contrib/unbound/services/mesh.c

   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 "util/edns.h"
  59 #include "sldns/sbuffer.h"
  60 #include "sldns/wire2str.h"
  61 #include "services/localzone.h"
  62 #include "util/data/dname.h"
  63 #include "respip/respip.h"
  64 #include "services/listen_dnsport.h"
  65
  66 /** subtract timers and the values do not overflow or become negative */
  67 static void
  68 timeval_subtract(struct timeval* d, const struct timeval* end, const struct timeval* start)
  69 {
  70 #ifndef S_SPLINT_S
  71         time_t end_usec = end->tv_usec;
  72         d->tv_sec = end->tv_sec - start->tv_sec;
  73         if(end_usec < start->tv_usec) {
  74                 end_usec += 1000000;
  75                 d->tv_sec--;
  76         }
  77         d->tv_usec = end_usec - start->tv_usec;
  78 #endif
  79 }
  80
  81 /** add timers and the values do not overflow or become negative */
  82 static void
  83 timeval_add(struct timeval* d, const struct timeval* add)
  84 {
  85 #ifndef S_SPLINT_S
  86         d->tv_sec += add->tv_sec;
  87         d->tv_usec += add->tv_usec;
  88         if(d->tv_usec > 1000000 ) {
  89                 d->tv_usec -= 1000000;
  90                 d->tv_sec++;
  91         }
  92 #endif
  93 }
  94
  95 /** divide sum of timers to get average */
  96 static void
  97 timeval_divide(struct timeval* avg, const struct timeval* sum, size_t d)
  98 {
  99 #ifndef S_SPLINT_S
 100         size_t leftover;
 101         if(d == 0) {
 102                 avg->tv_sec = 0;
 103                 avg->tv_usec = 0;
 104                 return;
 105         }
 106         avg->tv_sec = sum->tv_sec / d;
 107         avg->tv_usec = sum->tv_usec / d;
 108         /* handle fraction from seconds divide */
 109         leftover = sum->tv_sec - avg->tv_sec*d;
 110         avg->tv_usec += (leftover*1000000)/d;
 111 #endif
 112 }
 113
 114 /** histogram compare of time values */
 115 static int
 116 timeval_smaller(const struct timeval* x, const struct timeval* y)
 117 {
 118 #ifndef S_SPLINT_S
 119         if(x->tv_sec < y->tv_sec)
 120                 return 1;
 121         else if(x->tv_sec == y->tv_sec) {
 122                 if(x->tv_usec <= y->tv_usec)
 123                         return 1;
 124                 else    return 0;
 125         }
 126         else    return 0;
 127 #endif
 128 }
 129
 130 /*
 131  * Compare two response-ip client info entries for the purpose of mesh state
 132  * compare.  It returns 0 if ci_a and ci_b are considered equal; otherwise
 133  * 1 or -1 (they mean 'ci_a is larger/smaller than ci_b', respectively, but
 134  * in practice it should be only used to mean they are different).
 135  * We cannot share the mesh state for two queries if different response-ip
 136  * actions can apply in the end, even if those queries are otherwise identical.
 137  * For this purpose we compare tag lists and tag action lists; they should be
 138  * identical to share the same state.
 139  * For tag data, we don't look into the data content, as it can be
 140  * expensive; unless tag data are not defined for both or they point to the
 141  * exact same data in memory (i.e., they come from the same ACL entry), we
 142  * consider these data different.
 143  * Likewise, if the client info is associated with views, we don't look into
 144  * the views.  They are considered different unless they are exactly the same
 145  * even if the views only differ in the names.
 146  */
 147 static int
 148 client_info_compare(const struct respip_client_info* ci_a,
 149         const struct respip_client_info* ci_b)
 150 {
 151         int cmp;
 152
 153         if(!ci_a && !ci_b)
 154                 return 0;
 155         if(ci_a && !ci_b)
 156                 return -1;
 157         if(!ci_a && ci_b)
 158                 return 1;
 159         if(ci_a->taglen != ci_b->taglen)
 160                 return (ci_a->taglen < ci_b->taglen) ? -1 : 1;
 161         cmp = memcmp(ci_a->taglist, ci_b->taglist, ci_a->taglen);
 162         if(cmp != 0)
 163                 return cmp;
 164         if(ci_a->tag_actions_size != ci_b->tag_actions_size)
 165                 return (ci_a->tag_actions_size < ci_b->tag_actions_size) ?
 166                         -1 : 1;
 167         cmp = memcmp(ci_a->tag_actions, ci_b->tag_actions,
 168                 ci_a->tag_actions_size);
 169         if(cmp != 0)
 170                 return cmp;
 171         if(ci_a->tag_datas != ci_b->tag_datas)
 172                 return ci_a->tag_datas < ci_b->tag_datas ? -1 : 1;
 173         if(ci_a->view != ci_b->view)
 174                 return ci_a->view < ci_b->view ? -1 : 1;
 175         /* For the unbound daemon these should be non-NULL and identical,
 176          * but we check that just in case. */
 177         if(ci_a->respip_set != ci_b->respip_set)
 178                 return ci_a->respip_set < ci_b->respip_set ? -1 : 1;
 179         return 0;
 180 }
 181
 182 int
 183 mesh_state_compare(const void* ap, const void* bp)
 184 {
 185         struct mesh_state* a = (struct mesh_state*)ap;
 186         struct mesh_state* b = (struct mesh_state*)bp;
 187         int cmp;
 188
 189         if(a->unique < b->unique)
 190                 return -1;
 191         if(a->unique > b->unique)
 192                 return 1;
 193
 194         if(a->s.is_priming && !b->s.is_priming)
 195                 return -1;
 196         if(!a->s.is_priming && b->s.is_priming)
 197                 return 1;
 198
 199         if(a->s.is_valrec && !b->s.is_valrec)
 200                 return -1;
 201         if(!a->s.is_valrec && b->s.is_valrec)
 202                 return 1;
 203
 204         if((a->s.query_flags&BIT_RD) && !(b->s.query_flags&BIT_RD))
 205                 return -1;
 206         if(!(a->s.query_flags&BIT_RD) && (b->s.query_flags&BIT_RD))
 207                 return 1;
 208
 209         if((a->s.query_flags&BIT_CD) && !(b->s.query_flags&BIT_CD))
 210                 return -1;
 211         if(!(a->s.query_flags&BIT_CD) && (b->s.query_flags&BIT_CD))
 212                 return 1;
 213
 214         cmp = query_info_compare(&a->s.qinfo, &b->s.qinfo);
 215         if(cmp != 0)
 216                 return cmp;
 217         return client_info_compare(a->s.client_info, b->s.client_info);
 218 }
 219
 220 int
 221 mesh_state_ref_compare(const void* ap, const void* bp)
 222 {
 223         struct mesh_state_ref* a = (struct mesh_state_ref*)ap;
 224         struct mesh_state_ref* b = (struct mesh_state_ref*)bp;
 225         return mesh_state_compare(a->s, b->s);
 226 }
 227
 228 struct mesh_area*
 229 mesh_create(struct module_stack* stack, struct module_env* env)
 230 {
 231         struct mesh_area* mesh = calloc(1, sizeof(struct mesh_area));
 232         if(!mesh) {
 233                 log_err("mesh area alloc: out of memory");
 234                 return NULL;
 235         }
 236         mesh->histogram = timehist_setup();
 237         mesh->qbuf_bak = sldns_buffer_new(env->cfg->msg_buffer_size);
 238         if(!mesh->histogram || !mesh->qbuf_bak) {
 239                 free(mesh);
 240                 log_err("mesh area alloc: out of memory");
 241                 return NULL;
 242         }
 243         mesh->mods = *stack;
 244         mesh->env = env;
 245         rbtree_init(&mesh->run, &mesh_state_compare);
 246         rbtree_init(&mesh->all, &mesh_state_compare);
 247         mesh->num_reply_addrs = 0;
 248         mesh->num_reply_states = 0;
 249         mesh->num_detached_states = 0;
 250         mesh->num_forever_states = 0;
 251         mesh->stats_jostled = 0;
 252         mesh->stats_dropped = 0;
 253         mesh->max_reply_states = env->cfg->num_queries_per_thread;
 254         mesh->max_forever_states = (mesh->max_reply_states+1)/2;
 255 #ifndef S_SPLINT_S
 256         mesh->jostle_max.tv_sec = (time_t)(env->cfg->jostle_time / 1000);
 257         mesh->jostle_max.tv_usec = (time_t)((env->cfg->jostle_time % 1000)
 258                 *1000);
 259 #endif
 260         return mesh;
 261 }
 262
 263 /** help mesh delete delete mesh states */
 264 static void
 265 mesh_delete_helper(rbnode_type* n)
 266 {
 267         struct mesh_state* mstate = (struct mesh_state*)n->key;
 268         /* perform a full delete, not only 'cleanup' routine,
 269          * because other callbacks expect a clean state in the mesh.
 270          * For 're-entrant' calls */
 271         mesh_state_delete(&mstate->s);
 272         /* but because these delete the items from the tree, postorder
 273          * traversal and rbtree rebalancing do not work together */
 274 }
 275
 276 void
 277 mesh_delete(struct mesh_area* mesh)
 278 {
 279         if(!mesh)
 280                 return;
 281         /* free all query states */
 282         while(mesh->all.count)
 283                 mesh_delete_helper(mesh->all.root);
 284         timehist_delete(mesh->histogram);
 285         sldns_buffer_free(mesh->qbuf_bak);
 286         free(mesh);
 287 }
 288
 289 void
 290 mesh_delete_all(struct mesh_area* mesh)
 291 {
 292         /* free all query states */
 293         while(mesh->all.count)
 294                 mesh_delete_helper(mesh->all.root);
 295         mesh->stats_dropped += mesh->num_reply_addrs;
 296         /* clear mesh area references */
 297         rbtree_init(&mesh->run, &mesh_state_compare);
 298         rbtree_init(&mesh->all, &mesh_state_compare);
 299         mesh->num_reply_addrs = 0;
 300         mesh->num_reply_states = 0;
 301         mesh->num_detached_states = 0;
 302         mesh->num_forever_states = 0;
 303         mesh->forever_first = NULL;
 304         mesh->forever_last = NULL;
 305         mesh->jostle_first = NULL;
 306         mesh->jostle_last = NULL;
 307 }
 308
 309 int mesh_make_new_space(struct mesh_area* mesh, sldns_buffer* qbuf)
 310 {
 311         struct mesh_state* m = mesh->jostle_first;
 312         /* free space is available */
 313         if(mesh->num_reply_states < mesh->max_reply_states)
 314                 return 1;
 315         /* try to kick out a jostle-list item */
 316         if(m && m->reply_list && m->list_select == mesh_jostle_list) {
 317                 /* how old is it? */
 318                 struct timeval age;
 319                 timeval_subtract(&age, mesh->env->now_tv,
 320                         &m->reply_list->start_time);
 321                 if(timeval_smaller(&mesh->jostle_max, &age)) {
 322                         /* its a goner */
 323                         log_nametypeclass(VERB_ALGO, "query jostled out to "
 324                                 "make space for a new one",
 325                                 m->s.qinfo.qname, m->s.qinfo.qtype,
 326                                 m->s.qinfo.qclass);
 327                         /* backup the query */
 328                         if(qbuf) sldns_buffer_copy(mesh->qbuf_bak, qbuf);
 329                         /* notify supers */
 330                         if(m->super_set.count > 0) {
 331                                 verbose(VERB_ALGO, "notify supers of failure");
 332                                 m->s.return_msg = NULL;
 333                                 m->s.return_rcode = LDNS_RCODE_SERVFAIL;
 334                                 mesh_walk_supers(mesh, m);
 335                         }
 336                         mesh->stats_jostled ++;
 337                         mesh_state_delete(&m->s);
 338                         /* restore the query - note that the qinfo ptr to
 339                          * the querybuffer is then correct again. */
 340                         if(qbuf) sldns_buffer_copy(qbuf, mesh->qbuf_bak);
 341                         return 1;
 342                 }
 343         }
 344         /* no space for new item */
 345         return 0;
 346 }
 347
 348 void mesh_new_client(struct mesh_area* mesh, struct query_info* qinfo,
 349         struct respip_client_info* cinfo, uint16_t qflags,
 350         struct edns_data* edns, struct comm_reply* rep, uint16_t qid)
 351 {
 352         struct mesh_state* s = NULL;
 353         int unique = unique_mesh_state(edns->opt_list, mesh->env);
 354         int was_detached = 0;
 355         int was_noreply = 0;
 356         int added = 0;
 357         struct sldns_buffer* r_buffer = rep->c->buffer;
 358         if(rep->c->tcp_req_info) {
 359                 r_buffer = rep->c->tcp_req_info->spool_buffer;
 360         }
 361         if(!unique)
 362                 s = mesh_area_find(mesh, cinfo, qinfo, qflags&(BIT_RD|BIT_CD), 0, 0);
 363         /* does this create a new reply state? */
 364         if(!s || s->list_select == mesh_no_list) {
 365                 if(!mesh_make_new_space(mesh, rep->c->buffer)) {
 366                         verbose(VERB_ALGO, "Too many queries. dropping "
 367                                 "incoming query.");
 368                         comm_point_drop_reply(rep);
 369                         mesh->stats_dropped ++;
 370                         return;
 371                 }
 372                 /* for this new reply state, the reply address is free,
 373                  * so the limit of reply addresses does not stop reply states*/
 374         } else {
 375                 /* protect our memory usage from storing reply addresses */
 376                 if(mesh->num_reply_addrs > mesh->max_reply_states*16) {
 377                         verbose(VERB_ALGO, "Too many requests queued. "
 378                                 "dropping incoming query.");
 379                         mesh->stats_dropped++;
 380                         comm_point_drop_reply(rep);
 381                         return;
 382                 }
 383         }
 384         /* see if it already exists, if not, create one */
 385         if(!s) {
 386 #ifdef UNBOUND_DEBUG
 387                 struct rbnode_type* n;
 388 #endif
 389                 s = mesh_state_create(mesh->env, qinfo, cinfo,
 390                         qflags&(BIT_RD|BIT_CD), 0, 0);
 391                 if(!s) {
 392                         log_err("mesh_state_create: out of memory; SERVFAIL");
 393                         if(!inplace_cb_reply_servfail_call(mesh->env, qinfo, NULL, NULL,
 394                                 LDNS_RCODE_SERVFAIL, edns, rep, mesh->env->scratch))
 395                                         edns->opt_list = NULL;
 396                         error_encode(r_buffer, LDNS_RCODE_SERVFAIL,
 397                                 qinfo, qid, qflags, edns);
 398                         comm_point_send_reply(rep);
 399                         return;
 400                 }
 401                 if(unique)
 402                         mesh_state_make_unique(s);
 403                 /* copy the edns options we got from the front */
 404                 if(edns->opt_list) {
 405                         s->s.edns_opts_front_in = edns_opt_copy_region(edns->opt_list,
 406                                 s->s.region);
 407                         if(!s->s.edns_opts_front_in) {
 408                                 log_err("mesh_state_create: out of memory; SERVFAIL");
 409                                 if(!inplace_cb_reply_servfail_call(mesh->env, qinfo, NULL,
 410                                         NULL, LDNS_RCODE_SERVFAIL, edns, rep, mesh->env->scratch))
 411                                                 edns->opt_list = NULL;
 412                                 error_encode(r_buffer, LDNS_RCODE_SERVFAIL,
 413                                         qinfo, qid, qflags, edns);
 414                                 comm_point_send_reply(rep);
 415                                 return;
 416                         }
 417                 }
 418
 419 #ifdef UNBOUND_DEBUG
 420                 n =
 421 #else
 422                 (void)
 423 #endif
 424                 rbtree_insert(&mesh->all, &s->node);
 425                 log_assert(n != NULL);
 426                 /* set detached (it is now) */
 427                 mesh->num_detached_states++;
 428                 added = 1;
 429         }
 430         if(!s->reply_list && !s->cb_list && s->super_set.count == 0)
 431                 was_detached = 1;
 432         if(!s->reply_list && !s->cb_list)
 433                 was_noreply = 1;
 434         /* add reply to s */
 435         if(!mesh_state_add_reply(s, edns, rep, qid, qflags, qinfo)) {
 436                         log_err("mesh_new_client: out of memory; SERVFAIL");
 437                 servfail_mem:
 438                         if(!inplace_cb_reply_servfail_call(mesh->env, qinfo, &s->s,
 439                                 NULL, LDNS_RCODE_SERVFAIL, edns, rep, mesh->env->scratch))
 440                                         edns->opt_list = NULL;
 441                         error_encode(r_buffer, LDNS_RCODE_SERVFAIL,
 442                                 qinfo, qid, qflags, edns);
 443                         comm_point_send_reply(rep);
 444                         if(added)
 445                                 mesh_state_delete(&s->s);
 446                         return;
 447         }
 448         if(rep->c->tcp_req_info) {
 449                 if(!tcp_req_info_add_meshstate(rep->c->tcp_req_info, mesh, s)) {
 450                         log_err("mesh_new_client: out of memory add tcpreqinfo");
 451                         goto servfail_mem;
 452                 }
 453         }
 454         /* update statistics */
 455         if(was_detached) {
 456                 log_assert(mesh->num_detached_states > 0);
 457                 mesh->num_detached_states--;
 458         }
 459         if(was_noreply) {
 460                 mesh->num_reply_states ++;
 461         }
 462         mesh->num_reply_addrs++;
 463         if(s->list_select == mesh_no_list) {
 464                 /* move to either the forever or the jostle_list */
 465                 if(mesh->num_forever_states < mesh->max_forever_states) {
 466                         mesh->num_forever_states ++;
 467                         mesh_list_insert(s, &mesh->forever_first,
 468                                 &mesh->forever_last);
 469                         s->list_select = mesh_forever_list;
 470                 } else {
 471                         mesh_list_insert(s, &mesh->jostle_first,
 472                                 &mesh->jostle_last);
 473                         s->list_select = mesh_jostle_list;
 474                 }
 475         }
 476         if(added)
 477                 mesh_run(mesh, s, module_event_new, NULL);
 478 }
 479
 480 int
 481 mesh_new_callback(struct mesh_area* mesh, struct query_info* qinfo,
 482         uint16_t qflags, struct edns_data* edns, sldns_buffer* buf,
 483         uint16_t qid, mesh_cb_func_type cb, void* cb_arg)
 484 {
 485         struct mesh_state* s = NULL;
 486         int unique = unique_mesh_state(edns->opt_list, mesh->env);
 487         int was_detached = 0;
 488         int was_noreply = 0;
 489         int added = 0;
 490         if(!unique)
 491                 s = mesh_area_find(mesh, NULL, qinfo, qflags&(BIT_RD|BIT_CD), 0, 0);
 492
 493         /* there are no limits on the number of callbacks */
 494
 495         /* see if it already exists, if not, create one */
 496         if(!s) {
 497 #ifdef UNBOUND_DEBUG
 498                 struct rbnode_type* n;
 499 #endif
 500                 s = mesh_state_create(mesh->env, qinfo, NULL,
 501                         qflags&(BIT_RD|BIT_CD), 0, 0);
 502                 if(!s) {
 503                         return 0;
 504                 }
 505                 if(unique)
 506                         mesh_state_make_unique(s);
 507                 if(edns->opt_list) {
 508                         s->s.edns_opts_front_in = edns_opt_copy_region(edns->opt_list,
 509                                 s->s.region);
 510                         if(!s->s.edns_opts_front_in) {
 511                                 return 0;
 512                         }
 513                 }
 514 #ifdef UNBOUND_DEBUG
 515                 n =
 516 #else
 517                 (void)
 518 #endif
 519                 rbtree_insert(&mesh->all, &s->node);
 520                 log_assert(n != NULL);
 521                 /* set detached (it is now) */
 522                 mesh->num_detached_states++;
 523                 added = 1;
 524         }
 525         if(!s->reply_list && !s->cb_list && s->super_set.count == 0)
 526                 was_detached = 1;
 527         if(!s->reply_list && !s->cb_list)
 528                 was_noreply = 1;
 529         /* add reply to s */
 530         if(!mesh_state_add_cb(s, edns, buf, cb, cb_arg, qid, qflags)) {
 531                         if(added)
 532                                 mesh_state_delete(&s->s);
 533                         return 0;
 534         }
 535         /* update statistics */
 536         if(was_detached) {
 537                 log_assert(mesh->num_detached_states > 0);
 538                 mesh->num_detached_states--;
 539         }
 540         if(was_noreply) {
 541                 mesh->num_reply_states ++;
 542         }
 543         mesh->num_reply_addrs++;
 544         if(added)
 545                 mesh_run(mesh, s, module_event_new, NULL);
 546         return 1;
 547 }
 548
 549 static void mesh_schedule_prefetch(struct mesh_area* mesh,
 550         struct query_info* qinfo, uint16_t qflags, time_t leeway, int run);
 551
 552 void mesh_new_prefetch(struct mesh_area* mesh, struct query_info* qinfo,
 553         uint16_t qflags, time_t leeway)
 554 {
 555         mesh_schedule_prefetch(mesh, qinfo, qflags, leeway, 1);
 556 }
 557
 558 /* Internal backend routine of mesh_new_prefetch().  It takes one additional
 559  * parameter, 'run', which controls whether to run the prefetch state
 560  * immediately.  When this function is called internally 'run' could be
 561  * 0 (false), in which case the new state is only made runnable so it
 562  * will not be run recursively on top of the current state. */
 563 static void mesh_schedule_prefetch(struct mesh_area* mesh,
 564         struct query_info* qinfo, uint16_t qflags, time_t leeway, int run)
 565 {
 566         struct mesh_state* s = mesh_area_find(mesh, NULL, qinfo,
 567                 qflags&(BIT_RD|BIT_CD), 0, 0);
 568 #ifdef UNBOUND_DEBUG
 569         struct rbnode_type* n;
 570 #endif
 571         /* already exists, and for a different purpose perhaps.
 572          * if mesh_no_list, keep it that way. */
 573         if(s) {
 574                 /* make it ignore the cache from now on */
 575                 if(!s->s.blacklist)
 576                         sock_list_insert(&s->s.blacklist, NULL, 0, s->s.region);
 577                 if(s->s.prefetch_leeway < leeway)
 578                         s->s.prefetch_leeway = leeway;
 579                 return;
 580         }
 581         if(!mesh_make_new_space(mesh, NULL)) {
 582                 verbose(VERB_ALGO, "Too many queries. dropped prefetch.");
 583                 mesh->stats_dropped ++;
 584                 return;
 585         }
 586
 587         s = mesh_state_create(mesh->env, qinfo, NULL,
 588                 qflags&(BIT_RD|BIT_CD), 0, 0);
 589         if(!s) {
 590                 log_err("prefetch mesh_state_create: out of memory");
 591                 return;
 592         }
 593 #ifdef UNBOUND_DEBUG
 594         n =
 595 #else
 596         (void)
 597 #endif
 598         rbtree_insert(&mesh->all, &s->node);
 599         log_assert(n != NULL);
 600         /* set detached (it is now) */
 601         mesh->num_detached_states++;
 602         /* make it ignore the cache */
 603         sock_list_insert(&s->s.blacklist, NULL, 0, s->s.region);
 604         s->s.prefetch_leeway = leeway;
 605
 606         if(s->list_select == mesh_no_list) {
 607                 /* move to either the forever or the jostle_list */
 608                 if(mesh->num_forever_states < mesh->max_forever_states) {
 609                         mesh->num_forever_states ++;
 610                         mesh_list_insert(s, &mesh->forever_first,
 611                                 &mesh->forever_last);
 612                         s->list_select = mesh_forever_list;
 613                 } else {
 614                         mesh_list_insert(s, &mesh->jostle_first,
 615                                 &mesh->jostle_last);
 616                         s->list_select = mesh_jostle_list;
 617                 }
 618         }
 619
 620         if(!run) {
 621 #ifdef UNBOUND_DEBUG
 622                 n =
 623 #else
 624                 (void)
 625 #endif
 626                 rbtree_insert(&mesh->run, &s->run_node);
 627                 log_assert(n != NULL);
 628                 return;
 629         }
 630
 631         mesh_run(mesh, s, module_event_new, NULL);
 632 }
 633
 634 void mesh_report_reply(struct mesh_area* mesh, struct outbound_entry* e,
 635         struct comm_reply* reply, int what)
 636 {
 637         enum module_ev event = module_event_reply;
 638         e->qstate->reply = reply;
 639         if(what != NETEVENT_NOERROR) {
 640                 event = module_event_noreply;
 641                 if(what == NETEVENT_CAPSFAIL)
 642                         event = module_event_capsfail;
 643         }
 644         mesh_run(mesh, e->qstate->mesh_info, event, e);
 645 }
 646
 647 struct mesh_state*
 648 mesh_state_create(struct module_env* env, struct query_info* qinfo,
 649         struct respip_client_info* cinfo, uint16_t qflags, int prime,
 650         int valrec)
 651 {
 652         struct regional* region = alloc_reg_obtain(env->alloc);
 653         struct mesh_state* mstate;
 654         int i;
 655         if(!region)
 656                 return NULL;
 657         mstate = (struct mesh_state*)regional_alloc(region,
 658                 sizeof(struct mesh_state));
 659         if(!mstate) {
 660                 alloc_reg_release(env->alloc, region);
 661                 return NULL;
 662         }
 663         memset(mstate, 0, sizeof(*mstate));
 664         mstate->node = *RBTREE_NULL;
 665         mstate->run_node = *RBTREE_NULL;
 666         mstate->node.key = mstate;
 667         mstate->run_node.key = mstate;
 668         mstate->reply_list = NULL;
 669         mstate->list_select = mesh_no_list;
 670         mstate->replies_sent = 0;
 671         rbtree_init(&mstate->super_set, &mesh_state_ref_compare);
 672         rbtree_init(&mstate->sub_set, &mesh_state_ref_compare);
 673         mstate->num_activated = 0;
 674         mstate->unique = NULL;
 675         /* init module qstate */
 676         mstate->s.qinfo.qtype = qinfo->qtype;
 677         mstate->s.qinfo.qclass = qinfo->qclass;
 678         mstate->s.qinfo.local_alias = NULL;
 679         mstate->s.qinfo.qname_len = qinfo->qname_len;
 680         mstate->s.qinfo.qname = regional_alloc_init(region, qinfo->qname,
 681                 qinfo->qname_len);
 682         if(!mstate->s.qinfo.qname) {
 683                 alloc_reg_release(env->alloc, region);
 684                 return NULL;
 685         }
 686         if(cinfo) {
 687                 mstate->s.client_info = regional_alloc_init(region, cinfo,
 688                         sizeof(*cinfo));
 689                 if(!mstate->s.client_info) {
 690                         alloc_reg_release(env->alloc, region);
 691                         return NULL;
 692                 }
 693         }
 694         /* remove all weird bits from qflags */
 695         mstate->s.query_flags = (qflags & (BIT_RD|BIT_CD));
 696         mstate->s.is_priming = prime;
 697         mstate->s.is_valrec = valrec;
 698         mstate->s.reply = NULL;
 699         mstate->s.region = region;
 700         mstate->s.curmod = 0;
 701         mstate->s.return_msg = 0;
 702         mstate->s.return_rcode = LDNS_RCODE_NOERROR;
 703         mstate->s.env = env;
 704         mstate->s.mesh_info = mstate;
 705         mstate->s.prefetch_leeway = 0;
 706         mstate->s.no_cache_lookup = 0;
 707         mstate->s.no_cache_store = 0;
 708         mstate->s.need_refetch = 0;
 709         mstate->s.was_ratelimited = 0;
 710
 711         /* init modules */
 712         for(i=0; i<env->mesh->mods.num; i++) {
 713                 mstate->s.minfo[i] = NULL;
 714                 mstate->s.ext_state[i] = module_state_initial;
 715         }
 716         /* init edns option lists */
 717         mstate->s.edns_opts_front_in = NULL;
 718         mstate->s.edns_opts_back_out = NULL;
 719         mstate->s.edns_opts_back_in = NULL;
 720         mstate->s.edns_opts_front_out = NULL;
 721
 722         return mstate;
 723 }
 724
 725 int
 726 mesh_state_is_unique(struct mesh_state* mstate)
 727 {
 728         return mstate->unique != NULL;
 729 }
 730
 731 void
 732 mesh_state_make_unique(struct mesh_state* mstate)
 733 {
 734         mstate->unique = mstate;
 735 }
 736
 737 void
 738 mesh_state_cleanup(struct mesh_state* mstate)
 739 {
 740         struct mesh_area* mesh;
 741         int i;
 742         if(!mstate)
 743                 return;
 744         mesh = mstate->s.env->mesh;
 745         /* drop unsent replies */
 746         if(!mstate->replies_sent) {
 747                 struct mesh_reply* rep = mstate->reply_list;
 748                 struct mesh_cb* cb;
 749                 /* in tcp_req_info, the mstates linked are removed, but
 750                  * the reply_list is now NULL, so the remove-from-empty-list
 751                  * takes no time and also it does not do the mesh accounting */
 752                 mstate->reply_list = NULL;
 753                 for(; rep; rep=rep->next) {
 754                         comm_point_drop_reply(&rep->query_reply);
 755                         mesh->num_reply_addrs--;
 756                 }
 757                 while((cb = mstate->cb_list)!=NULL) {
 758                         mstate->cb_list = cb->next;
 759                         fptr_ok(fptr_whitelist_mesh_cb(cb->cb));
 760                         (*cb->cb)(cb->cb_arg, LDNS_RCODE_SERVFAIL, NULL,
 761                                 sec_status_unchecked, NULL, 0);
 762                         mesh->num_reply_addrs--;
 763                 }
 764         }
 765
 766         /* de-init modules */
 767         for(i=0; i<mesh->mods.num; i++) {
 768                 fptr_ok(fptr_whitelist_mod_clear(mesh->mods.mod[i]->clear));
 769                 (*mesh->mods.mod[i]->clear)(&mstate->s, i);
 770                 mstate->s.minfo[i] = NULL;
 771                 mstate->s.ext_state[i] = module_finished;
 772         }
 773         alloc_reg_release(mstate->s.env->alloc, mstate->s.region);
 774 }
 775
 776 void
 777 mesh_state_delete(struct module_qstate* qstate)
 778 {
 779         struct mesh_area* mesh;
 780         struct mesh_state_ref* super, ref;
 781         struct mesh_state* mstate;
 782         if(!qstate)
 783                 return;
 784         mstate = qstate->mesh_info;
 785         mesh = mstate->s.env->mesh;
 786         mesh_detach_subs(&mstate->s);
 787         if(mstate->list_select == mesh_forever_list) {
 788                 mesh->num_forever_states --;
 789                 mesh_list_remove(mstate, &mesh->forever_first,
 790                         &mesh->forever_last);
 791         } else if(mstate->list_select == mesh_jostle_list) {
 792                 mesh_list_remove(mstate, &mesh->jostle_first,
 793                         &mesh->jostle_last);
 794         }
 795         if(!mstate->reply_list && !mstate->cb_list
 796                 && mstate->super_set.count == 0) {
 797                 log_assert(mesh->num_detached_states > 0);
 798                 mesh->num_detached_states--;
 799         }
 800         if(mstate->reply_list || mstate->cb_list) {
 801                 log_assert(mesh->num_reply_states > 0);
 802                 mesh->num_reply_states--;
 803         }
 804         ref.node.key = &ref;
 805         ref.s = mstate;
 806         RBTREE_FOR(super, struct mesh_state_ref*, &mstate->super_set) {
 807                 (void)rbtree_delete(&super->s->sub_set, &ref);
 808         }
 809         (void)rbtree_delete(&mesh->run, mstate);
 810         (void)rbtree_delete(&mesh->all, mstate);
 811         mesh_state_cleanup(mstate);
 812 }
 813
 814 /** helper recursive rbtree find routine */
 815 static int
 816 find_in_subsub(struct mesh_state* m, struct mesh_state* tofind, size_t *c)
 817 {
 818         struct mesh_state_ref* r;
 819         if((*c)++ > MESH_MAX_SUBSUB)
 820                 return 1;
 821         RBTREE_FOR(r, struct mesh_state_ref*, &m->sub_set) {
 822                 if(r->s == tofind || find_in_subsub(r->s, tofind, c))
 823                         return 1;
 824         }
 825         return 0;
 826 }
 827
 828 /** find cycle for already looked up mesh_state */
 829 static int
 830 mesh_detect_cycle_found(struct module_qstate* qstate, struct mesh_state* dep_m)
 831 {
 832         struct mesh_state* cyc_m = qstate->mesh_info;
 833         size_t counter = 0;
 834         if(!dep_m)
 835                 return 0;
 836         if(dep_m == cyc_m || find_in_subsub(dep_m, cyc_m, &counter)) {
 837                 if(counter > MESH_MAX_SUBSUB)
 838                         return 2;
 839                 return 1;
 840         }
 841         return 0;
 842 }
 843
 844 void mesh_detach_subs(struct module_qstate* qstate)
 845 {
 846         struct mesh_area* mesh = qstate->env->mesh;
 847         struct mesh_state_ref* ref, lookup;
 848 #ifdef UNBOUND_DEBUG
 849         struct rbnode_type* n;
 850 #endif
 851         lookup.node.key = &lookup;
 852         lookup.s = qstate->mesh_info;
 853         RBTREE_FOR(ref, struct mesh_state_ref*, &qstate->mesh_info->sub_set) {
 854 #ifdef UNBOUND_DEBUG
 855                 n =
 856 #else
 857                 (void)
 858 #endif
 859                 rbtree_delete(&ref->s->super_set, &lookup);
 860                 log_assert(n != NULL); /* must have been present */
 861                 if(!ref->s->reply_list && !ref->s->cb_list
 862                         && ref->s->super_set.count == 0) {
 863                         mesh->num_detached_states++;
 864                         log_assert(mesh->num_detached_states +
 865                                 mesh->num_reply_states <= mesh->all.count);
 866                 }
 867         }
 868         rbtree_init(&qstate->mesh_info->sub_set, &mesh_state_ref_compare);
 869 }
 870
 871 int mesh_add_sub(struct module_qstate* qstate, struct query_info* qinfo,
 872         uint16_t qflags, int prime, int valrec, struct module_qstate** newq,
 873         struct mesh_state** sub)
 874 {
 875         /* find it, if not, create it */
 876         struct mesh_area* mesh = qstate->env->mesh;
 877         *sub = mesh_area_find(mesh, NULL, qinfo, qflags,
 878                 prime, valrec);
 879         if(mesh_detect_cycle_found(qstate, *sub)) {
 880                 verbose(VERB_ALGO, "attach failed, cycle detected");
 881                 return 0;
 882         }
 883         if(!*sub) {
 884 #ifdef UNBOUND_DEBUG
 885                 struct rbnode_type* n;
 886 #endif
 887                 /* create a new one */
 888                 *sub = mesh_state_create(qstate->env, qinfo, NULL, qflags, prime,
 889                         valrec);
 890                 if(!*sub) {
 891                         log_err("mesh_attach_sub: out of memory");
 892                         return 0;
 893                 }
 894 #ifdef UNBOUND_DEBUG
 895                 n =
 896 #else
 897                 (void)
 898 #endif
 899                 rbtree_insert(&mesh->all, &(*sub)->node);
 900                 log_assert(n != NULL);
 901                 /* set detached (it is now) */
 902                 mesh->num_detached_states++;
 903                 /* set new query state to run */
 904 #ifdef UNBOUND_DEBUG
 905                 n =
 906 #else
 907                 (void)
 908 #endif
 909                 rbtree_insert(&mesh->run, &(*sub)->run_node);
 910                 log_assert(n != NULL);
 911                 *newq = &(*sub)->s;
 912         } else
 913                 *newq = NULL;
 914         return 1;
 915 }
 916
 917 int mesh_attach_sub(struct module_qstate* qstate, struct query_info* qinfo,
 918         uint16_t qflags, int prime, int valrec, struct module_qstate** newq)
 919 {
 920         struct mesh_area* mesh = qstate->env->mesh;
 921         struct mesh_state* sub = NULL;
 922         int was_detached;
 923         if(!mesh_add_sub(qstate, qinfo, qflags, prime, valrec, newq, &sub))
 924                 return 0;
 925         was_detached = (sub->super_set.count == 0);
 926         if(!mesh_state_attachment(qstate->mesh_info, sub))
 927                 return 0;
 928         /* if it was a duplicate  attachment, the count was not zero before */
 929         if(!sub->reply_list && !sub->cb_list && was_detached &&
 930                 sub->super_set.count == 1) {
 931                 /* it used to be detached, before this one got added */
 932                 log_assert(mesh->num_detached_states > 0);
 933                 mesh->num_detached_states--;
 934         }
 935         /* *newq will be run when inited after the current module stops */
 936         return 1;
 937 }
 938
 939 int mesh_state_attachment(struct mesh_state* super, struct mesh_state* sub)
 940 {
 941 #ifdef UNBOUND_DEBUG
 942         struct rbnode_type* n;
 943 #endif
 944         struct mesh_state_ref* subref; /* points to sub, inserted in super */
 945         struct mesh_state_ref* superref; /* points to super, inserted in sub */
 946         if( !(subref = regional_alloc(super->s.region,
 947                 sizeof(struct mesh_state_ref))) ||
 948                 !(superref = regional_alloc(sub->s.region,
 949                 sizeof(struct mesh_state_ref))) ) {
 950                 log_err("mesh_state_attachment: out of memory");
 951                 return 0;
 952         }
 953         superref->node.key = superref;
 954         superref->s = super;
 955         subref->node.key = subref;
 956         subref->s = sub;
 957         if(!rbtree_insert(&sub->super_set, &superref->node)) {
 958                 /* this should not happen, iterator and validator do not
 959                  * attach subqueries that are identical. */
 960                 /* already attached, we are done, nothing todo.
 961                  * since superref and subref already allocated in region,
 962                  * we cannot free them */
 963                 return 1;
 964         }
 965 #ifdef UNBOUND_DEBUG
 966         n =
 967 #else
 968         (void)
 969 #endif
 970         rbtree_insert(&super->sub_set, &subref->node);
 971         log_assert(n != NULL); /* we checked above if statement, the reverse
 972           administration should not fail now, unless they are out of sync */
 973         return 1;
 974 }
 975
 976 /**
 977  * callback results to mesh cb entry
 978  * @param m: mesh state to send it for.
 979  * @param rcode: if not 0, error code.
 980  * @param rep: reply to send (or NULL if rcode is set).
 981  * @param r: callback entry
 982  */
 983 static void
 984 mesh_do_callback(struct mesh_state* m, int rcode, struct reply_info* rep,
 985         struct mesh_cb* r)
 986 {
 987         int secure;
 988         char* reason = NULL;
 989         int was_ratelimited = m->s.was_ratelimited;
 990         /* bogus messages are not made into servfail, sec_status passed
 991          * to the callback function */
 992         if(rep && rep->security == sec_status_secure)
 993                 secure = 1;
 994         else    secure = 0;
 995         if(!rep && rcode == LDNS_RCODE_NOERROR)
 996                 rcode = LDNS_RCODE_SERVFAIL;
 997         if(!rcode && (rep->security == sec_status_bogus ||
 998                 rep->security == sec_status_secure_sentinel_fail)) {
 999                 if(!(reason = errinf_to_str_bogus(&m->s)))
1000                         rcode = LDNS_RCODE_SERVFAIL;
1001         }
1002         /* send the reply */
1003         if(rcode) {
1004                 if(rcode == LDNS_RCODE_SERVFAIL) {
1005                         if(!inplace_cb_reply_servfail_call(m->s.env, &m->s.qinfo, &m->s,
1006                                 rep, rcode, &r->edns, NULL, m->s.region))
1007                                         r->edns.opt_list = NULL;
1008                 } else {
1009                         if(!inplace_cb_reply_call(m->s.env, &m->s.qinfo, &m->s, rep, rcode,
1010                                 &r->edns, NULL, m->s.region))
1011                                         r->edns.opt_list = NULL;
1012                 }
1013                 fptr_ok(fptr_whitelist_mesh_cb(r->cb));
1014                 (*r->cb)(r->cb_arg, rcode, r->buf, sec_status_unchecked, NULL,
1015                         was_ratelimited);
1016         } else {
1017                 size_t udp_size = r->edns.udp_size;
1018                 sldns_buffer_clear(r->buf);
1019                 r->edns.edns_version = EDNS_ADVERTISED_VERSION;
1020                 r->edns.udp_size = EDNS_ADVERTISED_SIZE;
1021                 r->edns.ext_rcode = 0;
1022                 r->edns.bits &= EDNS_DO;
1023
1024                 if(!inplace_cb_reply_call(m->s.env, &m->s.qinfo, &m->s, rep,
1025                         LDNS_RCODE_NOERROR, &r->edns, NULL, m->s.region) ||
1026                         !reply_info_answer_encode(&m->s.qinfo, rep, r->qid,
1027                         r->qflags, r->buf, 0, 1,
1028                         m->s.env->scratch, udp_size, &r->edns,
1029                         (int)(r->edns.bits & EDNS_DO), secure))
1030                 {
1031                         fptr_ok(fptr_whitelist_mesh_cb(r->cb));
1032                         (*r->cb)(r->cb_arg, LDNS_RCODE_SERVFAIL, r->buf,
1033                                 sec_status_unchecked, NULL, 0);
1034                 } else {
1035                         fptr_ok(fptr_whitelist_mesh_cb(r->cb));
1036                         (*r->cb)(r->cb_arg, LDNS_RCODE_NOERROR, r->buf,
1037                                 rep->security, reason, was_ratelimited);
1038                 }
1039         }
1040         free(reason);
1041         m->s.env->mesh->num_reply_addrs--;
1042 }
1043
1044 /**
1045  * Send reply to mesh reply entry
1046  * @param m: mesh state to send it for.
1047  * @param rcode: if not 0, error code.
1048  * @param rep: reply to send (or NULL if rcode is set).
1049  * @param r: reply entry
1050  * @param r_buffer: buffer to use for reply entry.
1051  * @param prev: previous reply, already has its answer encoded in buffer.
1052  * @param prev_buffer: buffer for previous reply.
1053  */
1054 static void
1055 mesh_send_reply(struct mesh_state* m, int rcode, struct reply_info* rep,
1056         struct mesh_reply* r, struct sldns_buffer* r_buffer,
1057         struct mesh_reply* prev, struct sldns_buffer* prev_buffer)
1058 {
1059         struct timeval end_time;
1060         struct timeval duration;
1061         int secure;
1062         /* Copy the client's EDNS for later restore, to make sure the edns
1063          * compare is with the correct edns options. */
1064         struct edns_data edns_bak = r->edns;
1065         /* examine security status */
1066         if(m->s.env->need_to_validate && (!(r->qflags&BIT_CD) ||
1067                 m->s.env->cfg->ignore_cd) && rep &&
1068                 (rep->security <= sec_status_bogus ||
1069                 rep->security == sec_status_secure_sentinel_fail)) {
1070                 rcode = LDNS_RCODE_SERVFAIL;
1071                 if(m->s.env->cfg->stat_extended)
1072                         m->s.env->mesh->ans_bogus++;
1073         }
1074         if(rep && rep->security == sec_status_secure)
1075                 secure = 1;
1076         else    secure = 0;
1077         if(!rep && rcode == LDNS_RCODE_NOERROR)
1078                 rcode = LDNS_RCODE_SERVFAIL;
1079         /* send the reply */
1080         /* We don't reuse the encoded answer if either the previous or current
1081          * response has a local alias.  We could compare the alias records
1082          * and still reuse the previous answer if they are the same, but that
1083          * would be complicated and error prone for the relatively minor case.
1084          * So we err on the side of safety. */
1085         if(prev && prev_buffer && prev->qflags == r->qflags &&
1086                 !prev->local_alias && !r->local_alias &&
1087                 prev->edns.edns_present == r->edns.edns_present &&
1088                 prev->edns.bits == r->edns.bits &&
1089                 prev->edns.udp_size == r->edns.udp_size &&
1090                 edns_opt_list_compare(prev->edns.opt_list, r->edns.opt_list)
1091                 == 0) {
1092                 /* if the previous reply is identical to this one, fix ID */
1093                 if(prev_buffer != r_buffer)
1094                         sldns_buffer_copy(r_buffer, prev_buffer);
1095                 sldns_buffer_write_at(r_buffer, 0, &r->qid, sizeof(uint16_t));
1096                 sldns_buffer_write_at(r_buffer, 12, r->qname,
1097                         m->s.qinfo.qname_len);
1098                 comm_point_send_reply(&r->query_reply);
1099         } else if(rcode) {
1100                 m->s.qinfo.qname = r->qname;
1101                 m->s.qinfo.local_alias = r->local_alias;
1102                 if(rcode == LDNS_RCODE_SERVFAIL) {
1103                         if(!inplace_cb_reply_servfail_call(m->s.env, &m->s.qinfo, &m->s,
1104                                 rep, rcode, &r->edns, NULL, m->s.region))
1105                                         r->edns.opt_list = NULL;
1106                 } else {
1107                         if(!inplace_cb_reply_call(m->s.env, &m->s.qinfo, &m->s, rep, rcode,
1108                                 &r->edns, NULL, m->s.region))
1109                                         r->edns.opt_list = NULL;
1110                 }
1111                 error_encode(r_buffer, rcode, &m->s.qinfo, r->qid,
1112                         r->qflags, &r->edns);
1113                 comm_point_send_reply(&r->query_reply);
1114         } else {
1115                 size_t udp_size = r->edns.udp_size;
1116                 r->edns.edns_version = EDNS_ADVERTISED_VERSION;
1117                 r->edns.udp_size = EDNS_ADVERTISED_SIZE;
1118                 r->edns.ext_rcode = 0;
1119                 r->edns.bits &= EDNS_DO;
1120                 m->s.qinfo.qname = r->qname;
1121                 m->s.qinfo.local_alias = r->local_alias;
1122                 if(!inplace_cb_reply_call(m->s.env, &m->s.qinfo, &m->s, rep,
1123                         LDNS_RCODE_NOERROR, &r->edns, NULL, m->s.region) ||
1124                         !apply_edns_options(&r->edns, &edns_bak,
1125                                 m->s.env->cfg, r->query_reply.c,
1126                                 m->s.region) ||
1127                         !reply_info_answer_encode(&m->s.qinfo, rep, r->qid,
1128                         r->qflags, r_buffer, 0, 1, m->s.env->scratch,
1129                         udp_size, &r->edns, (int)(r->edns.bits & EDNS_DO),
1130                         secure))
1131                 {
1132                         if(!inplace_cb_reply_servfail_call(m->s.env, &m->s.qinfo, &m->s,
1133                         rep, LDNS_RCODE_SERVFAIL, &r->edns, NULL, m->s.region))
1134                                 r->edns.opt_list = NULL;
1135                         error_encode(r_buffer, LDNS_RCODE_SERVFAIL,
1136                                 &m->s.qinfo, r->qid, r->qflags, &r->edns);
1137                 }
1138                 r->edns = edns_bak;
1139                 comm_point_send_reply(&r->query_reply);
1140         }
1141         /* account */
1142         m->s.env->mesh->num_reply_addrs--;
1143         end_time = *m->s.env->now_tv;
1144         timeval_subtract(&duration, &end_time, &r->start_time);
1145         verbose(VERB_ALGO, "query took " ARG_LL "d.%6.6d sec",
1146                 (long long)duration.tv_sec, (int)duration.tv_usec);
1147         m->s.env->mesh->replies_sent++;
1148         timeval_add(&m->s.env->mesh->replies_sum_wait, &duration);
1149         timehist_insert(m->s.env->mesh->histogram, &duration);
1150         if(m->s.env->cfg->stat_extended) {
1151                 uint16_t rc = FLAGS_GET_RCODE(sldns_buffer_read_u16_at(
1152                         r_buffer, 2));
1153                 if(secure) m->s.env->mesh->ans_secure++;
1154                 m->s.env->mesh->ans_rcode[ rc ] ++;
1155                 if(rc == 0 && LDNS_ANCOUNT(sldns_buffer_begin(r_buffer)) == 0)
1156                         m->s.env->mesh->ans_nodata++;
1157         }
1158         /* Log reply sent */
1159         if(m->s.env->cfg->log_replies) {
1160                 log_reply_info(0, &m->s.qinfo, &r->query_reply.addr,
1161                         r->query_reply.addrlen, duration, 0, r_buffer);
1162         }
1163 }
1164
1165 void mesh_query_done(struct mesh_state* mstate)
1166 {
1167         struct mesh_reply* r;
1168         struct mesh_reply* prev = NULL;
1169         struct sldns_buffer* prev_buffer = NULL;
1170         struct mesh_cb* c;
1171         struct reply_info* rep = (mstate->s.return_msg?
1172                 mstate->s.return_msg->rep:NULL);
1173         if((mstate->s.return_rcode == LDNS_RCODE_SERVFAIL ||
1174                 (rep && FLAGS_GET_RCODE(rep->flags) == LDNS_RCODE_SERVFAIL))
1175                 && mstate->s.env->cfg->log_servfail
1176                 && !mstate->s.env->cfg->val_log_squelch) {
1177                 char* err = errinf_to_str_servfail(&mstate->s);
1178                 if(err)
1179                         log_err("%s", err);
1180                 free(err);
1181         }
1182         for(r = mstate->reply_list; r; r = r->next) {
1183                 /* if a response-ip address block has been stored the
1184                  *  information should be logged for each client. */
1185                 if(mstate->s.respip_action_info &&
1186                         mstate->s.respip_action_info->addrinfo) {
1187                         respip_inform_print(mstate->s.respip_action_info->addrinfo,
1188                                 r->qname, mstate->s.qinfo.qtype,
1189                                 mstate->s.qinfo.qclass, r->local_alias,
1190                                 &r->query_reply);
1191                 }
1192
1193                 /* if this query is determined to be dropped during the
1194                  * mesh processing, this is the point to take that action. */
1195                 if(mstate->s.is_drop)
1196                         comm_point_drop_reply(&r->query_reply);
1197                 else {
1198                         struct sldns_buffer* r_buffer = r->query_reply.c->buffer;
1199                         if(r->query_reply.c->tcp_req_info) {
1200                                 r_buffer = r->query_reply.c->tcp_req_info->spool_buffer;
1201                                 prev_buffer = NULL;
1202                         }
1203                         mesh_send_reply(mstate, mstate->s.return_rcode, rep,
1204                                 r, r_buffer, prev, prev_buffer);
1205                         if(r->query_reply.c->tcp_req_info) {
1206                                 tcp_req_info_remove_mesh_state(r->query_reply.c->tcp_req_info, mstate);
1207                                 r_buffer = NULL;
1208                         }
1209                         prev = r;
1210                         prev_buffer = r_buffer;
1211                 }
1212         }
1213         mstate->replies_sent = 1;
1214         while((c = mstate->cb_list) != NULL) {
1215                 /* take this cb off the list; so that the list can be
1216                  * changed, eg. by adds from the callback routine */
1217                 if(!mstate->reply_list && mstate->cb_list && !c->next) {
1218                         /* was a reply state, not anymore */
1219                         mstate->s.env->mesh->num_reply_states--;
1220                 }
1221                 mstate->cb_list = c->next;
1222                 if(!mstate->reply_list && !mstate->cb_list &&
1223                         mstate->super_set.count == 0)
1224                         mstate->s.env->mesh->num_detached_states++;
1225                 mesh_do_callback(mstate, mstate->s.return_rcode, rep, c);
1226         }
1227 }
1228
1229 void mesh_walk_supers(struct mesh_area* mesh, struct mesh_state* mstate)
1230 {
1231         struct mesh_state_ref* ref;
1232         RBTREE_FOR(ref, struct mesh_state_ref*, &mstate->super_set)
1233         {
1234                 /* make super runnable */
1235                 (void)rbtree_insert(&mesh->run, &ref->s->run_node);
1236                 /* callback the function to inform super of result */
1237                 fptr_ok(fptr_whitelist_mod_inform_super(
1238                         mesh->mods.mod[ref->s->s.curmod]->inform_super));
1239                 (*mesh->mods.mod[ref->s->s.curmod]->inform_super)(&mstate->s,
1240                         ref->s->s.curmod, &ref->s->s);
1241                 /* copy state that is always relevant to super */
1242                 copy_state_to_super(&mstate->s, ref->s->s.curmod, &ref->s->s);
1243         }
1244 }
1245
1246 struct mesh_state* mesh_area_find(struct mesh_area* mesh,
1247         struct respip_client_info* cinfo, struct query_info* qinfo,
1248         uint16_t qflags, int prime, int valrec)
1249 {
1250         struct mesh_state key;
1251         struct mesh_state* result;
1252
1253         key.node.key = &key;
1254         key.s.is_priming = prime;
1255         key.s.is_valrec = valrec;
1256         key.s.qinfo = *qinfo;
1257         key.s.query_flags = qflags;
1258         /* We are searching for a similar mesh state when we DO want to
1259          * aggregate the state. Thus unique is set to NULL. (default when we
1260          * desire aggregation).*/
1261         key.unique = NULL;
1262         key.s.client_info = cinfo;
1263
1264         result = (struct mesh_state*)rbtree_search(&mesh->all, &key);
1265         return result;
1266 }
1267
1268 int mesh_state_add_cb(struct mesh_state* s, struct edns_data* edns,
1269         sldns_buffer* buf, mesh_cb_func_type cb, void* cb_arg,
1270         uint16_t qid, uint16_t qflags)
1271 {
1272         struct mesh_cb* r = regional_alloc(s->s.region,
1273                 sizeof(struct mesh_cb));
1274         if(!r)
1275                 return 0;
1276         r->buf = buf;
1277         log_assert(fptr_whitelist_mesh_cb(cb)); /* early failure ifmissing*/
1278         r->cb = cb;
1279         r->cb_arg = cb_arg;
1280         r->edns = *edns;
1281         if(edns->opt_list) {
1282                 r->edns.opt_list = edns_opt_copy_region(edns->opt_list,
1283                         s->s.region);
1284                 if(!r->edns.opt_list)
1285                         return 0;
1286         }
1287         r->qid = qid;
1288         r->qflags = qflags;
1289         r->next = s->cb_list;
1290         s->cb_list = r;
1291         return 1;
1292
1293 }
1294
1295 int mesh_state_add_reply(struct mesh_state* s, struct edns_data* edns,
1296         struct comm_reply* rep, uint16_t qid, uint16_t qflags,
1297         const struct query_info* qinfo)
1298 {
1299         struct mesh_reply* r = regional_alloc(s->s.region,
1300                 sizeof(struct mesh_reply));
1301         if(!r)
1302                 return 0;
1303         r->query_reply = *rep;
1304         r->edns = *edns;
1305         if(edns->opt_list) {
1306                 r->edns.opt_list = edns_opt_copy_region(edns->opt_list,
1307                         s->s.region);
1308                 if(!r->edns.opt_list)
1309                         return 0;
1310         }
1311         r->qid = qid;
1312         r->qflags = qflags;
1313         r->start_time = *s->s.env->now_tv;
1314         r->next = s->reply_list;
1315         r->qname = regional_alloc_init(s->s.region, qinfo->qname,
1316                 s->s.qinfo.qname_len);
1317         if(!r->qname)
1318                 return 0;
1319
1320         /* Data related to local alias stored in 'qinfo' (if any) is ephemeral
1321          * and can be different for different original queries (even if the
1322          * replaced query name is the same).  So we need to make a deep copy
1323          * and store the copy for each reply info. */
1324         if(qinfo->local_alias) {
1325                 struct packed_rrset_data* d;
1326                 struct packed_rrset_data* dsrc;
1327                 r->local_alias = regional_alloc_zero(s->s.region,
1328                         sizeof(*qinfo->local_alias));
1329                 if(!r->local_alias)
1330                         return 0;
1331                 r->local_alias->rrset = regional_alloc_init(s->s.region,
1332                         qinfo->local_alias->rrset,
1333                         sizeof(*qinfo->local_alias->rrset));
1334                 if(!r->local_alias->rrset)
1335                         return 0;
1336                 dsrc = qinfo->local_alias->rrset->entry.data;
1337
1338                 /* In the current implementation, a local alias must be
1339                  * a single CNAME RR (see worker_handle_request()). */
1340                 log_assert(!qinfo->local_alias->next && dsrc->count == 1 &&
1341                         qinfo->local_alias->rrset->rk.type ==
1342                         htons(LDNS_RR_TYPE_CNAME));
1343                 /* Technically, we should make a local copy for the owner
1344                  * name of the RRset, but in the case of the first (and
1345                  * currently only) local alias RRset, the owner name should
1346                  * point to the qname of the corresponding query, which should
1347                  * be valid throughout the lifetime of this mesh_reply.  So
1348                  * we can skip copying. */
1349                 log_assert(qinfo->local_alias->rrset->rk.dname ==
1350                         sldns_buffer_at(rep->c->buffer, LDNS_HEADER_SIZE));
1351
1352                 /* the rrset is not packed, like in the cache, but it is
1353                  * individualy allocated with an allocator from localzone. */
1354                 d = regional_alloc_zero(s->s.region, sizeof(*d));
1355                 if(!d)
1356                         return 0;
1357                 r->local_alias->rrset->entry.data = d;
1358                 if(!rrset_insert_rr(s->s.region, d, dsrc->rr_data[0],
1359                         dsrc->rr_len[0], dsrc->rr_ttl[0], "CNAME local alias"))
1360                         return 0;
1361         } else
1362                 r->local_alias = NULL;
1363
1364         s->reply_list = r;
1365         return 1;
1366 }
1367
1368 /* Extract the query info and flags from 'mstate' into '*qinfop' and '*qflags'.
1369  * Since this is only used for internal refetch of otherwise-expired answer,
1370  * we simply ignore the rare failure mode when memory allocation fails. */
1371 static void
1372 mesh_copy_qinfo(struct mesh_state* mstate, struct query_info** qinfop,
1373         uint16_t* qflags)
1374 {
1375         struct regional* region = mstate->s.env->scratch;
1376         struct query_info* qinfo;
1377
1378         qinfo = regional_alloc_init(region, &mstate->s.qinfo, sizeof(*qinfo));
1379         if(!qinfo)
1380                 return;
1381         qinfo->qname = regional_alloc_init(region, qinfo->qname,
1382                 qinfo->qname_len);
1383         if(!qinfo->qname)
1384                 return;
1385         *qinfop = qinfo;
1386         *qflags = mstate->s.query_flags;
1387 }
1388
1389 /**
1390  * Continue processing the mesh state at another module.
1391  * Handles module to modules transfer of control.
1392  * Handles module finished.
1393  * @param mesh: the mesh area.
1394  * @param mstate: currently active mesh state.
1395  *      Deleted if finished, calls _done and _supers to
1396  *      send replies to clients and inform other mesh states.
1397  *      This in turn may create additional runnable mesh states.
1398  * @param s: state at which the current module exited.
1399  * @param ev: the event sent to the module.
1400  *      returned is the event to send to the next module.
1401  * @return true if continue processing at the new module.
1402  *      false if not continued processing is needed.
1403  */
1404 static int
1405 mesh_continue(struct mesh_area* mesh, struct mesh_state* mstate,
1406         enum module_ext_state s, enum module_ev* ev)
1407 {
1408         mstate->num_activated++;
1409         if(mstate->num_activated > MESH_MAX_ACTIVATION) {
1410                 /* module is looping. Stop it. */
1411                 log_err("internal error: looping module (%s) stopped",
1412                         mesh->mods.mod[mstate->s.curmod]->name);
1413                 log_query_info(0, "pass error for qstate",
1414                         &mstate->s.qinfo);
1415                 s = module_error;
1416         }
1417         if(s == module_wait_module || s == module_restart_next) {
1418                 /* start next module */
1419                 mstate->s.curmod++;
1420                 if(mesh->mods.num == mstate->s.curmod) {
1421                         log_err("Cannot pass to next module; at last module");
1422                         log_query_info(VERB_QUERY, "pass error for qstate",
1423                                 &mstate->s.qinfo);
1424                         mstate->s.curmod--;
1425                         return mesh_continue(mesh, mstate, module_error, ev);
1426                 }
1427                 if(s == module_restart_next) {
1428                         int curmod = mstate->s.curmod;
1429                         for(; mstate->s.curmod < mesh->mods.num;
1430                                 mstate->s.curmod++) {
1431                                 fptr_ok(fptr_whitelist_mod_clear(
1432                                         mesh->mods.mod[mstate->s.curmod]->clear));
1433                                 (*mesh->mods.mod[mstate->s.curmod]->clear)
1434                                         (&mstate->s, mstate->s.curmod);
1435                                 mstate->s.minfo[mstate->s.curmod] = NULL;
1436                         }
1437                         mstate->s.curmod = curmod;
1438                 }
1439                 *ev = module_event_pass;
1440                 return 1;
1441         }
1442         if(s == module_wait_subquery && mstate->sub_set.count == 0) {
1443                 log_err("module cannot wait for subquery, subquery list empty");
1444                 log_query_info(VERB_QUERY, "pass error for qstate",
1445                         &mstate->s.qinfo);
1446                 s = module_error;
1447         }
1448         if(s == module_error && mstate->s.return_rcode == LDNS_RCODE_NOERROR) {
1449                 /* error is bad, handle pass back up below */
1450                 mstate->s.return_rcode = LDNS_RCODE_SERVFAIL;
1451         }
1452         if(s == module_error) {
1453                 mesh_query_done(mstate);
1454                 mesh_walk_supers(mesh, mstate);
1455                 mesh_state_delete(&mstate->s);
1456                 return 0;
1457         }
1458         if(s == module_finished) {
1459                 if(mstate->s.curmod == 0) {
1460                         struct query_info* qinfo = NULL;
1461                         uint16_t qflags;
1462
1463                         mesh_query_done(mstate);
1464                         mesh_walk_supers(mesh, mstate);
1465
1466                         /* If the answer to the query needs to be refetched
1467                          * from an external DNS server, we'll need to schedule
1468                          * a prefetch after removing the current state, so
1469                          * we need to make a copy of the query info here. */
1470                         if(mstate->s.need_refetch)
1471                                 mesh_copy_qinfo(mstate, &qinfo, &qflags);
1472
1473                         mesh_state_delete(&mstate->s);
1474                         if(qinfo) {
1475                                 mesh_schedule_prefetch(mesh, qinfo, qflags,
1476                                         0, 1);
1477                         }
1478                         return 0;
1479                 }
1480                 /* pass along the locus of control */
1481                 mstate->s.curmod --;
1482                 *ev = module_event_moddone;
1483                 return 1;
1484         }
1485         return 0;
1486 }
1487
1488 void mesh_run(struct mesh_area* mesh, struct mesh_state* mstate,
1489         enum module_ev ev, struct outbound_entry* e)
1490 {
1491         enum module_ext_state s;
1492         verbose(VERB_ALGO, "mesh_run: start");
1493         while(mstate) {
1494                 /* run the module */
1495                 fptr_ok(fptr_whitelist_mod_operate(
1496                         mesh->mods.mod[mstate->s.curmod]->operate));
1497                 (*mesh->mods.mod[mstate->s.curmod]->operate)
1498                         (&mstate->s, ev, mstate->s.curmod, e);
1499
1500                 /* examine results */
1501                 mstate->s.reply = NULL;
1502                 regional_free_all(mstate->s.env->scratch);
1503                 s = mstate->s.ext_state[mstate->s.curmod];
1504                 verbose(VERB_ALGO, "mesh_run: %s module exit state is %s",
1505                         mesh->mods.mod[mstate->s.curmod]->name, strextstate(s));
1506                 e = NULL;
1507                 if(mesh_continue(mesh, mstate, s, &ev))
1508                         continue;
1509
1510                 /* run more modules */
1511                 ev = module_event_pass;
1512                 if(mesh->run.count > 0) {
1513                         /* pop random element off the runnable tree */
1514                         mstate = (struct mesh_state*)mesh->run.root->key;
1515                         (void)rbtree_delete(&mesh->run, mstate);
1516                 } else mstate = NULL;
1517         }
1518         if(verbosity >= VERB_ALGO) {
1519                 mesh_stats(mesh, "mesh_run: end");
1520                 mesh_log_list(mesh);
1521         }
1522 }
1523
1524 void
1525 mesh_log_list(struct mesh_area* mesh)
1526 {
1527         char buf[30];
1528         struct mesh_state* m;
1529         int num = 0;
1530         RBTREE_FOR(m, struct mesh_state*, &mesh->all) {
1531                 snprintf(buf, sizeof(buf), "%d%s%s%s%s%s%s mod%d %s%s",
1532                         num++, (m->s.is_priming)?"p":"",  /* prime */
1533                         (m->s.is_valrec)?"v":"",  /* prime */
1534                         (m->s.query_flags&BIT_RD)?"RD":"",
1535                         (m->s.query_flags&BIT_CD)?"CD":"",
1536                         (m->super_set.count==0)?"d":"", /* detached */
1537                         (m->sub_set.count!=0)?"c":"",  /* children */
1538                         m->s.curmod, (m->reply_list)?"rep":"", /*hasreply*/
1539                         (m->cb_list)?"cb":"" /* callbacks */
1540                         );
1541                 log_query_info(VERB_ALGO, buf, &m->s.qinfo);
1542         }
1543 }
1544
1545 void
1546 mesh_stats(struct mesh_area* mesh, const char* str)
1547 {
1548         verbose(VERB_DETAIL, "%s %u recursion states (%u with reply, "
1549                 "%u detached), %u waiting replies, %u recursion replies "
1550                 "sent, %d replies dropped, %d states jostled out",
1551                 str, (unsigned)mesh->all.count,
1552                 (unsigned)mesh->num_reply_states,
1553                 (unsigned)mesh->num_detached_states,
1554                 (unsigned)mesh->num_reply_addrs,
1555                 (unsigned)mesh->replies_sent,
1556                 (unsigned)mesh->stats_dropped,
1557                 (unsigned)mesh->stats_jostled);
1558         if(mesh->replies_sent > 0) {
1559                 struct timeval avg;
1560                 timeval_divide(&avg, &mesh->replies_sum_wait,
1561                         mesh->replies_sent);
1562                 log_info("average recursion processing time "
1563                         ARG_LL "d.%6.6d sec",
1564                         (long long)avg.tv_sec, (int)avg.tv_usec);
1565                 log_info("histogram of recursion processing times");
1566                 timehist_log(mesh->histogram, "recursions");
1567         }
1568 }
1569
1570 void
1571 mesh_stats_clear(struct mesh_area* mesh)
1572 {
1573         if(!mesh)
1574                 return;
1575         mesh->replies_sent = 0;
1576         mesh->replies_sum_wait.tv_sec = 0;
1577         mesh->replies_sum_wait.tv_usec = 0;
1578         mesh->stats_jostled = 0;
1579         mesh->stats_dropped = 0;
1580         timehist_clear(mesh->histogram);
1581         mesh->ans_secure = 0;
1582         mesh->ans_bogus = 0;
1583         memset(&mesh->ans_rcode[0], 0, sizeof(size_t)*16);
1584         mesh->ans_nodata = 0;
1585 }
1586
1587 size_t
1588 mesh_get_mem(struct mesh_area* mesh)
1589 {
1590         struct mesh_state* m;
1591         size_t s = sizeof(*mesh) + sizeof(struct timehist) +
1592                 sizeof(struct th_buck)*mesh->histogram->num +
1593                 sizeof(sldns_buffer) + sldns_buffer_capacity(mesh->qbuf_bak);
1594         RBTREE_FOR(m, struct mesh_state*, &mesh->all) {
1595                 /* all, including m itself allocated in qstate region */
1596                 s += regional_get_mem(m->s.region);
1597         }
1598         return s;
1599 }
1600
1601 int
1602 mesh_detect_cycle(struct module_qstate* qstate, struct query_info* qinfo,
1603         uint16_t flags, int prime, int valrec)
1604 {
1605         struct mesh_area* mesh = qstate->env->mesh;
1606         struct mesh_state* dep_m = NULL;
1607         if(!mesh_state_is_unique(qstate->mesh_info))
1608                 dep_m = mesh_area_find(mesh, NULL, qinfo, flags, prime, valrec);
1609         return mesh_detect_cycle_found(qstate, dep_m);
1610 }
1611
1612 void mesh_list_insert(struct mesh_state* m, struct mesh_state** fp,
1613         struct mesh_state** lp)
1614 {
1615         /* insert as last element */
1616         m->prev = *lp;
1617         m->next = NULL;
1618         if(*lp)
1619                 (*lp)->next = m;
1620         else    *fp = m;
1621         *lp = m;
1622 }
1623
1624 void mesh_list_remove(struct mesh_state* m, struct mesh_state** fp,
1625         struct mesh_state** lp)
1626 {
1627         if(m->next)
1628                 m->next->prev = m->prev;
1629         else    *lp = m->prev;
1630         if(m->prev)
1631                 m->prev->next = m->next;
1632         else    *fp = m->next;
1633 }
1634
1635 void mesh_state_remove_reply(struct mesh_area* mesh, struct mesh_state* m,
1636         struct comm_point* cp)
1637 {
1638         struct mesh_reply* n, *prev = NULL;
1639         n = m->reply_list;
1640         /* when in mesh_cleanup, it sets the reply_list to NULL, so that
1641          * there is no accounting twice */
1642         if(!n) return; /* nothing to remove, also no accounting needed */
1643         while(n) {
1644                 if(n->query_reply.c == cp) {
1645                         /* unlink it */
1646                         if(prev) prev->next = n->next;
1647                         else m->reply_list = n->next;
1648                         /* delete it, but allocated in m region */
1649                         mesh->num_reply_addrs--;
1650
1651                         /* prev = prev; */
1652                         n = n->next;
1653                         continue;
1654                 }
1655                 prev = n;
1656                 n = n->next;
1657         }
1658         /* it was not detached (because it had a reply list), could be now */
1659         if(!m->reply_list && !m->cb_list
1660                 && m->super_set.count == 0) {
1661                 mesh->num_detached_states++;
1662         }
1663         /* if not replies any more in mstate, it is no longer a reply_state */
1664         if(!m->reply_list && !m->cb_list) {
1665                 log_assert(mesh->num_reply_states > 0);
1666                 mesh->num_reply_states--;
1667         }
1668 }