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_secalgo.h"
  71 #include <ctype.h>
  72
  73 /** bytes to use for NSEC3 hash buffer. 20 for sha1 */
  74 #define N3HASHBUFLEN 32
  75 /** max number of CNAMEs we are willing to follow (in one answer) */
  76 #define MAX_CNAME_CHAIN 8
  77 /** timeout for probe packets for SOA */
  78 #define AUTH_PROBE_TIMEOUT 100 /* msec */
  79 /** when to stop with SOA probes (when exponential timeouts exceed this) */
  80 #define AUTH_PROBE_TIMEOUT_STOP 1000 /* msec */
  81 /* auth transfer timeout for TCP connections, in msec */
  82 #define AUTH_TRANSFER_TIMEOUT 10000 /* msec */
  83 /* auth transfer max backoff for failed tranfers and probes */
  84 #define AUTH_TRANSFER_MAX_BACKOFF 86400 /* sec */
  85 /* auth http port number */
  86 #define AUTH_HTTP_PORT 80
  87 /* auth https port number */
  88 #define AUTH_HTTPS_PORT 443
  89 /* max depth for nested $INCLUDEs */
  90 #define MAX_INCLUDE_DEPTH 10
  91 /** number of timeouts before we fallback from IXFR to AXFR,
  92  * because some versions of servers (eg. dnsmasq) drop IXFR packets. */
  93 #define NUM_TIMEOUTS_FALLBACK_IXFR 3
  94
  95 /** pick up nextprobe task to start waiting to perform transfer actions */
  96 static void xfr_set_timeout(struct auth_xfer* xfr, struct module_env* env,
  97         int failure, int lookup_only);
  98 /** move to sending the probe packets, next if fails. task_probe */
  99 static void xfr_probe_send_or_end(struct auth_xfer* xfr,
 100         struct module_env* env);
 101 /** pick up probe task with specified(or NULL) destination first,
 102  * or transfer task if nothing to probe, or false if already in progress */
 103 static int xfr_start_probe(struct auth_xfer* xfr, struct module_env* env,
 104         struct auth_master* spec);
 105 /** delete xfer structure (not its tree entry) */
 106 void auth_xfer_delete(struct auth_xfer* xfr);
 107
 108 /** create new dns_msg */
 109 static struct dns_msg*
 110 msg_create(struct regional* region, struct query_info* qinfo)
 111 {
 112         struct dns_msg* msg = (struct dns_msg*)regional_alloc(region,
 113                 sizeof(struct dns_msg));
 114         if(!msg)
 115                 return NULL;
 116         msg->qinfo.qname = regional_alloc_init(region, qinfo->qname,
 117                 qinfo->qname_len);
 118         if(!msg->qinfo.qname)
 119                 return NULL;
 120         msg->qinfo.qname_len = qinfo->qname_len;
 121         msg->qinfo.qtype = qinfo->qtype;
 122         msg->qinfo.qclass = qinfo->qclass;
 123         msg->qinfo.local_alias = NULL;
 124         /* non-packed reply_info, because it needs to grow the array */
 125         msg->rep = (struct reply_info*)regional_alloc_zero(region,
 126                 sizeof(struct reply_info)-sizeof(struct rrset_ref));
 127         if(!msg->rep)
 128                 return NULL;
 129         msg->rep->flags = (uint16_t)(BIT_QR | BIT_AA);
 130         msg->rep->authoritative = 1;
 131         msg->rep->qdcount = 1;
 132         /* rrsets is NULL, no rrsets yet */
 133         return msg;
 134 }
 135
 136 /** grow rrset array by one in msg */
 137 static int
 138 msg_grow_array(struct regional* region, struct dns_msg* msg)
 139 {
 140         if(msg->rep->rrsets == NULL) {
 141                 msg->rep->rrsets = regional_alloc_zero(region,
 142                         sizeof(struct ub_packed_rrset_key*)*(msg->rep->rrset_count+1));
 143                 if(!msg->rep->rrsets)
 144                         return 0;
 145         } else {
 146                 struct ub_packed_rrset_key** rrsets_old = msg->rep->rrsets;
 147                 msg->rep->rrsets = regional_alloc_zero(region,
 148                         sizeof(struct ub_packed_rrset_key*)*(msg->rep->rrset_count+1));
 149                 if(!msg->rep->rrsets)
 150                         return 0;
 151                 memmove(msg->rep->rrsets, rrsets_old,
 152                         sizeof(struct ub_packed_rrset_key*)*msg->rep->rrset_count);
 153         }
 154         return 1;
 155 }
 156
 157 /** get ttl of rrset */
 158 static time_t
 159 get_rrset_ttl(struct ub_packed_rrset_key* k)
 160 {
 161         struct packed_rrset_data* d = (struct packed_rrset_data*)
 162                 k->entry.data;
 163         return d->ttl;
 164 }
 165
 166 /** Copy rrset into region from domain-datanode and packet rrset */
 167 static struct ub_packed_rrset_key*
 168 auth_packed_rrset_copy_region(struct auth_zone* z, struct auth_data* node,
 169         struct auth_rrset* rrset, struct regional* region, time_t adjust)
 170 {
 171         struct ub_packed_rrset_key key;
 172         memset(&key, 0, sizeof(key));
 173         key.entry.key = &key;
 174         key.entry.data = rrset->data;
 175         key.rk.dname = node->name;
 176         key.rk.dname_len = node->namelen;
 177         key.rk.type = htons(rrset->type);
 178         key.rk.rrset_class = htons(z->dclass);
 179         key.entry.hash = rrset_key_hash(&key.rk);
 180         return packed_rrset_copy_region(&key, region, adjust);
 181 }
 182
 183 /** fix up msg->rep TTL and prefetch ttl */
 184 static void
 185 msg_ttl(struct dns_msg* msg)
 186 {
 187         if(msg->rep->rrset_count == 0) return;
 188         if(msg->rep->rrset_count == 1) {
 189                 msg->rep->ttl = get_rrset_ttl(msg->rep->rrsets[0]);
 190                 msg->rep->prefetch_ttl = PREFETCH_TTL_CALC(msg->rep->ttl);
 191                 msg->rep->serve_expired_ttl = msg->rep->ttl + SERVE_EXPIRED_TTL;
 192         } else if(get_rrset_ttl(msg->rep->rrsets[msg->rep->rrset_count-1]) <
 193                 msg->rep->ttl) {
 194                 msg->rep->ttl = get_rrset_ttl(msg->rep->rrsets[
 195                         msg->rep->rrset_count-1]);
 196                 msg->rep->prefetch_ttl = PREFETCH_TTL_CALC(msg->rep->ttl);
 197                 msg->rep->serve_expired_ttl = msg->rep->ttl + SERVE_EXPIRED_TTL;
 198         }
 199 }
 200
 201 /** see if rrset is a duplicate in the answer message */
 202 static int
 203 msg_rrset_duplicate(struct dns_msg* msg, uint8_t* nm, size_t nmlen,
 204         uint16_t type, uint16_t dclass)
 205 {
 206         size_t i;
 207         for(i=0; i<msg->rep->rrset_count; i++) {
 208                 struct ub_packed_rrset_key* k = msg->rep->rrsets[i];
 209                 if(ntohs(k->rk.type) == type && k->rk.dname_len == nmlen &&
 210                         ntohs(k->rk.rrset_class) == dclass &&
 211                         query_dname_compare(k->rk.dname, nm) == 0)
 212                         return 1;
 213         }
 214         return 0;
 215 }
 216
 217 /** add rrset to answer section (no auth, add rrsets yet) */
 218 static int
 219 msg_add_rrset_an(struct auth_zone* z, struct regional* region,
 220         struct dns_msg* msg, struct auth_data* node, struct auth_rrset* rrset)
 221 {
 222         log_assert(msg->rep->ns_numrrsets == 0);
 223         log_assert(msg->rep->ar_numrrsets == 0);
 224         if(!rrset || !node)
 225                 return 1;
 226         if(msg_rrset_duplicate(msg, node->name, node->namelen, rrset->type,
 227                 z->dclass))
 228                 return 1;
 229         /* grow array */
 230         if(!msg_grow_array(region, msg))
 231                 return 0;
 232         /* copy it */
 233         if(!(msg->rep->rrsets[msg->rep->rrset_count] =
 234                 auth_packed_rrset_copy_region(z, node, rrset, region, 0)))
 235                 return 0;
 236         msg->rep->rrset_count++;
 237         msg->rep->an_numrrsets++;
 238         msg_ttl(msg);
 239         return 1;
 240 }
 241
 242 /** add rrset to authority section (no additonal section rrsets yet) */
 243 static int
 244 msg_add_rrset_ns(struct auth_zone* z, struct regional* region,
 245         struct dns_msg* msg, struct auth_data* node, struct auth_rrset* rrset)
 246 {
 247         log_assert(msg->rep->ar_numrrsets == 0);
 248         if(!rrset || !node)
 249                 return 1;
 250         if(msg_rrset_duplicate(msg, node->name, node->namelen, rrset->type,
 251                 z->dclass))
 252                 return 1;
 253         /* grow array */
 254         if(!msg_grow_array(region, msg))
 255                 return 0;
 256         /* copy it */
 257         if(!(msg->rep->rrsets[msg->rep->rrset_count] =
 258                 auth_packed_rrset_copy_region(z, node, rrset, region, 0)))
 259                 return 0;
 260         msg->rep->rrset_count++;
 261         msg->rep->ns_numrrsets++;
 262         msg_ttl(msg);
 263         return 1;
 264 }
 265
 266 /** add rrset to additional section */
 267 static int
 268 msg_add_rrset_ar(struct auth_zone* z, struct regional* region,
 269         struct dns_msg* msg, struct auth_data* node, struct auth_rrset* rrset)
 270 {
 271         if(!rrset || !node)
 272                 return 1;
 273         if(msg_rrset_duplicate(msg, node->name, node->namelen, rrset->type,
 274                 z->dclass))
 275                 return 1;
 276         /* grow array */
 277         if(!msg_grow_array(region, msg))
 278                 return 0;
 279         /* copy it */
 280         if(!(msg->rep->rrsets[msg->rep->rrset_count] =
 281                 auth_packed_rrset_copy_region(z, node, rrset, region, 0)))
 282                 return 0;
 283         msg->rep->rrset_count++;
 284         msg->rep->ar_numrrsets++;
 285         msg_ttl(msg);
 286         return 1;
 287 }
 288
 289 struct auth_zones* auth_zones_create(void)
 290 {
 291         struct auth_zones* az = (struct auth_zones*)calloc(1, sizeof(*az));
 292         if(!az) {
 293                 log_err("out of memory");
 294                 return NULL;
 295         }
 296         rbtree_init(&az->ztree, &auth_zone_cmp);
 297         rbtree_init(&az->xtree, &auth_xfer_cmp);
 298         lock_rw_init(&az->lock);
 299         lock_protect(&az->lock, &az->ztree, sizeof(az->ztree));
 300         lock_protect(&az->lock, &az->xtree, sizeof(az->xtree));
 301         /* also lock protects the rbnode's in struct auth_zone, auth_xfer */
 302         return az;
 303 }
 304
 305 int auth_zone_cmp(const void* z1, const void* z2)
 306 {
 307         /* first sort on class, so that hierarchy can be maintained within
 308          * a class */
 309         struct auth_zone* a = (struct auth_zone*)z1;
 310         struct auth_zone* b = (struct auth_zone*)z2;
 311         int m;
 312         if(a->dclass != b->dclass) {
 313                 if(a->dclass < b->dclass)
 314                         return -1;
 315                 return 1;
 316         }
 317         /* sorted such that higher zones sort before lower zones (their
 318          * contents) */
 319         return dname_lab_cmp(a->name, a->namelabs, b->name, b->namelabs, &m);
 320 }
 321
 322 int auth_data_cmp(const void* z1, const void* z2)
 323 {
 324         struct auth_data* a = (struct auth_data*)z1;
 325         struct auth_data* b = (struct auth_data*)z2;
 326         int m;
 327         /* canonical sort, because DNSSEC needs that */
 328         return dname_canon_lab_cmp(a->name, a->namelabs, b->name,
 329                 b->namelabs, &m);
 330 }
 331
 332 int auth_xfer_cmp(const void* z1, const void* z2)
 333 {
 334         /* first sort on class, so that hierarchy can be maintained within
 335          * a class */
 336         struct auth_xfer* a = (struct auth_xfer*)z1;
 337         struct auth_xfer* b = (struct auth_xfer*)z2;
 338         int m;
 339         if(a->dclass != b->dclass) {
 340                 if(a->dclass < b->dclass)
 341                         return -1;
 342                 return 1;
 343         }
 344         /* sorted such that higher zones sort before lower zones (their
 345          * contents) */
 346         return dname_lab_cmp(a->name, a->namelabs, b->name, b->namelabs, &m);
 347 }
 348
 349 /** delete auth rrset node */
 350 static void
 351 auth_rrset_delete(struct auth_rrset* rrset)
 352 {
 353         if(!rrset) return;
 354         free(rrset->data);
 355         free(rrset);
 356 }
 357
 358 /** delete auth data domain node */
 359 static void
 360 auth_data_delete(struct auth_data* n)
 361 {
 362         struct auth_rrset* p, *np;
 363         if(!n) return;
 364         p = n->rrsets;
 365         while(p) {
 366                 np = p->next;
 367                 auth_rrset_delete(p);
 368                 p = np;
 369         }
 370         free(n->name);
 371         free(n);
 372 }
 373
 374 /** helper traverse to delete zones */
 375 static void
 376 auth_data_del(rbnode_type* n, void* ATTR_UNUSED(arg))
 377 {
 378         struct auth_data* z = (struct auth_data*)n->key;
 379         auth_data_delete(z);
 380 }
 381
 382 /** delete an auth zone structure (tree remove must be done elsewhere) */
 383 static void
 384 auth_zone_delete(struct auth_zone* z)
 385 {
 386         if(!z) return;
 387         lock_rw_destroy(&z->lock);
 388         traverse_postorder(&z->data, auth_data_del, NULL);
 389         free(z->name);
 390         free(z->zonefile);
 391         free(z);
 392 }
 393
 394 struct auth_zone*
 395 auth_zone_create(struct auth_zones* az, uint8_t* nm, size_t nmlen,
 396         uint16_t dclass)
 397 {
 398         struct auth_zone* z = (struct auth_zone*)calloc(1, sizeof(*z));
 399         if(!z) {
 400                 return NULL;
 401         }
 402         z->node.key = z;
 403         z->dclass = dclass;
 404         z->namelen = nmlen;
 405         z->namelabs = dname_count_labels(nm);
 406         z->name = memdup(nm, nmlen);
 407         if(!z->name) {
 408                 free(z);
 409                 return NULL;
 410         }
 411         rbtree_init(&z->data, &auth_data_cmp);
 412         lock_rw_init(&z->lock);
 413         lock_protect(&z->lock, &z->name, sizeof(*z)-sizeof(rbnode_type));
 414         lock_rw_wrlock(&z->lock);
 415         /* z lock protects all, except rbtree itself, which is az->lock */
 416         if(!rbtree_insert(&az->ztree, &z->node)) {
 417                 lock_rw_unlock(&z->lock);
 418                 auth_zone_delete(z);
 419                 log_warn("duplicate auth zone");
 420                 return NULL;
 421         }
 422         return z;
 423 }
 424
 425 struct auth_zone*
 426 auth_zone_find(struct auth_zones* az, uint8_t* nm, size_t nmlen,
 427         uint16_t dclass)
 428 {
 429         struct auth_zone key;
 430         key.node.key = &key;
 431         key.dclass = dclass;
 432         key.name = nm;
 433         key.namelen = nmlen;
 434         key.namelabs = dname_count_labels(nm);
 435         return (struct auth_zone*)rbtree_search(&az->ztree, &key);
 436 }
 437
 438 struct auth_xfer*
 439 auth_xfer_find(struct auth_zones* az, uint8_t* nm, size_t nmlen,
 440         uint16_t dclass)
 441 {
 442         struct auth_xfer key;
 443         key.node.key = &key;
 444         key.dclass = dclass;
 445         key.name = nm;
 446         key.namelen = nmlen;
 447         key.namelabs = dname_count_labels(nm);
 448         return (struct auth_xfer*)rbtree_search(&az->xtree, &key);
 449 }
 450
 451 /** find an auth zone or sorted less-or-equal, return true if exact */
 452 static int
 453 auth_zone_find_less_equal(struct auth_zones* az, uint8_t* nm, size_t nmlen,
 454         uint16_t dclass, struct auth_zone** z)
 455 {
 456         struct auth_zone key;
 457         key.node.key = &key;
 458         key.dclass = dclass;
 459         key.name = nm;
 460         key.namelen = nmlen;
 461         key.namelabs = dname_count_labels(nm);
 462         return rbtree_find_less_equal(&az->ztree, &key, (rbnode_type**)z);
 463 }
 464
 465
 466 /** find the auth zone that is above the given name */
 467 struct auth_zone*
 468 auth_zones_find_zone(struct auth_zones* az, uint8_t* name, size_t name_len,
 469         uint16_t dclass)
 470 {
 471         uint8_t* nm = name;
 472         size_t nmlen = name_len;
 473         struct auth_zone* z;
 474         if(auth_zone_find_less_equal(az, nm, nmlen, dclass, &z)) {
 475                 /* exact match */
 476                 return z;
 477         } else {
 478                 /* less-or-nothing */
 479                 if(!z) return NULL; /* nothing smaller, nothing above it */
 480                 /* we found smaller name; smaller may be above the name,
 481                  * but not below it. */
 482                 nm = dname_get_shared_topdomain(z->name, name);
 483                 dname_count_size_labels(nm, &nmlen);
 484                 z = NULL;
 485         }
 486
 487         /* search up */
 488         while(!z) {
 489                 z = auth_zone_find(az, nm, nmlen, dclass);
 490                 if(z) return z;
 491                 if(dname_is_root(nm)) break;
 492                 dname_remove_label(&nm, &nmlen);
 493         }
 494         return NULL;
 495 }
 496
 497 /** find or create zone with name str. caller must have lock on az.
 498  * returns a wrlocked zone */
 499 static struct auth_zone*
 500 auth_zones_find_or_add_zone(struct auth_zones* az, char* name)
 501 {
 502         uint8_t nm[LDNS_MAX_DOMAINLEN+1];
 503         size_t nmlen = sizeof(nm);
 504         struct auth_zone* z;
 505
 506         if(sldns_str2wire_dname_buf(name, nm, &nmlen) != 0) {
 507                 log_err("cannot parse auth zone name: %s", name);
 508                 return 0;
 509         }
 510         z = auth_zone_find(az, nm, nmlen, LDNS_RR_CLASS_IN);
 511         if(!z) {
 512                 /* not found, create the zone */
 513                 z = auth_zone_create(az, nm, nmlen, LDNS_RR_CLASS_IN);
 514         } else {
 515                 lock_rw_wrlock(&z->lock);
 516         }
 517         return z;
 518 }
 519
 520 /** find or create xfer zone with name str. caller must have lock on az.
 521  * returns a locked xfer */
 522 static struct auth_xfer*
 523 auth_zones_find_or_add_xfer(struct auth_zones* az, struct auth_zone* z)
 524 {
 525         struct auth_xfer* x;
 526         x = auth_xfer_find(az, z->name, z->namelen, z->dclass);
 527         if(!x) {
 528                 /* not found, create the zone */
 529                 x = auth_xfer_create(az, z);
 530         } else {
 531                 lock_basic_lock(&x->lock);
 532         }
 533         return x;
 534 }
 535
 536 int
 537 auth_zone_set_zonefile(struct auth_zone* z, char* zonefile)
 538 {
 539         if(z->zonefile) free(z->zonefile);
 540         if(zonefile == NULL) {
 541                 z->zonefile = NULL;
 542         } else {
 543                 z->zonefile = strdup(zonefile);
 544                 if(!z->zonefile) {
 545                         log_err("malloc failure");
 546                         return 0;
 547                 }
 548         }
 549         return 1;
 550 }
 551
 552 /** set auth zone fallback. caller must have lock on zone */
 553 int
 554 auth_zone_set_fallback(struct auth_zone* z, char* fallbackstr)
 555 {
 556         if(strcmp(fallbackstr, "yes") != 0 && strcmp(fallbackstr, "no") != 0){
 557                 log_err("auth zone fallback, expected yes or no, got %s",
 558                         fallbackstr);
 559                 return 0;
 560         }
 561         z->fallback_enabled = (strcmp(fallbackstr, "yes")==0);
 562         return 1;
 563 }
 564
 565 /** create domain with the given name */
 566 static struct auth_data*
 567 az_domain_create(struct auth_zone* z, uint8_t* nm, size_t nmlen)
 568 {
 569         struct auth_data* n = (struct auth_data*)malloc(sizeof(*n));
 570         if(!n) return NULL;
 571         memset(n, 0, sizeof(*n));
 572         n->node.key = n;
 573         n->name = memdup(nm, nmlen);
 574         if(!n->name) {
 575                 free(n);
 576                 return NULL;
 577         }
 578         n->namelen = nmlen;
 579         n->namelabs = dname_count_labels(nm);
 580         if(!rbtree_insert(&z->data, &n->node)) {
 581                 log_warn("duplicate auth domain name");
 582                 free(n->name);
 583                 free(n);
 584                 return NULL;
 585         }
 586         return n;
 587 }
 588
 589 /** find domain with exactly the given name */
 590 static struct auth_data*
 591 az_find_name(struct auth_zone* z, uint8_t* nm, size_t nmlen)
 592 {
 593         struct auth_zone key;
 594         key.node.key = &key;
 595         key.name = nm;
 596         key.namelen = nmlen;
 597         key.namelabs = dname_count_labels(nm);
 598         return (struct auth_data*)rbtree_search(&z->data, &key);
 599 }
 600
 601 /** Find domain name (or closest match) */
 602 static void
 603 az_find_domain(struct auth_zone* z, struct query_info* qinfo, int* node_exact,
 604         struct auth_data** node)
 605 {
 606         struct auth_zone key;
 607         key.node.key = &key;
 608         key.name = qinfo->qname;
 609         key.namelen = qinfo->qname_len;
 610         key.namelabs = dname_count_labels(key.name);
 611         *node_exact = rbtree_find_less_equal(&z->data, &key,
 612                 (rbnode_type**)node);
 613 }
 614
 615 /** find or create domain with name in zone */
 616 static struct auth_data*
 617 az_domain_find_or_create(struct auth_zone* z, uint8_t* dname,
 618         size_t dname_len)
 619 {
 620         struct auth_data* n = az_find_name(z, dname, dname_len);
 621         if(!n) {
 622                 n = az_domain_create(z, dname, dname_len);
 623         }
 624         return n;
 625 }
 626
 627 /** find rrset of given type in the domain */
 628 static struct auth_rrset*
 629 az_domain_rrset(struct auth_data* n, uint16_t t)
 630 {
 631         struct auth_rrset* rrset;
 632         if(!n) return NULL;
 633         rrset = n->rrsets;
 634         while(rrset) {
 635                 if(rrset->type == t)
 636                         return rrset;
 637                 rrset = rrset->next;
 638         }
 639         return NULL;
 640 }
 641
 642 /** remove rrset of this type from domain */
 643 static void
 644 domain_remove_rrset(struct auth_data* node, uint16_t rr_type)
 645 {
 646         struct auth_rrset* rrset, *prev;
 647         if(!node) return;
 648         prev = NULL;
 649         rrset = node->rrsets;
 650         while(rrset) {
 651                 if(rrset->type == rr_type) {
 652                         /* found it, now delete it */
 653                         if(prev) prev->next = rrset->next;
 654                         else    node->rrsets = rrset->next;
 655                         auth_rrset_delete(rrset);
 656                         return;
 657                 }
 658                 prev = rrset;
 659                 rrset = rrset->next;
 660         }
 661 }
 662
 663 /** find an rr index in the rrset.  returns true if found */
 664 static int
 665 az_rrset_find_rr(struct packed_rrset_data* d, uint8_t* rdata, size_t len,
 666         size_t* index)
 667 {
 668         size_t i;
 669         for(i=0; i<d->count; i++) {
 670                 if(d->rr_len[i] != len)
 671                         continue;
 672                 if(memcmp(d->rr_data[i], rdata, len) == 0) {
 673                         *index = i;
 674                         return 1;
 675                 }
 676         }
 677         return 0;
 678 }
 679
 680 /** find an rrsig index in the rrset.  returns true if found */
 681 static int
 682 az_rrset_find_rrsig(struct packed_rrset_data* d, uint8_t* rdata, size_t len,
 683         size_t* index)
 684 {
 685         size_t i;
 686         for(i=d->count; i<d->count + d->rrsig_count; i++) {
 687                 if(d->rr_len[i] != len)
 688                         continue;
 689                 if(memcmp(d->rr_data[i], rdata, len) == 0) {
 690                         *index = i;
 691                         return 1;
 692                 }
 693         }
 694         return 0;
 695 }
 696
 697 /** see if rdata is duplicate */
 698 static int
 699 rdata_duplicate(struct packed_rrset_data* d, uint8_t* rdata, size_t len)
 700 {
 701         size_t i;
 702         for(i=0; i<d->count + d->rrsig_count; i++) {
 703                 if(d->rr_len[i] != len)
 704                         continue;
 705                 if(memcmp(d->rr_data[i], rdata, len) == 0)
 706                         return 1;
 707         }
 708         return 0;
 709 }
 710
 711 /** get rrsig type covered from rdata.
 712  * @param rdata: rdata in wireformat, starting with 16bit rdlength.
 713  * @param rdatalen: length of rdata buffer.
 714  * @return type covered (or 0).
 715  */
 716 static uint16_t
 717 rrsig_rdata_get_type_covered(uint8_t* rdata, size_t rdatalen)
 718 {
 719         if(rdatalen < 4)
 720                 return 0;
 721         return sldns_read_uint16(rdata+2);
 722 }
 723
 724 /** remove RR from existing RRset. Also sig, if it is a signature.
 725  * reallocates the packed rrset for a new one, false on alloc failure */
 726 static int
 727 rrset_remove_rr(struct auth_rrset* rrset, size_t index)
 728 {
 729         struct packed_rrset_data* d, *old = rrset->data;
 730         size_t i;
 731         if(index >= old->count + old->rrsig_count)
 732                 return 0; /* index out of bounds */
 733         d = (struct packed_rrset_data*)calloc(1, packed_rrset_sizeof(old) - (
 734                 sizeof(size_t) + sizeof(uint8_t*) + sizeof(time_t) +
 735                 old->rr_len[index]));
 736         if(!d) {
 737                 log_err("malloc failure");
 738                 return 0;
 739         }
 740         d->ttl = old->ttl;
 741         d->count = old->count;
 742         d->rrsig_count = old->rrsig_count;
 743         if(index < d->count) d->count--;
 744         else d->rrsig_count--;
 745         d->trust = old->trust;
 746         d->security = old->security;
 747
 748         /* set rr_len, needed for ptr_fixup */
 749         d->rr_len = (size_t*)((uint8_t*)d +
 750                 sizeof(struct packed_rrset_data));
 751         if(index > 0)
 752                 memmove(d->rr_len, old->rr_len, (index)*sizeof(size_t));
 753         if(index+1 < old->count+old->rrsig_count)
 754                 memmove(&d->rr_len[index], &old->rr_len[index+1],
 755                 (old->count+old->rrsig_count - (index+1))*sizeof(size_t));
 756         packed_rrset_ptr_fixup(d);
 757
 758         /* move over ttls */
 759         if(index > 0)
 760                 memmove(d->rr_ttl, old->rr_ttl, (index)*sizeof(time_t));
 761         if(index+1 < old->count+old->rrsig_count)
 762                 memmove(&d->rr_ttl[index], &old->rr_ttl[index+1],
 763                 (old->count+old->rrsig_count - (index+1))*sizeof(time_t));
 764
 765         /* move over rr_data */
 766         for(i=0; i<d->count+d->rrsig_count; i++) {
 767                 size_t oldi;
 768                 if(i < index) oldi = i;
 769                 else oldi = i+1;
 770                 memmove(d->rr_data[i], old->rr_data[oldi], d->rr_len[i]);
 771         }
 772
 773         /* recalc ttl (lowest of remaining RR ttls) */
 774         if(d->count + d->rrsig_count > 0)
 775                 d->ttl = d->rr_ttl[0];
 776         for(i=0; i<d->count+d->rrsig_count; i++) {
 777                 if(d->rr_ttl[i] < d->ttl)
 778                         d->ttl = d->rr_ttl[i];
 779         }
 780
 781         free(rrset->data);
 782         rrset->data = d;
 783         return 1;
 784 }
 785
 786 /** add RR to existing RRset. If insert_sig is true, add to rrsigs.
 787  * This reallocates the packed rrset for a new one */
 788 static int
 789 rrset_add_rr(struct auth_rrset* rrset, uint32_t rr_ttl, uint8_t* rdata,
 790         size_t rdatalen, int insert_sig)
 791 {
 792         struct packed_rrset_data* d, *old = rrset->data;
 793         size_t total, old_total;
 794
 795         d = (struct packed_rrset_data*)calloc(1, packed_rrset_sizeof(old)
 796                 + sizeof(size_t) + sizeof(uint8_t*) + sizeof(time_t)
 797                 + rdatalen);
 798         if(!d) {
 799                 log_err("out of memory");
 800                 return 0;
 801         }
 802         /* copy base values */
 803         memcpy(d, old, sizeof(struct packed_rrset_data));
 804         if(!insert_sig) {
 805                 d->count++;
 806         } else {
 807                 d->rrsig_count++;
 808         }
 809         old_total = old->count + old->rrsig_count;
 810         total = d->count + d->rrsig_count;
 811         /* set rr_len, needed for ptr_fixup */
 812         d->rr_len = (size_t*)((uint8_t*)d +
 813                 sizeof(struct packed_rrset_data));
 814         if(old->count != 0)
 815                 memmove(d->rr_len, old->rr_len, old->count*sizeof(size_t));
 816         if(old->rrsig_count != 0)
 817                 memmove(d->rr_len+d->count, old->rr_len+old->count,
 818                         old->rrsig_count*sizeof(size_t));
 819         if(!insert_sig)
 820                 d->rr_len[d->count-1] = rdatalen;
 821         else    d->rr_len[total-1] = rdatalen;
 822         packed_rrset_ptr_fixup(d);
 823         if((time_t)rr_ttl < d->ttl)
 824                 d->ttl = rr_ttl;
 825
 826         /* copy old values into new array */
 827         if(old->count != 0) {
 828                 memmove(d->rr_ttl, old->rr_ttl, old->count*sizeof(time_t));
 829                 /* all the old rr pieces are allocated sequential, so we
 830                  * can copy them in one go */
 831                 memmove(d->rr_data[0], old->rr_data[0],
 832                         (old->rr_data[old->count-1] - old->rr_data[0]) +
 833                         old->rr_len[old->count-1]);
 834         }
 835         if(old->rrsig_count != 0) {
 836                 memmove(d->rr_ttl+d->count, old->rr_ttl+old->count,
 837                         old->rrsig_count*sizeof(time_t));
 838                 memmove(d->rr_data[d->count], old->rr_data[old->count],
 839                         (old->rr_data[old_total-1] - old->rr_data[old->count]) +
 840                         old->rr_len[old_total-1]);
 841         }
 842
 843         /* insert new value */
 844         if(!insert_sig) {
 845                 d->rr_ttl[d->count-1] = rr_ttl;
 846                 memmove(d->rr_data[d->count-1], rdata, rdatalen);
 847         } else {
 848                 d->rr_ttl[total-1] = rr_ttl;
 849                 memmove(d->rr_data[total-1], rdata, rdatalen);
 850         }
 851
 852         rrset->data = d;
 853         free(old);
 854         return 1;
 855 }
 856
 857 /** Create new rrset for node with packed rrset with one RR element */
 858 static struct auth_rrset*
 859 rrset_create(struct auth_data* node, uint16_t rr_type, uint32_t rr_ttl,
 860         uint8_t* rdata, size_t rdatalen)
 861 {
 862         struct auth_rrset* rrset = (struct auth_rrset*)calloc(1,
 863                 sizeof(*rrset));
 864         struct auth_rrset* p, *prev;
 865         struct packed_rrset_data* d;
 866         if(!rrset) {
 867                 log_err("out of memory");
 868                 return NULL;
 869         }
 870         rrset->type = rr_type;
 871
 872         /* the rrset data structure, with one RR */
 873         d = (struct packed_rrset_data*)calloc(1,
 874                 sizeof(struct packed_rrset_data) + sizeof(size_t) +
 875                 sizeof(uint8_t*) + sizeof(time_t) + rdatalen);
 876         if(!d) {
 877                 free(rrset);
 878                 log_err("out of memory");
 879                 return NULL;
 880         }
 881         rrset->data = d;
 882         d->ttl = rr_ttl;
 883         d->trust = rrset_trust_prim_noglue;
 884         d->rr_len = (size_t*)((uint8_t*)d + sizeof(struct packed_rrset_data));
 885         d->rr_data = (uint8_t**)&(d->rr_len[1]);
 886         d->rr_ttl = (time_t*)&(d->rr_data[1]);
 887         d->rr_data[0] = (uint8_t*)&(d->rr_ttl[1]);
 888
 889         /* insert the RR */
 890         d->rr_len[0] = rdatalen;
 891         d->rr_ttl[0] = rr_ttl;
 892         memmove(d->rr_data[0], rdata, rdatalen);
 893         d->count++;
 894
 895         /* insert rrset into linked list for domain */
 896         /* find sorted place to link the rrset into the list */
 897         prev = NULL;
 898         p = node->rrsets;
 899         while(p && p->type<=rr_type) {
 900                 prev = p;
 901                 p = p->next;
 902         }
 903         /* so, prev is smaller, and p is larger than rr_type */
 904         rrset->next = p;
 905         if(prev) prev->next = rrset;
 906         else node->rrsets = rrset;
 907         return rrset;
 908 }
 909
 910 /** count number (and size) of rrsigs that cover a type */
 911 static size_t
 912 rrsig_num_that_cover(struct auth_rrset* rrsig, uint16_t rr_type, size_t* sigsz)
 913 {
 914         struct packed_rrset_data* d = rrsig->data;
 915         size_t i, num = 0;
 916         *sigsz = 0;
 917         log_assert(d && rrsig->type == LDNS_RR_TYPE_RRSIG);
 918         for(i=0; i<d->count+d->rrsig_count; i++) {
 919                 if(rrsig_rdata_get_type_covered(d->rr_data[i],
 920                         d->rr_len[i]) == rr_type) {
 921                         num++;
 922                         (*sigsz) += d->rr_len[i];
 923                 }
 924         }
 925         return num;
 926 }
 927
 928 /** See if rrsig set has covered sigs for rrset and move them over */
 929 static int
 930 rrset_moveover_rrsigs(struct auth_data* node, uint16_t rr_type,
 931         struct auth_rrset* rrset, struct auth_rrset* rrsig)
 932 {
 933         size_t sigs, sigsz, i, j, total;
 934         struct packed_rrset_data* sigold = rrsig->data;
 935         struct packed_rrset_data* old = rrset->data;
 936         struct packed_rrset_data* d, *sigd;
 937
 938         log_assert(rrset->type == rr_type);
 939         log_assert(rrsig->type == LDNS_RR_TYPE_RRSIG);
 940         sigs = rrsig_num_that_cover(rrsig, rr_type, &sigsz);
 941         if(sigs == 0) {
 942                 /* 0 rrsigs to move over, done */
 943                 return 1;
 944         }
 945
 946         /* allocate rrset sigsz larger for extra sigs elements, and
 947          * allocate rrsig sigsz smaller for less sigs elements. */
 948         d = (struct packed_rrset_data*)calloc(1, packed_rrset_sizeof(old)
 949                 + sigs*(sizeof(size_t) + sizeof(uint8_t*) + sizeof(time_t))
 950                 + sigsz);
 951         if(!d) {
 952                 log_err("out of memory");
 953                 return 0;
 954         }
 955         /* copy base values */
 956         total = old->count + old->rrsig_count;
 957         memcpy(d, old, sizeof(struct packed_rrset_data));
 958         d->rrsig_count += sigs;
 959         /* setup rr_len */
 960         d->rr_len = (size_t*)((uint8_t*)d +
 961                 sizeof(struct packed_rrset_data));
 962         if(total != 0)
 963                 memmove(d->rr_len, old->rr_len, total*sizeof(size_t));
 964         j = d->count+d->rrsig_count-sigs;
 965         for(i=0; i<sigold->count+sigold->rrsig_count; i++) {
 966                 if(rrsig_rdata_get_type_covered(sigold->rr_data[i],
 967                         sigold->rr_len[i]) == rr_type) {
 968                         d->rr_len[j] = sigold->rr_len[i];
 969                         j++;
 970                 }
 971         }
 972         packed_rrset_ptr_fixup(d);
 973
 974         /* copy old values into new array */
 975         if(total != 0) {
 976                 memmove(d->rr_ttl, old->rr_ttl, total*sizeof(time_t));
 977                 /* all the old rr pieces are allocated sequential, so we
 978                  * can copy them in one go */
 979                 memmove(d->rr_data[0], old->rr_data[0],
 980                         (old->rr_data[total-1] - old->rr_data[0]) +
 981                         old->rr_len[total-1]);
 982         }
 983
 984         /* move over the rrsigs to the larger rrset*/
 985         j = d->count+d->rrsig_count-sigs;
 986         for(i=0; i<sigold->count+sigold->rrsig_count; i++) {
 987                 if(rrsig_rdata_get_type_covered(sigold->rr_data[i],
 988                         sigold->rr_len[i]) == rr_type) {
 989                         /* move this one over to location j */
 990                         d->rr_ttl[j] = sigold->rr_ttl[i];
 991                         memmove(d->rr_data[j], sigold->rr_data[i],
 992                                 sigold->rr_len[i]);
 993                         if(d->rr_ttl[j] < d->ttl)
 994                                 d->ttl = d->rr_ttl[j];
 995                         j++;
 996                 }
 997         }
 998
 999         /* put it in and deallocate the old rrset */
1000         rrset->data = d;
1001         free(old);
1002
1003         /* now make rrsig set smaller */
1004         if(sigold->count+sigold->rrsig_count == sigs) {
1005                 /* remove all sigs from rrsig, remove it entirely */
1006                 domain_remove_rrset(node, LDNS_RR_TYPE_RRSIG);
1007                 return 1;
1008         }
1009         log_assert(packed_rrset_sizeof(sigold) > sigs*(sizeof(size_t) +
1010                 sizeof(uint8_t*) + sizeof(time_t)) + sigsz);
1011         sigd = (struct packed_rrset_data*)calloc(1, packed_rrset_sizeof(sigold)
1012                 - sigs*(sizeof(size_t) + sizeof(uint8_t*) + sizeof(time_t))
1013                 - sigsz);
1014         if(!sigd) {
1015                 /* no need to free up d, it has already been placed in the
1016                  * node->rrset structure */
1017                 log_err("out of memory");
1018                 return 0;
1019         }
1020         /* copy base values */
1021         memcpy(sigd, sigold, sizeof(struct packed_rrset_data));
1022         /* in sigd the RRSIGs are stored in the base of the RR, in count */
1023         sigd->count -= sigs;
1024         /* setup rr_len */
1025         sigd->rr_len = (size_t*)((uint8_t*)sigd +
1026                 sizeof(struct packed_rrset_data));
1027         j = 0;
1028         for(i=0; i<sigold->count+sigold->rrsig_count; i++) {
1029                 if(rrsig_rdata_get_type_covered(sigold->rr_data[i],
1030                         sigold->rr_len[i]) != rr_type) {
1031                         sigd->rr_len[j] = sigold->rr_len[i];
1032                         j++;
1033                 }
1034         }
1035         packed_rrset_ptr_fixup(sigd);
1036
1037         /* copy old values into new rrsig array */
1038         j = 0;
1039         for(i=0; i<sigold->count+sigold->rrsig_count; i++) {
1040                 if(rrsig_rdata_get_type_covered(sigold->rr_data[i],
1041                         sigold->rr_len[i]) != rr_type) {
1042                         /* move this one over to location j */
1043                         sigd->rr_ttl[j] = sigold->rr_ttl[i];
1044                         memmove(sigd->rr_data[j], sigold->rr_data[i],
1045                                 sigold->rr_len[i]);
1046                         if(j==0) sigd->ttl = sigd->rr_ttl[j];
1047                         else {
1048                                 if(sigd->rr_ttl[j] < sigd->ttl)
1049                                         sigd->ttl = sigd->rr_ttl[j];
1050                         }
1051                         j++;
1052                 }
1053         }
1054
1055         /* put it in and deallocate the old rrset */
1056         rrsig->data = sigd;
1057         free(sigold);
1058
1059         return 1;
1060 }
1061
1062 /** copy the rrsigs from the rrset to the rrsig rrset, because the rrset
1063  * is going to be deleted.  reallocates the RRSIG rrset data. */
1064 static int
1065 rrsigs_copy_from_rrset_to_rrsigset(struct auth_rrset* rrset,
1066         struct auth_rrset* rrsigset)
1067 {
1068         size_t i;
1069         if(rrset->data->rrsig_count == 0)
1070                 return 1;
1071
1072         /* move them over one by one, because there might be duplicates,
1073          * duplicates are ignored */
1074         for(i=rrset->data->count;
1075                 i<rrset->data->count+rrset->data->rrsig_count; i++) {
1076                 uint8_t* rdata = rrset->data->rr_data[i];
1077                 size_t rdatalen = rrset->data->rr_len[i];
1078                 time_t rr_ttl  = rrset->data->rr_ttl[i];
1079
1080                 if(rdata_duplicate(rrsigset->data, rdata, rdatalen)) {
1081                         continue;
1082                 }
1083                 if(!rrset_add_rr(rrsigset, rr_ttl, rdata, rdatalen, 0))
1084                         return 0;
1085         }
1086         return 1;
1087 }
1088
1089 /** Add rr to node, ignores duplicate RRs,
1090  * rdata points to buffer with rdatalen octets, starts with 2bytelength. */
1091 static int
1092 az_domain_add_rr(struct auth_data* node, uint16_t rr_type, uint32_t rr_ttl,
1093         uint8_t* rdata, size_t rdatalen, int* duplicate)
1094 {
1095         struct auth_rrset* rrset;
1096         /* packed rrsets have their rrsigs along with them, sort them out */
1097         if(rr_type == LDNS_RR_TYPE_RRSIG) {
1098                 uint16_t ctype = rrsig_rdata_get_type_covered(rdata, rdatalen);
1099                 if((rrset=az_domain_rrset(node, ctype))!= NULL) {
1100                         /* a node of the correct type exists, add the RRSIG
1101                          * to the rrset of the covered data type */
1102                         if(rdata_duplicate(rrset->data, rdata, rdatalen)) {
1103                                 if(duplicate) *duplicate = 1;
1104                                 return 1;
1105                         }
1106                         if(!rrset_add_rr(rrset, rr_ttl, rdata, rdatalen, 1))
1107                                 return 0;
1108                 } else if((rrset=az_domain_rrset(node, rr_type))!= NULL) {
1109                         /* add RRSIG to rrset of type RRSIG */
1110                         if(rdata_duplicate(rrset->data, rdata, rdatalen)) {
1111                                 if(duplicate) *duplicate = 1;
1112                                 return 1;
1113                         }
1114                         if(!rrset_add_rr(rrset, rr_ttl, rdata, rdatalen, 0))
1115                                 return 0;
1116                 } else {
1117                         /* create rrset of type RRSIG */
1118                         if(!rrset_create(node, rr_type, rr_ttl, rdata,
1119                                 rdatalen))
1120                                 return 0;
1121                 }
1122         } else {
1123                 /* normal RR type */
1124                 if((rrset=az_domain_rrset(node, rr_type))!= NULL) {
1125                         /* add data to existing node with data type */
1126                         if(rdata_duplicate(rrset->data, rdata, rdatalen)) {
1127                                 if(duplicate) *duplicate = 1;
1128                                 return 1;
1129                         }
1130                         if(!rrset_add_rr(rrset, rr_ttl, rdata, rdatalen, 0))
1131                                 return 0;
1132                 } else {
1133                         struct auth_rrset* rrsig;
1134                         /* create new node with data type */
1135                         if(!(rrset=rrset_create(node, rr_type, rr_ttl, rdata,
1136                                 rdatalen)))
1137                                 return 0;
1138
1139                         /* see if node of type RRSIG has signatures that
1140                          * cover the data type, and move them over */
1141                         /* and then make the RRSIG type smaller */
1142                         if((rrsig=az_domain_rrset(node, LDNS_RR_TYPE_RRSIG))
1143                                 != NULL) {
1144                                 if(!rrset_moveover_rrsigs(node, rr_type,
1145                                         rrset, rrsig))
1146                                         return 0;
1147                         }
1148                 }
1149         }
1150         return 1;
1151 }
1152
1153 /** insert RR into zone, ignore duplicates */
1154 static int
1155 az_insert_rr(struct auth_zone* z, uint8_t* rr, size_t rr_len,
1156         size_t dname_len, int* duplicate)
1157 {
1158         struct auth_data* node;
1159         uint8_t* dname = rr;
1160         uint16_t rr_type = sldns_wirerr_get_type(rr, rr_len, dname_len);
1161         uint16_t rr_class = sldns_wirerr_get_class(rr, rr_len, dname_len);
1162         uint32_t rr_ttl = sldns_wirerr_get_ttl(rr, rr_len, dname_len);
1163         size_t rdatalen = ((size_t)sldns_wirerr_get_rdatalen(rr, rr_len,
1164                 dname_len))+2;
1165         /* rdata points to rdata prefixed with uint16 rdatalength */
1166         uint8_t* rdata = sldns_wirerr_get_rdatawl(rr, rr_len, dname_len);
1167
1168         if(rr_class != z->dclass) {
1169                 log_err("wrong class for RR");
1170                 return 0;
1171         }
1172         if(!(node=az_domain_find_or_create(z, dname, dname_len))) {
1173                 log_err("cannot create domain");
1174                 return 0;
1175         }
1176         if(!az_domain_add_rr(node, rr_type, rr_ttl, rdata, rdatalen,
1177                 duplicate)) {
1178                 log_err("cannot add RR to domain");
1179                 return 0;
1180         }
1181         return 1;
1182 }
1183
1184 /** Remove rr from node, ignores nonexisting RRs,
1185  * rdata points to buffer with rdatalen octets, starts with 2bytelength. */
1186 static int
1187 az_domain_remove_rr(struct auth_data* node, uint16_t rr_type,
1188         uint8_t* rdata, size_t rdatalen, int* nonexist)
1189 {
1190         struct auth_rrset* rrset;
1191         size_t index = 0;
1192
1193         /* find the plain RR of the given type */
1194         if((rrset=az_domain_rrset(node, rr_type))!= NULL) {
1195                 if(az_rrset_find_rr(rrset->data, rdata, rdatalen, &index)) {
1196                         if(rrset->data->count == 1 &&
1197                                 rrset->data->rrsig_count == 0) {
1198                                 /* last RR, delete the rrset */
1199                                 domain_remove_rrset(node, rr_type);
1200                         } else if(rrset->data->count == 1 &&
1201                                 rrset->data->rrsig_count != 0) {
1202                                 /* move RRSIGs to the RRSIG rrset, or
1203                                  * this one becomes that RRset */
1204                                 struct auth_rrset* rrsigset = az_domain_rrset(
1205                                         node, LDNS_RR_TYPE_RRSIG);
1206                                 if(rrsigset) {
1207                                         /* move left over rrsigs to the
1208                                          * existing rrset of type RRSIG */
1209                                         rrsigs_copy_from_rrset_to_rrsigset(
1210                                                 rrset, rrsigset);
1211                                         /* and then delete the rrset */
1212                                         domain_remove_rrset(node, rr_type);
1213                                 } else {
1214                                         /* no rrset of type RRSIG, this
1215                                          * set is now of that type,
1216                                          * just remove the rr */
1217                                         if(!rrset_remove_rr(rrset, index))
1218                                                 return 0;
1219                                         rrset->type = LDNS_RR_TYPE_RRSIG;
1220                                         rrset->data->count = rrset->data->rrsig_count;
1221                                         rrset->data->rrsig_count = 0;
1222                                 }
1223                         } else {
1224                                 /* remove the RR from the rrset */
1225                                 if(!rrset_remove_rr(rrset, index))
1226                                         return 0;
1227                         }
1228                         return 1;
1229                 }
1230                 /* rr not found in rrset */
1231         }
1232
1233         /* is it a type RRSIG, look under the covered type */
1234         if(rr_type == LDNS_RR_TYPE_RRSIG) {
1235                 uint16_t ctype = rrsig_rdata_get_type_covered(rdata, rdatalen);
1236                 if((rrset=az_domain_rrset(node, ctype))!= NULL) {
1237                         if(az_rrset_find_rrsig(rrset->data, rdata, rdatalen,
1238                                 &index)) {
1239                                 /* rrsig should have d->count > 0, be
1240                                  * over some rr of that type */
1241                                 /* remove the rrsig from the rrsigs list of the
1242                                  * rrset */
1243                                 if(!rrset_remove_rr(rrset, index))
1244                                         return 0;
1245                                 return 1;
1246                         }
1247                 }
1248                 /* also RRSIG not found */
1249         }
1250
1251         /* nothing found to delete */
1252         if(nonexist) *nonexist = 1;
1253         return 1;
1254 }
1255
1256 /** remove RR from zone, ignore if it does not exist, false on alloc failure*/
1257 static int
1258 az_remove_rr(struct auth_zone* z, uint8_t* rr, size_t rr_len,
1259         size_t dname_len, int* nonexist)
1260 {
1261         struct auth_data* node;
1262         uint8_t* dname = rr;
1263         uint16_t rr_type = sldns_wirerr_get_type(rr, rr_len, dname_len);
1264         uint16_t rr_class = sldns_wirerr_get_class(rr, rr_len, dname_len);
1265         size_t rdatalen = ((size_t)sldns_wirerr_get_rdatalen(rr, rr_len,
1266                 dname_len))+2;
1267         /* rdata points to rdata prefixed with uint16 rdatalength */
1268         uint8_t* rdata = sldns_wirerr_get_rdatawl(rr, rr_len, dname_len);
1269
1270         if(rr_class != z->dclass) {
1271                 log_err("wrong class for RR");
1272                 /* really also a nonexisting entry, because no records
1273                  * of that class in the zone, but return an error because
1274                  * getting records of the wrong class is a failure of the
1275                  * zone transfer */
1276                 return 0;
1277         }
1278         node = az_find_name(z, dname, dname_len);
1279         if(!node) {
1280                 /* node with that name does not exist */
1281                 /* nonexisting entry, because no such name */
1282                 *nonexist = 1;
1283                 return 1;
1284         }
1285         if(!az_domain_remove_rr(node, rr_type, rdata, rdatalen, nonexist)) {
1286                 /* alloc failure or so */
1287                 return 0;
1288         }
1289         /* remove the node, if necessary */
1290         /* an rrsets==NULL entry is not kept around for empty nonterminals,
1291          * and also parent nodes are not kept around, so we just delete it */
1292         if(node->rrsets == NULL) {
1293                 (void)rbtree_delete(&z->data, node);
1294                 auth_data_delete(node);
1295         }
1296         return 1;
1297 }
1298
1299 /** decompress an RR into the buffer where it'll be an uncompressed RR
1300  * with uncompressed dname and uncompressed rdata (dnames) */
1301 static int
1302 decompress_rr_into_buffer(struct sldns_buffer* buf, uint8_t* pkt,
1303         size_t pktlen, uint8_t* dname, uint16_t rr_type, uint16_t rr_class,
1304         uint32_t rr_ttl, uint8_t* rr_data, uint16_t rr_rdlen)
1305 {
1306         sldns_buffer pktbuf;
1307         size_t dname_len = 0;
1308         size_t rdlenpos;
1309         size_t rdlen;
1310         uint8_t* rd;
1311         const sldns_rr_descriptor* desc;
1312         sldns_buffer_init_frm_data(&pktbuf, pkt, pktlen);
1313         sldns_buffer_clear(buf);
1314
1315         /* decompress dname */
1316         sldns_buffer_set_position(&pktbuf,
1317                 (size_t)(dname - sldns_buffer_current(&pktbuf)));
1318         dname_len = pkt_dname_len(&pktbuf);
1319         if(dname_len == 0) return 0; /* parse fail on dname */
1320         if(!sldns_buffer_available(buf, dname_len)) return 0;
1321         dname_pkt_copy(&pktbuf, sldns_buffer_current(buf), dname);
1322         sldns_buffer_skip(buf, (ssize_t)dname_len);
1323
1324         /* type, class, ttl and rdatalength fields */
1325         if(!sldns_buffer_available(buf, 10)) return 0;
1326         sldns_buffer_write_u16(buf, rr_type);
1327         sldns_buffer_write_u16(buf, rr_class);
1328         sldns_buffer_write_u32(buf, rr_ttl);
1329         rdlenpos = sldns_buffer_position(buf);
1330         sldns_buffer_write_u16(buf, 0); /* rd length position */
1331
1332         /* decompress rdata */
1333         desc = sldns_rr_descript(rr_type);
1334         rd = rr_data;
1335         rdlen = rr_rdlen;
1336         if(rdlen > 0 && desc && desc->_dname_count > 0) {
1337                 int count = (int)desc->_dname_count;
1338                 int rdf = 0;
1339                 size_t len; /* how much rdata to plain copy */
1340                 size_t uncompressed_len, compressed_len;
1341                 size_t oldpos;
1342                 /* decompress dnames. */
1343                 while(rdlen > 0 && count) {
1344                         switch(desc->_wireformat[rdf]) {
1345                         case LDNS_RDF_TYPE_DNAME:
1346                                 sldns_buffer_set_position(&pktbuf,
1347                                         (size_t)(rd -
1348                                         sldns_buffer_begin(&pktbuf)));
1349                                 oldpos = sldns_buffer_position(&pktbuf);
1350                                 /* moves pktbuf to right after the
1351                                  * compressed dname, and returns uncompressed
1352                                  * dname length */
1353                                 uncompressed_len = pkt_dname_len(&pktbuf);
1354                                 if(!uncompressed_len)
1355                                         return 0; /* parse error in dname */
1356                                 if(!sldns_buffer_available(buf,
1357                                         uncompressed_len))
1358                                         /* dname too long for buffer */
1359                                         return 0;
1360                                 dname_pkt_copy(&pktbuf,
1361                                         sldns_buffer_current(buf), rd);
1362                                 sldns_buffer_skip(buf, (ssize_t)uncompressed_len);
1363                                 compressed_len = sldns_buffer_position(
1364                                         &pktbuf) - oldpos;
1365                                 rd += compressed_len;
1366                                 rdlen -= compressed_len;
1367                                 count--;
1368                                 len = 0;
1369                                 break;
1370                         case LDNS_RDF_TYPE_STR:
1371                                 len = rd[0] + 1;
1372                                 break;
1373                         default:
1374                                 len = get_rdf_size(desc->_wireformat[rdf]);
1375                                 break;
1376                         }
1377                         if(len) {
1378                                 if(!sldns_buffer_available(buf, len))
1379                                         return 0; /* too long for buffer */
1380                                 sldns_buffer_write(buf, rd, len);
1381                                 rd += len;
1382                                 rdlen -= len;
1383                         }
1384                         rdf++;
1385                 }
1386         }
1387         /* copy remaining data */
1388         if(rdlen > 0) {
1389                 if(!sldns_buffer_available(buf, rdlen)) return 0;
1390                 sldns_buffer_write(buf, rd, rdlen);
1391         }
1392         /* fixup rdlength */
1393         sldns_buffer_write_u16_at(buf, rdlenpos,
1394                 sldns_buffer_position(buf)-rdlenpos-2);
1395         sldns_buffer_flip(buf);
1396         return 1;
1397 }
1398
1399 /** insert RR into zone, from packet, decompress RR,
1400  * if duplicate is nonNULL set the flag but otherwise ignore duplicates */
1401 static int
1402 az_insert_rr_decompress(struct auth_zone* z, uint8_t* pkt, size_t pktlen,
1403         struct sldns_buffer* scratch_buffer, uint8_t* dname, uint16_t rr_type,
1404         uint16_t rr_class, uint32_t rr_ttl, uint8_t* rr_data,
1405         uint16_t rr_rdlen, int* duplicate)
1406 {
1407         uint8_t* rr;
1408         size_t rr_len;
1409         size_t dname_len;
1410         if(!decompress_rr_into_buffer(scratch_buffer, pkt, pktlen, dname,
1411                 rr_type, rr_class, rr_ttl, rr_data, rr_rdlen)) {
1412                 log_err("could not decompress RR");
1413                 return 0;
1414         }
1415         rr = sldns_buffer_begin(scratch_buffer);
1416         rr_len = sldns_buffer_limit(scratch_buffer);
1417         dname_len = dname_valid(rr, rr_len);
1418         return az_insert_rr(z, rr, rr_len, dname_len, duplicate);
1419 }
1420
1421 /** remove RR from zone, from packet, decompress RR,
1422  * if nonexist is nonNULL set the flag but otherwise ignore nonexisting entries*/
1423 static int
1424 az_remove_rr_decompress(struct auth_zone* z, uint8_t* pkt, size_t pktlen,
1425         struct sldns_buffer* scratch_buffer, uint8_t* dname, uint16_t rr_type,
1426         uint16_t rr_class, uint32_t rr_ttl, uint8_t* rr_data,
1427         uint16_t rr_rdlen, int* nonexist)
1428 {
1429         uint8_t* rr;
1430         size_t rr_len;
1431         size_t dname_len;
1432         if(!decompress_rr_into_buffer(scratch_buffer, pkt, pktlen, dname,
1433                 rr_type, rr_class, rr_ttl, rr_data, rr_rdlen)) {
1434                 log_err("could not decompress RR");
1435                 return 0;
1436         }
1437         rr = sldns_buffer_begin(scratch_buffer);
1438         rr_len = sldns_buffer_limit(scratch_buffer);
1439         dname_len = dname_valid(rr, rr_len);
1440         return az_remove_rr(z, rr, rr_len, dname_len, nonexist);
1441 }
1442
1443 /**
1444  * Parse zonefile
1445  * @param z: zone to read in.
1446  * @param in: file to read from (just opened).
1447  * @param rr: buffer to use for RRs, 64k.
1448  *      passed so that recursive includes can use the same buffer and do
1449  *      not grow the stack too much.
1450  * @param rrbuflen: sizeof rr buffer.
1451  * @param state: parse state with $ORIGIN, $TTL and 'prev-dname' and so on,
1452  *      that is kept between includes.
1453  *      The lineno is set at 1 and then increased by the function.
1454  * @param fname: file name.
1455  * @param depth: recursion depth for includes
1456  * @param cfg: config for chroot.
1457  * returns false on failure, has printed an error message
1458  */
1459 static int
1460 az_parse_file(struct auth_zone* z, FILE* in, uint8_t* rr, size_t rrbuflen,
1461         struct sldns_file_parse_state* state, char* fname, int depth,
1462         struct config_file* cfg)
1463 {
1464         size_t rr_len, dname_len;
1465         int status;
1466         state->lineno = 1;
1467
1468         while(!feof(in)) {
1469                 rr_len = rrbuflen;
1470                 dname_len = 0;
1471                 status = sldns_fp2wire_rr_buf(in, rr, &rr_len, &dname_len,
1472                         state);
1473                 if(status == LDNS_WIREPARSE_ERR_INCLUDE && rr_len == 0) {
1474                         /* we have $INCLUDE or $something */
1475                         if(strncmp((char*)rr, "$INCLUDE ", 9) == 0 ||
1476                            strncmp((char*)rr, "$INCLUDE\t", 9) == 0) {
1477                                 FILE* inc;
1478                                 int lineno_orig = state->lineno;
1479                                 char* incfile = (char*)rr + 8;
1480                                 if(depth > MAX_INCLUDE_DEPTH) {
1481                                         log_err("%s:%d max include depth"
1482                                           "exceeded", fname, state->lineno);
1483                                         return 0;
1484                                 }
1485                                 /* skip spaces */
1486                                 while(*incfile == ' ' || *incfile == '\t')
1487                                         incfile++;
1488                                 /* adjust for chroot on include file */
1489                                 if(cfg->chrootdir && cfg->chrootdir[0] &&
1490                                         strncmp(incfile, cfg->chrootdir,
1491                                                 strlen(cfg->chrootdir)) == 0)
1492                                         incfile += strlen(cfg->chrootdir);
1493                                 incfile = strdup(incfile);
1494                                 if(!incfile) {
1495                                         log_err("malloc failure");
1496                                         return 0;
1497                                 }
1498                                 verbose(VERB_ALGO, "opening $INCLUDE %s",
1499                                         incfile);
1500                                 inc = fopen(incfile, "r");
1501                                 if(!inc) {
1502                                         log_err("%s:%d cannot open include "
1503                                                 "file %s: %s", fname,
1504                                                 lineno_orig, incfile,
1505                                                 strerror(errno));
1506                                         free(incfile);
1507                                         return 0;
1508                                 }
1509                                 /* recurse read that file now */
1510                                 if(!az_parse_file(z, inc, rr, rrbuflen,
1511                                         state, incfile, depth+1, cfg)) {
1512                                         log_err("%s:%d cannot parse include "
1513                                                 "file %s", fname,
1514                                                 lineno_orig, incfile);
1515                                         fclose(inc);
1516                                         free(incfile);
1517                                         return 0;
1518                                 }
1519                                 fclose(inc);
1520                                 verbose(VERB_ALGO, "done with $INCLUDE %s",
1521                                         incfile);
1522                                 free(incfile);
1523                                 state->lineno = lineno_orig;
1524                         }
1525                         continue;
1526                 }
1527                 if(status != 0) {
1528                         log_err("parse error %s %d:%d: %s", fname,
1529                                 state->lineno, LDNS_WIREPARSE_OFFSET(status),
1530                                 sldns_get_errorstr_parse(status));
1531                         return 0;
1532                 }
1533                 if(rr_len == 0) {
1534                         /* EMPTY line, TTL or ORIGIN */
1535                         continue;
1536                 }
1537                 /* insert wirerr in rrbuf */
1538                 if(!az_insert_rr(z, rr, rr_len, dname_len, NULL)) {
1539                         char buf[17];
1540                         sldns_wire2str_type_buf(sldns_wirerr_get_type(rr,
1541                                 rr_len, dname_len), buf, sizeof(buf));
1542                         log_err("%s:%d cannot insert RR of type %s",
1543                                 fname, state->lineno, buf);
1544                         return 0;
1545                 }
1546         }
1547         return 1;
1548 }
1549
1550 int
1551 auth_zone_read_zonefile(struct auth_zone* z, struct config_file* cfg)
1552 {
1553         uint8_t rr[LDNS_RR_BUF_SIZE];
1554         struct sldns_file_parse_state state;
1555         char* zfilename;
1556         FILE* in;
1557         if(!z || !z->zonefile || z->zonefile[0]==0)
1558                 return 1; /* no file, or "", nothing to read */
1559
1560         zfilename = z->zonefile;
1561         if(cfg->chrootdir && cfg->chrootdir[0] && strncmp(zfilename,
1562                 cfg->chrootdir, strlen(cfg->chrootdir)) == 0)
1563                 zfilename += strlen(cfg->chrootdir);
1564         if(verbosity >= VERB_ALGO) {
1565                 char nm[255+1];
1566                 dname_str(z->name, nm);
1567                 verbose(VERB_ALGO, "read zonefile %s for %s", zfilename, nm);
1568         }
1569         in = fopen(zfilename, "r");
1570         if(!in) {
1571                 char* n = sldns_wire2str_dname(z->name, z->namelen);
1572                 if(z->zone_is_slave && errno == ENOENT) {
1573                         /* we fetch the zone contents later, no file yet */
1574                         verbose(VERB_ALGO, "no zonefile %s for %s",
1575                                 zfilename, n?n:"error");
1576                         free(n);
1577                         return 1;
1578                 }
1579                 log_err("cannot open zonefile %s for %s: %s",
1580                         zfilename, n?n:"error", strerror(errno));
1581                 free(n);
1582                 return 0;
1583         }
1584
1585         /* clear the data tree */
1586         traverse_postorder(&z->data, auth_data_del, NULL);
1587         rbtree_init(&z->data, &auth_data_cmp);
1588
1589         memset(&state, 0, sizeof(state));
1590         /* default TTL to 3600 */
1591         state.default_ttl = 3600;
1592         /* set $ORIGIN to the zone name */
1593         if(z->namelen <= sizeof(state.origin)) {
1594                 memcpy(state.origin, z->name, z->namelen);
1595                 state.origin_len = z->namelen;
1596         }
1597         /* parse the (toplevel) file */
1598         if(!az_parse_file(z, in, rr, sizeof(rr), &state, zfilename, 0, cfg)) {
1599                 char* n = sldns_wire2str_dname(z->name, z->namelen);
1600                 log_err("error parsing zonefile %s for %s",
1601                         zfilename, n?n:"error");
1602                 free(n);
1603                 fclose(in);
1604                 return 0;
1605         }
1606         fclose(in);
1607         return 1;
1608 }
1609
1610 /** write buffer to file and check return codes */
1611 static int
1612 write_out(FILE* out, const char* str, size_t len)
1613 {
1614         size_t r;
1615         if(len == 0)
1616                 return 1;
1617         r = fwrite(str, 1, len, out);
1618         if(r == 0) {
1619                 log_err("write failed: %s", strerror(errno));
1620                 return 0;
1621         } else if(r < len) {
1622                 log_err("write failed: too short (disk full?)");
1623                 return 0;
1624         }
1625         return 1;
1626 }
1627
1628 /** convert auth rr to string */
1629 static int
1630 auth_rr_to_string(uint8_t* nm, size_t nmlen, uint16_t tp, uint16_t cl,
1631         struct packed_rrset_data* data, size_t i, char* s, size_t buflen)
1632 {
1633         int w = 0;
1634         size_t slen = buflen, datlen;
1635         uint8_t* dat;
1636         if(i >= data->count) tp = LDNS_RR_TYPE_RRSIG;
1637         dat = nm;
1638         datlen = nmlen;
1639         w += sldns_wire2str_dname_scan(&dat, &datlen, &s, &slen, NULL, 0);
1640         w += sldns_str_print(&s, &slen, "\t");
1641         w += sldns_str_print(&s, &slen, "%lu\t", (unsigned long)data->rr_ttl[i]);
1642         w += sldns_wire2str_class_print(&s, &slen, cl);
1643         w += sldns_str_print(&s, &slen, "\t");
1644         w += sldns_wire2str_type_print(&s, &slen, tp);
1645         w += sldns_str_print(&s, &slen, "\t");
1646         datlen = data->rr_len[i]-2;
1647         dat = data->rr_data[i]+2;
1648         w += sldns_wire2str_rdata_scan(&dat, &datlen, &s, &slen, tp, NULL, 0);
1649
1650         if(tp == LDNS_RR_TYPE_DNSKEY) {
1651                 w += sldns_str_print(&s, &slen, " ;{id = %u}",
1652                         sldns_calc_keytag_raw(data->rr_data[i]+2,
1653                                 data->rr_len[i]-2));
1654         }
1655         w += sldns_str_print(&s, &slen, "\n");
1656
1657         if(w > (int)buflen) {
1658                 log_nametypeclass(0, "RR too long to print", nm, tp, cl);
1659                 return 0;
1660         }
1661         return 1;
1662 }
1663
1664 /** write rrset to file */
1665 static int
1666 auth_zone_write_rrset(struct auth_zone* z, struct auth_data* node,
1667         struct auth_rrset* r, FILE* out)
1668 {
1669         size_t i, count = r->data->count + r->data->rrsig_count;
1670         char buf[LDNS_RR_BUF_SIZE];
1671         for(i=0; i<count; i++) {
1672                 if(!auth_rr_to_string(node->name, node->namelen, r->type,
1673                         z->dclass, r->data, i, buf, sizeof(buf))) {
1674                         verbose(VERB_ALGO, "failed to rr2str rr %d", (int)i);
1675                         continue;
1676                 }
1677                 if(!write_out(out, buf, strlen(buf)))
1678                         return 0;
1679         }
1680         return 1;
1681 }
1682
1683 /** write domain to file */
1684 static int
1685 auth_zone_write_domain(struct auth_zone* z, struct auth_data* n, FILE* out)
1686 {
1687         struct auth_rrset* r;
1688         /* if this is zone apex, write SOA first */
1689         if(z->namelen == n->namelen) {
1690                 struct auth_rrset* soa = az_domain_rrset(n, LDNS_RR_TYPE_SOA);
1691                 if(soa) {
1692                         if(!auth_zone_write_rrset(z, n, soa, out))
1693                                 return 0;
1694                 }
1695         }
1696         /* write all the RRsets for this domain */
1697         for(r = n->rrsets; r; r = r->next) {
1698                 if(z->namelen == n->namelen &&
1699                         r->type == LDNS_RR_TYPE_SOA)
1700                         continue; /* skip SOA here */
1701                 if(!auth_zone_write_rrset(z, n, r, out))
1702                         return 0;
1703         }
1704         return 1;
1705 }
1706
1707 int auth_zone_write_file(struct auth_zone* z, const char* fname)
1708 {
1709         FILE* out;
1710         struct auth_data* n;
1711         out = fopen(fname, "w");
1712         if(!out) {
1713                 log_err("could not open %s: %s", fname, strerror(errno));
1714                 return 0;
1715         }
1716         RBTREE_FOR(n, struct auth_data*, &z->data) {
1717                 if(!auth_zone_write_domain(z, n, out)) {
1718                         log_err("could not write domain to %s", fname);
1719                         fclose(out);
1720                         return 0;
1721                 }
1722         }
1723         fclose(out);
1724         return 1;
1725 }
1726
1727 /** read all auth zones from file (if they have) */
1728 static int
1729 auth_zones_read_zones(struct auth_zones* az, struct config_file* cfg)
1730 {
1731         struct auth_zone* z;
1732         lock_rw_wrlock(&az->lock);
1733         RBTREE_FOR(z, struct auth_zone*, &az->ztree) {
1734                 lock_rw_wrlock(&z->lock);
1735                 if(!auth_zone_read_zonefile(z, cfg)) {
1736                         lock_rw_unlock(&z->lock);
1737                         lock_rw_unlock(&az->lock);
1738                         return 0;
1739                 }
1740                 lock_rw_unlock(&z->lock);
1741         }
1742         lock_rw_unlock(&az->lock);
1743         return 1;
1744 }
1745
1746 /** find serial number of zone or false if none */
1747 int
1748 auth_zone_get_serial(struct auth_zone* z, uint32_t* serial)
1749 {
1750         struct auth_data* apex;
1751         struct auth_rrset* soa;
1752         struct packed_rrset_data* d;
1753         apex = az_find_name(z, z->name, z->namelen);
1754         if(!apex) return 0;
1755         soa = az_domain_rrset(apex, LDNS_RR_TYPE_SOA);
1756         if(!soa || soa->data->count==0)
1757                 return 0; /* no RRset or no RRs in rrset */
1758         if(soa->data->rr_len[0] < 2+4*5) return 0; /* SOA too short */
1759         d = soa->data;
1760         *serial = sldns_read_uint32(d->rr_data[0]+(d->rr_len[0]-20));
1761         return 1;
1762 }
1763
1764 /** Find auth_zone SOA and populate the values in xfr(soa values). */
1765 static int
1766 xfr_find_soa(struct auth_zone* z, struct auth_xfer* xfr)
1767 {
1768         struct auth_data* apex;
1769         struct auth_rrset* soa;
1770         struct packed_rrset_data* d;
1771         apex = az_find_name(z, z->name, z->namelen);
1772         if(!apex) return 0;
1773         soa = az_domain_rrset(apex, LDNS_RR_TYPE_SOA);
1774         if(!soa || soa->data->count==0)
1775                 return 0; /* no RRset or no RRs in rrset */
1776         if(soa->data->rr_len[0] < 2+4*5) return 0; /* SOA too short */
1777         /* SOA record ends with serial, refresh, retry, expiry, minimum,
1778          * as 4 byte fields */
1779         d = soa->data;
1780         xfr->have_zone = 1;
1781         xfr->serial = sldns_read_uint32(d->rr_data[0]+(d->rr_len[0]-20));
1782         xfr->refresh = sldns_read_uint32(d->rr_data[0]+(d->rr_len[0]-16));
1783         xfr->retry = sldns_read_uint32(d->rr_data[0]+(d->rr_len[0]-12));
1784         xfr->expiry = sldns_read_uint32(d->rr_data[0]+(d->rr_len[0]-8));
1785         /* soa minimum at d->rr_len[0]-4 */
1786         return 1;
1787 }
1788
1789 /**
1790  * Setup auth_xfer zone
1791  * This populates the have_zone, soa values, and so on times.
1792  * Doesn't do network traffic yet, can set option flags.
1793  * @param z: locked by caller, and modified for setup
1794  * @param x: locked by caller, and modified.
1795  * @return false on failure.
1796  */
1797 static int
1798 auth_xfer_setup(struct auth_zone* z, struct auth_xfer* x)
1799 {
1800         /* for a zone without zone transfers, x==NULL, so skip them,
1801          * i.e. the zone config is fixed with no masters or urls */
1802         if(!z || !x) return 1;
1803         if(!xfr_find_soa(z, x)) {
1804                 return 1;
1805         }
1806         /* nothing for probe, nextprobe and transfer tasks */
1807         return 1;
1808 }
1809
1810 /**
1811  * Setup all zones
1812  * @param az: auth zones structure
1813  * @return false on failure.
1814  */
1815 static int
1816 auth_zones_setup_zones(struct auth_zones* az)
1817 {
1818         struct auth_zone* z;
1819         struct auth_xfer* x;
1820         lock_rw_wrlock(&az->lock);
1821         RBTREE_FOR(z, struct auth_zone*, &az->ztree) {
1822                 lock_rw_wrlock(&z->lock);
1823                 x = auth_xfer_find(az, z->name, z->namelen, z->dclass);
1824                 if(x) {
1825                         lock_basic_lock(&x->lock);
1826                 }
1827                 if(!auth_xfer_setup(z, x)) {
1828                         if(x) {
1829                                 lock_basic_unlock(&x->lock);
1830                         }
1831                         lock_rw_unlock(&z->lock);
1832                         lock_rw_unlock(&az->lock);
1833                         return 0;
1834                 }
1835                 if(x) {
1836                         lock_basic_unlock(&x->lock);
1837                 }
1838                 lock_rw_unlock(&z->lock);
1839         }
1840         lock_rw_unlock(&az->lock);
1841         return 1;
1842 }
1843
1844 /** set config items and create zones */
1845 static int
1846 auth_zones_cfg(struct auth_zones* az, struct config_auth* c)
1847 {
1848         struct auth_zone* z;
1849         struct auth_xfer* x = NULL;
1850
1851         /* create zone */
1852         lock_rw_wrlock(&az->lock);
1853         if(!(z=auth_zones_find_or_add_zone(az, c->name))) {
1854                 lock_rw_unlock(&az->lock);
1855                 return 0;
1856         }
1857         if(c->masters || c->urls) {
1858                 if(!(x=auth_zones_find_or_add_xfer(az, z))) {
1859                         lock_rw_unlock(&az->lock);
1860                         lock_rw_unlock(&z->lock);
1861                         return 0;
1862                 }
1863         }
1864         if(c->for_downstream)
1865                 az->have_downstream = 1;
1866         lock_rw_unlock(&az->lock);
1867
1868         /* set options */
1869         z->zone_deleted = 0;
1870         if(!auth_zone_set_zonefile(z, c->zonefile)) {
1871                 if(x) {
1872                         lock_basic_unlock(&x->lock);
1873                 }
1874                 lock_rw_unlock(&z->lock);
1875                 return 0;
1876         }
1877         z->for_downstream = c->for_downstream;
1878         z->for_upstream = c->for_upstream;
1879         z->fallback_enabled = c->fallback_enabled;
1880
1881         /* xfer zone */
1882         if(x) {
1883                 z->zone_is_slave = 1;
1884                 /* set options on xfer zone */
1885                 if(!xfer_set_masters(&x->task_probe->masters, c, 0)) {
1886                         lock_basic_unlock(&x->lock);
1887                         lock_rw_unlock(&z->lock);
1888                         return 0;
1889                 }
1890                 if(!xfer_set_masters(&x->task_transfer->masters, c, 1)) {
1891                         lock_basic_unlock(&x->lock);
1892                         lock_rw_unlock(&z->lock);
1893                         return 0;
1894                 }
1895                 lock_basic_unlock(&x->lock);
1896         }
1897
1898         lock_rw_unlock(&z->lock);
1899         return 1;
1900 }
1901
1902 /** set all auth zones deleted, then in auth_zones_cfg, it marks them
1903  * as nondeleted (if they are still in the config), and then later
1904  * we can find deleted zones */
1905 static void
1906 az_setall_deleted(struct auth_zones* az)
1907 {
1908         struct auth_zone* z;
1909         lock_rw_wrlock(&az->lock);
1910         RBTREE_FOR(z, struct auth_zone*, &az->ztree) {
1911                 lock_rw_wrlock(&z->lock);
1912                 z->zone_deleted = 1;
1913                 lock_rw_unlock(&z->lock);
1914         }
1915         lock_rw_unlock(&az->lock);
1916 }
1917
1918 /** find zones that are marked deleted and delete them.
1919  * This is called from apply_cfg, and there are no threads and no
1920  * workers, so the xfr can just be deleted. */
1921 static void
1922 az_delete_deleted_zones(struct auth_zones* az)
1923 {
1924         struct auth_zone* z;
1925         struct auth_zone* delete_list = NULL, *next;
1926         struct auth_xfer* xfr;
1927         lock_rw_wrlock(&az->lock);
1928         RBTREE_FOR(z, struct auth_zone*, &az->ztree) {
1929                 lock_rw_wrlock(&z->lock);
1930                 if(z->zone_deleted) {
1931                         /* we cannot alter the rbtree right now, but
1932                          * we can put it on a linked list and then
1933                          * delete it */
1934                         z->delete_next = delete_list;
1935                         delete_list = z;
1936                 }
1937                 lock_rw_unlock(&z->lock);
1938         }
1939         /* now we are out of the tree loop and we can loop and delete
1940          * the zones */
1941         z = delete_list;
1942         while(z) {
1943                 next = z->delete_next;
1944                 xfr = auth_xfer_find(az, z->name, z->namelen, z->dclass);
1945                 if(xfr) {
1946                         (void)rbtree_delete(&az->xtree, &xfr->node);
1947                         auth_xfer_delete(xfr);
1948                 }
1949                 (void)rbtree_delete(&az->ztree, &z->node);
1950                 auth_zone_delete(z);
1951                 z = next;
1952         }
1953         lock_rw_unlock(&az->lock);
1954 }
1955
1956 int auth_zones_apply_cfg(struct auth_zones* az, struct config_file* cfg,
1957         int setup)
1958 {
1959         struct config_auth* p;
1960         az_setall_deleted(az);
1961         for(p = cfg->auths; p; p = p->next) {
1962                 if(!p->name || p->name[0] == 0) {
1963                         log_warn("auth-zone without a name, skipped");
1964                         continue;
1965                 }
1966                 if(!auth_zones_cfg(az, p)) {
1967                         log_err("cannot config auth zone %s", p->name);
1968                         return 0;
1969                 }
1970         }
1971         az_delete_deleted_zones(az);
1972         if(!auth_zones_read_zones(az, cfg))
1973                 return 0;
1974         if(setup) {
1975                 if(!auth_zones_setup_zones(az))
1976                         return 0;
1977         }
1978         return 1;
1979 }
1980
1981 /** delete chunks
1982  * @param at: transfer structure with chunks list.  The chunks and their
1983  *      data are freed.
1984  */
1985 static void
1986 auth_chunks_delete(struct auth_transfer* at)
1987 {
1988         if(at->chunks_first) {
1989                 struct auth_chunk* c, *cn;
1990                 c = at->chunks_first;
1991                 while(c) {
1992                         cn = c->next;
1993                         free(c->data);
1994                         free(c);
1995                         c = cn;
1996                 }
1997         }
1998         at->chunks_first = NULL;
1999         at->chunks_last = NULL;
2000 }
2001
2002 /** free master addr list */
2003 static void
2004 auth_free_master_addrs(struct auth_addr* list)
2005 {
2006         struct auth_addr *n;
2007         while(list) {
2008                 n = list->next;
2009                 free(list);
2010                 list = n;
2011         }
2012 }
2013
2014 /** free the masters list */
2015 static void
2016 auth_free_masters(struct auth_master* list)
2017 {
2018         struct auth_master* n;
2019         while(list) {
2020                 n = list->next;
2021                 auth_free_master_addrs(list->list);
2022                 free(list->host);
2023                 free(list->file);
2024                 free(list);
2025                 list = n;
2026         }
2027 }
2028
2029 /** delete auth xfer structure
2030  * @param xfr: delete this xfer and its tasks.
2031  */
2032 void
2033 auth_xfer_delete(struct auth_xfer* xfr)
2034 {
2035         if(!xfr) return;
2036         lock_basic_destroy(&xfr->lock);
2037         free(xfr->name);
2038         if(xfr->task_nextprobe) {
2039                 comm_timer_delete(xfr->task_nextprobe->timer);
2040                 free(xfr->task_nextprobe);
2041         }
2042         if(xfr->task_probe) {
2043                 auth_free_masters(xfr->task_probe->masters);
2044                 comm_point_delete(xfr->task_probe->cp);
2045                 free(xfr->task_probe);
2046         }
2047         if(xfr->task_transfer) {
2048                 auth_free_masters(xfr->task_transfer->masters);
2049                 comm_point_delete(xfr->task_transfer->cp);
2050                 if(xfr->task_transfer->chunks_first) {
2051                         auth_chunks_delete(xfr->task_transfer);
2052                 }
2053                 free(xfr->task_transfer);
2054         }
2055         auth_free_masters(xfr->allow_notify_list);
2056         free(xfr);
2057 }
2058
2059 /** helper traverse to delete zones */
2060 static void
2061 auth_zone_del(rbnode_type* n, void* ATTR_UNUSED(arg))
2062 {
2063         struct auth_zone* z = (struct auth_zone*)n->key;
2064         auth_zone_delete(z);
2065 }
2066
2067 /** helper traverse to delete xfer zones */
2068 static void
2069 auth_xfer_del(rbnode_type* n, void* ATTR_UNUSED(arg))
2070 {
2071         struct auth_xfer* z = (struct auth_xfer*)n->key;
2072         auth_xfer_delete(z);
2073 }
2074
2075 void auth_zones_delete(struct auth_zones* az)
2076 {
2077         if(!az) return;
2078         lock_rw_destroy(&az->lock);
2079         traverse_postorder(&az->ztree, auth_zone_del, NULL);
2080         traverse_postorder(&az->xtree, auth_xfer_del, NULL);
2081         free(az);
2082 }
2083
2084 /** true if domain has only nsec3 */
2085 static int
2086 domain_has_only_nsec3(struct auth_data* n)
2087 {
2088         struct auth_rrset* rrset = n->rrsets;
2089         int nsec3_seen = 0;
2090         while(rrset) {
2091                 if(rrset->type == LDNS_RR_TYPE_NSEC3) {
2092                         nsec3_seen = 1;
2093                 } else if(rrset->type != LDNS_RR_TYPE_RRSIG) {
2094                         return 0;
2095                 }
2096                 rrset = rrset->next;
2097         }
2098         return nsec3_seen;
2099 }
2100
2101 /** see if the domain has a wildcard child '*.domain' */
2102 static struct auth_data*
2103 az_find_wildcard_domain(struct auth_zone* z, uint8_t* nm, size_t nmlen)
2104 {
2105         uint8_t wc[LDNS_MAX_DOMAINLEN];
2106         if(nmlen+2 > sizeof(wc))
2107                 return NULL; /* result would be too long */
2108         wc[0] = 1; /* length of wildcard label */
2109         wc[1] = (uint8_t)'*'; /* wildcard label */
2110         memmove(wc+2, nm, nmlen);
2111         return az_find_name(z, wc, nmlen+2);
2112 }
2113
2114 /** find wildcard between qname and cename */
2115 static struct auth_data*
2116 az_find_wildcard(struct auth_zone* z, struct query_info* qinfo,
2117         struct auth_data* ce)
2118 {
2119         uint8_t* nm = qinfo->qname;
2120         size_t nmlen = qinfo->qname_len;
2121         struct auth_data* node;
2122         if(!dname_subdomain_c(nm, z->name))
2123                 return NULL; /* out of zone */
2124         while((node=az_find_wildcard_domain(z, nm, nmlen))==NULL) {
2125                 /* see if we can go up to find the wildcard */
2126                 if(nmlen == z->namelen)
2127                         return NULL; /* top of zone reached */
2128                 if(ce && nmlen == ce->namelen)
2129                         return NULL; /* ce reached */
2130                 if(dname_is_root(nm))
2131                         return NULL; /* cannot go up */
2132                 dname_remove_label(&nm, &nmlen);
2133         }
2134         return node;
2135 }
2136
2137 /** domain is not exact, find first candidate ce (name that matches
2138  * a part of qname) in tree */
2139 static struct auth_data*
2140 az_find_candidate_ce(struct auth_zone* z, struct query_info* qinfo,
2141         struct auth_data* n)
2142 {
2143         uint8_t* nm;
2144         size_t nmlen;
2145         if(n) {
2146                 nm = dname_get_shared_topdomain(qinfo->qname, n->name);
2147         } else {
2148                 nm = qinfo->qname;
2149         }
2150         dname_count_size_labels(nm, &nmlen);
2151         n = az_find_name(z, nm, nmlen);
2152         /* delete labels and go up on name */
2153         while(!n) {
2154                 if(dname_is_root(nm))
2155                         return NULL; /* cannot go up */
2156                 dname_remove_label(&nm, &nmlen);
2157                 n = az_find_name(z, nm, nmlen);
2158         }
2159         return n;
2160 }
2161
2162 /** go up the auth tree to next existing name. */
2163 static struct auth_data*
2164 az_domain_go_up(struct auth_zone* z, struct auth_data* n)
2165 {
2166         uint8_t* nm = n->name;
2167         size_t nmlen = n->namelen;
2168         while(!dname_is_root(nm)) {
2169                 dname_remove_label(&nm, &nmlen);
2170                 if((n=az_find_name(z, nm, nmlen)) != NULL)
2171                         return n;
2172         }
2173         return NULL;
2174 }
2175
2176 /** Find the closest encloser, an name that exists and is above the
2177  * qname.
2178  * return true if the node (param node) is existing, nonobscured and
2179  *      can be used to generate answers from.  It is then also node_exact.
2180  * returns false if the node is not good enough (or it wasn't node_exact)
2181  *      in this case the ce can be filled.
2182  *      if ce is NULL, no ce exists, and likely the zone is completely empty,
2183  *      not even with a zone apex.
2184  *      if ce is nonNULL it is the closest enclosing upper name (that exists
2185  *      itself for answer purposes).  That name may have DNAME, NS or wildcard
2186  *      rrset is the closest DNAME or NS rrset that was found.
2187  */
2188 static int
2189 az_find_ce(struct auth_zone* z, struct query_info* qinfo,
2190         struct auth_data* node, int node_exact, struct auth_data** ce,
2191         struct auth_rrset** rrset)
2192 {
2193         struct auth_data* n = node;
2194         *ce = NULL;
2195         *rrset = NULL;
2196         if(!node_exact) {
2197                 /* if not exact, lookup closest exact match */
2198                 n = az_find_candidate_ce(z, qinfo, n);
2199         } else {
2200                 /* if exact, the node itself is the first candidate ce */
2201                 *ce = n;
2202         }
2203
2204         /* no direct answer from nsec3-only domains */
2205         if(n && domain_has_only_nsec3(n)) {
2206                 node_exact = 0;
2207                 *ce = NULL;
2208         }
2209
2210         /* with exact matches, walk up the labels until we find the
2211          * delegation, or DNAME or zone end */
2212         while(n) {
2213                 /* see if the current candidate has issues */
2214                 /* not zone apex and has type NS */
2215                 if(n->namelen != z->namelen &&
2216                         (*rrset=az_domain_rrset(n, LDNS_RR_TYPE_NS)) &&
2217                         /* delegate here, but DS at exact the dp has notype */
2218                         (qinfo->qtype != LDNS_RR_TYPE_DS ||
2219                         n->namelen != qinfo->qname_len)) {
2220                         /* referral */
2221                         /* this is ce and the lowernode is nonexisting */
2222                         *ce = n;
2223                         return 0;
2224                 }
2225                 /* not equal to qname and has type DNAME */
2226                 if(n->namelen != qinfo->qname_len &&
2227                         (*rrset=az_domain_rrset(n, LDNS_RR_TYPE_DNAME))) {
2228                         /* this is ce and the lowernode is nonexisting */
2229                         *ce = n;
2230                         return 0;
2231                 }
2232
2233                 if(*ce == NULL && !domain_has_only_nsec3(n)) {
2234                         /* if not found yet, this exact name must be
2235                          * our lowest match (but not nsec3onlydomain) */
2236                         *ce = n;
2237                 }
2238
2239                 /* walk up the tree by removing labels from name and lookup */
2240                 n = az_domain_go_up(z, n);
2241         }
2242         /* found no problems, if it was an exact node, it is fine to use */
2243         return node_exact;
2244 }
2245
2246 /** add additional A/AAAA from domain names in rrset rdata (+offset)
2247  * offset is number of bytes in rdata where the dname is located. */
2248 static int
2249 az_add_additionals_from(struct auth_zone* z, struct regional* region,
2250         struct dns_msg* msg, struct auth_rrset* rrset, size_t offset)
2251 {
2252         struct packed_rrset_data* d = rrset->data;
2253         size_t i;
2254         if(!d) return 0;
2255         for(i=0; i<d->count; i++) {
2256                 size_t dlen;
2257                 struct auth_data* domain;
2258                 struct auth_rrset* ref;
2259                 if(d->rr_len[i] < 2+offset)
2260                         continue; /* too short */
2261                 if(!(dlen = dname_valid(d->rr_data[i]+2+offset,
2262                         d->rr_len[i]-2-offset)))
2263                         continue; /* malformed */
2264                 domain = az_find_name(z, d->rr_data[i]+2+offset, dlen);
2265                 if(!domain)
2266                         continue;
2267                 if((ref=az_domain_rrset(domain, LDNS_RR_TYPE_A)) != NULL) {
2268                         if(!msg_add_rrset_ar(z, region, msg, domain, ref))
2269                                 return 0;
2270                 }
2271                 if((ref=az_domain_rrset(domain, LDNS_RR_TYPE_AAAA)) != NULL) {
2272                         if(!msg_add_rrset_ar(z, region, msg, domain, ref))
2273                                 return 0;
2274                 }
2275         }
2276         return 1;
2277 }
2278
2279 /** add negative SOA record (with negative TTL) */
2280 static int
2281 az_add_negative_soa(struct auth_zone* z, struct regional* region,
2282         struct dns_msg* msg)
2283 {
2284         uint32_t minimum;
2285         struct packed_rrset_data* d;
2286         struct auth_rrset* soa;
2287         struct auth_data* apex = az_find_name(z, z->name, z->namelen);
2288         if(!apex) return 0;
2289         soa = az_domain_rrset(apex, LDNS_RR_TYPE_SOA);
2290         if(!soa) return 0;
2291         /* must be first to put in message; we want to fix the TTL with
2292          * one RRset here, otherwise we'd need to loop over the RRs to get
2293          * the resulting lower TTL */
2294         log_assert(msg->rep->rrset_count == 0);
2295         if(!msg_add_rrset_ns(z, region, msg, apex, soa)) return 0;
2296         /* fixup TTL */
2297         d = (struct packed_rrset_data*)msg->rep->rrsets[msg->rep->rrset_count-1]->entry.data;
2298         /* last 4 bytes are minimum ttl in network format */
2299         if(d->count == 0) return 0;
2300         if(d->rr_len[0] < 2+4) return 0;
2301         minimum = sldns_read_uint32(d->rr_data[0]+(d->rr_len[0]-4));
2302         d->ttl = (time_t)minimum;
2303         d->rr_ttl[0] = (time_t)minimum;
2304         msg->rep->ttl = get_rrset_ttl(msg->rep->rrsets[0]);
2305         msg->rep->prefetch_ttl = PREFETCH_TTL_CALC(msg->rep->ttl);
2306         msg->rep->serve_expired_ttl = msg->rep->ttl + SERVE_EXPIRED_TTL;
2307         return 1;
2308 }
2309
2310 /** See if the query goes to empty nonterminal (that has no auth_data,
2311  * but there are nodes underneath.  We already checked that there are
2312  * not NS, or DNAME above, so that we only need to check if some node
2313  * exists below (with nonempty rr list), return true if emptynonterminal */
2314 static int
2315 az_empty_nonterminal(struct auth_zone* z, struct query_info* qinfo,
2316         struct auth_data* node)
2317 {
2318         struct auth_data* next;
2319         if(!node) {
2320                 /* no smaller was found, use first (smallest) node as the
2321                  * next one */
2322                 next = (struct auth_data*)rbtree_first(&z->data);
2323         } else {
2324                 next = (struct auth_data*)rbtree_next(&node->node);
2325         }
2326         while(next && (rbnode_type*)next != RBTREE_NULL && next->rrsets == NULL) {
2327                 /* the next name has empty rrsets, is an empty nonterminal
2328                  * itself, see if there exists something below it */
2329                 next = (struct auth_data*)rbtree_next(&node->node);
2330         }
2331         if((rbnode_type*)next == RBTREE_NULL || !next) {
2332                 /* there is no next node, so something below it cannot
2333                  * exist */
2334                 return 0;
2335         }
2336         /* a next node exists, if there was something below the query,
2337          * this node has to be it.  See if it is below the query name */
2338         if(dname_strict_subdomain_c(next->name, qinfo->qname))
2339                 return 1;
2340         return 0;
2341 }
2342
2343 /** create synth cname target name in buffer, or fail if too long */
2344 static size_t
2345 synth_cname_buf(uint8_t* qname, size_t qname_len, size_t dname_len,
2346         uint8_t* dtarg, size_t dtarglen, uint8_t* buf, size_t buflen)
2347 {
2348         size_t newlen = qname_len + dtarglen - dname_len;
2349         if(newlen > buflen) {
2350                 /* YXDOMAIN error */
2351                 return 0;
2352         }
2353         /* new name is concatenation of qname front (without DNAME owner)
2354          * and DNAME target name */
2355         memcpy(buf, qname, qname_len-dname_len);
2356         memmove(buf+(qname_len-dname_len), dtarg, dtarglen);
2357         return newlen;
2358 }
2359
2360 /** create synthetic CNAME rrset for in a DNAME answer in region,
2361  * false on alloc failure, cname==NULL when name too long. */
2362 static int
2363 create_synth_cname(uint8_t* qname, size_t qname_len, struct regional* region,
2364         struct auth_data* node, struct auth_rrset* dname, uint16_t dclass,
2365         struct ub_packed_rrset_key** cname)
2366 {
2367         uint8_t buf[LDNS_MAX_DOMAINLEN];
2368         uint8_t* dtarg;
2369         size_t dtarglen, newlen;
2370         struct packed_rrset_data* d;
2371
2372         /* get DNAME target name */
2373         if(dname->data->count < 1) return 0;
2374         if(dname->data->rr_len[0] < 3) return 0; /* at least rdatalen +1 */
2375         dtarg = dname->data->rr_data[0]+2;
2376         dtarglen = dname->data->rr_len[0]-2;
2377         if(sldns_read_uint16(dname->data->rr_data[0]) != dtarglen)
2378                 return 0; /* rdatalen in DNAME rdata is malformed */
2379         if(dname_valid(dtarg, dtarglen) != dtarglen)
2380                 return 0; /* DNAME RR has malformed rdata */
2381
2382         /* synthesize a CNAME */
2383         newlen = synth_cname_buf(qname, qname_len, node->namelen,
2384                 dtarg, dtarglen, buf, sizeof(buf));
2385         if(newlen == 0) {
2386                 /* YXDOMAIN error */
2387                 *cname = NULL;
2388                 return 1;
2389         }
2390         *cname = (struct ub_packed_rrset_key*)regional_alloc(region,
2391                 sizeof(struct ub_packed_rrset_key));
2392         if(!*cname)
2393                 return 0; /* out of memory */
2394         memset(&(*cname)->entry, 0, sizeof((*cname)->entry));
2395         (*cname)->entry.key = (*cname);
2396         (*cname)->rk.type = htons(LDNS_RR_TYPE_CNAME);
2397         (*cname)->rk.rrset_class = htons(dclass);
2398         (*cname)->rk.flags = 0;
2399         (*cname)->rk.dname = regional_alloc_init(region, qname, qname_len);
2400         if(!(*cname)->rk.dname)
2401                 return 0; /* out of memory */
2402         (*cname)->rk.dname_len = qname_len;
2403         (*cname)->entry.hash = rrset_key_hash(&(*cname)->rk);
2404         d = (struct packed_rrset_data*)regional_alloc_zero(region,
2405                 sizeof(struct packed_rrset_data) + sizeof(size_t) +
2406                 sizeof(uint8_t*) + sizeof(time_t) + sizeof(uint16_t)
2407                 + newlen);
2408         if(!d)
2409                 return 0; /* out of memory */
2410         (*cname)->entry.data = d;
2411         d->ttl = 0; /* 0 for synthesized CNAME TTL */
2412         d->count = 1;
2413         d->rrsig_count = 0;
2414         d->trust = rrset_trust_ans_noAA;
2415         d->rr_len = (size_t*)((uint8_t*)d +
2416                 sizeof(struct packed_rrset_data));
2417         d->rr_len[0] = newlen + sizeof(uint16_t);
2418         packed_rrset_ptr_fixup(d);
2419         d->rr_ttl[0] = d->ttl;
2420         sldns_write_uint16(d->rr_data[0], newlen);
2421         memmove(d->rr_data[0] + sizeof(uint16_t), buf, newlen);
2422         return 1;
2423 }
2424
2425 /** add a synthesized CNAME to the answer section */
2426 static int
2427 add_synth_cname(struct auth_zone* z, uint8_t* qname, size_t qname_len,
2428         struct regional* region, struct dns_msg* msg, struct auth_data* dname,
2429         struct auth_rrset* rrset)
2430 {
2431         struct ub_packed_rrset_key* cname;
2432         /* synthesize a CNAME */
2433         if(!create_synth_cname(qname, qname_len, region, dname, rrset,
2434                 z->dclass, &cname)) {
2435                 /* out of memory */
2436                 return 0;
2437         }
2438         if(!cname) {
2439                 /* cname cannot be create because of YXDOMAIN */
2440                 msg->rep->flags |= LDNS_RCODE_YXDOMAIN;
2441                 return 1;
2442         }
2443         /* add cname to message */
2444         if(!msg_grow_array(region, msg))
2445                 return 0;
2446         msg->rep->rrsets[msg->rep->rrset_count] = cname;
2447         msg->rep->rrset_count++;
2448         msg->rep->an_numrrsets++;
2449         msg_ttl(msg);
2450         return 1;
2451 }
2452
2453 /** Change a dname to a different one, for wildcard namechange */
2454 static void
2455 az_change_dnames(struct dns_msg* msg, uint8_t* oldname, uint8_t* newname,
2456         size_t newlen, int an_only)
2457 {
2458         size_t i;
2459         size_t start = 0, end = msg->rep->rrset_count;
2460         if(!an_only) start = msg->rep->an_numrrsets;
2461         if(an_only) end = msg->rep->an_numrrsets;
2462         for(i=start; i<end; i++) {
2463                 /* allocated in region so we can change the ptrs */
2464                 if(query_dname_compare(msg->rep->rrsets[i]->rk.dname, oldname)
2465                         == 0) {
2466                         msg->rep->rrsets[i]->rk.dname = newname;
2467                         msg->rep->rrsets[i]->rk.dname_len = newlen;
2468                 }
2469         }
2470 }
2471
2472 /** find NSEC record covering the query */
2473 static struct auth_rrset*
2474 az_find_nsec_cover(struct auth_zone* z, struct auth_data** node)
2475 {
2476         uint8_t* nm = (*node)->name;
2477         size_t nmlen = (*node)->namelen;
2478         struct auth_rrset* rrset;
2479         /* find the NSEC for the smallest-or-equal node */
2480         /* if node == NULL, we did not find a smaller name.  But the zone
2481          * name is the smallest name and should have an NSEC. So there is
2482          * no NSEC to return (for a properly signed zone) */
2483         /* for empty nonterminals, the auth-data node should not exist,
2484          * and thus we don't need to go rbtree_previous here to find
2485          * a domain with an NSEC record */
2486         /* but there could be glue, and if this is node, then it has no NSEC.
2487          * Go up to find nonglue (previous) NSEC-holding nodes */
2488         while((rrset=az_domain_rrset(*node, LDNS_RR_TYPE_NSEC)) == NULL) {
2489                 if(dname_is_root(nm)) return NULL;
2490                 if(nmlen == z->namelen) return NULL;
2491                 dname_remove_label(&nm, &nmlen);
2492                 /* adjust *node for the nsec rrset to find in */
2493                 *node = az_find_name(z, nm, nmlen);
2494         }
2495         return rrset;
2496 }
2497
2498 /** Find NSEC and add for wildcard denial */
2499 static int
2500 az_nsec_wildcard_denial(struct auth_zone* z, struct regional* region,
2501         struct dns_msg* msg, uint8_t* cenm, size_t cenmlen)
2502 {
2503         struct query_info qinfo;
2504         int node_exact;
2505         struct auth_data* node;
2506         struct auth_rrset* nsec;
2507         uint8_t wc[LDNS_MAX_DOMAINLEN];
2508         if(cenmlen+2 > sizeof(wc))
2509                 return 0; /* result would be too long */
2510         wc[0] = 1; /* length of wildcard label */
2511         wc[1] = (uint8_t)'*'; /* wildcard label */
2512         memmove(wc+2, cenm, cenmlen);
2513
2514         /* we have '*.ce' in wc wildcard name buffer */
2515         /* get nsec cover for that */
2516         qinfo.qname = wc;
2517         qinfo.qname_len = cenmlen+2;
2518         qinfo.qtype = 0;
2519         qinfo.qclass = 0;
2520         az_find_domain(z, &qinfo, &node_exact, &node);
2521         if((nsec=az_find_nsec_cover(z, &node)) != NULL) {
2522                 if(!msg_add_rrset_ns(z, region, msg, node, nsec)) return 0;
2523         }
2524         return 1;
2525 }
2526
2527 /** Find the NSEC3PARAM rrset (if any) and if true you have the parameters */
2528 static int
2529 az_nsec3_param(struct auth_zone* z, int* algo, size_t* iter, uint8_t** salt,
2530         size_t* saltlen)
2531 {
2532         struct auth_data* apex;
2533         struct auth_rrset* param;
2534         size_t i;
2535         apex = az_find_name(z, z->name, z->namelen);
2536         if(!apex) return 0;
2537         param = az_domain_rrset(apex, LDNS_RR_TYPE_NSEC3PARAM);
2538         if(!param || param->data->count==0)
2539                 return 0; /* no RRset or no RRs in rrset */
2540         /* find out which NSEC3PARAM RR has supported parameters */
2541         /* skip unknown flags (dynamic signer is recalculating nsec3 chain) */
2542         for(i=0; i<param->data->count; i++) {
2543                 uint8_t* rdata = param->data->rr_data[i]+2;
2544                 size_t rdatalen = param->data->rr_len[i];
2545                 if(rdatalen < 2+5)
2546                         continue; /* too short */
2547                 if(!nsec3_hash_algo_size_supported((int)(rdata[0])))
2548                         continue; /* unsupported algo */
2549                 if(rdatalen < (size_t)(2+5+(size_t)rdata[4]))
2550                         continue; /* salt missing */
2551                 if((rdata[1]&NSEC3_UNKNOWN_FLAGS)!=0)
2552                         continue; /* unknown flags */
2553                 *algo = (int)(rdata[0]);
2554                 *iter = sldns_read_uint16(rdata+2);
2555                 *saltlen = rdata[4];
2556                 if(*saltlen == 0)
2557                         *salt = NULL;
2558                 else    *salt = rdata+5;
2559                 return 1;
2560         }
2561         /* no supported params */
2562         return 0;
2563 }
2564
2565 /** Hash a name with nsec3param into buffer, it has zone name appended.
2566  * return length of hash */
2567 static size_t
2568 az_nsec3_hash(uint8_t* buf, size_t buflen, uint8_t* nm, size_t nmlen,
2569         int algo, size_t iter, uint8_t* salt, size_t saltlen)
2570 {
2571         size_t hlen = nsec3_hash_algo_size_supported(algo);
2572         /* buffer has domain name, nsec3hash, and 256 is for max saltlen
2573          * (salt has 0-255 length) */
2574         unsigned char p[LDNS_MAX_DOMAINLEN+1+N3HASHBUFLEN+256];
2575         size_t i;
2576         if(nmlen+saltlen > sizeof(p) || hlen+saltlen > sizeof(p))
2577                 return 0;
2578         if(hlen > buflen)
2579                 return 0; /* somehow too large for destination buffer */
2580         /* hashfunc(name, salt) */
2581         memmove(p, nm, nmlen);
2582         query_dname_tolower(p);
2583         memmove(p+nmlen, salt, saltlen);
2584         (void)secalgo_nsec3_hash(algo, p, nmlen+saltlen, (unsigned char*)buf);
2585         for(i=0; i<iter; i++) {
2586                 /* hashfunc(hash, salt) */
2587                 memmove(p, buf, hlen);
2588                 memmove(p+hlen, salt, saltlen);
2589                 (void)secalgo_nsec3_hash(algo, p, hlen+saltlen,
2590                         (unsigned char*)buf);
2591         }
2592         return hlen;
2593 }
2594
2595 /** Hash name and return b32encoded hashname for lookup, zone name appended */
2596 static int
2597 az_nsec3_hashname(struct auth_zone* z, uint8_t* hashname, size_t* hashnmlen,
2598         uint8_t* nm, size_t nmlen, int algo, size_t iter, uint8_t* salt,
2599         size_t saltlen)
2600 {
2601         uint8_t hash[N3HASHBUFLEN];
2602         size_t hlen;
2603         int ret;
2604         hlen = az_nsec3_hash(hash, sizeof(hash), nm, nmlen, algo, iter,
2605                 salt, saltlen);
2606         if(!hlen) return 0;
2607         /* b32 encode */
2608         if(*hashnmlen < hlen*2+1+z->namelen) /* approx b32 as hexb16 */
2609                 return 0;
2610         ret = sldns_b32_ntop_extended_hex(hash, hlen, (char*)(hashname+1),
2611                 (*hashnmlen)-1);
2612         if(ret<1)
2613                 return 0;
2614         hashname[0] = (uint8_t)ret;
2615         ret++;
2616         if((*hashnmlen) - ret < z->namelen)
2617                 return 0;
2618         memmove(hashname+ret, z->name, z->namelen);
2619         *hashnmlen = z->namelen+(size_t)ret;
2620         return 1;
2621 }
2622
2623 /** Find the datanode that covers the nsec3hash-name */
2624 static struct auth_data*
2625 az_nsec3_findnode(struct auth_zone* z, uint8_t* hashnm, size_t hashnmlen)
2626 {
2627         struct query_info qinfo;
2628         struct auth_data* node;
2629         int node_exact;
2630         qinfo.qclass = 0;
2631         qinfo.qtype = 0;
2632         qinfo.qname = hashnm;
2633         qinfo.qname_len = hashnmlen;
2634         /* because canonical ordering and b32 nsec3 ordering are the same.
2635          * this is a good lookup to find the nsec3 name. */
2636         az_find_domain(z, &qinfo, &node_exact, &node);
2637         /* but we may have to skip non-nsec3 nodes */
2638         /* this may be a lot, the way to speed that up is to have a
2639          * separate nsec3 tree with nsec3 nodes */
2640         while(node && (rbnode_type*)node != RBTREE_NULL &&
2641                 !az_domain_rrset(node, LDNS_RR_TYPE_NSEC3)) {
2642                 node = (struct auth_data*)rbtree_previous(&node->node);
2643         }
2644         if((rbnode_type*)node == RBTREE_NULL)
2645                 node = NULL;
2646         return node;
2647 }
2648
2649 /** Find cover for hashed(nm, nmlen) (or NULL) */
2650 static struct auth_data*
2651 az_nsec3_find_cover(struct auth_zone* z, uint8_t* nm, size_t nmlen,
2652         int algo, size_t iter, uint8_t* salt, size_t saltlen)
2653 {
2654         struct auth_data* node;
2655         uint8_t hname[LDNS_MAX_DOMAINLEN];
2656         size_t hlen = sizeof(hname);
2657         if(!az_nsec3_hashname(z, hname, &hlen, nm, nmlen, algo, iter,
2658                 salt, saltlen))
2659                 return NULL;
2660         node = az_nsec3_findnode(z, hname, hlen);
2661         if(node)
2662                 return node;
2663         /* we did not find any, perhaps because the NSEC3 hash is before
2664          * the first hash, we have to find the 'last hash' in the zone */
2665         node = (struct auth_data*)rbtree_last(&z->data);
2666         while(node && (rbnode_type*)node != RBTREE_NULL &&
2667                 !az_domain_rrset(node, LDNS_RR_TYPE_NSEC3)) {
2668                 node = (struct auth_data*)rbtree_previous(&node->node);
2669         }
2670         if((rbnode_type*)node == RBTREE_NULL)
2671                 node = NULL;
2672         return node;
2673 }
2674
2675 /** Find exact match for hashed(nm, nmlen) NSEC3 record or NULL */
2676 static struct auth_data*
2677 az_nsec3_find_exact(struct auth_zone* z, uint8_t* nm, size_t nmlen,
2678         int algo, size_t iter, uint8_t* salt, size_t saltlen)
2679 {
2680         struct auth_data* node;
2681         uint8_t hname[LDNS_MAX_DOMAINLEN];
2682         size_t hlen = sizeof(hname);
2683         if(!az_nsec3_hashname(z, hname, &hlen, nm, nmlen, algo, iter,
2684                 salt, saltlen))
2685                 return NULL;
2686         node = az_find_name(z, hname, hlen);
2687         if(az_domain_rrset(node, LDNS_RR_TYPE_NSEC3))
2688                 return node;
2689         return NULL;
2690 }
2691
2692 /** Return nextcloser name (as a ref into the qname).  This is one label
2693  * more than the cenm (cename must be a suffix of qname) */
2694 static void
2695 az_nsec3_get_nextcloser(uint8_t* cenm, uint8_t* qname, size_t qname_len,
2696         uint8_t** nx, size_t* nxlen)
2697 {
2698         int celabs = dname_count_labels(cenm);
2699         int qlabs = dname_count_labels(qname);
2700         int strip = qlabs - celabs -1;
2701         log_assert(dname_strict_subdomain(qname, qlabs, cenm, celabs));
2702         *nx = qname;
2703         *nxlen = qname_len;
2704         if(strip>0)
2705                 dname_remove_labels(nx, nxlen, strip);
2706 }
2707
2708 /** Find the closest encloser that has exact NSEC3.
2709  * updated cenm to the new name. If it went up no-exact-ce is true. */
2710 static struct auth_data*
2711 az_nsec3_find_ce(struct auth_zone* z, uint8_t** cenm, size_t* cenmlen,
2712         int* no_exact_ce, int algo, size_t iter, uint8_t* salt, size_t saltlen)
2713 {
2714         struct auth_data* node;
2715         while((node = az_nsec3_find_exact(z, *cenm, *cenmlen,
2716                 algo, iter, salt, saltlen)) == NULL) {
2717                 if(*cenmlen == z->namelen) {
2718                         /* next step up would take us out of the zone. fail */
2719                         return NULL;
2720                 }
2721                 *no_exact_ce = 1;
2722                 dname_remove_label(cenm, cenmlen);
2723         }
2724         return node;
2725 }
2726
2727 /* Insert NSEC3 record in authority section, if NULL does nothing */
2728 static int
2729 az_nsec3_insert(struct auth_zone* z, struct regional* region,
2730         struct dns_msg* msg, struct auth_data* node)
2731 {
2732         struct auth_rrset* nsec3;
2733         if(!node) return 1; /* no node, skip this */
2734         nsec3 = az_domain_rrset(node, LDNS_RR_TYPE_NSEC3);
2735         if(!nsec3) return 1; /* if no nsec3 RR, skip it */
2736         if(!msg_add_rrset_ns(z, region, msg, node, nsec3)) return 0;
2737         return 1;
2738 }
2739
2740 /** add NSEC3 records to the zone for the nsec3 proof.
2741  * Specify with the flags with parts of the proof are required.
2742  * the ce is the exact matching name (for notype) but also delegation points.
2743  * qname is the one where the nextcloser name can be derived from.
2744  * If NSEC3 is not properly there (in the zone) nothing is added.
2745  * always enabled: include nsec3 proving about the Closest Encloser.
2746  *      that is an exact match that should exist for it.
2747  *      If that does not exist, a higher exact match + nxproof is enabled
2748  *      (for some sort of opt-out empty nonterminal cases).
2749  * ceproof: include ce proof NSEC3 (omitted for wildcard replies).
2750  * nxproof: include denial of the qname.
2751  * wcproof: include denial of wildcard (wildcard.ce).
2752  */
2753 static int
2754 az_add_nsec3_proof(struct auth_zone* z, struct regional* region,
2755         struct dns_msg* msg, uint8_t* cenm, size_t cenmlen, uint8_t* qname,
2756         size_t qname_len, int ceproof, int nxproof, int wcproof)
2757 {
2758         int algo;
2759         size_t iter, saltlen;
2760         uint8_t* salt;
2761         int no_exact_ce = 0;
2762         struct auth_data* node;
2763
2764         /* find parameters of nsec3 proof */
2765         if(!az_nsec3_param(z, &algo, &iter, &salt, &saltlen))
2766                 return 1; /* no nsec3 */
2767         /* find ce that has an NSEC3 */
2768         if(ceproof) {
2769                 node = az_nsec3_find_ce(z, &cenm, &cenmlen, &no_exact_ce,
2770                         algo, iter, salt, saltlen);
2771                 if(no_exact_ce) nxproof = 1;
2772                 if(!az_nsec3_insert(z, region, msg, node))
2773                         return 0;
2774         }
2775
2776         if(nxproof) {
2777                 uint8_t* nx;
2778                 size_t nxlen;
2779                 /* create nextcloser domain name */
2780                 az_nsec3_get_nextcloser(cenm, qname, qname_len, &nx, &nxlen);
2781                 /* find nsec3 that matches or covers it */
2782                 node = az_nsec3_find_cover(z, nx, nxlen, algo, iter, salt,
2783                         saltlen);
2784                 if(!az_nsec3_insert(z, region, msg, node))
2785                         return 0;
2786         }
2787         if(wcproof) {
2788                 /* create wildcard name *.ce */
2789                 uint8_t wc[LDNS_MAX_DOMAINLEN];
2790                 size_t wclen;
2791                 if(cenmlen+2 > sizeof(wc))
2792                         return 0; /* result would be too long */
2793                 wc[0] = 1; /* length of wildcard label */
2794                 wc[1] = (uint8_t)'*'; /* wildcard label */
2795                 memmove(wc+2, cenm, cenmlen);
2796                 wclen = cenmlen+2;
2797                 /* find nsec3 that matches or covers it */
2798                 node = az_nsec3_find_cover(z, wc, wclen, algo, iter, salt,
2799                         saltlen);
2800                 if(!az_nsec3_insert(z, region, msg, node))
2801                         return 0;
2802         }
2803         return 1;
2804 }
2805
2806 /** generate answer for positive answer */
2807 static int
2808 az_generate_positive_answer(struct auth_zone* z, struct regional* region,
2809         struct dns_msg* msg, struct auth_data* node, struct auth_rrset* rrset)
2810 {
2811         if(!msg_add_rrset_an(z, region, msg, node, rrset)) return 0;
2812         /* see if we want additional rrs */
2813         if(rrset->type == LDNS_RR_TYPE_MX) {
2814                 if(!az_add_additionals_from(z, region, msg, rrset, 2))
2815                         return 0;
2816         } else if(rrset->type == LDNS_RR_TYPE_SRV) {
2817                 if(!az_add_additionals_from(z, region, msg, rrset, 6))
2818                         return 0;
2819         } else if(rrset->type == LDNS_RR_TYPE_NS) {
2820                 if(!az_add_additionals_from(z, region, msg, rrset, 0))
2821                         return 0;
2822         }
2823         return 1;
2824 }
2825
2826 /** generate answer for type ANY answer */
2827 static int
2828 az_generate_any_answer(struct auth_zone* z, struct regional* region,
2829         struct dns_msg* msg, struct auth_data* node)
2830 {
2831         struct auth_rrset* rrset;
2832         int added = 0;
2833         /* add a couple (at least one) RRs */
2834         if((rrset=az_domain_rrset(node, LDNS_RR_TYPE_SOA)) != NULL) {
2835                 if(!msg_add_rrset_an(z, region, msg, node, rrset)) return 0;
2836                 added++;
2837         }
2838         if((rrset=az_domain_rrset(node, LDNS_RR_TYPE_MX)) != NULL) {
2839                 if(!msg_add_rrset_an(z, region, msg, node, rrset)) return 0;
2840                 added++;
2841         }
2842         if((rrset=az_domain_rrset(node, LDNS_RR_TYPE_A)) != NULL) {
2843                 if(!msg_add_rrset_an(z, region, msg, node, rrset)) return 0;
2844                 added++;
2845         }
2846         if((rrset=az_domain_rrset(node, LDNS_RR_TYPE_AAAA)) != NULL) {
2847                 if(!msg_add_rrset_an(z, region, msg, node, rrset)) return 0;
2848                 added++;
2849         }
2850         if(added == 0 && node && node->rrsets) {
2851                 if(!msg_add_rrset_an(z, region, msg, node,
2852                         node->rrsets)) return 0;
2853         }
2854         return 1;
2855 }
2856
2857 /** follow cname chain and add more data to the answer section */
2858 static int
2859 follow_cname_chain(struct auth_zone* z, uint16_t qtype,
2860         struct regional* region, struct dns_msg* msg,
2861         struct packed_rrset_data* d)
2862 {
2863         int maxchain = 0;
2864         /* see if we can add the target of the CNAME into the answer */
2865         while(maxchain++ < MAX_CNAME_CHAIN) {
2866                 struct auth_data* node;
2867                 struct auth_rrset* rrset;
2868                 size_t clen;
2869                 /* d has cname rdata */
2870                 if(d->count == 0) break; /* no CNAME */
2871                 if(d->rr_len[0] < 2+1) break; /* too small */
2872                 if((clen=dname_valid(d->rr_data[0]+2, d->rr_len[0]-2))==0)
2873                         break; /* malformed */
2874                 if(!dname_subdomain_c(d->rr_data[0]+2, z->name))
2875                         break; /* target out of zone */
2876                 if((node = az_find_name(z, d->rr_data[0]+2, clen))==NULL)
2877                         break; /* no such target name */
2878                 if((rrset=az_domain_rrset(node, qtype))!=NULL) {
2879                         /* done we found the target */
2880                         if(!msg_add_rrset_an(z, region, msg, node, rrset))
2881                                 return 0;
2882                         break;
2883                 }
2884                 if((rrset=az_domain_rrset(node, LDNS_RR_TYPE_CNAME))==NULL)
2885                         break; /* no further CNAME chain, notype */
2886                 if(!msg_add_rrset_an(z, region, msg, node, rrset)) return 0;
2887                 d = rrset->data;
2888         }
2889         return 1;
2890 }
2891
2892 /** generate answer for cname answer */
2893 static int
2894 az_generate_cname_answer(struct auth_zone* z, struct query_info* qinfo,
2895         struct regional* region, struct dns_msg* msg,
2896         struct auth_data* node, struct auth_rrset* rrset)
2897 {
2898         if(!msg_add_rrset_an(z, region, msg, node, rrset)) return 0;
2899         if(!rrset) return 1;
2900         if(!follow_cname_chain(z, qinfo->qtype, region, msg, rrset->data))
2901                 return 0;
2902         return 1;
2903 }
2904
2905 /** generate answer for notype answer */
2906 static int
2907 az_generate_notype_answer(struct auth_zone* z, struct regional* region,
2908         struct dns_msg* msg, struct auth_data* node)
2909 {
2910         struct auth_rrset* rrset;
2911         if(!az_add_negative_soa(z, region, msg)) return 0;
2912         /* DNSSEC denial NSEC */
2913         if((rrset=az_domain_rrset(node, LDNS_RR_TYPE_NSEC))!=NULL) {
2914                 if(!msg_add_rrset_ns(z, region, msg, node, rrset)) return 0;
2915         } else if(node) {
2916                 /* DNSSEC denial NSEC3 */
2917                 if(!az_add_nsec3_proof(z, region, msg, node->name,
2918                         node->namelen, msg->qinfo.qname,
2919                         msg->qinfo.qname_len, 1, 0, 0))
2920                         return 0;
2921         }
2922         return 1;
2923 }
2924
2925 /** generate answer for referral answer */
2926 static int
2927 az_generate_referral_answer(struct auth_zone* z, struct regional* region,
2928         struct dns_msg* msg, struct auth_data* ce, struct auth_rrset* rrset)
2929 {
2930         struct auth_rrset* ds, *nsec;
2931         /* turn off AA flag, referral is nonAA because it leaves the zone */
2932         log_assert(ce);
2933         msg->rep->flags &= ~BIT_AA;
2934         if(!msg_add_rrset_ns(z, region, msg, ce, rrset)) return 0;
2935         /* add DS or deny it */
2936         if((ds=az_domain_rrset(ce, LDNS_RR_TYPE_DS))!=NULL) {
2937                 if(!msg_add_rrset_ns(z, region, msg, ce, ds)) return 0;
2938         } else {
2939                 /* deny the DS */
2940                 if((nsec=az_domain_rrset(ce, LDNS_RR_TYPE_NSEC))!=NULL) {
2941                         if(!msg_add_rrset_ns(z, region, msg, ce, nsec))
2942                                 return 0;
2943                 } else {
2944                         if(!az_add_nsec3_proof(z, region, msg, ce->name,
2945                                 ce->namelen, msg->qinfo.qname,
2946                                 msg->qinfo.qname_len, 1, 0, 0))
2947                                 return 0;
2948                 }
2949         }
2950         /* add additional rrs for type NS */
2951         if(!az_add_additionals_from(z, region, msg, rrset, 0)) return 0;
2952         return 1;
2953 }
2954
2955 /** generate answer for DNAME answer */
2956 static int
2957 az_generate_dname_answer(struct auth_zone* z, struct query_info* qinfo,
2958         struct regional* region, struct dns_msg* msg, struct auth_data* ce,
2959         struct auth_rrset* rrset)
2960 {
2961         log_assert(ce);
2962         /* add the DNAME and then a CNAME */
2963         if(!msg_add_rrset_an(z, region, msg, ce, rrset)) return 0;
2964         if(!add_synth_cname(z, qinfo->qname, qinfo->qname_len, region,
2965                 msg, ce, rrset)) return 0;
2966         if(FLAGS_GET_RCODE(msg->rep->flags) == LDNS_RCODE_YXDOMAIN)
2967                 return 1;
2968         if(msg->rep->rrset_count == 0 ||
2969                 !msg->rep->rrsets[msg->rep->rrset_count-1])
2970                 return 0;
2971         if(!follow_cname_chain(z, qinfo->qtype, region, msg,
2972                 (struct packed_rrset_data*)msg->rep->rrsets[
2973                 msg->rep->rrset_count-1]->entry.data))
2974                 return 0;
2975         return 1;
2976 }
2977
2978 /** generate answer for wildcard answer */
2979 static int
2980 az_generate_wildcard_answer(struct auth_zone* z, struct query_info* qinfo,
2981         struct regional* region, struct dns_msg* msg, struct auth_data* ce,
2982         struct auth_data* wildcard, struct auth_data* node)
2983 {
2984         struct auth_rrset* rrset, *nsec;
2985         if((rrset=az_domain_rrset(wildcard, qinfo->qtype)) != NULL) {
2986                 /* wildcard has type, add it */
2987                 if(!msg_add_rrset_an(z, region, msg, wildcard, rrset))
2988                         return 0;
2989                 az_change_dnames(msg, wildcard->name, msg->qinfo.qname,
2990                         msg->qinfo.qname_len, 1);
2991         } else if((rrset=az_domain_rrset(wildcard, LDNS_RR_TYPE_CNAME))!=NULL) {
2992                 /* wildcard has cname instead, do that */
2993                 if(!msg_add_rrset_an(z, region, msg, wildcard, rrset))
2994                         return 0;
2995                 az_change_dnames(msg, wildcard->name, msg->qinfo.qname,
2996                         msg->qinfo.qname_len, 1);
2997                 if(!follow_cname_chain(z, qinfo->qtype, region, msg,
2998                         rrset->data))
2999                         return 0;
3000         } else if(qinfo->qtype == LDNS_RR_TYPE_ANY && wildcard->rrsets) {
3001                 /* add ANY rrsets from wildcard node */
3002                 if(!az_generate_any_answer(z, region, msg, wildcard))
3003                         return 0;
3004                 az_change_dnames(msg, wildcard->name, msg->qinfo.qname,
3005                         msg->qinfo.qname_len, 1);
3006         } else {
3007                 /* wildcard has nodata, notype answer */
3008                 /* call other notype routine for dnssec notype denials */
3009                 if(!az_generate_notype_answer(z, region, msg, wildcard))
3010                         return 0;
3011         }
3012
3013         /* ce and node for dnssec denial of wildcard original name */
3014         if((nsec=az_find_nsec_cover(z, &node)) != NULL) {
3015                 if(!msg_add_rrset_ns(z, region, msg, node, nsec)) return 0;
3016         } else if(ce) {
3017                 uint8_t* wildup = wildcard->name;
3018                 size_t wilduplen= wildcard->namelen;
3019                 dname_remove_label(&wildup, &wilduplen);
3020                 if(!az_add_nsec3_proof(z, region, msg, wildup,
3021                         wilduplen, msg->qinfo.qname,
3022                         msg->qinfo.qname_len, 0, 1, 0))
3023                         return 0;
3024         }
3025
3026         /* fixup name of wildcard from *.zone to qname, use already allocated
3027          * pointer to msg qname */
3028         az_change_dnames(msg, wildcard->name, msg->qinfo.qname,
3029                 msg->qinfo.qname_len, 0);
3030         return 1;
3031 }
3032
3033 /** generate answer for nxdomain answer */
3034 static int
3035 az_generate_nxdomain_answer(struct auth_zone* z, struct regional* region,
3036         struct dns_msg* msg, struct auth_data* ce, struct auth_data* node)
3037 {
3038         struct auth_rrset* nsec;
3039         msg->rep->flags |= LDNS_RCODE_NXDOMAIN;
3040         if(!az_add_negative_soa(z, region, msg)) return 0;
3041         if((nsec=az_find_nsec_cover(z, &node)) != NULL) {
3042                 if(!msg_add_rrset_ns(z, region, msg, node, nsec)) return 0;
3043                 if(ce && !az_nsec_wildcard_denial(z, region, msg, ce->name,
3044                         ce->namelen)) return 0;
3045         } else if(ce) {
3046                 if(!az_add_nsec3_proof(z, region, msg, ce->name,
3047                         ce->namelen, msg->qinfo.qname,
3048                         msg->qinfo.qname_len, 1, 1, 1))
3049                         return 0;
3050         }
3051         return 1;
3052 }
3053
3054 /** Create answers when an exact match exists for the domain name */
3055 static int
3056 az_generate_answer_with_node(struct auth_zone* z, struct query_info* qinfo,
3057         struct regional* region, struct dns_msg* msg, struct auth_data* node)
3058 {
3059         struct auth_rrset* rrset;
3060         /* positive answer, rrset we are looking for exists */
3061         if((rrset=az_domain_rrset(node, qinfo->qtype)) != NULL) {
3062                 return az_generate_positive_answer(z, region, msg, node, rrset);
3063         }
3064         /* CNAME? */
3065         if((rrset=az_domain_rrset(node, LDNS_RR_TYPE_CNAME)) != NULL) {
3066                 return az_generate_cname_answer(z, qinfo, region, msg,
3067                         node, rrset);
3068         }
3069         /* type ANY ? */
3070         if(qinfo->qtype == LDNS_RR_TYPE_ANY) {
3071                 return az_generate_any_answer(z, region, msg, node);
3072         }
3073         /* NOERROR/NODATA (no such type at domain name) */
3074         return az_generate_notype_answer(z, region, msg, node);
3075 }
3076
3077 /** Generate answer without an existing-node that we can use.
3078  * So it'll be a referral, DNAME or nxdomain */
3079 static int
3080 az_generate_answer_nonexistnode(struct auth_zone* z, struct query_info* qinfo,
3081         struct regional* region, struct dns_msg* msg, struct auth_data* ce,
3082         struct auth_rrset* rrset, struct auth_data* node)
3083 {
3084         struct auth_data* wildcard;
3085
3086         /* we do not have an exact matching name (that exists) */
3087         /* see if we have a NS or DNAME in the ce */
3088         if(ce && rrset && rrset->type == LDNS_RR_TYPE_NS) {
3089                 return az_generate_referral_answer(z, region, msg, ce, rrset);
3090         }
3091         if(ce && rrset && rrset->type == LDNS_RR_TYPE_DNAME) {
3092                 return az_generate_dname_answer(z, qinfo, region, msg, ce,
3093                         rrset);
3094         }
3095         /* if there is an empty nonterminal, wildcard and nxdomain don't
3096          * happen, it is a notype answer */
3097         if(az_empty_nonterminal(z, qinfo, node)) {
3098                 return az_generate_notype_answer(z, region, msg, node);
3099         }
3100         /* see if we have a wildcard under the ce */
3101         if((wildcard=az_find_wildcard(z, qinfo, ce)) != NULL) {
3102                 return az_generate_wildcard_answer(z, qinfo, region, msg,
3103                         ce, wildcard, node);
3104         }
3105         /* generate nxdomain answer */
3106         return az_generate_nxdomain_answer(z, region, msg, ce, node);
3107 }
3108
3109 /** Lookup answer in a zone. */
3110 static int
3111 auth_zone_generate_answer(struct auth_zone* z, struct query_info* qinfo,
3112         struct regional* region, struct dns_msg** msg, int* fallback)
3113 {
3114         struct auth_data* node, *ce;
3115         struct auth_rrset* rrset;
3116         int node_exact, node_exists;
3117         /* does the zone want fallback in case of failure? */
3118         *fallback = z->fallback_enabled;
3119         if(!(*msg=msg_create(region, qinfo))) return 0;
3120
3121         /* lookup if there is a matching domain name for the query */
3122         az_find_domain(z, qinfo, &node_exact, &node);
3123
3124         /* see if node exists for generating answers from (i.e. not glue and
3125          * obscured by NS or DNAME or NSEC3-only), and also return the
3126          * closest-encloser from that, closest node that should be used
3127          * to generate answers from that is above the query */
3128         node_exists = az_find_ce(z, qinfo, node, node_exact, &ce, &rrset);
3129
3130         if(verbosity >= VERB_ALGO) {
3131                 char zname[256], qname[256], nname[256], cename[256],
3132                         tpstr[32], rrstr[32];
3133                 sldns_wire2str_dname_buf(qinfo->qname, qinfo->qname_len, qname,
3134                         sizeof(qname));
3135                 sldns_wire2str_type_buf(qinfo->qtype, tpstr, sizeof(tpstr));
3136                 sldns_wire2str_dname_buf(z->name, z->namelen, zname,
3137                         sizeof(zname));
3138                 if(node)
3139                         sldns_wire2str_dname_buf(node->name, node->namelen,
3140                                 nname, sizeof(nname));
3141                 else    snprintf(nname, sizeof(nname), "NULL");
3142                 if(ce)
3143                         sldns_wire2str_dname_buf(ce->name, ce->namelen,
3144                                 cename, sizeof(cename));
3145                 else    snprintf(cename, sizeof(cename), "NULL");
3146                 if(rrset) sldns_wire2str_type_buf(rrset->type, rrstr,
3147                         sizeof(rrstr));
3148                 else    snprintf(rrstr, sizeof(rrstr), "NULL");
3149                 log_info("auth_zone %s query %s %s, domain %s %s %s, "
3150                         "ce %s, rrset %s", zname, qname, tpstr, nname,
3151                         (node_exact?"exact":"notexact"),
3152                         (node_exists?"exist":"notexist"), cename, rrstr);
3153         }
3154
3155         if(node_exists) {
3156                 /* the node is fine, generate answer from node */
3157                 return az_generate_answer_with_node(z, qinfo, region, *msg,
3158                         node);
3159         }
3160         return az_generate_answer_nonexistnode(z, qinfo, region, *msg,
3161                 ce, rrset, node);
3162 }
3163
3164 int auth_zones_lookup(struct auth_zones* az, struct query_info* qinfo,
3165         struct regional* region, struct dns_msg** msg, int* fallback,
3166         uint8_t* dp_nm, size_t dp_nmlen)
3167 {
3168         int r;
3169         struct auth_zone* z;
3170         /* find the zone that should contain the answer. */
3171         lock_rw_rdlock(&az->lock);
3172         z = auth_zone_find(az, dp_nm, dp_nmlen, qinfo->qclass);
3173         if(!z) {
3174                 lock_rw_unlock(&az->lock);
3175                 /* no auth zone, fallback to internet */
3176                 *fallback = 1;
3177                 return 0;
3178         }
3179         lock_rw_rdlock(&z->lock);
3180         lock_rw_unlock(&az->lock);
3181
3182         /* if not for upstream queries, fallback */
3183         if(!z->for_upstream) {
3184                 lock_rw_unlock(&z->lock);
3185                 *fallback = 1;
3186                 return 0;
3187         }
3188         if(z->zone_expired) {
3189                 *fallback = z->fallback_enabled;
3190                 lock_rw_unlock(&z->lock);
3191                 return 0;
3192         }
3193         /* see what answer that zone would generate */
3194         r = auth_zone_generate_answer(z, qinfo, region, msg, fallback);
3195         lock_rw_unlock(&z->lock);
3196         return r;
3197 }
3198
3199 /** encode auth answer */
3200 static void
3201 auth_answer_encode(struct query_info* qinfo, struct module_env* env,
3202         struct edns_data* edns, struct comm_reply* repinfo, sldns_buffer* buf,
3203         struct regional* temp, struct dns_msg* msg)
3204 {
3205         uint16_t udpsize;
3206         udpsize = edns->udp_size;
3207         edns->edns_version = EDNS_ADVERTISED_VERSION;
3208         edns->udp_size = EDNS_ADVERTISED_SIZE;
3209         edns->ext_rcode = 0;
3210         edns->bits &= EDNS_DO;
3211
3212         if(!inplace_cb_reply_local_call(env, qinfo, NULL, msg->rep,
3213                 (int)FLAGS_GET_RCODE(msg->rep->flags), edns, repinfo, temp)
3214                 || !reply_info_answer_encode(qinfo, msg->rep,
3215                 *(uint16_t*)sldns_buffer_begin(buf),
3216                 sldns_buffer_read_u16_at(buf, 2),
3217                 buf, 0, 0, temp, udpsize, edns,
3218                 (int)(edns->bits&EDNS_DO), 0)) {
3219                 error_encode(buf, (LDNS_RCODE_SERVFAIL|BIT_AA), qinfo,
3220                         *(uint16_t*)sldns_buffer_begin(buf),
3221                         sldns_buffer_read_u16_at(buf, 2), edns);
3222         }
3223 }
3224
3225 /** encode auth error answer */
3226 static void
3227 auth_error_encode(struct query_info* qinfo, struct module_env* env,
3228         struct edns_data* edns, struct comm_reply* repinfo, sldns_buffer* buf,
3229         struct regional* temp, int rcode)
3230 {
3231         edns->edns_version = EDNS_ADVERTISED_VERSION;
3232         edns->udp_size = EDNS_ADVERTISED_SIZE;
3233         edns->ext_rcode = 0;
3234         edns->bits &= EDNS_DO;
3235
3236         if(!inplace_cb_reply_local_call(env, qinfo, NULL, NULL,
3237                 rcode, edns, repinfo, temp))
3238                 edns->opt_list = NULL;
3239         error_encode(buf, rcode|BIT_AA, qinfo,
3240                 *(uint16_t*)sldns_buffer_begin(buf),
3241                 sldns_buffer_read_u16_at(buf, 2), edns);
3242 }
3243
3244 int auth_zones_answer(struct auth_zones* az, struct module_env* env,
3245         struct query_info* qinfo, struct edns_data* edns,
3246         struct comm_reply* repinfo, struct sldns_buffer* buf, struct regional* temp)
3247 {
3248         struct dns_msg* msg = NULL;
3249         struct auth_zone* z;
3250         int r;
3251         int fallback = 0;
3252
3253         lock_rw_rdlock(&az->lock);
3254         if(!az->have_downstream) {
3255                 /* no downstream auth zones */
3256                 lock_rw_unlock(&az->lock);
3257                 return 0;
3258         }
3259         if(qinfo->qtype == LDNS_RR_TYPE_DS) {
3260                 uint8_t* delname = qinfo->qname;
3261                 size_t delnamelen = qinfo->qname_len;
3262                 dname_remove_label(&delname, &delnamelen);
3263                 z = auth_zones_find_zone(az, delname, delnamelen,
3264                         qinfo->qclass);
3265         } else {
3266                 z = auth_zones_find_zone(az, qinfo->qname, qinfo->qname_len,
3267                         qinfo->qclass);
3268         }
3269         if(!z) {
3270                 /* no zone above it */
3271                 lock_rw_unlock(&az->lock);
3272                 return 0;
3273         }
3274         lock_rw_rdlock(&z->lock);
3275         lock_rw_unlock(&az->lock);
3276         if(!z->for_downstream) {
3277                 lock_rw_unlock(&z->lock);
3278                 return 0;
3279         }
3280         if(z->zone_expired) {
3281                 if(z->fallback_enabled) {
3282                         lock_rw_unlock(&z->lock);
3283                         return 0;
3284                 }
3285                 lock_rw_unlock(&z->lock);
3286                 lock_rw_wrlock(&az->lock);
3287                 az->num_query_down++;
3288                 lock_rw_unlock(&az->lock);
3289                 auth_error_encode(qinfo, env, edns, repinfo, buf, temp,
3290                         LDNS_RCODE_SERVFAIL);
3291                 return 1;
3292         }
3293
3294         /* answer it from zone z */
3295         r = auth_zone_generate_answer(z, qinfo, temp, &msg, &fallback);
3296         lock_rw_unlock(&z->lock);
3297         if(!r && fallback) {
3298                 /* fallback to regular answering (recursive) */
3299                 return 0;
3300         }
3301         lock_rw_wrlock(&az->lock);
3302         az->num_query_down++;
3303         lock_rw_unlock(&az->lock);
3304
3305         /* encode answer */
3306         if(!r)
3307                 auth_error_encode(qinfo, env, edns, repinfo, buf, temp,
3308                         LDNS_RCODE_SERVFAIL);
3309         else    auth_answer_encode(qinfo, env, edns, repinfo, buf, temp, msg);
3310
3311         return 1;
3312 }
3313
3314 int auth_zones_can_fallback(struct auth_zones* az, uint8_t* nm, size_t nmlen,
3315         uint16_t dclass)
3316 {
3317         int r;
3318         struct auth_zone* z;
3319         lock_rw_rdlock(&az->lock);
3320         z = auth_zone_find(az, nm, nmlen, dclass);
3321         if(!z) {
3322                 lock_rw_unlock(&az->lock);
3323                 /* no such auth zone, fallback */
3324                 return 1;
3325         }
3326         lock_rw_rdlock(&z->lock);
3327         lock_rw_unlock(&az->lock);
3328         r = z->fallback_enabled || (!z->for_upstream);
3329         lock_rw_unlock(&z->lock);
3330         return r;
3331 }
3332
3333 int
3334 auth_zone_parse_notify_serial(sldns_buffer* pkt, uint32_t *serial)
3335 {
3336         struct query_info q;
3337         uint16_t rdlen;
3338         memset(&q, 0, sizeof(q));
3339         sldns_buffer_set_position(pkt, 0);
3340         if(!query_info_parse(&q, pkt)) return 0;
3341         if(LDNS_ANCOUNT(sldns_buffer_begin(pkt)) == 0) return 0;
3342         /* skip name of RR in answer section */
3343         if(sldns_buffer_remaining(pkt) < 1) return 0;
3344         if(pkt_dname_len(pkt) == 0) return 0;
3345         /* check type */
3346         if(sldns_buffer_remaining(pkt) < 10 /* type,class,ttl,rdatalen*/)
3347                 return 0;
3348         if(sldns_buffer_read_u16(pkt) != LDNS_RR_TYPE_SOA) return 0;
3349         sldns_buffer_skip(pkt, 2); /* class */
3350         sldns_buffer_skip(pkt, 4); /* ttl */
3351         rdlen = sldns_buffer_read_u16(pkt); /* rdatalen */
3352         if(sldns_buffer_remaining(pkt) < rdlen) return 0;
3353         if(rdlen < 22) return 0; /* bad soa length */
3354         sldns_buffer_skip(pkt, (ssize_t)(rdlen-20));
3355         *serial = sldns_buffer_read_u32(pkt);
3356         /* return true when has serial in answer section */
3357         return 1;
3358 }
3359
3360 /** see if addr appears in the list */
3361 static int
3362 addr_in_list(struct auth_addr* list, struct sockaddr_storage* addr,
3363         socklen_t addrlen)
3364 {
3365         struct auth_addr* p;
3366         for(p=list; p; p=p->next) {
3367                 if(sockaddr_cmp_addr(addr, addrlen, &p->addr, p->addrlen)==0)
3368                         return 1;
3369         }
3370         return 0;
3371 }
3372
3373 /** check if an address matches a master specification (or one of its
3374  * addresses in the addr list) */
3375 static int
3376 addr_matches_master(struct auth_master* master, struct sockaddr_storage* addr,
3377         socklen_t addrlen, struct auth_master** fromhost)
3378 {
3379         struct sockaddr_storage a;
3380         socklen_t alen = 0;
3381         int net = 0;
3382         if(addr_in_list(master->list, addr, addrlen)) {
3383                 *fromhost = master;
3384                 return 1;
3385         }
3386         /* compare address (but not port number, that is the destination
3387          * port of the master, the port number of the received notify is
3388          * allowed to by any port on that master) */
3389         if(extstrtoaddr(master->host, &a, &alen) &&
3390                 sockaddr_cmp_addr(addr, addrlen, &a, alen)==0) {
3391                 *fromhost = master;
3392                 return 1;
3393         }
3394         /* prefixes, addr/len, like 10.0.0.0/8 */
3395         /* not http and has a / and there is one / */
3396         if(master->allow_notify && !master->http &&
3397                 strchr(master->host, '/') != NULL &&
3398                 strchr(master->host, '/') == strrchr(master->host, '/') &&
3399                 netblockstrtoaddr(master->host, UNBOUND_DNS_PORT, &a, &alen,
3400                 &net) && alen == addrlen) {
3401                 if(addr_in_common(addr, (addr_is_ip6(addr, addrlen)?128:32),
3402                         &a, net, alen) >= net) {
3403                         *fromhost = NULL; /* prefix does not have destination
3404                                 to send the probe or transfer with */
3405                         return 1; /* matches the netblock */
3406                 }
3407         }
3408         return 0;
3409 }
3410
3411 /** check access list for notifies */
3412 static int
3413 az_xfr_allowed_notify(struct auth_xfer* xfr, struct sockaddr_storage* addr,
3414         socklen_t addrlen, struct auth_master** fromhost)
3415 {
3416         struct auth_master* p;
3417         for(p=xfr->allow_notify_list; p; p=p->next) {
3418                 if(addr_matches_master(p, addr, addrlen, fromhost)) {
3419                         return 1;
3420                 }
3421         }
3422         return 0;
3423 }
3424
3425 /** see if the serial means the zone has to be updated, i.e. the serial
3426  * is newer than the zone serial, or we have no zone */
3427 static int
3428 xfr_serial_means_update(struct auth_xfer* xfr, uint32_t serial)
3429 {
3430         if(!xfr->have_zone)
3431                 return 1; /* no zone, anything is better */
3432         if(xfr->zone_expired)
3433                 return 1; /* expired, the sent serial is better than expired
3434                         data */
3435         if(compare_serial(xfr->serial, serial) < 0)
3436                 return 1; /* our serial is smaller than the sent serial,
3437                         the data is newer, fetch it */
3438         return 0;
3439 }
3440
3441 /** note notify serial, updates the notify information in the xfr struct */
3442 static void
3443 xfr_note_notify_serial(struct auth_xfer* xfr, int has_serial, uint32_t serial)
3444 {
3445         if(xfr->notify_received && xfr->notify_has_serial && has_serial) {
3446                 /* see if this serial is newer */
3447                 if(compare_serial(xfr->notify_serial, serial) < 0)
3448                         xfr->notify_serial = serial;
3449         } else if(xfr->notify_received && xfr->notify_has_serial &&
3450                 !has_serial) {
3451                 /* remove serial, we have notify without serial */
3452                 xfr->notify_has_serial = 0;
3453                 xfr->notify_serial = 0;
3454         } else if(xfr->notify_received && !xfr->notify_has_serial) {
3455                 /* we already have notify without serial, keep it
3456                  * that way; no serial check when current operation
3457                  * is done */
3458         } else {
3459                 xfr->notify_received = 1;
3460                 xfr->notify_has_serial = has_serial;
3461                 xfr->notify_serial = serial;
3462         }
3463 }
3464
3465 /** process a notify serial, start new probe or note serial. xfr is locked */
3466 static void
3467 xfr_process_notify(struct auth_xfer* xfr, struct module_env* env,
3468         int has_serial, uint32_t serial, struct auth_master* fromhost)
3469 {
3470         /* if the serial of notify is older than we have, don't fetch
3471          * a zone, we already have it */
3472         if(has_serial && !xfr_serial_means_update(xfr, serial)) {
3473                 lock_basic_unlock(&xfr->lock);
3474                 return;
3475         }
3476         /* start new probe with this addr src, or note serial */
3477         if(!xfr_start_probe(xfr, env, fromhost)) {
3478                 /* not started because already in progress, note the serial */
3479                 xfr_note_notify_serial(xfr, has_serial, serial);
3480                 lock_basic_unlock(&xfr->lock);
3481         }
3482         /* successful end of start_probe unlocked xfr->lock */
3483 }
3484
3485 int auth_zones_notify(struct auth_zones* az, struct module_env* env,
3486         uint8_t* nm, size_t nmlen, uint16_t dclass,
3487         struct sockaddr_storage* addr, socklen_t addrlen, int has_serial,
3488         uint32_t serial, int* refused)
3489 {
3490         struct auth_xfer* xfr;
3491         struct auth_master* fromhost = NULL;
3492         /* see which zone this is */
3493         lock_rw_rdlock(&az->lock);
3494         xfr = auth_xfer_find(az, nm, nmlen, dclass);
3495         if(!xfr) {
3496                 lock_rw_unlock(&az->lock);
3497                 /* no such zone, refuse the notify */
3498                 *refused = 1;
3499                 return 0;
3500         }
3501         lock_basic_lock(&xfr->lock);
3502         lock_rw_unlock(&az->lock);
3503
3504         /* check access list for notifies */
3505         if(!az_xfr_allowed_notify(xfr, addr, addrlen, &fromhost)) {
3506                 lock_basic_unlock(&xfr->lock);
3507                 /* notify not allowed, refuse the notify */
3508                 *refused = 1;
3509                 return 0;
3510         }
3511
3512         /* process the notify */
3513         xfr_process_notify(xfr, env, has_serial, serial, fromhost);
3514         return 1;
3515 }
3516
3517 int auth_zones_startprobesequence(struct auth_zones* az,
3518         struct module_env* env, uint8_t* nm, size_t nmlen, uint16_t dclass)
3519 {
3520         struct auth_xfer* xfr;
3521         lock_rw_rdlock(&az->lock);
3522         xfr = auth_xfer_find(az, nm, nmlen, dclass);
3523         if(!xfr) {
3524                 lock_rw_unlock(&az->lock);
3525                 return 0;
3526         }
3527         lock_basic_lock(&xfr->lock);
3528         lock_rw_unlock(&az->lock);
3529
3530         xfr_process_notify(xfr, env, 0, 0, NULL);
3531         return 1;
3532 }
3533
3534 /** set a zone expired */
3535 static void
3536 auth_xfer_set_expired(struct auth_xfer* xfr, struct module_env* env,
3537         int expired)
3538 {
3539         struct auth_zone* z;
3540
3541         /* expire xfr */
3542         lock_basic_lock(&xfr->lock);
3543         xfr->zone_expired = expired;
3544         lock_basic_unlock(&xfr->lock);
3545
3546         /* find auth_zone */
3547         lock_rw_rdlock(&env->auth_zones->lock);
3548         z = auth_zone_find(env->auth_zones, xfr->name, xfr->namelen,
3549                 xfr->dclass);
3550         if(!z) {
3551                 lock_rw_unlock(&env->auth_zones->lock);
3552                 return;
3553         }
3554         lock_rw_wrlock(&z->lock);
3555         lock_rw_unlock(&env->auth_zones->lock);
3556
3557         /* expire auth_zone */
3558         z->zone_expired = expired;
3559         lock_rw_unlock(&z->lock);
3560 }
3561
3562 /** find master (from notify or probe) in list of masters */
3563 static struct auth_master*
3564 find_master_by_host(struct auth_master* list, char* host)
3565 {
3566         struct auth_master* p;
3567         for(p=list; p; p=p->next) {
3568                 if(strcmp(p->host, host) == 0)
3569                         return p;
3570         }
3571         return NULL;
3572 }
3573
3574 /** delete the looked up auth_addrs for all the masters in the list */
3575 static void
3576 xfr_masterlist_free_addrs(struct auth_master* list)
3577 {
3578         struct auth_master* m;
3579         for(m=list; m; m=m->next) {
3580                 if(m->list) {
3581                         auth_free_master_addrs(m->list);
3582                         m->list = NULL;
3583                 }
3584         }
3585 }
3586
3587 /** copy a list of auth_addrs */
3588 static struct auth_addr*
3589 auth_addr_list_copy(struct auth_addr* source)
3590 {
3591         struct auth_addr* list = NULL, *last = NULL;
3592         struct auth_addr* p;
3593         for(p=source; p; p=p->next) {
3594                 struct auth_addr* a = (struct auth_addr*)memdup(p, sizeof(*p));
3595                 if(!a) {
3596                         log_err("malloc failure");
3597                         auth_free_master_addrs(list);
3598                         return NULL;
3599                 }
3600                 a->next = NULL;
3601                 if(last) last->next = a;
3602                 if(!list) list = a;
3603                 last = a;
3604         }
3605         return list;
3606 }
3607
3608 /** copy a master to a new structure, NULL on alloc failure */
3609 static struct auth_master*
3610 auth_master_copy(struct auth_master* o)
3611 {
3612         struct auth_master* m;
3613         if(!o) return NULL;
3614         m = (struct auth_master*)memdup(o, sizeof(*o));
3615         if(!m) {
3616                 log_err("malloc failure");
3617                 return NULL;
3618         }
3619         m->next = NULL;
3620         if(m->host) {
3621                 m->host = strdup(m->host);
3622                 if(!m->host) {
3623                         free(m);
3624                         log_err("malloc failure");
3625                         return NULL;
3626                 }
3627         }
3628         if(m->file) {
3629                 m->file = strdup(m->file);
3630                 if(!m->file) {
3631                         free(m->host);
3632                         free(m);
3633                         log_err("malloc failure");
3634                         return NULL;
3635                 }
3636         }
3637         if(m->list) {
3638                 m->list = auth_addr_list_copy(m->list);
3639                 if(!m->list) {
3640                         free(m->file);
3641                         free(m->host);
3642                         free(m);
3643                         return NULL;
3644                 }
3645         }
3646         return m;
3647 }
3648
3649 /** copy the master addresses from the task_probe lookups to the allow_notify
3650  * list of masters */
3651 static void
3652 probe_copy_masters_for_allow_notify(struct auth_xfer* xfr)
3653 {
3654         struct auth_master* list = NULL, *last = NULL;
3655         struct auth_master* p;
3656         /* build up new list with copies */
3657         for(p = xfr->task_probe->masters; p; p=p->next) {
3658                 struct auth_master* m = auth_master_copy(p);
3659                 if(!m) {
3660                         auth_free_masters(list);
3661                         /* failed because of malloc failure, use old list */
3662                         return;
3663                 }
3664                 m->next = NULL;
3665                 if(last) last->next = m;
3666                 if(!list) list = m;
3667                 last = m;
3668         }
3669         /* success, replace list */
3670         auth_free_masters(xfr->allow_notify_list);
3671         xfr->allow_notify_list = list;
3672 }
3673
3674 /** start the lookups for task_transfer */
3675 static void
3676 xfr_transfer_start_lookups(struct auth_xfer* xfr)
3677 {
3678         /* delete all the looked up addresses in the list */
3679         xfr_masterlist_free_addrs(xfr->task_transfer->masters);
3680
3681         /* start lookup at the first master */
3682         xfr->task_transfer->lookup_target = xfr->task_transfer->masters;
3683         xfr->task_transfer->lookup_aaaa = 0;
3684 }
3685
3686 /** move to the next lookup of hostname for task_transfer */
3687 static void
3688 xfr_transfer_move_to_next_lookup(struct auth_xfer* xfr, struct module_env* env)
3689 {
3690         if(!xfr->task_transfer->lookup_target)
3691                 return; /* already at end of list */
3692         if(!xfr->task_transfer->lookup_aaaa && env->cfg->do_ip6) {
3693                 /* move to lookup AAAA */
3694                 xfr->task_transfer->lookup_aaaa = 1;
3695                 return;
3696         }
3697         xfr->task_transfer->lookup_target =
3698                 xfr->task_transfer->lookup_target->next;
3699         xfr->task_transfer->lookup_aaaa = 0;
3700         if(!env->cfg->do_ip4 && xfr->task_transfer->lookup_target!=NULL)
3701                 xfr->task_transfer->lookup_aaaa = 1;
3702 }
3703
3704 /** start the lookups for task_probe */
3705 static void
3706 xfr_probe_start_lookups(struct auth_xfer* xfr)
3707 {
3708         /* delete all the looked up addresses in the list */
3709         xfr_masterlist_free_addrs(xfr->task_probe->masters);
3710
3711         /* start lookup at the first master */
3712         xfr->task_probe->lookup_target = xfr->task_probe->masters;
3713         xfr->task_probe->lookup_aaaa = 0;
3714 }
3715
3716 /** move to the next lookup of hostname for task_probe */
3717 static void
3718 xfr_probe_move_to_next_lookup(struct auth_xfer* xfr, struct module_env* env)
3719 {
3720         if(!xfr->task_probe->lookup_target)
3721                 return; /* already at end of list */
3722         if(!xfr->task_probe->lookup_aaaa && env->cfg->do_ip6) {
3723                 /* move to lookup AAAA */
3724                 xfr->task_probe->lookup_aaaa = 1;
3725                 return;
3726         }
3727         xfr->task_probe->lookup_target = xfr->task_probe->lookup_target->next;
3728         xfr->task_probe->lookup_aaaa = 0;
3729         if(!env->cfg->do_ip4 && xfr->task_probe->lookup_target!=NULL)
3730                 xfr->task_probe->lookup_aaaa = 1;
3731 }
3732
3733 /** start the iteration of the task_transfer list of masters */
3734 static void
3735 xfr_transfer_start_list(struct auth_xfer* xfr, struct auth_master* spec)
3736 {
3737         if(spec) {
3738                 xfr->task_transfer->scan_specific = find_master_by_host(
3739                         xfr->task_transfer->masters, spec->host);
3740                 if(xfr->task_transfer->scan_specific) {
3741                         xfr->task_transfer->scan_target = NULL;
3742                         xfr->task_transfer->scan_addr = NULL;
3743                         if(xfr->task_transfer->scan_specific->list)
3744                                 xfr->task_transfer->scan_addr =
3745                                         xfr->task_transfer->scan_specific->list;
3746                         return;
3747                 }
3748         }
3749         /* no specific (notified) host to scan */
3750         xfr->task_transfer->scan_specific = NULL;
3751         xfr->task_transfer->scan_addr = NULL;
3752         /* pick up first scan target */
3753         xfr->task_transfer->scan_target = xfr->task_transfer->masters;
3754         if(xfr->task_transfer->scan_target && xfr->task_transfer->
3755                 scan_target->list)
3756                 xfr->task_transfer->scan_addr =
3757                         xfr->task_transfer->scan_target->list;
3758 }
3759
3760 /** start the iteration of the task_probe list of masters */
3761 static void
3762 xfr_probe_start_list(struct auth_xfer* xfr, struct auth_master* spec)
3763 {
3764         if(spec) {
3765                 xfr->task_probe->scan_specific = find_master_by_host(
3766                         xfr->task_probe->masters, spec->host);
3767                 if(xfr->task_probe->scan_specific) {
3768                         xfr->task_probe->scan_target = NULL;
3769                         xfr->task_probe->scan_addr = NULL;
3770                         if(xfr->task_probe->scan_specific->list)
3771                                 xfr->task_probe->scan_addr =
3772                                         xfr->task_probe->scan_specific->list;
3773                         return;
3774                 }
3775         }
3776         /* no specific (notified) host to scan */
3777         xfr->task_probe->scan_specific = NULL;
3778         xfr->task_probe->scan_addr = NULL;
3779         /* pick up first scan target */
3780         xfr->task_probe->scan_target = xfr->task_probe->masters;
3781         if(xfr->task_probe->scan_target && xfr->task_probe->scan_target->list)
3782                 xfr->task_probe->scan_addr =
3783                         xfr->task_probe->scan_target->list;
3784 }
3785
3786 /** pick up the master that is being scanned right now, task_transfer */
3787 static struct auth_master*
3788 xfr_transfer_current_master(struct auth_xfer* xfr)
3789 {
3790         if(xfr->task_transfer->scan_specific)
3791                 return xfr->task_transfer->scan_specific;
3792         return xfr->task_transfer->scan_target;
3793 }
3794
3795 /** pick up the master that is being scanned right now, task_probe */
3796 static struct auth_master*
3797 xfr_probe_current_master(struct auth_xfer* xfr)
3798 {
3799         if(xfr->task_probe->scan_specific)
3800                 return xfr->task_probe->scan_specific;
3801         return xfr->task_probe->scan_target;
3802 }
3803
3804 /** true if at end of list, task_transfer */
3805 static int
3806 xfr_transfer_end_of_list(struct auth_xfer* xfr)
3807 {
3808         return !xfr->task_transfer->scan_specific &&
3809                 !xfr->task_transfer->scan_target;
3810 }
3811
3812 /** true if at end of list, task_probe */
3813 static int
3814 xfr_probe_end_of_list(struct auth_xfer* xfr)
3815 {
3816         return !xfr->task_probe->scan_specific && !xfr->task_probe->scan_target;
3817 }
3818
3819 /** move to next master in list, task_transfer */
3820 static void
3821 xfr_transfer_nextmaster(struct auth_xfer* xfr)
3822 {
3823         if(!xfr->task_transfer->scan_specific &&
3824                 !xfr->task_transfer->scan_target)
3825                 return;
3826         if(xfr->task_transfer->scan_addr) {
3827                 xfr->task_transfer->scan_addr =
3828                         xfr->task_transfer->scan_addr->next;
3829                 if(xfr->task_transfer->scan_addr)
3830                         return;
3831         }
3832         if(xfr->task_transfer->scan_specific) {
3833                 xfr->task_transfer->scan_specific = NULL;
3834                 xfr->task_transfer->scan_target = xfr->task_transfer->masters;
3835                 if(xfr->task_transfer->scan_target && xfr->task_transfer->
3836                         scan_target->list)
3837                         xfr->task_transfer->scan_addr =
3838                                 xfr->task_transfer->scan_target->list;
3839                 return;
3840         }
3841         if(!xfr->task_transfer->scan_target)
3842                 return;
3843         xfr->task_transfer->scan_target = xfr->task_transfer->scan_target->next;
3844         if(xfr->task_transfer->scan_target && xfr->task_transfer->
3845                 scan_target->list)
3846                 xfr->task_transfer->scan_addr =
3847                         xfr->task_transfer->scan_target->list;
3848         return;
3849 }
3850
3851 /** move to next master in list, task_probe */
3852 static void
3853 xfr_probe_nextmaster(struct auth_xfer* xfr)
3854 {
3855         if(!xfr->task_probe->scan_specific && !xfr->task_probe->scan_target)
3856                 return;
3857         if(xfr->task_probe->scan_addr) {
3858                 xfr->task_probe->scan_addr = xfr->task_probe->scan_addr->next;
3859                 if(xfr->task_probe->scan_addr)
3860                         return;
3861         }
3862         if(xfr->task_probe->scan_specific) {
3863                 xfr->task_probe->scan_specific = NULL;
3864                 xfr->task_probe->scan_target = xfr->task_probe->masters;
3865                 if(xfr->task_probe->scan_target && xfr->task_probe->
3866                         scan_target->list)
3867                         xfr->task_probe->scan_addr =
3868                                 xfr->task_probe->scan_target->list;
3869                 return;
3870         }
3871         if(!xfr->task_probe->scan_target)
3872                 return;
3873         xfr->task_probe->scan_target = xfr->task_probe->scan_target->next;
3874         if(xfr->task_probe->scan_target && xfr->task_probe->
3875                 scan_target->list)
3876                 xfr->task_probe->scan_addr =
3877                         xfr->task_probe->scan_target->list;
3878         return;
3879 }
3880
3881 /** create SOA probe packet for xfr */
3882 static void
3883 xfr_create_soa_probe_packet(struct auth_xfer* xfr, sldns_buffer* buf,
3884         uint16_t id)
3885 {
3886         struct query_info qinfo;
3887
3888         memset(&qinfo, 0, sizeof(qinfo));
3889         qinfo.qname = xfr->name;
3890         qinfo.qname_len = xfr->namelen;
3891         qinfo.qtype = LDNS_RR_TYPE_SOA;
3892         qinfo.qclass = xfr->dclass;
3893         qinfo_query_encode(buf, &qinfo);
3894         sldns_buffer_write_u16_at(buf, 0, id);
3895 }
3896
3897 /** create IXFR/AXFR packet for xfr */
3898 static void
3899 xfr_create_ixfr_packet(struct auth_xfer* xfr, sldns_buffer* buf, uint16_t id,
3900         struct auth_master* master)
3901 {
3902         struct query_info qinfo;
3903         uint32_t serial;
3904         int have_zone;
3905         have_zone = xfr->have_zone;
3906         serial = xfr->serial;
3907
3908         memset(&qinfo, 0, sizeof(qinfo));
3909         qinfo.qname = xfr->name;
3910         qinfo.qname_len = xfr->namelen;
3911         xfr->task_transfer->got_xfr_serial = 0;
3912         xfr->task_transfer->rr_scan_num = 0;
3913         xfr->task_transfer->incoming_xfr_serial = 0;
3914         xfr->task_transfer->on_ixfr_is_axfr = 0;
3915         xfr->task_transfer->on_ixfr = 1;
3916         qinfo.qtype = LDNS_RR_TYPE_IXFR;
3917         if(!have_zone || xfr->task_transfer->ixfr_fail || !master->ixfr) {
3918                 qinfo.qtype = LDNS_RR_TYPE_AXFR;
3919                 xfr->task_transfer->ixfr_fail = 0;
3920                 xfr->task_transfer->on_ixfr = 0;
3921         }
3922
3923         qinfo.qclass = xfr->dclass;
3924         qinfo_query_encode(buf, &qinfo);
3925         sldns_buffer_write_u16_at(buf, 0, id);
3926
3927         /* append serial for IXFR */
3928         if(qinfo.qtype == LDNS_RR_TYPE_IXFR) {
3929                 size_t end = sldns_buffer_limit(buf);
3930                 sldns_buffer_clear(buf);
3931                 sldns_buffer_set_position(buf, end);
3932                 /* auth section count 1 */
3933                 sldns_buffer_write_u16_at(buf, LDNS_NSCOUNT_OFF, 1);
3934                 /* write SOA */
3935                 sldns_buffer_write_u8(buf, 0xC0); /* compressed ptr to qname */
3936                 sldns_buffer_write_u8(buf, 0x0C);
3937                 sldns_buffer_write_u16(buf, LDNS_RR_TYPE_SOA);
3938                 sldns_buffer_write_u16(buf, qinfo.qclass);
3939                 sldns_buffer_write_u32(buf, 0); /* ttl */
3940                 sldns_buffer_write_u16(buf, 22); /* rdata length */
3941                 sldns_buffer_write_u8(buf, 0); /* . */
3942                 sldns_buffer_write_u8(buf, 0); /* . */
3943                 sldns_buffer_write_u32(buf, serial); /* serial */
3944                 sldns_buffer_write_u32(buf, 0); /* refresh */
3945                 sldns_buffer_write_u32(buf, 0); /* retry */
3946                 sldns_buffer_write_u32(buf, 0); /* expire */
3947                 sldns_buffer_write_u32(buf, 0); /* minimum */
3948                 sldns_buffer_flip(buf);
3949         }
3950 }
3951
3952 /** check if returned packet is OK */
3953 static int
3954 check_packet_ok(sldns_buffer* pkt, uint16_t qtype, struct auth_xfer* xfr,
3955         uint32_t* serial)
3956 {
3957         /* parse to see if packet worked, valid reply */
3958
3959         /* check serial number of SOA */
3960         if(sldns_buffer_limit(pkt) < LDNS_HEADER_SIZE)
3961                 return 0;
3962
3963         /* check ID */
3964         if(LDNS_ID_WIRE(sldns_buffer_begin(pkt)) != xfr->task_probe->id)
3965                 return 0;
3966
3967         /* check flag bits and rcode */
3968         if(!LDNS_QR_WIRE(sldns_buffer_begin(pkt)))
3969                 return 0;
3970         if(LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt)) != LDNS_PACKET_QUERY)
3971                 return 0;
3972         if(LDNS_RCODE_WIRE(sldns_buffer_begin(pkt)) != LDNS_RCODE_NOERROR)
3973                 return 0;
3974
3975         /* check qname */
3976         if(LDNS_QDCOUNT(sldns_buffer_begin(pkt)) != 1)
3977                 return 0;
3978         sldns_buffer_skip(pkt, LDNS_HEADER_SIZE);
3979         if(sldns_buffer_remaining(pkt) < xfr->namelen)
3980                 return 0;
3981         if(query_dname_compare(sldns_buffer_current(pkt), xfr->name) != 0)
3982                 return 0;
3983         sldns_buffer_skip(pkt, (ssize_t)xfr->namelen);
3984
3985         /* check qtype, qclass */
3986         if(sldns_buffer_remaining(pkt) < 4)
3987                 return 0;
3988         if(sldns_buffer_read_u16(pkt) != qtype)
3989                 return 0;
3990         if(sldns_buffer_read_u16(pkt) != xfr->dclass)
3991                 return 0;
3992
3993         if(serial) {
3994                 uint16_t rdlen;
3995                 /* read serial number, from answer section SOA */
3996                 if(LDNS_ANCOUNT(sldns_buffer_begin(pkt)) == 0)
3997                         return 0;
3998                 /* read from first record SOA record */
3999                 if(sldns_buffer_remaining(pkt) < 1)
4000                         return 0;
4001                 if(dname_pkt_compare(pkt, sldns_buffer_current(pkt),
4002                         xfr->name) != 0)
4003                         return 0;
4004                 if(!pkt_dname_len(pkt))
4005                         return 0;
4006                 /* type, class, ttl, rdatalen */
4007                 if(sldns_buffer_remaining(pkt) < 4+4+2)
4008                         return 0;
4009                 if(sldns_buffer_read_u16(pkt) != qtype)
4010                         return 0;
4011                 if(sldns_buffer_read_u16(pkt) != xfr->dclass)
4012                         return 0;
4013                 sldns_buffer_skip(pkt, 4); /* ttl */
4014                 rdlen = sldns_buffer_read_u16(pkt);
4015                 if(sldns_buffer_remaining(pkt) < rdlen)
4016                         return 0;
4017                 if(sldns_buffer_remaining(pkt) < 1)
4018                         return 0;
4019                 if(!pkt_dname_len(pkt)) /* soa name */
4020                         return 0;
4021                 if(sldns_buffer_remaining(pkt) < 1)
4022                         return 0;
4023                 if(!pkt_dname_len(pkt)) /* soa name */
4024                         return 0;
4025                 if(sldns_buffer_remaining(pkt) < 20)
4026                         return 0;
4027                 *serial = sldns_buffer_read_u32(pkt);
4028         }
4029         return 1;
4030 }
4031
4032 /** read one line from chunks into buffer at current position */
4033 static int
4034 chunkline_get_line(struct auth_chunk** chunk, size_t* chunk_pos,
4035         sldns_buffer* buf)
4036 {
4037         int readsome = 0;
4038         while(*chunk) {
4039                 /* more text in this chunk? */
4040                 if(*chunk_pos < (*chunk)->len) {
4041                         readsome = 1;
4042                         while(*chunk_pos < (*chunk)->len) {
4043                                 char c = (char)((*chunk)->data[*chunk_pos]);
4044                                 (*chunk_pos)++;
4045                                 if(sldns_buffer_remaining(buf) < 2) {
4046                                         /* buffer too short */
4047                                         verbose(VERB_ALGO, "http chunkline, "
4048                                                 "line too long");
4049                                         return 0;
4050                                 }
4051                                 sldns_buffer_write_u8(buf, (uint8_t)c);
4052                                 if(c == '\n') {
4053                                         /* we are done */
4054                                         return 1;
4055                                 }
4056                         }
4057                 }
4058                 /* move to next chunk */
4059                 *chunk = (*chunk)->next;
4060                 *chunk_pos = 0;
4061         }
4062         /* no more text */
4063         if(readsome) return 1;
4064         return 0;
4065 }
4066
4067 /** count number of open and closed parenthesis in a chunkline */
4068 static int
4069 chunkline_count_parens(sldns_buffer* buf, size_t start)
4070 {
4071         size_t end = sldns_buffer_position(buf);
4072         size_t i;
4073         int count = 0;
4074         int squote = 0, dquote = 0;
4075         for(i=start; i<end; i++) {
4076                 char c = (char)sldns_buffer_read_u8_at(buf, i);
4077                 if(squote && c != '\'') continue;
4078                 if(dquote && c != '"') continue;
4079                 if(c == '"')
4080                         dquote = !dquote; /* skip quoted part */
4081                 else if(c == '\'')
4082                         squote = !squote; /* skip quoted part */
4083                 else if(c == '(')
4084                         count ++;
4085                 else if(c == ')')
4086                         count --;
4087                 else if(c == ';') {
4088                         /* rest is a comment */
4089                         return count;
4090                 }
4091         }
4092         return count;
4093 }
4094
4095 /** remove trailing ;... comment from a line in the chunkline buffer */
4096 static void
4097 chunkline_remove_trailcomment(sldns_buffer* buf, size_t start)
4098 {
4099         size_t end = sldns_buffer_position(buf);
4100         size_t i;
4101         int squote = 0, dquote = 0;
4102         for(i=start; i<end; i++) {
4103                 char c = (char)sldns_buffer_read_u8_at(buf, i);
4104                 if(squote && c != '\'') continue;
4105                 if(dquote && c != '"') continue;
4106                 if(c == '"')
4107                         dquote = !dquote; /* skip quoted part */
4108                 else if(c == '\'')
4109                         squote = !squote; /* skip quoted part */
4110                 else if(c == ';') {
4111                         /* rest is a comment */
4112                         sldns_buffer_set_position(buf, i);
4113                         return;
4114                 }
4115         }
4116         /* nothing to remove */
4117 }
4118
4119 /** see if a chunkline is a comment line (or empty line) */
4120 static int
4121 chunkline_is_comment_line_or_empty(sldns_buffer* buf)
4122 {
4123         size_t i, end = sldns_buffer_limit(buf);
4124         for(i=0; i<end; i++) {
4125                 char c = (char)sldns_buffer_read_u8_at(buf, i);
4126                 if(c == ';')
4127                         return 1; /* comment */
4128                 else if(c != ' ' && c != '\t' && c != '\r' && c != '\n')
4129                         return 0; /* not a comment */
4130         }
4131         return 1; /* empty */
4132 }
4133
4134 /** find a line with ( ) collated */
4135 static int
4136 chunkline_get_line_collated(struct auth_chunk** chunk, size_t* chunk_pos,
4137         sldns_buffer* buf)
4138 {
4139         size_t pos;
4140         int parens = 0;
4141         sldns_buffer_clear(buf);
4142         pos = sldns_buffer_position(buf);
4143         if(!chunkline_get_line(chunk, chunk_pos, buf)) {
4144                 if(sldns_buffer_position(buf) < sldns_buffer_limit(buf))
4145                         sldns_buffer_write_u8_at(buf, sldns_buffer_position(buf), 0);
4146                 else sldns_buffer_write_u8_at(buf, sldns_buffer_position(buf)-1, 0);
4147                 sldns_buffer_flip(buf);
4148                 return 0;
4149         }
4150         parens += chunkline_count_parens(buf, pos);
4151         while(parens > 0) {
4152                 chunkline_remove_trailcomment(buf, pos);
4153                 pos = sldns_buffer_position(buf);
4154                 if(!chunkline_get_line(chunk, chunk_pos, buf)) {
4155                         if(sldns_buffer_position(buf) < sldns_buffer_limit(buf))
4156                                 sldns_buffer_write_u8_at(buf, sldns_buffer_position(buf), 0);
4157                         else sldns_buffer_write_u8_at(buf, sldns_buffer_position(buf)-1, 0);
4158                         sldns_buffer_flip(buf);
4159                         return 0;
4160                 }
4161                 parens += chunkline_count_parens(buf, pos);
4162         }
4163
4164         if(sldns_buffer_remaining(buf) < 1) {
4165                 verbose(VERB_ALGO, "http chunkline: "
4166                         "line too long");
4167                 return 0;
4168         }
4169         sldns_buffer_write_u8_at(buf, sldns_buffer_position(buf), 0);
4170         sldns_buffer_flip(buf);
4171         return 1;
4172 }
4173
4174 /** process $ORIGIN for http */
4175 static int
4176 http_parse_origin(sldns_buffer* buf, struct sldns_file_parse_state* pstate)
4177 {
4178         char* line = (char*)sldns_buffer_begin(buf);
4179         if(strncmp(line, "$ORIGIN", 7) == 0 &&
4180                 isspace((unsigned char)line[7])) {
4181                 int s;
4182                 pstate->origin_len = sizeof(pstate->origin);
4183                 s = sldns_str2wire_dname_buf(sldns_strip_ws(line+8),
4184                         pstate->origin, &pstate->origin_len);
4185                 if(s) pstate->origin_len = 0;
4186                 return 1;
4187         }
4188         return 0;
4189 }
4190
4191 /** process $TTL for http */
4192 static int
4193 http_parse_ttl(sldns_buffer* buf, struct sldns_file_parse_state* pstate)
4194 {
4195         char* line = (char*)sldns_buffer_begin(buf);
4196         if(strncmp(line, "$TTL", 4) == 0 &&
4197                 isspace((unsigned char)line[4])) {
4198                 const char* end = NULL;
4199                 pstate->default_ttl = sldns_str2period(
4200                         sldns_strip_ws(line+5), &end);
4201                 return 1;
4202         }
4203         return 0;
4204 }
4205
4206 /** find noncomment RR line in chunks, collates lines if ( ) format */
4207 static int
4208 chunkline_non_comment_RR(struct auth_chunk** chunk, size_t* chunk_pos,
4209         sldns_buffer* buf, struct sldns_file_parse_state* pstate)
4210 {
4211         while(chunkline_get_line_collated(chunk, chunk_pos, buf)) {
4212                 if(chunkline_is_comment_line_or_empty(buf)) {
4213                         /* a comment, go to next line */
4214                         continue;
4215                 }
4216                 if(http_parse_origin(buf, pstate)) {
4217                         continue; /* $ORIGIN has been handled */
4218                 }
4219                 if(http_parse_ttl(buf, pstate)) {
4220                         continue; /* $TTL has been handled */
4221                 }
4222                 return 1;
4223         }
4224         /* no noncomments, fail */
4225         return 0;
4226 }
4227
4228 /** check syntax of chunklist zonefile, parse first RR, return false on
4229  * failure and return a string in the scratch buffer (first RR string)
4230  * on failure. */
4231 static int
4232 http_zonefile_syntax_check(struct auth_xfer* xfr, sldns_buffer* buf)
4233 {
4234         uint8_t rr[LDNS_RR_BUF_SIZE];
4235         size_t rr_len, dname_len = 0;
4236         struct sldns_file_parse_state pstate;
4237         struct auth_chunk* chunk;
4238         size_t chunk_pos;
4239         int e;
4240         memset(&pstate, 0, sizeof(pstate));
4241         pstate.default_ttl = 3600;
4242         if(xfr->namelen < sizeof(pstate.origin)) {
4243                 pstate.origin_len = xfr->namelen;
4244                 memmove(pstate.origin, xfr->name, xfr->namelen);
4245         }
4246         chunk = xfr->task_transfer->chunks_first;
4247         chunk_pos = 0;
4248         if(!chunkline_non_comment_RR(&chunk, &chunk_pos, buf, &pstate)) {
4249                 return 0;
4250         }
4251         rr_len = sizeof(rr);
4252         e=sldns_str2wire_rr_buf((char*)sldns_buffer_begin(buf), rr, &rr_len,
4253                 &dname_len, pstate.default_ttl,
4254                 pstate.origin_len?pstate.origin:NULL, pstate.origin_len,
4255                 pstate.prev_rr_len?pstate.prev_rr:NULL, pstate.prev_rr_len);
4256         if(e != 0) {
4257                 log_err("parse failure on first RR[%d]: %s",
4258                         LDNS_WIREPARSE_OFFSET(e),
4259                         sldns_get_errorstr_parse(LDNS_WIREPARSE_ERROR(e)));
4260                 return 0;
4261         }
4262         /* check that class is correct */
4263         if(sldns_wirerr_get_class(rr, rr_len, dname_len) != xfr->dclass) {
4264                 log_err("parse failure: first record in downloaded zonefile "
4265                         "from wrong RR class");
4266                 return 0;
4267         }
4268         return 1;
4269 }
4270
4271 /** sum sizes of chunklist */
4272 static size_t
4273 chunklist_sum(struct auth_chunk* list)
4274 {
4275         struct auth_chunk* p;
4276         size_t s = 0;
4277         for(p=list; p; p=p->next) {
4278                 s += p->len;
4279         }
4280         return s;
4281 }
4282
4283 /** remove newlines from collated line */
4284 static void
4285 chunkline_newline_removal(sldns_buffer* buf)
4286 {
4287         size_t i, end=sldns_buffer_limit(buf);
4288         for(i=0; i<end; i++) {
4289                 char c = (char)sldns_buffer_read_u8_at(buf, i);
4290                 if(c == '\n' && i==end-1) {
4291                         sldns_buffer_write_u8_at(buf, i, 0);
4292                         sldns_buffer_set_limit(buf, end-1);
4293                         return;
4294                 }
4295                 if(c == '\n')
4296                         sldns_buffer_write_u8_at(buf, i, (uint8_t)' ');
4297         }
4298 }
4299
4300 /** for http download, parse and add RR to zone */
4301 static int
4302 http_parse_add_rr(struct auth_xfer* xfr, struct auth_zone* z,
4303         sldns_buffer* buf, struct sldns_file_parse_state* pstate)
4304 {
4305         uint8_t rr[LDNS_RR_BUF_SIZE];
4306         size_t rr_len, dname_len = 0;
4307         int e;
4308         char* line = (char*)sldns_buffer_begin(buf);
4309         rr_len = sizeof(rr);
4310         e = sldns_str2wire_rr_buf(line, rr, &rr_len, &dname_len,
4311                 pstate->default_ttl,
4312                 pstate->origin_len?pstate->origin:NULL, pstate->origin_len,
4313                 pstate->prev_rr_len?pstate->prev_rr:NULL, pstate->prev_rr_len);
4314         if(e != 0) {
4315                 log_err("%s/%s parse failure RR[%d]: %s in '%s'",
4316                         xfr->task_transfer->master->host,
4317                         xfr->task_transfer->master->file,
4318                         LDNS_WIREPARSE_OFFSET(e),
4319                         sldns_get_errorstr_parse(LDNS_WIREPARSE_ERROR(e)),
4320                         line);
4321                 return 0;
4322         }
4323         if(rr_len == 0)
4324                 return 1; /* empty line or so */
4325
4326         /* set prev */
4327         if(dname_len < sizeof(pstate->prev_rr)) {
4328                 memmove(pstate->prev_rr, rr, dname_len);
4329                 pstate->prev_rr_len = dname_len;
4330         }
4331
4332         return az_insert_rr(z, rr, rr_len, dname_len, NULL);
4333 }
4334
4335 /** RR list iterator, returns RRs from answer section one by one from the
4336  * dns packets in the chunklist */
4337 static void
4338 chunk_rrlist_start(struct auth_xfer* xfr, struct auth_chunk** rr_chunk,
4339         int* rr_num, size_t* rr_pos)
4340 {
4341         *rr_chunk = xfr->task_transfer->chunks_first;
4342         *rr_num = 0;
4343         *rr_pos = 0;
4344 }
4345
4346 /** RR list iterator, see if we are at the end of the list */
4347 static int
4348 chunk_rrlist_end(struct auth_chunk* rr_chunk, int rr_num)
4349 {
4350         while(rr_chunk) {
4351                 if(rr_chunk->len < LDNS_HEADER_SIZE)
4352                         return 1;
4353                 if(rr_num < (int)LDNS_ANCOUNT(rr_chunk->data))
4354                         return 0;
4355                 /* no more RRs in this chunk */
4356                 /* continue with next chunk, see if it has RRs */
4357                 rr_chunk = rr_chunk->next;
4358                 rr_num = 0;
4359         }
4360         return 1;
4361 }
4362
4363 /** RR list iterator, move to next RR */
4364 static void
4365 chunk_rrlist_gonext(struct auth_chunk** rr_chunk, int* rr_num,
4366         size_t* rr_pos, size_t rr_nextpos)
4367 {
4368         /* already at end of chunks? */
4369         if(!*rr_chunk)
4370                 return;
4371         /* move within this chunk */
4372         if((*rr_chunk)->len >= LDNS_HEADER_SIZE &&
4373                 (*rr_num)+1 < (int)LDNS_ANCOUNT((*rr_chunk)->data)) {
4374                 (*rr_num) += 1;
4375                 *rr_pos = rr_nextpos;
4376                 return;
4377         }
4378         /* no more RRs in this chunk */
4379         /* continue with next chunk, see if it has RRs */
4380         if(*rr_chunk)
4381                 *rr_chunk = (*rr_chunk)->next;
4382         while(*rr_chunk) {
4383                 *rr_num = 0;
4384                 *rr_pos = 0;
4385                 if((*rr_chunk)->len >= LDNS_HEADER_SIZE &&
4386                         LDNS_ANCOUNT((*rr_chunk)->data) > 0) {
4387                         return;
4388                 }
4389                 *rr_chunk = (*rr_chunk)->next;
4390         }
4391 }
4392
4393 /** RR iterator, get current RR information, false on parse error */
4394 static int
4395 chunk_rrlist_get_current(struct auth_chunk* rr_chunk, int rr_num,
4396         size_t rr_pos, uint8_t** rr_dname, uint16_t* rr_type,
4397         uint16_t* rr_class, uint32_t* rr_ttl, uint16_t* rr_rdlen,
4398         uint8_t** rr_rdata, size_t* rr_nextpos)
4399 {
4400         sldns_buffer pkt;
4401         /* integrity checks on position */
4402         if(!rr_chunk) return 0;
4403         if(rr_chunk->len < LDNS_HEADER_SIZE) return 0;
4404         if(rr_num >= (int)LDNS_ANCOUNT(rr_chunk->data)) return 0;
4405         if(rr_pos >= rr_chunk->len) return 0;
4406
4407         /* fetch rr information */
4408         sldns_buffer_init_frm_data(&pkt, rr_chunk->data, rr_chunk->len);
4409         if(rr_pos == 0) {
4410                 size_t i;
4411                 /* skip question section */
4412                 sldns_buffer_set_position(&pkt, LDNS_HEADER_SIZE);
4413                 for(i=0; i<LDNS_QDCOUNT(rr_chunk->data); i++) {
4414                         if(pkt_dname_len(&pkt) == 0) return 0;
4415                         if(sldns_buffer_remaining(&pkt) < 4) return 0;
4416                         sldns_buffer_skip(&pkt, 4); /* type and class */
4417                 }
4418         } else  {
4419                 sldns_buffer_set_position(&pkt, rr_pos);
4420         }
4421         *rr_dname = sldns_buffer_current(&pkt);
4422         if(pkt_dname_len(&pkt) == 0) return 0;
4423         if(sldns_buffer_remaining(&pkt) < 10) return 0;
4424         *rr_type = sldns_buffer_read_u16(&pkt);
4425         *rr_class = sldns_buffer_read_u16(&pkt);
4426         *rr_ttl = sldns_buffer_read_u32(&pkt);
4427         *rr_rdlen = sldns_buffer_read_u16(&pkt);
4428         if(sldns_buffer_remaining(&pkt) < (*rr_rdlen)) return 0;
4429         *rr_rdata = sldns_buffer_current(&pkt);
4430         sldns_buffer_skip(&pkt, (ssize_t)(*rr_rdlen));
4431         *rr_nextpos = sldns_buffer_position(&pkt);
4432         return 1;
4433 }
4434
4435 /** print log message where we are in parsing the zone transfer */
4436 static void
4437 log_rrlist_position(const char* label, struct auth_chunk* rr_chunk,
4438         uint8_t* rr_dname, uint16_t rr_type, size_t rr_counter)
4439 {
4440         sldns_buffer pkt;
4441         size_t dlen;
4442         uint8_t buf[256];
4443         char str[256];
4444         char typestr[32];
4445         sldns_buffer_init_frm_data(&pkt, rr_chunk->data, rr_chunk->len);
4446         sldns_buffer_set_position(&pkt, (size_t)(rr_dname -
4447                 sldns_buffer_begin(&pkt)));
4448         if((dlen=pkt_dname_len(&pkt)) == 0) return;
4449         if(dlen >= sizeof(buf)) return;
4450         dname_pkt_copy(&pkt, buf, rr_dname);
4451         dname_str(buf, str);
4452         (void)sldns_wire2str_type_buf(rr_type, typestr, sizeof(typestr));
4453         verbose(VERB_ALGO, "%s at[%d] %s %s", label, (int)rr_counter,
4454                 str, typestr);
4455 }
4456
4457 /** check that start serial is OK for ixfr. we are at rr_counter == 0,
4458  * and we are going to check rr_counter == 1 (has to be type SOA) serial */
4459 static int
4460 ixfr_start_serial(struct auth_chunk* rr_chunk, int rr_num, size_t rr_pos,
4461         uint8_t* rr_dname, uint16_t rr_type, uint16_t rr_class,
4462         uint32_t rr_ttl, uint16_t rr_rdlen, uint8_t* rr_rdata,
4463         size_t rr_nextpos, uint32_t transfer_serial, uint32_t xfr_serial)
4464 {
4465         uint32_t startserial;
4466         /* move forward on RR */
4467         chunk_rrlist_gonext(&rr_chunk, &rr_num, &rr_pos, rr_nextpos);
4468         if(chunk_rrlist_end(rr_chunk, rr_num)) {
4469                 /* no second SOA */
4470                 verbose(VERB_OPS, "IXFR has no second SOA record");
4471                 return 0;
4472         }
4473         if(!chunk_rrlist_get_current(rr_chunk, rr_num, rr_pos,
4474                 &rr_dname, &rr_type, &rr_class, &rr_ttl, &rr_rdlen,
4475                 &rr_rdata, &rr_nextpos)) {
4476                 verbose(VERB_OPS, "IXFR cannot parse second SOA record");
4477                 /* failed to parse RR */
4478                 return 0;
4479         }
4480         if(rr_type != LDNS_RR_TYPE_SOA) {
4481                 verbose(VERB_OPS, "IXFR second record is not type SOA");
4482                 return 0;
4483         }
4484         if(rr_rdlen < 22) {
4485                 verbose(VERB_OPS, "IXFR, second SOA has short rdlength");
4486                 return 0; /* bad SOA rdlen */
4487         }
4488         startserial = sldns_read_uint32(rr_rdata+rr_rdlen-20);
4489         if(startserial == transfer_serial) {
4490                 /* empty AXFR, not an IXFR */
4491                 verbose(VERB_OPS, "IXFR second serial same as first");
4492                 return 0;
4493         }
4494         if(startserial != xfr_serial) {
4495                 /* wrong start serial, it does not match the serial in
4496                  * memory */
4497                 verbose(VERB_OPS, "IXFR is from serial %u to %u but %u "
4498                         "in memory, rejecting the zone transfer",
4499                         (unsigned)startserial, (unsigned)transfer_serial,
4500                         (unsigned)xfr_serial);
4501                 return 0;
4502         }
4503         /* everything OK in second SOA serial */
4504         return 1;
4505 }
4506
4507 /** apply IXFR to zone in memory. z is locked. false on failure(mallocfail) */
4508 static int
4509 apply_ixfr(struct auth_xfer* xfr, struct auth_zone* z,
4510         struct sldns_buffer* scratch_buffer)
4511 {
4512         struct auth_chunk* rr_chunk;
4513         int rr_num;
4514         size_t rr_pos;
4515         uint8_t* rr_dname, *rr_rdata;
4516         uint16_t rr_type, rr_class, rr_rdlen;
4517         uint32_t rr_ttl;
4518         size_t rr_nextpos;
4519         int have_transfer_serial = 0;
4520         uint32_t transfer_serial = 0;
4521         size_t rr_counter = 0;
4522         int delmode = 0;
4523         int softfail = 0;
4524
4525         /* start RR iterator over chunklist of packets */
4526         chunk_rrlist_start(xfr, &rr_chunk, &rr_num, &rr_pos);
4527         while(!chunk_rrlist_end(rr_chunk, rr_num)) {
4528                 if(!chunk_rrlist_get_current(rr_chunk, rr_num, rr_pos,
4529                         &rr_dname, &rr_type, &rr_class, &rr_ttl, &rr_rdlen,
4530                         &rr_rdata, &rr_nextpos)) {
4531                         /* failed to parse RR */
4532                         return 0;
4533                 }
4534                 if(verbosity>=7) log_rrlist_position("apply ixfr",
4535                         rr_chunk, rr_dname, rr_type, rr_counter);
4536                 /* twiddle add/del mode and check for start and end */
4537                 if(rr_counter == 0 && rr_type != LDNS_RR_TYPE_SOA)
4538                         return 0;
4539                 if(rr_counter == 1 && rr_type != LDNS_RR_TYPE_SOA) {
4540                         /* this is an AXFR returned from the IXFR master */
4541                         /* but that should already have been detected, by
4542                          * on_ixfr_is_axfr */
4543                         return 0;
4544                 }
4545                 if(rr_type == LDNS_RR_TYPE_SOA) {
4546                         uint32_t serial;
4547                         if(rr_rdlen < 22) return 0; /* bad SOA rdlen */
4548                         serial = sldns_read_uint32(rr_rdata+rr_rdlen-20);
4549                         if(have_transfer_serial == 0) {
4550                                 have_transfer_serial = 1;
4551                                 transfer_serial = serial;
4552                                 delmode = 1; /* gets negated below */
4553                                 /* check second RR before going any further */
4554                                 if(!ixfr_start_serial(rr_chunk, rr_num, rr_pos,
4555                                         rr_dname, rr_type, rr_class, rr_ttl,
4556                                         rr_rdlen, rr_rdata, rr_nextpos,
4557                                         transfer_serial, xfr->serial)) {
4558                                         return 0;
4559                                 }
4560                         } else if(transfer_serial == serial) {
4561                                 have_transfer_serial++;
4562                                 if(rr_counter == 1) {
4563                                         /* empty AXFR, with SOA; SOA; */
4564                                         /* should have been detected by
4565                                          * on_ixfr_is_axfr */
4566                                         return 0;
4567                                 }
4568                                 if(have_transfer_serial == 3) {
4569                                         /* see serial three times for end */
4570                                         /* eg. IXFR:
4571                                          *  SOA 3 start
4572                                          *  SOA 1 second RR, followed by del
4573                                          *  SOA 2 followed by add
4574                                          *  SOA 2 followed by del
4575                                          *  SOA 3 followed by add
4576                                          *  SOA 3 end */
4577                                         /* ended by SOA record */
4578                                         xfr->serial = transfer_serial;
4579                                         break;
4580                                 }
4581                         }
4582                         /* twiddle add/del mode */
4583                         /* switch from delete part to add part and back again
4584                          * just before the soa, it gets deleted and added too
4585                          * this means we switch to delete mode for the final
4586                          * SOA(so skip that one) */
4587                         delmode = !delmode;
4588                 }
4589                 /* process this RR */
4590                 /* if the RR is deleted twice or added twice, then we
4591                  * softfail, and continue with the rest of the IXFR, so
4592                  * that we serve something fairly nice during the refetch */
4593                 if(verbosity>=7) log_rrlist_position((delmode?"del":"add"),
4594                         rr_chunk, rr_dname, rr_type, rr_counter);
4595                 if(delmode) {
4596                         /* delete this RR */
4597                         int nonexist = 0;
4598                         if(!az_remove_rr_decompress(z, rr_chunk->data,
4599                                 rr_chunk->len, scratch_buffer, rr_dname,
4600                                 rr_type, rr_class, rr_ttl, rr_rdata, rr_rdlen,
4601                                 &nonexist)) {
4602                                 /* failed, malloc error or so */
4603                                 return 0;
4604                         }
4605                         if(nonexist) {
4606                                 /* it was removal of a nonexisting RR */
4607                                 if(verbosity>=4) log_rrlist_position(
4608                                         "IXFR error nonexistent RR",
4609                                         rr_chunk, rr_dname, rr_type, rr_counter);
4610                                 softfail = 1;
4611                         }
4612                 } else if(rr_counter != 0) {
4613                         /* skip first SOA RR for addition, it is added in
4614                          * the addition part near the end of the ixfr, when
4615                          * that serial is seen the second time. */
4616                         int duplicate = 0;
4617                         /* add this RR */
4618                         if(!az_insert_rr_decompress(z, rr_chunk->data,
4619                                 rr_chunk->len, scratch_buffer, rr_dname,
4620                                 rr_type, rr_class, rr_ttl, rr_rdata, rr_rdlen,
4621                                 &duplicate)) {
4622                                 /* failed, malloc error or so */
4623                                 return 0;
4624                         }
4625                         if(duplicate) {
4626                                 /* it was a duplicate */
4627                                 if(verbosity>=4) log_rrlist_position(
4628                                         "IXFR error duplicate RR",
4629                                         rr_chunk, rr_dname, rr_type, rr_counter);
4630                                 softfail = 1;
4631                         }
4632                 }
4633
4634                 rr_counter++;
4635                 chunk_rrlist_gonext(&rr_chunk, &rr_num, &rr_pos, rr_nextpos);
4636         }
4637         if(softfail) {
4638                 verbose(VERB_ALGO, "IXFR did not apply cleanly, fetching full zone");
4639                 return 0;
4640         }
4641         return 1;
4642 }
4643
4644 /** apply AXFR to zone in memory. z is locked. false on failure(mallocfail) */
4645 static int
4646 apply_axfr(struct auth_xfer* xfr, struct auth_zone* z,
4647         struct sldns_buffer* scratch_buffer)
4648 {
4649         struct auth_chunk* rr_chunk;
4650         int rr_num;
4651         size_t rr_pos;
4652         uint8_t* rr_dname, *rr_rdata;
4653         uint16_t rr_type, rr_class, rr_rdlen;
4654         uint32_t rr_ttl;
4655         uint32_t serial = 0;
4656         size_t rr_nextpos;
4657         size_t rr_counter = 0;
4658         int have_end_soa = 0;
4659
4660         /* clear the data tree */
4661         traverse_postorder(&z->data, auth_data_del, NULL);
4662         rbtree_init(&z->data, &auth_data_cmp);
4663         xfr->have_zone = 0;
4664         xfr->serial = 0;
4665
4666         /* insert all RRs in to the zone */
4667         /* insert the SOA only once, skip the last one */
4668         /* start RR iterator over chunklist of packets */
4669         chunk_rrlist_start(xfr, &rr_chunk, &rr_num, &rr_pos);
4670         while(!chunk_rrlist_end(rr_chunk, rr_num)) {
4671                 if(!chunk_rrlist_get_current(rr_chunk, rr_num, rr_pos,
4672                         &rr_dname, &rr_type, &rr_class, &rr_ttl, &rr_rdlen,
4673                         &rr_rdata, &rr_nextpos)) {
4674                         /* failed to parse RR */
4675                         return 0;
4676                 }
4677                 if(verbosity>=7) log_rrlist_position("apply_axfr",
4678                         rr_chunk, rr_dname, rr_type, rr_counter);
4679                 if(rr_type == LDNS_RR_TYPE_SOA) {
4680                         if(rr_counter != 0) {
4681                                 /* end of the axfr */
4682                                 have_end_soa = 1;
4683                                 break;
4684                         }
4685                         if(rr_rdlen < 22) return 0; /* bad SOA rdlen */
4686                         serial = sldns_read_uint32(rr_rdata+rr_rdlen-20);
4687                 }
4688
4689                 /* add this RR */
4690                 if(!az_insert_rr_decompress(z, rr_chunk->data, rr_chunk->len,
4691                         scratch_buffer, rr_dname, rr_type, rr_class, rr_ttl,
4692                         rr_rdata, rr_rdlen, NULL)) {
4693                         /* failed, malloc error or so */
4694                         return 0;
4695                 }
4696
4697                 rr_counter++;
4698                 chunk_rrlist_gonext(&rr_chunk, &rr_num, &rr_pos, rr_nextpos);
4699         }
4700         if(!have_end_soa) {
4701                 log_err("no end SOA record for AXFR");
4702                 return 0;
4703         }
4704
4705         xfr->serial = serial;
4706         xfr->have_zone = 1;
4707         return 1;
4708 }
4709
4710 /** apply HTTP to zone in memory. z is locked. false on failure(mallocfail) */
4711 static int
4712 apply_http(struct auth_xfer* xfr, struct auth_zone* z,
4713         struct sldns_buffer* scratch_buffer)
4714 {
4715         /* parse data in chunks */
4716         /* parse RR's and read into memory. ignore $INCLUDE from the
4717          * downloaded file*/
4718         struct sldns_file_parse_state pstate;
4719         struct auth_chunk* chunk;
4720         size_t chunk_pos;
4721         memset(&pstate, 0, sizeof(pstate));
4722         pstate.default_ttl = 3600;
4723         if(xfr->namelen < sizeof(pstate.origin)) {
4724                 pstate.origin_len = xfr->namelen;
4725                 memmove(pstate.origin, xfr->name, xfr->namelen);
4726         }
4727
4728         if(verbosity >= VERB_ALGO)
4729                 verbose(VERB_ALGO, "http download %s of size %d",
4730                 xfr->task_transfer->master->file,
4731                 (int)chunklist_sum(xfr->task_transfer->chunks_first));
4732         if(xfr->task_transfer->chunks_first && verbosity >= VERB_ALGO) {
4733                 char preview[1024];
4734                 if(xfr->task_transfer->chunks_first->len+1 > sizeof(preview)) {
4735                         memmove(preview, xfr->task_transfer->chunks_first->data,
4736                                 sizeof(preview)-1);
4737                         preview[sizeof(preview)-1]=0;
4738                 } else {
4739                         memmove(preview, xfr->task_transfer->chunks_first->data,
4740                                 xfr->task_transfer->chunks_first->len);
4741                         preview[xfr->task_transfer->chunks_first->len]=0;
4742                 }
4743                 log_info("auth zone http downloaded content preview: %s",
4744                         preview);
4745         }
4746
4747         /* perhaps a little syntax check before we try to apply the data? */
4748         if(!http_zonefile_syntax_check(xfr, scratch_buffer)) {
4749                 log_err("http download %s/%s does not contain a zonefile, "
4750                         "but got '%s'", xfr->task_transfer->master->host,
4751                         xfr->task_transfer->master->file,
4752                         sldns_buffer_begin(scratch_buffer));
4753                 return 0;
4754         }
4755
4756         /* clear the data tree */
4757         traverse_postorder(&z->data, auth_data_del, NULL);
4758         rbtree_init(&z->data, &auth_data_cmp);
4759         xfr->have_zone = 0;
4760         xfr->serial = 0;
4761
4762         chunk = xfr->task_transfer->chunks_first;
4763         chunk_pos = 0;
4764         pstate.lineno = 0;
4765         while(chunkline_get_line_collated(&chunk, &chunk_pos, scratch_buffer)) {
4766                 /* process this line */
4767                 pstate.lineno++;
4768                 chunkline_newline_removal(scratch_buffer);
4769                 if(chunkline_is_comment_line_or_empty(scratch_buffer)) {
4770                         continue;
4771                 }
4772                 /* parse line and add RR */
4773                 if(http_parse_origin(scratch_buffer, &pstate)) {
4774                         continue; /* $ORIGIN has been handled */
4775                 }
4776                 if(http_parse_ttl(scratch_buffer, &pstate)) {
4777                         continue; /* $TTL has been handled */
4778                 }
4779                 if(!http_parse_add_rr(xfr, z, scratch_buffer, &pstate)) {
4780                         verbose(VERB_ALGO, "error parsing line [%s:%d] %s",
4781                                 xfr->task_transfer->master->file,
4782                                 pstate.lineno,
4783                                 sldns_buffer_begin(scratch_buffer));
4784                         return 0;
4785                 }
4786         }
4787         return 1;
4788 }
4789
4790 /** write http chunks to zonefile to create downloaded file */
4791 static int
4792 auth_zone_write_chunks(struct auth_xfer* xfr, const char* fname)
4793 {
4794         FILE* out;
4795         struct auth_chunk* p;
4796         out = fopen(fname, "w");
4797         if(!out) {
4798                 log_err("could not open %s: %s", fname, strerror(errno));
4799                 return 0;
4800         }
4801         for(p = xfr->task_transfer->chunks_first; p ; p = p->next) {
4802                 if(!write_out(out, (char*)p->data, p->len)) {
4803                         log_err("could not write http download to %s", fname);
4804                         fclose(out);
4805                         return 0;
4806                 }
4807         }
4808         fclose(out);
4809         return 1;
4810 }
4811
4812 /** write to zonefile after zone has been updated */
4813 static void
4814 xfr_write_after_update(struct auth_xfer* xfr, struct module_env* env)
4815 {
4816         struct config_file* cfg = env->cfg;
4817         struct auth_zone* z;
4818         char tmpfile[1024];
4819         char* zfilename;
4820         lock_basic_unlock(&xfr->lock);
4821
4822         /* get lock again, so it is a readlock and concurrently queries
4823          * can be answered */
4824         lock_rw_rdlock(&env->auth_zones->lock);
4825         z = auth_zone_find(env->auth_zones, xfr->name, xfr->namelen,
4826                 xfr->dclass);
4827         if(!z) {
4828                 lock_rw_unlock(&env->auth_zones->lock);
4829                 /* the zone is gone, ignore xfr results */
4830                 lock_basic_lock(&xfr->lock);
4831                 return;
4832         }
4833         lock_rw_rdlock(&z->lock);
4834         lock_basic_lock(&xfr->lock);
4835         lock_rw_unlock(&env->auth_zones->lock);
4836
4837         if(z->zonefile == NULL || z->zonefile[0] == 0) {
4838                 lock_rw_unlock(&z->lock);
4839                 /* no write needed, no zonefile set */
4840                 return;
4841         }
4842         zfilename = z->zonefile;
4843         if(cfg->chrootdir && cfg->chrootdir[0] && strncmp(zfilename,
4844                 cfg->chrootdir, strlen(cfg->chrootdir)) == 0)
4845                 zfilename += strlen(cfg->chrootdir);
4846
4847         /* write to tempfile first */
4848         if((size_t)strlen(zfilename) + 16 > sizeof(tmpfile)) {
4849                 verbose(VERB_ALGO, "tmpfilename too long, cannot update "
4850                         " zonefile %s", zfilename);
4851                 lock_rw_unlock(&z->lock);
4852                 return;
4853         }
4854         snprintf(tmpfile, sizeof(tmpfile), "%s.tmp%u", zfilename,
4855                 (unsigned)getpid());
4856         if(xfr->task_transfer->master->http) {
4857                 /* use the stored chunk list to write them */
4858                 if(!auth_zone_write_chunks(xfr, tmpfile)) {
4859                         unlink(tmpfile);
4860                         lock_rw_unlock(&z->lock);
4861                 }
4862         } else if(!auth_zone_write_file(z, tmpfile)) {
4863                 unlink(tmpfile);
4864                 lock_rw_unlock(&z->lock);
4865                 return;
4866         }
4867         if(rename(tmpfile, zfilename) < 0) {
4868                 log_err("could not rename(%s, %s): %s", tmpfile, zfilename,
4869                         strerror(errno));
4870                 unlink(tmpfile);
4871                 lock_rw_unlock(&z->lock);
4872                 return;
4873         }
4874         lock_rw_unlock(&z->lock);
4875 }
4876
4877 /** process chunk list and update zone in memory,
4878  * return false if it did not work */
4879 static int
4880 xfr_process_chunk_list(struct auth_xfer* xfr, struct module_env* env,
4881         int* ixfr_fail)
4882 {
4883         struct auth_zone* z;
4884
4885         /* obtain locks and structures */
4886         /* release xfr lock, then, while holding az->lock grab both
4887          * z->lock and xfr->lock */
4888         lock_basic_unlock(&xfr->lock);
4889         lock_rw_rdlock(&env->auth_zones->lock);
4890         z = auth_zone_find(env->auth_zones, xfr->name, xfr->namelen,
4891                 xfr->dclass);
4892         if(!z) {
4893                 lock_rw_unlock(&env->auth_zones->lock);
4894                 /* the zone is gone, ignore xfr results */
4895                 lock_basic_lock(&xfr->lock);
4896                 return 0;
4897         }
4898         lock_rw_wrlock(&z->lock);
4899         lock_basic_lock(&xfr->lock);
4900         lock_rw_unlock(&env->auth_zones->lock);
4901
4902         /* apply data */
4903         if(xfr->task_transfer->master->http) {
4904                 if(!apply_http(xfr, z, env->scratch_buffer)) {
4905                         lock_rw_unlock(&z->lock);
4906                         verbose(VERB_ALGO, "http from %s: could not store data",
4907                                 xfr->task_transfer->master->host);
4908                         return 0;
4909                 }
4910         } else if(xfr->task_transfer->on_ixfr &&
4911                 !xfr->task_transfer->on_ixfr_is_axfr) {
4912                 if(!apply_ixfr(xfr, z, env->scratch_buffer)) {
4913                         lock_rw_unlock(&z->lock);
4914                         verbose(VERB_ALGO, "xfr from %s: could not store IXFR"
4915                                 " data", xfr->task_transfer->master->host);
4916                         *ixfr_fail = 1;
4917                         return 0;
4918                 }
4919         } else {
4920                 if(!apply_axfr(xfr, z, env->scratch_buffer)) {
4921                         lock_rw_unlock(&z->lock);
4922                         verbose(VERB_ALGO, "xfr from %s: could not store AXFR"
4923                                 " data", xfr->task_transfer->master->host);
4924                         return 0;
4925                 }
4926         }
4927         xfr->zone_expired = 0;
4928         z->zone_expired = 0;
4929         if(!xfr_find_soa(z, xfr)) {
4930                 lock_rw_unlock(&z->lock);
4931                 verbose(VERB_ALGO, "xfr from %s: no SOA in zone after update"
4932                         " (or malformed RR)", xfr->task_transfer->master->host);
4933                 return 0;
4934         }
4935         if(xfr->have_zone)
4936                 xfr->lease_time = *env->now;
4937
4938         /* unlock */
4939         lock_rw_unlock(&z->lock);
4940
4941         if(verbosity >= VERB_QUERY && xfr->have_zone) {
4942                 char zname[256];
4943                 dname_str(xfr->name, zname);
4944                 verbose(VERB_QUERY, "auth zone %s updated to serial %u", zname,
4945                         (unsigned)xfr->serial);
4946         }
4947         /* see if we need to write to a zonefile */
4948         xfr_write_after_update(xfr, env);
4949         return 1;
4950 }
4951
4952 /** disown task_transfer.  caller must hold xfr.lock */
4953 static void
4954 xfr_transfer_disown(struct auth_xfer* xfr)
4955 {
4956         /* remove the commpoint */
4957         comm_point_delete(xfr->task_transfer->cp);
4958         xfr->task_transfer->cp = NULL;
4959         /* we don't own this item anymore */
4960         xfr->task_transfer->worker = NULL;
4961         xfr->task_transfer->env = NULL;
4962 }
4963
4964 /** lookup a host name for its addresses, if needed */
4965 static int
4966 xfr_transfer_lookup_host(struct auth_xfer* xfr, struct module_env* env)
4967 {
4968         struct sockaddr_storage addr;
4969         socklen_t addrlen = 0;
4970         struct auth_master* master = xfr->task_transfer->lookup_target;
4971         struct query_info qinfo;
4972         uint16_t qflags = BIT_RD;
4973         uint8_t dname[LDNS_MAX_DOMAINLEN+1];
4974         struct edns_data edns;
4975         sldns_buffer* buf = env->scratch_buffer;
4976         if(!master) return 0;
4977         if(extstrtoaddr(master->host, &addr, &addrlen)) {
4978                 /* not needed, host is in IP addr format */
4979                 return 0;
4980         }
4981         if(master->allow_notify)
4982                 return 0; /* allow-notifies are not transferred from, no
4983                 lookup is needed */
4984
4985         /* use mesh_new_callback to probe for non-addr hosts,
4986          * and then wait for them to be looked up (in cache, or query) */
4987         qinfo.qname_len = sizeof(dname);
4988         if(sldns_str2wire_dname_buf(master->host, dname, &qinfo.qname_len)
4989                 != 0) {
4990                 log_err("cannot parse host name of master %s", master->host);
4991                 return 0;
4992         }
4993         qinfo.qname = dname;
4994         qinfo.qclass = xfr->dclass;
4995         qinfo.qtype = LDNS_RR_TYPE_A;
4996         if(xfr->task_transfer->lookup_aaaa)
4997                 qinfo.qtype = LDNS_RR_TYPE_AAAA;
4998         qinfo.local_alias = NULL;
4999         if(verbosity >= VERB_ALGO) {
5000                 char buf1[512];
5001                 char buf2[LDNS_MAX_DOMAINLEN+1];
5002                 dname_str(xfr->name, buf2);
5003                 snprintf(buf1, sizeof(buf1), "auth zone %s: master lookup"
5004                         " for task_transfer", buf2);
5005                 log_query_info(VERB_ALGO, buf1, &qinfo);
5006         }
5007         edns.edns_present = 1;
5008         edns.ext_rcode = 0;
5009         edns.edns_version = 0;
5010         edns.bits = EDNS_DO;
5011         edns.opt_list = NULL;
5012         if(sldns_buffer_capacity(buf) < 65535)
5013                 edns.udp_size = (uint16_t)sldns_buffer_capacity(buf);
5014         else    edns.udp_size = 65535;
5015
5016         /* unlock xfr during mesh_new_callback() because the callback can be
5017          * called straight away */
5018         lock_basic_unlock(&xfr->lock);
5019         if(!mesh_new_callback(env->mesh, &qinfo, qflags, &edns, buf, 0,
5020                 &auth_xfer_transfer_lookup_callback, xfr)) {
5021                 lock_basic_lock(&xfr->lock);
5022                 log_err("out of memory lookup up master %s", master->host);
5023                 return 0;
5024         }
5025         lock_basic_lock(&xfr->lock);
5026         return 1;
5027 }
5028
5029 /** initiate TCP to the target and fetch zone.
5030  * returns true if that was successfully started, and timeout setup. */
5031 static int
5032 xfr_transfer_init_fetch(struct auth_xfer* xfr, struct module_env* env)
5033 {
5034         struct sockaddr_storage addr;
5035         socklen_t addrlen = 0;
5036         struct auth_master* master = xfr->task_transfer->master;
5037         if(!master) return 0;
5038         if(master->allow_notify) return 0; /* only for notify */
5039
5040         /* get master addr */
5041         if(xfr->task_transfer->scan_addr) {
5042                 addrlen = xfr->task_transfer->scan_addr->addrlen;
5043                 memmove(&addr, &xfr->task_transfer->scan_addr->addr, addrlen);
5044         } else {
5045                 if(!extstrtoaddr(master->host, &addr, &addrlen)) {
5046                         /* the ones that are not in addr format are supposed
5047                          * to be looked up.  The lookup has failed however,
5048                          * so skip them */
5049                         char zname[255+1];
5050                         dname_str(xfr->name, zname);
5051                         log_err("%s: failed lookup, cannot transfer from master %s",
5052                                 zname, master->host);
5053                         return 0;
5054                 }
5055         }
5056
5057         /* remove previous TCP connection (if any) */
5058         if(xfr->task_transfer->cp) {
5059                 comm_point_delete(xfr->task_transfer->cp);
5060                 xfr->task_transfer->cp = NULL;
5061         }
5062
5063         if(master->http) {
5064                 /* perform http fetch */
5065                 /* store http port number into sockaddr,
5066                  * unless someone used unbound's host@port notation */
5067                 if(strchr(master->host, '@') == NULL)
5068                         sockaddr_store_port(&addr, addrlen, master->port);
5069                 xfr->task_transfer->cp = outnet_comm_point_for_http(
5070                         env->outnet, auth_xfer_transfer_http_callback, xfr,
5071                         &addr, addrlen, AUTH_TRANSFER_TIMEOUT, master->ssl,
5072                         master->host, master->file);
5073                 if(!xfr->task_transfer->cp) {
5074                         char zname[255+1];
5075                         dname_str(xfr->name, zname);
5076                         verbose(VERB_ALGO, "cannot create http cp "
5077                                 "connection for %s to %s", zname,
5078                                 master->host);
5079                         return 0;
5080                 }
5081                 return 1;
5082         }
5083
5084         /* perform AXFR/IXFR */
5085         /* set the packet to be written */
5086         /* create new ID */
5087         xfr->task_transfer->id = (uint16_t)(ub_random(env->rnd)&0xffff);
5088         xfr_create_ixfr_packet(xfr, env->scratch_buffer,
5089                 xfr->task_transfer->id, master);
5090
5091         /* connect on fd */
5092         xfr->task_transfer->cp = outnet_comm_point_for_tcp(env->outnet,
5093                 auth_xfer_transfer_tcp_callback, xfr, &addr, addrlen,
5094                 env->scratch_buffer, AUTH_TRANSFER_TIMEOUT);
5095         if(!xfr->task_transfer->cp) {
5096                 char zname[255+1];
5097                 dname_str(xfr->name, zname);
5098                 verbose(VERB_ALGO, "cannot create tcp cp connection for "
5099                         "xfr %s to %s", zname, master->host);
5100                 return 0;
5101         }
5102         return 1;
5103 }
5104
5105 /** perform next lookup, next transfer TCP, or end and resume wait time task */
5106 static void
5107 xfr_transfer_nexttarget_or_end(struct auth_xfer* xfr, struct module_env* env)
5108 {
5109         log_assert(xfr->task_transfer->worker == env->worker);
5110
5111         /* are we performing lookups? */
5112         while(xfr->task_transfer->lookup_target) {
5113                 if(xfr_transfer_lookup_host(xfr, env)) {
5114                         /* wait for lookup to finish,
5115                          * note that the hostname may be in unbound's cache
5116                          * and we may then get an instant cache response,
5117                          * and that calls the callback just like a full
5118                          * lookup and lookup failures also call callback */
5119                         lock_basic_unlock(&xfr->lock);
5120                         return;
5121                 }
5122                 xfr_transfer_move_to_next_lookup(xfr, env);
5123         }
5124
5125         /* initiate TCP and fetch the zone from the master */
5126         /* and set timeout on it */
5127         while(!xfr_transfer_end_of_list(xfr)) {
5128                 xfr->task_transfer->master = xfr_transfer_current_master(xfr);
5129                 if(xfr_transfer_init_fetch(xfr, env)) {
5130                         /* successfully started, wait for callback */
5131                         lock_basic_unlock(&xfr->lock);
5132                         return;
5133                 }
5134                 /* failed to fetch, next master */
5135                 xfr_transfer_nextmaster(xfr);
5136         }
5137
5138         /* we failed to fetch the zone, move to wait task
5139          * use the shorter retry timeout */
5140         xfr_transfer_disown(xfr);
5141
5142         /* pick up the nextprobe task and wait */
5143         if(xfr->task_nextprobe->worker == NULL)
5144                 xfr_set_timeout(xfr, env, 1, 0);
5145         lock_basic_unlock(&xfr->lock);
5146 }
5147
5148 /** add addrs from A or AAAA rrset to the master */
5149 static void
5150 xfr_master_add_addrs(struct auth_master* m, struct ub_packed_rrset_key* rrset,
5151         uint16_t rrtype)
5152 {
5153         size_t i;
5154         struct packed_rrset_data* data;
5155         if(!m || !rrset) return;
5156         if(rrtype != LDNS_RR_TYPE_A && rrtype != LDNS_RR_TYPE_AAAA)
5157                 return;
5158         data = (struct packed_rrset_data*)rrset->entry.data;
5159         for(i=0; i<data->count; i++) {
5160                 struct auth_addr* a;
5161                 size_t len = data->rr_len[i] - 2;
5162                 uint8_t* rdata = data->rr_data[i]+2;
5163                 if(rrtype == LDNS_RR_TYPE_A && len != INET_SIZE)
5164                         continue; /* wrong length for A */
5165                 if(rrtype == LDNS_RR_TYPE_AAAA && len != INET6_SIZE)
5166                         continue; /* wrong length for AAAA */
5167
5168                 /* add and alloc it */
5169                 a = (struct auth_addr*)calloc(1, sizeof(*a));
5170                 if(!a) {
5171                         log_err("out of memory");
5172                         return;
5173                 }
5174                 if(rrtype == LDNS_RR_TYPE_A) {
5175                         struct sockaddr_in* sa;
5176                         a->addrlen = (socklen_t)sizeof(*sa);
5177                         sa = (struct sockaddr_in*)&a->addr;
5178                         sa->sin_family = AF_INET;
5179                         sa->sin_port = (in_port_t)htons(UNBOUND_DNS_PORT);
5180                         memmove(&sa->sin_addr, rdata, INET_SIZE);
5181                 } else {
5182                         struct sockaddr_in6* sa;
5183                         a->addrlen = (socklen_t)sizeof(*sa);
5184                         sa = (struct sockaddr_in6*)&a->addr;
5185                         sa->sin6_family = AF_INET6;
5186                         sa->sin6_port = (in_port_t)htons(UNBOUND_DNS_PORT);
5187                         memmove(&sa->sin6_addr, rdata, INET6_SIZE);
5188                 }
5189                 if(verbosity >= VERB_ALGO) {
5190                         char s[64];
5191                         addr_to_str(&a->addr, a->addrlen, s, sizeof(s));
5192                         verbose(VERB_ALGO, "auth host %s lookup %s",
5193                                 m->host, s);
5194                 }
5195                 /* append to list */
5196                 a->next = m->list;
5197                 m->list = a;
5198         }
5199 }
5200
5201 /** callback for task_transfer lookup of host name, of A or AAAA */
5202 void auth_xfer_transfer_lookup_callback(void* arg, int rcode, sldns_buffer* buf,
5203         enum sec_status ATTR_UNUSED(sec), char* ATTR_UNUSED(why_bogus),
5204         int ATTR_UNUSED(was_ratelimited))
5205 {
5206         struct auth_xfer* xfr = (struct auth_xfer*)arg;
5207         struct module_env* env;
5208         log_assert(xfr->task_transfer);
5209         lock_basic_lock(&xfr->lock);
5210         env = xfr->task_transfer->env;
5211         if(env->outnet->want_to_quit) {
5212                 lock_basic_unlock(&xfr->lock);
5213                 return; /* stop on quit */
5214         }
5215
5216         /* process result */
5217         if(rcode == LDNS_RCODE_NOERROR) {
5218                 uint16_t wanted_qtype = LDNS_RR_TYPE_A;
5219                 struct regional* temp = env->scratch;
5220                 struct query_info rq;
5221                 struct reply_info* rep;
5222                 if(xfr->task_transfer->lookup_aaaa)
5223                         wanted_qtype = LDNS_RR_TYPE_AAAA;
5224                 memset(&rq, 0, sizeof(rq));
5225                 rep = parse_reply_in_temp_region(buf, temp, &rq);
5226                 if(rep && rq.qtype == wanted_qtype &&
5227                         FLAGS_GET_RCODE(rep->flags) == LDNS_RCODE_NOERROR) {
5228                         /* parsed successfully */
5229                         struct ub_packed_rrset_key* answer =
5230                                 reply_find_answer_rrset(&rq, rep);
5231                         if(answer) {
5232                                 xfr_master_add_addrs(xfr->task_transfer->
5233                                         lookup_target, answer, wanted_qtype);
5234                         }
5235                 }
5236         }
5237         if(xfr->task_transfer->lookup_target->list &&
5238                 xfr->task_transfer->lookup_target == xfr_transfer_current_master(xfr))
5239                 xfr->task_transfer->scan_addr = xfr->task_transfer->lookup_target->list;
5240
5241         /* move to lookup AAAA after A lookup, move to next hostname lookup,
5242          * or move to fetch the zone, or, if nothing to do, end task_transfer */
5243         xfr_transfer_move_to_next_lookup(xfr, env);
5244         xfr_transfer_nexttarget_or_end(xfr, env);
5245 }
5246
5247 /** check if xfer (AXFR or IXFR) packet is OK.
5248  * return false if we lost connection (SERVFAIL, or unreadable).
5249  * return false if we need to move from IXFR to AXFR, with gonextonfail
5250  *      set to false, so the same master is tried again, but with AXFR.
5251  * return true if fine to link into data.
5252  * return true with transferdone=true when the transfer has ended.
5253  */
5254 static int
5255 check_xfer_packet(sldns_buffer* pkt, struct auth_xfer* xfr,
5256         int* gonextonfail, int* transferdone)
5257 {
5258         uint8_t* wire = sldns_buffer_begin(pkt);
5259         int i;
5260         if(sldns_buffer_limit(pkt) < LDNS_HEADER_SIZE) {
5261                 verbose(VERB_ALGO, "xfr to %s failed, packet too small",
5262                         xfr->task_transfer->master->host);
5263                 return 0;
5264         }
5265         if(!LDNS_QR_WIRE(wire)) {
5266                 verbose(VERB_ALGO, "xfr to %s failed, packet has no QR flag",
5267                         xfr->task_transfer->master->host);
5268                 return 0;
5269         }
5270         if(LDNS_TC_WIRE(wire)) {
5271                 verbose(VERB_ALGO, "xfr to %s failed, packet has TC flag",
5272                         xfr->task_transfer->master->host);
5273                 return 0;
5274         }
5275         /* check ID */
5276         if(LDNS_ID_WIRE(wire) != xfr->task_transfer->id) {
5277                 verbose(VERB_ALGO, "xfr to %s failed, packet wrong ID",
5278                         xfr->task_transfer->master->host);
5279                 return 0;
5280         }
5281         if(LDNS_RCODE_WIRE(wire) != LDNS_RCODE_NOERROR) {
5282                 char rcode[32];
5283                 sldns_wire2str_rcode_buf((int)LDNS_RCODE_WIRE(wire), rcode,
5284                         sizeof(rcode));
5285                 /* if we are doing IXFR, check for fallback */
5286                 if(xfr->task_transfer->on_ixfr) {
5287                         if(LDNS_RCODE_WIRE(wire) == LDNS_RCODE_NOTIMPL ||
5288                                 LDNS_RCODE_WIRE(wire) == LDNS_RCODE_SERVFAIL ||
5289                                 LDNS_RCODE_WIRE(wire) == LDNS_RCODE_REFUSED ||
5290                                 LDNS_RCODE_WIRE(wire) == LDNS_RCODE_FORMERR) {
5291                                 verbose(VERB_ALGO, "xfr to %s, fallback "
5292                                         "from IXFR to AXFR (with rcode %s)",
5293                                         xfr->task_transfer->master->host,
5294                                         rcode);
5295                                 xfr->task_transfer->ixfr_fail = 1;
5296                                 *gonextonfail = 0;
5297                                 return 0;
5298                         }
5299                 }
5300                 verbose(VERB_ALGO, "xfr to %s failed, packet with rcode %s",
5301                         xfr->task_transfer->master->host, rcode);
5302                 return 0;
5303         }
5304         if(LDNS_OPCODE_WIRE(wire) != LDNS_PACKET_QUERY) {
5305                 verbose(VERB_ALGO, "xfr to %s failed, packet with bad opcode",
5306                         xfr->task_transfer->master->host);
5307                 return 0;
5308         }
5309         if(LDNS_QDCOUNT(wire) > 1) {
5310                 verbose(VERB_ALGO, "xfr to %s failed, packet has qdcount %d",
5311                         xfr->task_transfer->master->host,
5312                         (int)LDNS_QDCOUNT(wire));
5313                 return 0;
5314         }
5315
5316         /* check qname */
5317         sldns_buffer_set_position(pkt, LDNS_HEADER_SIZE);
5318         for(i=0; i<(int)LDNS_QDCOUNT(wire); i++) {
5319                 size_t pos = sldns_buffer_position(pkt);
5320                 uint16_t qtype, qclass;
5321                 if(pkt_dname_len(pkt) == 0) {
5322                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5323                                 "malformed dname",
5324                                 xfr->task_transfer->master->host);
5325                         return 0;
5326                 }
5327                 if(dname_pkt_compare(pkt, sldns_buffer_at(pkt, pos),
5328                         xfr->name) != 0) {
5329                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5330                                 "wrong qname",
5331                                 xfr->task_transfer->master->host);
5332                         return 0;
5333                 }
5334                 if(sldns_buffer_remaining(pkt) < 4) {
5335                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5336                                 "truncated query RR",
5337                                 xfr->task_transfer->master->host);
5338                         return 0;
5339                 }
5340                 qtype = sldns_buffer_read_u16(pkt);
5341                 qclass = sldns_buffer_read_u16(pkt);
5342                 if(qclass != xfr->dclass) {
5343                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5344                                 "wrong qclass",
5345                                 xfr->task_transfer->master->host);
5346                         return 0;
5347                 }
5348                 if(xfr->task_transfer->on_ixfr) {
5349                         if(qtype != LDNS_RR_TYPE_IXFR) {
5350                                 verbose(VERB_ALGO, "xfr to %s failed, packet "
5351                                         "with wrong qtype, expected IXFR",
5352                                 xfr->task_transfer->master->host);
5353                                 return 0;
5354                         }
5355                 } else {
5356                         if(qtype != LDNS_RR_TYPE_AXFR) {
5357                                 verbose(VERB_ALGO, "xfr to %s failed, packet "
5358                                         "with wrong qtype, expected AXFR",
5359                                 xfr->task_transfer->master->host);
5360                                 return 0;
5361                         }
5362                 }
5363         }
5364
5365         /* check parse of RRs in packet, store first SOA serial
5366          * to be able to detect last SOA (with that serial) to see if done */
5367         /* also check for IXFR 'zone up to date' reply */
5368         for(i=0; i<(int)LDNS_ANCOUNT(wire); i++) {
5369                 size_t pos = sldns_buffer_position(pkt);
5370                 uint16_t tp, rdlen;
5371                 if(pkt_dname_len(pkt) == 0) {
5372                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5373                                 "malformed dname in answer section",
5374                                 xfr->task_transfer->master->host);
5375                         return 0;
5376                 }
5377                 if(sldns_buffer_remaining(pkt) < 10) {
5378                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5379                                 "truncated RR",
5380                                 xfr->task_transfer->master->host);
5381                         return 0;
5382                 }
5383                 tp = sldns_buffer_read_u16(pkt);
5384                 (void)sldns_buffer_read_u16(pkt); /* class */
5385                 (void)sldns_buffer_read_u32(pkt); /* ttl */
5386                 rdlen = sldns_buffer_read_u16(pkt);
5387                 if(sldns_buffer_remaining(pkt) < rdlen) {
5388                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5389                                 "truncated RR rdata",
5390                                 xfr->task_transfer->master->host);
5391                         return 0;
5392                 }
5393
5394                 /* RR parses (haven't checked rdata itself), now look at
5395                  * SOA records to see serial number */
5396                 if(xfr->task_transfer->rr_scan_num == 0 &&
5397                         tp != LDNS_RR_TYPE_SOA) {
5398                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5399                                 "malformed zone transfer, no start SOA",
5400                                 xfr->task_transfer->master->host);
5401                         return 0;
5402                 }
5403                 if(xfr->task_transfer->rr_scan_num == 1 &&
5404                         tp != LDNS_RR_TYPE_SOA) {
5405                         /* second RR is not a SOA record, this is not an IXFR
5406                          * the master is replying with an AXFR */
5407                         xfr->task_transfer->on_ixfr_is_axfr = 1;
5408                 }
5409                 if(tp == LDNS_RR_TYPE_SOA) {
5410                         uint32_t serial;
5411                         if(rdlen < 22) {
5412                                 verbose(VERB_ALGO, "xfr to %s failed, packet "
5413                                         "with SOA with malformed rdata",
5414                                         xfr->task_transfer->master->host);
5415                                 return 0;
5416                         }
5417                         if(dname_pkt_compare(pkt, sldns_buffer_at(pkt, pos),
5418                                 xfr->name) != 0) {
5419                                 verbose(VERB_ALGO, "xfr to %s failed, packet "
5420                                         "with SOA with wrong dname",
5421                                         xfr->task_transfer->master->host);
5422                                 return 0;
5423                         }
5424
5425                         /* read serial number of SOA */
5426                         serial = sldns_buffer_read_u32_at(pkt,
5427                                 sldns_buffer_position(pkt)+rdlen-20);
5428
5429                         /* check for IXFR 'zone has SOA x' reply */
5430                         if(xfr->task_transfer->on_ixfr &&
5431                                 xfr->task_transfer->rr_scan_num == 0 &&
5432                                 LDNS_ANCOUNT(wire)==1) {
5433                                 verbose(VERB_ALGO, "xfr to %s ended, "
5434                                         "IXFR reply that zone has serial %u",
5435                                         xfr->task_transfer->master->host,
5436                                         (unsigned)serial);
5437                                 return 0;
5438                         }
5439
5440                         /* if first SOA, store serial number */
5441                         if(xfr->task_transfer->got_xfr_serial == 0) {
5442                                 xfr->task_transfer->got_xfr_serial = 1;
5443                                 xfr->task_transfer->incoming_xfr_serial =
5444                                         serial;
5445                                 verbose(VERB_ALGO, "xfr %s: contains "
5446                                         "SOA serial %u",
5447                                         xfr->task_transfer->master->host,
5448                                         (unsigned)serial);
5449                         /* see if end of AXFR */
5450                         } else if(!xfr->task_transfer->on_ixfr ||
5451                                 xfr->task_transfer->on_ixfr_is_axfr) {
5452                                 /* second SOA with serial is the end
5453                                  * for AXFR */
5454                                 *transferdone = 1;
5455                                 verbose(VERB_ALGO, "xfr %s: last AXFR packet",
5456                                         xfr->task_transfer->master->host);
5457                         /* for IXFR, count SOA records with that serial */
5458                         } else if(xfr->task_transfer->incoming_xfr_serial ==
5459                                 serial && xfr->task_transfer->got_xfr_serial
5460                                 == 1) {
5461                                 xfr->task_transfer->got_xfr_serial++;
5462                         /* if not first soa, if serial==firstserial, the
5463                          * third time we are at the end, for IXFR */
5464                         } else if(xfr->task_transfer->incoming_xfr_serial ==
5465                                 serial && xfr->task_transfer->got_xfr_serial
5466                                 == 2) {
5467                                 verbose(VERB_ALGO, "xfr %s: last IXFR packet",
5468                                         xfr->task_transfer->master->host);
5469                                 *transferdone = 1;
5470                                 /* continue parse check, if that succeeds,
5471                                  * transfer is done */
5472                         }
5473                 }
5474                 xfr->task_transfer->rr_scan_num++;
5475
5476                 /* skip over RR rdata to go to the next RR */
5477                 sldns_buffer_skip(pkt, (ssize_t)rdlen);
5478         }
5479
5480         /* check authority section */
5481         /* we skip over the RRs checking packet format */
5482         for(i=0; i<(int)LDNS_NSCOUNT(wire); i++) {
5483                 uint16_t rdlen;
5484                 if(pkt_dname_len(pkt) == 0) {
5485                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5486                                 "malformed dname in authority section",
5487                                 xfr->task_transfer->master->host);
5488                         return 0;
5489                 }
5490                 if(sldns_buffer_remaining(pkt) < 10) {
5491                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5492                                 "truncated RR",
5493                                 xfr->task_transfer->master->host);
5494                         return 0;
5495                 }
5496                 (void)sldns_buffer_read_u16(pkt); /* type */
5497                 (void)sldns_buffer_read_u16(pkt); /* class */
5498                 (void)sldns_buffer_read_u32(pkt); /* ttl */
5499                 rdlen = sldns_buffer_read_u16(pkt);
5500                 if(sldns_buffer_remaining(pkt) < rdlen) {
5501                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5502                                 "truncated RR rdata",
5503                                 xfr->task_transfer->master->host);
5504                         return 0;
5505                 }
5506                 /* skip over RR rdata to go to the next RR */
5507                 sldns_buffer_skip(pkt, (ssize_t)rdlen);
5508         }
5509
5510         /* check additional section */
5511         for(i=0; i<(int)LDNS_ARCOUNT(wire); i++) {
5512                 uint16_t rdlen;
5513                 if(pkt_dname_len(pkt) == 0) {
5514                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5515                                 "malformed dname in additional section",
5516                                 xfr->task_transfer->master->host);
5517                         return 0;
5518                 }
5519                 if(sldns_buffer_remaining(pkt) < 10) {
5520                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5521                                 "truncated RR",
5522                                 xfr->task_transfer->master->host);
5523                         return 0;
5524                 }
5525                 (void)sldns_buffer_read_u16(pkt); /* type */
5526                 (void)sldns_buffer_read_u16(pkt); /* class */
5527                 (void)sldns_buffer_read_u32(pkt); /* ttl */
5528                 rdlen = sldns_buffer_read_u16(pkt);
5529                 if(sldns_buffer_remaining(pkt) < rdlen) {
5530                         verbose(VERB_ALGO, "xfr to %s failed, packet with "
5531                                 "truncated RR rdata",
5532                                 xfr->task_transfer->master->host);
5533                         return 0;
5534                 }
5535                 /* skip over RR rdata to go to the next RR */
5536                 sldns_buffer_skip(pkt, (ssize_t)rdlen);
5537         }
5538
5539         return 1;
5540 }
5541
5542 /** Link the data from this packet into the worklist of transferred data */
5543 static int
5544 xfer_link_data(sldns_buffer* pkt, struct auth_xfer* xfr)
5545 {
5546         /* alloc it */
5547         struct auth_chunk* e;
5548         e = (struct auth_chunk*)calloc(1, sizeof(*e));
5549         if(!e) return 0;
5550         e->next = NULL;
5551         e->len = sldns_buffer_limit(pkt);
5552         e->data = memdup(sldns_buffer_begin(pkt), e->len);
5553         if(!e->data) {
5554                 free(e);
5555                 return 0;
5556         }
5557
5558         /* alloc succeeded, link into list */
5559         if(!xfr->task_transfer->chunks_first)
5560                 xfr->task_transfer->chunks_first = e;
5561         if(xfr->task_transfer->chunks_last)
5562                 xfr->task_transfer->chunks_last->next = e;
5563         xfr->task_transfer->chunks_last = e;
5564         return 1;
5565 }
5566
5567 /** task transfer.  the list of data is complete. process it and if failed
5568  * move to next master, if succeeded, end the task transfer */
5569 static void
5570 process_list_end_transfer(struct auth_xfer* xfr, struct module_env* env)
5571 {
5572         int ixfr_fail = 0;
5573         if(xfr_process_chunk_list(xfr, env, &ixfr_fail)) {
5574                 /* it worked! */
5575                 auth_chunks_delete(xfr->task_transfer);
5576
5577                 /* we fetched the zone, move to wait task */
5578                 xfr_transfer_disown(xfr);
5579
5580                 if(xfr->notify_received && (!xfr->notify_has_serial ||
5581                         (xfr->notify_has_serial &&
5582                         xfr_serial_means_update(xfr, xfr->notify_serial)))) {
5583                         uint32_t sr = xfr->notify_serial;
5584                         int has_sr = xfr->notify_has_serial;
5585                         /* we received a notify while probe/transfer was
5586                          * in progress.  start a new probe and transfer */
5587                         xfr->notify_received = 0;
5588                         xfr->notify_has_serial = 0;
5589                         xfr->notify_serial = 0;
5590                         if(!xfr_start_probe(xfr, env, NULL)) {
5591                                 /* if we couldn't start it, already in
5592                                  * progress; restore notify serial,
5593                                  * while xfr still locked */
5594                                 xfr->notify_received = 1;
5595                                 xfr->notify_has_serial = has_sr;
5596                                 xfr->notify_serial = sr;
5597                                 lock_basic_unlock(&xfr->lock);
5598                         }
5599                         return;
5600                 } else {
5601                         /* pick up the nextprobe task and wait (normail wait time) */
5602                         if(xfr->task_nextprobe->worker == NULL)
5603                                 xfr_set_timeout(xfr, env, 0, 0);
5604                 }
5605                 lock_basic_unlock(&xfr->lock);
5606                 return;
5607         }
5608         /* processing failed */
5609         /* when done, delete data from list */
5610         auth_chunks_delete(xfr->task_transfer);
5611         if(ixfr_fail) {
5612                 xfr->task_transfer->ixfr_fail = 1;
5613         } else {
5614                 xfr_transfer_nextmaster(xfr);
5615         }
5616         xfr_transfer_nexttarget_or_end(xfr, env);
5617 }
5618
5619 /** callback for task_transfer tcp connections */
5620 int
5621 auth_xfer_transfer_tcp_callback(struct comm_point* c, void* arg, int err,
5622         struct comm_reply* ATTR_UNUSED(repinfo))
5623 {
5624         struct auth_xfer* xfr = (struct auth_xfer*)arg;
5625         struct module_env* env;
5626         int gonextonfail = 1;
5627         int transferdone = 0;
5628         log_assert(xfr->task_transfer);
5629         lock_basic_lock(&xfr->lock);
5630         env = xfr->task_transfer->env;
5631         if(env->outnet->want_to_quit) {
5632                 lock_basic_unlock(&xfr->lock);
5633                 return 0; /* stop on quit */
5634         }
5635
5636         if(err != NETEVENT_NOERROR) {
5637                 /* connection failed, closed, or timeout */
5638                 /* stop this transfer, cleanup
5639                  * and continue task_transfer*/
5640                 verbose(VERB_ALGO, "xfr stopped, connection lost to %s",
5641                         xfr->task_transfer->master->host);
5642
5643                 /* see if IXFR caused the failure, if so, try AXFR */
5644                 if(xfr->task_transfer->on_ixfr) {
5645                         xfr->task_transfer->ixfr_possible_timeout_count++;
5646                         if(xfr->task_transfer->ixfr_possible_timeout_count >=
5647                                 NUM_TIMEOUTS_FALLBACK_IXFR) {
5648                                 verbose(VERB_ALGO, "xfr to %s, fallback "
5649                                         "from IXFR to AXFR (because of timeouts)",
5650                                         xfr->task_transfer->master->host);
5651                                 xfr->task_transfer->ixfr_fail = 1;
5652                                 gonextonfail = 0;
5653                         }
5654                 }
5655
5656         failed:
5657                 /* delete transferred data from list */
5658                 auth_chunks_delete(xfr->task_transfer);
5659                 comm_point_delete(xfr->task_transfer->cp);
5660                 xfr->task_transfer->cp = NULL;
5661                 if(gonextonfail)
5662                         xfr_transfer_nextmaster(xfr);
5663                 xfr_transfer_nexttarget_or_end(xfr, env);
5664                 return 0;
5665         }
5666         /* note that IXFR worked without timeout */
5667         if(xfr->task_transfer->on_ixfr)
5668                 xfr->task_transfer->ixfr_possible_timeout_count = 0;
5669
5670         /* handle returned packet */
5671         /* if it fails, cleanup and end this transfer */
5672         /* if it needs to fallback from IXFR to AXFR, do that */
5673         if(!check_xfer_packet(c->buffer, xfr, &gonextonfail, &transferdone)) {
5674                 goto failed;
5675         }
5676         /* if it is good, link it into the list of data */
5677         /* if the link into list of data fails (malloc fail) cleanup and end */
5678         if(!xfer_link_data(c->buffer, xfr)) {
5679                 verbose(VERB_ALGO, "xfr stopped to %s, malloc failed",
5680                         xfr->task_transfer->master->host);
5681                 goto failed;
5682         }
5683         /* if the transfer is done now, disconnect and process the list */
5684         if(transferdone) {
5685                 comm_point_delete(xfr->task_transfer->cp);
5686                 xfr->task_transfer->cp = NULL;
5687                 process_list_end_transfer(xfr, env);
5688                 return 0;
5689         }
5690
5691         /* if we want to read more messages, setup the commpoint to read
5692          * a DNS packet, and the timeout */
5693         lock_basic_unlock(&xfr->lock);
5694         c->tcp_is_reading = 1;
5695         sldns_buffer_clear(c->buffer);
5696         comm_point_start_listening(c, -1, AUTH_TRANSFER_TIMEOUT);
5697         return 0;
5698 }
5699
5700 /** callback for task_transfer http connections */
5701 int
5702 auth_xfer_transfer_http_callback(struct comm_point* c, void* arg, int err,
5703         struct comm_reply* repinfo)
5704 {
5705         struct auth_xfer* xfr = (struct auth_xfer*)arg;
5706         struct module_env* env;
5707         log_assert(xfr->task_transfer);
5708         lock_basic_lock(&xfr->lock);
5709         env = xfr->task_transfer->env;
5710         if(env->outnet->want_to_quit) {
5711                 lock_basic_unlock(&xfr->lock);
5712                 return 0; /* stop on quit */
5713         }
5714         verbose(VERB_ALGO, "auth zone transfer http callback");
5715
5716         if(err != NETEVENT_NOERROR && err != NETEVENT_DONE) {
5717                 /* connection failed, closed, or timeout */
5718                 /* stop this transfer, cleanup
5719                  * and continue task_transfer*/
5720                 verbose(VERB_ALGO, "http stopped, connection lost to %s",
5721                         xfr->task_transfer->master->host);
5722         failed:
5723                 /* delete transferred data from list */
5724                 auth_chunks_delete(xfr->task_transfer);
5725                 if(repinfo) repinfo->c = NULL; /* signal cp deleted to
5726                                 the routine calling this callback */
5727                 comm_point_delete(xfr->task_transfer->cp);
5728                 xfr->task_transfer->cp = NULL;
5729                 xfr_transfer_nextmaster(xfr);
5730                 xfr_transfer_nexttarget_or_end(xfr, env);
5731                 return 0;
5732         }
5733
5734         /* if it is good, link it into the list of data */
5735         /* if the link into list of data fails (malloc fail) cleanup and end */
5736         if(sldns_buffer_limit(c->buffer) > 0) {
5737                 verbose(VERB_ALGO, "auth zone http queued up %d bytes",
5738                         (int)sldns_buffer_limit(c->buffer));
5739                 if(!xfer_link_data(c->buffer, xfr)) {
5740                         verbose(VERB_ALGO, "http stopped to %s, malloc failed",
5741                                 xfr->task_transfer->master->host);
5742                         goto failed;
5743                 }
5744         }
5745         /* if the transfer is done now, disconnect and process the list */
5746         if(err == NETEVENT_DONE) {
5747                 if(repinfo) repinfo->c = NULL; /* signal cp deleted to
5748                                 the routine calling this callback */
5749                 comm_point_delete(xfr->task_transfer->cp);
5750                 xfr->task_transfer->cp = NULL;
5751                 process_list_end_transfer(xfr, env);
5752                 return 0;
5753         }
5754
5755         /* if we want to read more messages, setup the commpoint to read
5756          * a DNS packet, and the timeout */
5757         lock_basic_unlock(&xfr->lock);
5758         c->tcp_is_reading = 1;
5759         sldns_buffer_clear(c->buffer);
5760         comm_point_start_listening(c, -1, AUTH_TRANSFER_TIMEOUT);
5761         return 0;
5762 }
5763
5764
5765 /** start transfer task by this worker , xfr is locked. */
5766 static void
5767 xfr_start_transfer(struct auth_xfer* xfr, struct module_env* env,
5768         struct auth_master* master)
5769 {
5770         log_assert(xfr->task_transfer != NULL);
5771         log_assert(xfr->task_transfer->worker == NULL);
5772         log_assert(xfr->task_transfer->chunks_first == NULL);
5773         log_assert(xfr->task_transfer->chunks_last == NULL);
5774         xfr->task_transfer->worker = env->worker;
5775         xfr->task_transfer->env = env;
5776
5777         /* init transfer process */
5778         /* find that master in the transfer's list of masters? */
5779         xfr_transfer_start_list(xfr, master);
5780         /* start lookup for hostnames in transfer master list */
5781         xfr_transfer_start_lookups(xfr);
5782
5783         /* initiate TCP, and set timeout on it */
5784         xfr_transfer_nexttarget_or_end(xfr, env);
5785 }
5786
5787 /** disown task_probe.  caller must hold xfr.lock */
5788 static void
5789 xfr_probe_disown(struct auth_xfer* xfr)
5790 {
5791         /* remove timer (from this worker's event base) */
5792         comm_timer_delete(xfr->task_probe->timer);
5793         xfr->task_probe->timer = NULL;
5794         /* remove the commpoint */
5795         comm_point_delete(xfr->task_probe->cp);
5796         xfr->task_probe->cp = NULL;
5797         /* we don't own this item anymore */
5798         xfr->task_probe->worker = NULL;
5799         xfr->task_probe->env = NULL;
5800 }
5801
5802 /** send the UDP probe to the master, this is part of task_probe */
5803 static int
5804 xfr_probe_send_probe(struct auth_xfer* xfr, struct module_env* env,
5805         int timeout)
5806 {
5807         struct sockaddr_storage addr;
5808         socklen_t addrlen = 0;
5809         struct timeval t;
5810         /* pick master */
5811         struct auth_master* master = xfr_probe_current_master(xfr);
5812         if(!master) return 0;
5813         if(master->allow_notify) return 0; /* only for notify */
5814         if(master->http) return 0; /* only masters get SOA UDP probe,
5815                 not urls, if those are in this list */
5816
5817         /* get master addr */
5818         if(xfr->task_probe->scan_addr) {
5819                 addrlen = xfr->task_probe->scan_addr->addrlen;
5820                 memmove(&addr, &xfr->task_probe->scan_addr->addr, addrlen);
5821         } else {
5822                 if(!extstrtoaddr(master->host, &addr, &addrlen)) {
5823                         /* the ones that are not in addr format are supposed
5824                          * to be looked up.  The lookup has failed however,
5825                          * so skip them */
5826                         char zname[255+1];
5827                         dname_str(xfr->name, zname);
5828                         log_err("%s: failed lookup, cannot probe to master %s",
5829                                 zname, master->host);
5830                         return 0;
5831                 }
5832         }
5833
5834         /* create packet */
5835         /* create new ID for new probes, but not on timeout retries,
5836          * this means we'll accept replies to previous retries to same ip */
5837         if(timeout == AUTH_PROBE_TIMEOUT)
5838                 xfr->task_probe->id = (uint16_t)(ub_random(env->rnd)&0xffff);
5839         xfr_create_soa_probe_packet(xfr, env->scratch_buffer,
5840                 xfr->task_probe->id);
5841         if(!xfr->task_probe->cp) {
5842                 xfr->task_probe->cp = outnet_comm_point_for_udp(env->outnet,
5843                         auth_xfer_probe_udp_callback, xfr, &addr, addrlen);
5844                 if(!xfr->task_probe->cp) {
5845                         char zname[255+1];
5846                         dname_str(xfr->name, zname);
5847                         verbose(VERB_ALGO, "cannot create udp cp for "
5848                                 "probe %s to %s", zname, master->host);
5849                         return 0;
5850                 }
5851         }
5852         if(!xfr->task_probe->timer) {
5853                 xfr->task_probe->timer = comm_timer_create(env->worker_base,
5854                         auth_xfer_probe_timer_callback, xfr);
5855                 if(!xfr->task_probe->timer) {
5856                         log_err("malloc failure");
5857                         return 0;
5858                 }
5859         }
5860
5861         /* send udp packet */
5862         if(!comm_point_send_udp_msg(xfr->task_probe->cp, env->scratch_buffer,
5863                 (struct sockaddr*)&addr, addrlen)) {
5864                 char zname[255+1];
5865                 dname_str(xfr->name, zname);
5866                 verbose(VERB_ALGO, "failed to send soa probe for %s to %s",
5867                         zname, master->host);
5868                 return 0;
5869         }
5870         xfr->task_probe->timeout = timeout;
5871 #ifndef S_SPLINT_S
5872         t.tv_sec = timeout/1000;
5873         t.tv_usec = (timeout%1000)*1000;
5874 #endif
5875         comm_timer_set(xfr->task_probe->timer, &t);
5876
5877         return 1;
5878 }
5879
5880 /** callback for task_probe timer */
5881 void
5882 auth_xfer_probe_timer_callback(void* arg)
5883 {
5884         struct auth_xfer* xfr = (struct auth_xfer*)arg;
5885         struct module_env* env;
5886         log_assert(xfr->task_probe);
5887         lock_basic_lock(&xfr->lock);
5888         env = xfr->task_probe->env;
5889         if(env->outnet->want_to_quit) {
5890                 lock_basic_unlock(&xfr->lock);
5891                 return; /* stop on quit */
5892         }
5893
5894         if(xfr->task_probe->timeout <= AUTH_PROBE_TIMEOUT_STOP) {
5895                 /* try again with bigger timeout */
5896                 if(xfr_probe_send_probe(xfr, env, xfr->task_probe->timeout*2)) {
5897                         lock_basic_unlock(&xfr->lock);
5898                         return;
5899                 }
5900         }
5901         /* delete commpoint so a new one is created, with a fresh port nr */
5902         comm_point_delete(xfr->task_probe->cp);
5903         xfr->task_probe->cp = NULL;
5904
5905         /* too many timeouts (or fail to send), move to next or end */
5906         xfr_probe_nextmaster(xfr);
5907         xfr_probe_send_or_end(xfr, env);
5908 }
5909
5910 /** callback for task_probe udp packets */
5911 int
5912 auth_xfer_probe_udp_callback(struct comm_point* c, void* arg, int err,
5913         struct comm_reply* repinfo)
5914 {
5915         struct auth_xfer* xfr = (struct auth_xfer*)arg;
5916         struct module_env* env;
5917         log_assert(xfr->task_probe);
5918         lock_basic_lock(&xfr->lock);
5919         env = xfr->task_probe->env;
5920         if(env->outnet->want_to_quit) {
5921                 lock_basic_unlock(&xfr->lock);
5922                 return 0; /* stop on quit */
5923         }
5924
5925         /* the comm_point_udp_callback is in a for loop for NUM_UDP_PER_SELECT
5926          * and we set rep.c=NULL to stop if from looking inside the commpoint*/
5927         repinfo->c = NULL;
5928         /* stop the timer */
5929         comm_timer_disable(xfr->task_probe->timer);
5930
5931         /* see if we got a packet and what that means */
5932         if(err == NETEVENT_NOERROR) {
5933                 uint32_t serial = 0;
5934                 if(check_packet_ok(c->buffer, LDNS_RR_TYPE_SOA, xfr,
5935                         &serial)) {
5936                         /* successful lookup */
5937                         if(verbosity >= VERB_ALGO) {
5938                                 char buf[256];
5939                                 dname_str(xfr->name, buf);
5940                                 verbose(VERB_ALGO, "auth zone %s: soa probe "
5941                                         "serial is %u", buf, (unsigned)serial);
5942                         }
5943                         /* see if this serial indicates that the zone has
5944                          * to be updated */
5945                         if(xfr_serial_means_update(xfr, serial)) {
5946                                 /* if updated, start the transfer task, if needed */
5947                                 verbose(VERB_ALGO, "auth_zone updated, start transfer");
5948                                 if(xfr->task_transfer->worker == NULL) {
5949                                         struct auth_master* master =
5950                                                 xfr_probe_current_master(xfr);
5951                                         /* if we have download URLs use them
5952                                          * in preference to this master we
5953                                          * just probed the SOA from */
5954                                         if(xfr->task_transfer->masters &&
5955                                                 xfr->task_transfer->masters->http)
5956                                                 master = NULL;
5957                                         xfr_probe_disown(xfr);
5958                                         xfr_start_transfer(xfr, env, master);
5959                                         return 0;
5960
5961                                 }
5962                                 /* other tasks are running, we don't do this anymore */
5963                                 xfr_probe_disown(xfr);
5964                                 lock_basic_unlock(&xfr->lock);
5965                                 /* return, we don't sent a reply to this udp packet,
5966                                  * and we setup the tasks to do next */
5967                                 return 0;
5968                         } else {
5969                                 verbose(VERB_ALGO, "auth_zone master reports unchanged soa serial");
5970                                 /* we if cannot find updates amongst the
5971                                  * masters, this means we then have a new lease
5972                                  * on the zone */
5973                                 xfr->task_probe->have_new_lease = 1;
5974                         }
5975                 } else {
5976                         if(verbosity >= VERB_ALGO) {
5977                                 char buf[256];
5978                                 dname_str(xfr->name, buf);
5979                                 verbose(VERB_ALGO, "auth zone %s: bad reply to soa probe", buf);
5980                         }
5981                 }
5982         } else {
5983                 if(verbosity >= VERB_ALGO) {
5984                         char buf[256];
5985                         dname_str(xfr->name, buf);
5986                         verbose(VERB_ALGO, "auth zone %s: soa probe failed", buf);
5987                 }
5988         }
5989
5990         /* failed lookup or not an update */
5991         /* delete commpoint so a new one is created, with a fresh port nr */
5992         comm_point_delete(xfr->task_probe->cp);
5993         xfr->task_probe->cp = NULL;
5994
5995         /* if the result was not a successfull probe, we need
5996          * to send the next one */
5997         xfr_probe_nextmaster(xfr);
5998         xfr_probe_send_or_end(xfr, env);
5999         return 0;
6000 }
6001
6002 /** lookup a host name for its addresses, if needed */
6003 static int
6004 xfr_probe_lookup_host(struct auth_xfer* xfr, struct module_env* env)
6005 {
6006         struct sockaddr_storage addr;
6007         socklen_t addrlen = 0;
6008         struct auth_master* master = xfr->task_probe->lookup_target;
6009         struct query_info qinfo;
6010         uint16_t qflags = BIT_RD;
6011         uint8_t dname[LDNS_MAX_DOMAINLEN+1];
6012         struct edns_data edns;
6013         sldns_buffer* buf = env->scratch_buffer;
6014         if(!master) return 0;
6015         if(extstrtoaddr(master->host, &addr, &addrlen)) {
6016                 /* not needed, host is in IP addr format */
6017                 return 0;
6018         }
6019         if(master->allow_notify && !master->http &&
6020                 strchr(master->host, '/') != NULL &&
6021                 strchr(master->host, '/') == strrchr(master->host, '/')) {
6022                 return 0; /* is IP/prefix format, not something to look up */
6023         }
6024
6025         /* use mesh_new_callback to probe for non-addr hosts,
6026          * and then wait for them to be looked up (in cache, or query) */
6027         qinfo.qname_len = sizeof(dname);
6028         if(sldns_str2wire_dname_buf(master->host, dname, &qinfo.qname_len)
6029                 != 0) {
6030                 log_err("cannot parse host name of master %s", master->host);
6031                 return 0;
6032         }
6033         qinfo.qname = dname;
6034         qinfo.qclass = xfr->dclass;
6035         qinfo.qtype = LDNS_RR_TYPE_A;
6036         if(xfr->task_probe->lookup_aaaa)
6037                 qinfo.qtype = LDNS_RR_TYPE_AAAA;
6038         qinfo.local_alias = NULL;
6039         if(verbosity >= VERB_ALGO) {
6040                 char buf1[512];
6041                 char buf2[LDNS_MAX_DOMAINLEN+1];
6042                 dname_str(xfr->name, buf2);
6043                 snprintf(buf1, sizeof(buf1), "auth zone %s: master lookup"
6044                         " for task_probe", buf2);
6045                 log_query_info(VERB_ALGO, buf1, &qinfo);
6046         }
6047         edns.edns_present = 1;
6048         edns.ext_rcode = 0;
6049         edns.edns_version = 0;
6050         edns.bits = EDNS_DO;
6051         edns.opt_list = NULL;
6052         if(sldns_buffer_capacity(buf) < 65535)
6053                 edns.udp_size = (uint16_t)sldns_buffer_capacity(buf);
6054         else    edns.udp_size = 65535;
6055
6056         /* unlock xfr during mesh_new_callback() because the callback can be
6057          * called straight away */
6058         lock_basic_unlock(&xfr->lock);
6059         if(!mesh_new_callback(env->mesh, &qinfo, qflags, &edns, buf, 0,
6060                 &auth_xfer_probe_lookup_callback, xfr)) {
6061                 lock_basic_lock(&xfr->lock);
6062                 log_err("out of memory lookup up master %s", master->host);
6063                 return 0;
6064         }
6065         lock_basic_lock(&xfr->lock);
6066         return 1;
6067 }
6068
6069 /** move to sending the probe packets, next if fails. task_probe */
6070 static void
6071 xfr_probe_send_or_end(struct auth_xfer* xfr, struct module_env* env)
6072 {
6073         /* are we doing hostname lookups? */
6074         while(xfr->task_probe->lookup_target) {
6075                 if(xfr_probe_lookup_host(xfr, env)) {
6076                         /* wait for lookup to finish,
6077                          * note that the hostname may be in unbound's cache
6078                          * and we may then get an instant cache response,
6079                          * and that calls the callback just like a full
6080                          * lookup and lookup failures also call callback */
6081                         lock_basic_unlock(&xfr->lock);
6082                         return;
6083                 }
6084                 xfr_probe_move_to_next_lookup(xfr, env);
6085         }
6086         /* probe of list has ended.  Create or refresh the list of of
6087          * allow_notify addrs */
6088         probe_copy_masters_for_allow_notify(xfr);
6089         if(xfr->task_probe->only_lookup) {
6090                 /* only wanted lookups for copy, stop probe and start wait */
6091                 xfr->task_probe->only_lookup = 0;
6092                 xfr_probe_disown(xfr);
6093                 if(xfr->task_nextprobe->worker == NULL)
6094                         xfr_set_timeout(xfr, env, 0, 0);
6095                 lock_basic_unlock(&xfr->lock);
6096                 return;
6097         }
6098
6099         /* send probe packets */
6100         while(!xfr_probe_end_of_list(xfr)) {
6101                 if(xfr_probe_send_probe(xfr, env, AUTH_PROBE_TIMEOUT)) {
6102                         /* successfully sent probe, wait for callback */
6103                         lock_basic_unlock(&xfr->lock);
6104                         return;
6105                 }
6106                 /* failed to send probe, next master */
6107                 xfr_probe_nextmaster(xfr);
6108         }
6109
6110         /* done with probe sequence, wait */
6111         if(xfr->task_probe->have_new_lease) {
6112                 /* if zone not updated, start the wait timer again */
6113                 verbose(VERB_ALGO, "auth_zone unchanged, new lease, wait");
6114                 xfr_probe_disown(xfr);
6115                 if(xfr->have_zone)
6116                         xfr->lease_time = *env->now;
6117                 if(xfr->task_nextprobe->worker == NULL)
6118                         xfr_set_timeout(xfr, env, 0, 0);
6119         } else {
6120                 /* we failed to send this as well, move to the wait task,
6121                  * use the shorter retry timeout */
6122                 xfr_probe_disown(xfr);
6123                 /* pick up the nextprobe task and wait */
6124                 if(xfr->task_nextprobe->worker == NULL)
6125                         xfr_set_timeout(xfr, env, 1, 0);
6126         }
6127
6128         lock_basic_unlock(&xfr->lock);
6129 }
6130
6131 /** callback for task_probe lookup of host name, of A or AAAA */
6132 void auth_xfer_probe_lookup_callback(void* arg, int rcode, sldns_buffer* buf,
6133         enum sec_status ATTR_UNUSED(sec), char* ATTR_UNUSED(why_bogus),
6134         int ATTR_UNUSED(was_ratelimited))
6135 {
6136         struct auth_xfer* xfr = (struct auth_xfer*)arg;
6137         struct module_env* env;
6138         log_assert(xfr->task_probe);
6139         lock_basic_lock(&xfr->lock);
6140         env = xfr->task_probe->env;
6141         if(env->outnet->want_to_quit) {
6142                 lock_basic_unlock(&xfr->lock);
6143                 return; /* stop on quit */
6144         }
6145
6146         /* process result */
6147         if(rcode == LDNS_RCODE_NOERROR) {
6148                 uint16_t wanted_qtype = LDNS_RR_TYPE_A;
6149                 struct regional* temp = env->scratch;
6150                 struct query_info rq;
6151                 struct reply_info* rep;
6152                 if(xfr->task_probe->lookup_aaaa)
6153                         wanted_qtype = LDNS_RR_TYPE_AAAA;
6154                 memset(&rq, 0, sizeof(rq));
6155                 rep = parse_reply_in_temp_region(buf, temp, &rq);
6156                 if(rep && rq.qtype == wanted_qtype &&
6157                         FLAGS_GET_RCODE(rep->flags) == LDNS_RCODE_NOERROR) {
6158                         /* parsed successfully */
6159                         struct ub_packed_rrset_key* answer =
6160                                 reply_find_answer_rrset(&rq, rep);
6161                         if(answer) {
6162                                 xfr_master_add_addrs(xfr->task_probe->
6163                                         lookup_target, answer, wanted_qtype);
6164                         }
6165                 }
6166         }
6167         if(xfr->task_probe->lookup_target->list &&
6168                 xfr->task_probe->lookup_target == xfr_probe_current_master(xfr))
6169                 xfr->task_probe->scan_addr = xfr->task_probe->lookup_target->list;
6170
6171         /* move to lookup AAAA after A lookup, move to next hostname lookup,
6172          * or move to send the probes, or, if nothing to do, end task_probe */
6173         xfr_probe_move_to_next_lookup(xfr, env);
6174         xfr_probe_send_or_end(xfr, env);
6175 }
6176
6177 /** disown task_nextprobe.  caller must hold xfr.lock */
6178 static void
6179 xfr_nextprobe_disown(struct auth_xfer* xfr)
6180 {
6181         /* delete the timer, because the next worker to pick this up may
6182          * not have the same event base */
6183         comm_timer_delete(xfr->task_nextprobe->timer);
6184         xfr->task_nextprobe->timer = NULL;
6185         xfr->task_nextprobe->next_probe = 0;
6186         /* we don't own this item anymore */
6187         xfr->task_nextprobe->worker = NULL;
6188         xfr->task_nextprobe->env = NULL;
6189 }
6190
6191 /** xfer nextprobe timeout callback, this is part of task_nextprobe */
6192 void
6193 auth_xfer_timer(void* arg)
6194 {
6195         struct auth_xfer* xfr = (struct auth_xfer*)arg;
6196         struct module_env* env;
6197         log_assert(xfr->task_nextprobe);
6198         lock_basic_lock(&xfr->lock);
6199         env = xfr->task_nextprobe->env;
6200         if(env->outnet->want_to_quit) {
6201                 lock_basic_unlock(&xfr->lock);
6202                 return; /* stop on quit */
6203         }
6204
6205         /* see if zone has expired, and if so, also set auth_zone expired */
6206         if(xfr->have_zone && !xfr->zone_expired &&
6207            *env->now >= xfr->lease_time + xfr->expiry) {
6208                 lock_basic_unlock(&xfr->lock);
6209                 auth_xfer_set_expired(xfr, env, 1);
6210                 lock_basic_lock(&xfr->lock);
6211         }
6212
6213         xfr_nextprobe_disown(xfr);
6214
6215         if(!xfr_start_probe(xfr, env, NULL)) {
6216                 /* not started because already in progress */
6217                 lock_basic_unlock(&xfr->lock);
6218         }
6219 }
6220
6221 /** return true if there are probe (SOA UDP query) targets in the master list*/
6222 static int
6223 have_probe_targets(struct auth_master* list)
6224 {
6225         struct auth_master* p;
6226         for(p=list; p; p = p->next) {
6227                 if(!p->allow_notify && p->host)
6228                         return 1;
6229         }
6230         return 0;
6231 }
6232
6233 /** start task_probe if possible, if no masters for probe start task_transfer
6234  * returns true if task has been started, and false if the task is already
6235  * in progress. */
6236 static int
6237 xfr_start_probe(struct auth_xfer* xfr, struct module_env* env,
6238         struct auth_master* spec)
6239 {
6240         /* see if we need to start a probe (or maybe it is already in
6241          * progress (due to notify)) */
6242         if(xfr->task_probe->worker == NULL) {
6243                 if(!have_probe_targets(xfr->task_probe->masters) &&
6244                         !(xfr->task_probe->only_lookup &&
6245                         xfr->task_probe->masters != NULL)) {
6246                         /* useless to pick up task_probe, no masters to
6247                          * probe. Instead attempt to pick up task transfer */
6248                         if(xfr->task_transfer->worker == NULL) {
6249                                 xfr_start_transfer(xfr, env, spec);
6250                                 return 1;
6251                         }
6252                         /* task transfer already in progress */
6253                         return 0;
6254                 }
6255
6256                 /* pick up the probe task ourselves */
6257                 xfr->task_probe->worker = env->worker;
6258                 xfr->task_probe->env = env;
6259                 xfr->task_probe->cp = NULL;
6260
6261                 /* start the task */
6262                 /* have not seen a new lease yet, this scan */
6263                 xfr->task_probe->have_new_lease = 0;
6264                 /* if this was a timeout, no specific first master to scan */
6265                 /* otherwise, spec is nonNULL the notified master, scan
6266                  * first and also transfer first from it */
6267                 xfr_probe_start_list(xfr, spec);
6268                 /* setup to start the lookup of hostnames of masters afresh */
6269                 xfr_probe_start_lookups(xfr);
6270                 /* send the probe packet or next send, or end task */
6271                 xfr_probe_send_or_end(xfr, env);
6272                 return 1;
6273         }
6274         return 0;
6275 }
6276
6277 /** for task_nextprobe.
6278  * determine next timeout for auth_xfer. Also (re)sets timer.
6279  * @param xfr: task structure
6280  * @param env: module environment, with worker and time.
6281  * @param failure: set true if timer should be set for failure retry.
6282  * @param lookup_only: only perform lookups when timer done, 0 sec timeout
6283  */
6284 static void
6285 xfr_set_timeout(struct auth_xfer* xfr, struct module_env* env,
6286         int failure, int lookup_only)
6287 {
6288         struct timeval tv;
6289         log_assert(xfr->task_nextprobe != NULL);
6290         log_assert(xfr->task_nextprobe->worker == NULL ||
6291                 xfr->task_nextprobe->worker == env->worker);
6292         /* normally, nextprobe = startoflease + refresh,
6293          * but if expiry is sooner, use that one.
6294          * after a failure, use the retry timer instead. */
6295         xfr->task_nextprobe->next_probe = *env->now;
6296         if(xfr->lease_time && !failure)
6297                 xfr->task_nextprobe->next_probe = xfr->lease_time;
6298
6299         if(!failure) {
6300                 xfr->task_nextprobe->backoff = 0;
6301         } else {
6302                 if(xfr->task_nextprobe->backoff == 0)
6303                                 xfr->task_nextprobe->backoff = 3;
6304                 else    xfr->task_nextprobe->backoff *= 2;
6305                 if(xfr->task_nextprobe->backoff > AUTH_TRANSFER_MAX_BACKOFF)
6306                         xfr->task_nextprobe->backoff =
6307                                 AUTH_TRANSFER_MAX_BACKOFF;
6308         }
6309
6310         if(xfr->have_zone) {
6311                 time_t wait = xfr->refresh;
6312                 if(failure) wait = xfr->retry;
6313                 if(xfr->expiry < wait)
6314                         xfr->task_nextprobe->next_probe += xfr->expiry;
6315                 else    xfr->task_nextprobe->next_probe += wait;
6316                 if(failure)
6317                         xfr->task_nextprobe->next_probe +=
6318                                 xfr->task_nextprobe->backoff;
6319                 /* put the timer exactly on expiry, if possible */
6320                 if(xfr->lease_time && xfr->lease_time+xfr->expiry <
6321                         xfr->task_nextprobe->next_probe &&
6322                         xfr->lease_time+xfr->expiry > *env->now)
6323                         xfr->task_nextprobe->next_probe =
6324                                 xfr->lease_time+xfr->expiry;
6325         } else {
6326                 xfr->task_nextprobe->next_probe +=
6327                         xfr->task_nextprobe->backoff;
6328         }
6329
6330         if(!xfr->task_nextprobe->timer) {
6331                 xfr->task_nextprobe->timer = comm_timer_create(
6332                         env->worker_base, auth_xfer_timer, xfr);
6333                 if(!xfr->task_nextprobe->timer) {
6334                         /* failed to malloc memory. likely zone transfer
6335                          * also fails for that. skip the timeout */
6336                         char zname[255+1];
6337                         dname_str(xfr->name, zname);
6338                         log_err("cannot allocate timer, no refresh for %s",
6339                                 zname);
6340                         return;
6341                 }
6342         }
6343         xfr->task_nextprobe->worker = env->worker;
6344         xfr->task_nextprobe->env = env;
6345         if(*(xfr->task_nextprobe->env->now) <= xfr->task_nextprobe->next_probe)
6346                 tv.tv_sec = xfr->task_nextprobe->next_probe -
6347                         *(xfr->task_nextprobe->env->now);
6348         else    tv.tv_sec = 0;
6349         if(tv.tv_sec != 0 && lookup_only && xfr->task_probe->masters) {
6350                 /* don't lookup_only, if lookup timeout is 0 anyway,
6351                  * or if we don't have masters to lookup */
6352                 tv.tv_sec = 0;
6353                 if(xfr->task_probe && xfr->task_probe->worker == NULL)
6354                         xfr->task_probe->only_lookup = 1;
6355         }
6356         if(verbosity >= VERB_ALGO) {
6357                 char zname[255+1];
6358                 dname_str(xfr->name, zname);
6359                 verbose(VERB_ALGO, "auth zone %s timeout in %d seconds",
6360                         zname, (int)tv.tv_sec);
6361         }
6362         tv.tv_usec = 0;
6363         comm_timer_set(xfr->task_nextprobe->timer, &tv);
6364 }
6365
6366 /** initial pick up of worker timeouts, ties events to worker event loop */
6367 void
6368 auth_xfer_pickup_initial(struct auth_zones* az, struct module_env* env)
6369 {
6370         struct auth_xfer* x;
6371         lock_rw_wrlock(&az->lock);
6372         RBTREE_FOR(x, struct auth_xfer*, &az->xtree) {
6373                 lock_basic_lock(&x->lock);
6374                 /* set lease_time, because we now have timestamp in env,
6375                  * (not earlier during startup and apply_cfg), and this
6376                  * notes the start time when the data was acquired */
6377                 if(x->have_zone)
6378                         x->lease_time = *env->now;
6379                 if(x->task_nextprobe && x->task_nextprobe->worker == NULL) {
6380                         xfr_set_timeout(x, env, 0, 1);
6381                 }
6382                 lock_basic_unlock(&x->lock);
6383         }
6384         lock_rw_unlock(&az->lock);
6385 }
6386
6387 void auth_zones_cleanup(struct auth_zones* az)
6388 {
6389         struct auth_xfer* x;
6390         lock_rw_wrlock(&az->lock);
6391         RBTREE_FOR(x, struct auth_xfer*, &az->xtree) {
6392                 lock_basic_lock(&x->lock);
6393                 if(x->task_nextprobe && x->task_nextprobe->worker != NULL) {
6394                         xfr_nextprobe_disown(x);
6395                 }
6396                 if(x->task_probe && x->task_probe->worker != NULL) {
6397                         xfr_probe_disown(x);
6398                 }
6399                 if(x->task_transfer && x->task_transfer->worker != NULL) {
6400                         auth_chunks_delete(x->task_transfer);
6401                         xfr_transfer_disown(x);
6402                 }
6403                 lock_basic_unlock(&x->lock);
6404         }
6405         lock_rw_unlock(&az->lock);
6406 }
6407
6408 /**
6409  * malloc the xfer and tasks
6410  * @param z: auth_zone with name of zone.
6411  */
6412 static struct auth_xfer*
6413 auth_xfer_new(struct auth_zone* z)
6414 {
6415         struct auth_xfer* xfr;
6416         xfr = (struct auth_xfer*)calloc(1, sizeof(*xfr));
6417         if(!xfr) return NULL;
6418         xfr->name = memdup(z->name, z->namelen);
6419         if(!xfr->name) {
6420                 free(xfr);
6421                 return NULL;
6422         }
6423         xfr->node.key = xfr;
6424         xfr->namelen = z->namelen;
6425         xfr->namelabs = z->namelabs;
6426         xfr->dclass = z->dclass;
6427
6428         xfr->task_nextprobe = (struct auth_nextprobe*)calloc(1,
6429                 sizeof(struct auth_nextprobe));
6430         if(!xfr->task_nextprobe) {
6431                 free(xfr->name);
6432                 free(xfr);
6433                 return NULL;
6434         }
6435         xfr->task_probe = (struct auth_probe*)calloc(1,
6436                 sizeof(struct auth_probe));
6437         if(!xfr->task_probe) {
6438                 free(xfr->task_nextprobe);
6439                 free(xfr->name);
6440                 free(xfr);
6441                 return NULL;
6442         }
6443         xfr->task_transfer = (struct auth_transfer*)calloc(1,
6444                 sizeof(struct auth_transfer));
6445         if(!xfr->task_transfer) {
6446                 free(xfr->task_probe);
6447                 free(xfr->task_nextprobe);
6448                 free(xfr->name);
6449                 free(xfr);
6450                 return NULL;
6451         }
6452
6453         lock_basic_init(&xfr->lock);
6454         lock_protect(&xfr->lock, &xfr->name, sizeof(xfr->name));
6455         lock_protect(&xfr->lock, &xfr->namelen, sizeof(xfr->namelen));
6456         lock_protect(&xfr->lock, xfr->name, xfr->namelen);
6457         lock_protect(&xfr->lock, &xfr->namelabs, sizeof(xfr->namelabs));
6458         lock_protect(&xfr->lock, &xfr->dclass, sizeof(xfr->dclass));
6459         lock_protect(&xfr->lock, &xfr->notify_received, sizeof(xfr->notify_received));
6460         lock_protect(&xfr->lock, &xfr->notify_serial, sizeof(xfr->notify_serial));
6461         lock_protect(&xfr->lock, &xfr->zone_expired, sizeof(xfr->zone_expired));
6462         lock_protect(&xfr->lock, &xfr->have_zone, sizeof(xfr->have_zone));
6463         lock_protect(&xfr->lock, &xfr->serial, sizeof(xfr->serial));
6464         lock_protect(&xfr->lock, &xfr->retry, sizeof(xfr->retry));
6465         lock_protect(&xfr->lock, &xfr->refresh, sizeof(xfr->refresh));
6466         lock_protect(&xfr->lock, &xfr->expiry, sizeof(xfr->expiry));
6467         lock_protect(&xfr->lock, &xfr->lease_time, sizeof(xfr->lease_time));
6468         lock_protect(&xfr->lock, &xfr->task_nextprobe->worker,
6469                 sizeof(xfr->task_nextprobe->worker));
6470         lock_protect(&xfr->lock, &xfr->task_probe->worker,
6471                 sizeof(xfr->task_probe->worker));
6472         lock_protect(&xfr->lock, &xfr->task_transfer->worker,
6473                 sizeof(xfr->task_transfer->worker));
6474         lock_basic_lock(&xfr->lock);
6475         return xfr;
6476 }
6477
6478 /** Create auth_xfer structure.
6479  * This populates the have_zone, soa values, and so on times.
6480  * and sets the timeout, if a zone transfer is needed a short timeout is set.
6481  * For that the auth_zone itself must exist (and read in zonefile)
6482  * returns false on alloc failure. */
6483 struct auth_xfer*
6484 auth_xfer_create(struct auth_zones* az, struct auth_zone* z)
6485 {
6486         struct auth_xfer* xfr;
6487
6488         /* malloc it */
6489         xfr = auth_xfer_new(z);
6490         if(!xfr) {
6491                 log_err("malloc failure");
6492                 return NULL;
6493         }
6494         /* insert in tree */
6495         (void)rbtree_insert(&az->xtree, &xfr->node);
6496         return xfr;
6497 }
6498
6499 /** create new auth_master structure */
6500 static struct auth_master*
6501 auth_master_new(struct auth_master*** list)
6502 {
6503         struct auth_master *m;
6504         m = (struct auth_master*)calloc(1, sizeof(*m));
6505         if(!m) {
6506                 log_err("malloc failure");
6507                 return NULL;
6508         }
6509         /* set first pointer to m, or next pointer of previous element to m */
6510         (**list) = m;
6511         /* store m's next pointer as future point to store at */
6512         (*list) = &(m->next);
6513         return m;
6514 }
6515
6516 /** dup_prefix : create string from initial part of other string, malloced */
6517 static char*
6518 dup_prefix(char* str, size_t num)
6519 {
6520         char* result;
6521         size_t len = strlen(str);
6522         if(len < num) num = len; /* not more than strlen */
6523         result = (char*)malloc(num+1);
6524         if(!result) {
6525                 log_err("malloc failure");
6526                 return result;
6527         }
6528         memmove(result, str, num);
6529         result[num] = 0;
6530         return result;
6531 }
6532
6533 /** dup string and print error on error */
6534 static char*
6535 dup_all(char* str)
6536 {
6537         char* result = strdup(str);
6538         if(!result) {
6539                 log_err("malloc failure");
6540                 return NULL;
6541         }
6542         return result;
6543 }
6544
6545 /** find first of two characters */
6546 static char*
6547 str_find_first_of_chars(char* s, char a, char b)
6548 {
6549         char* ra = strchr(s, a);
6550         char* rb = strchr(s, b);
6551         if(!ra) return rb;
6552         if(!rb) return ra;
6553         if(ra < rb) return ra;
6554         return rb;
6555 }
6556
6557 /** parse URL into host and file parts, false on malloc or parse error */
6558 static int
6559 parse_url(char* url, char** host, char** file, int* port, int* ssl)
6560 {
6561         char* p = url;
6562         /* parse http://www.example.com/file.htm
6563          * or http://127.0.0.1   (index.html)
6564          * or https://[::1@1234]/a/b/c/d */
6565         *ssl = 1;
6566         *port = AUTH_HTTPS_PORT;
6567
6568         /* parse http:// or https:// */
6569         if(strncmp(p, "http://", 7) == 0) {
6570                 p += 7;
6571                 *ssl = 0;
6572                 *port = AUTH_HTTP_PORT;
6573         } else if(strncmp(p, "https://", 8) == 0) {
6574                 p += 8;
6575         } else if(strstr(p, "://") && strchr(p, '/') > strstr(p, "://") &&
6576                 strchr(p, ':') >= strstr(p, "://")) {
6577                 char* uri = dup_prefix(p, (size_t)(strstr(p, "://")-p));
6578                 log_err("protocol %s:// not supported (for url %s)",
6579                         uri?uri:"", p);
6580                 free(uri);
6581                 return 0;
6582         }
6583
6584         /* parse hostname part */
6585         if(p[0] == '[') {
6586                 char* end = strchr(p, ']');
6587                 p++; /* skip over [ */
6588                 if(end) {
6589                         *host = dup_prefix(p, (size_t)(end-p));
6590                         if(!*host) return 0;
6591                         p = end+1; /* skip over ] */
6592                 } else {
6593                         *host = dup_all(p);
6594                         if(!*host) return 0;
6595                         p = end;
6596                 }
6597         } else {
6598                 char* end = str_find_first_of_chars(p, ':', '/');
6599                 if(end) {
6600                         *host = dup_prefix(p, (size_t)(end-p));
6601                         if(!*host) return 0;
6602                 } else {
6603                         *host = dup_all(p);
6604                         if(!*host) return 0;
6605                 }
6606                 p = end; /* at next : or / or NULL */
6607         }
6608
6609         /* parse port number */
6610         if(p && p[0] == ':') {
6611                 char* end = NULL;
6612                 *port = strtol(p+1, &end, 10);
6613                 p = end;
6614         }
6615
6616         /* parse filename part */
6617         while(p && *p == '/')
6618                 p++;
6619         if(!p || p[0] == 0)
6620                 *file = strdup("index.html");
6621         else    *file = strdup(p);
6622         if(!*file) {
6623                 log_err("malloc failure");
6624                 return 0;
6625         }
6626         return 1;
6627 }
6628
6629 int
6630 xfer_set_masters(struct auth_master** list, struct config_auth* c,
6631         int with_http)
6632 {
6633         struct auth_master* m;
6634         struct config_strlist* p;
6635         /* list points to the first, or next pointer for the new element */
6636         while(*list) {
6637                 list = &( (*list)->next );
6638         }
6639         if(with_http)
6640           for(p = c->urls; p; p = p->next) {
6641                 m = auth_master_new(&list);
6642                 m->http = 1;
6643                 if(!parse_url(p->str, &m->host, &m->file, &m->port, &m->ssl))
6644                         return 0;
6645         }
6646         for(p = c->masters; p; p = p->next) {
6647                 m = auth_master_new(&list);
6648                 m->ixfr = 1; /* this flag is not configurable */
6649                 m->host = strdup(p->str);
6650                 if(!m->host) {
6651                         log_err("malloc failure");
6652                         return 0;
6653                 }
6654         }
6655         for(p = c->allow_notify; p; p = p->next) {
6656                 m = auth_master_new(&list);
6657                 m->allow_notify = 1;
6658                 m->host = strdup(p->str);
6659                 if(!m->host) {
6660                         log_err("malloc failure");
6661                         return 0;
6662                 }
6663         }
6664         return 1;
6665 }
6666
6667 #define SERIAL_BITS     32
6668 int
6669 compare_serial(uint32_t a, uint32_t b)
6670 {
6671         const uint32_t cutoff = ((uint32_t) 1 << (SERIAL_BITS - 1));
6672
6673         if (a == b) {
6674                 return 0;
6675         } else if ((a < b && b - a < cutoff) || (a > b && a - b > cutoff)) {
6676                 return -1;
6677         } else {
6678                 return 1;
6679         }
6680 }