]> CyberLeo.Net >> Repos - FreeBSD/FreeBSD.git/blob - contrib/unbound/services/authzone.c
usr.bin/bc: update to version 5.3.1
[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->namelen, dname, dname_len, rr_type,
1310                         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                 }
2760         }
2761 }
2762
2763 /** find NSEC record covering the query */
2764 static struct auth_rrset*
2765 az_find_nsec_cover(struct auth_zone* z, struct auth_data** node)
2766 {
2767         uint8_t* nm = (*node)->name;
2768         size_t nmlen = (*node)->namelen;
2769         struct auth_rrset* rrset;
2770         /* find the NSEC for the smallest-or-equal node */
2771         /* if node == NULL, we did not find a smaller name.  But the zone
2772          * name is the smallest name and should have an NSEC. So there is
2773          * no NSEC to return (for a properly signed zone) */
2774         /* for empty nonterminals, the auth-data node should not exist,
2775          * and thus we don't need to go rbtree_previous here to find
2776          * a domain with an NSEC record */
2777         /* but there could be glue, and if this is node, then it has no NSEC.
2778          * Go up to find nonglue (previous) NSEC-holding nodes */
2779         while((rrset=az_domain_rrset(*node, LDNS_RR_TYPE_NSEC)) == NULL) {
2780                 if(dname_is_root(nm)) return NULL;
2781                 if(nmlen == z->namelen) return NULL;
2782                 dname_remove_label(&nm, &nmlen);
2783                 /* adjust *node for the nsec rrset to find in */
2784                 *node = az_find_name(z, nm, nmlen);
2785         }
2786         return rrset;
2787 }
2788
2789 /** Find NSEC and add for wildcard denial */
2790 static int
2791 az_nsec_wildcard_denial(struct auth_zone* z, struct regional* region,
2792         struct dns_msg* msg, uint8_t* cenm, size_t cenmlen)
2793 {
2794         struct query_info qinfo;
2795         int node_exact;
2796         struct auth_data* node;
2797         struct auth_rrset* nsec;
2798         uint8_t wc[LDNS_MAX_DOMAINLEN];
2799         if(cenmlen+2 > sizeof(wc))
2800                 return 0; /* result would be too long */
2801         wc[0] = 1; /* length of wildcard label */
2802         wc[1] = (uint8_t)'*'; /* wildcard label */
2803         memmove(wc+2, cenm, cenmlen);
2804
2805         /* we have '*.ce' in wc wildcard name buffer */
2806         /* get nsec cover for that */
2807         qinfo.qname = wc;
2808         qinfo.qname_len = cenmlen+2;
2809         qinfo.qtype = 0;
2810         qinfo.qclass = 0;
2811         az_find_domain(z, &qinfo, &node_exact, &node);
2812         if((nsec=az_find_nsec_cover(z, &node)) != NULL) {
2813                 if(!msg_add_rrset_ns(z, region, msg, node, nsec)) return 0;
2814         }
2815         return 1;
2816 }
2817
2818 /** Find the NSEC3PARAM rrset (if any) and if true you have the parameters */
2819 static int
2820 az_nsec3_param(struct auth_zone* z, int* algo, size_t* iter, uint8_t** salt,
2821         size_t* saltlen)
2822 {
2823         struct auth_data* apex;
2824         struct auth_rrset* param;
2825         size_t i;
2826         apex = az_find_name(z, z->name, z->namelen);
2827         if(!apex) return 0;
2828         param = az_domain_rrset(apex, LDNS_RR_TYPE_NSEC3PARAM);
2829         if(!param || param->data->count==0)
2830                 return 0; /* no RRset or no RRs in rrset */
2831         /* find out which NSEC3PARAM RR has supported parameters */
2832         /* skip unknown flags (dynamic signer is recalculating nsec3 chain) */
2833         for(i=0; i<param->data->count; i++) {
2834                 uint8_t* rdata = param->data->rr_data[i]+2;
2835                 size_t rdatalen = param->data->rr_len[i];
2836                 if(rdatalen < 2+5)
2837                         continue; /* too short */
2838                 if(!nsec3_hash_algo_size_supported((int)(rdata[0])))
2839                         continue; /* unsupported algo */
2840                 if(rdatalen < (size_t)(2+5+(size_t)rdata[4]))
2841                         continue; /* salt missing */
2842                 if((rdata[1]&NSEC3_UNKNOWN_FLAGS)!=0)
2843                         continue; /* unknown flags */
2844                 *algo = (int)(rdata[0]);
2845                 *iter = sldns_read_uint16(rdata+2);
2846                 *saltlen = rdata[4];
2847                 if(*saltlen == 0)
2848                         *salt = NULL;
2849                 else    *salt = rdata+5;
2850                 return 1;
2851         }
2852         /* no supported params */
2853         return 0;
2854 }
2855
2856 /** Hash a name with nsec3param into buffer, it has zone name appended.
2857  * return length of hash */
2858 static size_t
2859 az_nsec3_hash(uint8_t* buf, size_t buflen, uint8_t* nm, size_t nmlen,
2860         int algo, size_t iter, uint8_t* salt, size_t saltlen)
2861 {
2862         size_t hlen = nsec3_hash_algo_size_supported(algo);
2863         /* buffer has domain name, nsec3hash, and 256 is for max saltlen
2864          * (salt has 0-255 length) */
2865         unsigned char p[LDNS_MAX_DOMAINLEN+1+N3HASHBUFLEN+256];
2866         size_t i;
2867         if(nmlen+saltlen > sizeof(p) || hlen+saltlen > sizeof(p))
2868                 return 0;
2869         if(hlen > buflen)
2870                 return 0; /* somehow too large for destination buffer */
2871         /* hashfunc(name, salt) */
2872         memmove(p, nm, nmlen);
2873         query_dname_tolower(p);
2874         if(salt && saltlen > 0)
2875                 memmove(p+nmlen, salt, saltlen);
2876         (void)secalgo_nsec3_hash(algo, p, nmlen+saltlen, (unsigned char*)buf);
2877         for(i=0; i<iter; i++) {
2878                 /* hashfunc(hash, salt) */
2879                 memmove(p, buf, hlen);
2880                 if(salt && saltlen > 0)
2881                         memmove(p+hlen, salt, saltlen);
2882                 (void)secalgo_nsec3_hash(algo, p, hlen+saltlen,
2883                         (unsigned char*)buf);
2884         }
2885         return hlen;
2886 }
2887
2888 /** Hash name and return b32encoded hashname for lookup, zone name appended */
2889 static int
2890 az_nsec3_hashname(struct auth_zone* z, uint8_t* hashname, size_t* hashnmlen,
2891         uint8_t* nm, size_t nmlen, int algo, size_t iter, uint8_t* salt,
2892         size_t saltlen)
2893 {
2894         uint8_t hash[N3HASHBUFLEN];
2895         size_t hlen;
2896         int ret;
2897         hlen = az_nsec3_hash(hash, sizeof(hash), nm, nmlen, algo, iter,
2898                 salt, saltlen);
2899         if(!hlen) return 0;
2900         /* b32 encode */
2901         if(*hashnmlen < hlen*2+1+z->namelen) /* approx b32 as hexb16 */
2902                 return 0;
2903         ret = sldns_b32_ntop_extended_hex(hash, hlen, (char*)(hashname+1),
2904                 (*hashnmlen)-1);
2905         if(ret<1)
2906                 return 0;
2907         hashname[0] = (uint8_t)ret;
2908         ret++;
2909         if((*hashnmlen) - ret < z->namelen)
2910                 return 0;
2911         memmove(hashname+ret, z->name, z->namelen);
2912         *hashnmlen = z->namelen+(size_t)ret;
2913         return 1;
2914 }
2915
2916 /** Find the datanode that covers the nsec3hash-name */
2917 static struct auth_data*
2918 az_nsec3_findnode(struct auth_zone* z, uint8_t* hashnm, size_t hashnmlen)
2919 {
2920         struct query_info qinfo;
2921         struct auth_data* node;
2922         int node_exact;
2923         qinfo.qclass = 0;
2924         qinfo.qtype = 0;
2925         qinfo.qname = hashnm;
2926         qinfo.qname_len = hashnmlen;
2927         /* because canonical ordering and b32 nsec3 ordering are the same.
2928          * this is a good lookup to find the nsec3 name. */
2929         az_find_domain(z, &qinfo, &node_exact, &node);
2930         /* but we may have to skip non-nsec3 nodes */
2931         /* this may be a lot, the way to speed that up is to have a
2932          * separate nsec3 tree with nsec3 nodes */
2933         while(node && (rbnode_type*)node != RBTREE_NULL &&
2934                 !az_domain_rrset(node, LDNS_RR_TYPE_NSEC3)) {
2935                 node = (struct auth_data*)rbtree_previous(&node->node);
2936         }
2937         if((rbnode_type*)node == RBTREE_NULL)
2938                 node = NULL;
2939         return node;
2940 }
2941
2942 /** Find cover for hashed(nm, nmlen) (or NULL) */
2943 static struct auth_data*
2944 az_nsec3_find_cover(struct auth_zone* z, uint8_t* nm, size_t nmlen,
2945         int algo, size_t iter, uint8_t* salt, size_t saltlen)
2946 {
2947         struct auth_data* node;
2948         uint8_t hname[LDNS_MAX_DOMAINLEN];
2949         size_t hlen = sizeof(hname);
2950         if(!az_nsec3_hashname(z, hname, &hlen, nm, nmlen, algo, iter,
2951                 salt, saltlen))
2952                 return NULL;
2953         node = az_nsec3_findnode(z, hname, hlen);
2954         if(node)
2955                 return node;
2956         /* we did not find any, perhaps because the NSEC3 hash is before
2957          * the first hash, we have to find the 'last hash' in the zone */
2958         node = (struct auth_data*)rbtree_last(&z->data);
2959         while(node && (rbnode_type*)node != RBTREE_NULL &&
2960                 !az_domain_rrset(node, LDNS_RR_TYPE_NSEC3)) {
2961                 node = (struct auth_data*)rbtree_previous(&node->node);
2962         }
2963         if((rbnode_type*)node == RBTREE_NULL)
2964                 node = NULL;
2965         return node;
2966 }
2967
2968 /** Find exact match for hashed(nm, nmlen) NSEC3 record or NULL */
2969 static struct auth_data*
2970 az_nsec3_find_exact(struct auth_zone* z, uint8_t* nm, size_t nmlen,
2971         int algo, size_t iter, uint8_t* salt, size_t saltlen)
2972 {
2973         struct auth_data* node;
2974         uint8_t hname[LDNS_MAX_DOMAINLEN];
2975         size_t hlen = sizeof(hname);
2976         if(!az_nsec3_hashname(z, hname, &hlen, nm, nmlen, algo, iter,
2977                 salt, saltlen))
2978                 return NULL;
2979         node = az_find_name(z, hname, hlen);
2980         if(az_domain_rrset(node, LDNS_RR_TYPE_NSEC3))
2981                 return node;
2982         return NULL;
2983 }
2984
2985 /** Return nextcloser name (as a ref into the qname).  This is one label
2986  * more than the cenm (cename must be a suffix of qname) */
2987 static void
2988 az_nsec3_get_nextcloser(uint8_t* cenm, uint8_t* qname, size_t qname_len,
2989         uint8_t** nx, size_t* nxlen)
2990 {
2991         int celabs = dname_count_labels(cenm);
2992         int qlabs = dname_count_labels(qname);
2993         int strip = qlabs - celabs -1;
2994         log_assert(dname_strict_subdomain(qname, qlabs, cenm, celabs));
2995         *nx = qname;
2996         *nxlen = qname_len;
2997         if(strip>0)
2998                 dname_remove_labels(nx, nxlen, strip);
2999 }
3000
3001 /** Find the closest encloser that has exact NSEC3.
3002  * updated cenm to the new name. If it went up no-exact-ce is true. */
3003 static struct auth_data*
3004 az_nsec3_find_ce(struct auth_zone* z, uint8_t** cenm, size_t* cenmlen,
3005         int* no_exact_ce, int algo, size_t iter, uint8_t* salt, size_t saltlen)
3006 {
3007         struct auth_data* node;
3008         while((node = az_nsec3_find_exact(z, *cenm, *cenmlen,
3009                 algo, iter, salt, saltlen)) == NULL) {
3010                 if(*cenmlen == z->namelen) {
3011                         /* next step up would take us out of the zone. fail */
3012                         return NULL;
3013                 }
3014                 *no_exact_ce = 1;
3015                 dname_remove_label(cenm, cenmlen);
3016         }
3017         return node;
3018 }
3019
3020 /* Insert NSEC3 record in authority section, if NULL does nothing */
3021 static int
3022 az_nsec3_insert(struct auth_zone* z, struct regional* region,
3023         struct dns_msg* msg, struct auth_data* node)
3024 {
3025         struct auth_rrset* nsec3;
3026         if(!node) return 1; /* no node, skip this */
3027         nsec3 = az_domain_rrset(node, LDNS_RR_TYPE_NSEC3);
3028         if(!nsec3) return 1; /* if no nsec3 RR, skip it */
3029         if(!msg_add_rrset_ns(z, region, msg, node, nsec3)) return 0;
3030         return 1;
3031 }
3032
3033 /** add NSEC3 records to the zone for the nsec3 proof.
3034  * Specify with the flags with parts of the proof are required.
3035  * the ce is the exact matching name (for notype) but also delegation points.
3036  * qname is the one where the nextcloser name can be derived from.
3037  * If NSEC3 is not properly there (in the zone) nothing is added.
3038  * always enabled: include nsec3 proving about the Closest Encloser.
3039  *      that is an exact match that should exist for it.
3040  *      If that does not exist, a higher exact match + nxproof is enabled
3041  *      (for some sort of opt-out empty nonterminal cases).
3042  * nodataproof: search for exact match and include that instead.
3043  * ceproof: include ce proof NSEC3 (omitted for wildcard replies).
3044  * nxproof: include denial of the qname.
3045  * wcproof: include denial of wildcard (wildcard.ce).
3046  */
3047 static int
3048 az_add_nsec3_proof(struct auth_zone* z, struct regional* region,
3049         struct dns_msg* msg, uint8_t* cenm, size_t cenmlen, uint8_t* qname,
3050         size_t qname_len, int nodataproof, int ceproof, int nxproof,
3051         int wcproof)
3052 {
3053         int algo;
3054         size_t iter, saltlen;
3055         uint8_t* salt;
3056         int no_exact_ce = 0;
3057         struct auth_data* node;
3058
3059         /* find parameters of nsec3 proof */
3060         if(!az_nsec3_param(z, &algo, &iter, &salt, &saltlen))
3061                 return 1; /* no nsec3 */
3062         if(nodataproof) {
3063                 /* see if the node has a hash of itself for the nodata
3064                  * proof nsec3, this has to be an exact match nsec3. */
3065                 struct auth_data* match;
3066                 match = az_nsec3_find_exact(z, qname, qname_len, algo,
3067                         iter, salt, saltlen);
3068                 if(match) {
3069                         if(!az_nsec3_insert(z, region, msg, match))
3070                                 return 0;
3071                         /* only nodata NSEC3 needed, no CE or others. */
3072                         return 1;
3073                 }
3074         }
3075         /* find ce that has an NSEC3 */
3076         if(ceproof) {
3077                 node = az_nsec3_find_ce(z, &cenm, &cenmlen, &no_exact_ce,
3078                         algo, iter, salt, saltlen);
3079                 if(no_exact_ce) nxproof = 1;
3080                 if(!az_nsec3_insert(z, region, msg, node))
3081                         return 0;
3082         }
3083
3084         if(nxproof) {
3085                 uint8_t* nx;
3086                 size_t nxlen;
3087                 /* create nextcloser domain name */
3088                 az_nsec3_get_nextcloser(cenm, qname, qname_len, &nx, &nxlen);
3089                 /* find nsec3 that matches or covers it */
3090                 node = az_nsec3_find_cover(z, nx, nxlen, algo, iter, salt,
3091                         saltlen);
3092                 if(!az_nsec3_insert(z, region, msg, node))
3093                         return 0;
3094         }
3095         if(wcproof) {
3096                 /* create wildcard name *.ce */
3097                 uint8_t wc[LDNS_MAX_DOMAINLEN];
3098                 size_t wclen;
3099                 if(cenmlen+2 > sizeof(wc))
3100                         return 0; /* result would be too long */
3101                 wc[0] = 1; /* length of wildcard label */
3102                 wc[1] = (uint8_t)'*'; /* wildcard label */
3103                 memmove(wc+2, cenm, cenmlen);
3104                 wclen = cenmlen+2;
3105                 /* find nsec3 that matches or covers it */
3106                 node = az_nsec3_find_cover(z, wc, wclen, algo, iter, salt,
3107                         saltlen);
3108                 if(!az_nsec3_insert(z, region, msg, node))
3109                         return 0;
3110         }
3111         return 1;
3112 }
3113
3114 /** generate answer for positive answer */
3115 static int
3116 az_generate_positive_answer(struct auth_zone* z, struct regional* region,
3117         struct dns_msg* msg, struct auth_data* node, struct auth_rrset* rrset)
3118 {
3119         if(!msg_add_rrset_an(z, region, msg, node, rrset)) return 0;
3120         /* see if we want additional rrs */
3121         if(rrset->type == LDNS_RR_TYPE_MX) {
3122                 if(!az_add_additionals_from(z, region, msg, rrset, 2))
3123                         return 0;
3124         } else if(rrset->type == LDNS_RR_TYPE_SRV) {
3125                 if(!az_add_additionals_from(z, region, msg, rrset, 6))
3126                         return 0;
3127         } else if(rrset->type == LDNS_RR_TYPE_NS) {
3128                 if(!az_add_additionals_from(z, region, msg, rrset, 0))
3129                         return 0;
3130         }
3131         return 1;
3132 }
3133
3134 /** generate answer for type ANY answer */
3135 static int
3136 az_generate_any_answer(struct auth_zone* z, struct regional* region,
3137         struct dns_msg* msg, struct auth_data* node)
3138 {
3139         struct auth_rrset* rrset;
3140         int added = 0;
3141         /* add a couple (at least one) RRs */
3142         if((rrset=az_domain_rrset(node, LDNS_RR_TYPE_SOA)) != NULL) {
3143                 if(!msg_add_rrset_an(z, region, msg, node, rrset)) return 0;
3144                 added++;
3145         }
3146         if((rrset=az_domain_rrset(node, LDNS_RR_TYPE_MX)) != NULL) {
3147                 if(!msg_add_rrset_an(z, region, msg, node, rrset)) return 0;
3148                 added++;
3149         }
3150         if((rrset=az_domain_rrset(node, LDNS_RR_TYPE_A)) != NULL) {
3151                 if(!msg_add_rrset_an(z, region, msg, node, rrset)) return 0;
3152                 added++;
3153         }
3154         if((rrset=az_domain_rrset(node, LDNS_RR_TYPE_AAAA)) != NULL) {
3155                 if(!msg_add_rrset_an(z, region, msg, node, rrset)) return 0;
3156                 added++;
3157         }
3158         if(added == 0 && node && node->rrsets) {
3159                 if(!msg_add_rrset_an(z, region, msg, node,
3160                         node->rrsets)) return 0;
3161         }
3162         return 1;
3163 }
3164
3165 /** follow cname chain and add more data to the answer section */
3166 static int
3167 follow_cname_chain(struct auth_zone* z, uint16_t qtype,
3168         struct regional* region, struct dns_msg* msg,
3169         struct packed_rrset_data* d)
3170 {
3171         int maxchain = 0;
3172         /* see if we can add the target of the CNAME into the answer */
3173         while(maxchain++ < MAX_CNAME_CHAIN) {
3174                 struct auth_data* node;
3175                 struct auth_rrset* rrset;
3176                 size_t clen;
3177                 /* d has cname rdata */
3178                 if(d->count == 0) break; /* no CNAME */
3179                 if(d->rr_len[0] < 2+1) break; /* too small */
3180                 if((clen=dname_valid(d->rr_data[0]+2, d->rr_len[0]-2))==0)
3181                         break; /* malformed */
3182                 if(!dname_subdomain_c(d->rr_data[0]+2, z->name))
3183                         break; /* target out of zone */
3184                 if((node = az_find_name(z, d->rr_data[0]+2, clen))==NULL)
3185                         break; /* no such target name */
3186                 if((rrset=az_domain_rrset(node, qtype))!=NULL) {
3187                         /* done we found the target */
3188                         if(!msg_add_rrset_an(z, region, msg, node, rrset))
3189                                 return 0;
3190                         break;
3191                 }
3192                 if((rrset=az_domain_rrset(node, LDNS_RR_TYPE_CNAME))==NULL)
3193                         break; /* no further CNAME chain, notype */
3194                 if(!msg_add_rrset_an(z, region, msg, node, rrset)) return 0;
3195                 d = rrset->data;
3196         }
3197         return 1;
3198 }
3199
3200 /** generate answer for cname answer */
3201 static int
3202 az_generate_cname_answer(struct auth_zone* z, struct query_info* qinfo,
3203         struct regional* region, struct dns_msg* msg,
3204         struct auth_data* node, struct auth_rrset* rrset)
3205 {
3206         if(!msg_add_rrset_an(z, region, msg, node, rrset)) return 0;
3207         if(!rrset) return 1;
3208         if(!follow_cname_chain(z, qinfo->qtype, region, msg, rrset->data))
3209                 return 0;
3210         return 1;
3211 }
3212
3213 /** generate answer for notype answer */
3214 static int
3215 az_generate_notype_answer(struct auth_zone* z, struct regional* region,
3216         struct dns_msg* msg, struct auth_data* node)
3217 {
3218         struct auth_rrset* rrset;
3219         if(!az_add_negative_soa(z, region, msg)) return 0;
3220         /* DNSSEC denial NSEC */
3221         if((rrset=az_domain_rrset(node, LDNS_RR_TYPE_NSEC))!=NULL) {
3222                 if(!msg_add_rrset_ns(z, region, msg, node, rrset)) return 0;
3223         } else if(node) {
3224                 /* DNSSEC denial NSEC3 */
3225                 if(!az_add_nsec3_proof(z, region, msg, node->name,
3226                         node->namelen, msg->qinfo.qname,
3227                         msg->qinfo.qname_len, 1, 1, 0, 0))
3228                         return 0;
3229         }
3230         return 1;
3231 }
3232
3233 /** generate answer for referral answer */
3234 static int
3235 az_generate_referral_answer(struct auth_zone* z, struct regional* region,
3236         struct dns_msg* msg, struct auth_data* ce, struct auth_rrset* rrset)
3237 {
3238         struct auth_rrset* ds, *nsec;
3239         /* turn off AA flag, referral is nonAA because it leaves the zone */
3240         log_assert(ce);
3241         msg->rep->flags &= ~BIT_AA;
3242         if(!msg_add_rrset_ns(z, region, msg, ce, rrset)) return 0;
3243         /* add DS or deny it */
3244         if((ds=az_domain_rrset(ce, LDNS_RR_TYPE_DS))!=NULL) {
3245                 if(!msg_add_rrset_ns(z, region, msg, ce, ds)) return 0;
3246         } else {
3247                 /* deny the DS */
3248                 if((nsec=az_domain_rrset(ce, LDNS_RR_TYPE_NSEC))!=NULL) {
3249                         if(!msg_add_rrset_ns(z, region, msg, ce, nsec))
3250                                 return 0;
3251                 } else {
3252                         if(!az_add_nsec3_proof(z, region, msg, ce->name,
3253                                 ce->namelen, msg->qinfo.qname,
3254                                 msg->qinfo.qname_len, 1, 1, 0, 0))
3255                                 return 0;
3256                 }
3257         }
3258         /* add additional rrs for type NS */
3259         if(!az_add_additionals_from(z, region, msg, rrset, 0)) return 0;
3260         return 1;
3261 }
3262
3263 /** generate answer for DNAME answer */
3264 static int
3265 az_generate_dname_answer(struct auth_zone* z, struct query_info* qinfo,
3266         struct regional* region, struct dns_msg* msg, struct auth_data* ce,
3267         struct auth_rrset* rrset)
3268 {
3269         log_assert(ce);
3270         /* add the DNAME and then a CNAME */
3271         if(!msg_add_rrset_an(z, region, msg, ce, rrset)) return 0;
3272         if(!add_synth_cname(z, qinfo->qname, qinfo->qname_len, region,
3273                 msg, ce, rrset)) return 0;
3274         if(FLAGS_GET_RCODE(msg->rep->flags) == LDNS_RCODE_YXDOMAIN)
3275                 return 1;
3276         if(msg->rep->rrset_count == 0 ||
3277                 !msg->rep->rrsets[msg->rep->rrset_count-1])
3278                 return 0;
3279         if(!follow_cname_chain(z, qinfo->qtype, region, msg, 
3280                 (struct packed_rrset_data*)msg->rep->rrsets[
3281                 msg->rep->rrset_count-1]->entry.data))
3282                 return 0;
3283         return 1;
3284 }
3285
3286 /** generate answer for wildcard answer */
3287 static int
3288 az_generate_wildcard_answer(struct auth_zone* z, struct query_info* qinfo,
3289         struct regional* region, struct dns_msg* msg, struct auth_data* ce,
3290         struct auth_data* wildcard, struct auth_data* node)
3291 {
3292         struct auth_rrset* rrset, *nsec;
3293         int insert_ce = 0;
3294         if((rrset=az_domain_rrset(wildcard, qinfo->qtype)) != NULL) {
3295                 /* wildcard has type, add it */
3296                 if(!msg_add_rrset_an(z, region, msg, wildcard, rrset))
3297                         return 0;
3298                 az_change_dnames(msg, wildcard->name, msg->qinfo.qname,
3299                         msg->qinfo.qname_len, 1);
3300         } else if((rrset=az_domain_rrset(wildcard, LDNS_RR_TYPE_CNAME))!=NULL) {
3301                 /* wildcard has cname instead, do that */
3302                 if(!msg_add_rrset_an(z, region, msg, wildcard, rrset))
3303                         return 0;
3304                 az_change_dnames(msg, wildcard->name, msg->qinfo.qname,
3305                         msg->qinfo.qname_len, 1);
3306                 if(!follow_cname_chain(z, qinfo->qtype, region, msg,
3307                         rrset->data))
3308                         return 0;
3309         } else if(qinfo->qtype == LDNS_RR_TYPE_ANY && wildcard->rrsets) {
3310                 /* add ANY rrsets from wildcard node */
3311                 if(!az_generate_any_answer(z, region, msg, wildcard))
3312                         return 0;
3313                 az_change_dnames(msg, wildcard->name, msg->qinfo.qname,
3314                         msg->qinfo.qname_len, 1);
3315         } else {
3316                 /* wildcard has nodata, notype answer */
3317                 /* call other notype routine for dnssec notype denials */
3318                 if(!az_generate_notype_answer(z, region, msg, wildcard))
3319                         return 0;
3320                 /* because the notype, there is no positive data with an
3321                  * RRSIG that indicates the wildcard position.  Thus the
3322                  * wildcard qname denial needs to have a CE nsec3. */
3323                 insert_ce = 1;
3324         }
3325
3326         /* ce and node for dnssec denial of wildcard original name */
3327         if((nsec=az_find_nsec_cover(z, &node)) != NULL) {
3328                 if(!msg_add_rrset_ns(z, region, msg, node, nsec)) return 0;
3329         } else if(ce) {
3330                 uint8_t* wildup = wildcard->name;
3331                 size_t wilduplen= wildcard->namelen;
3332                 dname_remove_label(&wildup, &wilduplen);
3333                 if(!az_add_nsec3_proof(z, region, msg, wildup,
3334                         wilduplen, msg->qinfo.qname,
3335                         msg->qinfo.qname_len, 0, insert_ce, 1, 0))
3336                         return 0;
3337         }
3338
3339         /* fixup name of wildcard from *.zone to qname, use already allocated
3340          * pointer to msg qname */
3341         az_change_dnames(msg, wildcard->name, msg->qinfo.qname,
3342                 msg->qinfo.qname_len, 0);
3343         return 1;
3344 }
3345
3346 /** generate answer for nxdomain answer */
3347 static int
3348 az_generate_nxdomain_answer(struct auth_zone* z, struct regional* region,
3349         struct dns_msg* msg, struct auth_data* ce, struct auth_data* node)
3350 {
3351         struct auth_rrset* nsec;
3352         msg->rep->flags |= LDNS_RCODE_NXDOMAIN;
3353         if(!az_add_negative_soa(z, region, msg)) return 0;
3354         if((nsec=az_find_nsec_cover(z, &node)) != NULL) {
3355                 if(!msg_add_rrset_ns(z, region, msg, node, nsec)) return 0;
3356                 if(ce && !az_nsec_wildcard_denial(z, region, msg, ce->name,
3357                         ce->namelen)) return 0;
3358         } else if(ce) {
3359                 if(!az_add_nsec3_proof(z, region, msg, ce->name,
3360                         ce->namelen, msg->qinfo.qname,
3361                         msg->qinfo.qname_len, 0, 1, 1, 1))
3362                         return 0;
3363         }
3364         return 1;
3365 }
3366
3367 /** Create answers when an exact match exists for the domain name */
3368 static int
3369 az_generate_answer_with_node(struct auth_zone* z, struct query_info* qinfo,
3370         struct regional* region, struct dns_msg* msg, struct auth_data* node)
3371 {
3372         struct auth_rrset* rrset;
3373         /* positive answer, rrset we are looking for exists */
3374         if((rrset=az_domain_rrset(node, qinfo->qtype)) != NULL) {
3375                 return az_generate_positive_answer(z, region, msg, node, rrset);
3376         }
3377         /* CNAME? */
3378         if((rrset=az_domain_rrset(node, LDNS_RR_TYPE_CNAME)) != NULL) {
3379                 return az_generate_cname_answer(z, qinfo, region, msg,
3380                         node, rrset);
3381         }
3382         /* type ANY ? */
3383         if(qinfo->qtype == LDNS_RR_TYPE_ANY) {
3384                 return az_generate_any_answer(z, region, msg, node);
3385         }
3386         /* NOERROR/NODATA (no such type at domain name) */
3387         return az_generate_notype_answer(z, region, msg, node);
3388 }
3389
3390 /** Generate answer without an existing-node that we can use.
3391  * So it'll be a referral, DNAME or nxdomain */
3392 static int
3393 az_generate_answer_nonexistnode(struct auth_zone* z, struct query_info* qinfo,
3394         struct regional* region, struct dns_msg* msg, struct auth_data* ce,
3395         struct auth_rrset* rrset, struct auth_data* node)
3396 {
3397         struct auth_data* wildcard;
3398
3399         /* we do not have an exact matching name (that exists) */
3400         /* see if we have a NS or DNAME in the ce */
3401         if(ce && rrset && rrset->type == LDNS_RR_TYPE_NS) {
3402                 return az_generate_referral_answer(z, region, msg, ce, rrset);
3403         }
3404         if(ce && rrset && rrset->type == LDNS_RR_TYPE_DNAME) {
3405                 return az_generate_dname_answer(z, qinfo, region, msg, ce,
3406                         rrset);
3407         }
3408         /* if there is an empty nonterminal, wildcard and nxdomain don't
3409          * happen, it is a notype answer */
3410         if(az_empty_nonterminal(z, qinfo, node)) {
3411                 return az_generate_notype_answer(z, region, msg, node);
3412         }
3413         /* see if we have a wildcard under the ce */
3414         if((wildcard=az_find_wildcard(z, qinfo, ce)) != NULL) {
3415                 return az_generate_wildcard_answer(z, qinfo, region, msg,
3416                         ce, wildcard, node);
3417         }
3418         /* generate nxdomain answer */
3419         return az_generate_nxdomain_answer(z, region, msg, ce, node);
3420 }
3421
3422 /** Lookup answer in a zone. */
3423 static int
3424 auth_zone_generate_answer(struct auth_zone* z, struct query_info* qinfo,
3425         struct regional* region, struct dns_msg** msg, int* fallback)
3426 {
3427         struct auth_data* node, *ce;
3428         struct auth_rrset* rrset;
3429         int node_exact, node_exists;
3430         /* does the zone want fallback in case of failure? */
3431         *fallback = z->fallback_enabled;
3432         if(!(*msg=msg_create(region, qinfo))) return 0;
3433
3434         /* lookup if there is a matching domain name for the query */
3435         az_find_domain(z, qinfo, &node_exact, &node);
3436
3437         /* see if node exists for generating answers from (i.e. not glue and
3438          * obscured by NS or DNAME or NSEC3-only), and also return the
3439          * closest-encloser from that, closest node that should be used
3440          * to generate answers from that is above the query */
3441         node_exists = az_find_ce(z, qinfo, node, node_exact, &ce, &rrset);
3442
3443         if(verbosity >= VERB_ALGO) {
3444                 char zname[256], qname[256], nname[256], cename[256],
3445                         tpstr[32], rrstr[32];
3446                 sldns_wire2str_dname_buf(qinfo->qname, qinfo->qname_len, qname,
3447                         sizeof(qname));
3448                 sldns_wire2str_type_buf(qinfo->qtype, tpstr, sizeof(tpstr));
3449                 sldns_wire2str_dname_buf(z->name, z->namelen, zname,
3450                         sizeof(zname));
3451                 if(node)
3452                         sldns_wire2str_dname_buf(node->name, node->namelen,
3453                                 nname, sizeof(nname));
3454                 else    snprintf(nname, sizeof(nname), "NULL");
3455                 if(ce)
3456                         sldns_wire2str_dname_buf(ce->name, ce->namelen,
3457                                 cename, sizeof(cename));
3458                 else    snprintf(cename, sizeof(cename), "NULL");
3459                 if(rrset) sldns_wire2str_type_buf(rrset->type, rrstr,
3460                         sizeof(rrstr));
3461                 else    snprintf(rrstr, sizeof(rrstr), "NULL");
3462                 log_info("auth_zone %s query %s %s, domain %s %s %s, "
3463                         "ce %s, rrset %s", zname, qname, tpstr, nname,
3464                         (node_exact?"exact":"notexact"),
3465                         (node_exists?"exist":"notexist"), cename, rrstr);
3466         }
3467
3468         if(node_exists) {
3469                 /* the node is fine, generate answer from node */
3470                 return az_generate_answer_with_node(z, qinfo, region, *msg,
3471                         node);
3472         }
3473         return az_generate_answer_nonexistnode(z, qinfo, region, *msg,
3474                 ce, rrset, node);
3475 }
3476
3477 int auth_zones_lookup(struct auth_zones* az, struct query_info* qinfo,
3478         struct regional* region, struct dns_msg** msg, int* fallback,
3479         uint8_t* dp_nm, size_t dp_nmlen)
3480 {
3481         int r;
3482         struct auth_zone* z;
3483         /* find the zone that should contain the answer. */
3484         lock_rw_rdlock(&az->lock);
3485         z = auth_zone_find(az, dp_nm, dp_nmlen, qinfo->qclass);
3486         if(!z) {
3487                 lock_rw_unlock(&az->lock);
3488                 /* no auth zone, fallback to internet */
3489                 *fallback = 1;
3490                 return 0;
3491         }
3492         lock_rw_rdlock(&z->lock);
3493         lock_rw_unlock(&az->lock);
3494
3495         /* if not for upstream queries, fallback */
3496         if(!z->for_upstream) {
3497                 lock_rw_unlock(&z->lock);
3498                 *fallback = 1;
3499                 return 0;
3500         }
3501         if(z->zone_expired) {
3502                 *fallback = z->fallback_enabled;
3503                 lock_rw_unlock(&z->lock);
3504                 return 0;
3505         }
3506         /* see what answer that zone would generate */
3507         r = auth_zone_generate_answer(z, qinfo, region, msg, fallback);
3508         lock_rw_unlock(&z->lock);
3509         return r;
3510 }
3511
3512 /** encode auth answer */
3513 static void
3514 auth_answer_encode(struct query_info* qinfo, struct module_env* env,
3515         struct edns_data* edns, struct comm_reply* repinfo, sldns_buffer* buf,
3516         struct regional* temp, struct dns_msg* msg)
3517 {
3518         uint16_t udpsize;
3519         udpsize = edns->udp_size;
3520         edns->edns_version = EDNS_ADVERTISED_VERSION;
3521         edns->udp_size = EDNS_ADVERTISED_SIZE;
3522         edns->ext_rcode = 0;
3523         edns->bits &= EDNS_DO;
3524
3525         if(!inplace_cb_reply_local_call(env, qinfo, NULL, msg->rep,
3526                 (int)FLAGS_GET_RCODE(msg->rep->flags), edns, repinfo, temp, env->now_tv)
3527                 || !reply_info_answer_encode(qinfo, msg->rep,
3528                 *(uint16_t*)sldns_buffer_begin(buf),
3529                 sldns_buffer_read_u16_at(buf, 2),
3530                 buf, 0, 0, temp, udpsize, edns,
3531                 (int)(edns->bits&EDNS_DO), 0)) {
3532                 error_encode(buf, (LDNS_RCODE_SERVFAIL|BIT_AA), qinfo,
3533                         *(uint16_t*)sldns_buffer_begin(buf),
3534                         sldns_buffer_read_u16_at(buf, 2), edns);
3535         }
3536 }
3537
3538 /** encode auth error answer */
3539 static void
3540 auth_error_encode(struct query_info* qinfo, struct module_env* env,
3541         struct edns_data* edns, struct comm_reply* repinfo, sldns_buffer* buf,
3542         struct regional* temp, int rcode)
3543 {
3544         edns->edns_version = EDNS_ADVERTISED_VERSION;
3545         edns->udp_size = EDNS_ADVERTISED_SIZE;
3546         edns->ext_rcode = 0;
3547         edns->bits &= EDNS_DO;
3548
3549         if(!inplace_cb_reply_local_call(env, qinfo, NULL, NULL,
3550                 rcode, edns, repinfo, temp, env->now_tv))
3551                 edns->opt_list_inplace_cb_out = NULL;
3552         error_encode(buf, rcode|BIT_AA, qinfo,
3553                 *(uint16_t*)sldns_buffer_begin(buf),
3554                 sldns_buffer_read_u16_at(buf, 2), edns);
3555 }
3556
3557 int auth_zones_answer(struct auth_zones* az, struct module_env* env,
3558         struct query_info* qinfo, struct edns_data* edns,
3559         struct comm_reply* repinfo, struct sldns_buffer* buf, struct regional* temp)
3560 {
3561         struct dns_msg* msg = NULL;
3562         struct auth_zone* z;
3563         int r;
3564         int fallback = 0;
3565
3566         lock_rw_rdlock(&az->lock);
3567         if(!az->have_downstream) {
3568                 /* no downstream auth zones */
3569                 lock_rw_unlock(&az->lock);
3570                 return 0;
3571         }
3572         if(qinfo->qtype == LDNS_RR_TYPE_DS) {
3573                 uint8_t* delname = qinfo->qname;
3574                 size_t delnamelen = qinfo->qname_len;
3575                 dname_remove_label(&delname, &delnamelen);
3576                 z = auth_zones_find_zone(az, delname, delnamelen,
3577                         qinfo->qclass);
3578         } else {
3579                 z = auth_zones_find_zone(az, qinfo->qname, qinfo->qname_len,
3580                         qinfo->qclass);
3581         }
3582         if(!z) {
3583                 /* no zone above it */
3584                 lock_rw_unlock(&az->lock);
3585                 return 0;
3586         }
3587         lock_rw_rdlock(&z->lock);
3588         lock_rw_unlock(&az->lock);
3589         if(!z->for_downstream) {
3590                 lock_rw_unlock(&z->lock);
3591                 return 0;
3592         }
3593         if(z->zone_expired) {
3594                 if(z->fallback_enabled) {
3595                         lock_rw_unlock(&z->lock);
3596                         return 0;
3597                 }
3598                 lock_rw_unlock(&z->lock);
3599                 lock_rw_wrlock(&az->lock);
3600                 az->num_query_down++;
3601                 lock_rw_unlock(&az->lock);
3602                 auth_error_encode(qinfo, env, edns, repinfo, buf, temp,
3603                         LDNS_RCODE_SERVFAIL);
3604                 return 1;
3605         }
3606
3607         /* answer it from zone z */
3608         r = auth_zone_generate_answer(z, qinfo, temp, &msg, &fallback);
3609         lock_rw_unlock(&z->lock);
3610         if(!r && fallback) {
3611                 /* fallback to regular answering (recursive) */
3612                 return 0;
3613         }
3614         lock_rw_wrlock(&az->lock);
3615         az->num_query_down++;
3616         lock_rw_unlock(&az->lock);
3617
3618         /* encode answer */
3619         if(!r)
3620                 auth_error_encode(qinfo, env, edns, repinfo, buf, temp,
3621                         LDNS_RCODE_SERVFAIL);
3622         else    auth_answer_encode(qinfo, env, edns, repinfo, buf, temp, msg);
3623
3624         return 1;
3625 }
3626
3627 int auth_zones_can_fallback(struct auth_zones* az, uint8_t* nm, size_t nmlen,
3628         uint16_t dclass)
3629 {
3630         int r;
3631         struct auth_zone* z;
3632         lock_rw_rdlock(&az->lock);
3633         z = auth_zone_find(az, nm, nmlen, dclass);
3634         if(!z) {
3635                 lock_rw_unlock(&az->lock);
3636                 /* no such auth zone, fallback */
3637                 return 1;
3638         }
3639         lock_rw_rdlock(&z->lock);
3640         lock_rw_unlock(&az->lock);
3641         r = z->fallback_enabled || (!z->for_upstream);
3642         lock_rw_unlock(&z->lock);
3643         return r;
3644 }
3645
3646 int
3647 auth_zone_parse_notify_serial(sldns_buffer* pkt, uint32_t *serial)
3648 {
3649         struct query_info q;
3650         uint16_t rdlen;
3651         memset(&q, 0, sizeof(q));
3652         sldns_buffer_set_position(pkt, 0);
3653         if(!query_info_parse(&q, pkt)) return 0;
3654         if(LDNS_ANCOUNT(sldns_buffer_begin(pkt)) == 0) return 0;
3655         /* skip name of RR in answer section */
3656         if(sldns_buffer_remaining(pkt) < 1) return 0;
3657         if(pkt_dname_len(pkt) == 0) return 0;
3658         /* check type */
3659         if(sldns_buffer_remaining(pkt) < 10 /* type,class,ttl,rdatalen*/)
3660                 return 0;
3661         if(sldns_buffer_read_u16(pkt) != LDNS_RR_TYPE_SOA) return 0;
3662         sldns_buffer_skip(pkt, 2); /* class */
3663         sldns_buffer_skip(pkt, 4); /* ttl */
3664         rdlen = sldns_buffer_read_u16(pkt); /* rdatalen */
3665         if(sldns_buffer_remaining(pkt) < rdlen) return 0;
3666         if(rdlen < 22) return 0; /* bad soa length */
3667         sldns_buffer_skip(pkt, (ssize_t)(rdlen-20));
3668         *serial = sldns_buffer_read_u32(pkt);
3669         /* return true when has serial in answer section */
3670         return 1;
3671 }
3672
3673 /** see if addr appears in the list */
3674 static int
3675 addr_in_list(struct auth_addr* list, struct sockaddr_storage* addr,
3676         socklen_t addrlen)
3677 {
3678         struct auth_addr* p;
3679         for(p=list; p; p=p->next) {
3680                 if(sockaddr_cmp_addr(addr, addrlen, &p->addr, p->addrlen)==0)
3681                         return 1;
3682         }
3683         return 0;
3684 }
3685
3686 /** check if an address matches a master specification (or one of its
3687  * addresses in the addr list) */
3688 static int
3689 addr_matches_master(struct auth_master* master, struct sockaddr_storage* addr,
3690         socklen_t addrlen, struct auth_master** fromhost)
3691 {
3692         struct sockaddr_storage a;
3693         socklen_t alen = 0;
3694         int net = 0;
3695         if(addr_in_list(master->list, addr, addrlen)) {
3696                 *fromhost = master;
3697                 return 1;       
3698         }
3699         /* compare address (but not port number, that is the destination
3700          * port of the master, the port number of the received notify is
3701          * allowed to by any port on that master) */
3702         if(extstrtoaddr(master->host, &a, &alen) &&
3703                 sockaddr_cmp_addr(addr, addrlen, &a, alen)==0) {
3704                 *fromhost = master;
3705                 return 1;
3706         }
3707         /* prefixes, addr/len, like 10.0.0.0/8 */
3708         /* not http and has a / and there is one / */
3709         if(master->allow_notify && !master->http &&
3710                 strchr(master->host, '/') != NULL &&
3711                 strchr(master->host, '/') == strrchr(master->host, '/') &&
3712                 netblockstrtoaddr(master->host, UNBOUND_DNS_PORT, &a, &alen,
3713                 &net) && alen == addrlen) {
3714                 if(addr_in_common(addr, (addr_is_ip6(addr, addrlen)?128:32),
3715                         &a, net, alen) >= net) {
3716                         *fromhost = NULL; /* prefix does not have destination
3717                                 to send the probe or transfer with */
3718                         return 1; /* matches the netblock */
3719                 }
3720         }
3721         return 0;
3722 }
3723
3724 /** check access list for notifies */
3725 static int
3726 az_xfr_allowed_notify(struct auth_xfer* xfr, struct sockaddr_storage* addr,
3727         socklen_t addrlen, struct auth_master** fromhost)
3728 {
3729         struct auth_master* p;
3730         for(p=xfr->allow_notify_list; p; p=p->next) {
3731                 if(addr_matches_master(p, addr, addrlen, fromhost)) {
3732                         return 1;
3733                 }
3734         }
3735         return 0;
3736 }
3737
3738 /** see if the serial means the zone has to be updated, i.e. the serial
3739  * is newer than the zone serial, or we have no zone */
3740 static int
3741 xfr_serial_means_update(struct auth_xfer* xfr, uint32_t serial)
3742 {
3743         if(!xfr->have_zone)
3744                 return 1; /* no zone, anything is better */
3745         if(xfr->zone_expired)
3746                 return 1; /* expired, the sent serial is better than expired
3747                         data */
3748         if(compare_serial(xfr->serial, serial) < 0)
3749                 return 1; /* our serial is smaller than the sent serial,
3750                         the data is newer, fetch it */
3751         return 0;
3752 }
3753
3754 /** note notify serial, updates the notify information in the xfr struct */
3755 static void
3756 xfr_note_notify_serial(struct auth_xfer* xfr, int has_serial, uint32_t serial)
3757 {
3758         if(xfr->notify_received && xfr->notify_has_serial && has_serial) {
3759                 /* see if this serial is newer */
3760                 if(compare_serial(xfr->notify_serial, serial) < 0)
3761                         xfr->notify_serial = serial;
3762         } else if(xfr->notify_received && xfr->notify_has_serial &&
3763                 !has_serial) {
3764                 /* remove serial, we have notify without serial */
3765                 xfr->notify_has_serial = 0;
3766                 xfr->notify_serial = 0;
3767         } else if(xfr->notify_received && !xfr->notify_has_serial) {
3768                 /* we already have notify without serial, keep it
3769                  * that way; no serial check when current operation
3770                  * is done */
3771         } else {
3772                 xfr->notify_received = 1;
3773                 xfr->notify_has_serial = has_serial;
3774                 xfr->notify_serial = serial;
3775         }
3776 }
3777
3778 /** process a notify serial, start new probe or note serial. xfr is locked */
3779 static void
3780 xfr_process_notify(struct auth_xfer* xfr, struct module_env* env,
3781         int has_serial, uint32_t serial, struct auth_master* fromhost)
3782 {
3783         /* if the serial of notify is older than we have, don't fetch
3784          * a zone, we already have it */
3785         if(has_serial && !xfr_serial_means_update(xfr, serial)) {
3786                 lock_basic_unlock(&xfr->lock);
3787                 return;
3788         }
3789         /* start new probe with this addr src, or note serial */
3790         if(!xfr_start_probe(xfr, env, fromhost)) {
3791                 /* not started because already in progress, note the serial */
3792                 xfr_note_notify_serial(xfr, has_serial, serial);
3793                 lock_basic_unlock(&xfr->lock);
3794         }
3795         /* successful end of start_probe unlocked xfr->lock */
3796 }
3797
3798 int auth_zones_notify(struct auth_zones* az, struct module_env* env,
3799         uint8_t* nm, size_t nmlen, uint16_t dclass,
3800         struct sockaddr_storage* addr, socklen_t addrlen, int has_serial,
3801         uint32_t serial, int* refused)
3802 {
3803         struct auth_xfer* xfr;
3804         struct auth_master* fromhost = NULL;
3805         /* see which zone this is */
3806         lock_rw_rdlock(&az->lock);
3807         xfr = auth_xfer_find(az, nm, nmlen, dclass);
3808         if(!xfr) {
3809                 lock_rw_unlock(&az->lock);
3810                 /* no such zone, refuse the notify */
3811                 *refused = 1;
3812                 return 0;
3813         }
3814         lock_basic_lock(&xfr->lock);
3815         lock_rw_unlock(&az->lock);
3816         
3817         /* check access list for notifies */
3818         if(!az_xfr_allowed_notify(xfr, addr, addrlen, &fromhost)) {
3819                 lock_basic_unlock(&xfr->lock);
3820                 /* notify not allowed, refuse the notify */
3821                 *refused = 1;
3822                 return 0;
3823         }
3824
3825         /* process the notify */
3826         xfr_process_notify(xfr, env, has_serial, serial, fromhost);
3827         return 1;
3828 }
3829
3830 int auth_zones_startprobesequence(struct auth_zones* az,
3831         struct module_env* env, uint8_t* nm, size_t nmlen, uint16_t dclass)
3832 {
3833         struct auth_xfer* xfr;
3834         lock_rw_rdlock(&az->lock);
3835         xfr = auth_xfer_find(az, nm, nmlen, dclass);
3836         if(!xfr) {
3837                 lock_rw_unlock(&az->lock);
3838                 return 0;
3839         }
3840         lock_basic_lock(&xfr->lock);
3841         lock_rw_unlock(&az->lock);
3842
3843         xfr_process_notify(xfr, env, 0, 0, NULL);
3844         return 1;
3845 }
3846
3847 /** set a zone expired */
3848 static void
3849 auth_xfer_set_expired(struct auth_xfer* xfr, struct module_env* env,
3850         int expired)
3851 {
3852         struct auth_zone* z;
3853
3854         /* expire xfr */
3855         lock_basic_lock(&xfr->lock);
3856         xfr->zone_expired = expired;
3857         lock_basic_unlock(&xfr->lock);
3858
3859         /* find auth_zone */
3860         lock_rw_rdlock(&env->auth_zones->lock);
3861         z = auth_zone_find(env->auth_zones, xfr->name, xfr->namelen,
3862                 xfr->dclass);
3863         if(!z) {
3864                 lock_rw_unlock(&env->auth_zones->lock);
3865                 return;
3866         }
3867         lock_rw_wrlock(&z->lock);
3868         lock_rw_unlock(&env->auth_zones->lock);
3869
3870         /* expire auth_zone */
3871         z->zone_expired = expired;
3872         lock_rw_unlock(&z->lock);
3873 }
3874
3875 /** find master (from notify or probe) in list of masters */
3876 static struct auth_master*
3877 find_master_by_host(struct auth_master* list, char* host)
3878 {
3879         struct auth_master* p;
3880         for(p=list; p; p=p->next) {
3881                 if(strcmp(p->host, host) == 0)
3882                         return p;
3883         }
3884         return NULL;
3885 }
3886
3887 /** delete the looked up auth_addrs for all the masters in the list */
3888 static void
3889 xfr_masterlist_free_addrs(struct auth_master* list)
3890 {
3891         struct auth_master* m;
3892         for(m=list; m; m=m->next) {
3893                 if(m->list) {
3894                         auth_free_master_addrs(m->list);
3895                         m->list = NULL;
3896                 }
3897         }
3898 }
3899
3900 /** copy a list of auth_addrs */
3901 static struct auth_addr*
3902 auth_addr_list_copy(struct auth_addr* source)
3903 {
3904         struct auth_addr* list = NULL, *last = NULL;
3905         struct auth_addr* p;
3906         for(p=source; p; p=p->next) {
3907                 struct auth_addr* a = (struct auth_addr*)memdup(p, sizeof(*p));
3908                 if(!a) {
3909                         log_err("malloc failure");
3910                         auth_free_master_addrs(list);
3911                         return NULL;
3912                 }
3913                 a->next = NULL;
3914                 if(last) last->next = a;
3915                 if(!list) list = a;
3916                 last = a;
3917         }
3918         return list;
3919 }
3920
3921 /** copy a master to a new structure, NULL on alloc failure */
3922 static struct auth_master*
3923 auth_master_copy(struct auth_master* o)
3924 {
3925         struct auth_master* m;
3926         if(!o) return NULL;
3927         m = (struct auth_master*)memdup(o, sizeof(*o));
3928         if(!m) {
3929                 log_err("malloc failure");
3930                 return NULL;
3931         }
3932         m->next = NULL;
3933         if(m->host) {
3934                 m->host = strdup(m->host);
3935                 if(!m->host) {
3936                         free(m);
3937                         log_err("malloc failure");
3938                         return NULL;
3939                 }
3940         }
3941         if(m->file) {
3942                 m->file = strdup(m->file);
3943                 if(!m->file) {
3944                         free(m->host);
3945                         free(m);
3946                         log_err("malloc failure");
3947                         return NULL;
3948                 }
3949         }
3950         if(m->list) {
3951                 m->list = auth_addr_list_copy(m->list);
3952                 if(!m->list) {
3953                         free(m->file);
3954                         free(m->host);
3955                         free(m);
3956                         return NULL;
3957                 }
3958         }
3959         return m;
3960 }
3961
3962 /** copy the master addresses from the task_probe lookups to the allow_notify
3963  * list of masters */
3964 static void
3965 probe_copy_masters_for_allow_notify(struct auth_xfer* xfr)
3966 {
3967         struct auth_master* list = NULL, *last = NULL;
3968         struct auth_master* p;
3969         /* build up new list with copies */
3970         for(p = xfr->task_probe->masters; p; p=p->next) {
3971                 struct auth_master* m = auth_master_copy(p);
3972                 if(!m) {
3973                         auth_free_masters(list);
3974                         /* failed because of malloc failure, use old list */
3975                         return;
3976                 }
3977                 m->next = NULL;
3978                 if(last) last->next = m;
3979                 if(!list) list = m;
3980                 last = m;
3981         }
3982         /* success, replace list */
3983         auth_free_masters(xfr->allow_notify_list);
3984         xfr->allow_notify_list = list;
3985 }
3986
3987 /** start the lookups for task_transfer */
3988 static void
3989 xfr_transfer_start_lookups(struct auth_xfer* xfr)
3990 {
3991         /* delete all the looked up addresses in the list */
3992         xfr->task_transfer->scan_addr = NULL;
3993         xfr_masterlist_free_addrs(xfr->task_transfer->masters);
3994
3995         /* start lookup at the first master */
3996         xfr->task_transfer->lookup_target = xfr->task_transfer->masters;
3997         xfr->task_transfer->lookup_aaaa = 0;
3998 }
3999
4000 /** move to the next lookup of hostname for task_transfer */
4001 static void
4002 xfr_transfer_move_to_next_lookup(struct auth_xfer* xfr, struct module_env* env)
4003 {
4004         if(!xfr->task_transfer->lookup_target)
4005                 return; /* already at end of list */
4006         if(!xfr->task_transfer->lookup_aaaa && env->cfg->do_ip6) {
4007                 /* move to lookup AAAA */
4008                 xfr->task_transfer->lookup_aaaa = 1;
4009                 return;
4010         }
4011         xfr->task_transfer->lookup_target = 
4012                 xfr->task_transfer->lookup_target->next;
4013         xfr->task_transfer->lookup_aaaa = 0;
4014         if(!env->cfg->do_ip4 && xfr->task_transfer->lookup_target!=NULL)
4015                 xfr->task_transfer->lookup_aaaa = 1;
4016 }
4017
4018 /** start the lookups for task_probe */
4019 static void
4020 xfr_probe_start_lookups(struct auth_xfer* xfr)
4021 {
4022         /* delete all the looked up addresses in the list */
4023         xfr->task_probe->scan_addr = NULL;
4024         xfr_masterlist_free_addrs(xfr->task_probe->masters);
4025
4026         /* start lookup at the first master */
4027         xfr->task_probe->lookup_target = xfr->task_probe->masters;
4028         xfr->task_probe->lookup_aaaa = 0;
4029 }
4030
4031 /** move to the next lookup of hostname for task_probe */
4032 static void
4033 xfr_probe_move_to_next_lookup(struct auth_xfer* xfr, struct module_env* env)
4034 {
4035         if(!xfr->task_probe->lookup_target)
4036                 return; /* already at end of list */
4037         if(!xfr->task_probe->lookup_aaaa && env->cfg->do_ip6) {
4038                 /* move to lookup AAAA */
4039                 xfr->task_probe->lookup_aaaa = 1;
4040                 return;
4041         }
4042         xfr->task_probe->lookup_target = xfr->task_probe->lookup_target->next;
4043         xfr->task_probe->lookup_aaaa = 0;
4044         if(!env->cfg->do_ip4 && xfr->task_probe->lookup_target!=NULL)
4045                 xfr->task_probe->lookup_aaaa = 1;
4046 }
4047
4048 /** start the iteration of the task_transfer list of masters */
4049 static void
4050 xfr_transfer_start_list(struct auth_xfer* xfr, struct auth_master* spec) 
4051 {
4052         if(spec) {
4053                 xfr->task_transfer->scan_specific = find_master_by_host(
4054                         xfr->task_transfer->masters, spec->host);
4055                 if(xfr->task_transfer->scan_specific) {
4056                         xfr->task_transfer->scan_target = NULL;
4057                         xfr->task_transfer->scan_addr = NULL;
4058                         if(xfr->task_transfer->scan_specific->list)
4059                                 xfr->task_transfer->scan_addr =
4060                                         xfr->task_transfer->scan_specific->list;
4061                         return;
4062                 }
4063         }
4064         /* no specific (notified) host to scan */
4065         xfr->task_transfer->scan_specific = NULL;
4066         xfr->task_transfer->scan_addr = NULL;
4067         /* pick up first scan target */
4068         xfr->task_transfer->scan_target = xfr->task_transfer->masters;
4069         if(xfr->task_transfer->scan_target && xfr->task_transfer->
4070                 scan_target->list)
4071                 xfr->task_transfer->scan_addr =
4072                         xfr->task_transfer->scan_target->list;
4073 }
4074
4075 /** start the iteration of the task_probe list of masters */
4076 static void
4077 xfr_probe_start_list(struct auth_xfer* xfr, struct auth_master* spec) 
4078 {
4079         if(spec) {
4080                 xfr->task_probe->scan_specific = find_master_by_host(
4081                         xfr->task_probe->masters, spec->host);
4082                 if(xfr->task_probe->scan_specific) {
4083                         xfr->task_probe->scan_target = NULL;
4084                         xfr->task_probe->scan_addr = NULL;
4085                         if(xfr->task_probe->scan_specific->list)
4086                                 xfr->task_probe->scan_addr =
4087                                         xfr->task_probe->scan_specific->list;
4088                         return;
4089                 }
4090         }
4091         /* no specific (notified) host to scan */
4092         xfr->task_probe->scan_specific = NULL;
4093         xfr->task_probe->scan_addr = NULL;
4094         /* pick up first scan target */
4095         xfr->task_probe->scan_target = xfr->task_probe->masters;
4096         if(xfr->task_probe->scan_target && xfr->task_probe->scan_target->list)
4097                 xfr->task_probe->scan_addr =
4098                         xfr->task_probe->scan_target->list;
4099 }
4100
4101 /** pick up the master that is being scanned right now, task_transfer */
4102 static struct auth_master*
4103 xfr_transfer_current_master(struct auth_xfer* xfr)
4104 {
4105         if(xfr->task_transfer->scan_specific)
4106                 return xfr->task_transfer->scan_specific;
4107         return xfr->task_transfer->scan_target;
4108 }
4109
4110 /** pick up the master that is being scanned right now, task_probe */
4111 static struct auth_master*
4112 xfr_probe_current_master(struct auth_xfer* xfr)
4113 {
4114         if(xfr->task_probe->scan_specific)
4115                 return xfr->task_probe->scan_specific;
4116         return xfr->task_probe->scan_target;
4117 }
4118
4119 /** true if at end of list, task_transfer */
4120 static int
4121 xfr_transfer_end_of_list(struct auth_xfer* xfr)
4122 {
4123         return !xfr->task_transfer->scan_specific &&
4124                 !xfr->task_transfer->scan_target;
4125 }
4126
4127 /** true if at end of list, task_probe */
4128 static int
4129 xfr_probe_end_of_list(struct auth_xfer* xfr)
4130 {
4131         return !xfr->task_probe->scan_specific && !xfr->task_probe->scan_target;
4132 }
4133
4134 /** move to next master in list, task_transfer */
4135 static void
4136 xfr_transfer_nextmaster(struct auth_xfer* xfr)
4137 {
4138         if(!xfr->task_transfer->scan_specific &&
4139                 !xfr->task_transfer->scan_target)
4140                 return;
4141         if(xfr->task_transfer->scan_addr) {
4142                 xfr->task_transfer->scan_addr =
4143                         xfr->task_transfer->scan_addr->next;
4144                 if(xfr->task_transfer->scan_addr)
4145                         return;
4146         }
4147         if(xfr->task_transfer->scan_specific) {
4148                 xfr->task_transfer->scan_specific = NULL;
4149                 xfr->task_transfer->scan_target = xfr->task_transfer->masters;
4150                 if(xfr->task_transfer->scan_target && xfr->task_transfer->
4151                         scan_target->list)
4152                         xfr->task_transfer->scan_addr =
4153                                 xfr->task_transfer->scan_target->list;
4154                 return;
4155         }
4156         if(!xfr->task_transfer->scan_target)
4157                 return;
4158         xfr->task_transfer->scan_target = xfr->task_transfer->scan_target->next;
4159         if(xfr->task_transfer->scan_target && xfr->task_transfer->
4160                 scan_target->list)
4161                 xfr->task_transfer->scan_addr =
4162                         xfr->task_transfer->scan_target->list;
4163         return;
4164 }
4165
4166 /** move to next master in list, task_probe */
4167 static void
4168 xfr_probe_nextmaster(struct auth_xfer* xfr)
4169 {
4170         if(!xfr->task_probe->scan_specific && !xfr->task_probe->scan_target)
4171                 return;
4172         if(xfr->task_probe->scan_addr) {
4173                 xfr->task_probe->scan_addr = xfr->task_probe->scan_addr->next;
4174                 if(xfr->task_probe->scan_addr)
4175                         return;
4176         }
4177         if(xfr->task_probe->scan_specific) {
4178                 xfr->task_probe->scan_specific = NULL;
4179                 xfr->task_probe->scan_target = xfr->task_probe->masters;
4180                 if(xfr->task_probe->scan_target && xfr->task_probe->
4181                         scan_target->list)
4182                         xfr->task_probe->scan_addr =
4183                                 xfr->task_probe->scan_target->list;
4184                 return;
4185         }
4186         if(!xfr->task_probe->scan_target)
4187                 return;
4188         xfr->task_probe->scan_target = xfr->task_probe->scan_target->next;
4189         if(xfr->task_probe->scan_target && xfr->task_probe->
4190                 scan_target->list)
4191                 xfr->task_probe->scan_addr =
4192                         xfr->task_probe->scan_target->list;
4193         return;
4194 }
4195
4196 /** create SOA probe packet for xfr */
4197 static void
4198 xfr_create_soa_probe_packet(struct auth_xfer* xfr, sldns_buffer* buf, 
4199         uint16_t id)
4200 {
4201         struct query_info qinfo;
4202
4203         memset(&qinfo, 0, sizeof(qinfo));
4204         qinfo.qname = xfr->name;
4205         qinfo.qname_len = xfr->namelen;
4206         qinfo.qtype = LDNS_RR_TYPE_SOA;
4207         qinfo.qclass = xfr->dclass;
4208         qinfo_query_encode(buf, &qinfo);
4209         sldns_buffer_write_u16_at(buf, 0, id);
4210 }
4211
4212 /** create IXFR/AXFR packet for xfr */
4213 static void
4214 xfr_create_ixfr_packet(struct auth_xfer* xfr, sldns_buffer* buf, uint16_t id,
4215         struct auth_master* master)
4216 {
4217         struct query_info qinfo;
4218         uint32_t serial;
4219         int have_zone;
4220         have_zone = xfr->have_zone;
4221         serial = xfr->serial;
4222
4223         memset(&qinfo, 0, sizeof(qinfo));
4224         qinfo.qname = xfr->name;
4225         qinfo.qname_len = xfr->namelen;
4226         xfr->task_transfer->got_xfr_serial = 0;
4227         xfr->task_transfer->rr_scan_num = 0;
4228         xfr->task_transfer->incoming_xfr_serial = 0;
4229         xfr->task_transfer->on_ixfr_is_axfr = 0;
4230         xfr->task_transfer->on_ixfr = 1;
4231         qinfo.qtype = LDNS_RR_TYPE_IXFR;
4232         if(!have_zone || xfr->task_transfer->ixfr_fail || !master->ixfr) {
4233                 qinfo.qtype = LDNS_RR_TYPE_AXFR;
4234                 xfr->task_transfer->ixfr_fail = 0;
4235                 xfr->task_transfer->on_ixfr = 0;
4236         }
4237
4238         qinfo.qclass = xfr->dclass;
4239         qinfo_query_encode(buf, &qinfo);
4240         sldns_buffer_write_u16_at(buf, 0, id);
4241
4242         /* append serial for IXFR */
4243         if(qinfo.qtype == LDNS_RR_TYPE_IXFR) {
4244                 size_t end = sldns_buffer_limit(buf);
4245                 sldns_buffer_clear(buf);
4246                 sldns_buffer_set_position(buf, end);
4247                 /* auth section count 1 */
4248                 sldns_buffer_write_u16_at(buf, LDNS_NSCOUNT_OFF, 1);
4249                 /* write SOA */
4250                 sldns_buffer_write_u8(buf, 0xC0); /* compressed ptr to qname */
4251                 sldns_buffer_write_u8(buf, 0x0C);
4252                 sldns_buffer_write_u16(buf, LDNS_RR_TYPE_SOA);
4253                 sldns_buffer_write_u16(buf, qinfo.qclass);
4254                 sldns_buffer_write_u32(buf, 0); /* ttl */
4255                 sldns_buffer_write_u16(buf, 22); /* rdata length */
4256                 sldns_buffer_write_u8(buf, 0); /* . */
4257                 sldns_buffer_write_u8(buf, 0); /* . */
4258                 sldns_buffer_write_u32(buf, serial); /* serial */
4259                 sldns_buffer_write_u32(buf, 0); /* refresh */
4260                 sldns_buffer_write_u32(buf, 0); /* retry */
4261                 sldns_buffer_write_u32(buf, 0); /* expire */
4262                 sldns_buffer_write_u32(buf, 0); /* minimum */
4263                 sldns_buffer_flip(buf);
4264         }
4265 }
4266
4267 /** check if returned packet is OK */
4268 static int
4269 check_packet_ok(sldns_buffer* pkt, uint16_t qtype, struct auth_xfer* xfr,
4270         uint32_t* serial)
4271 {
4272         /* parse to see if packet worked, valid reply */
4273
4274         /* check serial number of SOA */
4275         if(sldns_buffer_limit(pkt) < LDNS_HEADER_SIZE)
4276                 return 0;
4277
4278         /* check ID */
4279         if(LDNS_ID_WIRE(sldns_buffer_begin(pkt)) != xfr->task_probe->id)
4280                 return 0;
4281
4282         /* check flag bits and rcode */
4283         if(!LDNS_QR_WIRE(sldns_buffer_begin(pkt)))
4284                 return 0;
4285         if(LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt)) != LDNS_PACKET_QUERY)
4286                 return 0;
4287         if(LDNS_RCODE_WIRE(sldns_buffer_begin(pkt)) != LDNS_RCODE_NOERROR)
4288                 return 0;
4289
4290         /* check qname */
4291         if(LDNS_QDCOUNT(sldns_buffer_begin(pkt)) != 1)
4292                 return 0;
4293         sldns_buffer_skip(pkt, LDNS_HEADER_SIZE);
4294         if(sldns_buffer_remaining(pkt) < xfr->namelen)
4295                 return 0;
4296         if(query_dname_compare(sldns_buffer_current(pkt), xfr->name) != 0)
4297                 return 0;
4298         sldns_buffer_skip(pkt, (ssize_t)xfr->namelen);
4299
4300         /* check qtype, qclass */
4301         if(sldns_buffer_remaining(pkt) < 4)
4302                 return 0;
4303         if(sldns_buffer_read_u16(pkt) != qtype)
4304                 return 0;
4305         if(sldns_buffer_read_u16(pkt) != xfr->dclass)
4306                 return 0;
4307
4308         if(serial) {
4309                 uint16_t rdlen;
4310                 /* read serial number, from answer section SOA */
4311                 if(LDNS_ANCOUNT(sldns_buffer_begin(pkt)) == 0)
4312                         return 0;
4313                 /* read from first record SOA record */
4314                 if(sldns_buffer_remaining(pkt) < 1)
4315                         return 0;
4316                 if(dname_pkt_compare(pkt, sldns_buffer_current(pkt),
4317                         xfr->name) != 0)
4318                         return 0;
4319                 if(!pkt_dname_len(pkt))
4320                         return 0;
4321                 /* type, class, ttl, rdatalen */
4322                 if(sldns_buffer_remaining(pkt) < 4+4+2)
4323                         return 0;
4324                 if(sldns_buffer_read_u16(pkt) != qtype)
4325                         return 0;
4326                 if(sldns_buffer_read_u16(pkt) != xfr->dclass)
4327                         return 0;
4328                 sldns_buffer_skip(pkt, 4); /* ttl */
4329                 rdlen = sldns_buffer_read_u16(pkt);
4330                 if(sldns_buffer_remaining(pkt) < rdlen)
4331                         return 0;
4332                 if(sldns_buffer_remaining(pkt) < 1)
4333                         return 0;
4334                 if(!pkt_dname_len(pkt)) /* soa name */
4335                         return 0;
4336                 if(sldns_buffer_remaining(pkt) < 1)
4337                         return 0;
4338                 if(!pkt_dname_len(pkt)) /* soa name */
4339                         return 0;
4340                 if(sldns_buffer_remaining(pkt) < 20)
4341                         return 0;
4342                 *serial = sldns_buffer_read_u32(pkt);
4343         }
4344         return 1;
4345 }
4346
4347 /** read one line from chunks into buffer at current position */
4348 static int
4349 chunkline_get_line(struct auth_chunk** chunk, size_t* chunk_pos,
4350         sldns_buffer* buf)
4351 {
4352         int readsome = 0;
4353         while(*chunk) {
4354                 /* more text in this chunk? */
4355                 if(*chunk_pos < (*chunk)->len) {
4356                         readsome = 1;
4357                         while(*chunk_pos < (*chunk)->len) {
4358                                 char c = (char)((*chunk)->data[*chunk_pos]);
4359                                 (*chunk_pos)++;
4360                                 if(sldns_buffer_remaining(buf) < 2) {
4361                                         /* buffer too short */
4362                                         verbose(VERB_ALGO, "http chunkline, "
4363                                                 "line too long");
4364                                         return 0;
4365                                 }
4366                                 sldns_buffer_write_u8(buf, (uint8_t)c);
4367                                 if(c == '\n') {
4368                                         /* we are done */
4369                                         return 1;
4370                                 }
4371                         }
4372                 }
4373                 /* move to next chunk */
4374                 *chunk = (*chunk)->next;
4375                 *chunk_pos = 0;
4376         }
4377         /* no more text */
4378         if(readsome) return 1;
4379         return 0;
4380 }
4381
4382 /** count number of open and closed parenthesis in a chunkline */
4383 static int
4384 chunkline_count_parens(sldns_buffer* buf, size_t start)
4385 {
4386         size_t end = sldns_buffer_position(buf);
4387         size_t i;
4388         int count = 0;
4389         int squote = 0, dquote = 0;
4390         for(i=start; i<end; i++) {
4391                 char c = (char)sldns_buffer_read_u8_at(buf, i);
4392                 if(squote && c != '\'') continue;
4393                 if(dquote && c != '"') continue;
4394                 if(c == '"')
4395                         dquote = !dquote; /* skip quoted part */
4396                 else if(c == '\'')
4397                         squote = !squote; /* skip quoted part */
4398                 else if(c == '(')
4399                         count ++;
4400                 else if(c == ')')
4401                         count --;
4402                 else if(c == ';') {
4403                         /* rest is a comment */
4404                         return count;
4405                 }
4406         }
4407         return count;
4408 }
4409
4410 /** remove trailing ;... comment from a line in the chunkline buffer */
4411 static void
4412 chunkline_remove_trailcomment(sldns_buffer* buf, size_t start)
4413 {
4414         size_t end = sldns_buffer_position(buf);
4415         size_t i;
4416         int squote = 0, dquote = 0;
4417         for(i=start; i<end; i++) {
4418                 char c = (char)sldns_buffer_read_u8_at(buf, i);
4419                 if(squote && c != '\'') continue;
4420                 if(dquote && c != '"') continue;
4421                 if(c == '"')
4422                         dquote = !dquote; /* skip quoted part */
4423                 else if(c == '\'')
4424                         squote = !squote; /* skip quoted part */
4425                 else if(c == ';') {
4426                         /* rest is a comment */
4427                         sldns_buffer_set_position(buf, i);
4428                         return;
4429                 }
4430         }
4431         /* nothing to remove */
4432 }
4433
4434 /** see if a chunkline is a comment line (or empty line) */
4435 static int
4436 chunkline_is_comment_line_or_empty(sldns_buffer* buf)
4437 {
4438         size_t i, end = sldns_buffer_limit(buf);
4439         for(i=0; i<end; i++) {
4440                 char c = (char)sldns_buffer_read_u8_at(buf, i);
4441                 if(c == ';')
4442                         return 1; /* comment */
4443                 else if(c != ' ' && c != '\t' && c != '\r' && c != '\n')
4444                         return 0; /* not a comment */
4445         }
4446         return 1; /* empty */
4447 }
4448
4449 /** find a line with ( ) collated */
4450 static int
4451 chunkline_get_line_collated(struct auth_chunk** chunk, size_t* chunk_pos,
4452         sldns_buffer* buf)
4453 {
4454         size_t pos;
4455         int parens = 0;
4456         sldns_buffer_clear(buf);
4457         pos = sldns_buffer_position(buf);
4458         if(!chunkline_get_line(chunk, chunk_pos, buf)) {
4459                 if(sldns_buffer_position(buf) < sldns_buffer_limit(buf))
4460                         sldns_buffer_write_u8_at(buf, sldns_buffer_position(buf), 0);
4461                 else sldns_buffer_write_u8_at(buf, sldns_buffer_position(buf)-1, 0);
4462                 sldns_buffer_flip(buf);
4463                 return 0;
4464         }
4465         parens += chunkline_count_parens(buf, pos);
4466         while(parens > 0) {
4467                 chunkline_remove_trailcomment(buf, pos);
4468                 pos = sldns_buffer_position(buf);
4469                 if(!chunkline_get_line(chunk, chunk_pos, buf)) {
4470                         if(sldns_buffer_position(buf) < sldns_buffer_limit(buf))
4471                                 sldns_buffer_write_u8_at(buf, sldns_buffer_position(buf), 0);
4472                         else sldns_buffer_write_u8_at(buf, sldns_buffer_position(buf)-1, 0);
4473                         sldns_buffer_flip(buf);
4474                         return 0;
4475                 }
4476                 parens += chunkline_count_parens(buf, pos);
4477         }
4478
4479         if(sldns_buffer_remaining(buf) < 1) {
4480                 verbose(VERB_ALGO, "http chunkline: "
4481                         "line too long");
4482                 return 0;
4483         }
4484         sldns_buffer_write_u8_at(buf, sldns_buffer_position(buf), 0);
4485         sldns_buffer_flip(buf);
4486         return 1;
4487 }
4488
4489 /** process $ORIGIN for http, 0 nothing, 1 done, 2 error */
4490 static int
4491 http_parse_origin(sldns_buffer* buf, struct sldns_file_parse_state* pstate)
4492 {
4493         char* line = (char*)sldns_buffer_begin(buf);
4494         if(strncmp(line, "$ORIGIN", 7) == 0 &&
4495                 isspace((unsigned char)line[7])) {
4496                 int s;
4497                 pstate->origin_len = sizeof(pstate->origin);
4498                 s = sldns_str2wire_dname_buf(sldns_strip_ws(line+8),
4499                         pstate->origin, &pstate->origin_len);
4500                 if(s) {
4501                         pstate->origin_len = 0;
4502                         return 2;
4503                 }
4504                 return 1;
4505         }
4506         return 0;
4507 }
4508
4509 /** process $TTL for http, 0 nothing, 1 done, 2 error */
4510 static int
4511 http_parse_ttl(sldns_buffer* buf, struct sldns_file_parse_state* pstate)
4512 {
4513         char* line = (char*)sldns_buffer_begin(buf);
4514         if(strncmp(line, "$TTL", 4) == 0 &&
4515                 isspace((unsigned char)line[4])) {
4516                 const char* end = NULL;
4517                 int overflow = 0;
4518                 pstate->default_ttl = sldns_str2period(
4519                         sldns_strip_ws(line+5), &end, &overflow);
4520                 if(overflow) {
4521                         return 2;
4522                 }
4523                 return 1;
4524         }
4525         return 0;
4526 }
4527
4528 /** find noncomment RR line in chunks, collates lines if ( ) format */
4529 static int
4530 chunkline_non_comment_RR(struct auth_chunk** chunk, size_t* chunk_pos,
4531         sldns_buffer* buf, struct sldns_file_parse_state* pstate)
4532 {
4533         int ret;
4534         while(chunkline_get_line_collated(chunk, chunk_pos, buf)) {
4535                 if(chunkline_is_comment_line_or_empty(buf)) {
4536                         /* a comment, go to next line */
4537                         continue;
4538                 }
4539                 if((ret=http_parse_origin(buf, pstate))!=0) {
4540                         if(ret == 2)
4541                                 return 0;
4542                         continue; /* $ORIGIN has been handled */
4543                 }
4544                 if((ret=http_parse_ttl(buf, pstate))!=0) {
4545                         if(ret == 2)
4546                                 return 0;
4547                         continue; /* $TTL has been handled */
4548                 }
4549                 return 1;
4550         }
4551         /* no noncomments, fail */
4552         return 0;
4553 }
4554
4555 /** check syntax of chunklist zonefile, parse first RR, return false on
4556  * failure and return a string in the scratch buffer (first RR string)
4557  * on failure. */
4558 static int
4559 http_zonefile_syntax_check(struct auth_xfer* xfr, sldns_buffer* buf)
4560 {
4561         uint8_t rr[LDNS_RR_BUF_SIZE];
4562         size_t rr_len, dname_len = 0;
4563         struct sldns_file_parse_state pstate;
4564         struct auth_chunk* chunk;
4565         size_t chunk_pos;
4566         int e;
4567         memset(&pstate, 0, sizeof(pstate));
4568         pstate.default_ttl = 3600;
4569         if(xfr->namelen < sizeof(pstate.origin)) {
4570                 pstate.origin_len = xfr->namelen;
4571                 memmove(pstate.origin, xfr->name, xfr->namelen);
4572         }
4573         chunk = xfr->task_transfer->chunks_first;
4574         chunk_pos = 0;
4575         if(!chunkline_non_comment_RR(&chunk, &chunk_pos, buf, &pstate)) {
4576                 return 0;
4577         }
4578         rr_len = sizeof(rr);
4579         e=sldns_str2wire_rr_buf((char*)sldns_buffer_begin(buf), rr, &rr_len,
4580                 &dname_len, pstate.default_ttl,
4581                 pstate.origin_len?pstate.origin:NULL, pstate.origin_len,
4582                 pstate.prev_rr_len?pstate.prev_rr:NULL, pstate.prev_rr_len);
4583         if(e != 0) {
4584                 log_err("parse failure on first RR[%d]: %s",
4585                         LDNS_WIREPARSE_OFFSET(e),
4586                         sldns_get_errorstr_parse(LDNS_WIREPARSE_ERROR(e)));
4587                 return 0;
4588         }
4589         /* check that class is correct */
4590         if(sldns_wirerr_get_class(rr, rr_len, dname_len) != xfr->dclass) {
4591                 log_err("parse failure: first record in downloaded zonefile "
4592                         "from wrong RR class");
4593                 return 0;
4594         }
4595         return 1;
4596 }
4597
4598 /** sum sizes of chunklist */
4599 static size_t
4600 chunklist_sum(struct auth_chunk* list)
4601 {
4602         struct auth_chunk* p;
4603         size_t s = 0;
4604         for(p=list; p; p=p->next) {
4605                 s += p->len;
4606         }
4607         return s;
4608 }
4609
4610 /** remove newlines from collated line */
4611 static void
4612 chunkline_newline_removal(sldns_buffer* buf)
4613 {
4614         size_t i, end=sldns_buffer_limit(buf);
4615         for(i=0; i<end; i++) {
4616                 char c = (char)sldns_buffer_read_u8_at(buf, i);
4617                 if(c == '\n' && i==end-1) {
4618                         sldns_buffer_write_u8_at(buf, i, 0);
4619                         sldns_buffer_set_limit(buf, end-1);
4620                         return;
4621                 }
4622                 if(c == '\n')
4623                         sldns_buffer_write_u8_at(buf, i, (uint8_t)' ');
4624         }
4625 }
4626
4627 /** for http download, parse and add RR to zone */
4628 static int
4629 http_parse_add_rr(struct auth_xfer* xfr, struct auth_zone* z,
4630         sldns_buffer* buf, struct sldns_file_parse_state* pstate)
4631 {
4632         uint8_t rr[LDNS_RR_BUF_SIZE];
4633         size_t rr_len, dname_len = 0;
4634         int e;
4635         char* line = (char*)sldns_buffer_begin(buf);
4636         rr_len = sizeof(rr);
4637         e = sldns_str2wire_rr_buf(line, rr, &rr_len, &dname_len,
4638                 pstate->default_ttl,
4639                 pstate->origin_len?pstate->origin:NULL, pstate->origin_len,
4640                 pstate->prev_rr_len?pstate->prev_rr:NULL, pstate->prev_rr_len);
4641         if(e != 0) {
4642                 log_err("%s/%s parse failure RR[%d]: %s in '%s'",
4643                         xfr->task_transfer->master->host,
4644                         xfr->task_transfer->master->file,
4645                         LDNS_WIREPARSE_OFFSET(e),
4646                         sldns_get_errorstr_parse(LDNS_WIREPARSE_ERROR(e)),
4647                         line);
4648                 return 0;
4649         }
4650         if(rr_len == 0)
4651                 return 1; /* empty line or so */
4652
4653         /* set prev */
4654         if(dname_len < sizeof(pstate->prev_rr)) {
4655                 memmove(pstate->prev_rr, rr, dname_len);
4656                 pstate->prev_rr_len = dname_len;
4657         }
4658
4659         return az_insert_rr(z, rr, rr_len, dname_len, NULL);
4660 }
4661
4662 /** RR list iterator, returns RRs from answer section one by one from the
4663  * dns packets in the chunklist */
4664 static void
4665 chunk_rrlist_start(struct auth_xfer* xfr, struct auth_chunk** rr_chunk,
4666         int* rr_num, size_t* rr_pos)
4667 {
4668         *rr_chunk = xfr->task_transfer->chunks_first;
4669         *rr_num = 0;
4670         *rr_pos = 0;
4671 }
4672
4673 /** RR list iterator, see if we are at the end of the list */
4674 static int
4675 chunk_rrlist_end(struct auth_chunk* rr_chunk, int rr_num)
4676 {
4677         while(rr_chunk) {
4678                 if(rr_chunk->len < LDNS_HEADER_SIZE)
4679                         return 1;
4680                 if(rr_num < (int)LDNS_ANCOUNT(rr_chunk->data))
4681                         return 0;
4682                 /* no more RRs in this chunk */
4683                 /* continue with next chunk, see if it has RRs */
4684                 rr_chunk = rr_chunk->next;
4685                 rr_num = 0;
4686         }
4687         return 1;
4688 }
4689
4690 /** RR list iterator, move to next RR */
4691 static void
4692 chunk_rrlist_gonext(struct auth_chunk** rr_chunk, int* rr_num,
4693         size_t* rr_pos, size_t rr_nextpos)
4694 {
4695         /* already at end of chunks? */
4696         if(!*rr_chunk)
4697                 return;
4698         /* move within this chunk */
4699         if((*rr_chunk)->len >= LDNS_HEADER_SIZE &&
4700                 (*rr_num)+1 < (int)LDNS_ANCOUNT((*rr_chunk)->data)) {
4701                 (*rr_num) += 1;
4702                 *rr_pos = rr_nextpos;
4703                 return;
4704         }
4705         /* no more RRs in this chunk */
4706         /* continue with next chunk, see if it has RRs */
4707         if(*rr_chunk)
4708                 *rr_chunk = (*rr_chunk)->next;
4709         while(*rr_chunk) {
4710                 *rr_num = 0;
4711                 *rr_pos = 0;
4712                 if((*rr_chunk)->len >= LDNS_HEADER_SIZE &&
4713                         LDNS_ANCOUNT((*rr_chunk)->data) > 0) {
4714                         return;
4715                 }
4716                 *rr_chunk = (*rr_chunk)->next;
4717         }
4718 }
4719
4720 /** RR iterator, get current RR information, false on parse error */
4721 static int
4722 chunk_rrlist_get_current(struct auth_chunk* rr_chunk, int rr_num,
4723         size_t rr_pos, uint8_t** rr_dname, uint16_t* rr_type,
4724         uint16_t* rr_class, uint32_t* rr_ttl, uint16_t* rr_rdlen,
4725         uint8_t** rr_rdata, size_t* rr_nextpos)
4726 {
4727         sldns_buffer pkt;
4728         /* integrity checks on position */
4729         if(!rr_chunk) return 0;
4730         if(rr_chunk->len < LDNS_HEADER_SIZE) return 0;
4731         if(rr_num >= (int)LDNS_ANCOUNT(rr_chunk->data)) return 0;
4732         if(rr_pos >= rr_chunk->len) return 0;
4733
4734         /* fetch rr information */
4735         sldns_buffer_init_frm_data(&pkt, rr_chunk->data, rr_chunk->len);
4736         if(rr_pos == 0) {
4737                 size_t i;
4738                 /* skip question section */
4739                 sldns_buffer_set_position(&pkt, LDNS_HEADER_SIZE);
4740                 for(i=0; i<LDNS_QDCOUNT(rr_chunk->data); i++) {
4741                         if(pkt_dname_len(&pkt) == 0) return 0;
4742                         if(sldns_buffer_remaining(&pkt) < 4) return 0;
4743                         sldns_buffer_skip(&pkt, 4); /* type and class */
4744                 }
4745         } else  {
4746                 sldns_buffer_set_position(&pkt, rr_pos);
4747         }
4748         *rr_dname = sldns_buffer_current(&pkt);
4749         if(pkt_dname_len(&pkt) == 0) return 0;
4750         if(sldns_buffer_remaining(&pkt) < 10) return 0;
4751         *rr_type = sldns_buffer_read_u16(&pkt);
4752         *rr_class = sldns_buffer_read_u16(&pkt);
4753         *rr_ttl = sldns_buffer_read_u32(&pkt);
4754         *rr_rdlen = sldns_buffer_read_u16(&pkt);
4755         if(sldns_buffer_remaining(&pkt) < (*rr_rdlen)) return 0;
4756         *rr_rdata = sldns_buffer_current(&pkt);
4757         sldns_buffer_skip(&pkt, (ssize_t)(*rr_rdlen));
4758         *rr_nextpos = sldns_buffer_position(&pkt);
4759         return 1;
4760 }
4761
4762 /** print log message where we are in parsing the zone transfer */
4763 static void
4764 log_rrlist_position(const char* label, struct auth_chunk* rr_chunk,
4765         uint8_t* rr_dname, uint16_t rr_type, size_t rr_counter)
4766 {
4767         sldns_buffer pkt;
4768         size_t dlen;
4769         uint8_t buf[256];
4770         char str[256];
4771         char typestr[32];
4772         sldns_buffer_init_frm_data(&pkt, rr_chunk->data, rr_chunk->len);
4773         sldns_buffer_set_position(&pkt, (size_t)(rr_dname -
4774                 sldns_buffer_begin(&pkt)));
4775         if((dlen=pkt_dname_len(&pkt)) == 0) return;
4776         if(dlen >= sizeof(buf)) return;
4777         dname_pkt_copy(&pkt, buf, rr_dname);
4778         dname_str(buf, str);
4779         (void)sldns_wire2str_type_buf(rr_type, typestr, sizeof(typestr));
4780         verbose(VERB_ALGO, "%s at[%d] %s %s", label, (int)rr_counter,
4781                 str, typestr);
4782 }
4783
4784 /** check that start serial is OK for ixfr. we are at rr_counter == 0,
4785  * and we are going to check rr_counter == 1 (has to be type SOA) serial */
4786 static int
4787 ixfr_start_serial(struct auth_chunk* rr_chunk, int rr_num, size_t rr_pos,
4788         uint8_t* rr_dname, uint16_t rr_type, uint16_t rr_class,
4789         uint32_t rr_ttl, uint16_t rr_rdlen, uint8_t* rr_rdata,
4790         size_t rr_nextpos, uint32_t transfer_serial, uint32_t xfr_serial)
4791 {
4792         uint32_t startserial;
4793         /* move forward on RR */
4794         chunk_rrlist_gonext(&rr_chunk, &rr_num, &rr_pos, rr_nextpos);
4795         if(chunk_rrlist_end(rr_chunk, rr_num)) {
4796                 /* no second SOA */
4797                 verbose(VERB_OPS, "IXFR has no second SOA record");
4798                 return 0;
4799         }
4800         if(!chunk_rrlist_get_current(rr_chunk, rr_num, rr_pos,
4801                 &rr_dname, &rr_type, &rr_class, &rr_ttl, &rr_rdlen,
4802                 &rr_rdata, &rr_nextpos)) {
4803                 verbose(VERB_OPS, "IXFR cannot parse second SOA record");
4804                 /* failed to parse RR */
4805                 return 0;
4806         }
4807         if(rr_type != LDNS_RR_TYPE_SOA) {
4808                 verbose(VERB_OPS, "IXFR second record is not type SOA");
4809                 return 0;
4810         }
4811         if(rr_rdlen < 22) {
4812                 verbose(VERB_OPS, "IXFR, second SOA has short rdlength");
4813                 return 0; /* bad SOA rdlen */
4814         }
4815         startserial = sldns_read_uint32(rr_rdata+rr_rdlen-20);
4816         if(startserial == transfer_serial) {
4817                 /* empty AXFR, not an IXFR */
4818                 verbose(VERB_OPS, "IXFR second serial same as first");
4819                 return 0;
4820         }
4821         if(startserial != xfr_serial) {
4822                 /* wrong start serial, it does not match the serial in
4823                  * memory */
4824                 verbose(VERB_OPS, "IXFR is from serial %u to %u but %u "
4825                         "in memory, rejecting the zone transfer",
4826                         (unsigned)startserial, (unsigned)transfer_serial,
4827                         (unsigned)xfr_serial);
4828                 return 0;
4829         }
4830         /* everything OK in second SOA serial */
4831         return 1;
4832 }
4833
4834 /** apply IXFR to zone in memory. z is locked. false on failure(mallocfail) */
4835 static int
4836 apply_ixfr(struct auth_xfer* xfr, struct auth_zone* z,
4837         struct sldns_buffer* scratch_buffer)
4838 {
4839         struct auth_chunk* rr_chunk;
4840         int rr_num;
4841         size_t rr_pos;
4842         uint8_t* rr_dname, *rr_rdata;
4843         uint16_t rr_type, rr_class, rr_rdlen;
4844         uint32_t rr_ttl;
4845         size_t rr_nextpos;
4846         int have_transfer_serial = 0;
4847         uint32_t transfer_serial = 0;
4848         size_t rr_counter = 0;
4849         int delmode = 0;
4850         int softfail = 0;
4851
4852         /* start RR iterator over chunklist of packets */
4853         chunk_rrlist_start(xfr, &rr_chunk, &rr_num, &rr_pos);
4854         while(!chunk_rrlist_end(rr_chunk, rr_num)) {
4855                 if(!chunk_rrlist_get_current(rr_chunk, rr_num, rr_pos,
4856                         &rr_dname, &rr_type, &rr_class, &rr_ttl, &rr_rdlen,
4857                         &rr_rdata, &rr_nextpos)) {
4858                         /* failed to parse RR */
4859                         return 0;
4860                 }
4861                 if(verbosity>=7) log_rrlist_position("apply ixfr",
4862                         rr_chunk, rr_dname, rr_type, rr_counter);
4863                 /* twiddle add/del mode and check for start and end */
4864                 if(rr_counter == 0 && rr_type != LDNS_RR_TYPE_SOA)
4865                         return 0;
4866                 if(rr_counter == 1 && rr_type != LDNS_RR_TYPE_SOA) {
4867                         /* this is an AXFR returned from the IXFR master */
4868                         /* but that should already have been detected, by
4869                          * on_ixfr_is_axfr */
4870                         return 0;
4871                 }
4872                 if(rr_type == LDNS_RR_TYPE_SOA) {
4873                         uint32_t serial;
4874                         if(rr_rdlen < 22) return 0; /* bad SOA rdlen */
4875                         serial = sldns_read_uint32(rr_rdata+rr_rdlen-20);
4876                         if(have_transfer_serial == 0) {
4877                                 have_transfer_serial = 1;
4878                                 transfer_serial = serial;
4879                                 delmode = 1; /* gets negated below */
4880                                 /* check second RR before going any further */
4881                                 if(!ixfr_start_serial(rr_chunk, rr_num, rr_pos,
4882                                         rr_dname, rr_type, rr_class, rr_ttl,
4883                                         rr_rdlen, rr_rdata, rr_nextpos,
4884                                         transfer_serial, xfr->serial)) {
4885                                         return 0;
4886                                 }
4887                         } else if(transfer_serial == serial) {
4888                                 have_transfer_serial++;
4889                                 if(rr_counter == 1) {
4890                                         /* empty AXFR, with SOA; SOA; */
4891                                         /* should have been detected by
4892                                          * on_ixfr_is_axfr */
4893                                         return 0;
4894                                 }
4895                                 if(have_transfer_serial == 3) {
4896                                         /* see serial three times for end */
4897                                         /* eg. IXFR:
4898                                          *  SOA 3 start
4899                                          *  SOA 1 second RR, followed by del
4900                                          *  SOA 2 followed by add
4901                                          *  SOA 2 followed by del
4902                                          *  SOA 3 followed by add
4903                                          *  SOA 3 end */
4904                                         /* ended by SOA record */
4905                                         xfr->serial = transfer_serial;
4906                                         break;
4907                                 }
4908                         }
4909                         /* twiddle add/del mode */
4910                         /* switch from delete part to add part and back again
4911                          * just before the soa, it gets deleted and added too
4912                          * this means we switch to delete mode for the final
4913                          * SOA(so skip that one) */
4914                         delmode = !delmode;
4915                 }
4916                 /* process this RR */
4917                 /* if the RR is deleted twice or added twice, then we 
4918                  * softfail, and continue with the rest of the IXFR, so
4919                  * that we serve something fairly nice during the refetch */
4920                 if(verbosity>=7) log_rrlist_position((delmode?"del":"add"),
4921                         rr_chunk, rr_dname, rr_type, rr_counter);
4922                 if(delmode) {
4923                         /* delete this RR */
4924                         int nonexist = 0;
4925                         if(!az_remove_rr_decompress(z, rr_chunk->data,
4926                                 rr_chunk->len, scratch_buffer, rr_dname,
4927                                 rr_type, rr_class, rr_ttl, rr_rdata, rr_rdlen,
4928                                 &nonexist)) {
4929                                 /* failed, malloc error or so */
4930                                 return 0;
4931                         }
4932                         if(nonexist) {
4933                                 /* it was removal of a nonexisting RR */
4934                                 if(verbosity>=4) log_rrlist_position(
4935                                         "IXFR error nonexistent RR",
4936                                         rr_chunk, rr_dname, rr_type, rr_counter);
4937                                 softfail = 1;
4938                         }
4939                 } else if(rr_counter != 0) {
4940                         /* skip first SOA RR for addition, it is added in
4941                          * the addition part near the end of the ixfr, when
4942                          * that serial is seen the second time. */
4943                         int duplicate = 0;
4944                         /* add this RR */
4945                         if(!az_insert_rr_decompress(z, rr_chunk->data,
4946                                 rr_chunk->len, scratch_buffer, rr_dname,
4947                                 rr_type, rr_class, rr_ttl, rr_rdata, rr_rdlen,
4948                                 &duplicate)) {
4949                                 /* failed, malloc error or so */
4950                                 return 0;
4951                         }
4952                         if(duplicate) {
4953                                 /* it was a duplicate */
4954                                 if(verbosity>=4) log_rrlist_position(
4955                                         "IXFR error duplicate RR",
4956                                         rr_chunk, rr_dname, rr_type, rr_counter);
4957                                 softfail = 1;
4958                         }
4959                 }
4960
4961                 rr_counter++;
4962                 chunk_rrlist_gonext(&rr_chunk, &rr_num, &rr_pos, rr_nextpos);
4963         }
4964         if(softfail) {
4965                 verbose(VERB_ALGO, "IXFR did not apply cleanly, fetching full zone");
4966                 return 0;
4967         }
4968         return 1;
4969 }
4970
4971 /** apply AXFR to zone in memory. z is locked. false on failure(mallocfail) */
4972 static int
4973 apply_axfr(struct auth_xfer* xfr, struct auth_zone* z,
4974         struct sldns_buffer* scratch_buffer)
4975 {
4976         struct auth_chunk* rr_chunk;
4977         int rr_num;
4978         size_t rr_pos;
4979         uint8_t* rr_dname, *rr_rdata;
4980         uint16_t rr_type, rr_class, rr_rdlen;
4981         uint32_t rr_ttl;
4982         uint32_t serial = 0;
4983         size_t rr_nextpos;
4984         size_t rr_counter = 0;
4985         int have_end_soa = 0;
4986
4987         /* clear the data tree */
4988         traverse_postorder(&z->data, auth_data_del, NULL);
4989         rbtree_init(&z->data, &auth_data_cmp);
4990         /* clear the RPZ policies */
4991         if(z->rpz)
4992                 rpz_clear(z->rpz);
4993
4994         xfr->have_zone = 0;
4995         xfr->serial = 0;
4996
4997         /* insert all RRs in to the zone */
4998         /* insert the SOA only once, skip the last one */
4999         /* start RR iterator over chunklist of packets */
5000         chunk_rrlist_start(xfr, &rr_chunk, &rr_num, &rr_pos);
5001         while(!chunk_rrlist_end(rr_chunk, rr_num)) {
5002                 if(!chunk_rrlist_get_current(rr_chunk, rr_num, rr_pos,
5003                         &rr_dname, &rr_type, &rr_class, &rr_ttl, &rr_rdlen,
5004                         &rr_rdata, &rr_nextpos)) {
5005                         /* failed to parse RR */
5006                         return 0;
5007                 }
5008                 if(verbosity>=7) log_rrlist_position("apply_axfr",
5009                         rr_chunk, rr_dname, rr_type, rr_counter);
5010                 if(rr_type == LDNS_RR_TYPE_SOA) {
5011                         if(rr_counter != 0) {
5012                                 /* end of the axfr */
5013                                 have_end_soa = 1;
5014                                 break;
5015                         }
5016                         if(rr_rdlen < 22) return 0; /* bad SOA rdlen */
5017                         serial = sldns_read_uint32(rr_rdata+rr_rdlen-20);
5018                 }
5019
5020                 /* add this RR */
5021                 if(!az_insert_rr_decompress(z, rr_chunk->data, rr_chunk->len,
5022                         scratch_buffer, rr_dname, rr_type, rr_class, rr_ttl,
5023                         rr_rdata, rr_rdlen, NULL)) {
5024                         /* failed, malloc error or so */
5025                         return 0;
5026                 }
5027
5028                 rr_counter++;
5029                 chunk_rrlist_gonext(&rr_chunk, &rr_num, &rr_pos, rr_nextpos);
5030         }
5031         if(!have_end_soa) {
5032                 log_err("no end SOA record for AXFR");
5033                 return 0;
5034         }
5035
5036         xfr->serial = serial;
5037         xfr->have_zone = 1;
5038         return 1;
5039 }
5040
5041 /** apply HTTP to zone in memory. z is locked. false on failure(mallocfail) */
5042 static int
5043 apply_http(struct auth_xfer* xfr, struct auth_zone* z,
5044         struct sldns_buffer* scratch_buffer)
5045 {
5046         /* parse data in chunks */
5047         /* parse RR's and read into memory. ignore $INCLUDE from the
5048          * downloaded file*/
5049         struct sldns_file_parse_state pstate;
5050         struct auth_chunk* chunk;
5051         size_t chunk_pos;
5052         int ret;
5053         memset(&pstate, 0, sizeof(pstate));
5054         pstate.default_ttl = 3600;
5055         if(xfr->namelen < sizeof(pstate.origin)) {
5056                 pstate.origin_len = xfr->namelen;
5057                 memmove(pstate.origin, xfr->name, xfr->namelen);
5058         }
5059
5060         if(verbosity >= VERB_ALGO)
5061                 verbose(VERB_ALGO, "http download %s of size %d",
5062                 xfr->task_transfer->master->file,
5063                 (int)chunklist_sum(xfr->task_transfer->chunks_first));
5064         if(xfr->task_transfer->chunks_first && verbosity >= VERB_ALGO) {
5065                 char preview[1024];
5066                 if(xfr->task_transfer->chunks_first->len+1 > sizeof(preview)) {
5067                         memmove(preview, xfr->task_transfer->chunks_first->data,
5068                                 sizeof(preview)-1);
5069                         preview[sizeof(preview)-1]=0;
5070                 } else {
5071                         memmove(preview, xfr->task_transfer->chunks_first->data,
5072                                 xfr->task_transfer->chunks_first->len);
5073                         preview[xfr->task_transfer->chunks_first->len]=0;
5074                 }
5075                 log_info("auth zone http downloaded content preview: %s",
5076                         preview);
5077         }
5078
5079         /* perhaps a little syntax check before we try to apply the data? */
5080         if(!http_zonefile_syntax_check(xfr, scratch_buffer)) {
5081                 log_err("http download %s/%s does not contain a zonefile, "
5082                         "but got '%s'", xfr->task_transfer->master->host,
5083                         xfr->task_transfer->master->file,
5084                         sldns_buffer_begin(scratch_buffer));
5085                 return 0;
5086         }
5087
5088         /* clear the data tree */
5089         traverse_postorder(&z->data, auth_data_del, NULL);
5090         rbtree_init(&z->data, &auth_data_cmp);
5091         /* clear the RPZ policies */
5092         if(z->rpz)
5093                 rpz_clear(z->rpz);
5094
5095         xfr->have_zone = 0;
5096         xfr->serial = 0;
5097
5098         chunk = xfr->task_transfer->chunks_first;
5099         chunk_pos = 0;
5100         pstate.lineno = 0;
5101         while(chunkline_get_line_collated(&chunk, &chunk_pos, scratch_buffer)) {
5102                 /* process this line */
5103                 pstate.lineno++;
5104                 chunkline_newline_removal(scratch_buffer);
5105                 if(chunkline_is_comment_line_or_empty(scratch_buffer)) {
5106                         continue;
5107                 }
5108                 /* parse line and add RR */
5109                 if((ret=http_parse_origin(scratch_buffer, &pstate))!=0) {
5110                         if(ret == 2) {
5111                                 verbose(VERB_ALGO, "error parsing ORIGIN on line [%s:%d] %s",
5112                                         xfr->task_transfer->master->file,
5113                                         pstate.lineno,
5114                                         sldns_buffer_begin(scratch_buffer));
5115                                 return 0;
5116                         }
5117                         continue; /* $ORIGIN has been handled */
5118                 }
5119                 if((ret=http_parse_ttl(scratch_buffer, &pstate))!=0) {
5120                         if(ret == 2) {
5121                                 verbose(VERB_ALGO, "error parsing TTL on line [%s:%d] %s",
5122                                         xfr->task_transfer->master->file,
5123                                         pstate.lineno,
5124                                         sldns_buffer_begin(scratch_buffer));
5125                                 return 0;
5126                         }
5127                         continue; /* $TTL has been handled */
5128                 }
5129                 if(!http_parse_add_rr(xfr, z, scratch_buffer, &pstate)) {
5130                         verbose(VERB_ALGO, "error parsing line [%s:%d] %s",
5131                                 xfr->task_transfer->master->file,
5132                                 pstate.lineno,
5133                                 sldns_buffer_begin(scratch_buffer));
5134                         return 0;
5135                 }
5136         }
5137         return 1;
5138 }
5139
5140 /** write http chunks to zonefile to create downloaded file */
5141 static int
5142 auth_zone_write_chunks(struct auth_xfer* xfr, const char* fname)
5143 {
5144         FILE* out;
5145         struct auth_chunk* p;
5146         out = fopen(fname, "w");
5147         if(!out) {
5148                 log_err("could not open %s: %s", fname, strerror(errno));
5149                 return 0;
5150         }
5151         for(p = xfr->task_transfer->chunks_first; p ; p = p->next) {
5152                 if(!write_out(out, (char*)p->data, p->len)) {
5153                         log_err("could not write http download to %s", fname);
5154                         fclose(out);
5155                         return 0;
5156                 }
5157         }
5158         fclose(out);
5159         return 1;
5160 }
5161
5162 /** write to zonefile after zone has been updated */
5163 static void
5164 xfr_write_after_update(struct auth_xfer* xfr, struct module_env* env)
5165 {
5166         struct config_file* cfg = env->cfg;
5167         struct auth_zone* z;
5168         char tmpfile[1024];
5169         char* zfilename;
5170         lock_basic_unlock(&xfr->lock);
5171
5172         /* get lock again, so it is a readlock and concurrently queries
5173          * can be answered */
5174         lock_rw_rdlock(&env->auth_zones->lock);
5175         z = auth_zone_find(env->auth_zones, xfr->name, xfr->namelen,
5176                 xfr->dclass);
5177         if(!z) {
5178                 lock_rw_unlock(&env->auth_zones->lock);
5179                 /* the zone is gone, ignore xfr results */
5180                 lock_basic_lock(&xfr->lock);
5181                 return;
5182         }
5183         lock_rw_rdlock(&z->lock);
5184         lock_basic_lock(&xfr->lock);
5185         lock_rw_unlock(&env->auth_zones->lock);
5186
5187         if(z->zonefile == NULL || z->zonefile[0] == 0) {
5188                 lock_rw_unlock(&z->lock);
5189                 /* no write needed, no zonefile set */
5190                 return;
5191         }
5192         zfilename = z->zonefile;
5193         if(cfg->chrootdir && cfg->chrootdir[0] && strncmp(zfilename,
5194                 cfg->chrootdir, strlen(cfg->chrootdir)) == 0)
5195                 zfilename += strlen(cfg->chrootdir);
5196         if(verbosity >= VERB_ALGO) {
5197                 char nm[255+1];
5198                 dname_str(z->name, nm);
5199                 verbose(VERB_ALGO, "write zonefile %s for %s", zfilename, nm);
5200         }
5201
5202         /* write to tempfile first */
5203         if((size_t)strlen(zfilename) + 16 > sizeof(tmpfile)) {
5204                 verbose(VERB_ALGO, "tmpfilename too long, cannot update "
5205                         " zonefile %s", zfilename);
5206                 lock_rw_unlock(&z->lock);
5207                 return;
5208         }
5209         snprintf(tmpfile, sizeof(tmpfile), "%s.tmp%u", zfilename,
5210                 (unsigned)getpid());
5211         if(xfr->task_transfer->master->http) {
5212                 /* use the stored chunk list to write them */
5213                 if(!auth_zone_write_chunks(xfr, tmpfile)) {
5214                         unlink(tmpfile);
5215                         lock_rw_unlock(&z->lock);
5216                         return;
5217                 }
5218         } else if(!auth_zone_write_file(z, tmpfile)) {
5219                 unlink(tmpfile);
5220                 lock_rw_unlock(&z->lock);
5221                 return;
5222         }
5223 #ifdef UB_ON_WINDOWS
5224         (void)unlink(zfilename); /* windows does not replace file with rename() */
5225 #endif
5226         if(rename(tmpfile, zfilename) < 0) {
5227                 log_err("could not rename(%s, %s): %s", tmpfile, zfilename,
5228                         strerror(errno));
5229                 unlink(tmpfile);
5230                 lock_rw_unlock(&z->lock);
5231                 return;
5232         }
5233         lock_rw_unlock(&z->lock);
5234 }
5235
5236 /** reacquire locks and structures. Starts with no locks, ends
5237  * with xfr and z locks, if fail, no z lock */
5238 static int xfr_process_reacquire_locks(struct auth_xfer* xfr,
5239         struct module_env* env, struct auth_zone** z)
5240 {
5241         /* release xfr lock, then, while holding az->lock grab both
5242          * z->lock and xfr->lock */
5243         lock_rw_rdlock(&env->auth_zones->lock);
5244         *z = auth_zone_find(env->auth_zones, xfr->name, xfr->namelen,
5245                 xfr->dclass);
5246         if(!*z) {
5247                 lock_rw_unlock(&env->auth_zones->lock);
5248                 lock_basic_lock(&xfr->lock);
5249                 *z = NULL;
5250                 return 0;
5251         }
5252         lock_rw_wrlock(&(*z)->lock);
5253         lock_basic_lock(&xfr->lock);
5254         lock_rw_unlock(&env->auth_zones->lock);
5255         return 1;
5256 }
5257
5258 /** process chunk list and update zone in memory,
5259  * return false if it did not work */
5260 static int
5261 xfr_process_chunk_list(struct auth_xfer* xfr, struct module_env* env,
5262         int* ixfr_fail)
5263 {
5264         struct auth_zone* z;
5265
5266         /* obtain locks and structures */
5267         lock_basic_unlock(&xfr->lock);
5268         if(!xfr_process_reacquire_locks(xfr, env, &z)) {
5269                 /* the zone is gone, ignore xfr results */
5270                 return 0;
5271         }
5272         /* holding xfr and z locks */
5273
5274         /* apply data */
5275         if(xfr->task_transfer->master->http) {
5276                 if(!apply_http(xfr, z, env->scratch_buffer)) {
5277                         lock_rw_unlock(&z->lock);
5278                         verbose(VERB_ALGO, "http from %s: could not store data",
5279                                 xfr->task_transfer->master->host);
5280                         return 0;
5281                 }
5282         } else if(xfr->task_transfer->on_ixfr &&
5283                 !xfr->task_transfer->on_ixfr_is_axfr) {
5284                 if(!apply_ixfr(xfr, z, env->scratch_buffer)) {
5285                         lock_rw_unlock(&z->lock);
5286                         verbose(VERB_ALGO, "xfr from %s: could not store IXFR"
5287                                 " data", xfr->task_transfer->master->host);
5288                         *ixfr_fail = 1;
5289                         return 0;
5290                 }
5291         } else {
5292                 if(!apply_axfr(xfr, z, env->scratch_buffer)) {
5293                         lock_rw_unlock(&z->lock);
5294                         verbose(VERB_ALGO, "xfr from %s: could not store AXFR"
5295                                 " data", xfr->task_transfer->master->host);
5296                         return 0;
5297                 }
5298         }
5299         xfr->zone_expired = 0;
5300         z->zone_expired = 0;
5301         if(!xfr_find_soa(z, xfr)) {
5302                 lock_rw_unlock(&z->lock);
5303                 verbose(VERB_ALGO, "xfr from %s: no SOA in zone after update"
5304                         " (or malformed RR)", xfr->task_transfer->master->host);
5305                 return 0;
5306         }
5307
5308         /* release xfr lock while verifying zonemd because it may have
5309          * to spawn lookups in the state machines */
5310         lock_basic_unlock(&xfr->lock);
5311         /* holding z lock */
5312         auth_zone_verify_zonemd(z, env, &env->mesh->mods, NULL, 0, 0);
5313         if(z->zone_expired) {
5314                 char zname[256];
5315                 /* ZONEMD must have failed */
5316                 /* reacquire locks, so we hold xfr lock on exit of routine,
5317                  * and both xfr and z again after releasing xfr for potential
5318                  * state machine mesh callbacks */
5319                 lock_rw_unlock(&z->lock);
5320                 if(!xfr_process_reacquire_locks(xfr, env, &z))
5321                         return 0;
5322                 dname_str(xfr->name, zname);
5323                 verbose(VERB_ALGO, "xfr from %s: ZONEMD failed for %s, transfer is failed", xfr->task_transfer->master->host, zname);
5324                 xfr->zone_expired = 1;
5325                 lock_rw_unlock(&z->lock);
5326                 return 0;
5327         }
5328         /* reacquire locks, so we hold xfr lock on exit of routine,
5329          * and both xfr and z again after releasing xfr for potential
5330          * state machine mesh callbacks */
5331         lock_rw_unlock(&z->lock);
5332         if(!xfr_process_reacquire_locks(xfr, env, &z))
5333                 return 0;
5334         /* holding xfr and z locks */
5335
5336         if(xfr->have_zone)
5337                 xfr->lease_time = *env->now;
5338
5339         if(z->rpz)
5340                 rpz_finish_config(z->rpz);
5341
5342         /* unlock */
5343         lock_rw_unlock(&z->lock);
5344
5345         if(verbosity >= VERB_QUERY && xfr->have_zone) {
5346                 char zname[256];
5347                 dname_str(xfr->name, zname);
5348                 verbose(VERB_QUERY, "auth zone %s updated to serial %u", zname,
5349                         (unsigned)xfr->serial);
5350         }
5351         /* see if we need to write to a zonefile */
5352         xfr_write_after_update(xfr, env);
5353         return 1;
5354 }
5355
5356 /** disown task_transfer.  caller must hold xfr.lock */
5357 static void
5358 xfr_transfer_disown(struct auth_xfer* xfr)
5359 {
5360         /* remove timer (from this worker's event base) */
5361         comm_timer_delete(xfr->task_transfer->timer);
5362         xfr->task_transfer->timer = NULL;
5363         /* remove the commpoint */
5364         comm_point_delete(xfr->task_transfer->cp);
5365         xfr->task_transfer->cp = NULL;
5366         /* we don't own this item anymore */
5367         xfr->task_transfer->worker = NULL;
5368         xfr->task_transfer->env = NULL;
5369 }
5370
5371 /** lookup a host name for its addresses, if needed */
5372 static int
5373 xfr_transfer_lookup_host(struct auth_xfer* xfr, struct module_env* env)
5374 {
5375         struct sockaddr_storage addr;
5376         socklen_t addrlen = 0;
5377         struct auth_master* master = xfr->task_transfer->lookup_target;
5378         struct query_info qinfo;
5379         uint16_t qflags = BIT_RD;
5380         uint8_t dname[LDNS_MAX_DOMAINLEN+1];
5381         struct edns_data edns;
5382         sldns_buffer* buf = env->scratch_buffer;
5383         if(!master) return 0;
5384         if(extstrtoaddr(master->host, &addr, &addrlen)) {
5385                 /* not needed, host is in IP addr format */
5386                 return 0;
5387         }
5388         if(master->allow_notify)
5389                 return 0; /* allow-notifies are not transferred from, no
5390                 lookup is needed */
5391
5392         /* use mesh_new_callback to probe for non-addr hosts,
5393          * and then wait for them to be looked up (in cache, or query) */
5394         qinfo.qname_len = sizeof(dname);
5395         if(sldns_str2wire_dname_buf(master->host, dname, &qinfo.qname_len)
5396                 != 0) {
5397                 log_err("cannot parse host name of master %s", master->host);
5398                 return 0;
5399         }
5400         qinfo.qname = dname;
5401         qinfo.qclass = xfr->dclass;
5402         qinfo.qtype = LDNS_RR_TYPE_A;
5403         if(xfr->task_transfer->lookup_aaaa)
5404                 qinfo.qtype = LDNS_RR_TYPE_AAAA;
5405         qinfo.local_alias = NULL;
5406         if(verbosity >= VERB_ALGO) {
5407                 char buf1[512];
5408                 char buf2[LDNS_MAX_DOMAINLEN+1];
5409                 dname_str(xfr->name, buf2);
5410                 snprintf(buf1, sizeof(buf1), "auth zone %s: master lookup"
5411                         " for task_transfer", buf2);
5412                 log_query_info(VERB_ALGO, buf1, &qinfo);
5413         }
5414         edns.edns_present = 1;
5415         edns.ext_rcode = 0;
5416         edns.edns_version = 0;
5417         edns.bits = EDNS_DO;
5418         edns.opt_list_in = NULL;
5419         edns.opt_list_out = NULL;
5420         edns.opt_list_inplace_cb_out = NULL;
5421         edns.padding_block_size = 0;
5422         if(sldns_buffer_capacity(buf) < 65535)
5423                 edns.udp_size = (uint16_t)sldns_buffer_capacity(buf);
5424         else    edns.udp_size = 65535;
5425
5426         /* unlock xfr during mesh_new_callback() because the callback can be
5427          * called straight away */
5428         lock_basic_unlock(&xfr->lock);
5429         if(!mesh_new_callback(env->mesh, &qinfo, qflags, &edns, buf, 0,
5430                 &auth_xfer_transfer_lookup_callback, xfr, 0)) {
5431                 lock_basic_lock(&xfr->lock);
5432                 log_err("out of memory lookup up master %s", master->host);
5433                 return 0;
5434         }
5435         lock_basic_lock(&xfr->lock);
5436         return 1;
5437 }
5438
5439 /** initiate TCP to the target and fetch zone.
5440  * returns true if that was successfully started, and timeout setup. */
5441 static int
5442 xfr_transfer_init_fetch(struct auth_xfer* xfr, struct module_env* env)
5443 {
5444         struct sockaddr_storage addr;
5445         socklen_t addrlen = 0;
5446         struct auth_master* master = xfr->task_transfer->master;
5447         char *auth_name = NULL;
5448         struct timeval t;
5449         int timeout;
5450         if(!master) return 0;
5451         if(master->allow_notify) return 0; /* only for notify */
5452
5453         /* get master addr */
5454         if(xfr->task_transfer->scan_addr) {
5455                 addrlen = xfr->task_transfer->scan_addr->addrlen;
5456                 memmove(&addr, &xfr->task_transfer->scan_addr->addr, addrlen);
5457         } else {
5458                 if(!authextstrtoaddr(master->host, &addr, &addrlen, &auth_name)) {
5459                         /* the ones that are not in addr format are supposed
5460                          * to be looked up.  The lookup has failed however,
5461                          * so skip them */
5462                         char zname[255+1];
5463                         dname_str(xfr->name, zname);
5464                         log_err("%s: failed lookup, cannot transfer from master %s",
5465                                 zname, master->host);
5466                         return 0;
5467                 }
5468         }
5469
5470         /* remove previous TCP connection (if any) */
5471         if(xfr->task_transfer->cp) {
5472                 comm_point_delete(xfr->task_transfer->cp);
5473                 xfr->task_transfer->cp = NULL;
5474         }
5475         if(!xfr->task_transfer->timer) {
5476                 xfr->task_transfer->timer = comm_timer_create(env->worker_base,
5477                         auth_xfer_transfer_timer_callback, xfr);
5478                 if(!xfr->task_transfer->timer) {
5479                         log_err("malloc failure");
5480                         return 0;
5481                 }
5482         }
5483         timeout = AUTH_TRANSFER_TIMEOUT;
5484 #ifndef S_SPLINT_S
5485         t.tv_sec = timeout/1000;
5486         t.tv_usec = (timeout%1000)*1000;
5487 #endif
5488
5489         if(master->http) {
5490                 /* perform http fetch */
5491                 /* store http port number into sockaddr,
5492                  * unless someone used unbound's host@port notation */
5493                 xfr->task_transfer->on_ixfr = 0;
5494                 if(strchr(master->host, '@') == NULL)
5495                         sockaddr_store_port(&addr, addrlen, master->port);
5496                 xfr->task_transfer->cp = outnet_comm_point_for_http(
5497                         env->outnet, auth_xfer_transfer_http_callback, xfr,
5498                         &addr, addrlen, -1, master->ssl, master->host,
5499                         master->file, env->cfg);
5500                 if(!xfr->task_transfer->cp) {
5501                         char zname[255+1], as[256];
5502                         dname_str(xfr->name, zname);
5503                         addr_to_str(&addr, addrlen, as, sizeof(as));
5504                         verbose(VERB_ALGO, "cannot create http cp "
5505                                 "connection for %s to %s", zname, as);
5506                         return 0;
5507                 }
5508                 comm_timer_set(xfr->task_transfer->timer, &t);
5509                 if(verbosity >= VERB_ALGO) {
5510                         char zname[255+1], as[256];
5511                         dname_str(xfr->name, zname);
5512                         addr_to_str(&addr, addrlen, as, sizeof(as));
5513                         verbose(VERB_ALGO, "auth zone %s transfer next HTTP fetch from %s started", zname, as);
5514                 }
5515                 return 1;
5516         }
5517
5518         /* perform AXFR/IXFR */
5519         /* set the packet to be written */
5520         /* create new ID */
5521         xfr->task_transfer->id = GET_RANDOM_ID(env->rnd);
5522         xfr_create_ixfr_packet(xfr, env->scratch_buffer,
5523                 xfr->task_transfer->id, master);
5524
5525         /* connect on fd */
5526         xfr->task_transfer->cp = outnet_comm_point_for_tcp(env->outnet,
5527                 auth_xfer_transfer_tcp_callback, xfr, &addr, addrlen,
5528                 env->scratch_buffer, -1,
5529                 auth_name != NULL, auth_name);
5530         if(!xfr->task_transfer->cp) {
5531                 char zname[255+1], as[256];
5532                 dname_str(xfr->name, zname);
5533                 addr_to_str(&addr, addrlen, as, sizeof(as));
5534                 verbose(VERB_ALGO, "cannot create tcp cp connection for "
5535                         "xfr %s to %s", zname, as);
5536                 return 0;
5537         }
5538         comm_timer_set(xfr->task_transfer->timer, &t);
5539         if(verbosity >= VERB_ALGO) {
5540                 char zname[255+1], as[256];
5541                 dname_str(xfr->name, zname);
5542                 addr_to_str(&addr, addrlen, as, sizeof(as));
5543                 verbose(VERB_ALGO, "auth zone %s transfer next %s fetch from %s started", zname, 
5544                         (xfr->task_transfer->on_ixfr?"IXFR":"AXFR"), as);
5545         }
5546         return 1;
5547 }
5548
5549 /** perform next lookup, next transfer TCP, or end and resume wait time task */
5550 static void
5551 xfr_transfer_nexttarget_or_end(struct auth_xfer* xfr, struct module_env* env)
5552 {
5553         log_assert(xfr->task_transfer->worker == env->worker);
5554
5555         /* are we performing lookups? */
5556         while(xfr->task_transfer->lookup_target) {
5557                 if(xfr_transfer_lookup_host(xfr, env)) {
5558                         /* wait for lookup to finish,
5559                          * note that the hostname may be in unbound's cache
5560                          * and we may then get an instant cache response,
5561                          * and that calls the callback just like a full
5562                          * lookup and lookup failures also call callback */
5563                         if(verbosity >= VERB_ALGO) {
5564                                 char zname[255+1];
5565                                 dname_str(xfr->name, zname);
5566                                 verbose(VERB_ALGO, "auth zone %s transfer next target lookup", zname);
5567                         }
5568                         lock_basic_unlock(&xfr->lock);
5569                         return;
5570                 }
5571                 xfr_transfer_move_to_next_lookup(xfr, env);
5572         }
5573
5574         /* initiate TCP and fetch the zone from the master */
5575         /* and set timeout on it */
5576         while(!xfr_transfer_end_of_list(xfr)) {
5577                 xfr->task_transfer->master = xfr_transfer_current_master(xfr);
5578                 if(xfr_transfer_init_fetch(xfr, env)) {
5579                         /* successfully started, wait for callback */
5580                         lock_basic_unlock(&xfr->lock);
5581                         return;
5582                 }
5583                 /* failed to fetch, next master */
5584                 xfr_transfer_nextmaster(xfr);
5585         }
5586         if(verbosity >= VERB_ALGO) {
5587                 char zname[255+1];
5588                 dname_str(xfr->name, zname);
5589                 verbose(VERB_ALGO, "auth zone %s transfer failed, wait", zname);
5590         }
5591
5592         /* we failed to fetch the zone, move to wait task
5593          * use the shorter retry timeout */
5594         xfr_transfer_disown(xfr);
5595
5596         /* pick up the nextprobe task and wait */
5597         if(xfr->task_nextprobe->worker == NULL)
5598                 xfr_set_timeout(xfr, env, 1, 0);
5599         lock_basic_unlock(&xfr->lock);
5600 }
5601
5602 /** add addrs from A or AAAA rrset to the master */
5603 static void
5604 xfr_master_add_addrs(struct auth_master* m, struct ub_packed_rrset_key* rrset,
5605         uint16_t rrtype)
5606 {
5607         size_t i;
5608         struct packed_rrset_data* data;
5609         if(!m || !rrset) return;
5610         if(rrtype != LDNS_RR_TYPE_A && rrtype != LDNS_RR_TYPE_AAAA)
5611                 return;
5612         data = (struct packed_rrset_data*)rrset->entry.data;
5613         for(i=0; i<data->count; i++) {
5614                 struct auth_addr* a;
5615                 size_t len = data->rr_len[i] - 2;
5616                 uint8_t* rdata = data->rr_data[i]+2;
5617                 if(rrtype == LDNS_RR_TYPE_A && len != INET_SIZE)
5618                         continue; /* wrong length for A */
5619                 if(rrtype == LDNS_RR_TYPE_AAAA && len != INET6_SIZE)
5620                         continue; /* wrong length for AAAA */
5621                 
5622                 /* add and alloc it */
5623                 a = (struct auth_addr*)calloc(1, sizeof(*a));
5624                 if(!a) {
5625                         log_err("out of memory");
5626                         return;
5627                 }
5628                 if(rrtype == LDNS_RR_TYPE_A) {
5629                         struct sockaddr_in* sa;
5630                         a->addrlen = (socklen_t)sizeof(*sa);
5631                         sa = (struct sockaddr_in*)&a->addr;
5632                         sa->sin_family = AF_INET;
5633                         sa->sin_port = (in_port_t)htons(UNBOUND_DNS_PORT);
5634                         memmove(&sa->sin_addr, rdata, INET_SIZE);
5635                 } else {
5636                         struct sockaddr_in6* sa;
5637                         a->addrlen = (socklen_t)sizeof(*sa);
5638                         sa = (struct sockaddr_in6*)&a->addr;
5639                         sa->sin6_family = AF_INET6;
5640                         sa->sin6_port = (in_port_t)htons(UNBOUND_DNS_PORT);
5641                         memmove(&sa->sin6_addr, rdata, INET6_SIZE);
5642                 }
5643                 if(verbosity >= VERB_ALGO) {
5644                         char s[64];
5645                         addr_to_str(&a->addr, a->addrlen, s, sizeof(s));
5646                         verbose(VERB_ALGO, "auth host %s lookup %s",
5647                                 m->host, s);
5648                 }
5649                 /* append to list */
5650                 a->next = m->list;
5651                 m->list = a;
5652         }
5653 }
5654
5655 /** callback for task_transfer lookup of host name, of A or AAAA */
5656 void auth_xfer_transfer_lookup_callback(void* arg, int rcode, sldns_buffer* buf,
5657         enum sec_status ATTR_UNUSED(sec), char* ATTR_UNUSED(why_bogus),
5658         int ATTR_UNUSED(was_ratelimited))
5659 {
5660         struct auth_xfer* xfr = (struct auth_xfer*)arg;
5661         struct module_env* env;
5662         log_assert(xfr->task_transfer);
5663         lock_basic_lock(&xfr->lock);
5664         env = xfr->task_transfer->env;
5665         if(!env || env->outnet->want_to_quit) {
5666                 lock_basic_unlock(&xfr->lock);
5667                 return; /* stop on quit */
5668         }
5669
5670         /* process result */
5671         if(rcode == LDNS_RCODE_NOERROR) {
5672                 uint16_t wanted_qtype = LDNS_RR_TYPE_A;
5673                 struct regional* temp = env->scratch;
5674                 struct query_info rq;
5675                 struct reply_info* rep;
5676                 if(xfr->task_transfer->lookup_aaaa)
5677                         wanted_qtype = LDNS_RR_TYPE_AAAA;
5678                 memset(&rq, 0, sizeof(rq));
5679                 rep = parse_reply_in_temp_region(buf, temp, &rq);
5680                 if(rep && rq.qtype == wanted_qtype &&
5681                         FLAGS_GET_RCODE(rep->flags) == LDNS_RCODE_NOERROR) {
5682                         /* parsed successfully */
5683                         struct ub_packed_rrset_key* answer =
5684                                 reply_find_answer_rrset(&rq, rep);
5685                         if(answer) {
5686                                 xfr_master_add_addrs(xfr->task_transfer->
5687                                         lookup_target, answer, wanted_qtype);
5688                         } else {
5689                                 if(verbosity >= VERB_ALGO) {
5690                                         char zname[255+1];
5691                                         dname_str(xfr->name, zname);
5692                                         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"));
5693                                 }
5694                         }
5695                 } else {
5696                         if(verbosity >= VERB_ALGO) {
5697                                 char zname[255+1];
5698                                 dname_str(xfr->name, zname);
5699                                 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"));
5700                         }
5701                 }
5702                 regional_free_all(temp);
5703         } else {
5704                 if(verbosity >= VERB_ALGO) {
5705                         char zname[255+1];
5706                         dname_str(xfr->name, zname);
5707                         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"));
5708                 }
5709         }
5710         if(xfr->task_transfer->lookup_target->list &&
5711                 xfr->task_transfer->lookup_target == xfr_transfer_current_master(xfr))
5712                 xfr->task_transfer->scan_addr = xfr->task_transfer->lookup_target->list;
5713
5714         /* move to lookup AAAA after A lookup, move to next hostname lookup,
5715          * or move to fetch the zone, or, if nothing to do, end task_transfer */
5716         xfr_transfer_move_to_next_lookup(xfr, env);
5717         xfr_transfer_nexttarget_or_end(xfr, env);
5718 }
5719
5720 /** check if xfer (AXFR or IXFR) packet is OK.
5721  * return false if we lost connection (SERVFAIL, or unreadable).
5722  * return false if we need to move from IXFR to AXFR, with gonextonfail
5723  *      set to false, so the same master is tried again, but with AXFR.
5724  * return true if fine to link into data.
5725  * return true with transferdone=true when the transfer has ended.
5726  */
5727 static int
5728 check_xfer_packet(sldns_buffer* pkt, struct auth_xfer* xfr,
5729         int* gonextonfail, int* transferdone)
5730 {
5731         uint8_t* wire = sldns_buffer_begin(pkt);
5732         int i;
5733         if(sldns_buffer_limit(pkt) < LDNS_HEADER_SIZE) {
5734                 verbose(VERB_ALGO, "xfr to %s failed, packet too small",
5735                         xfr->task_transfer->master->host);
5736                 return 0;
5737         }
5738         if(!LDNS_QR_WIRE(wire)) {
5739                 verbose(VERB_ALGO, "xfr to %s failed, packet has no QR flag",
5740                         xfr->task_transfer->master->host);
5741                 return 0;
5742         }
5743         if(LDNS_TC_WIRE(wire)) {
5744                 verbose(VERB_ALGO, "xfr to %s failed, packet has TC flag",
5745                         xfr->task_transfer->master->host);
5746                 return 0;
5747         }
5748         /* check ID */
5749         if(LDNS_ID_WIRE(wire) != xfr->task_transfer->id) {
5750                 verbose(VERB_ALGO, "xfr to %s failed, packet wrong ID",
5751                         xfr->task_transfer->master->host);
5752                 return 0;
5753         }
5754         if(LDNS_RCODE_WIRE(wire) != LDNS_RCODE_NOERROR) {
5755                 char rcode[32];
5756                 sldns_wire2str_rcode_buf((int)LDNS_RCODE_WIRE(wire), rcode,
5757                         sizeof(rcode));
5758                 /* if we are doing IXFR, check for fallback */
5759                 if(xfr->task_transfer->on_ixfr) {
5760                         if(LDNS_RCODE_WIRE(wire) == LDNS_RCODE_NOTIMPL ||
5761                                 LDNS_RCODE_WIRE(wire) == LDNS_RCODE_SERVFAIL ||
5762                                 LDNS_RCODE_WIRE(wire) == LDNS_RCODE_REFUSED ||
5763                                 LDNS_RCODE_WIRE(wire) == LDNS_RCODE_FORMERR) {
5764                                 verbose(VERB_ALGO, "xfr to %s, fallback "
5765                                         "from IXFR to AXFR (with rcode %s)",
5766                                         xfr->task_transfer->master->host,
5767                                         rcode);
5768                                 xfr->task_transfer->ixfr_fail = 1;
5769                                 *gonextonfail = 0;
5770                                 return 0;
5771                         }
5772                 }
5773                 verbose(VERB_ALGO, "xfr to %s failed, packet with rcode %s",
5774                         xfr->task_transfer->master->host, rcode);
5775                 return 0;
5776         }
5777         if(LDNS_OPCODE_WIRE(wire) != LDNS_PACKET_QUERY) {
5778                 verbose(VERB_ALGO, "xfr to %s failed, packet with bad opcode",
5779                         xfr->task_transfer->master->host);
5780                 return 0;
5781         }
5782         if(LDNS_QDCOUNT(wire) > 1) {
5783                 verbose(VERB_ALGO, "xfr to %s failed, packet has qdcount %d",
5784                         xfr->task_transfer->master->host,
5785                         (int)LDNS_QDCOUNT(wire));
5786                 return 0;
5787         }
5788
5789         /* check qname */
5790         sldns_buffer_set_position(pkt, LDNS_HEADER_SIZE);
5791         for(i=0; i<(int)LDNS_QDCOUNT(wire); i++) {
5792                 size_t pos = sldns_buffer_position(pkt);
5793                 uint16_t qtype, qclass;
5794                 if(pkt_dname_len(pkt) == 0) {
5795                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5796                                 "malformed dname",
5797                                 xfr->task_transfer->master->host);
5798                         return 0;
5799                 }
5800                 if(dname_pkt_compare(pkt, sldns_buffer_at(pkt, pos),
5801                         xfr->name) != 0) {
5802                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5803                                 "wrong qname",
5804                                 xfr->task_transfer->master->host);
5805                         return 0;
5806                 }
5807                 if(sldns_buffer_remaining(pkt) < 4) {
5808                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5809                                 "truncated query RR",
5810                                 xfr->task_transfer->master->host);
5811                         return 0;
5812                 }
5813                 qtype = sldns_buffer_read_u16(pkt);
5814                 qclass = sldns_buffer_read_u16(pkt);
5815                 if(qclass != xfr->dclass) {
5816                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5817                                 "wrong qclass",
5818                                 xfr->task_transfer->master->host);
5819                         return 0;
5820                 }
5821                 if(xfr->task_transfer->on_ixfr) {
5822                         if(qtype != LDNS_RR_TYPE_IXFR) {
5823                                 verbose(VERB_ALGO, "xfr to %s failed, packet "
5824                                         "with wrong qtype, expected IXFR",
5825                                 xfr->task_transfer->master->host);
5826                                 return 0;
5827                         }
5828                 } else {
5829                         if(qtype != LDNS_RR_TYPE_AXFR) {
5830                                 verbose(VERB_ALGO, "xfr to %s failed, packet "
5831                                         "with wrong qtype, expected AXFR",
5832                                 xfr->task_transfer->master->host);
5833                                 return 0;
5834                         }
5835                 }
5836         }
5837
5838         /* check parse of RRs in packet, store first SOA serial
5839          * to be able to detect last SOA (with that serial) to see if done */
5840         /* also check for IXFR 'zone up to date' reply */
5841         for(i=0; i<(int)LDNS_ANCOUNT(wire); i++) {
5842                 size_t pos = sldns_buffer_position(pkt);
5843                 uint16_t tp, rdlen;
5844                 if(pkt_dname_len(pkt) == 0) {
5845                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5846                                 "malformed dname in answer section",
5847                                 xfr->task_transfer->master->host);
5848                         return 0;
5849                 }
5850                 if(sldns_buffer_remaining(pkt) < 10) {
5851                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5852                                 "truncated RR",
5853                                 xfr->task_transfer->master->host);
5854                         return 0;
5855                 }
5856                 tp = sldns_buffer_read_u16(pkt);
5857                 (void)sldns_buffer_read_u16(pkt); /* class */
5858                 (void)sldns_buffer_read_u32(pkt); /* ttl */
5859                 rdlen = sldns_buffer_read_u16(pkt);
5860                 if(sldns_buffer_remaining(pkt) < rdlen) {
5861                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5862                                 "truncated RR rdata",
5863                                 xfr->task_transfer->master->host);
5864                         return 0;
5865                 }
5866
5867                 /* RR parses (haven't checked rdata itself), now look at
5868                  * SOA records to see serial number */
5869                 if(xfr->task_transfer->rr_scan_num == 0 &&
5870                         tp != LDNS_RR_TYPE_SOA) {
5871                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5872                                 "malformed zone transfer, no start SOA",
5873                                 xfr->task_transfer->master->host);
5874                         return 0;
5875                 }
5876                 if(xfr->task_transfer->rr_scan_num == 1 &&
5877                         tp != LDNS_RR_TYPE_SOA) {
5878                         /* second RR is not a SOA record, this is not an IXFR
5879                          * the master is replying with an AXFR */
5880                         xfr->task_transfer->on_ixfr_is_axfr = 1;
5881                 }
5882                 if(tp == LDNS_RR_TYPE_SOA) {
5883                         uint32_t serial;
5884                         if(rdlen < 22) {
5885                                 verbose(VERB_ALGO, "xfr to %s failed, packet "
5886                                         "with SOA with malformed rdata",
5887                                         xfr->task_transfer->master->host);
5888                                 return 0;
5889                         }
5890                         if(dname_pkt_compare(pkt, sldns_buffer_at(pkt, pos),
5891                                 xfr->name) != 0) {
5892                                 verbose(VERB_ALGO, "xfr to %s failed, packet "
5893                                         "with SOA with wrong dname",
5894                                         xfr->task_transfer->master->host);
5895                                 return 0;
5896                         }
5897
5898                         /* read serial number of SOA */
5899                         serial = sldns_buffer_read_u32_at(pkt,
5900                                 sldns_buffer_position(pkt)+rdlen-20);
5901
5902                         /* check for IXFR 'zone has SOA x' reply */
5903                         if(xfr->task_transfer->on_ixfr &&
5904                                 xfr->task_transfer->rr_scan_num == 0 &&
5905                                 LDNS_ANCOUNT(wire)==1) {
5906                                 verbose(VERB_ALGO, "xfr to %s ended, "
5907                                         "IXFR reply that zone has serial %u,"
5908                                         " fallback from IXFR to AXFR",
5909                                         xfr->task_transfer->master->host,
5910                                         (unsigned)serial);
5911                                 xfr->task_transfer->ixfr_fail = 1;
5912                                 *gonextonfail = 0;
5913                                 return 0;
5914                         }
5915
5916                         /* if first SOA, store serial number */
5917                         if(xfr->task_transfer->got_xfr_serial == 0) {
5918                                 xfr->task_transfer->got_xfr_serial = 1;
5919                                 xfr->task_transfer->incoming_xfr_serial =
5920                                         serial;
5921                                 verbose(VERB_ALGO, "xfr %s: contains "
5922                                         "SOA serial %u",
5923                                         xfr->task_transfer->master->host,
5924                                         (unsigned)serial);
5925                         /* see if end of AXFR */
5926                         } else if(!xfr->task_transfer->on_ixfr ||
5927                                 xfr->task_transfer->on_ixfr_is_axfr) {
5928                                 /* second SOA with serial is the end
5929                                  * for AXFR */
5930                                 *transferdone = 1;
5931                                 verbose(VERB_ALGO, "xfr %s: last AXFR packet",
5932                                         xfr->task_transfer->master->host);
5933                         /* for IXFR, count SOA records with that serial */
5934                         } else if(xfr->task_transfer->incoming_xfr_serial ==
5935                                 serial && xfr->task_transfer->got_xfr_serial
5936                                 == 1) {
5937                                 xfr->task_transfer->got_xfr_serial++;
5938                         /* if not first soa, if serial==firstserial, the
5939                          * third time we are at the end, for IXFR */
5940                         } else if(xfr->task_transfer->incoming_xfr_serial ==
5941                                 serial && xfr->task_transfer->got_xfr_serial
5942                                 == 2) {
5943                                 verbose(VERB_ALGO, "xfr %s: last IXFR packet",
5944                                         xfr->task_transfer->master->host);
5945                                 *transferdone = 1;
5946                                 /* continue parse check, if that succeeds,
5947                                  * transfer is done */
5948                         }
5949                 }
5950                 xfr->task_transfer->rr_scan_num++;
5951
5952                 /* skip over RR rdata to go to the next RR */
5953                 sldns_buffer_skip(pkt, (ssize_t)rdlen);
5954         }
5955
5956         /* check authority section */
5957         /* we skip over the RRs checking packet format */
5958         for(i=0; i<(int)LDNS_NSCOUNT(wire); i++) {
5959                 uint16_t rdlen;
5960                 if(pkt_dname_len(pkt) == 0) {
5961                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5962                                 "malformed dname in authority section",
5963                                 xfr->task_transfer->master->host);
5964                         return 0;
5965                 }
5966                 if(sldns_buffer_remaining(pkt) < 10) {
5967                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5968                                 "truncated RR",
5969                                 xfr->task_transfer->master->host);
5970                         return 0;
5971                 }
5972                 (void)sldns_buffer_read_u16(pkt); /* type */
5973                 (void)sldns_buffer_read_u16(pkt); /* class */
5974                 (void)sldns_buffer_read_u32(pkt); /* ttl */
5975                 rdlen = sldns_buffer_read_u16(pkt);
5976                 if(sldns_buffer_remaining(pkt) < rdlen) {
5977                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5978                                 "truncated RR rdata",
5979                                 xfr->task_transfer->master->host);
5980                         return 0;
5981                 }
5982                 /* skip over RR rdata to go to the next RR */
5983                 sldns_buffer_skip(pkt, (ssize_t)rdlen);
5984         }
5985
5986         /* check additional section */
5987         for(i=0; i<(int)LDNS_ARCOUNT(wire); i++) {
5988                 uint16_t rdlen;
5989                 if(pkt_dname_len(pkt) == 0) {
5990                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5991                                 "malformed dname in additional section",
5992                                 xfr->task_transfer->master->host);
5993                         return 0;
5994                 }
5995                 if(sldns_buffer_remaining(pkt) < 10) {
5996                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5997                                 "truncated RR",
5998                                 xfr->task_transfer->master->host);
5999                         return 0;
6000                 }
6001                 (void)sldns_buffer_read_u16(pkt); /* type */
6002                 (void)sldns_buffer_read_u16(pkt); /* class */
6003                 (void)sldns_buffer_read_u32(pkt); /* ttl */
6004                 rdlen = sldns_buffer_read_u16(pkt);
6005                 if(sldns_buffer_remaining(pkt) < rdlen) {
6006                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
6007                                 "truncated RR rdata",
6008                                 xfr->task_transfer->master->host);
6009                         return 0;
6010                 }
6011                 /* skip over RR rdata to go to the next RR */
6012                 sldns_buffer_skip(pkt, (ssize_t)rdlen);
6013         }
6014
6015         return 1;
6016 }
6017
6018 /** Link the data from this packet into the worklist of transferred data */
6019 static int
6020 xfer_link_data(sldns_buffer* pkt, struct auth_xfer* xfr)
6021 {
6022         /* alloc it */
6023         struct auth_chunk* e;
6024         e = (struct auth_chunk*)calloc(1, sizeof(*e));
6025         if(!e) return 0;
6026         e->next = NULL;
6027         e->len = sldns_buffer_limit(pkt);
6028         e->data = memdup(sldns_buffer_begin(pkt), e->len);
6029         if(!e->data) {
6030                 free(e);
6031                 return 0;
6032         }
6033
6034         /* alloc succeeded, link into list */
6035         if(!xfr->task_transfer->chunks_first)
6036                 xfr->task_transfer->chunks_first = e;
6037         if(xfr->task_transfer->chunks_last)
6038                 xfr->task_transfer->chunks_last->next = e;
6039         xfr->task_transfer->chunks_last = e;
6040         return 1;
6041 }
6042
6043 /** task transfer.  the list of data is complete. process it and if failed
6044  * move to next master, if succeeded, end the task transfer */
6045 static void
6046 process_list_end_transfer(struct auth_xfer* xfr, struct module_env* env)
6047 {
6048         int ixfr_fail = 0;
6049         if(xfr_process_chunk_list(xfr, env, &ixfr_fail)) {
6050                 /* it worked! */
6051                 auth_chunks_delete(xfr->task_transfer);
6052
6053                 /* we fetched the zone, move to wait task */
6054                 xfr_transfer_disown(xfr);
6055
6056                 if(xfr->notify_received && (!xfr->notify_has_serial ||
6057                         (xfr->notify_has_serial && 
6058                         xfr_serial_means_update(xfr, xfr->notify_serial)))) {
6059                         uint32_t sr = xfr->notify_serial;
6060                         int has_sr = xfr->notify_has_serial;
6061                         /* we received a notify while probe/transfer was
6062                          * in progress.  start a new probe and transfer */
6063                         xfr->notify_received = 0;
6064                         xfr->notify_has_serial = 0;
6065                         xfr->notify_serial = 0;
6066                         if(!xfr_start_probe(xfr, env, NULL)) {
6067                                 /* if we couldn't start it, already in
6068                                  * progress; restore notify serial,
6069                                  * while xfr still locked */
6070                                 xfr->notify_received = 1;
6071                                 xfr->notify_has_serial = has_sr;
6072                                 xfr->notify_serial = sr;
6073                                 lock_basic_unlock(&xfr->lock);
6074                         }
6075                         return;
6076                 } else {
6077                         /* pick up the nextprobe task and wait (normail wait time) */
6078                         if(xfr->task_nextprobe->worker == NULL)
6079                                 xfr_set_timeout(xfr, env, 0, 0);
6080                 }
6081                 lock_basic_unlock(&xfr->lock);
6082                 return;
6083         }
6084         /* processing failed */
6085         /* when done, delete data from list */
6086         auth_chunks_delete(xfr->task_transfer);
6087         if(ixfr_fail) {
6088                 xfr->task_transfer->ixfr_fail = 1;
6089         } else {
6090                 xfr_transfer_nextmaster(xfr);
6091         }
6092         xfr_transfer_nexttarget_or_end(xfr, env);
6093 }
6094
6095 /** callback for the task_transfer timer */
6096 void
6097 auth_xfer_transfer_timer_callback(void* arg)
6098 {
6099         struct auth_xfer* xfr = (struct auth_xfer*)arg;
6100         struct module_env* env;
6101         int gonextonfail = 1;
6102         log_assert(xfr->task_transfer);
6103         lock_basic_lock(&xfr->lock);
6104         env = xfr->task_transfer->env;
6105         if(!env || env->outnet->want_to_quit) {
6106                 lock_basic_unlock(&xfr->lock);
6107                 return; /* stop on quit */
6108         }
6109
6110         verbose(VERB_ALGO, "xfr stopped, connection timeout to %s",
6111                 xfr->task_transfer->master->host);
6112
6113         /* see if IXFR caused the failure, if so, try AXFR */
6114         if(xfr->task_transfer->on_ixfr) {
6115                 xfr->task_transfer->ixfr_possible_timeout_count++;
6116                 if(xfr->task_transfer->ixfr_possible_timeout_count >=
6117                         NUM_TIMEOUTS_FALLBACK_IXFR) {
6118                         verbose(VERB_ALGO, "xfr to %s, fallback "
6119                                 "from IXFR to AXFR (because of timeouts)",
6120                                 xfr->task_transfer->master->host);
6121                         xfr->task_transfer->ixfr_fail = 1;
6122                         gonextonfail = 0;
6123                 }
6124         }
6125
6126         /* delete transferred data from list */
6127         auth_chunks_delete(xfr->task_transfer);
6128         comm_point_delete(xfr->task_transfer->cp);
6129         xfr->task_transfer->cp = NULL;
6130         if(gonextonfail)
6131                 xfr_transfer_nextmaster(xfr);
6132         xfr_transfer_nexttarget_or_end(xfr, env);
6133 }
6134
6135 /** callback for task_transfer tcp connections */
6136 int
6137 auth_xfer_transfer_tcp_callback(struct comm_point* c, void* arg, int err,
6138         struct comm_reply* ATTR_UNUSED(repinfo))
6139 {
6140         struct auth_xfer* xfr = (struct auth_xfer*)arg;
6141         struct module_env* env;
6142         int gonextonfail = 1;
6143         int transferdone = 0;
6144         log_assert(xfr->task_transfer);
6145         lock_basic_lock(&xfr->lock);
6146         env = xfr->task_transfer->env;
6147         if(!env || env->outnet->want_to_quit) {
6148                 lock_basic_unlock(&xfr->lock);
6149                 return 0; /* stop on quit */
6150         }
6151         /* stop the timer */
6152         comm_timer_disable(xfr->task_transfer->timer);
6153
6154         if(err != NETEVENT_NOERROR) {
6155                 /* connection failed, closed, or timeout */
6156                 /* stop this transfer, cleanup 
6157                  * and continue task_transfer*/
6158                 verbose(VERB_ALGO, "xfr stopped, connection lost to %s",
6159                         xfr->task_transfer->master->host);
6160
6161                 /* see if IXFR caused the failure, if so, try AXFR */
6162                 if(xfr->task_transfer->on_ixfr) {
6163                         xfr->task_transfer->ixfr_possible_timeout_count++;
6164                         if(xfr->task_transfer->ixfr_possible_timeout_count >=
6165                                 NUM_TIMEOUTS_FALLBACK_IXFR) {
6166                                 verbose(VERB_ALGO, "xfr to %s, fallback "
6167                                         "from IXFR to AXFR (because of timeouts)",
6168                                         xfr->task_transfer->master->host);
6169                                 xfr->task_transfer->ixfr_fail = 1;
6170                                 gonextonfail = 0;
6171                         }
6172                 }
6173
6174         failed:
6175                 /* delete transferred data from list */
6176                 auth_chunks_delete(xfr->task_transfer);
6177                 comm_point_delete(xfr->task_transfer->cp);
6178                 xfr->task_transfer->cp = NULL;
6179                 if(gonextonfail)
6180                         xfr_transfer_nextmaster(xfr);
6181                 xfr_transfer_nexttarget_or_end(xfr, env);
6182                 return 0;
6183         }
6184         /* note that IXFR worked without timeout */
6185         if(xfr->task_transfer->on_ixfr)
6186                 xfr->task_transfer->ixfr_possible_timeout_count = 0;
6187
6188         /* handle returned packet */
6189         /* if it fails, cleanup and end this transfer */
6190         /* if it needs to fallback from IXFR to AXFR, do that */
6191         if(!check_xfer_packet(c->buffer, xfr, &gonextonfail, &transferdone)) {
6192                 goto failed;
6193         }
6194         /* if it is good, link it into the list of data */
6195         /* if the link into list of data fails (malloc fail) cleanup and end */
6196         if(!xfer_link_data(c->buffer, xfr)) {
6197                 verbose(VERB_ALGO, "xfr stopped to %s, malloc failed",
6198                         xfr->task_transfer->master->host);
6199                 goto failed;
6200         }
6201         /* if the transfer is done now, disconnect and process the list */
6202         if(transferdone) {
6203                 comm_point_delete(xfr->task_transfer->cp);
6204                 xfr->task_transfer->cp = NULL;
6205                 process_list_end_transfer(xfr, env);
6206                 return 0;
6207         }
6208
6209         /* if we want to read more messages, setup the commpoint to read
6210          * a DNS packet, and the timeout */
6211         lock_basic_unlock(&xfr->lock);
6212         c->tcp_is_reading = 1;
6213         sldns_buffer_clear(c->buffer);
6214         comm_point_start_listening(c, -1, AUTH_TRANSFER_TIMEOUT);
6215         return 0;
6216 }
6217
6218 /** callback for task_transfer http connections */
6219 int
6220 auth_xfer_transfer_http_callback(struct comm_point* c, void* arg, int err,
6221         struct comm_reply* repinfo)
6222 {
6223         struct auth_xfer* xfr = (struct auth_xfer*)arg;
6224         struct module_env* env;
6225         log_assert(xfr->task_transfer);
6226         lock_basic_lock(&xfr->lock);
6227         env = xfr->task_transfer->env;
6228         if(!env || env->outnet->want_to_quit) {
6229                 lock_basic_unlock(&xfr->lock);
6230                 return 0; /* stop on quit */
6231         }
6232         verbose(VERB_ALGO, "auth zone transfer http callback");
6233         /* stop the timer */
6234         comm_timer_disable(xfr->task_transfer->timer);
6235
6236         if(err != NETEVENT_NOERROR && err != NETEVENT_DONE) {
6237                 /* connection failed, closed, or timeout */
6238                 /* stop this transfer, cleanup 
6239                  * and continue task_transfer*/
6240                 verbose(VERB_ALGO, "http stopped, connection lost to %s",
6241                         xfr->task_transfer->master->host);
6242         failed:
6243                 /* delete transferred data from list */
6244                 auth_chunks_delete(xfr->task_transfer);
6245                 if(repinfo) repinfo->c = NULL; /* signal cp deleted to
6246                                 the routine calling this callback */
6247                 comm_point_delete(xfr->task_transfer->cp);
6248                 xfr->task_transfer->cp = NULL;
6249                 xfr_transfer_nextmaster(xfr);
6250                 xfr_transfer_nexttarget_or_end(xfr, env);
6251                 return 0;
6252         }
6253
6254         /* if it is good, link it into the list of data */
6255         /* if the link into list of data fails (malloc fail) cleanup and end */
6256         if(sldns_buffer_limit(c->buffer) > 0) {
6257                 verbose(VERB_ALGO, "auth zone http queued up %d bytes",
6258                         (int)sldns_buffer_limit(c->buffer));
6259                 if(!xfer_link_data(c->buffer, xfr)) {
6260                         verbose(VERB_ALGO, "http stopped to %s, malloc failed",
6261                                 xfr->task_transfer->master->host);
6262                         goto failed;
6263                 }
6264         }
6265         /* if the transfer is done now, disconnect and process the list */
6266         if(err == NETEVENT_DONE) {
6267                 if(repinfo) repinfo->c = NULL; /* signal cp deleted to
6268                                 the routine calling this callback */
6269                 comm_point_delete(xfr->task_transfer->cp);
6270                 xfr->task_transfer->cp = NULL;
6271                 process_list_end_transfer(xfr, env);
6272                 return 0;
6273         }
6274
6275         /* if we want to read more messages, setup the commpoint to read
6276          * a DNS packet, and the timeout */
6277         lock_basic_unlock(&xfr->lock);
6278         c->tcp_is_reading = 1;
6279         sldns_buffer_clear(c->buffer);
6280         comm_point_start_listening(c, -1, AUTH_TRANSFER_TIMEOUT);
6281         return 0;
6282 }
6283
6284
6285 /** start transfer task by this worker , xfr is locked. */
6286 static void
6287 xfr_start_transfer(struct auth_xfer* xfr, struct module_env* env,
6288         struct auth_master* master)
6289 {
6290         log_assert(xfr->task_transfer != NULL);
6291         log_assert(xfr->task_transfer->worker == NULL);
6292         log_assert(xfr->task_transfer->chunks_first == NULL);
6293         log_assert(xfr->task_transfer->chunks_last == NULL);
6294         xfr->task_transfer->worker = env->worker;
6295         xfr->task_transfer->env = env;
6296
6297         /* init transfer process */
6298         /* find that master in the transfer's list of masters? */
6299         xfr_transfer_start_list(xfr, master);
6300         /* start lookup for hostnames in transfer master list */
6301         xfr_transfer_start_lookups(xfr);
6302
6303         /* initiate TCP, and set timeout on it */
6304         xfr_transfer_nexttarget_or_end(xfr, env);
6305 }
6306
6307 /** disown task_probe.  caller must hold xfr.lock */
6308 static void
6309 xfr_probe_disown(struct auth_xfer* xfr)
6310 {
6311         /* remove timer (from this worker's event base) */
6312         comm_timer_delete(xfr->task_probe->timer);
6313         xfr->task_probe->timer = NULL;
6314         /* remove the commpoint */
6315         comm_point_delete(xfr->task_probe->cp);
6316         xfr->task_probe->cp = NULL;
6317         /* we don't own this item anymore */
6318         xfr->task_probe->worker = NULL;
6319         xfr->task_probe->env = NULL;
6320 }
6321
6322 /** send the UDP probe to the master, this is part of task_probe */
6323 static int
6324 xfr_probe_send_probe(struct auth_xfer* xfr, struct module_env* env,
6325         int timeout)
6326 {
6327         struct sockaddr_storage addr;
6328         socklen_t addrlen = 0;
6329         struct timeval t;
6330         /* pick master */
6331         struct auth_master* master = xfr_probe_current_master(xfr);
6332         char *auth_name = NULL;
6333         if(!master) return 0;
6334         if(master->allow_notify) return 0; /* only for notify */
6335         if(master->http) return 0; /* only masters get SOA UDP probe,
6336                 not urls, if those are in this list */
6337
6338         /* get master addr */
6339         if(xfr->task_probe->scan_addr) {
6340                 addrlen = xfr->task_probe->scan_addr->addrlen;
6341                 memmove(&addr, &xfr->task_probe->scan_addr->addr, addrlen);
6342         } else {
6343                 if(!authextstrtoaddr(master->host, &addr, &addrlen, &auth_name)) {
6344                         /* the ones that are not in addr format are supposed
6345                          * to be looked up.  The lookup has failed however,
6346                          * so skip them */
6347                         char zname[255+1];
6348                         dname_str(xfr->name, zname);
6349                         log_err("%s: failed lookup, cannot probe to master %s",
6350                                 zname, master->host);
6351                         return 0;
6352                 }
6353                 if (auth_name != NULL) {
6354                         if (addr.ss_family == AF_INET
6355                         &&  (int)ntohs(((struct sockaddr_in *)&addr)->sin_port)
6356                             == env->cfg->ssl_port)
6357                                 ((struct sockaddr_in *)&addr)->sin_port
6358                                         = htons((uint16_t)env->cfg->port);
6359                         else if (addr.ss_family == AF_INET6
6360                         &&  (int)ntohs(((struct sockaddr_in6 *)&addr)->sin6_port)
6361                             == env->cfg->ssl_port)
6362                                 ((struct sockaddr_in6 *)&addr)->sin6_port
6363                                         = htons((uint16_t)env->cfg->port);
6364                 }
6365         }
6366
6367         /* create packet */
6368         /* create new ID for new probes, but not on timeout retries,
6369          * this means we'll accept replies to previous retries to same ip */
6370         if(timeout == AUTH_PROBE_TIMEOUT)
6371                 xfr->task_probe->id = GET_RANDOM_ID(env->rnd);
6372         xfr_create_soa_probe_packet(xfr, env->scratch_buffer, 
6373                 xfr->task_probe->id);
6374         /* we need to remove the cp if we have a different ip4/ip6 type now */
6375         if(xfr->task_probe->cp &&
6376                 ((xfr->task_probe->cp_is_ip6 && !addr_is_ip6(&addr, addrlen)) ||
6377                 (!xfr->task_probe->cp_is_ip6 && addr_is_ip6(&addr, addrlen)))
6378                 ) {
6379                 comm_point_delete(xfr->task_probe->cp);
6380                 xfr->task_probe->cp = NULL;
6381         }
6382         if(!xfr->task_probe->cp) {
6383                 if(addr_is_ip6(&addr, addrlen))
6384                         xfr->task_probe->cp_is_ip6 = 1;
6385                 else    xfr->task_probe->cp_is_ip6 = 0;
6386                 xfr->task_probe->cp = outnet_comm_point_for_udp(env->outnet,
6387                         auth_xfer_probe_udp_callback, xfr, &addr, addrlen);
6388                 if(!xfr->task_probe->cp) {
6389                         char zname[255+1], as[256];
6390                         dname_str(xfr->name, zname);
6391                         addr_to_str(&addr, addrlen, as, sizeof(as));
6392                         verbose(VERB_ALGO, "cannot create udp cp for "
6393                                 "probe %s to %s", zname, as);
6394                         return 0;
6395                 }
6396         }
6397         if(!xfr->task_probe->timer) {
6398                 xfr->task_probe->timer = comm_timer_create(env->worker_base,
6399                         auth_xfer_probe_timer_callback, xfr);
6400                 if(!xfr->task_probe->timer) {
6401                         log_err("malloc failure");
6402                         return 0;
6403                 }
6404         }
6405
6406         /* send udp packet */
6407         if(!comm_point_send_udp_msg(xfr->task_probe->cp, env->scratch_buffer,
6408                 (struct sockaddr*)&addr, addrlen, 0)) {
6409                 char zname[255+1], as[256];
6410                 dname_str(xfr->name, zname);
6411                 addr_to_str(&addr, addrlen, as, sizeof(as));
6412                 verbose(VERB_ALGO, "failed to send soa probe for %s to %s",
6413                         zname, as);
6414                 return 0;
6415         }
6416         if(verbosity >= VERB_ALGO) {
6417                 char zname[255+1], as[256];
6418                 dname_str(xfr->name, zname);
6419                 addr_to_str(&addr, addrlen, as, sizeof(as));
6420                 verbose(VERB_ALGO, "auth zone %s soa probe sent to %s", zname,
6421                         as);
6422         }
6423         xfr->task_probe->timeout = timeout;
6424 #ifndef S_SPLINT_S
6425         t.tv_sec = timeout/1000;
6426         t.tv_usec = (timeout%1000)*1000;
6427 #endif
6428         comm_timer_set(xfr->task_probe->timer, &t);
6429
6430         return 1;
6431 }
6432
6433 /** callback for task_probe timer */
6434 void
6435 auth_xfer_probe_timer_callback(void* arg)
6436 {
6437         struct auth_xfer* xfr = (struct auth_xfer*)arg;
6438         struct module_env* env;
6439         log_assert(xfr->task_probe);
6440         lock_basic_lock(&xfr->lock);
6441         env = xfr->task_probe->env;
6442         if(!env || env->outnet->want_to_quit) {
6443                 lock_basic_unlock(&xfr->lock);
6444                 return; /* stop on quit */
6445         }
6446
6447         if(verbosity >= VERB_ALGO) {
6448                 char zname[255+1];
6449                 dname_str(xfr->name, zname);
6450                 verbose(VERB_ALGO, "auth zone %s soa probe timeout", zname);
6451         }
6452         if(xfr->task_probe->timeout <= AUTH_PROBE_TIMEOUT_STOP) {
6453                 /* try again with bigger timeout */
6454                 if(xfr_probe_send_probe(xfr, env, xfr->task_probe->timeout*2)) {
6455                         lock_basic_unlock(&xfr->lock);
6456                         return;
6457                 }
6458         }
6459         /* delete commpoint so a new one is created, with a fresh port nr */
6460         comm_point_delete(xfr->task_probe->cp);
6461         xfr->task_probe->cp = NULL;
6462
6463         /* too many timeouts (or fail to send), move to next or end */
6464         xfr_probe_nextmaster(xfr);
6465         xfr_probe_send_or_end(xfr, env);
6466 }
6467
6468 /** callback for task_probe udp packets */
6469 int
6470 auth_xfer_probe_udp_callback(struct comm_point* c, void* arg, int err,
6471         struct comm_reply* repinfo)
6472 {
6473         struct auth_xfer* xfr = (struct auth_xfer*)arg;
6474         struct module_env* env;
6475         log_assert(xfr->task_probe);
6476         lock_basic_lock(&xfr->lock);
6477         env = xfr->task_probe->env;
6478         if(!env || env->outnet->want_to_quit) {
6479                 lock_basic_unlock(&xfr->lock);
6480                 return 0; /* stop on quit */
6481         }
6482
6483         /* the comm_point_udp_callback is in a for loop for NUM_UDP_PER_SELECT
6484          * and we set rep.c=NULL to stop if from looking inside the commpoint*/
6485         repinfo->c = NULL;
6486         /* stop the timer */
6487         comm_timer_disable(xfr->task_probe->timer);
6488
6489         /* see if we got a packet and what that means */
6490         if(err == NETEVENT_NOERROR) {
6491                 uint32_t serial = 0;
6492                 if(check_packet_ok(c->buffer, LDNS_RR_TYPE_SOA, xfr,
6493                         &serial)) {
6494                         /* successful lookup */
6495                         if(verbosity >= VERB_ALGO) {
6496                                 char buf[256];
6497                                 dname_str(xfr->name, buf);
6498                                 verbose(VERB_ALGO, "auth zone %s: soa probe "
6499                                         "serial is %u", buf, (unsigned)serial);
6500                         }
6501                         /* see if this serial indicates that the zone has
6502                          * to be updated */
6503                         if(xfr_serial_means_update(xfr, serial)) {
6504                                 /* if updated, start the transfer task, if needed */
6505                                 verbose(VERB_ALGO, "auth_zone updated, start transfer");
6506                                 if(xfr->task_transfer->worker == NULL) {
6507                                         struct auth_master* master =
6508                                                 xfr_probe_current_master(xfr);
6509                                         /* if we have download URLs use them
6510                                          * in preference to this master we
6511                                          * just probed the SOA from */
6512                                         if(xfr->task_transfer->masters &&
6513                                                 xfr->task_transfer->masters->http)
6514                                                 master = NULL;
6515                                         xfr_probe_disown(xfr);
6516                                         xfr_start_transfer(xfr, env, master);
6517                                         return 0;
6518
6519                                 }
6520                                 /* other tasks are running, we don't do this anymore */
6521                                 xfr_probe_disown(xfr);
6522                                 lock_basic_unlock(&xfr->lock);
6523                                 /* return, we don't sent a reply to this udp packet,
6524                                  * and we setup the tasks to do next */
6525                                 return 0;
6526                         } else {
6527                                 verbose(VERB_ALGO, "auth_zone master reports unchanged soa serial");
6528                                 /* we if cannot find updates amongst the
6529                                  * masters, this means we then have a new lease
6530                                  * on the zone */
6531                                 xfr->task_probe->have_new_lease = 1;
6532                         }
6533                 } else {
6534                         if(verbosity >= VERB_ALGO) {
6535                                 char buf[256];
6536                                 dname_str(xfr->name, buf);
6537                                 verbose(VERB_ALGO, "auth zone %s: bad reply to soa probe", buf);
6538                         }
6539                 }
6540         } else {
6541                 if(verbosity >= VERB_ALGO) {
6542                         char buf[256];
6543                         dname_str(xfr->name, buf);
6544                         verbose(VERB_ALGO, "auth zone %s: soa probe failed", buf);
6545                 }
6546         }
6547         
6548         /* failed lookup or not an update */
6549         /* delete commpoint so a new one is created, with a fresh port nr */
6550         comm_point_delete(xfr->task_probe->cp);
6551         xfr->task_probe->cp = NULL;
6552
6553         /* if the result was not a successful probe, we need
6554          * to send the next one */
6555         xfr_probe_nextmaster(xfr);
6556         xfr_probe_send_or_end(xfr, env);
6557         return 0;
6558 }
6559
6560 /** lookup a host name for its addresses, if needed */
6561 static int
6562 xfr_probe_lookup_host(struct auth_xfer* xfr, struct module_env* env)
6563 {
6564         struct sockaddr_storage addr;
6565         socklen_t addrlen = 0;
6566         struct auth_master* master = xfr->task_probe->lookup_target;
6567         struct query_info qinfo;
6568         uint16_t qflags = BIT_RD;
6569         uint8_t dname[LDNS_MAX_DOMAINLEN+1];
6570         struct edns_data edns;
6571         sldns_buffer* buf = env->scratch_buffer;
6572         if(!master) return 0;
6573         if(extstrtoaddr(master->host, &addr, &addrlen)) {
6574                 /* not needed, host is in IP addr format */
6575                 return 0;
6576         }
6577         if(master->allow_notify && !master->http &&
6578                 strchr(master->host, '/') != NULL &&
6579                 strchr(master->host, '/') == strrchr(master->host, '/')) {
6580                 return 0; /* is IP/prefix format, not something to look up */
6581         }
6582
6583         /* use mesh_new_callback to probe for non-addr hosts,
6584          * and then wait for them to be looked up (in cache, or query) */
6585         qinfo.qname_len = sizeof(dname);
6586         if(sldns_str2wire_dname_buf(master->host, dname, &qinfo.qname_len)
6587                 != 0) {
6588                 log_err("cannot parse host name of master %s", master->host);
6589                 return 0;
6590         }
6591         qinfo.qname = dname;
6592         qinfo.qclass = xfr->dclass;
6593         qinfo.qtype = LDNS_RR_TYPE_A;
6594         if(xfr->task_probe->lookup_aaaa)
6595                 qinfo.qtype = LDNS_RR_TYPE_AAAA;
6596         qinfo.local_alias = NULL;
6597         if(verbosity >= VERB_ALGO) {
6598                 char buf1[512];
6599                 char buf2[LDNS_MAX_DOMAINLEN+1];
6600                 dname_str(xfr->name, buf2);
6601                 snprintf(buf1, sizeof(buf1), "auth zone %s: master lookup"
6602                         " for task_probe", buf2);
6603                 log_query_info(VERB_ALGO, buf1, &qinfo);
6604         }
6605         edns.edns_present = 1;
6606         edns.ext_rcode = 0;
6607         edns.edns_version = 0;
6608         edns.bits = EDNS_DO;
6609         edns.opt_list_in = NULL;
6610         edns.opt_list_out = NULL;
6611         edns.opt_list_inplace_cb_out = NULL;
6612         edns.padding_block_size = 0;
6613         if(sldns_buffer_capacity(buf) < 65535)
6614                 edns.udp_size = (uint16_t)sldns_buffer_capacity(buf);
6615         else    edns.udp_size = 65535;
6616
6617         /* unlock xfr during mesh_new_callback() because the callback can be
6618          * called straight away */
6619         lock_basic_unlock(&xfr->lock);
6620         if(!mesh_new_callback(env->mesh, &qinfo, qflags, &edns, buf, 0,
6621                 &auth_xfer_probe_lookup_callback, xfr, 0)) {
6622                 lock_basic_lock(&xfr->lock);
6623                 log_err("out of memory lookup up master %s", master->host);
6624                 return 0;
6625         }
6626         lock_basic_lock(&xfr->lock);
6627         return 1;
6628 }
6629
6630 /** move to sending the probe packets, next if fails. task_probe */
6631 static void
6632 xfr_probe_send_or_end(struct auth_xfer* xfr, struct module_env* env)
6633 {
6634         /* are we doing hostname lookups? */
6635         while(xfr->task_probe->lookup_target) {
6636                 if(xfr_probe_lookup_host(xfr, env)) {
6637                         /* wait for lookup to finish,
6638                          * note that the hostname may be in unbound's cache
6639                          * and we may then get an instant cache response,
6640                          * and that calls the callback just like a full
6641                          * lookup and lookup failures also call callback */
6642                         if(verbosity >= VERB_ALGO) {
6643                                 char zname[255+1];
6644                                 dname_str(xfr->name, zname);
6645                                 verbose(VERB_ALGO, "auth zone %s probe next target lookup", zname);
6646                         }
6647                         lock_basic_unlock(&xfr->lock);
6648                         return;
6649                 }
6650                 xfr_probe_move_to_next_lookup(xfr, env);
6651         }
6652         /* probe of list has ended.  Create or refresh the list of of
6653          * allow_notify addrs */
6654         probe_copy_masters_for_allow_notify(xfr);
6655         if(verbosity >= VERB_ALGO) {
6656                 char zname[255+1];
6657                 dname_str(xfr->name, zname);
6658                 verbose(VERB_ALGO, "auth zone %s probe: notify addrs updated", zname);
6659         }
6660         if(xfr->task_probe->only_lookup) {
6661                 /* only wanted lookups for copy, stop probe and start wait */
6662                 xfr->task_probe->only_lookup = 0;
6663                 if(verbosity >= VERB_ALGO) {
6664                         char zname[255+1];
6665                         dname_str(xfr->name, zname);
6666                         verbose(VERB_ALGO, "auth zone %s probe: finished only_lookup", zname);
6667                 }
6668                 xfr_probe_disown(xfr);
6669                 if(xfr->task_nextprobe->worker == NULL)
6670                         xfr_set_timeout(xfr, env, 0, 0);
6671                 lock_basic_unlock(&xfr->lock);
6672                 return;
6673         }
6674
6675         /* send probe packets */
6676         while(!xfr_probe_end_of_list(xfr)) {
6677                 if(xfr_probe_send_probe(xfr, env, AUTH_PROBE_TIMEOUT)) {
6678                         /* successfully sent probe, wait for callback */
6679                         lock_basic_unlock(&xfr->lock);
6680                         return;
6681                 }
6682                 /* failed to send probe, next master */
6683                 xfr_probe_nextmaster(xfr);
6684         }
6685
6686         /* done with probe sequence, wait */
6687         if(xfr->task_probe->have_new_lease) {
6688                 /* if zone not updated, start the wait timer again */
6689                 if(verbosity >= VERB_ALGO) {
6690                         char zname[255+1];
6691                         dname_str(xfr->name, zname);
6692                         verbose(VERB_ALGO, "auth_zone %s unchanged, new lease, wait", zname);
6693                 }
6694                 xfr_probe_disown(xfr);
6695                 if(xfr->have_zone)
6696                         xfr->lease_time = *env->now;
6697                 if(xfr->task_nextprobe->worker == NULL)
6698                         xfr_set_timeout(xfr, env, 0, 0);
6699         } else {
6700                 if(verbosity >= VERB_ALGO) {
6701                         char zname[255+1];
6702                         dname_str(xfr->name, zname);
6703                         verbose(VERB_ALGO, "auth zone %s soa probe failed, wait to retry", zname);
6704                 }
6705                 /* we failed to send this as well, move to the wait task,
6706                  * use the shorter retry timeout */
6707                 xfr_probe_disown(xfr);
6708                 /* pick up the nextprobe task and wait */
6709                 if(xfr->task_nextprobe->worker == NULL)
6710                         xfr_set_timeout(xfr, env, 1, 0);
6711         }
6712
6713         lock_basic_unlock(&xfr->lock);
6714 }
6715
6716 /** callback for task_probe lookup of host name, of A or AAAA */
6717 void auth_xfer_probe_lookup_callback(void* arg, int rcode, sldns_buffer* buf,
6718         enum sec_status ATTR_UNUSED(sec), char* ATTR_UNUSED(why_bogus),
6719         int ATTR_UNUSED(was_ratelimited))
6720 {
6721         struct auth_xfer* xfr = (struct auth_xfer*)arg;
6722         struct module_env* env;
6723         log_assert(xfr->task_probe);
6724         lock_basic_lock(&xfr->lock);
6725         env = xfr->task_probe->env;
6726         if(!env || env->outnet->want_to_quit) {
6727                 lock_basic_unlock(&xfr->lock);
6728                 return; /* stop on quit */
6729         }
6730
6731         /* process result */
6732         if(rcode == LDNS_RCODE_NOERROR) {
6733                 uint16_t wanted_qtype = LDNS_RR_TYPE_A;
6734                 struct regional* temp = env->scratch;
6735                 struct query_info rq;
6736                 struct reply_info* rep;
6737                 if(xfr->task_probe->lookup_aaaa)
6738                         wanted_qtype = LDNS_RR_TYPE_AAAA;
6739                 memset(&rq, 0, sizeof(rq));
6740                 rep = parse_reply_in_temp_region(buf, temp, &rq);
6741                 if(rep && rq.qtype == wanted_qtype &&
6742                         FLAGS_GET_RCODE(rep->flags) == LDNS_RCODE_NOERROR) {
6743                         /* parsed successfully */
6744                         struct ub_packed_rrset_key* answer =
6745                                 reply_find_answer_rrset(&rq, rep);
6746                         if(answer) {
6747                                 xfr_master_add_addrs(xfr->task_probe->
6748                                         lookup_target, answer, wanted_qtype);
6749                         } else {
6750                                 if(verbosity >= VERB_ALGO) {
6751                                         char zname[255+1];
6752                                         dname_str(xfr->name, zname);
6753                                         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"));
6754                                 }
6755                         }
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 no address", zname, xfr->task_probe->lookup_target->host, (xfr->task_probe->lookup_aaaa?"AAAA":"A"));
6761                         }
6762                 }
6763                 regional_free_all(temp);
6764         } else {
6765                 if(verbosity >= VERB_ALGO) {
6766                         char zname[255+1];
6767                         dname_str(xfr->name, zname);
6768                         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"));
6769                 }
6770         }
6771         if(xfr->task_probe->lookup_target->list &&
6772                 xfr->task_probe->lookup_target == xfr_probe_current_master(xfr))
6773                 xfr->task_probe->scan_addr = xfr->task_probe->lookup_target->list;
6774
6775         /* move to lookup AAAA after A lookup, move to next hostname lookup,
6776          * or move to send the probes, or, if nothing to do, end task_probe */
6777         xfr_probe_move_to_next_lookup(xfr, env);
6778         xfr_probe_send_or_end(xfr, env);
6779 }
6780
6781 /** disown task_nextprobe.  caller must hold xfr.lock */
6782 static void
6783 xfr_nextprobe_disown(struct auth_xfer* xfr)
6784 {
6785         /* delete the timer, because the next worker to pick this up may
6786          * not have the same event base */
6787         comm_timer_delete(xfr->task_nextprobe->timer);
6788         xfr->task_nextprobe->timer = NULL;
6789         xfr->task_nextprobe->next_probe = 0;
6790         /* we don't own this item anymore */
6791         xfr->task_nextprobe->worker = NULL;
6792         xfr->task_nextprobe->env = NULL;
6793 }
6794
6795 /** xfer nextprobe timeout callback, this is part of task_nextprobe */
6796 void
6797 auth_xfer_timer(void* arg)
6798 {
6799         struct auth_xfer* xfr = (struct auth_xfer*)arg;
6800         struct module_env* env;
6801         log_assert(xfr->task_nextprobe);
6802         lock_basic_lock(&xfr->lock);
6803         env = xfr->task_nextprobe->env;
6804         if(!env || env->outnet->want_to_quit) {
6805                 lock_basic_unlock(&xfr->lock);
6806                 return; /* stop on quit */
6807         }
6808
6809         /* see if zone has expired, and if so, also set auth_zone expired */
6810         if(xfr->have_zone && !xfr->zone_expired &&
6811            *env->now >= xfr->lease_time + xfr->expiry) {
6812                 lock_basic_unlock(&xfr->lock);
6813                 auth_xfer_set_expired(xfr, env, 1);
6814                 lock_basic_lock(&xfr->lock);
6815         }
6816
6817         xfr_nextprobe_disown(xfr);
6818
6819         if(!xfr_start_probe(xfr, env, NULL)) {
6820                 /* not started because already in progress */
6821                 lock_basic_unlock(&xfr->lock);
6822         }
6823 }
6824
6825 /** return true if there are probe (SOA UDP query) targets in the master list*/
6826 static int
6827 have_probe_targets(struct auth_master* list)
6828 {
6829         struct auth_master* p;
6830         for(p=list; p; p = p->next) {
6831                 if(!p->allow_notify && p->host)
6832                         return 1;
6833         }
6834         return 0;
6835 }
6836
6837 /** start task_probe if possible, if no masters for probe start task_transfer
6838  * returns true if task has been started, and false if the task is already
6839  * in progress. */
6840 static int
6841 xfr_start_probe(struct auth_xfer* xfr, struct module_env* env,
6842         struct auth_master* spec)
6843 {
6844         /* see if we need to start a probe (or maybe it is already in
6845          * progress (due to notify)) */
6846         if(xfr->task_probe->worker == NULL) {
6847                 if(!have_probe_targets(xfr->task_probe->masters) &&
6848                         !(xfr->task_probe->only_lookup &&
6849                         xfr->task_probe->masters != NULL)) {
6850                         /* useless to pick up task_probe, no masters to
6851                          * probe. Instead attempt to pick up task transfer */
6852                         if(xfr->task_transfer->worker == NULL) {
6853                                 xfr_start_transfer(xfr, env, spec);
6854                                 return 1;
6855                         }
6856                         /* task transfer already in progress */
6857                         return 0;
6858                 }
6859
6860                 /* pick up the probe task ourselves */
6861                 xfr->task_probe->worker = env->worker;
6862                 xfr->task_probe->env = env;
6863                 xfr->task_probe->cp = NULL;
6864
6865                 /* start the task */
6866                 /* have not seen a new lease yet, this scan */
6867                 xfr->task_probe->have_new_lease = 0;
6868                 /* if this was a timeout, no specific first master to scan */
6869                 /* otherwise, spec is nonNULL the notified master, scan
6870                  * first and also transfer first from it */
6871                 xfr_probe_start_list(xfr, spec);
6872                 /* setup to start the lookup of hostnames of masters afresh */
6873                 xfr_probe_start_lookups(xfr);
6874                 /* send the probe packet or next send, or end task */
6875                 xfr_probe_send_or_end(xfr, env);
6876                 return 1;
6877         }
6878         return 0;
6879 }
6880
6881 /** for task_nextprobe.
6882  * determine next timeout for auth_xfer. Also (re)sets timer.
6883  * @param xfr: task structure
6884  * @param env: module environment, with worker and time.
6885  * @param failure: set true if timer should be set for failure retry.
6886  * @param lookup_only: only perform lookups when timer done, 0 sec timeout
6887  */
6888 static void
6889 xfr_set_timeout(struct auth_xfer* xfr, struct module_env* env,
6890         int failure, int lookup_only)
6891 {
6892         struct timeval tv;
6893         log_assert(xfr->task_nextprobe != NULL);
6894         log_assert(xfr->task_nextprobe->worker == NULL ||
6895                 xfr->task_nextprobe->worker == env->worker);
6896         /* normally, nextprobe = startoflease + refresh,
6897          * but if expiry is sooner, use that one.
6898          * after a failure, use the retry timer instead. */
6899         xfr->task_nextprobe->next_probe = *env->now;
6900         if(xfr->lease_time && !failure)
6901                 xfr->task_nextprobe->next_probe = xfr->lease_time;
6902         
6903         if(!failure) {
6904                 xfr->task_nextprobe->backoff = 0;
6905         } else {
6906                 if(xfr->task_nextprobe->backoff == 0)
6907                                 xfr->task_nextprobe->backoff = 3;
6908                 else    xfr->task_nextprobe->backoff *= 2;
6909                 if(xfr->task_nextprobe->backoff > AUTH_TRANSFER_MAX_BACKOFF)
6910                         xfr->task_nextprobe->backoff =
6911                                 AUTH_TRANSFER_MAX_BACKOFF;
6912         }
6913
6914         if(xfr->have_zone) {
6915                 time_t wait = xfr->refresh;
6916                 if(failure) wait = xfr->retry;
6917                 if(xfr->expiry < wait)
6918                         xfr->task_nextprobe->next_probe += xfr->expiry;
6919                 else    xfr->task_nextprobe->next_probe += wait;
6920                 if(failure)
6921                         xfr->task_nextprobe->next_probe +=
6922                                 xfr->task_nextprobe->backoff;
6923                 /* put the timer exactly on expiry, if possible */
6924                 if(xfr->lease_time && xfr->lease_time+xfr->expiry <
6925                         xfr->task_nextprobe->next_probe &&
6926                         xfr->lease_time+xfr->expiry > *env->now)
6927                         xfr->task_nextprobe->next_probe =
6928                                 xfr->lease_time+xfr->expiry;
6929         } else {
6930                 xfr->task_nextprobe->next_probe +=
6931                         xfr->task_nextprobe->backoff;
6932         }
6933
6934         if(!xfr->task_nextprobe->timer) {
6935                 xfr->task_nextprobe->timer = comm_timer_create(
6936                         env->worker_base, auth_xfer_timer, xfr);
6937                 if(!xfr->task_nextprobe->timer) {
6938                         /* failed to malloc memory. likely zone transfer
6939                          * also fails for that. skip the timeout */
6940                         char zname[255+1];
6941                         dname_str(xfr->name, zname);
6942                         log_err("cannot allocate timer, no refresh for %s",
6943                                 zname);
6944                         return;
6945                 }
6946         }
6947         xfr->task_nextprobe->worker = env->worker;
6948         xfr->task_nextprobe->env = env;
6949         if(*(xfr->task_nextprobe->env->now) <= xfr->task_nextprobe->next_probe)
6950                 tv.tv_sec = xfr->task_nextprobe->next_probe - 
6951                         *(xfr->task_nextprobe->env->now);
6952         else    tv.tv_sec = 0;
6953         if(tv.tv_sec != 0 && lookup_only && xfr->task_probe->masters) {
6954                 /* don't lookup_only, if lookup timeout is 0 anyway,
6955                  * or if we don't have masters to lookup */
6956                 tv.tv_sec = 0;
6957                 if(xfr->task_probe->worker == NULL)
6958                         xfr->task_probe->only_lookup = 1;
6959         }
6960         if(verbosity >= VERB_ALGO) {
6961                 char zname[255+1];
6962                 dname_str(xfr->name, zname);
6963                 verbose(VERB_ALGO, "auth zone %s timeout in %d seconds",
6964                         zname, (int)tv.tv_sec);
6965         }
6966         tv.tv_usec = 0;
6967         comm_timer_set(xfr->task_nextprobe->timer, &tv);
6968 }
6969
6970 /** initial pick up of worker timeouts, ties events to worker event loop */
6971 void
6972 auth_xfer_pickup_initial(struct auth_zones* az, struct module_env* env)
6973 {
6974         struct auth_xfer* x;
6975         lock_rw_wrlock(&az->lock);
6976         RBTREE_FOR(x, struct auth_xfer*, &az->xtree) {
6977                 lock_basic_lock(&x->lock);
6978                 /* set lease_time, because we now have timestamp in env,
6979                  * (not earlier during startup and apply_cfg), and this
6980                  * notes the start time when the data was acquired */
6981                 if(x->have_zone)
6982                         x->lease_time = *env->now;
6983                 if(x->task_nextprobe && x->task_nextprobe->worker == NULL) {
6984                         xfr_set_timeout(x, env, 0, 1);
6985                 }
6986                 lock_basic_unlock(&x->lock);
6987         }
6988         lock_rw_unlock(&az->lock);
6989 }
6990
6991 void auth_zones_cleanup(struct auth_zones* az)
6992 {
6993         struct auth_xfer* x;
6994         lock_rw_wrlock(&az->lock);
6995         RBTREE_FOR(x, struct auth_xfer*, &az->xtree) {
6996                 lock_basic_lock(&x->lock);
6997                 if(x->task_nextprobe && x->task_nextprobe->worker != NULL) {
6998                         xfr_nextprobe_disown(x);
6999                 }
7000                 if(x->task_probe && x->task_probe->worker != NULL) {
7001                         xfr_probe_disown(x);
7002                 }
7003                 if(x->task_transfer && x->task_transfer->worker != NULL) {
7004                         auth_chunks_delete(x->task_transfer);
7005                         xfr_transfer_disown(x);
7006                 }
7007                 lock_basic_unlock(&x->lock);
7008         }
7009         lock_rw_unlock(&az->lock);
7010 }
7011
7012 /**
7013  * malloc the xfer and tasks
7014  * @param z: auth_zone with name of zone.
7015  */
7016 static struct auth_xfer*
7017 auth_xfer_new(struct auth_zone* z)
7018 {
7019         struct auth_xfer* xfr;
7020         xfr = (struct auth_xfer*)calloc(1, sizeof(*xfr));
7021         if(!xfr) return NULL;
7022         xfr->name = memdup(z->name, z->namelen);
7023         if(!xfr->name) {
7024                 free(xfr);
7025                 return NULL;
7026         }
7027         xfr->node.key = xfr;
7028         xfr->namelen = z->namelen;
7029         xfr->namelabs = z->namelabs;
7030         xfr->dclass = z->dclass;
7031
7032         xfr->task_nextprobe = (struct auth_nextprobe*)calloc(1,
7033                 sizeof(struct auth_nextprobe));
7034         if(!xfr->task_nextprobe) {
7035                 free(xfr->name);
7036                 free(xfr);
7037                 return NULL;
7038         }
7039         xfr->task_probe = (struct auth_probe*)calloc(1,
7040                 sizeof(struct auth_probe));
7041         if(!xfr->task_probe) {
7042                 free(xfr->task_nextprobe);
7043                 free(xfr->name);
7044                 free(xfr);
7045                 return NULL;
7046         }
7047         xfr->task_transfer = (struct auth_transfer*)calloc(1,
7048                 sizeof(struct auth_transfer));
7049         if(!xfr->task_transfer) {
7050                 free(xfr->task_probe);
7051                 free(xfr->task_nextprobe);
7052                 free(xfr->name);
7053                 free(xfr);
7054                 return NULL;
7055         }
7056
7057         lock_basic_init(&xfr->lock);
7058         lock_protect(&xfr->lock, &xfr->name, sizeof(xfr->name));
7059         lock_protect(&xfr->lock, &xfr->namelen, sizeof(xfr->namelen));
7060         lock_protect(&xfr->lock, xfr->name, xfr->namelen);
7061         lock_protect(&xfr->lock, &xfr->namelabs, sizeof(xfr->namelabs));
7062         lock_protect(&xfr->lock, &xfr->dclass, sizeof(xfr->dclass));
7063         lock_protect(&xfr->lock, &xfr->notify_received, sizeof(xfr->notify_received));
7064         lock_protect(&xfr->lock, &xfr->notify_serial, sizeof(xfr->notify_serial));
7065         lock_protect(&xfr->lock, &xfr->zone_expired, sizeof(xfr->zone_expired));
7066         lock_protect(&xfr->lock, &xfr->have_zone, sizeof(xfr->have_zone));
7067         lock_protect(&xfr->lock, &xfr->serial, sizeof(xfr->serial));
7068         lock_protect(&xfr->lock, &xfr->retry, sizeof(xfr->retry));
7069         lock_protect(&xfr->lock, &xfr->refresh, sizeof(xfr->refresh));
7070         lock_protect(&xfr->lock, &xfr->expiry, sizeof(xfr->expiry));
7071         lock_protect(&xfr->lock, &xfr->lease_time, sizeof(xfr->lease_time));
7072         lock_protect(&xfr->lock, &xfr->task_nextprobe->worker,
7073                 sizeof(xfr->task_nextprobe->worker));
7074         lock_protect(&xfr->lock, &xfr->task_probe->worker,
7075                 sizeof(xfr->task_probe->worker));
7076         lock_protect(&xfr->lock, &xfr->task_transfer->worker,
7077                 sizeof(xfr->task_transfer->worker));
7078         lock_basic_lock(&xfr->lock);
7079         return xfr;
7080 }
7081
7082 /** Create auth_xfer structure.
7083  * This populates the have_zone, soa values, and so on times.
7084  * and sets the timeout, if a zone transfer is needed a short timeout is set.
7085  * For that the auth_zone itself must exist (and read in zonefile)
7086  * returns false on alloc failure. */
7087 struct auth_xfer*
7088 auth_xfer_create(struct auth_zones* az, struct auth_zone* z)
7089 {
7090         struct auth_xfer* xfr;
7091
7092         /* malloc it */
7093         xfr = auth_xfer_new(z);
7094         if(!xfr) {
7095                 log_err("malloc failure");
7096                 return NULL;
7097         }
7098         /* insert in tree */
7099         (void)rbtree_insert(&az->xtree, &xfr->node);
7100         return xfr;
7101 }
7102
7103 /** create new auth_master structure */
7104 static struct auth_master*
7105 auth_master_new(struct auth_master*** list)
7106 {
7107         struct auth_master *m;
7108         m = (struct auth_master*)calloc(1, sizeof(*m));
7109         if(!m) {
7110                 log_err("malloc failure");
7111                 return NULL;
7112         }
7113         /* set first pointer to m, or next pointer of previous element to m */
7114         (**list) = m;
7115         /* store m's next pointer as future point to store at */
7116         (*list) = &(m->next);
7117         return m;
7118 }
7119
7120 /** dup_prefix : create string from initial part of other string, malloced */
7121 static char*
7122 dup_prefix(char* str, size_t num)
7123 {
7124         char* result;
7125         size_t len = strlen(str);
7126         if(len < num) num = len; /* not more than strlen */
7127         result = (char*)malloc(num+1);
7128         if(!result) {
7129                 log_err("malloc failure");
7130                 return result;
7131         }
7132         memmove(result, str, num);
7133         result[num] = 0;
7134         return result;
7135 }
7136
7137 /** dup string and print error on error */
7138 static char*
7139 dup_all(char* str)
7140 {
7141         char* result = strdup(str);
7142         if(!result) {
7143                 log_err("malloc failure");
7144                 return NULL;
7145         }
7146         return result;
7147 }
7148
7149 /** find first of two characters */
7150 static char*
7151 str_find_first_of_chars(char* s, char a, char b)
7152 {
7153         char* ra = strchr(s, a);
7154         char* rb = strchr(s, b);
7155         if(!ra) return rb;
7156         if(!rb) return ra;
7157         if(ra < rb) return ra;
7158         return rb;
7159 }
7160
7161 /** parse URL into host and file parts, false on malloc or parse error */
7162 static int
7163 parse_url(char* url, char** host, char** file, int* port, int* ssl)
7164 {
7165         char* p = url;
7166         /* parse http://www.example.com/file.htm
7167          * or http://127.0.0.1   (index.html)
7168          * or https://[::1@1234]/a/b/c/d */
7169         *ssl = 1;
7170         *port = AUTH_HTTPS_PORT;
7171
7172         /* parse http:// or https:// */
7173         if(strncmp(p, "http://", 7) == 0) {
7174                 p += 7;
7175                 *ssl = 0;
7176                 *port = AUTH_HTTP_PORT;
7177         } else if(strncmp(p, "https://", 8) == 0) {
7178                 p += 8;
7179         } else if(strstr(p, "://") && strchr(p, '/') > strstr(p, "://") &&
7180                 strchr(p, ':') >= strstr(p, "://")) {
7181                 char* uri = dup_prefix(p, (size_t)(strstr(p, "://")-p));
7182                 log_err("protocol %s:// not supported (for url %s)",
7183                         uri?uri:"", p);
7184                 free(uri);
7185                 return 0;
7186         }
7187
7188         /* parse hostname part */
7189         if(p[0] == '[') {
7190                 char* end = strchr(p, ']');
7191                 p++; /* skip over [ */
7192                 if(end) {
7193                         *host = dup_prefix(p, (size_t)(end-p));
7194                         if(!*host) return 0;
7195                         p = end+1; /* skip over ] */
7196                 } else {
7197                         *host = dup_all(p);
7198                         if(!*host) return 0;
7199                         p = end;
7200                 }
7201         } else {
7202                 char* end = str_find_first_of_chars(p, ':', '/');
7203                 if(end) {
7204                         *host = dup_prefix(p, (size_t)(end-p));
7205                         if(!*host) return 0;
7206                 } else {
7207                         *host = dup_all(p);
7208                         if(!*host) return 0;
7209                 }
7210                 p = end; /* at next : or / or NULL */
7211         }
7212
7213         /* parse port number */
7214         if(p && p[0] == ':') {
7215                 char* end = NULL;
7216                 *port = strtol(p+1, &end, 10);
7217                 p = end;
7218         }
7219
7220         /* parse filename part */
7221         while(p && *p == '/')
7222                 p++;
7223         if(!p || p[0] == 0)
7224                 *file = strdup("/");
7225         else    *file = strdup(p);
7226         if(!*file) {
7227                 log_err("malloc failure");
7228                 return 0;
7229         }
7230         return 1;
7231 }
7232
7233 int
7234 xfer_set_masters(struct auth_master** list, struct config_auth* c,
7235         int with_http)
7236 {
7237         struct auth_master* m;
7238         struct config_strlist* p;
7239         /* list points to the first, or next pointer for the new element */
7240         while(*list) {
7241                 list = &( (*list)->next );
7242         }
7243         if(with_http)
7244           for(p = c->urls; p; p = p->next) {
7245                 m = auth_master_new(&list);
7246                 if(!m) return 0;
7247                 m->http = 1;
7248                 if(!parse_url(p->str, &m->host, &m->file, &m->port, &m->ssl))
7249                         return 0;
7250         }
7251         for(p = c->masters; p; p = p->next) {
7252                 m = auth_master_new(&list);
7253                 if(!m) return 0;
7254                 m->ixfr = 1; /* this flag is not configurable */
7255                 m->host = strdup(p->str);
7256                 if(!m->host) {
7257                         log_err("malloc failure");
7258                         return 0;
7259                 }
7260         }
7261         for(p = c->allow_notify; p; p = p->next) {
7262                 m = auth_master_new(&list);
7263                 if(!m) return 0;
7264                 m->allow_notify = 1;
7265                 m->host = strdup(p->str);
7266                 if(!m->host) {
7267                         log_err("malloc failure");
7268                         return 0;
7269                 }
7270         }
7271         return 1;
7272 }
7273
7274 #define SERIAL_BITS     32
7275 int
7276 compare_serial(uint32_t a, uint32_t b)
7277 {
7278         const uint32_t cutoff = ((uint32_t) 1 << (SERIAL_BITS - 1));
7279
7280         if (a == b) {
7281                 return 0;
7282         } else if ((a < b && b - a < cutoff) || (a > b && a - b > cutoff)) {
7283                 return -1;
7284         } else {
7285                 return 1;
7286         }
7287 }
7288
7289 int zonemd_hashalgo_supported(int hashalgo)
7290 {
7291         if(hashalgo == ZONEMD_ALGO_SHA384) return 1;
7292         if(hashalgo == ZONEMD_ALGO_SHA512) return 1;
7293         return 0;
7294 }
7295
7296 int zonemd_scheme_supported(int scheme)
7297 {
7298         if(scheme == ZONEMD_SCHEME_SIMPLE) return 1;
7299         return 0;
7300 }
7301
7302 /** initialize hash for hashing with zonemd hash algo */
7303 static struct secalgo_hash* zonemd_digest_init(int hashalgo, char** reason)
7304 {
7305         struct secalgo_hash *h;
7306         if(hashalgo == ZONEMD_ALGO_SHA384) {
7307                 /* sha384 */
7308                 h = secalgo_hash_create_sha384();
7309                 if(!h)
7310                         *reason = "digest sha384 could not be created";
7311                 return h;
7312         } else if(hashalgo == ZONEMD_ALGO_SHA512) {
7313                 /* sha512 */
7314                 h = secalgo_hash_create_sha512();
7315                 if(!h)
7316                         *reason = "digest sha512 could not be created";
7317                 return h;
7318         }
7319         /* unknown hash algo */
7320         *reason = "unsupported algorithm";
7321         return NULL;
7322 }
7323
7324 /** update the hash for zonemd */
7325 static int zonemd_digest_update(int hashalgo, struct secalgo_hash* h,
7326         uint8_t* data, size_t len, char** reason)
7327 {
7328         if(hashalgo == ZONEMD_ALGO_SHA384) {
7329                 if(!secalgo_hash_update(h, data, len)) {
7330                         *reason = "digest sha384 failed";
7331                         return 0;
7332                 }
7333                 return 1;
7334         } else if(hashalgo == ZONEMD_ALGO_SHA512) {
7335                 if(!secalgo_hash_update(h, data, len)) {
7336                         *reason = "digest sha512 failed";
7337                         return 0;
7338                 }
7339                 return 1;
7340         }
7341         /* unknown hash algo */
7342         *reason = "unsupported algorithm";
7343         return 0;
7344 }
7345
7346 /** finish the hash for zonemd */
7347 static int zonemd_digest_finish(int hashalgo, struct secalgo_hash* h,
7348         uint8_t* result, size_t hashlen, size_t* resultlen, char** reason)
7349 {
7350         if(hashalgo == ZONEMD_ALGO_SHA384) {
7351                 if(hashlen < 384/8) {
7352                         *reason = "digest buffer too small for sha384";
7353                         return 0;
7354                 }
7355                 if(!secalgo_hash_final(h, result, hashlen, resultlen)) {
7356                         *reason = "digest sha384 finish failed";
7357                         return 0;
7358                 }
7359                 return 1;
7360         } else if(hashalgo == ZONEMD_ALGO_SHA512) {
7361                 if(hashlen < 512/8) {
7362                         *reason = "digest buffer too small for sha512";
7363                         return 0;
7364                 }
7365                 if(!secalgo_hash_final(h, result, hashlen, resultlen)) {
7366                         *reason = "digest sha512 finish failed";
7367                         return 0;
7368                 }
7369                 return 1;
7370         }
7371         /* unknown algo */
7372         *reason = "unsupported algorithm";
7373         return 0;
7374 }
7375
7376 /** add rrsets from node to the list */
7377 static size_t authdata_rrsets_to_list(struct auth_rrset** array,
7378         size_t arraysize, struct auth_rrset* first)
7379 {
7380         struct auth_rrset* rrset = first;
7381         size_t num = 0;
7382         while(rrset) {
7383                 if(num >= arraysize)
7384                         return num;
7385                 array[num] = rrset;
7386                 num++;
7387                 rrset = rrset->next;
7388         }
7389         return num;
7390 }
7391
7392 /** compare rr list entries */
7393 static int rrlist_compare(const void* arg1, const void* arg2)
7394 {
7395         struct auth_rrset* r1 = *(struct auth_rrset**)arg1;
7396         struct auth_rrset* r2 = *(struct auth_rrset**)arg2;
7397         uint16_t t1, t2;
7398         if(r1 == NULL) t1 = LDNS_RR_TYPE_RRSIG;
7399         else t1 = r1->type;
7400         if(r2 == NULL) t2 = LDNS_RR_TYPE_RRSIG;
7401         else t2 = r2->type;
7402         if(t1 < t2)
7403                 return -1;
7404         if(t1 > t2)
7405                 return 1;
7406         return 0;
7407 }
7408
7409 /** add type RRSIG to rr list if not one there already,
7410  * this is to perform RRSIG collate processing at that point. */
7411 static void addrrsigtype_if_needed(struct auth_rrset** array,
7412         size_t arraysize, size_t* rrnum, struct auth_data* node)
7413 {
7414         if(az_domain_rrset(node, LDNS_RR_TYPE_RRSIG))
7415                 return; /* already one there */
7416         if((*rrnum) >= arraysize)
7417                 return; /* array too small? */
7418         array[*rrnum] = NULL; /* nothing there, but need entry in list */
7419         (*rrnum)++;
7420 }
7421
7422 /** collate the RRs in an RRset using the simple scheme */
7423 static int zonemd_simple_rrset(struct auth_zone* z, int hashalgo,
7424         struct secalgo_hash* h, struct auth_data* node,
7425         struct auth_rrset* rrset, struct regional* region,
7426         struct sldns_buffer* buf, char** reason)
7427 {
7428         /* canonicalize */
7429         struct ub_packed_rrset_key key;
7430         memset(&key, 0, sizeof(key));
7431         key.entry.key = &key;
7432         key.entry.data = rrset->data;
7433         key.rk.dname = node->name;
7434         key.rk.dname_len = node->namelen;
7435         key.rk.type = htons(rrset->type);
7436         key.rk.rrset_class = htons(z->dclass);
7437         if(!rrset_canonicalize_to_buffer(region, buf, &key)) {
7438                 *reason = "out of memory";
7439                 return 0;
7440         }
7441         regional_free_all(region);
7442
7443         /* hash */
7444         if(!zonemd_digest_update(hashalgo, h, sldns_buffer_begin(buf),
7445                 sldns_buffer_limit(buf), reason)) {
7446                 return 0;
7447         }
7448         return 1;
7449 }
7450
7451 /** count number of RRSIGs in a domain name rrset list */
7452 static size_t zonemd_simple_count_rrsig(struct auth_rrset* rrset,
7453         struct auth_rrset** rrlist, size_t rrnum,
7454         struct auth_zone* z, struct auth_data* node)
7455 {
7456         size_t i, count = 0;
7457         if(rrset) {
7458                 size_t j;
7459                 for(j = 0; j<rrset->data->count; j++) {
7460                         if(rrsig_rdata_get_type_covered(rrset->data->
7461                                 rr_data[j], rrset->data->rr_len[j]) ==
7462                                 LDNS_RR_TYPE_ZONEMD &&
7463                                 query_dname_compare(z->name, node->name)==0) {
7464                                 /* omit RRSIGs over type ZONEMD at apex */
7465                                 continue;
7466                         }
7467                         count++;
7468                 }
7469         }
7470         for(i=0; i<rrnum; i++) {
7471                 if(rrlist[i] && rrlist[i]->type == LDNS_RR_TYPE_ZONEMD &&
7472                         query_dname_compare(z->name, node->name)==0) {
7473                         /* omit RRSIGs over type ZONEMD at apex */
7474                         continue;
7475                 }
7476                 count += (rrlist[i]?rrlist[i]->data->rrsig_count:0);
7477         }
7478         return count;
7479 }
7480
7481 /** allocate sparse rrset data for the number of entries in tepm region */
7482 static int zonemd_simple_rrsig_allocs(struct regional* region,
7483         struct packed_rrset_data* data, size_t count)
7484 {
7485         data->rr_len = regional_alloc(region, sizeof(*data->rr_len) * count);
7486         if(!data->rr_len) {
7487                 return 0;
7488         }
7489         data->rr_ttl = regional_alloc(region, sizeof(*data->rr_ttl) * count);
7490         if(!data->rr_ttl) {
7491                 return 0;
7492         }
7493         data->rr_data = regional_alloc(region, sizeof(*data->rr_data) * count);
7494         if(!data->rr_data) {
7495                 return 0;
7496         }
7497         return 1;
7498 }
7499
7500 /** add the RRSIGs from the rrs in the domain into the data */
7501 static void add_rrlist_rrsigs_into_data(struct packed_rrset_data* data,
7502         size_t* done, struct auth_rrset** rrlist, size_t rrnum,
7503         struct auth_zone* z, struct auth_data* node)
7504 {
7505         size_t i;
7506         for(i=0; i<rrnum; i++) {
7507                 size_t j;
7508                 if(!rrlist[i])
7509                         continue;
7510                 if(rrlist[i] && rrlist[i]->type == LDNS_RR_TYPE_ZONEMD &&
7511                         query_dname_compare(z->name, node->name)==0) {
7512                         /* omit RRSIGs over type ZONEMD at apex */
7513                         continue;
7514                 }
7515                 for(j = 0; j<rrlist[i]->data->rrsig_count; j++) {
7516                         data->rr_len[*done] = rrlist[i]->data->rr_len[rrlist[i]->data->count + j];
7517                         data->rr_ttl[*done] = rrlist[i]->data->rr_ttl[rrlist[i]->data->count + j];
7518                         /* reference the rdata in the rrset, no need to
7519                          * copy it, it is no longer needed at the end of
7520                          * the routine */
7521                         data->rr_data[*done] = rrlist[i]->data->rr_data[rrlist[i]->data->count + j];
7522                         (*done)++;
7523                 }
7524         }
7525 }
7526
7527 static void add_rrset_into_data(struct packed_rrset_data* data,
7528         size_t* done, struct auth_rrset* rrset,
7529         struct auth_zone* z, struct auth_data* node)
7530 {
7531         if(rrset) {
7532                 size_t j;
7533                 for(j = 0; j<rrset->data->count; j++) {
7534                         if(rrsig_rdata_get_type_covered(rrset->data->
7535                                 rr_data[j], rrset->data->rr_len[j]) ==
7536                                 LDNS_RR_TYPE_ZONEMD &&
7537                                 query_dname_compare(z->name, node->name)==0) {
7538                                 /* omit RRSIGs over type ZONEMD at apex */
7539                                 continue;
7540                         }
7541                         data->rr_len[*done] = rrset->data->rr_len[j];
7542                         data->rr_ttl[*done] = rrset->data->rr_ttl[j];
7543                         /* reference the rdata in the rrset, no need to
7544                          * copy it, it is no longer need at the end of
7545                          * the routine */
7546                         data->rr_data[*done] = rrset->data->rr_data[j];
7547                         (*done)++;
7548                 }
7549         }
7550 }
7551
7552 /** collate the RRSIGs using the simple scheme */
7553 static int zonemd_simple_rrsig(struct auth_zone* z, int hashalgo,
7554         struct secalgo_hash* h, struct auth_data* node,
7555         struct auth_rrset* rrset, struct auth_rrset** rrlist, size_t rrnum,
7556         struct regional* region, struct sldns_buffer* buf, char** reason)
7557 {
7558         /* the rrset pointer can be NULL, this means it is type RRSIG and
7559          * there is no ordinary type RRSIG there.  The RRSIGs are stored
7560          * with the RRsets in their data.
7561          *
7562          * The RRset pointer can be nonNULL. This happens if there is
7563          * no RR that is covered by the RRSIG for the domain.  Then this
7564          * RRSIG RR is stored in an rrset of type RRSIG. The other RRSIGs
7565          * are stored in the rrset entries for the RRs in the rr list for
7566          * the domain node.  We need to collate the rrset's data, if any, and
7567          * the rrlist's rrsigs */
7568         /* if this is the apex, omit RRSIGs that cover type ZONEMD */
7569         /* build rrsig rrset */
7570         size_t done = 0;
7571         struct ub_packed_rrset_key key;
7572         struct packed_rrset_data data;
7573         memset(&key, 0, sizeof(key));
7574         memset(&data, 0, sizeof(data));
7575         key.entry.key = &key;
7576         key.entry.data = &data;
7577         key.rk.dname = node->name;
7578         key.rk.dname_len = node->namelen;
7579         key.rk.type = htons(LDNS_RR_TYPE_RRSIG);
7580         key.rk.rrset_class = htons(z->dclass);
7581         data.count = zonemd_simple_count_rrsig(rrset, rrlist, rrnum, z, node);
7582         if(!zonemd_simple_rrsig_allocs(region, &data, data.count)) {
7583                 *reason = "out of memory";
7584                 regional_free_all(region);
7585                 return 0;
7586         }
7587         /* all the RRSIGs stored in the other rrsets for this domain node */
7588         add_rrlist_rrsigs_into_data(&data, &done, rrlist, rrnum, z, node);
7589         /* plus the RRSIGs stored in an rrset of type RRSIG for this node */
7590         add_rrset_into_data(&data, &done, rrset, z, node);
7591
7592         /* canonicalize */
7593         if(!rrset_canonicalize_to_buffer(region, buf, &key)) {
7594                 *reason = "out of memory";
7595                 regional_free_all(region);
7596                 return 0;
7597         }
7598         regional_free_all(region);
7599
7600         /* hash */
7601         if(!zonemd_digest_update(hashalgo, h, sldns_buffer_begin(buf),
7602                 sldns_buffer_limit(buf), reason)) {
7603                 return 0;
7604         }
7605         return 1;
7606 }
7607
7608 /** collate a domain's rrsets using the simple scheme */
7609 static int zonemd_simple_domain(struct auth_zone* z, int hashalgo,
7610         struct secalgo_hash* h, struct auth_data* node,
7611         struct regional* region, struct sldns_buffer* buf, char** reason)
7612 {
7613         const size_t rrlistsize = 65536;
7614         struct auth_rrset* rrlist[rrlistsize];
7615         size_t i, rrnum = 0;
7616         /* see if the domain is out of scope, the zone origin,
7617          * that would be omitted */
7618         if(!dname_subdomain_c(node->name, z->name))
7619                 return 1; /* continue */
7620         /* loop over the rrsets in ascending order. */
7621         rrnum = authdata_rrsets_to_list(rrlist, rrlistsize, node->rrsets);
7622         addrrsigtype_if_needed(rrlist, rrlistsize, &rrnum, node);
7623         qsort(rrlist, rrnum, sizeof(*rrlist), rrlist_compare);
7624         for(i=0; i<rrnum; i++) {
7625                 if(rrlist[i] && rrlist[i]->type == LDNS_RR_TYPE_ZONEMD &&
7626                         query_dname_compare(z->name, node->name) == 0) {
7627                         /* omit type ZONEMD at apex */
7628                         continue;
7629                 }
7630                 if(rrlist[i] == NULL || rrlist[i]->type ==
7631                         LDNS_RR_TYPE_RRSIG) {
7632                         if(!zonemd_simple_rrsig(z, hashalgo, h, node,
7633                                 rrlist[i], rrlist, rrnum, region, buf, reason))
7634                                 return 0;
7635                 } else if(!zonemd_simple_rrset(z, hashalgo, h, node,
7636                         rrlist[i], region, buf, reason)) {
7637                         return 0;
7638                 }
7639         }
7640         return 1;
7641 }
7642
7643 /** collate the zone using the simple scheme */
7644 static int zonemd_simple_collate(struct auth_zone* z, int hashalgo,
7645         struct secalgo_hash* h, struct regional* region,
7646         struct sldns_buffer* buf, char** reason)
7647 {
7648         /* our tree is sorted in canonical order, so we can just loop over
7649          * the tree */
7650         struct auth_data* n;
7651         RBTREE_FOR(n, struct auth_data*, &z->data) {
7652                 if(!zonemd_simple_domain(z, hashalgo, h, n, region, buf,
7653                         reason))
7654                         return 0;
7655         }
7656         return 1;
7657 }
7658
7659 int auth_zone_generate_zonemd_hash(struct auth_zone* z, int scheme,
7660         int hashalgo, uint8_t* hash, size_t hashlen, size_t* resultlen,
7661         struct regional* region, struct sldns_buffer* buf, char** reason)
7662 {
7663         struct secalgo_hash* h = zonemd_digest_init(hashalgo, reason);
7664         if(!h) {
7665                 if(!*reason)
7666                         *reason = "digest init fail";
7667                 return 0;
7668         }
7669         if(scheme == ZONEMD_SCHEME_SIMPLE) {
7670                 if(!zonemd_simple_collate(z, hashalgo, h, region, buf, reason)) {
7671                         if(!*reason) *reason = "scheme simple collate fail";
7672                         secalgo_hash_delete(h);
7673                         return 0;
7674                 }
7675         }
7676         if(!zonemd_digest_finish(hashalgo, h, hash, hashlen, resultlen,
7677                 reason)) {
7678                 secalgo_hash_delete(h);
7679                 *reason = "digest finish fail";
7680                 return 0;
7681         }
7682         secalgo_hash_delete(h);
7683         return 1;
7684 }
7685
7686 int auth_zone_generate_zonemd_check(struct auth_zone* z, int scheme,
7687         int hashalgo, uint8_t* hash, size_t hashlen, struct regional* region,
7688         struct sldns_buffer* buf, char** reason)
7689 {
7690         uint8_t gen[512];
7691         size_t genlen = 0;
7692         *reason = NULL;
7693         if(!zonemd_hashalgo_supported(hashalgo)) {
7694                 /* allow it */
7695                 *reason = "unsupported algorithm";
7696                 return 1;
7697         }
7698         if(!zonemd_scheme_supported(scheme)) {
7699                 /* allow it */
7700                 *reason = "unsupported scheme";
7701                 return 1;
7702         }
7703         if(hashlen < 12) {
7704                 /* the ZONEMD draft requires digests to fail if too small */
7705                 *reason = "digest length too small, less than 12";
7706                 return 0;
7707         }
7708         /* generate digest */
7709         if(!auth_zone_generate_zonemd_hash(z, scheme, hashalgo, gen,
7710                 sizeof(gen), &genlen, region, buf, reason)) {
7711                 /* reason filled in by zonemd hash routine */
7712                 return 0;
7713         }
7714         /* check digest length */
7715         if(hashlen != genlen) {
7716                 *reason = "incorrect digest length";
7717                 if(verbosity >= VERB_ALGO) {
7718                         verbose(VERB_ALGO, "zonemd scheme=%d hashalgo=%d",
7719                                 scheme, hashalgo);
7720                         log_hex("ZONEMD should be  ", gen, genlen);
7721                         log_hex("ZONEMD to check is", hash, hashlen);
7722                 }
7723                 return 0;
7724         }
7725         /* check digest */
7726         if(memcmp(hash, gen, genlen) != 0) {
7727                 *reason = "incorrect digest";
7728                 if(verbosity >= VERB_ALGO) {
7729                         verbose(VERB_ALGO, "zonemd scheme=%d hashalgo=%d",
7730                                 scheme, hashalgo);
7731                         log_hex("ZONEMD should be  ", gen, genlen);
7732                         log_hex("ZONEMD to check is", hash, hashlen);
7733                 }
7734                 return 0;
7735         }
7736         return 1;
7737 }
7738
7739 /** log auth zone message with zone name in front. */
7740 static void auth_zone_log(uint8_t* name, enum verbosity_value level,
7741         const char* format, ...) ATTR_FORMAT(printf, 3, 4);
7742 static void auth_zone_log(uint8_t* name, enum verbosity_value level,
7743         const char* format, ...)
7744 {
7745         va_list args;
7746         va_start(args, format);
7747         if(verbosity >= level) {
7748                 char str[255+1];
7749                 char msg[MAXSYSLOGMSGLEN];
7750                 dname_str(name, str);
7751                 vsnprintf(msg, sizeof(msg), format, args);
7752                 verbose(level, "auth zone %s %s", str, msg);
7753         }
7754         va_end(args);
7755 }
7756
7757 /** ZONEMD, dnssec verify the rrset with the dnskey */
7758 static int zonemd_dnssec_verify_rrset(struct auth_zone* z,
7759         struct module_env* env, struct module_stack* mods,
7760         struct ub_packed_rrset_key* dnskey, struct auth_data* node,
7761         struct auth_rrset* rrset, char** why_bogus, uint8_t* sigalg)
7762 {
7763         struct ub_packed_rrset_key pk;
7764         enum sec_status sec;
7765         struct val_env* ve;
7766         int m;
7767         m = modstack_find(mods, "validator");
7768         if(m == -1) {
7769                 auth_zone_log(z->name, VERB_ALGO, "zonemd dnssec verify: have "
7770                         "DNSKEY chain of trust, but no validator module");
7771                 return 0;
7772         }
7773         ve = (struct val_env*)env->modinfo[m];
7774
7775         memset(&pk, 0, sizeof(pk));
7776         pk.entry.key = &pk;
7777         pk.entry.data = rrset->data;
7778         pk.rk.dname = node->name;
7779         pk.rk.dname_len = node->namelen;
7780         pk.rk.type = htons(rrset->type);
7781         pk.rk.rrset_class = htons(z->dclass);
7782         if(verbosity >= VERB_ALGO) {
7783                 char typestr[32];
7784                 typestr[0]=0;
7785                 sldns_wire2str_type_buf(rrset->type, typestr, sizeof(typestr));
7786                 auth_zone_log(z->name, VERB_ALGO,
7787                         "zonemd: verify %s RRset with DNSKEY", typestr);
7788         }
7789         sec = dnskeyset_verify_rrset(env, ve, &pk, dnskey, sigalg, why_bogus, NULL,
7790                 LDNS_SECTION_ANSWER, NULL);
7791         if(sec == sec_status_secure) {
7792                 return 1;
7793         }
7794         if(why_bogus)
7795                 auth_zone_log(z->name, VERB_ALGO, "DNSSEC verify was bogus: %s", *why_bogus);
7796         return 0;
7797 }
7798
7799 /** check for nsec3, the RR with params equal, if bitmap has the type */
7800 static int nsec3_of_param_has_type(struct auth_rrset* nsec3, int algo,
7801         size_t iter, uint8_t* salt, size_t saltlen, uint16_t rrtype)
7802 {
7803         int i, count = (int)nsec3->data->count;
7804         struct ub_packed_rrset_key pk;
7805         memset(&pk, 0, sizeof(pk));
7806         pk.entry.data = nsec3->data;
7807         for(i=0; i<count; i++) {
7808                 int rralgo;
7809                 size_t rriter, rrsaltlen;
7810                 uint8_t* rrsalt;
7811                 if(!nsec3_get_params(&pk, i, &rralgo, &rriter, &rrsalt,
7812                         &rrsaltlen))
7813                         continue; /* no parameters, malformed */
7814                 if(rralgo != algo || rriter != iter || rrsaltlen != saltlen)
7815                         continue; /* different parameters */
7816                 if(saltlen != 0) {
7817                         if(rrsalt == NULL || salt == NULL)
7818                                 continue;
7819                         if(memcmp(rrsalt, salt, saltlen) != 0)
7820                                 continue; /* different salt parameters */
7821                 }
7822                 if(nsec3_has_type(&pk, i, rrtype))
7823                         return 1;
7824         }
7825         return 0;
7826 }
7827
7828 /** Verify the absence of ZONEMD with DNSSEC by checking NSEC, NSEC3 type flag.
7829  * return false on failure, reason contains description of failure. */
7830 static int zonemd_check_dnssec_absence(struct auth_zone* z,
7831         struct module_env* env, struct module_stack* mods,
7832         struct ub_packed_rrset_key* dnskey, struct auth_data* apex,
7833         char** reason, char** why_bogus, uint8_t* sigalg)
7834 {
7835         struct auth_rrset* nsec = NULL;
7836         if(!apex) {
7837                 *reason = "zone has no apex domain but ZONEMD missing";
7838                 return 0;
7839         }
7840         nsec = az_domain_rrset(apex, LDNS_RR_TYPE_NSEC);
7841         if(nsec) {
7842                 struct ub_packed_rrset_key pk;
7843                 /* dnssec verify the NSEC */
7844                 if(!zonemd_dnssec_verify_rrset(z, env, mods, dnskey, apex,
7845                         nsec, why_bogus, sigalg)) {
7846                         *reason = "DNSSEC verify failed for NSEC RRset";
7847                         return 0;
7848                 }
7849                 /* check type bitmap */
7850                 memset(&pk, 0, sizeof(pk));
7851                 pk.entry.data = nsec->data;
7852                 if(nsec_has_type(&pk, LDNS_RR_TYPE_ZONEMD)) {
7853                         *reason = "DNSSEC NSEC bitmap says type ZONEMD exists";
7854                         return 0;
7855                 }
7856                 auth_zone_log(z->name, VERB_ALGO, "zonemd DNSSEC NSEC verification of absence of ZONEMD secure");
7857         } else {
7858                 /* NSEC3 perhaps ? */
7859                 int algo;
7860                 size_t iter, saltlen;
7861                 uint8_t* salt;
7862                 struct auth_rrset* nsec3param = az_domain_rrset(apex,
7863                         LDNS_RR_TYPE_NSEC3PARAM);
7864                 struct auth_data* match;
7865                 struct auth_rrset* nsec3;
7866                 if(!nsec3param) {
7867                         *reason = "zone has no NSEC information but ZONEMD missing";
7868                         return 0;
7869                 }
7870                 if(!az_nsec3_param(z, &algo, &iter, &salt, &saltlen)) {
7871                         *reason = "zone has no NSEC information but ZONEMD missing";
7872                         return 0;
7873                 }
7874                 /* find the NSEC3 record */
7875                 match = az_nsec3_find_exact(z, z->name, z->namelen, algo,
7876                         iter, salt, saltlen);
7877                 if(!match) {
7878                         *reason = "zone has no NSEC3 domain for the apex but ZONEMD missing";
7879                         return 0;
7880                 }
7881                 nsec3 = az_domain_rrset(match, LDNS_RR_TYPE_NSEC3);
7882                 if(!nsec3) {
7883                         *reason = "zone has no NSEC3 RRset for the apex but ZONEMD missing";
7884                         return 0;
7885                 }
7886                 /* dnssec verify the NSEC3 */
7887                 if(!zonemd_dnssec_verify_rrset(z, env, mods, dnskey, match,
7888                         nsec3, why_bogus, sigalg)) {
7889                         *reason = "DNSSEC verify failed for NSEC3 RRset";
7890                         return 0;
7891                 }
7892                 /* check type bitmap */
7893                 if(nsec3_of_param_has_type(nsec3, algo, iter, salt, saltlen,
7894                         LDNS_RR_TYPE_ZONEMD)) {
7895                         *reason = "DNSSEC NSEC3 bitmap says type ZONEMD exists";
7896                         return 0;
7897                 }
7898                 auth_zone_log(z->name, VERB_ALGO, "zonemd DNSSEC NSEC3 verification of absence of ZONEMD secure");
7899         }
7900
7901         return 1;
7902 }
7903
7904 /** Verify the SOA and ZONEMD DNSSEC signatures.
7905  * return false on failure, reason contains description of failure. */
7906 static int zonemd_check_dnssec_soazonemd(struct auth_zone* z,
7907         struct module_env* env, struct module_stack* mods,
7908         struct ub_packed_rrset_key* dnskey, struct auth_data* apex,
7909         struct auth_rrset* zonemd_rrset, char** reason, char** why_bogus,
7910         uint8_t* sigalg)
7911 {
7912         struct auth_rrset* soa;
7913         if(!apex) {
7914                 *reason = "zone has no apex domain";
7915                 return 0;
7916         }
7917         soa = az_domain_rrset(apex, LDNS_RR_TYPE_SOA);
7918         if(!soa) {
7919                 *reason = "zone has no SOA RRset";
7920                 return 0;
7921         }
7922         if(!zonemd_dnssec_verify_rrset(z, env, mods, dnskey, apex, soa,
7923                 why_bogus, sigalg)) {
7924                 *reason = "DNSSEC verify failed for SOA RRset";
7925                 return 0;
7926         }
7927         if(!zonemd_dnssec_verify_rrset(z, env, mods, dnskey, apex,
7928                 zonemd_rrset, why_bogus, sigalg)) {
7929                 *reason = "DNSSEC verify failed for ZONEMD RRset";
7930                 return 0;
7931         }
7932         auth_zone_log(z->name, VERB_ALGO, "zonemd DNSSEC verification of SOA and ZONEMD RRsets secure");
7933         return 1;
7934 }
7935
7936 /**
7937  * Fail the ZONEMD verification.
7938  * @param z: auth zone that fails.
7939  * @param env: environment with config, to ignore failure or not.
7940  * @param reason: failure string description.
7941  * @param why_bogus: failure string for DNSSEC verification failure.
7942  * @param result: strdup result in here if not NULL.
7943  */
7944 static void auth_zone_zonemd_fail(struct auth_zone* z, struct module_env* env,
7945         char* reason, char* why_bogus, char** result)
7946 {
7947         char zstr[255+1];
7948         /* if fail: log reason, and depending on config also take action
7949          * and drop the zone, eg. it is gone from memory, set zone_expired */
7950         dname_str(z->name, zstr);
7951         if(!reason) reason = "verification failed";
7952         if(result) {
7953                 if(why_bogus) {
7954                         char res[1024];
7955                         snprintf(res, sizeof(res), "%s: %s", reason,
7956                                 why_bogus);
7957                         *result = strdup(res);
7958                 } else {
7959                         *result = strdup(reason);
7960                 }
7961                 if(!*result) log_err("out of memory");
7962         } else {
7963                 log_warn("auth zone %s: ZONEMD verification failed: %s", zstr, reason);
7964         }
7965
7966         if(env->cfg->zonemd_permissive_mode) {
7967                 verbose(VERB_ALGO, "zonemd-permissive-mode enabled, "
7968                         "not blocking zone %s", zstr);
7969                 return;
7970         }
7971
7972         /* expired means the zone gives servfail and is not used by
7973          * lookup if fallback_enabled*/
7974         z->zone_expired = 1;
7975 }
7976
7977 /**
7978  * Verify the zonemd with DNSSEC and hash check, with given key.
7979  * @param z: auth zone.
7980  * @param env: environment with config and temp buffers.
7981  * @param mods: module stack with validator env for verification.
7982  * @param dnskey: dnskey that we can use, or NULL.  If nonnull, the key
7983  *      has been verified and is the start of the chain of trust.
7984  * @param is_insecure: if true, the dnskey is not used, the zone is insecure.
7985  *      And dnssec is not used.  It is DNSSEC secure insecure or not under
7986  *      a trust anchor.
7987  * @param sigalg: if nonNULL provide algorithm downgrade protection.
7988  *      Otherwise one algorithm is enough. Must have space of ALGO_NEEDS_MAX+1.
7989  * @param result: if not NULL result reason copied here.
7990  */
7991 static void
7992 auth_zone_verify_zonemd_with_key(struct auth_zone* z, struct module_env* env,
7993         struct module_stack* mods, struct ub_packed_rrset_key* dnskey,
7994         int is_insecure, char** result, uint8_t* sigalg)
7995 {
7996         char* reason = NULL, *why_bogus = NULL;
7997         struct auth_data* apex = NULL;
7998         struct auth_rrset* zonemd_rrset = NULL;
7999         int zonemd_absent = 0, zonemd_absence_dnssecok = 0;
8000
8001         /* see if ZONEMD is present or absent. */
8002         apex = az_find_name(z, z->name, z->namelen);
8003         if(!apex) {
8004                 zonemd_absent = 1;
8005         } else {
8006                 zonemd_rrset = az_domain_rrset(apex, LDNS_RR_TYPE_ZONEMD);
8007                 if(!zonemd_rrset || zonemd_rrset->data->count==0) {
8008                         zonemd_absent = 1;
8009                         zonemd_rrset = NULL;
8010                 }
8011         }
8012
8013         /* if no DNSSEC, done. */
8014         /* if no ZONEMD, and DNSSEC, use DNSKEY to verify NSEC or NSEC3 for
8015          * zone apex.  Check ZONEMD bit is turned off or else fail */
8016         /* if ZONEMD, and DNSSEC, check DNSSEC signature on SOA and ZONEMD,
8017          * or else fail */
8018         if(!dnskey && !is_insecure) {
8019                 auth_zone_zonemd_fail(z, env, "DNSKEY missing", NULL, result);
8020                 return;
8021         } else if(!zonemd_rrset && dnskey && !is_insecure) {
8022                 /* fetch, DNSSEC verify, and check NSEC/NSEC3 */
8023                 if(!zonemd_check_dnssec_absence(z, env, mods, dnskey, apex,
8024                         &reason, &why_bogus, sigalg)) {
8025                         auth_zone_zonemd_fail(z, env, reason, why_bogus, result);
8026                         return;
8027                 }
8028                 zonemd_absence_dnssecok = 1;
8029         } else if(zonemd_rrset && dnskey && !is_insecure) {
8030                 /* check DNSSEC verify of SOA and ZONEMD */
8031                 if(!zonemd_check_dnssec_soazonemd(z, env, mods, dnskey, apex,
8032                         zonemd_rrset, &reason, &why_bogus, sigalg)) {
8033                         auth_zone_zonemd_fail(z, env, reason, why_bogus, result);
8034                         return;
8035                 }
8036         }
8037
8038         if(zonemd_absent && z->zonemd_reject_absence) {
8039                 auth_zone_zonemd_fail(z, env, "ZONEMD absent and that is not allowed by config", NULL, result);
8040                 return;
8041         }
8042         if(zonemd_absent && zonemd_absence_dnssecok) {
8043                 auth_zone_log(z->name, VERB_ALGO, "DNSSEC verified nonexistence of ZONEMD");
8044                 if(result) {
8045                         *result = strdup("DNSSEC verified nonexistence of ZONEMD");
8046                         if(!*result) log_err("out of memory");
8047                 }
8048                 return;
8049         }
8050         if(zonemd_absent) {
8051                 auth_zone_log(z->name, VERB_ALGO, "no ZONEMD present");
8052                 if(result) {
8053                         *result = strdup("no ZONEMD present");
8054                         if(!*result) log_err("out of memory");
8055                 }
8056                 return;
8057         }
8058
8059         /* check ZONEMD checksum and report or else fail. */
8060         if(!auth_zone_zonemd_check_hash(z, env, &reason)) {
8061                 auth_zone_zonemd_fail(z, env, reason, NULL, result);
8062                 return;
8063         }
8064
8065         /* success! log the success */
8066         if(reason)
8067                 auth_zone_log(z->name, VERB_ALGO, "ZONEMD %s", reason);
8068         else    auth_zone_log(z->name, VERB_ALGO, "ZONEMD verification successful");
8069         if(result) {
8070                 if(reason)
8071                         *result = strdup(reason);
8072                 else    *result = strdup("ZONEMD verification successful");
8073                 if(!*result) log_err("out of memory");
8074         }
8075 }
8076
8077 /**
8078  * verify the zone DNSKEY rrset from the trust anchor
8079  * This is possible because the anchor is for the zone itself, and can
8080  * thus apply straight to the zone DNSKEY set.
8081  * @param z: the auth zone.
8082  * @param env: environment with time and temp buffers.
8083  * @param mods: module stack for validator environment for dnssec validation.
8084  * @param anchor: trust anchor to use
8085  * @param is_insecure: returned, true if the zone is securely insecure.
8086  * @param why_bogus: if the routine fails, returns the failure reason.
8087  * @param keystorage: where to store the ub_packed_rrset_key that is created
8088  *      on success. A pointer to it is returned on success.
8089  * @return the dnskey RRset, reference to zone data and keystorage, or
8090  *      NULL on failure.
8091  */
8092 static struct ub_packed_rrset_key*
8093 zonemd_get_dnskey_from_anchor(struct auth_zone* z, struct module_env* env,
8094         struct module_stack* mods, struct trust_anchor* anchor,
8095         int* is_insecure, char** why_bogus,
8096         struct ub_packed_rrset_key* keystorage)
8097 {
8098         struct auth_data* apex;
8099         struct auth_rrset* dnskey_rrset;
8100         enum sec_status sec;
8101         struct val_env* ve;
8102         int m;
8103
8104         apex = az_find_name(z, z->name, z->namelen);
8105         if(!apex) {
8106                 *why_bogus = "have trust anchor, but zone has no apex domain for DNSKEY";
8107                 return 0;
8108         }
8109         dnskey_rrset = az_domain_rrset(apex, LDNS_RR_TYPE_DNSKEY);
8110         if(!dnskey_rrset || dnskey_rrset->data->count==0) {
8111                 *why_bogus = "have trust anchor, but zone has no DNSKEY";
8112                 return 0;
8113         }
8114
8115         m = modstack_find(mods, "validator");
8116         if(m == -1) {
8117                 *why_bogus = "have trust anchor, but no validator module";
8118                 return 0;
8119         }
8120         ve = (struct val_env*)env->modinfo[m];
8121
8122         memset(keystorage, 0, sizeof(*keystorage));
8123         keystorage->entry.key = keystorage;
8124         keystorage->entry.data = dnskey_rrset->data;
8125         keystorage->rk.dname = apex->name;
8126         keystorage->rk.dname_len = apex->namelen;
8127         keystorage->rk.type = htons(LDNS_RR_TYPE_DNSKEY);
8128         keystorage->rk.rrset_class = htons(z->dclass);
8129         auth_zone_log(z->name, VERB_QUERY,
8130                 "zonemd: verify DNSKEY RRset with trust anchor");
8131         sec = val_verify_DNSKEY_with_TA(env, ve, keystorage, anchor->ds_rrset,
8132                 anchor->dnskey_rrset, NULL, why_bogus, NULL, NULL);
8133         regional_free_all(env->scratch);
8134         if(sec == sec_status_secure) {
8135                 /* success */
8136                 *is_insecure = 0;
8137                 return keystorage;
8138         } else if(sec == sec_status_insecure) {
8139                 /* insecure */
8140                 *is_insecure = 1;
8141         } else {
8142                 /* bogus */
8143                 *is_insecure = 0;
8144                 auth_zone_log(z->name, VERB_ALGO,
8145                         "zonemd: verify DNSKEY RRset with trust anchor failed: %s", *why_bogus);
8146         }
8147         return NULL;
8148 }
8149
8150 /** verify the DNSKEY from the zone with looked up DS record */
8151 static struct ub_packed_rrset_key*
8152 auth_zone_verify_zonemd_key_with_ds(struct auth_zone* z,
8153         struct module_env* env, struct module_stack* mods,
8154         struct ub_packed_rrset_key* ds, int* is_insecure, char** why_bogus,
8155         struct ub_packed_rrset_key* keystorage, uint8_t* sigalg)
8156 {
8157         struct auth_data* apex;
8158         struct auth_rrset* dnskey_rrset;
8159         enum sec_status sec;
8160         struct val_env* ve;
8161         int m;
8162
8163         /* fetch DNSKEY from zone data */
8164         apex = az_find_name(z, z->name, z->namelen);
8165         if(!apex) {
8166                 *why_bogus = "in verifywithDS, zone has no apex";
8167                 return NULL;
8168         }
8169         dnskey_rrset = az_domain_rrset(apex, LDNS_RR_TYPE_DNSKEY);
8170         if(!dnskey_rrset || dnskey_rrset->data->count==0) {
8171                 *why_bogus = "in verifywithDS, zone has no DNSKEY";
8172                 return NULL;
8173         }
8174
8175         m = modstack_find(mods, "validator");
8176         if(m == -1) {
8177                 *why_bogus = "in verifywithDS, have no validator module";
8178                 return NULL;
8179         }
8180         ve = (struct val_env*)env->modinfo[m];
8181
8182         memset(keystorage, 0, sizeof(*keystorage));
8183         keystorage->entry.key = keystorage;
8184         keystorage->entry.data = dnskey_rrset->data;
8185         keystorage->rk.dname = apex->name;
8186         keystorage->rk.dname_len = apex->namelen;
8187         keystorage->rk.type = htons(LDNS_RR_TYPE_DNSKEY);
8188         keystorage->rk.rrset_class = htons(z->dclass);
8189         auth_zone_log(z->name, VERB_QUERY, "zonemd: verify zone DNSKEY with DS");
8190         // @TODO add EDE here? we currently just pass NULL
8191         sec = val_verify_DNSKEY_with_DS(env, ve, keystorage, ds, sigalg,
8192                 why_bogus, NULL, NULL);
8193         regional_free_all(env->scratch);
8194         if(sec == sec_status_secure) {
8195                 /* success */
8196                 return keystorage;
8197         } else if(sec == sec_status_insecure) {
8198                 /* insecure */
8199                 *is_insecure = 1;
8200         } else {
8201                 /* bogus */
8202                 *is_insecure = 0;
8203                 if(*why_bogus == NULL)
8204                         *why_bogus = "verify failed";
8205                 auth_zone_log(z->name, VERB_ALGO,
8206                         "zonemd: verify DNSKEY RRset with DS failed: %s",
8207                         *why_bogus);
8208         }
8209         return NULL;
8210 }
8211
8212 /** callback for ZONEMD lookup of DNSKEY */
8213 void auth_zonemd_dnskey_lookup_callback(void* arg, int rcode, sldns_buffer* buf,
8214         enum sec_status sec, char* why_bogus, int ATTR_UNUSED(was_ratelimited))
8215 {
8216         struct auth_zone* z = (struct auth_zone*)arg;
8217         struct module_env* env;
8218         char* reason = NULL, *ds_bogus = NULL, *typestr="DNSKEY";
8219         struct ub_packed_rrset_key* dnskey = NULL, *ds = NULL;
8220         int is_insecure = 0, downprot;
8221         struct ub_packed_rrset_key keystorage;
8222         uint8_t sigalg[ALGO_NEEDS_MAX+1];
8223
8224         lock_rw_wrlock(&z->lock);
8225         env = z->zonemd_callback_env;
8226         /* release the env variable so another worker can pick up the
8227          * ZONEMD verification task if it wants to */
8228         z->zonemd_callback_env = NULL;
8229         if(!env || env->outnet->want_to_quit || z->zone_deleted) {
8230                 lock_rw_unlock(&z->lock);
8231                 return; /* stop on quit */
8232         }
8233         if(z->zonemd_callback_qtype == LDNS_RR_TYPE_DS)
8234                 typestr = "DS";
8235         downprot = env->cfg->harden_algo_downgrade;
8236
8237         /* process result */
8238         if(sec == sec_status_bogus) {
8239                 reason = why_bogus;
8240                 if(!reason) {
8241                         if(z->zonemd_callback_qtype == LDNS_RR_TYPE_DNSKEY)
8242                                 reason = "lookup of DNSKEY was bogus";
8243                         else    reason = "lookup of DS was bogus";
8244                 }
8245                 auth_zone_log(z->name, VERB_ALGO,
8246                         "zonemd lookup of %s was bogus: %s", typestr, reason);
8247         } else if(rcode == LDNS_RCODE_NOERROR) {
8248                 uint16_t wanted_qtype = z->zonemd_callback_qtype;
8249                 struct regional* temp = env->scratch;
8250                 struct query_info rq;
8251                 struct reply_info* rep;
8252                 memset(&rq, 0, sizeof(rq));
8253                 rep = parse_reply_in_temp_region(buf, temp, &rq);
8254                 if(rep && rq.qtype == wanted_qtype &&
8255                         query_dname_compare(z->name, rq.qname) == 0 &&
8256                         FLAGS_GET_RCODE(rep->flags) == LDNS_RCODE_NOERROR) {
8257                         /* parsed successfully */
8258                         struct ub_packed_rrset_key* answer =
8259                                 reply_find_answer_rrset(&rq, rep);
8260                         if(answer && sec == sec_status_secure) {
8261                                 if(z->zonemd_callback_qtype == LDNS_RR_TYPE_DNSKEY)
8262                                         dnskey = answer;
8263                                 else    ds = answer;
8264                                 auth_zone_log(z->name, VERB_ALGO,
8265                                         "zonemd lookup of %s was secure", typestr);
8266                         } else if(sec == sec_status_secure && !answer) {
8267                                 is_insecure = 1;
8268                                 auth_zone_log(z->name, VERB_ALGO,
8269                                         "zonemd lookup of %s has no content, but is secure, treat as insecure", typestr);
8270                         } else if(sec == sec_status_insecure) {
8271                                 is_insecure = 1;
8272                                 auth_zone_log(z->name, VERB_ALGO,
8273                                         "zonemd lookup of %s was insecure", typestr);
8274                         } else if(sec == sec_status_indeterminate) {
8275                                 is_insecure = 1;
8276                                 auth_zone_log(z->name, VERB_ALGO,
8277                                         "zonemd lookup of %s was indeterminate, treat as insecure", typestr);
8278                         } else {
8279                                 auth_zone_log(z->name, VERB_ALGO,
8280                                         "zonemd lookup of %s has nodata", typestr);
8281                                 if(z->zonemd_callback_qtype == LDNS_RR_TYPE_DNSKEY)
8282                                         reason = "lookup of DNSKEY has nodata";
8283                                 else    reason = "lookup of DS has nodata";
8284                         }
8285                 } else if(rep && rq.qtype == wanted_qtype &&
8286                         query_dname_compare(z->name, rq.qname) == 0 &&
8287                         FLAGS_GET_RCODE(rep->flags) == LDNS_RCODE_NXDOMAIN &&
8288                         sec == sec_status_secure) {
8289                         /* secure nxdomain, so the zone is like some RPZ zone
8290                          * that does not exist in the wider internet, with
8291                          * a secure nxdomain answer outside of it. So we
8292                          * treat the zonemd zone without a dnssec chain of
8293                          * trust, as insecure. */
8294                         is_insecure = 1;
8295                         auth_zone_log(z->name, VERB_ALGO,
8296                                 "zonemd lookup of %s was secure NXDOMAIN, treat as insecure", typestr);
8297                 } else if(rep && rq.qtype == wanted_qtype &&
8298                         query_dname_compare(z->name, rq.qname) == 0 &&
8299                         FLAGS_GET_RCODE(rep->flags) == LDNS_RCODE_NXDOMAIN &&
8300                         sec == sec_status_insecure) {
8301                         is_insecure = 1;
8302                         auth_zone_log(z->name, VERB_ALGO,
8303                                 "zonemd lookup of %s was insecure NXDOMAIN, treat as insecure", typestr);
8304                 } else if(rep && rq.qtype == wanted_qtype &&
8305                         query_dname_compare(z->name, rq.qname) == 0 &&
8306                         FLAGS_GET_RCODE(rep->flags) == LDNS_RCODE_NXDOMAIN &&
8307                         sec == sec_status_indeterminate) {
8308                         is_insecure = 1;
8309                         auth_zone_log(z->name, VERB_ALGO,
8310                                 "zonemd lookup of %s was indeterminate NXDOMAIN, treat as insecure", typestr);
8311                 } else {
8312                         auth_zone_log(z->name, VERB_ALGO,
8313                                 "zonemd lookup of %s has no answer", typestr);
8314                         if(z->zonemd_callback_qtype == LDNS_RR_TYPE_DNSKEY)
8315                                 reason = "lookup of DNSKEY has no answer";
8316                         else    reason = "lookup of DS has no answer";
8317                 }
8318         } else {
8319                 auth_zone_log(z->name, VERB_ALGO,
8320                         "zonemd lookup of %s failed", typestr);
8321                 if(z->zonemd_callback_qtype == LDNS_RR_TYPE_DNSKEY)
8322                         reason = "lookup of DNSKEY failed";
8323                 else    reason = "lookup of DS failed";
8324         }
8325
8326         if(!reason && !is_insecure && !dnskey && ds) {
8327                 dnskey = auth_zone_verify_zonemd_key_with_ds(z, env,
8328                         &env->mesh->mods, ds, &is_insecure, &ds_bogus,
8329                         &keystorage, downprot?sigalg:NULL);
8330                 if(!dnskey && !is_insecure && !reason)
8331                         reason = "DNSKEY verify with DS failed";
8332         }
8333
8334         if(reason) {
8335                 auth_zone_zonemd_fail(z, env, reason, ds_bogus, NULL);
8336                 lock_rw_unlock(&z->lock);
8337                 return;
8338         }
8339
8340         auth_zone_verify_zonemd_with_key(z, env, &env->mesh->mods, dnskey,
8341                 is_insecure, NULL, downprot?sigalg:NULL);
8342         regional_free_all(env->scratch);
8343         lock_rw_unlock(&z->lock);
8344 }
8345
8346 /** lookup DNSKEY for ZONEMD verification */
8347 static int
8348 zonemd_lookup_dnskey(struct auth_zone* z, struct module_env* env)
8349 {
8350         struct query_info qinfo;
8351         uint16_t qflags = BIT_RD;
8352         struct edns_data edns;
8353         sldns_buffer* buf = env->scratch_buffer;
8354         int fetch_ds = 0;
8355
8356         if(!z->fallback_enabled) {
8357                 /* we cannot actually get the DNSKEY, because it is in the
8358                  * zone we have ourselves, and it is not served yet
8359                  * (possibly), so fetch type DS */
8360                 fetch_ds = 1;
8361         }
8362         if(z->zonemd_callback_env) {
8363                 /* another worker is already working on the callback
8364                  * for the DNSKEY lookup for ZONEMD verification.
8365                  * We do not also have to do ZONEMD verification, let that
8366                  * worker do it */
8367                 auth_zone_log(z->name, VERB_ALGO,
8368                         "zonemd needs lookup of %s and that already is worked on by another worker", (fetch_ds?"DS":"DNSKEY"));
8369                 return 1;
8370         }
8371
8372         /* use mesh_new_callback to lookup the DNSKEY,
8373          * and then wait for them to be looked up (in cache, or query) */
8374         qinfo.qname_len = z->namelen;
8375         qinfo.qname = z->name;
8376         qinfo.qclass = z->dclass;
8377         if(fetch_ds)
8378                 qinfo.qtype = LDNS_RR_TYPE_DS;
8379         else    qinfo.qtype = LDNS_RR_TYPE_DNSKEY;
8380         qinfo.local_alias = NULL;
8381         if(verbosity >= VERB_ALGO) {
8382                 char buf1[512];
8383                 char buf2[LDNS_MAX_DOMAINLEN+1];
8384                 dname_str(z->name, buf2);
8385                 snprintf(buf1, sizeof(buf1), "auth zone %s: lookup %s "
8386                         "for zonemd verification", buf2,
8387                         (fetch_ds?"DS":"DNSKEY"));
8388                 log_query_info(VERB_ALGO, buf1, &qinfo);
8389         }
8390         edns.edns_present = 1;
8391         edns.ext_rcode = 0;
8392         edns.edns_version = 0;
8393         edns.bits = EDNS_DO;
8394         edns.opt_list_in = NULL;
8395         edns.opt_list_out = NULL;
8396         edns.opt_list_inplace_cb_out = NULL;
8397         if(sldns_buffer_capacity(buf) < 65535)
8398                 edns.udp_size = (uint16_t)sldns_buffer_capacity(buf);
8399         else    edns.udp_size = 65535;
8400
8401         /* store the worker-specific module env for the callback.
8402          * We can then reference this when the callback executes */
8403         z->zonemd_callback_env = env;
8404         z->zonemd_callback_qtype = qinfo.qtype;
8405         /* the callback can be called straight away */
8406         lock_rw_unlock(&z->lock);
8407         if(!mesh_new_callback(env->mesh, &qinfo, qflags, &edns, buf, 0,
8408                 &auth_zonemd_dnskey_lookup_callback, z, 0)) {
8409                 lock_rw_wrlock(&z->lock);
8410                 log_err("out of memory lookup of %s for zonemd",
8411                         (fetch_ds?"DS":"DNSKEY"));
8412                 return 0;
8413         }
8414         lock_rw_wrlock(&z->lock);
8415         return 1;
8416 }
8417
8418 void auth_zone_verify_zonemd(struct auth_zone* z, struct module_env* env,
8419         struct module_stack* mods, char** result, int offline, int only_online)
8420 {
8421         char* reason = NULL, *why_bogus = NULL;
8422         struct trust_anchor* anchor = NULL;
8423         struct ub_packed_rrset_key* dnskey = NULL;
8424         struct ub_packed_rrset_key keystorage;
8425         int is_insecure = 0;
8426         /* verify the ZONEMD if present.
8427          * If not present check if absence is allowed by DNSSEC */
8428         if(!z->zonemd_check)
8429                 return;
8430         if(z->data.count == 0)
8431                 return; /* no data */
8432
8433         /* if zone is under a trustanchor */
8434         /* is it equal to trustanchor - get dnskey's verified */
8435         /* else, find chain of trust by fetching DNSKEYs lookup for zone */
8436         /* result if that, if insecure, means no DNSSEC for the ZONEMD,
8437          * otherwise we have the zone DNSKEY for the DNSSEC verification. */
8438         if(env->anchors)
8439                 anchor = anchors_lookup(env->anchors, z->name, z->namelen,
8440                         z->dclass);
8441         if(anchor && anchor->numDS == 0 && anchor->numDNSKEY == 0) {
8442                 /* domain-insecure trust anchor for unsigned zones */
8443                 lock_basic_unlock(&anchor->lock);
8444                 if(only_online)
8445                         return;
8446                 dnskey = NULL;
8447                 is_insecure = 1;
8448         } else if(anchor && query_dname_compare(z->name, anchor->name) == 0) {
8449                 if(only_online) {
8450                         lock_basic_unlock(&anchor->lock);
8451                         return;
8452                 }
8453                 /* equal to trustanchor, no need for online lookups */
8454                 dnskey = zonemd_get_dnskey_from_anchor(z, env, mods, anchor,
8455                         &is_insecure, &why_bogus, &keystorage);
8456                 lock_basic_unlock(&anchor->lock);
8457                 if(!dnskey && !reason && !is_insecure) {
8458                         reason = "verify DNSKEY RRset with trust anchor failed";
8459                 }
8460         } else if(anchor) {
8461                 lock_basic_unlock(&anchor->lock);
8462                 /* perform online lookups */
8463                 if(offline)
8464                         return;
8465                 /* setup online lookups, and wait for them */
8466                 if(zonemd_lookup_dnskey(z, env)) {
8467                         /* wait for the lookup */
8468                         return;
8469                 }
8470                 reason = "could not lookup DNSKEY for chain of trust";
8471         } else {
8472                 /* the zone is not under a trust anchor */
8473                 if(only_online)
8474                         return;
8475                 dnskey = NULL;
8476                 is_insecure = 1;
8477         }
8478
8479         if(reason) {
8480                 auth_zone_zonemd_fail(z, env, reason, why_bogus, result);
8481                 return;
8482         }
8483
8484         auth_zone_verify_zonemd_with_key(z, env, mods, dnskey, is_insecure,
8485                 result, NULL);
8486         regional_free_all(env->scratch);
8487 }
8488
8489 void auth_zones_pickup_zonemd_verify(struct auth_zones* az,
8490         struct module_env* env)
8491 {
8492         struct auth_zone key;
8493         uint8_t savezname[255+1];
8494         size_t savezname_len;
8495         struct auth_zone* z;
8496         key.node.key = &key;
8497         lock_rw_rdlock(&az->lock);
8498         RBTREE_FOR(z, struct auth_zone*, &az->ztree) {
8499                 lock_rw_wrlock(&z->lock);
8500                 if(!z->zonemd_check) {
8501                         lock_rw_unlock(&z->lock);
8502                         continue;
8503                 }
8504                 key.dclass = z->dclass;
8505                 key.namelabs = z->namelabs;
8506                 if(z->namelen > sizeof(savezname)) {
8507                         lock_rw_unlock(&z->lock);
8508                         log_err("auth_zones_pickup_zonemd_verify: zone name too long");
8509                         continue;
8510                 }
8511                 savezname_len = z->namelen;
8512                 memmove(savezname, z->name, z->namelen);
8513                 lock_rw_unlock(&az->lock);
8514                 auth_zone_verify_zonemd(z, env, &env->mesh->mods, NULL, 0, 1);
8515                 lock_rw_unlock(&z->lock);
8516                 lock_rw_rdlock(&az->lock);
8517                 /* find the zone we had before, it is not deleted,
8518                  * because we have a flag for that that is processed at
8519                  * apply_cfg time */
8520                 key.namelen = savezname_len;
8521                 key.name = savezname;
8522                 z = (struct auth_zone*)rbtree_search(&az->ztree, &key);
8523                 if(!z)
8524                         break;
8525         }
8526         lock_rw_unlock(&az->lock);
8527 }