xref: /netbsd-src/sys/netinet6/in6_src.c (revision 23c8222edbfb0f0932d88a8351d3a0cf817dfb9e)
1 /*	$NetBSD: in6_src.c,v 1.18 2003/12/10 11:46:33 itojun 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 1.18 2003/12/10 11:46:33 itojun 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 #include "loop.h"
101 extern struct ifnet loif[NLOOP];
102 
103 /*
104  * Return an IPv6 address, which is the most appropriate for a given
105  * destination and user specified options.
106  * If necessary, this function lookups the routing table and returns
107  * an entry to the caller for later use.
108  */
109 struct in6_addr *
110 in6_selectsrc(dstsock, opts, mopts, ro, laddr, errorp)
111 	struct sockaddr_in6 *dstsock;
112 	struct ip6_pktopts *opts;
113 	struct ip6_moptions *mopts;
114 	struct route_in6 *ro;
115 	struct in6_addr *laddr;
116 	int *errorp;
117 {
118 	struct in6_addr *dst;
119 	struct in6_ifaddr *ia6 = 0;
120 	struct in6_pktinfo *pi = NULL;
121 
122 	dst = &dstsock->sin6_addr;
123 	*errorp = 0;
124 
125 	/*
126 	 * If the source address is explicitly specified by the caller,
127 	 * use it.
128 	 */
129 	if (opts && (pi = opts->ip6po_pktinfo) &&
130 	    !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr))
131 		return (&pi->ipi6_addr);
132 
133 	/*
134 	 * If the source address is not specified but the socket(if any)
135 	 * is already bound, use the bound address.
136 	 */
137 	if (laddr && !IN6_IS_ADDR_UNSPECIFIED(laddr))
138 		return (laddr);
139 
140 	/*
141 	 * If the caller doesn't specify the source address but
142 	 * the outgoing interface, use an address associated with
143 	 * the interface.
144 	 */
145 	if (pi && pi->ipi6_ifindex) {
146 		/* XXX boundary check is assumed to be already done. */
147 		ia6 = in6_ifawithscope(ifindex2ifnet[pi->ipi6_ifindex],
148 				       dst);
149 		if (ia6 == 0) {
150 			*errorp = EADDRNOTAVAIL;
151 			return (0);
152 		}
153 		return (&satosin6(&ia6->ia_addr)->sin6_addr);
154 	}
155 
156 	/*
157 	 * If the destination address is a link-local unicast address or
158 	 * a multicast address, and if the outgoing interface is specified
159 	 * by the sin6_scope_id filed, use an address associated with the
160 	 * interface.
161 	 * XXX: We're now trying to define more specific semantics of
162 	 *      sin6_scope_id field, so this part will be rewritten in
163 	 *      the near future.
164 	 */
165 	if ((IN6_IS_ADDR_LINKLOCAL(dst) || IN6_IS_ADDR_MULTICAST(dst)) &&
166 	    dstsock->sin6_scope_id) {
167 		/*
168 		 * I'm not sure if boundary check for scope_id is done
169 		 * somewhere...
170 		 */
171 		if (dstsock->sin6_scope_id < 0 ||
172 		    if_indexlim <= dstsock->sin6_scope_id ||
173 		    !ifindex2ifnet[dstsock->sin6_scope_id]) {
174 			*errorp = ENXIO; /* XXX: better error? */
175 			return (0);
176 		}
177 		ia6 = in6_ifawithscope(ifindex2ifnet[dstsock->sin6_scope_id],
178 				       dst);
179 		if (ia6 == 0) {
180 			*errorp = EADDRNOTAVAIL;
181 			return (0);
182 		}
183 		return (&satosin6(&ia6->ia_addr)->sin6_addr);
184 	}
185 
186 	/*
187 	 * If the destination address is a multicast address and
188 	 * the outgoing interface for the address is specified
189 	 * by the caller, use an address associated with the interface.
190 	 * There is a sanity check here; if the destination has node-local
191 	 * scope, the outgoing interfacde should be a loopback address.
192 	 * Even if the outgoing interface is not specified, we also
193 	 * choose a loopback interface as the outgoing interface.
194 	 */
195 	if (IN6_IS_ADDR_MULTICAST(dst)) {
196 		struct ifnet *ifp = mopts ? mopts->im6o_multicast_ifp : NULL;
197 
198 		if (ifp == NULL && IN6_IS_ADDR_MC_NODELOCAL(dst)) {
199 			ifp = &loif[0];
200 		}
201 
202 		if (ifp) {
203 			ia6 = in6_ifawithscope(ifp, dst);
204 			if (ia6 == 0) {
205 				*errorp = EADDRNOTAVAIL;
206 				return (0);
207 			}
208 			return (&satosin6(&ia6->ia_addr)->sin6_addr);
209 		}
210 	}
211 
212 	/*
213 	 * If the next hop address for the packet is specified
214 	 * by caller, use an address associated with the route
215 	 * to the next hop.
216 	 */
217 	{
218 		struct sockaddr_in6 *sin6_next;
219 		struct rtentry *rt;
220 
221 		if (opts && opts->ip6po_nexthop) {
222 			sin6_next = satosin6(opts->ip6po_nexthop);
223 			rt = nd6_lookup(&sin6_next->sin6_addr, 1, NULL);
224 			if (rt) {
225 				ia6 = in6_ifawithscope(rt->rt_ifp, dst);
226 				if (ia6 == 0)
227 					ia6 = ifatoia6(rt->rt_ifa);
228 			}
229 			if (ia6 == 0) {
230 				*errorp = EADDRNOTAVAIL;
231 				return (0);
232 			}
233 			return (&satosin6(&ia6->ia_addr)->sin6_addr);
234 		}
235 	}
236 
237 	/*
238 	 * If route is known or can be allocated now,
239 	 * our src addr is taken from the i/f, else punt.
240 	 * Note that we should check the address family of the
241 	 * cached destination, in case of sharing the cache with IPv4.
242 	 */
243 	if (ro) {
244 		if (ro->ro_rt &&
245 		    (ro->ro_dst.sin6_family != AF_INET6 ||
246 		     !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr, dst))) {
247 			RTFREE(ro->ro_rt);
248 			ro->ro_rt = (struct rtentry *)0;
249 		}
250 		if (ro->ro_rt == (struct rtentry *)0 ||
251 		    ro->ro_rt->rt_ifp == (struct ifnet *)0) {
252 			struct sockaddr_in6 *sa6;
253 
254 			/* No route yet, so try to acquire one */
255 			bzero(&ro->ro_dst, sizeof(struct sockaddr_in6));
256 			sa6 = (struct sockaddr_in6 *)&ro->ro_dst;
257 			sa6->sin6_family = AF_INET6;
258 			sa6->sin6_len = sizeof(struct sockaddr_in6);
259 			sa6->sin6_addr = *dst;
260 			sa6->sin6_scope_id = dstsock->sin6_scope_id;
261 			if (IN6_IS_ADDR_MULTICAST(dst)) {
262 				ro->ro_rt = rtalloc1(&((struct route *)ro)
263 						     ->ro_dst, 0);
264 			} else {
265 				rtalloc((struct route *)ro);
266 			}
267 		}
268 
269 		/*
270 		 * in_pcbconnect() checks out IFF_LOOPBACK to skip using
271 		 * the address. But we don't know why it does so.
272 		 * It is necessary to ensure the scope even for lo0
273 		 * so doesn't check out IFF_LOOPBACK.
274 		 */
275 
276 		if (ro->ro_rt) {
277 			ia6 = in6_ifawithscope(ro->ro_rt->rt_ifa->ifa_ifp, dst);
278 			if (ia6 == 0) /* xxx scope error ?*/
279 				ia6 = ifatoia6(ro->ro_rt->rt_ifa);
280 		}
281 #if 0
282 		/*
283 		 * xxx The followings are necessary? (kazu)
284 		 * I don't think so.
285 		 * It's for SO_DONTROUTE option in IPv4.(jinmei)
286 		 */
287 		if (ia6 == 0) {
288 			struct sockaddr_in6 sin6 = {sizeof(sin6), AF_INET6, 0};
289 
290 			sin6->sin6_addr = *dst;
291 
292 			ia6 = ifatoia6(ifa_ifwithdstaddr(sin6tosa(&sin6)));
293 			if (ia6 == 0)
294 				ia6 = ifatoia6(ifa_ifwithnet(sin6tosa(&sin6)));
295 			if (ia6 == 0)
296 				return (0);
297 			return (&satosin6(&ia6->ia_addr)->sin6_addr);
298 		}
299 #endif /* 0 */
300 		if (ia6 == 0) {
301 			*errorp = EHOSTUNREACH;	/* no route */
302 			return (0);
303 		}
304 		return (&satosin6(&ia6->ia_addr)->sin6_addr);
305 	}
306 
307 	*errorp = EADDRNOTAVAIL;
308 	return (0);
309 }
310 
311 /*
312  * Default hop limit selection. The precedence is as follows:
313  * 1. Hoplimit value specified via ioctl.
314  * 2. (If the outgoing interface is detected) the current
315  *     hop limit of the interface specified by router advertisement.
316  * 3. The system default hoplimit.
317 */
318 int
319 in6_selecthlim(in6p, ifp)
320 	struct in6pcb *in6p;
321 	struct ifnet *ifp;
322 {
323 	if (in6p && in6p->in6p_af != AF_INET6)
324 		return (-1);
325 
326 	if (in6p && in6p->in6p_hops >= 0)
327 		return (in6p->in6p_hops);
328 	else if (ifp)
329 		return (ND_IFINFO(ifp)->chlim);
330 	else
331 		return (ip6_defhlim);
332 }
333 
334 /*
335  * Find an empty port and set it to the specified PCB.
336  */
337 int
338 in6_pcbsetport(laddr, in6p, p)
339 	struct in6_addr *laddr;
340 	struct in6pcb *in6p;
341 	struct proc *p;
342 {
343 	struct socket *so = in6p->in6p_socket;
344 	struct inpcbtable *table = in6p->in6p_table;
345 	int cnt;
346 	u_int16_t min, max;
347 	u_int16_t lport, *lastport;
348 	int wild = 0;
349 	void *t;
350 
351 	if (in6p->in6p_af != AF_INET6)
352 		return (EINVAL);
353 
354 	/* XXX: this is redundant when called from in6_pcbbind */
355 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0 &&
356 	   ((so->so_proto->pr_flags & PR_CONNREQUIRED) == 0 ||
357 	    (so->so_options & SO_ACCEPTCONN) == 0))
358 		wild = 1;
359 
360 	if (in6p->in6p_flags & IN6P_LOWPORT) {
361 #ifndef IPNOPRIVPORTS
362 		if (p == 0 || (suser(p->p_ucred, &p->p_acflag) != 0))
363 			return (EACCES);
364 #endif
365 		min = ip6_lowportmin;
366 		max = ip6_lowportmax;
367 		lastport = &table->inpt_lastlow;
368 	} else {
369 		min = ip6_anonportmin;
370 		max = ip6_anonportmax;
371 		lastport = &table->inpt_lastport;
372 	}
373 	if (min > max) {	/* sanity check */
374 		u_int16_t swp;
375 
376 		swp = min;
377 		min = max;
378 		max = swp;
379 	}
380 
381 	lport = *lastport - 1;
382 	for (cnt = max - min + 1; cnt; cnt--, lport--) {
383 		if (lport < min || lport > max)
384 			lport = max;
385 #ifdef INET
386 		if (IN6_IS_ADDR_V4MAPPED(laddr)) {
387 			t = in_pcblookup_port(table,
388 			    *(struct in_addr *)&in6p->in6p_laddr.s6_addr32[3],
389 			    lport, wild);
390 		} else
391 #endif
392 		{
393 			t = in6_pcblookup_port(table, laddr, lport, wild);
394 		}
395 		if (t == 0)
396 			goto found;
397 	}
398 
399 	return (EAGAIN);
400 
401 found:
402 	in6p->in6p_flags |= IN6P_ANONPORT;
403 	*lastport = lport;
404 	in6p->in6p_lport = htons(lport);
405 	in6_pcbstate(in6p, IN6P_BOUND);
406 	return (0);		/* success */
407 }
408 
409 /*
410  * generate kernel-internal form (scopeid embedded into s6_addr16[1]).
411  * If the address scope of is link-local, embed the interface index in the
412  * address.  The routine determines our precedence
413  * between advanced API scope/interface specification and basic API
414  * specification.
415  *
416  * this function should be nuked in the future, when we get rid of
417  * embedded scopeid thing.
418  *
419  * XXX actually, it is over-specification to return ifp against sin6_scope_id.
420  * there can be multiple interfaces that belong to a particular scope zone
421  * (in specification, we have 1:N mapping between a scope zone and interfaces).
422  * we may want to change the function to return something other than ifp.
423  */
424 int
425 in6_embedscope(in6, sin6, in6p, ifpp)
426 	struct in6_addr *in6;
427 	const struct sockaddr_in6 *sin6;
428 	struct in6pcb *in6p;
429 	struct ifnet **ifpp;
430 {
431 	struct ifnet *ifp = NULL;
432 	u_int32_t scopeid;
433 
434 	*in6 = sin6->sin6_addr;
435 	scopeid = sin6->sin6_scope_id;
436 	if (ifpp)
437 		*ifpp = NULL;
438 
439 	/*
440 	 * don't try to read sin6->sin6_addr beyond here, since the caller may
441 	 * ask us to overwrite existing sockaddr_in6
442 	 */
443 
444 #ifdef ENABLE_DEFAULT_SCOPE
445 	if (scopeid == 0)
446 		scopeid = scope6_addr2default(in6);
447 #endif
448 
449 	if (IN6_IS_SCOPE_LINKLOCAL(in6)) {
450 		struct in6_pktinfo *pi;
451 
452 		/*
453 		 * KAME assumption: link id == interface id
454 		 */
455 
456 		if (in6p && in6p->in6p_outputopts &&
457 		    (pi = in6p->in6p_outputopts->ip6po_pktinfo) &&
458 		    pi->ipi6_ifindex) {
459 			ifp = ifindex2ifnet[pi->ipi6_ifindex];
460 			in6->s6_addr16[1] = htons(pi->ipi6_ifindex);
461 		} else if (in6p && IN6_IS_ADDR_MULTICAST(in6) &&
462 			   in6p->in6p_moptions &&
463 			   in6p->in6p_moptions->im6o_multicast_ifp) {
464 			ifp = in6p->in6p_moptions->im6o_multicast_ifp;
465 			in6->s6_addr16[1] = htons(ifp->if_index);
466 		} else if (scopeid) {
467 			/* boundary check */
468 			if (scopeid < 0 || if_indexlim <= scopeid ||
469 			    !ifindex2ifnet[scopeid])
470 				return ENXIO;  /* XXX EINVAL? */
471 			ifp = ifindex2ifnet[scopeid];
472 			/* XXX assignment to 16bit from 32bit variable */
473 			in6->s6_addr16[1] = htons(scopeid & 0xffff);
474 		}
475 
476 		if (ifpp)
477 			*ifpp = ifp;
478 	}
479 
480 	return 0;
481 }
482 
483 /*
484  * generate standard sockaddr_in6 from embedded form.
485  * touches sin6_addr and sin6_scope_id only.
486  *
487  * this function should be nuked in the future, when we get rid of
488  * embedded scopeid thing.
489  */
490 int
491 in6_recoverscope(sin6, in6, ifp)
492 	struct sockaddr_in6 *sin6;
493 	const struct in6_addr *in6;
494 	struct ifnet *ifp;
495 {
496 	u_int32_t scopeid;
497 
498 	sin6->sin6_addr = *in6;
499 
500 	/*
501 	 * don't try to read *in6 beyond here, since the caller may
502 	 * ask us to overwrite existing sockaddr_in6
503 	 */
504 
505 	sin6->sin6_scope_id = 0;
506 	if (IN6_IS_SCOPE_LINKLOCAL(in6)) {
507 		/*
508 		 * KAME assumption: link id == interface id
509 		 */
510 		scopeid = ntohs(sin6->sin6_addr.s6_addr16[1]);
511 		if (scopeid) {
512 			/* sanity check */
513 			if (scopeid < 0 || if_indexlim <= scopeid ||
514 			    !ifindex2ifnet[scopeid])
515 				return ENXIO;
516 			if (ifp && ifp->if_index != scopeid)
517 				return ENXIO;
518 			sin6->sin6_addr.s6_addr16[1] = 0;
519 			sin6->sin6_scope_id = scopeid;
520 		}
521 	}
522 
523 	return 0;
524 }
525 
526 /*
527  * just clear the embedded scope identifer.
528  */
529 void
530 in6_clearscope(addr)
531 	struct in6_addr *addr;
532 {
533 	if (IN6_IS_SCOPE_LINKLOCAL(addr))
534 		addr->s6_addr16[1] = 0;
535 }
536