res_send.c revision 186090
1/*
2 * Copyright (c) 1985, 1989, 1993
3 *    The Regents of the University of California.  All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 *    may be used to endorse or promote products derived from this software
15 *    without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 */
29
30/*
31 * Portions Copyright (c) 1993 by Digital Equipment Corporation.
32 *
33 * Permission to use, copy, modify, and distribute this software for any
34 * purpose with or without fee is hereby granted, provided that the above
35 * copyright notice and this permission notice appear in all copies, and that
36 * the name of Digital Equipment Corporation not be used in advertising or
37 * publicity pertaining to distribution of the document or software without
38 * specific, written prior permission.
39 *
40 * THE SOFTWARE IS PROVIDED "AS IS" AND DIGITAL EQUIPMENT CORP. DISCLAIMS ALL
41 * WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES
42 * OF MERCHANTABILITY AND FITNESS.   IN NO EVENT SHALL DIGITAL EQUIPMENT
43 * CORPORATION BE LIABLE FOR ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL
44 * DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR
45 * PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
46 * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS
47 * SOFTWARE.
48 */
49
50/*
51 * Copyright (c) 2005 by Internet Systems Consortium, Inc. ("ISC")
52 * Portions Copyright (c) 1996-1999 by Internet Software Consortium.
53 *
54 * Permission to use, copy, modify, and distribute this software for any
55 * purpose with or without fee is hereby granted, provided that the above
56 * copyright notice and this permission notice appear in all copies.
57 *
58 * THE SOFTWARE IS PROVIDED "AS IS" AND ISC DISCLAIMS ALL WARRANTIES
59 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
60 * MERCHANTABILITY AND FITNESS.  IN NO EVENT SHALL ISC BE LIABLE FOR
61 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
62 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
63 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT
64 * OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS 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 1.9.18.10 2008/01/27 02:06:26 marka Exp $";
70#endif /* LIBC_SCCS and not lint */
71#include <sys/cdefs.h>
72__FBSDID("$FreeBSD: head/lib/libc/resolv/res_send.c 186090 2008-12-14 19:39:53Z 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 (_getsockname(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, 0);
664#if !defined(USE_POLL) && !defined(USE_KQUEUE)
665		if (statp->_vcsock > highestFD) {
666			res_nclose(statp);
667			errno = ENOTSOCK;
668		}
669#endif
670		if (statp->_vcsock < 0) {
671			switch (errno) {
672			case EPROTONOSUPPORT:
673#ifdef EPFNOSUPPORT
674			case EPFNOSUPPORT:
675#endif
676			case EAFNOSUPPORT:
677				Perror(statp, stderr, "socket(vc)", errno);
678				return (0);
679			default:
680				*terrno = errno;
681				Perror(statp, stderr, "socket(vc)", errno);
682				return (-1);
683			}
684		}
685#ifdef SO_NOSIGPIPE
686		/*
687		 * Disable generation of SIGPIPE when writing to a closed
688		 * socket.  Write should return -1 and set errno to EPIPE
689		 * instead.
690		 *
691		 * Push on even if setsockopt(SO_NOSIGPIPE) fails.
692		 */
693		(void)_setsockopt(statp->_vcsock, SOL_SOCKET, SO_NOSIGPIPE, &on,
694			         sizeof(on));
695#endif
696		errno = 0;
697		if (_connect(statp->_vcsock, nsap, nsaplen) < 0) {
698			*terrno = errno;
699			Aerror(statp, stderr, "connect/vc", errno, nsap,
700			    nsaplen);
701			res_nclose(statp);
702			return (0);
703		}
704		statp->_flags |= RES_F_VC;
705	}
706
707	/*
708	 * Send length & message
709	 */
710	ns_put16((u_short)buflen, (u_char*)&len);
711	iov[0] = evConsIovec(&len, INT16SZ);
712	DE_CONST(buf, tmp);
713	iov[1] = evConsIovec(tmp, buflen);
714	if (_writev(statp->_vcsock, iov, 2) != (INT16SZ + buflen)) {
715		*terrno = errno;
716		Perror(statp, stderr, "write failed", errno);
717		res_nclose(statp);
718		return (0);
719	}
720	/*
721	 * Receive length & response
722	 */
723 read_len:
724	cp = ans;
725	len = INT16SZ;
726	while ((n = _read(statp->_vcsock, (char *)cp, (int)len)) > 0) {
727		cp += n;
728		if ((len -= n) == 0)
729			break;
730	}
731	if (n <= 0) {
732		*terrno = errno;
733		Perror(statp, stderr, "read failed", errno);
734		res_nclose(statp);
735		/*
736		 * A long running process might get its TCP
737		 * connection reset if the remote server was
738		 * restarted.  Requery the server instead of
739		 * trying a new one.  When there is only one
740		 * server, this means that a query might work
741		 * instead of failing.  We only allow one reset
742		 * per query to prevent looping.
743		 */
744		if (*terrno == ECONNRESET && !connreset) {
745			connreset = 1;
746			res_nclose(statp);
747			goto same_ns;
748		}
749		res_nclose(statp);
750		return (0);
751	}
752	resplen = ns_get16(ans);
753	if (resplen > anssiz) {
754		Dprint(statp->options & RES_DEBUG,
755		       (stdout, ";; response truncated\n")
756		       );
757		truncating = 1;
758		len = anssiz;
759	} else
760		len = resplen;
761	if (len < HFIXEDSZ) {
762		/*
763		 * Undersized message.
764		 */
765		Dprint(statp->options & RES_DEBUG,
766		       (stdout, ";; undersized: %d\n", len));
767		*terrno = EMSGSIZE;
768		res_nclose(statp);
769		return (0);
770	}
771	cp = ans;
772	while (len != 0 &&
773	    (n = _read(statp->_vcsock, (char *)cp, (int)len)) > 0) {
774		cp += n;
775		len -= n;
776	}
777	if (n <= 0) {
778		*terrno = errno;
779		Perror(statp, stderr, "read(vc)", errno);
780		res_nclose(statp);
781		return (0);
782	}
783	if (truncating) {
784		/*
785		 * Flush rest of answer so connection stays in synch.
786		 */
787		anhp->tc = 1;
788		len = resplen - anssiz;
789		while (len != 0) {
790			char junk[PACKETSZ];
791
792			n = _read(statp->_vcsock, junk,
793			    (len > sizeof junk) ? sizeof junk : len);
794			if (n > 0)
795				len -= n;
796			else
797				break;
798		}
799	}
800	/*
801	 * If the calling applicating has bailed out of
802	 * a previous call and failed to arrange to have
803	 * the circuit closed or the server has got
804	 * itself confused, then drop the packet and
805	 * wait for the correct one.
806	 */
807	if (hp->id != anhp->id) {
808		DprintQ((statp->options & RES_DEBUG) ||
809			(statp->pfcode & RES_PRF_REPLY),
810			(stdout, ";; old answer (unexpected):\n"),
811			ans, (resplen > anssiz) ? anssiz: resplen);
812		goto read_len;
813	}
814
815	/*
816	 * All is well, or the error is fatal.  Signal that the
817	 * next nameserver ought not be tried.
818	 */
819	return (resplen);
820}
821
822static int
823send_dg(res_state statp,
824#ifdef USE_KQUEUE
825	int kq,
826#endif
827	const u_char *buf, int buflen, u_char *ans,
828	int anssiz, int *terrno, int ns, int tries, int *v_circuit,
829	int *gotsomewhere)
830{
831	const HEADER *hp = (const HEADER *) buf;
832	HEADER *anhp = (HEADER *) ans;
833	const struct sockaddr *nsap;
834	int nsaplen;
835	struct timespec now, timeout, finish;
836	struct sockaddr_storage from;
837	ISC_SOCKLEN_T fromlen;
838	int resplen, seconds, n, s;
839#ifdef USE_KQUEUE
840	struct kevent kv;
841#else
842#ifdef USE_POLL
843	int     polltimeout;
844	struct pollfd   pollfd;
845#else
846	fd_set dsmask;
847#endif
848#endif
849
850	nsap = get_nsaddr(statp, ns);
851	nsaplen = get_salen(nsap);
852	if (EXT(statp).nssocks[ns] == -1) {
853		EXT(statp).nssocks[ns] = _socket(nsap->sa_family,
854		    SOCK_DGRAM, 0);
855#if !defined(USE_POLL) && !defined(USE_KQUEUE)
856		if (EXT(statp).nssocks[ns] > highestFD) {
857			res_nclose(statp);
858			errno = ENOTSOCK;
859		}
860#endif
861		if (EXT(statp).nssocks[ns] < 0) {
862			switch (errno) {
863			case EPROTONOSUPPORT:
864#ifdef EPFNOSUPPORT
865			case EPFNOSUPPORT:
866#endif
867			case EAFNOSUPPORT:
868				Perror(statp, stderr, "socket(dg)", errno);
869				return (0);
870			default:
871				*terrno = errno;
872				Perror(statp, stderr, "socket(dg)", errno);
873				return (-1);
874			}
875		}
876#ifndef CANNOT_CONNECT_DGRAM
877		/*
878		 * On a 4.3BSD+ machine (client and server,
879		 * actually), sending to a nameserver datagram
880		 * port with no nameserver will cause an
881		 * ICMP port unreachable message to be returned.
882		 * If our datagram socket is "connected" to the
883		 * server, we get an ECONNREFUSED error on the next
884		 * socket operation, and select returns if the
885		 * error message is received.  We can thus detect
886		 * the absence of a nameserver without timing out.
887		 *
888		 * When the option "insecure1" is specified, we'd
889		 * rather expect to see responses from an "unknown"
890		 * address.  In order to let the kernel accept such
891		 * responses, do not connect the socket here.
892		 * XXX: or do we need an explicit option to disable
893		 * connecting?
894		 */
895		if (!(statp->options & RES_INSECURE1) &&
896		    _connect(EXT(statp).nssocks[ns], nsap, nsaplen) < 0) {
897			Aerror(statp, stderr, "connect(dg)", errno, nsap,
898			    nsaplen);
899			res_nclose(statp);
900			return (0);
901		}
902#endif /* !CANNOT_CONNECT_DGRAM */
903		Dprint(statp->options & RES_DEBUG,
904		       (stdout, ";; new DG socket\n"))
905	}
906	s = EXT(statp).nssocks[ns];
907#ifndef CANNOT_CONNECT_DGRAM
908	if (statp->options & RES_INSECURE1) {
909		if (_sendto(s,
910		    (const char*)buf, buflen, 0, nsap, nsaplen) != buflen) {
911			Aerror(statp, stderr, "sendto", errno, nsap, nsaplen);
912			res_nclose(statp);
913			return (0);
914		}
915	} else if (send(s, (const char*)buf, buflen, 0) != buflen) {
916		Perror(statp, stderr, "send", errno);
917		res_nclose(statp);
918		return (0);
919	}
920#else /* !CANNOT_CONNECT_DGRAM */
921	if (_sendto(s, (const char*)buf, buflen, 0, nsap, nsaplen) != buflen)
922	{
923		Aerror(statp, stderr, "sendto", errno, nsap, nsaplen);
924		res_nclose(statp);
925		return (0);
926	}
927#endif /* !CANNOT_CONNECT_DGRAM */
928
929	/*
930	 * Wait for reply.
931	 */
932	seconds = (statp->retrans << tries);
933	if (ns > 0)
934		seconds /= statp->nscount;
935	if (seconds <= 0)
936		seconds = 1;
937	now = evNowTime();
938	timeout = evConsTime(seconds, 0);
939	finish = evAddTime(now, timeout);
940	goto nonow;
941 wait:
942	now = evNowTime();
943 nonow:
944#ifndef USE_POLL
945	if (evCmpTime(finish, now) > 0)
946		timeout = evSubTime(finish, now);
947	else
948		timeout = evConsTime(0, 0);
949#ifdef USE_KQUEUE
950	EV_SET(&kv, s, EVFILT_READ, EV_ADD | EV_ONESHOT, 0, 0, 0);
951	n = _kevent(kq, &kv, 1, &kv, 1, &timeout);
952#else
953	FD_ZERO(&dsmask);
954	FD_SET(s, &dsmask);
955	n = pselect(s + 1, &dsmask, NULL, NULL, &timeout, NULL);
956#endif
957#else
958	timeout = evSubTime(finish, now);
959	if (timeout.tv_sec < 0)
960		timeout = evConsTime(0, 0);
961	polltimeout = 1000*timeout.tv_sec +
962		timeout.tv_nsec/1000000;
963	pollfd.fd = s;
964	pollfd.events = POLLRDNORM;
965	n = poll(&pollfd, 1, polltimeout);
966#endif /* USE_POLL */
967
968	if (n == 0) {
969		Dprint(statp->options & RES_DEBUG, (stdout, ";; timeout\n"));
970		*gotsomewhere = 1;
971		return (0);
972	}
973	if (n < 0) {
974		if (errno == EINTR)
975			goto wait;
976#ifdef USE_KQUEUE
977		Perror(statp, stderr, "kevent", errno);
978#else
979#ifndef USE_POLL
980		Perror(statp, stderr, "select", errno);
981#else
982		Perror(statp, stderr, "poll", errno);
983#endif /* USE_POLL */
984#endif
985		res_nclose(statp);
986		return (0);
987	}
988#ifdef USE_KQUEUE
989	if (kv.ident != s)
990		goto wait;
991#endif
992	errno = 0;
993	fromlen = sizeof(from);
994	resplen = _recvfrom(s, (char*)ans, anssiz,0,
995			   (struct sockaddr *)&from, &fromlen);
996	if (resplen <= 0) {
997		Perror(statp, stderr, "recvfrom", errno);
998		res_nclose(statp);
999		return (0);
1000	}
1001	*gotsomewhere = 1;
1002	if (resplen < HFIXEDSZ) {
1003		/*
1004		 * Undersized message.
1005		 */
1006		Dprint(statp->options & RES_DEBUG,
1007		       (stdout, ";; undersized: %d\n",
1008			resplen));
1009		*terrno = EMSGSIZE;
1010		res_nclose(statp);
1011		return (0);
1012	}
1013	if (hp->id != anhp->id) {
1014		/*
1015		 * response from old query, ignore it.
1016		 * XXX - potential security hazard could
1017		 *	 be detected here.
1018		 */
1019		DprintQ((statp->options & RES_DEBUG) ||
1020			(statp->pfcode & RES_PRF_REPLY),
1021			(stdout, ";; old answer:\n"),
1022			ans, (resplen > anssiz) ? anssiz : resplen);
1023		goto wait;
1024	}
1025	if (!(statp->options & RES_INSECURE1) &&
1026	    !res_ourserver_p(statp, (struct sockaddr *)&from)) {
1027		/*
1028		 * response from wrong server? ignore it.
1029		 * XXX - potential security hazard could
1030		 *	 be detected here.
1031		 */
1032		DprintQ((statp->options & RES_DEBUG) ||
1033			(statp->pfcode & RES_PRF_REPLY),
1034			(stdout, ";; not our server:\n"),
1035			ans, (resplen > anssiz) ? anssiz : resplen);
1036		goto wait;
1037	}
1038#ifdef RES_USE_EDNS0
1039	if (anhp->rcode == FORMERR && (statp->options & RES_USE_EDNS0) != 0U) {
1040		/*
1041		 * Do not retry if the server do not understand EDNS0.
1042		 * The case has to be captured here, as FORMERR packet do not
1043		 * carry query section, hence res_queriesmatch() returns 0.
1044		 */
1045		DprintQ(statp->options & RES_DEBUG,
1046			(stdout, "server rejected query with EDNS0:\n"),
1047			ans, (resplen > anssiz) ? anssiz : resplen);
1048		/* record the error */
1049		statp->_flags |= RES_F_EDNS0ERR;
1050		res_nclose(statp);
1051		return (0);
1052	}
1053#endif
1054	if (!(statp->options & RES_INSECURE2) &&
1055	    !res_queriesmatch(buf, buf + buflen,
1056			      ans, ans + anssiz)) {
1057		/*
1058		 * response contains wrong query? ignore it.
1059		 * XXX - potential security hazard could
1060		 *	 be detected here.
1061		 */
1062		DprintQ((statp->options & RES_DEBUG) ||
1063			(statp->pfcode & RES_PRF_REPLY),
1064			(stdout, ";; wrong query name:\n"),
1065			ans, (resplen > anssiz) ? anssiz : resplen);
1066		goto wait;
1067	}
1068	if (anhp->rcode == SERVFAIL ||
1069	    anhp->rcode == NOTIMP ||
1070	    anhp->rcode == REFUSED) {
1071		DprintQ(statp->options & RES_DEBUG,
1072			(stdout, "server rejected query:\n"),
1073			ans, (resplen > anssiz) ? anssiz : resplen);
1074		res_nclose(statp);
1075		/* don't retry if called from dig */
1076		if (!statp->pfcode)
1077			return (0);
1078	}
1079	if (!(statp->options & RES_IGNTC) && anhp->tc) {
1080		/*
1081		 * To get the rest of answer,
1082		 * use TCP with same server.
1083		 */
1084		Dprint(statp->options & RES_DEBUG,
1085		       (stdout, ";; truncated answer\n"));
1086		*v_circuit = 1;
1087		res_nclose(statp);
1088		return (1);
1089	}
1090	/*
1091	 * All is well, or the error is fatal.  Signal that the
1092	 * next nameserver ought not be tried.
1093	 */
1094	return (resplen);
1095}
1096
1097static void
1098Aerror(const res_state statp, FILE *file, const char *string, int error,
1099       const struct sockaddr *address, int alen)
1100{
1101	int save = errno;
1102	char hbuf[NI_MAXHOST];
1103	char sbuf[NI_MAXSERV];
1104
1105	alen = alen;
1106
1107	if ((statp->options & RES_DEBUG) != 0U) {
1108		if (getnameinfo(address, alen, hbuf, sizeof(hbuf),
1109		    sbuf, sizeof(sbuf), niflags)) {
1110			strncpy(hbuf, "?", sizeof(hbuf) - 1);
1111			hbuf[sizeof(hbuf) - 1] = '\0';
1112			strncpy(sbuf, "?", sizeof(sbuf) - 1);
1113			sbuf[sizeof(sbuf) - 1] = '\0';
1114		}
1115		fprintf(file, "res_send: %s ([%s].%s): %s\n",
1116			string, hbuf, sbuf, strerror(error));
1117	}
1118	errno = save;
1119}
1120
1121static void
1122Perror(const res_state statp, FILE *file, const char *string, int error) {
1123	int save = errno;
1124
1125	if ((statp->options & RES_DEBUG) != 0U)
1126		fprintf(file, "res_send: %s: %s\n",
1127			string, strerror(error));
1128	errno = save;
1129}
1130
1131static int
1132sock_eq(struct sockaddr *a, struct sockaddr *b) {
1133	struct sockaddr_in *a4, *b4;
1134	struct sockaddr_in6 *a6, *b6;
1135
1136	if (a->sa_family != b->sa_family)
1137		return 0;
1138	switch (a->sa_family) {
1139	case AF_INET:
1140		a4 = (struct sockaddr_in *)a;
1141		b4 = (struct sockaddr_in *)b;
1142		return a4->sin_port == b4->sin_port &&
1143		    a4->sin_addr.s_addr == b4->sin_addr.s_addr;
1144	case AF_INET6:
1145		a6 = (struct sockaddr_in6 *)a;
1146		b6 = (struct sockaddr_in6 *)b;
1147		return a6->sin6_port == b6->sin6_port &&
1148#ifdef HAVE_SIN6_SCOPE_ID
1149		    a6->sin6_scope_id == b6->sin6_scope_id &&
1150#endif
1151		    IN6_ARE_ADDR_EQUAL(&a6->sin6_addr, &b6->sin6_addr);
1152	default:
1153		return 0;
1154	}
1155}
1156
1157#if defined(NEED_PSELECT) && !defined(USE_POLL) && !defined(USE_KQUEUE)
1158/* XXX needs to move to the porting library. */
1159static int
1160pselect(int nfds, void *rfds, void *wfds, void *efds,
1161	struct timespec *tsp, const sigset_t *sigmask)
1162{
1163	struct timeval tv, *tvp;
1164	sigset_t sigs;
1165	int n;
1166
1167	if (tsp) {
1168		tvp = &tv;
1169		tv = evTimeVal(*tsp);
1170	} else
1171		tvp = NULL;
1172	if (sigmask)
1173		sigprocmask(SIG_SETMASK, sigmask, &sigs);
1174	n = select(nfds, rfds, wfds, efds, tvp);
1175	if (sigmask)
1176		sigprocmask(SIG_SETMASK, &sigs, NULL);
1177	if (tsp)
1178		*tsp = evTimeSpec(tv);
1179	return (n);
1180}
1181#endif
1182