contrib/unbound/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 "libunbound/context.h"
  73 #include "libunbound/libworker.h"
  74 #include "ldns/sbuffer.h"
  75
  76 #ifdef HAVE_SYS_TYPES_H
  77 #  include <sys/types.h>
  78 #endif
  79 #ifdef HAVE_NETDB_H
  80 #include <netdb.h>
  81 #endif
  82 #include <signal.h>
  83 #ifdef UB_ON_WINDOWS
  84 #include "winrc/win_svc.h"
  85 #endif
  86
  87 /** Size of an UDP datagram */
  88 #define NORMAL_UDP_SIZE 512 /* bytes */
  89
  90 /**
  91  * seconds to add to prefetch leeway.  This is a TTL that expires old rrsets
  92  * earlier than they should in order to put the new update into the cache.
  93  * This additional value is to make sure that if not all TTLs are equal in
  94  * the message to be updated(and replaced), that rrsets with up to this much
  95  * extra TTL are also replaced.  This means that the resulting new message
  96  * will have (most likely) this TTL at least, avoiding very small 'split
  97  * second' TTLs due to operators choosing relative primes for TTLs (or so).
  98  * Also has to be at least one to break ties (and overwrite cached entry).
  99  */
 100 #define PREFETCH_EXPIRY_ADD 60
 101
 102 #ifdef UNBOUND_ALLOC_STATS
 103 /** measure memory leakage */
 104 static void
 105 debug_memleak(size_t accounted, size_t heap,
 106         size_t total_alloc, size_t total_free)
 107 {
 108         static int init = 0;
 109         static size_t base_heap, base_accounted, base_alloc, base_free;
 110         size_t base_af, cur_af, grow_af, grow_acc;
 111         if(!init) {
 112                 init = 1;
 113                 base_heap = heap;
 114                 base_accounted = accounted;
 115                 base_alloc = total_alloc;
 116                 base_free = total_free;
 117         }
 118         base_af = base_alloc - base_free;
 119         cur_af = total_alloc - total_free;
 120         grow_af = cur_af - base_af;
 121         grow_acc = accounted - base_accounted;
 122         log_info("Leakage: %d leaked. growth: %u use, %u acc, %u heap",
 123                 (int)(grow_af - grow_acc), (unsigned)grow_af,
 124                 (unsigned)grow_acc, (unsigned)(heap - base_heap));
 125 }
 126
 127 /** give debug heap size indication */
 128 static void
 129 debug_total_mem(size_t calctotal)
 130 {
 131 #ifdef HAVE_SBRK
 132         extern void* unbound_start_brk;
 133         extern size_t unbound_mem_alloc, unbound_mem_freed;
 134         void* cur = sbrk(0);
 135         int total = cur-unbound_start_brk;
 136         log_info("Total heap memory estimate: %u  total-alloc: %u  "
 137                 "total-free: %u", (unsigned)total,
 138                 (unsigned)unbound_mem_alloc, (unsigned)unbound_mem_freed);
 139         debug_memleak(calctotal, (size_t)total,
 140                 unbound_mem_alloc, unbound_mem_freed);
 141 #else
 142         (void)calctotal;
 143 #endif /* HAVE_SBRK */
 144 }
 145 #endif /* UNBOUND_ALLOC_STATS */
 146
 147 /** Report on memory usage by this thread and global */
 148 static void
 149 worker_mem_report(struct worker* ATTR_UNUSED(worker),
 150         struct serviced_query* ATTR_UNUSED(cur_serv))
 151 {
 152 #ifdef UNBOUND_ALLOC_STATS
 153         /* debug func in validator module */
 154         size_t total, front, back, mesh, msg, rrset, infra, ac, superac;
 155         size_t me, iter, val, anch;
 156         int i;
 157         if(verbosity < VERB_ALGO)
 158                 return;
 159         front = listen_get_mem(worker->front);
 160         back = outnet_get_mem(worker->back);
 161         msg = slabhash_get_mem(worker->env.msg_cache);
 162         rrset = slabhash_get_mem(&worker->env.rrset_cache->table);
 163         infra = infra_get_mem(worker->env.infra_cache);
 164         mesh = mesh_get_mem(worker->env.mesh);
 165         ac = alloc_get_mem(&worker->alloc);
 166         superac = alloc_get_mem(&worker->daemon->superalloc);
 167         anch = anchors_get_mem(worker->env.anchors);
 168         iter = 0;
 169         val = 0;
 170         for(i=0; i<worker->env.mesh->mods.num; i++) {
 171                 fptr_ok(fptr_whitelist_mod_get_mem(worker->env.mesh->
 172                         mods.mod[i]->get_mem));
 173                 if(strcmp(worker->env.mesh->mods.mod[i]->name, "validator")==0)
 174                         val += (*worker->env.mesh->mods.mod[i]->get_mem)
 175                                 (&worker->env, i);
 176                 else    iter += (*worker->env.mesh->mods.mod[i]->get_mem)
 177                                 (&worker->env, i);
 178         }
 179         me = sizeof(*worker) + sizeof(*worker->base) + sizeof(*worker->comsig)
 180                 + comm_point_get_mem(worker->cmd_com)
 181                 + sizeof(worker->rndstate)
 182                 + regional_get_mem(worker->scratchpad)
 183                 + sizeof(*worker->env.scratch_buffer)
 184                 + sldns_buffer_capacity(worker->env.scratch_buffer)
 185                 + forwards_get_mem(worker->env.fwds)
 186                 + hints_get_mem(worker->env.hints);
 187         if(worker->thread_num == 0)
 188                 me += acl_list_get_mem(worker->daemon->acl);
 189         if(cur_serv) {
 190                 me += serviced_get_mem(cur_serv);
 191         }
 192         total = front+back+mesh+msg+rrset+infra+iter+val+ac+superac+me;
 193         log_info("Memory conditions: %u front=%u back=%u mesh=%u msg=%u "
 194                 "rrset=%u infra=%u iter=%u val=%u anchors=%u "
 195                 "alloccache=%u globalalloccache=%u me=%u",
 196                 (unsigned)total, (unsigned)front, (unsigned)back,
 197                 (unsigned)mesh, (unsigned)msg, (unsigned)rrset,
 198                 (unsigned)infra, (unsigned)iter, (unsigned)val, (unsigned)anch,
 199                 (unsigned)ac, (unsigned)superac, (unsigned)me);
 200         debug_total_mem(total);
 201 #else /* no UNBOUND_ALLOC_STATS */
 202         size_t val = 0;
 203         int i;
 204         if(verbosity < VERB_QUERY)
 205                 return;
 206         for(i=0; i<worker->env.mesh->mods.num; i++) {
 207                 fptr_ok(fptr_whitelist_mod_get_mem(worker->env.mesh->
 208                         mods.mod[i]->get_mem));
 209                 if(strcmp(worker->env.mesh->mods.mod[i]->name, "validator")==0)
 210                         val += (*worker->env.mesh->mods.mod[i]->get_mem)
 211                                 (&worker->env, i);
 212         }
 213         verbose(VERB_QUERY, "cache memory msg=%u rrset=%u infra=%u val=%u",
 214                 (unsigned)slabhash_get_mem(worker->env.msg_cache),
 215                 (unsigned)slabhash_get_mem(&worker->env.rrset_cache->table),
 216                 (unsigned)infra_get_mem(worker->env.infra_cache),
 217                 (unsigned)val);
 218 #endif /* UNBOUND_ALLOC_STATS */
 219 }
 220
 221 void
 222 worker_send_cmd(struct worker* worker, enum worker_commands cmd)
 223 {
 224         uint32_t c = (uint32_t)htonl(cmd);
 225         if(!tube_write_msg(worker->cmd, (uint8_t*)&c, sizeof(c), 0)) {
 226                 log_err("worker send cmd %d failed", (int)cmd);
 227         }
 228 }
 229
 230 int
 231 worker_handle_reply(struct comm_point* c, void* arg, int error,
 232         struct comm_reply* reply_info)
 233 {
 234         struct module_qstate* q = (struct module_qstate*)arg;
 235         struct worker* worker = q->env->worker;
 236         struct outbound_entry e;
 237         e.qstate = q;
 238         e.qsent = NULL;
 239
 240         if(error != 0) {
 241                 mesh_report_reply(worker->env.mesh, &e, reply_info, error);
 242                 worker_mem_report(worker, NULL);
 243                 return 0;
 244         }
 245         /* sanity check. */
 246         if(!LDNS_QR_WIRE(sldns_buffer_begin(c->buffer))
 247                 || LDNS_OPCODE_WIRE(sldns_buffer_begin(c->buffer)) !=
 248                         LDNS_PACKET_QUERY
 249                 || LDNS_QDCOUNT(sldns_buffer_begin(c->buffer)) > 1) {
 250                 /* error becomes timeout for the module as if this reply
 251                  * never arrived. */
 252                 mesh_report_reply(worker->env.mesh, &e, reply_info,
 253                         NETEVENT_TIMEOUT);
 254                 worker_mem_report(worker, NULL);
 255                 return 0;
 256         }
 257         mesh_report_reply(worker->env.mesh, &e, reply_info, NETEVENT_NOERROR);
 258         worker_mem_report(worker, NULL);
 259         return 0;
 260 }
 261
 262 int
 263 worker_handle_service_reply(struct comm_point* c, void* arg, int error,
 264         struct comm_reply* reply_info)
 265 {
 266         struct outbound_entry* e = (struct outbound_entry*)arg;
 267         struct worker* worker = e->qstate->env->worker;
 268         struct serviced_query *sq = e->qsent;
 269
 270         verbose(VERB_ALGO, "worker svcd callback for qstate %p", e->qstate);
 271         if(error != 0) {
 272                 mesh_report_reply(worker->env.mesh, e, reply_info, error);
 273                 worker_mem_report(worker, sq);
 274                 return 0;
 275         }
 276         /* sanity check. */
 277         if(!LDNS_QR_WIRE(sldns_buffer_begin(c->buffer))
 278                 || LDNS_OPCODE_WIRE(sldns_buffer_begin(c->buffer)) !=
 279                         LDNS_PACKET_QUERY
 280                 || LDNS_QDCOUNT(sldns_buffer_begin(c->buffer)) > 1) {
 281                 /* error becomes timeout for the module as if this reply
 282                  * never arrived. */
 283                 verbose(VERB_ALGO, "worker: bad reply handled as timeout");
 284                 mesh_report_reply(worker->env.mesh, e, reply_info,
 285                         NETEVENT_TIMEOUT);
 286                 worker_mem_report(worker, sq);
 287                 return 0;
 288         }
 289         mesh_report_reply(worker->env.mesh, e, reply_info, NETEVENT_NOERROR);
 290         worker_mem_report(worker, sq);
 291         return 0;
 292 }
 293
 294 /** check request sanity.
 295  * @param pkt: the wire packet to examine for sanity.
 296  * @param worker: parameters for checking.
 297  * @return error code, 0 OK, or -1 discard.
 298 */
 299 static int
 300 worker_check_request(sldns_buffer* pkt, struct worker* worker)
 301 {
 302         if(sldns_buffer_limit(pkt) < LDNS_HEADER_SIZE) {
 303                 verbose(VERB_QUERY, "request too short, discarded");
 304                 return -1;
 305         }
 306         if(sldns_buffer_limit(pkt) > NORMAL_UDP_SIZE &&
 307                 worker->daemon->cfg->harden_large_queries) {
 308                 verbose(VERB_QUERY, "request too large, discarded");
 309                 return -1;
 310         }
 311         if(LDNS_QR_WIRE(sldns_buffer_begin(pkt))) {
 312                 verbose(VERB_QUERY, "request has QR bit on, discarded");
 313                 return -1;
 314         }
 315         if(LDNS_TC_WIRE(sldns_buffer_begin(pkt))) {
 316                 LDNS_TC_CLR(sldns_buffer_begin(pkt));
 317                 verbose(VERB_QUERY, "request bad, has TC bit on");
 318                 return LDNS_RCODE_FORMERR;
 319         }
 320         if(LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt)) != LDNS_PACKET_QUERY) {
 321                 verbose(VERB_QUERY, "request unknown opcode %d",
 322                         LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt)));
 323                 return LDNS_RCODE_NOTIMPL;
 324         }
 325         if(LDNS_QDCOUNT(sldns_buffer_begin(pkt)) != 1) {
 326                 verbose(VERB_QUERY, "request wrong nr qd=%d",
 327                         LDNS_QDCOUNT(sldns_buffer_begin(pkt)));
 328                 return LDNS_RCODE_FORMERR;
 329         }
 330         if(LDNS_ANCOUNT(sldns_buffer_begin(pkt)) != 0) {
 331                 verbose(VERB_QUERY, "request wrong nr an=%d",
 332                         LDNS_ANCOUNT(sldns_buffer_begin(pkt)));
 333                 return LDNS_RCODE_FORMERR;
 334         }
 335         if(LDNS_NSCOUNT(sldns_buffer_begin(pkt)) != 0) {
 336                 verbose(VERB_QUERY, "request wrong nr ns=%d",
 337                         LDNS_NSCOUNT(sldns_buffer_begin(pkt)));
 338                 return LDNS_RCODE_FORMERR;
 339         }
 340         if(LDNS_ARCOUNT(sldns_buffer_begin(pkt)) > 1) {
 341                 verbose(VERB_QUERY, "request wrong nr ar=%d",
 342                         LDNS_ARCOUNT(sldns_buffer_begin(pkt)));
 343                 return LDNS_RCODE_FORMERR;
 344         }
 345         return 0;
 346 }
 347
 348 void
 349 worker_handle_control_cmd(struct tube* ATTR_UNUSED(tube), uint8_t* msg,
 350         size_t len, int error, void* arg)
 351 {
 352         struct worker* worker = (struct worker*)arg;
 353         enum worker_commands cmd;
 354         if(error != NETEVENT_NOERROR) {
 355                 free(msg);
 356                 if(error == NETEVENT_CLOSED)
 357                         comm_base_exit(worker->base);
 358                 else    log_info("control event: %d", error);
 359                 return;
 360         }
 361         if(len != sizeof(uint32_t)) {
 362                 fatal_exit("bad control msg length %d", (int)len);
 363         }
 364         cmd = sldns_read_uint32(msg);
 365         free(msg);
 366         switch(cmd) {
 367         case worker_cmd_quit:
 368                 verbose(VERB_ALGO, "got control cmd quit");
 369                 comm_base_exit(worker->base);
 370                 break;
 371         case worker_cmd_stats:
 372                 verbose(VERB_ALGO, "got control cmd stats");
 373                 server_stats_reply(worker, 1);
 374                 break;
 375         case worker_cmd_stats_noreset:
 376                 verbose(VERB_ALGO, "got control cmd stats_noreset");
 377                 server_stats_reply(worker, 0);
 378                 break;
 379         case worker_cmd_remote:
 380                 verbose(VERB_ALGO, "got control cmd remote");
 381                 daemon_remote_exec(worker);
 382                 break;
 383         default:
 384                 log_err("bad command %d", (int)cmd);
 385                 break;
 386         }
 387 }
 388
 389 /** check if a delegation is secure */
 390 static enum sec_status
 391 check_delegation_secure(struct reply_info *rep)
 392 {
 393         /* return smallest security status */
 394         size_t i;
 395         enum sec_status sec = sec_status_secure;
 396         enum sec_status s;
 397         size_t num = rep->an_numrrsets + rep->ns_numrrsets;
 398         /* check if answer and authority are OK */
 399         for(i=0; i<num; i++) {
 400                 s = ((struct packed_rrset_data*)rep->rrsets[i]->entry.data)
 401                         ->security;
 402                 if(s < sec)
 403                         sec = s;
 404         }
 405         /* in additional, only unchecked triggers revalidation */
 406         for(i=num; i<rep->rrset_count; i++) {
 407                 s = ((struct packed_rrset_data*)rep->rrsets[i]->entry.data)
 408                         ->security;
 409                 if(s == sec_status_unchecked)
 410                         return s;
 411         }
 412         return sec;
 413 }
 414
 415 /** remove nonsecure from a delegation referral additional section */
 416 static void
 417 deleg_remove_nonsecure_additional(struct reply_info* rep)
 418 {
 419         /* we can simply edit it, since we are working in the scratch region */
 420         size_t i;
 421         enum sec_status s;
 422
 423         for(i = rep->an_numrrsets+rep->ns_numrrsets; i<rep->rrset_count; i++) {
 424                 s = ((struct packed_rrset_data*)rep->rrsets[i]->entry.data)
 425                         ->security;
 426                 if(s != sec_status_secure) {
 427                         memmove(rep->rrsets+i, rep->rrsets+i+1,
 428                                 sizeof(struct ub_packed_rrset_key*)*
 429                                 (rep->rrset_count - i - 1));
 430                         rep->ar_numrrsets--;
 431                         rep->rrset_count--;
 432                         i--;
 433                 }
 434         }
 435 }
 436
 437 /** answer nonrecursive query from the cache */
 438 static int
 439 answer_norec_from_cache(struct worker* worker, struct query_info* qinfo,
 440         uint16_t id, uint16_t flags, struct comm_reply* repinfo,
 441         struct edns_data* edns)
 442 {
 443         /* for a nonrecursive query return either:
 444          *      o an error (servfail; we try to avoid this)
 445          *      o a delegation (closest we have; this routine tries that)
 446          *      o the answer (checked by answer_from_cache)
 447          *
 448          * So, grab a delegation from the rrset cache.
 449          * Then check if it needs validation, if so, this routine fails,
 450          * so that iterator can prime and validator can verify rrsets.
 451          */
 452         uint16_t udpsize = edns->udp_size;
 453         int secure = 0;
 454         time_t timenow = *worker->env.now;
 455         int must_validate = (!(flags&BIT_CD) || worker->env.cfg->ignore_cd)
 456                 && worker->env.need_to_validate;
 457         struct dns_msg *msg = NULL;
 458         struct delegpt *dp;
 459
 460         dp = dns_cache_find_delegation(&worker->env, qinfo->qname,
 461                 qinfo->qname_len, qinfo->qtype, qinfo->qclass,
 462                 worker->scratchpad, &msg, timenow);
 463         if(!dp) { /* no delegation, need to reprime */
 464                 regional_free_all(worker->scratchpad);
 465                 return 0;
 466         }
 467         if(must_validate) {
 468                 switch(check_delegation_secure(msg->rep)) {
 469                 case sec_status_unchecked:
 470                         /* some rrsets have not been verified yet, go and
 471                          * let validator do that */
 472                         regional_free_all(worker->scratchpad);
 473                         return 0;
 474                 case sec_status_bogus:
 475                         /* some rrsets are bogus, reply servfail */
 476                         edns->edns_version = EDNS_ADVERTISED_VERSION;
 477                         edns->udp_size = EDNS_ADVERTISED_SIZE;
 478                         edns->ext_rcode = 0;
 479                         edns->bits &= EDNS_DO;
 480                         error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL,
 481                                 &msg->qinfo, id, flags, edns);
 482                         regional_free_all(worker->scratchpad);
 483                         if(worker->stats.extended) {
 484                                 worker->stats.ans_bogus++;
 485                                 worker->stats.ans_rcode[LDNS_RCODE_SERVFAIL]++;
 486                         }
 487                         return 1;
 488                 case sec_status_secure:
 489                         /* all rrsets are secure */
 490                         /* remove non-secure rrsets from the add. section*/
 491                         if(worker->env.cfg->val_clean_additional)
 492                                 deleg_remove_nonsecure_additional(msg->rep);
 493                         secure = 1;
 494                         break;
 495                 case sec_status_indeterminate:
 496                 case sec_status_insecure:
 497                 default:
 498                         /* not secure */
 499                         secure = 0;
 500                         break;
 501                 }
 502         }
 503         /* return this delegation from the cache */
 504         edns->edns_version = EDNS_ADVERTISED_VERSION;
 505         edns->udp_size = EDNS_ADVERTISED_SIZE;
 506         edns->ext_rcode = 0;
 507         edns->bits &= EDNS_DO;
 508         msg->rep->flags |= BIT_QR|BIT_RA;
 509         if(!reply_info_answer_encode(&msg->qinfo, msg->rep, id, flags,
 510                 repinfo->c->buffer, 0, 1, worker->scratchpad,
 511                 udpsize, edns, (int)(edns->bits & EDNS_DO), secure)) {
 512                 error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL,
 513                         &msg->qinfo, id, flags, edns);
 514         }
 515         regional_free_all(worker->scratchpad);
 516         if(worker->stats.extended) {
 517                 if(secure) worker->stats.ans_secure++;
 518                 server_stats_insrcode(&worker->stats, repinfo->c->buffer);
 519         }
 520         return 1;
 521 }
 522
 523 /** answer query from the cache */
 524 static int
 525 answer_from_cache(struct worker* worker, struct query_info* qinfo,
 526         struct reply_info* rep, uint16_t id, uint16_t flags,
 527         struct comm_reply* repinfo, struct edns_data* edns)
 528 {
 529         time_t timenow = *worker->env.now;
 530         uint16_t udpsize = edns->udp_size;
 531         int secure;
 532         int must_validate = (!(flags&BIT_CD) || worker->env.cfg->ignore_cd)
 533                 && worker->env.need_to_validate;
 534         /* see if it is possible */
 535         if(rep->ttl < timenow) {
 536                 /* the rrsets may have been updated in the meantime.
 537                  * we will refetch the message format from the
 538                  * authoritative server
 539                  */
 540                 return 0;
 541         }
 542         if(!rrset_array_lock(rep->ref, rep->rrset_count, timenow))
 543                 return 0;
 544         /* locked and ids and ttls are OK. */
 545         /* check CNAME chain (if any) */
 546         if(rep->an_numrrsets > 0 && (rep->rrsets[0]->rk.type ==
 547                 htons(LDNS_RR_TYPE_CNAME) || rep->rrsets[0]->rk.type ==
 548                 htons(LDNS_RR_TYPE_DNAME))) {
 549                 if(!reply_check_cname_chain(rep)) {
 550                         /* cname chain invalid, redo iterator steps */
 551                         verbose(VERB_ALGO, "Cache reply: cname chain broken");
 552                 bail_out:
 553                         rrset_array_unlock_touch(worker->env.rrset_cache,
 554                                 worker->scratchpad, rep->ref, rep->rrset_count);
 555                         regional_free_all(worker->scratchpad);
 556                         return 0;
 557                 }
 558         }
 559         /* check security status of the cached answer */
 560         if( rep->security == sec_status_bogus && must_validate) {
 561                 /* BAD cached */
 562                 edns->edns_version = EDNS_ADVERTISED_VERSION;
 563                 edns->udp_size = EDNS_ADVERTISED_SIZE;
 564                 edns->ext_rcode = 0;
 565                 edns->bits &= EDNS_DO;
 566                 error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL,
 567                         qinfo, id, flags, edns);
 568                 rrset_array_unlock_touch(worker->env.rrset_cache,
 569                         worker->scratchpad, rep->ref, rep->rrset_count);
 570                 regional_free_all(worker->scratchpad);
 571                 if(worker->stats.extended) {
 572                         worker->stats.ans_bogus ++;
 573                         worker->stats.ans_rcode[LDNS_RCODE_SERVFAIL] ++;
 574                 }
 575                 return 1;
 576         } else if( rep->security == sec_status_unchecked && must_validate) {
 577                 verbose(VERB_ALGO, "Cache reply: unchecked entry needs "
 578                         "validation");
 579                 goto bail_out; /* need to validate cache entry first */
 580         } else if(rep->security == sec_status_secure) {
 581                 if(reply_all_rrsets_secure(rep))
 582                         secure = 1;
 583                 else    {
 584                         if(must_validate) {
 585                                 verbose(VERB_ALGO, "Cache reply: secure entry"
 586                                         " changed status");
 587                                 goto bail_out; /* rrset changed, re-verify */
 588                         }
 589                         secure = 0;
 590                 }
 591         } else  secure = 0;
 592
 593         edns->edns_version = EDNS_ADVERTISED_VERSION;
 594         edns->udp_size = EDNS_ADVERTISED_SIZE;
 595         edns->ext_rcode = 0;
 596         edns->bits &= EDNS_DO;
 597         if(!reply_info_answer_encode(qinfo, rep, id, flags,
 598                 repinfo->c->buffer, timenow, 1, worker->scratchpad,
 599                 udpsize, edns, (int)(edns->bits & EDNS_DO), secure)) {
 600                 error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL,
 601                         qinfo, id, flags, edns);
 602         }
 603         /* cannot send the reply right now, because blocking network syscall
 604          * is bad while holding locks. */
 605         rrset_array_unlock_touch(worker->env.rrset_cache, worker->scratchpad,
 606                 rep->ref, rep->rrset_count);
 607         regional_free_all(worker->scratchpad);
 608         if(worker->stats.extended) {
 609                 if(secure) worker->stats.ans_secure++;
 610                 server_stats_insrcode(&worker->stats, repinfo->c->buffer);
 611         }
 612         /* go and return this buffer to the client */
 613         return 1;
 614 }
 615
 616 /** Reply to client and perform prefetch to keep cache up to date */
 617 static void
 618 reply_and_prefetch(struct worker* worker, struct query_info* qinfo,
 619         uint16_t flags, struct comm_reply* repinfo, time_t leeway)
 620 {
 621         /* first send answer to client to keep its latency
 622          * as small as a cachereply */
 623         comm_point_send_reply(repinfo);
 624         server_stats_prefetch(&worker->stats, worker);
 625
 626         /* create the prefetch in the mesh as a normal lookup without
 627          * client addrs waiting, which has the cache blacklisted (to bypass
 628          * the cache and go to the network for the data). */
 629         /* this (potentially) runs the mesh for the new query */
 630         mesh_new_prefetch(worker->env.mesh, qinfo, flags, leeway +
 631                 PREFETCH_EXPIRY_ADD);
 632 }
 633
 634 /**
 635  * Fill CH class answer into buffer. Keeps query.
 636  * @param pkt: buffer
 637  * @param str: string to put into text record (<255).
 638  * @param edns: edns reply information.
 639  */
 640 static void
 641 chaos_replystr(sldns_buffer* pkt, const char* str, struct edns_data* edns)
 642 {
 643         size_t len = strlen(str);
 644         unsigned int rd = LDNS_RD_WIRE(sldns_buffer_begin(pkt));
 645         unsigned int cd = LDNS_CD_WIRE(sldns_buffer_begin(pkt));
 646         if(len>255) len=255; /* cap size of TXT record */
 647         sldns_buffer_clear(pkt);
 648         sldns_buffer_skip(pkt, (ssize_t)sizeof(uint16_t)); /* skip id */
 649         sldns_buffer_write_u16(pkt, (uint16_t)(BIT_QR|BIT_RA));
 650         if(rd) LDNS_RD_SET(sldns_buffer_begin(pkt));
 651         if(cd) LDNS_CD_SET(sldns_buffer_begin(pkt));
 652         sldns_buffer_write_u16(pkt, 1); /* qdcount */
 653         sldns_buffer_write_u16(pkt, 1); /* ancount */
 654         sldns_buffer_write_u16(pkt, 0); /* nscount */
 655         sldns_buffer_write_u16(pkt, 0); /* arcount */
 656         (void)query_dname_len(pkt); /* skip qname */
 657         sldns_buffer_skip(pkt, (ssize_t)sizeof(uint16_t)); /* skip qtype */
 658         sldns_buffer_skip(pkt, (ssize_t)sizeof(uint16_t)); /* skip qclass */
 659         sldns_buffer_write_u16(pkt, 0xc00c); /* compr ptr to query */
 660         sldns_buffer_write_u16(pkt, LDNS_RR_TYPE_TXT);
 661         sldns_buffer_write_u16(pkt, LDNS_RR_CLASS_CH);
 662         sldns_buffer_write_u32(pkt, 0); /* TTL */
 663         sldns_buffer_write_u16(pkt, sizeof(uint8_t) + len);
 664         sldns_buffer_write_u8(pkt, len);
 665         sldns_buffer_write(pkt, str, len);
 666         sldns_buffer_flip(pkt);
 667         edns->edns_version = EDNS_ADVERTISED_VERSION;
 668         edns->udp_size = EDNS_ADVERTISED_SIZE;
 669         edns->bits &= EDNS_DO;
 670         attach_edns_record(pkt, edns);
 671 }
 672
 673 /**
 674  * Answer CH class queries.
 675  * @param w: worker
 676  * @param qinfo: query info. Pointer into packet buffer.
 677  * @param edns: edns info from query.
 678  * @param pkt: packet buffer.
 679  * @return: true if a reply is to be sent.
 680  */
 681 static int
 682 answer_chaos(struct worker* w, struct query_info* qinfo,
 683         struct edns_data* edns, sldns_buffer* pkt)
 684 {
 685         struct config_file* cfg = w->env.cfg;
 686         if(qinfo->qtype != LDNS_RR_TYPE_ANY && qinfo->qtype != LDNS_RR_TYPE_TXT)
 687                 return 0;
 688         if(query_dname_compare(qinfo->qname,
 689                 (uint8_t*)"\002id\006server") == 0 ||
 690                 query_dname_compare(qinfo->qname,
 691                 (uint8_t*)"\010hostname\004bind") == 0)
 692         {
 693                 if(cfg->hide_identity)
 694                         return 0;
 695                 if(cfg->identity==NULL || cfg->identity[0]==0) {
 696                         char buf[MAXHOSTNAMELEN+1];
 697                         if (gethostname(buf, MAXHOSTNAMELEN) == 0) {
 698                                 buf[MAXHOSTNAMELEN] = 0;
 699                                 chaos_replystr(pkt, buf, edns);
 700                         } else  {
 701                                 log_err("gethostname: %s", strerror(errno));
 702                                 chaos_replystr(pkt, "no hostname", edns);
 703                         }
 704                 }
 705                 else    chaos_replystr(pkt, cfg->identity, edns);
 706                 return 1;
 707         }
 708         if(query_dname_compare(qinfo->qname,
 709                 (uint8_t*)"\007version\006server") == 0 ||
 710                 query_dname_compare(qinfo->qname,
 711                 (uint8_t*)"\007version\004bind") == 0)
 712         {
 713                 if(cfg->hide_version)
 714                         return 0;
 715                 if(cfg->version==NULL || cfg->version[0]==0)
 716                         chaos_replystr(pkt, PACKAGE_STRING, edns);
 717                 else    chaos_replystr(pkt, cfg->version, edns);
 718                 return 1;
 719         }
 720         return 0;
 721 }
 722
 723 static int
 724 deny_refuse(struct comm_point* c, enum acl_access acl,
 725         enum acl_access deny, enum acl_access refuse,
 726         struct worker* worker, struct comm_reply* repinfo)
 727 {
 728         if(acl == deny) {
 729                 comm_point_drop_reply(repinfo);
 730                 if(worker->stats.extended)
 731                         worker->stats.unwanted_queries++;
 732                 return 0;
 733         } else if(acl == refuse) {
 734                 log_addr(VERB_ALGO, "refused query from",
 735                         &repinfo->addr, repinfo->addrlen);
 736                 log_buf(VERB_ALGO, "refuse", c->buffer);
 737                 if(worker->stats.extended)
 738                         worker->stats.unwanted_queries++;
 739                 if(worker_check_request(c->buffer, worker) == -1) {
 740                         comm_point_drop_reply(repinfo);
 741                         return 0; /* discard this */
 742                 }
 743                 sldns_buffer_set_limit(c->buffer, LDNS_HEADER_SIZE);
 744                 sldns_buffer_write_at(c->buffer, 4,
 745                         (uint8_t*)"\0\0\0\0\0\0\0\0", 8);
 746                 LDNS_QR_SET(sldns_buffer_begin(c->buffer));
 747                 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer),
 748                         LDNS_RCODE_REFUSED);
 749                 return 1;
 750         }
 751
 752         return -1;
 753 }
 754
 755 static int
 756 deny_refuse_all(struct comm_point* c, enum acl_access acl,
 757         struct worker* worker, struct comm_reply* repinfo)
 758 {
 759         return deny_refuse(c, acl, acl_deny, acl_refuse, worker, repinfo);
 760 }
 761
 762 static int
 763 deny_refuse_non_local(struct comm_point* c, enum acl_access acl,
 764         struct worker* worker, struct comm_reply* repinfo)
 765 {
 766         return deny_refuse(c, acl, acl_deny_non_local, acl_refuse_non_local, worker, repinfo);
 767 }
 768
 769 int
 770 worker_handle_request(struct comm_point* c, void* arg, int error,
 771         struct comm_reply* repinfo)
 772 {
 773         struct worker* worker = (struct worker*)arg;
 774         int ret;
 775         hashvalue_t h;
 776         struct lruhash_entry* e;
 777         struct query_info qinfo;
 778         struct edns_data edns;
 779         enum acl_access acl;
 780         int rc = 0;
 781
 782         if(error != NETEVENT_NOERROR) {
 783                 /* some bad tcp query DNS formats give these error calls */
 784                 verbose(VERB_ALGO, "handle request called with err=%d", error);
 785                 return 0;
 786         }
 787 #ifdef USE_DNSTAP
 788         if(worker->dtenv.log_client_query_messages)
 789                 dt_msg_send_client_query(&worker->dtenv, &repinfo->addr, c->type,
 790                         c->buffer);
 791 #endif
 792         acl = acl_list_lookup(worker->daemon->acl, &repinfo->addr,
 793                 repinfo->addrlen);
 794         if((ret=deny_refuse_all(c, acl, worker, repinfo)) != -1)
 795         {
 796                 if(ret == 1)
 797                         goto send_reply;
 798                 return ret;
 799         }
 800         if((ret=worker_check_request(c->buffer, worker)) != 0) {
 801                 verbose(VERB_ALGO, "worker check request: bad query.");
 802                 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen);
 803                 if(ret != -1) {
 804                         LDNS_QR_SET(sldns_buffer_begin(c->buffer));
 805                         LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), ret);
 806                         return 1;
 807                 }
 808                 comm_point_drop_reply(repinfo);
 809                 return 0;
 810         }
 811         worker->stats.num_queries++;
 812         /* see if query is in the cache */
 813         if(!query_info_parse(&qinfo, c->buffer)) {
 814                 verbose(VERB_ALGO, "worker parse request: formerror.");
 815                 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen);
 816                 sldns_buffer_rewind(c->buffer);
 817                 LDNS_QR_SET(sldns_buffer_begin(c->buffer));
 818                 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer),
 819                         LDNS_RCODE_FORMERR);
 820                 server_stats_insrcode(&worker->stats, c->buffer);
 821                 goto send_reply;
 822         }
 823         if(worker->env.cfg->log_queries) {
 824                 char ip[128];
 825                 addr_to_str(&repinfo->addr, repinfo->addrlen, ip, sizeof(ip));
 826                 log_nametypeclass(0, ip, qinfo.qname, qinfo.qtype, qinfo.qclass);
 827         }
 828         if(qinfo.qtype == LDNS_RR_TYPE_AXFR ||
 829                 qinfo.qtype == LDNS_RR_TYPE_IXFR) {
 830                 verbose(VERB_ALGO, "worker request: refused zone transfer.");
 831                 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen);
 832                 sldns_buffer_rewind(c->buffer);
 833                 LDNS_QR_SET(sldns_buffer_begin(c->buffer));
 834                 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer),
 835                         LDNS_RCODE_REFUSED);
 836                 if(worker->stats.extended) {
 837                         worker->stats.qtype[qinfo.qtype]++;
 838                         server_stats_insrcode(&worker->stats, c->buffer);
 839                 }
 840                 goto send_reply;
 841         }
 842         if((ret=parse_edns_from_pkt(c->buffer, &edns)) != 0) {
 843                 verbose(VERB_ALGO, "worker parse edns: formerror.");
 844                 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen);
 845                 sldns_buffer_rewind(c->buffer);
 846                 LDNS_QR_SET(sldns_buffer_begin(c->buffer));
 847                 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), ret);
 848                 server_stats_insrcode(&worker->stats, c->buffer);
 849                 goto send_reply;
 850         }
 851         if(edns.edns_present && edns.edns_version != 0) {
 852                 edns.ext_rcode = (uint8_t)(EDNS_RCODE_BADVERS>>4);
 853                 edns.edns_version = EDNS_ADVERTISED_VERSION;
 854                 edns.udp_size = EDNS_ADVERTISED_SIZE;
 855                 edns.bits &= EDNS_DO;
 856                 verbose(VERB_ALGO, "query with bad edns version.");
 857                 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen);
 858                 error_encode(c->buffer, EDNS_RCODE_BADVERS&0xf, &qinfo,
 859                         *(uint16_t*)(void *)sldns_buffer_begin(c->buffer),
 860                         sldns_buffer_read_u16_at(c->buffer, 2), NULL);
 861                 attach_edns_record(c->buffer, &edns);
 862                 goto send_reply;
 863         }
 864         if(edns.edns_present && edns.udp_size < NORMAL_UDP_SIZE &&
 865                 worker->daemon->cfg->harden_short_bufsize) {
 866                 verbose(VERB_QUERY, "worker request: EDNS bufsize %d ignored",
 867                         (int)edns.udp_size);
 868                 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen);
 869                 edns.udp_size = NORMAL_UDP_SIZE;
 870         }
 871         if(edns.udp_size > worker->daemon->cfg->max_udp_size &&
 872                 c->type == comm_udp) {
 873                 verbose(VERB_QUERY,
 874                         "worker request: max UDP reply size modified"
 875                         " (%d to max-udp-size)", (int)edns.udp_size);
 876                 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen);
 877                 edns.udp_size = worker->daemon->cfg->max_udp_size;
 878         }
 879         if(edns.udp_size < LDNS_HEADER_SIZE) {
 880                 verbose(VERB_ALGO, "worker request: edns is too small.");
 881                 log_addr(VERB_CLIENT, "from", &repinfo->addr, repinfo->addrlen);
 882                 LDNS_QR_SET(sldns_buffer_begin(c->buffer));
 883                 LDNS_TC_SET(sldns_buffer_begin(c->buffer));
 884                 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer),
 885                         LDNS_RCODE_SERVFAIL);
 886                 sldns_buffer_set_position(c->buffer, LDNS_HEADER_SIZE);
 887                 sldns_buffer_write_at(c->buffer, 4,
 888                         (uint8_t*)"\0\0\0\0\0\0\0\0", 8);
 889                 sldns_buffer_flip(c->buffer);
 890                 goto send_reply;
 891         }
 892         if(worker->stats.extended)
 893                 server_stats_insquery(&worker->stats, c, qinfo.qtype,
 894                         qinfo.qclass, &edns, repinfo);
 895         if(c->type != comm_udp)
 896                 edns.udp_size = 65535; /* max size for TCP replies */
 897         if(qinfo.qclass == LDNS_RR_CLASS_CH && answer_chaos(worker, &qinfo,
 898                 &edns, c->buffer)) {
 899                 server_stats_insrcode(&worker->stats, c->buffer);
 900                 goto send_reply;
 901         }
 902         if(local_zones_answer(worker->daemon->local_zones, &qinfo, &edns,
 903                 c->buffer, worker->scratchpad, repinfo)) {
 904                 regional_free_all(worker->scratchpad);
 905                 if(sldns_buffer_limit(c->buffer) == 0) {
 906                         comm_point_drop_reply(repinfo);
 907                         return 0;
 908                 }
 909                 server_stats_insrcode(&worker->stats, c->buffer);
 910                 goto send_reply;
 911         }
 912
 913         /* We've looked in our local zones. If the answer isn't there, we
 914          * might need to bail out based on ACLs now. */
 915         if((ret=deny_refuse_non_local(c, acl, worker, repinfo)) != -1)
 916         {
 917                 if(ret == 1)
 918                         goto send_reply;
 919                 return ret;
 920         }
 921
 922         /* If this request does not have the recursion bit set, verify
 923          * ACLs allow the snooping. */
 924         if(!(LDNS_RD_WIRE(sldns_buffer_begin(c->buffer))) &&
 925                 acl != acl_allow_snoop ) {
 926                 sldns_buffer_set_limit(c->buffer, LDNS_HEADER_SIZE);
 927                 sldns_buffer_write_at(c->buffer, 4,
 928                         (uint8_t*)"\0\0\0\0\0\0\0\0", 8);
 929                 LDNS_QR_SET(sldns_buffer_begin(c->buffer));
 930                 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer),
 931                         LDNS_RCODE_REFUSED);
 932                 sldns_buffer_flip(c->buffer);
 933                 server_stats_insrcode(&worker->stats, c->buffer);
 934                 log_addr(VERB_ALGO, "refused nonrec (cache snoop) query from",
 935                         &repinfo->addr, repinfo->addrlen);
 936                 goto send_reply;
 937         }
 938         h = query_info_hash(&qinfo, sldns_buffer_read_u16_at(c->buffer, 2));
 939         if((e=slabhash_lookup(worker->env.msg_cache, h, &qinfo, 0))) {
 940                 /* answer from cache - we have acquired a readlock on it */
 941                 if(answer_from_cache(worker, &qinfo,
 942                         (struct reply_info*)e->data,
 943                         *(uint16_t*)(void *)sldns_buffer_begin(c->buffer),
 944                         sldns_buffer_read_u16_at(c->buffer, 2), repinfo,
 945                         &edns)) {
 946                         /* prefetch it if the prefetch TTL expired */
 947                         if(worker->env.cfg->prefetch && *worker->env.now >=
 948                                 ((struct reply_info*)e->data)->prefetch_ttl) {
 949                                 time_t leeway = ((struct reply_info*)e->
 950                                         data)->ttl - *worker->env.now;
 951                                 lock_rw_unlock(&e->lock);
 952                                 reply_and_prefetch(worker, &qinfo,
 953                                         sldns_buffer_read_u16_at(c->buffer, 2),
 954                                         repinfo, leeway);
 955                                 rc = 0;
 956                                 goto send_reply_rc;
 957                         }
 958                         lock_rw_unlock(&e->lock);
 959                         goto send_reply;
 960                 }
 961                 verbose(VERB_ALGO, "answer from the cache failed");
 962                 lock_rw_unlock(&e->lock);
 963         }
 964         if(!LDNS_RD_WIRE(sldns_buffer_begin(c->buffer))) {
 965                 if(answer_norec_from_cache(worker, &qinfo,
 966                         *(uint16_t*)(void *)sldns_buffer_begin(c->buffer),
 967                         sldns_buffer_read_u16_at(c->buffer, 2), repinfo,
 968                         &edns)) {
 969                         goto send_reply;
 970                 }
 971                 verbose(VERB_ALGO, "answer norec from cache -- "
 972                         "need to validate or not primed");
 973         }
 974         sldns_buffer_rewind(c->buffer);
 975         server_stats_querymiss(&worker->stats, worker);
 976
 977         if(verbosity >= VERB_CLIENT) {
 978                 if(c->type == comm_udp)
 979                         log_addr(VERB_CLIENT, "udp request from",
 980                                 &repinfo->addr, repinfo->addrlen);
 981                 else    log_addr(VERB_CLIENT, "tcp request from",
 982                                 &repinfo->addr, repinfo->addrlen);
 983         }
 984
 985         /* grab a work request structure for this new request */
 986         mesh_new_client(worker->env.mesh, &qinfo,
 987                 sldns_buffer_read_u16_at(c->buffer, 2),
 988                 &edns, repinfo, *(uint16_t*)(void *)sldns_buffer_begin(c->buffer));
 989         worker_mem_report(worker, NULL);
 990         return 0;
 991
 992 send_reply:
 993         rc = 1;
 994 send_reply_rc:
 995 #ifdef USE_DNSTAP
 996         if(worker->dtenv.log_client_response_messages)
 997                 dt_msg_send_client_response(&worker->dtenv, &repinfo->addr,
 998                         c->type, c->buffer);
 999 #endif
1000         return rc;
1001 }
1002
1003 void
1004 worker_sighandler(int sig, void* arg)
1005 {
1006         /* note that log, print, syscalls here give race conditions.
1007          * And cause hangups if the log-lock is held by the application. */
1008         struct worker* worker = (struct worker*)arg;
1009         switch(sig) {
1010 #ifdef SIGHUP
1011                 case SIGHUP:
1012                         comm_base_exit(worker->base);
1013                         break;
1014 #endif
1015                 case SIGINT:
1016                         worker->need_to_exit = 1;
1017                         comm_base_exit(worker->base);
1018                         break;
1019 #ifdef SIGQUIT
1020                 case SIGQUIT:
1021                         worker->need_to_exit = 1;
1022                         comm_base_exit(worker->base);
1023                         break;
1024 #endif
1025                 case SIGTERM:
1026                         worker->need_to_exit = 1;
1027                         comm_base_exit(worker->base);
1028                         break;
1029                 default:
1030                         /* unknown signal, ignored */
1031                         break;
1032         }
1033 }
1034
1035 /** restart statistics timer for worker, if enabled */
1036 static void
1037 worker_restart_timer(struct worker* worker)
1038 {
1039         if(worker->env.cfg->stat_interval > 0) {
1040                 struct timeval tv;
1041 #ifndef S_SPLINT_S
1042                 tv.tv_sec = worker->env.cfg->stat_interval;
1043                 tv.tv_usec = 0;
1044 #endif
1045                 comm_timer_set(worker->stat_timer, &tv);
1046         }
1047 }
1048
1049 void worker_stat_timer_cb(void* arg)
1050 {
1051         struct worker* worker = (struct worker*)arg;
1052         server_stats_log(&worker->stats, worker, worker->thread_num);
1053         mesh_stats(worker->env.mesh, "mesh has");
1054         worker_mem_report(worker, NULL);
1055         if(!worker->daemon->cfg->stat_cumulative) {
1056                 worker_stats_clear(worker);
1057         }
1058         /* start next timer */
1059         worker_restart_timer(worker);
1060 }
1061
1062 void worker_probe_timer_cb(void* arg)
1063 {
1064         struct worker* worker = (struct worker*)arg;
1065         struct timeval tv;
1066 #ifndef S_SPLINT_S
1067         tv.tv_sec = (time_t)autr_probe_timer(&worker->env);
1068         tv.tv_usec = 0;
1069 #endif
1070         if(tv.tv_sec != 0)
1071                 comm_timer_set(worker->env.probe_timer, &tv);
1072 }
1073
1074 struct worker*
1075 worker_create(struct daemon* daemon, int id, int* ports, int n)
1076 {
1077         unsigned int seed;
1078         struct worker* worker = (struct worker*)calloc(1,
1079                 sizeof(struct worker));
1080         if(!worker)
1081                 return NULL;
1082         worker->numports = n;
1083         worker->ports = (int*)memdup(ports, sizeof(int)*n);
1084         if(!worker->ports) {
1085                 free(worker);
1086                 return NULL;
1087         }
1088         worker->daemon = daemon;
1089         worker->thread_num = id;
1090         if(!(worker->cmd = tube_create())) {
1091                 free(worker->ports);
1092                 free(worker);
1093                 return NULL;
1094         }
1095         /* create random state here to avoid locking trouble in RAND_bytes */
1096         seed = (unsigned int)time(NULL) ^ (unsigned int)getpid() ^
1097                 (((unsigned int)worker->thread_num)<<17);
1098                 /* shift thread_num so it does not match out pid bits */
1099         if(!(worker->rndstate = ub_initstate(seed, daemon->rand))) {
1100                 seed = 0;
1101                 log_err("could not init random numbers.");
1102                 tube_delete(worker->cmd);
1103                 free(worker->ports);
1104                 free(worker);
1105                 return NULL;
1106         }
1107         seed = 0;
1108 #ifdef USE_DNSTAP
1109         if(daemon->cfg->dnstap) {
1110                 log_assert(daemon->dtenv != NULL);
1111                 memcpy(&worker->dtenv, daemon->dtenv, sizeof(struct dt_env));
1112                 if(!dt_init(&worker->dtenv))
1113                         fatal_exit("dt_init failed");
1114         }
1115 #endif
1116         return worker;
1117 }
1118
1119 int
1120 worker_init(struct worker* worker, struct config_file *cfg,
1121         struct listen_port* ports, int do_sigs)
1122 {
1123 #ifdef USE_DNSTAP
1124         struct dt_env* dtenv = &worker->dtenv;
1125 #else
1126         void* dtenv = NULL;
1127 #endif
1128         worker->need_to_exit = 0;
1129         worker->base = comm_base_create(do_sigs);
1130         if(!worker->base) {
1131                 log_err("could not create event handling base");
1132                 worker_delete(worker);
1133                 return 0;
1134         }
1135         comm_base_set_slow_accept_handlers(worker->base, &worker_stop_accept,
1136                 &worker_start_accept, worker);
1137         if(do_sigs) {
1138 #ifdef SIGHUP
1139                 ub_thread_sig_unblock(SIGHUP);
1140 #endif
1141                 ub_thread_sig_unblock(SIGINT);
1142 #ifdef SIGQUIT
1143                 ub_thread_sig_unblock(SIGQUIT);
1144 #endif
1145                 ub_thread_sig_unblock(SIGTERM);
1146 #ifndef LIBEVENT_SIGNAL_PROBLEM
1147                 worker->comsig = comm_signal_create(worker->base,
1148                         worker_sighandler, worker);
1149                 if(!worker->comsig
1150 #ifdef SIGHUP
1151                         || !comm_signal_bind(worker->comsig, SIGHUP)
1152 #endif
1153 #ifdef SIGQUIT
1154                         || !comm_signal_bind(worker->comsig, SIGQUIT)
1155 #endif
1156                         || !comm_signal_bind(worker->comsig, SIGTERM)
1157                         || !comm_signal_bind(worker->comsig, SIGINT)) {
1158                         log_err("could not create signal handlers");
1159                         worker_delete(worker);
1160                         return 0;
1161                 }
1162 #endif /* LIBEVENT_SIGNAL_PROBLEM */
1163                 if(!daemon_remote_open_accept(worker->daemon->rc,
1164                         worker->daemon->rc_ports, worker)) {
1165                         worker_delete(worker);
1166                         return 0;
1167                 }
1168 #ifdef UB_ON_WINDOWS
1169                 wsvc_setup_worker(worker);
1170 #endif /* UB_ON_WINDOWS */
1171         } else { /* !do_sigs */
1172                 worker->comsig = NULL;
1173         }
1174         worker->front = listen_create(worker->base, ports,
1175                 cfg->msg_buffer_size, (int)cfg->incoming_num_tcp,
1176                 worker->daemon->listen_sslctx, dtenv, worker_handle_request,
1177                 worker);
1178         if(!worker->front) {
1179                 log_err("could not create listening sockets");
1180                 worker_delete(worker);
1181                 return 0;
1182         }
1183         worker->back = outside_network_create(worker->base,
1184                 cfg->msg_buffer_size, (size_t)cfg->outgoing_num_ports,
1185                 cfg->out_ifs, cfg->num_out_ifs, cfg->do_ip4, cfg->do_ip6,
1186                 cfg->do_tcp?cfg->outgoing_num_tcp:0,
1187                 worker->daemon->env->infra_cache, worker->rndstate,
1188                 cfg->use_caps_bits_for_id, worker->ports, worker->numports,
1189                 cfg->unwanted_threshold, &worker_alloc_cleanup, worker,
1190                 cfg->do_udp, worker->daemon->connect_sslctx, cfg->delay_close,
1191                 dtenv);
1192         if(!worker->back) {
1193                 log_err("could not create outgoing sockets");
1194                 worker_delete(worker);
1195                 return 0;
1196         }
1197         /* start listening to commands */
1198         if(!tube_setup_bg_listen(worker->cmd, worker->base,
1199                 &worker_handle_control_cmd, worker)) {
1200                 log_err("could not create control compt.");
1201                 worker_delete(worker);
1202                 return 0;
1203         }
1204         worker->stat_timer = comm_timer_create(worker->base,
1205                 worker_stat_timer_cb, worker);
1206         if(!worker->stat_timer) {
1207                 log_err("could not create statistics timer");
1208         }
1209
1210         /* we use the msg_buffer_size as a good estimate for what the
1211          * user wants for memory usage sizes */
1212         worker->scratchpad = regional_create_custom(cfg->msg_buffer_size);
1213         if(!worker->scratchpad) {
1214                 log_err("malloc failure");
1215                 worker_delete(worker);
1216                 return 0;
1217         }
1218
1219         server_stats_init(&worker->stats, cfg);
1220         alloc_init(&worker->alloc, &worker->daemon->superalloc,
1221                 worker->thread_num);
1222         alloc_set_id_cleanup(&worker->alloc, &worker_alloc_cleanup, worker);
1223         worker->env = *worker->daemon->env;
1224         comm_base_timept(worker->base, &worker->env.now, &worker->env.now_tv);
1225         if(worker->thread_num == 0)
1226                 log_set_time(worker->env.now);
1227         worker->env.worker = worker;
1228         worker->env.send_query = &worker_send_query;
1229         worker->env.alloc = &worker->alloc;
1230         worker->env.rnd = worker->rndstate;
1231         worker->env.scratch = worker->scratchpad;
1232         worker->env.mesh = mesh_create(&worker->daemon->mods, &worker->env);
1233         worker->env.detach_subs = &mesh_detach_subs;
1234         worker->env.attach_sub = &mesh_attach_sub;
1235         worker->env.kill_sub = &mesh_state_delete;
1236         worker->env.detect_cycle = &mesh_detect_cycle;
1237         worker->env.scratch_buffer = sldns_buffer_new(cfg->msg_buffer_size);
1238         if(!(worker->env.fwds = forwards_create()) ||
1239                 !forwards_apply_cfg(worker->env.fwds, cfg)) {
1240                 log_err("Could not set forward zones");
1241                 worker_delete(worker);
1242                 return 0;
1243         }
1244         if(!(worker->env.hints = hints_create()) ||
1245                 !hints_apply_cfg(worker->env.hints, cfg)) {
1246                 log_err("Could not set root or stub hints");
1247                 worker_delete(worker);
1248                 return 0;
1249         }
1250         /* one probe timer per process -- if we have 5011 anchors */
1251         if(autr_get_num_anchors(worker->env.anchors) > 0
1252 #ifndef THREADS_DISABLED
1253                 && worker->thread_num == 0
1254 #endif
1255                 ) {
1256                 struct timeval tv;
1257                 tv.tv_sec = 0;
1258                 tv.tv_usec = 0;
1259                 worker->env.probe_timer = comm_timer_create(worker->base,
1260                         worker_probe_timer_cb, worker);
1261                 if(!worker->env.probe_timer) {
1262                         log_err("could not create 5011-probe timer");
1263                 } else {
1264                         /* let timer fire, then it can reset itself */
1265                         comm_timer_set(worker->env.probe_timer, &tv);
1266                 }
1267         }
1268         if(!worker->env.mesh || !worker->env.scratch_buffer) {
1269                 worker_delete(worker);
1270                 return 0;
1271         }
1272         worker_mem_report(worker, NULL);
1273         /* if statistics enabled start timer */
1274         if(worker->env.cfg->stat_interval > 0) {
1275                 verbose(VERB_ALGO, "set statistics interval %d secs",
1276                         worker->env.cfg->stat_interval);
1277                 worker_restart_timer(worker);
1278         }
1279         return 1;
1280 }
1281
1282 void
1283 worker_work(struct worker* worker)
1284 {
1285         comm_base_dispatch(worker->base);
1286 }
1287
1288 void
1289 worker_delete(struct worker* worker)
1290 {
1291         if(!worker)
1292                 return;
1293         if(worker->env.mesh && verbosity >= VERB_OPS) {
1294                 server_stats_log(&worker->stats, worker, worker->thread_num);
1295                 mesh_stats(worker->env.mesh, "mesh has");
1296                 worker_mem_report(worker, NULL);
1297         }
1298         outside_network_quit_prepare(worker->back);
1299         mesh_delete(worker->env.mesh);
1300         sldns_buffer_free(worker->env.scratch_buffer);
1301         forwards_delete(worker->env.fwds);
1302         hints_delete(worker->env.hints);
1303         listen_delete(worker->front);
1304         outside_network_delete(worker->back);
1305         comm_signal_delete(worker->comsig);
1306         tube_delete(worker->cmd);
1307         comm_timer_delete(worker->stat_timer);
1308         comm_timer_delete(worker->env.probe_timer);
1309         free(worker->ports);
1310         if(worker->thread_num == 0) {
1311                 log_set_time(NULL);
1312 #ifdef UB_ON_WINDOWS
1313                 wsvc_desetup_worker(worker);
1314 #endif /* UB_ON_WINDOWS */
1315         }
1316         comm_base_delete(worker->base);
1317         ub_randfree(worker->rndstate);
1318         alloc_clear(&worker->alloc);
1319         regional_destroy(worker->scratchpad);
1320         free(worker);
1321 }
1322
1323 struct outbound_entry*
1324 worker_send_query(uint8_t* qname, size_t qnamelen, uint16_t qtype,
1325         uint16_t qclass, uint16_t flags, int dnssec, int want_dnssec,
1326         int nocaps, struct sockaddr_storage* addr, socklen_t addrlen,
1327         uint8_t* zone, size_t zonelen, struct module_qstate* q)
1328 {
1329         struct worker* worker = q->env->worker;
1330         struct outbound_entry* e = (struct outbound_entry*)regional_alloc(
1331                 q->region, sizeof(*e));
1332         if(!e)
1333                 return NULL;
1334         e->qstate = q;
1335         e->qsent = outnet_serviced_query(worker->back, qname,
1336                 qnamelen, qtype, qclass, flags, dnssec, want_dnssec, nocaps,
1337                 q->env->cfg->tcp_upstream, q->env->cfg->ssl_upstream, addr,
1338                 addrlen, zone, zonelen, worker_handle_service_reply, e,
1339                 worker->back->udp_buff);
1340         if(!e->qsent) {
1341                 return NULL;
1342         }
1343         return e;
1344 }
1345
1346 void
1347 worker_alloc_cleanup(void* arg)
1348 {
1349         struct worker* worker = (struct worker*)arg;
1350         slabhash_clear(&worker->env.rrset_cache->table);
1351         slabhash_clear(worker->env.msg_cache);
1352 }
1353
1354 void worker_stats_clear(struct worker* worker)
1355 {
1356         server_stats_init(&worker->stats, worker->env.cfg);
1357         mesh_stats_clear(worker->env.mesh);
1358         worker->back->unwanted_replies = 0;
1359         worker->back->num_tcp_outgoing = 0;
1360 }
1361
1362 void worker_start_accept(void* arg)
1363 {
1364         struct worker* worker = (struct worker*)arg;
1365         listen_start_accept(worker->front);
1366         if(worker->thread_num == 0)
1367                 daemon_remote_start_accept(worker->daemon->rc);
1368 }
1369
1370 void worker_stop_accept(void* arg)
1371 {
1372         struct worker* worker = (struct worker*)arg;
1373         listen_stop_accept(worker->front);
1374         if(worker->thread_num == 0)
1375                 daemon_remote_stop_accept(worker->daemon->rc);
1376 }
1377
1378 /* --- fake callbacks for fptr_wlist to work --- */
1379 struct outbound_entry* libworker_send_query(uint8_t* ATTR_UNUSED(qname),
1380         size_t ATTR_UNUSED(qnamelen), uint16_t ATTR_UNUSED(qtype),
1381         uint16_t ATTR_UNUSED(qclass), uint16_t ATTR_UNUSED(flags),
1382         int ATTR_UNUSED(dnssec), int ATTR_UNUSED(want_dnssec),
1383         int ATTR_UNUSED(nocaps), struct sockaddr_storage* ATTR_UNUSED(addr),
1384         socklen_t ATTR_UNUSED(addrlen), uint8_t* ATTR_UNUSED(zone),
1385         size_t ATTR_UNUSED(zonelen), struct module_qstate* ATTR_UNUSED(q))
1386 {
1387         log_assert(0);
1388         return 0;
1389 }
1390
1391 int libworker_handle_reply(struct comm_point* ATTR_UNUSED(c),
1392         void* ATTR_UNUSED(arg), int ATTR_UNUSED(error),
1393         struct comm_reply* ATTR_UNUSED(reply_info))
1394 {
1395         log_assert(0);
1396         return 0;
1397 }
1398
1399 int libworker_handle_service_reply(struct comm_point* ATTR_UNUSED(c),
1400         void* ATTR_UNUSED(arg), int ATTR_UNUSED(error),
1401         struct comm_reply* ATTR_UNUSED(reply_info))
1402 {
1403         log_assert(0);
1404         return 0;
1405 }
1406
1407 void libworker_handle_control_cmd(struct tube* ATTR_UNUSED(tube),
1408         uint8_t* ATTR_UNUSED(buffer), size_t ATTR_UNUSED(len),
1409         int ATTR_UNUSED(error), void* ATTR_UNUSED(arg))
1410 {
1411         log_assert(0);
1412 }
1413
1414 void libworker_fg_done_cb(void* ATTR_UNUSED(arg), int ATTR_UNUSED(rcode),
1415         sldns_buffer* ATTR_UNUSED(buf), enum sec_status ATTR_UNUSED(s),
1416         char* ATTR_UNUSED(why_bogus))
1417 {
1418         log_assert(0);
1419 }
1420
1421 void libworker_bg_done_cb(void* ATTR_UNUSED(arg), int ATTR_UNUSED(rcode),
1422         sldns_buffer* ATTR_UNUSED(buf), enum sec_status ATTR_UNUSED(s),
1423         char* ATTR_UNUSED(why_bogus))
1424 {
1425         log_assert(0);
1426 }
1427
1428 void libworker_event_done_cb(void* ATTR_UNUSED(arg), int ATTR_UNUSED(rcode),
1429         sldns_buffer* ATTR_UNUSED(buf), enum sec_status ATTR_UNUSED(s),
1430         char* ATTR_UNUSED(why_bogus))
1431 {
1432         log_assert(0);
1433 }
1434
1435 int context_query_cmp(const void* ATTR_UNUSED(a), const void* ATTR_UNUSED(b))
1436 {
1437         log_assert(0);
1438         return 0;
1439 }
1440
1441 int order_lock_cmp(const void* ATTR_UNUSED(e1), const void* ATTR_UNUSED(e2))
1442 {
1443         log_assert(0);
1444         return 0;
1445 }
1446
1447 int codeline_cmp(const void* ATTR_UNUSED(a), const void* ATTR_UNUSED(b))
1448 {
1449         log_assert(0);
1450         return 0;
1451 }
1452