xref: /freebsd/contrib/unbound/services/mesh.c (revision 0572ccaa4543b0abef8ef81e384c1d04de9f3da1)
1 /*
2  * services/mesh.c - deal with mesh of query states and handle events for that.
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 functions to assist in dealing with a mesh of
40  * query states. This mesh is supposed to be thread-specific.
41  * It consists of query states (per qname, qtype, qclass) and connections
42  * between query states and the super and subquery states, and replies to
43  * send back to clients.
44  */
45 #include "config.h"
46 #include "services/mesh.h"
47 #include "services/outbound_list.h"
48 #include "services/cache/dns.h"
49 #include "util/log.h"
50 #include "util/net_help.h"
51 #include "util/module.h"
52 #include "util/regional.h"
53 #include "util/data/msgencode.h"
54 #include "util/timehist.h"
55 #include "util/fptr_wlist.h"
56 #include "util/alloc.h"
57 #include "util/config_file.h"
58 #include "ldns/sbuffer.h"
59 
60 /** subtract timers and the values do not overflow or become negative */
61 static void
62 timeval_subtract(struct timeval* d, const struct timeval* end, const struct timeval* start)
63 {
64 #ifndef S_SPLINT_S
65 	time_t end_usec = end->tv_usec;
66 	d->tv_sec = end->tv_sec - start->tv_sec;
67 	if(end_usec < start->tv_usec) {
68 		end_usec += 1000000;
69 		d->tv_sec--;
70 	}
71 	d->tv_usec = end_usec - start->tv_usec;
72 #endif
73 }
74 
75 /** add timers and the values do not overflow or become negative */
76 static void
77 timeval_add(struct timeval* d, const struct timeval* add)
78 {
79 #ifndef S_SPLINT_S
80 	d->tv_sec += add->tv_sec;
81 	d->tv_usec += add->tv_usec;
82 	if(d->tv_usec > 1000000 ) {
83 		d->tv_usec -= 1000000;
84 		d->tv_sec++;
85 	}
86 #endif
87 }
88 
89 /** divide sum of timers to get average */
90 static void
91 timeval_divide(struct timeval* avg, const struct timeval* sum, size_t d)
92 {
93 #ifndef S_SPLINT_S
94 	size_t leftover;
95 	if(d == 0) {
96 		avg->tv_sec = 0;
97 		avg->tv_usec = 0;
98 		return;
99 	}
100 	avg->tv_sec = sum->tv_sec / d;
101 	avg->tv_usec = sum->tv_usec / d;
102 	/* handle fraction from seconds divide */
103 	leftover = sum->tv_sec - avg->tv_sec*d;
104 	avg->tv_usec += (leftover*1000000)/d;
105 #endif
106 }
107 
108 /** histogram compare of time values */
109 static int
110 timeval_smaller(const struct timeval* x, const struct timeval* y)
111 {
112 #ifndef S_SPLINT_S
113 	if(x->tv_sec < y->tv_sec)
114 		return 1;
115 	else if(x->tv_sec == y->tv_sec) {
116 		if(x->tv_usec <= y->tv_usec)
117 			return 1;
118 		else	return 0;
119 	}
120 	else	return 0;
121 #endif
122 }
123 
124 int
125 mesh_state_compare(const void* ap, const void* bp)
126 {
127 	struct mesh_state* a = (struct mesh_state*)ap;
128 	struct mesh_state* b = (struct mesh_state*)bp;
129 
130 	if(a->s.is_priming && !b->s.is_priming)
131 		return -1;
132 	if(!a->s.is_priming && b->s.is_priming)
133 		return 1;
134 
135 	if((a->s.query_flags&BIT_RD) && !(b->s.query_flags&BIT_RD))
136 		return -1;
137 	if(!(a->s.query_flags&BIT_RD) && (b->s.query_flags&BIT_RD))
138 		return 1;
139 
140 	if((a->s.query_flags&BIT_CD) && !(b->s.query_flags&BIT_CD))
141 		return -1;
142 	if(!(a->s.query_flags&BIT_CD) && (b->s.query_flags&BIT_CD))
143 		return 1;
144 
145 	return query_info_compare(&a->s.qinfo, &b->s.qinfo);
146 }
147 
148 int
149 mesh_state_ref_compare(const void* ap, const void* bp)
150 {
151 	struct mesh_state_ref* a = (struct mesh_state_ref*)ap;
152 	struct mesh_state_ref* b = (struct mesh_state_ref*)bp;
153 	return mesh_state_compare(a->s, b->s);
154 }
155 
156 struct mesh_area*
157 mesh_create(struct module_stack* stack, struct module_env* env)
158 {
159 	struct mesh_area* mesh = calloc(1, sizeof(struct mesh_area));
160 	if(!mesh) {
161 		log_err("mesh area alloc: out of memory");
162 		return NULL;
163 	}
164 	mesh->histogram = timehist_setup();
165 	mesh->qbuf_bak = sldns_buffer_new(env->cfg->msg_buffer_size);
166 	if(!mesh->histogram || !mesh->qbuf_bak) {
167 		free(mesh);
168 		log_err("mesh area alloc: out of memory");
169 		return NULL;
170 	}
171 	mesh->mods = *stack;
172 	mesh->env = env;
173 	rbtree_init(&mesh->run, &mesh_state_compare);
174 	rbtree_init(&mesh->all, &mesh_state_compare);
175 	mesh->num_reply_addrs = 0;
176 	mesh->num_reply_states = 0;
177 	mesh->num_detached_states = 0;
178 	mesh->num_forever_states = 0;
179 	mesh->stats_jostled = 0;
180 	mesh->stats_dropped = 0;
181 	mesh->max_reply_states = env->cfg->num_queries_per_thread;
182 	mesh->max_forever_states = (mesh->max_reply_states+1)/2;
183 #ifndef S_SPLINT_S
184 	mesh->jostle_max.tv_sec = (time_t)(env->cfg->jostle_time / 1000);
185 	mesh->jostle_max.tv_usec = (time_t)((env->cfg->jostle_time % 1000)
186 		*1000);
187 #endif
188 	return mesh;
189 }
190 
191 /** help mesh delete delete mesh states */
192 static void
193 mesh_delete_helper(rbnode_t* n)
194 {
195 	struct mesh_state* mstate = (struct mesh_state*)n->key;
196 	/* perform a full delete, not only 'cleanup' routine,
197 	 * because other callbacks expect a clean state in the mesh.
198 	 * For 're-entrant' calls */
199 	mesh_state_delete(&mstate->s);
200 	/* but because these delete the items from the tree, postorder
201 	 * traversal and rbtree rebalancing do not work together */
202 }
203 
204 void
205 mesh_delete(struct mesh_area* mesh)
206 {
207 	if(!mesh)
208 		return;
209 	/* free all query states */
210 	while(mesh->all.count)
211 		mesh_delete_helper(mesh->all.root);
212 	timehist_delete(mesh->histogram);
213 	sldns_buffer_free(mesh->qbuf_bak);
214 	free(mesh);
215 }
216 
217 void
218 mesh_delete_all(struct mesh_area* mesh)
219 {
220 	/* free all query states */
221 	while(mesh->all.count)
222 		mesh_delete_helper(mesh->all.root);
223 	mesh->stats_dropped += mesh->num_reply_addrs;
224 	/* clear mesh area references */
225 	rbtree_init(&mesh->run, &mesh_state_compare);
226 	rbtree_init(&mesh->all, &mesh_state_compare);
227 	mesh->num_reply_addrs = 0;
228 	mesh->num_reply_states = 0;
229 	mesh->num_detached_states = 0;
230 	mesh->num_forever_states = 0;
231 	mesh->forever_first = NULL;
232 	mesh->forever_last = NULL;
233 	mesh->jostle_first = NULL;
234 	mesh->jostle_last = NULL;
235 }
236 
237 int mesh_make_new_space(struct mesh_area* mesh, sldns_buffer* qbuf)
238 {
239 	struct mesh_state* m = mesh->jostle_first;
240 	/* free space is available */
241 	if(mesh->num_reply_states < mesh->max_reply_states)
242 		return 1;
243 	/* try to kick out a jostle-list item */
244 	if(m && m->reply_list && m->list_select == mesh_jostle_list) {
245 		/* how old is it? */
246 		struct timeval age;
247 		timeval_subtract(&age, mesh->env->now_tv,
248 			&m->reply_list->start_time);
249 		if(timeval_smaller(&mesh->jostle_max, &age)) {
250 			/* its a goner */
251 			log_nametypeclass(VERB_ALGO, "query jostled out to "
252 				"make space for a new one",
253 				m->s.qinfo.qname, m->s.qinfo.qtype,
254 				m->s.qinfo.qclass);
255 			/* backup the query */
256 			if(qbuf) sldns_buffer_copy(mesh->qbuf_bak, qbuf);
257 			/* notify supers */
258 			if(m->super_set.count > 0) {
259 				verbose(VERB_ALGO, "notify supers of failure");
260 				m->s.return_msg = NULL;
261 				m->s.return_rcode = LDNS_RCODE_SERVFAIL;
262 				mesh_walk_supers(mesh, m);
263 			}
264 			mesh->stats_jostled ++;
265 			mesh_state_delete(&m->s);
266 			/* restore the query - note that the qinfo ptr to
267 			 * the querybuffer is then correct again. */
268 			if(qbuf) sldns_buffer_copy(qbuf, mesh->qbuf_bak);
269 			return 1;
270 		}
271 	}
272 	/* no space for new item */
273 	return 0;
274 }
275 
276 void mesh_new_client(struct mesh_area* mesh, struct query_info* qinfo,
277         uint16_t qflags, struct edns_data* edns, struct comm_reply* rep,
278         uint16_t qid)
279 {
280 	/* do not use CD flag from user for mesh state, we want the CD-query
281 	 * to receive validation anyway, to protect out cache contents and
282 	 * avoid bad-data in this cache that a downstream validator cannot
283 	 * remove from this cache */
284 	struct mesh_state* s = mesh_area_find(mesh, qinfo, qflags&BIT_RD, 0);
285 	int was_detached = 0;
286 	int was_noreply = 0;
287 	int added = 0;
288 	/* does this create a new reply state? */
289 	if(!s || s->list_select == mesh_no_list) {
290 		if(!mesh_make_new_space(mesh, rep->c->buffer)) {
291 			verbose(VERB_ALGO, "Too many queries. dropping "
292 				"incoming query.");
293 			comm_point_drop_reply(rep);
294 			mesh->stats_dropped ++;
295 			return;
296 		}
297 		/* for this new reply state, the reply address is free,
298 		 * so the limit of reply addresses does not stop reply states*/
299 	} else {
300 		/* protect our memory usage from storing reply addresses */
301 		if(mesh->num_reply_addrs > mesh->max_reply_states*16) {
302 			verbose(VERB_ALGO, "Too many requests queued. "
303 				"dropping incoming query.");
304 			mesh->stats_dropped++;
305 			comm_point_drop_reply(rep);
306 			return;
307 		}
308 	}
309 	/* see if it already exists, if not, create one */
310 	if(!s) {
311 #ifdef UNBOUND_DEBUG
312 		struct rbnode_t* n;
313 #endif
314 		s = mesh_state_create(mesh->env, qinfo, qflags&BIT_RD, 0);
315 		if(!s) {
316 			log_err("mesh_state_create: out of memory; SERVFAIL");
317 			error_encode(rep->c->buffer, LDNS_RCODE_SERVFAIL,
318 				qinfo, qid, qflags, edns);
319 			comm_point_send_reply(rep);
320 			return;
321 		}
322 #ifdef UNBOUND_DEBUG
323 		n =
324 #else
325 		(void)
326 #endif
327 		rbtree_insert(&mesh->all, &s->node);
328 		log_assert(n != NULL);
329 		/* set detached (it is now) */
330 		mesh->num_detached_states++;
331 		added = 1;
332 	}
333 	if(!s->reply_list && !s->cb_list && s->super_set.count == 0)
334 		was_detached = 1;
335 	if(!s->reply_list && !s->cb_list)
336 		was_noreply = 1;
337 	/* add reply to s */
338 	if(!mesh_state_add_reply(s, edns, rep, qid, qflags, qinfo->qname)) {
339 			log_err("mesh_new_client: out of memory; SERVFAIL");
340 			error_encode(rep->c->buffer, LDNS_RCODE_SERVFAIL,
341 				qinfo, qid, qflags, edns);
342 			comm_point_send_reply(rep);
343 			if(added)
344 				mesh_state_delete(&s->s);
345 			return;
346 	}
347 	/* update statistics */
348 	if(was_detached) {
349 		log_assert(mesh->num_detached_states > 0);
350 		mesh->num_detached_states--;
351 	}
352 	if(was_noreply) {
353 		mesh->num_reply_states ++;
354 	}
355 	mesh->num_reply_addrs++;
356 	if(s->list_select == mesh_no_list) {
357 		/* move to either the forever or the jostle_list */
358 		if(mesh->num_forever_states < mesh->max_forever_states) {
359 			mesh->num_forever_states ++;
360 			mesh_list_insert(s, &mesh->forever_first,
361 				&mesh->forever_last);
362 			s->list_select = mesh_forever_list;
363 		} else {
364 			mesh_list_insert(s, &mesh->jostle_first,
365 				&mesh->jostle_last);
366 			s->list_select = mesh_jostle_list;
367 		}
368 	}
369 	if(added)
370 		mesh_run(mesh, s, module_event_new, NULL);
371 }
372 
373 int
374 mesh_new_callback(struct mesh_area* mesh, struct query_info* qinfo,
375 	uint16_t qflags, struct edns_data* edns, sldns_buffer* buf,
376 	uint16_t qid, mesh_cb_func_t cb, void* cb_arg)
377 {
378 	struct mesh_state* s = mesh_area_find(mesh, qinfo, qflags&BIT_RD, 0);
379 	int was_detached = 0;
380 	int was_noreply = 0;
381 	int added = 0;
382 	/* there are no limits on the number of callbacks */
383 
384 	/* see if it already exists, if not, create one */
385 	if(!s) {
386 #ifdef UNBOUND_DEBUG
387 		struct rbnode_t* n;
388 #endif
389 		s = mesh_state_create(mesh->env, qinfo, qflags&BIT_RD, 0);
390 		if(!s) {
391 			return 0;
392 		}
393 #ifdef UNBOUND_DEBUG
394 		n =
395 #else
396 		(void)
397 #endif
398 		rbtree_insert(&mesh->all, &s->node);
399 		log_assert(n != NULL);
400 		/* set detached (it is now) */
401 		mesh->num_detached_states++;
402 		added = 1;
403 	}
404 	if(!s->reply_list && !s->cb_list && s->super_set.count == 0)
405 		was_detached = 1;
406 	if(!s->reply_list && !s->cb_list)
407 		was_noreply = 1;
408 	/* add reply to s */
409 	if(!mesh_state_add_cb(s, edns, buf, cb, cb_arg, qid, qflags)) {
410 			if(added)
411 				mesh_state_delete(&s->s);
412 			return 0;
413 	}
414 	/* update statistics */
415 	if(was_detached) {
416 		log_assert(mesh->num_detached_states > 0);
417 		mesh->num_detached_states--;
418 	}
419 	if(was_noreply) {
420 		mesh->num_reply_states ++;
421 	}
422 	mesh->num_reply_addrs++;
423 	if(added)
424 		mesh_run(mesh, s, module_event_new, NULL);
425 	return 1;
426 }
427 
428 void mesh_new_prefetch(struct mesh_area* mesh, struct query_info* qinfo,
429         uint16_t qflags, time_t leeway)
430 {
431 	struct mesh_state* s = mesh_area_find(mesh, qinfo, qflags&BIT_RD, 0);
432 #ifdef UNBOUND_DEBUG
433 	struct rbnode_t* n;
434 #endif
435 	/* already exists, and for a different purpose perhaps.
436 	 * if mesh_no_list, keep it that way. */
437 	if(s) {
438 		/* make it ignore the cache from now on */
439 		if(!s->s.blacklist)
440 			sock_list_insert(&s->s.blacklist, NULL, 0, s->s.region);
441 		if(s->s.prefetch_leeway < leeway)
442 			s->s.prefetch_leeway = leeway;
443 		return;
444 	}
445 	if(!mesh_make_new_space(mesh, NULL)) {
446 		verbose(VERB_ALGO, "Too many queries. dropped prefetch.");
447 		mesh->stats_dropped ++;
448 		return;
449 	}
450 	s = mesh_state_create(mesh->env, qinfo, qflags&BIT_RD, 0);
451 	if(!s) {
452 		log_err("prefetch mesh_state_create: out of memory");
453 		return;
454 	}
455 #ifdef UNBOUND_DEBUG
456 	n =
457 #else
458 	(void)
459 #endif
460 	rbtree_insert(&mesh->all, &s->node);
461 	log_assert(n != NULL);
462 	/* set detached (it is now) */
463 	mesh->num_detached_states++;
464 	/* make it ignore the cache */
465 	sock_list_insert(&s->s.blacklist, NULL, 0, s->s.region);
466 	s->s.prefetch_leeway = leeway;
467 
468 	if(s->list_select == mesh_no_list) {
469 		/* move to either the forever or the jostle_list */
470 		if(mesh->num_forever_states < mesh->max_forever_states) {
471 			mesh->num_forever_states ++;
472 			mesh_list_insert(s, &mesh->forever_first,
473 				&mesh->forever_last);
474 			s->list_select = mesh_forever_list;
475 		} else {
476 			mesh_list_insert(s, &mesh->jostle_first,
477 				&mesh->jostle_last);
478 			s->list_select = mesh_jostle_list;
479 		}
480 	}
481 	mesh_run(mesh, s, module_event_new, NULL);
482 }
483 
484 void mesh_report_reply(struct mesh_area* mesh, struct outbound_entry* e,
485         struct comm_reply* reply, int what)
486 {
487 	enum module_ev event = module_event_reply;
488 	e->qstate->reply = reply;
489 	if(what != NETEVENT_NOERROR) {
490 		event = module_event_noreply;
491 		if(what == NETEVENT_CAPSFAIL)
492 			event = module_event_capsfail;
493 	}
494 	mesh_run(mesh, e->qstate->mesh_info, event, e);
495 }
496 
497 struct mesh_state*
498 mesh_state_create(struct module_env* env, struct query_info* qinfo,
499 	uint16_t qflags, int prime)
500 {
501 	struct regional* region = alloc_reg_obtain(env->alloc);
502 	struct mesh_state* mstate;
503 	int i;
504 	if(!region)
505 		return NULL;
506 	mstate = (struct mesh_state*)regional_alloc(region,
507 		sizeof(struct mesh_state));
508 	if(!mstate) {
509 		alloc_reg_release(env->alloc, region);
510 		return NULL;
511 	}
512 	memset(mstate, 0, sizeof(*mstate));
513 	mstate->node = *RBTREE_NULL;
514 	mstate->run_node = *RBTREE_NULL;
515 	mstate->node.key = mstate;
516 	mstate->run_node.key = mstate;
517 	mstate->reply_list = NULL;
518 	mstate->list_select = mesh_no_list;
519 	mstate->replies_sent = 0;
520 	rbtree_init(&mstate->super_set, &mesh_state_ref_compare);
521 	rbtree_init(&mstate->sub_set, &mesh_state_ref_compare);
522 	mstate->num_activated = 0;
523 	/* init module qstate */
524 	mstate->s.qinfo.qtype = qinfo->qtype;
525 	mstate->s.qinfo.qclass = qinfo->qclass;
526 	mstate->s.qinfo.qname_len = qinfo->qname_len;
527 	mstate->s.qinfo.qname = regional_alloc_init(region, qinfo->qname,
528 		qinfo->qname_len);
529 	if(!mstate->s.qinfo.qname) {
530 		alloc_reg_release(env->alloc, region);
531 		return NULL;
532 	}
533 	/* remove all weird bits from qflags */
534 	mstate->s.query_flags = (qflags & (BIT_RD|BIT_CD));
535 	mstate->s.is_priming = prime;
536 	mstate->s.reply = NULL;
537 	mstate->s.region = region;
538 	mstate->s.curmod = 0;
539 	mstate->s.return_msg = 0;
540 	mstate->s.return_rcode = LDNS_RCODE_NOERROR;
541 	mstate->s.env = env;
542 	mstate->s.mesh_info = mstate;
543 	mstate->s.prefetch_leeway = 0;
544 	/* init modules */
545 	for(i=0; i<env->mesh->mods.num; i++) {
546 		mstate->s.minfo[i] = NULL;
547 		mstate->s.ext_state[i] = module_state_initial;
548 	}
549 	return mstate;
550 }
551 
552 void
553 mesh_state_cleanup(struct mesh_state* mstate)
554 {
555 	struct mesh_area* mesh;
556 	int i;
557 	if(!mstate)
558 		return;
559 	mesh = mstate->s.env->mesh;
560 	/* drop unsent replies */
561 	if(!mstate->replies_sent) {
562 		struct mesh_reply* rep;
563 		struct mesh_cb* cb;
564 		for(rep=mstate->reply_list; rep; rep=rep->next) {
565 			comm_point_drop_reply(&rep->query_reply);
566 			mesh->num_reply_addrs--;
567 		}
568 		for(cb=mstate->cb_list; cb; cb=cb->next) {
569 			fptr_ok(fptr_whitelist_mesh_cb(cb->cb));
570 			(*cb->cb)(cb->cb_arg, LDNS_RCODE_SERVFAIL, NULL,
571 				sec_status_unchecked, NULL);
572 			mesh->num_reply_addrs--;
573 		}
574 	}
575 
576 	/* de-init modules */
577 	for(i=0; i<mesh->mods.num; i++) {
578 		fptr_ok(fptr_whitelist_mod_clear(mesh->mods.mod[i]->clear));
579 		(*mesh->mods.mod[i]->clear)(&mstate->s, i);
580 		mstate->s.minfo[i] = NULL;
581 		mstate->s.ext_state[i] = module_finished;
582 	}
583 	alloc_reg_release(mstate->s.env->alloc, mstate->s.region);
584 }
585 
586 void
587 mesh_state_delete(struct module_qstate* qstate)
588 {
589 	struct mesh_area* mesh;
590 	struct mesh_state_ref* super, ref;
591 	struct mesh_state* mstate;
592 	if(!qstate)
593 		return;
594 	mstate = qstate->mesh_info;
595 	mesh = mstate->s.env->mesh;
596 	mesh_detach_subs(&mstate->s);
597 	if(mstate->list_select == mesh_forever_list) {
598 		mesh->num_forever_states --;
599 		mesh_list_remove(mstate, &mesh->forever_first,
600 			&mesh->forever_last);
601 	} else if(mstate->list_select == mesh_jostle_list) {
602 		mesh_list_remove(mstate, &mesh->jostle_first,
603 			&mesh->jostle_last);
604 	}
605 	if(!mstate->reply_list && !mstate->cb_list
606 		&& mstate->super_set.count == 0) {
607 		log_assert(mesh->num_detached_states > 0);
608 		mesh->num_detached_states--;
609 	}
610 	if(mstate->reply_list || mstate->cb_list) {
611 		log_assert(mesh->num_reply_states > 0);
612 		mesh->num_reply_states--;
613 	}
614 	ref.node.key = &ref;
615 	ref.s = mstate;
616 	RBTREE_FOR(super, struct mesh_state_ref*, &mstate->super_set) {
617 		(void)rbtree_delete(&super->s->sub_set, &ref);
618 	}
619 	(void)rbtree_delete(&mesh->run, mstate);
620 	(void)rbtree_delete(&mesh->all, mstate);
621 	mesh_state_cleanup(mstate);
622 }
623 
624 /** helper recursive rbtree find routine */
625 static int
626 find_in_subsub(struct mesh_state* m, struct mesh_state* tofind, size_t *c)
627 {
628 	struct mesh_state_ref* r;
629 	if((*c)++ > MESH_MAX_SUBSUB)
630 		return 1;
631 	RBTREE_FOR(r, struct mesh_state_ref*, &m->sub_set) {
632 		if(r->s == tofind || find_in_subsub(r->s, tofind, c))
633 			return 1;
634 	}
635 	return 0;
636 }
637 
638 /** find cycle for already looked up mesh_state */
639 static int
640 mesh_detect_cycle_found(struct module_qstate* qstate, struct mesh_state* dep_m)
641 {
642 	struct mesh_state* cyc_m = qstate->mesh_info;
643 	size_t counter = 0;
644 	if(!dep_m)
645 		return 0;
646 	if(dep_m == cyc_m || find_in_subsub(dep_m, cyc_m, &counter)) {
647 		if(counter > MESH_MAX_SUBSUB)
648 			return 2;
649 		return 1;
650 	}
651 	return 0;
652 }
653 
654 void mesh_detach_subs(struct module_qstate* qstate)
655 {
656 	struct mesh_area* mesh = qstate->env->mesh;
657 	struct mesh_state_ref* ref, lookup;
658 #ifdef UNBOUND_DEBUG
659 	struct rbnode_t* n;
660 #endif
661 	lookup.node.key = &lookup;
662 	lookup.s = qstate->mesh_info;
663 	RBTREE_FOR(ref, struct mesh_state_ref*, &qstate->mesh_info->sub_set) {
664 #ifdef UNBOUND_DEBUG
665 		n =
666 #else
667 		(void)
668 #endif
669 		rbtree_delete(&ref->s->super_set, &lookup);
670 		log_assert(n != NULL); /* must have been present */
671 		if(!ref->s->reply_list && !ref->s->cb_list
672 			&& ref->s->super_set.count == 0) {
673 			mesh->num_detached_states++;
674 			log_assert(mesh->num_detached_states +
675 				mesh->num_reply_states <= mesh->all.count);
676 		}
677 	}
678 	rbtree_init(&qstate->mesh_info->sub_set, &mesh_state_ref_compare);
679 }
680 
681 int mesh_attach_sub(struct module_qstate* qstate, struct query_info* qinfo,
682         uint16_t qflags, int prime, struct module_qstate** newq)
683 {
684 	/* find it, if not, create it */
685 	struct mesh_area* mesh = qstate->env->mesh;
686 	struct mesh_state* sub = mesh_area_find(mesh, qinfo, qflags, prime);
687 	int was_detached;
688 	if(mesh_detect_cycle_found(qstate, sub)) {
689 		verbose(VERB_ALGO, "attach failed, cycle detected");
690 		return 0;
691 	}
692 	if(!sub) {
693 #ifdef UNBOUND_DEBUG
694 		struct rbnode_t* n;
695 #endif
696 		/* create a new one */
697 		sub = mesh_state_create(qstate->env, qinfo, qflags, prime);
698 		if(!sub) {
699 			log_err("mesh_attach_sub: out of memory");
700 			return 0;
701 		}
702 #ifdef UNBOUND_DEBUG
703 		n =
704 #else
705 		(void)
706 #endif
707 		rbtree_insert(&mesh->all, &sub->node);
708 		log_assert(n != NULL);
709 		/* set detached (it is now) */
710 		mesh->num_detached_states++;
711 		/* set new query state to run */
712 #ifdef UNBOUND_DEBUG
713 		n =
714 #else
715 		(void)
716 #endif
717 		rbtree_insert(&mesh->run, &sub->run_node);
718 		log_assert(n != NULL);
719 		*newq = &sub->s;
720 	} else
721 		*newq = NULL;
722 	was_detached = (sub->super_set.count == 0);
723 	if(!mesh_state_attachment(qstate->mesh_info, sub))
724 		return 0;
725 	/* if it was a duplicate  attachment, the count was not zero before */
726 	if(!sub->reply_list && !sub->cb_list && was_detached &&
727 		sub->super_set.count == 1) {
728 		/* it used to be detached, before this one got added */
729 		log_assert(mesh->num_detached_states > 0);
730 		mesh->num_detached_states--;
731 	}
732 	/* *newq will be run when inited after the current module stops */
733 	return 1;
734 }
735 
736 int mesh_state_attachment(struct mesh_state* super, struct mesh_state* sub)
737 {
738 #ifdef UNBOUND_DEBUG
739 	struct rbnode_t* n;
740 #endif
741 	struct mesh_state_ref* subref; /* points to sub, inserted in super */
742 	struct mesh_state_ref* superref; /* points to super, inserted in sub */
743 	if( !(subref = regional_alloc(super->s.region,
744 		sizeof(struct mesh_state_ref))) ||
745 		!(superref = regional_alloc(sub->s.region,
746 		sizeof(struct mesh_state_ref))) ) {
747 		log_err("mesh_state_attachment: out of memory");
748 		return 0;
749 	}
750 	superref->node.key = superref;
751 	superref->s = super;
752 	subref->node.key = subref;
753 	subref->s = sub;
754 	if(!rbtree_insert(&sub->super_set, &superref->node)) {
755 		/* this should not happen, iterator and validator do not
756 		 * attach subqueries that are identical. */
757 		/* already attached, we are done, nothing todo.
758 		 * since superref and subref already allocated in region,
759 		 * we cannot free them */
760 		return 1;
761 	}
762 #ifdef UNBOUND_DEBUG
763 	n =
764 #else
765 	(void)
766 #endif
767 	rbtree_insert(&super->sub_set, &subref->node);
768 	log_assert(n != NULL); /* we checked above if statement, the reverse
769 	  administration should not fail now, unless they are out of sync */
770 	return 1;
771 }
772 
773 /**
774  * callback results to mesh cb entry
775  * @param m: mesh state to send it for.
776  * @param rcode: if not 0, error code.
777  * @param rep: reply to send (or NULL if rcode is set).
778  * @param r: callback entry
779  */
780 static void
781 mesh_do_callback(struct mesh_state* m, int rcode, struct reply_info* rep,
782 	struct mesh_cb* r)
783 {
784 	int secure;
785 	char* reason = NULL;
786 	/* bogus messages are not made into servfail, sec_status passed
787 	 * to the callback function */
788 	if(rep && rep->security == sec_status_secure)
789 		secure = 1;
790 	else	secure = 0;
791 	if(!rep && rcode == LDNS_RCODE_NOERROR)
792 		rcode = LDNS_RCODE_SERVFAIL;
793 	if(!rcode && rep->security == sec_status_bogus) {
794 		if(!(reason = errinf_to_str(&m->s)))
795 			rcode = LDNS_RCODE_SERVFAIL;
796 	}
797 	/* send the reply */
798 	if(rcode) {
799 		fptr_ok(fptr_whitelist_mesh_cb(r->cb));
800 		(*r->cb)(r->cb_arg, rcode, r->buf, sec_status_unchecked, NULL);
801 	} else {
802 		size_t udp_size = r->edns.udp_size;
803 		sldns_buffer_clear(r->buf);
804 		r->edns.edns_version = EDNS_ADVERTISED_VERSION;
805 		r->edns.udp_size = EDNS_ADVERTISED_SIZE;
806 		r->edns.ext_rcode = 0;
807 		r->edns.bits &= EDNS_DO;
808 		if(!reply_info_answer_encode(&m->s.qinfo, rep, r->qid,
809 			r->qflags, r->buf, 0, 1,
810 			m->s.env->scratch, udp_size, &r->edns,
811 			(int)(r->edns.bits & EDNS_DO), secure))
812 		{
813 			fptr_ok(fptr_whitelist_mesh_cb(r->cb));
814 			(*r->cb)(r->cb_arg, LDNS_RCODE_SERVFAIL, r->buf,
815 				sec_status_unchecked, NULL);
816 		} else {
817 			fptr_ok(fptr_whitelist_mesh_cb(r->cb));
818 			(*r->cb)(r->cb_arg, LDNS_RCODE_NOERROR, r->buf,
819 				rep->security, reason);
820 		}
821 	}
822 	free(reason);
823 	m->s.env->mesh->num_reply_addrs--;
824 }
825 
826 /**
827  * Send reply to mesh reply entry
828  * @param m: mesh state to send it for.
829  * @param rcode: if not 0, error code.
830  * @param rep: reply to send (or NULL if rcode is set).
831  * @param r: reply entry
832  * @param prev: previous reply, already has its answer encoded in buffer.
833  */
834 static void
835 mesh_send_reply(struct mesh_state* m, int rcode, struct reply_info* rep,
836 	struct mesh_reply* r, struct mesh_reply* prev)
837 {
838 	struct timeval end_time;
839 	struct timeval duration;
840 	int secure;
841 	/* examine security status */
842 	if(m->s.env->need_to_validate && (!(r->qflags&BIT_CD) ||
843 		m->s.env->cfg->ignore_cd) && rep &&
844 		rep->security <= sec_status_bogus) {
845 		rcode = LDNS_RCODE_SERVFAIL;
846 		if(m->s.env->cfg->stat_extended)
847 			m->s.env->mesh->ans_bogus++;
848 	}
849 	if(rep && rep->security == sec_status_secure)
850 		secure = 1;
851 	else	secure = 0;
852 	if(!rep && rcode == LDNS_RCODE_NOERROR)
853 		rcode = LDNS_RCODE_SERVFAIL;
854 	/* send the reply */
855 	if(prev && prev->qflags == r->qflags &&
856 		prev->edns.edns_present == r->edns.edns_present &&
857 		prev->edns.bits == r->edns.bits &&
858 		prev->edns.udp_size == r->edns.udp_size) {
859 		/* if the previous reply is identical to this one, fix ID */
860 		if(prev->query_reply.c->buffer != r->query_reply.c->buffer)
861 			sldns_buffer_copy(r->query_reply.c->buffer,
862 				prev->query_reply.c->buffer);
863 		sldns_buffer_write_at(r->query_reply.c->buffer, 0,
864 			&r->qid, sizeof(uint16_t));
865 		sldns_buffer_write_at(r->query_reply.c->buffer, 12,
866 			r->qname, m->s.qinfo.qname_len);
867 		comm_point_send_reply(&r->query_reply);
868 	} else if(rcode) {
869 		m->s.qinfo.qname = r->qname;
870 		error_encode(r->query_reply.c->buffer, rcode, &m->s.qinfo,
871 			r->qid, r->qflags, &r->edns);
872 		comm_point_send_reply(&r->query_reply);
873 	} else {
874 		size_t udp_size = r->edns.udp_size;
875 		r->edns.edns_version = EDNS_ADVERTISED_VERSION;
876 		r->edns.udp_size = EDNS_ADVERTISED_SIZE;
877 		r->edns.ext_rcode = 0;
878 		r->edns.bits &= EDNS_DO;
879 		m->s.qinfo.qname = r->qname;
880 		if(!reply_info_answer_encode(&m->s.qinfo, rep, r->qid,
881 			r->qflags, r->query_reply.c->buffer, 0, 1,
882 			m->s.env->scratch, udp_size, &r->edns,
883 			(int)(r->edns.bits & EDNS_DO), secure))
884 		{
885 			error_encode(r->query_reply.c->buffer,
886 				LDNS_RCODE_SERVFAIL, &m->s.qinfo, r->qid,
887 				r->qflags, &r->edns);
888 		}
889 		comm_point_send_reply(&r->query_reply);
890 	}
891 	/* account */
892 	m->s.env->mesh->num_reply_addrs--;
893 	end_time = *m->s.env->now_tv;
894 	timeval_subtract(&duration, &end_time, &r->start_time);
895 	verbose(VERB_ALGO, "query took " ARG_LL "d.%6.6d sec",
896 		(long long)duration.tv_sec, (int)duration.tv_usec);
897 	m->s.env->mesh->replies_sent++;
898 	timeval_add(&m->s.env->mesh->replies_sum_wait, &duration);
899 	timehist_insert(m->s.env->mesh->histogram, &duration);
900 	if(m->s.env->cfg->stat_extended) {
901 		uint16_t rc = FLAGS_GET_RCODE(sldns_buffer_read_u16_at(r->
902 			query_reply.c->buffer, 2));
903 		if(secure) m->s.env->mesh->ans_secure++;
904 		m->s.env->mesh->ans_rcode[ rc ] ++;
905 		if(rc == 0 && LDNS_ANCOUNT(sldns_buffer_begin(r->
906 			query_reply.c->buffer)) == 0)
907 			m->s.env->mesh->ans_nodata++;
908 	}
909 }
910 
911 void mesh_query_done(struct mesh_state* mstate)
912 {
913 	struct mesh_reply* r;
914 	struct mesh_reply* prev = NULL;
915 	struct mesh_cb* c;
916 	struct reply_info* rep = (mstate->s.return_msg?
917 		mstate->s.return_msg->rep:NULL);
918 	for(r = mstate->reply_list; r; r = r->next) {
919 		mesh_send_reply(mstate, mstate->s.return_rcode, rep, r, prev);
920 		prev = r;
921 	}
922 	mstate->replies_sent = 1;
923 	for(c = mstate->cb_list; c; c = c->next) {
924 		mesh_do_callback(mstate, mstate->s.return_rcode, rep, c);
925 	}
926 }
927 
928 void mesh_walk_supers(struct mesh_area* mesh, struct mesh_state* mstate)
929 {
930 	struct mesh_state_ref* ref;
931 	RBTREE_FOR(ref, struct mesh_state_ref*, &mstate->super_set)
932 	{
933 		/* make super runnable */
934 		(void)rbtree_insert(&mesh->run, &ref->s->run_node);
935 		/* callback the function to inform super of result */
936 		fptr_ok(fptr_whitelist_mod_inform_super(
937 			mesh->mods.mod[ref->s->s.curmod]->inform_super));
938 		(*mesh->mods.mod[ref->s->s.curmod]->inform_super)(&mstate->s,
939 			ref->s->s.curmod, &ref->s->s);
940 	}
941 }
942 
943 struct mesh_state* mesh_area_find(struct mesh_area* mesh,
944 	struct query_info* qinfo, uint16_t qflags, int prime)
945 {
946 	struct mesh_state key;
947 	struct mesh_state* result;
948 
949 	key.node.key = &key;
950 	key.s.is_priming = prime;
951 	key.s.qinfo = *qinfo;
952 	key.s.query_flags = qflags;
953 
954 	result = (struct mesh_state*)rbtree_search(&mesh->all, &key);
955 	return result;
956 }
957 
958 int mesh_state_add_cb(struct mesh_state* s, struct edns_data* edns,
959         sldns_buffer* buf, mesh_cb_func_t cb, void* cb_arg,
960 	uint16_t qid, uint16_t qflags)
961 {
962 	struct mesh_cb* r = regional_alloc(s->s.region,
963 		sizeof(struct mesh_cb));
964 	if(!r)
965 		return 0;
966 	r->buf = buf;
967 	log_assert(fptr_whitelist_mesh_cb(cb)); /* early failure ifmissing*/
968 	r->cb = cb;
969 	r->cb_arg = cb_arg;
970 	r->edns = *edns;
971 	r->qid = qid;
972 	r->qflags = qflags;
973 	r->next = s->cb_list;
974 	s->cb_list = r;
975 	return 1;
976 
977 }
978 
979 int mesh_state_add_reply(struct mesh_state* s, struct edns_data* edns,
980         struct comm_reply* rep, uint16_t qid, uint16_t qflags, uint8_t* qname)
981 {
982 	struct mesh_reply* r = regional_alloc(s->s.region,
983 		sizeof(struct mesh_reply));
984 	if(!r)
985 		return 0;
986 	r->query_reply = *rep;
987 	r->edns = *edns;
988 	r->qid = qid;
989 	r->qflags = qflags;
990 	r->start_time = *s->s.env->now_tv;
991 	r->next = s->reply_list;
992 	r->qname = regional_alloc_init(s->s.region, qname,
993 		s->s.qinfo.qname_len);
994 	if(!r->qname)
995 		return 0;
996 	s->reply_list = r;
997 	return 1;
998 
999 }
1000 
1001 /**
1002  * Continue processing the mesh state at another module.
1003  * Handles module to modules tranfer of control.
1004  * Handles module finished.
1005  * @param mesh: the mesh area.
1006  * @param mstate: currently active mesh state.
1007  * 	Deleted if finished, calls _done and _supers to
1008  * 	send replies to clients and inform other mesh states.
1009  * 	This in turn may create additional runnable mesh states.
1010  * @param s: state at which the current module exited.
1011  * @param ev: the event sent to the module.
1012  * 	returned is the event to send to the next module.
1013  * @return true if continue processing at the new module.
1014  * 	false if not continued processing is needed.
1015  */
1016 static int
1017 mesh_continue(struct mesh_area* mesh, struct mesh_state* mstate,
1018 	enum module_ext_state s, enum module_ev* ev)
1019 {
1020 	mstate->num_activated++;
1021 	if(mstate->num_activated > MESH_MAX_ACTIVATION) {
1022 		/* module is looping. Stop it. */
1023 		log_err("internal error: looping module stopped");
1024 		log_query_info(VERB_QUERY, "pass error for qstate",
1025 			&mstate->s.qinfo);
1026 		s = module_error;
1027 	}
1028 	if(s == module_wait_module || s == module_restart_next) {
1029 		/* start next module */
1030 		mstate->s.curmod++;
1031 		if(mesh->mods.num == mstate->s.curmod) {
1032 			log_err("Cannot pass to next module; at last module");
1033 			log_query_info(VERB_QUERY, "pass error for qstate",
1034 				&mstate->s.qinfo);
1035 			mstate->s.curmod--;
1036 			return mesh_continue(mesh, mstate, module_error, ev);
1037 		}
1038 		if(s == module_restart_next) {
1039 			fptr_ok(fptr_whitelist_mod_clear(
1040 				mesh->mods.mod[mstate->s.curmod]->clear));
1041 			(*mesh->mods.mod[mstate->s.curmod]->clear)
1042 				(&mstate->s, mstate->s.curmod);
1043 			mstate->s.minfo[mstate->s.curmod] = NULL;
1044 		}
1045 		*ev = module_event_pass;
1046 		return 1;
1047 	}
1048 	if(s == module_error && mstate->s.return_rcode == LDNS_RCODE_NOERROR) {
1049 		/* error is bad, handle pass back up below */
1050 		mstate->s.return_rcode = LDNS_RCODE_SERVFAIL;
1051 	}
1052 	if(s == module_error || s == module_finished) {
1053 		if(mstate->s.curmod == 0) {
1054 			mesh_query_done(mstate);
1055 			mesh_walk_supers(mesh, mstate);
1056 			mesh_state_delete(&mstate->s);
1057 			return 0;
1058 		}
1059 		/* pass along the locus of control */
1060 		mstate->s.curmod --;
1061 		*ev = module_event_moddone;
1062 		return 1;
1063 	}
1064 	return 0;
1065 }
1066 
1067 void mesh_run(struct mesh_area* mesh, struct mesh_state* mstate,
1068 	enum module_ev ev, struct outbound_entry* e)
1069 {
1070 	enum module_ext_state s;
1071 	verbose(VERB_ALGO, "mesh_run: start");
1072 	while(mstate) {
1073 		/* run the module */
1074 		fptr_ok(fptr_whitelist_mod_operate(
1075 			mesh->mods.mod[mstate->s.curmod]->operate));
1076 		(*mesh->mods.mod[mstate->s.curmod]->operate)
1077 			(&mstate->s, ev, mstate->s.curmod, e);
1078 
1079 		/* examine results */
1080 		mstate->s.reply = NULL;
1081 		regional_free_all(mstate->s.env->scratch);
1082 		s = mstate->s.ext_state[mstate->s.curmod];
1083 		verbose(VERB_ALGO, "mesh_run: %s module exit state is %s",
1084 			mesh->mods.mod[mstate->s.curmod]->name, strextstate(s));
1085 		e = NULL;
1086 		if(mesh_continue(mesh, mstate, s, &ev))
1087 			continue;
1088 
1089 		/* run more modules */
1090 		ev = module_event_pass;
1091 		if(mesh->run.count > 0) {
1092 			/* pop random element off the runnable tree */
1093 			mstate = (struct mesh_state*)mesh->run.root->key;
1094 			(void)rbtree_delete(&mesh->run, mstate);
1095 		} else mstate = NULL;
1096 	}
1097 	if(verbosity >= VERB_ALGO) {
1098 		mesh_stats(mesh, "mesh_run: end");
1099 		mesh_log_list(mesh);
1100 	}
1101 }
1102 
1103 void
1104 mesh_log_list(struct mesh_area* mesh)
1105 {
1106 	char buf[30];
1107 	struct mesh_state* m;
1108 	int num = 0;
1109 	RBTREE_FOR(m, struct mesh_state*, &mesh->all) {
1110 		snprintf(buf, sizeof(buf), "%d%s%s%s%s%s mod%d %s%s",
1111 			num++, (m->s.is_priming)?"p":"",  /* prime */
1112 			(m->s.query_flags&BIT_RD)?"RD":"",
1113 			(m->s.query_flags&BIT_CD)?"CD":"",
1114 			(m->super_set.count==0)?"d":"", /* detached */
1115 			(m->sub_set.count!=0)?"c":"",  /* children */
1116 			m->s.curmod, (m->reply_list)?"rep":"", /*hasreply*/
1117 			(m->cb_list)?"cb":"" /* callbacks */
1118 			);
1119 		log_query_info(VERB_ALGO, buf, &m->s.qinfo);
1120 	}
1121 }
1122 
1123 void
1124 mesh_stats(struct mesh_area* mesh, const char* str)
1125 {
1126 	verbose(VERB_DETAIL, "%s %u recursion states (%u with reply, "
1127 		"%u detached), %u waiting replies, %u recursion replies "
1128 		"sent, %d replies dropped, %d states jostled out",
1129 		str, (unsigned)mesh->all.count,
1130 		(unsigned)mesh->num_reply_states,
1131 		(unsigned)mesh->num_detached_states,
1132 		(unsigned)mesh->num_reply_addrs,
1133 		(unsigned)mesh->replies_sent,
1134 		(unsigned)mesh->stats_dropped,
1135 		(unsigned)mesh->stats_jostled);
1136 	if(mesh->replies_sent > 0) {
1137 		struct timeval avg;
1138 		timeval_divide(&avg, &mesh->replies_sum_wait,
1139 			mesh->replies_sent);
1140 		log_info("average recursion processing time "
1141 			ARG_LL "d.%6.6d sec",
1142 			(long long)avg.tv_sec, (int)avg.tv_usec);
1143 		log_info("histogram of recursion processing times");
1144 		timehist_log(mesh->histogram, "recursions");
1145 	}
1146 }
1147 
1148 void
1149 mesh_stats_clear(struct mesh_area* mesh)
1150 {
1151 	if(!mesh)
1152 		return;
1153 	mesh->replies_sent = 0;
1154 	mesh->replies_sum_wait.tv_sec = 0;
1155 	mesh->replies_sum_wait.tv_usec = 0;
1156 	mesh->stats_jostled = 0;
1157 	mesh->stats_dropped = 0;
1158 	timehist_clear(mesh->histogram);
1159 	mesh->ans_secure = 0;
1160 	mesh->ans_bogus = 0;
1161 	memset(&mesh->ans_rcode[0], 0, sizeof(size_t)*16);
1162 	mesh->ans_nodata = 0;
1163 }
1164 
1165 size_t
1166 mesh_get_mem(struct mesh_area* mesh)
1167 {
1168 	struct mesh_state* m;
1169 	size_t s = sizeof(*mesh) + sizeof(struct timehist) +
1170 		sizeof(struct th_buck)*mesh->histogram->num +
1171 		sizeof(sldns_buffer) + sldns_buffer_capacity(mesh->qbuf_bak);
1172 	RBTREE_FOR(m, struct mesh_state*, &mesh->all) {
1173 		/* all, including m itself allocated in qstate region */
1174 		s += regional_get_mem(m->s.region);
1175 	}
1176 	return s;
1177 }
1178 
1179 int
1180 mesh_detect_cycle(struct module_qstate* qstate, struct query_info* qinfo,
1181 	uint16_t flags, int prime)
1182 {
1183 	struct mesh_area* mesh = qstate->env->mesh;
1184 	struct mesh_state* dep_m = mesh_area_find(mesh, qinfo, flags, prime);
1185 	return mesh_detect_cycle_found(qstate, dep_m);
1186 }
1187 
1188 void mesh_list_insert(struct mesh_state* m, struct mesh_state** fp,
1189         struct mesh_state** lp)
1190 {
1191 	/* insert as last element */
1192 	m->prev = *lp;
1193 	m->next = NULL;
1194 	if(*lp)
1195 		(*lp)->next = m;
1196 	else	*fp = m;
1197 	*lp = m;
1198 }
1199 
1200 void mesh_list_remove(struct mesh_state* m, struct mesh_state** fp,
1201         struct mesh_state** lp)
1202 {
1203 	if(m->next)
1204 		m->next->prev = m->prev;
1205 	else	*lp = m->prev;
1206 	if(m->prev)
1207 		m->prev->next = m->next;
1208 	else	*fp = m->next;
1209 }
1210