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 "iterator/iter_fwd.h" 72 #include "iterator/iter_hints.h" 73 #include "validator/autotrust.h" 74 #include "validator/val_anchor.h" 75 #include "respip/respip.h" 76 #include "libunbound/context.h" 77 #include "libunbound/libworker.h" 78 #include "sldns/sbuffer.h" 79 #include "sldns/wire2str.h" 80 #include "util/shm_side/shm_main.h" 81 #include "dnscrypt/dnscrypt.h" 82 83 #ifdef HAVE_SYS_TYPES_H 84 # include <sys/types.h> 85 #endif 86 #ifdef HAVE_NETDB_H 87 #include <netdb.h> 88 #endif 89 #include <signal.h> 90 #ifdef UB_ON_WINDOWS 91 #include "winrc/win_svc.h" 92 #endif 93 94 /** Size of an UDP datagram */ 95 #define NORMAL_UDP_SIZE 512 /* bytes */ 96 /** ratelimit for error responses */ 97 #define ERROR_RATELIMIT 100 /* qps */ 98 99 /** 100 * seconds to add to prefetch leeway. This is a TTL that expires old rrsets 101 * earlier than they should in order to put the new update into the cache. 102 * This additional value is to make sure that if not all TTLs are equal in 103 * the message to be updated(and replaced), that rrsets with up to this much 104 * extra TTL are also replaced. This means that the resulting new message 105 * will have (most likely) this TTL at least, avoiding very small 'split 106 * second' TTLs due to operators choosing relative primes for TTLs (or so). 107 * Also has to be at least one to break ties (and overwrite cached entry). 108 */ 109 #define PREFETCH_EXPIRY_ADD 60 110 111 /** Report on memory usage by this thread and global */ 112 static void 113 worker_mem_report(struct worker* ATTR_UNUSED(worker), 114 struct serviced_query* ATTR_UNUSED(cur_serv)) 115 { 116 #ifdef UNBOUND_ALLOC_STATS 117 /* measure memory leakage */ 118 extern size_t unbound_mem_alloc, unbound_mem_freed; 119 /* debug func in validator module */ 120 size_t total, front, back, mesh, msg, rrset, infra, ac, superac; 121 size_t me, iter, val, anch; 122 int i; 123 #ifdef CLIENT_SUBNET 124 size_t subnet = 0; 125 #endif /* CLIENT_SUBNET */ 126 if(verbosity < VERB_ALGO) 127 return; 128 front = listen_get_mem(worker->front); 129 back = outnet_get_mem(worker->back); 130 msg = slabhash_get_mem(worker->env.msg_cache); 131 rrset = slabhash_get_mem(&worker->env.rrset_cache->table); 132 infra = infra_get_mem(worker->env.infra_cache); 133 mesh = mesh_get_mem(worker->env.mesh); 134 ac = alloc_get_mem(&worker->alloc); 135 superac = alloc_get_mem(&worker->daemon->superalloc); 136 anch = anchors_get_mem(worker->env.anchors); 137 iter = 0; 138 val = 0; 139 for(i=0; i<worker->env.mesh->mods.num; i++) { 140 fptr_ok(fptr_whitelist_mod_get_mem(worker->env.mesh-> 141 mods.mod[i]->get_mem)); 142 if(strcmp(worker->env.mesh->mods.mod[i]->name, "validator")==0) 143 val += (*worker->env.mesh->mods.mod[i]->get_mem) 144 (&worker->env, i); 145 #ifdef CLIENT_SUBNET 146 else if(strcmp(worker->env.mesh->mods.mod[i]->name, 147 "subnet")==0) 148 subnet += (*worker->env.mesh->mods.mod[i]->get_mem) 149 (&worker->env, i); 150 #endif /* CLIENT_SUBNET */ 151 else iter += (*worker->env.mesh->mods.mod[i]->get_mem) 152 (&worker->env, i); 153 } 154 me = sizeof(*worker) + sizeof(*worker->base) + sizeof(*worker->comsig) 155 + comm_point_get_mem(worker->cmd_com) 156 + sizeof(worker->rndstate) 157 + regional_get_mem(worker->scratchpad) 158 + sizeof(*worker->env.scratch_buffer) 159 + sldns_buffer_capacity(worker->env.scratch_buffer) 160 + forwards_get_mem(worker->env.fwds) 161 + hints_get_mem(worker->env.hints); 162 if(worker->thread_num == 0) 163 me += acl_list_get_mem(worker->daemon->acl); 164 if(cur_serv) { 165 me += serviced_get_mem(cur_serv); 166 } 167 total = front+back+mesh+msg+rrset+infra+iter+val+ac+superac+me; 168 #ifdef CLIENT_SUBNET 169 total += subnet; 170 log_info("Memory conditions: %u front=%u back=%u mesh=%u msg=%u " 171 "rrset=%u infra=%u iter=%u val=%u subnet=%u anchors=%u " 172 "alloccache=%u globalalloccache=%u me=%u", 173 (unsigned)total, (unsigned)front, (unsigned)back, 174 (unsigned)mesh, (unsigned)msg, (unsigned)rrset, (unsigned)infra, 175 (unsigned)iter, (unsigned)val, 176 (unsigned)subnet, (unsigned)anch, (unsigned)ac, 177 (unsigned)superac, (unsigned)me); 178 #else /* no CLIENT_SUBNET */ 179 log_info("Memory conditions: %u front=%u back=%u mesh=%u msg=%u " 180 "rrset=%u infra=%u iter=%u val=%u anchors=%u " 181 "alloccache=%u globalalloccache=%u me=%u", 182 (unsigned)total, (unsigned)front, (unsigned)back, 183 (unsigned)mesh, (unsigned)msg, (unsigned)rrset, 184 (unsigned)infra, (unsigned)iter, (unsigned)val, (unsigned)anch, 185 (unsigned)ac, (unsigned)superac, (unsigned)me); 186 #endif /* CLIENT_SUBNET */ 187 log_info("Total heap memory estimate: %u total-alloc: %u " 188 "total-free: %u", (unsigned)total, 189 (unsigned)unbound_mem_alloc, (unsigned)unbound_mem_freed); 190 #else /* no UNBOUND_ALLOC_STATS */ 191 size_t val = 0; 192 #ifdef CLIENT_SUBNET 193 size_t subnet = 0; 194 #endif /* CLIENT_SUBNET */ 195 int i; 196 if(verbosity < VERB_QUERY) 197 return; 198 for(i=0; i<worker->env.mesh->mods.num; i++) { 199 fptr_ok(fptr_whitelist_mod_get_mem(worker->env.mesh-> 200 mods.mod[i]->get_mem)); 201 if(strcmp(worker->env.mesh->mods.mod[i]->name, "validator")==0) 202 val += (*worker->env.mesh->mods.mod[i]->get_mem) 203 (&worker->env, i); 204 #ifdef CLIENT_SUBNET 205 else if(strcmp(worker->env.mesh->mods.mod[i]->name, 206 "subnet")==0) 207 subnet += (*worker->env.mesh->mods.mod[i]->get_mem) 208 (&worker->env, i); 209 #endif /* CLIENT_SUBNET */ 210 } 211 #ifdef CLIENT_SUBNET 212 verbose(VERB_QUERY, "cache memory msg=%u rrset=%u infra=%u val=%u " 213 "subnet=%u", 214 (unsigned)slabhash_get_mem(worker->env.msg_cache), 215 (unsigned)slabhash_get_mem(&worker->env.rrset_cache->table), 216 (unsigned)infra_get_mem(worker->env.infra_cache), 217 (unsigned)val, (unsigned)subnet); 218 #else /* no CLIENT_SUBNET */ 219 verbose(VERB_QUERY, "cache memory msg=%u rrset=%u infra=%u val=%u", 220 (unsigned)slabhash_get_mem(worker->env.msg_cache), 221 (unsigned)slabhash_get_mem(&worker->env.rrset_cache->table), 222 (unsigned)infra_get_mem(worker->env.infra_cache), 223 (unsigned)val); 224 #endif /* CLIENT_SUBNET */ 225 #endif /* UNBOUND_ALLOC_STATS */ 226 } 227 228 void 229 worker_send_cmd(struct worker* worker, enum worker_commands cmd) 230 { 231 uint32_t c = (uint32_t)htonl(cmd); 232 if(!tube_write_msg(worker->cmd, (uint8_t*)&c, sizeof(c), 0)) { 233 log_err("worker send cmd %d failed", (int)cmd); 234 } 235 } 236 237 int 238 worker_handle_reply(struct comm_point* c, void* arg, int error, 239 struct comm_reply* reply_info) 240 { 241 struct module_qstate* q = (struct module_qstate*)arg; 242 struct worker* worker = q->env->worker; 243 struct outbound_entry e; 244 e.qstate = q; 245 e.qsent = NULL; 246 247 if(error != 0) { 248 mesh_report_reply(worker->env.mesh, &e, reply_info, error); 249 worker_mem_report(worker, NULL); 250 return 0; 251 } 252 /* sanity check. */ 253 if(!LDNS_QR_WIRE(sldns_buffer_begin(c->buffer)) 254 || LDNS_OPCODE_WIRE(sldns_buffer_begin(c->buffer)) != 255 LDNS_PACKET_QUERY 256 || LDNS_QDCOUNT(sldns_buffer_begin(c->buffer)) > 1) { 257 /* error becomes timeout for the module as if this reply 258 * never arrived. */ 259 mesh_report_reply(worker->env.mesh, &e, reply_info, 260 NETEVENT_TIMEOUT); 261 worker_mem_report(worker, NULL); 262 return 0; 263 } 264 mesh_report_reply(worker->env.mesh, &e, reply_info, NETEVENT_NOERROR); 265 worker_mem_report(worker, NULL); 266 return 0; 267 } 268 269 int 270 worker_handle_service_reply(struct comm_point* c, void* arg, int error, 271 struct comm_reply* reply_info) 272 { 273 struct outbound_entry* e = (struct outbound_entry*)arg; 274 struct worker* worker = e->qstate->env->worker; 275 struct serviced_query *sq = e->qsent; 276 277 verbose(VERB_ALGO, "worker svcd callback for qstate %p", e->qstate); 278 if(error != 0) { 279 mesh_report_reply(worker->env.mesh, e, reply_info, error); 280 worker_mem_report(worker, sq); 281 return 0; 282 } 283 /* sanity check. */ 284 if(!LDNS_QR_WIRE(sldns_buffer_begin(c->buffer)) 285 || LDNS_OPCODE_WIRE(sldns_buffer_begin(c->buffer)) != 286 LDNS_PACKET_QUERY 287 || LDNS_QDCOUNT(sldns_buffer_begin(c->buffer)) > 1) { 288 /* error becomes timeout for the module as if this reply 289 * never arrived. */ 290 verbose(VERB_ALGO, "worker: bad reply handled as timeout"); 291 mesh_report_reply(worker->env.mesh, e, reply_info, 292 NETEVENT_TIMEOUT); 293 worker_mem_report(worker, sq); 294 return 0; 295 } 296 mesh_report_reply(worker->env.mesh, e, reply_info, NETEVENT_NOERROR); 297 worker_mem_report(worker, sq); 298 return 0; 299 } 300 301 /** ratelimit error replies 302 * @param worker: the worker struct with ratelimit counter 303 * @param err: error code that would be wanted. 304 * @return value of err if okay, or -1 if it should be discarded instead. 305 */ 306 static int 307 worker_err_ratelimit(struct worker* worker, int err) 308 { 309 if(worker->err_limit_time == *worker->env.now) { 310 /* see if limit is exceeded for this second */ 311 if(worker->err_limit_count++ > ERROR_RATELIMIT) 312 return -1; 313 } else { 314 /* new second, new limits */ 315 worker->err_limit_time = *worker->env.now; 316 worker->err_limit_count = 1; 317 } 318 return err; 319 } 320 321 /** check request sanity. 322 * @param pkt: the wire packet to examine for sanity. 323 * @param worker: parameters for checking. 324 * @return error code, 0 OK, or -1 discard. 325 */ 326 static int 327 worker_check_request(sldns_buffer* pkt, struct worker* worker) 328 { 329 if(sldns_buffer_limit(pkt) < LDNS_HEADER_SIZE) { 330 verbose(VERB_QUERY, "request too short, discarded"); 331 return -1; 332 } 333 if(sldns_buffer_limit(pkt) > NORMAL_UDP_SIZE && 334 worker->daemon->cfg->harden_large_queries) { 335 verbose(VERB_QUERY, "request too large, discarded"); 336 return -1; 337 } 338 if(LDNS_QR_WIRE(sldns_buffer_begin(pkt))) { 339 verbose(VERB_QUERY, "request has QR bit on, discarded"); 340 return -1; 341 } 342 if(LDNS_TC_WIRE(sldns_buffer_begin(pkt))) { 343 LDNS_TC_CLR(sldns_buffer_begin(pkt)); 344 verbose(VERB_QUERY, "request bad, has TC bit on"); 345 return worker_err_ratelimit(worker, LDNS_RCODE_FORMERR); 346 } 347 if(LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt)) != LDNS_PACKET_QUERY && 348 LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt)) != LDNS_PACKET_NOTIFY) { 349 verbose(VERB_QUERY, "request unknown opcode %d", 350 LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt))); 351 return worker_err_ratelimit(worker, LDNS_RCODE_NOTIMPL); 352 } 353 if(LDNS_QDCOUNT(sldns_buffer_begin(pkt)) != 1) { 354 verbose(VERB_QUERY, "request wrong nr qd=%d", 355 LDNS_QDCOUNT(sldns_buffer_begin(pkt))); 356 return worker_err_ratelimit(worker, LDNS_RCODE_FORMERR); 357 } 358 if(LDNS_ANCOUNT(sldns_buffer_begin(pkt)) != 0 && 359 (LDNS_ANCOUNT(sldns_buffer_begin(pkt)) != 1 || 360 LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt)) != LDNS_PACKET_NOTIFY)) { 361 verbose(VERB_QUERY, "request wrong nr an=%d", 362 LDNS_ANCOUNT(sldns_buffer_begin(pkt))); 363 return worker_err_ratelimit(worker, LDNS_RCODE_FORMERR); 364 } 365 if(LDNS_NSCOUNT(sldns_buffer_begin(pkt)) != 0) { 366 verbose(VERB_QUERY, "request wrong nr ns=%d", 367 LDNS_NSCOUNT(sldns_buffer_begin(pkt))); 368 return worker_err_ratelimit(worker, LDNS_RCODE_FORMERR); 369 } 370 if(LDNS_ARCOUNT(sldns_buffer_begin(pkt)) > 1) { 371 verbose(VERB_QUERY, "request wrong nr ar=%d", 372 LDNS_ARCOUNT(sldns_buffer_begin(pkt))); 373 return worker_err_ratelimit(worker, LDNS_RCODE_FORMERR); 374 } 375 return 0; 376 } 377 378 void 379 worker_handle_control_cmd(struct tube* ATTR_UNUSED(tube), uint8_t* msg, 380 size_t len, int error, void* arg) 381 { 382 struct worker* worker = (struct worker*)arg; 383 enum worker_commands cmd; 384 if(error != NETEVENT_NOERROR) { 385 free(msg); 386 if(error == NETEVENT_CLOSED) 387 comm_base_exit(worker->base); 388 else log_info("control event: %d", error); 389 return; 390 } 391 if(len != sizeof(uint32_t)) { 392 fatal_exit("bad control msg length %d", (int)len); 393 } 394 cmd = sldns_read_uint32(msg); 395 free(msg); 396 switch(cmd) { 397 case worker_cmd_quit: 398 verbose(VERB_ALGO, "got control cmd quit"); 399 comm_base_exit(worker->base); 400 break; 401 case worker_cmd_stats: 402 verbose(VERB_ALGO, "got control cmd stats"); 403 server_stats_reply(worker, 1); 404 break; 405 case worker_cmd_stats_noreset: 406 verbose(VERB_ALGO, "got control cmd stats_noreset"); 407 server_stats_reply(worker, 0); 408 break; 409 case worker_cmd_remote: 410 verbose(VERB_ALGO, "got control cmd remote"); 411 daemon_remote_exec(worker); 412 break; 413 default: 414 log_err("bad command %d", (int)cmd); 415 break; 416 } 417 } 418 419 /** check if a delegation is secure */ 420 static enum sec_status 421 check_delegation_secure(struct reply_info *rep) 422 { 423 /* return smallest security status */ 424 size_t i; 425 enum sec_status sec = sec_status_secure; 426 enum sec_status s; 427 size_t num = rep->an_numrrsets + rep->ns_numrrsets; 428 /* check if answer and authority are OK */ 429 for(i=0; i<num; i++) { 430 s = ((struct packed_rrset_data*)rep->rrsets[i]->entry.data) 431 ->security; 432 if(s < sec) 433 sec = s; 434 } 435 /* in additional, only unchecked triggers revalidation */ 436 for(i=num; i<rep->rrset_count; i++) { 437 s = ((struct packed_rrset_data*)rep->rrsets[i]->entry.data) 438 ->security; 439 if(s == sec_status_unchecked) 440 return s; 441 } 442 return sec; 443 } 444 445 /** remove nonsecure from a delegation referral additional section */ 446 static void 447 deleg_remove_nonsecure_additional(struct reply_info* rep) 448 { 449 /* we can simply edit it, since we are working in the scratch region */ 450 size_t i; 451 enum sec_status s; 452 453 for(i = rep->an_numrrsets+rep->ns_numrrsets; i<rep->rrset_count; i++) { 454 s = ((struct packed_rrset_data*)rep->rrsets[i]->entry.data) 455 ->security; 456 if(s != sec_status_secure) { 457 memmove(rep->rrsets+i, rep->rrsets+i+1, 458 sizeof(struct ub_packed_rrset_key*)* 459 (rep->rrset_count - i - 1)); 460 rep->ar_numrrsets--; 461 rep->rrset_count--; 462 i--; 463 } 464 } 465 } 466 467 /** answer nonrecursive query from the cache */ 468 static int 469 answer_norec_from_cache(struct worker* worker, struct query_info* qinfo, 470 uint16_t id, uint16_t flags, struct comm_reply* repinfo, 471 struct edns_data* edns) 472 { 473 /* for a nonrecursive query return either: 474 * o an error (servfail; we try to avoid this) 475 * o a delegation (closest we have; this routine tries that) 476 * o the answer (checked by answer_from_cache) 477 * 478 * So, grab a delegation from the rrset cache. 479 * Then check if it needs validation, if so, this routine fails, 480 * so that iterator can prime and validator can verify rrsets. 481 */ 482 struct edns_data edns_bak; 483 uint16_t udpsize = edns->udp_size; 484 int secure = 0; 485 time_t timenow = *worker->env.now; 486 int must_validate = (!(flags&BIT_CD) || worker->env.cfg->ignore_cd) 487 && worker->env.need_to_validate; 488 struct dns_msg *msg = NULL; 489 struct delegpt *dp; 490 491 dp = dns_cache_find_delegation(&worker->env, qinfo->qname, 492 qinfo->qname_len, qinfo->qtype, qinfo->qclass, 493 worker->scratchpad, &msg, timenow); 494 if(!dp) { /* no delegation, need to reprime */ 495 return 0; 496 } 497 /* In case we have a local alias, copy it into the delegation message. 498 * Shallow copy should be fine, as we'll be done with msg in this 499 * function. */ 500 msg->qinfo.local_alias = qinfo->local_alias; 501 if(must_validate) { 502 switch(check_delegation_secure(msg->rep)) { 503 case sec_status_unchecked: 504 /* some rrsets have not been verified yet, go and 505 * let validator do that */ 506 return 0; 507 case sec_status_bogus: 508 case sec_status_secure_sentinel_fail: 509 /* some rrsets are bogus, reply servfail */ 510 edns->edns_version = EDNS_ADVERTISED_VERSION; 511 edns->udp_size = EDNS_ADVERTISED_SIZE; 512 edns->ext_rcode = 0; 513 edns->bits &= EDNS_DO; 514 if(!inplace_cb_reply_servfail_call(&worker->env, qinfo, NULL, 515 msg->rep, LDNS_RCODE_SERVFAIL, edns, repinfo, worker->scratchpad)) 516 return 0; 517 error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL, 518 &msg->qinfo, id, flags, edns); 519 if(worker->stats.extended) { 520 worker->stats.ans_bogus++; 521 worker->stats.ans_rcode[LDNS_RCODE_SERVFAIL]++; 522 } 523 return 1; 524 case sec_status_secure: 525 /* all rrsets are secure */ 526 /* remove non-secure rrsets from the add. section*/ 527 if(worker->env.cfg->val_clean_additional) 528 deleg_remove_nonsecure_additional(msg->rep); 529 secure = 1; 530 break; 531 case sec_status_indeterminate: 532 case sec_status_insecure: 533 default: 534 /* not secure */ 535 secure = 0; 536 break; 537 } 538 } 539 /* return this delegation from the cache */ 540 edns_bak = *edns; 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 return 0; 548 msg->rep->flags |= BIT_QR|BIT_RA; 549 if(!apply_edns_options(edns, &edns_bak, worker->env.cfg, 550 repinfo->c, worker->scratchpad) || 551 !reply_info_answer_encode(&msg->qinfo, msg->rep, id, flags, 552 repinfo->c->buffer, 0, 1, worker->scratchpad, 553 udpsize, edns, (int)(edns->bits & EDNS_DO), secure)) { 554 if(!inplace_cb_reply_servfail_call(&worker->env, qinfo, NULL, NULL, 555 LDNS_RCODE_SERVFAIL, edns, repinfo, worker->scratchpad)) 556 edns->opt_list = NULL; 557 error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL, 558 &msg->qinfo, id, flags, edns); 559 } 560 if(worker->stats.extended) { 561 if(secure) worker->stats.ans_secure++; 562 server_stats_insrcode(&worker->stats, repinfo->c->buffer); 563 } 564 return 1; 565 } 566 567 /** Apply, if applicable, a response IP action to a cached answer. 568 * If the answer is rewritten as a result of an action, '*encode_repp' will 569 * point to the reply info containing the modified answer. '*encode_repp' will 570 * be intact otherwise. 571 * It returns 1 on success, 0 otherwise. */ 572 static int 573 apply_respip_action(struct worker* worker, const struct query_info* qinfo, 574 struct respip_client_info* cinfo, struct reply_info* rep, 575 struct comm_reply* repinfo, struct ub_packed_rrset_key** alias_rrset, 576 struct reply_info** encode_repp, struct auth_zones* az) 577 { 578 struct respip_action_info actinfo = {0}; 579 actinfo.action = respip_none; 580 581 if(qinfo->qtype != LDNS_RR_TYPE_A && 582 qinfo->qtype != LDNS_RR_TYPE_AAAA && 583 qinfo->qtype != LDNS_RR_TYPE_ANY) 584 return 1; 585 586 if(!respip_rewrite_reply(qinfo, cinfo, rep, encode_repp, &actinfo, 587 alias_rrset, 0, worker->scratchpad, az)) 588 return 0; 589 590 /* xxx_deny actions mean dropping the reply, unless the original reply 591 * was redirected to response-ip data. */ 592 if((actinfo.action == respip_deny || 593 actinfo.action == respip_inform_deny) && 594 *encode_repp == rep) 595 *encode_repp = NULL; 596 597 /* If address info is returned, it means the action should be an 598 * 'inform' variant and the information should be logged. */ 599 if(actinfo.addrinfo) { 600 respip_inform_print(&actinfo, qinfo->qname, 601 qinfo->qtype, qinfo->qclass, qinfo->local_alias, 602 repinfo); 603 604 if(worker->stats.extended && actinfo.rpz_used) { 605 if(actinfo.rpz_disabled) 606 worker->stats.rpz_action[RPZ_DISABLED_ACTION]++; 607 if(actinfo.rpz_cname_override) 608 worker->stats.rpz_action[RPZ_CNAME_OVERRIDE_ACTION]++; 609 else 610 worker->stats.rpz_action[ 611 respip_action_to_rpz_action(actinfo.action)]++; 612 } 613 } 614 615 return 1; 616 } 617 618 /** answer query from the cache. 619 * Normally, the answer message will be built in repinfo->c->buffer; if the 620 * answer is supposed to be suppressed or the answer is supposed to be an 621 * incomplete CNAME chain, the buffer is explicitly cleared to signal the 622 * caller as such. In the latter case *partial_rep will point to the incomplete 623 * reply, and this function is (possibly) supposed to be called again with that 624 * *partial_rep value to complete the chain. In addition, if the query should 625 * be completely dropped, '*need_drop' will be set to 1. */ 626 static int 627 answer_from_cache(struct worker* worker, struct query_info* qinfo, 628 struct respip_client_info* cinfo, int* need_drop, int* is_expired_answer, 629 int* is_secure_answer, struct ub_packed_rrset_key** alias_rrset, 630 struct reply_info** partial_repp, 631 struct reply_info* rep, uint16_t id, uint16_t flags, 632 struct comm_reply* repinfo, struct edns_data* edns) 633 { 634 struct edns_data edns_bak; 635 time_t timenow = *worker->env.now; 636 uint16_t udpsize = edns->udp_size; 637 struct reply_info* encode_rep = rep; 638 struct reply_info* partial_rep = *partial_repp; 639 int must_validate = (!(flags&BIT_CD) || worker->env.cfg->ignore_cd) 640 && worker->env.need_to_validate; 641 *partial_repp = NULL; /* avoid accidental further pass */ 642 643 /* Check TTL */ 644 if(rep->ttl < timenow) { 645 /* Check if we need to serve expired now */ 646 if(worker->env.cfg->serve_expired && 647 !worker->env.cfg->serve_expired_client_timeout) { 648 if(worker->env.cfg->serve_expired_ttl && 649 rep->serve_expired_ttl < timenow) 650 return 0; 651 if(!rrset_array_lock(rep->ref, rep->rrset_count, 0)) 652 return 0; 653 *is_expired_answer = 1; 654 } else { 655 /* the rrsets may have been updated in the meantime. 656 * we will refetch the message format from the 657 * authoritative server 658 */ 659 return 0; 660 } 661 } else { 662 if(!rrset_array_lock(rep->ref, rep->rrset_count, timenow)) 663 return 0; 664 } 665 /* locked and ids and ttls are OK. */ 666 667 /* check CNAME chain (if any) */ 668 if(rep->an_numrrsets > 0 && (rep->rrsets[0]->rk.type == 669 htons(LDNS_RR_TYPE_CNAME) || rep->rrsets[0]->rk.type == 670 htons(LDNS_RR_TYPE_DNAME))) { 671 if(!reply_check_cname_chain(qinfo, rep)) { 672 /* cname chain invalid, redo iterator steps */ 673 verbose(VERB_ALGO, "Cache reply: cname chain broken"); 674 goto bail_out; 675 } 676 } 677 /* check security status of the cached answer */ 678 if(must_validate && (rep->security == sec_status_bogus || 679 rep->security == sec_status_secure_sentinel_fail)) { 680 /* BAD cached */ 681 edns->edns_version = EDNS_ADVERTISED_VERSION; 682 edns->udp_size = EDNS_ADVERTISED_SIZE; 683 edns->ext_rcode = 0; 684 edns->bits &= EDNS_DO; 685 if(!inplace_cb_reply_servfail_call(&worker->env, qinfo, NULL, rep, 686 LDNS_RCODE_SERVFAIL, edns, repinfo, worker->scratchpad)) 687 goto bail_out; 688 error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL, 689 qinfo, id, flags, edns); 690 rrset_array_unlock_touch(worker->env.rrset_cache, 691 worker->scratchpad, rep->ref, rep->rrset_count); 692 if(worker->stats.extended) { 693 worker->stats.ans_bogus ++; 694 worker->stats.ans_rcode[LDNS_RCODE_SERVFAIL] ++; 695 } 696 return 1; 697 } else if(rep->security == sec_status_unchecked && must_validate) { 698 verbose(VERB_ALGO, "Cache reply: unchecked entry needs " 699 "validation"); 700 goto bail_out; /* need to validate cache entry first */ 701 } else if(rep->security == sec_status_secure) { 702 if(reply_all_rrsets_secure(rep)) { 703 *is_secure_answer = 1; 704 } else { 705 if(must_validate) { 706 verbose(VERB_ALGO, "Cache reply: secure entry" 707 " changed status"); 708 goto bail_out; /* rrset changed, re-verify */ 709 } 710 *is_secure_answer = 0; 711 } 712 } else *is_secure_answer = 0; 713 714 edns_bak = *edns; 715 edns->edns_version = EDNS_ADVERTISED_VERSION; 716 edns->udp_size = EDNS_ADVERTISED_SIZE; 717 edns->ext_rcode = 0; 718 edns->bits &= EDNS_DO; 719 if(!inplace_cb_reply_cache_call(&worker->env, qinfo, NULL, rep, 720 (int)(flags&LDNS_RCODE_MASK), edns, repinfo, worker->scratchpad)) 721 goto bail_out; 722 *alias_rrset = NULL; /* avoid confusion if caller set it to non-NULL */ 723 if((worker->daemon->use_response_ip || worker->daemon->use_rpz) && 724 !partial_rep && !apply_respip_action(worker, qinfo, cinfo, rep, 725 repinfo, alias_rrset, 726 &encode_rep, worker->env.auth_zones)) { 727 goto bail_out; 728 } else if(partial_rep && 729 !respip_merge_cname(partial_rep, qinfo, rep, cinfo, 730 must_validate, &encode_rep, worker->scratchpad, 731 worker->env.auth_zones)) { 732 goto bail_out; 733 } 734 if(encode_rep != rep) { 735 /* if rewritten, it can't be considered "secure" */ 736 *is_secure_answer = 0; 737 } 738 if(!encode_rep || *alias_rrset) { 739 if(!encode_rep) 740 *need_drop = 1; 741 else { 742 /* If a partial CNAME chain is found, we first need to 743 * make a copy of the reply in the scratchpad so we 744 * can release the locks and lookup the cache again. */ 745 *partial_repp = reply_info_copy(encode_rep, NULL, 746 worker->scratchpad); 747 if(!*partial_repp) 748 goto bail_out; 749 } 750 } else if(!apply_edns_options(edns, &edns_bak, worker->env.cfg, 751 repinfo->c, worker->scratchpad) || 752 !reply_info_answer_encode(qinfo, encode_rep, id, flags, 753 repinfo->c->buffer, timenow, 1, worker->scratchpad, 754 udpsize, edns, (int)(edns->bits & EDNS_DO), *is_secure_answer)) { 755 if(!inplace_cb_reply_servfail_call(&worker->env, qinfo, NULL, NULL, 756 LDNS_RCODE_SERVFAIL, edns, repinfo, worker->scratchpad)) 757 edns->opt_list = NULL; 758 error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL, 759 qinfo, id, flags, edns); 760 } 761 /* cannot send the reply right now, because blocking network syscall 762 * is bad while holding locks. */ 763 rrset_array_unlock_touch(worker->env.rrset_cache, worker->scratchpad, 764 rep->ref, rep->rrset_count); 765 /* go and return this buffer to the client */ 766 return 1; 767 768 bail_out: 769 rrset_array_unlock_touch(worker->env.rrset_cache, 770 worker->scratchpad, rep->ref, rep->rrset_count); 771 return 0; 772 } 773 774 /** Reply to client and perform prefetch to keep cache up to date. */ 775 static void 776 reply_and_prefetch(struct worker* worker, struct query_info* qinfo, 777 uint16_t flags, struct comm_reply* repinfo, time_t leeway, int noreply) 778 { 779 /* first send answer to client to keep its latency 780 * as small as a cachereply */ 781 if(!noreply) { 782 if(repinfo->c->tcp_req_info) { 783 sldns_buffer_copy( 784 repinfo->c->tcp_req_info->spool_buffer, 785 repinfo->c->buffer); 786 } 787 comm_point_send_reply(repinfo); 788 } 789 server_stats_prefetch(&worker->stats, worker); 790 791 /* create the prefetch in the mesh as a normal lookup without 792 * client addrs waiting, which has the cache blacklisted (to bypass 793 * the cache and go to the network for the data). */ 794 /* this (potentially) runs the mesh for the new query */ 795 mesh_new_prefetch(worker->env.mesh, qinfo, flags, leeway + 796 PREFETCH_EXPIRY_ADD); 797 } 798 799 /** 800 * Fill CH class answer into buffer. Keeps query. 801 * @param pkt: buffer 802 * @param str: string to put into text record (<255). 803 * array of strings, every string becomes a text record. 804 * @param num: number of strings in array. 805 * @param edns: edns reply information. 806 * @param worker: worker with scratch region. 807 * @param repinfo: reply information for a communication point. 808 */ 809 static void 810 chaos_replystr(sldns_buffer* pkt, char** str, int num, struct edns_data* edns, 811 struct worker* worker, struct comm_reply* repinfo) 812 { 813 int i; 814 unsigned int rd = LDNS_RD_WIRE(sldns_buffer_begin(pkt)); 815 unsigned int cd = LDNS_CD_WIRE(sldns_buffer_begin(pkt)); 816 sldns_buffer_clear(pkt); 817 sldns_buffer_skip(pkt, (ssize_t)sizeof(uint16_t)); /* skip id */ 818 sldns_buffer_write_u16(pkt, (uint16_t)(BIT_QR|BIT_RA)); 819 if(rd) LDNS_RD_SET(sldns_buffer_begin(pkt)); 820 if(cd) LDNS_CD_SET(sldns_buffer_begin(pkt)); 821 sldns_buffer_write_u16(pkt, 1); /* qdcount */ 822 sldns_buffer_write_u16(pkt, (uint16_t)num); /* ancount */ 823 sldns_buffer_write_u16(pkt, 0); /* nscount */ 824 sldns_buffer_write_u16(pkt, 0); /* arcount */ 825 (void)query_dname_len(pkt); /* skip qname */ 826 sldns_buffer_skip(pkt, (ssize_t)sizeof(uint16_t)); /* skip qtype */ 827 sldns_buffer_skip(pkt, (ssize_t)sizeof(uint16_t)); /* skip qclass */ 828 for(i=0; i<num; i++) { 829 size_t len = strlen(str[i]); 830 if(len>255) len=255; /* cap size of TXT record */ 831 sldns_buffer_write_u16(pkt, 0xc00c); /* compr ptr to query */ 832 sldns_buffer_write_u16(pkt, LDNS_RR_TYPE_TXT); 833 sldns_buffer_write_u16(pkt, LDNS_RR_CLASS_CH); 834 sldns_buffer_write_u32(pkt, 0); /* TTL */ 835 sldns_buffer_write_u16(pkt, sizeof(uint8_t) + len); 836 sldns_buffer_write_u8(pkt, len); 837 sldns_buffer_write(pkt, str[i], len); 838 } 839 sldns_buffer_flip(pkt); 840 edns->edns_version = EDNS_ADVERTISED_VERSION; 841 edns->udp_size = EDNS_ADVERTISED_SIZE; 842 edns->bits &= EDNS_DO; 843 if(!inplace_cb_reply_local_call(&worker->env, NULL, NULL, NULL, 844 LDNS_RCODE_NOERROR, edns, repinfo, worker->scratchpad)) 845 edns->opt_list = NULL; 846 if(sldns_buffer_capacity(pkt) >= 847 sldns_buffer_limit(pkt)+calc_edns_field_size(edns)) 848 attach_edns_record(pkt, edns); 849 } 850 851 /** Reply with one string */ 852 static void 853 chaos_replyonestr(sldns_buffer* pkt, const char* str, struct edns_data* edns, 854 struct worker* worker, struct comm_reply* repinfo) 855 { 856 chaos_replystr(pkt, (char**)&str, 1, edns, worker, repinfo); 857 } 858 859 /** 860 * Create CH class trustanchor answer. 861 * @param pkt: buffer 862 * @param edns: edns reply information. 863 * @param w: worker with scratch region. 864 * @param repinfo: reply information for a communication point. 865 */ 866 static void 867 chaos_trustanchor(sldns_buffer* pkt, struct edns_data* edns, struct worker* w, 868 struct comm_reply* repinfo) 869 { 870 #define TA_RESPONSE_MAX_TXT 16 /* max number of TXT records */ 871 #define TA_RESPONSE_MAX_TAGS 32 /* max number of tags printed per zone */ 872 char* str_array[TA_RESPONSE_MAX_TXT]; 873 uint16_t tags[TA_RESPONSE_MAX_TAGS]; 874 int num = 0; 875 struct trust_anchor* ta; 876 877 if(!w->env.need_to_validate) { 878 /* no validator module, reply no trustanchors */ 879 chaos_replystr(pkt, NULL, 0, edns, w, repinfo); 880 return; 881 } 882 883 /* fill the string with contents */ 884 lock_basic_lock(&w->env.anchors->lock); 885 RBTREE_FOR(ta, struct trust_anchor*, w->env.anchors->tree) { 886 char* str; 887 size_t i, numtag, str_len = 255; 888 if(num == TA_RESPONSE_MAX_TXT) continue; 889 str = (char*)regional_alloc(w->scratchpad, str_len); 890 if(!str) continue; 891 lock_basic_lock(&ta->lock); 892 numtag = anchor_list_keytags(ta, tags, TA_RESPONSE_MAX_TAGS); 893 if(numtag == 0) { 894 /* empty, insecure point */ 895 lock_basic_unlock(&ta->lock); 896 continue; 897 } 898 str_array[num] = str; 899 num++; 900 901 /* spool name of anchor */ 902 (void)sldns_wire2str_dname_buf(ta->name, ta->namelen, str, str_len); 903 str_len -= strlen(str); str += strlen(str); 904 /* spool tags */ 905 for(i=0; i<numtag; i++) { 906 snprintf(str, str_len, " %u", (unsigned)tags[i]); 907 str_len -= strlen(str); str += strlen(str); 908 } 909 lock_basic_unlock(&ta->lock); 910 } 911 lock_basic_unlock(&w->env.anchors->lock); 912 913 chaos_replystr(pkt, str_array, num, edns, w, repinfo); 914 regional_free_all(w->scratchpad); 915 } 916 917 /** 918 * Answer CH class queries. 919 * @param w: worker 920 * @param qinfo: query info. Pointer into packet buffer. 921 * @param edns: edns info from query. 922 * @param repinfo: reply information for a communication point. 923 * @param pkt: packet buffer. 924 * @return: true if a reply is to be sent. 925 */ 926 static int 927 answer_chaos(struct worker* w, struct query_info* qinfo, 928 struct edns_data* edns, struct comm_reply* repinfo, sldns_buffer* pkt) 929 { 930 struct config_file* cfg = w->env.cfg; 931 if(qinfo->qtype != LDNS_RR_TYPE_ANY && qinfo->qtype != LDNS_RR_TYPE_TXT) 932 return 0; 933 if(query_dname_compare(qinfo->qname, 934 (uint8_t*)"\002id\006server") == 0 || 935 query_dname_compare(qinfo->qname, 936 (uint8_t*)"\010hostname\004bind") == 0) 937 { 938 if(cfg->hide_identity) 939 return 0; 940 if(cfg->identity==NULL || cfg->identity[0]==0) { 941 char buf[MAXHOSTNAMELEN+1]; 942 if (gethostname(buf, MAXHOSTNAMELEN) == 0) { 943 buf[MAXHOSTNAMELEN] = 0; 944 chaos_replyonestr(pkt, buf, edns, w, repinfo); 945 } else { 946 log_err("gethostname: %s", strerror(errno)); 947 chaos_replyonestr(pkt, "no hostname", edns, w, repinfo); 948 } 949 } 950 else chaos_replyonestr(pkt, cfg->identity, edns, w, repinfo); 951 return 1; 952 } 953 if(query_dname_compare(qinfo->qname, 954 (uint8_t*)"\007version\006server") == 0 || 955 query_dname_compare(qinfo->qname, 956 (uint8_t*)"\007version\004bind") == 0) 957 { 958 if(cfg->hide_version) 959 return 0; 960 if(cfg->version==NULL || cfg->version[0]==0) 961 chaos_replyonestr(pkt, PACKAGE_STRING, edns, w, repinfo); 962 else chaos_replyonestr(pkt, cfg->version, edns, w, repinfo); 963 return 1; 964 } 965 if(query_dname_compare(qinfo->qname, 966 (uint8_t*)"\013trustanchor\007unbound") == 0) 967 { 968 if(cfg->hide_trustanchor) 969 return 0; 970 chaos_trustanchor(pkt, edns, w, repinfo); 971 return 1; 972 } 973 974 return 0; 975 } 976 977 /** 978 * Answer notify queries. These are notifies for authoritative zones, 979 * the reply is an ack that the notify has been received. We need to check 980 * access permission here. 981 * @param w: worker 982 * @param qinfo: query info. Pointer into packet buffer. 983 * @param edns: edns info from query. 984 * @param repinfo: reply info with source address. 985 * @param pkt: packet buffer. 986 */ 987 static void 988 answer_notify(struct worker* w, struct query_info* qinfo, 989 struct edns_data* edns, sldns_buffer* pkt, struct comm_reply* repinfo) 990 { 991 int refused = 0; 992 int rcode = LDNS_RCODE_NOERROR; 993 uint32_t serial = 0; 994 int has_serial; 995 if(!w->env.auth_zones) return; 996 has_serial = auth_zone_parse_notify_serial(pkt, &serial); 997 if(auth_zones_notify(w->env.auth_zones, &w->env, qinfo->qname, 998 qinfo->qname_len, qinfo->qclass, &repinfo->addr, 999 repinfo->addrlen, has_serial, serial, &refused)) { 1000 rcode = LDNS_RCODE_NOERROR; 1001 } else { 1002 if(refused) 1003 rcode = LDNS_RCODE_REFUSED; 1004 else rcode = LDNS_RCODE_SERVFAIL; 1005 } 1006 1007 if(verbosity >= VERB_DETAIL) { 1008 char buf[380]; 1009 char zname[255+1]; 1010 char sr[25]; 1011 dname_str(qinfo->qname, zname); 1012 sr[0]=0; 1013 if(has_serial) 1014 snprintf(sr, sizeof(sr), "serial %u ", 1015 (unsigned)serial); 1016 if(rcode == LDNS_RCODE_REFUSED) 1017 snprintf(buf, sizeof(buf), 1018 "refused NOTIFY %sfor %s from", sr, zname); 1019 else if(rcode == LDNS_RCODE_SERVFAIL) 1020 snprintf(buf, sizeof(buf), 1021 "servfail for NOTIFY %sfor %s from", sr, zname); 1022 else snprintf(buf, sizeof(buf), 1023 "received NOTIFY %sfor %s from", sr, zname); 1024 log_addr(VERB_DETAIL, buf, &repinfo->addr, repinfo->addrlen); 1025 } 1026 edns->edns_version = EDNS_ADVERTISED_VERSION; 1027 edns->udp_size = EDNS_ADVERTISED_SIZE; 1028 edns->ext_rcode = 0; 1029 edns->bits &= EDNS_DO; 1030 edns->opt_list = NULL; 1031 error_encode(pkt, rcode, qinfo, 1032 *(uint16_t*)(void *)sldns_buffer_begin(pkt), 1033 sldns_buffer_read_u16_at(pkt, 2), edns); 1034 LDNS_OPCODE_SET(sldns_buffer_begin(pkt), LDNS_PACKET_NOTIFY); 1035 } 1036 1037 static int 1038 deny_refuse(struct comm_point* c, enum acl_access acl, 1039 enum acl_access deny, enum acl_access refuse, 1040 struct worker* worker, struct comm_reply* repinfo) 1041 { 1042 if(acl == deny) { 1043 comm_point_drop_reply(repinfo); 1044 if(worker->stats.extended) 1045 worker->stats.unwanted_queries++; 1046 return 0; 1047 } else if(acl == refuse) { 1048 log_addr(VERB_ALGO, "refused query from", 1049 &repinfo->addr, repinfo->addrlen); 1050 log_buf(VERB_ALGO, "refuse", c->buffer); 1051 if(worker->stats.extended) 1052 worker->stats.unwanted_queries++; 1053 if(worker_check_request(c->buffer, worker) == -1) { 1054 comm_point_drop_reply(repinfo); 1055 return 0; /* discard this */ 1056 } 1057 sldns_buffer_set_limit(c->buffer, LDNS_HEADER_SIZE); 1058 sldns_buffer_write_at(c->buffer, 4, 1059 (uint8_t*)"\0\0\0\0\0\0\0\0", 8); 1060 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1061 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1062 LDNS_RCODE_REFUSED); 1063 sldns_buffer_set_position(c->buffer, LDNS_HEADER_SIZE); 1064 sldns_buffer_flip(c->buffer); 1065 return 1; 1066 } 1067 1068 return -1; 1069 } 1070 1071 static int 1072 deny_refuse_all(struct comm_point* c, enum acl_access acl, 1073 struct worker* worker, struct comm_reply* repinfo) 1074 { 1075 return deny_refuse(c, acl, acl_deny, acl_refuse, worker, repinfo); 1076 } 1077 1078 static int 1079 deny_refuse_non_local(struct comm_point* c, enum acl_access acl, 1080 struct worker* worker, struct comm_reply* repinfo) 1081 { 1082 return deny_refuse(c, acl, acl_deny_non_local, acl_refuse_non_local, worker, repinfo); 1083 } 1084 1085 int 1086 worker_handle_request(struct comm_point* c, void* arg, int error, 1087 struct comm_reply* repinfo) 1088 { 1089 struct worker* worker = (struct worker*)arg; 1090 int ret; 1091 hashvalue_type h; 1092 struct lruhash_entry* e; 1093 struct query_info qinfo; 1094 struct edns_data edns; 1095 enum acl_access acl; 1096 struct acl_addr* acladdr; 1097 int rc = 0; 1098 int need_drop = 0; 1099 int is_expired_answer = 0; 1100 int is_secure_answer = 0; 1101 /* We might have to chase a CNAME chain internally, in which case 1102 * we'll have up to two replies and combine them to build a complete 1103 * answer. These variables control this case. */ 1104 struct ub_packed_rrset_key* alias_rrset = NULL; 1105 struct reply_info* partial_rep = NULL; 1106 struct query_info* lookup_qinfo = &qinfo; 1107 struct query_info qinfo_tmp; /* placeholder for lookup_qinfo */ 1108 struct respip_client_info* cinfo = NULL, cinfo_tmp; 1109 memset(&qinfo, 0, sizeof(qinfo)); 1110 1111 if(error != NETEVENT_NOERROR || !repinfo) { 1112 /* some bad tcp query DNS formats give these error calls */ 1113 verbose(VERB_ALGO, "handle request called with err=%d", error); 1114 return 0; 1115 } 1116 #ifdef USE_DNSCRYPT 1117 repinfo->max_udp_size = worker->daemon->cfg->max_udp_size; 1118 if(!dnsc_handle_curved_request(worker->daemon->dnscenv, repinfo)) { 1119 worker->stats.num_query_dnscrypt_crypted_malformed++; 1120 return 0; 1121 } 1122 if(c->dnscrypt && !repinfo->is_dnscrypted) { 1123 char buf[LDNS_MAX_DOMAINLEN+1]; 1124 /* Check if this is unencrypted and asking for certs */ 1125 if(worker_check_request(c->buffer, worker) != 0) { 1126 verbose(VERB_ALGO, 1127 "dnscrypt: worker check request: bad query."); 1128 log_addr(VERB_CLIENT,"from",&repinfo->addr, 1129 repinfo->addrlen); 1130 comm_point_drop_reply(repinfo); 1131 return 0; 1132 } 1133 if(!query_info_parse(&qinfo, c->buffer)) { 1134 verbose(VERB_ALGO, 1135 "dnscrypt: worker parse request: formerror."); 1136 log_addr(VERB_CLIENT, "from", &repinfo->addr, 1137 repinfo->addrlen); 1138 comm_point_drop_reply(repinfo); 1139 return 0; 1140 } 1141 dname_str(qinfo.qname, buf); 1142 if(!(qinfo.qtype == LDNS_RR_TYPE_TXT && 1143 strcasecmp(buf, 1144 worker->daemon->dnscenv->provider_name) == 0)) { 1145 verbose(VERB_ALGO, 1146 "dnscrypt: not TXT \"%s\". Received: %s \"%s\"", 1147 worker->daemon->dnscenv->provider_name, 1148 sldns_rr_descript(qinfo.qtype)->_name, 1149 buf); 1150 comm_point_drop_reply(repinfo); 1151 worker->stats.num_query_dnscrypt_cleartext++; 1152 return 0; 1153 } 1154 worker->stats.num_query_dnscrypt_cert++; 1155 sldns_buffer_rewind(c->buffer); 1156 } else if(c->dnscrypt && repinfo->is_dnscrypted) { 1157 worker->stats.num_query_dnscrypt_crypted++; 1158 } 1159 #endif 1160 #ifdef USE_DNSTAP 1161 if(worker->dtenv.log_client_query_messages) 1162 dt_msg_send_client_query(&worker->dtenv, &repinfo->addr, c->type, 1163 c->buffer); 1164 #endif 1165 acladdr = acl_addr_lookup(worker->daemon->acl, &repinfo->addr, 1166 repinfo->addrlen); 1167 acl = acl_get_control(acladdr); 1168 if((ret=deny_refuse_all(c, acl, worker, repinfo)) != -1) 1169 { 1170 if(ret == 1) 1171 goto send_reply; 1172 return ret; 1173 } 1174 if((ret=worker_check_request(c->buffer, worker)) != 0) { 1175 verbose(VERB_ALGO, "worker check request: bad query."); 1176 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen); 1177 if(ret != -1) { 1178 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1179 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), ret); 1180 return 1; 1181 } 1182 comm_point_drop_reply(repinfo); 1183 return 0; 1184 } 1185 1186 worker->stats.num_queries++; 1187 1188 /* check if this query should be dropped based on source ip rate limiting */ 1189 if(!infra_ip_ratelimit_inc(worker->env.infra_cache, repinfo, 1190 *worker->env.now, c->buffer)) { 1191 /* See if we are passed through with slip factor */ 1192 if(worker->env.cfg->ip_ratelimit_factor != 0 && 1193 ub_random_max(worker->env.rnd, 1194 worker->env.cfg->ip_ratelimit_factor) == 0) { 1195 1196 char addrbuf[128]; 1197 addr_to_str(&repinfo->addr, repinfo->addrlen, 1198 addrbuf, sizeof(addrbuf)); 1199 verbose(VERB_QUERY, "ip_ratelimit allowed through for ip address %s because of slip in ip_ratelimit_factor", 1200 addrbuf); 1201 } else { 1202 worker->stats.num_queries_ip_ratelimited++; 1203 comm_point_drop_reply(repinfo); 1204 return 0; 1205 } 1206 } 1207 1208 /* see if query is in the cache */ 1209 if(!query_info_parse(&qinfo, c->buffer)) { 1210 verbose(VERB_ALGO, "worker parse request: formerror."); 1211 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen); 1212 memset(&qinfo, 0, sizeof(qinfo)); /* zero qinfo.qname */ 1213 if(worker_err_ratelimit(worker, LDNS_RCODE_FORMERR) == -1) { 1214 comm_point_drop_reply(repinfo); 1215 return 0; 1216 } 1217 sldns_buffer_rewind(c->buffer); 1218 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1219 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1220 LDNS_RCODE_FORMERR); 1221 server_stats_insrcode(&worker->stats, c->buffer); 1222 goto send_reply; 1223 } 1224 if(worker->env.cfg->log_queries) { 1225 char ip[128]; 1226 addr_to_str(&repinfo->addr, repinfo->addrlen, ip, sizeof(ip)); 1227 log_query_in(ip, qinfo.qname, qinfo.qtype, qinfo.qclass); 1228 } 1229 if(qinfo.qtype == LDNS_RR_TYPE_AXFR || 1230 qinfo.qtype == LDNS_RR_TYPE_IXFR) { 1231 verbose(VERB_ALGO, "worker request: refused zone transfer."); 1232 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen); 1233 sldns_buffer_rewind(c->buffer); 1234 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1235 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1236 LDNS_RCODE_REFUSED); 1237 if(worker->stats.extended) { 1238 worker->stats.qtype[qinfo.qtype]++; 1239 server_stats_insrcode(&worker->stats, c->buffer); 1240 } 1241 goto send_reply; 1242 } 1243 if(qinfo.qtype == LDNS_RR_TYPE_OPT || 1244 qinfo.qtype == LDNS_RR_TYPE_TSIG || 1245 qinfo.qtype == LDNS_RR_TYPE_TKEY || 1246 qinfo.qtype == LDNS_RR_TYPE_MAILA || 1247 qinfo.qtype == LDNS_RR_TYPE_MAILB || 1248 (qinfo.qtype >= 128 && qinfo.qtype <= 248)) { 1249 verbose(VERB_ALGO, "worker request: formerror for meta-type."); 1250 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen); 1251 if(worker_err_ratelimit(worker, LDNS_RCODE_FORMERR) == -1) { 1252 comm_point_drop_reply(repinfo); 1253 return 0; 1254 } 1255 sldns_buffer_rewind(c->buffer); 1256 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1257 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1258 LDNS_RCODE_FORMERR); 1259 if(worker->stats.extended) { 1260 worker->stats.qtype[qinfo.qtype]++; 1261 server_stats_insrcode(&worker->stats, c->buffer); 1262 } 1263 goto send_reply; 1264 } 1265 if((ret=parse_edns_from_pkt(c->buffer, &edns, worker->scratchpad)) != 0) { 1266 struct edns_data reply_edns; 1267 verbose(VERB_ALGO, "worker parse edns: formerror."); 1268 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen); 1269 memset(&reply_edns, 0, sizeof(reply_edns)); 1270 reply_edns.edns_present = 1; 1271 reply_edns.udp_size = EDNS_ADVERTISED_SIZE; 1272 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), ret); 1273 error_encode(c->buffer, ret, &qinfo, 1274 *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 1275 sldns_buffer_read_u16_at(c->buffer, 2), &reply_edns); 1276 regional_free_all(worker->scratchpad); 1277 server_stats_insrcode(&worker->stats, c->buffer); 1278 goto send_reply; 1279 } 1280 if(edns.edns_present) { 1281 struct edns_option* edns_opt; 1282 if(edns.edns_version != 0) { 1283 edns.ext_rcode = (uint8_t)(EDNS_RCODE_BADVERS>>4); 1284 edns.edns_version = EDNS_ADVERTISED_VERSION; 1285 edns.udp_size = EDNS_ADVERTISED_SIZE; 1286 edns.bits &= EDNS_DO; 1287 edns.opt_list = NULL; 1288 verbose(VERB_ALGO, "query with bad edns version."); 1289 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen); 1290 error_encode(c->buffer, EDNS_RCODE_BADVERS&0xf, &qinfo, 1291 *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 1292 sldns_buffer_read_u16_at(c->buffer, 2), NULL); 1293 if(sldns_buffer_capacity(c->buffer) >= 1294 sldns_buffer_limit(c->buffer)+calc_edns_field_size(&edns)) 1295 attach_edns_record(c->buffer, &edns); 1296 regional_free_all(worker->scratchpad); 1297 goto send_reply; 1298 } 1299 if(edns.udp_size < NORMAL_UDP_SIZE && 1300 worker->daemon->cfg->harden_short_bufsize) { 1301 verbose(VERB_QUERY, "worker request: EDNS bufsize %d ignored", 1302 (int)edns.udp_size); 1303 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen); 1304 edns.udp_size = NORMAL_UDP_SIZE; 1305 } 1306 if(c->type != comm_udp) { 1307 edns_opt = edns_opt_list_find(edns.opt_list, LDNS_EDNS_KEEPALIVE); 1308 if(edns_opt && edns_opt->opt_len > 0) { 1309 edns.ext_rcode = 0; 1310 edns.edns_version = EDNS_ADVERTISED_VERSION; 1311 edns.udp_size = EDNS_ADVERTISED_SIZE; 1312 edns.bits &= EDNS_DO; 1313 edns.opt_list = NULL; 1314 verbose(VERB_ALGO, "query with bad edns keepalive."); 1315 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen); 1316 error_encode(c->buffer, LDNS_RCODE_FORMERR, &qinfo, 1317 *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 1318 sldns_buffer_read_u16_at(c->buffer, 2), NULL); 1319 if(sldns_buffer_capacity(c->buffer) >= 1320 sldns_buffer_limit(c->buffer)+calc_edns_field_size(&edns)) 1321 attach_edns_record(c->buffer, &edns); 1322 regional_free_all(worker->scratchpad); 1323 goto send_reply; 1324 } 1325 } 1326 } 1327 if(edns.udp_size > worker->daemon->cfg->max_udp_size && 1328 c->type == comm_udp) { 1329 verbose(VERB_QUERY, 1330 "worker request: max UDP reply size modified" 1331 " (%d to max-udp-size)", (int)edns.udp_size); 1332 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen); 1333 edns.udp_size = worker->daemon->cfg->max_udp_size; 1334 } 1335 if(edns.udp_size < LDNS_HEADER_SIZE) { 1336 verbose(VERB_ALGO, "worker request: edns is too small."); 1337 log_addr(VERB_CLIENT, "from", &repinfo->addr, repinfo->addrlen); 1338 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1339 LDNS_TC_SET(sldns_buffer_begin(c->buffer)); 1340 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1341 LDNS_RCODE_SERVFAIL); 1342 sldns_buffer_set_position(c->buffer, LDNS_HEADER_SIZE); 1343 sldns_buffer_write_at(c->buffer, 4, 1344 (uint8_t*)"\0\0\0\0\0\0\0\0", 8); 1345 sldns_buffer_flip(c->buffer); 1346 regional_free_all(worker->scratchpad); 1347 goto send_reply; 1348 } 1349 if(worker->stats.extended) 1350 server_stats_insquery(&worker->stats, c, qinfo.qtype, 1351 qinfo.qclass, &edns, repinfo); 1352 if(c->type != comm_udp) 1353 edns.udp_size = 65535; /* max size for TCP replies */ 1354 if(qinfo.qclass == LDNS_RR_CLASS_CH && answer_chaos(worker, &qinfo, 1355 &edns, repinfo, c->buffer)) { 1356 server_stats_insrcode(&worker->stats, c->buffer); 1357 regional_free_all(worker->scratchpad); 1358 goto send_reply; 1359 } 1360 if(LDNS_OPCODE_WIRE(sldns_buffer_begin(c->buffer)) == 1361 LDNS_PACKET_NOTIFY) { 1362 answer_notify(worker, &qinfo, &edns, c->buffer, repinfo); 1363 regional_free_all(worker->scratchpad); 1364 goto send_reply; 1365 } 1366 if(local_zones_answer(worker->daemon->local_zones, &worker->env, &qinfo, 1367 &edns, c->buffer, worker->scratchpad, repinfo, acladdr->taglist, 1368 acladdr->taglen, acladdr->tag_actions, 1369 acladdr->tag_actions_size, acladdr->tag_datas, 1370 acladdr->tag_datas_size, worker->daemon->cfg->tagname, 1371 worker->daemon->cfg->num_tags, acladdr->view)) { 1372 regional_free_all(worker->scratchpad); 1373 if(sldns_buffer_limit(c->buffer) == 0) { 1374 comm_point_drop_reply(repinfo); 1375 return 0; 1376 } 1377 server_stats_insrcode(&worker->stats, c->buffer); 1378 goto send_reply; 1379 } 1380 if(worker->env.auth_zones && 1381 rpz_apply_qname_trigger(worker->env.auth_zones, 1382 &worker->env, &qinfo, &edns, c->buffer, worker->scratchpad, 1383 repinfo, acladdr->taglist, acladdr->taglen, &worker->stats)) { 1384 regional_free_all(worker->scratchpad); 1385 if(sldns_buffer_limit(c->buffer) == 0) { 1386 comm_point_drop_reply(repinfo); 1387 return 0; 1388 } 1389 server_stats_insrcode(&worker->stats, c->buffer); 1390 goto send_reply; 1391 } 1392 if(worker->env.auth_zones && 1393 auth_zones_answer(worker->env.auth_zones, &worker->env, 1394 &qinfo, &edns, repinfo, c->buffer, worker->scratchpad)) { 1395 regional_free_all(worker->scratchpad); 1396 if(sldns_buffer_limit(c->buffer) == 0) { 1397 comm_point_drop_reply(repinfo); 1398 return 0; 1399 } 1400 /* set RA for everyone that can have recursion (based on 1401 * access control list) */ 1402 if(LDNS_RD_WIRE(sldns_buffer_begin(c->buffer)) && 1403 acl != acl_deny_non_local && acl != acl_refuse_non_local) 1404 LDNS_RA_SET(sldns_buffer_begin(c->buffer)); 1405 server_stats_insrcode(&worker->stats, c->buffer); 1406 goto send_reply; 1407 } 1408 1409 /* We've looked in our local zones. If the answer isn't there, we 1410 * might need to bail out based on ACLs now. */ 1411 if((ret=deny_refuse_non_local(c, acl, worker, repinfo)) != -1) 1412 { 1413 regional_free_all(worker->scratchpad); 1414 if(ret == 1) 1415 goto send_reply; 1416 return ret; 1417 } 1418 1419 /* If this request does not have the recursion bit set, verify 1420 * ACLs allow the recursion bit to be treated as set. */ 1421 if(!(LDNS_RD_WIRE(sldns_buffer_begin(c->buffer))) && 1422 acl == acl_allow_setrd ) { 1423 LDNS_RD_SET(sldns_buffer_begin(c->buffer)); 1424 } 1425 1426 /* If this request does not have the recursion bit set, verify 1427 * ACLs allow the snooping. */ 1428 if(!(LDNS_RD_WIRE(sldns_buffer_begin(c->buffer))) && 1429 acl != acl_allow_snoop ) { 1430 error_encode(c->buffer, LDNS_RCODE_REFUSED, &qinfo, 1431 *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 1432 sldns_buffer_read_u16_at(c->buffer, 2), NULL); 1433 regional_free_all(worker->scratchpad); 1434 server_stats_insrcode(&worker->stats, c->buffer); 1435 log_addr(VERB_ALGO, "refused nonrec (cache snoop) query from", 1436 &repinfo->addr, repinfo->addrlen); 1437 goto send_reply; 1438 } 1439 1440 /* If we've found a local alias, replace the qname with the alias 1441 * target before resolving it. */ 1442 if(qinfo.local_alias) { 1443 struct ub_packed_rrset_key* rrset = qinfo.local_alias->rrset; 1444 struct packed_rrset_data* d = rrset->entry.data; 1445 1446 /* Sanity check: our current implementation only supports 1447 * a single CNAME RRset as a local alias. */ 1448 if(qinfo.local_alias->next || 1449 rrset->rk.type != htons(LDNS_RR_TYPE_CNAME) || 1450 d->count != 1) { 1451 log_err("assumption failure: unexpected local alias"); 1452 regional_free_all(worker->scratchpad); 1453 return 0; /* drop it */ 1454 } 1455 qinfo.qname = d->rr_data[0] + 2; 1456 qinfo.qname_len = d->rr_len[0] - 2; 1457 } 1458 1459 /* If we may apply IP-based actions to the answer, build the client 1460 * information. As this can be expensive, skip it if there is 1461 * absolutely no possibility of it. */ 1462 if((worker->daemon->use_response_ip || worker->daemon->use_rpz) && 1463 (qinfo.qtype == LDNS_RR_TYPE_A || 1464 qinfo.qtype == LDNS_RR_TYPE_AAAA || 1465 qinfo.qtype == LDNS_RR_TYPE_ANY)) { 1466 cinfo_tmp.taglist = acladdr->taglist; 1467 cinfo_tmp.taglen = acladdr->taglen; 1468 cinfo_tmp.tag_actions = acladdr->tag_actions; 1469 cinfo_tmp.tag_actions_size = acladdr->tag_actions_size; 1470 cinfo_tmp.tag_datas = acladdr->tag_datas; 1471 cinfo_tmp.tag_datas_size = acladdr->tag_datas_size; 1472 cinfo_tmp.view = acladdr->view; 1473 cinfo_tmp.respip_set = worker->daemon->respip_set; 1474 cinfo = &cinfo_tmp; 1475 } 1476 1477 lookup_cache: 1478 /* Lookup the cache. In case we chase an intermediate CNAME chain 1479 * this is a two-pass operation, and lookup_qinfo is different for 1480 * each pass. We should still pass the original qinfo to 1481 * answer_from_cache(), however, since it's used to build the reply. */ 1482 if(!edns_bypass_cache_stage(edns.opt_list, &worker->env)) { 1483 is_expired_answer = 0; 1484 is_secure_answer = 0; 1485 h = query_info_hash(lookup_qinfo, sldns_buffer_read_u16_at(c->buffer, 2)); 1486 if((e=slabhash_lookup(worker->env.msg_cache, h, lookup_qinfo, 0))) { 1487 /* answer from cache - we have acquired a readlock on it */ 1488 if(answer_from_cache(worker, &qinfo, 1489 cinfo, &need_drop, &is_expired_answer, &is_secure_answer, 1490 &alias_rrset, &partial_rep, (struct reply_info*)e->data, 1491 *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 1492 sldns_buffer_read_u16_at(c->buffer, 2), repinfo, 1493 &edns)) { 1494 /* prefetch it if the prefetch TTL expired. 1495 * Note that if there is more than one pass 1496 * its qname must be that used for cache 1497 * lookup. */ 1498 if((worker->env.cfg->prefetch && *worker->env.now >= 1499 ((struct reply_info*)e->data)->prefetch_ttl) || 1500 (worker->env.cfg->serve_expired && 1501 *worker->env.now >= ((struct reply_info*)e->data)->ttl)) { 1502 1503 time_t leeway = ((struct reply_info*)e-> 1504 data)->ttl - *worker->env.now; 1505 if(((struct reply_info*)e->data)->ttl 1506 < *worker->env.now) 1507 leeway = 0; 1508 lock_rw_unlock(&e->lock); 1509 reply_and_prefetch(worker, lookup_qinfo, 1510 sldns_buffer_read_u16_at(c->buffer, 2), 1511 repinfo, leeway, 1512 (partial_rep || need_drop)); 1513 if(!partial_rep) { 1514 rc = 0; 1515 regional_free_all(worker->scratchpad); 1516 goto send_reply_rc; 1517 } 1518 } else if(!partial_rep) { 1519 lock_rw_unlock(&e->lock); 1520 regional_free_all(worker->scratchpad); 1521 goto send_reply; 1522 } else { 1523 /* Note that we've already released the 1524 * lock if we're here after prefetch. */ 1525 lock_rw_unlock(&e->lock); 1526 } 1527 /* We've found a partial reply ending with an 1528 * alias. Replace the lookup qinfo for the 1529 * alias target and lookup the cache again to 1530 * (possibly) complete the reply. As we're 1531 * passing the "base" reply, there will be no 1532 * more alias chasing. */ 1533 memset(&qinfo_tmp, 0, sizeof(qinfo_tmp)); 1534 get_cname_target(alias_rrset, &qinfo_tmp.qname, 1535 &qinfo_tmp.qname_len); 1536 if(!qinfo_tmp.qname) { 1537 log_err("unexpected: invalid answer alias"); 1538 regional_free_all(worker->scratchpad); 1539 return 0; /* drop query */ 1540 } 1541 qinfo_tmp.qtype = qinfo.qtype; 1542 qinfo_tmp.qclass = qinfo.qclass; 1543 lookup_qinfo = &qinfo_tmp; 1544 goto lookup_cache; 1545 } 1546 verbose(VERB_ALGO, "answer from the cache failed"); 1547 lock_rw_unlock(&e->lock); 1548 } 1549 if(!LDNS_RD_WIRE(sldns_buffer_begin(c->buffer))) { 1550 if(answer_norec_from_cache(worker, &qinfo, 1551 *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 1552 sldns_buffer_read_u16_at(c->buffer, 2), repinfo, 1553 &edns)) { 1554 regional_free_all(worker->scratchpad); 1555 goto send_reply; 1556 } 1557 verbose(VERB_ALGO, "answer norec from cache -- " 1558 "need to validate or not primed"); 1559 } 1560 } 1561 sldns_buffer_rewind(c->buffer); 1562 server_stats_querymiss(&worker->stats, worker); 1563 1564 if(verbosity >= VERB_CLIENT) { 1565 if(c->type == comm_udp) 1566 log_addr(VERB_CLIENT, "udp request from", 1567 &repinfo->addr, repinfo->addrlen); 1568 else log_addr(VERB_CLIENT, "tcp request from", 1569 &repinfo->addr, repinfo->addrlen); 1570 } 1571 1572 /* grab a work request structure for this new request */ 1573 mesh_new_client(worker->env.mesh, &qinfo, cinfo, 1574 sldns_buffer_read_u16_at(c->buffer, 2), 1575 &edns, repinfo, *(uint16_t*)(void *)sldns_buffer_begin(c->buffer)); 1576 regional_free_all(worker->scratchpad); 1577 worker_mem_report(worker, NULL); 1578 return 0; 1579 1580 send_reply: 1581 rc = 1; 1582 send_reply_rc: 1583 if(need_drop) { 1584 comm_point_drop_reply(repinfo); 1585 return 0; 1586 } 1587 if(is_expired_answer) { 1588 worker->stats.ans_expired++; 1589 } 1590 if(worker->stats.extended) { 1591 if(is_secure_answer) worker->stats.ans_secure++; 1592 server_stats_insrcode(&worker->stats, repinfo->c->buffer); 1593 } 1594 #ifdef USE_DNSTAP 1595 if(worker->dtenv.log_client_response_messages) 1596 dt_msg_send_client_response(&worker->dtenv, &repinfo->addr, 1597 c->type, c->buffer); 1598 #endif 1599 if(worker->env.cfg->log_replies) 1600 { 1601 struct timeval tv; 1602 memset(&tv, 0, sizeof(tv)); 1603 if(qinfo.local_alias && qinfo.local_alias->rrset && 1604 qinfo.local_alias->rrset->rk.dname) { 1605 /* log original qname, before the local alias was 1606 * used to resolve that CNAME to something else */ 1607 qinfo.qname = qinfo.local_alias->rrset->rk.dname; 1608 log_reply_info(NO_VERBOSE, &qinfo, &repinfo->addr, repinfo->addrlen, 1609 tv, 1, c->buffer); 1610 } else { 1611 log_reply_info(NO_VERBOSE, &qinfo, &repinfo->addr, repinfo->addrlen, 1612 tv, 1, c->buffer); 1613 } 1614 } 1615 #ifdef USE_DNSCRYPT 1616 if(!dnsc_handle_uncurved_request(repinfo)) { 1617 return 0; 1618 } 1619 #endif 1620 return rc; 1621 } 1622 1623 void 1624 worker_sighandler(int sig, void* arg) 1625 { 1626 /* note that log, print, syscalls here give race conditions. 1627 * And cause hangups if the log-lock is held by the application. */ 1628 struct worker* worker = (struct worker*)arg; 1629 switch(sig) { 1630 #ifdef SIGHUP 1631 case SIGHUP: 1632 comm_base_exit(worker->base); 1633 break; 1634 #endif 1635 case SIGINT: 1636 worker->need_to_exit = 1; 1637 comm_base_exit(worker->base); 1638 break; 1639 #ifdef SIGQUIT 1640 case SIGQUIT: 1641 worker->need_to_exit = 1; 1642 comm_base_exit(worker->base); 1643 break; 1644 #endif 1645 case SIGTERM: 1646 worker->need_to_exit = 1; 1647 comm_base_exit(worker->base); 1648 break; 1649 default: 1650 /* unknown signal, ignored */ 1651 break; 1652 } 1653 } 1654 1655 /** restart statistics timer for worker, if enabled */ 1656 static void 1657 worker_restart_timer(struct worker* worker) 1658 { 1659 if(worker->env.cfg->stat_interval > 0) { 1660 struct timeval tv; 1661 #ifndef S_SPLINT_S 1662 tv.tv_sec = worker->env.cfg->stat_interval; 1663 tv.tv_usec = 0; 1664 #endif 1665 comm_timer_set(worker->stat_timer, &tv); 1666 } 1667 } 1668 1669 void worker_stat_timer_cb(void* arg) 1670 { 1671 struct worker* worker = (struct worker*)arg; 1672 server_stats_log(&worker->stats, worker, worker->thread_num); 1673 mesh_stats(worker->env.mesh, "mesh has"); 1674 worker_mem_report(worker, NULL); 1675 /* SHM is enabled, process data to SHM */ 1676 if (worker->daemon->cfg->shm_enable) { 1677 shm_main_run(worker); 1678 } 1679 if(!worker->daemon->cfg->stat_cumulative) { 1680 worker_stats_clear(worker); 1681 } 1682 /* start next timer */ 1683 worker_restart_timer(worker); 1684 } 1685 1686 void worker_probe_timer_cb(void* arg) 1687 { 1688 struct worker* worker = (struct worker*)arg; 1689 struct timeval tv; 1690 #ifndef S_SPLINT_S 1691 tv.tv_sec = (time_t)autr_probe_timer(&worker->env); 1692 tv.tv_usec = 0; 1693 #endif 1694 if(tv.tv_sec != 0) 1695 comm_timer_set(worker->env.probe_timer, &tv); 1696 } 1697 1698 struct worker* 1699 worker_create(struct daemon* daemon, int id, int* ports, int n) 1700 { 1701 unsigned int seed; 1702 struct worker* worker = (struct worker*)calloc(1, 1703 sizeof(struct worker)); 1704 if(!worker) 1705 return NULL; 1706 worker->numports = n; 1707 worker->ports = (int*)memdup(ports, sizeof(int)*n); 1708 if(!worker->ports) { 1709 free(worker); 1710 return NULL; 1711 } 1712 worker->daemon = daemon; 1713 worker->thread_num = id; 1714 if(!(worker->cmd = tube_create())) { 1715 free(worker->ports); 1716 free(worker); 1717 return NULL; 1718 } 1719 /* create random state here to avoid locking trouble in RAND_bytes */ 1720 if(!(worker->rndstate = ub_initstate(daemon->rand))) { 1721 log_err("could not init random numbers."); 1722 tube_delete(worker->cmd); 1723 free(worker->ports); 1724 free(worker); 1725 return NULL; 1726 } 1727 explicit_bzero(&seed, sizeof(seed)); 1728 #ifdef USE_DNSTAP 1729 if(daemon->cfg->dnstap) { 1730 log_assert(daemon->dtenv != NULL); 1731 memcpy(&worker->dtenv, daemon->dtenv, sizeof(struct dt_env)); 1732 if(!dt_init(&worker->dtenv)) 1733 fatal_exit("dt_init failed"); 1734 } 1735 #endif 1736 return worker; 1737 } 1738 1739 int 1740 worker_init(struct worker* worker, struct config_file *cfg, 1741 struct listen_port* ports, int do_sigs) 1742 { 1743 #ifdef USE_DNSTAP 1744 struct dt_env* dtenv = &worker->dtenv; 1745 #else 1746 void* dtenv = NULL; 1747 #endif 1748 worker->need_to_exit = 0; 1749 worker->base = comm_base_create(do_sigs); 1750 if(!worker->base) { 1751 log_err("could not create event handling base"); 1752 worker_delete(worker); 1753 return 0; 1754 } 1755 comm_base_set_slow_accept_handlers(worker->base, &worker_stop_accept, 1756 &worker_start_accept, worker); 1757 if(do_sigs) { 1758 #ifdef SIGHUP 1759 ub_thread_sig_unblock(SIGHUP); 1760 #endif 1761 ub_thread_sig_unblock(SIGINT); 1762 #ifdef SIGQUIT 1763 ub_thread_sig_unblock(SIGQUIT); 1764 #endif 1765 ub_thread_sig_unblock(SIGTERM); 1766 #ifndef LIBEVENT_SIGNAL_PROBLEM 1767 worker->comsig = comm_signal_create(worker->base, 1768 worker_sighandler, worker); 1769 if(!worker->comsig 1770 #ifdef SIGHUP 1771 || !comm_signal_bind(worker->comsig, SIGHUP) 1772 #endif 1773 #ifdef SIGQUIT 1774 || !comm_signal_bind(worker->comsig, SIGQUIT) 1775 #endif 1776 || !comm_signal_bind(worker->comsig, SIGTERM) 1777 || !comm_signal_bind(worker->comsig, SIGINT)) { 1778 log_err("could not create signal handlers"); 1779 worker_delete(worker); 1780 return 0; 1781 } 1782 #endif /* LIBEVENT_SIGNAL_PROBLEM */ 1783 if(!daemon_remote_open_accept(worker->daemon->rc, 1784 worker->daemon->rc_ports, worker)) { 1785 worker_delete(worker); 1786 return 0; 1787 } 1788 #ifdef UB_ON_WINDOWS 1789 wsvc_setup_worker(worker); 1790 #endif /* UB_ON_WINDOWS */ 1791 } else { /* !do_sigs */ 1792 worker->comsig = NULL; 1793 } 1794 worker->front = listen_create(worker->base, ports, 1795 cfg->msg_buffer_size, (int)cfg->incoming_num_tcp, 1796 cfg->do_tcp_keepalive 1797 ? cfg->tcp_keepalive_timeout 1798 : cfg->tcp_idle_timeout, 1799 worker->daemon->tcl, 1800 worker->daemon->listen_sslctx, 1801 dtenv, worker_handle_request, worker); 1802 if(!worker->front) { 1803 log_err("could not create listening sockets"); 1804 worker_delete(worker); 1805 return 0; 1806 } 1807 worker->back = outside_network_create(worker->base, 1808 cfg->msg_buffer_size, (size_t)cfg->outgoing_num_ports, 1809 cfg->out_ifs, cfg->num_out_ifs, cfg->do_ip4, cfg->do_ip6, 1810 cfg->do_tcp?cfg->outgoing_num_tcp:0, 1811 worker->daemon->env->infra_cache, worker->rndstate, 1812 cfg->use_caps_bits_for_id, worker->ports, worker->numports, 1813 cfg->unwanted_threshold, cfg->outgoing_tcp_mss, 1814 &worker_alloc_cleanup, worker, 1815 cfg->do_udp || cfg->udp_upstream_without_downstream, 1816 worker->daemon->connect_sslctx, cfg->delay_close, 1817 dtenv); 1818 if(!worker->back) { 1819 log_err("could not create outgoing sockets"); 1820 worker_delete(worker); 1821 return 0; 1822 } 1823 /* start listening to commands */ 1824 if(!tube_setup_bg_listen(worker->cmd, worker->base, 1825 &worker_handle_control_cmd, worker)) { 1826 log_err("could not create control compt."); 1827 worker_delete(worker); 1828 return 0; 1829 } 1830 worker->stat_timer = comm_timer_create(worker->base, 1831 worker_stat_timer_cb, worker); 1832 if(!worker->stat_timer) { 1833 log_err("could not create statistics timer"); 1834 } 1835 1836 /* we use the msg_buffer_size as a good estimate for what the 1837 * user wants for memory usage sizes */ 1838 worker->scratchpad = regional_create_custom(cfg->msg_buffer_size); 1839 if(!worker->scratchpad) { 1840 log_err("malloc failure"); 1841 worker_delete(worker); 1842 return 0; 1843 } 1844 1845 server_stats_init(&worker->stats, cfg); 1846 alloc_init(&worker->alloc, &worker->daemon->superalloc, 1847 worker->thread_num); 1848 alloc_set_id_cleanup(&worker->alloc, &worker_alloc_cleanup, worker); 1849 worker->env = *worker->daemon->env; 1850 comm_base_timept(worker->base, &worker->env.now, &worker->env.now_tv); 1851 worker->env.worker = worker; 1852 worker->env.worker_base = worker->base; 1853 worker->env.send_query = &worker_send_query; 1854 worker->env.alloc = &worker->alloc; 1855 worker->env.outnet = worker->back; 1856 worker->env.rnd = worker->rndstate; 1857 /* If case prefetch is triggered, the corresponding mesh will clear 1858 * the scratchpad for the module env in the middle of request handling. 1859 * It would be prone to a use-after-free kind of bug, so we avoid 1860 * sharing it with worker's own scratchpad at the cost of having 1861 * one more pad per worker. */ 1862 worker->env.scratch = regional_create_custom(cfg->msg_buffer_size); 1863 if(!worker->env.scratch) { 1864 log_err("malloc failure"); 1865 worker_delete(worker); 1866 return 0; 1867 } 1868 worker->env.mesh = mesh_create(&worker->daemon->mods, &worker->env); 1869 /* Pass on daemon variables that we would need in the mesh area */ 1870 worker->env.mesh->use_response_ip = worker->daemon->use_response_ip; 1871 worker->env.mesh->use_rpz = worker->daemon->use_rpz; 1872 1873 worker->env.detach_subs = &mesh_detach_subs; 1874 worker->env.attach_sub = &mesh_attach_sub; 1875 worker->env.add_sub = &mesh_add_sub; 1876 worker->env.kill_sub = &mesh_state_delete; 1877 worker->env.detect_cycle = &mesh_detect_cycle; 1878 worker->env.scratch_buffer = sldns_buffer_new(cfg->msg_buffer_size); 1879 if(!(worker->env.fwds = forwards_create()) || 1880 !forwards_apply_cfg(worker->env.fwds, cfg)) { 1881 log_err("Could not set forward zones"); 1882 worker_delete(worker); 1883 return 0; 1884 } 1885 if(!(worker->env.hints = hints_create()) || 1886 !hints_apply_cfg(worker->env.hints, cfg)) { 1887 log_err("Could not set root or stub hints"); 1888 worker_delete(worker); 1889 return 0; 1890 } 1891 /* one probe timer per process -- if we have 5011 anchors */ 1892 if(autr_get_num_anchors(worker->env.anchors) > 0 1893 #ifndef THREADS_DISABLED 1894 && worker->thread_num == 0 1895 #endif 1896 ) { 1897 struct timeval tv; 1898 tv.tv_sec = 0; 1899 tv.tv_usec = 0; 1900 worker->env.probe_timer = comm_timer_create(worker->base, 1901 worker_probe_timer_cb, worker); 1902 if(!worker->env.probe_timer) { 1903 log_err("could not create 5011-probe timer"); 1904 } else { 1905 /* let timer fire, then it can reset itself */ 1906 comm_timer_set(worker->env.probe_timer, &tv); 1907 } 1908 } 1909 /* zone transfer tasks, setup once per process, if any */ 1910 if(worker->env.auth_zones 1911 #ifndef THREADS_DISABLED 1912 && worker->thread_num == 0 1913 #endif 1914 ) { 1915 auth_xfer_pickup_initial(worker->env.auth_zones, &worker->env); 1916 } 1917 if(!worker->env.mesh || !worker->env.scratch_buffer) { 1918 worker_delete(worker); 1919 return 0; 1920 } 1921 worker_mem_report(worker, NULL); 1922 /* if statistics enabled start timer */ 1923 if(worker->env.cfg->stat_interval > 0) { 1924 verbose(VERB_ALGO, "set statistics interval %d secs", 1925 worker->env.cfg->stat_interval); 1926 worker_restart_timer(worker); 1927 } 1928 return 1; 1929 } 1930 1931 void 1932 worker_work(struct worker* worker) 1933 { 1934 comm_base_dispatch(worker->base); 1935 } 1936 1937 void 1938 worker_delete(struct worker* worker) 1939 { 1940 if(!worker) 1941 return; 1942 if(worker->env.mesh && verbosity >= VERB_OPS) { 1943 server_stats_log(&worker->stats, worker, worker->thread_num); 1944 mesh_stats(worker->env.mesh, "mesh has"); 1945 worker_mem_report(worker, NULL); 1946 } 1947 outside_network_quit_prepare(worker->back); 1948 mesh_delete(worker->env.mesh); 1949 sldns_buffer_free(worker->env.scratch_buffer); 1950 forwards_delete(worker->env.fwds); 1951 hints_delete(worker->env.hints); 1952 listen_delete(worker->front); 1953 outside_network_delete(worker->back); 1954 comm_signal_delete(worker->comsig); 1955 tube_delete(worker->cmd); 1956 comm_timer_delete(worker->stat_timer); 1957 comm_timer_delete(worker->env.probe_timer); 1958 free(worker->ports); 1959 if(worker->thread_num == 0) { 1960 #ifdef UB_ON_WINDOWS 1961 wsvc_desetup_worker(worker); 1962 #endif /* UB_ON_WINDOWS */ 1963 } 1964 comm_base_delete(worker->base); 1965 ub_randfree(worker->rndstate); 1966 alloc_clear(&worker->alloc); 1967 regional_destroy(worker->env.scratch); 1968 regional_destroy(worker->scratchpad); 1969 free(worker); 1970 } 1971 1972 struct outbound_entry* 1973 worker_send_query(struct query_info* qinfo, uint16_t flags, int dnssec, 1974 int want_dnssec, int nocaps, struct sockaddr_storage* addr, 1975 socklen_t addrlen, uint8_t* zone, size_t zonelen, int ssl_upstream, 1976 char* tls_auth_name, struct module_qstate* q) 1977 { 1978 struct worker* worker = q->env->worker; 1979 struct outbound_entry* e = (struct outbound_entry*)regional_alloc( 1980 q->region, sizeof(*e)); 1981 if(!e) 1982 return NULL; 1983 e->qstate = q; 1984 e->qsent = outnet_serviced_query(worker->back, qinfo, flags, dnssec, 1985 want_dnssec, nocaps, q->env->cfg->tcp_upstream, 1986 ssl_upstream, tls_auth_name, addr, addrlen, zone, zonelen, q, 1987 worker_handle_service_reply, e, worker->back->udp_buff, q->env); 1988 if(!e->qsent) { 1989 return NULL; 1990 } 1991 return e; 1992 } 1993 1994 void 1995 worker_alloc_cleanup(void* arg) 1996 { 1997 struct worker* worker = (struct worker*)arg; 1998 slabhash_clear(&worker->env.rrset_cache->table); 1999 slabhash_clear(worker->env.msg_cache); 2000 } 2001 2002 void worker_stats_clear(struct worker* worker) 2003 { 2004 server_stats_init(&worker->stats, worker->env.cfg); 2005 mesh_stats_clear(worker->env.mesh); 2006 worker->back->unwanted_replies = 0; 2007 worker->back->num_tcp_outgoing = 0; 2008 } 2009 2010 void worker_start_accept(void* arg) 2011 { 2012 struct worker* worker = (struct worker*)arg; 2013 listen_start_accept(worker->front); 2014 if(worker->thread_num == 0) 2015 daemon_remote_start_accept(worker->daemon->rc); 2016 } 2017 2018 void worker_stop_accept(void* arg) 2019 { 2020 struct worker* worker = (struct worker*)arg; 2021 listen_stop_accept(worker->front); 2022 if(worker->thread_num == 0) 2023 daemon_remote_stop_accept(worker->daemon->rc); 2024 } 2025 2026 /* --- fake callbacks for fptr_wlist to work --- */ 2027 struct outbound_entry* libworker_send_query( 2028 struct query_info* ATTR_UNUSED(qinfo), 2029 uint16_t ATTR_UNUSED(flags), int ATTR_UNUSED(dnssec), 2030 int ATTR_UNUSED(want_dnssec), int ATTR_UNUSED(nocaps), 2031 struct sockaddr_storage* ATTR_UNUSED(addr), socklen_t ATTR_UNUSED(addrlen), 2032 uint8_t* ATTR_UNUSED(zone), size_t ATTR_UNUSED(zonelen), 2033 int ATTR_UNUSED(ssl_upstream), char* ATTR_UNUSED(tls_auth_name), 2034 struct module_qstate* ATTR_UNUSED(q)) 2035 { 2036 log_assert(0); 2037 return 0; 2038 } 2039 2040 int libworker_handle_reply(struct comm_point* ATTR_UNUSED(c), 2041 void* ATTR_UNUSED(arg), int ATTR_UNUSED(error), 2042 struct comm_reply* ATTR_UNUSED(reply_info)) 2043 { 2044 log_assert(0); 2045 return 0; 2046 } 2047 2048 int libworker_handle_service_reply(struct comm_point* ATTR_UNUSED(c), 2049 void* ATTR_UNUSED(arg), int ATTR_UNUSED(error), 2050 struct comm_reply* ATTR_UNUSED(reply_info)) 2051 { 2052 log_assert(0); 2053 return 0; 2054 } 2055 2056 void libworker_handle_control_cmd(struct tube* ATTR_UNUSED(tube), 2057 uint8_t* ATTR_UNUSED(buffer), size_t ATTR_UNUSED(len), 2058 int ATTR_UNUSED(error), void* ATTR_UNUSED(arg)) 2059 { 2060 log_assert(0); 2061 } 2062 2063 void libworker_fg_done_cb(void* ATTR_UNUSED(arg), int ATTR_UNUSED(rcode), 2064 sldns_buffer* ATTR_UNUSED(buf), enum sec_status ATTR_UNUSED(s), 2065 char* ATTR_UNUSED(why_bogus), int ATTR_UNUSED(was_ratelimited)) 2066 { 2067 log_assert(0); 2068 } 2069 2070 void libworker_bg_done_cb(void* ATTR_UNUSED(arg), int ATTR_UNUSED(rcode), 2071 sldns_buffer* ATTR_UNUSED(buf), enum sec_status ATTR_UNUSED(s), 2072 char* ATTR_UNUSED(why_bogus), int ATTR_UNUSED(was_ratelimited)) 2073 { 2074 log_assert(0); 2075 } 2076 2077 void libworker_event_done_cb(void* ATTR_UNUSED(arg), int ATTR_UNUSED(rcode), 2078 sldns_buffer* ATTR_UNUSED(buf), enum sec_status ATTR_UNUSED(s), 2079 char* ATTR_UNUSED(why_bogus), int ATTR_UNUSED(was_ratelimited)) 2080 { 2081 log_assert(0); 2082 } 2083 2084 int context_query_cmp(const void* ATTR_UNUSED(a), const void* ATTR_UNUSED(b)) 2085 { 2086 log_assert(0); 2087 return 0; 2088 } 2089 2090 int order_lock_cmp(const void* ATTR_UNUSED(e1), const void* ATTR_UNUSED(e2)) 2091 { 2092 log_assert(0); 2093 return 0; 2094 } 2095 2096 int codeline_cmp(const void* ATTR_UNUSED(a), const void* ATTR_UNUSED(b)) 2097 { 2098 log_assert(0); 2099 return 0; 2100 } 2101 2102