xref: /freebsd/sys/netinet6/in6_src.c (revision 1f4bcc459a76b7aa664f3fd557684cd0ba6da352)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: in6_src.c,v 1.132 2003/08/26 04:42:27 keiichi Exp $
30  */
31 
32 /*-
33  * Copyright (c) 1982, 1986, 1991, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)in_pcb.c	8.2 (Berkeley) 1/4/94
61  */
62 
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
65 
66 #include "opt_inet.h"
67 #include "opt_inet6.h"
68 #include "opt_mpath.h"
69 
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/if_llatbl.h>
93 #ifdef RADIX_MPATH
94 #include <net/radix_mpath.h>
95 #endif
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 static VNET_DEFINE(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 rtentry **, int, u_int);
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 	rule = (r);	\
162 	/* { \
163 	char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
164 	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)); \
165 	} */ \
166 	goto replace; \
167 } while(0)
168 #define NEXT(r) do {\
169 	/* { \
170 	char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
171 	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)); \
172 	} */ \
173 	goto next;		/* XXX: we can't use 'continue' here */ \
174 } while(0)
175 #define BREAK(r) do { \
176 	IP6STAT_INC(ip6s_sources_rule[(r)]); \
177 	rule = (r);	\
178 	goto out;		/* XXX: we can't use 'break' here */ \
179 } while(0)
180 
181 static int
182 in6_selectsrc(uint32_t fibnum, struct sockaddr_in6 *dstsock,
183     struct ip6_pktopts *opts, struct inpcb *inp, struct ucred *cred,
184     struct ifnet **ifpp, struct in6_addr *srcp)
185 {
186 	struct rm_priotracker in6_ifa_tracker;
187 	struct in6_addr dst, tmp;
188 	struct ifnet *ifp = NULL, *oifp = NULL;
189 	struct in6_ifaddr *ia = NULL, *ia_best = NULL;
190 	struct in6_pktinfo *pi = NULL;
191 	int dst_scope = -1, best_scope = -1, best_matchlen = -1;
192 	struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL;
193 	u_int32_t odstzone;
194 	int prefer_tempaddr;
195 	int error, rule;
196 	struct ip6_moptions *mopts;
197 
198 	KASSERT(srcp != NULL, ("%s: srcp is NULL", __func__));
199 
200 	dst = dstsock->sin6_addr; /* make a copy for local operation */
201 	if (ifpp) {
202 		/*
203 		 * Save a possibly passed in ifp for in6_selectsrc. Only
204 		 * neighbor discovery code should use this feature, where
205 		 * we may know the interface but not the FIB number holding
206 		 * the connected subnet in case someone deleted it from the
207 		 * default FIB and we need to check the interface.
208 		 */
209 		if (*ifpp != NULL)
210 			oifp = *ifpp;
211 		*ifpp = NULL;
212 	}
213 
214 	if (inp != NULL) {
215 		INP_LOCK_ASSERT(inp);
216 		mopts = inp->in6p_moptions;
217 	} else {
218 		mopts = NULL;
219 	}
220 
221 	/*
222 	 * If the source address is explicitly specified by the caller,
223 	 * check if the requested source address is indeed a unicast address
224 	 * assigned to the node, and can be used as the packet's source
225 	 * address.  If everything is okay, use the address as source.
226 	 */
227 	if (opts && (pi = opts->ip6po_pktinfo) &&
228 	    !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) {
229 		struct sockaddr_in6 srcsock;
230 		struct in6_ifaddr *ia6;
231 
232 		/* get the outgoing interface */
233 		if ((error = in6_selectif(dstsock, opts, mopts, &ifp, oifp,
234 		    fibnum))
235 		    != 0)
236 			return (error);
237 
238 		/*
239 		 * determine the appropriate zone id of the source based on
240 		 * the zone of the destination and the outgoing interface.
241 		 * If the specified address is ambiguous wrt the scope zone,
242 		 * the interface must be specified; otherwise, ifa_ifwithaddr()
243 		 * will fail matching the address.
244 		 */
245 		bzero(&srcsock, sizeof(srcsock));
246 		srcsock.sin6_family = AF_INET6;
247 		srcsock.sin6_len = sizeof(srcsock);
248 		srcsock.sin6_addr = pi->ipi6_addr;
249 		if (ifp) {
250 			error = in6_setscope(&srcsock.sin6_addr, ifp, NULL);
251 			if (error)
252 				return (error);
253 		}
254 		if (cred != NULL && (error = prison_local_ip6(cred,
255 		    &srcsock.sin6_addr, (inp != NULL &&
256 		    (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
257 			return (error);
258 
259 		ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(
260 		    (struct sockaddr *)&srcsock);
261 		if (ia6 == NULL ||
262 		    (ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY))) {
263 			if (ia6 != NULL)
264 				ifa_free(&ia6->ia_ifa);
265 			return (EADDRNOTAVAIL);
266 		}
267 		pi->ipi6_addr = srcsock.sin6_addr; /* XXX: this overrides pi */
268 		if (ifpp)
269 			*ifpp = ifp;
270 		bcopy(&ia6->ia_addr.sin6_addr, srcp, sizeof(*srcp));
271 		ifa_free(&ia6->ia_ifa);
272 		return (0);
273 	}
274 
275 	/*
276 	 * Otherwise, if the socket has already bound the source, just use it.
277 	 */
278 	if (inp != NULL && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
279 		if (cred != NULL &&
280 		    (error = prison_local_ip6(cred, &inp->in6p_laddr,
281 		    ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
282 			return (error);
283 		bcopy(&inp->in6p_laddr, srcp, sizeof(*srcp));
284 		return (0);
285 	}
286 
287 	/*
288 	 * Bypass source address selection and use the primary jail IP
289 	 * if requested.
290 	 */
291 	if (cred != NULL && !prison_saddrsel_ip6(cred, srcp))
292 		return (0);
293 
294 	/*
295 	 * If the address is not specified, choose the best one based on
296 	 * the outgoing interface and the destination address.
297 	 */
298 	/* get the outgoing interface */
299 	if ((error = in6_selectif(dstsock, opts, mopts, &ifp, oifp,
300 	    (inp != NULL) ? inp->inp_inc.inc_fibnum : RT_DEFAULT_FIB)) != 0)
301 		return (error);
302 
303 #ifdef DIAGNOSTIC
304 	if (ifp == NULL)	/* this should not happen */
305 		panic("in6_selectsrc: NULL ifp");
306 #endif
307 	error = in6_setscope(&dst, ifp, &odstzone);
308 	if (error)
309 		return (error);
310 
311 	rule = 0;
312 	IN6_IFADDR_RLOCK(&in6_ifa_tracker);
313 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
314 		int new_scope = -1, new_matchlen = -1;
315 		struct in6_addrpolicy *new_policy = NULL;
316 		u_int32_t srczone, osrczone, dstzone;
317 		struct in6_addr src;
318 		struct ifnet *ifp1 = ia->ia_ifp;
319 
320 		/*
321 		 * We'll never take an address that breaks the scope zone
322 		 * of the destination.  We also skip an address if its zone
323 		 * does not contain the outgoing interface.
324 		 * XXX: we should probably use sin6_scope_id here.
325 		 */
326 		if (in6_setscope(&dst, ifp1, &dstzone) ||
327 		    odstzone != dstzone) {
328 			continue;
329 		}
330 		src = ia->ia_addr.sin6_addr;
331 		if (in6_setscope(&src, ifp, &osrczone) ||
332 		    in6_setscope(&src, ifp1, &srczone) ||
333 		    osrczone != srczone) {
334 			continue;
335 		}
336 
337 		/* avoid unusable addresses */
338 		if ((ia->ia6_flags &
339 		     (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) {
340 				continue;
341 		}
342 		if (!V_ip6_use_deprecated && IFA6_IS_DEPRECATED(ia))
343 			continue;
344 
345 		/* If jailed only take addresses of the jail into account. */
346 		if (cred != NULL &&
347 		    prison_check_ip6(cred, &ia->ia_addr.sin6_addr) != 0)
348 			continue;
349 
350 		/* Rule 1: Prefer same address */
351 		if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) {
352 			ia_best = ia;
353 			BREAK(1); /* there should be no better candidate */
354 		}
355 
356 		if (ia_best == NULL)
357 			REPLACE(0);
358 
359 		/* Rule 2: Prefer appropriate scope */
360 		if (dst_scope < 0)
361 			dst_scope = in6_addrscope(&dst);
362 		new_scope = in6_addrscope(&ia->ia_addr.sin6_addr);
363 		if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) {
364 			if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0)
365 				REPLACE(2);
366 			NEXT(2);
367 		} else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) {
368 			if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0)
369 				NEXT(2);
370 			REPLACE(2);
371 		}
372 
373 		/*
374 		 * Rule 3: Avoid deprecated addresses.  Note that the case of
375 		 * !ip6_use_deprecated is already rejected above.
376 		 */
377 		if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia))
378 			NEXT(3);
379 		if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia))
380 			REPLACE(3);
381 
382 		/* Rule 4: Prefer home addresses */
383 		/*
384 		 * XXX: This is a TODO.  We should probably merge the MIP6
385 		 * case above.
386 		 */
387 
388 		/* Rule 5: Prefer outgoing interface */
389 		if (!(ND_IFINFO(ifp)->flags & ND6_IFF_NO_PREFER_IFACE)) {
390 			if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp)
391 				NEXT(5);
392 			if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp)
393 				REPLACE(5);
394 		}
395 
396 		/*
397 		 * Rule 6: Prefer matching label
398 		 * Note that best_policy should be non-NULL here.
399 		 */
400 		if (dst_policy == NULL)
401 			dst_policy = lookup_addrsel_policy(dstsock);
402 		if (dst_policy->label != ADDR_LABEL_NOTAPP) {
403 			new_policy = lookup_addrsel_policy(&ia->ia_addr);
404 			if (dst_policy->label == best_policy->label &&
405 			    dst_policy->label != new_policy->label)
406 				NEXT(6);
407 			if (dst_policy->label != best_policy->label &&
408 			    dst_policy->label == new_policy->label)
409 				REPLACE(6);
410 		}
411 
412 		/*
413 		 * Rule 7: Prefer public addresses.
414 		 * We allow users to reverse the logic by configuring
415 		 * a sysctl variable, so that privacy conscious users can
416 		 * always prefer temporary addresses.
417 		 */
418 		if (opts == NULL ||
419 		    opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) {
420 			prefer_tempaddr = V_ip6_prefer_tempaddr;
421 		} else if (opts->ip6po_prefer_tempaddr ==
422 		    IP6PO_TEMPADDR_NOTPREFER) {
423 			prefer_tempaddr = 0;
424 		} else
425 			prefer_tempaddr = 1;
426 		if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
427 		    (ia->ia6_flags & IN6_IFF_TEMPORARY)) {
428 			if (prefer_tempaddr)
429 				REPLACE(7);
430 			else
431 				NEXT(7);
432 		}
433 		if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
434 		    !(ia->ia6_flags & IN6_IFF_TEMPORARY)) {
435 			if (prefer_tempaddr)
436 				NEXT(7);
437 			else
438 				REPLACE(7);
439 		}
440 
441 		/*
442 		 * Rule 8: prefer addresses on alive interfaces.
443 		 * This is a KAME specific rule.
444 		 */
445 		if ((ia_best->ia_ifp->if_flags & IFF_UP) &&
446 		    !(ia->ia_ifp->if_flags & IFF_UP))
447 			NEXT(8);
448 		if (!(ia_best->ia_ifp->if_flags & IFF_UP) &&
449 		    (ia->ia_ifp->if_flags & IFF_UP))
450 			REPLACE(8);
451 
452 		/*
453 		 * Rule 9: prefer address with better virtual status.
454 		 */
455 		if (ifa_preferred(&ia_best->ia_ifa, &ia->ia_ifa))
456 			REPLACE(9);
457 		if (ifa_preferred(&ia->ia_ifa, &ia_best->ia_ifa))
458 			NEXT(9);
459 
460 		/*
461 		 * Rule 10: prefer address with `prefer_source' flag.
462 		 */
463 		if ((ia_best->ia6_flags & IN6_IFF_PREFER_SOURCE) == 0 &&
464 		    (ia->ia6_flags & IN6_IFF_PREFER_SOURCE) != 0)
465 			REPLACE(10);
466 		if ((ia_best->ia6_flags & IN6_IFF_PREFER_SOURCE) != 0 &&
467 		    (ia->ia6_flags & IN6_IFF_PREFER_SOURCE) == 0)
468 			NEXT(10);
469 
470 		/*
471 		 * Rule 14: Use longest matching prefix.
472 		 * Note: in the address selection draft, this rule is
473 		 * documented as "Rule 8".  However, since it is also
474 		 * documented that this rule can be overridden, we assign
475 		 * a large number so that it is easy to assign smaller numbers
476 		 * to more preferred rules.
477 		 */
478 		new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst);
479 		if (best_matchlen < new_matchlen)
480 			REPLACE(14);
481 		if (new_matchlen < best_matchlen)
482 			NEXT(14);
483 
484 		/* Rule 15 is reserved. */
485 
486 		/*
487 		 * Last resort: just keep the current candidate.
488 		 * Or, do we need more rules?
489 		 */
490 		continue;
491 
492 	  replace:
493 		ia_best = ia;
494 		best_scope = (new_scope >= 0 ? new_scope :
495 			      in6_addrscope(&ia_best->ia_addr.sin6_addr));
496 		best_policy = (new_policy ? new_policy :
497 			       lookup_addrsel_policy(&ia_best->ia_addr));
498 		best_matchlen = (new_matchlen >= 0 ? new_matchlen :
499 				 in6_matchlen(&ia_best->ia_addr.sin6_addr,
500 					      &dst));
501 
502 	  next:
503 		continue;
504 
505 	  out:
506 		break;
507 	}
508 
509 	if ((ia = ia_best) == NULL) {
510 		IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
511 		IP6STAT_INC(ip6s_sources_none);
512 		return (EADDRNOTAVAIL);
513 	}
514 
515 	/*
516 	 * At this point at least one of the addresses belonged to the jail
517 	 * but it could still be, that we want to further restrict it, e.g.
518 	 * theoratically IN6_IS_ADDR_LOOPBACK.
519 	 * It must not be IN6_IS_ADDR_UNSPECIFIED anymore.
520 	 * prison_local_ip6() will fix an IN6_IS_ADDR_LOOPBACK but should
521 	 * let all others previously selected pass.
522 	 * Use tmp to not change ::1 on lo0 to the primary jail address.
523 	 */
524 	tmp = ia->ia_addr.sin6_addr;
525 	if (cred != NULL && prison_local_ip6(cred, &tmp, (inp != NULL &&
526 	    (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0) {
527 		IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
528 		IP6STAT_INC(ip6s_sources_none);
529 		return (EADDRNOTAVAIL);
530 	}
531 
532 	if (ifpp)
533 		*ifpp = ifp;
534 
535 	bcopy(&tmp, srcp, sizeof(*srcp));
536 	if (ia->ia_ifp == ifp)
537 		IP6STAT_INC(ip6s_sources_sameif[best_scope]);
538 	else
539 		IP6STAT_INC(ip6s_sources_otherif[best_scope]);
540 	if (dst_scope == best_scope)
541 		IP6STAT_INC(ip6s_sources_samescope[best_scope]);
542 	else
543 		IP6STAT_INC(ip6s_sources_otherscope[best_scope]);
544 	if (IFA6_IS_DEPRECATED(ia))
545 		IP6STAT_INC(ip6s_sources_deprecated[best_scope]);
546 	IN6_IFADDR_RUNLOCK(&in6_ifa_tracker);
547 	return (0);
548 }
549 
550 /*
551  * Select source address based on @inp, @dstsock and @opts.
552  * Stores selected address to @srcp. If @scope_ambiguous is set,
553  * embed scope from selected outgoing interface. If @hlim pointer
554  * is provided, stores calculated hop limit there.
555  * Returns 0 on success.
556  */
557 int
558 in6_selectsrc_socket(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
559     struct inpcb *inp, struct ucred *cred, int scope_ambiguous,
560     struct in6_addr *srcp, int *hlim)
561 {
562 	struct ifnet *retifp;
563 	uint32_t fibnum;
564 	int error;
565 
566 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : RT_DEFAULT_FIB;
567 	retifp = NULL;
568 
569 	error = in6_selectsrc(fibnum, dstsock, opts, inp, cred, &retifp, srcp);
570 	if (error != 0)
571 		return (error);
572 
573 	if (hlim != NULL)
574 		*hlim = in6_selecthlim(inp, retifp);
575 
576 	if (retifp == NULL || scope_ambiguous == 0)
577 		return (0);
578 
579 	/*
580 	 * Application should provide a proper zone ID or the use of
581 	 * default zone IDs should be enabled.  Unfortunately, some
582 	 * applications do not behave as it should, so we need a
583 	 * workaround.  Even if an appropriate ID is not determined
584 	 * (when it's required), if we can determine the outgoing
585 	 * interface. determine the zone ID based on the interface.
586 	 */
587 	error = in6_setscope(&dstsock->sin6_addr, retifp, NULL);
588 
589 	return (error);
590 }
591 
592 /*
593  * Select source address based on @fibnum, @dst and @scopeid.
594  * Stores selected address to @srcp.
595  * Returns 0 on success.
596  *
597  * Used by non-socket based consumers (ND code mostly)
598  */
599 int
600 in6_selectsrc_addr(uint32_t fibnum, const struct in6_addr *dst,
601     uint32_t scopeid, struct ifnet *ifp, struct in6_addr *srcp,
602     int *hlim)
603 {
604 	struct ifnet *retifp;
605 	struct sockaddr_in6 dst_sa;
606 	int error;
607 
608 	retifp = ifp;
609 	bzero(&dst_sa, sizeof(dst_sa));
610 	dst_sa.sin6_family = AF_INET6;
611 	dst_sa.sin6_len = sizeof(dst_sa);
612 	dst_sa.sin6_addr = *dst;
613 	dst_sa.sin6_scope_id = scopeid;
614 	sa6_embedscope(&dst_sa, 0);
615 
616 	error = in6_selectsrc(fibnum, &dst_sa, NULL, NULL, NULL, &retifp, srcp);
617 	if (hlim != NULL)
618 		*hlim = in6_selecthlim(NULL, retifp);
619 
620 	return (error);
621 }
622 
623 /*
624  * clone - meaningful only for bsdi and freebsd
625  */
626 static int
627 selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
628     struct ip6_moptions *mopts, struct route_in6 *ro,
629     struct ifnet **retifp, struct rtentry **retrt, int norouteok, u_int fibnum)
630 {
631 	int error = 0;
632 	struct ifnet *ifp = NULL;
633 	struct rtentry *rt = NULL;
634 	struct sockaddr_in6 *sin6_next;
635 	struct in6_pktinfo *pi = NULL;
636 	struct in6_addr *dst = &dstsock->sin6_addr;
637 	uint32_t zoneid;
638 #if 0
639 	char ip6buf[INET6_ADDRSTRLEN];
640 
641 	if (dstsock->sin6_addr.s6_addr32[0] == 0 &&
642 	    dstsock->sin6_addr.s6_addr32[1] == 0 &&
643 	    !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) {
644 		printf("in6_selectroute: strange destination %s\n",
645 		       ip6_sprintf(ip6buf, &dstsock->sin6_addr));
646 	} else {
647 		printf("in6_selectroute: destination = %s%%%d\n",
648 		       ip6_sprintf(ip6buf, &dstsock->sin6_addr),
649 		       dstsock->sin6_scope_id); /* for debug */
650 	}
651 #endif
652 
653 	/* If the caller specify the outgoing interface explicitly, use it. */
654 	if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) {
655 		/* XXX boundary check is assumed to be already done. */
656 		ifp = ifnet_byindex(pi->ipi6_ifindex);
657 		if (ifp != NULL &&
658 		    (norouteok || retrt == NULL ||
659 		    IN6_IS_ADDR_MULTICAST(dst))) {
660 			/*
661 			 * we do not have to check or get the route for
662 			 * multicast.
663 			 */
664 			goto done;
665 		} else
666 			goto getroute;
667 	}
668 	/*
669 	 * If the destination address is a multicast address and the outgoing
670 	 * interface for the address is specified by the caller, use it.
671 	 */
672 	if (IN6_IS_ADDR_MULTICAST(dst) &&
673 	    mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) {
674 		goto done; /* we do not need a route for multicast. */
675 	}
676 	/*
677 	 * If destination address is LLA or link- or node-local multicast,
678 	 * use it's embedded scope zone id to determine outgoing interface.
679 	 */
680 	if (IN6_IS_ADDR_MC_LINKLOCAL(dst) ||
681 	    IN6_IS_ADDR_MC_NODELOCAL(dst)) {
682 		zoneid = ntohs(in6_getscope(dst));
683 		if (zoneid > 0) {
684 			ifp = in6_getlinkifnet(zoneid);
685 			goto done;
686 		}
687 	}
688 
689   getroute:
690 	/*
691 	 * If the next hop address for the packet is specified by the caller,
692 	 * use it as the gateway.
693 	 */
694 	if (opts && opts->ip6po_nexthop) {
695 		struct route_in6 *ron;
696 
697 		sin6_next = satosin6(opts->ip6po_nexthop);
698 		if (IN6_IS_ADDR_LINKLOCAL(&sin6_next->sin6_addr)) {
699 			/*
700 			 * Next hop is LLA, thus it should be neighbor.
701 			 * Determine outgoing interface by zone index.
702 			 */
703 			zoneid = ntohs(in6_getscope(&sin6_next->sin6_addr));
704 			if (zoneid > 0) {
705 				ifp = in6_getlinkifnet(zoneid);
706 				goto done;
707 			}
708 		}
709 		ron = &opts->ip6po_nextroute;
710 		/* Use a cached route if it exists and is valid. */
711 		if (ron->ro_rt != NULL && (
712 		    (ron->ro_rt->rt_flags & RTF_UP) == 0 ||
713 		    ron->ro_dst.sin6_family != AF_INET6 ||
714 		    !IN6_ARE_ADDR_EQUAL(&ron->ro_dst.sin6_addr,
715 			&sin6_next->sin6_addr)))
716 			RO_RTFREE(ron);
717 		if (ron->ro_rt == NULL) {
718 			ron->ro_dst = *sin6_next;
719 			in6_rtalloc(ron, fibnum); /* multi path case? */
720 		}
721 		/*
722 		 * The node identified by that address must be a
723 		 * neighbor of the sending host.
724 		 */
725 		if (ron->ro_rt == NULL ||
726 		    (ron->ro_rt->rt_flags & RTF_GATEWAY) != 0)
727 			error = EHOSTUNREACH;
728 		goto done;
729 	}
730 
731 	/*
732 	 * Use a cached route if it exists and is valid, else try to allocate
733 	 * a new one.  Note that we should check the address family of the
734 	 * cached destination, in case of sharing the cache with IPv4.
735 	 */
736 	if (ro) {
737 		if (ro->ro_rt &&
738 		    (!(ro->ro_rt->rt_flags & RTF_UP) ||
739 		     ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 ||
740 		     !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr,
741 		     dst))) {
742 			RTFREE(ro->ro_rt);
743 			ro->ro_rt = (struct rtentry *)NULL;
744 		}
745 		if (ro->ro_rt == (struct rtentry *)NULL) {
746 			struct sockaddr_in6 *sa6;
747 
748 			/* No route yet, so try to acquire one */
749 			bzero(&ro->ro_dst, sizeof(struct sockaddr_in6));
750 			sa6 = (struct sockaddr_in6 *)&ro->ro_dst;
751 			*sa6 = *dstsock;
752 			sa6->sin6_scope_id = 0;
753 
754 #ifdef RADIX_MPATH
755 				rtalloc_mpath_fib((struct route *)ro,
756 				    ntohl(sa6->sin6_addr.s6_addr32[3]), fibnum);
757 #else
758 				ro->ro_rt = in6_rtalloc1((struct sockaddr *)
759 				    &ro->ro_dst, 0, 0UL, fibnum);
760 				if (ro->ro_rt)
761 					RT_UNLOCK(ro->ro_rt);
762 #endif
763 		}
764 
765 		/*
766 		 * do not care about the result if we have the nexthop
767 		 * explicitly specified.
768 		 */
769 		if (opts && opts->ip6po_nexthop)
770 			goto done;
771 
772 		if (ro->ro_rt) {
773 			ifp = ro->ro_rt->rt_ifp;
774 
775 			if (ifp == NULL) { /* can this really happen? */
776 				RTFREE(ro->ro_rt);
777 				ro->ro_rt = NULL;
778 			}
779 		}
780 		if (ro->ro_rt == NULL)
781 			error = EHOSTUNREACH;
782 		rt = ro->ro_rt;
783 
784 		/*
785 		 * Check if the outgoing interface conflicts with
786 		 * the interface specified by ipi6_ifindex (if specified).
787 		 * Note that loopback interface is always okay.
788 		 * (this may happen when we are sending a packet to one of
789 		 *  our own addresses.)
790 		 */
791 		if (ifp && opts && opts->ip6po_pktinfo &&
792 		    opts->ip6po_pktinfo->ipi6_ifindex) {
793 			if (!(ifp->if_flags & IFF_LOOPBACK) &&
794 			    ifp->if_index !=
795 			    opts->ip6po_pktinfo->ipi6_ifindex) {
796 				error = EHOSTUNREACH;
797 				goto done;
798 			}
799 		}
800 	}
801 
802   done:
803 	if (ifp == NULL && rt == NULL) {
804 		/*
805 		 * This can happen if the caller did not pass a cached route
806 		 * nor any other hints.  We treat this case an error.
807 		 */
808 		error = EHOSTUNREACH;
809 	}
810 	if (error == EHOSTUNREACH)
811 		IP6STAT_INC(ip6s_noroute);
812 
813 	if (retifp != NULL) {
814 		*retifp = ifp;
815 
816 		/*
817 		 * Adjust the "outgoing" interface.  If we're going to loop
818 		 * the packet back to ourselves, the ifp would be the loopback
819 		 * interface. However, we'd rather know the interface associated
820 		 * to the destination address (which should probably be one of
821 		 * our own addresses.)
822 		 */
823 		if (rt) {
824 			if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) &&
825 			    (rt->rt_gateway->sa_family == AF_LINK))
826 				*retifp =
827 					ifnet_byindex(((struct sockaddr_dl *)
828 						       rt->rt_gateway)->sdl_index);
829 		}
830 	}
831 
832 	if (retrt != NULL)
833 		*retrt = rt;	/* rt may be NULL */
834 
835 	return (error);
836 }
837 
838 static int
839 in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
840     struct ip6_moptions *mopts, struct ifnet **retifp,
841     struct ifnet *oifp, u_int fibnum)
842 {
843 	int error;
844 	struct route_in6 sro;
845 	struct rtentry *rt = NULL;
846 	int rt_flags;
847 
848 	KASSERT(retifp != NULL, ("%s: retifp is NULL", __func__));
849 
850 	bzero(&sro, sizeof(sro));
851 	rt_flags = 0;
852 
853 	error = selectroute(dstsock, opts, mopts, &sro, retifp, &rt, 1, fibnum);
854 
855 	if (rt)
856 		rt_flags = rt->rt_flags;
857 	if (rt && rt == sro.ro_rt)
858 		RTFREE(rt);
859 
860 	if (error != 0) {
861 		/* Help ND. See oifp comment in in6_selectsrc(). */
862 		if (oifp != NULL && fibnum == RT_DEFAULT_FIB) {
863 			*retifp = oifp;
864 			error = 0;
865 		}
866 		return (error);
867 	}
868 
869 	/*
870 	 * do not use a rejected or black hole route.
871 	 * XXX: this check should be done in the L2 output routine.
872 	 * However, if we skipped this check here, we'd see the following
873 	 * scenario:
874 	 * - install a rejected route for a scoped address prefix
875 	 *   (like fe80::/10)
876 	 * - send a packet to a destination that matches the scoped prefix,
877 	 *   with ambiguity about the scope zone.
878 	 * - pick the outgoing interface from the route, and disambiguate the
879 	 *   scope zone with the interface.
880 	 * - ip6_output() would try to get another route with the "new"
881 	 *   destination, which may be valid.
882 	 * - we'd see no error on output.
883 	 * Although this may not be very harmful, it should still be confusing.
884 	 * We thus reject the case here.
885 	 */
886 
887 	if (rt_flags & (RTF_REJECT | RTF_BLACKHOLE)) {
888 		error = (rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
889 		return (error);
890 	}
891 
892 	return (0);
893 }
894 
895 /*
896  * Public wrapper function to selectroute().
897  *
898  * XXX-BZ in6_selectroute() should and will grow the FIB argument. The
899  * in6_selectroute_fib() function is only there for backward compat on stable.
900  */
901 int
902 in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
903     struct ip6_moptions *mopts, struct route_in6 *ro,
904     struct ifnet **retifp, struct rtentry **retrt)
905 {
906 
907 	return (selectroute(dstsock, opts, mopts, ro, retifp,
908 	    retrt, 0, RT_DEFAULT_FIB));
909 }
910 
911 #ifndef BURN_BRIDGES
912 int
913 in6_selectroute_fib(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
914     struct ip6_moptions *mopts, struct route_in6 *ro,
915     struct ifnet **retifp, struct rtentry **retrt, u_int fibnum)
916 {
917 
918 	return (selectroute(dstsock, opts, mopts, ro, retifp,
919 	    retrt, 0, fibnum));
920 }
921 #endif
922 
923 /*
924  * Default hop limit selection. The precedence is as follows:
925  * 1. Hoplimit value specified via ioctl.
926  * 2. (If the outgoing interface is detected) the current
927  *     hop limit of the interface specified by router advertisement.
928  * 3. The system default hoplimit.
929  */
930 int
931 in6_selecthlim(struct inpcb *in6p, struct ifnet *ifp)
932 {
933 
934 	if (in6p && in6p->in6p_hops >= 0)
935 		return (in6p->in6p_hops);
936 	else if (ifp)
937 		return (ND_IFINFO(ifp)->chlim);
938 	else if (in6p && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
939 		struct nhop6_basic nh6;
940 		struct in6_addr dst;
941 		uint32_t fibnum, scopeid;
942 		int hlim;
943 
944 		fibnum = in6p->inp_inc.inc_fibnum;
945 		in6_splitscope(&in6p->in6p_faddr, &dst, &scopeid);
946 		if (fib6_lookup_nh_basic(fibnum, &dst, scopeid, 0, 0, &nh6)==0){
947 			hlim = ND_IFINFO(nh6.nh_ifp)->chlim;
948 			return (hlim);
949 		}
950 	}
951 	return (V_ip6_defhlim);
952 }
953 
954 /*
955  * XXX: this is borrowed from in6_pcbbind(). If possible, we should
956  * share this function by all *bsd*...
957  */
958 int
959 in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct ucred *cred)
960 {
961 	struct socket *so = inp->inp_socket;
962 	u_int16_t lport = 0;
963 	int error, lookupflags = 0;
964 #ifdef INVARIANTS
965 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
966 #endif
967 
968 	INP_WLOCK_ASSERT(inp);
969 	INP_HASH_WLOCK_ASSERT(pcbinfo);
970 
971 	error = prison_local_ip6(cred, laddr,
972 	    ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0));
973 	if (error)
974 		return(error);
975 
976 	/* XXX: this is redundant when called from in6_pcbbind */
977 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
978 		lookupflags = INPLOOKUP_WILDCARD;
979 
980 	inp->inp_flags |= INP_ANONPORT;
981 
982 	error = in_pcb_lport(inp, NULL, &lport, cred, lookupflags);
983 	if (error != 0)
984 		return (error);
985 
986 	inp->inp_lport = lport;
987 	if (in_pcbinshash(inp) != 0) {
988 		inp->in6p_laddr = in6addr_any;
989 		inp->inp_lport = 0;
990 		return (EAGAIN);
991 	}
992 
993 	return (0);
994 }
995 
996 void
997 addrsel_policy_init(void)
998 {
999 
1000 	init_policy_queue();
1001 
1002 	/* initialize the "last resort" policy */
1003 	bzero(&V_defaultaddrpolicy, sizeof(V_defaultaddrpolicy));
1004 	V_defaultaddrpolicy.label = ADDR_LABEL_NOTAPP;
1005 
1006 	if (!IS_DEFAULT_VNET(curvnet))
1007 		return;
1008 
1009 	ADDRSEL_LOCK_INIT();
1010 	ADDRSEL_SXLOCK_INIT();
1011 }
1012 
1013 static struct in6_addrpolicy *
1014 lookup_addrsel_policy(struct sockaddr_in6 *key)
1015 {
1016 	struct in6_addrpolicy *match = NULL;
1017 
1018 	ADDRSEL_LOCK();
1019 	match = match_addrsel_policy(key);
1020 
1021 	if (match == NULL)
1022 		match = &V_defaultaddrpolicy;
1023 	else
1024 		match->use++;
1025 	ADDRSEL_UNLOCK();
1026 
1027 	return (match);
1028 }
1029 
1030 /*
1031  * Subroutines to manage the address selection policy table via sysctl.
1032  */
1033 struct walkarg {
1034 	struct sysctl_req *w_req;
1035 };
1036 
1037 static int in6_src_sysctl(SYSCTL_HANDLER_ARGS);
1038 SYSCTL_DECL(_net_inet6_ip6);
1039 static SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy,
1040 	CTLFLAG_RD, in6_src_sysctl, "");
1041 
1042 static int
1043 in6_src_sysctl(SYSCTL_HANDLER_ARGS)
1044 {
1045 	struct walkarg w;
1046 
1047 	if (req->newptr)
1048 		return EPERM;
1049 
1050 	bzero(&w, sizeof(w));
1051 	w.w_req = req;
1052 
1053 	return (walk_addrsel_policy(dump_addrsel_policyent, &w));
1054 }
1055 
1056 int
1057 in6_src_ioctl(u_long cmd, caddr_t data)
1058 {
1059 	struct in6_addrpolicy ent0;
1060 
1061 	if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY)
1062 		return (EOPNOTSUPP); /* check for safety */
1063 
1064 	ent0 = *(struct in6_addrpolicy *)data;
1065 
1066 	if (ent0.label == ADDR_LABEL_NOTAPP)
1067 		return (EINVAL);
1068 	/* check if the prefix mask is consecutive. */
1069 	if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0)
1070 		return (EINVAL);
1071 	/* clear trailing garbages (if any) of the prefix address. */
1072 	IN6_MASK_ADDR(&ent0.addr.sin6_addr, &ent0.addrmask.sin6_addr);
1073 	ent0.use = 0;
1074 
1075 	switch (cmd) {
1076 	case SIOCAADDRCTL_POLICY:
1077 		return (add_addrsel_policyent(&ent0));
1078 	case SIOCDADDRCTL_POLICY:
1079 		return (delete_addrsel_policyent(&ent0));
1080 	}
1081 
1082 	return (0);		/* XXX: compromise compilers */
1083 }
1084 
1085 /*
1086  * The followings are implementation of the policy table using a
1087  * simple tail queue.
1088  * XXX such details should be hidden.
1089  * XXX implementation using binary tree should be more efficient.
1090  */
1091 struct addrsel_policyent {
1092 	TAILQ_ENTRY(addrsel_policyent) ape_entry;
1093 	struct in6_addrpolicy ape_policy;
1094 };
1095 
1096 TAILQ_HEAD(addrsel_policyhead, addrsel_policyent);
1097 
1098 static VNET_DEFINE(struct addrsel_policyhead, addrsel_policytab);
1099 #define	V_addrsel_policytab		VNET(addrsel_policytab)
1100 
1101 static void
1102 init_policy_queue(void)
1103 {
1104 
1105 	TAILQ_INIT(&V_addrsel_policytab);
1106 }
1107 
1108 static int
1109 add_addrsel_policyent(struct in6_addrpolicy *newpolicy)
1110 {
1111 	struct addrsel_policyent *new, *pol;
1112 
1113 	new = malloc(sizeof(*new), M_IFADDR,
1114 	       M_WAITOK);
1115 	ADDRSEL_XLOCK();
1116 	ADDRSEL_LOCK();
1117 
1118 	/* duplication check */
1119 	TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1120 		if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr,
1121 				       &pol->ape_policy.addr.sin6_addr) &&
1122 		    IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr,
1123 				       &pol->ape_policy.addrmask.sin6_addr)) {
1124 			ADDRSEL_UNLOCK();
1125 			ADDRSEL_XUNLOCK();
1126 			free(new, M_IFADDR);
1127 			return (EEXIST);	/* or override it? */
1128 		}
1129 	}
1130 
1131 	bzero(new, sizeof(*new));
1132 
1133 	/* XXX: should validate entry */
1134 	new->ape_policy = *newpolicy;
1135 
1136 	TAILQ_INSERT_TAIL(&V_addrsel_policytab, new, ape_entry);
1137 	ADDRSEL_UNLOCK();
1138 	ADDRSEL_XUNLOCK();
1139 
1140 	return (0);
1141 }
1142 
1143 static int
1144 delete_addrsel_policyent(struct in6_addrpolicy *key)
1145 {
1146 	struct addrsel_policyent *pol;
1147 
1148 	ADDRSEL_XLOCK();
1149 	ADDRSEL_LOCK();
1150 
1151 	/* search for the entry in the table */
1152 	TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1153 		if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr,
1154 		    &pol->ape_policy.addr.sin6_addr) &&
1155 		    IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr,
1156 		    &pol->ape_policy.addrmask.sin6_addr)) {
1157 			break;
1158 		}
1159 	}
1160 	if (pol == NULL) {
1161 		ADDRSEL_UNLOCK();
1162 		ADDRSEL_XUNLOCK();
1163 		return (ESRCH);
1164 	}
1165 
1166 	TAILQ_REMOVE(&V_addrsel_policytab, pol, ape_entry);
1167 	ADDRSEL_UNLOCK();
1168 	ADDRSEL_XUNLOCK();
1169 	free(pol, M_IFADDR);
1170 
1171 	return (0);
1172 }
1173 
1174 static int
1175 walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *), void *w)
1176 {
1177 	struct addrsel_policyent *pol;
1178 	int error = 0;
1179 
1180 	ADDRSEL_SLOCK();
1181 	TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1182 		if ((error = (*callback)(&pol->ape_policy, w)) != 0) {
1183 			ADDRSEL_SUNLOCK();
1184 			return (error);
1185 		}
1186 	}
1187 	ADDRSEL_SUNLOCK();
1188 	return (error);
1189 }
1190 
1191 static int
1192 dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg)
1193 {
1194 	int error = 0;
1195 	struct walkarg *w = arg;
1196 
1197 	error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol));
1198 
1199 	return (error);
1200 }
1201 
1202 static struct in6_addrpolicy *
1203 match_addrsel_policy(struct sockaddr_in6 *key)
1204 {
1205 	struct addrsel_policyent *pent;
1206 	struct in6_addrpolicy *bestpol = NULL, *pol;
1207 	int matchlen, bestmatchlen = -1;
1208 	u_char *mp, *ep, *k, *p, m;
1209 
1210 	TAILQ_FOREACH(pent, &V_addrsel_policytab, ape_entry) {
1211 		matchlen = 0;
1212 
1213 		pol = &pent->ape_policy;
1214 		mp = (u_char *)&pol->addrmask.sin6_addr;
1215 		ep = mp + 16;	/* XXX: scope field? */
1216 		k = (u_char *)&key->sin6_addr;
1217 		p = (u_char *)&pol->addr.sin6_addr;
1218 		for (; mp < ep && *mp; mp++, k++, p++) {
1219 			m = *mp;
1220 			if ((*k & m) != *p)
1221 				goto next; /* not match */
1222 			if (m == 0xff) /* short cut for a typical case */
1223 				matchlen += 8;
1224 			else {
1225 				while (m >= 0x80) {
1226 					matchlen++;
1227 					m <<= 1;
1228 				}
1229 			}
1230 		}
1231 
1232 		/* matched.  check if this is better than the current best. */
1233 		if (bestpol == NULL ||
1234 		    matchlen > bestmatchlen) {
1235 			bestpol = pol;
1236 			bestmatchlen = matchlen;
1237 		}
1238 
1239 	  next:
1240 		continue;
1241 	}
1242 
1243 	return (bestpol);
1244 }
1245