xref: /openbsd-src/sys/netinet6/in6_src.c (revision 0b7734b3d77bb9b21afec6f4621cae6c805dbd45)
1 /*	$OpenBSD: in6_src.c,v 1.78 2016/07/22 11:14:41 mpi Exp $	*/
2 /*	$KAME: in6_src.c,v 1.36 2001/02/06 04:08:17 itojun Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1991, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)in_pcb.c	8.2 (Berkeley) 1/4/94
62  */
63 
64 #include <sys/param.h>
65 #include <sys/systm.h>
66 #include <sys/mbuf.h>
67 #include <sys/protosw.h>
68 #include <sys/socket.h>
69 #include <sys/socketvar.h>
70 #include <sys/ioctl.h>
71 #include <sys/errno.h>
72 #include <sys/time.h>
73 
74 #include <net/if.h>
75 #include <net/if_var.h>
76 #include <net/route.h>
77 
78 #include <netinet/in.h>
79 #include <netinet/ip.h>
80 #include <netinet/in_pcb.h>
81 #include <netinet6/in6_var.h>
82 #include <netinet/ip6.h>
83 #include <netinet6/ip6_var.h>
84 #include <netinet6/nd6.h>
85 
86 int in6_selectif(struct sockaddr_in6 *, struct ip6_pktopts *,
87     struct ip6_moptions *, struct route_in6 *, struct ifnet **, u_int);
88 
89 /*
90  * Return an IPv6 address, which is the most appropriate for a given
91  * destination and user specified options.
92  * If necessary, this function lookups the routing table and returns
93  * an entry to the caller for later use.
94  */
95 int
96 in6_selectsrc(struct in6_addr **in6src, struct sockaddr_in6 *dstsock,
97     struct ip6_pktopts *opts, struct ip6_moptions *mopts,
98     struct route_in6 *ro, struct in6_addr *laddr, u_int rtableid)
99 {
100 	struct ifnet *ifp = NULL;
101 	struct in6_addr *dst;
102 	struct in6_ifaddr *ia6 = NULL;
103 	struct in6_pktinfo *pi = NULL;
104 	int	error;
105 
106 	dst = &dstsock->sin6_addr;
107 
108 	/*
109 	 * If the source address is explicitly specified by the caller,
110 	 * check if the requested source address is indeed a unicast address
111 	 * assigned to the node, and can be used as the packet's source
112 	 * address.  If everything is okay, use the address as source.
113 	 */
114 	if (opts && (pi = opts->ip6po_pktinfo) &&
115 	    !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr)) {
116 		struct sockaddr_in6 sa6;
117 
118 		/* get the outgoing interface */
119 		error = in6_selectif(dstsock, opts, mopts, ro, &ifp, rtableid);
120 		if (error)
121 			return (error);
122 
123 		bzero(&sa6, sizeof(sa6));
124 		sa6.sin6_family = AF_INET6;
125 		sa6.sin6_len = sizeof(sa6);
126 		sa6.sin6_addr = pi->ipi6_addr;
127 
128 		if (ifp && IN6_IS_SCOPE_EMBED(&sa6.sin6_addr))
129 			sa6.sin6_addr.s6_addr16[1] = htons(ifp->if_index);
130 		if_put(ifp); /* put reference from in6_selectif */
131 
132 		ia6 = ifatoia6(ifa_ifwithaddr(sin6tosa(&sa6), rtableid));
133 		if (ia6 == NULL || (ia6->ia6_flags &
134 		     (IN6_IFF_ANYCAST|IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED)))
135 			return (EADDRNOTAVAIL);
136 
137 		pi->ipi6_addr = sa6.sin6_addr; /* XXX: this overrides pi */
138 
139 		*in6src = &pi->ipi6_addr;
140 		return (0);
141 	}
142 
143 	/*
144 	 * If the source address is not specified but the socket(if any)
145 	 * is already bound, use the bound address.
146 	 */
147 	if (laddr && !IN6_IS_ADDR_UNSPECIFIED(laddr)) {
148 		*in6src = laddr;
149 		return (0);
150 	}
151 
152 	/*
153 	 * If the caller doesn't specify the source address but
154 	 * the outgoing interface, use an address associated with
155 	 * the interface.
156 	 */
157 	if (pi && pi->ipi6_ifindex) {
158 		ifp = if_get(pi->ipi6_ifindex);
159 		if (ifp == NULL)
160 			return (ENXIO); /* XXX: better error? */
161 
162 		ia6 = in6_ifawithscope(ifp, dst, rtableid);
163 		if_put(ifp);
164 
165 		if (ia6 == NULL)
166 			return (EADDRNOTAVAIL);
167 
168 		*in6src = &ia6->ia_addr.sin6_addr;
169 		return (0);
170 	}
171 
172 	/*
173 	 * If the destination address is a link-local unicast address or
174 	 * a link/interface-local multicast address, and if the outgoing
175 	 * interface is specified by the sin6_scope_id filed, use an address
176 	 * associated with the interface.
177 	 * XXX: We're now trying to define more specific semantics of
178 	 *      sin6_scope_id field, so this part will be rewritten in
179 	 *      the near future.
180 	 */
181 	if ((IN6_IS_ADDR_LINKLOCAL(dst) || IN6_IS_ADDR_MC_LINKLOCAL(dst) ||
182 	     IN6_IS_ADDR_MC_INTFACELOCAL(dst)) && dstsock->sin6_scope_id) {
183 		ifp = if_get(dstsock->sin6_scope_id);
184 		if (ifp == NULL)
185 			return (ENXIO); /* XXX: better error? */
186 
187 		ia6 = in6_ifawithscope(ifp, dst, rtableid);
188 		if_put(ifp);
189 
190 		if (ia6 == NULL)
191 			return (EADDRNOTAVAIL);
192 
193 		*in6src = &ia6->ia_addr.sin6_addr;
194 		return (0);
195 	}
196 
197 	/*
198 	 * If the destination address is a multicast address and
199 	 * the outgoing interface for the address is specified
200 	 * by the caller, use an address associated with the interface.
201 	 * Even if the outgoing interface is not specified, we also
202 	 * choose a loopback interface as the outgoing interface.
203 	 */
204 	if (IN6_IS_ADDR_MULTICAST(dst)) {
205 		ifp = mopts ? if_get(mopts->im6o_ifidx) : NULL;
206 
207 		if (!ifp && dstsock->sin6_scope_id)
208 			ifp = if_get(htons(dstsock->sin6_scope_id));
209 
210 		if (ifp) {
211 			ia6 = in6_ifawithscope(ifp, dst, rtableid);
212 			if_put(ifp);
213 
214 			if (ia6 == NULL)
215 				return (EADDRNOTAVAIL);
216 
217 			*in6src = &ia6->ia_addr.sin6_addr;
218 			return (0);
219 		}
220 	}
221 
222 	/*
223 	 * If route is known or can be allocated now,
224 	 * our src addr is taken from the i/f, else punt.
225 	 */
226 	if (ro) {
227 		if (!rtisvalid(ro->ro_rt) || (ro->ro_tableid != rtableid) ||
228 		    !IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, dst)) {
229 			rtfree(ro->ro_rt);
230 			ro->ro_rt = NULL;
231 		}
232 		if (ro->ro_rt == NULL) {
233 			struct sockaddr_in6 *sa6;
234 
235 			/* No route yet, so try to acquire one */
236 			bzero(&ro->ro_dst, sizeof(struct sockaddr_in6));
237 			ro->ro_tableid = rtableid;
238 			sa6 = &ro->ro_dst;
239 			sa6->sin6_family = AF_INET6;
240 			sa6->sin6_len = sizeof(struct sockaddr_in6);
241 			sa6->sin6_addr = *dst;
242 			sa6->sin6_scope_id = dstsock->sin6_scope_id;
243 			ro->ro_rt = rtalloc(sin6tosa(&ro->ro_dst),
244 			    RT_RESOLVE, ro->ro_tableid);
245 		}
246 
247 		/*
248 		 * in_pcbconnect() checks out IFF_LOOPBACK to skip using
249 		 * the address. But we don't know why it does so.
250 		 * It is necessary to ensure the scope even for lo0
251 		 * so doesn't check out IFF_LOOPBACK.
252 		 */
253 
254 		if (ro->ro_rt) {
255 			ifp = if_get(ro->ro_rt->rt_ifidx);
256 			if (ifp != NULL) {
257 				ia6 = in6_ifawithscope(ifp, dst, rtableid);
258 				if_put(ifp);
259 			}
260 			if (ia6 == NULL) /* xxx scope error ?*/
261 				ia6 = ifatoia6(ro->ro_rt->rt_ifa);
262 		}
263 		if (ia6 == NULL)
264 			return (EHOSTUNREACH);	/* no route */
265 
266 		*in6src = &ia6->ia_addr.sin6_addr;
267 		return (0);
268 	}
269 
270 	return (EADDRNOTAVAIL);
271 }
272 
273 struct rtentry *
274 in6_selectroute(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
275     struct route_in6 *ro, unsigned int rtableid)
276 {
277 	struct in6_addr *dst;
278 
279 	dst = &dstsock->sin6_addr;
280 
281 	/*
282 	 * Use a cached route if it exists and is valid, else try to allocate
283 	 * a new one.  Note that we should check the address family of the
284 	 * cached destination, in case of sharing the cache with IPv4.
285 	 */
286 	if (ro) {
287 		if (!rtisvalid(ro->ro_rt) ||
288 		     sin6tosa(&ro->ro_dst)->sa_family != AF_INET6 ||
289 		     !IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, dst)) {
290 			rtfree(ro->ro_rt);
291 			ro->ro_rt = NULL;
292 		}
293 		if (ro->ro_rt == NULL) {
294 			struct sockaddr_in6 *sa6;
295 
296 			/* No route yet, so try to acquire one */
297 			bzero(&ro->ro_dst, sizeof(struct sockaddr_in6));
298 			ro->ro_tableid = rtableid;
299 			sa6 = &ro->ro_dst;
300 			*sa6 = *dstsock;
301 			sa6->sin6_scope_id = 0;
302 			ro->ro_tableid = rtableid;
303 			ro->ro_rt = rtalloc_mpath(sin6tosa(&ro->ro_dst),
304 			    NULL, ro->ro_tableid);
305 		}
306 
307 		/*
308 		 * Check if the outgoing interface conflicts with
309 		 * the interface specified by ipi6_ifindex (if specified).
310 		 * Note that loopback interface is always okay.
311 		 * (this may happen when we are sending a packet to one of
312 		 *  our own addresses.)
313 		 */
314 		if (opts && opts->ip6po_pktinfo &&
315 		    opts->ip6po_pktinfo->ipi6_ifindex) {
316 			if (ro->ro_rt != NULL &&
317 			    !ISSET(ro->ro_rt->rt_flags, RTF_LOCAL) &&
318 			    ro->ro_rt->rt_ifidx !=
319 			    opts->ip6po_pktinfo->ipi6_ifindex) {
320 			    	return (NULL);
321 			}
322 		}
323 
324 		return (ro->ro_rt);
325 	}
326 
327 	return (NULL);
328 }
329 
330 int
331 in6_selectif(struct sockaddr_in6 *dstsock, struct ip6_pktopts *opts,
332     struct ip6_moptions *mopts, struct route_in6 *ro, struct ifnet **retifp,
333     u_int rtableid)
334 {
335 	struct rtentry *rt = NULL;
336 	struct in6_pktinfo *pi = NULL;
337 
338 	/* If the caller specify the outgoing interface explicitly, use it. */
339 	if (opts && (pi = opts->ip6po_pktinfo) != NULL && pi->ipi6_ifindex) {
340 		*retifp = if_get(pi->ipi6_ifindex);
341 		if (*retifp != NULL)
342 			return (0);
343 	}
344 
345 	/*
346 	 * If the destination address is a multicast address and the outgoing
347 	 * interface for the address is specified by the caller, use it.
348 	 */
349 	if (IN6_IS_ADDR_MULTICAST(&dstsock->sin6_addr) &&
350 	    mopts != NULL && (*retifp = if_get(mopts->im6o_ifidx)) != NULL)
351 	    	return (0);
352 
353 	rt = in6_selectroute(dstsock, opts, ro, rtableid);
354 	if (rt == NULL)
355 		return (EHOSTUNREACH);
356 
357 	/*
358 	 * do not use a rejected or black hole route.
359 	 * XXX: this check should be done in the L2 output routine.
360 	 * However, if we skipped this check here, we'd see the following
361 	 * scenario:
362 	 * - install a rejected route for a scoped address prefix
363 	 *   (like fe80::/10)
364 	 * - send a packet to a destination that matches the scoped prefix,
365 	 *   with ambiguity about the scope zone.
366 	 * - pick the outgoing interface from the route, and disambiguate the
367 	 *   scope zone with the interface.
368 	 * - ip6_output() would try to get another route with the "new"
369 	 *   destination, which may be valid.
370 	 * - we'd see no error on output.
371 	 * Although this may not be very harmful, it should still be confusing.
372 	 * We thus reject the case here.
373 	 */
374 	if (rt && (rt->rt_flags & (RTF_REJECT | RTF_BLACKHOLE)))
375 		return (rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
376 
377 	if (rt != NULL)
378 		*retifp = if_get(rt->rt_ifidx);
379 
380 	return (0);
381 }
382 
383 int
384 in6_selecthlim(struct inpcb *in6p)
385 {
386 	if (in6p && in6p->inp_hops >= 0)
387 		return (in6p->inp_hops);
388 
389 	return (ip6_defhlim);
390 }
391 
392 /*
393  * generate kernel-internal form (scopeid embedded into s6_addr16[1]).
394  * If the address scope of is link-local, embed the interface index in the
395  * address.  The routine determines our precedence
396  * between advanced API scope/interface specification and basic API
397  * specification.
398  *
399  * this function should be nuked in the future, when we get rid of
400  * embedded scopeid thing.
401  *
402  * XXX actually, it is over-specification to return ifp against sin6_scope_id.
403  * there can be multiple interfaces that belong to a particular scope zone
404  * (in specification, we have 1:N mapping between a scope zone and interfaces).
405  * we may want to change the function to return something other than ifp.
406  */
407 int
408 in6_embedscope(struct in6_addr *in6, const struct sockaddr_in6 *sin6,
409     struct inpcb *in6p)
410 {
411 	struct ifnet *ifp = NULL;
412 	u_int32_t scopeid;
413 
414 	*in6 = sin6->sin6_addr;
415 	scopeid = sin6->sin6_scope_id;
416 
417 	/*
418 	 * don't try to read sin6->sin6_addr beyond here, since the caller may
419 	 * ask us to overwrite existing sockaddr_in6
420 	 */
421 
422 	if (IN6_IS_SCOPE_EMBED(in6)) {
423 		struct in6_pktinfo *pi;
424 
425 		/*
426 		 * KAME assumption: link id == interface id
427 		 */
428 
429 		if (in6p && in6p->inp_outputopts6 &&
430 		    (pi = in6p->inp_outputopts6->ip6po_pktinfo) &&
431 		    pi->ipi6_ifindex) {
432 			ifp = if_get(pi->ipi6_ifindex);
433 			if (ifp == NULL)
434 				return ENXIO;  /* XXX EINVAL? */
435 			in6->s6_addr16[1] = htons(pi->ipi6_ifindex);
436 		} else if (in6p && IN6_IS_ADDR_MULTICAST(in6) &&
437 		    in6p->inp_moptions6 &&
438 		    (ifp = if_get(in6p->inp_moptions6->im6o_ifidx))) {
439 			in6->s6_addr16[1] = htons(ifp->if_index);
440 		} else if (scopeid) {
441 			ifp = if_get(scopeid);
442 			if (ifp == NULL)
443 				return ENXIO;  /* XXX EINVAL? */
444 			/*XXX assignment to 16bit from 32bit variable */
445 			in6->s6_addr16[1] = htons(scopeid & 0xffff);
446 		}
447 		if_put(ifp);
448 	}
449 
450 	return 0;
451 }
452 
453 /*
454  * generate standard sockaddr_in6 from embedded form.
455  * touches sin6_addr and sin6_scope_id only.
456  *
457  * this function should be nuked in the future, when we get rid of
458  * embedded scopeid thing.
459  */
460 void
461 in6_recoverscope(struct sockaddr_in6 *sin6, const struct in6_addr *in6)
462 {
463 	u_int32_t scopeid;
464 
465 	sin6->sin6_addr = *in6;
466 
467 	/*
468 	 * don't try to read *in6 beyond here, since the caller may
469 	 * ask us to overwrite existing sockaddr_in6
470 	 */
471 
472 	sin6->sin6_scope_id = 0;
473 	if (IN6_IS_SCOPE_EMBED(in6)) {
474 		/*
475 		 * KAME assumption: link id == interface id
476 		 */
477 		scopeid = ntohs(sin6->sin6_addr.s6_addr16[1]);
478 		if (scopeid) {
479 			sin6->sin6_addr.s6_addr16[1] = 0;
480 			sin6->sin6_scope_id = scopeid;
481 		}
482 	}
483 }
484 
485 /*
486  * just clear the embedded scope identifer.
487  */
488 void
489 in6_clearscope(struct in6_addr *addr)
490 {
491 	if (IN6_IS_SCOPE_EMBED(addr))
492 		addr->s6_addr16[1] = 0;
493 }
494