xref: /freebsd/sys/netinet6/in6_src.c (revision 9a14aa017b21c292740c00ee098195cd46642730)
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 __P((struct sockaddr_in6 *, struct ip6_pktopts *,
131 	struct ip6_moptions *, struct route_in6 *, struct ifnet **,
132 	struct rtentry **, int));
133 static int in6_selectif __P((struct sockaddr_in6 *, struct ip6_pktopts *,
134 	struct ip6_moptions *, struct route_in6 *ro, struct ifnet **));
135 
136 static struct in6_addrpolicy *lookup_addrsel_policy(struct sockaddr_in6 *);
137 
138 static void init_policy_queue(void);
139 static int add_addrsel_policyent(struct in6_addrpolicy *);
140 static int delete_addrsel_policyent(struct in6_addrpolicy *);
141 static int walk_addrsel_policy __P((int (*)(struct in6_addrpolicy *, void *),
142 				    void *));
143 static int dump_addrsel_policyent(struct in6_addrpolicy *, void *);
144 static struct in6_addrpolicy *match_addrsel_policy(struct sockaddr_in6 *);
145 
146 /*
147  * Return an IPv6 address, which is the most appropriate for a given
148  * destination and user specified options.
149  * If necessary, this function lookups the routing table and returns
150  * an entry to the caller for later use.
151  */
152 #define REPLACE(r) do {\
153 	if ((r) < sizeof(V_ip6stat.ip6s_sources_rule) / \
154 		sizeof(V_ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
155 		V_ip6stat.ip6s_sources_rule[(r)]++; \
156 	/* { \
157 	char ip6buf[INET6_ADDRSTRLEN], ip6b[INET6_ADDRSTRLEN]; \
158 	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)); \
159 	} */ \
160 	goto replace; \
161 } while(0)
162 #define NEXT(r) do {\
163 	if ((r) < sizeof(V_ip6stat.ip6s_sources_rule) / \
164 		sizeof(V_ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
165 		V_ip6stat.ip6s_sources_rule[(r)]++; \
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 	if ((r) < sizeof(V_ip6stat.ip6s_sources_rule) / \
174 		sizeof(V_ip6stat.ip6s_sources_rule[0])) /* check for safety */ \
175 		V_ip6stat.ip6s_sources_rule[(r)]++; \
176 	goto out;		/* XXX: we can't use 'break' here */ \
177 } while(0)
178 
179 int
180 in6_selectsrc(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
181     struct inpcb *inp, struct route_in6 *ro, struct ucred *cred,
182     struct ifnet **ifpp, struct in6_addr *srcp)
183 {
184 	struct in6_addr dst, tmp;
185 	struct ifnet *ifp = NULL;
186 	struct in6_ifaddr *ia = NULL, *ia_best = NULL;
187 	struct in6_pktinfo *pi = NULL;
188 	int dst_scope = -1, best_scope = -1, best_matchlen = -1;
189 	struct in6_addrpolicy *dst_policy = NULL, *best_policy = NULL;
190 	u_int32_t odstzone;
191 	int prefer_tempaddr;
192 	int error;
193 	struct ip6_moptions *mopts;
194 
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 		*ifpp = NULL;
200 
201 	if (inp != NULL) {
202 		INP_LOCK_ASSERT(inp);
203 		mopts = inp->in6p_moptions;
204 	} else {
205 		mopts = NULL;
206 	}
207 
208 	/*
209 	 * If the source address is explicitly specified by the caller,
210 	 * check if the requested source address is indeed a unicast address
211 	 * assigned to the node, and can be used as the packet's source
212 	 * address.  If everything is okay, use the address as source.
213 	 */
214 	if (opts && (pi = opts->ip6po_pktinfo) &&
215 	    !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) {
216 		struct sockaddr_in6 srcsock;
217 		struct in6_ifaddr *ia6;
218 
219 		/* get the outgoing interface */
220 		if ((error = in6_selectif(dstsock, opts, mopts, ro, &ifp)) != 0)
221 			return (error);
222 
223 		/*
224 		 * determine the appropriate zone id of the source based on
225 		 * the zone of the destination and the outgoing interface.
226 		 * If the specified address is ambiguous wrt the scope zone,
227 		 * the interface must be specified; otherwise, ifa_ifwithaddr()
228 		 * will fail matching the address.
229 		 */
230 		bzero(&srcsock, sizeof(srcsock));
231 		srcsock.sin6_family = AF_INET6;
232 		srcsock.sin6_len = sizeof(srcsock);
233 		srcsock.sin6_addr = pi->ipi6_addr;
234 		if (ifp) {
235 			error = in6_setscope(&srcsock.sin6_addr, ifp, NULL);
236 			if (error)
237 				return (error);
238 		}
239 		if (cred != NULL && (error = prison_local_ip6(cred,
240 		    &srcsock.sin6_addr, (inp != NULL &&
241 		    (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
242 			return (error);
243 
244 		ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(
245 		    (struct sockaddr *)&srcsock);
246 		if (ia6 == NULL ||
247 		    (ia6->ia6_flags & (IN6_IFF_ANYCAST | IN6_IFF_NOTREADY))) {
248 			if (ia6 != NULL)
249 				ifa_free(&ia6->ia_ifa);
250 			return (EADDRNOTAVAIL);
251 		}
252 		pi->ipi6_addr = srcsock.sin6_addr; /* XXX: this overrides pi */
253 		if (ifpp)
254 			*ifpp = ifp;
255 		bcopy(&ia6->ia_addr.sin6_addr, srcp, sizeof(*srcp));
256 		ifa_free(&ia6->ia_ifa);
257 		return (0);
258 	}
259 
260 	/*
261 	 * Otherwise, if the socket has already bound the source, just use it.
262 	 */
263 	if (inp != NULL && !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
264 		if (cred != NULL &&
265 		    (error = prison_local_ip6(cred, &inp->in6p_laddr,
266 		    ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
267 			return (error);
268 		bcopy(&inp->in6p_laddr, srcp, sizeof(*srcp));
269 		return (0);
270 	}
271 
272 	/*
273 	 * Bypass source address selection and use the primary jail IP
274 	 * if requested.
275 	 */
276 	if (cred != NULL && !prison_saddrsel_ip6(cred, srcp))
277 		return (0);
278 
279 	/*
280 	 * If the address is not specified, choose the best one based on
281 	 * the outgoing interface and the destination address.
282 	 */
283 	/* get the outgoing interface */
284 	if ((error = in6_selectif(dstsock, opts, mopts, ro, &ifp)) != 0)
285 		return (error);
286 
287 #ifdef DIAGNOSTIC
288 	if (ifp == NULL)	/* this should not happen */
289 		panic("in6_selectsrc: NULL ifp");
290 #endif
291 	error = in6_setscope(&dst, ifp, &odstzone);
292 	if (error)
293 		return (error);
294 
295 	IN6_IFADDR_RLOCK();
296 	TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) {
297 		int new_scope = -1, new_matchlen = -1;
298 		struct in6_addrpolicy *new_policy = NULL;
299 		u_int32_t srczone, osrczone, dstzone;
300 		struct in6_addr src;
301 		struct ifnet *ifp1 = ia->ia_ifp;
302 
303 		/*
304 		 * We'll never take an address that breaks the scope zone
305 		 * of the destination.  We also skip an address if its zone
306 		 * does not contain the outgoing interface.
307 		 * XXX: we should probably use sin6_scope_id here.
308 		 */
309 		if (in6_setscope(&dst, ifp1, &dstzone) ||
310 		    odstzone != dstzone) {
311 			continue;
312 		}
313 		src = ia->ia_addr.sin6_addr;
314 		if (in6_setscope(&src, ifp, &osrczone) ||
315 		    in6_setscope(&src, ifp1, &srczone) ||
316 		    osrczone != srczone) {
317 			continue;
318 		}
319 
320 		/* avoid unusable addresses */
321 		if ((ia->ia6_flags &
322 		     (IN6_IFF_NOTREADY | IN6_IFF_ANYCAST | IN6_IFF_DETACHED))) {
323 				continue;
324 		}
325 		if (!V_ip6_use_deprecated && IFA6_IS_DEPRECATED(ia))
326 			continue;
327 
328 		/* If jailed only take addresses of the jail into account. */
329 		if (cred != NULL &&
330 		    prison_check_ip6(cred, &ia->ia_addr.sin6_addr) != 0)
331 			continue;
332 
333 		/* Rule 1: Prefer same address */
334 		if (IN6_ARE_ADDR_EQUAL(&dst, &ia->ia_addr.sin6_addr)) {
335 			ia_best = ia;
336 			BREAK(1); /* there should be no better candidate */
337 		}
338 
339 		if (ia_best == NULL)
340 			REPLACE(0);
341 
342 		/* Rule 2: Prefer appropriate scope */
343 		if (dst_scope < 0)
344 			dst_scope = in6_addrscope(&dst);
345 		new_scope = in6_addrscope(&ia->ia_addr.sin6_addr);
346 		if (IN6_ARE_SCOPE_CMP(best_scope, new_scope) < 0) {
347 			if (IN6_ARE_SCOPE_CMP(best_scope, dst_scope) < 0)
348 				REPLACE(2);
349 			NEXT(2);
350 		} else if (IN6_ARE_SCOPE_CMP(new_scope, best_scope) < 0) {
351 			if (IN6_ARE_SCOPE_CMP(new_scope, dst_scope) < 0)
352 				NEXT(2);
353 			REPLACE(2);
354 		}
355 
356 		/*
357 		 * Rule 3: Avoid deprecated addresses.  Note that the case of
358 		 * !ip6_use_deprecated is already rejected above.
359 		 */
360 		if (!IFA6_IS_DEPRECATED(ia_best) && IFA6_IS_DEPRECATED(ia))
361 			NEXT(3);
362 		if (IFA6_IS_DEPRECATED(ia_best) && !IFA6_IS_DEPRECATED(ia))
363 			REPLACE(3);
364 
365 		/* Rule 4: Prefer home addresses */
366 		/*
367 		 * XXX: This is a TODO.  We should probably merge the MIP6
368 		 * case above.
369 		 */
370 
371 		/* Rule 5: Prefer outgoing interface */
372 		if (ia_best->ia_ifp == ifp && ia->ia_ifp != ifp)
373 			NEXT(5);
374 		if (ia_best->ia_ifp != ifp && ia->ia_ifp == ifp)
375 			REPLACE(5);
376 
377 		/*
378 		 * Rule 6: Prefer matching label
379 		 * Note that best_policy should be non-NULL here.
380 		 */
381 		if (dst_policy == NULL)
382 			dst_policy = lookup_addrsel_policy(dstsock);
383 		if (dst_policy->label != ADDR_LABEL_NOTAPP) {
384 			new_policy = lookup_addrsel_policy(&ia->ia_addr);
385 			if (dst_policy->label == best_policy->label &&
386 			    dst_policy->label != new_policy->label)
387 				NEXT(6);
388 			if (dst_policy->label != best_policy->label &&
389 			    dst_policy->label == new_policy->label)
390 				REPLACE(6);
391 		}
392 
393 		/*
394 		 * Rule 7: Prefer public addresses.
395 		 * We allow users to reverse the logic by configuring
396 		 * a sysctl variable, so that privacy conscious users can
397 		 * always prefer temporary addresses.
398 		 */
399 		if (opts == NULL ||
400 		    opts->ip6po_prefer_tempaddr == IP6PO_TEMPADDR_SYSTEM) {
401 			prefer_tempaddr = V_ip6_prefer_tempaddr;
402 		} else if (opts->ip6po_prefer_tempaddr ==
403 		    IP6PO_TEMPADDR_NOTPREFER) {
404 			prefer_tempaddr = 0;
405 		} else
406 			prefer_tempaddr = 1;
407 		if (!(ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
408 		    (ia->ia6_flags & IN6_IFF_TEMPORARY)) {
409 			if (prefer_tempaddr)
410 				REPLACE(7);
411 			else
412 				NEXT(7);
413 		}
414 		if ((ia_best->ia6_flags & IN6_IFF_TEMPORARY) &&
415 		    !(ia->ia6_flags & IN6_IFF_TEMPORARY)) {
416 			if (prefer_tempaddr)
417 				NEXT(7);
418 			else
419 				REPLACE(7);
420 		}
421 
422 		/*
423 		 * Rule 8: prefer addresses on alive interfaces.
424 		 * This is a KAME specific rule.
425 		 */
426 		if ((ia_best->ia_ifp->if_flags & IFF_UP) &&
427 		    !(ia->ia_ifp->if_flags & IFF_UP))
428 			NEXT(8);
429 		if (!(ia_best->ia_ifp->if_flags & IFF_UP) &&
430 		    (ia->ia_ifp->if_flags & IFF_UP))
431 			REPLACE(8);
432 
433 		/*
434 		 * Rule 14: Use longest matching prefix.
435 		 * Note: in the address selection draft, this rule is
436 		 * documented as "Rule 8".  However, since it is also
437 		 * documented that this rule can be overridden, we assign
438 		 * a large number so that it is easy to assign smaller numbers
439 		 * to more preferred rules.
440 		 */
441 		new_matchlen = in6_matchlen(&ia->ia_addr.sin6_addr, &dst);
442 		if (best_matchlen < new_matchlen)
443 			REPLACE(14);
444 		if (new_matchlen < best_matchlen)
445 			NEXT(14);
446 
447 		/* Rule 15 is reserved. */
448 
449 		/*
450 		 * Last resort: just keep the current candidate.
451 		 * Or, do we need more rules?
452 		 */
453 		continue;
454 
455 	  replace:
456 		ia_best = ia;
457 		best_scope = (new_scope >= 0 ? new_scope :
458 			      in6_addrscope(&ia_best->ia_addr.sin6_addr));
459 		best_policy = (new_policy ? new_policy :
460 			       lookup_addrsel_policy(&ia_best->ia_addr));
461 		best_matchlen = (new_matchlen >= 0 ? new_matchlen :
462 				 in6_matchlen(&ia_best->ia_addr.sin6_addr,
463 					      &dst));
464 
465 	  next:
466 		continue;
467 
468 	  out:
469 		break;
470 	}
471 
472 	if ((ia = ia_best) == NULL) {
473 		IN6_IFADDR_RUNLOCK();
474 		return (EADDRNOTAVAIL);
475 	}
476 
477 	/*
478 	 * At this point at least one of the addresses belonged to the jail
479 	 * but it could still be, that we want to further restrict it, e.g.
480 	 * theoratically IN6_IS_ADDR_LOOPBACK.
481 	 * It must not be IN6_IS_ADDR_UNSPECIFIED anymore.
482 	 * prison_local_ip6() will fix an IN6_IS_ADDR_LOOPBACK but should
483 	 * let all others previously selected pass.
484 	 * Use tmp to not change ::1 on lo0 to the primary jail address.
485 	 */
486 	tmp = ia->ia_addr.sin6_addr;
487 	if (cred != NULL && prison_local_ip6(cred, &tmp, (inp != NULL &&
488 	    (inp->inp_flags & IN6P_IPV6_V6ONLY) != 0)) != 0) {
489 		IN6_IFADDR_RUNLOCK();
490 		return (EADDRNOTAVAIL);
491 	}
492 
493 	if (ifpp)
494 		*ifpp = ifp;
495 
496 	bcopy(&tmp, srcp, sizeof(*srcp));
497 	IN6_IFADDR_RUNLOCK();
498 	return (0);
499 }
500 
501 /*
502  * clone - meaningful only for bsdi and freebsd
503  */
504 static int
505 selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
506     struct ip6_moptions *mopts, struct route_in6 *ro,
507     struct ifnet **retifp, struct rtentry **retrt, int norouteok)
508 {
509 	int error = 0;
510 	struct ifnet *ifp = NULL;
511 	struct rtentry *rt = NULL;
512 	struct sockaddr_in6 *sin6_next;
513 	struct in6_pktinfo *pi = NULL;
514 	struct in6_addr *dst = &dstsock->sin6_addr;
515 #if 0
516 	char ip6buf[INET6_ADDRSTRLEN];
517 
518 	if (dstsock->sin6_addr.s6_addr32[0] == 0 &&
519 	    dstsock->sin6_addr.s6_addr32[1] == 0 &&
520 	    !IN6_IS_ADDR_LOOPBACK(&dstsock->sin6_addr)) {
521 		printf("in6_selectroute: strange destination %s\n",
522 		       ip6_sprintf(ip6buf, &dstsock->sin6_addr));
523 	} else {
524 		printf("in6_selectroute: destination = %s%%%d\n",
525 		       ip6_sprintf(ip6buf, &dstsock->sin6_addr),
526 		       dstsock->sin6_scope_id); /* for debug */
527 	}
528 #endif
529 
530 	/* If the caller specify the outgoing interface explicitly, use it. */
531 	if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) {
532 		/* XXX boundary check is assumed to be already done. */
533 		ifp = ifnet_byindex(pi->ipi6_ifindex);
534 		if (ifp != NULL &&
535 		    (norouteok || retrt == NULL ||
536 		    IN6_IS_ADDR_MULTICAST(dst))) {
537 			/*
538 			 * we do not have to check or get the route for
539 			 * multicast.
540 			 */
541 			goto done;
542 		} else
543 			goto getroute;
544 	}
545 
546 	/*
547 	 * If the destination address is a multicast address and the outgoing
548 	 * interface for the address is specified by the caller, use it.
549 	 */
550 	if (IN6_IS_ADDR_MULTICAST(dst) &&
551 	    mopts != NULL && (ifp = mopts->im6o_multicast_ifp) != NULL) {
552 		goto done; /* we do not need a route for multicast. */
553 	}
554 
555   getroute:
556 	/*
557 	 * If the next hop address for the packet is specified by the caller,
558 	 * use it as the gateway.
559 	 */
560 	if (opts && opts->ip6po_nexthop) {
561 		struct route_in6 *ron;
562 		struct llentry *la;
563 
564 		sin6_next = satosin6(opts->ip6po_nexthop);
565 
566 		/* at this moment, we only support AF_INET6 next hops */
567 		if (sin6_next->sin6_family != AF_INET6) {
568 			error = EAFNOSUPPORT; /* or should we proceed? */
569 			goto done;
570 		}
571 
572 		/*
573 		 * If the next hop is an IPv6 address, then the node identified
574 		 * by that address must be a neighbor of the sending host.
575 		 */
576 		ron = &opts->ip6po_nextroute;
577 		/*
578 		 * XXX what do we do here?
579 		 * PLZ to be fixing
580 		 */
581 
582 
583 		if (ron->ro_rt == NULL) {
584 			rtalloc((struct route *)ron); /* multi path case? */
585 			if (ron->ro_rt == NULL) {
586 				if (ron->ro_rt) {
587 					RTFREE(ron->ro_rt);
588 					ron->ro_rt = NULL;
589 				}
590 				error = EHOSTUNREACH;
591 				goto done;
592 			}
593 		}
594 
595 		rt = ron->ro_rt;
596 		ifp = rt->rt_ifp;
597 		IF_AFDATA_LOCK(ifp);
598 		la = lla_lookup(LLTABLE6(ifp), 0, (struct sockaddr *)&sin6_next->sin6_addr);
599 		IF_AFDATA_UNLOCK(ifp);
600 		if (la != NULL)
601 			LLE_RUNLOCK(la);
602 		else {
603 			error = EHOSTUNREACH;
604 			goto done;
605 		}
606 #if 0
607 		if ((ron->ro_rt &&
608 		     (ron->ro_rt->rt_flags & (RTF_UP | RTF_LLINFO)) !=
609 		     (RTF_UP | RTF_LLINFO)) ||
610 		    !IN6_ARE_ADDR_EQUAL(&satosin6(&ron->ro_dst)->sin6_addr,
611 		    &sin6_next->sin6_addr)) {
612 			if (ron->ro_rt) {
613 				RTFREE(ron->ro_rt);
614 				ron->ro_rt = NULL;
615 			}
616 			*satosin6(&ron->ro_dst) = *sin6_next;
617 		}
618 		if (ron->ro_rt == NULL) {
619 			rtalloc((struct route *)ron); /* multi path case? */
620 			if (ron->ro_rt == NULL ||
621 			    !(ron->ro_rt->rt_flags & RTF_LLINFO)) {
622 				if (ron->ro_rt) {
623 					RTFREE(ron->ro_rt);
624 					ron->ro_rt = NULL;
625 				}
626 				error = EHOSTUNREACH;
627 				goto done;
628 			}
629 		}
630 #endif
631 
632 		/*
633 		 * When cloning is required, try to allocate a route to the
634 		 * destination so that the caller can store path MTU
635 		 * information.
636 		 */
637 		goto done;
638 	}
639 
640 	/*
641 	 * Use a cached route if it exists and is valid, else try to allocate
642 	 * a new one.  Note that we should check the address family of the
643 	 * cached destination, in case of sharing the cache with IPv4.
644 	 */
645 	if (ro) {
646 		if (ro->ro_rt &&
647 		    (!(ro->ro_rt->rt_flags & RTF_UP) ||
648 		     ((struct sockaddr *)(&ro->ro_dst))->sa_family != AF_INET6 ||
649 		     !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr,
650 		     dst))) {
651 			RTFREE(ro->ro_rt);
652 			ro->ro_rt = (struct rtentry *)NULL;
653 		}
654 		if (ro->ro_rt == (struct rtentry *)NULL) {
655 			struct sockaddr_in6 *sa6;
656 
657 			/* No route yet, so try to acquire one */
658 			bzero(&ro->ro_dst, sizeof(struct sockaddr_in6));
659 			sa6 = (struct sockaddr_in6 *)&ro->ro_dst;
660 			*sa6 = *dstsock;
661 			sa6->sin6_scope_id = 0;
662 
663 #ifdef RADIX_MPATH
664 				rtalloc_mpath((struct route *)ro,
665 				    ntohl(sa6->sin6_addr.s6_addr32[3]));
666 #else
667 				ro->ro_rt = rtalloc1(&((struct route *)ro)
668 				    ->ro_dst, 0, 0UL);
669 				if (ro->ro_rt)
670 					RT_UNLOCK(ro->ro_rt);
671 #endif
672 		}
673 
674 		/*
675 		 * do not care about the result if we have the nexthop
676 		 * explicitly specified.
677 		 */
678 		if (opts && opts->ip6po_nexthop)
679 			goto done;
680 
681 		if (ro->ro_rt) {
682 			ifp = ro->ro_rt->rt_ifp;
683 
684 			if (ifp == NULL) { /* can this really happen? */
685 				RTFREE(ro->ro_rt);
686 				ro->ro_rt = NULL;
687 			}
688 		}
689 		if (ro->ro_rt == NULL)
690 			error = EHOSTUNREACH;
691 		rt = ro->ro_rt;
692 
693 		/*
694 		 * Check if the outgoing interface conflicts with
695 		 * the interface specified by ipi6_ifindex (if specified).
696 		 * Note that loopback interface is always okay.
697 		 * (this may happen when we are sending a packet to one of
698 		 *  our own addresses.)
699 		 */
700 		if (ifp && opts && opts->ip6po_pktinfo &&
701 		    opts->ip6po_pktinfo->ipi6_ifindex) {
702 			if (!(ifp->if_flags & IFF_LOOPBACK) &&
703 			    ifp->if_index !=
704 			    opts->ip6po_pktinfo->ipi6_ifindex) {
705 				error = EHOSTUNREACH;
706 				goto done;
707 			}
708 		}
709 	}
710 
711   done:
712 	if (ifp == NULL && rt == NULL) {
713 		/*
714 		 * This can happen if the caller did not pass a cached route
715 		 * nor any other hints.  We treat this case an error.
716 		 */
717 		error = EHOSTUNREACH;
718 	}
719 	if (error == EHOSTUNREACH)
720 		V_ip6stat.ip6s_noroute++;
721 
722 	if (retifp != NULL) {
723 		*retifp = ifp;
724 
725 		/*
726 		 * Adjust the "outgoing" interface.  If we're going to loop
727 		 * the packet back to ourselves, the ifp would be the loopback
728 		 * interface. However, we'd rather know the interface associated
729 		 * to the destination address (which should probably be one of
730 		 * our own addresses.)
731 		 */
732 		if (rt) {
733 			if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) &&
734 			    (rt->rt_gateway->sa_family == AF_LINK))
735 				*retifp =
736 					ifnet_byindex(((struct sockaddr_dl *)
737 						       rt->rt_gateway)->sdl_index);
738 		}
739 	}
740 
741 	if (retrt != NULL)
742 		*retrt = rt;	/* rt may be NULL */
743 
744 	return (error);
745 }
746 
747 static int
748 in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
749     struct ip6_moptions *mopts, struct route_in6 *ro, struct ifnet **retifp)
750 {
751 	int error;
752 	struct route_in6 sro;
753 	struct rtentry *rt = NULL;
754 
755 	if (ro == NULL) {
756 		bzero(&sro, sizeof(sro));
757 		ro = &sro;
758 	}
759 
760 	if ((error = selectroute(dstsock, opts, mopts, ro, retifp,
761 				     &rt, 1)) != 0) {
762 		if (ro == &sro && rt && rt == sro.ro_rt)
763 			RTFREE(rt);
764 		return (error);
765 	}
766 
767 	/*
768 	 * do not use a rejected or black hole route.
769 	 * XXX: this check should be done in the L2 output routine.
770 	 * However, if we skipped this check here, we'd see the following
771 	 * scenario:
772 	 * - install a rejected route for a scoped address prefix
773 	 *   (like fe80::/10)
774 	 * - send a packet to a destination that matches the scoped prefix,
775 	 *   with ambiguity about the scope zone.
776 	 * - pick the outgoing interface from the route, and disambiguate the
777 	 *   scope zone with the interface.
778 	 * - ip6_output() would try to get another route with the "new"
779 	 *   destination, which may be valid.
780 	 * - we'd see no error on output.
781 	 * Although this may not be very harmful, it should still be confusing.
782 	 * We thus reject the case here.
783 	 */
784 	if (rt && (rt->rt_flags & (RTF_REJECT | RTF_BLACKHOLE))) {
785 		int flags = (rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
786 
787 		if (ro == &sro && rt && rt == sro.ro_rt)
788 			RTFREE(rt);
789 		return (flags);
790 	}
791 
792 	if (ro == &sro && rt && rt == sro.ro_rt)
793 		RTFREE(rt);
794 	return (0);
795 }
796 
797 /*
798  * clone - meaningful only for bsdi and freebsd
799  */
800 int
801 in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
802     struct ip6_moptions *mopts, struct route_in6 *ro,
803     struct ifnet **retifp, struct rtentry **retrt)
804 {
805 
806 	return (selectroute(dstsock, opts, mopts, ro, retifp,
807 	    retrt, 0));
808 }
809 
810 /*
811  * Default hop limit selection. The precedence is as follows:
812  * 1. Hoplimit value specified via ioctl.
813  * 2. (If the outgoing interface is detected) the current
814  *     hop limit of the interface specified by router advertisement.
815  * 3. The system default hoplimit.
816  */
817 int
818 in6_selecthlim(struct inpcb *in6p, struct ifnet *ifp)
819 {
820 
821 	if (in6p && in6p->in6p_hops >= 0)
822 		return (in6p->in6p_hops);
823 	else if (ifp)
824 		return (ND_IFINFO(ifp)->chlim);
825 	else if (in6p && !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
826 		struct route_in6 ro6;
827 		struct ifnet *lifp;
828 
829 		bzero(&ro6, sizeof(ro6));
830 		ro6.ro_dst.sin6_family = AF_INET6;
831 		ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
832 		ro6.ro_dst.sin6_addr = in6p->in6p_faddr;
833 		rtalloc((struct route *)&ro6);
834 		if (ro6.ro_rt) {
835 			lifp = ro6.ro_rt->rt_ifp;
836 			RTFREE(ro6.ro_rt);
837 			if (lifp)
838 				return (ND_IFINFO(lifp)->chlim);
839 		} else
840 			return (V_ip6_defhlim);
841 	}
842 	return (V_ip6_defhlim);
843 }
844 
845 /*
846  * XXX: this is borrowed from in6_pcbbind(). If possible, we should
847  * share this function by all *bsd*...
848  */
849 int
850 in6_pcbsetport(struct in6_addr *laddr, struct inpcb *inp, struct ucred *cred)
851 {
852 	struct socket *so = inp->inp_socket;
853 	u_int16_t lport = 0;
854 	int error, lookupflags = 0;
855 #ifdef INVARIANTS
856 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
857 #endif
858 
859 	INP_WLOCK_ASSERT(inp);
860 	INP_HASH_WLOCK_ASSERT(pcbinfo);
861 
862 	error = prison_local_ip6(cred, laddr,
863 	    ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0));
864 	if (error)
865 		return(error);
866 
867 	/* XXX: this is redundant when called from in6_pcbbind */
868 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
869 		lookupflags = INPLOOKUP_WILDCARD;
870 
871 	inp->inp_flags |= INP_ANONPORT;
872 
873 	error = in_pcb_lport(inp, NULL, &lport, cred, lookupflags);
874 	if (error != 0)
875 		return (error);
876 
877 	inp->inp_lport = lport;
878 	if (in_pcbinshash(inp) != 0) {
879 		inp->in6p_laddr = in6addr_any;
880 		inp->inp_lport = 0;
881 		return (EAGAIN);
882 	}
883 
884 	return (0);
885 }
886 
887 void
888 addrsel_policy_init(void)
889 {
890 
891 	init_policy_queue();
892 
893 	/* initialize the "last resort" policy */
894 	bzero(&V_defaultaddrpolicy, sizeof(V_defaultaddrpolicy));
895 	V_defaultaddrpolicy.label = ADDR_LABEL_NOTAPP;
896 
897 	if (!IS_DEFAULT_VNET(curvnet))
898 		return;
899 
900 	ADDRSEL_LOCK_INIT();
901 	ADDRSEL_SXLOCK_INIT();
902 }
903 
904 static struct in6_addrpolicy *
905 lookup_addrsel_policy(struct sockaddr_in6 *key)
906 {
907 	struct in6_addrpolicy *match = NULL;
908 
909 	ADDRSEL_LOCK();
910 	match = match_addrsel_policy(key);
911 
912 	if (match == NULL)
913 		match = &V_defaultaddrpolicy;
914 	else
915 		match->use++;
916 	ADDRSEL_UNLOCK();
917 
918 	return (match);
919 }
920 
921 /*
922  * Subroutines to manage the address selection policy table via sysctl.
923  */
924 struct walkarg {
925 	struct sysctl_req *w_req;
926 };
927 
928 static int in6_src_sysctl(SYSCTL_HANDLER_ARGS);
929 SYSCTL_DECL(_net_inet6_ip6);
930 static SYSCTL_NODE(_net_inet6_ip6, IPV6CTL_ADDRCTLPOLICY, addrctlpolicy,
931 	CTLFLAG_RD, in6_src_sysctl, "");
932 
933 static int
934 in6_src_sysctl(SYSCTL_HANDLER_ARGS)
935 {
936 	struct walkarg w;
937 
938 	if (req->newptr)
939 		return EPERM;
940 
941 	bzero(&w, sizeof(w));
942 	w.w_req = req;
943 
944 	return (walk_addrsel_policy(dump_addrsel_policyent, &w));
945 }
946 
947 int
948 in6_src_ioctl(u_long cmd, caddr_t data)
949 {
950 	int i;
951 	struct in6_addrpolicy ent0;
952 
953 	if (cmd != SIOCAADDRCTL_POLICY && cmd != SIOCDADDRCTL_POLICY)
954 		return (EOPNOTSUPP); /* check for safety */
955 
956 	ent0 = *(struct in6_addrpolicy *)data;
957 
958 	if (ent0.label == ADDR_LABEL_NOTAPP)
959 		return (EINVAL);
960 	/* check if the prefix mask is consecutive. */
961 	if (in6_mask2len(&ent0.addrmask.sin6_addr, NULL) < 0)
962 		return (EINVAL);
963 	/* clear trailing garbages (if any) of the prefix address. */
964 	for (i = 0; i < 4; i++) {
965 		ent0.addr.sin6_addr.s6_addr32[i] &=
966 			ent0.addrmask.sin6_addr.s6_addr32[i];
967 	}
968 	ent0.use = 0;
969 
970 	switch (cmd) {
971 	case SIOCAADDRCTL_POLICY:
972 		return (add_addrsel_policyent(&ent0));
973 	case SIOCDADDRCTL_POLICY:
974 		return (delete_addrsel_policyent(&ent0));
975 	}
976 
977 	return (0);		/* XXX: compromise compilers */
978 }
979 
980 /*
981  * The followings are implementation of the policy table using a
982  * simple tail queue.
983  * XXX such details should be hidden.
984  * XXX implementation using binary tree should be more efficient.
985  */
986 struct addrsel_policyent {
987 	TAILQ_ENTRY(addrsel_policyent) ape_entry;
988 	struct in6_addrpolicy ape_policy;
989 };
990 
991 TAILQ_HEAD(addrsel_policyhead, addrsel_policyent);
992 
993 static VNET_DEFINE(struct addrsel_policyhead, addrsel_policytab);
994 #define	V_addrsel_policytab		VNET(addrsel_policytab)
995 
996 static void
997 init_policy_queue(void)
998 {
999 
1000 	TAILQ_INIT(&V_addrsel_policytab);
1001 }
1002 
1003 static int
1004 add_addrsel_policyent(struct in6_addrpolicy *newpolicy)
1005 {
1006 	struct addrsel_policyent *new, *pol;
1007 
1008 	new = malloc(sizeof(*new), M_IFADDR,
1009 	       M_WAITOK);
1010 	ADDRSEL_XLOCK();
1011 	ADDRSEL_LOCK();
1012 
1013 	/* duplication check */
1014 	TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1015 		if (IN6_ARE_ADDR_EQUAL(&newpolicy->addr.sin6_addr,
1016 				       &pol->ape_policy.addr.sin6_addr) &&
1017 		    IN6_ARE_ADDR_EQUAL(&newpolicy->addrmask.sin6_addr,
1018 				       &pol->ape_policy.addrmask.sin6_addr)) {
1019 			ADDRSEL_UNLOCK();
1020 			ADDRSEL_XUNLOCK();
1021 			free(new, M_IFADDR);
1022 			return (EEXIST);	/* or override it? */
1023 		}
1024 	}
1025 
1026 	bzero(new, sizeof(*new));
1027 
1028 	/* XXX: should validate entry */
1029 	new->ape_policy = *newpolicy;
1030 
1031 	TAILQ_INSERT_TAIL(&V_addrsel_policytab, new, ape_entry);
1032 	ADDRSEL_UNLOCK();
1033 	ADDRSEL_XUNLOCK();
1034 
1035 	return (0);
1036 }
1037 
1038 static int
1039 delete_addrsel_policyent(struct in6_addrpolicy *key)
1040 {
1041 	struct addrsel_policyent *pol;
1042 
1043 	ADDRSEL_XLOCK();
1044 	ADDRSEL_LOCK();
1045 
1046 	/* search for the entry in the table */
1047 	TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1048 		if (IN6_ARE_ADDR_EQUAL(&key->addr.sin6_addr,
1049 		    &pol->ape_policy.addr.sin6_addr) &&
1050 		    IN6_ARE_ADDR_EQUAL(&key->addrmask.sin6_addr,
1051 		    &pol->ape_policy.addrmask.sin6_addr)) {
1052 			break;
1053 		}
1054 	}
1055 	if (pol == NULL) {
1056 		ADDRSEL_UNLOCK();
1057 		ADDRSEL_XUNLOCK();
1058 		return (ESRCH);
1059 	}
1060 
1061 	TAILQ_REMOVE(&V_addrsel_policytab, pol, ape_entry);
1062 	ADDRSEL_UNLOCK();
1063 	ADDRSEL_XUNLOCK();
1064 
1065 	return (0);
1066 }
1067 
1068 static int
1069 walk_addrsel_policy(int (*callback)(struct in6_addrpolicy *, void *),
1070     void *w)
1071 {
1072 	struct addrsel_policyent *pol;
1073 	int error = 0;
1074 
1075 	ADDRSEL_SLOCK();
1076 	TAILQ_FOREACH(pol, &V_addrsel_policytab, ape_entry) {
1077 		if ((error = (*callback)(&pol->ape_policy, w)) != 0) {
1078 			ADDRSEL_SUNLOCK();
1079 			return (error);
1080 		}
1081 	}
1082 	ADDRSEL_SUNLOCK();
1083 	return (error);
1084 }
1085 
1086 static int
1087 dump_addrsel_policyent(struct in6_addrpolicy *pol, void *arg)
1088 {
1089 	int error = 0;
1090 	struct walkarg *w = arg;
1091 
1092 	error = SYSCTL_OUT(w->w_req, pol, sizeof(*pol));
1093 
1094 	return (error);
1095 }
1096 
1097 static struct in6_addrpolicy *
1098 match_addrsel_policy(struct sockaddr_in6 *key)
1099 {
1100 	struct addrsel_policyent *pent;
1101 	struct in6_addrpolicy *bestpol = NULL, *pol;
1102 	int matchlen, bestmatchlen = -1;
1103 	u_char *mp, *ep, *k, *p, m;
1104 
1105 	TAILQ_FOREACH(pent, &V_addrsel_policytab, ape_entry) {
1106 		matchlen = 0;
1107 
1108 		pol = &pent->ape_policy;
1109 		mp = (u_char *)&pol->addrmask.sin6_addr;
1110 		ep = mp + 16;	/* XXX: scope field? */
1111 		k = (u_char *)&key->sin6_addr;
1112 		p = (u_char *)&pol->addr.sin6_addr;
1113 		for (; mp < ep && *mp; mp++, k++, p++) {
1114 			m = *mp;
1115 			if ((*k & m) != *p)
1116 				goto next; /* not match */
1117 			if (m == 0xff) /* short cut for a typical case */
1118 				matchlen += 8;
1119 			else {
1120 				while (m >= 0x80) {
1121 					matchlen++;
1122 					m <<= 1;
1123 				}
1124 			}
1125 		}
1126 
1127 		/* matched.  check if this is better than the current best. */
1128 		if (bestpol == NULL ||
1129 		    matchlen > bestmatchlen) {
1130 			bestpol = pol;
1131 			bestmatchlen = matchlen;
1132 		}
1133 
1134 	  next:
1135 		continue;
1136 	}
1137 
1138 	return (bestpol);
1139 }
1140