xref: /netbsd-src/sys/netinet6/in6_src.c (revision 946379e7b37692fc43f68eb0d1c10daa0a7f3b6c)
1 /*	$NetBSD: in6_src.c,v 1.59 2015/12/12 23:34:25 christos Exp $	*/
2 /*	$KAME: in6_src.c,v 1.159 2005/10/19 01:40:32 t-momose Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1991, 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  *	@(#)in_pcb.c	8.2 (Berkeley) 1/4/94
66  */
67 
68 #include <sys/cdefs.h>
69 __KERNEL_RCSID(0, "$NetBSD: in6_src.c,v 1.59 2015/12/12 23:34:25 christos Exp $");
70 
71 #ifdef _KERNEL_OPT
72 #include "opt_inet.h"
73 #endif
74 
75 #include <sys/param.h>
76 #include <sys/systm.h>
77 #include <sys/malloc.h>
78 #include <sys/mbuf.h>
79 #include <sys/protosw.h>
80 #include <sys/socket.h>
81 #include <sys/socketvar.h>
82 #include <sys/ioctl.h>
83 #include <sys/errno.h>
84 #include <sys/time.h>
85 #include <sys/kernel.h>
86 #include <sys/proc.h>
87 #include <sys/kauth.h>
88 
89 #include <net/if.h>
90 #include <net/if_types.h>
91 #include <net/route.h>
92 
93 #include <netinet/in.h>
94 #include <netinet/in_var.h>
95 #include <netinet/in_systm.h>
96 #include <netinet/ip.h>
97 #include <netinet/in_pcb.h>
98 #include <netinet/portalgo.h>
99 #include <netinet6/in6_var.h>
100 #include <netinet/ip6.h>
101 #include <netinet6/in6_pcb.h>
102 #include <netinet6/ip6_var.h>
103 #include <netinet6/ip6_private.h>
104 #include <netinet6/nd6.h>
105 #include <netinet6/scope6_var.h>
106 
107 #include <net/net_osdep.h>
108 
109 #ifdef MIP6
110 #include <netinet6/mip6.h>
111 #include <netinet6/mip6_var.h>
112 #include "mip.h"
113 #if NMIP > 0
114 #include <net/if_mip.h>
115 #endif /* NMIP > 0 */
116 #endif /* MIP6 */
117 
118 #include <netinet/tcp_vtw.h>
119 
120 #define ADDR_LABEL_NOTAPP (-1)
121 struct in6_addrpolicy defaultaddrpolicy;
122 
123 int ip6_prefer_tempaddr = 0;
124 
125 static int selectroute(struct sockaddr_in6 *, struct ip6_pktopts *,
126 	struct ip6_moptions *, struct route *, struct ifnet **,
127 	struct rtentry **, int, int);
128 static int in6_selectif(struct sockaddr_in6 *, struct ip6_pktopts *,
129 	struct ip6_moptions *, struct route *, struct ifnet **);
130 
131 static struct in6_addrpolicy *lookup_addrsel_policy(struct sockaddr_in6 *);
132 
133 static void init_policy_queue(void);
134 static int add_addrsel_policyent(struct in6_addrpolicy *);
135 static int delete_addrsel_policyent(struct in6_addrpolicy *);
136 static int walk_addrsel_policy(int (*)(struct in6_addrpolicy *, void *),
137 				    void *);
138 static int dump_addrsel_policyent(struct in6_addrpolicy *, void *);
139 static struct in6_addrpolicy *match_addrsel_policy(struct sockaddr_in6 *);
140 
141 /*
142  * Return an IPv6 address, which is the most appropriate for a given
143  * destination and user specified options.
144  * If necessary, this function lookups the routing table and returns
145  * an entry to the caller for later use.
146  */
147 #if 0				/* diabled ad-hoc */
148 #define REPLACE(r) do {\
149 	if ((r) < sizeof(ip6stat.ip6s_sources_rule) / \
150 		sizeof(ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
151 		ip6stat.ip6s_sources_rule[(r)]++; \
152 	/* printf("in6_selectsrc: replace %s with %s by %d\n", ia_best ? ip6_sprintf(&ia_best->ia_addr.sin6_addr) : "none", ip6_sprintf(&ia->ia_addr.sin6_addr), (r)); */ \
153 	goto replace; \
154 } while(0)
155 #define NEXT(r) do {\
156 	if ((r) < sizeof(ip6stat.ip6s_sources_rule) / \
157 		sizeof(ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
158 		ip6stat.ip6s_sources_rule[(r)]++; \
159 	/* printf("in6_selectsrc: keep %s against %s by %d\n", ia_best ? ip6_sprintf(&ia_best->ia_addr.sin6_addr) : "none", ip6_sprintf(&ia->ia_addr.sin6_addr), (r)); */ \
160 	goto next; 		/* XXX: we can't use 'continue' here */ \
161 } while(0)
162 #define BREAK(r) do { \
163 	if ((r) < sizeof(ip6stat.ip6s_sources_rule) / \
164 		sizeof(ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
165 		ip6stat.ip6s_sources_rule[(r)]++; \
166 	goto out; 		/* XXX: we can't use 'break' here */ \
167 } while(0)
168 #else
169 #define REPLACE(r) goto replace
170 #define NEXT(r) goto next
171 #define BREAK(r) goto out
172 #endif
173 
174 struct in6_addr *
175 in6_selectsrc(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
176 	struct ip6_moptions *mopts, struct route *ro, struct in6_addr *laddr,
177 	struct ifnet **ifpp, int *errorp)
178 {
179 	struct in6_addr dst;
180 	struct ifnet *ifp = NULL;
181 	struct in6_ifaddr *ia = NULL, *ia_best = NULL;
182 	struct in6_pktinfo *pi = NULL;
183 	int dst_scope = -1, best_scope = -1, best_matchlen = -1;
184 	struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL;
185 	u_int32_t odstzone;
186 	int error;
187 	int prefer_tempaddr;
188 #if defined(MIP6) && NMIP > 0
189 	u_int8_t ip6po_usecoa = 0;
190 #endif /* MIP6 && NMIP > 0 */
191 
192 	dst = dstsock->sin6_addr; /* make a copy for local operation */
193 	*errorp = 0;
194 	if (ifpp)
195 		*ifpp = NULL;
196 
197 	/*
198 	 * Try to determine the outgoing interface for the given destination.
199 	 * We do this regardless of whether the socket is bound, since the
200 	 * caller may need this information as a side effect of the call
201 	 * to this function (e.g., for identifying the appropriate scope zone
202 	 * ID).
203 	 */
204 	error = in6_selectif(dstsock, opts, mopts, ro, &ifp);
205 	if (ifpp)
206 		*ifpp = ifp;
207 
208 	/*
209 	 * If the source address is explicitly specified by the caller,
210 	 * check if the requested source address is indeed a unicast address
211 	 * assigned to the node, and can be used as the packet's source
212 	 * address.  If everything is okay, use the address as source.
213 	 */
214 	if (opts && (pi = opts->ip6po_pktinfo) &&
215 	    !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) {
216 		struct sockaddr_in6 srcsock;
217 		struct in6_ifaddr *ia6;
218 
219 		/*
220 		 * Determine the appropriate zone id of the source based on
221 		 * the zone of the destination and the outgoing interface.
222 		 * If the specified address is ambiguous wrt the scope zone,
223 		 * the interface must be specified; otherwise, ifa_ifwithaddr()
224 		 * will fail matching the address.
225 		 */
226 		memset(&srcsock, 0, sizeof(srcsock));
227 		srcsock.sin6_family = AF_INET6;
228 		srcsock.sin6_len = sizeof(srcsock);
229 		srcsock.sin6_addr = pi->ipi6_addr;
230 		if (ifp) {
231 			*errorp = in6_setscope(&srcsock.sin6_addr, ifp, NULL);
232 			if (*errorp != 0)
233 				return (NULL);
234 		}
235 
236 		ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)(&srcsock));
237 		if (ia6 == NULL ||
238 		    (ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY))) {
239 			*errorp = EADDRNOTAVAIL;
240 			return (NULL);
241 		}
242 		pi->ipi6_addr = srcsock.sin6_addr; /* XXX: this overrides pi */
243 		if (ifpp)
244 			*ifpp = ifp;
245 		return (&ia6->ia_addr.sin6_addr);
246 	}
247 
248 	/*
249 	 * If the socket has already bound the source, just use it.  We don't
250 	 * care at the moment whether in6_selectif() succeeded above, even
251 	 * though it would eventually cause an error.
252 	 */
253 	if (laddr && !IN6_IS_ADDR_UNSPECIFIED(laddr))
254 		return (laddr);
255 
256 	/*
257 	 * The outgoing interface is crucial in the general selection procedure
258 	 * below.  If it is not known at this point, we fail.
259 	 */
260 	if (ifp == NULL) {
261 		*errorp = error;
262 		return (NULL);
263 	}
264 
265 	/*
266 	 * If the address is not yet determined, choose the best one based on
267 	 * the outgoing interface and the destination address.
268 	 */
269 
270 #if defined(MIP6) && NMIP > 0
271 	/*
272 	 * a caller can specify IP6PO_USECOA to not to use a home
273 	 * address.  for example, the case that the neighbour
274 	 * unreachability detection to the global address.
275 	 */
276 	if (opts != NULL &&
277 	    (opts->ip6po_flags & IP6PO_USECOA) != 0) {
278 		ip6po_usecoa = 1;
279 	}
280 #endif /* MIP6 && NMIP > 0 */
281 
282 #ifdef DIAGNOSTIC
283 	if (ifp == NULL)	/* this should not happen */
284 		panic("in6_selectsrc: NULL ifp");
285 #endif
286 	*errorp = in6_setscope(&dst, ifp, &odstzone);
287 	if (*errorp != 0)
288 		return (NULL);
289 
290 	for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
291 		int new_scope = -1, new_matchlen = -1;
292 		struct in6_addrpolicy *new_policy = NULL;
293 		u_int32_t srczone, osrczone, dstzone;
294 		struct in6_addr src;
295 		struct ifnet *ifp1 = ia->ia_ifp;
296 
297 		/*
298 		 * We'll never take an address that breaks the scope zone
299 		 * of the destination.  We also skip an address if its zone
300 		 * does not contain the outgoing interface.
301 		 * XXX: we should probably use sin6_scope_id here.
302 		 */
303 		if (in6_setscope(&dst, ifp1, &dstzone) ||
304 		    odstzone != dstzone) {
305 			continue;
306 		}
307 		src = ia->ia_addr.sin6_addr;
308 		if (in6_setscope(&src, ifp, &osrczone) ||
309 		    in6_setscope(&src, ifp1, &srczone) ||
310 		    osrczone != srczone) {
311 			continue;
312 		}
313 
314 		/* avoid unusable addresses */
315 		if ((ia->ia6_flags &
316 		     (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) {
317 				continue;
318 		}
319 		if (!ip6_use_deprecated && IFA6_IS_DEPRECATED(ia))
320 			continue;
321 
322 #if defined(MIP6) && NMIP > 0
323 		/* avoid unusable home addresses. */
324 		if ((ia->ia6_flags & IN6_IFF_HOME) &&
325 		    !mip6_ifa6_is_addr_valid_hoa(ia))
326 			continue;
327 #endif /* MIP6 && NMIP > 0 */
328 
329 		/* Rule 1: Prefer same address */
330 		if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) {
331 			ia_best = ia;
332 			BREAK(1); /* there should be no better candidate */
333 		}
334 
335 		if (ia_best == NULL)
336 			REPLACE(0);
337 
338 		/* Rule 2: Prefer appropriate scope */
339 		if (dst_scope < 0)
340 			dst_scope = in6_addrscope(&dst);
341 		new_scope = in6_addrscope(&ia->ia_addr.sin6_addr);
342 		if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) {
343 			if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0)
344 				REPLACE(2);
345 			NEXT(2);
346 		} else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) {
347 			if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0)
348 				NEXT(2);
349 			REPLACE(2);
350 		}
351 
352 		/*
353 		 * Rule 3: Avoid deprecated addresses.  Note that the case of
354 		 * !ip6_use_deprecated is already rejected above.
355 		 */
356 		if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia))
357 			NEXT(3);
358 		if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia))
359 			REPLACE(3);
360 
361 		/* Rule 4: Prefer home addresses */
362 #if defined(MIP6) && NMIP > 0
363 		if (!MIP6_IS_MN)
364 			goto skip_rule4;
365 
366 		if ((ia_best->ia6_flags & IN6_IFF_HOME) == 0 &&
367 		    (ia->ia6_flags & IN6_IFF_HOME) == 0) {
368 			/* both address are not home addresses. */
369 			goto skip_rule4;
370 		}
371 
372 		/*
373 		 * If SA is simultaneously a home address and care-of
374 		 * address and SB is not, then prefer SA. Similarly,
375 		 * if SB is simultaneously a home address and care-of
376 		 * address and SA is not, then prefer SB.
377 		 */
378 		if (((ia_best->ia6_flags & IN6_IFF_HOME) != 0 &&
379 			ia_best->ia_ifp->if_type != IFT_MIP)
380 		    &&
381 		    ((ia->ia6_flags & IN6_IFF_HOME) != 0 &&
382 			ia->ia_ifp->if_type == IFT_MIP))
383 			NEXT(4);
384 		if (((ia_best->ia6_flags & IN6_IFF_HOME) != 0 &&
385 			ia_best->ia_ifp->if_type == IFT_MIP)
386 		    &&
387 		    ((ia->ia6_flags & IN6_IFF_HOME) != 0 &&
388 			ia->ia_ifp->if_type != IFT_MIP))
389 			REPLACE(4);
390 		if (ip6po_usecoa == 0) {
391 			/*
392 			 * If SA is just a home address and SB is just
393 			 * a care-of address, then prefer
394 			 * SA. Similarly, if SB is just a home address
395 			 * and SA is just a care-of address, then
396 			 * prefer SB.
397 			 */
398 			if ((ia_best->ia6_flags & IN6_IFF_HOME) != 0 &&
399 			    (ia->ia6_flags & IN6_IFF_HOME) == 0) {
400 				NEXT(4);
401 			}
402 			if ((ia_best->ia6_flags & IN6_IFF_HOME) == 0 &&
403 			    (ia->ia6_flags & IN6_IFF_HOME) != 0) {
404 				REPLACE(4);
405 			}
406 		} else {
407 			/*
408 			 * a sender don't want to use a home address
409 			 * because:
410 			 *
411 			 * 1) we cannot use.  (ex. NS or NA to global
412 			 * addresses.)
413 			 *
414 			 * 2) a user specified not to use.
415 			 * (ex. mip6control -u)
416 			 */
417 			if ((ia_best->ia6_flags & IN6_IFF_HOME) == 0 &&
418 			    (ia->ia6_flags & IN6_IFF_HOME) != 0) {
419 				/* XXX breaks stat */
420 				NEXT(0);
421 			}
422 			if ((ia_best->ia6_flags & IN6_IFF_HOME) != 0 &&
423 			    (ia->ia6_flags & IN6_IFF_HOME) == 0) {
424 				/* XXX breaks stat */
425 				REPLACE(0);
426 			}
427 		}
428 	skip_rule4:
429 #endif /* MIP6 && NMIP > 0 */
430 
431 		/* Rule 5: Prefer outgoing interface */
432 		if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp)
433 			NEXT(5);
434 		if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp)
435 			REPLACE(5);
436 
437 		/*
438 		 * Rule 6: Prefer matching label
439 		 * Note that best_policy should be non-NULL here.
440 		 */
441 		if (dst_policy == NULL)
442 			dst_policy = lookup_addrsel_policy(dstsock);
443 		if (dst_policy->label != ADDR_LABEL_NOTAPP) {
444 			new_policy = lookup_addrsel_policy(&ia->ia_addr);
445 			if (dst_policy->label == best_policy->label &&
446 			    dst_policy->label != new_policy->label)
447 				NEXT(6);
448 			if (dst_policy->label != best_policy->label &&
449 			    dst_policy->label == new_policy->label)
450 				REPLACE(6);
451 		}
452 
453 		/*
454 		 * Rule 7: Prefer public addresses.
455 		 * We allow users to reverse the logic by configuring
456 		 * a sysctl variable, so that privacy conscious users can
457 		 * always prefer temporary addresses.
458 		 */
459 		if (opts == NULL ||
460 		    opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) {
461 			prefer_tempaddr = ip6_prefer_tempaddr;
462 		} else if (opts->ip6po_prefer_tempaddr ==
463 		    IP6PO_TEMPADDR_NOTPREFER) {
464 			prefer_tempaddr = 0;
465 		} else
466 			prefer_tempaddr = 1;
467 		if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
468 		    (ia->ia6_flags & IN6_IFF_TEMPORARY)) {
469 			if (prefer_tempaddr)
470 				REPLACE(7);
471 			else
472 				NEXT(7);
473 		}
474 		if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
475 		    !(ia->ia6_flags & IN6_IFF_TEMPORARY)) {
476 			if (prefer_tempaddr)
477 				NEXT(7);
478 			else
479 				REPLACE(7);
480 		}
481 
482 		/*
483 		 * Rule 8: prefer addresses on alive interfaces.
484 		 * This is a KAME specific rule.
485 		 */
486 		if ((ia_best->ia_ifp->if_flags & IFF_UP) &&
487 		    !(ia->ia_ifp->if_flags & IFF_UP))
488 			NEXT(8);
489 		if (!(ia_best->ia_ifp->if_flags & IFF_UP) &&
490 		    (ia->ia_ifp->if_flags & IFF_UP))
491 			REPLACE(8);
492 
493 		/*
494 		 * Rule 9: prefer addresses on "preferred" interfaces.
495 		 * This is a KAME specific rule.
496 		 */
497 #ifdef notyet			/* until introducing address selection */
498 #define NDI_BEST ND_IFINFO(ia_best->ia_ifp)
499 #define NDI_NEW  ND_IFINFO(ia->ia_ifp)
500 		if ((NDI_BEST->flags & ND6_IFF_PREFER_SOURCE) &&
501 		    !(NDI_NEW->flags & ND6_IFF_PREFER_SOURCE))
502 			NEXT(9);
503 		if (!(NDI_BEST->flags & ND6_IFF_PREFER_SOURCE) &&
504 		    (NDI_NEW->flags & ND6_IFF_PREFER_SOURCE))
505 			REPLACE(9);
506 #undef NDI_BEST
507 #undef NDI_NEW
508 #endif
509 
510 		/*
511 		 * Rule 14: Use longest matching prefix.
512 		 * Note: in the address selection draft, this rule is
513 		 * documented as "Rule 8".  However, since it is also
514 		 * documented that this rule can be overridden, we assign
515 		 * a large number so that it is easy to assign smaller numbers
516 		 * to more preferred rules.
517 		 */
518 		new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst);
519 		if (best_matchlen < new_matchlen)
520 			REPLACE(14);
521 		if (new_matchlen < best_matchlen)
522 			NEXT(14);
523 
524 		/* Rule 15 is reserved. */
525 
526 		/*
527 		 * Last resort: just keep the current candidate.
528 		 * Or, do we need more rules?
529 		 */
530 		continue;
531 
532 	  replace:
533 		ia_best = ia;
534 		best_scope = (new_scope >= 0 ? new_scope :
535 			      in6_addrscope(&ia_best->ia_addr.sin6_addr));
536 		best_policy = (new_policy ? new_policy :
537 			       lookup_addrsel_policy(&ia_best->ia_addr));
538 		best_matchlen = (new_matchlen >= 0 ? new_matchlen :
539 				 in6_matchlen(&ia_best->ia_addr.sin6_addr,
540 					      &dst));
541 
542 	  next:
543 		continue;
544 
545 	  out:
546 		break;
547 	}
548 
549 	if ((ia = ia_best) == NULL) {
550 		*errorp = EADDRNOTAVAIL;
551 		return (NULL);
552 	}
553 
554 	return (&ia->ia_addr.sin6_addr);
555 }
556 #undef REPLACE
557 #undef BREAK
558 #undef NEXT
559 
560 static int
561 selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
562 	struct ip6_moptions *mopts, struct route *ro, struct ifnet **retifp,
563 	struct rtentry **retrt, int clone, int norouteok)
564 {
565 	int error = 0;
566 	struct ifnet *ifp = NULL;
567 	struct rtentry *rt = NULL;
568 	struct sockaddr_in6 *sin6_next;
569 	struct in6_pktinfo *pi = NULL;
570 	struct in6_addr *dst;
571 
572 	dst = &dstsock->sin6_addr;
573 
574 #if 0
575 	if (dstsock->sin6_addr.s6_addr32[0] == 0 &&
576 	    dstsock->sin6_addr.s6_addr32[1] == 0 &&
577 	    !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) {
578 		printf("in6_selectroute: strange destination %s\n",
579 		       ip6_sprintf(&dstsock->sin6_addr));
580 	} else {
581 		printf("in6_selectroute: destination = %s%%%d\n",
582 		       ip6_sprintf(&dstsock->sin6_addr),
583 		       dstsock->sin6_scope_id); /* for debug */
584 	}
585 #endif
586 
587 	/* If the caller specify the outgoing interface explicitly, use it. */
588 	if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) {
589 		/* XXX boundary check is assumed to be already done. */
590 		ifp = if_byindex(pi->ipi6_ifindex);
591 		if (ifp != NULL &&
592 		    (norouteok || retrt == NULL ||
593 		    IN6_IS_ADDR_MULTICAST(dst))) {
594 			/*
595 			 * we do not have to check or get the route for
596 			 * multicast.
597 			 */
598 			goto done;
599 		} else
600 			goto getroute;
601 	}
602 
603 	/*
604 	 * If the destination address is a multicast address and the outgoing
605 	 * interface for the address is specified by the caller, use it.
606 	 */
607 	if (IN6_IS_ADDR_MULTICAST(dst) &&
608 	    mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) {
609 		goto done; /* we do not need a route for multicast. */
610 	}
611 
612   getroute:
613 	/*
614 	 * If the next hop address for the packet is specified by the caller,
615 	 * use it as the gateway.
616 	 */
617 	if (opts && opts->ip6po_nexthop) {
618 		struct route *ron;
619 
620 		sin6_next = satosin6(opts->ip6po_nexthop);
621 
622 		/* at this moment, we only support AF_INET6 next hops */
623 		if (sin6_next->sin6_family != AF_INET6) {
624 			error = EAFNOSUPPORT; /* or should we proceed? */
625 			goto done;
626 		}
627 
628 		/*
629 		 * If the next hop is an IPv6 address, then the node identified
630 		 * by that address must be a neighbor of the sending host.
631 		 */
632 		ron = &opts->ip6po_nextroute;
633 		if ((rt = rtcache_lookup(ron, sin6tosa(sin6_next))) == NULL ||
634 		    (rt->rt_flags & RTF_GATEWAY) != 0 ||
635 		    !nd6_is_addr_neighbor(sin6_next, rt->rt_ifp)) {
636 			rtcache_free(ron);
637 			error = EHOSTUNREACH;
638 			goto done;
639 		}
640 		ifp = rt->rt_ifp;
641 
642 		/*
643 		 * When cloning is required, try to allocate a route to the
644 		 * destination so that the caller can store path MTU
645 		 * information.
646 		 */
647 		if (!clone)
648 			goto done;
649 	}
650 
651 	/*
652 	 * Use a cached route if it exists and is valid, else try to allocate
653 	 * a new one.  Note that we should check the address family of the
654 	 * cached destination, in case of sharing the cache with IPv4.
655 	 */
656 	if (ro != NULL) {
657 		union {
658 			struct sockaddr		dst;
659 			struct sockaddr_in6	dst6;
660 		} u;
661 
662 		/* No route yet, so try to acquire one */
663 		u.dst6 = *dstsock;
664 		u.dst6.sin6_scope_id = 0;
665 		rt = rtcache_lookup1(ro, &u.dst, clone);
666 
667 		/*
668 		 * do not care about the result if we have the nexthop
669 		 * explicitly specified.
670 		 */
671 		if (opts && opts->ip6po_nexthop)
672 			goto done;
673 
674 		if (rt == NULL)
675 			error = EHOSTUNREACH;
676 		else
677 			ifp = rt->rt_ifp;
678 
679 		/*
680 		 * Check if the outgoing interface conflicts with
681 		 * the interface specified by ipi6_ifindex (if specified).
682 		 * Note that loopback interface is always okay.
683 		 * (this may happen when we are sending a packet to one of
684 		 *  our own addresses.)
685 		 */
686 		if (opts && opts->ip6po_pktinfo &&
687 		    opts->ip6po_pktinfo->ipi6_ifindex) {
688 			if (!(ifp->if_flags & IFF_LOOPBACK) &&
689 			    ifp->if_index !=
690 			    opts->ip6po_pktinfo->ipi6_ifindex) {
691 				error = EHOSTUNREACH;
692 				goto done;
693 			}
694 		}
695 	}
696 
697   done:
698 	if (ifp == NULL && rt == NULL) {
699 		/*
700 		 * This can happen if the caller did not pass a cached route
701 		 * nor any other hints.  We treat this case an error.
702 		 */
703 		error = EHOSTUNREACH;
704 	}
705 	if (error == EHOSTUNREACH)
706 		IP6_STATINC(IP6_STAT_NOROUTE);
707 
708 	if (retifp != NULL)
709 		*retifp = ifp;
710 	if (retrt != NULL)
711 		*retrt = rt;	/* rt may be NULL */
712 
713 	return (error);
714 }
715 
716 static int
717 in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
718 	struct ip6_moptions *mopts, struct route *ro, struct ifnet **retifp)
719 {
720 	int error, clone;
721 	struct rtentry *rt = NULL;
722 
723 	clone = IN6_IS_ADDR_MULTICAST(&dstsock->sin6_addr) ? 0 : 1;
724 	if ((error = selectroute(dstsock, opts, mopts, ro, retifp,
725 	    &rt, clone, 1)) != 0) {
726 		return (error);
727 	}
728 
729 	/*
730 	 * do not use a rejected or black hole route.
731 	 * XXX: this check should be done in the L2 output routine.
732 	 * However, if we skipped this check here, we'd see the following
733 	 * scenario:
734 	 * - install a rejected route for a scoped address prefix
735 	 *   (like fe80::/10)
736 	 * - send a packet to a destination that matches the scoped prefix,
737 	 *   with ambiguity about the scope zone.
738 	 * - pick the outgoing interface from the route, and disambiguate the
739 	 *   scope zone with the interface.
740 	 * - ip6_output() would try to get another route with the "new"
741 	 *   destination, which may be valid.
742 	 * - we'd see no error on output.
743 	 * Although this may not be very harmful, it should still be confusing.
744 	 * We thus reject the case here.
745 	 */
746 	if (rt && (rt->rt_flags & (RTF_REJECT | RTF_BLACKHOLE)))
747 		return (rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
748 
749 	/*
750 	 * Adjust the "outgoing" interface.  If we're going to loop the packet
751 	 * back to ourselves, the ifp would be the loopback interface.
752 	 * However, we'd rather know the interface associated to the
753 	 * destination address (which should probably be one of our own
754 	 * addresses.)
755 	 */
756 	if (rt && rt->rt_ifa && rt->rt_ifa->ifa_ifp)
757 		*retifp = rt->rt_ifa->ifa_ifp;
758 
759 	return (0);
760 }
761 
762 /*
763  * close - meaningful only for bsdi and freebsd.
764  */
765 
766 int
767 in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
768 	struct ip6_moptions *mopts, struct route *ro, struct ifnet **retifp,
769 	struct rtentry **retrt, int clone)
770 {
771 	return selectroute(dstsock, opts, mopts, ro, retifp,
772 	    retrt, clone, 0);
773 }
774 
775 /*
776  * Default hop limit selection. The precedence is as follows:
777  * 1. Hoplimit value specified via ioctl.
778  * 2. (If the outgoing interface is detected) the current
779  *     hop limit of the interface specified by router advertisement.
780  * 3. The system default hoplimit.
781 */
782 int
783 in6_selecthlim(struct in6pcb *in6p, struct ifnet *ifp)
784 {
785 	if (in6p && in6p->in6p_hops >= 0)
786 		return (in6p->in6p_hops);
787 	else if (ifp)
788 		return (ND_IFINFO(ifp)->chlim);
789 	else
790 		return (ip6_defhlim);
791 }
792 
793 int
794 in6_selecthlim_rt(struct in6pcb *in6p)
795 {
796 	struct rtentry *rt;
797 
798 	if (in6p == NULL)
799 		return in6_selecthlim(in6p, NULL);
800 
801 	rt = rtcache_validate(&in6p->in6p_route);
802 	if (rt != NULL)
803 		return in6_selecthlim(in6p, rt->rt_ifp);
804 	else
805 		return in6_selecthlim(in6p, NULL);
806 }
807 
808 /*
809  * Find an empty port and set it to the specified PCB.
810  */
811 int
812 in6_pcbsetport(struct sockaddr_in6 *sin6, struct in6pcb *in6p, struct lwp *l)
813 {
814 	struct socket *so = in6p->in6p_socket;
815 	struct inpcbtable *table = in6p->in6p_table;
816 	u_int16_t lport, *lastport;
817 	enum kauth_network_req req;
818 	int error = 0;
819 
820 	if (in6p->in6p_flags & IN6P_LOWPORT) {
821 #ifndef IPNOPRIVPORTS
822 		req = KAUTH_REQ_NETWORK_BIND_PRIVPORT;
823 #else
824 		req = KAUTH_REQ_NETWORK_BIND_PORT;
825 #endif
826 		lastport = &table->inpt_lastlow;
827 	} else {
828 		req = KAUTH_REQ_NETWORK_BIND_PORT;
829 
830 		lastport = &table->inpt_lastport;
831 	}
832 
833 	/* XXX-kauth: KAUTH_REQ_NETWORK_BIND_AUTOASSIGN_{,PRIV}PORT */
834 	error = kauth_authorize_network(l->l_cred, KAUTH_NETWORK_BIND, req, so,
835 	    sin6, NULL);
836 	if (error)
837 		return (EACCES);
838 
839        /*
840         * Use RFC6056 randomized port selection
841         */
842 	error = portalgo_randport(&lport, &in6p->in6p_head, l->l_cred);
843 	if (error)
844 		return error;
845 
846 	in6p->in6p_flags |= IN6P_ANONPORT;
847 	*lastport = lport;
848 	in6p->in6p_lport = htons(lport);
849 	in6_pcbstate(in6p, IN6P_BOUND);
850 	return (0);		/* success */
851 }
852 
853 void
854 addrsel_policy_init(void)
855 {
856 	init_policy_queue();
857 
858 	/* initialize the "last resort" policy */
859 	memset(&defaultaddrpolicy, 0, sizeof(defaultaddrpolicy));
860 	defaultaddrpolicy.label = ADDR_LABEL_NOTAPP;
861 }
862 
863 static struct in6_addrpolicy *
864 lookup_addrsel_policy(struct sockaddr_in6 *key)
865 {
866 	struct in6_addrpolicy *match = NULL;
867 
868 	match = match_addrsel_policy(key);
869 
870 	if (match == NULL)
871 		match = &defaultaddrpolicy;
872 	else
873 		match->use++;
874 
875 	return (match);
876 }
877 
878 /*
879  * Subroutines to manage the address selection policy table via sysctl.
880  */
881 struct sel_walkarg {
882 	size_t	w_total;
883 	size_t	w_given;
884 	void *	w_where;
885 	void *w_limit;
886 };
887 
888 int sysctl_net_inet6_addrctlpolicy(SYSCTLFN_ARGS);
889 int
890 sysctl_net_inet6_addrctlpolicy(SYSCTLFN_ARGS)
891 {
892 	int error = 0;
893 	int s;
894 
895 	s = splsoftnet();
896 
897 	if (newp) {
898 		error = EPERM;
899 		goto end;
900 	}
901 	if (oldp && oldlenp == NULL) {
902 		error = EINVAL;
903 		goto end;
904 	}
905 	if (oldp || oldlenp) {
906 		struct sel_walkarg w;
907 		size_t oldlen = *oldlenp;
908 
909 		memset(&w, 0, sizeof(w));
910 		w.w_given = oldlen;
911 		w.w_where = oldp;
912 		if (oldp)
913 			w.w_limit = (char *)oldp + oldlen;
914 
915 		error = walk_addrsel_policy(dump_addrsel_policyent, &w);
916 
917 		*oldlenp = w.w_total;
918 		if (oldp && w.w_total > oldlen && error == 0)
919 			error = ENOMEM;
920 	}
921 
922   end:
923 	splx(s);
924 
925 	return (error);
926 }
927 
928 int
929 in6_src_ioctl(u_long cmd, void *data)
930 {
931 	int i;
932 	struct in6_addrpolicy ent0;
933 
934 	if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY)
935 		return (EOPNOTSUPP); /* check for safety */
936 
937 	ent0 = *(struct in6_addrpolicy *)data;
938 
939 	if (ent0.label == ADDR_LABEL_NOTAPP)
940 		return (EINVAL);
941 	/* check if the prefix mask is consecutive. */
942 	if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0)
943 		return (EINVAL);
944 	/* clear trailing garbages (if any) of the prefix address. */
945 	for (i = 0; i < 4; i++) {
946 		ent0.addr.sin6_addr.s6_addr32[i] &=
947 			ent0.addrmask.sin6_addr.s6_addr32[i];
948 	}
949 	ent0.use = 0;
950 
951 	switch (cmd) {
952 	case SIOCAADDRCTL_POLICY:
953 		return (add_addrsel_policyent(&ent0));
954 	case SIOCDADDRCTL_POLICY:
955 		return (delete_addrsel_policyent(&ent0));
956 	}
957 
958 	return (0);		/* XXX: compromise compilers */
959 }
960 
961 /*
962  * The followings are implementation of the policy table using a
963  * simple tail queue.
964  * XXX such details should be hidden.
965  * XXX implementation using binary tree should be more efficient.
966  */
967 struct addrsel_policyent {
968 	TAILQ_ENTRY(addrsel_policyent) ape_entry;
969 	struct in6_addrpolicy ape_policy;
970 };
971 
972 TAILQ_HEAD(addrsel_policyhead, addrsel_policyent);
973 
974 struct addrsel_policyhead addrsel_policytab;
975 
976 static void
977 init_policy_queue(void)
978 {
979 	TAILQ_INIT(&addrsel_policytab);
980 }
981 
982 static int
983 add_addrsel_policyent(struct in6_addrpolicy *newpolicy)
984 {
985 	struct addrsel_policyent *newpol, *pol;
986 
987 	/* duplication check */
988 	TAILQ_FOREACH(pol, &addrsel_policytab, ape_entry) {
989 		if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr,
990 		    &pol->ape_policy.addr.sin6_addr) &&
991 		    IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr,
992 		    &pol->ape_policy.addrmask.sin6_addr)) {
993 			return (EEXIST);	/* or override it? */
994 		}
995 	}
996 
997 	newpol = malloc(sizeof(*newpol), M_IFADDR, M_WAITOK|M_ZERO);
998 
999 	/* XXX: should validate entry */
1000 	newpol->ape_policy = *newpolicy;
1001 
1002 	TAILQ_INSERT_TAIL(&addrsel_policytab, newpol, ape_entry);
1003 
1004 	return (0);
1005 }
1006 
1007 static int
1008 delete_addrsel_policyent(struct in6_addrpolicy *key)
1009 {
1010 	struct addrsel_policyent *pol;
1011 
1012 	/* search for the entry in the table */
1013 	for (pol = TAILQ_FIRST(&addrsel_policytab); pol;
1014 	     pol = TAILQ_NEXT(pol, ape_entry)) {
1015 		if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr,
1016 		    &pol->ape_policy.addr.sin6_addr) &&
1017 		    IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr,
1018 		    &pol->ape_policy.addrmask.sin6_addr)) {
1019 			break;
1020 		}
1021 	}
1022 	if (pol == NULL) {
1023 		return (ESRCH);
1024 	}
1025 
1026 	TAILQ_REMOVE(&addrsel_policytab, pol, ape_entry);
1027 
1028 	return (0);
1029 }
1030 
1031 static int
1032 walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *), void *w)
1033 {
1034 	struct addrsel_policyent *pol;
1035 	int error = 0;
1036 
1037 	TAILQ_FOREACH(pol, &addrsel_policytab, ape_entry) {
1038 		if ((error = (*callback)(&pol->ape_policy, w)) != 0)
1039 			return error;
1040 	}
1041 
1042 	return error;
1043 }
1044 
1045 static int
1046 dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg)
1047 {
1048 	int error = 0;
1049 	struct sel_walkarg *w = arg;
1050 
1051 	if (w->w_where && (char *)w->w_where + sizeof(*pol) <= (char *)w->w_limit) {
1052 		if ((error = copyout(pol, w->w_where, sizeof(*pol))) != 0)
1053 			return error;
1054 		w->w_where = (char *)w->w_where + sizeof(*pol);
1055 	}
1056 	w->w_total += sizeof(*pol);
1057 
1058 	return error;
1059 }
1060 
1061 static struct in6_addrpolicy *
1062 match_addrsel_policy(struct sockaddr_in6 *key)
1063 {
1064 	struct addrsel_policyent *pent;
1065 	struct in6_addrpolicy *bestpol = NULL, *pol;
1066 	int matchlen, bestmatchlen = -1;
1067 	u_char *mp, *ep, *k, *p, m;
1068 
1069 	for (pent = TAILQ_FIRST(&addrsel_policytab); pent;
1070 	     pent = TAILQ_NEXT(pent, ape_entry)) {
1071 		matchlen = 0;
1072 
1073 		pol = &pent->ape_policy;
1074 		mp = (u_char *)&pol->addrmask.sin6_addr;
1075 		ep = mp + 16;	/* XXX: scope field? */
1076 		k = (u_char *)&key->sin6_addr;
1077 		p = (u_char *)&pol->addr.sin6_addr;
1078 		for (; mp < ep && *mp; mp++, k++, p++) {
1079 			m = *mp;
1080 			if ((*k & m) != *p)
1081 				goto next; /* not match */
1082 			if (m == 0xff) /* short cut for a typical case */
1083 				matchlen += 8;
1084 			else {
1085 				while (m >= 0x80) {
1086 					matchlen++;
1087 					m <<= 1;
1088 				}
1089 			}
1090 		}
1091 
1092 		/* matched.  check if this is better than the current best. */
1093 		if (bestpol == NULL ||
1094 		    matchlen > bestmatchlen) {
1095 			bestpol = pol;
1096 			bestmatchlen = matchlen;
1097 		}
1098 
1099 	  next:
1100 		continue;
1101 	}
1102 
1103 	return (bestpol);
1104 }
1105