }
679			}
680			regional_free_all(env->scratch);
681		}
682	}
683	return 0;
684}
685
686int
687iter_msg_has_dnssec(struct dns_msg* msg)
688{
689	size_t i;
690	if(!msg || !msg->rep)
691		return 0;
692	for(i=0; i<msg->rep->an_numrrsets + msg->rep->ns_numrrsets; i++) {
693		if(((struct packed_rrset_data*)msg->rep->rrsets[i]->
694			entry.data)->rrsig_count > 0)
695			return 1;
696	}
697	/* empty message has no DNSSEC info, with DNSSEC the reply is
698	 * not empty (NSEC) */
699	return 0;
700}
701
702int iter_msg_from_zone(struct dns_msg* msg, struct delegpt* dp,
703        enum response_type type, uint16_t dclass)
704{
705	if(!msg || !dp || !msg->rep || !dp->name)
706		return 0;
707	/* SOA RRset - always from reply zone */
708	if(reply_find_rrset_section_an(msg->rep, dp->name, dp->namelen,
709		LDNS_RR_TYPE_SOA, dclass) ||
710	   reply_find_rrset_section_ns(msg->rep, dp->name, dp->namelen,
711		LDNS_RR_TYPE_SOA, dclass))
712		return 1;
713	if(type == RESPONSE_TYPE_REFERRAL) {
714		size_t i;
715		/* if it adds a single label, i.e. we expect .com,
716		 * and referral to example.com. NS ... , then origin zone
717		 * is .com. For a referral to sub.example.com. NS ... then
718		 * we do not know, since example.com. may be in between. */
719		for(i=0; i<msg->rep->an_numrrsets+msg->rep->ns_numrrsets;
720			i++) {
721			struct ub_packed_rrset_key* s = msg->rep->rrsets[i];
722			if(ntohs(s->rk.type) == LDNS_RR_TYPE_NS &&
723				ntohs(s->rk.rrset_class) == dclass) {
724				int l = dname_count_labels(s->rk.dname);
725				if(l == dp->namelabs + 1 &&
726					dname_strict_subdomain(s->rk.dname,
727					l, dp->name, dp->namelabs))
728					return 1;
729			}
730		}
731		return 0;
732	}
733	log_assert(type==RESPONSE_TYPE_ANSWER || type==RESPONSE_TYPE_CNAME);
734	/* not a referral, and not lame delegation (upwards), so,
735	 * any NS rrset must be from the zone itself */
736	if(reply_find_rrset_section_an(msg->rep, dp->name, dp->namelen,
737		LDNS_RR_TYPE_NS, dclass) ||
738	   reply_find_rrset_section_ns(msg->rep, dp->name, dp->namelen,
739		LDNS_RR_TYPE_NS, dclass))
740		return 1;
741	/* a DNSKEY set is expected at the zone apex as well */
742	/* this is for 'minimal responses' for DNSKEYs */
743	if(reply_find_rrset_section_an(msg->rep, dp->name, dp->namelen,
744		LDNS_RR_TYPE_DNSKEY, dclass))
745		return 1;
746	return 0;
747}
748
749/**
750 * check equality of two rrsets
751 * @param k1: rrset
752 * @param k2: rrset
753 * @return true if equal
754 */
755static int
756rrset_equal(struct ub_packed_rrset_key* k1, struct ub_packed_rrset_key* k2)
757{
758	struct packed_rrset_data* d1 = (struct packed_rrset_data*)
759		k1->entry.data;
760	struct packed_rrset_data* d2 = (struct packed_rrset_data*)
761		k2->entry.data;
762	size_t i, t;
763	if(k1->rk.dname_len != k2->rk.dname_len ||
764		k1->rk.flags != k2->rk.flags ||
765		k1->rk.type != k2->rk.type ||
766		k1->rk.rrset_class != k2->rk.rrset_class ||
767		query_dname_compare(k1->rk.dname, k2->rk.dname) != 0)
768		return 0;
769	if(	/* do not check ttl: d1->ttl != d2->ttl || */
770		d1->count != d2->count ||
771		d1->rrsig_count != d2->rrsig_count ||
772		d1->trust != d2->trust ||
773		d1->security != d2->security)
774		return 0;
775	t = d1->count + d1->rrsig_count;
776	for(i=0; i<t; i++) {
777		if(d1->rr_len[i] != d2->rr_len[i] ||
778			/* no ttl check: d1->rr_ttl[i] != d2->rr_ttl[i] ||*/
779			memcmp(d1->rr_data[i], d2->rr_data[i],
780				d1->rr_len[i]) != 0)
781			return 0;
782	}
783	return 1;
784}
785
786int
787reply_equal(struct reply_info* p, struct reply_info* q, struct regional* region)
788{
789	size_t i;
790	if(p->flags != q->flags ||
791		p->qdcount != q->qdcount ||
792		/* do not check TTL, this may differ */
793		/*
794		p->ttl != q->ttl ||
795		p->prefetch_ttl != q->prefetch_ttl ||
796		*/
797		p->security != q->security ||
798		p->an_numrrsets != q->an_numrrsets ||
799		p->ns_numrrsets != q->ns_numrrsets ||
800		p->ar_numrrsets != q->ar_numrrsets ||
801		p->rrset_count != q->rrset_count)
802		return 0;
803	for(i=0; i<p->rrset_count; i++) {
804		if(!rrset_equal(p->rrsets[i], q->rrsets[i])) {
805			if(!rrset_canonical_equal(region, p->rrsets[i],
806				q->rrsets[i])) {
807				regional_free_all(region);
808				return 0;
809			}
810			regional_free_all(region);
811		}
812	}
813	return 1;
814}
815
816void
817caps_strip_reply(struct reply_info* rep)
818{
819	size_t i;
820	if(!rep) return;
821	/* see if message is a referral, in which case the additional and
822	 * NS record cannot be removed */
823	/* referrals have the AA flag unset (strict check, not elsewhere in
824	 * unbound, but for 0x20 this is very convenient). */
825	if(!(rep->flags&BIT_AA))
826		return;
827	/* remove the additional section from the reply */
828	if(rep->ar_numrrsets != 0) {
829		verbose(VERB_ALGO, "caps fallback: removing additional section");
830		rep->rrset_count -= rep->ar_numrrsets;
831		rep->ar_numrrsets = 0;
832	}
833	/* is there an NS set in the authority section to remove? */
834	/* the failure case (Cisco firewalls) only has one rrset in authsec */
835	for(i=rep->an_numrrsets; i<rep->an_numrrsets+rep->ns_numrrsets; i++) {
836		struct ub_packed_rrset_key* s = rep->rrsets[i];
837		if(ntohs(s->rk.type) == LDNS_RR_TYPE_NS) {
838			/* remove NS rrset and break from loop (loop limits
839			 * have changed) */
840			/* move last rrset into this position (there is no
841			 * additional section any more) */
842			verbose(VERB_ALGO, "caps fallback: removing NS rrset");
843			if(i < rep->rrset_count-1)
844				rep->rrsets[i]=rep->rrsets[rep->rrset_count-1];
845			rep->rrset_count --;
846			rep->ns_numrrsets --;
847			break;
848		}
849	}
850}
851
852int caps_failed_rcode(struct reply_info* rep)
853{
854	return !(FLAGS_GET_RCODE(rep->flags) == LDNS_RCODE_NOERROR ||
855		FLAGS_GET_RCODE(rep->flags) == LDNS_RCODE_NXDOMAIN);
856}
857
858void
859iter_store_parentside_rrset(struct module_env* env,
860	struct ub_packed_rrset_key* rrset)
861{
862	struct rrset_ref ref;
863	rrset = packed_rrset_copy_alloc(rrset, env->alloc, *env->now);
864	if(!rrset) {
865		log_err("malloc failure in store_parentside_rrset");
866		return;
867	}
868	rrset->rk.flags |= PACKED_RRSET_PARENT_SIDE;
869	rrset->entry.hash = rrset_key_hash(&rrset->rk);
870	ref.key = rrset;
871	ref.id = rrset->id;
872	/* ignore ret: if it was in the cache, ref updated */
873	(void)rrset_cache_update(env->rrset_cache, &ref, env->alloc, *env->now);
874}
875
876/** fetch NS record from reply, if any */
877static struct ub_packed_rrset_key*
878reply_get_NS_rrset(struct reply_info* rep)
879{
880	size_t i;
881	for(i=0; i<rep->rrset_count; i++) {
882		if(rep->rrsets[i]->rk.type == htons(LDNS_RR_TYPE_NS)) {
883			return rep->rrsets[i];
884		}
885	}
886	return NULL;
887}
888
889void
890iter_store_parentside_NS(struct module_env* env, struct reply_info* rep)
891{
892	struct ub_packed_rrset_key* rrset = reply_get_NS_rrset(rep);
893	if(rrset) {
894		log_rrset_key(VERB_ALGO, "store parent-side NS", rrset);
895		iter_store_parentside_rrset(env, rrset);
896	}
897}
898
899void iter_store_parentside_neg(struct module_env* env,
900        struct query_info* qinfo, struct reply_info* rep)
901{
902	/* TTL: NS from referral in iq->deleg_msg,
903	 *      or first RR from iq->response,
904	 *      or servfail5secs if !iq->response */
905	time_t ttl = NORR_TTL;
906	struct ub_packed_rrset_key* neg;
907	struct packed_rrset_data* newd;
908	if(rep) {
909		struct ub_packed_rrset_key* rrset = reply_get_NS_rrset(rep);
910		if(!rrset && rep->rrset_count != 0) rrset = rep->rrsets[0];
911		if(rrset) ttl = ub_packed_rrset_ttl(rrset);
912	}
913	/* create empty rrset to store */
914	neg = (struct ub_packed_rrset_key*)regional_alloc(env->scratch,
915	                sizeof(struct ub_packed_rrset_key));
916	if(!neg) {
917		log_err("out of memory in store_parentside_neg");
918		return;
919	}
920	memset(&neg->entry, 0, sizeof(neg->entry));
921	neg->entry.key = neg;
922	neg->rk.type = htons(qinfo->qtype);
923	neg->rk.rrset_class = htons(qinfo->qclass);
924	neg->rk.flags = 0;
925	neg->rk.dname = regional_alloc_init(env->scratch, qinfo->qname,
926		qinfo->qname_len);
927	if(!neg->rk.dname) {
928		log_err("out of memory in store_parentside_neg");
929		return;
930	}
931	neg->rk.dname_len = qinfo->qname_len;
932	neg->entry.hash = rrset_key_hash(&neg->rk);
933	newd = (struct packed_rrset_data*)regional_alloc_zero(env->scratch,
934		sizeof(struct packed_rrset_data) + sizeof(size_t) +
935		sizeof(uint8_t*) + sizeof(time_t) + sizeof(uint16_t));
936	if(!newd) {
937		log_err("out of memory in store_parentside_neg");
938		return;
939	}
940	neg->entry.data = newd;
941	newd->ttl = ttl;
942	/* entry must have one RR, otherwise not valid in cache.
943	 * put in one RR with empty rdata: those are ignored as nameserver */
944	newd->count = 1;
945	newd->rrsig_count = 0;
946	newd->trust = rrset_trust_ans_noAA;
947	newd->rr_len = (size_t*)((uint8_t*)newd +
948		sizeof(struct packed_rrset_data));
949	newd->rr_len[0] = 0 /* zero len rdata */ + sizeof(uint16_t);
950	packed_rrset_ptr_fixup(newd);
951	newd->rr_ttl[0] = newd->ttl;
952	sldns_write_uint16(newd->rr_data[0], 0 /* zero len rdata */);
953	/* store it */
954	log_rrset_key(VERB_ALGO, "store parent-side negative", neg);
955	iter_store_parentside_rrset(env, neg);
956}
957
958int
959iter_lookup_parent_NS_from_cache(struct module_env* env, struct delegpt* dp,
960	struct regional* region, struct query_info* qinfo)
961{
962	struct ub_packed_rrset_key* akey;
963	akey = rrset_cache_lookup(env->rrset_cache, dp->name,
964		dp->namelen, LDNS_RR_TYPE_NS, qinfo->qclass,
965		PACKED_RRSET_PARENT_SIDE, *env->now, 0);
966	if(akey) {
967		log_rrset_key(VERB_ALGO, "found parent-side NS in cache", akey);
968		dp->has_parent_side_NS = 1;
969		/* and mark the new names as lame */
970		if(!delegpt_rrset_add_ns(dp, region, akey, 1)) {
971			lock_rw_unlock(&akey->entry.lock);
972			return 0;
973		}
974		lock_rw_unlock(&akey->entry.lock);
975	}
976	return 1;
977}
978
979int iter_lookup_parent_glue_from_cache(struct module_env* env,
980        struct delegpt* dp, struct regional* region, struct query_info* qinfo)
981{
982	struct ub_packed_rrset_key* akey;
983	struct delegpt_ns* ns;
984	size_t num = delegpt_count_targets(dp);
985	for(ns = dp->nslist; ns; ns = ns->next) {
986		/* get cached parentside A */
987		akey = rrset_cache_lookup(env->rrset_cache, ns->name,
988			ns->namelen, LDNS_RR_TYPE_A, qinfo->qclass,
989			PACKED_RRSET_PARENT_SIDE, *env->now, 0);
990		if(akey) {
991			log_rrset_key(VERB_ALGO, "found parent-side", akey);
992			ns->done_pside4 = 1;
993			/* a negative-cache-element has no addresses it adds */
994			if(!delegpt_add_rrset_A(dp, region, akey, 1))
995				log_err("malloc failure in lookup_parent_glue");
996			lock_rw_unlock(&akey->entry.lock);
997		}
998		/* get cached parentside AAAA */
999		akey = rrset_cache_lookup(env->rrset_cache, ns->name,
1000			ns->namelen, LDNS_RR_TYPE_AAAA, qinfo->qclass,
1001			PACKED_RRSET_PARENT_SIDE, *env->now, 0);
1002		if(akey) {
1003			log_rrset_key(VERB_ALGO, "found parent-side", akey);
1004			ns->done_pside6 = 1;
1005			/* a negative-cache-element has no addresses it adds */
1006			if(!delegpt_add_rrset_AAAA(dp, region, akey, 1))
1007				log_err("malloc failure in lookup_parent_glue");
1008			lock_rw_unlock(&akey->entry.lock);
1009		}
1010	}
1011	/* see if new (but lame) addresses have become available */
1012	return delegpt_count_targets(dp) != num;
1013}
1014
1015int
1016iter_get_next_root(struct iter_hints* hints, struct iter_forwards* fwd,
1017	uint16_t* c)
1018{
1019	uint16_t c1 = *c, c2 = *c;
1020	int r1 = hints_next_root(hints, &c1);
1021	int r2 = forwards_next_root(fwd, &c2);
1022	if(!r1 && !r2) /* got none, end of list */
1023		return 0;
1024	else if(!r1) /* got one, return that */
1025		*c = c2;
1026	else if(!r2)
1027		*c = c1;
1028	else if(c1 < c2) /* got both take smallest */
1029		*c = c1;
1030	else	*c = c2;
1031	return 1;
1032}
1033
1034void
1035iter_scrub_ds(struct dns_msg* msg, struct ub_packed_rrset_key* ns, uint8_t* z)
1036{
1037	/* Only the DS record for the delegation itself is expected.
1038	 * We allow DS for everything between the bailiwick and the
1039	 * zonecut, thus DS records must be at or above the zonecut.
1040	 * And the DS records must be below the server authority zone.
1041	 * The answer section is already scrubbed. */
1042	size_t i = msg->rep->an_numrrsets;
1043	while(i < (msg->rep->an_numrrsets + msg->rep->ns_numrrsets)) {
1044		struct ub_packed_rrset_key* s = msg->rep->rrsets[i];
1045		if(ntohs(s->rk.type) == LDNS_RR_TYPE_DS &&
1046			(!ns || !dname_subdomain_c(ns->rk.dname, s->rk.dname)
1047			|| query_dname_compare(z, s->rk.dname) == 0)) {
1048			log_nametypeclass(VERB_ALGO, "removing irrelevant DS",
1049				s->rk.dname, ntohs(s->rk.type),
1050				ntohs(s->rk.rrset_class));
1051			memmove(msg->rep->rrsets+i, msg->rep->rrsets+i+1,
1052				sizeof(struct ub_packed_rrset_key*) *
1053				(msg->rep->rrset_count-i-1));
1054			msg->rep->ns_numrrsets--;
1055			msg->rep->rrset_count--;
1056			/* stay at same i, but new record */
1057			continue;
1058		}
1059		i++;
1060	}
1061}
1062
1063void iter_dec_attempts(struct delegpt* dp, int d)
1064{
1065	struct delegpt_addr* a;
1066	for(a=dp->target_list; a; a = a->next_target) {
1067		if(a->attempts >= OUTBOUND_MSG_RETRY) {
1068			/* add back to result list */
1069			a->next_result = dp->result_list;
1070			dp->result_list = a;
1071		}
1072		if(a->attempts > d)
1073			a->attempts -= d;
1074		else a->attempts = 0;
1075	}
1076}
1077
1078void iter_merge_retry_counts(struct delegpt* dp, struct delegpt* old)
1079{
1080	struct delegpt_addr* a, *o, *prev;
1081	for(a=dp->target_list; a; a = a->next_target) {
1082		o = delegpt_find_addr(old, &a->addr, a->addrlen);
1083		if(o) {
1084			log_addr(VERB_ALGO, "copy attempt count previous dp",
1085				&a->addr, a->addrlen);
1086			a->attempts = o->attempts;
1087		}
1088	}
1089	prev = NULL;
1090	a = dp->usable_list;
1091	while(a) {
1092		if(a->attempts >= OUTBOUND_MSG_RETRY) {
1093			log_addr(VERB_ALGO, "remove from usable list dp",
1094				&a->addr, a->addrlen);
1095			/* remove from result list */
1096			if(prev)
1097				prev->next_usable = a->next_usable;
1098			else	dp->usable_list = a->next_usable;
1099			/* prev stays the same */
1100			a = a->next_usable;
1101			continue;
1102		}
1103		prev = a;
1104		a = a->next_usable;
1105	}
1106}
1107
1108int
1109iter_ds_toolow(struct dns_msg* msg, struct delegpt* dp)
1110{
1111	/* if for query example.com, there is example.com SOA or a subdomain
1112	 * of example.com, then we are too low and need to fetch NS. */
1113	size_t i;
1114	/* if we have a DNAME or CNAME we are probably wrong */
1115	/* if we have a qtype DS in the answer section, its fine */
1116	for(i=0; i < msg->rep->an_numrrsets; i++) {
1117		struct ub_packed_rrset_key* s = msg->rep->rrsets[i];
1118		if(ntohs(s->rk.type) == LDNS_RR_TYPE_DNAME ||
1119			ntohs(s->rk.type) == LDNS_RR_TYPE_CNAME) {
1120			/* not the right answer, maybe too low, check the
1121			 * RRSIG signer name (if there is any) for a hint
1122			 * that it is from the dp zone anyway */
1123			uint8_t* sname;
1124			size_t slen;
1125			val_find_rrset_signer(s, &sname, &slen);
1126			if(sname && query_dname_compare(dp->name, sname)==0)
1127				return 0; /* it is fine, from the right dp */
1128			return 1;
1129		}
1130		if(ntohs(s->rk.type) == LDNS_RR_TYPE_DS)
1131			return 0; /* fine, we have a DS record */
1132	}
1133	for(i=msg->rep->an_numrrsets;
1134		i < msg->rep->an_numrrsets + msg->rep->ns_numrrsets; i++) {
1135		struct ub_packed_rrset_key* s = msg->rep->rrsets[i];
1136		if(ntohs(s->rk.type) == LDNS_RR_TYPE_SOA) {
1137			if(dname_subdomain_c(s->rk.dname, msg->qinfo.qname))
1138				return 1; /* point is too low */
1139			if(query_dname_compare(s->rk.dname, dp->name)==0)
1140				return 0; /* right dp */
1141		}
1142		if(ntohs(s->rk.type) == LDNS_RR_TYPE_NSEC ||
1143			ntohs(s->rk.type) == LDNS_RR_TYPE_NSEC3) {
1144			uint8_t* sname;
1145			size_t slen;
1146			val_find_rrset_signer(s, &sname, &slen);
1147			if(sname && query_dname_compare(dp->name, sname)==0)
1148				return 0; /* it is fine, from the right dp */
1149			return 1;
1150		}
1151	}
1152	/* we do not know */
1153	return 1;
1154}
1155
1156int iter_dp_cangodown(struct query_info* qinfo, struct delegpt* dp)
1157{
1158	/* no delegation point, do not see how we can go down,
1159	 * robust check, it should really exist */
1160	if(!dp) return 0;
1161
1162	/* see if dp equals the qname, then we cannot go down further */
1163	if(query_dname_compare(qinfo->qname, dp->name) == 0)
1164		return 0;
1165	/* if dp is one label above the name we also cannot go down further */
1166	if(dname_count_labels(qinfo->qname) == dp->namelabs+1)
1167		return 0;
1168	return 1;
1169}
1170