xref: /freebsd/sys/netinet6/in6_src.c (revision 0a3ed02ab8527f5f4513116da55f4e7500fe806b)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (C) 1995, 1996, 1997, and 1998 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  *	$KAME: in6_src.c,v 1.132 2003/08/26 04:42:27 keiichi Exp $
32  */
33 
34 /*-
35  * Copyright (c) 1982, 1986, 1991, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. Neither the name of the University nor the names of its contributors
47  *    may be used to endorse or promote products derived from this software
48  *    without specific prior written permission.
49  *
50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60  * SUCH DAMAGE.
61  *
62  *	@(#)in_pcb.c	8.2 (Berkeley) 1/4/94
63  */
64 
65 #include <sys/cdefs.h>
66 __FBSDID("$FreeBSD$");
67 
68 #include "opt_inet.h"
69 #include "opt_inet6.h"
70 #include "opt_mpath.h"
71 
72 #include <sys/param.h>
73 #include <sys/systm.h>
74 #include <sys/lock.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/priv.h>
78 #include <sys/protosw.h>
79 #include <sys/socket.h>
80 #include <sys/socketvar.h>
81 #include <sys/sockio.h>
82 #include <sys/sysctl.h>
83 #include <sys/errno.h>
84 #include <sys/time.h>
85 #include <sys/jail.h>
86 #include <sys/kernel.h>
87 #include <sys/rmlock.h>
88 #include <sys/sx.h>
89 
90 #include <net/if.h>
91 #include <net/if_var.h>
92 #include <net/if_dl.h>
93 #include <net/route.h>
94 #include <net/route/nhop.h>
95 #include <net/if_llatbl.h>
96 
97 #include <netinet/in.h>
98 #include <netinet/in_var.h>
99 #include <netinet/in_systm.h>
100 #include <netinet/ip.h>
101 #include <netinet/in_pcb.h>
102 #include <netinet/ip_var.h>
103 #include <netinet/udp.h>
104 #include <netinet/udp_var.h>
105 
106 #include <netinet6/in6_var.h>
107 #include <netinet/ip6.h>
108 #include <netinet6/in6_fib.h>
109 #include <netinet6/in6_pcb.h>
110 #include <netinet6/ip6_var.h>
111 #include <netinet6/scope6_var.h>
112 #include <netinet6/nd6.h>
113 
114 static struct mtx addrsel_lock;
115 #define	ADDRSEL_LOCK_INIT()	mtx_init(&addrsel_lock, "addrsel_lock", NULL, MTX_DEF)
116 #define	ADDRSEL_LOCK()		mtx_lock(&addrsel_lock)
117 #define	ADDRSEL_UNLOCK()	mtx_unlock(&addrsel_lock)
118 #define	ADDRSEL_LOCK_ASSERT()	mtx_assert(&addrsel_lock, MA_OWNED)
119 
120 static struct sx addrsel_sxlock;
121 #define	ADDRSEL_SXLOCK_INIT()	sx_init(&addrsel_sxlock, "addrsel_sxlock")
122 #define	ADDRSEL_SLOCK()		sx_slock(&addrsel_sxlock)
123 #define	ADDRSEL_SUNLOCK()	sx_sunlock(&addrsel_sxlock)
124 #define	ADDRSEL_XLOCK()		sx_xlock(&addrsel_sxlock)
125 #define	ADDRSEL_XUNLOCK()	sx_xunlock(&addrsel_sxlock)
126 
127 #define ADDR_LABEL_NOTAPP (-1)
128 VNET_DEFINE_STATIC(struct in6_addrpolicy, defaultaddrpolicy);
129 #define	V_defaultaddrpolicy		VNET(defaultaddrpolicy)
130 
131 VNET_DEFINE(int, ip6_prefer_tempaddr) = 0;
132 
133 static int selectroute(struct sockaddr_in6 *, struct ip6_pktopts *,
134 	struct ip6_moptions *, struct route_in6 *, struct ifnet **,
135 	struct nhop_object **, int, u_int, uint32_t);
136 static int in6_selectif(struct sockaddr_in6 *, struct ip6_pktopts *,
137 	struct ip6_moptions *, struct ifnet **,
138 	struct ifnet *, u_int);
139 static int in6_selectsrc(uint32_t, struct sockaddr_in6 *,
140 	struct ip6_pktopts *, struct inpcb *, struct ucred *,
141 	struct ifnet **, struct in6_addr *);
142 
143 static struct in6_addrpolicy *lookup_addrsel_policy(struct sockaddr_in6 *);
144 
145 static void init_policy_queue(void);
146 static int add_addrsel_policyent(struct in6_addrpolicy *);
147 static int delete_addrsel_policyent(struct in6_addrpolicy *);
148 static int walk_addrsel_policy(int (*)(struct in6_addrpolicy *, void *),
149 	void *);
150 static int dump_addrsel_policyent(struct in6_addrpolicy *, void *);
151 static struct in6_addrpolicy *match_addrsel_policy(struct sockaddr_in6 *);
152 
153 /*
154  * Return an IPv6 address, which is the most appropriate for a given
155  * destination and user specified options.
156  * If necessary, this function lookups the routing table and returns
157  * an entry to the caller for later use.
158  */
159 #define REPLACE(r) do {\
160 	IP6STAT_INC(ip6s_sources_rule[(r)]); \
161 	/* { \
162 	char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
163 	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)); \
164 	} */ \
165 	goto replace; \
166 } while(0)
167 #define NEXT(r) do {\
168 	/* { \
169 	char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
170 	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)); \
171 	} */ \
172 	goto next;		/* XXX: we can't use 'continue' here */ \
173 } while(0)
174 #define BREAK(r) do { \
175 	IP6STAT_INC(ip6s_sources_rule[(r)]); \
176 	goto out;		/* XXX: we can't use 'break' here */ \
177 } while(0)
178 
179 static int
180 in6_selectsrc(uint32_t fibnum, struct sockaddr_in6 *dstsock,
181     struct ip6_pktopts *opts, struct inpcb *inp, struct ucred *cred,
182     struct ifnet **ifpp, struct in6_addr *srcp)
183 {
184 	struct rm_priotracker in6_ifa_tracker;
185 	struct in6_addr dst, tmp;
186 	struct ifnet *ifp = NULL, *oifp = NULL;
187 	struct in6_ifaddr *ia = NULL, *ia_best = NULL;
188 	struct in6_pktinfo *pi = NULL;
189 	int dst_scope = -1, best_scope = -1, best_matchlen = -1;
190 	struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL;
191 	u_int32_t odstzone;
192 	int prefer_tempaddr;
193 	int error;
194 	struct ip6_moptions *mopts;
195 
196 	KASSERT(srcp != NULL, ("%s: srcp is NULL", __func__));
197 
198 	dst = dstsock->sin6_addr; /* make a copy for local operation */
199 	if (ifpp) {
200 		/*
201 		 * Save a possibly passed in ifp for in6_selectsrc. Only
202 		 * neighbor discovery code should use this feature, where
203 		 * we may know the interface but not the FIB number holding
204 		 * the connected subnet in case someone deleted it from the
205 		 * default FIB and we need to check the interface.
206 		 */
207 		if (*ifpp != NULL)
208 			oifp = *ifpp;
209 		*ifpp = NULL;
210 	}
211 
212 	if (inp != NULL) {
213 		INP_LOCK_ASSERT(inp);
214 		mopts = inp->in6p_moptions;
215 	} else {
216 		mopts = NULL;
217 	}
218 
219 	/*
220 	 * If the source address is explicitly specified by the caller,
221 	 * check if the requested source address is indeed a unicast address
222 	 * assigned to the node, and can be used as the packet's source
223 	 * address.  If everything is okay, use the address as source.
224 	 */
225 	if (opts && (pi = opts->ip6po_pktinfo) &&
226 	    !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) {
227 		/* get the outgoing interface */
228 		if ((error = in6_selectif(dstsock, opts, mopts, &ifp, oifp,
229 		    fibnum))
230 		    != 0)
231 			return (error);
232 
233 		/*
234 		 * determine the appropriate zone id of the source based on
235 		 * the zone of the destination and the outgoing interface.
236 		 * If the specified address is ambiguous wrt the scope zone,
237 		 * the interface must be specified; otherwise, ifa_ifwithaddr()
238 		 * will fail matching the address.
239 		 */
240 		tmp = pi->ipi6_addr;
241 		if (ifp) {
242 			error = in6_setscope(&tmp, ifp, &odstzone);
243 			if (error)
244 				return (error);
245 		}
246 		if (cred != NULL && (error = prison_local_ip6(cred,
247 		    &tmp, (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0)
248 			return (error);
249 
250 		/*
251 		 * If IPV6_BINDANY socket option is set, we allow to specify
252 		 * non local addresses as source address in IPV6_PKTINFO
253 		 * ancillary data.
254 		 */
255 		if ((inp->inp_flags & INP_BINDANY) == 0) {
256 			ia = in6ifa_ifwithaddr(&tmp, 0 /* XXX */);
257 			if (ia == NULL || (ia->ia6_flags & (IN6_IFF_ANYCAST |
258 			    IN6_IFF_NOTREADY))) {
259 				if (ia != NULL)
260 					ifa_free(&ia->ia_ifa);
261 				return (EADDRNOTAVAIL);
262 			}
263 			bcopy(&ia->ia_addr.sin6_addr, srcp, sizeof(*srcp));
264 			ifa_free(&ia->ia_ifa);
265 		} else
266 			bcopy(&tmp, srcp, sizeof(*srcp));
267 		pi->ipi6_addr = tmp; /* XXX: this overrides pi */
268 		if (ifpp)
269 			*ifpp = ifp;
270 		return (0);
271 	}
272 
273 	/*
274 	 * Otherwise, if the socket has already bound the source, just use it.
275 	 */
276 	if (inp != NULL && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
277 		if (cred != NULL &&
278 		    (error = prison_local_ip6(cred, &inp->in6p_laddr,
279 		    ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
280 			return (error);
281 		bcopy(&inp->in6p_laddr, srcp, sizeof(*srcp));
282 		return (0);
283 	}
284 
285 	/*
286 	 * Bypass source address selection and use the primary jail IP
287 	 * if requested.
288 	 */
289 	if (cred != NULL && !prison_saddrsel_ip6(cred, srcp))
290 		return (0);
291 
292 	/*
293 	 * If the address is not specified, choose the best one based on
294 	 * the outgoing interface and the destination address.
295 	 */
296 	/* get the outgoing interface */
297 	if ((error = in6_selectif(dstsock, opts, mopts, &ifp, oifp,
298 	    (inp != NULL) ? inp->inp_inc.inc_fibnum : fibnum)) != 0)
299 		return (error);
300 
301 #ifdef DIAGNOSTIC
302 	if (ifp == NULL)	/* this should not happen */
303 		panic("in6_selectsrc: NULL ifp");
304 #endif
305 	error = in6_setscope(&dst, ifp, &odstzone);
306 	if (error)
307 		return (error);
308 
309 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
310 	CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
311 		int new_scope = -1, new_matchlen = -1;
312 		struct in6_addrpolicy *new_policy = NULL;
313 		u_int32_t srczone, osrczone, dstzone;
314 		struct in6_addr src;
315 		struct ifnet *ifp1 = ia->ia_ifp;
316 
317 		/*
318 		 * We'll never take an address that breaks the scope zone
319 		 * of the destination.  We also skip an address if its zone
320 		 * does not contain the outgoing interface.
321 		 * XXX: we should probably use sin6_scope_id here.
322 		 */
323 		if (in6_setscope(&dst, ifp1, &dstzone) ||
324 		    odstzone != dstzone) {
325 			continue;
326 		}
327 		src = ia->ia_addr.sin6_addr;
328 		if (in6_setscope(&src, ifp, &osrczone) ||
329 		    in6_setscope(&src, ifp1, &srczone) ||
330 		    osrczone != srczone) {
331 			continue;
332 		}
333 
334 		/* avoid unusable addresses */
335 		if ((ia->ia6_flags &
336 		     (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) {
337 				continue;
338 		}
339 		if (!V_ip6_use_deprecated && IFA6_IS_DEPRECATED(ia))
340 			continue;
341 
342 		/* If jailed only take addresses of the jail into account. */
343 		if (cred != NULL &&
344 		    prison_check_ip6(cred, &ia->ia_addr.sin6_addr) != 0)
345 			continue;
346 
347 		/* Rule 1: Prefer same address */
348 		if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) {
349 			ia_best = ia;
350 			BREAK(1); /* there should be no better candidate */
351 		}
352 
353 		if (ia_best == NULL)
354 			REPLACE(0);
355 
356 		/* Rule 2: Prefer appropriate scope */
357 		if (dst_scope < 0)
358 			dst_scope = in6_addrscope(&dst);
359 		new_scope = in6_addrscope(&ia->ia_addr.sin6_addr);
360 		if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) {
361 			if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0)
362 				REPLACE(2);
363 			NEXT(2);
364 		} else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) {
365 			if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0)
366 				NEXT(2);
367 			REPLACE(2);
368 		}
369 
370 		/*
371 		 * Rule 3: Avoid deprecated addresses.  Note that the case of
372 		 * !ip6_use_deprecated is already rejected above.
373 		 */
374 		if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia))
375 			NEXT(3);
376 		if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia))
377 			REPLACE(3);
378 
379 		/* Rule 4: Prefer home addresses */
380 		/*
381 		 * XXX: This is a TODO.  We should probably merge the MIP6
382 		 * case above.
383 		 */
384 
385 		/* Rule 5: Prefer outgoing interface */
386 		if (!(ND_IFINFO(ifp)->flags & ND6_IFF_NO_PREFER_IFACE)) {
387 			if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp)
388 				NEXT(5);
389 			if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp)
390 				REPLACE(5);
391 		}
392 
393 		/*
394 		 * Rule 6: Prefer matching label
395 		 * Note that best_policy should be non-NULL here.
396 		 */
397 		if (dst_policy == NULL)
398 			dst_policy = lookup_addrsel_policy(dstsock);
399 		if (dst_policy->label != ADDR_LABEL_NOTAPP) {
400 			new_policy = lookup_addrsel_policy(&ia->ia_addr);
401 			if (dst_policy->label == best_policy->label &&
402 			    dst_policy->label != new_policy->label)
403 				NEXT(6);
404 			if (dst_policy->label != best_policy->label &&
405 			    dst_policy->label == new_policy->label)
406 				REPLACE(6);
407 		}
408 
409 		/*
410 		 * Rule 7: Prefer public addresses.
411 		 * We allow users to reverse the logic by configuring
412 		 * a sysctl variable, so that privacy conscious users can
413 		 * always prefer temporary addresses.
414 		 */
415 		if (opts == NULL ||
416 		    opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) {
417 			prefer_tempaddr = V_ip6_prefer_tempaddr;
418 		} else if (opts->ip6po_prefer_tempaddr ==
419 		    IP6PO_TEMPADDR_NOTPREFER) {
420 			prefer_tempaddr = 0;
421 		} else
422 			prefer_tempaddr = 1;
423 		if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
424 		    (ia->ia6_flags & IN6_IFF_TEMPORARY)) {
425 			if (prefer_tempaddr)
426 				REPLACE(7);
427 			else
428 				NEXT(7);
429 		}
430 		if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
431 		    !(ia->ia6_flags & IN6_IFF_TEMPORARY)) {
432 			if (prefer_tempaddr)
433 				NEXT(7);
434 			else
435 				REPLACE(7);
436 		}
437 
438 		/*
439 		 * Rule 8: prefer addresses on alive interfaces.
440 		 * This is a KAME specific rule.
441 		 */
442 		if ((ia_best->ia_ifp->if_flags & IFF_UP) &&
443 		    !(ia->ia_ifp->if_flags & IFF_UP))
444 			NEXT(8);
445 		if (!(ia_best->ia_ifp->if_flags & IFF_UP) &&
446 		    (ia->ia_ifp->if_flags & IFF_UP))
447 			REPLACE(8);
448 
449 		/*
450 		 * Rule 9: prefer address with better virtual status.
451 		 */
452 		if (ifa_preferred(&ia_best->ia_ifa, &ia->ia_ifa))
453 			REPLACE(9);
454 		if (ifa_preferred(&ia->ia_ifa, &ia_best->ia_ifa))
455 			NEXT(9);
456 
457 		/*
458 		 * Rule 10: prefer address with `prefer_source' flag.
459 		 */
460 		if ((ia_best->ia6_flags & IN6_IFF_PREFER_SOURCE) == 0 &&
461 		    (ia->ia6_flags & IN6_IFF_PREFER_SOURCE) != 0)
462 			REPLACE(10);
463 		if ((ia_best->ia6_flags & IN6_IFF_PREFER_SOURCE) != 0 &&
464 		    (ia->ia6_flags & IN6_IFF_PREFER_SOURCE) == 0)
465 			NEXT(10);
466 
467 		/*
468 		 * Rule 14: Use longest matching prefix.
469 		 * Note: in the address selection draft, this rule is
470 		 * documented as "Rule 8".  However, since it is also
471 		 * documented that this rule can be overridden, we assign
472 		 * a large number so that it is easy to assign smaller numbers
473 		 * to more preferred rules.
474 		 */
475 		new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst);
476 		if (best_matchlen < new_matchlen)
477 			REPLACE(14);
478 		if (new_matchlen < best_matchlen)
479 			NEXT(14);
480 
481 		/* Rule 15 is reserved. */
482 
483 		/*
484 		 * Last resort: just keep the current candidate.
485 		 * Or, do we need more rules?
486 		 */
487 		continue;
488 
489 	  replace:
490 		ia_best = ia;
491 		best_scope = (new_scope >= 0 ? new_scope :
492 			      in6_addrscope(&ia_best->ia_addr.sin6_addr));
493 		best_policy = (new_policy ? new_policy :
494 			       lookup_addrsel_policy(&ia_best->ia_addr));
495 		best_matchlen = (new_matchlen >= 0 ? new_matchlen :
496 				 in6_matchlen(&ia_best->ia_addr.sin6_addr,
497 					      &dst));
498 
499 	  next:
500 		continue;
501 
502 	  out:
503 		break;
504 	}
505 
506 	if ((ia = ia_best) == NULL) {
507 		IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
508 		IP6STAT_INC(ip6s_sources_none);
509 		return (EADDRNOTAVAIL);
510 	}
511 
512 	/*
513 	 * At this point at least one of the addresses belonged to the jail
514 	 * but it could still be, that we want to further restrict it, e.g.
515 	 * theoratically IN6_IS_ADDR_LOOPBACK.
516 	 * It must not be IN6_IS_ADDR_UNSPECIFIED anymore.
517 	 * prison_local_ip6() will fix an IN6_IS_ADDR_LOOPBACK but should
518 	 * let all others previously selected pass.
519 	 * Use tmp to not change ::1 on lo0 to the primary jail address.
520 	 */
521 	tmp = ia->ia_addr.sin6_addr;
522 	if (cred != NULL && prison_local_ip6(cred, &tmp, (inp != NULL &&
523 	    (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0) {
524 		IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
525 		IP6STAT_INC(ip6s_sources_none);
526 		return (EADDRNOTAVAIL);
527 	}
528 
529 	if (ifpp)
530 		*ifpp = ifp;
531 
532 	bcopy(&tmp, srcp, sizeof(*srcp));
533 	if (ia->ia_ifp == ifp)
534 		IP6STAT_INC(ip6s_sources_sameif[best_scope]);
535 	else
536 		IP6STAT_INC(ip6s_sources_otherif[best_scope]);
537 	if (dst_scope == best_scope)
538 		IP6STAT_INC(ip6s_sources_samescope[best_scope]);
539 	else
540 		IP6STAT_INC(ip6s_sources_otherscope[best_scope]);
541 	if (IFA6_IS_DEPRECATED(ia))
542 		IP6STAT_INC(ip6s_sources_deprecated[best_scope]);
543 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
544 	return (0);
545 }
546 
547 /*
548  * Select source address based on @inp, @dstsock and @opts.
549  * Stores selected address to @srcp. If @scope_ambiguous is set,
550  * embed scope from selected outgoing interface. If @hlim pointer
551  * is provided, stores calculated hop limit there.
552  * Returns 0 on success.
553  */
554 int
555 in6_selectsrc_socket(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
556     struct inpcb *inp, struct ucred *cred, int scope_ambiguous,
557     struct in6_addr *srcp, int *hlim)
558 {
559 	struct ifnet *retifp;
560 	uint32_t fibnum;
561 	int error;
562 
563 	fibnum = inp->inp_inc.inc_fibnum;
564 	retifp = NULL;
565 
566 	error = in6_selectsrc(fibnum, dstsock, opts, inp, cred, &retifp, srcp);
567 	if (error != 0)
568 		return (error);
569 
570 	if (hlim != NULL)
571 		*hlim = in6_selecthlim(inp, retifp);
572 
573 	if (retifp == NULL || scope_ambiguous == 0)
574 		return (0);
575 
576 	/*
577 	 * Application should provide a proper zone ID or the use of
578 	 * default zone IDs should be enabled.  Unfortunately, some
579 	 * applications do not behave as it should, so we need a
580 	 * workaround.  Even if an appropriate ID is not determined
581 	 * (when it's required), if we can determine the outgoing
582 	 * interface. determine the zone ID based on the interface.
583 	 */
584 	error = in6_setscope(&dstsock->sin6_addr, retifp, NULL);
585 
586 	return (error);
587 }
588 
589 /*
590  * Select source address based on @fibnum, @dst and @scopeid.
591  * Stores selected address to @srcp.
592  * Returns 0 on success.
593  *
594  * Used by non-socket based consumers (ND code mostly)
595  */
596 int
597 in6_selectsrc_addr(uint32_t fibnum, const struct in6_addr *dst,
598     uint32_t scopeid, struct ifnet *ifp, struct in6_addr *srcp,
599     int *hlim)
600 {
601 	struct ifnet *retifp;
602 	struct sockaddr_in6 dst_sa;
603 	int error;
604 
605 	retifp = ifp;
606 	bzero(&dst_sa, sizeof(dst_sa));
607 	dst_sa.sin6_family = AF_INET6;
608 	dst_sa.sin6_len = sizeof(dst_sa);
609 	dst_sa.sin6_addr = *dst;
610 	dst_sa.sin6_scope_id = scopeid;
611 	sa6_embedscope(&dst_sa, 0);
612 
613 	error = in6_selectsrc(fibnum, &dst_sa, NULL, NULL, NULL, &retifp, srcp);
614 	if (hlim != NULL)
615 		*hlim = in6_selecthlim(NULL, retifp);
616 
617 	return (error);
618 }
619 
620 /*
621  * clone - meaningful only for bsdi and freebsd
622  */
623 static int
624 selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
625     struct ip6_moptions *mopts, struct route_in6 *ro,
626     struct ifnet **retifp, struct nhop_object **retnh, int norouteok,
627     u_int fibnum, uint32_t flowid)
628 {
629 	int error = 0;
630 	struct ifnet *ifp = NULL;
631 	struct nhop_object *nh = NULL;
632 	struct sockaddr_in6 *sin6_next;
633 	struct in6_pktinfo *pi = NULL;
634 	struct in6_addr *dst = &dstsock->sin6_addr;
635 	uint32_t zoneid;
636 #if 0
637 	char ip6buf[INET6_ADDRSTRLEN];
638 
639 	if (dstsock->sin6_addr.s6_addr32[0] == 0 &&
640 	    dstsock->sin6_addr.s6_addr32[1] == 0 &&
641 	    !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) {
642 		printf("%s: strange destination %s\n", __func__,
643 		       ip6_sprintf(ip6buf, &dstsock->sin6_addr));
644 	} else {
645 		printf("%s: destination = %s%%%d\n", __func__,
646 		       ip6_sprintf(ip6buf, &dstsock->sin6_addr),
647 		       dstsock->sin6_scope_id); /* for debug */
648 	}
649 #endif
650 
651 	/* If the caller specify the outgoing interface explicitly, use it. */
652 	if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) {
653 		/* XXX boundary check is assumed to be already done. */
654 		ifp = ifnet_byindex(pi->ipi6_ifindex);
655 		if (ifp != NULL &&
656 		    (norouteok || retnh == NULL ||
657 		    IN6_IS_ADDR_MULTICAST(dst))) {
658 			/*
659 			 * we do not have to check or get the route for
660 			 * multicast.
661 			 */
662 			goto done;
663 		} else
664 			goto getroute;
665 	}
666 	/*
667 	 * If the destination address is a multicast address and the outgoing
668 	 * interface for the address is specified by the caller, use it.
669 	 */
670 	if (IN6_IS_ADDR_MULTICAST(dst) &&
671 	    mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) {
672 		goto done; /* we do not need a route for multicast. */
673 	}
674 	/*
675 	 * If destination address is LLA or link- or node-local multicast,
676 	 * use it's embedded scope zone id to determine outgoing interface.
677 	 */
678 	if (IN6_IS_ADDR_MC_LINKLOCAL(dst) ||
679 	    IN6_IS_ADDR_MC_NODELOCAL(dst)) {
680 		zoneid = ntohs(in6_getscope(dst));
681 		if (zoneid > 0) {
682 			ifp = in6_getlinkifnet(zoneid);
683 			goto done;
684 		}
685 	}
686 
687   getroute:
688 	/*
689 	 * If the next hop address for the packet is specified by the caller,
690 	 * use it as the gateway.
691 	 */
692 	if (opts && opts->ip6po_nexthop) {
693 		struct route_in6 *ron;
694 
695 		sin6_next = satosin6(opts->ip6po_nexthop);
696 		if (IN6_IS_ADDR_LINKLOCAL(&sin6_next->sin6_addr)) {
697 			/*
698 			 * Next hop is LLA, thus it should be neighbor.
699 			 * Determine outgoing interface by zone index.
700 			 */
701 			zoneid = ntohs(in6_getscope(&sin6_next->sin6_addr));
702 			if (zoneid > 0) {
703 				ifp = in6_getlinkifnet(zoneid);
704 				goto done;
705 			}
706 		}
707 		ron = &opts->ip6po_nextroute;
708 		/* Use a cached route if it exists and is valid. */
709 		if (ron->ro_nh != NULL && (
710 		    !NH_IS_VALID(ron->ro_nh) ||
711 		    ron->ro_dst.sin6_family != AF_INET6 ||
712 		    !IN6_ARE_ADDR_EQUAL(&ron->ro_dst.sin6_addr,
713 			&sin6_next->sin6_addr)))
714 			RO_NHFREE(ron);
715 		if (ron->ro_nh == NULL) {
716 			ron->ro_dst = *sin6_next;
717 			/*
718 			 * sin6_next is not link-local OR scopeid is 0,
719 			 * no need to clear scope
720 			 */
721 			ron->ro_nh = fib6_lookup(fibnum,
722 			    &sin6_next->sin6_addr, 0, NHR_REF, flowid);
723 		}
724 		/*
725 		 * The node identified by that address must be a
726 		 * neighbor of the sending host.
727 		 */
728 		if (ron->ro_nh == NULL ||
729 		    (ron->ro_nh->nh_flags & NHF_GATEWAY) != 0)
730 			error = EHOSTUNREACH;
731 		else {
732 			nh = ron->ro_nh;
733 			ifp = nh->nh_ifp;
734 		}
735 		goto done;
736 	}
737 
738 	/*
739 	 * Use a cached route if it exists and is valid, else try to allocate
740 	 * a new one.  Note that we should check the address family of the
741 	 * cached destination, in case of sharing the cache with IPv4.
742 	 */
743 	if (ro) {
744 		if (ro->ro_nh &&
745 		    (!NH_IS_VALID(ro->ro_nh) ||
746 		     ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 ||
747 		     !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr,
748 		     dst))) {
749 			RO_NHFREE(ro);
750 		}
751 		if (ro->ro_nh == (struct nhop_object *)NULL) {
752 			struct sockaddr_in6 *sa6;
753 
754 			/* No route yet, so try to acquire one */
755 			bzero(&ro->ro_dst, sizeof(struct sockaddr_in6));
756 			sa6 = (struct sockaddr_in6 *)&ro->ro_dst;
757 			*sa6 = *dstsock;
758 			sa6->sin6_scope_id = 0;
759 
760 			/*
761 			 * Currently dst has scopeid embedded iff it is LL.
762 			 * New routing API accepts scopeid as a separate argument.
763 			 * Convert dst before/after doing lookup
764 			 */
765 			uint32_t scopeid = 0;
766 			if (IN6_IS_SCOPE_LINKLOCAL(&sa6->sin6_addr)) {
767 				/* Unwrap in6_getscope() and in6_clearscope() */
768 				scopeid = ntohs(sa6->sin6_addr.s6_addr16[1]);
769 				sa6->sin6_addr.s6_addr16[1] = 0;
770 
771 			}
772 
773 			ro->ro_nh = fib6_lookup(fibnum,
774 			    &sa6->sin6_addr, scopeid, NHR_REF, flowid);
775 
776 			if (IN6_IS_SCOPE_LINKLOCAL(&sa6->sin6_addr))
777 				sa6->sin6_addr.s6_addr16[1] = htons(scopeid);
778 		}
779 
780 		/*
781 		 * do not care about the result if we have the nexthop
782 		 * explicitly specified.
783 		 */
784 		if (opts && opts->ip6po_nexthop)
785 			goto done;
786 
787 		if (ro->ro_nh)
788 			ifp = ro->ro_nh->nh_ifp;
789 		else
790 			error = EHOSTUNREACH;
791 		nh = ro->ro_nh;
792 
793 		/*
794 		 * Check if the outgoing interface conflicts with
795 		 * the interface specified by ipi6_ifindex (if specified).
796 		 * Note that loopback interface is always okay.
797 		 * (this may happen when we are sending a packet to one of
798 		 *  our own addresses.)
799 		 */
800 		if (ifp && opts && opts->ip6po_pktinfo &&
801 		    opts->ip6po_pktinfo->ipi6_ifindex) {
802 			if (!(ifp->if_flags & IFF_LOOPBACK) &&
803 			    ifp->if_index !=
804 			    opts->ip6po_pktinfo->ipi6_ifindex) {
805 				error = EHOSTUNREACH;
806 				goto done;
807 			}
808 		}
809 	}
810 
811   done:
812 	if (ifp == NULL && nh == NULL) {
813 		/*
814 		 * This can happen if the caller did not pass a cached route
815 		 * nor any other hints.  We treat this case an error.
816 		 */
817 		error = EHOSTUNREACH;
818 	}
819 	if (error == EHOSTUNREACH)
820 		IP6STAT_INC(ip6s_noroute);
821 
822 	if (retifp != NULL) {
823 		if (nh != NULL)
824 			*retifp = nh->nh_aifp;
825 		else
826 			*retifp = ifp;
827 	}
828 
829 	if (retnh != NULL)
830 		*retnh = nh;	/* nh may be NULL */
831 
832 	return (error);
833 }
834 
835 static int
836 in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
837     struct ip6_moptions *mopts, struct ifnet **retifp,
838     struct ifnet *oifp, u_int fibnum)
839 {
840 	int error;
841 	struct route_in6 sro;
842 	struct nhop_object *nh = NULL;
843 	uint16_t nh_flags;
844 
845 	KASSERT(retifp != NULL, ("%s: retifp is NULL", __func__));
846 
847 	bzero(&sro, sizeof(sro));
848 	nh_flags = 0;
849 
850 	error = selectroute(dstsock, opts, mopts, &sro, retifp, &nh, 1, fibnum, 0);
851 
852 	if (nh != NULL)
853 		nh_flags = nh->nh_flags;
854 	if (nh != NULL && nh == sro.ro_nh)
855 		NH_FREE(nh);
856 
857 	if (error != 0) {
858 		/* Help ND. See oifp comment in in6_selectsrc(). */
859 		if (oifp != NULL && fibnum == RT_DEFAULT_FIB) {
860 			*retifp = oifp;
861 			error = 0;
862 		}
863 		return (error);
864 	}
865 
866 	/*
867 	 * do not use a rejected or black hole route.
868 	 * XXX: this check should be done in the L2 output routine.
869 	 * However, if we skipped this check here, we'd see the following
870 	 * scenario:
871 	 * - install a rejected route for a scoped address prefix
872 	 *   (like fe80::/10)
873 	 * - send a packet to a destination that matches the scoped prefix,
874 	 *   with ambiguity about the scope zone.
875 	 * - pick the outgoing interface from the route, and disambiguate the
876 	 *   scope zone with the interface.
877 	 * - ip6_output() would try to get another route with the "new"
878 	 *   destination, which may be valid.
879 	 * - we'd see no error on output.
880 	 * Although this may not be very harmful, it should still be confusing.
881 	 * We thus reject the case here.
882 	 */
883 
884 	if (nh_flags & (NHF_REJECT | NHF_BLACKHOLE)) {
885 		error = (nh_flags & NHF_HOST ? EHOSTUNREACH : ENETUNREACH);
886 		return (error);
887 	}
888 
889 	return (0);
890 }
891 
892 /* Public wrapper function to selectroute(). */
893 int
894 in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
895     struct ip6_moptions *mopts, struct route_in6 *ro,
896     struct ifnet **retifp, struct nhop_object **retnh, u_int fibnum, uint32_t flowid)
897 {
898 
899 	return (selectroute(dstsock, opts, mopts, ro, retifp,
900 	    retnh, 0, fibnum, flowid));
901 }
902 
903 /*
904  * Default hop limit selection. The precedence is as follows:
905  * 1. Hoplimit value specified via ioctl.
906  * 2. (If the outgoing interface is detected) the current
907  *     hop limit of the interface specified by router advertisement.
908  * 3. The system default hoplimit.
909  */
910 int
911 in6_selecthlim(struct inpcb *inp, struct ifnet *ifp)
912 {
913 
914 	if (inp && inp->in6p_hops >= 0)
915 		return (inp->in6p_hops);
916 	else if (ifp)
917 		return (ND_IFINFO(ifp)->chlim);
918 	else if (inp && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
919 		struct nhop6_basic nh6;
920 		struct in6_addr dst;
921 		uint32_t fibnum, scopeid;
922 		int hlim;
923 
924 		fibnum = inp->inp_inc.inc_fibnum;
925 		in6_splitscope(&inp->in6p_faddr, &dst, &scopeid);
926 		if (fib6_lookup_nh_basic(fibnum, &dst, scopeid, 0, 0, &nh6)==0){
927 			hlim = ND_IFINFO(nh6.nh_ifp)->chlim;
928 			return (hlim);
929 		}
930 	}
931 	return (V_ip6_defhlim);
932 }
933 
934 /*
935  * XXX: this is borrowed from in6_pcbbind(). If possible, we should
936  * share this function by all *bsd*...
937  */
938 int
939 in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct ucred *cred)
940 {
941 	struct socket *so = inp->inp_socket;
942 	u_int16_t lport = 0;
943 	int error, lookupflags = 0;
944 #ifdef INVARIANTS
945 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
946 #endif
947 
948 	INP_WLOCK_ASSERT(inp);
949 	INP_HASH_WLOCK_ASSERT(pcbinfo);
950 
951 	error = prison_local_ip6(cred, laddr,
952 	    ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0));
953 	if (error)
954 		return(error);
955 
956 	/* XXX: this is redundant when called from in6_pcbbind */
957 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT|SO_REUSEPORT_LB)) == 0)
958 		lookupflags = INPLOOKUP_WILDCARD;
959 
960 	inp->inp_flags |= INP_ANONPORT;
961 
962 	error = in_pcb_lport(inp, NULL, &lport, cred, lookupflags);
963 	if (error != 0)
964 		return (error);
965 
966 	inp->inp_lport = lport;
967 	if (in_pcbinshash(inp) != 0) {
968 		inp->in6p_laddr = in6addr_any;
969 		inp->inp_lport = 0;
970 		return (EAGAIN);
971 	}
972 
973 	return (0);
974 }
975 
976 void
977 addrsel_policy_init(void)
978 {
979 
980 	init_policy_queue();
981 
982 	/* initialize the "last resort" policy */
983 	bzero(&V_defaultaddrpolicy, sizeof(V_defaultaddrpolicy));
984 	V_defaultaddrpolicy.label = ADDR_LABEL_NOTAPP;
985 
986 	if (!IS_DEFAULT_VNET(curvnet))
987 		return;
988 
989 	ADDRSEL_LOCK_INIT();
990 	ADDRSEL_SXLOCK_INIT();
991 }
992 
993 static struct in6_addrpolicy *
994 lookup_addrsel_policy(struct sockaddr_in6 *key)
995 {
996 	struct in6_addrpolicy *match = NULL;
997 
998 	ADDRSEL_LOCK();
999 	match = match_addrsel_policy(key);
1000 
1001 	if (match == NULL)
1002 		match = &V_defaultaddrpolicy;
1003 	else
1004 		match->use++;
1005 	ADDRSEL_UNLOCK();
1006 
1007 	return (match);
1008 }
1009 
1010 /*
1011  * Subroutines to manage the address selection policy table via sysctl.
1012  */
1013 struct walkarg {
1014 	struct sysctl_req *w_req;
1015 };
1016 
1017 static int in6_src_sysctl(SYSCTL_HANDLER_ARGS);
1018 SYSCTL_DECL(_net_inet6_ip6);
1019 static SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy,
1020     CTLFLAG_RD | CTLFLAG_MPSAFE, in6_src_sysctl,
1021     "");
1022 
1023 static int
1024 in6_src_sysctl(SYSCTL_HANDLER_ARGS)
1025 {
1026 	struct walkarg w;
1027 
1028 	if (req->newptr)
1029 		return EPERM;
1030 
1031 	bzero(&w, sizeof(w));
1032 	w.w_req = req;
1033 
1034 	return (walk_addrsel_policy(dump_addrsel_policyent, &w));
1035 }
1036 
1037 int
1038 in6_src_ioctl(u_long cmd, caddr_t data)
1039 {
1040 	struct in6_addrpolicy ent0;
1041 
1042 	if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY)
1043 		return (EOPNOTSUPP); /* check for safety */
1044 
1045 	ent0 = *(struct in6_addrpolicy *)data;
1046 
1047 	if (ent0.label == ADDR_LABEL_NOTAPP)
1048 		return (EINVAL);
1049 	/* check if the prefix mask is consecutive. */
1050 	if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0)
1051 		return (EINVAL);
1052 	/* clear trailing garbages (if any) of the prefix address. */
1053 	IN6_MASK_ADDR(&ent0.addr.sin6_addr, &ent0.addrmask.sin6_addr);
1054 	ent0.use = 0;
1055 
1056 	switch (cmd) {
1057 	case SIOCAADDRCTL_POLICY:
1058 		return (add_addrsel_policyent(&ent0));
1059 	case SIOCDADDRCTL_POLICY:
1060 		return (delete_addrsel_policyent(&ent0));
1061 	}
1062 
1063 	return (0);		/* XXX: compromise compilers */
1064 }
1065 
1066 /*
1067  * The followings are implementation of the policy table using a
1068  * simple tail queue.
1069  * XXX such details should be hidden.
1070  * XXX implementation using binary tree should be more efficient.
1071  */
1072 struct addrsel_policyent {
1073 	TAILQ_ENTRY(addrsel_policyent) ape_entry;
1074 	struct in6_addrpolicy ape_policy;
1075 };
1076 
1077 TAILQ_HEAD(addrsel_policyhead, addrsel_policyent);
1078 
1079 VNET_DEFINE_STATIC(struct addrsel_policyhead, addrsel_policytab);
1080 #define	V_addrsel_policytab		VNET(addrsel_policytab)
1081 
1082 static void
1083 init_policy_queue(void)
1084 {
1085 
1086 	TAILQ_INIT(&V_addrsel_policytab);
1087 }
1088 
1089 static int
1090 add_addrsel_policyent(struct in6_addrpolicy *newpolicy)
1091 {
1092 	struct addrsel_policyent *new, *pol;
1093 
1094 	new = malloc(sizeof(*new), M_IFADDR,
1095 	       M_WAITOK);
1096 	ADDRSEL_XLOCK();
1097 	ADDRSEL_LOCK();
1098 
1099 	/* duplication check */
1100 	TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1101 		if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr,
1102 				       &pol->ape_policy.addr.sin6_addr) &&
1103 		    IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr,
1104 				       &pol->ape_policy.addrmask.sin6_addr)) {
1105 			ADDRSEL_UNLOCK();
1106 			ADDRSEL_XUNLOCK();
1107 			free(new, M_IFADDR);
1108 			return (EEXIST);	/* or override it? */
1109 		}
1110 	}
1111 
1112 	bzero(new, sizeof(*new));
1113 
1114 	/* XXX: should validate entry */
1115 	new->ape_policy = *newpolicy;
1116 
1117 	TAILQ_INSERT_TAIL(&V_addrsel_policytab, new, ape_entry);
1118 	ADDRSEL_UNLOCK();
1119 	ADDRSEL_XUNLOCK();
1120 
1121 	return (0);
1122 }
1123 
1124 static int
1125 delete_addrsel_policyent(struct in6_addrpolicy *key)
1126 {
1127 	struct addrsel_policyent *pol;
1128 
1129 	ADDRSEL_XLOCK();
1130 	ADDRSEL_LOCK();
1131 
1132 	/* search for the entry in the table */
1133 	TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1134 		if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr,
1135 		    &pol->ape_policy.addr.sin6_addr) &&
1136 		    IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr,
1137 		    &pol->ape_policy.addrmask.sin6_addr)) {
1138 			break;
1139 		}
1140 	}
1141 	if (pol == NULL) {
1142 		ADDRSEL_UNLOCK();
1143 		ADDRSEL_XUNLOCK();
1144 		return (ESRCH);
1145 	}
1146 
1147 	TAILQ_REMOVE(&V_addrsel_policytab, pol, ape_entry);
1148 	ADDRSEL_UNLOCK();
1149 	ADDRSEL_XUNLOCK();
1150 	free(pol, M_IFADDR);
1151 
1152 	return (0);
1153 }
1154 
1155 static int
1156 walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *), void *w)
1157 {
1158 	struct addrsel_policyent *pol;
1159 	int error = 0;
1160 
1161 	ADDRSEL_SLOCK();
1162 	TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1163 		if ((error = (*callback)(&pol->ape_policy, w)) != 0) {
1164 			ADDRSEL_SUNLOCK();
1165 			return (error);
1166 		}
1167 	}
1168 	ADDRSEL_SUNLOCK();
1169 	return (error);
1170 }
1171 
1172 static int
1173 dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg)
1174 {
1175 	int error = 0;
1176 	struct walkarg *w = arg;
1177 
1178 	error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol));
1179 
1180 	return (error);
1181 }
1182 
1183 static struct in6_addrpolicy *
1184 match_addrsel_policy(struct sockaddr_in6 *key)
1185 {
1186 	struct addrsel_policyent *pent;
1187 	struct in6_addrpolicy *bestpol = NULL, *pol;
1188 	int matchlen, bestmatchlen = -1;
1189 	u_char *mp, *ep, *k, *p, m;
1190 
1191 	TAILQ_FOREACH(pent, &V_addrsel_policytab, ape_entry) {
1192 		matchlen = 0;
1193 
1194 		pol = &pent->ape_policy;
1195 		mp = (u_char *)&pol->addrmask.sin6_addr;
1196 		ep = mp + 16;	/* XXX: scope field? */
1197 		k = (u_char *)&key->sin6_addr;
1198 		p = (u_char *)&pol->addr.sin6_addr;
1199 		for (; mp < ep && *mp; mp++, k++, p++) {
1200 			m = *mp;
1201 			if ((*k & m) != *p)
1202 				goto next; /* not match */
1203 			if (m == 0xff) /* short cut for a typical case */
1204 				matchlen += 8;
1205 			else {
1206 				while (m >= 0x80) {
1207 					matchlen++;
1208 					m <<= 1;
1209 				}
1210 			}
1211 		}
1212 
1213 		/* matched.  check if this is better than the current best. */
1214 		if (bestpol == NULL ||
1215 		    matchlen > bestmatchlen) {
1216 			bestpol = pol;
1217 			bestmatchlen = matchlen;
1218 		}
1219 
1220 	  next:
1221 		continue;
1222 	}
1223 
1224 	return (bestpol);
1225 }
1226