1238106Sdes/* 2238106Sdes * services/mesh.c - deal with mesh of query states and handle events for that. 3238106Sdes * 4238106Sdes * Copyright (c) 2007, NLnet Labs. All rights reserved. 5238106Sdes * 6238106Sdes * This software is open source. 7238106Sdes * 8238106Sdes * Redistribution and use in source and binary forms, with or without 9238106Sdes * modification, are permitted provided that the following conditions 10238106Sdes * are met: 11238106Sdes * 12238106Sdes * Redistributions of source code must retain the above copyright notice, 13238106Sdes * this list of conditions and the following disclaimer. 14238106Sdes * 15238106Sdes * Redistributions in binary form must reproduce the above copyright notice, 16238106Sdes * this list of conditions and the following disclaimer in the documentation 17238106Sdes * and/or other materials provided with the distribution. 18238106Sdes * 19238106Sdes * Neither the name of the NLNET LABS nor the names of its contributors may 20238106Sdes * be used to endorse or promote products derived from this software without 21238106Sdes * specific prior written permission. 22238106Sdes * 23238106Sdes * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 24269257Sdes * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 25269257Sdes * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 26269257Sdes * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 27269257Sdes * HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 28269257Sdes * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED 29269257Sdes * TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR 30269257Sdes * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF 31269257Sdes * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING 32269257Sdes * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 33269257Sdes * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 34238106Sdes */ 35238106Sdes 36238106Sdes/** 37238106Sdes * \file 38238106Sdes * 39238106Sdes * This file contains functions to assist in dealing with a mesh of 40238106Sdes * query states. This mesh is supposed to be thread-specific. 41238106Sdes * It consists of query states (per qname, qtype, qclass) and connections 42238106Sdes * between query states and the super and subquery states, and replies to 43238106Sdes * send back to clients. 44238106Sdes */ 45238106Sdes#include "config.h" 46238106Sdes#include "services/mesh.h" 47238106Sdes#include "services/outbound_list.h" 48238106Sdes#include "services/cache/dns.h" 49238106Sdes#include "util/log.h" 50238106Sdes#include "util/net_help.h" 51238106Sdes#include "util/module.h" 52238106Sdes#include "util/regional.h" 53238106Sdes#include "util/data/msgencode.h" 54238106Sdes#include "util/timehist.h" 55238106Sdes#include "util/fptr_wlist.h" 56238106Sdes#include "util/alloc.h" 57238106Sdes#include "util/config_file.h" 58269257Sdes#include "ldns/sbuffer.h" 59238106Sdes 60238106Sdes/** subtract timers and the values do not overflow or become negative */ 61238106Sdesstatic void 62238106Sdestimeval_subtract(struct timeval* d, const struct timeval* end, const struct timeval* start) 63238106Sdes{ 64238106Sdes#ifndef S_SPLINT_S 65238106Sdes time_t end_usec = end->tv_usec; 66238106Sdes d->tv_sec = end->tv_sec - start->tv_sec; 67238106Sdes if(end_usec < start->tv_usec) { 68238106Sdes end_usec += 1000000; 69238106Sdes d->tv_sec--; 70238106Sdes } 71238106Sdes d->tv_usec = end_usec - start->tv_usec; 72238106Sdes#endif 73238106Sdes} 74238106Sdes 75238106Sdes/** add timers and the values do not overflow or become negative */ 76238106Sdesstatic void 77238106Sdestimeval_add(struct timeval* d, const struct timeval* add) 78238106Sdes{ 79238106Sdes#ifndef S_SPLINT_S 80238106Sdes d->tv_sec += add->tv_sec; 81238106Sdes d->tv_usec += add->tv_usec; 82238106Sdes if(d->tv_usec > 1000000 ) { 83238106Sdes d->tv_usec -= 1000000; 84238106Sdes d->tv_sec++; 85238106Sdes } 86238106Sdes#endif 87238106Sdes} 88238106Sdes 89238106Sdes/** divide sum of timers to get average */ 90238106Sdesstatic void 91238106Sdestimeval_divide(struct timeval* avg, const struct timeval* sum, size_t d) 92238106Sdes{ 93238106Sdes#ifndef S_SPLINT_S 94238106Sdes size_t leftover; 95238106Sdes if(d == 0) { 96238106Sdes avg->tv_sec = 0; 97238106Sdes avg->tv_usec = 0; 98238106Sdes return; 99238106Sdes } 100238106Sdes avg->tv_sec = sum->tv_sec / d; 101238106Sdes avg->tv_usec = sum->tv_usec / d; 102238106Sdes /* handle fraction from seconds divide */ 103238106Sdes leftover = sum->tv_sec - avg->tv_sec*d; 104238106Sdes avg->tv_usec += (leftover*1000000)/d; 105238106Sdes#endif 106238106Sdes} 107238106Sdes 108238106Sdes/** histogram compare of time values */ 109238106Sdesstatic int 110238106Sdestimeval_smaller(const struct timeval* x, const struct timeval* y) 111238106Sdes{ 112238106Sdes#ifndef S_SPLINT_S 113238106Sdes if(x->tv_sec < y->tv_sec) 114238106Sdes return 1; 115238106Sdes else if(x->tv_sec == y->tv_sec) { 116238106Sdes if(x->tv_usec <= y->tv_usec) 117238106Sdes return 1; 118238106Sdes else return 0; 119238106Sdes } 120238106Sdes else return 0; 121238106Sdes#endif 122238106Sdes} 123238106Sdes 124238106Sdesint 125238106Sdesmesh_state_compare(const void* ap, const void* bp) 126238106Sdes{ 127238106Sdes struct mesh_state* a = (struct mesh_state*)ap; 128238106Sdes struct mesh_state* b = (struct mesh_state*)bp; 129238106Sdes 130238106Sdes if(a->s.is_priming && !b->s.is_priming) 131238106Sdes return -1; 132238106Sdes if(!a->s.is_priming && b->s.is_priming) 133238106Sdes return 1; 134238106Sdes 135238106Sdes if((a->s.query_flags&BIT_RD) && !(b->s.query_flags&BIT_RD)) 136238106Sdes return -1; 137238106Sdes if(!(a->s.query_flags&BIT_RD) && (b->s.query_flags&BIT_RD)) 138238106Sdes return 1; 139238106Sdes 140238106Sdes if((a->s.query_flags&BIT_CD) && !(b->s.query_flags&BIT_CD)) 141238106Sdes return -1; 142238106Sdes if(!(a->s.query_flags&BIT_CD) && (b->s.query_flags&BIT_CD)) 143238106Sdes return 1; 144238106Sdes 145238106Sdes return query_info_compare(&a->s.qinfo, &b->s.qinfo); 146238106Sdes} 147238106Sdes 148238106Sdesint 149238106Sdesmesh_state_ref_compare(const void* ap, const void* bp) 150238106Sdes{ 151238106Sdes struct mesh_state_ref* a = (struct mesh_state_ref*)ap; 152238106Sdes struct mesh_state_ref* b = (struct mesh_state_ref*)bp; 153238106Sdes return mesh_state_compare(a->s, b->s); 154238106Sdes} 155238106Sdes 156238106Sdesstruct mesh_area* 157238106Sdesmesh_create(struct module_stack* stack, struct module_env* env) 158238106Sdes{ 159238106Sdes struct mesh_area* mesh = calloc(1, sizeof(struct mesh_area)); 160238106Sdes if(!mesh) { 161238106Sdes log_err("mesh area alloc: out of memory"); 162238106Sdes return NULL; 163238106Sdes } 164238106Sdes mesh->histogram = timehist_setup(); 165269257Sdes mesh->qbuf_bak = sldns_buffer_new(env->cfg->msg_buffer_size); 166238106Sdes if(!mesh->histogram || !mesh->qbuf_bak) { 167238106Sdes free(mesh); 168238106Sdes log_err("mesh area alloc: out of memory"); 169238106Sdes return NULL; 170238106Sdes } 171238106Sdes mesh->mods = *stack; 172238106Sdes mesh->env = env; 173238106Sdes rbtree_init(&mesh->run, &mesh_state_compare); 174238106Sdes rbtree_init(&mesh->all, &mesh_state_compare); 175238106Sdes mesh->num_reply_addrs = 0; 176238106Sdes mesh->num_reply_states = 0; 177238106Sdes mesh->num_detached_states = 0; 178238106Sdes mesh->num_forever_states = 0; 179238106Sdes mesh->stats_jostled = 0; 180238106Sdes mesh->stats_dropped = 0; 181238106Sdes mesh->max_reply_states = env->cfg->num_queries_per_thread; 182238106Sdes mesh->max_forever_states = (mesh->max_reply_states+1)/2; 183238106Sdes#ifndef S_SPLINT_S 184238106Sdes mesh->jostle_max.tv_sec = (time_t)(env->cfg->jostle_time / 1000); 185238106Sdes mesh->jostle_max.tv_usec = (time_t)((env->cfg->jostle_time % 1000) 186238106Sdes *1000); 187238106Sdes#endif 188238106Sdes return mesh; 189238106Sdes} 190238106Sdes 191238106Sdes/** help mesh delete delete mesh states */ 192238106Sdesstatic void 193238106Sdesmesh_delete_helper(rbnode_t* n) 194238106Sdes{ 195238106Sdes struct mesh_state* mstate = (struct mesh_state*)n->key; 196238106Sdes /* perform a full delete, not only 'cleanup' routine, 197238106Sdes * because other callbacks expect a clean state in the mesh. 198238106Sdes * For 're-entrant' calls */ 199238106Sdes mesh_state_delete(&mstate->s); 200238106Sdes /* but because these delete the items from the tree, postorder 201238106Sdes * traversal and rbtree rebalancing do not work together */ 202238106Sdes} 203238106Sdes 204238106Sdesvoid 205238106Sdesmesh_delete(struct mesh_area* mesh) 206238106Sdes{ 207238106Sdes if(!mesh) 208238106Sdes return; 209238106Sdes /* free all query states */ 210238106Sdes while(mesh->all.count) 211238106Sdes mesh_delete_helper(mesh->all.root); 212238106Sdes timehist_delete(mesh->histogram); 213269257Sdes sldns_buffer_free(mesh->qbuf_bak); 214238106Sdes free(mesh); 215238106Sdes} 216238106Sdes 217238106Sdesvoid 218238106Sdesmesh_delete_all(struct mesh_area* mesh) 219238106Sdes{ 220238106Sdes /* free all query states */ 221238106Sdes while(mesh->all.count) 222238106Sdes mesh_delete_helper(mesh->all.root); 223238106Sdes mesh->stats_dropped += mesh->num_reply_addrs; 224238106Sdes /* clear mesh area references */ 225238106Sdes rbtree_init(&mesh->run, &mesh_state_compare); 226238106Sdes rbtree_init(&mesh->all, &mesh_state_compare); 227238106Sdes mesh->num_reply_addrs = 0; 228238106Sdes mesh->num_reply_states = 0; 229238106Sdes mesh->num_detached_states = 0; 230238106Sdes mesh->num_forever_states = 0; 231238106Sdes mesh->forever_first = NULL; 232238106Sdes mesh->forever_last = NULL; 233238106Sdes mesh->jostle_first = NULL; 234238106Sdes mesh->jostle_last = NULL; 235238106Sdes} 236238106Sdes 237269257Sdesint mesh_make_new_space(struct mesh_area* mesh, sldns_buffer* qbuf) 238238106Sdes{ 239238106Sdes struct mesh_state* m = mesh->jostle_first; 240238106Sdes /* free space is available */ 241238106Sdes if(mesh->num_reply_states < mesh->max_reply_states) 242238106Sdes return 1; 243238106Sdes /* try to kick out a jostle-list item */ 244238106Sdes if(m && m->reply_list && m->list_select == mesh_jostle_list) { 245238106Sdes /* how old is it? */ 246238106Sdes struct timeval age; 247238106Sdes timeval_subtract(&age, mesh->env->now_tv, 248238106Sdes &m->reply_list->start_time); 249238106Sdes if(timeval_smaller(&mesh->jostle_max, &age)) { 250238106Sdes /* its a goner */ 251238106Sdes log_nametypeclass(VERB_ALGO, "query jostled out to " 252238106Sdes "make space for a new one", 253238106Sdes m->s.qinfo.qname, m->s.qinfo.qtype, 254238106Sdes m->s.qinfo.qclass); 255238106Sdes /* backup the query */ 256269257Sdes if(qbuf) sldns_buffer_copy(mesh->qbuf_bak, qbuf); 257238106Sdes /* notify supers */ 258238106Sdes if(m->super_set.count > 0) { 259238106Sdes verbose(VERB_ALGO, "notify supers of failure"); 260238106Sdes m->s.return_msg = NULL; 261238106Sdes m->s.return_rcode = LDNS_RCODE_SERVFAIL; 262238106Sdes mesh_walk_supers(mesh, m); 263238106Sdes } 264238106Sdes mesh->stats_jostled ++; 265238106Sdes mesh_state_delete(&m->s); 266238106Sdes /* restore the query - note that the qinfo ptr to 267238106Sdes * the querybuffer is then correct again. */ 268269257Sdes if(qbuf) sldns_buffer_copy(qbuf, mesh->qbuf_bak); 269238106Sdes return 1; 270238106Sdes } 271238106Sdes } 272238106Sdes /* no space for new item */ 273238106Sdes return 0; 274238106Sdes} 275238106Sdes 276238106Sdesvoid mesh_new_client(struct mesh_area* mesh, struct query_info* qinfo, 277238106Sdes uint16_t qflags, struct edns_data* edns, struct comm_reply* rep, 278238106Sdes uint16_t qid) 279238106Sdes{ 280238106Sdes /* do not use CD flag from user for mesh state, we want the CD-query 281238106Sdes * to receive validation anyway, to protect out cache contents and 282238106Sdes * avoid bad-data in this cache that a downstream validator cannot 283238106Sdes * remove from this cache */ 284238106Sdes struct mesh_state* s = mesh_area_find(mesh, qinfo, qflags&BIT_RD, 0); 285238106Sdes int was_detached = 0; 286238106Sdes int was_noreply = 0; 287238106Sdes int added = 0; 288238106Sdes /* does this create a new reply state? */ 289238106Sdes if(!s || s->list_select == mesh_no_list) { 290238106Sdes if(!mesh_make_new_space(mesh, rep->c->buffer)) { 291238106Sdes verbose(VERB_ALGO, "Too many queries. dropping " 292238106Sdes "incoming query."); 293238106Sdes comm_point_drop_reply(rep); 294238106Sdes mesh->stats_dropped ++; 295238106Sdes return; 296238106Sdes } 297238106Sdes /* for this new reply state, the reply address is free, 298238106Sdes * so the limit of reply addresses does not stop reply states*/ 299238106Sdes } else { 300238106Sdes /* protect our memory usage from storing reply addresses */ 301238106Sdes if(mesh->num_reply_addrs > mesh->max_reply_states*16) { 302238106Sdes verbose(VERB_ALGO, "Too many requests queued. " 303238106Sdes "dropping incoming query."); 304238106Sdes mesh->stats_dropped++; 305238106Sdes comm_point_drop_reply(rep); 306238106Sdes return; 307238106Sdes } 308238106Sdes } 309238106Sdes /* see if it already exists, if not, create one */ 310238106Sdes if(!s) { 311238106Sdes#ifdef UNBOUND_DEBUG 312238106Sdes struct rbnode_t* n; 313238106Sdes#endif 314238106Sdes s = mesh_state_create(mesh->env, qinfo, qflags&BIT_RD, 0); 315238106Sdes if(!s) { 316238106Sdes log_err("mesh_state_create: out of memory; SERVFAIL"); 317238106Sdes error_encode(rep->c->buffer, LDNS_RCODE_SERVFAIL, 318238106Sdes qinfo, qid, qflags, edns); 319238106Sdes comm_point_send_reply(rep); 320238106Sdes return; 321238106Sdes } 322238106Sdes#ifdef UNBOUND_DEBUG 323238106Sdes n = 324269257Sdes#else 325269257Sdes (void) 326238106Sdes#endif 327238106Sdes rbtree_insert(&mesh->all, &s->node); 328238106Sdes log_assert(n != NULL); 329238106Sdes /* set detached (it is now) */ 330238106Sdes mesh->num_detached_states++; 331238106Sdes added = 1; 332238106Sdes } 333238106Sdes if(!s->reply_list && !s->cb_list && s->super_set.count == 0) 334238106Sdes was_detached = 1; 335238106Sdes if(!s->reply_list && !s->cb_list) 336238106Sdes was_noreply = 1; 337238106Sdes /* add reply to s */ 338238106Sdes if(!mesh_state_add_reply(s, edns, rep, qid, qflags, qinfo->qname)) { 339238106Sdes log_err("mesh_new_client: out of memory; SERVFAIL"); 340238106Sdes error_encode(rep->c->buffer, LDNS_RCODE_SERVFAIL, 341238106Sdes qinfo, qid, qflags, edns); 342238106Sdes comm_point_send_reply(rep); 343238106Sdes if(added) 344238106Sdes mesh_state_delete(&s->s); 345238106Sdes return; 346238106Sdes } 347238106Sdes /* update statistics */ 348238106Sdes if(was_detached) { 349238106Sdes log_assert(mesh->num_detached_states > 0); 350238106Sdes mesh->num_detached_states--; 351238106Sdes } 352238106Sdes if(was_noreply) { 353238106Sdes mesh->num_reply_states ++; 354238106Sdes } 355238106Sdes mesh->num_reply_addrs++; 356238106Sdes if(s->list_select == mesh_no_list) { 357238106Sdes /* move to either the forever or the jostle_list */ 358238106Sdes if(mesh->num_forever_states < mesh->max_forever_states) { 359238106Sdes mesh->num_forever_states ++; 360238106Sdes mesh_list_insert(s, &mesh->forever_first, 361238106Sdes &mesh->forever_last); 362238106Sdes s->list_select = mesh_forever_list; 363238106Sdes } else { 364238106Sdes mesh_list_insert(s, &mesh->jostle_first, 365238106Sdes &mesh->jostle_last); 366238106Sdes s->list_select = mesh_jostle_list; 367238106Sdes } 368238106Sdes } 369238106Sdes if(added) 370238106Sdes mesh_run(mesh, s, module_event_new, NULL); 371238106Sdes} 372238106Sdes 373238106Sdesint 374238106Sdesmesh_new_callback(struct mesh_area* mesh, struct query_info* qinfo, 375269257Sdes uint16_t qflags, struct edns_data* edns, sldns_buffer* buf, 376238106Sdes uint16_t qid, mesh_cb_func_t cb, void* cb_arg) 377238106Sdes{ 378238106Sdes struct mesh_state* s = mesh_area_find(mesh, qinfo, qflags&BIT_RD, 0); 379238106Sdes int was_detached = 0; 380238106Sdes int was_noreply = 0; 381238106Sdes int added = 0; 382238106Sdes /* there are no limits on the number of callbacks */ 383238106Sdes 384238106Sdes /* see if it already exists, if not, create one */ 385238106Sdes if(!s) { 386238106Sdes#ifdef UNBOUND_DEBUG 387238106Sdes struct rbnode_t* n; 388238106Sdes#endif 389238106Sdes s = mesh_state_create(mesh->env, qinfo, qflags&BIT_RD, 0); 390238106Sdes if(!s) { 391238106Sdes return 0; 392238106Sdes } 393238106Sdes#ifdef UNBOUND_DEBUG 394238106Sdes n = 395269257Sdes#else 396269257Sdes (void) 397238106Sdes#endif 398238106Sdes rbtree_insert(&mesh->all, &s->node); 399238106Sdes log_assert(n != NULL); 400238106Sdes /* set detached (it is now) */ 401238106Sdes mesh->num_detached_states++; 402238106Sdes added = 1; 403238106Sdes } 404238106Sdes if(!s->reply_list && !s->cb_list && s->super_set.count == 0) 405238106Sdes was_detached = 1; 406238106Sdes if(!s->reply_list && !s->cb_list) 407238106Sdes was_noreply = 1; 408238106Sdes /* add reply to s */ 409238106Sdes if(!mesh_state_add_cb(s, edns, buf, cb, cb_arg, qid, qflags)) { 410238106Sdes if(added) 411238106Sdes mesh_state_delete(&s->s); 412238106Sdes return 0; 413238106Sdes } 414238106Sdes /* update statistics */ 415238106Sdes if(was_detached) { 416238106Sdes log_assert(mesh->num_detached_states > 0); 417238106Sdes mesh->num_detached_states--; 418238106Sdes } 419238106Sdes if(was_noreply) { 420238106Sdes mesh->num_reply_states ++; 421238106Sdes } 422238106Sdes mesh->num_reply_addrs++; 423238106Sdes if(added) 424238106Sdes mesh_run(mesh, s, module_event_new, NULL); 425238106Sdes return 1; 426238106Sdes} 427238106Sdes 428238106Sdesvoid mesh_new_prefetch(struct mesh_area* mesh, struct query_info* qinfo, 429269257Sdes uint16_t qflags, time_t leeway) 430238106Sdes{ 431238106Sdes struct mesh_state* s = mesh_area_find(mesh, qinfo, qflags&BIT_RD, 0); 432238106Sdes#ifdef UNBOUND_DEBUG 433238106Sdes struct rbnode_t* n; 434238106Sdes#endif 435238106Sdes /* already exists, and for a different purpose perhaps. 436238106Sdes * if mesh_no_list, keep it that way. */ 437238106Sdes if(s) { 438238106Sdes /* make it ignore the cache from now on */ 439238106Sdes if(!s->s.blacklist) 440238106Sdes sock_list_insert(&s->s.blacklist, NULL, 0, s->s.region); 441238106Sdes if(s->s.prefetch_leeway < leeway) 442238106Sdes s->s.prefetch_leeway = leeway; 443238106Sdes return; 444238106Sdes } 445238106Sdes if(!mesh_make_new_space(mesh, NULL)) { 446238106Sdes verbose(VERB_ALGO, "Too many queries. dropped prefetch."); 447238106Sdes mesh->stats_dropped ++; 448238106Sdes return; 449238106Sdes } 450238106Sdes s = mesh_state_create(mesh->env, qinfo, qflags&BIT_RD, 0); 451238106Sdes if(!s) { 452238106Sdes log_err("prefetch mesh_state_create: out of memory"); 453238106Sdes return; 454238106Sdes } 455238106Sdes#ifdef UNBOUND_DEBUG 456238106Sdes n = 457269257Sdes#else 458269257Sdes (void) 459238106Sdes#endif 460238106Sdes rbtree_insert(&mesh->all, &s->node); 461238106Sdes log_assert(n != NULL); 462238106Sdes /* set detached (it is now) */ 463238106Sdes mesh->num_detached_states++; 464238106Sdes /* make it ignore the cache */ 465238106Sdes sock_list_insert(&s->s.blacklist, NULL, 0, s->s.region); 466238106Sdes s->s.prefetch_leeway = leeway; 467238106Sdes 468238106Sdes if(s->list_select == mesh_no_list) { 469238106Sdes /* move to either the forever or the jostle_list */ 470238106Sdes if(mesh->num_forever_states < mesh->max_forever_states) { 471238106Sdes mesh->num_forever_states ++; 472238106Sdes mesh_list_insert(s, &mesh->forever_first, 473238106Sdes &mesh->forever_last); 474238106Sdes s->list_select = mesh_forever_list; 475238106Sdes } else { 476238106Sdes mesh_list_insert(s, &mesh->jostle_first, 477238106Sdes &mesh->jostle_last); 478238106Sdes s->list_select = mesh_jostle_list; 479238106Sdes } 480238106Sdes } 481238106Sdes mesh_run(mesh, s, module_event_new, NULL); 482238106Sdes} 483238106Sdes 484238106Sdesvoid mesh_report_reply(struct mesh_area* mesh, struct outbound_entry* e, 485238106Sdes struct comm_reply* reply, int what) 486238106Sdes{ 487238106Sdes enum module_ev event = module_event_reply; 488238106Sdes e->qstate->reply = reply; 489238106Sdes if(what != NETEVENT_NOERROR) { 490238106Sdes event = module_event_noreply; 491238106Sdes if(what == NETEVENT_CAPSFAIL) 492238106Sdes event = module_event_capsfail; 493238106Sdes } 494238106Sdes mesh_run(mesh, e->qstate->mesh_info, event, e); 495238106Sdes} 496238106Sdes 497238106Sdesstruct mesh_state* 498238106Sdesmesh_state_create(struct module_env* env, struct query_info* qinfo, 499238106Sdes uint16_t qflags, int prime) 500238106Sdes{ 501238106Sdes struct regional* region = alloc_reg_obtain(env->alloc); 502238106Sdes struct mesh_state* mstate; 503238106Sdes int i; 504238106Sdes if(!region) 505238106Sdes return NULL; 506238106Sdes mstate = (struct mesh_state*)regional_alloc(region, 507238106Sdes sizeof(struct mesh_state)); 508238106Sdes if(!mstate) { 509238106Sdes alloc_reg_release(env->alloc, region); 510238106Sdes return NULL; 511238106Sdes } 512238106Sdes memset(mstate, 0, sizeof(*mstate)); 513238106Sdes mstate->node = *RBTREE_NULL; 514238106Sdes mstate->run_node = *RBTREE_NULL; 515238106Sdes mstate->node.key = mstate; 516238106Sdes mstate->run_node.key = mstate; 517238106Sdes mstate->reply_list = NULL; 518238106Sdes mstate->list_select = mesh_no_list; 519238106Sdes mstate->replies_sent = 0; 520238106Sdes rbtree_init(&mstate->super_set, &mesh_state_ref_compare); 521238106Sdes rbtree_init(&mstate->sub_set, &mesh_state_ref_compare); 522238106Sdes mstate->num_activated = 0; 523238106Sdes /* init module qstate */ 524238106Sdes mstate->s.qinfo.qtype = qinfo->qtype; 525238106Sdes mstate->s.qinfo.qclass = qinfo->qclass; 526238106Sdes mstate->s.qinfo.qname_len = qinfo->qname_len; 527238106Sdes mstate->s.qinfo.qname = regional_alloc_init(region, qinfo->qname, 528238106Sdes qinfo->qname_len); 529238106Sdes if(!mstate->s.qinfo.qname) { 530238106Sdes alloc_reg_release(env->alloc, region); 531238106Sdes return NULL; 532238106Sdes } 533238106Sdes /* remove all weird bits from qflags */ 534238106Sdes mstate->s.query_flags = (qflags & (BIT_RD|BIT_CD)); 535238106Sdes mstate->s.is_priming = prime; 536238106Sdes mstate->s.reply = NULL; 537238106Sdes mstate->s.region = region; 538238106Sdes mstate->s.curmod = 0; 539238106Sdes mstate->s.return_msg = 0; 540238106Sdes mstate->s.return_rcode = LDNS_RCODE_NOERROR; 541238106Sdes mstate->s.env = env; 542238106Sdes mstate->s.mesh_info = mstate; 543238106Sdes mstate->s.prefetch_leeway = 0; 544238106Sdes /* init modules */ 545238106Sdes for(i=0; i<env->mesh->mods.num; i++) { 546238106Sdes mstate->s.minfo[i] = NULL; 547238106Sdes mstate->s.ext_state[i] = module_state_initial; 548238106Sdes } 549238106Sdes return mstate; 550238106Sdes} 551238106Sdes 552238106Sdesvoid 553238106Sdesmesh_state_cleanup(struct mesh_state* mstate) 554238106Sdes{ 555238106Sdes struct mesh_area* mesh; 556238106Sdes int i; 557238106Sdes if(!mstate) 558238106Sdes return; 559238106Sdes mesh = mstate->s.env->mesh; 560238106Sdes /* drop unsent replies */ 561238106Sdes if(!mstate->replies_sent) { 562238106Sdes struct mesh_reply* rep; 563238106Sdes struct mesh_cb* cb; 564238106Sdes for(rep=mstate->reply_list; rep; rep=rep->next) { 565238106Sdes comm_point_drop_reply(&rep->query_reply); 566238106Sdes mesh->num_reply_addrs--; 567238106Sdes } 568238106Sdes for(cb=mstate->cb_list; cb; cb=cb->next) { 569238106Sdes fptr_ok(fptr_whitelist_mesh_cb(cb->cb)); 570238106Sdes (*cb->cb)(cb->cb_arg, LDNS_RCODE_SERVFAIL, NULL, 571238106Sdes sec_status_unchecked, NULL); 572238106Sdes mesh->num_reply_addrs--; 573238106Sdes } 574238106Sdes } 575238106Sdes 576238106Sdes /* de-init modules */ 577238106Sdes for(i=0; i<mesh->mods.num; i++) { 578238106Sdes fptr_ok(fptr_whitelist_mod_clear(mesh->mods.mod[i]->clear)); 579238106Sdes (*mesh->mods.mod[i]->clear)(&mstate->s, i); 580238106Sdes mstate->s.minfo[i] = NULL; 581238106Sdes mstate->s.ext_state[i] = module_finished; 582238106Sdes } 583238106Sdes alloc_reg_release(mstate->s.env->alloc, mstate->s.region); 584238106Sdes} 585238106Sdes 586238106Sdesvoid 587238106Sdesmesh_state_delete(struct module_qstate* qstate) 588238106Sdes{ 589238106Sdes struct mesh_area* mesh; 590238106Sdes struct mesh_state_ref* super, ref; 591238106Sdes struct mesh_state* mstate; 592238106Sdes if(!qstate) 593238106Sdes return; 594238106Sdes mstate = qstate->mesh_info; 595238106Sdes mesh = mstate->s.env->mesh; 596238106Sdes mesh_detach_subs(&mstate->s); 597238106Sdes if(mstate->list_select == mesh_forever_list) { 598238106Sdes mesh->num_forever_states --; 599238106Sdes mesh_list_remove(mstate, &mesh->forever_first, 600238106Sdes &mesh->forever_last); 601238106Sdes } else if(mstate->list_select == mesh_jostle_list) { 602238106Sdes mesh_list_remove(mstate, &mesh->jostle_first, 603238106Sdes &mesh->jostle_last); 604238106Sdes } 605238106Sdes if(!mstate->reply_list && !mstate->cb_list 606238106Sdes && mstate->super_set.count == 0) { 607238106Sdes log_assert(mesh->num_detached_states > 0); 608238106Sdes mesh->num_detached_states--; 609238106Sdes } 610238106Sdes if(mstate->reply_list || mstate->cb_list) { 611238106Sdes log_assert(mesh->num_reply_states > 0); 612238106Sdes mesh->num_reply_states--; 613238106Sdes } 614238106Sdes ref.node.key = &ref; 615238106Sdes ref.s = mstate; 616238106Sdes RBTREE_FOR(super, struct mesh_state_ref*, &mstate->super_set) { 617238106Sdes (void)rbtree_delete(&super->s->sub_set, &ref); 618238106Sdes } 619238106Sdes (void)rbtree_delete(&mesh->run, mstate); 620238106Sdes (void)rbtree_delete(&mesh->all, mstate); 621238106Sdes mesh_state_cleanup(mstate); 622238106Sdes} 623238106Sdes 624238106Sdes/** helper recursive rbtree find routine */ 625238106Sdesstatic int 626238106Sdesfind_in_subsub(struct mesh_state* m, struct mesh_state* tofind, size_t *c) 627238106Sdes{ 628238106Sdes struct mesh_state_ref* r; 629238106Sdes if((*c)++ > MESH_MAX_SUBSUB) 630238106Sdes return 1; 631238106Sdes RBTREE_FOR(r, struct mesh_state_ref*, &m->sub_set) { 632238106Sdes if(r->s == tofind || find_in_subsub(r->s, tofind, c)) 633238106Sdes return 1; 634238106Sdes } 635238106Sdes return 0; 636238106Sdes} 637238106Sdes 638238106Sdes/** find cycle for already looked up mesh_state */ 639238106Sdesstatic int 640238106Sdesmesh_detect_cycle_found(struct module_qstate* qstate, struct mesh_state* dep_m) 641238106Sdes{ 642238106Sdes struct mesh_state* cyc_m = qstate->mesh_info; 643238106Sdes size_t counter = 0; 644238106Sdes if(!dep_m) 645238106Sdes return 0; 646238106Sdes if(dep_m == cyc_m || find_in_subsub(dep_m, cyc_m, &counter)) { 647238106Sdes if(counter > MESH_MAX_SUBSUB) 648238106Sdes return 2; 649238106Sdes return 1; 650238106Sdes } 651238106Sdes return 0; 652238106Sdes} 653238106Sdes 654238106Sdesvoid mesh_detach_subs(struct module_qstate* qstate) 655238106Sdes{ 656238106Sdes struct mesh_area* mesh = qstate->env->mesh; 657238106Sdes struct mesh_state_ref* ref, lookup; 658238106Sdes#ifdef UNBOUND_DEBUG 659238106Sdes struct rbnode_t* n; 660238106Sdes#endif 661238106Sdes lookup.node.key = &lookup; 662238106Sdes lookup.s = qstate->mesh_info; 663238106Sdes RBTREE_FOR(ref, struct mesh_state_ref*, &qstate->mesh_info->sub_set) { 664238106Sdes#ifdef UNBOUND_DEBUG 665238106Sdes n = 666269257Sdes#else 667269257Sdes (void) 668238106Sdes#endif 669238106Sdes rbtree_delete(&ref->s->super_set, &lookup); 670238106Sdes log_assert(n != NULL); /* must have been present */ 671238106Sdes if(!ref->s->reply_list && !ref->s->cb_list 672238106Sdes && ref->s->super_set.count == 0) { 673238106Sdes mesh->num_detached_states++; 674238106Sdes log_assert(mesh->num_detached_states + 675238106Sdes mesh->num_reply_states <= mesh->all.count); 676238106Sdes } 677238106Sdes } 678238106Sdes rbtree_init(&qstate->mesh_info->sub_set, &mesh_state_ref_compare); 679238106Sdes} 680238106Sdes 681238106Sdesint mesh_attach_sub(struct module_qstate* qstate, struct query_info* qinfo, 682238106Sdes uint16_t qflags, int prime, struct module_qstate** newq) 683238106Sdes{ 684238106Sdes /* find it, if not, create it */ 685238106Sdes struct mesh_area* mesh = qstate->env->mesh; 686238106Sdes struct mesh_state* sub = mesh_area_find(mesh, qinfo, qflags, prime); 687249141Sdes int was_detached; 688238106Sdes if(mesh_detect_cycle_found(qstate, sub)) { 689238106Sdes verbose(VERB_ALGO, "attach failed, cycle detected"); 690238106Sdes return 0; 691238106Sdes } 692238106Sdes if(!sub) { 693238106Sdes#ifdef UNBOUND_DEBUG 694238106Sdes struct rbnode_t* n; 695238106Sdes#endif 696238106Sdes /* create a new one */ 697238106Sdes sub = mesh_state_create(qstate->env, qinfo, qflags, prime); 698238106Sdes if(!sub) { 699238106Sdes log_err("mesh_attach_sub: out of memory"); 700238106Sdes return 0; 701238106Sdes } 702238106Sdes#ifdef UNBOUND_DEBUG 703238106Sdes n = 704269257Sdes#else 705269257Sdes (void) 706238106Sdes#endif 707238106Sdes rbtree_insert(&mesh->all, &sub->node); 708238106Sdes log_assert(n != NULL); 709238106Sdes /* set detached (it is now) */ 710238106Sdes mesh->num_detached_states++; 711238106Sdes /* set new query state to run */ 712238106Sdes#ifdef UNBOUND_DEBUG 713238106Sdes n = 714269257Sdes#else 715269257Sdes (void) 716238106Sdes#endif 717238106Sdes rbtree_insert(&mesh->run, &sub->run_node); 718238106Sdes log_assert(n != NULL); 719238106Sdes *newq = &sub->s; 720238106Sdes } else 721238106Sdes *newq = NULL; 722249141Sdes was_detached = (sub->super_set.count == 0); 723238106Sdes if(!mesh_state_attachment(qstate->mesh_info, sub)) 724238106Sdes return 0; 725249141Sdes /* if it was a duplicate attachment, the count was not zero before */ 726249141Sdes if(!sub->reply_list && !sub->cb_list && was_detached && 727249141Sdes sub->super_set.count == 1) { 728238106Sdes /* it used to be detached, before this one got added */ 729238106Sdes log_assert(mesh->num_detached_states > 0); 730238106Sdes mesh->num_detached_states--; 731238106Sdes } 732238106Sdes /* *newq will be run when inited after the current module stops */ 733238106Sdes return 1; 734238106Sdes} 735238106Sdes 736238106Sdesint mesh_state_attachment(struct mesh_state* super, struct mesh_state* sub) 737238106Sdes{ 738238106Sdes#ifdef UNBOUND_DEBUG 739238106Sdes struct rbnode_t* n; 740238106Sdes#endif 741238106Sdes struct mesh_state_ref* subref; /* points to sub, inserted in super */ 742238106Sdes struct mesh_state_ref* superref; /* points to super, inserted in sub */ 743238106Sdes if( !(subref = regional_alloc(super->s.region, 744238106Sdes sizeof(struct mesh_state_ref))) || 745238106Sdes !(superref = regional_alloc(sub->s.region, 746238106Sdes sizeof(struct mesh_state_ref))) ) { 747238106Sdes log_err("mesh_state_attachment: out of memory"); 748238106Sdes return 0; 749238106Sdes } 750238106Sdes superref->node.key = superref; 751238106Sdes superref->s = super; 752238106Sdes subref->node.key = subref; 753238106Sdes subref->s = sub; 754249141Sdes if(!rbtree_insert(&sub->super_set, &superref->node)) { 755249141Sdes /* this should not happen, iterator and validator do not 756249141Sdes * attach subqueries that are identical. */ 757249141Sdes /* already attached, we are done, nothing todo. 758249141Sdes * since superref and subref already allocated in region, 759249141Sdes * we cannot free them */ 760249141Sdes return 1; 761249141Sdes } 762238106Sdes#ifdef UNBOUND_DEBUG 763238106Sdes n = 764269257Sdes#else 765269257Sdes (void) 766238106Sdes#endif 767238106Sdes rbtree_insert(&super->sub_set, &subref->node); 768249141Sdes log_assert(n != NULL); /* we checked above if statement, the reverse 769249141Sdes administration should not fail now, unless they are out of sync */ 770238106Sdes return 1; 771238106Sdes} 772238106Sdes 773238106Sdes/** 774238106Sdes * callback results to mesh cb entry 775238106Sdes * @param m: mesh state to send it for. 776238106Sdes * @param rcode: if not 0, error code. 777238106Sdes * @param rep: reply to send (or NULL if rcode is set). 778238106Sdes * @param r: callback entry 779238106Sdes */ 780238106Sdesstatic void 781238106Sdesmesh_do_callback(struct mesh_state* m, int rcode, struct reply_info* rep, 782238106Sdes struct mesh_cb* r) 783238106Sdes{ 784238106Sdes int secure; 785238106Sdes char* reason = NULL; 786238106Sdes /* bogus messages are not made into servfail, sec_status passed 787238106Sdes * to the callback function */ 788238106Sdes if(rep && rep->security == sec_status_secure) 789238106Sdes secure = 1; 790238106Sdes else secure = 0; 791238106Sdes if(!rep && rcode == LDNS_RCODE_NOERROR) 792238106Sdes rcode = LDNS_RCODE_SERVFAIL; 793238106Sdes if(!rcode && rep->security == sec_status_bogus) { 794238106Sdes if(!(reason = errinf_to_str(&m->s))) 795238106Sdes rcode = LDNS_RCODE_SERVFAIL; 796238106Sdes } 797238106Sdes /* send the reply */ 798238106Sdes if(rcode) { 799238106Sdes fptr_ok(fptr_whitelist_mesh_cb(r->cb)); 800238106Sdes (*r->cb)(r->cb_arg, rcode, r->buf, sec_status_unchecked, NULL); 801238106Sdes } else { 802238106Sdes size_t udp_size = r->edns.udp_size; 803269257Sdes sldns_buffer_clear(r->buf); 804238106Sdes r->edns.edns_version = EDNS_ADVERTISED_VERSION; 805238106Sdes r->edns.udp_size = EDNS_ADVERTISED_SIZE; 806238106Sdes r->edns.ext_rcode = 0; 807238106Sdes r->edns.bits &= EDNS_DO; 808238106Sdes if(!reply_info_answer_encode(&m->s.qinfo, rep, r->qid, 809238106Sdes r->qflags, r->buf, 0, 1, 810238106Sdes m->s.env->scratch, udp_size, &r->edns, 811238106Sdes (int)(r->edns.bits & EDNS_DO), secure)) 812238106Sdes { 813238106Sdes fptr_ok(fptr_whitelist_mesh_cb(r->cb)); 814238106Sdes (*r->cb)(r->cb_arg, LDNS_RCODE_SERVFAIL, r->buf, 815238106Sdes sec_status_unchecked, NULL); 816238106Sdes } else { 817238106Sdes fptr_ok(fptr_whitelist_mesh_cb(r->cb)); 818238106Sdes (*r->cb)(r->cb_arg, LDNS_RCODE_NOERROR, r->buf, 819238106Sdes rep->security, reason); 820238106Sdes } 821238106Sdes } 822238106Sdes free(reason); 823238106Sdes m->s.env->mesh->num_reply_addrs--; 824238106Sdes} 825238106Sdes 826238106Sdes/** 827238106Sdes * Send reply to mesh reply entry 828238106Sdes * @param m: mesh state to send it for. 829238106Sdes * @param rcode: if not 0, error code. 830238106Sdes * @param rep: reply to send (or NULL if rcode is set). 831238106Sdes * @param r: reply entry 832238106Sdes * @param prev: previous reply, already has its answer encoded in buffer. 833238106Sdes */ 834238106Sdesstatic void 835238106Sdesmesh_send_reply(struct mesh_state* m, int rcode, struct reply_info* rep, 836238106Sdes struct mesh_reply* r, struct mesh_reply* prev) 837238106Sdes{ 838238106Sdes struct timeval end_time; 839238106Sdes struct timeval duration; 840238106Sdes int secure; 841238106Sdes /* examine security status */ 842238106Sdes if(m->s.env->need_to_validate && (!(r->qflags&BIT_CD) || 843238106Sdes m->s.env->cfg->ignore_cd) && rep && 844238106Sdes rep->security <= sec_status_bogus) { 845238106Sdes rcode = LDNS_RCODE_SERVFAIL; 846238106Sdes if(m->s.env->cfg->stat_extended) 847238106Sdes m->s.env->mesh->ans_bogus++; 848238106Sdes } 849238106Sdes if(rep && rep->security == sec_status_secure) 850238106Sdes secure = 1; 851238106Sdes else secure = 0; 852238106Sdes if(!rep && rcode == LDNS_RCODE_NOERROR) 853238106Sdes rcode = LDNS_RCODE_SERVFAIL; 854238106Sdes /* send the reply */ 855238106Sdes if(prev && prev->qflags == r->qflags && 856238106Sdes prev->edns.edns_present == r->edns.edns_present && 857238106Sdes prev->edns.bits == r->edns.bits && 858238106Sdes prev->edns.udp_size == r->edns.udp_size) { 859238106Sdes /* if the previous reply is identical to this one, fix ID */ 860238106Sdes if(prev->query_reply.c->buffer != r->query_reply.c->buffer) 861269257Sdes sldns_buffer_copy(r->query_reply.c->buffer, 862238106Sdes prev->query_reply.c->buffer); 863269257Sdes sldns_buffer_write_at(r->query_reply.c->buffer, 0, 864238106Sdes &r->qid, sizeof(uint16_t)); 865269257Sdes sldns_buffer_write_at(r->query_reply.c->buffer, 12, 866238106Sdes r->qname, m->s.qinfo.qname_len); 867238106Sdes comm_point_send_reply(&r->query_reply); 868238106Sdes } else if(rcode) { 869238106Sdes m->s.qinfo.qname = r->qname; 870238106Sdes error_encode(r->query_reply.c->buffer, rcode, &m->s.qinfo, 871238106Sdes r->qid, r->qflags, &r->edns); 872238106Sdes comm_point_send_reply(&r->query_reply); 873238106Sdes } else { 874238106Sdes size_t udp_size = r->edns.udp_size; 875238106Sdes r->edns.edns_version = EDNS_ADVERTISED_VERSION; 876238106Sdes r->edns.udp_size = EDNS_ADVERTISED_SIZE; 877238106Sdes r->edns.ext_rcode = 0; 878238106Sdes r->edns.bits &= EDNS_DO; 879238106Sdes m->s.qinfo.qname = r->qname; 880238106Sdes if(!reply_info_answer_encode(&m->s.qinfo, rep, r->qid, 881238106Sdes r->qflags, r->query_reply.c->buffer, 0, 1, 882238106Sdes m->s.env->scratch, udp_size, &r->edns, 883238106Sdes (int)(r->edns.bits & EDNS_DO), secure)) 884238106Sdes { 885238106Sdes error_encode(r->query_reply.c->buffer, 886238106Sdes LDNS_RCODE_SERVFAIL, &m->s.qinfo, r->qid, 887238106Sdes r->qflags, &r->edns); 888238106Sdes } 889238106Sdes comm_point_send_reply(&r->query_reply); 890238106Sdes } 891238106Sdes /* account */ 892238106Sdes m->s.env->mesh->num_reply_addrs--; 893238106Sdes end_time = *m->s.env->now_tv; 894238106Sdes timeval_subtract(&duration, &end_time, &r->start_time); 895269257Sdes verbose(VERB_ALGO, "query took " ARG_LL "d.%6.6d sec", 896269257Sdes (long long)duration.tv_sec, (int)duration.tv_usec); 897238106Sdes m->s.env->mesh->replies_sent++; 898238106Sdes timeval_add(&m->s.env->mesh->replies_sum_wait, &duration); 899238106Sdes timehist_insert(m->s.env->mesh->histogram, &duration); 900238106Sdes if(m->s.env->cfg->stat_extended) { 901269257Sdes uint16_t rc = FLAGS_GET_RCODE(sldns_buffer_read_u16_at(r-> 902238106Sdes query_reply.c->buffer, 2)); 903238106Sdes if(secure) m->s.env->mesh->ans_secure++; 904238106Sdes m->s.env->mesh->ans_rcode[ rc ] ++; 905269257Sdes if(rc == 0 && LDNS_ANCOUNT(sldns_buffer_begin(r-> 906238106Sdes query_reply.c->buffer)) == 0) 907238106Sdes m->s.env->mesh->ans_nodata++; 908238106Sdes } 909238106Sdes} 910238106Sdes 911238106Sdesvoid mesh_query_done(struct mesh_state* mstate) 912238106Sdes{ 913238106Sdes struct mesh_reply* r; 914238106Sdes struct mesh_reply* prev = NULL; 915238106Sdes struct mesh_cb* c; 916238106Sdes struct reply_info* rep = (mstate->s.return_msg? 917238106Sdes mstate->s.return_msg->rep:NULL); 918238106Sdes for(r = mstate->reply_list; r; r = r->next) { 919238106Sdes mesh_send_reply(mstate, mstate->s.return_rcode, rep, r, prev); 920238106Sdes prev = r; 921238106Sdes } 922238106Sdes mstate->replies_sent = 1; 923238106Sdes for(c = mstate->cb_list; c; c = c->next) { 924238106Sdes mesh_do_callback(mstate, mstate->s.return_rcode, rep, c); 925238106Sdes } 926238106Sdes} 927238106Sdes 928238106Sdesvoid mesh_walk_supers(struct mesh_area* mesh, struct mesh_state* mstate) 929238106Sdes{ 930238106Sdes struct mesh_state_ref* ref; 931238106Sdes RBTREE_FOR(ref, struct mesh_state_ref*, &mstate->super_set) 932238106Sdes { 933238106Sdes /* make super runnable */ 934238106Sdes (void)rbtree_insert(&mesh->run, &ref->s->run_node); 935238106Sdes /* callback the function to inform super of result */ 936238106Sdes fptr_ok(fptr_whitelist_mod_inform_super( 937238106Sdes mesh->mods.mod[ref->s->s.curmod]->inform_super)); 938238106Sdes (*mesh->mods.mod[ref->s->s.curmod]->inform_super)(&mstate->s, 939238106Sdes ref->s->s.curmod, &ref->s->s); 940238106Sdes } 941238106Sdes} 942238106Sdes 943238106Sdesstruct mesh_state* mesh_area_find(struct mesh_area* mesh, 944238106Sdes struct query_info* qinfo, uint16_t qflags, int prime) 945238106Sdes{ 946238106Sdes struct mesh_state key; 947238106Sdes struct mesh_state* result; 948238106Sdes 949238106Sdes key.node.key = &key; 950238106Sdes key.s.is_priming = prime; 951238106Sdes key.s.qinfo = *qinfo; 952238106Sdes key.s.query_flags = qflags; 953238106Sdes 954238106Sdes result = (struct mesh_state*)rbtree_search(&mesh->all, &key); 955238106Sdes return result; 956238106Sdes} 957238106Sdes 958238106Sdesint mesh_state_add_cb(struct mesh_state* s, struct edns_data* edns, 959269257Sdes sldns_buffer* buf, mesh_cb_func_t cb, void* cb_arg, 960238106Sdes uint16_t qid, uint16_t qflags) 961238106Sdes{ 962238106Sdes struct mesh_cb* r = regional_alloc(s->s.region, 963238106Sdes sizeof(struct mesh_cb)); 964238106Sdes if(!r) 965238106Sdes return 0; 966238106Sdes r->buf = buf; 967238106Sdes log_assert(fptr_whitelist_mesh_cb(cb)); /* early failure ifmissing*/ 968238106Sdes r->cb = cb; 969238106Sdes r->cb_arg = cb_arg; 970238106Sdes r->edns = *edns; 971238106Sdes r->qid = qid; 972238106Sdes r->qflags = qflags; 973238106Sdes r->next = s->cb_list; 974238106Sdes s->cb_list = r; 975238106Sdes return 1; 976238106Sdes 977238106Sdes} 978238106Sdes 979238106Sdesint mesh_state_add_reply(struct mesh_state* s, struct edns_data* edns, 980238106Sdes struct comm_reply* rep, uint16_t qid, uint16_t qflags, uint8_t* qname) 981238106Sdes{ 982238106Sdes struct mesh_reply* r = regional_alloc(s->s.region, 983238106Sdes sizeof(struct mesh_reply)); 984238106Sdes if(!r) 985238106Sdes return 0; 986238106Sdes r->query_reply = *rep; 987238106Sdes r->edns = *edns; 988238106Sdes r->qid = qid; 989238106Sdes r->qflags = qflags; 990238106Sdes r->start_time = *s->s.env->now_tv; 991238106Sdes r->next = s->reply_list; 992238106Sdes r->qname = regional_alloc_init(s->s.region, qname, 993238106Sdes s->s.qinfo.qname_len); 994238106Sdes if(!r->qname) 995238106Sdes return 0; 996238106Sdes s->reply_list = r; 997238106Sdes return 1; 998238106Sdes 999238106Sdes} 1000238106Sdes 1001238106Sdes/** 1002238106Sdes * Continue processing the mesh state at another module. 1003238106Sdes * Handles module to modules tranfer of control. 1004238106Sdes * Handles module finished. 1005238106Sdes * @param mesh: the mesh area. 1006238106Sdes * @param mstate: currently active mesh state. 1007238106Sdes * Deleted if finished, calls _done and _supers to 1008238106Sdes * send replies to clients and inform other mesh states. 1009238106Sdes * This in turn may create additional runnable mesh states. 1010238106Sdes * @param s: state at which the current module exited. 1011238106Sdes * @param ev: the event sent to the module. 1012238106Sdes * returned is the event to send to the next module. 1013238106Sdes * @return true if continue processing at the new module. 1014238106Sdes * false if not continued processing is needed. 1015238106Sdes */ 1016238106Sdesstatic int 1017238106Sdesmesh_continue(struct mesh_area* mesh, struct mesh_state* mstate, 1018238106Sdes enum module_ext_state s, enum module_ev* ev) 1019238106Sdes{ 1020238106Sdes mstate->num_activated++; 1021238106Sdes if(mstate->num_activated > MESH_MAX_ACTIVATION) { 1022238106Sdes /* module is looping. Stop it. */ 1023238106Sdes log_err("internal error: looping module stopped"); 1024238106Sdes log_query_info(VERB_QUERY, "pass error for qstate", 1025238106Sdes &mstate->s.qinfo); 1026238106Sdes s = module_error; 1027238106Sdes } 1028238106Sdes if(s == module_wait_module || s == module_restart_next) { 1029238106Sdes /* start next module */ 1030238106Sdes mstate->s.curmod++; 1031238106Sdes if(mesh->mods.num == mstate->s.curmod) { 1032238106Sdes log_err("Cannot pass to next module; at last module"); 1033238106Sdes log_query_info(VERB_QUERY, "pass error for qstate", 1034238106Sdes &mstate->s.qinfo); 1035238106Sdes mstate->s.curmod--; 1036238106Sdes return mesh_continue(mesh, mstate, module_error, ev); 1037238106Sdes } 1038238106Sdes if(s == module_restart_next) { 1039238106Sdes fptr_ok(fptr_whitelist_mod_clear( 1040238106Sdes mesh->mods.mod[mstate->s.curmod]->clear)); 1041238106Sdes (*mesh->mods.mod[mstate->s.curmod]->clear) 1042238106Sdes (&mstate->s, mstate->s.curmod); 1043238106Sdes mstate->s.minfo[mstate->s.curmod] = NULL; 1044238106Sdes } 1045238106Sdes *ev = module_event_pass; 1046238106Sdes return 1; 1047238106Sdes } 1048238106Sdes if(s == module_error && mstate->s.return_rcode == LDNS_RCODE_NOERROR) { 1049238106Sdes /* error is bad, handle pass back up below */ 1050238106Sdes mstate->s.return_rcode = LDNS_RCODE_SERVFAIL; 1051238106Sdes } 1052238106Sdes if(s == module_error || s == module_finished) { 1053238106Sdes if(mstate->s.curmod == 0) { 1054238106Sdes mesh_query_done(mstate); 1055238106Sdes mesh_walk_supers(mesh, mstate); 1056238106Sdes mesh_state_delete(&mstate->s); 1057238106Sdes return 0; 1058238106Sdes } 1059238106Sdes /* pass along the locus of control */ 1060238106Sdes mstate->s.curmod --; 1061238106Sdes *ev = module_event_moddone; 1062238106Sdes return 1; 1063238106Sdes } 1064238106Sdes return 0; 1065238106Sdes} 1066238106Sdes 1067238106Sdesvoid mesh_run(struct mesh_area* mesh, struct mesh_state* mstate, 1068238106Sdes enum module_ev ev, struct outbound_entry* e) 1069238106Sdes{ 1070238106Sdes enum module_ext_state s; 1071238106Sdes verbose(VERB_ALGO, "mesh_run: start"); 1072238106Sdes while(mstate) { 1073238106Sdes /* run the module */ 1074238106Sdes fptr_ok(fptr_whitelist_mod_operate( 1075238106Sdes mesh->mods.mod[mstate->s.curmod]->operate)); 1076238106Sdes (*mesh->mods.mod[mstate->s.curmod]->operate) 1077238106Sdes (&mstate->s, ev, mstate->s.curmod, e); 1078238106Sdes 1079238106Sdes /* examine results */ 1080238106Sdes mstate->s.reply = NULL; 1081238106Sdes regional_free_all(mstate->s.env->scratch); 1082238106Sdes s = mstate->s.ext_state[mstate->s.curmod]; 1083238106Sdes verbose(VERB_ALGO, "mesh_run: %s module exit state is %s", 1084238106Sdes mesh->mods.mod[mstate->s.curmod]->name, strextstate(s)); 1085238106Sdes e = NULL; 1086238106Sdes if(mesh_continue(mesh, mstate, s, &ev)) 1087238106Sdes continue; 1088238106Sdes 1089238106Sdes /* run more modules */ 1090238106Sdes ev = module_event_pass; 1091238106Sdes if(mesh->run.count > 0) { 1092238106Sdes /* pop random element off the runnable tree */ 1093238106Sdes mstate = (struct mesh_state*)mesh->run.root->key; 1094238106Sdes (void)rbtree_delete(&mesh->run, mstate); 1095238106Sdes } else mstate = NULL; 1096238106Sdes } 1097238106Sdes if(verbosity >= VERB_ALGO) { 1098238106Sdes mesh_stats(mesh, "mesh_run: end"); 1099238106Sdes mesh_log_list(mesh); 1100238106Sdes } 1101238106Sdes} 1102238106Sdes 1103238106Sdesvoid 1104238106Sdesmesh_log_list(struct mesh_area* mesh) 1105238106Sdes{ 1106238106Sdes char buf[30]; 1107238106Sdes struct mesh_state* m; 1108238106Sdes int num = 0; 1109238106Sdes RBTREE_FOR(m, struct mesh_state*, &mesh->all) { 1110238106Sdes snprintf(buf, sizeof(buf), "%d%s%s%s%s%s mod%d %s%s", 1111238106Sdes num++, (m->s.is_priming)?"p":"", /* prime */ 1112238106Sdes (m->s.query_flags&BIT_RD)?"RD":"", 1113238106Sdes (m->s.query_flags&BIT_CD)?"CD":"", 1114238106Sdes (m->super_set.count==0)?"d":"", /* detached */ 1115238106Sdes (m->sub_set.count!=0)?"c":"", /* children */ 1116238106Sdes m->s.curmod, (m->reply_list)?"rep":"", /*hasreply*/ 1117238106Sdes (m->cb_list)?"cb":"" /* callbacks */ 1118238106Sdes ); 1119238106Sdes log_query_info(VERB_ALGO, buf, &m->s.qinfo); 1120238106Sdes } 1121238106Sdes} 1122238106Sdes 1123238106Sdesvoid 1124238106Sdesmesh_stats(struct mesh_area* mesh, const char* str) 1125238106Sdes{ 1126238106Sdes verbose(VERB_DETAIL, "%s %u recursion states (%u with reply, " 1127238106Sdes "%u detached), %u waiting replies, %u recursion replies " 1128238106Sdes "sent, %d replies dropped, %d states jostled out", 1129238106Sdes str, (unsigned)mesh->all.count, 1130238106Sdes (unsigned)mesh->num_reply_states, 1131238106Sdes (unsigned)mesh->num_detached_states, 1132238106Sdes (unsigned)mesh->num_reply_addrs, 1133238106Sdes (unsigned)mesh->replies_sent, 1134238106Sdes (unsigned)mesh->stats_dropped, 1135238106Sdes (unsigned)mesh->stats_jostled); 1136238106Sdes if(mesh->replies_sent > 0) { 1137238106Sdes struct timeval avg; 1138238106Sdes timeval_divide(&avg, &mesh->replies_sum_wait, 1139238106Sdes mesh->replies_sent); 1140238106Sdes log_info("average recursion processing time " 1141269257Sdes ARG_LL "d.%6.6d sec", 1142269257Sdes (long long)avg.tv_sec, (int)avg.tv_usec); 1143238106Sdes log_info("histogram of recursion processing times"); 1144238106Sdes timehist_log(mesh->histogram, "recursions"); 1145238106Sdes } 1146238106Sdes} 1147238106Sdes 1148238106Sdesvoid 1149238106Sdesmesh_stats_clear(struct mesh_area* mesh) 1150238106Sdes{ 1151238106Sdes if(!mesh) 1152238106Sdes return; 1153238106Sdes mesh->replies_sent = 0; 1154238106Sdes mesh->replies_sum_wait.tv_sec = 0; 1155238106Sdes mesh->replies_sum_wait.tv_usec = 0; 1156238106Sdes mesh->stats_jostled = 0; 1157238106Sdes mesh->stats_dropped = 0; 1158238106Sdes timehist_clear(mesh->histogram); 1159238106Sdes mesh->ans_secure = 0; 1160238106Sdes mesh->ans_bogus = 0; 1161238106Sdes memset(&mesh->ans_rcode[0], 0, sizeof(size_t)*16); 1162238106Sdes mesh->ans_nodata = 0; 1163238106Sdes} 1164238106Sdes 1165238106Sdessize_t 1166238106Sdesmesh_get_mem(struct mesh_area* mesh) 1167238106Sdes{ 1168238106Sdes struct mesh_state* m; 1169238106Sdes size_t s = sizeof(*mesh) + sizeof(struct timehist) + 1170238106Sdes sizeof(struct th_buck)*mesh->histogram->num + 1171269257Sdes sizeof(sldns_buffer) + sldns_buffer_capacity(mesh->qbuf_bak); 1172238106Sdes RBTREE_FOR(m, struct mesh_state*, &mesh->all) { 1173238106Sdes /* all, including m itself allocated in qstate region */ 1174238106Sdes s += regional_get_mem(m->s.region); 1175238106Sdes } 1176238106Sdes return s; 1177238106Sdes} 1178238106Sdes 1179238106Sdesint 1180238106Sdesmesh_detect_cycle(struct module_qstate* qstate, struct query_info* qinfo, 1181238106Sdes uint16_t flags, int prime) 1182238106Sdes{ 1183238106Sdes struct mesh_area* mesh = qstate->env->mesh; 1184238106Sdes struct mesh_state* dep_m = mesh_area_find(mesh, qinfo, flags, prime); 1185238106Sdes return mesh_detect_cycle_found(qstate, dep_m); 1186238106Sdes} 1187238106Sdes 1188238106Sdesvoid mesh_list_insert(struct mesh_state* m, struct mesh_state** fp, 1189238106Sdes struct mesh_state** lp) 1190238106Sdes{ 1191238106Sdes /* insert as last element */ 1192238106Sdes m->prev = *lp; 1193238106Sdes m->next = NULL; 1194238106Sdes if(*lp) 1195238106Sdes (*lp)->next = m; 1196238106Sdes else *fp = m; 1197238106Sdes *lp = m; 1198238106Sdes} 1199238106Sdes 1200238106Sdesvoid mesh_list_remove(struct mesh_state* m, struct mesh_state** fp, 1201238106Sdes struct mesh_state** lp) 1202238106Sdes{ 1203238106Sdes if(m->next) 1204238106Sdes m->next->prev = m->prev; 1205238106Sdes else *lp = m->prev; 1206238106Sdes if(m->prev) 1207238106Sdes m->prev->next = m->next; 1208238106Sdes else *fp = m->next; 1209238106Sdes} 1210