1 /* 2 * iterator/iterator.h - iterative resolver DNS query response module 3 * 4 * Copyright (c) 2007, NLnet Labs. All rights reserved. 5 * 6 * This software is open source. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 12 * Redistributions of source code must retain the above copyright notice, 13 * this list of conditions and the following disclaimer. 14 * 15 * Redistributions in binary form must reproduce the above copyright notice, 16 * this list of conditions and the following disclaimer in the documentation 17 * and/or other materials provided with the distribution. 18 * 19 * Neither the name of the NLNET LABS nor the names of its contributors may 20 * be used to endorse or promote products derived from this software without 21 * specific prior written permission. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 24 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 25 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 26 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 27 * HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 28 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED 29 * TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR 30 * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF 31 * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING 32 * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 33 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 34 */ 35 36 /** 37 * \file 38 * 39 * This file contains a module that performs recursive iterative DNS query 40 * processing. 41 */ 42 43 #ifndef ITERATOR_ITERATOR_H 44 #define ITERATOR_ITERATOR_H 45 #include "services/outbound_list.h" 46 #include "util/data/msgreply.h" 47 #include "util/module.h" 48 struct delegpt; 49 struct iter_hints; 50 struct iter_forwards; 51 struct iter_donotq; 52 struct iter_prep_list; 53 struct iter_priv; 54 struct rbtree_type; 55 56 /** max number of targets spawned for a query and its subqueries */ 57 #define MAX_TARGET_COUNT 64 58 /** max number of upstream queries for a query and its subqueries, it is 59 * never reset. */ 60 extern int MAX_GLOBAL_QUOTA; 61 /** max number of target lookups per qstate, per delegation point */ 62 #define MAX_DP_TARGET_COUNT 16 63 /** max number of nxdomains allowed for target lookups for a query and 64 * its subqueries */ 65 #define MAX_TARGET_NX 5 66 /** max number of nxdomains allowed for target lookups for a query and 67 * its subqueries when fallback has kicked in */ 68 #define MAX_TARGET_NX_FALLBACK (MAX_TARGET_NX*2) 69 /** max number of referrals. Makes sure resolver does not run away */ 70 #define MAX_REFERRAL_COUNT 130 71 /** max number of queries for which to perform dnsseclameness detection, 72 * (rrsigs missing detection) after that, just pick up that response */ 73 #define DNSSEC_LAME_DETECT_COUNT 4 74 /** 75 * max number of QNAME minimisation iterations. Limits number of queries for 76 * QNAMEs with a lot of labels. 77 */ 78 #define MAX_MINIMISE_COUNT 10 79 /* max number of time-outs for minimised query. Prevents resolving failures 80 * when the QNAME minimisation QTYPE is blocked. */ 81 #define MAX_MINIMISE_TIMEOUT_COUNT 3 82 /** 83 * number of labels from QNAME that are always send individually when using 84 * QNAME minimisation, even when the number of labels of the QNAME is bigger 85 * than MAX_MINIMISE_COUNT */ 86 #define MINIMISE_ONE_LAB 4 87 #define MINIMISE_MULTIPLE_LABS (MAX_MINIMISE_COUNT - MINIMISE_ONE_LAB) 88 /** at what query-sent-count to stop target fetch policy */ 89 #define TARGET_FETCH_STOP 3 90 /** how nice is a server without further information, in msec 91 * Equals rtt initial timeout value. 92 */ 93 extern int UNKNOWN_SERVER_NICENESS; 94 /** maximum timeout before a host is deemed unsuitable, in msec. 95 * After host_ttl this will be timed out and the host will be tried again. 96 * Equals RTT_MAX_TIMEOUT, and thus when RTT_MAX_TIMEOUT is overwritten by 97 * config infra_cache_max_rtt, it will be overwritten as well. */ 98 extern int USEFUL_SERVER_TOP_TIMEOUT; 99 /** penalty to validation failed blacklisted IPs 100 * Equals USEFUL_SERVER_TOP_TIMEOUT*4, and thus when RTT_MAX_TIMEOUT is 101 * overwritten by config infra_cache_max_rtt, it will be overwritten as well. */ 102 extern int BLACKLIST_PENALTY; 103 /** RTT band, within this amount from the best, servers are chosen randomly. 104 * Chosen so that the UNKNOWN_SERVER_NICENESS falls within the band of a 105 * fast server, this causes server exploration as a side benefit. msec. */ 106 #define RTT_BAND 400 107 /** Number of retries for empty nodata packets before it is accepted. */ 108 #define EMPTY_NODATA_RETRY_COUNT 2 109 110 /** 111 * Global state for the iterator. 112 */ 113 struct iter_env { 114 /** A flag to indicate whether or not we have an IPv6 route */ 115 int supports_ipv6; 116 117 /** A flag to indicate whether or not we have an IPv4 route */ 118 int supports_ipv4; 119 120 /** A flag to locally apply NAT64 to make IPv4 addrs into IPv6 */ 121 int use_nat64; 122 123 /** NAT64 prefix address, cf. dns64_env->prefix_addr */ 124 struct sockaddr_storage nat64_prefix_addr; 125 126 /** sizeof(sockaddr_in6) */ 127 socklen_t nat64_prefix_addrlen; 128 129 /** CIDR mask length of NAT64 prefix */ 130 int nat64_prefix_net; 131 132 /** A set of inetaddrs that should never be queried. */ 133 struct iter_donotq* donotq; 134 135 /** private address space and private domains */ 136 struct iter_priv* priv; 137 138 /** whitelist for capsforid names */ 139 struct rbtree_type* caps_white; 140 141 /** The maximum dependency depth that this resolver will pursue. */ 142 int max_dependency_depth; 143 144 /** 145 * The target fetch policy for each dependency level. This is 146 * described as a simple number (per dependency level): 147 * negative numbers (usually just -1) mean fetch-all, 148 * 0 means only fetch on demand, and 149 * positive numbers mean to fetch at most that many targets. 150 * array of max_dependency_depth+1 size. 151 */ 152 int* target_fetch_policy; 153 154 /** lock on ratelimit counter */ 155 lock_basic_type queries_ratelimit_lock; 156 /** number of queries that have been ratelimited */ 157 size_t num_queries_ratelimited; 158 159 /** number of retries on outgoing queries */ 160 int outbound_msg_retry; 161 162 /** number of queries_sent */ 163 int max_sent_count; 164 165 /** max number of query restarts to limit length of CNAME chain */ 166 int max_query_restarts; 167 }; 168 169 /** 170 * QNAME minimisation state 171 */ 172 enum minimisation_state { 173 /** 174 * (Re)start minimisation. Outgoing QNAME should be set to dp->name. 175 * State entered on new query or after following referral or CNAME. 176 */ 177 INIT_MINIMISE_STATE = 0, 178 /** 179 * QNAME minimisation ongoing. Increase QNAME on every iteration. 180 */ 181 MINIMISE_STATE, 182 /** 183 * Don't increment QNAME this iteration 184 */ 185 SKIP_MINIMISE_STATE, 186 /** 187 * Send out full QNAME + original QTYPE 188 */ 189 DONOT_MINIMISE_STATE, 190 }; 191 192 /** 193 * State of the iterator for a query. 194 */ 195 enum iter_state { 196 /** 197 * Externally generated queries start at this state. Query restarts are 198 * reset to this state. 199 */ 200 INIT_REQUEST_STATE = 0, 201 202 /** 203 * Root priming events reactivate here, most other events pass 204 * through this naturally as the 2nd part of the INIT_REQUEST_STATE. 205 */ 206 INIT_REQUEST_2_STATE, 207 208 /** 209 * Stub priming events reactivate here, most other events pass 210 * through this naturally as the 3rd part of the INIT_REQUEST_STATE. 211 */ 212 INIT_REQUEST_3_STATE, 213 214 /** 215 * Each time a delegation point changes for a given query or a 216 * query times out and/or wakes up, this state is (re)visited. 217 * This state is responsible for iterating through a list of 218 * nameserver targets. 219 */ 220 QUERYTARGETS_STATE, 221 222 /** 223 * Responses to queries start at this state. This state handles 224 * the decision tree associated with handling responses. 225 */ 226 QUERY_RESP_STATE, 227 228 /** Responses to priming queries finish at this state. */ 229 PRIME_RESP_STATE, 230 231 /** Collecting query class information, for qclass=ANY, when 232 * it spawns off queries for every class, it returns here. */ 233 COLLECT_CLASS_STATE, 234 235 /** Find NS record to resolve DS record from, walking to the right 236 * NS spot until we find it */ 237 DSNS_FIND_STATE, 238 239 /** Responses that are to be returned upstream end at this state. 240 * As well as responses to target queries. */ 241 FINISHED_STATE 242 }; 243 244 /** 245 * Shared counters for queries. 246 */ 247 enum target_count_variables { 248 /** Reference count for the shared iter_qstate->target_count. */ 249 TARGET_COUNT_REF = 0, 250 /** Number of target queries spawned for the query and subqueries. */ 251 TARGET_COUNT_QUERIES, 252 /** Number of nxdomain responses encountered. */ 253 TARGET_COUNT_NX, 254 /** Global quota on number of queries to upstream servers per 255 * client request, that is never reset. */ 256 TARGET_COUNT_GLOBAL_QUOTA, 257 258 /** This should stay last here, it is used for the allocation */ 259 TARGET_COUNT_MAX, 260 }; 261 262 /** 263 * Per query state for the iterator module. 264 */ 265 struct iter_qstate { 266 /** 267 * State of the iterator module. 268 * This is the state that event is in or should sent to -- all 269 * requests should start with the INIT_REQUEST_STATE. All 270 * responses should start with QUERY_RESP_STATE. Subsequent 271 * processing of the event will change this state. 272 */ 273 enum iter_state state; 274 275 /** 276 * Final state for the iterator module. 277 * This is the state that responses should be routed to once the 278 * response is final. For externally initiated queries, this 279 * will be FINISHED_STATE, locally initiated queries will have 280 * different final states. 281 */ 282 enum iter_state final_state; 283 284 /** 285 * The depth of this query, this means the depth of recursion. 286 * This address is needed for another query, which is an address 287 * needed for another query, etc. Original client query has depth 0. 288 */ 289 int depth; 290 291 /** 292 * The response 293 */ 294 struct dns_msg* response; 295 296 /** 297 * This is a list of RRsets that must be prepended to the 298 * ANSWER section of a response before being sent upstream. 299 */ 300 struct iter_prep_list* an_prepend_list; 301 /** Last element of the prepend list */ 302 struct iter_prep_list* an_prepend_last; 303 304 /** 305 * This is the list of RRsets that must be prepended to the 306 * AUTHORITY section of the response before being sent upstream. 307 */ 308 struct iter_prep_list* ns_prepend_list; 309 /** Last element of the authority prepend list */ 310 struct iter_prep_list* ns_prepend_last; 311 312 /** query name used for chasing the results. Initially the same as 313 * the state qinfo, but after CNAMEs this will be different. 314 * The query info used to elicit the results needed. */ 315 struct query_info qchase; 316 /** query flags to use when chasing the answer (i.e. RD flag) */ 317 uint16_t chase_flags; 318 /** true if we set RD bit because of last resort recursion lame query*/ 319 int chase_to_rd; 320 321 /** 322 * This is the current delegation point for an in-progress query. This 323 * object retains state as to which delegation targets need to be 324 * (sub)queried for vs which ones have already been visited. 325 */ 326 struct delegpt* dp; 327 328 /** state for 0x20 fallback when capsfail happens, 0 not a fallback */ 329 int caps_fallback; 330 /** state for capsfail: current server number to try */ 331 size_t caps_server; 332 /** state for capsfail: stored query for comparisons. Can be NULL if 333 * no response had been seen prior to starting the fallback. */ 334 struct reply_info* caps_reply; 335 struct dns_msg* caps_response; 336 337 /** Current delegation message - returned for non-RD queries */ 338 struct dns_msg* deleg_msg; 339 340 /** number of outstanding target sub queries */ 341 int num_target_queries; 342 343 /** outstanding direct queries */ 344 int num_current_queries; 345 346 /** the number of times this query has been restarted. */ 347 int query_restart_count; 348 349 /** the number of times this query has followed a referral. */ 350 int referral_count; 351 352 /** number of queries fired off */ 353 int sent_count; 354 355 /** malloced-array shared with this query and its subqueries. It keeps 356 * track of the defined enum target_count_variables counters. */ 357 int* target_count; 358 359 /** number of target lookups per delegation point. Reset to 0 after 360 * receiving referral answer. Not shared with subqueries. */ 361 int dp_target_count; 362 363 /** Delegation point that triggered the NXNS fallback; shared with 364 * this query and its subqueries, count-referenced by the reference 365 * counter in target_count. 366 * This also marks the fallback activation. */ 367 uint8_t** nxns_dp; 368 369 /** if true, already tested for ratelimiting and passed the test */ 370 int ratelimit_ok; 371 372 /** 373 * The query must store NS records from referrals as parentside RRs 374 * Enabled once it hits resolution problems, to throttle retries. 375 * If enabled it is the pointer to the old delegation point with 376 * the old retry counts for bad-nameserver-addresses. 377 */ 378 struct delegpt* store_parent_NS; 379 380 /** 381 * The query is for parent-side glue(A or AAAA) for a nameserver. 382 * If the item is seen as glue in a referral, and pside_glue is NULL, 383 * then it is stored in pside_glue for later. 384 * If it was never seen, at the end, then a negative caching element 385 * must be created. 386 * The (data or negative) RR cache element then throttles retries. 387 */ 388 int query_for_pside_glue; 389 /** the parent-side-glue element (NULL if none, its first match) */ 390 struct ub_packed_rrset_key* pside_glue; 391 392 /** If nonNULL we are walking upwards from DS query to find NS */ 393 uint8_t* dsns_point; 394 /** length of the dname in dsns_point */ 395 size_t dsns_point_len; 396 397 /** 398 * expected dnssec information for this iteration step. 399 * If dnssec rrsigs are expected and not given, the server is marked 400 * lame (dnssec-lame). 401 */ 402 int dnssec_expected; 403 404 /** 405 * We are expecting dnssec information, but we also know the server 406 * is DNSSEC lame. The response need not be marked dnssec-lame again. 407 */ 408 int dnssec_lame_query; 409 410 /** 411 * This is flag that, if true, means that this event is 412 * waiting for a stub priming query. 413 */ 414 int wait_priming_stub; 415 416 /** 417 * This is a flag that, if true, means that this query is 418 * for (re)fetching glue from a zone. Since the address should 419 * have been glue, query again to the servers that should have 420 * been returning it as glue. 421 * The delegation point must be set to the one that should *not* 422 * be used when creating the state. A higher one will be attempted. 423 */ 424 int refetch_glue; 425 426 /** 427 * This flag detects that a completely empty nodata was received, 428 * already so that it is accepted later. */ 429 int empty_nodata_found; 430 431 /** list of pending queries to authoritative servers. */ 432 struct outbound_list outlist; 433 434 /** QNAME minimisation state, RFC9156 */ 435 enum minimisation_state minimisation_state; 436 437 /** State for capsfail: QNAME minimisation state for comparisons. */ 438 enum minimisation_state caps_minimisation_state; 439 440 /** 441 * The query info that is sent upstream. Will be a subset of qchase 442 * when qname minimisation is enabled. 443 */ 444 struct query_info qinfo_out; 445 446 /** 447 * Count number of QNAME minimisation iterations. Used to limit number of 448 * outgoing queries when QNAME minimisation is enabled. 449 */ 450 int minimise_count; 451 452 /** 453 * Count number of time-outs. Used to prevent resolving failures when 454 * the QNAME minimisation QTYPE is blocked. Used to determine if 455 * capsforid fallback should be started.*/ 456 int timeout_count; 457 458 /** True if the current response is from auth_zone */ 459 int auth_zone_response; 460 /** True if the auth_zones should not be consulted for the query */ 461 int auth_zone_avoid; 462 /** true if there have been scrubbing failures of reply packets */ 463 int scrub_failures; 464 /** true if there have been parse failures of reply packets */ 465 int parse_failures; 466 /** a failure printout address for last received answer */ 467 union { 468 struct in_addr in; 469 #ifdef AF_INET6 470 struct in6_addr in6; 471 #endif 472 } fail_addr; 473 /** which fail_addr, 0 is nothing, 4 or 6 */ 474 int fail_addr_type; 475 }; 476 477 /** 478 * List of prepend items 479 */ 480 struct iter_prep_list { 481 /** next in list */ 482 struct iter_prep_list* next; 483 /** rrset */ 484 struct ub_packed_rrset_key* rrset; 485 }; 486 487 /** 488 * Get the iterator function block. 489 * @return: function block with function pointers to iterator methods. 490 */ 491 struct module_func_block* iter_get_funcblock(void); 492 493 /** 494 * Get iterator state as a string 495 * @param state: to convert 496 * @return constant string that is printable. 497 */ 498 const char* iter_state_to_string(enum iter_state state); 499 500 /** 501 * See if iterator state is a response state 502 * @param s: to inspect 503 * @return true if response state. 504 */ 505 int iter_state_is_responsestate(enum iter_state s); 506 507 /** iterator init */ 508 int iter_init(struct module_env* env, int id); 509 510 /** iterator deinit */ 511 void iter_deinit(struct module_env* env, int id); 512 513 /** iterator operate on a query */ 514 void iter_operate(struct module_qstate* qstate, enum module_ev event, int id, 515 struct outbound_entry* outbound); 516 517 /** 518 * Return priming query results to interested super querystates. 519 * 520 * Sets the delegation point and delegation message (not nonRD queries). 521 * This is a callback from walk_supers. 522 * 523 * @param qstate: query state that finished. 524 * @param id: module id. 525 * @param super: the qstate to inform. 526 */ 527 void iter_inform_super(struct module_qstate* qstate, int id, 528 struct module_qstate* super); 529 530 /** iterator cleanup query state */ 531 void iter_clear(struct module_qstate* qstate, int id); 532 533 /** iterator alloc size routine */ 534 size_t iter_get_mem(struct module_env* env, int id); 535 536 #endif /* ITERATOR_ITERATOR_H */ 537