xref: /freebsd/sys/netinet6/in6_src.c (revision 995dc984471c92c03daad19a1d35af46c086ef3e)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: in6_src.c,v 1.132 2003/08/26 04:42:27 keiichi Exp $
30  */
31 
32 /*-
33  * Copyright (c) 1982, 1986, 1991, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)in_pcb.c	8.2 (Berkeley) 1/4/94
61  */
62 
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
65 
66 #include "opt_inet.h"
67 #include "opt_inet6.h"
68 
69 #include <sys/param.h>
70 #include <sys/systm.h>
71 #include <sys/lock.h>
72 #include <sys/malloc.h>
73 #include <sys/mbuf.h>
74 #include <sys/priv.h>
75 #include <sys/protosw.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/sockio.h>
79 #include <sys/sysctl.h>
80 #include <sys/errno.h>
81 #include <sys/time.h>
82 #include <sys/kernel.h>
83 #include <sys/sx.h>
84 
85 #include <net/if.h>
86 #include <net/route.h>
87 
88 #include <netinet/in.h>
89 #include <netinet/in_var.h>
90 #include <netinet/in_systm.h>
91 #include <netinet/ip.h>
92 #include <netinet/in_pcb.h>
93 #include <netinet6/in6_var.h>
94 #include <netinet/ip6.h>
95 #include <netinet6/in6_pcb.h>
96 #include <netinet6/ip6_var.h>
97 #include <netinet6/scope6_var.h>
98 #include <netinet6/nd6.h>
99 
100 static struct mtx addrsel_lock;
101 #define	ADDRSEL_LOCK_INIT()	mtx_init(&addrsel_lock, "addrsel_lock", NULL, MTX_DEF)
102 #define	ADDRSEL_LOCK()		mtx_lock(&addrsel_lock)
103 #define	ADDRSEL_UNLOCK()	mtx_unlock(&addrsel_lock)
104 #define	ADDRSEL_LOCK_ASSERT()	mtx_assert(&addrsel_lock, MA_OWNED)
105 
106 static struct sx addrsel_sxlock;
107 #define	ADDRSEL_SXLOCK_INIT()	sx_init(&addrsel_sxlock, "addrsel_sxlock")
108 #define	ADDRSEL_SLOCK()		sx_slock(&addrsel_sxlock)
109 #define	ADDRSEL_SUNLOCK()	sx_sunlock(&addrsel_sxlock)
110 #define	ADDRSEL_XLOCK()		sx_xlock(&addrsel_sxlock)
111 #define	ADDRSEL_XUNLOCK()	sx_xunlock(&addrsel_sxlock)
112 
113 #define ADDR_LABEL_NOTAPP (-1)
114 struct in6_addrpolicy defaultaddrpolicy;
115 
116 int ip6_prefer_tempaddr = 0;
117 
118 static int selectroute __P((struct sockaddr_in6 *, struct ip6_pktopts *,
119 	struct ip6_moptions *, struct route_in6 *, struct ifnet **,
120 	struct rtentry **, int, int));
121 static int in6_selectif __P((struct sockaddr_in6 *, struct ip6_pktopts *,
122 	struct ip6_moptions *, struct route_in6 *ro, struct ifnet **));
123 
124 static struct in6_addrpolicy *lookup_addrsel_policy(struct sockaddr_in6 *);
125 
126 static void init_policy_queue(void);
127 static int add_addrsel_policyent(struct in6_addrpolicy *);
128 static int delete_addrsel_policyent(struct in6_addrpolicy *);
129 static int walk_addrsel_policy __P((int (*)(struct in6_addrpolicy *, void *),
130 				    void *));
131 static int dump_addrsel_policyent(struct in6_addrpolicy *, void *);
132 static struct in6_addrpolicy *match_addrsel_policy(struct sockaddr_in6 *);
133 
134 /*
135  * Return an IPv6 address, which is the most appropriate for a given
136  * destination and user specified options.
137  * If necessary, this function lookups the routing table and returns
138  * an entry to the caller for later use.
139  */
140 #define REPLACE(r) do {\
141 	if ((r) < sizeof(ip6stat.ip6s_sources_rule) / \
142 		sizeof(ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
143 		ip6stat.ip6s_sources_rule[(r)]++; \
144 	/* { \
145 	char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
146 	printf("in6_selectsrc: replace %s with %s by %d\n", ia_best ? ip6_sprintf(ip6buf, &ia_best->ia_addr.sin6_addr) : "none", ip6_sprintf(ip6b, &ia->ia_addr.sin6_addr), (r)); \
147 	} */ \
148 	goto replace; \
149 } while(0)
150 #define NEXT(r) do {\
151 	if ((r) < sizeof(ip6stat.ip6s_sources_rule) / \
152 		sizeof(ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
153 		ip6stat.ip6s_sources_rule[(r)]++; \
154 	/* { \
155 	char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
156 	printf("in6_selectsrc: keep %s against %s by %d\n", ia_best ? ip6_sprintf(ip6buf, &ia_best->ia_addr.sin6_addr) : "none", ip6_sprintf(ip6b, &ia->ia_addr.sin6_addr), (r)); \
157 	} */ \
158 	goto next;		/* XXX: we can't use 'continue' here */ \
159 } while(0)
160 #define BREAK(r) do { \
161 	if ((r) < sizeof(ip6stat.ip6s_sources_rule) / \
162 		sizeof(ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
163 		ip6stat.ip6s_sources_rule[(r)]++; \
164 	goto out;		/* XXX: we can't use 'break' here */ \
165 } while(0)
166 
167 struct in6_addr *
168 in6_selectsrc(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
169     struct ip6_moptions *mopts, struct route_in6 *ro,
170     struct in6_addr *laddr, struct ifnet **ifpp, int *errorp)
171 {
172 	struct in6_addr dst;
173 	struct ifnet *ifp = NULL;
174 	struct in6_ifaddr *ia = NULL, *ia_best = NULL;
175 	struct in6_pktinfo *pi = NULL;
176 	int dst_scope = -1, best_scope = -1, best_matchlen = -1;
177 	struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL;
178 	u_int32_t odstzone;
179 	int prefer_tempaddr;
180 
181 	dst = dstsock->sin6_addr; /* make a copy for local operation */
182 	*errorp = 0;
183 	if (ifpp)
184 		*ifpp = NULL;
185 
186 	/*
187 	 * If the source address is explicitly specified by the caller,
188 	 * check if the requested source address is indeed a unicast address
189 	 * assigned to the node, and can be used as the packet's source
190 	 * address.  If everything is okay, use the address as source.
191 	 */
192 	if (opts && (pi = opts->ip6po_pktinfo) &&
193 	    !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) {
194 		struct sockaddr_in6 srcsock;
195 		struct in6_ifaddr *ia6;
196 
197 		/* get the outgoing interface */
198 		if ((*errorp = in6_selectif(dstsock, opts, mopts, ro, &ifp))
199 		    != 0) {
200 			return (NULL);
201 		}
202 
203 		/*
204 		 * determine the appropriate zone id of the source based on
205 		 * the zone of the destination and the outgoing interface.
206 		 * If the specified address is ambiguous wrt the scope zone,
207 		 * the interface must be specified; otherwise, ifa_ifwithaddr()
208 		 * will fail matching the address.
209 		 */
210 		bzero(&srcsock, sizeof(srcsock));
211 		srcsock.sin6_family = AF_INET6;
212 		srcsock.sin6_len = sizeof(srcsock);
213 		srcsock.sin6_addr = pi->ipi6_addr;
214 		if (ifp) {
215 			*errorp = in6_setscope(&srcsock.sin6_addr, ifp, NULL);
216 			if (*errorp != 0)
217 				return (NULL);
218 		}
219 
220 		ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)(&srcsock));
221 		if (ia6 == NULL ||
222 		    (ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY))) {
223 			*errorp = EADDRNOTAVAIL;
224 			return (NULL);
225 		}
226 		pi->ipi6_addr = srcsock.sin6_addr; /* XXX: this overrides pi */
227 		if (ifpp)
228 			*ifpp = ifp;
229 		return (&ia6->ia_addr.sin6_addr);
230 	}
231 
232 	/*
233 	 * Otherwise, if the socket has already bound the source, just use it.
234 	 */
235 	if (laddr && !IN6_IS_ADDR_UNSPECIFIED(laddr))
236 		return (laddr);
237 
238 	/*
239 	 * If the address is not specified, choose the best one based on
240 	 * the outgoing interface and the destination address.
241 	 */
242 	/* get the outgoing interface */
243 	if ((*errorp = in6_selectif(dstsock, opts, mopts, ro, &ifp)) != 0)
244 		return (NULL);
245 
246 #ifdef DIAGNOSTIC
247 	if (ifp == NULL)	/* this should not happen */
248 		panic("in6_selectsrc: NULL ifp");
249 #endif
250 	*errorp = in6_setscope(&dst, ifp, &odstzone);
251 	if (*errorp != 0)
252 		return (NULL);
253 
254 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
255 		int new_scope = -1, new_matchlen = -1;
256 		struct in6_addrpolicy *new_policy = NULL;
257 		u_int32_t srczone, osrczone, dstzone;
258 		struct in6_addr src;
259 		struct ifnet *ifp1 = ia->ia_ifp;
260 
261 		/*
262 		 * We'll never take an address that breaks the scope zone
263 		 * of the destination.  We also skip an address if its zone
264 		 * does not contain the outgoing interface.
265 		 * XXX: we should probably use sin6_scope_id here.
266 		 */
267 		if (in6_setscope(&dst, ifp1, &dstzone) ||
268 		    odstzone != dstzone) {
269 			continue;
270 		}
271 		src = ia->ia_addr.sin6_addr;
272 		if (in6_setscope(&src, ifp, &osrczone) ||
273 		    in6_setscope(&src, ifp1, &srczone) ||
274 		    osrczone != srczone) {
275 			continue;
276 		}
277 
278 		/* avoid unusable addresses */
279 		if ((ia->ia6_flags &
280 		     (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) {
281 				continue;
282 		}
283 		if (!ip6_use_deprecated && IFA6_IS_DEPRECATED(ia))
284 			continue;
285 
286 		/* Rule 1: Prefer same address */
287 		if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) {
288 			ia_best = ia;
289 			BREAK(1); /* there should be no better candidate */
290 		}
291 
292 		if (ia_best == NULL)
293 			REPLACE(0);
294 
295 		/* Rule 2: Prefer appropriate scope */
296 		if (dst_scope < 0)
297 			dst_scope = in6_addrscope(&dst);
298 		new_scope = in6_addrscope(&ia->ia_addr.sin6_addr);
299 		if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) {
300 			if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0)
301 				REPLACE(2);
302 			NEXT(2);
303 		} else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) {
304 			if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0)
305 				NEXT(2);
306 			REPLACE(2);
307 		}
308 
309 		/*
310 		 * Rule 3: Avoid deprecated addresses.  Note that the case of
311 		 * !ip6_use_deprecated is already rejected above.
312 		 */
313 		if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia))
314 			NEXT(3);
315 		if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia))
316 			REPLACE(3);
317 
318 		/* Rule 4: Prefer home addresses */
319 		/*
320 		 * XXX: This is a TODO.  We should probably merge the MIP6
321 		 * case above.
322 		 */
323 
324 		/* Rule 5: Prefer outgoing interface */
325 		if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp)
326 			NEXT(5);
327 		if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp)
328 			REPLACE(5);
329 
330 		/*
331 		 * Rule 6: Prefer matching label
332 		 * Note that best_policy should be non-NULL here.
333 		 */
334 		if (dst_policy == NULL)
335 			dst_policy = lookup_addrsel_policy(dstsock);
336 		if (dst_policy->label != ADDR_LABEL_NOTAPP) {
337 			new_policy = lookup_addrsel_policy(&ia->ia_addr);
338 			if (dst_policy->label == best_policy->label &&
339 			    dst_policy->label != new_policy->label)
340 				NEXT(6);
341 			if (dst_policy->label != best_policy->label &&
342 			    dst_policy->label == new_policy->label)
343 				REPLACE(6);
344 		}
345 
346 		/*
347 		 * Rule 7: Prefer public addresses.
348 		 * We allow users to reverse the logic by configuring
349 		 * a sysctl variable, so that privacy conscious users can
350 		 * always prefer temporary addresses.
351 		 */
352 		if (opts == NULL ||
353 		    opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) {
354 			prefer_tempaddr = ip6_prefer_tempaddr;
355 		} else if (opts->ip6po_prefer_tempaddr ==
356 		    IP6PO_TEMPADDR_NOTPREFER) {
357 			prefer_tempaddr = 0;
358 		} else
359 			prefer_tempaddr = 1;
360 		if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
361 		    (ia->ia6_flags & IN6_IFF_TEMPORARY)) {
362 			if (prefer_tempaddr)
363 				REPLACE(7);
364 			else
365 				NEXT(7);
366 		}
367 		if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
368 		    !(ia->ia6_flags & IN6_IFF_TEMPORARY)) {
369 			if (prefer_tempaddr)
370 				NEXT(7);
371 			else
372 				REPLACE(7);
373 		}
374 
375 		/*
376 		 * Rule 8: prefer addresses on alive interfaces.
377 		 * This is a KAME specific rule.
378 		 */
379 		if ((ia_best->ia_ifp->if_flags & IFF_UP) &&
380 		    !(ia->ia_ifp->if_flags & IFF_UP))
381 			NEXT(8);
382 		if (!(ia_best->ia_ifp->if_flags & IFF_UP) &&
383 		    (ia->ia_ifp->if_flags & IFF_UP))
384 			REPLACE(8);
385 
386 		/*
387 		 * Rule 14: Use longest matching prefix.
388 		 * Note: in the address selection draft, this rule is
389 		 * documented as "Rule 8".  However, since it is also
390 		 * documented that this rule can be overridden, we assign
391 		 * a large number so that it is easy to assign smaller numbers
392 		 * to more preferred rules.
393 		 */
394 		new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst);
395 		if (best_matchlen < new_matchlen)
396 			REPLACE(14);
397 		if (new_matchlen < best_matchlen)
398 			NEXT(14);
399 
400 		/* Rule 15 is reserved. */
401 
402 		/*
403 		 * Last resort: just keep the current candidate.
404 		 * Or, do we need more rules?
405 		 */
406 		continue;
407 
408 	  replace:
409 		ia_best = ia;
410 		best_scope = (new_scope >= 0 ? new_scope :
411 			      in6_addrscope(&ia_best->ia_addr.sin6_addr));
412 		best_policy = (new_policy ? new_policy :
413 			       lookup_addrsel_policy(&ia_best->ia_addr));
414 		best_matchlen = (new_matchlen >= 0 ? new_matchlen :
415 				 in6_matchlen(&ia_best->ia_addr.sin6_addr,
416 					      &dst));
417 
418 	  next:
419 		continue;
420 
421 	  out:
422 		break;
423 	}
424 
425 	if ((ia = ia_best) == NULL) {
426 		*errorp = EADDRNOTAVAIL;
427 		return (NULL);
428 	}
429 
430 	if (ifpp)
431 		*ifpp = ifp;
432 
433 	return (&ia->ia_addr.sin6_addr);
434 }
435 
436 /*
437  * clone - meaningful only for bsdi and freebsd
438  */
439 static int
440 selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
441     struct ip6_moptions *mopts, struct route_in6 *ro,
442     struct ifnet **retifp, struct rtentry **retrt, int clone,
443     int norouteok)
444 {
445 	int error = 0;
446 	struct ifnet *ifp = NULL;
447 	struct rtentry *rt = NULL;
448 	struct sockaddr_in6 *sin6_next;
449 	struct in6_pktinfo *pi = NULL;
450 	struct in6_addr *dst = &dstsock->sin6_addr;
451 #if 0
452 	char ip6buf[INET6_ADDRSTRLEN];
453 
454 	if (dstsock->sin6_addr.s6_addr32[0] == 0 &&
455 	    dstsock->sin6_addr.s6_addr32[1] == 0 &&
456 	    !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) {
457 		printf("in6_selectroute: strange destination %s\n",
458 		       ip6_sprintf(ip6buf, &dstsock->sin6_addr));
459 	} else {
460 		printf("in6_selectroute: destination = %s%%%d\n",
461 		       ip6_sprintf(ip6buf, &dstsock->sin6_addr),
462 		       dstsock->sin6_scope_id); /* for debug */
463 	}
464 #endif
465 
466 	/* If the caller specify the outgoing interface explicitly, use it. */
467 	if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) {
468 		/* XXX boundary check is assumed to be already done. */
469 		ifp = ifnet_byindex(pi->ipi6_ifindex);
470 		if (ifp != NULL &&
471 		    (norouteok || retrt == NULL ||
472 		    IN6_IS_ADDR_MULTICAST(dst))) {
473 			/*
474 			 * we do not have to check or get the route for
475 			 * multicast.
476 			 */
477 			goto done;
478 		} else
479 			goto getroute;
480 	}
481 
482 	/*
483 	 * If the destination address is a multicast address and the outgoing
484 	 * interface for the address is specified by the caller, use it.
485 	 */
486 	if (IN6_IS_ADDR_MULTICAST(dst) &&
487 	    mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) {
488 		goto done; /* we do not need a route for multicast. */
489 	}
490 
491   getroute:
492 	/*
493 	 * If the next hop address for the packet is specified by the caller,
494 	 * use it as the gateway.
495 	 */
496 	if (opts && opts->ip6po_nexthop) {
497 		struct route_in6 *ron;
498 
499 		sin6_next = satosin6(opts->ip6po_nexthop);
500 
501 		/* at this moment, we only support AF_INET6 next hops */
502 		if (sin6_next->sin6_family != AF_INET6) {
503 			error = EAFNOSUPPORT; /* or should we proceed? */
504 			goto done;
505 		}
506 
507 		/*
508 		 * If the next hop is an IPv6 address, then the node identified
509 		 * by that address must be a neighbor of the sending host.
510 		 */
511 		ron = &opts->ip6po_nextroute;
512 		if ((ron->ro_rt &&
513 		     (ron->ro_rt->rt_flags & (RTF_UP | RTF_LLINFO)) !=
514 		     (RTF_UP | RTF_LLINFO)) ||
515 		    !IN6_ARE_ADDR_EQUAL(&satosin6(&ron->ro_dst)->sin6_addr,
516 		    &sin6_next->sin6_addr)) {
517 			if (ron->ro_rt) {
518 				RTFREE(ron->ro_rt);
519 				ron->ro_rt = NULL;
520 			}
521 			*satosin6(&ron->ro_dst) = *sin6_next;
522 		}
523 		if (ron->ro_rt == NULL) {
524 			rtalloc((struct route *)ron); /* multi path case? */
525 			if (ron->ro_rt == NULL ||
526 			    !(ron->ro_rt->rt_flags & RTF_LLINFO)) {
527 				if (ron->ro_rt) {
528 					RTFREE(ron->ro_rt);
529 					ron->ro_rt = NULL;
530 				}
531 				error = EHOSTUNREACH;
532 				goto done;
533 			}
534 		}
535 		rt = ron->ro_rt;
536 		ifp = rt->rt_ifp;
537 
538 		/*
539 		 * When cloning is required, try to allocate a route to the
540 		 * destination so that the caller can store path MTU
541 		 * information.
542 		 */
543 		if (!clone)
544 			goto done;
545 	}
546 
547 	/*
548 	 * Use a cached route if it exists and is valid, else try to allocate
549 	 * a new one.  Note that we should check the address family of the
550 	 * cached destination, in case of sharing the cache with IPv4.
551 	 */
552 	if (ro) {
553 		if (ro->ro_rt &&
554 		    (!(ro->ro_rt->rt_flags & RTF_UP) ||
555 		     ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 ||
556 		     !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr,
557 		     dst))) {
558 			RTFREE(ro->ro_rt);
559 			ro->ro_rt = (struct rtentry *)NULL;
560 		}
561 		if (ro->ro_rt == (struct rtentry *)NULL) {
562 			struct sockaddr_in6 *sa6;
563 
564 			/* No route yet, so try to acquire one */
565 			bzero(&ro->ro_dst, sizeof(struct sockaddr_in6));
566 			sa6 = (struct sockaddr_in6 *)&ro->ro_dst;
567 			*sa6 = *dstsock;
568 			sa6->sin6_scope_id = 0;
569 
570 			if (clone) {
571 				rtalloc((struct route *)ro);
572 			} else {
573 				ro->ro_rt = rtalloc1(&((struct route *)ro)
574 						     ->ro_dst, 0, 0UL);
575 				if (ro->ro_rt)
576 					RT_UNLOCK(ro->ro_rt);
577 			}
578 		}
579 
580 		/*
581 		 * do not care about the result if we have the nexthop
582 		 * explicitly specified.
583 		 */
584 		if (opts && opts->ip6po_nexthop)
585 			goto done;
586 
587 		if (ro->ro_rt) {
588 			ifp = ro->ro_rt->rt_ifp;
589 
590 			if (ifp == NULL) { /* can this really happen? */
591 				RTFREE(ro->ro_rt);
592 				ro->ro_rt = NULL;
593 			}
594 		}
595 		if (ro->ro_rt == NULL)
596 			error = EHOSTUNREACH;
597 		rt = ro->ro_rt;
598 
599 		/*
600 		 * Check if the outgoing interface conflicts with
601 		 * the interface specified by ipi6_ifindex (if specified).
602 		 * Note that loopback interface is always okay.
603 		 * (this may happen when we are sending a packet to one of
604 		 *  our own addresses.)
605 		 */
606 		if (ifp && opts && opts->ip6po_pktinfo &&
607 		    opts->ip6po_pktinfo->ipi6_ifindex) {
608 			if (!(ifp->if_flags & IFF_LOOPBACK) &&
609 			    ifp->if_index !=
610 			    opts->ip6po_pktinfo->ipi6_ifindex) {
611 				error = EHOSTUNREACH;
612 				goto done;
613 			}
614 		}
615 	}
616 
617   done:
618 	if (ifp == NULL && rt == NULL) {
619 		/*
620 		 * This can happen if the caller did not pass a cached route
621 		 * nor any other hints.  We treat this case an error.
622 		 */
623 		error = EHOSTUNREACH;
624 	}
625 	if (error == EHOSTUNREACH)
626 		ip6stat.ip6s_noroute++;
627 
628 	if (retifp != NULL)
629 		*retifp = ifp;
630 	if (retrt != NULL)
631 		*retrt = rt;	/* rt may be NULL */
632 
633 	return (error);
634 }
635 
636 static int
637 in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
638     struct ip6_moptions *mopts, struct route_in6 *ro, struct ifnet **retifp)
639 {
640 	int error;
641 	struct route_in6 sro;
642 	struct rtentry *rt = NULL;
643 
644 	if (ro == NULL) {
645 		bzero(&sro, sizeof(sro));
646 		ro = &sro;
647 	}
648 
649 	if ((error = selectroute(dstsock, opts, mopts, ro, retifp,
650 				     &rt, 0, 1)) != 0) {
651 		if (ro == &sro && rt && rt == sro.ro_rt)
652 			RTFREE(rt);
653 		return (error);
654 	}
655 
656 	/*
657 	 * do not use a rejected or black hole route.
658 	 * XXX: this check should be done in the L2 output routine.
659 	 * However, if we skipped this check here, we'd see the following
660 	 * scenario:
661 	 * - install a rejected route for a scoped address prefix
662 	 *   (like fe80::/10)
663 	 * - send a packet to a destination that matches the scoped prefix,
664 	 *   with ambiguity about the scope zone.
665 	 * - pick the outgoing interface from the route, and disambiguate the
666 	 *   scope zone with the interface.
667 	 * - ip6_output() would try to get another route with the "new"
668 	 *   destination, which may be valid.
669 	 * - we'd see no error on output.
670 	 * Although this may not be very harmful, it should still be confusing.
671 	 * We thus reject the case here.
672 	 */
673 	if (rt && (rt->rt_flags & (RTF_REJECT | RTF_BLACKHOLE))) {
674 		int flags = (rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
675 
676 		if (ro == &sro && rt && rt == sro.ro_rt)
677 			RTFREE(rt);
678 		return (flags);
679 	}
680 
681 	/*
682 	 * Adjust the "outgoing" interface.  If we're going to loop the packet
683 	 * back to ourselves, the ifp would be the loopback interface.
684 	 * However, we'd rather know the interface associated to the
685 	 * destination address (which should probably be one of our own
686 	 * addresses.)
687 	 */
688 	if (rt && rt->rt_ifa && rt->rt_ifa->ifa_ifp)
689 		*retifp = rt->rt_ifa->ifa_ifp;
690 
691 	if (ro == &sro && rt && rt == sro.ro_rt)
692 		RTFREE(rt);
693 	return (0);
694 }
695 
696 /*
697  * clone - meaningful only for bsdi and freebsd
698  */
699 int
700 in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
701     struct ip6_moptions *mopts, struct route_in6 *ro,
702     struct ifnet **retifp, struct rtentry **retrt, int clone)
703 {
704 
705 	return (selectroute(dstsock, opts, mopts, ro, retifp,
706 	    retrt, clone, 0));
707 }
708 
709 /*
710  * Default hop limit selection. The precedence is as follows:
711  * 1. Hoplimit value specified via ioctl.
712  * 2. (If the outgoing interface is detected) the current
713  *     hop limit of the interface specified by router advertisement.
714  * 3. The system default hoplimit.
715  */
716 int
717 in6_selecthlim(struct in6pcb *in6p, struct ifnet *ifp)
718 {
719 
720 	if (in6p && in6p->in6p_hops >= 0)
721 		return (in6p->in6p_hops);
722 	else if (ifp)
723 		return (ND_IFINFO(ifp)->chlim);
724 	else if (in6p && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
725 		struct route_in6 ro6;
726 		struct ifnet *lifp;
727 
728 		bzero(&ro6, sizeof(ro6));
729 		ro6.ro_dst.sin6_family = AF_INET6;
730 		ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
731 		ro6.ro_dst.sin6_addr = in6p->in6p_faddr;
732 		rtalloc((struct route *)&ro6);
733 		if (ro6.ro_rt) {
734 			lifp = ro6.ro_rt->rt_ifp;
735 			RTFREE(ro6.ro_rt);
736 			if (lifp)
737 				return (ND_IFINFO(lifp)->chlim);
738 		} else
739 			return (ip6_defhlim);
740 	}
741 	return (ip6_defhlim);
742 }
743 
744 /*
745  * XXX: this is borrowed from in6_pcbbind(). If possible, we should
746  * share this function by all *bsd*...
747  */
748 int
749 in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct ucred *cred)
750 {
751 	struct socket *so = inp->inp_socket;
752 	u_int16_t lport = 0, first, last, *lastport;
753 	int count, error = 0, wild = 0;
754 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
755 
756 	INP_INFO_WLOCK_ASSERT(pcbinfo);
757 	INP_LOCK_ASSERT(inp);
758 
759 	/* XXX: this is redundant when called from in6_pcbbind */
760 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
761 		wild = INPLOOKUP_WILDCARD;
762 
763 	inp->inp_flags |= INP_ANONPORT;
764 
765 	if (inp->inp_flags & INP_HIGHPORT) {
766 		first = ipport_hifirstauto;	/* sysctl */
767 		last  = ipport_hilastauto;
768 		lastport = &pcbinfo->ipi_lasthi;
769 	} else if (inp->inp_flags & INP_LOWPORT) {
770 		error = priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0);
771 		if (error)
772 			return error;
773 		first = ipport_lowfirstauto;	/* 1023 */
774 		last  = ipport_lowlastauto;	/* 600 */
775 		lastport = &pcbinfo->ipi_lastlow;
776 	} else {
777 		first = ipport_firstauto;	/* sysctl */
778 		last  = ipport_lastauto;
779 		lastport = &pcbinfo->ipi_lastport;
780 	}
781 	/*
782 	 * Simple check to ensure all ports are not used up causing
783 	 * a deadlock here.
784 	 *
785 	 * We split the two cases (up and down) so that the direction
786 	 * is not being tested on each round of the loop.
787 	 */
788 	if (first > last) {
789 		/*
790 		 * counting down
791 		 */
792 		count = first - last;
793 
794 		do {
795 			if (count-- < 0) {	/* completely used? */
796 				/*
797 				 * Undo any address bind that may have
798 				 * occurred above.
799 				 */
800 				inp->in6p_laddr = in6addr_any;
801 				return (EAGAIN);
802 			}
803 			--*lastport;
804 			if (*lastport > first || *lastport < last)
805 				*lastport = first;
806 			lport = htons(*lastport);
807 		} while (in6_pcblookup_local(pcbinfo, &inp->in6p_laddr,
808 					     lport, wild));
809 	} else {
810 		/*
811 			 * counting up
812 			 */
813 		count = last - first;
814 
815 		do {
816 			if (count-- < 0) {	/* completely used? */
817 				/*
818 				 * Undo any address bind that may have
819 				 * occurred above.
820 				 */
821 				inp->in6p_laddr = in6addr_any;
822 				return (EAGAIN);
823 			}
824 			++*lastport;
825 			if (*lastport < first || *lastport > last)
826 				*lastport = first;
827 			lport = htons(*lastport);
828 		} while (in6_pcblookup_local(pcbinfo,
829 					     &inp->in6p_laddr, lport, wild));
830 	}
831 
832 	inp->inp_lport = lport;
833 	if (in_pcbinshash(inp) != 0) {
834 		inp->in6p_laddr = in6addr_any;
835 		inp->inp_lport = 0;
836 		return (EAGAIN);
837 	}
838 
839 	return (0);
840 }
841 
842 void
843 addrsel_policy_init(void)
844 {
845 	ADDRSEL_LOCK_INIT();
846 	ADDRSEL_SXLOCK_INIT();
847 
848 	init_policy_queue();
849 
850 	/* initialize the "last resort" policy */
851 	bzero(&defaultaddrpolicy, sizeof(defaultaddrpolicy));
852 	defaultaddrpolicy.label = ADDR_LABEL_NOTAPP;
853 }
854 
855 static struct in6_addrpolicy *
856 lookup_addrsel_policy(struct sockaddr_in6 *key)
857 {
858 	struct in6_addrpolicy *match = NULL;
859 
860 	ADDRSEL_LOCK();
861 	match = match_addrsel_policy(key);
862 
863 	if (match == NULL)
864 		match = &defaultaddrpolicy;
865 	else
866 		match->use++;
867 	ADDRSEL_UNLOCK();
868 
869 	return (match);
870 }
871 
872 /*
873  * Subroutines to manage the address selection policy table via sysctl.
874  */
875 struct walkarg {
876 	struct sysctl_req *w_req;
877 };
878 
879 static int in6_src_sysctl(SYSCTL_HANDLER_ARGS);
880 SYSCTL_DECL(_net_inet6_ip6);
881 SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy,
882 	CTLFLAG_RD, in6_src_sysctl, "");
883 
884 static int
885 in6_src_sysctl(SYSCTL_HANDLER_ARGS)
886 {
887 	struct walkarg w;
888 
889 	if (req->newptr)
890 		return EPERM;
891 
892 	bzero(&w, sizeof(w));
893 	w.w_req = req;
894 
895 	return (walk_addrsel_policy(dump_addrsel_policyent, &w));
896 }
897 
898 int
899 in6_src_ioctl(u_long cmd, caddr_t data)
900 {
901 	int i;
902 	struct in6_addrpolicy ent0;
903 
904 	if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY)
905 		return (EOPNOTSUPP); /* check for safety */
906 
907 	ent0 = *(struct in6_addrpolicy *)data;
908 
909 	if (ent0.label == ADDR_LABEL_NOTAPP)
910 		return (EINVAL);
911 	/* check if the prefix mask is consecutive. */
912 	if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0)
913 		return (EINVAL);
914 	/* clear trailing garbages (if any) of the prefix address. */
915 	for (i = 0; i < 4; i++) {
916 		ent0.addr.sin6_addr.s6_addr32[i] &=
917 			ent0.addrmask.sin6_addr.s6_addr32[i];
918 	}
919 	ent0.use = 0;
920 
921 	switch (cmd) {
922 	case SIOCAADDRCTL_POLICY:
923 		return (add_addrsel_policyent(&ent0));
924 	case SIOCDADDRCTL_POLICY:
925 		return (delete_addrsel_policyent(&ent0));
926 	}
927 
928 	return (0);		/* XXX: compromise compilers */
929 }
930 
931 /*
932  * The followings are implementation of the policy table using a
933  * simple tail queue.
934  * XXX such details should be hidden.
935  * XXX implementation using binary tree should be more efficient.
936  */
937 struct addrsel_policyent {
938 	TAILQ_ENTRY(addrsel_policyent) ape_entry;
939 	struct in6_addrpolicy ape_policy;
940 };
941 
942 TAILQ_HEAD(addrsel_policyhead, addrsel_policyent);
943 
944 struct addrsel_policyhead addrsel_policytab;
945 
946 static void
947 init_policy_queue(void)
948 {
949 
950 	TAILQ_INIT(&addrsel_policytab);
951 }
952 
953 static int
954 add_addrsel_policyent(struct in6_addrpolicy *newpolicy)
955 {
956 	struct addrsel_policyent *new, *pol;
957 
958 	MALLOC(new, struct addrsel_policyent *, sizeof(*new), M_IFADDR,
959 	       M_WAITOK);
960 	ADDRSEL_XLOCK();
961 	ADDRSEL_LOCK();
962 
963 	/* duplication check */
964 	TAILQ_FOREACH(pol, &addrsel_policytab, ape_entry) {
965 		if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr,
966 				       &pol->ape_policy.addr.sin6_addr) &&
967 		    IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr,
968 				       &pol->ape_policy.addrmask.sin6_addr)) {
969 			ADDRSEL_UNLOCK();
970 			ADDRSEL_XUNLOCK();
971 			FREE(new, M_IFADDR);
972 			return (EEXIST);	/* or override it? */
973 		}
974 	}
975 
976 	bzero(new, sizeof(*new));
977 
978 	/* XXX: should validate entry */
979 	new->ape_policy = *newpolicy;
980 
981 	TAILQ_INSERT_TAIL(&addrsel_policytab, new, ape_entry);
982 	ADDRSEL_UNLOCK();
983 	ADDRSEL_XUNLOCK();
984 
985 	return (0);
986 }
987 
988 static int
989 delete_addrsel_policyent(struct in6_addrpolicy *key)
990 {
991 	struct addrsel_policyent *pol;
992 
993 	ADDRSEL_XLOCK();
994 	ADDRSEL_LOCK();
995 
996 	/* search for the entry in the table */
997 	TAILQ_FOREACH(pol, &addrsel_policytab, ape_entry) {
998 		if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr,
999 		    &pol->ape_policy.addr.sin6_addr) &&
1000 		    IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr,
1001 		    &pol->ape_policy.addrmask.sin6_addr)) {
1002 			break;
1003 		}
1004 	}
1005 	if (pol == NULL) {
1006 		ADDRSEL_UNLOCK();
1007 		ADDRSEL_XUNLOCK();
1008 		return (ESRCH);
1009 	}
1010 
1011 	TAILQ_REMOVE(&addrsel_policytab, pol, ape_entry);
1012 	ADDRSEL_UNLOCK();
1013 	ADDRSEL_XUNLOCK();
1014 
1015 	return (0);
1016 }
1017 
1018 static int
1019 walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *),
1020     void *w)
1021 {
1022 	struct addrsel_policyent *pol;
1023 	int error = 0;
1024 
1025 	ADDRSEL_SLOCK();
1026 	TAILQ_FOREACH(pol, &addrsel_policytab, ape_entry) {
1027 		if ((error = (*callback)(&pol->ape_policy, w)) != 0) {
1028 			ADDRSEL_SUNLOCK();
1029 			return (error);
1030 		}
1031 	}
1032 	ADDRSEL_SUNLOCK();
1033 	return (error);
1034 }
1035 
1036 static int
1037 dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg)
1038 {
1039 	int error = 0;
1040 	struct walkarg *w = arg;
1041 
1042 	error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol));
1043 
1044 	return (error);
1045 }
1046 
1047 static struct in6_addrpolicy *
1048 match_addrsel_policy(struct sockaddr_in6 *key)
1049 {
1050 	struct addrsel_policyent *pent;
1051 	struct in6_addrpolicy *bestpol = NULL, *pol;
1052 	int matchlen, bestmatchlen = -1;
1053 	u_char *mp, *ep, *k, *p, m;
1054 
1055 	TAILQ_FOREACH(pent, &addrsel_policytab, ape_entry) {
1056 		matchlen = 0;
1057 
1058 		pol = &pent->ape_policy;
1059 		mp = (u_char *)&pol->addrmask.sin6_addr;
1060 		ep = mp + 16;	/* XXX: scope field? */
1061 		k = (u_char *)&key->sin6_addr;
1062 		p = (u_char *)&pol->addr.sin6_addr;
1063 		for (; mp < ep && *mp; mp++, k++, p++) {
1064 			m = *mp;
1065 			if ((*k & m) != *p)
1066 				goto next; /* not match */
1067 			if (m == 0xff) /* short cut for a typical case */
1068 				matchlen += 8;
1069 			else {
1070 				while (m >= 0x80) {
1071 					matchlen++;
1072 					m <<= 1;
1073 				}
1074 			}
1075 		}
1076 
1077 		/* matched.  check if this is better than the current best. */
1078 		if (bestpol == NULL ||
1079 		    matchlen > bestmatchlen) {
1080 			bestpol = pol;
1081 			bestmatchlen = matchlen;
1082 		}
1083 
1084 	  next:
1085 		continue;
1086 	}
1087 
1088 	return (bestpol);
1089 }
1090