daemon/worker.c

   1 /*
   2  * daemon/worker.c - worker that handles a pending list of requests.
   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 implements the worker that handles callbacks on events, for
  40  * pending requests.
  41  */
  42 #include "config.h"
  43 #include "util/log.h"
  44 #include "util/net_help.h"
  45 #include "util/random.h"
  46 #include "daemon/worker.h"
  47 #include "daemon/daemon.h"
  48 #include "daemon/remote.h"
  49 #include "daemon/acl_list.h"
  50 #include "util/netevent.h"
  51 #include "util/config_file.h"
  52 #include "util/module.h"
  53 #include "util/regional.h"
  54 #include "util/storage/slabhash.h"
  55 #include "services/listen_dnsport.h"
  56 #include "services/outside_network.h"
  57 #include "services/outbound_list.h"
  58 #include "services/cache/rrset.h"
  59 #include "services/cache/infra.h"
  60 #include "services/cache/dns.h"
  61 #include "services/mesh.h"
  62 #include "services/localzone.h"
  63 #include "util/data/msgparse.h"
  64 #include "util/data/msgencode.h"
  65 #include "util/data/dname.h"
  66 #include "util/fptr_wlist.h"
  67 #include "util/tube.h"
  68 #include "iterator/iter_fwd.h"
  69 #include "iterator/iter_hints.h"
  70 #include "validator/autotrust.h"
  71 #include "validator/val_anchor.h"
  72 #include "ldns/sbuffer.h"
  73
  74 #ifdef HAVE_SYS_TYPES_H
  75 #  include <sys/types.h>
  76 #endif
  77 #ifdef HAVE_NETDB_H
  78 #include <netdb.h>
  79 #endif
  80 #include <signal.h>
  81 #ifdef UB_ON_WINDOWS
  82 #include "winrc/win_svc.h"
  83 #endif
  84
  85 /** Size of an UDP datagram */
  86 #define NORMAL_UDP_SIZE 512 /* bytes */
  87
  88 /**
  89  * seconds to add to prefetch leeway.  This is a TTL that expires old rrsets
  90  * earlier than they should in order to put the new update into the cache.
  91  * This additional value is to make sure that if not all TTLs are equal in
  92  * the message to be updated(and replaced), that rrsets with up to this much
  93  * extra TTL are also replaced.  This means that the resulting new message
  94  * will have (most likely) this TTL at least, avoiding very small 'split
  95  * second' TTLs due to operators choosing relative primes for TTLs (or so).
  96  * Also has to be at least one to break ties (and overwrite cached entry).
  97  */
  98 #define PREFETCH_EXPIRY_ADD 60
  99
 100 #ifdef UNBOUND_ALLOC_STATS
 101 /** measure memory leakage */
 102 static void
 103 debug_memleak(size_t accounted, size_t heap,
 104         size_t total_alloc, size_t total_free)
 105 {
 106         static int init = 0;
 107         static size_t base_heap, base_accounted, base_alloc, base_free;
 108         size_t base_af, cur_af, grow_af, grow_acc;
 109         if(!init) {
 110                 init = 1;
 111                 base_heap = heap;
 112                 base_accounted = accounted;
 113                 base_alloc = total_alloc;
 114                 base_free = total_free;
 115         }
 116         base_af = base_alloc - base_free;
 117         cur_af = total_alloc - total_free;
 118         grow_af = cur_af - base_af;
 119         grow_acc = accounted - base_accounted;
 120         log_info("Leakage: %d leaked. growth: %u use, %u acc, %u heap",
 121                 (int)(grow_af - grow_acc), (unsigned)grow_af,
 122                 (unsigned)grow_acc, (unsigned)(heap - base_heap));
 123 }
 124
 125 /** give debug heap size indication */
 126 static void
 127 debug_total_mem(size_t calctotal)
 128 {
 129 #ifdef HAVE_SBRK
 130         extern void* unbound_start_brk;
 131         extern size_t unbound_mem_alloc, unbound_mem_freed;
 132         void* cur = sbrk(0);
 133         int total = cur-unbound_start_brk;
 134         log_info("Total heap memory estimate: %u  total-alloc: %u  "
 135                 "total-free: %u", (unsigned)total,
 136                 (unsigned)unbound_mem_alloc, (unsigned)unbound_mem_freed);
 137         debug_memleak(calctotal, (size_t)total,
 138                 unbound_mem_alloc, unbound_mem_freed);
 139 #else
 140         (void)calctotal;
 141 #endif /* HAVE_SBRK */
 142 }
 143 #endif /* UNBOUND_ALLOC_STATS */
 144
 145 /** Report on memory usage by this thread and global */
 146 static void
 147 worker_mem_report(struct worker* ATTR_UNUSED(worker),
 148         struct serviced_query* ATTR_UNUSED(cur_serv))
 149 {
 150 #ifdef UNBOUND_ALLOC_STATS
 151         /* debug func in validator module */
 152         size_t total, front, back, mesh, msg, rrset, infra, ac, superac;
 153         size_t me, iter, val, anch;
 154         int i;
 155         if(verbosity < VERB_ALGO)
 156                 return;
 157         front = listen_get_mem(worker->front);
 158         back = outnet_get_mem(worker->back);
 159         msg = slabhash_get_mem(worker->env.msg_cache);
 160         rrset = slabhash_get_mem(&worker->env.rrset_cache->table);
 161         infra = infra_get_mem(worker->env.infra_cache);
 162         mesh = mesh_get_mem(worker->env.mesh);
 163         ac = alloc_get_mem(&worker->alloc);
 164         superac = alloc_get_mem(&worker->daemon->superalloc);
 165         anch = anchors_get_mem(worker->env.anchors);
 166         iter = 0;
 167         val = 0;
 168         for(i=0; i<worker->env.mesh->mods.num; i++) {
 169                 fptr_ok(fptr_whitelist_mod_get_mem(worker->env.mesh->
 170                         mods.mod[i]->get_mem));
 171                 if(strcmp(worker->env.mesh->mods.mod[i]->name, "validator")==0)
 172                         val += (*worker->env.mesh->mods.mod[i]->get_mem)
 173                                 (&worker->env, i);
 174                 else    iter += (*worker->env.mesh->mods.mod[i]->get_mem)
 175                                 (&worker->env, i);
 176         }
 177         me = sizeof(*worker) + sizeof(*worker->base) + sizeof(*worker->comsig)
 178                 + comm_point_get_mem(worker->cmd_com)
 179                 + sizeof(worker->rndstate)
 180                 + regional_get_mem(worker->scratchpad)
 181                 + sizeof(*worker->env.scratch_buffer)
 182                 + sldns_buffer_capacity(worker->env.scratch_buffer)
 183                 + forwards_get_mem(worker->env.fwds)
 184                 + hints_get_mem(worker->env.hints);
 185         if(worker->thread_num == 0)
 186                 me += acl_list_get_mem(worker->daemon->acl);
 187         if(cur_serv) {
 188                 me += serviced_get_mem(cur_serv);
 189         }
 190         total = front+back+mesh+msg+rrset+infra+iter+val+ac+superac+me;
 191         log_info("Memory conditions: %u front=%u back=%u mesh=%u msg=%u "
 192                 "rrset=%u infra=%u iter=%u val=%u anchors=%u "
 193                 "alloccache=%u globalalloccache=%u me=%u",
 194                 (unsigned)total, (unsigned)front, (unsigned)back,
 195                 (unsigned)mesh, (unsigned)msg, (unsigned)rrset,
 196                 (unsigned)infra, (unsigned)iter, (unsigned)val, (unsigned)anch,
 197                 (unsigned)ac, (unsigned)superac, (unsigned)me);
 198         debug_total_mem(total);
 199 #else /* no UNBOUND_ALLOC_STATS */
 200         size_t val = 0;
 201         int i;
 202         if(verbosity < VERB_QUERY)
 203                 return;
 204         for(i=0; i<worker->env.mesh->mods.num; i++) {
 205                 fptr_ok(fptr_whitelist_mod_get_mem(worker->env.mesh->
 206                         mods.mod[i]->get_mem));
 207                 if(strcmp(worker->env.mesh->mods.mod[i]->name, "validator")==0)
 208                         val += (*worker->env.mesh->mods.mod[i]->get_mem)
 209                                 (&worker->env, i);
 210         }
 211         verbose(VERB_QUERY, "cache memory msg=%u rrset=%u infra=%u val=%u",
 212                 (unsigned)slabhash_get_mem(worker->env.msg_cache),
 213                 (unsigned)slabhash_get_mem(&worker->env.rrset_cache->table),
 214                 (unsigned)infra_get_mem(worker->env.infra_cache),
 215                 (unsigned)val);
 216 #endif /* UNBOUND_ALLOC_STATS */
 217 }
 218
 219 void
 220 worker_send_cmd(struct worker* worker, enum worker_commands cmd)
 221 {
 222         uint32_t c = (uint32_t)htonl(cmd);
 223         if(!tube_write_msg(worker->cmd, (uint8_t*)&c, sizeof(c), 0)) {
 224                 log_err("worker send cmd %d failed", (int)cmd);
 225         }
 226 }
 227
 228 int
 229 worker_handle_reply(struct comm_point* c, void* arg, int error,
 230         struct comm_reply* reply_info)
 231 {
 232         struct module_qstate* q = (struct module_qstate*)arg;
 233         struct worker* worker = q->env->worker;
 234         struct outbound_entry e;
 235         e.qstate = q;
 236         e.qsent = NULL;
 237
 238         if(error != 0) {
 239                 mesh_report_reply(worker->env.mesh, &e, reply_info, error);
 240                 worker_mem_report(worker, NULL);
 241                 return 0;
 242         }
 243         /* sanity check. */
 244         if(!LDNS_QR_WIRE(sldns_buffer_begin(c->buffer))
 245                 || LDNS_OPCODE_WIRE(sldns_buffer_begin(c->buffer)) !=
 246                         LDNS_PACKET_QUERY
 247                 || LDNS_QDCOUNT(sldns_buffer_begin(c->buffer)) > 1) {
 248                 /* error becomes timeout for the module as if this reply
 249                  * never arrived. */
 250                 mesh_report_reply(worker->env.mesh, &e, reply_info,
 251                         NETEVENT_TIMEOUT);
 252                 worker_mem_report(worker, NULL);
 253                 return 0;
 254         }
 255         mesh_report_reply(worker->env.mesh, &e, reply_info, NETEVENT_NOERROR);
 256         worker_mem_report(worker, NULL);
 257         return 0;
 258 }
 259
 260 int
 261 worker_handle_service_reply(struct comm_point* c, void* arg, int error,
 262         struct comm_reply* reply_info)
 263 {
 264         struct outbound_entry* e = (struct outbound_entry*)arg;
 265         struct worker* worker = e->qstate->env->worker;
 266         struct serviced_query *sq = e->qsent;
 267
 268         verbose(VERB_ALGO, "worker svcd callback for qstate %p", e->qstate);
 269         if(error != 0) {
 270                 mesh_report_reply(worker->env.mesh, e, reply_info, error);
 271                 worker_mem_report(worker, sq);
 272                 return 0;
 273         }
 274         /* sanity check. */
 275         if(!LDNS_QR_WIRE(sldns_buffer_begin(c->buffer))
 276                 || LDNS_OPCODE_WIRE(sldns_buffer_begin(c->buffer)) !=
 277                         LDNS_PACKET_QUERY
 278                 || LDNS_QDCOUNT(sldns_buffer_begin(c->buffer)) > 1) {
 279                 /* error becomes timeout for the module as if this reply
 280                  * never arrived. */
 281                 verbose(VERB_ALGO, "worker: bad reply handled as timeout");
 282                 mesh_report_reply(worker->env.mesh, e, reply_info,
 283                         NETEVENT_TIMEOUT);
 284                 worker_mem_report(worker, sq);
 285                 return 0;
 286         }
 287         mesh_report_reply(worker->env.mesh, e, reply_info, NETEVENT_NOERROR);
 288         worker_mem_report(worker, sq);
 289         return 0;
 290 }
 291
 292 /** check request sanity.
 293  * @param pkt: the wire packet to examine for sanity.
 294  * @param worker: parameters for checking.
 295  * @return error code, 0 OK, or -1 discard.
 296 */
 297 static int
 298 worker_check_request(sldns_buffer* pkt, struct worker* worker)
 299 {
 300         if(sldns_buffer_limit(pkt) < LDNS_HEADER_SIZE) {
 301                 verbose(VERB_QUERY, "request too short, discarded");
 302                 return -1;
 303         }
 304         if(sldns_buffer_limit(pkt) > NORMAL_UDP_SIZE &&
 305                 worker->daemon->cfg->harden_large_queries) {
 306                 verbose(VERB_QUERY, "request too large, discarded");
 307                 return -1;
 308         }
 309         if(LDNS_QR_WIRE(sldns_buffer_begin(pkt))) {
 310                 verbose(VERB_QUERY, "request has QR bit on, discarded");
 311                 return -1;
 312         }
 313         if(LDNS_TC_WIRE(sldns_buffer_begin(pkt))) {
 314                 LDNS_TC_CLR(sldns_buffer_begin(pkt));
 315                 verbose(VERB_QUERY, "request bad, has TC bit on");
 316                 return LDNS_RCODE_FORMERR;
 317         }
 318         if(LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt)) != LDNS_PACKET_QUERY) {
 319                 verbose(VERB_QUERY, "request unknown opcode %d",
 320                         LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt)));
 321                 return LDNS_RCODE_NOTIMPL;
 322         }
 323         if(LDNS_QDCOUNT(sldns_buffer_begin(pkt)) != 1) {
 324                 verbose(VERB_QUERY, "request wrong nr qd=%d",
 325                         LDNS_QDCOUNT(sldns_buffer_begin(pkt)));
 326                 return LDNS_RCODE_FORMERR;
 327         }
 328         if(LDNS_ANCOUNT(sldns_buffer_begin(pkt)) != 0) {
 329                 verbose(VERB_QUERY, "request wrong nr an=%d",
 330                         LDNS_ANCOUNT(sldns_buffer_begin(pkt)));
 331                 return LDNS_RCODE_FORMERR;
 332         }
 333         if(LDNS_NSCOUNT(sldns_buffer_begin(pkt)) != 0) {
 334                 verbose(VERB_QUERY, "request wrong nr ns=%d",
 335                         LDNS_NSCOUNT(sldns_buffer_begin(pkt)));
 336                 return LDNS_RCODE_FORMERR;
 337         }
 338         if(LDNS_ARCOUNT(sldns_buffer_begin(pkt)) > 1) {
 339                 verbose(VERB_QUERY, "request wrong nr ar=%d",
 340                         LDNS_ARCOUNT(sldns_buffer_begin(pkt)));
 341                 return LDNS_RCODE_FORMERR;
 342         }
 343         return 0;
 344 }
 345
 346 void
 347 worker_handle_control_cmd(struct tube* ATTR_UNUSED(tube), uint8_t* msg,
 348         size_t len, int error, void* arg)
 349 {
 350         struct worker* worker = (struct worker*)arg;
 351         enum worker_commands cmd;
 352         if(error != NETEVENT_NOERROR) {
 353                 free(msg);
 354                 if(error == NETEVENT_CLOSED)
 355                         comm_base_exit(worker->base);
 356                 else    log_info("control event: %d", error);
 357                 return;
 358         }
 359         if(len != sizeof(uint32_t)) {
 360                 fatal_exit("bad control msg length %d", (int)len);
 361         }
 362         cmd = sldns_read_uint32(msg);
 363         free(msg);
 364         switch(cmd) {
 365         case worker_cmd_quit:
 366                 verbose(VERB_ALGO, "got control cmd quit");
 367                 comm_base_exit(worker->base);
 368                 break;
 369         case worker_cmd_stats:
 370                 verbose(VERB_ALGO, "got control cmd stats");
 371                 server_stats_reply(worker, 1);
 372                 break;
 373         case worker_cmd_stats_noreset:
 374                 verbose(VERB_ALGO, "got control cmd stats_noreset");
 375                 server_stats_reply(worker, 0);
 376                 break;
 377         case worker_cmd_remote:
 378                 verbose(VERB_ALGO, "got control cmd remote");
 379                 daemon_remote_exec(worker);
 380                 break;
 381         default:
 382                 log_err("bad command %d", (int)cmd);
 383                 break;
 384         }
 385 }
 386
 387 /** check if a delegation is secure */
 388 static enum sec_status
 389 check_delegation_secure(struct reply_info *rep)
 390 {
 391         /* return smallest security status */
 392         size_t i;
 393         enum sec_status sec = sec_status_secure;
 394         enum sec_status s;
 395         size_t num = rep->an_numrrsets + rep->ns_numrrsets;
 396         /* check if answer and authority are OK */
 397         for(i=0; i<num; i++) {
 398                 s = ((struct packed_rrset_data*)rep->rrsets[i]->entry.data)
 399                         ->security;
 400                 if(s < sec)
 401                         sec = s;
 402         }
 403         /* in additional, only unchecked triggers revalidation */
 404         for(i=num; i<rep->rrset_count; i++) {
 405                 s = ((struct packed_rrset_data*)rep->rrsets[i]->entry.data)
 406                         ->security;
 407                 if(s == sec_status_unchecked)
 408                         return s;
 409         }
 410         return sec;
 411 }
 412
 413 /** remove nonsecure from a delegation referral additional section */
 414 static void
 415 deleg_remove_nonsecure_additional(struct reply_info* rep)
 416 {
 417         /* we can simply edit it, since we are working in the scratch region */
 418         size_t i;
 419         enum sec_status s;
 420
 421         for(i = rep->an_numrrsets+rep->ns_numrrsets; i<rep->rrset_count; i++) {
 422                 s = ((struct packed_rrset_data*)rep->rrsets[i]->entry.data)
 423                         ->security;
 424                 if(s != sec_status_secure) {
 425                         memmove(rep->rrsets+i, rep->rrsets+i+1,
 426                                 sizeof(struct ub_packed_rrset_key*)*
 427                                 (rep->rrset_count - i - 1));
 428                         rep->ar_numrrsets--;
 429                         rep->rrset_count--;
 430                         i--;
 431                 }
 432         }
 433 }
 434
 435 /** answer nonrecursive query from the cache */
 436 static int
 437 answer_norec_from_cache(struct worker* worker, struct query_info* qinfo,
 438         uint16_t id, uint16_t flags, struct comm_reply* repinfo,
 439         struct edns_data* edns)
 440 {
 441         /* for a nonrecursive query return either:
 442          *      o an error (servfail; we try to avoid this)
 443          *      o a delegation (closest we have; this routine tries that)
 444          *      o the answer (checked by answer_from_cache)
 445          *
 446          * So, grab a delegation from the rrset cache.
 447          * Then check if it needs validation, if so, this routine fails,
 448          * so that iterator can prime and validator can verify rrsets.
 449          */
 450         uint16_t udpsize = edns->udp_size;
 451         int secure = 0;
 452         time_t timenow = *worker->env.now;
 453         int must_validate = (!(flags&BIT_CD) || worker->env.cfg->ignore_cd)
 454                 && worker->env.need_to_validate;
 455         struct dns_msg *msg = NULL;
 456         struct delegpt *dp;
 457
 458         dp = dns_cache_find_delegation(&worker->env, qinfo->qname,
 459                 qinfo->qname_len, qinfo->qtype, qinfo->qclass,
 460                 worker->scratchpad, &msg, timenow);
 461         if(!dp) { /* no delegation, need to reprime */
 462                 regional_free_all(worker->scratchpad);
 463                 return 0;
 464         }
 465         if(must_validate) {
 466                 switch(check_delegation_secure(msg->rep)) {
 467                 case sec_status_unchecked:
 468                         /* some rrsets have not been verified yet, go and
 469                          * let validator do that */
 470                         regional_free_all(worker->scratchpad);
 471                         return 0;
 472                 case sec_status_bogus:
 473                         /* some rrsets are bogus, reply servfail */
 474                         edns->edns_version = EDNS_ADVERTISED_VERSION;
 475                         edns->udp_size = EDNS_ADVERTISED_SIZE;
 476                         edns->ext_rcode = 0;
 477                         edns->bits &= EDNS_DO;
 478                         error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL,
 479                                 &msg->qinfo, id, flags, edns);
 480                         regional_free_all(worker->scratchpad);
 481                         if(worker->stats.extended) {
 482                                 worker->stats.ans_bogus++;
 483                                 worker->stats.ans_rcode[LDNS_RCODE_SERVFAIL]++;
 484                         }
 485                         return 1;
 486                 case sec_status_secure:
 487                         /* all rrsets are secure */
 488                         /* remove non-secure rrsets from the add. section*/
 489                         if(worker->env.cfg->val_clean_additional)
 490                                 deleg_remove_nonsecure_additional(msg->rep);
 491                         secure = 1;
 492                         break;
 493                 case sec_status_indeterminate:
 494                 case sec_status_insecure:
 495                 default:
 496                         /* not secure */
 497                         secure = 0;
 498                         break;
 499                 }
 500         }
 501         /* return this delegation from the cache */
 502         edns->edns_version = EDNS_ADVERTISED_VERSION;
 503         edns->udp_size = EDNS_ADVERTISED_SIZE;
 504         edns->ext_rcode = 0;
 505         edns->bits &= EDNS_DO;
 506         msg->rep->flags |= BIT_QR|BIT_RA;
 507         if(!reply_info_answer_encode(&msg->qinfo, msg->rep, id, flags,
 508                 repinfo->c->buffer, 0, 1, worker->scratchpad,
 509                 udpsize, edns, (int)(edns->bits & EDNS_DO), secure)) {
 510                 error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL,
 511                         &msg->qinfo, id, flags, edns);
 512         }
 513         regional_free_all(worker->scratchpad);
 514         if(worker->stats.extended) {
 515                 if(secure) worker->stats.ans_secure++;
 516                 server_stats_insrcode(&worker->stats, repinfo->c->buffer);
 517         }
 518         return 1;
 519 }
 520
 521 /** answer query from the cache */
 522 static int
 523 answer_from_cache(struct worker* worker, struct query_info* qinfo,
 524         struct reply_info* rep, uint16_t id, uint16_t flags,
 525         struct comm_reply* repinfo, struct edns_data* edns)
 526 {
 527         time_t timenow = *worker->env.now;
 528         uint16_t udpsize = edns->udp_size;
 529         int secure;
 530         int must_validate = (!(flags&BIT_CD) || worker->env.cfg->ignore_cd)
 531                 && worker->env.need_to_validate;
 532         /* see if it is possible */
 533         if(rep->ttl < timenow) {
 534                 /* the rrsets may have been updated in the meantime.
 535                  * we will refetch the message format from the
 536                  * authoritative server
 537                  */
 538                 return 0;
 539         }
 540         if(!rrset_array_lock(rep->ref, rep->rrset_count, timenow))
 541                 return 0;
 542         /* locked and ids and ttls are OK. */
 543         /* check CNAME chain (if any) */
 544         if(rep->an_numrrsets > 0 && (rep->rrsets[0]->rk.type ==
 545                 htons(LDNS_RR_TYPE_CNAME) || rep->rrsets[0]->rk.type ==
 546                 htons(LDNS_RR_TYPE_DNAME))) {
 547                 if(!reply_check_cname_chain(rep)) {
 548                         /* cname chain invalid, redo iterator steps */
 549                         verbose(VERB_ALGO, "Cache reply: cname chain broken");
 550                 bail_out:
 551                         rrset_array_unlock_touch(worker->env.rrset_cache,
 552                                 worker->scratchpad, rep->ref, rep->rrset_count);
 553                         regional_free_all(worker->scratchpad);
 554                         return 0;
 555                 }
 556         }
 557         /* check security status of the cached answer */
 558         if( rep->security == sec_status_bogus && must_validate) {
 559                 /* BAD cached */
 560                 edns->edns_version = EDNS_ADVERTISED_VERSION;
 561                 edns->udp_size = EDNS_ADVERTISED_SIZE;
 562                 edns->ext_rcode = 0;
 563                 edns->bits &= EDNS_DO;
 564                 error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL,
 565                         qinfo, id, flags, edns);
 566                 rrset_array_unlock_touch(worker->env.rrset_cache,
 567                         worker->scratchpad, rep->ref, rep->rrset_count);
 568                 regional_free_all(worker->scratchpad);
 569                 if(worker->stats.extended) {
 570                         worker->stats.ans_bogus ++;
 571                         worker->stats.ans_rcode[LDNS_RCODE_SERVFAIL] ++;
 572                 }
 573                 return 1;
 574         } else if( rep->security == sec_status_unchecked && must_validate) {
 575                 verbose(VERB_ALGO, "Cache reply: unchecked entry needs "
 576                         "validation");
 577                 goto bail_out; /* need to validate cache entry first */
 578         } else if(rep->security == sec_status_secure) {
 579                 if(reply_all_rrsets_secure(rep))
 580                         secure = 1;
 581                 else    {
 582                         if(must_validate) {
 583                                 verbose(VERB_ALGO, "Cache reply: secure entry"
 584                                         " changed status");
 585                                 goto bail_out; /* rrset changed, re-verify */
 586                         }
 587                         secure = 0;
 588                 }
 589         } else  secure = 0;
 590
 591         edns->edns_version = EDNS_ADVERTISED_VERSION;
 592         edns->udp_size = EDNS_ADVERTISED_SIZE;
 593         edns->ext_rcode = 0;
 594         edns->bits &= EDNS_DO;
 595         if(!reply_info_answer_encode(qinfo, rep, id, flags,
 596                 repinfo->c->buffer, timenow, 1, worker->scratchpad,
 597                 udpsize, edns, (int)(edns->bits & EDNS_DO), secure)) {
 598                 error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL,
 599                         qinfo, id, flags, edns);
 600         }
 601         /* cannot send the reply right now, because blocking network syscall
 602          * is bad while holding locks. */
 603         rrset_array_unlock_touch(worker->env.rrset_cache, worker->scratchpad,
 604                 rep->ref, rep->rrset_count);
 605         regional_free_all(worker->scratchpad);
 606         if(worker->stats.extended) {
 607                 if(secure) worker->stats.ans_secure++;
 608                 server_stats_insrcode(&worker->stats, repinfo->c->buffer);
 609         }
 610         /* go and return this buffer to the client */
 611         return 1;
 612 }
 613
 614 /** Reply to client and perform prefetch to keep cache up to date */
 615 static void
 616 reply_and_prefetch(struct worker* worker, struct query_info* qinfo,
 617         uint16_t flags, struct comm_reply* repinfo, time_t leeway)
 618 {
 619         /* first send answer to client to keep its latency
 620          * as small as a cachereply */
 621         comm_point_send_reply(repinfo);
 622         server_stats_prefetch(&worker->stats, worker);
 623
 624         /* create the prefetch in the mesh as a normal lookup without
 625          * client addrs waiting, which has the cache blacklisted (to bypass
 626          * the cache and go to the network for the data). */
 627         /* this (potentially) runs the mesh for the new query */
 628         mesh_new_prefetch(worker->env.mesh, qinfo, flags, leeway +
 629                 PREFETCH_EXPIRY_ADD);
 630 }
 631
 632 /**
 633  * Fill CH class answer into buffer. Keeps query.
 634  * @param pkt: buffer
 635  * @param str: string to put into text record (<255).
 636  * @param edns: edns reply information.
 637  */
 638 static void
 639 chaos_replystr(sldns_buffer* pkt, const char* str, struct edns_data* edns)
 640 {
 641         size_t len = strlen(str);
 642         unsigned int rd = LDNS_RD_WIRE(sldns_buffer_begin(pkt));
 643         unsigned int cd = LDNS_CD_WIRE(sldns_buffer_begin(pkt));
 644         if(len>255) len=255; /* cap size of TXT record */
 645         sldns_buffer_clear(pkt);
 646         sldns_buffer_skip(pkt, (ssize_t)sizeof(uint16_t)); /* skip id */
 647         sldns_buffer_write_u16(pkt, (uint16_t)(BIT_QR|BIT_RA));
 648         if(rd) LDNS_RD_SET(sldns_buffer_begin(pkt));
 649         if(cd) LDNS_CD_SET(sldns_buffer_begin(pkt));
 650         sldns_buffer_write_u16(pkt, 1); /* qdcount */
 651         sldns_buffer_write_u16(pkt, 1); /* ancount */
 652         sldns_buffer_write_u16(pkt, 0); /* nscount */
 653         sldns_buffer_write_u16(pkt, 0); /* arcount */
 654         (void)query_dname_len(pkt); /* skip qname */
 655         sldns_buffer_skip(pkt, (ssize_t)sizeof(uint16_t)); /* skip qtype */
 656         sldns_buffer_skip(pkt, (ssize_t)sizeof(uint16_t)); /* skip qclass */
 657         sldns_buffer_write_u16(pkt, 0xc00c); /* compr ptr to query */
 658         sldns_buffer_write_u16(pkt, LDNS_RR_TYPE_TXT);
 659         sldns_buffer_write_u16(pkt, LDNS_RR_CLASS_CH);
 660         sldns_buffer_write_u32(pkt, 0); /* TTL */
 661         sldns_buffer_write_u16(pkt, sizeof(uint8_t) + len);
 662         sldns_buffer_write_u8(pkt, len);
 663         sldns_buffer_write(pkt, str, len);
 664         sldns_buffer_flip(pkt);
 665         edns->edns_version = EDNS_ADVERTISED_VERSION;
 666         edns->udp_size = EDNS_ADVERTISED_SIZE;
 667         edns->bits &= EDNS_DO;
 668         attach_edns_record(pkt, edns);
 669 }
 670
 671 /**
 672  * Answer CH class queries.
 673  * @param w: worker
 674  * @param qinfo: query info. Pointer into packet buffer.
 675  * @param edns: edns info from query.
 676  * @param pkt: packet buffer.
 677  * @return: true if a reply is to be sent.
 678  */
 679 static int
 680 answer_chaos(struct worker* w, struct query_info* qinfo,
 681         struct edns_data* edns, sldns_buffer* pkt)
 682 {
 683         struct config_file* cfg = w->env.cfg;
 684         if(qinfo->qtype != LDNS_RR_TYPE_ANY && qinfo->qtype != LDNS_RR_TYPE_TXT)
 685                 return 0;
 686         if(query_dname_compare(qinfo->qname,
 687                 (uint8_t*)"\002id\006server") == 0 ||
 688                 query_dname_compare(qinfo->qname,
 689                 (uint8_t*)"\010hostname\004bind") == 0)
 690         {
 691                 if(cfg->hide_identity)
 692                         return 0;
 693                 if(cfg->identity==NULL || cfg->identity[0]==0) {
 694                         char buf[MAXHOSTNAMELEN+1];
 695                         if (gethostname(buf, MAXHOSTNAMELEN) == 0) {
 696                                 buf[MAXHOSTNAMELEN] = 0;
 697                                 chaos_replystr(pkt, buf, edns);
 698                         } else  {
 699                                 log_err("gethostname: %s", strerror(errno));
 700                                 chaos_replystr(pkt, "no hostname", edns);
 701                         }
 702                 }
 703                 else    chaos_replystr(pkt, cfg->identity, edns);
 704                 return 1;
 705         }
 706         if(query_dname_compare(qinfo->qname,
 707                 (uint8_t*)"\007version\006server") == 0 ||
 708                 query_dname_compare(qinfo->qname,
 709                 (uint8_t*)"\007version\004bind") == 0)
 710         {
 711                 if(cfg->hide_version)
 712                         return 0;
 713                 if(cfg->version==NULL || cfg->version[0]==0)
 714                         chaos_replystr(pkt, PACKAGE_STRING, edns);
 715                 else    chaos_replystr(pkt, cfg->version, edns);
 716                 return 1;
 717         }
 718         return 0;
 719 }
 720
 721 int
 722 deny_refuse(struct comm_point* c, enum acl_access acl,
 723         enum acl_access deny, enum acl_access refuse,
 724         struct worker* worker, struct comm_reply* repinfo)
 725 {
 726         if(acl == deny) {
 727                 comm_point_drop_reply(repinfo);
 728                 if(worker->stats.extended)
 729                         worker->stats.unwanted_queries++;
 730                 return 0;
 731         } else if(acl == refuse) {
 732                 log_addr(VERB_ALGO, "refused query from",
 733                         &repinfo->addr, repinfo->addrlen);
 734                 log_buf(VERB_ALGO, "refuse", c->buffer);
 735                 if(worker->stats.extended)
 736                         worker->stats.unwanted_queries++;
 737                 if(worker_check_request(c->buffer, worker) == -1) {
 738                         comm_point_drop_reply(repinfo);
 739                         return 0; /* discard this */
 740                 }
 741                 sldns_buffer_set_limit(c->buffer, LDNS_HEADER_SIZE);
 742                 sldns_buffer_write_at(c->buffer, 4,
 743                         (uint8_t*)"\0\0\0\0\0\0\0\0", 8);
 744                 LDNS_QR_SET(sldns_buffer_begin(c->buffer));
 745                 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer),
 746                         LDNS_RCODE_REFUSED);
 747                 return 1;
 748         }
 749
 750         return -1;
 751 }
 752
 753 int
 754 deny_refuse_all(struct comm_point* c, enum acl_access acl,
 755         struct worker* worker, struct comm_reply* repinfo)
 756 {
 757         return deny_refuse(c, acl, acl_deny, acl_refuse, worker, repinfo);
 758 }
 759
 760 int
 761 deny_refuse_non_local(struct comm_point* c, enum acl_access acl,
 762         struct worker* worker, struct comm_reply* repinfo)
 763 {
 764         return deny_refuse(c, acl, acl_deny_non_local, acl_refuse_non_local, worker, repinfo);
 765 }
 766
 767 int
 768 worker_handle_request(struct comm_point* c, void* arg, int error,
 769         struct comm_reply* repinfo)
 770 {
 771         struct worker* worker = (struct worker*)arg;
 772         int ret;
 773         hashvalue_t h;
 774         struct lruhash_entry* e;
 775         struct query_info qinfo;
 776         struct edns_data edns;
 777         enum acl_access acl;
 778
 779         if(error != NETEVENT_NOERROR) {
 780                 /* some bad tcp query DNS formats give these error calls */
 781                 verbose(VERB_ALGO, "handle request called with err=%d", error);
 782                 return 0;
 783         }
 784         acl = acl_list_lookup(worker->daemon->acl, &repinfo->addr,
 785                 repinfo->addrlen);
 786         if((ret=deny_refuse_all(c, acl, worker, repinfo)) != -1)
 787         {
 788                 return ret;
 789         }
 790         if((ret=worker_check_request(c->buffer, worker)) != 0) {
 791                 verbose(VERB_ALGO, "worker check request: bad query.");
 792                 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen);
 793                 if(ret != -1) {
 794                         LDNS_QR_SET(sldns_buffer_begin(c->buffer));
 795                         LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), ret);
 796                         return 1;
 797                 }
 798                 comm_point_drop_reply(repinfo);
 799                 return 0;
 800         }
 801         worker->stats.num_queries++;
 802         /* see if query is in the cache */
 803         if(!query_info_parse(&qinfo, c->buffer)) {
 804                 verbose(VERB_ALGO, "worker parse request: formerror.");
 805                 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen);
 806                 sldns_buffer_rewind(c->buffer);
 807                 LDNS_QR_SET(sldns_buffer_begin(c->buffer));
 808                 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer),
 809                         LDNS_RCODE_FORMERR);
 810                 server_stats_insrcode(&worker->stats, c->buffer);
 811                 return 1;
 812         }
 813         if(worker->env.cfg->log_queries) {
 814                 char ip[128];
 815                 addr_to_str(&repinfo->addr, repinfo->addrlen, ip, sizeof(ip));
 816                 log_nametypeclass(0, ip, qinfo.qname, qinfo.qtype, qinfo.qclass);
 817         }
 818         if(qinfo.qtype == LDNS_RR_TYPE_AXFR ||
 819                 qinfo.qtype == LDNS_RR_TYPE_IXFR) {
 820                 verbose(VERB_ALGO, "worker request: refused zone transfer.");
 821                 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen);
 822                 sldns_buffer_rewind(c->buffer);
 823                 LDNS_QR_SET(sldns_buffer_begin(c->buffer));
 824                 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer),
 825                         LDNS_RCODE_REFUSED);
 826                 if(worker->stats.extended) {
 827                         worker->stats.qtype[qinfo.qtype]++;
 828                         server_stats_insrcode(&worker->stats, c->buffer);
 829                 }
 830                 return 1;
 831         }
 832         if((ret=parse_edns_from_pkt(c->buffer, &edns)) != 0) {
 833                 verbose(VERB_ALGO, "worker parse edns: formerror.");
 834                 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen);
 835                 sldns_buffer_rewind(c->buffer);
 836                 LDNS_QR_SET(sldns_buffer_begin(c->buffer));
 837                 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), ret);
 838                 server_stats_insrcode(&worker->stats, c->buffer);
 839                 return 1;
 840         }
 841         if(edns.edns_present && edns.edns_version != 0) {
 842                 edns.ext_rcode = (uint8_t)(EDNS_RCODE_BADVERS>>4);
 843                 edns.edns_version = EDNS_ADVERTISED_VERSION;
 844                 edns.udp_size = EDNS_ADVERTISED_SIZE;
 845                 edns.bits &= EDNS_DO;
 846                 verbose(VERB_ALGO, "query with bad edns version.");
 847                 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen);
 848                 error_encode(c->buffer, EDNS_RCODE_BADVERS&0xf, &qinfo,
 849                         *(uint16_t*)sldns_buffer_begin(c->buffer),
 850                         sldns_buffer_read_u16_at(c->buffer, 2), NULL);
 851                 attach_edns_record(c->buffer, &edns);
 852                 return 1;
 853         }
 854         if(edns.edns_present && edns.udp_size < NORMAL_UDP_SIZE &&
 855                 worker->daemon->cfg->harden_short_bufsize) {
 856                 verbose(VERB_QUERY, "worker request: EDNS bufsize %d ignored",
 857                         (int)edns.udp_size);
 858                 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen);
 859                 edns.udp_size = NORMAL_UDP_SIZE;
 860         }
 861         if(edns.udp_size > worker->daemon->cfg->max_udp_size &&
 862                 c->type == comm_udp) {
 863                 verbose(VERB_QUERY,
 864                         "worker request: max UDP reply size modified"
 865                         " (%d to max-udp-size)", (int)edns.udp_size);
 866                 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen);
 867                 edns.udp_size = worker->daemon->cfg->max_udp_size;
 868         }
 869         if(edns.udp_size < LDNS_HEADER_SIZE) {
 870                 verbose(VERB_ALGO, "worker request: edns is too small.");
 871                 log_addr(VERB_CLIENT, "from", &repinfo->addr, repinfo->addrlen);
 872                 LDNS_QR_SET(sldns_buffer_begin(c->buffer));
 873                 LDNS_TC_SET(sldns_buffer_begin(c->buffer));
 874                 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer),
 875                         LDNS_RCODE_SERVFAIL);
 876                 sldns_buffer_set_position(c->buffer, LDNS_HEADER_SIZE);
 877                 sldns_buffer_write_at(c->buffer, 4,
 878                         (uint8_t*)"\0\0\0\0\0\0\0\0", 8);
 879                 sldns_buffer_flip(c->buffer);
 880                 return 1;
 881         }
 882         if(worker->stats.extended)
 883                 server_stats_insquery(&worker->stats, c, qinfo.qtype,
 884                         qinfo.qclass, &edns, repinfo);
 885         if(c->type != comm_udp)
 886                 edns.udp_size = 65535; /* max size for TCP replies */
 887         if(qinfo.qclass == LDNS_RR_CLASS_CH && answer_chaos(worker, &qinfo,
 888                 &edns, c->buffer)) {
 889                 server_stats_insrcode(&worker->stats, c->buffer);
 890                 return 1;
 891         }
 892         if(local_zones_answer(worker->daemon->local_zones, &qinfo, &edns,
 893                 c->buffer, worker->scratchpad)) {
 894                 regional_free_all(worker->scratchpad);
 895                 if(sldns_buffer_limit(c->buffer) == 0) {
 896                         comm_point_drop_reply(repinfo);
 897                         return 0;
 898                 }
 899                 server_stats_insrcode(&worker->stats, c->buffer);
 900                 return 1;
 901         }
 902
 903         /* We've looked in our local zones. If the answer isn't there, we
 904          * might need to bail out based on ACLs now. */
 905         if((ret=deny_refuse_non_local(c, acl, worker, repinfo)) != -1)
 906         {
 907                 return ret;
 908         }
 909
 910         /* If this request does not have the recursion bit set, verify
 911          * ACLs allow the snooping. */
 912         if(!(LDNS_RD_WIRE(sldns_buffer_begin(c->buffer))) &&
 913                 acl != acl_allow_snoop ) {
 914                 sldns_buffer_set_limit(c->buffer, LDNS_HEADER_SIZE);
 915                 sldns_buffer_write_at(c->buffer, 4,
 916                         (uint8_t*)"\0\0\0\0\0\0\0\0", 8);
 917                 LDNS_QR_SET(sldns_buffer_begin(c->buffer));
 918                 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer),
 919                         LDNS_RCODE_REFUSED);
 920                 sldns_buffer_flip(c->buffer);
 921                 server_stats_insrcode(&worker->stats, c->buffer);
 922                 log_addr(VERB_ALGO, "refused nonrec (cache snoop) query from",
 923                         &repinfo->addr, repinfo->addrlen);
 924                 return 1;
 925         }
 926         h = query_info_hash(&qinfo);
 927         if((e=slabhash_lookup(worker->env.msg_cache, h, &qinfo, 0))) {
 928                 /* answer from cache - we have acquired a readlock on it */
 929                 if(answer_from_cache(worker, &qinfo,
 930                         (struct reply_info*)e->data,
 931                         *(uint16_t*)sldns_buffer_begin(c->buffer),
 932                         sldns_buffer_read_u16_at(c->buffer, 2), repinfo,
 933                         &edns)) {
 934                         /* prefetch it if the prefetch TTL expired */
 935                         if(worker->env.cfg->prefetch && *worker->env.now >=
 936                                 ((struct reply_info*)e->data)->prefetch_ttl) {
 937                                 time_t leeway = ((struct reply_info*)e->
 938                                         data)->ttl - *worker->env.now;
 939                                 lock_rw_unlock(&e->lock);
 940                                 reply_and_prefetch(worker, &qinfo,
 941                                         sldns_buffer_read_u16_at(c->buffer, 2),
 942                                         repinfo, leeway);
 943                                 return 0;
 944                         }
 945                         lock_rw_unlock(&e->lock);
 946                         return 1;
 947                 }
 948                 verbose(VERB_ALGO, "answer from the cache failed");
 949                 lock_rw_unlock(&e->lock);
 950         }
 951         if(!LDNS_RD_WIRE(sldns_buffer_begin(c->buffer))) {
 952                 if(answer_norec_from_cache(worker, &qinfo,
 953                         *(uint16_t*)sldns_buffer_begin(c->buffer),
 954                         sldns_buffer_read_u16_at(c->buffer, 2), repinfo,
 955                         &edns)) {
 956                         return 1;
 957                 }
 958                 verbose(VERB_ALGO, "answer norec from cache -- "
 959                         "need to validate or not primed");
 960         }
 961         sldns_buffer_rewind(c->buffer);
 962         server_stats_querymiss(&worker->stats, worker);
 963
 964         if(verbosity >= VERB_CLIENT) {
 965                 if(c->type == comm_udp)
 966                         log_addr(VERB_CLIENT, "udp request from",
 967                                 &repinfo->addr, repinfo->addrlen);
 968                 else    log_addr(VERB_CLIENT, "tcp request from",
 969                                 &repinfo->addr, repinfo->addrlen);
 970         }
 971
 972         /* grab a work request structure for this new request */
 973         mesh_new_client(worker->env.mesh, &qinfo,
 974                 sldns_buffer_read_u16_at(c->buffer, 2),
 975                 &edns, repinfo, *(uint16_t*)sldns_buffer_begin(c->buffer));
 976         worker_mem_report(worker, NULL);
 977         return 0;
 978 }
 979
 980 void
 981 worker_sighandler(int sig, void* arg)
 982 {
 983         /* note that log, print, syscalls here give race conditions. */
 984         /* we still print DETAIL logs, because this is extensive per message
 985          * logging anyway, and the operator may then have an interest
 986          * in the cause for unbound to exit */
 987         struct worker* worker = (struct worker*)arg;
 988         switch(sig) {
 989 #ifdef SIGHUP
 990                 case SIGHUP:
 991                         verbose(VERB_QUERY, "caught signal SIGHUP");
 992                         comm_base_exit(worker->base);
 993                         break;
 994 #endif
 995                 case SIGINT:
 996                         verbose(VERB_QUERY, "caught signal SIGINT");
 997                         worker->need_to_exit = 1;
 998                         comm_base_exit(worker->base);
 999                         break;
1000 #ifdef SIGQUIT
1001                 case SIGQUIT:
1002                         verbose(VERB_QUERY, "caught signal SIGQUIT");
1003                         worker->need_to_exit = 1;
1004                         comm_base_exit(worker->base);
1005                         break;
1006 #endif
1007                 case SIGTERM:
1008                         verbose(VERB_QUERY, "caught signal SIGTERM");
1009                         worker->need_to_exit = 1;
1010                         comm_base_exit(worker->base);
1011                         break;
1012                 default:
1013                         log_err("unknown signal: %d, ignored", sig);
1014                         break;
1015         }
1016 }
1017
1018 /** restart statistics timer for worker, if enabled */
1019 static void
1020 worker_restart_timer(struct worker* worker)
1021 {
1022         if(worker->env.cfg->stat_interval > 0) {
1023                 struct timeval tv;
1024 #ifndef S_SPLINT_S
1025                 tv.tv_sec = worker->env.cfg->stat_interval;
1026                 tv.tv_usec = 0;
1027 #endif
1028                 comm_timer_set(worker->stat_timer, &tv);
1029         }
1030 }
1031
1032 void worker_stat_timer_cb(void* arg)
1033 {
1034         struct worker* worker = (struct worker*)arg;
1035         server_stats_log(&worker->stats, worker, worker->thread_num);
1036         mesh_stats(worker->env.mesh, "mesh has");
1037         worker_mem_report(worker, NULL);
1038         if(!worker->daemon->cfg->stat_cumulative) {
1039                 worker_stats_clear(worker);
1040         }
1041         /* start next timer */
1042         worker_restart_timer(worker);
1043 }
1044
1045 void worker_probe_timer_cb(void* arg)
1046 {
1047         struct worker* worker = (struct worker*)arg;
1048         struct timeval tv;
1049 #ifndef S_SPLINT_S
1050         tv.tv_sec = (time_t)autr_probe_timer(&worker->env);
1051         tv.tv_usec = 0;
1052 #endif
1053         if(tv.tv_sec != 0)
1054                 comm_timer_set(worker->env.probe_timer, &tv);
1055 }
1056
1057 struct worker*
1058 worker_create(struct daemon* daemon, int id, int* ports, int n)
1059 {
1060         unsigned int seed;
1061         struct worker* worker = (struct worker*)calloc(1,
1062                 sizeof(struct worker));
1063         if(!worker)
1064                 return NULL;
1065         worker->numports = n;
1066         worker->ports = (int*)memdup(ports, sizeof(int)*n);
1067         if(!worker->ports) {
1068                 free(worker);
1069                 return NULL;
1070         }
1071         worker->daemon = daemon;
1072         worker->thread_num = id;
1073         if(!(worker->cmd = tube_create())) {
1074                 free(worker->ports);
1075                 free(worker);
1076                 return NULL;
1077         }
1078         /* create random state here to avoid locking trouble in RAND_bytes */
1079         seed = (unsigned int)time(NULL) ^ (unsigned int)getpid() ^
1080                 (((unsigned int)worker->thread_num)<<17);
1081                 /* shift thread_num so it does not match out pid bits */
1082         if(!(worker->rndstate = ub_initstate(seed, daemon->rand))) {
1083                 seed = 0;
1084                 log_err("could not init random numbers.");
1085                 tube_delete(worker->cmd);
1086                 free(worker->ports);
1087                 free(worker);
1088                 return NULL;
1089         }
1090         seed = 0;
1091         return worker;
1092 }
1093
1094 int
1095 worker_init(struct worker* worker, struct config_file *cfg,
1096         struct listen_port* ports, int do_sigs)
1097 {
1098         worker->need_to_exit = 0;
1099         worker->base = comm_base_create(do_sigs);
1100         if(!worker->base) {
1101                 log_err("could not create event handling base");
1102                 worker_delete(worker);
1103                 return 0;
1104         }
1105         comm_base_set_slow_accept_handlers(worker->base, &worker_stop_accept,
1106                 &worker_start_accept, worker);
1107         if(do_sigs) {
1108 #ifdef SIGHUP
1109                 ub_thread_sig_unblock(SIGHUP);
1110 #endif
1111                 ub_thread_sig_unblock(SIGINT);
1112 #ifdef SIGQUIT
1113                 ub_thread_sig_unblock(SIGQUIT);
1114 #endif
1115                 ub_thread_sig_unblock(SIGTERM);
1116 #ifndef LIBEVENT_SIGNAL_PROBLEM
1117                 worker->comsig = comm_signal_create(worker->base,
1118                         worker_sighandler, worker);
1119                 if(!worker->comsig
1120 #ifdef SIGHUP
1121                         || !comm_signal_bind(worker->comsig, SIGHUP)
1122 #endif
1123 #ifdef SIGQUIT
1124                         || !comm_signal_bind(worker->comsig, SIGQUIT)
1125 #endif
1126                         || !comm_signal_bind(worker->comsig, SIGTERM)
1127                         || !comm_signal_bind(worker->comsig, SIGINT)) {
1128                         log_err("could not create signal handlers");
1129                         worker_delete(worker);
1130                         return 0;
1131                 }
1132 #endif /* LIBEVENT_SIGNAL_PROBLEM */
1133                 if(!daemon_remote_open_accept(worker->daemon->rc,
1134                         worker->daemon->rc_ports, worker)) {
1135                         worker_delete(worker);
1136                         return 0;
1137                 }
1138 #ifdef UB_ON_WINDOWS
1139                 wsvc_setup_worker(worker);
1140 #endif /* UB_ON_WINDOWS */
1141         } else { /* !do_sigs */
1142                 worker->comsig = NULL;
1143         }
1144         worker->front = listen_create(worker->base, ports,
1145                 cfg->msg_buffer_size, (int)cfg->incoming_num_tcp,
1146                 worker->daemon->listen_sslctx, worker_handle_request, worker);
1147         if(!worker->front) {
1148                 log_err("could not create listening sockets");
1149                 worker_delete(worker);
1150                 return 0;
1151         }
1152         worker->back = outside_network_create(worker->base,
1153                 cfg->msg_buffer_size, (size_t)cfg->outgoing_num_ports,
1154                 cfg->out_ifs, cfg->num_out_ifs, cfg->do_ip4, cfg->do_ip6,
1155                 cfg->do_tcp?cfg->outgoing_num_tcp:0,
1156                 worker->daemon->env->infra_cache, worker->rndstate,
1157                 cfg->use_caps_bits_for_id, worker->ports, worker->numports,
1158                 cfg->unwanted_threshold, &worker_alloc_cleanup, worker,
1159                 cfg->do_udp, worker->daemon->connect_sslctx, cfg->delay_close);
1160         if(!worker->back) {
1161                 log_err("could not create outgoing sockets");
1162                 worker_delete(worker);
1163                 return 0;
1164         }
1165         /* start listening to commands */
1166         if(!tube_setup_bg_listen(worker->cmd, worker->base,
1167                 &worker_handle_control_cmd, worker)) {
1168                 log_err("could not create control compt.");
1169                 worker_delete(worker);
1170                 return 0;
1171         }
1172         worker->stat_timer = comm_timer_create(worker->base,
1173                 worker_stat_timer_cb, worker);
1174         if(!worker->stat_timer) {
1175                 log_err("could not create statistics timer");
1176         }
1177
1178         /* we use the msg_buffer_size as a good estimate for what the
1179          * user wants for memory usage sizes */
1180         worker->scratchpad = regional_create_custom(cfg->msg_buffer_size);
1181         if(!worker->scratchpad) {
1182                 log_err("malloc failure");
1183                 worker_delete(worker);
1184                 return 0;
1185         }
1186
1187         server_stats_init(&worker->stats, cfg);
1188         alloc_init(&worker->alloc, &worker->daemon->superalloc,
1189                 worker->thread_num);
1190         alloc_set_id_cleanup(&worker->alloc, &worker_alloc_cleanup, worker);
1191         worker->env = *worker->daemon->env;
1192         comm_base_timept(worker->base, &worker->env.now, &worker->env.now_tv);
1193         if(worker->thread_num == 0)
1194                 log_set_time(worker->env.now);
1195         worker->env.worker = worker;
1196         worker->env.send_query = &worker_send_query;
1197         worker->env.alloc = &worker->alloc;
1198         worker->env.rnd = worker->rndstate;
1199         worker->env.scratch = worker->scratchpad;
1200         worker->env.mesh = mesh_create(&worker->daemon->mods, &worker->env);
1201         worker->env.detach_subs = &mesh_detach_subs;
1202         worker->env.attach_sub = &mesh_attach_sub;
1203         worker->env.kill_sub = &mesh_state_delete;
1204         worker->env.detect_cycle = &mesh_detect_cycle;
1205         worker->env.scratch_buffer = sldns_buffer_new(cfg->msg_buffer_size);
1206         if(!(worker->env.fwds = forwards_create()) ||
1207                 !forwards_apply_cfg(worker->env.fwds, cfg)) {
1208                 log_err("Could not set forward zones");
1209                 worker_delete(worker);
1210                 return 0;
1211         }
1212         if(!(worker->env.hints = hints_create()) ||
1213                 !hints_apply_cfg(worker->env.hints, cfg)) {
1214                 log_err("Could not set root or stub hints");
1215                 worker_delete(worker);
1216                 return 0;
1217         }
1218         /* one probe timer per process -- if we have 5011 anchors */
1219         if(autr_get_num_anchors(worker->env.anchors) > 0
1220 #ifndef THREADS_DISABLED
1221                 && worker->thread_num == 0
1222 #endif
1223                 ) {
1224                 struct timeval tv;
1225                 tv.tv_sec = 0;
1226                 tv.tv_usec = 0;
1227                 worker->env.probe_timer = comm_timer_create(worker->base,
1228                         worker_probe_timer_cb, worker);
1229                 if(!worker->env.probe_timer) {
1230                         log_err("could not create 5011-probe timer");
1231                 } else {
1232                         /* let timer fire, then it can reset itself */
1233                         comm_timer_set(worker->env.probe_timer, &tv);
1234                 }
1235         }
1236         if(!worker->env.mesh || !worker->env.scratch_buffer) {
1237                 worker_delete(worker);
1238                 return 0;
1239         }
1240         worker_mem_report(worker, NULL);
1241         /* if statistics enabled start timer */
1242         if(worker->env.cfg->stat_interval > 0) {
1243                 verbose(VERB_ALGO, "set statistics interval %d secs",
1244                         worker->env.cfg->stat_interval);
1245                 worker_restart_timer(worker);
1246         }
1247         return 1;
1248 }
1249
1250 void
1251 worker_work(struct worker* worker)
1252 {
1253         comm_base_dispatch(worker->base);
1254 }
1255
1256 void
1257 worker_delete(struct worker* worker)
1258 {
1259         if(!worker)
1260                 return;
1261         if(worker->env.mesh && verbosity >= VERB_OPS) {
1262                 server_stats_log(&worker->stats, worker, worker->thread_num);
1263                 mesh_stats(worker->env.mesh, "mesh has");
1264                 worker_mem_report(worker, NULL);
1265         }
1266         outside_network_quit_prepare(worker->back);
1267         mesh_delete(worker->env.mesh);
1268         sldns_buffer_free(worker->env.scratch_buffer);
1269         forwards_delete(worker->env.fwds);
1270         hints_delete(worker->env.hints);
1271         listen_delete(worker->front);
1272         outside_network_delete(worker->back);
1273         comm_signal_delete(worker->comsig);
1274         tube_delete(worker->cmd);
1275         comm_timer_delete(worker->stat_timer);
1276         comm_timer_delete(worker->env.probe_timer);
1277         free(worker->ports);
1278         if(worker->thread_num == 0) {
1279                 log_set_time(NULL);
1280 #ifdef UB_ON_WINDOWS
1281                 wsvc_desetup_worker(worker);
1282 #endif /* UB_ON_WINDOWS */
1283         }
1284         comm_base_delete(worker->base);
1285         ub_randfree(worker->rndstate);
1286         alloc_clear(&worker->alloc);
1287         regional_destroy(worker->scratchpad);
1288         free(worker);
1289 }
1290
1291 struct outbound_entry*
1292 worker_send_query(uint8_t* qname, size_t qnamelen, uint16_t qtype,
1293         uint16_t qclass, uint16_t flags, int dnssec, int want_dnssec,
1294         struct sockaddr_storage* addr, socklen_t addrlen, uint8_t* zone,
1295         size_t zonelen, struct module_qstate* q)
1296 {
1297         struct worker* worker = q->env->worker;
1298         struct outbound_entry* e = (struct outbound_entry*)regional_alloc(
1299                 q->region, sizeof(*e));
1300         if(!e)
1301                 return NULL;
1302         e->qstate = q;
1303         e->qsent = outnet_serviced_query(worker->back, qname,
1304                 qnamelen, qtype, qclass, flags, dnssec, want_dnssec,
1305                 q->env->cfg->tcp_upstream, q->env->cfg->ssl_upstream, addr,
1306                 addrlen, zone, zonelen, worker_handle_service_reply, e,
1307                 worker->back->udp_buff);
1308         if(!e->qsent) {
1309                 return NULL;
1310         }
1311         return e;
1312 }
1313
1314 void
1315 worker_alloc_cleanup(void* arg)
1316 {
1317         struct worker* worker = (struct worker*)arg;
1318         slabhash_clear(&worker->env.rrset_cache->table);
1319         slabhash_clear(worker->env.msg_cache);
1320 }
1321
1322 void worker_stats_clear(struct worker* worker)
1323 {
1324         server_stats_init(&worker->stats, worker->env.cfg);
1325         mesh_stats_clear(worker->env.mesh);
1326         worker->back->unwanted_replies = 0;
1327 }
1328
1329 void worker_start_accept(void* arg)
1330 {
1331         struct worker* worker = (struct worker*)arg;
1332         listen_start_accept(worker->front);
1333         if(worker->thread_num == 0)
1334                 daemon_remote_start_accept(worker->daemon->rc);
1335 }
1336
1337 void worker_stop_accept(void* arg)
1338 {
1339         struct worker* worker = (struct worker*)arg;
1340         listen_stop_accept(worker->front);
1341         if(worker->thread_num == 0)
1342                 daemon_remote_stop_accept(worker->daemon->rc);
1343 }
1344
1345 /* --- fake callbacks for fptr_wlist to work --- */
1346 struct outbound_entry* libworker_send_query(uint8_t* ATTR_UNUSED(qname),
1347         size_t ATTR_UNUSED(qnamelen), uint16_t ATTR_UNUSED(qtype),
1348         uint16_t ATTR_UNUSED(qclass), uint16_t ATTR_UNUSED(flags),
1349         int ATTR_UNUSED(dnssec), int ATTR_UNUSED(want_dnssec),
1350         struct sockaddr_storage* ATTR_UNUSED(addr),
1351         socklen_t ATTR_UNUSED(addrlen), struct module_qstate* ATTR_UNUSED(q))
1352 {
1353         log_assert(0);
1354         return 0;
1355 }
1356
1357 int libworker_handle_reply(struct comm_point* ATTR_UNUSED(c),
1358         void* ATTR_UNUSED(arg), int ATTR_UNUSED(error),
1359         struct comm_reply* ATTR_UNUSED(reply_info))
1360 {
1361         log_assert(0);
1362         return 0;
1363 }
1364
1365 int libworker_handle_service_reply(struct comm_point* ATTR_UNUSED(c),
1366         void* ATTR_UNUSED(arg), int ATTR_UNUSED(error),
1367         struct comm_reply* ATTR_UNUSED(reply_info))
1368 {
1369         log_assert(0);
1370         return 0;
1371 }
1372
1373 void libworker_handle_control_cmd(struct tube* ATTR_UNUSED(tube),
1374         uint8_t* ATTR_UNUSED(buffer), size_t ATTR_UNUSED(len),
1375         int ATTR_UNUSED(error), void* ATTR_UNUSED(arg))
1376 {
1377         log_assert(0);
1378 }
1379
1380 void libworker_fg_done_cb(void* ATTR_UNUSED(arg), int ATTR_UNUSED(rcode),
1381         sldns_buffer* ATTR_UNUSED(buf), enum sec_status ATTR_UNUSED(s),
1382         char* ATTR_UNUSED(why_bogus))
1383 {
1384         log_assert(0);
1385 }
1386
1387 void libworker_bg_done_cb(void* ATTR_UNUSED(arg), int ATTR_UNUSED(rcode),
1388         sldns_buffer* ATTR_UNUSED(buf), enum sec_status ATTR_UNUSED(s),
1389         char* ATTR_UNUSED(why_bogus))
1390 {
1391         log_assert(0);
1392 }
1393
1394 void libworker_event_done_cb(void* ATTR_UNUSED(arg), int ATTR_UNUSED(rcode),
1395         sldns_buffer* ATTR_UNUSED(buf), enum sec_status ATTR_UNUSED(s),
1396         char* ATTR_UNUSED(why_bogus))
1397 {
1398         log_assert(0);
1399 }
1400
1401 int context_query_cmp(const void* ATTR_UNUSED(a), const void* ATTR_UNUSED(b))
1402 {
1403         log_assert(0);
1404         return 0;
1405 }
1406
1407 int order_lock_cmp(const void* ATTR_UNUSED(e1), const void* ATTR_UNUSED(e2))
1408 {
1409         log_assert(0);
1410         return 0;
1411 }
1412
1413 int codeline_cmp(const void* ATTR_UNUSED(a), const void* ATTR_UNUSED(b))
1414 {
1415         log_assert(0);
1416         return 0;
1417 }
1418