res_send.c revision 156956
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 * 3. All advertising materials mentioning features or use of this software
14 *    must display the following acknowledgement:
15 * 	This product includes software developed by the University of
16 * 	California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 *    may be used to endorse or promote products derived from this software
19 *    without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 */
33
34/*
35 * Portions Copyright (c) 1993 by Digital Equipment Corporation.
36 *
37 * Permission to use, copy, modify, and distribute this software for any
38 * purpose with or without fee is hereby granted, provided that the above
39 * copyright notice and this permission notice appear in all copies, and that
40 * the name of Digital Equipment Corporation not be used in advertising or
41 * publicity pertaining to distribution of the document or software without
42 * specific, written prior permission.
43 *
44 * THE SOFTWARE IS PROVIDED "AS IS" AND DIGITAL EQUIPMENT CORP. DISCLAIMS ALL
45 * WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES
46 * OF MERCHANTABILITY AND FITNESS.   IN NO EVENT SHALL DIGITAL EQUIPMENT
47 * CORPORATION BE LIABLE FOR ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL
48 * DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR
49 * PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
50 * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS
51 * SOFTWARE.
52 */
53
54/*
55 * Copyright (c) 2005 by Internet Systems Consortium, Inc. ("ISC")
56 * Portions Copyright (c) 1996-1999 by Internet Software Consortium.
57 *
58 * Permission to use, copy, modify, and distribute this software for any
59 * purpose with or without fee is hereby granted, provided that the above
60 * copyright notice and this permission notice appear in all copies.
61 *
62 * THE SOFTWARE IS PROVIDED "AS IS" AND ISC DISCLAIMS ALL WARRANTIES
63 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
64 * MERCHANTABILITY AND FITNESS.  IN NO EVENT SHALL ISC BE LIABLE FOR
65 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
66 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
67 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT
68 * OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
69 */
70
71#if defined(LIBC_SCCS) && !defined(lint)
72static const char sccsid[] = "@(#)res_send.c	8.1 (Berkeley) 6/4/93";
73static const char rcsid[] = "$Id: res_send.c,v 1.5.2.2.4.7 2005/08/15 02:04:41 marka Exp $";
74#endif /* LIBC_SCCS and not lint */
75#include <sys/cdefs.h>
76__FBSDID("$FreeBSD: head/lib/libc/resolv/res_send.c 156956 2006-03-21 15:37:16Z ume $");
77
78/*
79 * Send query to name server and wait for reply.
80 */
81
82#include "port_before.h"
83#ifndef USE_KQUEUE
84#include "fd_setsize.h"
85#endif
86
87#include "namespace.h"
88#include <sys/types.h>
89#include <sys/param.h>
90#include <sys/time.h>
91#include <sys/socket.h>
92#include <sys/uio.h>
93
94#include <netinet/in.h>
95#include <arpa/nameser.h>
96#include <arpa/inet.h>
97
98#include <errno.h>
99#include <netdb.h>
100#include <resolv.h>
101#include <signal.h>
102#include <stdio.h>
103#include <stdlib.h>
104#include <string.h>
105#include <unistd.h>
106
107#include <isc/eventlib.h>
108
109#include "port_after.h"
110
111#ifdef USE_KQUEUE
112#include <sys/event.h>
113#else
114#ifdef USE_POLL
115#ifdef HAVE_STROPTS_H
116#include <stropts.h>
117#endif
118#include <poll.h>
119#endif /* USE_POLL */
120#endif
121
122#include "un-namespace.h"
123
124/* Options.  Leave them on. */
125#define DEBUG
126#include "res_debug.h"
127#include "res_private.h"
128
129#define EXT(res) ((res)->_u._ext)
130
131#ifndef USE_POLL
132static const int highestFD = FD_SETSIZE - 1;
133#else
134static int highestFD = 0;
135#endif
136
137/* Forward. */
138
139static int		get_salen(const struct sockaddr *);
140static struct sockaddr * get_nsaddr(res_state, size_t);
141static int		send_vc(res_state, const u_char *, int,
142				u_char *, int, int *, int);
143static int		send_dg(res_state,
144#ifdef USE_KQUEUE
145				int kq,
146#endif
147				const u_char *, int,
148				u_char *, int, int *, int,
149				int *, int *);
150static void		Aerror(const res_state, FILE *, const char *, int,
151			       const struct sockaddr *, int);
152static void		Perror(const res_state, FILE *, const char *, int);
153static int		sock_eq(struct sockaddr *, struct sockaddr *);
154#if defined(NEED_PSELECT) && !defined(USE_POLL) && !defined(USE_KQUEUE)
155static int		pselect(int, void *, void *, void *,
156				struct timespec *,
157				const sigset_t *);
158#endif
159void res_pquery(const res_state, const u_char *, int, FILE *);
160
161static const int niflags = NI_NUMERICHOST | NI_NUMERICSERV;
162
163/* Public. */
164
165/* int
166 * res_isourserver(ina)
167 *	looks up "ina" in _res.ns_addr_list[]
168 * returns:
169 *	0  : not found
170 *	>0 : found
171 * author:
172 *	paul vixie, 29may94
173 */
174int
175res_ourserver_p(const res_state statp, const struct sockaddr *sa) {
176	const struct sockaddr_in *inp, *srv;
177	const struct sockaddr_in6 *in6p, *srv6;
178	int ns;
179
180	switch (sa->sa_family) {
181	case AF_INET:
182		inp = (const struct sockaddr_in *)sa;
183		for (ns = 0;  ns < statp->nscount;  ns++) {
184			srv = (struct sockaddr_in *)get_nsaddr(statp, ns);
185			if (srv->sin_family == inp->sin_family &&
186			    srv->sin_port == inp->sin_port &&
187			    (srv->sin_addr.s_addr == INADDR_ANY ||
188			     srv->sin_addr.s_addr == inp->sin_addr.s_addr))
189				return (1);
190		}
191		break;
192	case AF_INET6:
193		if (EXT(statp).ext == NULL)
194			break;
195		in6p = (const struct sockaddr_in6 *)sa;
196		for (ns = 0;  ns < statp->nscount;  ns++) {
197			srv6 = (struct sockaddr_in6 *)get_nsaddr(statp, ns);
198			if (srv6->sin6_family == in6p->sin6_family &&
199			    srv6->sin6_port == in6p->sin6_port &&
200#ifdef HAVE_SIN6_SCOPE_ID
201			    (srv6->sin6_scope_id == 0 ||
202			     srv6->sin6_scope_id == in6p->sin6_scope_id) &&
203#endif
204			    (IN6_IS_ADDR_UNSPECIFIED(&srv6->sin6_addr) ||
205			     IN6_ARE_ADDR_EQUAL(&srv6->sin6_addr, &in6p->sin6_addr)))
206				return (1);
207		}
208		break;
209	default:
210		break;
211	}
212	return (0);
213}
214
215/* int
216 * res_nameinquery(name, type, class, buf, eom)
217 *	look for (name,type,class) in the query section of packet (buf,eom)
218 * requires:
219 *	buf + HFIXEDSZ <= eom
220 * returns:
221 *	-1 : format error
222 *	0  : not found
223 *	>0 : found
224 * author:
225 *	paul vixie, 29may94
226 */
227int
228res_nameinquery(const char *name, int type, int class,
229		const u_char *buf, const u_char *eom)
230{
231	const u_char *cp = buf + HFIXEDSZ;
232	int qdcount = ntohs(((const HEADER*)buf)->qdcount);
233
234	while (qdcount-- > 0) {
235		char tname[MAXDNAME+1];
236		int n, ttype, tclass;
237
238		n = dn_expand(buf, eom, cp, tname, sizeof tname);
239		if (n < 0)
240			return (-1);
241		cp += n;
242		if (cp + 2 * INT16SZ > eom)
243			return (-1);
244		ttype = ns_get16(cp); cp += INT16SZ;
245		tclass = ns_get16(cp); cp += INT16SZ;
246		if (ttype == type && tclass == class &&
247		    ns_samename(tname, name) == 1)
248			return (1);
249	}
250	return (0);
251}
252
253/* int
254 * res_queriesmatch(buf1, eom1, buf2, eom2)
255 *	is there a 1:1 mapping of (name,type,class)
256 *	in (buf1,eom1) and (buf2,eom2)?
257 * returns:
258 *	-1 : format error
259 *	0  : not a 1:1 mapping
260 *	>0 : is a 1:1 mapping
261 * author:
262 *	paul vixie, 29may94
263 */
264int
265res_queriesmatch(const u_char *buf1, const u_char *eom1,
266		 const u_char *buf2, const u_char *eom2)
267{
268	const u_char *cp = buf1 + HFIXEDSZ;
269	int qdcount = ntohs(((const HEADER*)buf1)->qdcount);
270
271	if (buf1 + HFIXEDSZ > eom1 || buf2 + HFIXEDSZ > eom2)
272		return (-1);
273
274	/*
275	 * Only header section present in replies to
276	 * dynamic update packets.
277	 */
278	if ((((const HEADER *)buf1)->opcode == ns_o_update) &&
279	    (((const HEADER *)buf2)->opcode == ns_o_update))
280		return (1);
281
282	if (qdcount != ntohs(((const HEADER*)buf2)->qdcount))
283		return (0);
284	while (qdcount-- > 0) {
285		char tname[MAXDNAME+1];
286		int n, ttype, tclass;
287
288		n = dn_expand(buf1, eom1, cp, tname, sizeof tname);
289		if (n < 0)
290			return (-1);
291		cp += n;
292		if (cp + 2 * INT16SZ > eom1)
293			return (-1);
294		ttype = ns_get16(cp);	cp += INT16SZ;
295		tclass = ns_get16(cp); cp += INT16SZ;
296		if (!res_nameinquery(tname, ttype, tclass, buf2, eom2))
297			return (0);
298	}
299	return (1);
300}
301
302int
303res_nsend(res_state statp,
304	  const u_char *buf, int buflen, u_char *ans, int anssiz)
305{
306	int gotsomewhere, terrno, try, v_circuit, resplen, ns, n;
307#ifdef USE_KQUEUE
308	int kq;
309#endif
310	char abuf[NI_MAXHOST];
311
312#ifdef USE_POLL
313	highestFD = sysconf(_SC_OPEN_MAX) - 1;
314#endif
315
316	if (statp->nscount == 0) {
317		errno = ESRCH;
318		return (-1);
319	}
320	if (anssiz < HFIXEDSZ) {
321		errno = EINVAL;
322		return (-1);
323	}
324	DprintQ((statp->options & RES_DEBUG) || (statp->pfcode & RES_PRF_QUERY),
325		(stdout, ";; res_send()\n"), buf, buflen);
326	v_circuit = (statp->options & RES_USEVC) || buflen > PACKETSZ;
327	gotsomewhere = 0;
328	terrno = ETIMEDOUT;
329
330#ifdef USE_KQUEUE
331	if ((kq = kqueue()) < 0) {
332		Perror(statp, stderr, "kqueue", errno);
333		return (-1);
334	}
335#endif
336
337	/*
338	 * If the ns_addr_list in the resolver context has changed, then
339	 * invalidate our cached copy and the associated timing data.
340	 */
341	if (EXT(statp).nscount != 0) {
342		int needclose = 0;
343		struct sockaddr_storage peer;
344		ISC_SOCKLEN_T peerlen;
345
346		if (EXT(statp).nscount != statp->nscount)
347			needclose++;
348		else
349			for (ns = 0; ns < statp->nscount; ns++) {
350				if (statp->nsaddr_list[ns].sin_family &&
351				    !sock_eq((struct sockaddr *)&statp->nsaddr_list[ns],
352					     (struct sockaddr *)&EXT(statp).ext->nsaddrs[ns])) {
353					needclose++;
354					break;
355				}
356
357				if (EXT(statp).nssocks[ns] == -1)
358					continue;
359				peerlen = sizeof(peer);
360				if (_getsockname(EXT(statp).nssocks[ns],
361				    (struct sockaddr *)&peer, &peerlen) < 0) {
362					needclose++;
363					break;
364				}
365				if (!sock_eq((struct sockaddr *)&peer,
366				    get_nsaddr(statp, ns))) {
367					needclose++;
368					break;
369				}
370			}
371		if (needclose) {
372			res_nclose(statp);
373			EXT(statp).nscount = 0;
374		}
375	}
376
377	/*
378	 * Maybe initialize our private copy of the ns_addr_list.
379	 */
380	if (EXT(statp).nscount == 0) {
381		for (ns = 0; ns < statp->nscount; ns++) {
382			EXT(statp).nstimes[ns] = RES_MAXTIME;
383			EXT(statp).nssocks[ns] = -1;
384			if (!statp->nsaddr_list[ns].sin_family)
385				continue;
386			EXT(statp).ext->nsaddrs[ns].sin =
387				 statp->nsaddr_list[ns];
388		}
389		EXT(statp).nscount = statp->nscount;
390	}
391
392	/*
393	 * Some resolvers want to even out the load on their nameservers.
394	 * Note that RES_BLAST overrides RES_ROTATE.
395	 */
396	if ((statp->options & RES_ROTATE) != 0U &&
397	    (statp->options & RES_BLAST) == 0U) {
398		union res_sockaddr_union inu;
399		struct sockaddr_in ina;
400		int lastns = statp->nscount - 1;
401		int fd;
402		u_int16_t nstime;
403
404		if (EXT(statp).ext != NULL)
405			inu = EXT(statp).ext->nsaddrs[0];
406		ina = statp->nsaddr_list[0];
407		fd = EXT(statp).nssocks[0];
408		nstime = EXT(statp).nstimes[0];
409		for (ns = 0; ns < lastns; ns++) {
410			if (EXT(statp).ext != NULL)
411                                EXT(statp).ext->nsaddrs[ns] =
412					EXT(statp).ext->nsaddrs[ns + 1];
413			statp->nsaddr_list[ns] = statp->nsaddr_list[ns + 1];
414			EXT(statp).nssocks[ns] = EXT(statp).nssocks[ns + 1];
415			EXT(statp).nstimes[ns] = EXT(statp).nstimes[ns + 1];
416		}
417		if (EXT(statp).ext != NULL)
418			EXT(statp).ext->nsaddrs[lastns] = inu;
419		statp->nsaddr_list[lastns] = ina;
420		EXT(statp).nssocks[lastns] = fd;
421		EXT(statp).nstimes[lastns] = nstime;
422	}
423
424	/*
425	 * Send request, RETRY times, or until successful.
426	 */
427	for (try = 0; try < statp->retry; try++) {
428	    for (ns = 0; ns < statp->nscount; ns++) {
429		struct sockaddr *nsap;
430		int nsaplen;
431		nsap = get_nsaddr(statp, ns);
432		nsaplen = get_salen(nsap);
433		statp->_flags &= ~RES_F_LASTMASK;
434		statp->_flags |= (ns << RES_F_LASTSHIFT);
435 same_ns:
436		if (statp->qhook) {
437			int done = 0, loops = 0;
438
439			do {
440				res_sendhookact act;
441
442				act = (*statp->qhook)(&nsap, &buf, &buflen,
443						      ans, anssiz, &resplen);
444				switch (act) {
445				case res_goahead:
446					done = 1;
447					break;
448				case res_nextns:
449					res_nclose(statp);
450					goto next_ns;
451				case res_done:
452#ifdef USE_KQUEUE
453					_close(kq);
454#endif
455					return (resplen);
456				case res_modified:
457					/* give the hook another try */
458					if (++loops < 42) /*doug adams*/
459						break;
460					/*FALLTHROUGH*/
461				case res_error:
462					/*FALLTHROUGH*/
463				default:
464					goto fail;
465				}
466			} while (!done);
467		}
468
469		Dprint(((statp->options & RES_DEBUG) &&
470			getnameinfo(nsap, nsaplen, abuf, sizeof(abuf),
471				    NULL, 0, niflags) == 0),
472		       (stdout, ";; Querying server (# %d) address = %s\n",
473			ns + 1, abuf));
474
475
476		if (v_circuit) {
477			/* Use VC; at most one attempt per server. */
478			try = statp->retry;
479			n = send_vc(statp, buf, buflen, ans, anssiz, &terrno,
480				    ns);
481			if (n < 0)
482				goto fail;
483			if (n == 0)
484				goto next_ns;
485			resplen = n;
486		} else {
487			/* Use datagrams. */
488			n = send_dg(statp,
489#ifdef USE_KQUEUE
490				    kq,
491#endif
492				    buf, buflen, ans, anssiz, &terrno,
493				    ns, &v_circuit, &gotsomewhere);
494			if (n < 0)
495				goto fail;
496			if (n == 0)
497				goto next_ns;
498			if (v_circuit)
499				goto same_ns;
500			resplen = n;
501		}
502
503		Dprint((statp->options & RES_DEBUG) ||
504		       ((statp->pfcode & RES_PRF_REPLY) &&
505			(statp->pfcode & RES_PRF_HEAD1)),
506		       (stdout, ";; got answer:\n"));
507
508		DprintQ((statp->options & RES_DEBUG) ||
509			(statp->pfcode & RES_PRF_REPLY),
510			(stdout, "%s", ""),
511			ans, (resplen > anssiz) ? anssiz : resplen);
512
513		/*
514		 * If we have temporarily opened a virtual circuit,
515		 * or if we haven't been asked to keep a socket open,
516		 * close the socket.
517		 */
518		if ((v_circuit && (statp->options & RES_USEVC) == 0U) ||
519		    (statp->options & RES_STAYOPEN) == 0U) {
520			res_nclose(statp);
521		}
522		if (statp->rhook) {
523			int done = 0, loops = 0;
524
525			do {
526				res_sendhookact act;
527
528				act = (*statp->rhook)(nsap, buf, buflen,
529						      ans, anssiz, &resplen);
530				switch (act) {
531				case res_goahead:
532				case res_done:
533					done = 1;
534					break;
535				case res_nextns:
536					res_nclose(statp);
537					goto next_ns;
538				case res_modified:
539					/* give the hook another try */
540					if (++loops < 42) /*doug adams*/
541						break;
542					/*FALLTHROUGH*/
543				case res_error:
544					/*FALLTHROUGH*/
545				default:
546					goto fail;
547				}
548			} while (!done);
549
550		}
551#ifdef USE_KQUEUE
552		_close(kq);
553#endif
554		return (resplen);
555 next_ns: ;
556	   } /*foreach ns*/
557	} /*foreach retry*/
558	res_nclose(statp);
559#ifdef USE_KQUEUE
560	_close(kq);
561#endif
562	if (!v_circuit) {
563		if (!gotsomewhere)
564			errno = ECONNREFUSED;	/* no nameservers found */
565		else
566			errno = ETIMEDOUT;	/* no answer obtained */
567	} else
568		errno = terrno;
569	return (-1);
570 fail:
571	res_nclose(statp);
572#ifdef USE_KQUEUE
573	_close(kq);
574#endif
575	return (-1);
576}
577
578/* Private */
579
580static int
581get_salen(sa)
582	const struct sockaddr *sa;
583{
584
585#ifdef HAVE_SA_LEN
586	/* There are people do not set sa_len.  Be forgiving to them. */
587	if (sa->sa_len)
588		return (sa->sa_len);
589#endif
590
591	if (sa->sa_family == AF_INET)
592		return (sizeof(struct sockaddr_in));
593	else if (sa->sa_family == AF_INET6)
594		return (sizeof(struct sockaddr_in6));
595	else
596		return (0);	/* unknown, die on connect */
597}
598
599/*
600 * pick appropriate nsaddr_list for use.  see res_init() for initialization.
601 */
602static struct sockaddr *
603get_nsaddr(statp, n)
604	res_state statp;
605	size_t n;
606{
607
608	if (!statp->nsaddr_list[n].sin_family && EXT(statp).ext) {
609		/*
610		 * - EXT(statp).ext->nsaddrs[n] holds an address that is larger
611		 *   than struct sockaddr, and
612		 * - user code did not update statp->nsaddr_list[n].
613		 */
614		return (struct sockaddr *)(void *)&EXT(statp).ext->nsaddrs[n];
615	} else {
616		/*
617		 * - user code updated statp->nsaddr_list[n], or
618		 * - statp->nsaddr_list[n] has the same content as
619		 *   EXT(statp).ext->nsaddrs[n].
620		 */
621		return (struct sockaddr *)(void *)&statp->nsaddr_list[n];
622	}
623}
624
625static int
626send_vc(res_state statp,
627	const u_char *buf, int buflen, u_char *ans, int anssiz,
628	int *terrno, int ns)
629{
630	const HEADER *hp = (const HEADER *) buf;
631	HEADER *anhp = (HEADER *) ans;
632	struct sockaddr *nsap;
633	int nsaplen;
634	int truncating, connreset, resplen, n;
635	struct iovec iov[2];
636	u_short len;
637	u_char *cp;
638	void *tmp;
639
640	nsap = get_nsaddr(statp, ns);
641	nsaplen = get_salen(nsap);
642
643	connreset = 0;
644 same_ns:
645	truncating = 0;
646
647	/* Are we still talking to whom we want to talk to? */
648	if (statp->_vcsock >= 0 && (statp->_flags & RES_F_VC) != 0) {
649		struct sockaddr_storage peer;
650		ISC_SOCKLEN_T size = sizeof peer;
651
652		if (_getpeername(statp->_vcsock,
653				(struct sockaddr *)&peer, &size) < 0 ||
654		    !sock_eq((struct sockaddr *)&peer, nsap)) {
655			res_nclose(statp);
656			statp->_flags &= ~RES_F_VC;
657		}
658	}
659
660	if (statp->_vcsock < 0 || (statp->_flags & RES_F_VC) == 0) {
661		if (statp->_vcsock >= 0)
662			res_nclose(statp);
663
664		statp->_vcsock = _socket(nsap->sa_family, SOCK_STREAM, 0);
665		if (statp->_vcsock > highestFD) {
666			res_nclose(statp);
667			errno = ENOTSOCK;
668		}
669		if (statp->_vcsock < 0) {
670			switch (errno) {
671			case EPROTONOSUPPORT:
672#ifdef EPFNOSUPPORT
673			case EPFNOSUPPORT:
674#endif
675			case EAFNOSUPPORT:
676				Perror(statp, stderr, "socket(vc)", errno);
677				return (0);
678			default:
679				*terrno = errno;
680				Perror(statp, stderr, "socket(vc)", errno);
681				return (-1);
682			}
683		}
684		errno = 0;
685		if (_connect(statp->_vcsock, nsap, nsaplen) < 0) {
686			*terrno = errno;
687			Aerror(statp, stderr, "connect/vc", errno, nsap,
688			    nsaplen);
689			res_nclose(statp);
690			return (0);
691		}
692		statp->_flags |= RES_F_VC;
693	}
694
695	/*
696	 * Send length & message
697	 */
698	ns_put16((u_short)buflen, (u_char*)&len);
699	iov[0] = evConsIovec(&len, INT16SZ);
700	DE_CONST(buf, tmp);
701	iov[1] = evConsIovec(tmp, buflen);
702	if (_writev(statp->_vcsock, iov, 2) != (INT16SZ + buflen)) {
703		*terrno = errno;
704		Perror(statp, stderr, "write failed", errno);
705		res_nclose(statp);
706		return (0);
707	}
708	/*
709	 * Receive length & response
710	 */
711 read_len:
712	cp = ans;
713	len = INT16SZ;
714	while ((n = _read(statp->_vcsock, (char *)cp, (int)len)) > 0) {
715		cp += n;
716		if ((len -= n) == 0)
717			break;
718	}
719	if (n <= 0) {
720		*terrno = errno;
721		Perror(statp, stderr, "read failed", errno);
722		res_nclose(statp);
723		/*
724		 * A long running process might get its TCP
725		 * connection reset if the remote server was
726		 * restarted.  Requery the server instead of
727		 * trying a new one.  When there is only one
728		 * server, this means that a query might work
729		 * instead of failing.  We only allow one reset
730		 * per query to prevent looping.
731		 */
732		if (*terrno == ECONNRESET && !connreset) {
733			connreset = 1;
734			res_nclose(statp);
735			goto same_ns;
736		}
737		res_nclose(statp);
738		return (0);
739	}
740	resplen = ns_get16(ans);
741	if (resplen > anssiz) {
742		Dprint(statp->options & RES_DEBUG,
743		       (stdout, ";; response truncated\n")
744		       );
745		truncating = 1;
746		len = anssiz;
747	} else
748		len = resplen;
749	if (len < HFIXEDSZ) {
750		/*
751		 * Undersized message.
752		 */
753		Dprint(statp->options & RES_DEBUG,
754		       (stdout, ";; undersized: %d\n", len));
755		*terrno = EMSGSIZE;
756		res_nclose(statp);
757		return (0);
758	}
759	cp = ans;
760	while (len != 0 &&
761	    (n = _read(statp->_vcsock, (char *)cp, (int)len)) > 0) {
762		cp += n;
763		len -= n;
764	}
765	if (n <= 0) {
766		*terrno = errno;
767		Perror(statp, stderr, "read(vc)", errno);
768		res_nclose(statp);
769		return (0);
770	}
771	if (truncating) {
772		/*
773		 * Flush rest of answer so connection stays in synch.
774		 */
775		anhp->tc = 1;
776		len = resplen - anssiz;
777		while (len != 0) {
778			char junk[PACKETSZ];
779
780			n = _read(statp->_vcsock, junk,
781			    (len > sizeof junk) ? sizeof junk : len);
782			if (n > 0)
783				len -= n;
784			else
785				break;
786		}
787	}
788	/*
789	 * If the calling applicating has bailed out of
790	 * a previous call and failed to arrange to have
791	 * the circuit closed or the server has got
792	 * itself confused, then drop the packet and
793	 * wait for the correct one.
794	 */
795	if (hp->id != anhp->id) {
796		DprintQ((statp->options & RES_DEBUG) ||
797			(statp->pfcode & RES_PRF_REPLY),
798			(stdout, ";; old answer (unexpected):\n"),
799			ans, (resplen > anssiz) ? anssiz: resplen);
800		goto read_len;
801	}
802
803	/*
804	 * All is well, or the error is fatal.  Signal that the
805	 * next nameserver ought not be tried.
806	 */
807	return (resplen);
808}
809
810static int
811send_dg(res_state statp,
812#ifdef USE_KQUEUE
813	int kq,
814#endif
815	const u_char *buf, int buflen, u_char *ans, int anssiz,
816	int *terrno, int ns, int *v_circuit, int *gotsomewhere)
817{
818	const HEADER *hp = (const HEADER *) buf;
819	HEADER *anhp = (HEADER *) ans;
820	const struct sockaddr *nsap;
821	int nsaplen;
822	struct timespec now, timeout, finish;
823	struct sockaddr_storage from;
824	ISC_SOCKLEN_T fromlen;
825	int resplen, seconds, n, s;
826#ifdef USE_KQUEUE
827	struct kevent kv;
828#else
829#ifdef USE_POLL
830	int     polltimeout;
831	struct pollfd   pollfd;
832#else
833	fd_set dsmask;
834#endif
835#endif
836
837	nsap = get_nsaddr(statp, ns);
838	nsaplen = get_salen(nsap);
839	if (EXT(statp).nssocks[ns] == -1) {
840		EXT(statp).nssocks[ns] = _socket(nsap->sa_family,
841		    SOCK_DGRAM, 0);
842		if (EXT(statp).nssocks[ns] > highestFD) {
843			res_nclose(statp);
844			errno = ENOTSOCK;
845		}
846		if (EXT(statp).nssocks[ns] < 0) {
847			switch (errno) {
848			case EPROTONOSUPPORT:
849#ifdef EPFNOSUPPORT
850			case EPFNOSUPPORT:
851#endif
852			case EAFNOSUPPORT:
853				Perror(statp, stderr, "socket(dg)", errno);
854				return (0);
855			default:
856				*terrno = errno;
857				Perror(statp, stderr, "socket(dg)", errno);
858				return (-1);
859			}
860		}
861#ifndef CANNOT_CONNECT_DGRAM
862		/*
863		 * On a 4.3BSD+ machine (client and server,
864		 * actually), sending to a nameserver datagram
865		 * port with no nameserver will cause an
866		 * ICMP port unreachable message to be returned.
867		 * If our datagram socket is "connected" to the
868		 * server, we get an ECONNREFUSED error on the next
869		 * socket operation, and select returns if the
870		 * error message is received.  We can thus detect
871		 * the absence of a nameserver without timing out.
872		 *
873		 * When the option "insecure1" is specified, we'd
874		 * rather expect to see responses from an "unknown"
875		 * address.  In order to let the kernel accept such
876		 * responses, do not connect the socket here.
877		 * XXX: or do we need an explicit option to disable
878		 * connecting?
879		 */
880		if (!(statp->options & RES_INSECURE1) &&
881		    _connect(EXT(statp).nssocks[ns], nsap, nsaplen) < 0) {
882			Aerror(statp, stderr, "connect(dg)", errno, nsap,
883			    nsaplen);
884			res_nclose(statp);
885			return (0);
886		}
887#endif /* !CANNOT_CONNECT_DGRAM */
888		Dprint(statp->options & RES_DEBUG,
889		       (stdout, ";; new DG socket\n"))
890	}
891	s = EXT(statp).nssocks[ns];
892#ifndef CANNOT_CONNECT_DGRAM
893	if (statp->options & RES_INSECURE1) {
894		if (_sendto(s,
895		    (const char*)buf, buflen, 0, nsap, nsaplen) != buflen) {
896			Aerror(statp, stderr, "sendto", errno, nsap, nsaplen);
897			res_nclose(statp);
898			return (0);
899		}
900	} else if (send(s, (const char*)buf, buflen, 0) != buflen) {
901		Perror(statp, stderr, "send", errno);
902		res_nclose(statp);
903		return (0);
904	}
905#else /* !CANNOT_CONNECT_DGRAM */
906	if (_sendto(s, (const char*)buf, buflen, 0, nsap, nsaplen) != buflen)
907	{
908		Aerror(statp, stderr, "sendto", errno, nsap, nsaplen);
909		res_nclose(statp);
910		return (0);
911	}
912#endif /* !CANNOT_CONNECT_DGRAM */
913
914	/*
915	 * Wait for reply.
916	 */
917	seconds = (statp->retrans << ns);
918	if (ns > 0)
919		seconds /= statp->nscount;
920	if (seconds <= 0)
921		seconds = 1;
922	now = evNowTime();
923	timeout = evConsTime(seconds, 0);
924	finish = evAddTime(now, timeout);
925	goto nonow;
926 wait:
927	now = evNowTime();
928 nonow:
929#ifndef USE_POLL
930	if (evCmpTime(finish, now) > 0)
931		timeout = evSubTime(finish, now);
932	else
933		timeout = evConsTime(0, 0);
934#ifdef USE_KQUEUE
935	EV_SET(&kv, s, EVFILT_READ, EV_ADD | EV_ONESHOT, 0, 0, 0);
936	n = _kevent(kq, &kv, 1, &kv, 1, &timeout);
937#else
938	FD_ZERO(&dsmask);
939	FD_SET(s, &dsmask);
940	n = pselect(s + 1, &dsmask, NULL, NULL, &timeout, NULL);
941#endif
942#else
943	timeout = evSubTime(finish, now);
944	if (timeout.tv_sec < 0)
945		timeout = evConsTime(0, 0);
946	polltimeout = 1000*timeout.tv_sec +
947		timeout.tv_nsec/1000000;
948	pollfd.fd = s;
949	pollfd.events = POLLRDNORM;
950	n = poll(&pollfd, 1, polltimeout);
951#endif /* USE_POLL */
952
953	if (n == 0) {
954		Dprint(statp->options & RES_DEBUG, (stdout, ";; timeout\n"));
955		*gotsomewhere = 1;
956		return (0);
957	}
958	if (n < 0) {
959		if (errno == EINTR)
960			goto wait;
961#ifdef USE_KQUEUE
962		Perror(statp, stderr, "kevent", errno);
963#else
964#ifndef USE_POLL
965		Perror(statp, stderr, "select", errno);
966#else
967		Perror(statp, stderr, "poll", errno);
968#endif /* USE_POLL */
969#endif
970		res_nclose(statp);
971		return (0);
972	}
973	errno = 0;
974	fromlen = sizeof(from);
975	resplen = _recvfrom(s, (char*)ans, anssiz,0,
976			   (struct sockaddr *)&from, &fromlen);
977	if (resplen <= 0) {
978		Perror(statp, stderr, "recvfrom", errno);
979		res_nclose(statp);
980		return (0);
981	}
982	*gotsomewhere = 1;
983	if (resplen < HFIXEDSZ) {
984		/*
985		 * Undersized message.
986		 */
987		Dprint(statp->options & RES_DEBUG,
988		       (stdout, ";; undersized: %d\n",
989			resplen));
990		*terrno = EMSGSIZE;
991		res_nclose(statp);
992		return (0);
993	}
994	if (hp->id != anhp->id) {
995		/*
996		 * response from old query, ignore it.
997		 * XXX - potential security hazard could
998		 *	 be detected here.
999		 */
1000		DprintQ((statp->options & RES_DEBUG) ||
1001			(statp->pfcode & RES_PRF_REPLY),
1002			(stdout, ";; old answer:\n"),
1003			ans, (resplen > anssiz) ? anssiz : resplen);
1004		goto wait;
1005	}
1006	if (!(statp->options & RES_INSECURE1) &&
1007	    !res_ourserver_p(statp, (struct sockaddr *)&from)) {
1008		/*
1009		 * response from wrong server? ignore it.
1010		 * XXX - potential security hazard could
1011		 *	 be detected here.
1012		 */
1013		DprintQ((statp->options & RES_DEBUG) ||
1014			(statp->pfcode & RES_PRF_REPLY),
1015			(stdout, ";; not our server:\n"),
1016			ans, (resplen > anssiz) ? anssiz : resplen);
1017		goto wait;
1018	}
1019#ifdef RES_USE_EDNS0
1020	if (anhp->rcode == FORMERR && (statp->options & RES_USE_EDNS0) != 0U) {
1021		/*
1022		 * Do not retry if the server do not understand EDNS0.
1023		 * The case has to be captured here, as FORMERR packet do not
1024		 * carry query section, hence res_queriesmatch() returns 0.
1025		 */
1026		DprintQ(statp->options & RES_DEBUG,
1027			(stdout, "server rejected query with EDNS0:\n"),
1028			ans, (resplen > anssiz) ? anssiz : resplen);
1029		/* record the error */
1030		statp->_flags |= RES_F_EDNS0ERR;
1031		res_nclose(statp);
1032		return (0);
1033	}
1034#endif
1035	if (!(statp->options & RES_INSECURE2) &&
1036	    !res_queriesmatch(buf, buf + buflen,
1037			      ans, ans + anssiz)) {
1038		/*
1039		 * response contains wrong query? ignore it.
1040		 * XXX - potential security hazard could
1041		 *	 be detected here.
1042		 */
1043		DprintQ((statp->options & RES_DEBUG) ||
1044			(statp->pfcode & RES_PRF_REPLY),
1045			(stdout, ";; wrong query name:\n"),
1046			ans, (resplen > anssiz) ? anssiz : resplen);
1047		goto wait;
1048	}
1049	if (anhp->rcode == SERVFAIL ||
1050	    anhp->rcode == NOTIMP ||
1051	    anhp->rcode == REFUSED) {
1052		DprintQ(statp->options & RES_DEBUG,
1053			(stdout, "server rejected query:\n"),
1054			ans, (resplen > anssiz) ? anssiz : resplen);
1055		res_nclose(statp);
1056		/* don't retry if called from dig */
1057		if (!statp->pfcode)
1058			return (0);
1059	}
1060	if (!(statp->options & RES_IGNTC) && anhp->tc) {
1061		/*
1062		 * To get the rest of answer,
1063		 * use TCP with same server.
1064		 */
1065		Dprint(statp->options & RES_DEBUG,
1066		       (stdout, ";; truncated answer\n"));
1067		*v_circuit = 1;
1068		res_nclose(statp);
1069		return (1);
1070	}
1071	/*
1072	 * All is well, or the error is fatal.  Signal that the
1073	 * next nameserver ought not be tried.
1074	 */
1075	return (resplen);
1076}
1077
1078static void
1079Aerror(const res_state statp, FILE *file, const char *string, int error,
1080       const struct sockaddr *address, int alen)
1081{
1082	int save = errno;
1083	char hbuf[NI_MAXHOST];
1084	char sbuf[NI_MAXSERV];
1085
1086	alen = alen;
1087
1088	if ((statp->options & RES_DEBUG) != 0U) {
1089		if (getnameinfo(address, alen, hbuf, sizeof(hbuf),
1090		    sbuf, sizeof(sbuf), niflags)) {
1091			strncpy(hbuf, "?", sizeof(hbuf) - 1);
1092			hbuf[sizeof(hbuf) - 1] = '\0';
1093			strncpy(sbuf, "?", sizeof(sbuf) - 1);
1094			sbuf[sizeof(sbuf) - 1] = '\0';
1095		}
1096		fprintf(file, "res_send: %s ([%s].%s): %s\n",
1097			string, hbuf, sbuf, strerror(error));
1098	}
1099	errno = save;
1100}
1101
1102static void
1103Perror(const res_state statp, FILE *file, const char *string, int error) {
1104	int save = errno;
1105
1106	if ((statp->options & RES_DEBUG) != 0U)
1107		fprintf(file, "res_send: %s: %s\n",
1108			string, strerror(error));
1109	errno = save;
1110}
1111
1112static int
1113sock_eq(struct sockaddr *a, struct sockaddr *b) {
1114	struct sockaddr_in *a4, *b4;
1115	struct sockaddr_in6 *a6, *b6;
1116
1117	if (a->sa_family != b->sa_family)
1118		return 0;
1119	switch (a->sa_family) {
1120	case AF_INET:
1121		a4 = (struct sockaddr_in *)a;
1122		b4 = (struct sockaddr_in *)b;
1123		return a4->sin_port == b4->sin_port &&
1124		    a4->sin_addr.s_addr == b4->sin_addr.s_addr;
1125	case AF_INET6:
1126		a6 = (struct sockaddr_in6 *)a;
1127		b6 = (struct sockaddr_in6 *)b;
1128		return a6->sin6_port == b6->sin6_port &&
1129#ifdef HAVE_SIN6_SCOPE_ID
1130		    a6->sin6_scope_id == b6->sin6_scope_id &&
1131#endif
1132		    IN6_ARE_ADDR_EQUAL(&a6->sin6_addr, &b6->sin6_addr);
1133	default:
1134		return 0;
1135	}
1136}
1137
1138#if defined(NEED_PSELECT) && !defined(USE_POLL) && !defined(USE_KQUEUE)
1139/* XXX needs to move to the porting library. */
1140static int
1141pselect(int nfds, void *rfds, void *wfds, void *efds,
1142	struct timespec *tsp, const sigset_t *sigmask)
1143{
1144	struct timeval tv, *tvp;
1145	sigset_t sigs;
1146	int n;
1147
1148	if (tsp) {
1149		tvp = &tv;
1150		tv = evTimeVal(*tsp);
1151	} else
1152		tvp = NULL;
1153	if (sigmask)
1154		sigprocmask(SIG_SETMASK, sigmask, &sigs);
1155	n = select(nfds, rfds, wfds, efds, tvp);
1156	if (sigmask)
1157		sigprocmask(SIG_SETMASK, &sigs, NULL);
1158	if (tsp)
1159		*tsp = evTimeSpec(tv);
1160	return (n);
1161}
1162#endif
1163