1/*	$KAME: name6.c,v 1.25 2000/06/26 16:44:40 itojun Exp $	*/
2
3/*
4 * Copyright (C) 1995, 1996, 1997, 1998, and 1999 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 * ++Copyright++ 1985, 1988, 1993
33 * -
34 * Copyright (c) 1985, 1988, 1993
35 *    The Regents of the University of California.  All rights reserved.
36 *
37 * Redistribution and use in source and binary forms, with or without
38 * modification, are permitted provided that the following conditions
39 * are met:
40 * 1. Redistributions of source code must retain the above copyright
41 *    notice, this list of conditions and the following disclaimer.
42 * 2. Redistributions in binary form must reproduce the above copyright
43 *    notice, this list of conditions and the following disclaimer in the
44 *    documentation and/or other materials provided with the distribution.
45 * 3. Neither the name of the University nor the names of its contributors
46 *    may be used to endorse or promote products derived from this software
47 *    without specific prior written permission.
48 *
49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59 * SUCH DAMAGE.
60 * -
61 * Portions Copyright (c) 1993 by Digital Equipment Corporation.
62 *
63 * Permission to use, copy, modify, and distribute this software for any
64 * purpose with or without fee is hereby granted, provided that the above
65 * copyright notice and this permission notice appear in all copies, and that
66 * the name of Digital Equipment Corporation not be used in advertising or
67 * publicity pertaining to distribution of the document or software without
68 * specific, written prior permission.
69 *
70 * THE SOFTWARE IS PROVIDED "AS IS" AND DIGITAL EQUIPMENT CORP. DISCLAIMS ALL
71 * WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES
72 * OF MERCHANTABILITY AND FITNESS.   IN NO EVENT SHALL DIGITAL EQUIPMENT
73 * CORPORATION BE LIABLE FOR ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL
74 * DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR
75 * PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
76 * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS
77 * SOFTWARE.
78 * -
79 * --Copyright--
80 */
81
82/*
83 *	Atsushi Onoe <onoe@sm.sony.co.jp>
84 */
85
86#include <sys/cdefs.h>
87__FBSDID("$FreeBSD$");
88
89#include "namespace.h"
90#include <sys/param.h>
91#include <sys/socket.h>
92#include <sys/time.h>
93#include <sys/queue.h>
94#include <netinet/in.h>
95#ifdef INET6
96#include <net/if.h>
97#include <sys/sysctl.h>
98#include <sys/ioctl.h>
99#include <netinet6/in6_var.h>	/* XXX */
100#endif
101
102#include <arpa/inet.h>
103#include <arpa/nameser.h>
104
105#include <errno.h>
106#include <netdb.h>
107#include <resolv.h>
108#include <stdio.h>
109#include <stdlib.h>
110#include <string.h>
111#include <stdarg.h>
112#include <nsswitch.h>
113#include <unistd.h>
114#include "un-namespace.h"
115#include "netdb_private.h"
116#include "res_private.h"
117
118#ifndef MAXALIASES
119#define	MAXALIASES	10
120#endif
121#ifndef	MAXADDRS
122#define	MAXADDRS	20
123#endif
124#ifndef MAXDNAME
125#define	MAXDNAME	1025
126#endif
127
128#ifdef INET6
129#define	ADDRLEN(af)	((af) == AF_INET6 ? sizeof(struct in6_addr) : \
130					    sizeof(struct in_addr))
131#else
132#define	ADDRLEN(af)	sizeof(struct in_addr)
133#endif
134
135#define	MAPADDR(ab, ina) \
136do {									\
137	memcpy(&(ab)->map_inaddr, ina, sizeof(struct in_addr));		\
138	memset((ab)->map_zero, 0, sizeof((ab)->map_zero));		\
139	memset((ab)->map_one, 0xff, sizeof((ab)->map_one));		\
140} while (0)
141#define	MAPADDRENABLED(flags) \
142	(((flags) & AI_V4MAPPED) || \
143	 (((flags) & AI_V4MAPPED_CFG)))
144
145union inx_addr {
146	struct in_addr	in_addr;
147#ifdef INET6
148	struct in6_addr	in6_addr;
149#endif
150	struct {
151		u_char	mau_zero[10];
152		u_char	mau_one[2];
153		struct in_addr mau_inaddr;
154	}		map_addr_un;
155#define	map_zero	map_addr_un.mau_zero
156#define	map_one		map_addr_un.mau_one
157#define	map_inaddr	map_addr_un.mau_inaddr
158};
159
160struct policyqueue {
161	TAILQ_ENTRY(policyqueue) pc_entry;
162#ifdef INET6
163	struct in6_addrpolicy pc_policy;
164#endif
165};
166TAILQ_HEAD(policyhead, policyqueue);
167
168#define AIO_SRCFLAG_DEPRECATED	0x1
169
170struct hp_order {
171	union {
172		struct sockaddr_storage aiou_ss;
173		struct sockaddr aiou_sa;
174	} aio_src_un;
175#define aio_srcsa aio_src_un.aiou_sa
176	u_int32_t aio_srcflag;
177	int aio_srcscope;
178	int aio_dstscope;
179	struct policyqueue *aio_srcpolicy;
180	struct policyqueue *aio_dstpolicy;
181	union {
182		struct sockaddr_storage aiou_ss;
183		struct sockaddr aiou_sa;
184	} aio_un;
185#define aio_sa aio_un.aiou_sa
186	int aio_matchlen;
187	char *aio_h_addr;
188	int aio_initial_sequence;
189};
190
191static struct	 hostent *_hpcopy(struct hostent *, int *);
192static struct	 hostent *_hpaddr(int, const char *, void *, int *);
193#ifdef INET6
194static struct	 hostent *_hpmerge(struct hostent *, struct hostent *, int *);
195static struct	 hostent *_hpmapv6(struct hostent *, int *);
196#endif
197static struct	 hostent *_hpsort(struct hostent *, res_state);
198
199#ifdef INET6
200static struct	 hostent *_hpreorder(struct hostent *);
201static int	 get_addrselectpolicy(struct policyhead *);
202static void	 free_addrselectpolicy(struct policyhead *);
203static struct	 policyqueue *match_addrselectpolicy(struct sockaddr *,
204	struct policyhead *);
205static void	 set_source(struct hp_order *, struct policyhead *);
206static int	 matchlen(struct sockaddr *, struct sockaddr *);
207static int	 comp_dst(const void *, const void *);
208static int	 gai_addr2scopetype(struct sockaddr *);
209#endif
210
211/*
212 * Functions defined in RFC2553
213 *	getipnodebyname, getipnodebyaddr, freehostent
214 */
215
216struct hostent *
217getipnodebyname(const char *name, int af, int flags, int *errp)
218{
219	struct hostent *hp;
220	union inx_addr addrbuf;
221	res_state statp;
222	u_long options;
223
224	switch (af) {
225	case AF_INET:
226#ifdef INET6
227	case AF_INET6:
228#endif
229		break;
230	default:
231		*errp = NO_RECOVERY;
232		return NULL;
233	}
234
235	if (flags & AI_ADDRCONFIG) {
236		int s;
237
238		if ((s = _socket(af, SOCK_DGRAM | SOCK_CLOEXEC, 0)) < 0)
239			return NULL;
240		/*
241		 * TODO:
242		 * Note that implementation dependent test for address
243		 * configuration should be done every time called
244		 * (or appropriate interval),
245		 * because addresses will be dynamically assigned or deleted.
246		 */
247		_close(s);
248	}
249
250#ifdef INET6
251	/* special case for literal address */
252	if (inet_pton(AF_INET6, name, &addrbuf) == 1) {
253		if (af != AF_INET6) {
254			*errp = HOST_NOT_FOUND;
255			return NULL;
256		}
257		return _hpaddr(af, name, &addrbuf, errp);
258	}
259#endif
260	if (inet_aton(name, (struct in_addr *)&addrbuf) == 1) {
261		if (af != AF_INET) {
262			if (MAPADDRENABLED(flags)) {
263				MAPADDR(&addrbuf, &addrbuf.in_addr);
264			} else {
265				*errp = HOST_NOT_FOUND;
266				return NULL;
267			}
268		}
269		return _hpaddr(af, name, &addrbuf, errp);
270	}
271
272
273	statp = __res_state();
274	if ((statp->options & RES_INIT) == 0) {
275		if (res_ninit(statp) < 0) {
276			*errp = NETDB_INTERNAL;
277			return NULL;
278		}
279	}
280
281	options = statp->options;
282	statp->options &= ~RES_USE_INET6;
283
284	hp = gethostbyname2(name, af);
285	hp = _hpcopy(hp, errp);
286#ifdef INET6
287	if (af == AF_INET6)
288		hp = _hpreorder(hp);
289
290	if (af == AF_INET6 && ((flags & AI_ALL) || hp == NULL) &&
291	    MAPADDRENABLED(flags)) {
292		struct hostent *hp2 = gethostbyname2(name, AF_INET);
293		if (hp == NULL)
294			if (hp2 == NULL)
295				*errp = statp->res_h_errno;
296			else
297				hp = _hpmapv6(hp2, errp);
298		else {
299			if (hp2 && strcmp(hp->h_name, hp2->h_name) == 0) {
300				struct hostent *hpb = hp;
301				hp = _hpmerge(hpb, hp2, errp);
302				freehostent(hpb);
303			}
304		}
305	}
306#endif
307
308	if (hp == NULL)
309		*errp = statp->res_h_errno;
310
311	statp->options = options;
312	return _hpsort(hp, statp);
313}
314
315struct hostent *
316getipnodebyaddr(const void *src, size_t len, int af, int *errp)
317{
318	struct hostent *hp;
319	res_state statp;
320	u_long options;
321
322#ifdef INET6
323	struct in6_addr addrbuf;
324#else
325	struct in_addr addrbuf;
326#endif
327
328	switch (af) {
329	case AF_INET:
330		if (len != sizeof(struct in_addr)) {
331			*errp = NO_RECOVERY;
332			return NULL;
333		}
334		if (rounddown2((long)src, sizeof(struct in_addr))) {
335			memcpy(&addrbuf, src, len);
336			src = &addrbuf;
337		}
338		if (((struct in_addr *)src)->s_addr == 0)
339			return NULL;
340		break;
341#ifdef INET6
342	case AF_INET6:
343		if (len != sizeof(struct in6_addr)) {
344			*errp = NO_RECOVERY;
345			return NULL;
346		}
347		if (rounddown2((long)src, sizeof(struct in6_addr) / 2)) {
348			/* XXX */
349			memcpy(&addrbuf, src, len);
350			src = &addrbuf;
351		}
352		if (IN6_IS_ADDR_UNSPECIFIED((struct in6_addr *)src))
353			return NULL;
354		if (IN6_IS_ADDR_V4MAPPED((struct in6_addr *)src)
355		||  IN6_IS_ADDR_V4COMPAT((struct in6_addr *)src)) {
356			src = (char *)src +
357			    (sizeof(struct in6_addr) - sizeof(struct in_addr));
358			af = AF_INET;
359			len = sizeof(struct in_addr);
360		}
361		break;
362#endif
363	default:
364		*errp = NO_RECOVERY;
365		return NULL;
366	}
367
368	statp = __res_state();
369	if ((statp->options & RES_INIT) == 0) {
370		if (res_ninit(statp) < 0) {
371			RES_SET_H_ERRNO(statp, NETDB_INTERNAL);
372			return NULL;
373		}
374	}
375
376	options = statp->options;
377	statp->options &= ~RES_USE_INET6;
378
379	hp = gethostbyaddr(src, len, af);
380	if (hp == NULL)
381		*errp = statp->res_h_errno;
382
383	statp->options = options;
384	return (_hpcopy(hp, errp));
385}
386
387void
388freehostent(struct hostent *ptr)
389{
390	free(ptr);
391}
392
393/*
394 * Private utility functions
395 */
396
397/*
398 * _hpcopy: allocate and copy hostent structure
399 */
400static struct hostent *
401_hpcopy(struct hostent *hp, int *errp)
402{
403	struct hostent *nhp;
404	char *cp, **pp;
405	int size, addrsize;
406	int nalias = 0, naddr = 0;
407	int al_off;
408	int i;
409
410	if (hp == NULL)
411		return hp;
412
413	/* count size to be allocated */
414	size = sizeof(struct hostent);
415	if (hp->h_name != NULL)
416		size += strlen(hp->h_name) + 1;
417	if ((pp = hp->h_aliases) != NULL) {
418		for (i = 0; *pp != NULL; i++, pp++) {
419			if (**pp != '\0') {
420				size += strlen(*pp) + 1;
421				nalias++;
422			}
423		}
424	}
425	/* adjust alignment */
426	size = ALIGN(size);
427	al_off = size;
428	size += sizeof(char *) * (nalias + 1);
429	addrsize = ALIGN(hp->h_length);
430	if ((pp = hp->h_addr_list) != NULL) {
431		while (*pp++ != NULL)
432			naddr++;
433	}
434	size += addrsize * naddr;
435	size += sizeof(char *) * (naddr + 1);
436
437	/* copy */
438	if ((nhp = (struct hostent *)malloc(size)) == NULL) {
439		*errp = TRY_AGAIN;
440		return NULL;
441	}
442	cp = (char *)&nhp[1];
443	if (hp->h_name != NULL) {
444		nhp->h_name = cp;
445		strcpy(cp, hp->h_name);
446		cp += strlen(cp) + 1;
447	} else
448		nhp->h_name = NULL;
449	nhp->h_aliases = (char **)((char *)nhp + al_off);
450	if ((pp = hp->h_aliases) != NULL) {
451		for (i = 0; *pp != NULL; pp++) {
452			if (**pp != '\0') {
453				nhp->h_aliases[i++] = cp;
454				strcpy(cp, *pp);
455				cp += strlen(cp) + 1;
456			}
457		}
458	}
459	nhp->h_aliases[nalias] = NULL;
460	cp = (char *)&nhp->h_aliases[nalias + 1];
461	nhp->h_addrtype = hp->h_addrtype;
462	nhp->h_length = hp->h_length;
463	nhp->h_addr_list = (char **)cp;
464	if ((pp = hp->h_addr_list) != NULL) {
465		cp = (char *)&nhp->h_addr_list[naddr + 1];
466		for (i = 0; *pp != NULL; pp++) {
467			nhp->h_addr_list[i++] = cp;
468			memcpy(cp, *pp, hp->h_length);
469			cp += addrsize;
470		}
471	}
472	nhp->h_addr_list[naddr] = NULL;
473	return nhp;
474}
475
476/*
477 * _hpaddr: construct hostent structure with one address
478 */
479static struct hostent *
480_hpaddr(int af, const char *name, void *addr, int *errp)
481{
482	struct hostent *hp, hpbuf;
483	char *addrs[2];
484
485	hp = &hpbuf;
486	hp->h_name = (char *)name;
487	hp->h_aliases = NULL;
488	hp->h_addrtype = af;
489	hp->h_length = ADDRLEN(af);
490	hp->h_addr_list = addrs;
491	addrs[0] = (char *)addr;
492	addrs[1] = NULL;
493	return (_hpcopy(hp, errp));
494}
495
496#ifdef INET6
497/*
498 * _hpmerge: merge 2 hostent structure, arguments will be freed
499 */
500static struct hostent *
501_hpmerge(struct hostent *hp1, struct hostent *hp2, int *errp)
502{
503	int i, j;
504	int naddr, nalias;
505	char **pp;
506	struct hostent *hp, hpbuf;
507	char *aliases[MAXALIASES + 1], *addrs[MAXADDRS + 1];
508	union inx_addr addrbuf[MAXADDRS];
509
510	if (hp1 == NULL)
511		return _hpcopy(hp2, errp);
512	if (hp2 == NULL)
513		return _hpcopy(hp1, errp);
514
515#define	HP(i)	(i == 1 ? hp1 : hp2)
516	hp = &hpbuf;
517	hp->h_name = (hp1->h_name != NULL ? hp1->h_name : hp2->h_name);
518	hp->h_aliases = aliases;
519	nalias = 0;
520	for (i = 1; i <= 2; i++) {
521		if ((pp = HP(i)->h_aliases) == NULL)
522			continue;
523		for (; nalias < MAXALIASES && *pp != NULL; pp++) {
524			/* check duplicates */
525			for (j = 0; j < nalias; j++)
526				if (strcasecmp(*pp, aliases[j]) == 0)
527					break;
528			if (j == nalias)
529				aliases[nalias++] = *pp;
530		}
531	}
532	aliases[nalias] = NULL;
533	if (hp1->h_length != hp2->h_length) {
534		hp->h_addrtype = AF_INET6;
535		hp->h_length = sizeof(struct in6_addr);
536	} else {
537		hp->h_addrtype = hp1->h_addrtype;
538		hp->h_length = hp1->h_length;
539	}
540
541	hp->h_addr_list = addrs;
542	naddr = 0;
543	for (i = 1; i <= 2; i++) {
544		if ((pp = HP(i)->h_addr_list) == NULL)
545			continue;
546		if (HP(i)->h_length == hp->h_length) {
547			while (naddr < MAXADDRS && *pp != NULL)
548				addrs[naddr++] = *pp++;
549		} else {
550			/* copy IPv4 addr as mapped IPv6 addr */
551			while (naddr < MAXADDRS && *pp != NULL) {
552				MAPADDR(&addrbuf[naddr], *pp++);
553				addrs[naddr] = (char *)&addrbuf[naddr];
554				naddr++;
555			}
556		}
557	}
558	addrs[naddr] = NULL;
559	return (_hpcopy(hp, errp));
560}
561#endif
562
563/*
564 * _hpmapv6: convert IPv4 hostent into IPv4-mapped IPv6 addresses
565 */
566#ifdef INET6
567static struct hostent *
568_hpmapv6(struct hostent *hp, int *errp)
569{
570	struct hostent hp6;
571
572	if (hp == NULL)
573		return NULL;
574	if (hp->h_addrtype == AF_INET6)
575		return _hpcopy(hp, errp);
576
577	memset(&hp6, 0, sizeof(struct hostent));
578	hp6.h_addrtype = AF_INET6;
579	hp6.h_length = sizeof(struct in6_addr);
580	return _hpmerge(&hp6, hp, errp);
581}
582#endif
583
584/*
585 * _hpsort: sort address by sortlist
586 */
587static struct hostent *
588_hpsort(struct hostent *hp, res_state statp)
589{
590	int i, j, n;
591	u_char *ap, *sp, *mp, **pp;
592	char t;
593	char order[MAXADDRS];
594	int nsort = statp->nsort;
595
596	if (hp == NULL || hp->h_addr_list[1] == NULL || nsort == 0)
597		return hp;
598	for (i = 0; (ap = (u_char *)hp->h_addr_list[i]); i++) {
599		for (j = 0; j < nsort; j++) {
600#ifdef INET6
601			if (statp->_u._ext.ext->sort_list[j].af !=
602			    hp->h_addrtype)
603				continue;
604			sp = (u_char *)&statp->_u._ext.ext->sort_list[j].addr;
605			mp = (u_char *)&statp->_u._ext.ext->sort_list[j].mask;
606#else
607			sp = (u_char *)&statp->sort_list[j].addr;
608			mp = (u_char *)&statp->sort_list[j].mask;
609#endif
610			for (n = 0; n < hp->h_length; n++) {
611				if ((ap[n] & mp[n]) != sp[n])
612					break;
613			}
614			if (n == hp->h_length)
615				break;
616		}
617		order[i] = j;
618	}
619	n = i;
620	pp = (u_char **)hp->h_addr_list;
621	for (i = 0; i < n - 1; i++) {
622		for (j = i + 1; j < n; j++) {
623			if (order[i] > order[j]) {
624				ap = pp[i];
625				pp[i] = pp[j];
626				pp[j] = ap;
627				t = order[i];
628				order[i] = order[j];
629				order[j] = t;
630			}
631		}
632	}
633	return hp;
634}
635
636#ifdef INET6
637/*
638 * _hpreorder: sort address by default address selection
639 */
640static struct hostent *
641_hpreorder(struct hostent *hp)
642{
643	struct hp_order *aio;
644	int i, n;
645	char *ap;
646	struct sockaddr *sa;
647	struct policyhead policyhead;
648
649	if (hp == NULL)
650		return hp;
651
652	switch (hp->h_addrtype) {
653	case AF_INET:
654#ifdef INET6
655	case AF_INET6:
656#endif
657		break;
658	default:
659		return hp;
660	}
661
662	/* count the number of addrinfo elements for sorting. */
663	for (n = 0; hp->h_addr_list[n] != NULL; n++)
664		;
665
666	/*
667	 * If the number is small enough, we can skip the reordering process.
668	 */
669	if (n <= 1)
670		return hp;
671
672	/* allocate a temporary array for sort and initialization of it. */
673	if ((aio = malloc(sizeof(*aio) * n)) == NULL)
674		return hp;	/* give up reordering */
675	memset(aio, 0, sizeof(*aio) * n);
676
677	/* retrieve address selection policy from the kernel */
678	TAILQ_INIT(&policyhead);
679	if (!get_addrselectpolicy(&policyhead)) {
680		/* no policy is installed into kernel, we don't sort. */
681		free(aio);
682		return hp;
683	}
684
685	for (i = 0; i < n; i++) {
686		ap = hp->h_addr_list[i];
687		aio[i].aio_h_addr = ap;
688		sa = &aio[i].aio_sa;
689		switch (hp->h_addrtype) {
690		case AF_INET:
691			sa->sa_family = AF_INET;
692			sa->sa_len = sizeof(struct sockaddr_in);
693			memcpy(&((struct sockaddr_in *)sa)->sin_addr, ap,
694			    sizeof(struct in_addr));
695			break;
696#ifdef INET6
697		case AF_INET6:
698			if (IN6_IS_ADDR_V4MAPPED((struct in6_addr *)ap)) {
699				sa->sa_family = AF_INET;
700				sa->sa_len = sizeof(struct sockaddr_in);
701				memcpy(&((struct sockaddr_in *)sa)->sin_addr,
702				    &ap[12], sizeof(struct in_addr));
703			} else {
704				sa->sa_family = AF_INET6;
705				sa->sa_len = sizeof(struct sockaddr_in6);
706				memcpy(&((struct sockaddr_in6 *)sa)->sin6_addr,
707				    ap, sizeof(struct in6_addr));
708			}
709			break;
710#endif
711		}
712		aio[i].aio_dstscope = gai_addr2scopetype(sa);
713		aio[i].aio_dstpolicy = match_addrselectpolicy(sa, &policyhead);
714		set_source(&aio[i], &policyhead);
715		aio[i].aio_initial_sequence = i;
716	}
717
718	/* perform sorting. */
719	qsort(aio, n, sizeof(*aio), comp_dst);
720
721	/* reorder the h_addr_list. */
722	for (i = 0; i < n; i++)
723		hp->h_addr_list[i] = aio[i].aio_h_addr;
724
725	/* cleanup and return */
726	free(aio);
727	free_addrselectpolicy(&policyhead);
728	return hp;
729}
730
731static int
732get_addrselectpolicy(struct policyhead *head)
733{
734#ifdef INET6
735	int mib[] = { CTL_NET, PF_INET6, IPPROTO_IPV6, IPV6CTL_ADDRCTLPOLICY };
736	size_t l;
737	char *buf;
738	struct in6_addrpolicy *pol, *ep;
739
740	if (sysctl(mib, nitems(mib), NULL, &l, NULL, 0) < 0)
741		return (0);
742	if ((buf = malloc(l)) == NULL)
743		return (0);
744	if (sysctl(mib, nitems(mib), buf, &l, NULL, 0) < 0) {
745		free(buf);
746		return (0);
747	}
748
749	ep = (struct in6_addrpolicy *)(buf + l);
750	for (pol = (struct in6_addrpolicy *)buf; pol + 1 <= ep; pol++) {
751		struct policyqueue *new;
752
753		if ((new = malloc(sizeof(*new))) == NULL) {
754			free_addrselectpolicy(head); /* make the list empty */
755			break;
756		}
757		new->pc_policy = *pol;
758		TAILQ_INSERT_TAIL(head, new, pc_entry);
759	}
760
761	free(buf);
762	return (1);
763#else
764	return (0);
765#endif
766}
767
768static void
769free_addrselectpolicy(struct policyhead *head)
770{
771	struct policyqueue *ent, *nent;
772
773	for (ent = TAILQ_FIRST(head); ent; ent = nent) {
774		nent = TAILQ_NEXT(ent, pc_entry);
775		TAILQ_REMOVE(head, ent, pc_entry);
776		free(ent);
777	}
778}
779
780static struct policyqueue *
781match_addrselectpolicy(struct sockaddr *addr, struct policyhead *head)
782{
783#ifdef INET6
784	struct policyqueue *ent, *bestent = NULL;
785	struct in6_addrpolicy *pol;
786	int matchlen, bestmatchlen = -1;
787	u_char *mp, *ep, *k, *p, m;
788	struct sockaddr_in6 key;
789
790	switch(addr->sa_family) {
791	case AF_INET6:
792		key = *(struct sockaddr_in6 *)addr;
793		break;
794	case AF_INET:
795		/* convert the address into IPv4-mapped IPv6 address. */
796		memset(&key, 0, sizeof(key));
797		key.sin6_family = AF_INET6;
798		key.sin6_len = sizeof(key);
799		_map_v4v6_address(
800		    (char *)&((struct sockaddr_in *)addr)->sin_addr,
801		    (char *)&key.sin6_addr);
802		break;
803	default:
804		return(NULL);
805	}
806
807	for (ent = TAILQ_FIRST(head); ent; ent = TAILQ_NEXT(ent, pc_entry)) {
808		pol = &ent->pc_policy;
809		matchlen = 0;
810
811		mp = (u_char *)&pol->addrmask.sin6_addr;
812		ep = mp + 16;	/* XXX: scope field? */
813		k = (u_char *)&key.sin6_addr;
814		p = (u_char *)&pol->addr.sin6_addr;
815		for (; mp < ep && *mp; mp++, k++, p++) {
816			m = *mp;
817			if ((*k & m) != *p)
818				goto next; /* not match */
819			if (m == 0xff) /* short cut for a typical case */
820				matchlen += 8;
821			else {
822				while (m >= 0x80) {
823					matchlen++;
824					m <<= 1;
825				}
826			}
827		}
828
829		/* matched.  check if this is better than the current best. */
830		if (matchlen > bestmatchlen) {
831			bestent = ent;
832			bestmatchlen = matchlen;
833		}
834
835	  next:
836		continue;
837	}
838
839	return(bestent);
840#else
841	return(NULL);
842#endif
843
844}
845
846static void
847set_source(struct hp_order *aio, struct policyhead *ph)
848{
849	struct sockaddr_storage ss = aio->aio_un.aiou_ss;
850	socklen_t srclen;
851	int s;
852
853	/* set unspec ("no source is available"), just in case */
854	aio->aio_srcsa.sa_family = AF_UNSPEC;
855	aio->aio_srcscope = -1;
856
857	switch(ss.ss_family) {
858	case AF_INET:
859		((struct sockaddr_in *)&ss)->sin_port = htons(1);
860		break;
861#ifdef INET6
862	case AF_INET6:
863		((struct sockaddr_in6 *)&ss)->sin6_port = htons(1);
864		break;
865#endif
866	default:		/* ignore unsupported AFs explicitly */
867		return;
868	}
869
870	/* open a socket to get the source address for the given dst */
871	if ((s = _socket(ss.ss_family, SOCK_DGRAM | SOCK_CLOEXEC,
872	    IPPROTO_UDP)) < 0)
873		return;		/* give up */
874	if (_connect(s, (struct sockaddr *)&ss, ss.ss_len) < 0)
875		goto cleanup;
876	srclen = ss.ss_len;
877	if (_getsockname(s, &aio->aio_srcsa, &srclen) < 0) {
878		aio->aio_srcsa.sa_family = AF_UNSPEC;
879		goto cleanup;
880	}
881	aio->aio_srcscope = gai_addr2scopetype(&aio->aio_srcsa);
882	aio->aio_srcpolicy = match_addrselectpolicy(&aio->aio_srcsa, ph);
883	aio->aio_matchlen = matchlen(&aio->aio_srcsa, (struct sockaddr *)&ss);
884#ifdef INET6
885	if (ss.ss_family == AF_INET6) {
886		struct in6_ifreq ifr6;
887		u_int32_t flags6;
888
889		memset(&ifr6, 0, sizeof(ifr6));
890		memcpy(&ifr6.ifr_addr, &ss, ss.ss_len);
891		if (_ioctl(s, SIOCGIFAFLAG_IN6, &ifr6) == 0) {
892			flags6 = ifr6.ifr_ifru.ifru_flags6;
893			if ((flags6 & IN6_IFF_DEPRECATED))
894				aio->aio_srcflag |= AIO_SRCFLAG_DEPRECATED;
895		}
896	}
897#endif
898
899  cleanup:
900	_close(s);
901	return;
902}
903
904static int
905matchlen(struct sockaddr *src, struct sockaddr *dst)
906{
907	int match = 0;
908	u_char *s, *d;
909	u_char *lim, r;
910	int addrlen;
911
912	switch (src->sa_family) {
913#ifdef INET6
914	case AF_INET6:
915		s = (u_char *)&((struct sockaddr_in6 *)src)->sin6_addr;
916		d = (u_char *)&((struct sockaddr_in6 *)dst)->sin6_addr;
917		addrlen = sizeof(struct in6_addr);
918		lim = s + addrlen;
919		break;
920#endif
921	case AF_INET:
922		s = (u_char *)&((struct sockaddr_in *)src)->sin_addr;
923		d = (u_char *)&((struct sockaddr_in *)dst)->sin_addr;
924		addrlen = sizeof(struct in_addr);
925		lim = s + addrlen;
926		break;
927	default:
928		return(0);
929	}
930
931	while (s < lim)
932		if ((r = (*d++ ^ *s++)) != 0) {
933			while ((r & 0x80) == 0) {
934				match++;
935				r <<= 1;
936			}
937			break;
938		} else
939			match += 8;
940	return(match);
941}
942
943static int
944comp_dst(const void *arg1, const void *arg2)
945{
946	const struct hp_order *dst1 = arg1, *dst2 = arg2;
947
948	/*
949	 * Rule 1: Avoid unusable destinations.
950	 * XXX: we currently do not consider if an appropriate route exists.
951	 */
952	if (dst1->aio_srcsa.sa_family != AF_UNSPEC &&
953	    dst2->aio_srcsa.sa_family == AF_UNSPEC) {
954		return(-1);
955	}
956	if (dst1->aio_srcsa.sa_family == AF_UNSPEC &&
957	    dst2->aio_srcsa.sa_family != AF_UNSPEC) {
958		return(1);
959	}
960
961	/* Rule 2: Prefer matching scope. */
962	if (dst1->aio_dstscope == dst1->aio_srcscope &&
963	    dst2->aio_dstscope != dst2->aio_srcscope) {
964		return(-1);
965	}
966	if (dst1->aio_dstscope != dst1->aio_srcscope &&
967	    dst2->aio_dstscope == dst2->aio_srcscope) {
968		return(1);
969	}
970
971	/* Rule 3: Avoid deprecated addresses. */
972	if (dst1->aio_srcsa.sa_family != AF_UNSPEC &&
973	    dst2->aio_srcsa.sa_family != AF_UNSPEC) {
974		if (!(dst1->aio_srcflag & AIO_SRCFLAG_DEPRECATED) &&
975		    (dst2->aio_srcflag & AIO_SRCFLAG_DEPRECATED)) {
976			return(-1);
977		}
978		if ((dst1->aio_srcflag & AIO_SRCFLAG_DEPRECATED) &&
979		    !(dst2->aio_srcflag & AIO_SRCFLAG_DEPRECATED)) {
980			return(1);
981		}
982	}
983
984	/* Rule 4: Prefer home addresses. */
985	/* XXX: not implemented yet */
986
987	/* Rule 5: Prefer matching label. */
988#ifdef INET6
989	if (dst1->aio_srcpolicy && dst1->aio_dstpolicy &&
990	    dst1->aio_srcpolicy->pc_policy.label ==
991	    dst1->aio_dstpolicy->pc_policy.label &&
992	    (dst2->aio_srcpolicy == NULL || dst2->aio_dstpolicy == NULL ||
993	     dst2->aio_srcpolicy->pc_policy.label !=
994	     dst2->aio_dstpolicy->pc_policy.label)) {
995		return(-1);
996	}
997	if (dst2->aio_srcpolicy && dst2->aio_dstpolicy &&
998	    dst2->aio_srcpolicy->pc_policy.label ==
999	    dst2->aio_dstpolicy->pc_policy.label &&
1000	    (dst1->aio_srcpolicy == NULL || dst1->aio_dstpolicy == NULL ||
1001	     dst1->aio_srcpolicy->pc_policy.label !=
1002	     dst1->aio_dstpolicy->pc_policy.label)) {
1003		return(1);
1004	}
1005#endif
1006
1007	/* Rule 6: Prefer higher precedence. */
1008#ifdef INET6
1009	if (dst1->aio_dstpolicy &&
1010	    (dst2->aio_dstpolicy == NULL ||
1011	     dst1->aio_dstpolicy->pc_policy.preced >
1012	     dst2->aio_dstpolicy->pc_policy.preced)) {
1013		return(-1);
1014	}
1015	if (dst2->aio_dstpolicy &&
1016	    (dst1->aio_dstpolicy == NULL ||
1017	     dst2->aio_dstpolicy->pc_policy.preced >
1018	     dst1->aio_dstpolicy->pc_policy.preced)) {
1019		return(1);
1020	}
1021#endif
1022
1023	/* Rule 7: Prefer native transport. */
1024	/* XXX: not implemented yet */
1025
1026	/* Rule 8: Prefer smaller scope. */
1027	if (dst1->aio_dstscope >= 0 &&
1028	    dst1->aio_dstscope < dst2->aio_dstscope) {
1029		return(-1);
1030	}
1031	if (dst2->aio_dstscope >= 0 &&
1032	    dst2->aio_dstscope < dst1->aio_dstscope) {
1033		return(1);
1034	}
1035
1036	/*
1037	 * Rule 9: Use longest matching prefix.
1038	 * We compare the match length in a same AF only.
1039	 */
1040	if (dst1->aio_sa.sa_family == dst2->aio_sa.sa_family) {
1041		if (dst1->aio_matchlen > dst2->aio_matchlen) {
1042			return(-1);
1043		}
1044		if (dst1->aio_matchlen < dst2->aio_matchlen) {
1045			return(1);
1046		}
1047	}
1048
1049	/* Rule 10: Otherwise, leave the order unchanged. */
1050
1051	/*
1052	 * Note that qsort is unstable; so, we can't return zero and
1053	 * expect the order to be unchanged.
1054	 * That also means we can't depend on the current position of
1055	 * dst2 being after dst1.  We must enforce the initial order
1056	 * with an explicit compare on the original position.
1057	 * The qsort specification requires that "When the same objects
1058	 * (consisting of width bytes, irrespective of their current
1059	 * positions in the array) are passed more than once to the
1060	 * comparison function, the results shall be consistent with one
1061	 * another."
1062	 * In other words, If A < B, then we must also return B > A.
1063	 */
1064	if (dst2->aio_initial_sequence < dst1->aio_initial_sequence)
1065		return(1);
1066
1067	return(-1);
1068}
1069
1070/*
1071 * Copy from scope.c.
1072 * XXX: we should standardize the functions and link them as standard
1073 * library.
1074 */
1075static int
1076gai_addr2scopetype(struct sockaddr *sa)
1077{
1078#ifdef INET6
1079	struct sockaddr_in6 *sa6;
1080#endif
1081	struct sockaddr_in *sa4;
1082
1083	switch(sa->sa_family) {
1084#ifdef INET6
1085	case AF_INET6:
1086		sa6 = (struct sockaddr_in6 *)sa;
1087		if (IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
1088			/* just use the scope field of the multicast address */
1089			return(sa6->sin6_addr.s6_addr[2] & 0x0f);
1090		}
1091		/*
1092		 * Unicast addresses: map scope type to corresponding scope
1093		 * value defined for multcast addresses.
1094		 * XXX: hardcoded scope type values are bad...
1095		 */
1096		if (IN6_IS_ADDR_LOOPBACK(&sa6->sin6_addr))
1097			return(1); /* node local scope */
1098		if (IN6_IS_ADDR_LINKLOCAL(&sa6->sin6_addr))
1099			return(2); /* link-local scope */
1100		if (IN6_IS_ADDR_SITELOCAL(&sa6->sin6_addr))
1101			return(5); /* site-local scope */
1102		return(14);	/* global scope */
1103		break;
1104#endif
1105	case AF_INET:
1106		/*
1107		 * IPv4 pseudo scoping according to RFC 3484.
1108		 */
1109		sa4 = (struct sockaddr_in *)sa;
1110		/* IPv4 autoconfiguration addresses have link-local scope. */
1111		if (((u_char *)&sa4->sin_addr)[0] == 169 &&
1112		    ((u_char *)&sa4->sin_addr)[1] == 254)
1113			return(2);
1114		/* Private addresses have site-local scope. */
1115		if (((u_char *)&sa4->sin_addr)[0] == 10 ||
1116		    (((u_char *)&sa4->sin_addr)[0] == 172 &&
1117		     (((u_char *)&sa4->sin_addr)[1] & 0xf0) == 16) ||
1118		    (((u_char *)&sa4->sin_addr)[0] == 192 &&
1119		     ((u_char *)&sa4->sin_addr)[1] == 168))
1120			return(14);	/* XXX: It should be 5 unless NAT */
1121		/* Loopback addresses have link-local scope. */
1122		if (((u_char *)&sa4->sin_addr)[0] == 127)
1123			return(2);
1124		return(14);
1125		break;
1126	default:
1127		errno = EAFNOSUPPORT; /* is this a good error? */
1128		return(-1);
1129	}
1130}
1131#endif
1132