xref: /freebsd/sys/netinet6/in6_src.c (revision 48c5129f93c5eb5419c87b08e4677d51513f1dc0)
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/sx.h>
86 
87 #include <net/if.h>
88 #include <net/if_dl.h>
89 #include <net/route.h>
90 #include <net/if_llatbl.h>
91 #ifdef RADIX_MPATH
92 #include <net/radix_mpath.h>
93 #endif
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_pcb.h>
107 #include <netinet6/ip6_var.h>
108 #include <netinet6/scope6_var.h>
109 #include <netinet6/nd6.h>
110 
111 static struct mtx addrsel_lock;
112 #define	ADDRSEL_LOCK_INIT()	mtx_init(&addrsel_lock, "addrsel_lock", NULL, MTX_DEF)
113 #define	ADDRSEL_LOCK()		mtx_lock(&addrsel_lock)
114 #define	ADDRSEL_UNLOCK()	mtx_unlock(&addrsel_lock)
115 #define	ADDRSEL_LOCK_ASSERT()	mtx_assert(&addrsel_lock, MA_OWNED)
116 
117 static struct sx addrsel_sxlock;
118 #define	ADDRSEL_SXLOCK_INIT()	sx_init(&addrsel_sxlock, "addrsel_sxlock")
119 #define	ADDRSEL_SLOCK()		sx_slock(&addrsel_sxlock)
120 #define	ADDRSEL_SUNLOCK()	sx_sunlock(&addrsel_sxlock)
121 #define	ADDRSEL_XLOCK()		sx_xlock(&addrsel_sxlock)
122 #define	ADDRSEL_XUNLOCK()	sx_xunlock(&addrsel_sxlock)
123 
124 #define ADDR_LABEL_NOTAPP (-1)
125 static VNET_DEFINE(struct in6_addrpolicy, defaultaddrpolicy);
126 #define	V_defaultaddrpolicy		VNET(defaultaddrpolicy)
127 
128 VNET_DEFINE(int, ip6_prefer_tempaddr) = 0;
129 
130 static int selectroute(struct sockaddr_in6 *, struct ip6_pktopts *,
131 	struct ip6_moptions *, struct route_in6 *, struct ifnet **,
132 	struct rtentry **, int, u_int);
133 static int in6_selectif(struct sockaddr_in6 *, struct ip6_pktopts *,
134 	struct ip6_moptions *, struct route_in6 *ro, struct ifnet **,
135 	struct ifnet *, u_int);
136 
137 static struct in6_addrpolicy *lookup_addrsel_policy(struct sockaddr_in6 *);
138 
139 static void init_policy_queue(void);
140 static int add_addrsel_policyent(struct in6_addrpolicy *);
141 static int delete_addrsel_policyent(struct in6_addrpolicy *);
142 static int walk_addrsel_policy(int (*)(struct in6_addrpolicy *, void *),
143 	void *);
144 static int dump_addrsel_policyent(struct in6_addrpolicy *, void *);
145 static struct in6_addrpolicy *match_addrsel_policy(struct sockaddr_in6 *);
146 
147 /*
148  * Return an IPv6 address, which is the most appropriate for a given
149  * destination and user specified options.
150  * If necessary, this function lookups the routing table and returns
151  * an entry to the caller for later use.
152  */
153 #define REPLACE(r) do {\
154 	if ((r) < sizeof(V_ip6stat.ip6s_sources_rule) / \
155 		sizeof(V_ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
156 		V_ip6stat.ip6s_sources_rule[(r)]++; \
157 	/* { \
158 	char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
159 	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)); \
160 	} */ \
161 	goto replace; \
162 } while(0)
163 #define NEXT(r) do {\
164 	if ((r) < sizeof(V_ip6stat.ip6s_sources_rule) / \
165 		sizeof(V_ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
166 		V_ip6stat.ip6s_sources_rule[(r)]++; \
167 	/* { \
168 	char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
169 	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)); \
170 	} */ \
171 	goto next;		/* XXX: we can't use 'continue' here */ \
172 } while(0)
173 #define BREAK(r) do { \
174 	if ((r) < sizeof(V_ip6stat.ip6s_sources_rule) / \
175 		sizeof(V_ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
176 		V_ip6stat.ip6s_sources_rule[(r)]++; \
177 	goto out;		/* XXX: we can't use 'break' here */ \
178 } while(0)
179 
180 int
181 in6_selectsrc(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
182     struct inpcb *inp, struct route_in6 *ro, struct ucred *cred,
183     struct ifnet **ifpp, struct in6_addr *srcp)
184 {
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 		struct sockaddr_in6 srcsock;
228 		struct in6_ifaddr *ia6;
229 
230 		/* get the outgoing interface */
231 		if ((error = in6_selectif(dstsock, opts, mopts, ro, &ifp, oifp,
232 		    (inp != NULL) ? inp->inp_inc.inc_fibnum : RT_DEFAULT_FIB))
233 		    != 0)
234 			return (error);
235 
236 		/*
237 		 * determine the appropriate zone id of the source based on
238 		 * the zone of the destination and the outgoing interface.
239 		 * If the specified address is ambiguous wrt the scope zone,
240 		 * the interface must be specified; otherwise, ifa_ifwithaddr()
241 		 * will fail matching the address.
242 		 */
243 		bzero(&srcsock, sizeof(srcsock));
244 		srcsock.sin6_family = AF_INET6;
245 		srcsock.sin6_len = sizeof(srcsock);
246 		srcsock.sin6_addr = pi->ipi6_addr;
247 		if (ifp) {
248 			error = in6_setscope(&srcsock.sin6_addr, ifp, NULL);
249 			if (error)
250 				return (error);
251 		}
252 		if (cred != NULL && (error = prison_local_ip6(cred,
253 		    &srcsock.sin6_addr, (inp != NULL &&
254 		    (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
255 			return (error);
256 
257 		ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(
258 		    (struct sockaddr *)&srcsock);
259 		if (ia6 == NULL ||
260 		    (ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY))) {
261 			if (ia6 != NULL)
262 				ifa_free(&ia6->ia_ifa);
263 			return (EADDRNOTAVAIL);
264 		}
265 		pi->ipi6_addr = srcsock.sin6_addr; /* XXX: this overrides pi */
266 		if (ifpp)
267 			*ifpp = ifp;
268 		bcopy(&ia6->ia_addr.sin6_addr, srcp, sizeof(*srcp));
269 		ifa_free(&ia6->ia_ifa);
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, ro, &ifp, oifp,
298 	    (inp != NULL) ? inp->inp_inc.inc_fibnum : RT_DEFAULT_FIB)) != 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();
310 	TAILQ_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 (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp)
387 			NEXT(5);
388 		if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp)
389 			REPLACE(5);
390 
391 		/*
392 		 * Rule 6: Prefer matching label
393 		 * Note that best_policy should be non-NULL here.
394 		 */
395 		if (dst_policy == NULL)
396 			dst_policy = lookup_addrsel_policy(dstsock);
397 		if (dst_policy->label != ADDR_LABEL_NOTAPP) {
398 			new_policy = lookup_addrsel_policy(&ia->ia_addr);
399 			if (dst_policy->label == best_policy->label &&
400 			    dst_policy->label != new_policy->label)
401 				NEXT(6);
402 			if (dst_policy->label != best_policy->label &&
403 			    dst_policy->label == new_policy->label)
404 				REPLACE(6);
405 		}
406 
407 		/*
408 		 * Rule 7: Prefer public addresses.
409 		 * We allow users to reverse the logic by configuring
410 		 * a sysctl variable, so that privacy conscious users can
411 		 * always prefer temporary addresses.
412 		 */
413 		if (opts == NULL ||
414 		    opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) {
415 			prefer_tempaddr = V_ip6_prefer_tempaddr;
416 		} else if (opts->ip6po_prefer_tempaddr ==
417 		    IP6PO_TEMPADDR_NOTPREFER) {
418 			prefer_tempaddr = 0;
419 		} else
420 			prefer_tempaddr = 1;
421 		if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
422 		    (ia->ia6_flags & IN6_IFF_TEMPORARY)) {
423 			if (prefer_tempaddr)
424 				REPLACE(7);
425 			else
426 				NEXT(7);
427 		}
428 		if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
429 		    !(ia->ia6_flags & IN6_IFF_TEMPORARY)) {
430 			if (prefer_tempaddr)
431 				NEXT(7);
432 			else
433 				REPLACE(7);
434 		}
435 
436 		/*
437 		 * Rule 8: prefer addresses on alive interfaces.
438 		 * This is a KAME specific rule.
439 		 */
440 		if ((ia_best->ia_ifp->if_flags & IFF_UP) &&
441 		    !(ia->ia_ifp->if_flags & IFF_UP))
442 			NEXT(8);
443 		if (!(ia_best->ia_ifp->if_flags & IFF_UP) &&
444 		    (ia->ia_ifp->if_flags & IFF_UP))
445 			REPLACE(8);
446 
447 		/*
448 		 * Rule 14: Use longest matching prefix.
449 		 * Note: in the address selection draft, this rule is
450 		 * documented as "Rule 8".  However, since it is also
451 		 * documented that this rule can be overridden, we assign
452 		 * a large number so that it is easy to assign smaller numbers
453 		 * to more preferred rules.
454 		 */
455 		new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst);
456 		if (best_matchlen < new_matchlen)
457 			REPLACE(14);
458 		if (new_matchlen < best_matchlen)
459 			NEXT(14);
460 
461 		/* Rule 15 is reserved. */
462 
463 		/*
464 		 * Last resort: just keep the current candidate.
465 		 * Or, do we need more rules?
466 		 */
467 		continue;
468 
469 	  replace:
470 		ia_best = ia;
471 		best_scope = (new_scope >= 0 ? new_scope :
472 			      in6_addrscope(&ia_best->ia_addr.sin6_addr));
473 		best_policy = (new_policy ? new_policy :
474 			       lookup_addrsel_policy(&ia_best->ia_addr));
475 		best_matchlen = (new_matchlen >= 0 ? new_matchlen :
476 				 in6_matchlen(&ia_best->ia_addr.sin6_addr,
477 					      &dst));
478 
479 	  next:
480 		continue;
481 
482 	  out:
483 		break;
484 	}
485 
486 	if ((ia = ia_best) == NULL) {
487 		IN6_IFADDR_RUNLOCK();
488 		return (EADDRNOTAVAIL);
489 	}
490 
491 	/*
492 	 * At this point at least one of the addresses belonged to the jail
493 	 * but it could still be, that we want to further restrict it, e.g.
494 	 * theoratically IN6_IS_ADDR_LOOPBACK.
495 	 * It must not be IN6_IS_ADDR_UNSPECIFIED anymore.
496 	 * prison_local_ip6() will fix an IN6_IS_ADDR_LOOPBACK but should
497 	 * let all others previously selected pass.
498 	 * Use tmp to not change ::1 on lo0 to the primary jail address.
499 	 */
500 	tmp = ia->ia_addr.sin6_addr;
501 	if (cred != NULL && prison_local_ip6(cred, &tmp, (inp != NULL &&
502 	    (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0) {
503 		IN6_IFADDR_RUNLOCK();
504 		return (EADDRNOTAVAIL);
505 	}
506 
507 	if (ifpp)
508 		*ifpp = ifp;
509 
510 	bcopy(&tmp, srcp, sizeof(*srcp));
511 	IN6_IFADDR_RUNLOCK();
512 	return (0);
513 }
514 
515 /*
516  * clone - meaningful only for bsdi and freebsd
517  */
518 static int
519 selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
520     struct ip6_moptions *mopts, struct route_in6 *ro,
521     struct ifnet **retifp, struct rtentry **retrt, int norouteok, u_int fibnum)
522 {
523 	int error = 0;
524 	struct ifnet *ifp = NULL;
525 	struct rtentry *rt = NULL;
526 	struct sockaddr_in6 *sin6_next;
527 	struct in6_pktinfo *pi = NULL;
528 	struct in6_addr *dst = &dstsock->sin6_addr;
529 #if 0
530 	char ip6buf[INET6_ADDRSTRLEN];
531 
532 	if (dstsock->sin6_addr.s6_addr32[0] == 0 &&
533 	    dstsock->sin6_addr.s6_addr32[1] == 0 &&
534 	    !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) {
535 		printf("in6_selectroute: strange destination %s\n",
536 		       ip6_sprintf(ip6buf, &dstsock->sin6_addr));
537 	} else {
538 		printf("in6_selectroute: destination = %s%%%d\n",
539 		       ip6_sprintf(ip6buf, &dstsock->sin6_addr),
540 		       dstsock->sin6_scope_id); /* for debug */
541 	}
542 #endif
543 
544 	/* If the caller specify the outgoing interface explicitly, use it. */
545 	if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) {
546 		/* XXX boundary check is assumed to be already done. */
547 		ifp = ifnet_byindex(pi->ipi6_ifindex);
548 		if (ifp != NULL &&
549 		    (norouteok || retrt == NULL ||
550 		    IN6_IS_ADDR_MULTICAST(dst))) {
551 			/*
552 			 * we do not have to check or get the route for
553 			 * multicast.
554 			 */
555 			goto done;
556 		} else
557 			goto getroute;
558 	}
559 
560 	/*
561 	 * If the destination address is a multicast address and the outgoing
562 	 * interface for the address is specified by the caller, use it.
563 	 */
564 	if (IN6_IS_ADDR_MULTICAST(dst) &&
565 	    mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) {
566 		goto done; /* we do not need a route for multicast. */
567 	}
568 
569   getroute:
570 	/*
571 	 * If the next hop address for the packet is specified by the caller,
572 	 * use it as the gateway.
573 	 */
574 	if (opts && opts->ip6po_nexthop) {
575 		struct route_in6 *ron;
576 		struct llentry *la;
577 
578 		sin6_next = satosin6(opts->ip6po_nexthop);
579 
580 		/* at this moment, we only support AF_INET6 next hops */
581 		if (sin6_next->sin6_family != AF_INET6) {
582 			error = EAFNOSUPPORT; /* or should we proceed? */
583 			goto done;
584 		}
585 
586 		/*
587 		 * If the next hop is an IPv6 address, then the node identified
588 		 * by that address must be a neighbor of the sending host.
589 		 */
590 		ron = &opts->ip6po_nextroute;
591 		/*
592 		 * XXX what do we do here?
593 		 * PLZ to be fixing
594 		 */
595 
596 
597 		if (ron->ro_rt == NULL) {
598 			in6_rtalloc(ron, fibnum); /* multi path case? */
599 			if (ron->ro_rt == NULL) {
600 				/* XXX-BZ WT.? */
601 				if (ron->ro_rt) {
602 					RTFREE(ron->ro_rt);
603 					ron->ro_rt = NULL;
604 				}
605 				error = EHOSTUNREACH;
606 				goto done;
607 			}
608 		}
609 
610 		rt = ron->ro_rt;
611 		ifp = rt->rt_ifp;
612 		IF_AFDATA_LOCK(ifp);
613 		la = lla_lookup(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6_next->sin6_addr);
614 		IF_AFDATA_UNLOCK(ifp);
615 		if (la != NULL)
616 			LLE_RUNLOCK(la);
617 		else {
618 			error = EHOSTUNREACH;
619 			goto done;
620 		}
621 #if 0
622 		if ((ron->ro_rt &&
623 		     (ron->ro_rt->rt_flags & (RTF_UP | RTF_LLINFO)) !=
624 		     (RTF_UP | RTF_LLINFO)) ||
625 		    !IN6_ARE_ADDR_EQUAL(&satosin6(&ron->ro_dst)->sin6_addr,
626 		    &sin6_next->sin6_addr)) {
627 			if (ron->ro_rt) {
628 				RTFREE(ron->ro_rt);
629 				ron->ro_rt = NULL;
630 			}
631 			*satosin6(&ron->ro_dst) = *sin6_next;
632 		}
633 		if (ron->ro_rt == NULL) {
634 			in6_rtalloc(ron, fibnum); /* multi path case? */
635 			if (ron->ro_rt == NULL ||
636 			    !(ron->ro_rt->rt_flags & RTF_LLINFO)) {
637 				if (ron->ro_rt) {
638 					RTFREE(ron->ro_rt);
639 					ron->ro_rt = NULL;
640 				}
641 				error = EHOSTUNREACH;
642 				goto done;
643 			}
644 		}
645 #endif
646 
647 		/*
648 		 * When cloning is required, try to allocate a route to the
649 		 * destination so that the caller can store path MTU
650 		 * information.
651 		 */
652 		goto done;
653 	}
654 
655 	/*
656 	 * Use a cached route if it exists and is valid, else try to allocate
657 	 * a new one.  Note that we should check the address family of the
658 	 * cached destination, in case of sharing the cache with IPv4.
659 	 */
660 	if (ro) {
661 		if (ro->ro_rt &&
662 		    (!(ro->ro_rt->rt_flags & RTF_UP) ||
663 		     ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 ||
664 		     !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr,
665 		     dst))) {
666 			RTFREE(ro->ro_rt);
667 			ro->ro_rt = (struct rtentry *)NULL;
668 		}
669 		if (ro->ro_rt == (struct rtentry *)NULL) {
670 			struct sockaddr_in6 *sa6;
671 
672 			/* No route yet, so try to acquire one */
673 			bzero(&ro->ro_dst, sizeof(struct sockaddr_in6));
674 			sa6 = (struct sockaddr_in6 *)&ro->ro_dst;
675 			*sa6 = *dstsock;
676 			sa6->sin6_scope_id = 0;
677 
678 #ifdef RADIX_MPATH
679 				rtalloc_mpath_fib((struct route *)ro,
680 				    ntohl(sa6->sin6_addr.s6_addr32[3]), fibnum);
681 #else
682 				ro->ro_rt = in6_rtalloc1((struct sockaddr *)
683 				    &ro->ro_dst, 0, 0UL, fibnum);
684 				if (ro->ro_rt)
685 					RT_UNLOCK(ro->ro_rt);
686 #endif
687 		}
688 
689 		/*
690 		 * do not care about the result if we have the nexthop
691 		 * explicitly specified.
692 		 */
693 		if (opts && opts->ip6po_nexthop)
694 			goto done;
695 
696 		if (ro->ro_rt) {
697 			ifp = ro->ro_rt->rt_ifp;
698 
699 			if (ifp == NULL) { /* can this really happen? */
700 				RTFREE(ro->ro_rt);
701 				ro->ro_rt = NULL;
702 			}
703 		}
704 		if (ro->ro_rt == NULL)
705 			error = EHOSTUNREACH;
706 		rt = ro->ro_rt;
707 
708 		/*
709 		 * Check if the outgoing interface conflicts with
710 		 * the interface specified by ipi6_ifindex (if specified).
711 		 * Note that loopback interface is always okay.
712 		 * (this may happen when we are sending a packet to one of
713 		 *  our own addresses.)
714 		 */
715 		if (ifp && opts && opts->ip6po_pktinfo &&
716 		    opts->ip6po_pktinfo->ipi6_ifindex) {
717 			if (!(ifp->if_flags & IFF_LOOPBACK) &&
718 			    ifp->if_index !=
719 			    opts->ip6po_pktinfo->ipi6_ifindex) {
720 				error = EHOSTUNREACH;
721 				goto done;
722 			}
723 		}
724 	}
725 
726   done:
727 	if (ifp == NULL && rt == NULL) {
728 		/*
729 		 * This can happen if the caller did not pass a cached route
730 		 * nor any other hints.  We treat this case an error.
731 		 */
732 		error = EHOSTUNREACH;
733 	}
734 	if (error == EHOSTUNREACH)
735 		V_ip6stat.ip6s_noroute++;
736 
737 	if (retifp != NULL) {
738 		*retifp = ifp;
739 
740 		/*
741 		 * Adjust the "outgoing" interface.  If we're going to loop
742 		 * the packet back to ourselves, the ifp would be the loopback
743 		 * interface. However, we'd rather know the interface associated
744 		 * to the destination address (which should probably be one of
745 		 * our own addresses.)
746 		 */
747 		if (rt) {
748 			if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) &&
749 			    (rt->rt_gateway->sa_family == AF_LINK))
750 				*retifp =
751 					ifnet_byindex(((struct sockaddr_dl *)
752 						       rt->rt_gateway)->sdl_index);
753 		}
754 	}
755 
756 	if (retrt != NULL)
757 		*retrt = rt;	/* rt may be NULL */
758 
759 	return (error);
760 }
761 
762 static int
763 in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
764     struct ip6_moptions *mopts, struct route_in6 *ro, struct ifnet **retifp,
765     struct ifnet *oifp, u_int fibnum)
766 {
767 	int error;
768 	struct route_in6 sro;
769 	struct rtentry *rt = NULL;
770 
771 	KASSERT(retifp != NULL, ("%s: retifp is NULL", __func__));
772 
773 	if (ro == NULL) {
774 		bzero(&sro, sizeof(sro));
775 		ro = &sro;
776 	}
777 
778 	if ((error = selectroute(dstsock, opts, mopts, ro, retifp,
779 	    &rt, 1, fibnum)) != 0) {
780 		if (ro == &sro && rt && rt == sro.ro_rt)
781 			RTFREE(rt);
782 		/* Help ND. See oifp comment in in6_selectsrc(). */
783 		if (oifp != NULL && fibnum == RT_DEFAULT_FIB) {
784 			*retifp = oifp;
785 			error = 0;
786 		}
787 		return (error);
788 	}
789 
790 	/*
791 	 * do not use a rejected or black hole route.
792 	 * XXX: this check should be done in the L2 output routine.
793 	 * However, if we skipped this check here, we'd see the following
794 	 * scenario:
795 	 * - install a rejected route for a scoped address prefix
796 	 *   (like fe80::/10)
797 	 * - send a packet to a destination that matches the scoped prefix,
798 	 *   with ambiguity about the scope zone.
799 	 * - pick the outgoing interface from the route, and disambiguate the
800 	 *   scope zone with the interface.
801 	 * - ip6_output() would try to get another route with the "new"
802 	 *   destination, which may be valid.
803 	 * - we'd see no error on output.
804 	 * Although this may not be very harmful, it should still be confusing.
805 	 * We thus reject the case here.
806 	 */
807 	if (rt && (rt->rt_flags & (RTF_REJECT | RTF_BLACKHOLE))) {
808 		int flags = (rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
809 
810 		if (ro == &sro && rt && rt == sro.ro_rt)
811 			RTFREE(rt);
812 		return (flags);
813 	}
814 
815 	if (ro == &sro && rt && rt == sro.ro_rt)
816 		RTFREE(rt);
817 	return (0);
818 }
819 
820 /*
821  * Public wrapper function to selectroute().
822  *
823  * XXX-BZ in6_selectroute() should and will grow the FIB argument. The
824  * in6_selectroute_fib() function is only there for backward compat on stable.
825  */
826 int
827 in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
828     struct ip6_moptions *mopts, struct route_in6 *ro,
829     struct ifnet **retifp, struct rtentry **retrt)
830 {
831 
832 	return (selectroute(dstsock, opts, mopts, ro, retifp,
833 	    retrt, 0, RT_DEFAULT_FIB));
834 }
835 
836 #ifndef BURN_BRIDGES
837 int
838 in6_selectroute_fib(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
839     struct ip6_moptions *mopts, struct route_in6 *ro,
840     struct ifnet **retifp, struct rtentry **retrt, u_int fibnum)
841 {
842 
843 	return (selectroute(dstsock, opts, mopts, ro, retifp,
844 	    retrt, 0, fibnum));
845 }
846 #endif
847 
848 /*
849  * Default hop limit selection. The precedence is as follows:
850  * 1. Hoplimit value specified via ioctl.
851  * 2. (If the outgoing interface is detected) the current
852  *     hop limit of the interface specified by router advertisement.
853  * 3. The system default hoplimit.
854  */
855 int
856 in6_selecthlim(struct inpcb *in6p, struct ifnet *ifp)
857 {
858 
859 	if (in6p && in6p->in6p_hops >= 0)
860 		return (in6p->in6p_hops);
861 	else if (ifp)
862 		return (ND_IFINFO(ifp)->chlim);
863 	else if (in6p && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
864 		struct route_in6 ro6;
865 		struct ifnet *lifp;
866 
867 		bzero(&ro6, sizeof(ro6));
868 		ro6.ro_dst.sin6_family = AF_INET6;
869 		ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
870 		ro6.ro_dst.sin6_addr = in6p->in6p_faddr;
871 		in6_rtalloc(&ro6, in6p->inp_inc.inc_fibnum);
872 		if (ro6.ro_rt) {
873 			lifp = ro6.ro_rt->rt_ifp;
874 			RTFREE(ro6.ro_rt);
875 			if (lifp)
876 				return (ND_IFINFO(lifp)->chlim);
877 		}
878 	}
879 	return (V_ip6_defhlim);
880 }
881 
882 /*
883  * XXX: this is borrowed from in6_pcbbind(). If possible, we should
884  * share this function by all *bsd*...
885  */
886 int
887 in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct ucred *cred)
888 {
889 	struct socket *so = inp->inp_socket;
890 	u_int16_t lport = 0;
891 	int error, lookupflags = 0;
892 #ifdef INVARIANTS
893 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
894 #endif
895 
896 	INP_WLOCK_ASSERT(inp);
897 	INP_HASH_WLOCK_ASSERT(pcbinfo);
898 
899 	error = prison_local_ip6(cred, laddr,
900 	    ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0));
901 	if (error)
902 		return(error);
903 
904 	/* XXX: this is redundant when called from in6_pcbbind */
905 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
906 		lookupflags = INPLOOKUP_WILDCARD;
907 
908 	inp->inp_flags |= INP_ANONPORT;
909 
910 	error = in_pcb_lport(inp, NULL, &lport, cred, lookupflags);
911 	if (error != 0)
912 		return (error);
913 
914 	inp->inp_lport = lport;
915 	if (in_pcbinshash(inp) != 0) {
916 		inp->in6p_laddr = in6addr_any;
917 		inp->inp_lport = 0;
918 		return (EAGAIN);
919 	}
920 
921 	return (0);
922 }
923 
924 void
925 addrsel_policy_init(void)
926 {
927 
928 	init_policy_queue();
929 
930 	/* initialize the "last resort" policy */
931 	bzero(&V_defaultaddrpolicy, sizeof(V_defaultaddrpolicy));
932 	V_defaultaddrpolicy.label = ADDR_LABEL_NOTAPP;
933 
934 	if (!IS_DEFAULT_VNET(curvnet))
935 		return;
936 
937 	ADDRSEL_LOCK_INIT();
938 	ADDRSEL_SXLOCK_INIT();
939 }
940 
941 static struct in6_addrpolicy *
942 lookup_addrsel_policy(struct sockaddr_in6 *key)
943 {
944 	struct in6_addrpolicy *match = NULL;
945 
946 	ADDRSEL_LOCK();
947 	match = match_addrsel_policy(key);
948 
949 	if (match == NULL)
950 		match = &V_defaultaddrpolicy;
951 	else
952 		match->use++;
953 	ADDRSEL_UNLOCK();
954 
955 	return (match);
956 }
957 
958 /*
959  * Subroutines to manage the address selection policy table via sysctl.
960  */
961 struct walkarg {
962 	struct sysctl_req *w_req;
963 };
964 
965 static int in6_src_sysctl(SYSCTL_HANDLER_ARGS);
966 SYSCTL_DECL(_net_inet6_ip6);
967 static SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy,
968 	CTLFLAG_RD, in6_src_sysctl, "");
969 
970 static int
971 in6_src_sysctl(SYSCTL_HANDLER_ARGS)
972 {
973 	struct walkarg w;
974 
975 	if (req->newptr)
976 		return EPERM;
977 
978 	bzero(&w, sizeof(w));
979 	w.w_req = req;
980 
981 	return (walk_addrsel_policy(dump_addrsel_policyent, &w));
982 }
983 
984 int
985 in6_src_ioctl(u_long cmd, caddr_t data)
986 {
987 	int i;
988 	struct in6_addrpolicy ent0;
989 
990 	if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY)
991 		return (EOPNOTSUPP); /* check for safety */
992 
993 	ent0 = *(struct in6_addrpolicy *)data;
994 
995 	if (ent0.label == ADDR_LABEL_NOTAPP)
996 		return (EINVAL);
997 	/* check if the prefix mask is consecutive. */
998 	if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0)
999 		return (EINVAL);
1000 	/* clear trailing garbages (if any) of the prefix address. */
1001 	for (i = 0; i < 4; i++) {
1002 		ent0.addr.sin6_addr.s6_addr32[i] &=
1003 			ent0.addrmask.sin6_addr.s6_addr32[i];
1004 	}
1005 	ent0.use = 0;
1006 
1007 	switch (cmd) {
1008 	case SIOCAADDRCTL_POLICY:
1009 		return (add_addrsel_policyent(&ent0));
1010 	case SIOCDADDRCTL_POLICY:
1011 		return (delete_addrsel_policyent(&ent0));
1012 	}
1013 
1014 	return (0);		/* XXX: compromise compilers */
1015 }
1016 
1017 /*
1018  * The followings are implementation of the policy table using a
1019  * simple tail queue.
1020  * XXX such details should be hidden.
1021  * XXX implementation using binary tree should be more efficient.
1022  */
1023 struct addrsel_policyent {
1024 	TAILQ_ENTRY(addrsel_policyent) ape_entry;
1025 	struct in6_addrpolicy ape_policy;
1026 };
1027 
1028 TAILQ_HEAD(addrsel_policyhead, addrsel_policyent);
1029 
1030 static VNET_DEFINE(struct addrsel_policyhead, addrsel_policytab);
1031 #define	V_addrsel_policytab		VNET(addrsel_policytab)
1032 
1033 static void
1034 init_policy_queue(void)
1035 {
1036 
1037 	TAILQ_INIT(&V_addrsel_policytab);
1038 }
1039 
1040 static int
1041 add_addrsel_policyent(struct in6_addrpolicy *newpolicy)
1042 {
1043 	struct addrsel_policyent *new, *pol;
1044 
1045 	new = malloc(sizeof(*new), M_IFADDR,
1046 	       M_WAITOK);
1047 	ADDRSEL_XLOCK();
1048 	ADDRSEL_LOCK();
1049 
1050 	/* duplication check */
1051 	TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1052 		if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr,
1053 				       &pol->ape_policy.addr.sin6_addr) &&
1054 		    IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr,
1055 				       &pol->ape_policy.addrmask.sin6_addr)) {
1056 			ADDRSEL_UNLOCK();
1057 			ADDRSEL_XUNLOCK();
1058 			free(new, M_IFADDR);
1059 			return (EEXIST);	/* or override it? */
1060 		}
1061 	}
1062 
1063 	bzero(new, sizeof(*new));
1064 
1065 	/* XXX: should validate entry */
1066 	new->ape_policy = *newpolicy;
1067 
1068 	TAILQ_INSERT_TAIL(&V_addrsel_policytab, new, ape_entry);
1069 	ADDRSEL_UNLOCK();
1070 	ADDRSEL_XUNLOCK();
1071 
1072 	return (0);
1073 }
1074 
1075 static int
1076 delete_addrsel_policyent(struct in6_addrpolicy *key)
1077 {
1078 	struct addrsel_policyent *pol;
1079 
1080 	ADDRSEL_XLOCK();
1081 	ADDRSEL_LOCK();
1082 
1083 	/* search for the entry in the table */
1084 	TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1085 		if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr,
1086 		    &pol->ape_policy.addr.sin6_addr) &&
1087 		    IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr,
1088 		    &pol->ape_policy.addrmask.sin6_addr)) {
1089 			break;
1090 		}
1091 	}
1092 	if (pol == NULL) {
1093 		ADDRSEL_UNLOCK();
1094 		ADDRSEL_XUNLOCK();
1095 		return (ESRCH);
1096 	}
1097 
1098 	TAILQ_REMOVE(&V_addrsel_policytab, pol, ape_entry);
1099 	ADDRSEL_UNLOCK();
1100 	ADDRSEL_XUNLOCK();
1101 
1102 	return (0);
1103 }
1104 
1105 static int
1106 walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *), void *w)
1107 {
1108 	struct addrsel_policyent *pol;
1109 	int error = 0;
1110 
1111 	ADDRSEL_SLOCK();
1112 	TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1113 		if ((error = (*callback)(&pol->ape_policy, w)) != 0) {
1114 			ADDRSEL_SUNLOCK();
1115 			return (error);
1116 		}
1117 	}
1118 	ADDRSEL_SUNLOCK();
1119 	return (error);
1120 }
1121 
1122 static int
1123 dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg)
1124 {
1125 	int error = 0;
1126 	struct walkarg *w = arg;
1127 
1128 	error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol));
1129 
1130 	return (error);
1131 }
1132 
1133 static struct in6_addrpolicy *
1134 match_addrsel_policy(struct sockaddr_in6 *key)
1135 {
1136 	struct addrsel_policyent *pent;
1137 	struct in6_addrpolicy *bestpol = NULL, *pol;
1138 	int matchlen, bestmatchlen = -1;
1139 	u_char *mp, *ep, *k, *p, m;
1140 
1141 	TAILQ_FOREACH(pent, &V_addrsel_policytab, ape_entry) {
1142 		matchlen = 0;
1143 
1144 		pol = &pent->ape_policy;
1145 		mp = (u_char *)&pol->addrmask.sin6_addr;
1146 		ep = mp + 16;	/* XXX: scope field? */
1147 		k = (u_char *)&key->sin6_addr;
1148 		p = (u_char *)&pol->addr.sin6_addr;
1149 		for (; mp < ep && *mp; mp++, k++, p++) {
1150 			m = *mp;
1151 			if ((*k & m) != *p)
1152 				goto next; /* not match */
1153 			if (m == 0xff) /* short cut for a typical case */
1154 				matchlen += 8;
1155 			else {
1156 				while (m >= 0x80) {
1157 					matchlen++;
1158 					m <<= 1;
1159 				}
1160 			}
1161 		}
1162 
1163 		/* matched.  check if this is better than the current best. */
1164 		if (bestpol == NULL ||
1165 		    matchlen > bestmatchlen) {
1166 			bestpol = pol;
1167 			bestmatchlen = matchlen;
1168 		}
1169 
1170 	  next:
1171 		continue;
1172 	}
1173 
1174 	return (bestpol);
1175 }
1176