res_send.c revision 165903
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 1.5.2.2.4.9 2006/10/16 23:00:50 marka Exp $";
70#endif /* LIBC_SCCS and not lint */
71#include <sys/cdefs.h>
72__FBSDID("$FreeBSD: head/lib/libc/resolv/res_send.c 165903 2007-01-09 00:28:16Z imp $");
73
74/*
75 * Send query to name server and wait for reply.
76 */
77
78#include "port_before.h"
79#ifndef USE_KQUEUE
80#include "fd_setsize.h"
81#endif
82
83#include "namespace.h"
84#include <sys/types.h>
85#include <sys/param.h>
86#include <sys/time.h>
87#include <sys/socket.h>
88#include <sys/uio.h>
89
90#include <netinet/in.h>
91#include <arpa/nameser.h>
92#include <arpa/inet.h>
93
94#include <errno.h>
95#include <netdb.h>
96#include <resolv.h>
97#include <signal.h>
98#include <stdio.h>
99#include <stdlib.h>
100#include <string.h>
101#include <unistd.h>
102
103#include <isc/eventlib.h>
104
105#include "port_after.h"
106
107#ifdef USE_KQUEUE
108#include <sys/event.h>
109#else
110#ifdef USE_POLL
111#ifdef HAVE_STROPTS_H
112#include <stropts.h>
113#endif
114#include <poll.h>
115#endif /* USE_POLL */
116#endif
117
118#include "un-namespace.h"
119
120/* Options.  Leave them on. */
121#define DEBUG
122#include "res_debug.h"
123#include "res_private.h"
124
125#define EXT(res) ((res)->_u._ext)
126
127#ifndef USE_POLL
128static const int highestFD = FD_SETSIZE - 1;
129#else
130static int highestFD = 0;
131#endif
132
133/* Forward. */
134
135static int		get_salen(const struct sockaddr *);
136static struct sockaddr * get_nsaddr(res_state, size_t);
137static int		send_vc(res_state, const u_char *, int,
138				u_char *, int, int *, int);
139static int		send_dg(res_state,
140#ifdef USE_KQUEUE
141				int kq,
142#endif
143				const u_char *, int,
144				u_char *, int, int *, int, int,
145				int *, int *);
146static void		Aerror(const res_state, FILE *, const char *, int,
147			       const struct sockaddr *, int);
148static void		Perror(const res_state, FILE *, const char *, int);
149static int		sock_eq(struct sockaddr *, struct sockaddr *);
150#if defined(NEED_PSELECT) && !defined(USE_POLL) && !defined(USE_KQUEUE)
151static int		pselect(int, void *, void *, void *,
152				struct timespec *,
153				const sigset_t *);
154#endif
155void res_pquery(const res_state, const u_char *, int, FILE *);
156
157static const int niflags = NI_NUMERICHOST | NI_NUMERICSERV;
158
159/* Public. */
160
161/* int
162 * res_isourserver(ina)
163 *	looks up "ina" in _res.ns_addr_list[]
164 * returns:
165 *	0  : not found
166 *	>0 : found
167 * author:
168 *	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/* int
212 * res_nameinquery(name, type, class, buf, eom)
213 *	look for (name,type,class) in the query section of packet (buf,eom)
214 * requires:
215 *	buf + HFIXEDSZ <= eom
216 * returns:
217 *	-1 : format error
218 *	0  : not found
219 *	>0 : found
220 * author:
221 *	paul vixie, 29may94
222 */
223int
224res_nameinquery(const char *name, int type, int class,
225		const u_char *buf, const u_char *eom)
226{
227	const u_char *cp = buf + HFIXEDSZ;
228	int qdcount = ntohs(((const HEADER*)buf)->qdcount);
229
230	while (qdcount-- > 0) {
231		char tname[MAXDNAME+1];
232		int n, ttype, tclass;
233
234		n = dn_expand(buf, eom, cp, tname, sizeof tname);
235		if (n < 0)
236			return (-1);
237		cp += n;
238		if (cp + 2 * INT16SZ > eom)
239			return (-1);
240		ttype = ns_get16(cp); cp += INT16SZ;
241		tclass = ns_get16(cp); cp += INT16SZ;
242		if (ttype == type && tclass == class &&
243		    ns_samename(tname, name) == 1)
244			return (1);
245	}
246	return (0);
247}
248
249/* int
250 * res_queriesmatch(buf1, eom1, buf2, eom2)
251 *	is there a 1:1 mapping of (name,type,class)
252 *	in (buf1,eom1) and (buf2,eom2)?
253 * returns:
254 *	-1 : format error
255 *	0  : not a 1:1 mapping
256 *	>0 : is a 1:1 mapping
257 * author:
258 *	paul vixie, 29may94
259 */
260int
261res_queriesmatch(const u_char *buf1, const u_char *eom1,
262		 const u_char *buf2, const u_char *eom2)
263{
264	const u_char *cp = buf1 + HFIXEDSZ;
265	int qdcount = ntohs(((const HEADER*)buf1)->qdcount);
266
267	if (buf1 + HFIXEDSZ > eom1 || buf2 + HFIXEDSZ > eom2)
268		return (-1);
269
270	/*
271	 * Only header section present in replies to
272	 * dynamic update packets.
273	 */
274	if ((((const HEADER *)buf1)->opcode == ns_o_update) &&
275	    (((const HEADER *)buf2)->opcode == ns_o_update))
276		return (1);
277
278	if (qdcount != ntohs(((const HEADER*)buf2)->qdcount))
279		return (0);
280	while (qdcount-- > 0) {
281		char tname[MAXDNAME+1];
282		int n, ttype, tclass;
283
284		n = dn_expand(buf1, eom1, cp, tname, sizeof tname);
285		if (n < 0)
286			return (-1);
287		cp += n;
288		if (cp + 2 * INT16SZ > eom1)
289			return (-1);
290		ttype = ns_get16(cp);	cp += INT16SZ;
291		tclass = ns_get16(cp); cp += INT16SZ;
292		if (!res_nameinquery(tname, ttype, tclass, buf2, eom2))
293			return (0);
294	}
295	return (1);
296}
297
298int
299res_nsend(res_state statp,
300	  const u_char *buf, int buflen, u_char *ans, int anssiz)
301{
302	int gotsomewhere, terrno, try, v_circuit, resplen, ns, n;
303#ifdef USE_KQUEUE
304	int kq;
305#endif
306	char abuf[NI_MAXHOST];
307
308#ifdef USE_POLL
309	highestFD = sysconf(_SC_OPEN_MAX) - 1;
310#endif
311
312	/* No name servers or res_init() failure */
313	if (statp->nscount == 0 || EXT(statp).ext == NULL) {
314		errno = ESRCH;
315		return (-1);
316	}
317	if (anssiz < HFIXEDSZ) {
318		errno = EINVAL;
319		return (-1);
320	}
321	DprintQ((statp->options & RES_DEBUG) || (statp->pfcode & RES_PRF_QUERY),
322		(stdout, ";; res_send()\n"), buf, buflen);
323	v_circuit = (statp->options & RES_USEVC) || buflen > PACKETSZ;
324	gotsomewhere = 0;
325	terrno = ETIMEDOUT;
326
327#ifdef USE_KQUEUE
328	if ((kq = kqueue()) < 0) {
329		Perror(statp, stderr, "kqueue", errno);
330		return (-1);
331	}
332#endif
333
334	/*
335	 * If the ns_addr_list in the resolver context has changed, then
336	 * invalidate our cached copy and the associated timing data.
337	 */
338	if (EXT(statp).nscount != 0) {
339		int needclose = 0;
340		struct sockaddr_storage peer;
341		ISC_SOCKLEN_T peerlen;
342
343		if (EXT(statp).nscount != statp->nscount)
344			needclose++;
345		else
346			for (ns = 0; ns < statp->nscount; ns++) {
347				if (statp->nsaddr_list[ns].sin_family &&
348				    !sock_eq((struct sockaddr *)&statp->nsaddr_list[ns],
349					     (struct sockaddr *)&EXT(statp).ext->nsaddrs[ns])) {
350					needclose++;
351					break;
352				}
353
354				if (EXT(statp).nssocks[ns] == -1)
355					continue;
356				peerlen = sizeof(peer);
357				if (_getsockname(EXT(statp).nssocks[ns],
358				    (struct sockaddr *)&peer, &peerlen) < 0) {
359					needclose++;
360					break;
361				}
362				if (!sock_eq((struct sockaddr *)&peer,
363				    get_nsaddr(statp, ns))) {
364					needclose++;
365					break;
366				}
367			}
368		if (needclose) {
369			res_nclose(statp);
370			EXT(statp).nscount = 0;
371		}
372	}
373
374	/*
375	 * Maybe initialize our private copy of the ns_addr_list.
376	 */
377	if (EXT(statp).nscount == 0) {
378		for (ns = 0; ns < statp->nscount; ns++) {
379			EXT(statp).nstimes[ns] = RES_MAXTIME;
380			EXT(statp).nssocks[ns] = -1;
381			if (!statp->nsaddr_list[ns].sin_family)
382				continue;
383			EXT(statp).ext->nsaddrs[ns].sin =
384				 statp->nsaddr_list[ns];
385		}
386		EXT(statp).nscount = statp->nscount;
387	}
388
389	/*
390	 * Some resolvers want to even out the load on their nameservers.
391	 * Note that RES_BLAST overrides RES_ROTATE.
392	 */
393	if ((statp->options & RES_ROTATE) != 0U &&
394	    (statp->options & RES_BLAST) == 0U) {
395		union res_sockaddr_union inu;
396		struct sockaddr_in ina;
397		int lastns = statp->nscount - 1;
398		int fd;
399		u_int16_t nstime;
400
401		if (EXT(statp).ext != NULL)
402			inu = EXT(statp).ext->nsaddrs[0];
403		ina = statp->nsaddr_list[0];
404		fd = EXT(statp).nssocks[0];
405		nstime = EXT(statp).nstimes[0];
406		for (ns = 0; ns < lastns; ns++) {
407			if (EXT(statp).ext != NULL)
408                                EXT(statp).ext->nsaddrs[ns] =
409					EXT(statp).ext->nsaddrs[ns + 1];
410			statp->nsaddr_list[ns] = statp->nsaddr_list[ns + 1];
411			EXT(statp).nssocks[ns] = EXT(statp).nssocks[ns + 1];
412			EXT(statp).nstimes[ns] = EXT(statp).nstimes[ns + 1];
413		}
414		if (EXT(statp).ext != NULL)
415			EXT(statp).ext->nsaddrs[lastns] = inu;
416		statp->nsaddr_list[lastns] = ina;
417		EXT(statp).nssocks[lastns] = fd;
418		EXT(statp).nstimes[lastns] = nstime;
419	}
420
421	/*
422	 * Send request, RETRY times, or until successful.
423	 */
424	for (try = 0; try < statp->retry; try++) {
425	    for (ns = 0; ns < statp->nscount; ns++) {
426		struct sockaddr *nsap;
427		int nsaplen;
428		nsap = get_nsaddr(statp, ns);
429		nsaplen = get_salen(nsap);
430		statp->_flags &= ~RES_F_LASTMASK;
431		statp->_flags |= (ns << RES_F_LASTSHIFT);
432 same_ns:
433		if (statp->qhook) {
434			int done = 0, loops = 0;
435
436			do {
437				res_sendhookact act;
438
439				act = (*statp->qhook)(&nsap, &buf, &buflen,
440						      ans, anssiz, &resplen);
441				switch (act) {
442				case res_goahead:
443					done = 1;
444					break;
445				case res_nextns:
446					res_nclose(statp);
447					goto next_ns;
448				case res_done:
449#ifdef USE_KQUEUE
450					_close(kq);
451#endif
452					return (resplen);
453				case res_modified:
454					/* give the hook another try */
455					if (++loops < 42) /*doug adams*/
456						break;
457					/*FALLTHROUGH*/
458				case res_error:
459					/*FALLTHROUGH*/
460				default:
461					goto fail;
462				}
463			} while (!done);
464		}
465
466		Dprint(((statp->options & RES_DEBUG) &&
467			getnameinfo(nsap, nsaplen, abuf, sizeof(abuf),
468				    NULL, 0, niflags) == 0),
469		       (stdout, ";; Querying server (# %d) address = %s\n",
470			ns + 1, abuf));
471
472
473		if (v_circuit) {
474			/* Use VC; at most one attempt per server. */
475			try = statp->retry;
476			n = send_vc(statp, buf, buflen, ans, anssiz, &terrno,
477				    ns);
478			if (n < 0)
479				goto fail;
480			if (n == 0)
481				goto next_ns;
482			resplen = n;
483		} else {
484			/* Use datagrams. */
485			n = send_dg(statp,
486#ifdef USE_KQUEUE
487				    kq,
488#endif
489				    buf, buflen, ans, anssiz, &terrno,
490				    ns, try, &v_circuit, &gotsomewhere);
491			if (n < 0)
492				goto fail;
493			if (n == 0)
494				goto next_ns;
495			if (v_circuit)
496				goto same_ns;
497			resplen = n;
498		}
499
500		Dprint((statp->options & RES_DEBUG) ||
501		       ((statp->pfcode & RES_PRF_REPLY) &&
502			(statp->pfcode & RES_PRF_HEAD1)),
503		       (stdout, ";; got answer:\n"));
504
505		DprintQ((statp->options & RES_DEBUG) ||
506			(statp->pfcode & RES_PRF_REPLY),
507			(stdout, "%s", ""),
508			ans, (resplen > anssiz) ? anssiz : resplen);
509
510		/*
511		 * If we have temporarily opened a virtual circuit,
512		 * or if we haven't been asked to keep a socket open,
513		 * close the socket.
514		 */
515		if ((v_circuit && (statp->options & RES_USEVC) == 0U) ||
516		    (statp->options & RES_STAYOPEN) == 0U) {
517			res_nclose(statp);
518		}
519		if (statp->rhook) {
520			int done = 0, loops = 0;
521
522			do {
523				res_sendhookact act;
524
525				act = (*statp->rhook)(nsap, buf, buflen,
526						      ans, anssiz, &resplen);
527				switch (act) {
528				case res_goahead:
529				case res_done:
530					done = 1;
531					break;
532				case res_nextns:
533					res_nclose(statp);
534					goto next_ns;
535				case res_modified:
536					/* give the hook another try */
537					if (++loops < 42) /*doug adams*/
538						break;
539					/*FALLTHROUGH*/
540				case res_error:
541					/*FALLTHROUGH*/
542				default:
543					goto fail;
544				}
545			} while (!done);
546
547		}
548#ifdef USE_KQUEUE
549		_close(kq);
550#endif
551		return (resplen);
552 next_ns: ;
553	   } /*foreach ns*/
554	} /*foreach retry*/
555	res_nclose(statp);
556#ifdef USE_KQUEUE
557	_close(kq);
558#endif
559	if (!v_circuit) {
560		if (!gotsomewhere)
561			errno = ECONNREFUSED;	/* no nameservers found */
562		else
563			errno = ETIMEDOUT;	/* no answer obtained */
564	} else
565		errno = terrno;
566	return (-1);
567 fail:
568	res_nclose(statp);
569#ifdef USE_KQUEUE
570	_close(kq);
571#endif
572	return (-1);
573}
574
575/* Private */
576
577static int
578get_salen(sa)
579	const struct sockaddr *sa;
580{
581
582#ifdef HAVE_SA_LEN
583	/* There are people do not set sa_len.  Be forgiving to them. */
584	if (sa->sa_len)
585		return (sa->sa_len);
586#endif
587
588	if (sa->sa_family == AF_INET)
589		return (sizeof(struct sockaddr_in));
590	else if (sa->sa_family == AF_INET6)
591		return (sizeof(struct sockaddr_in6));
592	else
593		return (0);	/* unknown, die on connect */
594}
595
596/*
597 * pick appropriate nsaddr_list for use.  see res_init() for initialization.
598 */
599static struct sockaddr *
600get_nsaddr(statp, n)
601	res_state statp;
602	size_t n;
603{
604
605	if (!statp->nsaddr_list[n].sin_family && EXT(statp).ext) {
606		/*
607		 * - EXT(statp).ext->nsaddrs[n] holds an address that is larger
608		 *   than struct sockaddr, and
609		 * - user code did not update statp->nsaddr_list[n].
610		 */
611		return (struct sockaddr *)(void *)&EXT(statp).ext->nsaddrs[n];
612	} else {
613		/*
614		 * - user code updated statp->nsaddr_list[n], or
615		 * - statp->nsaddr_list[n] has the same content as
616		 *   EXT(statp).ext->nsaddrs[n].
617		 */
618		return (struct sockaddr *)(void *)&statp->nsaddr_list[n];
619	}
620}
621
622static int
623send_vc(res_state statp,
624	const u_char *buf, int buflen, u_char *ans, int anssiz,
625	int *terrno, int ns)
626{
627	const HEADER *hp = (const HEADER *) buf;
628	HEADER *anhp = (HEADER *) ans;
629	struct sockaddr *nsap;
630	int nsaplen;
631	int truncating, connreset, resplen, n;
632	struct iovec iov[2];
633	u_short len;
634	u_char *cp;
635	void *tmp;
636
637	nsap = get_nsaddr(statp, ns);
638	nsaplen = get_salen(nsap);
639
640	connreset = 0;
641 same_ns:
642	truncating = 0;
643
644	/* Are we still talking to whom we want to talk to? */
645	if (statp->_vcsock >= 0 && (statp->_flags & RES_F_VC) != 0) {
646		struct sockaddr_storage peer;
647		ISC_SOCKLEN_T size = sizeof peer;
648
649		if (_getpeername(statp->_vcsock,
650				(struct sockaddr *)&peer, &size) < 0 ||
651		    !sock_eq((struct sockaddr *)&peer, nsap)) {
652			res_nclose(statp);
653			statp->_flags &= ~RES_F_VC;
654		}
655	}
656
657	if (statp->_vcsock < 0 || (statp->_flags & RES_F_VC) == 0) {
658		if (statp->_vcsock >= 0)
659			res_nclose(statp);
660
661		statp->_vcsock = _socket(nsap->sa_family, SOCK_STREAM, 0);
662		if (statp->_vcsock > highestFD) {
663			res_nclose(statp);
664			errno = ENOTSOCK;
665		}
666		if (statp->_vcsock < 0) {
667			switch (errno) {
668			case EPROTONOSUPPORT:
669#ifdef EPFNOSUPPORT
670			case EPFNOSUPPORT:
671#endif
672			case EAFNOSUPPORT:
673				Perror(statp, stderr, "socket(vc)", errno);
674				return (0);
675			default:
676				*terrno = errno;
677				Perror(statp, stderr, "socket(vc)", errno);
678				return (-1);
679			}
680		}
681		errno = 0;
682		if (_connect(statp->_vcsock, nsap, nsaplen) < 0) {
683			*terrno = errno;
684			Aerror(statp, stderr, "connect/vc", errno, nsap,
685			    nsaplen);
686			res_nclose(statp);
687			return (0);
688		}
689		statp->_flags |= RES_F_VC;
690	}
691
692	/*
693	 * Send length & message
694	 */
695	ns_put16((u_short)buflen, (u_char*)&len);
696	iov[0] = evConsIovec(&len, INT16SZ);
697	DE_CONST(buf, tmp);
698	iov[1] = evConsIovec(tmp, buflen);
699	if (_writev(statp->_vcsock, iov, 2) != (INT16SZ + buflen)) {
700		*terrno = errno;
701		Perror(statp, stderr, "write failed", errno);
702		res_nclose(statp);
703		return (0);
704	}
705	/*
706	 * Receive length & response
707	 */
708 read_len:
709	cp = ans;
710	len = INT16SZ;
711	while ((n = _read(statp->_vcsock, (char *)cp, (int)len)) > 0) {
712		cp += n;
713		if ((len -= n) == 0)
714			break;
715	}
716	if (n <= 0) {
717		*terrno = errno;
718		Perror(statp, stderr, "read failed", errno);
719		res_nclose(statp);
720		/*
721		 * A long running process might get its TCP
722		 * connection reset if the remote server was
723		 * restarted.  Requery the server instead of
724		 * trying a new one.  When there is only one
725		 * server, this means that a query might work
726		 * instead of failing.  We only allow one reset
727		 * per query to prevent looping.
728		 */
729		if (*terrno == ECONNRESET && !connreset) {
730			connreset = 1;
731			res_nclose(statp);
732			goto same_ns;
733		}
734		res_nclose(statp);
735		return (0);
736	}
737	resplen = ns_get16(ans);
738	if (resplen > anssiz) {
739		Dprint(statp->options & RES_DEBUG,
740		       (stdout, ";; response truncated\n")
741		       );
742		truncating = 1;
743		len = anssiz;
744	} else
745		len = resplen;
746	if (len < HFIXEDSZ) {
747		/*
748		 * Undersized message.
749		 */
750		Dprint(statp->options & RES_DEBUG,
751		       (stdout, ";; undersized: %d\n", len));
752		*terrno = EMSGSIZE;
753		res_nclose(statp);
754		return (0);
755	}
756	cp = ans;
757	while (len != 0 &&
758	    (n = _read(statp->_vcsock, (char *)cp, (int)len)) > 0) {
759		cp += n;
760		len -= n;
761	}
762	if (n <= 0) {
763		*terrno = errno;
764		Perror(statp, stderr, "read(vc)", errno);
765		res_nclose(statp);
766		return (0);
767	}
768	if (truncating) {
769		/*
770		 * Flush rest of answer so connection stays in synch.
771		 */
772		anhp->tc = 1;
773		len = resplen - anssiz;
774		while (len != 0) {
775			char junk[PACKETSZ];
776
777			n = _read(statp->_vcsock, junk,
778			    (len > sizeof junk) ? sizeof junk : len);
779			if (n > 0)
780				len -= n;
781			else
782				break;
783		}
784	}
785	/*
786	 * If the calling applicating has bailed out of
787	 * a previous call and failed to arrange to have
788	 * the circuit closed or the server has got
789	 * itself confused, then drop the packet and
790	 * wait for the correct one.
791	 */
792	if (hp->id != anhp->id) {
793		DprintQ((statp->options & RES_DEBUG) ||
794			(statp->pfcode & RES_PRF_REPLY),
795			(stdout, ";; old answer (unexpected):\n"),
796			ans, (resplen > anssiz) ? anssiz: resplen);
797		goto read_len;
798	}
799
800	/*
801	 * All is well, or the error is fatal.  Signal that the
802	 * next nameserver ought not be tried.
803	 */
804	return (resplen);
805}
806
807static int
808send_dg(res_state statp,
809#ifdef USE_KQUEUE
810	int kq,
811#endif
812	const u_char *buf, int buflen, u_char *ans,
813	int anssiz, int *terrno, int ns, int try, int *v_circuit,
814	int *gotsomewhere)
815{
816	const HEADER *hp = (const HEADER *) buf;
817	HEADER *anhp = (HEADER *) ans;
818	const struct sockaddr *nsap;
819	int nsaplen;
820	struct timespec now, timeout, finish;
821	struct sockaddr_storage from;
822	ISC_SOCKLEN_T fromlen;
823	int resplen, seconds, n, s;
824#ifdef USE_KQUEUE
825	struct kevent kv;
826#else
827#ifdef USE_POLL
828	int     polltimeout;
829	struct pollfd   pollfd;
830#else
831	fd_set dsmask;
832#endif
833#endif
834
835	nsap = get_nsaddr(statp, ns);
836	nsaplen = get_salen(nsap);
837	if (EXT(statp).nssocks[ns] == -1) {
838		EXT(statp).nssocks[ns] = _socket(nsap->sa_family,
839		    SOCK_DGRAM, 0);
840		if (EXT(statp).nssocks[ns] > highestFD) {
841			res_nclose(statp);
842			errno = ENOTSOCK;
843		}
844		if (EXT(statp).nssocks[ns] < 0) {
845			switch (errno) {
846			case EPROTONOSUPPORT:
847#ifdef EPFNOSUPPORT
848			case EPFNOSUPPORT:
849#endif
850			case EAFNOSUPPORT:
851				Perror(statp, stderr, "socket(dg)", errno);
852				return (0);
853			default:
854				*terrno = errno;
855				Perror(statp, stderr, "socket(dg)", errno);
856				return (-1);
857			}
858		}
859#ifndef CANNOT_CONNECT_DGRAM
860		/*
861		 * On a 4.3BSD+ machine (client and server,
862		 * actually), sending to a nameserver datagram
863		 * port with no nameserver will cause an
864		 * ICMP port unreachable message to be returned.
865		 * If our datagram socket is "connected" to the
866		 * server, we get an ECONNREFUSED error on the next
867		 * socket operation, and select returns if the
868		 * error message is received.  We can thus detect
869		 * the absence of a nameserver without timing out.
870		 *
871		 * When the option "insecure1" is specified, we'd
872		 * rather expect to see responses from an "unknown"
873		 * address.  In order to let the kernel accept such
874		 * responses, do not connect the socket here.
875		 * XXX: or do we need an explicit option to disable
876		 * connecting?
877		 */
878		if (!(statp->options & RES_INSECURE1) &&
879		    _connect(EXT(statp).nssocks[ns], nsap, nsaplen) < 0) {
880			Aerror(statp, stderr, "connect(dg)", errno, nsap,
881			    nsaplen);
882			res_nclose(statp);
883			return (0);
884		}
885#endif /* !CANNOT_CONNECT_DGRAM */
886		Dprint(statp->options & RES_DEBUG,
887		       (stdout, ";; new DG socket\n"))
888	}
889	s = EXT(statp).nssocks[ns];
890#ifndef CANNOT_CONNECT_DGRAM
891	if (statp->options & RES_INSECURE1) {
892		if (_sendto(s,
893		    (const char*)buf, buflen, 0, nsap, nsaplen) != buflen) {
894			Aerror(statp, stderr, "sendto", errno, nsap, nsaplen);
895			res_nclose(statp);
896			return (0);
897		}
898	} else if (send(s, (const char*)buf, buflen, 0) != buflen) {
899		Perror(statp, stderr, "send", errno);
900		res_nclose(statp);
901		return (0);
902	}
903#else /* !CANNOT_CONNECT_DGRAM */
904	if (_sendto(s, (const char*)buf, buflen, 0, nsap, nsaplen) != buflen)
905	{
906		Aerror(statp, stderr, "sendto", errno, nsap, nsaplen);
907		res_nclose(statp);
908		return (0);
909	}
910#endif /* !CANNOT_CONNECT_DGRAM */
911
912	/*
913	 * Wait for reply.
914	 */
915	seconds = (statp->retrans << try);
916	if (ns > 0)
917		seconds /= statp->nscount;
918	if (seconds <= 0)
919		seconds = 1;
920	now = evNowTime();
921	timeout = evConsTime(seconds, 0);
922	finish = evAddTime(now, timeout);
923	goto nonow;
924 wait:
925	now = evNowTime();
926 nonow:
927#ifndef USE_POLL
928	if (evCmpTime(finish, now) > 0)
929		timeout = evSubTime(finish, now);
930	else
931		timeout = evConsTime(0, 0);
932#ifdef USE_KQUEUE
933	EV_SET(&kv, s, EVFILT_READ, EV_ADD | EV_ONESHOT, 0, 0, 0);
934	n = _kevent(kq, &kv, 1, &kv, 1, &timeout);
935#else
936	FD_ZERO(&dsmask);
937	FD_SET(s, &dsmask);
938	n = pselect(s + 1, &dsmask, NULL, NULL, &timeout, NULL);
939#endif
940#else
941	timeout = evSubTime(finish, now);
942	if (timeout.tv_sec < 0)
943		timeout = evConsTime(0, 0);
944	polltimeout = 1000*timeout.tv_sec +
945		timeout.tv_nsec/1000000;
946	pollfd.fd = s;
947	pollfd.events = POLLRDNORM;
948	n = poll(&pollfd, 1, polltimeout);
949#endif /* USE_POLL */
950
951	if (n == 0) {
952		Dprint(statp->options & RES_DEBUG, (stdout, ";; timeout\n"));
953		*gotsomewhere = 1;
954		return (0);
955	}
956	if (n < 0) {
957		if (errno == EINTR)
958			goto wait;
959#ifdef USE_KQUEUE
960		Perror(statp, stderr, "kevent", errno);
961#else
962#ifndef USE_POLL
963		Perror(statp, stderr, "select", errno);
964#else
965		Perror(statp, stderr, "poll", errno);
966#endif /* USE_POLL */
967#endif
968		res_nclose(statp);
969		return (0);
970	}
971#ifdef USE_KQUEUE
972	if (kv.ident != s)
973		goto wait;
974#endif
975	errno = 0;
976	fromlen = sizeof(from);
977	resplen = _recvfrom(s, (char*)ans, anssiz,0,
978			   (struct sockaddr *)&from, &fromlen);
979	if (resplen <= 0) {
980		Perror(statp, stderr, "recvfrom", errno);
981		res_nclose(statp);
982		return (0);
983	}
984	*gotsomewhere = 1;
985	if (resplen < HFIXEDSZ) {
986		/*
987		 * Undersized message.
988		 */
989		Dprint(statp->options & RES_DEBUG,
990		       (stdout, ";; undersized: %d\n",
991			resplen));
992		*terrno = EMSGSIZE;
993		res_nclose(statp);
994		return (0);
995	}
996	if (hp->id != anhp->id) {
997		/*
998		 * response from old query, ignore it.
999		 * XXX - potential security hazard could
1000		 *	 be detected here.
1001		 */
1002		DprintQ((statp->options & RES_DEBUG) ||
1003			(statp->pfcode & RES_PRF_REPLY),
1004			(stdout, ";; old answer:\n"),
1005			ans, (resplen > anssiz) ? anssiz : resplen);
1006		goto wait;
1007	}
1008	if (!(statp->options & RES_INSECURE1) &&
1009	    !res_ourserver_p(statp, (struct sockaddr *)&from)) {
1010		/*
1011		 * response from wrong server? ignore it.
1012		 * XXX - potential security hazard could
1013		 *	 be detected here.
1014		 */
1015		DprintQ((statp->options & RES_DEBUG) ||
1016			(statp->pfcode & RES_PRF_REPLY),
1017			(stdout, ";; not our server:\n"),
1018			ans, (resplen > anssiz) ? anssiz : resplen);
1019		goto wait;
1020	}
1021#ifdef RES_USE_EDNS0
1022	if (anhp->rcode == FORMERR && (statp->options & RES_USE_EDNS0) != 0U) {
1023		/*
1024		 * Do not retry if the server do not understand EDNS0.
1025		 * The case has to be captured here, as FORMERR packet do not
1026		 * carry query section, hence res_queriesmatch() returns 0.
1027		 */
1028		DprintQ(statp->options & RES_DEBUG,
1029			(stdout, "server rejected query with EDNS0:\n"),
1030			ans, (resplen > anssiz) ? anssiz : resplen);
1031		/* record the error */
1032		statp->_flags |= RES_F_EDNS0ERR;
1033		res_nclose(statp);
1034		return (0);
1035	}
1036#endif
1037	if (!(statp->options & RES_INSECURE2) &&
1038	    !res_queriesmatch(buf, buf + buflen,
1039			      ans, ans + anssiz)) {
1040		/*
1041		 * response contains wrong query? ignore it.
1042		 * XXX - potential security hazard could
1043		 *	 be detected here.
1044		 */
1045		DprintQ((statp->options & RES_DEBUG) ||
1046			(statp->pfcode & RES_PRF_REPLY),
1047			(stdout, ";; wrong query name:\n"),
1048			ans, (resplen > anssiz) ? anssiz : resplen);
1049		goto wait;
1050	}
1051	if (anhp->rcode == SERVFAIL ||
1052	    anhp->rcode == NOTIMP ||
1053	    anhp->rcode == REFUSED) {
1054		DprintQ(statp->options & RES_DEBUG,
1055			(stdout, "server rejected query:\n"),
1056			ans, (resplen > anssiz) ? anssiz : resplen);
1057		res_nclose(statp);
1058		/* don't retry if called from dig */
1059		if (!statp->pfcode)
1060			return (0);
1061	}
1062	if (!(statp->options & RES_IGNTC) && anhp->tc) {
1063		/*
1064		 * To get the rest of answer,
1065		 * use TCP with same server.
1066		 */
1067		Dprint(statp->options & RES_DEBUG,
1068		       (stdout, ";; truncated answer\n"));
1069		*v_circuit = 1;
1070		res_nclose(statp);
1071		return (1);
1072	}
1073	/*
1074	 * All is well, or the error is fatal.  Signal that the
1075	 * next nameserver ought not be tried.
1076	 */
1077	return (resplen);
1078}
1079
1080static void
1081Aerror(const res_state statp, FILE *file, const char *string, int error,
1082       const struct sockaddr *address, int alen)
1083{
1084	int save = errno;
1085	char hbuf[NI_MAXHOST];
1086	char sbuf[NI_MAXSERV];
1087
1088	alen = alen;
1089
1090	if ((statp->options & RES_DEBUG) != 0U) {
1091		if (getnameinfo(address, alen, hbuf, sizeof(hbuf),
1092		    sbuf, sizeof(sbuf), niflags)) {
1093			strncpy(hbuf, "?", sizeof(hbuf) - 1);
1094			hbuf[sizeof(hbuf) - 1] = '\0';
1095			strncpy(sbuf, "?", sizeof(sbuf) - 1);
1096			sbuf[sizeof(sbuf) - 1] = '\0';
1097		}
1098		fprintf(file, "res_send: %s ([%s].%s): %s\n",
1099			string, hbuf, sbuf, strerror(error));
1100	}
1101	errno = save;
1102}
1103
1104static void
1105Perror(const res_state statp, FILE *file, const char *string, int error) {
1106	int save = errno;
1107
1108	if ((statp->options & RES_DEBUG) != 0U)
1109		fprintf(file, "res_send: %s: %s\n",
1110			string, strerror(error));
1111	errno = save;
1112}
1113
1114static int
1115sock_eq(struct sockaddr *a, struct sockaddr *b) {
1116	struct sockaddr_in *a4, *b4;
1117	struct sockaddr_in6 *a6, *b6;
1118
1119	if (a->sa_family != b->sa_family)
1120		return 0;
1121	switch (a->sa_family) {
1122	case AF_INET:
1123		a4 = (struct sockaddr_in *)a;
1124		b4 = (struct sockaddr_in *)b;
1125		return a4->sin_port == b4->sin_port &&
1126		    a4->sin_addr.s_addr == b4->sin_addr.s_addr;
1127	case AF_INET6:
1128		a6 = (struct sockaddr_in6 *)a;
1129		b6 = (struct sockaddr_in6 *)b;
1130		return a6->sin6_port == b6->sin6_port &&
1131#ifdef HAVE_SIN6_SCOPE_ID
1132		    a6->sin6_scope_id == b6->sin6_scope_id &&
1133#endif
1134		    IN6_ARE_ADDR_EQUAL(&a6->sin6_addr, &b6->sin6_addr);
1135	default:
1136		return 0;
1137	}
1138}
1139
1140#if defined(NEED_PSELECT) && !defined(USE_POLL) && !defined(USE_KQUEUE)
1141/* XXX needs to move to the porting library. */
1142static int
1143pselect(int nfds, void *rfds, void *wfds, void *efds,
1144	struct timespec *tsp, const sigset_t *sigmask)
1145{
1146	struct timeval tv, *tvp;
1147	sigset_t sigs;
1148	int n;
1149
1150	if (tsp) {
1151		tvp = &tv;
1152		tv = evTimeVal(*tsp);
1153	} else
1154		tvp = NULL;
1155	if (sigmask)
1156		sigprocmask(SIG_SETMASK, sigmask, &sigs);
1157	n = select(nfds, rfds, wfds, efds, tvp);
1158	if (sigmask)
1159		sigprocmask(SIG_SETMASK, &sigs, NULL);
1160	if (tsp)
1161		*tsp = evTimeSpec(tv);
1162	return (n);
1163}
1164#endif
1165