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 /* Add ID */ 534 query_addtxt(q, 535 buffer_begin(q->packet) + QHEADERSZ, 536 CLASS_CH, 537 0, 538 nsd->identity); 539 ANCOUNT_SET(q->packet, ANCOUNT(q->packet) + 1); 540 } else if ((q->qname->name_size == 16 541 && memcmp(dname_name(q->qname), "\007version\006server", 16) == 0) || 542 (q->qname->name_size == 14 543 && memcmp(dname_name(q->qname), "\007version\004bind", 14) == 0)) 544 { 545 if(!nsd->options->hide_version) { 546 /* Add version */ 547 query_addtxt(q, 548 buffer_begin(q->packet) + QHEADERSZ, 549 CLASS_CH, 550 0, 551 nsd->version); 552 ANCOUNT_SET(q->packet, ANCOUNT(q->packet) + 1); 553 } else { 554 RCODE_SET(q->packet, RCODE_REFUSE); 555 } 556 } else { 557 RCODE_SET(q->packet, RCODE_REFUSE); 558 } 559 break; 560 default: 561 RCODE_SET(q->packet, RCODE_REFUSE); 562 break; 563 } 564 565 return QUERY_PROCESSED; 566 } 567 568 569 /* 570 * Find the covering NSEC for a non-existent domain name. Normally 571 * the NSEC will be located at CLOSEST_MATCH, except when it is an 572 * empty non-terminal. In this case the NSEC may be located at the 573 * previous domain name (in canonical ordering). 574 */ 575 static domain_type * 576 find_covering_nsec(domain_type *closest_match, 577 zone_type *zone, 578 rrset_type **nsec_rrset) 579 { 580 assert(closest_match); 581 assert(nsec_rrset); 582 583 /* loop away temporary created domains. For real ones it is &RBTREE_NULL */ 584 #ifdef USE_RADIX_TREE 585 while (closest_match->rnode == NULL) 586 #else 587 while (closest_match->node.parent == NULL) 588 #endif 589 closest_match = closest_match->parent; 590 while (closest_match) { 591 *nsec_rrset = domain_find_rrset(closest_match, zone, TYPE_NSEC); 592 if (*nsec_rrset) { 593 return closest_match; 594 } 595 if (closest_match == zone->apex) { 596 /* Don't look outside the current zone. */ 597 return NULL; 598 } 599 closest_match = domain_previous(closest_match); 600 } 601 return NULL; 602 } 603 604 605 struct additional_rr_types 606 { 607 uint16_t rr_type; 608 rr_section_type rr_section; 609 }; 610 611 struct additional_rr_types default_additional_rr_types[] = { 612 { TYPE_A, ADDITIONAL_A_SECTION }, 613 { TYPE_AAAA, ADDITIONAL_AAAA_SECTION }, 614 { 0, (rr_section_type) 0 } 615 }; 616 617 struct additional_rr_types swap_aaaa_additional_rr_types[] = { 618 { TYPE_AAAA, ADDITIONAL_A_SECTION }, 619 { TYPE_A, ADDITIONAL_AAAA_SECTION }, 620 { 0, (rr_section_type) 0 } 621 }; 622 623 struct additional_rr_types rt_additional_rr_types[] = { 624 { TYPE_A, ADDITIONAL_A_SECTION }, 625 { TYPE_AAAA, ADDITIONAL_AAAA_SECTION }, 626 { TYPE_X25, ADDITIONAL_OTHER_SECTION }, 627 { TYPE_ISDN, ADDITIONAL_OTHER_SECTION }, 628 { 0, (rr_section_type) 0 } 629 }; 630 631 static void 632 add_additional_rrsets(struct query *query, answer_type *answer, 633 rrset_type *master_rrset, size_t rdata_index, 634 int allow_glue, struct additional_rr_types types[]) 635 { 636 size_t i; 637 638 assert(query); 639 assert(answer); 640 assert(master_rrset); 641 assert(rdata_atom_is_domain(rrset_rrtype(master_rrset), rdata_index)); 642 643 for (i = 0; i < master_rrset->rr_count; ++i) { 644 int j; 645 domain_type *additional = rdata_atom_domain(master_rrset->rrs[i].rdatas[rdata_index]); 646 domain_type *match = additional; 647 648 assert(additional); 649 650 if (!allow_glue && domain_is_glue(match, query->zone)) 651 continue; 652 653 /* 654 * Check to see if we need to generate the dependent 655 * based on a wildcard domain. 656 */ 657 while (!match->is_existing) { 658 match = match->parent; 659 } 660 if (additional != match && domain_wildcard_child(match)) { 661 domain_type *wildcard_child = domain_wildcard_child(match); 662 domain_type *temp = (domain_type *) region_alloc( 663 query->region, sizeof(domain_type)); 664 #ifdef USE_RADIX_TREE 665 temp->rnode = NULL; 666 temp->dname = additional->dname; 667 #else 668 memcpy(&temp->node, &additional->node, sizeof(rbnode_type)); 669 temp->node.parent = NULL; 670 #endif 671 temp->number = additional->number; 672 temp->parent = match; 673 temp->wildcard_child_closest_match = temp; 674 temp->rrsets = wildcard_child->rrsets; 675 temp->is_existing = wildcard_child->is_existing; 676 additional = temp; 677 } 678 679 for (j = 0; types[j].rr_type != 0; ++j) { 680 rrset_type *rrset = domain_find_rrset( 681 additional, query->zone, types[j].rr_type); 682 if (rrset) { 683 answer_add_rrset(answer, types[j].rr_section, 684 additional, rrset); 685 } 686 } 687 } 688 } 689 690 static int 691 answer_needs_ns(struct query* query) 692 { 693 assert(query); 694 /* Currently, only troublesome for DNSKEY and DS, 695 * cuz their RRSETs are quite large. */ 696 return (query->qtype != TYPE_DNSKEY && query->qtype != TYPE_DS 697 && query->qtype != TYPE_ANY); 698 } 699 700 static int 701 add_rrset(struct query *query, 702 answer_type *answer, 703 rr_section_type section, 704 domain_type *owner, 705 rrset_type *rrset) 706 { 707 int result; 708 709 assert(query); 710 assert(answer); 711 assert(owner); 712 assert(rrset); 713 assert(rrset_rrclass(rrset) == CLASS_IN); 714 715 result = answer_add_rrset(answer, section, owner, rrset); 716 if(minimal_responses && section != AUTHORITY_SECTION && 717 query->qtype != TYPE_NS) 718 return result; 719 switch (rrset_rrtype(rrset)) { 720 case TYPE_NS: 721 #if defined(INET6) 722 /* if query over IPv6, swap A and AAAA; put AAAA first */ 723 add_additional_rrsets(query, answer, rrset, 0, 1, 724 (query->addr.ss_family == AF_INET6)? 725 swap_aaaa_additional_rr_types: 726 default_additional_rr_types); 727 #else 728 add_additional_rrsets(query, answer, rrset, 0, 1, 729 default_additional_rr_types); 730 #endif 731 break; 732 case TYPE_MB: 733 add_additional_rrsets(query, answer, rrset, 0, 0, 734 default_additional_rr_types); 735 break; 736 case TYPE_MX: 737 case TYPE_KX: 738 add_additional_rrsets(query, answer, rrset, 1, 0, 739 default_additional_rr_types); 740 break; 741 case TYPE_RT: 742 add_additional_rrsets(query, answer, rrset, 1, 0, 743 rt_additional_rr_types); 744 break; 745 case TYPE_SRV: 746 add_additional_rrsets(query, answer, rrset, 3, 0, 747 default_additional_rr_types); 748 break; 749 default: 750 break; 751 } 752 753 return result; 754 } 755 756 757 /* returns 0 on error, or the domain number for to_name. 758 from_name is changes to to_name by the DNAME rr. 759 DNAME rr is from src to dest. 760 closest encloser encloses the to_name. */ 761 static size_t 762 query_synthesize_cname(struct query* q, struct answer* answer, const dname_type* from_name, 763 const dname_type* to_name, domain_type* src, domain_type* to_closest_encloser, 764 domain_type** to_closest_match, uint32_t ttl) 765 { 766 /* add temporary domains for from_name and to_name and all 767 their (not allocated yet) parents */ 768 /* any domains below src are not_existing (because of DNAME at src) */ 769 int i; 770 domain_type* cname_domain; 771 domain_type* cname_dest; 772 rrset_type* rrset; 773 774 /* allocate source part */ 775 domain_type* lastparent = src; 776 assert(q && answer && from_name && to_name && src && to_closest_encloser); 777 assert(to_closest_match); 778 for(i=0; i < from_name->label_count - domain_dname(src)->label_count; i++) 779 { 780 domain_type* newdom = query_get_tempdomain(q); 781 if(!newdom) 782 return 0; 783 newdom->is_existing = 1; 784 newdom->parent = lastparent; 785 #ifdef USE_RADIX_TREE 786 newdom->dname 787 #else 788 newdom->node.key 789 #endif 790 = dname_partial_copy(q->region, 791 from_name, domain_dname(src)->label_count + i + 1); 792 if(dname_compare(domain_dname(newdom), q->qname) == 0) { 793 /* 0 good for query name, otherwise new number */ 794 newdom->number = 0; 795 } 796 DEBUG(DEBUG_QUERY,2, (LOG_INFO, "created temp domain src %d. %s nr %d", i, 797 domain_to_string(newdom), (int)newdom->number)); 798 lastparent = newdom; 799 } 800 cname_domain = lastparent; 801 802 /* allocate dest part */ 803 lastparent = to_closest_encloser; 804 for(i=0; i < to_name->label_count - domain_dname(to_closest_encloser)->label_count; 805 i++) 806 { 807 domain_type* newdom = query_get_tempdomain(q); 808 if(!newdom) 809 return 0; 810 newdom->is_existing = 0; 811 newdom->parent = lastparent; 812 #ifdef USE_RADIX_TREE 813 newdom->dname 814 #else 815 newdom->node.key 816 #endif 817 = dname_partial_copy(q->region, 818 to_name, domain_dname(to_closest_encloser)->label_count + i + 1); 819 DEBUG(DEBUG_QUERY,2, (LOG_INFO, "created temp domain dest %d. %s nr %d", i, 820 domain_to_string(newdom), (int)newdom->number)); 821 lastparent = newdom; 822 } 823 cname_dest = lastparent; 824 *to_closest_match = cname_dest; 825 826 /* allocate the CNAME RR */ 827 rrset = (rrset_type*) region_alloc(q->region, sizeof(rrset_type)); 828 memset(rrset, 0, sizeof(rrset_type)); 829 rrset->zone = q->zone; 830 rrset->rr_count = 1; 831 rrset->rrs = (rr_type*) region_alloc(q->region, sizeof(rr_type)); 832 memset(rrset->rrs, 0, sizeof(rr_type)); 833 rrset->rrs->owner = cname_domain; 834 rrset->rrs->ttl = ttl; 835 rrset->rrs->type = TYPE_CNAME; 836 rrset->rrs->klass = CLASS_IN; 837 rrset->rrs->rdata_count = 1; 838 rrset->rrs->rdatas = (rdata_atom_type*)region_alloc(q->region, 839 sizeof(rdata_atom_type)); 840 rrset->rrs->rdatas->domain = cname_dest; 841 842 if(!add_rrset(q, answer, ANSWER_SECTION, cname_domain, rrset)) { 843 log_msg(LOG_ERR, "could not add synthesized CNAME rrset to packet"); 844 } 845 846 return cname_dest->number; 847 } 848 849 /* 850 * Answer delegation information. 851 * 852 * DNSSEC: Include the DS RRset if present. Otherwise include an NSEC 853 * record proving the DS RRset does not exist. 854 */ 855 static void 856 answer_delegation(query_type *query, answer_type *answer) 857 { 858 assert(answer); 859 assert(query->delegation_domain); 860 assert(query->delegation_rrset); 861 862 if (query->cname_count == 0) { 863 AA_CLR(query->packet); 864 } else { 865 AA_SET(query->packet); 866 } 867 868 add_rrset(query, 869 answer, 870 AUTHORITY_SECTION, 871 query->delegation_domain, 872 query->delegation_rrset); 873 if (query->edns.dnssec_ok && zone_is_secure(query->zone)) { 874 rrset_type *rrset; 875 if ((rrset = domain_find_rrset(query->delegation_domain, query->zone, TYPE_DS))) { 876 add_rrset(query, answer, AUTHORITY_SECTION, 877 query->delegation_domain, rrset); 878 #ifdef NSEC3 879 } else if (query->zone->nsec3_param) { 880 nsec3_answer_delegation(query, answer); 881 #endif 882 } else if ((rrset = domain_find_rrset(query->delegation_domain, query->zone, TYPE_NSEC))) { 883 add_rrset(query, answer, AUTHORITY_SECTION, 884 query->delegation_domain, rrset); 885 } 886 } 887 } 888 889 890 /* 891 * Answer SOA information. 892 */ 893 static void 894 answer_soa(struct query *query, answer_type *answer) 895 { 896 if (query->qclass != CLASS_ANY) { 897 add_rrset(query, answer, 898 AUTHORITY_SECTION, 899 query->zone->apex, 900 query->zone->soa_nx_rrset); 901 } 902 } 903 904 905 /* 906 * Answer that the domain name exists but there is no RRset with the 907 * requested type. 908 * 909 * DNSSEC: Include the correct NSEC record proving that the type does 910 * not exist. In the wildcard no data (3.1.3.4) case the wildcard IS 911 * NOT expanded, so the ORIGINAL parameter must point to the original 912 * wildcard entry, not to the generated entry. 913 */ 914 static void 915 answer_nodata(struct query *query, answer_type *answer, domain_type *original) 916 { 917 if (query->cname_count == 0) { 918 answer_soa(query, answer); 919 } 920 921 #ifdef NSEC3 922 if (query->edns.dnssec_ok && query->zone->nsec3_param) { 923 nsec3_answer_nodata(query, answer, original); 924 } else 925 #endif 926 if (query->edns.dnssec_ok && zone_is_secure(query->zone)) { 927 domain_type *nsec_domain; 928 rrset_type *nsec_rrset; 929 930 nsec_domain = find_covering_nsec(original, query->zone, &nsec_rrset); 931 if (nsec_domain) { 932 add_rrset(query, answer, AUTHORITY_SECTION, nsec_domain, nsec_rrset); 933 } 934 } 935 } 936 937 static void 938 answer_nxdomain(query_type *query, answer_type *answer) 939 { 940 RCODE_SET(query->packet, RCODE_NXDOMAIN); 941 answer_soa(query, answer); 942 } 943 944 945 /* 946 * Answer domain information (or SOA if we do not have an RRset for 947 * the type specified by the query). 948 */ 949 static void 950 answer_domain(struct nsd* nsd, struct query *q, answer_type *answer, 951 domain_type *domain, domain_type *original) 952 { 953 rrset_type *rrset; 954 955 if (q->qtype == TYPE_ANY) { 956 int added = 0; 957 for (rrset = domain_find_any_rrset(domain, q->zone); rrset; rrset = rrset->next) { 958 if (rrset->zone == q->zone 959 #ifdef NSEC3 960 && rrset_rrtype(rrset) != TYPE_NSEC3 961 #endif 962 /* 963 * Don't include the RRSIG RRset when 964 * DNSSEC is used, because it is added 965 * automatically on an per-RRset basis. 966 */ 967 && !(q->edns.dnssec_ok 968 && zone_is_secure(q->zone) 969 && rrset_rrtype(rrset) == TYPE_RRSIG)) 970 { 971 add_rrset(q, answer, ANSWER_SECTION, domain, rrset); 972 ++added; 973 #ifdef NOTYET 974 /* minimize response size with one RR, 975 * according to RFC 8482(4.1). */ 976 break; 977 #endif 978 } 979 } 980 if (added == 0) { 981 answer_nodata(q, answer, original); 982 return; 983 } 984 #ifdef NSEC3 985 } else if (q->qtype == TYPE_NSEC3) { 986 answer_nodata(q, answer, original); 987 return; 988 #endif 989 } else if ((rrset = domain_find_rrset(domain, q->zone, q->qtype))) { 990 add_rrset(q, answer, ANSWER_SECTION, domain, rrset); 991 } else if ((rrset = domain_find_rrset(domain, q->zone, TYPE_CNAME))) { 992 int added; 993 994 /* 995 * If the CNAME is not added it is already in the 996 * answer, so we have a CNAME loop. Don't follow the 997 * CNAME target in this case. 998 */ 999 added = add_rrset(q, answer, ANSWER_SECTION, domain, rrset); 1000 assert(rrset->rr_count > 0); 1001 if (added) { 1002 /* only process first CNAME record */ 1003 domain_type *closest_match = rdata_atom_domain(rrset->rrs[0].rdatas[0]); 1004 domain_type *closest_encloser = closest_match; 1005 zone_type* origzone = q->zone; 1006 ++q->cname_count; 1007 1008 answer_lookup_zone(nsd, q, answer, closest_match->number, 1009 closest_match == closest_encloser, 1010 closest_match, closest_encloser, 1011 domain_dname(closest_match)); 1012 q->zone = origzone; 1013 } 1014 return; 1015 } else { 1016 answer_nodata(q, answer, original); 1017 return; 1018 } 1019 1020 if (q->qclass != CLASS_ANY && q->zone->ns_rrset && answer_needs_ns(q) 1021 && !minimal_responses) { 1022 add_rrset(q, answer, OPTIONAL_AUTHORITY_SECTION, q->zone->apex, 1023 q->zone->ns_rrset); 1024 } 1025 } 1026 1027 1028 /* 1029 * Answer with authoritative data. If a wildcard is matched the owner 1030 * name will be expanded to the domain name specified by 1031 * DOMAIN_NUMBER. DOMAIN_NUMBER 0 (zero) is reserved for the original 1032 * query name. 1033 * 1034 * DNSSEC: Include the necessary NSEC records in case the request 1035 * domain name does not exist and/or a wildcard match does not exist. 1036 */ 1037 static void 1038 answer_authoritative(struct nsd *nsd, 1039 struct query *q, 1040 answer_type *answer, 1041 size_t domain_number, 1042 int exact, 1043 domain_type *closest_match, 1044 domain_type *closest_encloser, 1045 const dname_type *qname) 1046 { 1047 domain_type *match; 1048 domain_type *original = closest_match; 1049 domain_type *dname_ce; 1050 rrset_type *rrset; 1051 1052 #ifdef NSEC3 1053 if(exact && domain_has_only_NSEC3(closest_match, q->zone)) { 1054 exact = 0; /* pretend it does not exist */ 1055 if(closest_encloser->parent) 1056 closest_encloser = closest_encloser->parent; 1057 } 1058 #endif /* NSEC3 */ 1059 if((dname_ce = find_dname_above(closest_encloser, q->zone)) != NULL) { 1060 /* occlude the found data, the DNAME is closest_encloser */ 1061 closest_encloser = dname_ce; 1062 exact = 0; 1063 } 1064 1065 if (exact) { 1066 match = closest_match; 1067 } else if ((rrset=domain_find_rrset(closest_encloser, q->zone, TYPE_DNAME))) { 1068 /* process DNAME */ 1069 const dname_type* name = qname; 1070 domain_type *dest = rdata_atom_domain(rrset->rrs[0].rdatas[0]); 1071 int added; 1072 assert(rrset->rr_count > 0); 1073 if(domain_number != 0) /* we followed CNAMEs or DNAMEs */ 1074 name = domain_dname(closest_match); 1075 DEBUG(DEBUG_QUERY,2, (LOG_INFO, "expanding DNAME for q=%s", dname_to_string(name, NULL))); 1076 DEBUG(DEBUG_QUERY,2, (LOG_INFO, "->src is %s", 1077 domain_to_string(closest_encloser))); 1078 DEBUG(DEBUG_QUERY,2, (LOG_INFO, "->dest is %s", 1079 domain_to_string(dest))); 1080 /* if the DNAME set is not added we have a loop, do not follow */ 1081 added = add_rrset(q, answer, ANSWER_SECTION, closest_encloser, rrset); 1082 if(added) { 1083 domain_type* src = closest_encloser; 1084 const dname_type* newname = dname_replace(q->region, name, 1085 domain_dname(src), domain_dname(dest)); 1086 size_t newnum = 0; 1087 zone_type* origzone = q->zone; 1088 ++q->cname_count; 1089 if(!newname) { /* newname too long */ 1090 RCODE_SET(q->packet, RCODE_YXDOMAIN); 1091 return; 1092 } 1093 DEBUG(DEBUG_QUERY,2, (LOG_INFO, "->result is %s", dname_to_string(newname, NULL))); 1094 /* follow the DNAME */ 1095 (void)namedb_lookup(nsd->db, newname, &closest_match, &closest_encloser); 1096 /* synthesize CNAME record */ 1097 newnum = query_synthesize_cname(q, answer, name, newname, 1098 src, closest_encloser, &closest_match, rrset->rrs[0].ttl); 1099 if(!newnum) { 1100 /* could not synthesize the CNAME. */ 1101 /* return previous CNAMEs to make resolver recurse for us */ 1102 return; 1103 } 1104 1105 answer_lookup_zone(nsd, q, answer, newnum, 1106 closest_match == closest_encloser, 1107 closest_match, closest_encloser, newname); 1108 q->zone = origzone; 1109 } 1110 if(!added) /* log the error so operator can find looping recursors */ 1111 log_msg(LOG_INFO, "DNAME processing stopped due to loop, qname %s", 1112 dname_to_string(q->qname, NULL)); 1113 return; 1114 } else if (domain_wildcard_child(closest_encloser)) { 1115 /* Generate the domain from the wildcard. */ 1116 domain_type *wildcard_child = domain_wildcard_child(closest_encloser); 1117 #ifdef RATELIMIT 1118 q->wildcard_domain = wildcard_child; 1119 #endif 1120 1121 match = (domain_type *) region_alloc(q->region, 1122 sizeof(domain_type)); 1123 #ifdef USE_RADIX_TREE 1124 match->rnode = NULL; 1125 match->dname = wildcard_child->dname; 1126 #else 1127 memcpy(&match->node, &wildcard_child->node, sizeof(rbnode_type)); 1128 match->node.parent = NULL; 1129 #endif 1130 match->parent = closest_encloser; 1131 match->wildcard_child_closest_match = match; 1132 match->number = domain_number; 1133 match->rrsets = wildcard_child->rrsets; 1134 match->is_existing = wildcard_child->is_existing; 1135 #ifdef NSEC3 1136 match->nsec3 = wildcard_child->nsec3; 1137 /* copy over these entries: 1138 match->nsec3_is_exact = wildcard_child->nsec3_is_exact; 1139 match->nsec3_cover = wildcard_child->nsec3_cover; 1140 match->nsec3_wcard_child_cover = wildcard_child->nsec3_wcard_child_cover; 1141 match->nsec3_ds_parent_is_exact = wildcard_child->nsec3_ds_parent_is_exact; 1142 match->nsec3_ds_parent_cover = wildcard_child->nsec3_ds_parent_cover; 1143 */ 1144 1145 if (q->edns.dnssec_ok && q->zone->nsec3_param) { 1146 /* Only add nsec3 wildcard data when do bit is set */ 1147 nsec3_answer_wildcard(q, answer, wildcard_child, qname); 1148 } 1149 #endif 1150 1151 /* 1152 * Remember the original domain in case a Wildcard No 1153 * Data (3.1.3.4) response needs to be generated. In 1154 * this particular case the wildcard IS NOT 1155 * expanded. 1156 */ 1157 original = wildcard_child; 1158 } else { 1159 match = NULL; 1160 } 1161 1162 /* Authoritative zone. */ 1163 #ifdef NSEC3 1164 if (q->edns.dnssec_ok && q->zone->nsec3_param) { 1165 nsec3_answer_authoritative(&match, q, answer, 1166 closest_encloser, qname); 1167 } else 1168 #endif 1169 if (q->edns.dnssec_ok && zone_is_secure(q->zone)) { 1170 if (match != closest_encloser) { 1171 domain_type *nsec_domain; 1172 rrset_type *nsec_rrset; 1173 1174 /* 1175 * No match found or generated from wildcard, 1176 * include NSEC record. 1177 */ 1178 nsec_domain = find_covering_nsec(closest_match, q->zone, &nsec_rrset); 1179 if (nsec_domain) { 1180 add_rrset(q, answer, AUTHORITY_SECTION, nsec_domain, nsec_rrset); 1181 } 1182 } 1183 if (!match) { 1184 domain_type *nsec_domain; 1185 rrset_type *nsec_rrset; 1186 1187 /* 1188 * No match and no wildcard. Include NSEC 1189 * proving there is no wildcard. 1190 */ 1191 if(closest_encloser && (nsec_domain = 1192 find_covering_nsec(closest_encloser-> 1193 wildcard_child_closest_match, q->zone, 1194 &nsec_rrset)) != NULL) { 1195 add_rrset(q, answer, AUTHORITY_SECTION, nsec_domain, nsec_rrset); 1196 } 1197 } 1198 } 1199 1200 #ifdef NSEC3 1201 if (RCODE(q->packet)!=RCODE_OK) { 1202 return; /* nsec3 collision failure */ 1203 } 1204 #endif 1205 if (match) { 1206 answer_domain(nsd, q, answer, match, original); 1207 } else { 1208 answer_nxdomain(q, answer); 1209 } 1210 } 1211 1212 /* 1213 * qname may be different after CNAMEs have been followed from query->qname. 1214 */ 1215 static void 1216 answer_lookup_zone(struct nsd *nsd, struct query *q, answer_type *answer, 1217 size_t domain_number, int exact, domain_type *closest_match, 1218 domain_type *closest_encloser, const dname_type *qname) 1219 { 1220 q->zone = domain_find_zone(nsd->db, closest_encloser); 1221 if (!q->zone) { 1222 /* no zone for this */ 1223 if(q->cname_count == 0) 1224 RCODE_SET(q->packet, RCODE_REFUSE); 1225 return; 1226 } 1227 assert(closest_encloser); /* otherwise, no q->zone would be found */ 1228 if(!q->zone->apex || !q->zone->soa_rrset) { 1229 /* zone is configured but not loaded */ 1230 if(q->cname_count == 0) 1231 RCODE_SET(q->packet, RCODE_SERVFAIL); 1232 return; 1233 } 1234 /* now move up the closest encloser until it exists, previous 1235 * (possibly empty) closest encloser was useful to finding the zone 1236 * (for empty zones too), but now we want actual data nodes */ 1237 if (closest_encloser && !closest_encloser->is_existing) { 1238 exact = 0; 1239 while (closest_encloser != NULL && !closest_encloser->is_existing) 1240 closest_encloser = closest_encloser->parent; 1241 } 1242 1243 /* 1244 * See RFC 4035 (DNSSEC protocol) section 3.1.4.1 Responding 1245 * to Queries for DS RRs. 1246 */ 1247 if (exact && q->qtype == TYPE_DS && closest_encloser == q->zone->apex) { 1248 /* 1249 * Type DS query at a zone cut, use the responsible 1250 * parent zone to generate the answer if we are 1251 * authoritative for the parent zone. 1252 */ 1253 zone_type *zone = domain_find_parent_zone(nsd->db, q->zone); 1254 if (zone) { 1255 q->zone = zone; 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 1265 /* see if the zone has expired (for secondary zones) */ 1266 if(q->zone && q->zone->opts && q->zone->opts->pattern && 1267 q->zone->opts->pattern->request_xfr != 0 && !q->zone->is_ok) { 1268 if(q->cname_count == 0) 1269 RCODE_SET(q->packet, RCODE_SERVFAIL); 1270 return; 1271 } 1272 1273 if (exact && q->qtype == TYPE_DS && closest_encloser == q->zone->apex) { 1274 /* 1275 * Type DS query at the zone apex (and the server is 1276 * not authoritative for the parent zone). 1277 */ 1278 if (q->qclass == CLASS_ANY) { 1279 AA_CLR(q->packet); 1280 } else { 1281 AA_SET(q->packet); 1282 } 1283 answer_nodata(q, answer, closest_encloser); 1284 } else { 1285 q->delegation_domain = domain_find_ns_rrsets( 1286 closest_encloser, q->zone, &q->delegation_rrset); 1287 if(q->delegation_domain && find_dname_above(q->delegation_domain, q->zone)) { 1288 q->delegation_domain = NULL; /* use higher DNAME */ 1289 } 1290 1291 if (!q->delegation_domain 1292 || (exact && q->qtype == TYPE_DS && closest_encloser == q->delegation_domain)) 1293 { 1294 if (q->qclass == CLASS_ANY) { 1295 AA_CLR(q->packet); 1296 } else { 1297 AA_SET(q->packet); 1298 } 1299 answer_authoritative(nsd, q, answer, domain_number, exact, 1300 closest_match, closest_encloser, qname); 1301 } 1302 else { 1303 answer_delegation(q, answer); 1304 } 1305 } 1306 } 1307 1308 static void 1309 answer_query(struct nsd *nsd, struct query *q) 1310 { 1311 domain_type *closest_match; 1312 domain_type *closest_encloser; 1313 int exact; 1314 uint16_t offset; 1315 answer_type answer; 1316 1317 answer_init(&answer); 1318 1319 exact = namedb_lookup(nsd->db, q->qname, &closest_match, &closest_encloser); 1320 1321 answer_lookup_zone(nsd, q, &answer, 0, exact, closest_match, 1322 closest_encloser, q->qname); 1323 ZTATUP2(nsd, q->zone, opcode, q->opcode); 1324 ZTATUP2(nsd, q->zone, qtype, q->qtype); 1325 ZTATUP2(nsd, q->zone, qclass, q->qclass); 1326 1327 offset = dname_label_offsets(q->qname)[domain_dname(closest_encloser)->label_count - 1] + QHEADERSZ; 1328 query_add_compression_domain(q, closest_encloser, offset); 1329 encode_answer(q, &answer); 1330 query_clear_compression_tables(q); 1331 } 1332 1333 void 1334 query_prepare_response(query_type *q) 1335 { 1336 uint16_t flags; 1337 1338 /* 1339 * Preserve the data up-to the current packet's limit. 1340 */ 1341 buffer_set_position(q->packet, buffer_limit(q->packet)); 1342 buffer_set_limit(q->packet, buffer_capacity(q->packet)); 1343 1344 /* 1345 * Reserve space for the EDNS records if required. 1346 */ 1347 q->reserved_space = edns_reserved_space(&q->edns); 1348 q->reserved_space += tsig_reserved_space(&q->tsig); 1349 1350 /* Update the flags. */ 1351 flags = FLAGS(q->packet); 1352 flags &= 0x0100U; /* Preserve the RD flag. */ 1353 /* CD flag must be cleared for auth answers */ 1354 flags |= 0x8000U; /* Set the QR flag. */ 1355 FLAGS_SET(q->packet, flags); 1356 } 1357 1358 /* 1359 * Processes the query. 1360 * 1361 */ 1362 query_state_type 1363 query_process(query_type *q, nsd_type *nsd) 1364 { 1365 /* The query... */ 1366 nsd_rc_type rc; 1367 query_state_type query_state; 1368 uint16_t arcount; 1369 1370 /* Sanity checks */ 1371 if (buffer_limit(q->packet) < QHEADERSZ) { 1372 /* packet too small to contain DNS header. 1373 Now packet investigation macros will work without problems. */ 1374 return QUERY_DISCARDED; 1375 } 1376 if (QR(q->packet)) { 1377 /* Not a query? Drop it on the floor. */ 1378 return QUERY_DISCARDED; 1379 } 1380 1381 /* check opcode early on, because new opcodes may have different 1382 * specification of the meaning of the rest of the packet */ 1383 q->opcode = OPCODE(q->packet); 1384 if(q->opcode != OPCODE_QUERY && q->opcode != OPCODE_NOTIFY) { 1385 if(query_ratelimit_err(nsd)) 1386 return QUERY_DISCARDED; 1387 return query_error(q, NSD_RC_IMPL); 1388 } 1389 1390 if (RCODE(q->packet) != RCODE_OK || !process_query_section(q)) { 1391 return query_formerr(q, nsd); 1392 } 1393 1394 /* Update statistics. */ 1395 STATUP2(nsd, opcode, q->opcode); 1396 STATUP2(nsd, qtype, q->qtype); 1397 STATUP2(nsd, qclass, q->qclass); 1398 1399 if (q->opcode != OPCODE_QUERY) { 1400 if (q->opcode == OPCODE_NOTIFY) { 1401 return answer_notify(nsd, q); 1402 } else { 1403 if(query_ratelimit_err(nsd)) 1404 return QUERY_DISCARDED; 1405 return query_error(q, NSD_RC_IMPL); 1406 } 1407 } 1408 1409 /* Dont bother to answer more than one question at once... */ 1410 if (QDCOUNT(q->packet) != 1) { 1411 FLAGS_SET(q->packet, 0); 1412 return query_formerr(q, nsd); 1413 } 1414 /* Ignore settings of flags */ 1415 1416 /* Dont allow any records in the answer or authority section... 1417 except for IXFR queries. */ 1418 if (ANCOUNT(q->packet) != 0 || 1419 (q->qtype!=TYPE_IXFR && NSCOUNT(q->packet) != 0)) { 1420 return query_formerr(q, nsd); 1421 } 1422 if(q->qtype==TYPE_IXFR && NSCOUNT(q->packet) > 0) { 1423 int i; /* skip ixfr soa information data here */ 1424 for(i=0; i< NSCOUNT(q->packet); i++) 1425 if(!packet_skip_rr(q->packet, 0)) 1426 return query_formerr(q, nsd); 1427 } 1428 1429 arcount = ARCOUNT(q->packet); 1430 if (arcount > 0) { 1431 /* According to draft-ietf-dnsext-rfc2671bis-edns0-10: 1432 * "The placement flexibility for the OPT RR does not 1433 * override the need for the TSIG or SIG(0) RRs to be 1434 * the last in the additional section whenever they are 1435 * present." 1436 * So we should not have to check for TSIG RR before 1437 * OPT RR. Keep the code for backwards compatibility. 1438 */ 1439 1440 /* see if tsig is before edns record */ 1441 if (!tsig_parse_rr(&q->tsig, q->packet)) 1442 return query_formerr(q, nsd); 1443 if(q->tsig.status != TSIG_NOT_PRESENT) 1444 --arcount; 1445 } 1446 /* See if there is an OPT RR. */ 1447 if (arcount > 0) { 1448 if (edns_parse_record(&q->edns, q->packet, q, nsd)) 1449 --arcount; 1450 } 1451 /* See if there is a TSIG RR. */ 1452 if (arcount > 0 && q->tsig.status == TSIG_NOT_PRESENT) { 1453 /* see if tsig is after the edns record */ 1454 if (!tsig_parse_rr(&q->tsig, q->packet)) 1455 return query_formerr(q, nsd); 1456 if(q->tsig.status != TSIG_NOT_PRESENT) 1457 --arcount; 1458 } 1459 /* If more RRs left in Add. Section, FORMERR. */ 1460 if (arcount > 0) { 1461 return query_formerr(q, nsd); 1462 } 1463 1464 /* Do we have any trailing garbage? */ 1465 #ifdef STRICT_MESSAGE_PARSE 1466 if (buffer_remaining(q->packet) > 0) { 1467 /* If we're strict.... */ 1468 return query_formerr(q, nsd); 1469 } 1470 #endif 1471 /* Remove trailing garbage. */ 1472 buffer_set_limit(q->packet, buffer_position(q->packet)); 1473 1474 rc = process_tsig(q); 1475 if (rc != NSD_RC_OK) { 1476 return query_error(q, rc); 1477 } 1478 rc = process_edns(nsd, q); 1479 if (rc != NSD_RC_OK) { 1480 /* We should not return FORMERR, but BADVERS (=16). 1481 * BADVERS is created with Ext. RCODE, followed by RCODE. 1482 * Ext. RCODE is set to 1, RCODE must be 0 (getting 0x10 = 16). 1483 * Thus RCODE = NOERROR = NSD_RC_OK. */ 1484 return query_error(q, NSD_RC_OK); 1485 } 1486 1487 query_prepare_response(q); 1488 1489 if (q->qclass != CLASS_IN && q->qclass != CLASS_ANY) { 1490 if (q->qclass == CLASS_CH) { 1491 return answer_chaos(nsd, q); 1492 } else { 1493 return query_error(q, NSD_RC_REFUSE); 1494 } 1495 } 1496 1497 query_state = answer_axfr_ixfr(nsd, q); 1498 if (query_state == QUERY_PROCESSED || query_state == QUERY_IN_AXFR) { 1499 return query_state; 1500 } 1501 if(q->qtype == TYPE_ANY && nsd->options->refuse_any && !q->tcp) { 1502 TC_SET(q->packet); 1503 return query_error(q, NSD_RC_OK); 1504 } 1505 1506 answer_query(nsd, q); 1507 1508 return QUERY_PROCESSED; 1509 } 1510 1511 void 1512 query_add_optional(query_type *q, nsd_type *nsd) 1513 { 1514 struct edns_data *edns = &nsd->edns_ipv4; 1515 #if defined(INET6) 1516 if (q->addr.ss_family == AF_INET6) { 1517 edns = &nsd->edns_ipv6; 1518 } 1519 #endif 1520 if (RCODE(q->packet) == RCODE_FORMAT) { 1521 return; 1522 } 1523 switch (q->edns.status) { 1524 case EDNS_NOT_PRESENT: 1525 break; 1526 case EDNS_OK: 1527 if (q->edns.dnssec_ok) edns->ok[7] = 0x80; 1528 else edns->ok[7] = 0x00; 1529 buffer_write(q->packet, edns->ok, OPT_LEN); 1530 if(q->edns.opt_reserved_space == 0 || !buffer_available( 1531 q->packet, 2+q->edns.opt_reserved_space)) { 1532 /* fill with NULLs */ 1533 buffer_write(q->packet, edns->rdata_none, OPT_RDATA); 1534 } else { 1535 /* rdata length */ 1536 buffer_write_u16(q->packet, q->edns.opt_reserved_space); 1537 /* edns options */ 1538 if(q->edns.nsid) { 1539 /* nsid opt header */ 1540 buffer_write(q->packet, edns->nsid, OPT_HDR); 1541 /* nsid payload */ 1542 buffer_write(q->packet, nsd->nsid, nsd->nsid_len); 1543 } 1544 } 1545 ARCOUNT_SET(q->packet, ARCOUNT(q->packet) + 1); 1546 STATUP(nsd, edns); 1547 ZTATUP(nsd, q->zone, edns); 1548 break; 1549 case EDNS_ERROR: 1550 if (q->edns.dnssec_ok) edns->error[7] = 0x80; 1551 else edns->error[7] = 0x00; 1552 buffer_write(q->packet, edns->error, OPT_LEN); 1553 buffer_write(q->packet, edns->rdata_none, OPT_RDATA); 1554 ARCOUNT_SET(q->packet, ARCOUNT(q->packet) + 1); 1555 STATUP(nsd, ednserr); 1556 ZTATUP(nsd, q->zone, ednserr); 1557 break; 1558 } 1559 1560 if (q->tsig.status != TSIG_NOT_PRESENT) { 1561 if (q->tsig.status == TSIG_ERROR || 1562 q->tsig.error_code != TSIG_ERROR_NOERROR) { 1563 tsig_error_reply(&q->tsig); 1564 tsig_append_rr(&q->tsig, q->packet); 1565 ARCOUNT_SET(q->packet, ARCOUNT(q->packet) + 1); 1566 } else if(q->tsig.status == TSIG_OK && 1567 q->tsig.error_code == TSIG_ERROR_NOERROR) 1568 { 1569 if(q->tsig_prepare_it) 1570 tsig_prepare(&q->tsig); 1571 if(q->tsig_update_it) 1572 tsig_update(&q->tsig, q->packet, buffer_position(q->packet)); 1573 if(q->tsig_sign_it) { 1574 tsig_sign(&q->tsig); 1575 tsig_append_rr(&q->tsig, q->packet); 1576 ARCOUNT_SET(q->packet, ARCOUNT(q->packet) + 1); 1577 } 1578 } 1579 } 1580 } 1581