getaddrinfo.c revision 160593
1/*	$KAME: getaddrinfo.c,v 1.15 2000/07/09 04:37:24 itojun Exp $	*/
2
3/*
4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 *    notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in the
14 *    documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the project nor the names of its contributors
16 *    may be used to endorse or promote products derived from this software
17 *    without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 */
31
32/*
33 * "#ifdef FAITH" part is local hack for supporting IPv4-v6 translator.
34 *
35 * Issues to be discussed:
36 * - Return values.  There are nonstandard return values defined and used
37 *   in the source code.  This is because RFC2553 is silent about which error
38 *   code must be returned for which situation.
39 * - freeaddrinfo(NULL).  RFC2553 is silent about it.  XNET 5.2 says it is
40 *   invalid.  current code - SEGV on freeaddrinfo(NULL)
41 *
42 * Note:
43 * - The code filters out AFs that are not supported by the kernel,
44 *   when globbing NULL hostname (to loopback, or wildcard).  Is it the right
45 *   thing to do?  What is the relationship with post-RFC2553 AI_ADDRCONFIG
46 *   in ai_flags?
47 * - (post-2553) semantics of AI_ADDRCONFIG itself is too vague.
48 *   (1) what should we do against numeric hostname (2) what should we do
49 *   against NULL hostname (3) what is AI_ADDRCONFIG itself.  AF not ready?
50 *   non-loopback address configured?  global address configured?
51 *
52 * OS specific notes for freebsd4:
53 * - FreeBSD supported $GAI.  The code does not.
54 */
55
56#include <sys/cdefs.h>
57__FBSDID("$FreeBSD: head/lib/libc/net/getaddrinfo.c 160593 2006-07-23 15:31:03Z ume $");
58
59#include "namespace.h"
60#include <sys/types.h>
61#include <sys/param.h>
62#include <sys/socket.h>
63#include <net/if.h>
64#include <netinet/in.h>
65#include <sys/queue.h>
66#ifdef INET6
67#include <net/if_var.h>
68#include <sys/sysctl.h>
69#include <sys/ioctl.h>
70#include <netinet6/in6_var.h>	/* XXX */
71#endif
72#include <arpa/inet.h>
73#include <arpa/nameser.h>
74#include <rpc/rpc.h>
75#include <rpcsvc/yp_prot.h>
76#include <rpcsvc/ypclnt.h>
77#include <netdb.h>
78#include <resolv.h>
79#include <string.h>
80#include <stdlib.h>
81#include <stddef.h>
82#include <ctype.h>
83#include <unistd.h>
84#include <stdio.h>
85#include <errno.h>
86
87#include "res_config.h"
88
89#ifdef DEBUG
90#include <syslog.h>
91#endif
92
93#include <stdarg.h>
94#include <nsswitch.h>
95#include "un-namespace.h"
96#include "libc_private.h"
97#ifdef NS_CACHING
98#include "nscache.h"
99#endif
100
101#if defined(__KAME__) && defined(INET6)
102# define FAITH
103#endif
104
105#define SUCCESS 0
106#define ANY 0
107#define YES 1
108#define NO  0
109
110static const char in_addrany[] = { 0, 0, 0, 0 };
111static const char in_loopback[] = { 127, 0, 0, 1 };
112#ifdef INET6
113static const char in6_addrany[] = {
114	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
115};
116static const char in6_loopback[] = {
117	0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1
118};
119#endif
120
121struct policyqueue {
122	TAILQ_ENTRY(policyqueue) pc_entry;
123#ifdef INET6
124	struct in6_addrpolicy pc_policy;
125#endif
126};
127TAILQ_HEAD(policyhead, policyqueue);
128
129static const struct afd {
130	int a_af;
131	int a_addrlen;
132	socklen_t a_socklen;
133	int a_off;
134	const char *a_addrany;
135	const char *a_loopback;
136	int a_scoped;
137} afdl [] = {
138#ifdef INET6
139#define	N_INET6 0
140	{PF_INET6, sizeof(struct in6_addr),
141	 sizeof(struct sockaddr_in6),
142	 offsetof(struct sockaddr_in6, sin6_addr),
143	 in6_addrany, in6_loopback, 1},
144#define	N_INET 1
145#else
146#define	N_INET 0
147#endif
148	{PF_INET, sizeof(struct in_addr),
149	 sizeof(struct sockaddr_in),
150	 offsetof(struct sockaddr_in, sin_addr),
151	 in_addrany, in_loopback, 0},
152	{0, 0, 0, 0, NULL, NULL, 0},
153};
154
155struct explore {
156	int e_af;
157	int e_socktype;
158	int e_protocol;
159	const char *e_protostr;
160	int e_wild;
161#define WILD_AF(ex)		((ex)->e_wild & 0x01)
162#define WILD_SOCKTYPE(ex)	((ex)->e_wild & 0x02)
163#define WILD_PROTOCOL(ex)	((ex)->e_wild & 0x04)
164};
165
166static const struct explore explore[] = {
167#if 0
168	{ PF_LOCAL, 0, ANY, ANY, NULL, 0x01 },
169#endif
170#ifdef INET6
171	{ PF_INET6, SOCK_DGRAM, IPPROTO_UDP, "udp", 0x07 },
172	{ PF_INET6, SOCK_STREAM, IPPROTO_TCP, "tcp", 0x07 },
173	{ PF_INET6, SOCK_RAW, ANY, NULL, 0x05 },
174#endif
175	{ PF_INET, SOCK_DGRAM, IPPROTO_UDP, "udp", 0x07 },
176	{ PF_INET, SOCK_STREAM, IPPROTO_TCP, "tcp", 0x07 },
177	{ PF_INET, SOCK_RAW, ANY, NULL, 0x05 },
178	{ PF_UNSPEC, SOCK_DGRAM, IPPROTO_UDP, "udp", 0x07 },
179	{ PF_UNSPEC, SOCK_STREAM, IPPROTO_TCP, "tcp", 0x07 },
180	{ PF_UNSPEC, SOCK_RAW, ANY, NULL, 0x05 },
181	{ -1, 0, 0, NULL, 0 },
182};
183
184#ifdef INET6
185#define PTON_MAX	16
186#else
187#define PTON_MAX	4
188#endif
189
190#define AIO_SRCFLAG_DEPRECATED	0x1
191
192struct ai_order {
193	union {
194		struct sockaddr_storage aiou_ss;
195		struct sockaddr aiou_sa;
196	} aio_src_un;
197#define aio_srcsa aio_src_un.aiou_sa
198	u_int32_t aio_srcflag;
199	int aio_srcscope;
200	int aio_dstscope;
201	struct policyqueue *aio_srcpolicy;
202	struct policyqueue *aio_dstpolicy;
203	struct addrinfo *aio_ai;
204	int aio_matchlen;
205};
206
207static const ns_src default_dns_files[] = {
208	{ NSSRC_FILES, 	NS_SUCCESS },
209	{ NSSRC_DNS, 	NS_SUCCESS },
210	{ 0 }
211};
212
213struct res_target {
214	struct res_target *next;
215	const char *name;	/* domain name */
216	int qclass, qtype;	/* class and type of query */
217	u_char *answer;		/* buffer to put answer */
218	int anslen;		/* size of answer buffer */
219	int n;			/* result length */
220};
221
222#define MAXPACKET	(64*1024)
223
224typedef union {
225	HEADER hdr;
226	u_char buf[MAXPACKET];
227} querybuf;
228
229static int str2number(const char *, int *);
230static int explore_null(const struct addrinfo *,
231	const char *, struct addrinfo **);
232static int explore_numeric(const struct addrinfo *, const char *,
233	const char *, struct addrinfo **, const char *);
234static int explore_numeric_scope(const struct addrinfo *, const char *,
235	const char *, struct addrinfo **);
236static int get_canonname(const struct addrinfo *,
237	struct addrinfo *, const char *);
238static struct addrinfo *get_ai(const struct addrinfo *,
239	const struct afd *, const char *);
240static int get_portmatch(const struct addrinfo *, const char *);
241static int get_port(struct addrinfo *, const char *, int);
242static const struct afd *find_afd(int);
243static int addrconfig(struct addrinfo *);
244static void set_source(struct ai_order *, struct policyhead *);
245static int comp_dst(const void *, const void *);
246#ifdef INET6
247static int ip6_str2scopeid(char *, struct sockaddr_in6 *, u_int32_t *);
248#endif
249static int gai_addr2scopetype(struct sockaddr *);
250
251static int explore_fqdn(const struct addrinfo *, const char *,
252	const char *, struct addrinfo **);
253
254static int reorder(struct addrinfo *);
255static int get_addrselectpolicy(struct policyhead *);
256static void free_addrselectpolicy(struct policyhead *);
257static struct policyqueue *match_addrselectpolicy(struct sockaddr *,
258	struct policyhead *);
259static int matchlen(struct sockaddr *, struct sockaddr *);
260
261static struct addrinfo *getanswer(const querybuf *, int, const char *, int,
262	const struct addrinfo *, res_state);
263#if defined(RESOLVSORT)
264static int addr4sort(struct addrinfo *, res_state);
265#endif
266static int _dns_getaddrinfo(void *, void *, va_list);
267static void _sethtent(FILE **);
268static void _endhtent(FILE **);
269static struct addrinfo *_gethtent(FILE **, const char *,
270	const struct addrinfo *);
271static int _files_getaddrinfo(void *, void *, va_list);
272#ifdef YP
273static struct addrinfo *_yphostent(char *, const struct addrinfo *);
274static int _yp_getaddrinfo(void *, void *, va_list);
275#endif
276#ifdef NS_CACHING
277static int addrinfo_id_func(char *, size_t *, va_list, void *);
278static int addrinfo_marshal_func(char *, size_t *, void *, va_list, void *);
279static int addrinfo_unmarshal_func(char *, size_t, void *, va_list, void *);
280#endif
281
282static int res_queryN(const char *, struct res_target *, res_state);
283static int res_searchN(const char *, struct res_target *, res_state);
284static int res_querydomainN(const char *, const char *,
285	struct res_target *, res_state);
286
287/* XXX macros that make external reference is BAD. */
288
289#define GET_AI(ai, afd, addr) \
290do { \
291	/* external reference: pai, error, and label free */ \
292	(ai) = get_ai(pai, (afd), (addr)); \
293	if ((ai) == NULL) { \
294		error = EAI_MEMORY; \
295		goto free; \
296	} \
297} while (/*CONSTCOND*/0)
298
299#define GET_PORT(ai, serv) \
300do { \
301	/* external reference: error and label free */ \
302	error = get_port((ai), (serv), 0); \
303	if (error != 0) \
304		goto free; \
305} while (/*CONSTCOND*/0)
306
307#define GET_CANONNAME(ai, str) \
308do { \
309	/* external reference: pai, error and label free */ \
310	error = get_canonname(pai, (ai), (str)); \
311	if (error != 0) \
312		goto free; \
313} while (/*CONSTCOND*/0)
314
315#define ERR(err) \
316do { \
317	/* external reference: error, and label bad */ \
318	error = (err); \
319	goto bad; \
320	/*NOTREACHED*/ \
321} while (/*CONSTCOND*/0)
322
323#define MATCH_FAMILY(x, y, w) \
324	((x) == (y) || (/*CONSTCOND*/(w) && ((x) == PF_UNSPEC || (y) == PF_UNSPEC)))
325#define MATCH(x, y, w) \
326	((x) == (y) || (/*CONSTCOND*/(w) && ((x) == ANY || (y) == ANY)))
327
328void
329freeaddrinfo(struct addrinfo *ai)
330{
331	struct addrinfo *next;
332
333	do {
334		next = ai->ai_next;
335		if (ai->ai_canonname)
336			free(ai->ai_canonname);
337		/* no need to free(ai->ai_addr) */
338		free(ai);
339		ai = next;
340	} while (ai);
341}
342
343static int
344str2number(const char *p, int *portp)
345{
346	char *ep;
347	unsigned long v;
348
349	if (*p == '\0')
350		return -1;
351	ep = NULL;
352	errno = 0;
353	v = strtoul(p, &ep, 10);
354	if (errno == 0 && ep && *ep == '\0' && v <= UINT_MAX) {
355		*portp = v;
356		return 0;
357	} else
358		return -1;
359}
360
361int
362getaddrinfo(const char *hostname, const char *servname,
363    const struct addrinfo *hints, struct addrinfo **res)
364{
365	struct addrinfo sentinel;
366	struct addrinfo *cur;
367	int error = 0;
368	struct addrinfo ai;
369	struct addrinfo ai0;
370	struct addrinfo *pai;
371	const struct explore *ex;
372	int numeric = 0;
373
374	memset(&sentinel, 0, sizeof(sentinel));
375	cur = &sentinel;
376	pai = &ai;
377	pai->ai_flags = 0;
378	pai->ai_family = PF_UNSPEC;
379	pai->ai_socktype = ANY;
380	pai->ai_protocol = ANY;
381	pai->ai_addrlen = 0;
382	pai->ai_canonname = NULL;
383	pai->ai_addr = NULL;
384	pai->ai_next = NULL;
385
386	if (hostname == NULL && servname == NULL)
387		return EAI_NONAME;
388	if (hints) {
389		/* error check for hints */
390		if (hints->ai_addrlen || hints->ai_canonname ||
391		    hints->ai_addr || hints->ai_next)
392			ERR(EAI_BADHINTS); /* xxx */
393		if (hints->ai_flags & ~AI_MASK)
394			ERR(EAI_BADFLAGS);
395		switch (hints->ai_family) {
396		case PF_UNSPEC:
397		case PF_INET:
398#ifdef INET6
399		case PF_INET6:
400#endif
401			break;
402		default:
403			ERR(EAI_FAMILY);
404		}
405		memcpy(pai, hints, sizeof(*pai));
406
407		/*
408		 * if both socktype/protocol are specified, check if they
409		 * are meaningful combination.
410		 */
411		if (pai->ai_socktype != ANY && pai->ai_protocol != ANY) {
412			for (ex = explore; ex->e_af >= 0; ex++) {
413				if (pai->ai_family != ex->e_af)
414					continue;
415				if (ex->e_socktype == ANY)
416					continue;
417				if (ex->e_protocol == ANY)
418					continue;
419				if (pai->ai_socktype == ex->e_socktype &&
420				    pai->ai_protocol != ex->e_protocol) {
421					ERR(EAI_BADHINTS);
422				}
423			}
424		}
425	}
426
427	/*
428	 * post-2553: AI_ALL and AI_V4MAPPED are effective only against
429	 * AF_INET6 query.  They need to be ignored if specified in other
430	 * occassions.
431	 */
432	switch (pai->ai_flags & (AI_ALL | AI_V4MAPPED)) {
433	case AI_V4MAPPED:
434	case AI_ALL | AI_V4MAPPED:
435		if (pai->ai_family != AF_INET6)
436			pai->ai_flags &= ~(AI_ALL | AI_V4MAPPED);
437		break;
438	case AI_ALL:
439#if 1
440		/* illegal */
441		ERR(EAI_BADFLAGS);
442#else
443		pai->ai_flags &= ~(AI_ALL | AI_V4MAPPED);
444#endif
445		break;
446	}
447
448	/*
449	 * check for special cases.  (1) numeric servname is disallowed if
450	 * socktype/protocol are left unspecified. (2) servname is disallowed
451	 * for raw and other inet{,6} sockets.
452	 */
453	if (MATCH_FAMILY(pai->ai_family, PF_INET, 1)
454#ifdef PF_INET6
455	    || MATCH_FAMILY(pai->ai_family, PF_INET6, 1)
456#endif
457	    ) {
458		ai0 = *pai;	/* backup *pai */
459
460		if (pai->ai_family == PF_UNSPEC) {
461#ifdef PF_INET6
462			pai->ai_family = PF_INET6;
463#else
464			pai->ai_family = PF_INET;
465#endif
466		}
467		error = get_portmatch(pai, servname);
468		if (error)
469			ERR(error);
470
471		*pai = ai0;
472	}
473
474	ai0 = *pai;
475
476	/* NULL hostname, or numeric hostname */
477	for (ex = explore; ex->e_af >= 0; ex++) {
478		*pai = ai0;
479
480		/* PF_UNSPEC entries are prepared for DNS queries only */
481		if (ex->e_af == PF_UNSPEC)
482			continue;
483
484		if (!MATCH_FAMILY(pai->ai_family, ex->e_af, WILD_AF(ex)))
485			continue;
486		if (!MATCH(pai->ai_socktype, ex->e_socktype, WILD_SOCKTYPE(ex)))
487			continue;
488		if (!MATCH(pai->ai_protocol, ex->e_protocol, WILD_PROTOCOL(ex)))
489			continue;
490
491		if (pai->ai_family == PF_UNSPEC)
492			pai->ai_family = ex->e_af;
493		if (pai->ai_socktype == ANY && ex->e_socktype != ANY)
494			pai->ai_socktype = ex->e_socktype;
495		if (pai->ai_protocol == ANY && ex->e_protocol != ANY)
496			pai->ai_protocol = ex->e_protocol;
497
498		if (hostname == NULL)
499			error = explore_null(pai, servname, &cur->ai_next);
500		else
501			error = explore_numeric_scope(pai, hostname, servname,
502			    &cur->ai_next);
503
504		if (error)
505			goto free;
506
507		while (cur && cur->ai_next)
508			cur = cur->ai_next;
509	}
510
511	/*
512	 * XXX
513	 * If numreic representation of AF1 can be interpreted as FQDN
514	 * representation of AF2, we need to think again about the code below.
515	 */
516	if (sentinel.ai_next) {
517		numeric = 1;
518		goto good;
519	}
520
521	if (hostname == NULL)
522		ERR(EAI_NONAME);	/* used to be EAI_NODATA */
523	if (pai->ai_flags & AI_NUMERICHOST)
524		ERR(EAI_NONAME);
525
526	if ((pai->ai_flags & AI_ADDRCONFIG) != 0 && !addrconfig(&ai0))
527		ERR(EAI_FAIL);
528
529	/*
530	 * hostname as alphabetical name.
531	 * we would like to prefer AF_INET6 than AF_INET, so we'll make a
532	 * outer loop by AFs.
533	 */
534	for (ex = explore; ex->e_af >= 0; ex++) {
535		*pai = ai0;
536
537		/* require exact match for family field */
538		if (pai->ai_family != ex->e_af)
539			continue;
540
541		if (!MATCH(pai->ai_socktype, ex->e_socktype,
542				WILD_SOCKTYPE(ex))) {
543			continue;
544		}
545		if (!MATCH(pai->ai_protocol, ex->e_protocol,
546				WILD_PROTOCOL(ex))) {
547			continue;
548		}
549
550		if (pai->ai_socktype == ANY && ex->e_socktype != ANY)
551			pai->ai_socktype = ex->e_socktype;
552		if (pai->ai_protocol == ANY && ex->e_protocol != ANY)
553			pai->ai_protocol = ex->e_protocol;
554
555		error = explore_fqdn(pai, hostname, servname,
556			&cur->ai_next);
557
558		while (cur && cur->ai_next)
559			cur = cur->ai_next;
560	}
561
562	/* XXX inhibit errors if we have the result */
563	if (sentinel.ai_next)
564		error = 0;
565
566good:
567	/*
568	 * ensure we return either:
569	 * - error == 0, non-NULL *res
570	 * - error != 0, NULL *res
571	 */
572	if (error == 0) {
573		if (sentinel.ai_next) {
574			/*
575			 * If the returned entry is for an active connection,
576			 * and the given name is not numeric, reorder the
577			 * list, so that the application would try the list
578			 * in the most efficient order.
579			 */
580			if (hints == NULL || !(hints->ai_flags & AI_PASSIVE)) {
581				if (!numeric)
582					(void)reorder(&sentinel);
583			}
584			*res = sentinel.ai_next;
585			return SUCCESS;
586		} else
587			error = EAI_FAIL;
588	}
589free:
590bad:
591	if (sentinel.ai_next)
592		freeaddrinfo(sentinel.ai_next);
593	*res = NULL;
594	return error;
595}
596
597static int
598reorder(struct addrinfo *sentinel)
599{
600	struct addrinfo *ai, **aip;
601	struct ai_order *aio;
602	int i, n;
603	struct policyhead policyhead;
604
605	/* count the number of addrinfo elements for sorting. */
606	for (n = 0, ai = sentinel->ai_next; ai != NULL; ai = ai->ai_next, n++)
607		;
608
609	/*
610	 * If the number is small enough, we can skip the reordering process.
611	 */
612	if (n <= 1)
613		return(n);
614
615	/* allocate a temporary array for sort and initialization of it. */
616	if ((aio = malloc(sizeof(*aio) * n)) == NULL)
617		return(n);	/* give up reordering */
618	memset(aio, 0, sizeof(*aio) * n);
619
620	/* retrieve address selection policy from the kernel */
621	TAILQ_INIT(&policyhead);
622	if (!get_addrselectpolicy(&policyhead)) {
623		/* no policy is installed into kernel, we don't sort. */
624		free(aio);
625		return (n);
626	}
627
628	for (i = 0, ai = sentinel->ai_next; i < n; ai = ai->ai_next, i++) {
629		aio[i].aio_ai = ai;
630		aio[i].aio_dstscope = gai_addr2scopetype(ai->ai_addr);
631		aio[i].aio_dstpolicy = match_addrselectpolicy(ai->ai_addr,
632							      &policyhead);
633		set_source(&aio[i], &policyhead);
634	}
635
636	/* perform sorting. */
637	qsort(aio, n, sizeof(*aio), comp_dst);
638
639	/* reorder the addrinfo chain. */
640	for (i = 0, aip = &sentinel->ai_next; i < n; i++) {
641		*aip = aio[i].aio_ai;
642		aip = &aio[i].aio_ai->ai_next;
643	}
644	*aip = NULL;
645
646	/* cleanup and return */
647	free(aio);
648	free_addrselectpolicy(&policyhead);
649	return(n);
650}
651
652static int
653get_addrselectpolicy(struct policyhead *head)
654{
655#ifdef INET6
656	int mib[] = { CTL_NET, PF_INET6, IPPROTO_IPV6, IPV6CTL_ADDRCTLPOLICY };
657	size_t l;
658	char *buf;
659	struct in6_addrpolicy *pol, *ep;
660
661	if (sysctl(mib, sizeof(mib) / sizeof(mib[0]), NULL, &l, NULL, 0) < 0)
662		return (0);
663	if ((buf = malloc(l)) == NULL)
664		return (0);
665	if (sysctl(mib, sizeof(mib) / sizeof(mib[0]), buf, &l, NULL, 0) < 0) {
666		free(buf);
667		return (0);
668	}
669
670	ep = (struct in6_addrpolicy *)(buf + l);
671	for (pol = (struct in6_addrpolicy *)buf; pol + 1 <= ep; pol++) {
672		struct policyqueue *new;
673
674		if ((new = malloc(sizeof(*new))) == NULL) {
675			free_addrselectpolicy(head); /* make the list empty */
676			break;
677		}
678		new->pc_policy = *pol;
679		TAILQ_INSERT_TAIL(head, new, pc_entry);
680	}
681
682	free(buf);
683	return (1);
684#else
685	return (0);
686#endif
687}
688
689static void
690free_addrselectpolicy(struct policyhead *head)
691{
692	struct policyqueue *ent, *nent;
693
694	for (ent = TAILQ_FIRST(head); ent; ent = nent) {
695		nent = TAILQ_NEXT(ent, pc_entry);
696		TAILQ_REMOVE(head, ent, pc_entry);
697		free(ent);
698	}
699}
700
701static struct policyqueue *
702match_addrselectpolicy(struct sockaddr *addr, struct policyhead *head)
703{
704#ifdef INET6
705	struct policyqueue *ent, *bestent = NULL;
706	struct in6_addrpolicy *pol;
707	int matchlen, bestmatchlen = -1;
708	u_char *mp, *ep, *k, *p, m;
709	struct sockaddr_in6 key;
710
711	switch(addr->sa_family) {
712	case AF_INET6:
713		key = *(struct sockaddr_in6 *)addr;
714		break;
715	case AF_INET:
716		/* convert the address into IPv4-mapped IPv6 address. */
717		memset(&key, 0, sizeof(key));
718		key.sin6_family = AF_INET6;
719		key.sin6_len = sizeof(key);
720		key.sin6_addr.s6_addr[10] = 0xff;
721		key.sin6_addr.s6_addr[11] = 0xff;
722		memcpy(&key.sin6_addr.s6_addr[12],
723		       &((struct sockaddr_in *)addr)->sin_addr, 4);
724		break;
725	default:
726		return(NULL);
727	}
728
729	for (ent = TAILQ_FIRST(head); ent; ent = TAILQ_NEXT(ent, pc_entry)) {
730		pol = &ent->pc_policy;
731		matchlen = 0;
732
733		mp = (u_char *)&pol->addrmask.sin6_addr;
734		ep = mp + 16;	/* XXX: scope field? */
735		k = (u_char *)&key.sin6_addr;
736		p = (u_char *)&pol->addr.sin6_addr;
737		for (; mp < ep && *mp; mp++, k++, p++) {
738			m = *mp;
739			if ((*k & m) != *p)
740				goto next; /* not match */
741			if (m == 0xff) /* short cut for a typical case */
742				matchlen += 8;
743			else {
744				while (m >= 0x80) {
745					matchlen++;
746					m <<= 1;
747				}
748			}
749		}
750
751		/* matched.  check if this is better than the current best. */
752		if (matchlen > bestmatchlen) {
753			bestent = ent;
754			bestmatchlen = matchlen;
755		}
756
757	  next:
758		continue;
759	}
760
761	return(bestent);
762#else
763	return(NULL);
764#endif
765
766}
767
768static void
769set_source(struct ai_order *aio, struct policyhead *ph)
770{
771	struct addrinfo ai = *aio->aio_ai;
772	struct sockaddr_storage ss;
773	socklen_t srclen;
774	int s;
775
776	/* set unspec ("no source is available"), just in case */
777	aio->aio_srcsa.sa_family = AF_UNSPEC;
778	aio->aio_srcscope = -1;
779
780	switch(ai.ai_family) {
781	case AF_INET:
782#ifdef INET6
783	case AF_INET6:
784#endif
785		break;
786	default:		/* ignore unsupported AFs explicitly */
787		return;
788	}
789
790	/* XXX: make a dummy addrinfo to call connect() */
791	ai.ai_socktype = SOCK_DGRAM;
792	ai.ai_protocol = IPPROTO_UDP; /* is UDP too specific? */
793	ai.ai_next = NULL;
794	memset(&ss, 0, sizeof(ss));
795	memcpy(&ss, ai.ai_addr, ai.ai_addrlen);
796	ai.ai_addr = (struct sockaddr *)&ss;
797	get_port(&ai, "1", 0);
798
799	/* open a socket to get the source address for the given dst */
800	if ((s = _socket(ai.ai_family, ai.ai_socktype, ai.ai_protocol)) < 0)
801		return;		/* give up */
802	if (_connect(s, ai.ai_addr, ai.ai_addrlen) < 0)
803		goto cleanup;
804	srclen = ai.ai_addrlen;
805	if (_getsockname(s, &aio->aio_srcsa, &srclen) < 0) {
806		aio->aio_srcsa.sa_family = AF_UNSPEC;
807		goto cleanup;
808	}
809	aio->aio_srcscope = gai_addr2scopetype(&aio->aio_srcsa);
810	aio->aio_srcpolicy = match_addrselectpolicy(&aio->aio_srcsa, ph);
811	aio->aio_matchlen = matchlen(&aio->aio_srcsa, aio->aio_ai->ai_addr);
812#ifdef INET6
813	if (ai.ai_family == AF_INET6) {
814		struct in6_ifreq ifr6;
815		u_int32_t flags6;
816
817		/* XXX: interface name should not be hardcoded */
818		strncpy(ifr6.ifr_name, "lo0", sizeof(ifr6.ifr_name));
819		memset(&ifr6, 0, sizeof(ifr6));
820		memcpy(&ifr6.ifr_addr, ai.ai_addr, ai.ai_addrlen);
821		if (_ioctl(s, SIOCGIFAFLAG_IN6, &ifr6) == 0) {
822			flags6 = ifr6.ifr_ifru.ifru_flags6;
823			if ((flags6 & IN6_IFF_DEPRECATED))
824				aio->aio_srcflag |= AIO_SRCFLAG_DEPRECATED;
825		}
826	}
827#endif
828
829  cleanup:
830	_close(s);
831	return;
832}
833
834static int
835matchlen(struct sockaddr *src, struct sockaddr *dst)
836{
837	int match = 0;
838	u_char *s, *d;
839	u_char *lim, r;
840	int addrlen;
841
842	switch (src->sa_family) {
843#ifdef INET6
844	case AF_INET6:
845		s = (u_char *)&((struct sockaddr_in6 *)src)->sin6_addr;
846		d = (u_char *)&((struct sockaddr_in6 *)dst)->sin6_addr;
847		addrlen = sizeof(struct in6_addr);
848		lim = s + addrlen;
849		break;
850#endif
851	case AF_INET:
852		s = (u_char *)&((struct sockaddr_in *)src)->sin_addr;
853		d = (u_char *)&((struct sockaddr_in *)dst)->sin_addr;
854		addrlen = sizeof(struct in_addr);
855		lim = s + addrlen;
856		break;
857	default:
858		return(0);
859	}
860
861	while (s < lim)
862		if ((r = (*d++ ^ *s++)) != 0) {
863			while (r < addrlen * 8) {
864				match++;
865				r <<= 1;
866			}
867			break;
868		} else
869			match += 8;
870	return(match);
871}
872
873static int
874comp_dst(const void *arg1, const void *arg2)
875{
876	const struct ai_order *dst1 = arg1, *dst2 = arg2;
877
878	/*
879	 * Rule 1: Avoid unusable destinations.
880	 * XXX: we currently do not consider if an appropriate route exists.
881	 */
882	if (dst1->aio_srcsa.sa_family != AF_UNSPEC &&
883	    dst2->aio_srcsa.sa_family == AF_UNSPEC) {
884		return(-1);
885	}
886	if (dst1->aio_srcsa.sa_family == AF_UNSPEC &&
887	    dst2->aio_srcsa.sa_family != AF_UNSPEC) {
888		return(1);
889	}
890
891	/* Rule 2: Prefer matching scope. */
892	if (dst1->aio_dstscope == dst1->aio_srcscope &&
893	    dst2->aio_dstscope != dst2->aio_srcscope) {
894		return(-1);
895	}
896	if (dst1->aio_dstscope != dst1->aio_srcscope &&
897	    dst2->aio_dstscope == dst2->aio_srcscope) {
898		return(1);
899	}
900
901	/* Rule 3: Avoid deprecated addresses. */
902	if (dst1->aio_srcsa.sa_family != AF_UNSPEC &&
903	    dst2->aio_srcsa.sa_family != AF_UNSPEC) {
904		if (!(dst1->aio_srcflag & AIO_SRCFLAG_DEPRECATED) &&
905		    (dst2->aio_srcflag & AIO_SRCFLAG_DEPRECATED)) {
906			return(-1);
907		}
908		if ((dst1->aio_srcflag & AIO_SRCFLAG_DEPRECATED) &&
909		    !(dst2->aio_srcflag & AIO_SRCFLAG_DEPRECATED)) {
910			return(1);
911		}
912	}
913
914	/* Rule 4: Prefer home addresses. */
915	/* XXX: not implemented yet */
916
917	/* Rule 5: Prefer matching label. */
918#ifdef INET6
919	if (dst1->aio_srcpolicy && dst1->aio_dstpolicy &&
920	    dst1->aio_srcpolicy->pc_policy.label ==
921	    dst1->aio_dstpolicy->pc_policy.label &&
922	    (dst2->aio_srcpolicy == NULL || dst2->aio_dstpolicy == NULL ||
923	     dst2->aio_srcpolicy->pc_policy.label !=
924	     dst2->aio_dstpolicy->pc_policy.label)) {
925		return(-1);
926	}
927	if (dst2->aio_srcpolicy && dst2->aio_dstpolicy &&
928	    dst2->aio_srcpolicy->pc_policy.label ==
929	    dst2->aio_dstpolicy->pc_policy.label &&
930	    (dst1->aio_srcpolicy == NULL || dst1->aio_dstpolicy == NULL ||
931	     dst1->aio_srcpolicy->pc_policy.label !=
932	     dst1->aio_dstpolicy->pc_policy.label)) {
933		return(1);
934	}
935#endif
936
937	/* Rule 6: Prefer higher precedence. */
938#ifdef INET6
939	if (dst1->aio_dstpolicy &&
940	    (dst2->aio_dstpolicy == NULL ||
941	     dst1->aio_dstpolicy->pc_policy.preced >
942	     dst2->aio_dstpolicy->pc_policy.preced)) {
943		return(-1);
944	}
945	if (dst2->aio_dstpolicy &&
946	    (dst1->aio_dstpolicy == NULL ||
947	     dst2->aio_dstpolicy->pc_policy.preced >
948	     dst1->aio_dstpolicy->pc_policy.preced)) {
949		return(1);
950	}
951#endif
952
953	/* Rule 7: Prefer native transport. */
954	/* XXX: not implemented yet */
955
956	/* Rule 8: Prefer smaller scope. */
957	if (dst1->aio_dstscope >= 0 &&
958	    dst1->aio_dstscope < dst2->aio_dstscope) {
959		return(-1);
960	}
961	if (dst2->aio_dstscope >= 0 &&
962	    dst2->aio_dstscope < dst1->aio_dstscope) {
963		return(1);
964	}
965
966	/*
967	 * Rule 9: Use longest matching prefix.
968	 * We compare the match length in a same AF only.
969	 */
970	if (dst1->aio_ai->ai_addr->sa_family ==
971	    dst2->aio_ai->ai_addr->sa_family) {
972		if (dst1->aio_matchlen > dst2->aio_matchlen) {
973			return(-1);
974		}
975		if (dst1->aio_matchlen < dst2->aio_matchlen) {
976			return(1);
977		}
978	}
979
980	/* Rule 10: Otherwise, leave the order unchanged. */
981	return(-1);
982}
983
984/*
985 * Copy from scope.c.
986 * XXX: we should standardize the functions and link them as standard
987 * library.
988 */
989static int
990gai_addr2scopetype(struct sockaddr *sa)
991{
992#ifdef INET6
993	struct sockaddr_in6 *sa6;
994#endif
995	struct sockaddr_in *sa4;
996
997	switch(sa->sa_family) {
998#ifdef INET6
999	case AF_INET6:
1000		sa6 = (struct sockaddr_in6 *)sa;
1001		if (IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
1002			/* just use the scope field of the multicast address */
1003			return(sa6->sin6_addr.s6_addr[2] & 0x0f);
1004		}
1005		/*
1006		 * Unicast addresses: map scope type to corresponding scope
1007		 * value defined for multcast addresses.
1008		 * XXX: hardcoded scope type values are bad...
1009		 */
1010		if (IN6_IS_ADDR_LOOPBACK(&sa6->sin6_addr))
1011			return(1); /* node local scope */
1012		if (IN6_IS_ADDR_LINKLOCAL(&sa6->sin6_addr))
1013			return(2); /* link-local scope */
1014		if (IN6_IS_ADDR_SITELOCAL(&sa6->sin6_addr))
1015			return(5); /* site-local scope */
1016		return(14);	/* global scope */
1017		break;
1018#endif
1019	case AF_INET:
1020		/*
1021		 * IPv4 pseudo scoping according to RFC 3484.
1022		 */
1023		sa4 = (struct sockaddr_in *)sa;
1024		/* IPv4 autoconfiguration addresses have link-local scope. */
1025		if (((u_char *)&sa4->sin_addr)[0] == 169 &&
1026		    ((u_char *)&sa4->sin_addr)[1] == 254)
1027			return(2);
1028		/* Private addresses have site-local scope. */
1029		if (((u_char *)&sa4->sin_addr)[0] == 10 ||
1030		    (((u_char *)&sa4->sin_addr)[0] == 172 &&
1031		     (((u_char *)&sa4->sin_addr)[1] & 0xf0) == 16) ||
1032		    (((u_char *)&sa4->sin_addr)[0] == 192 &&
1033		     ((u_char *)&sa4->sin_addr)[1] == 168))
1034			return(14);	/* XXX: It should be 5 unless NAT */
1035		/* Loopback addresses have link-local scope. */
1036		if (((u_char *)&sa4->sin_addr)[0] == 127)
1037			return(2);
1038		return(14);
1039		break;
1040	default:
1041		errno = EAFNOSUPPORT; /* is this a good error? */
1042		return(-1);
1043	}
1044}
1045
1046/*
1047 * hostname == NULL.
1048 * passive socket -> anyaddr (0.0.0.0 or ::)
1049 * non-passive socket -> localhost (127.0.0.1 or ::1)
1050 */
1051static int
1052explore_null(const struct addrinfo *pai, const char *servname,
1053    struct addrinfo **res)
1054{
1055	int s;
1056	const struct afd *afd;
1057	struct addrinfo *ai;
1058	int error;
1059
1060	*res = NULL;
1061	ai = NULL;
1062
1063	/*
1064	 * filter out AFs that are not supported by the kernel
1065	 * XXX errno?
1066	 */
1067	s = _socket(pai->ai_family, SOCK_DGRAM, 0);
1068	if (s < 0) {
1069		if (errno != EMFILE)
1070			return 0;
1071	} else
1072		_close(s);
1073
1074	/*
1075	 * if the servname does not match socktype/protocol, ignore it.
1076	 */
1077	if (get_portmatch(pai, servname) != 0)
1078		return 0;
1079
1080	afd = find_afd(pai->ai_family);
1081	if (afd == NULL)
1082		return 0;
1083
1084	if (pai->ai_flags & AI_PASSIVE) {
1085		GET_AI(ai, afd, afd->a_addrany);
1086		GET_PORT(ai, servname);
1087	} else {
1088		GET_AI(ai, afd, afd->a_loopback);
1089		GET_PORT(ai, servname);
1090	}
1091
1092	*res = ai;
1093	return 0;
1094
1095free:
1096	if (ai != NULL)
1097		freeaddrinfo(ai);
1098	return error;
1099}
1100
1101/*
1102 * numeric hostname
1103 */
1104static int
1105explore_numeric(const struct addrinfo *pai, const char *hostname,
1106    const char *servname, struct addrinfo **res, const char *canonname)
1107{
1108	const struct afd *afd;
1109	struct addrinfo *ai;
1110	int error;
1111	char pton[PTON_MAX];
1112
1113	*res = NULL;
1114	ai = NULL;
1115
1116	/*
1117	 * if the servname does not match socktype/protocol, ignore it.
1118	 */
1119	if (get_portmatch(pai, servname) != 0)
1120		return 0;
1121
1122	afd = find_afd(pai->ai_family);
1123	if (afd == NULL)
1124		return 0;
1125
1126	switch (afd->a_af) {
1127	case AF_INET:
1128		/*
1129		 * RFC3493 requires getaddrinfo() to accept AF_INET formats
1130		 * that are accepted by inet_addr() and its family.  The
1131		 * accepted forms includes the "classful" one, which inet_pton
1132		 * does not accept.  So we need to separate the case for
1133		 * AF_INET.
1134		 */
1135		if (inet_aton(hostname, (struct in_addr *)pton) != 1)
1136			return 0;
1137		break;
1138	default:
1139		if (inet_pton(afd->a_af, hostname, pton) != 1)
1140			return 0;
1141		break;
1142	}
1143
1144	if (pai->ai_family == afd->a_af) {
1145		GET_AI(ai, afd, pton);
1146		GET_PORT(ai, servname);
1147		if ((pai->ai_flags & AI_CANONNAME)) {
1148			/*
1149			 * Set the numeric address itself as the canonical
1150			 * name, based on a clarification in RFC3493.
1151			 */
1152			GET_CANONNAME(ai, canonname);
1153		}
1154	} else {
1155		/*
1156		 * XXX: This should not happen since we already matched the AF
1157		 * by find_afd.
1158		 */
1159		ERR(EAI_FAMILY);
1160	}
1161
1162	*res = ai;
1163	return 0;
1164
1165free:
1166bad:
1167	if (ai != NULL)
1168		freeaddrinfo(ai);
1169	return error;
1170}
1171
1172/*
1173 * numeric hostname with scope
1174 */
1175static int
1176explore_numeric_scope(const struct addrinfo *pai, const char *hostname,
1177    const char *servname, struct addrinfo **res)
1178{
1179#if !defined(SCOPE_DELIMITER) || !defined(INET6)
1180	return explore_numeric(pai, hostname, servname, res, hostname);
1181#else
1182	const struct afd *afd;
1183	struct addrinfo *cur;
1184	int error;
1185	char *cp, *hostname2 = NULL, *scope, *addr;
1186	struct sockaddr_in6 *sin6;
1187
1188	/*
1189	 * if the servname does not match socktype/protocol, ignore it.
1190	 */
1191	if (get_portmatch(pai, servname) != 0)
1192		return 0;
1193
1194	afd = find_afd(pai->ai_family);
1195	if (afd == NULL)
1196		return 0;
1197
1198	if (!afd->a_scoped)
1199		return explore_numeric(pai, hostname, servname, res, hostname);
1200
1201	cp = strchr(hostname, SCOPE_DELIMITER);
1202	if (cp == NULL)
1203		return explore_numeric(pai, hostname, servname, res, hostname);
1204
1205	/*
1206	 * Handle special case of <scoped_address><delimiter><scope id>
1207	 */
1208	hostname2 = strdup(hostname);
1209	if (hostname2 == NULL)
1210		return EAI_MEMORY;
1211	/* terminate at the delimiter */
1212	hostname2[cp - hostname] = '\0';
1213	addr = hostname2;
1214	scope = cp + 1;
1215
1216	error = explore_numeric(pai, addr, servname, res, hostname);
1217	if (error == 0) {
1218		u_int32_t scopeid;
1219
1220		for (cur = *res; cur; cur = cur->ai_next) {
1221			if (cur->ai_family != AF_INET6)
1222				continue;
1223			sin6 = (struct sockaddr_in6 *)(void *)cur->ai_addr;
1224			if (ip6_str2scopeid(scope, sin6, &scopeid) == -1) {
1225				free(hostname2);
1226				return(EAI_NONAME); /* XXX: is return OK? */
1227			}
1228			sin6->sin6_scope_id = scopeid;
1229		}
1230	}
1231
1232	free(hostname2);
1233
1234	return error;
1235#endif
1236}
1237
1238static int
1239get_canonname(const struct addrinfo *pai, struct addrinfo *ai, const char *str)
1240{
1241	if ((pai->ai_flags & AI_CANONNAME) != 0) {
1242		ai->ai_canonname = strdup(str);
1243		if (ai->ai_canonname == NULL)
1244			return EAI_MEMORY;
1245	}
1246	return 0;
1247}
1248
1249static struct addrinfo *
1250get_ai(const struct addrinfo *pai, const struct afd *afd, const char *addr)
1251{
1252	char *p;
1253	struct addrinfo *ai;
1254#ifdef FAITH
1255	struct in6_addr faith_prefix;
1256	char *fp_str;
1257	int translate = 0;
1258#endif
1259
1260#ifdef FAITH
1261	/*
1262	 * Transfrom an IPv4 addr into a special IPv6 addr format for
1263	 * IPv6->IPv4 translation gateway. (only TCP is supported now)
1264	 *
1265	 * +-----------------------------------+------------+
1266	 * | faith prefix part (12 bytes)      | embedded   |
1267	 * |                                   | IPv4 addr part (4 bytes)
1268	 * +-----------------------------------+------------+
1269	 *
1270	 * faith prefix part is specified as ascii IPv6 addr format
1271	 * in environmental variable GAI.
1272	 * For FAITH to work correctly, routing to faith prefix must be
1273	 * setup toward a machine where a FAITH daemon operates.
1274	 * Also, the machine must enable some mechanizm
1275	 * (e.g. faith interface hack) to divert those packet with
1276	 * faith prefixed destination addr to user-land FAITH daemon.
1277	 */
1278	fp_str = getenv("GAI");
1279	if (fp_str && inet_pton(AF_INET6, fp_str, &faith_prefix) == 1 &&
1280	    afd->a_af == AF_INET && pai->ai_socktype == SOCK_STREAM) {
1281		u_int32_t v4a;
1282		u_int8_t v4a_top;
1283
1284		memcpy(&v4a, addr, sizeof v4a);
1285		v4a_top = v4a >> IN_CLASSA_NSHIFT;
1286		if (!IN_MULTICAST(v4a) && !IN_EXPERIMENTAL(v4a) &&
1287		    v4a_top != 0 && v4a != IN_LOOPBACKNET) {
1288			afd = &afdl[N_INET6];
1289			memcpy(&faith_prefix.s6_addr[12], addr,
1290			       sizeof(struct in_addr));
1291			translate = 1;
1292		}
1293	}
1294#endif
1295
1296	ai = (struct addrinfo *)malloc(sizeof(struct addrinfo)
1297		+ (afd->a_socklen));
1298	if (ai == NULL)
1299		return NULL;
1300
1301	memcpy(ai, pai, sizeof(struct addrinfo));
1302	ai->ai_addr = (struct sockaddr *)(void *)(ai + 1);
1303	memset(ai->ai_addr, 0, (size_t)afd->a_socklen);
1304	ai->ai_addr->sa_len = afd->a_socklen;
1305	ai->ai_addrlen = afd->a_socklen;
1306	ai->ai_addr->sa_family = ai->ai_family = afd->a_af;
1307	p = (char *)(void *)(ai->ai_addr);
1308#ifdef FAITH
1309	if (translate == 1)
1310		memcpy(p + afd->a_off, &faith_prefix, (size_t)afd->a_addrlen);
1311	else
1312#endif
1313	memcpy(p + afd->a_off, addr, (size_t)afd->a_addrlen);
1314	return ai;
1315}
1316
1317static int
1318get_portmatch(const struct addrinfo *ai, const char *servname)
1319{
1320
1321	/* get_port does not touch first argument when matchonly == 1. */
1322	/* LINTED const cast */
1323	return get_port((struct addrinfo *)ai, servname, 1);
1324}
1325
1326static int
1327get_port(struct addrinfo *ai, const char *servname, int matchonly)
1328{
1329	const char *proto;
1330	struct servent *sp;
1331	int port, error;
1332	int allownumeric;
1333
1334	if (servname == NULL)
1335		return 0;
1336	switch (ai->ai_family) {
1337	case AF_INET:
1338#ifdef AF_INET6
1339	case AF_INET6:
1340#endif
1341		break;
1342	default:
1343		return 0;
1344	}
1345
1346	switch (ai->ai_socktype) {
1347	case SOCK_RAW:
1348		return EAI_SERVICE;
1349	case SOCK_DGRAM:
1350	case SOCK_STREAM:
1351		allownumeric = 1;
1352		break;
1353	case ANY:
1354		allownumeric = 0;
1355		break;
1356	default:
1357		return EAI_SOCKTYPE;
1358	}
1359
1360	error = str2number(servname, &port);
1361	if (error == 0) {
1362		if (!allownumeric)
1363			return EAI_SERVICE;
1364		if (port < 0 || port > 65535)
1365			return EAI_SERVICE;
1366		port = htons(port);
1367	} else {
1368		if (ai->ai_flags & AI_NUMERICSERV)
1369			return EAI_NONAME;
1370		switch (ai->ai_socktype) {
1371		case SOCK_DGRAM:
1372			proto = "udp";
1373			break;
1374		case SOCK_STREAM:
1375			proto = "tcp";
1376			break;
1377		default:
1378			proto = NULL;
1379			break;
1380		}
1381
1382		if ((sp = getservbyname(servname, proto)) == NULL)
1383			return EAI_SERVICE;
1384		port = sp->s_port;
1385	}
1386
1387	if (!matchonly) {
1388		switch (ai->ai_family) {
1389		case AF_INET:
1390			((struct sockaddr_in *)(void *)
1391			    ai->ai_addr)->sin_port = port;
1392			break;
1393#ifdef INET6
1394		case AF_INET6:
1395			((struct sockaddr_in6 *)(void *)
1396			    ai->ai_addr)->sin6_port = port;
1397			break;
1398#endif
1399		}
1400	}
1401
1402	return 0;
1403}
1404
1405static const struct afd *
1406find_afd(int af)
1407{
1408	const struct afd *afd;
1409
1410	if (af == PF_UNSPEC)
1411		return NULL;
1412	for (afd = afdl; afd->a_af; afd++) {
1413		if (afd->a_af == af)
1414			return afd;
1415	}
1416	return NULL;
1417}
1418
1419/*
1420 * post-2553: AI_ADDRCONFIG check.  if we use getipnodeby* as backend, backend
1421 * will take care of it.
1422 * the semantics of AI_ADDRCONFIG is not defined well.  we are not sure
1423 * if the code is right or not.
1424 *
1425 * XXX PF_UNSPEC -> PF_INET6 + PF_INET mapping needs to be in sync with
1426 * _dns_getaddrinfo.
1427 */
1428static int
1429addrconfig(struct addrinfo *pai)
1430{
1431	int s, af;
1432
1433	/*
1434	 * TODO:
1435	 * Note that implementation dependent test for address
1436	 * configuration should be done everytime called
1437	 * (or apropriate interval),
1438	 * because addresses will be dynamically assigned or deleted.
1439	 */
1440	af = pai->ai_family;
1441	if (af == AF_UNSPEC) {
1442		if ((s = _socket(AF_INET6, SOCK_DGRAM, 0)) < 0)
1443			af = AF_INET;
1444		else {
1445			_close(s);
1446			if ((s = _socket(AF_INET, SOCK_DGRAM, 0)) < 0)
1447				af = AF_INET6;
1448			else
1449				_close(s);
1450		}
1451	}
1452	if (af != AF_UNSPEC) {
1453		if ((s = _socket(af, SOCK_DGRAM, 0)) < 0)
1454			return 0;
1455		_close(s);
1456	}
1457	pai->ai_family = af;
1458	return 1;
1459}
1460
1461#ifdef INET6
1462/* convert a string to a scope identifier. XXX: IPv6 specific */
1463static int
1464ip6_str2scopeid(char *scope, struct sockaddr_in6 *sin6, u_int32_t *scopeid)
1465{
1466	u_long lscopeid;
1467	struct in6_addr *a6;
1468	char *ep;
1469
1470	a6 = &sin6->sin6_addr;
1471
1472	/* empty scopeid portion is invalid */
1473	if (*scope == '\0')
1474		return -1;
1475
1476	if (IN6_IS_ADDR_LINKLOCAL(a6) || IN6_IS_ADDR_MC_LINKLOCAL(a6)) {
1477		/*
1478		 * We currently assume a one-to-one mapping between links
1479		 * and interfaces, so we simply use interface indices for
1480		 * like-local scopes.
1481		 */
1482		*scopeid = if_nametoindex(scope);
1483		if (*scopeid == 0)
1484			goto trynumeric;
1485		return 0;
1486	}
1487
1488	/* still unclear about literal, allow numeric only - placeholder */
1489	if (IN6_IS_ADDR_SITELOCAL(a6) || IN6_IS_ADDR_MC_SITELOCAL(a6))
1490		goto trynumeric;
1491	if (IN6_IS_ADDR_MC_ORGLOCAL(a6))
1492		goto trynumeric;
1493	else
1494		goto trynumeric;	/* global */
1495
1496	/* try to convert to a numeric id as a last resort */
1497  trynumeric:
1498	errno = 0;
1499	lscopeid = strtoul(scope, &ep, 10);
1500	*scopeid = (u_int32_t)(lscopeid & 0xffffffffUL);
1501	if (errno == 0 && ep && *ep == '\0' && *scopeid == lscopeid)
1502		return 0;
1503	else
1504		return -1;
1505}
1506#endif
1507
1508
1509#ifdef NS_CACHING
1510static int
1511addrinfo_id_func(char *buffer, size_t *buffer_size, va_list ap,
1512    void *cache_mdata)
1513{
1514	res_state statp;
1515	u_long res_options;
1516
1517	const int op_id = 0;	/* identifies the getaddrinfo for the cache */
1518	char *hostname;
1519	struct addrinfo *hints;
1520
1521	char *p;
1522	int ai_flags, ai_family, ai_socktype, ai_protocol;
1523	size_t desired_size, size;
1524
1525	statp = __res_state();
1526	res_options = statp->options & (RES_RECURSE | RES_DEFNAMES |
1527	    RES_DNSRCH | RES_NOALIASES | RES_USE_INET6);
1528
1529	hostname = va_arg(ap, char *);
1530	hints = va_arg(ap, struct addrinfo *);
1531
1532	desired_size = sizeof(res_options) + sizeof(int) + sizeof(int) * 4;
1533	if (hostname != NULL) {
1534		size = strlen(hostname);
1535		desired_size += size + 1;
1536	} else
1537		size = 0;
1538
1539	if (desired_size > *buffer_size) {
1540		*buffer_size = desired_size;
1541		return (NS_RETURN);
1542	}
1543
1544	if (hints == NULL)
1545		ai_flags = ai_family = ai_socktype = ai_protocol = 0;
1546	else {
1547		ai_flags = hints->ai_flags;
1548		ai_family = hints->ai_family;
1549		ai_socktype = hints->ai_socktype;
1550		ai_protocol = hints->ai_protocol;
1551	}
1552
1553	p = buffer;
1554	memcpy(p, &res_options, sizeof(res_options));
1555	p += sizeof(res_options);
1556
1557	memcpy(p, &op_id, sizeof(int));
1558	p += sizeof(int);
1559
1560	memcpy(p, &ai_flags, sizeof(int));
1561	p += sizeof(int);
1562
1563	memcpy(p, &ai_family, sizeof(int));
1564	p += sizeof(int);
1565
1566	memcpy(p, &ai_socktype, sizeof(int));
1567	p += sizeof(int);
1568
1569	memcpy(p, &ai_protocol, sizeof(int));
1570	p += sizeof(int);
1571
1572	if (hostname != NULL)
1573		memcpy(p, hostname, size);
1574
1575	*buffer_size = desired_size;
1576	return (NS_SUCCESS);
1577}
1578
1579static int
1580addrinfo_marshal_func(char *buffer, size_t *buffer_size, void *retval,
1581    va_list ap, void *cache_mdata)
1582{
1583	struct addrinfo	*ai, *cai;
1584	char *p;
1585	size_t desired_size, size, ai_size;
1586
1587	ai = *((struct addrinfo **)retval);
1588
1589	desired_size = sizeof(size_t);
1590	ai_size = 0;
1591	for (cai = ai; cai != NULL; cai = cai->ai_next) {
1592		desired_size += sizeof(struct addrinfo) + cai->ai_addrlen;
1593		if (cai->ai_canonname != NULL)
1594			desired_size += sizeof(size_t) +
1595			    strlen(cai->ai_canonname);
1596		++ai_size;
1597	}
1598
1599	if (desired_size > *buffer_size) {
1600		/* this assignment is here for future use */
1601		errno = ERANGE;
1602		*buffer_size = desired_size;
1603		return (NS_RETURN);
1604	}
1605
1606	memset(buffer, 0, desired_size);
1607	p = buffer;
1608
1609	memcpy(p, &ai_size, sizeof(size_t));
1610	p += sizeof(size_t);
1611	for (cai = ai; cai != NULL; cai = cai->ai_next) {
1612		memcpy(p, cai, sizeof(struct addrinfo));
1613		p += sizeof(struct addrinfo);
1614
1615		memcpy(p, cai->ai_addr, cai->ai_addrlen);
1616		p += cai->ai_addrlen;
1617
1618		if (cai->ai_canonname != NULL) {
1619			size = strlen(cai->ai_canonname);
1620			memcpy(p, &size, sizeof(size_t));
1621			p += sizeof(size_t);
1622
1623			memcpy(p, cai->ai_canonname, size);
1624			p += size;
1625		}
1626	}
1627
1628	return (NS_SUCCESS);
1629}
1630
1631static int
1632addrinfo_unmarshal_func(char *buffer, size_t buffer_size, void *retval,
1633    va_list ap, void *cache_mdata)
1634{
1635	struct addrinfo	new_ai, *result, *sentinel, *lasts;
1636
1637	char *p;
1638	size_t ai_size, ai_i, size;
1639
1640	p = buffer;
1641	memcpy(&ai_size, p, sizeof(size_t));
1642	p += sizeof(size_t);
1643
1644	result = NULL;
1645	lasts = NULL;
1646	for (ai_i = 0; ai_i < ai_size; ++ai_i) {
1647		memcpy(&new_ai, p, sizeof(struct addrinfo));
1648		p += sizeof(struct addrinfo);
1649		size = new_ai.ai_addrlen + sizeof(struct addrinfo) +
1650			_ALIGNBYTES;
1651
1652		sentinel = (struct addrinfo *)malloc(size);
1653		memset(sentinel, 0, size);
1654
1655		memcpy(sentinel, &new_ai, sizeof(struct addrinfo));
1656		sentinel->ai_addr = (struct sockaddr *)_ALIGN((char *)sentinel +
1657		    sizeof(struct addrinfo));
1658
1659		memcpy(sentinel->ai_addr, p, new_ai.ai_addrlen);
1660		p += new_ai.ai_addrlen;
1661
1662		if (new_ai.ai_canonname != NULL) {
1663			memcpy(&size, p, sizeof(size_t));
1664			p += sizeof(size_t);
1665
1666			sentinel->ai_canonname = (char *)malloc(size + 1);
1667			memset(sentinel->ai_canonname, 0, size + 1);
1668
1669			memcpy(sentinel->ai_canonname, p, size);
1670			p += size;
1671		}
1672
1673		if (result == NULL) {
1674			result = sentinel;
1675			lasts = sentinel;
1676		} else {
1677			lasts->ai_next = sentinel;
1678			lasts = sentinel;
1679		}
1680	}
1681
1682	*((struct addrinfo **)retval) = result;
1683	return (NS_SUCCESS);
1684}
1685#endif /* NS_CACHING */
1686
1687/*
1688 * FQDN hostname, DNS lookup
1689 */
1690static int
1691explore_fqdn(const struct addrinfo *pai, const char *hostname,
1692    const char *servname, struct addrinfo **res)
1693{
1694	struct addrinfo *result;
1695	struct addrinfo *cur;
1696	int error = 0;
1697
1698#ifdef NS_CACHING
1699	static const nss_cache_info cache_info =
1700	NS_COMMON_CACHE_INFO_INITIALIZER(
1701		hosts, NULL, addrinfo_id_func, addrinfo_marshal_func,
1702		addrinfo_unmarshal_func);
1703#endif
1704	static const ns_dtab dtab[] = {
1705		NS_FILES_CB(_files_getaddrinfo, NULL)
1706		{ NSSRC_DNS, _dns_getaddrinfo, NULL },	/* force -DHESIOD */
1707		NS_NIS_CB(_yp_getaddrinfo, NULL)
1708#ifdef NS_CACHING
1709		NS_CACHE_CB(&cache_info)
1710#endif
1711		{ 0 }
1712	};
1713
1714	result = NULL;
1715
1716	/*
1717	 * if the servname does not match socktype/protocol, ignore it.
1718	 */
1719	if (get_portmatch(pai, servname) != 0)
1720		return 0;
1721
1722	switch (_nsdispatch(&result, dtab, NSDB_HOSTS, "getaddrinfo",
1723			default_dns_files, hostname, pai)) {
1724	case NS_TRYAGAIN:
1725		error = EAI_AGAIN;
1726		goto free;
1727	case NS_UNAVAIL:
1728		error = EAI_FAIL;
1729		goto free;
1730	case NS_NOTFOUND:
1731		error = EAI_NONAME;
1732		goto free;
1733	case NS_SUCCESS:
1734		error = 0;
1735		for (cur = result; cur; cur = cur->ai_next) {
1736			GET_PORT(cur, servname);
1737			/* canonname should be filled already */
1738		}
1739		break;
1740	}
1741
1742	*res = result;
1743
1744	return 0;
1745
1746free:
1747	if (result)
1748		freeaddrinfo(result);
1749	return error;
1750}
1751
1752#ifdef DEBUG
1753static const char AskedForGot[] =
1754	"gethostby*.getanswer: asked for \"%s\", got \"%s\"";
1755#endif
1756
1757static struct addrinfo *
1758getanswer(const querybuf *answer, int anslen, const char *qname, int qtype,
1759    const struct addrinfo *pai, res_state res)
1760{
1761	struct addrinfo sentinel, *cur;
1762	struct addrinfo ai;
1763	const struct afd *afd;
1764	char *canonname;
1765	const HEADER *hp;
1766	const u_char *cp;
1767	int n;
1768	const u_char *eom;
1769	char *bp, *ep;
1770	int type, class, ancount, qdcount;
1771	int haveanswer, had_error;
1772	char tbuf[MAXDNAME];
1773	int (*name_ok)(const char *);
1774	char hostbuf[8*1024];
1775
1776	memset(&sentinel, 0, sizeof(sentinel));
1777	cur = &sentinel;
1778
1779	canonname = NULL;
1780	eom = answer->buf + anslen;
1781	switch (qtype) {
1782	case T_A:
1783	case T_AAAA:
1784	case T_ANY:	/*use T_ANY only for T_A/T_AAAA lookup*/
1785		name_ok = res_hnok;
1786		break;
1787	default:
1788		return (NULL);	/* XXX should be abort(); */
1789	}
1790	/*
1791	 * find first satisfactory answer
1792	 */
1793	hp = &answer->hdr;
1794	ancount = ntohs(hp->ancount);
1795	qdcount = ntohs(hp->qdcount);
1796	bp = hostbuf;
1797	ep = hostbuf + sizeof hostbuf;
1798	cp = answer->buf + HFIXEDSZ;
1799	if (qdcount != 1) {
1800		RES_SET_H_ERRNO(res, NO_RECOVERY);
1801		return (NULL);
1802	}
1803	n = dn_expand(answer->buf, eom, cp, bp, ep - bp);
1804	if ((n < 0) || !(*name_ok)(bp)) {
1805		RES_SET_H_ERRNO(res, NO_RECOVERY);
1806		return (NULL);
1807	}
1808	cp += n + QFIXEDSZ;
1809	if (qtype == T_A || qtype == T_AAAA || qtype == T_ANY) {
1810		/* res_send() has already verified that the query name is the
1811		 * same as the one we sent; this just gets the expanded name
1812		 * (i.e., with the succeeding search-domain tacked on).
1813		 */
1814		n = strlen(bp) + 1;		/* for the \0 */
1815		if (n >= MAXHOSTNAMELEN) {
1816			RES_SET_H_ERRNO(res, NO_RECOVERY);
1817			return (NULL);
1818		}
1819		canonname = bp;
1820		bp += n;
1821		/* The qname can be abbreviated, but h_name is now absolute. */
1822		qname = canonname;
1823	}
1824	haveanswer = 0;
1825	had_error = 0;
1826	while (ancount-- > 0 && cp < eom && !had_error) {
1827		n = dn_expand(answer->buf, eom, cp, bp, ep - bp);
1828		if ((n < 0) || !(*name_ok)(bp)) {
1829			had_error++;
1830			continue;
1831		}
1832		cp += n;			/* name */
1833		type = _getshort(cp);
1834 		cp += INT16SZ;			/* type */
1835		class = _getshort(cp);
1836 		cp += INT16SZ + INT32SZ;	/* class, TTL */
1837		n = _getshort(cp);
1838		cp += INT16SZ;			/* len */
1839		if (class != C_IN) {
1840			/* XXX - debug? syslog? */
1841			cp += n;
1842			continue;		/* XXX - had_error++ ? */
1843		}
1844		if ((qtype == T_A || qtype == T_AAAA || qtype == T_ANY) &&
1845		    type == T_CNAME) {
1846			n = dn_expand(answer->buf, eom, cp, tbuf, sizeof tbuf);
1847			if ((n < 0) || !(*name_ok)(tbuf)) {
1848				had_error++;
1849				continue;
1850			}
1851			cp += n;
1852			/* Get canonical name. */
1853			n = strlen(tbuf) + 1;	/* for the \0 */
1854			if (n > ep - bp || n >= MAXHOSTNAMELEN) {
1855				had_error++;
1856				continue;
1857			}
1858			strlcpy(bp, tbuf, ep - bp);
1859			canonname = bp;
1860			bp += n;
1861			continue;
1862		}
1863		if (qtype == T_ANY) {
1864			if (!(type == T_A || type == T_AAAA)) {
1865				cp += n;
1866				continue;
1867			}
1868		} else if (type != qtype) {
1869#ifdef DEBUG
1870			if (type != T_KEY && type != T_SIG)
1871				syslog(LOG_NOTICE|LOG_AUTH,
1872	       "gethostby*.getanswer: asked for \"%s %s %s\", got type \"%s\"",
1873				       qname, p_class(C_IN), p_type(qtype),
1874				       p_type(type));
1875#endif
1876			cp += n;
1877			continue;		/* XXX - had_error++ ? */
1878		}
1879		switch (type) {
1880		case T_A:
1881		case T_AAAA:
1882			if (strcasecmp(canonname, bp) != 0) {
1883#ifdef DEBUG
1884				syslog(LOG_NOTICE|LOG_AUTH,
1885				       AskedForGot, canonname, bp);
1886#endif
1887				cp += n;
1888				continue;	/* XXX - had_error++ ? */
1889			}
1890			if (type == T_A && n != INADDRSZ) {
1891				cp += n;
1892				continue;
1893			}
1894			if (type == T_AAAA && n != IN6ADDRSZ) {
1895				cp += n;
1896				continue;
1897			}
1898#ifdef FILTER_V4MAPPED
1899			if (type == T_AAAA) {
1900				struct in6_addr in6;
1901				memcpy(&in6, cp, sizeof(in6));
1902				if (IN6_IS_ADDR_V4MAPPED(&in6)) {
1903					cp += n;
1904					continue;
1905				}
1906			}
1907#endif
1908			if (!haveanswer) {
1909				int nn;
1910
1911				canonname = bp;
1912				nn = strlen(bp) + 1;	/* for the \0 */
1913				bp += nn;
1914			}
1915
1916			/* don't overwrite pai */
1917			ai = *pai;
1918			ai.ai_family = (type == T_A) ? AF_INET : AF_INET6;
1919			afd = find_afd(ai.ai_family);
1920			if (afd == NULL) {
1921				cp += n;
1922				continue;
1923			}
1924			cur->ai_next = get_ai(&ai, afd, (const char *)cp);
1925			if (cur->ai_next == NULL)
1926				had_error++;
1927			while (cur && cur->ai_next)
1928				cur = cur->ai_next;
1929			cp += n;
1930			break;
1931		default:
1932			abort();
1933		}
1934		if (!had_error)
1935			haveanswer++;
1936	}
1937	if (haveanswer) {
1938#if defined(RESOLVSORT)
1939		/*
1940		 * We support only IPv4 address for backward
1941		 * compatibility against gethostbyname(3).
1942		 */
1943		if (res->nsort && qtype == T_A) {
1944			if (addr4sort(&sentinel, res) < 0) {
1945				freeaddrinfo(sentinel.ai_next);
1946				RES_SET_H_ERRNO(res, NO_RECOVERY);
1947				return NULL;
1948			}
1949		}
1950#endif /*RESOLVSORT*/
1951		if (!canonname)
1952			(void)get_canonname(pai, sentinel.ai_next, qname);
1953		else
1954			(void)get_canonname(pai, sentinel.ai_next, canonname);
1955		RES_SET_H_ERRNO(res, NETDB_SUCCESS);
1956		return sentinel.ai_next;
1957	}
1958
1959	RES_SET_H_ERRNO(res, NO_RECOVERY);
1960	return NULL;
1961}
1962
1963#ifdef RESOLVSORT
1964struct addr_ptr {
1965	struct addrinfo *ai;
1966	int aval;
1967};
1968
1969static int
1970addr4sort(struct addrinfo *sentinel, res_state res)
1971{
1972	struct addrinfo *ai;
1973	struct addr_ptr *addrs, addr;
1974	struct sockaddr_in *sin;
1975	int naddrs, i, j;
1976	int needsort = 0;
1977
1978	if (!sentinel)
1979		return -1;
1980	naddrs = 0;
1981	for (ai = sentinel->ai_next; ai; ai = ai->ai_next)
1982		naddrs++;
1983	if (naddrs < 2)
1984		return 0;		/* We don't need sorting. */
1985	if ((addrs = malloc(sizeof(struct addr_ptr) * naddrs)) == NULL)
1986		return -1;
1987	i = 0;
1988	for (ai = sentinel->ai_next; ai; ai = ai->ai_next) {
1989		sin = (struct sockaddr_in *)ai->ai_addr;
1990		for (j = 0; (unsigned)j < res->nsort; j++) {
1991			if (res->sort_list[j].addr.s_addr ==
1992			    (sin->sin_addr.s_addr & res->sort_list[j].mask))
1993				break;
1994		}
1995		addrs[i].ai = ai;
1996		addrs[i].aval = j;
1997		if (needsort == 0 && i > 0 && j < addrs[i - 1].aval)
1998			needsort = i;
1999		i++;
2000	}
2001	if (!needsort) {
2002		free(addrs);
2003		return 0;
2004	}
2005
2006	while (needsort < naddrs) {
2007		for (j = needsort - 1; j >= 0; j--) {
2008			if (addrs[j].aval > addrs[j+1].aval) {
2009				addr = addrs[j];
2010				addrs[j] = addrs[j + 1];
2011				addrs[j + 1] = addr;
2012			} else
2013				break;
2014		}
2015		needsort++;
2016	}
2017
2018	ai = sentinel;
2019	for (i = 0; i < naddrs; ++i) {
2020		ai->ai_next = addrs[i].ai;
2021		ai = ai->ai_next;
2022	}
2023	ai->ai_next = NULL;
2024	free(addrs);
2025	return 0;
2026}
2027#endif /*RESOLVSORT*/
2028
2029/*ARGSUSED*/
2030static int
2031_dns_getaddrinfo(void *rv, void *cb_data, va_list ap)
2032{
2033	struct addrinfo *ai;
2034	querybuf *buf, *buf2;
2035	const char *hostname;
2036	const struct addrinfo *pai;
2037	struct addrinfo sentinel, *cur;
2038	struct res_target q, q2;
2039	res_state res;
2040
2041	hostname = va_arg(ap, char *);
2042	pai = va_arg(ap, const struct addrinfo *);
2043
2044	memset(&q, 0, sizeof(q));
2045	memset(&q2, 0, sizeof(q2));
2046	memset(&sentinel, 0, sizeof(sentinel));
2047	cur = &sentinel;
2048
2049	buf = malloc(sizeof(*buf));
2050	if (!buf) {
2051		RES_SET_H_ERRNO(res, NETDB_INTERNAL);
2052		return NS_NOTFOUND;
2053	}
2054	buf2 = malloc(sizeof(*buf2));
2055	if (!buf2) {
2056		free(buf);
2057		RES_SET_H_ERRNO(res, NETDB_INTERNAL);
2058		return NS_NOTFOUND;
2059	}
2060
2061	switch (pai->ai_family) {
2062	case AF_UNSPEC:
2063		q.name = hostname;
2064		q.qclass = C_IN;
2065		q.qtype = T_A;
2066		q.answer = buf->buf;
2067		q.anslen = sizeof(buf->buf);
2068		q.next = &q2;
2069		q2.name = hostname;
2070		q2.qclass = C_IN;
2071		q2.qtype = T_AAAA;
2072		q2.answer = buf2->buf;
2073		q2.anslen = sizeof(buf2->buf);
2074		break;
2075	case AF_INET:
2076		q.name = hostname;
2077		q.qclass = C_IN;
2078		q.qtype = T_A;
2079		q.answer = buf->buf;
2080		q.anslen = sizeof(buf->buf);
2081		break;
2082	case AF_INET6:
2083		q.name = hostname;
2084		q.qclass = C_IN;
2085		q.qtype = T_AAAA;
2086		q.answer = buf->buf;
2087		q.anslen = sizeof(buf->buf);
2088		break;
2089	default:
2090		free(buf);
2091		free(buf2);
2092		return NS_UNAVAIL;
2093	}
2094
2095	res = __res_state();
2096	if ((res->options & RES_INIT) == 0 && res_ninit(res) == -1) {
2097		RES_SET_H_ERRNO(res, NETDB_INTERNAL);
2098		free(buf);
2099		free(buf2);
2100		return NS_NOTFOUND;
2101	}
2102
2103	if (res_searchN(hostname, &q, res) < 0) {
2104		free(buf);
2105		free(buf2);
2106		return NS_NOTFOUND;
2107	}
2108	/* prefer IPv6 */
2109	if (q.next) {
2110		ai = getanswer(buf2, q2.n, q2.name, q2.qtype, pai, res);
2111		if (ai) {
2112			cur->ai_next = ai;
2113			while (cur && cur->ai_next)
2114				cur = cur->ai_next;
2115		}
2116	}
2117	ai = getanswer(buf, q.n, q.name, q.qtype, pai, res);
2118	if (ai)
2119		cur->ai_next = ai;
2120	free(buf);
2121	free(buf2);
2122	if (sentinel.ai_next == NULL)
2123		switch (res->res_h_errno) {
2124		case HOST_NOT_FOUND:
2125			return NS_NOTFOUND;
2126		case TRY_AGAIN:
2127			return NS_TRYAGAIN;
2128		default:
2129			return NS_UNAVAIL;
2130		}
2131	*((struct addrinfo **)rv) = sentinel.ai_next;
2132	return NS_SUCCESS;
2133}
2134
2135static void
2136_sethtent(FILE **hostf)
2137{
2138	if (!*hostf)
2139		*hostf = fopen(_PATH_HOSTS, "r");
2140	else
2141		rewind(*hostf);
2142}
2143
2144static void
2145_endhtent(FILE **hostf)
2146{
2147	if (*hostf) {
2148		(void) fclose(*hostf);
2149		*hostf = NULL;
2150	}
2151}
2152
2153static struct addrinfo *
2154_gethtent(FILE **hostf, const char *name, const struct addrinfo *pai)
2155{
2156	char *p;
2157	char *cp, *tname, *cname;
2158	struct addrinfo hints, *res0, *res;
2159	int error;
2160	const char *addr;
2161	char hostbuf[8*1024];
2162
2163	if (!*hostf && !(*hostf = fopen(_PATH_HOSTS, "r")))
2164		return (NULL);
2165again:
2166	if (!(p = fgets(hostbuf, sizeof hostbuf, *hostf)))
2167		return (NULL);
2168	if (*p == '#')
2169		goto again;
2170	cp = strpbrk(p, "#\n");
2171	if (cp != NULL)
2172		*cp = '\0';
2173	if (!(cp = strpbrk(p, " \t")))
2174		goto again;
2175	*cp++ = '\0';
2176	addr = p;
2177	cname = NULL;
2178	/* if this is not something we're looking for, skip it. */
2179	while (cp && *cp) {
2180		if (*cp == ' ' || *cp == '\t') {
2181			cp++;
2182			continue;
2183		}
2184		tname = cp;
2185		if (cname == NULL)
2186			cname = cp;
2187		if ((cp = strpbrk(cp, " \t")) != NULL)
2188			*cp++ = '\0';
2189		if (strcasecmp(name, tname) == 0)
2190			goto found;
2191	}
2192	goto again;
2193
2194found:
2195	/* we should not glob socktype/protocol here */
2196	memset(&hints, 0, sizeof(hints));
2197	hints.ai_family = pai->ai_family;
2198	hints.ai_socktype = SOCK_DGRAM;
2199	hints.ai_protocol = 0;
2200	hints.ai_flags = AI_NUMERICHOST;
2201	error = getaddrinfo(addr, "0", &hints, &res0);
2202	if (error)
2203		goto again;
2204#ifdef FILTER_V4MAPPED
2205	/* XXX should check all items in the chain */
2206	if (res0->ai_family == AF_INET6 &&
2207	    IN6_IS_ADDR_V4MAPPED(&((struct sockaddr_in6 *)res0->ai_addr)->sin6_addr)) {
2208		freeaddrinfo(res0);
2209		goto again;
2210	}
2211#endif
2212	for (res = res0; res; res = res->ai_next) {
2213		/* cover it up */
2214		res->ai_flags = pai->ai_flags;
2215		res->ai_socktype = pai->ai_socktype;
2216		res->ai_protocol = pai->ai_protocol;
2217
2218		if (pai->ai_flags & AI_CANONNAME) {
2219			if (get_canonname(pai, res, cname) != 0) {
2220				freeaddrinfo(res0);
2221				goto again;
2222			}
2223		}
2224	}
2225	return res0;
2226}
2227
2228/*ARGSUSED*/
2229static int
2230_files_getaddrinfo(void *rv, void *cb_data, va_list ap)
2231{
2232	const char *name;
2233	const struct addrinfo *pai;
2234	struct addrinfo sentinel, *cur;
2235	struct addrinfo *p;
2236	FILE *hostf = NULL;
2237
2238	name = va_arg(ap, char *);
2239	pai = va_arg(ap, struct addrinfo *);
2240
2241	memset(&sentinel, 0, sizeof(sentinel));
2242	cur = &sentinel;
2243
2244	_sethtent(&hostf);
2245	while ((p = _gethtent(&hostf, name, pai)) != NULL) {
2246		cur->ai_next = p;
2247		while (cur && cur->ai_next)
2248			cur = cur->ai_next;
2249	}
2250	_endhtent(&hostf);
2251
2252	*((struct addrinfo **)rv) = sentinel.ai_next;
2253	if (sentinel.ai_next == NULL)
2254		return NS_NOTFOUND;
2255	return NS_SUCCESS;
2256}
2257
2258#ifdef YP
2259/*ARGSUSED*/
2260static struct addrinfo *
2261_yphostent(char *line, const struct addrinfo *pai)
2262{
2263	struct addrinfo sentinel, *cur;
2264	struct addrinfo hints, *res, *res0;
2265	int error;
2266	char *p = line;
2267	const char *addr, *canonname;
2268	char *nextline;
2269	char *cp;
2270
2271	addr = canonname = NULL;
2272
2273	memset(&sentinel, 0, sizeof(sentinel));
2274	cur = &sentinel;
2275
2276nextline:
2277	/* terminate line */
2278	cp = strchr(p, '\n');
2279	if (cp) {
2280		*cp++ = '\0';
2281		nextline = cp;
2282	} else
2283		nextline = NULL;
2284
2285	cp = strpbrk(p, " \t");
2286	if (cp == NULL) {
2287		if (canonname == NULL)
2288			return (NULL);
2289		else
2290			goto done;
2291	}
2292	*cp++ = '\0';
2293
2294	addr = p;
2295
2296	while (cp && *cp) {
2297		if (*cp == ' ' || *cp == '\t') {
2298			cp++;
2299			continue;
2300		}
2301		if (!canonname)
2302			canonname = cp;
2303		if ((cp = strpbrk(cp, " \t")) != NULL)
2304			*cp++ = '\0';
2305	}
2306
2307	hints = *pai;
2308	hints.ai_flags = AI_NUMERICHOST;
2309	error = getaddrinfo(addr, NULL, &hints, &res0);
2310	if (error == 0) {
2311		for (res = res0; res; res = res->ai_next) {
2312			/* cover it up */
2313			res->ai_flags = pai->ai_flags;
2314
2315			if (pai->ai_flags & AI_CANONNAME)
2316				(void)get_canonname(pai, res, canonname);
2317		}
2318	} else
2319		res0 = NULL;
2320	if (res0) {
2321		cur->ai_next = res0;
2322		while (cur && cur->ai_next)
2323			cur = cur->ai_next;
2324	}
2325
2326	if (nextline) {
2327		p = nextline;
2328		goto nextline;
2329	}
2330
2331done:
2332	return sentinel.ai_next;
2333}
2334
2335/*ARGSUSED*/
2336static int
2337_yp_getaddrinfo(void *rv, void *cb_data, va_list ap)
2338{
2339	struct addrinfo sentinel, *cur;
2340	struct addrinfo *ai = NULL;
2341	char *ypbuf;
2342	int ypbuflen, r;
2343	const char *name;
2344	const struct addrinfo *pai;
2345	char *ypdomain;
2346
2347	if (_yp_check(&ypdomain) == 0)
2348		return NS_UNAVAIL;
2349
2350	name = va_arg(ap, char *);
2351	pai = va_arg(ap, const struct addrinfo *);
2352
2353	memset(&sentinel, 0, sizeof(sentinel));
2354	cur = &sentinel;
2355
2356	/* hosts.byname is only for IPv4 (Solaris8) */
2357	if (pai->ai_family == PF_UNSPEC || pai->ai_family == PF_INET) {
2358		r = yp_match(ypdomain, "hosts.byname", name,
2359			(int)strlen(name), &ypbuf, &ypbuflen);
2360		if (r == 0) {
2361			struct addrinfo ai4;
2362
2363			ai4 = *pai;
2364			ai4.ai_family = AF_INET;
2365			ai = _yphostent(ypbuf, &ai4);
2366			if (ai) {
2367				cur->ai_next = ai;
2368				while (cur && cur->ai_next)
2369					cur = cur->ai_next;
2370			}
2371			free(ypbuf);
2372		}
2373	}
2374
2375	/* ipnodes.byname can hold both IPv4/v6 */
2376	r = yp_match(ypdomain, "ipnodes.byname", name,
2377		(int)strlen(name), &ypbuf, &ypbuflen);
2378	if (r == 0) {
2379		ai = _yphostent(ypbuf, pai);
2380		if (ai)
2381			cur->ai_next = ai;
2382		free(ypbuf);
2383	}
2384
2385	if (sentinel.ai_next == NULL) {
2386		RES_SET_H_ERRNO(__res_state(), HOST_NOT_FOUND);
2387		return NS_NOTFOUND;
2388	}
2389	*((struct addrinfo **)rv) = sentinel.ai_next;
2390	return NS_SUCCESS;
2391}
2392#endif
2393
2394/* resolver logic */
2395
2396/*
2397 * Formulate a normal query, send, and await answer.
2398 * Returned answer is placed in supplied buffer "answer".
2399 * Perform preliminary check of answer, returning success only
2400 * if no error is indicated and the answer count is nonzero.
2401 * Return the size of the response on success, -1 on error.
2402 * Error number is left in h_errno.
2403 *
2404 * Caller must parse answer and determine whether it answers the question.
2405 */
2406static int
2407res_queryN(const char *name, struct res_target *target, res_state res)
2408{
2409	u_char *buf;
2410	HEADER *hp;
2411	int n;
2412	u_int oflags;
2413	struct res_target *t;
2414	int rcode;
2415	int ancount;
2416
2417	rcode = NOERROR;
2418	ancount = 0;
2419
2420	buf = malloc(MAXPACKET);
2421	if (!buf) {
2422		RES_SET_H_ERRNO(res, NETDB_INTERNAL);
2423		return -1;
2424	}
2425
2426	for (t = target; t; t = t->next) {
2427		int class, type;
2428		u_char *answer;
2429		int anslen;
2430
2431		hp = (HEADER *)(void *)t->answer;
2432
2433		/* make it easier... */
2434		class = t->qclass;
2435		type = t->qtype;
2436		answer = t->answer;
2437		anslen = t->anslen;
2438
2439		oflags = res->_flags;
2440
2441again:
2442		hp->rcode = NOERROR;	/* default */
2443
2444#ifdef DEBUG
2445		if (res->options & RES_DEBUG)
2446			printf(";; res_query(%s, %d, %d)\n", name, class, type);
2447#endif
2448
2449		n = res_nmkquery(res, QUERY, name, class, type, NULL, 0, NULL,
2450		    buf, MAXPACKET);
2451		if (n > 0 && (res->_flags & RES_F_EDNS0ERR) == 0 &&
2452		    (res->options & (RES_USE_EDNS0|RES_USE_DNSSEC)) != 0U)
2453			n = res_nopt(res, n, buf, MAXPACKET, anslen);
2454		if (n <= 0) {
2455#ifdef DEBUG
2456			if (res->options & RES_DEBUG)
2457				printf(";; res_query: mkquery failed\n");
2458#endif
2459			free(buf);
2460			RES_SET_H_ERRNO(res, NO_RECOVERY);
2461			return (n);
2462		}
2463		n = res_nsend(res, buf, n, answer, anslen);
2464		if (n < 0) {
2465			/*
2466			 * if the query choked with EDNS0, retry
2467			 * without EDNS0
2468			 */
2469			if ((res->options & (RES_USE_EDNS0|RES_USE_DNSSEC))
2470			    != 0U &&
2471			    ((oflags ^ res->_flags) & RES_F_EDNS0ERR) != 0) {
2472				res->_flags |= RES_F_EDNS0ERR;
2473				if (res->options & RES_DEBUG)
2474					printf(";; res_nquery: retry without EDNS0\n");
2475				goto again;
2476			}
2477			rcode = hp->rcode;	/* record most recent error */
2478#ifdef DEBUG
2479			if (res->options & RES_DEBUG)
2480				printf(";; res_query: send error\n");
2481#endif
2482			continue;
2483		}
2484
2485		if (n > anslen)
2486			hp->rcode = FORMERR; /* XXX not very informative */
2487		if (hp->rcode != NOERROR || ntohs(hp->ancount) == 0) {
2488			rcode = hp->rcode;	/* record most recent error */
2489#ifdef DEBUG
2490			if (res->options & RES_DEBUG)
2491				printf(";; rcode = %u, ancount=%u\n", hp->rcode,
2492				    ntohs(hp->ancount));
2493#endif
2494			continue;
2495		}
2496
2497		ancount += ntohs(hp->ancount);
2498
2499		t->n = n;
2500	}
2501
2502	free(buf);
2503
2504	if (ancount == 0) {
2505		switch (rcode) {
2506		case NXDOMAIN:
2507			RES_SET_H_ERRNO(res, HOST_NOT_FOUND);
2508			break;
2509		case SERVFAIL:
2510			RES_SET_H_ERRNO(res, TRY_AGAIN);
2511			break;
2512		case NOERROR:
2513			RES_SET_H_ERRNO(res, NO_DATA);
2514			break;
2515		case FORMERR:
2516		case NOTIMP:
2517		case REFUSED:
2518		default:
2519			RES_SET_H_ERRNO(res, NO_RECOVERY);
2520			break;
2521		}
2522		return (-1);
2523	}
2524	return (ancount);
2525}
2526
2527/*
2528 * Formulate a normal query, send, and retrieve answer in supplied buffer.
2529 * Return the size of the response on success, -1 on error.
2530 * If enabled, implement search rules until answer or unrecoverable failure
2531 * is detected.  Error code, if any, is left in h_errno.
2532 */
2533static int
2534res_searchN(const char *name, struct res_target *target, res_state res)
2535{
2536	const char *cp, * const *domain;
2537	HEADER *hp = (HEADER *)(void *)target->answer;	/*XXX*/
2538	u_int dots;
2539	int trailing_dot, ret, saved_herrno;
2540	int got_nodata = 0, got_servfail = 0, root_on_list = 0;
2541	int tried_as_is = 0;
2542	int searched = 0;
2543	char abuf[MAXDNAME];
2544
2545	errno = 0;
2546	RES_SET_H_ERRNO(res, HOST_NOT_FOUND); /* default, if we never query */
2547	dots = 0;
2548	for (cp = name; *cp; cp++)
2549		dots += (*cp == '.');
2550	trailing_dot = 0;
2551	if (cp > name && *--cp == '.')
2552		trailing_dot++;
2553
2554	/*
2555	 * if there aren't any dots, it could be a user-level alias
2556	 */
2557	if (!dots &&
2558	    (cp = res_hostalias(res, name, abuf, sizeof(abuf))) != NULL)
2559		return (res_queryN(cp, target, res));
2560
2561	/*
2562	 * If there are enough dots in the name, let's just give it a
2563	 * try 'as is'. The threshold can be set with the "ndots" option.
2564	 * Also, query 'as is', if there is a trailing dot in the name.
2565	 */
2566	saved_herrno = -1;
2567	if (dots >= res->ndots || trailing_dot) {
2568		ret = res_querydomainN(name, NULL, target, res);
2569		if (ret > 0 || trailing_dot)
2570			return (ret);
2571		if (errno == ECONNREFUSED) {
2572			RES_SET_H_ERRNO(res, TRY_AGAIN);
2573			return (-1);
2574		}
2575		switch (res->res_h_errno) {
2576		case NO_DATA:
2577		case HOST_NOT_FOUND:
2578			break;
2579		case TRY_AGAIN:
2580			if (hp->rcode == SERVFAIL)
2581				break;
2582			/* FALLTHROUGH */
2583		default:
2584			return (-1);
2585		}
2586		saved_herrno = res->res_h_errno;
2587		tried_as_is++;
2588	}
2589
2590	/*
2591	 * We do at least one level of search if
2592	 *	- there is no dot and RES_DEFNAME is set, or
2593	 *	- there is at least one dot, there is no trailing dot,
2594	 *	  and RES_DNSRCH is set.
2595	 */
2596	if ((!dots && (res->options & RES_DEFNAMES)) ||
2597	    (dots && !trailing_dot && (res->options & RES_DNSRCH))) {
2598		int done = 0;
2599
2600		for (domain = (const char * const *)res->dnsrch;
2601		   *domain && !done;
2602		   domain++) {
2603			searched = 1;
2604
2605			if (domain[0][0] == '\0' ||
2606			    (domain[0][0] == '.' && domain[0][1] == '\0'))
2607				root_on_list++;
2608
2609			if (root_on_list && tried_as_is)
2610				continue;
2611
2612			ret = res_querydomainN(name, *domain, target, res);
2613			if (ret > 0)
2614				return (ret);
2615
2616			/*
2617			 * If no server present, give up.
2618			 * If name isn't found in this domain,
2619			 * keep trying higher domains in the search list
2620			 * (if that's enabled).
2621			 * On a NO_DATA error, keep trying, otherwise
2622			 * a wildcard entry of another type could keep us
2623			 * from finding this entry higher in the domain.
2624			 * If we get some other error (negative answer or
2625			 * server failure), then stop searching up,
2626			 * but try the input name below in case it's
2627			 * fully-qualified.
2628			 */
2629			if (errno == ECONNREFUSED) {
2630				RES_SET_H_ERRNO(res, TRY_AGAIN);
2631				return (-1);
2632			}
2633
2634			switch (res->res_h_errno) {
2635			case NO_DATA:
2636				got_nodata++;
2637				/* FALLTHROUGH */
2638			case HOST_NOT_FOUND:
2639				/* keep trying */
2640				break;
2641			case TRY_AGAIN:
2642				got_servfail++;
2643				if (hp->rcode == SERVFAIL) {
2644					/* try next search element, if any */
2645					break;
2646				}
2647				/* FALLTHROUGH */
2648			default:
2649				/* anything else implies that we're done */
2650				done++;
2651			}
2652			/*
2653			 * if we got here for some reason other than DNSRCH,
2654			 * we only wanted one iteration of the loop, so stop.
2655			 */
2656			if (!(res->options & RES_DNSRCH))
2657			        done++;
2658		}
2659	}
2660
2661	switch (res->res_h_errno) {
2662	case NO_DATA:
2663	case HOST_NOT_FOUND:
2664		break;
2665	case TRY_AGAIN:
2666		if (hp->rcode == SERVFAIL)
2667			break;
2668		/* FALLTHROUGH */
2669	default:
2670		goto giveup;
2671	}
2672
2673	/*
2674	 * If the query has not already been tried as is then try it
2675	 * unless RES_NOTLDQUERY is set and there were no dots.
2676	 */
2677	if ((dots || !searched || !(res->options & RES_NOTLDQUERY)) &&
2678	    !(tried_as_is || root_on_list)) {
2679		ret = res_querydomainN(name, NULL, target, res);
2680		if (ret > 0)
2681			return (ret);
2682	}
2683
2684	/*
2685	 * if we got here, we didn't satisfy the search.
2686	 * if we did an initial full query, return that query's h_errno
2687	 * (note that we wouldn't be here if that query had succeeded).
2688	 * else if we ever got a nodata, send that back as the reason.
2689	 * else send back meaningless h_errno, that being the one from
2690	 * the last DNSRCH we did.
2691	 */
2692giveup:
2693	if (saved_herrno != -1)
2694		RES_SET_H_ERRNO(res, saved_herrno);
2695	else if (got_nodata)
2696		RES_SET_H_ERRNO(res, NO_DATA);
2697	else if (got_servfail)
2698		RES_SET_H_ERRNO(res, TRY_AGAIN);
2699	return (-1);
2700}
2701
2702/*
2703 * Perform a call on res_query on the concatenation of name and domain,
2704 * removing a trailing dot from name if domain is NULL.
2705 */
2706static int
2707res_querydomainN(const char *name, const char *domain,
2708    struct res_target *target, res_state res)
2709{
2710	char nbuf[MAXDNAME];
2711	const char *longname = nbuf;
2712	size_t n, d;
2713
2714#ifdef DEBUG
2715	if (res->options & RES_DEBUG)
2716		printf(";; res_querydomain(%s, %s)\n",
2717			name, domain?domain:"<Nil>");
2718#endif
2719	if (domain == NULL) {
2720		/*
2721		 * Check for trailing '.';
2722		 * copy without '.' if present.
2723		 */
2724		n = strlen(name);
2725		if (n >= MAXDNAME) {
2726			RES_SET_H_ERRNO(res, NO_RECOVERY);
2727			return (-1);
2728		}
2729		if (n > 0 && name[--n] == '.') {
2730			strncpy(nbuf, name, n);
2731			nbuf[n] = '\0';
2732		} else
2733			longname = name;
2734	} else {
2735		n = strlen(name);
2736		d = strlen(domain);
2737		if (n + d + 1 >= MAXDNAME) {
2738			RES_SET_H_ERRNO(res, NO_RECOVERY);
2739			return (-1);
2740		}
2741		snprintf(nbuf, sizeof(nbuf), "%s.%s", name, domain);
2742	}
2743	return (res_queryN(longname, target, res));
2744}
2745