1 /* 2 * query.c -- nsd(8) the resolver. 3 * 4 * Copyright (c) 2001-2006, NLnet Labs. All rights reserved. 5 * 6 * See LICENSE for the license. 7 * 8 */ 9 10 #include "config.h" 11 12 #include <sys/types.h> 13 #include <sys/socket.h> 14 #include <netinet/in.h> 15 #include <arpa/inet.h> 16 #include <assert.h> 17 #include <ctype.h> 18 #include <errno.h> 19 #include <limits.h> 20 #include <stddef.h> 21 #include <stdio.h> 22 #include <stdlib.h> 23 #include <string.h> 24 #include <time.h> 25 #include <unistd.h> 26 #include <netdb.h> 27 28 #include "answer.h" 29 #include "axfr.h" 30 #include "dns.h" 31 #include "dname.h" 32 #include "nsd.h" 33 #include "namedb.h" 34 #include "query.h" 35 #include "util.h" 36 #include "options.h" 37 #include "nsec3.h" 38 #include "tsig.h" 39 40 /* [Bug #253] Adding unnecessary NS RRset may lead to undesired truncation. 41 * This function determines if the final response packet needs the NS RRset 42 * included. Currently, it will only return negative if QTYPE == DNSKEY|DS. 43 * This way, resolvers won't fallback to TCP unnecessarily when priming 44 * trust anchors. 45 */ 46 static int answer_needs_ns(struct query *query); 47 48 static int add_rrset(struct query *query, 49 answer_type *answer, 50 rr_section_type section, 51 domain_type *owner, 52 rrset_type *rrset); 53 54 static void answer_authoritative(struct nsd *nsd, 55 struct query *q, 56 answer_type *answer, 57 size_t domain_number, 58 int exact, 59 domain_type *closest_match, 60 domain_type *closest_encloser, 61 const dname_type *qname); 62 63 static void answer_lookup_zone(struct nsd *nsd, struct query *q, 64 answer_type *answer, size_t domain_number, 65 int exact, domain_type *closest_match, 66 domain_type *closest_encloser, 67 const dname_type *qname); 68 69 void 70 query_put_dname_offset(struct query *q, domain_type *domain, uint16_t offset) 71 { 72 assert(q); 73 assert(domain); 74 assert(domain->number > 0); 75 76 if (offset > MAX_COMPRESSION_OFFSET) 77 return; 78 if (q->compressed_dname_count >= MAX_COMPRESSED_DNAMES) 79 return; 80 81 q->compressed_dname_offsets[domain->number] = offset; 82 q->compressed_dnames[q->compressed_dname_count] = domain; 83 ++q->compressed_dname_count; 84 } 85 86 void 87 query_clear_dname_offsets(struct query *q, size_t max_offset) 88 { 89 while (q->compressed_dname_count > 0 90 && (q->compressed_dname_offsets[q->compressed_dnames[q->compressed_dname_count - 1]->number] 91 >= max_offset)) 92 { 93 q->compressed_dname_offsets[q->compressed_dnames[q->compressed_dname_count - 1]->number] = 0; 94 --q->compressed_dname_count; 95 } 96 } 97 98 void 99 query_clear_compression_tables(struct query *q) 100 { 101 uint16_t i; 102 103 for (i = 0; i < q->compressed_dname_count; ++i) { 104 assert(q->compressed_dnames); 105 q->compressed_dname_offsets[q->compressed_dnames[i]->number] = 0; 106 } 107 q->compressed_dname_count = 0; 108 } 109 110 void 111 query_add_compression_domain(struct query *q, domain_type *domain, uint16_t offset) 112 { 113 while (domain->parent) { 114 DEBUG(DEBUG_NAME_COMPRESSION, 2, 115 (LOG_INFO, "query dname: %s, number: %lu, offset: %u\n", 116 domain_to_string(domain), 117 (unsigned long) domain->number, 118 offset)); 119 query_put_dname_offset(q, domain, offset); 120 offset += label_length(dname_name(domain_dname(domain))) + 1; 121 domain = domain->parent; 122 } 123 } 124 125 /* 126 * Generate an error response with the specified RCODE. 127 */ 128 query_state_type 129 query_error (struct query *q, nsd_rc_type rcode) 130 { 131 if (rcode == NSD_RC_DISCARD) { 132 return QUERY_DISCARDED; 133 } 134 135 buffer_clear(q->packet); 136 137 QR_SET(q->packet); /* This is an answer. */ 138 AD_CLR(q->packet); 139 RCODE_SET(q->packet, (int) rcode); /* Error code. */ 140 141 /* Truncate the question as well... */ 142 QDCOUNT_SET(q->packet, 0); 143 ANCOUNT_SET(q->packet, 0); 144 NSCOUNT_SET(q->packet, 0); 145 ARCOUNT_SET(q->packet, 0); 146 buffer_set_position(q->packet, QHEADERSZ); 147 return QUERY_PROCESSED; 148 } 149 150 static int 151 query_ratelimit_err(nsd_type* nsd) 152 { 153 time_t now = time(NULL); 154 if(nsd->err_limit_time == now) { 155 /* see if limit is exceeded for this second */ 156 if(nsd->err_limit_count++ > ERROR_RATELIMIT) 157 return 1; 158 } else { 159 /* new second, new limits */ 160 nsd->err_limit_time = now; 161 nsd->err_limit_count = 1; 162 } 163 return 0; 164 } 165 166 static query_state_type 167 query_formerr (struct query *query, nsd_type* nsd) 168 { 169 int opcode = OPCODE(query->packet); 170 if(query_ratelimit_err(nsd)) 171 return QUERY_DISCARDED; 172 FLAGS_SET(query->packet, FLAGS(query->packet) & 0x0100U); 173 /* Preserve the RD flag. Clear the rest. */ 174 OPCODE_SET(query->packet, opcode); 175 return query_error(query, NSD_RC_FORMAT); 176 } 177 178 static void 179 query_cleanup(void *data) 180 { 181 query_type *query = (query_type *) data; 182 region_destroy(query->region); 183 } 184 185 query_type * 186 query_create(region_type *region, uint16_t *compressed_dname_offsets, 187 size_t compressed_dname_size, domain_type **compressed_dnames) 188 { 189 query_type *query 190 = (query_type *) region_alloc_zero(region, sizeof(query_type)); 191 /* create region with large block size, because the initial chunk 192 saves many mallocs in the server */ 193 query->region = region_create_custom(xalloc, free, 16384, 16384/8, 32, 0); 194 query->compressed_dname_offsets = compressed_dname_offsets; 195 query->compressed_dnames = compressed_dnames; 196 query->packet = buffer_create(region, QIOBUFSZ); 197 region_add_cleanup(region, query_cleanup, query); 198 query->compressed_dname_offsets_size = compressed_dname_size; 199 tsig_create_record(&query->tsig, region); 200 query->tsig_prepare_it = 1; 201 query->tsig_update_it = 1; 202 query->tsig_sign_it = 1; 203 return query; 204 } 205 206 void 207 query_reset(query_type *q, size_t maxlen, int is_tcp) 208 { 209 /* 210 * As long as less than 4Kb (region block size) has been used, 211 * this call to free_all is free, the block is saved for re-use, 212 * so no malloc() or free() calls are done. 213 * at present use of the region is for: 214 * o query qname dname_type (255 max). 215 * o wildcard expansion domain_type (7*ptr+u32+2bytes)+(5*ptr nsec3) 216 * o wildcard expansion for additional section domain_type. 217 * o nsec3 hashed name(s) (3 dnames for a nonexist_proof, 218 * one proof per wildcard and for nx domain). 219 */ 220 region_free_all(q->region); 221 q->addrlen = sizeof(q->addr); 222 q->maxlen = maxlen; 223 q->reserved_space = 0; 224 buffer_clear(q->packet); 225 edns_init_record(&q->edns); 226 tsig_init_record(&q->tsig, NULL, NULL); 227 q->tsig_prepare_it = 1; 228 q->tsig_update_it = 1; 229 q->tsig_sign_it = 1; 230 q->tcp = is_tcp; 231 q->qname = NULL; 232 q->qtype = 0; 233 q->qclass = 0; 234 q->zone = NULL; 235 q->opcode = 0; 236 q->cname_count = 0; 237 q->delegation_domain = NULL; 238 q->delegation_rrset = NULL; 239 q->compressed_dname_count = 0; 240 q->number_temporary_domains = 0; 241 242 q->axfr_is_done = 0; 243 q->axfr_zone = NULL; 244 q->axfr_current_domain = NULL; 245 q->axfr_current_rrset = NULL; 246 q->axfr_current_rr = 0; 247 248 #ifdef RATELIMIT 249 q->wildcard_domain = NULL; 250 #endif 251 } 252 253 /* get a temporary domain number (or 0=failure) */ 254 static domain_type* 255 query_get_tempdomain(struct query *q) 256 { 257 static domain_type d[EXTRA_DOMAIN_NUMBERS]; 258 if(q->number_temporary_domains >= EXTRA_DOMAIN_NUMBERS) 259 return 0; 260 q->number_temporary_domains ++; 261 memset(&d[q->number_temporary_domains-1], 0, sizeof(domain_type)); 262 d[q->number_temporary_domains-1].number = q->compressed_dname_offsets_size + 263 q->number_temporary_domains - 1; 264 return &d[q->number_temporary_domains-1]; 265 } 266 267 static void 268 query_addtxt(struct query *q, 269 const uint8_t *dname, 270 uint16_t klass, 271 uint32_t ttl, 272 const char *txt) 273 { 274 size_t txt_length = strlen(txt); 275 uint8_t len = (uint8_t) txt_length; 276 277 assert(txt_length <= UCHAR_MAX); 278 279 /* Add the dname */ 280 if (dname >= buffer_begin(q->packet) 281 && dname <= buffer_current(q->packet)) 282 { 283 buffer_write_u16(q->packet, 284 0xc000 | (dname - buffer_begin(q->packet))); 285 } else { 286 buffer_write(q->packet, dname + 1, *dname); 287 } 288 289 buffer_write_u16(q->packet, TYPE_TXT); 290 buffer_write_u16(q->packet, klass); 291 buffer_write_u32(q->packet, ttl); 292 buffer_write_u16(q->packet, len + 1); 293 buffer_write_u8(q->packet, len); 294 buffer_write(q->packet, txt, len); 295 } 296 297 /* 298 * Parse the question section of a query. The normalized query name 299 * is stored in QUERY->name, the class in QUERY->klass, and the type 300 * in QUERY->type. 301 */ 302 static int 303 process_query_section(query_type *query) 304 { 305 uint8_t qnamebuf[MAXDOMAINLEN]; 306 307 buffer_set_position(query->packet, QHEADERSZ); 308 /* Lets parse the query name and convert it to lower case. */ 309 if(!packet_read_query_section(query->packet, qnamebuf, 310 &query->qtype, &query->qclass)) 311 return 0; 312 query->qname = dname_make(query->region, qnamebuf, 1); 313 return 1; 314 } 315 316 317 /* 318 * Process an optional EDNS OPT record. Sets QUERY->EDNS to 0 if 319 * there was no EDNS record, to -1 if there was an invalid or 320 * unsupported EDNS record, and to 1 otherwise. Updates QUERY->MAXLEN 321 * if the EDNS record specifies a maximum supported response length. 322 * 323 * Return NSD_RC_FORMAT on failure, NSD_RC_OK on success. 324 */ 325 static nsd_rc_type 326 process_edns(nsd_type* nsd, struct query *q) 327 { 328 if (q->edns.status == EDNS_ERROR) { 329 /* The only error is VERSION not implemented */ 330 return NSD_RC_FORMAT; 331 } 332 333 if (q->edns.status == EDNS_OK) { 334 /* Only care about UDP size larger than normal... */ 335 if (!q->tcp && q->edns.maxlen > UDP_MAX_MESSAGE_LEN) { 336 size_t edns_size; 337 #if defined(INET6) 338 if (q->addr.ss_family == AF_INET6) { 339 edns_size = nsd->ipv6_edns_size; 340 } else 341 #endif 342 edns_size = nsd->ipv4_edns_size; 343 344 if (q->edns.maxlen < edns_size) { 345 q->maxlen = q->edns.maxlen; 346 } else { 347 q->maxlen = edns_size; 348 } 349 350 #if defined(INET6) && !defined(IPV6_USE_MIN_MTU) && !defined(IPV6_MTU) 351 /* 352 * Use IPv6 minimum MTU to avoid sending 353 * packets that are too large for some links. 354 * IPv6 will not automatically fragment in 355 * this case (unlike IPv4). 356 */ 357 if (q->addr.ss_family == AF_INET6 358 && q->maxlen > IPV6_MIN_MTU) 359 { 360 q->maxlen = IPV6_MIN_MTU; 361 } 362 #endif 363 } 364 365 /* Strip the OPT resource record off... */ 366 buffer_set_position(q->packet, q->edns.position); 367 buffer_set_limit(q->packet, q->edns.position); 368 ARCOUNT_SET(q->packet, ARCOUNT(q->packet) - 1); 369 } 370 return NSD_RC_OK; 371 } 372 373 /* 374 * Processes TSIG. 375 * Sets error when tsig does not verify on the query. 376 */ 377 static nsd_rc_type 378 process_tsig(struct query* q) 379 { 380 if(q->tsig.status == TSIG_ERROR) 381 return NSD_RC_FORMAT; 382 if(q->tsig.status == TSIG_OK) { 383 if(!tsig_from_query(&q->tsig)) { 384 char a[128]; 385 addr2str(&q->addr, a, sizeof(a)); 386 log_msg(LOG_ERR, "query: bad tsig (%s) for key %s from %s", 387 tsig_error(q->tsig.error_code), 388 dname_to_string(q->tsig.key_name, NULL), a); 389 return NSD_RC_NOTAUTH; 390 } 391 buffer_set_limit(q->packet, q->tsig.position); 392 ARCOUNT_SET(q->packet, ARCOUNT(q->packet) - 1); 393 tsig_prepare(&q->tsig); 394 tsig_update(&q->tsig, q->packet, buffer_limit(q->packet)); 395 if(!tsig_verify(&q->tsig)) { 396 char a[128]; 397 addr2str(&q->addr, a, sizeof(a)); 398 log_msg(LOG_ERR, "query: bad tsig signature for key %s from %s", 399 dname_to_string(q->tsig.key->name, NULL), a); 400 return NSD_RC_NOTAUTH; 401 } 402 DEBUG(DEBUG_XFRD,1, (LOG_INFO, "query good tsig signature for %s", 403 dname_to_string(q->tsig.key->name, NULL))); 404 } 405 return NSD_RC_OK; 406 } 407 408 /* 409 * Check notify acl and forward to xfrd (or return an error). 410 */ 411 static query_state_type 412 answer_notify(struct nsd* nsd, struct query *query) 413 { 414 int acl_num, acl_num_xfr; 415 struct acl_options *why; 416 nsd_rc_type rc; 417 418 struct zone_options* zone_opt; 419 DEBUG(DEBUG_XFRD,1, (LOG_INFO, "got notify %s processing acl", 420 dname_to_string(query->qname, NULL))); 421 422 zone_opt = zone_options_find(nsd->options, query->qname); 423 if(!zone_opt) 424 return query_error(query, NSD_RC_NXDOMAIN); 425 426 if(!nsd->this_child) /* we are in debug mode or something */ 427 return query_error(query, NSD_RC_SERVFAIL); 428 429 if(!tsig_find_rr(&query->tsig, query->packet)) { 430 DEBUG(DEBUG_XFRD,2, (LOG_ERR, "bad tsig RR format")); 431 return query_error(query, NSD_RC_FORMAT); 432 } 433 rc = process_tsig(query); 434 if(rc != NSD_RC_OK) 435 return query_error(query, rc); 436 437 /* check if it passes acl */ 438 if((acl_num = acl_check_incoming(zone_opt->pattern->allow_notify, query, 439 &why)) != -1) 440 { 441 sig_atomic_t mode = NSD_PASS_TO_XFRD; 442 int s = nsd->this_child->parent_fd; 443 uint16_t sz; 444 uint32_t acl_send = htonl(acl_num); 445 uint32_t acl_xfr; 446 size_t pos; 447 448 /* Find priority candidate for request XFR. -1 if no match */ 449 acl_num_xfr = acl_check_incoming( 450 zone_opt->pattern->request_xfr, query, NULL); 451 452 acl_xfr = htonl(acl_num_xfr); 453 454 assert(why); 455 DEBUG(DEBUG_XFRD,1, (LOG_INFO, "got notify %s passed acl %s %s", 456 dname_to_string(query->qname, NULL), 457 why->ip_address_spec, 458 why->nokey?"NOKEY": 459 (why->blocked?"BLOCKED":why->key_name))); 460 sz = buffer_limit(query->packet); 461 if(buffer_limit(query->packet) > MAX_PACKET_SIZE) 462 return query_error(query, NSD_RC_SERVFAIL); 463 /* forward to xfrd for processing 464 Note. Blocking IPC I/O, but acl is OK. */ 465 sz = htons(sz); 466 if(!write_socket(s, &mode, sizeof(mode)) || 467 !write_socket(s, &sz, sizeof(sz)) || 468 !write_socket(s, buffer_begin(query->packet), 469 buffer_limit(query->packet)) || 470 !write_socket(s, &acl_send, sizeof(acl_send)) || 471 !write_socket(s, &acl_xfr, sizeof(acl_xfr))) { 472 log_msg(LOG_ERR, "error in IPC notify server2main, %s", 473 strerror(errno)); 474 return query_error(query, NSD_RC_SERVFAIL); 475 } 476 if(verbosity >= 1) { 477 uint32_t serial = 0; 478 char address[128]; 479 addr2str(&query->addr, address, sizeof(address)); 480 if(packet_find_notify_serial(query->packet, &serial)) 481 VERBOSITY(1, (LOG_INFO, "notify for %s from %s serial %u", 482 dname_to_string(query->qname, NULL), address, 483 (unsigned)serial)); 484 else 485 VERBOSITY(1, (LOG_INFO, "notify for %s from %s", 486 dname_to_string(query->qname, NULL), address)); 487 } 488 489 /* create notify reply - keep same query contents */ 490 QR_SET(query->packet); /* This is an answer. */ 491 AA_SET(query->packet); /* we are authoritative. */ 492 ANCOUNT_SET(query->packet, 0); 493 NSCOUNT_SET(query->packet, 0); 494 ARCOUNT_SET(query->packet, 0); 495 RCODE_SET(query->packet, RCODE_OK); /* Error code. */ 496 /* position is right after the query */ 497 pos = buffer_position(query->packet); 498 buffer_clear(query->packet); 499 buffer_set_position(query->packet, pos); 500 /* tsig is added in add_additional later (if needed) */ 501 return QUERY_PROCESSED; 502 } 503 504 if (verbosity >= 2) { 505 char address[128]; 506 addr2str(&query->addr, address, sizeof(address)); 507 VERBOSITY(2, (LOG_INFO, "notify for %s from %s refused, %s%s", 508 dname_to_string(query->qname, NULL), 509 address, 510 why?why->key_name:"no acl matches", 511 why?why->ip_address_spec:".")); 512 } 513 514 return query_error(query, NSD_RC_REFUSE); 515 } 516 517 518 /* 519 * Answer a query in the CHAOS class. 520 */ 521 static query_state_type 522 answer_chaos(struct nsd *nsd, query_type *q) 523 { 524 AA_CLR(q->packet); 525 switch (q->qtype) { 526 case TYPE_ANY: 527 case TYPE_TXT: 528 if ((q->qname->name_size == 11 529 && memcmp(dname_name(q->qname), "\002id\006server", 11) == 0) || 530 (q->qname->name_size == 15 531 && memcmp(dname_name(q->qname), "\010hostname\004bind", 15) == 0)) 532 { 533 if(!nsd->options->hide_identity) { 534 /* Add ID */ 535 query_addtxt(q, 536 buffer_begin(q->packet) + QHEADERSZ, 537 CLASS_CH, 538 0, 539 nsd->identity); 540 ANCOUNT_SET(q->packet, ANCOUNT(q->packet) + 1); 541 } else { 542 RCODE_SET(q->packet, RCODE_REFUSE); 543 } 544 } else if ((q->qname->name_size == 16 545 && memcmp(dname_name(q->qname), "\007version\006server", 16) == 0) || 546 (q->qname->name_size == 14 547 && memcmp(dname_name(q->qname), "\007version\004bind", 14) == 0)) 548 { 549 if(!nsd->options->hide_version) { 550 /* Add version */ 551 query_addtxt(q, 552 buffer_begin(q->packet) + QHEADERSZ, 553 CLASS_CH, 554 0, 555 nsd->version); 556 ANCOUNT_SET(q->packet, ANCOUNT(q->packet) + 1); 557 } else { 558 RCODE_SET(q->packet, RCODE_REFUSE); 559 } 560 } else { 561 RCODE_SET(q->packet, RCODE_REFUSE); 562 } 563 break; 564 default: 565 RCODE_SET(q->packet, RCODE_REFUSE); 566 break; 567 } 568 569 return QUERY_PROCESSED; 570 } 571 572 573 /* 574 * Find the covering NSEC for a non-existent domain name. Normally 575 * the NSEC will be located at CLOSEST_MATCH, except when it is an 576 * empty non-terminal. In this case the NSEC may be located at the 577 * previous domain name (in canonical ordering). 578 */ 579 static domain_type * 580 find_covering_nsec(domain_type *closest_match, 581 zone_type *zone, 582 rrset_type **nsec_rrset) 583 { 584 assert(closest_match); 585 assert(nsec_rrset); 586 587 /* loop away temporary created domains. For real ones it is &RBTREE_NULL */ 588 #ifdef USE_RADIX_TREE 589 while (closest_match->rnode == NULL) 590 #else 591 while (closest_match->node.parent == NULL) 592 #endif 593 closest_match = closest_match->parent; 594 while (closest_match) { 595 *nsec_rrset = domain_find_rrset(closest_match, zone, TYPE_NSEC); 596 if (*nsec_rrset) { 597 return closest_match; 598 } 599 if (closest_match == zone->apex) { 600 /* Don't look outside the current zone. */ 601 return NULL; 602 } 603 closest_match = domain_previous(closest_match); 604 } 605 return NULL; 606 } 607 608 609 struct additional_rr_types 610 { 611 uint16_t rr_type; 612 rr_section_type rr_section; 613 }; 614 615 struct additional_rr_types default_additional_rr_types[] = { 616 { TYPE_A, ADDITIONAL_A_SECTION }, 617 { TYPE_AAAA, ADDITIONAL_AAAA_SECTION }, 618 { 0, (rr_section_type) 0 } 619 }; 620 621 struct additional_rr_types swap_aaaa_additional_rr_types[] = { 622 { TYPE_AAAA, ADDITIONAL_A_SECTION }, 623 { TYPE_A, ADDITIONAL_AAAA_SECTION }, 624 { 0, (rr_section_type) 0 } 625 }; 626 627 struct additional_rr_types rt_additional_rr_types[] = { 628 { TYPE_A, ADDITIONAL_A_SECTION }, 629 { TYPE_AAAA, ADDITIONAL_AAAA_SECTION }, 630 { TYPE_X25, ADDITIONAL_OTHER_SECTION }, 631 { TYPE_ISDN, ADDITIONAL_OTHER_SECTION }, 632 { 0, (rr_section_type) 0 } 633 }; 634 635 static void 636 add_additional_rrsets(struct query *query, answer_type *answer, 637 rrset_type *master_rrset, size_t rdata_index, 638 int allow_glue, struct additional_rr_types types[]) 639 { 640 size_t i; 641 642 assert(query); 643 assert(answer); 644 assert(master_rrset); 645 assert(rdata_atom_is_domain(rrset_rrtype(master_rrset), rdata_index)); 646 647 for (i = 0; i < master_rrset->rr_count; ++i) { 648 int j; 649 domain_type *additional = rdata_atom_domain(master_rrset->rrs[i].rdatas[rdata_index]); 650 domain_type *match = additional; 651 652 assert(additional); 653 654 if (!allow_glue && domain_is_glue(match, query->zone)) 655 continue; 656 657 /* 658 * Check to see if we need to generate the dependent 659 * based on a wildcard domain. 660 */ 661 while (!match->is_existing) { 662 match = match->parent; 663 } 664 if (additional != match && domain_wildcard_child(match)) { 665 domain_type *wildcard_child = domain_wildcard_child(match); 666 domain_type *temp = (domain_type *) region_alloc( 667 query->region, sizeof(domain_type)); 668 #ifdef USE_RADIX_TREE 669 temp->rnode = NULL; 670 temp->dname = additional->dname; 671 #else 672 memcpy(&temp->node, &additional->node, sizeof(rbnode_type)); 673 temp->node.parent = NULL; 674 #endif 675 temp->number = additional->number; 676 temp->parent = match; 677 temp->wildcard_child_closest_match = temp; 678 temp->rrsets = wildcard_child->rrsets; 679 temp->is_existing = wildcard_child->is_existing; 680 additional = temp; 681 } 682 683 for (j = 0; types[j].rr_type != 0; ++j) { 684 rrset_type *rrset = domain_find_rrset( 685 additional, query->zone, types[j].rr_type); 686 if (rrset) { 687 answer_add_rrset(answer, types[j].rr_section, 688 additional, rrset); 689 } 690 } 691 } 692 } 693 694 static int 695 answer_needs_ns(struct query* query) 696 { 697 assert(query); 698 /* Currently, only troublesome for DNSKEY and DS, 699 * cuz their RRSETs are quite large. */ 700 return (query->qtype != TYPE_DNSKEY && query->qtype != TYPE_DS 701 && query->qtype != TYPE_ANY); 702 } 703 704 static int 705 add_rrset(struct query *query, 706 answer_type *answer, 707 rr_section_type section, 708 domain_type *owner, 709 rrset_type *rrset) 710 { 711 int result; 712 713 assert(query); 714 assert(answer); 715 assert(owner); 716 assert(rrset); 717 assert(rrset_rrclass(rrset) == CLASS_IN); 718 719 result = answer_add_rrset(answer, section, owner, rrset); 720 if(minimal_responses && section != AUTHORITY_SECTION && 721 query->qtype != TYPE_NS) 722 return result; 723 switch (rrset_rrtype(rrset)) { 724 case TYPE_NS: 725 #if defined(INET6) 726 /* if query over IPv6, swap A and AAAA; put AAAA first */ 727 add_additional_rrsets(query, answer, rrset, 0, 1, 728 (query->addr.ss_family == AF_INET6)? 729 swap_aaaa_additional_rr_types: 730 default_additional_rr_types); 731 #else 732 add_additional_rrsets(query, answer, rrset, 0, 1, 733 default_additional_rr_types); 734 #endif 735 break; 736 case TYPE_MB: 737 add_additional_rrsets(query, answer, rrset, 0, 0, 738 default_additional_rr_types); 739 break; 740 case TYPE_MX: 741 case TYPE_KX: 742 add_additional_rrsets(query, answer, rrset, 1, 0, 743 default_additional_rr_types); 744 break; 745 case TYPE_RT: 746 add_additional_rrsets(query, answer, rrset, 1, 0, 747 rt_additional_rr_types); 748 break; 749 case TYPE_SRV: 750 add_additional_rrsets(query, answer, rrset, 3, 0, 751 default_additional_rr_types); 752 break; 753 default: 754 break; 755 } 756 757 return result; 758 } 759 760 761 /* returns 0 on error, or the domain number for to_name. 762 from_name is changes to to_name by the DNAME rr. 763 DNAME rr is from src to dest. 764 closest encloser encloses the to_name. */ 765 static size_t 766 query_synthesize_cname(struct query* q, struct answer* answer, const dname_type* from_name, 767 const dname_type* to_name, domain_type* src, domain_type* to_closest_encloser, 768 domain_type** to_closest_match, uint32_t ttl) 769 { 770 /* add temporary domains for from_name and to_name and all 771 their (not allocated yet) parents */ 772 /* any domains below src are not_existing (because of DNAME at src) */ 773 int i; 774 domain_type* cname_domain; 775 domain_type* cname_dest; 776 rrset_type* rrset; 777 778 /* allocate source part */ 779 domain_type* lastparent = src; 780 assert(q && answer && from_name && to_name && src && to_closest_encloser); 781 assert(to_closest_match); 782 for(i=0; i < from_name->label_count - domain_dname(src)->label_count; i++) 783 { 784 domain_type* newdom = query_get_tempdomain(q); 785 if(!newdom) 786 return 0; 787 newdom->is_existing = 1; 788 newdom->parent = lastparent; 789 #ifdef USE_RADIX_TREE 790 newdom->dname 791 #else 792 newdom->node.key 793 #endif 794 = dname_partial_copy(q->region, 795 from_name, domain_dname(src)->label_count + i + 1); 796 if(dname_compare(domain_dname(newdom), q->qname) == 0) { 797 /* 0 good for query name, otherwise new number */ 798 newdom->number = 0; 799 } 800 DEBUG(DEBUG_QUERY,2, (LOG_INFO, "created temp domain src %d. %s nr %d", i, 801 domain_to_string(newdom), (int)newdom->number)); 802 lastparent = newdom; 803 } 804 cname_domain = lastparent; 805 806 /* allocate dest part */ 807 lastparent = to_closest_encloser; 808 for(i=0; i < to_name->label_count - domain_dname(to_closest_encloser)->label_count; 809 i++) 810 { 811 domain_type* newdom = query_get_tempdomain(q); 812 if(!newdom) 813 return 0; 814 newdom->is_existing = 0; 815 newdom->parent = lastparent; 816 #ifdef USE_RADIX_TREE 817 newdom->dname 818 #else 819 newdom->node.key 820 #endif 821 = dname_partial_copy(q->region, 822 to_name, domain_dname(to_closest_encloser)->label_count + i + 1); 823 DEBUG(DEBUG_QUERY,2, (LOG_INFO, "created temp domain dest %d. %s nr %d", i, 824 domain_to_string(newdom), (int)newdom->number)); 825 lastparent = newdom; 826 } 827 cname_dest = lastparent; 828 *to_closest_match = cname_dest; 829 830 /* allocate the CNAME RR */ 831 rrset = (rrset_type*) region_alloc(q->region, sizeof(rrset_type)); 832 memset(rrset, 0, sizeof(rrset_type)); 833 rrset->zone = q->zone; 834 rrset->rr_count = 1; 835 rrset->rrs = (rr_type*) region_alloc(q->region, sizeof(rr_type)); 836 memset(rrset->rrs, 0, sizeof(rr_type)); 837 rrset->rrs->owner = cname_domain; 838 rrset->rrs->ttl = ttl; 839 rrset->rrs->type = TYPE_CNAME; 840 rrset->rrs->klass = CLASS_IN; 841 rrset->rrs->rdata_count = 1; 842 rrset->rrs->rdatas = (rdata_atom_type*)region_alloc(q->region, 843 sizeof(rdata_atom_type)); 844 rrset->rrs->rdatas->domain = cname_dest; 845 846 if(!add_rrset(q, answer, ANSWER_SECTION, cname_domain, rrset)) { 847 log_msg(LOG_ERR, "could not add synthesized CNAME rrset to packet"); 848 } 849 850 return cname_dest->number; 851 } 852 853 /* 854 * Answer delegation information. 855 * 856 * DNSSEC: Include the DS RRset if present. Otherwise include an NSEC 857 * record proving the DS RRset does not exist. 858 */ 859 static void 860 answer_delegation(query_type *query, answer_type *answer) 861 { 862 assert(answer); 863 assert(query->delegation_domain); 864 assert(query->delegation_rrset); 865 866 if (query->cname_count == 0) { 867 AA_CLR(query->packet); 868 } else { 869 AA_SET(query->packet); 870 } 871 872 add_rrset(query, 873 answer, 874 AUTHORITY_SECTION, 875 query->delegation_domain, 876 query->delegation_rrset); 877 if (query->edns.dnssec_ok && zone_is_secure(query->zone)) { 878 rrset_type *rrset; 879 if ((rrset = domain_find_rrset(query->delegation_domain, query->zone, TYPE_DS))) { 880 add_rrset(query, answer, AUTHORITY_SECTION, 881 query->delegation_domain, rrset); 882 #ifdef NSEC3 883 } else if (query->zone->nsec3_param) { 884 nsec3_answer_delegation(query, answer); 885 #endif 886 } else if ((rrset = domain_find_rrset(query->delegation_domain, query->zone, TYPE_NSEC))) { 887 add_rrset(query, answer, AUTHORITY_SECTION, 888 query->delegation_domain, rrset); 889 } 890 } 891 } 892 893 894 /* 895 * Answer SOA information. 896 */ 897 static void 898 answer_soa(struct query *query, answer_type *answer) 899 { 900 if (query->qclass != CLASS_ANY) { 901 add_rrset(query, answer, 902 AUTHORITY_SECTION, 903 query->zone->apex, 904 query->zone->soa_nx_rrset); 905 } 906 } 907 908 909 /* 910 * Answer that the domain name exists but there is no RRset with the 911 * requested type. 912 * 913 * DNSSEC: Include the correct NSEC record proving that the type does 914 * not exist. In the wildcard no data (3.1.3.4) case the wildcard IS 915 * NOT expanded, so the ORIGINAL parameter must point to the original 916 * wildcard entry, not to the generated entry. 917 */ 918 static void 919 answer_nodata(struct query *query, answer_type *answer, domain_type *original) 920 { 921 answer_soa(query, answer); 922 923 #ifdef NSEC3 924 if (query->edns.dnssec_ok && query->zone->nsec3_param) { 925 nsec3_answer_nodata(query, answer, original); 926 } else 927 #endif 928 if (query->edns.dnssec_ok && zone_is_secure(query->zone)) { 929 domain_type *nsec_domain; 930 rrset_type *nsec_rrset; 931 932 nsec_domain = find_covering_nsec(original, query->zone, &nsec_rrset); 933 if (nsec_domain) { 934 add_rrset(query, answer, AUTHORITY_SECTION, nsec_domain, nsec_rrset); 935 } 936 } 937 } 938 939 static void 940 answer_nxdomain(query_type *query, answer_type *answer) 941 { 942 RCODE_SET(query->packet, RCODE_NXDOMAIN); 943 answer_soa(query, answer); 944 } 945 946 947 /* 948 * Answer domain information (or SOA if we do not have an RRset for 949 * the type specified by the query). 950 */ 951 static void 952 answer_domain(struct nsd* nsd, struct query *q, answer_type *answer, 953 domain_type *domain, domain_type *original) 954 { 955 rrset_type *rrset; 956 957 if (q->qtype == TYPE_ANY) { 958 rrset_type *preferred_rrset = NULL; 959 rrset_type *normal_rrset = NULL; 960 rrset_type *non_preferred_rrset = NULL; 961 962 /* 963 * Minimize response size for ANY, with one RRset 964 * according to RFC 8482(4.1). 965 * Prefers popular and not large rtypes (A,AAAA,...) 966 * lowering large ones (DNSKEY,RRSIG,...). 967 */ 968 for (rrset = domain_find_any_rrset(domain, q->zone); rrset; rrset = rrset->next) { 969 if (rrset->zone == q->zone 970 #ifdef NSEC3 971 && rrset_rrtype(rrset) != TYPE_NSEC3 972 #endif 973 /* 974 * Don't include the RRSIG RRset when 975 * DNSSEC is used, because it is added 976 * automatically on an per-RRset basis. 977 */ 978 && !(q->edns.dnssec_ok 979 && zone_is_secure(q->zone) 980 && rrset_rrtype(rrset) == TYPE_RRSIG)) 981 { 982 switch(rrset_rrtype(rrset)) { 983 case TYPE_A: 984 case TYPE_AAAA: 985 case TYPE_SOA: 986 case TYPE_MX: 987 case TYPE_PTR: 988 preferred_rrset = rrset; 989 break; 990 case TYPE_DNSKEY: 991 case TYPE_RRSIG: 992 case TYPE_NSEC: 993 non_preferred_rrset = rrset; 994 break; 995 default: 996 normal_rrset = rrset; 997 } 998 if (preferred_rrset) break; 999 } 1000 } 1001 if (preferred_rrset) { 1002 add_rrset(q, answer, ANSWER_SECTION, domain, preferred_rrset); 1003 } else if (normal_rrset) { 1004 add_rrset(q, answer, ANSWER_SECTION, domain, normal_rrset); 1005 } else if (non_preferred_rrset) { 1006 add_rrset(q, answer, ANSWER_SECTION, domain, non_preferred_rrset); 1007 } else { 1008 answer_nodata(q, answer, original); 1009 return; 1010 } 1011 #ifdef NSEC3 1012 } else if (q->qtype == TYPE_NSEC3) { 1013 answer_nodata(q, answer, original); 1014 return; 1015 #endif 1016 } else if ((rrset = domain_find_rrset(domain, q->zone, q->qtype))) { 1017 add_rrset(q, answer, ANSWER_SECTION, domain, rrset); 1018 } else if ((rrset = domain_find_rrset(domain, q->zone, TYPE_CNAME))) { 1019 int added; 1020 1021 /* 1022 * If the CNAME is not added it is already in the 1023 * answer, so we have a CNAME loop. Don't follow the 1024 * CNAME target in this case. 1025 */ 1026 added = add_rrset(q, answer, ANSWER_SECTION, domain, rrset); 1027 assert(rrset->rr_count > 0); 1028 if (added) { 1029 /* only process first CNAME record */ 1030 domain_type *closest_match = rdata_atom_domain(rrset->rrs[0].rdatas[0]); 1031 domain_type *closest_encloser = closest_match; 1032 zone_type* origzone = q->zone; 1033 ++q->cname_count; 1034 1035 answer_lookup_zone(nsd, q, answer, closest_match->number, 1036 closest_match == closest_encloser, 1037 closest_match, closest_encloser, 1038 domain_dname(closest_match)); 1039 q->zone = origzone; 1040 } 1041 return; 1042 } else { 1043 answer_nodata(q, answer, original); 1044 return; 1045 } 1046 1047 if (q->qclass != CLASS_ANY && q->zone->ns_rrset && answer_needs_ns(q) 1048 && !minimal_responses) { 1049 add_rrset(q, answer, OPTIONAL_AUTHORITY_SECTION, q->zone->apex, 1050 q->zone->ns_rrset); 1051 } 1052 } 1053 1054 1055 /* 1056 * Answer with authoritative data. If a wildcard is matched the owner 1057 * name will be expanded to the domain name specified by 1058 * DOMAIN_NUMBER. DOMAIN_NUMBER 0 (zero) is reserved for the original 1059 * query name. 1060 * 1061 * DNSSEC: Include the necessary NSEC records in case the request 1062 * domain name does not exist and/or a wildcard match does not exist. 1063 */ 1064 static void 1065 answer_authoritative(struct nsd *nsd, 1066 struct query *q, 1067 answer_type *answer, 1068 size_t domain_number, 1069 int exact, 1070 domain_type *closest_match, 1071 domain_type *closest_encloser, 1072 const dname_type *qname) 1073 { 1074 domain_type *match; 1075 domain_type *original = closest_match; 1076 domain_type *dname_ce; 1077 rrset_type *rrset; 1078 1079 #ifdef NSEC3 1080 if(exact && domain_has_only_NSEC3(closest_match, q->zone)) { 1081 exact = 0; /* pretend it does not exist */ 1082 if(closest_encloser->parent) 1083 closest_encloser = closest_encloser->parent; 1084 } 1085 #endif /* NSEC3 */ 1086 if((dname_ce = find_dname_above(closest_encloser, q->zone)) != NULL) { 1087 /* occlude the found data, the DNAME is closest_encloser */ 1088 closest_encloser = dname_ce; 1089 exact = 0; 1090 } 1091 1092 if (exact) { 1093 match = closest_match; 1094 } else if ((rrset=domain_find_rrset(closest_encloser, q->zone, TYPE_DNAME))) { 1095 /* process DNAME */ 1096 const dname_type* name = qname; 1097 domain_type *dest = rdata_atom_domain(rrset->rrs[0].rdatas[0]); 1098 int added; 1099 assert(rrset->rr_count > 0); 1100 if(domain_number != 0) /* we followed CNAMEs or DNAMEs */ 1101 name = domain_dname(closest_match); 1102 DEBUG(DEBUG_QUERY,2, (LOG_INFO, "expanding DNAME for q=%s", dname_to_string(name, NULL))); 1103 DEBUG(DEBUG_QUERY,2, (LOG_INFO, "->src is %s", 1104 domain_to_string(closest_encloser))); 1105 DEBUG(DEBUG_QUERY,2, (LOG_INFO, "->dest is %s", 1106 domain_to_string(dest))); 1107 /* if the DNAME set is not added we have a loop, do not follow */ 1108 added = add_rrset(q, answer, ANSWER_SECTION, closest_encloser, rrset); 1109 if(added) { 1110 domain_type* src = closest_encloser; 1111 const dname_type* newname = dname_replace(q->region, name, 1112 domain_dname(src), domain_dname(dest)); 1113 size_t newnum = 0; 1114 zone_type* origzone = q->zone; 1115 ++q->cname_count; 1116 if(!newname) { /* newname too long */ 1117 RCODE_SET(q->packet, RCODE_YXDOMAIN); 1118 return; 1119 } 1120 DEBUG(DEBUG_QUERY,2, (LOG_INFO, "->result is %s", dname_to_string(newname, NULL))); 1121 /* follow the DNAME */ 1122 (void)namedb_lookup(nsd->db, newname, &closest_match, &closest_encloser); 1123 /* synthesize CNAME record */ 1124 newnum = query_synthesize_cname(q, answer, name, newname, 1125 src, closest_encloser, &closest_match, rrset->rrs[0].ttl); 1126 if(!newnum) { 1127 /* could not synthesize the CNAME. */ 1128 /* return previous CNAMEs to make resolver recurse for us */ 1129 return; 1130 } 1131 1132 answer_lookup_zone(nsd, q, answer, newnum, 1133 closest_match == closest_encloser, 1134 closest_match, closest_encloser, newname); 1135 q->zone = origzone; 1136 } 1137 if(!added) /* log the error so operator can find looping recursors */ 1138 log_msg(LOG_INFO, "DNAME processing stopped due to loop, qname %s", 1139 dname_to_string(q->qname, NULL)); 1140 return; 1141 } else if (domain_wildcard_child(closest_encloser)) { 1142 /* Generate the domain from the wildcard. */ 1143 domain_type *wildcard_child = domain_wildcard_child(closest_encloser); 1144 #ifdef RATELIMIT 1145 q->wildcard_domain = wildcard_child; 1146 #endif 1147 1148 match = (domain_type *) region_alloc(q->region, 1149 sizeof(domain_type)); 1150 #ifdef USE_RADIX_TREE 1151 match->rnode = NULL; 1152 match->dname = wildcard_child->dname; 1153 #else 1154 memcpy(&match->node, &wildcard_child->node, sizeof(rbnode_type)); 1155 match->node.parent = NULL; 1156 #endif 1157 match->parent = closest_encloser; 1158 match->wildcard_child_closest_match = match; 1159 match->number = domain_number; 1160 match->rrsets = wildcard_child->rrsets; 1161 match->is_existing = wildcard_child->is_existing; 1162 #ifdef NSEC3 1163 match->nsec3 = wildcard_child->nsec3; 1164 /* copy over these entries: 1165 match->nsec3_is_exact = wildcard_child->nsec3_is_exact; 1166 match->nsec3_cover = wildcard_child->nsec3_cover; 1167 match->nsec3_wcard_child_cover = wildcard_child->nsec3_wcard_child_cover; 1168 match->nsec3_ds_parent_is_exact = wildcard_child->nsec3_ds_parent_is_exact; 1169 match->nsec3_ds_parent_cover = wildcard_child->nsec3_ds_parent_cover; 1170 */ 1171 1172 if (q->edns.dnssec_ok && q->zone->nsec3_param) { 1173 /* Only add nsec3 wildcard data when do bit is set */ 1174 nsec3_answer_wildcard(q, answer, wildcard_child, qname); 1175 } 1176 #endif 1177 1178 /* 1179 * Remember the original domain in case a Wildcard No 1180 * Data (3.1.3.4) response needs to be generated. In 1181 * this particular case the wildcard IS NOT 1182 * expanded. 1183 */ 1184 original = wildcard_child; 1185 } else { 1186 match = NULL; 1187 } 1188 1189 /* Authoritative zone. */ 1190 #ifdef NSEC3 1191 if (q->edns.dnssec_ok && q->zone->nsec3_param) { 1192 nsec3_answer_authoritative(&match, q, answer, 1193 closest_encloser, qname); 1194 } else 1195 #endif 1196 if (q->edns.dnssec_ok && zone_is_secure(q->zone)) { 1197 if (match != closest_encloser) { 1198 domain_type *nsec_domain; 1199 rrset_type *nsec_rrset; 1200 1201 /* 1202 * No match found or generated from wildcard, 1203 * include NSEC record. 1204 */ 1205 nsec_domain = find_covering_nsec(closest_match, q->zone, &nsec_rrset); 1206 if (nsec_domain) { 1207 add_rrset(q, answer, AUTHORITY_SECTION, nsec_domain, nsec_rrset); 1208 } 1209 } 1210 if (!match) { 1211 domain_type *nsec_domain; 1212 rrset_type *nsec_rrset; 1213 1214 /* 1215 * No match and no wildcard. Include NSEC 1216 * proving there is no wildcard. 1217 */ 1218 if(closest_encloser && (nsec_domain = 1219 find_covering_nsec(closest_encloser-> 1220 wildcard_child_closest_match, q->zone, 1221 &nsec_rrset)) != NULL) { 1222 add_rrset(q, answer, AUTHORITY_SECTION, nsec_domain, nsec_rrset); 1223 } 1224 } 1225 } 1226 1227 #ifdef NSEC3 1228 if (RCODE(q->packet)!=RCODE_OK) { 1229 return; /* nsec3 collision failure */ 1230 } 1231 #endif 1232 if (match) { 1233 answer_domain(nsd, q, answer, match, original); 1234 } else { 1235 answer_nxdomain(q, answer); 1236 } 1237 } 1238 1239 /* 1240 * qname may be different after CNAMEs have been followed from query->qname. 1241 */ 1242 static void 1243 answer_lookup_zone(struct nsd *nsd, struct query *q, answer_type *answer, 1244 size_t domain_number, int exact, domain_type *closest_match, 1245 domain_type *closest_encloser, const dname_type *qname) 1246 { 1247 zone_type* origzone = q->zone; 1248 q->zone = domain_find_zone(nsd->db, closest_encloser); 1249 if (!q->zone) { 1250 /* no zone for this */ 1251 if(q->cname_count == 0) 1252 RCODE_SET(q->packet, RCODE_REFUSE); 1253 return; 1254 } 1255 assert(closest_encloser); /* otherwise, no q->zone would be found */ 1256 if(!q->zone->apex || !q->zone->soa_rrset) { 1257 /* zone is configured but not loaded */ 1258 if(q->cname_count == 0) 1259 RCODE_SET(q->packet, RCODE_SERVFAIL); 1260 return; 1261 } 1262 1263 /* 1264 * If confine-to-zone is set to yes do not return additional 1265 * information for a zone with a different apex from the query zone. 1266 */ 1267 if (nsd->options->confine_to_zone && 1268 (origzone != NULL && dname_compare(domain_dname(origzone->apex), domain_dname(q->zone->apex)) != 0)) { 1269 return; 1270 } 1271 1272 /* now move up the closest encloser until it exists, previous 1273 * (possibly empty) closest encloser was useful to finding the zone 1274 * (for empty zones too), but now we want actual data nodes */ 1275 if (closest_encloser && !closest_encloser->is_existing) { 1276 exact = 0; 1277 while (closest_encloser != NULL && !closest_encloser->is_existing) 1278 closest_encloser = closest_encloser->parent; 1279 } 1280 1281 /* 1282 * See RFC 4035 (DNSSEC protocol) section 3.1.4.1 Responding 1283 * to Queries for DS RRs. 1284 */ 1285 if (exact && q->qtype == TYPE_DS && closest_encloser == q->zone->apex) { 1286 /* 1287 * Type DS query at a zone cut, use the responsible 1288 * parent zone to generate the answer if we are 1289 * authoritative for the parent zone. 1290 */ 1291 zone_type *zone = domain_find_parent_zone(nsd->db, q->zone); 1292 if (zone) { 1293 q->zone = zone; 1294 if(!q->zone->apex || !q->zone->soa_rrset) { 1295 /* zone is configured but not loaded */ 1296 if(q->cname_count == 0) 1297 RCODE_SET(q->packet, RCODE_SERVFAIL); 1298 return; 1299 } 1300 } 1301 } 1302 1303 /* see if the zone has expired (for secondary zones) */ 1304 if(q->zone && q->zone->opts && q->zone->opts->pattern && 1305 q->zone->opts->pattern->request_xfr != 0 && !q->zone->is_ok) { 1306 if(q->cname_count == 0) 1307 RCODE_SET(q->packet, RCODE_SERVFAIL); 1308 return; 1309 } 1310 1311 if (exact && q->qtype == TYPE_DS && closest_encloser == q->zone->apex) { 1312 /* 1313 * Type DS query at the zone apex (and the server is 1314 * not authoritative for the parent zone). 1315 */ 1316 if (q->qclass == CLASS_ANY) { 1317 AA_CLR(q->packet); 1318 } else { 1319 AA_SET(q->packet); 1320 } 1321 answer_nodata(q, answer, closest_encloser); 1322 } else { 1323 q->delegation_domain = domain_find_ns_rrsets( 1324 closest_encloser, q->zone, &q->delegation_rrset); 1325 if(q->delegation_domain && find_dname_above(q->delegation_domain, q->zone)) { 1326 q->delegation_domain = NULL; /* use higher DNAME */ 1327 } 1328 1329 if (!q->delegation_domain 1330 || (exact && q->qtype == TYPE_DS && closest_encloser == q->delegation_domain)) 1331 { 1332 if (q->qclass == CLASS_ANY) { 1333 AA_CLR(q->packet); 1334 } else { 1335 AA_SET(q->packet); 1336 } 1337 answer_authoritative(nsd, q, answer, domain_number, exact, 1338 closest_match, closest_encloser, qname); 1339 } 1340 else { 1341 answer_delegation(q, answer); 1342 } 1343 } 1344 } 1345 1346 static void 1347 answer_query(struct nsd *nsd, struct query *q) 1348 { 1349 domain_type *closest_match; 1350 domain_type *closest_encloser; 1351 int exact; 1352 uint16_t offset; 1353 answer_type answer; 1354 1355 answer_init(&answer); 1356 1357 exact = namedb_lookup(nsd->db, q->qname, &closest_match, &closest_encloser); 1358 1359 answer_lookup_zone(nsd, q, &answer, 0, exact, closest_match, 1360 closest_encloser, q->qname); 1361 ZTATUP2(nsd, q->zone, opcode, q->opcode); 1362 ZTATUP2(nsd, q->zone, qtype, q->qtype); 1363 ZTATUP2(nsd, q->zone, qclass, q->qclass); 1364 1365 offset = dname_label_offsets(q->qname)[domain_dname(closest_encloser)->label_count - 1] + QHEADERSZ; 1366 query_add_compression_domain(q, closest_encloser, offset); 1367 encode_answer(q, &answer); 1368 query_clear_compression_tables(q); 1369 } 1370 1371 void 1372 query_prepare_response(query_type *q) 1373 { 1374 uint16_t flags; 1375 1376 /* 1377 * Preserve the data up-to the current packet's limit. 1378 */ 1379 buffer_set_position(q->packet, buffer_limit(q->packet)); 1380 buffer_set_limit(q->packet, buffer_capacity(q->packet)); 1381 1382 /* 1383 * Reserve space for the EDNS records if required. 1384 */ 1385 q->reserved_space = edns_reserved_space(&q->edns); 1386 q->reserved_space += tsig_reserved_space(&q->tsig); 1387 1388 /* Update the flags. */ 1389 flags = FLAGS(q->packet); 1390 flags &= 0x0100U; /* Preserve the RD flag. */ 1391 /* CD flag must be cleared for auth answers */ 1392 flags |= 0x8000U; /* Set the QR flag. */ 1393 FLAGS_SET(q->packet, flags); 1394 } 1395 1396 /* 1397 * Processes the query. 1398 * 1399 */ 1400 query_state_type 1401 query_process(query_type *q, nsd_type *nsd) 1402 { 1403 /* The query... */ 1404 nsd_rc_type rc; 1405 query_state_type query_state; 1406 uint16_t arcount; 1407 1408 /* Sanity checks */ 1409 if (buffer_limit(q->packet) < QHEADERSZ) { 1410 /* packet too small to contain DNS header. 1411 Now packet investigation macros will work without problems. */ 1412 return QUERY_DISCARDED; 1413 } 1414 if (QR(q->packet)) { 1415 /* Not a query? Drop it on the floor. */ 1416 return QUERY_DISCARDED; 1417 } 1418 1419 /* check opcode early on, because new opcodes may have different 1420 * specification of the meaning of the rest of the packet */ 1421 q->opcode = OPCODE(q->packet); 1422 if(q->opcode != OPCODE_QUERY && q->opcode != OPCODE_NOTIFY) { 1423 if(query_ratelimit_err(nsd)) 1424 return QUERY_DISCARDED; 1425 if(nsd->options->drop_updates && q->opcode == OPCODE_UPDATE) 1426 return QUERY_DISCARDED; 1427 return query_error(q, NSD_RC_IMPL); 1428 } 1429 1430 if (RCODE(q->packet) != RCODE_OK || !process_query_section(q)) { 1431 return query_formerr(q, nsd); 1432 } 1433 1434 /* Update statistics. */ 1435 STATUP2(nsd, opcode, q->opcode); 1436 STATUP2(nsd, qtype, q->qtype); 1437 STATUP2(nsd, qclass, q->qclass); 1438 1439 if (q->opcode != OPCODE_QUERY) { 1440 if (q->opcode == OPCODE_NOTIFY) { 1441 return answer_notify(nsd, q); 1442 } else { 1443 if(query_ratelimit_err(nsd)) 1444 return QUERY_DISCARDED; 1445 return query_error(q, NSD_RC_IMPL); 1446 } 1447 } 1448 1449 /* Dont bother to answer more than one question at once... */ 1450 if (QDCOUNT(q->packet) != 1) { 1451 if(QDCOUNT(q->packet) == 0 && ANCOUNT(q->packet) == 0 && 1452 NSCOUNT(q->packet) == 0 && ARCOUNT(q->packet) == 1 && 1453 buffer_limit(q->packet) >= QHEADERSZ+OPT_LEN+ 1454 OPT_RDATA) { 1455 /* add edns section to answer */ 1456 buffer_set_position(q->packet, QHEADERSZ); 1457 if (edns_parse_record(&q->edns, q->packet, q, nsd)) { 1458 if(process_edns(nsd, q) == NSD_RC_OK) { 1459 int opcode = OPCODE(q->packet); 1460 (void)query_error(q, NSD_RC_FORMAT); 1461 query_add_optional(q, nsd); 1462 FLAGS_SET(q->packet, FLAGS(q->packet) & 0x0100U); 1463 /* Preserve the RD flag. Clear the rest. */ 1464 OPCODE_SET(q->packet, opcode); 1465 QR_SET(q->packet); 1466 return QUERY_PROCESSED; 1467 } 1468 } 1469 } 1470 FLAGS_SET(q->packet, 0); 1471 return query_formerr(q, nsd); 1472 } 1473 /* Ignore settings of flags */ 1474 1475 /* Dont allow any records in the answer or authority section... 1476 except for IXFR queries. */ 1477 if (ANCOUNT(q->packet) != 0 || 1478 (q->qtype!=TYPE_IXFR && NSCOUNT(q->packet) != 0)) { 1479 return query_formerr(q, nsd); 1480 } 1481 if(q->qtype==TYPE_IXFR && NSCOUNT(q->packet) > 0) { 1482 unsigned int i; /* skip ixfr soa information data here */ 1483 unsigned int nscount = (unsigned)NSCOUNT(q->packet); 1484 /* define a bound on the number of extraneous records allowed, 1485 * we expect 1, a SOA serial record, and no more. 1486 * perhaps RRSIGs (but not needed), otherwise we do not 1487 * understand what this means. We do not want too many 1488 * because the high iteration counts slow down. */ 1489 if(nscount > 64) return query_formerr(q, nsd); 1490 for(i=0; i< nscount; i++) 1491 if(!packet_skip_rr(q->packet, 0)) 1492 return query_formerr(q, nsd); 1493 } 1494 1495 arcount = ARCOUNT(q->packet); 1496 if (arcount > 0) { 1497 /* According to draft-ietf-dnsext-rfc2671bis-edns0-10: 1498 * "The placement flexibility for the OPT RR does not 1499 * override the need for the TSIG or SIG(0) RRs to be 1500 * the last in the additional section whenever they are 1501 * present." 1502 * So we should not have to check for TSIG RR before 1503 * OPT RR. Keep the code for backwards compatibility. 1504 */ 1505 1506 /* see if tsig is before edns record */ 1507 if (!tsig_parse_rr(&q->tsig, q->packet)) 1508 return query_formerr(q, nsd); 1509 if(q->tsig.status != TSIG_NOT_PRESENT) 1510 --arcount; 1511 } 1512 /* See if there is an OPT RR. */ 1513 if (arcount > 0) { 1514 if (edns_parse_record(&q->edns, q->packet, q, nsd)) 1515 --arcount; 1516 } 1517 /* See if there is a TSIG RR. */ 1518 if (arcount > 0 && q->tsig.status == TSIG_NOT_PRESENT) { 1519 /* see if tsig is after the edns record */ 1520 if (!tsig_parse_rr(&q->tsig, q->packet)) 1521 return query_formerr(q, nsd); 1522 if(q->tsig.status != TSIG_NOT_PRESENT) 1523 --arcount; 1524 } 1525 /* If more RRs left in Add. Section, FORMERR. */ 1526 if (arcount > 0) { 1527 return query_formerr(q, nsd); 1528 } 1529 1530 /* Do we have any trailing garbage? */ 1531 #ifdef STRICT_MESSAGE_PARSE 1532 if (buffer_remaining(q->packet) > 0) { 1533 /* If we're strict.... */ 1534 return query_formerr(q, nsd); 1535 } 1536 #endif 1537 /* Remove trailing garbage. */ 1538 buffer_set_limit(q->packet, buffer_position(q->packet)); 1539 1540 rc = process_tsig(q); 1541 if (rc != NSD_RC_OK) { 1542 return query_error(q, rc); 1543 } 1544 rc = process_edns(nsd, q); 1545 if (rc != NSD_RC_OK) { 1546 /* We should not return FORMERR, but BADVERS (=16). 1547 * BADVERS is created with Ext. RCODE, followed by RCODE. 1548 * Ext. RCODE is set to 1, RCODE must be 0 (getting 0x10 = 16). 1549 * Thus RCODE = NOERROR = NSD_RC_OK. */ 1550 RCODE_SET(q->packet, NSD_RC_OK); 1551 buffer_clear(q->packet); 1552 buffer_set_position(q->packet, 1553 QHEADERSZ + 4 + q->qname->name_size); 1554 QR_SET(q->packet); 1555 AD_CLR(q->packet); 1556 QDCOUNT_SET(q->packet, 1); 1557 ANCOUNT_SET(q->packet, 0); 1558 NSCOUNT_SET(q->packet, 0); 1559 ARCOUNT_SET(q->packet, 0); 1560 return QUERY_PROCESSED; 1561 } 1562 1563 query_prepare_response(q); 1564 1565 if (q->qclass != CLASS_IN && q->qclass != CLASS_ANY) { 1566 if (q->qclass == CLASS_CH) { 1567 return answer_chaos(nsd, q); 1568 } else { 1569 return query_error(q, NSD_RC_REFUSE); 1570 } 1571 } 1572 1573 query_state = answer_axfr_ixfr(nsd, q); 1574 if (query_state == QUERY_PROCESSED || query_state == QUERY_IN_AXFR) { 1575 return query_state; 1576 } 1577 if(q->qtype == TYPE_ANY && nsd->options->refuse_any && !q->tcp) { 1578 TC_SET(q->packet); 1579 return query_error(q, NSD_RC_OK); 1580 } 1581 1582 answer_query(nsd, q); 1583 1584 return QUERY_PROCESSED; 1585 } 1586 1587 void 1588 query_add_optional(query_type *q, nsd_type *nsd) 1589 { 1590 struct edns_data *edns = &nsd->edns_ipv4; 1591 #if defined(INET6) 1592 if (q->addr.ss_family == AF_INET6) { 1593 edns = &nsd->edns_ipv6; 1594 } 1595 #endif 1596 if (RCODE(q->packet) == RCODE_FORMAT) { 1597 return; 1598 } 1599 switch (q->edns.status) { 1600 case EDNS_NOT_PRESENT: 1601 break; 1602 case EDNS_OK: 1603 if (q->edns.dnssec_ok) edns->ok[7] = 0x80; 1604 else edns->ok[7] = 0x00; 1605 buffer_write(q->packet, edns->ok, OPT_LEN); 1606 if(q->edns.opt_reserved_space == 0 || !buffer_available( 1607 q->packet, 2+q->edns.opt_reserved_space)) { 1608 /* fill with NULLs */ 1609 buffer_write(q->packet, edns->rdata_none, OPT_RDATA); 1610 } else { 1611 /* rdata length */ 1612 buffer_write_u16(q->packet, q->edns.opt_reserved_space); 1613 /* edns options */ 1614 if(q->edns.nsid) { 1615 /* nsid opt header */ 1616 buffer_write(q->packet, edns->nsid, OPT_HDR); 1617 /* nsid payload */ 1618 buffer_write(q->packet, nsd->nsid, nsd->nsid_len); 1619 } 1620 } 1621 ARCOUNT_SET(q->packet, ARCOUNT(q->packet) + 1); 1622 STATUP(nsd, edns); 1623 ZTATUP(nsd, q->zone, edns); 1624 break; 1625 case EDNS_ERROR: 1626 if (q->edns.dnssec_ok) edns->error[7] = 0x80; 1627 else edns->error[7] = 0x00; 1628 buffer_write(q->packet, edns->error, OPT_LEN); 1629 buffer_write(q->packet, edns->rdata_none, OPT_RDATA); 1630 ARCOUNT_SET(q->packet, ARCOUNT(q->packet) + 1); 1631 STATUP(nsd, ednserr); 1632 ZTATUP(nsd, q->zone, ednserr); 1633 break; 1634 } 1635 1636 if (q->tsig.status != TSIG_NOT_PRESENT) { 1637 if (q->tsig.status == TSIG_ERROR || 1638 q->tsig.error_code != TSIG_ERROR_NOERROR) { 1639 tsig_error_reply(&q->tsig); 1640 tsig_append_rr(&q->tsig, q->packet); 1641 ARCOUNT_SET(q->packet, ARCOUNT(q->packet) + 1); 1642 } else if(q->tsig.status == TSIG_OK && 1643 q->tsig.error_code == TSIG_ERROR_NOERROR) 1644 { 1645 if(q->tsig_prepare_it) 1646 tsig_prepare(&q->tsig); 1647 if(q->tsig_update_it) 1648 tsig_update(&q->tsig, q->packet, buffer_position(q->packet)); 1649 if(q->tsig_sign_it) { 1650 tsig_sign(&q->tsig); 1651 tsig_append_rr(&q->tsig, q->packet); 1652 ARCOUNT_SET(q->packet, ARCOUNT(q->packet) + 1); 1653 } 1654 } 1655 } 1656 } 1657