xref: /freebsd/lib/libc/net/name6.c (revision dd41de95a84d979615a2ef11df6850622bf6184e)
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 <sys/cdefs.h>
89 __FBSDID("$FreeBSD$");
90 
91 #include "namespace.h"
92 #include <sys/param.h>
93 #include <sys/socket.h>
94 #include <sys/time.h>
95 #include <sys/queue.h>
96 #include <netinet/in.h>
97 #ifdef INET6
98 #include <net/if.h>
99 #include <sys/sysctl.h>
100 #include <sys/ioctl.h>
101 #include <netinet6/in6_var.h>	/* XXX */
102 #endif
103 
104 #include <arpa/inet.h>
105 #include <arpa/nameser.h>
106 
107 #include <errno.h>
108 #include <netdb.h>
109 #include <resolv.h>
110 #include <stdio.h>
111 #include <stdlib.h>
112 #include <string.h>
113 #include <stdarg.h>
114 #include <nsswitch.h>
115 #include <unistd.h>
116 #include "un-namespace.h"
117 #include "netdb_private.h"
118 #include "res_private.h"
119 
120 #ifndef MAXALIASES
121 #define	MAXALIASES	10
122 #endif
123 #ifndef	MAXADDRS
124 #define	MAXADDRS	20
125 #endif
126 #ifndef MAXDNAME
127 #define	MAXDNAME	1025
128 #endif
129 
130 #ifdef INET6
131 #define	ADDRLEN(af)	((af) == AF_INET6 ? sizeof(struct in6_addr) : \
132 					    sizeof(struct in_addr))
133 #else
134 #define	ADDRLEN(af)	sizeof(struct in_addr)
135 #endif
136 
137 #define	MAPADDR(ab, ina) \
138 do {									\
139 	memcpy(&(ab)->map_inaddr, ina, sizeof(struct in_addr));		\
140 	memset((ab)->map_zero, 0, sizeof((ab)->map_zero));		\
141 	memset((ab)->map_one, 0xff, sizeof((ab)->map_one));		\
142 } while (0)
143 #define	MAPADDRENABLED(flags) \
144 	(((flags) & AI_V4MAPPED) || \
145 	 (((flags) & AI_V4MAPPED_CFG)))
146 
147 union inx_addr {
148 	struct in_addr	in_addr;
149 #ifdef INET6
150 	struct in6_addr	in6_addr;
151 #endif
152 	struct {
153 		u_char	mau_zero[10];
154 		u_char	mau_one[2];
155 		struct in_addr mau_inaddr;
156 	}		map_addr_un;
157 #define	map_zero	map_addr_un.mau_zero
158 #define	map_one		map_addr_un.mau_one
159 #define	map_inaddr	map_addr_un.mau_inaddr
160 };
161 
162 struct policyqueue {
163 	TAILQ_ENTRY(policyqueue) pc_entry;
164 #ifdef INET6
165 	struct in6_addrpolicy pc_policy;
166 #endif
167 };
168 TAILQ_HEAD(policyhead, policyqueue);
169 
170 #define AIO_SRCFLAG_DEPRECATED	0x1
171 
172 struct hp_order {
173 	union {
174 		struct sockaddr_storage aiou_ss;
175 		struct sockaddr aiou_sa;
176 	} aio_src_un;
177 #define aio_srcsa aio_src_un.aiou_sa
178 	u_int32_t aio_srcflag;
179 	int aio_srcscope;
180 	int aio_dstscope;
181 	struct policyqueue *aio_srcpolicy;
182 	struct policyqueue *aio_dstpolicy;
183 	union {
184 		struct sockaddr_storage aiou_ss;
185 		struct sockaddr aiou_sa;
186 	} aio_un;
187 #define aio_sa aio_un.aiou_sa
188 	int aio_matchlen;
189 	char *aio_h_addr;
190 	int aio_initial_sequence;
191 };
192 
193 static struct	 hostent *_hpcopy(struct hostent *, int *);
194 static struct	 hostent *_hpaddr(int, const char *, void *, int *);
195 #ifdef INET6
196 static struct	 hostent *_hpmerge(struct hostent *, struct hostent *, int *);
197 static struct	 hostent *_hpmapv6(struct hostent *, int *);
198 #endif
199 static struct	 hostent *_hpsort(struct hostent *, res_state);
200 
201 #ifdef INET6
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 #endif
212 
213 /*
214  * Functions defined in RFC2553
215  *	getipnodebyname, getipnodebyaddr, freehostent
216  */
217 
218 struct hostent *
219 getipnodebyname(const char *name, int af, int flags, int *errp)
220 {
221 	struct hostent *hp;
222 	union inx_addr addrbuf;
223 	res_state statp;
224 	u_long options;
225 
226 	switch (af) {
227 	case AF_INET:
228 #ifdef INET6
229 	case AF_INET6:
230 #endif
231 		break;
232 	default:
233 		*errp = NO_RECOVERY;
234 		return NULL;
235 	}
236 
237 	if (flags & AI_ADDRCONFIG) {
238 		int s;
239 
240 		if ((s = _socket(af, SOCK_DGRAM | SOCK_CLOEXEC, 0)) < 0)
241 			return NULL;
242 		/*
243 		 * TODO:
244 		 * Note that implementation dependent test for address
245 		 * configuration should be done every time called
246 		 * (or appropriate interval),
247 		 * because addresses will be dynamically assigned or deleted.
248 		 */
249 		_close(s);
250 	}
251 
252 #ifdef INET6
253 	/* special case for literal address */
254 	if (inet_pton(AF_INET6, name, &addrbuf) == 1) {
255 		if (af != AF_INET6) {
256 			*errp = HOST_NOT_FOUND;
257 			return NULL;
258 		}
259 		return _hpaddr(af, name, &addrbuf, errp);
260 	}
261 #endif
262 	if (inet_aton(name, (struct in_addr *)&addrbuf) == 1) {
263 		if (af != AF_INET) {
264 			if (MAPADDRENABLED(flags)) {
265 				MAPADDR(&addrbuf, &addrbuf.in_addr);
266 			} else {
267 				*errp = HOST_NOT_FOUND;
268 				return NULL;
269 			}
270 		}
271 		return _hpaddr(af, name, &addrbuf, errp);
272 	}
273 
274 
275 	statp = __res_state();
276 	if ((statp->options & RES_INIT) == 0) {
277 		if (res_ninit(statp) < 0) {
278 			*errp = NETDB_INTERNAL;
279 			return NULL;
280 		}
281 	}
282 
283 	options = statp->options;
284 	statp->options &= ~RES_USE_INET6;
285 
286 	hp = gethostbyname2(name, af);
287 	hp = _hpcopy(hp, errp);
288 #ifdef INET6
289 	if (af == AF_INET6)
290 		hp = _hpreorder(hp);
291 
292 	if (af == AF_INET6 && ((flags & AI_ALL) || hp == NULL) &&
293 	    MAPADDRENABLED(flags)) {
294 		struct hostent *hp2 = gethostbyname2(name, AF_INET);
295 		if (hp == NULL)
296 			if (hp2 == NULL)
297 				*errp = statp->res_h_errno;
298 			else
299 				hp = _hpmapv6(hp2, errp);
300 		else {
301 			if (hp2 && strcmp(hp->h_name, hp2->h_name) == 0) {
302 				struct hostent *hpb = hp;
303 				hp = _hpmerge(hpb, hp2, errp);
304 				freehostent(hpb);
305 			}
306 		}
307 	}
308 #endif
309 
310 	if (hp == NULL)
311 		*errp = statp->res_h_errno;
312 
313 	statp->options = options;
314 	return _hpsort(hp, statp);
315 }
316 
317 struct hostent *
318 getipnodebyaddr(const void *src, size_t len, int af, int *errp)
319 {
320 	struct hostent *hp;
321 	res_state statp;
322 	u_long options;
323 
324 #ifdef INET6
325 	struct in6_addr addrbuf;
326 #else
327 	struct in_addr addrbuf;
328 #endif
329 
330 	switch (af) {
331 	case AF_INET:
332 		if (len != sizeof(struct in_addr)) {
333 			*errp = NO_RECOVERY;
334 			return NULL;
335 		}
336 		if (rounddown2((long)src, sizeof(struct in_addr))) {
337 			memcpy(&addrbuf, src, len);
338 			src = &addrbuf;
339 		}
340 		if (((struct in_addr *)src)->s_addr == 0)
341 			return NULL;
342 		break;
343 #ifdef INET6
344 	case AF_INET6:
345 		if (len != sizeof(struct in6_addr)) {
346 			*errp = NO_RECOVERY;
347 			return NULL;
348 		}
349 		if (rounddown2((long)src, sizeof(struct in6_addr) / 2)) {
350 			/* XXX */
351 			memcpy(&addrbuf, src, len);
352 			src = &addrbuf;
353 		}
354 		if (IN6_IS_ADDR_UNSPECIFIED((struct in6_addr *)src))
355 			return NULL;
356 		if (IN6_IS_ADDR_V4MAPPED((struct in6_addr *)src)
357 		||  IN6_IS_ADDR_V4COMPAT((struct in6_addr *)src)) {
358 			src = (char *)src +
359 			    (sizeof(struct in6_addr) - sizeof(struct in_addr));
360 			af = AF_INET;
361 			len = sizeof(struct in_addr);
362 		}
363 		break;
364 #endif
365 	default:
366 		*errp = NO_RECOVERY;
367 		return NULL;
368 	}
369 
370 	statp = __res_state();
371 	if ((statp->options & RES_INIT) == 0) {
372 		if (res_ninit(statp) < 0) {
373 			RES_SET_H_ERRNO(statp, NETDB_INTERNAL);
374 			return NULL;
375 		}
376 	}
377 
378 	options = statp->options;
379 	statp->options &= ~RES_USE_INET6;
380 
381 	hp = gethostbyaddr(src, len, af);
382 	if (hp == NULL)
383 		*errp = statp->res_h_errno;
384 
385 	statp->options = options;
386 	return (_hpcopy(hp, errp));
387 }
388 
389 void
390 freehostent(struct hostent *ptr)
391 {
392 	free(ptr);
393 }
394 
395 /*
396  * Private utility functions
397  */
398 
399 /*
400  * _hpcopy: allocate and copy hostent structure
401  */
402 static struct hostent *
403 _hpcopy(struct hostent *hp, int *errp)
404 {
405 	struct hostent *nhp;
406 	char *cp, **pp;
407 	int size, addrsize;
408 	int nalias = 0, naddr = 0;
409 	int al_off;
410 	int i;
411 
412 	if (hp == NULL)
413 		return hp;
414 
415 	/* count size to be allocated */
416 	size = sizeof(struct hostent);
417 	if (hp->h_name != NULL)
418 		size += strlen(hp->h_name) + 1;
419 	if ((pp = hp->h_aliases) != NULL) {
420 		for (i = 0; *pp != NULL; i++, pp++) {
421 			if (**pp != '\0') {
422 				size += strlen(*pp) + 1;
423 				nalias++;
424 			}
425 		}
426 	}
427 	/* adjust alignment */
428 	size = ALIGN(size);
429 	al_off = size;
430 	size += sizeof(char *) * (nalias + 1);
431 	addrsize = ALIGN(hp->h_length);
432 	if ((pp = hp->h_addr_list) != NULL) {
433 		while (*pp++ != NULL)
434 			naddr++;
435 	}
436 	size += addrsize * naddr;
437 	size += sizeof(char *) * (naddr + 1);
438 
439 	/* copy */
440 	if ((nhp = (struct hostent *)malloc(size)) == NULL) {
441 		*errp = TRY_AGAIN;
442 		return NULL;
443 	}
444 	cp = (char *)&nhp[1];
445 	if (hp->h_name != NULL) {
446 		nhp->h_name = cp;
447 		strcpy(cp, hp->h_name);
448 		cp += strlen(cp) + 1;
449 	} else
450 		nhp->h_name = NULL;
451 	nhp->h_aliases = (char **)((char *)nhp + al_off);
452 	if ((pp = hp->h_aliases) != NULL) {
453 		for (i = 0; *pp != NULL; pp++) {
454 			if (**pp != '\0') {
455 				nhp->h_aliases[i++] = cp;
456 				strcpy(cp, *pp);
457 				cp += strlen(cp) + 1;
458 			}
459 		}
460 	}
461 	nhp->h_aliases[nalias] = NULL;
462 	cp = (char *)&nhp->h_aliases[nalias + 1];
463 	nhp->h_addrtype = hp->h_addrtype;
464 	nhp->h_length = hp->h_length;
465 	nhp->h_addr_list = (char **)cp;
466 	if ((pp = hp->h_addr_list) != NULL) {
467 		cp = (char *)&nhp->h_addr_list[naddr + 1];
468 		for (i = 0; *pp != NULL; pp++) {
469 			nhp->h_addr_list[i++] = cp;
470 			memcpy(cp, *pp, hp->h_length);
471 			cp += addrsize;
472 		}
473 	}
474 	nhp->h_addr_list[naddr] = NULL;
475 	return nhp;
476 }
477 
478 /*
479  * _hpaddr: construct hostent structure with one address
480  */
481 static struct hostent *
482 _hpaddr(int af, const char *name, void *addr, int *errp)
483 {
484 	struct hostent *hp, hpbuf;
485 	char *addrs[2];
486 
487 	hp = &hpbuf;
488 	hp->h_name = (char *)name;
489 	hp->h_aliases = NULL;
490 	hp->h_addrtype = af;
491 	hp->h_length = ADDRLEN(af);
492 	hp->h_addr_list = addrs;
493 	addrs[0] = (char *)addr;
494 	addrs[1] = NULL;
495 	return (_hpcopy(hp, errp));
496 }
497 
498 #ifdef INET6
499 /*
500  * _hpmerge: merge 2 hostent structure, arguments will be freed
501  */
502 static struct hostent *
503 _hpmerge(struct hostent *hp1, struct hostent *hp2, int *errp)
504 {
505 	int i, j;
506 	int naddr, nalias;
507 	char **pp;
508 	struct hostent *hp, hpbuf;
509 	char *aliases[MAXALIASES + 1], *addrs[MAXADDRS + 1];
510 	union inx_addr addrbuf[MAXADDRS];
511 
512 	if (hp1 == NULL)
513 		return _hpcopy(hp2, errp);
514 	if (hp2 == NULL)
515 		return _hpcopy(hp1, errp);
516 
517 #define	HP(i)	(i == 1 ? hp1 : hp2)
518 	hp = &hpbuf;
519 	hp->h_name = (hp1->h_name != NULL ? hp1->h_name : hp2->h_name);
520 	hp->h_aliases = aliases;
521 	nalias = 0;
522 	for (i = 1; i <= 2; i++) {
523 		if ((pp = HP(i)->h_aliases) == NULL)
524 			continue;
525 		for (; nalias < MAXALIASES && *pp != NULL; pp++) {
526 			/* check duplicates */
527 			for (j = 0; j < nalias; j++)
528 				if (strcasecmp(*pp, aliases[j]) == 0)
529 					break;
530 			if (j == nalias)
531 				aliases[nalias++] = *pp;
532 		}
533 	}
534 	aliases[nalias] = NULL;
535 	if (hp1->h_length != hp2->h_length) {
536 		hp->h_addrtype = AF_INET6;
537 		hp->h_length = sizeof(struct in6_addr);
538 	} else {
539 		hp->h_addrtype = hp1->h_addrtype;
540 		hp->h_length = hp1->h_length;
541 	}
542 
543 	hp->h_addr_list = addrs;
544 	naddr = 0;
545 	for (i = 1; i <= 2; i++) {
546 		if ((pp = HP(i)->h_addr_list) == NULL)
547 			continue;
548 		if (HP(i)->h_length == hp->h_length) {
549 			while (naddr < MAXADDRS && *pp != NULL)
550 				addrs[naddr++] = *pp++;
551 		} else {
552 			/* copy IPv4 addr as mapped IPv6 addr */
553 			while (naddr < MAXADDRS && *pp != NULL) {
554 				MAPADDR(&addrbuf[naddr], *pp++);
555 				addrs[naddr] = (char *)&addrbuf[naddr];
556 				naddr++;
557 			}
558 		}
559 	}
560 	addrs[naddr] = NULL;
561 	return (_hpcopy(hp, errp));
562 }
563 #endif
564 
565 /*
566  * _hpmapv6: convert IPv4 hostent into IPv4-mapped IPv6 addresses
567  */
568 #ifdef INET6
569 static struct hostent *
570 _hpmapv6(struct hostent *hp, int *errp)
571 {
572 	struct hostent hp6;
573 
574 	if (hp == NULL)
575 		return NULL;
576 	if (hp->h_addrtype == AF_INET6)
577 		return _hpcopy(hp, errp);
578 
579 	memset(&hp6, 0, sizeof(struct hostent));
580 	hp6.h_addrtype = AF_INET6;
581 	hp6.h_length = sizeof(struct in6_addr);
582 	return _hpmerge(&hp6, hp, errp);
583 }
584 #endif
585 
586 /*
587  * _hpsort: sort address by sortlist
588  */
589 static struct hostent *
590 _hpsort(struct hostent *hp, res_state statp)
591 {
592 	int i, j, n;
593 	u_char *ap, *sp, *mp, **pp;
594 	char t;
595 	char order[MAXADDRS];
596 	int nsort = statp->nsort;
597 
598 	if (hp == NULL || hp->h_addr_list[1] == NULL || nsort == 0)
599 		return hp;
600 	for (i = 0; (ap = (u_char *)hp->h_addr_list[i]); i++) {
601 		for (j = 0; j < nsort; j++) {
602 #ifdef INET6
603 			if (statp->_u._ext.ext->sort_list[j].af !=
604 			    hp->h_addrtype)
605 				continue;
606 			sp = (u_char *)&statp->_u._ext.ext->sort_list[j].addr;
607 			mp = (u_char *)&statp->_u._ext.ext->sort_list[j].mask;
608 #else
609 			sp = (u_char *)&statp->sort_list[j].addr;
610 			mp = (u_char *)&statp->sort_list[j].mask;
611 #endif
612 			for (n = 0; n < hp->h_length; n++) {
613 				if ((ap[n] & mp[n]) != sp[n])
614 					break;
615 			}
616 			if (n == hp->h_length)
617 				break;
618 		}
619 		order[i] = j;
620 	}
621 	n = i;
622 	pp = (u_char **)hp->h_addr_list;
623 	for (i = 0; i < n - 1; i++) {
624 		for (j = i + 1; j < n; j++) {
625 			if (order[i] > order[j]) {
626 				ap = pp[i];
627 				pp[i] = pp[j];
628 				pp[j] = ap;
629 				t = order[i];
630 				order[i] = order[j];
631 				order[j] = t;
632 			}
633 		}
634 	}
635 	return hp;
636 }
637 
638 #ifdef INET6
639 /*
640  * _hpreorder: sort address by default address selection
641  */
642 static struct hostent *
643 _hpreorder(struct hostent *hp)
644 {
645 	struct hp_order *aio;
646 	int i, n;
647 	char *ap;
648 	struct sockaddr *sa;
649 	struct policyhead policyhead;
650 
651 	if (hp == NULL)
652 		return hp;
653 
654 	switch (hp->h_addrtype) {
655 	case AF_INET:
656 #ifdef INET6
657 	case AF_INET6:
658 #endif
659 		break;
660 	default:
661 		return hp;
662 	}
663 
664 	/* count the number of addrinfo elements for sorting. */
665 	for (n = 0; hp->h_addr_list[n] != NULL; n++)
666 		;
667 
668 	/*
669 	 * If the number is small enough, we can skip the reordering process.
670 	 */
671 	if (n <= 1)
672 		return hp;
673 
674 	/* allocate a temporary array for sort and initialization of it. */
675 	if ((aio = malloc(sizeof(*aio) * n)) == NULL)
676 		return hp;	/* give up reordering */
677 	memset(aio, 0, sizeof(*aio) * n);
678 
679 	/* retrieve address selection policy from the kernel */
680 	TAILQ_INIT(&policyhead);
681 	if (!get_addrselectpolicy(&policyhead)) {
682 		/* no policy is installed into kernel, we don't sort. */
683 		free(aio);
684 		return hp;
685 	}
686 
687 	for (i = 0; i < n; i++) {
688 		ap = hp->h_addr_list[i];
689 		aio[i].aio_h_addr = ap;
690 		sa = &aio[i].aio_sa;
691 		switch (hp->h_addrtype) {
692 		case AF_INET:
693 			sa->sa_family = AF_INET;
694 			sa->sa_len = sizeof(struct sockaddr_in);
695 			memcpy(&((struct sockaddr_in *)sa)->sin_addr, ap,
696 			    sizeof(struct in_addr));
697 			break;
698 #ifdef INET6
699 		case AF_INET6:
700 			if (IN6_IS_ADDR_V4MAPPED((struct in6_addr *)ap)) {
701 				sa->sa_family = AF_INET;
702 				sa->sa_len = sizeof(struct sockaddr_in);
703 				memcpy(&((struct sockaddr_in *)sa)->sin_addr,
704 				    &ap[12], sizeof(struct in_addr));
705 			} else {
706 				sa->sa_family = AF_INET6;
707 				sa->sa_len = sizeof(struct sockaddr_in6);
708 				memcpy(&((struct sockaddr_in6 *)sa)->sin6_addr,
709 				    ap, sizeof(struct in6_addr));
710 			}
711 			break;
712 #endif
713 		}
714 		aio[i].aio_dstscope = gai_addr2scopetype(sa);
715 		aio[i].aio_dstpolicy = match_addrselectpolicy(sa, &policyhead);
716 		set_source(&aio[i], &policyhead);
717 		aio[i].aio_initial_sequence = i;
718 	}
719 
720 	/* perform sorting. */
721 	qsort(aio, n, sizeof(*aio), comp_dst);
722 
723 	/* reorder the h_addr_list. */
724 	for (i = 0; i < n; i++)
725 		hp->h_addr_list[i] = aio[i].aio_h_addr;
726 
727 	/* cleanup and return */
728 	free(aio);
729 	free_addrselectpolicy(&policyhead);
730 	return hp;
731 }
732 
733 static int
734 get_addrselectpolicy(struct policyhead *head)
735 {
736 #ifdef INET6
737 	int mib[] = { CTL_NET, PF_INET6, IPPROTO_IPV6, IPV6CTL_ADDRCTLPOLICY };
738 	size_t l;
739 	char *buf;
740 	struct in6_addrpolicy *pol, *ep;
741 
742 	if (sysctl(mib, nitems(mib), NULL, &l, NULL, 0) < 0)
743 		return (0);
744 	if ((buf = malloc(l)) == NULL)
745 		return (0);
746 	if (sysctl(mib, nitems(mib), buf, &l, NULL, 0) < 0) {
747 		free(buf);
748 		return (0);
749 	}
750 
751 	ep = (struct in6_addrpolicy *)(buf + l);
752 	for (pol = (struct in6_addrpolicy *)buf; pol + 1 <= ep; pol++) {
753 		struct policyqueue *new;
754 
755 		if ((new = malloc(sizeof(*new))) == NULL) {
756 			free_addrselectpolicy(head); /* make the list empty */
757 			break;
758 		}
759 		new->pc_policy = *pol;
760 		TAILQ_INSERT_TAIL(head, new, pc_entry);
761 	}
762 
763 	free(buf);
764 	return (1);
765 #else
766 	return (0);
767 #endif
768 }
769 
770 static void
771 free_addrselectpolicy(struct policyhead *head)
772 {
773 	struct policyqueue *ent, *nent;
774 
775 	for (ent = TAILQ_FIRST(head); ent; ent = nent) {
776 		nent = TAILQ_NEXT(ent, pc_entry);
777 		TAILQ_REMOVE(head, ent, pc_entry);
778 		free(ent);
779 	}
780 }
781 
782 static struct policyqueue *
783 match_addrselectpolicy(struct sockaddr *addr, struct policyhead *head)
784 {
785 #ifdef INET6
786 	struct policyqueue *ent, *bestent = NULL;
787 	struct in6_addrpolicy *pol;
788 	int matchlen, bestmatchlen = -1;
789 	u_char *mp, *ep, *k, *p, m;
790 	struct sockaddr_in6 key;
791 
792 	switch(addr->sa_family) {
793 	case AF_INET6:
794 		key = *(struct sockaddr_in6 *)addr;
795 		break;
796 	case AF_INET:
797 		/* convert the address into IPv4-mapped IPv6 address. */
798 		memset(&key, 0, sizeof(key));
799 		key.sin6_family = AF_INET6;
800 		key.sin6_len = sizeof(key);
801 		_map_v4v6_address(
802 		    (char *)&((struct sockaddr_in *)addr)->sin_addr,
803 		    (char *)&key.sin6_addr);
804 		break;
805 	default:
806 		return(NULL);
807 	}
808 
809 	for (ent = TAILQ_FIRST(head); ent; ent = TAILQ_NEXT(ent, pc_entry)) {
810 		pol = &ent->pc_policy;
811 		matchlen = 0;
812 
813 		mp = (u_char *)&pol->addrmask.sin6_addr;
814 		ep = mp + 16;	/* XXX: scope field? */
815 		k = (u_char *)&key.sin6_addr;
816 		p = (u_char *)&pol->addr.sin6_addr;
817 		for (; mp < ep && *mp; mp++, k++, p++) {
818 			m = *mp;
819 			if ((*k & m) != *p)
820 				goto next; /* not match */
821 			if (m == 0xff) /* short cut for a typical case */
822 				matchlen += 8;
823 			else {
824 				while (m >= 0x80) {
825 					matchlen++;
826 					m <<= 1;
827 				}
828 			}
829 		}
830 
831 		/* matched.  check if this is better than the current best. */
832 		if (matchlen > bestmatchlen) {
833 			bestent = ent;
834 			bestmatchlen = matchlen;
835 		}
836 
837 	  next:
838 		continue;
839 	}
840 
841 	return(bestent);
842 #else
843 	return(NULL);
844 #endif
845 
846 }
847 
848 static void
849 set_source(struct hp_order *aio, struct policyhead *ph)
850 {
851 	struct sockaddr_storage ss = aio->aio_un.aiou_ss;
852 	socklen_t srclen;
853 	int s;
854 
855 	/* set unspec ("no source is available"), just in case */
856 	aio->aio_srcsa.sa_family = AF_UNSPEC;
857 	aio->aio_srcscope = -1;
858 
859 	switch(ss.ss_family) {
860 	case AF_INET:
861 		((struct sockaddr_in *)&ss)->sin_port = htons(1);
862 		break;
863 #ifdef INET6
864 	case AF_INET6:
865 		((struct sockaddr_in6 *)&ss)->sin6_port = htons(1);
866 		break;
867 #endif
868 	default:		/* ignore unsupported AFs explicitly */
869 		return;
870 	}
871 
872 	/* open a socket to get the source address for the given dst */
873 	if ((s = _socket(ss.ss_family, SOCK_DGRAM | SOCK_CLOEXEC,
874 	    IPPROTO_UDP)) < 0)
875 		return;		/* give up */
876 	if (_connect(s, (struct sockaddr *)&ss, ss.ss_len) < 0)
877 		goto cleanup;
878 	srclen = ss.ss_len;
879 	if (_getsockname(s, &aio->aio_srcsa, &srclen) < 0) {
880 		aio->aio_srcsa.sa_family = AF_UNSPEC;
881 		goto cleanup;
882 	}
883 	aio->aio_srcscope = gai_addr2scopetype(&aio->aio_srcsa);
884 	aio->aio_srcpolicy = match_addrselectpolicy(&aio->aio_srcsa, ph);
885 	aio->aio_matchlen = matchlen(&aio->aio_srcsa, (struct sockaddr *)&ss);
886 #ifdef INET6
887 	if (ss.ss_family == AF_INET6) {
888 		struct in6_ifreq ifr6;
889 		u_int32_t flags6;
890 
891 		memset(&ifr6, 0, sizeof(ifr6));
892 		memcpy(&ifr6.ifr_addr, &ss, ss.ss_len);
893 		if (_ioctl(s, SIOCGIFAFLAG_IN6, &ifr6) == 0) {
894 			flags6 = ifr6.ifr_ifru.ifru_flags6;
895 			if ((flags6 & IN6_IFF_DEPRECATED))
896 				aio->aio_srcflag |= AIO_SRCFLAG_DEPRECATED;
897 		}
898 	}
899 #endif
900 
901   cleanup:
902 	_close(s);
903 	return;
904 }
905 
906 static int
907 matchlen(struct sockaddr *src, struct sockaddr *dst)
908 {
909 	int match = 0;
910 	u_char *s, *d;
911 	u_char *lim, r;
912 	int addrlen;
913 
914 	switch (src->sa_family) {
915 #ifdef INET6
916 	case AF_INET6:
917 		s = (u_char *)&((struct sockaddr_in6 *)src)->sin6_addr;
918 		d = (u_char *)&((struct sockaddr_in6 *)dst)->sin6_addr;
919 		addrlen = sizeof(struct in6_addr);
920 		lim = s + addrlen;
921 		break;
922 #endif
923 	case AF_INET:
924 		s = (u_char *)&((struct sockaddr_in *)src)->sin_addr;
925 		d = (u_char *)&((struct sockaddr_in *)dst)->sin_addr;
926 		addrlen = sizeof(struct in_addr);
927 		lim = s + addrlen;
928 		break;
929 	default:
930 		return(0);
931 	}
932 
933 	while (s < lim)
934 		if ((r = (*d++ ^ *s++)) != 0) {
935 			while ((r & 0x80) == 0) {
936 				match++;
937 				r <<= 1;
938 			}
939 			break;
940 		} else
941 			match += 8;
942 	return(match);
943 }
944 
945 static int
946 comp_dst(const void *arg1, const void *arg2)
947 {
948 	const struct hp_order *dst1 = arg1, *dst2 = arg2;
949 
950 	/*
951 	 * Rule 1: Avoid unusable destinations.
952 	 * XXX: we currently do not consider if an appropriate route exists.
953 	 */
954 	if (dst1->aio_srcsa.sa_family != AF_UNSPEC &&
955 	    dst2->aio_srcsa.sa_family == AF_UNSPEC) {
956 		return(-1);
957 	}
958 	if (dst1->aio_srcsa.sa_family == AF_UNSPEC &&
959 	    dst2->aio_srcsa.sa_family != AF_UNSPEC) {
960 		return(1);
961 	}
962 
963 	/* Rule 2: Prefer matching scope. */
964 	if (dst1->aio_dstscope == dst1->aio_srcscope &&
965 	    dst2->aio_dstscope != dst2->aio_srcscope) {
966 		return(-1);
967 	}
968 	if (dst1->aio_dstscope != dst1->aio_srcscope &&
969 	    dst2->aio_dstscope == dst2->aio_srcscope) {
970 		return(1);
971 	}
972 
973 	/* Rule 3: Avoid deprecated addresses. */
974 	if (dst1->aio_srcsa.sa_family != AF_UNSPEC &&
975 	    dst2->aio_srcsa.sa_family != AF_UNSPEC) {
976 		if (!(dst1->aio_srcflag & AIO_SRCFLAG_DEPRECATED) &&
977 		    (dst2->aio_srcflag & AIO_SRCFLAG_DEPRECATED)) {
978 			return(-1);
979 		}
980 		if ((dst1->aio_srcflag & AIO_SRCFLAG_DEPRECATED) &&
981 		    !(dst2->aio_srcflag & AIO_SRCFLAG_DEPRECATED)) {
982 			return(1);
983 		}
984 	}
985 
986 	/* Rule 4: Prefer home addresses. */
987 	/* XXX: not implemented yet */
988 
989 	/* Rule 5: Prefer matching label. */
990 #ifdef INET6
991 	if (dst1->aio_srcpolicy && dst1->aio_dstpolicy &&
992 	    dst1->aio_srcpolicy->pc_policy.label ==
993 	    dst1->aio_dstpolicy->pc_policy.label &&
994 	    (dst2->aio_srcpolicy == NULL || dst2->aio_dstpolicy == NULL ||
995 	     dst2->aio_srcpolicy->pc_policy.label !=
996 	     dst2->aio_dstpolicy->pc_policy.label)) {
997 		return(-1);
998 	}
999 	if (dst2->aio_srcpolicy && dst2->aio_dstpolicy &&
1000 	    dst2->aio_srcpolicy->pc_policy.label ==
1001 	    dst2->aio_dstpolicy->pc_policy.label &&
1002 	    (dst1->aio_srcpolicy == NULL || dst1->aio_dstpolicy == NULL ||
1003 	     dst1->aio_srcpolicy->pc_policy.label !=
1004 	     dst1->aio_dstpolicy->pc_policy.label)) {
1005 		return(1);
1006 	}
1007 #endif
1008 
1009 	/* Rule 6: Prefer higher precedence. */
1010 #ifdef INET6
1011 	if (dst1->aio_dstpolicy &&
1012 	    (dst2->aio_dstpolicy == NULL ||
1013 	     dst1->aio_dstpolicy->pc_policy.preced >
1014 	     dst2->aio_dstpolicy->pc_policy.preced)) {
1015 		return(-1);
1016 	}
1017 	if (dst2->aio_dstpolicy &&
1018 	    (dst1->aio_dstpolicy == NULL ||
1019 	     dst2->aio_dstpolicy->pc_policy.preced >
1020 	     dst1->aio_dstpolicy->pc_policy.preced)) {
1021 		return(1);
1022 	}
1023 #endif
1024 
1025 	/* Rule 7: Prefer native transport. */
1026 	/* XXX: not implemented yet */
1027 
1028 	/* Rule 8: Prefer smaller scope. */
1029 	if (dst1->aio_dstscope >= 0 &&
1030 	    dst1->aio_dstscope < dst2->aio_dstscope) {
1031 		return(-1);
1032 	}
1033 	if (dst2->aio_dstscope >= 0 &&
1034 	    dst2->aio_dstscope < dst1->aio_dstscope) {
1035 		return(1);
1036 	}
1037 
1038 	/*
1039 	 * Rule 9: Use longest matching prefix.
1040 	 * We compare the match length in a same AF only.
1041 	 */
1042 	if (dst1->aio_sa.sa_family == dst2->aio_sa.sa_family) {
1043 		if (dst1->aio_matchlen > dst2->aio_matchlen) {
1044 			return(-1);
1045 		}
1046 		if (dst1->aio_matchlen < dst2->aio_matchlen) {
1047 			return(1);
1048 		}
1049 	}
1050 
1051 	/* Rule 10: Otherwise, leave the order unchanged. */
1052 
1053 	/*
1054 	 * Note that qsort is unstable; so, we can't return zero and
1055 	 * expect the order to be unchanged.
1056 	 * That also means we can't depend on the current position of
1057 	 * dst2 being after dst1.  We must enforce the initial order
1058 	 * with an explicit compare on the original position.
1059 	 * The qsort specification requires that "When the same objects
1060 	 * (consisting of width bytes, irrespective of their current
1061 	 * positions in the array) are passed more than once to the
1062 	 * comparison function, the results shall be consistent with one
1063 	 * another."
1064 	 * In other words, If A < B, then we must also return B > A.
1065 	 */
1066 	if (dst2->aio_initial_sequence < dst1->aio_initial_sequence)
1067 		return(1);
1068 
1069 	return(-1);
1070 }
1071 
1072 /*
1073  * Copy from scope.c.
1074  * XXX: we should standardize the functions and link them as standard
1075  * library.
1076  */
1077 static int
1078 gai_addr2scopetype(struct sockaddr *sa)
1079 {
1080 #ifdef INET6
1081 	struct sockaddr_in6 *sa6;
1082 #endif
1083 	struct sockaddr_in *sa4;
1084 
1085 	switch(sa->sa_family) {
1086 #ifdef INET6
1087 	case AF_INET6:
1088 		sa6 = (struct sockaddr_in6 *)sa;
1089 		if (IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
1090 			/* just use the scope field of the multicast address */
1091 			return(sa6->sin6_addr.s6_addr[2] & 0x0f);
1092 		}
1093 		/*
1094 		 * Unicast addresses: map scope type to corresponding scope
1095 		 * value defined for multcast addresses.
1096 		 * XXX: hardcoded scope type values are bad...
1097 		 */
1098 		if (IN6_IS_ADDR_LOOPBACK(&sa6->sin6_addr))
1099 			return(1); /* node local scope */
1100 		if (IN6_IS_ADDR_LINKLOCAL(&sa6->sin6_addr))
1101 			return(2); /* link-local scope */
1102 		if (IN6_IS_ADDR_SITELOCAL(&sa6->sin6_addr))
1103 			return(5); /* site-local scope */
1104 		return(14);	/* global scope */
1105 		break;
1106 #endif
1107 	case AF_INET:
1108 		/*
1109 		 * IPv4 pseudo scoping according to RFC 3484.
1110 		 */
1111 		sa4 = (struct sockaddr_in *)sa;
1112 		/* IPv4 autoconfiguration addresses have link-local scope. */
1113 		if (((u_char *)&sa4->sin_addr)[0] == 169 &&
1114 		    ((u_char *)&sa4->sin_addr)[1] == 254)
1115 			return(2);
1116 		/* Private addresses have site-local scope. */
1117 		if (((u_char *)&sa4->sin_addr)[0] == 10 ||
1118 		    (((u_char *)&sa4->sin_addr)[0] == 172 &&
1119 		     (((u_char *)&sa4->sin_addr)[1] & 0xf0) == 16) ||
1120 		    (((u_char *)&sa4->sin_addr)[0] == 192 &&
1121 		     ((u_char *)&sa4->sin_addr)[1] == 168))
1122 			return(14);	/* XXX: It should be 5 unless NAT */
1123 		/* Loopback addresses have link-local scope. */
1124 		if (((u_char *)&sa4->sin_addr)[0] == 127)
1125 			return(2);
1126 		return(14);
1127 		break;
1128 	default:
1129 		errno = EAFNOSUPPORT; /* is this a good error? */
1130 		return(-1);
1131 	}
1132 }
1133 #endif
1134