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 "sldns/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 /** ratelimit for error responses */ 90 #define ERROR_RATELIMIT 100 /* qps */ 91 92 /** 93 * seconds to add to prefetch leeway. This is a TTL that expires old rrsets 94 * earlier than they should in order to put the new update into the cache. 95 * This additional value is to make sure that if not all TTLs are equal in 96 * the message to be updated(and replaced), that rrsets with up to this much 97 * extra TTL are also replaced. This means that the resulting new message 98 * will have (most likely) this TTL at least, avoiding very small 'split 99 * second' TTLs due to operators choosing relative primes for TTLs (or so). 100 * Also has to be at least one to break ties (and overwrite cached entry). 101 */ 102 #define PREFETCH_EXPIRY_ADD 60 103 104 #ifdef UNBOUND_ALLOC_STATS 105 /** measure memory leakage */ 106 static void 107 debug_memleak(size_t accounted, size_t heap, 108 size_t total_alloc, size_t total_free) 109 { 110 static int init = 0; 111 static size_t base_heap, base_accounted, base_alloc, base_free; 112 size_t base_af, cur_af, grow_af, grow_acc; 113 if(!init) { 114 init = 1; 115 base_heap = heap; 116 base_accounted = accounted; 117 base_alloc = total_alloc; 118 base_free = total_free; 119 } 120 base_af = base_alloc - base_free; 121 cur_af = total_alloc - total_free; 122 grow_af = cur_af - base_af; 123 grow_acc = accounted - base_accounted; 124 log_info("Leakage: %d leaked. growth: %u use, %u acc, %u heap", 125 (int)(grow_af - grow_acc), (unsigned)grow_af, 126 (unsigned)grow_acc, (unsigned)(heap - base_heap)); 127 } 128 129 /** give debug heap size indication */ 130 static void 131 debug_total_mem(size_t calctotal) 132 { 133 #ifdef HAVE_SBRK 134 extern void* unbound_start_brk; 135 extern size_t unbound_mem_alloc, unbound_mem_freed; 136 void* cur = sbrk(0); 137 int total = cur-unbound_start_brk; 138 log_info("Total heap memory estimate: %u total-alloc: %u " 139 "total-free: %u", (unsigned)total, 140 (unsigned)unbound_mem_alloc, (unsigned)unbound_mem_freed); 141 debug_memleak(calctotal, (size_t)total, 142 unbound_mem_alloc, unbound_mem_freed); 143 #else 144 (void)calctotal; 145 #endif /* HAVE_SBRK */ 146 } 147 #endif /* UNBOUND_ALLOC_STATS */ 148 149 /** Report on memory usage by this thread and global */ 150 static void 151 worker_mem_report(struct worker* ATTR_UNUSED(worker), 152 struct serviced_query* ATTR_UNUSED(cur_serv)) 153 { 154 #ifdef UNBOUND_ALLOC_STATS 155 /* debug func in validator module */ 156 size_t total, front, back, mesh, msg, rrset, infra, ac, superac; 157 size_t me, iter, val, anch; 158 int i; 159 if(verbosity < VERB_ALGO) 160 return; 161 front = listen_get_mem(worker->front); 162 back = outnet_get_mem(worker->back); 163 msg = slabhash_get_mem(worker->env.msg_cache); 164 rrset = slabhash_get_mem(&worker->env.rrset_cache->table); 165 infra = infra_get_mem(worker->env.infra_cache); 166 mesh = mesh_get_mem(worker->env.mesh); 167 ac = alloc_get_mem(&worker->alloc); 168 superac = alloc_get_mem(&worker->daemon->superalloc); 169 anch = anchors_get_mem(worker->env.anchors); 170 iter = 0; 171 val = 0; 172 for(i=0; i<worker->env.mesh->mods.num; i++) { 173 fptr_ok(fptr_whitelist_mod_get_mem(worker->env.mesh-> 174 mods.mod[i]->get_mem)); 175 if(strcmp(worker->env.mesh->mods.mod[i]->name, "validator")==0) 176 val += (*worker->env.mesh->mods.mod[i]->get_mem) 177 (&worker->env, i); 178 else iter += (*worker->env.mesh->mods.mod[i]->get_mem) 179 (&worker->env, i); 180 } 181 me = sizeof(*worker) + sizeof(*worker->base) + sizeof(*worker->comsig) 182 + comm_point_get_mem(worker->cmd_com) 183 + sizeof(worker->rndstate) 184 + regional_get_mem(worker->scratchpad) 185 + sizeof(*worker->env.scratch_buffer) 186 + sldns_buffer_capacity(worker->env.scratch_buffer) 187 + forwards_get_mem(worker->env.fwds) 188 + hints_get_mem(worker->env.hints); 189 if(worker->thread_num == 0) 190 me += acl_list_get_mem(worker->daemon->acl); 191 if(cur_serv) { 192 me += serviced_get_mem(cur_serv); 193 } 194 total = front+back+mesh+msg+rrset+infra+iter+val+ac+superac+me; 195 log_info("Memory conditions: %u front=%u back=%u mesh=%u msg=%u " 196 "rrset=%u infra=%u iter=%u val=%u anchors=%u " 197 "alloccache=%u globalalloccache=%u me=%u", 198 (unsigned)total, (unsigned)front, (unsigned)back, 199 (unsigned)mesh, (unsigned)msg, (unsigned)rrset, 200 (unsigned)infra, (unsigned)iter, (unsigned)val, (unsigned)anch, 201 (unsigned)ac, (unsigned)superac, (unsigned)me); 202 debug_total_mem(total); 203 #else /* no UNBOUND_ALLOC_STATS */ 204 size_t val = 0; 205 int i; 206 if(verbosity < VERB_QUERY) 207 return; 208 for(i=0; i<worker->env.mesh->mods.num; i++) { 209 fptr_ok(fptr_whitelist_mod_get_mem(worker->env.mesh-> 210 mods.mod[i]->get_mem)); 211 if(strcmp(worker->env.mesh->mods.mod[i]->name, "validator")==0) 212 val += (*worker->env.mesh->mods.mod[i]->get_mem) 213 (&worker->env, i); 214 } 215 verbose(VERB_QUERY, "cache memory msg=%u rrset=%u infra=%u val=%u", 216 (unsigned)slabhash_get_mem(worker->env.msg_cache), 217 (unsigned)slabhash_get_mem(&worker->env.rrset_cache->table), 218 (unsigned)infra_get_mem(worker->env.infra_cache), 219 (unsigned)val); 220 #endif /* UNBOUND_ALLOC_STATS */ 221 } 222 223 void 224 worker_send_cmd(struct worker* worker, enum worker_commands cmd) 225 { 226 uint32_t c = (uint32_t)htonl(cmd); 227 if(!tube_write_msg(worker->cmd, (uint8_t*)&c, sizeof(c), 0)) { 228 log_err("worker send cmd %d failed", (int)cmd); 229 } 230 } 231 232 int 233 worker_handle_reply(struct comm_point* c, void* arg, int error, 234 struct comm_reply* reply_info) 235 { 236 struct module_qstate* q = (struct module_qstate*)arg; 237 struct worker* worker = q->env->worker; 238 struct outbound_entry e; 239 e.qstate = q; 240 e.qsent = NULL; 241 242 if(error != 0) { 243 mesh_report_reply(worker->env.mesh, &e, reply_info, error); 244 worker_mem_report(worker, NULL); 245 return 0; 246 } 247 /* sanity check. */ 248 if(!LDNS_QR_WIRE(sldns_buffer_begin(c->buffer)) 249 || LDNS_OPCODE_WIRE(sldns_buffer_begin(c->buffer)) != 250 LDNS_PACKET_QUERY 251 || LDNS_QDCOUNT(sldns_buffer_begin(c->buffer)) > 1) { 252 /* error becomes timeout for the module as if this reply 253 * never arrived. */ 254 mesh_report_reply(worker->env.mesh, &e, reply_info, 255 NETEVENT_TIMEOUT); 256 worker_mem_report(worker, NULL); 257 return 0; 258 } 259 mesh_report_reply(worker->env.mesh, &e, reply_info, NETEVENT_NOERROR); 260 worker_mem_report(worker, NULL); 261 return 0; 262 } 263 264 int 265 worker_handle_service_reply(struct comm_point* c, void* arg, int error, 266 struct comm_reply* reply_info) 267 { 268 struct outbound_entry* e = (struct outbound_entry*)arg; 269 struct worker* worker = e->qstate->env->worker; 270 struct serviced_query *sq = e->qsent; 271 272 verbose(VERB_ALGO, "worker svcd callback for qstate %p", e->qstate); 273 if(error != 0) { 274 mesh_report_reply(worker->env.mesh, e, reply_info, error); 275 worker_mem_report(worker, sq); 276 return 0; 277 } 278 /* sanity check. */ 279 if(!LDNS_QR_WIRE(sldns_buffer_begin(c->buffer)) 280 || LDNS_OPCODE_WIRE(sldns_buffer_begin(c->buffer)) != 281 LDNS_PACKET_QUERY 282 || LDNS_QDCOUNT(sldns_buffer_begin(c->buffer)) > 1) { 283 /* error becomes timeout for the module as if this reply 284 * never arrived. */ 285 verbose(VERB_ALGO, "worker: bad reply handled as timeout"); 286 mesh_report_reply(worker->env.mesh, e, reply_info, 287 NETEVENT_TIMEOUT); 288 worker_mem_report(worker, sq); 289 return 0; 290 } 291 mesh_report_reply(worker->env.mesh, e, reply_info, NETEVENT_NOERROR); 292 worker_mem_report(worker, sq); 293 return 0; 294 } 295 296 /** ratelimit error replies 297 * @param worker: the worker struct with ratelimit counter 298 * @param err: error code that would be wanted. 299 * @return value of err if okay, or -1 if it should be discarded instead. 300 */ 301 static int 302 worker_err_ratelimit(struct worker* worker, int err) 303 { 304 if(worker->err_limit_time == *worker->env.now) { 305 /* see if limit is exceeded for this second */ 306 if(worker->err_limit_count++ > ERROR_RATELIMIT) 307 return -1; 308 } else { 309 /* new second, new limits */ 310 worker->err_limit_time = *worker->env.now; 311 worker->err_limit_count = 1; 312 } 313 return err; 314 } 315 316 /** check request sanity. 317 * @param pkt: the wire packet to examine for sanity. 318 * @param worker: parameters for checking. 319 * @return error code, 0 OK, or -1 discard. 320 */ 321 static int 322 worker_check_request(sldns_buffer* pkt, struct worker* worker) 323 { 324 if(sldns_buffer_limit(pkt) < LDNS_HEADER_SIZE) { 325 verbose(VERB_QUERY, "request too short, discarded"); 326 return -1; 327 } 328 if(sldns_buffer_limit(pkt) > NORMAL_UDP_SIZE && 329 worker->daemon->cfg->harden_large_queries) { 330 verbose(VERB_QUERY, "request too large, discarded"); 331 return -1; 332 } 333 if(LDNS_QR_WIRE(sldns_buffer_begin(pkt))) { 334 verbose(VERB_QUERY, "request has QR bit on, discarded"); 335 return -1; 336 } 337 if(LDNS_TC_WIRE(sldns_buffer_begin(pkt))) { 338 LDNS_TC_CLR(sldns_buffer_begin(pkt)); 339 verbose(VERB_QUERY, "request bad, has TC bit on"); 340 return worker_err_ratelimit(worker, LDNS_RCODE_FORMERR); 341 } 342 if(LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt)) != LDNS_PACKET_QUERY) { 343 verbose(VERB_QUERY, "request unknown opcode %d", 344 LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt))); 345 return worker_err_ratelimit(worker, LDNS_RCODE_NOTIMPL); 346 } 347 if(LDNS_QDCOUNT(sldns_buffer_begin(pkt)) != 1) { 348 verbose(VERB_QUERY, "request wrong nr qd=%d", 349 LDNS_QDCOUNT(sldns_buffer_begin(pkt))); 350 return worker_err_ratelimit(worker, LDNS_RCODE_FORMERR); 351 } 352 if(LDNS_ANCOUNT(sldns_buffer_begin(pkt)) != 0) { 353 verbose(VERB_QUERY, "request wrong nr an=%d", 354 LDNS_ANCOUNT(sldns_buffer_begin(pkt))); 355 return worker_err_ratelimit(worker, LDNS_RCODE_FORMERR); 356 } 357 if(LDNS_NSCOUNT(sldns_buffer_begin(pkt)) != 0) { 358 verbose(VERB_QUERY, "request wrong nr ns=%d", 359 LDNS_NSCOUNT(sldns_buffer_begin(pkt))); 360 return worker_err_ratelimit(worker, LDNS_RCODE_FORMERR); 361 } 362 if(LDNS_ARCOUNT(sldns_buffer_begin(pkt)) > 1) { 363 verbose(VERB_QUERY, "request wrong nr ar=%d", 364 LDNS_ARCOUNT(sldns_buffer_begin(pkt))); 365 return worker_err_ratelimit(worker, LDNS_RCODE_FORMERR); 366 } 367 return 0; 368 } 369 370 void 371 worker_handle_control_cmd(struct tube* ATTR_UNUSED(tube), uint8_t* msg, 372 size_t len, int error, void* arg) 373 { 374 struct worker* worker = (struct worker*)arg; 375 enum worker_commands cmd; 376 if(error != NETEVENT_NOERROR) { 377 free(msg); 378 if(error == NETEVENT_CLOSED) 379 comm_base_exit(worker->base); 380 else log_info("control event: %d", error); 381 return; 382 } 383 if(len != sizeof(uint32_t)) { 384 fatal_exit("bad control msg length %d", (int)len); 385 } 386 cmd = sldns_read_uint32(msg); 387 free(msg); 388 switch(cmd) { 389 case worker_cmd_quit: 390 verbose(VERB_ALGO, "got control cmd quit"); 391 comm_base_exit(worker->base); 392 break; 393 case worker_cmd_stats: 394 verbose(VERB_ALGO, "got control cmd stats"); 395 server_stats_reply(worker, 1); 396 break; 397 case worker_cmd_stats_noreset: 398 verbose(VERB_ALGO, "got control cmd stats_noreset"); 399 server_stats_reply(worker, 0); 400 break; 401 case worker_cmd_remote: 402 verbose(VERB_ALGO, "got control cmd remote"); 403 daemon_remote_exec(worker); 404 break; 405 default: 406 log_err("bad command %d", (int)cmd); 407 break; 408 } 409 } 410 411 /** check if a delegation is secure */ 412 static enum sec_status 413 check_delegation_secure(struct reply_info *rep) 414 { 415 /* return smallest security status */ 416 size_t i; 417 enum sec_status sec = sec_status_secure; 418 enum sec_status s; 419 size_t num = rep->an_numrrsets + rep->ns_numrrsets; 420 /* check if answer and authority are OK */ 421 for(i=0; i<num; i++) { 422 s = ((struct packed_rrset_data*)rep->rrsets[i]->entry.data) 423 ->security; 424 if(s < sec) 425 sec = s; 426 } 427 /* in additional, only unchecked triggers revalidation */ 428 for(i=num; i<rep->rrset_count; i++) { 429 s = ((struct packed_rrset_data*)rep->rrsets[i]->entry.data) 430 ->security; 431 if(s == sec_status_unchecked) 432 return s; 433 } 434 return sec; 435 } 436 437 /** remove nonsecure from a delegation referral additional section */ 438 static void 439 deleg_remove_nonsecure_additional(struct reply_info* rep) 440 { 441 /* we can simply edit it, since we are working in the scratch region */ 442 size_t i; 443 enum sec_status s; 444 445 for(i = rep->an_numrrsets+rep->ns_numrrsets; i<rep->rrset_count; i++) { 446 s = ((struct packed_rrset_data*)rep->rrsets[i]->entry.data) 447 ->security; 448 if(s != sec_status_secure) { 449 memmove(rep->rrsets+i, rep->rrsets+i+1, 450 sizeof(struct ub_packed_rrset_key*)* 451 (rep->rrset_count - i - 1)); 452 rep->ar_numrrsets--; 453 rep->rrset_count--; 454 i--; 455 } 456 } 457 } 458 459 /** answer nonrecursive query from the cache */ 460 static int 461 answer_norec_from_cache(struct worker* worker, struct query_info* qinfo, 462 uint16_t id, uint16_t flags, struct comm_reply* repinfo, 463 struct edns_data* edns) 464 { 465 /* for a nonrecursive query return either: 466 * o an error (servfail; we try to avoid this) 467 * o a delegation (closest we have; this routine tries that) 468 * o the answer (checked by answer_from_cache) 469 * 470 * So, grab a delegation from the rrset cache. 471 * Then check if it needs validation, if so, this routine fails, 472 * so that iterator can prime and validator can verify rrsets. 473 */ 474 uint16_t udpsize = edns->udp_size; 475 int secure = 0; 476 time_t timenow = *worker->env.now; 477 int must_validate = (!(flags&BIT_CD) || worker->env.cfg->ignore_cd) 478 && worker->env.need_to_validate; 479 struct dns_msg *msg = NULL; 480 struct delegpt *dp; 481 482 dp = dns_cache_find_delegation(&worker->env, qinfo->qname, 483 qinfo->qname_len, qinfo->qtype, qinfo->qclass, 484 worker->scratchpad, &msg, timenow); 485 if(!dp) { /* no delegation, need to reprime */ 486 return 0; 487 } 488 if(must_validate) { 489 switch(check_delegation_secure(msg->rep)) { 490 case sec_status_unchecked: 491 /* some rrsets have not been verified yet, go and 492 * let validator do that */ 493 return 0; 494 case sec_status_bogus: 495 /* some rrsets are bogus, reply servfail */ 496 edns->edns_version = EDNS_ADVERTISED_VERSION; 497 edns->udp_size = EDNS_ADVERTISED_SIZE; 498 edns->ext_rcode = 0; 499 edns->bits &= EDNS_DO; 500 if(!edns_opt_inplace_reply(edns, worker->scratchpad)) 501 return 0; 502 error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL, 503 &msg->qinfo, id, flags, edns); 504 if(worker->stats.extended) { 505 worker->stats.ans_bogus++; 506 worker->stats.ans_rcode[LDNS_RCODE_SERVFAIL]++; 507 } 508 return 1; 509 case sec_status_secure: 510 /* all rrsets are secure */ 511 /* remove non-secure rrsets from the add. section*/ 512 if(worker->env.cfg->val_clean_additional) 513 deleg_remove_nonsecure_additional(msg->rep); 514 secure = 1; 515 break; 516 case sec_status_indeterminate: 517 case sec_status_insecure: 518 default: 519 /* not secure */ 520 secure = 0; 521 break; 522 } 523 } 524 /* return this delegation from the cache */ 525 edns->edns_version = EDNS_ADVERTISED_VERSION; 526 edns->udp_size = EDNS_ADVERTISED_SIZE; 527 edns->ext_rcode = 0; 528 edns->bits &= EDNS_DO; 529 if(!edns_opt_inplace_reply(edns, worker->scratchpad)) 530 return 0; 531 msg->rep->flags |= BIT_QR|BIT_RA; 532 if(!reply_info_answer_encode(&msg->qinfo, msg->rep, id, flags, 533 repinfo->c->buffer, 0, 1, worker->scratchpad, 534 udpsize, edns, (int)(edns->bits & EDNS_DO), secure)) { 535 error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL, 536 &msg->qinfo, id, flags, edns); 537 } 538 if(worker->stats.extended) { 539 if(secure) worker->stats.ans_secure++; 540 server_stats_insrcode(&worker->stats, repinfo->c->buffer); 541 } 542 return 1; 543 } 544 545 /** answer query from the cache */ 546 static int 547 answer_from_cache(struct worker* worker, struct query_info* qinfo, 548 struct reply_info* rep, uint16_t id, uint16_t flags, 549 struct comm_reply* repinfo, struct edns_data* edns) 550 { 551 time_t timenow = *worker->env.now; 552 uint16_t udpsize = edns->udp_size; 553 int secure; 554 int must_validate = (!(flags&BIT_CD) || worker->env.cfg->ignore_cd) 555 && worker->env.need_to_validate; 556 /* see if it is possible */ 557 if(rep->ttl < timenow) { 558 /* the rrsets may have been updated in the meantime. 559 * we will refetch the message format from the 560 * authoritative server 561 */ 562 return 0; 563 } 564 if(!rrset_array_lock(rep->ref, rep->rrset_count, timenow)) 565 return 0; 566 /* locked and ids and ttls are OK. */ 567 /* check CNAME chain (if any) */ 568 if(rep->an_numrrsets > 0 && (rep->rrsets[0]->rk.type == 569 htons(LDNS_RR_TYPE_CNAME) || rep->rrsets[0]->rk.type == 570 htons(LDNS_RR_TYPE_DNAME))) { 571 if(!reply_check_cname_chain(qinfo, rep)) { 572 /* cname chain invalid, redo iterator steps */ 573 verbose(VERB_ALGO, "Cache reply: cname chain broken"); 574 bail_out: 575 rrset_array_unlock_touch(worker->env.rrset_cache, 576 worker->scratchpad, rep->ref, rep->rrset_count); 577 return 0; 578 } 579 } 580 /* check security status of the cached answer */ 581 if( rep->security == sec_status_bogus && must_validate) { 582 /* BAD cached */ 583 edns->edns_version = EDNS_ADVERTISED_VERSION; 584 edns->udp_size = EDNS_ADVERTISED_SIZE; 585 edns->ext_rcode = 0; 586 edns->bits &= EDNS_DO; 587 if(!edns_opt_inplace_reply(edns, worker->scratchpad)) 588 return 0; 589 error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL, 590 qinfo, id, flags, edns); 591 rrset_array_unlock_touch(worker->env.rrset_cache, 592 worker->scratchpad, rep->ref, rep->rrset_count); 593 if(worker->stats.extended) { 594 worker->stats.ans_bogus ++; 595 worker->stats.ans_rcode[LDNS_RCODE_SERVFAIL] ++; 596 } 597 return 1; 598 } else if( rep->security == sec_status_unchecked && must_validate) { 599 verbose(VERB_ALGO, "Cache reply: unchecked entry needs " 600 "validation"); 601 goto bail_out; /* need to validate cache entry first */ 602 } else if(rep->security == sec_status_secure) { 603 if(reply_all_rrsets_secure(rep)) 604 secure = 1; 605 else { 606 if(must_validate) { 607 verbose(VERB_ALGO, "Cache reply: secure entry" 608 " changed status"); 609 goto bail_out; /* rrset changed, re-verify */ 610 } 611 secure = 0; 612 } 613 } else secure = 0; 614 615 edns->edns_version = EDNS_ADVERTISED_VERSION; 616 edns->udp_size = EDNS_ADVERTISED_SIZE; 617 edns->ext_rcode = 0; 618 edns->bits &= EDNS_DO; 619 if(!edns_opt_inplace_reply(edns, worker->scratchpad)) 620 return 0; 621 if(!reply_info_answer_encode(qinfo, rep, id, flags, 622 repinfo->c->buffer, timenow, 1, worker->scratchpad, 623 udpsize, edns, (int)(edns->bits & EDNS_DO), secure)) { 624 error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL, 625 qinfo, id, flags, edns); 626 } 627 /* cannot send the reply right now, because blocking network syscall 628 * is bad while holding locks. */ 629 rrset_array_unlock_touch(worker->env.rrset_cache, worker->scratchpad, 630 rep->ref, rep->rrset_count); 631 if(worker->stats.extended) { 632 if(secure) worker->stats.ans_secure++; 633 server_stats_insrcode(&worker->stats, repinfo->c->buffer); 634 } 635 /* go and return this buffer to the client */ 636 return 1; 637 } 638 639 /** Reply to client and perform prefetch to keep cache up to date */ 640 static void 641 reply_and_prefetch(struct worker* worker, struct query_info* qinfo, 642 uint16_t flags, struct comm_reply* repinfo, time_t leeway) 643 { 644 /* first send answer to client to keep its latency 645 * as small as a cachereply */ 646 comm_point_send_reply(repinfo); 647 server_stats_prefetch(&worker->stats, worker); 648 649 /* create the prefetch in the mesh as a normal lookup without 650 * client addrs waiting, which has the cache blacklisted (to bypass 651 * the cache and go to the network for the data). */ 652 /* this (potentially) runs the mesh for the new query */ 653 mesh_new_prefetch(worker->env.mesh, qinfo, flags, leeway + 654 PREFETCH_EXPIRY_ADD); 655 } 656 657 /** 658 * Fill CH class answer into buffer. Keeps query. 659 * @param pkt: buffer 660 * @param str: string to put into text record (<255). 661 * @param edns: edns reply information. 662 * @param worker: worker with scratch region. 663 */ 664 static void 665 chaos_replystr(sldns_buffer* pkt, const char* str, struct edns_data* edns, 666 struct worker* worker) 667 { 668 size_t len = strlen(str); 669 unsigned int rd = LDNS_RD_WIRE(sldns_buffer_begin(pkt)); 670 unsigned int cd = LDNS_CD_WIRE(sldns_buffer_begin(pkt)); 671 if(len>255) len=255; /* cap size of TXT record */ 672 sldns_buffer_clear(pkt); 673 sldns_buffer_skip(pkt, (ssize_t)sizeof(uint16_t)); /* skip id */ 674 sldns_buffer_write_u16(pkt, (uint16_t)(BIT_QR|BIT_RA)); 675 if(rd) LDNS_RD_SET(sldns_buffer_begin(pkt)); 676 if(cd) LDNS_CD_SET(sldns_buffer_begin(pkt)); 677 sldns_buffer_write_u16(pkt, 1); /* qdcount */ 678 sldns_buffer_write_u16(pkt, 1); /* ancount */ 679 sldns_buffer_write_u16(pkt, 0); /* nscount */ 680 sldns_buffer_write_u16(pkt, 0); /* arcount */ 681 (void)query_dname_len(pkt); /* skip qname */ 682 sldns_buffer_skip(pkt, (ssize_t)sizeof(uint16_t)); /* skip qtype */ 683 sldns_buffer_skip(pkt, (ssize_t)sizeof(uint16_t)); /* skip qclass */ 684 sldns_buffer_write_u16(pkt, 0xc00c); /* compr ptr to query */ 685 sldns_buffer_write_u16(pkt, LDNS_RR_TYPE_TXT); 686 sldns_buffer_write_u16(pkt, LDNS_RR_CLASS_CH); 687 sldns_buffer_write_u32(pkt, 0); /* TTL */ 688 sldns_buffer_write_u16(pkt, sizeof(uint8_t) + len); 689 sldns_buffer_write_u8(pkt, len); 690 sldns_buffer_write(pkt, str, len); 691 sldns_buffer_flip(pkt); 692 edns->edns_version = EDNS_ADVERTISED_VERSION; 693 edns->udp_size = EDNS_ADVERTISED_SIZE; 694 edns->bits &= EDNS_DO; 695 if(!edns_opt_inplace_reply(edns, worker->scratchpad)) 696 edns->opt_list = NULL; 697 attach_edns_record(pkt, edns); 698 } 699 700 /** 701 * Answer CH class queries. 702 * @param w: worker 703 * @param qinfo: query info. Pointer into packet buffer. 704 * @param edns: edns info from query. 705 * @param pkt: packet buffer. 706 * @return: true if a reply is to be sent. 707 */ 708 static int 709 answer_chaos(struct worker* w, struct query_info* qinfo, 710 struct edns_data* edns, sldns_buffer* pkt) 711 { 712 struct config_file* cfg = w->env.cfg; 713 if(qinfo->qtype != LDNS_RR_TYPE_ANY && qinfo->qtype != LDNS_RR_TYPE_TXT) 714 return 0; 715 if(query_dname_compare(qinfo->qname, 716 (uint8_t*)"\002id\006server") == 0 || 717 query_dname_compare(qinfo->qname, 718 (uint8_t*)"\010hostname\004bind") == 0) 719 { 720 if(cfg->hide_identity) 721 return 0; 722 if(cfg->identity==NULL || cfg->identity[0]==0) { 723 char buf[MAXHOSTNAMELEN+1]; 724 if (gethostname(buf, MAXHOSTNAMELEN) == 0) { 725 buf[MAXHOSTNAMELEN] = 0; 726 chaos_replystr(pkt, buf, edns, w); 727 } else { 728 log_err("gethostname: %s", strerror(errno)); 729 chaos_replystr(pkt, "no hostname", edns, w); 730 } 731 } 732 else chaos_replystr(pkt, cfg->identity, edns, w); 733 return 1; 734 } 735 if(query_dname_compare(qinfo->qname, 736 (uint8_t*)"\007version\006server") == 0 || 737 query_dname_compare(qinfo->qname, 738 (uint8_t*)"\007version\004bind") == 0) 739 { 740 if(cfg->hide_version) 741 return 0; 742 if(cfg->version==NULL || cfg->version[0]==0) 743 chaos_replystr(pkt, PACKAGE_STRING, edns, w); 744 else chaos_replystr(pkt, cfg->version, edns, w); 745 return 1; 746 } 747 return 0; 748 } 749 750 static int 751 deny_refuse(struct comm_point* c, enum acl_access acl, 752 enum acl_access deny, enum acl_access refuse, 753 struct worker* worker, struct comm_reply* repinfo) 754 { 755 if(acl == deny) { 756 comm_point_drop_reply(repinfo); 757 if(worker->stats.extended) 758 worker->stats.unwanted_queries++; 759 return 0; 760 } else if(acl == refuse) { 761 log_addr(VERB_ALGO, "refused query from", 762 &repinfo->addr, repinfo->addrlen); 763 log_buf(VERB_ALGO, "refuse", c->buffer); 764 if(worker->stats.extended) 765 worker->stats.unwanted_queries++; 766 if(worker_check_request(c->buffer, worker) == -1) { 767 comm_point_drop_reply(repinfo); 768 return 0; /* discard this */ 769 } 770 sldns_buffer_set_limit(c->buffer, LDNS_HEADER_SIZE); 771 sldns_buffer_write_at(c->buffer, 4, 772 (uint8_t*)"\0\0\0\0\0\0\0\0", 8); 773 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 774 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 775 LDNS_RCODE_REFUSED); 776 sldns_buffer_set_position(c->buffer, LDNS_HEADER_SIZE); 777 sldns_buffer_flip(c->buffer); 778 return 1; 779 } 780 781 return -1; 782 } 783 784 static int 785 deny_refuse_all(struct comm_point* c, enum acl_access acl, 786 struct worker* worker, struct comm_reply* repinfo) 787 { 788 return deny_refuse(c, acl, acl_deny, acl_refuse, worker, repinfo); 789 } 790 791 static int 792 deny_refuse_non_local(struct comm_point* c, enum acl_access acl, 793 struct worker* worker, struct comm_reply* repinfo) 794 { 795 return deny_refuse(c, acl, acl_deny_non_local, acl_refuse_non_local, worker, repinfo); 796 } 797 798 int 799 worker_handle_request(struct comm_point* c, void* arg, int error, 800 struct comm_reply* repinfo) 801 { 802 struct worker* worker = (struct worker*)arg; 803 int ret; 804 hashvalue_t h; 805 struct lruhash_entry* e; 806 struct query_info qinfo; 807 struct edns_data edns; 808 enum acl_access acl; 809 struct acl_addr* acladdr; 810 int rc = 0; 811 812 if(error != NETEVENT_NOERROR) { 813 /* some bad tcp query DNS formats give these error calls */ 814 verbose(VERB_ALGO, "handle request called with err=%d", error); 815 return 0; 816 } 817 #ifdef USE_DNSTAP 818 if(worker->dtenv.log_client_query_messages) 819 dt_msg_send_client_query(&worker->dtenv, &repinfo->addr, c->type, 820 c->buffer); 821 #endif 822 acladdr = acl_addr_lookup(worker->daemon->acl, &repinfo->addr, 823 repinfo->addrlen); 824 acl = acl_get_control(acladdr); 825 if((ret=deny_refuse_all(c, acl, worker, repinfo)) != -1) 826 { 827 if(ret == 1) 828 goto send_reply; 829 return ret; 830 } 831 if((ret=worker_check_request(c->buffer, worker)) != 0) { 832 verbose(VERB_ALGO, "worker check request: bad query."); 833 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen); 834 if(ret != -1) { 835 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 836 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), ret); 837 return 1; 838 } 839 comm_point_drop_reply(repinfo); 840 return 0; 841 } 842 worker->stats.num_queries++; 843 /* see if query is in the cache */ 844 if(!query_info_parse(&qinfo, c->buffer)) { 845 verbose(VERB_ALGO, "worker parse request: formerror."); 846 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen); 847 if(worker_err_ratelimit(worker, LDNS_RCODE_FORMERR) == -1) { 848 comm_point_drop_reply(repinfo); 849 return 0; 850 } 851 sldns_buffer_rewind(c->buffer); 852 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 853 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 854 LDNS_RCODE_FORMERR); 855 server_stats_insrcode(&worker->stats, c->buffer); 856 goto send_reply; 857 } 858 if(worker->env.cfg->log_queries) { 859 char ip[128]; 860 addr_to_str(&repinfo->addr, repinfo->addrlen, ip, sizeof(ip)); 861 log_nametypeclass(0, ip, qinfo.qname, qinfo.qtype, qinfo.qclass); 862 } 863 if(qinfo.qtype == LDNS_RR_TYPE_AXFR || 864 qinfo.qtype == LDNS_RR_TYPE_IXFR) { 865 verbose(VERB_ALGO, "worker request: refused zone transfer."); 866 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen); 867 sldns_buffer_rewind(c->buffer); 868 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 869 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 870 LDNS_RCODE_REFUSED); 871 if(worker->stats.extended) { 872 worker->stats.qtype[qinfo.qtype]++; 873 server_stats_insrcode(&worker->stats, c->buffer); 874 } 875 goto send_reply; 876 } 877 if((ret=parse_edns_from_pkt(c->buffer, &edns, worker->scratchpad)) != 0) { 878 struct edns_data reply_edns; 879 verbose(VERB_ALGO, "worker parse edns: formerror."); 880 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen); 881 memset(&reply_edns, 0, sizeof(reply_edns)); 882 reply_edns.edns_present = 1; 883 reply_edns.udp_size = EDNS_ADVERTISED_SIZE; 884 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), ret); 885 error_encode(c->buffer, ret, &qinfo, 886 *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 887 sldns_buffer_read_u16_at(c->buffer, 2), &reply_edns); 888 regional_free_all(worker->scratchpad); 889 server_stats_insrcode(&worker->stats, c->buffer); 890 goto send_reply; 891 } 892 if(edns.edns_present && edns.edns_version != 0) { 893 edns.ext_rcode = (uint8_t)(EDNS_RCODE_BADVERS>>4); 894 edns.edns_version = EDNS_ADVERTISED_VERSION; 895 edns.udp_size = EDNS_ADVERTISED_SIZE; 896 edns.bits &= EDNS_DO; 897 edns.opt_list = NULL; 898 verbose(VERB_ALGO, "query with bad edns version."); 899 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen); 900 error_encode(c->buffer, EDNS_RCODE_BADVERS&0xf, &qinfo, 901 *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 902 sldns_buffer_read_u16_at(c->buffer, 2), NULL); 903 attach_edns_record(c->buffer, &edns); 904 regional_free_all(worker->scratchpad); 905 goto send_reply; 906 } 907 if(edns.edns_present && edns.udp_size < NORMAL_UDP_SIZE && 908 worker->daemon->cfg->harden_short_bufsize) { 909 verbose(VERB_QUERY, "worker request: EDNS bufsize %d ignored", 910 (int)edns.udp_size); 911 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen); 912 edns.udp_size = NORMAL_UDP_SIZE; 913 } 914 if(edns.udp_size > worker->daemon->cfg->max_udp_size && 915 c->type == comm_udp) { 916 verbose(VERB_QUERY, 917 "worker request: max UDP reply size modified" 918 " (%d to max-udp-size)", (int)edns.udp_size); 919 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen); 920 edns.udp_size = worker->daemon->cfg->max_udp_size; 921 } 922 if(edns.udp_size < LDNS_HEADER_SIZE) { 923 verbose(VERB_ALGO, "worker request: edns is too small."); 924 log_addr(VERB_CLIENT, "from", &repinfo->addr, repinfo->addrlen); 925 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 926 LDNS_TC_SET(sldns_buffer_begin(c->buffer)); 927 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 928 LDNS_RCODE_SERVFAIL); 929 sldns_buffer_set_position(c->buffer, LDNS_HEADER_SIZE); 930 sldns_buffer_write_at(c->buffer, 4, 931 (uint8_t*)"\0\0\0\0\0\0\0\0", 8); 932 sldns_buffer_flip(c->buffer); 933 regional_free_all(worker->scratchpad); 934 goto send_reply; 935 } 936 if(worker->stats.extended) 937 server_stats_insquery(&worker->stats, c, qinfo.qtype, 938 qinfo.qclass, &edns, repinfo); 939 if(c->type != comm_udp) 940 edns.udp_size = 65535; /* max size for TCP replies */ 941 if(qinfo.qclass == LDNS_RR_CLASS_CH && answer_chaos(worker, &qinfo, 942 &edns, c->buffer)) { 943 server_stats_insrcode(&worker->stats, c->buffer); 944 regional_free_all(worker->scratchpad); 945 goto send_reply; 946 } 947 if(local_zones_answer(worker->daemon->local_zones, &qinfo, &edns, 948 c->buffer, worker->scratchpad, repinfo, 949 acladdr->taglist, acladdr->taglen, acladdr->tag_actions, 950 acladdr->tag_actions_size, acladdr->tag_datas, 951 acladdr->tag_datas_size, worker->daemon->cfg->tagname, 952 worker->daemon->cfg->num_tags)) { 953 regional_free_all(worker->scratchpad); 954 if(sldns_buffer_limit(c->buffer) == 0) { 955 comm_point_drop_reply(repinfo); 956 return 0; 957 } 958 server_stats_insrcode(&worker->stats, c->buffer); 959 goto send_reply; 960 } 961 962 /* We've looked in our local zones. If the answer isn't there, we 963 * might need to bail out based on ACLs now. */ 964 if((ret=deny_refuse_non_local(c, acl, worker, repinfo)) != -1) 965 { 966 regional_free_all(worker->scratchpad); 967 if(ret == 1) 968 goto send_reply; 969 return ret; 970 } 971 972 /* If this request does not have the recursion bit set, verify 973 * ACLs allow the snooping. */ 974 if(!(LDNS_RD_WIRE(sldns_buffer_begin(c->buffer))) && 975 acl != acl_allow_snoop ) { 976 sldns_buffer_set_limit(c->buffer, LDNS_HEADER_SIZE); 977 sldns_buffer_write_at(c->buffer, 4, 978 (uint8_t*)"\0\0\0\0\0\0\0\0", 8); 979 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 980 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 981 LDNS_RCODE_REFUSED); 982 sldns_buffer_flip(c->buffer); 983 regional_free_all(worker->scratchpad); 984 server_stats_insrcode(&worker->stats, c->buffer); 985 log_addr(VERB_ALGO, "refused nonrec (cache snoop) query from", 986 &repinfo->addr, repinfo->addrlen); 987 goto send_reply; 988 } 989 h = query_info_hash(&qinfo, sldns_buffer_read_u16_at(c->buffer, 2)); 990 if((e=slabhash_lookup(worker->env.msg_cache, h, &qinfo, 0))) { 991 /* answer from cache - we have acquired a readlock on it */ 992 if(answer_from_cache(worker, &qinfo, 993 (struct reply_info*)e->data, 994 *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 995 sldns_buffer_read_u16_at(c->buffer, 2), repinfo, 996 &edns)) { 997 /* prefetch it if the prefetch TTL expired */ 998 if(worker->env.cfg->prefetch && *worker->env.now >= 999 ((struct reply_info*)e->data)->prefetch_ttl) { 1000 time_t leeway = ((struct reply_info*)e-> 1001 data)->ttl - *worker->env.now; 1002 lock_rw_unlock(&e->lock); 1003 reply_and_prefetch(worker, &qinfo, 1004 sldns_buffer_read_u16_at(c->buffer, 2), 1005 repinfo, leeway); 1006 rc = 0; 1007 regional_free_all(worker->scratchpad); 1008 goto send_reply_rc; 1009 } 1010 lock_rw_unlock(&e->lock); 1011 regional_free_all(worker->scratchpad); 1012 goto send_reply; 1013 } 1014 verbose(VERB_ALGO, "answer from the cache failed"); 1015 lock_rw_unlock(&e->lock); 1016 } 1017 if(!LDNS_RD_WIRE(sldns_buffer_begin(c->buffer))) { 1018 if(answer_norec_from_cache(worker, &qinfo, 1019 *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 1020 sldns_buffer_read_u16_at(c->buffer, 2), repinfo, 1021 &edns)) { 1022 regional_free_all(worker->scratchpad); 1023 goto send_reply; 1024 } 1025 verbose(VERB_ALGO, "answer norec from cache -- " 1026 "need to validate or not primed"); 1027 } 1028 sldns_buffer_rewind(c->buffer); 1029 server_stats_querymiss(&worker->stats, worker); 1030 1031 if(verbosity >= VERB_CLIENT) { 1032 if(c->type == comm_udp) 1033 log_addr(VERB_CLIENT, "udp request from", 1034 &repinfo->addr, repinfo->addrlen); 1035 else log_addr(VERB_CLIENT, "tcp request from", 1036 &repinfo->addr, repinfo->addrlen); 1037 } 1038 1039 /* grab a work request structure for this new request */ 1040 mesh_new_client(worker->env.mesh, &qinfo, 1041 sldns_buffer_read_u16_at(c->buffer, 2), 1042 &edns, repinfo, *(uint16_t*)(void *)sldns_buffer_begin(c->buffer)); 1043 regional_free_all(worker->scratchpad); 1044 worker_mem_report(worker, NULL); 1045 return 0; 1046 1047 send_reply: 1048 rc = 1; 1049 send_reply_rc: 1050 #ifdef USE_DNSTAP 1051 if(worker->dtenv.log_client_response_messages) 1052 dt_msg_send_client_response(&worker->dtenv, &repinfo->addr, 1053 c->type, c->buffer); 1054 #endif 1055 return rc; 1056 } 1057 1058 void 1059 worker_sighandler(int sig, void* arg) 1060 { 1061 /* note that log, print, syscalls here give race conditions. 1062 * And cause hangups if the log-lock is held by the application. */ 1063 struct worker* worker = (struct worker*)arg; 1064 switch(sig) { 1065 #ifdef SIGHUP 1066 case SIGHUP: 1067 comm_base_exit(worker->base); 1068 break; 1069 #endif 1070 case SIGINT: 1071 worker->need_to_exit = 1; 1072 comm_base_exit(worker->base); 1073 break; 1074 #ifdef SIGQUIT 1075 case SIGQUIT: 1076 worker->need_to_exit = 1; 1077 comm_base_exit(worker->base); 1078 break; 1079 #endif 1080 case SIGTERM: 1081 worker->need_to_exit = 1; 1082 comm_base_exit(worker->base); 1083 break; 1084 default: 1085 /* unknown signal, ignored */ 1086 break; 1087 } 1088 } 1089 1090 /** restart statistics timer for worker, if enabled */ 1091 static void 1092 worker_restart_timer(struct worker* worker) 1093 { 1094 if(worker->env.cfg->stat_interval > 0) { 1095 struct timeval tv; 1096 #ifndef S_SPLINT_S 1097 tv.tv_sec = worker->env.cfg->stat_interval; 1098 tv.tv_usec = 0; 1099 #endif 1100 comm_timer_set(worker->stat_timer, &tv); 1101 } 1102 } 1103 1104 void worker_stat_timer_cb(void* arg) 1105 { 1106 struct worker* worker = (struct worker*)arg; 1107 server_stats_log(&worker->stats, worker, worker->thread_num); 1108 mesh_stats(worker->env.mesh, "mesh has"); 1109 worker_mem_report(worker, NULL); 1110 if(!worker->daemon->cfg->stat_cumulative) { 1111 worker_stats_clear(worker); 1112 } 1113 /* start next timer */ 1114 worker_restart_timer(worker); 1115 } 1116 1117 void worker_probe_timer_cb(void* arg) 1118 { 1119 struct worker* worker = (struct worker*)arg; 1120 struct timeval tv; 1121 #ifndef S_SPLINT_S 1122 tv.tv_sec = (time_t)autr_probe_timer(&worker->env); 1123 tv.tv_usec = 0; 1124 #endif 1125 if(tv.tv_sec != 0) 1126 comm_timer_set(worker->env.probe_timer, &tv); 1127 } 1128 1129 struct worker* 1130 worker_create(struct daemon* daemon, int id, int* ports, int n) 1131 { 1132 unsigned int seed; 1133 struct worker* worker = (struct worker*)calloc(1, 1134 sizeof(struct worker)); 1135 if(!worker) 1136 return NULL; 1137 worker->numports = n; 1138 worker->ports = (int*)memdup(ports, sizeof(int)*n); 1139 if(!worker->ports) { 1140 free(worker); 1141 return NULL; 1142 } 1143 worker->daemon = daemon; 1144 worker->thread_num = id; 1145 if(!(worker->cmd = tube_create())) { 1146 free(worker->ports); 1147 free(worker); 1148 return NULL; 1149 } 1150 /* create random state here to avoid locking trouble in RAND_bytes */ 1151 seed = (unsigned int)time(NULL) ^ (unsigned int)getpid() ^ 1152 (((unsigned int)worker->thread_num)<<17); 1153 /* shift thread_num so it does not match out pid bits */ 1154 if(!(worker->rndstate = ub_initstate(seed, daemon->rand))) { 1155 seed = 0; 1156 log_err("could not init random numbers."); 1157 tube_delete(worker->cmd); 1158 free(worker->ports); 1159 free(worker); 1160 return NULL; 1161 } 1162 seed = 0; 1163 #ifdef USE_DNSTAP 1164 if(daemon->cfg->dnstap) { 1165 log_assert(daemon->dtenv != NULL); 1166 memcpy(&worker->dtenv, daemon->dtenv, sizeof(struct dt_env)); 1167 if(!dt_init(&worker->dtenv)) 1168 fatal_exit("dt_init failed"); 1169 } 1170 #endif 1171 return worker; 1172 } 1173 1174 int 1175 worker_init(struct worker* worker, struct config_file *cfg, 1176 struct listen_port* ports, int do_sigs) 1177 { 1178 #ifdef USE_DNSTAP 1179 struct dt_env* dtenv = &worker->dtenv; 1180 #else 1181 void* dtenv = NULL; 1182 #endif 1183 worker->need_to_exit = 0; 1184 worker->base = comm_base_create(do_sigs); 1185 if(!worker->base) { 1186 log_err("could not create event handling base"); 1187 worker_delete(worker); 1188 return 0; 1189 } 1190 comm_base_set_slow_accept_handlers(worker->base, &worker_stop_accept, 1191 &worker_start_accept, worker); 1192 if(do_sigs) { 1193 #ifdef SIGHUP 1194 ub_thread_sig_unblock(SIGHUP); 1195 #endif 1196 ub_thread_sig_unblock(SIGINT); 1197 #ifdef SIGQUIT 1198 ub_thread_sig_unblock(SIGQUIT); 1199 #endif 1200 ub_thread_sig_unblock(SIGTERM); 1201 #ifndef LIBEVENT_SIGNAL_PROBLEM 1202 worker->comsig = comm_signal_create(worker->base, 1203 worker_sighandler, worker); 1204 if(!worker->comsig 1205 #ifdef SIGHUP 1206 || !comm_signal_bind(worker->comsig, SIGHUP) 1207 #endif 1208 #ifdef SIGQUIT 1209 || !comm_signal_bind(worker->comsig, SIGQUIT) 1210 #endif 1211 || !comm_signal_bind(worker->comsig, SIGTERM) 1212 || !comm_signal_bind(worker->comsig, SIGINT)) { 1213 log_err("could not create signal handlers"); 1214 worker_delete(worker); 1215 return 0; 1216 } 1217 #endif /* LIBEVENT_SIGNAL_PROBLEM */ 1218 if(!daemon_remote_open_accept(worker->daemon->rc, 1219 worker->daemon->rc_ports, worker)) { 1220 worker_delete(worker); 1221 return 0; 1222 } 1223 #ifdef UB_ON_WINDOWS 1224 wsvc_setup_worker(worker); 1225 #endif /* UB_ON_WINDOWS */ 1226 } else { /* !do_sigs */ 1227 worker->comsig = NULL; 1228 } 1229 worker->front = listen_create(worker->base, ports, 1230 cfg->msg_buffer_size, (int)cfg->incoming_num_tcp, 1231 worker->daemon->listen_sslctx, dtenv, worker_handle_request, 1232 worker); 1233 if(!worker->front) { 1234 log_err("could not create listening sockets"); 1235 worker_delete(worker); 1236 return 0; 1237 } 1238 worker->back = outside_network_create(worker->base, 1239 cfg->msg_buffer_size, (size_t)cfg->outgoing_num_ports, 1240 cfg->out_ifs, cfg->num_out_ifs, cfg->do_ip4, cfg->do_ip6, 1241 cfg->do_tcp?cfg->outgoing_num_tcp:0, 1242 worker->daemon->env->infra_cache, worker->rndstate, 1243 cfg->use_caps_bits_for_id, worker->ports, worker->numports, 1244 cfg->unwanted_threshold, cfg->outgoing_tcp_mss, 1245 &worker_alloc_cleanup, worker, 1246 cfg->do_udp, worker->daemon->connect_sslctx, cfg->delay_close, 1247 dtenv); 1248 if(!worker->back) { 1249 log_err("could not create outgoing sockets"); 1250 worker_delete(worker); 1251 return 0; 1252 } 1253 /* start listening to commands */ 1254 if(!tube_setup_bg_listen(worker->cmd, worker->base, 1255 &worker_handle_control_cmd, worker)) { 1256 log_err("could not create control compt."); 1257 worker_delete(worker); 1258 return 0; 1259 } 1260 worker->stat_timer = comm_timer_create(worker->base, 1261 worker_stat_timer_cb, worker); 1262 if(!worker->stat_timer) { 1263 log_err("could not create statistics timer"); 1264 } 1265 1266 /* we use the msg_buffer_size as a good estimate for what the 1267 * user wants for memory usage sizes */ 1268 worker->scratchpad = regional_create_custom(cfg->msg_buffer_size); 1269 if(!worker->scratchpad) { 1270 log_err("malloc failure"); 1271 worker_delete(worker); 1272 return 0; 1273 } 1274 1275 server_stats_init(&worker->stats, cfg); 1276 alloc_init(&worker->alloc, &worker->daemon->superalloc, 1277 worker->thread_num); 1278 alloc_set_id_cleanup(&worker->alloc, &worker_alloc_cleanup, worker); 1279 worker->env = *worker->daemon->env; 1280 comm_base_timept(worker->base, &worker->env.now, &worker->env.now_tv); 1281 if(worker->thread_num == 0) 1282 log_set_time(worker->env.now); 1283 worker->env.worker = worker; 1284 worker->env.send_query = &worker_send_query; 1285 worker->env.alloc = &worker->alloc; 1286 worker->env.rnd = worker->rndstate; 1287 worker->env.scratch = worker->scratchpad; 1288 worker->env.mesh = mesh_create(&worker->daemon->mods, &worker->env); 1289 worker->env.detach_subs = &mesh_detach_subs; 1290 worker->env.attach_sub = &mesh_attach_sub; 1291 worker->env.kill_sub = &mesh_state_delete; 1292 worker->env.detect_cycle = &mesh_detect_cycle; 1293 worker->env.scratch_buffer = sldns_buffer_new(cfg->msg_buffer_size); 1294 if(!(worker->env.fwds = forwards_create()) || 1295 !forwards_apply_cfg(worker->env.fwds, cfg)) { 1296 log_err("Could not set forward zones"); 1297 worker_delete(worker); 1298 return 0; 1299 } 1300 if(!(worker->env.hints = hints_create()) || 1301 !hints_apply_cfg(worker->env.hints, cfg)) { 1302 log_err("Could not set root or stub hints"); 1303 worker_delete(worker); 1304 return 0; 1305 } 1306 /* one probe timer per process -- if we have 5011 anchors */ 1307 if(autr_get_num_anchors(worker->env.anchors) > 0 1308 #ifndef THREADS_DISABLED 1309 && worker->thread_num == 0 1310 #endif 1311 ) { 1312 struct timeval tv; 1313 tv.tv_sec = 0; 1314 tv.tv_usec = 0; 1315 worker->env.probe_timer = comm_timer_create(worker->base, 1316 worker_probe_timer_cb, worker); 1317 if(!worker->env.probe_timer) { 1318 log_err("could not create 5011-probe timer"); 1319 } else { 1320 /* let timer fire, then it can reset itself */ 1321 comm_timer_set(worker->env.probe_timer, &tv); 1322 } 1323 } 1324 if(!worker->env.mesh || !worker->env.scratch_buffer) { 1325 worker_delete(worker); 1326 return 0; 1327 } 1328 worker_mem_report(worker, NULL); 1329 /* if statistics enabled start timer */ 1330 if(worker->env.cfg->stat_interval > 0) { 1331 verbose(VERB_ALGO, "set statistics interval %d secs", 1332 worker->env.cfg->stat_interval); 1333 worker_restart_timer(worker); 1334 } 1335 return 1; 1336 } 1337 1338 void 1339 worker_work(struct worker* worker) 1340 { 1341 comm_base_dispatch(worker->base); 1342 } 1343 1344 void 1345 worker_delete(struct worker* worker) 1346 { 1347 if(!worker) 1348 return; 1349 if(worker->env.mesh && verbosity >= VERB_OPS) { 1350 server_stats_log(&worker->stats, worker, worker->thread_num); 1351 mesh_stats(worker->env.mesh, "mesh has"); 1352 worker_mem_report(worker, NULL); 1353 } 1354 outside_network_quit_prepare(worker->back); 1355 mesh_delete(worker->env.mesh); 1356 sldns_buffer_free(worker->env.scratch_buffer); 1357 forwards_delete(worker->env.fwds); 1358 hints_delete(worker->env.hints); 1359 listen_delete(worker->front); 1360 outside_network_delete(worker->back); 1361 comm_signal_delete(worker->comsig); 1362 tube_delete(worker->cmd); 1363 comm_timer_delete(worker->stat_timer); 1364 comm_timer_delete(worker->env.probe_timer); 1365 free(worker->ports); 1366 if(worker->thread_num == 0) { 1367 log_set_time(NULL); 1368 #ifdef UB_ON_WINDOWS 1369 wsvc_desetup_worker(worker); 1370 #endif /* UB_ON_WINDOWS */ 1371 } 1372 comm_base_delete(worker->base); 1373 ub_randfree(worker->rndstate); 1374 alloc_clear(&worker->alloc); 1375 regional_destroy(worker->scratchpad); 1376 free(worker); 1377 } 1378 1379 struct outbound_entry* 1380 worker_send_query(uint8_t* qname, size_t qnamelen, uint16_t qtype, 1381 uint16_t qclass, uint16_t flags, int dnssec, int want_dnssec, 1382 int nocaps, struct edns_option* opt_list, 1383 struct sockaddr_storage* addr, socklen_t addrlen, uint8_t* zone, 1384 size_t zonelen, struct module_qstate* q) 1385 { 1386 struct worker* worker = q->env->worker; 1387 struct outbound_entry* e = (struct outbound_entry*)regional_alloc( 1388 q->region, sizeof(*e)); 1389 if(!e) 1390 return NULL; 1391 e->qstate = q; 1392 e->qsent = outnet_serviced_query(worker->back, qname, 1393 qnamelen, qtype, qclass, flags, dnssec, want_dnssec, nocaps, 1394 q->env->cfg->tcp_upstream, q->env->cfg->ssl_upstream, opt_list, 1395 addr, addrlen, zone, zonelen, worker_handle_service_reply, e, 1396 worker->back->udp_buff); 1397 if(!e->qsent) { 1398 return NULL; 1399 } 1400 return e; 1401 } 1402 1403 void 1404 worker_alloc_cleanup(void* arg) 1405 { 1406 struct worker* worker = (struct worker*)arg; 1407 slabhash_clear(&worker->env.rrset_cache->table); 1408 slabhash_clear(worker->env.msg_cache); 1409 } 1410 1411 void worker_stats_clear(struct worker* worker) 1412 { 1413 server_stats_init(&worker->stats, worker->env.cfg); 1414 mesh_stats_clear(worker->env.mesh); 1415 worker->back->unwanted_replies = 0; 1416 worker->back->num_tcp_outgoing = 0; 1417 } 1418 1419 void worker_start_accept(void* arg) 1420 { 1421 struct worker* worker = (struct worker*)arg; 1422 listen_start_accept(worker->front); 1423 if(worker->thread_num == 0) 1424 daemon_remote_start_accept(worker->daemon->rc); 1425 } 1426 1427 void worker_stop_accept(void* arg) 1428 { 1429 struct worker* worker = (struct worker*)arg; 1430 listen_stop_accept(worker->front); 1431 if(worker->thread_num == 0) 1432 daemon_remote_stop_accept(worker->daemon->rc); 1433 } 1434 1435 /* --- fake callbacks for fptr_wlist to work --- */ 1436 struct outbound_entry* libworker_send_query(uint8_t* ATTR_UNUSED(qname), 1437 size_t ATTR_UNUSED(qnamelen), uint16_t ATTR_UNUSED(qtype), 1438 uint16_t ATTR_UNUSED(qclass), uint16_t ATTR_UNUSED(flags), 1439 int ATTR_UNUSED(dnssec), int ATTR_UNUSED(want_dnssec), 1440 int ATTR_UNUSED(nocaps), struct edns_option* ATTR_UNUSED(opt_list), 1441 struct sockaddr_storage* ATTR_UNUSED(addr), 1442 socklen_t ATTR_UNUSED(addrlen), uint8_t* ATTR_UNUSED(zone), 1443 size_t ATTR_UNUSED(zonelen), struct module_qstate* ATTR_UNUSED(q)) 1444 { 1445 log_assert(0); 1446 return 0; 1447 } 1448 1449 int libworker_handle_reply(struct comm_point* ATTR_UNUSED(c), 1450 void* ATTR_UNUSED(arg), int ATTR_UNUSED(error), 1451 struct comm_reply* ATTR_UNUSED(reply_info)) 1452 { 1453 log_assert(0); 1454 return 0; 1455 } 1456 1457 int libworker_handle_service_reply(struct comm_point* ATTR_UNUSED(c), 1458 void* ATTR_UNUSED(arg), int ATTR_UNUSED(error), 1459 struct comm_reply* ATTR_UNUSED(reply_info)) 1460 { 1461 log_assert(0); 1462 return 0; 1463 } 1464 1465 void libworker_handle_control_cmd(struct tube* ATTR_UNUSED(tube), 1466 uint8_t* ATTR_UNUSED(buffer), size_t ATTR_UNUSED(len), 1467 int ATTR_UNUSED(error), void* ATTR_UNUSED(arg)) 1468 { 1469 log_assert(0); 1470 } 1471 1472 void libworker_fg_done_cb(void* ATTR_UNUSED(arg), int ATTR_UNUSED(rcode), 1473 sldns_buffer* ATTR_UNUSED(buf), enum sec_status ATTR_UNUSED(s), 1474 char* ATTR_UNUSED(why_bogus)) 1475 { 1476 log_assert(0); 1477 } 1478 1479 void libworker_bg_done_cb(void* ATTR_UNUSED(arg), int ATTR_UNUSED(rcode), 1480 sldns_buffer* ATTR_UNUSED(buf), enum sec_status ATTR_UNUSED(s), 1481 char* ATTR_UNUSED(why_bogus)) 1482 { 1483 log_assert(0); 1484 } 1485 1486 void libworker_event_done_cb(void* ATTR_UNUSED(arg), int ATTR_UNUSED(rcode), 1487 sldns_buffer* ATTR_UNUSED(buf), enum sec_status ATTR_UNUSED(s), 1488 char* ATTR_UNUSED(why_bogus)) 1489 { 1490 log_assert(0); 1491 } 1492 1493 int context_query_cmp(const void* ATTR_UNUSED(a), const void* ATTR_UNUSED(b)) 1494 { 1495 log_assert(0); 1496 return 0; 1497 } 1498 1499 int order_lock_cmp(const void* ATTR_UNUSED(e1), const void* ATTR_UNUSED(e2)) 1500 { 1501 log_assert(0); 1502 return 0; 1503 } 1504 1505 int codeline_cmp(const void* ATTR_UNUSED(a), const void* ATTR_UNUSED(b)) 1506 { 1507 log_assert(0); 1508 return 0; 1509 } 1510 1511