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