xref: /freebsd/sys/netinet6/in6_src.c (revision 1b6c76a2fe091c74f08427e6c870851025a9cf67)
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(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_MULTICAST(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_index < dstsock->sin6_scope_id) {
170 			*errorp = ENXIO; /* XXX: better error? */
171 			return(0);
172 		}
173 		ia6 = in6_ifawithscope(ifindex2ifnet[dstsock->sin6_scope_id],
174 				       dst);
175 		if (ia6 == 0) {
176 			*errorp = EADDRNOTAVAIL;
177 			return(0);
178 		}
179 		return(&satosin6(&ia6->ia_addr)->sin6_addr);
180 	}
181 
182 	/*
183 	 * If the destination address is a multicast address and
184 	 * the outgoing interface for the address is specified
185 	 * by the caller, use an address associated with the interface.
186 	 * There is a sanity check here; if the destination has node-local
187 	 * scope, the outgoing interfacde should be a loopback address.
188 	 * Even if the outgoing interface is not specified, we also
189 	 * choose a loopback interface as the outgoing interface.
190 	 */
191 	if (IN6_IS_ADDR_MULTICAST(dst)) {
192 		struct ifnet *ifp = mopts ? mopts->im6o_multicast_ifp : NULL;
193 
194 		if (ifp == NULL && IN6_IS_ADDR_MC_NODELOCAL(dst)) {
195 			ifp = &loif[0];
196 		}
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 	 */
237 	if (ro) {
238 		if (ro->ro_rt &&
239 		    !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr, dst)) {
240 			RTFREE(ro->ro_rt);
241 			ro->ro_rt = (struct rtentry *)0;
242 		}
243 		if (ro->ro_rt == (struct rtentry *)0 ||
244 		    ro->ro_rt->rt_ifp == (struct ifnet *)0) {
245 			struct sockaddr_in6 *sa6;
246 
247 			/* No route yet, so try to acquire one */
248 			bzero(&ro->ro_dst, sizeof(struct sockaddr_in6));
249 			sa6 = (struct sockaddr_in6 *)&ro->ro_dst;
250 			sa6->sin6_family = AF_INET6;
251 			sa6->sin6_len = sizeof(struct sockaddr_in6);
252 			sa6->sin6_addr = *dst;
253 			sa6->sin6_scope_id = dstsock->sin6_scope_id;
254 			if (IN6_IS_ADDR_MULTICAST(dst)) {
255 				ro->ro_rt = rtalloc1(&((struct route *)ro)
256 						     ->ro_dst, 0, 0UL);
257 			} else {
258 				rtalloc((struct route *)ro);
259 			}
260 		}
261 
262 		/*
263 		 * in_pcbconnect() checks out IFF_LOOPBACK to skip using
264 		 * the address. But we don't know why it does so.
265 		 * It is necessary to ensure the scope even for lo0
266 		 * so doesn't check out IFF_LOOPBACK.
267 		 */
268 
269 		if (ro->ro_rt) {
270 			ia6 = in6_ifawithscope(ro->ro_rt->rt_ifa->ifa_ifp, dst);
271 			if (ia6 == 0) /* xxx scope error ?*/
272 				ia6 = ifatoia6(ro->ro_rt->rt_ifa);
273 		}
274 #if 0
275 		/*
276 		 * xxx The followings are necessary? (kazu)
277 		 * I don't think so.
278 		 * It's for SO_DONTROUTE option in IPv4.(jinmei)
279 		 */
280 		if (ia6 == 0) {
281 			struct sockaddr_in6 sin6 = {sizeof(sin6), AF_INET6, 0};
282 
283 			sin6->sin6_addr = *dst;
284 
285 			ia6 = ifatoia6(ifa_ifwithdstaddr(sin6tosa(&sin6)));
286 			if (ia6 == 0)
287 				ia6 = ifatoia6(ifa_ifwithnet(sin6tosa(&sin6)));
288 			if (ia6 == 0)
289 				return(0);
290 			return(&satosin6(&ia6->ia_addr)->sin6_addr);
291 		}
292 #endif /* 0 */
293 		if (ia6 == 0) {
294 			*errorp = EHOSTUNREACH;	/* no route */
295 			return(0);
296 		}
297 		return(&satosin6(&ia6->ia_addr)->sin6_addr);
298 	}
299 
300 	*errorp = EADDRNOTAVAIL;
301 	return(0);
302 }
303 
304 /*
305  * Default hop limit selection. The precedence is as follows:
306  * 1. Hoplimit value specified via ioctl.
307  * 2. (If the outgoing interface is detected) the current
308  *     hop limit of the interface specified by router advertisement.
309  * 3. The system default hoplimit.
310 */
311 int
312 in6_selecthlim(in6p, ifp)
313 	struct in6pcb *in6p;
314 	struct ifnet *ifp;
315 {
316 	if (in6p && in6p->in6p_hops >= 0)
317 		return(in6p->in6p_hops);
318 	else if (ifp)
319 		return(nd_ifinfo[ifp->if_index].chlim);
320 	else
321 		return(ip6_defhlim);
322 }
323 
324 /*
325  * XXX: this is borrowed from in6_pcbbind(). If possible, we should
326  * share this function by all *bsd*...
327  */
328 int
329 in6_pcbsetport(laddr, inp, p)
330 	struct in6_addr *laddr;
331 	struct inpcb *inp;
332 	struct proc *p;
333 {
334 	struct socket *so = inp->inp_socket;
335 	u_int16_t lport = 0, first, last, *lastport;
336 	int count, error = 0, wild = 0;
337 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
338 
339 	/* XXX: this is redundant when called from in6_pcbbind */
340 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
341 		wild = INPLOOKUP_WILDCARD;
342 
343 	inp->inp_flags |= INP_ANONPORT;
344 
345 	if (inp->inp_flags & INP_HIGHPORT) {
346 		first = ipport_hifirstauto;	/* sysctl */
347 		last  = ipport_hilastauto;
348 		lastport = &pcbinfo->lasthi;
349 	} else if (inp->inp_flags & INP_LOWPORT) {
350 		if (p && (error = suser(p)))
351 			return error;
352 		first = ipport_lowfirstauto;	/* 1023 */
353 		last  = ipport_lowlastauto;	/* 600 */
354 		lastport = &pcbinfo->lastlow;
355 	} else {
356 		first = ipport_firstauto;	/* sysctl */
357 		last  = ipport_lastauto;
358 		lastport = &pcbinfo->lastport;
359 	}
360 	/*
361 	 * Simple check to ensure all ports are not used up causing
362 	 * a deadlock here.
363 	 *
364 	 * We split the two cases (up and down) so that the direction
365 	 * is not being tested on each round of the loop.
366 	 */
367 	if (first > last) {
368 		/*
369 		 * counting down
370 		 */
371 		count = first - last;
372 
373 		do {
374 			if (count-- < 0) {	/* completely used? */
375 				/*
376 				 * Undo any address bind that may have
377 				 * occurred above.
378 				 */
379 				inp->in6p_laddr = in6addr_any;
380 				return (EAGAIN);
381 			}
382 			--*lastport;
383 			if (*lastport > first || *lastport < last)
384 				*lastport = first;
385 			lport = htons(*lastport);
386 		} while (in6_pcblookup_local(pcbinfo,
387 					     &inp->in6p_laddr, lport, wild));
388 	} else {
389 		/*
390 			 * counting up
391 			 */
392 		count = last - first;
393 
394 		do {
395 			if (count-- < 0) {	/* completely used? */
396 				/*
397 				 * Undo any address bind that may have
398 				 * occurred above.
399 				 */
400 				inp->in6p_laddr = in6addr_any;
401 				return (EAGAIN);
402 			}
403 			++*lastport;
404 			if (*lastport < first || *lastport > last)
405 				*lastport = first;
406 			lport = htons(*lastport);
407 		} while (in6_pcblookup_local(pcbinfo,
408 					     &inp->in6p_laddr, lport, wild));
409 	}
410 
411 	inp->inp_lport = lport;
412 	if (in_pcbinshash(inp) != 0) {
413 		inp->in6p_laddr = in6addr_any;
414 		inp->inp_lport = 0;
415 		return (EAGAIN);
416 	}
417 
418 	return(0);
419 }
420 
421 /*
422  * generate kernel-internal form (scopeid embedded into s6_addr16[1]).
423  * If the address scope of is link-local, embed the interface index in the
424  * address.  The routine determines our precedence
425  * between advanced API scope/interface specification and basic API
426  * specification.
427  *
428  * this function should be nuked in the future, when we get rid of
429  * embedded scopeid thing.
430  *
431  * XXX actually, it is over-specification to return ifp against sin6_scope_id.
432  * there can be multiple interfaces that belong to a particular scope zone
433  * (in specification, we have 1:N mapping between a scope zone and interfaces).
434  * we may want to change the function to return something other than ifp.
435  */
436 int
437 in6_embedscope(in6, sin6, in6p, ifpp)
438 	struct in6_addr *in6;
439 	const struct sockaddr_in6 *sin6;
440 #ifdef HAVE_NRL_INPCB
441 	struct inpcb *in6p;
442 #define in6p_outputopts	inp_outputopts6
443 #define in6p_moptions	inp_moptions6
444 #else
445 	struct in6pcb *in6p;
446 #endif
447 	struct ifnet **ifpp;
448 {
449 	struct ifnet *ifp = NULL;
450 	u_int32_t scopeid;
451 
452 	*in6 = sin6->sin6_addr;
453 	scopeid = sin6->sin6_scope_id;
454 	if (ifpp)
455 		*ifpp = NULL;
456 
457 	/*
458 	 * don't try to read sin6->sin6_addr beyond here, since the caller may
459 	 * ask us to overwrite existing sockaddr_in6
460 	 */
461 
462 #ifdef ENABLE_DEFAULT_SCOPE
463 	if (scopeid == 0)
464 		scopeid = scope6_addr2default(in6);
465 #endif
466 
467 	if (IN6_IS_SCOPE_LINKLOCAL(in6)) {
468 		struct in6_pktinfo *pi;
469 
470 		/*
471 		 * KAME assumption: link id == interface id
472 		 */
473 
474 		if (in6p && in6p->in6p_outputopts &&
475 		    (pi = in6p->in6p_outputopts->ip6po_pktinfo) &&
476 		    pi->ipi6_ifindex) {
477 			ifp = ifindex2ifnet[pi->ipi6_ifindex];
478 			in6->s6_addr16[1] = htons(pi->ipi6_ifindex);
479 		} else if (in6p && IN6_IS_ADDR_MULTICAST(in6) &&
480 			   in6p->in6p_moptions &&
481 			   in6p->in6p_moptions->im6o_multicast_ifp) {
482 			ifp = in6p->in6p_moptions->im6o_multicast_ifp;
483 			in6->s6_addr16[1] = htons(ifp->if_index);
484 		} else if (scopeid) {
485 			/* boundary check */
486 			if (scopeid < 0 || if_index < scopeid)
487 				return ENXIO;  /* XXX EINVAL? */
488 			ifp = ifindex2ifnet[scopeid];
489 			/*XXX assignment to 16bit from 32bit variable */
490 			in6->s6_addr16[1] = htons(scopeid & 0xffff);
491 		}
492 
493 		if (ifpp)
494 			*ifpp = ifp;
495 	}
496 
497 	return 0;
498 }
499 #ifdef HAVE_NRL_INPCB
500 #undef in6p_outputopts
501 #undef in6p_moptions
502 #endif
503 
504 /*
505  * generate standard sockaddr_in6 from embedded form.
506  * touches sin6_addr and sin6_scope_id only.
507  *
508  * this function should be nuked in the future, when we get rid of
509  * embedded scopeid thing.
510  */
511 int
512 in6_recoverscope(sin6, in6, ifp)
513 	struct sockaddr_in6 *sin6;
514 	const struct in6_addr *in6;
515 	struct ifnet *ifp;
516 {
517 	u_int32_t scopeid;
518 
519 	sin6->sin6_addr = *in6;
520 
521 	/*
522 	 * don't try to read *in6 beyond here, since the caller may
523 	 * ask us to overwrite existing sockaddr_in6
524 	 */
525 
526 	sin6->sin6_scope_id = 0;
527 	if (IN6_IS_SCOPE_LINKLOCAL(in6)) {
528 		/*
529 		 * KAME assumption: link id == interface id
530 		 */
531 		scopeid = ntohs(sin6->sin6_addr.s6_addr16[1]);
532 		if (scopeid) {
533 			/* sanity check */
534 			if (scopeid < 0 || if_index < scopeid)
535 				return ENXIO;
536 			if (ifp && ifp->if_index != scopeid)
537 				return ENXIO;
538 			sin6->sin6_addr.s6_addr16[1] = 0;
539 			sin6->sin6_scope_id = scopeid;
540 		}
541 	}
542 
543 	return 0;
544 }
545 
546 /*
547  * just clear the embedded scope identifer.
548  * XXX: currently used for bsdi4 only as a supplement function.
549  */
550 void
551 in6_clearscope(addr)
552 	struct in6_addr *addr;
553 {
554 	if (IN6_IS_SCOPE_LINKLOCAL(addr))
555 		addr->s6_addr16[1] = 0;
556 }
557