]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/blob - contrib/unbound/services/authzone.c
unbound: Vendor import 1.18.0
[FreeBSD/FreeBSD.git] / contrib / unbound / services / authzone.c
1 /*
2  * services/authzone.c - authoritative zone that is locally hosted.
3  *
4  * Copyright (c) 2017, 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 the functions for an authority zone.  This zone
40  * is queried by the iterator, just like a stub or forward zone, but then
41  * the data is locally held.
42  */
43
44 #include "config.h"
45 #include "services/authzone.h"
46 #include "util/data/dname.h"
47 #include "util/data/msgparse.h"
48 #include "util/data/msgreply.h"
49 #include "util/data/msgencode.h"
50 #include "util/data/packed_rrset.h"
51 #include "util/regional.h"
52 #include "util/net_help.h"
53 #include "util/netevent.h"
54 #include "util/config_file.h"
55 #include "util/log.h"
56 #include "util/module.h"
57 #include "util/random.h"
58 #include "services/cache/dns.h"
59 #include "services/outside_network.h"
60 #include "services/listen_dnsport.h"
61 #include "services/mesh.h"
62 #include "sldns/rrdef.h"
63 #include "sldns/pkthdr.h"
64 #include "sldns/sbuffer.h"
65 #include "sldns/str2wire.h"
66 #include "sldns/wire2str.h"
67 #include "sldns/parseutil.h"
68 #include "sldns/keyraw.h"
69 #include "validator/val_nsec3.h"
70 #include "validator/val_nsec.h"
71 #include "validator/val_secalgo.h"
72 #include "validator/val_sigcrypt.h"
73 #include "validator/val_anchor.h"
74 #include "validator/val_utils.h"
75 #include <ctype.h>
76
77 /** bytes to use for NSEC3 hash buffer. 20 for sha1 */
78 #define N3HASHBUFLEN 32
79 /** max number of CNAMEs we are willing to follow (in one answer) */
80 #define MAX_CNAME_CHAIN 8
81 /** timeout for probe packets for SOA */
82 #define AUTH_PROBE_TIMEOUT 100 /* msec */
83 /** when to stop with SOA probes (when exponential timeouts exceed this) */
84 #define AUTH_PROBE_TIMEOUT_STOP 1000 /* msec */
85 /* auth transfer timeout for TCP connections, in msec */
86 #define AUTH_TRANSFER_TIMEOUT 10000 /* msec */
87 /* auth transfer max backoff for failed transfers and probes */
88 #define AUTH_TRANSFER_MAX_BACKOFF 86400 /* sec */
89 /* auth http port number */
90 #define AUTH_HTTP_PORT 80
91 /* auth https port number */
92 #define AUTH_HTTPS_PORT 443
93 /* max depth for nested $INCLUDEs */
94 #define MAX_INCLUDE_DEPTH 10
95 /** number of timeouts before we fallback from IXFR to AXFR,
96  * because some versions of servers (eg. dnsmasq) drop IXFR packets. */
97 #define NUM_TIMEOUTS_FALLBACK_IXFR 3
98
99 /** pick up nextprobe task to start waiting to perform transfer actions */
100 static void xfr_set_timeout(struct auth_xfer* xfr, struct module_env* env,
101         int failure, int lookup_only);
102 /** move to sending the probe packets, next if fails. task_probe */
103 static void xfr_probe_send_or_end(struct auth_xfer* xfr,
104         struct module_env* env);
105 /** pick up probe task with specified(or NULL) destination first,
106  * or transfer task if nothing to probe, or false if already in progress */
107 static int xfr_start_probe(struct auth_xfer* xfr, struct module_env* env,
108         struct auth_master* spec);
109 /** delete xfer structure (not its tree entry) */
110 static void auth_xfer_delete(struct auth_xfer* xfr);
111
112 /** create new dns_msg */
113 static struct dns_msg*
114 msg_create(struct regional* region, struct query_info* qinfo)
115 {
116         struct dns_msg* msg = (struct dns_msg*)regional_alloc(region,
117                 sizeof(struct dns_msg));
118         if(!msg)
119                 return NULL;
120         msg->qinfo.qname = regional_alloc_init(region, qinfo->qname,
121                 qinfo->qname_len);
122         if(!msg->qinfo.qname)
123                 return NULL;
124         msg->qinfo.qname_len = qinfo->qname_len;
125         msg->qinfo.qtype = qinfo->qtype;
126         msg->qinfo.qclass = qinfo->qclass;
127         msg->qinfo.local_alias = NULL;
128         /* non-packed reply_info, because it needs to grow the array */
129         msg->rep = (struct reply_info*)regional_alloc_zero(region,
130                 sizeof(struct reply_info)-sizeof(struct rrset_ref));
131         if(!msg->rep)
132                 return NULL;
133         msg->rep->flags = (uint16_t)(BIT_QR | BIT_AA);
134         msg->rep->authoritative = 1;
135         msg->rep->reason_bogus = LDNS_EDE_NONE;
136         msg->rep->qdcount = 1;
137         /* rrsets is NULL, no rrsets yet */
138         return msg;
139 }
140
141 /** grow rrset array by one in msg */
142 static int
143 msg_grow_array(struct regional* region, struct dns_msg* msg)
144 {
145         if(msg->rep->rrsets == NULL) {
146                 msg->rep->rrsets = regional_alloc_zero(region,
147                         sizeof(struct ub_packed_rrset_key*)*(msg->rep->rrset_count+1));
148                 if(!msg->rep->rrsets)
149                         return 0;
150         } else {
151                 struct ub_packed_rrset_key** rrsets_old = msg->rep->rrsets;
152                 msg->rep->rrsets = regional_alloc_zero(region,
153                         sizeof(struct ub_packed_rrset_key*)*(msg->rep->rrset_count+1));
154                 if(!msg->rep->rrsets)
155                         return 0;
156                 memmove(msg->rep->rrsets, rrsets_old,
157                         sizeof(struct ub_packed_rrset_key*)*msg->rep->rrset_count);
158         }
159         return 1;
160 }
161
162 /** get ttl of rrset */
163 static time_t
164 get_rrset_ttl(struct ub_packed_rrset_key* k)
165 {
166         struct packed_rrset_data* d = (struct packed_rrset_data*)
167                 k->entry.data;
168         return d->ttl;
169 }
170
171 /** Copy rrset into region from domain-datanode and packet rrset */
172 static struct ub_packed_rrset_key*
173 auth_packed_rrset_copy_region(struct auth_zone* z, struct auth_data* node,
174         struct auth_rrset* rrset, struct regional* region, time_t adjust)
175 {
176         struct ub_packed_rrset_key key;
177         memset(&key, 0, sizeof(key));
178         key.entry.key = &key;
179         key.entry.data = rrset->data;
180         key.rk.dname = node->name;
181         key.rk.dname_len = node->namelen;
182         key.rk.type = htons(rrset->type);
183         key.rk.rrset_class = htons(z->dclass);
184         key.entry.hash = rrset_key_hash(&key.rk);
185         return packed_rrset_copy_region(&key, region, adjust);
186 }
187
188 /** fix up msg->rep TTL and prefetch ttl */
189 static void
190 msg_ttl(struct dns_msg* msg)
191 {
192         if(msg->rep->rrset_count == 0) return;
193         if(msg->rep->rrset_count == 1) {
194                 msg->rep->ttl = get_rrset_ttl(msg->rep->rrsets[0]);
195                 msg->rep->prefetch_ttl = PREFETCH_TTL_CALC(msg->rep->ttl);
196                 msg->rep->serve_expired_ttl = msg->rep->ttl + SERVE_EXPIRED_TTL;
197         } else if(get_rrset_ttl(msg->rep->rrsets[msg->rep->rrset_count-1]) <
198                 msg->rep->ttl) {
199                 msg->rep->ttl = get_rrset_ttl(msg->rep->rrsets[
200                         msg->rep->rrset_count-1]);
201                 msg->rep->prefetch_ttl = PREFETCH_TTL_CALC(msg->rep->ttl);
202                 msg->rep->serve_expired_ttl = msg->rep->ttl + SERVE_EXPIRED_TTL;
203         }
204 }
205
206 /** see if rrset is a duplicate in the answer message */
207 static int
208 msg_rrset_duplicate(struct dns_msg* msg, uint8_t* nm, size_t nmlen,
209         uint16_t type, uint16_t dclass)
210 {
211         size_t i;
212         for(i=0; i<msg->rep->rrset_count; i++) {
213                 struct ub_packed_rrset_key* k = msg->rep->rrsets[i];
214                 if(ntohs(k->rk.type) == type && k->rk.dname_len == nmlen &&
215                         ntohs(k->rk.rrset_class) == dclass &&
216                         query_dname_compare(k->rk.dname, nm) == 0)
217                         return 1;
218         }
219         return 0;
220 }
221
222 /** add rrset to answer section (no auth, add rrsets yet) */
223 static int
224 msg_add_rrset_an(struct auth_zone* z, struct regional* region,
225         struct dns_msg* msg, struct auth_data* node, struct auth_rrset* rrset)
226 {
227         log_assert(msg->rep->ns_numrrsets == 0);
228         log_assert(msg->rep->ar_numrrsets == 0);
229         if(!rrset || !node)
230                 return 1;
231         if(msg_rrset_duplicate(msg, node->name, node->namelen, rrset->type,
232                 z->dclass))
233                 return 1;
234         /* grow array */
235         if(!msg_grow_array(region, msg))
236                 return 0;
237         /* copy it */
238         if(!(msg->rep->rrsets[msg->rep->rrset_count] =
239                 auth_packed_rrset_copy_region(z, node, rrset, region, 0)))
240                 return 0;
241         msg->rep->rrset_count++;
242         msg->rep->an_numrrsets++;
243         msg_ttl(msg);
244         return 1;
245 }
246
247 /** add rrset to authority section (no additional section rrsets yet) */
248 static int
249 msg_add_rrset_ns(struct auth_zone* z, struct regional* region,
250         struct dns_msg* msg, struct auth_data* node, struct auth_rrset* rrset)
251 {
252         log_assert(msg->rep->ar_numrrsets == 0);
253         if(!rrset || !node)
254                 return 1;
255         if(msg_rrset_duplicate(msg, node->name, node->namelen, rrset->type,
256                 z->dclass))
257                 return 1;
258         /* grow array */
259         if(!msg_grow_array(region, msg))
260                 return 0;
261         /* copy it */
262         if(!(msg->rep->rrsets[msg->rep->rrset_count] =
263                 auth_packed_rrset_copy_region(z, node, rrset, region, 0)))
264                 return 0;
265         msg->rep->rrset_count++;
266         msg->rep->ns_numrrsets++;
267         msg_ttl(msg);
268         return 1;
269 }
270
271 /** add rrset to additional section */
272 static int
273 msg_add_rrset_ar(struct auth_zone* z, struct regional* region,
274         struct dns_msg* msg, struct auth_data* node, struct auth_rrset* rrset)
275 {
276         if(!rrset || !node)
277                 return 1;
278         if(msg_rrset_duplicate(msg, node->name, node->namelen, rrset->type,
279                 z->dclass))
280                 return 1;
281         /* grow array */
282         if(!msg_grow_array(region, msg))
283                 return 0;
284         /* copy it */
285         if(!(msg->rep->rrsets[msg->rep->rrset_count] =
286                 auth_packed_rrset_copy_region(z, node, rrset, region, 0)))
287                 return 0;
288         msg->rep->rrset_count++;
289         msg->rep->ar_numrrsets++;
290         msg_ttl(msg);
291         return 1;
292 }
293
294 struct auth_zones* auth_zones_create(void)
295 {
296         struct auth_zones* az = (struct auth_zones*)calloc(1, sizeof(*az));
297         if(!az) {
298                 log_err("out of memory");
299                 return NULL;
300         }
301         rbtree_init(&az->ztree, &auth_zone_cmp);
302         rbtree_init(&az->xtree, &auth_xfer_cmp);
303         lock_rw_init(&az->lock);
304         lock_protect(&az->lock, &az->ztree, sizeof(az->ztree));
305         lock_protect(&az->lock, &az->xtree, sizeof(az->xtree));
306         /* also lock protects the rbnode's in struct auth_zone, auth_xfer */
307         lock_rw_init(&az->rpz_lock);
308         lock_protect(&az->rpz_lock, &az->rpz_first, sizeof(az->rpz_first));
309         return az;
310 }
311
312 int auth_zone_cmp(const void* z1, const void* z2)
313 {
314         /* first sort on class, so that hierarchy can be maintained within
315          * a class */
316         struct auth_zone* a = (struct auth_zone*)z1;
317         struct auth_zone* b = (struct auth_zone*)z2;
318         int m;
319         if(a->dclass != b->dclass) {
320                 if(a->dclass < b->dclass)
321                         return -1;
322                 return 1;
323         }
324         /* sorted such that higher zones sort before lower zones (their
325          * contents) */
326         return dname_lab_cmp(a->name, a->namelabs, b->name, b->namelabs, &m);
327 }
328
329 int auth_data_cmp(const void* z1, const void* z2)
330 {
331         struct auth_data* a = (struct auth_data*)z1;
332         struct auth_data* b = (struct auth_data*)z2;
333         int m;
334         /* canonical sort, because DNSSEC needs that */
335         return dname_canon_lab_cmp(a->name, a->namelabs, b->name,
336                 b->namelabs, &m);
337 }
338
339 int auth_xfer_cmp(const void* z1, const void* z2)
340 {
341         /* first sort on class, so that hierarchy can be maintained within
342          * a class */
343         struct auth_xfer* a = (struct auth_xfer*)z1;
344         struct auth_xfer* b = (struct auth_xfer*)z2;
345         int m;
346         if(a->dclass != b->dclass) {
347                 if(a->dclass < b->dclass)
348                         return -1;
349                 return 1;
350         }
351         /* sorted such that higher zones sort before lower zones (their
352          * contents) */
353         return dname_lab_cmp(a->name, a->namelabs, b->name, b->namelabs, &m);
354 }
355
356 /** delete auth rrset node */
357 static void
358 auth_rrset_delete(struct auth_rrset* rrset)
359 {
360         if(!rrset) return;
361         free(rrset->data);
362         free(rrset);
363 }
364
365 /** delete auth data domain node */
366 static void
367 auth_data_delete(struct auth_data* n)
368 {
369         struct auth_rrset* p, *np;
370         if(!n) return;
371         p = n->rrsets;
372         while(p) {
373                 np = p->next;
374                 auth_rrset_delete(p);
375                 p = np;
376         }
377         free(n->name);
378         free(n);
379 }
380
381 /** helper traverse to delete zones */
382 static void
383 auth_data_del(rbnode_type* n, void* ATTR_UNUSED(arg))
384 {
385         struct auth_data* z = (struct auth_data*)n->key;
386         auth_data_delete(z);
387 }
388
389 /** delete an auth zone structure (tree remove must be done elsewhere) */
390 static void
391 auth_zone_delete(struct auth_zone* z, struct auth_zones* az)
392 {
393         if(!z) return;
394         lock_rw_destroy(&z->lock);
395         traverse_postorder(&z->data, auth_data_del, NULL);
396
397         if(az && z->rpz) {
398                 /* keep RPZ linked list intact */
399                 lock_rw_wrlock(&az->rpz_lock);
400                 if(z->rpz_az_prev)
401                         z->rpz_az_prev->rpz_az_next = z->rpz_az_next;
402                 else
403                         az->rpz_first = z->rpz_az_next;
404                 if(z->rpz_az_next)
405                         z->rpz_az_next->rpz_az_prev = z->rpz_az_prev;
406                 lock_rw_unlock(&az->rpz_lock);
407         }
408         if(z->rpz)
409                 rpz_delete(z->rpz);
410         free(z->name);
411         free(z->zonefile);
412         free(z);
413 }
414
415 struct auth_zone*
416 auth_zone_create(struct auth_zones* az, uint8_t* nm, size_t nmlen,
417         uint16_t dclass)
418 {
419         struct auth_zone* z = (struct auth_zone*)calloc(1, sizeof(*z));
420         if(!z) {
421                 return NULL;
422         }
423         z->node.key = z;
424         z->dclass = dclass;
425         z->namelen = nmlen;
426         z->namelabs = dname_count_labels(nm);
427         z->name = memdup(nm, nmlen);
428         if(!z->name) {
429                 free(z);
430                 return NULL;
431         }
432         rbtree_init(&z->data, &auth_data_cmp);
433         lock_rw_init(&z->lock);
434         lock_protect(&z->lock, &z->name, sizeof(*z)-sizeof(rbnode_type)-
435                         sizeof(&z->rpz_az_next)-sizeof(&z->rpz_az_prev));
436         lock_rw_wrlock(&z->lock);
437         /* z lock protects all, except rbtree itself and the rpz linked list
438          * pointers, which are protected using az->lock */
439         if(!rbtree_insert(&az->ztree, &z->node)) {
440                 lock_rw_unlock(&z->lock);
441                 auth_zone_delete(z, NULL);
442                 log_warn("duplicate auth zone");
443                 return NULL;
444         }
445         return z;
446 }
447
448 struct auth_zone*
449 auth_zone_find(struct auth_zones* az, uint8_t* nm, size_t nmlen,
450         uint16_t dclass)
451 {
452         struct auth_zone key;
453         key.node.key = &key;
454         key.dclass = dclass;
455         key.name = nm;
456         key.namelen = nmlen;
457         key.namelabs = dname_count_labels(nm);
458         return (struct auth_zone*)rbtree_search(&az->ztree, &key);
459 }
460
461 struct auth_xfer*
462 auth_xfer_find(struct auth_zones* az, uint8_t* nm, size_t nmlen,
463         uint16_t dclass)
464 {
465         struct auth_xfer key;
466         key.node.key = &key;
467         key.dclass = dclass;
468         key.name = nm;
469         key.namelen = nmlen;
470         key.namelabs = dname_count_labels(nm);
471         return (struct auth_xfer*)rbtree_search(&az->xtree, &key);
472 }
473
474 /** find an auth zone or sorted less-or-equal, return true if exact */
475 static int
476 auth_zone_find_less_equal(struct auth_zones* az, uint8_t* nm, size_t nmlen,
477         uint16_t dclass, struct auth_zone** z)
478 {
479         struct auth_zone key;
480         key.node.key = &key;
481         key.dclass = dclass;
482         key.name = nm;
483         key.namelen = nmlen;
484         key.namelabs = dname_count_labels(nm);
485         return rbtree_find_less_equal(&az->ztree, &key, (rbnode_type**)z);
486 }
487
488
489 /** find the auth zone that is above the given name */
490 struct auth_zone*
491 auth_zones_find_zone(struct auth_zones* az, uint8_t* name, size_t name_len,
492         uint16_t dclass)
493 {
494         uint8_t* nm = name;
495         size_t nmlen = name_len;
496         struct auth_zone* z;
497         if(auth_zone_find_less_equal(az, nm, nmlen, dclass, &z)) {
498                 /* exact match */
499                 return z;
500         } else {
501                 /* less-or-nothing */
502                 if(!z) return NULL; /* nothing smaller, nothing above it */
503                 /* we found smaller name; smaller may be above the name,
504                  * but not below it. */
505                 nm = dname_get_shared_topdomain(z->name, name);
506                 dname_count_size_labels(nm, &nmlen);
507                 z = NULL;
508         }
509
510         /* search up */
511         while(!z) {
512                 z = auth_zone_find(az, nm, nmlen, dclass);
513                 if(z) return z;
514                 if(dname_is_root(nm)) break;
515                 dname_remove_label(&nm, &nmlen);
516         }
517         return NULL;
518 }
519
520 /** find or create zone with name str. caller must have lock on az. 
521  * returns a wrlocked zone */
522 static struct auth_zone*
523 auth_zones_find_or_add_zone(struct auth_zones* az, char* name)
524 {
525         uint8_t nm[LDNS_MAX_DOMAINLEN+1];
526         size_t nmlen = sizeof(nm);
527         struct auth_zone* z;
528
529         if(sldns_str2wire_dname_buf(name, nm, &nmlen) != 0) {
530                 log_err("cannot parse auth zone name: %s", name);
531                 return 0;
532         }
533         z = auth_zone_find(az, nm, nmlen, LDNS_RR_CLASS_IN);
534         if(!z) {
535                 /* not found, create the zone */
536                 z = auth_zone_create(az, nm, nmlen, LDNS_RR_CLASS_IN);
537         } else {
538                 lock_rw_wrlock(&z->lock);
539         }
540         return z;
541 }
542
543 /** find or create xfer zone with name str. caller must have lock on az. 
544  * returns a locked xfer */
545 static struct auth_xfer*
546 auth_zones_find_or_add_xfer(struct auth_zones* az, struct auth_zone* z)
547 {
548         struct auth_xfer* x;
549         x = auth_xfer_find(az, z->name, z->namelen, z->dclass);
550         if(!x) {
551                 /* not found, create the zone */
552                 x = auth_xfer_create(az, z);
553         } else {
554                 lock_basic_lock(&x->lock);
555         }
556         return x;
557 }
558
559 int
560 auth_zone_set_zonefile(struct auth_zone* z, char* zonefile)
561 {
562         if(z->zonefile) free(z->zonefile);
563         if(zonefile == NULL) {
564                 z->zonefile = NULL;
565         } else {
566                 z->zonefile = strdup(zonefile);
567                 if(!z->zonefile) {
568                         log_err("malloc failure");
569                         return 0;
570                 }
571         }
572         return 1;
573 }
574
575 /** set auth zone fallback. caller must have lock on zone */
576 int
577 auth_zone_set_fallback(struct auth_zone* z, char* fallbackstr)
578 {
579         if(strcmp(fallbackstr, "yes") != 0 && strcmp(fallbackstr, "no") != 0){
580                 log_err("auth zone fallback, expected yes or no, got %s",
581                         fallbackstr);
582                 return 0;
583         }
584         z->fallback_enabled = (strcmp(fallbackstr, "yes")==0);
585         return 1;
586 }
587
588 /** create domain with the given name */
589 static struct auth_data*
590 az_domain_create(struct auth_zone* z, uint8_t* nm, size_t nmlen)
591 {
592         struct auth_data* n = (struct auth_data*)malloc(sizeof(*n));
593         if(!n) return NULL;
594         memset(n, 0, sizeof(*n));
595         n->node.key = n;
596         n->name = memdup(nm, nmlen);
597         if(!n->name) {
598                 free(n);
599                 return NULL;
600         }
601         n->namelen = nmlen;
602         n->namelabs = dname_count_labels(nm);
603         if(!rbtree_insert(&z->data, &n->node)) {
604                 log_warn("duplicate auth domain name");
605                 free(n->name);
606                 free(n);
607                 return NULL;
608         }
609         return n;
610 }
611
612 /** find domain with exactly the given name */
613 static struct auth_data*
614 az_find_name(struct auth_zone* z, uint8_t* nm, size_t nmlen)
615 {
616         struct auth_zone key;
617         key.node.key = &key;
618         key.name = nm;
619         key.namelen = nmlen;
620         key.namelabs = dname_count_labels(nm);
621         return (struct auth_data*)rbtree_search(&z->data, &key);
622 }
623
624 /** Find domain name (or closest match) */
625 static void
626 az_find_domain(struct auth_zone* z, struct query_info* qinfo, int* node_exact,
627         struct auth_data** node)
628 {
629         struct auth_zone key;
630         key.node.key = &key;
631         key.name = qinfo->qname;
632         key.namelen = qinfo->qname_len;
633         key.namelabs = dname_count_labels(key.name);
634         *node_exact = rbtree_find_less_equal(&z->data, &key,
635                 (rbnode_type**)node);
636 }
637
638 /** find or create domain with name in zone */
639 static struct auth_data*
640 az_domain_find_or_create(struct auth_zone* z, uint8_t* dname,
641         size_t dname_len)
642 {
643         struct auth_data* n = az_find_name(z, dname, dname_len);
644         if(!n) {
645                 n = az_domain_create(z, dname, dname_len);
646         }
647         return n;
648 }
649
650 /** find rrset of given type in the domain */
651 static struct auth_rrset*
652 az_domain_rrset(struct auth_data* n, uint16_t t)
653 {
654         struct auth_rrset* rrset;
655         if(!n) return NULL;
656         rrset = n->rrsets;
657         while(rrset) {
658                 if(rrset->type == t)
659                         return rrset;
660                 rrset = rrset->next;
661         }
662         return NULL;
663 }
664
665 /** remove rrset of this type from domain */
666 static void
667 domain_remove_rrset(struct auth_data* node, uint16_t rr_type)
668 {
669         struct auth_rrset* rrset, *prev;
670         if(!node) return;
671         prev = NULL;
672         rrset = node->rrsets;
673         while(rrset) {
674                 if(rrset->type == rr_type) {
675                         /* found it, now delete it */
676                         if(prev) prev->next = rrset->next;
677                         else    node->rrsets = rrset->next;
678                         auth_rrset_delete(rrset);
679                         return;
680                 }
681                 prev = rrset;
682                 rrset = rrset->next;
683         }
684 }
685
686 /** find an rrsig index in the rrset.  returns true if found */
687 static int
688 az_rrset_find_rrsig(struct packed_rrset_data* d, uint8_t* rdata, size_t len,
689         size_t* index)
690 {
691         size_t i;
692         for(i=d->count; i<d->count + d->rrsig_count; i++) {
693                 if(d->rr_len[i] != len)
694                         continue;
695                 if(memcmp(d->rr_data[i], rdata, len) == 0) {
696                         *index = i;
697                         return 1;
698                 }
699         }
700         return 0;
701 }
702
703 /** see if rdata is duplicate */
704 static int
705 rdata_duplicate(struct packed_rrset_data* d, uint8_t* rdata, size_t len)
706 {
707         size_t i;
708         for(i=0; i<d->count + d->rrsig_count; i++) {
709                 if(d->rr_len[i] != len)
710                         continue;
711                 if(memcmp(d->rr_data[i], rdata, len) == 0)
712                         return 1;
713         }
714         return 0;
715 }
716
717 /** get rrsig type covered from rdata.
718  * @param rdata: rdata in wireformat, starting with 16bit rdlength.
719  * @param rdatalen: length of rdata buffer.
720  * @return type covered (or 0).
721  */
722 static uint16_t
723 rrsig_rdata_get_type_covered(uint8_t* rdata, size_t rdatalen)
724 {
725         if(rdatalen < 4)
726                 return 0;
727         return sldns_read_uint16(rdata+2);
728 }
729
730 /** remove RR from existing RRset. Also sig, if it is a signature.
731  * reallocates the packed rrset for a new one, false on alloc failure */
732 static int
733 rrset_remove_rr(struct auth_rrset* rrset, size_t index)
734 {
735         struct packed_rrset_data* d, *old = rrset->data;
736         size_t i;
737         if(index >= old->count + old->rrsig_count)
738                 return 0; /* index out of bounds */
739         d = (struct packed_rrset_data*)calloc(1, packed_rrset_sizeof(old) - (
740                 sizeof(size_t) + sizeof(uint8_t*) + sizeof(time_t) +
741                 old->rr_len[index]));
742         if(!d) {
743                 log_err("malloc failure");
744                 return 0;
745         }
746         d->ttl = old->ttl;
747         d->count = old->count;
748         d->rrsig_count = old->rrsig_count;
749         if(index < d->count) d->count--;
750         else d->rrsig_count--;
751         d->trust = old->trust;
752         d->security = old->security;
753
754         /* set rr_len, needed for ptr_fixup */
755         d->rr_len = (size_t*)((uint8_t*)d +
756                 sizeof(struct packed_rrset_data));
757         if(index > 0)
758                 memmove(d->rr_len, old->rr_len, (index)*sizeof(size_t));
759         if(index+1 < old->count+old->rrsig_count)
760                 memmove(&d->rr_len[index], &old->rr_len[index+1],
761                 (old->count+old->rrsig_count - (index+1))*sizeof(size_t));
762         packed_rrset_ptr_fixup(d);
763
764         /* move over ttls */
765         if(index > 0)
766                 memmove(d->rr_ttl, old->rr_ttl, (index)*sizeof(time_t));
767         if(index+1 < old->count+old->rrsig_count)
768                 memmove(&d->rr_ttl[index], &old->rr_ttl[index+1],
769                 (old->count+old->rrsig_count - (index+1))*sizeof(time_t));
770         
771         /* move over rr_data */
772         for(i=0; i<d->count+d->rrsig_count; i++) {
773                 size_t oldi;
774                 if(i < index) oldi = i;
775                 else oldi = i+1;
776                 memmove(d->rr_data[i], old->rr_data[oldi], d->rr_len[i]);
777         }
778
779         /* recalc ttl (lowest of remaining RR ttls) */
780         if(d->count + d->rrsig_count > 0)
781                 d->ttl = d->rr_ttl[0];
782         for(i=0; i<d->count+d->rrsig_count; i++) {
783                 if(d->rr_ttl[i] < d->ttl)
784                         d->ttl = d->rr_ttl[i];
785         }
786
787         free(rrset->data);
788         rrset->data = d;
789         return 1;
790 }
791
792 /** add RR to existing RRset. If insert_sig is true, add to rrsigs. 
793  * This reallocates the packed rrset for a new one */
794 static int
795 rrset_add_rr(struct auth_rrset* rrset, uint32_t rr_ttl, uint8_t* rdata,
796         size_t rdatalen, int insert_sig)
797 {
798         struct packed_rrset_data* d, *old = rrset->data;
799         size_t total, old_total;
800
801         d = (struct packed_rrset_data*)calloc(1, packed_rrset_sizeof(old)
802                 + sizeof(size_t) + sizeof(uint8_t*) + sizeof(time_t)
803                 + rdatalen);
804         if(!d) {
805                 log_err("out of memory");
806                 return 0;
807         }
808         /* copy base values */
809         memcpy(d, old, sizeof(struct packed_rrset_data));
810         if(!insert_sig) {
811                 d->count++;
812         } else {
813                 d->rrsig_count++;
814         }
815         old_total = old->count + old->rrsig_count;
816         total = d->count + d->rrsig_count;
817         /* set rr_len, needed for ptr_fixup */
818         d->rr_len = (size_t*)((uint8_t*)d +
819                 sizeof(struct packed_rrset_data));
820         if(old->count != 0)
821                 memmove(d->rr_len, old->rr_len, old->count*sizeof(size_t));
822         if(old->rrsig_count != 0)
823                 memmove(d->rr_len+d->count, old->rr_len+old->count,
824                         old->rrsig_count*sizeof(size_t));
825         if(!insert_sig)
826                 d->rr_len[d->count-1] = rdatalen;
827         else    d->rr_len[total-1] = rdatalen;
828         packed_rrset_ptr_fixup(d);
829         if((time_t)rr_ttl < d->ttl)
830                 d->ttl = rr_ttl;
831
832         /* copy old values into new array */
833         if(old->count != 0) {
834                 memmove(d->rr_ttl, old->rr_ttl, old->count*sizeof(time_t));
835                 /* all the old rr pieces are allocated sequential, so we
836                  * can copy them in one go */
837                 memmove(d->rr_data[0], old->rr_data[0],
838                         (old->rr_data[old->count-1] - old->rr_data[0]) +
839                         old->rr_len[old->count-1]);
840         }
841         if(old->rrsig_count != 0) {
842                 memmove(d->rr_ttl+d->count, old->rr_ttl+old->count,
843                         old->rrsig_count*sizeof(time_t));
844                 memmove(d->rr_data[d->count], old->rr_data[old->count],
845                         (old->rr_data[old_total-1] - old->rr_data[old->count]) +
846                         old->rr_len[old_total-1]);
847         }
848
849         /* insert new value */
850         if(!insert_sig) {
851                 d->rr_ttl[d->count-1] = rr_ttl;
852                 memmove(d->rr_data[d->count-1], rdata, rdatalen);
853         } else {
854                 d->rr_ttl[total-1] = rr_ttl;
855                 memmove(d->rr_data[total-1], rdata, rdatalen);
856         }
857
858         rrset->data = d;
859         free(old);
860         return 1;
861 }
862
863 /** Create new rrset for node with packed rrset with one RR element */
864 static struct auth_rrset*
865 rrset_create(struct auth_data* node, uint16_t rr_type, uint32_t rr_ttl,
866         uint8_t* rdata, size_t rdatalen)
867 {
868         struct auth_rrset* rrset = (struct auth_rrset*)calloc(1,
869                 sizeof(*rrset));
870         struct auth_rrset* p, *prev;
871         struct packed_rrset_data* d;
872         if(!rrset) {
873                 log_err("out of memory");
874                 return NULL;
875         }
876         rrset->type = rr_type;
877
878         /* the rrset data structure, with one RR */
879         d = (struct packed_rrset_data*)calloc(1,
880                 sizeof(struct packed_rrset_data) + sizeof(size_t) +
881                 sizeof(uint8_t*) + sizeof(time_t) + rdatalen);
882         if(!d) {
883                 free(rrset);
884                 log_err("out of memory");
885                 return NULL;
886         }
887         rrset->data = d;
888         d->ttl = rr_ttl;
889         d->trust = rrset_trust_prim_noglue;
890         d->rr_len = (size_t*)((uint8_t*)d + sizeof(struct packed_rrset_data));
891         d->rr_data = (uint8_t**)&(d->rr_len[1]);
892         d->rr_ttl = (time_t*)&(d->rr_data[1]);
893         d->rr_data[0] = (uint8_t*)&(d->rr_ttl[1]);
894
895         /* insert the RR */
896         d->rr_len[0] = rdatalen;
897         d->rr_ttl[0] = rr_ttl;
898         memmove(d->rr_data[0], rdata, rdatalen);
899         d->count++;
900
901         /* insert rrset into linked list for domain */
902         /* find sorted place to link the rrset into the list */
903         prev = NULL;
904         p = node->rrsets;
905         while(p && p->type<=rr_type) {
906                 prev = p;
907                 p = p->next;
908         }
909         /* so, prev is smaller, and p is larger than rr_type */
910         rrset->next = p;
911         if(prev) prev->next = rrset;
912         else node->rrsets = rrset;
913         return rrset;
914 }
915
916 /** count number (and size) of rrsigs that cover a type */
917 static size_t
918 rrsig_num_that_cover(struct auth_rrset* rrsig, uint16_t rr_type, size_t* sigsz)
919 {
920         struct packed_rrset_data* d = rrsig->data;
921         size_t i, num = 0;
922         *sigsz = 0;
923         log_assert(d && rrsig->type == LDNS_RR_TYPE_RRSIG);
924         for(i=0; i<d->count+d->rrsig_count; i++) {
925                 if(rrsig_rdata_get_type_covered(d->rr_data[i],
926                         d->rr_len[i]) == rr_type) {
927                         num++;
928                         (*sigsz) += d->rr_len[i];
929                 }
930         }
931         return num;
932 }
933
934 /** See if rrsig set has covered sigs for rrset and move them over */
935 static int
936 rrset_moveover_rrsigs(struct auth_data* node, uint16_t rr_type,
937         struct auth_rrset* rrset, struct auth_rrset* rrsig)
938 {
939         size_t sigs, sigsz, i, j, total;
940         struct packed_rrset_data* sigold = rrsig->data;
941         struct packed_rrset_data* old = rrset->data;
942         struct packed_rrset_data* d, *sigd;
943
944         log_assert(rrset->type == rr_type);
945         log_assert(rrsig->type == LDNS_RR_TYPE_RRSIG);
946         sigs = rrsig_num_that_cover(rrsig, rr_type, &sigsz);
947         if(sigs == 0) {
948                 /* 0 rrsigs to move over, done */
949                 return 1;
950         }
951
952         /* allocate rrset sigsz larger for extra sigs elements, and
953          * allocate rrsig sigsz smaller for less sigs elements. */
954         d = (struct packed_rrset_data*)calloc(1, packed_rrset_sizeof(old)
955                 + sigs*(sizeof(size_t) + sizeof(uint8_t*) + sizeof(time_t))
956                 + sigsz);
957         if(!d) {
958                 log_err("out of memory");
959                 return 0;
960         }
961         /* copy base values */
962         total = old->count + old->rrsig_count;
963         memcpy(d, old, sizeof(struct packed_rrset_data));
964         d->rrsig_count += sigs;
965         /* setup rr_len */
966         d->rr_len = (size_t*)((uint8_t*)d +
967                 sizeof(struct packed_rrset_data));
968         if(total != 0)
969                 memmove(d->rr_len, old->rr_len, total*sizeof(size_t));
970         j = d->count+d->rrsig_count-sigs;
971         for(i=0; i<sigold->count+sigold->rrsig_count; i++) {
972                 if(rrsig_rdata_get_type_covered(sigold->rr_data[i],
973                         sigold->rr_len[i]) == rr_type) {
974                         d->rr_len[j] = sigold->rr_len[i];
975                         j++;
976                 }
977         }
978         packed_rrset_ptr_fixup(d);
979
980         /* copy old values into new array */
981         if(total != 0) {
982                 memmove(d->rr_ttl, old->rr_ttl, total*sizeof(time_t));
983                 /* all the old rr pieces are allocated sequential, so we
984                  * can copy them in one go */
985                 memmove(d->rr_data[0], old->rr_data[0],
986                         (old->rr_data[total-1] - old->rr_data[0]) +
987                         old->rr_len[total-1]);
988         }
989
990         /* move over the rrsigs to the larger rrset*/
991         j = d->count+d->rrsig_count-sigs;
992         for(i=0; i<sigold->count+sigold->rrsig_count; i++) {
993                 if(rrsig_rdata_get_type_covered(sigold->rr_data[i],
994                         sigold->rr_len[i]) == rr_type) {
995                         /* move this one over to location j */
996                         d->rr_ttl[j] = sigold->rr_ttl[i];
997                         memmove(d->rr_data[j], sigold->rr_data[i],
998                                 sigold->rr_len[i]);
999                         if(d->rr_ttl[j] < d->ttl)
1000                                 d->ttl = d->rr_ttl[j];
1001                         j++;
1002                 }
1003         }
1004
1005         /* put it in and deallocate the old rrset */
1006         rrset->data = d;
1007         free(old);
1008
1009         /* now make rrsig set smaller */
1010         if(sigold->count+sigold->rrsig_count == sigs) {
1011                 /* remove all sigs from rrsig, remove it entirely */
1012                 domain_remove_rrset(node, LDNS_RR_TYPE_RRSIG);
1013                 return 1;
1014         }
1015         log_assert(packed_rrset_sizeof(sigold) > sigs*(sizeof(size_t) +
1016                 sizeof(uint8_t*) + sizeof(time_t)) + sigsz);
1017         sigd = (struct packed_rrset_data*)calloc(1, packed_rrset_sizeof(sigold)
1018                 - sigs*(sizeof(size_t) + sizeof(uint8_t*) + sizeof(time_t))
1019                 - sigsz);
1020         if(!sigd) {
1021                 /* no need to free up d, it has already been placed in the
1022                  * node->rrset structure */
1023                 log_err("out of memory");
1024                 return 0;
1025         }
1026         /* copy base values */
1027         memcpy(sigd, sigold, sizeof(struct packed_rrset_data));
1028         /* in sigd the RRSIGs are stored in the base of the RR, in count */
1029         sigd->count -= sigs;
1030         /* setup rr_len */
1031         sigd->rr_len = (size_t*)((uint8_t*)sigd +
1032                 sizeof(struct packed_rrset_data));
1033         j = 0;
1034         for(i=0; i<sigold->count+sigold->rrsig_count; i++) {
1035                 if(rrsig_rdata_get_type_covered(sigold->rr_data[i],
1036                         sigold->rr_len[i]) != rr_type) {
1037                         sigd->rr_len[j] = sigold->rr_len[i];
1038                         j++;
1039                 }
1040         }
1041         packed_rrset_ptr_fixup(sigd);
1042
1043         /* copy old values into new rrsig array */
1044         j = 0;
1045         for(i=0; i<sigold->count+sigold->rrsig_count; i++) {
1046                 if(rrsig_rdata_get_type_covered(sigold->rr_data[i],
1047                         sigold->rr_len[i]) != rr_type) {
1048                         /* move this one over to location j */
1049                         sigd->rr_ttl[j] = sigold->rr_ttl[i];
1050                         memmove(sigd->rr_data[j], sigold->rr_data[i],
1051                                 sigold->rr_len[i]);
1052                         if(j==0) sigd->ttl = sigd->rr_ttl[j];
1053                         else {
1054                                 if(sigd->rr_ttl[j] < sigd->ttl)
1055                                         sigd->ttl = sigd->rr_ttl[j];
1056                         }
1057                         j++;
1058                 }
1059         }
1060
1061         /* put it in and deallocate the old rrset */
1062         rrsig->data = sigd;
1063         free(sigold);
1064
1065         return 1;
1066 }
1067
1068 /** copy the rrsigs from the rrset to the rrsig rrset, because the rrset
1069  * is going to be deleted.  reallocates the RRSIG rrset data. */
1070 static int
1071 rrsigs_copy_from_rrset_to_rrsigset(struct auth_rrset* rrset,
1072         struct auth_rrset* rrsigset)
1073 {
1074         size_t i;
1075         if(rrset->data->rrsig_count == 0)
1076                 return 1;
1077
1078         /* move them over one by one, because there might be duplicates,
1079          * duplicates are ignored */
1080         for(i=rrset->data->count;
1081                 i<rrset->data->count+rrset->data->rrsig_count; i++) {
1082                 uint8_t* rdata = rrset->data->rr_data[i];
1083                 size_t rdatalen = rrset->data->rr_len[i];
1084                 time_t rr_ttl  = rrset->data->rr_ttl[i];
1085
1086                 if(rdata_duplicate(rrsigset->data, rdata, rdatalen)) {
1087                         continue;
1088                 }
1089                 if(!rrset_add_rr(rrsigset, rr_ttl, rdata, rdatalen, 0))
1090                         return 0;
1091         }
1092         return 1;
1093 }
1094
1095 /** Add rr to node, ignores duplicate RRs,
1096  * rdata points to buffer with rdatalen octets, starts with 2bytelength. */
1097 static int
1098 az_domain_add_rr(struct auth_data* node, uint16_t rr_type, uint32_t rr_ttl,
1099         uint8_t* rdata, size_t rdatalen, int* duplicate)
1100 {
1101         struct auth_rrset* rrset;
1102         /* packed rrsets have their rrsigs along with them, sort them out */
1103         if(rr_type == LDNS_RR_TYPE_RRSIG) {
1104                 uint16_t ctype = rrsig_rdata_get_type_covered(rdata, rdatalen);
1105                 if((rrset=az_domain_rrset(node, ctype))!= NULL) {
1106                         /* a node of the correct type exists, add the RRSIG
1107                          * to the rrset of the covered data type */
1108                         if(rdata_duplicate(rrset->data, rdata, rdatalen)) {
1109                                 if(duplicate) *duplicate = 1;
1110                                 return 1;
1111                         }
1112                         if(!rrset_add_rr(rrset, rr_ttl, rdata, rdatalen, 1))
1113                                 return 0;
1114                 } else if((rrset=az_domain_rrset(node, rr_type))!= NULL) {
1115                         /* add RRSIG to rrset of type RRSIG */
1116                         if(rdata_duplicate(rrset->data, rdata, rdatalen)) {
1117                                 if(duplicate) *duplicate = 1;
1118                                 return 1;
1119                         }
1120                         if(!rrset_add_rr(rrset, rr_ttl, rdata, rdatalen, 0))
1121                                 return 0;
1122                 } else {
1123                         /* create rrset of type RRSIG */
1124                         if(!rrset_create(node, rr_type, rr_ttl, rdata,
1125                                 rdatalen))
1126                                 return 0;
1127                 }
1128         } else {
1129                 /* normal RR type */
1130                 if((rrset=az_domain_rrset(node, rr_type))!= NULL) {
1131                         /* add data to existing node with data type */
1132                         if(rdata_duplicate(rrset->data, rdata, rdatalen)) {
1133                                 if(duplicate) *duplicate = 1;
1134                                 return 1;
1135                         }
1136                         if(!rrset_add_rr(rrset, rr_ttl, rdata, rdatalen, 0))
1137                                 return 0;
1138                 } else {
1139                         struct auth_rrset* rrsig;
1140                         /* create new node with data type */
1141                         if(!(rrset=rrset_create(node, rr_type, rr_ttl, rdata,
1142                                 rdatalen)))
1143                                 return 0;
1144
1145                         /* see if node of type RRSIG has signatures that
1146                          * cover the data type, and move them over */
1147                         /* and then make the RRSIG type smaller */
1148                         if((rrsig=az_domain_rrset(node, LDNS_RR_TYPE_RRSIG))
1149                                 != NULL) {
1150                                 if(!rrset_moveover_rrsigs(node, rr_type,
1151                                         rrset, rrsig))
1152                                         return 0;
1153                         }
1154                 }
1155         }
1156         return 1;
1157 }
1158
1159 /** insert RR into zone, ignore duplicates */
1160 static int
1161 az_insert_rr(struct auth_zone* z, uint8_t* rr, size_t rr_len,
1162         size_t dname_len, int* duplicate)
1163 {
1164         struct auth_data* node;
1165         uint8_t* dname = rr;
1166         uint16_t rr_type = sldns_wirerr_get_type(rr, rr_len, dname_len);
1167         uint16_t rr_class = sldns_wirerr_get_class(rr, rr_len, dname_len);
1168         uint32_t rr_ttl = sldns_wirerr_get_ttl(rr, rr_len, dname_len);
1169         size_t rdatalen = ((size_t)sldns_wirerr_get_rdatalen(rr, rr_len,
1170                 dname_len))+2;
1171         /* rdata points to rdata prefixed with uint16 rdatalength */
1172         uint8_t* rdata = sldns_wirerr_get_rdatawl(rr, rr_len, dname_len);
1173
1174         if(rr_class != z->dclass) {
1175                 log_err("wrong class for RR");
1176                 return 0;
1177         }
1178         if(!(node=az_domain_find_or_create(z, dname, dname_len))) {
1179                 log_err("cannot create domain");
1180                 return 0;
1181         }
1182         if(!az_domain_add_rr(node, rr_type, rr_ttl, rdata, rdatalen,
1183                 duplicate)) {
1184                 log_err("cannot add RR to domain");
1185                 return 0;
1186         }
1187         if(z->rpz) {
1188                 if(!(rpz_insert_rr(z->rpz, z->name, z->namelen, dname,
1189                         dname_len, rr_type, rr_class, rr_ttl, rdata, rdatalen,
1190                         rr, rr_len)))
1191                         return 0;
1192         }
1193         return 1;
1194 }
1195
1196 /** Remove rr from node, ignores nonexisting RRs,
1197  * rdata points to buffer with rdatalen octets, starts with 2bytelength. */
1198 static int
1199 az_domain_remove_rr(struct auth_data* node, uint16_t rr_type,
1200         uint8_t* rdata, size_t rdatalen, int* nonexist)
1201 {
1202         struct auth_rrset* rrset;
1203         size_t index = 0;
1204
1205         /* find the plain RR of the given type */
1206         if((rrset=az_domain_rrset(node, rr_type))!= NULL) {
1207                 if(packed_rrset_find_rr(rrset->data, rdata, rdatalen, &index)) {
1208                         if(rrset->data->count == 1 &&
1209                                 rrset->data->rrsig_count == 0) {
1210                                 /* last RR, delete the rrset */
1211                                 domain_remove_rrset(node, rr_type);
1212                         } else if(rrset->data->count == 1 &&
1213                                 rrset->data->rrsig_count != 0) {
1214                                 /* move RRSIGs to the RRSIG rrset, or
1215                                  * this one becomes that RRset */
1216                                 struct auth_rrset* rrsigset = az_domain_rrset(
1217                                         node, LDNS_RR_TYPE_RRSIG);
1218                                 if(rrsigset) {
1219                                         /* move left over rrsigs to the
1220                                          * existing rrset of type RRSIG */
1221                                         rrsigs_copy_from_rrset_to_rrsigset(
1222                                                 rrset, rrsigset);
1223                                         /* and then delete the rrset */
1224                                         domain_remove_rrset(node, rr_type);
1225                                 } else {
1226                                         /* no rrset of type RRSIG, this
1227                                          * set is now of that type,
1228                                          * just remove the rr */
1229                                         if(!rrset_remove_rr(rrset, index))
1230                                                 return 0;
1231                                         rrset->type = LDNS_RR_TYPE_RRSIG;
1232                                         rrset->data->count = rrset->data->rrsig_count;
1233                                         rrset->data->rrsig_count = 0;
1234                                 }
1235                         } else {
1236                                 /* remove the RR from the rrset */
1237                                 if(!rrset_remove_rr(rrset, index))
1238                                         return 0;
1239                         }
1240                         return 1;
1241                 }
1242                 /* rr not found in rrset */
1243         }
1244
1245         /* is it a type RRSIG, look under the covered type */
1246         if(rr_type == LDNS_RR_TYPE_RRSIG) {
1247                 uint16_t ctype = rrsig_rdata_get_type_covered(rdata, rdatalen);
1248                 if((rrset=az_domain_rrset(node, ctype))!= NULL) {
1249                         if(az_rrset_find_rrsig(rrset->data, rdata, rdatalen,
1250                                 &index)) {
1251                                 /* rrsig should have d->count > 0, be
1252                                  * over some rr of that type */
1253                                 /* remove the rrsig from the rrsigs list of the
1254                                  * rrset */
1255                                 if(!rrset_remove_rr(rrset, index))
1256                                         return 0;
1257                                 return 1;
1258                         }
1259                 }
1260                 /* also RRSIG not found */
1261         }
1262
1263         /* nothing found to delete */
1264         if(nonexist) *nonexist = 1;
1265         return 1;
1266 }
1267
1268 /** remove RR from zone, ignore if it does not exist, false on alloc failure*/
1269 static int
1270 az_remove_rr(struct auth_zone* z, uint8_t* rr, size_t rr_len,
1271         size_t dname_len, int* nonexist)
1272 {
1273         struct auth_data* node;
1274         uint8_t* dname = rr;
1275         uint16_t rr_type = sldns_wirerr_get_type(rr, rr_len, dname_len);
1276         uint16_t rr_class = sldns_wirerr_get_class(rr, rr_len, dname_len);
1277         size_t rdatalen = ((size_t)sldns_wirerr_get_rdatalen(rr, rr_len,
1278                 dname_len))+2;
1279         /* rdata points to rdata prefixed with uint16 rdatalength */
1280         uint8_t* rdata = sldns_wirerr_get_rdatawl(rr, rr_len, dname_len);
1281
1282         if(rr_class != z->dclass) {
1283                 log_err("wrong class for RR");
1284                 /* really also a nonexisting entry, because no records
1285                  * of that class in the zone, but return an error because
1286                  * getting records of the wrong class is a failure of the
1287                  * zone transfer */
1288                 return 0;
1289         }
1290         node = az_find_name(z, dname, dname_len);
1291         if(!node) {
1292                 /* node with that name does not exist */
1293                 /* nonexisting entry, because no such name */
1294                 *nonexist = 1;
1295                 return 1;
1296         }
1297         if(!az_domain_remove_rr(node, rr_type, rdata, rdatalen, nonexist)) {
1298                 /* alloc failure or so */
1299                 return 0;
1300         }
1301         /* remove the node, if necessary */
1302         /* an rrsets==NULL entry is not kept around for empty nonterminals,
1303          * and also parent nodes are not kept around, so we just delete it */
1304         if(node->rrsets == NULL) {
1305                 (void)rbtree_delete(&z->data, node);
1306                 auth_data_delete(node);
1307         }
1308         if(z->rpz) {
1309                 rpz_remove_rr(z->rpz, z->name, z->namelen, dname, dname_len,
1310                         rr_type, rr_class, rdata, rdatalen);
1311         }
1312         return 1;
1313 }
1314
1315 /** decompress an RR into the buffer where it'll be an uncompressed RR
1316  * with uncompressed dname and uncompressed rdata (dnames) */
1317 static int
1318 decompress_rr_into_buffer(struct sldns_buffer* buf, uint8_t* pkt,
1319         size_t pktlen, uint8_t* dname, uint16_t rr_type, uint16_t rr_class,
1320         uint32_t rr_ttl, uint8_t* rr_data, uint16_t rr_rdlen)
1321 {
1322         sldns_buffer pktbuf;
1323         size_t dname_len = 0;
1324         size_t rdlenpos;
1325         size_t rdlen;
1326         uint8_t* rd;
1327         const sldns_rr_descriptor* desc;
1328         sldns_buffer_init_frm_data(&pktbuf, pkt, pktlen);
1329         sldns_buffer_clear(buf);
1330
1331         /* decompress dname */
1332         sldns_buffer_set_position(&pktbuf,
1333                 (size_t)(dname - sldns_buffer_current(&pktbuf)));
1334         dname_len = pkt_dname_len(&pktbuf);
1335         if(dname_len == 0) return 0; /* parse fail on dname */
1336         if(!sldns_buffer_available(buf, dname_len)) return 0;
1337         dname_pkt_copy(&pktbuf, sldns_buffer_current(buf), dname);
1338         sldns_buffer_skip(buf, (ssize_t)dname_len);
1339
1340         /* type, class, ttl and rdatalength fields */
1341         if(!sldns_buffer_available(buf, 10)) return 0;
1342         sldns_buffer_write_u16(buf, rr_type);
1343         sldns_buffer_write_u16(buf, rr_class);
1344         sldns_buffer_write_u32(buf, rr_ttl);
1345         rdlenpos = sldns_buffer_position(buf);
1346         sldns_buffer_write_u16(buf, 0); /* rd length position */
1347
1348         /* decompress rdata */
1349         desc = sldns_rr_descript(rr_type);
1350         rd = rr_data;
1351         rdlen = rr_rdlen;
1352         if(rdlen > 0 && desc && desc->_dname_count > 0) {
1353                 int count = (int)desc->_dname_count;
1354                 int rdf = 0;
1355                 size_t len; /* how much rdata to plain copy */
1356                 size_t uncompressed_len, compressed_len;
1357                 size_t oldpos;
1358                 /* decompress dnames. */
1359                 while(rdlen > 0 && count) {
1360                         switch(desc->_wireformat[rdf]) {
1361                         case LDNS_RDF_TYPE_DNAME:
1362                                 sldns_buffer_set_position(&pktbuf,
1363                                         (size_t)(rd -
1364                                         sldns_buffer_begin(&pktbuf)));
1365                                 oldpos = sldns_buffer_position(&pktbuf);
1366                                 /* moves pktbuf to right after the
1367                                  * compressed dname, and returns uncompressed
1368                                  * dname length */
1369                                 uncompressed_len = pkt_dname_len(&pktbuf);
1370                                 if(!uncompressed_len)
1371                                         return 0; /* parse error in dname */
1372                                 if(!sldns_buffer_available(buf,
1373                                         uncompressed_len))
1374                                         /* dname too long for buffer */
1375                                         return 0;
1376                                 dname_pkt_copy(&pktbuf, 
1377                                         sldns_buffer_current(buf), rd);
1378                                 sldns_buffer_skip(buf, (ssize_t)uncompressed_len);
1379                                 compressed_len = sldns_buffer_position(
1380                                         &pktbuf) - oldpos;
1381                                 rd += compressed_len;
1382                                 rdlen -= compressed_len;
1383                                 count--;
1384                                 len = 0;
1385                                 break;
1386                         case LDNS_RDF_TYPE_STR:
1387                                 len = rd[0] + 1;
1388                                 break;
1389                         default:
1390                                 len = get_rdf_size(desc->_wireformat[rdf]);
1391                                 break;
1392                         }
1393                         if(len) {
1394                                 if(!sldns_buffer_available(buf, len))
1395                                         return 0; /* too long for buffer */
1396                                 sldns_buffer_write(buf, rd, len);
1397                                 rd += len;
1398                                 rdlen -= len;
1399                         }
1400                         rdf++;
1401                 }
1402         }
1403         /* copy remaining data */
1404         if(rdlen > 0) {
1405                 if(!sldns_buffer_available(buf, rdlen)) return 0;
1406                 sldns_buffer_write(buf, rd, rdlen);
1407         }
1408         /* fixup rdlength */
1409         sldns_buffer_write_u16_at(buf, rdlenpos,
1410                 sldns_buffer_position(buf)-rdlenpos-2);
1411         sldns_buffer_flip(buf);
1412         return 1;
1413 }
1414
1415 /** insert RR into zone, from packet, decompress RR,
1416  * if duplicate is nonNULL set the flag but otherwise ignore duplicates */
1417 static int
1418 az_insert_rr_decompress(struct auth_zone* z, uint8_t* pkt, size_t pktlen,
1419         struct sldns_buffer* scratch_buffer, uint8_t* dname, uint16_t rr_type,
1420         uint16_t rr_class, uint32_t rr_ttl, uint8_t* rr_data,
1421         uint16_t rr_rdlen, int* duplicate)
1422 {
1423         uint8_t* rr;
1424         size_t rr_len;
1425         size_t dname_len;
1426         if(!decompress_rr_into_buffer(scratch_buffer, pkt, pktlen, dname,
1427                 rr_type, rr_class, rr_ttl, rr_data, rr_rdlen)) {
1428                 log_err("could not decompress RR");
1429                 return 0;
1430         }
1431         rr = sldns_buffer_begin(scratch_buffer);
1432         rr_len = sldns_buffer_limit(scratch_buffer);
1433         dname_len = dname_valid(rr, rr_len);
1434         return az_insert_rr(z, rr, rr_len, dname_len, duplicate);
1435 }
1436
1437 /** remove RR from zone, from packet, decompress RR,
1438  * if nonexist is nonNULL set the flag but otherwise ignore nonexisting entries*/
1439 static int
1440 az_remove_rr_decompress(struct auth_zone* z, uint8_t* pkt, size_t pktlen,
1441         struct sldns_buffer* scratch_buffer, uint8_t* dname, uint16_t rr_type,
1442         uint16_t rr_class, uint32_t rr_ttl, uint8_t* rr_data,
1443         uint16_t rr_rdlen, int* nonexist)
1444 {
1445         uint8_t* rr;
1446         size_t rr_len;
1447         size_t dname_len;
1448         if(!decompress_rr_into_buffer(scratch_buffer, pkt, pktlen, dname,
1449                 rr_type, rr_class, rr_ttl, rr_data, rr_rdlen)) {
1450                 log_err("could not decompress RR");
1451                 return 0;
1452         }
1453         rr = sldns_buffer_begin(scratch_buffer);
1454         rr_len = sldns_buffer_limit(scratch_buffer);
1455         dname_len = dname_valid(rr, rr_len);
1456         return az_remove_rr(z, rr, rr_len, dname_len, nonexist);
1457 }
1458
1459 /** 
1460  * Parse zonefile
1461  * @param z: zone to read in.
1462  * @param in: file to read from (just opened).
1463  * @param rr: buffer to use for RRs, 64k.
1464  *      passed so that recursive includes can use the same buffer and do
1465  *      not grow the stack too much.
1466  * @param rrbuflen: sizeof rr buffer.
1467  * @param state: parse state with $ORIGIN, $TTL and 'prev-dname' and so on,
1468  *      that is kept between includes.
1469  *      The lineno is set at 1 and then increased by the function.
1470  * @param fname: file name.
1471  * @param depth: recursion depth for includes
1472  * @param cfg: config for chroot.
1473  * returns false on failure, has printed an error message
1474  */
1475 static int
1476 az_parse_file(struct auth_zone* z, FILE* in, uint8_t* rr, size_t rrbuflen,
1477         struct sldns_file_parse_state* state, char* fname, int depth,
1478         struct config_file* cfg)
1479 {
1480         size_t rr_len, dname_len;
1481         int status;
1482         state->lineno = 1;
1483
1484         while(!feof(in)) {
1485                 rr_len = rrbuflen;
1486                 dname_len = 0;
1487                 status = sldns_fp2wire_rr_buf(in, rr, &rr_len, &dname_len,
1488                         state);
1489                 if(status == LDNS_WIREPARSE_ERR_INCLUDE && rr_len == 0) {
1490                         /* we have $INCLUDE or $something */
1491                         if(strncmp((char*)rr, "$INCLUDE ", 9) == 0 ||
1492                            strncmp((char*)rr, "$INCLUDE\t", 9) == 0) {
1493                                 FILE* inc;
1494                                 int lineno_orig = state->lineno;
1495                                 char* incfile = (char*)rr + 8;
1496                                 if(depth > MAX_INCLUDE_DEPTH) {
1497                                         log_err("%s:%d max include depth"
1498                                           "exceeded", fname, state->lineno);
1499                                         return 0;
1500                                 }
1501                                 /* skip spaces */
1502                                 while(*incfile == ' ' || *incfile == '\t')
1503                                         incfile++;
1504                                 /* adjust for chroot on include file */
1505                                 if(cfg->chrootdir && cfg->chrootdir[0] &&
1506                                         strncmp(incfile, cfg->chrootdir,
1507                                                 strlen(cfg->chrootdir)) == 0)
1508                                         incfile += strlen(cfg->chrootdir);
1509                                 incfile = strdup(incfile);
1510                                 if(!incfile) {
1511                                         log_err("malloc failure");
1512                                         return 0;
1513                                 }
1514                                 verbose(VERB_ALGO, "opening $INCLUDE %s",
1515                                         incfile);
1516                                 inc = fopen(incfile, "r");
1517                                 if(!inc) {
1518                                         log_err("%s:%d cannot open include "
1519                                                 "file %s: %s", fname,
1520                                                 lineno_orig, incfile,
1521                                                 strerror(errno));
1522                                         free(incfile);
1523                                         return 0;
1524                                 }
1525                                 /* recurse read that file now */
1526                                 if(!az_parse_file(z, inc, rr, rrbuflen,
1527                                         state, incfile, depth+1, cfg)) {
1528                                         log_err("%s:%d cannot parse include "
1529                                                 "file %s", fname,
1530                                                 lineno_orig, incfile);
1531                                         fclose(inc);
1532                                         free(incfile);
1533                                         return 0;
1534                                 }
1535                                 fclose(inc);
1536                                 verbose(VERB_ALGO, "done with $INCLUDE %s",
1537                                         incfile);
1538                                 free(incfile);
1539                                 state->lineno = lineno_orig;
1540                         }
1541                         continue;
1542                 }
1543                 if(status != 0) {
1544                         log_err("parse error %s %d:%d: %s", fname,
1545                                 state->lineno, LDNS_WIREPARSE_OFFSET(status),
1546                                 sldns_get_errorstr_parse(status));
1547                         return 0;
1548                 }
1549                 if(rr_len == 0) {
1550                         /* EMPTY line, TTL or ORIGIN */
1551                         continue;
1552                 }
1553                 /* insert wirerr in rrbuf */
1554                 if(!az_insert_rr(z, rr, rr_len, dname_len, NULL)) {
1555                         char buf[17];
1556                         sldns_wire2str_type_buf(sldns_wirerr_get_type(rr,
1557                                 rr_len, dname_len), buf, sizeof(buf));
1558                         log_err("%s:%d cannot insert RR of type %s",
1559                                 fname, state->lineno, buf);
1560                         return 0;
1561                 }
1562         }
1563         return 1;
1564 }
1565
1566 int
1567 auth_zone_read_zonefile(struct auth_zone* z, struct config_file* cfg)
1568 {
1569         uint8_t rr[LDNS_RR_BUF_SIZE];
1570         struct sldns_file_parse_state state;
1571         char* zfilename;
1572         FILE* in;
1573         if(!z || !z->zonefile || z->zonefile[0]==0)
1574                 return 1; /* no file, or "", nothing to read */
1575         
1576         zfilename = z->zonefile;
1577         if(cfg->chrootdir && cfg->chrootdir[0] && strncmp(zfilename,
1578                 cfg->chrootdir, strlen(cfg->chrootdir)) == 0)
1579                 zfilename += strlen(cfg->chrootdir);
1580         if(verbosity >= VERB_ALGO) {
1581                 char nm[255+1];
1582                 dname_str(z->name, nm);
1583                 verbose(VERB_ALGO, "read zonefile %s for %s", zfilename, nm);
1584         }
1585         in = fopen(zfilename, "r");
1586         if(!in) {
1587                 char* n = sldns_wire2str_dname(z->name, z->namelen);
1588                 if(z->zone_is_slave && errno == ENOENT) {
1589                         /* we fetch the zone contents later, no file yet */
1590                         verbose(VERB_ALGO, "no zonefile %s for %s",
1591                                 zfilename, n?n:"error");
1592                         free(n);
1593                         return 1;
1594                 }
1595                 log_err("cannot open zonefile %s for %s: %s",
1596                         zfilename, n?n:"error", strerror(errno));
1597                 free(n);
1598                 return 0;
1599         }
1600
1601         /* clear the data tree */
1602         traverse_postorder(&z->data, auth_data_del, NULL);
1603         rbtree_init(&z->data, &auth_data_cmp);
1604         /* clear the RPZ policies */
1605         if(z->rpz)
1606                 rpz_clear(z->rpz);
1607
1608         memset(&state, 0, sizeof(state));
1609         /* default TTL to 3600 */
1610         state.default_ttl = 3600;
1611         /* set $ORIGIN to the zone name */
1612         if(z->namelen <= sizeof(state.origin)) {
1613                 memcpy(state.origin, z->name, z->namelen);
1614                 state.origin_len = z->namelen;
1615         }
1616         /* parse the (toplevel) file */
1617         if(!az_parse_file(z, in, rr, sizeof(rr), &state, zfilename, 0, cfg)) {
1618                 char* n = sldns_wire2str_dname(z->name, z->namelen);
1619                 log_err("error parsing zonefile %s for %s",
1620                         zfilename, n?n:"error");
1621                 free(n);
1622                 fclose(in);
1623                 return 0;
1624         }
1625         fclose(in);
1626
1627         if(z->rpz)
1628                 rpz_finish_config(z->rpz);
1629         return 1;
1630 }
1631
1632 /** write buffer to file and check return codes */
1633 static int
1634 write_out(FILE* out, const char* str, size_t len)
1635 {
1636         size_t r;
1637         if(len == 0)
1638                 return 1;
1639         r = fwrite(str, 1, len, out);
1640         if(r == 0) {
1641                 log_err("write failed: %s", strerror(errno));
1642                 return 0;
1643         } else if(r < len) {
1644                 log_err("write failed: too short (disk full?)");
1645                 return 0;
1646         }
1647         return 1;
1648 }
1649
1650 /** convert auth rr to string */
1651 static int
1652 auth_rr_to_string(uint8_t* nm, size_t nmlen, uint16_t tp, uint16_t cl,
1653         struct packed_rrset_data* data, size_t i, char* s, size_t buflen)
1654 {
1655         int w = 0;
1656         size_t slen = buflen, datlen;
1657         uint8_t* dat;
1658         if(i >= data->count) tp = LDNS_RR_TYPE_RRSIG;
1659         dat = nm;
1660         datlen = nmlen;
1661         w += sldns_wire2str_dname_scan(&dat, &datlen, &s, &slen, NULL, 0, NULL);
1662         w += sldns_str_print(&s, &slen, "\t");
1663         w += sldns_str_print(&s, &slen, "%lu\t", (unsigned long)data->rr_ttl[i]);
1664         w += sldns_wire2str_class_print(&s, &slen, cl);
1665         w += sldns_str_print(&s, &slen, "\t");
1666         w += sldns_wire2str_type_print(&s, &slen, tp);
1667         w += sldns_str_print(&s, &slen, "\t");
1668         datlen = data->rr_len[i]-2;
1669         dat = data->rr_data[i]+2;
1670         w += sldns_wire2str_rdata_scan(&dat, &datlen, &s, &slen, tp, NULL, 0, NULL);
1671
1672         if(tp == LDNS_RR_TYPE_DNSKEY) {
1673                 w += sldns_str_print(&s, &slen, " ;{id = %u}",
1674                         sldns_calc_keytag_raw(data->rr_data[i]+2,
1675                                 data->rr_len[i]-2));
1676         }
1677         w += sldns_str_print(&s, &slen, "\n");
1678
1679         if(w >= (int)buflen) {
1680                 log_nametypeclass(NO_VERBOSE, "RR too long to print", nm, tp, cl);
1681                 return 0;
1682         }
1683         return 1;
1684 }
1685
1686 /** write rrset to file */
1687 static int
1688 auth_zone_write_rrset(struct auth_zone* z, struct auth_data* node,
1689         struct auth_rrset* r, FILE* out)
1690 {
1691         size_t i, count = r->data->count + r->data->rrsig_count;
1692         char buf[LDNS_RR_BUF_SIZE];
1693         for(i=0; i<count; i++) {
1694                 if(!auth_rr_to_string(node->name, node->namelen, r->type,
1695                         z->dclass, r->data, i, buf, sizeof(buf))) {
1696                         verbose(VERB_ALGO, "failed to rr2str rr %d", (int)i);
1697                         continue;
1698                 }
1699                 if(!write_out(out, buf, strlen(buf)))
1700                         return 0;
1701         }
1702         return 1;
1703 }
1704
1705 /** write domain to file */
1706 static int
1707 auth_zone_write_domain(struct auth_zone* z, struct auth_data* n, FILE* out)
1708 {
1709         struct auth_rrset* r;
1710         /* if this is zone apex, write SOA first */
1711         if(z->namelen == n->namelen) {
1712                 struct auth_rrset* soa = az_domain_rrset(n, LDNS_RR_TYPE_SOA);
1713                 if(soa) {
1714                         if(!auth_zone_write_rrset(z, n, soa, out))
1715                                 return 0;
1716                 }
1717         }
1718         /* write all the RRsets for this domain */
1719         for(r = n->rrsets; r; r = r->next) {
1720                 if(z->namelen == n->namelen &&
1721                         r->type == LDNS_RR_TYPE_SOA)
1722                         continue; /* skip SOA here */
1723                 if(!auth_zone_write_rrset(z, n, r, out))
1724                         return 0;
1725         }
1726         return 1;
1727 }
1728
1729 int auth_zone_write_file(struct auth_zone* z, const char* fname)
1730 {
1731         FILE* out;
1732         struct auth_data* n;
1733         out = fopen(fname, "w");
1734         if(!out) {
1735                 log_err("could not open %s: %s", fname, strerror(errno));
1736                 return 0;
1737         }
1738         RBTREE_FOR(n, struct auth_data*, &z->data) {
1739                 if(!auth_zone_write_domain(z, n, out)) {
1740                         log_err("could not write domain to %s", fname);
1741                         fclose(out);
1742                         return 0;
1743                 }
1744         }
1745         fclose(out);
1746         return 1;
1747 }
1748
1749 /** offline verify for zonemd, while reading a zone file to immediately
1750  * spot bad hashes in zonefile as they are read.
1751  * Creates temp buffers, but uses anchors and validation environment
1752  * from the module_env. */
1753 static void
1754 zonemd_offline_verify(struct auth_zone* z, struct module_env* env_for_val,
1755         struct module_stack* mods)
1756 {
1757         struct module_env env;
1758         time_t now = 0;
1759         if(!z->zonemd_check)
1760                 return;
1761         env = *env_for_val;
1762         env.scratch_buffer = sldns_buffer_new(env.cfg->msg_buffer_size);
1763         if(!env.scratch_buffer) {
1764                 log_err("out of memory");
1765                 goto clean_exit;
1766         }
1767         env.scratch = regional_create();
1768         if(!env.now) {
1769                 env.now = &now;
1770                 now = time(NULL);
1771         }
1772         if(!env.scratch) {
1773                 log_err("out of memory");
1774                 goto clean_exit;
1775         }
1776         auth_zone_verify_zonemd(z, &env, mods, NULL, 1, 0);
1777
1778 clean_exit:
1779         /* clean up and exit */
1780         sldns_buffer_free(env.scratch_buffer);
1781         regional_destroy(env.scratch);
1782 }
1783
1784 /** read all auth zones from file (if they have) */
1785 static int
1786 auth_zones_read_zones(struct auth_zones* az, struct config_file* cfg,
1787         struct module_env* env, struct module_stack* mods)
1788 {
1789         struct auth_zone* z;
1790         lock_rw_wrlock(&az->lock);
1791         RBTREE_FOR(z, struct auth_zone*, &az->ztree) {
1792                 lock_rw_wrlock(&z->lock);
1793                 if(!auth_zone_read_zonefile(z, cfg)) {
1794                         lock_rw_unlock(&z->lock);
1795                         lock_rw_unlock(&az->lock);
1796                         return 0;
1797                 }
1798                 if(z->zonefile && z->zonefile[0]!=0 && env)
1799                         zonemd_offline_verify(z, env, mods);
1800                 lock_rw_unlock(&z->lock);
1801         }
1802         lock_rw_unlock(&az->lock);
1803         return 1;
1804 }
1805
1806 /** fetch the content of a ZONEMD RR from the rdata */
1807 static int zonemd_fetch_parameters(struct auth_rrset* zonemd_rrset, size_t i,
1808         uint32_t* serial, int* scheme, int* hashalgo, uint8_t** hash,
1809         size_t* hashlen)
1810 {
1811         size_t rr_len;
1812         uint8_t* rdata;
1813         if(i >= zonemd_rrset->data->count)
1814                 return 0;
1815         rr_len = zonemd_rrset->data->rr_len[i];
1816         if(rr_len < 2+4+1+1)
1817                 return 0; /* too short, for rdlen+serial+scheme+algo */
1818         rdata = zonemd_rrset->data->rr_data[i];
1819         *serial = sldns_read_uint32(rdata+2);
1820         *scheme = rdata[6];
1821         *hashalgo = rdata[7];
1822         *hashlen = rr_len - 8;
1823         if(*hashlen == 0)
1824                 *hash = NULL;
1825         else    *hash = rdata+8;
1826         return 1;
1827 }
1828
1829 /**
1830  * See if the ZONEMD scheme, hash occurs more than once.
1831  * @param zonemd_rrset: the zonemd rrset to check with the RRs in it.
1832  * @param index: index of the original, this is allowed to have that
1833  *      scheme and hashalgo, but other RRs should not have it.
1834  * @param scheme: the scheme to check for.
1835  * @param hashalgo: the hash algorithm to check for.
1836  * @return true if it occurs more than once.
1837  */
1838 static int zonemd_is_duplicate_scheme_hash(struct auth_rrset* zonemd_rrset,
1839         size_t index, int scheme, int hashalgo)
1840 {
1841         size_t j;
1842         for(j=0; j<zonemd_rrset->data->count; j++) {
1843                 uint32_t serial2 = 0;
1844                 int scheme2 = 0, hashalgo2 = 0;
1845                 uint8_t* hash2 = NULL;
1846                 size_t hashlen2 = 0;
1847                 if(index == j) {
1848                         /* this is the original */
1849                         continue;
1850                 }
1851                 if(!zonemd_fetch_parameters(zonemd_rrset, j, &serial2,
1852                         &scheme2, &hashalgo2, &hash2, &hashlen2)) {
1853                         /* malformed, skip it */
1854                         continue;
1855                 }
1856                 if(scheme == scheme2 && hashalgo == hashalgo2) {
1857                         /* duplicate scheme, hash */
1858                         verbose(VERB_ALGO, "zonemd duplicate for scheme %d "
1859                                 "and hash %d", scheme, hashalgo);
1860                         return 1;
1861                 }
1862         }
1863         return 0;
1864 }
1865
1866 /**
1867  * Check ZONEMDs if present for the auth zone.  Depending on config
1868  * it can warn or fail on that.  Checks the hash of the ZONEMD.
1869  * @param z: auth zone to check for.
1870  *      caller must hold lock on zone.
1871  * @param env: module env for temp buffers.
1872  * @param reason: returned on failure.
1873  * @return false on failure, true if hash checks out.
1874  */
1875 static int auth_zone_zonemd_check_hash(struct auth_zone* z,
1876         struct module_env* env, char** reason)
1877 {
1878         /* loop over ZONEMDs and see which one is valid. if not print
1879          * failure (depending on config) */
1880         struct auth_data* apex;
1881         struct auth_rrset* zonemd_rrset;
1882         size_t i;
1883         struct regional* region = NULL;
1884         struct sldns_buffer* buf = NULL;
1885         uint32_t soa_serial = 0;
1886         char* unsupported_reason = NULL;
1887         int only_unsupported = 1;
1888         region = env->scratch;
1889         regional_free_all(region);
1890         buf = env->scratch_buffer;
1891         if(!auth_zone_get_serial(z, &soa_serial)) {
1892                 *reason = "zone has no SOA serial";
1893                 return 0;
1894         }
1895
1896         apex = az_find_name(z, z->name, z->namelen);
1897         if(!apex) {
1898                 *reason = "zone has no apex";
1899                 return 0;
1900         }
1901         zonemd_rrset = az_domain_rrset(apex, LDNS_RR_TYPE_ZONEMD);
1902         if(!zonemd_rrset || zonemd_rrset->data->count==0) {
1903                 *reason = "zone has no ZONEMD";
1904                 return 0; /* no RRset or no RRs in rrset */
1905         }
1906
1907         /* we have a ZONEMD, check if it is correct */
1908         for(i=0; i<zonemd_rrset->data->count; i++) {
1909                 uint32_t serial = 0;
1910                 int scheme = 0, hashalgo = 0;
1911                 uint8_t* hash = NULL;
1912                 size_t hashlen = 0;
1913                 if(!zonemd_fetch_parameters(zonemd_rrset, i, &serial, &scheme,
1914                         &hashalgo, &hash, &hashlen)) {
1915                         /* malformed RR */
1916                         *reason = "ZONEMD rdata malformed";
1917                         only_unsupported = 0;
1918                         continue;
1919                 }
1920                 /* check for duplicates */
1921                 if(zonemd_is_duplicate_scheme_hash(zonemd_rrset, i, scheme,
1922                         hashalgo)) {
1923                         /* duplicate hash of the same scheme,hash
1924                          * is not allowed. */
1925                         *reason = "ZONEMD RRSet contains more than one RR "
1926                                 "with the same scheme and hash algorithm";
1927                         only_unsupported = 0;
1928                         continue;
1929                 }
1930                 regional_free_all(region);
1931                 if(serial != soa_serial) {
1932                         *reason = "ZONEMD serial is wrong";
1933                         only_unsupported = 0;
1934                         continue;
1935                 }
1936                 *reason = NULL;
1937                 if(auth_zone_generate_zonemd_check(z, scheme, hashalgo,
1938                         hash, hashlen, region, buf, reason)) {
1939                         /* success */
1940                         if(*reason) {
1941                                 if(!unsupported_reason)
1942                                         unsupported_reason = *reason;
1943                                 /* continue to check for valid ZONEMD */
1944                                 if(verbosity >= VERB_ALGO) {
1945                                         char zstr[255+1];
1946                                         dname_str(z->name, zstr);
1947                                         verbose(VERB_ALGO, "auth-zone %s ZONEMD %d %d is unsupported: %s", zstr, (int)scheme, (int)hashalgo, *reason);
1948                                 }
1949                                 *reason = NULL;
1950                                 continue;
1951                         }
1952                         if(verbosity >= VERB_ALGO) {
1953                                 char zstr[255+1];
1954                                 dname_str(z->name, zstr);
1955                                 if(!*reason)
1956                                         verbose(VERB_ALGO, "auth-zone %s ZONEMD hash is correct", zstr);
1957                         }
1958                         return 1;
1959                 }
1960                 only_unsupported = 0;
1961                 /* try next one */
1962         }
1963         /* have we seen no failures but only unsupported algo,
1964          * and one unsupported algorithm, or more. */
1965         if(only_unsupported && unsupported_reason) {
1966                 /* only unsupported algorithms, with valid serial, not
1967                  * malformed. Did not see supported algorithms, failed or
1968                  * successful ones. */
1969                 *reason = unsupported_reason;
1970                 return 1;
1971         }
1972         /* fail, we may have reason */
1973         if(!*reason)
1974                 *reason = "no ZONEMD records found";
1975         if(verbosity >= VERB_ALGO) {
1976                 char zstr[255+1];
1977                 dname_str(z->name, zstr);
1978                 verbose(VERB_ALGO, "auth-zone %s ZONEMD failed: %s", zstr, *reason);
1979         }
1980         return 0;
1981 }
1982
1983 /** find the apex SOA RRset, if it exists */
1984 struct auth_rrset* auth_zone_get_soa_rrset(struct auth_zone* z)
1985 {
1986         struct auth_data* apex;
1987         struct auth_rrset* soa;
1988         apex = az_find_name(z, z->name, z->namelen);
1989         if(!apex) return NULL;
1990         soa = az_domain_rrset(apex, LDNS_RR_TYPE_SOA);
1991         return soa;
1992 }
1993
1994 /** find serial number of zone or false if none */
1995 int
1996 auth_zone_get_serial(struct auth_zone* z, uint32_t* serial)
1997 {
1998         struct auth_data* apex;
1999         struct auth_rrset* soa;
2000         struct packed_rrset_data* d;
2001         apex = az_find_name(z, z->name, z->namelen);
2002         if(!apex) return 0;
2003         soa = az_domain_rrset(apex, LDNS_RR_TYPE_SOA);
2004         if(!soa || soa->data->count==0)
2005                 return 0; /* no RRset or no RRs in rrset */
2006         if(soa->data->rr_len[0] < 2+4*5) return 0; /* SOA too short */
2007         d = soa->data;
2008         *serial = sldns_read_uint32(d->rr_data[0]+(d->rr_len[0]-20));
2009         return 1;
2010 }
2011
2012 /** Find auth_zone SOA and populate the values in xfr(soa values). */
2013 int
2014 xfr_find_soa(struct auth_zone* z, struct auth_xfer* xfr)
2015 {
2016         struct auth_data* apex;
2017         struct auth_rrset* soa;
2018         struct packed_rrset_data* d;
2019         apex = az_find_name(z, z->name, z->namelen);
2020         if(!apex) return 0;
2021         soa = az_domain_rrset(apex, LDNS_RR_TYPE_SOA);
2022         if(!soa || soa->data->count==0)
2023                 return 0; /* no RRset or no RRs in rrset */
2024         if(soa->data->rr_len[0] < 2+4*5) return 0; /* SOA too short */
2025         /* SOA record ends with serial, refresh, retry, expiry, minimum,
2026          * as 4 byte fields */
2027         d = soa->data;
2028         xfr->have_zone = 1;
2029         xfr->serial = sldns_read_uint32(d->rr_data[0]+(d->rr_len[0]-20));
2030         xfr->refresh = sldns_read_uint32(d->rr_data[0]+(d->rr_len[0]-16));
2031         xfr->retry = sldns_read_uint32(d->rr_data[0]+(d->rr_len[0]-12));
2032         xfr->expiry = sldns_read_uint32(d->rr_data[0]+(d->rr_len[0]-8));
2033         /* soa minimum at d->rr_len[0]-4 */
2034         return 1;
2035 }
2036
2037 /** 
2038  * Setup auth_xfer zone
2039  * This populates the have_zone, soa values, and so on times.
2040  * Doesn't do network traffic yet, can set option flags.
2041  * @param z: locked by caller, and modified for setup
2042  * @param x: locked by caller, and modified.
2043  * @return false on failure.
2044  */
2045 static int
2046 auth_xfer_setup(struct auth_zone* z, struct auth_xfer* x)
2047 {
2048         /* for a zone without zone transfers, x==NULL, so skip them,
2049          * i.e. the zone config is fixed with no masters or urls */
2050         if(!z || !x) return 1;
2051         if(!xfr_find_soa(z, x)) {
2052                 return 1;
2053         }
2054         /* nothing for probe, nextprobe and transfer tasks */
2055         return 1;
2056 }
2057
2058 /**
2059  * Setup all zones
2060  * @param az: auth zones structure
2061  * @return false on failure.
2062  */
2063 static int
2064 auth_zones_setup_zones(struct auth_zones* az)
2065 {
2066         struct auth_zone* z;
2067         struct auth_xfer* x;
2068         lock_rw_wrlock(&az->lock);
2069         RBTREE_FOR(z, struct auth_zone*, &az->ztree) {
2070                 lock_rw_wrlock(&z->lock);
2071                 x = auth_xfer_find(az, z->name, z->namelen, z->dclass);
2072                 if(x) {
2073                         lock_basic_lock(&x->lock);
2074                 }
2075                 if(!auth_xfer_setup(z, x)) {
2076                         if(x) {
2077                                 lock_basic_unlock(&x->lock);
2078                         }
2079                         lock_rw_unlock(&z->lock);
2080                         lock_rw_unlock(&az->lock);
2081                         return 0;
2082                 }
2083                 if(x) {
2084                         lock_basic_unlock(&x->lock);
2085                 }
2086                 lock_rw_unlock(&z->lock);
2087         }
2088         lock_rw_unlock(&az->lock);
2089         return 1;
2090 }
2091
2092 /** set config items and create zones */
2093 static int
2094 auth_zones_cfg(struct auth_zones* az, struct config_auth* c)
2095 {
2096         struct auth_zone* z;
2097         struct auth_xfer* x = NULL;
2098
2099         /* create zone */
2100         if(c->isrpz) {
2101                 /* if the rpz lock is needed, grab it before the other
2102                  * locks to avoid a lock dependency cycle */
2103                 lock_rw_wrlock(&az->rpz_lock);
2104         }
2105         lock_rw_wrlock(&az->lock);
2106         if(!(z=auth_zones_find_or_add_zone(az, c->name))) {
2107                 lock_rw_unlock(&az->lock);
2108                 if(c->isrpz) {
2109                         lock_rw_unlock(&az->rpz_lock);
2110                 }
2111                 return 0;
2112         }
2113         if(c->masters || c->urls) {
2114                 if(!(x=auth_zones_find_or_add_xfer(az, z))) {
2115                         lock_rw_unlock(&az->lock);
2116                         lock_rw_unlock(&z->lock);
2117                         if(c->isrpz) {
2118                                 lock_rw_unlock(&az->rpz_lock);
2119                         }
2120                         return 0;
2121                 }
2122         }
2123         if(c->for_downstream)
2124                 az->have_downstream = 1;
2125         lock_rw_unlock(&az->lock);
2126
2127         /* set options */
2128         z->zone_deleted = 0;
2129         if(!auth_zone_set_zonefile(z, c->zonefile)) {
2130                 if(x) {
2131                         lock_basic_unlock(&x->lock);
2132                 }
2133                 lock_rw_unlock(&z->lock);
2134                 if(c->isrpz) {
2135                         lock_rw_unlock(&az->rpz_lock);
2136                 }
2137                 return 0;
2138         }
2139         z->for_downstream = c->for_downstream;
2140         z->for_upstream = c->for_upstream;
2141         z->fallback_enabled = c->fallback_enabled;
2142         z->zonemd_check = c->zonemd_check;
2143         z->zonemd_reject_absence = c->zonemd_reject_absence;
2144         if(c->isrpz && !z->rpz){
2145                 if(!(z->rpz = rpz_create(c))){
2146                         fatal_exit("Could not setup RPZ zones");
2147                         return 0;
2148                 }
2149                 lock_protect(&z->lock, &z->rpz->local_zones, sizeof(*z->rpz));
2150                 /* the az->rpz_lock is locked above */
2151                 z->rpz_az_next = az->rpz_first;
2152                 if(az->rpz_first)
2153                         az->rpz_first->rpz_az_prev = z;
2154                 az->rpz_first = z;
2155         }
2156         if(c->isrpz) {
2157                 lock_rw_unlock(&az->rpz_lock);
2158         }
2159
2160         /* xfer zone */
2161         if(x) {
2162                 z->zone_is_slave = 1;
2163                 /* set options on xfer zone */
2164                 if(!xfer_set_masters(&x->task_probe->masters, c, 0)) {
2165                         lock_basic_unlock(&x->lock);
2166                         lock_rw_unlock(&z->lock);
2167                         return 0;
2168                 }
2169                 if(!xfer_set_masters(&x->task_transfer->masters, c, 1)) {
2170                         lock_basic_unlock(&x->lock);
2171                         lock_rw_unlock(&z->lock);
2172                         return 0;
2173                 }
2174                 lock_basic_unlock(&x->lock);
2175         }
2176
2177         lock_rw_unlock(&z->lock);
2178         return 1;
2179 }
2180
2181 /** set all auth zones deleted, then in auth_zones_cfg, it marks them
2182  * as nondeleted (if they are still in the config), and then later
2183  * we can find deleted zones */
2184 static void
2185 az_setall_deleted(struct auth_zones* az)
2186 {
2187         struct auth_zone* z;
2188         lock_rw_wrlock(&az->lock);
2189         RBTREE_FOR(z, struct auth_zone*, &az->ztree) {
2190                 lock_rw_wrlock(&z->lock);
2191                 z->zone_deleted = 1;
2192                 lock_rw_unlock(&z->lock);
2193         }
2194         lock_rw_unlock(&az->lock);
2195 }
2196
2197 /** find zones that are marked deleted and delete them.
2198  * This is called from apply_cfg, and there are no threads and no
2199  * workers, so the xfr can just be deleted. */
2200 static void
2201 az_delete_deleted_zones(struct auth_zones* az)
2202 {
2203         struct auth_zone* z;
2204         struct auth_zone* delete_list = NULL, *next;
2205         struct auth_xfer* xfr;
2206         lock_rw_wrlock(&az->lock);
2207         RBTREE_FOR(z, struct auth_zone*, &az->ztree) {
2208                 lock_rw_wrlock(&z->lock);
2209                 if(z->zone_deleted) {
2210                         /* we cannot alter the rbtree right now, but
2211                          * we can put it on a linked list and then
2212                          * delete it */
2213                         z->delete_next = delete_list;
2214                         delete_list = z;
2215                 }
2216                 lock_rw_unlock(&z->lock);
2217         }
2218         /* now we are out of the tree loop and we can loop and delete
2219          * the zones */
2220         z = delete_list;
2221         while(z) {
2222                 next = z->delete_next;
2223                 xfr = auth_xfer_find(az, z->name, z->namelen, z->dclass);
2224                 if(xfr) {
2225                         (void)rbtree_delete(&az->xtree, &xfr->node);
2226                         auth_xfer_delete(xfr);
2227                 }
2228                 (void)rbtree_delete(&az->ztree, &z->node);
2229                 auth_zone_delete(z, az);
2230                 z = next;
2231         }
2232         lock_rw_unlock(&az->lock);
2233 }
2234
2235 int auth_zones_apply_cfg(struct auth_zones* az, struct config_file* cfg,
2236         int setup, int* is_rpz, struct module_env* env,
2237         struct module_stack* mods)
2238 {
2239         struct config_auth* p;
2240         az_setall_deleted(az);
2241         for(p = cfg->auths; p; p = p->next) {
2242                 if(!p->name || p->name[0] == 0) {
2243                         log_warn("auth-zone without a name, skipped");
2244                         continue;
2245                 }
2246                 *is_rpz = (*is_rpz || p->isrpz);
2247                 if(!auth_zones_cfg(az, p)) {
2248                         log_err("cannot config auth zone %s", p->name);
2249                         return 0;
2250                 }
2251         }
2252         az_delete_deleted_zones(az);
2253         if(!auth_zones_read_zones(az, cfg, env, mods))
2254                 return 0;
2255         if(setup) {
2256                 if(!auth_zones_setup_zones(az))
2257                         return 0;
2258         }
2259         return 1;
2260 }
2261
2262 /** delete chunks
2263  * @param at: transfer structure with chunks list.  The chunks and their
2264  *      data are freed.
2265  */
2266 static void
2267 auth_chunks_delete(struct auth_transfer* at)
2268 {
2269         if(at->chunks_first) {
2270                 struct auth_chunk* c, *cn;
2271                 c = at->chunks_first;
2272                 while(c) {
2273                         cn = c->next;
2274                         free(c->data);
2275                         free(c);
2276                         c = cn;
2277                 }
2278         }
2279         at->chunks_first = NULL;
2280         at->chunks_last = NULL;
2281 }
2282
2283 /** free master addr list */
2284 static void
2285 auth_free_master_addrs(struct auth_addr* list)
2286 {
2287         struct auth_addr *n;
2288         while(list) {
2289                 n = list->next;
2290                 free(list);
2291                 list = n;
2292         }
2293 }
2294
2295 /** free the masters list */
2296 static void
2297 auth_free_masters(struct auth_master* list)
2298 {
2299         struct auth_master* n;
2300         while(list) {
2301                 n = list->next;
2302                 auth_free_master_addrs(list->list);
2303                 free(list->host);
2304                 free(list->file);
2305                 free(list);
2306                 list = n;
2307         }
2308 }
2309
2310 /** delete auth xfer structure
2311  * @param xfr: delete this xfer and its tasks.
2312  */
2313 static void
2314 auth_xfer_delete(struct auth_xfer* xfr)
2315 {
2316         if(!xfr) return;
2317         lock_basic_destroy(&xfr->lock);
2318         free(xfr->name);
2319         if(xfr->task_nextprobe) {
2320                 comm_timer_delete(xfr->task_nextprobe->timer);
2321                 free(xfr->task_nextprobe);
2322         }
2323         if(xfr->task_probe) {
2324                 auth_free_masters(xfr->task_probe->masters);
2325                 comm_point_delete(xfr->task_probe->cp);
2326                 comm_timer_delete(xfr->task_probe->timer);
2327                 free(xfr->task_probe);
2328         }
2329         if(xfr->task_transfer) {
2330                 auth_free_masters(xfr->task_transfer->masters);
2331                 comm_point_delete(xfr->task_transfer->cp);
2332                 comm_timer_delete(xfr->task_transfer->timer);
2333                 if(xfr->task_transfer->chunks_first) {
2334                         auth_chunks_delete(xfr->task_transfer);
2335                 }
2336                 free(xfr->task_transfer);
2337         }
2338         auth_free_masters(xfr->allow_notify_list);
2339         free(xfr);
2340 }
2341
2342 /** helper traverse to delete zones */
2343 static void
2344 auth_zone_del(rbnode_type* n, void* ATTR_UNUSED(arg))
2345 {
2346         struct auth_zone* z = (struct auth_zone*)n->key;
2347         auth_zone_delete(z, NULL);
2348 }
2349
2350 /** helper traverse to delete xfer zones */
2351 static void
2352 auth_xfer_del(rbnode_type* n, void* ATTR_UNUSED(arg))
2353 {
2354         struct auth_xfer* z = (struct auth_xfer*)n->key;
2355         auth_xfer_delete(z);
2356 }
2357
2358 void auth_zones_delete(struct auth_zones* az)
2359 {
2360         if(!az) return;
2361         lock_rw_destroy(&az->lock);
2362         lock_rw_destroy(&az->rpz_lock);
2363         traverse_postorder(&az->ztree, auth_zone_del, NULL);
2364         traverse_postorder(&az->xtree, auth_xfer_del, NULL);
2365         free(az);
2366 }
2367
2368 /** true if domain has only nsec3 */
2369 static int
2370 domain_has_only_nsec3(struct auth_data* n)
2371 {
2372         struct auth_rrset* rrset = n->rrsets;
2373         int nsec3_seen = 0;
2374         while(rrset) {
2375                 if(rrset->type == LDNS_RR_TYPE_NSEC3) {
2376                         nsec3_seen = 1;
2377                 } else if(rrset->type != LDNS_RR_TYPE_RRSIG) {
2378                         return 0;
2379                 }
2380                 rrset = rrset->next;
2381         }
2382         return nsec3_seen;
2383 }
2384
2385 /** see if the domain has a wildcard child '*.domain' */
2386 static struct auth_data*
2387 az_find_wildcard_domain(struct auth_zone* z, uint8_t* nm, size_t nmlen)
2388 {
2389         uint8_t wc[LDNS_MAX_DOMAINLEN];
2390         if(nmlen+2 > sizeof(wc))
2391                 return NULL; /* result would be too long */
2392         wc[0] = 1; /* length of wildcard label */
2393         wc[1] = (uint8_t)'*'; /* wildcard label */
2394         memmove(wc+2, nm, nmlen);
2395         return az_find_name(z, wc, nmlen+2);
2396 }
2397
2398 /** find wildcard between qname and cename */
2399 static struct auth_data*
2400 az_find_wildcard(struct auth_zone* z, struct query_info* qinfo,
2401         struct auth_data* ce)
2402 {
2403         uint8_t* nm = qinfo->qname;
2404         size_t nmlen = qinfo->qname_len;
2405         struct auth_data* node;
2406         if(!dname_subdomain_c(nm, z->name))
2407                 return NULL; /* out of zone */
2408         while((node=az_find_wildcard_domain(z, nm, nmlen))==NULL) {
2409                 /* see if we can go up to find the wildcard */
2410                 if(nmlen == z->namelen)
2411                         return NULL; /* top of zone reached */
2412                 if(ce && nmlen == ce->namelen)
2413                         return NULL; /* ce reached */
2414                 if(dname_is_root(nm))
2415                         return NULL; /* cannot go up */
2416                 dname_remove_label(&nm, &nmlen);
2417         }
2418         return node;
2419 }
2420
2421 /** domain is not exact, find first candidate ce (name that matches
2422  * a part of qname) in tree */
2423 static struct auth_data*
2424 az_find_candidate_ce(struct auth_zone* z, struct query_info* qinfo,
2425         struct auth_data* n)
2426 {
2427         uint8_t* nm;
2428         size_t nmlen;
2429         if(n) {
2430                 nm = dname_get_shared_topdomain(qinfo->qname, n->name);
2431         } else {
2432                 nm = qinfo->qname;
2433         }
2434         dname_count_size_labels(nm, &nmlen);
2435         n = az_find_name(z, nm, nmlen);
2436         /* delete labels and go up on name */
2437         while(!n) {
2438                 if(dname_is_root(nm))
2439                         return NULL; /* cannot go up */
2440                 dname_remove_label(&nm, &nmlen);
2441                 n = az_find_name(z, nm, nmlen);
2442         }
2443         return n;
2444 }
2445
2446 /** go up the auth tree to next existing name. */
2447 static struct auth_data*
2448 az_domain_go_up(struct auth_zone* z, struct auth_data* n)
2449 {
2450         uint8_t* nm = n->name;
2451         size_t nmlen = n->namelen;
2452         while(!dname_is_root(nm)) {
2453                 dname_remove_label(&nm, &nmlen);
2454                 if((n=az_find_name(z, nm, nmlen)) != NULL)
2455                         return n;
2456         }
2457         return NULL;
2458 }
2459
2460 /** Find the closest encloser, an name that exists and is above the
2461  * qname.
2462  * return true if the node (param node) is existing, nonobscured and
2463  *      can be used to generate answers from.  It is then also node_exact.
2464  * returns false if the node is not good enough (or it wasn't node_exact)
2465  *      in this case the ce can be filled.
2466  *      if ce is NULL, no ce exists, and likely the zone is completely empty,
2467  *      not even with a zone apex.
2468  *      if ce is nonNULL it is the closest enclosing upper name (that exists
2469  *      itself for answer purposes).  That name may have DNAME, NS or wildcard
2470  *      rrset is the closest DNAME or NS rrset that was found.
2471  */
2472 static int
2473 az_find_ce(struct auth_zone* z, struct query_info* qinfo,
2474         struct auth_data* node, int node_exact, struct auth_data** ce,
2475         struct auth_rrset** rrset)
2476 {
2477         struct auth_data* n = node;
2478         *ce = NULL;
2479         *rrset = NULL;
2480         if(!node_exact) {
2481                 /* if not exact, lookup closest exact match */
2482                 n = az_find_candidate_ce(z, qinfo, n);
2483         } else {
2484                 /* if exact, the node itself is the first candidate ce */
2485                 *ce = n;
2486         }
2487
2488         /* no direct answer from nsec3-only domains */
2489         if(n && domain_has_only_nsec3(n)) {
2490                 node_exact = 0;
2491                 *ce = NULL;
2492         }
2493
2494         /* with exact matches, walk up the labels until we find the
2495          * delegation, or DNAME or zone end */
2496         while(n) {
2497                 /* see if the current candidate has issues */
2498                 /* not zone apex and has type NS */
2499                 if(n->namelen != z->namelen &&
2500                         (*rrset=az_domain_rrset(n, LDNS_RR_TYPE_NS)) &&
2501                         /* delegate here, but DS at exact the dp has notype */
2502                         (qinfo->qtype != LDNS_RR_TYPE_DS || 
2503                         n->namelen != qinfo->qname_len)) {
2504                         /* referral */
2505                         /* this is ce and the lowernode is nonexisting */
2506                         *ce = n;
2507                         return 0;
2508                 }
2509                 /* not equal to qname and has type DNAME */
2510                 if(n->namelen != qinfo->qname_len &&
2511                         (*rrset=az_domain_rrset(n, LDNS_RR_TYPE_DNAME))) {
2512                         /* this is ce and the lowernode is nonexisting */
2513                         *ce = n;
2514                         return 0;
2515                 }
2516
2517                 if(*ce == NULL && !domain_has_only_nsec3(n)) {
2518                         /* if not found yet, this exact name must be
2519                          * our lowest match (but not nsec3onlydomain) */
2520                         *ce = n;
2521                 }
2522
2523                 /* walk up the tree by removing labels from name and lookup */
2524                 n = az_domain_go_up(z, n);
2525         }
2526         /* found no problems, if it was an exact node, it is fine to use */
2527         return node_exact;
2528 }
2529
2530 /** add additional A/AAAA from domain names in rrset rdata (+offset)
2531  * offset is number of bytes in rdata where the dname is located. */
2532 static int
2533 az_add_additionals_from(struct auth_zone* z, struct regional* region,
2534         struct dns_msg* msg, struct auth_rrset* rrset, size_t offset)
2535 {
2536         struct packed_rrset_data* d = rrset->data;
2537         size_t i;
2538         if(!d) return 0;
2539         for(i=0; i<d->count; i++) {
2540                 size_t dlen;
2541                 struct auth_data* domain;
2542                 struct auth_rrset* ref;
2543                 if(d->rr_len[i] < 2+offset)
2544                         continue; /* too short */
2545                 if(!(dlen = dname_valid(d->rr_data[i]+2+offset,
2546                         d->rr_len[i]-2-offset)))
2547                         continue; /* malformed */
2548                 domain = az_find_name(z, d->rr_data[i]+2+offset, dlen);
2549                 if(!domain)
2550                         continue;
2551                 if((ref=az_domain_rrset(domain, LDNS_RR_TYPE_A)) != NULL) {
2552                         if(!msg_add_rrset_ar(z, region, msg, domain, ref))
2553                                 return 0;
2554                 }
2555                 if((ref=az_domain_rrset(domain, LDNS_RR_TYPE_AAAA)) != NULL) {
2556                         if(!msg_add_rrset_ar(z, region, msg, domain, ref))
2557                                 return 0;
2558                 }
2559         }
2560         return 1;
2561 }
2562
2563 /** add negative SOA record (with negative TTL) */
2564 static int
2565 az_add_negative_soa(struct auth_zone* z, struct regional* region,
2566         struct dns_msg* msg)
2567 {
2568         time_t minimum;
2569         size_t i;
2570         struct packed_rrset_data* d;
2571         struct auth_rrset* soa;
2572         struct auth_data* apex = az_find_name(z, z->name, z->namelen);
2573         if(!apex) return 0;
2574         soa = az_domain_rrset(apex, LDNS_RR_TYPE_SOA);
2575         if(!soa) return 0;
2576         /* must be first to put in message; we want to fix the TTL with
2577          * one RRset here, otherwise we'd need to loop over the RRs to get
2578          * the resulting lower TTL */
2579         log_assert(msg->rep->rrset_count == 0);
2580         if(!msg_add_rrset_ns(z, region, msg, apex, soa)) return 0;
2581         /* fixup TTL */
2582         d = (struct packed_rrset_data*)msg->rep->rrsets[msg->rep->rrset_count-1]->entry.data;
2583         /* last 4 bytes are minimum ttl in network format */
2584         if(d->count == 0) return 0;
2585         if(d->rr_len[0] < 2+4) return 0;
2586         minimum = (time_t)sldns_read_uint32(d->rr_data[0]+(d->rr_len[0]-4));
2587         minimum = d->ttl<minimum?d->ttl:minimum;
2588         d->ttl = minimum;
2589         for(i=0; i < d->count + d->rrsig_count; i++)
2590                 d->rr_ttl[i] = minimum;
2591         msg->rep->ttl = get_rrset_ttl(msg->rep->rrsets[0]);
2592         msg->rep->prefetch_ttl = PREFETCH_TTL_CALC(msg->rep->ttl);
2593         msg->rep->serve_expired_ttl = msg->rep->ttl + SERVE_EXPIRED_TTL;
2594         return 1;
2595 }
2596
2597 /** See if the query goes to empty nonterminal (that has no auth_data,
2598  * but there are nodes underneath.  We already checked that there are
2599  * not NS, or DNAME above, so that we only need to check if some node
2600  * exists below (with nonempty rr list), return true if emptynonterminal */
2601 static int
2602 az_empty_nonterminal(struct auth_zone* z, struct query_info* qinfo,
2603         struct auth_data* node)
2604 {
2605         struct auth_data* next;
2606         if(!node) {
2607                 /* no smaller was found, use first (smallest) node as the
2608                  * next one */
2609                 next = (struct auth_data*)rbtree_first(&z->data);
2610         } else {
2611                 next = (struct auth_data*)rbtree_next(&node->node);
2612         }
2613         while(next && (rbnode_type*)next != RBTREE_NULL && next->rrsets == NULL) {
2614                 /* the next name has empty rrsets, is an empty nonterminal
2615                  * itself, see if there exists something below it */
2616                 next = (struct auth_data*)rbtree_next(&node->node);
2617         }
2618         if((rbnode_type*)next == RBTREE_NULL || !next) {
2619                 /* there is no next node, so something below it cannot
2620                  * exist */
2621                 return 0;
2622         }
2623         /* a next node exists, if there was something below the query,
2624          * this node has to be it.  See if it is below the query name */
2625         if(dname_strict_subdomain_c(next->name, qinfo->qname))
2626                 return 1;
2627         return 0;
2628 }
2629
2630 /** create synth cname target name in buffer, or fail if too long */
2631 static size_t
2632 synth_cname_buf(uint8_t* qname, size_t qname_len, size_t dname_len,
2633         uint8_t* dtarg, size_t dtarglen, uint8_t* buf, size_t buflen)
2634 {
2635         size_t newlen = qname_len + dtarglen - dname_len;
2636         if(newlen > buflen) {
2637                 /* YXDOMAIN error */
2638                 return 0;
2639         }
2640         /* new name is concatenation of qname front (without DNAME owner)
2641          * and DNAME target name */
2642         memcpy(buf, qname, qname_len-dname_len);
2643         memmove(buf+(qname_len-dname_len), dtarg, dtarglen);
2644         return newlen;
2645 }
2646
2647 /** create synthetic CNAME rrset for in a DNAME answer in region,
2648  * false on alloc failure, cname==NULL when name too long. */
2649 static int
2650 create_synth_cname(uint8_t* qname, size_t qname_len, struct regional* region,
2651         struct auth_data* node, struct auth_rrset* dname, uint16_t dclass,
2652         struct ub_packed_rrset_key** cname)
2653 {
2654         uint8_t buf[LDNS_MAX_DOMAINLEN];
2655         uint8_t* dtarg;
2656         size_t dtarglen, newlen;
2657         struct packed_rrset_data* d;
2658
2659         /* get DNAME target name */
2660         if(dname->data->count < 1) return 0;
2661         if(dname->data->rr_len[0] < 3) return 0; /* at least rdatalen +1 */
2662         dtarg = dname->data->rr_data[0]+2;
2663         dtarglen = dname->data->rr_len[0]-2;
2664         if(sldns_read_uint16(dname->data->rr_data[0]) != dtarglen)
2665                 return 0; /* rdatalen in DNAME rdata is malformed */
2666         if(dname_valid(dtarg, dtarglen) != dtarglen)
2667                 return 0; /* DNAME RR has malformed rdata */
2668         if(qname_len == 0)
2669                 return 0; /* too short */
2670         if(qname_len <= node->namelen)
2671                 return 0; /* qname too short for dname removal */
2672
2673         /* synthesize a CNAME */
2674         newlen = synth_cname_buf(qname, qname_len, node->namelen,
2675                 dtarg, dtarglen, buf, sizeof(buf));
2676         if(newlen == 0) {
2677                 /* YXDOMAIN error */
2678                 *cname = NULL;
2679                 return 1;
2680         }
2681         *cname = (struct ub_packed_rrset_key*)regional_alloc(region,
2682                 sizeof(struct ub_packed_rrset_key));
2683         if(!*cname)
2684                 return 0; /* out of memory */
2685         memset(&(*cname)->entry, 0, sizeof((*cname)->entry));
2686         (*cname)->entry.key = (*cname);
2687         (*cname)->rk.type = htons(LDNS_RR_TYPE_CNAME);
2688         (*cname)->rk.rrset_class = htons(dclass);
2689         (*cname)->rk.flags = 0;
2690         (*cname)->rk.dname = regional_alloc_init(region, qname, qname_len);
2691         if(!(*cname)->rk.dname)
2692                 return 0; /* out of memory */
2693         (*cname)->rk.dname_len = qname_len;
2694         (*cname)->entry.hash = rrset_key_hash(&(*cname)->rk);
2695         d = (struct packed_rrset_data*)regional_alloc_zero(region,
2696                 sizeof(struct packed_rrset_data) + sizeof(size_t) +
2697                 sizeof(uint8_t*) + sizeof(time_t) + sizeof(uint16_t)
2698                 + newlen);
2699         if(!d)
2700                 return 0; /* out of memory */
2701         (*cname)->entry.data = d;
2702         d->ttl = 0; /* 0 for synthesized CNAME TTL */
2703         d->count = 1;
2704         d->rrsig_count = 0;
2705         d->trust = rrset_trust_ans_noAA;
2706         d->rr_len = (size_t*)((uint8_t*)d +
2707                 sizeof(struct packed_rrset_data));
2708         d->rr_len[0] = newlen + sizeof(uint16_t);
2709         packed_rrset_ptr_fixup(d);
2710         d->rr_ttl[0] = d->ttl;
2711         sldns_write_uint16(d->rr_data[0], newlen);
2712         memmove(d->rr_data[0] + sizeof(uint16_t), buf, newlen);
2713         return 1;
2714 }
2715
2716 /** add a synthesized CNAME to the answer section */
2717 static int
2718 add_synth_cname(struct auth_zone* z, uint8_t* qname, size_t qname_len,
2719         struct regional* region, struct dns_msg* msg, struct auth_data* dname,
2720         struct auth_rrset* rrset)
2721 {
2722         struct ub_packed_rrset_key* cname;
2723         /* synthesize a CNAME */
2724         if(!create_synth_cname(qname, qname_len, region, dname, rrset,
2725                 z->dclass, &cname)) {
2726                 /* out of memory */
2727                 return 0;
2728         }
2729         if(!cname) {
2730                 /* cname cannot be create because of YXDOMAIN */
2731                 msg->rep->flags |= LDNS_RCODE_YXDOMAIN;
2732                 return 1;
2733         }
2734         /* add cname to message */
2735         if(!msg_grow_array(region, msg))
2736                 return 0;
2737         msg->rep->rrsets[msg->rep->rrset_count] = cname;
2738         msg->rep->rrset_count++;
2739         msg->rep->an_numrrsets++;
2740         msg_ttl(msg);
2741         return 1;
2742 }
2743
2744 /** Change a dname to a different one, for wildcard namechange */
2745 static void
2746 az_change_dnames(struct dns_msg* msg, uint8_t* oldname, uint8_t* newname,
2747         size_t newlen, int an_only)
2748 {
2749         size_t i;
2750         size_t start = 0, end = msg->rep->rrset_count;
2751         if(!an_only) start = msg->rep->an_numrrsets;
2752         if(an_only) end = msg->rep->an_numrrsets;
2753         for(i=start; i<end; i++) {
2754                 /* allocated in region so we can change the ptrs */
2755                 if(query_dname_compare(msg->rep->rrsets[i]->rk.dname, oldname)
2756                         == 0) {
2757                         msg->rep->rrsets[i]->rk.dname = newname;
2758                         msg->rep->rrsets[i]->rk.dname_len = newlen;
2759                         msg->rep->rrsets[i]->entry.hash = rrset_key_hash(&msg->rep->rrsets[i]->rk);
2760                 }
2761         }
2762 }
2763
2764 /** find NSEC record covering the query */
2765 static struct auth_rrset*
2766 az_find_nsec_cover(struct auth_zone* z, struct auth_data** node)
2767 {
2768         uint8_t* nm = (*node)->name;
2769         size_t nmlen = (*node)->namelen;
2770         struct auth_rrset* rrset;
2771         /* find the NSEC for the smallest-or-equal node */
2772         /* if node == NULL, we did not find a smaller name.  But the zone
2773          * name is the smallest name and should have an NSEC. So there is
2774          * no NSEC to return (for a properly signed zone) */
2775         /* for empty nonterminals, the auth-data node should not exist,
2776          * and thus we don't need to go rbtree_previous here to find
2777          * a domain with an NSEC record */
2778         /* but there could be glue, and if this is node, then it has no NSEC.
2779          * Go up to find nonglue (previous) NSEC-holding nodes */
2780         while((rrset=az_domain_rrset(*node, LDNS_RR_TYPE_NSEC)) == NULL) {
2781                 if(dname_is_root(nm)) return NULL;
2782                 if(nmlen == z->namelen) return NULL;
2783                 dname_remove_label(&nm, &nmlen);
2784                 /* adjust *node for the nsec rrset to find in */
2785                 *node = az_find_name(z, nm, nmlen);
2786         }
2787         return rrset;
2788 }
2789
2790 /** Find NSEC and add for wildcard denial */
2791 static int
2792 az_nsec_wildcard_denial(struct auth_zone* z, struct regional* region,
2793         struct dns_msg* msg, uint8_t* cenm, size_t cenmlen)
2794 {
2795         struct query_info qinfo;
2796         int node_exact;
2797         struct auth_data* node;
2798         struct auth_rrset* nsec;
2799         uint8_t wc[LDNS_MAX_DOMAINLEN];
2800         if(cenmlen+2 > sizeof(wc))
2801                 return 0; /* result would be too long */
2802         wc[0] = 1; /* length of wildcard label */
2803         wc[1] = (uint8_t)'*'; /* wildcard label */
2804         memmove(wc+2, cenm, cenmlen);
2805
2806         /* we have '*.ce' in wc wildcard name buffer */
2807         /* get nsec cover for that */
2808         qinfo.qname = wc;
2809         qinfo.qname_len = cenmlen+2;
2810         qinfo.qtype = 0;
2811         qinfo.qclass = 0;
2812         az_find_domain(z, &qinfo, &node_exact, &node);
2813         if((nsec=az_find_nsec_cover(z, &node)) != NULL) {
2814                 if(!msg_add_rrset_ns(z, region, msg, node, nsec)) return 0;
2815         }
2816         return 1;
2817 }
2818
2819 /** Find the NSEC3PARAM rrset (if any) and if true you have the parameters */
2820 static int
2821 az_nsec3_param(struct auth_zone* z, int* algo, size_t* iter, uint8_t** salt,
2822         size_t* saltlen)
2823 {
2824         struct auth_data* apex;
2825         struct auth_rrset* param;
2826         size_t i;
2827         apex = az_find_name(z, z->name, z->namelen);
2828         if(!apex) return 0;
2829         param = az_domain_rrset(apex, LDNS_RR_TYPE_NSEC3PARAM);
2830         if(!param || param->data->count==0)
2831                 return 0; /* no RRset or no RRs in rrset */
2832         /* find out which NSEC3PARAM RR has supported parameters */
2833         /* skip unknown flags (dynamic signer is recalculating nsec3 chain) */
2834         for(i=0; i<param->data->count; i++) {
2835                 uint8_t* rdata = param->data->rr_data[i]+2;
2836                 size_t rdatalen = param->data->rr_len[i];
2837                 if(rdatalen < 2+5)
2838                         continue; /* too short */
2839                 if(!nsec3_hash_algo_size_supported((int)(rdata[0])))
2840                         continue; /* unsupported algo */
2841                 if(rdatalen < (size_t)(2+5+(size_t)rdata[4]))
2842                         continue; /* salt missing */
2843                 if((rdata[1]&NSEC3_UNKNOWN_FLAGS)!=0)
2844                         continue; /* unknown flags */
2845                 *algo = (int)(rdata[0]);
2846                 *iter = sldns_read_uint16(rdata+2);
2847                 *saltlen = rdata[4];
2848                 if(*saltlen == 0)
2849                         *salt = NULL;
2850                 else    *salt = rdata+5;
2851                 return 1;
2852         }
2853         /* no supported params */
2854         return 0;
2855 }
2856
2857 /** Hash a name with nsec3param into buffer, it has zone name appended.
2858  * return length of hash */
2859 static size_t
2860 az_nsec3_hash(uint8_t* buf, size_t buflen, uint8_t* nm, size_t nmlen,
2861         int algo, size_t iter, uint8_t* salt, size_t saltlen)
2862 {
2863         size_t hlen = nsec3_hash_algo_size_supported(algo);
2864         /* buffer has domain name, nsec3hash, and 256 is for max saltlen
2865          * (salt has 0-255 length) */
2866         unsigned char p[LDNS_MAX_DOMAINLEN+1+N3HASHBUFLEN+256];
2867         size_t i;
2868         if(nmlen+saltlen > sizeof(p) || hlen+saltlen > sizeof(p))
2869                 return 0;
2870         if(hlen > buflen)
2871                 return 0; /* somehow too large for destination buffer */
2872         /* hashfunc(name, salt) */
2873         memmove(p, nm, nmlen);
2874         query_dname_tolower(p);
2875         if(salt && saltlen > 0)
2876                 memmove(p+nmlen, salt, saltlen);
2877         (void)secalgo_nsec3_hash(algo, p, nmlen+saltlen, (unsigned char*)buf);
2878         for(i=0; i<iter; i++) {
2879                 /* hashfunc(hash, salt) */
2880                 memmove(p, buf, hlen);
2881                 if(salt && saltlen > 0)
2882                         memmove(p+hlen, salt, saltlen);
2883                 (void)secalgo_nsec3_hash(algo, p, hlen+saltlen,
2884                         (unsigned char*)buf);
2885         }
2886         return hlen;
2887 }
2888
2889 /** Hash name and return b32encoded hashname for lookup, zone name appended */
2890 static int
2891 az_nsec3_hashname(struct auth_zone* z, uint8_t* hashname, size_t* hashnmlen,
2892         uint8_t* nm, size_t nmlen, int algo, size_t iter, uint8_t* salt,
2893         size_t saltlen)
2894 {
2895         uint8_t hash[N3HASHBUFLEN];
2896         size_t hlen;
2897         int ret;
2898         hlen = az_nsec3_hash(hash, sizeof(hash), nm, nmlen, algo, iter,
2899                 salt, saltlen);
2900         if(!hlen) return 0;
2901         /* b32 encode */
2902         if(*hashnmlen < hlen*2+1+z->namelen) /* approx b32 as hexb16 */
2903                 return 0;
2904         ret = sldns_b32_ntop_extended_hex(hash, hlen, (char*)(hashname+1),
2905                 (*hashnmlen)-1);
2906         if(ret<1)
2907                 return 0;
2908         hashname[0] = (uint8_t)ret;
2909         ret++;
2910         if((*hashnmlen) - ret < z->namelen)
2911                 return 0;
2912         memmove(hashname+ret, z->name, z->namelen);
2913         *hashnmlen = z->namelen+(size_t)ret;
2914         return 1;
2915 }
2916
2917 /** Find the datanode that covers the nsec3hash-name */
2918 static struct auth_data*
2919 az_nsec3_findnode(struct auth_zone* z, uint8_t* hashnm, size_t hashnmlen)
2920 {
2921         struct query_info qinfo;
2922         struct auth_data* node;
2923         int node_exact;
2924         qinfo.qclass = 0;
2925         qinfo.qtype = 0;
2926         qinfo.qname = hashnm;
2927         qinfo.qname_len = hashnmlen;
2928         /* because canonical ordering and b32 nsec3 ordering are the same.
2929          * this is a good lookup to find the nsec3 name. */
2930         az_find_domain(z, &qinfo, &node_exact, &node);
2931         /* but we may have to skip non-nsec3 nodes */
2932         /* this may be a lot, the way to speed that up is to have a
2933          * separate nsec3 tree with nsec3 nodes */
2934         while(node && (rbnode_type*)node != RBTREE_NULL &&
2935                 !az_domain_rrset(node, LDNS_RR_TYPE_NSEC3)) {
2936                 node = (struct auth_data*)rbtree_previous(&node->node);
2937         }
2938         if((rbnode_type*)node == RBTREE_NULL)
2939                 node = NULL;
2940         return node;
2941 }
2942
2943 /** Find cover for hashed(nm, nmlen) (or NULL) */
2944 static struct auth_data*
2945 az_nsec3_find_cover(struct auth_zone* z, uint8_t* nm, size_t nmlen,
2946         int algo, size_t iter, uint8_t* salt, size_t saltlen)
2947 {
2948         struct auth_data* node;
2949         uint8_t hname[LDNS_MAX_DOMAINLEN];
2950         size_t hlen = sizeof(hname);
2951         if(!az_nsec3_hashname(z, hname, &hlen, nm, nmlen, algo, iter,
2952                 salt, saltlen))
2953                 return NULL;
2954         node = az_nsec3_findnode(z, hname, hlen);
2955         if(node)
2956                 return node;
2957         /* we did not find any, perhaps because the NSEC3 hash is before
2958          * the first hash, we have to find the 'last hash' in the zone */
2959         node = (struct auth_data*)rbtree_last(&z->data);
2960         while(node && (rbnode_type*)node != RBTREE_NULL &&
2961                 !az_domain_rrset(node, LDNS_RR_TYPE_NSEC3)) {
2962                 node = (struct auth_data*)rbtree_previous(&node->node);
2963         }
2964         if((rbnode_type*)node == RBTREE_NULL)
2965                 node = NULL;
2966         return node;
2967 }
2968
2969 /** Find exact match for hashed(nm, nmlen) NSEC3 record or NULL */
2970 static struct auth_data*
2971 az_nsec3_find_exact(struct auth_zone* z, uint8_t* nm, size_t nmlen,
2972         int algo, size_t iter, uint8_t* salt, size_t saltlen)
2973 {
2974         struct auth_data* node;
2975         uint8_t hname[LDNS_MAX_DOMAINLEN];
2976         size_t hlen = sizeof(hname);
2977         if(!az_nsec3_hashname(z, hname, &hlen, nm, nmlen, algo, iter,
2978                 salt, saltlen))
2979                 return NULL;
2980         node = az_find_name(z, hname, hlen);
2981         if(az_domain_rrset(node, LDNS_RR_TYPE_NSEC3))
2982                 return node;
2983         return NULL;
2984 }
2985
2986 /** Return nextcloser name (as a ref into the qname).  This is one label
2987  * more than the cenm (cename must be a suffix of qname) */
2988 static void
2989 az_nsec3_get_nextcloser(uint8_t* cenm, uint8_t* qname, size_t qname_len,
2990         uint8_t** nx, size_t* nxlen)
2991 {
2992         int celabs = dname_count_labels(cenm);
2993         int qlabs = dname_count_labels(qname);
2994         int strip = qlabs - celabs -1;
2995         log_assert(dname_strict_subdomain(qname, qlabs, cenm, celabs));
2996         *nx = qname;
2997         *nxlen = qname_len;
2998         if(strip>0)
2999                 dname_remove_labels(nx, nxlen, strip);
3000 }
3001
3002 /** Find the closest encloser that has exact NSEC3.
3003  * updated cenm to the new name. If it went up no-exact-ce is true. */
3004 static struct auth_data*
3005 az_nsec3_find_ce(struct auth_zone* z, uint8_t** cenm, size_t* cenmlen,
3006         int* no_exact_ce, int algo, size_t iter, uint8_t* salt, size_t saltlen)
3007 {
3008         struct auth_data* node;
3009         while((node = az_nsec3_find_exact(z, *cenm, *cenmlen,
3010                 algo, iter, salt, saltlen)) == NULL) {
3011                 if(*cenmlen == z->namelen) {
3012                         /* next step up would take us out of the zone. fail */
3013                         return NULL;
3014                 }
3015                 *no_exact_ce = 1;
3016                 dname_remove_label(cenm, cenmlen);
3017         }
3018         return node;
3019 }
3020
3021 /* Insert NSEC3 record in authority section, if NULL does nothing */
3022 static int
3023 az_nsec3_insert(struct auth_zone* z, struct regional* region,
3024         struct dns_msg* msg, struct auth_data* node)
3025 {
3026         struct auth_rrset* nsec3;
3027         if(!node) return 1; /* no node, skip this */
3028         nsec3 = az_domain_rrset(node, LDNS_RR_TYPE_NSEC3);
3029         if(!nsec3) return 1; /* if no nsec3 RR, skip it */
3030         if(!msg_add_rrset_ns(z, region, msg, node, nsec3)) return 0;
3031         return 1;
3032 }
3033
3034 /** add NSEC3 records to the zone for the nsec3 proof.
3035  * Specify with the flags with parts of the proof are required.
3036  * the ce is the exact matching name (for notype) but also delegation points.
3037  * qname is the one where the nextcloser name can be derived from.
3038  * If NSEC3 is not properly there (in the zone) nothing is added.
3039  * always enabled: include nsec3 proving about the Closest Encloser.
3040  *      that is an exact match that should exist for it.
3041  *      If that does not exist, a higher exact match + nxproof is enabled
3042  *      (for some sort of opt-out empty nonterminal cases).
3043  * nodataproof: search for exact match and include that instead.
3044  * ceproof: include ce proof NSEC3 (omitted for wildcard replies).
3045  * nxproof: include denial of the qname.
3046  * wcproof: include denial of wildcard (wildcard.ce).
3047  */
3048 static int
3049 az_add_nsec3_proof(struct auth_zone* z, struct regional* region,
3050         struct dns_msg* msg, uint8_t* cenm, size_t cenmlen, uint8_t* qname,
3051         size_t qname_len, int nodataproof, int ceproof, int nxproof,
3052         int wcproof)
3053 {
3054         int algo;
3055         size_t iter, saltlen;
3056         uint8_t* salt;
3057         int no_exact_ce = 0;
3058         struct auth_data* node;
3059
3060         /* find parameters of nsec3 proof */
3061         if(!az_nsec3_param(z, &algo, &iter, &salt, &saltlen))
3062                 return 1; /* no nsec3 */
3063         if(nodataproof) {
3064                 /* see if the node has a hash of itself for the nodata
3065                  * proof nsec3, this has to be an exact match nsec3. */
3066                 struct auth_data* match;
3067                 match = az_nsec3_find_exact(z, qname, qname_len, algo,
3068                         iter, salt, saltlen);
3069                 if(match) {
3070                         if(!az_nsec3_insert(z, region, msg, match))
3071                                 return 0;
3072                         /* only nodata NSEC3 needed, no CE or others. */
3073                         return 1;
3074                 }
3075         }
3076         /* find ce that has an NSEC3 */
3077         if(ceproof) {
3078                 node = az_nsec3_find_ce(z, &cenm, &cenmlen, &no_exact_ce,
3079                         algo, iter, salt, saltlen);
3080                 if(no_exact_ce) nxproof = 1;
3081                 if(!az_nsec3_insert(z, region, msg, node))
3082                         return 0;
3083         }
3084
3085         if(nxproof) {
3086                 uint8_t* nx;
3087                 size_t nxlen;
3088                 /* create nextcloser domain name */
3089                 az_nsec3_get_nextcloser(cenm, qname, qname_len, &nx, &nxlen);
3090                 /* find nsec3 that matches or covers it */
3091                 node = az_nsec3_find_cover(z, nx, nxlen, algo, iter, salt,
3092                         saltlen);
3093                 if(!az_nsec3_insert(z, region, msg, node))
3094                         return 0;
3095         }
3096         if(wcproof) {
3097                 /* create wildcard name *.ce */
3098                 uint8_t wc[LDNS_MAX_DOMAINLEN];
3099                 size_t wclen;
3100                 if(cenmlen+2 > sizeof(wc))
3101                         return 0; /* result would be too long */
3102                 wc[0] = 1; /* length of wildcard label */
3103                 wc[1] = (uint8_t)'*'; /* wildcard label */
3104                 memmove(wc+2, cenm, cenmlen);
3105                 wclen = cenmlen+2;
3106                 /* find nsec3 that matches or covers it */
3107                 node = az_nsec3_find_cover(z, wc, wclen, algo, iter, salt,
3108                         saltlen);
3109                 if(!az_nsec3_insert(z, region, msg, node))
3110                         return 0;
3111         }
3112         return 1;
3113 }
3114
3115 /** generate answer for positive answer */
3116 static int
3117 az_generate_positive_answer(struct auth_zone* z, struct regional* region,
3118         struct dns_msg* msg, struct auth_data* node, struct auth_rrset* rrset)
3119 {
3120         if(!msg_add_rrset_an(z, region, msg, node, rrset)) return 0;
3121         /* see if we want additional rrs */
3122         if(rrset->type == LDNS_RR_TYPE_MX) {
3123                 if(!az_add_additionals_from(z, region, msg, rrset, 2))
3124                         return 0;
3125         } else if(rrset->type == LDNS_RR_TYPE_SRV) {
3126                 if(!az_add_additionals_from(z, region, msg, rrset, 6))
3127                         return 0;
3128         } else if(rrset->type == LDNS_RR_TYPE_NS) {
3129                 if(!az_add_additionals_from(z, region, msg, rrset, 0))
3130                         return 0;
3131         }
3132         return 1;
3133 }
3134
3135 /** generate answer for type ANY answer */
3136 static int
3137 az_generate_any_answer(struct auth_zone* z, struct regional* region,
3138         struct dns_msg* msg, struct auth_data* node)
3139 {
3140         struct auth_rrset* rrset;
3141         int added = 0;
3142         /* add a couple (at least one) RRs */
3143         if((rrset=az_domain_rrset(node, LDNS_RR_TYPE_SOA)) != NULL) {
3144                 if(!msg_add_rrset_an(z, region, msg, node, rrset)) return 0;
3145                 added++;
3146         }
3147         if((rrset=az_domain_rrset(node, LDNS_RR_TYPE_MX)) != NULL) {
3148                 if(!msg_add_rrset_an(z, region, msg, node, rrset)) return 0;
3149                 added++;
3150         }
3151         if((rrset=az_domain_rrset(node, LDNS_RR_TYPE_A)) != NULL) {
3152                 if(!msg_add_rrset_an(z, region, msg, node, rrset)) return 0;
3153                 added++;
3154         }
3155         if((rrset=az_domain_rrset(node, LDNS_RR_TYPE_AAAA)) != NULL) {
3156                 if(!msg_add_rrset_an(z, region, msg, node, rrset)) return 0;
3157                 added++;
3158         }
3159         if(added == 0 && node && node->rrsets) {
3160                 if(!msg_add_rrset_an(z, region, msg, node,
3161                         node->rrsets)) return 0;
3162         }
3163         return 1;
3164 }
3165
3166 /** follow cname chain and add more data to the answer section */
3167 static int
3168 follow_cname_chain(struct auth_zone* z, uint16_t qtype,
3169         struct regional* region, struct dns_msg* msg,
3170         struct packed_rrset_data* d)
3171 {
3172         int maxchain = 0;
3173         /* see if we can add the target of the CNAME into the answer */
3174         while(maxchain++ < MAX_CNAME_CHAIN) {
3175                 struct auth_data* node;
3176                 struct auth_rrset* rrset;
3177                 size_t clen;
3178                 /* d has cname rdata */
3179                 if(d->count == 0) break; /* no CNAME */
3180                 if(d->rr_len[0] < 2+1) break; /* too small */
3181                 if((clen=dname_valid(d->rr_data[0]+2, d->rr_len[0]-2))==0)
3182                         break; /* malformed */
3183                 if(!dname_subdomain_c(d->rr_data[0]+2, z->name))
3184                         break; /* target out of zone */
3185                 if((node = az_find_name(z, d->rr_data[0]+2, clen))==NULL)
3186                         break; /* no such target name */
3187                 if((rrset=az_domain_rrset(node, qtype))!=NULL) {
3188                         /* done we found the target */
3189                         if(!msg_add_rrset_an(z, region, msg, node, rrset))
3190                                 return 0;
3191                         break;
3192                 }
3193                 if((rrset=az_domain_rrset(node, LDNS_RR_TYPE_CNAME))==NULL)
3194                         break; /* no further CNAME chain, notype */
3195                 if(!msg_add_rrset_an(z, region, msg, node, rrset)) return 0;
3196                 d = rrset->data;
3197         }
3198         return 1;
3199 }
3200
3201 /** generate answer for cname answer */
3202 static int
3203 az_generate_cname_answer(struct auth_zone* z, struct query_info* qinfo,
3204         struct regional* region, struct dns_msg* msg,
3205         struct auth_data* node, struct auth_rrset* rrset)
3206 {
3207         if(!msg_add_rrset_an(z, region, msg, node, rrset)) return 0;
3208         if(!rrset) return 1;
3209         if(!follow_cname_chain(z, qinfo->qtype, region, msg, rrset->data))
3210                 return 0;
3211         return 1;
3212 }
3213
3214 /** generate answer for notype answer */
3215 static int
3216 az_generate_notype_answer(struct auth_zone* z, struct regional* region,
3217         struct dns_msg* msg, struct auth_data* node)
3218 {
3219         struct auth_rrset* rrset;
3220         if(!az_add_negative_soa(z, region, msg)) return 0;
3221         /* DNSSEC denial NSEC */
3222         if((rrset=az_domain_rrset(node, LDNS_RR_TYPE_NSEC))!=NULL) {
3223                 if(!msg_add_rrset_ns(z, region, msg, node, rrset)) return 0;
3224         } else if(node) {
3225                 /* DNSSEC denial NSEC3 */
3226                 if(!az_add_nsec3_proof(z, region, msg, node->name,
3227                         node->namelen, msg->qinfo.qname,
3228                         msg->qinfo.qname_len, 1, 1, 0, 0))
3229                         return 0;
3230         }
3231         return 1;
3232 }
3233
3234 /** generate answer for referral answer */
3235 static int
3236 az_generate_referral_answer(struct auth_zone* z, struct regional* region,
3237         struct dns_msg* msg, struct auth_data* ce, struct auth_rrset* rrset)
3238 {
3239         struct auth_rrset* ds, *nsec;
3240         /* turn off AA flag, referral is nonAA because it leaves the zone */
3241         log_assert(ce);
3242         msg->rep->flags &= ~BIT_AA;
3243         if(!msg_add_rrset_ns(z, region, msg, ce, rrset)) return 0;
3244         /* add DS or deny it */
3245         if((ds=az_domain_rrset(ce, LDNS_RR_TYPE_DS))!=NULL) {
3246                 if(!msg_add_rrset_ns(z, region, msg, ce, ds)) return 0;
3247         } else {
3248                 /* deny the DS */
3249                 if((nsec=az_domain_rrset(ce, LDNS_RR_TYPE_NSEC))!=NULL) {
3250                         if(!msg_add_rrset_ns(z, region, msg, ce, nsec))
3251                                 return 0;
3252                 } else {
3253                         if(!az_add_nsec3_proof(z, region, msg, ce->name,
3254                                 ce->namelen, msg->qinfo.qname,
3255                                 msg->qinfo.qname_len, 1, 1, 0, 0))
3256                                 return 0;
3257                 }
3258         }
3259         /* add additional rrs for type NS */
3260         if(!az_add_additionals_from(z, region, msg, rrset, 0)) return 0;
3261         return 1;
3262 }
3263
3264 /** generate answer for DNAME answer */
3265 static int
3266 az_generate_dname_answer(struct auth_zone* z, struct query_info* qinfo,
3267         struct regional* region, struct dns_msg* msg, struct auth_data* ce,
3268         struct auth_rrset* rrset)
3269 {
3270         log_assert(ce);
3271         /* add the DNAME and then a CNAME */
3272         if(!msg_add_rrset_an(z, region, msg, ce, rrset)) return 0;
3273         if(!add_synth_cname(z, qinfo->qname, qinfo->qname_len, region,
3274                 msg, ce, rrset)) return 0;
3275         if(FLAGS_GET_RCODE(msg->rep->flags) == LDNS_RCODE_YXDOMAIN)
3276                 return 1;
3277         if(msg->rep->rrset_count == 0 ||
3278                 !msg->rep->rrsets[msg->rep->rrset_count-1])
3279                 return 0;
3280         if(!follow_cname_chain(z, qinfo->qtype, region, msg, 
3281                 (struct packed_rrset_data*)msg->rep->rrsets[
3282                 msg->rep->rrset_count-1]->entry.data))
3283                 return 0;
3284         return 1;
3285 }
3286
3287 /** generate answer for wildcard answer */
3288 static int
3289 az_generate_wildcard_answer(struct auth_zone* z, struct query_info* qinfo,
3290         struct regional* region, struct dns_msg* msg, struct auth_data* ce,
3291         struct auth_data* wildcard, struct auth_data* node)
3292 {
3293         struct auth_rrset* rrset, *nsec;
3294         int insert_ce = 0;
3295         if((rrset=az_domain_rrset(wildcard, qinfo->qtype)) != NULL) {
3296                 /* wildcard has type, add it */
3297                 if(!msg_add_rrset_an(z, region, msg, wildcard, rrset))
3298                         return 0;
3299                 az_change_dnames(msg, wildcard->name, msg->qinfo.qname,
3300                         msg->qinfo.qname_len, 1);
3301         } else if((rrset=az_domain_rrset(wildcard, LDNS_RR_TYPE_CNAME))!=NULL) {
3302                 /* wildcard has cname instead, do that */
3303                 if(!msg_add_rrset_an(z, region, msg, wildcard, rrset))
3304                         return 0;
3305                 az_change_dnames(msg, wildcard->name, msg->qinfo.qname,
3306                         msg->qinfo.qname_len, 1);
3307                 if(!follow_cname_chain(z, qinfo->qtype, region, msg,
3308                         rrset->data))
3309                         return 0;
3310         } else if(qinfo->qtype == LDNS_RR_TYPE_ANY && wildcard->rrsets) {
3311                 /* add ANY rrsets from wildcard node */
3312                 if(!az_generate_any_answer(z, region, msg, wildcard))
3313                         return 0;
3314                 az_change_dnames(msg, wildcard->name, msg->qinfo.qname,
3315                         msg->qinfo.qname_len, 1);
3316         } else {
3317                 /* wildcard has nodata, notype answer */
3318                 /* call other notype routine for dnssec notype denials */
3319                 if(!az_generate_notype_answer(z, region, msg, wildcard))
3320                         return 0;
3321                 /* because the notype, there is no positive data with an
3322                  * RRSIG that indicates the wildcard position.  Thus the
3323                  * wildcard qname denial needs to have a CE nsec3. */
3324                 insert_ce = 1;
3325         }
3326
3327         /* ce and node for dnssec denial of wildcard original name */
3328         if((nsec=az_find_nsec_cover(z, &node)) != NULL) {
3329                 if(!msg_add_rrset_ns(z, region, msg, node, nsec)) return 0;
3330         } else if(ce) {
3331                 uint8_t* wildup = wildcard->name;
3332                 size_t wilduplen= wildcard->namelen;
3333                 dname_remove_label(&wildup, &wilduplen);
3334                 if(!az_add_nsec3_proof(z, region, msg, wildup,
3335                         wilduplen, msg->qinfo.qname,
3336                         msg->qinfo.qname_len, 0, insert_ce, 1, 0))
3337                         return 0;
3338         }
3339
3340         /* fixup name of wildcard from *.zone to qname, use already allocated
3341          * pointer to msg qname */
3342         az_change_dnames(msg, wildcard->name, msg->qinfo.qname,
3343                 msg->qinfo.qname_len, 0);
3344         return 1;
3345 }
3346
3347 /** generate answer for nxdomain answer */
3348 static int
3349 az_generate_nxdomain_answer(struct auth_zone* z, struct regional* region,
3350         struct dns_msg* msg, struct auth_data* ce, struct auth_data* node)
3351 {
3352         struct auth_rrset* nsec;
3353         msg->rep->flags |= LDNS_RCODE_NXDOMAIN;
3354         if(!az_add_negative_soa(z, region, msg)) return 0;
3355         if((nsec=az_find_nsec_cover(z, &node)) != NULL) {
3356                 if(!msg_add_rrset_ns(z, region, msg, node, nsec)) return 0;
3357                 if(ce && !az_nsec_wildcard_denial(z, region, msg, ce->name,
3358                         ce->namelen)) return 0;
3359         } else if(ce) {
3360                 if(!az_add_nsec3_proof(z, region, msg, ce->name,
3361                         ce->namelen, msg->qinfo.qname,
3362                         msg->qinfo.qname_len, 0, 1, 1, 1))
3363                         return 0;
3364         }
3365         return 1;
3366 }
3367
3368 /** Create answers when an exact match exists for the domain name */
3369 static int
3370 az_generate_answer_with_node(struct auth_zone* z, struct query_info* qinfo,
3371         struct regional* region, struct dns_msg* msg, struct auth_data* node)
3372 {
3373         struct auth_rrset* rrset;
3374         /* positive answer, rrset we are looking for exists */
3375         if((rrset=az_domain_rrset(node, qinfo->qtype)) != NULL) {
3376                 return az_generate_positive_answer(z, region, msg, node, rrset);
3377         }
3378         /* CNAME? */
3379         if((rrset=az_domain_rrset(node, LDNS_RR_TYPE_CNAME)) != NULL) {
3380                 return az_generate_cname_answer(z, qinfo, region, msg,
3381                         node, rrset);
3382         }
3383         /* type ANY ? */
3384         if(qinfo->qtype == LDNS_RR_TYPE_ANY) {
3385                 return az_generate_any_answer(z, region, msg, node);
3386         }
3387         /* NOERROR/NODATA (no such type at domain name) */
3388         return az_generate_notype_answer(z, region, msg, node);
3389 }
3390
3391 /** Generate answer without an existing-node that we can use.
3392  * So it'll be a referral, DNAME or nxdomain */
3393 static int
3394 az_generate_answer_nonexistnode(struct auth_zone* z, struct query_info* qinfo,
3395         struct regional* region, struct dns_msg* msg, struct auth_data* ce,
3396         struct auth_rrset* rrset, struct auth_data* node)
3397 {
3398         struct auth_data* wildcard;
3399
3400         /* we do not have an exact matching name (that exists) */
3401         /* see if we have a NS or DNAME in the ce */
3402         if(ce && rrset && rrset->type == LDNS_RR_TYPE_NS) {
3403                 return az_generate_referral_answer(z, region, msg, ce, rrset);
3404         }
3405         if(ce && rrset && rrset->type == LDNS_RR_TYPE_DNAME) {
3406                 return az_generate_dname_answer(z, qinfo, region, msg, ce,
3407                         rrset);
3408         }
3409         /* if there is an empty nonterminal, wildcard and nxdomain don't
3410          * happen, it is a notype answer */
3411         if(az_empty_nonterminal(z, qinfo, node)) {
3412                 return az_generate_notype_answer(z, region, msg, node);
3413         }
3414         /* see if we have a wildcard under the ce */
3415         if((wildcard=az_find_wildcard(z, qinfo, ce)) != NULL) {
3416                 return az_generate_wildcard_answer(z, qinfo, region, msg,
3417                         ce, wildcard, node);
3418         }
3419         /* generate nxdomain answer */
3420         return az_generate_nxdomain_answer(z, region, msg, ce, node);
3421 }
3422
3423 /** Lookup answer in a zone. */
3424 static int
3425 auth_zone_generate_answer(struct auth_zone* z, struct query_info* qinfo,
3426         struct regional* region, struct dns_msg** msg, int* fallback)
3427 {
3428         struct auth_data* node, *ce;
3429         struct auth_rrset* rrset;
3430         int node_exact, node_exists;
3431         /* does the zone want fallback in case of failure? */
3432         *fallback = z->fallback_enabled;
3433         if(!(*msg=msg_create(region, qinfo))) return 0;
3434
3435         /* lookup if there is a matching domain name for the query */
3436         az_find_domain(z, qinfo, &node_exact, &node);
3437
3438         /* see if node exists for generating answers from (i.e. not glue and
3439          * obscured by NS or DNAME or NSEC3-only), and also return the
3440          * closest-encloser from that, closest node that should be used
3441          * to generate answers from that is above the query */
3442         node_exists = az_find_ce(z, qinfo, node, node_exact, &ce, &rrset);
3443
3444         if(verbosity >= VERB_ALGO) {
3445                 char zname[256], qname[256], nname[256], cename[256],
3446                         tpstr[32], rrstr[32];
3447                 sldns_wire2str_dname_buf(qinfo->qname, qinfo->qname_len, qname,
3448                         sizeof(qname));
3449                 sldns_wire2str_type_buf(qinfo->qtype, tpstr, sizeof(tpstr));
3450                 sldns_wire2str_dname_buf(z->name, z->namelen, zname,
3451                         sizeof(zname));
3452                 if(node)
3453                         sldns_wire2str_dname_buf(node->name, node->namelen,
3454                                 nname, sizeof(nname));
3455                 else    snprintf(nname, sizeof(nname), "NULL");
3456                 if(ce)
3457                         sldns_wire2str_dname_buf(ce->name, ce->namelen,
3458                                 cename, sizeof(cename));
3459                 else    snprintf(cename, sizeof(cename), "NULL");
3460                 if(rrset) sldns_wire2str_type_buf(rrset->type, rrstr,
3461                         sizeof(rrstr));
3462                 else    snprintf(rrstr, sizeof(rrstr), "NULL");
3463                 log_info("auth_zone %s query %s %s, domain %s %s %s, "
3464                         "ce %s, rrset %s", zname, qname, tpstr, nname,
3465                         (node_exact?"exact":"notexact"),
3466                         (node_exists?"exist":"notexist"), cename, rrstr);
3467         }
3468
3469         if(node_exists) {
3470                 /* the node is fine, generate answer from node */
3471                 return az_generate_answer_with_node(z, qinfo, region, *msg,
3472                         node);
3473         }
3474         return az_generate_answer_nonexistnode(z, qinfo, region, *msg,
3475                 ce, rrset, node);
3476 }
3477
3478 int auth_zones_lookup(struct auth_zones* az, struct query_info* qinfo,
3479         struct regional* region, struct dns_msg** msg, int* fallback,
3480         uint8_t* dp_nm, size_t dp_nmlen)
3481 {
3482         int r;
3483         struct auth_zone* z;
3484         /* find the zone that should contain the answer. */
3485         lock_rw_rdlock(&az->lock);
3486         z = auth_zone_find(az, dp_nm, dp_nmlen, qinfo->qclass);
3487         if(!z) {
3488                 lock_rw_unlock(&az->lock);
3489                 /* no auth zone, fallback to internet */
3490                 *fallback = 1;
3491                 return 0;
3492         }
3493         lock_rw_rdlock(&z->lock);
3494         lock_rw_unlock(&az->lock);
3495
3496         /* if not for upstream queries, fallback */
3497         if(!z->for_upstream) {
3498                 lock_rw_unlock(&z->lock);
3499                 *fallback = 1;
3500                 return 0;
3501         }
3502         if(z->zone_expired) {
3503                 *fallback = z->fallback_enabled;
3504                 lock_rw_unlock(&z->lock);
3505                 return 0;
3506         }
3507         /* see what answer that zone would generate */
3508         r = auth_zone_generate_answer(z, qinfo, region, msg, fallback);
3509         lock_rw_unlock(&z->lock);
3510         return r;
3511 }
3512
3513 /** encode auth answer */
3514 static void
3515 auth_answer_encode(struct query_info* qinfo, struct module_env* env,
3516         struct edns_data* edns, struct comm_reply* repinfo, sldns_buffer* buf,
3517         struct regional* temp, struct dns_msg* msg)
3518 {
3519         uint16_t udpsize;
3520         udpsize = edns->udp_size;
3521         edns->edns_version = EDNS_ADVERTISED_VERSION;
3522         edns->udp_size = EDNS_ADVERTISED_SIZE;
3523         edns->ext_rcode = 0;
3524         edns->bits &= EDNS_DO;
3525
3526         if(!inplace_cb_reply_local_call(env, qinfo, NULL, msg->rep,
3527                 (int)FLAGS_GET_RCODE(msg->rep->flags), edns, repinfo, temp, env->now_tv)
3528                 || !reply_info_answer_encode(qinfo, msg->rep,
3529                 *(uint16_t*)sldns_buffer_begin(buf),
3530                 sldns_buffer_read_u16_at(buf, 2),
3531                 buf, 0, 0, temp, udpsize, edns,
3532                 (int)(edns->bits&EDNS_DO), 0)) {
3533                 error_encode(buf, (LDNS_RCODE_SERVFAIL|BIT_AA), qinfo,
3534                         *(uint16_t*)sldns_buffer_begin(buf),
3535                         sldns_buffer_read_u16_at(buf, 2), edns);
3536         }
3537 }
3538
3539 /** encode auth error answer */
3540 static void
3541 auth_error_encode(struct query_info* qinfo, struct module_env* env,
3542         struct edns_data* edns, struct comm_reply* repinfo, sldns_buffer* buf,
3543         struct regional* temp, int rcode)
3544 {
3545         edns->edns_version = EDNS_ADVERTISED_VERSION;
3546         edns->udp_size = EDNS_ADVERTISED_SIZE;
3547         edns->ext_rcode = 0;
3548         edns->bits &= EDNS_DO;
3549
3550         if(!inplace_cb_reply_local_call(env, qinfo, NULL, NULL,
3551                 rcode, edns, repinfo, temp, env->now_tv))
3552                 edns->opt_list_inplace_cb_out = NULL;
3553         error_encode(buf, rcode|BIT_AA, qinfo,
3554                 *(uint16_t*)sldns_buffer_begin(buf),
3555                 sldns_buffer_read_u16_at(buf, 2), edns);
3556 }
3557
3558 int auth_zones_answer(struct auth_zones* az, struct module_env* env,
3559         struct query_info* qinfo, struct edns_data* edns,
3560         struct comm_reply* repinfo, struct sldns_buffer* buf, struct regional* temp)
3561 {
3562         struct dns_msg* msg = NULL;
3563         struct auth_zone* z;
3564         int r;
3565         int fallback = 0;
3566
3567         lock_rw_rdlock(&az->lock);
3568         if(!az->have_downstream) {
3569                 /* no downstream auth zones */
3570                 lock_rw_unlock(&az->lock);
3571                 return 0;
3572         }
3573         if(qinfo->qtype == LDNS_RR_TYPE_DS) {
3574                 uint8_t* delname = qinfo->qname;
3575                 size_t delnamelen = qinfo->qname_len;
3576                 dname_remove_label(&delname, &delnamelen);
3577                 z = auth_zones_find_zone(az, delname, delnamelen,
3578                         qinfo->qclass);
3579         } else {
3580                 z = auth_zones_find_zone(az, qinfo->qname, qinfo->qname_len,
3581                         qinfo->qclass);
3582         }
3583         if(!z) {
3584                 /* no zone above it */
3585                 lock_rw_unlock(&az->lock);
3586                 return 0;
3587         }
3588         lock_rw_rdlock(&z->lock);
3589         lock_rw_unlock(&az->lock);
3590         if(!z->for_downstream) {
3591                 lock_rw_unlock(&z->lock);
3592                 return 0;
3593         }
3594         if(z->zone_expired) {
3595                 if(z->fallback_enabled) {
3596                         lock_rw_unlock(&z->lock);
3597                         return 0;
3598                 }
3599                 lock_rw_unlock(&z->lock);
3600                 lock_rw_wrlock(&az->lock);
3601                 az->num_query_down++;
3602                 lock_rw_unlock(&az->lock);
3603                 auth_error_encode(qinfo, env, edns, repinfo, buf, temp,
3604                         LDNS_RCODE_SERVFAIL);
3605                 return 1;
3606         }
3607
3608         /* answer it from zone z */
3609         r = auth_zone_generate_answer(z, qinfo, temp, &msg, &fallback);
3610         lock_rw_unlock(&z->lock);
3611         if(!r && fallback) {
3612                 /* fallback to regular answering (recursive) */
3613                 return 0;
3614         }
3615         lock_rw_wrlock(&az->lock);
3616         az->num_query_down++;
3617         lock_rw_unlock(&az->lock);
3618
3619         /* encode answer */
3620         if(!r)
3621                 auth_error_encode(qinfo, env, edns, repinfo, buf, temp,
3622                         LDNS_RCODE_SERVFAIL);
3623         else    auth_answer_encode(qinfo, env, edns, repinfo, buf, temp, msg);
3624
3625         return 1;
3626 }
3627
3628 int auth_zones_can_fallback(struct auth_zones* az, uint8_t* nm, size_t nmlen,
3629         uint16_t dclass)
3630 {
3631         int r;
3632         struct auth_zone* z;
3633         lock_rw_rdlock(&az->lock);
3634         z = auth_zone_find(az, nm, nmlen, dclass);
3635         if(!z) {
3636                 lock_rw_unlock(&az->lock);
3637                 /* no such auth zone, fallback */
3638                 return 1;
3639         }
3640         lock_rw_rdlock(&z->lock);
3641         lock_rw_unlock(&az->lock);
3642         r = z->fallback_enabled || (!z->for_upstream);
3643         lock_rw_unlock(&z->lock);
3644         return r;
3645 }
3646
3647 int
3648 auth_zone_parse_notify_serial(sldns_buffer* pkt, uint32_t *serial)
3649 {
3650         struct query_info q;
3651         uint16_t rdlen;
3652         memset(&q, 0, sizeof(q));
3653         sldns_buffer_set_position(pkt, 0);
3654         if(!query_info_parse(&q, pkt)) return 0;
3655         if(LDNS_ANCOUNT(sldns_buffer_begin(pkt)) == 0) return 0;
3656         /* skip name of RR in answer section */
3657         if(sldns_buffer_remaining(pkt) < 1) return 0;
3658         if(pkt_dname_len(pkt) == 0) return 0;
3659         /* check type */
3660         if(sldns_buffer_remaining(pkt) < 10 /* type,class,ttl,rdatalen*/)
3661                 return 0;
3662         if(sldns_buffer_read_u16(pkt) != LDNS_RR_TYPE_SOA) return 0;
3663         sldns_buffer_skip(pkt, 2); /* class */
3664         sldns_buffer_skip(pkt, 4); /* ttl */
3665         rdlen = sldns_buffer_read_u16(pkt); /* rdatalen */
3666         if(sldns_buffer_remaining(pkt) < rdlen) return 0;
3667         if(rdlen < 22) return 0; /* bad soa length */
3668         sldns_buffer_skip(pkt, (ssize_t)(rdlen-20));
3669         *serial = sldns_buffer_read_u32(pkt);
3670         /* return true when has serial in answer section */
3671         return 1;
3672 }
3673
3674 /** see if addr appears in the list */
3675 static int
3676 addr_in_list(struct auth_addr* list, struct sockaddr_storage* addr,
3677         socklen_t addrlen)
3678 {
3679         struct auth_addr* p;
3680         for(p=list; p; p=p->next) {
3681                 if(sockaddr_cmp_addr(addr, addrlen, &p->addr, p->addrlen)==0)
3682                         return 1;
3683         }
3684         return 0;
3685 }
3686
3687 /** check if an address matches a master specification (or one of its
3688  * addresses in the addr list) */
3689 static int
3690 addr_matches_master(struct auth_master* master, struct sockaddr_storage* addr,
3691         socklen_t addrlen, struct auth_master** fromhost)
3692 {
3693         struct sockaddr_storage a;
3694         socklen_t alen = 0;
3695         int net = 0;
3696         if(addr_in_list(master->list, addr, addrlen)) {
3697                 *fromhost = master;
3698                 return 1;       
3699         }
3700         /* compare address (but not port number, that is the destination
3701          * port of the master, the port number of the received notify is
3702          * allowed to by any port on that master) */
3703         if(extstrtoaddr(master->host, &a, &alen, UNBOUND_DNS_PORT) &&
3704                 sockaddr_cmp_addr(addr, addrlen, &a, alen)==0) {
3705                 *fromhost = master;
3706                 return 1;
3707         }
3708         /* prefixes, addr/len, like 10.0.0.0/8 */
3709         /* not http and has a / and there is one / */
3710         if(master->allow_notify && !master->http &&
3711                 strchr(master->host, '/') != NULL &&
3712                 strchr(master->host, '/') == strrchr(master->host, '/') &&
3713                 netblockstrtoaddr(master->host, UNBOUND_DNS_PORT, &a, &alen,
3714                 &net) && alen == addrlen) {
3715                 if(addr_in_common(addr, (addr_is_ip6(addr, addrlen)?128:32),
3716                         &a, net, alen) >= net) {
3717                         *fromhost = NULL; /* prefix does not have destination
3718                                 to send the probe or transfer with */
3719                         return 1; /* matches the netblock */
3720                 }
3721         }
3722         return 0;
3723 }
3724
3725 /** check access list for notifies */
3726 static int
3727 az_xfr_allowed_notify(struct auth_xfer* xfr, struct sockaddr_storage* addr,
3728         socklen_t addrlen, struct auth_master** fromhost)
3729 {
3730         struct auth_master* p;
3731         for(p=xfr->allow_notify_list; p; p=p->next) {
3732                 if(addr_matches_master(p, addr, addrlen, fromhost)) {
3733                         return 1;
3734                 }
3735         }
3736         return 0;
3737 }
3738
3739 /** see if the serial means the zone has to be updated, i.e. the serial
3740  * is newer than the zone serial, or we have no zone */
3741 static int
3742 xfr_serial_means_update(struct auth_xfer* xfr, uint32_t serial)
3743 {
3744         if(!xfr->have_zone)
3745                 return 1; /* no zone, anything is better */
3746         if(xfr->zone_expired)
3747                 return 1; /* expired, the sent serial is better than expired
3748                         data */
3749         if(compare_serial(xfr->serial, serial) < 0)
3750                 return 1; /* our serial is smaller than the sent serial,
3751                         the data is newer, fetch it */
3752         return 0;
3753 }
3754
3755 /** note notify serial, updates the notify information in the xfr struct */
3756 static void
3757 xfr_note_notify_serial(struct auth_xfer* xfr, int has_serial, uint32_t serial)
3758 {
3759         if(xfr->notify_received && xfr->notify_has_serial && has_serial) {
3760                 /* see if this serial is newer */
3761                 if(compare_serial(xfr->notify_serial, serial) < 0)
3762                         xfr->notify_serial = serial;
3763         } else if(xfr->notify_received && xfr->notify_has_serial &&
3764                 !has_serial) {
3765                 /* remove serial, we have notify without serial */
3766                 xfr->notify_has_serial = 0;
3767                 xfr->notify_serial = 0;
3768         } else if(xfr->notify_received && !xfr->notify_has_serial) {
3769                 /* we already have notify without serial, keep it
3770                  * that way; no serial check when current operation
3771                  * is done */
3772         } else {
3773                 xfr->notify_received = 1;
3774                 xfr->notify_has_serial = has_serial;
3775                 xfr->notify_serial = serial;
3776         }
3777 }
3778
3779 /** process a notify serial, start new probe or note serial. xfr is locked */
3780 static void
3781 xfr_process_notify(struct auth_xfer* xfr, struct module_env* env,
3782         int has_serial, uint32_t serial, struct auth_master* fromhost)
3783 {
3784         /* if the serial of notify is older than we have, don't fetch
3785          * a zone, we already have it */
3786         if(has_serial && !xfr_serial_means_update(xfr, serial)) {
3787                 lock_basic_unlock(&xfr->lock);
3788                 return;
3789         }
3790         /* start new probe with this addr src, or note serial */
3791         if(!xfr_start_probe(xfr, env, fromhost)) {
3792                 /* not started because already in progress, note the serial */
3793                 xfr_note_notify_serial(xfr, has_serial, serial);
3794                 lock_basic_unlock(&xfr->lock);
3795         }
3796         /* successful end of start_probe unlocked xfr->lock */
3797 }
3798
3799 int auth_zones_notify(struct auth_zones* az, struct module_env* env,
3800         uint8_t* nm, size_t nmlen, uint16_t dclass,
3801         struct sockaddr_storage* addr, socklen_t addrlen, int has_serial,
3802         uint32_t serial, int* refused)
3803 {
3804         struct auth_xfer* xfr;
3805         struct auth_master* fromhost = NULL;
3806         /* see which zone this is */
3807         lock_rw_rdlock(&az->lock);
3808         xfr = auth_xfer_find(az, nm, nmlen, dclass);
3809         if(!xfr) {
3810                 lock_rw_unlock(&az->lock);
3811                 /* no such zone, refuse the notify */
3812                 *refused = 1;
3813                 return 0;
3814         }
3815         lock_basic_lock(&xfr->lock);
3816         lock_rw_unlock(&az->lock);
3817         
3818         /* check access list for notifies */
3819         if(!az_xfr_allowed_notify(xfr, addr, addrlen, &fromhost)) {
3820                 lock_basic_unlock(&xfr->lock);
3821                 /* notify not allowed, refuse the notify */
3822                 *refused = 1;
3823                 return 0;
3824         }
3825
3826         /* process the notify */
3827         xfr_process_notify(xfr, env, has_serial, serial, fromhost);
3828         return 1;
3829 }
3830
3831 int auth_zones_startprobesequence(struct auth_zones* az,
3832         struct module_env* env, uint8_t* nm, size_t nmlen, uint16_t dclass)
3833 {
3834         struct auth_xfer* xfr;
3835         lock_rw_rdlock(&az->lock);
3836         xfr = auth_xfer_find(az, nm, nmlen, dclass);
3837         if(!xfr) {
3838                 lock_rw_unlock(&az->lock);
3839                 return 0;
3840         }
3841         lock_basic_lock(&xfr->lock);
3842         lock_rw_unlock(&az->lock);
3843
3844         xfr_process_notify(xfr, env, 0, 0, NULL);
3845         return 1;
3846 }
3847
3848 /** set a zone expired */
3849 static void
3850 auth_xfer_set_expired(struct auth_xfer* xfr, struct module_env* env,
3851         int expired)
3852 {
3853         struct auth_zone* z;
3854
3855         /* expire xfr */
3856         lock_basic_lock(&xfr->lock);
3857         xfr->zone_expired = expired;
3858         lock_basic_unlock(&xfr->lock);
3859
3860         /* find auth_zone */
3861         lock_rw_rdlock(&env->auth_zones->lock);
3862         z = auth_zone_find(env->auth_zones, xfr->name, xfr->namelen,
3863                 xfr->dclass);
3864         if(!z) {
3865                 lock_rw_unlock(&env->auth_zones->lock);
3866                 return;
3867         }
3868         lock_rw_wrlock(&z->lock);
3869         lock_rw_unlock(&env->auth_zones->lock);
3870
3871         /* expire auth_zone */
3872         z->zone_expired = expired;
3873         lock_rw_unlock(&z->lock);
3874 }
3875
3876 /** find master (from notify or probe) in list of masters */
3877 static struct auth_master*
3878 find_master_by_host(struct auth_master* list, char* host)
3879 {
3880         struct auth_master* p;
3881         for(p=list; p; p=p->next) {
3882                 if(strcmp(p->host, host) == 0)
3883                         return p;
3884         }
3885         return NULL;
3886 }
3887
3888 /** delete the looked up auth_addrs for all the masters in the list */
3889 static void
3890 xfr_masterlist_free_addrs(struct auth_master* list)
3891 {
3892         struct auth_master* m;
3893         for(m=list; m; m=m->next) {
3894                 if(m->list) {
3895                         auth_free_master_addrs(m->list);
3896                         m->list = NULL;
3897                 }
3898         }
3899 }
3900
3901 /** copy a list of auth_addrs */
3902 static struct auth_addr*
3903 auth_addr_list_copy(struct auth_addr* source)
3904 {
3905         struct auth_addr* list = NULL, *last = NULL;
3906         struct auth_addr* p;
3907         for(p=source; p; p=p->next) {
3908                 struct auth_addr* a = (struct auth_addr*)memdup(p, sizeof(*p));
3909                 if(!a) {
3910                         log_err("malloc failure");
3911                         auth_free_master_addrs(list);
3912                         return NULL;
3913                 }
3914                 a->next = NULL;
3915                 if(last) last->next = a;
3916                 if(!list) list = a;
3917                 last = a;
3918         }
3919         return list;
3920 }
3921
3922 /** copy a master to a new structure, NULL on alloc failure */
3923 static struct auth_master*
3924 auth_master_copy(struct auth_master* o)
3925 {
3926         struct auth_master* m;
3927         if(!o) return NULL;
3928         m = (struct auth_master*)memdup(o, sizeof(*o));
3929         if(!m) {
3930                 log_err("malloc failure");
3931                 return NULL;
3932         }
3933         m->next = NULL;
3934         if(m->host) {
3935                 m->host = strdup(m->host);
3936                 if(!m->host) {
3937                         free(m);
3938                         log_err("malloc failure");
3939                         return NULL;
3940                 }
3941         }
3942         if(m->file) {
3943                 m->file = strdup(m->file);
3944                 if(!m->file) {
3945                         free(m->host);
3946                         free(m);
3947                         log_err("malloc failure");
3948                         return NULL;
3949                 }
3950         }
3951         if(m->list) {
3952                 m->list = auth_addr_list_copy(m->list);
3953                 if(!m->list) {
3954                         free(m->file);
3955                         free(m->host);
3956                         free(m);
3957                         return NULL;
3958                 }
3959         }
3960         return m;
3961 }
3962
3963 /** copy the master addresses from the task_probe lookups to the allow_notify
3964  * list of masters */
3965 static void
3966 probe_copy_masters_for_allow_notify(struct auth_xfer* xfr)
3967 {
3968         struct auth_master* list = NULL, *last = NULL;
3969         struct auth_master* p;
3970         /* build up new list with copies */
3971         for(p = xfr->task_transfer->masters; p; p=p->next) {
3972                 struct auth_master* m = auth_master_copy(p);
3973                 if(!m) {
3974                         auth_free_masters(list);
3975                         /* failed because of malloc failure, use old list */
3976                         return;
3977                 }
3978                 m->next = NULL;
3979                 if(last) last->next = m;
3980                 if(!list) list = m;
3981                 last = m;
3982         }
3983         /* success, replace list */
3984         auth_free_masters(xfr->allow_notify_list);
3985         xfr->allow_notify_list = list;
3986 }
3987
3988 /** start the lookups for task_transfer */
3989 static void
3990 xfr_transfer_start_lookups(struct auth_xfer* xfr)
3991 {
3992         /* delete all the looked up addresses in the list */
3993         xfr->task_transfer->scan_addr = NULL;
3994         xfr_masterlist_free_addrs(xfr->task_transfer->masters);
3995
3996         /* start lookup at the first master */
3997         xfr->task_transfer->lookup_target = xfr->task_transfer->masters;
3998         xfr->task_transfer->lookup_aaaa = 0;
3999 }
4000
4001 /** move to the next lookup of hostname for task_transfer */
4002 static void
4003 xfr_transfer_move_to_next_lookup(struct auth_xfer* xfr, struct module_env* env)
4004 {
4005         if(!xfr->task_transfer->lookup_target)
4006                 return; /* already at end of list */
4007         if(!xfr->task_transfer->lookup_aaaa && env->cfg->do_ip6) {
4008                 /* move to lookup AAAA */
4009                 xfr->task_transfer->lookup_aaaa = 1;
4010                 return;
4011         }
4012         xfr->task_transfer->lookup_target = 
4013                 xfr->task_transfer->lookup_target->next;
4014         xfr->task_transfer->lookup_aaaa = 0;
4015         if(!env->cfg->do_ip4 && xfr->task_transfer->lookup_target!=NULL)
4016                 xfr->task_transfer->lookup_aaaa = 1;
4017 }
4018
4019 /** start the lookups for task_probe */
4020 static void
4021 xfr_probe_start_lookups(struct auth_xfer* xfr)
4022 {
4023         /* delete all the looked up addresses in the list */
4024         xfr->task_probe->scan_addr = NULL;
4025         xfr_masterlist_free_addrs(xfr->task_probe->masters);
4026
4027         /* start lookup at the first master */
4028         xfr->task_probe->lookup_target = xfr->task_probe->masters;
4029         xfr->task_probe->lookup_aaaa = 0;
4030 }
4031
4032 /** move to the next lookup of hostname for task_probe */
4033 static void
4034 xfr_probe_move_to_next_lookup(struct auth_xfer* xfr, struct module_env* env)
4035 {
4036         if(!xfr->task_probe->lookup_target)
4037                 return; /* already at end of list */
4038         if(!xfr->task_probe->lookup_aaaa && env->cfg->do_ip6) {
4039                 /* move to lookup AAAA */
4040                 xfr->task_probe->lookup_aaaa = 1;
4041                 return;
4042         }
4043         xfr->task_probe->lookup_target = xfr->task_probe->lookup_target->next;
4044         xfr->task_probe->lookup_aaaa = 0;
4045         if(!env->cfg->do_ip4 && xfr->task_probe->lookup_target!=NULL)
4046                 xfr->task_probe->lookup_aaaa = 1;
4047 }
4048
4049 /** start the iteration of the task_transfer list of masters */
4050 static void
4051 xfr_transfer_start_list(struct auth_xfer* xfr, struct auth_master* spec) 
4052 {
4053         if(spec) {
4054                 xfr->task_transfer->scan_specific = find_master_by_host(
4055                         xfr->task_transfer->masters, spec->host);
4056                 if(xfr->task_transfer->scan_specific) {
4057                         xfr->task_transfer->scan_target = NULL;
4058                         xfr->task_transfer->scan_addr = NULL;
4059                         if(xfr->task_transfer->scan_specific->list)
4060                                 xfr->task_transfer->scan_addr =
4061                                         xfr->task_transfer->scan_specific->list;
4062                         return;
4063                 }
4064         }
4065         /* no specific (notified) host to scan */
4066         xfr->task_transfer->scan_specific = NULL;
4067         xfr->task_transfer->scan_addr = NULL;
4068         /* pick up first scan target */
4069         xfr->task_transfer->scan_target = xfr->task_transfer->masters;
4070         if(xfr->task_transfer->scan_target && xfr->task_transfer->
4071                 scan_target->list)
4072                 xfr->task_transfer->scan_addr =
4073                         xfr->task_transfer->scan_target->list;
4074 }
4075
4076 /** start the iteration of the task_probe list of masters */
4077 static void
4078 xfr_probe_start_list(struct auth_xfer* xfr, struct auth_master* spec) 
4079 {
4080         if(spec) {
4081                 xfr->task_probe->scan_specific = find_master_by_host(
4082                         xfr->task_probe->masters, spec->host);
4083                 if(xfr->task_probe->scan_specific) {
4084                         xfr->task_probe->scan_target = NULL;
4085                         xfr->task_probe->scan_addr = NULL;
4086                         if(xfr->task_probe->scan_specific->list)
4087                                 xfr->task_probe->scan_addr =
4088                                         xfr->task_probe->scan_specific->list;
4089                         return;
4090                 }
4091         }
4092         /* no specific (notified) host to scan */
4093         xfr->task_probe->scan_specific = NULL;
4094         xfr->task_probe->scan_addr = NULL;
4095         /* pick up first scan target */
4096         xfr->task_probe->scan_target = xfr->task_probe->masters;
4097         if(xfr->task_probe->scan_target && xfr->task_probe->scan_target->list)
4098                 xfr->task_probe->scan_addr =
4099                         xfr->task_probe->scan_target->list;
4100 }
4101
4102 /** pick up the master that is being scanned right now, task_transfer */
4103 static struct auth_master*
4104 xfr_transfer_current_master(struct auth_xfer* xfr)
4105 {
4106         if(xfr->task_transfer->scan_specific)
4107                 return xfr->task_transfer->scan_specific;
4108         return xfr->task_transfer->scan_target;
4109 }
4110
4111 /** pick up the master that is being scanned right now, task_probe */
4112 static struct auth_master*
4113 xfr_probe_current_master(struct auth_xfer* xfr)
4114 {
4115         if(xfr->task_probe->scan_specific)
4116                 return xfr->task_probe->scan_specific;
4117         return xfr->task_probe->scan_target;
4118 }
4119
4120 /** true if at end of list, task_transfer */
4121 static int
4122 xfr_transfer_end_of_list(struct auth_xfer* xfr)
4123 {
4124         return !xfr->task_transfer->scan_specific &&
4125                 !xfr->task_transfer->scan_target;
4126 }
4127
4128 /** true if at end of list, task_probe */
4129 static int
4130 xfr_probe_end_of_list(struct auth_xfer* xfr)
4131 {
4132         return !xfr->task_probe->scan_specific && !xfr->task_probe->scan_target;
4133 }
4134
4135 /** move to next master in list, task_transfer */
4136 static void
4137 xfr_transfer_nextmaster(struct auth_xfer* xfr)
4138 {
4139         if(!xfr->task_transfer->scan_specific &&
4140                 !xfr->task_transfer->scan_target)
4141                 return;
4142         if(xfr->task_transfer->scan_addr) {
4143                 xfr->task_transfer->scan_addr =
4144                         xfr->task_transfer->scan_addr->next;
4145                 if(xfr->task_transfer->scan_addr)
4146                         return;
4147         }
4148         if(xfr->task_transfer->scan_specific) {
4149                 xfr->task_transfer->scan_specific = NULL;
4150                 xfr->task_transfer->scan_target = xfr->task_transfer->masters;
4151                 if(xfr->task_transfer->scan_target && xfr->task_transfer->
4152                         scan_target->list)
4153                         xfr->task_transfer->scan_addr =
4154                                 xfr->task_transfer->scan_target->list;
4155                 return;
4156         }
4157         if(!xfr->task_transfer->scan_target)
4158                 return;
4159         xfr->task_transfer->scan_target = xfr->task_transfer->scan_target->next;
4160         if(xfr->task_transfer->scan_target && xfr->task_transfer->
4161                 scan_target->list)
4162                 xfr->task_transfer->scan_addr =
4163                         xfr->task_transfer->scan_target->list;
4164         return;
4165 }
4166
4167 /** move to next master in list, task_probe */
4168 static void
4169 xfr_probe_nextmaster(struct auth_xfer* xfr)
4170 {
4171         if(!xfr->task_probe->scan_specific && !xfr->task_probe->scan_target)
4172                 return;
4173         if(xfr->task_probe->scan_addr) {
4174                 xfr->task_probe->scan_addr = xfr->task_probe->scan_addr->next;
4175                 if(xfr->task_probe->scan_addr)
4176                         return;
4177         }
4178         if(xfr->task_probe->scan_specific) {
4179                 xfr->task_probe->scan_specific = NULL;
4180                 xfr->task_probe->scan_target = xfr->task_probe->masters;
4181                 if(xfr->task_probe->scan_target && xfr->task_probe->
4182                         scan_target->list)
4183                         xfr->task_probe->scan_addr =
4184                                 xfr->task_probe->scan_target->list;
4185                 return;
4186         }
4187         if(!xfr->task_probe->scan_target)
4188                 return;
4189         xfr->task_probe->scan_target = xfr->task_probe->scan_target->next;
4190         if(xfr->task_probe->scan_target && xfr->task_probe->
4191                 scan_target->list)
4192                 xfr->task_probe->scan_addr =
4193                         xfr->task_probe->scan_target->list;
4194         return;
4195 }
4196
4197 /** create SOA probe packet for xfr */
4198 static void
4199 xfr_create_soa_probe_packet(struct auth_xfer* xfr, sldns_buffer* buf, 
4200         uint16_t id)
4201 {
4202         struct query_info qinfo;
4203
4204         memset(&qinfo, 0, sizeof(qinfo));
4205         qinfo.qname = xfr->name;
4206         qinfo.qname_len = xfr->namelen;
4207         qinfo.qtype = LDNS_RR_TYPE_SOA;
4208         qinfo.qclass = xfr->dclass;
4209         qinfo_query_encode(buf, &qinfo);
4210         sldns_buffer_write_u16_at(buf, 0, id);
4211 }
4212
4213 /** create IXFR/AXFR packet for xfr */
4214 static void
4215 xfr_create_ixfr_packet(struct auth_xfer* xfr, sldns_buffer* buf, uint16_t id,
4216         struct auth_master* master)
4217 {
4218         struct query_info qinfo;
4219         uint32_t serial;
4220         int have_zone;
4221         have_zone = xfr->have_zone;
4222         serial = xfr->serial;
4223
4224         memset(&qinfo, 0, sizeof(qinfo));
4225         qinfo.qname = xfr->name;
4226         qinfo.qname_len = xfr->namelen;
4227         xfr->task_transfer->got_xfr_serial = 0;
4228         xfr->task_transfer->rr_scan_num = 0;
4229         xfr->task_transfer->incoming_xfr_serial = 0;
4230         xfr->task_transfer->on_ixfr_is_axfr = 0;
4231         xfr->task_transfer->on_ixfr = 1;
4232         qinfo.qtype = LDNS_RR_TYPE_IXFR;
4233         if(!have_zone || xfr->task_transfer->ixfr_fail || !master->ixfr) {
4234                 qinfo.qtype = LDNS_RR_TYPE_AXFR;
4235                 xfr->task_transfer->ixfr_fail = 0;
4236                 xfr->task_transfer->on_ixfr = 0;
4237         }
4238
4239         qinfo.qclass = xfr->dclass;
4240         qinfo_query_encode(buf, &qinfo);
4241         sldns_buffer_write_u16_at(buf, 0, id);
4242
4243         /* append serial for IXFR */
4244         if(qinfo.qtype == LDNS_RR_TYPE_IXFR) {
4245                 size_t end = sldns_buffer_limit(buf);
4246                 sldns_buffer_clear(buf);
4247                 sldns_buffer_set_position(buf, end);
4248                 /* auth section count 1 */
4249                 sldns_buffer_write_u16_at(buf, LDNS_NSCOUNT_OFF, 1);
4250                 /* write SOA */
4251                 sldns_buffer_write_u8(buf, 0xC0); /* compressed ptr to qname */
4252                 sldns_buffer_write_u8(buf, 0x0C);
4253                 sldns_buffer_write_u16(buf, LDNS_RR_TYPE_SOA);
4254                 sldns_buffer_write_u16(buf, qinfo.qclass);
4255                 sldns_buffer_write_u32(buf, 0); /* ttl */
4256                 sldns_buffer_write_u16(buf, 22); /* rdata length */
4257                 sldns_buffer_write_u8(buf, 0); /* . */
4258                 sldns_buffer_write_u8(buf, 0); /* . */
4259                 sldns_buffer_write_u32(buf, serial); /* serial */
4260                 sldns_buffer_write_u32(buf, 0); /* refresh */
4261                 sldns_buffer_write_u32(buf, 0); /* retry */
4262                 sldns_buffer_write_u32(buf, 0); /* expire */
4263                 sldns_buffer_write_u32(buf, 0); /* minimum */
4264                 sldns_buffer_flip(buf);
4265         }
4266 }
4267
4268 /** check if returned packet is OK */
4269 static int
4270 check_packet_ok(sldns_buffer* pkt, uint16_t qtype, struct auth_xfer* xfr,
4271         uint32_t* serial)
4272 {
4273         /* parse to see if packet worked, valid reply */
4274
4275         /* check serial number of SOA */
4276         if(sldns_buffer_limit(pkt) < LDNS_HEADER_SIZE)
4277                 return 0;
4278
4279         /* check ID */
4280         if(LDNS_ID_WIRE(sldns_buffer_begin(pkt)) != xfr->task_probe->id)
4281                 return 0;
4282
4283         /* check flag bits and rcode */
4284         if(!LDNS_QR_WIRE(sldns_buffer_begin(pkt)))
4285                 return 0;
4286         if(LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt)) != LDNS_PACKET_QUERY)
4287                 return 0;
4288         if(LDNS_RCODE_WIRE(sldns_buffer_begin(pkt)) != LDNS_RCODE_NOERROR)
4289                 return 0;
4290
4291         /* check qname */
4292         if(LDNS_QDCOUNT(sldns_buffer_begin(pkt)) != 1)
4293                 return 0;
4294         sldns_buffer_skip(pkt, LDNS_HEADER_SIZE);
4295         if(sldns_buffer_remaining(pkt) < xfr->namelen)
4296                 return 0;
4297         if(query_dname_compare(sldns_buffer_current(pkt), xfr->name) != 0)
4298                 return 0;
4299         sldns_buffer_skip(pkt, (ssize_t)xfr->namelen);
4300
4301         /* check qtype, qclass */
4302         if(sldns_buffer_remaining(pkt) < 4)
4303                 return 0;
4304         if(sldns_buffer_read_u16(pkt) != qtype)
4305                 return 0;
4306         if(sldns_buffer_read_u16(pkt) != xfr->dclass)
4307                 return 0;
4308
4309         if(serial) {
4310                 uint16_t rdlen;
4311                 /* read serial number, from answer section SOA */
4312                 if(LDNS_ANCOUNT(sldns_buffer_begin(pkt)) == 0)
4313                         return 0;
4314                 /* read from first record SOA record */
4315                 if(sldns_buffer_remaining(pkt) < 1)
4316                         return 0;
4317                 if(dname_pkt_compare(pkt, sldns_buffer_current(pkt),
4318                         xfr->name) != 0)
4319                         return 0;
4320                 if(!pkt_dname_len(pkt))
4321                         return 0;
4322                 /* type, class, ttl, rdatalen */
4323                 if(sldns_buffer_remaining(pkt) < 4+4+2)
4324                         return 0;
4325                 if(sldns_buffer_read_u16(pkt) != qtype)
4326                         return 0;
4327                 if(sldns_buffer_read_u16(pkt) != xfr->dclass)
4328                         return 0;
4329                 sldns_buffer_skip(pkt, 4); /* ttl */
4330                 rdlen = sldns_buffer_read_u16(pkt);
4331                 if(sldns_buffer_remaining(pkt) < rdlen)
4332                         return 0;
4333                 if(sldns_buffer_remaining(pkt) < 1)
4334                         return 0;
4335                 if(!pkt_dname_len(pkt)) /* soa name */
4336                         return 0;
4337                 if(sldns_buffer_remaining(pkt) < 1)
4338                         return 0;
4339                 if(!pkt_dname_len(pkt)) /* soa name */
4340                         return 0;
4341                 if(sldns_buffer_remaining(pkt) < 20)
4342                         return 0;
4343                 *serial = sldns_buffer_read_u32(pkt);
4344         }
4345         return 1;
4346 }
4347
4348 /** read one line from chunks into buffer at current position */
4349 static int
4350 chunkline_get_line(struct auth_chunk** chunk, size_t* chunk_pos,
4351         sldns_buffer* buf)
4352 {
4353         int readsome = 0;
4354         while(*chunk) {
4355                 /* more text in this chunk? */
4356                 if(*chunk_pos < (*chunk)->len) {
4357                         readsome = 1;
4358                         while(*chunk_pos < (*chunk)->len) {
4359                                 char c = (char)((*chunk)->data[*chunk_pos]);
4360                                 (*chunk_pos)++;
4361                                 if(sldns_buffer_remaining(buf) < 2) {
4362                                         /* buffer too short */
4363                                         verbose(VERB_ALGO, "http chunkline, "
4364                                                 "line too long");
4365                                         return 0;
4366                                 }
4367                                 sldns_buffer_write_u8(buf, (uint8_t)c);
4368                                 if(c == '\n') {
4369                                         /* we are done */
4370                                         return 1;
4371                                 }
4372                         }
4373                 }
4374                 /* move to next chunk */
4375                 *chunk = (*chunk)->next;
4376                 *chunk_pos = 0;
4377         }
4378         /* no more text */
4379         if(readsome) return 1;
4380         return 0;
4381 }
4382
4383 /** count number of open and closed parenthesis in a chunkline */
4384 static int
4385 chunkline_count_parens(sldns_buffer* buf, size_t start)
4386 {
4387         size_t end = sldns_buffer_position(buf);
4388         size_t i;
4389         int count = 0;
4390         int squote = 0, dquote = 0;
4391         for(i=start; i<end; i++) {
4392                 char c = (char)sldns_buffer_read_u8_at(buf, i);
4393                 if(squote && c != '\'') continue;
4394                 if(dquote && c != '"') continue;
4395                 if(c == '"')
4396                         dquote = !dquote; /* skip quoted part */
4397                 else if(c == '\'')
4398                         squote = !squote; /* skip quoted part */
4399                 else if(c == '(')
4400                         count ++;
4401                 else if(c == ')')
4402                         count --;
4403                 else if(c == ';') {
4404                         /* rest is a comment */
4405                         return count;
4406                 }
4407         }
4408         return count;
4409 }
4410
4411 /** remove trailing ;... comment from a line in the chunkline buffer */
4412 static void
4413 chunkline_remove_trailcomment(sldns_buffer* buf, size_t start)
4414 {
4415         size_t end = sldns_buffer_position(buf);
4416         size_t i;
4417         int squote = 0, dquote = 0;
4418         for(i=start; i<end; i++) {
4419                 char c = (char)sldns_buffer_read_u8_at(buf, i);
4420                 if(squote && c != '\'') continue;
4421                 if(dquote && c != '"') continue;
4422                 if(c == '"')
4423                         dquote = !dquote; /* skip quoted part */
4424                 else if(c == '\'')
4425                         squote = !squote; /* skip quoted part */
4426                 else if(c == ';') {
4427                         /* rest is a comment */
4428                         sldns_buffer_set_position(buf, i);
4429                         return;
4430                 }
4431         }
4432         /* nothing to remove */
4433 }
4434
4435 /** see if a chunkline is a comment line (or empty line) */
4436 static int
4437 chunkline_is_comment_line_or_empty(sldns_buffer* buf)
4438 {
4439         size_t i, end = sldns_buffer_limit(buf);
4440         for(i=0; i<end; i++) {
4441                 char c = (char)sldns_buffer_read_u8_at(buf, i);
4442                 if(c == ';')
4443                         return 1; /* comment */
4444                 else if(c != ' ' && c != '\t' && c != '\r' && c != '\n')
4445                         return 0; /* not a comment */
4446         }
4447         return 1; /* empty */
4448 }
4449
4450 /** find a line with ( ) collated */
4451 static int
4452 chunkline_get_line_collated(struct auth_chunk** chunk, size_t* chunk_pos,
4453         sldns_buffer* buf)
4454 {
4455         size_t pos;
4456         int parens = 0;
4457         sldns_buffer_clear(buf);
4458         pos = sldns_buffer_position(buf);
4459         if(!chunkline_get_line(chunk, chunk_pos, buf)) {
4460                 if(sldns_buffer_position(buf) < sldns_buffer_limit(buf))
4461                         sldns_buffer_write_u8_at(buf, sldns_buffer_position(buf), 0);
4462                 else sldns_buffer_write_u8_at(buf, sldns_buffer_position(buf)-1, 0);
4463                 sldns_buffer_flip(buf);
4464                 return 0;
4465         }
4466         parens += chunkline_count_parens(buf, pos);
4467         while(parens > 0) {
4468                 chunkline_remove_trailcomment(buf, pos);
4469                 pos = sldns_buffer_position(buf);
4470                 if(!chunkline_get_line(chunk, chunk_pos, buf)) {
4471                         if(sldns_buffer_position(buf) < sldns_buffer_limit(buf))
4472                                 sldns_buffer_write_u8_at(buf, sldns_buffer_position(buf), 0);
4473                         else sldns_buffer_write_u8_at(buf, sldns_buffer_position(buf)-1, 0);
4474                         sldns_buffer_flip(buf);
4475                         return 0;
4476                 }
4477                 parens += chunkline_count_parens(buf, pos);
4478         }
4479
4480         if(sldns_buffer_remaining(buf) < 1) {
4481                 verbose(VERB_ALGO, "http chunkline: "
4482                         "line too long");
4483                 return 0;
4484         }
4485         sldns_buffer_write_u8_at(buf, sldns_buffer_position(buf), 0);
4486         sldns_buffer_flip(buf);
4487         return 1;
4488 }
4489
4490 /** process $ORIGIN for http, 0 nothing, 1 done, 2 error */
4491 static int
4492 http_parse_origin(sldns_buffer* buf, struct sldns_file_parse_state* pstate)
4493 {
4494         char* line = (char*)sldns_buffer_begin(buf);
4495         if(strncmp(line, "$ORIGIN", 7) == 0 &&
4496                 isspace((unsigned char)line[7])) {
4497                 int s;
4498                 pstate->origin_len = sizeof(pstate->origin);
4499                 s = sldns_str2wire_dname_buf(sldns_strip_ws(line+8),
4500                         pstate->origin, &pstate->origin_len);
4501                 if(s) {
4502                         pstate->origin_len = 0;
4503                         return 2;
4504                 }
4505                 return 1;
4506         }
4507         return 0;
4508 }
4509
4510 /** process $TTL for http, 0 nothing, 1 done, 2 error */
4511 static int
4512 http_parse_ttl(sldns_buffer* buf, struct sldns_file_parse_state* pstate)
4513 {
4514         char* line = (char*)sldns_buffer_begin(buf);
4515         if(strncmp(line, "$TTL", 4) == 0 &&
4516                 isspace((unsigned char)line[4])) {
4517                 const char* end = NULL;
4518                 int overflow = 0;
4519                 pstate->default_ttl = sldns_str2period(
4520                         sldns_strip_ws(line+5), &end, &overflow);
4521                 if(overflow) {
4522                         return 2;
4523                 }
4524                 return 1;
4525         }
4526         return 0;
4527 }
4528
4529 /** find noncomment RR line in chunks, collates lines if ( ) format */
4530 static int
4531 chunkline_non_comment_RR(struct auth_chunk** chunk, size_t* chunk_pos,
4532         sldns_buffer* buf, struct sldns_file_parse_state* pstate)
4533 {
4534         int ret;
4535         while(chunkline_get_line_collated(chunk, chunk_pos, buf)) {
4536                 if(chunkline_is_comment_line_or_empty(buf)) {
4537                         /* a comment, go to next line */
4538                         continue;
4539                 }
4540                 if((ret=http_parse_origin(buf, pstate))!=0) {
4541                         if(ret == 2)
4542                                 return 0;
4543                         continue; /* $ORIGIN has been handled */
4544                 }
4545                 if((ret=http_parse_ttl(buf, pstate))!=0) {
4546                         if(ret == 2)
4547                                 return 0;
4548                         continue; /* $TTL has been handled */
4549                 }
4550                 return 1;
4551         }
4552         /* no noncomments, fail */
4553         return 0;
4554 }
4555
4556 /** check syntax of chunklist zonefile, parse first RR, return false on
4557  * failure and return a string in the scratch buffer (first RR string)
4558  * on failure. */
4559 static int
4560 http_zonefile_syntax_check(struct auth_xfer* xfr, sldns_buffer* buf)
4561 {
4562         uint8_t rr[LDNS_RR_BUF_SIZE];
4563         size_t rr_len, dname_len = 0;
4564         struct sldns_file_parse_state pstate;
4565         struct auth_chunk* chunk;
4566         size_t chunk_pos;
4567         int e;
4568         memset(&pstate, 0, sizeof(pstate));
4569         pstate.default_ttl = 3600;
4570         if(xfr->namelen < sizeof(pstate.origin)) {
4571                 pstate.origin_len = xfr->namelen;
4572                 memmove(pstate.origin, xfr->name, xfr->namelen);
4573         }
4574         chunk = xfr->task_transfer->chunks_first;
4575         chunk_pos = 0;
4576         if(!chunkline_non_comment_RR(&chunk, &chunk_pos, buf, &pstate)) {
4577                 return 0;
4578         }
4579         rr_len = sizeof(rr);
4580         e=sldns_str2wire_rr_buf((char*)sldns_buffer_begin(buf), rr, &rr_len,
4581                 &dname_len, pstate.default_ttl,
4582                 pstate.origin_len?pstate.origin:NULL, pstate.origin_len,
4583                 pstate.prev_rr_len?pstate.prev_rr:NULL, pstate.prev_rr_len);
4584         if(e != 0) {
4585                 log_err("parse failure on first RR[%d]: %s",
4586                         LDNS_WIREPARSE_OFFSET(e),
4587                         sldns_get_errorstr_parse(LDNS_WIREPARSE_ERROR(e)));
4588                 return 0;
4589         }
4590         /* check that class is correct */
4591         if(sldns_wirerr_get_class(rr, rr_len, dname_len) != xfr->dclass) {
4592                 log_err("parse failure: first record in downloaded zonefile "
4593                         "from wrong RR class");
4594                 return 0;
4595         }
4596         return 1;
4597 }
4598
4599 /** sum sizes of chunklist */
4600 static size_t
4601 chunklist_sum(struct auth_chunk* list)
4602 {
4603         struct auth_chunk* p;
4604         size_t s = 0;
4605         for(p=list; p; p=p->next) {
4606                 s += p->len;
4607         }
4608         return s;
4609 }
4610
4611 /** remove newlines from collated line */
4612 static void
4613 chunkline_newline_removal(sldns_buffer* buf)
4614 {
4615         size_t i, end=sldns_buffer_limit(buf);
4616         for(i=0; i<end; i++) {
4617                 char c = (char)sldns_buffer_read_u8_at(buf, i);
4618                 if(c == '\n' && i==end-1) {
4619                         sldns_buffer_write_u8_at(buf, i, 0);
4620                         sldns_buffer_set_limit(buf, end-1);
4621                         return;
4622                 }
4623                 if(c == '\n')
4624                         sldns_buffer_write_u8_at(buf, i, (uint8_t)' ');
4625         }
4626 }
4627
4628 /** for http download, parse and add RR to zone */
4629 static int
4630 http_parse_add_rr(struct auth_xfer* xfr, struct auth_zone* z,
4631         sldns_buffer* buf, struct sldns_file_parse_state* pstate)
4632 {
4633         uint8_t rr[LDNS_RR_BUF_SIZE];
4634         size_t rr_len, dname_len = 0;
4635         int e;
4636         char* line = (char*)sldns_buffer_begin(buf);
4637         rr_len = sizeof(rr);
4638         e = sldns_str2wire_rr_buf(line, rr, &rr_len, &dname_len,
4639                 pstate->default_ttl,
4640                 pstate->origin_len?pstate->origin:NULL, pstate->origin_len,
4641                 pstate->prev_rr_len?pstate->prev_rr:NULL, pstate->prev_rr_len);
4642         if(e != 0) {
4643                 log_err("%s/%s parse failure RR[%d]: %s in '%s'",
4644                         xfr->task_transfer->master->host,
4645                         xfr->task_transfer->master->file,
4646                         LDNS_WIREPARSE_OFFSET(e),
4647                         sldns_get_errorstr_parse(LDNS_WIREPARSE_ERROR(e)),
4648                         line);
4649                 return 0;
4650         }
4651         if(rr_len == 0)
4652                 return 1; /* empty line or so */
4653
4654         /* set prev */
4655         if(dname_len < sizeof(pstate->prev_rr)) {
4656                 memmove(pstate->prev_rr, rr, dname_len);
4657                 pstate->prev_rr_len = dname_len;
4658         }
4659
4660         return az_insert_rr(z, rr, rr_len, dname_len, NULL);
4661 }
4662
4663 /** RR list iterator, returns RRs from answer section one by one from the
4664  * dns packets in the chunklist */
4665 static void
4666 chunk_rrlist_start(struct auth_xfer* xfr, struct auth_chunk** rr_chunk,
4667         int* rr_num, size_t* rr_pos)
4668 {
4669         *rr_chunk = xfr->task_transfer->chunks_first;
4670         *rr_num = 0;
4671         *rr_pos = 0;
4672 }
4673
4674 /** RR list iterator, see if we are at the end of the list */
4675 static int
4676 chunk_rrlist_end(struct auth_chunk* rr_chunk, int rr_num)
4677 {
4678         while(rr_chunk) {
4679                 if(rr_chunk->len < LDNS_HEADER_SIZE)
4680                         return 1;
4681                 if(rr_num < (int)LDNS_ANCOUNT(rr_chunk->data))
4682                         return 0;
4683                 /* no more RRs in this chunk */
4684                 /* continue with next chunk, see if it has RRs */
4685                 rr_chunk = rr_chunk->next;
4686                 rr_num = 0;
4687         }
4688         return 1;
4689 }
4690
4691 /** RR list iterator, move to next RR */
4692 static void
4693 chunk_rrlist_gonext(struct auth_chunk** rr_chunk, int* rr_num,
4694         size_t* rr_pos, size_t rr_nextpos)
4695 {
4696         /* already at end of chunks? */
4697         if(!*rr_chunk)
4698                 return;
4699         /* move within this chunk */
4700         if((*rr_chunk)->len >= LDNS_HEADER_SIZE &&
4701                 (*rr_num)+1 < (int)LDNS_ANCOUNT((*rr_chunk)->data)) {
4702                 (*rr_num) += 1;
4703                 *rr_pos = rr_nextpos;
4704                 return;
4705         }
4706         /* no more RRs in this chunk */
4707         /* continue with next chunk, see if it has RRs */
4708         if(*rr_chunk)
4709                 *rr_chunk = (*rr_chunk)->next;
4710         while(*rr_chunk) {
4711                 *rr_num = 0;
4712                 *rr_pos = 0;
4713                 if((*rr_chunk)->len >= LDNS_HEADER_SIZE &&
4714                         LDNS_ANCOUNT((*rr_chunk)->data) > 0) {
4715                         return;
4716                 }
4717                 *rr_chunk = (*rr_chunk)->next;
4718         }
4719 }
4720
4721 /** RR iterator, get current RR information, false on parse error */
4722 static int
4723 chunk_rrlist_get_current(struct auth_chunk* rr_chunk, int rr_num,
4724         size_t rr_pos, uint8_t** rr_dname, uint16_t* rr_type,
4725         uint16_t* rr_class, uint32_t* rr_ttl, uint16_t* rr_rdlen,
4726         uint8_t** rr_rdata, size_t* rr_nextpos)
4727 {
4728         sldns_buffer pkt;
4729         /* integrity checks on position */
4730         if(!rr_chunk) return 0;
4731         if(rr_chunk->len < LDNS_HEADER_SIZE) return 0;
4732         if(rr_num >= (int)LDNS_ANCOUNT(rr_chunk->data)) return 0;
4733         if(rr_pos >= rr_chunk->len) return 0;
4734
4735         /* fetch rr information */
4736         sldns_buffer_init_frm_data(&pkt, rr_chunk->data, rr_chunk->len);
4737         if(rr_pos == 0) {
4738                 size_t i;
4739                 /* skip question section */
4740                 sldns_buffer_set_position(&pkt, LDNS_HEADER_SIZE);
4741                 for(i=0; i<LDNS_QDCOUNT(rr_chunk->data); i++) {
4742                         if(pkt_dname_len(&pkt) == 0) return 0;
4743                         if(sldns_buffer_remaining(&pkt) < 4) return 0;
4744                         sldns_buffer_skip(&pkt, 4); /* type and class */
4745                 }
4746         } else  {
4747                 sldns_buffer_set_position(&pkt, rr_pos);
4748         }
4749         *rr_dname = sldns_buffer_current(&pkt);
4750         if(pkt_dname_len(&pkt) == 0) return 0;
4751         if(sldns_buffer_remaining(&pkt) < 10) return 0;
4752         *rr_type = sldns_buffer_read_u16(&pkt);
4753         *rr_class = sldns_buffer_read_u16(&pkt);
4754         *rr_ttl = sldns_buffer_read_u32(&pkt);
4755         *rr_rdlen = sldns_buffer_read_u16(&pkt);
4756         if(sldns_buffer_remaining(&pkt) < (*rr_rdlen)) return 0;
4757         *rr_rdata = sldns_buffer_current(&pkt);
4758         sldns_buffer_skip(&pkt, (ssize_t)(*rr_rdlen));
4759         *rr_nextpos = sldns_buffer_position(&pkt);
4760         return 1;
4761 }
4762
4763 /** print log message where we are in parsing the zone transfer */
4764 static void
4765 log_rrlist_position(const char* label, struct auth_chunk* rr_chunk,
4766         uint8_t* rr_dname, uint16_t rr_type, size_t rr_counter)
4767 {
4768         sldns_buffer pkt;
4769         size_t dlen;
4770         uint8_t buf[256];
4771         char str[256];
4772         char typestr[32];
4773         sldns_buffer_init_frm_data(&pkt, rr_chunk->data, rr_chunk->len);
4774         sldns_buffer_set_position(&pkt, (size_t)(rr_dname -
4775                 sldns_buffer_begin(&pkt)));
4776         if((dlen=pkt_dname_len(&pkt)) == 0) return;
4777         if(dlen >= sizeof(buf)) return;
4778         dname_pkt_copy(&pkt, buf, rr_dname);
4779         dname_str(buf, str);
4780         (void)sldns_wire2str_type_buf(rr_type, typestr, sizeof(typestr));
4781         verbose(VERB_ALGO, "%s at[%d] %s %s", label, (int)rr_counter,
4782                 str, typestr);
4783 }
4784
4785 /** check that start serial is OK for ixfr. we are at rr_counter == 0,
4786  * and we are going to check rr_counter == 1 (has to be type SOA) serial */
4787 static int
4788 ixfr_start_serial(struct auth_chunk* rr_chunk, int rr_num, size_t rr_pos,
4789         uint8_t* rr_dname, uint16_t rr_type, uint16_t rr_class,
4790         uint32_t rr_ttl, uint16_t rr_rdlen, uint8_t* rr_rdata,
4791         size_t rr_nextpos, uint32_t transfer_serial, uint32_t xfr_serial)
4792 {
4793         uint32_t startserial;
4794         /* move forward on RR */
4795         chunk_rrlist_gonext(&rr_chunk, &rr_num, &rr_pos, rr_nextpos);
4796         if(chunk_rrlist_end(rr_chunk, rr_num)) {
4797                 /* no second SOA */
4798                 verbose(VERB_OPS, "IXFR has no second SOA record");
4799                 return 0;
4800         }
4801         if(!chunk_rrlist_get_current(rr_chunk, rr_num, rr_pos,
4802                 &rr_dname, &rr_type, &rr_class, &rr_ttl, &rr_rdlen,
4803                 &rr_rdata, &rr_nextpos)) {
4804                 verbose(VERB_OPS, "IXFR cannot parse second SOA record");
4805                 /* failed to parse RR */
4806                 return 0;
4807         }
4808         if(rr_type != LDNS_RR_TYPE_SOA) {
4809                 verbose(VERB_OPS, "IXFR second record is not type SOA");
4810                 return 0;
4811         }
4812         if(rr_rdlen < 22) {
4813                 verbose(VERB_OPS, "IXFR, second SOA has short rdlength");
4814                 return 0; /* bad SOA rdlen */
4815         }
4816         startserial = sldns_read_uint32(rr_rdata+rr_rdlen-20);
4817         if(startserial == transfer_serial) {
4818                 /* empty AXFR, not an IXFR */
4819                 verbose(VERB_OPS, "IXFR second serial same as first");
4820                 return 0;
4821         }
4822         if(startserial != xfr_serial) {
4823                 /* wrong start serial, it does not match the serial in
4824                  * memory */
4825                 verbose(VERB_OPS, "IXFR is from serial %u to %u but %u "
4826                         "in memory, rejecting the zone transfer",
4827                         (unsigned)startserial, (unsigned)transfer_serial,
4828                         (unsigned)xfr_serial);
4829                 return 0;
4830         }
4831         /* everything OK in second SOA serial */
4832         return 1;
4833 }
4834
4835 /** apply IXFR to zone in memory. z is locked. false on failure(mallocfail) */
4836 static int
4837 apply_ixfr(struct auth_xfer* xfr, struct auth_zone* z,
4838         struct sldns_buffer* scratch_buffer)
4839 {
4840         struct auth_chunk* rr_chunk;
4841         int rr_num;
4842         size_t rr_pos;
4843         uint8_t* rr_dname, *rr_rdata;
4844         uint16_t rr_type, rr_class, rr_rdlen;
4845         uint32_t rr_ttl;
4846         size_t rr_nextpos;
4847         int have_transfer_serial = 0;
4848         uint32_t transfer_serial = 0;
4849         size_t rr_counter = 0;
4850         int delmode = 0;
4851         int softfail = 0;
4852
4853         /* start RR iterator over chunklist of packets */
4854         chunk_rrlist_start(xfr, &rr_chunk, &rr_num, &rr_pos);
4855         while(!chunk_rrlist_end(rr_chunk, rr_num)) {
4856                 if(!chunk_rrlist_get_current(rr_chunk, rr_num, rr_pos,
4857                         &rr_dname, &rr_type, &rr_class, &rr_ttl, &rr_rdlen,
4858                         &rr_rdata, &rr_nextpos)) {
4859                         /* failed to parse RR */
4860                         return 0;
4861                 }
4862                 if(verbosity>=7) log_rrlist_position("apply ixfr",
4863                         rr_chunk, rr_dname, rr_type, rr_counter);
4864                 /* twiddle add/del mode and check for start and end */
4865                 if(rr_counter == 0 && rr_type != LDNS_RR_TYPE_SOA)
4866                         return 0;
4867                 if(rr_counter == 1 && rr_type != LDNS_RR_TYPE_SOA) {
4868                         /* this is an AXFR returned from the IXFR master */
4869                         /* but that should already have been detected, by
4870                          * on_ixfr_is_axfr */
4871                         return 0;
4872                 }
4873                 if(rr_type == LDNS_RR_TYPE_SOA) {
4874                         uint32_t serial;
4875                         if(rr_rdlen < 22) return 0; /* bad SOA rdlen */
4876                         serial = sldns_read_uint32(rr_rdata+rr_rdlen-20);
4877                         if(have_transfer_serial == 0) {
4878                                 have_transfer_serial = 1;
4879                                 transfer_serial = serial;
4880                                 delmode = 1; /* gets negated below */
4881                                 /* check second RR before going any further */
4882                                 if(!ixfr_start_serial(rr_chunk, rr_num, rr_pos,
4883                                         rr_dname, rr_type, rr_class, rr_ttl,
4884                                         rr_rdlen, rr_rdata, rr_nextpos,
4885                                         transfer_serial, xfr->serial)) {
4886                                         return 0;
4887                                 }
4888                         } else if(transfer_serial == serial) {
4889                                 have_transfer_serial++;
4890                                 if(rr_counter == 1) {
4891                                         /* empty AXFR, with SOA; SOA; */
4892                                         /* should have been detected by
4893                                          * on_ixfr_is_axfr */
4894                                         return 0;
4895                                 }
4896                                 if(have_transfer_serial == 3) {
4897                                         /* see serial three times for end */
4898                                         /* eg. IXFR:
4899                                          *  SOA 3 start
4900                                          *  SOA 1 second RR, followed by del
4901                                          *  SOA 2 followed by add
4902                                          *  SOA 2 followed by del
4903                                          *  SOA 3 followed by add
4904                                          *  SOA 3 end */
4905                                         /* ended by SOA record */
4906                                         xfr->serial = transfer_serial;
4907                                         break;
4908                                 }
4909                         }
4910                         /* twiddle add/del mode */
4911                         /* switch from delete part to add part and back again
4912                          * just before the soa, it gets deleted and added too
4913                          * this means we switch to delete mode for the final
4914                          * SOA(so skip that one) */
4915                         delmode = !delmode;
4916                 }
4917                 /* process this RR */
4918                 /* if the RR is deleted twice or added twice, then we 
4919                  * softfail, and continue with the rest of the IXFR, so
4920                  * that we serve something fairly nice during the refetch */
4921                 if(verbosity>=7) log_rrlist_position((delmode?"del":"add"),
4922                         rr_chunk, rr_dname, rr_type, rr_counter);
4923                 if(delmode) {
4924                         /* delete this RR */
4925                         int nonexist = 0;
4926                         if(!az_remove_rr_decompress(z, rr_chunk->data,
4927                                 rr_chunk->len, scratch_buffer, rr_dname,
4928                                 rr_type, rr_class, rr_ttl, rr_rdata, rr_rdlen,
4929                                 &nonexist)) {
4930                                 /* failed, malloc error or so */
4931                                 return 0;
4932                         }
4933                         if(nonexist) {
4934                                 /* it was removal of a nonexisting RR */
4935                                 if(verbosity>=4) log_rrlist_position(
4936                                         "IXFR error nonexistent RR",
4937                                         rr_chunk, rr_dname, rr_type, rr_counter);
4938                                 softfail = 1;
4939                         }
4940                 } else if(rr_counter != 0) {
4941                         /* skip first SOA RR for addition, it is added in
4942                          * the addition part near the end of the ixfr, when
4943                          * that serial is seen the second time. */
4944                         int duplicate = 0;
4945                         /* add this RR */
4946                         if(!az_insert_rr_decompress(z, rr_chunk->data,
4947                                 rr_chunk->len, scratch_buffer, rr_dname,
4948                                 rr_type, rr_class, rr_ttl, rr_rdata, rr_rdlen,
4949                                 &duplicate)) {
4950                                 /* failed, malloc error or so */
4951                                 return 0;
4952                         }
4953                         if(duplicate) {
4954                                 /* it was a duplicate */
4955                                 if(verbosity>=4) log_rrlist_position(
4956                                         "IXFR error duplicate RR",
4957                                         rr_chunk, rr_dname, rr_type, rr_counter);
4958                                 softfail = 1;
4959                         }
4960                 }
4961
4962                 rr_counter++;
4963                 chunk_rrlist_gonext(&rr_chunk, &rr_num, &rr_pos, rr_nextpos);
4964         }
4965         if(softfail) {
4966                 verbose(VERB_ALGO, "IXFR did not apply cleanly, fetching full zone");
4967                 return 0;
4968         }
4969         return 1;
4970 }
4971
4972 /** apply AXFR to zone in memory. z is locked. false on failure(mallocfail) */
4973 static int
4974 apply_axfr(struct auth_xfer* xfr, struct auth_zone* z,
4975         struct sldns_buffer* scratch_buffer)
4976 {
4977         struct auth_chunk* rr_chunk;
4978         int rr_num;
4979         size_t rr_pos;
4980         uint8_t* rr_dname, *rr_rdata;
4981         uint16_t rr_type, rr_class, rr_rdlen;
4982         uint32_t rr_ttl;
4983         uint32_t serial = 0;
4984         size_t rr_nextpos;
4985         size_t rr_counter = 0;
4986         int have_end_soa = 0;
4987
4988         /* clear the data tree */
4989         traverse_postorder(&z->data, auth_data_del, NULL);
4990         rbtree_init(&z->data, &auth_data_cmp);
4991         /* clear the RPZ policies */
4992         if(z->rpz)
4993                 rpz_clear(z->rpz);
4994
4995         xfr->have_zone = 0;
4996         xfr->serial = 0;
4997
4998         /* insert all RRs in to the zone */
4999         /* insert the SOA only once, skip the last one */
5000         /* start RR iterator over chunklist of packets */
5001         chunk_rrlist_start(xfr, &rr_chunk, &rr_num, &rr_pos);
5002         while(!chunk_rrlist_end(rr_chunk, rr_num)) {
5003                 if(!chunk_rrlist_get_current(rr_chunk, rr_num, rr_pos,
5004                         &rr_dname, &rr_type, &rr_class, &rr_ttl, &rr_rdlen,
5005                         &rr_rdata, &rr_nextpos)) {
5006                         /* failed to parse RR */
5007                         return 0;
5008                 }
5009                 if(verbosity>=7) log_rrlist_position("apply_axfr",
5010                         rr_chunk, rr_dname, rr_type, rr_counter);
5011                 if(rr_type == LDNS_RR_TYPE_SOA) {
5012                         if(rr_counter != 0) {
5013                                 /* end of the axfr */
5014                                 have_end_soa = 1;
5015                                 break;
5016                         }
5017                         if(rr_rdlen < 22) return 0; /* bad SOA rdlen */
5018                         serial = sldns_read_uint32(rr_rdata+rr_rdlen-20);
5019                 }
5020
5021                 /* add this RR */
5022                 if(!az_insert_rr_decompress(z, rr_chunk->data, rr_chunk->len,
5023                         scratch_buffer, rr_dname, rr_type, rr_class, rr_ttl,
5024                         rr_rdata, rr_rdlen, NULL)) {
5025                         /* failed, malloc error or so */
5026                         return 0;
5027                 }
5028
5029                 rr_counter++;
5030                 chunk_rrlist_gonext(&rr_chunk, &rr_num, &rr_pos, rr_nextpos);
5031         }
5032         if(!have_end_soa) {
5033                 log_err("no end SOA record for AXFR");
5034                 return 0;
5035         }
5036
5037         xfr->serial = serial;
5038         xfr->have_zone = 1;
5039         return 1;
5040 }
5041
5042 /** apply HTTP to zone in memory. z is locked. false on failure(mallocfail) */
5043 static int
5044 apply_http(struct auth_xfer* xfr, struct auth_zone* z,
5045         struct sldns_buffer* scratch_buffer)
5046 {
5047         /* parse data in chunks */
5048         /* parse RR's and read into memory. ignore $INCLUDE from the
5049          * downloaded file*/
5050         struct sldns_file_parse_state pstate;
5051         struct auth_chunk* chunk;
5052         size_t chunk_pos;
5053         int ret;
5054         memset(&pstate, 0, sizeof(pstate));
5055         pstate.default_ttl = 3600;
5056         if(xfr->namelen < sizeof(pstate.origin)) {
5057                 pstate.origin_len = xfr->namelen;
5058                 memmove(pstate.origin, xfr->name, xfr->namelen);
5059         }
5060
5061         if(verbosity >= VERB_ALGO)
5062                 verbose(VERB_ALGO, "http download %s of size %d",
5063                 xfr->task_transfer->master->file,
5064                 (int)chunklist_sum(xfr->task_transfer->chunks_first));
5065         if(xfr->task_transfer->chunks_first && verbosity >= VERB_ALGO) {
5066                 char preview[1024];
5067                 if(xfr->task_transfer->chunks_first->len+1 > sizeof(preview)) {
5068                         memmove(preview, xfr->task_transfer->chunks_first->data,
5069                                 sizeof(preview)-1);
5070                         preview[sizeof(preview)-1]=0;
5071                 } else {
5072                         memmove(preview, xfr->task_transfer->chunks_first->data,
5073                                 xfr->task_transfer->chunks_first->len);
5074                         preview[xfr->task_transfer->chunks_first->len]=0;
5075                 }
5076                 log_info("auth zone http downloaded content preview: %s",
5077                         preview);
5078         }
5079
5080         /* perhaps a little syntax check before we try to apply the data? */
5081         if(!http_zonefile_syntax_check(xfr, scratch_buffer)) {
5082                 log_err("http download %s/%s does not contain a zonefile, "
5083                         "but got '%s'", xfr->task_transfer->master->host,
5084                         xfr->task_transfer->master->file,
5085                         sldns_buffer_begin(scratch_buffer));
5086                 return 0;
5087         }
5088
5089         /* clear the data tree */
5090         traverse_postorder(&z->data, auth_data_del, NULL);
5091         rbtree_init(&z->data, &auth_data_cmp);
5092         /* clear the RPZ policies */
5093         if(z->rpz)
5094                 rpz_clear(z->rpz);
5095
5096         xfr->have_zone = 0;
5097         xfr->serial = 0;
5098
5099         chunk = xfr->task_transfer->chunks_first;
5100         chunk_pos = 0;
5101         pstate.lineno = 0;
5102         while(chunkline_get_line_collated(&chunk, &chunk_pos, scratch_buffer)) {
5103                 /* process this line */
5104                 pstate.lineno++;
5105                 chunkline_newline_removal(scratch_buffer);
5106                 if(chunkline_is_comment_line_or_empty(scratch_buffer)) {
5107                         continue;
5108                 }
5109                 /* parse line and add RR */
5110                 if((ret=http_parse_origin(scratch_buffer, &pstate))!=0) {
5111                         if(ret == 2) {
5112                                 verbose(VERB_ALGO, "error parsing ORIGIN on line [%s:%d] %s",
5113                                         xfr->task_transfer->master->file,
5114                                         pstate.lineno,
5115                                         sldns_buffer_begin(scratch_buffer));
5116                                 return 0;
5117                         }
5118                         continue; /* $ORIGIN has been handled */
5119                 }
5120                 if((ret=http_parse_ttl(scratch_buffer, &pstate))!=0) {
5121                         if(ret == 2) {
5122                                 verbose(VERB_ALGO, "error parsing TTL on line [%s:%d] %s",
5123                                         xfr->task_transfer->master->file,
5124                                         pstate.lineno,
5125                                         sldns_buffer_begin(scratch_buffer));
5126                                 return 0;
5127                         }
5128                         continue; /* $TTL has been handled */
5129                 }
5130                 if(!http_parse_add_rr(xfr, z, scratch_buffer, &pstate)) {
5131                         verbose(VERB_ALGO, "error parsing line [%s:%d] %s",
5132                                 xfr->task_transfer->master->file,
5133                                 pstate.lineno,
5134                                 sldns_buffer_begin(scratch_buffer));
5135                         return 0;
5136                 }
5137         }
5138         return 1;
5139 }
5140
5141 /** write http chunks to zonefile to create downloaded file */
5142 static int
5143 auth_zone_write_chunks(struct auth_xfer* xfr, const char* fname)
5144 {
5145         FILE* out;
5146         struct auth_chunk* p;
5147         out = fopen(fname, "w");
5148         if(!out) {
5149                 log_err("could not open %s: %s", fname, strerror(errno));
5150                 return 0;
5151         }
5152         for(p = xfr->task_transfer->chunks_first; p ; p = p->next) {
5153                 if(!write_out(out, (char*)p->data, p->len)) {
5154                         log_err("could not write http download to %s", fname);
5155                         fclose(out);
5156                         return 0;
5157                 }
5158         }
5159         fclose(out);
5160         return 1;
5161 }
5162
5163 /** write to zonefile after zone has been updated */
5164 static void
5165 xfr_write_after_update(struct auth_xfer* xfr, struct module_env* env)
5166 {
5167         struct config_file* cfg = env->cfg;
5168         struct auth_zone* z;
5169         char tmpfile[1024];
5170         char* zfilename;
5171         lock_basic_unlock(&xfr->lock);
5172
5173         /* get lock again, so it is a readlock and concurrently queries
5174          * can be answered */
5175         lock_rw_rdlock(&env->auth_zones->lock);
5176         z = auth_zone_find(env->auth_zones, xfr->name, xfr->namelen,
5177                 xfr->dclass);
5178         if(!z) {
5179                 lock_rw_unlock(&env->auth_zones->lock);
5180                 /* the zone is gone, ignore xfr results */
5181                 lock_basic_lock(&xfr->lock);
5182                 return;
5183         }
5184         lock_rw_rdlock(&z->lock);
5185         lock_basic_lock(&xfr->lock);
5186         lock_rw_unlock(&env->auth_zones->lock);
5187
5188         if(z->zonefile == NULL || z->zonefile[0] == 0) {
5189                 lock_rw_unlock(&z->lock);
5190                 /* no write needed, no zonefile set */
5191                 return;
5192         }
5193         zfilename = z->zonefile;
5194         if(cfg->chrootdir && cfg->chrootdir[0] && strncmp(zfilename,
5195                 cfg->chrootdir, strlen(cfg->chrootdir)) == 0)
5196                 zfilename += strlen(cfg->chrootdir);
5197         if(verbosity >= VERB_ALGO) {
5198                 char nm[255+1];
5199                 dname_str(z->name, nm);
5200                 verbose(VERB_ALGO, "write zonefile %s for %s", zfilename, nm);
5201         }
5202
5203         /* write to tempfile first */
5204         if((size_t)strlen(zfilename) + 16 > sizeof(tmpfile)) {
5205                 verbose(VERB_ALGO, "tmpfilename too long, cannot update "
5206                         " zonefile %s", zfilename);
5207                 lock_rw_unlock(&z->lock);
5208                 return;
5209         }
5210         snprintf(tmpfile, sizeof(tmpfile), "%s.tmp%u", zfilename,
5211                 (unsigned)getpid());
5212         if(xfr->task_transfer->master->http) {
5213                 /* use the stored chunk list to write them */
5214                 if(!auth_zone_write_chunks(xfr, tmpfile)) {
5215                         unlink(tmpfile);
5216                         lock_rw_unlock(&z->lock);
5217                         return;
5218                 }
5219         } else if(!auth_zone_write_file(z, tmpfile)) {
5220                 unlink(tmpfile);
5221                 lock_rw_unlock(&z->lock);
5222                 return;
5223         }
5224 #ifdef UB_ON_WINDOWS
5225         (void)unlink(zfilename); /* windows does not replace file with rename() */
5226 #endif
5227         if(rename(tmpfile, zfilename) < 0) {
5228                 log_err("could not rename(%s, %s): %s", tmpfile, zfilename,
5229                         strerror(errno));
5230                 unlink(tmpfile);
5231                 lock_rw_unlock(&z->lock);
5232                 return;
5233         }
5234         lock_rw_unlock(&z->lock);
5235 }
5236
5237 /** reacquire locks and structures. Starts with no locks, ends
5238  * with xfr and z locks, if fail, no z lock */
5239 static int xfr_process_reacquire_locks(struct auth_xfer* xfr,
5240         struct module_env* env, struct auth_zone** z)
5241 {
5242         /* release xfr lock, then, while holding az->lock grab both
5243          * z->lock and xfr->lock */
5244         lock_rw_rdlock(&env->auth_zones->lock);
5245         *z = auth_zone_find(env->auth_zones, xfr->name, xfr->namelen,
5246                 xfr->dclass);
5247         if(!*z) {
5248                 lock_rw_unlock(&env->auth_zones->lock);
5249                 lock_basic_lock(&xfr->lock);
5250                 *z = NULL;
5251                 return 0;
5252         }
5253         lock_rw_wrlock(&(*z)->lock);
5254         lock_basic_lock(&xfr->lock);
5255         lock_rw_unlock(&env->auth_zones->lock);
5256         return 1;
5257 }
5258
5259 /** process chunk list and update zone in memory,
5260  * return false if it did not work */
5261 static int
5262 xfr_process_chunk_list(struct auth_xfer* xfr, struct module_env* env,
5263         int* ixfr_fail)
5264 {
5265         struct auth_zone* z;
5266
5267         /* obtain locks and structures */
5268         lock_basic_unlock(&xfr->lock);
5269         if(!xfr_process_reacquire_locks(xfr, env, &z)) {
5270                 /* the zone is gone, ignore xfr results */
5271                 return 0;
5272         }
5273         /* holding xfr and z locks */
5274
5275         /* apply data */
5276         if(xfr->task_transfer->master->http) {
5277                 if(!apply_http(xfr, z, env->scratch_buffer)) {
5278                         lock_rw_unlock(&z->lock);
5279                         verbose(VERB_ALGO, "http from %s: could not store data",
5280                                 xfr->task_transfer->master->host);
5281                         return 0;
5282                 }
5283         } else if(xfr->task_transfer->on_ixfr &&
5284                 !xfr->task_transfer->on_ixfr_is_axfr) {
5285                 if(!apply_ixfr(xfr, z, env->scratch_buffer)) {
5286                         lock_rw_unlock(&z->lock);
5287                         verbose(VERB_ALGO, "xfr from %s: could not store IXFR"
5288                                 " data", xfr->task_transfer->master->host);
5289                         *ixfr_fail = 1;
5290                         return 0;
5291                 }
5292         } else {
5293                 if(!apply_axfr(xfr, z, env->scratch_buffer)) {
5294                         lock_rw_unlock(&z->lock);
5295                         verbose(VERB_ALGO, "xfr from %s: could not store AXFR"
5296                                 " data", xfr->task_transfer->master->host);
5297                         return 0;
5298                 }
5299         }
5300         xfr->zone_expired = 0;
5301         z->zone_expired = 0;
5302         if(!xfr_find_soa(z, xfr)) {
5303                 lock_rw_unlock(&z->lock);
5304                 verbose(VERB_ALGO, "xfr from %s: no SOA in zone after update"
5305                         " (or malformed RR)", xfr->task_transfer->master->host);
5306                 return 0;
5307         }
5308
5309         /* release xfr lock while verifying zonemd because it may have
5310          * to spawn lookups in the state machines */
5311         lock_basic_unlock(&xfr->lock);
5312         /* holding z lock */
5313         auth_zone_verify_zonemd(z, env, &env->mesh->mods, NULL, 0, 0);
5314         if(z->zone_expired) {
5315                 char zname[256];
5316                 /* ZONEMD must have failed */
5317                 /* reacquire locks, so we hold xfr lock on exit of routine,
5318                  * and both xfr and z again after releasing xfr for potential
5319                  * state machine mesh callbacks */
5320                 lock_rw_unlock(&z->lock);
5321                 if(!xfr_process_reacquire_locks(xfr, env, &z))
5322                         return 0;
5323                 dname_str(xfr->name, zname);
5324                 verbose(VERB_ALGO, "xfr from %s: ZONEMD failed for %s, transfer is failed", xfr->task_transfer->master->host, zname);
5325                 xfr->zone_expired = 1;
5326                 lock_rw_unlock(&z->lock);
5327                 return 0;
5328         }
5329         /* reacquire locks, so we hold xfr lock on exit of routine,
5330          * and both xfr and z again after releasing xfr for potential
5331          * state machine mesh callbacks */
5332         lock_rw_unlock(&z->lock);
5333         if(!xfr_process_reacquire_locks(xfr, env, &z))
5334                 return 0;
5335         /* holding xfr and z locks */
5336
5337         if(xfr->have_zone)
5338                 xfr->lease_time = *env->now;
5339
5340         if(z->rpz)
5341                 rpz_finish_config(z->rpz);
5342
5343         /* unlock */
5344         lock_rw_unlock(&z->lock);
5345
5346         if(verbosity >= VERB_QUERY && xfr->have_zone) {
5347                 char zname[256];
5348                 dname_str(xfr->name, zname);
5349                 verbose(VERB_QUERY, "auth zone %s updated to serial %u", zname,
5350                         (unsigned)xfr->serial);
5351         }
5352         /* see if we need to write to a zonefile */
5353         xfr_write_after_update(xfr, env);
5354         return 1;
5355 }
5356
5357 /** disown task_transfer.  caller must hold xfr.lock */
5358 static void
5359 xfr_transfer_disown(struct auth_xfer* xfr)
5360 {
5361         /* remove timer (from this worker's event base) */
5362         comm_timer_delete(xfr->task_transfer->timer);
5363         xfr->task_transfer->timer = NULL;
5364         /* remove the commpoint */
5365         comm_point_delete(xfr->task_transfer->cp);
5366         xfr->task_transfer->cp = NULL;
5367         /* we don't own this item anymore */
5368         xfr->task_transfer->worker = NULL;
5369         xfr->task_transfer->env = NULL;
5370 }
5371
5372 /** lookup a host name for its addresses, if needed */
5373 static int
5374 xfr_transfer_lookup_host(struct auth_xfer* xfr, struct module_env* env)
5375 {
5376         struct sockaddr_storage addr;
5377         socklen_t addrlen = 0;
5378         struct auth_master* master = xfr->task_transfer->lookup_target;
5379         struct query_info qinfo;
5380         uint16_t qflags = BIT_RD;
5381         uint8_t dname[LDNS_MAX_DOMAINLEN+1];
5382         struct edns_data edns;
5383         sldns_buffer* buf = env->scratch_buffer;
5384         if(!master) return 0;
5385         if(extstrtoaddr(master->host, &addr, &addrlen, UNBOUND_DNS_PORT)) {
5386                 /* not needed, host is in IP addr format */
5387                 return 0;
5388         }
5389         if(master->allow_notify)
5390                 return 0; /* allow-notifies are not transferred from, no
5391                 lookup is needed */
5392
5393         /* use mesh_new_callback to probe for non-addr hosts,
5394          * and then wait for them to be looked up (in cache, or query) */
5395         qinfo.qname_len = sizeof(dname);
5396         if(sldns_str2wire_dname_buf(master->host, dname, &qinfo.qname_len)
5397                 != 0) {
5398                 log_err("cannot parse host name of master %s", master->host);
5399                 return 0;
5400         }
5401         qinfo.qname = dname;
5402         qinfo.qclass = xfr->dclass;
5403         qinfo.qtype = LDNS_RR_TYPE_A;
5404         if(xfr->task_transfer->lookup_aaaa)
5405                 qinfo.qtype = LDNS_RR_TYPE_AAAA;
5406         qinfo.local_alias = NULL;
5407         if(verbosity >= VERB_ALGO) {
5408                 char buf1[512];
5409                 char buf2[LDNS_MAX_DOMAINLEN+1];
5410                 dname_str(xfr->name, buf2);
5411                 snprintf(buf1, sizeof(buf1), "auth zone %s: master lookup"
5412                         " for task_transfer", buf2);
5413                 log_query_info(VERB_ALGO, buf1, &qinfo);
5414         }
5415         edns.edns_present = 1;
5416         edns.ext_rcode = 0;
5417         edns.edns_version = 0;
5418         edns.bits = EDNS_DO;
5419         edns.opt_list_in = NULL;
5420         edns.opt_list_out = NULL;
5421         edns.opt_list_inplace_cb_out = NULL;
5422         edns.padding_block_size = 0;
5423         edns.cookie_present = 0;
5424         edns.cookie_valid = 0;
5425         if(sldns_buffer_capacity(buf) < 65535)
5426                 edns.udp_size = (uint16_t)sldns_buffer_capacity(buf);
5427         else    edns.udp_size = 65535;
5428
5429         /* unlock xfr during mesh_new_callback() because the callback can be
5430          * called straight away */
5431         lock_basic_unlock(&xfr->lock);
5432         if(!mesh_new_callback(env->mesh, &qinfo, qflags, &edns, buf, 0,
5433                 &auth_xfer_transfer_lookup_callback, xfr, 0)) {
5434                 lock_basic_lock(&xfr->lock);
5435                 log_err("out of memory lookup up master %s", master->host);
5436                 return 0;
5437         }
5438         lock_basic_lock(&xfr->lock);
5439         return 1;
5440 }
5441
5442 /** initiate TCP to the target and fetch zone.
5443  * returns true if that was successfully started, and timeout setup. */
5444 static int
5445 xfr_transfer_init_fetch(struct auth_xfer* xfr, struct module_env* env)
5446 {
5447         struct sockaddr_storage addr;
5448         socklen_t addrlen = 0;
5449         struct auth_master* master = xfr->task_transfer->master;
5450         char *auth_name = NULL;
5451         struct timeval t;
5452         int timeout;
5453         if(!master) return 0;
5454         if(master->allow_notify) return 0; /* only for notify */
5455
5456         /* get master addr */
5457         if(xfr->task_transfer->scan_addr) {
5458                 addrlen = xfr->task_transfer->scan_addr->addrlen;
5459                 memmove(&addr, &xfr->task_transfer->scan_addr->addr, addrlen);
5460         } else {
5461                 if(!authextstrtoaddr(master->host, &addr, &addrlen, &auth_name)) {
5462                         /* the ones that are not in addr format are supposed
5463                          * to be looked up.  The lookup has failed however,
5464                          * so skip them */
5465                         char zname[255+1];
5466                         dname_str(xfr->name, zname);
5467                         log_err("%s: failed lookup, cannot transfer from master %s",
5468                                 zname, master->host);
5469                         return 0;
5470                 }
5471         }
5472
5473         /* remove previous TCP connection (if any) */
5474         if(xfr->task_transfer->cp) {
5475                 comm_point_delete(xfr->task_transfer->cp);
5476                 xfr->task_transfer->cp = NULL;
5477         }
5478         if(!xfr->task_transfer->timer) {
5479                 xfr->task_transfer->timer = comm_timer_create(env->worker_base,
5480                         auth_xfer_transfer_timer_callback, xfr);
5481                 if(!xfr->task_transfer->timer) {
5482                         log_err("malloc failure");
5483                         return 0;
5484                 }
5485         }
5486         timeout = AUTH_TRANSFER_TIMEOUT;
5487 #ifndef S_SPLINT_S
5488         t.tv_sec = timeout/1000;
5489         t.tv_usec = (timeout%1000)*1000;
5490 #endif
5491
5492         if(master->http) {
5493                 /* perform http fetch */
5494                 /* store http port number into sockaddr,
5495                  * unless someone used unbound's host@port notation */
5496                 xfr->task_transfer->on_ixfr = 0;
5497                 if(strchr(master->host, '@') == NULL)
5498                         sockaddr_store_port(&addr, addrlen, master->port);
5499                 xfr->task_transfer->cp = outnet_comm_point_for_http(
5500                         env->outnet, auth_xfer_transfer_http_callback, xfr,
5501                         &addr, addrlen, -1, master->ssl, master->host,
5502                         master->file, env->cfg);
5503                 if(!xfr->task_transfer->cp) {
5504                         char zname[255+1], as[256];
5505                         dname_str(xfr->name, zname);
5506                         addr_to_str(&addr, addrlen, as, sizeof(as));
5507                         verbose(VERB_ALGO, "cannot create http cp "
5508                                 "connection for %s to %s", zname, as);
5509                         return 0;
5510                 }
5511                 comm_timer_set(xfr->task_transfer->timer, &t);
5512                 if(verbosity >= VERB_ALGO) {
5513                         char zname[255+1], as[256];
5514                         dname_str(xfr->name, zname);
5515                         addr_to_str(&addr, addrlen, as, sizeof(as));
5516                         verbose(VERB_ALGO, "auth zone %s transfer next HTTP fetch from %s started", zname, as);
5517                 }
5518                 /* Create or refresh the list of allow_notify addrs */
5519                 probe_copy_masters_for_allow_notify(xfr);
5520                 return 1;
5521         }
5522
5523         /* perform AXFR/IXFR */
5524         /* set the packet to be written */
5525         /* create new ID */
5526         xfr->task_transfer->id = GET_RANDOM_ID(env->rnd);
5527         xfr_create_ixfr_packet(xfr, env->scratch_buffer,
5528                 xfr->task_transfer->id, master);
5529
5530         /* connect on fd */
5531         xfr->task_transfer->cp = outnet_comm_point_for_tcp(env->outnet,
5532                 auth_xfer_transfer_tcp_callback, xfr, &addr, addrlen,
5533                 env->scratch_buffer, -1,
5534                 auth_name != NULL, auth_name);
5535         if(!xfr->task_transfer->cp) {
5536                 char zname[255+1], as[256];
5537                 dname_str(xfr->name, zname);
5538                 addr_to_str(&addr, addrlen, as, sizeof(as));
5539                 verbose(VERB_ALGO, "cannot create tcp cp connection for "
5540                         "xfr %s to %s", zname, as);
5541                 return 0;
5542         }
5543         comm_timer_set(xfr->task_transfer->timer, &t);
5544         if(verbosity >= VERB_ALGO) {
5545                 char zname[255+1], as[256];
5546                 dname_str(xfr->name, zname);
5547                 addr_to_str(&addr, addrlen, as, sizeof(as));
5548                 verbose(VERB_ALGO, "auth zone %s transfer next %s fetch from %s started", zname, 
5549                         (xfr->task_transfer->on_ixfr?"IXFR":"AXFR"), as);
5550         }
5551         return 1;
5552 }
5553
5554 /** perform next lookup, next transfer TCP, or end and resume wait time task */
5555 static void
5556 xfr_transfer_nexttarget_or_end(struct auth_xfer* xfr, struct module_env* env)
5557 {
5558         log_assert(xfr->task_transfer->worker == env->worker);
5559
5560         /* are we performing lookups? */
5561         while(xfr->task_transfer->lookup_target) {
5562                 if(xfr_transfer_lookup_host(xfr, env)) {
5563                         /* wait for lookup to finish,
5564                          * note that the hostname may be in unbound's cache
5565                          * and we may then get an instant cache response,
5566                          * and that calls the callback just like a full
5567                          * lookup and lookup failures also call callback */
5568                         if(verbosity >= VERB_ALGO) {
5569                                 char zname[255+1];
5570                                 dname_str(xfr->name, zname);
5571                                 verbose(VERB_ALGO, "auth zone %s transfer next target lookup", zname);
5572                         }
5573                         lock_basic_unlock(&xfr->lock);
5574                         return;
5575                 }
5576                 xfr_transfer_move_to_next_lookup(xfr, env);
5577         }
5578
5579         /* initiate TCP and fetch the zone from the master */
5580         /* and set timeout on it */
5581         while(!xfr_transfer_end_of_list(xfr)) {
5582                 xfr->task_transfer->master = xfr_transfer_current_master(xfr);
5583                 if(xfr_transfer_init_fetch(xfr, env)) {
5584                         /* successfully started, wait for callback */
5585                         lock_basic_unlock(&xfr->lock);
5586                         return;
5587                 }
5588                 /* failed to fetch, next master */
5589                 xfr_transfer_nextmaster(xfr);
5590         }
5591         if(verbosity >= VERB_ALGO) {
5592                 char zname[255+1];
5593                 dname_str(xfr->name, zname);
5594                 verbose(VERB_ALGO, "auth zone %s transfer failed, wait", zname);
5595         }
5596
5597         /* we failed to fetch the zone, move to wait task
5598          * use the shorter retry timeout */
5599         xfr_transfer_disown(xfr);
5600
5601         /* pick up the nextprobe task and wait */
5602         if(xfr->task_nextprobe->worker == NULL)
5603                 xfr_set_timeout(xfr, env, 1, 0);
5604         lock_basic_unlock(&xfr->lock);
5605 }
5606
5607 /** add addrs from A or AAAA rrset to the master */
5608 static void
5609 xfr_master_add_addrs(struct auth_master* m, struct ub_packed_rrset_key* rrset,
5610         uint16_t rrtype)
5611 {
5612         size_t i;
5613         struct packed_rrset_data* data;
5614         if(!m || !rrset) return;
5615         if(rrtype != LDNS_RR_TYPE_A && rrtype != LDNS_RR_TYPE_AAAA)
5616                 return;
5617         data = (struct packed_rrset_data*)rrset->entry.data;
5618         for(i=0; i<data->count; i++) {
5619                 struct auth_addr* a;
5620                 size_t len = data->rr_len[i] - 2;
5621                 uint8_t* rdata = data->rr_data[i]+2;
5622                 if(rrtype == LDNS_RR_TYPE_A && len != INET_SIZE)
5623                         continue; /* wrong length for A */
5624                 if(rrtype == LDNS_RR_TYPE_AAAA && len != INET6_SIZE)
5625                         continue; /* wrong length for AAAA */
5626                 
5627                 /* add and alloc it */
5628                 a = (struct auth_addr*)calloc(1, sizeof(*a));
5629                 if(!a) {
5630                         log_err("out of memory");
5631                         return;
5632                 }
5633                 if(rrtype == LDNS_RR_TYPE_A) {
5634                         struct sockaddr_in* sa;
5635                         a->addrlen = (socklen_t)sizeof(*sa);
5636                         sa = (struct sockaddr_in*)&a->addr;
5637                         sa->sin_family = AF_INET;
5638                         sa->sin_port = (in_port_t)htons(UNBOUND_DNS_PORT);
5639                         memmove(&sa->sin_addr, rdata, INET_SIZE);
5640                 } else {
5641                         struct sockaddr_in6* sa;
5642                         a->addrlen = (socklen_t)sizeof(*sa);
5643                         sa = (struct sockaddr_in6*)&a->addr;
5644                         sa->sin6_family = AF_INET6;
5645                         sa->sin6_port = (in_port_t)htons(UNBOUND_DNS_PORT);
5646                         memmove(&sa->sin6_addr, rdata, INET6_SIZE);
5647                 }
5648                 if(verbosity >= VERB_ALGO) {
5649                         char s[64];
5650                         addr_to_str(&a->addr, a->addrlen, s, sizeof(s));
5651                         verbose(VERB_ALGO, "auth host %s lookup %s",
5652                                 m->host, s);
5653                 }
5654                 /* append to list */
5655                 a->next = m->list;
5656                 m->list = a;
5657         }
5658 }
5659
5660 /** callback for task_transfer lookup of host name, of A or AAAA */
5661 void auth_xfer_transfer_lookup_callback(void* arg, int rcode, sldns_buffer* buf,
5662         enum sec_status ATTR_UNUSED(sec), char* ATTR_UNUSED(why_bogus),
5663         int ATTR_UNUSED(was_ratelimited))
5664 {
5665         struct auth_xfer* xfr = (struct auth_xfer*)arg;
5666         struct module_env* env;
5667         log_assert(xfr->task_transfer);
5668         lock_basic_lock(&xfr->lock);
5669         env = xfr->task_transfer->env;
5670         if(!env || env->outnet->want_to_quit) {
5671                 lock_basic_unlock(&xfr->lock);
5672                 return; /* stop on quit */
5673         }
5674
5675         /* process result */
5676         if(rcode == LDNS_RCODE_NOERROR) {
5677                 uint16_t wanted_qtype = LDNS_RR_TYPE_A;
5678                 struct regional* temp = env->scratch;
5679                 struct query_info rq;
5680                 struct reply_info* rep;
5681                 if(xfr->task_transfer->lookup_aaaa)
5682                         wanted_qtype = LDNS_RR_TYPE_AAAA;
5683                 memset(&rq, 0, sizeof(rq));
5684                 rep = parse_reply_in_temp_region(buf, temp, &rq);
5685                 if(rep && rq.qtype == wanted_qtype &&
5686                         FLAGS_GET_RCODE(rep->flags) == LDNS_RCODE_NOERROR) {
5687                         /* parsed successfully */
5688                         struct ub_packed_rrset_key* answer =
5689                                 reply_find_answer_rrset(&rq, rep);
5690                         if(answer) {
5691                                 xfr_master_add_addrs(xfr->task_transfer->
5692                                         lookup_target, answer, wanted_qtype);
5693                         } else {
5694                                 if(verbosity >= VERB_ALGO) {
5695                                         char zname[255+1];
5696                                         dname_str(xfr->name, zname);
5697                                         verbose(VERB_ALGO, "auth zone %s host %s type %s transfer lookup has nodata", zname, xfr->task_transfer->lookup_target->host, (xfr->task_transfer->lookup_aaaa?"AAAA":"A"));
5698                                 }
5699                         }
5700                 } else {
5701                         if(verbosity >= VERB_ALGO) {
5702                                 char zname[255+1];
5703                                 dname_str(xfr->name, zname);
5704                                 verbose(VERB_ALGO, "auth zone %s host %s type %s transfer lookup has no answer", zname, xfr->task_transfer->lookup_target->host, (xfr->task_transfer->lookup_aaaa?"AAAA":"A"));
5705                         }
5706                 }
5707                 regional_free_all(temp);
5708         } else {
5709                 if(verbosity >= VERB_ALGO) {
5710                         char zname[255+1];
5711                         dname_str(xfr->name, zname);
5712                         verbose(VERB_ALGO, "auth zone %s host %s type %s transfer lookup failed", zname, xfr->task_transfer->lookup_target->host, (xfr->task_transfer->lookup_aaaa?"AAAA":"A"));
5713                 }
5714         }
5715         if(xfr->task_transfer->lookup_target->list &&
5716                 xfr->task_transfer->lookup_target == xfr_transfer_current_master(xfr))
5717                 xfr->task_transfer->scan_addr = xfr->task_transfer->lookup_target->list;
5718
5719         /* move to lookup AAAA after A lookup, move to next hostname lookup,
5720          * or move to fetch the zone, or, if nothing to do, end task_transfer */
5721         xfr_transfer_move_to_next_lookup(xfr, env);
5722         xfr_transfer_nexttarget_or_end(xfr, env);
5723 }
5724
5725 /** check if xfer (AXFR or IXFR) packet is OK.
5726  * return false if we lost connection (SERVFAIL, or unreadable).
5727  * return false if we need to move from IXFR to AXFR, with gonextonfail
5728  *      set to false, so the same master is tried again, but with AXFR.
5729  * return true if fine to link into data.
5730  * return true with transferdone=true when the transfer has ended.
5731  */
5732 static int
5733 check_xfer_packet(sldns_buffer* pkt, struct auth_xfer* xfr,
5734         int* gonextonfail, int* transferdone)
5735 {
5736         uint8_t* wire = sldns_buffer_begin(pkt);
5737         int i;
5738         if(sldns_buffer_limit(pkt) < LDNS_HEADER_SIZE) {
5739                 verbose(VERB_ALGO, "xfr to %s failed, packet too small",
5740                         xfr->task_transfer->master->host);
5741                 return 0;
5742         }
5743         if(!LDNS_QR_WIRE(wire)) {
5744                 verbose(VERB_ALGO, "xfr to %s failed, packet has no QR flag",
5745                         xfr->task_transfer->master->host);
5746                 return 0;
5747         }
5748         if(LDNS_TC_WIRE(wire)) {
5749                 verbose(VERB_ALGO, "xfr to %s failed, packet has TC flag",
5750                         xfr->task_transfer->master->host);
5751                 return 0;
5752         }
5753         /* check ID */
5754         if(LDNS_ID_WIRE(wire) != xfr->task_transfer->id) {
5755                 verbose(VERB_ALGO, "xfr to %s failed, packet wrong ID",
5756                         xfr->task_transfer->master->host);
5757                 return 0;
5758         }
5759         if(LDNS_RCODE_WIRE(wire) != LDNS_RCODE_NOERROR) {
5760                 char rcode[32];
5761                 sldns_wire2str_rcode_buf((int)LDNS_RCODE_WIRE(wire), rcode,
5762                         sizeof(rcode));
5763                 /* if we are doing IXFR, check for fallback */
5764                 if(xfr->task_transfer->on_ixfr) {
5765                         if(LDNS_RCODE_WIRE(wire) == LDNS_RCODE_NOTIMPL ||
5766                                 LDNS_RCODE_WIRE(wire) == LDNS_RCODE_SERVFAIL ||
5767                                 LDNS_RCODE_WIRE(wire) == LDNS_RCODE_REFUSED ||
5768                                 LDNS_RCODE_WIRE(wire) == LDNS_RCODE_FORMERR) {
5769                                 verbose(VERB_ALGO, "xfr to %s, fallback "
5770                                         "from IXFR to AXFR (with rcode %s)",
5771                                         xfr->task_transfer->master->host,
5772                                         rcode);
5773                                 xfr->task_transfer->ixfr_fail = 1;
5774                                 *gonextonfail = 0;
5775                                 return 0;
5776                         }
5777                 }
5778                 verbose(VERB_ALGO, "xfr to %s failed, packet with rcode %s",
5779                         xfr->task_transfer->master->host, rcode);
5780                 return 0;
5781         }
5782         if(LDNS_OPCODE_WIRE(wire) != LDNS_PACKET_QUERY) {
5783                 verbose(VERB_ALGO, "xfr to %s failed, packet with bad opcode",
5784                         xfr->task_transfer->master->host);
5785                 return 0;
5786         }
5787         if(LDNS_QDCOUNT(wire) > 1) {
5788                 verbose(VERB_ALGO, "xfr to %s failed, packet has qdcount %d",
5789                         xfr->task_transfer->master->host,
5790                         (int)LDNS_QDCOUNT(wire));
5791                 return 0;
5792         }
5793
5794         /* check qname */
5795         sldns_buffer_set_position(pkt, LDNS_HEADER_SIZE);
5796         for(i=0; i<(int)LDNS_QDCOUNT(wire); i++) {
5797                 size_t pos = sldns_buffer_position(pkt);
5798                 uint16_t qtype, qclass;
5799                 if(pkt_dname_len(pkt) == 0) {
5800                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5801                                 "malformed dname",
5802                                 xfr->task_transfer->master->host);
5803                         return 0;
5804                 }
5805                 if(dname_pkt_compare(pkt, sldns_buffer_at(pkt, pos),
5806                         xfr->name) != 0) {
5807                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5808                                 "wrong qname",
5809                                 xfr->task_transfer->master->host);
5810                         return 0;
5811                 }
5812                 if(sldns_buffer_remaining(pkt) < 4) {
5813                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5814                                 "truncated query RR",
5815                                 xfr->task_transfer->master->host);
5816                         return 0;
5817                 }
5818                 qtype = sldns_buffer_read_u16(pkt);
5819                 qclass = sldns_buffer_read_u16(pkt);
5820                 if(qclass != xfr->dclass) {
5821                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5822                                 "wrong qclass",
5823                                 xfr->task_transfer->master->host);
5824                         return 0;
5825                 }
5826                 if(xfr->task_transfer->on_ixfr) {
5827                         if(qtype != LDNS_RR_TYPE_IXFR) {
5828                                 verbose(VERB_ALGO, "xfr to %s failed, packet "
5829                                         "with wrong qtype, expected IXFR",
5830                                 xfr->task_transfer->master->host);
5831                                 return 0;
5832                         }
5833                 } else {
5834                         if(qtype != LDNS_RR_TYPE_AXFR) {
5835                                 verbose(VERB_ALGO, "xfr to %s failed, packet "
5836                                         "with wrong qtype, expected AXFR",
5837                                 xfr->task_transfer->master->host);
5838                                 return 0;
5839                         }
5840                 }
5841         }
5842
5843         /* check parse of RRs in packet, store first SOA serial
5844          * to be able to detect last SOA (with that serial) to see if done */
5845         /* also check for IXFR 'zone up to date' reply */
5846         for(i=0; i<(int)LDNS_ANCOUNT(wire); i++) {
5847                 size_t pos = sldns_buffer_position(pkt);
5848                 uint16_t tp, rdlen;
5849                 if(pkt_dname_len(pkt) == 0) {
5850                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5851                                 "malformed dname in answer section",
5852                                 xfr->task_transfer->master->host);
5853                         return 0;
5854                 }
5855                 if(sldns_buffer_remaining(pkt) < 10) {
5856                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5857                                 "truncated RR",
5858                                 xfr->task_transfer->master->host);
5859                         return 0;
5860                 }
5861                 tp = sldns_buffer_read_u16(pkt);
5862                 (void)sldns_buffer_read_u16(pkt); /* class */
5863                 (void)sldns_buffer_read_u32(pkt); /* ttl */
5864                 rdlen = sldns_buffer_read_u16(pkt);
5865                 if(sldns_buffer_remaining(pkt) < rdlen) {
5866                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5867                                 "truncated RR rdata",
5868                                 xfr->task_transfer->master->host);
5869                         return 0;
5870                 }
5871
5872                 /* RR parses (haven't checked rdata itself), now look at
5873                  * SOA records to see serial number */
5874                 if(xfr->task_transfer->rr_scan_num == 0 &&
5875                         tp != LDNS_RR_TYPE_SOA) {
5876                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5877                                 "malformed zone transfer, no start SOA",
5878                                 xfr->task_transfer->master->host);
5879                         return 0;
5880                 }
5881                 if(xfr->task_transfer->rr_scan_num == 1 &&
5882                         tp != LDNS_RR_TYPE_SOA) {
5883                         /* second RR is not a SOA record, this is not an IXFR
5884                          * the master is replying with an AXFR */
5885                         xfr->task_transfer->on_ixfr_is_axfr = 1;
5886                 }
5887                 if(tp == LDNS_RR_TYPE_SOA) {
5888                         uint32_t serial;
5889                         if(rdlen < 22) {
5890                                 verbose(VERB_ALGO, "xfr to %s failed, packet "
5891                                         "with SOA with malformed rdata",
5892                                         xfr->task_transfer->master->host);
5893                                 return 0;
5894                         }
5895                         if(dname_pkt_compare(pkt, sldns_buffer_at(pkt, pos),
5896                                 xfr->name) != 0) {
5897                                 verbose(VERB_ALGO, "xfr to %s failed, packet "
5898                                         "with SOA with wrong dname",
5899                                         xfr->task_transfer->master->host);
5900                                 return 0;
5901                         }
5902
5903                         /* read serial number of SOA */
5904                         serial = sldns_buffer_read_u32_at(pkt,
5905                                 sldns_buffer_position(pkt)+rdlen-20);
5906
5907                         /* check for IXFR 'zone has SOA x' reply */
5908                         if(xfr->task_transfer->on_ixfr &&
5909                                 xfr->task_transfer->rr_scan_num == 0 &&
5910                                 LDNS_ANCOUNT(wire)==1) {
5911                                 verbose(VERB_ALGO, "xfr to %s ended, "
5912                                         "IXFR reply that zone has serial %u,"
5913                                         " fallback from IXFR to AXFR",
5914                                         xfr->task_transfer->master->host,
5915                                         (unsigned)serial);
5916                                 xfr->task_transfer->ixfr_fail = 1;
5917                                 *gonextonfail = 0;
5918                                 return 0;
5919                         }
5920
5921                         /* if first SOA, store serial number */
5922                         if(xfr->task_transfer->got_xfr_serial == 0) {
5923                                 xfr->task_transfer->got_xfr_serial = 1;
5924                                 xfr->task_transfer->incoming_xfr_serial =
5925                                         serial;
5926                                 verbose(VERB_ALGO, "xfr %s: contains "
5927                                         "SOA serial %u",
5928                                         xfr->task_transfer->master->host,
5929                                         (unsigned)serial);
5930                         /* see if end of AXFR */
5931                         } else if(!xfr->task_transfer->on_ixfr ||
5932                                 xfr->task_transfer->on_ixfr_is_axfr) {
5933                                 /* second SOA with serial is the end
5934                                  * for AXFR */
5935                                 *transferdone = 1;
5936                                 verbose(VERB_ALGO, "xfr %s: last AXFR packet",
5937                                         xfr->task_transfer->master->host);
5938                         /* for IXFR, count SOA records with that serial */
5939                         } else if(xfr->task_transfer->incoming_xfr_serial ==
5940                                 serial && xfr->task_transfer->got_xfr_serial
5941                                 == 1) {
5942                                 xfr->task_transfer->got_xfr_serial++;
5943                         /* if not first soa, if serial==firstserial, the
5944                          * third time we are at the end, for IXFR */
5945                         } else if(xfr->task_transfer->incoming_xfr_serial ==
5946                                 serial && xfr->task_transfer->got_xfr_serial
5947                                 == 2) {
5948                                 verbose(VERB_ALGO, "xfr %s: last IXFR packet",
5949                                         xfr->task_transfer->master->host);
5950                                 *transferdone = 1;
5951                                 /* continue parse check, if that succeeds,
5952                                  * transfer is done */
5953                         }
5954                 }
5955                 xfr->task_transfer->rr_scan_num++;
5956
5957                 /* skip over RR rdata to go to the next RR */
5958                 sldns_buffer_skip(pkt, (ssize_t)rdlen);
5959         }
5960
5961         /* check authority section */
5962         /* we skip over the RRs checking packet format */
5963         for(i=0; i<(int)LDNS_NSCOUNT(wire); i++) {
5964                 uint16_t rdlen;
5965                 if(pkt_dname_len(pkt) == 0) {
5966                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5967                                 "malformed dname in authority section",
5968                                 xfr->task_transfer->master->host);
5969                         return 0;
5970                 }
5971                 if(sldns_buffer_remaining(pkt) < 10) {
5972                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5973                                 "truncated RR",
5974                                 xfr->task_transfer->master->host);
5975                         return 0;
5976                 }
5977                 (void)sldns_buffer_read_u16(pkt); /* type */
5978                 (void)sldns_buffer_read_u16(pkt); /* class */
5979                 (void)sldns_buffer_read_u32(pkt); /* ttl */
5980                 rdlen = sldns_buffer_read_u16(pkt);
5981                 if(sldns_buffer_remaining(pkt) < rdlen) {
5982                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5983                                 "truncated RR rdata",
5984                                 xfr->task_transfer->master->host);
5985                         return 0;
5986                 }
5987                 /* skip over RR rdata to go to the next RR */
5988                 sldns_buffer_skip(pkt, (ssize_t)rdlen);
5989         }
5990
5991         /* check additional section */
5992         for(i=0; i<(int)LDNS_ARCOUNT(wire); i++) {
5993                 uint16_t rdlen;
5994                 if(pkt_dname_len(pkt) == 0) {
5995                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5996                                 "malformed dname in additional section",
5997                                 xfr->task_transfer->master->host);
5998                         return 0;
5999                 }
6000                 if(sldns_buffer_remaining(pkt) < 10) {
6001                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
6002                                 "truncated RR",
6003                                 xfr->task_transfer->master->host);
6004                         return 0;
6005                 }
6006                 (void)sldns_buffer_read_u16(pkt); /* type */
6007                 (void)sldns_buffer_read_u16(pkt); /* class */
6008                 (void)sldns_buffer_read_u32(pkt); /* ttl */
6009                 rdlen = sldns_buffer_read_u16(pkt);
6010                 if(sldns_buffer_remaining(pkt) < rdlen) {
6011                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
6012                                 "truncated RR rdata",
6013                                 xfr->task_transfer->master->host);
6014                         return 0;
6015                 }
6016                 /* skip over RR rdata to go to the next RR */
6017                 sldns_buffer_skip(pkt, (ssize_t)rdlen);
6018         }
6019
6020         return 1;
6021 }
6022
6023 /** Link the data from this packet into the worklist of transferred data */
6024 static int
6025 xfer_link_data(sldns_buffer* pkt, struct auth_xfer* xfr)
6026 {
6027         /* alloc it */
6028         struct auth_chunk* e;
6029         e = (struct auth_chunk*)calloc(1, sizeof(*e));
6030         if(!e) return 0;
6031         e->next = NULL;
6032         e->len = sldns_buffer_limit(pkt);
6033         e->data = memdup(sldns_buffer_begin(pkt), e->len);
6034         if(!e->data) {
6035                 free(e);
6036                 return 0;
6037         }
6038
6039         /* alloc succeeded, link into list */
6040         if(!xfr->task_transfer->chunks_first)
6041                 xfr->task_transfer->chunks_first = e;
6042         if(xfr->task_transfer->chunks_last)
6043                 xfr->task_transfer->chunks_last->next = e;
6044         xfr->task_transfer->chunks_last = e;
6045         return 1;
6046 }
6047
6048 /** task transfer.  the list of data is complete. process it and if failed
6049  * move to next master, if succeeded, end the task transfer */
6050 static void
6051 process_list_end_transfer(struct auth_xfer* xfr, struct module_env* env)
6052 {
6053         int ixfr_fail = 0;
6054         if(xfr_process_chunk_list(xfr, env, &ixfr_fail)) {
6055                 /* it worked! */
6056                 auth_chunks_delete(xfr->task_transfer);
6057
6058                 /* we fetched the zone, move to wait task */
6059                 xfr_transfer_disown(xfr);
6060
6061                 if(xfr->notify_received && (!xfr->notify_has_serial ||
6062                         (xfr->notify_has_serial && 
6063                         xfr_serial_means_update(xfr, xfr->notify_serial)))) {
6064                         uint32_t sr = xfr->notify_serial;
6065                         int has_sr = xfr->notify_has_serial;
6066                         /* we received a notify while probe/transfer was
6067                          * in progress.  start a new probe and transfer */
6068                         xfr->notify_received = 0;
6069                         xfr->notify_has_serial = 0;
6070                         xfr->notify_serial = 0;
6071                         if(!xfr_start_probe(xfr, env, NULL)) {
6072                                 /* if we couldn't start it, already in
6073                                  * progress; restore notify serial,
6074                                  * while xfr still locked */
6075                                 xfr->notify_received = 1;
6076                                 xfr->notify_has_serial = has_sr;
6077                                 xfr->notify_serial = sr;
6078                                 lock_basic_unlock(&xfr->lock);
6079                         }
6080                         return;
6081                 } else {
6082                         /* pick up the nextprobe task and wait (normail wait time) */
6083                         if(xfr->task_nextprobe->worker == NULL)
6084                                 xfr_set_timeout(xfr, env, 0, 0);
6085                 }
6086                 lock_basic_unlock(&xfr->lock);
6087                 return;
6088         }
6089         /* processing failed */
6090         /* when done, delete data from list */
6091         auth_chunks_delete(xfr->task_transfer);
6092         if(ixfr_fail) {
6093                 xfr->task_transfer->ixfr_fail = 1;
6094         } else {
6095                 xfr_transfer_nextmaster(xfr);
6096         }
6097         xfr_transfer_nexttarget_or_end(xfr, env);
6098 }
6099
6100 /** callback for the task_transfer timer */
6101 void
6102 auth_xfer_transfer_timer_callback(void* arg)
6103 {
6104         struct auth_xfer* xfr = (struct auth_xfer*)arg;
6105         struct module_env* env;
6106         int gonextonfail = 1;
6107         log_assert(xfr->task_transfer);
6108         lock_basic_lock(&xfr->lock);
6109         env = xfr->task_transfer->env;
6110         if(!env || env->outnet->want_to_quit) {
6111                 lock_basic_unlock(&xfr->lock);
6112                 return; /* stop on quit */
6113         }
6114
6115         verbose(VERB_ALGO, "xfr stopped, connection timeout to %s",
6116                 xfr->task_transfer->master->host);
6117
6118         /* see if IXFR caused the failure, if so, try AXFR */
6119         if(xfr->task_transfer->on_ixfr) {
6120                 xfr->task_transfer->ixfr_possible_timeout_count++;
6121                 if(xfr->task_transfer->ixfr_possible_timeout_count >=
6122                         NUM_TIMEOUTS_FALLBACK_IXFR) {
6123                         verbose(VERB_ALGO, "xfr to %s, fallback "
6124                                 "from IXFR to AXFR (because of timeouts)",
6125                                 xfr->task_transfer->master->host);
6126                         xfr->task_transfer->ixfr_fail = 1;
6127                         gonextonfail = 0;
6128                 }
6129         }
6130
6131         /* delete transferred data from list */
6132         auth_chunks_delete(xfr->task_transfer);
6133         comm_point_delete(xfr->task_transfer->cp);
6134         xfr->task_transfer->cp = NULL;
6135         if(gonextonfail)
6136                 xfr_transfer_nextmaster(xfr);
6137         xfr_transfer_nexttarget_or_end(xfr, env);
6138 }
6139
6140 /** callback for task_transfer tcp connections */
6141 int
6142 auth_xfer_transfer_tcp_callback(struct comm_point* c, void* arg, int err,
6143         struct comm_reply* ATTR_UNUSED(repinfo))
6144 {
6145         struct auth_xfer* xfr = (struct auth_xfer*)arg;
6146         struct module_env* env;
6147         int gonextonfail = 1;
6148         int transferdone = 0;
6149         log_assert(xfr->task_transfer);
6150         lock_basic_lock(&xfr->lock);
6151         env = xfr->task_transfer->env;
6152         if(!env || env->outnet->want_to_quit) {
6153                 lock_basic_unlock(&xfr->lock);
6154                 return 0; /* stop on quit */
6155         }
6156         /* stop the timer */
6157         comm_timer_disable(xfr->task_transfer->timer);
6158
6159         if(err != NETEVENT_NOERROR) {
6160                 /* connection failed, closed, or timeout */
6161                 /* stop this transfer, cleanup 
6162                  * and continue task_transfer*/
6163                 verbose(VERB_ALGO, "xfr stopped, connection lost to %s",
6164                         xfr->task_transfer->master->host);
6165
6166                 /* see if IXFR caused the failure, if so, try AXFR */
6167                 if(xfr->task_transfer->on_ixfr) {
6168                         xfr->task_transfer->ixfr_possible_timeout_count++;
6169                         if(xfr->task_transfer->ixfr_possible_timeout_count >=
6170                                 NUM_TIMEOUTS_FALLBACK_IXFR) {
6171                                 verbose(VERB_ALGO, "xfr to %s, fallback "
6172                                         "from IXFR to AXFR (because of timeouts)",
6173                                         xfr->task_transfer->master->host);
6174                                 xfr->task_transfer->ixfr_fail = 1;
6175                                 gonextonfail = 0;
6176                         }
6177                 }
6178
6179         failed:
6180                 /* delete transferred data from list */
6181                 auth_chunks_delete(xfr->task_transfer);
6182                 comm_point_delete(xfr->task_transfer->cp);
6183                 xfr->task_transfer->cp = NULL;
6184                 if(gonextonfail)
6185                         xfr_transfer_nextmaster(xfr);
6186                 xfr_transfer_nexttarget_or_end(xfr, env);
6187                 return 0;
6188         }
6189         /* note that IXFR worked without timeout */
6190         if(xfr->task_transfer->on_ixfr)
6191                 xfr->task_transfer->ixfr_possible_timeout_count = 0;
6192
6193         /* handle returned packet */
6194         /* if it fails, cleanup and end this transfer */
6195         /* if it needs to fallback from IXFR to AXFR, do that */
6196         if(!check_xfer_packet(c->buffer, xfr, &gonextonfail, &transferdone)) {
6197                 goto failed;
6198         }
6199         /* if it is good, link it into the list of data */
6200         /* if the link into list of data fails (malloc fail) cleanup and end */
6201         if(!xfer_link_data(c->buffer, xfr)) {
6202                 verbose(VERB_ALGO, "xfr stopped to %s, malloc failed",
6203                         xfr->task_transfer->master->host);
6204                 goto failed;
6205         }
6206         /* if the transfer is done now, disconnect and process the list */
6207         if(transferdone) {
6208                 comm_point_delete(xfr->task_transfer->cp);
6209                 xfr->task_transfer->cp = NULL;
6210                 process_list_end_transfer(xfr, env);
6211                 return 0;
6212         }
6213
6214         /* if we want to read more messages, setup the commpoint to read
6215          * a DNS packet, and the timeout */
6216         lock_basic_unlock(&xfr->lock);
6217         c->tcp_is_reading = 1;
6218         sldns_buffer_clear(c->buffer);
6219         comm_point_start_listening(c, -1, AUTH_TRANSFER_TIMEOUT);
6220         return 0;
6221 }
6222
6223 /** callback for task_transfer http connections */
6224 int
6225 auth_xfer_transfer_http_callback(struct comm_point* c, void* arg, int err,
6226         struct comm_reply* repinfo)
6227 {
6228         struct auth_xfer* xfr = (struct auth_xfer*)arg;
6229         struct module_env* env;
6230         log_assert(xfr->task_transfer);
6231         lock_basic_lock(&xfr->lock);
6232         env = xfr->task_transfer->env;
6233         if(!env || env->outnet->want_to_quit) {
6234                 lock_basic_unlock(&xfr->lock);
6235                 return 0; /* stop on quit */
6236         }
6237         verbose(VERB_ALGO, "auth zone transfer http callback");
6238         /* stop the timer */
6239         comm_timer_disable(xfr->task_transfer->timer);
6240
6241         if(err != NETEVENT_NOERROR && err != NETEVENT_DONE) {
6242                 /* connection failed, closed, or timeout */
6243                 /* stop this transfer, cleanup 
6244                  * and continue task_transfer*/
6245                 verbose(VERB_ALGO, "http stopped, connection lost to %s",
6246                         xfr->task_transfer->master->host);
6247         failed:
6248                 /* delete transferred data from list */
6249                 auth_chunks_delete(xfr->task_transfer);
6250                 if(repinfo) repinfo->c = NULL; /* signal cp deleted to
6251                                 the routine calling this callback */
6252                 comm_point_delete(xfr->task_transfer->cp);
6253                 xfr->task_transfer->cp = NULL;
6254                 xfr_transfer_nextmaster(xfr);
6255                 xfr_transfer_nexttarget_or_end(xfr, env);
6256                 return 0;
6257         }
6258
6259         /* if it is good, link it into the list of data */
6260         /* if the link into list of data fails (malloc fail) cleanup and end */
6261         if(sldns_buffer_limit(c->buffer) > 0) {
6262                 verbose(VERB_ALGO, "auth zone http queued up %d bytes",
6263                         (int)sldns_buffer_limit(c->buffer));
6264                 if(!xfer_link_data(c->buffer, xfr)) {
6265                         verbose(VERB_ALGO, "http stopped to %s, malloc failed",
6266                                 xfr->task_transfer->master->host);
6267                         goto failed;
6268                 }
6269         }
6270         /* if the transfer is done now, disconnect and process the list */
6271         if(err == NETEVENT_DONE) {
6272                 if(repinfo) repinfo->c = NULL; /* signal cp deleted to
6273                                 the routine calling this callback */
6274                 comm_point_delete(xfr->task_transfer->cp);
6275                 xfr->task_transfer->cp = NULL;
6276                 process_list_end_transfer(xfr, env);
6277                 return 0;
6278         }
6279
6280         /* if we want to read more messages, setup the commpoint to read
6281          * a DNS packet, and the timeout */
6282         lock_basic_unlock(&xfr->lock);
6283         c->tcp_is_reading = 1;
6284         sldns_buffer_clear(c->buffer);
6285         comm_point_start_listening(c, -1, AUTH_TRANSFER_TIMEOUT);
6286         return 0;
6287 }
6288
6289
6290 /** start transfer task by this worker , xfr is locked. */
6291 static void
6292 xfr_start_transfer(struct auth_xfer* xfr, struct module_env* env,
6293         struct auth_master* master)
6294 {
6295         log_assert(xfr->task_transfer != NULL);
6296         log_assert(xfr->task_transfer->worker == NULL);
6297         log_assert(xfr->task_transfer->chunks_first == NULL);
6298         log_assert(xfr->task_transfer->chunks_last == NULL);
6299         xfr->task_transfer->worker = env->worker;
6300         xfr->task_transfer->env = env;
6301
6302         /* init transfer process */
6303         /* find that master in the transfer's list of masters? */
6304         xfr_transfer_start_list(xfr, master);
6305         /* start lookup for hostnames in transfer master list */
6306         xfr_transfer_start_lookups(xfr);
6307
6308         /* initiate TCP, and set timeout on it */
6309         xfr_transfer_nexttarget_or_end(xfr, env);
6310 }
6311
6312 /** disown task_probe.  caller must hold xfr.lock */
6313 static void
6314 xfr_probe_disown(struct auth_xfer* xfr)
6315 {
6316         /* remove timer (from this worker's event base) */
6317         comm_timer_delete(xfr->task_probe->timer);
6318         xfr->task_probe->timer = NULL;
6319         /* remove the commpoint */
6320         comm_point_delete(xfr->task_probe->cp);
6321         xfr->task_probe->cp = NULL;
6322         /* we don't own this item anymore */
6323         xfr->task_probe->worker = NULL;
6324         xfr->task_probe->env = NULL;
6325 }
6326
6327 /** send the UDP probe to the master, this is part of task_probe */
6328 static int
6329 xfr_probe_send_probe(struct auth_xfer* xfr, struct module_env* env,
6330         int timeout)
6331 {
6332         struct sockaddr_storage addr;
6333         socklen_t addrlen = 0;
6334         struct timeval t;
6335         /* pick master */
6336         struct auth_master* master = xfr_probe_current_master(xfr);
6337         char *auth_name = NULL;
6338         if(!master) return 0;
6339         if(master->allow_notify) return 0; /* only for notify */
6340         if(master->http) return 0; /* only masters get SOA UDP probe,
6341                 not urls, if those are in this list */
6342
6343         /* get master addr */
6344         if(xfr->task_probe->scan_addr) {
6345                 addrlen = xfr->task_probe->scan_addr->addrlen;
6346                 memmove(&addr, &xfr->task_probe->scan_addr->addr, addrlen);
6347         } else {
6348                 if(!authextstrtoaddr(master->host, &addr, &addrlen, &auth_name)) {
6349                         /* the ones that are not in addr format are supposed
6350                          * to be looked up.  The lookup has failed however,
6351                          * so skip them */
6352                         char zname[255+1];
6353                         dname_str(xfr->name, zname);
6354                         log_err("%s: failed lookup, cannot probe to master %s",
6355                                 zname, master->host);
6356                         return 0;
6357                 }
6358                 if (auth_name != NULL) {
6359                         if (addr.ss_family == AF_INET
6360                         &&  (int)ntohs(((struct sockaddr_in *)&addr)->sin_port)
6361                             == env->cfg->ssl_port)
6362                                 ((struct sockaddr_in *)&addr)->sin_port
6363                                         = htons((uint16_t)env->cfg->port);
6364                         else if (addr.ss_family == AF_INET6
6365                         &&  (int)ntohs(((struct sockaddr_in6 *)&addr)->sin6_port)
6366                             == env->cfg->ssl_port)
6367                                 ((struct sockaddr_in6 *)&addr)->sin6_port
6368                                         = htons((uint16_t)env->cfg->port);
6369                 }
6370         }
6371
6372         /* create packet */
6373         /* create new ID for new probes, but not on timeout retries,
6374          * this means we'll accept replies to previous retries to same ip */
6375         if(timeout == AUTH_PROBE_TIMEOUT)
6376                 xfr->task_probe->id = GET_RANDOM_ID(env->rnd);
6377         xfr_create_soa_probe_packet(xfr, env->scratch_buffer, 
6378                 xfr->task_probe->id);
6379         /* we need to remove the cp if we have a different ip4/ip6 type now */
6380         if(xfr->task_probe->cp &&
6381                 ((xfr->task_probe->cp_is_ip6 && !addr_is_ip6(&addr, addrlen)) ||
6382                 (!xfr->task_probe->cp_is_ip6 && addr_is_ip6(&addr, addrlen)))
6383                 ) {
6384                 comm_point_delete(xfr->task_probe->cp);
6385                 xfr->task_probe->cp = NULL;
6386         }
6387         if(!xfr->task_probe->cp) {
6388                 if(addr_is_ip6(&addr, addrlen))
6389                         xfr->task_probe->cp_is_ip6 = 1;
6390                 else    xfr->task_probe->cp_is_ip6 = 0;
6391                 xfr->task_probe->cp = outnet_comm_point_for_udp(env->outnet,
6392                         auth_xfer_probe_udp_callback, xfr, &addr, addrlen);
6393                 if(!xfr->task_probe->cp) {
6394                         char zname[255+1], as[256];
6395                         dname_str(xfr->name, zname);
6396                         addr_to_str(&addr, addrlen, as, sizeof(as));
6397                         verbose(VERB_ALGO, "cannot create udp cp for "
6398                                 "probe %s to %s", zname, as);
6399                         return 0;
6400                 }
6401         }
6402         if(!xfr->task_probe->timer) {
6403                 xfr->task_probe->timer = comm_timer_create(env->worker_base,
6404                         auth_xfer_probe_timer_callback, xfr);
6405                 if(!xfr->task_probe->timer) {
6406                         log_err("malloc failure");
6407                         return 0;
6408                 }
6409         }
6410
6411         /* send udp packet */
6412         if(!comm_point_send_udp_msg(xfr->task_probe->cp, env->scratch_buffer,
6413                 (struct sockaddr*)&addr, addrlen, 0)) {
6414                 char zname[255+1], as[256];
6415                 dname_str(xfr->name, zname);
6416                 addr_to_str(&addr, addrlen, as, sizeof(as));
6417                 verbose(VERB_ALGO, "failed to send soa probe for %s to %s",
6418                         zname, as);
6419                 return 0;
6420         }
6421         if(verbosity >= VERB_ALGO) {
6422                 char zname[255+1], as[256];
6423                 dname_str(xfr->name, zname);
6424                 addr_to_str(&addr, addrlen, as, sizeof(as));
6425                 verbose(VERB_ALGO, "auth zone %s soa probe sent to %s", zname,
6426                         as);
6427         }
6428         xfr->task_probe->timeout = timeout;
6429 #ifndef S_SPLINT_S
6430         t.tv_sec = timeout/1000;
6431         t.tv_usec = (timeout%1000)*1000;
6432 #endif
6433         comm_timer_set(xfr->task_probe->timer, &t);
6434
6435         return 1;
6436 }
6437
6438 /** callback for task_probe timer */
6439 void
6440 auth_xfer_probe_timer_callback(void* arg)
6441 {
6442         struct auth_xfer* xfr = (struct auth_xfer*)arg;
6443         struct module_env* env;
6444         log_assert(xfr->task_probe);
6445         lock_basic_lock(&xfr->lock);
6446         env = xfr->task_probe->env;
6447         if(!env || env->outnet->want_to_quit) {
6448                 lock_basic_unlock(&xfr->lock);
6449                 return; /* stop on quit */
6450         }
6451
6452         if(verbosity >= VERB_ALGO) {
6453                 char zname[255+1];
6454                 dname_str(xfr->name, zname);
6455                 verbose(VERB_ALGO, "auth zone %s soa probe timeout", zname);
6456         }
6457         if(xfr->task_probe->timeout <= AUTH_PROBE_TIMEOUT_STOP) {
6458                 /* try again with bigger timeout */
6459                 if(xfr_probe_send_probe(xfr, env, xfr->task_probe->timeout*2)) {
6460                         lock_basic_unlock(&xfr->lock);
6461                         return;
6462                 }
6463         }
6464         /* delete commpoint so a new one is created, with a fresh port nr */
6465         comm_point_delete(xfr->task_probe->cp);
6466         xfr->task_probe->cp = NULL;
6467
6468         /* too many timeouts (or fail to send), move to next or end */
6469         xfr_probe_nextmaster(xfr);
6470         xfr_probe_send_or_end(xfr, env);
6471 }
6472
6473 /** callback for task_probe udp packets */
6474 int
6475 auth_xfer_probe_udp_callback(struct comm_point* c, void* arg, int err,
6476         struct comm_reply* repinfo)
6477 {
6478         struct auth_xfer* xfr = (struct auth_xfer*)arg;
6479         struct module_env* env;
6480         log_assert(xfr->task_probe);
6481         lock_basic_lock(&xfr->lock);
6482         env = xfr->task_probe->env;
6483         if(!env || env->outnet->want_to_quit) {
6484                 lock_basic_unlock(&xfr->lock);
6485                 return 0; /* stop on quit */
6486         }
6487
6488         /* the comm_point_udp_callback is in a for loop for NUM_UDP_PER_SELECT
6489          * and we set rep.c=NULL to stop if from looking inside the commpoint*/
6490         repinfo->c = NULL;
6491         /* stop the timer */
6492         comm_timer_disable(xfr->task_probe->timer);
6493
6494         /* see if we got a packet and what that means */
6495         if(err == NETEVENT_NOERROR) {
6496                 uint32_t serial = 0;
6497                 if(check_packet_ok(c->buffer, LDNS_RR_TYPE_SOA, xfr,
6498                         &serial)) {
6499                         /* successful lookup */
6500                         if(verbosity >= VERB_ALGO) {
6501                                 char buf[256];
6502                                 dname_str(xfr->name, buf);
6503                                 verbose(VERB_ALGO, "auth zone %s: soa probe "
6504                                         "serial is %u", buf, (unsigned)serial);
6505                         }
6506                         /* see if this serial indicates that the zone has
6507                          * to be updated */
6508                         if(xfr_serial_means_update(xfr, serial)) {
6509                                 /* if updated, start the transfer task, if needed */
6510                                 verbose(VERB_ALGO, "auth_zone updated, start transfer");
6511                                 if(xfr->task_transfer->worker == NULL) {
6512                                         struct auth_master* master =
6513                                                 xfr_probe_current_master(xfr);
6514                                         /* if we have download URLs use them
6515                                          * in preference to this master we
6516                                          * just probed the SOA from */
6517                                         if(xfr->task_transfer->masters &&
6518                                                 xfr->task_transfer->masters->http)
6519                                                 master = NULL;
6520                                         xfr_probe_disown(xfr);
6521                                         xfr_start_transfer(xfr, env, master);
6522                                         return 0;
6523
6524                                 }
6525                                 /* other tasks are running, we don't do this anymore */
6526                                 xfr_probe_disown(xfr);
6527                                 lock_basic_unlock(&xfr->lock);
6528                                 /* return, we don't sent a reply to this udp packet,
6529                                  * and we setup the tasks to do next */
6530                                 return 0;
6531                         } else {
6532                                 verbose(VERB_ALGO, "auth_zone master reports unchanged soa serial");
6533                                 /* we if cannot find updates amongst the
6534                                  * masters, this means we then have a new lease
6535                                  * on the zone */
6536                                 xfr->task_probe->have_new_lease = 1;
6537                         }
6538                 } else {
6539                         if(verbosity >= VERB_ALGO) {
6540                                 char buf[256];
6541                                 dname_str(xfr->name, buf);
6542                                 verbose(VERB_ALGO, "auth zone %s: bad reply to soa probe", buf);
6543                         }
6544                 }
6545         } else {
6546                 if(verbosity >= VERB_ALGO) {
6547                         char buf[256];
6548                         dname_str(xfr->name, buf);
6549                         verbose(VERB_ALGO, "auth zone %s: soa probe failed", buf);
6550                 }
6551         }
6552         
6553         /* failed lookup or not an update */
6554         /* delete commpoint so a new one is created, with a fresh port nr */
6555         comm_point_delete(xfr->task_probe->cp);
6556         xfr->task_probe->cp = NULL;
6557
6558         /* if the result was not a successful probe, we need
6559          * to send the next one */
6560         xfr_probe_nextmaster(xfr);
6561         xfr_probe_send_or_end(xfr, env);
6562         return 0;
6563 }
6564
6565 /** lookup a host name for its addresses, if needed */
6566 static int
6567 xfr_probe_lookup_host(struct auth_xfer* xfr, struct module_env* env)
6568 {
6569         struct sockaddr_storage addr;
6570         socklen_t addrlen = 0;
6571         struct auth_master* master = xfr->task_probe->lookup_target;
6572         struct query_info qinfo;
6573         uint16_t qflags = BIT_RD;
6574         uint8_t dname[LDNS_MAX_DOMAINLEN+1];
6575         struct edns_data edns;
6576         sldns_buffer* buf = env->scratch_buffer;
6577         if(!master) return 0;
6578         if(extstrtoaddr(master->host, &addr, &addrlen, UNBOUND_DNS_PORT)) {
6579                 /* not needed, host is in IP addr format */
6580                 return 0;
6581         }
6582         if(master->allow_notify && !master->http &&
6583                 strchr(master->host, '/') != NULL &&
6584                 strchr(master->host, '/') == strrchr(master->host, '/')) {
6585                 return 0; /* is IP/prefix format, not something to look up */
6586         }
6587
6588         /* use mesh_new_callback to probe for non-addr hosts,
6589          * and then wait for them to be looked up (in cache, or query) */
6590         qinfo.qname_len = sizeof(dname);
6591         if(sldns_str2wire_dname_buf(master->host, dname, &qinfo.qname_len)
6592                 != 0) {
6593                 log_err("cannot parse host name of master %s", master->host);
6594                 return 0;
6595         }
6596         qinfo.qname = dname;
6597         qinfo.qclass = xfr->dclass;
6598         qinfo.qtype = LDNS_RR_TYPE_A;
6599         if(xfr->task_probe->lookup_aaaa)
6600                 qinfo.qtype = LDNS_RR_TYPE_AAAA;
6601         qinfo.local_alias = NULL;
6602         if(verbosity >= VERB_ALGO) {
6603                 char buf1[512];
6604                 char buf2[LDNS_MAX_DOMAINLEN+1];
6605                 dname_str(xfr->name, buf2);
6606                 snprintf(buf1, sizeof(buf1), "auth zone %s: master lookup"
6607                         " for task_probe", buf2);
6608                 log_query_info(VERB_ALGO, buf1, &qinfo);
6609         }
6610         edns.edns_present = 1;
6611         edns.ext_rcode = 0;
6612         edns.edns_version = 0;
6613         edns.bits = EDNS_DO;
6614         edns.opt_list_in = NULL;
6615         edns.opt_list_out = NULL;
6616         edns.opt_list_inplace_cb_out = NULL;
6617         edns.padding_block_size = 0;
6618         edns.cookie_present = 0;
6619         edns.cookie_valid = 0;
6620         if(sldns_buffer_capacity(buf) < 65535)
6621                 edns.udp_size = (uint16_t)sldns_buffer_capacity(buf);
6622         else    edns.udp_size = 65535;
6623
6624         /* unlock xfr during mesh_new_callback() because the callback can be
6625          * called straight away */
6626         lock_basic_unlock(&xfr->lock);
6627         if(!mesh_new_callback(env->mesh, &qinfo, qflags, &edns, buf, 0,
6628                 &auth_xfer_probe_lookup_callback, xfr, 0)) {
6629                 lock_basic_lock(&xfr->lock);
6630                 log_err("out of memory lookup up master %s", master->host);
6631                 return 0;
6632         }
6633         lock_basic_lock(&xfr->lock);
6634         return 1;
6635 }
6636
6637 /** move to sending the probe packets, next if fails. task_probe */
6638 static void
6639 xfr_probe_send_or_end(struct auth_xfer* xfr, struct module_env* env)
6640 {
6641         /* are we doing hostname lookups? */
6642         while(xfr->task_probe->lookup_target) {
6643                 if(xfr_probe_lookup_host(xfr, env)) {
6644                         /* wait for lookup to finish,
6645                          * note that the hostname may be in unbound's cache
6646                          * and we may then get an instant cache response,
6647                          * and that calls the callback just like a full
6648                          * lookup and lookup failures also call callback */
6649                         if(verbosity >= VERB_ALGO) {
6650                                 char zname[255+1];
6651                                 dname_str(xfr->name, zname);
6652                                 verbose(VERB_ALGO, "auth zone %s probe next target lookup", zname);
6653                         }
6654                         lock_basic_unlock(&xfr->lock);
6655                         return;
6656                 }
6657                 xfr_probe_move_to_next_lookup(xfr, env);
6658         }
6659         /* probe of list has ended.  Create or refresh the list of of
6660          * allow_notify addrs */
6661         probe_copy_masters_for_allow_notify(xfr);
6662         if(verbosity >= VERB_ALGO) {
6663                 char zname[255+1];
6664                 dname_str(xfr->name, zname);
6665                 verbose(VERB_ALGO, "auth zone %s probe: notify addrs updated", zname);
6666         }
6667         if(xfr->task_probe->only_lookup) {
6668                 /* only wanted lookups for copy, stop probe and start wait */
6669                 xfr->task_probe->only_lookup = 0;
6670                 if(verbosity >= VERB_ALGO) {
6671                         char zname[255+1];
6672                         dname_str(xfr->name, zname);
6673                         verbose(VERB_ALGO, "auth zone %s probe: finished only_lookup", zname);
6674                 }
6675                 xfr_probe_disown(xfr);
6676                 if(xfr->task_nextprobe->worker == NULL)
6677                         xfr_set_timeout(xfr, env, 0, 0);
6678                 lock_basic_unlock(&xfr->lock);
6679                 return;
6680         }
6681
6682         /* send probe packets */
6683         while(!xfr_probe_end_of_list(xfr)) {
6684                 if(xfr_probe_send_probe(xfr, env, AUTH_PROBE_TIMEOUT)) {
6685                         /* successfully sent probe, wait for callback */
6686                         lock_basic_unlock(&xfr->lock);
6687                         return;
6688                 }
6689                 /* failed to send probe, next master */
6690                 xfr_probe_nextmaster(xfr);
6691         }
6692
6693         /* done with probe sequence, wait */
6694         if(xfr->task_probe->have_new_lease) {
6695                 /* if zone not updated, start the wait timer again */
6696                 if(verbosity >= VERB_ALGO) {
6697                         char zname[255+1];
6698                         dname_str(xfr->name, zname);
6699                         verbose(VERB_ALGO, "auth_zone %s unchanged, new lease, wait", zname);
6700                 }
6701                 xfr_probe_disown(xfr);
6702                 if(xfr->have_zone)
6703                         xfr->lease_time = *env->now;
6704                 if(xfr->task_nextprobe->worker == NULL)
6705                         xfr_set_timeout(xfr, env, 0, 0);
6706         } else {
6707                 if(verbosity >= VERB_ALGO) {
6708                         char zname[255+1];
6709                         dname_str(xfr->name, zname);
6710                         verbose(VERB_ALGO, "auth zone %s soa probe failed, wait to retry", zname);
6711                 }
6712                 /* we failed to send this as well, move to the wait task,
6713                  * use the shorter retry timeout */
6714                 xfr_probe_disown(xfr);
6715                 /* pick up the nextprobe task and wait */
6716                 if(xfr->task_nextprobe->worker == NULL)
6717                         xfr_set_timeout(xfr, env, 1, 0);
6718         }
6719
6720         lock_basic_unlock(&xfr->lock);
6721 }
6722
6723 /** callback for task_probe lookup of host name, of A or AAAA */
6724 void auth_xfer_probe_lookup_callback(void* arg, int rcode, sldns_buffer* buf,
6725         enum sec_status ATTR_UNUSED(sec), char* ATTR_UNUSED(why_bogus),
6726         int ATTR_UNUSED(was_ratelimited))
6727 {
6728         struct auth_xfer* xfr = (struct auth_xfer*)arg;
6729         struct module_env* env;
6730         log_assert(xfr->task_probe);
6731         lock_basic_lock(&xfr->lock);
6732         env = xfr->task_probe->env;
6733         if(!env || env->outnet->want_to_quit) {
6734                 lock_basic_unlock(&xfr->lock);
6735                 return; /* stop on quit */
6736         }
6737
6738         /* process result */
6739         if(rcode == LDNS_RCODE_NOERROR) {
6740                 uint16_t wanted_qtype = LDNS_RR_TYPE_A;
6741                 struct regional* temp = env->scratch;
6742                 struct query_info rq;
6743                 struct reply_info* rep;
6744                 if(xfr->task_probe->lookup_aaaa)
6745                         wanted_qtype = LDNS_RR_TYPE_AAAA;
6746                 memset(&rq, 0, sizeof(rq));
6747                 rep = parse_reply_in_temp_region(buf, temp, &rq);
6748                 if(rep && rq.qtype == wanted_qtype &&
6749                         FLAGS_GET_RCODE(rep->flags) == LDNS_RCODE_NOERROR) {
6750                         /* parsed successfully */
6751                         struct ub_packed_rrset_key* answer =
6752                                 reply_find_answer_rrset(&rq, rep);
6753                         if(answer) {
6754                                 xfr_master_add_addrs(xfr->task_probe->
6755                                         lookup_target, answer, wanted_qtype);
6756                         } else {
6757                                 if(verbosity >= VERB_ALGO) {
6758                                         char zname[255+1];
6759                                         dname_str(xfr->name, zname);
6760                                         verbose(VERB_ALGO, "auth zone %s host %s type %s probe lookup has nodata", zname, xfr->task_probe->lookup_target->host, (xfr->task_probe->lookup_aaaa?"AAAA":"A"));
6761                                 }
6762                         }
6763                 } else {
6764                         if(verbosity >= VERB_ALGO) {
6765                                 char zname[255+1];
6766                                 dname_str(xfr->name, zname);
6767                                 verbose(VERB_ALGO, "auth zone %s host %s type %s probe lookup has no address", zname, xfr->task_probe->lookup_target->host, (xfr->task_probe->lookup_aaaa?"AAAA":"A"));
6768                         }
6769                 }
6770                 regional_free_all(temp);
6771         } else {
6772                 if(verbosity >= VERB_ALGO) {
6773                         char zname[255+1];
6774                         dname_str(xfr->name, zname);
6775                         verbose(VERB_ALGO, "auth zone %s host %s type %s probe lookup failed", zname, xfr->task_probe->lookup_target->host, (xfr->task_probe->lookup_aaaa?"AAAA":"A"));
6776                 }
6777         }
6778         if(xfr->task_probe->lookup_target->list &&
6779                 xfr->task_probe->lookup_target == xfr_probe_current_master(xfr))
6780                 xfr->task_probe->scan_addr = xfr->task_probe->lookup_target->list;
6781
6782         /* move to lookup AAAA after A lookup, move to next hostname lookup,
6783          * or move to send the probes, or, if nothing to do, end task_probe */
6784         xfr_probe_move_to_next_lookup(xfr, env);
6785         xfr_probe_send_or_end(xfr, env);
6786 }
6787
6788 /** disown task_nextprobe.  caller must hold xfr.lock */
6789 static void
6790 xfr_nextprobe_disown(struct auth_xfer* xfr)
6791 {
6792         /* delete the timer, because the next worker to pick this up may
6793          * not have the same event base */
6794         comm_timer_delete(xfr->task_nextprobe->timer);
6795         xfr->task_nextprobe->timer = NULL;
6796         xfr->task_nextprobe->next_probe = 0;
6797         /* we don't own this item anymore */
6798         xfr->task_nextprobe->worker = NULL;
6799         xfr->task_nextprobe->env = NULL;
6800 }
6801
6802 /** xfer nextprobe timeout callback, this is part of task_nextprobe */
6803 void
6804 auth_xfer_timer(void* arg)
6805 {
6806         struct auth_xfer* xfr = (struct auth_xfer*)arg;
6807         struct module_env* env;
6808         log_assert(xfr->task_nextprobe);
6809         lock_basic_lock(&xfr->lock);
6810         env = xfr->task_nextprobe->env;
6811         if(!env || env->outnet->want_to_quit) {
6812                 lock_basic_unlock(&xfr->lock);
6813                 return; /* stop on quit */
6814         }
6815
6816         /* see if zone has expired, and if so, also set auth_zone expired */
6817         if(xfr->have_zone && !xfr->zone_expired &&
6818            *env->now >= xfr->lease_time + xfr->expiry) {
6819                 lock_basic_unlock(&xfr->lock);
6820                 auth_xfer_set_expired(xfr, env, 1);
6821                 lock_basic_lock(&xfr->lock);
6822         }
6823
6824         xfr_nextprobe_disown(xfr);
6825
6826         if(!xfr_start_probe(xfr, env, NULL)) {
6827                 /* not started because already in progress */
6828                 lock_basic_unlock(&xfr->lock);
6829         }
6830 }
6831
6832 /** return true if there are probe (SOA UDP query) targets in the master list*/
6833 static int
6834 have_probe_targets(struct auth_master* list)
6835 {
6836         struct auth_master* p;
6837         for(p=list; p; p = p->next) {
6838                 if(!p->allow_notify && p->host)
6839                         return 1;
6840         }
6841         return 0;
6842 }
6843
6844 /** start task_probe if possible, if no masters for probe start task_transfer
6845  * returns true if task has been started, and false if the task is already
6846  * in progress. */
6847 static int
6848 xfr_start_probe(struct auth_xfer* xfr, struct module_env* env,
6849         struct auth_master* spec)
6850 {
6851         /* see if we need to start a probe (or maybe it is already in
6852          * progress (due to notify)) */
6853         if(xfr->task_probe->worker == NULL) {
6854                 if(!have_probe_targets(xfr->task_probe->masters) &&
6855                         !(xfr->task_probe->only_lookup &&
6856                         xfr->task_probe->masters != NULL)) {
6857                         /* useless to pick up task_probe, no masters to
6858                          * probe. Instead attempt to pick up task transfer */
6859                         if(xfr->task_transfer->worker == NULL) {
6860                                 xfr_start_transfer(xfr, env, spec);
6861                                 return 1;
6862                         }
6863                         /* task transfer already in progress */
6864                         return 0;
6865                 }
6866
6867                 /* pick up the probe task ourselves */
6868                 xfr->task_probe->worker = env->worker;
6869                 xfr->task_probe->env = env;
6870                 xfr->task_probe->cp = NULL;
6871
6872                 /* start the task */
6873                 /* have not seen a new lease yet, this scan */
6874                 xfr->task_probe->have_new_lease = 0;
6875                 /* if this was a timeout, no specific first master to scan */
6876                 /* otherwise, spec is nonNULL the notified master, scan
6877                  * first and also transfer first from it */
6878                 xfr_probe_start_list(xfr, spec);
6879                 /* setup to start the lookup of hostnames of masters afresh */
6880                 xfr_probe_start_lookups(xfr);
6881                 /* send the probe packet or next send, or end task */
6882                 xfr_probe_send_or_end(xfr, env);
6883                 return 1;
6884         }
6885         return 0;
6886 }
6887
6888 /** for task_nextprobe.
6889  * determine next timeout for auth_xfer. Also (re)sets timer.
6890  * @param xfr: task structure
6891  * @param env: module environment, with worker and time.
6892  * @param failure: set true if timer should be set for failure retry.
6893  * @param lookup_only: only perform lookups when timer done, 0 sec timeout
6894  */
6895 static void
6896 xfr_set_timeout(struct auth_xfer* xfr, struct module_env* env,
6897         int failure, int lookup_only)
6898 {
6899         struct timeval tv;
6900         log_assert(xfr->task_nextprobe != NULL);
6901         log_assert(xfr->task_nextprobe->worker == NULL ||
6902                 xfr->task_nextprobe->worker == env->worker);
6903         /* normally, nextprobe = startoflease + refresh,
6904          * but if expiry is sooner, use that one.
6905          * after a failure, use the retry timer instead. */
6906         xfr->task_nextprobe->next_probe = *env->now;
6907         if(xfr->lease_time && !failure)
6908                 xfr->task_nextprobe->next_probe = xfr->lease_time;
6909         
6910         if(!failure) {
6911                 xfr->task_nextprobe->backoff = 0;
6912         } else {
6913                 if(xfr->task_nextprobe->backoff == 0)
6914                                 xfr->task_nextprobe->backoff = 3;
6915                 else    xfr->task_nextprobe->backoff *= 2;
6916                 if(xfr->task_nextprobe->backoff > AUTH_TRANSFER_MAX_BACKOFF)
6917                         xfr->task_nextprobe->backoff =
6918                                 AUTH_TRANSFER_MAX_BACKOFF;
6919         }
6920
6921         if(xfr->have_zone) {
6922                 time_t wait = xfr->refresh;
6923                 if(failure) wait = xfr->retry;
6924                 if(xfr->expiry < wait)
6925                         xfr->task_nextprobe->next_probe += xfr->expiry;
6926                 else    xfr->task_nextprobe->next_probe += wait;
6927                 if(failure)
6928                         xfr->task_nextprobe->next_probe +=
6929                                 xfr->task_nextprobe->backoff;
6930                 /* put the timer exactly on expiry, if possible */
6931                 if(xfr->lease_time && xfr->lease_time+xfr->expiry <
6932                         xfr->task_nextprobe->next_probe &&
6933                         xfr->lease_time+xfr->expiry > *env->now)
6934                         xfr->task_nextprobe->next_probe =
6935                                 xfr->lease_time+xfr->expiry;
6936         } else {
6937                 xfr->task_nextprobe->next_probe +=
6938                         xfr->task_nextprobe->backoff;
6939         }
6940
6941         if(!xfr->task_nextprobe->timer) {
6942                 xfr->task_nextprobe->timer = comm_timer_create(
6943                         env->worker_base, auth_xfer_timer, xfr);
6944                 if(!xfr->task_nextprobe->timer) {
6945                         /* failed to malloc memory. likely zone transfer
6946                          * also fails for that. skip the timeout */
6947                         char zname[255+1];
6948                         dname_str(xfr->name, zname);
6949                         log_err("cannot allocate timer, no refresh for %s",
6950                                 zname);
6951                         return;
6952                 }
6953         }
6954         xfr->task_nextprobe->worker = env->worker;
6955         xfr->task_nextprobe->env = env;
6956         if(*(xfr->task_nextprobe->env->now) <= xfr->task_nextprobe->next_probe)
6957                 tv.tv_sec = xfr->task_nextprobe->next_probe - 
6958                         *(xfr->task_nextprobe->env->now);
6959         else    tv.tv_sec = 0;
6960         if(tv.tv_sec != 0 && lookup_only && xfr->task_probe->masters) {
6961                 /* don't lookup_only, if lookup timeout is 0 anyway,
6962                  * or if we don't have masters to lookup */
6963                 tv.tv_sec = 0;
6964                 if(xfr->task_probe->worker == NULL)
6965                         xfr->task_probe->only_lookup = 1;
6966         }
6967         if(verbosity >= VERB_ALGO) {
6968                 char zname[255+1];
6969                 dname_str(xfr->name, zname);
6970                 verbose(VERB_ALGO, "auth zone %s timeout in %d seconds",
6971                         zname, (int)tv.tv_sec);
6972         }
6973         tv.tv_usec = 0;
6974         comm_timer_set(xfr->task_nextprobe->timer, &tv);
6975 }
6976
6977 /** initial pick up of worker timeouts, ties events to worker event loop */
6978 void
6979 auth_xfer_pickup_initial(struct auth_zones* az, struct module_env* env)
6980 {
6981         struct auth_xfer* x;
6982         lock_rw_wrlock(&az->lock);
6983         RBTREE_FOR(x, struct auth_xfer*, &az->xtree) {
6984                 lock_basic_lock(&x->lock);
6985                 /* set lease_time, because we now have timestamp in env,
6986                  * (not earlier during startup and apply_cfg), and this
6987                  * notes the start time when the data was acquired */
6988                 if(x->have_zone)
6989                         x->lease_time = *env->now;
6990                 if(x->task_nextprobe && x->task_nextprobe->worker == NULL) {
6991                         xfr_set_timeout(x, env, 0, 1);
6992                 }
6993                 lock_basic_unlock(&x->lock);
6994         }
6995         lock_rw_unlock(&az->lock);
6996 }
6997
6998 void auth_zones_cleanup(struct auth_zones* az)
6999 {
7000         struct auth_xfer* x;
7001         lock_rw_wrlock(&az->lock);
7002         RBTREE_FOR(x, struct auth_xfer*, &az->xtree) {
7003                 lock_basic_lock(&x->lock);
7004                 if(x->task_nextprobe && x->task_nextprobe->worker != NULL) {
7005                         xfr_nextprobe_disown(x);
7006                 }
7007                 if(x->task_probe && x->task_probe->worker != NULL) {
7008                         xfr_probe_disown(x);
7009                 }
7010                 if(x->task_transfer && x->task_transfer->worker != NULL) {
7011                         auth_chunks_delete(x->task_transfer);
7012                         xfr_transfer_disown(x);
7013                 }
7014                 lock_basic_unlock(&x->lock);
7015         }
7016         lock_rw_unlock(&az->lock);
7017 }
7018
7019 /**
7020  * malloc the xfer and tasks
7021  * @param z: auth_zone with name of zone.
7022  */
7023 static struct auth_xfer*
7024 auth_xfer_new(struct auth_zone* z)
7025 {
7026         struct auth_xfer* xfr;
7027         xfr = (struct auth_xfer*)calloc(1, sizeof(*xfr));
7028         if(!xfr) return NULL;
7029         xfr->name = memdup(z->name, z->namelen);
7030         if(!xfr->name) {
7031                 free(xfr);
7032                 return NULL;
7033         }
7034         xfr->node.key = xfr;
7035         xfr->namelen = z->namelen;
7036         xfr->namelabs = z->namelabs;
7037         xfr->dclass = z->dclass;
7038
7039         xfr->task_nextprobe = (struct auth_nextprobe*)calloc(1,
7040                 sizeof(struct auth_nextprobe));
7041         if(!xfr->task_nextprobe) {
7042                 free(xfr->name);
7043                 free(xfr);
7044                 return NULL;
7045         }
7046         xfr->task_probe = (struct auth_probe*)calloc(1,
7047                 sizeof(struct auth_probe));
7048         if(!xfr->task_probe) {
7049                 free(xfr->task_nextprobe);
7050                 free(xfr->name);
7051                 free(xfr);
7052                 return NULL;
7053         }
7054         xfr->task_transfer = (struct auth_transfer*)calloc(1,
7055                 sizeof(struct auth_transfer));
7056         if(!xfr->task_transfer) {
7057                 free(xfr->task_probe);
7058                 free(xfr->task_nextprobe);
7059                 free(xfr->name);
7060                 free(xfr);
7061                 return NULL;
7062         }
7063
7064         lock_basic_init(&xfr->lock);
7065         lock_protect(&xfr->lock, &xfr->name, sizeof(xfr->name));
7066         lock_protect(&xfr->lock, &xfr->namelen, sizeof(xfr->namelen));
7067         lock_protect(&xfr->lock, xfr->name, xfr->namelen);
7068         lock_protect(&xfr->lock, &xfr->namelabs, sizeof(xfr->namelabs));
7069         lock_protect(&xfr->lock, &xfr->dclass, sizeof(xfr->dclass));
7070         lock_protect(&xfr->lock, &xfr->notify_received, sizeof(xfr->notify_received));
7071         lock_protect(&xfr->lock, &xfr->notify_serial, sizeof(xfr->notify_serial));
7072         lock_protect(&xfr->lock, &xfr->zone_expired, sizeof(xfr->zone_expired));
7073         lock_protect(&xfr->lock, &xfr->have_zone, sizeof(xfr->have_zone));
7074         lock_protect(&xfr->lock, &xfr->serial, sizeof(xfr->serial));
7075         lock_protect(&xfr->lock, &xfr->retry, sizeof(xfr->retry));
7076         lock_protect(&xfr->lock, &xfr->refresh, sizeof(xfr->refresh));
7077         lock_protect(&xfr->lock, &xfr->expiry, sizeof(xfr->expiry));
7078         lock_protect(&xfr->lock, &xfr->lease_time, sizeof(xfr->lease_time));
7079         lock_protect(&xfr->lock, &xfr->task_nextprobe->worker,
7080                 sizeof(xfr->task_nextprobe->worker));
7081         lock_protect(&xfr->lock, &xfr->task_probe->worker,
7082                 sizeof(xfr->task_probe->worker));
7083         lock_protect(&xfr->lock, &xfr->task_transfer->worker,
7084                 sizeof(xfr->task_transfer->worker));
7085         lock_basic_lock(&xfr->lock);
7086         return xfr;
7087 }
7088
7089 /** Create auth_xfer structure.
7090  * This populates the have_zone, soa values, and so on times.
7091  * and sets the timeout, if a zone transfer is needed a short timeout is set.
7092  * For that the auth_zone itself must exist (and read in zonefile)
7093  * returns false on alloc failure. */
7094 struct auth_xfer*
7095 auth_xfer_create(struct auth_zones* az, struct auth_zone* z)
7096 {
7097         struct auth_xfer* xfr;
7098
7099         /* malloc it */
7100         xfr = auth_xfer_new(z);
7101         if(!xfr) {
7102                 log_err("malloc failure");
7103                 return NULL;
7104         }
7105         /* insert in tree */
7106         (void)rbtree_insert(&az->xtree, &xfr->node);
7107         return xfr;
7108 }
7109
7110 /** create new auth_master structure */
7111 static struct auth_master*
7112 auth_master_new(struct auth_master*** list)
7113 {
7114         struct auth_master *m;
7115         m = (struct auth_master*)calloc(1, sizeof(*m));
7116         if(!m) {
7117                 log_err("malloc failure");
7118                 return NULL;
7119         }
7120         /* set first pointer to m, or next pointer of previous element to m */
7121         (**list) = m;
7122         /* store m's next pointer as future point to store at */
7123         (*list) = &(m->next);
7124         return m;
7125 }
7126
7127 /** dup_prefix : create string from initial part of other string, malloced */
7128 static char*
7129 dup_prefix(char* str, size_t num)
7130 {
7131         char* result;
7132         size_t len = strlen(str);
7133         if(len < num) num = len; /* not more than strlen */
7134         result = (char*)malloc(num+1);
7135         if(!result) {
7136                 log_err("malloc failure");
7137                 return result;
7138         }
7139         memmove(result, str, num);
7140         result[num] = 0;
7141         return result;
7142 }
7143
7144 /** dup string and print error on error */
7145 static char*
7146 dup_all(char* str)
7147 {
7148         char* result = strdup(str);
7149         if(!result) {
7150                 log_err("malloc failure");
7151                 return NULL;
7152         }
7153         return result;
7154 }
7155
7156 /** find first of two characters */
7157 static char*
7158 str_find_first_of_chars(char* s, char a, char b)
7159 {
7160         char* ra = strchr(s, a);
7161         char* rb = strchr(s, b);
7162         if(!ra) return rb;
7163         if(!rb) return ra;
7164         if(ra < rb) return ra;
7165         return rb;
7166 }
7167
7168 /** parse URL into host and file parts, false on malloc or parse error */
7169 static int
7170 parse_url(char* url, char** host, char** file, int* port, int* ssl)
7171 {
7172         char* p = url;
7173         /* parse http://www.example.com/file.htm
7174          * or http://127.0.0.1   (index.html)
7175          * or https://[::1@1234]/a/b/c/d */
7176         *ssl = 1;
7177         *port = AUTH_HTTPS_PORT;
7178
7179         /* parse http:// or https:// */
7180         if(strncmp(p, "http://", 7) == 0) {
7181                 p += 7;
7182                 *ssl = 0;
7183                 *port = AUTH_HTTP_PORT;
7184         } else if(strncmp(p, "https://", 8) == 0) {
7185                 p += 8;
7186         } else if(strstr(p, "://") && strchr(p, '/') > strstr(p, "://") &&
7187                 strchr(p, ':') >= strstr(p, "://")) {
7188                 char* uri = dup_prefix(p, (size_t)(strstr(p, "://")-p));
7189                 log_err("protocol %s:// not supported (for url %s)",
7190                         uri?uri:"", p);
7191                 free(uri);
7192                 return 0;
7193         }
7194
7195         /* parse hostname part */
7196         if(p[0] == '[') {
7197                 char* end = strchr(p, ']');
7198                 p++; /* skip over [ */
7199                 if(end) {
7200                         *host = dup_prefix(p, (size_t)(end-p));
7201                         if(!*host) return 0;
7202                         p = end+1; /* skip over ] */
7203                 } else {
7204                         *host = dup_all(p);
7205                         if(!*host) return 0;
7206                         p = end;
7207                 }
7208         } else {
7209                 char* end = str_find_first_of_chars(p, ':', '/');
7210                 if(end) {
7211                         *host = dup_prefix(p, (size_t)(end-p));
7212                         if(!*host) return 0;
7213                 } else {
7214                         *host = dup_all(p);
7215                         if(!*host) return 0;
7216                 }
7217                 p = end; /* at next : or / or NULL */
7218         }
7219
7220         /* parse port number */
7221         if(p && p[0] == ':') {
7222                 char* end = NULL;
7223                 *port = strtol(p+1, &end, 10);
7224                 p = end;
7225         }
7226
7227         /* parse filename part */
7228         while(p && *p == '/')
7229                 p++;
7230         if(!p || p[0] == 0)
7231                 *file = strdup("/");
7232         else    *file = strdup(p);
7233         if(!*file) {
7234                 log_err("malloc failure");
7235                 return 0;
7236         }
7237         return 1;
7238 }
7239
7240 int
7241 xfer_set_masters(struct auth_master** list, struct config_auth* c,
7242         int with_http)
7243 {
7244         struct auth_master* m;
7245         struct config_strlist* p;
7246         /* list points to the first, or next pointer for the new element */
7247         while(*list) {
7248                 list = &( (*list)->next );
7249         }
7250         if(with_http)
7251           for(p = c->urls; p; p = p->next) {
7252                 m = auth_master_new(&list);
7253                 if(!m) return 0;
7254                 m->http = 1;
7255                 if(!parse_url(p->str, &m->host, &m->file, &m->port, &m->ssl))
7256                         return 0;
7257         }
7258         for(p = c->masters; p; p = p->next) {
7259                 m = auth_master_new(&list);
7260                 if(!m) return 0;
7261                 m->ixfr = 1; /* this flag is not configurable */
7262                 m->host = strdup(p->str);
7263                 if(!m->host) {
7264                         log_err("malloc failure");
7265                         return 0;
7266                 }
7267         }
7268         for(p = c->allow_notify; p; p = p->next) {
7269                 m = auth_master_new(&list);
7270                 if(!m) return 0;
7271                 m->allow_notify = 1;
7272                 m->host = strdup(p->str);
7273                 if(!m->host) {
7274                         log_err("malloc failure");
7275                         return 0;
7276                 }
7277         }
7278         return 1;
7279 }
7280
7281 #define SERIAL_BITS     32
7282 int
7283 compare_serial(uint32_t a, uint32_t b)
7284 {
7285         const uint32_t cutoff = ((uint32_t) 1 << (SERIAL_BITS - 1));
7286
7287         if (a == b) {
7288                 return 0;
7289         } else if ((a < b && b - a < cutoff) || (a > b && a - b > cutoff)) {
7290                 return -1;
7291         } else {
7292                 return 1;
7293         }
7294 }
7295
7296 int zonemd_hashalgo_supported(int hashalgo)
7297 {
7298         if(hashalgo == ZONEMD_ALGO_SHA384) return 1;
7299         if(hashalgo == ZONEMD_ALGO_SHA512) return 1;
7300         return 0;
7301 }
7302
7303 int zonemd_scheme_supported(int scheme)
7304 {
7305         if(scheme == ZONEMD_SCHEME_SIMPLE) return 1;
7306         return 0;
7307 }
7308
7309 /** initialize hash for hashing with zonemd hash algo */
7310 static struct secalgo_hash* zonemd_digest_init(int hashalgo, char** reason)
7311 {
7312         struct secalgo_hash *h;
7313         if(hashalgo == ZONEMD_ALGO_SHA384) {
7314                 /* sha384 */
7315                 h = secalgo_hash_create_sha384();
7316                 if(!h)
7317                         *reason = "digest sha384 could not be created";
7318                 return h;
7319         } else if(hashalgo == ZONEMD_ALGO_SHA512) {
7320                 /* sha512 */
7321                 h = secalgo_hash_create_sha512();
7322                 if(!h)
7323                         *reason = "digest sha512 could not be created";
7324                 return h;
7325         }
7326         /* unknown hash algo */
7327         *reason = "unsupported algorithm";
7328         return NULL;
7329 }
7330
7331 /** update the hash for zonemd */
7332 static int zonemd_digest_update(int hashalgo, struct secalgo_hash* h,
7333         uint8_t* data, size_t len, char** reason)
7334 {
7335         if(hashalgo == ZONEMD_ALGO_SHA384) {
7336                 if(!secalgo_hash_update(h, data, len)) {
7337                         *reason = "digest sha384 failed";
7338                         return 0;
7339                 }
7340                 return 1;
7341         } else if(hashalgo == ZONEMD_ALGO_SHA512) {
7342                 if(!secalgo_hash_update(h, data, len)) {
7343                         *reason = "digest sha512 failed";
7344                         return 0;
7345                 }
7346                 return 1;
7347         }
7348         /* unknown hash algo */
7349         *reason = "unsupported algorithm";
7350         return 0;
7351 }
7352
7353 /** finish the hash for zonemd */
7354 static int zonemd_digest_finish(int hashalgo, struct secalgo_hash* h,
7355         uint8_t* result, size_t hashlen, size_t* resultlen, char** reason)
7356 {
7357         if(hashalgo == ZONEMD_ALGO_SHA384) {
7358                 if(hashlen < 384/8) {
7359                         *reason = "digest buffer too small for sha384";
7360                         return 0;
7361                 }
7362                 if(!secalgo_hash_final(h, result, hashlen, resultlen)) {
7363                         *reason = "digest sha384 finish failed";
7364                         return 0;
7365                 }
7366                 return 1;
7367         } else if(hashalgo == ZONEMD_ALGO_SHA512) {
7368                 if(hashlen < 512/8) {
7369                         *reason = "digest buffer too small for sha512";
7370                         return 0;
7371                 }
7372                 if(!secalgo_hash_final(h, result, hashlen, resultlen)) {
7373                         *reason = "digest sha512 finish failed";
7374                         return 0;
7375                 }
7376                 return 1;
7377         }
7378         /* unknown algo */
7379         *reason = "unsupported algorithm";
7380         return 0;
7381 }
7382
7383 /** add rrsets from node to the list */
7384 static size_t authdata_rrsets_to_list(struct auth_rrset** array,
7385         size_t arraysize, struct auth_rrset* first)
7386 {
7387         struct auth_rrset* rrset = first;
7388         size_t num = 0;
7389         while(rrset) {
7390                 if(num >= arraysize)
7391                         return num;
7392                 array[num] = rrset;
7393                 num++;
7394                 rrset = rrset->next;
7395         }
7396         return num;
7397 }
7398
7399 /** compare rr list entries */
7400 static int rrlist_compare(const void* arg1, const void* arg2)
7401 {
7402         struct auth_rrset* r1 = *(struct auth_rrset**)arg1;
7403         struct auth_rrset* r2 = *(struct auth_rrset**)arg2;
7404         uint16_t t1, t2;
7405         if(r1 == NULL) t1 = LDNS_RR_TYPE_RRSIG;
7406         else t1 = r1->type;
7407         if(r2 == NULL) t2 = LDNS_RR_TYPE_RRSIG;
7408         else t2 = r2->type;
7409         if(t1 < t2)
7410                 return -1;
7411         if(t1 > t2)
7412                 return 1;
7413         return 0;
7414 }
7415
7416 /** add type RRSIG to rr list if not one there already,
7417  * this is to perform RRSIG collate processing at that point. */
7418 static void addrrsigtype_if_needed(struct auth_rrset** array,
7419         size_t arraysize, size_t* rrnum, struct auth_data* node)
7420 {
7421         if(az_domain_rrset(node, LDNS_RR_TYPE_RRSIG))
7422                 return; /* already one there */
7423         if((*rrnum) >= arraysize)
7424                 return; /* array too small? */
7425         array[*rrnum] = NULL; /* nothing there, but need entry in list */
7426         (*rrnum)++;
7427 }
7428
7429 /** collate the RRs in an RRset using the simple scheme */
7430 static int zonemd_simple_rrset(struct auth_zone* z, int hashalgo,
7431         struct secalgo_hash* h, struct auth_data* node,
7432         struct auth_rrset* rrset, struct regional* region,
7433         struct sldns_buffer* buf, char** reason)
7434 {
7435         /* canonicalize */
7436         struct ub_packed_rrset_key key;
7437         memset(&key, 0, sizeof(key));
7438         key.entry.key = &key;
7439         key.entry.data = rrset->data;
7440         key.rk.dname = node->name;
7441         key.rk.dname_len = node->namelen;
7442         key.rk.type = htons(rrset->type);
7443         key.rk.rrset_class = htons(z->dclass);
7444         if(!rrset_canonicalize_to_buffer(region, buf, &key)) {
7445                 *reason = "out of memory";
7446                 return 0;
7447         }
7448         regional_free_all(region);
7449
7450         /* hash */
7451         if(!zonemd_digest_update(hashalgo, h, sldns_buffer_begin(buf),
7452                 sldns_buffer_limit(buf), reason)) {
7453                 return 0;
7454         }
7455         return 1;
7456 }
7457
7458 /** count number of RRSIGs in a domain name rrset list */
7459 static size_t zonemd_simple_count_rrsig(struct auth_rrset* rrset,
7460         struct auth_rrset** rrlist, size_t rrnum,
7461         struct auth_zone* z, struct auth_data* node)
7462 {
7463         size_t i, count = 0;
7464         if(rrset) {
7465                 size_t j;
7466                 for(j = 0; j<rrset->data->count; j++) {
7467                         if(rrsig_rdata_get_type_covered(rrset->data->
7468                                 rr_data[j], rrset->data->rr_len[j]) ==
7469                                 LDNS_RR_TYPE_ZONEMD &&
7470                                 query_dname_compare(z->name, node->name)==0) {
7471                                 /* omit RRSIGs over type ZONEMD at apex */
7472                                 continue;
7473                         }
7474                         count++;
7475                 }
7476         }
7477         for(i=0; i<rrnum; i++) {
7478                 if(rrlist[i] && rrlist[i]->type == LDNS_RR_TYPE_ZONEMD &&
7479                         query_dname_compare(z->name, node->name)==0) {
7480                         /* omit RRSIGs over type ZONEMD at apex */
7481                         continue;
7482                 }
7483                 count += (rrlist[i]?rrlist[i]->data->rrsig_count:0);
7484         }
7485         return count;
7486 }
7487
7488 /** allocate sparse rrset data for the number of entries in tepm region */
7489 static int zonemd_simple_rrsig_allocs(struct regional* region,
7490         struct packed_rrset_data* data, size_t count)
7491 {
7492         data->rr_len = regional_alloc(region, sizeof(*data->rr_len) * count);
7493         if(!data->rr_len) {
7494                 return 0;
7495         }
7496         data->rr_ttl = regional_alloc(region, sizeof(*data->rr_ttl) * count);
7497         if(!data->rr_ttl) {
7498                 return 0;
7499         }
7500         data->rr_data = regional_alloc(region, sizeof(*data->rr_data) * count);
7501         if(!data->rr_data) {
7502                 return 0;
7503         }
7504         return 1;
7505 }
7506
7507 /** add the RRSIGs from the rrs in the domain into the data */
7508 static void add_rrlist_rrsigs_into_data(struct packed_rrset_data* data,
7509         size_t* done, struct auth_rrset** rrlist, size_t rrnum,
7510         struct auth_zone* z, struct auth_data* node)
7511 {
7512         size_t i;
7513         for(i=0; i<rrnum; i++) {
7514                 size_t j;
7515                 if(!rrlist[i])
7516                         continue;
7517                 if(rrlist[i]->type == LDNS_RR_TYPE_ZONEMD &&
7518                         query_dname_compare(z->name, node->name)==0) {
7519                         /* omit RRSIGs over type ZONEMD at apex */
7520                         continue;
7521                 }
7522                 for(j = 0; j<rrlist[i]->data->rrsig_count; j++) {
7523                         data->rr_len[*done] = rrlist[i]->data->rr_len[rrlist[i]->data->count + j];
7524                         data->rr_ttl[*done] = rrlist[i]->data->rr_ttl[rrlist[i]->data->count + j];
7525                         /* reference the rdata in the rrset, no need to
7526                          * copy it, it is no longer needed at the end of
7527                          * the routine */
7528                         data->rr_data[*done] = rrlist[i]->data->rr_data[rrlist[i]->data->count + j];
7529                         (*done)++;
7530                 }
7531         }
7532 }
7533
7534 static void add_rrset_into_data(struct packed_rrset_data* data,
7535         size_t* done, struct auth_rrset* rrset,
7536         struct auth_zone* z, struct auth_data* node)
7537 {
7538         if(rrset) {
7539                 size_t j;
7540                 for(j = 0; j<rrset->data->count; j++) {
7541                         if(rrsig_rdata_get_type_covered(rrset->data->
7542                                 rr_data[j], rrset->data->rr_len[j]) ==
7543                                 LDNS_RR_TYPE_ZONEMD &&
7544                                 query_dname_compare(z->name, node->name)==0) {
7545                                 /* omit RRSIGs over type ZONEMD at apex */
7546                                 continue;
7547                         }
7548                         data->rr_len[*done] = rrset->data->rr_len[j];
7549                         data->rr_ttl[*done] = rrset->data->rr_ttl[j];
7550                         /* reference the rdata in the rrset, no need to
7551                          * copy it, it is no longer need at the end of
7552                          * the routine */
7553                         data->rr_data[*done] = rrset->data->rr_data[j];
7554                         (*done)++;
7555                 }
7556         }
7557 }
7558
7559 /** collate the RRSIGs using the simple scheme */
7560 static int zonemd_simple_rrsig(struct auth_zone* z, int hashalgo,
7561         struct secalgo_hash* h, struct auth_data* node,
7562         struct auth_rrset* rrset, struct auth_rrset** rrlist, size_t rrnum,
7563         struct regional* region, struct sldns_buffer* buf, char** reason)
7564 {
7565         /* the rrset pointer can be NULL, this means it is type RRSIG and
7566          * there is no ordinary type RRSIG there.  The RRSIGs are stored
7567          * with the RRsets in their data.
7568          *
7569          * The RRset pointer can be nonNULL. This happens if there is
7570          * no RR that is covered by the RRSIG for the domain.  Then this
7571          * RRSIG RR is stored in an rrset of type RRSIG. The other RRSIGs
7572          * are stored in the rrset entries for the RRs in the rr list for
7573          * the domain node.  We need to collate the rrset's data, if any, and
7574          * the rrlist's rrsigs */
7575         /* if this is the apex, omit RRSIGs that cover type ZONEMD */
7576         /* build rrsig rrset */
7577         size_t done = 0;
7578         struct ub_packed_rrset_key key;
7579         struct packed_rrset_data data;
7580         memset(&key, 0, sizeof(key));
7581         memset(&data, 0, sizeof(data));
7582         key.entry.key = &key;
7583         key.entry.data = &data;
7584         key.rk.dname = node->name;
7585         key.rk.dname_len = node->namelen;
7586         key.rk.type = htons(LDNS_RR_TYPE_RRSIG);
7587         key.rk.rrset_class = htons(z->dclass);
7588         data.count = zonemd_simple_count_rrsig(rrset, rrlist, rrnum, z, node);
7589         if(!zonemd_simple_rrsig_allocs(region, &data, data.count)) {
7590                 *reason = "out of memory";
7591                 regional_free_all(region);
7592                 return 0;
7593         }
7594         /* all the RRSIGs stored in the other rrsets for this domain node */
7595         add_rrlist_rrsigs_into_data(&data, &done, rrlist, rrnum, z, node);
7596         /* plus the RRSIGs stored in an rrset of type RRSIG for this node */
7597         add_rrset_into_data(&data, &done, rrset, z, node);
7598
7599         /* canonicalize */
7600         if(!rrset_canonicalize_to_buffer(region, buf, &key)) {
7601                 *reason = "out of memory";
7602                 regional_free_all(region);
7603                 return 0;
7604         }
7605         regional_free_all(region);
7606
7607         /* hash */
7608         if(!zonemd_digest_update(hashalgo, h, sldns_buffer_begin(buf),
7609                 sldns_buffer_limit(buf), reason)) {
7610                 return 0;
7611         }
7612         return 1;
7613 }
7614
7615 /** collate a domain's rrsets using the simple scheme */
7616 static int zonemd_simple_domain(struct auth_zone* z, int hashalgo,
7617         struct secalgo_hash* h, struct auth_data* node,
7618         struct regional* region, struct sldns_buffer* buf, char** reason)
7619 {
7620         const size_t rrlistsize = 65536;
7621         struct auth_rrset* rrlist[rrlistsize];
7622         size_t i, rrnum = 0;
7623         /* see if the domain is out of scope, the zone origin,
7624          * that would be omitted */
7625         if(!dname_subdomain_c(node->name, z->name))
7626                 return 1; /* continue */
7627         /* loop over the rrsets in ascending order. */
7628         rrnum = authdata_rrsets_to_list(rrlist, rrlistsize, node->rrsets);
7629         addrrsigtype_if_needed(rrlist, rrlistsize, &rrnum, node);
7630         qsort(rrlist, rrnum, sizeof(*rrlist), rrlist_compare);
7631         for(i=0; i<rrnum; i++) {
7632                 if(rrlist[i] && rrlist[i]->type == LDNS_RR_TYPE_ZONEMD &&
7633                         query_dname_compare(z->name, node->name) == 0) {
7634                         /* omit type ZONEMD at apex */
7635                         continue;
7636                 }
7637                 if(rrlist[i] == NULL || rrlist[i]->type ==
7638                         LDNS_RR_TYPE_RRSIG) {
7639                         if(!zonemd_simple_rrsig(z, hashalgo, h, node,
7640                                 rrlist[i], rrlist, rrnum, region, buf, reason))
7641                                 return 0;
7642                 } else if(!zonemd_simple_rrset(z, hashalgo, h, node,
7643                         rrlist[i], region, buf, reason)) {
7644                         return 0;
7645                 }
7646         }
7647         return 1;
7648 }
7649
7650 /** collate the zone using the simple scheme */
7651 static int zonemd_simple_collate(struct auth_zone* z, int hashalgo,
7652         struct secalgo_hash* h, struct regional* region,
7653         struct sldns_buffer* buf, char** reason)
7654 {
7655         /* our tree is sorted in canonical order, so we can just loop over
7656          * the tree */
7657         struct auth_data* n;
7658         RBTREE_FOR(n, struct auth_data*, &z->data) {
7659                 if(!zonemd_simple_domain(z, hashalgo, h, n, region, buf,
7660                         reason))
7661                         return 0;
7662         }
7663         return 1;
7664 }
7665
7666 int auth_zone_generate_zonemd_hash(struct auth_zone* z, int scheme,
7667         int hashalgo, uint8_t* hash, size_t hashlen, size_t* resultlen,
7668         struct regional* region, struct sldns_buffer* buf, char** reason)
7669 {
7670         struct secalgo_hash* h = zonemd_digest_init(hashalgo, reason);
7671         if(!h) {
7672                 if(!*reason)
7673                         *reason = "digest init fail";
7674                 return 0;
7675         }
7676         if(scheme == ZONEMD_SCHEME_SIMPLE) {
7677                 if(!zonemd_simple_collate(z, hashalgo, h, region, buf, reason)) {
7678                         if(!*reason) *reason = "scheme simple collate fail";
7679                         secalgo_hash_delete(h);
7680                         return 0;
7681                 }
7682         }
7683         if(!zonemd_digest_finish(hashalgo, h, hash, hashlen, resultlen,
7684                 reason)) {
7685                 secalgo_hash_delete(h);
7686                 *reason = "digest finish fail";
7687                 return 0;
7688         }
7689         secalgo_hash_delete(h);
7690         return 1;
7691 }
7692
7693 int auth_zone_generate_zonemd_check(struct auth_zone* z, int scheme,
7694         int hashalgo, uint8_t* hash, size_t hashlen, struct regional* region,
7695         struct sldns_buffer* buf, char** reason)
7696 {
7697         uint8_t gen[512];
7698         size_t genlen = 0;
7699         *reason = NULL;
7700         if(!zonemd_hashalgo_supported(hashalgo)) {
7701                 /* allow it */
7702                 *reason = "unsupported algorithm";
7703                 return 1;
7704         }
7705         if(!zonemd_scheme_supported(scheme)) {
7706                 /* allow it */
7707                 *reason = "unsupported scheme";
7708                 return 1;
7709         }
7710         if(hashlen < 12) {
7711                 /* the ZONEMD draft requires digests to fail if too small */
7712                 *reason = "digest length too small, less than 12";
7713                 return 0;
7714         }
7715         /* generate digest */
7716         if(!auth_zone_generate_zonemd_hash(z, scheme, hashalgo, gen,
7717                 sizeof(gen), &genlen, region, buf, reason)) {
7718                 /* reason filled in by zonemd hash routine */
7719                 return 0;
7720         }
7721         /* check digest length */
7722         if(hashlen != genlen) {
7723                 *reason = "incorrect digest length";
7724                 if(verbosity >= VERB_ALGO) {
7725                         verbose(VERB_ALGO, "zonemd scheme=%d hashalgo=%d",
7726                                 scheme, hashalgo);
7727                         log_hex("ZONEMD should be  ", gen, genlen);
7728                         log_hex("ZONEMD to check is", hash, hashlen);
7729                 }
7730                 return 0;
7731         }
7732         /* check digest */
7733         if(memcmp(hash, gen, genlen) != 0) {
7734                 *reason = "incorrect digest";
7735                 if(verbosity >= VERB_ALGO) {
7736                         verbose(VERB_ALGO, "zonemd scheme=%d hashalgo=%d",
7737                                 scheme, hashalgo);
7738                         log_hex("ZONEMD should be  ", gen, genlen);
7739                         log_hex("ZONEMD to check is", hash, hashlen);
7740                 }
7741                 return 0;
7742         }
7743         return 1;
7744 }
7745
7746 /** log auth zone message with zone name in front. */
7747 static void auth_zone_log(uint8_t* name, enum verbosity_value level,
7748         const char* format, ...) ATTR_FORMAT(printf, 3, 4);
7749 static void auth_zone_log(uint8_t* name, enum verbosity_value level,
7750         const char* format, ...)
7751 {
7752         va_list args;
7753         va_start(args, format);
7754         if(verbosity >= level) {
7755                 char str[255+1];
7756                 char msg[MAXSYSLOGMSGLEN];
7757                 dname_str(name, str);
7758                 vsnprintf(msg, sizeof(msg), format, args);
7759                 verbose(level, "auth zone %s %s", str, msg);
7760         }
7761         va_end(args);
7762 }
7763
7764 /** ZONEMD, dnssec verify the rrset with the dnskey */
7765 static int zonemd_dnssec_verify_rrset(struct auth_zone* z,
7766         struct module_env* env, struct module_stack* mods,
7767         struct ub_packed_rrset_key* dnskey, struct auth_data* node,
7768         struct auth_rrset* rrset, char** why_bogus, uint8_t* sigalg)
7769 {
7770         struct ub_packed_rrset_key pk;
7771         enum sec_status sec;
7772         struct val_env* ve;
7773         int m;
7774         m = modstack_find(mods, "validator");
7775         if(m == -1) {
7776                 auth_zone_log(z->name, VERB_ALGO, "zonemd dnssec verify: have "
7777                         "DNSKEY chain of trust, but no validator module");
7778                 return 0;
7779         }
7780         ve = (struct val_env*)env->modinfo[m];
7781
7782         memset(&pk, 0, sizeof(pk));
7783         pk.entry.key = &pk;
7784         pk.entry.data = rrset->data;
7785         pk.rk.dname = node->name;
7786         pk.rk.dname_len = node->namelen;
7787         pk.rk.type = htons(rrset->type);
7788         pk.rk.rrset_class = htons(z->dclass);
7789         if(verbosity >= VERB_ALGO) {
7790                 char typestr[32];
7791                 typestr[0]=0;
7792                 sldns_wire2str_type_buf(rrset->type, typestr, sizeof(typestr));
7793                 auth_zone_log(z->name, VERB_ALGO,
7794                         "zonemd: verify %s RRset with DNSKEY", typestr);
7795         }
7796         sec = dnskeyset_verify_rrset(env, ve, &pk, dnskey, sigalg, why_bogus, NULL,
7797                 LDNS_SECTION_ANSWER, NULL);
7798         if(sec == sec_status_secure) {
7799                 return 1;
7800         }
7801         if(why_bogus)
7802                 auth_zone_log(z->name, VERB_ALGO, "DNSSEC verify was bogus: %s", *why_bogus);
7803         return 0;
7804 }
7805
7806 /** check for nsec3, the RR with params equal, if bitmap has the type */
7807 static int nsec3_of_param_has_type(struct auth_rrset* nsec3, int algo,
7808         size_t iter, uint8_t* salt, size_t saltlen, uint16_t rrtype)
7809 {
7810         int i, count = (int)nsec3->data->count;
7811         struct ub_packed_rrset_key pk;
7812         memset(&pk, 0, sizeof(pk));
7813         pk.entry.data = nsec3->data;
7814         for(i=0; i<count; i++) {
7815                 int rralgo;
7816                 size_t rriter, rrsaltlen;
7817                 uint8_t* rrsalt;
7818                 if(!nsec3_get_params(&pk, i, &rralgo, &rriter, &rrsalt,
7819                         &rrsaltlen))
7820                         continue; /* no parameters, malformed */
7821                 if(rralgo != algo || rriter != iter || rrsaltlen != saltlen)
7822                         continue; /* different parameters */
7823                 if(saltlen != 0) {
7824                         if(rrsalt == NULL || salt == NULL)
7825                                 continue;
7826                         if(memcmp(rrsalt, salt, saltlen) != 0)
7827                                 continue; /* different salt parameters */
7828                 }
7829                 if(nsec3_has_type(&pk, i, rrtype))
7830                         return 1;
7831         }
7832         return 0;
7833 }
7834
7835 /** Verify the absence of ZONEMD with DNSSEC by checking NSEC, NSEC3 type flag.
7836  * return false on failure, reason contains description of failure. */
7837 static int zonemd_check_dnssec_absence(struct auth_zone* z,
7838         struct module_env* env, struct module_stack* mods,
7839         struct ub_packed_rrset_key* dnskey, struct auth_data* apex,
7840         char** reason, char** why_bogus, uint8_t* sigalg)
7841 {
7842         struct auth_rrset* nsec = NULL;
7843         if(!apex) {
7844                 *reason = "zone has no apex domain but ZONEMD missing";
7845                 return 0;
7846         }
7847         nsec = az_domain_rrset(apex, LDNS_RR_TYPE_NSEC);
7848         if(nsec) {
7849                 struct ub_packed_rrset_key pk;
7850                 /* dnssec verify the NSEC */
7851                 if(!zonemd_dnssec_verify_rrset(z, env, mods, dnskey, apex,
7852                         nsec, why_bogus, sigalg)) {
7853                         *reason = "DNSSEC verify failed for NSEC RRset";
7854                         return 0;
7855                 }
7856                 /* check type bitmap */
7857                 memset(&pk, 0, sizeof(pk));
7858                 pk.entry.data = nsec->data;
7859                 if(nsec_has_type(&pk, LDNS_RR_TYPE_ZONEMD)) {
7860                         *reason = "DNSSEC NSEC bitmap says type ZONEMD exists";
7861                         return 0;
7862                 }
7863                 auth_zone_log(z->name, VERB_ALGO, "zonemd DNSSEC NSEC verification of absence of ZONEMD secure");
7864         } else {
7865                 /* NSEC3 perhaps ? */
7866                 int algo;
7867                 size_t iter, saltlen;
7868                 uint8_t* salt;
7869                 struct auth_rrset* nsec3param = az_domain_rrset(apex,
7870                         LDNS_RR_TYPE_NSEC3PARAM);
7871                 struct auth_data* match;
7872                 struct auth_rrset* nsec3;
7873                 if(!nsec3param) {
7874                         *reason = "zone has no NSEC information but ZONEMD missing";
7875                         return 0;
7876                 }
7877                 if(!az_nsec3_param(z, &algo, &iter, &salt, &saltlen)) {
7878                         *reason = "zone has no NSEC information but ZONEMD missing";
7879                         return 0;
7880                 }
7881                 /* find the NSEC3 record */
7882                 match = az_nsec3_find_exact(z, z->name, z->namelen, algo,
7883                         iter, salt, saltlen);
7884                 if(!match) {
7885                         *reason = "zone has no NSEC3 domain for the apex but ZONEMD missing";
7886                         return 0;
7887                 }
7888                 nsec3 = az_domain_rrset(match, LDNS_RR_TYPE_NSEC3);
7889                 if(!nsec3) {
7890                         *reason = "zone has no NSEC3 RRset for the apex but ZONEMD missing";
7891                         return 0;
7892                 }
7893                 /* dnssec verify the NSEC3 */
7894                 if(!zonemd_dnssec_verify_rrset(z, env, mods, dnskey, match,
7895                         nsec3, why_bogus, sigalg)) {
7896                         *reason = "DNSSEC verify failed for NSEC3 RRset";
7897                         return 0;
7898                 }
7899                 /* check type bitmap */
7900                 if(nsec3_of_param_has_type(nsec3, algo, iter, salt, saltlen,
7901                         LDNS_RR_TYPE_ZONEMD)) {
7902                         *reason = "DNSSEC NSEC3 bitmap says type ZONEMD exists";
7903                         return 0;
7904                 }
7905                 auth_zone_log(z->name, VERB_ALGO, "zonemd DNSSEC NSEC3 verification of absence of ZONEMD secure");
7906         }
7907
7908         return 1;
7909 }
7910
7911 /** Verify the SOA and ZONEMD DNSSEC signatures.
7912  * return false on failure, reason contains description of failure. */
7913 static int zonemd_check_dnssec_soazonemd(struct auth_zone* z,
7914         struct module_env* env, struct module_stack* mods,
7915         struct ub_packed_rrset_key* dnskey, struct auth_data* apex,
7916         struct auth_rrset* zonemd_rrset, char** reason, char** why_bogus,
7917         uint8_t* sigalg)
7918 {
7919         struct auth_rrset* soa;
7920         if(!apex) {
7921                 *reason = "zone has no apex domain";
7922                 return 0;
7923         }
7924         soa = az_domain_rrset(apex, LDNS_RR_TYPE_SOA);
7925         if(!soa) {
7926                 *reason = "zone has no SOA RRset";
7927                 return 0;
7928         }
7929         if(!zonemd_dnssec_verify_rrset(z, env, mods, dnskey, apex, soa,
7930                 why_bogus, sigalg)) {
7931                 *reason = "DNSSEC verify failed for SOA RRset";
7932                 return 0;
7933         }
7934         if(!zonemd_dnssec_verify_rrset(z, env, mods, dnskey, apex,
7935                 zonemd_rrset, why_bogus, sigalg)) {
7936                 *reason = "DNSSEC verify failed for ZONEMD RRset";
7937                 return 0;
7938         }
7939         auth_zone_log(z->name, VERB_ALGO, "zonemd DNSSEC verification of SOA and ZONEMD RRsets secure");
7940         return 1;
7941 }
7942
7943 /**
7944  * Fail the ZONEMD verification.
7945  * @param z: auth zone that fails.
7946  * @param env: environment with config, to ignore failure or not.
7947  * @param reason: failure string description.
7948  * @param why_bogus: failure string for DNSSEC verification failure.
7949  * @param result: strdup result in here if not NULL.
7950  */
7951 static void auth_zone_zonemd_fail(struct auth_zone* z, struct module_env* env,
7952         char* reason, char* why_bogus, char** result)
7953 {
7954         char zstr[255+1];
7955         /* if fail: log reason, and depending on config also take action
7956          * and drop the zone, eg. it is gone from memory, set zone_expired */
7957         dname_str(z->name, zstr);
7958         if(!reason) reason = "verification failed";
7959         if(result) {
7960                 if(why_bogus) {
7961                         char res[1024];
7962                         snprintf(res, sizeof(res), "%s: %s", reason,
7963                                 why_bogus);
7964                         *result = strdup(res);
7965                 } else {
7966                         *result = strdup(reason);
7967                 }
7968                 if(!*result) log_err("out of memory");
7969         } else {
7970                 log_warn("auth zone %s: ZONEMD verification failed: %s", zstr, reason);
7971         }
7972
7973         if(env->cfg->zonemd_permissive_mode) {
7974                 verbose(VERB_ALGO, "zonemd-permissive-mode enabled, "
7975                         "not blocking zone %s", zstr);
7976                 return;
7977         }
7978
7979         /* expired means the zone gives servfail and is not used by
7980          * lookup if fallback_enabled*/
7981         z->zone_expired = 1;
7982 }
7983
7984 /**
7985  * Verify the zonemd with DNSSEC and hash check, with given key.
7986  * @param z: auth zone.
7987  * @param env: environment with config and temp buffers.
7988  * @param mods: module stack with validator env for verification.
7989  * @param dnskey: dnskey that we can use, or NULL.  If nonnull, the key
7990  *      has been verified and is the start of the chain of trust.
7991  * @param is_insecure: if true, the dnskey is not used, the zone is insecure.
7992  *      And dnssec is not used.  It is DNSSEC secure insecure or not under
7993  *      a trust anchor.
7994  * @param sigalg: if nonNULL provide algorithm downgrade protection.
7995  *      Otherwise one algorithm is enough. Must have space of ALGO_NEEDS_MAX+1.
7996  * @param result: if not NULL result reason copied here.
7997  */
7998 static void
7999 auth_zone_verify_zonemd_with_key(struct auth_zone* z, struct module_env* env,
8000         struct module_stack* mods, struct ub_packed_rrset_key* dnskey,
8001         int is_insecure, char** result, uint8_t* sigalg)
8002 {
8003         char* reason = NULL, *why_bogus = NULL;
8004         struct auth_data* apex = NULL;
8005         struct auth_rrset* zonemd_rrset = NULL;
8006         int zonemd_absent = 0, zonemd_absence_dnssecok = 0;
8007
8008         /* see if ZONEMD is present or absent. */
8009         apex = az_find_name(z, z->name, z->namelen);
8010         if(!apex) {
8011                 zonemd_absent = 1;
8012         } else {
8013                 zonemd_rrset = az_domain_rrset(apex, LDNS_RR_TYPE_ZONEMD);
8014                 if(!zonemd_rrset || zonemd_rrset->data->count==0) {
8015                         zonemd_absent = 1;
8016                         zonemd_rrset = NULL;
8017                 }
8018         }
8019
8020         /* if no DNSSEC, done. */
8021         /* if no ZONEMD, and DNSSEC, use DNSKEY to verify NSEC or NSEC3 for
8022          * zone apex.  Check ZONEMD bit is turned off or else fail */
8023         /* if ZONEMD, and DNSSEC, check DNSSEC signature on SOA and ZONEMD,
8024          * or else fail */
8025         if(!dnskey && !is_insecure) {
8026                 auth_zone_zonemd_fail(z, env, "DNSKEY missing", NULL, result);
8027                 return;
8028         } else if(!zonemd_rrset && dnskey && !is_insecure) {
8029                 /* fetch, DNSSEC verify, and check NSEC/NSEC3 */
8030                 if(!zonemd_check_dnssec_absence(z, env, mods, dnskey, apex,
8031                         &reason, &why_bogus, sigalg)) {
8032                         auth_zone_zonemd_fail(z, env, reason, why_bogus, result);
8033                         return;
8034                 }
8035                 zonemd_absence_dnssecok = 1;
8036         } else if(zonemd_rrset && dnskey && !is_insecure) {
8037                 /* check DNSSEC verify of SOA and ZONEMD */
8038                 if(!zonemd_check_dnssec_soazonemd(z, env, mods, dnskey, apex,
8039                         zonemd_rrset, &reason, &why_bogus, sigalg)) {
8040                         auth_zone_zonemd_fail(z, env, reason, why_bogus, result);
8041                         return;
8042                 }
8043         }
8044
8045         if(zonemd_absent && z->zonemd_reject_absence) {
8046                 auth_zone_zonemd_fail(z, env, "ZONEMD absent and that is not allowed by config", NULL, result);
8047                 return;
8048         }
8049         if(zonemd_absent && zonemd_absence_dnssecok) {
8050                 auth_zone_log(z->name, VERB_ALGO, "DNSSEC verified nonexistence of ZONEMD");
8051                 if(result) {
8052                         *result = strdup("DNSSEC verified nonexistence of ZONEMD");
8053                         if(!*result) log_err("out of memory");
8054                 }
8055                 return;
8056         }
8057         if(zonemd_absent) {
8058                 auth_zone_log(z->name, VERB_ALGO, "no ZONEMD present");
8059                 if(result) {
8060                         *result = strdup("no ZONEMD present");
8061                         if(!*result) log_err("out of memory");
8062                 }
8063                 return;
8064         }
8065
8066         /* check ZONEMD checksum and report or else fail. */
8067         if(!auth_zone_zonemd_check_hash(z, env, &reason)) {
8068                 auth_zone_zonemd_fail(z, env, reason, NULL, result);
8069                 return;
8070         }
8071
8072         /* success! log the success */
8073         if(reason)
8074                 auth_zone_log(z->name, VERB_ALGO, "ZONEMD %s", reason);
8075         else    auth_zone_log(z->name, VERB_ALGO, "ZONEMD verification successful");
8076         if(result) {
8077                 if(reason)
8078                         *result = strdup(reason);
8079                 else    *result = strdup("ZONEMD verification successful");
8080                 if(!*result) log_err("out of memory");
8081         }
8082 }
8083
8084 /**
8085  * verify the zone DNSKEY rrset from the trust anchor
8086  * This is possible because the anchor is for the zone itself, and can
8087  * thus apply straight to the zone DNSKEY set.
8088  * @param z: the auth zone.
8089  * @param env: environment with time and temp buffers.
8090  * @param mods: module stack for validator environment for dnssec validation.
8091  * @param anchor: trust anchor to use
8092  * @param is_insecure: returned, true if the zone is securely insecure.
8093  * @param why_bogus: if the routine fails, returns the failure reason.
8094  * @param keystorage: where to store the ub_packed_rrset_key that is created
8095  *      on success. A pointer to it is returned on success.
8096  * @return the dnskey RRset, reference to zone data and keystorage, or
8097  *      NULL on failure.
8098  */
8099 static struct ub_packed_rrset_key*
8100 zonemd_get_dnskey_from_anchor(struct auth_zone* z, struct module_env* env,
8101         struct module_stack* mods, struct trust_anchor* anchor,
8102         int* is_insecure, char** why_bogus,
8103         struct ub_packed_rrset_key* keystorage)
8104 {
8105         struct auth_data* apex;
8106         struct auth_rrset* dnskey_rrset;
8107         enum sec_status sec;
8108         struct val_env* ve;
8109         int m;
8110
8111         apex = az_find_name(z, z->name, z->namelen);
8112         if(!apex) {
8113                 *why_bogus = "have trust anchor, but zone has no apex domain for DNSKEY";
8114                 return 0;
8115         }
8116         dnskey_rrset = az_domain_rrset(apex, LDNS_RR_TYPE_DNSKEY);
8117         if(!dnskey_rrset || dnskey_rrset->data->count==0) {
8118                 *why_bogus = "have trust anchor, but zone has no DNSKEY";
8119                 return 0;
8120         }
8121
8122         m = modstack_find(mods, "validator");
8123         if(m == -1) {
8124                 *why_bogus = "have trust anchor, but no validator module";
8125                 return 0;
8126         }
8127         ve = (struct val_env*)env->modinfo[m];
8128
8129         memset(keystorage, 0, sizeof(*keystorage));
8130         keystorage->entry.key = keystorage;
8131         keystorage->entry.data = dnskey_rrset->data;
8132         keystorage->rk.dname = apex->name;
8133         keystorage->rk.dname_len = apex->namelen;
8134         keystorage->rk.type = htons(LDNS_RR_TYPE_DNSKEY);
8135         keystorage->rk.rrset_class = htons(z->dclass);
8136         auth_zone_log(z->name, VERB_QUERY,
8137                 "zonemd: verify DNSKEY RRset with trust anchor");
8138         sec = val_verify_DNSKEY_with_TA(env, ve, keystorage, anchor->ds_rrset,
8139                 anchor->dnskey_rrset, NULL, why_bogus, NULL, NULL);
8140         regional_free_all(env->scratch);
8141         if(sec == sec_status_secure) {
8142                 /* success */
8143                 *is_insecure = 0;
8144                 return keystorage;
8145         } else if(sec == sec_status_insecure) {
8146                 /* insecure */
8147                 *is_insecure = 1;
8148         } else {
8149                 /* bogus */
8150                 *is_insecure = 0;
8151                 auth_zone_log(z->name, VERB_ALGO,
8152                         "zonemd: verify DNSKEY RRset with trust anchor failed: %s", *why_bogus);
8153         }
8154         return NULL;
8155 }
8156
8157 /** verify the DNSKEY from the zone with looked up DS record */
8158 static struct ub_packed_rrset_key*
8159 auth_zone_verify_zonemd_key_with_ds(struct auth_zone* z,
8160         struct module_env* env, struct module_stack* mods,
8161         struct ub_packed_rrset_key* ds, int* is_insecure, char** why_bogus,
8162         struct ub_packed_rrset_key* keystorage, uint8_t* sigalg)
8163 {
8164         struct auth_data* apex;
8165         struct auth_rrset* dnskey_rrset;
8166         enum sec_status sec;
8167         struct val_env* ve;
8168         int m;
8169
8170         /* fetch DNSKEY from zone data */
8171         apex = az_find_name(z, z->name, z->namelen);
8172         if(!apex) {
8173                 *why_bogus = "in verifywithDS, zone has no apex";
8174                 return NULL;
8175         }
8176         dnskey_rrset = az_domain_rrset(apex, LDNS_RR_TYPE_DNSKEY);
8177         if(!dnskey_rrset || dnskey_rrset->data->count==0) {
8178                 *why_bogus = "in verifywithDS, zone has no DNSKEY";
8179                 return NULL;
8180         }
8181
8182         m = modstack_find(mods, "validator");
8183         if(m == -1) {
8184                 *why_bogus = "in verifywithDS, have no validator module";
8185                 return NULL;
8186         }
8187         ve = (struct val_env*)env->modinfo[m];
8188
8189         memset(keystorage, 0, sizeof(*keystorage));
8190         keystorage->entry.key = keystorage;
8191         keystorage->entry.data = dnskey_rrset->data;
8192         keystorage->rk.dname = apex->name;
8193         keystorage->rk.dname_len = apex->namelen;
8194         keystorage->rk.type = htons(LDNS_RR_TYPE_DNSKEY);
8195         keystorage->rk.rrset_class = htons(z->dclass);
8196         auth_zone_log(z->name, VERB_QUERY, "zonemd: verify zone DNSKEY with DS");
8197         sec = val_verify_DNSKEY_with_DS(env, ve, keystorage, ds, sigalg,
8198                 why_bogus, NULL, NULL);
8199         regional_free_all(env->scratch);
8200         if(sec == sec_status_secure) {
8201                 /* success */
8202                 return keystorage;
8203         } else if(sec == sec_status_insecure) {
8204                 /* insecure */
8205                 *is_insecure = 1;
8206         } else {
8207                 /* bogus */
8208                 *is_insecure = 0;
8209                 if(*why_bogus == NULL)
8210                         *why_bogus = "verify failed";
8211                 auth_zone_log(z->name, VERB_ALGO,
8212                         "zonemd: verify DNSKEY RRset with DS failed: %s",
8213                         *why_bogus);
8214         }
8215         return NULL;
8216 }
8217
8218 /** callback for ZONEMD lookup of DNSKEY */
8219 void auth_zonemd_dnskey_lookup_callback(void* arg, int rcode, sldns_buffer* buf,
8220         enum sec_status sec, char* why_bogus, int ATTR_UNUSED(was_ratelimited))
8221 {
8222         struct auth_zone* z = (struct auth_zone*)arg;
8223         struct module_env* env;
8224         char* reason = NULL, *ds_bogus = NULL, *typestr="DNSKEY";
8225         struct ub_packed_rrset_key* dnskey = NULL, *ds = NULL;
8226         int is_insecure = 0, downprot;
8227         struct ub_packed_rrset_key keystorage;
8228         uint8_t sigalg[ALGO_NEEDS_MAX+1];
8229
8230         lock_rw_wrlock(&z->lock);
8231         env = z->zonemd_callback_env;
8232         /* release the env variable so another worker can pick up the
8233          * ZONEMD verification task if it wants to */
8234         z->zonemd_callback_env = NULL;
8235         if(!env || env->outnet->want_to_quit || z->zone_deleted) {
8236                 lock_rw_unlock(&z->lock);
8237                 return; /* stop on quit */
8238         }
8239         if(z->zonemd_callback_qtype == LDNS_RR_TYPE_DS)
8240                 typestr = "DS";
8241         downprot = env->cfg->harden_algo_downgrade;
8242
8243         /* process result */
8244         if(sec == sec_status_bogus) {
8245                 reason = why_bogus;
8246                 if(!reason) {
8247                         if(z->zonemd_callback_qtype == LDNS_RR_TYPE_DNSKEY)
8248                                 reason = "lookup of DNSKEY was bogus";
8249                         else    reason = "lookup of DS was bogus";
8250                 }
8251                 auth_zone_log(z->name, VERB_ALGO,
8252                         "zonemd lookup of %s was bogus: %s", typestr, reason);
8253         } else if(rcode == LDNS_RCODE_NOERROR) {
8254                 uint16_t wanted_qtype = z->zonemd_callback_qtype;
8255                 struct regional* temp = env->scratch;
8256                 struct query_info rq;
8257                 struct reply_info* rep;
8258                 memset(&rq, 0, sizeof(rq));
8259                 rep = parse_reply_in_temp_region(buf, temp, &rq);
8260                 if(rep && rq.qtype == wanted_qtype &&
8261                         query_dname_compare(z->name, rq.qname) == 0 &&
8262                         FLAGS_GET_RCODE(rep->flags) == LDNS_RCODE_NOERROR) {
8263                         /* parsed successfully */
8264                         struct ub_packed_rrset_key* answer =
8265                                 reply_find_answer_rrset(&rq, rep);
8266                         if(answer && sec == sec_status_secure) {
8267                                 if(z->zonemd_callback_qtype == LDNS_RR_TYPE_DNSKEY)
8268                                         dnskey = answer;
8269                                 else    ds = answer;
8270                                 auth_zone_log(z->name, VERB_ALGO,
8271                                         "zonemd lookup of %s was secure", typestr);
8272                         } else if(sec == sec_status_secure && !answer) {
8273                                 is_insecure = 1;
8274                                 auth_zone_log(z->name, VERB_ALGO,
8275                                         "zonemd lookup of %s has no content, but is secure, treat as insecure", typestr);
8276                         } else if(sec == sec_status_insecure) {
8277                                 is_insecure = 1;
8278                                 auth_zone_log(z->name, VERB_ALGO,
8279                                         "zonemd lookup of %s was insecure", typestr);
8280                         } else if(sec == sec_status_indeterminate) {
8281                                 is_insecure = 1;
8282                                 auth_zone_log(z->name, VERB_ALGO,
8283                                         "zonemd lookup of %s was indeterminate, treat as insecure", typestr);
8284                         } else {
8285                                 auth_zone_log(z->name, VERB_ALGO,
8286                                         "zonemd lookup of %s has nodata", typestr);
8287                                 if(z->zonemd_callback_qtype == LDNS_RR_TYPE_DNSKEY)
8288                                         reason = "lookup of DNSKEY has nodata";
8289                                 else    reason = "lookup of DS has nodata";
8290                         }
8291                 } else if(rep && rq.qtype == wanted_qtype &&
8292                         query_dname_compare(z->name, rq.qname) == 0 &&
8293                         FLAGS_GET_RCODE(rep->flags) == LDNS_RCODE_NXDOMAIN &&
8294                         sec == sec_status_secure) {
8295                         /* secure nxdomain, so the zone is like some RPZ zone
8296                          * that does not exist in the wider internet, with
8297                          * a secure nxdomain answer outside of it. So we
8298                          * treat the zonemd zone without a dnssec chain of
8299                          * trust, as insecure. */
8300                         is_insecure = 1;
8301                         auth_zone_log(z->name, VERB_ALGO,
8302                                 "zonemd lookup of %s was secure NXDOMAIN, treat as insecure", typestr);
8303                 } else if(rep && rq.qtype == wanted_qtype &&
8304                         query_dname_compare(z->name, rq.qname) == 0 &&
8305                         FLAGS_GET_RCODE(rep->flags) == LDNS_RCODE_NXDOMAIN &&
8306                         sec == sec_status_insecure) {
8307                         is_insecure = 1;
8308                         auth_zone_log(z->name, VERB_ALGO,
8309                                 "zonemd lookup of %s was insecure NXDOMAIN, treat as insecure", typestr);
8310                 } else if(rep && rq.qtype == wanted_qtype &&
8311                         query_dname_compare(z->name, rq.qname) == 0 &&
8312                         FLAGS_GET_RCODE(rep->flags) == LDNS_RCODE_NXDOMAIN &&
8313                         sec == sec_status_indeterminate) {
8314                         is_insecure = 1;
8315                         auth_zone_log(z->name, VERB_ALGO,
8316                                 "zonemd lookup of %s was indeterminate NXDOMAIN, treat as insecure", typestr);
8317                 } else {
8318                         auth_zone_log(z->name, VERB_ALGO,
8319                                 "zonemd lookup of %s has no answer", typestr);
8320                         if(z->zonemd_callback_qtype == LDNS_RR_TYPE_DNSKEY)
8321                                 reason = "lookup of DNSKEY has no answer";
8322                         else    reason = "lookup of DS has no answer";
8323                 }
8324         } else {
8325                 auth_zone_log(z->name, VERB_ALGO,
8326                         "zonemd lookup of %s failed", typestr);
8327                 if(z->zonemd_callback_qtype == LDNS_RR_TYPE_DNSKEY)
8328                         reason = "lookup of DNSKEY failed";
8329                 else    reason = "lookup of DS failed";
8330         }
8331
8332         if(!reason && !is_insecure && !dnskey && ds) {
8333                 dnskey = auth_zone_verify_zonemd_key_with_ds(z, env,
8334                         &env->mesh->mods, ds, &is_insecure, &ds_bogus,
8335                         &keystorage, downprot?sigalg:NULL);
8336                 if(!dnskey && !is_insecure && !reason)
8337                         reason = "DNSKEY verify with DS failed";
8338         }
8339
8340         if(reason) {
8341                 auth_zone_zonemd_fail(z, env, reason, ds_bogus, NULL);
8342                 lock_rw_unlock(&z->lock);
8343                 return;
8344         }
8345
8346         auth_zone_verify_zonemd_with_key(z, env, &env->mesh->mods, dnskey,
8347                 is_insecure, NULL, downprot?sigalg:NULL);
8348         regional_free_all(env->scratch);
8349         lock_rw_unlock(&z->lock);
8350 }
8351
8352 /** lookup DNSKEY for ZONEMD verification */
8353 static int
8354 zonemd_lookup_dnskey(struct auth_zone* z, struct module_env* env)
8355 {
8356         struct query_info qinfo;
8357         uint16_t qflags = BIT_RD;
8358         struct edns_data edns;
8359         sldns_buffer* buf = env->scratch_buffer;
8360         int fetch_ds = 0;
8361
8362         if(!z->fallback_enabled) {
8363                 /* we cannot actually get the DNSKEY, because it is in the
8364                  * zone we have ourselves, and it is not served yet
8365                  * (possibly), so fetch type DS */
8366                 fetch_ds = 1;
8367         }
8368         if(z->zonemd_callback_env) {
8369                 /* another worker is already working on the callback
8370                  * for the DNSKEY lookup for ZONEMD verification.
8371                  * We do not also have to do ZONEMD verification, let that
8372                  * worker do it */
8373                 auth_zone_log(z->name, VERB_ALGO,
8374                         "zonemd needs lookup of %s and that already is worked on by another worker", (fetch_ds?"DS":"DNSKEY"));
8375                 return 1;
8376         }
8377
8378         /* use mesh_new_callback to lookup the DNSKEY,
8379          * and then wait for them to be looked up (in cache, or query) */
8380         qinfo.qname_len = z->namelen;
8381         qinfo.qname = z->name;
8382         qinfo.qclass = z->dclass;
8383         if(fetch_ds)
8384                 qinfo.qtype = LDNS_RR_TYPE_DS;
8385         else    qinfo.qtype = LDNS_RR_TYPE_DNSKEY;
8386         qinfo.local_alias = NULL;
8387         if(verbosity >= VERB_ALGO) {
8388                 char buf1[512];
8389                 char buf2[LDNS_MAX_DOMAINLEN+1];
8390                 dname_str(z->name, buf2);
8391                 snprintf(buf1, sizeof(buf1), "auth zone %s: lookup %s "
8392                         "for zonemd verification", buf2,
8393                         (fetch_ds?"DS":"DNSKEY"));
8394                 log_query_info(VERB_ALGO, buf1, &qinfo);
8395         }
8396         edns.edns_present = 1;
8397         edns.ext_rcode = 0;
8398         edns.edns_version = 0;
8399         edns.bits = EDNS_DO;
8400         edns.opt_list_in = NULL;
8401         edns.opt_list_out = NULL;
8402         edns.opt_list_inplace_cb_out = NULL;
8403         if(sldns_buffer_capacity(buf) < 65535)
8404                 edns.udp_size = (uint16_t)sldns_buffer_capacity(buf);
8405         else    edns.udp_size = 65535;
8406
8407         /* store the worker-specific module env for the callback.
8408          * We can then reference this when the callback executes */
8409         z->zonemd_callback_env = env;
8410         z->zonemd_callback_qtype = qinfo.qtype;
8411         /* the callback can be called straight away */
8412         lock_rw_unlock(&z->lock);
8413         if(!mesh_new_callback(env->mesh, &qinfo, qflags, &edns, buf, 0,
8414                 &auth_zonemd_dnskey_lookup_callback, z, 0)) {
8415                 lock_rw_wrlock(&z->lock);
8416                 log_err("out of memory lookup of %s for zonemd",
8417                         (fetch_ds?"DS":"DNSKEY"));
8418                 return 0;
8419         }
8420         lock_rw_wrlock(&z->lock);
8421         return 1;
8422 }
8423
8424 void auth_zone_verify_zonemd(struct auth_zone* z, struct module_env* env,
8425         struct module_stack* mods, char** result, int offline, int only_online)
8426 {
8427         char* reason = NULL, *why_bogus = NULL;
8428         struct trust_anchor* anchor = NULL;
8429         struct ub_packed_rrset_key* dnskey = NULL;
8430         struct ub_packed_rrset_key keystorage;
8431         int is_insecure = 0;
8432         /* verify the ZONEMD if present.
8433          * If not present check if absence is allowed by DNSSEC */
8434         if(!z->zonemd_check)
8435                 return;
8436         if(z->data.count == 0)
8437                 return; /* no data */
8438
8439         /* if zone is under a trustanchor */
8440         /* is it equal to trustanchor - get dnskey's verified */
8441         /* else, find chain of trust by fetching DNSKEYs lookup for zone */
8442         /* result if that, if insecure, means no DNSSEC for the ZONEMD,
8443          * otherwise we have the zone DNSKEY for the DNSSEC verification. */
8444         if(env->anchors)
8445                 anchor = anchors_lookup(env->anchors, z->name, z->namelen,
8446                         z->dclass);
8447         if(anchor && anchor->numDS == 0 && anchor->numDNSKEY == 0) {
8448                 /* domain-insecure trust anchor for unsigned zones */
8449                 lock_basic_unlock(&anchor->lock);
8450                 if(only_online)
8451                         return;
8452                 dnskey = NULL;
8453                 is_insecure = 1;
8454         } else if(anchor && query_dname_compare(z->name, anchor->name) == 0) {
8455                 if(only_online) {
8456                         lock_basic_unlock(&anchor->lock);
8457                         return;
8458                 }
8459                 /* equal to trustanchor, no need for online lookups */
8460                 dnskey = zonemd_get_dnskey_from_anchor(z, env, mods, anchor,
8461                         &is_insecure, &why_bogus, &keystorage);
8462                 lock_basic_unlock(&anchor->lock);
8463                 if(!dnskey && !reason && !is_insecure) {
8464                         reason = "verify DNSKEY RRset with trust anchor failed";
8465                 }
8466         } else if(anchor) {
8467                 lock_basic_unlock(&anchor->lock);
8468                 /* perform online lookups */
8469                 if(offline)
8470                         return;
8471                 /* setup online lookups, and wait for them */
8472                 if(zonemd_lookup_dnskey(z, env)) {
8473                         /* wait for the lookup */
8474                         return;
8475                 }
8476                 reason = "could not lookup DNSKEY for chain of trust";
8477         } else {
8478                 /* the zone is not under a trust anchor */
8479                 if(only_online)
8480                         return;
8481                 dnskey = NULL;
8482                 is_insecure = 1;
8483         }
8484
8485         if(reason) {
8486                 auth_zone_zonemd_fail(z, env, reason, why_bogus, result);
8487                 return;
8488         }
8489
8490         auth_zone_verify_zonemd_with_key(z, env, mods, dnskey, is_insecure,
8491                 result, NULL);
8492         regional_free_all(env->scratch);
8493 }
8494
8495 void auth_zones_pickup_zonemd_verify(struct auth_zones* az,
8496         struct module_env* env)
8497 {
8498         struct auth_zone key;
8499         uint8_t savezname[255+1];
8500         size_t savezname_len;
8501         struct auth_zone* z;
8502         key.node.key = &key;
8503         lock_rw_rdlock(&az->lock);
8504         RBTREE_FOR(z, struct auth_zone*, &az->ztree) {
8505                 lock_rw_wrlock(&z->lock);
8506                 if(!z->zonemd_check) {
8507                         lock_rw_unlock(&z->lock);
8508                         continue;
8509                 }
8510                 key.dclass = z->dclass;
8511                 key.namelabs = z->namelabs;
8512                 if(z->namelen > sizeof(savezname)) {
8513                         lock_rw_unlock(&z->lock);
8514                         log_err("auth_zones_pickup_zonemd_verify: zone name too long");
8515                         continue;
8516                 }
8517                 savezname_len = z->namelen;
8518                 memmove(savezname, z->name, z->namelen);
8519                 lock_rw_unlock(&az->lock);
8520                 auth_zone_verify_zonemd(z, env, &env->mesh->mods, NULL, 0, 1);
8521                 lock_rw_unlock(&z->lock);
8522                 lock_rw_rdlock(&az->lock);
8523                 /* find the zone we had before, it is not deleted,
8524                  * because we have a flag for that that is processed at
8525                  * apply_cfg time */
8526                 key.namelen = savezname_len;
8527                 key.name = savezname;
8528                 z = (struct auth_zone*)rbtree_search(&az->ztree, &key);
8529                 if(!z)
8530                         break;
8531         }
8532         lock_rw_unlock(&az->lock);
8533 }