xref: /freebsd/contrib/unbound/iterator/iterator.h (revision 7a789145f88a6aceacc59029a0cafe7de7aeefea)
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_donotq;
50 struct iter_prep_list;
51 struct iter_priv;
52 struct rbtree_type;
53 
54 /** max number of targets spawned for a query and its subqueries */
55 #define MAX_TARGET_COUNT	64
56 /** max number of upstream queries for a query and its subqueries, it is
57  * never reset. */
58 extern int MAX_GLOBAL_QUOTA;
59 /** max number of target lookups per qstate, per delegation point */
60 #define MAX_DP_TARGET_COUNT	16
61 /** max number of nxdomains allowed for target lookups for a query and
62  * its subqueries */
63 #define MAX_TARGET_NX		5
64 /** max number of nxdomains allowed for target lookups for a query and
65  * its subqueries when fallback has kicked in */
66 #define MAX_TARGET_NX_FALLBACK	(MAX_TARGET_NX*2)
67 /** max number of referrals. Makes sure resolver does not run away */
68 #define MAX_REFERRAL_COUNT	130
69 /** max number of queries for which to perform dnsseclameness detection,
70  * (rrsigs missing detection) after that, just pick up that response */
71 #define DNSSEC_LAME_DETECT_COUNT 4
72 /**
73  * max number of QNAME minimisation iterations. Limits number of queries for
74  * QNAMEs with a lot of labels.
75 */
76 #define MAX_MINIMISE_COUNT	10
77 /* max number of time-outs for minimised query. Prevents resolving failures
78  * when the QNAME minimisation QTYPE is blocked. */
79 #define MAX_MINIMISE_TIMEOUT_COUNT 3
80 /**
81  * number of labels from QNAME that are always send individually when using
82  * QNAME minimisation, even when the number of labels of the QNAME is bigger
83  * than MAX_MINIMISE_COUNT */
84 #define MINIMISE_ONE_LAB	4
85 #define MINIMISE_MULTIPLE_LABS	(MAX_MINIMISE_COUNT - MINIMISE_ONE_LAB)
86 /** at what query-sent-count to stop target fetch policy */
87 #define TARGET_FETCH_STOP	3
88 /** how nice is a server without further information, in msec
89  * Equals rtt initial timeout value.
90  */
91 extern int UNKNOWN_SERVER_NICENESS;
92 /** maximum timeout before a host is deemed unsuitable, in msec.
93  * After host_ttl this will be timed out and the host will be tried again.
94  * Equals RTT_MAX_TIMEOUT, and thus when RTT_MAX_TIMEOUT is overwritten by
95  * config infra_cache_max_rtt, it will be overwritten as well. */
96 extern int USEFUL_SERVER_TOP_TIMEOUT;
97 /** penalty to validation failed blacklisted IPs
98  * Equals USEFUL_SERVER_TOP_TIMEOUT*4, and thus when RTT_MAX_TIMEOUT is
99  * overwritten by config infra_cache_max_rtt, it will be overwritten as well. */
100 extern int BLACKLIST_PENALTY;
101 /** RTT band, within this amount from the best, servers are chosen randomly.
102  * Chosen so that the UNKNOWN_SERVER_NICENESS falls within the band of a
103  * fast server, this causes server exploration as a side benefit. msec. */
104 #define RTT_BAND 400
105 /** Number of retries for empty nodata packets before it is accepted. */
106 #define EMPTY_NODATA_RETRY_COUNT 2
107 /** max label-strip iterations in DSNS_FIND_STATE (RFC 4035 4.2 parent-NS
108  * search) before giving up; bounds upstream NS sends per client DS.
109  * Means the max number of labels in grandchild to the grandparent zone that
110  * are co-hosted. */
111 #define MAX_DSNS_FIND_COUNT    20
112 
113 /**
114  * Iterator global state for nat64.
115  */
116 struct iter_nat64 {
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 
130 /**
131  * Global state for the iterator.
132  */
133 struct iter_env {
134 	/** A flag to indicate whether or not we have an IPv6 route */
135 	int supports_ipv6;
136 
137 	/** A flag to indicate whether or not we have an IPv4 route */
138 	int supports_ipv4;
139 
140 	/** State for nat64 */
141 	struct iter_nat64 nat64;
142 
143 	/** A set of inetaddrs that should never be queried. */
144 	struct iter_donotq* donotq;
145 
146 	/** private address space and private domains */
147 	struct iter_priv* priv;
148 
149 	/** whitelist for capsforid names */
150 	struct rbtree_type* caps_white;
151 
152 	/** The maximum dependency depth that this resolver will pursue. */
153 	int max_dependency_depth;
154 
155 	/**
156 	 * The target fetch policy for each dependency level. This is
157 	 * described as a simple number (per dependency level):
158 	 *	negative numbers (usually just -1) mean fetch-all,
159 	 *	0 means only fetch on demand, and
160 	 *	positive numbers mean to fetch at most that many targets.
161 	 * array of max_dependency_depth+1 size.
162 	 */
163 	int* target_fetch_policy;
164 
165 	/** lock on ratelimit counter */
166 	lock_basic_type queries_ratelimit_lock;
167 	/** number of queries that have been ratelimited */
168 	size_t num_queries_ratelimited;
169 
170 	/** number of retries on outgoing queries */
171 	int outbound_msg_retry;
172 
173 	/** number of queries_sent */
174 	int max_sent_count;
175 
176 	/** max number of query restarts to limit length of CNAME chain */
177 	int max_query_restarts;
178 };
179 
180 /**
181  * QNAME minimisation state
182  */
183 enum minimisation_state {
184 	/**
185 	 * (Re)start minimisation. Outgoing QNAME should be set to dp->name.
186 	 * State entered on new query or after following referral or CNAME.
187 	 */
188 	INIT_MINIMISE_STATE = 0,
189 	/**
190 	 * QNAME minimisation ongoing. Increase QNAME on every iteration.
191 	 */
192 	MINIMISE_STATE,
193 	/**
194 	 * Don't increment QNAME this iteration
195 	 */
196 	SKIP_MINIMISE_STATE,
197 	/**
198 	 * Send out full QNAME + original QTYPE
199 	 */
200 	DONOT_MINIMISE_STATE,
201 };
202 
203 /**
204  * State of the iterator for a query.
205  */
206 enum iter_state {
207 	/**
208 	 * Externally generated queries start at this state. Query restarts are
209 	 * reset to this state.
210 	 */
211 	INIT_REQUEST_STATE = 0,
212 
213 	/**
214 	 * Root priming events reactivate here, most other events pass
215 	 * through this naturally as the 2nd part of the INIT_REQUEST_STATE.
216 	 */
217 	INIT_REQUEST_2_STATE,
218 
219 	/**
220 	 * Stub priming events reactivate here, most other events pass
221 	 * through this naturally as the 3rd part of the INIT_REQUEST_STATE.
222 	 */
223 	INIT_REQUEST_3_STATE,
224 
225 	/**
226 	 * Each time a delegation point changes for a given query or a
227 	 * query times out and/or wakes up, this state is (re)visited.
228 	 * This state is responsible for iterating through a list of
229 	 * nameserver targets.
230 	 */
231 	QUERYTARGETS_STATE,
232 
233 	/**
234 	 * Responses to queries start at this state. This state handles
235 	 * the decision tree associated with handling responses.
236 	 */
237 	QUERY_RESP_STATE,
238 
239 	/** Responses to priming queries finish at this state. */
240 	PRIME_RESP_STATE,
241 
242 	/** Collecting query class information, for qclass=ANY, when
243 	 * it spawns off queries for every class, it returns here. */
244 	COLLECT_CLASS_STATE,
245 
246 	/** Find NS record to resolve DS record from, walking to the right
247 	 * NS spot until we find it */
248 	DSNS_FIND_STATE,
249 
250 	/** Responses that are to be returned upstream end at this state.
251 	 * As well as responses to target queries. */
252 	FINISHED_STATE
253 };
254 
255 /**
256  * Shared counters for queries.
257  */
258 enum target_count_variables {
259 	/** Reference count for the shared iter_qstate->target_count. */
260 	TARGET_COUNT_REF = 0,
261 	/** Number of target queries spawned for the query and subqueries. */
262 	TARGET_COUNT_QUERIES,
263 	/** Number of nxdomain responses encountered. */
264 	TARGET_COUNT_NX,
265 	/** Global quota on number of queries to upstream servers per
266 	 * client request, that is never reset. */
267 	TARGET_COUNT_GLOBAL_QUOTA,
268 
269 	/** This should stay last here, it is used for the allocation */
270 	TARGET_COUNT_MAX,
271 };
272 
273 /**
274  * Per query state for the iterator module.
275  */
276 struct iter_qstate {
277 	/**
278 	 * State of the iterator module.
279 	 * This is the state that event is in or should sent to -- all
280 	 * requests should start with the INIT_REQUEST_STATE. All
281 	 * responses should start with QUERY_RESP_STATE. Subsequent
282 	 * processing of the event will change this state.
283 	 */
284 	enum iter_state state;
285 
286 	/**
287 	 * Final state for the iterator module.
288 	 * This is the state that responses should be routed to once the
289 	 * response is final. For externally initiated queries, this
290 	 * will be FINISHED_STATE, locally initiated queries will have
291 	 * different final states.
292 	 */
293 	enum iter_state final_state;
294 
295 	/**
296 	 * The depth of this query, this means the depth of recursion.
297 	 * This address is needed for another query, which is an address
298 	 * needed for another query, etc. Original client query has depth 0.
299 	 */
300 	int depth;
301 
302 	/**
303 	 * The response
304 	 */
305 	struct dns_msg* response;
306 
307 	/**
308 	 * This is a list of RRsets that must be prepended to the
309 	 * ANSWER section of a response before being sent upstream.
310 	 */
311 	struct iter_prep_list* an_prepend_list;
312 	/** Last element of the prepend list */
313 	struct iter_prep_list* an_prepend_last;
314 
315 	/**
316 	 * This is the list of RRsets that must be prepended to the
317 	 * AUTHORITY section of the response before being sent upstream.
318 	 */
319 	struct iter_prep_list* ns_prepend_list;
320 	/** Last element of the authority prepend list */
321 	struct iter_prep_list* ns_prepend_last;
322 
323 	/** query name used for chasing the results. Initially the same as
324 	 * the state qinfo, but after CNAMEs this will be different.
325 	 * The query info used to elicit the results needed. */
326 	struct query_info qchase;
327 	/** query flags to use when chasing the answer (i.e. RD flag) */
328 	uint16_t chase_flags;
329 	/** true if we set RD bit because of last resort recursion lame query*/
330 	int chase_to_rd;
331 
332 	/**
333 	 * This is the current delegation point for an in-progress query. This
334 	 * object retains state as to which delegation targets need to be
335 	 * (sub)queried for vs which ones have already been visited.
336 	 */
337 	struct delegpt* dp;
338 
339 	/** state for 0x20 fallback when capsfail happens, 0 not a fallback */
340 	int caps_fallback;
341 	/** state for capsfail: current server number to try */
342 	size_t caps_server;
343 	/** state for capsfail: stored query for comparisons. Can be NULL if
344 	 * no response had been seen prior to starting the fallback. */
345 	struct reply_info* caps_reply;
346 	struct dns_msg* caps_response;
347 
348 	/** Current delegation message - returned for non-RD queries */
349 	struct dns_msg* deleg_msg;
350 
351 	/** number of outstanding target sub queries */
352 	int num_target_queries;
353 
354 	/** outstanding direct queries */
355 	int num_current_queries;
356 
357 	/** the number of times this query has been restarted. */
358 	int query_restart_count;
359 
360 	/** the number of times this query has followed a referral. */
361 	int referral_count;
362 
363 	/** number of queries fired off */
364 	int sent_count;
365 
366 	/** malloced-array shared with this query and its subqueries. It keeps
367 	 * track of the defined enum target_count_variables counters. */
368 	int* target_count;
369 
370 	/** number of target lookups per delegation point. Reset to 0 after
371 	 * receiving referral answer. Not shared with subqueries. */
372 	int dp_target_count;
373 
374 	/** Delegation point that triggered the NXNS fallback; shared with
375 	 * this query and its subqueries, count-referenced by the reference
376 	 * counter in target_count.
377 	 * This also marks the fallback activation. */
378 	uint8_t** nxns_dp;
379 
380 	/** if true, already tested for ratelimiting and passed the test */
381 	int ratelimit_ok;
382 
383 	/** If the last query, that may be a referral, incremented the
384 	 * ratelimit counter. */
385 	int ratelimit_incremented;
386 
387 	/**
388 	 * The query must store NS records from referrals as parentside RRs
389 	 * Enabled once it hits resolution problems, to throttle retries.
390 	 * If enabled it is the pointer to the old delegation point with
391 	 * the old retry counts for bad-nameserver-addresses.
392 	 */
393 	struct delegpt* store_parent_NS;
394 
395 	/**
396 	 * The query is for parent-side glue(A or AAAA) for a nameserver.
397 	 * If the item is seen as glue in a referral, and pside_glue is NULL,
398 	 * then it is stored in pside_glue for later.
399 	 * If it was never seen, at the end, then a negative caching element
400 	 * must be created.
401 	 * The (data or negative) RR cache element then throttles retries.
402 	 */
403 	int query_for_pside_glue;
404 	/** the parent-side-glue element (NULL if none, its first match) */
405 	struct ub_packed_rrset_key* pside_glue;
406 
407 	/** If nonNULL we are walking upwards from DS query to find NS */
408 	uint8_t* dsns_point;
409 	/** length of the dname in dsns_point */
410 	size_t dsns_point_len;
411 	/** number of label-strip iterations performed in DSNS_FIND_STATE */
412 	int dsns_count;
413 
414 	/**
415 	 * expected dnssec information for this iteration step.
416 	 * If dnssec rrsigs are expected and not given, the server is marked
417 	 * lame (dnssec-lame).
418 	 */
419 	int dnssec_expected;
420 
421 	/**
422 	 * We are expecting dnssec information, but we also know the server
423 	 * is DNSSEC lame.  The response need not be marked dnssec-lame again.
424 	 */
425 	int dnssec_lame_query;
426 
427 	/**
428 	 * This is flag that, if true, means that this event is
429 	 * waiting for a stub priming query.
430 	 */
431 	int wait_priming_stub;
432 
433 	/**
434 	 * This is a flag that, if true, means that this query is
435 	 * for (re)fetching glue from a zone. Since the address should
436 	 * have been glue, query again to the servers that should have
437 	 * been returning it as glue.
438 	 * The delegation point must be set to the one that should *not*
439 	 * be used when creating the state. A higher one will be attempted.
440 	 */
441 	int refetch_glue;
442 
443 	/**
444 	 * This flag detects that a completely empty nodata was received,
445 	 * already so that it is accepted later. */
446 	int empty_nodata_found;
447 
448 	/** Store if the answer was empty, but lame, before it became empty.*/
449 	int msg_lame_empty;
450 
451 	/** Store if the answer was a referral, to self, before scrub. So the
452 	 * it is not some sort of answer. */
453 	int msg_lame_referral;
454 
455 	/** list of pending queries to authoritative servers. */
456 	struct outbound_list outlist;
457 
458 	/** QNAME minimisation state, RFC9156 */
459 	enum minimisation_state minimisation_state;
460 
461 	/** State for capsfail: QNAME minimisation state for comparisons. */
462 	enum minimisation_state caps_minimisation_state;
463 
464 	/**
465 	 * The query info that is sent upstream. Will be a subset of qchase
466 	 * when qname minimisation is enabled.
467 	 */
468 	struct query_info qinfo_out;
469 
470 	/**
471 	 * Count number of QNAME minimisation iterations. Used to limit number of
472 	 * outgoing queries when QNAME minimisation is enabled.
473 	 */
474 	int minimise_count;
475 
476 	/**
477 	 * Count number of time-outs. Used to prevent resolving failures when
478 	 * the QNAME minimisation QTYPE is blocked. Used to determine if
479 	 * capsforid fallback should be started.*/
480 	int timeout_count;
481 
482 	/** True if the current response is from auth_zone */
483 	int auth_zone_response;
484 	/** True if the auth_zones should not be consulted for the query */
485 	int auth_zone_avoid;
486 	/** true if there have been scrubbing failures of reply packets */
487 	int scrub_failures;
488 	/** true if there have been parse failures of reply packets */
489 	int parse_failures;
490 	/** a failure printout address for last received answer */
491 	union {
492 		struct in_addr in;
493 #ifdef AF_INET6
494 		struct in6_addr in6;
495 #endif
496 	} fail_addr;
497 	/** which fail_addr, 0 is nothing, 4 or 6 */
498 	int fail_addr_type;
499 };
500 
501 /**
502  * List of prepend items
503  */
504 struct iter_prep_list {
505 	/** next in list */
506 	struct iter_prep_list* next;
507 	/** rrset */
508 	struct ub_packed_rrset_key* rrset;
509 };
510 
511 /**
512  * Get the iterator function block.
513  * @return: function block with function pointers to iterator methods.
514  */
515 struct module_func_block* iter_get_funcblock(void);
516 
517 /**
518  * Get iterator state as a string
519  * @param state: to convert
520  * @return constant string that is printable.
521  */
522 const char* iter_state_to_string(enum iter_state state);
523 
524 /**
525  * See if iterator state is a response state
526  * @param s: to inspect
527  * @return true if response state.
528  */
529 int iter_state_is_responsestate(enum iter_state s);
530 
531 /** iterator init */
532 int iter_init(struct module_env* env, int id);
533 
534 /** iterator deinit */
535 void iter_deinit(struct module_env* env, int id);
536 
537 /** iterator operate on a query */
538 void iter_operate(struct module_qstate* qstate, enum module_ev event, int id,
539 	struct outbound_entry* outbound);
540 
541 /**
542  * Return priming query results to interested super querystates.
543  *
544  * Sets the delegation point and delegation message (not nonRD queries).
545  * This is a callback from walk_supers.
546  *
547  * @param qstate: query state that finished.
548  * @param id: module id.
549  * @param super: the qstate to inform.
550  */
551 void iter_inform_super(struct module_qstate* qstate, int id,
552 	struct module_qstate* super);
553 
554 /** iterator cleanup query state */
555 void iter_clear(struct module_qstate* qstate, int id);
556 
557 /** iterator alloc size routine */
558 size_t iter_get_mem(struct module_env* env, int id);
559 
560 #endif /* ITERATOR_ITERATOR_H */
561