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/tube.h" 70 #include "util/edns.h" 71 #include "util/timeval_func.h" 72 #include "iterator/iter_fwd.h" 73 #include "iterator/iter_hints.h" 74 #include "iterator/iter_utils.h" 75 #include "validator/autotrust.h" 76 #include "validator/val_anchor.h" 77 #include "respip/respip.h" 78 #include "libunbound/context.h" 79 #include "libunbound/libworker.h" 80 #include "sldns/sbuffer.h" 81 #include "sldns/wire2str.h" 82 #include "util/shm_side/shm_main.h" 83 #include "dnscrypt/dnscrypt.h" 84 #include "dnstap/dtstream.h" 85 86 #ifdef HAVE_SYS_TYPES_H 87 # include <sys/types.h> 88 #endif 89 #ifdef HAVE_NETDB_H 90 #include <netdb.h> 91 #endif 92 #include <signal.h> 93 #ifdef UB_ON_WINDOWS 94 #include "winrc/win_svc.h" 95 #endif 96 97 /** Size of an UDP datagram */ 98 #define NORMAL_UDP_SIZE 512 /* bytes */ 99 /** ratelimit for error responses */ 100 #define ERROR_RATELIMIT 100 /* qps */ 101 102 /** 103 * seconds to add to prefetch leeway. This is a TTL that expires old rrsets 104 * earlier than they should in order to put the new update into the cache. 105 * This additional value is to make sure that if not all TTLs are equal in 106 * the message to be updated(and replaced), that rrsets with up to this much 107 * extra TTL are also replaced. This means that the resulting new message 108 * will have (most likely) this TTL at least, avoiding very small 'split 109 * second' TTLs due to operators choosing relative primes for TTLs (or so). 110 * Also has to be at least one to break ties (and overwrite cached entry). 111 */ 112 #define PREFETCH_EXPIRY_ADD 60 113 114 /** Report on memory usage by this thread and global */ 115 static void 116 worker_mem_report(struct worker* ATTR_UNUSED(worker), 117 struct serviced_query* ATTR_UNUSED(cur_serv)) 118 { 119 #ifdef UNBOUND_ALLOC_STATS 120 /* measure memory leakage */ 121 extern size_t unbound_mem_alloc, unbound_mem_freed; 122 /* debug func in validator module */ 123 size_t total, front, back, mesh, msg, rrset, infra, ac, superac; 124 size_t me, iter, val, anch; 125 int i; 126 #ifdef CLIENT_SUBNET 127 size_t subnet = 0; 128 #endif /* CLIENT_SUBNET */ 129 if(verbosity < VERB_ALGO) 130 return; 131 front = listen_get_mem(worker->front); 132 back = outnet_get_mem(worker->back); 133 msg = slabhash_get_mem(worker->env.msg_cache); 134 rrset = slabhash_get_mem(&worker->env.rrset_cache->table); 135 infra = infra_get_mem(worker->env.infra_cache); 136 mesh = mesh_get_mem(worker->env.mesh); 137 ac = alloc_get_mem(worker->alloc); 138 superac = alloc_get_mem(&worker->daemon->superalloc); 139 anch = anchors_get_mem(worker->env.anchors); 140 iter = 0; 141 val = 0; 142 for(i=0; i<worker->env.mesh->mods.num; i++) { 143 fptr_ok(fptr_whitelist_mod_get_mem(worker->env.mesh-> 144 mods.mod[i]->get_mem)); 145 if(strcmp(worker->env.mesh->mods.mod[i]->name, "validator")==0) 146 val += (*worker->env.mesh->mods.mod[i]->get_mem) 147 (&worker->env, i); 148 #ifdef CLIENT_SUBNET 149 else if(strcmp(worker->env.mesh->mods.mod[i]->name, 150 "subnetcache")==0) 151 subnet += (*worker->env.mesh->mods.mod[i]->get_mem) 152 (&worker->env, i); 153 #endif /* CLIENT_SUBNET */ 154 else iter += (*worker->env.mesh->mods.mod[i]->get_mem) 155 (&worker->env, i); 156 } 157 me = sizeof(*worker) + sizeof(*worker->base) + sizeof(*worker->comsig) 158 + comm_point_get_mem(worker->cmd_com) 159 + sizeof(worker->rndstate) 160 + regional_get_mem(worker->scratchpad) 161 + sizeof(*worker->env.scratch_buffer) 162 + sldns_buffer_capacity(worker->env.scratch_buffer) 163 + forwards_get_mem(worker->env.fwds) 164 + hints_get_mem(worker->env.hints); 165 if(worker->thread_num == 0) 166 me += acl_list_get_mem(worker->daemon->acl); 167 if(cur_serv) { 168 me += serviced_get_mem(cur_serv); 169 } 170 total = front+back+mesh+msg+rrset+infra+iter+val+ac+superac+me; 171 #ifdef CLIENT_SUBNET 172 total += subnet; 173 log_info("Memory conditions: %u front=%u back=%u mesh=%u msg=%u " 174 "rrset=%u infra=%u iter=%u val=%u subnet=%u anchors=%u " 175 "alloccache=%u globalalloccache=%u me=%u", 176 (unsigned)total, (unsigned)front, (unsigned)back, 177 (unsigned)mesh, (unsigned)msg, (unsigned)rrset, (unsigned)infra, 178 (unsigned)iter, (unsigned)val, 179 (unsigned)subnet, (unsigned)anch, (unsigned)ac, 180 (unsigned)superac, (unsigned)me); 181 #else /* no CLIENT_SUBNET */ 182 log_info("Memory conditions: %u front=%u back=%u mesh=%u msg=%u " 183 "rrset=%u infra=%u iter=%u val=%u anchors=%u " 184 "alloccache=%u globalalloccache=%u me=%u", 185 (unsigned)total, (unsigned)front, (unsigned)back, 186 (unsigned)mesh, (unsigned)msg, (unsigned)rrset, 187 (unsigned)infra, (unsigned)iter, (unsigned)val, (unsigned)anch, 188 (unsigned)ac, (unsigned)superac, (unsigned)me); 189 #endif /* CLIENT_SUBNET */ 190 log_info("Total heap memory estimate: %u total-alloc: %u " 191 "total-free: %u", (unsigned)total, 192 (unsigned)unbound_mem_alloc, (unsigned)unbound_mem_freed); 193 #else /* no UNBOUND_ALLOC_STATS */ 194 size_t val = 0; 195 #ifdef CLIENT_SUBNET 196 size_t subnet = 0; 197 #endif /* CLIENT_SUBNET */ 198 int i; 199 if(verbosity < VERB_QUERY) 200 return; 201 for(i=0; i<worker->env.mesh->mods.num; i++) { 202 fptr_ok(fptr_whitelist_mod_get_mem(worker->env.mesh-> 203 mods.mod[i]->get_mem)); 204 if(strcmp(worker->env.mesh->mods.mod[i]->name, "validator")==0) 205 val += (*worker->env.mesh->mods.mod[i]->get_mem) 206 (&worker->env, i); 207 #ifdef CLIENT_SUBNET 208 else if(strcmp(worker->env.mesh->mods.mod[i]->name, 209 "subnetcache")==0) 210 subnet += (*worker->env.mesh->mods.mod[i]->get_mem) 211 (&worker->env, i); 212 #endif /* CLIENT_SUBNET */ 213 } 214 #ifdef CLIENT_SUBNET 215 verbose(VERB_QUERY, "cache memory msg=%u rrset=%u infra=%u val=%u " 216 "subnet=%u", 217 (unsigned)slabhash_get_mem(worker->env.msg_cache), 218 (unsigned)slabhash_get_mem(&worker->env.rrset_cache->table), 219 (unsigned)infra_get_mem(worker->env.infra_cache), 220 (unsigned)val, (unsigned)subnet); 221 #else /* no CLIENT_SUBNET */ 222 verbose(VERB_QUERY, "cache memory msg=%u rrset=%u infra=%u val=%u", 223 (unsigned)slabhash_get_mem(worker->env.msg_cache), 224 (unsigned)slabhash_get_mem(&worker->env.rrset_cache->table), 225 (unsigned)infra_get_mem(worker->env.infra_cache), 226 (unsigned)val); 227 #endif /* CLIENT_SUBNET */ 228 #endif /* UNBOUND_ALLOC_STATS */ 229 } 230 231 void 232 worker_send_cmd(struct worker* worker, enum worker_commands cmd) 233 { 234 uint32_t c = (uint32_t)htonl(cmd); 235 if(!tube_write_msg(worker->cmd, (uint8_t*)&c, sizeof(c), 0)) { 236 log_err("worker send cmd %d failed", (int)cmd); 237 } 238 } 239 240 int 241 worker_handle_service_reply(struct comm_point* c, void* arg, int error, 242 struct comm_reply* reply_info) 243 { 244 struct outbound_entry* e = (struct outbound_entry*)arg; 245 struct worker* worker = e->qstate->env->worker; 246 struct serviced_query *sq = e->qsent; 247 248 verbose(VERB_ALGO, "worker svcd callback for qstate %p", e->qstate); 249 if(error != 0) { 250 mesh_report_reply(worker->env.mesh, e, reply_info, error); 251 worker_mem_report(worker, sq); 252 return 0; 253 } 254 /* sanity check. */ 255 if(!LDNS_QR_WIRE(sldns_buffer_begin(c->buffer)) 256 || LDNS_OPCODE_WIRE(sldns_buffer_begin(c->buffer)) != 257 LDNS_PACKET_QUERY 258 || LDNS_QDCOUNT(sldns_buffer_begin(c->buffer)) > 1) { 259 /* error becomes timeout for the module as if this reply 260 * never arrived. */ 261 verbose(VERB_ALGO, "worker: bad reply handled as timeout"); 262 mesh_report_reply(worker->env.mesh, e, reply_info, 263 NETEVENT_TIMEOUT); 264 worker_mem_report(worker, sq); 265 return 0; 266 } 267 mesh_report_reply(worker->env.mesh, e, reply_info, NETEVENT_NOERROR); 268 worker_mem_report(worker, sq); 269 return 0; 270 } 271 272 /** ratelimit error replies 273 * @param worker: the worker struct with ratelimit counter 274 * @param err: error code that would be wanted. 275 * @return value of err if okay, or -1 if it should be discarded instead. 276 */ 277 static int 278 worker_err_ratelimit(struct worker* worker, int err) 279 { 280 if(worker->err_limit_time == *worker->env.now) { 281 /* see if limit is exceeded for this second */ 282 if(worker->err_limit_count++ > ERROR_RATELIMIT) 283 return -1; 284 } else { 285 /* new second, new limits */ 286 worker->err_limit_time = *worker->env.now; 287 worker->err_limit_count = 1; 288 } 289 return err; 290 } 291 292 /** 293 * Structure holding the result of the worker_check_request function. 294 * Based on configuration it could be called up to four times; ideally should 295 * be called once. 296 */ 297 struct check_request_result { 298 int checked; 299 int value; 300 }; 301 /** check request sanity. 302 * @param pkt: the wire packet to examine for sanity. 303 * @param worker: parameters for checking. 304 * @param out: struct to update with the result. 305 */ 306 static void 307 worker_check_request(sldns_buffer* pkt, struct worker* worker, 308 struct check_request_result* out) 309 { 310 if(out->checked) return; 311 out->checked = 1; 312 if(sldns_buffer_limit(pkt) < LDNS_HEADER_SIZE) { 313 verbose(VERB_QUERY, "request too short, discarded"); 314 out->value = -1; 315 return; 316 } 317 if(sldns_buffer_limit(pkt) > NORMAL_UDP_SIZE && 318 worker->daemon->cfg->harden_large_queries) { 319 verbose(VERB_QUERY, "request too large, discarded"); 320 out->value = -1; 321 return; 322 } 323 if(LDNS_QR_WIRE(sldns_buffer_begin(pkt))) { 324 verbose(VERB_QUERY, "request has QR bit on, discarded"); 325 out->value = -1; 326 return; 327 } 328 if(LDNS_TC_WIRE(sldns_buffer_begin(pkt))) { 329 LDNS_TC_CLR(sldns_buffer_begin(pkt)); 330 verbose(VERB_QUERY, "request bad, has TC bit on"); 331 out->value = worker_err_ratelimit(worker, LDNS_RCODE_FORMERR); 332 return; 333 } 334 if(LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt)) != LDNS_PACKET_QUERY && 335 LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt)) != LDNS_PACKET_NOTIFY) { 336 verbose(VERB_QUERY, "request unknown opcode %d", 337 LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt))); 338 out->value = worker_err_ratelimit(worker, LDNS_RCODE_NOTIMPL); 339 return; 340 } 341 if(LDNS_QDCOUNT(sldns_buffer_begin(pkt)) != 1) { 342 verbose(VERB_QUERY, "request wrong nr qd=%d", 343 LDNS_QDCOUNT(sldns_buffer_begin(pkt))); 344 out->value = worker_err_ratelimit(worker, LDNS_RCODE_FORMERR); 345 return; 346 } 347 if(LDNS_ANCOUNT(sldns_buffer_begin(pkt)) != 0 && 348 (LDNS_ANCOUNT(sldns_buffer_begin(pkt)) != 1 || 349 LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt)) != LDNS_PACKET_NOTIFY)) { 350 verbose(VERB_QUERY, "request wrong nr an=%d", 351 LDNS_ANCOUNT(sldns_buffer_begin(pkt))); 352 out->value = worker_err_ratelimit(worker, LDNS_RCODE_FORMERR); 353 return; 354 } 355 if(LDNS_NSCOUNT(sldns_buffer_begin(pkt)) != 0) { 356 verbose(VERB_QUERY, "request wrong nr ns=%d", 357 LDNS_NSCOUNT(sldns_buffer_begin(pkt))); 358 out->value = worker_err_ratelimit(worker, LDNS_RCODE_FORMERR); 359 return; 360 } 361 if(LDNS_ARCOUNT(sldns_buffer_begin(pkt)) > 1) { 362 verbose(VERB_QUERY, "request wrong nr ar=%d", 363 LDNS_ARCOUNT(sldns_buffer_begin(pkt))); 364 out->value = worker_err_ratelimit(worker, LDNS_RCODE_FORMERR); 365 return; 366 } 367 out->value = 0; 368 return; 369 } 370 371 void 372 worker_handle_control_cmd(struct tube* ATTR_UNUSED(tube), uint8_t* msg, 373 size_t len, int error, void* arg) 374 { 375 struct worker* worker = (struct worker*)arg; 376 enum worker_commands cmd; 377 if(error != NETEVENT_NOERROR) { 378 free(msg); 379 if(error == NETEVENT_CLOSED) 380 comm_base_exit(worker->base); 381 else log_info("control event: %d", error); 382 return; 383 } 384 if(len != sizeof(uint32_t)) { 385 fatal_exit("bad control msg length %d", (int)len); 386 } 387 cmd = sldns_read_uint32(msg); 388 free(msg); 389 switch(cmd) { 390 case worker_cmd_quit: 391 verbose(VERB_ALGO, "got control cmd quit"); 392 comm_base_exit(worker->base); 393 break; 394 case worker_cmd_stats: 395 verbose(VERB_ALGO, "got control cmd stats"); 396 server_stats_reply(worker, 1); 397 break; 398 case worker_cmd_stats_noreset: 399 verbose(VERB_ALGO, "got control cmd stats_noreset"); 400 server_stats_reply(worker, 0); 401 break; 402 case worker_cmd_remote: 403 verbose(VERB_ALGO, "got control cmd remote"); 404 daemon_remote_exec(worker); 405 break; 406 default: 407 log_err("bad command %d", (int)cmd); 408 break; 409 } 410 } 411 412 /** check if a delegation is secure */ 413 static enum sec_status 414 check_delegation_secure(struct reply_info *rep) 415 { 416 /* return smallest security status */ 417 size_t i; 418 enum sec_status sec = sec_status_secure; 419 enum sec_status s; 420 size_t num = rep->an_numrrsets + rep->ns_numrrsets; 421 /* check if answer and authority are OK */ 422 for(i=0; i<num; i++) { 423 s = ((struct packed_rrset_data*)rep->rrsets[i]->entry.data) 424 ->security; 425 if(s < sec) 426 sec = s; 427 } 428 /* in additional, only unchecked triggers revalidation */ 429 for(i=num; i<rep->rrset_count; i++) { 430 s = ((struct packed_rrset_data*)rep->rrsets[i]->entry.data) 431 ->security; 432 if(s == sec_status_unchecked) 433 return s; 434 } 435 return sec; 436 } 437 438 /** remove nonsecure from a delegation referral additional section */ 439 static void 440 deleg_remove_nonsecure_additional(struct reply_info* rep) 441 { 442 /* we can simply edit it, since we are working in the scratch region */ 443 size_t i; 444 enum sec_status s; 445 446 for(i = rep->an_numrrsets+rep->ns_numrrsets; i<rep->rrset_count; i++) { 447 s = ((struct packed_rrset_data*)rep->rrsets[i]->entry.data) 448 ->security; 449 if(s != sec_status_secure) { 450 memmove(rep->rrsets+i, rep->rrsets+i+1, 451 sizeof(struct ub_packed_rrset_key*)* 452 (rep->rrset_count - i - 1)); 453 rep->ar_numrrsets--; 454 rep->rrset_count--; 455 i--; 456 } 457 } 458 } 459 460 /** answer nonrecursive query from the cache */ 461 static int 462 answer_norec_from_cache(struct worker* worker, struct query_info* qinfo, 463 uint16_t id, uint16_t flags, struct comm_reply* repinfo, 464 struct edns_data* edns) 465 { 466 /* for a nonrecursive query return either: 467 * o an error (servfail; we try to avoid this) 468 * o a delegation (closest we have; this routine tries that) 469 * o the answer (checked by answer_from_cache) 470 * 471 * So, grab a delegation from the rrset cache. 472 * Then check if it needs validation, if so, this routine fails, 473 * so that iterator can prime and validator can verify rrsets. 474 */ 475 uint16_t udpsize = edns->udp_size; 476 int secure = 0; 477 time_t timenow = *worker->env.now; 478 int has_cd_bit = (flags&BIT_CD); 479 int must_validate = (!has_cd_bit || worker->env.cfg->ignore_cd) 480 && worker->env.need_to_validate; 481 struct dns_msg *msg = NULL; 482 struct delegpt *dp; 483 484 dp = dns_cache_find_delegation(&worker->env, qinfo->qname, 485 qinfo->qname_len, qinfo->qtype, qinfo->qclass, 486 worker->scratchpad, &msg, timenow, 0, NULL, 0); 487 if(!dp) { /* no delegation, need to reprime */ 488 return 0; 489 } 490 /* In case we have a local alias, copy it into the delegation message. 491 * Shallow copy should be fine, as we'll be done with msg in this 492 * function. */ 493 msg->qinfo.local_alias = qinfo->local_alias; 494 if(must_validate) { 495 switch(check_delegation_secure(msg->rep)) { 496 case sec_status_unchecked: 497 /* some rrsets have not been verified yet, go and 498 * let validator do that */ 499 return 0; 500 case sec_status_bogus: 501 case sec_status_secure_sentinel_fail: 502 /* some rrsets are bogus, reply servfail */ 503 edns->edns_version = EDNS_ADVERTISED_VERSION; 504 edns->udp_size = EDNS_ADVERTISED_SIZE; 505 edns->ext_rcode = 0; 506 edns->bits &= EDNS_DO; 507 if(!inplace_cb_reply_servfail_call(&worker->env, qinfo, NULL, 508 msg->rep, LDNS_RCODE_SERVFAIL, edns, repinfo, worker->scratchpad, 509 worker->env.now_tv)) 510 return 0; 511 /* Attach the cached EDE (RFC8914) */ 512 if(worker->env.cfg->ede && 513 msg->rep->reason_bogus != LDNS_EDE_NONE) { 514 edns_opt_list_append_ede(&edns->opt_list_out, 515 worker->scratchpad, msg->rep->reason_bogus, 516 msg->rep->reason_bogus_str); 517 } 518 error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL, 519 &msg->qinfo, id, flags, edns); 520 if(worker->stats.extended) { 521 worker->stats.ans_bogus++; 522 worker->stats.ans_rcode[LDNS_RCODE_SERVFAIL]++; 523 } 524 return 1; 525 case sec_status_secure: 526 /* all rrsets are secure */ 527 /* remove non-secure rrsets from the add. section*/ 528 if(worker->env.cfg->val_clean_additional) 529 deleg_remove_nonsecure_additional(msg->rep); 530 secure = 1; 531 break; 532 case sec_status_indeterminate: 533 case sec_status_insecure: 534 default: 535 /* not secure */ 536 secure = 0; 537 break; 538 } 539 } 540 /* return this delegation from the cache */ 541 edns->edns_version = EDNS_ADVERTISED_VERSION; 542 edns->udp_size = EDNS_ADVERTISED_SIZE; 543 edns->ext_rcode = 0; 544 edns->bits &= EDNS_DO; 545 if(!inplace_cb_reply_cache_call(&worker->env, qinfo, NULL, msg->rep, 546 (int)(flags&LDNS_RCODE_MASK), edns, repinfo, worker->scratchpad, 547 worker->env.now_tv)) 548 return 0; 549 msg->rep->flags |= BIT_QR|BIT_RA; 550 /* Attach the cached EDE (RFC8914) if CD bit is set and the answer is 551 * bogus. */ 552 if(worker->env.cfg->ede && has_cd_bit && 553 (check_delegation_secure(msg->rep) == sec_status_bogus || 554 check_delegation_secure(msg->rep) == sec_status_secure_sentinel_fail) && 555 msg->rep->reason_bogus != LDNS_EDE_NONE) { 556 edns_opt_list_append_ede(&edns->opt_list_out, 557 worker->scratchpad, msg->rep->reason_bogus, 558 msg->rep->reason_bogus_str); 559 } 560 if(!reply_info_answer_encode(&msg->qinfo, msg->rep, id, flags, 561 repinfo->c->buffer, 0, 1, worker->scratchpad, 562 udpsize, edns, (int)(edns->bits & EDNS_DO), secure)) { 563 if(!inplace_cb_reply_servfail_call(&worker->env, qinfo, NULL, NULL, 564 LDNS_RCODE_SERVFAIL, edns, repinfo, worker->scratchpad, 565 worker->env.now_tv)) 566 edns->opt_list_inplace_cb_out = NULL; 567 error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL, 568 &msg->qinfo, id, flags, edns); 569 } 570 if(worker->stats.extended) { 571 if(secure) worker->stats.ans_secure++; 572 server_stats_insrcode(&worker->stats, repinfo->c->buffer); 573 } 574 return 1; 575 } 576 577 /** Apply, if applicable, a response IP action to a cached answer. 578 * If the answer is rewritten as a result of an action, '*encode_repp' will 579 * point to the reply info containing the modified answer. '*encode_repp' will 580 * be intact otherwise. 581 * It returns 1 on success, 0 otherwise. */ 582 static int 583 apply_respip_action(struct worker* worker, const struct query_info* qinfo, 584 struct respip_client_info* cinfo, struct reply_info* rep, 585 struct sockaddr_storage* addr, socklen_t addrlen, 586 struct ub_packed_rrset_key** alias_rrset, 587 struct reply_info** encode_repp, struct auth_zones* az) 588 { 589 struct respip_action_info actinfo = {0, 0, 0, 0, NULL, 0, NULL}; 590 actinfo.action = respip_none; 591 592 if(qinfo->qtype != LDNS_RR_TYPE_A && 593 qinfo->qtype != LDNS_RR_TYPE_AAAA && 594 qinfo->qtype != LDNS_RR_TYPE_ANY) 595 return 1; 596 597 if(!respip_rewrite_reply(qinfo, cinfo, rep, encode_repp, &actinfo, 598 alias_rrset, 0, worker->scratchpad, az, NULL)) 599 return 0; 600 601 /* xxx_deny actions mean dropping the reply, unless the original reply 602 * was redirected to response-ip data. */ 603 if(actinfo.action == respip_always_deny || 604 ((actinfo.action == respip_deny || 605 actinfo.action == respip_inform_deny) && 606 *encode_repp == rep)) 607 *encode_repp = NULL; 608 609 /* If address info is returned, it means the action should be an 610 * 'inform' variant and the information should be logged. */ 611 if(actinfo.addrinfo) { 612 respip_inform_print(&actinfo, qinfo->qname, 613 qinfo->qtype, qinfo->qclass, qinfo->local_alias, 614 addr, addrlen); 615 616 if(worker->stats.extended && actinfo.rpz_used) { 617 if(actinfo.rpz_disabled) 618 worker->stats.rpz_action[RPZ_DISABLED_ACTION]++; 619 if(actinfo.rpz_cname_override) 620 worker->stats.rpz_action[RPZ_CNAME_OVERRIDE_ACTION]++; 621 else 622 worker->stats.rpz_action[ 623 respip_action_to_rpz_action(actinfo.action)]++; 624 } 625 } 626 627 return 1; 628 } 629 630 /** answer query from the cache. 631 * Normally, the answer message will be built in repinfo->c->buffer; if the 632 * answer is supposed to be suppressed or the answer is supposed to be an 633 * incomplete CNAME chain, the buffer is explicitly cleared to signal the 634 * caller as such. In the latter case *partial_rep will point to the incomplete 635 * reply, and this function is (possibly) supposed to be called again with that 636 * *partial_rep value to complete the chain. In addition, if the query should 637 * be completely dropped, '*need_drop' will be set to 1. */ 638 static int 639 answer_from_cache(struct worker* worker, struct query_info* qinfo, 640 struct respip_client_info* cinfo, int* need_drop, int* is_expired_answer, 641 int* is_secure_answer, struct ub_packed_rrset_key** alias_rrset, 642 struct reply_info** partial_repp, 643 struct reply_info* rep, uint16_t id, uint16_t flags, 644 struct comm_reply* repinfo, struct edns_data* edns) 645 { 646 time_t timenow = *worker->env.now; 647 uint16_t udpsize = edns->udp_size; 648 struct reply_info* encode_rep = rep; 649 struct reply_info* partial_rep = *partial_repp; 650 int has_cd_bit = (flags&BIT_CD); 651 int must_validate = (!has_cd_bit || worker->env.cfg->ignore_cd) 652 && worker->env.need_to_validate; 653 *partial_repp = NULL; /* avoid accidental further pass */ 654 655 /* Check TTL */ 656 if(rep->ttl < timenow) { 657 /* Check if we need to serve expired now */ 658 if(worker->env.cfg->serve_expired && 659 !worker->env.cfg->serve_expired_client_timeout) { 660 if(worker->env.cfg->serve_expired_ttl && 661 rep->serve_expired_ttl < timenow) 662 return 0; 663 /* Ignore expired failure answers */ 664 if(FLAGS_GET_RCODE(rep->flags) != 665 LDNS_RCODE_NOERROR && 666 FLAGS_GET_RCODE(rep->flags) != 667 LDNS_RCODE_NXDOMAIN && 668 FLAGS_GET_RCODE(rep->flags) != 669 LDNS_RCODE_YXDOMAIN) 670 return 0; 671 if(!rrset_array_lock(rep->ref, rep->rrset_count, 0)) 672 return 0; 673 *is_expired_answer = 1; 674 } else { 675 /* the rrsets may have been updated in the meantime. 676 * we will refetch the message format from the 677 * authoritative server 678 */ 679 return 0; 680 } 681 } else { 682 if(!rrset_array_lock(rep->ref, rep->rrset_count, timenow)) 683 return 0; 684 } 685 /* locked and ids and ttls are OK. */ 686 687 /* check CNAME chain (if any) */ 688 if(rep->an_numrrsets > 0 && (rep->rrsets[0]->rk.type == 689 htons(LDNS_RR_TYPE_CNAME) || rep->rrsets[0]->rk.type == 690 htons(LDNS_RR_TYPE_DNAME))) { 691 if(!reply_check_cname_chain(qinfo, rep)) { 692 /* cname chain invalid, redo iterator steps */ 693 verbose(VERB_ALGO, "Cache reply: cname chain broken"); 694 goto bail_out; 695 } 696 } 697 /* check security status of the cached answer */ 698 if(must_validate && (rep->security == sec_status_bogus || 699 rep->security == sec_status_secure_sentinel_fail)) { 700 /* BAD cached */ 701 edns->edns_version = EDNS_ADVERTISED_VERSION; 702 edns->udp_size = EDNS_ADVERTISED_SIZE; 703 edns->ext_rcode = 0; 704 edns->bits &= EDNS_DO; 705 if(!inplace_cb_reply_servfail_call(&worker->env, qinfo, NULL, rep, 706 LDNS_RCODE_SERVFAIL, edns, repinfo, worker->scratchpad, 707 worker->env.now_tv)) 708 goto bail_out; 709 /* Attach the cached EDE (RFC8914) */ 710 if(worker->env.cfg->ede && rep->reason_bogus != LDNS_EDE_NONE) { 711 edns_opt_list_append_ede(&edns->opt_list_out, 712 worker->scratchpad, rep->reason_bogus, 713 rep->reason_bogus_str); 714 } 715 error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL, 716 qinfo, id, flags, edns); 717 rrset_array_unlock_touch(worker->env.rrset_cache, 718 worker->scratchpad, rep->ref, rep->rrset_count); 719 if(worker->stats.extended) { 720 worker->stats.ans_bogus ++; 721 worker->stats.ans_rcode[LDNS_RCODE_SERVFAIL] ++; 722 } 723 return 1; 724 } else if(rep->security == sec_status_unchecked && must_validate) { 725 verbose(VERB_ALGO, "Cache reply: unchecked entry needs " 726 "validation"); 727 goto bail_out; /* need to validate cache entry first */ 728 } else if(rep->security == sec_status_secure) { 729 if(reply_all_rrsets_secure(rep)) { 730 *is_secure_answer = 1; 731 } else { 732 if(must_validate) { 733 verbose(VERB_ALGO, "Cache reply: secure entry" 734 " changed status"); 735 goto bail_out; /* rrset changed, re-verify */ 736 } 737 *is_secure_answer = 0; 738 } 739 } else *is_secure_answer = 0; 740 741 edns->edns_version = EDNS_ADVERTISED_VERSION; 742 edns->udp_size = EDNS_ADVERTISED_SIZE; 743 edns->ext_rcode = 0; 744 edns->bits &= EDNS_DO; 745 *alias_rrset = NULL; /* avoid confusion if caller set it to non-NULL */ 746 if((worker->daemon->use_response_ip || worker->daemon->use_rpz) && 747 !partial_rep && !apply_respip_action(worker, qinfo, cinfo, rep, 748 &repinfo->client_addr, repinfo->client_addrlen, alias_rrset, 749 &encode_rep, worker->env.auth_zones)) { 750 goto bail_out; 751 } else if(partial_rep && 752 !respip_merge_cname(partial_rep, qinfo, rep, cinfo, 753 must_validate, &encode_rep, worker->scratchpad, 754 worker->env.auth_zones)) { 755 goto bail_out; 756 } 757 if(encode_rep != rep) { 758 /* if rewritten, it can't be considered "secure" */ 759 *is_secure_answer = 0; 760 } 761 if(!encode_rep || *alias_rrset) { 762 if(!encode_rep) 763 *need_drop = 1; 764 else { 765 /* If a partial CNAME chain is found, we first need to 766 * make a copy of the reply in the scratchpad so we 767 * can release the locks and lookup the cache again. */ 768 *partial_repp = reply_info_copy(encode_rep, NULL, 769 worker->scratchpad); 770 if(!*partial_repp) 771 goto bail_out; 772 } 773 } else { 774 if(*is_expired_answer == 1 && 775 worker->env.cfg->ede_serve_expired && worker->env.cfg->ede) { 776 EDNS_OPT_LIST_APPEND_EDE(&edns->opt_list_out, 777 worker->scratchpad, LDNS_EDE_STALE_ANSWER, ""); 778 } 779 /* Attach the cached EDE (RFC8914) if CD bit is set and the 780 * answer is bogus. */ 781 if(*is_secure_answer == 0 && 782 worker->env.cfg->ede && has_cd_bit && 783 encode_rep->reason_bogus != LDNS_EDE_NONE) { 784 edns_opt_list_append_ede(&edns->opt_list_out, 785 worker->scratchpad, encode_rep->reason_bogus, 786 encode_rep->reason_bogus_str); 787 } 788 if(!inplace_cb_reply_cache_call(&worker->env, qinfo, NULL, encode_rep, 789 (int)(flags&LDNS_RCODE_MASK), edns, repinfo, worker->scratchpad, 790 worker->env.now_tv)) 791 goto bail_out; 792 if(!reply_info_answer_encode(qinfo, encode_rep, id, flags, 793 repinfo->c->buffer, timenow, 1, worker->scratchpad, 794 udpsize, edns, (int)(edns->bits & EDNS_DO), 795 *is_secure_answer)) { 796 if(!inplace_cb_reply_servfail_call(&worker->env, qinfo, 797 NULL, NULL, LDNS_RCODE_SERVFAIL, edns, repinfo, 798 worker->scratchpad, worker->env.now_tv)) 799 edns->opt_list_inplace_cb_out = NULL; 800 error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL, 801 qinfo, id, flags, edns); 802 } 803 } 804 /* cannot send the reply right now, because blocking network syscall 805 * is bad while holding locks. */ 806 rrset_array_unlock_touch(worker->env.rrset_cache, worker->scratchpad, 807 rep->ref, rep->rrset_count); 808 /* go and return this buffer to the client */ 809 return 1; 810 811 bail_out: 812 rrset_array_unlock_touch(worker->env.rrset_cache, 813 worker->scratchpad, rep->ref, rep->rrset_count); 814 return 0; 815 } 816 817 /** Reply to client and perform prefetch to keep cache up to date. */ 818 static void 819 reply_and_prefetch(struct worker* worker, struct query_info* qinfo, 820 uint16_t flags, struct comm_reply* repinfo, time_t leeway, int noreply, 821 int rpz_passthru, struct edns_option* opt_list) 822 { 823 (void)opt_list; 824 /* first send answer to client to keep its latency 825 * as small as a cachereply */ 826 if(!noreply) { 827 if(repinfo->c->tcp_req_info) { 828 sldns_buffer_copy( 829 repinfo->c->tcp_req_info->spool_buffer, 830 repinfo->c->buffer); 831 } 832 comm_point_send_reply(repinfo); 833 } 834 server_stats_prefetch(&worker->stats, worker); 835 #ifdef CLIENT_SUBNET 836 /* Check if the subnet module is enabled. In that case pass over the 837 * comm_reply information for ECS generation later. The mesh states are 838 * unique when subnet is enabled. */ 839 if(modstack_find(&worker->env.mesh->mods, "subnetcache") != -1 840 && worker->env.unique_mesh) { 841 mesh_new_prefetch(worker->env.mesh, qinfo, flags, leeway + 842 PREFETCH_EXPIRY_ADD, rpz_passthru, 843 &repinfo->client_addr, opt_list); 844 return; 845 } 846 #endif 847 /* create the prefetch in the mesh as a normal lookup without 848 * client addrs waiting, which has the cache blacklisted (to bypass 849 * the cache and go to the network for the data). */ 850 /* this (potentially) runs the mesh for the new query */ 851 mesh_new_prefetch(worker->env.mesh, qinfo, flags, leeway + 852 PREFETCH_EXPIRY_ADD, rpz_passthru, NULL, NULL); 853 } 854 855 /** 856 * Fill CH class answer into buffer. Keeps query. 857 * @param pkt: buffer 858 * @param str: string to put into text record (<255). 859 * array of strings, every string becomes a text record. 860 * @param num: number of strings in array. 861 * @param edns: edns reply information. 862 * @param worker: worker with scratch region. 863 * @param repinfo: reply information for a communication point. 864 */ 865 static void 866 chaos_replystr(sldns_buffer* pkt, char** str, int num, struct edns_data* edns, 867 struct worker* worker, struct comm_reply* repinfo) 868 { 869 int i; 870 unsigned int rd = LDNS_RD_WIRE(sldns_buffer_begin(pkt)); 871 unsigned int cd = LDNS_CD_WIRE(sldns_buffer_begin(pkt)); 872 size_t udpsize = edns->udp_size; 873 edns->edns_version = EDNS_ADVERTISED_VERSION; 874 edns->udp_size = EDNS_ADVERTISED_SIZE; 875 edns->bits &= EDNS_DO; 876 if(!inplace_cb_reply_local_call(&worker->env, NULL, NULL, NULL, 877 LDNS_RCODE_NOERROR, edns, repinfo, worker->scratchpad, 878 worker->env.now_tv)) 879 edns->opt_list_inplace_cb_out = NULL; 880 sldns_buffer_clear(pkt); 881 sldns_buffer_skip(pkt, (ssize_t)sizeof(uint16_t)); /* skip id */ 882 sldns_buffer_write_u16(pkt, (uint16_t)(BIT_QR|BIT_RA)); 883 if(rd) LDNS_RD_SET(sldns_buffer_begin(pkt)); 884 if(cd) LDNS_CD_SET(sldns_buffer_begin(pkt)); 885 sldns_buffer_write_u16(pkt, 1); /* qdcount */ 886 sldns_buffer_write_u16(pkt, (uint16_t)num); /* ancount */ 887 sldns_buffer_write_u16(pkt, 0); /* nscount */ 888 sldns_buffer_write_u16(pkt, 0); /* arcount */ 889 (void)query_dname_len(pkt); /* skip qname */ 890 sldns_buffer_skip(pkt, (ssize_t)sizeof(uint16_t)); /* skip qtype */ 891 sldns_buffer_skip(pkt, (ssize_t)sizeof(uint16_t)); /* skip qclass */ 892 for(i=0; i<num; i++) { 893 size_t len = strlen(str[i]); 894 if(len>255) len=255; /* cap size of TXT record */ 895 if(sldns_buffer_position(pkt)+2+2+2+4+2+1+len+ 896 calc_edns_field_size(edns) > udpsize) { 897 sldns_buffer_write_u16_at(pkt, 6, i); /* ANCOUNT */ 898 LDNS_TC_SET(sldns_buffer_begin(pkt)); 899 break; 900 } 901 sldns_buffer_write_u16(pkt, 0xc00c); /* compr ptr to query */ 902 sldns_buffer_write_u16(pkt, LDNS_RR_TYPE_TXT); 903 sldns_buffer_write_u16(pkt, LDNS_RR_CLASS_CH); 904 sldns_buffer_write_u32(pkt, 0); /* TTL */ 905 sldns_buffer_write_u16(pkt, sizeof(uint8_t) + len); 906 sldns_buffer_write_u8(pkt, len); 907 sldns_buffer_write(pkt, str[i], len); 908 } 909 sldns_buffer_flip(pkt); 910 if(sldns_buffer_capacity(pkt) >= 911 sldns_buffer_limit(pkt)+calc_edns_field_size(edns)) 912 attach_edns_record(pkt, edns); 913 } 914 915 /** Reply with one string */ 916 static void 917 chaos_replyonestr(sldns_buffer* pkt, const char* str, struct edns_data* edns, 918 struct worker* worker, struct comm_reply* repinfo) 919 { 920 chaos_replystr(pkt, (char**)&str, 1, edns, worker, repinfo); 921 } 922 923 /** 924 * Create CH class trustanchor answer. 925 * @param pkt: buffer 926 * @param edns: edns reply information. 927 * @param w: worker with scratch region. 928 * @param repinfo: reply information for a communication point. 929 */ 930 static void 931 chaos_trustanchor(sldns_buffer* pkt, struct edns_data* edns, struct worker* w, 932 struct comm_reply* repinfo) 933 { 934 #define TA_RESPONSE_MAX_TXT 16 /* max number of TXT records */ 935 #define TA_RESPONSE_MAX_TAGS 32 /* max number of tags printed per zone */ 936 char* str_array[TA_RESPONSE_MAX_TXT]; 937 uint16_t tags[TA_RESPONSE_MAX_TAGS]; 938 int num = 0; 939 struct trust_anchor* ta; 940 941 if(!w->env.need_to_validate) { 942 /* no validator module, reply no trustanchors */ 943 chaos_replystr(pkt, NULL, 0, edns, w, repinfo); 944 return; 945 } 946 947 /* fill the string with contents */ 948 lock_basic_lock(&w->env.anchors->lock); 949 RBTREE_FOR(ta, struct trust_anchor*, w->env.anchors->tree) { 950 char* str; 951 size_t i, numtag, str_len = 255; 952 if(num == TA_RESPONSE_MAX_TXT) continue; 953 str = (char*)regional_alloc(w->scratchpad, str_len); 954 if(!str) continue; 955 lock_basic_lock(&ta->lock); 956 numtag = anchor_list_keytags(ta, tags, TA_RESPONSE_MAX_TAGS); 957 if(numtag == 0) { 958 /* empty, insecure point */ 959 lock_basic_unlock(&ta->lock); 960 continue; 961 } 962 str_array[num] = str; 963 num++; 964 965 /* spool name of anchor */ 966 (void)sldns_wire2str_dname_buf(ta->name, ta->namelen, str, str_len); 967 str_len -= strlen(str); str += strlen(str); 968 /* spool tags */ 969 for(i=0; i<numtag; i++) { 970 snprintf(str, str_len, " %u", (unsigned)tags[i]); 971 str_len -= strlen(str); str += strlen(str); 972 } 973 lock_basic_unlock(&ta->lock); 974 } 975 lock_basic_unlock(&w->env.anchors->lock); 976 977 chaos_replystr(pkt, str_array, num, edns, w, repinfo); 978 regional_free_all(w->scratchpad); 979 } 980 981 /** 982 * Answer CH class queries. 983 * @param w: worker 984 * @param qinfo: query info. Pointer into packet buffer. 985 * @param edns: edns info from query. 986 * @param repinfo: reply information for a communication point. 987 * @param pkt: packet buffer. 988 * @return: true if a reply is to be sent. 989 */ 990 static int 991 answer_chaos(struct worker* w, struct query_info* qinfo, 992 struct edns_data* edns, struct comm_reply* repinfo, sldns_buffer* pkt) 993 { 994 struct config_file* cfg = w->env.cfg; 995 if(qinfo->qtype != LDNS_RR_TYPE_ANY && qinfo->qtype != LDNS_RR_TYPE_TXT) 996 return 0; 997 if(query_dname_compare(qinfo->qname, 998 (uint8_t*)"\002id\006server") == 0 || 999 query_dname_compare(qinfo->qname, 1000 (uint8_t*)"\010hostname\004bind") == 0) 1001 { 1002 if(cfg->hide_identity) 1003 return 0; 1004 if(cfg->identity==NULL || cfg->identity[0]==0) { 1005 char buf[MAXHOSTNAMELEN+1]; 1006 if (gethostname(buf, MAXHOSTNAMELEN) == 0) { 1007 buf[MAXHOSTNAMELEN] = 0; 1008 chaos_replyonestr(pkt, buf, edns, w, repinfo); 1009 } else { 1010 log_err("gethostname: %s", strerror(errno)); 1011 chaos_replyonestr(pkt, "no hostname", edns, w, repinfo); 1012 } 1013 } 1014 else chaos_replyonestr(pkt, cfg->identity, edns, w, repinfo); 1015 return 1; 1016 } 1017 if(query_dname_compare(qinfo->qname, 1018 (uint8_t*)"\007version\006server") == 0 || 1019 query_dname_compare(qinfo->qname, 1020 (uint8_t*)"\007version\004bind") == 0) 1021 { 1022 if(cfg->hide_version) 1023 return 0; 1024 if(cfg->version==NULL || cfg->version[0]==0) 1025 chaos_replyonestr(pkt, PACKAGE_STRING, edns, w, repinfo); 1026 else chaos_replyonestr(pkt, cfg->version, edns, w, repinfo); 1027 return 1; 1028 } 1029 if(query_dname_compare(qinfo->qname, 1030 (uint8_t*)"\013trustanchor\007unbound") == 0) 1031 { 1032 if(cfg->hide_trustanchor) 1033 return 0; 1034 chaos_trustanchor(pkt, edns, w, repinfo); 1035 return 1; 1036 } 1037 1038 return 0; 1039 } 1040 1041 /** 1042 * Answer notify queries. These are notifies for authoritative zones, 1043 * the reply is an ack that the notify has been received. We need to check 1044 * access permission here. 1045 * @param w: worker 1046 * @param qinfo: query info. Pointer into packet buffer. 1047 * @param edns: edns info from query. 1048 * @param addr: client address. 1049 * @param addrlen: client address length. 1050 * @param pkt: packet buffer. 1051 */ 1052 static void 1053 answer_notify(struct worker* w, struct query_info* qinfo, 1054 struct edns_data* edns, sldns_buffer* pkt, 1055 struct sockaddr_storage* addr, socklen_t addrlen) 1056 { 1057 int refused = 0; 1058 int rcode = LDNS_RCODE_NOERROR; 1059 uint32_t serial = 0; 1060 int has_serial; 1061 if(!w->env.auth_zones) return; 1062 has_serial = auth_zone_parse_notify_serial(pkt, &serial); 1063 if(auth_zones_notify(w->env.auth_zones, &w->env, qinfo->qname, 1064 qinfo->qname_len, qinfo->qclass, addr, 1065 addrlen, has_serial, serial, &refused)) { 1066 rcode = LDNS_RCODE_NOERROR; 1067 } else { 1068 if(refused) 1069 rcode = LDNS_RCODE_REFUSED; 1070 else rcode = LDNS_RCODE_SERVFAIL; 1071 } 1072 1073 if(verbosity >= VERB_DETAIL) { 1074 char buf[380]; 1075 char zname[255+1]; 1076 char sr[25]; 1077 dname_str(qinfo->qname, zname); 1078 sr[0]=0; 1079 if(has_serial) 1080 snprintf(sr, sizeof(sr), "serial %u ", 1081 (unsigned)serial); 1082 if(rcode == LDNS_RCODE_REFUSED) 1083 snprintf(buf, sizeof(buf), 1084 "refused NOTIFY %sfor %s from", sr, zname); 1085 else if(rcode == LDNS_RCODE_SERVFAIL) 1086 snprintf(buf, sizeof(buf), 1087 "servfail for NOTIFY %sfor %s from", sr, zname); 1088 else snprintf(buf, sizeof(buf), 1089 "received NOTIFY %sfor %s from", sr, zname); 1090 log_addr(VERB_DETAIL, buf, addr, addrlen); 1091 } 1092 edns->edns_version = EDNS_ADVERTISED_VERSION; 1093 edns->udp_size = EDNS_ADVERTISED_SIZE; 1094 edns->ext_rcode = 0; 1095 edns->bits &= EDNS_DO; 1096 error_encode(pkt, rcode, qinfo, 1097 *(uint16_t*)(void *)sldns_buffer_begin(pkt), 1098 sldns_buffer_read_u16_at(pkt, 2), edns); 1099 LDNS_OPCODE_SET(sldns_buffer_begin(pkt), LDNS_PACKET_NOTIFY); 1100 } 1101 1102 static int 1103 deny_refuse(struct comm_point* c, enum acl_access acl, 1104 enum acl_access deny, enum acl_access refuse, 1105 struct worker* worker, struct comm_reply* repinfo, 1106 struct acl_addr* acladdr, int ede, 1107 struct check_request_result* check_result) 1108 { 1109 if(acl == deny) { 1110 if(verbosity >= VERB_ALGO) { 1111 log_acl_action("dropped", &repinfo->client_addr, 1112 repinfo->client_addrlen, acl, acladdr); 1113 log_buf(VERB_ALGO, "dropped", c->buffer); 1114 } 1115 comm_point_drop_reply(repinfo); 1116 if(worker->stats.extended) 1117 worker->stats.unwanted_queries++; 1118 return 0; 1119 } else if(acl == refuse) { 1120 size_t opt_rr_mark; 1121 1122 if(verbosity >= VERB_ALGO) { 1123 log_acl_action("refused", &repinfo->client_addr, 1124 repinfo->client_addrlen, acl, acladdr); 1125 log_buf(VERB_ALGO, "refuse", c->buffer); 1126 } 1127 1128 if(worker->stats.extended) 1129 worker->stats.unwanted_queries++; 1130 worker_check_request(c->buffer, worker, check_result); 1131 if(check_result->value != 0) { 1132 if(check_result->value != -1) { 1133 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1134 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1135 check_result->value); 1136 return 1; 1137 } 1138 comm_point_drop_reply(repinfo); 1139 return 0; 1140 } 1141 /* worker_check_request() above guarantees that the buffer contains at 1142 * least a header and that qdcount == 1 1143 */ 1144 log_assert(sldns_buffer_limit(c->buffer) >= LDNS_HEADER_SIZE 1145 && LDNS_QDCOUNT(sldns_buffer_begin(c->buffer)) == 1); 1146 1147 sldns_buffer_skip(c->buffer, LDNS_HEADER_SIZE); /* skip header */ 1148 1149 /* check additional section is present and that we respond with EDEs */ 1150 if(LDNS_ARCOUNT(sldns_buffer_begin(c->buffer)) != 1 1151 || !ede) { 1152 LDNS_QDCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1153 LDNS_ANCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1154 LDNS_NSCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1155 LDNS_ARCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1156 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1157 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1158 LDNS_RCODE_REFUSED); 1159 sldns_buffer_flip(c->buffer); 1160 return 1; 1161 } 1162 1163 if (!query_dname_len(c->buffer)) { 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_FORMERR); 1171 sldns_buffer_set_position(c->buffer, LDNS_HEADER_SIZE); 1172 sldns_buffer_flip(c->buffer); 1173 return 1; 1174 } 1175 /* space available for query type and class? */ 1176 if (sldns_buffer_remaining(c->buffer) < 2 * sizeof(uint16_t)) { 1177 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1178 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1179 LDNS_RCODE_FORMERR); 1180 LDNS_QDCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1181 LDNS_ANCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1182 LDNS_NSCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1183 LDNS_ARCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1184 sldns_buffer_set_position(c->buffer, LDNS_HEADER_SIZE); 1185 sldns_buffer_flip(c->buffer); 1186 return 1; 1187 } 1188 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1189 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1190 LDNS_RCODE_REFUSED); 1191 1192 sldns_buffer_skip(c->buffer, (ssize_t)sizeof(uint16_t)); /* skip qtype */ 1193 1194 sldns_buffer_skip(c->buffer, (ssize_t)sizeof(uint16_t)); /* skip qclass */ 1195 1196 /* The OPT RR to be returned should come directly after 1197 * the query, so mark this spot. 1198 */ 1199 opt_rr_mark = sldns_buffer_position(c->buffer); 1200 1201 /* Skip through the RR records */ 1202 if(LDNS_ANCOUNT(sldns_buffer_begin(c->buffer)) != 0 || 1203 LDNS_NSCOUNT(sldns_buffer_begin(c->buffer)) != 0) { 1204 if(!skip_pkt_rrs(c->buffer, 1205 ((int)LDNS_ANCOUNT(sldns_buffer_begin(c->buffer)))+ 1206 ((int)LDNS_NSCOUNT(sldns_buffer_begin(c->buffer))))) { 1207 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1208 LDNS_RCODE_FORMERR); 1209 LDNS_ANCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1210 LDNS_NSCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1211 LDNS_ARCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1212 sldns_buffer_set_position(c->buffer, opt_rr_mark); 1213 sldns_buffer_flip(c->buffer); 1214 return 1; 1215 } 1216 } 1217 /* Do we have a valid OPT RR here? If not return REFUSED (could be a valid TSIG or something so no FORMERR) */ 1218 /* domain name must be the root of length 1. */ 1219 if(sldns_buffer_remaining(c->buffer) < 1 || *sldns_buffer_current(c->buffer) != 0) { 1220 LDNS_ANCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1221 LDNS_NSCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1222 LDNS_ARCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1223 sldns_buffer_set_position(c->buffer, opt_rr_mark); 1224 sldns_buffer_flip(c->buffer); 1225 return 1; 1226 } else { 1227 sldns_buffer_skip(c->buffer, 1); /* skip root label */ 1228 } 1229 if(sldns_buffer_remaining(c->buffer) < 2 || 1230 sldns_buffer_read_u16(c->buffer) != LDNS_RR_TYPE_OPT) { 1231 LDNS_ANCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1232 LDNS_NSCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1233 LDNS_ARCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1234 sldns_buffer_set_position(c->buffer, opt_rr_mark); 1235 sldns_buffer_flip(c->buffer); 1236 return 1; 1237 } 1238 /* Write OPT RR directly after the query, 1239 * so without the (possibly skipped) Answer and NS RRs 1240 */ 1241 LDNS_ANCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1242 LDNS_NSCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1243 sldns_buffer_clear(c->buffer); /* reset write limit */ 1244 sldns_buffer_set_position(c->buffer, opt_rr_mark); 1245 1246 /* Check if OPT record can be written 1247 * 17 == root label (1) + RR type (2) + UDP Size (2) 1248 * + Fields (4) + rdata len (2) + EDE Option code (2) 1249 * + EDE Option length (2) + EDE info-code (2) 1250 */ 1251 if (sldns_buffer_available(c->buffer, 17) == 0) { 1252 LDNS_ARCOUNT_SET(sldns_buffer_begin(c->buffer), 0); 1253 sldns_buffer_flip(c->buffer); 1254 return 1; 1255 } 1256 1257 LDNS_ARCOUNT_SET(sldns_buffer_begin(c->buffer), 1); 1258 1259 /* root label */ 1260 sldns_buffer_write_u8(c->buffer, 0); 1261 sldns_buffer_write_u16(c->buffer, LDNS_RR_TYPE_OPT); 1262 sldns_buffer_write_u16(c->buffer, EDNS_ADVERTISED_SIZE); 1263 1264 /* write OPT Record TTL Field */ 1265 sldns_buffer_write_u32(c->buffer, 0); 1266 1267 /* write rdata len: EDE option + length + info-code */ 1268 sldns_buffer_write_u16(c->buffer, 6); 1269 1270 /* write OPTIONS; add EDE option code */ 1271 sldns_buffer_write_u16(c->buffer, LDNS_EDNS_EDE); 1272 1273 /* write single EDE option length (for just 1 info-code) */ 1274 sldns_buffer_write_u16(c->buffer, 2); 1275 1276 /* write single EDE info-code */ 1277 sldns_buffer_write_u16(c->buffer, LDNS_EDE_PROHIBITED); 1278 1279 sldns_buffer_flip(c->buffer); 1280 1281 verbose(VERB_ALGO, "attached EDE code: %d", LDNS_EDE_PROHIBITED); 1282 1283 return 1; 1284 1285 } 1286 1287 return -1; 1288 } 1289 1290 static int 1291 deny_refuse_all(struct comm_point* c, enum acl_access* acl, 1292 struct worker* worker, struct comm_reply* repinfo, 1293 struct acl_addr** acladdr, int ede, int check_proxy, 1294 struct check_request_result* check_result) 1295 { 1296 if(check_proxy) { 1297 *acladdr = acl_addr_lookup(worker->daemon->acl, 1298 &repinfo->remote_addr, repinfo->remote_addrlen); 1299 } else { 1300 *acladdr = acl_addr_lookup(worker->daemon->acl, 1301 &repinfo->client_addr, repinfo->client_addrlen); 1302 } 1303 /* If there is no ACL based on client IP use the interface ACL. */ 1304 if(!(*acladdr) && c->socket) { 1305 *acladdr = c->socket->acl; 1306 } 1307 *acl = acl_get_control(*acladdr); 1308 return deny_refuse(c, *acl, acl_deny, acl_refuse, worker, repinfo, 1309 *acladdr, ede, check_result); 1310 } 1311 1312 static int 1313 deny_refuse_non_local(struct comm_point* c, enum acl_access acl, 1314 struct worker* worker, struct comm_reply* repinfo, 1315 struct acl_addr* acladdr, int ede, 1316 struct check_request_result* check_result) 1317 { 1318 return deny_refuse(c, acl, acl_deny_non_local, acl_refuse_non_local, 1319 worker, repinfo, acladdr, ede, check_result); 1320 } 1321 1322 /* Returns 1 if the ip rate limit check can happen before EDNS parsing, 1323 * else 0 */ 1324 static int 1325 pre_edns_ip_ratelimit_check(enum acl_access acl) 1326 { 1327 if(acl == acl_allow_cookie) return 0; 1328 return 1; 1329 } 1330 1331 /* Check if the query is blocked by source IP rate limiting. 1332 * Returns 1 if it passes the check, 0 otherwise. */ 1333 static int 1334 check_ip_ratelimit(struct worker* worker, struct sockaddr_storage* addr, 1335 socklen_t addrlen, int has_cookie, sldns_buffer* pkt) 1336 { 1337 if(!infra_ip_ratelimit_inc(worker->env.infra_cache, addr, addrlen, 1338 *worker->env.now, has_cookie, 1339 worker->env.cfg->ip_ratelimit_backoff, pkt)) { 1340 /* See if we can pass through with slip factor */ 1341 if(!has_cookie && worker->env.cfg->ip_ratelimit_factor != 0 && 1342 ub_random_max(worker->env.rnd, 1343 worker->env.cfg->ip_ratelimit_factor) == 0) { 1344 char addrbuf[128]; 1345 addr_to_str(addr, addrlen, addrbuf, sizeof(addrbuf)); 1346 verbose(VERB_QUERY, "ip_ratelimit allowed through for " 1347 "ip address %s because of slip in " 1348 "ip_ratelimit_factor", addrbuf); 1349 return 1; 1350 } 1351 return 0; 1352 } 1353 return 1; 1354 } 1355 1356 int 1357 worker_handle_request(struct comm_point* c, void* arg, int error, 1358 struct comm_reply* repinfo) 1359 { 1360 struct worker* worker = (struct worker*)arg; 1361 int ret; 1362 hashvalue_type h; 1363 struct lruhash_entry* e; 1364 struct query_info qinfo; 1365 struct edns_data edns; 1366 struct edns_option* original_edns_list = NULL; 1367 enum acl_access acl; 1368 struct acl_addr* acladdr; 1369 int pre_edns_ip_ratelimit = 1; 1370 int rc = 0; 1371 int need_drop = 0; 1372 int is_expired_answer = 0; 1373 int is_secure_answer = 0; 1374 int rpz_passthru = 0; 1375 long long wait_queue_time = 0; 1376 /* We might have to chase a CNAME chain internally, in which case 1377 * we'll have up to two replies and combine them to build a complete 1378 * answer. These variables control this case. */ 1379 struct ub_packed_rrset_key* alias_rrset = NULL; 1380 struct reply_info* partial_rep = NULL; 1381 struct query_info* lookup_qinfo = &qinfo; 1382 struct query_info qinfo_tmp; /* placeholder for lookup_qinfo */ 1383 struct respip_client_info* cinfo = NULL, cinfo_tmp; 1384 struct timeval wait_time; 1385 struct check_request_result check_result = {0,0}; 1386 memset(&qinfo, 0, sizeof(qinfo)); 1387 1388 if((error != NETEVENT_NOERROR && error != NETEVENT_DONE)|| !repinfo) { 1389 /* some bad tcp query DNS formats give these error calls */ 1390 verbose(VERB_ALGO, "handle request called with err=%d", error); 1391 return 0; 1392 } 1393 1394 if (worker->env.cfg->sock_queue_timeout && timeval_isset(&c->recv_tv)) { 1395 timeval_subtract(&wait_time, worker->env.now_tv, &c->recv_tv); 1396 wait_queue_time = wait_time.tv_sec * 1000000 + wait_time.tv_usec; 1397 if (worker->stats.max_query_time_us < wait_queue_time) 1398 worker->stats.max_query_time_us = wait_queue_time; 1399 if(wait_queue_time > 1400 (long long)(worker->env.cfg->sock_queue_timeout * 1000000)) { 1401 /* count and drop queries that were sitting in the socket queue too long */ 1402 worker->stats.num_queries_timed_out++; 1403 return 0; 1404 } 1405 } 1406 1407 #ifdef USE_DNSCRYPT 1408 repinfo->max_udp_size = worker->daemon->cfg->max_udp_size; 1409 if(!dnsc_handle_curved_request(worker->daemon->dnscenv, repinfo)) { 1410 worker->stats.num_query_dnscrypt_crypted_malformed++; 1411 return 0; 1412 } 1413 if(c->dnscrypt && !repinfo->is_dnscrypted) { 1414 char buf[LDNS_MAX_DOMAINLEN+1]; 1415 /* Check if this is unencrypted and asking for certs */ 1416 worker_check_request(c->buffer, worker, &check_result); 1417 if(check_result.value != 0) { 1418 verbose(VERB_ALGO, 1419 "dnscrypt: worker check request: bad query."); 1420 log_addr(VERB_CLIENT,"from",&repinfo->client_addr, 1421 repinfo->client_addrlen); 1422 comm_point_drop_reply(repinfo); 1423 return 0; 1424 } 1425 if(!query_info_parse(&qinfo, c->buffer)) { 1426 verbose(VERB_ALGO, 1427 "dnscrypt: worker parse request: formerror."); 1428 log_addr(VERB_CLIENT, "from", &repinfo->client_addr, 1429 repinfo->client_addrlen); 1430 comm_point_drop_reply(repinfo); 1431 return 0; 1432 } 1433 dname_str(qinfo.qname, buf); 1434 if(!(qinfo.qtype == LDNS_RR_TYPE_TXT && 1435 strcasecmp(buf, 1436 worker->daemon->dnscenv->provider_name) == 0)) { 1437 verbose(VERB_ALGO, 1438 "dnscrypt: not TXT \"%s\". Received: %s \"%s\"", 1439 worker->daemon->dnscenv->provider_name, 1440 sldns_rr_descript(qinfo.qtype)->_name, 1441 buf); 1442 comm_point_drop_reply(repinfo); 1443 worker->stats.num_query_dnscrypt_cleartext++; 1444 return 0; 1445 } 1446 worker->stats.num_query_dnscrypt_cert++; 1447 sldns_buffer_rewind(c->buffer); 1448 } else if(c->dnscrypt && repinfo->is_dnscrypted) { 1449 worker->stats.num_query_dnscrypt_crypted++; 1450 } 1451 #endif 1452 #ifdef USE_DNSTAP 1453 /* 1454 * sending src (client)/dst (local service) addresses over DNSTAP from incoming request handler 1455 */ 1456 if(worker->dtenv.log_client_query_messages) { 1457 log_addr(VERB_ALGO, "request from client", &repinfo->client_addr, repinfo->client_addrlen); 1458 log_addr(VERB_ALGO, "to local addr", (void*)repinfo->c->socket->addr->ai_addr, repinfo->c->socket->addr->ai_addrlen); 1459 dt_msg_send_client_query(&worker->dtenv, &repinfo->client_addr, (void*)repinfo->c->socket->addr->ai_addr, c->type, c->buffer, 1460 ((worker->env.cfg->sock_queue_timeout && timeval_isset(&c->recv_tv))?&c->recv_tv:NULL)); 1461 } 1462 #endif 1463 /* Check deny/refuse ACLs */ 1464 if(repinfo->is_proxied) { 1465 if((ret=deny_refuse_all(c, &acl, worker, repinfo, &acladdr, 1466 worker->env.cfg->ede, 1, &check_result)) != -1) { 1467 if(ret == 1) 1468 goto send_reply; 1469 return ret; 1470 } 1471 } 1472 if((ret=deny_refuse_all(c, &acl, worker, repinfo, &acladdr, 1473 worker->env.cfg->ede, 0, &check_result)) != -1) { 1474 if(ret == 1) 1475 goto send_reply; 1476 return ret; 1477 } 1478 1479 worker_check_request(c->buffer, worker, &check_result); 1480 if(check_result.value != 0) { 1481 verbose(VERB_ALGO, "worker check request: bad query."); 1482 log_addr(VERB_CLIENT,"from",&repinfo->client_addr, repinfo->client_addrlen); 1483 if(check_result.value != -1) { 1484 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1485 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1486 check_result.value); 1487 return 1; 1488 } 1489 comm_point_drop_reply(repinfo); 1490 return 0; 1491 } 1492 1493 worker->stats.num_queries++; 1494 pre_edns_ip_ratelimit = pre_edns_ip_ratelimit_check(acl); 1495 1496 /* If the IP rate limiting check needs extra EDNS information (e.g., 1497 * DNS Cookies) postpone the check until after EDNS is parsed. */ 1498 if(pre_edns_ip_ratelimit) { 1499 /* NOTE: we always check the repinfo->client_address. 1500 * IP ratelimiting is implicitly disabled for proxies. */ 1501 if(!check_ip_ratelimit(worker, &repinfo->client_addr, 1502 repinfo->client_addrlen, 0, c->buffer)) { 1503 worker->stats.num_queries_ip_ratelimited++; 1504 comm_point_drop_reply(repinfo); 1505 return 0; 1506 } 1507 } 1508 1509 if(!query_info_parse(&qinfo, c->buffer)) { 1510 verbose(VERB_ALGO, "worker parse request: formerror."); 1511 log_addr(VERB_CLIENT, "from", &repinfo->client_addr, 1512 repinfo->client_addrlen); 1513 memset(&qinfo, 0, sizeof(qinfo)); /* zero qinfo.qname */ 1514 if(worker_err_ratelimit(worker, LDNS_RCODE_FORMERR) == -1) { 1515 comm_point_drop_reply(repinfo); 1516 return 0; 1517 } 1518 sldns_buffer_rewind(c->buffer); 1519 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1520 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1521 LDNS_RCODE_FORMERR); 1522 goto send_reply; 1523 } 1524 if(worker->env.cfg->log_queries) { 1525 char ip[128]; 1526 addr_to_str(&repinfo->client_addr, repinfo->client_addrlen, ip, sizeof(ip)); 1527 log_query_in(ip, qinfo.qname, qinfo.qtype, qinfo.qclass); 1528 } 1529 if(qinfo.qtype == LDNS_RR_TYPE_AXFR || 1530 qinfo.qtype == LDNS_RR_TYPE_IXFR) { 1531 verbose(VERB_ALGO, "worker request: refused zone transfer."); 1532 log_addr(VERB_CLIENT, "from", &repinfo->client_addr, 1533 repinfo->client_addrlen); 1534 sldns_buffer_rewind(c->buffer); 1535 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1536 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1537 LDNS_RCODE_REFUSED); 1538 if(worker->stats.extended) { 1539 worker->stats.qtype[qinfo.qtype]++; 1540 } 1541 goto send_reply; 1542 } 1543 if(qinfo.qtype == LDNS_RR_TYPE_OPT || 1544 qinfo.qtype == LDNS_RR_TYPE_TSIG || 1545 qinfo.qtype == LDNS_RR_TYPE_TKEY || 1546 qinfo.qtype == LDNS_RR_TYPE_MAILA || 1547 qinfo.qtype == LDNS_RR_TYPE_MAILB || 1548 (qinfo.qtype >= 128 && qinfo.qtype <= 248)) { 1549 verbose(VERB_ALGO, "worker request: formerror for meta-type."); 1550 log_addr(VERB_CLIENT, "from", &repinfo->client_addr, 1551 repinfo->client_addrlen); 1552 if(worker_err_ratelimit(worker, LDNS_RCODE_FORMERR) == -1) { 1553 comm_point_drop_reply(repinfo); 1554 return 0; 1555 } 1556 sldns_buffer_rewind(c->buffer); 1557 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1558 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1559 LDNS_RCODE_FORMERR); 1560 if(worker->stats.extended) { 1561 worker->stats.qtype[qinfo.qtype]++; 1562 } 1563 goto send_reply; 1564 } 1565 if((ret=parse_edns_from_query_pkt( 1566 c->buffer, &edns, worker->env.cfg, c, repinfo, 1567 (worker->env.now ? *worker->env.now : time(NULL)), 1568 worker->scratchpad)) != 0) { 1569 struct edns_data reply_edns; 1570 verbose(VERB_ALGO, "worker parse edns: formerror."); 1571 log_addr(VERB_CLIENT, "from", &repinfo->client_addr, 1572 repinfo->client_addrlen); 1573 memset(&reply_edns, 0, sizeof(reply_edns)); 1574 reply_edns.edns_present = 1; 1575 error_encode(c->buffer, ret, &qinfo, 1576 *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 1577 sldns_buffer_read_u16_at(c->buffer, 2), &reply_edns); 1578 regional_free_all(worker->scratchpad); 1579 goto send_reply; 1580 } 1581 if(edns.edns_present) { 1582 if(edns.edns_version != 0) { 1583 edns.opt_list_in = NULL; 1584 edns.opt_list_out = NULL; 1585 edns.opt_list_inplace_cb_out = NULL; 1586 verbose(VERB_ALGO, "query with bad edns version."); 1587 log_addr(VERB_CLIENT, "from", &repinfo->client_addr, 1588 repinfo->client_addrlen); 1589 extended_error_encode(c->buffer, EDNS_RCODE_BADVERS, &qinfo, 1590 *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 1591 sldns_buffer_read_u16_at(c->buffer, 2), 0, &edns); 1592 regional_free_all(worker->scratchpad); 1593 goto send_reply; 1594 } 1595 if(edns.udp_size < NORMAL_UDP_SIZE && 1596 worker->daemon->cfg->harden_short_bufsize) { 1597 verbose(VERB_QUERY, "worker request: EDNS bufsize %d ignored", 1598 (int)edns.udp_size); 1599 log_addr(VERB_CLIENT, "from", &repinfo->client_addr, 1600 repinfo->client_addrlen); 1601 edns.udp_size = NORMAL_UDP_SIZE; 1602 } 1603 } 1604 1605 /* Get stats for cookies */ 1606 server_stats_downstream_cookie(&worker->stats, &edns); 1607 1608 /* If the IP rate limiting check was postponed, check now. */ 1609 if(!pre_edns_ip_ratelimit) { 1610 /* NOTE: we always check the repinfo->client_address. 1611 * IP ratelimiting is implicitly disabled for proxies. */ 1612 if(!check_ip_ratelimit(worker, &repinfo->client_addr, 1613 repinfo->client_addrlen, edns.cookie_valid, 1614 c->buffer)) { 1615 worker->stats.num_queries_ip_ratelimited++; 1616 comm_point_drop_reply(repinfo); 1617 return 0; 1618 } 1619 } 1620 1621 /* "if, else if" sequence below deals with downstream DNS Cookies */ 1622 if(acl != acl_allow_cookie) 1623 ; /* pass; No cookie downstream processing whatsoever */ 1624 1625 else if(edns.cookie_valid) 1626 ; /* pass; Valid cookie is good! */ 1627 1628 else if(c->type != comm_udp) 1629 ; /* pass; Stateful transport */ 1630 1631 else if(edns.cookie_present) { 1632 /* Cookie present, but not valid: Cookie was bad! */ 1633 extended_error_encode(c->buffer, 1634 LDNS_EXT_RCODE_BADCOOKIE, &qinfo, 1635 *(uint16_t*)(void *) 1636 sldns_buffer_begin(c->buffer), 1637 sldns_buffer_read_u16_at(c->buffer, 2), 1638 0, &edns); 1639 regional_free_all(worker->scratchpad); 1640 goto send_reply; 1641 } else { 1642 /* Cookie required, but no cookie present on UDP */ 1643 verbose(VERB_ALGO, "worker request: " 1644 "need cookie or stateful transport"); 1645 log_addr(VERB_ALGO, "from",&repinfo->remote_addr 1646 , repinfo->remote_addrlen); 1647 EDNS_OPT_LIST_APPEND_EDE(&edns.opt_list_out, 1648 worker->scratchpad, LDNS_EDE_OTHER, 1649 "DNS Cookie needed for UDP replies"); 1650 error_encode(c->buffer, 1651 (LDNS_RCODE_REFUSED|BIT_TC), &qinfo, 1652 *(uint16_t*)(void *) 1653 sldns_buffer_begin(c->buffer), 1654 sldns_buffer_read_u16_at(c->buffer, 2), 1655 &edns); 1656 regional_free_all(worker->scratchpad); 1657 goto send_reply; 1658 } 1659 1660 if(edns.udp_size > worker->daemon->cfg->max_udp_size && 1661 c->type == comm_udp) { 1662 verbose(VERB_QUERY, 1663 "worker request: max UDP reply size modified" 1664 " (%d to max-udp-size)", (int)edns.udp_size); 1665 log_addr(VERB_CLIENT, "from", &repinfo->client_addr, 1666 repinfo->client_addrlen); 1667 edns.udp_size = worker->daemon->cfg->max_udp_size; 1668 } 1669 if(edns.udp_size < LDNS_HEADER_SIZE) { 1670 verbose(VERB_ALGO, "worker request: edns is too small."); 1671 log_addr(VERB_CLIENT, "from", &repinfo->client_addr, 1672 repinfo->client_addrlen); 1673 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1674 LDNS_TC_SET(sldns_buffer_begin(c->buffer)); 1675 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1676 LDNS_RCODE_SERVFAIL); 1677 sldns_buffer_set_position(c->buffer, LDNS_HEADER_SIZE); 1678 sldns_buffer_write_at(c->buffer, 4, 1679 (uint8_t*)"\0\0\0\0\0\0\0\0", 8); 1680 sldns_buffer_flip(c->buffer); 1681 regional_free_all(worker->scratchpad); 1682 goto send_reply; 1683 } 1684 if(worker->stats.extended) 1685 server_stats_insquery(&worker->stats, c, qinfo.qtype, 1686 qinfo.qclass, &edns, repinfo); 1687 if(c->type != comm_udp) 1688 edns.udp_size = 65535; /* max size for TCP replies */ 1689 if(qinfo.qclass == LDNS_RR_CLASS_CH && answer_chaos(worker, &qinfo, 1690 &edns, repinfo, c->buffer)) { 1691 regional_free_all(worker->scratchpad); 1692 goto send_reply; 1693 } 1694 if(LDNS_OPCODE_WIRE(sldns_buffer_begin(c->buffer)) == 1695 LDNS_PACKET_NOTIFY) { 1696 answer_notify(worker, &qinfo, &edns, c->buffer, 1697 &repinfo->client_addr, repinfo->client_addrlen); 1698 regional_free_all(worker->scratchpad); 1699 goto send_reply; 1700 } 1701 if(local_zones_answer(worker->daemon->local_zones, &worker->env, &qinfo, 1702 &edns, c->buffer, worker->scratchpad, repinfo, acladdr->taglist, 1703 acladdr->taglen, acladdr->tag_actions, 1704 acladdr->tag_actions_size, acladdr->tag_datas, 1705 acladdr->tag_datas_size, worker->daemon->cfg->tagname, 1706 worker->daemon->cfg->num_tags, acladdr->view)) { 1707 regional_free_all(worker->scratchpad); 1708 if(sldns_buffer_limit(c->buffer) == 0) { 1709 comm_point_drop_reply(repinfo); 1710 return 0; 1711 } 1712 goto send_reply; 1713 } 1714 if(worker->env.auth_zones && 1715 rpz_callback_from_worker_request(worker->env.auth_zones, 1716 &worker->env, &qinfo, &edns, c->buffer, worker->scratchpad, 1717 repinfo, acladdr->taglist, acladdr->taglen, &worker->stats, 1718 &rpz_passthru)) { 1719 regional_free_all(worker->scratchpad); 1720 if(sldns_buffer_limit(c->buffer) == 0) { 1721 comm_point_drop_reply(repinfo); 1722 return 0; 1723 } 1724 goto send_reply; 1725 } 1726 if(worker->env.auth_zones && 1727 auth_zones_answer(worker->env.auth_zones, &worker->env, 1728 &qinfo, &edns, repinfo, c->buffer, worker->scratchpad)) { 1729 regional_free_all(worker->scratchpad); 1730 if(sldns_buffer_limit(c->buffer) == 0) { 1731 comm_point_drop_reply(repinfo); 1732 return 0; 1733 } 1734 /* set RA for everyone that can have recursion (based on 1735 * access control list) */ 1736 if(LDNS_RD_WIRE(sldns_buffer_begin(c->buffer)) && 1737 acl != acl_deny_non_local && acl != acl_refuse_non_local) 1738 LDNS_RA_SET(sldns_buffer_begin(c->buffer)); 1739 goto send_reply; 1740 } 1741 1742 /* We've looked in our local zones. If the answer isn't there, we 1743 * might need to bail out based on ACLs now. */ 1744 if((ret=deny_refuse_non_local(c, acl, worker, repinfo, acladdr, 1745 worker->env.cfg->ede, &check_result)) != -1) 1746 { 1747 regional_free_all(worker->scratchpad); 1748 if(ret == 1) 1749 goto send_reply; 1750 return ret; 1751 } 1752 1753 /* If this request does not have the recursion bit set, verify 1754 * ACLs allow the recursion bit to be treated as set. */ 1755 if(!(LDNS_RD_WIRE(sldns_buffer_begin(c->buffer))) && 1756 acl == acl_allow_setrd ) { 1757 LDNS_RD_SET(sldns_buffer_begin(c->buffer)); 1758 } 1759 1760 /* If this request does not have the recursion bit set, verify 1761 * ACLs allow the snooping. */ 1762 if(!(LDNS_RD_WIRE(sldns_buffer_begin(c->buffer))) && 1763 acl != acl_allow_snoop ) { 1764 if(worker->env.cfg->ede) { 1765 EDNS_OPT_LIST_APPEND_EDE(&edns.opt_list_out, 1766 worker->scratchpad, LDNS_EDE_NOT_AUTHORITATIVE, ""); 1767 } 1768 error_encode(c->buffer, LDNS_RCODE_REFUSED, &qinfo, 1769 *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 1770 sldns_buffer_read_u16_at(c->buffer, 2), &edns); 1771 regional_free_all(worker->scratchpad); 1772 log_addr(VERB_ALGO, "refused nonrec (cache snoop) query from", 1773 &repinfo->client_addr, repinfo->client_addrlen); 1774 1775 goto send_reply; 1776 } 1777 1778 /* If we've found a local alias, replace the qname with the alias 1779 * target before resolving it. */ 1780 if(qinfo.local_alias) { 1781 struct ub_packed_rrset_key* rrset = qinfo.local_alias->rrset; 1782 struct packed_rrset_data* d = rrset->entry.data; 1783 1784 /* Sanity check: our current implementation only supports 1785 * a single CNAME RRset as a local alias. */ 1786 if(qinfo.local_alias->next || 1787 rrset->rk.type != htons(LDNS_RR_TYPE_CNAME) || 1788 d->count != 1) { 1789 log_err("assumption failure: unexpected local alias"); 1790 regional_free_all(worker->scratchpad); 1791 return 0; /* drop it */ 1792 } 1793 qinfo.qname = d->rr_data[0] + 2; 1794 qinfo.qname_len = d->rr_len[0] - 2; 1795 } 1796 1797 /* If we may apply IP-based actions to the answer, build the client 1798 * information. As this can be expensive, skip it if there is 1799 * absolutely no possibility of it. */ 1800 if((worker->daemon->use_response_ip || worker->daemon->use_rpz) && 1801 (qinfo.qtype == LDNS_RR_TYPE_A || 1802 qinfo.qtype == LDNS_RR_TYPE_AAAA || 1803 qinfo.qtype == LDNS_RR_TYPE_ANY)) { 1804 cinfo_tmp.taglist = acladdr->taglist; 1805 cinfo_tmp.taglen = acladdr->taglen; 1806 cinfo_tmp.tag_actions = acladdr->tag_actions; 1807 cinfo_tmp.tag_actions_size = acladdr->tag_actions_size; 1808 cinfo_tmp.tag_datas = acladdr->tag_datas; 1809 cinfo_tmp.tag_datas_size = acladdr->tag_datas_size; 1810 cinfo_tmp.view = acladdr->view; 1811 cinfo_tmp.respip_set = worker->daemon->respip_set; 1812 cinfo = &cinfo_tmp; 1813 } 1814 1815 /* Keep the original edns list around. The pointer could change if there is 1816 * a cached answer (through the inplace callback function there). 1817 * No need to actually copy the contents as they shouldn't change. 1818 * Used while prefetching and subnet is enabled. */ 1819 original_edns_list = edns.opt_list_in; 1820 lookup_cache: 1821 /* Lookup the cache. In case we chase an intermediate CNAME chain 1822 * this is a two-pass operation, and lookup_qinfo is different for 1823 * each pass. We should still pass the original qinfo to 1824 * answer_from_cache(), however, since it's used to build the reply. */ 1825 if(!edns_bypass_cache_stage(edns.opt_list_in, &worker->env)) { 1826 is_expired_answer = 0; 1827 is_secure_answer = 0; 1828 h = query_info_hash(lookup_qinfo, sldns_buffer_read_u16_at(c->buffer, 2)); 1829 if((e=slabhash_lookup(worker->env.msg_cache, h, lookup_qinfo, 0))) { 1830 struct reply_info* rep = (struct reply_info*)e->data; 1831 /* answer from cache - we have acquired a readlock on it */ 1832 if(answer_from_cache(worker, &qinfo, cinfo, &need_drop, 1833 &is_expired_answer, &is_secure_answer, 1834 &alias_rrset, &partial_rep, rep, 1835 *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 1836 sldns_buffer_read_u16_at(c->buffer, 2), repinfo, 1837 &edns)) { 1838 /* prefetch it if the prefetch TTL expired. 1839 * Note that if there is more than one pass 1840 * its qname must be that used for cache 1841 * lookup. */ 1842 if((worker->env.cfg->prefetch && 1843 *worker->env.now >= rep->prefetch_ttl) || 1844 (worker->env.cfg->serve_expired && 1845 *worker->env.now > rep->ttl)) { 1846 1847 time_t leeway = rep->ttl - *worker->env.now; 1848 if(rep->ttl < *worker->env.now) 1849 leeway = 0; 1850 lock_rw_unlock(&e->lock); 1851 1852 reply_and_prefetch(worker, lookup_qinfo, 1853 sldns_buffer_read_u16_at(c->buffer, 2), 1854 repinfo, leeway, 1855 (partial_rep || need_drop), 1856 rpz_passthru, 1857 original_edns_list); 1858 if(!partial_rep) { 1859 rc = 0; 1860 regional_free_all(worker->scratchpad); 1861 goto send_reply_rc; 1862 } 1863 } else if(!partial_rep) { 1864 lock_rw_unlock(&e->lock); 1865 regional_free_all(worker->scratchpad); 1866 goto send_reply; 1867 } else { 1868 /* Note that we've already released the 1869 * lock if we're here after prefetch. */ 1870 lock_rw_unlock(&e->lock); 1871 } 1872 /* We've found a partial reply ending with an 1873 * alias. Replace the lookup qinfo for the 1874 * alias target and lookup the cache again to 1875 * (possibly) complete the reply. As we're 1876 * passing the "base" reply, there will be no 1877 * more alias chasing. */ 1878 memset(&qinfo_tmp, 0, sizeof(qinfo_tmp)); 1879 get_cname_target(alias_rrset, &qinfo_tmp.qname, 1880 &qinfo_tmp.qname_len); 1881 if(!qinfo_tmp.qname) { 1882 log_err("unexpected: invalid answer alias"); 1883 regional_free_all(worker->scratchpad); 1884 return 0; /* drop query */ 1885 } 1886 qinfo_tmp.qtype = qinfo.qtype; 1887 qinfo_tmp.qclass = qinfo.qclass; 1888 lookup_qinfo = &qinfo_tmp; 1889 goto lookup_cache; 1890 } 1891 verbose(VERB_ALGO, "answer from the cache failed"); 1892 lock_rw_unlock(&e->lock); 1893 } 1894 1895 if(!LDNS_RD_WIRE(sldns_buffer_begin(c->buffer))) { 1896 if(answer_norec_from_cache(worker, &qinfo, 1897 *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 1898 sldns_buffer_read_u16_at(c->buffer, 2), repinfo, 1899 &edns)) { 1900 regional_free_all(worker->scratchpad); 1901 goto send_reply; 1902 } 1903 verbose(VERB_ALGO, "answer norec from cache -- " 1904 "need to validate or not primed"); 1905 } 1906 } 1907 sldns_buffer_rewind(c->buffer); 1908 server_stats_querymiss(&worker->stats, worker); 1909 1910 if(verbosity >= VERB_CLIENT) { 1911 if(c->type == comm_udp) 1912 log_addr(VERB_CLIENT, "udp request from", 1913 &repinfo->client_addr, repinfo->client_addrlen); 1914 else log_addr(VERB_CLIENT, "tcp request from", 1915 &repinfo->client_addr, repinfo->client_addrlen); 1916 } 1917 1918 /* grab a work request structure for this new request */ 1919 mesh_new_client(worker->env.mesh, &qinfo, cinfo, 1920 sldns_buffer_read_u16_at(c->buffer, 2), 1921 &edns, repinfo, *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 1922 rpz_passthru); 1923 regional_free_all(worker->scratchpad); 1924 worker_mem_report(worker, NULL); 1925 return 0; 1926 1927 send_reply: 1928 rc = 1; 1929 send_reply_rc: 1930 if(need_drop) { 1931 comm_point_drop_reply(repinfo); 1932 return 0; 1933 } 1934 if(is_expired_answer) { 1935 worker->stats.ans_expired++; 1936 } 1937 server_stats_insrcode(&worker->stats, c->buffer); 1938 if(worker->stats.extended) { 1939 if(is_secure_answer) worker->stats.ans_secure++; 1940 } 1941 #ifdef USE_DNSTAP 1942 /* 1943 * sending src (client)/dst (local service) addresses over DNSTAP from send_reply code label (when we serviced local zone for ex.) 1944 */ 1945 if(worker->dtenv.log_client_response_messages) { 1946 log_addr(VERB_ALGO, "from local addr", (void*)repinfo->c->socket->addr->ai_addr, repinfo->c->socket->addr->ai_addrlen); 1947 log_addr(VERB_ALGO, "response to client", &repinfo->client_addr, repinfo->client_addrlen); 1948 dt_msg_send_client_response(&worker->dtenv, &repinfo->client_addr, (void*)repinfo->c->socket->addr->ai_addr, c->type, c->buffer); 1949 } 1950 #endif 1951 if(worker->env.cfg->log_replies) 1952 { 1953 struct timeval tv; 1954 memset(&tv, 0, sizeof(tv)); 1955 if(qinfo.local_alias && qinfo.local_alias->rrset && 1956 qinfo.local_alias->rrset->rk.dname) { 1957 /* log original qname, before the local alias was 1958 * used to resolve that CNAME to something else */ 1959 qinfo.qname = qinfo.local_alias->rrset->rk.dname; 1960 log_reply_info(NO_VERBOSE, &qinfo, 1961 &repinfo->client_addr, repinfo->client_addrlen, 1962 tv, 1, c->buffer); 1963 } else { 1964 log_reply_info(NO_VERBOSE, &qinfo, 1965 &repinfo->client_addr, repinfo->client_addrlen, 1966 tv, 1, c->buffer); 1967 } 1968 } 1969 #ifdef USE_DNSCRYPT 1970 if(!dnsc_handle_uncurved_request(repinfo)) { 1971 return 0; 1972 } 1973 #endif 1974 return rc; 1975 } 1976 1977 void 1978 worker_sighandler(int sig, void* arg) 1979 { 1980 /* note that log, print, syscalls here give race conditions. 1981 * And cause hangups if the log-lock is held by the application. */ 1982 struct worker* worker = (struct worker*)arg; 1983 switch(sig) { 1984 #ifdef SIGHUP 1985 case SIGHUP: 1986 comm_base_exit(worker->base); 1987 break; 1988 #endif 1989 #ifdef SIGBREAK 1990 case SIGBREAK: 1991 #endif 1992 case SIGINT: 1993 worker->need_to_exit = 1; 1994 comm_base_exit(worker->base); 1995 break; 1996 #ifdef SIGQUIT 1997 case SIGQUIT: 1998 worker->need_to_exit = 1; 1999 comm_base_exit(worker->base); 2000 break; 2001 #endif 2002 case SIGTERM: 2003 worker->need_to_exit = 1; 2004 comm_base_exit(worker->base); 2005 break; 2006 default: 2007 /* unknown signal, ignored */ 2008 break; 2009 } 2010 } 2011 2012 /** restart statistics timer for worker, if enabled */ 2013 static void 2014 worker_restart_timer(struct worker* worker) 2015 { 2016 if(worker->env.cfg->stat_interval > 0) { 2017 struct timeval tv; 2018 #ifndef S_SPLINT_S 2019 tv.tv_sec = worker->env.cfg->stat_interval; 2020 tv.tv_usec = 0; 2021 #endif 2022 comm_timer_set(worker->stat_timer, &tv); 2023 } 2024 } 2025 2026 void worker_stat_timer_cb(void* arg) 2027 { 2028 struct worker* worker = (struct worker*)arg; 2029 server_stats_log(&worker->stats, worker, worker->thread_num); 2030 mesh_stats(worker->env.mesh, "mesh has"); 2031 worker_mem_report(worker, NULL); 2032 /* SHM is enabled, process data to SHM */ 2033 if (worker->daemon->cfg->shm_enable) { 2034 shm_main_run(worker); 2035 } 2036 if(!worker->daemon->cfg->stat_cumulative) { 2037 worker_stats_clear(worker); 2038 } 2039 /* start next timer */ 2040 worker_restart_timer(worker); 2041 } 2042 2043 void worker_probe_timer_cb(void* arg) 2044 { 2045 struct worker* worker = (struct worker*)arg; 2046 struct timeval tv; 2047 #ifndef S_SPLINT_S 2048 tv.tv_sec = (time_t)autr_probe_timer(&worker->env); 2049 tv.tv_usec = 0; 2050 #endif 2051 if(tv.tv_sec != 0) 2052 comm_timer_set(worker->env.probe_timer, &tv); 2053 } 2054 2055 struct worker* 2056 worker_create(struct daemon* daemon, int id, int* ports, int n) 2057 { 2058 unsigned int seed; 2059 struct worker* worker = (struct worker*)calloc(1, 2060 sizeof(struct worker)); 2061 if(!worker) 2062 return NULL; 2063 worker->numports = n; 2064 worker->ports = (int*)memdup(ports, sizeof(int)*n); 2065 if(!worker->ports) { 2066 free(worker); 2067 return NULL; 2068 } 2069 worker->daemon = daemon; 2070 worker->thread_num = id; 2071 if(!(worker->cmd = tube_create())) { 2072 free(worker->ports); 2073 free(worker); 2074 return NULL; 2075 } 2076 /* create random state here to avoid locking trouble in RAND_bytes */ 2077 if(!(worker->rndstate = ub_initstate(daemon->rand))) { 2078 log_err("could not init random numbers."); 2079 tube_delete(worker->cmd); 2080 free(worker->ports); 2081 free(worker); 2082 return NULL; 2083 } 2084 explicit_bzero(&seed, sizeof(seed)); 2085 return worker; 2086 } 2087 2088 int 2089 worker_init(struct worker* worker, struct config_file *cfg, 2090 struct listen_port* ports, int do_sigs) 2091 { 2092 #ifdef USE_DNSTAP 2093 struct dt_env* dtenv = &worker->dtenv; 2094 #else 2095 void* dtenv = NULL; 2096 #endif 2097 #ifdef HAVE_GETTID 2098 worker->thread_tid = gettid(); 2099 #endif 2100 worker->need_to_exit = 0; 2101 worker->base = comm_base_create(do_sigs); 2102 if(!worker->base) { 2103 log_err("could not create event handling base"); 2104 worker_delete(worker); 2105 return 0; 2106 } 2107 comm_base_set_slow_accept_handlers(worker->base, &worker_stop_accept, 2108 &worker_start_accept, worker); 2109 if(do_sigs) { 2110 #ifdef SIGHUP 2111 ub_thread_sig_unblock(SIGHUP); 2112 #endif 2113 #ifdef SIGBREAK 2114 ub_thread_sig_unblock(SIGBREAK); 2115 #endif 2116 ub_thread_sig_unblock(SIGINT); 2117 #ifdef SIGQUIT 2118 ub_thread_sig_unblock(SIGQUIT); 2119 #endif 2120 ub_thread_sig_unblock(SIGTERM); 2121 #ifndef LIBEVENT_SIGNAL_PROBLEM 2122 worker->comsig = comm_signal_create(worker->base, 2123 worker_sighandler, worker); 2124 if(!worker->comsig 2125 #ifdef SIGHUP 2126 || !comm_signal_bind(worker->comsig, SIGHUP) 2127 #endif 2128 #ifdef SIGQUIT 2129 || !comm_signal_bind(worker->comsig, SIGQUIT) 2130 #endif 2131 || !comm_signal_bind(worker->comsig, SIGTERM) 2132 #ifdef SIGBREAK 2133 || !comm_signal_bind(worker->comsig, SIGBREAK) 2134 #endif 2135 || !comm_signal_bind(worker->comsig, SIGINT)) { 2136 log_err("could not create signal handlers"); 2137 worker_delete(worker); 2138 return 0; 2139 } 2140 #endif /* LIBEVENT_SIGNAL_PROBLEM */ 2141 if(!daemon_remote_open_accept(worker->daemon->rc, 2142 worker->daemon->rc_ports, worker)) { 2143 worker_delete(worker); 2144 return 0; 2145 } 2146 #ifdef UB_ON_WINDOWS 2147 wsvc_setup_worker(worker); 2148 #endif /* UB_ON_WINDOWS */ 2149 } else { /* !do_sigs */ 2150 worker->comsig = NULL; 2151 } 2152 #ifdef USE_DNSTAP 2153 if(cfg->dnstap) { 2154 log_assert(worker->daemon->dtenv != NULL); 2155 memcpy(&worker->dtenv, worker->daemon->dtenv, sizeof(struct dt_env)); 2156 if(!dt_init(&worker->dtenv, worker->base)) 2157 fatal_exit("dt_init failed"); 2158 } 2159 #endif 2160 worker->front = listen_create(worker->base, ports, 2161 cfg->msg_buffer_size, (int)cfg->incoming_num_tcp, 2162 cfg->do_tcp_keepalive 2163 ? cfg->tcp_keepalive_timeout 2164 : cfg->tcp_idle_timeout, 2165 cfg->harden_large_queries, cfg->http_max_streams, 2166 cfg->http_endpoint, cfg->http_notls_downstream, 2167 worker->daemon->tcl, worker->daemon->listen_sslctx, 2168 dtenv, worker_handle_request, worker); 2169 if(!worker->front) { 2170 log_err("could not create listening sockets"); 2171 worker_delete(worker); 2172 return 0; 2173 } 2174 worker->back = outside_network_create(worker->base, 2175 cfg->msg_buffer_size, (size_t)cfg->outgoing_num_ports, 2176 cfg->out_ifs, cfg->num_out_ifs, cfg->do_ip4, cfg->do_ip6, 2177 cfg->do_tcp?cfg->outgoing_num_tcp:0, cfg->ip_dscp, 2178 worker->daemon->env->infra_cache, worker->rndstate, 2179 cfg->use_caps_bits_for_id, worker->ports, worker->numports, 2180 cfg->unwanted_threshold, cfg->outgoing_tcp_mss, 2181 &worker_alloc_cleanup, worker, 2182 cfg->do_udp || cfg->udp_upstream_without_downstream, 2183 worker->daemon->connect_sslctx, cfg->delay_close, 2184 cfg->tls_use_sni, dtenv, cfg->udp_connect, 2185 cfg->max_reuse_tcp_queries, cfg->tcp_reuse_timeout, 2186 cfg->tcp_auth_query_timeout); 2187 if(!worker->back) { 2188 log_err("could not create outgoing sockets"); 2189 worker_delete(worker); 2190 return 0; 2191 } 2192 iterator_set_ip46_support(&worker->daemon->mods, worker->daemon->env, 2193 worker->back); 2194 /* start listening to commands */ 2195 if(!tube_setup_bg_listen(worker->cmd, worker->base, 2196 &worker_handle_control_cmd, worker)) { 2197 log_err("could not create control compt."); 2198 worker_delete(worker); 2199 return 0; 2200 } 2201 worker->stat_timer = comm_timer_create(worker->base, 2202 worker_stat_timer_cb, worker); 2203 if(!worker->stat_timer) { 2204 log_err("could not create statistics timer"); 2205 } 2206 2207 /* we use the msg_buffer_size as a good estimate for what the 2208 * user wants for memory usage sizes */ 2209 worker->scratchpad = regional_create_custom(cfg->msg_buffer_size); 2210 if(!worker->scratchpad) { 2211 log_err("malloc failure"); 2212 worker_delete(worker); 2213 return 0; 2214 } 2215 2216 server_stats_init(&worker->stats, cfg); 2217 worker->alloc = worker->daemon->worker_allocs[worker->thread_num]; 2218 alloc_set_id_cleanup(worker->alloc, &worker_alloc_cleanup, worker); 2219 worker->env = *worker->daemon->env; 2220 comm_base_timept(worker->base, &worker->env.now, &worker->env.now_tv); 2221 worker->env.worker = worker; 2222 worker->env.worker_base = worker->base; 2223 worker->env.send_query = &worker_send_query; 2224 worker->env.alloc = worker->alloc; 2225 worker->env.outnet = worker->back; 2226 worker->env.rnd = worker->rndstate; 2227 /* If case prefetch is triggered, the corresponding mesh will clear 2228 * the scratchpad for the module env in the middle of request handling. 2229 * It would be prone to a use-after-free kind of bug, so we avoid 2230 * sharing it with worker's own scratchpad at the cost of having 2231 * one more pad per worker. */ 2232 worker->env.scratch = regional_create_custom(cfg->msg_buffer_size); 2233 if(!worker->env.scratch) { 2234 log_err("malloc failure"); 2235 worker_delete(worker); 2236 return 0; 2237 } 2238 worker->env.mesh = mesh_create(&worker->daemon->mods, &worker->env); 2239 if(!worker->env.mesh) { 2240 log_err("malloc failure"); 2241 worker_delete(worker); 2242 return 0; 2243 } 2244 /* Pass on daemon variables that we would need in the mesh area */ 2245 worker->env.mesh->use_response_ip = worker->daemon->use_response_ip; 2246 worker->env.mesh->use_rpz = worker->daemon->use_rpz; 2247 2248 worker->env.detach_subs = &mesh_detach_subs; 2249 worker->env.attach_sub = &mesh_attach_sub; 2250 worker->env.add_sub = &mesh_add_sub; 2251 worker->env.kill_sub = &mesh_state_delete; 2252 worker->env.detect_cycle = &mesh_detect_cycle; 2253 worker->env.scratch_buffer = sldns_buffer_new(cfg->msg_buffer_size); 2254 if(!worker->env.scratch_buffer) { 2255 log_err("malloc failure"); 2256 worker_delete(worker); 2257 return 0; 2258 } 2259 if(!(worker->env.fwds = forwards_create()) || 2260 !forwards_apply_cfg(worker->env.fwds, cfg)) { 2261 log_err("Could not set forward zones"); 2262 worker_delete(worker); 2263 return 0; 2264 } 2265 if(!(worker->env.hints = hints_create()) || 2266 !hints_apply_cfg(worker->env.hints, cfg)) { 2267 log_err("Could not set root or stub hints"); 2268 worker_delete(worker); 2269 return 0; 2270 } 2271 /* one probe timer per process -- if we have 5011 anchors */ 2272 if(autr_get_num_anchors(worker->env.anchors) > 0 2273 #ifndef THREADS_DISABLED 2274 && worker->thread_num == 0 2275 #endif 2276 ) { 2277 struct timeval tv; 2278 tv.tv_sec = 0; 2279 tv.tv_usec = 0; 2280 worker->env.probe_timer = comm_timer_create(worker->base, 2281 worker_probe_timer_cb, worker); 2282 if(!worker->env.probe_timer) { 2283 log_err("could not create 5011-probe timer"); 2284 } else { 2285 /* let timer fire, then it can reset itself */ 2286 comm_timer_set(worker->env.probe_timer, &tv); 2287 } 2288 } 2289 /* zone transfer tasks, setup once per process, if any */ 2290 if(worker->env.auth_zones 2291 #ifndef THREADS_DISABLED 2292 && worker->thread_num == 0 2293 #endif 2294 ) { 2295 auth_xfer_pickup_initial(worker->env.auth_zones, &worker->env); 2296 auth_zones_pickup_zonemd_verify(worker->env.auth_zones, 2297 &worker->env); 2298 } 2299 #ifdef USE_DNSTAP 2300 if(worker->daemon->cfg->dnstap 2301 #ifndef THREADS_DISABLED 2302 && worker->thread_num == 0 2303 #endif 2304 ) { 2305 if(!dt_io_thread_start(dtenv->dtio, comm_base_internal( 2306 worker->base), worker->daemon->num)) { 2307 log_err("could not start dnstap io thread"); 2308 worker_delete(worker); 2309 return 0; 2310 } 2311 } 2312 #endif /* USE_DNSTAP */ 2313 worker_mem_report(worker, NULL); 2314 /* if statistics enabled start timer */ 2315 if(worker->env.cfg->stat_interval > 0) { 2316 verbose(VERB_ALGO, "set statistics interval %d secs", 2317 worker->env.cfg->stat_interval); 2318 worker_restart_timer(worker); 2319 } 2320 return 1; 2321 } 2322 2323 void 2324 worker_work(struct worker* worker) 2325 { 2326 comm_base_dispatch(worker->base); 2327 } 2328 2329 void 2330 worker_delete(struct worker* worker) 2331 { 2332 if(!worker) 2333 return; 2334 if(worker->env.mesh && verbosity >= VERB_OPS) { 2335 server_stats_log(&worker->stats, worker, worker->thread_num); 2336 mesh_stats(worker->env.mesh, "mesh has"); 2337 worker_mem_report(worker, NULL); 2338 } 2339 outside_network_quit_prepare(worker->back); 2340 mesh_delete(worker->env.mesh); 2341 sldns_buffer_free(worker->env.scratch_buffer); 2342 forwards_delete(worker->env.fwds); 2343 hints_delete(worker->env.hints); 2344 listen_delete(worker->front); 2345 outside_network_delete(worker->back); 2346 comm_signal_delete(worker->comsig); 2347 tube_delete(worker->cmd); 2348 comm_timer_delete(worker->stat_timer); 2349 comm_timer_delete(worker->env.probe_timer); 2350 free(worker->ports); 2351 if(worker->thread_num == 0) { 2352 #ifdef UB_ON_WINDOWS 2353 wsvc_desetup_worker(worker); 2354 #endif /* UB_ON_WINDOWS */ 2355 } 2356 #ifdef USE_DNSTAP 2357 if(worker->daemon->cfg->dnstap 2358 #ifndef THREADS_DISABLED 2359 && worker->thread_num == 0 2360 #endif 2361 ) { 2362 dt_io_thread_stop(worker->dtenv.dtio); 2363 } 2364 dt_deinit(&worker->dtenv); 2365 #endif /* USE_DNSTAP */ 2366 comm_base_delete(worker->base); 2367 ub_randfree(worker->rndstate); 2368 /* don't touch worker->alloc, as it's maintained in daemon */ 2369 regional_destroy(worker->env.scratch); 2370 regional_destroy(worker->scratchpad); 2371 free(worker); 2372 } 2373 2374 struct outbound_entry* 2375 worker_send_query(struct query_info* qinfo, uint16_t flags, int dnssec, 2376 int want_dnssec, int nocaps, int check_ratelimit, 2377 struct sockaddr_storage* addr, socklen_t addrlen, uint8_t* zone, 2378 size_t zonelen, int tcp_upstream, int ssl_upstream, char* tls_auth_name, 2379 struct module_qstate* q, int* was_ratelimited) 2380 { 2381 struct worker* worker = q->env->worker; 2382 struct outbound_entry* e = (struct outbound_entry*)regional_alloc( 2383 q->region, sizeof(*e)); 2384 if(!e) 2385 return NULL; 2386 e->qstate = q; 2387 e->qsent = outnet_serviced_query(worker->back, qinfo, flags, dnssec, 2388 want_dnssec, nocaps, check_ratelimit, tcp_upstream, 2389 ssl_upstream, tls_auth_name, addr, addrlen, zone, zonelen, q, 2390 worker_handle_service_reply, e, worker->back->udp_buff, q->env, 2391 was_ratelimited); 2392 if(!e->qsent) { 2393 return NULL; 2394 } 2395 return e; 2396 } 2397 2398 void 2399 worker_alloc_cleanup(void* arg) 2400 { 2401 struct worker* worker = (struct worker*)arg; 2402 slabhash_clear(&worker->env.rrset_cache->table); 2403 slabhash_clear(worker->env.msg_cache); 2404 } 2405 2406 void worker_stats_clear(struct worker* worker) 2407 { 2408 server_stats_init(&worker->stats, worker->env.cfg); 2409 mesh_stats_clear(worker->env.mesh); 2410 worker->back->unwanted_replies = 0; 2411 worker->back->num_tcp_outgoing = 0; 2412 worker->back->num_udp_outgoing = 0; 2413 } 2414 2415 void worker_start_accept(void* arg) 2416 { 2417 struct worker* worker = (struct worker*)arg; 2418 listen_start_accept(worker->front); 2419 if(worker->thread_num == 0) 2420 daemon_remote_start_accept(worker->daemon->rc); 2421 } 2422 2423 void worker_stop_accept(void* arg) 2424 { 2425 struct worker* worker = (struct worker*)arg; 2426 listen_stop_accept(worker->front); 2427 if(worker->thread_num == 0) 2428 daemon_remote_stop_accept(worker->daemon->rc); 2429 } 2430 2431 /* --- fake callbacks for fptr_wlist to work --- */ 2432 struct outbound_entry* libworker_send_query( 2433 struct query_info* ATTR_UNUSED(qinfo), 2434 uint16_t ATTR_UNUSED(flags), int ATTR_UNUSED(dnssec), 2435 int ATTR_UNUSED(want_dnssec), int ATTR_UNUSED(nocaps), 2436 int ATTR_UNUSED(check_ratelimit), 2437 struct sockaddr_storage* ATTR_UNUSED(addr), socklen_t ATTR_UNUSED(addrlen), 2438 uint8_t* ATTR_UNUSED(zone), size_t ATTR_UNUSED(zonelen), int ATTR_UNUSED(tcp_upstream), 2439 int ATTR_UNUSED(ssl_upstream), char* ATTR_UNUSED(tls_auth_name), 2440 struct module_qstate* ATTR_UNUSED(q), int* ATTR_UNUSED(was_ratelimited)) 2441 { 2442 log_assert(0); 2443 return 0; 2444 } 2445 2446 int libworker_handle_service_reply(struct comm_point* ATTR_UNUSED(c), 2447 void* ATTR_UNUSED(arg), int ATTR_UNUSED(error), 2448 struct comm_reply* ATTR_UNUSED(reply_info)) 2449 { 2450 log_assert(0); 2451 return 0; 2452 } 2453 2454 void libworker_handle_control_cmd(struct tube* ATTR_UNUSED(tube), 2455 uint8_t* ATTR_UNUSED(buffer), size_t ATTR_UNUSED(len), 2456 int ATTR_UNUSED(error), void* ATTR_UNUSED(arg)) 2457 { 2458 log_assert(0); 2459 } 2460 2461 void libworker_fg_done_cb(void* ATTR_UNUSED(arg), int ATTR_UNUSED(rcode), 2462 sldns_buffer* ATTR_UNUSED(buf), enum sec_status ATTR_UNUSED(s), 2463 char* ATTR_UNUSED(why_bogus), int ATTR_UNUSED(was_ratelimited)) 2464 { 2465 log_assert(0); 2466 } 2467 2468 void libworker_bg_done_cb(void* ATTR_UNUSED(arg), int ATTR_UNUSED(rcode), 2469 sldns_buffer* ATTR_UNUSED(buf), enum sec_status ATTR_UNUSED(s), 2470 char* ATTR_UNUSED(why_bogus), int ATTR_UNUSED(was_ratelimited)) 2471 { 2472 log_assert(0); 2473 } 2474 2475 void libworker_event_done_cb(void* ATTR_UNUSED(arg), int ATTR_UNUSED(rcode), 2476 sldns_buffer* ATTR_UNUSED(buf), enum sec_status ATTR_UNUSED(s), 2477 char* ATTR_UNUSED(why_bogus), int ATTR_UNUSED(was_ratelimited)) 2478 { 2479 log_assert(0); 2480 } 2481 2482 int context_query_cmp(const void* ATTR_UNUSED(a), const void* ATTR_UNUSED(b)) 2483 { 2484 log_assert(0); 2485 return 0; 2486 } 2487 2488 int order_lock_cmp(const void* ATTR_UNUSED(e1), const void* ATTR_UNUSED(e2)) 2489 { 2490 log_assert(0); 2491 return 0; 2492 } 2493 2494 int codeline_cmp(const void* ATTR_UNUSED(a), const void* ATTR_UNUSED(b)) 2495 { 2496 log_assert(0); 2497 return 0; 2498 } 2499 2500 #ifdef USE_DNSTAP 2501 void dtio_tap_callback(int ATTR_UNUSED(fd), short ATTR_UNUSED(ev), 2502 void* ATTR_UNUSED(arg)) 2503 { 2504 log_assert(0); 2505 } 2506 #endif 2507 2508 #ifdef USE_DNSTAP 2509 void dtio_mainfdcallback(int ATTR_UNUSED(fd), short ATTR_UNUSED(ev), 2510 void* ATTR_UNUSED(arg)) 2511 { 2512 log_assert(0); 2513 } 2514 #endif 2515