xref: /freebsd/sys/netinet6/in6_src.c (revision 41466b50c1d5bfd1cf6adaae547a579a75d7c04e)
1 /*	$FreeBSD$	*/
2 /*	$KAME: in6_src.c,v 1.37 2001/03/29 05:34:31 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. All advertising materials mentioning features or use of this software
46  *    must display the following acknowledgement:
47  *	This product includes software developed by the University of
48  *	California, Berkeley and its contributors.
49  * 4. Neither the name of the University nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63  * SUCH DAMAGE.
64  *
65  *	@(#)in_pcb.c	8.2 (Berkeley) 1/4/94
66  */
67 
68 #include "opt_inet.h"
69 #include "opt_inet6.h"
70 
71 #include <sys/param.h>
72 #include <sys/systm.h>
73 #include <sys/malloc.h>
74 #include <sys/mbuf.h>
75 #include <sys/protosw.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/errno.h>
79 #include <sys/time.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(ifnet_byindex(pi->ipi6_ifindex), dst);
145 		if (ia6 == 0) {
146 			*errorp = EADDRNOTAVAIL;
147 			return(0);
148 		}
149 		return(&satosin6(&ia6->ia_addr)->sin6_addr);
150 	}
151 
152 	/*
153 	 * If the destination address is a link-local unicast address or
154 	 * a multicast address, and if the outgoing interface is specified
155 	 * by the sin6_scope_id filed, use an address associated with the
156 	 * interface.
157 	 * XXX: We're now trying to define more specific semantics of
158 	 *      sin6_scope_id field, so this part will be rewritten in
159 	 *      the near future.
160 	 */
161 	if ((IN6_IS_ADDR_LINKLOCAL(dst) || IN6_IS_ADDR_MULTICAST(dst)) &&
162 	    dstsock->sin6_scope_id) {
163 		/*
164 		 * I'm not sure if boundary check for scope_id is done
165 		 * somewhere...
166 		 */
167 		if (dstsock->sin6_scope_id < 0 ||
168 		    if_index < dstsock->sin6_scope_id) {
169 			*errorp = ENXIO; /* XXX: better error? */
170 			return(0);
171 		}
172 		ia6 = in6_ifawithscope(ifnet_byindex(dstsock->sin6_scope_id),
173 				       dst);
174 		if (ia6 == 0) {
175 			*errorp = EADDRNOTAVAIL;
176 			return(0);
177 		}
178 		return(&satosin6(&ia6->ia_addr)->sin6_addr);
179 	}
180 
181 	/*
182 	 * If the destination address is a multicast address and
183 	 * the outgoing interface for the address is specified
184 	 * by the caller, use an address associated with the interface.
185 	 * There is a sanity check here; if the destination has node-local
186 	 * scope, the outgoing interfacde should be a loopback address.
187 	 * Even if the outgoing interface is not specified, we also
188 	 * choose a loopback interface as the outgoing interface.
189 	 */
190 	if (IN6_IS_ADDR_MULTICAST(dst)) {
191 		struct ifnet *ifp = mopts ? mopts->im6o_multicast_ifp : NULL;
192 
193 		if (ifp == NULL && IN6_IS_ADDR_MC_NODELOCAL(dst)) {
194 			ifp = &loif[0];
195 		}
196 
197 		if (ifp) {
198 			ia6 = in6_ifawithscope(ifp, dst);
199 			if (ia6 == 0) {
200 				*errorp = EADDRNOTAVAIL;
201 				return(0);
202 			}
203 			return(&satosin6(&ia6->ia_addr)->sin6_addr);
204 		}
205 	}
206 
207 	/*
208 	 * If the next hop address for the packet is specified
209 	 * by caller, use an address associated with the route
210 	 * to the next hop.
211 	 */
212 	{
213 		struct sockaddr_in6 *sin6_next;
214 		struct rtentry *rt;
215 
216 		if (opts && opts->ip6po_nexthop) {
217 			sin6_next = satosin6(opts->ip6po_nexthop);
218 			rt = nd6_lookup(&sin6_next->sin6_addr, 1, NULL);
219 			if (rt) {
220 				ia6 = in6_ifawithscope(rt->rt_ifp, dst);
221 				if (ia6 == 0)
222 					ia6 = ifatoia6(rt->rt_ifa);
223 			}
224 			if (ia6 == 0) {
225 				*errorp = EADDRNOTAVAIL;
226 				return(0);
227 			}
228 			return(&satosin6(&ia6->ia_addr)->sin6_addr);
229 		}
230 	}
231 
232 	/*
233 	 * If route is known or can be allocated now,
234 	 * our src addr is taken from the i/f, else punt.
235 	 */
236 	if (ro) {
237 		if (ro->ro_rt &&
238 		    !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr, dst)) {
239 			RTFREE(ro->ro_rt);
240 			ro->ro_rt = (struct rtentry *)0;
241 		}
242 		if (ro->ro_rt == (struct rtentry *)0 ||
243 		    ro->ro_rt->rt_ifp == (struct ifnet *)0) {
244 			struct sockaddr_in6 *sa6;
245 
246 			/* No route yet, so try to acquire one */
247 			bzero(&ro->ro_dst, sizeof(struct sockaddr_in6));
248 			sa6 = (struct sockaddr_in6 *)&ro->ro_dst;
249 			sa6->sin6_family = AF_INET6;
250 			sa6->sin6_len = sizeof(struct sockaddr_in6);
251 			sa6->sin6_addr = *dst;
252 			sa6->sin6_scope_id = dstsock->sin6_scope_id;
253 			if (IN6_IS_ADDR_MULTICAST(dst)) {
254 				ro->ro_rt = rtalloc1(&((struct route *)ro)
255 						     ->ro_dst, 0, 0UL);
256 			} else {
257 				rtalloc((struct route *)ro);
258 			}
259 		}
260 
261 		/*
262 		 * in_pcbconnect() checks out IFF_LOOPBACK to skip using
263 		 * the address. But we don't know why it does so.
264 		 * It is necessary to ensure the scope even for lo0
265 		 * so doesn't check out IFF_LOOPBACK.
266 		 */
267 
268 		if (ro->ro_rt) {
269 			ia6 = in6_ifawithscope(ro->ro_rt->rt_ifa->ifa_ifp, dst);
270 			if (ia6 == 0) /* xxx scope error ?*/
271 				ia6 = ifatoia6(ro->ro_rt->rt_ifa);
272 		}
273 #if 0
274 		/*
275 		 * xxx The followings are necessary? (kazu)
276 		 * I don't think so.
277 		 * It's for SO_DONTROUTE option in IPv4.(jinmei)
278 		 */
279 		if (ia6 == 0) {
280 			struct sockaddr_in6 sin6 = {sizeof(sin6), AF_INET6, 0};
281 
282 			sin6->sin6_addr = *dst;
283 
284 			ia6 = ifatoia6(ifa_ifwithdstaddr(sin6tosa(&sin6)));
285 			if (ia6 == 0)
286 				ia6 = ifatoia6(ifa_ifwithnet(sin6tosa(&sin6)));
287 			if (ia6 == 0)
288 				return(0);
289 			return(&satosin6(&ia6->ia_addr)->sin6_addr);
290 		}
291 #endif /* 0 */
292 		if (ia6 == 0) {
293 			*errorp = EHOSTUNREACH;	/* no route */
294 			return(0);
295 		}
296 		return(&satosin6(&ia6->ia_addr)->sin6_addr);
297 	}
298 
299 	*errorp = EADDRNOTAVAIL;
300 	return(0);
301 }
302 
303 /*
304  * Default hop limit selection. The precedence is as follows:
305  * 1. Hoplimit value specified via ioctl.
306  * 2. (If the outgoing interface is detected) the current
307  *     hop limit of the interface specified by router advertisement.
308  * 3. The system default hoplimit.
309 */
310 int
311 in6_selecthlim(in6p, ifp)
312 	struct in6pcb *in6p;
313 	struct ifnet *ifp;
314 {
315 	if (in6p && in6p->in6p_hops >= 0)
316 		return(in6p->in6p_hops);
317 	else if (ifp)
318 		return(nd_ifinfo[ifp->if_index].chlim);
319 	else
320 		return(ip6_defhlim);
321 }
322 
323 /*
324  * XXX: this is borrowed from in6_pcbbind(). If possible, we should
325  * share this function by all *bsd*...
326  */
327 int
328 in6_pcbsetport(laddr, inp, td)
329 	struct in6_addr *laddr;
330 	struct inpcb *inp;
331 	struct thread *td;
332 {
333 	struct socket *so = inp->inp_socket;
334 	u_int16_t lport = 0, first, last, *lastport;
335 	int count, error = 0, wild = 0;
336 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
337 
338 	/* XXX: this is redundant when called from in6_pcbbind */
339 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
340 		wild = INPLOOKUP_WILDCARD;
341 
342 	inp->inp_flags |= INP_ANONPORT;
343 
344 	if (inp->inp_flags & INP_HIGHPORT) {
345 		first = ipport_hifirstauto;	/* sysctl */
346 		last  = ipport_hilastauto;
347 		lastport = &pcbinfo->lasthi;
348 	} else if (inp->inp_flags & INP_LOWPORT) {
349 		if (td && (error = suser_td(td)))
350 			return error;
351 		first = ipport_lowfirstauto;	/* 1023 */
352 		last  = ipport_lowlastauto;	/* 600 */
353 		lastport = &pcbinfo->lastlow;
354 	} else {
355 		first = ipport_firstauto;	/* sysctl */
356 		last  = ipport_lastauto;
357 		lastport = &pcbinfo->lastport;
358 	}
359 	/*
360 	 * Simple check to ensure all ports are not used up causing
361 	 * a deadlock here.
362 	 *
363 	 * We split the two cases (up and down) so that the direction
364 	 * is not being tested on each round of the loop.
365 	 */
366 	if (first > last) {
367 		/*
368 		 * counting down
369 		 */
370 		count = first - last;
371 
372 		do {
373 			if (count-- < 0) {	/* completely used? */
374 				/*
375 				 * Undo any address bind that may have
376 				 * occurred above.
377 				 */
378 				inp->in6p_laddr = in6addr_any;
379 				return (EAGAIN);
380 			}
381 			--*lastport;
382 			if (*lastport > first || *lastport < last)
383 				*lastport = first;
384 			lport = htons(*lastport);
385 		} while (in6_pcblookup_local(pcbinfo,
386 					     &inp->in6p_laddr, lport, wild));
387 	} else {
388 		/*
389 			 * counting up
390 			 */
391 		count = last - first;
392 
393 		do {
394 			if (count-- < 0) {	/* completely used? */
395 				/*
396 				 * Undo any address bind that may have
397 				 * occurred above.
398 				 */
399 				inp->in6p_laddr = in6addr_any;
400 				return (EAGAIN);
401 			}
402 			++*lastport;
403 			if (*lastport < first || *lastport > last)
404 				*lastport = first;
405 			lport = htons(*lastport);
406 		} while (in6_pcblookup_local(pcbinfo,
407 					     &inp->in6p_laddr, lport, wild));
408 	}
409 
410 	inp->inp_lport = lport;
411 	if (in_pcbinshash(inp) != 0) {
412 		inp->in6p_laddr = in6addr_any;
413 		inp->inp_lport = 0;
414 		return (EAGAIN);
415 	}
416 
417 	return(0);
418 }
419 
420 /*
421  * generate kernel-internal form (scopeid embedded into s6_addr16[1]).
422  * If the address scope of is link-local, embed the interface index in the
423  * address.  The routine determines our precedence
424  * between advanced API scope/interface specification and basic API
425  * specification.
426  *
427  * this function should be nuked in the future, when we get rid of
428  * embedded scopeid thing.
429  *
430  * XXX actually, it is over-specification to return ifp against sin6_scope_id.
431  * there can be multiple interfaces that belong to a particular scope zone
432  * (in specification, we have 1:N mapping between a scope zone and interfaces).
433  * we may want to change the function to return something other than ifp.
434  */
435 int
436 in6_embedscope(in6, sin6, in6p, ifpp)
437 	struct in6_addr *in6;
438 	const struct sockaddr_in6 *sin6;
439 #ifdef HAVE_NRL_INPCB
440 	struct inpcb *in6p;
441 #define in6p_outputopts	inp_outputopts6
442 #define in6p_moptions	inp_moptions6
443 #else
444 	struct in6pcb *in6p;
445 #endif
446 	struct ifnet **ifpp;
447 {
448 	struct ifnet *ifp = NULL;
449 	u_int32_t scopeid;
450 
451 	*in6 = sin6->sin6_addr;
452 	scopeid = sin6->sin6_scope_id;
453 	if (ifpp)
454 		*ifpp = NULL;
455 
456 	/*
457 	 * don't try to read sin6->sin6_addr beyond here, since the caller may
458 	 * ask us to overwrite existing sockaddr_in6
459 	 */
460 
461 #ifdef ENABLE_DEFAULT_SCOPE
462 	if (scopeid == 0)
463 		scopeid = scope6_addr2default(in6);
464 #endif
465 
466 	if (IN6_IS_SCOPE_LINKLOCAL(in6)) {
467 		struct in6_pktinfo *pi;
468 
469 		/*
470 		 * KAME assumption: link id == interface id
471 		 */
472 
473 		if (in6p && in6p->in6p_outputopts &&
474 		    (pi = in6p->in6p_outputopts->ip6po_pktinfo) &&
475 		    pi->ipi6_ifindex) {
476 			ifp = ifnet_byindex(pi->ipi6_ifindex);
477 			in6->s6_addr16[1] = htons(pi->ipi6_ifindex);
478 		} else if (in6p && IN6_IS_ADDR_MULTICAST(in6) &&
479 			   in6p->in6p_moptions &&
480 			   in6p->in6p_moptions->im6o_multicast_ifp) {
481 			ifp = in6p->in6p_moptions->im6o_multicast_ifp;
482 			in6->s6_addr16[1] = htons(ifp->if_index);
483 		} else if (scopeid) {
484 			/* boundary check */
485 			if (scopeid < 0 || if_index < scopeid)
486 				return ENXIO;  /* XXX EINVAL? */
487 			ifp = ifnet_byindex(scopeid);
488 			/*XXX assignment to 16bit from 32bit variable */
489 			in6->s6_addr16[1] = htons(scopeid & 0xffff);
490 		}
491 
492 		if (ifpp)
493 			*ifpp = ifp;
494 	}
495 
496 	return 0;
497 }
498 #ifdef HAVE_NRL_INPCB
499 #undef in6p_outputopts
500 #undef in6p_moptions
501 #endif
502 
503 /*
504  * generate standard sockaddr_in6 from embedded form.
505  * touches sin6_addr and sin6_scope_id only.
506  *
507  * this function should be nuked in the future, when we get rid of
508  * embedded scopeid thing.
509  */
510 int
511 in6_recoverscope(sin6, in6, ifp)
512 	struct sockaddr_in6 *sin6;
513 	const struct in6_addr *in6;
514 	struct ifnet *ifp;
515 {
516 	u_int32_t scopeid;
517 
518 	sin6->sin6_addr = *in6;
519 
520 	/*
521 	 * don't try to read *in6 beyond here, since the caller may
522 	 * ask us to overwrite existing sockaddr_in6
523 	 */
524 
525 	sin6->sin6_scope_id = 0;
526 	if (IN6_IS_SCOPE_LINKLOCAL(in6)) {
527 		/*
528 		 * KAME assumption: link id == interface id
529 		 */
530 		scopeid = ntohs(sin6->sin6_addr.s6_addr16[1]);
531 		if (scopeid) {
532 			/* sanity check */
533 			if (scopeid < 0 || if_index < scopeid)
534 				return ENXIO;
535 			if (ifp && ifp->if_index != scopeid)
536 				return ENXIO;
537 			sin6->sin6_addr.s6_addr16[1] = 0;
538 			sin6->sin6_scope_id = scopeid;
539 		}
540 	}
541 
542 	return 0;
543 }
544 
545 /*
546  * just clear the embedded scope identifer.
547  * XXX: currently used for bsdi4 only as a supplement function.
548  */
549 void
550 in6_clearscope(addr)
551 	struct in6_addr *addr;
552 {
553 	if (IN6_IS_SCOPE_LINKLOCAL(addr))
554 		addr->s6_addr16[1] = 0;
555 }
556