xref: /freebsd/lib/libc/net/name6.c (revision 1f4bcc459a76b7aa664f3fd557684cd0ba6da352)
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) \
136 do {									\
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 
145 union 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 
160 struct policyqueue {
161 	TAILQ_ENTRY(policyqueue) pc_entry;
162 #ifdef INET6
163 	struct in6_addrpolicy pc_policy;
164 #endif
165 };
166 TAILQ_HEAD(policyhead, policyqueue);
167 
168 #define AIO_SRCFLAG_DEPRECATED	0x1
169 
170 struct 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 };
189 
190 static struct	 hostent *_hpcopy(struct hostent *, int *);
191 static struct	 hostent *_hpaddr(int, const char *, void *, int *);
192 #ifdef INET6
193 static struct	 hostent *_hpmerge(struct hostent *, struct hostent *, int *);
194 static struct	 hostent *_hpmapv6(struct hostent *, int *);
195 #endif
196 static struct	 hostent *_hpsort(struct hostent *, res_state);
197 
198 #ifdef INET6
199 static struct	 hostent *_hpreorder(struct hostent *);
200 static int	 get_addrselectpolicy(struct policyhead *);
201 static void	 free_addrselectpolicy(struct policyhead *);
202 static struct	 policyqueue *match_addrselectpolicy(struct sockaddr *,
203 	struct policyhead *);
204 static void	 set_source(struct hp_order *, struct policyhead *);
205 static int	 matchlen(struct sockaddr *, struct sockaddr *);
206 static int	 comp_dst(const void *, const void *);
207 static int	 gai_addr2scopetype(struct sockaddr *);
208 #endif
209 
210 /*
211  * Functions defined in RFC2553
212  *	getipnodebyname, getipnodebyaddr, freehostent
213  */
214 
215 struct hostent *
216 getipnodebyname(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 everytime called
243 		 * (or apropriate 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 
314 struct hostent *
315 getipnodebyaddr(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 ((long)src & ~(sizeof(struct in_addr) - 1)) {
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 ((long)src & ~(sizeof(struct in6_addr) / 2 - 1)) {	/*XXX*/
347 			memcpy(&addrbuf, src, len);
348 			src = &addrbuf;
349 		}
350 		if (IN6_IS_ADDR_UNSPECIFIED((struct in6_addr *)src))
351 			return NULL;
352 		if (IN6_IS_ADDR_V4MAPPED((struct in6_addr *)src)
353 		||  IN6_IS_ADDR_V4COMPAT((struct in6_addr *)src)) {
354 			src = (char *)src +
355 			    (sizeof(struct in6_addr) - sizeof(struct in_addr));
356 			af = AF_INET;
357 			len = sizeof(struct in_addr);
358 		}
359 		break;
360 #endif
361 	default:
362 		*errp = NO_RECOVERY;
363 		return NULL;
364 	}
365 
366 	statp = __res_state();
367 	if ((statp->options & RES_INIT) == 0) {
368 		if (res_ninit(statp) < 0) {
369 			RES_SET_H_ERRNO(statp, NETDB_INTERNAL);
370 			return NULL;
371 		}
372 	}
373 
374 	options = statp->options;
375 	statp->options &= ~RES_USE_INET6;
376 
377 	hp = gethostbyaddr(src, len, af);
378 	if (hp == NULL)
379 		*errp = statp->res_h_errno;
380 
381 	statp->options = options;
382 	return (_hpcopy(hp, errp));
383 }
384 
385 void
386 freehostent(struct hostent *ptr)
387 {
388 	free(ptr);
389 }
390 
391 /*
392  * Private utility functions
393  */
394 
395 /*
396  * _hpcopy: allocate and copy hostent structure
397  */
398 static struct hostent *
399 _hpcopy(struct hostent *hp, int *errp)
400 {
401 	struct hostent *nhp;
402 	char *cp, **pp;
403 	int size, addrsize;
404 	int nalias = 0, naddr = 0;
405 	int al_off;
406 	int i;
407 
408 	if (hp == NULL)
409 		return hp;
410 
411 	/* count size to be allocated */
412 	size = sizeof(struct hostent);
413 	if (hp->h_name != NULL)
414 		size += strlen(hp->h_name) + 1;
415 	if ((pp = hp->h_aliases) != NULL) {
416 		for (i = 0; *pp != NULL; i++, pp++) {
417 			if (**pp != '\0') {
418 				size += strlen(*pp) + 1;
419 				nalias++;
420 			}
421 		}
422 	}
423 	/* adjust alignment */
424 	size = ALIGN(size);
425 	al_off = size;
426 	size += sizeof(char *) * (nalias + 1);
427 	addrsize = ALIGN(hp->h_length);
428 	if ((pp = hp->h_addr_list) != NULL) {
429 		while (*pp++ != NULL)
430 			naddr++;
431 	}
432 	size += addrsize * naddr;
433 	size += sizeof(char *) * (naddr + 1);
434 
435 	/* copy */
436 	if ((nhp = (struct hostent *)malloc(size)) == NULL) {
437 		*errp = TRY_AGAIN;
438 		return NULL;
439 	}
440 	cp = (char *)&nhp[1];
441 	if (hp->h_name != NULL) {
442 		nhp->h_name = cp;
443 		strcpy(cp, hp->h_name);
444 		cp += strlen(cp) + 1;
445 	} else
446 		nhp->h_name = NULL;
447 	nhp->h_aliases = (char **)((char *)nhp + al_off);
448 	if ((pp = hp->h_aliases) != NULL) {
449 		for (i = 0; *pp != NULL; pp++) {
450 			if (**pp != '\0') {
451 				nhp->h_aliases[i++] = cp;
452 				strcpy(cp, *pp);
453 				cp += strlen(cp) + 1;
454 			}
455 		}
456 	}
457 	nhp->h_aliases[nalias] = NULL;
458 	cp = (char *)&nhp->h_aliases[nalias + 1];
459 	nhp->h_addrtype = hp->h_addrtype;
460 	nhp->h_length = hp->h_length;
461 	nhp->h_addr_list = (char **)cp;
462 	if ((pp = hp->h_addr_list) != NULL) {
463 		cp = (char *)&nhp->h_addr_list[naddr + 1];
464 		for (i = 0; *pp != NULL; pp++) {
465 			nhp->h_addr_list[i++] = cp;
466 			memcpy(cp, *pp, hp->h_length);
467 			cp += addrsize;
468 		}
469 	}
470 	nhp->h_addr_list[naddr] = NULL;
471 	return nhp;
472 }
473 
474 /*
475  * _hpaddr: construct hostent structure with one address
476  */
477 static struct hostent *
478 _hpaddr(int af, const char *name, void *addr, int *errp)
479 {
480 	struct hostent *hp, hpbuf;
481 	char *addrs[2];
482 
483 	hp = &hpbuf;
484 	hp->h_name = (char *)name;
485 	hp->h_aliases = NULL;
486 	hp->h_addrtype = af;
487 	hp->h_length = ADDRLEN(af);
488 	hp->h_addr_list = addrs;
489 	addrs[0] = (char *)addr;
490 	addrs[1] = NULL;
491 	return (_hpcopy(hp, errp));
492 }
493 
494 #ifdef INET6
495 /*
496  * _hpmerge: merge 2 hostent structure, arguments will be freed
497  */
498 static struct hostent *
499 _hpmerge(struct hostent *hp1, struct hostent *hp2, int *errp)
500 {
501 	int i, j;
502 	int naddr, nalias;
503 	char **pp;
504 	struct hostent *hp, hpbuf;
505 	char *aliases[MAXALIASES + 1], *addrs[MAXADDRS + 1];
506 	union inx_addr addrbuf[MAXADDRS];
507 
508 	if (hp1 == NULL)
509 		return _hpcopy(hp2, errp);
510 	if (hp2 == NULL)
511 		return _hpcopy(hp1, errp);
512 
513 #define	HP(i)	(i == 1 ? hp1 : hp2)
514 	hp = &hpbuf;
515 	hp->h_name = (hp1->h_name != NULL ? hp1->h_name : hp2->h_name);
516 	hp->h_aliases = aliases;
517 	nalias = 0;
518 	for (i = 1; i <= 2; i++) {
519 		if ((pp = HP(i)->h_aliases) == NULL)
520 			continue;
521 		for (; nalias < MAXALIASES && *pp != NULL; pp++) {
522 			/* check duplicates */
523 			for (j = 0; j < nalias; j++)
524 				if (strcasecmp(*pp, aliases[j]) == 0)
525 					break;
526 			if (j == nalias)
527 				aliases[nalias++] = *pp;
528 		}
529 	}
530 	aliases[nalias] = NULL;
531 	if (hp1->h_length != hp2->h_length) {
532 		hp->h_addrtype = AF_INET6;
533 		hp->h_length = sizeof(struct in6_addr);
534 	} else {
535 		hp->h_addrtype = hp1->h_addrtype;
536 		hp->h_length = hp1->h_length;
537 	}
538 
539 	hp->h_addr_list = addrs;
540 	naddr = 0;
541 	for (i = 1; i <= 2; i++) {
542 		if ((pp = HP(i)->h_addr_list) == NULL)
543 			continue;
544 		if (HP(i)->h_length == hp->h_length) {
545 			while (naddr < MAXADDRS && *pp != NULL)
546 				addrs[naddr++] = *pp++;
547 		} else {
548 			/* copy IPv4 addr as mapped IPv6 addr */
549 			while (naddr < MAXADDRS && *pp != NULL) {
550 				MAPADDR(&addrbuf[naddr], *pp++);
551 				addrs[naddr] = (char *)&addrbuf[naddr];
552 				naddr++;
553 			}
554 		}
555 	}
556 	addrs[naddr] = NULL;
557 	return (_hpcopy(hp, errp));
558 }
559 #endif
560 
561 /*
562  * _hpmapv6: convert IPv4 hostent into IPv4-mapped IPv6 addresses
563  */
564 #ifdef INET6
565 static struct hostent *
566 _hpmapv6(struct hostent *hp, int *errp)
567 {
568 	struct hostent hp6;
569 
570 	if (hp == NULL)
571 		return NULL;
572 	if (hp->h_addrtype == AF_INET6)
573 		return _hpcopy(hp, errp);
574 
575 	memset(&hp6, 0, sizeof(struct hostent));
576 	hp6.h_addrtype = AF_INET6;
577 	hp6.h_length = sizeof(struct in6_addr);
578 	return _hpmerge(&hp6, hp, errp);
579 }
580 #endif
581 
582 /*
583  * _hpsort: sort address by sortlist
584  */
585 static struct hostent *
586 _hpsort(struct hostent *hp, res_state statp)
587 {
588 	int i, j, n;
589 	u_char *ap, *sp, *mp, **pp;
590 	char t;
591 	char order[MAXADDRS];
592 	int nsort = statp->nsort;
593 
594 	if (hp == NULL || hp->h_addr_list[1] == NULL || nsort == 0)
595 		return hp;
596 	for (i = 0; (ap = (u_char *)hp->h_addr_list[i]); i++) {
597 		for (j = 0; j < nsort; j++) {
598 #ifdef INET6
599 			if (statp->_u._ext.ext->sort_list[j].af !=
600 			    hp->h_addrtype)
601 				continue;
602 			sp = (u_char *)&statp->_u._ext.ext->sort_list[j].addr;
603 			mp = (u_char *)&statp->_u._ext.ext->sort_list[j].mask;
604 #else
605 			sp = (u_char *)&statp->sort_list[j].addr;
606 			mp = (u_char *)&statp->sort_list[j].mask;
607 #endif
608 			for (n = 0; n < hp->h_length; n++) {
609 				if ((ap[n] & mp[n]) != sp[n])
610 					break;
611 			}
612 			if (n == hp->h_length)
613 				break;
614 		}
615 		order[i] = j;
616 	}
617 	n = i;
618 	pp = (u_char **)hp->h_addr_list;
619 	for (i = 0; i < n - 1; i++) {
620 		for (j = i + 1; j < n; j++) {
621 			if (order[i] > order[j]) {
622 				ap = pp[i];
623 				pp[i] = pp[j];
624 				pp[j] = ap;
625 				t = order[i];
626 				order[i] = order[j];
627 				order[j] = t;
628 			}
629 		}
630 	}
631 	return hp;
632 }
633 
634 #ifdef INET6
635 /*
636  * _hpreorder: sort address by default address selection
637  */
638 static struct hostent *
639 _hpreorder(struct hostent *hp)
640 {
641 	struct hp_order *aio;
642 	int i, n;
643 	char *ap;
644 	struct sockaddr *sa;
645 	struct policyhead policyhead;
646 
647 	if (hp == NULL)
648 		return hp;
649 
650 	switch (hp->h_addrtype) {
651 	case AF_INET:
652 #ifdef INET6
653 	case AF_INET6:
654 #endif
655 		break;
656 	default:
657 		free_addrselectpolicy(&policyhead);
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 	}
715 
716 	/* perform sorting. */
717 	qsort(aio, n, sizeof(*aio), comp_dst);
718 
719 	/* reorder the h_addr_list. */
720 	for (i = 0; i < n; i++)
721 		hp->h_addr_list[i] = aio[i].aio_h_addr;
722 
723 	/* cleanup and return */
724 	free(aio);
725 	free_addrselectpolicy(&policyhead);
726 	return hp;
727 }
728 
729 static int
730 get_addrselectpolicy(struct policyhead *head)
731 {
732 #ifdef INET6
733 	int mib[] = { CTL_NET, PF_INET6, IPPROTO_IPV6, IPV6CTL_ADDRCTLPOLICY };
734 	size_t l;
735 	char *buf;
736 	struct in6_addrpolicy *pol, *ep;
737 
738 	if (sysctl(mib, sizeof(mib) / sizeof(mib[0]), NULL, &l, NULL, 0) < 0)
739 		return (0);
740 	if ((buf = malloc(l)) == NULL)
741 		return (0);
742 	if (sysctl(mib, sizeof(mib) / sizeof(mib[0]), buf, &l, NULL, 0) < 0) {
743 		free(buf);
744 		return (0);
745 	}
746 
747 	ep = (struct in6_addrpolicy *)(buf + l);
748 	for (pol = (struct in6_addrpolicy *)buf; pol + 1 <= ep; pol++) {
749 		struct policyqueue *new;
750 
751 		if ((new = malloc(sizeof(*new))) == NULL) {
752 			free_addrselectpolicy(head); /* make the list empty */
753 			break;
754 		}
755 		new->pc_policy = *pol;
756 		TAILQ_INSERT_TAIL(head, new, pc_entry);
757 	}
758 
759 	free(buf);
760 	return (1);
761 #else
762 	return (0);
763 #endif
764 }
765 
766 static void
767 free_addrselectpolicy(struct policyhead *head)
768 {
769 	struct policyqueue *ent, *nent;
770 
771 	for (ent = TAILQ_FIRST(head); ent; ent = nent) {
772 		nent = TAILQ_NEXT(ent, pc_entry);
773 		TAILQ_REMOVE(head, ent, pc_entry);
774 		free(ent);
775 	}
776 }
777 
778 static struct policyqueue *
779 match_addrselectpolicy(struct sockaddr *addr, struct policyhead *head)
780 {
781 #ifdef INET6
782 	struct policyqueue *ent, *bestent = NULL;
783 	struct in6_addrpolicy *pol;
784 	int matchlen, bestmatchlen = -1;
785 	u_char *mp, *ep, *k, *p, m;
786 	struct sockaddr_in6 key;
787 
788 	switch(addr->sa_family) {
789 	case AF_INET6:
790 		key = *(struct sockaddr_in6 *)addr;
791 		break;
792 	case AF_INET:
793 		/* convert the address into IPv4-mapped IPv6 address. */
794 		memset(&key, 0, sizeof(key));
795 		key.sin6_family = AF_INET6;
796 		key.sin6_len = sizeof(key);
797 		_map_v4v6_address(
798 		    (char *)&((struct sockaddr_in *)addr)->sin_addr,
799 		    (char *)&key.sin6_addr);
800 		break;
801 	default:
802 		return(NULL);
803 	}
804 
805 	for (ent = TAILQ_FIRST(head); ent; ent = TAILQ_NEXT(ent, pc_entry)) {
806 		pol = &ent->pc_policy;
807 		matchlen = 0;
808 
809 		mp = (u_char *)&pol->addrmask.sin6_addr;
810 		ep = mp + 16;	/* XXX: scope field? */
811 		k = (u_char *)&key.sin6_addr;
812 		p = (u_char *)&pol->addr.sin6_addr;
813 		for (; mp < ep && *mp; mp++, k++, p++) {
814 			m = *mp;
815 			if ((*k & m) != *p)
816 				goto next; /* not match */
817 			if (m == 0xff) /* short cut for a typical case */
818 				matchlen += 8;
819 			else {
820 				while (m >= 0x80) {
821 					matchlen++;
822 					m <<= 1;
823 				}
824 			}
825 		}
826 
827 		/* matched.  check if this is better than the current best. */
828 		if (matchlen > bestmatchlen) {
829 			bestent = ent;
830 			bestmatchlen = matchlen;
831 		}
832 
833 	  next:
834 		continue;
835 	}
836 
837 	return(bestent);
838 #else
839 	return(NULL);
840 #endif
841 
842 }
843 
844 static void
845 set_source(struct hp_order *aio, struct policyhead *ph)
846 {
847 	struct sockaddr_storage ss = aio->aio_un.aiou_ss;
848 	socklen_t srclen;
849 	int s;
850 
851 	/* set unspec ("no source is available"), just in case */
852 	aio->aio_srcsa.sa_family = AF_UNSPEC;
853 	aio->aio_srcscope = -1;
854 
855 	switch(ss.ss_family) {
856 	case AF_INET:
857 		((struct sockaddr_in *)&ss)->sin_port = htons(1);
858 		break;
859 #ifdef INET6
860 	case AF_INET6:
861 		((struct sockaddr_in6 *)&ss)->sin6_port = htons(1);
862 		break;
863 #endif
864 	default:		/* ignore unsupported AFs explicitly */
865 		return;
866 	}
867 
868 	/* open a socket to get the source address for the given dst */
869 	if ((s = _socket(ss.ss_family, SOCK_DGRAM | SOCK_CLOEXEC,
870 	    IPPROTO_UDP)) < 0)
871 		return;		/* give up */
872 	if (_connect(s, (struct sockaddr *)&ss, ss.ss_len) < 0)
873 		goto cleanup;
874 	srclen = ss.ss_len;
875 	if (_getsockname(s, &aio->aio_srcsa, &srclen) < 0) {
876 		aio->aio_srcsa.sa_family = AF_UNSPEC;
877 		goto cleanup;
878 	}
879 	aio->aio_srcscope = gai_addr2scopetype(&aio->aio_srcsa);
880 	aio->aio_srcpolicy = match_addrselectpolicy(&aio->aio_srcsa, ph);
881 	aio->aio_matchlen = matchlen(&aio->aio_srcsa, (struct sockaddr *)&ss);
882 #ifdef INET6
883 	if (ss.ss_family == AF_INET6) {
884 		struct in6_ifreq ifr6;
885 		u_int32_t flags6;
886 
887 		memset(&ifr6, 0, sizeof(ifr6));
888 		memcpy(&ifr6.ifr_addr, &ss, ss.ss_len);
889 		if (_ioctl(s, SIOCGIFAFLAG_IN6, &ifr6) == 0) {
890 			flags6 = ifr6.ifr_ifru.ifru_flags6;
891 			if ((flags6 & IN6_IFF_DEPRECATED))
892 				aio->aio_srcflag |= AIO_SRCFLAG_DEPRECATED;
893 		}
894 	}
895 #endif
896 
897   cleanup:
898 	_close(s);
899 	return;
900 }
901 
902 static int
903 matchlen(struct sockaddr *src, struct sockaddr *dst)
904 {
905 	int match = 0;
906 	u_char *s, *d;
907 	u_char *lim, r;
908 	int addrlen;
909 
910 	switch (src->sa_family) {
911 #ifdef INET6
912 	case AF_INET6:
913 		s = (u_char *)&((struct sockaddr_in6 *)src)->sin6_addr;
914 		d = (u_char *)&((struct sockaddr_in6 *)dst)->sin6_addr;
915 		addrlen = sizeof(struct in6_addr);
916 		lim = s + addrlen;
917 		break;
918 #endif
919 	case AF_INET:
920 		s = (u_char *)&((struct sockaddr_in *)src)->sin_addr;
921 		d = (u_char *)&((struct sockaddr_in *)dst)->sin_addr;
922 		addrlen = sizeof(struct in_addr);
923 		lim = s + addrlen;
924 		break;
925 	default:
926 		return(0);
927 	}
928 
929 	while (s < lim)
930 		if ((r = (*d++ ^ *s++)) != 0) {
931 			while (r < addrlen * 8) {
932 				match++;
933 				r <<= 1;
934 			}
935 			break;
936 		} else
937 			match += 8;
938 	return(match);
939 }
940 
941 static int
942 comp_dst(const void *arg1, const void *arg2)
943 {
944 	const struct hp_order *dst1 = arg1, *dst2 = arg2;
945 
946 	/*
947 	 * Rule 1: Avoid unusable destinations.
948 	 * XXX: we currently do not consider if an appropriate route exists.
949 	 */
950 	if (dst1->aio_srcsa.sa_family != AF_UNSPEC &&
951 	    dst2->aio_srcsa.sa_family == AF_UNSPEC) {
952 		return(-1);
953 	}
954 	if (dst1->aio_srcsa.sa_family == AF_UNSPEC &&
955 	    dst2->aio_srcsa.sa_family != AF_UNSPEC) {
956 		return(1);
957 	}
958 
959 	/* Rule 2: Prefer matching scope. */
960 	if (dst1->aio_dstscope == dst1->aio_srcscope &&
961 	    dst2->aio_dstscope != dst2->aio_srcscope) {
962 		return(-1);
963 	}
964 	if (dst1->aio_dstscope != dst1->aio_srcscope &&
965 	    dst2->aio_dstscope == dst2->aio_srcscope) {
966 		return(1);
967 	}
968 
969 	/* Rule 3: Avoid deprecated addresses. */
970 	if (dst1->aio_srcsa.sa_family != AF_UNSPEC &&
971 	    dst2->aio_srcsa.sa_family != AF_UNSPEC) {
972 		if (!(dst1->aio_srcflag & AIO_SRCFLAG_DEPRECATED) &&
973 		    (dst2->aio_srcflag & AIO_SRCFLAG_DEPRECATED)) {
974 			return(-1);
975 		}
976 		if ((dst1->aio_srcflag & AIO_SRCFLAG_DEPRECATED) &&
977 		    !(dst2->aio_srcflag & AIO_SRCFLAG_DEPRECATED)) {
978 			return(1);
979 		}
980 	}
981 
982 	/* Rule 4: Prefer home addresses. */
983 	/* XXX: not implemented yet */
984 
985 	/* Rule 5: Prefer matching label. */
986 #ifdef INET6
987 	if (dst1->aio_srcpolicy && dst1->aio_dstpolicy &&
988 	    dst1->aio_srcpolicy->pc_policy.label ==
989 	    dst1->aio_dstpolicy->pc_policy.label &&
990 	    (dst2->aio_srcpolicy == NULL || dst2->aio_dstpolicy == NULL ||
991 	     dst2->aio_srcpolicy->pc_policy.label !=
992 	     dst2->aio_dstpolicy->pc_policy.label)) {
993 		return(-1);
994 	}
995 	if (dst2->aio_srcpolicy && dst2->aio_dstpolicy &&
996 	    dst2->aio_srcpolicy->pc_policy.label ==
997 	    dst2->aio_dstpolicy->pc_policy.label &&
998 	    (dst1->aio_srcpolicy == NULL || dst1->aio_dstpolicy == NULL ||
999 	     dst1->aio_srcpolicy->pc_policy.label !=
1000 	     dst1->aio_dstpolicy->pc_policy.label)) {
1001 		return(1);
1002 	}
1003 #endif
1004 
1005 	/* Rule 6: Prefer higher precedence. */
1006 #ifdef INET6
1007 	if (dst1->aio_dstpolicy &&
1008 	    (dst2->aio_dstpolicy == NULL ||
1009 	     dst1->aio_dstpolicy->pc_policy.preced >
1010 	     dst2->aio_dstpolicy->pc_policy.preced)) {
1011 		return(-1);
1012 	}
1013 	if (dst2->aio_dstpolicy &&
1014 	    (dst1->aio_dstpolicy == NULL ||
1015 	     dst2->aio_dstpolicy->pc_policy.preced >
1016 	     dst1->aio_dstpolicy->pc_policy.preced)) {
1017 		return(1);
1018 	}
1019 #endif
1020 
1021 	/* Rule 7: Prefer native transport. */
1022 	/* XXX: not implemented yet */
1023 
1024 	/* Rule 8: Prefer smaller scope. */
1025 	if (dst1->aio_dstscope >= 0 &&
1026 	    dst1->aio_dstscope < dst2->aio_dstscope) {
1027 		return(-1);
1028 	}
1029 	if (dst2->aio_dstscope >= 0 &&
1030 	    dst2->aio_dstscope < dst1->aio_dstscope) {
1031 		return(1);
1032 	}
1033 
1034 	/*
1035 	 * Rule 9: Use longest matching prefix.
1036 	 * We compare the match length in a same AF only.
1037 	 */
1038 	if (dst1->aio_sa.sa_family == dst2->aio_sa.sa_family) {
1039 		if (dst1->aio_matchlen > dst2->aio_matchlen) {
1040 			return(-1);
1041 		}
1042 		if (dst1->aio_matchlen < dst2->aio_matchlen) {
1043 			return(1);
1044 		}
1045 	}
1046 
1047 	/* Rule 10: Otherwise, leave the order unchanged. */
1048 	return(-1);
1049 }
1050 
1051 /*
1052  * Copy from scope.c.
1053  * XXX: we should standardize the functions and link them as standard
1054  * library.
1055  */
1056 static int
1057 gai_addr2scopetype(struct sockaddr *sa)
1058 {
1059 #ifdef INET6
1060 	struct sockaddr_in6 *sa6;
1061 #endif
1062 	struct sockaddr_in *sa4;
1063 
1064 	switch(sa->sa_family) {
1065 #ifdef INET6
1066 	case AF_INET6:
1067 		sa6 = (struct sockaddr_in6 *)sa;
1068 		if (IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
1069 			/* just use the scope field of the multicast address */
1070 			return(sa6->sin6_addr.s6_addr[2] & 0x0f);
1071 		}
1072 		/*
1073 		 * Unicast addresses: map scope type to corresponding scope
1074 		 * value defined for multcast addresses.
1075 		 * XXX: hardcoded scope type values are bad...
1076 		 */
1077 		if (IN6_IS_ADDR_LOOPBACK(&sa6->sin6_addr))
1078 			return(1); /* node local scope */
1079 		if (IN6_IS_ADDR_LINKLOCAL(&sa6->sin6_addr))
1080 			return(2); /* link-local scope */
1081 		if (IN6_IS_ADDR_SITELOCAL(&sa6->sin6_addr))
1082 			return(5); /* site-local scope */
1083 		return(14);	/* global scope */
1084 		break;
1085 #endif
1086 	case AF_INET:
1087 		/*
1088 		 * IPv4 pseudo scoping according to RFC 3484.
1089 		 */
1090 		sa4 = (struct sockaddr_in *)sa;
1091 		/* IPv4 autoconfiguration addresses have link-local scope. */
1092 		if (((u_char *)&sa4->sin_addr)[0] == 169 &&
1093 		    ((u_char *)&sa4->sin_addr)[1] == 254)
1094 			return(2);
1095 		/* Private addresses have site-local scope. */
1096 		if (((u_char *)&sa4->sin_addr)[0] == 10 ||
1097 		    (((u_char *)&sa4->sin_addr)[0] == 172 &&
1098 		     (((u_char *)&sa4->sin_addr)[1] & 0xf0) == 16) ||
1099 		    (((u_char *)&sa4->sin_addr)[0] == 192 &&
1100 		     ((u_char *)&sa4->sin_addr)[1] == 168))
1101 			return(14);	/* XXX: It should be 5 unless NAT */
1102 		/* Loopback addresses have link-local scope. */
1103 		if (((u_char *)&sa4->sin_addr)[0] == 127)
1104 			return(2);
1105 		return(14);
1106 		break;
1107 	default:
1108 		errno = EAFNOSUPPORT; /* is this a good error? */
1109 		return(-1);
1110 	}
1111 }
1112 #endif
1113