1/*	$KAME: name6.c,v 1.25 2000/06/26 16:44:40 itojun Exp $	*/
2
3/*-
4 * SPDX-License-Identifier: BSD-3-Clause
5 *
6 * Copyright (C) 1995, 1996, 1997, 1998, and 1999 WIDE Project.
7 * All rights reserved.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 * 1. Redistributions of source code must retain the above copyright
13 *    notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 *    notice, this list of conditions and the following disclaimer in the
16 *    documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the project nor the names of its contributors
18 *    may be used to endorse or promote products derived from this software
19 *    without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 */
33/*
34 * ++Copyright++ 1985, 1988, 1993
35 * -
36 * Copyright (c) 1985, 1988, 1993
37 *    The Regents of the University of California.  All rights reserved.
38 *
39 * Redistribution and use in source and binary forms, with or without
40 * modification, are permitted provided that the following conditions
41 * are met:
42 * 1. Redistributions of source code must retain the above copyright
43 *    notice, this list of conditions and the following disclaimer.
44 * 2. Redistributions in binary form must reproduce the above copyright
45 *    notice, this list of conditions and the following disclaimer in the
46 *    documentation and/or other materials provided with the distribution.
47 * 3. Neither the name of the University nor the names of its contributors
48 *    may be used to endorse or promote products derived from this software
49 *    without specific prior written permission.
50 *
51 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
52 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
53 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
54 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
55 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
56 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
57 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
58 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
59 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
60 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
61 * SUCH DAMAGE.
62 * -
63 * Portions Copyright (c) 1993 by Digital Equipment Corporation.
64 *
65 * Permission to use, copy, modify, and distribute this software for any
66 * purpose with or without fee is hereby granted, provided that the above
67 * copyright notice and this permission notice appear in all copies, and that
68 * the name of Digital Equipment Corporation not be used in advertising or
69 * publicity pertaining to distribution of the document or software without
70 * specific, written prior permission.
71 *
72 * THE SOFTWARE IS PROVIDED "AS IS" AND DIGITAL EQUIPMENT CORP. DISCLAIMS ALL
73 * WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES
74 * OF MERCHANTABILITY AND FITNESS.   IN NO EVENT SHALL DIGITAL EQUIPMENT
75 * CORPORATION BE LIABLE FOR ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL
76 * DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR
77 * PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
78 * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS
79 * SOFTWARE.
80 * -
81 * --Copyright--
82 */
83
84/*
85 *	Atsushi Onoe <onoe@sm.sony.co.jp>
86 */
87
88#include "namespace.h"
89#include <sys/param.h>
90#include <sys/socket.h>
91#include <sys/time.h>
92#include <sys/queue.h>
93#include <netinet/in.h>
94#ifdef INET6
95#include <net/if.h>
96#include <sys/sysctl.h>
97#include <sys/ioctl.h>
98#include <netinet6/in6_var.h>	/* XXX */
99#endif
100
101#include <arpa/inet.h>
102#include <arpa/nameser.h>
103
104#include <errno.h>
105#include <netdb.h>
106#include <resolv.h>
107#include <stdio.h>
108#include <stdlib.h>
109#include <string.h>
110#include <stdarg.h>
111#include <nsswitch.h>
112#include <unistd.h>
113#include "un-namespace.h"
114#include "netdb_private.h"
115#include "res_private.h"
116
117#ifndef MAXALIASES
118#define	MAXALIASES	10
119#endif
120#ifndef	MAXADDRS
121#define	MAXADDRS	20
122#endif
123#ifndef MAXDNAME
124#define	MAXDNAME	1025
125#endif
126
127#ifdef INET6
128#define	ADDRLEN(af)	((af) == AF_INET6 ? sizeof(struct in6_addr) : \
129					    sizeof(struct in_addr))
130#else
131#define	ADDRLEN(af)	sizeof(struct in_addr)
132#endif
133
134#define	MAPADDR(ab, ina) \
135do {									\
136	memcpy(&(ab)->map_inaddr, ina, sizeof(struct in_addr));		\
137	memset((ab)->map_zero, 0, sizeof((ab)->map_zero));		\
138	memset((ab)->map_one, 0xff, sizeof((ab)->map_one));		\
139} while (0)
140#define	MAPADDRENABLED(flags) \
141	(((flags) & AI_V4MAPPED) || \
142	 (((flags) & AI_V4MAPPED_CFG)))
143
144union inx_addr {
145	struct in_addr	in_addr;
146#ifdef INET6
147	struct in6_addr	in6_addr;
148#endif
149	struct {
150		u_char	mau_zero[10];
151		u_char	mau_one[2];
152		struct in_addr mau_inaddr;
153	}		map_addr_un;
154#define	map_zero	map_addr_un.mau_zero
155#define	map_one		map_addr_un.mau_one
156#define	map_inaddr	map_addr_un.mau_inaddr
157};
158
159struct policyqueue {
160	TAILQ_ENTRY(policyqueue) pc_entry;
161#ifdef INET6
162	struct in6_addrpolicy pc_policy;
163#endif
164};
165TAILQ_HEAD(policyhead, policyqueue);
166
167#define AIO_SRCFLAG_DEPRECATED	0x1
168
169struct hp_order {
170	union {
171		struct sockaddr_storage aiou_ss;
172		struct sockaddr aiou_sa;
173	} aio_src_un;
174#define aio_srcsa aio_src_un.aiou_sa
175	u_int32_t aio_srcflag;
176	int aio_srcscope;
177	int aio_dstscope;
178	struct policyqueue *aio_srcpolicy;
179	struct policyqueue *aio_dstpolicy;
180	union {
181		struct sockaddr_storage aiou_ss;
182		struct sockaddr aiou_sa;
183	} aio_un;
184#define aio_sa aio_un.aiou_sa
185	int aio_matchlen;
186	char *aio_h_addr;
187	int aio_initial_sequence;
188};
189
190static struct	 hostent *_hpcopy(struct hostent *, int *);
191static struct	 hostent *_hpaddr(int, const char *, void *, int *);
192#ifdef INET6
193static struct	 hostent *_hpmerge(struct hostent *, struct hostent *, int *);
194static struct	 hostent *_hpmapv6(struct hostent *, int *);
195#endif
196static struct	 hostent *_hpsort(struct hostent *, res_state);
197
198#ifdef INET6
199static struct	 hostent *_hpreorder(struct hostent *);
200static int	 get_addrselectpolicy(struct policyhead *);
201static void	 free_addrselectpolicy(struct policyhead *);
202static struct	 policyqueue *match_addrselectpolicy(struct sockaddr *,
203	struct policyhead *);
204static void	 set_source(struct hp_order *, struct policyhead *);
205static int	 matchlen(struct sockaddr *, struct sockaddr *);
206static int	 comp_dst(const void *, const void *);
207static int	 gai_addr2scopetype(struct sockaddr *);
208#endif
209
210/*
211 * Functions defined in RFC2553
212 *	getipnodebyname, getipnodebyaddr, freehostent
213 */
214
215struct hostent *
216getipnodebyname(const char *name, int af, int flags, int *errp)
217{
218	struct hostent *hp;
219	union inx_addr addrbuf;
220	res_state statp;
221	u_long options;
222
223	switch (af) {
224	case AF_INET:
225#ifdef INET6
226	case AF_INET6:
227#endif
228		break;
229	default:
230		*errp = NO_RECOVERY;
231		return NULL;
232	}
233
234	if (flags & AI_ADDRCONFIG) {
235		int s;
236
237		if ((s = _socket(af, SOCK_DGRAM | SOCK_CLOEXEC, 0)) < 0)
238			return NULL;
239		/*
240		 * TODO:
241		 * Note that implementation dependent test for address
242		 * configuration should be done every time called
243		 * (or appropriate interval),
244		 * because addresses will be dynamically assigned or deleted.
245		 */
246		_close(s);
247	}
248
249#ifdef INET6
250	/* special case for literal address */
251	if (inet_pton(AF_INET6, name, &addrbuf) == 1) {
252		if (af != AF_INET6) {
253			*errp = HOST_NOT_FOUND;
254			return NULL;
255		}
256		return _hpaddr(af, name, &addrbuf, errp);
257	}
258#endif
259	if (inet_aton(name, (struct in_addr *)&addrbuf) == 1) {
260		if (af != AF_INET) {
261			if (MAPADDRENABLED(flags)) {
262				MAPADDR(&addrbuf, &addrbuf.in_addr);
263			} else {
264				*errp = HOST_NOT_FOUND;
265				return NULL;
266			}
267		}
268		return _hpaddr(af, name, &addrbuf, errp);
269	}
270
271
272	statp = __res_state();
273	if ((statp->options & RES_INIT) == 0) {
274		if (res_ninit(statp) < 0) {
275			*errp = NETDB_INTERNAL;
276			return NULL;
277		}
278	}
279
280	options = statp->options;
281	statp->options &= ~RES_USE_INET6;
282
283	hp = gethostbyname2(name, af);
284	hp = _hpcopy(hp, errp);
285#ifdef INET6
286	if (af == AF_INET6)
287		hp = _hpreorder(hp);
288
289	if (af == AF_INET6 && ((flags & AI_ALL) || hp == NULL) &&
290	    MAPADDRENABLED(flags)) {
291		struct hostent *hp2 = gethostbyname2(name, AF_INET);
292		if (hp == NULL)
293			if (hp2 == NULL)
294				*errp = statp->res_h_errno;
295			else
296				hp = _hpmapv6(hp2, errp);
297		else {
298			if (hp2 && strcmp(hp->h_name, hp2->h_name) == 0) {
299				struct hostent *hpb = hp;
300				hp = _hpmerge(hpb, hp2, errp);
301				freehostent(hpb);
302			}
303		}
304	}
305#endif
306
307	if (hp == NULL)
308		*errp = statp->res_h_errno;
309
310	statp->options = options;
311	return _hpsort(hp, statp);
312}
313
314struct hostent *
315getipnodebyaddr(const void *src, size_t len, int af, int *errp)
316{
317	struct hostent *hp;
318	res_state statp;
319	u_long options;
320
321#ifdef INET6
322	struct in6_addr addrbuf;
323#else
324	struct in_addr addrbuf;
325#endif
326
327	switch (af) {
328	case AF_INET:
329		if (len != sizeof(struct in_addr)) {
330			*errp = NO_RECOVERY;
331			return NULL;
332		}
333		if (rounddown2((long)src, sizeof(struct in_addr))) {
334			memcpy(&addrbuf, src, len);
335			src = &addrbuf;
336		}
337		if (((struct in_addr *)src)->s_addr == 0)
338			return NULL;
339		break;
340#ifdef INET6
341	case AF_INET6:
342		if (len != sizeof(struct in6_addr)) {
343			*errp = NO_RECOVERY;
344			return NULL;
345		}
346		if (rounddown2((long)src, sizeof(struct in6_addr) / 2)) {
347			/* XXX */
348			memcpy(&addrbuf, src, len);
349			src = &addrbuf;
350		}
351		if (IN6_IS_ADDR_UNSPECIFIED((struct in6_addr *)src))
352			return NULL;
353		if (IN6_IS_ADDR_V4MAPPED((struct in6_addr *)src)
354		||  IN6_IS_ADDR_V4COMPAT((struct in6_addr *)src)) {
355			src = (char *)src +
356			    (sizeof(struct in6_addr) - sizeof(struct in_addr));
357			af = AF_INET;
358			len = sizeof(struct in_addr);
359		}
360		break;
361#endif
362	default:
363		*errp = NO_RECOVERY;
364		return NULL;
365	}
366
367	statp = __res_state();
368	if ((statp->options & RES_INIT) == 0) {
369		if (res_ninit(statp) < 0) {
370			RES_SET_H_ERRNO(statp, NETDB_INTERNAL);
371			return NULL;
372		}
373	}
374
375	options = statp->options;
376	statp->options &= ~RES_USE_INET6;
377
378	hp = gethostbyaddr(src, len, af);
379	if (hp == NULL)
380		*errp = statp->res_h_errno;
381
382	statp->options = options;
383	return (_hpcopy(hp, errp));
384}
385
386void
387freehostent(struct hostent *ptr)
388{
389	free(ptr);
390}
391
392/*
393 * Private utility functions
394 */
395
396/*
397 * _hpcopy: allocate and copy hostent structure
398 */
399static struct hostent *
400_hpcopy(struct hostent *hp, int *errp)
401{
402	struct hostent *nhp;
403	char *cp, **pp;
404	int size, addrsize;
405	int nalias = 0, naddr = 0;
406	int al_off;
407	int i;
408
409	if (hp == NULL)
410		return hp;
411
412	/* count size to be allocated */
413	size = sizeof(struct hostent);
414	if (hp->h_name != NULL)
415		size += strlen(hp->h_name) + 1;
416	if ((pp = hp->h_aliases) != NULL) {
417		for (i = 0; *pp != NULL; i++, pp++) {
418			if (**pp != '\0') {
419				size += strlen(*pp) + 1;
420				nalias++;
421			}
422		}
423	}
424	/* adjust alignment */
425	size = ALIGN(size);
426	al_off = size;
427	size += sizeof(char *) * (nalias + 1);
428	addrsize = ALIGN(hp->h_length);
429	if ((pp = hp->h_addr_list) != NULL) {
430		while (*pp++ != NULL)
431			naddr++;
432	}
433	size += addrsize * naddr;
434	size += sizeof(char *) * (naddr + 1);
435
436	/* copy */
437	if ((nhp = (struct hostent *)malloc(size)) == NULL) {
438		*errp = TRY_AGAIN;
439		return NULL;
440	}
441	cp = (char *)&nhp[1];
442	if (hp->h_name != NULL) {
443		nhp->h_name = cp;
444		strcpy(cp, hp->h_name);
445		cp += strlen(cp) + 1;
446	} else
447		nhp->h_name = NULL;
448	nhp->h_aliases = (char **)((char *)nhp + al_off);
449	if ((pp = hp->h_aliases) != NULL) {
450		for (i = 0; *pp != NULL; pp++) {
451			if (**pp != '\0') {
452				nhp->h_aliases[i++] = cp;
453				strcpy(cp, *pp);
454				cp += strlen(cp) + 1;
455			}
456		}
457	}
458	nhp->h_aliases[nalias] = NULL;
459	cp = (char *)&nhp->h_aliases[nalias + 1];
460	nhp->h_addrtype = hp->h_addrtype;
461	nhp->h_length = hp->h_length;
462	nhp->h_addr_list = (char **)cp;
463	if ((pp = hp->h_addr_list) != NULL) {
464		cp = (char *)&nhp->h_addr_list[naddr + 1];
465		for (i = 0; *pp != NULL; pp++) {
466			nhp->h_addr_list[i++] = cp;
467			memcpy(cp, *pp, hp->h_length);
468			cp += addrsize;
469		}
470	}
471	nhp->h_addr_list[naddr] = NULL;
472	return nhp;
473}
474
475/*
476 * _hpaddr: construct hostent structure with one address
477 */
478static struct hostent *
479_hpaddr(int af, const char *name, void *addr, int *errp)
480{
481	struct hostent *hp, hpbuf;
482	char *addrs[2];
483
484	hp = &hpbuf;
485	hp->h_name = (char *)name;
486	hp->h_aliases = NULL;
487	hp->h_addrtype = af;
488	hp->h_length = ADDRLEN(af);
489	hp->h_addr_list = addrs;
490	addrs[0] = (char *)addr;
491	addrs[1] = NULL;
492	return (_hpcopy(hp, errp));
493}
494
495#ifdef INET6
496/*
497 * _hpmerge: merge 2 hostent structure, arguments will be freed
498 */
499static struct hostent *
500_hpmerge(struct hostent *hp1, struct hostent *hp2, int *errp)
501{
502	int i, j;
503	int naddr, nalias;
504	char **pp;
505	struct hostent *hp, hpbuf;
506	char *aliases[MAXALIASES + 1], *addrs[MAXADDRS + 1];
507	union inx_addr addrbuf[MAXADDRS];
508
509	if (hp1 == NULL)
510		return _hpcopy(hp2, errp);
511	if (hp2 == NULL)
512		return _hpcopy(hp1, errp);
513
514#define	HP(i)	(i == 1 ? hp1 : hp2)
515	hp = &hpbuf;
516	hp->h_name = (hp1->h_name != NULL ? hp1->h_name : hp2->h_name);
517	hp->h_aliases = aliases;
518	nalias = 0;
519	for (i = 1; i <= 2; i++) {
520		if ((pp = HP(i)->h_aliases) == NULL)
521			continue;
522		for (; nalias < MAXALIASES && *pp != NULL; pp++) {
523			/* check duplicates */
524			for (j = 0; j < nalias; j++)
525				if (strcasecmp(*pp, aliases[j]) == 0)
526					break;
527			if (j == nalias)
528				aliases[nalias++] = *pp;
529		}
530	}
531	aliases[nalias] = NULL;
532	if (hp1->h_length != hp2->h_length) {
533		hp->h_addrtype = AF_INET6;
534		hp->h_length = sizeof(struct in6_addr);
535	} else {
536		hp->h_addrtype = hp1->h_addrtype;
537		hp->h_length = hp1->h_length;
538	}
539
540	hp->h_addr_list = addrs;
541	naddr = 0;
542	for (i = 1; i <= 2; i++) {
543		if ((pp = HP(i)->h_addr_list) == NULL)
544			continue;
545		if (HP(i)->h_length == hp->h_length) {
546			while (naddr < MAXADDRS && *pp != NULL)
547				addrs[naddr++] = *pp++;
548		} else {
549			/* copy IPv4 addr as mapped IPv6 addr */
550			while (naddr < MAXADDRS && *pp != NULL) {
551				MAPADDR(&addrbuf[naddr], *pp++);
552				addrs[naddr] = (char *)&addrbuf[naddr];
553				naddr++;
554			}
555		}
556	}
557	addrs[naddr] = NULL;
558	return (_hpcopy(hp, errp));
559}
560#endif
561
562/*
563 * _hpmapv6: convert IPv4 hostent into IPv4-mapped IPv6 addresses
564 */
565#ifdef INET6
566static struct hostent *
567_hpmapv6(struct hostent *hp, int *errp)
568{
569	struct hostent hp6;
570
571	if (hp == NULL)
572		return NULL;
573	if (hp->h_addrtype == AF_INET6)
574		return _hpcopy(hp, errp);
575
576	memset(&hp6, 0, sizeof(struct hostent));
577	hp6.h_addrtype = AF_INET6;
578	hp6.h_length = sizeof(struct in6_addr);
579	return _hpmerge(&hp6, hp, errp);
580}
581#endif
582
583/*
584 * _hpsort: sort address by sortlist
585 */
586static struct hostent *
587_hpsort(struct hostent *hp, res_state statp)
588{
589	int i, j, n;
590	u_char *ap, *sp, *mp, **pp;
591	char t;
592	char order[MAXADDRS];
593	int nsort = statp->nsort;
594
595	if (hp == NULL || hp->h_addr_list[1] == NULL || nsort == 0)
596		return hp;
597	for (i = 0; (ap = (u_char *)hp->h_addr_list[i]); i++) {
598		for (j = 0; j < nsort; j++) {
599#ifdef INET6
600			if (statp->_u._ext.ext->sort_list[j].af !=
601			    hp->h_addrtype)
602				continue;
603			sp = (u_char *)&statp->_u._ext.ext->sort_list[j].addr;
604			mp = (u_char *)&statp->_u._ext.ext->sort_list[j].mask;
605#else
606			sp = (u_char *)&statp->sort_list[j].addr;
607			mp = (u_char *)&statp->sort_list[j].mask;
608#endif
609			for (n = 0; n < hp->h_length; n++) {
610				if ((ap[n] & mp[n]) != sp[n])
611					break;
612			}
613			if (n == hp->h_length)
614				break;
615		}
616		order[i] = j;
617	}
618	n = i;
619	pp = (u_char **)hp->h_addr_list;
620	for (i = 0; i < n - 1; i++) {
621		for (j = i + 1; j < n; j++) {
622			if (order[i] > order[j]) {
623				ap = pp[i];
624				pp[i] = pp[j];
625				pp[j] = ap;
626				t = order[i];
627				order[i] = order[j];
628				order[j] = t;
629			}
630		}
631	}
632	return hp;
633}
634
635#ifdef INET6
636/*
637 * _hpreorder: sort address by default address selection
638 */
639static struct hostent *
640_hpreorder(struct hostent *hp)
641{
642	struct hp_order *aio;
643	int i, n;
644	char *ap;
645	struct sockaddr *sa;
646	struct policyhead policyhead;
647
648	if (hp == NULL)
649		return hp;
650
651	switch (hp->h_addrtype) {
652	case AF_INET:
653#ifdef INET6
654	case AF_INET6:
655#endif
656		break;
657	default:
658		return hp;
659	}
660
661	/* count the number of addrinfo elements for sorting. */
662	for (n = 0; hp->h_addr_list[n] != NULL; n++)
663		;
664
665	/*
666	 * If the number is small enough, we can skip the reordering process.
667	 */
668	if (n <= 1)
669		return hp;
670
671	/* allocate a temporary array for sort and initialization of it. */
672	if ((aio = malloc(sizeof(*aio) * n)) == NULL)
673		return hp;	/* give up reordering */
674	memset(aio, 0, sizeof(*aio) * n);
675
676	/* retrieve address selection policy from the kernel */
677	TAILQ_INIT(&policyhead);
678	if (!get_addrselectpolicy(&policyhead)) {
679		/* no policy is installed into kernel, we don't sort. */
680		free(aio);
681		return hp;
682	}
683
684	for (i = 0; i < n; i++) {
685		ap = hp->h_addr_list[i];
686		aio[i].aio_h_addr = ap;
687		sa = &aio[i].aio_sa;
688		switch (hp->h_addrtype) {
689		case AF_INET:
690			sa->sa_family = AF_INET;
691			sa->sa_len = sizeof(struct sockaddr_in);
692			memcpy(&((struct sockaddr_in *)sa)->sin_addr, ap,
693			    sizeof(struct in_addr));
694			break;
695#ifdef INET6
696		case AF_INET6:
697			if (IN6_IS_ADDR_V4MAPPED((struct in6_addr *)ap)) {
698				sa->sa_family = AF_INET;
699				sa->sa_len = sizeof(struct sockaddr_in);
700				memcpy(&((struct sockaddr_in *)sa)->sin_addr,
701				    &ap[12], sizeof(struct in_addr));
702			} else {
703				sa->sa_family = AF_INET6;
704				sa->sa_len = sizeof(struct sockaddr_in6);
705				memcpy(&((struct sockaddr_in6 *)sa)->sin6_addr,
706				    ap, sizeof(struct in6_addr));
707			}
708			break;
709#endif
710		}
711		aio[i].aio_dstscope = gai_addr2scopetype(sa);
712		aio[i].aio_dstpolicy = match_addrselectpolicy(sa, &policyhead);
713		set_source(&aio[i], &policyhead);
714		aio[i].aio_initial_sequence = i;
715	}
716
717	/* perform sorting. */
718	qsort(aio, n, sizeof(*aio), comp_dst);
719
720	/* reorder the h_addr_list. */
721	for (i = 0; i < n; i++)
722		hp->h_addr_list[i] = aio[i].aio_h_addr;
723
724	/* cleanup and return */
725	free(aio);
726	free_addrselectpolicy(&policyhead);
727	return hp;
728}
729
730static int
731get_addrselectpolicy(struct policyhead *head)
732{
733#ifdef INET6
734	int mib[] = { CTL_NET, PF_INET6, IPPROTO_IPV6, IPV6CTL_ADDRCTLPOLICY };
735	size_t l;
736	char *buf;
737	struct in6_addrpolicy *pol, *ep;
738
739	if (sysctl(mib, nitems(mib), NULL, &l, NULL, 0) < 0)
740		return (0);
741	if ((buf = malloc(l)) == NULL)
742		return (0);
743	if (sysctl(mib, nitems(mib), buf, &l, NULL, 0) < 0) {
744		free(buf);
745		return (0);
746	}
747
748	ep = (struct in6_addrpolicy *)(buf + l);
749	for (pol = (struct in6_addrpolicy *)buf; pol + 1 <= ep; pol++) {
750		struct policyqueue *new;
751
752		if ((new = malloc(sizeof(*new))) == NULL) {
753			free_addrselectpolicy(head); /* make the list empty */
754			break;
755		}
756		new->pc_policy = *pol;
757		TAILQ_INSERT_TAIL(head, new, pc_entry);
758	}
759
760	free(buf);
761	return (1);
762#else
763	return (0);
764#endif
765}
766
767static void
768free_addrselectpolicy(struct policyhead *head)
769{
770	struct policyqueue *ent, *nent;
771
772	for (ent = TAILQ_FIRST(head); ent; ent = nent) {
773		nent = TAILQ_NEXT(ent, pc_entry);
774		TAILQ_REMOVE(head, ent, pc_entry);
775		free(ent);
776	}
777}
778
779static struct policyqueue *
780match_addrselectpolicy(struct sockaddr *addr, struct policyhead *head)
781{
782#ifdef INET6
783	struct policyqueue *ent, *bestent = NULL;
784	struct in6_addrpolicy *pol;
785	int matchlen, bestmatchlen = -1;
786	u_char *mp, *ep, *k, *p, m;
787	struct sockaddr_in6 key;
788
789	switch(addr->sa_family) {
790	case AF_INET6:
791		key = *(struct sockaddr_in6 *)addr;
792		break;
793	case AF_INET:
794		/* convert the address into IPv4-mapped IPv6 address. */
795		memset(&key, 0, sizeof(key));
796		key.sin6_family = AF_INET6;
797		key.sin6_len = sizeof(key);
798		_map_v4v6_address(
799		    (char *)&((struct sockaddr_in *)addr)->sin_addr,
800		    (char *)&key.sin6_addr);
801		break;
802	default:
803		return(NULL);
804	}
805
806	for (ent = TAILQ_FIRST(head); ent; ent = TAILQ_NEXT(ent, pc_entry)) {
807		pol = &ent->pc_policy;
808		matchlen = 0;
809
810		mp = (u_char *)&pol->addrmask.sin6_addr;
811		ep = mp + 16;	/* XXX: scope field? */
812		k = (u_char *)&key.sin6_addr;
813		p = (u_char *)&pol->addr.sin6_addr;
814		for (; mp < ep && *mp; mp++, k++, p++) {
815			m = *mp;
816			if ((*k & m) != *p)
817				goto next; /* not match */
818			if (m == 0xff) /* short cut for a typical case */
819				matchlen += 8;
820			else {
821				while (m >= 0x80) {
822					matchlen++;
823					m <<= 1;
824				}
825			}
826		}
827
828		/* matched.  check if this is better than the current best. */
829		if (matchlen > bestmatchlen) {
830			bestent = ent;
831			bestmatchlen = matchlen;
832		}
833
834	  next:
835		continue;
836	}
837
838	return(bestent);
839#else
840	return(NULL);
841#endif
842
843}
844
845static void
846set_source(struct hp_order *aio, struct policyhead *ph)
847{
848	struct sockaddr_storage ss = aio->aio_un.aiou_ss;
849	socklen_t srclen;
850	int s;
851
852	/* set unspec ("no source is available"), just in case */
853	aio->aio_srcsa.sa_family = AF_UNSPEC;
854	aio->aio_srcscope = -1;
855
856	switch(ss.ss_family) {
857	case AF_INET:
858		((struct sockaddr_in *)&ss)->sin_port = htons(1);
859		break;
860#ifdef INET6
861	case AF_INET6:
862		((struct sockaddr_in6 *)&ss)->sin6_port = htons(1);
863		break;
864#endif
865	default:		/* ignore unsupported AFs explicitly */
866		return;
867	}
868
869	/* open a socket to get the source address for the given dst */
870	if ((s = _socket(ss.ss_family, SOCK_DGRAM | SOCK_CLOEXEC,
871	    IPPROTO_UDP)) < 0)
872		return;		/* give up */
873	if (_connect(s, (struct sockaddr *)&ss, ss.ss_len) < 0)
874		goto cleanup;
875	srclen = ss.ss_len;
876	if (_getsockname(s, &aio->aio_srcsa, &srclen) < 0) {
877		aio->aio_srcsa.sa_family = AF_UNSPEC;
878		goto cleanup;
879	}
880	aio->aio_srcscope = gai_addr2scopetype(&aio->aio_srcsa);
881	aio->aio_srcpolicy = match_addrselectpolicy(&aio->aio_srcsa, ph);
882	aio->aio_matchlen = matchlen(&aio->aio_srcsa, (struct sockaddr *)&ss);
883#ifdef INET6
884	if (ss.ss_family == AF_INET6) {
885		struct in6_ifreq ifr6;
886		u_int32_t flags6;
887
888		memset(&ifr6, 0, sizeof(ifr6));
889		memcpy(&ifr6.ifr_addr, &ss, ss.ss_len);
890		if (_ioctl(s, SIOCGIFAFLAG_IN6, &ifr6) == 0) {
891			flags6 = ifr6.ifr_ifru.ifru_flags6;
892			if ((flags6 & IN6_IFF_DEPRECATED))
893				aio->aio_srcflag |= AIO_SRCFLAG_DEPRECATED;
894		}
895	}
896#endif
897
898  cleanup:
899	_close(s);
900	return;
901}
902
903static int
904matchlen(struct sockaddr *src, struct sockaddr *dst)
905{
906	int match = 0;
907	u_char *s, *d;
908	u_char *lim, r;
909	int addrlen;
910
911	switch (src->sa_family) {
912#ifdef INET6
913	case AF_INET6:
914		s = (u_char *)&((struct sockaddr_in6 *)src)->sin6_addr;
915		d = (u_char *)&((struct sockaddr_in6 *)dst)->sin6_addr;
916		addrlen = sizeof(struct in6_addr);
917		lim = s + addrlen;
918		break;
919#endif
920	case AF_INET:
921		s = (u_char *)&((struct sockaddr_in *)src)->sin_addr;
922		d = (u_char *)&((struct sockaddr_in *)dst)->sin_addr;
923		addrlen = sizeof(struct in_addr);
924		lim = s + addrlen;
925		break;
926	default:
927		return(0);
928	}
929
930	while (s < lim)
931		if ((r = (*d++ ^ *s++)) != 0) {
932			while ((r & 0x80) == 0) {
933				match++;
934				r <<= 1;
935			}
936			break;
937		} else
938			match += 8;
939	return(match);
940}
941
942static int
943comp_dst(const void *arg1, const void *arg2)
944{
945	const struct hp_order *dst1 = arg1, *dst2 = arg2;
946
947	/*
948	 * Rule 1: Avoid unusable destinations.
949	 * XXX: we currently do not consider if an appropriate route exists.
950	 */
951	if (dst1->aio_srcsa.sa_family != AF_UNSPEC &&
952	    dst2->aio_srcsa.sa_family == AF_UNSPEC) {
953		return(-1);
954	}
955	if (dst1->aio_srcsa.sa_family == AF_UNSPEC &&
956	    dst2->aio_srcsa.sa_family != AF_UNSPEC) {
957		return(1);
958	}
959
960	/* Rule 2: Prefer matching scope. */
961	if (dst1->aio_dstscope == dst1->aio_srcscope &&
962	    dst2->aio_dstscope != dst2->aio_srcscope) {
963		return(-1);
964	}
965	if (dst1->aio_dstscope != dst1->aio_srcscope &&
966	    dst2->aio_dstscope == dst2->aio_srcscope) {
967		return(1);
968	}
969
970	/* Rule 3: Avoid deprecated addresses. */
971	if (dst1->aio_srcsa.sa_family != AF_UNSPEC &&
972	    dst2->aio_srcsa.sa_family != AF_UNSPEC) {
973		if (!(dst1->aio_srcflag & AIO_SRCFLAG_DEPRECATED) &&
974		    (dst2->aio_srcflag & AIO_SRCFLAG_DEPRECATED)) {
975			return(-1);
976		}
977		if ((dst1->aio_srcflag & AIO_SRCFLAG_DEPRECATED) &&
978		    !(dst2->aio_srcflag & AIO_SRCFLAG_DEPRECATED)) {
979			return(1);
980		}
981	}
982
983	/* Rule 4: Prefer home addresses. */
984	/* XXX: not implemented yet */
985
986	/* Rule 5: Prefer matching label. */
987#ifdef INET6
988	if (dst1->aio_srcpolicy && dst1->aio_dstpolicy &&
989	    dst1->aio_srcpolicy->pc_policy.label ==
990	    dst1->aio_dstpolicy->pc_policy.label &&
991	    (dst2->aio_srcpolicy == NULL || dst2->aio_dstpolicy == NULL ||
992	     dst2->aio_srcpolicy->pc_policy.label !=
993	     dst2->aio_dstpolicy->pc_policy.label)) {
994		return(-1);
995	}
996	if (dst2->aio_srcpolicy && dst2->aio_dstpolicy &&
997	    dst2->aio_srcpolicy->pc_policy.label ==
998	    dst2->aio_dstpolicy->pc_policy.label &&
999	    (dst1->aio_srcpolicy == NULL || dst1->aio_dstpolicy == NULL ||
1000	     dst1->aio_srcpolicy->pc_policy.label !=
1001	     dst1->aio_dstpolicy->pc_policy.label)) {
1002		return(1);
1003	}
1004#endif
1005
1006	/* Rule 6: Prefer higher precedence. */
1007#ifdef INET6
1008	if (dst1->aio_dstpolicy &&
1009	    (dst2->aio_dstpolicy == NULL ||
1010	     dst1->aio_dstpolicy->pc_policy.preced >
1011	     dst2->aio_dstpolicy->pc_policy.preced)) {
1012		return(-1);
1013	}
1014	if (dst2->aio_dstpolicy &&
1015	    (dst1->aio_dstpolicy == NULL ||
1016	     dst2->aio_dstpolicy->pc_policy.preced >
1017	     dst1->aio_dstpolicy->pc_policy.preced)) {
1018		return(1);
1019	}
1020#endif
1021
1022	/* Rule 7: Prefer native transport. */
1023	/* XXX: not implemented yet */
1024
1025	/* Rule 8: Prefer smaller scope. */
1026	if (dst1->aio_dstscope >= 0 &&
1027	    dst1->aio_dstscope < dst2->aio_dstscope) {
1028		return(-1);
1029	}
1030	if (dst2->aio_dstscope >= 0 &&
1031	    dst2->aio_dstscope < dst1->aio_dstscope) {
1032		return(1);
1033	}
1034
1035	/*
1036	 * Rule 9: Use longest matching prefix.
1037	 * We compare the match length in a same AF only.
1038	 */
1039	if (dst1->aio_sa.sa_family == dst2->aio_sa.sa_family) {
1040		if (dst1->aio_matchlen > dst2->aio_matchlen) {
1041			return(-1);
1042		}
1043		if (dst1->aio_matchlen < dst2->aio_matchlen) {
1044			return(1);
1045		}
1046	}
1047
1048	/* Rule 10: Otherwise, leave the order unchanged. */
1049
1050	/*
1051	 * Note that qsort is unstable; so, we can't return zero and
1052	 * expect the order to be unchanged.
1053	 * That also means we can't depend on the current position of
1054	 * dst2 being after dst1.  We must enforce the initial order
1055	 * with an explicit compare on the original position.
1056	 * The qsort specification requires that "When the same objects
1057	 * (consisting of width bytes, irrespective of their current
1058	 * positions in the array) are passed more than once to the
1059	 * comparison function, the results shall be consistent with one
1060	 * another."
1061	 * In other words, If A < B, then we must also return B > A.
1062	 */
1063	if (dst2->aio_initial_sequence < dst1->aio_initial_sequence)
1064		return(1);
1065
1066	return(-1);
1067}
1068
1069/*
1070 * Copy from scope.c.
1071 * XXX: we should standardize the functions and link them as standard
1072 * library.
1073 */
1074static int
1075gai_addr2scopetype(struct sockaddr *sa)
1076{
1077#ifdef INET6
1078	struct sockaddr_in6 *sa6;
1079#endif
1080	struct sockaddr_in *sa4;
1081
1082	switch(sa->sa_family) {
1083#ifdef INET6
1084	case AF_INET6:
1085		sa6 = (struct sockaddr_in6 *)sa;
1086		if (IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
1087			/* just use the scope field of the multicast address */
1088			return(sa6->sin6_addr.s6_addr[2] & 0x0f);
1089		}
1090		/*
1091		 * Unicast addresses: map scope type to corresponding scope
1092		 * value defined for multcast addresses.
1093		 * XXX: hardcoded scope type values are bad...
1094		 */
1095		if (IN6_IS_ADDR_LOOPBACK(&sa6->sin6_addr))
1096			return(1); /* node local scope */
1097		if (IN6_IS_ADDR_LINKLOCAL(&sa6->sin6_addr))
1098			return(2); /* link-local scope */
1099		if (IN6_IS_ADDR_SITELOCAL(&sa6->sin6_addr))
1100			return(5); /* site-local scope */
1101		return(14);	/* global scope */
1102		break;
1103#endif
1104	case AF_INET:
1105		/*
1106		 * IPv4 pseudo scoping according to RFC 3484.
1107		 */
1108		sa4 = (struct sockaddr_in *)sa;
1109		/* IPv4 autoconfiguration addresses have link-local scope. */
1110		if (((u_char *)&sa4->sin_addr)[0] == 169 &&
1111		    ((u_char *)&sa4->sin_addr)[1] == 254)
1112			return(2);
1113		/* Private addresses have site-local scope. */
1114		if (((u_char *)&sa4->sin_addr)[0] == 10 ||
1115		    (((u_char *)&sa4->sin_addr)[0] == 172 &&
1116		     (((u_char *)&sa4->sin_addr)[1] & 0xf0) == 16) ||
1117		    (((u_char *)&sa4->sin_addr)[0] == 192 &&
1118		     ((u_char *)&sa4->sin_addr)[1] == 168))
1119			return(14);	/* XXX: It should be 5 unless NAT */
1120		/* Loopback addresses have link-local scope. */
1121		if (((u_char *)&sa4->sin_addr)[0] == 127)
1122			return(2);
1123		return(14);
1124		break;
1125	default:
1126		errno = EAFNOSUPPORT; /* is this a good error? */
1127		return(-1);
1128	}
1129}
1130#endif
1131