xref: /freebsd/contrib/unbound/validator/val_neg.c (revision 50caa0ea0c16499c40e785b5aa37053b180b2830)
1 /*
2  * validator/val_neg.c - validator aggressive negative caching functions.
3  *
4  * Copyright (c) 2008, 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 helper functions for the validator module.
40  * The functions help with aggressive negative caching.
41  * This creates new denials of existence, and proofs for absence of types
42  * from cached NSEC records.
43  */
44 #include "config.h"
45 #ifdef HAVE_OPENSSL_SSL_H
46 #include <openssl/ssl.h>
47 #define NSEC3_SHA_LEN SHA_DIGEST_LENGTH
48 #else
49 #define NSEC3_SHA_LEN 20
50 #endif
51 #include "validator/val_neg.h"
52 #include "validator/val_nsec.h"
53 #include "validator/val_nsec3.h"
54 #include "validator/val_utils.h"
55 #include "util/data/dname.h"
56 #include "util/data/msgreply.h"
57 #include "util/log.h"
58 #include "util/net_help.h"
59 #include "util/config_file.h"
60 #include "services/cache/rrset.h"
61 #include "services/cache/dns.h"
62 #include "sldns/rrdef.h"
63 #include "sldns/sbuffer.h"
64 
65 /**
66  * The maximum salt length that the negative cache is willing to use.
67  * Larger salt increases the computation time, while recommendations are
68  * for zero salt length for zones.
69  */
70 #define MAX_SALT_LENGTH 64
71 
val_neg_data_compare(const void * a,const void * b)72 int val_neg_data_compare(const void* a, const void* b)
73 {
74 	struct val_neg_data* x = (struct val_neg_data*)a;
75 	struct val_neg_data* y = (struct val_neg_data*)b;
76 	int m;
77 	return dname_canon_lab_cmp(x->name, x->labs, y->name, y->labs, &m);
78 }
79 
val_neg_zone_compare(const void * a,const void * b)80 int val_neg_zone_compare(const void* a, const void* b)
81 {
82 	struct val_neg_zone* x = (struct val_neg_zone*)a;
83 	struct val_neg_zone* y = (struct val_neg_zone*)b;
84 	int m;
85 	if(x->dclass != y->dclass) {
86 		if(x->dclass < y->dclass)
87 			return -1;
88 		return 1;
89 	}
90 	return dname_canon_lab_cmp(x->name, x->labs, y->name, y->labs, &m);
91 }
92 
val_neg_create(struct config_file * cfg,size_t maxiter)93 struct val_neg_cache* val_neg_create(struct config_file* cfg, size_t maxiter)
94 {
95 	struct val_neg_cache* neg = (struct val_neg_cache*)calloc(1,
96 		sizeof(*neg));
97 	if(!neg) {
98 		log_err("Could not create neg cache: out of memory");
99 		return NULL;
100 	}
101 	neg->nsec3_max_iter = maxiter;
102 	neg->max = 1024*1024; /* 1 M is thousands of entries */
103 	if(cfg) neg->max = cfg->neg_cache_size;
104 	rbtree_init(&neg->tree, &val_neg_zone_compare);
105 	lock_basic_init(&neg->lock);
106 	lock_protect(&neg->lock, neg, sizeof(*neg));
107 	return neg;
108 }
109 
val_neg_get_mem(struct val_neg_cache * neg)110 size_t val_neg_get_mem(struct val_neg_cache* neg)
111 {
112 	size_t result;
113 	lock_basic_lock(&neg->lock);
114 	result = sizeof(*neg) + neg->use;
115 	lock_basic_unlock(&neg->lock);
116 	return result;
117 }
118 
119 /** clear datas on cache deletion */
120 static void
neg_clear_datas(rbnode_type * n,void * ATTR_UNUSED (arg))121 neg_clear_datas(rbnode_type* n, void* ATTR_UNUSED(arg))
122 {
123 	struct val_neg_data* d = (struct val_neg_data*)n;
124 	free(d->name);
125 	free(d);
126 }
127 
128 /** clear zones on cache deletion */
129 static void
neg_clear_zones(rbnode_type * n,void * ATTR_UNUSED (arg))130 neg_clear_zones(rbnode_type* n, void* ATTR_UNUSED(arg))
131 {
132 	struct val_neg_zone* z = (struct val_neg_zone*)n;
133 	/* delete all the rrset entries in the tree */
134 	traverse_postorder(&z->tree, &neg_clear_datas, NULL);
135 	free(z->nsec3_salt);
136 	free(z->name);
137 	free(z);
138 }
139 
neg_cache_delete(struct val_neg_cache * neg)140 void neg_cache_delete(struct val_neg_cache* neg)
141 {
142 	if(!neg) return;
143 	lock_basic_destroy(&neg->lock);
144 	/* delete all the zones in the tree */
145 	traverse_postorder(&neg->tree, &neg_clear_zones, NULL);
146 	free(neg);
147 }
148 
149 /**
150  * Put data element at the front of the LRU list.
151  * @param neg: negative cache with LRU start and end.
152  * @param data: this data is fronted.
153  */
neg_lru_front(struct val_neg_cache * neg,struct val_neg_data * data)154 static void neg_lru_front(struct val_neg_cache* neg,
155 	struct val_neg_data* data)
156 {
157 	data->prev = NULL;
158 	data->next = neg->first;
159 	if(!neg->first)
160 		neg->last = data;
161 	else	neg->first->prev = data;
162 	neg->first = data;
163 }
164 
165 /**
166  * Remove data element from LRU list.
167  * @param neg: negative cache with LRU start and end.
168  * @param data: this data is removed from the list.
169  */
neg_lru_remove(struct val_neg_cache * neg,struct val_neg_data * data)170 static void neg_lru_remove(struct val_neg_cache* neg,
171 	struct val_neg_data* data)
172 {
173 	if(data->prev)
174 		data->prev->next = data->next;
175 	else	neg->first = data->next;
176 	if(data->next)
177 		data->next->prev = data->prev;
178 	else	neg->last = data->prev;
179 }
180 
181 /**
182  * Touch LRU for data element, put it at the start of the LRU list.
183  * @param neg: negative cache with LRU start and end.
184  * @param data: this data is used.
185  */
neg_lru_touch(struct val_neg_cache * neg,struct val_neg_data * data)186 static void neg_lru_touch(struct val_neg_cache* neg,
187 	struct val_neg_data* data)
188 {
189 	if(data == neg->first)
190 		return; /* nothing to do */
191 	/* remove from current lru position */
192 	neg_lru_remove(neg, data);
193 	/* add at front */
194 	neg_lru_front(neg, data);
195 }
196 
197 /**
198  * Delete a zone element from the negative cache.
199  * May delete other zone elements to keep tree coherent, or
200  * only mark the element as 'not in use'.
201  * @param neg: negative cache.
202  * @param z: zone element to delete.
203  */
neg_delete_zone(struct val_neg_cache * neg,struct val_neg_zone * z)204 static void neg_delete_zone(struct val_neg_cache* neg, struct val_neg_zone* z)
205 {
206 	struct val_neg_zone* p, *np;
207 	if(!z) return;
208 	log_assert(z->in_use);
209 	log_assert(z->count > 0);
210 	z->in_use = 0;
211 
212 	/* go up the tree and reduce counts */
213 	p = z;
214 	while(p) {
215 		log_assert(p->count > 0);
216 		p->count --;
217 		p = p->parent;
218 	}
219 
220 	/* remove zones with zero count */
221 	p = z;
222 	while(p && p->count == 0) {
223 		np = p->parent;
224 		(void)rbtree_delete(&neg->tree, &p->node);
225 		neg->use -= p->len + sizeof(*p);
226 		free(p->nsec3_salt);
227 		free(p->name);
228 		free(p);
229 		p = np;
230 	}
231 }
232 
neg_delete_data(struct val_neg_cache * neg,struct val_neg_data * el)233 void neg_delete_data(struct val_neg_cache* neg, struct val_neg_data* el)
234 {
235 	struct val_neg_zone* z;
236 	struct val_neg_data* p, *np;
237 	if(!el) return;
238 	z = el->zone;
239 	log_assert(el->in_use);
240 	log_assert(el->count > 0);
241 	el->in_use = 0;
242 
243 	/* remove it from the lru list */
244 	neg_lru_remove(neg, el);
245 	log_assert(neg->first != el && neg->last != el);
246 
247 	/* go up the tree and reduce counts */
248 	p = el;
249 	while(p) {
250 		log_assert(p->count > 0);
251 		p->count --;
252 		p = p->parent;
253 	}
254 
255 	/* delete 0 count items from tree */
256 	p = el;
257 	while(p && p->count == 0) {
258 		np = p->parent;
259 		(void)rbtree_delete(&z->tree, &p->node);
260 		neg->use -= p->len + sizeof(*p);
261 		free(p->name);
262 		free(p);
263 		p = np;
264 	}
265 
266 	/* check if the zone is now unused */
267 	if(z->tree.count == 0) {
268 		neg_delete_zone(neg, z);
269 	}
270 }
271 
272 /**
273  * Create more space in negative cache
274  * The oldest elements are deleted until enough space is present.
275  * Empty zones are deleted.
276  * @param neg: negative cache.
277  * @param need: how many bytes are needed.
278  */
neg_make_space(struct val_neg_cache * neg,size_t need)279 static void neg_make_space(struct val_neg_cache* neg, size_t need)
280 {
281 	/* delete elements until enough space or its empty */
282 	while(neg->last && neg->max < neg->use + need) {
283 		neg_delete_data(neg, neg->last);
284 	}
285 }
286 
neg_find_zone(struct val_neg_cache * neg,uint8_t * nm,size_t len,uint16_t dclass)287 struct val_neg_zone* neg_find_zone(struct val_neg_cache* neg,
288 	uint8_t* nm, size_t len, uint16_t dclass)
289 {
290 	struct val_neg_zone lookfor;
291 	struct val_neg_zone* result;
292 	lookfor.node.key = &lookfor;
293 	lookfor.name = nm;
294 	lookfor.len = len;
295 	lookfor.labs = dname_count_labels(lookfor.name);
296 	lookfor.dclass = dclass;
297 
298 	result = (struct val_neg_zone*)
299 		rbtree_search(&neg->tree, lookfor.node.key);
300 	return result;
301 }
302 
303 /**
304  * Find the given data
305  * @param zone: negative zone
306  * @param nm: what to look for.
307  * @param len: length of nm
308  * @param labs: labels in nm
309  * @return data or NULL if not found.
310  */
neg_find_data(struct val_neg_zone * zone,uint8_t * nm,size_t len,int labs)311 static struct val_neg_data* neg_find_data(struct val_neg_zone* zone,
312 	uint8_t* nm, size_t len, int labs)
313 {
314 	struct val_neg_data lookfor;
315 	struct val_neg_data* result;
316 	lookfor.node.key = &lookfor;
317 	lookfor.name = nm;
318 	lookfor.len = len;
319 	lookfor.labs = labs;
320 
321 	result = (struct val_neg_data*)
322 		rbtree_search(&zone->tree, lookfor.node.key);
323 	return result;
324 }
325 
326 /**
327  * Calculate space needed for the data and all its parents
328  * @param rep: NSEC entries.
329  * @return size.
330  */
calc_data_need(struct reply_info * rep)331 static size_t calc_data_need(struct reply_info* rep)
332 {
333 	uint8_t* d;
334 	size_t i, len, res = 0;
335 
336 	for(i=rep->an_numrrsets; i<rep->an_numrrsets+rep->ns_numrrsets; i++) {
337 		if(ntohs(rep->rrsets[i]->rk.type) == LDNS_RR_TYPE_NSEC) {
338 			d = rep->rrsets[i]->rk.dname;
339 			len = rep->rrsets[i]->rk.dname_len;
340 			res = sizeof(struct val_neg_data) + len;
341 			while(!dname_is_root(d)) {
342 				log_assert(len > 1); /* not root label */
343 				dname_remove_label(&d, &len);
344 				res += sizeof(struct val_neg_data) + len;
345 			}
346 		}
347 	}
348 	return res;
349 }
350 
351 /**
352  * Calculate space needed for zone and all its parents
353  * @param d: name of zone
354  * @param len: length of name
355  * @return size.
356  */
calc_zone_need(uint8_t * d,size_t len)357 static size_t calc_zone_need(uint8_t* d, size_t len)
358 {
359 	size_t res = sizeof(struct val_neg_zone) + len;
360 	while(!dname_is_root(d)) {
361 		log_assert(len > 1); /* not root label */
362 		dname_remove_label(&d, &len);
363 		res += sizeof(struct val_neg_zone) + len;
364 	}
365 	return res;
366 }
367 
368 /**
369  * Find closest existing parent zone of the given name.
370  * @param neg: negative cache.
371  * @param nm: name to look for
372  * @param nm_len: length of nm
373  * @param labs: labelcount of nm.
374  * @param qclass: class.
375  * @return the zone or NULL if none found.
376  */
neg_closest_zone_parent(struct val_neg_cache * neg,uint8_t * nm,size_t nm_len,int labs,uint16_t qclass)377 static struct val_neg_zone* neg_closest_zone_parent(struct val_neg_cache* neg,
378 	uint8_t* nm, size_t nm_len, int labs, uint16_t qclass)
379 {
380 	struct val_neg_zone key;
381 	struct val_neg_zone* result;
382 	rbnode_type* res = NULL;
383 	key.node.key = &key;
384 	key.name = nm;
385 	key.len = nm_len;
386 	key.labs = labs;
387 	key.dclass = qclass;
388 	if(rbtree_find_less_equal(&neg->tree, &key, &res)) {
389 		/* exact match */
390 		result = (struct val_neg_zone*)res;
391 	} else {
392 		/* smaller element (or no element) */
393 		int m;
394 		result = (struct val_neg_zone*)res;
395 		if(!result || result->dclass != qclass)
396 			return NULL;
397 		/* count number of labels matched */
398 		(void)dname_lab_cmp(result->name, result->labs, key.name,
399 			key.labs, &m);
400 		while(result) { /* go up until qname is subdomain of stub */
401 			if(result->labs <= m)
402 				break;
403 			result = result->parent;
404 		}
405 	}
406 	return result;
407 }
408 
409 /**
410  * Find closest existing parent data for the given name.
411  * @param zone: to look in.
412  * @param nm: name to look for
413  * @param nm_len: length of nm
414  * @param labs: labelcount of nm.
415  * @return the data or NULL if none found.
416  */
neg_closest_data_parent(struct val_neg_zone * zone,uint8_t * nm,size_t nm_len,int labs)417 static struct val_neg_data* neg_closest_data_parent(
418 	struct val_neg_zone* zone, uint8_t* nm, size_t nm_len, int labs)
419 {
420 	struct val_neg_data key;
421 	struct val_neg_data* result;
422 	rbnode_type* res = NULL;
423 	key.node.key = &key;
424 	key.name = nm;
425 	key.len = nm_len;
426 	key.labs = labs;
427 	if(rbtree_find_less_equal(&zone->tree, &key, &res)) {
428 		/* exact match */
429 		result = (struct val_neg_data*)res;
430 	} else {
431 		/* smaller element (or no element) */
432 		int m;
433 		result = (struct val_neg_data*)res;
434 		if(!result)
435 			return NULL;
436 		/* count number of labels matched */
437 		(void)dname_lab_cmp(result->name, result->labs, key.name,
438 			key.labs, &m);
439 		while(result) { /* go up until qname is subdomain of stub */
440 			if(result->labs <= m)
441 				break;
442 			result = result->parent;
443 		}
444 	}
445 	return result;
446 }
447 
448 /**
449  * Create a single zone node
450  * @param nm: name for zone (copied)
451  * @param nm_len: length of name
452  * @param labs: labels in name.
453  * @param dclass: class of zone, host order.
454  * @return new zone or NULL on failure
455  */
neg_setup_zone_node(uint8_t * nm,size_t nm_len,int labs,uint16_t dclass)456 static struct val_neg_zone* neg_setup_zone_node(
457 	uint8_t* nm, size_t nm_len, int labs, uint16_t dclass)
458 {
459 	struct val_neg_zone* zone =
460 		(struct val_neg_zone*)calloc(1, sizeof(*zone));
461 	if(!zone) {
462 		return NULL;
463 	}
464 	zone->node.key = zone;
465 	zone->name = memdup(nm, nm_len);
466 	if(!zone->name) {
467 		free(zone);
468 		return NULL;
469 	}
470 	zone->len = nm_len;
471 	zone->labs = labs;
472 	zone->dclass = dclass;
473 
474 	rbtree_init(&zone->tree, &val_neg_data_compare);
475 	return zone;
476 }
477 
478 /**
479  * Create a linked list of parent zones, starting at longname ending on
480  * the parent (can be NULL, creates to the root).
481  * @param nm: name for lowest in chain
482  * @param nm_len: length of name
483  * @param labs: labels in name.
484  * @param dclass: class of zone.
485  * @param parent: NULL for to root, else so it fits under here.
486  * @return zone; a chain of zones and their parents up to the parent.
487  *  	or NULL on malloc failure
488  */
neg_zone_chain(uint8_t * nm,size_t nm_len,int labs,uint16_t dclass,struct val_neg_zone * parent)489 static struct val_neg_zone* neg_zone_chain(
490 	uint8_t* nm, size_t nm_len, int labs, uint16_t dclass,
491 	struct val_neg_zone* parent)
492 {
493 	int i;
494 	int tolabs = parent?parent->labs:0;
495 	struct val_neg_zone* zone, *prev = NULL, *first = NULL;
496 
497 	/* create the new subtree, i is labelcount of current creation */
498 	/* this creates a 'first' to z->parent=NULL list of zones */
499 	for(i=labs; i!=tolabs; i--) {
500 		/* create new item */
501 		zone = neg_setup_zone_node(nm, nm_len, i, dclass);
502 		if(!zone) {
503 			/* need to delete other allocations in this routine!*/
504 			struct val_neg_zone* p=first, *np;
505 			while(p) {
506 				np = p->parent;
507 				free(p->name);
508 				free(p);
509 				p = np;
510 			}
511 			return NULL;
512 		}
513 		if(i == labs) {
514 			first = zone;
515 		} else {
516 			prev->parent = zone;
517 		}
518 		/* prepare for next name */
519 		prev = zone;
520 		dname_remove_label(&nm, &nm_len);
521 	}
522 	return first;
523 }
524 
val_neg_zone_take_inuse(struct val_neg_zone * zone)525 void val_neg_zone_take_inuse(struct val_neg_zone* zone)
526 {
527 	if(!zone->in_use) {
528 		struct val_neg_zone* p;
529 		zone->in_use = 1;
530 		/* increase usage count of all parents */
531 		for(p=zone; p; p = p->parent) {
532 			p->count++;
533 		}
534 	}
535 }
536 
neg_create_zone(struct val_neg_cache * neg,uint8_t * nm,size_t nm_len,uint16_t dclass)537 struct val_neg_zone* neg_create_zone(struct val_neg_cache* neg,
538 	uint8_t* nm, size_t nm_len, uint16_t dclass)
539 {
540 	struct val_neg_zone* zone;
541 	struct val_neg_zone* parent;
542 	struct val_neg_zone* p, *np;
543 	int labs = dname_count_labels(nm);
544 
545 	/* find closest enclosing parent zone that (still) exists */
546 	parent = neg_closest_zone_parent(neg, nm, nm_len, labs, dclass);
547 	if(parent && query_dname_compare(parent->name, nm) == 0)
548 		return parent; /* already exists, weird */
549 	/* if parent exists, it is in use */
550 	log_assert(!parent || parent->count > 0);
551 	zone = neg_zone_chain(nm, nm_len, labs, dclass, parent);
552 	if(!zone) {
553 		return NULL;
554 	}
555 
556 	/* insert the list of zones into the tree */
557 	p = zone;
558 	while(p) {
559 		np = p->parent;
560 		/* mem use */
561 		neg->use += sizeof(struct val_neg_zone) + p->len;
562 		/* insert in tree */
563 		(void)rbtree_insert(&neg->tree, &p->node);
564 		/* last one needs proper parent pointer */
565 		if(np == NULL)
566 			p->parent = parent;
567 		p = np;
568 	}
569 	return zone;
570 }
571 
572 /** find zone name of message, returns the SOA record */
reply_find_soa(struct reply_info * rep)573 static struct ub_packed_rrset_key* reply_find_soa(struct reply_info* rep)
574 {
575 	size_t i;
576 	for(i=rep->an_numrrsets; i< rep->an_numrrsets+rep->ns_numrrsets; i++){
577 		if(ntohs(rep->rrsets[i]->rk.type) == LDNS_RR_TYPE_SOA)
578 			return rep->rrsets[i];
579 	}
580 	return NULL;
581 }
582 
583 /** see if the reply has NSEC records worthy of caching */
reply_has_nsec(struct reply_info * rep)584 static int reply_has_nsec(struct reply_info* rep)
585 {
586 	size_t i;
587 	struct packed_rrset_data* d;
588 	if(rep->security != sec_status_secure)
589 		return 0;
590 	for(i=rep->an_numrrsets; i< rep->an_numrrsets+rep->ns_numrrsets; i++){
591 		if(ntohs(rep->rrsets[i]->rk.type) == LDNS_RR_TYPE_NSEC) {
592 			d = (struct packed_rrset_data*)rep->rrsets[i]->
593 				entry.data;
594 			if(d->security == sec_status_secure)
595 				return 1;
596 		}
597 	}
598 	return 0;
599 }
600 
601 
602 /**
603  * Create single node of data element.
604  * @param nm: name (copied)
605  * @param nm_len: length of name
606  * @param labs: labels in name.
607  * @return element with name nm, or NULL malloc failure.
608  */
neg_setup_data_node(uint8_t * nm,size_t nm_len,int labs)609 static struct val_neg_data* neg_setup_data_node(
610 	uint8_t* nm, size_t nm_len, int labs)
611 {
612 	struct val_neg_data* el;
613 	el = (struct val_neg_data*)calloc(1, sizeof(*el));
614 	if(!el) {
615 		return NULL;
616 	}
617 	el->node.key = el;
618 	el->name = memdup(nm, nm_len);
619 	if(!el->name) {
620 		free(el);
621 		return NULL;
622 	}
623 	el->len = nm_len;
624 	el->labs = labs;
625 	return el;
626 }
627 
628 /**
629  * Create chain of data element and parents
630  * @param nm: name
631  * @param nm_len: length of name
632  * @param labs: labels in name.
633  * @param parent: up to where to make, if NULL up to root label.
634  * @return lowest element with name nm, or NULL malloc failure.
635  */
neg_data_chain(uint8_t * nm,size_t nm_len,int labs,struct val_neg_data * parent)636 static struct val_neg_data* neg_data_chain(
637 	uint8_t* nm, size_t nm_len, int labs, struct val_neg_data* parent)
638 {
639 	int i;
640 	int tolabs = parent?parent->labs:0;
641 	struct val_neg_data* el, *first = NULL, *prev = NULL;
642 
643 	/* create the new subtree, i is labelcount of current creation */
644 	/* this creates a 'first' to z->parent=NULL list of zones */
645 	for(i=labs; i!=tolabs; i--) {
646 		/* create new item */
647 		el = neg_setup_data_node(nm, nm_len, i);
648 		if(!el) {
649 			/* need to delete other allocations in this routine!*/
650 			struct val_neg_data* p = first, *np;
651 			while(p) {
652 				np = p->parent;
653 				free(p->name);
654 				free(p);
655 				p = np;
656 			}
657 			return NULL;
658 		}
659 		if(i == labs) {
660 			first = el;
661 		} else {
662 			prev->parent = el;
663 		}
664 
665 		/* prepare for next name */
666 		prev = el;
667 		dname_remove_label(&nm, &nm_len);
668 	}
669 	return first;
670 }
671 
672 /**
673  * Remove NSEC records between start and end points.
674  * By walking the tree, the tree is sorted canonically.
675  * @param neg: negative cache.
676  * @param zone: the zone
677  * @param el: element to start walking at.
678  * @param nsec: the nsec record with the end point
679  */
wipeout(struct val_neg_cache * neg,struct val_neg_zone * zone,struct val_neg_data * el,struct ub_packed_rrset_key * nsec)680 static void wipeout(struct val_neg_cache* neg, struct val_neg_zone* zone,
681 	struct val_neg_data* el, struct ub_packed_rrset_key* nsec)
682 {
683 	struct packed_rrset_data* d = (struct packed_rrset_data*)nsec->
684 		entry.data;
685 	uint8_t* end;
686 	size_t end_len;
687 	int end_labs, m;
688 	rbnode_type* walk, *next;
689 	struct val_neg_data* cur;
690 	uint8_t buf[257];
691 	/* get endpoint */
692 	if(!d || d->count == 0 || d->rr_len[0] < 2+1)
693 		return;
694 	if(ntohs(nsec->rk.type) == LDNS_RR_TYPE_NSEC) {
695 		end = d->rr_data[0]+2;
696 		end_len = dname_valid(end, d->rr_len[0]-2);
697 		end_labs = dname_count_labels(end);
698 	} else {
699 		/* NSEC3 */
700 		if(!nsec3_get_nextowner_b32(nsec, 0, buf, sizeof(buf)))
701 			return;
702 		end = buf;
703 		end_labs = dname_count_size_labels(end, &end_len);
704 	}
705 
706 	/* sanity check, both owner and end must be below the zone apex */
707 	if(!dname_subdomain_c(el->name, zone->name) ||
708 		!dname_subdomain_c(end, zone->name))
709 		return;
710 
711 	/* detect end of zone NSEC ; wipe until the end of zone */
712 	if(query_dname_compare(end, zone->name) == 0) {
713 		end = NULL;
714 	}
715 
716 	walk = rbtree_next(&el->node);
717 	while(walk && walk != RBTREE_NULL) {
718 		cur = (struct val_neg_data*)walk;
719 		/* sanity check: must be larger than start */
720 		if(dname_canon_lab_cmp(cur->name, cur->labs,
721 			el->name, el->labs, &m) <= 0) {
722 			/* r == 0 skip original record. */
723 			/* r < 0  too small! */
724 			walk = rbtree_next(walk);
725 			continue;
726 		}
727 		/* stop at endpoint, also data at empty nonterminals must be
728 		 * removed (no NSECs there) so everything between
729 		 * start and end */
730 		if(end && dname_canon_lab_cmp(cur->name, cur->labs,
731 			end, end_labs, &m) >= 0) {
732 			break;
733 		}
734 		/* this element has to be deleted, but we cannot do it
735 		 * now, because we are walking the tree still ... */
736 		/* get the next element: */
737 		next = rbtree_next(walk);
738 		/* now delete the original element, this may trigger
739 		 * rbtree rebalances, but really, the next element is
740 		 * the one we need.
741 		 * But it may trigger delete of other data and the
742 		 * entire zone. However, if that happens, this is done
743 		 * by deleting the *parents* of the element for deletion,
744 		 * and maybe also the entire zone if it is empty.
745 		 * But parents are smaller in canonical compare, thus,
746 		 * if a larger element exists, then it is not a parent,
747 		 * it cannot get deleted, the zone cannot get empty.
748 		 * If the next==NULL, then zone can be empty. */
749 		if(cur->in_use)
750 			neg_delete_data(neg, cur);
751 		walk = next;
752 	}
753 }
754 
neg_insert_data(struct val_neg_cache * neg,struct val_neg_zone * zone,struct ub_packed_rrset_key * nsec)755 void neg_insert_data(struct val_neg_cache* neg,
756 	struct val_neg_zone* zone, struct ub_packed_rrset_key* nsec)
757 {
758 	struct packed_rrset_data* d;
759 	struct val_neg_data* parent;
760 	struct val_neg_data* el;
761 	uint8_t* nm = nsec->rk.dname;
762 	size_t nm_len = nsec->rk.dname_len;
763 	int labs = dname_count_labels(nsec->rk.dname);
764 
765 	d = (struct packed_rrset_data*)nsec->entry.data;
766 	if( !(d->security == sec_status_secure ||
767 		(d->security == sec_status_unchecked && d->rrsig_count > 0)))
768 		return;
769 	log_nametypeclass(VERB_ALGO, "negcache rr",
770 		nsec->rk.dname, ntohs(nsec->rk.type),
771 		ntohs(nsec->rk.rrset_class));
772 
773 	/* find closest enclosing parent data that (still) exists */
774 	parent = neg_closest_data_parent(zone, nm, nm_len, labs);
775 	if(parent && query_dname_compare(parent->name, nm) == 0) {
776 		/* perfect match already exists */
777 		log_assert(parent->count > 0);
778 		el = parent;
779 	} else {
780 		struct val_neg_data* p, *np;
781 
782 		/* create subtree for perfect match */
783 		/* if parent exists, it is in use */
784 		log_assert(!parent || parent->count > 0);
785 
786 		el = neg_data_chain(nm, nm_len, labs, parent);
787 		if(!el) {
788 			log_err("out of memory inserting NSEC negative cache");
789 			return;
790 		}
791 		el->in_use = 0; /* set on below */
792 
793 		/* insert the list of zones into the tree */
794 		p = el;
795 		while(p) {
796 			np = p->parent;
797 			/* mem use */
798 			neg->use += sizeof(struct val_neg_data) + p->len;
799 			/* insert in tree */
800 			p->zone = zone;
801 			(void)rbtree_insert(&zone->tree, &p->node);
802 			/* last one needs proper parent pointer */
803 			if(np == NULL)
804 				p->parent = parent;
805 			p = np;
806 		}
807 	}
808 
809 	if(!el->in_use) {
810 		struct val_neg_data* p;
811 
812 		el->in_use = 1;
813 		/* increase usage count of all parents */
814 		for(p=el; p; p = p->parent) {
815 			p->count++;
816 		}
817 
818 		neg_lru_front(neg, el);
819 	} else {
820 		/* in use, bring to front, lru */
821 		neg_lru_touch(neg, el);
822 	}
823 
824 	/* if nsec3 store last used parameters */
825 	if(ntohs(nsec->rk.type) == LDNS_RR_TYPE_NSEC3) {
826 		int h;
827 		uint8_t* s;
828 		size_t slen, it;
829 		if(nsec3_get_params(nsec, 0, &h, &it, &s, &slen) &&
830 			it <= neg->nsec3_max_iter &&
831 			(h != zone->nsec3_hash || it != zone->nsec3_iter ||
832 			slen != zone->nsec3_saltlen ||
833 			(slen != 0 && zone->nsec3_salt && s
834 			  && memcmp(zone->nsec3_salt, s, slen) != 0))) {
835 
836 			if(slen > MAX_SALT_LENGTH) {
837 				/* RFC 9276 s3.1: operators SHOULD NOT use a salt; large
838 				 * salts inflate per-hash block count. Decline to cache. */
839 				return;
840 			} else if(slen > 0) {
841 				uint8_t* sa = memdup(s, slen);
842 				if(sa) {
843 					free(zone->nsec3_salt);
844 					zone->nsec3_salt = sa;
845 					zone->nsec3_saltlen = slen;
846 					zone->nsec3_iter = it;
847 					zone->nsec3_hash = h;
848 				}
849 			} else {
850 				free(zone->nsec3_salt);
851 				zone->nsec3_salt = NULL;
852 				zone->nsec3_saltlen = 0;
853 				zone->nsec3_iter = it;
854 				zone->nsec3_hash = h;
855 			}
856 		}
857 	}
858 
859 	/* wipe out the cache items between NSEC start and end */
860 	wipeout(neg, zone, el, nsec);
861 }
862 
863 /** see if the reply has signed NSEC records and return the signer */
reply_nsec_signer(struct reply_info * rep,size_t * signer_len,uint16_t * dclass)864 static uint8_t* reply_nsec_signer(struct reply_info* rep, size_t* signer_len,
865 	uint16_t* dclass)
866 {
867 	size_t i;
868 	struct packed_rrset_data* d;
869 	uint8_t* s;
870 	for(i=rep->an_numrrsets; i< rep->an_numrrsets+rep->ns_numrrsets; i++){
871 		if(ntohs(rep->rrsets[i]->rk.type) == LDNS_RR_TYPE_NSEC ||
872 			ntohs(rep->rrsets[i]->rk.type) == LDNS_RR_TYPE_NSEC3) {
873 			d = (struct packed_rrset_data*)rep->rrsets[i]->
874 				entry.data;
875 			/* return first signer name of first NSEC */
876 			if(d->rrsig_count != 0) {
877 				val_find_rrset_signer(rep->rrsets[i],
878 					&s, signer_len);
879 				if(s && *signer_len) {
880 					*dclass = ntohs(rep->rrsets[i]->
881 						rk.rrset_class);
882 					return s;
883 				}
884 			}
885 		}
886 	}
887 	return 0;
888 }
889 
val_neg_addreply(struct val_neg_cache * neg,struct reply_info * rep)890 void val_neg_addreply(struct val_neg_cache* neg, struct reply_info* rep)
891 {
892 	size_t i, need;
893 	struct ub_packed_rrset_key* soa;
894 	uint8_t* dname = NULL;
895 	size_t dname_len;
896 	uint16_t rrset_class;
897 	struct val_neg_zone* zone;
898 	/* see if secure nsecs inside */
899 	if(!reply_has_nsec(rep))
900 		return;
901 	/* find the zone name in message */
902 	if((soa = reply_find_soa(rep))) {
903 		dname = soa->rk.dname;
904 		dname_len = soa->rk.dname_len;
905 		rrset_class = ntohs(soa->rk.rrset_class);
906 	}
907 	else {
908 		/* No SOA in positive (wildcard) answer. Use signer from the
909 		 * validated answer RRsets' signature. */
910 		if(!(dname = reply_nsec_signer(rep, &dname_len, &rrset_class)))
911 			return;
912 	}
913 
914 	log_nametypeclass(VERB_ALGO, "negcache insert for zone",
915 		dname, LDNS_RR_TYPE_SOA, rrset_class);
916 
917 	/* ask for enough space to store all of it */
918 	need = calc_data_need(rep) +
919 		calc_zone_need(dname, dname_len);
920 	lock_basic_lock(&neg->lock);
921 	neg_make_space(neg, need);
922 
923 	/* find or create the zone entry */
924 	zone = neg_find_zone(neg, dname, dname_len, rrset_class);
925 	if(!zone) {
926 		if(!(zone = neg_create_zone(neg, dname, dname_len,
927 			rrset_class))) {
928 			lock_basic_unlock(&neg->lock);
929 			log_err("out of memory adding negative zone");
930 			return;
931 		}
932 	}
933 	val_neg_zone_take_inuse(zone);
934 
935 	/* insert the NSECs */
936 	for(i=rep->an_numrrsets; i< rep->an_numrrsets+rep->ns_numrrsets; i++){
937 		if(ntohs(rep->rrsets[i]->rk.type) != LDNS_RR_TYPE_NSEC)
938 			continue;
939 		if(!dname_subdomain_c(rep->rrsets[i]->rk.dname,
940 			zone->name)) continue;
941 		/* insert NSEC into this zone's tree */
942 		neg_insert_data(neg, zone, rep->rrsets[i]);
943 	}
944 	if(zone->tree.count == 0) {
945 		/* remove empty zone if inserts failed */
946 		neg_delete_zone(neg, zone);
947 	}
948 	lock_basic_unlock(&neg->lock);
949 }
950 
951 /**
952  * Lookup closest data record. For NSEC denial.
953  * @param zone: zone to look in
954  * @param qname: name to look for.
955  * @param len: length of name
956  * @param labs: labels in name
957  * @param data: data element, exact or smaller or NULL
958  * @return true if exact match.
959  */
neg_closest_data(struct val_neg_zone * zone,uint8_t * qname,size_t len,int labs,struct val_neg_data ** data)960 static int neg_closest_data(struct val_neg_zone* zone,
961 	uint8_t* qname, size_t len, int labs, struct val_neg_data** data)
962 {
963 	struct val_neg_data key;
964 	rbnode_type* r;
965 	key.node.key = &key;
966 	key.name = qname;
967 	key.len = len;
968 	key.labs = labs;
969 	if(rbtree_find_less_equal(&zone->tree, &key, &r)) {
970 		/* exact match */
971 		*data = (struct val_neg_data*)r;
972 		return 1;
973 	} else {
974 		/* smaller match */
975 		*data = (struct val_neg_data*)r;
976 		return 0;
977 	}
978 }
979 
val_neg_addreferral(struct val_neg_cache * neg,struct reply_info * rep,uint8_t * zone_name)980 void val_neg_addreferral(struct val_neg_cache* neg, struct reply_info* rep,
981 	uint8_t* zone_name)
982 {
983 	size_t i, need;
984 	uint8_t* signer;
985 	size_t signer_len;
986 	uint16_t dclass;
987 	struct val_neg_zone* zone;
988 	/* no SOA in this message, find RRSIG over NSEC's signer name.
989 	 * note the NSEC records are maybe not validated yet */
990 	signer = reply_nsec_signer(rep, &signer_len, &dclass);
991 	if(!signer)
992 		return;
993 	if(!dname_subdomain_c(signer, zone_name)) {
994 		/* the signer is not in the bailiwick, throw it out */
995 		return;
996 	}
997 
998 	log_nametypeclass(VERB_ALGO, "negcache insert referral ",
999 		signer, LDNS_RR_TYPE_NS, dclass);
1000 
1001 	/* ask for enough space to store all of it */
1002 	need = calc_data_need(rep) + calc_zone_need(signer, signer_len);
1003 	lock_basic_lock(&neg->lock);
1004 	neg_make_space(neg, need);
1005 
1006 	/* find or create the zone entry */
1007 	zone = neg_find_zone(neg, signer, signer_len, dclass);
1008 	if(!zone) {
1009 		if(!(zone = neg_create_zone(neg, signer, signer_len,
1010 			dclass))) {
1011 			lock_basic_unlock(&neg->lock);
1012 			log_err("out of memory adding negative zone");
1013 			return;
1014 		}
1015 	}
1016 	val_neg_zone_take_inuse(zone);
1017 
1018 	/* insert the NSECs */
1019 	for(i=rep->an_numrrsets; i< rep->an_numrrsets+rep->ns_numrrsets; i++){
1020 		if(ntohs(rep->rrsets[i]->rk.type) != LDNS_RR_TYPE_NSEC &&
1021 			ntohs(rep->rrsets[i]->rk.type) != LDNS_RR_TYPE_NSEC3)
1022 			continue;
1023 		if(!dname_subdomain_c(rep->rrsets[i]->rk.dname,
1024 			zone->name)) continue;
1025 		/* insert NSEC into this zone's tree */
1026 		neg_insert_data(neg, zone, rep->rrsets[i]);
1027 	}
1028 	if(zone->tree.count == 0) {
1029 		/* remove empty zone if inserts failed */
1030 		neg_delete_zone(neg, zone);
1031 	}
1032 	lock_basic_unlock(&neg->lock);
1033 }
1034 
1035 /**
1036  * Check that an NSEC3 rrset does not have a type set.
1037  * None of the nsec3s in a hash-collision are allowed to have the type.
1038  * (since we do not know which one is the nsec3 looked at, flags, ..., we
1039  * ignore the cached item and let it bypass negative caching).
1040  * @param k: the nsec3 rrset to check.
1041  * @param t: type to check
1042  * @return true if no RRs have the type.
1043  */
nsec3_no_type(struct ub_packed_rrset_key * k,uint16_t t)1044 static int nsec3_no_type(struct ub_packed_rrset_key* k, uint16_t t)
1045 {
1046 	int count = (int)((struct packed_rrset_data*)k->entry.data)->count;
1047 	int i;
1048 	for(i=0; i<count; i++)
1049 		if(nsec3_has_type(k, i, t))
1050 			return 0;
1051 	return 1;
1052 }
1053 
1054 /**
1055  * See if rrset exists in rrset cache.
1056  * If it does, the bit is checked, and if not expired, it is returned
1057  * allocated in region.
1058  * @param rrset_cache: rrset cache
1059  * @param qname: to lookup rrset name
1060  * @param qname_len: length of qname.
1061  * @param qtype: type of rrset to lookup, host order
1062  * @param qclass: class of rrset to lookup, host order
1063  * @param flags: flags for rrset to lookup
1064  * @param region: where to alloc result
1065  * @param checkbit: if true, a bit in the nsec typemap is checked for absence.
1066  * @param checktype: which bit to check
1067  * @param now: to check ttl against
1068  * @return rrset or NULL
1069  */
1070 static struct ub_packed_rrset_key*
grab_nsec(struct rrset_cache * rrset_cache,uint8_t * qname,size_t qname_len,uint16_t qtype,uint16_t qclass,uint32_t flags,struct regional * region,int checkbit,uint16_t checktype,time_t now)1071 grab_nsec(struct rrset_cache* rrset_cache, uint8_t* qname, size_t qname_len,
1072 	uint16_t qtype, uint16_t qclass, uint32_t flags,
1073 	struct regional* region, int checkbit, uint16_t checktype,
1074 	time_t now)
1075 {
1076 	struct ub_packed_rrset_key* r, *k = rrset_cache_lookup(rrset_cache,
1077 		qname, qname_len, qtype, qclass, flags, now, 0);
1078 	struct packed_rrset_data* d;
1079 	if(!k) return NULL;
1080 	d = k->entry.data;
1081 	/* only secure or unchecked records that have signatures. */
1082 	if( ! ( d->security == sec_status_secure ||
1083 		(d->security == sec_status_unchecked &&
1084 		d->rrsig_count > 0) ) ) {
1085 		lock_rw_unlock(&k->entry.lock);
1086 		return NULL;
1087 	}
1088 	/* check if checktype is absent */
1089 	if(checkbit && (
1090 		(qtype == LDNS_RR_TYPE_NSEC && nsec_has_type(k, checktype)) ||
1091 		(qtype == LDNS_RR_TYPE_NSEC3 && !nsec3_no_type(k, checktype))
1092 		)) {
1093 		lock_rw_unlock(&k->entry.lock);
1094 		return NULL;
1095 	}
1096 	/* looks OK! copy to region and return it */
1097 	r = packed_rrset_copy_region(k, region, now);
1098 	/* if it failed, we return the NULL */
1099 	lock_rw_unlock(&k->entry.lock);
1100 	return r;
1101 }
1102 
1103 /**
1104  * Get best NSEC record for qname. Might be matching, covering or totally
1105  * useless.
1106  * @param neg_cache: neg cache
1107  * @param qname: to lookup rrset name
1108  * @param qname_len: length of qname.
1109  * @param qclass: class of rrset to lookup, host order
1110  * @param rrset_cache: rrset cache
1111  * @param now: to check ttl against
1112  * @param region: where to alloc result
1113  * @return rrset or NULL
1114  */
1115 static struct ub_packed_rrset_key*
neg_find_nsec(struct val_neg_cache * neg_cache,uint8_t * qname,size_t qname_len,uint16_t qclass,struct rrset_cache * rrset_cache,time_t now,struct regional * region)1116 neg_find_nsec(struct val_neg_cache* neg_cache, uint8_t* qname, size_t qname_len,
1117 	uint16_t qclass, struct rrset_cache* rrset_cache, time_t now,
1118 	struct regional* region)
1119 {
1120 	int labs;
1121 	uint32_t flags;
1122 	struct val_neg_zone* zone;
1123 	struct val_neg_data* data;
1124 	struct ub_packed_rrset_key* nsec;
1125 
1126 	labs = dname_count_labels(qname);
1127 	lock_basic_lock(&neg_cache->lock);
1128 	zone = neg_closest_zone_parent(neg_cache, qname, qname_len, labs,
1129 		qclass);
1130 	while(zone && !zone->in_use)
1131 		zone = zone->parent;
1132 	if(!zone) {
1133 		lock_basic_unlock(&neg_cache->lock);
1134 		return NULL;
1135 	}
1136 
1137 	/* NSEC only for now */
1138 	if(zone->nsec3_hash) {
1139 		lock_basic_unlock(&neg_cache->lock);
1140 		return NULL;
1141 	}
1142 
1143 	/* ignore return value, don't care if it is an exact or smaller match */
1144 	(void)neg_closest_data(zone, qname, qname_len, labs, &data);
1145 	if(!data) {
1146 		lock_basic_unlock(&neg_cache->lock);
1147 		return NULL;
1148 	}
1149 
1150 	/* ENT nodes are not in use, try the previous node. If the previous node
1151 	 * is not in use, we don't have an useful NSEC and give up. */
1152 	if(!data->in_use) {
1153 		data = (struct val_neg_data*)rbtree_previous((rbnode_type*)data);
1154 		if((rbnode_type*)data == RBTREE_NULL || !data->in_use) {
1155 			lock_basic_unlock(&neg_cache->lock);
1156 			return NULL;
1157 		}
1158 	}
1159 
1160 	flags = 0;
1161 	if(query_dname_compare(data->name, zone->name) == 0)
1162 		flags = PACKED_RRSET_NSEC_AT_APEX;
1163 
1164 	nsec = grab_nsec(rrset_cache, data->name, data->len, LDNS_RR_TYPE_NSEC,
1165 		zone->dclass, flags, region, 0, 0, now);
1166 	lock_basic_unlock(&neg_cache->lock);
1167 	return nsec;
1168 }
1169 
1170 /** find nsec3 closest encloser in neg cache */
1171 static struct val_neg_data*
neg_find_nsec3_ce(struct val_neg_zone * zone,uint8_t * qname,size_t qname_len,int qlabs,sldns_buffer * buf,uint8_t * hashnc,size_t * nclen)1172 neg_find_nsec3_ce(struct val_neg_zone* zone, uint8_t* qname, size_t qname_len,
1173 		int qlabs, sldns_buffer* buf, uint8_t* hashnc, size_t* nclen)
1174 {
1175 	struct val_neg_data* data;
1176 	uint8_t hashce[NSEC3_SHA_LEN];
1177 	uint8_t b32[257];
1178 	size_t celen, b32len;
1179 	int hashmax = MAX_NSEC3_CALCULATIONS;
1180 	if(qlabs > hashmax) {
1181 		/* strip leading labels so the walk costs at most
1182 		 * MAX_NSEC3_CALCULATIONS hashes, mirroring val_nsec3.c */
1183 		while(qlabs > hashmax) {
1184 			dname_remove_label(&qname, &qname_len);
1185 			qlabs--;
1186 		}
1187 	}
1188 
1189 	*nclen = 0;
1190 	while(qlabs > 0) {
1191 		/* hash */
1192 		if(!(celen=nsec3_get_hashed(buf, qname, qname_len,
1193 			zone->nsec3_hash, zone->nsec3_iter, zone->nsec3_salt,
1194 			zone->nsec3_saltlen, hashce, sizeof(hashce))))
1195 			return NULL;
1196 		if(!(b32len=nsec3_hash_to_b32(hashce, celen, zone->name,
1197 			zone->len, b32, sizeof(b32))))
1198 			return NULL;
1199 
1200 		/* lookup (exact match only) */
1201 		data = neg_find_data(zone, b32, b32len, zone->labs+1);
1202 		if(data && data->in_use) {
1203 			/* found ce match! */
1204 			return data;
1205 		}
1206 
1207 		*nclen = celen;
1208 		memmove(hashnc, hashce, celen);
1209 		dname_remove_label(&qname, &qname_len);
1210 		qlabs --;
1211 	}
1212 	return NULL;
1213 }
1214 
1215 /** check nsec3 parameters on nsec3 rrset with current zone values */
1216 static int
neg_params_ok(struct val_neg_zone * zone,struct ub_packed_rrset_key * rrset)1217 neg_params_ok(struct val_neg_zone* zone, struct ub_packed_rrset_key* rrset)
1218 {
1219 	int h;
1220 	uint8_t* s;
1221 	size_t slen, it;
1222 	if(!nsec3_get_params(rrset, 0, &h, &it, &s, &slen))
1223 		return 0;
1224 	return (h == zone->nsec3_hash && it == zone->nsec3_iter &&
1225 		slen == zone->nsec3_saltlen &&
1226 		(slen != 0 && zone->nsec3_salt && s
1227 		  && memcmp(zone->nsec3_salt, s, slen) == 0));
1228 }
1229 
1230 /** get next closer for nsec3 proof */
1231 static struct ub_packed_rrset_key*
neg_nsec3_getnc(struct val_neg_zone * zone,uint8_t * hashnc,size_t nclen,struct rrset_cache * rrset_cache,struct regional * region,time_t now,uint8_t * b32,size_t maxb32)1232 neg_nsec3_getnc(struct val_neg_zone* zone, uint8_t* hashnc, size_t nclen,
1233 	struct rrset_cache* rrset_cache, struct regional* region,
1234 	time_t now, uint8_t* b32, size_t maxb32)
1235 {
1236 	struct ub_packed_rrset_key* nc_rrset;
1237 	struct val_neg_data* data;
1238 	size_t b32len;
1239 
1240 	if(!(b32len=nsec3_hash_to_b32(hashnc, nclen, zone->name,
1241 		zone->len, b32, maxb32)))
1242 		return NULL;
1243 	(void)neg_closest_data(zone, b32, b32len, zone->labs+1, &data);
1244 	if(!data && zone->tree.count != 0) {
1245 		/* could be before the first entry ; return the last
1246 		 * entry (possibly the rollover nsec3 at end) */
1247 		data = (struct val_neg_data*)rbtree_last(&zone->tree);
1248 	}
1249 	while(data && !data->in_use)
1250 		data = data->parent;
1251 	if(!data)
1252 		return NULL;
1253 	/* got a data element in tree, grab it */
1254 	nc_rrset = grab_nsec(rrset_cache, data->name, data->len,
1255 		LDNS_RR_TYPE_NSEC3, zone->dclass, 0, region, 0, 0, now);
1256 	if(!nc_rrset)
1257 		return NULL;
1258 	if(!neg_params_ok(zone, nc_rrset))
1259 		return NULL;
1260 	return nc_rrset;
1261 }
1262 
1263 /** neg cache nsec3 proof procedure*/
1264 static struct dns_msg*
neg_nsec3_proof_ds(struct val_neg_zone * zone,uint8_t * qname,size_t qname_len,int qlabs,sldns_buffer * buf,struct rrset_cache * rrset_cache,struct regional * region,time_t now,uint8_t * topname)1265 neg_nsec3_proof_ds(struct val_neg_zone* zone, uint8_t* qname, size_t qname_len,
1266 		int qlabs, sldns_buffer* buf, struct rrset_cache* rrset_cache,
1267 		struct regional* region, time_t now, uint8_t* topname)
1268 {
1269 	struct dns_msg* msg;
1270 	struct val_neg_data* data;
1271 	uint8_t hashnc[NSEC3_SHA_LEN];
1272 	size_t nclen;
1273 	struct ub_packed_rrset_key* ce_rrset, *nc_rrset;
1274 	struct nsec3_cached_hash c;
1275 	uint8_t nc_b32[257];
1276 
1277 	/* for NSEC3 ; determine the closest encloser for which we
1278 	 * can find an exact match. Remember the hashed lower name,
1279 	 * since that is the one we need a closest match for.
1280 	 * If we find a match straight away, then it becomes NODATA.
1281 	 * Otherwise, NXDOMAIN or if OPTOUT, an insecure delegation.
1282 	 * Also check that parameters are the same on closest encloser
1283 	 * and on closest match.
1284 	 */
1285 	if(!zone->nsec3_hash)
1286 		return NULL; /* not nsec3 zone */
1287 
1288 	if(!topname && qlabs > zone->labs + 1)
1289 		return NULL; /* iterator caller; opt-out proof would be discarded
1290 			     * at the !topname check below anyway.
1291 			     * The qlabs check allows the exact-match for
1292 			     * the one-label-below-zone case. */
1293 
1294 	if(!(data=neg_find_nsec3_ce(zone, qname, qname_len, qlabs, buf,
1295 		hashnc, &nclen))) {
1296 		return NULL;
1297 	}
1298 
1299 	/* grab the ce rrset */
1300 	ce_rrset = grab_nsec(rrset_cache, data->name, data->len,
1301 		LDNS_RR_TYPE_NSEC3, zone->dclass, 0, region, 1,
1302 		LDNS_RR_TYPE_DS, now);
1303 	if(!ce_rrset)
1304 		return NULL;
1305 	if(!neg_params_ok(zone, ce_rrset))
1306 		return NULL;
1307 
1308 	if(nclen == 0) {
1309 		/* exact match, just check the type bits */
1310 		/* need: -SOA, -DS, +NS */
1311 		if(nsec3_has_type(ce_rrset, 0, LDNS_RR_TYPE_SOA) ||
1312 			nsec3_has_type(ce_rrset, 0, LDNS_RR_TYPE_DS) ||
1313 			!nsec3_has_type(ce_rrset, 0, LDNS_RR_TYPE_NS))
1314 			return NULL;
1315 		if(!(msg = dns_msg_create(qname, qname_len,
1316 			LDNS_RR_TYPE_DS, zone->dclass, region, 1)))
1317 			return NULL;
1318 		/* The cache response means recursion is available. */
1319 		msg->rep->flags |= BIT_RA;
1320 		/* TTL reduced in grab_nsec */
1321 		if(!dns_msg_authadd(msg, region, ce_rrset, 0))
1322 			return NULL;
1323 		return msg;
1324 	}
1325 
1326 	/* optout is not allowed without knowing the trust-anchor in use,
1327 	 * otherwise the optout could spoof away that anchor */
1328 	if(!topname)
1329 		return NULL;
1330 
1331 	/* if there is no exact match, it must be in an optout span
1332 	 * (an existing DS implies an NSEC3 must exist) */
1333 	nc_rrset = neg_nsec3_getnc(zone, hashnc, nclen, rrset_cache,
1334 		region, now, nc_b32, sizeof(nc_b32));
1335 	if(!nc_rrset)
1336 		return NULL;
1337 	if(!neg_params_ok(zone, nc_rrset))
1338 		return NULL;
1339 	if(!nsec3_has_optout(nc_rrset, 0))
1340 		return NULL;
1341 	c.hash = hashnc;
1342 	c.hash_len = nclen;
1343 	c.b32 = nc_b32+1;
1344 	c.b32_len = (size_t)nc_b32[0];
1345 	if(nsec3_covers(zone->name, &c, nc_rrset, 0, buf)) {
1346 		/* nc_rrset covers the next closer name.
1347 		 * ce_rrset equals a closer encloser.
1348 		 * nc_rrset is optout.
1349 		 * No need to check wildcard for type DS */
1350 		/* capacity=3: ce + nc + soa(if needed) */
1351 		if(!(msg = dns_msg_create(qname, qname_len,
1352 			LDNS_RR_TYPE_DS, zone->dclass, region, 3)))
1353 			return NULL;
1354 		/* The cache response means recursion is available. */
1355 		msg->rep->flags |= BIT_RA;
1356 		/* now=0 because TTL was reduced in grab_nsec */
1357 		if(!dns_msg_authadd(msg, region, ce_rrset, 0))
1358 			return NULL;
1359 		if(!dns_msg_authadd(msg, region, nc_rrset, 0))
1360 			return NULL;
1361 		return msg;
1362 	}
1363 	return NULL;
1364 }
1365 
1366 /**
1367  * Add SOA record for external responses.
1368  * @param rrset_cache: to look into.
1369  * @param now: current time.
1370  * @param region: where to perform the allocation
1371  * @param msg: current msg with NSEC.
1372  * @param zone: val_neg_zone if we have one.
1373  * @return false on lookup or alloc failure.
1374  */
add_soa(struct rrset_cache * rrset_cache,time_t now,struct regional * region,struct dns_msg * msg,struct val_neg_zone * zone)1375 static int add_soa(struct rrset_cache* rrset_cache, time_t now,
1376 	struct regional* region, struct dns_msg* msg, struct val_neg_zone* zone)
1377 {
1378 	struct ub_packed_rrset_key* soa;
1379 	uint8_t* nm;
1380 	size_t nmlen;
1381 	uint16_t dclass;
1382 	if(zone) {
1383 		nm = zone->name;
1384 		nmlen = zone->len;
1385 		dclass = zone->dclass;
1386 	} else {
1387 		/* Assumes the signer is the zone SOA to add */
1388 		nm = reply_nsec_signer(msg->rep, &nmlen, &dclass);
1389 		if(!nm)
1390 			return 0;
1391 	}
1392 	soa = rrset_cache_lookup(rrset_cache, nm, nmlen, LDNS_RR_TYPE_SOA,
1393 		dclass, PACKED_RRSET_SOA_NEG, now, 0);
1394 	if(!soa)
1395 		return 0;
1396 	if(!dns_msg_authadd(msg, region, soa, now)) {
1397 		lock_rw_unlock(&soa->entry.lock);
1398 		return 0;
1399 	}
1400 	lock_rw_unlock(&soa->entry.lock);
1401 	return 1;
1402 }
1403 
1404 struct dns_msg*
val_neg_getmsg(struct val_neg_cache * neg,struct query_info * qinfo,struct regional * region,struct rrset_cache * rrset_cache,sldns_buffer * buf,time_t now,int addsoa,uint8_t * topname,struct config_file * cfg)1405 val_neg_getmsg(struct val_neg_cache* neg, struct query_info* qinfo,
1406 	struct regional* region, struct rrset_cache* rrset_cache,
1407 	sldns_buffer* buf, time_t now, int addsoa, uint8_t* topname,
1408 	struct config_file* cfg)
1409 {
1410 	struct dns_msg* msg;
1411 	struct ub_packed_rrset_key* nsec; /* qname matching/covering nsec */
1412 	struct ub_packed_rrset_key* wcrr; /* wildcard record or nsec */
1413 	uint8_t* nodata_wc = NULL;
1414 	uint8_t* ce = NULL;
1415 	size_t ce_len;
1416 	uint8_t wc_ce[LDNS_MAX_DOMAINLEN+3];
1417 	struct query_info wc_qinfo;
1418 	struct ub_packed_rrset_key* cache_wc;
1419 	struct packed_rrset_data* wcrr_data;
1420 	int rcode = LDNS_RCODE_NOERROR;
1421 	uint8_t* zname;
1422 	size_t zname_len;
1423 	int zname_labs;
1424 	struct val_neg_zone* zone;
1425 
1426 	/* only for DS queries when aggressive use of NSEC is disabled */
1427 	if(qinfo->qtype != LDNS_RR_TYPE_DS && !cfg->aggressive_nsec)
1428 		return NULL;
1429 	log_assert(!topname || dname_subdomain_c(qinfo->qname, topname));
1430 
1431 	/* Get best available NSEC for qname */
1432 	nsec = neg_find_nsec(neg, qinfo->qname, qinfo->qname_len, qinfo->qclass,
1433 		rrset_cache, now, region);
1434 
1435 	/* Matching NSEC, use to generate No Data answer. Not creating answers
1436 	 * yet for No Data proven using wildcard. */
1437 	if(nsec && nsec_proves_nodata(nsec, qinfo, &nodata_wc) && !nodata_wc) {
1438 		/* do not create nodata answers for qtype ANY, it is a query
1439 		 * type, not an rrtype to disprove. Nameerrors are useful for
1440 		 * qtype ANY, in the else branch. */
1441 		if(qinfo->qtype == LDNS_RR_TYPE_ANY)
1442 			return NULL;
1443 		if(!(msg = dns_msg_create(qinfo->qname, qinfo->qname_len,
1444 			qinfo->qtype, qinfo->qclass, region, 2)))
1445 			return NULL;
1446 		/* The cache response means recursion is available. */
1447 		msg->rep->flags |= BIT_RA;
1448 		if(!dns_msg_authadd(msg, region, nsec, 0))
1449 			return NULL;
1450 		if(addsoa && !add_soa(rrset_cache, now, region, msg, NULL))
1451 			return NULL;
1452 
1453 		lock_basic_lock(&neg->lock);
1454 		neg->num_neg_cache_noerror++;
1455 		lock_basic_unlock(&neg->lock);
1456 		return msg;
1457 	} else if(nsec && val_nsec_proves_name_error(nsec, qinfo->qname)) {
1458 		if(!(msg = dns_msg_create(qinfo->qname, qinfo->qname_len,
1459 			qinfo->qtype, qinfo->qclass, region, 3)))
1460 			return NULL;
1461 		/* The cache response means recursion is available. */
1462 		msg->rep->flags |= BIT_RA;
1463 		if(!(ce = nsec_closest_encloser(qinfo->qname, nsec)))
1464 			return NULL;
1465 		dname_count_size_labels(ce, &ce_len);
1466 
1467 		/* No extra extra NSEC required if both nameerror qname and
1468 		 * nodata *.ce. are proven already. */
1469 		if(!nodata_wc || query_dname_compare(nodata_wc, ce) != 0) {
1470 			/* Qname proven non existing, get wildcard record for
1471 			 * QTYPE or NSEC covering or matching wildcard. */
1472 
1473 			/* Num labels in ce is always smaller than in qname,
1474 			 * therefore adding the wildcard label cannot overflow
1475 			 * buffer. */
1476 			wc_ce[0] = 1;
1477 			wc_ce[1] = (uint8_t)'*';
1478 			memmove(wc_ce+2, ce, ce_len);
1479 			wc_qinfo.qname = wc_ce;
1480 			wc_qinfo.qname_len = ce_len + 2;
1481 			wc_qinfo.qtype = qinfo->qtype;
1482 
1483 
1484 			if((cache_wc = rrset_cache_lookup(rrset_cache, wc_qinfo.qname,
1485 				wc_qinfo.qname_len, wc_qinfo.qtype,
1486 				qinfo->qclass, 0/*flags*/, now, 0/*read only*/))) {
1487 				/* Synthesize wildcard answer */
1488 				wcrr_data = (struct packed_rrset_data*)cache_wc->entry.data;
1489 				if(!(wcrr_data->security == sec_status_secure ||
1490 					(wcrr_data->security == sec_status_unchecked &&
1491 					wcrr_data->rrsig_count > 0))) {
1492 					lock_rw_unlock(&cache_wc->entry.lock);
1493 					return NULL;
1494 				}
1495 				if(!(wcrr = packed_rrset_copy_region(cache_wc,
1496 					region, now))) {
1497 					lock_rw_unlock(&cache_wc->entry.lock);
1498 					return NULL;
1499 				};
1500 				lock_rw_unlock(&cache_wc->entry.lock);
1501 				wcrr->rk.dname = qinfo->qname;
1502 				wcrr->rk.dname_len = qinfo->qname_len;
1503 				if(!dns_msg_ansadd(msg, region, wcrr, 0))
1504 					return NULL;
1505 				/* No SOA needed for wildcard synthesised
1506 				 * answer. */
1507 				addsoa = 0;
1508 			} else {
1509 				/* Get wildcard NSEC for possible non existence
1510 				 * proof */
1511 				if(!(wcrr = neg_find_nsec(neg, wc_qinfo.qname,
1512 					wc_qinfo.qname_len, qinfo->qclass,
1513 					rrset_cache, now, region)))
1514 					return NULL;
1515 
1516 				nodata_wc = NULL;
1517 				if(val_nsec_proves_name_error(wcrr, wc_ce))
1518 					rcode = LDNS_RCODE_NXDOMAIN;
1519 				else if(!nsec_proves_nodata(wcrr, &wc_qinfo,
1520 					&nodata_wc) || nodata_wc)
1521 					/* &nodata_wc shouldn't be set, wc_qinfo
1522 					 * already contains wildcard domain. */
1523 					/* NSEC doesn't prove anything for
1524 					 * wildcard. */
1525 					return NULL;
1526 				if(query_dname_compare(wcrr->rk.dname,
1527 					nsec->rk.dname) != 0)
1528 					if(!dns_msg_authadd(msg, region, wcrr, 0))
1529 						return NULL;
1530 			}
1531 		}
1532 
1533 		if(!dns_msg_authadd(msg, region, nsec, 0))
1534 			return NULL;
1535 		if(addsoa && !add_soa(rrset_cache, now, region, msg, NULL))
1536 			return NULL;
1537 
1538 		/* Increment statistic counters */
1539 		lock_basic_lock(&neg->lock);
1540 		if(rcode == LDNS_RCODE_NOERROR)
1541 			neg->num_neg_cache_noerror++;
1542 		else if(rcode == LDNS_RCODE_NXDOMAIN)
1543 			neg->num_neg_cache_nxdomain++;
1544 		lock_basic_unlock(&neg->lock);
1545 
1546 		FLAGS_SET_RCODE(msg->rep->flags, rcode);
1547 		return msg;
1548 	}
1549 
1550 	/* No aggressive use of NSEC3 for now, only proceed for DS types. */
1551 	if(qinfo->qtype != LDNS_RR_TYPE_DS){
1552 		return NULL;
1553 	}
1554 	/* check NSEC3 neg cache for type DS */
1555 	/* need to look one zone higher for DS type */
1556 	zname = qinfo->qname;
1557 	zname_len = qinfo->qname_len;
1558 	dname_remove_label(&zname, &zname_len);
1559 	zname_labs = dname_count_labels(zname);
1560 
1561 	/* lookup closest zone */
1562 	lock_basic_lock(&neg->lock);
1563 	zone = neg_closest_zone_parent(neg, zname, zname_len, zname_labs,
1564 		qinfo->qclass);
1565 	while(zone && !zone->in_use)
1566 		zone = zone->parent;
1567 	/* check that the zone is not too high up so that we do not pick data
1568 	 * out of a zone that is above the last-seen key (or trust-anchor). */
1569 	if(zone && topname) {
1570 		if(!dname_subdomain_c(zone->name, topname))
1571 			zone = NULL;
1572 	}
1573 	if(!zone) {
1574 		lock_basic_unlock(&neg->lock);
1575 		return NULL;
1576 	}
1577 
1578 	msg = neg_nsec3_proof_ds(zone, qinfo->qname, qinfo->qname_len,
1579 		zname_labs+1, buf, rrset_cache, region, now, topname);
1580 	if(msg && addsoa && !add_soa(rrset_cache, now, region, msg, zone)) {
1581 		lock_basic_unlock(&neg->lock);
1582 		return NULL;
1583 	}
1584 	lock_basic_unlock(&neg->lock);
1585 	return msg;
1586 }
1587 
1588 void
val_neg_adjust_size(struct val_neg_cache * neg,size_t max)1589 val_neg_adjust_size(struct val_neg_cache* neg, size_t max)
1590 {
1591 	lock_basic_lock(&neg->lock);
1592 	neg->max = max;
1593 	neg_make_space(neg, 0);
1594 	lock_basic_unlock(&neg->lock);
1595 }
1596