xref: /dragonfly/lib/libc/net/name6.c (revision 60233e58)
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. All advertising materials mentioning features or use of this software
46  *    must display the following acknowledgement:
47  * 	This product includes software developed by the University of
48  * 	California, Berkeley and its contributors.
49  * 4. Neither the name of the University nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63  * SUCH DAMAGE.
64  * -
65  * Portions Copyright (c) 1993 by Digital Equipment Corporation.
66  *
67  * Permission to use, copy, modify, and distribute this software for any
68  * purpose with or without fee is hereby granted, provided that the above
69  * copyright notice and this permission notice appear in all copies, and that
70  * the name of Digital Equipment Corporation not be used in advertising or
71  * publicity pertaining to distribution of the document or software without
72  * specific, written prior permission.
73  *
74  * THE SOFTWARE IS PROVIDED "AS IS" AND DIGITAL EQUIPMENT CORP. DISCLAIMS ALL
75  * WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES
76  * OF MERCHANTABILITY AND FITNESS.   IN NO EVENT SHALL DIGITAL EQUIPMENT
77  * CORPORATION BE LIABLE FOR ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL
78  * DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR
79  * PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
80  * ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS
81  * SOFTWARE.
82  * -
83  * --Copyright--
84  *
85  * $FreeBSD: src/lib/libc/net/name6.c,v 1.62 2007/07/31 16:09:41 bushman Exp $
86  */
87 
88 /*
89  *	Atsushi Onoe <onoe@sm.sony.co.jp>
90  */
91 
92 #include "namespace.h"
93 #include <sys/param.h>
94 #include <sys/socket.h>
95 #include <sys/time.h>
96 #include <sys/queue.h>
97 #include <netinet/in.h>
98 #ifdef INET6
99 #include <net/if.h>
100 #include <net/if_var.h>
101 #include <sys/sysctl.h>
102 #include <sys/ioctl.h>
103 #include <netinet6/in6_var.h>	/* XXX */
104 #endif
105 
106 #include <arpa/inet.h>
107 #include <arpa/nameser.h>
108 
109 #include <errno.h>
110 #include <netdb.h>
111 #include <resolv.h>
112 #include <stdio.h>
113 #include <stdlib.h>
114 #include <string.h>
115 #include <stdarg.h>
116 #include <nsswitch.h>
117 #include <unistd.h>
118 #include "un-namespace.h"
119 #include "netdb_private.h"
120 #include "res_private.h"
121 
122 #ifndef MAXALIASES
123 #define	MAXALIASES	10
124 #endif
125 #ifndef	MAXADDRS
126 #define	MAXADDRS	20
127 #endif
128 #ifndef MAXDNAME
129 #define	MAXDNAME	1025
130 #endif
131 
132 #ifdef INET6
133 #define	ADDRLEN(af)	((af) == AF_INET6 ? sizeof(struct in6_addr) : \
134 					    sizeof(struct in_addr))
135 #else
136 #define	ADDRLEN(af)	sizeof(struct in_addr)
137 #endif
138 
139 #define	MAPADDR(ab, ina) \
140 do {									\
141 	memcpy(&(ab)->map_inaddr, ina, sizeof(struct in_addr));		\
142 	memset((ab)->map_zero, 0, sizeof((ab)->map_zero));		\
143 	memset((ab)->map_one, 0xff, sizeof((ab)->map_one));		\
144 } while (0)
145 #define	MAPADDRENABLED(flags) \
146 	(((flags) & AI_V4MAPPED) || \
147 	 (((flags) & AI_V4MAPPED_CFG)))
148 
149 union inx_addr {
150 	struct in_addr	in_addr;
151 #ifdef INET6
152 	struct in6_addr	in6_addr;
153 #endif
154 	struct {
155 		u_char	mau_zero[10];
156 		u_char	mau_one[2];
157 		struct in_addr mau_inaddr;
158 	}		map_addr_un;
159 #define	map_zero	map_addr_un.mau_zero
160 #define	map_one		map_addr_un.mau_one
161 #define	map_inaddr	map_addr_un.mau_inaddr
162 };
163 
164 struct policyqueue {
165 	TAILQ_ENTRY(policyqueue) pc_entry;
166 #ifdef INET6
167 	struct in6_addrpolicy pc_policy;
168 #endif
169 };
170 TAILQ_HEAD(policyhead, policyqueue);
171 
172 #define AIO_SRCFLAG_DEPRECATED	0x1
173 
174 struct hp_order {
175 	union {
176 		struct sockaddr_storage aiou_ss;
177 		struct sockaddr aiou_sa;
178 	} aio_src_un;
179 #define aio_srcsa aio_src_un.aiou_sa
180 	u_int32_t aio_srcflag;
181 	int aio_srcscope;
182 	int aio_dstscope;
183 	struct policyqueue *aio_srcpolicy;
184 	struct policyqueue *aio_dstpolicy;
185 	union {
186 		struct sockaddr_storage aiou_ss;
187 		struct sockaddr aiou_sa;
188 	} aio_un;
189 #define aio_sa aio_un.aiou_sa
190 	int aio_matchlen;
191 	char *aio_h_addr;
192 };
193 
194 static struct hostent	*_hpcopy(struct hostent *, int *);
195 static struct hostent	*_hpaddr(int, const char *, void *, int *);
196 #ifdef INET6
197 static struct hostent	*_hpmerge(struct hostent *, struct hostent *, int *);
198 static struct hostent	*_hpmapv6(struct hostent *, int *);
199 #endif
200 static struct hostent	*_hpsort(struct hostent *, res_state);
201 
202 static struct hostent	*_hpreorder(struct hostent *);
203 static int		 get_addrselectpolicy(struct policyhead *);
204 static void		 free_addrselectpolicy(struct policyhead *);
205 static struct policyqueue *match_addrselectpolicy(struct sockaddr *,
206 						  struct policyhead *);
207 static void		 set_source(struct hp_order *, struct policyhead *);
208 static int		 matchlen(struct sockaddr *, struct sockaddr *);
209 static int		 comp_dst(const void *, const void *);
210 static int		 gai_addr2scopetype(struct sockaddr *);
211 
212 /*
213  * Functions defined in RFC2553
214  *	getipnodebyname, getipnodebyaddr, freehostent
215  */
216 
217 struct hostent *
218 getipnodebyname(const char *name, int af, int flags, int *errp)
219 {
220 	struct hostent *hp;
221 	union inx_addr addrbuf;
222 	res_state statp;
223 	u_long options;
224 
225 	switch (af) {
226 	case AF_INET:
227 #ifdef INET6
228 	case AF_INET6:
229 #endif
230 		break;
231 	default:
232 		*errp = NO_RECOVERY;
233 		return NULL;
234 	}
235 
236 	if (flags & AI_ADDRCONFIG) {
237 		int s;
238 
239 		if ((s = _socket(af, SOCK_DGRAM, 0)) < 0)
240 			return NULL;
241 		/*
242 		 * TODO:
243 		 * Note that implementation dependent test for address
244 		 * configuration should be done everytime called
245 		 * (or apropriate interval),
246 		 * because addresses will be dynamically assigned or deleted.
247 		 */
248 		_close(s);
249 	}
250 
251 #ifdef INET6
252 	/* special case for literal address */
253 	if (inet_pton(AF_INET6, name, &addrbuf) == 1) {
254 		if (af != AF_INET6) {
255 			*errp = HOST_NOT_FOUND;
256 			return NULL;
257 		}
258 		return _hpaddr(af, name, &addrbuf, errp);
259 	}
260 #endif
261 	if (inet_aton(name, (struct in_addr *)&addrbuf) == 1) {
262 		if (af != AF_INET) {
263 			if (MAPADDRENABLED(flags)) {
264 				MAPADDR(&addrbuf, &addrbuf.in_addr);
265 			} else {
266 				*errp = HOST_NOT_FOUND;
267 				return NULL;
268 			}
269 		}
270 		return _hpaddr(af, name, &addrbuf, errp);
271 	}
272 
273 
274 	statp = __res_state();
275 	if ((statp->options & RES_INIT) == 0) {
276 		if (res_ninit(statp) < 0) {
277 			*errp = NETDB_INTERNAL;
278 			return NULL;
279 		}
280 	}
281 
282 	options = statp->options;
283 	statp->options &= ~RES_USE_INET6;
284 
285 	hp = gethostbyname2(name, af);
286 	hp = _hpcopy(hp, errp);
287 
288 #ifdef INET6
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 _hpreorder(_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 /*
635  * _hpreorder: sort address by default address selection
636  */
637 static struct hostent *
638 _hpreorder(struct hostent *hp)
639 {
640 	struct hp_order *aio;
641 	int i, n;
642 	char *ap;
643 	struct sockaddr *sa;
644 	struct policyhead policyhead;
645 
646 	if (hp == NULL)
647 		return hp;
648 
649 	switch (hp->h_addrtype) {
650 	case AF_INET:
651 #ifdef INET6
652 	case AF_INET6:
653 #endif
654 		break;
655 	default:
656 		free_addrselectpolicy(&policyhead);
657 		return hp;
658 	}
659 
660 	/* count the number of addrinfo elements for sorting. */
661 	for (n = 0; hp->h_addr_list[n] != NULL; n++)
662 		;
663 
664 	/*
665 	 * If the number is small enough, we can skip the reordering process.
666 	 */
667 	if (n <= 1)
668 		return hp;
669 
670 	/* allocate a temporary array for sort and initialization of it. */
671 	if ((aio = malloc(sizeof(*aio) * n)) == NULL)
672 		return hp;	/* give up reordering */
673 	memset(aio, 0, sizeof(*aio) * n);
674 
675 	/* retrieve address selection policy from the kernel */
676 	TAILQ_INIT(&policyhead);
677 	if (!get_addrselectpolicy(&policyhead)) {
678 		/* no policy is installed into kernel, we don't sort. */
679 		free(aio);
680 		return hp;
681 	}
682 
683 	for (i = 0; i < n; i++) {
684 		ap = hp->h_addr_list[i];
685 		aio[i].aio_h_addr = ap;
686 		sa = &aio[i].aio_sa;
687 		switch (hp->h_addrtype) {
688 		case AF_INET:
689 			sa->sa_family = AF_INET;
690 			sa->sa_len = sizeof(struct sockaddr_in);
691 			memcpy(&((struct sockaddr_in *)sa)->sin_addr, ap,
692 			    sizeof(struct in_addr));
693 			break;
694 #ifdef INET6
695 		case AF_INET6:
696 			if (IN6_IS_ADDR_V4MAPPED((struct in6_addr *)ap)) {
697 				sa->sa_family = AF_INET;
698 				sa->sa_len = sizeof(struct sockaddr_in);
699 				memcpy(&((struct sockaddr_in *)sa)->sin_addr,
700 				    &ap[12], sizeof(struct in_addr));
701 			} else {
702 				sa->sa_family = AF_INET6;
703 				sa->sa_len = sizeof(struct sockaddr_in6);
704 				memcpy(&((struct sockaddr_in6 *)sa)->sin6_addr,
705 				    ap, sizeof(struct in6_addr));
706 			}
707 			break;
708 #endif
709 		}
710 		aio[i].aio_dstscope = gai_addr2scopetype(sa);
711 		aio[i].aio_dstpolicy = match_addrselectpolicy(sa, &policyhead);
712 		set_source(&aio[i], &policyhead);
713 	}
714 
715 	/* perform sorting. */
716 	qsort(aio, n, sizeof(*aio), comp_dst);
717 
718 	/* reorder the h_addr_list. */
719 	for (i = 0; i < n; i++)
720 		hp->h_addr_list[i] = aio[i].aio_h_addr;
721 
722 	/* cleanup and return */
723 	free(aio);
724 	free_addrselectpolicy(&policyhead);
725 	return hp;
726 }
727 
728 static int
729 get_addrselectpolicy(struct policyhead *head)
730 {
731 #ifdef INET6
732 	int mib[] = { CTL_NET, PF_INET6, IPPROTO_IPV6, IPV6CTL_ADDRCTLPOLICY };
733 	size_t l;
734 	char *buf;
735 	struct in6_addrpolicy *pol, *ep;
736 
737 	if (sysctl(mib, sizeof(mib) / sizeof(mib[0]), NULL, &l, NULL, 0) < 0)
738 		return (0);
739 	if ((buf = malloc(l)) == NULL)
740 		return (0);
741 	if (sysctl(mib, sizeof(mib) / sizeof(mib[0]), buf, &l, NULL, 0) < 0) {
742 		free(buf);
743 		return (0);
744 	}
745 
746 	ep = (struct in6_addrpolicy *)(buf + l);
747 	for (pol = (struct in6_addrpolicy *)buf; pol + 1 <= ep; pol++) {
748 		struct policyqueue *new;
749 
750 		if ((new = malloc(sizeof(*new))) == NULL) {
751 			free_addrselectpolicy(head); /* make the list empty */
752 			break;
753 		}
754 		new->pc_policy = *pol;
755 		TAILQ_INSERT_TAIL(head, new, pc_entry);
756 	}
757 
758 	free(buf);
759 	return (1);
760 #else
761 	return (0);
762 #endif
763 }
764 
765 static void
766 free_addrselectpolicy(struct policyhead *head)
767 {
768 	struct policyqueue *ent, *nent;
769 
770 	for (ent = TAILQ_FIRST(head); ent; ent = nent) {
771 		nent = TAILQ_NEXT(ent, pc_entry);
772 		TAILQ_REMOVE(head, ent, pc_entry);
773 		free(ent);
774 	}
775 }
776 
777 static struct policyqueue *
778 match_addrselectpolicy(struct sockaddr *addr, struct policyhead *head)
779 {
780 #ifdef INET6
781 	struct policyqueue *ent, *bestent = NULL;
782 	struct in6_addrpolicy *pol;
783 	int matchlen, bestmatchlen = -1;
784 	u_char *mp, *ep, *k, *p, m;
785 	struct sockaddr_in6 key;
786 
787 	switch(addr->sa_family) {
788 	case AF_INET6:
789 		key = *(struct sockaddr_in6 *)addr;
790 		break;
791 	case AF_INET:
792 		/* convert the address into IPv4-mapped IPv6 address. */
793 		memset(&key, 0, sizeof(key));
794 		key.sin6_family = AF_INET6;
795 		key.sin6_len = sizeof(key);
796 		key.sin6_addr.s6_addr[10] = 0xff;
797 		key.sin6_addr.s6_addr[11] = 0xff;
798 		memcpy(&key.sin6_addr.s6_addr[12],
799 		       &((struct sockaddr_in *)addr)->sin_addr, 4);
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, IPPROTO_UDP)) < 0)
870 		return;		/* give up */
871 	if (_connect(s, (struct sockaddr *)&ss, ss.ss_len) < 0)
872 		goto cleanup;
873 	srclen = ss.ss_len;
874 	if (_getsockname(s, &aio->aio_srcsa, &srclen) < 0) {
875 		aio->aio_srcsa.sa_family = AF_UNSPEC;
876 		goto cleanup;
877 	}
878 	aio->aio_srcscope = gai_addr2scopetype(&aio->aio_srcsa);
879 	aio->aio_srcpolicy = match_addrselectpolicy(&aio->aio_srcsa, ph);
880 	aio->aio_matchlen = matchlen(&aio->aio_srcsa, (struct sockaddr *)&ss);
881 #ifdef INET6
882 	if (ss.ss_family == AF_INET6) {
883 		struct in6_ifreq ifr6;
884 		u_int32_t flags6;
885 
886 		/* XXX: interface name should not be hardcoded */
887 		strncpy(ifr6.ifr_name, "lo0", sizeof(ifr6.ifr_name));
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 
903 static int
904 matchlen(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 < addrlen * 8) {
933 				match++;
934 				r <<= 1;
935 			}
936 			break;
937 		} else
938 			match += 8;
939 	return(match);
940 }
941 
942 static int
943 comp_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 	return(-1);
1050 }
1051 
1052 /*
1053  * Copy from scope.c.
1054  * XXX: we should standardize the functions and link them as standard
1055  * library.
1056  */
1057 static int
1058 gai_addr2scopetype(struct sockaddr *sa)
1059 {
1060 #ifdef INET6
1061 	struct sockaddr_in6 *sa6;
1062 #endif
1063 	struct sockaddr_in *sa4;
1064 
1065 	switch(sa->sa_family) {
1066 #ifdef INET6
1067 	case AF_INET6:
1068 		sa6 = (struct sockaddr_in6 *)sa;
1069 		if (IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
1070 			/* just use the scope field of the multicast address */
1071 			return(sa6->sin6_addr.s6_addr[2] & 0x0f);
1072 		}
1073 		/*
1074 		 * Unicast addresses: map scope type to corresponding scope
1075 		 * value defined for multcast addresses.
1076 		 * XXX: hardcoded scope type values are bad...
1077 		 */
1078 		if (IN6_IS_ADDR_LOOPBACK(&sa6->sin6_addr))
1079 			return(1); /* node local scope */
1080 		if (IN6_IS_ADDR_LINKLOCAL(&sa6->sin6_addr))
1081 			return(2); /* link-local scope */
1082 		if (IN6_IS_ADDR_SITELOCAL(&sa6->sin6_addr))
1083 			return(5); /* site-local scope */
1084 		return(14);	/* global scope */
1085 		break;
1086 #endif
1087 	case AF_INET:
1088 		/*
1089 		 * IPv4 pseudo scoping according to RFC 3484.
1090 		 */
1091 		sa4 = (struct sockaddr_in *)sa;
1092 		/* IPv4 autoconfiguration addresses have link-local scope. */
1093 		if (((u_char *)&sa4->sin_addr)[0] == 169 &&
1094 		    ((u_char *)&sa4->sin_addr)[1] == 254)
1095 			return(2);
1096 		/* Private addresses have site-local scope. */
1097 		if (((u_char *)&sa4->sin_addr)[0] == 10 ||
1098 		    (((u_char *)&sa4->sin_addr)[0] == 172 &&
1099 		     (((u_char *)&sa4->sin_addr)[1] & 0xf0) == 16) ||
1100 		    (((u_char *)&sa4->sin_addr)[0] == 192 &&
1101 		     ((u_char *)&sa4->sin_addr)[1] == 168))
1102 			return(14);	/* XXX: It should be 5 unless NAT */
1103 		/* Loopback addresses have link-local scope. */
1104 		if (((u_char *)&sa4->sin_addr)[0] == 127)
1105 			return(2);
1106 		return(14);
1107 		break;
1108 	default:
1109 		errno = EAFNOSUPPORT; /* is this a good error? */
1110 		return(-1);
1111 	}
1112 }
1113