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/authzone.h" 62 #include "services/mesh.h" 63 #include "services/localzone.h" 64 #include "services/rpz.h" 65 #include "util/data/msgparse.h" 66 #include "util/data/msgencode.h" 67 #include "util/data/dname.h" 68 #include "util/fptr_wlist.h" 69 #include "util/proxy_protocol.h" 70 #include "util/tube.h" 71 #include "util/edns.h" 72 #include "util/timeval_func.h" 73 #include "iterator/iter_fwd.h" 74 #include "iterator/iter_hints.h" 75 #include "iterator/iter_utils.h" 76 #include "validator/autotrust.h" 77 #include "validator/val_anchor.h" 78 #include "respip/respip.h" 79 #include "libunbound/context.h" 80 #include "libunbound/libworker.h" 81 #include "sldns/sbuffer.h" 82 #include "sldns/wire2str.h" 83 #include "util/shm_side/shm_main.h" 84 #include "dnscrypt/dnscrypt.h" 85 #include "dnstap/dtstream.h" 86 87 #ifdef HAVE_SYS_TYPES_H 88 # include <sys/types.h> 89 #endif 90 #ifdef HAVE_NETDB_H 91 #include <netdb.h> 92 #endif 93 #include <signal.h> 94 #ifdef UB_ON_WINDOWS 95 #include "winrc/win_svc.h" 96 #endif 97 98 /** Size of an UDP datagram */ 99 #define NORMAL_UDP_SIZE 512 /* bytes */ 100 /** ratelimit for error responses */ 101 #define ERROR_RATELIMIT 100 /* qps */ 102 103 /** 104 * seconds to add to prefetch leeway. This is a TTL that expires old rrsets 105 * earlier than they should in order to put the new update into the cache. 106 * This additional value is to make sure that if not all TTLs are equal in 107 * the message to be updated(and replaced), that rrsets with up to this much 108 * extra TTL are also replaced. This means that the resulting new message 109 * will have (most likely) this TTL at least, avoiding very small 'split 110 * second' TTLs due to operators choosing relative primes for TTLs (or so). 111 * Also has to be at least one to break ties (and overwrite cached entry). 112 */ 113 #define PREFETCH_EXPIRY_ADD 60 114 115 /** Report on memory usage by this thread and global */ 116 static void 117 worker_mem_report(struct worker* ATTR_UNUSED(worker), 118 struct serviced_query* ATTR_UNUSED(cur_serv)) 119 { 120 #ifdef UNBOUND_ALLOC_STATS 121 /* measure memory leakage */ 122 extern size_t unbound_mem_alloc, unbound_mem_freed; 123 /* debug func in validator module */ 124 size_t total, front, back, mesh, msg, rrset, infra, ac, superac; 125 size_t me, iter, val, anch; 126 int i; 127 #ifdef CLIENT_SUBNET 128 size_t subnet = 0; 129 #endif /* CLIENT_SUBNET */ 130 if(verbosity < VERB_ALGO) 131 return; 132 front = listen_get_mem(worker->front); 133 back = outnet_get_mem(worker->back); 134 msg = slabhash_get_mem(worker->env.msg_cache); 135 rrset = slabhash_get_mem(&worker->env.rrset_cache->table); 136 infra = infra_get_mem(worker->env.infra_cache); 137 mesh = mesh_get_mem(worker->env.mesh); 138 ac = alloc_get_mem(worker->alloc); 139 superac = alloc_get_mem(&worker->daemon->superalloc); 140 anch = anchors_get_mem(worker->env.anchors); 141 iter = 0; 142 val = 0; 143 for(i=0; i<worker->env.mesh->mods.num; i++) { 144 fptr_ok(fptr_whitelist_mod_get_mem(worker->env.mesh-> 145 mods.mod[i]->get_mem)); 146 if(strcmp(worker->env.mesh->mods.mod[i]->name, "validator")==0) 147 val += (*worker->env.mesh->mods.mod[i]->get_mem) 148 (&worker->env, i); 149 #ifdef CLIENT_SUBNET 150 else if(strcmp(worker->env.mesh->mods.mod[i]->name, 151 "subnetcache")==0) 152 subnet += (*worker->env.mesh->mods.mod[i]->get_mem) 153 (&worker->env, i); 154 #endif /* CLIENT_SUBNET */ 155 else iter += (*worker->env.mesh->mods.mod[i]->get_mem) 156 (&worker->env, i); 157 } 158 me = sizeof(*worker) + sizeof(*worker->base) + sizeof(*worker->comsig) 159 + comm_point_get_mem(worker->cmd_com) 160 + sizeof(worker->rndstate) 161 + regional_get_mem(worker->scratchpad) 162 + sizeof(*worker->env.scratch_buffer) 163 + sldns_buffer_capacity(worker->env.scratch_buffer) 164 + forwards_get_mem(worker->env.fwds) 165 + hints_get_mem(worker->env.hints); 166 if(worker->thread_num == 0) 167 me += acl_list_get_mem(worker->daemon->acl); 168 if(cur_serv) { 169 me += serviced_get_mem(cur_serv); 170 } 171 total = front+back+mesh+msg+rrset+infra+iter+val+ac+superac+me; 172 #ifdef CLIENT_SUBNET 173 total += subnet; 174 log_info("Memory conditions: %u front=%u back=%u mesh=%u msg=%u " 175 "rrset=%u infra=%u iter=%u val=%u subnet=%u anchors=%u " 176 "alloccache=%u globalalloccache=%u me=%u", 177 (unsigned)total, (unsigned)front, (unsigned)back, 178 (unsigned)mesh, (unsigned)msg, (unsigned)rrset, (unsigned)infra, 179 (unsigned)iter, (unsigned)val, 180 (unsigned)subnet, (unsigned)anch, (unsigned)ac, 181 (unsigned)superac, (unsigned)me); 182 #else /* no CLIENT_SUBNET */ 183 log_info("Memory conditions: %u front=%u back=%u mesh=%u msg=%u " 184 "rrset=%u infra=%u iter=%u val=%u anchors=%u " 185 "alloccache=%u globalalloccache=%u me=%u", 186 (unsigned)total, (unsigned)front, (unsigned)back, 187 (unsigned)mesh, (unsigned)msg, (unsigned)rrset, 188 (unsigned)infra, (unsigned)iter, (unsigned)val, (unsigned)anch, 189 (unsigned)ac, (unsigned)superac, (unsigned)me); 190 #endif /* CLIENT_SUBNET */ 191 log_info("Total heap memory estimate: %u total-alloc: %u " 192 "total-free: %u", (unsigned)total, 193 (unsigned)unbound_mem_alloc, (unsigned)unbound_mem_freed); 194 #else /* no UNBOUND_ALLOC_STATS */ 195 size_t val = 0; 196 #ifdef CLIENT_SUBNET 197 size_t subnet = 0; 198 #endif /* CLIENT_SUBNET */ 199 int i; 200 if(verbosity < VERB_QUERY) 201 return; 202 for(i=0; i<worker->env.mesh->mods.num; i++) { 203 fptr_ok(fptr_whitelist_mod_get_mem(worker->env.mesh-> 204 mods.mod[i]->get_mem)); 205 if(strcmp(worker->env.mesh->mods.mod[i]->name, "validator")==0) 206 val += (*worker->env.mesh->mods.mod[i]->get_mem) 207 (&worker->env, i); 208 #ifdef CLIENT_SUBNET 209 else if(strcmp(worker->env.mesh->mods.mod[i]->name, 210 "subnetcache")==0) 211 subnet += (*worker->env.mesh->mods.mod[i]->get_mem) 212 (&worker->env, i); 213 #endif /* CLIENT_SUBNET */ 214 } 215 #ifdef CLIENT_SUBNET 216 verbose(VERB_QUERY, "cache memory msg=%u rrset=%u infra=%u val=%u " 217 "subnet=%u", 218 (unsigned)slabhash_get_mem(worker->env.msg_cache), 219 (unsigned)slabhash_get_mem(&worker->env.rrset_cache->table), 220 (unsigned)infra_get_mem(worker->env.infra_cache), 221 (unsigned)val, (unsigned)subnet); 222 #else /* no CLIENT_SUBNET */ 223 verbose(VERB_QUERY, "cache memory msg=%u rrset=%u infra=%u val=%u", 224 (unsigned)slabhash_get_mem(worker->env.msg_cache), 225 (unsigned)slabhash_get_mem(&worker->env.rrset_cache->table), 226 (unsigned)infra_get_mem(worker->env.infra_cache), 227 (unsigned)val); 228 #endif /* CLIENT_SUBNET */ 229 #endif /* UNBOUND_ALLOC_STATS */ 230 } 231 232 void 233 worker_send_cmd(struct worker* worker, enum worker_commands cmd) 234 { 235 uint32_t c = (uint32_t)htonl(cmd); 236 if(!tube_write_msg(worker->cmd, (uint8_t*)&c, sizeof(c), 0)) { 237 log_err("worker send cmd %d failed", (int)cmd); 238 } 239 } 240 241 int 242 worker_handle_service_reply(struct comm_point* c, void* arg, int error, 243 struct comm_reply* reply_info) 244 { 245 struct outbound_entry* e = (struct outbound_entry*)arg; 246 struct worker* worker = e->qstate->env->worker; 247 struct serviced_query *sq = e->qsent; 248 249 verbose(VERB_ALGO, "worker svcd callback for qstate %p", e->qstate); 250 if(error != 0) { 251 mesh_report_reply(worker->env.mesh, e, reply_info, error); 252 worker_mem_report(worker, sq); 253 return 0; 254 } 255 /* sanity check. */ 256 if(!LDNS_QR_WIRE(sldns_buffer_begin(c->buffer)) 257 || LDNS_OPCODE_WIRE(sldns_buffer_begin(c->buffer)) != 258 LDNS_PACKET_QUERY 259 || LDNS_QDCOUNT(sldns_buffer_begin(c->buffer)) > 1) { 260 /* error becomes timeout for the module as if this reply 261 * never arrived. */ 262 verbose(VERB_ALGO, "worker: bad reply handled as timeout"); 263 mesh_report_reply(worker->env.mesh, e, reply_info, 264 NETEVENT_TIMEOUT); 265 worker_mem_report(worker, sq); 266 return 0; 267 } 268 mesh_report_reply(worker->env.mesh, e, reply_info, NETEVENT_NOERROR); 269 worker_mem_report(worker, sq); 270 return 0; 271 } 272 273 /** ratelimit error replies 274 * @param worker: the worker struct with ratelimit counter 275 * @param err: error code that would be wanted. 276 * @return value of err if okay, or -1 if it should be discarded instead. 277 */ 278 static int 279 worker_err_ratelimit(struct worker* worker, int err) 280 { 281 if(worker->err_limit_time == *worker->env.now) { 282 /* see if limit is exceeded for this second */ 283 if(worker->err_limit_count++ > ERROR_RATELIMIT) 284 return -1; 285 } else { 286 /* new second, new limits */ 287 worker->err_limit_time = *worker->env.now; 288 worker->err_limit_count = 1; 289 } 290 return err; 291 } 292 293 /** 294 * Structure holding the result of the worker_check_request function. 295 * Based on configuration it could be called up to four times; ideally should 296 * be called once. 297 */ 298 struct check_request_result { 299 int checked; 300 int value; 301 }; 302 /** check request sanity. 303 * @param pkt: the wire packet to examine for sanity. 304 * @param worker: parameters for checking. 305 * @param out: struct to update with the result. 306 */ 307 static void 308 worker_check_request(sldns_buffer* pkt, struct worker* worker, 309 struct check_request_result* out) 310 { 311 if(out->checked) return; 312 out->checked = 1; 313 if(sldns_buffer_limit(pkt) < LDNS_HEADER_SIZE) { 314 verbose(VERB_QUERY, "request too short, discarded"); 315 out->value = -1; 316 return; 317 } 318 if(sldns_buffer_limit(pkt) > NORMAL_UDP_SIZE && 319 worker->daemon->cfg->harden_large_queries) { 320 verbose(VERB_QUERY, "request too large, discarded"); 321 out->value = -1; 322 return; 323 } 324 if(LDNS_QR_WIRE(sldns_buffer_begin(pkt))) { 325 verbose(VERB_QUERY, "request has QR bit on, discarded"); 326 out->value = -1; 327 return; 328 } 329 if(LDNS_TC_WIRE(sldns_buffer_begin(pkt))) { 330 LDNS_TC_CLR(sldns_buffer_begin(pkt)); 331 verbose(VERB_QUERY, "request bad, has TC bit on"); 332 out->value = worker_err_ratelimit(worker, LDNS_RCODE_FORMERR); 333 return; 334 } 335 if(LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt)) != LDNS_PACKET_QUERY && 336 LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt)) != LDNS_PACKET_NOTIFY) { 337 verbose(VERB_QUERY, "request unknown opcode %d", 338 LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt))); 339 out->value = worker_err_ratelimit(worker, LDNS_RCODE_NOTIMPL); 340 return; 341 } 342 if(LDNS_QDCOUNT(sldns_buffer_begin(pkt)) != 1) { 343 verbose(VERB_QUERY, "request wrong nr qd=%d", 344 LDNS_QDCOUNT(sldns_buffer_begin(pkt))); 345 out->value = worker_err_ratelimit(worker, LDNS_RCODE_FORMERR); 346 return; 347 } 348 if(LDNS_ANCOUNT(sldns_buffer_begin(pkt)) != 0 && 349 (LDNS_ANCOUNT(sldns_buffer_begin(pkt)) != 1 || 350 LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt)) != LDNS_PACKET_NOTIFY)) { 351 verbose(VERB_QUERY, "request wrong nr an=%d", 352 LDNS_ANCOUNT(sldns_buffer_begin(pkt))); 353 out->value = worker_err_ratelimit(worker, LDNS_RCODE_FORMERR); 354 return; 355 } 356 if(LDNS_NSCOUNT(sldns_buffer_begin(pkt)) != 0) { 357 verbose(VERB_QUERY, "request wrong nr ns=%d", 358 LDNS_NSCOUNT(sldns_buffer_begin(pkt))); 359 out->value = worker_err_ratelimit(worker, LDNS_RCODE_FORMERR); 360 return; 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 out->value = worker_err_ratelimit(worker, LDNS_RCODE_FORMERR); 366 return; 367 } 368 out->value = 0; 369 return; 370 } 371 372 void 373 worker_handle_control_cmd(struct tube* ATTR_UNUSED(tube), uint8_t* msg, 374 size_t len, int error, void* arg) 375 { 376 struct worker* worker = (struct worker*)arg; 377 enum worker_commands cmd; 378 if(error != NETEVENT_NOERROR) { 379 free(msg); 380 if(error == NETEVENT_CLOSED) 381 comm_base_exit(worker->base); 382 else log_info("control event: %d", error); 383 return; 384 } 385 if(len != sizeof(uint32_t)) { 386 fatal_exit("bad control msg length %d", (int)len); 387 } 388 cmd = sldns_read_uint32(msg); 389 free(msg); 390 switch(cmd) { 391 case worker_cmd_quit: 392 verbose(VERB_ALGO, "got control cmd quit"); 393 comm_base_exit(worker->base); 394 break; 395 case worker_cmd_stats: 396 verbose(VERB_ALGO, "got control cmd stats"); 397 server_stats_reply(worker, 1); 398 break; 399 case worker_cmd_stats_noreset: 400 verbose(VERB_ALGO, "got control cmd stats_noreset"); 401 server_stats_reply(worker, 0); 402 break; 403 case worker_cmd_remote: 404 verbose(VERB_ALGO, "got control cmd remote"); 405 daemon_remote_exec(worker); 406 break; 407 default: 408 log_err("bad command %d", (int)cmd); 409 break; 410 } 411 } 412 413 /** check if a delegation is secure */ 414 static enum sec_status 415 check_delegation_secure(struct reply_info *rep) 416 { 417 /* return smallest security status */ 418 size_t i; 419 enum sec_status sec = sec_status_secure; 420 enum sec_status s; 421 size_t num = rep->an_numrrsets + rep->ns_numrrsets; 422 /* check if answer and authority are OK */ 423 for(i=0; i<num; i++) { 424 s = ((struct packed_rrset_data*)rep->rrsets[i]->entry.data) 425 ->security; 426 if(s < sec) 427 sec = s; 428 } 429 /* in additional, only unchecked triggers revalidation */ 430 for(i=num; i<rep->rrset_count; i++) { 431 s = ((struct packed_rrset_data*)rep->rrsets[i]->entry.data) 432 ->security; 433 if(s == sec_status_unchecked) 434 return s; 435 } 436 return sec; 437 } 438 439 /** remove nonsecure from a delegation referral additional section */ 440 static void 441 deleg_remove_nonsecure_additional(struct reply_info* rep) 442 { 443 /* we can simply edit it, since we are working in the scratch region */ 444 size_t i; 445 enum sec_status s; 446 447 for(i = rep->an_numrrsets+rep->ns_numrrsets; i<rep->rrset_count; i++) { 448 s = ((struct packed_rrset_data*)rep->rrsets[i]->entry.data) 449 ->security; 450 if(s != sec_status_secure) { 451 memmove(rep->rrsets+i, rep->rrsets+i+1, 452 sizeof(struct ub_packed_rrset_key*)* 453 (rep->rrset_count - i - 1)); 454 rep->ar_numrrsets--; 455 rep->rrset_count--; 456 i--; 457 } 458 } 459 } 460 461 /** answer nonrecursive query from the cache */ 462 static int 463 answer_norec_from_cache(struct worker* worker, struct query_info* qinfo, 464 uint16_t id, uint16_t flags, struct comm_reply* repinfo, 465 struct edns_data* edns) 466 { 467 /* for a nonrecursive query return either: 468 * o an error (servfail; we try to avoid this) 469 * o a delegation (closest we have; this routine tries that) 470 * o the answer (checked by answer_from_cache) 471 * 472 * So, grab a delegation from the rrset cache. 473 * Then check if it needs validation, if so, this routine fails, 474 * so that iterator can prime and validator can verify rrsets. 475 */ 476 uint16_t udpsize = edns->udp_size; 477 int secure = 0; 478 time_t timenow = *worker->env.now; 479 int has_cd_bit = (flags&BIT_CD); 480 int must_validate = (!has_cd_bit || worker->env.cfg->ignore_cd) 481 && worker->env.need_to_validate; 482 struct dns_msg *msg = NULL; 483 struct delegpt *dp; 484 485 dp = dns_cache_find_delegation(&worker->env, qinfo->qname, 486 qinfo->qname_len, qinfo->qtype, qinfo->qclass, 487 worker->scratchpad, &msg, timenow, 0, NULL, 0); 488 if(!dp) { /* no delegation, need to reprime */ 489 return 0; 490 } 491 /* In case we have a local alias, copy it into the delegation message. 492 * Shallow copy should be fine, as we'll be done with msg in this 493 * function. */ 494 msg->qinfo.local_alias = qinfo->local_alias; 495 if(must_validate) { 496 switch(check_delegation_secure(msg->rep)) { 497 case sec_status_unchecked: 498 /* some rrsets have not been verified yet, go and 499 * let validator do that */ 500 return 0; 501 case sec_status_bogus: 502 case sec_status_secure_sentinel_fail: 503 /* some rrsets are bogus, reply servfail */ 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 if(!inplace_cb_reply_servfail_call(&worker->env, qinfo, NULL, 509 msg->rep, LDNS_RCODE_SERVFAIL, edns, repinfo, worker->scratchpad, 510 worker->env.now_tv)) 511 return 0; 512 /* Attach the cached EDE (RFC8914) */ 513 if(worker->env.cfg->ede && 514 msg->rep->reason_bogus != LDNS_EDE_NONE) { 515 edns_opt_list_append_ede(&edns->opt_list_out, 516 worker->scratchpad, msg->rep->reason_bogus, 517 msg->rep->reason_bogus_str); 518 } 519 error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL, 520 &msg->qinfo, id, flags, edns); 521 if(worker->stats.extended) { 522 worker->stats.ans_bogus++; 523 worker->stats.ans_rcode[LDNS_RCODE_SERVFAIL]++; 524 } 525 return 1; 526 case sec_status_secure: 527 /* all rrsets are secure */ 528 /* remove non-secure rrsets from the add. section*/ 529 if(worker->env.cfg->val_clean_additional) 530 deleg_remove_nonsecure_additional(msg->rep); 531 secure = 1; 532 break; 533 case sec_status_indeterminate: 534 case sec_status_insecure: 535 default: 536 /* not secure */ 537 secure = 0; 538 break; 539 } 540 } 541 /* return this delegation from the cache */ 542 edns->edns_version = EDNS_ADVERTISED_VERSION; 543 edns->udp_size = EDNS_ADVERTISED_SIZE; 544 edns->ext_rcode = 0; 545 edns->bits &= EDNS_DO; 546 if(worker->env.cfg->disable_edns_do && (edns->bits & EDNS_DO)) 547 edns->edns_present = 0; 548 if(!inplace_cb_reply_cache_call(&worker->env, qinfo, NULL, msg->rep, 549 (int)(flags&LDNS_RCODE_MASK), edns, repinfo, worker->scratchpad, 550 worker->env.now_tv)) 551 return 0; 552 msg->rep->flags |= BIT_QR|BIT_RA; 553 /* Attach the cached EDE (RFC8914) if CD bit is set and the answer is 554 * bogus. */ 555 if(worker->env.cfg->ede && has_cd_bit && 556 (check_delegation_secure(msg->rep) == sec_status_bogus || 557 check_delegation_secure(msg->rep) == sec_status_secure_sentinel_fail) && 558 msg->rep->reason_bogus != LDNS_EDE_NONE) { 559 edns_opt_list_append_ede(&edns->opt_list_out, 560 worker->scratchpad, msg->rep->reason_bogus, 561 msg->rep->reason_bogus_str); 562 } 563 if(!reply_info_answer_encode(&msg->qinfo, msg->rep, id, flags, 564 repinfo->c->buffer, 0, 1, worker->scratchpad, 565 udpsize, edns, (int)(edns->bits & EDNS_DO), secure)) { 566 if(!inplace_cb_reply_servfail_call(&worker->env, qinfo, NULL, NULL, 567 LDNS_RCODE_SERVFAIL, edns, repinfo, worker->scratchpad, 568 worker->env.now_tv)) 569 edns->opt_list_inplace_cb_out = NULL; 570 error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL, 571 &msg->qinfo, id, flags, edns); 572 } 573 if(worker->stats.extended) { 574 if(secure) worker->stats.ans_secure++; 575 server_stats_insrcode(&worker->stats, repinfo->c->buffer); 576 } 577 return 1; 578 } 579 580 /** Apply, if applicable, a response IP action to a cached answer. 581 * If the answer is rewritten as a result of an action, '*encode_repp' will 582 * point to the reply info containing the modified answer. '*encode_repp' will 583 * be intact otherwise. 584 * It returns 1 on success, 0 otherwise. */ 585 static int 586 apply_respip_action(struct worker* worker, const struct query_info* qinfo, 587 struct respip_client_info* cinfo, struct reply_info* rep, 588 struct sockaddr_storage* addr, socklen_t addrlen, 589 struct ub_packed_rrset_key** alias_rrset, 590 struct reply_info** encode_repp, struct auth_zones* az) 591 { 592 struct respip_action_info actinfo = {0, 0, 0, 0, NULL, 0, NULL}; 593 actinfo.action = respip_none; 594 595 if(qinfo->qtype != LDNS_RR_TYPE_A && 596 qinfo->qtype != LDNS_RR_TYPE_AAAA && 597 qinfo->qtype != LDNS_RR_TYPE_ANY) 598 return 1; 599 600 if(!respip_rewrite_reply(qinfo, cinfo, rep, encode_repp, &actinfo, 601 alias_rrset, 0, worker->scratchpad, az, NULL)) 602 return 0; 603 604 /* xxx_deny actions mean dropping the reply, unless the original reply 605 * was redirected to response-ip data. */ 606 if(actinfo.action == respip_always_deny || 607 ((actinfo.action == respip_deny || 608 actinfo.action == respip_inform_deny) && 609 *encode_repp == rep)) 610 *encode_repp = NULL; 611 612 /* If address info is returned, it means the action should be an 613 * 'inform' variant and the information should be logged. */ 614 if(actinfo.addrinfo) { 615 respip_inform_print(&actinfo, qinfo->qname, 616 qinfo->qtype, qinfo->qclass, qinfo->local_alias, 617 addr, addrlen); 618 619 if(worker->stats.extended && actinfo.rpz_used) { 620 if(actinfo.rpz_disabled) 621 worker->stats.rpz_action[RPZ_DISABLED_ACTION]++; 622 if(actinfo.rpz_cname_override) 623 worker->stats.rpz_action[RPZ_CNAME_OVERRIDE_ACTION]++; 624 else 625 worker->stats.rpz_action[ 626 respip_action_to_rpz_action(actinfo.action)]++; 627 } 628 } 629 630 return 1; 631 } 632 633 /** answer query from the cache. 634 * Normally, the answer message will be built in repinfo->c->buffer; if the 635 * answer is supposed to be suppressed or the answer is supposed to be an 636 * incomplete CNAME chain, the buffer is explicitly cleared to signal the 637 * caller as such. In the latter case *partial_rep will point to the incomplete 638 * reply, and this function is (possibly) supposed to be called again with that 639 * *partial_rep value to complete the chain. In addition, if the query should 640 * be completely dropped, '*need_drop' will be set to 1. */ 641 static int 642 answer_from_cache(struct worker* worker, struct query_info* qinfo, 643 struct respip_client_info* cinfo, int* need_drop, int* is_expired_answer, 644 int* is_secure_answer, struct ub_packed_rrset_key** alias_rrset, 645 struct reply_info** partial_repp, 646 struct reply_info* rep, uint16_t id, uint16_t flags, 647 struct comm_reply* repinfo, struct edns_data* edns) 648 { 649 time_t timenow = *worker->env.now; 650 uint16_t udpsize = edns->udp_size; 651 struct reply_info* encode_rep = rep; 652 struct reply_info* partial_rep = *partial_repp; 653 int has_cd_bit = (flags&BIT_CD); 654 int must_validate = (!has_cd_bit || worker->env.cfg->ignore_cd) 655 && worker->env.need_to_validate; 656 *partial_repp = NULL; /* avoid accidental further pass */ 657 658 /* Check TTL */ 659 if(rep->ttl < timenow) { 660 /* Check if we need to serve expired now */ 661 if(worker->env.cfg->serve_expired && 662 !worker->env.cfg->serve_expired_client_timeout 663 #ifdef USE_CACHEDB 664 && !(worker->env.cachedb_enabled && 665 worker->env.cfg->cachedb_check_when_serve_expired) 666 #endif 667 ) { 668 if(worker->env.cfg->serve_expired_ttl && 669 rep->serve_expired_ttl < timenow) 670 return 0; 671 /* Ignore expired failure answers */ 672 if(FLAGS_GET_RCODE(rep->flags) != 673 LDNS_RCODE_NOERROR && 674 FLAGS_GET_RCODE(rep->flags) != 675 LDNS_RCODE_NXDOMAIN && 676 FLAGS_GET_RCODE(rep->flags) != 677 LDNS_RCODE_YXDOMAIN) 678 return 0; 679 if(!rrset_array_lock(rep->ref, rep->rrset_count, 0)) 680 return 0; 681 *is_expired_answer = 1; 682 } else { 683 /* the rrsets may have been updated in the meantime. 684 * we will refetch the message format from the 685 * authoritative server 686 */ 687 return 0; 688 } 689 } else { 690 if(!rrset_array_lock(rep->ref, rep->rrset_count, timenow)) 691 return 0; 692 } 693 /* locked and ids and ttls are OK. */ 694 695 /* check CNAME chain (if any) */ 696 if(rep->an_numrrsets > 0 && (rep->rrsets[0]->rk.type == 697 htons(LDNS_RR_TYPE_CNAME) || rep->rrsets[0]->rk.type == 698 htons(LDNS_RR_TYPE_DNAME))) { 699 if(!reply_check_cname_chain(qinfo, rep)) { 700 /* cname chain invalid, redo iterator steps */ 701 verbose(VERB_ALGO, "Cache reply: cname chain broken"); 702 goto bail_out; 703 } 704 } 705 /* check security status of the cached answer */ 706 if(must_validate && (rep->security == sec_status_bogus || 707 rep->security == sec_status_secure_sentinel_fail)) { 708 /* BAD cached */ 709 edns->edns_version = EDNS_ADVERTISED_VERSION; 710 edns->udp_size = EDNS_ADVERTISED_SIZE; 711 edns->ext_rcode = 0; 712 edns->bits &= EDNS_DO; 713 if(worker->env.cfg->disable_edns_do && (edns->bits & EDNS_DO)) 714 edns->edns_present = 0; 715 if(!inplace_cb_reply_servfail_call(&worker->env, qinfo, NULL, rep, 716 LDNS_RCODE_SERVFAIL, edns, repinfo, worker->scratchpad, 717 worker->env.now_tv)) 718 goto bail_out; 719 /* Attach the cached EDE (RFC8914) */ 720 if(worker->env.cfg->ede && rep->reason_bogus != LDNS_EDE_NONE) { 721 edns_opt_list_append_ede(&edns->opt_list_out, 722 worker->scratchpad, rep->reason_bogus, 723 rep->reason_bogus_str); 724 } 725 error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL, 726 qinfo, id, flags, edns); 727 rrset_array_unlock_touch(worker->env.rrset_cache, 728 worker->scratchpad, rep->ref, rep->rrset_count); 729 if(worker->stats.extended) { 730 worker->stats.ans_bogus ++; 731 worker->stats.ans_rcode[LDNS_RCODE_SERVFAIL] ++; 732 } 733 return 1; 734 } else if(rep->security == sec_status_unchecked && must_validate) { 735 verbose(VERB_ALGO, "Cache reply: unchecked entry needs " 736 "validation"); 737 goto bail_out; /* need to validate cache entry first */ 738 } else if(rep->security == sec_status_secure) { 739 if(reply_all_rrsets_secure(rep)) { 740 *is_secure_answer = 1; 741 } else { 742 if(must_validate) { 743 verbose(VERB_ALGO, "Cache reply: secure entry" 744 " changed status"); 745 goto bail_out; /* rrset changed, re-verify */ 746 } 747 *is_secure_answer = 0; 748 } 749 } else *is_secure_answer = 0; 750 751 edns->edns_version = EDNS_ADVERTISED_VERSION; 752 edns->udp_size = EDNS_ADVERTISED_SIZE; 753 edns->ext_rcode = 0; 754 edns->bits &= EDNS_DO; 755 if(worker->env.cfg->disable_edns_do && (edns->bits & EDNS_DO)) 756 edns->edns_present = 0; 757 *alias_rrset = NULL; /* avoid confusion if caller set it to non-NULL */ 758 if((worker->daemon->use_response_ip || worker->daemon->use_rpz) && 759 !partial_rep && !apply_respip_action(worker, qinfo, cinfo, rep, 760 &repinfo->client_addr, repinfo->client_addrlen, alias_rrset, 761 &encode_rep, worker->env.auth_zones)) { 762 goto bail_out; 763 } else if(partial_rep && 764 !respip_merge_cname(partial_rep, qinfo, rep, cinfo, 765 must_validate, &encode_rep, worker->scratchpad, 766 worker->env.auth_zones)) { 767 goto bail_out; 768 } 769 if(encode_rep != rep) { 770 /* if rewritten, it can't be considered "secure" */ 771 *is_secure_answer = 0; 772 } 773 if(!encode_rep || *alias_rrset) { 774 if(!encode_rep) 775 *need_drop = 1; 776 else { 777 /* If a partial CNAME chain is found, we first need to 778 * make a copy of the reply in the scratchpad so we 779 * can release the locks and lookup the cache again. */ 780 *partial_repp = reply_info_copy(encode_rep, NULL, 781 worker->scratchpad); 782 if(!*partial_repp) 783 goto bail_out; 784 } 785 } else { 786 if(*is_expired_answer == 1 && 787 worker->env.cfg->ede_serve_expired && worker->env.cfg->ede) { 788 EDNS_OPT_LIST_APPEND_EDE(&edns->opt_list_out, 789 worker->scratchpad, LDNS_EDE_STALE_ANSWER, ""); 790 } 791 /* Attach the cached EDE (RFC8914) if CD bit is set and the 792 * answer is bogus. */ 793 if(*is_secure_answer == 0 && 794 worker->env.cfg->ede && has_cd_bit && 795 encode_rep->reason_bogus != LDNS_EDE_NONE) { 796 edns_opt_list_append_ede(&edns->opt_list_out, 797 worker->scratchpad, encode_rep->reason_bogus, 798 encode_rep->reason_bogus_str); 799 } 800 if(!inplace_cb_reply_cache_call(&worker->env, qinfo, NULL, encode_rep, 801 (int)(flags&LDNS_RCODE_MASK), edns, repinfo, worker->scratchpad, 802 worker->env.now_tv)) 803 goto bail_out; 804 if(!reply_info_answer_encode(qinfo, encode_rep, id, flags, 805 repinfo->c->buffer, timenow, 1, worker->scratchpad, 806 udpsize, edns, (int)(edns->bits & EDNS_DO), 807 *is_secure_answer)) { 808 if(!inplace_cb_reply_servfail_call(&worker->env, qinfo, 809 NULL, NULL, LDNS_RCODE_SERVFAIL, edns, repinfo, 810 worker->scratchpad, worker->env.now_tv)) 811 edns->opt_list_inplace_cb_out = NULL; 812 error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL, 813 qinfo, id, flags, edns); 814 } 815 } 816 /* cannot send the reply right now, because blocking network syscall 817 * is bad while holding locks. */ 818 rrset_array_unlock_touch(worker->env.rrset_cache, worker->scratchpad, 819 rep->ref, rep->rrset_count); 820 /* go and return this buffer to the client */ 821 return 1; 822 823 bail_out: 824 rrset_array_unlock_touch(worker->env.rrset_cache, 825 worker->scratchpad, rep->ref, rep->rrset_count); 826 return 0; 827 } 828 829 /** Reply to client and perform prefetch to keep cache up to date. */ 830 static void 831 reply_and_prefetch(struct worker* worker, struct query_info* qinfo, 832 uint16_t flags, struct comm_reply* repinfo, time_t leeway, int noreply, 833 int rpz_passthru, struct edns_option* opt_list) 834 { 835 (void)opt_list; 836 /* first send answer to client to keep its latency 837 * as small as a cachereply */ 838 if(!noreply) { 839 if(repinfo->c->tcp_req_info) { 840 sldns_buffer_copy( 841 repinfo->c->tcp_req_info->spool_buffer, 842 repinfo->c->buffer); 843 } 844 comm_point_send_reply(repinfo); 845 } 846 server_stats_prefetch(&worker->stats, worker); 847 #ifdef CLIENT_SUBNET 848 /* Check if the subnet module is enabled. In that case pass over the 849 * comm_reply information for ECS generation later. The mesh states are 850 * unique when subnet is enabled. */ 851 if(modstack_find(&worker->env.mesh->mods, "subnetcache") != -1 852 && worker->env.unique_mesh) { 853 mesh_new_prefetch(worker->env.mesh, qinfo, flags, leeway + 854 PREFETCH_EXPIRY_ADD, rpz_passthru, 855 &repinfo->client_addr, opt_list); 856 return; 857 } 858 #endif 859 /* create the prefetch in the mesh as a normal lookup without 860 * client addrs waiting, which has the cache blacklisted (to bypass 861 * the cache and go to the network for the data). */ 862 /* this (potentially) runs the mesh for the new query */ 863 mesh_new_prefetch(worker->env.mesh, qinfo, flags, leeway + 864 PREFETCH_EXPIRY_ADD, rpz_passthru, NULL, NULL); 865 } 866 867 /** 868 * Fill CH class answer into buffer. Keeps query. 869 * @param pkt: buffer 870 * @param str: string to put into text record (<255). 871 * array of strings, every string becomes a text record. 872 * @param num: number of strings in array. 873 * @param edns: edns reply information. 874 * @param worker: worker with scratch region. 875 * @param repinfo: reply information for a communication point. 876 */ 877 static void 878 chaos_replystr(sldns_buffer* pkt, char** str, int num, struct edns_data* edns, 879 struct worker* worker, struct comm_reply* repinfo) 880 { 881 int i; 882 unsigned int rd = LDNS_RD_WIRE(sldns_buffer_begin(pkt)); 883 unsigned int cd = LDNS_CD_WIRE(sldns_buffer_begin(pkt)); 884 size_t udpsize = edns->udp_size; 885 edns->edns_version = EDNS_ADVERTISED_VERSION; 886 edns->udp_size = EDNS_ADVERTISED_SIZE; 887 edns->bits &= EDNS_DO; 888 if(!inplace_cb_reply_local_call(&worker->env, NULL, NULL, NULL, 889 LDNS_RCODE_NOERROR, edns, repinfo, worker->scratchpad, 890 worker->env.now_tv)) 891 edns->opt_list_inplace_cb_out = NULL; 892 sldns_buffer_clear(pkt); 893 sldns_buffer_skip(pkt, (ssize_t)sizeof(uint16_t)); /* skip id */ 894 sldns_buffer_write_u16(pkt, (uint16_t)(BIT_QR|BIT_RA)); 895 if(rd) LDNS_RD_SET(sldns_buffer_begin(pkt)); 896 if(cd) LDNS_CD_SET(sldns_buffer_begin(pkt)); 897 sldns_buffer_write_u16(pkt, 1); /* qdcount */ 898 sldns_buffer_write_u16(pkt, (uint16_t)num); /* ancount */ 899 sldns_buffer_write_u16(pkt, 0); /* nscount */ 900 sldns_buffer_write_u16(pkt, 0); /* arcount */ 901 (void)query_dname_len(pkt); /* skip qname */ 902 sldns_buffer_skip(pkt, (ssize_t)sizeof(uint16_t)); /* skip qtype */ 903 sldns_buffer_skip(pkt, (ssize_t)sizeof(uint16_t)); /* skip qclass */ 904 for(i=0; i<num; i++) { 905 size_t len = strlen(str[i]); 906 if(len>255) len=255; /* cap size of TXT record */ 907 if(sldns_buffer_position(pkt)+2+2+2+4+2+1+len+ 908 calc_edns_field_size(edns) > udpsize) { 909 sldns_buffer_write_u16_at(pkt, 6, i); /* ANCOUNT */ 910 LDNS_TC_SET(sldns_buffer_begin(pkt)); 911 break; 912 } 913 sldns_buffer_write_u16(pkt, 0xc00c); /* compr ptr to query */ 914 sldns_buffer_write_u16(pkt, LDNS_RR_TYPE_TXT); 915 sldns_buffer_write_u16(pkt, LDNS_RR_CLASS_CH); 916 sldns_buffer_write_u32(pkt, 0); /* TTL */ 917 sldns_buffer_write_u16(pkt, sizeof(uint8_t) + len); 918 sldns_buffer_write_u8(pkt, len); 919 sldns_buffer_write(pkt, str[i], len); 920 } 921 sldns_buffer_flip(pkt); 922 if(sldns_buffer_capacity(pkt) >= 923 sldns_buffer_limit(pkt)+calc_edns_field_size(edns)) 924 attach_edns_record(pkt, edns); 925 } 926 927 /** Reply with one string */ 928 static void 929 chaos_replyonestr(sldns_buffer* pkt, const char* str, struct edns_data* edns, 930 struct worker* worker, struct comm_reply* repinfo) 931 { 932 chaos_replystr(pkt, (char**)&str, 1, edns, worker, repinfo); 933 } 934 935 /** 936 * Create CH class trustanchor answer. 937 * @param pkt: buffer 938 * @param edns: edns reply information. 939 * @param w: worker with scratch region. 940 * @param repinfo: reply information for a communication point. 941 */ 942 static void 943 chaos_trustanchor(sldns_buffer* pkt, struct edns_data* edns, struct worker* w, 944 struct comm_reply* repinfo) 945 { 946 #define TA_RESPONSE_MAX_TXT 16 /* max number of TXT records */ 947 #define TA_RESPONSE_MAX_TAGS 32 /* max number of tags printed per zone */ 948 char* str_array[TA_RESPONSE_MAX_TXT]; 949 uint16_t tags[TA_RESPONSE_MAX_TAGS]; 950 int num = 0; 951 struct trust_anchor* ta; 952 953 if(!w->env.need_to_validate) { 954 /* no validator module, reply no trustanchors */ 955 chaos_replystr(pkt, NULL, 0, edns, w, repinfo); 956 return; 957 } 958 959 /* fill the string with contents */ 960 lock_basic_lock(&w->env.anchors->lock); 961 RBTREE_FOR(ta, struct trust_anchor*, w->env.anchors->tree) { 962 char* str; 963 size_t i, numtag, str_len = 255; 964 if(num == TA_RESPONSE_MAX_TXT) continue; 965 str = (char*)regional_alloc(w->scratchpad, str_len); 966 if(!str) continue; 967 lock_basic_lock(&ta->lock); 968 numtag = anchor_list_keytags(ta, tags, TA_RESPONSE_MAX_TAGS); 969 if(numtag == 0) { 970 /* empty, insecure point */ 971 lock_basic_unlock(&ta->lock); 972 continue; 973 } 974 str_array[num] = str; 975 num++; 976 977 /* spool name of anchor */ 978 (void)sldns_wire2str_dname_buf(ta->name, ta->namelen, str, str_len); 979 str_len -= strlen(str); str += strlen(str); 980 /* spool tags */ 981 for(i=0; i<numtag; i++) { 982 snprintf(str, str_len, " %u", (unsigned)tags[i]); 983 str_len -= strlen(str); str += strlen(str); 984 } 985 lock_basic_unlock(&ta->lock); 986 } 987 lock_basic_unlock(&w->env.anchors->lock); 988 989 chaos_replystr(pkt, str_array, num, edns, w, repinfo); 990 regional_free_all(w->scratchpad); 991 } 992 993 /** 994 * Answer CH class queries. 995 * @param w: worker 996 * @param qinfo: query info. Pointer into packet buffer. 997 * @param edns: edns info from query. 998 * @param repinfo: reply information for a communication point. 999 * @param pkt: packet buffer. 1000 * @return: true if a reply is to be sent. 1001 */ 1002 static int 1003 answer_chaos(struct worker* w, struct query_info* qinfo, 1004 struct edns_data* edns, struct comm_reply* repinfo, sldns_buffer* pkt) 1005 { 1006 struct config_file* cfg = w->env.cfg; 1007 if(qinfo->qtype != LDNS_RR_TYPE_ANY && qinfo->qtype != LDNS_RR_TYPE_TXT) 1008 return 0; 1009 if(query_dname_compare(qinfo->qname, 1010 (uint8_t*)"\002id\006server") == 0 || 1011 query_dname_compare(qinfo->qname, 1012 (uint8_t*)"\010hostname\004bind") == 0) 1013 { 1014 if(cfg->hide_identity) 1015 return 0; 1016 if(cfg->identity==NULL || cfg->identity[0]==0) { 1017 char buf[MAXHOSTNAMELEN+1]; 1018 if (gethostname(buf, MAXHOSTNAMELEN) == 0) { 1019 buf[MAXHOSTNAMELEN] = 0; 1020 chaos_replyonestr(pkt, buf, edns, w, repinfo); 1021 } else { 1022 log_err("gethostname: %s", strerror(errno)); 1023 chaos_replyonestr(pkt, "no hostname", edns, w, repinfo); 1024 } 1025 } 1026 else chaos_replyonestr(pkt, cfg->identity, edns, w, repinfo); 1027 return 1; 1028 } 1029 if(query_dname_compare(qinfo->qname, 1030 (uint8_t*)"\007version\006server") == 0 || 1031 query_dname_compare(qinfo->qname, 1032 (uint8_t*)"\007version\004bind") == 0) 1033 { 1034 if(cfg->hide_version) 1035 return 0; 1036 if(cfg->version==NULL || cfg->version[0]==0) 1037 chaos_replyonestr(pkt, PACKAGE_STRING, edns, w, repinfo); 1038 else chaos_replyonestr(pkt, cfg->version, edns, w, repinfo); 1039 return 1; 1040 } 1041 if(query_dname_compare(qinfo->qname, 1042 (uint8_t*)"\013trustanchor\007unbound") == 0) 1043 { 1044 if(cfg->hide_trustanchor) 1045 return 0; 1046 chaos_trustanchor(pkt, edns, w, repinfo); 1047 return 1; 1048 } 1049 1050 return 0; 1051 } 1052 1053 /** 1054 * Answer notify queries. These are notifies for authoritative zones, 1055 * the reply is an ack that the notify has been received. We need to check 1056 * access permission here. 1057 * @param w: worker 1058 * @param qinfo: query info. Pointer into packet buffer. 1059 * @param edns: edns info from query. 1060 * @param addr: client address. 1061 * @param addrlen: client address length. 1062 * @param pkt: packet buffer. 1063 */ 1064 static void 1065 answer_notify(struct worker* w, struct query_info* qinfo, 1066 struct edns_data* edns, sldns_buffer* pkt, 1067 struct sockaddr_storage* addr, socklen_t addrlen) 1068 { 1069 int refused = 0; 1070 int rcode = LDNS_RCODE_NOERROR; 1071 uint32_t serial = 0; 1072 int has_serial; 1073 if(!w->env.auth_zones) return; 1074 has_serial = auth_zone_parse_notify_serial(pkt, &serial); 1075 if(auth_zones_notify(w->env.auth_zones, &w->env, qinfo->qname, 1076 qinfo->qname_len, qinfo->qclass, addr, 1077 addrlen, has_serial, serial, &refused)) { 1078 rcode = LDNS_RCODE_NOERROR; 1079 } else { 1080 if(refused) 1081 rcode = LDNS_RCODE_REFUSED; 1082 else rcode = LDNS_RCODE_SERVFAIL; 1083 } 1084 1085 if(verbosity >= VERB_DETAIL) { 1086 char buf[380]; 1087 char zname[255+1]; 1088 char sr[25]; 1089 dname_str(qinfo->qname, zname); 1090 sr[0]=0; 1091 if(has_serial) 1092 snprintf(sr, sizeof(sr), "serial %u ", 1093 (unsigned)serial); 1094 if(rcode == LDNS_RCODE_REFUSED) 1095 snprintf(buf, sizeof(buf), 1096 "refused NOTIFY %sfor %s from", sr, zname); 1097 else if(rcode == LDNS_RCODE_SERVFAIL) 1098 snprintf(buf, sizeof(buf), 1099 "servfail for NOTIFY %sfor %s from", sr, zname); 1100 else snprintf(buf, sizeof(buf), 1101 "received NOTIFY %sfor %s from", sr, zname); 1102 log_addr(VERB_DETAIL, buf, addr, addrlen); 1103 } 1104 edns->edns_version = EDNS_ADVERTISED_VERSION; 1105 edns->udp_size = EDNS_ADVERTISED_SIZE; 1106 edns->ext_rcode = 0; 1107 edns->bits &= EDNS_DO; 1108 error_encode(pkt, rcode, qinfo, 1109 *(uint16_t*)(void *)sldns_buffer_begin(pkt), 1110 sldns_buffer_read_u16_at(pkt, 2), edns); 1111 LDNS_OPCODE_SET(sldns_buffer_begin(pkt), LDNS_PACKET_NOTIFY); 1112 } 1113 1114 static int 1115 deny_refuse(struct comm_point* c, enum acl_access acl, 1116 enum acl_access deny, enum acl_access refuse, 1117 struct worker* worker, struct comm_reply* repinfo, 1118 struct acl_addr* acladdr, int ede, 1119 struct check_request_result* check_result) 1120 { 1121 if(acl == deny) { 1122 if(verbosity >= VERB_ALGO) { 1123 log_acl_action("dropped", &repinfo->client_addr, 1124 repinfo->client_addrlen, acl, acladdr); 1125 log_buf(VERB_ALGO, "dropped", c->buffer); 1126 } 1127 comm_point_drop_reply(repinfo); 1128 if(worker->stats.extended) 1129 worker->stats.unwanted_queries++; 1130 return 0; 1131 } else if(acl == refuse) { 1132 size_t opt_rr_mark; 1133 1134 if(verbosity >= VERB_ALGO) { 1135 log_acl_action("refused", &repinfo->client_addr, 1136 repinfo->client_addrlen, acl, acladdr); 1137 log_buf(VERB_ALGO, "refuse", c->buffer); 1138 } 1139 1140 if(worker->stats.extended) 1141 worker->stats.unwanted_queries++; 1142 worker_check_request(c->buffer, worker, check_result); 1143 if(check_result->value != 0) { 1144 if(check_result->value != -1) { 1145 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1146 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1147 check_result->value); 1148 return 1; 1149 } 1150 comm_point_drop_reply(repinfo); 1151 return 0; 1152 } 1153 /* worker_check_request() above guarantees that the buffer contains at 1154 * least a header and that qdcount == 1 1155 */ 1156 log_assert(sldns_buffer_limit(c->buffer) >= LDNS_HEADER_SIZE 1157 && LDNS_QDCOUNT(sldns_buffer_begin(c->buffer)) == 1); 1158 1159 sldns_buffer_set_position(c->buffer, LDNS_HEADER_SIZE); /* skip header */ 1160 1161 /* check additional section is present and that we respond with EDEs */ 1162 if(LDNS_ARCOUNT(sldns_buffer_begin(c->buffer)) != 1 1163 || !ede) { 1164 LDNS_QDCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1165 LDNS_ANCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1166 LDNS_NSCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1167 LDNS_ARCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1168 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1169 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1170 LDNS_RCODE_REFUSED); 1171 sldns_buffer_set_position(c->buffer, LDNS_HEADER_SIZE); 1172 sldns_buffer_flip(c->buffer); 1173 return 1; 1174 } 1175 1176 if (!query_dname_len(c->buffer)) { 1177 LDNS_QDCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1178 LDNS_ANCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1179 LDNS_NSCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1180 LDNS_ARCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1181 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1182 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1183 LDNS_RCODE_FORMERR); 1184 sldns_buffer_set_position(c->buffer, LDNS_HEADER_SIZE); 1185 sldns_buffer_flip(c->buffer); 1186 return 1; 1187 } 1188 /* space available for query type and class? */ 1189 if (sldns_buffer_remaining(c->buffer) < 2 * sizeof(uint16_t)) { 1190 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1191 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1192 LDNS_RCODE_FORMERR); 1193 LDNS_QDCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1194 LDNS_ANCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1195 LDNS_NSCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1196 LDNS_ARCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1197 sldns_buffer_set_position(c->buffer, LDNS_HEADER_SIZE); 1198 sldns_buffer_flip(c->buffer); 1199 return 1; 1200 } 1201 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1202 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1203 LDNS_RCODE_REFUSED); 1204 1205 sldns_buffer_skip(c->buffer, (ssize_t)sizeof(uint16_t)); /* skip qtype */ 1206 1207 sldns_buffer_skip(c->buffer, (ssize_t)sizeof(uint16_t)); /* skip qclass */ 1208 1209 /* The OPT RR to be returned should come directly after 1210 * the query, so mark this spot. 1211 */ 1212 opt_rr_mark = sldns_buffer_position(c->buffer); 1213 1214 /* Skip through the RR records */ 1215 if(LDNS_ANCOUNT(sldns_buffer_begin(c->buffer)) != 0 || 1216 LDNS_NSCOUNT(sldns_buffer_begin(c->buffer)) != 0) { 1217 if(!skip_pkt_rrs(c->buffer, 1218 ((int)LDNS_ANCOUNT(sldns_buffer_begin(c->buffer)))+ 1219 ((int)LDNS_NSCOUNT(sldns_buffer_begin(c->buffer))))) { 1220 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1221 LDNS_RCODE_FORMERR); 1222 LDNS_ANCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1223 LDNS_NSCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1224 LDNS_ARCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1225 sldns_buffer_set_position(c->buffer, opt_rr_mark); 1226 sldns_buffer_flip(c->buffer); 1227 return 1; 1228 } 1229 } 1230 /* Do we have a valid OPT RR here? If not return REFUSED (could be a valid TSIG or something so no FORMERR) */ 1231 /* domain name must be the root of length 1. */ 1232 if(sldns_buffer_remaining(c->buffer) < 1 || *sldns_buffer_current(c->buffer) != 0) { 1233 LDNS_ANCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1234 LDNS_NSCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1235 LDNS_ARCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1236 sldns_buffer_set_position(c->buffer, opt_rr_mark); 1237 sldns_buffer_flip(c->buffer); 1238 return 1; 1239 } else { 1240 sldns_buffer_skip(c->buffer, 1); /* skip root label */ 1241 } 1242 if(sldns_buffer_remaining(c->buffer) < 2 || 1243 sldns_buffer_read_u16(c->buffer) != LDNS_RR_TYPE_OPT) { 1244 LDNS_ANCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1245 LDNS_NSCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1246 LDNS_ARCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1247 sldns_buffer_set_position(c->buffer, opt_rr_mark); 1248 sldns_buffer_flip(c->buffer); 1249 return 1; 1250 } 1251 /* Write OPT RR directly after the query, 1252 * so without the (possibly skipped) Answer and NS RRs 1253 */ 1254 LDNS_ANCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1255 LDNS_NSCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1256 sldns_buffer_clear(c->buffer); /* reset write limit */ 1257 sldns_buffer_set_position(c->buffer, opt_rr_mark); 1258 1259 /* Check if OPT record can be written 1260 * 17 == root label (1) + RR type (2) + UDP Size (2) 1261 * + Fields (4) + rdata len (2) + EDE Option code (2) 1262 * + EDE Option length (2) + EDE info-code (2) 1263 */ 1264 if (sldns_buffer_available(c->buffer, 17) == 0) { 1265 LDNS_ARCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1266 sldns_buffer_flip(c->buffer); 1267 return 1; 1268 } 1269 1270 LDNS_ARCOUNT_SET(sldns_buffer_begin(c->buffer), 1); 1271 1272 /* root label */ 1273 sldns_buffer_write_u8(c->buffer, 0); 1274 sldns_buffer_write_u16(c->buffer, LDNS_RR_TYPE_OPT); 1275 sldns_buffer_write_u16(c->buffer, EDNS_ADVERTISED_SIZE); 1276 1277 /* write OPT Record TTL Field */ 1278 sldns_buffer_write_u32(c->buffer, 0); 1279 1280 /* write rdata len: EDE option + length + info-code */ 1281 sldns_buffer_write_u16(c->buffer, 6); 1282 1283 /* write OPTIONS; add EDE option code */ 1284 sldns_buffer_write_u16(c->buffer, LDNS_EDNS_EDE); 1285 1286 /* write single EDE option length (for just 1 info-code) */ 1287 sldns_buffer_write_u16(c->buffer, 2); 1288 1289 /* write single EDE info-code */ 1290 sldns_buffer_write_u16(c->buffer, LDNS_EDE_PROHIBITED); 1291 1292 sldns_buffer_flip(c->buffer); 1293 1294 verbose(VERB_ALGO, "attached EDE code: %d", LDNS_EDE_PROHIBITED); 1295 1296 return 1; 1297 1298 } 1299 1300 return -1; 1301 } 1302 1303 static int 1304 deny_refuse_all(struct comm_point* c, enum acl_access* acl, 1305 struct worker* worker, struct comm_reply* repinfo, 1306 struct acl_addr** acladdr, int ede, int check_proxy, 1307 struct check_request_result* check_result) 1308 { 1309 if(check_proxy) { 1310 *acladdr = acl_addr_lookup(worker->daemon->acl, 1311 &repinfo->remote_addr, repinfo->remote_addrlen); 1312 } else { 1313 *acladdr = acl_addr_lookup(worker->daemon->acl, 1314 &repinfo->client_addr, repinfo->client_addrlen); 1315 } 1316 /* If there is no ACL based on client IP use the interface ACL. */ 1317 if(!(*acladdr) && c->socket) { 1318 *acladdr = c->socket->acl; 1319 } 1320 *acl = acl_get_control(*acladdr); 1321 return deny_refuse(c, *acl, acl_deny, acl_refuse, worker, repinfo, 1322 *acladdr, ede, check_result); 1323 } 1324 1325 static int 1326 deny_refuse_non_local(struct comm_point* c, enum acl_access acl, 1327 struct worker* worker, struct comm_reply* repinfo, 1328 struct acl_addr* acladdr, int ede, 1329 struct check_request_result* check_result) 1330 { 1331 return deny_refuse(c, acl, acl_deny_non_local, acl_refuse_non_local, 1332 worker, repinfo, acladdr, ede, check_result); 1333 } 1334 1335 /* Check if the query is blocked by source IP rate limiting. 1336 * Returns 1 if it passes the check, 0 otherwise. */ 1337 static int 1338 check_ip_ratelimit(struct worker* worker, struct sockaddr_storage* addr, 1339 socklen_t addrlen, int has_cookie, sldns_buffer* pkt) 1340 { 1341 if(!infra_ip_ratelimit_inc(worker->env.infra_cache, addr, addrlen, 1342 *worker->env.now, has_cookie, 1343 worker->env.cfg->ip_ratelimit_backoff, pkt)) { 1344 /* See if we can pass through with slip factor */ 1345 if(!has_cookie && worker->env.cfg->ip_ratelimit_factor != 0 && 1346 ub_random_max(worker->env.rnd, 1347 worker->env.cfg->ip_ratelimit_factor) == 0) { 1348 char addrbuf[128]; 1349 addr_to_str(addr, addrlen, addrbuf, sizeof(addrbuf)); 1350 verbose(VERB_QUERY, "ip_ratelimit allowed through for " 1351 "ip address %s because of slip in " 1352 "ip_ratelimit_factor", addrbuf); 1353 return 1; 1354 } 1355 return 0; 1356 } 1357 return 1; 1358 } 1359 1360 int 1361 worker_handle_request(struct comm_point* c, void* arg, int error, 1362 struct comm_reply* repinfo) 1363 { 1364 struct worker* worker = (struct worker*)arg; 1365 int ret; 1366 hashvalue_type h; 1367 struct lruhash_entry* e; 1368 struct query_info qinfo; 1369 struct edns_data edns; 1370 struct edns_option* original_edns_list = NULL; 1371 enum acl_access acl; 1372 struct acl_addr* acladdr; 1373 int pre_edns_ip_ratelimit = 1; 1374 int rc = 0; 1375 int need_drop = 0; 1376 int is_expired_answer = 0; 1377 int is_secure_answer = 0; 1378 int rpz_passthru = 0; 1379 long long wait_queue_time = 0; 1380 /* We might have to chase a CNAME chain internally, in which case 1381 * we'll have up to two replies and combine them to build a complete 1382 * answer. These variables control this case. */ 1383 struct ub_packed_rrset_key* alias_rrset = NULL; 1384 struct reply_info* partial_rep = NULL; 1385 struct query_info* lookup_qinfo = &qinfo; 1386 struct query_info qinfo_tmp; /* placeholder for lookup_qinfo */ 1387 struct respip_client_info* cinfo = NULL, cinfo_tmp; 1388 struct timeval wait_time; 1389 struct check_request_result check_result = {0,0}; 1390 memset(&qinfo, 0, sizeof(qinfo)); 1391 1392 if((error != NETEVENT_NOERROR && error != NETEVENT_DONE)|| !repinfo) { 1393 /* some bad tcp query DNS formats give these error calls */ 1394 verbose(VERB_ALGO, "handle request called with err=%d", error); 1395 return 0; 1396 } 1397 1398 if (worker->env.cfg->sock_queue_timeout && timeval_isset(&c->recv_tv)) { 1399 timeval_subtract(&wait_time, worker->env.now_tv, &c->recv_tv); 1400 wait_queue_time = wait_time.tv_sec * 1000000 + wait_time.tv_usec; 1401 if (worker->stats.max_query_time_us < wait_queue_time) 1402 worker->stats.max_query_time_us = wait_queue_time; 1403 if(wait_queue_time > 1404 (long long)(worker->env.cfg->sock_queue_timeout * 1000000)) { 1405 /* count and drop queries that were sitting in the socket queue too long */ 1406 worker->stats.num_queries_timed_out++; 1407 return 0; 1408 } 1409 } 1410 1411 #ifdef USE_DNSCRYPT 1412 repinfo->max_udp_size = worker->daemon->cfg->max_udp_size; 1413 if(!dnsc_handle_curved_request(worker->daemon->dnscenv, repinfo)) { 1414 worker->stats.num_query_dnscrypt_crypted_malformed++; 1415 return 0; 1416 } 1417 if(c->dnscrypt && !repinfo->is_dnscrypted) { 1418 char buf[LDNS_MAX_DOMAINLEN+1]; 1419 /* Check if this is unencrypted and asking for certs */ 1420 worker_check_request(c->buffer, worker, &check_result); 1421 if(check_result.value != 0) { 1422 verbose(VERB_ALGO, 1423 "dnscrypt: worker check request: bad query."); 1424 log_addr(VERB_CLIENT,"from",&repinfo->client_addr, 1425 repinfo->client_addrlen); 1426 comm_point_drop_reply(repinfo); 1427 return 0; 1428 } 1429 if(!query_info_parse(&qinfo, c->buffer)) { 1430 verbose(VERB_ALGO, 1431 "dnscrypt: worker parse request: formerror."); 1432 log_addr(VERB_CLIENT, "from", &repinfo->client_addr, 1433 repinfo->client_addrlen); 1434 comm_point_drop_reply(repinfo); 1435 return 0; 1436 } 1437 dname_str(qinfo.qname, buf); 1438 if(!(qinfo.qtype == LDNS_RR_TYPE_TXT && 1439 strcasecmp(buf, 1440 worker->daemon->dnscenv->provider_name) == 0)) { 1441 verbose(VERB_ALGO, 1442 "dnscrypt: not TXT \"%s\". Received: %s \"%s\"", 1443 worker->daemon->dnscenv->provider_name, 1444 sldns_rr_descript(qinfo.qtype)->_name, 1445 buf); 1446 comm_point_drop_reply(repinfo); 1447 worker->stats.num_query_dnscrypt_cleartext++; 1448 return 0; 1449 } 1450 worker->stats.num_query_dnscrypt_cert++; 1451 sldns_buffer_rewind(c->buffer); 1452 } else if(c->dnscrypt && repinfo->is_dnscrypted) { 1453 worker->stats.num_query_dnscrypt_crypted++; 1454 } 1455 #endif 1456 #ifdef USE_DNSTAP 1457 /* 1458 * sending src (client)/dst (local service) addresses over DNSTAP from incoming request handler 1459 */ 1460 if(worker->dtenv.log_client_query_messages) { 1461 log_addr(VERB_ALGO, "request from client", &repinfo->client_addr, repinfo->client_addrlen); 1462 log_addr(VERB_ALGO, "to local addr", (void*)repinfo->c->socket->addr, repinfo->c->socket->addrlen); 1463 dt_msg_send_client_query(&worker->dtenv, &repinfo->client_addr, (void*)repinfo->c->socket->addr, c->type, c->ssl, c->buffer, 1464 ((worker->env.cfg->sock_queue_timeout && timeval_isset(&c->recv_tv))?&c->recv_tv:NULL)); 1465 } 1466 #endif 1467 /* Check deny/refuse ACLs */ 1468 if(repinfo->is_proxied) { 1469 if((ret=deny_refuse_all(c, &acl, worker, repinfo, &acladdr, 1470 worker->env.cfg->ede, 1, &check_result)) != -1) { 1471 if(ret == 1) 1472 goto send_reply; 1473 return ret; 1474 } 1475 } 1476 if((ret=deny_refuse_all(c, &acl, worker, repinfo, &acladdr, 1477 worker->env.cfg->ede, 0, &check_result)) != -1) { 1478 if(ret == 1) 1479 goto send_reply; 1480 return ret; 1481 } 1482 1483 worker_check_request(c->buffer, worker, &check_result); 1484 if(check_result.value != 0) { 1485 verbose(VERB_ALGO, "worker check request: bad query."); 1486 log_addr(VERB_CLIENT,"from",&repinfo->client_addr, repinfo->client_addrlen); 1487 if(check_result.value != -1) { 1488 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1489 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1490 check_result.value); 1491 return 1; 1492 } 1493 comm_point_drop_reply(repinfo); 1494 return 0; 1495 } 1496 1497 worker->stats.num_queries++; 1498 pre_edns_ip_ratelimit = !worker->env.cfg->do_answer_cookie 1499 || sldns_buffer_limit(c->buffer) < LDNS_HEADER_SIZE 1500 || LDNS_ARCOUNT(sldns_buffer_begin(c->buffer)) == 0; 1501 1502 /* If the IP rate limiting check needs extra EDNS information (e.g., 1503 * DNS Cookies) postpone the check until after EDNS is parsed. */ 1504 if(pre_edns_ip_ratelimit) { 1505 /* NOTE: we always check the repinfo->client_address. 1506 * IP ratelimiting is implicitly disabled for proxies. */ 1507 if(!check_ip_ratelimit(worker, &repinfo->client_addr, 1508 repinfo->client_addrlen, 0, c->buffer)) { 1509 worker->stats.num_queries_ip_ratelimited++; 1510 comm_point_drop_reply(repinfo); 1511 return 0; 1512 } 1513 } 1514 1515 if(!query_info_parse(&qinfo, c->buffer)) { 1516 verbose(VERB_ALGO, "worker parse request: formerror."); 1517 log_addr(VERB_CLIENT, "from", &repinfo->client_addr, 1518 repinfo->client_addrlen); 1519 memset(&qinfo, 0, sizeof(qinfo)); /* zero qinfo.qname */ 1520 if(worker_err_ratelimit(worker, LDNS_RCODE_FORMERR) == -1) { 1521 comm_point_drop_reply(repinfo); 1522 return 0; 1523 } 1524 sldns_buffer_rewind(c->buffer); 1525 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1526 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1527 LDNS_RCODE_FORMERR); 1528 goto send_reply; 1529 } 1530 if(worker->env.cfg->log_queries) { 1531 char ip[128]; 1532 addr_to_str(&repinfo->client_addr, repinfo->client_addrlen, ip, sizeof(ip)); 1533 log_query_in(ip, qinfo.qname, qinfo.qtype, qinfo.qclass); 1534 } 1535 if(qinfo.qtype == LDNS_RR_TYPE_AXFR || 1536 qinfo.qtype == LDNS_RR_TYPE_IXFR) { 1537 verbose(VERB_ALGO, "worker request: refused zone transfer."); 1538 log_addr(VERB_CLIENT, "from", &repinfo->client_addr, 1539 repinfo->client_addrlen); 1540 sldns_buffer_rewind(c->buffer); 1541 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1542 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1543 LDNS_RCODE_REFUSED); 1544 if(worker->stats.extended) { 1545 worker->stats.qtype[qinfo.qtype]++; 1546 } 1547 goto send_reply; 1548 } 1549 if(qinfo.qtype == LDNS_RR_TYPE_OPT || 1550 qinfo.qtype == LDNS_RR_TYPE_TSIG || 1551 qinfo.qtype == LDNS_RR_TYPE_TKEY || 1552 qinfo.qtype == LDNS_RR_TYPE_MAILA || 1553 qinfo.qtype == LDNS_RR_TYPE_MAILB || 1554 (qinfo.qtype >= 128 && qinfo.qtype <= 248)) { 1555 verbose(VERB_ALGO, "worker request: formerror for meta-type."); 1556 log_addr(VERB_CLIENT, "from", &repinfo->client_addr, 1557 repinfo->client_addrlen); 1558 if(worker_err_ratelimit(worker, LDNS_RCODE_FORMERR) == -1) { 1559 comm_point_drop_reply(repinfo); 1560 return 0; 1561 } 1562 sldns_buffer_rewind(c->buffer); 1563 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1564 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1565 LDNS_RCODE_FORMERR); 1566 if(worker->stats.extended) { 1567 worker->stats.qtype[qinfo.qtype]++; 1568 } 1569 goto send_reply; 1570 } 1571 if((ret=parse_edns_from_query_pkt( 1572 c->buffer, &edns, worker->env.cfg, c, repinfo, 1573 (worker->env.now ? *worker->env.now : time(NULL)), 1574 worker->scratchpad)) != 0) { 1575 struct edns_data reply_edns; 1576 verbose(VERB_ALGO, "worker parse edns: formerror."); 1577 log_addr(VERB_CLIENT, "from", &repinfo->client_addr, 1578 repinfo->client_addrlen); 1579 memset(&reply_edns, 0, sizeof(reply_edns)); 1580 reply_edns.edns_present = 1; 1581 error_encode(c->buffer, ret, &qinfo, 1582 *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 1583 sldns_buffer_read_u16_at(c->buffer, 2), &reply_edns); 1584 regional_free_all(worker->scratchpad); 1585 goto send_reply; 1586 } 1587 if(edns.edns_present) { 1588 if(edns.edns_version != 0) { 1589 edns.opt_list_in = NULL; 1590 edns.opt_list_out = NULL; 1591 edns.opt_list_inplace_cb_out = NULL; 1592 verbose(VERB_ALGO, "query with bad edns version."); 1593 log_addr(VERB_CLIENT, "from", &repinfo->client_addr, 1594 repinfo->client_addrlen); 1595 extended_error_encode(c->buffer, EDNS_RCODE_BADVERS, &qinfo, 1596 *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 1597 sldns_buffer_read_u16_at(c->buffer, 2), 0, &edns); 1598 regional_free_all(worker->scratchpad); 1599 goto send_reply; 1600 } 1601 if(edns.udp_size < NORMAL_UDP_SIZE && 1602 worker->daemon->cfg->harden_short_bufsize) { 1603 verbose(VERB_QUERY, "worker request: EDNS bufsize %d ignored", 1604 (int)edns.udp_size); 1605 log_addr(VERB_CLIENT, "from", &repinfo->client_addr, 1606 repinfo->client_addrlen); 1607 edns.udp_size = NORMAL_UDP_SIZE; 1608 } 1609 } 1610 1611 /* Get stats for cookies */ 1612 server_stats_downstream_cookie(&worker->stats, &edns); 1613 1614 /* If the IP rate limiting check was postponed, check now. */ 1615 if(!pre_edns_ip_ratelimit) { 1616 /* NOTE: we always check the repinfo->client_address. 1617 * IP ratelimiting is implicitly disabled for proxies. */ 1618 if(!check_ip_ratelimit(worker, &repinfo->client_addr, 1619 repinfo->client_addrlen, edns.cookie_valid, 1620 c->buffer)) { 1621 worker->stats.num_queries_ip_ratelimited++; 1622 comm_point_drop_reply(repinfo); 1623 return 0; 1624 } 1625 } 1626 1627 /* "if, else if" sequence below deals with downstream DNS Cookies */ 1628 if(acl != acl_allow_cookie) 1629 ; /* pass; No cookie downstream processing whatsoever */ 1630 1631 else if(edns.cookie_valid) 1632 ; /* pass; Valid cookie is good! */ 1633 1634 else if(c->type != comm_udp) 1635 ; /* pass; Stateful transport */ 1636 1637 else if(edns.cookie_present) { 1638 /* Cookie present, but not valid: Cookie was bad! */ 1639 extended_error_encode(c->buffer, 1640 LDNS_EXT_RCODE_BADCOOKIE, &qinfo, 1641 *(uint16_t*)(void *) 1642 sldns_buffer_begin(c->buffer), 1643 sldns_buffer_read_u16_at(c->buffer, 2), 1644 0, &edns); 1645 regional_free_all(worker->scratchpad); 1646 goto send_reply; 1647 } else { 1648 /* Cookie required, but no cookie present on UDP */ 1649 verbose(VERB_ALGO, "worker request: " 1650 "need cookie or stateful transport"); 1651 log_addr(VERB_ALGO, "from",&repinfo->remote_addr 1652 , repinfo->remote_addrlen); 1653 EDNS_OPT_LIST_APPEND_EDE(&edns.opt_list_out, 1654 worker->scratchpad, LDNS_EDE_OTHER, 1655 "DNS Cookie needed for UDP replies"); 1656 error_encode(c->buffer, 1657 (LDNS_RCODE_REFUSED|BIT_TC), &qinfo, 1658 *(uint16_t*)(void *) 1659 sldns_buffer_begin(c->buffer), 1660 sldns_buffer_read_u16_at(c->buffer, 2), 1661 &edns); 1662 regional_free_all(worker->scratchpad); 1663 goto send_reply; 1664 } 1665 1666 if(edns.udp_size > worker->daemon->cfg->max_udp_size && 1667 c->type == comm_udp) { 1668 verbose(VERB_QUERY, 1669 "worker request: max UDP reply size modified" 1670 " (%d to max-udp-size)", (int)edns.udp_size); 1671 log_addr(VERB_CLIENT, "from", &repinfo->client_addr, 1672 repinfo->client_addrlen); 1673 edns.udp_size = worker->daemon->cfg->max_udp_size; 1674 } 1675 if(edns.udp_size < LDNS_HEADER_SIZE) { 1676 verbose(VERB_ALGO, "worker request: edns is too small."); 1677 log_addr(VERB_CLIENT, "from", &repinfo->client_addr, 1678 repinfo->client_addrlen); 1679 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1680 LDNS_TC_SET(sldns_buffer_begin(c->buffer)); 1681 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1682 LDNS_RCODE_SERVFAIL); 1683 sldns_buffer_set_position(c->buffer, LDNS_HEADER_SIZE); 1684 sldns_buffer_write_at(c->buffer, 4, 1685 (uint8_t*)"\0\0\0\0\0\0\0\0", 8); 1686 sldns_buffer_flip(c->buffer); 1687 regional_free_all(worker->scratchpad); 1688 goto send_reply; 1689 } 1690 if(worker->stats.extended) 1691 server_stats_insquery(&worker->stats, c, qinfo.qtype, 1692 qinfo.qclass, &edns, repinfo); 1693 if(c->type != comm_udp) 1694 edns.udp_size = 65535; /* max size for TCP replies */ 1695 if(qinfo.qclass == LDNS_RR_CLASS_CH && answer_chaos(worker, &qinfo, 1696 &edns, repinfo, c->buffer)) { 1697 regional_free_all(worker->scratchpad); 1698 goto send_reply; 1699 } 1700 if(LDNS_OPCODE_WIRE(sldns_buffer_begin(c->buffer)) == 1701 LDNS_PACKET_NOTIFY) { 1702 answer_notify(worker, &qinfo, &edns, c->buffer, 1703 &repinfo->client_addr, repinfo->client_addrlen); 1704 regional_free_all(worker->scratchpad); 1705 goto send_reply; 1706 } 1707 if(local_zones_answer(worker->daemon->local_zones, &worker->env, &qinfo, 1708 &edns, c->buffer, worker->scratchpad, repinfo, acladdr->taglist, 1709 acladdr->taglen, acladdr->tag_actions, 1710 acladdr->tag_actions_size, acladdr->tag_datas, 1711 acladdr->tag_datas_size, worker->daemon->cfg->tagname, 1712 worker->daemon->cfg->num_tags, acladdr->view)) { 1713 regional_free_all(worker->scratchpad); 1714 if(sldns_buffer_limit(c->buffer) == 0) { 1715 comm_point_drop_reply(repinfo); 1716 return 0; 1717 } 1718 goto send_reply; 1719 } 1720 if(worker->env.auth_zones && 1721 rpz_callback_from_worker_request(worker->env.auth_zones, 1722 &worker->env, &qinfo, &edns, c->buffer, worker->scratchpad, 1723 repinfo, acladdr->taglist, acladdr->taglen, &worker->stats, 1724 &rpz_passthru)) { 1725 regional_free_all(worker->scratchpad); 1726 if(sldns_buffer_limit(c->buffer) == 0) { 1727 comm_point_drop_reply(repinfo); 1728 return 0; 1729 } 1730 goto send_reply; 1731 } 1732 if(worker->env.auth_zones && 1733 auth_zones_answer(worker->env.auth_zones, &worker->env, 1734 &qinfo, &edns, repinfo, c->buffer, worker->scratchpad)) { 1735 regional_free_all(worker->scratchpad); 1736 if(sldns_buffer_limit(c->buffer) == 0) { 1737 comm_point_drop_reply(repinfo); 1738 return 0; 1739 } 1740 /* set RA for everyone that can have recursion (based on 1741 * access control list) */ 1742 if(LDNS_RD_WIRE(sldns_buffer_begin(c->buffer)) && 1743 acl != acl_deny_non_local && acl != acl_refuse_non_local) 1744 LDNS_RA_SET(sldns_buffer_begin(c->buffer)); 1745 goto send_reply; 1746 } 1747 1748 /* We've looked in our local zones. If the answer isn't there, we 1749 * might need to bail out based on ACLs now. */ 1750 if((ret=deny_refuse_non_local(c, acl, worker, repinfo, acladdr, 1751 worker->env.cfg->ede, &check_result)) != -1) 1752 { 1753 regional_free_all(worker->scratchpad); 1754 if(ret == 1) 1755 goto send_reply; 1756 return ret; 1757 } 1758 1759 /* If this request does not have the recursion bit set, verify 1760 * ACLs allow the recursion bit to be treated as set. */ 1761 if(!(LDNS_RD_WIRE(sldns_buffer_begin(c->buffer))) && 1762 acl == acl_allow_setrd ) { 1763 LDNS_RD_SET(sldns_buffer_begin(c->buffer)); 1764 } 1765 1766 /* If this request does not have the recursion bit set, verify 1767 * ACLs allow the snooping. */ 1768 if(!(LDNS_RD_WIRE(sldns_buffer_begin(c->buffer))) && 1769 acl != acl_allow_snoop ) { 1770 if(worker->env.cfg->ede) { 1771 EDNS_OPT_LIST_APPEND_EDE(&edns.opt_list_out, 1772 worker->scratchpad, LDNS_EDE_NOT_AUTHORITATIVE, ""); 1773 } 1774 error_encode(c->buffer, LDNS_RCODE_REFUSED, &qinfo, 1775 *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 1776 sldns_buffer_read_u16_at(c->buffer, 2), &edns); 1777 regional_free_all(worker->scratchpad); 1778 log_addr(VERB_ALGO, "refused nonrec (cache snoop) query from", 1779 &repinfo->client_addr, repinfo->client_addrlen); 1780 1781 goto send_reply; 1782 } 1783 1784 /* If we've found a local alias, replace the qname with the alias 1785 * target before resolving it. */ 1786 if(qinfo.local_alias) { 1787 struct ub_packed_rrset_key* rrset = qinfo.local_alias->rrset; 1788 struct packed_rrset_data* d = rrset->entry.data; 1789 1790 /* Sanity check: our current implementation only supports 1791 * a single CNAME RRset as a local alias. */ 1792 if(qinfo.local_alias->next || 1793 rrset->rk.type != htons(LDNS_RR_TYPE_CNAME) || 1794 d->count != 1) { 1795 log_err("assumption failure: unexpected local alias"); 1796 regional_free_all(worker->scratchpad); 1797 return 0; /* drop it */ 1798 } 1799 qinfo.qname = d->rr_data[0] + 2; 1800 qinfo.qname_len = d->rr_len[0] - 2; 1801 } 1802 1803 /* If we may apply IP-based actions to the answer, build the client 1804 * information. As this can be expensive, skip it if there is 1805 * absolutely no possibility of it. */ 1806 if((worker->daemon->use_response_ip || worker->daemon->use_rpz) && 1807 (qinfo.qtype == LDNS_RR_TYPE_A || 1808 qinfo.qtype == LDNS_RR_TYPE_AAAA || 1809 qinfo.qtype == LDNS_RR_TYPE_ANY)) { 1810 cinfo_tmp.taglist = acladdr->taglist; 1811 cinfo_tmp.taglen = acladdr->taglen; 1812 cinfo_tmp.tag_actions = acladdr->tag_actions; 1813 cinfo_tmp.tag_actions_size = acladdr->tag_actions_size; 1814 cinfo_tmp.tag_datas = acladdr->tag_datas; 1815 cinfo_tmp.tag_datas_size = acladdr->tag_datas_size; 1816 cinfo_tmp.view = acladdr->view; 1817 cinfo_tmp.respip_set = worker->daemon->respip_set; 1818 cinfo = &cinfo_tmp; 1819 } 1820 1821 /* Keep the original edns list around. The pointer could change if there is 1822 * a cached answer (through the inplace callback function there). 1823 * No need to actually copy the contents as they shouldn't change. 1824 * Used while prefetching and subnet is enabled. */ 1825 original_edns_list = edns.opt_list_in; 1826 lookup_cache: 1827 /* Lookup the cache. In case we chase an intermediate CNAME chain 1828 * this is a two-pass operation, and lookup_qinfo is different for 1829 * each pass. We should still pass the original qinfo to 1830 * answer_from_cache(), however, since it's used to build the reply. */ 1831 if(!edns_bypass_cache_stage(edns.opt_list_in, &worker->env)) { 1832 is_expired_answer = 0; 1833 is_secure_answer = 0; 1834 h = query_info_hash(lookup_qinfo, sldns_buffer_read_u16_at(c->buffer, 2)); 1835 if((e=slabhash_lookup(worker->env.msg_cache, h, lookup_qinfo, 0))) { 1836 struct reply_info* rep = (struct reply_info*)e->data; 1837 /* answer from cache - we have acquired a readlock on it */ 1838 if(answer_from_cache(worker, &qinfo, cinfo, &need_drop, 1839 &is_expired_answer, &is_secure_answer, 1840 &alias_rrset, &partial_rep, rep, 1841 *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 1842 sldns_buffer_read_u16_at(c->buffer, 2), repinfo, 1843 &edns)) { 1844 /* prefetch it if the prefetch TTL expired. 1845 * Note that if there is more than one pass 1846 * its qname must be that used for cache 1847 * lookup. */ 1848 if((worker->env.cfg->prefetch && 1849 *worker->env.now >= rep->prefetch_ttl) || 1850 (worker->env.cfg->serve_expired && 1851 *worker->env.now > rep->ttl)) { 1852 1853 time_t leeway = rep->ttl - *worker->env.now; 1854 if(rep->ttl < *worker->env.now) 1855 leeway = 0; 1856 lock_rw_unlock(&e->lock); 1857 1858 reply_and_prefetch(worker, lookup_qinfo, 1859 sldns_buffer_read_u16_at(c->buffer, 2), 1860 repinfo, leeway, 1861 (partial_rep || need_drop), 1862 rpz_passthru, 1863 original_edns_list); 1864 if(!partial_rep) { 1865 rc = 0; 1866 regional_free_all(worker->scratchpad); 1867 goto send_reply_rc; 1868 } 1869 } else if(!partial_rep) { 1870 lock_rw_unlock(&e->lock); 1871 regional_free_all(worker->scratchpad); 1872 goto send_reply; 1873 } else { 1874 /* Note that we've already released the 1875 * lock if we're here after prefetch. */ 1876 lock_rw_unlock(&e->lock); 1877 } 1878 /* We've found a partial reply ending with an 1879 * alias. Replace the lookup qinfo for the 1880 * alias target and lookup the cache again to 1881 * (possibly) complete the reply. As we're 1882 * passing the "base" reply, there will be no 1883 * more alias chasing. */ 1884 memset(&qinfo_tmp, 0, sizeof(qinfo_tmp)); 1885 get_cname_target(alias_rrset, &qinfo_tmp.qname, 1886 &qinfo_tmp.qname_len); 1887 if(!qinfo_tmp.qname) { 1888 log_err("unexpected: invalid answer alias"); 1889 regional_free_all(worker->scratchpad); 1890 return 0; /* drop query */ 1891 } 1892 qinfo_tmp.qtype = qinfo.qtype; 1893 qinfo_tmp.qclass = qinfo.qclass; 1894 lookup_qinfo = &qinfo_tmp; 1895 goto lookup_cache; 1896 } 1897 verbose(VERB_ALGO, "answer from the cache failed"); 1898 lock_rw_unlock(&e->lock); 1899 } 1900 1901 if(!LDNS_RD_WIRE(sldns_buffer_begin(c->buffer))) { 1902 if(answer_norec_from_cache(worker, &qinfo, 1903 *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 1904 sldns_buffer_read_u16_at(c->buffer, 2), repinfo, 1905 &edns)) { 1906 regional_free_all(worker->scratchpad); 1907 goto send_reply; 1908 } 1909 verbose(VERB_ALGO, "answer norec from cache -- " 1910 "need to validate or not primed"); 1911 } 1912 } 1913 sldns_buffer_rewind(c->buffer); 1914 server_stats_querymiss(&worker->stats, worker); 1915 1916 if(verbosity >= VERB_CLIENT) { 1917 if(c->type == comm_udp) 1918 log_addr(VERB_CLIENT, "udp request from", 1919 &repinfo->client_addr, repinfo->client_addrlen); 1920 else log_addr(VERB_CLIENT, "tcp request from", 1921 &repinfo->client_addr, repinfo->client_addrlen); 1922 } 1923 1924 /* grab a work request structure for this new request */ 1925 mesh_new_client(worker->env.mesh, &qinfo, cinfo, 1926 sldns_buffer_read_u16_at(c->buffer, 2), 1927 &edns, repinfo, *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 1928 rpz_passthru); 1929 regional_free_all(worker->scratchpad); 1930 worker_mem_report(worker, NULL); 1931 return 0; 1932 1933 send_reply: 1934 rc = 1; 1935 send_reply_rc: 1936 if(need_drop) { 1937 comm_point_drop_reply(repinfo); 1938 return 0; 1939 } 1940 if(is_expired_answer) { 1941 worker->stats.ans_expired++; 1942 } 1943 server_stats_insrcode(&worker->stats, c->buffer); 1944 if(worker->stats.extended) { 1945 if(is_secure_answer) worker->stats.ans_secure++; 1946 } 1947 #ifdef USE_DNSTAP 1948 /* 1949 * sending src (client)/dst (local service) addresses over DNSTAP from send_reply code label (when we serviced local zone for ex.) 1950 */ 1951 if(worker->dtenv.log_client_response_messages && rc !=0) { 1952 log_addr(VERB_ALGO, "from local addr", (void*)repinfo->c->socket->addr, repinfo->c->socket->addrlen); 1953 log_addr(VERB_ALGO, "response to client", &repinfo->client_addr, repinfo->client_addrlen); 1954 dt_msg_send_client_response(&worker->dtenv, &repinfo->client_addr, (void*)repinfo->c->socket->addr, c->type, c->ssl, c->buffer); 1955 } 1956 #endif 1957 if(worker->env.cfg->log_replies) 1958 { 1959 struct timeval tv; 1960 memset(&tv, 0, sizeof(tv)); 1961 if(qinfo.local_alias && qinfo.local_alias->rrset && 1962 qinfo.local_alias->rrset->rk.dname) { 1963 /* log original qname, before the local alias was 1964 * used to resolve that CNAME to something else */ 1965 qinfo.qname = qinfo.local_alias->rrset->rk.dname; 1966 log_reply_info(NO_VERBOSE, &qinfo, 1967 &repinfo->client_addr, repinfo->client_addrlen, 1968 tv, 1, c->buffer, 1969 (worker->env.cfg->log_destaddr?(void*)repinfo->c->socket->addr:NULL), 1970 c->type); 1971 } else { 1972 log_reply_info(NO_VERBOSE, &qinfo, 1973 &repinfo->client_addr, repinfo->client_addrlen, 1974 tv, 1, c->buffer, 1975 (worker->env.cfg->log_destaddr?(void*)repinfo->c->socket->addr:NULL), 1976 c->type); 1977 } 1978 } 1979 #ifdef USE_DNSCRYPT 1980 if(!dnsc_handle_uncurved_request(repinfo)) { 1981 return 0; 1982 } 1983 #endif 1984 return rc; 1985 } 1986 1987 void 1988 worker_sighandler(int sig, void* arg) 1989 { 1990 /* note that log, print, syscalls here give race conditions. 1991 * And cause hangups if the log-lock is held by the application. */ 1992 struct worker* worker = (struct worker*)arg; 1993 switch(sig) { 1994 #ifdef SIGHUP 1995 case SIGHUP: 1996 comm_base_exit(worker->base); 1997 break; 1998 #endif 1999 #ifdef SIGBREAK 2000 case SIGBREAK: 2001 #endif 2002 case SIGINT: 2003 worker->need_to_exit = 1; 2004 comm_base_exit(worker->base); 2005 break; 2006 #ifdef SIGQUIT 2007 case SIGQUIT: 2008 worker->need_to_exit = 1; 2009 comm_base_exit(worker->base); 2010 break; 2011 #endif 2012 case SIGTERM: 2013 worker->need_to_exit = 1; 2014 comm_base_exit(worker->base); 2015 break; 2016 default: 2017 /* unknown signal, ignored */ 2018 break; 2019 } 2020 } 2021 2022 /** restart statistics timer for worker, if enabled */ 2023 static void 2024 worker_restart_timer(struct worker* worker) 2025 { 2026 if(worker->env.cfg->stat_interval > 0) { 2027 struct timeval tv; 2028 #ifndef S_SPLINT_S 2029 tv.tv_sec = worker->env.cfg->stat_interval; 2030 tv.tv_usec = 0; 2031 #endif 2032 comm_timer_set(worker->stat_timer, &tv); 2033 } 2034 } 2035 2036 void worker_stat_timer_cb(void* arg) 2037 { 2038 struct worker* worker = (struct worker*)arg; 2039 server_stats_log(&worker->stats, worker, worker->thread_num); 2040 mesh_stats(worker->env.mesh, "mesh has"); 2041 worker_mem_report(worker, NULL); 2042 /* SHM is enabled, process data to SHM */ 2043 if (worker->daemon->cfg->shm_enable) { 2044 shm_main_run(worker); 2045 } 2046 if(!worker->daemon->cfg->stat_cumulative) { 2047 worker_stats_clear(worker); 2048 } 2049 /* start next timer */ 2050 worker_restart_timer(worker); 2051 } 2052 2053 void worker_probe_timer_cb(void* arg) 2054 { 2055 struct worker* worker = (struct worker*)arg; 2056 struct timeval tv; 2057 #ifndef S_SPLINT_S 2058 tv.tv_sec = (time_t)autr_probe_timer(&worker->env); 2059 tv.tv_usec = 0; 2060 #endif 2061 if(tv.tv_sec != 0) 2062 comm_timer_set(worker->env.probe_timer, &tv); 2063 } 2064 2065 struct worker* 2066 worker_create(struct daemon* daemon, int id, int* ports, int n) 2067 { 2068 unsigned int seed; 2069 struct worker* worker = (struct worker*)calloc(1, 2070 sizeof(struct worker)); 2071 if(!worker) 2072 return NULL; 2073 worker->numports = n; 2074 worker->ports = (int*)memdup(ports, sizeof(int)*n); 2075 if(!worker->ports) { 2076 free(worker); 2077 return NULL; 2078 } 2079 worker->daemon = daemon; 2080 worker->thread_num = id; 2081 if(!(worker->cmd = tube_create())) { 2082 free(worker->ports); 2083 free(worker); 2084 return NULL; 2085 } 2086 /* create random state here to avoid locking trouble in RAND_bytes */ 2087 if(!(worker->rndstate = ub_initstate(daemon->rand))) { 2088 log_err("could not init random numbers."); 2089 tube_delete(worker->cmd); 2090 free(worker->ports); 2091 free(worker); 2092 return NULL; 2093 } 2094 explicit_bzero(&seed, sizeof(seed)); 2095 return worker; 2096 } 2097 2098 int 2099 worker_init(struct worker* worker, struct config_file *cfg, 2100 struct listen_port* ports, int do_sigs) 2101 { 2102 #ifdef USE_DNSTAP 2103 struct dt_env* dtenv = &worker->dtenv; 2104 #else 2105 void* dtenv = NULL; 2106 #endif 2107 #ifdef HAVE_GETTID 2108 worker->thread_tid = gettid(); 2109 #endif 2110 worker->need_to_exit = 0; 2111 worker->base = comm_base_create(do_sigs); 2112 if(!worker->base) { 2113 log_err("could not create event handling base"); 2114 worker_delete(worker); 2115 return 0; 2116 } 2117 comm_base_set_slow_accept_handlers(worker->base, &worker_stop_accept, 2118 &worker_start_accept, worker); 2119 if(do_sigs) { 2120 #ifdef SIGHUP 2121 ub_thread_sig_unblock(SIGHUP); 2122 #endif 2123 #ifdef SIGBREAK 2124 ub_thread_sig_unblock(SIGBREAK); 2125 #endif 2126 ub_thread_sig_unblock(SIGINT); 2127 #ifdef SIGQUIT 2128 ub_thread_sig_unblock(SIGQUIT); 2129 #endif 2130 ub_thread_sig_unblock(SIGTERM); 2131 #ifndef LIBEVENT_SIGNAL_PROBLEM 2132 worker->comsig = comm_signal_create(worker->base, 2133 worker_sighandler, worker); 2134 if(!worker->comsig 2135 #ifdef SIGHUP 2136 || !comm_signal_bind(worker->comsig, SIGHUP) 2137 #endif 2138 #ifdef SIGQUIT 2139 || !comm_signal_bind(worker->comsig, SIGQUIT) 2140 #endif 2141 || !comm_signal_bind(worker->comsig, SIGTERM) 2142 #ifdef SIGBREAK 2143 || !comm_signal_bind(worker->comsig, SIGBREAK) 2144 #endif 2145 || !comm_signal_bind(worker->comsig, SIGINT)) { 2146 log_err("could not create signal handlers"); 2147 worker_delete(worker); 2148 return 0; 2149 } 2150 #endif /* LIBEVENT_SIGNAL_PROBLEM */ 2151 if(!daemon_remote_open_accept(worker->daemon->rc, 2152 worker->daemon->rc_ports, worker)) { 2153 worker_delete(worker); 2154 return 0; 2155 } 2156 #ifdef UB_ON_WINDOWS 2157 wsvc_setup_worker(worker); 2158 #endif /* UB_ON_WINDOWS */ 2159 } else { /* !do_sigs */ 2160 worker->comsig = NULL; 2161 } 2162 #ifdef USE_DNSTAP 2163 if(cfg->dnstap) { 2164 log_assert(worker->daemon->dtenv != NULL); 2165 memcpy(&worker->dtenv, worker->daemon->dtenv, sizeof(struct dt_env)); 2166 if(!dt_init(&worker->dtenv, worker->base)) 2167 fatal_exit("dt_init failed"); 2168 } 2169 #endif 2170 worker->front = listen_create(worker->base, ports, 2171 cfg->msg_buffer_size, (int)cfg->incoming_num_tcp, 2172 cfg->do_tcp_keepalive 2173 ? cfg->tcp_keepalive_timeout 2174 : cfg->tcp_idle_timeout, 2175 cfg->harden_large_queries, cfg->http_max_streams, 2176 cfg->http_endpoint, cfg->http_notls_downstream, 2177 worker->daemon->tcl, worker->daemon->listen_sslctx, 2178 dtenv, worker_handle_request, worker); 2179 if(!worker->front) { 2180 log_err("could not create listening sockets"); 2181 worker_delete(worker); 2182 return 0; 2183 } 2184 worker->back = outside_network_create(worker->base, 2185 cfg->msg_buffer_size, (size_t)cfg->outgoing_num_ports, 2186 cfg->out_ifs, cfg->num_out_ifs, cfg->do_ip4, cfg->do_ip6, 2187 cfg->do_tcp?cfg->outgoing_num_tcp:0, cfg->ip_dscp, 2188 worker->daemon->env->infra_cache, worker->rndstate, 2189 cfg->use_caps_bits_for_id, worker->ports, worker->numports, 2190 cfg->unwanted_threshold, cfg->outgoing_tcp_mss, 2191 &worker_alloc_cleanup, worker, 2192 cfg->do_udp || cfg->udp_upstream_without_downstream, 2193 worker->daemon->connect_sslctx, cfg->delay_close, 2194 cfg->tls_use_sni, dtenv, cfg->udp_connect, 2195 cfg->max_reuse_tcp_queries, cfg->tcp_reuse_timeout, 2196 cfg->tcp_auth_query_timeout); 2197 if(!worker->back) { 2198 log_err("could not create outgoing sockets"); 2199 worker_delete(worker); 2200 return 0; 2201 } 2202 iterator_set_ip46_support(&worker->daemon->mods, worker->daemon->env, 2203 worker->back); 2204 /* start listening to commands */ 2205 if(!tube_setup_bg_listen(worker->cmd, worker->base, 2206 &worker_handle_control_cmd, worker)) { 2207 log_err("could not create control compt."); 2208 worker_delete(worker); 2209 return 0; 2210 } 2211 worker->stat_timer = comm_timer_create(worker->base, 2212 worker_stat_timer_cb, worker); 2213 if(!worker->stat_timer) { 2214 log_err("could not create statistics timer"); 2215 } 2216 2217 /* we use the msg_buffer_size as a good estimate for what the 2218 * user wants for memory usage sizes */ 2219 worker->scratchpad = regional_create_custom(cfg->msg_buffer_size); 2220 if(!worker->scratchpad) { 2221 log_err("malloc failure"); 2222 worker_delete(worker); 2223 return 0; 2224 } 2225 2226 server_stats_init(&worker->stats, cfg); 2227 worker->alloc = worker->daemon->worker_allocs[worker->thread_num]; 2228 alloc_set_id_cleanup(worker->alloc, &worker_alloc_cleanup, worker); 2229 worker->env = *worker->daemon->env; 2230 comm_base_timept(worker->base, &worker->env.now, &worker->env.now_tv); 2231 worker->env.worker = worker; 2232 worker->env.worker_base = worker->base; 2233 worker->env.send_query = &worker_send_query; 2234 worker->env.alloc = worker->alloc; 2235 worker->env.outnet = worker->back; 2236 worker->env.rnd = worker->rndstate; 2237 /* If case prefetch is triggered, the corresponding mesh will clear 2238 * the scratchpad for the module env in the middle of request handling. 2239 * It would be prone to a use-after-free kind of bug, so we avoid 2240 * sharing it with worker's own scratchpad at the cost of having 2241 * one more pad per worker. */ 2242 worker->env.scratch = regional_create_custom(cfg->msg_buffer_size); 2243 if(!worker->env.scratch) { 2244 log_err("malloc failure"); 2245 worker_delete(worker); 2246 return 0; 2247 } 2248 worker->env.mesh = mesh_create(&worker->daemon->mods, &worker->env); 2249 if(!worker->env.mesh) { 2250 log_err("malloc failure"); 2251 worker_delete(worker); 2252 return 0; 2253 } 2254 /* Pass on daemon variables that we would need in the mesh area */ 2255 worker->env.mesh->use_response_ip = worker->daemon->use_response_ip; 2256 worker->env.mesh->use_rpz = worker->daemon->use_rpz; 2257 2258 worker->env.detach_subs = &mesh_detach_subs; 2259 worker->env.attach_sub = &mesh_attach_sub; 2260 worker->env.add_sub = &mesh_add_sub; 2261 worker->env.kill_sub = &mesh_state_delete; 2262 worker->env.detect_cycle = &mesh_detect_cycle; 2263 worker->env.scratch_buffer = sldns_buffer_new(cfg->msg_buffer_size); 2264 if(!worker->env.scratch_buffer) { 2265 log_err("malloc failure"); 2266 worker_delete(worker); 2267 return 0; 2268 } 2269 /* one probe timer per process -- if we have 5011 anchors */ 2270 if(autr_get_num_anchors(worker->env.anchors) > 0 2271 #ifndef THREADS_DISABLED 2272 && worker->thread_num == 0 2273 #endif 2274 ) { 2275 struct timeval tv; 2276 tv.tv_sec = 0; 2277 tv.tv_usec = 0; 2278 worker->env.probe_timer = comm_timer_create(worker->base, 2279 worker_probe_timer_cb, worker); 2280 if(!worker->env.probe_timer) { 2281 log_err("could not create 5011-probe timer"); 2282 } else { 2283 /* let timer fire, then it can reset itself */ 2284 comm_timer_set(worker->env.probe_timer, &tv); 2285 } 2286 } 2287 /* zone transfer tasks, setup once per process, if any */ 2288 if(worker->env.auth_zones 2289 #ifndef THREADS_DISABLED 2290 && worker->thread_num == 0 2291 #endif 2292 ) { 2293 auth_xfer_pickup_initial(worker->env.auth_zones, &worker->env); 2294 auth_zones_pickup_zonemd_verify(worker->env.auth_zones, 2295 &worker->env); 2296 } 2297 #ifdef USE_DNSTAP 2298 if(worker->daemon->cfg->dnstap 2299 #ifndef THREADS_DISABLED 2300 && worker->thread_num == 0 2301 #endif 2302 ) { 2303 if(!dt_io_thread_start(dtenv->dtio, comm_base_internal( 2304 worker->base), worker->daemon->num)) { 2305 log_err("could not start dnstap io thread"); 2306 worker_delete(worker); 2307 return 0; 2308 } 2309 } 2310 #endif /* USE_DNSTAP */ 2311 worker_mem_report(worker, NULL); 2312 /* if statistics enabled start timer */ 2313 if(worker->env.cfg->stat_interval > 0) { 2314 verbose(VERB_ALGO, "set statistics interval %d secs", 2315 worker->env.cfg->stat_interval); 2316 worker_restart_timer(worker); 2317 } 2318 pp_init(&sldns_write_uint16, &sldns_write_uint32); 2319 return 1; 2320 } 2321 2322 void 2323 worker_work(struct worker* worker) 2324 { 2325 comm_base_dispatch(worker->base); 2326 } 2327 2328 void 2329 worker_delete(struct worker* worker) 2330 { 2331 if(!worker) 2332 return; 2333 if(worker->env.mesh && verbosity >= VERB_OPS) { 2334 server_stats_log(&worker->stats, worker, worker->thread_num); 2335 mesh_stats(worker->env.mesh, "mesh has"); 2336 worker_mem_report(worker, NULL); 2337 } 2338 outside_network_quit_prepare(worker->back); 2339 mesh_delete(worker->env.mesh); 2340 sldns_buffer_free(worker->env.scratch_buffer); 2341 listen_delete(worker->front); 2342 outside_network_delete(worker->back); 2343 comm_signal_delete(worker->comsig); 2344 tube_delete(worker->cmd); 2345 comm_timer_delete(worker->stat_timer); 2346 comm_timer_delete(worker->env.probe_timer); 2347 free(worker->ports); 2348 if(worker->thread_num == 0) { 2349 #ifdef UB_ON_WINDOWS 2350 wsvc_desetup_worker(worker); 2351 #endif /* UB_ON_WINDOWS */ 2352 } 2353 #ifdef USE_DNSTAP 2354 if(worker->daemon->cfg->dnstap 2355 #ifndef THREADS_DISABLED 2356 && worker->thread_num == 0 2357 #endif 2358 ) { 2359 dt_io_thread_stop(worker->dtenv.dtio); 2360 } 2361 dt_deinit(&worker->dtenv); 2362 #endif /* USE_DNSTAP */ 2363 comm_base_delete(worker->base); 2364 ub_randfree(worker->rndstate); 2365 /* don't touch worker->alloc, as it's maintained in daemon */ 2366 regional_destroy(worker->env.scratch); 2367 regional_destroy(worker->scratchpad); 2368 free(worker); 2369 } 2370 2371 struct outbound_entry* 2372 worker_send_query(struct query_info* qinfo, uint16_t flags, int dnssec, 2373 int want_dnssec, int nocaps, int check_ratelimit, 2374 struct sockaddr_storage* addr, socklen_t addrlen, uint8_t* zone, 2375 size_t zonelen, int tcp_upstream, int ssl_upstream, char* tls_auth_name, 2376 struct module_qstate* q, int* was_ratelimited) 2377 { 2378 struct worker* worker = q->env->worker; 2379 struct outbound_entry* e = (struct outbound_entry*)regional_alloc( 2380 q->region, sizeof(*e)); 2381 if(!e) 2382 return NULL; 2383 e->qstate = q; 2384 e->qsent = outnet_serviced_query(worker->back, qinfo, flags, dnssec, 2385 want_dnssec, nocaps, check_ratelimit, tcp_upstream, 2386 ssl_upstream, tls_auth_name, addr, addrlen, zone, zonelen, q, 2387 worker_handle_service_reply, e, worker->back->udp_buff, q->env, 2388 was_ratelimited); 2389 if(!e->qsent) { 2390 return NULL; 2391 } 2392 return e; 2393 } 2394 2395 void 2396 worker_alloc_cleanup(void* arg) 2397 { 2398 struct worker* worker = (struct worker*)arg; 2399 slabhash_clear(&worker->env.rrset_cache->table); 2400 slabhash_clear(worker->env.msg_cache); 2401 } 2402 2403 void worker_stats_clear(struct worker* worker) 2404 { 2405 server_stats_init(&worker->stats, worker->env.cfg); 2406 mesh_stats_clear(worker->env.mesh); 2407 worker->back->unwanted_replies = 0; 2408 worker->back->num_tcp_outgoing = 0; 2409 worker->back->num_udp_outgoing = 0; 2410 } 2411 2412 void worker_start_accept(void* arg) 2413 { 2414 struct worker* worker = (struct worker*)arg; 2415 listen_start_accept(worker->front); 2416 if(worker->thread_num == 0) 2417 daemon_remote_start_accept(worker->daemon->rc); 2418 } 2419 2420 void worker_stop_accept(void* arg) 2421 { 2422 struct worker* worker = (struct worker*)arg; 2423 listen_stop_accept(worker->front); 2424 if(worker->thread_num == 0) 2425 daemon_remote_stop_accept(worker->daemon->rc); 2426 } 2427 2428 /* --- fake callbacks for fptr_wlist to work --- */ 2429 struct outbound_entry* libworker_send_query( 2430 struct query_info* ATTR_UNUSED(qinfo), 2431 uint16_t ATTR_UNUSED(flags), int ATTR_UNUSED(dnssec), 2432 int ATTR_UNUSED(want_dnssec), int ATTR_UNUSED(nocaps), 2433 int ATTR_UNUSED(check_ratelimit), 2434 struct sockaddr_storage* ATTR_UNUSED(addr), socklen_t ATTR_UNUSED(addrlen), 2435 uint8_t* ATTR_UNUSED(zone), size_t ATTR_UNUSED(zonelen), int ATTR_UNUSED(tcp_upstream), 2436 int ATTR_UNUSED(ssl_upstream), char* ATTR_UNUSED(tls_auth_name), 2437 struct module_qstate* ATTR_UNUSED(q), int* ATTR_UNUSED(was_ratelimited)) 2438 { 2439 log_assert(0); 2440 return 0; 2441 } 2442 2443 int libworker_handle_service_reply(struct comm_point* ATTR_UNUSED(c), 2444 void* ATTR_UNUSED(arg), int ATTR_UNUSED(error), 2445 struct comm_reply* ATTR_UNUSED(reply_info)) 2446 { 2447 log_assert(0); 2448 return 0; 2449 } 2450 2451 void libworker_handle_control_cmd(struct tube* ATTR_UNUSED(tube), 2452 uint8_t* ATTR_UNUSED(buffer), size_t ATTR_UNUSED(len), 2453 int ATTR_UNUSED(error), void* ATTR_UNUSED(arg)) 2454 { 2455 log_assert(0); 2456 } 2457 2458 void libworker_fg_done_cb(void* ATTR_UNUSED(arg), int ATTR_UNUSED(rcode), 2459 sldns_buffer* ATTR_UNUSED(buf), enum sec_status ATTR_UNUSED(s), 2460 char* ATTR_UNUSED(why_bogus), int ATTR_UNUSED(was_ratelimited)) 2461 { 2462 log_assert(0); 2463 } 2464 2465 void libworker_bg_done_cb(void* ATTR_UNUSED(arg), int ATTR_UNUSED(rcode), 2466 sldns_buffer* ATTR_UNUSED(buf), enum sec_status ATTR_UNUSED(s), 2467 char* ATTR_UNUSED(why_bogus), int ATTR_UNUSED(was_ratelimited)) 2468 { 2469 log_assert(0); 2470 } 2471 2472 void libworker_event_done_cb(void* ATTR_UNUSED(arg), int ATTR_UNUSED(rcode), 2473 sldns_buffer* ATTR_UNUSED(buf), enum sec_status ATTR_UNUSED(s), 2474 char* ATTR_UNUSED(why_bogus), int ATTR_UNUSED(was_ratelimited)) 2475 { 2476 log_assert(0); 2477 } 2478 2479 int context_query_cmp(const void* ATTR_UNUSED(a), const void* ATTR_UNUSED(b)) 2480 { 2481 log_assert(0); 2482 return 0; 2483 } 2484 2485 int order_lock_cmp(const void* ATTR_UNUSED(e1), const void* ATTR_UNUSED(e2)) 2486 { 2487 log_assert(0); 2488 return 0; 2489 } 2490 2491 int codeline_cmp(const void* ATTR_UNUSED(a), const void* ATTR_UNUSED(b)) 2492 { 2493 log_assert(0); 2494 return 0; 2495 } 2496 2497 #ifdef USE_DNSTAP 2498 void dtio_tap_callback(int ATTR_UNUSED(fd), short ATTR_UNUSED(ev), 2499 void* ATTR_UNUSED(arg)) 2500 { 2501 log_assert(0); 2502 } 2503 #endif 2504 2505 #ifdef USE_DNSTAP 2506 void dtio_mainfdcallback(int ATTR_UNUSED(fd), short ATTR_UNUSED(ev), 2507 void* ATTR_UNUSED(arg)) 2508 { 2509 log_assert(0); 2510 } 2511 #endif 2512