xref: /freebsd/sys/netinet6/in6_src.c (revision 410556f1f10fd35b350102725fd8504c3cb0afc8)
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 <sys/param.h>
71 #include <sys/systm.h>
72 #include <sys/lock.h>
73 #include <sys/malloc.h>
74 #include <sys/mbuf.h>
75 #include <sys/priv.h>
76 #include <sys/protosw.h>
77 #include <sys/socket.h>
78 #include <sys/socketvar.h>
79 #include <sys/sockio.h>
80 #include <sys/sysctl.h>
81 #include <sys/errno.h>
82 #include <sys/time.h>
83 #include <sys/jail.h>
84 #include <sys/kernel.h>
85 #include <sys/rmlock.h>
86 #include <sys/sx.h>
87 
88 #include <net/if.h>
89 #include <net/if_var.h>
90 #include <net/if_dl.h>
91 #include <net/route.h>
92 #include <net/route/nhop.h>
93 #include <net/if_llatbl.h>
94 
95 #include <netinet/in.h>
96 #include <netinet/in_var.h>
97 #include <netinet/in_systm.h>
98 #include <netinet/ip.h>
99 #include <netinet/in_pcb.h>
100 #include <netinet/ip_var.h>
101 #include <netinet/udp.h>
102 #include <netinet/udp_var.h>
103 
104 #include <netinet6/in6_var.h>
105 #include <netinet/ip6.h>
106 #include <netinet6/in6_fib.h>
107 #include <netinet6/in6_pcb.h>
108 #include <netinet6/ip6_var.h>
109 #include <netinet6/scope6_var.h>
110 #include <netinet6/nd6.h>
111 
112 static struct mtx addrsel_lock;
113 #define	ADDRSEL_LOCK_INIT()	mtx_init(&addrsel_lock, "addrsel_lock", NULL, MTX_DEF)
114 #define	ADDRSEL_LOCK()		mtx_lock(&addrsel_lock)
115 #define	ADDRSEL_UNLOCK()	mtx_unlock(&addrsel_lock)
116 #define	ADDRSEL_LOCK_ASSERT()	mtx_assert(&addrsel_lock, MA_OWNED)
117 
118 static struct sx addrsel_sxlock;
119 #define	ADDRSEL_SXLOCK_INIT()	sx_init(&addrsel_sxlock, "addrsel_sxlock")
120 #define	ADDRSEL_SLOCK()		sx_slock(&addrsel_sxlock)
121 #define	ADDRSEL_SUNLOCK()	sx_sunlock(&addrsel_sxlock)
122 #define	ADDRSEL_XLOCK()		sx_xlock(&addrsel_sxlock)
123 #define	ADDRSEL_XUNLOCK()	sx_xunlock(&addrsel_sxlock)
124 
125 #define ADDR_LABEL_NOTAPP (-1)
126 VNET_DEFINE_STATIC(struct in6_addrpolicy, defaultaddrpolicy);
127 #define	V_defaultaddrpolicy		VNET(defaultaddrpolicy)
128 
129 VNET_DEFINE(int, ip6_prefer_tempaddr) = 0;
130 
131 static int selectroute(struct sockaddr_in6 *, struct ip6_pktopts *,
132 	struct ip6_moptions *, struct route_in6 *, struct ifnet **,
133 	struct nhop_object **, int, u_int, uint32_t);
134 static int in6_selectif(struct sockaddr_in6 *, struct ip6_pktopts *,
135 	struct ip6_moptions *, struct ifnet **,
136 	struct ifnet *, u_int);
137 static int in6_selectsrc(uint32_t, struct sockaddr_in6 *,
138 	struct ip6_pktopts *, struct inpcb *, struct ucred *,
139 	struct ifnet **, struct in6_addr *);
140 
141 static struct in6_addrpolicy *lookup_addrsel_policy(struct sockaddr_in6 *);
142 
143 static void init_policy_queue(void);
144 static int add_addrsel_policyent(struct in6_addrpolicy *);
145 static int delete_addrsel_policyent(struct in6_addrpolicy *);
146 static int walk_addrsel_policy(int (*)(struct in6_addrpolicy *, void *),
147 	void *);
148 static int dump_addrsel_policyent(struct in6_addrpolicy *, void *);
149 static struct in6_addrpolicy *match_addrsel_policy(struct sockaddr_in6 *);
150 
151 /*
152  * Return an IPv6 address, which is the most appropriate for a given
153  * destination and user specified options.
154  * If necessary, this function lookups the routing table and returns
155  * an entry to the caller for later use.
156  */
157 #define REPLACE(r) do {\
158 	IP6STAT_INC(ip6s_sources_rule[(r)]); \
159 	/* { \
160 	char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
161 	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)); \
162 	} */ \
163 	goto replace; \
164 } while(0)
165 #define NEXT(r) do {\
166 	/* { \
167 	char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
168 	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)); \
169 	} */ \
170 	goto next;		/* XXX: we can't use 'continue' here */ \
171 } while(0)
172 #define BREAK(r) do { \
173 	IP6STAT_INC(ip6s_sources_rule[(r)]); \
174 	goto out;		/* XXX: we can't use 'break' here */ \
175 } while(0)
176 
177 static int
178 in6_selectsrc(uint32_t fibnum, struct sockaddr_in6 *dstsock,
179     struct ip6_pktopts *opts, struct inpcb *inp, struct ucred *cred,
180     struct ifnet **ifpp, struct in6_addr *srcp)
181 {
182 	struct rm_priotracker in6_ifa_tracker;
183 	struct in6_addr dst, tmp;
184 	struct ifnet *ifp = NULL, *oifp = NULL;
185 	struct in6_ifaddr *ia = NULL, *ia_best = NULL;
186 	struct in6_pktinfo *pi = NULL;
187 	int dst_scope = -1, best_scope = -1, best_matchlen = -1;
188 	struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL;
189 	u_int32_t odstzone;
190 	int prefer_tempaddr;
191 	int error;
192 	struct ip6_moptions *mopts;
193 
194 	NET_EPOCH_ASSERT();
195 	KASSERT(srcp != NULL, ("%s: srcp is NULL", __func__));
196 
197 	dst = dstsock->sin6_addr; /* make a copy for local operation */
198 	if (ifpp) {
199 		/*
200 		 * Save a possibly passed in ifp for in6_selectsrc. Only
201 		 * neighbor discovery code should use this feature, where
202 		 * we may know the interface but not the FIB number holding
203 		 * the connected subnet in case someone deleted it from the
204 		 * default FIB and we need to check the interface.
205 		 */
206 		if (*ifpp != NULL)
207 			oifp = *ifpp;
208 		*ifpp = NULL;
209 	}
210 
211 	if (inp != NULL) {
212 		INP_LOCK_ASSERT(inp);
213 		mopts = inp->in6p_moptions;
214 	} else {
215 		mopts = NULL;
216 	}
217 
218 	/*
219 	 * If the source address is explicitly specified by the caller,
220 	 * check if the requested source address is indeed a unicast address
221 	 * assigned to the node, and can be used as the packet's source
222 	 * address.  If everything is okay, use the address as source.
223 	 */
224 	if (opts && (pi = opts->ip6po_pktinfo) &&
225 	    !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) {
226 		/* get the outgoing interface */
227 		if ((error = in6_selectif(dstsock, opts, mopts, &ifp, oifp,
228 		    fibnum))
229 		    != 0)
230 			return (error);
231 
232 		/*
233 		 * determine the appropriate zone id of the source based on
234 		 * the zone of the destination and the outgoing interface.
235 		 * If the specified address is ambiguous wrt the scope zone,
236 		 * the interface must be specified; otherwise, ifa_ifwithaddr()
237 		 * will fail matching the address.
238 		 */
239 		tmp = pi->ipi6_addr;
240 		if (ifp) {
241 			error = in6_setscope(&tmp, ifp, &odstzone);
242 			if (error)
243 				return (error);
244 		}
245 		if (cred != NULL && (error = prison_local_ip6(cred,
246 		    &tmp, (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0)
247 			return (error);
248 
249 		/*
250 		 * If IPV6_BINDANY socket option is set, we allow to specify
251 		 * non local addresses as source address in IPV6_PKTINFO
252 		 * ancillary data.
253 		 */
254 		if ((inp->inp_flags & INP_BINDANY) == 0) {
255 			ia = in6ifa_ifwithaddr(&tmp, 0 /* XXX */, false);
256 			if (ia == NULL || (ia->ia6_flags & (IN6_IFF_ANYCAST |
257 			    IN6_IFF_NOTREADY)))
258 				return (EADDRNOTAVAIL);
259 			bcopy(&ia->ia_addr.sin6_addr, srcp, sizeof(*srcp));
260 		} else
261 			bcopy(&tmp, srcp, sizeof(*srcp));
262 		pi->ipi6_addr = tmp; /* XXX: this overrides pi */
263 		if (ifpp)
264 			*ifpp = ifp;
265 		return (0);
266 	}
267 
268 	/*
269 	 * Otherwise, if the socket has already bound the source, just use it.
270 	 */
271 	if (inp != NULL && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
272 		if (cred != NULL &&
273 		    (error = prison_local_ip6(cred, &inp->in6p_laddr,
274 		    ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
275 			return (error);
276 		bcopy(&inp->in6p_laddr, srcp, sizeof(*srcp));
277 		return (0);
278 	}
279 
280 	/*
281 	 * Bypass source address selection and use the primary jail IP
282 	 * if requested.
283 	 */
284 	if (cred != NULL && !prison_saddrsel_ip6(cred, srcp))
285 		return (0);
286 
287 	/*
288 	 * If the address is not specified, choose the best one based on
289 	 * the outgoing interface and the destination address.
290 	 */
291 	/* get the outgoing interface */
292 	if ((error = in6_selectif(dstsock, opts, mopts, &ifp, oifp,
293 	    (inp != NULL) ? inp->inp_inc.inc_fibnum : fibnum)) != 0)
294 		return (error);
295 
296 #ifdef DIAGNOSTIC
297 	if (ifp == NULL)	/* this should not happen */
298 		panic("in6_selectsrc: NULL ifp");
299 #endif
300 	error = in6_setscope(&dst, ifp, &odstzone);
301 	if (error)
302 		return (error);
303 
304 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
305 	CK_STAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
306 		int new_scope = -1, new_matchlen = -1;
307 		struct in6_addrpolicy *new_policy = NULL;
308 		u_int32_t srczone, osrczone, dstzone;
309 		struct in6_addr src;
310 		struct ifnet *ifp1 = ia->ia_ifp;
311 
312 		/*
313 		 * We'll never take an address that breaks the scope zone
314 		 * of the destination.  We also skip an address if its zone
315 		 * does not contain the outgoing interface.
316 		 * XXX: we should probably use sin6_scope_id here.
317 		 */
318 		if (in6_setscope(&dst, ifp1, &dstzone) ||
319 		    odstzone != dstzone) {
320 			continue;
321 		}
322 		src = ia->ia_addr.sin6_addr;
323 		if (in6_setscope(&src, ifp, &osrczone) ||
324 		    in6_setscope(&src, ifp1, &srczone) ||
325 		    osrczone != srczone) {
326 			continue;
327 		}
328 
329 		/* avoid unusable addresses */
330 		if ((ia->ia6_flags &
331 		     (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) {
332 				continue;
333 		}
334 		if (!V_ip6_use_deprecated && IFA6_IS_DEPRECATED(ia))
335 			continue;
336 
337 		/* If jailed only take addresses of the jail into account. */
338 		if (cred != NULL &&
339 		    prison_check_ip6(cred, &ia->ia_addr.sin6_addr) != 0)
340 			continue;
341 
342 		/* Rule 1: Prefer same address */
343 		if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) {
344 			ia_best = ia;
345 			BREAK(1); /* there should be no better candidate */
346 		}
347 
348 		if (ia_best == NULL)
349 			REPLACE(0);
350 
351 		/* Rule 2: Prefer appropriate scope */
352 		if (dst_scope < 0)
353 			dst_scope = in6_addrscope(&dst);
354 		new_scope = in6_addrscope(&ia->ia_addr.sin6_addr);
355 		if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) {
356 			if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0)
357 				REPLACE(2);
358 			NEXT(2);
359 		} else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) {
360 			if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0)
361 				NEXT(2);
362 			REPLACE(2);
363 		}
364 
365 		/*
366 		 * Rule 3: Avoid deprecated addresses.  Note that the case of
367 		 * !ip6_use_deprecated is already rejected above.
368 		 */
369 		if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia))
370 			NEXT(3);
371 		if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia))
372 			REPLACE(3);
373 
374 		/* Rule 4: Prefer home addresses */
375 		/*
376 		 * XXX: This is a TODO.  We should probably merge the MIP6
377 		 * case above.
378 		 */
379 
380 		/* Rule 5: Prefer outgoing interface */
381 		if (!(ND_IFINFO(ifp)->flags & ND6_IFF_NO_PREFER_IFACE)) {
382 			if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp)
383 				NEXT(5);
384 			if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp)
385 				REPLACE(5);
386 		}
387 
388 		/*
389 		 * Rule 6: Prefer matching label
390 		 * Note that best_policy should be non-NULL here.
391 		 */
392 		if (dst_policy == NULL)
393 			dst_policy = lookup_addrsel_policy(dstsock);
394 		if (dst_policy->label != ADDR_LABEL_NOTAPP) {
395 			new_policy = lookup_addrsel_policy(&ia->ia_addr);
396 			if (dst_policy->label == best_policy->label &&
397 			    dst_policy->label != new_policy->label)
398 				NEXT(6);
399 			if (dst_policy->label != best_policy->label &&
400 			    dst_policy->label == new_policy->label)
401 				REPLACE(6);
402 		}
403 
404 		/*
405 		 * Rule 7: Prefer public addresses.
406 		 * We allow users to reverse the logic by configuring
407 		 * a sysctl variable, so that privacy conscious users can
408 		 * always prefer temporary addresses.
409 		 */
410 		if (opts == NULL ||
411 		    opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) {
412 			prefer_tempaddr = V_ip6_prefer_tempaddr;
413 		} else if (opts->ip6po_prefer_tempaddr ==
414 		    IP6PO_TEMPADDR_NOTPREFER) {
415 			prefer_tempaddr = 0;
416 		} else
417 			prefer_tempaddr = 1;
418 		if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
419 		    (ia->ia6_flags & IN6_IFF_TEMPORARY)) {
420 			if (prefer_tempaddr)
421 				REPLACE(7);
422 			else
423 				NEXT(7);
424 		}
425 		if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
426 		    !(ia->ia6_flags & IN6_IFF_TEMPORARY)) {
427 			if (prefer_tempaddr)
428 				NEXT(7);
429 			else
430 				REPLACE(7);
431 		}
432 
433 		/*
434 		 * Rule 8: prefer addresses on alive interfaces.
435 		 * This is a KAME specific rule.
436 		 */
437 		if ((ia_best->ia_ifp->if_flags & IFF_UP) &&
438 		    !(ia->ia_ifp->if_flags & IFF_UP))
439 			NEXT(8);
440 		if (!(ia_best->ia_ifp->if_flags & IFF_UP) &&
441 		    (ia->ia_ifp->if_flags & IFF_UP))
442 			REPLACE(8);
443 
444 		/*
445 		 * Rule 9: prefer address with better virtual status.
446 		 */
447 		if (ifa_preferred(&ia_best->ia_ifa, &ia->ia_ifa))
448 			REPLACE(9);
449 		if (ifa_preferred(&ia->ia_ifa, &ia_best->ia_ifa))
450 			NEXT(9);
451 
452 		/*
453 		 * Rule 10: prefer address with `prefer_source' flag.
454 		 */
455 		if ((ia_best->ia6_flags & IN6_IFF_PREFER_SOURCE) == 0 &&
456 		    (ia->ia6_flags & IN6_IFF_PREFER_SOURCE) != 0)
457 			REPLACE(10);
458 		if ((ia_best->ia6_flags & IN6_IFF_PREFER_SOURCE) != 0 &&
459 		    (ia->ia6_flags & IN6_IFF_PREFER_SOURCE) == 0)
460 			NEXT(10);
461 
462 		/*
463 		 * Rule 14: Use longest matching prefix.
464 		 * Note: in the address selection draft, this rule is
465 		 * documented as "Rule 8".  However, since it is also
466 		 * documented that this rule can be overridden, we assign
467 		 * a large number so that it is easy to assign smaller numbers
468 		 * to more preferred rules.
469 		 */
470 		new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst);
471 		if (best_matchlen < new_matchlen)
472 			REPLACE(14);
473 		if (new_matchlen < best_matchlen)
474 			NEXT(14);
475 
476 		/* Rule 15 is reserved. */
477 
478 		/*
479 		 * Last resort: just keep the current candidate.
480 		 * Or, do we need more rules?
481 		 */
482 		continue;
483 
484 	  replace:
485 		ia_best = ia;
486 		best_scope = (new_scope >= 0 ? new_scope :
487 			      in6_addrscope(&ia_best->ia_addr.sin6_addr));
488 		best_policy = (new_policy ? new_policy :
489 			       lookup_addrsel_policy(&ia_best->ia_addr));
490 		best_matchlen = (new_matchlen >= 0 ? new_matchlen :
491 				 in6_matchlen(&ia_best->ia_addr.sin6_addr,
492 					      &dst));
493 
494 	  next:
495 		continue;
496 
497 	  out:
498 		break;
499 	}
500 
501 	if ((ia = ia_best) == NULL) {
502 		IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
503 		IP6STAT_INC(ip6s_sources_none);
504 		return (EADDRNOTAVAIL);
505 	}
506 
507 	/*
508 	 * At this point at least one of the addresses belonged to the jail
509 	 * but it could still be, that we want to further restrict it, e.g.
510 	 * theoratically IN6_IS_ADDR_LOOPBACK.
511 	 * It must not be IN6_IS_ADDR_UNSPECIFIED anymore.
512 	 * prison_local_ip6() will fix an IN6_IS_ADDR_LOOPBACK but should
513 	 * let all others previously selected pass.
514 	 * Use tmp to not change ::1 on lo0 to the primary jail address.
515 	 */
516 	tmp = ia->ia_addr.sin6_addr;
517 	if (cred != NULL && prison_local_ip6(cred, &tmp, (inp != NULL &&
518 	    (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0) {
519 		IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
520 		IP6STAT_INC(ip6s_sources_none);
521 		return (EADDRNOTAVAIL);
522 	}
523 
524 	if (ifpp)
525 		*ifpp = ifp;
526 
527 	bcopy(&tmp, srcp, sizeof(*srcp));
528 	if (ia->ia_ifp == ifp)
529 		IP6STAT_INC(ip6s_sources_sameif[best_scope]);
530 	else
531 		IP6STAT_INC(ip6s_sources_otherif[best_scope]);
532 	if (dst_scope == best_scope)
533 		IP6STAT_INC(ip6s_sources_samescope[best_scope]);
534 	else
535 		IP6STAT_INC(ip6s_sources_otherscope[best_scope]);
536 	if (IFA6_IS_DEPRECATED(ia))
537 		IP6STAT_INC(ip6s_sources_deprecated[best_scope]);
538 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
539 	return (0);
540 }
541 
542 /*
543  * Select source address based on @inp, @dstsock and @opts.
544  * Stores selected address to @srcp. If @scope_ambiguous is set,
545  * embed scope from selected outgoing interface. If @hlim pointer
546  * is provided, stores calculated hop limit there.
547  * Returns 0 on success.
548  */
549 int
550 in6_selectsrc_socket(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
551     struct inpcb *inp, struct ucred *cred, int scope_ambiguous,
552     struct in6_addr *srcp, int *hlim)
553 {
554 	struct ifnet *retifp;
555 	uint32_t fibnum;
556 	int error;
557 
558 	fibnum = inp->inp_inc.inc_fibnum;
559 	retifp = NULL;
560 
561 	error = in6_selectsrc(fibnum, dstsock, opts, inp, cred, &retifp, srcp);
562 	if (error != 0)
563 		return (error);
564 
565 	if (hlim != NULL)
566 		*hlim = in6_selecthlim(inp, retifp);
567 
568 	if (retifp == NULL || scope_ambiguous == 0)
569 		return (0);
570 
571 	/*
572 	 * Application should provide a proper zone ID or the use of
573 	 * default zone IDs should be enabled.  Unfortunately, some
574 	 * applications do not behave as it should, so we need a
575 	 * workaround.  Even if an appropriate ID is not determined
576 	 * (when it's required), if we can determine the outgoing
577 	 * interface. determine the zone ID based on the interface.
578 	 */
579 	error = in6_setscope(&dstsock->sin6_addr, retifp, NULL);
580 
581 	return (error);
582 }
583 
584 /*
585  * Select source address based on @fibnum, @dst and @scopeid.
586  * Stores selected address to @srcp.
587  * Returns 0 on success.
588  *
589  * Used by non-socket based consumers (ND code mostly)
590  */
591 int
592 in6_selectsrc_addr(uint32_t fibnum, const struct in6_addr *dst,
593     uint32_t scopeid, struct ifnet *ifp, struct in6_addr *srcp,
594     int *hlim)
595 {
596 	struct ifnet *retifp;
597 	struct sockaddr_in6 dst_sa;
598 	int error;
599 
600 	retifp = ifp;
601 	bzero(&dst_sa, sizeof(dst_sa));
602 	dst_sa.sin6_family = AF_INET6;
603 	dst_sa.sin6_len = sizeof(dst_sa);
604 	dst_sa.sin6_addr = *dst;
605 	dst_sa.sin6_scope_id = scopeid;
606 	sa6_embedscope(&dst_sa, 0);
607 
608 	error = in6_selectsrc(fibnum, &dst_sa, NULL, NULL, NULL, &retifp, srcp);
609 	if (hlim != NULL)
610 		*hlim = in6_selecthlim(NULL, retifp);
611 
612 	return (error);
613 }
614 
615 /*
616  * clone - meaningful only for bsdi and freebsd
617  */
618 static int
619 selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
620     struct ip6_moptions *mopts, struct route_in6 *ro,
621     struct ifnet **retifp, struct nhop_object **retnh, int norouteok,
622     u_int fibnum, uint32_t flowid)
623 {
624 	int error = 0;
625 	struct ifnet *ifp = NULL;
626 	struct nhop_object *nh = NULL;
627 	struct sockaddr_in6 *sin6_next;
628 	struct in6_pktinfo *pi = NULL;
629 	struct in6_addr *dst = &dstsock->sin6_addr;
630 	uint32_t zoneid;
631 #if 0
632 	char ip6buf[INET6_ADDRSTRLEN];
633 
634 	if (dstsock->sin6_addr.s6_addr32[0] == 0 &&
635 	    dstsock->sin6_addr.s6_addr32[1] == 0 &&
636 	    !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) {
637 		printf("%s: strange destination %s\n", __func__,
638 		       ip6_sprintf(ip6buf, &dstsock->sin6_addr));
639 	} else {
640 		printf("%s: destination = %s%%%d\n", __func__,
641 		       ip6_sprintf(ip6buf, &dstsock->sin6_addr),
642 		       dstsock->sin6_scope_id); /* for debug */
643 	}
644 #endif
645 
646 	/* If the caller specify the outgoing interface explicitly, use it. */
647 	if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) {
648 		/* XXX boundary check is assumed to be already done. */
649 		ifp = ifnet_byindex(pi->ipi6_ifindex);
650 		if (ifp != NULL &&
651 		    (norouteok || retnh == NULL ||
652 		    IN6_IS_ADDR_MULTICAST(dst))) {
653 			/*
654 			 * we do not have to check or get the route for
655 			 * multicast.
656 			 */
657 			goto done;
658 		} else
659 			goto getroute;
660 	}
661 	/*
662 	 * If the destination address is a multicast address and the outgoing
663 	 * interface for the address is specified by the caller, use it.
664 	 */
665 	if (IN6_IS_ADDR_MULTICAST(dst) &&
666 	    mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) {
667 		goto done; /* we do not need a route for multicast. */
668 	}
669 	/*
670 	 * If destination address is LLA or link- or node-local multicast,
671 	 * use it's embedded scope zone id to determine outgoing interface.
672 	 */
673 	if (IN6_IS_ADDR_MC_LINKLOCAL(dst) ||
674 	    IN6_IS_ADDR_MC_NODELOCAL(dst)) {
675 		zoneid = ntohs(in6_getscope(dst));
676 		if (zoneid > 0) {
677 			ifp = in6_getlinkifnet(zoneid);
678 			goto done;
679 		}
680 	}
681 
682   getroute:
683 	/*
684 	 * If the next hop address for the packet is specified by the caller,
685 	 * use it as the gateway.
686 	 */
687 	if (opts && opts->ip6po_nexthop) {
688 		struct route_in6 *ron;
689 
690 		sin6_next = satosin6(opts->ip6po_nexthop);
691 		if (IN6_IS_ADDR_LINKLOCAL(&sin6_next->sin6_addr)) {
692 			/*
693 			 * Next hop is LLA, thus it should be neighbor.
694 			 * Determine outgoing interface by zone index.
695 			 */
696 			zoneid = ntohs(in6_getscope(&sin6_next->sin6_addr));
697 			if (zoneid > 0) {
698 				ifp = in6_getlinkifnet(zoneid);
699 				goto done;
700 			}
701 		}
702 		ron = &opts->ip6po_nextroute;
703 		/* Use a cached route if it exists and is valid. */
704 		if (ron->ro_nh != NULL && (
705 		    !NH_IS_VALID(ron->ro_nh) ||
706 		    ron->ro_dst.sin6_family != AF_INET6 ||
707 		    !IN6_ARE_ADDR_EQUAL(&ron->ro_dst.sin6_addr,
708 			&sin6_next->sin6_addr)))
709 			RO_NHFREE(ron);
710 		if (ron->ro_nh == NULL) {
711 			ron->ro_dst = *sin6_next;
712 			/*
713 			 * sin6_next is not link-local OR scopeid is 0,
714 			 * no need to clear scope
715 			 */
716 			ron->ro_nh = fib6_lookup(fibnum,
717 			    &sin6_next->sin6_addr, 0, NHR_REF, flowid);
718 		}
719 		/*
720 		 * The node identified by that address must be a
721 		 * neighbor of the sending host.
722 		 */
723 		if (ron->ro_nh == NULL ||
724 		    (ron->ro_nh->nh_flags & NHF_GATEWAY) != 0)
725 			error = EHOSTUNREACH;
726 		else {
727 			nh = ron->ro_nh;
728 			ifp = nh->nh_ifp;
729 		}
730 		goto done;
731 	}
732 
733 	/*
734 	 * Use a cached route if it exists and is valid, else try to allocate
735 	 * a new one.  Note that we should check the address family of the
736 	 * cached destination, in case of sharing the cache with IPv4.
737 	 */
738 	if (ro) {
739 		if (ro->ro_nh &&
740 		    (!NH_IS_VALID(ro->ro_nh) ||
741 		     ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 ||
742 		     !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr,
743 		     dst))) {
744 			RO_NHFREE(ro);
745 		}
746 		if (ro->ro_nh == (struct nhop_object *)NULL) {
747 			struct sockaddr_in6 *sa6;
748 
749 			/* No route yet, so try to acquire one */
750 			bzero(&ro->ro_dst, sizeof(struct sockaddr_in6));
751 			sa6 = (struct sockaddr_in6 *)&ro->ro_dst;
752 			*sa6 = *dstsock;
753 			sa6->sin6_scope_id = 0;
754 
755 			/*
756 			 * Currently dst has scopeid embedded iff it is LL.
757 			 * New routing API accepts scopeid as a separate argument.
758 			 * Convert dst before/after doing lookup
759 			 */
760 			uint32_t scopeid = 0;
761 			if (IN6_IS_SCOPE_LINKLOCAL(&sa6->sin6_addr)) {
762 				/* Unwrap in6_getscope() and in6_clearscope() */
763 				scopeid = ntohs(sa6->sin6_addr.s6_addr16[1]);
764 				sa6->sin6_addr.s6_addr16[1] = 0;
765 			}
766 
767 			ro->ro_nh = fib6_lookup(fibnum,
768 			    &sa6->sin6_addr, scopeid, NHR_REF, flowid);
769 
770 			if (IN6_IS_SCOPE_LINKLOCAL(&sa6->sin6_addr))
771 				sa6->sin6_addr.s6_addr16[1] = htons(scopeid);
772 		}
773 
774 		/*
775 		 * do not care about the result if we have the nexthop
776 		 * explicitly specified.
777 		 */
778 		if (opts && opts->ip6po_nexthop)
779 			goto done;
780 
781 		if (ro->ro_nh)
782 			ifp = ro->ro_nh->nh_ifp;
783 		else
784 			error = EHOSTUNREACH;
785 		nh = ro->ro_nh;
786 
787 		/*
788 		 * Check if the outgoing interface conflicts with
789 		 * the interface specified by ipi6_ifindex (if specified).
790 		 * Note that loopback interface is always okay.
791 		 * (this may happen when we are sending a packet to one of
792 		 *  our own addresses.)
793 		 */
794 		if (ifp && opts && opts->ip6po_pktinfo &&
795 		    opts->ip6po_pktinfo->ipi6_ifindex) {
796 			if (!(ifp->if_flags & IFF_LOOPBACK) &&
797 			    ifp->if_index !=
798 			    opts->ip6po_pktinfo->ipi6_ifindex) {
799 				error = EHOSTUNREACH;
800 				goto done;
801 			}
802 		}
803 	}
804 
805   done:
806 	if (ifp == NULL && nh == NULL) {
807 		/*
808 		 * This can happen if the caller did not pass a cached route
809 		 * nor any other hints.  We treat this case an error.
810 		 */
811 		error = EHOSTUNREACH;
812 	}
813 	if (error == EHOSTUNREACH)
814 		IP6STAT_INC(ip6s_noroute);
815 
816 	if (retifp != NULL) {
817 		if (nh != NULL)
818 			*retifp = nh->nh_aifp;
819 		else
820 			*retifp = ifp;
821 	}
822 
823 	if (retnh != NULL)
824 		*retnh = nh;	/* nh may be NULL */
825 
826 	return (error);
827 }
828 
829 static int
830 in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
831     struct ip6_moptions *mopts, struct ifnet **retifp,
832     struct ifnet *oifp, u_int fibnum)
833 {
834 	int error;
835 	struct route_in6 sro;
836 	struct nhop_object *nh = NULL;
837 	uint16_t nh_flags;
838 
839 	KASSERT(retifp != NULL, ("%s: retifp is NULL", __func__));
840 
841 	bzero(&sro, sizeof(sro));
842 	nh_flags = 0;
843 
844 	error = selectroute(dstsock, opts, mopts, &sro, retifp, &nh, 1, fibnum, 0);
845 
846 	if (nh != NULL)
847 		nh_flags = nh->nh_flags;
848 	if (nh != NULL && nh == sro.ro_nh)
849 		NH_FREE(nh);
850 
851 	if (error != 0) {
852 		/* Help ND. See oifp comment in in6_selectsrc(). */
853 		if (oifp != NULL && fibnum == RT_DEFAULT_FIB) {
854 			*retifp = oifp;
855 			error = 0;
856 		}
857 		return (error);
858 	}
859 
860 	/*
861 	 * do not use a rejected or black hole route.
862 	 * XXX: this check should be done in the L2 output routine.
863 	 * However, if we skipped this check here, we'd see the following
864 	 * scenario:
865 	 * - install a rejected route for a scoped address prefix
866 	 *   (like fe80::/10)
867 	 * - send a packet to a destination that matches the scoped prefix,
868 	 *   with ambiguity about the scope zone.
869 	 * - pick the outgoing interface from the route, and disambiguate the
870 	 *   scope zone with the interface.
871 	 * - ip6_output() would try to get another route with the "new"
872 	 *   destination, which may be valid.
873 	 * - we'd see no error on output.
874 	 * Although this may not be very harmful, it should still be confusing.
875 	 * We thus reject the case here.
876 	 */
877 
878 	if (nh_flags & (NHF_REJECT | NHF_BLACKHOLE)) {
879 		error = (nh_flags & NHF_HOST ? EHOSTUNREACH : ENETUNREACH);
880 		return (error);
881 	}
882 
883 	return (0);
884 }
885 
886 /* Public wrapper function to selectroute(). */
887 int
888 in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
889     struct ip6_moptions *mopts, struct route_in6 *ro,
890     struct ifnet **retifp, struct nhop_object **retnh, u_int fibnum, uint32_t flowid)
891 {
892 
893 	return (selectroute(dstsock, opts, mopts, ro, retifp,
894 	    retnh, 0, fibnum, flowid));
895 }
896 
897 /*
898  * Default hop limit selection. The precedence is as follows:
899  * 1. Hoplimit value specified via ioctl.
900  * 2. (If the outgoing interface is detected) the current
901  *     hop limit of the interface specified by router advertisement.
902  * 3. The system default hoplimit.
903  */
904 int
905 in6_selecthlim(struct inpcb *inp, struct ifnet *ifp)
906 {
907 
908 	if (inp && inp->in6p_hops >= 0)
909 		return (inp->in6p_hops);
910 	else if (ifp)
911 		return (ND_IFINFO(ifp)->chlim);
912 	else if (inp && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
913 		struct nhop_object *nh;
914 		struct in6_addr dst;
915 		uint32_t fibnum, scopeid;
916 		int hlim;
917 
918 		fibnum = inp->inp_inc.inc_fibnum;
919 		in6_splitscope(&inp->in6p_faddr, &dst, &scopeid);
920 		nh = fib6_lookup(fibnum, &dst, scopeid, 0, 0);
921 		if (nh != NULL) {
922 			hlim = ND_IFINFO(nh->nh_ifp)->chlim;
923 			return (hlim);
924 		}
925 	}
926 	return (V_ip6_defhlim);
927 }
928 
929 /*
930  * XXX: this is borrowed from in6_pcbbind(). If possible, we should
931  * share this function by all *bsd*...
932  */
933 int
934 in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct ucred *cred)
935 {
936 	struct socket *so = inp->inp_socket;
937 	u_int16_t lport = 0;
938 	int error, lookupflags = 0;
939 #ifdef INVARIANTS
940 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
941 #endif
942 
943 	INP_WLOCK_ASSERT(inp);
944 	INP_HASH_WLOCK_ASSERT(pcbinfo);
945 
946 	error = prison_local_ip6(cred, laddr,
947 	    ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0));
948 	if (error)
949 		return(error);
950 
951 	/* XXX: this is redundant when called from in6_pcbbind */
952 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT|SO_REUSEPORT_LB)) == 0)
953 		lookupflags = INPLOOKUP_WILDCARD;
954 
955 	inp->inp_flags |= INP_ANONPORT;
956 
957 	error = in_pcb_lport(inp, NULL, &lport, cred, lookupflags);
958 	if (error != 0)
959 		return (error);
960 
961 	inp->inp_lport = lport;
962 	if (in_pcbinshash(inp) != 0) {
963 		inp->in6p_laddr = in6addr_any;
964 		inp->inp_lport = 0;
965 		return (EAGAIN);
966 	}
967 
968 	return (0);
969 }
970 
971 void
972 addrsel_policy_init(void)
973 {
974 
975 	init_policy_queue();
976 
977 	/* initialize the "last resort" policy */
978 	bzero(&V_defaultaddrpolicy, sizeof(V_defaultaddrpolicy));
979 	V_defaultaddrpolicy.label = ADDR_LABEL_NOTAPP;
980 
981 	if (!IS_DEFAULT_VNET(curvnet))
982 		return;
983 
984 	ADDRSEL_LOCK_INIT();
985 	ADDRSEL_SXLOCK_INIT();
986 }
987 
988 static struct in6_addrpolicy *
989 lookup_addrsel_policy(struct sockaddr_in6 *key)
990 {
991 	struct in6_addrpolicy *match = NULL;
992 
993 	ADDRSEL_LOCK();
994 	match = match_addrsel_policy(key);
995 
996 	if (match == NULL)
997 		match = &V_defaultaddrpolicy;
998 	else
999 		match->use++;
1000 	ADDRSEL_UNLOCK();
1001 
1002 	return (match);
1003 }
1004 
1005 /*
1006  * Subroutines to manage the address selection policy table via sysctl.
1007  */
1008 struct walkarg {
1009 	struct sysctl_req *w_req;
1010 };
1011 
1012 static int in6_src_sysctl(SYSCTL_HANDLER_ARGS);
1013 SYSCTL_DECL(_net_inet6_ip6);
1014 static SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy,
1015     CTLFLAG_RD | CTLFLAG_MPSAFE, in6_src_sysctl,
1016     "");
1017 
1018 static int
1019 in6_src_sysctl(SYSCTL_HANDLER_ARGS)
1020 {
1021 	struct walkarg w;
1022 
1023 	if (req->newptr)
1024 		return EPERM;
1025 
1026 	bzero(&w, sizeof(w));
1027 	w.w_req = req;
1028 
1029 	return (walk_addrsel_policy(dump_addrsel_policyent, &w));
1030 }
1031 
1032 int
1033 in6_src_ioctl(u_long cmd, caddr_t data)
1034 {
1035 	struct in6_addrpolicy ent0;
1036 
1037 	if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY)
1038 		return (EOPNOTSUPP); /* check for safety */
1039 
1040 	ent0 = *(struct in6_addrpolicy *)data;
1041 
1042 	if (ent0.label == ADDR_LABEL_NOTAPP)
1043 		return (EINVAL);
1044 	/* check if the prefix mask is consecutive. */
1045 	if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0)
1046 		return (EINVAL);
1047 	/* clear trailing garbages (if any) of the prefix address. */
1048 	IN6_MASK_ADDR(&ent0.addr.sin6_addr, &ent0.addrmask.sin6_addr);
1049 	ent0.use = 0;
1050 
1051 	switch (cmd) {
1052 	case SIOCAADDRCTL_POLICY:
1053 		return (add_addrsel_policyent(&ent0));
1054 	case SIOCDADDRCTL_POLICY:
1055 		return (delete_addrsel_policyent(&ent0));
1056 	}
1057 
1058 	return (0);		/* XXX: compromise compilers */
1059 }
1060 
1061 /*
1062  * The followings are implementation of the policy table using a
1063  * simple tail queue.
1064  * XXX such details should be hidden.
1065  * XXX implementation using binary tree should be more efficient.
1066  */
1067 struct addrsel_policyent {
1068 	TAILQ_ENTRY(addrsel_policyent) ape_entry;
1069 	struct in6_addrpolicy ape_policy;
1070 };
1071 
1072 TAILQ_HEAD(addrsel_policyhead, addrsel_policyent);
1073 
1074 VNET_DEFINE_STATIC(struct addrsel_policyhead, addrsel_policytab);
1075 #define	V_addrsel_policytab		VNET(addrsel_policytab)
1076 
1077 static void
1078 init_policy_queue(void)
1079 {
1080 
1081 	TAILQ_INIT(&V_addrsel_policytab);
1082 }
1083 
1084 static int
1085 add_addrsel_policyent(struct in6_addrpolicy *newpolicy)
1086 {
1087 	struct addrsel_policyent *new, *pol;
1088 
1089 	new = malloc(sizeof(*new), M_IFADDR,
1090 	       M_WAITOK);
1091 	ADDRSEL_XLOCK();
1092 	ADDRSEL_LOCK();
1093 
1094 	/* duplication check */
1095 	TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1096 		if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr,
1097 				       &pol->ape_policy.addr.sin6_addr) &&
1098 		    IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr,
1099 				       &pol->ape_policy.addrmask.sin6_addr)) {
1100 			ADDRSEL_UNLOCK();
1101 			ADDRSEL_XUNLOCK();
1102 			free(new, M_IFADDR);
1103 			return (EEXIST);	/* or override it? */
1104 		}
1105 	}
1106 
1107 	bzero(new, sizeof(*new));
1108 
1109 	/* XXX: should validate entry */
1110 	new->ape_policy = *newpolicy;
1111 
1112 	TAILQ_INSERT_TAIL(&V_addrsel_policytab, new, ape_entry);
1113 	ADDRSEL_UNLOCK();
1114 	ADDRSEL_XUNLOCK();
1115 
1116 	return (0);
1117 }
1118 
1119 static int
1120 delete_addrsel_policyent(struct in6_addrpolicy *key)
1121 {
1122 	struct addrsel_policyent *pol;
1123 
1124 	ADDRSEL_XLOCK();
1125 	ADDRSEL_LOCK();
1126 
1127 	/* search for the entry in the table */
1128 	TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1129 		if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr,
1130 		    &pol->ape_policy.addr.sin6_addr) &&
1131 		    IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr,
1132 		    &pol->ape_policy.addrmask.sin6_addr)) {
1133 			break;
1134 		}
1135 	}
1136 	if (pol == NULL) {
1137 		ADDRSEL_UNLOCK();
1138 		ADDRSEL_XUNLOCK();
1139 		return (ESRCH);
1140 	}
1141 
1142 	TAILQ_REMOVE(&V_addrsel_policytab, pol, ape_entry);
1143 	ADDRSEL_UNLOCK();
1144 	ADDRSEL_XUNLOCK();
1145 	free(pol, M_IFADDR);
1146 
1147 	return (0);
1148 }
1149 
1150 static int
1151 walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *), void *w)
1152 {
1153 	struct addrsel_policyent *pol;
1154 	int error = 0;
1155 
1156 	ADDRSEL_SLOCK();
1157 	TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1158 		if ((error = (*callback)(&pol->ape_policy, w)) != 0) {
1159 			ADDRSEL_SUNLOCK();
1160 			return (error);
1161 		}
1162 	}
1163 	ADDRSEL_SUNLOCK();
1164 	return (error);
1165 }
1166 
1167 static int
1168 dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg)
1169 {
1170 	int error = 0;
1171 	struct walkarg *w = arg;
1172 
1173 	error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol));
1174 
1175 	return (error);
1176 }
1177 
1178 static struct in6_addrpolicy *
1179 match_addrsel_policy(struct sockaddr_in6 *key)
1180 {
1181 	struct addrsel_policyent *pent;
1182 	struct in6_addrpolicy *bestpol = NULL, *pol;
1183 	int matchlen, bestmatchlen = -1;
1184 	u_char *mp, *ep, *k, *p, m;
1185 
1186 	TAILQ_FOREACH(pent, &V_addrsel_policytab, ape_entry) {
1187 		matchlen = 0;
1188 
1189 		pol = &pent->ape_policy;
1190 		mp = (u_char *)&pol->addrmask.sin6_addr;
1191 		ep = mp + 16;	/* XXX: scope field? */
1192 		k = (u_char *)&key->sin6_addr;
1193 		p = (u_char *)&pol->addr.sin6_addr;
1194 		for (; mp < ep && *mp; mp++, k++, p++) {
1195 			m = *mp;
1196 			if ((*k & m) != *p)
1197 				goto next; /* not match */
1198 			if (m == 0xff) /* short cut for a typical case */
1199 				matchlen += 8;
1200 			else {
1201 				while (m >= 0x80) {
1202 					matchlen++;
1203 					m <<= 1;
1204 				}
1205 			}
1206 		}
1207 
1208 		/* matched.  check if this is better than the current best. */
1209 		if (bestpol == NULL ||
1210 		    matchlen > bestmatchlen) {
1211 			bestpol = pol;
1212 			bestmatchlen = matchlen;
1213 		}
1214 
1215 	  next:
1216 		continue;
1217 	}
1218 
1219 	return (bestpol);
1220 }
1221