xref: /netbsd-src/sys/netinet6/in6_src.c (revision df0caa2637da0538ecdf6b878c4d08e684b43d8f)
1 /*	$NetBSD: in6_src.c,v 1.21 2005/05/29 21:43:51 christos 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/cdefs.h>
65 __KERNEL_RCSID(0, "$NetBSD: in6_src.c,v 1.21 2005/05/29 21:43:51 christos Exp $");
66 
67 #include "opt_inet.h"
68 
69 #include <sys/param.h>
70 #include <sys/systm.h>
71 #include <sys/malloc.h>
72 #include <sys/mbuf.h>
73 #include <sys/protosw.h>
74 #include <sys/socket.h>
75 #include <sys/socketvar.h>
76 #include <sys/ioctl.h>
77 #include <sys/errno.h>
78 #include <sys/time.h>
79 #include <sys/proc.h>
80 
81 #include <net/if.h>
82 #include <net/route.h>
83 
84 #include <netinet/in.h>
85 #include <netinet/in_var.h>
86 #include <netinet/in_systm.h>
87 #include <netinet/ip.h>
88 #include <netinet/in_pcb.h>
89 #include <netinet6/in6_var.h>
90 #include <netinet/ip6.h>
91 #include <netinet6/in6_pcb.h>
92 #include <netinet6/ip6_var.h>
93 #include <netinet6/nd6.h>
94 #ifdef ENABLE_DEFAULT_SCOPE
95 #include <netinet6/scope6_var.h>
96 #endif
97 
98 #include <net/net_osdep.h>
99 
100 /*
101  * Return an IPv6 address, which is the most appropriate for a given
102  * destination and user specified options.
103  * If necessary, this function lookups the routing table and returns
104  * an entry to the caller for later use.
105  */
106 struct in6_addr *
107 in6_selectsrc(dstsock, opts, mopts, ro, laddr, errorp)
108 	struct sockaddr_in6 *dstsock;
109 	struct ip6_pktopts *opts;
110 	struct ip6_moptions *mopts;
111 	struct route_in6 *ro;
112 	struct in6_addr *laddr;
113 	int *errorp;
114 {
115 	struct in6_addr *dst;
116 	struct in6_ifaddr *ia6 = 0;
117 	struct in6_pktinfo *pi = NULL;
118 
119 	dst = &dstsock->sin6_addr;
120 	*errorp = 0;
121 
122 	/*
123 	 * If the source address is explicitly specified by the caller,
124 	 * use it.
125 	 */
126 	if (opts && (pi = opts->ip6po_pktinfo) &&
127 	    !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr))
128 		return (&pi->ipi6_addr);
129 
130 	/*
131 	 * If the source address is not specified but the socket(if any)
132 	 * is already bound, use the bound address.
133 	 */
134 	if (laddr && !IN6_IS_ADDR_UNSPECIFIED(laddr))
135 		return (laddr);
136 
137 	/*
138 	 * If the caller doesn't specify the source address but
139 	 * the outgoing interface, use an address associated with
140 	 * the interface.
141 	 */
142 	if (pi && pi->ipi6_ifindex) {
143 		/* XXX boundary check is assumed to be already done. */
144 		ia6 = in6_ifawithscope(ifindex2ifnet[pi->ipi6_ifindex],
145 				       dst);
146 		if (ia6 == 0) {
147 			*errorp = EADDRNOTAVAIL;
148 			return (0);
149 		}
150 		return (&satosin6(&ia6->ia_addr)->sin6_addr);
151 	}
152 
153 	/*
154 	 * If the destination address is a link-local unicast address or
155 	 * a multicast address, and if the outgoing interface is specified
156 	 * by the sin6_scope_id filed, use an address associated with the
157 	 * interface.
158 	 * XXX: We're now trying to define more specific semantics of
159 	 *      sin6_scope_id field, so this part will be rewritten in
160 	 *      the near future.
161 	 */
162 	if ((IN6_IS_ADDR_LINKLOCAL(dst) || IN6_IS_ADDR_MC_LINKLOCAL(dst)) &&
163 	    dstsock->sin6_scope_id) {
164 		/*
165 		 * I'm not sure if boundary check for scope_id is done
166 		 * somewhere...
167 		 */
168 		if (dstsock->sin6_scope_id < 0 ||
169 		    if_indexlim <= dstsock->sin6_scope_id ||
170 		    !ifindex2ifnet[dstsock->sin6_scope_id]) {
171 			*errorp = ENXIO; /* XXX: better error? */
172 			return (0);
173 		}
174 		ia6 = in6_ifawithscope(ifindex2ifnet[dstsock->sin6_scope_id],
175 				       dst);
176 		if (ia6 == 0) {
177 			*errorp = EADDRNOTAVAIL;
178 			return (0);
179 		}
180 		return (&satosin6(&ia6->ia_addr)->sin6_addr);
181 	}
182 
183 	/*
184 	 * If the destination address is a multicast address and
185 	 * the outgoing interface for the address is specified
186 	 * by the caller, use an address associated with the interface.
187 	 * There is a sanity check here; if the destination has node-local
188 	 * scope, the outgoing interfacde should be a loopback address.
189 	 * Even if the outgoing interface is not specified, we also
190 	 * choose a loopback interface as the outgoing interface.
191 	 */
192 	if (IN6_IS_ADDR_MULTICAST(dst)) {
193 		struct ifnet *ifp = mopts ? mopts->im6o_multicast_ifp : NULL;
194 
195 		if (ifp == NULL && IN6_IS_ADDR_MC_NODELOCAL(dst))
196 			ifp = lo0ifp;
197 
198 		if (ifp) {
199 			ia6 = in6_ifawithscope(ifp, dst);
200 			if (ia6 == 0) {
201 				*errorp = EADDRNOTAVAIL;
202 				return (0);
203 			}
204 			return (&satosin6(&ia6->ia_addr)->sin6_addr);
205 		}
206 	}
207 
208 	/*
209 	 * If the next hop address for the packet is specified
210 	 * by caller, use an address associated with the route
211 	 * to the next hop.
212 	 */
213 	{
214 		struct sockaddr_in6 *sin6_next;
215 		struct rtentry *rt;
216 
217 		if (opts && opts->ip6po_nexthop) {
218 			sin6_next = satosin6(opts->ip6po_nexthop);
219 			rt = nd6_lookup(&sin6_next->sin6_addr, 1, NULL);
220 			if (rt) {
221 				ia6 = in6_ifawithscope(rt->rt_ifp, dst);
222 				if (ia6 == 0)
223 					ia6 = ifatoia6(rt->rt_ifa);
224 			}
225 			if (ia6 == 0) {
226 				*errorp = EADDRNOTAVAIL;
227 				return (0);
228 			}
229 			return (&satosin6(&ia6->ia_addr)->sin6_addr);
230 		}
231 	}
232 
233 	/*
234 	 * If route is known or can be allocated now,
235 	 * our src addr is taken from the i/f, else punt.
236 	 * Note that we should check the address family of the
237 	 * cached destination, in case of sharing the cache with IPv4.
238 	 */
239 	if (ro) {
240 		if (ro->ro_rt &&
241 		    (ro->ro_dst.sin6_family != AF_INET6 ||
242 		     !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr, dst))) {
243 			RTFREE(ro->ro_rt);
244 			ro->ro_rt = (struct rtentry *)0;
245 		}
246 		if (ro->ro_rt == (struct rtentry *)0 ||
247 		    ro->ro_rt->rt_ifp == (struct ifnet *)0) {
248 			struct sockaddr_in6 *sa6;
249 
250 			/* No route yet, so try to acquire one */
251 			bzero(&ro->ro_dst, sizeof(struct sockaddr_in6));
252 			sa6 = (struct sockaddr_in6 *)&ro->ro_dst;
253 			sa6->sin6_family = AF_INET6;
254 			sa6->sin6_len = sizeof(struct sockaddr_in6);
255 			sa6->sin6_addr = *dst;
256 			sa6->sin6_scope_id = dstsock->sin6_scope_id;
257 			if (IN6_IS_ADDR_MULTICAST(dst)) {
258 				ro->ro_rt = rtalloc1(&((struct route *)ro)
259 						     ->ro_dst, 0);
260 			} else {
261 				rtalloc((struct route *)ro);
262 			}
263 		}
264 
265 		/*
266 		 * in_pcbconnect() checks out IFF_LOOPBACK to skip using
267 		 * the address. But we don't know why it does so.
268 		 * It is necessary to ensure the scope even for lo0
269 		 * so doesn't check out IFF_LOOPBACK.
270 		 */
271 
272 		if (ro->ro_rt) {
273 			ia6 = in6_ifawithscope(ro->ro_rt->rt_ifa->ifa_ifp, dst);
274 			if (ia6 == 0) /* xxx scope error ?*/
275 				ia6 = ifatoia6(ro->ro_rt->rt_ifa);
276 		}
277 #if 0
278 		/*
279 		 * xxx The followings are necessary? (kazu)
280 		 * I don't think so.
281 		 * It's for SO_DONTROUTE option in IPv4.(jinmei)
282 		 */
283 		if (ia6 == 0) {
284 			struct sockaddr_in6 sin6 = {sizeof(sin6), AF_INET6, 0};
285 
286 			sin6->sin6_addr = *dst;
287 
288 			ia6 = ifatoia6(ifa_ifwithdstaddr(sin6tosa(&sin6)));
289 			if (ia6 == 0)
290 				ia6 = ifatoia6(ifa_ifwithnet(sin6tosa(&sin6)));
291 			if (ia6 == 0)
292 				return (0);
293 			return (&satosin6(&ia6->ia_addr)->sin6_addr);
294 		}
295 #endif /* 0 */
296 		if (ia6 == 0) {
297 			*errorp = EHOSTUNREACH;	/* no route */
298 			return (0);
299 		}
300 		return (&satosin6(&ia6->ia_addr)->sin6_addr);
301 	}
302 
303 	*errorp = EADDRNOTAVAIL;
304 	return (0);
305 }
306 
307 /*
308  * Default hop limit selection. The precedence is as follows:
309  * 1. Hoplimit value specified via ioctl.
310  * 2. (If the outgoing interface is detected) the current
311  *     hop limit of the interface specified by router advertisement.
312  * 3. The system default hoplimit.
313 */
314 int
315 in6_selecthlim(in6p, ifp)
316 	struct in6pcb *in6p;
317 	struct ifnet *ifp;
318 {
319 	if (in6p && in6p->in6p_af != AF_INET6)
320 		return (-1);
321 
322 	if (in6p && in6p->in6p_hops >= 0)
323 		return (in6p->in6p_hops);
324 	else if (ifp)
325 		return (ND_IFINFO(ifp)->chlim);
326 	else
327 		return (ip6_defhlim);
328 }
329 
330 /*
331  * Find an empty port and set it to the specified PCB.
332  */
333 int
334 in6_pcbsetport(laddr, in6p, p)
335 	struct in6_addr *laddr;
336 	struct in6pcb *in6p;
337 	struct proc *p;
338 {
339 	struct socket *so = in6p->in6p_socket;
340 	struct inpcbtable *table = in6p->in6p_table;
341 	int cnt;
342 	u_int16_t xmin, xmax;
343 	u_int16_t lport, *lastport;
344 	int wild = 0;
345 	void *t;
346 
347 	if (in6p->in6p_af != AF_INET6)
348 		return (EINVAL);
349 
350 	/* XXX: this is redundant when called from in6_pcbbind */
351 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0 &&
352 	   ((so->so_proto->pr_flags & PR_CONNREQUIRED) == 0 ||
353 	    (so->so_options & SO_ACCEPTCONN) == 0))
354 		wild = 1;
355 
356 	if (in6p->in6p_flags & IN6P_LOWPORT) {
357 #ifndef IPNOPRIVPORTS
358 		if (p == 0 || (suser(p->p_ucred, &p->p_acflag) != 0))
359 			return (EACCES);
360 #endif
361 		xmin = ip6_lowportmin;
362 		xmax = ip6_lowportmax;
363 		lastport = &table->inpt_lastlow;
364 	} else {
365 		xmin = ip6_anonportmin;
366 		xmax = ip6_anonportmax;
367 		lastport = &table->inpt_lastport;
368 	}
369 	if (xmin > xmax) {	/* sanity check */
370 		u_int16_t swp;
371 
372 		swp = xmin;
373 		xmin = xmax;
374 		xmax = swp;
375 	}
376 
377 	lport = *lastport - 1;
378 	for (cnt = xmax - xmin + 1; cnt; cnt--, lport--) {
379 		if (lport < xmin || lport > xmax)
380 			lport = xmax;
381 #ifdef INET
382 		if (IN6_IS_ADDR_V4MAPPED(laddr)) {
383 			t = in_pcblookup_port(table,
384 			    *(struct in_addr *)&in6p->in6p_laddr.s6_addr32[3],
385 			    lport, wild);
386 		} else
387 #endif
388 		{
389 			t = in6_pcblookup_port(table, laddr, lport, wild);
390 		}
391 		if (t == 0)
392 			goto found;
393 	}
394 
395 	return (EAGAIN);
396 
397 found:
398 	in6p->in6p_flags |= IN6P_ANONPORT;
399 	*lastport = lport;
400 	in6p->in6p_lport = htons(lport);
401 	in6_pcbstate(in6p, IN6P_BOUND);
402 	return (0);		/* success */
403 }
404 
405 /*
406  * generate kernel-internal form (scopeid embedded into s6_addr16[1]).
407  * If the address scope of is link-local, embed the interface index in the
408  * address.  The routine determines our precedence
409  * between advanced API scope/interface specification and basic API
410  * specification.
411  *
412  * this function should be nuked in the future, when we get rid of
413  * embedded scopeid thing.
414  *
415  * XXX actually, it is over-specification to return ifp against sin6_scope_id.
416  * there can be multiple interfaces that belong to a particular scope zone
417  * (in specification, we have 1:N mapping between a scope zone and interfaces).
418  * we may want to change the function to return something other than ifp.
419  */
420 int
421 in6_embedscope(in6, sin6, in6p, ifpp)
422 	struct in6_addr *in6;
423 	const struct sockaddr_in6 *sin6;
424 	struct in6pcb *in6p;
425 	struct ifnet **ifpp;
426 {
427 	struct ifnet *ifp = NULL;
428 	u_int32_t scopeid;
429 
430 	*in6 = sin6->sin6_addr;
431 	scopeid = sin6->sin6_scope_id;
432 	if (ifpp)
433 		*ifpp = NULL;
434 
435 	/*
436 	 * don't try to read sin6->sin6_addr beyond here, since the caller may
437 	 * ask us to overwrite existing sockaddr_in6
438 	 */
439 
440 #ifdef ENABLE_DEFAULT_SCOPE
441 	if (scopeid == 0)
442 		scopeid = scope6_addr2default(in6);
443 #endif
444 
445 	if (IN6_IS_SCOPE_LINKLOCAL(in6)) {
446 		struct in6_pktinfo *pi;
447 
448 		/*
449 		 * KAME assumption: link id == interface id
450 		 */
451 
452 		if (in6p && in6p->in6p_outputopts &&
453 		    (pi = in6p->in6p_outputopts->ip6po_pktinfo) &&
454 		    pi->ipi6_ifindex) {
455 			ifp = ifindex2ifnet[pi->ipi6_ifindex];
456 			in6->s6_addr16[1] = htons(pi->ipi6_ifindex);
457 		} else if (in6p && IN6_IS_ADDR_MULTICAST(in6) &&
458 			   in6p->in6p_moptions &&
459 			   in6p->in6p_moptions->im6o_multicast_ifp) {
460 			ifp = in6p->in6p_moptions->im6o_multicast_ifp;
461 			in6->s6_addr16[1] = htons(ifp->if_index);
462 		} else if (scopeid) {
463 			/* boundary check */
464 			if (scopeid < 0 || if_indexlim <= scopeid ||
465 			    !ifindex2ifnet[scopeid])
466 				return ENXIO;  /* XXX EINVAL? */
467 			ifp = ifindex2ifnet[scopeid];
468 			/* XXX assignment to 16bit from 32bit variable */
469 			in6->s6_addr16[1] = htons(scopeid & 0xffff);
470 		}
471 
472 		if (ifpp)
473 			*ifpp = ifp;
474 	}
475 
476 	return 0;
477 }
478 
479 /*
480  * generate standard sockaddr_in6 from embedded form.
481  * touches sin6_addr and sin6_scope_id only.
482  *
483  * this function should be nuked in the future, when we get rid of
484  * embedded scopeid thing.
485  */
486 int
487 in6_recoverscope(sin6, in6, ifp)
488 	struct sockaddr_in6 *sin6;
489 	const struct in6_addr *in6;
490 	struct ifnet *ifp;
491 {
492 	u_int32_t scopeid;
493 
494 	sin6->sin6_addr = *in6;
495 
496 	/*
497 	 * don't try to read *in6 beyond here, since the caller may
498 	 * ask us to overwrite existing sockaddr_in6
499 	 */
500 
501 	sin6->sin6_scope_id = 0;
502 	if (IN6_IS_SCOPE_LINKLOCAL(in6)) {
503 		/*
504 		 * KAME assumption: link id == interface id
505 		 */
506 		scopeid = ntohs(sin6->sin6_addr.s6_addr16[1]);
507 		if (scopeid) {
508 			/* sanity check */
509 			if (scopeid < 0 || if_indexlim <= scopeid ||
510 			    !ifindex2ifnet[scopeid])
511 				return ENXIO;
512 			if (ifp && ifp->if_index != scopeid)
513 				return ENXIO;
514 			sin6->sin6_addr.s6_addr16[1] = 0;
515 			sin6->sin6_scope_id = scopeid;
516 		}
517 	}
518 
519 	return 0;
520 }
521 
522 /*
523  * just clear the embedded scope identifer.
524  */
525 void
526 in6_clearscope(addr)
527 	struct in6_addr *addr;
528 {
529 	if (IN6_IS_SCOPE_LINKLOCAL(addr))
530 		addr->s6_addr16[1] = 0;
531 }
532