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 "iterator/iter_utils.h" 74 #include "validator/autotrust.h" 75 #include "validator/val_anchor.h" 76 #include "respip/respip.h" 77 #include "libunbound/context.h" 78 #include "libunbound/libworker.h" 79 #include "sldns/sbuffer.h" 80 #include "sldns/wire2str.h" 81 #include "util/shm_side/shm_main.h" 82 #include "dnscrypt/dnscrypt.h" 83 #include "dnstap/dtstream.h" 84 85 #ifdef HAVE_SYS_TYPES_H 86 # include <sys/types.h> 87 #endif 88 #ifdef HAVE_NETDB_H 89 #include <netdb.h> 90 #endif 91 #include <signal.h> 92 #ifdef UB_ON_WINDOWS 93 #include "winrc/win_svc.h" 94 #endif 95 96 /** Size of an UDP datagram */ 97 #define NORMAL_UDP_SIZE 512 /* bytes */ 98 /** ratelimit for error responses */ 99 #define ERROR_RATELIMIT 100 /* qps */ 100 101 /** 102 * seconds to add to prefetch leeway. This is a TTL that expires old rrsets 103 * earlier than they should in order to put the new update into the cache. 104 * This additional value is to make sure that if not all TTLs are equal in 105 * the message to be updated(and replaced), that rrsets with up to this much 106 * extra TTL are also replaced. This means that the resulting new message 107 * will have (most likely) this TTL at least, avoiding very small 'split 108 * second' TTLs due to operators choosing relative primes for TTLs (or so). 109 * Also has to be at least one to break ties (and overwrite cached entry). 110 */ 111 #define PREFETCH_EXPIRY_ADD 60 112 113 /** Report on memory usage by this thread and global */ 114 static void 115 worker_mem_report(struct worker* ATTR_UNUSED(worker), 116 struct serviced_query* ATTR_UNUSED(cur_serv)) 117 { 118 #ifdef UNBOUND_ALLOC_STATS 119 /* measure memory leakage */ 120 extern size_t unbound_mem_alloc, unbound_mem_freed; 121 /* debug func in validator module */ 122 size_t total, front, back, mesh, msg, rrset, infra, ac, superac; 123 size_t me, iter, val, anch; 124 int i; 125 #ifdef CLIENT_SUBNET 126 size_t subnet = 0; 127 #endif /* CLIENT_SUBNET */ 128 if(verbosity < VERB_ALGO) 129 return; 130 front = listen_get_mem(worker->front); 131 back = outnet_get_mem(worker->back); 132 msg = slabhash_get_mem(worker->env.msg_cache); 133 rrset = slabhash_get_mem(&worker->env.rrset_cache->table); 134 infra = infra_get_mem(worker->env.infra_cache); 135 mesh = mesh_get_mem(worker->env.mesh); 136 ac = alloc_get_mem(&worker->alloc); 137 superac = alloc_get_mem(&worker->daemon->superalloc); 138 anch = anchors_get_mem(worker->env.anchors); 139 iter = 0; 140 val = 0; 141 for(i=0; i<worker->env.mesh->mods.num; i++) { 142 fptr_ok(fptr_whitelist_mod_get_mem(worker->env.mesh-> 143 mods.mod[i]->get_mem)); 144 if(strcmp(worker->env.mesh->mods.mod[i]->name, "validator")==0) 145 val += (*worker->env.mesh->mods.mod[i]->get_mem) 146 (&worker->env, i); 147 #ifdef CLIENT_SUBNET 148 else if(strcmp(worker->env.mesh->mods.mod[i]->name, 149 "subnet")==0) 150 subnet += (*worker->env.mesh->mods.mod[i]->get_mem) 151 (&worker->env, i); 152 #endif /* CLIENT_SUBNET */ 153 else iter += (*worker->env.mesh->mods.mod[i]->get_mem) 154 (&worker->env, i); 155 } 156 me = sizeof(*worker) + sizeof(*worker->base) + sizeof(*worker->comsig) 157 + comm_point_get_mem(worker->cmd_com) 158 + sizeof(worker->rndstate) 159 + regional_get_mem(worker->scratchpad) 160 + sizeof(*worker->env.scratch_buffer) 161 + sldns_buffer_capacity(worker->env.scratch_buffer) 162 + forwards_get_mem(worker->env.fwds) 163 + hints_get_mem(worker->env.hints); 164 if(worker->thread_num == 0) 165 me += acl_list_get_mem(worker->daemon->acl); 166 if(cur_serv) { 167 me += serviced_get_mem(cur_serv); 168 } 169 total = front+back+mesh+msg+rrset+infra+iter+val+ac+superac+me; 170 #ifdef CLIENT_SUBNET 171 total += subnet; 172 log_info("Memory conditions: %u front=%u back=%u mesh=%u msg=%u " 173 "rrset=%u infra=%u iter=%u val=%u subnet=%u anchors=%u " 174 "alloccache=%u globalalloccache=%u me=%u", 175 (unsigned)total, (unsigned)front, (unsigned)back, 176 (unsigned)mesh, (unsigned)msg, (unsigned)rrset, (unsigned)infra, 177 (unsigned)iter, (unsigned)val, 178 (unsigned)subnet, (unsigned)anch, (unsigned)ac, 179 (unsigned)superac, (unsigned)me); 180 #else /* no CLIENT_SUBNET */ 181 log_info("Memory conditions: %u front=%u back=%u mesh=%u msg=%u " 182 "rrset=%u infra=%u iter=%u val=%u anchors=%u " 183 "alloccache=%u globalalloccache=%u me=%u", 184 (unsigned)total, (unsigned)front, (unsigned)back, 185 (unsigned)mesh, (unsigned)msg, (unsigned)rrset, 186 (unsigned)infra, (unsigned)iter, (unsigned)val, (unsigned)anch, 187 (unsigned)ac, (unsigned)superac, (unsigned)me); 188 #endif /* CLIENT_SUBNET */ 189 log_info("Total heap memory estimate: %u total-alloc: %u " 190 "total-free: %u", (unsigned)total, 191 (unsigned)unbound_mem_alloc, (unsigned)unbound_mem_freed); 192 #else /* no UNBOUND_ALLOC_STATS */ 193 size_t val = 0; 194 #ifdef CLIENT_SUBNET 195 size_t subnet = 0; 196 #endif /* CLIENT_SUBNET */ 197 int i; 198 if(verbosity < VERB_QUERY) 199 return; 200 for(i=0; i<worker->env.mesh->mods.num; i++) { 201 fptr_ok(fptr_whitelist_mod_get_mem(worker->env.mesh-> 202 mods.mod[i]->get_mem)); 203 if(strcmp(worker->env.mesh->mods.mod[i]->name, "validator")==0) 204 val += (*worker->env.mesh->mods.mod[i]->get_mem) 205 (&worker->env, i); 206 #ifdef CLIENT_SUBNET 207 else if(strcmp(worker->env.mesh->mods.mod[i]->name, 208 "subnet")==0) 209 subnet += (*worker->env.mesh->mods.mod[i]->get_mem) 210 (&worker->env, i); 211 #endif /* CLIENT_SUBNET */ 212 } 213 #ifdef CLIENT_SUBNET 214 verbose(VERB_QUERY, "cache memory msg=%u rrset=%u infra=%u val=%u " 215 "subnet=%u", 216 (unsigned)slabhash_get_mem(worker->env.msg_cache), 217 (unsigned)slabhash_get_mem(&worker->env.rrset_cache->table), 218 (unsigned)infra_get_mem(worker->env.infra_cache), 219 (unsigned)val, (unsigned)subnet); 220 #else /* no CLIENT_SUBNET */ 221 verbose(VERB_QUERY, "cache memory msg=%u rrset=%u infra=%u val=%u", 222 (unsigned)slabhash_get_mem(worker->env.msg_cache), 223 (unsigned)slabhash_get_mem(&worker->env.rrset_cache->table), 224 (unsigned)infra_get_mem(worker->env.infra_cache), 225 (unsigned)val); 226 #endif /* CLIENT_SUBNET */ 227 #endif /* UNBOUND_ALLOC_STATS */ 228 } 229 230 void 231 worker_send_cmd(struct worker* worker, enum worker_commands cmd) 232 { 233 uint32_t c = (uint32_t)htonl(cmd); 234 if(!tube_write_msg(worker->cmd, (uint8_t*)&c, sizeof(c), 0)) { 235 log_err("worker send cmd %d failed", (int)cmd); 236 } 237 } 238 239 int 240 worker_handle_service_reply(struct comm_point* c, void* arg, int error, 241 struct comm_reply* reply_info) 242 { 243 struct outbound_entry* e = (struct outbound_entry*)arg; 244 struct worker* worker = e->qstate->env->worker; 245 struct serviced_query *sq = e->qsent; 246 247 verbose(VERB_ALGO, "worker svcd callback for qstate %p", e->qstate); 248 if(error != 0) { 249 mesh_report_reply(worker->env.mesh, e, reply_info, error); 250 worker_mem_report(worker, sq); 251 return 0; 252 } 253 /* sanity check. */ 254 if(!LDNS_QR_WIRE(sldns_buffer_begin(c->buffer)) 255 || LDNS_OPCODE_WIRE(sldns_buffer_begin(c->buffer)) != 256 LDNS_PACKET_QUERY 257 || LDNS_QDCOUNT(sldns_buffer_begin(c->buffer)) > 1) { 258 /* error becomes timeout for the module as if this reply 259 * never arrived. */ 260 verbose(VERB_ALGO, "worker: bad reply handled as timeout"); 261 mesh_report_reply(worker->env.mesh, e, reply_info, 262 NETEVENT_TIMEOUT); 263 worker_mem_report(worker, sq); 264 return 0; 265 } 266 mesh_report_reply(worker->env.mesh, e, reply_info, NETEVENT_NOERROR); 267 worker_mem_report(worker, sq); 268 return 0; 269 } 270 271 /** ratelimit error replies 272 * @param worker: the worker struct with ratelimit counter 273 * @param err: error code that would be wanted. 274 * @return value of err if okay, or -1 if it should be discarded instead. 275 */ 276 static int 277 worker_err_ratelimit(struct worker* worker, int err) 278 { 279 if(worker->err_limit_time == *worker->env.now) { 280 /* see if limit is exceeded for this second */ 281 if(worker->err_limit_count++ > ERROR_RATELIMIT) 282 return -1; 283 } else { 284 /* new second, new limits */ 285 worker->err_limit_time = *worker->env.now; 286 worker->err_limit_count = 1; 287 } 288 return err; 289 } 290 291 /** check request sanity. 292 * @param pkt: the wire packet to examine for sanity. 293 * @param worker: parameters for checking. 294 * @return error code, 0 OK, or -1 discard. 295 */ 296 static int 297 worker_check_request(sldns_buffer* pkt, struct worker* worker) 298 { 299 if(sldns_buffer_limit(pkt) < LDNS_HEADER_SIZE) { 300 verbose(VERB_QUERY, "request too short, discarded"); 301 return -1; 302 } 303 if(sldns_buffer_limit(pkt) > NORMAL_UDP_SIZE && 304 worker->daemon->cfg->harden_large_queries) { 305 verbose(VERB_QUERY, "request too large, discarded"); 306 return -1; 307 } 308 if(LDNS_QR_WIRE(sldns_buffer_begin(pkt))) { 309 verbose(VERB_QUERY, "request has QR bit on, discarded"); 310 return -1; 311 } 312 if(LDNS_TC_WIRE(sldns_buffer_begin(pkt))) { 313 LDNS_TC_CLR(sldns_buffer_begin(pkt)); 314 verbose(VERB_QUERY, "request bad, has TC bit on"); 315 return worker_err_ratelimit(worker, LDNS_RCODE_FORMERR); 316 } 317 if(LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt)) != LDNS_PACKET_QUERY && 318 LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt)) != LDNS_PACKET_NOTIFY) { 319 verbose(VERB_QUERY, "request unknown opcode %d", 320 LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt))); 321 return worker_err_ratelimit(worker, LDNS_RCODE_NOTIMPL); 322 } 323 if(LDNS_QDCOUNT(sldns_buffer_begin(pkt)) != 1) { 324 verbose(VERB_QUERY, "request wrong nr qd=%d", 325 LDNS_QDCOUNT(sldns_buffer_begin(pkt))); 326 return worker_err_ratelimit(worker, LDNS_RCODE_FORMERR); 327 } 328 if(LDNS_ANCOUNT(sldns_buffer_begin(pkt)) != 0 && 329 (LDNS_ANCOUNT(sldns_buffer_begin(pkt)) != 1 || 330 LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt)) != LDNS_PACKET_NOTIFY)) { 331 verbose(VERB_QUERY, "request wrong nr an=%d", 332 LDNS_ANCOUNT(sldns_buffer_begin(pkt))); 333 return worker_err_ratelimit(worker, LDNS_RCODE_FORMERR); 334 } 335 if(LDNS_NSCOUNT(sldns_buffer_begin(pkt)) != 0) { 336 verbose(VERB_QUERY, "request wrong nr ns=%d", 337 LDNS_NSCOUNT(sldns_buffer_begin(pkt))); 338 return worker_err_ratelimit(worker, LDNS_RCODE_FORMERR); 339 } 340 if(LDNS_ARCOUNT(sldns_buffer_begin(pkt)) > 1) { 341 verbose(VERB_QUERY, "request wrong nr ar=%d", 342 LDNS_ARCOUNT(sldns_buffer_begin(pkt))); 343 return worker_err_ratelimit(worker, LDNS_RCODE_FORMERR); 344 } 345 return 0; 346 } 347 348 void 349 worker_handle_control_cmd(struct tube* ATTR_UNUSED(tube), uint8_t* msg, 350 size_t len, int error, void* arg) 351 { 352 struct worker* worker = (struct worker*)arg; 353 enum worker_commands cmd; 354 if(error != NETEVENT_NOERROR) { 355 free(msg); 356 if(error == NETEVENT_CLOSED) 357 comm_base_exit(worker->base); 358 else log_info("control event: %d", error); 359 return; 360 } 361 if(len != sizeof(uint32_t)) { 362 fatal_exit("bad control msg length %d", (int)len); 363 } 364 cmd = sldns_read_uint32(msg); 365 free(msg); 366 switch(cmd) { 367 case worker_cmd_quit: 368 verbose(VERB_ALGO, "got control cmd quit"); 369 comm_base_exit(worker->base); 370 break; 371 case worker_cmd_stats: 372 verbose(VERB_ALGO, "got control cmd stats"); 373 server_stats_reply(worker, 1); 374 break; 375 case worker_cmd_stats_noreset: 376 verbose(VERB_ALGO, "got control cmd stats_noreset"); 377 server_stats_reply(worker, 0); 378 break; 379 case worker_cmd_remote: 380 verbose(VERB_ALGO, "got control cmd remote"); 381 daemon_remote_exec(worker); 382 break; 383 default: 384 log_err("bad command %d", (int)cmd); 385 break; 386 } 387 } 388 389 /** check if a delegation is secure */ 390 static enum sec_status 391 check_delegation_secure(struct reply_info *rep) 392 { 393 /* return smallest security status */ 394 size_t i; 395 enum sec_status sec = sec_status_secure; 396 enum sec_status s; 397 size_t num = rep->an_numrrsets + rep->ns_numrrsets; 398 /* check if answer and authority are OK */ 399 for(i=0; i<num; i++) { 400 s = ((struct packed_rrset_data*)rep->rrsets[i]->entry.data) 401 ->security; 402 if(s < sec) 403 sec = s; 404 } 405 /* in additional, only unchecked triggers revalidation */ 406 for(i=num; i<rep->rrset_count; i++) { 407 s = ((struct packed_rrset_data*)rep->rrsets[i]->entry.data) 408 ->security; 409 if(s == sec_status_unchecked) 410 return s; 411 } 412 return sec; 413 } 414 415 /** remove nonsecure from a delegation referral additional section */ 416 static void 417 deleg_remove_nonsecure_additional(struct reply_info* rep) 418 { 419 /* we can simply edit it, since we are working in the scratch region */ 420 size_t i; 421 enum sec_status s; 422 423 for(i = rep->an_numrrsets+rep->ns_numrrsets; i<rep->rrset_count; i++) { 424 s = ((struct packed_rrset_data*)rep->rrsets[i]->entry.data) 425 ->security; 426 if(s != sec_status_secure) { 427 memmove(rep->rrsets+i, rep->rrsets+i+1, 428 sizeof(struct ub_packed_rrset_key*)* 429 (rep->rrset_count - i - 1)); 430 rep->ar_numrrsets--; 431 rep->rrset_count--; 432 i--; 433 } 434 } 435 } 436 437 /** answer nonrecursive query from the cache */ 438 static int 439 answer_norec_from_cache(struct worker* worker, struct query_info* qinfo, 440 uint16_t id, uint16_t flags, struct comm_reply* repinfo, 441 struct edns_data* edns) 442 { 443 /* for a nonrecursive query return either: 444 * o an error (servfail; we try to avoid this) 445 * o a delegation (closest we have; this routine tries that) 446 * o the answer (checked by answer_from_cache) 447 * 448 * So, grab a delegation from the rrset cache. 449 * Then check if it needs validation, if so, this routine fails, 450 * so that iterator can prime and validator can verify rrsets. 451 */ 452 struct edns_data edns_bak; 453 uint16_t udpsize = edns->udp_size; 454 int secure = 0; 455 time_t timenow = *worker->env.now; 456 int must_validate = (!(flags&BIT_CD) || worker->env.cfg->ignore_cd) 457 && worker->env.need_to_validate; 458 struct dns_msg *msg = NULL; 459 struct delegpt *dp; 460 461 dp = dns_cache_find_delegation(&worker->env, qinfo->qname, 462 qinfo->qname_len, qinfo->qtype, qinfo->qclass, 463 worker->scratchpad, &msg, timenow); 464 if(!dp) { /* no delegation, need to reprime */ 465 return 0; 466 } 467 /* In case we have a local alias, copy it into the delegation message. 468 * Shallow copy should be fine, as we'll be done with msg in this 469 * function. */ 470 msg->qinfo.local_alias = qinfo->local_alias; 471 if(must_validate) { 472 switch(check_delegation_secure(msg->rep)) { 473 case sec_status_unchecked: 474 /* some rrsets have not been verified yet, go and 475 * let validator do that */ 476 return 0; 477 case sec_status_bogus: 478 case sec_status_secure_sentinel_fail: 479 /* some rrsets are bogus, reply servfail */ 480 edns->edns_version = EDNS_ADVERTISED_VERSION; 481 edns->udp_size = EDNS_ADVERTISED_SIZE; 482 edns->ext_rcode = 0; 483 edns->bits &= EDNS_DO; 484 if(!inplace_cb_reply_servfail_call(&worker->env, qinfo, NULL, 485 msg->rep, LDNS_RCODE_SERVFAIL, edns, repinfo, worker->scratchpad, 486 worker->env.now_tv)) 487 return 0; 488 error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL, 489 &msg->qinfo, id, flags, edns); 490 if(worker->stats.extended) { 491 worker->stats.ans_bogus++; 492 worker->stats.ans_rcode[LDNS_RCODE_SERVFAIL]++; 493 } 494 return 1; 495 case sec_status_secure: 496 /* all rrsets are secure */ 497 /* remove non-secure rrsets from the add. section*/ 498 if(worker->env.cfg->val_clean_additional) 499 deleg_remove_nonsecure_additional(msg->rep); 500 secure = 1; 501 break; 502 case sec_status_indeterminate: 503 case sec_status_insecure: 504 default: 505 /* not secure */ 506 secure = 0; 507 break; 508 } 509 } 510 /* return this delegation from the cache */ 511 edns_bak = *edns; 512 edns->edns_version = EDNS_ADVERTISED_VERSION; 513 edns->udp_size = EDNS_ADVERTISED_SIZE; 514 edns->ext_rcode = 0; 515 edns->bits &= EDNS_DO; 516 if(!inplace_cb_reply_cache_call(&worker->env, qinfo, NULL, msg->rep, 517 (int)(flags&LDNS_RCODE_MASK), edns, repinfo, worker->scratchpad, 518 worker->env.now_tv)) 519 return 0; 520 msg->rep->flags |= BIT_QR|BIT_RA; 521 if(!apply_edns_options(edns, &edns_bak, worker->env.cfg, 522 repinfo->c, worker->scratchpad) || 523 !reply_info_answer_encode(&msg->qinfo, msg->rep, id, flags, 524 repinfo->c->buffer, 0, 1, worker->scratchpad, 525 udpsize, edns, (int)(edns->bits & EDNS_DO), secure)) { 526 if(!inplace_cb_reply_servfail_call(&worker->env, qinfo, NULL, NULL, 527 LDNS_RCODE_SERVFAIL, edns, repinfo, worker->scratchpad, 528 worker->env.now_tv)) 529 edns->opt_list = NULL; 530 error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL, 531 &msg->qinfo, id, flags, edns); 532 } 533 if(worker->stats.extended) { 534 if(secure) worker->stats.ans_secure++; 535 server_stats_insrcode(&worker->stats, repinfo->c->buffer); 536 } 537 return 1; 538 } 539 540 /** Apply, if applicable, a response IP action to a cached answer. 541 * If the answer is rewritten as a result of an action, '*encode_repp' will 542 * point to the reply info containing the modified answer. '*encode_repp' will 543 * be intact otherwise. 544 * It returns 1 on success, 0 otherwise. */ 545 static int 546 apply_respip_action(struct worker* worker, const struct query_info* qinfo, 547 struct respip_client_info* cinfo, struct reply_info* rep, 548 struct comm_reply* repinfo, struct ub_packed_rrset_key** alias_rrset, 549 struct reply_info** encode_repp, struct auth_zones* az) 550 { 551 struct respip_action_info actinfo = {0, 0, 0, 0, NULL, 0, NULL}; 552 actinfo.action = respip_none; 553 554 if(qinfo->qtype != LDNS_RR_TYPE_A && 555 qinfo->qtype != LDNS_RR_TYPE_AAAA && 556 qinfo->qtype != LDNS_RR_TYPE_ANY) 557 return 1; 558 559 if(!respip_rewrite_reply(qinfo, cinfo, rep, encode_repp, &actinfo, 560 alias_rrset, 0, worker->scratchpad, az)) 561 return 0; 562 563 /* xxx_deny actions mean dropping the reply, unless the original reply 564 * was redirected to response-ip data. */ 565 if((actinfo.action == respip_deny || 566 actinfo.action == respip_inform_deny) && 567 *encode_repp == rep) 568 *encode_repp = NULL; 569 570 /* If address info is returned, it means the action should be an 571 * 'inform' variant and the information should be logged. */ 572 if(actinfo.addrinfo) { 573 respip_inform_print(&actinfo, qinfo->qname, 574 qinfo->qtype, qinfo->qclass, qinfo->local_alias, 575 repinfo); 576 577 if(worker->stats.extended && actinfo.rpz_used) { 578 if(actinfo.rpz_disabled) 579 worker->stats.rpz_action[RPZ_DISABLED_ACTION]++; 580 if(actinfo.rpz_cname_override) 581 worker->stats.rpz_action[RPZ_CNAME_OVERRIDE_ACTION]++; 582 else 583 worker->stats.rpz_action[ 584 respip_action_to_rpz_action(actinfo.action)]++; 585 } 586 } 587 588 return 1; 589 } 590 591 /** answer query from the cache. 592 * Normally, the answer message will be built in repinfo->c->buffer; if the 593 * answer is supposed to be suppressed or the answer is supposed to be an 594 * incomplete CNAME chain, the buffer is explicitly cleared to signal the 595 * caller as such. In the latter case *partial_rep will point to the incomplete 596 * reply, and this function is (possibly) supposed to be called again with that 597 * *partial_rep value to complete the chain. In addition, if the query should 598 * be completely dropped, '*need_drop' will be set to 1. */ 599 static int 600 answer_from_cache(struct worker* worker, struct query_info* qinfo, 601 struct respip_client_info* cinfo, int* need_drop, int* is_expired_answer, 602 int* is_secure_answer, struct ub_packed_rrset_key** alias_rrset, 603 struct reply_info** partial_repp, 604 struct reply_info* rep, uint16_t id, uint16_t flags, 605 struct comm_reply* repinfo, struct edns_data* edns) 606 { 607 struct edns_data edns_bak; 608 time_t timenow = *worker->env.now; 609 uint16_t udpsize = edns->udp_size; 610 struct reply_info* encode_rep = rep; 611 struct reply_info* partial_rep = *partial_repp; 612 int must_validate = (!(flags&BIT_CD) || worker->env.cfg->ignore_cd) 613 && worker->env.need_to_validate; 614 *partial_repp = NULL; /* avoid accidental further pass */ 615 616 /* Check TTL */ 617 if(rep->ttl < timenow) { 618 /* Check if we need to serve expired now */ 619 if(worker->env.cfg->serve_expired && 620 !worker->env.cfg->serve_expired_client_timeout) { 621 if(worker->env.cfg->serve_expired_ttl && 622 rep->serve_expired_ttl < timenow) 623 return 0; 624 if(!rrset_array_lock(rep->ref, rep->rrset_count, 0)) 625 return 0; 626 *is_expired_answer = 1; 627 } else { 628 /* the rrsets may have been updated in the meantime. 629 * we will refetch the message format from the 630 * authoritative server 631 */ 632 return 0; 633 } 634 } else { 635 if(!rrset_array_lock(rep->ref, rep->rrset_count, timenow)) 636 return 0; 637 } 638 /* locked and ids and ttls are OK. */ 639 640 /* check CNAME chain (if any) */ 641 if(rep->an_numrrsets > 0 && (rep->rrsets[0]->rk.type == 642 htons(LDNS_RR_TYPE_CNAME) || rep->rrsets[0]->rk.type == 643 htons(LDNS_RR_TYPE_DNAME))) { 644 if(!reply_check_cname_chain(qinfo, rep)) { 645 /* cname chain invalid, redo iterator steps */ 646 verbose(VERB_ALGO, "Cache reply: cname chain broken"); 647 goto bail_out; 648 } 649 } 650 /* check security status of the cached answer */ 651 if(must_validate && (rep->security == sec_status_bogus || 652 rep->security == sec_status_secure_sentinel_fail)) { 653 /* BAD cached */ 654 edns->edns_version = EDNS_ADVERTISED_VERSION; 655 edns->udp_size = EDNS_ADVERTISED_SIZE; 656 edns->ext_rcode = 0; 657 edns->bits &= EDNS_DO; 658 if(!inplace_cb_reply_servfail_call(&worker->env, qinfo, NULL, rep, 659 LDNS_RCODE_SERVFAIL, edns, repinfo, worker->scratchpad, 660 worker->env.now_tv)) 661 goto bail_out; 662 error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL, 663 qinfo, id, flags, edns); 664 rrset_array_unlock_touch(worker->env.rrset_cache, 665 worker->scratchpad, rep->ref, rep->rrset_count); 666 if(worker->stats.extended) { 667 worker->stats.ans_bogus ++; 668 worker->stats.ans_rcode[LDNS_RCODE_SERVFAIL] ++; 669 } 670 return 1; 671 } else if(rep->security == sec_status_unchecked && must_validate) { 672 verbose(VERB_ALGO, "Cache reply: unchecked entry needs " 673 "validation"); 674 goto bail_out; /* need to validate cache entry first */ 675 } else if(rep->security == sec_status_secure) { 676 if(reply_all_rrsets_secure(rep)) { 677 *is_secure_answer = 1; 678 } else { 679 if(must_validate) { 680 verbose(VERB_ALGO, "Cache reply: secure entry" 681 " changed status"); 682 goto bail_out; /* rrset changed, re-verify */ 683 } 684 *is_secure_answer = 0; 685 } 686 } else *is_secure_answer = 0; 687 688 edns_bak = *edns; 689 edns->edns_version = EDNS_ADVERTISED_VERSION; 690 edns->udp_size = EDNS_ADVERTISED_SIZE; 691 edns->ext_rcode = 0; 692 edns->bits &= EDNS_DO; 693 if(!inplace_cb_reply_cache_call(&worker->env, qinfo, NULL, rep, 694 (int)(flags&LDNS_RCODE_MASK), edns, repinfo, worker->scratchpad, 695 worker->env.now_tv)) 696 goto bail_out; 697 *alias_rrset = NULL; /* avoid confusion if caller set it to non-NULL */ 698 if((worker->daemon->use_response_ip || worker->daemon->use_rpz) && 699 !partial_rep && !apply_respip_action(worker, qinfo, cinfo, rep, 700 repinfo, alias_rrset, 701 &encode_rep, worker->env.auth_zones)) { 702 goto bail_out; 703 } else if(partial_rep && 704 !respip_merge_cname(partial_rep, qinfo, rep, cinfo, 705 must_validate, &encode_rep, worker->scratchpad, 706 worker->env.auth_zones)) { 707 goto bail_out; 708 } 709 if(encode_rep != rep) { 710 /* if rewritten, it can't be considered "secure" */ 711 *is_secure_answer = 0; 712 } 713 if(!encode_rep || *alias_rrset) { 714 if(!encode_rep) 715 *need_drop = 1; 716 else { 717 /* If a partial CNAME chain is found, we first need to 718 * make a copy of the reply in the scratchpad so we 719 * can release the locks and lookup the cache again. */ 720 *partial_repp = reply_info_copy(encode_rep, NULL, 721 worker->scratchpad); 722 if(!*partial_repp) 723 goto bail_out; 724 } 725 } else if(!apply_edns_options(edns, &edns_bak, worker->env.cfg, 726 repinfo->c, worker->scratchpad) || 727 !reply_info_answer_encode(qinfo, encode_rep, id, flags, 728 repinfo->c->buffer, timenow, 1, worker->scratchpad, 729 udpsize, edns, (int)(edns->bits & EDNS_DO), *is_secure_answer)) { 730 if(!inplace_cb_reply_servfail_call(&worker->env, qinfo, NULL, NULL, 731 LDNS_RCODE_SERVFAIL, edns, repinfo, worker->scratchpad, 732 worker->env.now_tv)) 733 edns->opt_list = NULL; 734 error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL, 735 qinfo, id, flags, edns); 736 } 737 /* cannot send the reply right now, because blocking network syscall 738 * is bad while holding locks. */ 739 rrset_array_unlock_touch(worker->env.rrset_cache, worker->scratchpad, 740 rep->ref, rep->rrset_count); 741 /* go and return this buffer to the client */ 742 return 1; 743 744 bail_out: 745 rrset_array_unlock_touch(worker->env.rrset_cache, 746 worker->scratchpad, rep->ref, rep->rrset_count); 747 return 0; 748 } 749 750 /** Reply to client and perform prefetch to keep cache up to date. */ 751 static void 752 reply_and_prefetch(struct worker* worker, struct query_info* qinfo, 753 uint16_t flags, struct comm_reply* repinfo, time_t leeway, int noreply) 754 { 755 /* first send answer to client to keep its latency 756 * as small as a cachereply */ 757 if(!noreply) { 758 if(repinfo->c->tcp_req_info) { 759 sldns_buffer_copy( 760 repinfo->c->tcp_req_info->spool_buffer, 761 repinfo->c->buffer); 762 } 763 comm_point_send_reply(repinfo); 764 } 765 server_stats_prefetch(&worker->stats, worker); 766 767 /* create the prefetch in the mesh as a normal lookup without 768 * client addrs waiting, which has the cache blacklisted (to bypass 769 * the cache and go to the network for the data). */ 770 /* this (potentially) runs the mesh for the new query */ 771 mesh_new_prefetch(worker->env.mesh, qinfo, flags, leeway + 772 PREFETCH_EXPIRY_ADD); 773 } 774 775 /** 776 * Fill CH class answer into buffer. Keeps query. 777 * @param pkt: buffer 778 * @param str: string to put into text record (<255). 779 * array of strings, every string becomes a text record. 780 * @param num: number of strings in array. 781 * @param edns: edns reply information. 782 * @param worker: worker with scratch region. 783 * @param repinfo: reply information for a communication point. 784 */ 785 static void 786 chaos_replystr(sldns_buffer* pkt, char** str, int num, struct edns_data* edns, 787 struct worker* worker, struct comm_reply* repinfo) 788 { 789 int i; 790 unsigned int rd = LDNS_RD_WIRE(sldns_buffer_begin(pkt)); 791 unsigned int cd = LDNS_CD_WIRE(sldns_buffer_begin(pkt)); 792 sldns_buffer_clear(pkt); 793 sldns_buffer_skip(pkt, (ssize_t)sizeof(uint16_t)); /* skip id */ 794 sldns_buffer_write_u16(pkt, (uint16_t)(BIT_QR|BIT_RA)); 795 if(rd) LDNS_RD_SET(sldns_buffer_begin(pkt)); 796 if(cd) LDNS_CD_SET(sldns_buffer_begin(pkt)); 797 sldns_buffer_write_u16(pkt, 1); /* qdcount */ 798 sldns_buffer_write_u16(pkt, (uint16_t)num); /* ancount */ 799 sldns_buffer_write_u16(pkt, 0); /* nscount */ 800 sldns_buffer_write_u16(pkt, 0); /* arcount */ 801 (void)query_dname_len(pkt); /* skip qname */ 802 sldns_buffer_skip(pkt, (ssize_t)sizeof(uint16_t)); /* skip qtype */ 803 sldns_buffer_skip(pkt, (ssize_t)sizeof(uint16_t)); /* skip qclass */ 804 for(i=0; i<num; i++) { 805 size_t len = strlen(str[i]); 806 if(len>255) len=255; /* cap size of TXT record */ 807 sldns_buffer_write_u16(pkt, 0xc00c); /* compr ptr to query */ 808 sldns_buffer_write_u16(pkt, LDNS_RR_TYPE_TXT); 809 sldns_buffer_write_u16(pkt, LDNS_RR_CLASS_CH); 810 sldns_buffer_write_u32(pkt, 0); /* TTL */ 811 sldns_buffer_write_u16(pkt, sizeof(uint8_t) + len); 812 sldns_buffer_write_u8(pkt, len); 813 sldns_buffer_write(pkt, str[i], len); 814 } 815 sldns_buffer_flip(pkt); 816 edns->edns_version = EDNS_ADVERTISED_VERSION; 817 edns->udp_size = EDNS_ADVERTISED_SIZE; 818 edns->bits &= EDNS_DO; 819 if(!inplace_cb_reply_local_call(&worker->env, NULL, NULL, NULL, 820 LDNS_RCODE_NOERROR, edns, repinfo, worker->scratchpad, 821 worker->env.now_tv)) 822 edns->opt_list = NULL; 823 if(sldns_buffer_capacity(pkt) >= 824 sldns_buffer_limit(pkt)+calc_edns_field_size(edns)) 825 attach_edns_record(pkt, edns); 826 } 827 828 /** Reply with one string */ 829 static void 830 chaos_replyonestr(sldns_buffer* pkt, const char* str, struct edns_data* edns, 831 struct worker* worker, struct comm_reply* repinfo) 832 { 833 chaos_replystr(pkt, (char**)&str, 1, edns, worker, repinfo); 834 } 835 836 /** 837 * Create CH class trustanchor answer. 838 * @param pkt: buffer 839 * @param edns: edns reply information. 840 * @param w: worker with scratch region. 841 * @param repinfo: reply information for a communication point. 842 */ 843 static void 844 chaos_trustanchor(sldns_buffer* pkt, struct edns_data* edns, struct worker* w, 845 struct comm_reply* repinfo) 846 { 847 #define TA_RESPONSE_MAX_TXT 16 /* max number of TXT records */ 848 #define TA_RESPONSE_MAX_TAGS 32 /* max number of tags printed per zone */ 849 char* str_array[TA_RESPONSE_MAX_TXT]; 850 uint16_t tags[TA_RESPONSE_MAX_TAGS]; 851 int num = 0; 852 struct trust_anchor* ta; 853 854 if(!w->env.need_to_validate) { 855 /* no validator module, reply no trustanchors */ 856 chaos_replystr(pkt, NULL, 0, edns, w, repinfo); 857 return; 858 } 859 860 /* fill the string with contents */ 861 lock_basic_lock(&w->env.anchors->lock); 862 RBTREE_FOR(ta, struct trust_anchor*, w->env.anchors->tree) { 863 char* str; 864 size_t i, numtag, str_len = 255; 865 if(num == TA_RESPONSE_MAX_TXT) continue; 866 str = (char*)regional_alloc(w->scratchpad, str_len); 867 if(!str) continue; 868 lock_basic_lock(&ta->lock); 869 numtag = anchor_list_keytags(ta, tags, TA_RESPONSE_MAX_TAGS); 870 if(numtag == 0) { 871 /* empty, insecure point */ 872 lock_basic_unlock(&ta->lock); 873 continue; 874 } 875 str_array[num] = str; 876 num++; 877 878 /* spool name of anchor */ 879 (void)sldns_wire2str_dname_buf(ta->name, ta->namelen, str, str_len); 880 str_len -= strlen(str); str += strlen(str); 881 /* spool tags */ 882 for(i=0; i<numtag; i++) { 883 snprintf(str, str_len, " %u", (unsigned)tags[i]); 884 str_len -= strlen(str); str += strlen(str); 885 } 886 lock_basic_unlock(&ta->lock); 887 } 888 lock_basic_unlock(&w->env.anchors->lock); 889 890 chaos_replystr(pkt, str_array, num, edns, w, repinfo); 891 regional_free_all(w->scratchpad); 892 } 893 894 /** 895 * Answer CH class queries. 896 * @param w: worker 897 * @param qinfo: query info. Pointer into packet buffer. 898 * @param edns: edns info from query. 899 * @param repinfo: reply information for a communication point. 900 * @param pkt: packet buffer. 901 * @return: true if a reply is to be sent. 902 */ 903 static int 904 answer_chaos(struct worker* w, struct query_info* qinfo, 905 struct edns_data* edns, struct comm_reply* repinfo, sldns_buffer* pkt) 906 { 907 struct config_file* cfg = w->env.cfg; 908 if(qinfo->qtype != LDNS_RR_TYPE_ANY && qinfo->qtype != LDNS_RR_TYPE_TXT) 909 return 0; 910 if(query_dname_compare(qinfo->qname, 911 (uint8_t*)"\002id\006server") == 0 || 912 query_dname_compare(qinfo->qname, 913 (uint8_t*)"\010hostname\004bind") == 0) 914 { 915 if(cfg->hide_identity) 916 return 0; 917 if(cfg->identity==NULL || cfg->identity[0]==0) { 918 char buf[MAXHOSTNAMELEN+1]; 919 if (gethostname(buf, MAXHOSTNAMELEN) == 0) { 920 buf[MAXHOSTNAMELEN] = 0; 921 chaos_replyonestr(pkt, buf, edns, w, repinfo); 922 } else { 923 log_err("gethostname: %s", strerror(errno)); 924 chaos_replyonestr(pkt, "no hostname", edns, w, repinfo); 925 } 926 } 927 else chaos_replyonestr(pkt, cfg->identity, edns, w, repinfo); 928 return 1; 929 } 930 if(query_dname_compare(qinfo->qname, 931 (uint8_t*)"\007version\006server") == 0 || 932 query_dname_compare(qinfo->qname, 933 (uint8_t*)"\007version\004bind") == 0) 934 { 935 if(cfg->hide_version) 936 return 0; 937 if(cfg->version==NULL || cfg->version[0]==0) 938 chaos_replyonestr(pkt, PACKAGE_STRING, edns, w, repinfo); 939 else chaos_replyonestr(pkt, cfg->version, edns, w, repinfo); 940 return 1; 941 } 942 if(query_dname_compare(qinfo->qname, 943 (uint8_t*)"\013trustanchor\007unbound") == 0) 944 { 945 if(cfg->hide_trustanchor) 946 return 0; 947 chaos_trustanchor(pkt, edns, w, repinfo); 948 return 1; 949 } 950 951 return 0; 952 } 953 954 /** 955 * Answer notify queries. These are notifies for authoritative zones, 956 * the reply is an ack that the notify has been received. We need to check 957 * access permission here. 958 * @param w: worker 959 * @param qinfo: query info. Pointer into packet buffer. 960 * @param edns: edns info from query. 961 * @param repinfo: reply info with source address. 962 * @param pkt: packet buffer. 963 */ 964 static void 965 answer_notify(struct worker* w, struct query_info* qinfo, 966 struct edns_data* edns, sldns_buffer* pkt, struct comm_reply* repinfo) 967 { 968 int refused = 0; 969 int rcode = LDNS_RCODE_NOERROR; 970 uint32_t serial = 0; 971 int has_serial; 972 if(!w->env.auth_zones) return; 973 has_serial = auth_zone_parse_notify_serial(pkt, &serial); 974 if(auth_zones_notify(w->env.auth_zones, &w->env, qinfo->qname, 975 qinfo->qname_len, qinfo->qclass, &repinfo->addr, 976 repinfo->addrlen, has_serial, serial, &refused)) { 977 rcode = LDNS_RCODE_NOERROR; 978 } else { 979 if(refused) 980 rcode = LDNS_RCODE_REFUSED; 981 else rcode = LDNS_RCODE_SERVFAIL; 982 } 983 984 if(verbosity >= VERB_DETAIL) { 985 char buf[380]; 986 char zname[255+1]; 987 char sr[25]; 988 dname_str(qinfo->qname, zname); 989 sr[0]=0; 990 if(has_serial) 991 snprintf(sr, sizeof(sr), "serial %u ", 992 (unsigned)serial); 993 if(rcode == LDNS_RCODE_REFUSED) 994 snprintf(buf, sizeof(buf), 995 "refused NOTIFY %sfor %s from", sr, zname); 996 else if(rcode == LDNS_RCODE_SERVFAIL) 997 snprintf(buf, sizeof(buf), 998 "servfail for NOTIFY %sfor %s from", sr, zname); 999 else snprintf(buf, sizeof(buf), 1000 "received NOTIFY %sfor %s from", sr, zname); 1001 log_addr(VERB_DETAIL, buf, &repinfo->addr, repinfo->addrlen); 1002 } 1003 edns->edns_version = EDNS_ADVERTISED_VERSION; 1004 edns->udp_size = EDNS_ADVERTISED_SIZE; 1005 edns->ext_rcode = 0; 1006 edns->bits &= EDNS_DO; 1007 edns->opt_list = NULL; 1008 error_encode(pkt, rcode, qinfo, 1009 *(uint16_t*)(void *)sldns_buffer_begin(pkt), 1010 sldns_buffer_read_u16_at(pkt, 2), edns); 1011 LDNS_OPCODE_SET(sldns_buffer_begin(pkt), LDNS_PACKET_NOTIFY); 1012 } 1013 1014 static int 1015 deny_refuse(struct comm_point* c, enum acl_access acl, 1016 enum acl_access deny, enum acl_access refuse, 1017 struct worker* worker, struct comm_reply* repinfo) 1018 { 1019 if(acl == deny) { 1020 comm_point_drop_reply(repinfo); 1021 if(worker->stats.extended) 1022 worker->stats.unwanted_queries++; 1023 return 0; 1024 } else if(acl == refuse) { 1025 log_addr(VERB_ALGO, "refused query from", 1026 &repinfo->addr, repinfo->addrlen); 1027 log_buf(VERB_ALGO, "refuse", c->buffer); 1028 if(worker->stats.extended) 1029 worker->stats.unwanted_queries++; 1030 if(worker_check_request(c->buffer, worker) == -1) { 1031 comm_point_drop_reply(repinfo); 1032 return 0; /* discard this */ 1033 } 1034 sldns_buffer_set_limit(c->buffer, LDNS_HEADER_SIZE); 1035 sldns_buffer_write_at(c->buffer, 4, 1036 (uint8_t*)"\0\0\0\0\0\0\0\0", 8); 1037 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1038 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1039 LDNS_RCODE_REFUSED); 1040 sldns_buffer_set_position(c->buffer, LDNS_HEADER_SIZE); 1041 sldns_buffer_flip(c->buffer); 1042 return 1; 1043 } 1044 1045 return -1; 1046 } 1047 1048 static int 1049 deny_refuse_all(struct comm_point* c, enum acl_access acl, 1050 struct worker* worker, struct comm_reply* repinfo) 1051 { 1052 return deny_refuse(c, acl, acl_deny, acl_refuse, worker, repinfo); 1053 } 1054 1055 static int 1056 deny_refuse_non_local(struct comm_point* c, enum acl_access acl, 1057 struct worker* worker, struct comm_reply* repinfo) 1058 { 1059 return deny_refuse(c, acl, acl_deny_non_local, acl_refuse_non_local, worker, repinfo); 1060 } 1061 1062 int 1063 worker_handle_request(struct comm_point* c, void* arg, int error, 1064 struct comm_reply* repinfo) 1065 { 1066 struct worker* worker = (struct worker*)arg; 1067 int ret; 1068 hashvalue_type h; 1069 struct lruhash_entry* e; 1070 struct query_info qinfo; 1071 struct edns_data edns; 1072 enum acl_access acl; 1073 struct acl_addr* acladdr; 1074 int rc = 0; 1075 int need_drop = 0; 1076 int is_expired_answer = 0; 1077 int is_secure_answer = 0; 1078 /* We might have to chase a CNAME chain internally, in which case 1079 * we'll have up to two replies and combine them to build a complete 1080 * answer. These variables control this case. */ 1081 struct ub_packed_rrset_key* alias_rrset = NULL; 1082 struct reply_info* partial_rep = NULL; 1083 struct query_info* lookup_qinfo = &qinfo; 1084 struct query_info qinfo_tmp; /* placeholder for lookup_qinfo */ 1085 struct respip_client_info* cinfo = NULL, cinfo_tmp; 1086 memset(&qinfo, 0, sizeof(qinfo)); 1087 1088 if((error != NETEVENT_NOERROR && error != NETEVENT_DONE)|| !repinfo) { 1089 /* some bad tcp query DNS formats give these error calls */ 1090 verbose(VERB_ALGO, "handle request called with err=%d", error); 1091 return 0; 1092 } 1093 #ifdef USE_DNSCRYPT 1094 repinfo->max_udp_size = worker->daemon->cfg->max_udp_size; 1095 if(!dnsc_handle_curved_request(worker->daemon->dnscenv, repinfo)) { 1096 worker->stats.num_query_dnscrypt_crypted_malformed++; 1097 return 0; 1098 } 1099 if(c->dnscrypt && !repinfo->is_dnscrypted) { 1100 char buf[LDNS_MAX_DOMAINLEN+1]; 1101 /* Check if this is unencrypted and asking for certs */ 1102 if(worker_check_request(c->buffer, worker) != 0) { 1103 verbose(VERB_ALGO, 1104 "dnscrypt: worker check request: bad query."); 1105 log_addr(VERB_CLIENT,"from",&repinfo->addr, 1106 repinfo->addrlen); 1107 comm_point_drop_reply(repinfo); 1108 return 0; 1109 } 1110 if(!query_info_parse(&qinfo, c->buffer)) { 1111 verbose(VERB_ALGO, 1112 "dnscrypt: worker parse request: formerror."); 1113 log_addr(VERB_CLIENT, "from", &repinfo->addr, 1114 repinfo->addrlen); 1115 comm_point_drop_reply(repinfo); 1116 return 0; 1117 } 1118 dname_str(qinfo.qname, buf); 1119 if(!(qinfo.qtype == LDNS_RR_TYPE_TXT && 1120 strcasecmp(buf, 1121 worker->daemon->dnscenv->provider_name) == 0)) { 1122 verbose(VERB_ALGO, 1123 "dnscrypt: not TXT \"%s\". Received: %s \"%s\"", 1124 worker->daemon->dnscenv->provider_name, 1125 sldns_rr_descript(qinfo.qtype)->_name, 1126 buf); 1127 comm_point_drop_reply(repinfo); 1128 worker->stats.num_query_dnscrypt_cleartext++; 1129 return 0; 1130 } 1131 worker->stats.num_query_dnscrypt_cert++; 1132 sldns_buffer_rewind(c->buffer); 1133 } else if(c->dnscrypt && repinfo->is_dnscrypted) { 1134 worker->stats.num_query_dnscrypt_crypted++; 1135 } 1136 #endif 1137 #ifdef USE_DNSTAP 1138 /* 1139 * sending src (client)/dst (local service) addresses over DNSTAP from incoming request handler 1140 */ 1141 if(worker->dtenv.log_client_query_messages) { 1142 log_addr(VERB_ALGO, "request from client", &repinfo->addr, repinfo->addrlen); 1143 log_addr(VERB_ALGO, "to local addr", (void*)repinfo->c->socket->addr->ai_addr, repinfo->c->socket->addr->ai_addrlen); 1144 dt_msg_send_client_query(&worker->dtenv, &repinfo->addr, (void*)repinfo->c->socket->addr->ai_addr, c->type, c->buffer); 1145 } 1146 #endif 1147 acladdr = acl_addr_lookup(worker->daemon->acl, &repinfo->addr, 1148 repinfo->addrlen); 1149 acl = acl_get_control(acladdr); 1150 if((ret=deny_refuse_all(c, acl, worker, repinfo)) != -1) 1151 { 1152 if(ret == 1) 1153 goto send_reply; 1154 return ret; 1155 } 1156 if((ret=worker_check_request(c->buffer, worker)) != 0) { 1157 verbose(VERB_ALGO, "worker check request: bad query."); 1158 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen); 1159 if(ret != -1) { 1160 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1161 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), ret); 1162 return 1; 1163 } 1164 comm_point_drop_reply(repinfo); 1165 return 0; 1166 } 1167 1168 worker->stats.num_queries++; 1169 1170 /* check if this query should be dropped based on source ip rate limiting */ 1171 if(!infra_ip_ratelimit_inc(worker->env.infra_cache, repinfo, 1172 *worker->env.now, c->buffer)) { 1173 /* See if we are passed through with slip factor */ 1174 if(worker->env.cfg->ip_ratelimit_factor != 0 && 1175 ub_random_max(worker->env.rnd, 1176 worker->env.cfg->ip_ratelimit_factor) == 0) { 1177 1178 char addrbuf[128]; 1179 addr_to_str(&repinfo->addr, repinfo->addrlen, 1180 addrbuf, sizeof(addrbuf)); 1181 verbose(VERB_QUERY, "ip_ratelimit allowed through for ip address %s because of slip in ip_ratelimit_factor", 1182 addrbuf); 1183 } else { 1184 worker->stats.num_queries_ip_ratelimited++; 1185 comm_point_drop_reply(repinfo); 1186 return 0; 1187 } 1188 } 1189 1190 /* see if query is in the cache */ 1191 if(!query_info_parse(&qinfo, c->buffer)) { 1192 verbose(VERB_ALGO, "worker parse request: formerror."); 1193 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen); 1194 memset(&qinfo, 0, sizeof(qinfo)); /* zero qinfo.qname */ 1195 if(worker_err_ratelimit(worker, LDNS_RCODE_FORMERR) == -1) { 1196 comm_point_drop_reply(repinfo); 1197 return 0; 1198 } 1199 sldns_buffer_rewind(c->buffer); 1200 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1201 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1202 LDNS_RCODE_FORMERR); 1203 goto send_reply; 1204 } 1205 if(worker->env.cfg->log_queries) { 1206 char ip[128]; 1207 addr_to_str(&repinfo->addr, repinfo->addrlen, ip, sizeof(ip)); 1208 log_query_in(ip, qinfo.qname, qinfo.qtype, qinfo.qclass); 1209 } 1210 if(qinfo.qtype == LDNS_RR_TYPE_AXFR || 1211 qinfo.qtype == LDNS_RR_TYPE_IXFR) { 1212 verbose(VERB_ALGO, "worker request: refused zone transfer."); 1213 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen); 1214 sldns_buffer_rewind(c->buffer); 1215 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1216 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1217 LDNS_RCODE_REFUSED); 1218 if(worker->stats.extended) { 1219 worker->stats.qtype[qinfo.qtype]++; 1220 } 1221 goto send_reply; 1222 } 1223 if(qinfo.qtype == LDNS_RR_TYPE_OPT || 1224 qinfo.qtype == LDNS_RR_TYPE_TSIG || 1225 qinfo.qtype == LDNS_RR_TYPE_TKEY || 1226 qinfo.qtype == LDNS_RR_TYPE_MAILA || 1227 qinfo.qtype == LDNS_RR_TYPE_MAILB || 1228 (qinfo.qtype >= 128 && qinfo.qtype <= 248)) { 1229 verbose(VERB_ALGO, "worker request: formerror for meta-type."); 1230 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen); 1231 if(worker_err_ratelimit(worker, LDNS_RCODE_FORMERR) == -1) { 1232 comm_point_drop_reply(repinfo); 1233 return 0; 1234 } 1235 sldns_buffer_rewind(c->buffer); 1236 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1237 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1238 LDNS_RCODE_FORMERR); 1239 if(worker->stats.extended) { 1240 worker->stats.qtype[qinfo.qtype]++; 1241 } 1242 goto send_reply; 1243 } 1244 if((ret=parse_edns_from_pkt(c->buffer, &edns, worker->scratchpad)) != 0) { 1245 struct edns_data reply_edns; 1246 verbose(VERB_ALGO, "worker parse edns: formerror."); 1247 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen); 1248 memset(&reply_edns, 0, sizeof(reply_edns)); 1249 reply_edns.edns_present = 1; 1250 reply_edns.udp_size = EDNS_ADVERTISED_SIZE; 1251 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), ret); 1252 error_encode(c->buffer, ret, &qinfo, 1253 *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 1254 sldns_buffer_read_u16_at(c->buffer, 2), &reply_edns); 1255 regional_free_all(worker->scratchpad); 1256 goto send_reply; 1257 } 1258 if(edns.edns_present) { 1259 struct edns_option* edns_opt; 1260 if(edns.edns_version != 0) { 1261 edns.ext_rcode = (uint8_t)(EDNS_RCODE_BADVERS>>4); 1262 edns.edns_version = EDNS_ADVERTISED_VERSION; 1263 edns.udp_size = EDNS_ADVERTISED_SIZE; 1264 edns.bits &= EDNS_DO; 1265 edns.opt_list = NULL; 1266 edns.padding_block_size = 0; 1267 verbose(VERB_ALGO, "query with bad edns version."); 1268 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen); 1269 error_encode(c->buffer, EDNS_RCODE_BADVERS&0xf, &qinfo, 1270 *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 1271 sldns_buffer_read_u16_at(c->buffer, 2), NULL); 1272 if(sldns_buffer_capacity(c->buffer) >= 1273 sldns_buffer_limit(c->buffer)+calc_edns_field_size(&edns)) 1274 attach_edns_record(c->buffer, &edns); 1275 regional_free_all(worker->scratchpad); 1276 goto send_reply; 1277 } 1278 if(edns.udp_size < NORMAL_UDP_SIZE && 1279 worker->daemon->cfg->harden_short_bufsize) { 1280 verbose(VERB_QUERY, "worker request: EDNS bufsize %d ignored", 1281 (int)edns.udp_size); 1282 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen); 1283 edns.udp_size = NORMAL_UDP_SIZE; 1284 } 1285 if(c->type != comm_udp) { 1286 edns_opt = edns_opt_list_find(edns.opt_list, LDNS_EDNS_KEEPALIVE); 1287 if(edns_opt && edns_opt->opt_len > 0) { 1288 edns.ext_rcode = 0; 1289 edns.edns_version = EDNS_ADVERTISED_VERSION; 1290 edns.udp_size = EDNS_ADVERTISED_SIZE; 1291 edns.bits &= EDNS_DO; 1292 edns.opt_list = NULL; 1293 verbose(VERB_ALGO, "query with bad edns keepalive."); 1294 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen); 1295 error_encode(c->buffer, LDNS_RCODE_FORMERR, &qinfo, 1296 *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 1297 sldns_buffer_read_u16_at(c->buffer, 2), NULL); 1298 if(sldns_buffer_capacity(c->buffer) >= 1299 sldns_buffer_limit(c->buffer)+calc_edns_field_size(&edns)) 1300 attach_edns_record(c->buffer, &edns); 1301 regional_free_all(worker->scratchpad); 1302 goto send_reply; 1303 } 1304 } 1305 } 1306 if(edns.udp_size > worker->daemon->cfg->max_udp_size && 1307 c->type == comm_udp) { 1308 verbose(VERB_QUERY, 1309 "worker request: max UDP reply size modified" 1310 " (%d to max-udp-size)", (int)edns.udp_size); 1311 log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen); 1312 edns.udp_size = worker->daemon->cfg->max_udp_size; 1313 } 1314 if(edns.udp_size < LDNS_HEADER_SIZE) { 1315 verbose(VERB_ALGO, "worker request: edns is too small."); 1316 log_addr(VERB_CLIENT, "from", &repinfo->addr, repinfo->addrlen); 1317 LDNS_QR_SET(sldns_buffer_begin(c->buffer)); 1318 LDNS_TC_SET(sldns_buffer_begin(c->buffer)); 1319 LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), 1320 LDNS_RCODE_SERVFAIL); 1321 sldns_buffer_set_position(c->buffer, LDNS_HEADER_SIZE); 1322 sldns_buffer_write_at(c->buffer, 4, 1323 (uint8_t*)"\0\0\0\0\0\0\0\0", 8); 1324 sldns_buffer_flip(c->buffer); 1325 regional_free_all(worker->scratchpad); 1326 goto send_reply; 1327 } 1328 if(worker->stats.extended) 1329 server_stats_insquery(&worker->stats, c, qinfo.qtype, 1330 qinfo.qclass, &edns, repinfo); 1331 if(c->type != comm_udp) 1332 edns.udp_size = 65535; /* max size for TCP replies */ 1333 if(qinfo.qclass == LDNS_RR_CLASS_CH && answer_chaos(worker, &qinfo, 1334 &edns, repinfo, c->buffer)) { 1335 regional_free_all(worker->scratchpad); 1336 goto send_reply; 1337 } 1338 if(LDNS_OPCODE_WIRE(sldns_buffer_begin(c->buffer)) == 1339 LDNS_PACKET_NOTIFY) { 1340 answer_notify(worker, &qinfo, &edns, c->buffer, repinfo); 1341 regional_free_all(worker->scratchpad); 1342 goto send_reply; 1343 } 1344 if(local_zones_answer(worker->daemon->local_zones, &worker->env, &qinfo, 1345 &edns, c->buffer, worker->scratchpad, repinfo, acladdr->taglist, 1346 acladdr->taglen, acladdr->tag_actions, 1347 acladdr->tag_actions_size, acladdr->tag_datas, 1348 acladdr->tag_datas_size, worker->daemon->cfg->tagname, 1349 worker->daemon->cfg->num_tags, acladdr->view)) { 1350 regional_free_all(worker->scratchpad); 1351 if(sldns_buffer_limit(c->buffer) == 0) { 1352 comm_point_drop_reply(repinfo); 1353 return 0; 1354 } 1355 goto send_reply; 1356 } 1357 if(worker->env.auth_zones && 1358 rpz_apply_qname_trigger(worker->env.auth_zones, 1359 &worker->env, &qinfo, &edns, c->buffer, worker->scratchpad, 1360 repinfo, acladdr->taglist, acladdr->taglen, &worker->stats)) { 1361 regional_free_all(worker->scratchpad); 1362 if(sldns_buffer_limit(c->buffer) == 0) { 1363 comm_point_drop_reply(repinfo); 1364 return 0; 1365 } 1366 goto send_reply; 1367 } 1368 if(worker->env.auth_zones && 1369 auth_zones_answer(worker->env.auth_zones, &worker->env, 1370 &qinfo, &edns, repinfo, c->buffer, worker->scratchpad)) { 1371 regional_free_all(worker->scratchpad); 1372 if(sldns_buffer_limit(c->buffer) == 0) { 1373 comm_point_drop_reply(repinfo); 1374 return 0; 1375 } 1376 /* set RA for everyone that can have recursion (based on 1377 * access control list) */ 1378 if(LDNS_RD_WIRE(sldns_buffer_begin(c->buffer)) && 1379 acl != acl_deny_non_local && acl != acl_refuse_non_local) 1380 LDNS_RA_SET(sldns_buffer_begin(c->buffer)); 1381 goto send_reply; 1382 } 1383 1384 /* We've looked in our local zones. If the answer isn't there, we 1385 * might need to bail out based on ACLs now. */ 1386 if((ret=deny_refuse_non_local(c, acl, worker, repinfo)) != -1) 1387 { 1388 regional_free_all(worker->scratchpad); 1389 if(ret == 1) 1390 goto send_reply; 1391 return ret; 1392 } 1393 1394 /* If this request does not have the recursion bit set, verify 1395 * ACLs allow the recursion bit to be treated as set. */ 1396 if(!(LDNS_RD_WIRE(sldns_buffer_begin(c->buffer))) && 1397 acl == acl_allow_setrd ) { 1398 LDNS_RD_SET(sldns_buffer_begin(c->buffer)); 1399 } 1400 1401 /* If this request does not have the recursion bit set, verify 1402 * ACLs allow the snooping. */ 1403 if(!(LDNS_RD_WIRE(sldns_buffer_begin(c->buffer))) && 1404 acl != acl_allow_snoop ) { 1405 error_encode(c->buffer, LDNS_RCODE_REFUSED, &qinfo, 1406 *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 1407 sldns_buffer_read_u16_at(c->buffer, 2), NULL); 1408 regional_free_all(worker->scratchpad); 1409 log_addr(VERB_ALGO, "refused nonrec (cache snoop) query from", 1410 &repinfo->addr, repinfo->addrlen); 1411 goto send_reply; 1412 } 1413 1414 /* If we've found a local alias, replace the qname with the alias 1415 * target before resolving it. */ 1416 if(qinfo.local_alias) { 1417 struct ub_packed_rrset_key* rrset = qinfo.local_alias->rrset; 1418 struct packed_rrset_data* d = rrset->entry.data; 1419 1420 /* Sanity check: our current implementation only supports 1421 * a single CNAME RRset as a local alias. */ 1422 if(qinfo.local_alias->next || 1423 rrset->rk.type != htons(LDNS_RR_TYPE_CNAME) || 1424 d->count != 1) { 1425 log_err("assumption failure: unexpected local alias"); 1426 regional_free_all(worker->scratchpad); 1427 return 0; /* drop it */ 1428 } 1429 qinfo.qname = d->rr_data[0] + 2; 1430 qinfo.qname_len = d->rr_len[0] - 2; 1431 } 1432 1433 /* If we may apply IP-based actions to the answer, build the client 1434 * information. As this can be expensive, skip it if there is 1435 * absolutely no possibility of it. */ 1436 if((worker->daemon->use_response_ip || worker->daemon->use_rpz) && 1437 (qinfo.qtype == LDNS_RR_TYPE_A || 1438 qinfo.qtype == LDNS_RR_TYPE_AAAA || 1439 qinfo.qtype == LDNS_RR_TYPE_ANY)) { 1440 cinfo_tmp.taglist = acladdr->taglist; 1441 cinfo_tmp.taglen = acladdr->taglen; 1442 cinfo_tmp.tag_actions = acladdr->tag_actions; 1443 cinfo_tmp.tag_actions_size = acladdr->tag_actions_size; 1444 cinfo_tmp.tag_datas = acladdr->tag_datas; 1445 cinfo_tmp.tag_datas_size = acladdr->tag_datas_size; 1446 cinfo_tmp.view = acladdr->view; 1447 cinfo_tmp.respip_set = worker->daemon->respip_set; 1448 cinfo = &cinfo_tmp; 1449 } 1450 1451 lookup_cache: 1452 /* Lookup the cache. In case we chase an intermediate CNAME chain 1453 * this is a two-pass operation, and lookup_qinfo is different for 1454 * each pass. We should still pass the original qinfo to 1455 * answer_from_cache(), however, since it's used to build the reply. */ 1456 if(!edns_bypass_cache_stage(edns.opt_list, &worker->env)) { 1457 is_expired_answer = 0; 1458 is_secure_answer = 0; 1459 h = query_info_hash(lookup_qinfo, sldns_buffer_read_u16_at(c->buffer, 2)); 1460 if((e=slabhash_lookup(worker->env.msg_cache, h, lookup_qinfo, 0))) { 1461 /* answer from cache - we have acquired a readlock on it */ 1462 if(answer_from_cache(worker, &qinfo, 1463 cinfo, &need_drop, &is_expired_answer, &is_secure_answer, 1464 &alias_rrset, &partial_rep, (struct reply_info*)e->data, 1465 *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 1466 sldns_buffer_read_u16_at(c->buffer, 2), repinfo, 1467 &edns)) { 1468 /* prefetch it if the prefetch TTL expired. 1469 * Note that if there is more than one pass 1470 * its qname must be that used for cache 1471 * lookup. */ 1472 if((worker->env.cfg->prefetch && *worker->env.now >= 1473 ((struct reply_info*)e->data)->prefetch_ttl) || 1474 (worker->env.cfg->serve_expired && 1475 *worker->env.now >= ((struct reply_info*)e->data)->ttl)) { 1476 1477 time_t leeway = ((struct reply_info*)e-> 1478 data)->ttl - *worker->env.now; 1479 if(((struct reply_info*)e->data)->ttl 1480 < *worker->env.now) 1481 leeway = 0; 1482 lock_rw_unlock(&e->lock); 1483 reply_and_prefetch(worker, lookup_qinfo, 1484 sldns_buffer_read_u16_at(c->buffer, 2), 1485 repinfo, leeway, 1486 (partial_rep || need_drop)); 1487 if(!partial_rep) { 1488 rc = 0; 1489 regional_free_all(worker->scratchpad); 1490 goto send_reply_rc; 1491 } 1492 } else if(!partial_rep) { 1493 lock_rw_unlock(&e->lock); 1494 regional_free_all(worker->scratchpad); 1495 goto send_reply; 1496 } else { 1497 /* Note that we've already released the 1498 * lock if we're here after prefetch. */ 1499 lock_rw_unlock(&e->lock); 1500 } 1501 /* We've found a partial reply ending with an 1502 * alias. Replace the lookup qinfo for the 1503 * alias target and lookup the cache again to 1504 * (possibly) complete the reply. As we're 1505 * passing the "base" reply, there will be no 1506 * more alias chasing. */ 1507 memset(&qinfo_tmp, 0, sizeof(qinfo_tmp)); 1508 get_cname_target(alias_rrset, &qinfo_tmp.qname, 1509 &qinfo_tmp.qname_len); 1510 if(!qinfo_tmp.qname) { 1511 log_err("unexpected: invalid answer alias"); 1512 regional_free_all(worker->scratchpad); 1513 return 0; /* drop query */ 1514 } 1515 qinfo_tmp.qtype = qinfo.qtype; 1516 qinfo_tmp.qclass = qinfo.qclass; 1517 lookup_qinfo = &qinfo_tmp; 1518 goto lookup_cache; 1519 } 1520 verbose(VERB_ALGO, "answer from the cache failed"); 1521 lock_rw_unlock(&e->lock); 1522 } 1523 if(!LDNS_RD_WIRE(sldns_buffer_begin(c->buffer))) { 1524 if(answer_norec_from_cache(worker, &qinfo, 1525 *(uint16_t*)(void *)sldns_buffer_begin(c->buffer), 1526 sldns_buffer_read_u16_at(c->buffer, 2), repinfo, 1527 &edns)) { 1528 regional_free_all(worker->scratchpad); 1529 goto send_reply; 1530 } 1531 verbose(VERB_ALGO, "answer norec from cache -- " 1532 "need to validate or not primed"); 1533 } 1534 } 1535 sldns_buffer_rewind(c->buffer); 1536 server_stats_querymiss(&worker->stats, worker); 1537 1538 if(verbosity >= VERB_CLIENT) { 1539 if(c->type == comm_udp) 1540 log_addr(VERB_CLIENT, "udp request from", 1541 &repinfo->addr, repinfo->addrlen); 1542 else log_addr(VERB_CLIENT, "tcp request from", 1543 &repinfo->addr, repinfo->addrlen); 1544 } 1545 1546 /* grab a work request structure for this new request */ 1547 mesh_new_client(worker->env.mesh, &qinfo, cinfo, 1548 sldns_buffer_read_u16_at(c->buffer, 2), 1549 &edns, repinfo, *(uint16_t*)(void *)sldns_buffer_begin(c->buffer)); 1550 regional_free_all(worker->scratchpad); 1551 worker_mem_report(worker, NULL); 1552 return 0; 1553 1554 send_reply: 1555 rc = 1; 1556 send_reply_rc: 1557 if(need_drop) { 1558 comm_point_drop_reply(repinfo); 1559 return 0; 1560 } 1561 if(is_expired_answer) { 1562 worker->stats.ans_expired++; 1563 } 1564 server_stats_insrcode(&worker->stats, c->buffer); 1565 if(worker->stats.extended) { 1566 if(is_secure_answer) worker->stats.ans_secure++; 1567 } 1568 #ifdef USE_DNSTAP 1569 /* 1570 * sending src (client)/dst (local service) addresses over DNSTAP from send_reply code label (when we serviced local zone for ex.) 1571 */ 1572 if(worker->dtenv.log_client_response_messages) { 1573 log_addr(VERB_ALGO, "from local addr", (void*)repinfo->c->socket->addr->ai_addr, repinfo->c->socket->addr->ai_addrlen); 1574 log_addr(VERB_ALGO, "response to client", &repinfo->addr, repinfo->addrlen); 1575 dt_msg_send_client_response(&worker->dtenv, &repinfo->addr, (void*)repinfo->c->socket->addr->ai_addr, c->type, c->buffer); 1576 } 1577 #endif 1578 if(worker->env.cfg->log_replies) 1579 { 1580 struct timeval tv; 1581 memset(&tv, 0, sizeof(tv)); 1582 if(qinfo.local_alias && qinfo.local_alias->rrset && 1583 qinfo.local_alias->rrset->rk.dname) { 1584 /* log original qname, before the local alias was 1585 * used to resolve that CNAME to something else */ 1586 qinfo.qname = qinfo.local_alias->rrset->rk.dname; 1587 log_reply_info(NO_VERBOSE, &qinfo, &repinfo->addr, repinfo->addrlen, 1588 tv, 1, c->buffer); 1589 } else { 1590 log_reply_info(NO_VERBOSE, &qinfo, &repinfo->addr, repinfo->addrlen, 1591 tv, 1, c->buffer); 1592 } 1593 } 1594 #ifdef USE_DNSCRYPT 1595 if(!dnsc_handle_uncurved_request(repinfo)) { 1596 return 0; 1597 } 1598 #endif 1599 return rc; 1600 } 1601 1602 void 1603 worker_sighandler(int sig, void* arg) 1604 { 1605 /* note that log, print, syscalls here give race conditions. 1606 * And cause hangups if the log-lock is held by the application. */ 1607 struct worker* worker = (struct worker*)arg; 1608 switch(sig) { 1609 #ifdef SIGHUP 1610 case SIGHUP: 1611 comm_base_exit(worker->base); 1612 break; 1613 #endif 1614 case SIGINT: 1615 worker->need_to_exit = 1; 1616 comm_base_exit(worker->base); 1617 break; 1618 #ifdef SIGQUIT 1619 case SIGQUIT: 1620 worker->need_to_exit = 1; 1621 comm_base_exit(worker->base); 1622 break; 1623 #endif 1624 case SIGTERM: 1625 worker->need_to_exit = 1; 1626 comm_base_exit(worker->base); 1627 break; 1628 default: 1629 /* unknown signal, ignored */ 1630 break; 1631 } 1632 } 1633 1634 /** restart statistics timer for worker, if enabled */ 1635 static void 1636 worker_restart_timer(struct worker* worker) 1637 { 1638 if(worker->env.cfg->stat_interval > 0) { 1639 struct timeval tv; 1640 #ifndef S_SPLINT_S 1641 tv.tv_sec = worker->env.cfg->stat_interval; 1642 tv.tv_usec = 0; 1643 #endif 1644 comm_timer_set(worker->stat_timer, &tv); 1645 } 1646 } 1647 1648 void worker_stat_timer_cb(void* arg) 1649 { 1650 struct worker* worker = (struct worker*)arg; 1651 server_stats_log(&worker->stats, worker, worker->thread_num); 1652 mesh_stats(worker->env.mesh, "mesh has"); 1653 worker_mem_report(worker, NULL); 1654 /* SHM is enabled, process data to SHM */ 1655 if (worker->daemon->cfg->shm_enable) { 1656 shm_main_run(worker); 1657 } 1658 if(!worker->daemon->cfg->stat_cumulative) { 1659 worker_stats_clear(worker); 1660 } 1661 /* start next timer */ 1662 worker_restart_timer(worker); 1663 } 1664 1665 void worker_probe_timer_cb(void* arg) 1666 { 1667 struct worker* worker = (struct worker*)arg; 1668 struct timeval tv; 1669 #ifndef S_SPLINT_S 1670 tv.tv_sec = (time_t)autr_probe_timer(&worker->env); 1671 tv.tv_usec = 0; 1672 #endif 1673 if(tv.tv_sec != 0) 1674 comm_timer_set(worker->env.probe_timer, &tv); 1675 } 1676 1677 struct worker* 1678 worker_create(struct daemon* daemon, int id, int* ports, int n) 1679 { 1680 unsigned int seed; 1681 struct worker* worker = (struct worker*)calloc(1, 1682 sizeof(struct worker)); 1683 if(!worker) 1684 return NULL; 1685 worker->numports = n; 1686 worker->ports = (int*)memdup(ports, sizeof(int)*n); 1687 if(!worker->ports) { 1688 free(worker); 1689 return NULL; 1690 } 1691 worker->daemon = daemon; 1692 worker->thread_num = id; 1693 if(!(worker->cmd = tube_create())) { 1694 free(worker->ports); 1695 free(worker); 1696 return NULL; 1697 } 1698 /* create random state here to avoid locking trouble in RAND_bytes */ 1699 if(!(worker->rndstate = ub_initstate(daemon->rand))) { 1700 log_err("could not init random numbers."); 1701 tube_delete(worker->cmd); 1702 free(worker->ports); 1703 free(worker); 1704 return NULL; 1705 } 1706 explicit_bzero(&seed, sizeof(seed)); 1707 return worker; 1708 } 1709 1710 int 1711 worker_init(struct worker* worker, struct config_file *cfg, 1712 struct listen_port* ports, int do_sigs) 1713 { 1714 #ifdef USE_DNSTAP 1715 struct dt_env* dtenv = &worker->dtenv; 1716 #else 1717 void* dtenv = NULL; 1718 #endif 1719 worker->need_to_exit = 0; 1720 worker->base = comm_base_create(do_sigs); 1721 if(!worker->base) { 1722 log_err("could not create event handling base"); 1723 worker_delete(worker); 1724 return 0; 1725 } 1726 comm_base_set_slow_accept_handlers(worker->base, &worker_stop_accept, 1727 &worker_start_accept, worker); 1728 if(do_sigs) { 1729 #ifdef SIGHUP 1730 ub_thread_sig_unblock(SIGHUP); 1731 #endif 1732 ub_thread_sig_unblock(SIGINT); 1733 #ifdef SIGQUIT 1734 ub_thread_sig_unblock(SIGQUIT); 1735 #endif 1736 ub_thread_sig_unblock(SIGTERM); 1737 #ifndef LIBEVENT_SIGNAL_PROBLEM 1738 worker->comsig = comm_signal_create(worker->base, 1739 worker_sighandler, worker); 1740 if(!worker->comsig 1741 #ifdef SIGHUP 1742 || !comm_signal_bind(worker->comsig, SIGHUP) 1743 #endif 1744 #ifdef SIGQUIT 1745 || !comm_signal_bind(worker->comsig, SIGQUIT) 1746 #endif 1747 || !comm_signal_bind(worker->comsig, SIGTERM) 1748 || !comm_signal_bind(worker->comsig, SIGINT)) { 1749 log_err("could not create signal handlers"); 1750 worker_delete(worker); 1751 return 0; 1752 } 1753 #endif /* LIBEVENT_SIGNAL_PROBLEM */ 1754 if(!daemon_remote_open_accept(worker->daemon->rc, 1755 worker->daemon->rc_ports, worker)) { 1756 worker_delete(worker); 1757 return 0; 1758 } 1759 #ifdef UB_ON_WINDOWS 1760 wsvc_setup_worker(worker); 1761 #endif /* UB_ON_WINDOWS */ 1762 } else { /* !do_sigs */ 1763 worker->comsig = NULL; 1764 } 1765 #ifdef USE_DNSTAP 1766 if(cfg->dnstap) { 1767 log_assert(worker->daemon->dtenv != NULL); 1768 memcpy(&worker->dtenv, worker->daemon->dtenv, sizeof(struct dt_env)); 1769 if(!dt_init(&worker->dtenv, worker->base)) 1770 fatal_exit("dt_init failed"); 1771 } 1772 #endif 1773 worker->front = listen_create(worker->base, ports, 1774 cfg->msg_buffer_size, (int)cfg->incoming_num_tcp, 1775 cfg->do_tcp_keepalive 1776 ? cfg->tcp_keepalive_timeout 1777 : cfg->tcp_idle_timeout, 1778 cfg->harden_large_queries, cfg->http_max_streams, 1779 cfg->http_endpoint, cfg->http_notls_downstream, 1780 worker->daemon->tcl, worker->daemon->listen_sslctx, 1781 dtenv, worker_handle_request, worker); 1782 if(!worker->front) { 1783 log_err("could not create listening sockets"); 1784 worker_delete(worker); 1785 return 0; 1786 } 1787 worker->back = outside_network_create(worker->base, 1788 cfg->msg_buffer_size, (size_t)cfg->outgoing_num_ports, 1789 cfg->out_ifs, cfg->num_out_ifs, cfg->do_ip4, cfg->do_ip6, 1790 cfg->do_tcp?cfg->outgoing_num_tcp:0, cfg->ip_dscp, 1791 worker->daemon->env->infra_cache, worker->rndstate, 1792 cfg->use_caps_bits_for_id, worker->ports, worker->numports, 1793 cfg->unwanted_threshold, cfg->outgoing_tcp_mss, 1794 &worker_alloc_cleanup, worker, 1795 cfg->do_udp || cfg->udp_upstream_without_downstream, 1796 worker->daemon->connect_sslctx, cfg->delay_close, 1797 cfg->tls_use_sni, dtenv, cfg->udp_connect, 1798 cfg->max_reuse_tcp_queries, cfg->tcp_reuse_timeout, 1799 cfg->tcp_auth_query_timeout); 1800 if(!worker->back) { 1801 log_err("could not create outgoing sockets"); 1802 worker_delete(worker); 1803 return 0; 1804 } 1805 iterator_set_ip46_support(&worker->daemon->mods, worker->daemon->env, 1806 worker->back); 1807 /* start listening to commands */ 1808 if(!tube_setup_bg_listen(worker->cmd, worker->base, 1809 &worker_handle_control_cmd, worker)) { 1810 log_err("could not create control compt."); 1811 worker_delete(worker); 1812 return 0; 1813 } 1814 worker->stat_timer = comm_timer_create(worker->base, 1815 worker_stat_timer_cb, worker); 1816 if(!worker->stat_timer) { 1817 log_err("could not create statistics timer"); 1818 } 1819 1820 /* we use the msg_buffer_size as a good estimate for what the 1821 * user wants for memory usage sizes */ 1822 worker->scratchpad = regional_create_custom(cfg->msg_buffer_size); 1823 if(!worker->scratchpad) { 1824 log_err("malloc failure"); 1825 worker_delete(worker); 1826 return 0; 1827 } 1828 1829 server_stats_init(&worker->stats, cfg); 1830 alloc_init(&worker->alloc, &worker->daemon->superalloc, 1831 worker->thread_num); 1832 alloc_set_id_cleanup(&worker->alloc, &worker_alloc_cleanup, worker); 1833 worker->env = *worker->daemon->env; 1834 comm_base_timept(worker->base, &worker->env.now, &worker->env.now_tv); 1835 worker->env.worker = worker; 1836 worker->env.worker_base = worker->base; 1837 worker->env.send_query = &worker_send_query; 1838 worker->env.alloc = &worker->alloc; 1839 worker->env.outnet = worker->back; 1840 worker->env.rnd = worker->rndstate; 1841 /* If case prefetch is triggered, the corresponding mesh will clear 1842 * the scratchpad for the module env in the middle of request handling. 1843 * It would be prone to a use-after-free kind of bug, so we avoid 1844 * sharing it with worker's own scratchpad at the cost of having 1845 * one more pad per worker. */ 1846 worker->env.scratch = regional_create_custom(cfg->msg_buffer_size); 1847 if(!worker->env.scratch) { 1848 log_err("malloc failure"); 1849 worker_delete(worker); 1850 return 0; 1851 } 1852 worker->env.mesh = mesh_create(&worker->daemon->mods, &worker->env); 1853 if(!worker->env.mesh) { 1854 log_err("malloc failure"); 1855 worker_delete(worker); 1856 return 0; 1857 } 1858 /* Pass on daemon variables that we would need in the mesh area */ 1859 worker->env.mesh->use_response_ip = worker->daemon->use_response_ip; 1860 worker->env.mesh->use_rpz = worker->daemon->use_rpz; 1861 1862 worker->env.detach_subs = &mesh_detach_subs; 1863 worker->env.attach_sub = &mesh_attach_sub; 1864 worker->env.add_sub = &mesh_add_sub; 1865 worker->env.kill_sub = &mesh_state_delete; 1866 worker->env.detect_cycle = &mesh_detect_cycle; 1867 worker->env.scratch_buffer = sldns_buffer_new(cfg->msg_buffer_size); 1868 if(!worker->env.scratch_buffer) { 1869 log_err("malloc failure"); 1870 worker_delete(worker); 1871 return 0; 1872 } 1873 if(!(worker->env.fwds = forwards_create()) || 1874 !forwards_apply_cfg(worker->env.fwds, cfg)) { 1875 log_err("Could not set forward zones"); 1876 worker_delete(worker); 1877 return 0; 1878 } 1879 if(!(worker->env.hints = hints_create()) || 1880 !hints_apply_cfg(worker->env.hints, cfg)) { 1881 log_err("Could not set root or stub hints"); 1882 worker_delete(worker); 1883 return 0; 1884 } 1885 /* one probe timer per process -- if we have 5011 anchors */ 1886 if(autr_get_num_anchors(worker->env.anchors) > 0 1887 #ifndef THREADS_DISABLED 1888 && worker->thread_num == 0 1889 #endif 1890 ) { 1891 struct timeval tv; 1892 tv.tv_sec = 0; 1893 tv.tv_usec = 0; 1894 worker->env.probe_timer = comm_timer_create(worker->base, 1895 worker_probe_timer_cb, worker); 1896 if(!worker->env.probe_timer) { 1897 log_err("could not create 5011-probe timer"); 1898 } else { 1899 /* let timer fire, then it can reset itself */ 1900 comm_timer_set(worker->env.probe_timer, &tv); 1901 } 1902 } 1903 /* zone transfer tasks, setup once per process, if any */ 1904 if(worker->env.auth_zones 1905 #ifndef THREADS_DISABLED 1906 && worker->thread_num == 0 1907 #endif 1908 ) { 1909 auth_xfer_pickup_initial(worker->env.auth_zones, &worker->env); 1910 auth_zones_pickup_zonemd_verify(worker->env.auth_zones, 1911 &worker->env); 1912 } 1913 #ifdef USE_DNSTAP 1914 if(worker->daemon->cfg->dnstap 1915 #ifndef THREADS_DISABLED 1916 && worker->thread_num == 0 1917 #endif 1918 ) { 1919 if(!dt_io_thread_start(dtenv->dtio, comm_base_internal( 1920 worker->base), worker->daemon->num)) { 1921 log_err("could not start dnstap io thread"); 1922 worker_delete(worker); 1923 return 0; 1924 } 1925 } 1926 #endif /* USE_DNSTAP */ 1927 worker_mem_report(worker, NULL); 1928 /* if statistics enabled start timer */ 1929 if(worker->env.cfg->stat_interval > 0) { 1930 verbose(VERB_ALGO, "set statistics interval %d secs", 1931 worker->env.cfg->stat_interval); 1932 worker_restart_timer(worker); 1933 } 1934 return 1; 1935 } 1936 1937 void 1938 worker_work(struct worker* worker) 1939 { 1940 comm_base_dispatch(worker->base); 1941 } 1942 1943 void 1944 worker_delete(struct worker* worker) 1945 { 1946 if(!worker) 1947 return; 1948 if(worker->env.mesh && verbosity >= VERB_OPS) { 1949 server_stats_log(&worker->stats, worker, worker->thread_num); 1950 mesh_stats(worker->env.mesh, "mesh has"); 1951 worker_mem_report(worker, NULL); 1952 } 1953 outside_network_quit_prepare(worker->back); 1954 mesh_delete(worker->env.mesh); 1955 sldns_buffer_free(worker->env.scratch_buffer); 1956 forwards_delete(worker->env.fwds); 1957 hints_delete(worker->env.hints); 1958 listen_delete(worker->front); 1959 outside_network_delete(worker->back); 1960 comm_signal_delete(worker->comsig); 1961 tube_delete(worker->cmd); 1962 comm_timer_delete(worker->stat_timer); 1963 comm_timer_delete(worker->env.probe_timer); 1964 free(worker->ports); 1965 if(worker->thread_num == 0) { 1966 #ifdef UB_ON_WINDOWS 1967 wsvc_desetup_worker(worker); 1968 #endif /* UB_ON_WINDOWS */ 1969 } 1970 #ifdef USE_DNSTAP 1971 if(worker->daemon->cfg->dnstap 1972 #ifndef THREADS_DISABLED 1973 && worker->thread_num == 0 1974 #endif 1975 ) { 1976 dt_io_thread_stop(worker->dtenv.dtio); 1977 } 1978 dt_deinit(&worker->dtenv); 1979 #endif /* USE_DNSTAP */ 1980 comm_base_delete(worker->base); 1981 ub_randfree(worker->rndstate); 1982 alloc_clear(&worker->alloc); 1983 regional_destroy(worker->env.scratch); 1984 regional_destroy(worker->scratchpad); 1985 free(worker); 1986 } 1987 1988 struct outbound_entry* 1989 worker_send_query(struct query_info* qinfo, uint16_t flags, int dnssec, 1990 int want_dnssec, int nocaps, struct sockaddr_storage* addr, 1991 socklen_t addrlen, uint8_t* zone, size_t zonelen, int ssl_upstream, 1992 char* tls_auth_name, struct module_qstate* q) 1993 { 1994 struct worker* worker = q->env->worker; 1995 struct outbound_entry* e = (struct outbound_entry*)regional_alloc( 1996 q->region, sizeof(*e)); 1997 if(!e) 1998 return NULL; 1999 e->qstate = q; 2000 e->qsent = outnet_serviced_query(worker->back, qinfo, flags, dnssec, 2001 want_dnssec, nocaps, q->env->cfg->tcp_upstream, 2002 ssl_upstream, tls_auth_name, addr, addrlen, zone, zonelen, q, 2003 worker_handle_service_reply, e, worker->back->udp_buff, q->env); 2004 if(!e->qsent) { 2005 return NULL; 2006 } 2007 return e; 2008 } 2009 2010 void 2011 worker_alloc_cleanup(void* arg) 2012 { 2013 struct worker* worker = (struct worker*)arg; 2014 slabhash_clear(&worker->env.rrset_cache->table); 2015 slabhash_clear(worker->env.msg_cache); 2016 } 2017 2018 void worker_stats_clear(struct worker* worker) 2019 { 2020 server_stats_init(&worker->stats, worker->env.cfg); 2021 mesh_stats_clear(worker->env.mesh); 2022 worker->back->unwanted_replies = 0; 2023 worker->back->num_tcp_outgoing = 0; 2024 } 2025 2026 void worker_start_accept(void* arg) 2027 { 2028 struct worker* worker = (struct worker*)arg; 2029 listen_start_accept(worker->front); 2030 if(worker->thread_num == 0) 2031 daemon_remote_start_accept(worker->daemon->rc); 2032 } 2033 2034 void worker_stop_accept(void* arg) 2035 { 2036 struct worker* worker = (struct worker*)arg; 2037 listen_stop_accept(worker->front); 2038 if(worker->thread_num == 0) 2039 daemon_remote_stop_accept(worker->daemon->rc); 2040 } 2041 2042 /* --- fake callbacks for fptr_wlist to work --- */ 2043 struct outbound_entry* libworker_send_query( 2044 struct query_info* ATTR_UNUSED(qinfo), 2045 uint16_t ATTR_UNUSED(flags), int ATTR_UNUSED(dnssec), 2046 int ATTR_UNUSED(want_dnssec), int ATTR_UNUSED(nocaps), 2047 struct sockaddr_storage* ATTR_UNUSED(addr), socklen_t ATTR_UNUSED(addrlen), 2048 uint8_t* ATTR_UNUSED(zone), size_t ATTR_UNUSED(zonelen), 2049 int ATTR_UNUSED(ssl_upstream), char* ATTR_UNUSED(tls_auth_name), 2050 struct module_qstate* ATTR_UNUSED(q)) 2051 { 2052 log_assert(0); 2053 return 0; 2054 } 2055 2056 int libworker_handle_service_reply(struct comm_point* ATTR_UNUSED(c), 2057 void* ATTR_UNUSED(arg), int ATTR_UNUSED(error), 2058 struct comm_reply* ATTR_UNUSED(reply_info)) 2059 { 2060 log_assert(0); 2061 return 0; 2062 } 2063 2064 void libworker_handle_control_cmd(struct tube* ATTR_UNUSED(tube), 2065 uint8_t* ATTR_UNUSED(buffer), size_t ATTR_UNUSED(len), 2066 int ATTR_UNUSED(error), void* ATTR_UNUSED(arg)) 2067 { 2068 log_assert(0); 2069 } 2070 2071 void libworker_fg_done_cb(void* ATTR_UNUSED(arg), int ATTR_UNUSED(rcode), 2072 sldns_buffer* ATTR_UNUSED(buf), enum sec_status ATTR_UNUSED(s), 2073 char* ATTR_UNUSED(why_bogus), int ATTR_UNUSED(was_ratelimited)) 2074 { 2075 log_assert(0); 2076 } 2077 2078 void libworker_bg_done_cb(void* ATTR_UNUSED(arg), int ATTR_UNUSED(rcode), 2079 sldns_buffer* ATTR_UNUSED(buf), enum sec_status ATTR_UNUSED(s), 2080 char* ATTR_UNUSED(why_bogus), int ATTR_UNUSED(was_ratelimited)) 2081 { 2082 log_assert(0); 2083 } 2084 2085 void libworker_event_done_cb(void* ATTR_UNUSED(arg), int ATTR_UNUSED(rcode), 2086 sldns_buffer* ATTR_UNUSED(buf), enum sec_status ATTR_UNUSED(s), 2087 char* ATTR_UNUSED(why_bogus), int ATTR_UNUSED(was_ratelimited)) 2088 { 2089 log_assert(0); 2090 } 2091 2092 int context_query_cmp(const void* ATTR_UNUSED(a), const void* ATTR_UNUSED(b)) 2093 { 2094 log_assert(0); 2095 return 0; 2096 } 2097 2098 int order_lock_cmp(const void* ATTR_UNUSED(e1), const void* ATTR_UNUSED(e2)) 2099 { 2100 log_assert(0); 2101 return 0; 2102 } 2103 2104 int codeline_cmp(const void* ATTR_UNUSED(a), const void* ATTR_UNUSED(b)) 2105 { 2106 log_assert(0); 2107 return 0; 2108 } 2109 2110 #ifdef USE_DNSTAP 2111 void dtio_tap_callback(int ATTR_UNUSED(fd), short ATTR_UNUSED(ev), 2112 void* ATTR_UNUSED(arg)) 2113 { 2114 log_assert(0); 2115 } 2116 #endif 2117 2118 #ifdef USE_DNSTAP 2119 void dtio_mainfdcallback(int ATTR_UNUSED(fd), short ATTR_UNUSED(ev), 2120 void* ATTR_UNUSED(arg)) 2121 { 2122 log_assert(0); 2123 } 2124 #endif 2125