xref: /freebsd/contrib/unbound/iterator/iter_utils.c (revision 3078531de10dcae44b253a35125c949ff4235284)
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 "services/outside_network.h"
54 #include "util/net_help.h"
55 #include "util/module.h"
56 #include "util/log.h"
57 #include "util/config_file.h"
58 #include "util/regional.h"
59 #include "util/data/msgparse.h"
60 #include "util/data/dname.h"
61 #include "util/random.h"
62 #include "util/fptr_wlist.h"
63 #include "validator/val_anchor.h"
64 #include "validator/val_kcache.h"
65 #include "validator/val_kentry.h"
66 #include "validator/val_utils.h"
67 #include "validator/val_sigcrypt.h"
68 #include "sldns/sbuffer.h"
69 #include "sldns/str2wire.h"
70 
71 /** time when nameserver glue is said to be 'recent' */
72 #define SUSPICION_RECENT_EXPIRY 86400
73 
74 /** fillup fetch policy array */
75 static void
76 fetch_fill(struct iter_env* ie, const char* str)
77 {
78 	char* s = (char*)str, *e;
79 	int i;
80 	for(i=0; i<ie->max_dependency_depth+1; i++) {
81 		ie->target_fetch_policy[i] = strtol(s, &e, 10);
82 		if(s == e)
83 			fatal_exit("cannot parse fetch policy number %s", s);
84 		s = e;
85 	}
86 }
87 
88 /** Read config string that represents the target fetch policy */
89 static int
90 read_fetch_policy(struct iter_env* ie, const char* str)
91 {
92 	int count = cfg_count_numbers(str);
93 	if(count < 1) {
94 		log_err("Cannot parse target fetch policy: \"%s\"", str);
95 		return 0;
96 	}
97 	ie->max_dependency_depth = count - 1;
98 	ie->target_fetch_policy = (int*)calloc(
99 		(size_t)ie->max_dependency_depth+1, sizeof(int));
100 	if(!ie->target_fetch_policy) {
101 		log_err("alloc fetch policy: out of memory");
102 		return 0;
103 	}
104 	fetch_fill(ie, str);
105 	return 1;
106 }
107 
108 /** apply config caps whitelist items to name tree */
109 static int
110 caps_white_apply_cfg(rbtree_type* ntree, struct config_file* cfg)
111 {
112 	struct config_strlist* p;
113 	for(p=cfg->caps_whitelist; p; p=p->next) {
114 		struct name_tree_node* n;
115 		size_t len;
116 		uint8_t* nm = sldns_str2wire_dname(p->str, &len);
117 		if(!nm) {
118 			log_err("could not parse %s", p->str);
119 			return 0;
120 		}
121 		n = (struct name_tree_node*)calloc(1, sizeof(*n));
122 		if(!n) {
123 			log_err("out of memory");
124 			free(nm);
125 			return 0;
126 		}
127 		n->node.key = n;
128 		n->name = nm;
129 		n->len = len;
130 		n->labs = dname_count_labels(nm);
131 		n->dclass = LDNS_RR_CLASS_IN;
132 		if(!name_tree_insert(ntree, n, nm, len, n->labs, n->dclass)) {
133 			/* duplicate element ignored, idempotent */
134 			free(n->name);
135 			free(n);
136 		}
137 	}
138 	name_tree_init_parents(ntree);
139 	return 1;
140 }
141 
142 int
143 iter_apply_cfg(struct iter_env* iter_env, struct config_file* cfg)
144 {
145 	int i;
146 	/* target fetch policy */
147 	if(!read_fetch_policy(iter_env, cfg->target_fetch_policy))
148 		return 0;
149 	for(i=0; i<iter_env->max_dependency_depth+1; i++)
150 		verbose(VERB_QUERY, "target fetch policy for level %d is %d",
151 			i, iter_env->target_fetch_policy[i]);
152 
153 	if(!iter_env->donotq)
154 		iter_env->donotq = donotq_create();
155 	if(!iter_env->donotq || !donotq_apply_cfg(iter_env->donotq, cfg)) {
156 		log_err("Could not set donotqueryaddresses");
157 		return 0;
158 	}
159 	if(!iter_env->priv)
160 		iter_env->priv = priv_create();
161 	if(!iter_env->priv || !priv_apply_cfg(iter_env->priv, cfg)) {
162 		log_err("Could not set private addresses");
163 		return 0;
164 	}
165 	if(cfg->caps_whitelist) {
166 		if(!iter_env->caps_white)
167 			iter_env->caps_white = rbtree_create(name_tree_compare);
168 		if(!iter_env->caps_white || !caps_white_apply_cfg(
169 			iter_env->caps_white, cfg)) {
170 			log_err("Could not set capsforid whitelist");
171 			return 0;
172 		}
173 
174 	}
175 	iter_env->supports_ipv6 = cfg->do_ip6;
176 	iter_env->supports_ipv4 = cfg->do_ip4;
177 	iter_env->outbound_msg_retry = cfg->outbound_msg_retry;
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  */
225 static int
226 iter_filter_unsuitable(struct iter_env* iter_env, struct module_env* env,
227 	uint8_t* name, size_t namelen, uint16_t qtype, time_t now,
228 	struct delegpt_addr* a)
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) */
282 static int
283 iter_fill_rtt(struct iter_env* iter_env, struct module_env* env,
284 	uint8_t* name, size_t namelen, uint16_t qtype, time_t now,
285 	struct delegpt* dp, int* best_rtt, struct sock_list* blacklist,
286 	size_t* num_suitable_results)
287 {
288 	int got_it = 0;
289 	struct delegpt_addr* a;
290 	*num_suitable_results = 0;
291 
292 	if(dp->bogus)
293 		return 0; /* NS bogus, all bogus, nothing found */
294 	for(a=dp->result_list; a; a = a->next_result) {
295 		a->sel_rtt = iter_filter_unsuitable(iter_env, env,
296 			name, namelen, qtype, now, a);
297 		if(a->sel_rtt != -1) {
298 			if(sock_list_find(blacklist, &a->addr, a->addrlen))
299 				a->sel_rtt += BLACKLIST_PENALTY;
300 
301 			if(!got_it) {
302 				*best_rtt = a->sel_rtt;
303 				got_it = 1;
304 			} else if(a->sel_rtt < *best_rtt) {
305 				*best_rtt = a->sel_rtt;
306 			}
307 			(*num_suitable_results)++;
308 		}
309 	}
310 	return got_it;
311 }
312 
313 /** compare two rtts, return -1, 0 or 1 */
314 static int
315 rtt_compare(const void* x, const void* y)
316 {
317 	if(*(int*)x == *(int*)y)
318 		return 0;
319 	if(*(int*)x > *(int*)y)
320 		return 1;
321 	return -1;
322 }
323 
324 /** get RTT for the Nth fastest server */
325 static int
326 nth_rtt(struct delegpt_addr* result_list, size_t num_results, size_t n)
327 {
328 	int rtt_band;
329 	size_t i;
330 	int* rtt_list, *rtt_index;
331 
332 	if(num_results < 1 || n >= num_results) {
333 		return -1;
334 	}
335 
336 	rtt_list = calloc(num_results, sizeof(int));
337 	if(!rtt_list) {
338 		log_err("malloc failure: allocating rtt_list");
339 		return -1;
340 	}
341 	rtt_index = rtt_list;
342 
343 	for(i=0; i<num_results && result_list; i++) {
344 		if(result_list->sel_rtt != -1) {
345 			*rtt_index = result_list->sel_rtt;
346 			rtt_index++;
347 		}
348 		result_list=result_list->next_result;
349 	}
350 	qsort(rtt_list, num_results, sizeof(*rtt_list), rtt_compare);
351 
352 	log_assert(n > 0);
353 	rtt_band = rtt_list[n-1];
354 	free(rtt_list);
355 
356 	return rtt_band;
357 }
358 
359 /** filter the address list, putting best targets at front,
360  * returns number of best targets (or 0, no suitable targets) */
361 static int
362 iter_filter_order(struct iter_env* iter_env, struct module_env* env,
363 	uint8_t* name, size_t namelen, uint16_t qtype, time_t now,
364 	struct delegpt* dp, int* selected_rtt, int open_target,
365 	struct sock_list* blacklist, time_t prefetch)
366 {
367 	int got_num = 0, low_rtt = 0, swap_to_front, rtt_band = RTT_BAND, nth;
368 	int alllame = 0;
369 	size_t num_results;
370 	struct delegpt_addr* a, *n, *prev=NULL;
371 
372 	/* fillup sel_rtt and find best rtt in the bunch */
373 	got_num = iter_fill_rtt(iter_env, env, name, namelen, qtype, now, dp,
374 		&low_rtt, blacklist, &num_results);
375 	if(got_num == 0)
376 		return 0;
377 	if(low_rtt >= USEFUL_SERVER_TOP_TIMEOUT &&
378 		/* If all missing (or not fully resolved) targets are lame,
379 		 * then use the remaining lame address. */
380 		((delegpt_count_missing_targets(dp, &alllame) > 0 && !alllame) ||
381 		open_target > 0)) {
382 		verbose(VERB_ALGO, "Bad choices, trying to get more choice");
383 		return 0; /* we want more choice. The best choice is a bad one.
384 			     return 0 to force the caller to fetch more */
385 	}
386 
387 	if(env->cfg->fast_server_permil != 0 && prefetch == 0 &&
388 		num_results > env->cfg->fast_server_num &&
389 		ub_random_max(env->rnd, 1000) < env->cfg->fast_server_permil) {
390 		/* the query is not prefetch, but for a downstream client,
391 		 * there are more servers available then the fastest N we want
392 		 * to choose from. Limit our choice to the fastest servers. */
393 		nth = nth_rtt(dp->result_list, num_results,
394 			env->cfg->fast_server_num);
395 		if(nth > 0) {
396 			rtt_band = nth - low_rtt;
397 			if(rtt_band > RTT_BAND)
398 				rtt_band = RTT_BAND;
399 		}
400 	}
401 
402 	got_num = 0;
403 	a = dp->result_list;
404 	while(a) {
405 		/* skip unsuitable targets */
406 		if(a->sel_rtt == -1) {
407 			prev = a;
408 			a = a->next_result;
409 			continue;
410 		}
411 		/* classify the server address and determine what to do */
412 		swap_to_front = 0;
413 		if(a->sel_rtt >= low_rtt && a->sel_rtt - low_rtt <= rtt_band) {
414 			got_num++;
415 			swap_to_front = 1;
416 		} else if(a->sel_rtt<low_rtt && low_rtt-a->sel_rtt<=rtt_band) {
417 			got_num++;
418 			swap_to_front = 1;
419 		}
420 		/* swap to front if necessary, or move to next result */
421 		if(swap_to_front && prev) {
422 			n = a->next_result;
423 			prev->next_result = n;
424 			a->next_result = dp->result_list;
425 			dp->result_list = a;
426 			a = n;
427 		} else {
428 			prev = a;
429 			a = a->next_result;
430 		}
431 	}
432 	*selected_rtt = low_rtt;
433 
434 	if (env->cfg->prefer_ip6) {
435 		int got_num6 = 0;
436 		int low_rtt6 = 0;
437 		int i;
438 		int attempt = -1; /* filter to make sure addresses have
439 		  less attempts on them than the first, to force round
440 		  robin when all the IPv6 addresses fail */
441 		int num4ok = 0; /* number ip4 at low attempt count */
442 		int num4_lowrtt = 0;
443 		prev = NULL;
444 		a = dp->result_list;
445 		for(i = 0; i < got_num; i++) {
446 			if(!a) break; /* robustness */
447 			swap_to_front = 0;
448 			if(a->addr.ss_family != AF_INET6 && attempt == -1) {
449 				/* if we only have ip4 at low attempt count,
450 				 * then ip6 is failing, and we need to
451 				 * select one of the remaining IPv4 addrs */
452 				attempt = a->attempts;
453 				num4ok++;
454 				num4_lowrtt = a->sel_rtt;
455 			} else if(a->addr.ss_family != AF_INET6 && attempt == a->attempts) {
456 				num4ok++;
457 				if(num4_lowrtt == 0 || a->sel_rtt < num4_lowrtt) {
458 					num4_lowrtt = a->sel_rtt;
459 				}
460 			}
461 			if(a->addr.ss_family == AF_INET6) {
462 				if(attempt == -1) {
463 					attempt = a->attempts;
464 				} else if(a->attempts > attempt) {
465 					break;
466 				}
467 				got_num6++;
468 				swap_to_front = 1;
469 				if(low_rtt6 == 0 || a->sel_rtt < low_rtt6) {
470 					low_rtt6 = a->sel_rtt;
471 				}
472 			}
473 			/* swap to front if IPv6, or move to next result */
474 			if(swap_to_front && prev) {
475 				n = a->next_result;
476 				prev->next_result = n;
477 				a->next_result = dp->result_list;
478 				dp->result_list = a;
479 				a = n;
480 			} else {
481 				prev = a;
482 				a = a->next_result;
483 			}
484 		}
485 		if(got_num6 > 0) {
486 			got_num = got_num6;
487 			*selected_rtt = low_rtt6;
488 		} else if(num4ok > 0) {
489 			got_num = num4ok;
490 			*selected_rtt = num4_lowrtt;
491 		}
492 	} else if (env->cfg->prefer_ip4) {
493 		int got_num4 = 0;
494 		int low_rtt4 = 0;
495 		int i;
496 		int attempt = -1; /* filter to make sure addresses have
497 		  less attempts on them than the first, to force round
498 		  robin when all the IPv4 addresses fail */
499 		int num6ok = 0; /* number ip6 at low attempt count */
500 		int num6_lowrtt = 0;
501 		prev = NULL;
502 		a = dp->result_list;
503 		for(i = 0; i < got_num; i++) {
504 			if(!a) break; /* robustness */
505 			swap_to_front = 0;
506 			if(a->addr.ss_family != AF_INET && attempt == -1) {
507 				/* if we only have ip6 at low attempt count,
508 				 * then ip4 is failing, and we need to
509 				 * select one of the remaining IPv6 addrs */
510 				attempt = a->attempts;
511 				num6ok++;
512 				num6_lowrtt = a->sel_rtt;
513 			} else if(a->addr.ss_family != AF_INET && attempt == a->attempts) {
514 				num6ok++;
515 				if(num6_lowrtt == 0 || a->sel_rtt < num6_lowrtt) {
516 					num6_lowrtt = a->sel_rtt;
517 				}
518 			}
519 			if(a->addr.ss_family == AF_INET) {
520 				if(attempt == -1) {
521 					attempt = a->attempts;
522 				} else if(a->attempts > attempt) {
523 					break;
524 				}
525 				got_num4++;
526 				swap_to_front = 1;
527 				if(low_rtt4 == 0 || a->sel_rtt < low_rtt4) {
528 					low_rtt4 = a->sel_rtt;
529 				}
530 			}
531 			/* swap to front if IPv4, or move to next result */
532 			if(swap_to_front && prev) {
533 				n = a->next_result;
534 				prev->next_result = n;
535 				a->next_result = dp->result_list;
536 				dp->result_list = a;
537 				a = n;
538 			} else {
539 				prev = a;
540 				a = a->next_result;
541 			}
542 		}
543 		if(got_num4 > 0) {
544 			got_num = got_num4;
545 			*selected_rtt = low_rtt4;
546 		} else if(num6ok > 0) {
547 			got_num = num6ok;
548 			*selected_rtt = num6_lowrtt;
549 		}
550 	}
551 	return got_num;
552 }
553 
554 struct delegpt_addr*
555 iter_server_selection(struct iter_env* iter_env,
556 	struct module_env* env, struct delegpt* dp,
557 	uint8_t* name, size_t namelen, uint16_t qtype, int* dnssec_lame,
558 	int* chase_to_rd, int open_target, struct sock_list* blacklist,
559 	time_t prefetch)
560 {
561 	int sel;
562 	int selrtt;
563 	struct delegpt_addr* a, *prev;
564 	int num = iter_filter_order(iter_env, env, name, namelen, qtype,
565 		*env->now, dp, &selrtt, open_target, blacklist, prefetch);
566 
567 	if(num == 0)
568 		return NULL;
569 	verbose(VERB_ALGO, "selrtt %d", selrtt);
570 	if(selrtt > BLACKLIST_PENALTY) {
571 		if(selrtt-BLACKLIST_PENALTY > USEFUL_SERVER_TOP_TIMEOUT*3) {
572 			verbose(VERB_ALGO, "chase to "
573 				"blacklisted recursion lame server");
574 			*chase_to_rd = 1;
575 		}
576 		if(selrtt-BLACKLIST_PENALTY > USEFUL_SERVER_TOP_TIMEOUT*2) {
577 			verbose(VERB_ALGO, "chase to "
578 				"blacklisted dnssec lame server");
579 			*dnssec_lame = 1;
580 		}
581 	} else {
582 		if(selrtt > USEFUL_SERVER_TOP_TIMEOUT*3) {
583 			verbose(VERB_ALGO, "chase to recursion lame server");
584 			*chase_to_rd = 1;
585 		}
586 		if(selrtt > USEFUL_SERVER_TOP_TIMEOUT*2) {
587 			verbose(VERB_ALGO, "chase to dnssec lame server");
588 			*dnssec_lame = 1;
589 		}
590 		if(selrtt == USEFUL_SERVER_TOP_TIMEOUT) {
591 			verbose(VERB_ALGO, "chase to blacklisted lame server");
592 			return NULL;
593 		}
594 	}
595 
596 	if(num == 1) {
597 		a = dp->result_list;
598 		if(++a->attempts < iter_env->outbound_msg_retry)
599 			return a;
600 		dp->result_list = a->next_result;
601 		return a;
602 	}
603 
604 	/* randomly select a target from the list */
605 	log_assert(num > 1);
606 	/* grab secure random number, to pick unexpected server.
607 	 * also we need it to be threadsafe. */
608 	sel = ub_random_max(env->rnd, num);
609 	a = dp->result_list;
610 	prev = NULL;
611 	while(sel > 0 && a) {
612 		prev = a;
613 		a = a->next_result;
614 		sel--;
615 	}
616 	if(!a)  /* robustness */
617 		return NULL;
618 	if(++a->attempts < iter_env->outbound_msg_retry)
619 		return a;
620 	/* remove it from the delegation point result list */
621 	if(prev)
622 		prev->next_result = a->next_result;
623 	else	dp->result_list = a->next_result;
624 	return a;
625 }
626 
627 struct dns_msg*
628 dns_alloc_msg(sldns_buffer* pkt, struct msg_parse* msg,
629 	struct regional* region)
630 {
631 	struct dns_msg* m = (struct dns_msg*)regional_alloc(region,
632 		sizeof(struct dns_msg));
633 	if(!m)
634 		return NULL;
635 	memset(m, 0, sizeof(*m));
636 	if(!parse_create_msg(pkt, msg, NULL, &m->qinfo, &m->rep, region)) {
637 		log_err("malloc failure: allocating incoming dns_msg");
638 		return NULL;
639 	}
640 	return m;
641 }
642 
643 struct dns_msg*
644 dns_copy_msg(struct dns_msg* from, struct regional* region)
645 {
646 	struct dns_msg* m = (struct dns_msg*)regional_alloc(region,
647 		sizeof(struct dns_msg));
648 	if(!m)
649 		return NULL;
650 	m->qinfo = from->qinfo;
651 	if(!(m->qinfo.qname = regional_alloc_init(region, from->qinfo.qname,
652 		from->qinfo.qname_len)))
653 		return NULL;
654 	if(!(m->rep = reply_info_copy(from->rep, NULL, region)))
655 		return NULL;
656 	return m;
657 }
658 
659 void
660 iter_dns_store(struct module_env* env, struct query_info* msgqinf,
661 	struct reply_info* msgrep, int is_referral, time_t leeway, int pside,
662 	struct regional* region, uint16_t flags, time_t qstarttime)
663 {
664 	if(!dns_cache_store(env, msgqinf, msgrep, is_referral, leeway,
665 		pside, region, flags, qstarttime))
666 		log_err("out of memory: cannot store data in cache");
667 }
668 
669 int
670 iter_ns_probability(struct ub_randstate* rnd, int n, int m)
671 {
672 	int sel;
673 	if(n == m) /* 100% chance */
674 		return 1;
675 	/* we do not need secure random numbers here, but
676 	 * we do need it to be threadsafe, so we use this */
677 	sel = ub_random_max(rnd, m);
678 	return (sel < n);
679 }
680 
681 /** detect dependency cycle for query and target */
682 static int
683 causes_cycle(struct module_qstate* qstate, uint8_t* name, size_t namelen,
684 	uint16_t t, uint16_t c)
685 {
686 	struct query_info qinf;
687 	qinf.qname = name;
688 	qinf.qname_len = namelen;
689 	qinf.qtype = t;
690 	qinf.qclass = c;
691 	qinf.local_alias = NULL;
692 	fptr_ok(fptr_whitelist_modenv_detect_cycle(
693 		qstate->env->detect_cycle));
694 	return (*qstate->env->detect_cycle)(qstate, &qinf,
695 		(uint16_t)(BIT_RD|BIT_CD), qstate->is_priming,
696 		qstate->is_valrec);
697 }
698 
699 void
700 iter_mark_cycle_targets(struct module_qstate* qstate, struct delegpt* dp)
701 {
702 	struct delegpt_ns* ns;
703 	for(ns = dp->nslist; ns; ns = ns->next) {
704 		if(ns->resolved)
705 			continue;
706 		/* see if this ns as target causes dependency cycle */
707 		if(causes_cycle(qstate, ns->name, ns->namelen,
708 			LDNS_RR_TYPE_AAAA, qstate->qinfo.qclass) ||
709 		   causes_cycle(qstate, ns->name, ns->namelen,
710 			LDNS_RR_TYPE_A, qstate->qinfo.qclass)) {
711 			log_nametypeclass(VERB_QUERY, "skipping target due "
712 			 	"to dependency cycle (harden-glue: no may "
713 				"fix some of the cycles)",
714 				ns->name, LDNS_RR_TYPE_A,
715 				qstate->qinfo.qclass);
716 			ns->resolved = 1;
717 		}
718 	}
719 }
720 
721 void
722 iter_mark_pside_cycle_targets(struct module_qstate* qstate, struct delegpt* dp)
723 {
724 	struct delegpt_ns* ns;
725 	for(ns = dp->nslist; ns; ns = ns->next) {
726 		if(ns->done_pside4 && ns->done_pside6)
727 			continue;
728 		/* see if this ns as target causes dependency cycle */
729 		if(causes_cycle(qstate, ns->name, ns->namelen,
730 			LDNS_RR_TYPE_A, qstate->qinfo.qclass)) {
731 			log_nametypeclass(VERB_QUERY, "skipping target due "
732 			 	"to dependency cycle", ns->name,
733 				LDNS_RR_TYPE_A, qstate->qinfo.qclass);
734 			ns->done_pside4 = 1;
735 		}
736 		if(causes_cycle(qstate, ns->name, ns->namelen,
737 			LDNS_RR_TYPE_AAAA, qstate->qinfo.qclass)) {
738 			log_nametypeclass(VERB_QUERY, "skipping target due "
739 			 	"to dependency cycle", ns->name,
740 				LDNS_RR_TYPE_AAAA, qstate->qinfo.qclass);
741 			ns->done_pside6 = 1;
742 		}
743 	}
744 }
745 
746 int
747 iter_dp_is_useless(struct query_info* qinfo, uint16_t qflags,
748 	struct delegpt* dp, int supports_ipv4, int supports_ipv6)
749 {
750 	struct delegpt_ns* ns;
751 	struct delegpt_addr* a;
752 	/* check:
753 	 *      o RD qflag is on.
754 	 *      o no addresses are provided.
755 	 *      o all NS items are required glue.
756 	 * OR
757 	 *      o RD qflag is on.
758 	 *      o no addresses are provided.
759 	 *      o the query is for one of the nameservers in dp,
760 	 *        and that nameserver is a glue-name for this dp.
761 	 */
762 	if(!(qflags&BIT_RD))
763 		return 0;
764 	/* either available or unused targets,
765 	 * if they exist, the dp is not useless. */
766 	for(a = dp->usable_list; a; a = a->next_usable) {
767 		if(!addr_is_ip6(&a->addr, a->addrlen) && supports_ipv4)
768 			return 0;
769 		else if(addr_is_ip6(&a->addr, a->addrlen) && supports_ipv6)
770 			return 0;
771 	}
772 	for(a = dp->result_list; a; a = a->next_result) {
773 		if(!addr_is_ip6(&a->addr, a->addrlen) && supports_ipv4)
774 			return 0;
775 		else if(addr_is_ip6(&a->addr, a->addrlen) && supports_ipv6)
776 			return 0;
777 	}
778 
779 	/* see if query is for one of the nameservers, which is glue */
780 	if( ((qinfo->qtype == LDNS_RR_TYPE_A && supports_ipv4) ||
781 		(qinfo->qtype == LDNS_RR_TYPE_AAAA && supports_ipv6)) &&
782 		dname_subdomain_c(qinfo->qname, dp->name) &&
783 		delegpt_find_ns(dp, qinfo->qname, qinfo->qname_len))
784 		return 1;
785 
786 	for(ns = dp->nslist; ns; ns = ns->next) {
787 		if(ns->resolved) /* skip failed targets */
788 			continue;
789 		if(!dname_subdomain_c(ns->name, dp->name))
790 			return 0; /* one address is not required glue */
791 	}
792 	return 1;
793 }
794 
795 int
796 iter_qname_indicates_dnssec(struct module_env* env, struct query_info *qinfo)
797 {
798 	struct trust_anchor* a;
799 	if(!env || !env->anchors || !qinfo || !qinfo->qname)
800 		return 0;
801 	/* a trust anchor exists above the name? */
802 	if((a=anchors_lookup(env->anchors, qinfo->qname, qinfo->qname_len,
803 		qinfo->qclass))) {
804 		if(a->numDS == 0 && a->numDNSKEY == 0) {
805 			/* insecure trust point */
806 			lock_basic_unlock(&a->lock);
807 			return 0;
808 		}
809 		lock_basic_unlock(&a->lock);
810 		return 1;
811 	}
812 	/* no trust anchor above it. */
813 	return 0;
814 }
815 
816 int
817 iter_indicates_dnssec(struct module_env* env, struct delegpt* dp,
818         struct dns_msg* msg, uint16_t dclass)
819 {
820 	struct trust_anchor* a;
821 	/* information not available, !env->anchors can be common */
822 	if(!env || !env->anchors || !dp || !dp->name)
823 		return 0;
824 	/* a trust anchor exists with this name, RRSIGs expected */
825 	if((a=anchor_find(env->anchors, dp->name, dp->namelabs, dp->namelen,
826 		dclass))) {
827 		if(a->numDS == 0 && a->numDNSKEY == 0) {
828 			/* insecure trust point */
829 			lock_basic_unlock(&a->lock);
830 			return 0;
831 		}
832 		lock_basic_unlock(&a->lock);
833 		return 1;
834 	}
835 	/* see if DS rrset was given, in AUTH section */
836 	if(msg && msg->rep &&
837 		reply_find_rrset_section_ns(msg->rep, dp->name, dp->namelen,
838 		LDNS_RR_TYPE_DS, dclass))
839 		return 1;
840 	/* look in key cache */
841 	if(env->key_cache) {
842 		struct key_entry_key* kk = key_cache_obtain(env->key_cache,
843 			dp->name, dp->namelen, dclass, env->scratch, *env->now);
844 		if(kk) {
845 			if(query_dname_compare(kk->name, dp->name) == 0) {
846 			  if(key_entry_isgood(kk) || key_entry_isbad(kk)) {
847 				regional_free_all(env->scratch);
848 				return 1;
849 			  } else if(key_entry_isnull(kk)) {
850 				regional_free_all(env->scratch);
851 				return 0;
852 			  }
853 			}
854 			regional_free_all(env->scratch);
855 		}
856 	}
857 	return 0;
858 }
859 
860 int
861 iter_msg_has_dnssec(struct dns_msg* msg)
862 {
863 	size_t i;
864 	if(!msg || !msg->rep)
865 		return 0;
866 	for(i=0; i<msg->rep->an_numrrsets + msg->rep->ns_numrrsets; i++) {
867 		if(((struct packed_rrset_data*)msg->rep->rrsets[i]->
868 			entry.data)->rrsig_count > 0)
869 			return 1;
870 	}
871 	/* empty message has no DNSSEC info, with DNSSEC the reply is
872 	 * not empty (NSEC) */
873 	return 0;
874 }
875 
876 int iter_msg_from_zone(struct dns_msg* msg, struct delegpt* dp,
877         enum response_type type, uint16_t dclass)
878 {
879 	if(!msg || !dp || !msg->rep || !dp->name)
880 		return 0;
881 	/* SOA RRset - always from reply zone */
882 	if(reply_find_rrset_section_an(msg->rep, dp->name, dp->namelen,
883 		LDNS_RR_TYPE_SOA, dclass) ||
884 	   reply_find_rrset_section_ns(msg->rep, dp->name, dp->namelen,
885 		LDNS_RR_TYPE_SOA, dclass))
886 		return 1;
887 	if(type == RESPONSE_TYPE_REFERRAL) {
888 		size_t i;
889 		/* if it adds a single label, i.e. we expect .com,
890 		 * and referral to example.com. NS ... , then origin zone
891 		 * is .com. For a referral to sub.example.com. NS ... then
892 		 * we do not know, since example.com. may be in between. */
893 		for(i=0; i<msg->rep->an_numrrsets+msg->rep->ns_numrrsets;
894 			i++) {
895 			struct ub_packed_rrset_key* s = msg->rep->rrsets[i];
896 			if(ntohs(s->rk.type) == LDNS_RR_TYPE_NS &&
897 				ntohs(s->rk.rrset_class) == dclass) {
898 				int l = dname_count_labels(s->rk.dname);
899 				if(l == dp->namelabs + 1 &&
900 					dname_strict_subdomain(s->rk.dname,
901 					l, dp->name, dp->namelabs))
902 					return 1;
903 			}
904 		}
905 		return 0;
906 	}
907 	log_assert(type==RESPONSE_TYPE_ANSWER || type==RESPONSE_TYPE_CNAME);
908 	/* not a referral, and not lame delegation (upwards), so,
909 	 * any NS rrset must be from the zone itself */
910 	if(reply_find_rrset_section_an(msg->rep, dp->name, dp->namelen,
911 		LDNS_RR_TYPE_NS, dclass) ||
912 	   reply_find_rrset_section_ns(msg->rep, dp->name, dp->namelen,
913 		LDNS_RR_TYPE_NS, dclass))
914 		return 1;
915 	/* a DNSKEY set is expected at the zone apex as well */
916 	/* this is for 'minimal responses' for DNSKEYs */
917 	if(reply_find_rrset_section_an(msg->rep, dp->name, dp->namelen,
918 		LDNS_RR_TYPE_DNSKEY, dclass))
919 		return 1;
920 	return 0;
921 }
922 
923 /**
924  * check equality of two rrsets
925  * @param k1: rrset
926  * @param k2: rrset
927  * @return true if equal
928  */
929 static int
930 rrset_equal(struct ub_packed_rrset_key* k1, struct ub_packed_rrset_key* k2)
931 {
932 	struct packed_rrset_data* d1 = (struct packed_rrset_data*)
933 		k1->entry.data;
934 	struct packed_rrset_data* d2 = (struct packed_rrset_data*)
935 		k2->entry.data;
936 	size_t i, t;
937 	if(k1->rk.dname_len != k2->rk.dname_len ||
938 		k1->rk.flags != k2->rk.flags ||
939 		k1->rk.type != k2->rk.type ||
940 		k1->rk.rrset_class != k2->rk.rrset_class ||
941 		query_dname_compare(k1->rk.dname, k2->rk.dname) != 0)
942 		return 0;
943 	if(	/* do not check ttl: d1->ttl != d2->ttl || */
944 		d1->count != d2->count ||
945 		d1->rrsig_count != d2->rrsig_count ||
946 		d1->trust != d2->trust ||
947 		d1->security != d2->security)
948 		return 0;
949 	t = d1->count + d1->rrsig_count;
950 	for(i=0; i<t; i++) {
951 		if(d1->rr_len[i] != d2->rr_len[i] ||
952 			/* no ttl check: d1->rr_ttl[i] != d2->rr_ttl[i] ||*/
953 			memcmp(d1->rr_data[i], d2->rr_data[i],
954 				d1->rr_len[i]) != 0)
955 			return 0;
956 	}
957 	return 1;
958 }
959 
960 /** compare rrsets and sort canonically.  Compares rrset name, type, class.
961  * return 0 if equal, +1 if x > y, and -1 if x < y.
962  */
963 static int
964 rrset_canonical_sort_cmp(const void* x, const void* y)
965 {
966 	struct ub_packed_rrset_key* rrx = *(struct ub_packed_rrset_key**)x;
967 	struct ub_packed_rrset_key* rry = *(struct ub_packed_rrset_key**)y;
968 	int r = dname_canonical_compare(rrx->rk.dname, rry->rk.dname);
969 	if(r != 0)
970 		return r;
971 	if(rrx->rk.type != rry->rk.type) {
972 		if(ntohs(rrx->rk.type) > ntohs(rry->rk.type))
973 			return 1;
974 		else	return -1;
975 	}
976 	if(rrx->rk.rrset_class != rry->rk.rrset_class) {
977 		if(ntohs(rrx->rk.rrset_class) > ntohs(rry->rk.rrset_class))
978 			return 1;
979 		else	return -1;
980 	}
981 	return 0;
982 }
983 
984 int
985 reply_equal(struct reply_info* p, struct reply_info* q, struct regional* region)
986 {
987 	size_t i;
988 	struct ub_packed_rrset_key** sorted_p, **sorted_q;
989 	if(p->flags != q->flags ||
990 		p->qdcount != q->qdcount ||
991 		/* do not check TTL, this may differ */
992 		/*
993 		p->ttl != q->ttl ||
994 		p->prefetch_ttl != q->prefetch_ttl ||
995 		*/
996 		p->security != q->security ||
997 		p->an_numrrsets != q->an_numrrsets ||
998 		p->ns_numrrsets != q->ns_numrrsets ||
999 		p->ar_numrrsets != q->ar_numrrsets ||
1000 		p->rrset_count != q->rrset_count)
1001 		return 0;
1002 	/* sort the rrsets in the authority and additional sections before
1003 	 * compare, the query and answer sections are ordered in the sequence
1004 	 * they should have (eg. one after the other for aliases). */
1005 	sorted_p = (struct ub_packed_rrset_key**)regional_alloc_init(
1006 		region, p->rrsets, sizeof(*sorted_p)*p->rrset_count);
1007 	if(!sorted_p) return 0;
1008 	log_assert(p->an_numrrsets + p->ns_numrrsets + p->ar_numrrsets <=
1009 		p->rrset_count);
1010 	qsort(sorted_p + p->an_numrrsets, p->ns_numrrsets,
1011 		sizeof(*sorted_p), rrset_canonical_sort_cmp);
1012 	qsort(sorted_p + p->an_numrrsets + p->ns_numrrsets, p->ar_numrrsets,
1013 		sizeof(*sorted_p), rrset_canonical_sort_cmp);
1014 
1015 	sorted_q = (struct ub_packed_rrset_key**)regional_alloc_init(
1016 		region, q->rrsets, sizeof(*sorted_q)*q->rrset_count);
1017 	if(!sorted_q) {
1018 		regional_free_all(region);
1019 		return 0;
1020 	}
1021 	log_assert(q->an_numrrsets + q->ns_numrrsets + q->ar_numrrsets <=
1022 		q->rrset_count);
1023 	qsort(sorted_q + q->an_numrrsets, q->ns_numrrsets,
1024 		sizeof(*sorted_q), rrset_canonical_sort_cmp);
1025 	qsort(sorted_q + q->an_numrrsets + q->ns_numrrsets, q->ar_numrrsets,
1026 		sizeof(*sorted_q), rrset_canonical_sort_cmp);
1027 
1028 	/* compare the rrsets */
1029 	for(i=0; i<p->rrset_count; i++) {
1030 		if(!rrset_equal(sorted_p[i], sorted_q[i])) {
1031 			if(!rrset_canonical_equal(region, sorted_p[i],
1032 				sorted_q[i])) {
1033 				regional_free_all(region);
1034 				return 0;
1035 			}
1036 		}
1037 	}
1038 	regional_free_all(region);
1039 	return 1;
1040 }
1041 
1042 void
1043 caps_strip_reply(struct reply_info* rep)
1044 {
1045 	size_t i;
1046 	if(!rep) return;
1047 	/* see if message is a referral, in which case the additional and
1048 	 * NS record cannot be removed */
1049 	/* referrals have the AA flag unset (strict check, not elsewhere in
1050 	 * unbound, but for 0x20 this is very convenient). */
1051 	if(!(rep->flags&BIT_AA))
1052 		return;
1053 	/* remove the additional section from the reply */
1054 	if(rep->ar_numrrsets != 0) {
1055 		verbose(VERB_ALGO, "caps fallback: removing additional section");
1056 		rep->rrset_count -= rep->ar_numrrsets;
1057 		rep->ar_numrrsets = 0;
1058 	}
1059 	/* is there an NS set in the authority section to remove? */
1060 	/* the failure case (Cisco firewalls) only has one rrset in authsec */
1061 	for(i=rep->an_numrrsets; i<rep->an_numrrsets+rep->ns_numrrsets; i++) {
1062 		struct ub_packed_rrset_key* s = rep->rrsets[i];
1063 		if(ntohs(s->rk.type) == LDNS_RR_TYPE_NS) {
1064 			/* remove NS rrset and break from loop (loop limits
1065 			 * have changed) */
1066 			/* move last rrset into this position (there is no
1067 			 * additional section any more) */
1068 			verbose(VERB_ALGO, "caps fallback: removing NS rrset");
1069 			if(i < rep->rrset_count-1)
1070 				rep->rrsets[i]=rep->rrsets[rep->rrset_count-1];
1071 			rep->rrset_count --;
1072 			rep->ns_numrrsets --;
1073 			break;
1074 		}
1075 	}
1076 }
1077 
1078 int caps_failed_rcode(struct reply_info* rep)
1079 {
1080 	return !(FLAGS_GET_RCODE(rep->flags) == LDNS_RCODE_NOERROR ||
1081 		FLAGS_GET_RCODE(rep->flags) == LDNS_RCODE_NXDOMAIN);
1082 }
1083 
1084 void
1085 iter_store_parentside_rrset(struct module_env* env,
1086 	struct ub_packed_rrset_key* rrset)
1087 {
1088 	struct rrset_ref ref;
1089 	rrset = packed_rrset_copy_alloc(rrset, env->alloc, *env->now);
1090 	if(!rrset) {
1091 		log_err("malloc failure in store_parentside_rrset");
1092 		return;
1093 	}
1094 	rrset->rk.flags |= PACKED_RRSET_PARENT_SIDE;
1095 	rrset->entry.hash = rrset_key_hash(&rrset->rk);
1096 	ref.key = rrset;
1097 	ref.id = rrset->id;
1098 	/* ignore ret: if it was in the cache, ref updated */
1099 	(void)rrset_cache_update(env->rrset_cache, &ref, env->alloc, *env->now);
1100 }
1101 
1102 /** fetch NS record from reply, if any */
1103 static struct ub_packed_rrset_key*
1104 reply_get_NS_rrset(struct reply_info* rep)
1105 {
1106 	size_t i;
1107 	for(i=0; i<rep->rrset_count; i++) {
1108 		if(rep->rrsets[i]->rk.type == htons(LDNS_RR_TYPE_NS)) {
1109 			return rep->rrsets[i];
1110 		}
1111 	}
1112 	return NULL;
1113 }
1114 
1115 void
1116 iter_store_parentside_NS(struct module_env* env, struct reply_info* rep)
1117 {
1118 	struct ub_packed_rrset_key* rrset = reply_get_NS_rrset(rep);
1119 	if(rrset) {
1120 		log_rrset_key(VERB_ALGO, "store parent-side NS", rrset);
1121 		iter_store_parentside_rrset(env, rrset);
1122 	}
1123 }
1124 
1125 void iter_store_parentside_neg(struct module_env* env,
1126         struct query_info* qinfo, struct reply_info* rep)
1127 {
1128 	/* TTL: NS from referral in iq->deleg_msg,
1129 	 *      or first RR from iq->response,
1130 	 *      or servfail5secs if !iq->response */
1131 	time_t ttl = NORR_TTL;
1132 	struct ub_packed_rrset_key* neg;
1133 	struct packed_rrset_data* newd;
1134 	if(rep) {
1135 		struct ub_packed_rrset_key* rrset = reply_get_NS_rrset(rep);
1136 		if(!rrset && rep->rrset_count != 0) rrset = rep->rrsets[0];
1137 		if(rrset) ttl = ub_packed_rrset_ttl(rrset);
1138 	}
1139 	/* create empty rrset to store */
1140 	neg = (struct ub_packed_rrset_key*)regional_alloc(env->scratch,
1141 	                sizeof(struct ub_packed_rrset_key));
1142 	if(!neg) {
1143 		log_err("out of memory in store_parentside_neg");
1144 		return;
1145 	}
1146 	memset(&neg->entry, 0, sizeof(neg->entry));
1147 	neg->entry.key = neg;
1148 	neg->rk.type = htons(qinfo->qtype);
1149 	neg->rk.rrset_class = htons(qinfo->qclass);
1150 	neg->rk.flags = 0;
1151 	neg->rk.dname = regional_alloc_init(env->scratch, qinfo->qname,
1152 		qinfo->qname_len);
1153 	if(!neg->rk.dname) {
1154 		log_err("out of memory in store_parentside_neg");
1155 		return;
1156 	}
1157 	neg->rk.dname_len = qinfo->qname_len;
1158 	neg->entry.hash = rrset_key_hash(&neg->rk);
1159 	newd = (struct packed_rrset_data*)regional_alloc_zero(env->scratch,
1160 		sizeof(struct packed_rrset_data) + sizeof(size_t) +
1161 		sizeof(uint8_t*) + sizeof(time_t) + sizeof(uint16_t));
1162 	if(!newd) {
1163 		log_err("out of memory in store_parentside_neg");
1164 		return;
1165 	}
1166 	neg->entry.data = newd;
1167 	newd->ttl = ttl;
1168 	/* entry must have one RR, otherwise not valid in cache.
1169 	 * put in one RR with empty rdata: those are ignored as nameserver */
1170 	newd->count = 1;
1171 	newd->rrsig_count = 0;
1172 	newd->trust = rrset_trust_ans_noAA;
1173 	newd->rr_len = (size_t*)((uint8_t*)newd +
1174 		sizeof(struct packed_rrset_data));
1175 	newd->rr_len[0] = 0 /* zero len rdata */ + sizeof(uint16_t);
1176 	packed_rrset_ptr_fixup(newd);
1177 	newd->rr_ttl[0] = newd->ttl;
1178 	sldns_write_uint16(newd->rr_data[0], 0 /* zero len rdata */);
1179 	/* store it */
1180 	log_rrset_key(VERB_ALGO, "store parent-side negative", neg);
1181 	iter_store_parentside_rrset(env, neg);
1182 }
1183 
1184 int
1185 iter_lookup_parent_NS_from_cache(struct module_env* env, struct delegpt* dp,
1186 	struct regional* region, struct query_info* qinfo)
1187 {
1188 	struct ub_packed_rrset_key* akey;
1189 	akey = rrset_cache_lookup(env->rrset_cache, dp->name,
1190 		dp->namelen, LDNS_RR_TYPE_NS, qinfo->qclass,
1191 		PACKED_RRSET_PARENT_SIDE, *env->now, 0);
1192 	if(akey) {
1193 		log_rrset_key(VERB_ALGO, "found parent-side NS in cache", akey);
1194 		dp->has_parent_side_NS = 1;
1195 		/* and mark the new names as lame */
1196 		if(!delegpt_rrset_add_ns(dp, region, akey, 1)) {
1197 			lock_rw_unlock(&akey->entry.lock);
1198 			return 0;
1199 		}
1200 		lock_rw_unlock(&akey->entry.lock);
1201 	}
1202 	return 1;
1203 }
1204 
1205 int iter_lookup_parent_glue_from_cache(struct module_env* env,
1206         struct delegpt* dp, struct regional* region, struct query_info* qinfo)
1207 {
1208 	struct ub_packed_rrset_key* akey;
1209 	struct delegpt_ns* ns;
1210 	size_t num = delegpt_count_targets(dp);
1211 	for(ns = dp->nslist; ns; ns = ns->next) {
1212 		/* get cached parentside A */
1213 		akey = rrset_cache_lookup(env->rrset_cache, ns->name,
1214 			ns->namelen, LDNS_RR_TYPE_A, qinfo->qclass,
1215 			PACKED_RRSET_PARENT_SIDE, *env->now, 0);
1216 		if(akey) {
1217 			log_rrset_key(VERB_ALGO, "found parent-side", akey);
1218 			ns->done_pside4 = 1;
1219 			/* a negative-cache-element has no addresses it adds */
1220 			if(!delegpt_add_rrset_A(dp, region, akey, 1, NULL))
1221 				log_err("malloc failure in lookup_parent_glue");
1222 			lock_rw_unlock(&akey->entry.lock);
1223 		}
1224 		/* get cached parentside AAAA */
1225 		akey = rrset_cache_lookup(env->rrset_cache, ns->name,
1226 			ns->namelen, LDNS_RR_TYPE_AAAA, qinfo->qclass,
1227 			PACKED_RRSET_PARENT_SIDE, *env->now, 0);
1228 		if(akey) {
1229 			log_rrset_key(VERB_ALGO, "found parent-side", akey);
1230 			ns->done_pside6 = 1;
1231 			/* a negative-cache-element has no addresses it adds */
1232 			if(!delegpt_add_rrset_AAAA(dp, region, akey, 1, NULL))
1233 				log_err("malloc failure in lookup_parent_glue");
1234 			lock_rw_unlock(&akey->entry.lock);
1235 		}
1236 	}
1237 	/* see if new (but lame) addresses have become available */
1238 	return delegpt_count_targets(dp) != num;
1239 }
1240 
1241 int
1242 iter_get_next_root(struct iter_hints* hints, struct iter_forwards* fwd,
1243 	uint16_t* c)
1244 {
1245 	uint16_t c1 = *c, c2 = *c;
1246 	int r1 = hints_next_root(hints, &c1);
1247 	int r2 = forwards_next_root(fwd, &c2);
1248 	if(!r1 && !r2) /* got none, end of list */
1249 		return 0;
1250 	else if(!r1) /* got one, return that */
1251 		*c = c2;
1252 	else if(!r2)
1253 		*c = c1;
1254 	else if(c1 < c2) /* got both take smallest */
1255 		*c = c1;
1256 	else	*c = c2;
1257 	return 1;
1258 }
1259 
1260 void
1261 iter_scrub_ds(struct dns_msg* msg, struct ub_packed_rrset_key* ns, uint8_t* z)
1262 {
1263 	/* Only the DS record for the delegation itself is expected.
1264 	 * We allow DS for everything between the bailiwick and the
1265 	 * zonecut, thus DS records must be at or above the zonecut.
1266 	 * And the DS records must be below the server authority zone.
1267 	 * The answer section is already scrubbed. */
1268 	size_t i = msg->rep->an_numrrsets;
1269 	while(i < (msg->rep->an_numrrsets + msg->rep->ns_numrrsets)) {
1270 		struct ub_packed_rrset_key* s = msg->rep->rrsets[i];
1271 		if(ntohs(s->rk.type) == LDNS_RR_TYPE_DS &&
1272 			(!ns || !dname_subdomain_c(ns->rk.dname, s->rk.dname)
1273 			|| query_dname_compare(z, s->rk.dname) == 0)) {
1274 			log_nametypeclass(VERB_ALGO, "removing irrelevant DS",
1275 				s->rk.dname, ntohs(s->rk.type),
1276 				ntohs(s->rk.rrset_class));
1277 			memmove(msg->rep->rrsets+i, msg->rep->rrsets+i+1,
1278 				sizeof(struct ub_packed_rrset_key*) *
1279 				(msg->rep->rrset_count-i-1));
1280 			msg->rep->ns_numrrsets--;
1281 			msg->rep->rrset_count--;
1282 			/* stay at same i, but new record */
1283 			continue;
1284 		}
1285 		i++;
1286 	}
1287 }
1288 
1289 void
1290 iter_scrub_nxdomain(struct dns_msg* msg)
1291 {
1292 	if(msg->rep->an_numrrsets == 0)
1293 		return;
1294 
1295 	memmove(msg->rep->rrsets, msg->rep->rrsets+msg->rep->an_numrrsets,
1296 		sizeof(struct ub_packed_rrset_key*) *
1297 		(msg->rep->rrset_count-msg->rep->an_numrrsets));
1298 	msg->rep->rrset_count -= msg->rep->an_numrrsets;
1299 	msg->rep->an_numrrsets = 0;
1300 }
1301 
1302 void iter_dec_attempts(struct delegpt* dp, int d, int outbound_msg_retry)
1303 {
1304 	struct delegpt_addr* a;
1305 	for(a=dp->target_list; a; a = a->next_target) {
1306 		if(a->attempts >= outbound_msg_retry) {
1307 			/* add back to result list */
1308 			a->next_result = dp->result_list;
1309 			dp->result_list = a;
1310 		}
1311 		if(a->attempts > d)
1312 			a->attempts -= d;
1313 		else a->attempts = 0;
1314 	}
1315 }
1316 
1317 void iter_merge_retry_counts(struct delegpt* dp, struct delegpt* old,
1318 	int outbound_msg_retry)
1319 {
1320 	struct delegpt_addr* a, *o, *prev;
1321 	for(a=dp->target_list; a; a = a->next_target) {
1322 		o = delegpt_find_addr(old, &a->addr, a->addrlen);
1323 		if(o) {
1324 			log_addr(VERB_ALGO, "copy attempt count previous dp",
1325 				&a->addr, a->addrlen);
1326 			a->attempts = o->attempts;
1327 		}
1328 	}
1329 	prev = NULL;
1330 	a = dp->usable_list;
1331 	while(a) {
1332 		if(a->attempts >= outbound_msg_retry) {
1333 			log_addr(VERB_ALGO, "remove from usable list dp",
1334 				&a->addr, a->addrlen);
1335 			/* remove from result list */
1336 			if(prev)
1337 				prev->next_usable = a->next_usable;
1338 			else	dp->usable_list = a->next_usable;
1339 			/* prev stays the same */
1340 			a = a->next_usable;
1341 			continue;
1342 		}
1343 		prev = a;
1344 		a = a->next_usable;
1345 	}
1346 }
1347 
1348 int
1349 iter_ds_toolow(struct dns_msg* msg, struct delegpt* dp)
1350 {
1351 	/* if for query example.com, there is example.com SOA or a subdomain
1352 	 * of example.com, then we are too low and need to fetch NS. */
1353 	size_t i;
1354 	/* if we have a DNAME or CNAME we are probably wrong */
1355 	/* if we have a qtype DS in the answer section, its fine */
1356 	for(i=0; i < msg->rep->an_numrrsets; i++) {
1357 		struct ub_packed_rrset_key* s = msg->rep->rrsets[i];
1358 		if(ntohs(s->rk.type) == LDNS_RR_TYPE_DNAME ||
1359 			ntohs(s->rk.type) == LDNS_RR_TYPE_CNAME) {
1360 			/* not the right answer, maybe too low, check the
1361 			 * RRSIG signer name (if there is any) for a hint
1362 			 * that it is from the dp zone anyway */
1363 			uint8_t* sname;
1364 			size_t slen;
1365 			val_find_rrset_signer(s, &sname, &slen);
1366 			if(sname && query_dname_compare(dp->name, sname)==0)
1367 				return 0; /* it is fine, from the right dp */
1368 			return 1;
1369 		}
1370 		if(ntohs(s->rk.type) == LDNS_RR_TYPE_DS)
1371 			return 0; /* fine, we have a DS record */
1372 	}
1373 	for(i=msg->rep->an_numrrsets;
1374 		i < msg->rep->an_numrrsets + msg->rep->ns_numrrsets; i++) {
1375 		struct ub_packed_rrset_key* s = msg->rep->rrsets[i];
1376 		if(ntohs(s->rk.type) == LDNS_RR_TYPE_SOA) {
1377 			if(dname_subdomain_c(s->rk.dname, msg->qinfo.qname))
1378 				return 1; /* point is too low */
1379 			if(query_dname_compare(s->rk.dname, dp->name)==0)
1380 				return 0; /* right dp */
1381 		}
1382 		if(ntohs(s->rk.type) == LDNS_RR_TYPE_NSEC ||
1383 			ntohs(s->rk.type) == LDNS_RR_TYPE_NSEC3) {
1384 			uint8_t* sname;
1385 			size_t slen;
1386 			val_find_rrset_signer(s, &sname, &slen);
1387 			if(sname && query_dname_compare(dp->name, sname)==0)
1388 				return 0; /* it is fine, from the right dp */
1389 			return 1;
1390 		}
1391 	}
1392 	/* we do not know */
1393 	return 1;
1394 }
1395 
1396 int iter_dp_cangodown(struct query_info* qinfo, struct delegpt* dp)
1397 {
1398 	/* no delegation point, do not see how we can go down,
1399 	 * robust check, it should really exist */
1400 	if(!dp) return 0;
1401 
1402 	/* see if dp equals the qname, then we cannot go down further */
1403 	if(query_dname_compare(qinfo->qname, dp->name) == 0)
1404 		return 0;
1405 	/* if dp is one label above the name we also cannot go down further */
1406 	if(dname_count_labels(qinfo->qname) == dp->namelabs+1)
1407 		return 0;
1408 	return 1;
1409 }
1410 
1411 int
1412 iter_stub_fwd_no_cache(struct module_qstate *qstate, struct query_info *qinf,
1413 	uint8_t** retdpname, size_t* retdpnamelen)
1414 {
1415 	struct iter_hints_stub *stub;
1416 	struct delegpt *dp;
1417 
1418 	/* Check for stub. */
1419 	stub = hints_lookup_stub(qstate->env->hints, qinf->qname,
1420 	    qinf->qclass, NULL);
1421 	dp = forwards_lookup(qstate->env->fwds, qinf->qname, qinf->qclass);
1422 
1423 	/* see if forward or stub is more pertinent */
1424 	if(stub && stub->dp && dp) {
1425 		if(dname_strict_subdomain(dp->name, dp->namelabs,
1426 			stub->dp->name, stub->dp->namelabs)) {
1427 			stub = NULL; /* ignore stub, forward is lower */
1428 		} else {
1429 			dp = NULL; /* ignore forward, stub is lower */
1430 		}
1431 	}
1432 
1433 	/* check stub */
1434 	if (stub != NULL && stub->dp != NULL) {
1435 		if(stub->dp->no_cache) {
1436 			char qname[255+1];
1437 			char dpname[255+1];
1438 			dname_str(qinf->qname, qname);
1439 			dname_str(stub->dp->name, dpname);
1440 			verbose(VERB_ALGO, "stub for %s %s has no_cache", qname, dpname);
1441 		}
1442 		if(retdpname) {
1443 			*retdpname = stub->dp->name;
1444 			*retdpnamelen = stub->dp->namelen;
1445 		}
1446 		return (stub->dp->no_cache);
1447 	}
1448 
1449 	/* Check for forward. */
1450 	if (dp) {
1451 		if(dp->no_cache) {
1452 			char qname[255+1];
1453 			char dpname[255+1];
1454 			dname_str(qinf->qname, qname);
1455 			dname_str(dp->name, dpname);
1456 			verbose(VERB_ALGO, "forward for %s %s has no_cache", qname, dpname);
1457 		}
1458 		if(retdpname) {
1459 			*retdpname = dp->name;
1460 			*retdpnamelen = dp->namelen;
1461 		}
1462 		return (dp->no_cache);
1463 	}
1464 	if(retdpname) {
1465 		*retdpname = NULL;
1466 		*retdpnamelen = 0;
1467 	}
1468 	return 0;
1469 }
1470 
1471 void iterator_set_ip46_support(struct module_stack* mods,
1472 	struct module_env* env, struct outside_network* outnet)
1473 {
1474 	int m = modstack_find(mods, "iterator");
1475 	struct iter_env* ie = NULL;
1476 	if(m == -1)
1477 		return;
1478 	ie = (struct iter_env*)env->modinfo[m];
1479 	if(outnet->pending == NULL)
1480 		return; /* we are in testbound, no rbtree for UDP */
1481 	if(outnet->num_ip4 == 0)
1482 		ie->supports_ipv4 = 0;
1483 	if(outnet->num_ip6 == 0)
1484 		ie->supports_ipv6 = 0;
1485 }
1486