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#include "dnstap/dtstream.h"
83
84#ifdef HAVE_SYS_TYPES_H
85#  include <sys/types.h>
86#endif
87#ifdef HAVE_NETDB_H
88#include <netdb.h>
89#endif
90#include <signal.h>
91#ifdef UB_ON_WINDOWS
92#include "winrc/win_svc.h"
93#endif
94
95/** Size of an UDP datagram */
96#define NORMAL_UDP_SIZE	512 /* bytes */
97/** ratelimit for error responses */
98#define ERROR_RATELIMIT 100 /* qps */
99
100/**
101 * seconds to add to prefetch leeway.  This is a TTL that expires old rrsets
102 * earlier than they should in order to put the new update into the cache.
103 * This additional value is to make sure that if not all TTLs are equal in
104 * the message to be updated(and replaced), that rrsets with up to this much
105 * extra TTL are also replaced.  This means that the resulting new message
106 * will have (most likely) this TTL at least, avoiding very small 'split
107 * second' TTLs due to operators choosing relative primes for TTLs (or so).
108 * Also has to be at least one to break ties (and overwrite cached entry).
109 */
110#define PREFETCH_EXPIRY_ADD 60
111
112/** Report on memory usage by this thread and global */
113static void
114worker_mem_report(struct worker* ATTR_UNUSED(worker),
115	struct serviced_query* ATTR_UNUSED(cur_serv))
116{
117#ifdef UNBOUND_ALLOC_STATS
118	/* measure memory leakage */
119	extern size_t unbound_mem_alloc, unbound_mem_freed;
120	/* debug func in validator module */
121	size_t total, front, back, mesh, msg, rrset, infra, ac, superac;
122	size_t me, iter, val, anch;
123	int i;
124#ifdef CLIENT_SUBNET
125	size_t subnet = 0;
126#endif /* CLIENT_SUBNET */
127	if(verbosity < VERB_ALGO)
128		return;
129	front = listen_get_mem(worker->front);
130	back = outnet_get_mem(worker->back);
131	msg = slabhash_get_mem(worker->env.msg_cache);
132	rrset = slabhash_get_mem(&worker->env.rrset_cache->table);
133	infra = infra_get_mem(worker->env.infra_cache);
134	mesh = mesh_get_mem(worker->env.mesh);
135	ac = alloc_get_mem(&worker->alloc);
136	superac = alloc_get_mem(&worker->daemon->superalloc);
137	anch = anchors_get_mem(worker->env.anchors);
138	iter = 0;
139	val = 0;
140	for(i=0; i<worker->env.mesh->mods.num; i++) {
141		fptr_ok(fptr_whitelist_mod_get_mem(worker->env.mesh->
142			mods.mod[i]->get_mem));
143		if(strcmp(worker->env.mesh->mods.mod[i]->name, "validator")==0)
144			val += (*worker->env.mesh->mods.mod[i]->get_mem)
145				(&worker->env, i);
146#ifdef CLIENT_SUBNET
147		else if(strcmp(worker->env.mesh->mods.mod[i]->name,
148			"subnet")==0)
149			subnet += (*worker->env.mesh->mods.mod[i]->get_mem)
150				(&worker->env, i);
151#endif /* CLIENT_SUBNET */
152		else	iter += (*worker->env.mesh->mods.mod[i]->get_mem)
153				(&worker->env, i);
154	}
155	me = sizeof(*worker) + sizeof(*worker->base) + sizeof(*worker->comsig)
156		+ comm_point_get_mem(worker->cmd_com)
157		+ sizeof(worker->rndstate)
158		+ regional_get_mem(worker->scratchpad)
159		+ sizeof(*worker->env.scratch_buffer)
160		+ sldns_buffer_capacity(worker->env.scratch_buffer)
161		+ forwards_get_mem(worker->env.fwds)
162		+ hints_get_mem(worker->env.hints);
163	if(worker->thread_num == 0)
164		me += acl_list_get_mem(worker->daemon->acl);
165	if(cur_serv) {
166		me += serviced_get_mem(cur_serv);
167	}
168	total = front+back+mesh+msg+rrset+infra+iter+val+ac+superac+me;
169#ifdef CLIENT_SUBNET
170	total += subnet;
171	log_info("Memory conditions: %u front=%u back=%u mesh=%u msg=%u "
172		"rrset=%u infra=%u iter=%u val=%u subnet=%u anchors=%u "
173		"alloccache=%u globalalloccache=%u me=%u",
174		(unsigned)total, (unsigned)front, (unsigned)back,
175		(unsigned)mesh, (unsigned)msg, (unsigned)rrset, (unsigned)infra,
176		(unsigned)iter, (unsigned)val,
177		(unsigned)subnet, (unsigned)anch, (unsigned)ac,
178		(unsigned)superac, (unsigned)me);
179#else /* no CLIENT_SUBNET */
180	log_info("Memory conditions: %u front=%u back=%u mesh=%u msg=%u "
181		"rrset=%u infra=%u iter=%u val=%u anchors=%u "
182		"alloccache=%u globalalloccache=%u me=%u",
183		(unsigned)total, (unsigned)front, (unsigned)back,
184		(unsigned)mesh, (unsigned)msg, (unsigned)rrset,
185		(unsigned)infra, (unsigned)iter, (unsigned)val, (unsigned)anch,
186		(unsigned)ac, (unsigned)superac, (unsigned)me);
187#endif /* CLIENT_SUBNET */
188	log_info("Total heap memory estimate: %u  total-alloc: %u  "
189		"total-free: %u", (unsigned)total,
190		(unsigned)unbound_mem_alloc, (unsigned)unbound_mem_freed);
191#else /* no UNBOUND_ALLOC_STATS */
192	size_t val = 0;
193#ifdef CLIENT_SUBNET
194	size_t subnet = 0;
195#endif /* CLIENT_SUBNET */
196	int i;
197	if(verbosity < VERB_QUERY)
198		return;
199	for(i=0; i<worker->env.mesh->mods.num; i++) {
200		fptr_ok(fptr_whitelist_mod_get_mem(worker->env.mesh->
201			mods.mod[i]->get_mem));
202		if(strcmp(worker->env.mesh->mods.mod[i]->name, "validator")==0)
203			val += (*worker->env.mesh->mods.mod[i]->get_mem)
204				(&worker->env, i);
205#ifdef CLIENT_SUBNET
206		else if(strcmp(worker->env.mesh->mods.mod[i]->name,
207			"subnet")==0)
208			subnet += (*worker->env.mesh->mods.mod[i]->get_mem)
209				(&worker->env, i);
210#endif /* CLIENT_SUBNET */
211	}
212#ifdef CLIENT_SUBNET
213	verbose(VERB_QUERY, "cache memory msg=%u rrset=%u infra=%u val=%u "
214		"subnet=%u",
215		(unsigned)slabhash_get_mem(worker->env.msg_cache),
216		(unsigned)slabhash_get_mem(&worker->env.rrset_cache->table),
217		(unsigned)infra_get_mem(worker->env.infra_cache),
218		(unsigned)val, (unsigned)subnet);
219#else /* no CLIENT_SUBNET */
220	verbose(VERB_QUERY, "cache memory msg=%u rrset=%u infra=%u val=%u",
221		(unsigned)slabhash_get_mem(worker->env.msg_cache),
222		(unsigned)slabhash_get_mem(&worker->env.rrset_cache->table),
223		(unsigned)infra_get_mem(worker->env.infra_cache),
224		(unsigned)val);
225#endif /* CLIENT_SUBNET */
226#endif /* UNBOUND_ALLOC_STATS */
227}
228
229void
230worker_send_cmd(struct worker* worker, enum worker_commands cmd)
231{
232	uint32_t c = (uint32_t)htonl(cmd);
233	if(!tube_write_msg(worker->cmd, (uint8_t*)&c, sizeof(c), 0)) {
234		log_err("worker send cmd %d failed", (int)cmd);
235	}
236}
237
238int
239worker_handle_reply(struct comm_point* c, void* arg, int error,
240	struct comm_reply* reply_info)
241{
242	struct module_qstate* q = (struct module_qstate*)arg;
243	struct worker* worker = q->env->worker;
244	struct outbound_entry e;
245	e.qstate = q;
246	e.qsent = NULL;
247
248	if(error != 0) {
249		mesh_report_reply(worker->env.mesh, &e, reply_info, error);
250		worker_mem_report(worker, NULL);
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		mesh_report_reply(worker->env.mesh, &e, reply_info,
261			NETEVENT_TIMEOUT);
262		worker_mem_report(worker, NULL);
263		return 0;
264	}
265	mesh_report_reply(worker->env.mesh, &e, reply_info, NETEVENT_NOERROR);
266	worker_mem_report(worker, NULL);
267	return 0;
268}
269
270int
271worker_handle_service_reply(struct comm_point* c, void* arg, int error,
272	struct comm_reply* reply_info)
273{
274	struct outbound_entry* e = (struct outbound_entry*)arg;
275	struct worker* worker = e->qstate->env->worker;
276	struct serviced_query *sq = e->qsent;
277
278	verbose(VERB_ALGO, "worker svcd callback for qstate %p", e->qstate);
279	if(error != 0) {
280		mesh_report_reply(worker->env.mesh, e, reply_info, error);
281		worker_mem_report(worker, sq);
282		return 0;
283	}
284	/* sanity check. */
285	if(!LDNS_QR_WIRE(sldns_buffer_begin(c->buffer))
286		|| LDNS_OPCODE_WIRE(sldns_buffer_begin(c->buffer)) !=
287			LDNS_PACKET_QUERY
288		|| LDNS_QDCOUNT(sldns_buffer_begin(c->buffer)) > 1) {
289		/* error becomes timeout for the module as if this reply
290		 * never arrived. */
291		verbose(VERB_ALGO, "worker: bad reply handled as timeout");
292		mesh_report_reply(worker->env.mesh, e, reply_info,
293			NETEVENT_TIMEOUT);
294		worker_mem_report(worker, sq);
295		return 0;
296	}
297	mesh_report_reply(worker->env.mesh, e, reply_info, NETEVENT_NOERROR);
298	worker_mem_report(worker, sq);
299	return 0;
300}
301
302/** ratelimit error replies
303 * @param worker: the worker struct with ratelimit counter
304 * @param err: error code that would be wanted.
305 * @return value of err if okay, or -1 if it should be discarded instead.
306 */
307static int
308worker_err_ratelimit(struct worker* worker, int err)
309{
310	if(worker->err_limit_time == *worker->env.now) {
311		/* see if limit is exceeded for this second */
312		if(worker->err_limit_count++ > ERROR_RATELIMIT)
313			return -1;
314	} else {
315		/* new second, new limits */
316		worker->err_limit_time = *worker->env.now;
317		worker->err_limit_count = 1;
318	}
319	return err;
320}
321
322/** check request sanity.
323 * @param pkt: the wire packet to examine for sanity.
324 * @param worker: parameters for checking.
325 * @return error code, 0 OK, or -1 discard.
326*/
327static int
328worker_check_request(sldns_buffer* pkt, struct worker* worker)
329{
330	if(sldns_buffer_limit(pkt) < LDNS_HEADER_SIZE) {
331		verbose(VERB_QUERY, "request too short, discarded");
332		return -1;
333	}
334	if(sldns_buffer_limit(pkt) > NORMAL_UDP_SIZE &&
335		worker->daemon->cfg->harden_large_queries) {
336		verbose(VERB_QUERY, "request too large, discarded");
337		return -1;
338	}
339	if(LDNS_QR_WIRE(sldns_buffer_begin(pkt))) {
340		verbose(VERB_QUERY, "request has QR bit on, discarded");
341		return -1;
342	}
343	if(LDNS_TC_WIRE(sldns_buffer_begin(pkt))) {
344		LDNS_TC_CLR(sldns_buffer_begin(pkt));
345		verbose(VERB_QUERY, "request bad, has TC bit on");
346		return worker_err_ratelimit(worker, LDNS_RCODE_FORMERR);
347	}
348	if(LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt)) != LDNS_PACKET_QUERY &&
349		LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt)) != LDNS_PACKET_NOTIFY) {
350		verbose(VERB_QUERY, "request unknown opcode %d",
351			LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt)));
352		return worker_err_ratelimit(worker, LDNS_RCODE_NOTIMPL);
353	}
354	if(LDNS_QDCOUNT(sldns_buffer_begin(pkt)) != 1) {
355		verbose(VERB_QUERY, "request wrong nr qd=%d",
356			LDNS_QDCOUNT(sldns_buffer_begin(pkt)));
357		return worker_err_ratelimit(worker, LDNS_RCODE_FORMERR);
358	}
359	if(LDNS_ANCOUNT(sldns_buffer_begin(pkt)) != 0 &&
360		(LDNS_ANCOUNT(sldns_buffer_begin(pkt)) != 1 ||
361		LDNS_OPCODE_WIRE(sldns_buffer_begin(pkt)) != LDNS_PACKET_NOTIFY)) {
362		verbose(VERB_QUERY, "request wrong nr an=%d",
363			LDNS_ANCOUNT(sldns_buffer_begin(pkt)));
364		return worker_err_ratelimit(worker, LDNS_RCODE_FORMERR);
365	}
366	if(LDNS_NSCOUNT(sldns_buffer_begin(pkt)) != 0) {
367		verbose(VERB_QUERY, "request wrong nr ns=%d",
368			LDNS_NSCOUNT(sldns_buffer_begin(pkt)));
369		return worker_err_ratelimit(worker, LDNS_RCODE_FORMERR);
370	}
371	if(LDNS_ARCOUNT(sldns_buffer_begin(pkt)) > 1) {
372		verbose(VERB_QUERY, "request wrong nr ar=%d",
373			LDNS_ARCOUNT(sldns_buffer_begin(pkt)));
374		return worker_err_ratelimit(worker, LDNS_RCODE_FORMERR);
375	}
376	return 0;
377}
378
379void
380worker_handle_control_cmd(struct tube* ATTR_UNUSED(tube), uint8_t* msg,
381	size_t len, int error, void* arg)
382{
383	struct worker* worker = (struct worker*)arg;
384	enum worker_commands cmd;
385	if(error != NETEVENT_NOERROR) {
386		free(msg);
387		if(error == NETEVENT_CLOSED)
388			comm_base_exit(worker->base);
389		else	log_info("control event: %d", error);
390		return;
391	}
392	if(len != sizeof(uint32_t)) {
393		fatal_exit("bad control msg length %d", (int)len);
394	}
395	cmd = sldns_read_uint32(msg);
396	free(msg);
397	switch(cmd) {
398	case worker_cmd_quit:
399		verbose(VERB_ALGO, "got control cmd quit");
400		comm_base_exit(worker->base);
401		break;
402	case worker_cmd_stats:
403		verbose(VERB_ALGO, "got control cmd stats");
404		server_stats_reply(worker, 1);
405		break;
406	case worker_cmd_stats_noreset:
407		verbose(VERB_ALGO, "got control cmd stats_noreset");
408		server_stats_reply(worker, 0);
409		break;
410	case worker_cmd_remote:
411		verbose(VERB_ALGO, "got control cmd remote");
412		daemon_remote_exec(worker);
413		break;
414	default:
415		log_err("bad command %d", (int)cmd);
416		break;
417	}
418}
419
420/** check if a delegation is secure */
421static enum sec_status
422check_delegation_secure(struct reply_info *rep)
423{
424	/* return smallest security status */
425	size_t i;
426	enum sec_status sec = sec_status_secure;
427	enum sec_status s;
428	size_t num = rep->an_numrrsets + rep->ns_numrrsets;
429	/* check if answer and authority are OK */
430	for(i=0; i<num; i++) {
431		s = ((struct packed_rrset_data*)rep->rrsets[i]->entry.data)
432			->security;
433		if(s < sec)
434			sec = s;
435	}
436	/* in additional, only unchecked triggers revalidation */
437	for(i=num; i<rep->rrset_count; i++) {
438		s = ((struct packed_rrset_data*)rep->rrsets[i]->entry.data)
439			->security;
440		if(s == sec_status_unchecked)
441			return s;
442	}
443	return sec;
444}
445
446/** remove nonsecure from a delegation referral additional section */
447static void
448deleg_remove_nonsecure_additional(struct reply_info* rep)
449{
450	/* we can simply edit it, since we are working in the scratch region */
451	size_t i;
452	enum sec_status s;
453
454	for(i = rep->an_numrrsets+rep->ns_numrrsets; i<rep->rrset_count; i++) {
455		s = ((struct packed_rrset_data*)rep->rrsets[i]->entry.data)
456			->security;
457		if(s != sec_status_secure) {
458			memmove(rep->rrsets+i, rep->rrsets+i+1,
459				sizeof(struct ub_packed_rrset_key*)*
460				(rep->rrset_count - i - 1));
461			rep->ar_numrrsets--;
462			rep->rrset_count--;
463			i--;
464		}
465	}
466}
467
468/** answer nonrecursive query from the cache */
469static int
470answer_norec_from_cache(struct worker* worker, struct query_info* qinfo,
471	uint16_t id, uint16_t flags, struct comm_reply* repinfo,
472	struct edns_data* edns)
473{
474	/* for a nonrecursive query return either:
475	 * 	o an error (servfail; we try to avoid this)
476	 * 	o a delegation (closest we have; this routine tries that)
477	 * 	o the answer (checked by answer_from_cache)
478	 *
479	 * So, grab a delegation from the rrset cache.
480	 * Then check if it needs validation, if so, this routine fails,
481	 * so that iterator can prime and validator can verify rrsets.
482	 */
483	struct edns_data edns_bak;
484	uint16_t udpsize = edns->udp_size;
485	int secure = 0;
486	time_t timenow = *worker->env.now;
487	int must_validate = (!(flags&BIT_CD) || worker->env.cfg->ignore_cd)
488		&& worker->env.need_to_validate;
489	struct dns_msg *msg = NULL;
490	struct delegpt *dp;
491
492	dp = dns_cache_find_delegation(&worker->env, qinfo->qname,
493		qinfo->qname_len, qinfo->qtype, qinfo->qclass,
494		worker->scratchpad, &msg, timenow);
495	if(!dp) { /* no delegation, need to reprime */
496		return 0;
497	}
498	/* In case we have a local alias, copy it into the delegation message.
499	 * Shallow copy should be fine, as we'll be done with msg in this
500	 * function. */
501	msg->qinfo.local_alias = qinfo->local_alias;
502	if(must_validate) {
503		switch(check_delegation_secure(msg->rep)) {
504		case sec_status_unchecked:
505			/* some rrsets have not been verified yet, go and
506			 * let validator do that */
507			return 0;
508		case sec_status_bogus:
509		case sec_status_secure_sentinel_fail:
510			/* some rrsets are bogus, reply servfail */
511			edns->edns_version = EDNS_ADVERTISED_VERSION;
512			edns->udp_size = EDNS_ADVERTISED_SIZE;
513			edns->ext_rcode = 0;
514			edns->bits &= EDNS_DO;
515			if(!inplace_cb_reply_servfail_call(&worker->env, qinfo, NULL,
516				msg->rep, LDNS_RCODE_SERVFAIL, edns, repinfo, worker->scratchpad,
517				worker->env.now_tv))
518					return 0;
519			error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL,
520				&msg->qinfo, id, flags, edns);
521			if(worker->stats.extended) {
522				worker->stats.ans_bogus++;
523				worker->stats.ans_rcode[LDNS_RCODE_SERVFAIL]++;
524			}
525			return 1;
526		case sec_status_secure:
527			/* all rrsets are secure */
528			/* remove non-secure rrsets from the add. section*/
529			if(worker->env.cfg->val_clean_additional)
530				deleg_remove_nonsecure_additional(msg->rep);
531			secure = 1;
532			break;
533		case sec_status_indeterminate:
534		case sec_status_insecure:
535		default:
536			/* not secure */
537			secure = 0;
538			break;
539		}
540	}
541	/* return this delegation from the cache */
542	edns_bak = *edns;
543	edns->edns_version = EDNS_ADVERTISED_VERSION;
544	edns->udp_size = EDNS_ADVERTISED_SIZE;
545	edns->ext_rcode = 0;
546	edns->bits &= EDNS_DO;
547	if(!inplace_cb_reply_cache_call(&worker->env, qinfo, NULL, msg->rep,
548		(int)(flags&LDNS_RCODE_MASK), edns, repinfo, worker->scratchpad,
549		worker->env.now_tv))
550			return 0;
551	msg->rep->flags |= BIT_QR|BIT_RA;
552	if(!apply_edns_options(edns, &edns_bak, worker->env.cfg,
553		repinfo->c, worker->scratchpad) ||
554		!reply_info_answer_encode(&msg->qinfo, msg->rep, id, flags,
555		repinfo->c->buffer, 0, 1, worker->scratchpad,
556		udpsize, edns, (int)(edns->bits & EDNS_DO), secure)) {
557		if(!inplace_cb_reply_servfail_call(&worker->env, qinfo, NULL, NULL,
558			LDNS_RCODE_SERVFAIL, edns, repinfo, worker->scratchpad,
559			worker->env.now_tv))
560				edns->opt_list = NULL;
561		error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL,
562			&msg->qinfo, id, flags, edns);
563	}
564	if(worker->stats.extended) {
565		if(secure) worker->stats.ans_secure++;
566		server_stats_insrcode(&worker->stats, repinfo->c->buffer);
567	}
568	return 1;
569}
570
571/** Apply, if applicable, a response IP action to a cached answer.
572 * If the answer is rewritten as a result of an action, '*encode_repp' will
573 * point to the reply info containing the modified answer.  '*encode_repp' will
574 * be intact otherwise.
575 * It returns 1 on success, 0 otherwise. */
576static int
577apply_respip_action(struct worker* worker, const struct query_info* qinfo,
578	struct respip_client_info* cinfo, struct reply_info* rep,
579	struct comm_reply* repinfo, struct ub_packed_rrset_key** alias_rrset,
580	struct reply_info** encode_repp, struct auth_zones* az)
581{
582	struct respip_action_info actinfo = {0, 0, 0, 0, NULL, 0, NULL};
583	actinfo.action = respip_none;
584
585	if(qinfo->qtype != LDNS_RR_TYPE_A &&
586		qinfo->qtype != LDNS_RR_TYPE_AAAA &&
587		qinfo->qtype != LDNS_RR_TYPE_ANY)
588		return 1;
589
590	if(!respip_rewrite_reply(qinfo, cinfo, rep, encode_repp, &actinfo,
591		alias_rrset, 0, worker->scratchpad, az))
592		return 0;
593
594	/* xxx_deny actions mean dropping the reply, unless the original reply
595	 * was redirected to response-ip data. */
596	if((actinfo.action == respip_deny ||
597		actinfo.action == respip_inform_deny) &&
598		*encode_repp == rep)
599		*encode_repp = NULL;
600
601	/* If address info is returned, it means the action should be an
602	 * 'inform' variant and the information should be logged. */
603	if(actinfo.addrinfo) {
604		respip_inform_print(&actinfo, qinfo->qname,
605			qinfo->qtype, qinfo->qclass, qinfo->local_alias,
606			repinfo);
607
608		if(worker->stats.extended && actinfo.rpz_used) {
609			if(actinfo.rpz_disabled)
610				worker->stats.rpz_action[RPZ_DISABLED_ACTION]++;
611			if(actinfo.rpz_cname_override)
612				worker->stats.rpz_action[RPZ_CNAME_OVERRIDE_ACTION]++;
613			else
614				worker->stats.rpz_action[
615					respip_action_to_rpz_action(actinfo.action)]++;
616		}
617	}
618
619	return 1;
620}
621
622/** answer query from the cache.
623 * Normally, the answer message will be built in repinfo->c->buffer; if the
624 * answer is supposed to be suppressed or the answer is supposed to be an
625 * incomplete CNAME chain, the buffer is explicitly cleared to signal the
626 * caller as such.  In the latter case *partial_rep will point to the incomplete
627 * reply, and this function is (possibly) supposed to be called again with that
628 * *partial_rep value to complete the chain.  In addition, if the query should
629 * be completely dropped, '*need_drop' will be set to 1. */
630static int
631answer_from_cache(struct worker* worker, struct query_info* qinfo,
632	struct respip_client_info* cinfo, int* need_drop, int* is_expired_answer,
633	int* is_secure_answer, struct ub_packed_rrset_key** alias_rrset,
634	struct reply_info** partial_repp,
635	struct reply_info* rep, uint16_t id, uint16_t flags,
636	struct comm_reply* repinfo, struct edns_data* edns)
637{
638	struct edns_data edns_bak;
639	time_t timenow = *worker->env.now;
640	uint16_t udpsize = edns->udp_size;
641	struct reply_info* encode_rep = rep;
642	struct reply_info* partial_rep = *partial_repp;
643	int must_validate = (!(flags&BIT_CD) || worker->env.cfg->ignore_cd)
644		&& worker->env.need_to_validate;
645	*partial_repp = NULL;  /* avoid accidental further pass */
646
647	/* Check TTL */
648	if(rep->ttl < timenow) {
649		/* Check if we need to serve expired now */
650		if(worker->env.cfg->serve_expired &&
651			!worker->env.cfg->serve_expired_client_timeout) {
652				if(worker->env.cfg->serve_expired_ttl &&
653					rep->serve_expired_ttl < timenow)
654					return 0;
655				if(!rrset_array_lock(rep->ref, rep->rrset_count, 0))
656					return 0;
657				*is_expired_answer = 1;
658		} else {
659			/* the rrsets may have been updated in the meantime.
660			 * we will refetch the message format from the
661			 * authoritative server
662			 */
663			return 0;
664		}
665	} else {
666		if(!rrset_array_lock(rep->ref, rep->rrset_count, timenow))
667			return 0;
668	}
669	/* locked and ids and ttls are OK. */
670
671	/* check CNAME chain (if any) */
672	if(rep->an_numrrsets > 0 && (rep->rrsets[0]->rk.type ==
673		htons(LDNS_RR_TYPE_CNAME) || rep->rrsets[0]->rk.type ==
674		htons(LDNS_RR_TYPE_DNAME))) {
675		if(!reply_check_cname_chain(qinfo, rep)) {
676			/* cname chain invalid, redo iterator steps */
677			verbose(VERB_ALGO, "Cache reply: cname chain broken");
678			goto bail_out;
679		}
680	}
681	/* check security status of the cached answer */
682	if(must_validate && (rep->security == sec_status_bogus ||
683		rep->security == sec_status_secure_sentinel_fail)) {
684		/* BAD cached */
685		edns->edns_version = EDNS_ADVERTISED_VERSION;
686		edns->udp_size = EDNS_ADVERTISED_SIZE;
687		edns->ext_rcode = 0;
688		edns->bits &= EDNS_DO;
689		if(!inplace_cb_reply_servfail_call(&worker->env, qinfo, NULL, rep,
690			LDNS_RCODE_SERVFAIL, edns, repinfo, worker->scratchpad,
691			worker->env.now_tv))
692			goto bail_out;
693		error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL,
694			qinfo, id, flags, edns);
695		rrset_array_unlock_touch(worker->env.rrset_cache,
696			worker->scratchpad, rep->ref, rep->rrset_count);
697		if(worker->stats.extended) {
698			worker->stats.ans_bogus ++;
699			worker->stats.ans_rcode[LDNS_RCODE_SERVFAIL] ++;
700		}
701		return 1;
702	} else if(rep->security == sec_status_unchecked && must_validate) {
703		verbose(VERB_ALGO, "Cache reply: unchecked entry needs "
704			"validation");
705		goto bail_out; /* need to validate cache entry first */
706	} else if(rep->security == sec_status_secure) {
707		if(reply_all_rrsets_secure(rep)) {
708			*is_secure_answer = 1;
709		} else {
710			if(must_validate) {
711				verbose(VERB_ALGO, "Cache reply: secure entry"
712					" changed status");
713				goto bail_out; /* rrset changed, re-verify */
714			}
715			*is_secure_answer = 0;
716		}
717	} else *is_secure_answer = 0;
718
719	edns_bak = *edns;
720	edns->edns_version = EDNS_ADVERTISED_VERSION;
721	edns->udp_size = EDNS_ADVERTISED_SIZE;
722	edns->ext_rcode = 0;
723	edns->bits &= EDNS_DO;
724	if(!inplace_cb_reply_cache_call(&worker->env, qinfo, NULL, rep,
725		(int)(flags&LDNS_RCODE_MASK), edns, repinfo, worker->scratchpad,
726		worker->env.now_tv))
727		goto bail_out;
728	*alias_rrset = NULL; /* avoid confusion if caller set it to non-NULL */
729	if((worker->daemon->use_response_ip || worker->daemon->use_rpz) &&
730		!partial_rep && !apply_respip_action(worker, qinfo, cinfo, rep,
731		repinfo, alias_rrset,
732		&encode_rep, worker->env.auth_zones)) {
733		goto bail_out;
734	} else if(partial_rep &&
735		!respip_merge_cname(partial_rep, qinfo, rep, cinfo,
736		must_validate, &encode_rep, worker->scratchpad,
737		worker->env.auth_zones)) {
738		goto bail_out;
739	}
740	if(encode_rep != rep) {
741		/* if rewritten, it can't be considered "secure" */
742		*is_secure_answer = 0;
743	}
744	if(!encode_rep || *alias_rrset) {
745		if(!encode_rep)
746			*need_drop = 1;
747		else {
748			/* If a partial CNAME chain is found, we first need to
749			 * make a copy of the reply in the scratchpad so we
750			 * can release the locks and lookup the cache again. */
751			*partial_repp = reply_info_copy(encode_rep, NULL,
752				worker->scratchpad);
753			if(!*partial_repp)
754				goto bail_out;
755		}
756	} else if(!apply_edns_options(edns, &edns_bak, worker->env.cfg,
757		repinfo->c, worker->scratchpad) ||
758		!reply_info_answer_encode(qinfo, encode_rep, id, flags,
759		repinfo->c->buffer, timenow, 1, worker->scratchpad,
760		udpsize, edns, (int)(edns->bits & EDNS_DO), *is_secure_answer)) {
761		if(!inplace_cb_reply_servfail_call(&worker->env, qinfo, NULL, NULL,
762			LDNS_RCODE_SERVFAIL, edns, repinfo, worker->scratchpad,
763			worker->env.now_tv))
764				edns->opt_list = NULL;
765		error_encode(repinfo->c->buffer, LDNS_RCODE_SERVFAIL,
766			qinfo, id, flags, edns);
767	}
768	/* cannot send the reply right now, because blocking network syscall
769	 * is bad while holding locks. */
770	rrset_array_unlock_touch(worker->env.rrset_cache, worker->scratchpad,
771		rep->ref, rep->rrset_count);
772	/* go and return this buffer to the client */
773	return 1;
774
775bail_out:
776	rrset_array_unlock_touch(worker->env.rrset_cache,
777		worker->scratchpad, rep->ref, rep->rrset_count);
778	return 0;
779}
780
781/** Reply to client and perform prefetch to keep cache up to date. */
782static void
783reply_and_prefetch(struct worker* worker, struct query_info* qinfo,
784	uint16_t flags, struct comm_reply* repinfo, time_t leeway, int noreply)
785{
786	/* first send answer to client to keep its latency
787	 * as small as a cachereply */
788	if(!noreply) {
789		if(repinfo->c->tcp_req_info) {
790			sldns_buffer_copy(
791				repinfo->c->tcp_req_info->spool_buffer,
792				repinfo->c->buffer);
793		}
794		comm_point_send_reply(repinfo);
795	}
796	server_stats_prefetch(&worker->stats, worker);
797
798	/* create the prefetch in the mesh as a normal lookup without
799	 * client addrs waiting, which has the cache blacklisted (to bypass
800	 * the cache and go to the network for the data). */
801	/* this (potentially) runs the mesh for the new query */
802	mesh_new_prefetch(worker->env.mesh, qinfo, flags, leeway +
803		PREFETCH_EXPIRY_ADD);
804}
805
806/**
807 * Fill CH class answer into buffer. Keeps query.
808 * @param pkt: buffer
809 * @param str: string to put into text record (<255).
810 * 	array of strings, every string becomes a text record.
811 * @param num: number of strings in array.
812 * @param edns: edns reply information.
813 * @param worker: worker with scratch region.
814 * @param repinfo: reply information for a communication point.
815 */
816static void
817chaos_replystr(sldns_buffer* pkt, char** str, int num, struct edns_data* edns,
818	struct worker* worker, struct comm_reply* repinfo)
819{
820	int i;
821	unsigned int rd = LDNS_RD_WIRE(sldns_buffer_begin(pkt));
822	unsigned int cd = LDNS_CD_WIRE(sldns_buffer_begin(pkt));
823	sldns_buffer_clear(pkt);
824	sldns_buffer_skip(pkt, (ssize_t)sizeof(uint16_t)); /* skip id */
825	sldns_buffer_write_u16(pkt, (uint16_t)(BIT_QR|BIT_RA));
826	if(rd) LDNS_RD_SET(sldns_buffer_begin(pkt));
827	if(cd) LDNS_CD_SET(sldns_buffer_begin(pkt));
828	sldns_buffer_write_u16(pkt, 1); /* qdcount */
829	sldns_buffer_write_u16(pkt, (uint16_t)num); /* ancount */
830	sldns_buffer_write_u16(pkt, 0); /* nscount */
831	sldns_buffer_write_u16(pkt, 0); /* arcount */
832	(void)query_dname_len(pkt); /* skip qname */
833	sldns_buffer_skip(pkt, (ssize_t)sizeof(uint16_t)); /* skip qtype */
834	sldns_buffer_skip(pkt, (ssize_t)sizeof(uint16_t)); /* skip qclass */
835	for(i=0; i<num; i++) {
836		size_t len = strlen(str[i]);
837		if(len>255) len=255; /* cap size of TXT record */
838		sldns_buffer_write_u16(pkt, 0xc00c); /* compr ptr to query */
839		sldns_buffer_write_u16(pkt, LDNS_RR_TYPE_TXT);
840		sldns_buffer_write_u16(pkt, LDNS_RR_CLASS_CH);
841		sldns_buffer_write_u32(pkt, 0); /* TTL */
842		sldns_buffer_write_u16(pkt, sizeof(uint8_t) + len);
843		sldns_buffer_write_u8(pkt, len);
844		sldns_buffer_write(pkt, str[i], len);
845	}
846	sldns_buffer_flip(pkt);
847	edns->edns_version = EDNS_ADVERTISED_VERSION;
848	edns->udp_size = EDNS_ADVERTISED_SIZE;
849	edns->bits &= EDNS_DO;
850	if(!inplace_cb_reply_local_call(&worker->env, NULL, NULL, NULL,
851		LDNS_RCODE_NOERROR, edns, repinfo, worker->scratchpad,
852		worker->env.now_tv))
853			edns->opt_list = NULL;
854	if(sldns_buffer_capacity(pkt) >=
855		sldns_buffer_limit(pkt)+calc_edns_field_size(edns))
856		attach_edns_record(pkt, edns);
857}
858
859/** Reply with one string */
860static void
861chaos_replyonestr(sldns_buffer* pkt, const char* str, struct edns_data* edns,
862	struct worker* worker, struct comm_reply* repinfo)
863{
864	chaos_replystr(pkt, (char**)&str, 1, edns, worker, repinfo);
865}
866
867/**
868 * Create CH class trustanchor answer.
869 * @param pkt: buffer
870 * @param edns: edns reply information.
871 * @param w: worker with scratch region.
872 * @param repinfo: reply information for a communication point.
873 */
874static void
875chaos_trustanchor(sldns_buffer* pkt, struct edns_data* edns, struct worker* w,
876	struct comm_reply* repinfo)
877{
878#define TA_RESPONSE_MAX_TXT 16 /* max number of TXT records */
879#define TA_RESPONSE_MAX_TAGS 32 /* max number of tags printed per zone */
880	char* str_array[TA_RESPONSE_MAX_TXT];
881	uint16_t tags[TA_RESPONSE_MAX_TAGS];
882	int num = 0;
883	struct trust_anchor* ta;
884
885	if(!w->env.need_to_validate) {
886		/* no validator module, reply no trustanchors */
887		chaos_replystr(pkt, NULL, 0, edns, w, repinfo);
888		return;
889	}
890
891	/* fill the string with contents */
892	lock_basic_lock(&w->env.anchors->lock);
893	RBTREE_FOR(ta, struct trust_anchor*, w->env.anchors->tree) {
894		char* str;
895		size_t i, numtag, str_len = 255;
896		if(num == TA_RESPONSE_MAX_TXT) continue;
897		str = (char*)regional_alloc(w->scratchpad, str_len);
898		if(!str) continue;
899		lock_basic_lock(&ta->lock);
900		numtag = anchor_list_keytags(ta, tags, TA_RESPONSE_MAX_TAGS);
901		if(numtag == 0) {
902			/* empty, insecure point */
903			lock_basic_unlock(&ta->lock);
904			continue;
905		}
906		str_array[num] = str;
907		num++;
908
909		/* spool name of anchor */
910		(void)sldns_wire2str_dname_buf(ta->name, ta->namelen, str, str_len);
911		str_len -= strlen(str); str += strlen(str);
912		/* spool tags */
913		for(i=0; i<numtag; i++) {
914			snprintf(str, str_len, " %u", (unsigned)tags[i]);
915			str_len -= strlen(str); str += strlen(str);
916		}
917		lock_basic_unlock(&ta->lock);
918	}
919	lock_basic_unlock(&w->env.anchors->lock);
920
921	chaos_replystr(pkt, str_array, num, edns, w, repinfo);
922	regional_free_all(w->scratchpad);
923}
924
925/**
926 * Answer CH class queries.
927 * @param w: worker
928 * @param qinfo: query info. Pointer into packet buffer.
929 * @param edns: edns info from query.
930 * @param repinfo: reply information for a communication point.
931 * @param pkt: packet buffer.
932 * @return: true if a reply is to be sent.
933 */
934static int
935answer_chaos(struct worker* w, struct query_info* qinfo,
936	struct edns_data* edns, struct comm_reply* repinfo, sldns_buffer* pkt)
937{
938	struct config_file* cfg = w->env.cfg;
939	if(qinfo->qtype != LDNS_RR_TYPE_ANY && qinfo->qtype != LDNS_RR_TYPE_TXT)
940		return 0;
941	if(query_dname_compare(qinfo->qname,
942		(uint8_t*)"\002id\006server") == 0 ||
943		query_dname_compare(qinfo->qname,
944		(uint8_t*)"\010hostname\004bind") == 0)
945	{
946		if(cfg->hide_identity)
947			return 0;
948		if(cfg->identity==NULL || cfg->identity[0]==0) {
949			char buf[MAXHOSTNAMELEN+1];
950			if (gethostname(buf, MAXHOSTNAMELEN) == 0) {
951				buf[MAXHOSTNAMELEN] = 0;
952				chaos_replyonestr(pkt, buf, edns, w, repinfo);
953			} else 	{
954				log_err("gethostname: %s", strerror(errno));
955				chaos_replyonestr(pkt, "no hostname", edns, w, repinfo);
956			}
957		}
958		else 	chaos_replyonestr(pkt, cfg->identity, edns, w, repinfo);
959		return 1;
960	}
961	if(query_dname_compare(qinfo->qname,
962		(uint8_t*)"\007version\006server") == 0 ||
963		query_dname_compare(qinfo->qname,
964		(uint8_t*)"\007version\004bind") == 0)
965	{
966		if(cfg->hide_version)
967			return 0;
968		if(cfg->version==NULL || cfg->version[0]==0)
969			chaos_replyonestr(pkt, PACKAGE_STRING, edns, w, repinfo);
970		else 	chaos_replyonestr(pkt, cfg->version, edns, w, repinfo);
971		return 1;
972	}
973	if(query_dname_compare(qinfo->qname,
974		(uint8_t*)"\013trustanchor\007unbound") == 0)
975	{
976		if(cfg->hide_trustanchor)
977			return 0;
978		chaos_trustanchor(pkt, edns, w, repinfo);
979		return 1;
980	}
981
982	return 0;
983}
984
985/**
986 * Answer notify queries.  These are notifies for authoritative zones,
987 * the reply is an ack that the notify has been received.  We need to check
988 * access permission here.
989 * @param w: worker
990 * @param qinfo: query info. Pointer into packet buffer.
991 * @param edns: edns info from query.
992 * @param repinfo: reply info with source address.
993 * @param pkt: packet buffer.
994 */
995static void
996answer_notify(struct worker* w, struct query_info* qinfo,
997	struct edns_data* edns, sldns_buffer* pkt, struct comm_reply* repinfo)
998{
999	int refused = 0;
1000	int rcode = LDNS_RCODE_NOERROR;
1001	uint32_t serial = 0;
1002	int has_serial;
1003	if(!w->env.auth_zones) return;
1004	has_serial = auth_zone_parse_notify_serial(pkt, &serial);
1005	if(auth_zones_notify(w->env.auth_zones, &w->env, qinfo->qname,
1006		qinfo->qname_len, qinfo->qclass, &repinfo->addr,
1007		repinfo->addrlen, has_serial, serial, &refused)) {
1008		rcode = LDNS_RCODE_NOERROR;
1009	} else {
1010		if(refused)
1011			rcode = LDNS_RCODE_REFUSED;
1012		else	rcode = LDNS_RCODE_SERVFAIL;
1013	}
1014
1015	if(verbosity >= VERB_DETAIL) {
1016		char buf[380];
1017		char zname[255+1];
1018		char sr[25];
1019		dname_str(qinfo->qname, zname);
1020		sr[0]=0;
1021		if(has_serial)
1022			snprintf(sr, sizeof(sr), "serial %u ",
1023				(unsigned)serial);
1024		if(rcode == LDNS_RCODE_REFUSED)
1025			snprintf(buf, sizeof(buf),
1026				"refused NOTIFY %sfor %s from", sr, zname);
1027		else if(rcode == LDNS_RCODE_SERVFAIL)
1028			snprintf(buf, sizeof(buf),
1029				"servfail for NOTIFY %sfor %s from", sr, zname);
1030		else	snprintf(buf, sizeof(buf),
1031				"received NOTIFY %sfor %s from", sr, zname);
1032		log_addr(VERB_DETAIL, buf, &repinfo->addr, repinfo->addrlen);
1033	}
1034	edns->edns_version = EDNS_ADVERTISED_VERSION;
1035	edns->udp_size = EDNS_ADVERTISED_SIZE;
1036	edns->ext_rcode = 0;
1037	edns->bits &= EDNS_DO;
1038	edns->opt_list = NULL;
1039	error_encode(pkt, rcode, qinfo,
1040		*(uint16_t*)(void *)sldns_buffer_begin(pkt),
1041		sldns_buffer_read_u16_at(pkt, 2), edns);
1042	LDNS_OPCODE_SET(sldns_buffer_begin(pkt), LDNS_PACKET_NOTIFY);
1043}
1044
1045static int
1046deny_refuse(struct comm_point* c, enum acl_access acl,
1047	enum acl_access deny, enum acl_access refuse,
1048	struct worker* worker, struct comm_reply* repinfo)
1049{
1050	if(acl == deny) {
1051		comm_point_drop_reply(repinfo);
1052		if(worker->stats.extended)
1053			worker->stats.unwanted_queries++;
1054		return 0;
1055	} else if(acl == refuse) {
1056		log_addr(VERB_ALGO, "refused query from",
1057			&repinfo->addr, repinfo->addrlen);
1058		log_buf(VERB_ALGO, "refuse", c->buffer);
1059		if(worker->stats.extended)
1060			worker->stats.unwanted_queries++;
1061		if(worker_check_request(c->buffer, worker) == -1) {
1062			comm_point_drop_reply(repinfo);
1063			return 0; /* discard this */
1064		}
1065		sldns_buffer_set_limit(c->buffer, LDNS_HEADER_SIZE);
1066		sldns_buffer_write_at(c->buffer, 4,
1067			(uint8_t*)"\0\0\0\0\0\0\0\0", 8);
1068		LDNS_QR_SET(sldns_buffer_begin(c->buffer));
1069		LDNS_RCODE_SET(sldns_buffer_begin(c->buffer),
1070			LDNS_RCODE_REFUSED);
1071		sldns_buffer_set_position(c->buffer, LDNS_HEADER_SIZE);
1072		sldns_buffer_flip(c->buffer);
1073		return 1;
1074	}
1075
1076	return -1;
1077}
1078
1079static int
1080deny_refuse_all(struct comm_point* c, enum acl_access acl,
1081	struct worker* worker, struct comm_reply* repinfo)
1082{
1083	return deny_refuse(c, acl, acl_deny, acl_refuse, worker, repinfo);
1084}
1085
1086static int
1087deny_refuse_non_local(struct comm_point* c, enum acl_access acl,
1088	struct worker* worker, struct comm_reply* repinfo)
1089{
1090	return deny_refuse(c, acl, acl_deny_non_local, acl_refuse_non_local, worker, repinfo);
1091}
1092
1093int
1094worker_handle_request(struct comm_point* c, void* arg, int error,
1095	struct comm_reply* repinfo)
1096{
1097	struct worker* worker = (struct worker*)arg;
1098	int ret;
1099	hashvalue_type h;
1100	struct lruhash_entry* e;
1101	struct query_info qinfo;
1102	struct edns_data edns;
1103	enum acl_access acl;
1104	struct acl_addr* acladdr;
1105	int rc = 0;
1106	int need_drop = 0;
1107	int is_expired_answer = 0;
1108	int is_secure_answer = 0;
1109	/* We might have to chase a CNAME chain internally, in which case
1110	 * we'll have up to two replies and combine them to build a complete
1111	 * answer.  These variables control this case. */
1112	struct ub_packed_rrset_key* alias_rrset = NULL;
1113	struct reply_info* partial_rep = NULL;
1114	struct query_info* lookup_qinfo = &qinfo;
1115	struct query_info qinfo_tmp; /* placeholder for lookup_qinfo */
1116	struct respip_client_info* cinfo = NULL, cinfo_tmp;
1117	memset(&qinfo, 0, sizeof(qinfo));
1118
1119	if((error != NETEVENT_NOERROR && error != NETEVENT_DONE)|| !repinfo) {
1120		/* some bad tcp query DNS formats give these error calls */
1121		verbose(VERB_ALGO, "handle request called with err=%d", error);
1122		return 0;
1123	}
1124#ifdef USE_DNSCRYPT
1125	repinfo->max_udp_size = worker->daemon->cfg->max_udp_size;
1126	if(!dnsc_handle_curved_request(worker->daemon->dnscenv, repinfo)) {
1127		worker->stats.num_query_dnscrypt_crypted_malformed++;
1128		return 0;
1129	}
1130	if(c->dnscrypt && !repinfo->is_dnscrypted) {
1131		char buf[LDNS_MAX_DOMAINLEN+1];
1132		/* Check if this is unencrypted and asking for certs */
1133		if(worker_check_request(c->buffer, worker) != 0) {
1134			verbose(VERB_ALGO,
1135				"dnscrypt: worker check request: bad query.");
1136			log_addr(VERB_CLIENT,"from",&repinfo->addr,
1137				repinfo->addrlen);
1138			comm_point_drop_reply(repinfo);
1139			return 0;
1140		}
1141		if(!query_info_parse(&qinfo, c->buffer)) {
1142			verbose(VERB_ALGO,
1143				"dnscrypt: worker parse request: formerror.");
1144			log_addr(VERB_CLIENT, "from", &repinfo->addr,
1145				repinfo->addrlen);
1146			comm_point_drop_reply(repinfo);
1147			return 0;
1148		}
1149		dname_str(qinfo.qname, buf);
1150		if(!(qinfo.qtype == LDNS_RR_TYPE_TXT &&
1151			strcasecmp(buf,
1152			worker->daemon->dnscenv->provider_name) == 0)) {
1153			verbose(VERB_ALGO,
1154				"dnscrypt: not TXT \"%s\". Received: %s \"%s\"",
1155				worker->daemon->dnscenv->provider_name,
1156				sldns_rr_descript(qinfo.qtype)->_name,
1157				buf);
1158			comm_point_drop_reply(repinfo);
1159			worker->stats.num_query_dnscrypt_cleartext++;
1160			return 0;
1161		}
1162		worker->stats.num_query_dnscrypt_cert++;
1163		sldns_buffer_rewind(c->buffer);
1164	} else if(c->dnscrypt && repinfo->is_dnscrypted) {
1165		worker->stats.num_query_dnscrypt_crypted++;
1166	}
1167#endif
1168#ifdef USE_DNSTAP
1169	if(worker->dtenv.log_client_query_messages)
1170		dt_msg_send_client_query(&worker->dtenv, &repinfo->addr, c->type,
1171			c->buffer);
1172#endif
1173	acladdr = acl_addr_lookup(worker->daemon->acl, &repinfo->addr,
1174		repinfo->addrlen);
1175	acl = acl_get_control(acladdr);
1176	if((ret=deny_refuse_all(c, acl, worker, repinfo)) != -1)
1177	{
1178		if(ret == 1)
1179			goto send_reply;
1180		return ret;
1181	}
1182	if((ret=worker_check_request(c->buffer, worker)) != 0) {
1183		verbose(VERB_ALGO, "worker check request: bad query.");
1184		log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen);
1185		if(ret != -1) {
1186			LDNS_QR_SET(sldns_buffer_begin(c->buffer));
1187			LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), ret);
1188			return 1;
1189		}
1190		comm_point_drop_reply(repinfo);
1191		return 0;
1192	}
1193
1194	worker->stats.num_queries++;
1195
1196	/* check if this query should be dropped based on source ip rate limiting */
1197	if(!infra_ip_ratelimit_inc(worker->env.infra_cache, repinfo,
1198			*worker->env.now, c->buffer)) {
1199		/* See if we are passed through with slip factor */
1200		if(worker->env.cfg->ip_ratelimit_factor != 0 &&
1201			ub_random_max(worker->env.rnd,
1202						  worker->env.cfg->ip_ratelimit_factor) == 0) {
1203
1204			char addrbuf[128];
1205			addr_to_str(&repinfo->addr, repinfo->addrlen,
1206						addrbuf, sizeof(addrbuf));
1207		  verbose(VERB_QUERY, "ip_ratelimit allowed through for ip address %s because of slip in ip_ratelimit_factor",
1208				  addrbuf);
1209		} else {
1210			worker->stats.num_queries_ip_ratelimited++;
1211			comm_point_drop_reply(repinfo);
1212			return 0;
1213		}
1214	}
1215
1216	/* see if query is in the cache */
1217	if(!query_info_parse(&qinfo, c->buffer)) {
1218		verbose(VERB_ALGO, "worker parse request: formerror.");
1219		log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen);
1220		memset(&qinfo, 0, sizeof(qinfo)); /* zero qinfo.qname */
1221		if(worker_err_ratelimit(worker, LDNS_RCODE_FORMERR) == -1) {
1222			comm_point_drop_reply(repinfo);
1223			return 0;
1224		}
1225		sldns_buffer_rewind(c->buffer);
1226		LDNS_QR_SET(sldns_buffer_begin(c->buffer));
1227		LDNS_RCODE_SET(sldns_buffer_begin(c->buffer),
1228			LDNS_RCODE_FORMERR);
1229		goto send_reply;
1230	}
1231	if(worker->env.cfg->log_queries) {
1232		char ip[128];
1233		addr_to_str(&repinfo->addr, repinfo->addrlen, ip, sizeof(ip));
1234		log_query_in(ip, qinfo.qname, qinfo.qtype, qinfo.qclass);
1235	}
1236	if(qinfo.qtype == LDNS_RR_TYPE_AXFR ||
1237		qinfo.qtype == LDNS_RR_TYPE_IXFR) {
1238		verbose(VERB_ALGO, "worker request: refused zone transfer.");
1239		log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen);
1240		sldns_buffer_rewind(c->buffer);
1241		LDNS_QR_SET(sldns_buffer_begin(c->buffer));
1242		LDNS_RCODE_SET(sldns_buffer_begin(c->buffer),
1243			LDNS_RCODE_REFUSED);
1244		if(worker->stats.extended) {
1245			worker->stats.qtype[qinfo.qtype]++;
1246		}
1247		goto send_reply;
1248	}
1249	if(qinfo.qtype == LDNS_RR_TYPE_OPT ||
1250		qinfo.qtype == LDNS_RR_TYPE_TSIG ||
1251		qinfo.qtype == LDNS_RR_TYPE_TKEY ||
1252		qinfo.qtype == LDNS_RR_TYPE_MAILA ||
1253		qinfo.qtype == LDNS_RR_TYPE_MAILB ||
1254		(qinfo.qtype >= 128 && qinfo.qtype <= 248)) {
1255		verbose(VERB_ALGO, "worker request: formerror for meta-type.");
1256		log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen);
1257		if(worker_err_ratelimit(worker, LDNS_RCODE_FORMERR) == -1) {
1258			comm_point_drop_reply(repinfo);
1259			return 0;
1260		}
1261		sldns_buffer_rewind(c->buffer);
1262		LDNS_QR_SET(sldns_buffer_begin(c->buffer));
1263		LDNS_RCODE_SET(sldns_buffer_begin(c->buffer),
1264			LDNS_RCODE_FORMERR);
1265		if(worker->stats.extended) {
1266			worker->stats.qtype[qinfo.qtype]++;
1267		}
1268		goto send_reply;
1269	}
1270	if((ret=parse_edns_from_pkt(c->buffer, &edns, worker->scratchpad)) != 0) {
1271		struct edns_data reply_edns;
1272		verbose(VERB_ALGO, "worker parse edns: formerror.");
1273		log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen);
1274		memset(&reply_edns, 0, sizeof(reply_edns));
1275		reply_edns.edns_present = 1;
1276		reply_edns.udp_size = EDNS_ADVERTISED_SIZE;
1277		LDNS_RCODE_SET(sldns_buffer_begin(c->buffer), ret);
1278		error_encode(c->buffer, ret, &qinfo,
1279			*(uint16_t*)(void *)sldns_buffer_begin(c->buffer),
1280			sldns_buffer_read_u16_at(c->buffer, 2), &reply_edns);
1281		regional_free_all(worker->scratchpad);
1282		goto send_reply;
1283	}
1284	if(edns.edns_present) {
1285		struct edns_option* edns_opt;
1286		if(edns.edns_version != 0) {
1287			edns.ext_rcode = (uint8_t)(EDNS_RCODE_BADVERS>>4);
1288			edns.edns_version = EDNS_ADVERTISED_VERSION;
1289			edns.udp_size = EDNS_ADVERTISED_SIZE;
1290			edns.bits &= EDNS_DO;
1291			edns.opt_list = NULL;
1292			edns.padding_block_size = 0;
1293			verbose(VERB_ALGO, "query with bad edns version.");
1294			log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen);
1295			error_encode(c->buffer, EDNS_RCODE_BADVERS&0xf, &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		if(edns.udp_size < NORMAL_UDP_SIZE &&
1305		   worker->daemon->cfg->harden_short_bufsize) {
1306			verbose(VERB_QUERY, "worker request: EDNS bufsize %d ignored",
1307				(int)edns.udp_size);
1308			log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen);
1309			edns.udp_size = NORMAL_UDP_SIZE;
1310		}
1311		if(c->type != comm_udp) {
1312			edns_opt = edns_opt_list_find(edns.opt_list, LDNS_EDNS_KEEPALIVE);
1313			if(edns_opt && edns_opt->opt_len > 0) {
1314				edns.ext_rcode = 0;
1315				edns.edns_version = EDNS_ADVERTISED_VERSION;
1316				edns.udp_size = EDNS_ADVERTISED_SIZE;
1317				edns.bits &= EDNS_DO;
1318				edns.opt_list = NULL;
1319				verbose(VERB_ALGO, "query with bad edns keepalive.");
1320				log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen);
1321				error_encode(c->buffer, LDNS_RCODE_FORMERR, &qinfo,
1322					*(uint16_t*)(void *)sldns_buffer_begin(c->buffer),
1323					sldns_buffer_read_u16_at(c->buffer, 2), NULL);
1324				if(sldns_buffer_capacity(c->buffer) >=
1325				   sldns_buffer_limit(c->buffer)+calc_edns_field_size(&edns))
1326					attach_edns_record(c->buffer, &edns);
1327				regional_free_all(worker->scratchpad);
1328				goto send_reply;
1329			}
1330		}
1331	}
1332	if(edns.udp_size > worker->daemon->cfg->max_udp_size &&
1333		c->type == comm_udp) {
1334		verbose(VERB_QUERY,
1335			"worker request: max UDP reply size modified"
1336			" (%d to max-udp-size)", (int)edns.udp_size);
1337		log_addr(VERB_CLIENT,"from",&repinfo->addr, repinfo->addrlen);
1338		edns.udp_size = worker->daemon->cfg->max_udp_size;
1339	}
1340	if(edns.udp_size < LDNS_HEADER_SIZE) {
1341		verbose(VERB_ALGO, "worker request: edns is too small.");
1342		log_addr(VERB_CLIENT, "from", &repinfo->addr, repinfo->addrlen);
1343		LDNS_QR_SET(sldns_buffer_begin(c->buffer));
1344		LDNS_TC_SET(sldns_buffer_begin(c->buffer));
1345		LDNS_RCODE_SET(sldns_buffer_begin(c->buffer),
1346			LDNS_RCODE_SERVFAIL);
1347		sldns_buffer_set_position(c->buffer, LDNS_HEADER_SIZE);
1348		sldns_buffer_write_at(c->buffer, 4,
1349			(uint8_t*)"\0\0\0\0\0\0\0\0", 8);
1350		sldns_buffer_flip(c->buffer);
1351		regional_free_all(worker->scratchpad);
1352		goto send_reply;
1353	}
1354	if(worker->stats.extended)
1355		server_stats_insquery(&worker->stats, c, qinfo.qtype,
1356			qinfo.qclass, &edns, repinfo);
1357	if(c->type != comm_udp)
1358		edns.udp_size = 65535; /* max size for TCP replies */
1359	if(qinfo.qclass == LDNS_RR_CLASS_CH && answer_chaos(worker, &qinfo,
1360		&edns, repinfo, c->buffer)) {
1361		regional_free_all(worker->scratchpad);
1362		goto send_reply;
1363	}
1364	if(LDNS_OPCODE_WIRE(sldns_buffer_begin(c->buffer)) ==
1365		LDNS_PACKET_NOTIFY) {
1366		answer_notify(worker, &qinfo, &edns, c->buffer, repinfo);
1367		regional_free_all(worker->scratchpad);
1368		goto send_reply;
1369	}
1370	if(local_zones_answer(worker->daemon->local_zones, &worker->env, &qinfo,
1371		&edns, c->buffer, worker->scratchpad, repinfo, acladdr->taglist,
1372		acladdr->taglen, acladdr->tag_actions,
1373		acladdr->tag_actions_size, acladdr->tag_datas,
1374		acladdr->tag_datas_size, worker->daemon->cfg->tagname,
1375		worker->daemon->cfg->num_tags, acladdr->view)) {
1376		regional_free_all(worker->scratchpad);
1377		if(sldns_buffer_limit(c->buffer) == 0) {
1378			comm_point_drop_reply(repinfo);
1379			return 0;
1380		}
1381		goto send_reply;
1382	}
1383	if(worker->env.auth_zones &&
1384		rpz_apply_qname_trigger(worker->env.auth_zones,
1385		&worker->env, &qinfo, &edns, c->buffer, worker->scratchpad,
1386		repinfo, acladdr->taglist, acladdr->taglen, &worker->stats)) {
1387		regional_free_all(worker->scratchpad);
1388		if(sldns_buffer_limit(c->buffer) == 0) {
1389			comm_point_drop_reply(repinfo);
1390			return 0;
1391		}
1392		goto send_reply;
1393	}
1394	if(worker->env.auth_zones &&
1395		auth_zones_answer(worker->env.auth_zones, &worker->env,
1396		&qinfo, &edns, repinfo, c->buffer, worker->scratchpad)) {
1397		regional_free_all(worker->scratchpad);
1398		if(sldns_buffer_limit(c->buffer) == 0) {
1399			comm_point_drop_reply(repinfo);
1400			return 0;
1401		}
1402		/* set RA for everyone that can have recursion (based on
1403		 * access control list) */
1404		if(LDNS_RD_WIRE(sldns_buffer_begin(c->buffer)) &&
1405		   acl != acl_deny_non_local && acl != acl_refuse_non_local)
1406			LDNS_RA_SET(sldns_buffer_begin(c->buffer));
1407		goto send_reply;
1408	}
1409
1410	/* We've looked in our local zones. If the answer isn't there, we
1411	 * might need to bail out based on ACLs now. */
1412	if((ret=deny_refuse_non_local(c, acl, worker, repinfo)) != -1)
1413	{
1414		regional_free_all(worker->scratchpad);
1415		if(ret == 1)
1416			goto send_reply;
1417		return ret;
1418	}
1419
1420	/* If this request does not have the recursion bit set, verify
1421	 * ACLs allow the recursion bit to be treated as set. */
1422	if(!(LDNS_RD_WIRE(sldns_buffer_begin(c->buffer))) &&
1423		acl == acl_allow_setrd ) {
1424		LDNS_RD_SET(sldns_buffer_begin(c->buffer));
1425	}
1426
1427	/* If this request does not have the recursion bit set, verify
1428	 * ACLs allow the snooping. */
1429	if(!(LDNS_RD_WIRE(sldns_buffer_begin(c->buffer))) &&
1430		acl != acl_allow_snoop ) {
1431		error_encode(c->buffer, LDNS_RCODE_REFUSED, &qinfo,
1432			*(uint16_t*)(void *)sldns_buffer_begin(c->buffer),
1433			sldns_buffer_read_u16_at(c->buffer, 2), NULL);
1434		regional_free_all(worker->scratchpad);
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
1477lookup_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
1580send_reply:
1581	rc = 1;
1582send_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	server_stats_insrcode(&worker->stats, c->buffer);
1591	if(worker->stats.extended) {
1592		if(is_secure_answer) worker->stats.ans_secure++;
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
1623void
1624worker_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 */
1656static void
1657worker_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
1669void 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
1686void 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
1698struct worker*
1699worker_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	return worker;
1729}
1730
1731int
1732worker_init(struct worker* worker, struct config_file *cfg,
1733	struct listen_port* ports, int do_sigs)
1734{
1735#ifdef USE_DNSTAP
1736	struct dt_env* dtenv = &worker->dtenv;
1737#else
1738	void* dtenv = NULL;
1739#endif
1740	worker->need_to_exit = 0;
1741	worker->base = comm_base_create(do_sigs);
1742	if(!worker->base) {
1743		log_err("could not create event handling base");
1744		worker_delete(worker);
1745		return 0;
1746	}
1747	comm_base_set_slow_accept_handlers(worker->base, &worker_stop_accept,
1748		&worker_start_accept, worker);
1749	if(do_sigs) {
1750#ifdef SIGHUP
1751		ub_thread_sig_unblock(SIGHUP);
1752#endif
1753		ub_thread_sig_unblock(SIGINT);
1754#ifdef SIGQUIT
1755		ub_thread_sig_unblock(SIGQUIT);
1756#endif
1757		ub_thread_sig_unblock(SIGTERM);
1758#ifndef LIBEVENT_SIGNAL_PROBLEM
1759		worker->comsig = comm_signal_create(worker->base,
1760			worker_sighandler, worker);
1761		if(!worker->comsig
1762#ifdef SIGHUP
1763			|| !comm_signal_bind(worker->comsig, SIGHUP)
1764#endif
1765#ifdef SIGQUIT
1766			|| !comm_signal_bind(worker->comsig, SIGQUIT)
1767#endif
1768			|| !comm_signal_bind(worker->comsig, SIGTERM)
1769			|| !comm_signal_bind(worker->comsig, SIGINT)) {
1770			log_err("could not create signal handlers");
1771			worker_delete(worker);
1772			return 0;
1773		}
1774#endif /* LIBEVENT_SIGNAL_PROBLEM */
1775		if(!daemon_remote_open_accept(worker->daemon->rc,
1776			worker->daemon->rc_ports, worker)) {
1777			worker_delete(worker);
1778			return 0;
1779		}
1780#ifdef UB_ON_WINDOWS
1781		wsvc_setup_worker(worker);
1782#endif /* UB_ON_WINDOWS */
1783	} else { /* !do_sigs */
1784		worker->comsig = NULL;
1785	}
1786#ifdef USE_DNSTAP
1787	if(cfg->dnstap) {
1788		log_assert(worker->daemon->dtenv != NULL);
1789		memcpy(&worker->dtenv, worker->daemon->dtenv, sizeof(struct dt_env));
1790		if(!dt_init(&worker->dtenv, worker->base))
1791			fatal_exit("dt_init failed");
1792	}
1793#endif
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		cfg->harden_large_queries, cfg->http_max_streams,
1800		cfg->http_endpoint, cfg->http_notls_downstream,
1801		worker->daemon->tcl, worker->daemon->listen_sslctx,
1802		dtenv, worker_handle_request, worker);
1803	if(!worker->front) {
1804		log_err("could not create listening sockets");
1805		worker_delete(worker);
1806		return 0;
1807	}
1808	worker->back = outside_network_create(worker->base,
1809		cfg->msg_buffer_size, (size_t)cfg->outgoing_num_ports,
1810		cfg->out_ifs, cfg->num_out_ifs, cfg->do_ip4, cfg->do_ip6,
1811		cfg->do_tcp?cfg->outgoing_num_tcp:0, cfg->ip_dscp,
1812		worker->daemon->env->infra_cache, worker->rndstate,
1813		cfg->use_caps_bits_for_id, worker->ports, worker->numports,
1814		cfg->unwanted_threshold, cfg->outgoing_tcp_mss,
1815		&worker_alloc_cleanup, worker,
1816		cfg->do_udp || cfg->udp_upstream_without_downstream,
1817		worker->daemon->connect_sslctx, cfg->delay_close,
1818		cfg->tls_use_sni, dtenv, cfg->udp_connect);
1819	if(!worker->back) {
1820		log_err("could not create outgoing sockets");
1821		worker_delete(worker);
1822		return 0;
1823	}
1824	/* start listening to commands */
1825	if(!tube_setup_bg_listen(worker->cmd, worker->base,
1826		&worker_handle_control_cmd, worker)) {
1827		log_err("could not create control compt.");
1828		worker_delete(worker);
1829		return 0;
1830	}
1831	worker->stat_timer = comm_timer_create(worker->base,
1832		worker_stat_timer_cb, worker);
1833	if(!worker->stat_timer) {
1834		log_err("could not create statistics timer");
1835	}
1836
1837	/* we use the msg_buffer_size as a good estimate for what the
1838	 * user wants for memory usage sizes */
1839	worker->scratchpad = regional_create_custom(cfg->msg_buffer_size);
1840	if(!worker->scratchpad) {
1841		log_err("malloc failure");
1842		worker_delete(worker);
1843		return 0;
1844	}
1845
1846	server_stats_init(&worker->stats, cfg);
1847	alloc_init(&worker->alloc, &worker->daemon->superalloc,
1848		worker->thread_num);
1849	alloc_set_id_cleanup(&worker->alloc, &worker_alloc_cleanup, worker);
1850	worker->env = *worker->daemon->env;
1851	comm_base_timept(worker->base, &worker->env.now, &worker->env.now_tv);
1852	worker->env.worker = worker;
1853	worker->env.worker_base = worker->base;
1854	worker->env.send_query = &worker_send_query;
1855	worker->env.alloc = &worker->alloc;
1856	worker->env.outnet = worker->back;
1857	worker->env.rnd = worker->rndstate;
1858	/* If case prefetch is triggered, the corresponding mesh will clear
1859	 * the scratchpad for the module env in the middle of request handling.
1860	 * It would be prone to a use-after-free kind of bug, so we avoid
1861	 * sharing it with worker's own scratchpad at the cost of having
1862	 * one more pad per worker. */
1863	worker->env.scratch = regional_create_custom(cfg->msg_buffer_size);
1864	if(!worker->env.scratch) {
1865		log_err("malloc failure");
1866		worker_delete(worker);
1867		return 0;
1868	}
1869	worker->env.mesh = mesh_create(&worker->daemon->mods, &worker->env);
1870	/* Pass on daemon variables that we would need in the mesh area */
1871	worker->env.mesh->use_response_ip = worker->daemon->use_response_ip;
1872	worker->env.mesh->use_rpz = worker->daemon->use_rpz;
1873
1874	worker->env.detach_subs = &mesh_detach_subs;
1875	worker->env.attach_sub = &mesh_attach_sub;
1876	worker->env.add_sub = &mesh_add_sub;
1877	worker->env.kill_sub = &mesh_state_delete;
1878	worker->env.detect_cycle = &mesh_detect_cycle;
1879	worker->env.scratch_buffer = sldns_buffer_new(cfg->msg_buffer_size);
1880	if(!(worker->env.fwds = forwards_create()) ||
1881		!forwards_apply_cfg(worker->env.fwds, cfg)) {
1882		log_err("Could not set forward zones");
1883		worker_delete(worker);
1884		return 0;
1885	}
1886	if(!(worker->env.hints = hints_create()) ||
1887		!hints_apply_cfg(worker->env.hints, cfg)) {
1888		log_err("Could not set root or stub hints");
1889		worker_delete(worker);
1890		return 0;
1891	}
1892	/* one probe timer per process -- if we have 5011 anchors */
1893	if(autr_get_num_anchors(worker->env.anchors) > 0
1894#ifndef THREADS_DISABLED
1895		&& worker->thread_num == 0
1896#endif
1897		) {
1898		struct timeval tv;
1899		tv.tv_sec = 0;
1900		tv.tv_usec = 0;
1901		worker->env.probe_timer = comm_timer_create(worker->base,
1902			worker_probe_timer_cb, worker);
1903		if(!worker->env.probe_timer) {
1904			log_err("could not create 5011-probe timer");
1905		} else {
1906			/* let timer fire, then it can reset itself */
1907			comm_timer_set(worker->env.probe_timer, &tv);
1908		}
1909	}
1910	/* zone transfer tasks, setup once per process, if any */
1911	if(worker->env.auth_zones
1912#ifndef THREADS_DISABLED
1913		&& worker->thread_num == 0
1914#endif
1915		) {
1916		auth_xfer_pickup_initial(worker->env.auth_zones, &worker->env);
1917	}
1918#ifdef USE_DNSTAP
1919	if(worker->daemon->cfg->dnstap
1920#ifndef THREADS_DISABLED
1921		&& worker->thread_num == 0
1922#endif
1923		) {
1924		if(!dt_io_thread_start(dtenv->dtio, comm_base_internal(
1925			worker->base), worker->daemon->num)) {
1926			log_err("could not start dnstap io thread");
1927			worker_delete(worker);
1928			return 0;
1929		}
1930	}
1931#endif /* USE_DNSTAP */
1932	if(!worker->env.mesh || !worker->env.scratch_buffer) {
1933		worker_delete(worker);
1934		return 0;
1935	}
1936	worker_mem_report(worker, NULL);
1937	/* if statistics enabled start timer */
1938	if(worker->env.cfg->stat_interval > 0) {
1939		verbose(VERB_ALGO, "set statistics interval %d secs",
1940			worker->env.cfg->stat_interval);
1941		worker_restart_timer(worker);
1942	}
1943	return 1;
1944}
1945
1946void
1947worker_work(struct worker* worker)
1948{
1949	comm_base_dispatch(worker->base);
1950}
1951
1952void
1953worker_delete(struct worker* worker)
1954{
1955	if(!worker)
1956		return;
1957	if(worker->env.mesh && verbosity >= VERB_OPS) {
1958		server_stats_log(&worker->stats, worker, worker->thread_num);
1959		mesh_stats(worker->env.mesh, "mesh has");
1960		worker_mem_report(worker, NULL);
1961	}
1962	outside_network_quit_prepare(worker->back);
1963	mesh_delete(worker->env.mesh);
1964	sldns_buffer_free(worker->env.scratch_buffer);
1965	forwards_delete(worker->env.fwds);
1966	hints_delete(worker->env.hints);
1967	listen_delete(worker->front);
1968	outside_network_delete(worker->back);
1969	comm_signal_delete(worker->comsig);
1970	tube_delete(worker->cmd);
1971	comm_timer_delete(worker->stat_timer);
1972	comm_timer_delete(worker->env.probe_timer);
1973	free(worker->ports);
1974	if(worker->thread_num == 0) {
1975#ifdef UB_ON_WINDOWS
1976		wsvc_desetup_worker(worker);
1977#endif /* UB_ON_WINDOWS */
1978	}
1979#ifdef USE_DNSTAP
1980	if(worker->daemon->cfg->dnstap
1981#ifndef THREADS_DISABLED
1982		&& worker->thread_num == 0
1983#endif
1984		) {
1985		dt_io_thread_stop(worker->dtenv.dtio);
1986	}
1987	dt_deinit(&worker->dtenv);
1988#endif /* USE_DNSTAP */
1989	comm_base_delete(worker->base);
1990	ub_randfree(worker->rndstate);
1991	alloc_clear(&worker->alloc);
1992	regional_destroy(worker->env.scratch);
1993	regional_destroy(worker->scratchpad);
1994	free(worker);
1995}
1996
1997struct outbound_entry*
1998worker_send_query(struct query_info* qinfo, uint16_t flags, int dnssec,
1999	int want_dnssec, int nocaps, struct sockaddr_storage* addr,
2000	socklen_t addrlen, uint8_t* zone, size_t zonelen, int ssl_upstream,
2001	char* tls_auth_name, struct module_qstate* q)
2002{
2003	struct worker* worker = q->env->worker;
2004	struct outbound_entry* e = (struct outbound_entry*)regional_alloc(
2005		q->region, sizeof(*e));
2006	if(!e)
2007		return NULL;
2008	e->qstate = q;
2009	e->qsent = outnet_serviced_query(worker->back, qinfo, flags, dnssec,
2010		want_dnssec, nocaps, q->env->cfg->tcp_upstream,
2011		ssl_upstream, tls_auth_name, addr, addrlen, zone, zonelen, q,
2012		worker_handle_service_reply, e, worker->back->udp_buff, q->env);
2013	if(!e->qsent) {
2014		return NULL;
2015	}
2016	return e;
2017}
2018
2019void
2020worker_alloc_cleanup(void* arg)
2021{
2022	struct worker* worker = (struct worker*)arg;
2023	slabhash_clear(&worker->env.rrset_cache->table);
2024	slabhash_clear(worker->env.msg_cache);
2025}
2026
2027void worker_stats_clear(struct worker* worker)
2028{
2029	server_stats_init(&worker->stats, worker->env.cfg);
2030	mesh_stats_clear(worker->env.mesh);
2031	worker->back->unwanted_replies = 0;
2032	worker->back->num_tcp_outgoing = 0;
2033}
2034
2035void worker_start_accept(void* arg)
2036{
2037	struct worker* worker = (struct worker*)arg;
2038	listen_start_accept(worker->front);
2039	if(worker->thread_num == 0)
2040		daemon_remote_start_accept(worker->daemon->rc);
2041}
2042
2043void worker_stop_accept(void* arg)
2044{
2045	struct worker* worker = (struct worker*)arg;
2046	listen_stop_accept(worker->front);
2047	if(worker->thread_num == 0)
2048		daemon_remote_stop_accept(worker->daemon->rc);
2049}
2050
2051/* --- fake callbacks for fptr_wlist to work --- */
2052struct outbound_entry* libworker_send_query(
2053	struct query_info* ATTR_UNUSED(qinfo),
2054	uint16_t ATTR_UNUSED(flags), int ATTR_UNUSED(dnssec),
2055	int ATTR_UNUSED(want_dnssec), int ATTR_UNUSED(nocaps),
2056	struct sockaddr_storage* ATTR_UNUSED(addr), socklen_t ATTR_UNUSED(addrlen),
2057	uint8_t* ATTR_UNUSED(zone), size_t ATTR_UNUSED(zonelen),
2058	int ATTR_UNUSED(ssl_upstream), char* ATTR_UNUSED(tls_auth_name),
2059	struct module_qstate* ATTR_UNUSED(q))
2060{
2061	log_assert(0);
2062	return 0;
2063}
2064
2065int libworker_handle_reply(struct comm_point* ATTR_UNUSED(c),
2066	void* ATTR_UNUSED(arg), int ATTR_UNUSED(error),
2067        struct comm_reply* ATTR_UNUSED(reply_info))
2068{
2069	log_assert(0);
2070	return 0;
2071}
2072
2073int libworker_handle_service_reply(struct comm_point* ATTR_UNUSED(c),
2074	void* ATTR_UNUSED(arg), int ATTR_UNUSED(error),
2075        struct comm_reply* ATTR_UNUSED(reply_info))
2076{
2077	log_assert(0);
2078	return 0;
2079}
2080
2081void libworker_handle_control_cmd(struct tube* ATTR_UNUSED(tube),
2082        uint8_t* ATTR_UNUSED(buffer), size_t ATTR_UNUSED(len),
2083        int ATTR_UNUSED(error), void* ATTR_UNUSED(arg))
2084{
2085	log_assert(0);
2086}
2087
2088void libworker_fg_done_cb(void* ATTR_UNUSED(arg), int ATTR_UNUSED(rcode),
2089	sldns_buffer* ATTR_UNUSED(buf), enum sec_status ATTR_UNUSED(s),
2090	char* ATTR_UNUSED(why_bogus), int ATTR_UNUSED(was_ratelimited))
2091{
2092	log_assert(0);
2093}
2094
2095void libworker_bg_done_cb(void* ATTR_UNUSED(arg), int ATTR_UNUSED(rcode),
2096	sldns_buffer* ATTR_UNUSED(buf), enum sec_status ATTR_UNUSED(s),
2097	char* ATTR_UNUSED(why_bogus), int ATTR_UNUSED(was_ratelimited))
2098{
2099	log_assert(0);
2100}
2101
2102void libworker_event_done_cb(void* ATTR_UNUSED(arg), int ATTR_UNUSED(rcode),
2103	sldns_buffer* ATTR_UNUSED(buf), enum sec_status ATTR_UNUSED(s),
2104	char* ATTR_UNUSED(why_bogus), int ATTR_UNUSED(was_ratelimited))
2105{
2106	log_assert(0);
2107}
2108
2109int context_query_cmp(const void* ATTR_UNUSED(a), const void* ATTR_UNUSED(b))
2110{
2111	log_assert(0);
2112	return 0;
2113}
2114
2115int order_lock_cmp(const void* ATTR_UNUSED(e1), const void* ATTR_UNUSED(e2))
2116{
2117	log_assert(0);
2118	return 0;
2119}
2120
2121int codeline_cmp(const void* ATTR_UNUSED(a), const void* ATTR_UNUSED(b))
2122{
2123	log_assert(0);
2124	return 0;
2125}
2126
2127#ifdef USE_DNSTAP
2128void dtio_tap_callback(int ATTR_UNUSED(fd), short ATTR_UNUSED(ev),
2129	void* ATTR_UNUSED(arg))
2130{
2131	log_assert(0);
2132}
2133#endif
2134
2135#ifdef USE_DNSTAP
2136void dtio_mainfdcallback(int ATTR_UNUSED(fd), short ATTR_UNUSED(ev),
2137	void* ATTR_UNUSED(arg))
2138{
2139	log_assert(0);
2140}
2141#endif
2142