xref: /freebsd/lib/libc/net/name6.c (revision 80ff58b89dcacfe07fe20b045890df9db5ca0af0)
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 
86 /*
87  *	Atsushi Onoe <onoe@sm.sony.co.jp>
88  */
89 
90 #include <sys/cdefs.h>
91 __FBSDID("$FreeBSD$");
92 
93 #include "namespace.h"
94 #include <sys/param.h>
95 #include <sys/socket.h>
96 #include <sys/time.h>
97 #include <sys/queue.h>
98 #include <netinet/in.h>
99 #ifdef INET6
100 #include <net/if.h>
101 #include <net/if_var.h>
102 #include <sys/sysctl.h>
103 #include <sys/ioctl.h>
104 #include <netinet6/in6_var.h>	/* XXX */
105 #endif
106 
107 #include <arpa/inet.h>
108 #include <arpa/nameser.h>
109 
110 #include <errno.h>
111 #include <netdb.h>
112 #include <resolv.h>
113 #include <stdio.h>
114 #include <stdlib.h>
115 #include <string.h>
116 #include <stdarg.h>
117 #include <nsswitch.h>
118 #include <unistd.h>
119 #include "un-namespace.h"
120 #include "netdb_private.h"
121 #include "res_private.h"
122 
123 #ifndef MAXALIASES
124 #define	MAXALIASES	10
125 #endif
126 #ifndef	MAXADDRS
127 #define	MAXADDRS	20
128 #endif
129 #ifndef MAXDNAME
130 #define	MAXDNAME	1025
131 #endif
132 
133 #ifdef INET6
134 #define	ADDRLEN(af)	((af) == AF_INET6 ? sizeof(struct in6_addr) : \
135 					    sizeof(struct in_addr))
136 #else
137 #define	ADDRLEN(af)	sizeof(struct in_addr)
138 #endif
139 
140 #define	MAPADDR(ab, ina) \
141 do {									\
142 	memcpy(&(ab)->map_inaddr, ina, sizeof(struct in_addr));		\
143 	memset((ab)->map_zero, 0, sizeof((ab)->map_zero));		\
144 	memset((ab)->map_one, 0xff, sizeof((ab)->map_one));		\
145 } while (0)
146 #define	MAPADDRENABLED(flags) \
147 	(((flags) & AI_V4MAPPED) || \
148 	 (((flags) & AI_V4MAPPED_CFG)))
149 
150 union inx_addr {
151 	struct in_addr	in_addr;
152 #ifdef INET6
153 	struct in6_addr	in6_addr;
154 #endif
155 	struct {
156 		u_char	mau_zero[10];
157 		u_char	mau_one[2];
158 		struct in_addr mau_inaddr;
159 	}		map_addr_un;
160 #define	map_zero	map_addr_un.mau_zero
161 #define	map_one		map_addr_un.mau_one
162 #define	map_inaddr	map_addr_un.mau_inaddr
163 };
164 
165 struct policyqueue {
166 	TAILQ_ENTRY(policyqueue) pc_entry;
167 #ifdef INET6
168 	struct in6_addrpolicy pc_policy;
169 #endif
170 };
171 TAILQ_HEAD(policyhead, policyqueue);
172 
173 #define AIO_SRCFLAG_DEPRECATED	0x1
174 
175 struct hp_order {
176 	union {
177 		struct sockaddr_storage aiou_ss;
178 		struct sockaddr aiou_sa;
179 	} aio_src_un;
180 #define aio_srcsa aio_src_un.aiou_sa
181 	u_int32_t aio_srcflag;
182 	int aio_srcscope;
183 	int aio_dstscope;
184 	struct policyqueue *aio_srcpolicy;
185 	struct policyqueue *aio_dstpolicy;
186 	union {
187 		struct sockaddr_storage aiou_ss;
188 		struct sockaddr aiou_sa;
189 	} aio_un;
190 #define aio_sa aio_un.aiou_sa
191 	int aio_matchlen;
192 	char *aio_h_addr;
193 };
194 
195 static struct	 hostent *_hpcopy(struct hostent *, int *);
196 static struct	 hostent *_hpaddr(int, const char *, void *, int *);
197 #ifdef INET6
198 static struct	 hostent *_hpmerge(struct hostent *, struct hostent *, int *);
199 static struct	 hostent *_hpmapv6(struct hostent *, int *);
200 #endif
201 static struct	 hostent *_hpsort(struct hostent *, res_state);
202 
203 #ifdef INET6
204 static struct	 hostent *_hpreorder(struct hostent *);
205 static int	 get_addrselectpolicy(struct policyhead *);
206 static void	 free_addrselectpolicy(struct policyhead *);
207 static struct	 policyqueue *match_addrselectpolicy(struct sockaddr *,
208 	struct policyhead *);
209 static void	 set_source(struct hp_order *, struct policyhead *);
210 static int	 matchlen(struct sockaddr *, struct sockaddr *);
211 static int	 comp_dst(const void *, const void *);
212 static int	 gai_addr2scopetype(struct sockaddr *);
213 #endif
214 
215 /*
216  * Functions defined in RFC2553
217  *	getipnodebyname, getipnodebyaddr, freehostent
218  */
219 
220 struct hostent *
221 getipnodebyname(const char *name, int af, int flags, int *errp)
222 {
223 	struct hostent *hp;
224 	union inx_addr addrbuf;
225 	res_state statp;
226 	u_long options;
227 
228 	switch (af) {
229 	case AF_INET:
230 #ifdef INET6
231 	case AF_INET6:
232 #endif
233 		break;
234 	default:
235 		*errp = NO_RECOVERY;
236 		return NULL;
237 	}
238 
239 	if (flags & AI_ADDRCONFIG) {
240 		int s;
241 
242 		if ((s = _socket(af, SOCK_DGRAM, 0)) < 0)
243 			return NULL;
244 		/*
245 		 * TODO:
246 		 * Note that implementation dependent test for address
247 		 * configuration should be done everytime called
248 		 * (or apropriate interval),
249 		 * because addresses will be dynamically assigned or deleted.
250 		 */
251 		_close(s);
252 	}
253 
254 #ifdef INET6
255 	/* special case for literal address */
256 	if (inet_pton(AF_INET6, name, &addrbuf) == 1) {
257 		if (af != AF_INET6) {
258 			*errp = HOST_NOT_FOUND;
259 			return NULL;
260 		}
261 		return _hpaddr(af, name, &addrbuf, errp);
262 	}
263 #endif
264 	if (inet_aton(name, (struct in_addr *)&addrbuf) == 1) {
265 		if (af != AF_INET) {
266 			if (MAPADDRENABLED(flags)) {
267 				MAPADDR(&addrbuf, &addrbuf.in_addr);
268 			} else {
269 				*errp = HOST_NOT_FOUND;
270 				return NULL;
271 			}
272 		}
273 		return _hpaddr(af, name, &addrbuf, errp);
274 	}
275 
276 
277 	statp = __res_state();
278 	if ((statp->options & RES_INIT) == 0) {
279 		if (res_ninit(statp) < 0) {
280 			*errp = NETDB_INTERNAL;
281 			return NULL;
282 		}
283 	}
284 
285 	options = statp->options;
286 	statp->options &= ~RES_USE_INET6;
287 
288 	hp = gethostbyname2(name, af);
289 	hp = _hpcopy(hp, errp);
290 #ifdef INET6
291 	if (af == AF_INET6)
292 		hp = _hpreorder(hp);
293 
294 	if (af == AF_INET6 && ((flags & AI_ALL) || hp == NULL) &&
295 	    MAPADDRENABLED(flags)) {
296 		struct hostent *hp2 = gethostbyname2(name, AF_INET);
297 		if (hp == NULL)
298 			if (hp2 == NULL)
299 				*errp = statp->res_h_errno;
300 			else
301 				hp = _hpmapv6(hp2, errp);
302 		else {
303 			if (hp2 && strcmp(hp->h_name, hp2->h_name) == 0) {
304 				struct hostent *hpb = hp;
305 				hp = _hpmerge(hpb, hp2, errp);
306 				freehostent(hpb);
307 			}
308 		}
309 	}
310 #endif
311 
312 	if (hp == NULL)
313 		*errp = statp->res_h_errno;
314 
315 	statp->options = options;
316 	return _hpsort(hp, statp);
317 }
318 
319 struct hostent *
320 getipnodebyaddr(const void *src, size_t len, int af, int *errp)
321 {
322 	struct hostent *hp;
323 	res_state statp;
324 	u_long options;
325 
326 #ifdef INET6
327 	struct in6_addr addrbuf;
328 #else
329 	struct in_addr addrbuf;
330 #endif
331 
332 	switch (af) {
333 	case AF_INET:
334 		if (len != sizeof(struct in_addr)) {
335 			*errp = NO_RECOVERY;
336 			return NULL;
337 		}
338 		if ((long)src & ~(sizeof(struct in_addr) - 1)) {
339 			memcpy(&addrbuf, src, len);
340 			src = &addrbuf;
341 		}
342 		if (((struct in_addr *)src)->s_addr == 0)
343 			return NULL;
344 		break;
345 #ifdef INET6
346 	case AF_INET6:
347 		if (len != sizeof(struct in6_addr)) {
348 			*errp = NO_RECOVERY;
349 			return NULL;
350 		}
351 		if ((long)src & ~(sizeof(struct in6_addr) / 2 - 1)) {	/*XXX*/
352 			memcpy(&addrbuf, src, len);
353 			src = &addrbuf;
354 		}
355 		if (IN6_IS_ADDR_UNSPECIFIED((struct in6_addr *)src))
356 			return NULL;
357 		if (IN6_IS_ADDR_V4MAPPED((struct in6_addr *)src)
358 		||  IN6_IS_ADDR_V4COMPAT((struct in6_addr *)src)) {
359 			src = (char *)src +
360 			    (sizeof(struct in6_addr) - sizeof(struct in_addr));
361 			af = AF_INET;
362 			len = sizeof(struct in_addr);
363 		}
364 		break;
365 #endif
366 	default:
367 		*errp = NO_RECOVERY;
368 		return NULL;
369 	}
370 
371 	statp = __res_state();
372 	if ((statp->options & RES_INIT) == 0) {
373 		if (res_ninit(statp) < 0) {
374 			RES_SET_H_ERRNO(statp, NETDB_INTERNAL);
375 			return NULL;
376 		}
377 	}
378 
379 	options = statp->options;
380 	statp->options &= ~RES_USE_INET6;
381 
382 	hp = gethostbyaddr(src, len, af);
383 	if (hp == NULL)
384 		*errp = statp->res_h_errno;
385 
386 	statp->options = options;
387 	return (_hpcopy(hp, errp));
388 }
389 
390 void
391 freehostent(struct hostent *ptr)
392 {
393 	free(ptr);
394 }
395 
396 /*
397  * Private utility functions
398  */
399 
400 /*
401  * _hpcopy: allocate and copy hostent structure
402  */
403 static struct hostent *
404 _hpcopy(struct hostent *hp, int *errp)
405 {
406 	struct hostent *nhp;
407 	char *cp, **pp;
408 	int size, addrsize;
409 	int nalias = 0, naddr = 0;
410 	int al_off;
411 	int i;
412 
413 	if (hp == NULL)
414 		return hp;
415 
416 	/* count size to be allocated */
417 	size = sizeof(struct hostent);
418 	if (hp->h_name != NULL)
419 		size += strlen(hp->h_name) + 1;
420 	if ((pp = hp->h_aliases) != NULL) {
421 		for (i = 0; *pp != NULL; i++, pp++) {
422 			if (**pp != '\0') {
423 				size += strlen(*pp) + 1;
424 				nalias++;
425 			}
426 		}
427 	}
428 	/* adjust alignment */
429 	size = ALIGN(size);
430 	al_off = size;
431 	size += sizeof(char *) * (nalias + 1);
432 	addrsize = ALIGN(hp->h_length);
433 	if ((pp = hp->h_addr_list) != NULL) {
434 		while (*pp++ != NULL)
435 			naddr++;
436 	}
437 	size += addrsize * naddr;
438 	size += sizeof(char *) * (naddr + 1);
439 
440 	/* copy */
441 	if ((nhp = (struct hostent *)malloc(size)) == NULL) {
442 		*errp = TRY_AGAIN;
443 		return NULL;
444 	}
445 	cp = (char *)&nhp[1];
446 	if (hp->h_name != NULL) {
447 		nhp->h_name = cp;
448 		strcpy(cp, hp->h_name);
449 		cp += strlen(cp) + 1;
450 	} else
451 		nhp->h_name = NULL;
452 	nhp->h_aliases = (char **)((char *)nhp + al_off);
453 	if ((pp = hp->h_aliases) != NULL) {
454 		for (i = 0; *pp != NULL; pp++) {
455 			if (**pp != '\0') {
456 				nhp->h_aliases[i++] = cp;
457 				strcpy(cp, *pp);
458 				cp += strlen(cp) + 1;
459 			}
460 		}
461 	}
462 	nhp->h_aliases[nalias] = NULL;
463 	cp = (char *)&nhp->h_aliases[nalias + 1];
464 	nhp->h_addrtype = hp->h_addrtype;
465 	nhp->h_length = hp->h_length;
466 	nhp->h_addr_list = (char **)cp;
467 	if ((pp = hp->h_addr_list) != NULL) {
468 		cp = (char *)&nhp->h_addr_list[naddr + 1];
469 		for (i = 0; *pp != NULL; pp++) {
470 			nhp->h_addr_list[i++] = cp;
471 			memcpy(cp, *pp, hp->h_length);
472 			cp += addrsize;
473 		}
474 	}
475 	nhp->h_addr_list[naddr] = NULL;
476 	return nhp;
477 }
478 
479 /*
480  * _hpaddr: construct hostent structure with one address
481  */
482 static struct hostent *
483 _hpaddr(int af, const char *name, void *addr, int *errp)
484 {
485 	struct hostent *hp, hpbuf;
486 	char *addrs[2];
487 
488 	hp = &hpbuf;
489 	hp->h_name = (char *)name;
490 	hp->h_aliases = NULL;
491 	hp->h_addrtype = af;
492 	hp->h_length = ADDRLEN(af);
493 	hp->h_addr_list = addrs;
494 	addrs[0] = (char *)addr;
495 	addrs[1] = NULL;
496 	return (_hpcopy(hp, errp));
497 }
498 
499 #ifdef INET6
500 /*
501  * _hpmerge: merge 2 hostent structure, arguments will be freed
502  */
503 static struct hostent *
504 _hpmerge(struct hostent *hp1, struct hostent *hp2, int *errp)
505 {
506 	int i, j;
507 	int naddr, nalias;
508 	char **pp;
509 	struct hostent *hp, hpbuf;
510 	char *aliases[MAXALIASES + 1], *addrs[MAXADDRS + 1];
511 	union inx_addr addrbuf[MAXADDRS];
512 
513 	if (hp1 == NULL)
514 		return _hpcopy(hp2, errp);
515 	if (hp2 == NULL)
516 		return _hpcopy(hp1, errp);
517 
518 #define	HP(i)	(i == 1 ? hp1 : hp2)
519 	hp = &hpbuf;
520 	hp->h_name = (hp1->h_name != NULL ? hp1->h_name : hp2->h_name);
521 	hp->h_aliases = aliases;
522 	nalias = 0;
523 	for (i = 1; i <= 2; i++) {
524 		if ((pp = HP(i)->h_aliases) == NULL)
525 			continue;
526 		for (; nalias < MAXALIASES && *pp != NULL; pp++) {
527 			/* check duplicates */
528 			for (j = 0; j < nalias; j++)
529 				if (strcasecmp(*pp, aliases[j]) == 0)
530 					break;
531 			if (j == nalias)
532 				aliases[nalias++] = *pp;
533 		}
534 	}
535 	aliases[nalias] = NULL;
536 	if (hp1->h_length != hp2->h_length) {
537 		hp->h_addrtype = AF_INET6;
538 		hp->h_length = sizeof(struct in6_addr);
539 	} else {
540 		hp->h_addrtype = hp1->h_addrtype;
541 		hp->h_length = hp1->h_length;
542 	}
543 
544 	hp->h_addr_list = addrs;
545 	naddr = 0;
546 	for (i = 1; i <= 2; i++) {
547 		if ((pp = HP(i)->h_addr_list) == NULL)
548 			continue;
549 		if (HP(i)->h_length == hp->h_length) {
550 			while (naddr < MAXADDRS && *pp != NULL)
551 				addrs[naddr++] = *pp++;
552 		} else {
553 			/* copy IPv4 addr as mapped IPv6 addr */
554 			while (naddr < MAXADDRS && *pp != NULL) {
555 				MAPADDR(&addrbuf[naddr], *pp++);
556 				addrs[naddr] = (char *)&addrbuf[naddr];
557 				naddr++;
558 			}
559 		}
560 	}
561 	addrs[naddr] = NULL;
562 	return (_hpcopy(hp, errp));
563 }
564 #endif
565 
566 /*
567  * _hpmapv6: convert IPv4 hostent into IPv4-mapped IPv6 addresses
568  */
569 #ifdef INET6
570 static struct hostent *
571 _hpmapv6(struct hostent *hp, int *errp)
572 {
573 	struct hostent hp6;
574 
575 	if (hp == NULL)
576 		return NULL;
577 	if (hp->h_addrtype == AF_INET6)
578 		return _hpcopy(hp, errp);
579 
580 	memset(&hp6, 0, sizeof(struct hostent));
581 	hp6.h_addrtype = AF_INET6;
582 	hp6.h_length = sizeof(struct in6_addr);
583 	return _hpmerge(&hp6, hp, errp);
584 }
585 #endif
586 
587 /*
588  * _hpsort: sort address by sortlist
589  */
590 static struct hostent *
591 _hpsort(struct hostent *hp, res_state statp)
592 {
593 	int i, j, n;
594 	u_char *ap, *sp, *mp, **pp;
595 	char t;
596 	char order[MAXADDRS];
597 	int nsort = statp->nsort;
598 
599 	if (hp == NULL || hp->h_addr_list[1] == NULL || nsort == 0)
600 		return hp;
601 	for (i = 0; (ap = (u_char *)hp->h_addr_list[i]); i++) {
602 		for (j = 0; j < nsort; j++) {
603 #ifdef INET6
604 			if (statp->_u._ext.ext->sort_list[j].af !=
605 			    hp->h_addrtype)
606 				continue;
607 			sp = (u_char *)&statp->_u._ext.ext->sort_list[j].addr;
608 			mp = (u_char *)&statp->_u._ext.ext->sort_list[j].mask;
609 #else
610 			sp = (u_char *)&statp->sort_list[j].addr;
611 			mp = (u_char *)&statp->sort_list[j].mask;
612 #endif
613 			for (n = 0; n < hp->h_length; n++) {
614 				if ((ap[n] & mp[n]) != sp[n])
615 					break;
616 			}
617 			if (n == hp->h_length)
618 				break;
619 		}
620 		order[i] = j;
621 	}
622 	n = i;
623 	pp = (u_char **)hp->h_addr_list;
624 	for (i = 0; i < n - 1; i++) {
625 		for (j = i + 1; j < n; j++) {
626 			if (order[i] > order[j]) {
627 				ap = pp[i];
628 				pp[i] = pp[j];
629 				pp[j] = ap;
630 				t = order[i];
631 				order[i] = order[j];
632 				order[j] = t;
633 			}
634 		}
635 	}
636 	return hp;
637 }
638 
639 #ifdef INET6
640 /*
641  * _hpreorder: sort address by default address selection
642  */
643 static struct hostent *
644 _hpreorder(struct hostent *hp)
645 {
646 	struct hp_order *aio;
647 	int i, n;
648 	char *ap;
649 	struct sockaddr *sa;
650 	struct policyhead policyhead;
651 
652 	if (hp == NULL)
653 		return hp;
654 
655 	switch (hp->h_addrtype) {
656 	case AF_INET:
657 #ifdef INET6
658 	case AF_INET6:
659 #endif
660 		break;
661 	default:
662 		free_addrselectpolicy(&policyhead);
663 		return hp;
664 	}
665 
666 	/* count the number of addrinfo elements for sorting. */
667 	for (n = 0; hp->h_addr_list[n] != NULL; n++)
668 		;
669 
670 	/*
671 	 * If the number is small enough, we can skip the reordering process.
672 	 */
673 	if (n <= 1)
674 		return hp;
675 
676 	/* allocate a temporary array for sort and initialization of it. */
677 	if ((aio = malloc(sizeof(*aio) * n)) == NULL)
678 		return hp;	/* give up reordering */
679 	memset(aio, 0, sizeof(*aio) * n);
680 
681 	/* retrieve address selection policy from the kernel */
682 	TAILQ_INIT(&policyhead);
683 	if (!get_addrselectpolicy(&policyhead)) {
684 		/* no policy is installed into kernel, we don't sort. */
685 		free(aio);
686 		return hp;
687 	}
688 
689 	for (i = 0; i < n; i++) {
690 		ap = hp->h_addr_list[i];
691 		aio[i].aio_h_addr = ap;
692 		sa = &aio[i].aio_sa;
693 		switch (hp->h_addrtype) {
694 		case AF_INET:
695 			sa->sa_family = AF_INET;
696 			sa->sa_len = sizeof(struct sockaddr_in);
697 			memcpy(&((struct sockaddr_in *)sa)->sin_addr, ap,
698 			    sizeof(struct in_addr));
699 			break;
700 #ifdef INET6
701 		case AF_INET6:
702 			if (IN6_IS_ADDR_V4MAPPED((struct in6_addr *)ap)) {
703 				sa->sa_family = AF_INET;
704 				sa->sa_len = sizeof(struct sockaddr_in);
705 				memcpy(&((struct sockaddr_in *)sa)->sin_addr,
706 				    &ap[12], sizeof(struct in_addr));
707 			} else {
708 				sa->sa_family = AF_INET6;
709 				sa->sa_len = sizeof(struct sockaddr_in6);
710 				memcpy(&((struct sockaddr_in6 *)sa)->sin6_addr,
711 				    ap, sizeof(struct in6_addr));
712 			}
713 			break;
714 #endif
715 		}
716 		aio[i].aio_dstscope = gai_addr2scopetype(sa);
717 		aio[i].aio_dstpolicy = match_addrselectpolicy(sa, &policyhead);
718 		set_source(&aio[i], &policyhead);
719 	}
720 
721 	/* perform sorting. */
722 	qsort(aio, n, sizeof(*aio), comp_dst);
723 
724 	/* reorder the h_addr_list. */
725 	for (i = 0; i < n; i++)
726 		hp->h_addr_list[i] = aio[i].aio_h_addr;
727 
728 	/* cleanup and return */
729 	free(aio);
730 	free_addrselectpolicy(&policyhead);
731 	return hp;
732 }
733 
734 static int
735 get_addrselectpolicy(struct policyhead *head)
736 {
737 #ifdef INET6
738 	int mib[] = { CTL_NET, PF_INET6, IPPROTO_IPV6, IPV6CTL_ADDRCTLPOLICY };
739 	size_t l;
740 	char *buf;
741 	struct in6_addrpolicy *pol, *ep;
742 
743 	if (sysctl(mib, sizeof(mib) / sizeof(mib[0]), NULL, &l, NULL, 0) < 0)
744 		return (0);
745 	if ((buf = malloc(l)) == NULL)
746 		return (0);
747 	if (sysctl(mib, sizeof(mib) / sizeof(mib[0]), buf, &l, NULL, 0) < 0) {
748 		free(buf);
749 		return (0);
750 	}
751 
752 	ep = (struct in6_addrpolicy *)(buf + l);
753 	for (pol = (struct in6_addrpolicy *)buf; pol + 1 <= ep; pol++) {
754 		struct policyqueue *new;
755 
756 		if ((new = malloc(sizeof(*new))) == NULL) {
757 			free_addrselectpolicy(head); /* make the list empty */
758 			break;
759 		}
760 		new->pc_policy = *pol;
761 		TAILQ_INSERT_TAIL(head, new, pc_entry);
762 	}
763 
764 	free(buf);
765 	return (1);
766 #else
767 	return (0);
768 #endif
769 }
770 
771 static void
772 free_addrselectpolicy(struct policyhead *head)
773 {
774 	struct policyqueue *ent, *nent;
775 
776 	for (ent = TAILQ_FIRST(head); ent; ent = nent) {
777 		nent = TAILQ_NEXT(ent, pc_entry);
778 		TAILQ_REMOVE(head, ent, pc_entry);
779 		free(ent);
780 	}
781 }
782 
783 static struct policyqueue *
784 match_addrselectpolicy(struct sockaddr *addr, struct policyhead *head)
785 {
786 #ifdef INET6
787 	struct policyqueue *ent, *bestent = NULL;
788 	struct in6_addrpolicy *pol;
789 	int matchlen, bestmatchlen = -1;
790 	u_char *mp, *ep, *k, *p, m;
791 	struct sockaddr_in6 key;
792 
793 	switch(addr->sa_family) {
794 	case AF_INET6:
795 		key = *(struct sockaddr_in6 *)addr;
796 		break;
797 	case AF_INET:
798 		/* convert the address into IPv4-mapped IPv6 address. */
799 		memset(&key, 0, sizeof(key));
800 		key.sin6_family = AF_INET6;
801 		key.sin6_len = sizeof(key);
802 		key.sin6_addr.s6_addr[10] = 0xff;
803 		key.sin6_addr.s6_addr[11] = 0xff;
804 		memcpy(&key.sin6_addr.s6_addr[12],
805 		       &((struct sockaddr_in *)addr)->sin_addr, 4);
806 		break;
807 	default:
808 		return(NULL);
809 	}
810 
811 	for (ent = TAILQ_FIRST(head); ent; ent = TAILQ_NEXT(ent, pc_entry)) {
812 		pol = &ent->pc_policy;
813 		matchlen = 0;
814 
815 		mp = (u_char *)&pol->addrmask.sin6_addr;
816 		ep = mp + 16;	/* XXX: scope field? */
817 		k = (u_char *)&key.sin6_addr;
818 		p = (u_char *)&pol->addr.sin6_addr;
819 		for (; mp < ep && *mp; mp++, k++, p++) {
820 			m = *mp;
821 			if ((*k & m) != *p)
822 				goto next; /* not match */
823 			if (m == 0xff) /* short cut for a typical case */
824 				matchlen += 8;
825 			else {
826 				while (m >= 0x80) {
827 					matchlen++;
828 					m <<= 1;
829 				}
830 			}
831 		}
832 
833 		/* matched.  check if this is better than the current best. */
834 		if (matchlen > bestmatchlen) {
835 			bestent = ent;
836 			bestmatchlen = matchlen;
837 		}
838 
839 	  next:
840 		continue;
841 	}
842 
843 	return(bestent);
844 #else
845 	return(NULL);
846 #endif
847 
848 }
849 
850 static void
851 set_source(struct hp_order *aio, struct policyhead *ph)
852 {
853 	struct sockaddr_storage ss = aio->aio_un.aiou_ss;
854 	socklen_t srclen;
855 	int s;
856 
857 	/* set unspec ("no source is available"), just in case */
858 	aio->aio_srcsa.sa_family = AF_UNSPEC;
859 	aio->aio_srcscope = -1;
860 
861 	switch(ss.ss_family) {
862 	case AF_INET:
863 		((struct sockaddr_in *)&ss)->sin_port = htons(1);
864 		break;
865 #ifdef INET6
866 	case AF_INET6:
867 		((struct sockaddr_in6 *)&ss)->sin6_port = htons(1);
868 		break;
869 #endif
870 	default:		/* ignore unsupported AFs explicitly */
871 		return;
872 	}
873 
874 	/* open a socket to get the source address for the given dst */
875 	if ((s = _socket(ss.ss_family, SOCK_DGRAM, IPPROTO_UDP)) < 0)
876 		return;		/* give up */
877 	if (_connect(s, (struct sockaddr *)&ss, ss.ss_len) < 0)
878 		goto cleanup;
879 	srclen = ss.ss_len;
880 	if (_getsockname(s, &aio->aio_srcsa, &srclen) < 0) {
881 		aio->aio_srcsa.sa_family = AF_UNSPEC;
882 		goto cleanup;
883 	}
884 	aio->aio_srcscope = gai_addr2scopetype(&aio->aio_srcsa);
885 	aio->aio_srcpolicy = match_addrselectpolicy(&aio->aio_srcsa, ph);
886 	aio->aio_matchlen = matchlen(&aio->aio_srcsa, (struct sockaddr *)&ss);
887 #ifdef INET6
888 	if (ss.ss_family == AF_INET6) {
889 		struct in6_ifreq ifr6;
890 		u_int32_t flags6;
891 
892 		memset(&ifr6, 0, sizeof(ifr6));
893 		memcpy(&ifr6.ifr_addr, &ss, ss.ss_len);
894 		if (_ioctl(s, SIOCGIFAFLAG_IN6, &ifr6) == 0) {
895 			flags6 = ifr6.ifr_ifru.ifru_flags6;
896 			if ((flags6 & IN6_IFF_DEPRECATED))
897 				aio->aio_srcflag |= AIO_SRCFLAG_DEPRECATED;
898 		}
899 	}
900 #endif
901 
902   cleanup:
903 	_close(s);
904 	return;
905 }
906 
907 static int
908 matchlen(struct sockaddr *src, struct sockaddr *dst)
909 {
910 	int match = 0;
911 	u_char *s, *d;
912 	u_char *lim, r;
913 	int addrlen;
914 
915 	switch (src->sa_family) {
916 #ifdef INET6
917 	case AF_INET6:
918 		s = (u_char *)&((struct sockaddr_in6 *)src)->sin6_addr;
919 		d = (u_char *)&((struct sockaddr_in6 *)dst)->sin6_addr;
920 		addrlen = sizeof(struct in6_addr);
921 		lim = s + addrlen;
922 		break;
923 #endif
924 	case AF_INET:
925 		s = (u_char *)&((struct sockaddr_in *)src)->sin_addr;
926 		d = (u_char *)&((struct sockaddr_in *)dst)->sin_addr;
927 		addrlen = sizeof(struct in_addr);
928 		lim = s + addrlen;
929 		break;
930 	default:
931 		return(0);
932 	}
933 
934 	while (s < lim)
935 		if ((r = (*d++ ^ *s++)) != 0) {
936 			while (r < addrlen * 8) {
937 				match++;
938 				r <<= 1;
939 			}
940 			break;
941 		} else
942 			match += 8;
943 	return(match);
944 }
945 
946 static int
947 comp_dst(const void *arg1, const void *arg2)
948 {
949 	const struct hp_order *dst1 = arg1, *dst2 = arg2;
950 
951 	/*
952 	 * Rule 1: Avoid unusable destinations.
953 	 * XXX: we currently do not consider if an appropriate route exists.
954 	 */
955 	if (dst1->aio_srcsa.sa_family != AF_UNSPEC &&
956 	    dst2->aio_srcsa.sa_family == AF_UNSPEC) {
957 		return(-1);
958 	}
959 	if (dst1->aio_srcsa.sa_family == AF_UNSPEC &&
960 	    dst2->aio_srcsa.sa_family != AF_UNSPEC) {
961 		return(1);
962 	}
963 
964 	/* Rule 2: Prefer matching scope. */
965 	if (dst1->aio_dstscope == dst1->aio_srcscope &&
966 	    dst2->aio_dstscope != dst2->aio_srcscope) {
967 		return(-1);
968 	}
969 	if (dst1->aio_dstscope != dst1->aio_srcscope &&
970 	    dst2->aio_dstscope == dst2->aio_srcscope) {
971 		return(1);
972 	}
973 
974 	/* Rule 3: Avoid deprecated addresses. */
975 	if (dst1->aio_srcsa.sa_family != AF_UNSPEC &&
976 	    dst2->aio_srcsa.sa_family != AF_UNSPEC) {
977 		if (!(dst1->aio_srcflag & AIO_SRCFLAG_DEPRECATED) &&
978 		    (dst2->aio_srcflag & AIO_SRCFLAG_DEPRECATED)) {
979 			return(-1);
980 		}
981 		if ((dst1->aio_srcflag & AIO_SRCFLAG_DEPRECATED) &&
982 		    !(dst2->aio_srcflag & AIO_SRCFLAG_DEPRECATED)) {
983 			return(1);
984 		}
985 	}
986 
987 	/* Rule 4: Prefer home addresses. */
988 	/* XXX: not implemented yet */
989 
990 	/* Rule 5: Prefer matching label. */
991 #ifdef INET6
992 	if (dst1->aio_srcpolicy && dst1->aio_dstpolicy &&
993 	    dst1->aio_srcpolicy->pc_policy.label ==
994 	    dst1->aio_dstpolicy->pc_policy.label &&
995 	    (dst2->aio_srcpolicy == NULL || dst2->aio_dstpolicy == NULL ||
996 	     dst2->aio_srcpolicy->pc_policy.label !=
997 	     dst2->aio_dstpolicy->pc_policy.label)) {
998 		return(-1);
999 	}
1000 	if (dst2->aio_srcpolicy && dst2->aio_dstpolicy &&
1001 	    dst2->aio_srcpolicy->pc_policy.label ==
1002 	    dst2->aio_dstpolicy->pc_policy.label &&
1003 	    (dst1->aio_srcpolicy == NULL || dst1->aio_dstpolicy == NULL ||
1004 	     dst1->aio_srcpolicy->pc_policy.label !=
1005 	     dst1->aio_dstpolicy->pc_policy.label)) {
1006 		return(1);
1007 	}
1008 #endif
1009 
1010 	/* Rule 6: Prefer higher precedence. */
1011 #ifdef INET6
1012 	if (dst1->aio_dstpolicy &&
1013 	    (dst2->aio_dstpolicy == NULL ||
1014 	     dst1->aio_dstpolicy->pc_policy.preced >
1015 	     dst2->aio_dstpolicy->pc_policy.preced)) {
1016 		return(-1);
1017 	}
1018 	if (dst2->aio_dstpolicy &&
1019 	    (dst1->aio_dstpolicy == NULL ||
1020 	     dst2->aio_dstpolicy->pc_policy.preced >
1021 	     dst1->aio_dstpolicy->pc_policy.preced)) {
1022 		return(1);
1023 	}
1024 #endif
1025 
1026 	/* Rule 7: Prefer native transport. */
1027 	/* XXX: not implemented yet */
1028 
1029 	/* Rule 8: Prefer smaller scope. */
1030 	if (dst1->aio_dstscope >= 0 &&
1031 	    dst1->aio_dstscope < dst2->aio_dstscope) {
1032 		return(-1);
1033 	}
1034 	if (dst2->aio_dstscope >= 0 &&
1035 	    dst2->aio_dstscope < dst1->aio_dstscope) {
1036 		return(1);
1037 	}
1038 
1039 	/*
1040 	 * Rule 9: Use longest matching prefix.
1041 	 * We compare the match length in a same AF only.
1042 	 */
1043 	if (dst1->aio_sa.sa_family == dst2->aio_sa.sa_family) {
1044 		if (dst1->aio_matchlen > dst2->aio_matchlen) {
1045 			return(-1);
1046 		}
1047 		if (dst1->aio_matchlen < dst2->aio_matchlen) {
1048 			return(1);
1049 		}
1050 	}
1051 
1052 	/* Rule 10: Otherwise, leave the order unchanged. */
1053 	return(-1);
1054 }
1055 
1056 /*
1057  * Copy from scope.c.
1058  * XXX: we should standardize the functions and link them as standard
1059  * library.
1060  */
1061 static int
1062 gai_addr2scopetype(struct sockaddr *sa)
1063 {
1064 #ifdef INET6
1065 	struct sockaddr_in6 *sa6;
1066 #endif
1067 	struct sockaddr_in *sa4;
1068 
1069 	switch(sa->sa_family) {
1070 #ifdef INET6
1071 	case AF_INET6:
1072 		sa6 = (struct sockaddr_in6 *)sa;
1073 		if (IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
1074 			/* just use the scope field of the multicast address */
1075 			return(sa6->sin6_addr.s6_addr[2] & 0x0f);
1076 		}
1077 		/*
1078 		 * Unicast addresses: map scope type to corresponding scope
1079 		 * value defined for multcast addresses.
1080 		 * XXX: hardcoded scope type values are bad...
1081 		 */
1082 		if (IN6_IS_ADDR_LOOPBACK(&sa6->sin6_addr))
1083 			return(1); /* node local scope */
1084 		if (IN6_IS_ADDR_LINKLOCAL(&sa6->sin6_addr))
1085 			return(2); /* link-local scope */
1086 		if (IN6_IS_ADDR_SITELOCAL(&sa6->sin6_addr))
1087 			return(5); /* site-local scope */
1088 		return(14);	/* global scope */
1089 		break;
1090 #endif
1091 	case AF_INET:
1092 		/*
1093 		 * IPv4 pseudo scoping according to RFC 3484.
1094 		 */
1095 		sa4 = (struct sockaddr_in *)sa;
1096 		/* IPv4 autoconfiguration addresses have link-local scope. */
1097 		if (((u_char *)&sa4->sin_addr)[0] == 169 &&
1098 		    ((u_char *)&sa4->sin_addr)[1] == 254)
1099 			return(2);
1100 		/* Private addresses have site-local scope. */
1101 		if (((u_char *)&sa4->sin_addr)[0] == 10 ||
1102 		    (((u_char *)&sa4->sin_addr)[0] == 172 &&
1103 		     (((u_char *)&sa4->sin_addr)[1] & 0xf0) == 16) ||
1104 		    (((u_char *)&sa4->sin_addr)[0] == 192 &&
1105 		     ((u_char *)&sa4->sin_addr)[1] == 168))
1106 			return(14);	/* XXX: It should be 5 unless NAT */
1107 		/* Loopback addresses have link-local scope. */
1108 		if (((u_char *)&sa4->sin_addr)[0] == 127)
1109 			return(2);
1110 		return(14);
1111 		break;
1112 	default:
1113 		errno = EAFNOSUPPORT; /* is this a good error? */
1114 		return(-1);
1115 	}
1116 }
1117 #endif
1118