res_send.c revision 270838
1/*
2 * Portions Copyright (C) 2004-2009  Internet Systems Consortium, Inc. ("ISC")
3 * Portions Copyright (C) 1996-2003  Internet Software Consortium.
4 *
5 * Permission to use, copy, modify, and/or distribute this software for any
6 * purpose with or without fee is hereby granted, provided that the above
7 * copyright notice and this permission notice appear in all copies.
8 *
9 * THE SOFTWARE IS PROVIDED "AS IS" AND ISC DISCLAIMS ALL WARRANTIES WITH
10 * REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY
11 * AND FITNESS.  IN NO EVENT SHALL ISC BE LIABLE FOR ANY SPECIAL, DIRECT,
12 * INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM
13 * LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE
14 * OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR
15 * PERFORMANCE OF THIS SOFTWARE.
16 */
17
18/*
19 * Copyright (c) 1985, 1989, 1993
20 *    The Regents of the University of California.  All rights reserved.
21 *
22 * Redistribution and use in source and binary forms, with or without
23 * modification, are permitted provided that the following conditions
24 * are met:
25 * 1. Redistributions of source code must retain the above copyright
26 *    notice, this list of conditions and the following disclaimer.
27 * 2. Redistributions in binary form must reproduce the above copyright
28 *    notice, this list of conditions and the following disclaimer in the
29 *    documentation and/or other materials provided with the distribution.
30 * 4. Neither the name of the University nor the names of its contributors
31 *    may be used to endorse or promote products derived from this software
32 *    without specific prior written permission.
33 *
34 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
35 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
36 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
37 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
38 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
39 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
40 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
41 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
42 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
43 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
44 * SUCH DAMAGE.
45 */
46
47/*
48 * Portions Copyright (c) 1993 by Digital Equipment Corporation.
49 *
50 * Permission to use, copy, modify, and distribute this software for any
51 * purpose with or without fee is hereby granted, provided that the above
52 * copyright notice and this permission notice appear in all copies, and that
53 * the name of Digital Equipment Corporation not be used in advertising or
54 * publicity pertaining to distribution of the document or software without
55 * specific, written prior permission.
56 *
57 * THE SOFTWARE IS PROVIDED "AS IS" AND DIGITAL EQUIPMENT CORP. DISCLAIMS ALL
58 * WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES
59 * OF MERCHANTABILITY AND FITNESS.   IN NO EVENT SHALL DIGITAL EQUIPMENT
60 * CORPORATION BE LIABLE FOR ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL
61 * DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR
62 * PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
63 * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS
64 * SOFTWARE.
65 */
66
67#if defined(LIBC_SCCS) && !defined(lint)
68static const char sccsid[] = "@(#)res_send.c	8.1 (Berkeley) 6/4/93";
69static const char rcsid[] = "$Id: res_send.c,v 1.22 2009/01/22 23:49:23 tbox Exp $";
70#endif /* LIBC_SCCS and not lint */
71#include <sys/cdefs.h>
72__FBSDID("$FreeBSD: stable/10/lib/libc/resolv/res_send.c 270838 2014-08-30 10:16:25Z ume $");
73
74/*! \file
75 * \brief
76 * Send query to name server and wait for reply.
77 */
78
79#include "port_before.h"
80#ifndef USE_KQUEUE
81#include "fd_setsize.h"
82#endif
83
84#include "namespace.h"
85#include <sys/types.h>
86#include <sys/param.h>
87#include <sys/time.h>
88#include <sys/socket.h>
89#include <sys/uio.h>
90
91#include <netinet/in.h>
92#include <arpa/nameser.h>
93#include <arpa/inet.h>
94
95#include <errno.h>
96#include <netdb.h>
97#include <resolv.h>
98#include <signal.h>
99#include <stdio.h>
100#include <stdlib.h>
101#include <string.h>
102#include <unistd.h>
103
104#include <isc/eventlib.h>
105
106#include "port_after.h"
107
108#ifdef USE_KQUEUE
109#include <sys/event.h>
110#else
111#ifdef USE_POLL
112#ifdef HAVE_STROPTS_H
113#include <stropts.h>
114#endif
115#include <poll.h>
116#endif /* USE_POLL */
117#endif
118
119#include "un-namespace.h"
120
121/* Options.  Leave them on. */
122#define DEBUG
123#include "res_debug.h"
124#include "res_private.h"
125
126#define EXT(res) ((res)->_u._ext)
127
128#if !defined(USE_POLL) && !defined(USE_KQUEUE)
129static const int highestFD = FD_SETSIZE - 1;
130#endif
131
132/* Forward. */
133
134static int		get_salen(const struct sockaddr *);
135static struct sockaddr * get_nsaddr(res_state, size_t);
136static int		send_vc(res_state, const u_char *, int,
137				u_char *, int, int *, int);
138static int		send_dg(res_state,
139#ifdef USE_KQUEUE
140				int kq,
141#endif
142				const u_char *, int,
143				u_char *, int, int *, int, int,
144				int *, int *);
145static void		Aerror(const res_state, FILE *, const char *, int,
146			       const struct sockaddr *, int);
147static void		Perror(const res_state, FILE *, const char *, int);
148static int		sock_eq(struct sockaddr *, struct sockaddr *);
149#if defined(NEED_PSELECT) && !defined(USE_POLL) && !defined(USE_KQUEUE)
150static int		pselect(int, void *, void *, void *,
151				struct timespec *,
152				const sigset_t *);
153#endif
154void res_pquery(const res_state, const u_char *, int, FILE *);
155
156static const int niflags = NI_NUMERICHOST | NI_NUMERICSERV;
157
158/* Public. */
159
160/*%
161 *	looks up "ina" in _res.ns_addr_list[]
162 *
163 * returns:
164 *\li	0  : not found
165 *\li	>0 : found
166 *
167 * author:
168 *\li	paul vixie, 29may94
169 */
170int
171res_ourserver_p(const res_state statp, const struct sockaddr *sa) {
172	const struct sockaddr_in *inp, *srv;
173	const struct sockaddr_in6 *in6p, *srv6;
174	int ns;
175
176	switch (sa->sa_family) {
177	case AF_INET:
178		inp = (const struct sockaddr_in *)sa;
179		for (ns = 0;  ns < statp->nscount;  ns++) {
180			srv = (struct sockaddr_in *)get_nsaddr(statp, ns);
181			if (srv->sin_family == inp->sin_family &&
182			    srv->sin_port == inp->sin_port &&
183			    (srv->sin_addr.s_addr == INADDR_ANY ||
184			     srv->sin_addr.s_addr == inp->sin_addr.s_addr))
185				return (1);
186		}
187		break;
188	case AF_INET6:
189		if (EXT(statp).ext == NULL)
190			break;
191		in6p = (const struct sockaddr_in6 *)sa;
192		for (ns = 0;  ns < statp->nscount;  ns++) {
193			srv6 = (struct sockaddr_in6 *)get_nsaddr(statp, ns);
194			if (srv6->sin6_family == in6p->sin6_family &&
195			    srv6->sin6_port == in6p->sin6_port &&
196#ifdef HAVE_SIN6_SCOPE_ID
197			    (srv6->sin6_scope_id == 0 ||
198			     srv6->sin6_scope_id == in6p->sin6_scope_id) &&
199#endif
200			    (IN6_IS_ADDR_UNSPECIFIED(&srv6->sin6_addr) ||
201			     IN6_ARE_ADDR_EQUAL(&srv6->sin6_addr, &in6p->sin6_addr)))
202				return (1);
203		}
204		break;
205	default:
206		break;
207	}
208	return (0);
209}
210
211/*%
212 *	look for (name,type,class) in the query section of packet (buf,eom)
213 *
214 * requires:
215 *\li	buf + HFIXEDSZ <= eom
216 *
217 * returns:
218 *\li	-1 : format error
219 *\li	0  : not found
220 *\li	>0 : found
221 *
222 * author:
223 *\li	paul vixie, 29may94
224 */
225int
226res_nameinquery(const char *name, int type, int class,
227		const u_char *buf, const u_char *eom)
228{
229	const u_char *cp = buf + HFIXEDSZ;
230	int qdcount = ntohs(((const HEADER*)buf)->qdcount);
231
232	while (qdcount-- > 0) {
233		char tname[MAXDNAME+1];
234		int n, ttype, tclass;
235
236		n = dn_expand(buf, eom, cp, tname, sizeof tname);
237		if (n < 0)
238			return (-1);
239		cp += n;
240		if (cp + 2 * INT16SZ > eom)
241			return (-1);
242		ttype = ns_get16(cp); cp += INT16SZ;
243		tclass = ns_get16(cp); cp += INT16SZ;
244		if (ttype == type && tclass == class &&
245		    ns_samename(tname, name) == 1)
246			return (1);
247	}
248	return (0);
249}
250
251/*%
252 *	is there a 1:1 mapping of (name,type,class)
253 *	in (buf1,eom1) and (buf2,eom2)?
254 *
255 * returns:
256 *\li	-1 : format error
257 *\li	0  : not a 1:1 mapping
258 *\li	>0 : is a 1:1 mapping
259 *
260 * author:
261 *\li	paul vixie, 29may94
262 */
263int
264res_queriesmatch(const u_char *buf1, const u_char *eom1,
265		 const u_char *buf2, const u_char *eom2)
266{
267	const u_char *cp = buf1 + HFIXEDSZ;
268	int qdcount = ntohs(((const HEADER*)buf1)->qdcount);
269
270	if (buf1 + HFIXEDSZ > eom1 || buf2 + HFIXEDSZ > eom2)
271		return (-1);
272
273	/*
274	 * Only header section present in replies to
275	 * dynamic update packets.
276	 */
277	if ((((const HEADER *)buf1)->opcode == ns_o_update) &&
278	    (((const HEADER *)buf2)->opcode == ns_o_update))
279		return (1);
280
281	if (qdcount != ntohs(((const HEADER*)buf2)->qdcount))
282		return (0);
283	while (qdcount-- > 0) {
284		char tname[MAXDNAME+1];
285		int n, ttype, tclass;
286
287		n = dn_expand(buf1, eom1, cp, tname, sizeof tname);
288		if (n < 0)
289			return (-1);
290		cp += n;
291		if (cp + 2 * INT16SZ > eom1)
292			return (-1);
293		ttype = ns_get16(cp);	cp += INT16SZ;
294		tclass = ns_get16(cp); cp += INT16SZ;
295		if (!res_nameinquery(tname, ttype, tclass, buf2, eom2))
296			return (0);
297	}
298	return (1);
299}
300
301int
302res_nsend(res_state statp,
303	  const u_char *buf, int buflen, u_char *ans, int anssiz)
304{
305	int gotsomewhere, terrno, tries, v_circuit, resplen, ns, n;
306#ifdef USE_KQUEUE
307	int kq;
308#endif
309	char abuf[NI_MAXHOST];
310
311	/* No name servers or res_init() failure */
312	if (statp->nscount == 0 || EXT(statp).ext == NULL) {
313		errno = ESRCH;
314		return (-1);
315	}
316	if (anssiz < HFIXEDSZ) {
317		errno = EINVAL;
318		return (-1);
319	}
320	DprintQ((statp->options & RES_DEBUG) || (statp->pfcode & RES_PRF_QUERY),
321		(stdout, ";; res_send()\n"), buf, buflen);
322	v_circuit = (statp->options & RES_USEVC) || buflen > PACKETSZ;
323	gotsomewhere = 0;
324	terrno = ETIMEDOUT;
325
326#ifdef USE_KQUEUE
327	if ((kq = kqueue()) < 0) {
328		Perror(statp, stderr, "kqueue", errno);
329		return (-1);
330	}
331#endif
332
333	/*
334	 * If the ns_addr_list in the resolver context has changed, then
335	 * invalidate our cached copy and the associated timing data.
336	 */
337	if (EXT(statp).nscount != 0) {
338		int needclose = 0;
339		struct sockaddr_storage peer;
340		ISC_SOCKLEN_T peerlen;
341
342		if (EXT(statp).nscount != statp->nscount)
343			needclose++;
344		else
345			for (ns = 0; ns < statp->nscount; ns++) {
346				if (statp->nsaddr_list[ns].sin_family &&
347				    !sock_eq((struct sockaddr *)&statp->nsaddr_list[ns],
348					     (struct sockaddr *)&EXT(statp).ext->nsaddrs[ns])) {
349					needclose++;
350					break;
351				}
352
353				if (EXT(statp).nssocks[ns] == -1)
354					continue;
355				peerlen = sizeof(peer);
356				if (_getpeername(EXT(statp).nssocks[ns],
357				    (struct sockaddr *)&peer, &peerlen) < 0) {
358					needclose++;
359					break;
360				}
361				if (!sock_eq((struct sockaddr *)&peer,
362				    get_nsaddr(statp, ns))) {
363					needclose++;
364					break;
365				}
366			}
367		if (needclose) {
368			res_nclose(statp);
369			EXT(statp).nscount = 0;
370		}
371	}
372
373	/*
374	 * Maybe initialize our private copy of the ns_addr_list.
375	 */
376	if (EXT(statp).nscount == 0) {
377		for (ns = 0; ns < statp->nscount; ns++) {
378			EXT(statp).nstimes[ns] = RES_MAXTIME;
379			EXT(statp).nssocks[ns] = -1;
380			if (!statp->nsaddr_list[ns].sin_family)
381				continue;
382			EXT(statp).ext->nsaddrs[ns].sin =
383				 statp->nsaddr_list[ns];
384		}
385		EXT(statp).nscount = statp->nscount;
386	}
387
388	/*
389	 * Some resolvers want to even out the load on their nameservers.
390	 * Note that RES_BLAST overrides RES_ROTATE.
391	 */
392	if ((statp->options & RES_ROTATE) != 0U &&
393	    (statp->options & RES_BLAST) == 0U) {
394		union res_sockaddr_union inu;
395		struct sockaddr_in ina;
396		int lastns = statp->nscount - 1;
397		int fd;
398		u_int16_t nstime;
399
400		if (EXT(statp).ext != NULL)
401			inu = EXT(statp).ext->nsaddrs[0];
402		ina = statp->nsaddr_list[0];
403		fd = EXT(statp).nssocks[0];
404		nstime = EXT(statp).nstimes[0];
405		for (ns = 0; ns < lastns; ns++) {
406			if (EXT(statp).ext != NULL)
407				EXT(statp).ext->nsaddrs[ns] =
408					EXT(statp).ext->nsaddrs[ns + 1];
409			statp->nsaddr_list[ns] = statp->nsaddr_list[ns + 1];
410			EXT(statp).nssocks[ns] = EXT(statp).nssocks[ns + 1];
411			EXT(statp).nstimes[ns] = EXT(statp).nstimes[ns + 1];
412		}
413		if (EXT(statp).ext != NULL)
414			EXT(statp).ext->nsaddrs[lastns] = inu;
415		statp->nsaddr_list[lastns] = ina;
416		EXT(statp).nssocks[lastns] = fd;
417		EXT(statp).nstimes[lastns] = nstime;
418	}
419
420	/*
421	 * Send request, RETRY times, or until successful.
422	 */
423	for (tries = 0; tries < statp->retry; tries++) {
424	    for (ns = 0; ns < statp->nscount; ns++) {
425		struct sockaddr *nsap;
426		int nsaplen;
427		nsap = get_nsaddr(statp, ns);
428		nsaplen = get_salen(nsap);
429		statp->_flags &= ~RES_F_LASTMASK;
430		statp->_flags |= (ns << RES_F_LASTSHIFT);
431 same_ns:
432		if (statp->qhook) {
433			int done = 0, loops = 0;
434
435			do {
436				res_sendhookact act;
437
438				act = (*statp->qhook)(&nsap, &buf, &buflen,
439						      ans, anssiz, &resplen);
440				switch (act) {
441				case res_goahead:
442					done = 1;
443					break;
444				case res_nextns:
445					res_nclose(statp);
446					goto next_ns;
447				case res_done:
448#ifdef USE_KQUEUE
449					_close(kq);
450#endif
451					return (resplen);
452				case res_modified:
453					/* give the hook another try */
454					if (++loops < 42) /*doug adams*/
455						break;
456					/*FALLTHROUGH*/
457				case res_error:
458					/*FALLTHROUGH*/
459				default:
460					goto fail;
461				}
462			} while (!done);
463		}
464
465		Dprint(((statp->options & RES_DEBUG) &&
466			getnameinfo(nsap, nsaplen, abuf, sizeof(abuf),
467				    NULL, 0, niflags) == 0),
468		       (stdout, ";; Querying server (# %d) address = %s\n",
469			ns + 1, abuf));
470
471
472		if (v_circuit) {
473			/* Use VC; at most one attempt per server. */
474			tries = statp->retry;
475			n = send_vc(statp, buf, buflen, ans, anssiz, &terrno,
476				    ns);
477			if (n < 0)
478				goto fail;
479			if (n == 0)
480				goto next_ns;
481			resplen = n;
482		} else {
483			/* Use datagrams. */
484			n = send_dg(statp,
485#ifdef USE_KQUEUE
486				    kq,
487#endif
488				    buf, buflen, ans, anssiz, &terrno,
489				    ns, tries, &v_circuit, &gotsomewhere);
490			if (n < 0)
491				goto fail;
492			if (n == 0)
493				goto next_ns;
494			if (v_circuit)
495				goto same_ns;
496			resplen = n;
497		}
498
499		Dprint((statp->options & RES_DEBUG) ||
500		       ((statp->pfcode & RES_PRF_REPLY) &&
501			(statp->pfcode & RES_PRF_HEAD1)),
502		       (stdout, ";; got answer:\n"));
503
504		DprintQ((statp->options & RES_DEBUG) ||
505			(statp->pfcode & RES_PRF_REPLY),
506			(stdout, "%s", ""),
507			ans, (resplen > anssiz) ? anssiz : resplen);
508
509		/*
510		 * If we have temporarily opened a virtual circuit,
511		 * or if we haven't been asked to keep a socket open,
512		 * close the socket.
513		 */
514		if ((v_circuit && (statp->options & RES_USEVC) == 0U) ||
515		    (statp->options & RES_STAYOPEN) == 0U) {
516			res_nclose(statp);
517		}
518		if (statp->rhook) {
519			int done = 0, loops = 0;
520
521			do {
522				res_sendhookact act;
523
524				act = (*statp->rhook)(nsap, buf, buflen,
525						      ans, anssiz, &resplen);
526				switch (act) {
527				case res_goahead:
528				case res_done:
529					done = 1;
530					break;
531				case res_nextns:
532					res_nclose(statp);
533					goto next_ns;
534				case res_modified:
535					/* give the hook another try */
536					if (++loops < 42) /*doug adams*/
537						break;
538					/*FALLTHROUGH*/
539				case res_error:
540					/*FALLTHROUGH*/
541				default:
542					goto fail;
543				}
544			} while (!done);
545
546		}
547#ifdef USE_KQUEUE
548		_close(kq);
549#endif
550		return (resplen);
551 next_ns: ;
552	   } /*foreach ns*/
553	} /*foreach retry*/
554	res_nclose(statp);
555#ifdef USE_KQUEUE
556	_close(kq);
557#endif
558	if (!v_circuit) {
559		if (!gotsomewhere)
560			errno = ECONNREFUSED;	/*%< no nameservers found */
561		else
562			errno = ETIMEDOUT;	/*%< no answer obtained */
563	} else
564		errno = terrno;
565	return (-1);
566 fail:
567	res_nclose(statp);
568#ifdef USE_KQUEUE
569	_close(kq);
570#endif
571	return (-1);
572}
573
574/* Private */
575
576static int
577get_salen(sa)
578	const struct sockaddr *sa;
579{
580
581#ifdef HAVE_SA_LEN
582	/* There are people do not set sa_len.  Be forgiving to them. */
583	if (sa->sa_len)
584		return (sa->sa_len);
585#endif
586
587	if (sa->sa_family == AF_INET)
588		return (sizeof(struct sockaddr_in));
589	else if (sa->sa_family == AF_INET6)
590		return (sizeof(struct sockaddr_in6));
591	else
592		return (0);	/*%< unknown, die on connect */
593}
594
595/*%
596 * pick appropriate nsaddr_list for use.  see res_init() for initialization.
597 */
598static struct sockaddr *
599get_nsaddr(statp, n)
600	res_state statp;
601	size_t n;
602{
603
604	if (!statp->nsaddr_list[n].sin_family && EXT(statp).ext) {
605		/*
606		 * - EXT(statp).ext->nsaddrs[n] holds an address that is larger
607		 *   than struct sockaddr, and
608		 * - user code did not update statp->nsaddr_list[n].
609		 */
610		return (struct sockaddr *)(void *)&EXT(statp).ext->nsaddrs[n];
611	} else {
612		/*
613		 * - user code updated statp->nsaddr_list[n], or
614		 * - statp->nsaddr_list[n] has the same content as
615		 *   EXT(statp).ext->nsaddrs[n].
616		 */
617		return (struct sockaddr *)(void *)&statp->nsaddr_list[n];
618	}
619}
620
621static int
622send_vc(res_state statp,
623	const u_char *buf, int buflen, u_char *ans, int anssiz,
624	int *terrno, int ns)
625{
626	const HEADER *hp = (const HEADER *) buf;
627	HEADER *anhp = (HEADER *) ans;
628	struct sockaddr *nsap;
629	int nsaplen;
630	int truncating, connreset, resplen, n;
631	struct iovec iov[2];
632	u_short len;
633	u_char *cp;
634	void *tmp;
635#ifdef SO_NOSIGPIPE
636	int on = 1;
637#endif
638
639	nsap = get_nsaddr(statp, ns);
640	nsaplen = get_salen(nsap);
641
642	connreset = 0;
643 same_ns:
644	truncating = 0;
645
646	/* Are we still talking to whom we want to talk to? */
647	if (statp->_vcsock >= 0 && (statp->_flags & RES_F_VC) != 0) {
648		struct sockaddr_storage peer;
649		ISC_SOCKLEN_T size = sizeof peer;
650
651		if (_getpeername(statp->_vcsock,
652				(struct sockaddr *)&peer, &size) < 0 ||
653		    !sock_eq((struct sockaddr *)&peer, nsap)) {
654			res_nclose(statp);
655			statp->_flags &= ~RES_F_VC;
656		}
657	}
658
659	if (statp->_vcsock < 0 || (statp->_flags & RES_F_VC) == 0) {
660		if (statp->_vcsock >= 0)
661			res_nclose(statp);
662
663		statp->_vcsock = _socket(nsap->sa_family, SOCK_STREAM |
664		    SOCK_CLOEXEC, 0);
665#if !defined(USE_POLL) && !defined(USE_KQUEUE)
666		if (statp->_vcsock > highestFD) {
667			res_nclose(statp);
668			errno = ENOTSOCK;
669		}
670#endif
671		if (statp->_vcsock < 0) {
672			switch (errno) {
673			case EPROTONOSUPPORT:
674#ifdef EPFNOSUPPORT
675			case EPFNOSUPPORT:
676#endif
677			case EAFNOSUPPORT:
678				Perror(statp, stderr, "socket(vc)", errno);
679				return (0);
680			default:
681				*terrno = errno;
682				Perror(statp, stderr, "socket(vc)", errno);
683				return (-1);
684			}
685		}
686#ifdef SO_NOSIGPIPE
687		/*
688		 * Disable generation of SIGPIPE when writing to a closed
689		 * socket.  Write should return -1 and set errno to EPIPE
690		 * instead.
691		 *
692		 * Push on even if setsockopt(SO_NOSIGPIPE) fails.
693		 */
694		(void)_setsockopt(statp->_vcsock, SOL_SOCKET, SO_NOSIGPIPE, &on,
695				 sizeof(on));
696#endif
697		errno = 0;
698		if (_connect(statp->_vcsock, nsap, nsaplen) < 0) {
699			*terrno = errno;
700			Aerror(statp, stderr, "connect/vc", errno, nsap,
701			    nsaplen);
702			res_nclose(statp);
703			return (0);
704		}
705		statp->_flags |= RES_F_VC;
706	}
707
708	/*
709	 * Send length & message
710	 */
711	ns_put16((u_short)buflen, (u_char*)&len);
712	iov[0] = evConsIovec(&len, INT16SZ);
713	DE_CONST(buf, tmp);
714	iov[1] = evConsIovec(tmp, buflen);
715	if (_writev(statp->_vcsock, iov, 2) != (INT16SZ + buflen)) {
716		*terrno = errno;
717		Perror(statp, stderr, "write failed", errno);
718		res_nclose(statp);
719		return (0);
720	}
721	/*
722	 * Receive length & response
723	 */
724 read_len:
725	cp = ans;
726	len = INT16SZ;
727	while ((n = _read(statp->_vcsock, (char *)cp, (int)len)) > 0) {
728		cp += n;
729		if ((len -= n) == 0)
730			break;
731	}
732	if (n <= 0) {
733		*terrno = errno;
734		Perror(statp, stderr, "read failed", errno);
735		res_nclose(statp);
736		/*
737		 * A long running process might get its TCP
738		 * connection reset if the remote server was
739		 * restarted.  Requery the server instead of
740		 * trying a new one.  When there is only one
741		 * server, this means that a query might work
742		 * instead of failing.  We only allow one reset
743		 * per query to prevent looping.
744		 */
745		if (*terrno == ECONNRESET && !connreset) {
746			connreset = 1;
747			res_nclose(statp);
748			goto same_ns;
749		}
750		res_nclose(statp);
751		return (0);
752	}
753	resplen = ns_get16(ans);
754	if (resplen > anssiz) {
755		Dprint(statp->options & RES_DEBUG,
756		       (stdout, ";; response truncated\n")
757		       );
758		truncating = 1;
759		len = anssiz;
760	} else
761		len = resplen;
762	if (len < HFIXEDSZ) {
763		/*
764		 * Undersized message.
765		 */
766		Dprint(statp->options & RES_DEBUG,
767		       (stdout, ";; undersized: %d\n", len));
768		*terrno = EMSGSIZE;
769		res_nclose(statp);
770		return (0);
771	}
772	cp = ans;
773	while (len != 0 &&
774	    (n = _read(statp->_vcsock, (char *)cp, (int)len)) > 0) {
775		cp += n;
776		len -= n;
777	}
778	if (n <= 0) {
779		*terrno = errno;
780		Perror(statp, stderr, "read(vc)", errno);
781		res_nclose(statp);
782		return (0);
783	}
784	if (truncating) {
785		/*
786		 * Flush rest of answer so connection stays in synch.
787		 */
788		anhp->tc = 1;
789		len = resplen - anssiz;
790		while (len != 0) {
791			char junk[PACKETSZ];
792
793			n = _read(statp->_vcsock, junk,
794			    (len > sizeof junk) ? sizeof junk : len);
795			if (n > 0)
796				len -= n;
797			else
798				break;
799		}
800	}
801	/*
802	 * If the calling applicating has bailed out of
803	 * a previous call and failed to arrange to have
804	 * the circuit closed or the server has got
805	 * itself confused, then drop the packet and
806	 * wait for the correct one.
807	 */
808	if (hp->id != anhp->id) {
809		DprintQ((statp->options & RES_DEBUG) ||
810			(statp->pfcode & RES_PRF_REPLY),
811			(stdout, ";; old answer (unexpected):\n"),
812			ans, (resplen > anssiz) ? anssiz: resplen);
813		goto read_len;
814	}
815
816	/*
817	 * All is well, or the error is fatal.  Signal that the
818	 * next nameserver ought not be tried.
819	 */
820	return (resplen);
821}
822
823static int
824send_dg(res_state statp,
825#ifdef USE_KQUEUE
826	int kq,
827#endif
828	const u_char *buf, int buflen, u_char *ans,
829	int anssiz, int *terrno, int ns, int tries, int *v_circuit,
830	int *gotsomewhere)
831{
832	const HEADER *hp = (const HEADER *) buf;
833	HEADER *anhp = (HEADER *) ans;
834	const struct sockaddr *nsap;
835	int nsaplen;
836	struct timespec now, timeout, finish;
837	struct sockaddr_storage from;
838	ISC_SOCKLEN_T fromlen;
839	int resplen, seconds, n, s;
840#ifdef USE_KQUEUE
841	struct kevent kv;
842#else
843#ifdef USE_POLL
844	int     polltimeout;
845	struct pollfd   pollfd;
846#else
847	fd_set dsmask;
848#endif
849#endif
850
851	nsap = get_nsaddr(statp, ns);
852	nsaplen = get_salen(nsap);
853	if (EXT(statp).nssocks[ns] == -1) {
854		EXT(statp).nssocks[ns] = _socket(nsap->sa_family,
855		    SOCK_DGRAM | SOCK_CLOEXEC, 0);
856#if !defined(USE_POLL) && !defined(USE_KQUEUE)
857		if (EXT(statp).nssocks[ns] > highestFD) {
858			res_nclose(statp);
859			errno = ENOTSOCK;
860		}
861#endif
862		if (EXT(statp).nssocks[ns] < 0) {
863			switch (errno) {
864			case EPROTONOSUPPORT:
865#ifdef EPFNOSUPPORT
866			case EPFNOSUPPORT:
867#endif
868			case EAFNOSUPPORT:
869				Perror(statp, stderr, "socket(dg)", errno);
870				return (0);
871			default:
872				*terrno = errno;
873				Perror(statp, stderr, "socket(dg)", errno);
874				return (-1);
875			}
876		}
877#ifndef CANNOT_CONNECT_DGRAM
878		/*
879		 * On a 4.3BSD+ machine (client and server,
880		 * actually), sending to a nameserver datagram
881		 * port with no nameserver will cause an
882		 * ICMP port unreachable message to be returned.
883		 * If our datagram socket is "connected" to the
884		 * server, we get an ECONNREFUSED error on the next
885		 * socket operation, and select returns if the
886		 * error message is received.  We can thus detect
887		 * the absence of a nameserver without timing out.
888		 *
889		 * When the option "insecure1" is specified, we'd
890		 * rather expect to see responses from an "unknown"
891		 * address.  In order to let the kernel accept such
892		 * responses, do not connect the socket here.
893		 * XXX: or do we need an explicit option to disable
894		 * connecting?
895		 */
896		if (!(statp->options & RES_INSECURE1) &&
897		    _connect(EXT(statp).nssocks[ns], nsap, nsaplen) < 0) {
898			Aerror(statp, stderr, "connect(dg)", errno, nsap,
899			    nsaplen);
900			res_nclose(statp);
901			return (0);
902		}
903#endif /* !CANNOT_CONNECT_DGRAM */
904		Dprint(statp->options & RES_DEBUG,
905		       (stdout, ";; new DG socket\n"))
906	}
907	s = EXT(statp).nssocks[ns];
908#ifndef CANNOT_CONNECT_DGRAM
909	if (statp->options & RES_INSECURE1) {
910		if (_sendto(s,
911		    (const char*)buf, buflen, 0, nsap, nsaplen) != buflen) {
912			Aerror(statp, stderr, "sendto", errno, nsap, nsaplen);
913			res_nclose(statp);
914			return (0);
915		}
916	} else if (send(s, (const char*)buf, buflen, 0) != buflen) {
917		Perror(statp, stderr, "send", errno);
918		res_nclose(statp);
919		return (0);
920	}
921#else /* !CANNOT_CONNECT_DGRAM */
922	if (_sendto(s, (const char*)buf, buflen, 0, nsap, nsaplen) != buflen)
923	{
924		Aerror(statp, stderr, "sendto", errno, nsap, nsaplen);
925		res_nclose(statp);
926		return (0);
927	}
928#endif /* !CANNOT_CONNECT_DGRAM */
929
930	/*
931	 * Wait for reply.
932	 */
933	seconds = (statp->retrans << tries);
934	if (ns > 0)
935		seconds /= statp->nscount;
936	if (seconds <= 0)
937		seconds = 1;
938	now = evNowTime();
939	timeout = evConsTime(seconds, 0);
940	finish = evAddTime(now, timeout);
941	goto nonow;
942 wait:
943	now = evNowTime();
944 nonow:
945#ifndef USE_POLL
946	if (evCmpTime(finish, now) > 0)
947		timeout = evSubTime(finish, now);
948	else
949		timeout = evConsTime(0, 0);
950#ifdef USE_KQUEUE
951	EV_SET(&kv, s, EVFILT_READ, EV_ADD | EV_ONESHOT, 0, 0, 0);
952	n = _kevent(kq, &kv, 1, &kv, 1, &timeout);
953#else
954	FD_ZERO(&dsmask);
955	FD_SET(s, &dsmask);
956	n = pselect(s + 1, &dsmask, NULL, NULL, &timeout, NULL);
957#endif
958#else
959	timeout = evSubTime(finish, now);
960	if (timeout.tv_sec < 0)
961		timeout = evConsTime(0, 0);
962	polltimeout = 1000*timeout.tv_sec +
963		timeout.tv_nsec/1000000;
964	pollfd.fd = s;
965	pollfd.events = POLLRDNORM;
966	n = poll(&pollfd, 1, polltimeout);
967#endif /* USE_POLL */
968
969	if (n == 0) {
970		Dprint(statp->options & RES_DEBUG, (stdout, ";; timeout\n"));
971		*gotsomewhere = 1;
972		return (0);
973	}
974	if (n < 0) {
975		if (errno == EINTR)
976			goto wait;
977#ifdef USE_KQUEUE
978		Perror(statp, stderr, "kevent", errno);
979#else
980#ifndef USE_POLL
981		Perror(statp, stderr, "select", errno);
982#else
983		Perror(statp, stderr, "poll", errno);
984#endif /* USE_POLL */
985#endif
986		res_nclose(statp);
987		return (0);
988	}
989#ifdef USE_KQUEUE
990	if (kv.ident != s)
991		goto wait;
992#endif
993	errno = 0;
994	fromlen = sizeof(from);
995	resplen = _recvfrom(s, (char*)ans, anssiz,0,
996			   (struct sockaddr *)&from, &fromlen);
997	if (resplen <= 0) {
998		Perror(statp, stderr, "recvfrom", errno);
999		res_nclose(statp);
1000		return (0);
1001	}
1002	*gotsomewhere = 1;
1003	if (resplen < HFIXEDSZ) {
1004		/*
1005		 * Undersized message.
1006		 */
1007		Dprint(statp->options & RES_DEBUG,
1008		       (stdout, ";; undersized: %d\n",
1009			resplen));
1010		*terrno = EMSGSIZE;
1011		res_nclose(statp);
1012		return (0);
1013	}
1014	if (hp->id != anhp->id) {
1015		/*
1016		 * response from old query, ignore it.
1017		 * XXX - potential security hazard could
1018		 *	 be detected here.
1019		 */
1020		DprintQ((statp->options & RES_DEBUG) ||
1021			(statp->pfcode & RES_PRF_REPLY),
1022			(stdout, ";; old answer:\n"),
1023			ans, (resplen > anssiz) ? anssiz : resplen);
1024		goto wait;
1025	}
1026	if (!(statp->options & RES_INSECURE1) &&
1027	    !res_ourserver_p(statp, (struct sockaddr *)&from)) {
1028		/*
1029		 * response from wrong server? ignore it.
1030		 * XXX - potential security hazard could
1031		 *	 be detected here.
1032		 */
1033		DprintQ((statp->options & RES_DEBUG) ||
1034			(statp->pfcode & RES_PRF_REPLY),
1035			(stdout, ";; not our server:\n"),
1036			ans, (resplen > anssiz) ? anssiz : resplen);
1037		goto wait;
1038	}
1039#ifdef RES_USE_EDNS0
1040	if (anhp->rcode == FORMERR && (statp->options & RES_USE_EDNS0) != 0U) {
1041		/*
1042		 * Do not retry if the server do not understand EDNS0.
1043		 * The case has to be captured here, as FORMERR packet do not
1044		 * carry query section, hence res_queriesmatch() returns 0.
1045		 */
1046		DprintQ(statp->options & RES_DEBUG,
1047			(stdout, "server rejected query with EDNS0:\n"),
1048			ans, (resplen > anssiz) ? anssiz : resplen);
1049		/* record the error */
1050		statp->_flags |= RES_F_EDNS0ERR;
1051		res_nclose(statp);
1052		return (0);
1053	}
1054#endif
1055	if (!(statp->options & RES_INSECURE2) &&
1056	    !res_queriesmatch(buf, buf + buflen,
1057			      ans, ans + anssiz)) {
1058		/*
1059		 * response contains wrong query? ignore it.
1060		 * XXX - potential security hazard could
1061		 *	 be detected here.
1062		 */
1063		DprintQ((statp->options & RES_DEBUG) ||
1064			(statp->pfcode & RES_PRF_REPLY),
1065			(stdout, ";; wrong query name:\n"),
1066			ans, (resplen > anssiz) ? anssiz : resplen);
1067		goto wait;
1068	}
1069	if (anhp->rcode == SERVFAIL ||
1070	    anhp->rcode == NOTIMP ||
1071	    anhp->rcode == REFUSED) {
1072		DprintQ(statp->options & RES_DEBUG,
1073			(stdout, "server rejected query:\n"),
1074			ans, (resplen > anssiz) ? anssiz : resplen);
1075		res_nclose(statp);
1076		/* don't retry if called from dig */
1077		if (!statp->pfcode)
1078			return (0);
1079	}
1080	if (!(statp->options & RES_IGNTC) && anhp->tc) {
1081		/*
1082		 * To get the rest of answer,
1083		 * use TCP with same server.
1084		 */
1085		Dprint(statp->options & RES_DEBUG,
1086		       (stdout, ";; truncated answer\n"));
1087		*v_circuit = 1;
1088		res_nclose(statp);
1089		return (1);
1090	}
1091	/*
1092	 * All is well, or the error is fatal.  Signal that the
1093	 * next nameserver ought not be tried.
1094	 */
1095	return (resplen);
1096}
1097
1098static void
1099Aerror(const res_state statp, FILE *file, const char *string, int error,
1100       const struct sockaddr *address, int alen)
1101{
1102	int save = errno;
1103	char hbuf[NI_MAXHOST];
1104	char sbuf[NI_MAXSERV];
1105
1106	if ((statp->options & RES_DEBUG) != 0U) {
1107		if (getnameinfo(address, alen, hbuf, sizeof(hbuf),
1108		    sbuf, sizeof(sbuf), niflags)) {
1109			strncpy(hbuf, "?", sizeof(hbuf) - 1);
1110			hbuf[sizeof(hbuf) - 1] = '\0';
1111			strncpy(sbuf, "?", sizeof(sbuf) - 1);
1112			sbuf[sizeof(sbuf) - 1] = '\0';
1113		}
1114		fprintf(file, "res_send: %s ([%s].%s): %s\n",
1115			string, hbuf, sbuf, strerror(error));
1116	}
1117	errno = save;
1118}
1119
1120static void
1121Perror(const res_state statp, FILE *file, const char *string, int error) {
1122	int save = errno;
1123
1124	if ((statp->options & RES_DEBUG) != 0U)
1125		fprintf(file, "res_send: %s: %s\n",
1126			string, strerror(error));
1127	errno = save;
1128}
1129
1130static int
1131sock_eq(struct sockaddr *a, struct sockaddr *b) {
1132	struct sockaddr_in *a4, *b4;
1133	struct sockaddr_in6 *a6, *b6;
1134
1135	if (a->sa_family != b->sa_family)
1136		return 0;
1137	switch (a->sa_family) {
1138	case AF_INET:
1139		a4 = (struct sockaddr_in *)a;
1140		b4 = (struct sockaddr_in *)b;
1141		return a4->sin_port == b4->sin_port &&
1142		    a4->sin_addr.s_addr == b4->sin_addr.s_addr;
1143	case AF_INET6:
1144		a6 = (struct sockaddr_in6 *)a;
1145		b6 = (struct sockaddr_in6 *)b;
1146		return a6->sin6_port == b6->sin6_port &&
1147#ifdef HAVE_SIN6_SCOPE_ID
1148		    a6->sin6_scope_id == b6->sin6_scope_id &&
1149#endif
1150		    IN6_ARE_ADDR_EQUAL(&a6->sin6_addr, &b6->sin6_addr);
1151	default:
1152		return 0;
1153	}
1154}
1155
1156#if defined(NEED_PSELECT) && !defined(USE_POLL) && !defined(USE_KQUEUE)
1157/* XXX needs to move to the porting library. */
1158static int
1159pselect(int nfds, void *rfds, void *wfds, void *efds,
1160	struct timespec *tsp, const sigset_t *sigmask)
1161{
1162	struct timeval tv, *tvp;
1163	sigset_t sigs;
1164	int n;
1165
1166	if (tsp) {
1167		tvp = &tv;
1168		tv = evTimeVal(*tsp);
1169	} else
1170		tvp = NULL;
1171	if (sigmask)
1172		sigprocmask(SIG_SETMASK, sigmask, &sigs);
1173	n = select(nfds, rfds, wfds, efds, tvp);
1174	if (sigmask)
1175		sigprocmask(SIG_SETMASK, &sigs, NULL);
1176	if (tsp)
1177		*tsp = evTimeSpec(tv);
1178	return (n);
1179}
1180#endif
1181