1 /* $NetBSD: res_send.c,v 1.9 2006/01/24 17:41:25 christos Exp $ */
2
3 /*
4 * Copyright (c) 1985, 1989, 1993
5 * The Regents of the University of California. All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. All advertising materials mentioning features or use of this software
16 * must display the following acknowledgement:
17 * This product includes software developed by the University of
18 * California, Berkeley and its contributors.
19 * 4. Neither the name of the University nor the names of its contributors
20 * may be used to endorse or promote products derived from this software
21 * without specific prior written permission.
22 *
23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * SUCH DAMAGE.
34 */
35
36 /*
37 * Portions Copyright (c) 1993 by Digital Equipment Corporation.
38 *
39 * Permission to use, copy, modify, and distribute this software for any
40 * purpose with or without fee is hereby granted, provided that the above
41 * copyright notice and this permission notice appear in all copies, and that
42 * the name of Digital Equipment Corporation not be used in advertising or
43 * publicity pertaining to distribution of the document or software without
44 * specific, written prior permission.
45 *
46 * THE SOFTWARE IS PROVIDED "AS IS" AND DIGITAL EQUIPMENT CORP. DISCLAIMS ALL
47 * WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES
48 * OF MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL DIGITAL EQUIPMENT
49 * CORPORATION BE LIABLE FOR ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL
50 * DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR
51 * PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
52 * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS
53 * SOFTWARE.
54 */
55
56 /*
57 * Copyright (c) 2004 by Internet Systems Consortium, Inc. ("ISC")
58 * Portions Copyright (c) 1996-1999 by Internet Software Consortium.
59 *
60 * Permission to use, copy, modify, and distribute this software for any
61 * purpose with or without fee is hereby granted, provided that the above
62 * copyright notice and this permission notice appear in all copies.
63 *
64 * THE SOFTWARE IS PROVIDED "AS IS" AND ISC DISCLAIMS ALL WARRANTIES
65 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
66 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL ISC BE LIABLE FOR
67 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
68 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
69 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT
70 * OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
71 */
72
73 /*
74 * Send query to name server and wait for reply.
75 */
76
77 #define LOG_TAG "res_send"
78
79 #include <sys/param.h>
80 #include <sys/socket.h>
81 #include <sys/time.h>
82 #include <sys/types.h>
83 #include <sys/uio.h>
84
85 #include <arpa/inet.h>
86 #include <arpa/nameser.h>
87 #include <netinet/in.h>
88
89 #include <errno.h>
90 #include <fcntl.h>
91 #include <netdb.h>
92 #include <poll.h>
93 #include <signal.h>
94 #include <stdio.h>
95 #include <stdlib.h>
96 #include <string.h>
97 #include <time.h>
98 #include <unistd.h>
99
100 #include <android-base/logging.h>
101 #include <android/multinetwork.h> // ResNsendFlags
102
103 #include <netdutils/Slice.h>
104 #include "DnsTlsDispatcher.h"
105 #include "DnsTlsTransport.h"
106 #include "PrivateDnsConfiguration.h"
107 #include "netd_resolv/resolv.h"
108 #include "netd_resolv/stats.h"
109 #include "private/android_filesystem_config.h"
110 #include "res_state_ext.h"
111 #include "resolv_cache.h"
112 #include "resolv_private.h"
113
114 // TODO: use the namespace something like android::netd_resolv for libnetd_resolv
115 using namespace android::net;
116 using android::netdutils::Slice;
117
118 static DnsTlsDispatcher sDnsTlsDispatcher;
119
120 static int get_salen(const struct sockaddr*);
121 static struct sockaddr* get_nsaddr(res_state, size_t);
122 static int send_vc(res_state, res_params* params, const u_char*, int, u_char*, int, int*, int,
123 time_t*, int*, int*);
124 static int send_dg(res_state, res_params* params, const u_char*, int, u_char*, int, int*, int, int*,
125 int*, time_t*, int*, int*);
126 static void Aerror(const res_state, const char*, int, const struct sockaddr*, int);
127 static void Perror(const res_state, const char*, int);
128
129 static int sock_eq(struct sockaddr*, struct sockaddr*);
130 static int connect_with_timeout(int sock, const struct sockaddr* nsap, socklen_t salen,
131 const struct timespec timeout);
132 static int retrying_poll(const int sock, short events, const struct timespec* finish);
133 static int res_tls_send(res_state, const Slice query, const Slice answer, int* rcode,
134 bool* fallback);
135
136 /* BIONIC-BEGIN: implement source port randomization */
137
138 // BEGIN: Code copied from ISC eventlib
139 // TODO: move away from this code
140
141 #define BILLION 1000000000
142
evConsTime(time_t sec,long nsec)143 static struct timespec evConsTime(time_t sec, long nsec) {
144 struct timespec x;
145
146 x.tv_sec = sec;
147 x.tv_nsec = nsec;
148 return (x);
149 }
150
evAddTime(struct timespec addend1,struct timespec addend2)151 static struct timespec evAddTime(struct timespec addend1, struct timespec addend2) {
152 struct timespec x;
153
154 x.tv_sec = addend1.tv_sec + addend2.tv_sec;
155 x.tv_nsec = addend1.tv_nsec + addend2.tv_nsec;
156 if (x.tv_nsec >= BILLION) {
157 x.tv_sec++;
158 x.tv_nsec -= BILLION;
159 }
160 return (x);
161 }
162
evSubTime(struct timespec minuend,struct timespec subtrahend)163 static struct timespec evSubTime(struct timespec minuend, struct timespec subtrahend) {
164 struct timespec x;
165
166 x.tv_sec = minuend.tv_sec - subtrahend.tv_sec;
167 if (minuend.tv_nsec >= subtrahend.tv_nsec)
168 x.tv_nsec = minuend.tv_nsec - subtrahend.tv_nsec;
169 else {
170 x.tv_nsec = BILLION - subtrahend.tv_nsec + minuend.tv_nsec;
171 x.tv_sec--;
172 }
173 return (x);
174 }
175
evCmpTime(struct timespec a,struct timespec b)176 static int evCmpTime(struct timespec a, struct timespec b) {
177 #define SGN(x) ((x) < 0 ? (-1) : (x) > 0 ? (1) : (0));
178 time_t s = a.tv_sec - b.tv_sec;
179 long n;
180
181 if (s != 0) return SGN(s);
182
183 n = a.tv_nsec - b.tv_nsec;
184 return SGN(n);
185 }
186
evNowTime(void)187 static struct timespec evNowTime(void) {
188 struct timespec tsnow;
189 clock_gettime(CLOCK_REALTIME, &tsnow);
190 return tsnow;
191 }
192
evConsIovec(void * buf,size_t cnt)193 static struct iovec evConsIovec(void* buf, size_t cnt) {
194 struct iovec ret;
195
196 memset(&ret, 0xf5, sizeof ret);
197 ret.iov_base = buf;
198 ret.iov_len = cnt;
199 return ret;
200 }
201
202 // END: Code copied from ISC eventlib
203
random_bind(int s,int family)204 static int random_bind(int s, int family) {
205 sockaddr_union u;
206 int j;
207 socklen_t slen;
208
209 /* clear all, this also sets the IP4/6 address to 'any' */
210 memset(&u, 0, sizeof u);
211
212 switch (family) {
213 case AF_INET:
214 u.sin.sin_family = family;
215 slen = sizeof u.sin;
216 break;
217 case AF_INET6:
218 u.sin6.sin6_family = family;
219 slen = sizeof u.sin6;
220 break;
221 default:
222 errno = EPROTO;
223 return -1;
224 }
225
226 /* first try to bind to a random source port a few times */
227 for (j = 0; j < 10; j++) {
228 /* find a random port between 1025 .. 65534 */
229 int port = 1025 + (arc4random_uniform(65535 - 1025));
230 if (family == AF_INET)
231 u.sin.sin_port = htons(port);
232 else
233 u.sin6.sin6_port = htons(port);
234
235 if (!bind(s, &u.sa, slen)) return 0;
236 }
237
238 // nothing after 10 attempts, our network table is probably busy
239 // let the system decide which port is best
240 if (family == AF_INET)
241 u.sin.sin_port = 0;
242 else
243 u.sin6.sin6_port = 0;
244
245 return bind(s, &u.sa, slen);
246 }
247 /* BIONIC-END */
248
249 // Disables all nameservers other than selectedServer
res_set_usable_server(int selectedServer,int nscount,bool usable_servers[])250 static void res_set_usable_server(int selectedServer, int nscount, bool usable_servers[]) {
251 int usableIndex = 0;
252 for (int ns = 0; ns < nscount; ns++) {
253 if (usable_servers[ns]) ++usableIndex;
254 if (usableIndex != selectedServer) usable_servers[ns] = false;
255 }
256 }
257
258 /* int
259 * res_isourserver(ina)
260 * looks up "ina" in _res.ns_addr_list[]
261 * returns:
262 * 0 : not found
263 * >0 : found
264 * author:
265 * paul vixie, 29may94
266 */
res_ourserver_p(const res_state statp,const sockaddr * sa)267 static int res_ourserver_p(const res_state statp, const sockaddr* sa) {
268 const sockaddr_in *inp, *srv;
269 const sockaddr_in6 *in6p, *srv6;
270 int ns;
271
272 switch (sa->sa_family) {
273 case AF_INET:
274 inp = (const struct sockaddr_in*) (const void*) sa;
275 for (ns = 0; ns < statp->nscount; ns++) {
276 srv = (struct sockaddr_in*) (void*) get_nsaddr(statp, (size_t) ns);
277 if (srv->sin_family == inp->sin_family && srv->sin_port == inp->sin_port &&
278 (srv->sin_addr.s_addr == INADDR_ANY ||
279 srv->sin_addr.s_addr == inp->sin_addr.s_addr))
280 return 1;
281 }
282 break;
283 case AF_INET6:
284 if (statp->_u._ext.ext == NULL) break;
285 in6p = (const struct sockaddr_in6*) (const void*) sa;
286 for (ns = 0; ns < statp->nscount; ns++) {
287 srv6 = (struct sockaddr_in6*) (void*) get_nsaddr(statp, (size_t) ns);
288 if (srv6->sin6_family == in6p->sin6_family && srv6->sin6_port == in6p->sin6_port &&
289 #ifdef HAVE_SIN6_SCOPE_ID
290 (srv6->sin6_scope_id == 0 || srv6->sin6_scope_id == in6p->sin6_scope_id) &&
291 #endif
292 (IN6_IS_ADDR_UNSPECIFIED(&srv6->sin6_addr) ||
293 IN6_ARE_ADDR_EQUAL(&srv6->sin6_addr, &in6p->sin6_addr)))
294 return 1;
295 }
296 break;
297 default:
298 break;
299 }
300 return 0;
301 }
302
303 /* int
304 * res_nameinquery(name, type, cl, buf, eom)
305 * look for (name, type, cl) in the query section of packet (buf, eom)
306 * requires:
307 * buf + HFIXEDSZ <= eom
308 * returns:
309 * -1 : format error
310 * 0 : not found
311 * >0 : found
312 * author:
313 * paul vixie, 29may94
314 */
res_nameinquery(const char * name,int type,int cl,const u_char * buf,const u_char * eom)315 int res_nameinquery(const char* name, int type, int cl, const u_char* buf, const u_char* eom) {
316 const u_char* cp = buf + HFIXEDSZ;
317 int qdcount = ntohs(((const HEADER*) (const void*) buf)->qdcount);
318
319 while (qdcount-- > 0) {
320 char tname[MAXDNAME + 1];
321 int n = dn_expand(buf, eom, cp, tname, sizeof tname);
322 if (n < 0) return (-1);
323 cp += n;
324 if (cp + 2 * INT16SZ > eom) return (-1);
325 int ttype = ntohs(*reinterpret_cast<const uint16_t*>(cp));
326 cp += INT16SZ;
327 int tclass = ntohs(*reinterpret_cast<const uint16_t*>(cp));
328 cp += INT16SZ;
329 if (ttype == type && tclass == cl && ns_samename(tname, name) == 1) return (1);
330 }
331 return (0);
332 }
333
334 /* int
335 * res_queriesmatch(buf1, eom1, buf2, eom2)
336 * is there a 1:1 mapping of (name,type,class)
337 * in (buf1,eom1) and (buf2,eom2)?
338 * returns:
339 * -1 : format error
340 * 0 : not a 1:1 mapping
341 * >0 : is a 1:1 mapping
342 * author:
343 * paul vixie, 29may94
344 */
res_queriesmatch(const u_char * buf1,const u_char * eom1,const u_char * buf2,const u_char * eom2)345 int res_queriesmatch(const u_char* buf1, const u_char* eom1, const u_char* buf2,
346 const u_char* eom2) {
347 const u_char* cp = buf1 + HFIXEDSZ;
348 int qdcount = ntohs(((const HEADER*) (const void*) buf1)->qdcount);
349
350 if (buf1 + HFIXEDSZ > eom1 || buf2 + HFIXEDSZ > eom2) return (-1);
351
352 /*
353 * Only header section present in replies to
354 * dynamic update packets.
355 */
356 if ((((const HEADER*) (const void*) buf1)->opcode == ns_o_update) &&
357 (((const HEADER*) (const void*) buf2)->opcode == ns_o_update))
358 return (1);
359
360 if (qdcount != ntohs(((const HEADER*) (const void*) buf2)->qdcount)) return (0);
361 while (qdcount-- > 0) {
362 char tname[MAXDNAME + 1];
363 int n = dn_expand(buf1, eom1, cp, tname, sizeof tname);
364 if (n < 0) return (-1);
365 cp += n;
366 if (cp + 2 * INT16SZ > eom1) return (-1);
367 int ttype = ntohs(*reinterpret_cast<const uint16_t*>(cp));
368 cp += INT16SZ;
369 int tclass = ntohs(*reinterpret_cast<const uint16_t*>(cp));
370 cp += INT16SZ;
371 if (!res_nameinquery(tname, ttype, tclass, buf2, eom2)) return (0);
372 }
373 return (1);
374 }
375
res_nsend(res_state statp,const u_char * buf,int buflen,u_char * ans,int anssiz,int * rcode,uint32_t flags)376 int res_nsend(res_state statp, const u_char* buf, int buflen, u_char* ans, int anssiz, int* rcode,
377 uint32_t flags) {
378 int gotsomewhere, terrno, v_circuit, resplen, n;
379 ResolvCacheStatus cache_status = RESOLV_CACHE_UNSUPPORTED;
380
381 if (anssiz < HFIXEDSZ) {
382 // TODO: Remove errno once callers stop using it
383 errno = EINVAL;
384 return -EINVAL;
385 }
386 LOG(DEBUG) << __func__;
387 res_pquery(buf, buflen);
388
389 v_circuit = (statp->options & RES_USEVC) || buflen > PACKETSZ;
390 gotsomewhere = 0;
391 terrno = ETIMEDOUT;
392
393 int anslen = 0;
394 cache_status = _resolv_cache_lookup(statp->netid, buf, buflen, ans, anssiz, &anslen, flags);
395
396 if (cache_status == RESOLV_CACHE_FOUND) {
397 HEADER* hp = (HEADER*)(void*)ans;
398 *rcode = hp->rcode;
399 return anslen;
400 } else if (cache_status != RESOLV_CACHE_UNSUPPORTED) {
401 // had a cache miss for a known network, so populate the thread private
402 // data so the normal resolve path can do its thing
403 _resolv_populate_res_for_net(statp);
404 }
405 if (statp->nscount == 0) {
406 // We have no nameservers configured, so there's no point trying.
407 // Tell the cache the query failed, or any retries and anyone else asking the same
408 // question will block for PENDING_REQUEST_TIMEOUT seconds instead of failing fast.
409 _resolv_cache_query_failed(statp->netid, buf, buflen, flags);
410
411 // TODO: Remove errno once callers stop using it
412 errno = ESRCH;
413 return -ESRCH;
414 }
415
416 /*
417 * If the ns_addr_list in the resolver context has changed, then
418 * invalidate our cached copy and the associated timing data.
419 */
420 if (statp->_u._ext.nscount != 0) {
421 int needclose = 0;
422 struct sockaddr_storage peer;
423 socklen_t peerlen;
424
425 if (statp->_u._ext.nscount != statp->nscount) {
426 needclose++;
427 } else {
428 for (int ns = 0; ns < statp->nscount; ns++) {
429 if (statp->nsaddr_list[ns].sin_family &&
430 !sock_eq((struct sockaddr*) (void*) &statp->nsaddr_list[ns],
431 (struct sockaddr*) (void*) &statp->_u._ext.ext->nsaddrs[ns])) {
432 needclose++;
433 break;
434 }
435
436 if (statp->_u._ext.nssocks[ns] == -1) continue;
437 peerlen = sizeof(peer);
438 if (getpeername(statp->_u._ext.nssocks[ns], (struct sockaddr*) (void*) &peer,
439 &peerlen) < 0) {
440 needclose++;
441 break;
442 }
443 if (!sock_eq((struct sockaddr*) (void*) &peer, get_nsaddr(statp, (size_t) ns))) {
444 needclose++;
445 break;
446 }
447 }
448 }
449 if (needclose) {
450 res_nclose(statp);
451 statp->_u._ext.nscount = 0;
452 }
453 }
454
455 /*
456 * Maybe initialize our private copy of the ns_addr_list.
457 */
458 if (statp->_u._ext.nscount == 0) {
459 for (int ns = 0; ns < statp->nscount; ns++) {
460 statp->_u._ext.nstimes[ns] = RES_MAXTIME;
461 statp->_u._ext.nssocks[ns] = -1;
462 if (!statp->nsaddr_list[ns].sin_family) continue;
463 statp->_u._ext.ext->nsaddrs[ns].sin = statp->nsaddr_list[ns];
464 }
465 statp->_u._ext.nscount = statp->nscount;
466 }
467
468 /*
469 * Some resolvers want to even out the load on their nameservers.
470 * Note that RES_BLAST overrides RES_ROTATE.
471 */
472 if ((statp->options & RES_ROTATE) != 0U && (statp->options & RES_BLAST) == 0U) {
473 sockaddr_union inu;
474 struct sockaddr_in ina;
475 int lastns = statp->nscount - 1;
476 int fd;
477 u_int16_t nstime;
478
479 if (statp->_u._ext.ext != NULL) inu = statp->_u._ext.ext->nsaddrs[0];
480 ina = statp->nsaddr_list[0];
481 fd = statp->_u._ext.nssocks[0];
482 nstime = statp->_u._ext.nstimes[0];
483 for (int ns = 0; ns < lastns; ns++) {
484 if (statp->_u._ext.ext != NULL)
485 statp->_u._ext.ext->nsaddrs[ns] = statp->_u._ext.ext->nsaddrs[ns + 1];
486 statp->nsaddr_list[ns] = statp->nsaddr_list[ns + 1];
487 statp->_u._ext.nssocks[ns] = statp->_u._ext.nssocks[ns + 1];
488 statp->_u._ext.nstimes[ns] = statp->_u._ext.nstimes[ns + 1];
489 }
490 if (statp->_u._ext.ext != NULL) statp->_u._ext.ext->nsaddrs[lastns] = inu;
491 statp->nsaddr_list[lastns] = ina;
492 statp->_u._ext.nssocks[lastns] = fd;
493 statp->_u._ext.nstimes[lastns] = nstime;
494 }
495
496 res_stats stats[MAXNS];
497 res_params params;
498 int revision_id = resolv_cache_get_resolver_stats(statp->netid, ¶ms, stats);
499 if (revision_id < 0) {
500 // TODO: Remove errno once callers stop using it
501 errno = ESRCH;
502 return -ESRCH;
503 }
504 bool usable_servers[MAXNS];
505 int usableServersCount = android_net_res_stats_get_usable_servers(
506 ¶ms, stats, statp->nscount, usable_servers);
507
508 if ((flags & ANDROID_RESOLV_NO_RETRY) && usableServersCount > 1) {
509 auto hp = reinterpret_cast<const HEADER*>(buf);
510
511 // Select a random server based on the query id
512 int selectedServer = (hp->id % usableServersCount) + 1;
513 res_set_usable_server(selectedServer, statp->nscount, usable_servers);
514 }
515
516 /*
517 * Send request, RETRY times, or until successful.
518 */
519 int retryTimes = (flags & ANDROID_RESOLV_NO_RETRY) ? 1 : params.retry_count;
520
521 for (int attempt = 0; attempt < retryTimes; ++attempt) {
522
523 for (int ns = 0; ns < statp->nscount; ns++) {
524 if (!usable_servers[ns]) continue;
525 struct sockaddr* nsap;
526 int nsaplen;
527 time_t now = 0;
528 int delay = 0;
529 *rcode = RCODE_INTERNAL_ERROR;
530 nsap = get_nsaddr(statp, (size_t) ns);
531 nsaplen = get_salen(nsap);
532
533 same_ns:
534 // TODO: Since we expect there is only one DNS server being queried here while this
535 // function tries to query all of private DNS servers. Consider moving it to other
536 // reasonable place. In addition, maybe add stats for private DNS.
537 if (!statp->use_local_nameserver) {
538 bool fallback = false;
539 resplen = res_tls_send(statp, Slice(const_cast<u_char*>(buf), buflen),
540 Slice(ans, anssiz), rcode, &fallback);
541 if (resplen > 0) {
542 if (cache_status == RESOLV_CACHE_NOTFOUND) {
543 _resolv_cache_add(statp->netid, buf, buflen, ans, resplen);
544 }
545 return resplen;
546 }
547 if (!fallback) {
548 _resolv_cache_query_failed(statp->netid, buf, buflen, flags);
549 res_nclose(statp);
550 return -terrno;
551 }
552 }
553
554 [[maybe_unused]] static const int niflags = NI_NUMERICHOST | NI_NUMERICSERV;
555 [[maybe_unused]] char abuf[NI_MAXHOST];
556
557 if (getnameinfo(nsap, (socklen_t)nsaplen, abuf, sizeof(abuf), NULL, 0, niflags) == 0)
558 LOG(DEBUG) << __func__ << ": Querying server (# " << ns + 1
559 << ") address = " << abuf;
560
561 if (v_circuit) {
562 /* Use VC; at most one attempt per server. */
563 bool shouldRecordStats = (attempt == 0);
564 attempt = retryTimes;
565
566 n = send_vc(statp, ¶ms, buf, buflen, ans, anssiz, &terrno, ns, &now, rcode,
567 &delay);
568
569 /*
570 * Only record stats the first time we try a query. This ensures that
571 * queries that deterministically fail (e.g., a name that always returns
572 * SERVFAIL or times out) do not unduly affect the stats.
573 */
574 if (shouldRecordStats) {
575 res_sample sample;
576 _res_stats_set_sample(&sample, now, *rcode, delay);
577 _resolv_cache_add_resolver_stats_sample(statp->netid, revision_id, ns, &sample,
578 params.max_samples);
579 }
580
581 LOG(INFO) << __func__ << ": used send_vc " << n;
582
583 if (n < 0) {
584 _resolv_cache_query_failed(statp->netid, buf, buflen, flags);
585 res_nclose(statp);
586 return -terrno;
587 };
588 if (n == 0) goto next_ns;
589 resplen = n;
590 } else {
591 /* Use datagrams. */
592 LOG(INFO) << __func__ << ": using send_dg";
593
594 n = send_dg(statp, ¶ms, buf, buflen, ans, anssiz, &terrno, ns, &v_circuit,
595 &gotsomewhere, &now, rcode, &delay);
596
597 /* Only record stats the first time we try a query. See above. */
598 if (attempt == 0) {
599 res_sample sample;
600 _res_stats_set_sample(&sample, now, *rcode, delay);
601 _resolv_cache_add_resolver_stats_sample(statp->netid, revision_id, ns, &sample,
602 params.max_samples);
603 }
604
605 LOG(INFO) << __func__ << ": used send_dg " << n;
606
607 if (n < 0) {
608 _resolv_cache_query_failed(statp->netid, buf, buflen, flags);
609 res_nclose(statp);
610 return -terrno;
611 };
612 if (n == 0) goto next_ns;
613 if (v_circuit) goto same_ns;
614 resplen = n;
615 }
616
617 LOG(DEBUG) << __func__ << ": got answer:";
618 res_pquery(ans, (resplen > anssiz) ? anssiz : resplen);
619
620 if (cache_status == RESOLV_CACHE_NOTFOUND) {
621 _resolv_cache_add(statp->netid, buf, buflen, ans, resplen);
622 }
623 /*
624 * If we have temporarily opened a virtual circuit,
625 * or if we haven't been asked to keep a socket open,
626 * close the socket.
627 */
628 if ((v_circuit && (statp->options & RES_USEVC) == 0U) ||
629 (statp->options & RES_STAYOPEN) == 0U) {
630 res_nclose(statp);
631 }
632 return (resplen);
633 next_ns:;
634 } // for each ns
635 } // for each retry
636 res_nclose(statp);
637 if (!v_circuit) {
638 if (!gotsomewhere) {
639 // TODO: Remove errno once callers stop using it
640 errno = ECONNREFUSED; /* no nameservers found */
641 terrno = ECONNREFUSED;
642 } else {
643 // TODO: Remove errno once callers stop using it
644 errno = ETIMEDOUT; /* no answer obtained */
645 terrno = ETIMEDOUT;
646 }
647 } else {
648 errno = terrno;
649 }
650 _resolv_cache_query_failed(statp->netid, buf, buflen, flags);
651 return -terrno;
652 }
653
654 /* Private */
655
get_salen(const struct sockaddr * sa)656 static int get_salen(const struct sockaddr* sa) {
657 if (sa->sa_family == AF_INET)
658 return (sizeof(struct sockaddr_in));
659 else if (sa->sa_family == AF_INET6)
660 return (sizeof(struct sockaddr_in6));
661 else
662 return (0); /* unknown, die on connect */
663 }
664
665 /*
666 * pick appropriate nsaddr_list for use. see res_init() for initialization.
667 */
get_nsaddr(res_state statp,size_t n)668 static struct sockaddr* get_nsaddr(res_state statp, size_t n) {
669 if (!statp->nsaddr_list[n].sin_family && statp->_u._ext.ext) {
670 /*
671 * - statp->_u._ext.ext->nsaddrs[n] holds an address that is larger
672 * than struct sockaddr, and
673 * - user code did not update statp->nsaddr_list[n].
674 */
675 return (struct sockaddr*) (void*) &statp->_u._ext.ext->nsaddrs[n];
676 } else {
677 /*
678 * - user code updated statp->nsaddr_list[n], or
679 * - statp->nsaddr_list[n] has the same content as
680 * statp->_u._ext.ext->nsaddrs[n].
681 */
682 return (struct sockaddr*) (void*) &statp->nsaddr_list[n];
683 }
684 }
685
get_timeout(const res_state statp,const res_params * params,const int ns)686 static struct timespec get_timeout(const res_state statp, const res_params* params, const int ns) {
687 int msec;
688 // Legacy algorithm which scales the timeout by nameserver number.
689 // For instance, with 4 nameservers: 5s, 2.5s, 5s, 10s
690 // This has no effect with 1 or 2 nameservers
691 msec = params->base_timeout_msec << ns;
692 if (ns > 0) {
693 msec /= statp->nscount;
694 }
695 // For safety, don't allow OEMs and experiments to configure a timeout shorter than 1s.
696 if (msec < 1000) {
697 msec = 1000; // Use at least 1000ms
698 }
699 LOG(INFO) << __func__ << ": using timeout of " << msec << " msec";
700
701 struct timespec result;
702 result.tv_sec = msec / 1000;
703 result.tv_nsec = (msec % 1000) * 1000000;
704 return result;
705 }
706
send_vc(res_state statp,res_params * params,const u_char * buf,int buflen,u_char * ans,int anssiz,int * terrno,int ns,time_t * at,int * rcode,int * delay)707 static int send_vc(res_state statp, res_params* params, const u_char* buf, int buflen, u_char* ans,
708 int anssiz, int* terrno, int ns, time_t* at, int* rcode, int* delay) {
709 *at = time(NULL);
710 *delay = 0;
711 const HEADER* hp = (const HEADER*) (const void*) buf;
712 HEADER* anhp = (HEADER*) (void*) ans;
713 struct sockaddr* nsap;
714 int nsaplen;
715 int truncating, connreset, n;
716 struct iovec iov[2];
717 u_char* cp;
718
719 LOG(INFO) << __func__ << ": using send_vc";
720
721 nsap = get_nsaddr(statp, (size_t) ns);
722 nsaplen = get_salen(nsap);
723
724 connreset = 0;
725 same_ns:
726 truncating = 0;
727
728 struct timespec now = evNowTime();
729
730 /* Are we still talking to whom we want to talk to? */
731 if (statp->_vcsock >= 0 && (statp->_flags & RES_F_VC) != 0) {
732 struct sockaddr_storage peer;
733 socklen_t size = sizeof peer;
734 unsigned old_mark;
735 socklen_t mark_size = sizeof(old_mark);
736 if (getpeername(statp->_vcsock, (struct sockaddr*) (void*) &peer, &size) < 0 ||
737 !sock_eq((struct sockaddr*) (void*) &peer, nsap) ||
738 getsockopt(statp->_vcsock, SOL_SOCKET, SO_MARK, &old_mark, &mark_size) < 0 ||
739 old_mark != statp->_mark) {
740 res_nclose(statp);
741 statp->_flags &= ~RES_F_VC;
742 }
743 }
744
745 if (statp->_vcsock < 0 || (statp->_flags & RES_F_VC) == 0) {
746 if (statp->_vcsock >= 0) res_nclose(statp);
747
748 statp->_vcsock = socket(nsap->sa_family, SOCK_STREAM | SOCK_CLOEXEC, 0);
749 if (statp->_vcsock < 0) {
750 switch (errno) {
751 case EPROTONOSUPPORT:
752 case EPFNOSUPPORT:
753 case EAFNOSUPPORT:
754 Perror(statp, "socket(vc)", errno);
755 return 0;
756 default:
757 *terrno = errno;
758 Perror(statp, "socket(vc)", errno);
759 return -1;
760 }
761 }
762 fchown(statp->_vcsock, AID_DNS, -1);
763 if (statp->_mark != MARK_UNSET) {
764 if (setsockopt(statp->_vcsock, SOL_SOCKET, SO_MARK, &statp->_mark,
765 sizeof(statp->_mark)) < 0) {
766 *terrno = errno;
767 Perror(statp, "setsockopt", errno);
768 return -1;
769 }
770 }
771 errno = 0;
772 if (random_bind(statp->_vcsock, nsap->sa_family) < 0) {
773 *terrno = errno;
774 Aerror(statp, "bind/vc", errno, nsap, nsaplen);
775 res_nclose(statp);
776 return (0);
777 }
778 if (connect_with_timeout(statp->_vcsock, nsap, (socklen_t) nsaplen,
779 get_timeout(statp, params, ns)) < 0) {
780 *terrno = errno;
781 Aerror(statp, "connect/vc", errno, nsap, nsaplen);
782 res_nclose(statp);
783 /*
784 * The way connect_with_timeout() is implemented prevents us from reliably
785 * determining whether this was really a timeout or e.g. ECONNREFUSED. Since
786 * currently both cases are handled in the same way, there is no need to
787 * change this (yet). If we ever need to reliably distinguish between these
788 * cases, both connect_with_timeout() and retrying_poll() need to be
789 * modified, though.
790 */
791 *rcode = RCODE_TIMEOUT;
792 return (0);
793 }
794 statp->_flags |= RES_F_VC;
795 }
796
797 /*
798 * Send length & message
799 */
800 uint16_t len = htons(static_cast<uint16_t>(buflen));
801 iov[0] = evConsIovec(&len, INT16SZ);
802 iov[1] = evConsIovec((void*) buf, (size_t) buflen);
803 if (writev(statp->_vcsock, iov, 2) != (INT16SZ + buflen)) {
804 *terrno = errno;
805 Perror(statp, "write failed", errno);
806 res_nclose(statp);
807 return (0);
808 }
809 /*
810 * Receive length & response
811 */
812 read_len:
813 cp = ans;
814 len = INT16SZ;
815 while ((n = read(statp->_vcsock, (char*) cp, (size_t) len)) > 0) {
816 cp += n;
817 if ((len -= n) == 0) break;
818 }
819 if (n <= 0) {
820 *terrno = errno;
821 Perror(statp, "read failed", errno);
822 res_nclose(statp);
823 /*
824 * A long running process might get its TCP
825 * connection reset if the remote server was
826 * restarted. Requery the server instead of
827 * trying a new one. When there is only one
828 * server, this means that a query might work
829 * instead of failing. We only allow one reset
830 * per query to prevent looping.
831 */
832 if (*terrno == ECONNRESET && !connreset) {
833 connreset = 1;
834 res_nclose(statp);
835 goto same_ns;
836 }
837 res_nclose(statp);
838 return (0);
839 }
840 uint16_t resplen = ntohs(*reinterpret_cast<const uint16_t*>(ans));
841 if (resplen > anssiz) {
842 LOG(DEBUG) << __func__ << ": response truncated";
843 truncating = 1;
844 len = anssiz;
845 } else
846 len = resplen;
847 if (len < HFIXEDSZ) {
848 /*
849 * Undersized message.
850 */
851 LOG(DEBUG) << __func__ << ": undersized: " << len;
852 *terrno = EMSGSIZE;
853 res_nclose(statp);
854 return (0);
855 }
856 cp = ans;
857 while (len != 0 && (n = read(statp->_vcsock, (char*) cp, (size_t) len)) > 0) {
858 cp += n;
859 len -= n;
860 }
861 if (n <= 0) {
862 *terrno = errno;
863 Perror(statp, "read(vc)", errno);
864 res_nclose(statp);
865 return (0);
866 }
867
868 if (truncating) {
869 /*
870 * Flush rest of answer so connection stays in synch.
871 */
872 anhp->tc = 1;
873 len = resplen - anssiz;
874 while (len != 0) {
875 char junk[PACKETSZ];
876
877 n = read(statp->_vcsock, junk, (len > sizeof junk) ? sizeof junk : len);
878 if (n > 0)
879 len -= n;
880 else
881 break;
882 }
883 }
884 /*
885 * If the calling application has bailed out of
886 * a previous call and failed to arrange to have
887 * the circuit closed or the server has got
888 * itself confused, then drop the packet and
889 * wait for the correct one.
890 */
891 if (hp->id != anhp->id) {
892 LOG(DEBUG) << __func__ << ": ld answer (unexpected):";
893 res_pquery(ans, (resplen > anssiz) ? anssiz : resplen);
894 goto read_len;
895 }
896
897 /*
898 * All is well, or the error is fatal. Signal that the
899 * next nameserver ought not be tried.
900 */
901 if (resplen > 0) {
902 struct timespec done = evNowTime();
903 *delay = _res_stats_calculate_rtt(&done, &now);
904 *rcode = anhp->rcode;
905 }
906 return (resplen);
907 }
908
909 /* return -1 on error (errno set), 0 on success */
connect_with_timeout(int sock,const struct sockaddr * nsap,socklen_t salen,const struct timespec timeout)910 static int connect_with_timeout(int sock, const struct sockaddr* nsap, socklen_t salen,
911 const struct timespec timeout) {
912 int res, origflags;
913
914 origflags = fcntl(sock, F_GETFL, 0);
915 fcntl(sock, F_SETFL, origflags | O_NONBLOCK);
916
917 res = connect(sock, nsap, salen);
918 if (res < 0 && errno != EINPROGRESS) {
919 res = -1;
920 goto done;
921 }
922 if (res != 0) {
923 struct timespec now = evNowTime();
924 struct timespec finish = evAddTime(now, timeout);
925 LOG(INFO) << __func__ << ": " << sock << " send_vc";
926 res = retrying_poll(sock, POLLIN | POLLOUT, &finish);
927 if (res <= 0) {
928 res = -1;
929 }
930 }
931 done:
932 fcntl(sock, F_SETFL, origflags);
933 LOG(INFO) << __func__ << ": " << sock << " connect_with_const timeout returning " << res;
934 return res;
935 }
936
retrying_poll(const int sock,const short events,const struct timespec * finish)937 static int retrying_poll(const int sock, const short events, const struct timespec* finish) {
938 struct timespec now, timeout;
939
940 retry:
941 LOG(INFO) << __func__ << ": " << sock << " retrying_poll";
942
943 now = evNowTime();
944 if (evCmpTime(*finish, now) > 0)
945 timeout = evSubTime(*finish, now);
946 else
947 timeout = evConsTime(0L, 0L);
948 struct pollfd fds = {.fd = sock, .events = events};
949 int n = ppoll(&fds, 1, &timeout, /*sigmask=*/NULL);
950 if (n == 0) {
951 LOG(INFO) << __func__ << ": " << sock << "retrying_poll timeout";
952 errno = ETIMEDOUT;
953 return 0;
954 }
955 if (n < 0) {
956 if (errno == EINTR) goto retry;
957 PLOG(INFO) << __func__ << ": " << sock << " retrying_poll failed";
958 return n;
959 }
960 if (fds.revents & (POLLIN | POLLOUT | POLLERR)) {
961 int error;
962 socklen_t len = sizeof(error);
963 if (getsockopt(sock, SOL_SOCKET, SO_ERROR, &error, &len) < 0 || error) {
964 errno = error;
965 PLOG(INFO) << __func__ << ": " << sock << " retrying_poll getsockopt failed";
966 return -1;
967 }
968 }
969 LOG(INFO) << __func__ << ": " << sock << " retrying_poll returning " << n;
970 return n;
971 }
972
send_dg(res_state statp,res_params * params,const u_char * buf,int buflen,u_char * ans,int anssiz,int * terrno,int ns,int * v_circuit,int * gotsomewhere,time_t * at,int * rcode,int * delay)973 static int send_dg(res_state statp, res_params* params, const u_char* buf, int buflen, u_char* ans,
974 int anssiz, int* terrno, int ns, int* v_circuit, int* gotsomewhere, time_t* at,
975 int* rcode, int* delay) {
976 *at = time(NULL);
977 *delay = 0;
978 const HEADER* hp = (const HEADER*) (const void*) buf;
979 HEADER* anhp = (HEADER*) (void*) ans;
980 const struct sockaddr* nsap;
981 int nsaplen;
982 struct timespec now, timeout, finish, done;
983 struct sockaddr_storage from;
984 socklen_t fromlen;
985 int resplen, n, s;
986
987 nsap = get_nsaddr(statp, (size_t) ns);
988 nsaplen = get_salen(nsap);
989 if (statp->_u._ext.nssocks[ns] == -1) {
990 statp->_u._ext.nssocks[ns] = socket(nsap->sa_family, SOCK_DGRAM | SOCK_CLOEXEC, 0);
991 if (statp->_u._ext.nssocks[ns] < 0) {
992 switch (errno) {
993 case EPROTONOSUPPORT:
994 case EPFNOSUPPORT:
995 case EAFNOSUPPORT:
996 Perror(statp, "socket(dg)", errno);
997 return (0);
998 default:
999 *terrno = errno;
1000 Perror(statp, "socket(dg)", errno);
1001 return (-1);
1002 }
1003 }
1004
1005 fchown(statp->_u._ext.nssocks[ns], AID_DNS, -1);
1006 if (statp->_mark != MARK_UNSET) {
1007 if (setsockopt(statp->_u._ext.nssocks[ns], SOL_SOCKET, SO_MARK, &(statp->_mark),
1008 sizeof(statp->_mark)) < 0) {
1009 res_nclose(statp);
1010 return -1;
1011 }
1012 }
1013 #ifndef CANNOT_CONNECT_DGRAM
1014 /*
1015 * On a 4.3BSD+ machine (client and server,
1016 * actually), sending to a nameserver datagram
1017 * port with no nameserver will cause an
1018 * ICMP port unreachable message to be returned.
1019 * If our datagram socket is "connected" to the
1020 * server, we get an ECONNREFUSED error on the next
1021 * socket operation, and select returns if the
1022 * error message is received. We can thus detect
1023 * the absence of a nameserver without timing out.
1024 */
1025 if (random_bind(statp->_u._ext.nssocks[ns], nsap->sa_family) < 0) {
1026 Aerror(statp, "bind(dg)", errno, nsap, nsaplen);
1027 res_nclose(statp);
1028 return (0);
1029 }
1030 if (connect(statp->_u._ext.nssocks[ns], nsap, (socklen_t) nsaplen) < 0) {
1031 Aerror(statp, "connect(dg)", errno, nsap, nsaplen);
1032 res_nclose(statp);
1033 return (0);
1034 }
1035 #endif /* !CANNOT_CONNECT_DGRAM */
1036 LOG(DEBUG) << __func__ << ": new DG socket";
1037 }
1038 s = statp->_u._ext.nssocks[ns];
1039 #ifndef CANNOT_CONNECT_DGRAM
1040 if (send(s, (const char*) buf, (size_t) buflen, 0) != buflen) {
1041 Perror(statp, "send", errno);
1042 res_nclose(statp);
1043 return 0;
1044 }
1045 #else /* !CANNOT_CONNECT_DGRAM */
1046 if (sendto(s, (const char*) buf, buflen, 0, nsap, nsaplen) != buflen) {
1047 Aerror(statp, "sendto", errno, nsap, nsaplen);
1048 res_nclose(statp);
1049 return 0;
1050 }
1051 #endif /* !CANNOT_CONNECT_DGRAM */
1052
1053 // Wait for reply.
1054 timeout = get_timeout(statp, params, ns);
1055 now = evNowTime();
1056 finish = evAddTime(now, timeout);
1057 retry:
1058 n = retrying_poll(s, POLLIN, &finish);
1059
1060 if (n == 0) {
1061 *rcode = RCODE_TIMEOUT;
1062 LOG(DEBUG) << __func__ << ": timeout";
1063 *gotsomewhere = 1;
1064 return 0;
1065 }
1066 if (n < 0) {
1067 Perror(statp, "poll", errno);
1068 res_nclose(statp);
1069 return 0;
1070 }
1071 errno = 0;
1072 fromlen = sizeof(from);
1073 resplen = recvfrom(s, (char*) ans, (size_t) anssiz, 0, (struct sockaddr*) (void*) &from,
1074 &fromlen);
1075 if (resplen <= 0) {
1076 Perror(statp, "recvfrom", errno);
1077 res_nclose(statp);
1078 return 0;
1079 }
1080 *gotsomewhere = 1;
1081 if (resplen < HFIXEDSZ) {
1082 /*
1083 * Undersized message.
1084 */
1085 LOG(DEBUG) << __func__ << ": undersized: " << resplen;
1086 *terrno = EMSGSIZE;
1087 res_nclose(statp);
1088 return 0;
1089 }
1090 if (hp->id != anhp->id) {
1091 /*
1092 * response from old query, ignore it.
1093 * XXX - potential security hazard could
1094 * be detected here.
1095 */
1096 LOG(DEBUG) << __func__ << ": old answer:";
1097 res_pquery(ans, (resplen > anssiz) ? anssiz : resplen);
1098 goto retry;
1099 }
1100 if (!(statp->options & RES_INSECURE1) &&
1101 !res_ourserver_p(statp, (struct sockaddr*) (void*) &from)) {
1102 /*
1103 * response from wrong server? ignore it.
1104 * XXX - potential security hazard could
1105 * be detected here.
1106 */
1107 LOG(DEBUG) << __func__ << ": not our server:";
1108 res_pquery(ans, (resplen > anssiz) ? anssiz : resplen);
1109 goto retry;
1110 }
1111 if (anhp->rcode == FORMERR && (statp->options & RES_USE_EDNS0) != 0U) {
1112 /*
1113 * Do not retry if the server do not understand EDNS0.
1114 * The case has to be captured here, as FORMERR packet do not
1115 * carry query section, hence res_queriesmatch() returns 0.
1116 */
1117 LOG(DEBUG) << __func__ << ": server rejected query with EDNS0:";
1118 res_pquery(ans, (resplen > anssiz) ? anssiz : resplen);
1119 /* record the error */
1120 statp->_flags |= RES_F_EDNS0ERR;
1121 res_nclose(statp);
1122 return 0;
1123 }
1124 if (!(statp->options & RES_INSECURE2) &&
1125 !res_queriesmatch(buf, buf + buflen, ans, ans + anssiz)) {
1126 /*
1127 * response contains wrong query? ignore it.
1128 * XXX - potential security hazard could
1129 * be detected here.
1130 */
1131 LOG(DEBUG) << __func__ << ": wrong query name:";
1132 res_pquery(ans, (resplen > anssiz) ? anssiz : resplen);
1133 goto retry;
1134 }
1135 done = evNowTime();
1136 *delay = _res_stats_calculate_rtt(&done, &now);
1137 if (anhp->rcode == SERVFAIL || anhp->rcode == NOTIMP || anhp->rcode == REFUSED) {
1138 LOG(DEBUG) << __func__ << ": server rejected query:";
1139 res_pquery(ans, (resplen > anssiz) ? anssiz : resplen);
1140 res_nclose(statp);
1141 *rcode = anhp->rcode;
1142 return 0;
1143 }
1144 if (!(statp->options & RES_IGNTC) && anhp->tc) {
1145 /*
1146 * To get the rest of answer,
1147 * use TCP with same server.
1148 */
1149 LOG(DEBUG) << __func__ << ": truncated answer";
1150 *v_circuit = 1;
1151 res_nclose(statp);
1152 return 1;
1153 }
1154 /*
1155 * All is well, or the error is fatal. Signal that the
1156 * next nameserver ought not be tried.
1157 */
1158 if (resplen > 0) {
1159 *rcode = anhp->rcode;
1160 }
1161 return resplen;
1162 }
1163
Aerror(const res_state statp,const char * string,int error,const struct sockaddr * address,int alen)1164 static void Aerror(const res_state statp, const char* string, int error,
1165 const struct sockaddr* address, int alen) {
1166 const int save = errno;
1167 char hbuf[NI_MAXHOST];
1168 char sbuf[NI_MAXSERV];
1169 constexpr int niflags = NI_NUMERICHOST | NI_NUMERICSERV;
1170
1171 if ((statp->options & RES_DEBUG) != 0U) {
1172 if (getnameinfo(address, (socklen_t) alen, hbuf, sizeof(hbuf), sbuf, sizeof(sbuf),
1173 niflags)) {
1174 strncpy(hbuf, "?", sizeof(hbuf) - 1);
1175 hbuf[sizeof(hbuf) - 1] = '\0';
1176 strncpy(sbuf, "?", sizeof(sbuf) - 1);
1177 sbuf[sizeof(sbuf) - 1] = '\0';
1178 }
1179 LOG(DEBUG) << __func__ << ": " << string << " ([" << hbuf << "]." << sbuf
1180 << "): " << strerror(error);
1181 }
1182 errno = save;
1183 }
1184
Perror(const res_state statp,const char * string,int error)1185 static void Perror(const res_state statp, const char* string, int error) {
1186 if ((statp->options & RES_DEBUG) != 0U) {
1187 LOG(DEBUG) << __func__ << ": " << string << ": " << strerror(error);
1188 }
1189 }
1190
sock_eq(struct sockaddr * a,struct sockaddr * b)1191 static int sock_eq(struct sockaddr* a, struct sockaddr* b) {
1192 struct sockaddr_in *a4, *b4;
1193 struct sockaddr_in6 *a6, *b6;
1194
1195 if (a->sa_family != b->sa_family) return 0;
1196 switch (a->sa_family) {
1197 case AF_INET:
1198 a4 = (struct sockaddr_in*) (void*) a;
1199 b4 = (struct sockaddr_in*) (void*) b;
1200 return a4->sin_port == b4->sin_port && a4->sin_addr.s_addr == b4->sin_addr.s_addr;
1201 case AF_INET6:
1202 a6 = (struct sockaddr_in6*) (void*) a;
1203 b6 = (struct sockaddr_in6*) (void*) b;
1204 return a6->sin6_port == b6->sin6_port &&
1205 #ifdef HAVE_SIN6_SCOPE_ID
1206 a6->sin6_scope_id == b6->sin6_scope_id &&
1207 #endif
1208 IN6_ARE_ADDR_EQUAL(&a6->sin6_addr, &b6->sin6_addr);
1209 default:
1210 return 0;
1211 }
1212 }
1213
res_tls_send(res_state statp,const Slice query,const Slice answer,int * rcode,bool * fallback)1214 static int res_tls_send(res_state statp, const Slice query, const Slice answer, int* rcode,
1215 bool* fallback) {
1216 int resplen = 0;
1217 const unsigned netId = statp->netid;
1218 const unsigned mark = statp->_mark;
1219
1220 PrivateDnsStatus privateDnsStatus = gPrivateDnsConfiguration.getStatus(netId);
1221
1222 if (privateDnsStatus.mode == PrivateDnsMode::OFF) {
1223 *fallback = true;
1224 return -1;
1225 }
1226
1227 if (privateDnsStatus.validatedServers.empty()) {
1228 if (privateDnsStatus.mode == PrivateDnsMode::OPPORTUNISTIC) {
1229 *fallback = true;
1230 return -1;
1231 } else {
1232 // Sleep and iterate some small number of times checking for the
1233 // arrival of resolved and validated server IP addresses, instead
1234 // of returning an immediate error.
1235 // This is needed because as soon as a network becomes the default network, apps will
1236 // send DNS queries on that network. If no servers have yet validated, and we do not
1237 // block those queries, they would immediately fail, causing application-visible errors.
1238 // Note that this can happen even before the network validates, since an unvalidated
1239 // network can become the default network if no validated networks are available.
1240 //
1241 // TODO: see if there is a better way to address this problem, such as buffering the
1242 // queries in a queue or only blocking queries for the first few seconds after a default
1243 // network change.
1244 for (int i = 0; i < 42; i++) {
1245 std::this_thread::sleep_for(std::chrono::milliseconds(100));
1246 if (!gPrivateDnsConfiguration.getStatus(netId).validatedServers.empty()) {
1247 privateDnsStatus = gPrivateDnsConfiguration.getStatus(netId);
1248 break;
1249 }
1250 }
1251 if (privateDnsStatus.validatedServers.empty()) {
1252 return -1;
1253 }
1254 }
1255 }
1256
1257 LOG(INFO) << __func__ << ": performing query over TLS";
1258
1259 const auto response = sDnsTlsDispatcher.query(privateDnsStatus.validatedServers, mark, query,
1260 answer, &resplen);
1261
1262 LOG(INFO) << __func__ << ": TLS query result: " << static_cast<int>(response);
1263
1264 if (privateDnsStatus.mode == PrivateDnsMode::OPPORTUNISTIC) {
1265 // In opportunistic mode, handle falling back to cleartext in some
1266 // cases (DNS shouldn't fail if a validated opportunistic mode server
1267 // becomes unreachable for some reason).
1268 switch (response) {
1269 case DnsTlsTransport::Response::success:
1270 *rcode = reinterpret_cast<HEADER*>(answer.base())->rcode;
1271 return resplen;
1272 case DnsTlsTransport::Response::network_error:
1273 // No need to set the error timeout here since it will fallback to UDP.
1274 case DnsTlsTransport::Response::internal_error:
1275 // Note: this will cause cleartext queries to be emitted, with
1276 // all of the EDNS0 goodness enabled. Fingers crossed. :-/
1277 *fallback = true;
1278 [[fallthrough]];
1279 default:
1280 return -1;
1281 }
1282 } else {
1283 // Strict mode
1284 switch (response) {
1285 case DnsTlsTransport::Response::success:
1286 *rcode = reinterpret_cast<HEADER*>(answer.base())->rcode;
1287 return resplen;
1288 case DnsTlsTransport::Response::network_error:
1289 // This case happens when the query stored in DnsTlsTransport is expired since
1290 // either 1) the query has been tried for 3 times but no response or 2) fail to
1291 // establish the connection with the server.
1292 *rcode = RCODE_TIMEOUT;
1293 [[fallthrough]];
1294 default:
1295 return -1;
1296 }
1297 }
1298 }
1299
resolv_res_nsend(const android_net_context * netContext,const uint8_t * msg,int msgLen,uint8_t * ans,int ansLen,int * rcode,uint32_t flags)1300 int resolv_res_nsend(const android_net_context* netContext, const uint8_t* msg, int msgLen,
1301 uint8_t* ans, int ansLen, int* rcode, uint32_t flags) {
1302 res_state res = res_get_state();
1303 res_setnetcontext(res, netContext);
1304 _resolv_populate_res_for_net(res);
1305 *rcode = NOERROR;
1306 return res_nsend(res, msg, msgLen, ans, ansLen, rcode, flags);
1307 }
1308