xref: /freebsd/sys/netinet6/in6_pcb.c (revision f9218d3d4fd34f082473b3a021c6d4d109fb47cf)
1 /*	$FreeBSD$	*/
2 /*	$KAME: in6_pcb.c,v 1.31 2001/05/21 05:45:10 jinmei 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 /*
35  * Copyright (c) 1982, 1986, 1991, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. All advertising materials mentioning features or use of this software
47  *    must display the following acknowledgement:
48  *	This product includes software developed by the University of
49  *	California, Berkeley and its contributors.
50  * 4. Neither the name of the University nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64  * SUCH DAMAGE.
65  *
66  *	@(#)in_pcb.c	8.2 (Berkeley) 1/4/94
67  */
68 
69 #include "opt_inet.h"
70 #include "opt_inet6.h"
71 #include "opt_ipsec.h"
72 
73 #include <sys/param.h>
74 #include <sys/systm.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/domain.h>
78 #include <sys/protosw.h>
79 #include <sys/socket.h>
80 #include <sys/socketvar.h>
81 #include <sys/sockio.h>
82 #include <sys/errno.h>
83 #include <sys/time.h>
84 #include <sys/proc.h>
85 #include <sys/jail.h>
86 
87 #include <vm/uma.h>
88 
89 #include <net/if.h>
90 #include <net/if_types.h>
91 #include <net/route.h>
92 
93 #include <netinet/in.h>
94 #include <netinet/in_var.h>
95 #include <netinet/in_systm.h>
96 #include <netinet/tcp_var.h>
97 #include <netinet/ip6.h>
98 #include <netinet/ip_var.h>
99 #include <netinet6/ip6_var.h>
100 #include <netinet6/nd6.h>
101 #include <netinet/in_pcb.h>
102 #include <netinet6/in6_pcb.h>
103 
104 #ifdef IPSEC
105 #include <netinet6/ipsec.h>
106 #ifdef INET6
107 #include <netinet6/ipsec6.h>
108 #endif
109 #include <netinet6/ah.h>
110 #ifdef INET6
111 #include <netinet6/ah6.h>
112 #endif
113 #include <netkey/key.h>
114 #endif /* IPSEC */
115 
116 #ifdef FAST_IPSEC
117 #include <netipsec/ipsec.h>
118 #include <netipsec/ipsec6.h>
119 #include <netipsec/key.h>
120 #define	IPSEC
121 #endif /* FAST_IPSEC */
122 
123 struct	in6_addr zeroin6_addr;
124 
125 int
126 in6_pcbbind(inp, nam, td)
127 	register struct inpcb *inp;
128 	struct sockaddr *nam;
129 	struct thread *td;
130 {
131 	struct socket *so = inp->inp_socket;
132 	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)NULL;
133 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
134 	u_short	lport = 0;
135 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
136 
137 	if (!in6_ifaddr) /* XXX broken! */
138 		return (EADDRNOTAVAIL);
139 	if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
140 		return(EINVAL);
141 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
142 		wild = 1;
143 	if (nam) {
144 		sin6 = (struct sockaddr_in6 *)nam;
145 		if (nam->sa_len != sizeof(*sin6))
146 			return(EINVAL);
147 		/*
148 		 * family check.
149 		 */
150 		if (nam->sa_family != AF_INET6)
151 			return(EAFNOSUPPORT);
152 
153 		/* KAME hack: embed scopeid */
154 		if (in6_embedscope(&sin6->sin6_addr, sin6, inp, NULL) != 0)
155 			return EINVAL;
156 		/* this must be cleared for ifa_ifwithaddr() */
157 		sin6->sin6_scope_id = 0;
158 
159 		lport = sin6->sin6_port;
160 		if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
161 			/*
162 			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
163 			 * allow compepte duplication of binding if
164 			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
165 			 * and a multicast address is bound on both
166 			 * new and duplicated sockets.
167 			 */
168 			if (so->so_options & SO_REUSEADDR)
169 				reuseport = SO_REUSEADDR|SO_REUSEPORT;
170 		} else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
171 			struct ifaddr *ia = NULL;
172 
173 			sin6->sin6_port = 0;		/* yech... */
174 			if ((ia = ifa_ifwithaddr((struct sockaddr *)sin6)) == 0)
175 				return(EADDRNOTAVAIL);
176 
177 			/*
178 			 * XXX: bind to an anycast address might accidentally
179 			 * cause sending a packet with anycast source address.
180 			 * We should allow to bind to a deprecated address, since
181 			 * the application dare to use it.
182 			 */
183 			if (ia &&
184 			    ((struct in6_ifaddr *)ia)->ia6_flags &
185 			    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|IN6_IFF_DETACHED)) {
186 				return(EADDRNOTAVAIL);
187 			}
188 		}
189 		if (lport) {
190 			struct inpcb *t;
191 
192 			/* GROSS */
193 			if (ntohs(lport) < IPV6PORT_RESERVED && td &&
194 			    suser_cred(td->td_ucred, PRISON_ROOT))
195 				return(EACCES);
196 			if (so->so_cred->cr_uid != 0 &&
197 			    !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
198 				t = in6_pcblookup_local(pcbinfo,
199 				    &sin6->sin6_addr, lport,
200 				    INPLOOKUP_WILDCARD);
201 				if (t &&
202 				    (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
203 				     !IN6_IS_ADDR_UNSPECIFIED(&t->in6p_laddr) ||
204 				     (t->inp_socket->so_options &
205 				      SO_REUSEPORT) == 0) &&
206 				    (so->so_cred->cr_uid !=
207 				     t->inp_socket->so_cred->cr_uid))
208 					return (EADDRINUSE);
209 				if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
210 				    IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
211 					struct sockaddr_in sin;
212 
213 					in6_sin6_2_sin(&sin, sin6);
214 					t = in_pcblookup_local(pcbinfo,
215 						sin.sin_addr, lport,
216 						INPLOOKUP_WILDCARD);
217 					if (t &&
218 					    (so->so_cred->cr_uid !=
219 					     t->inp_socket->so_cred->cr_uid) &&
220 					    (ntohl(t->inp_laddr.s_addr) !=
221 					     INADDR_ANY ||
222 					     INP_SOCKAF(so) ==
223 					     INP_SOCKAF(t->inp_socket)))
224 						return (EADDRINUSE);
225 				}
226 			}
227 			t = in6_pcblookup_local(pcbinfo, &sin6->sin6_addr,
228 						lport, wild);
229 			if (t && (reuseport & t->inp_socket->so_options) == 0)
230 				return(EADDRINUSE);
231 			if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
232 			    IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
233 				struct sockaddr_in sin;
234 
235 				in6_sin6_2_sin(&sin, sin6);
236 				t = in_pcblookup_local(pcbinfo, sin.sin_addr,
237 						       lport, wild);
238 				if (t &&
239 				    (reuseport & t->inp_socket->so_options)
240 				    == 0 &&
241 				    (ntohl(t->inp_laddr.s_addr)
242 				     != INADDR_ANY ||
243 				     INP_SOCKAF(so) ==
244 				     INP_SOCKAF(t->inp_socket)))
245 					return (EADDRINUSE);
246 			}
247 		}
248 		inp->in6p_laddr = sin6->sin6_addr;
249 	}
250 	if (lport == 0) {
251 		int e;
252 		if ((e = in6_pcbsetport(&inp->in6p_laddr, inp, td)) != 0)
253 			return(e);
254 	}
255 	else {
256 		inp->inp_lport = lport;
257 		if (in_pcbinshash(inp) != 0) {
258 			inp->in6p_laddr = in6addr_any;
259 			inp->inp_lport = 0;
260 			return (EAGAIN);
261 		}
262 	}
263 	return(0);
264 }
265 
266 /*
267  *   Transform old in6_pcbconnect() into an inner subroutine for new
268  *   in6_pcbconnect(): Do some validity-checking on the remote
269  *   address (in mbuf 'nam') and then determine local host address
270  *   (i.e., which interface) to use to access that remote host.
271  *
272  *   This preserves definition of in6_pcbconnect(), while supporting a
273  *   slightly different version for T/TCP.  (This is more than
274  *   a bit of a kludge, but cleaning up the internal interfaces would
275  *   have forced minor changes in every protocol).
276  */
277 
278 int
279 in6_pcbladdr(inp, nam, plocal_addr6)
280 	register struct inpcb *inp;
281 	struct sockaddr *nam;
282 	struct in6_addr **plocal_addr6;
283 {
284 	register struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
285 	struct ifnet *ifp = NULL;
286 	int error = 0;
287 
288 	if (nam->sa_len != sizeof (*sin6))
289 		return (EINVAL);
290 	if (sin6->sin6_family != AF_INET6)
291 		return (EAFNOSUPPORT);
292 	if (sin6->sin6_port == 0)
293 		return (EADDRNOTAVAIL);
294 
295 	/* KAME hack: embed scopeid */
296 	if (in6_embedscope(&sin6->sin6_addr, sin6, inp, &ifp) != 0)
297 		return EINVAL;
298 
299 	if (in6_ifaddr) {
300 		/*
301 		 * If the destination address is UNSPECIFIED addr,
302 		 * use the loopback addr, e.g ::1.
303 		 */
304 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
305 			sin6->sin6_addr = in6addr_loopback;
306 	}
307 	{
308 		/*
309 		 * XXX: in6_selectsrc might replace the bound local address
310 		 * with the address specified by setsockopt(IPV6_PKTINFO).
311 		 * Is it the intended behavior?
312 		 */
313 		*plocal_addr6 = in6_selectsrc(sin6, inp->in6p_outputopts,
314 					      inp->in6p_moptions,
315 					      &inp->in6p_route,
316 					      &inp->in6p_laddr, &error);
317 		if (*plocal_addr6 == 0) {
318 			if (error == 0)
319 				error = EADDRNOTAVAIL;
320 			return(error);
321 		}
322 		/*
323 		 * Don't do pcblookup call here; return interface in
324 		 * plocal_addr6
325 		 * and exit to caller, that will do the lookup.
326 		 */
327 	}
328 
329 	if (inp->in6p_route.ro_rt)
330 		ifp = inp->in6p_route.ro_rt->rt_ifp;
331 
332 	return(0);
333 }
334 
335 /*
336  * Outer subroutine:
337  * Connect from a socket to a specified address.
338  * Both address and port must be specified in argument sin.
339  * If don't have a local address for this socket yet,
340  * then pick one.
341  */
342 int
343 in6_pcbconnect(inp, nam, td)
344 	register struct inpcb *inp;
345 	struct sockaddr *nam;
346 	struct thread *td;
347 {
348 	struct in6_addr *addr6;
349 	register struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
350 	int error;
351 
352 	/*
353 	 * Call inner routine, to assign local interface address.
354 	 * in6_pcbladdr() may automatically fill in sin6_scope_id.
355 	 */
356 	if ((error = in6_pcbladdr(inp, nam, &addr6)) != 0)
357 		return(error);
358 
359 	if (in6_pcblookup_hash(inp->inp_pcbinfo, &sin6->sin6_addr,
360 			       sin6->sin6_port,
361 			      IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
362 			      ? addr6 : &inp->in6p_laddr,
363 			      inp->inp_lport, 0, NULL) != NULL) {
364 		return (EADDRINUSE);
365 	}
366 	if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
367 		if (inp->inp_lport == 0) {
368 			error = in6_pcbbind(inp, (struct sockaddr *)0, td);
369 			if (error)
370 				return (error);
371 		}
372 		inp->in6p_laddr = *addr6;
373 	}
374 	inp->in6p_faddr = sin6->sin6_addr;
375 	inp->inp_fport = sin6->sin6_port;
376 	/* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
377 	inp->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK;
378 	if (inp->in6p_flags & IN6P_AUTOFLOWLABEL)
379 		inp->in6p_flowinfo |=
380 		    (htonl(ip6_flow_seq++) & IPV6_FLOWLABEL_MASK);
381 
382 	in_pcbrehash(inp);
383 	return (0);
384 }
385 
386 #if 0
387 /*
388  * Return an IPv6 address, which is the most appropriate for given
389  * destination and user specified options.
390  * If necessary, this function lookups the routing table and return
391  * an entry to the caller for later use.
392  */
393 struct in6_addr *
394 in6_selectsrc(dstsock, opts, mopts, ro, laddr, errorp)
395 	struct sockaddr_in6 *dstsock;
396 	struct ip6_pktopts *opts;
397 	struct ip6_moptions *mopts;
398 	struct route_in6 *ro;
399 	struct in6_addr *laddr;
400 	int *errorp;
401 {
402 	struct in6_addr *dst;
403 	struct in6_ifaddr *ia6 = 0;
404 	struct in6_pktinfo *pi = NULL;
405 
406 	dst = &dstsock->sin6_addr;
407 	*errorp = 0;
408 
409 	/*
410 	 * If the source address is explicitly specified by the caller,
411 	 * use it.
412 	 */
413 	if (opts && (pi = opts->ip6po_pktinfo) &&
414 	    !IN6_IS_ADDR_UNSPECIFIED(&pi->ipi6_addr))
415 		return(&pi->ipi6_addr);
416 
417 	/*
418 	 * If the source address is not specified but the socket(if any)
419 	 * is already bound, use the bound address.
420 	 */
421 	if (laddr && !IN6_IS_ADDR_UNSPECIFIED(laddr))
422 		return(laddr);
423 
424 	/*
425 	 * If the caller doesn't specify the source address but
426 	 * the outgoing interface, use an address associated with
427 	 * the interface.
428 	 */
429 	if (pi && pi->ipi6_ifindex) {
430 		/* XXX boundary check is assumed to be already done. */
431 		ia6 = in6_ifawithscope(ifnet_byindex(pi->ipi6_ifindex), dst);
432 		if (ia6 == 0) {
433 			*errorp = EADDRNOTAVAIL;
434 			return(0);
435 		}
436 		return(&satosin6(&ia6->ia_addr)->sin6_addr);
437 	}
438 
439 	/*
440 	 * If the destination address is a link-local unicast address or
441 	 * a multicast address, and if the outgoing interface is specified
442 	 * by the sin6_scope_id filed, use an address associated with the
443 	 * interface.
444 	 * XXX: We're now trying to define more specific semantics of
445 	 *      sin6_scope_id field, so this part will be rewritten in
446 	 *      the near future.
447 	 */
448 	if ((IN6_IS_ADDR_LINKLOCAL(dst) || IN6_IS_ADDR_MULTICAST(dst)) &&
449 	    dstsock->sin6_scope_id) {
450 		/*
451 		 * I'm not sure if boundary check for scope_id is done
452 		 * somewhere...
453 		 */
454 		if (dstsock->sin6_scope_id < 0 ||
455 		    if_index < dstsock->sin6_scope_id) {
456 			*errorp = ENXIO; /* XXX: better error? */
457 			return(0);
458 		}
459 		ia6 = in6_ifawithscope(ifnet_byindex(dstsock->sin6_scope_id),
460 				       dst);
461 		if (ia6 == 0) {
462 			*errorp = EADDRNOTAVAIL;
463 			return(0);
464 		}
465 		return(&satosin6(&ia6->ia_addr)->sin6_addr);
466 	}
467 
468 	/*
469 	 * If the destination address is a multicast address and
470 	 * the outgoing interface for the address is specified
471 	 * by the caller, use an address associated with the interface.
472 	 * There is a sanity check here; if the destination has node-local
473 	 * scope, the outgoing interfacde should be a loopback address.
474 	 * Even if the outgoing interface is not specified, we also
475 	 * choose a loopback interface as the outgoing interface.
476 	 */
477 	if (IN6_IS_ADDR_MULTICAST(dst)) {
478 		struct ifnet *ifp = mopts ? mopts->im6o_multicast_ifp : NULL;
479 
480 		if (ifp == NULL && IN6_IS_ADDR_MC_NODELOCAL(dst)) {
481 			ifp = &loif[0];
482 		}
483 
484 		if (ifp) {
485 			ia6 = in6_ifawithscope(ifp, dst);
486 			if (ia6 == 0) {
487 				*errorp = EADDRNOTAVAIL;
488 				return(0);
489 			}
490 			return(&ia6->ia_addr.sin6_addr);
491 		}
492 	}
493 
494 	/*
495 	 * If the next hop address for the packet is specified
496 	 * by caller, use an address associated with the route
497 	 * to the next hop.
498 	 */
499 	{
500 		struct sockaddr_in6 *sin6_next;
501 		struct rtentry *rt;
502 
503 		if (opts && opts->ip6po_nexthop) {
504 			sin6_next = satosin6(opts->ip6po_nexthop);
505 			rt = nd6_lookup(&sin6_next->sin6_addr, 1, NULL);
506 			if (rt) {
507 				ia6 = in6_ifawithscope(rt->rt_ifp, dst);
508 				if (ia6 == 0)
509 					ia6 = ifatoia6(rt->rt_ifa);
510 			}
511 			if (ia6 == 0) {
512 				*errorp = EADDRNOTAVAIL;
513 				return(0);
514 			}
515 			return(&satosin6(&ia6->ia_addr)->sin6_addr);
516 		}
517 	}
518 
519 	/*
520 	 * If route is known or can be allocated now,
521 	 * our src addr is taken from the i/f, else punt.
522 	 */
523 	if (ro) {
524 		if (ro->ro_rt &&
525 		    !IN6_ARE_ADDR_EQUAL(&satosin6(&ro->ro_dst)->sin6_addr, dst)) {
526 			RTFREE(ro->ro_rt);
527 			ro->ro_rt = (struct rtentry *)0;
528 		}
529 		if (ro->ro_rt == (struct rtentry *)0 ||
530 		    ro->ro_rt->rt_ifp == (struct ifnet *)0) {
531 			struct sockaddr_in6 *dst6;
532 
533 			/* No route yet, so try to acquire one */
534 			bzero(&ro->ro_dst, sizeof(struct sockaddr_in6));
535 			dst6 = (struct sockaddr_in6 *)&ro->ro_dst;
536 			dst6->sin6_family = AF_INET6;
537 			dst6->sin6_len = sizeof(struct sockaddr_in6);
538 			dst6->sin6_addr = *dst;
539 			if (IN6_IS_ADDR_MULTICAST(dst)) {
540 				ro->ro_rt = rtalloc1(&((struct route *)ro)
541 						     ->ro_dst, 0, 0UL);
542 			} else {
543 				rtalloc((struct route *)ro);
544 			}
545 		}
546 
547 		/*
548 		 * in_pcbconnect() checks out IFF_LOOPBACK to skip using
549 		 * the address. But we don't know why it does so.
550 		 * It is necessary to ensure the scope even for lo0
551 		 * so doesn't check out IFF_LOOPBACK.
552 		 */
553 
554 		if (ro->ro_rt) {
555 			ia6 = in6_ifawithscope(ro->ro_rt->rt_ifa->ifa_ifp, dst);
556 			if (ia6 == 0) /* xxx scope error ?*/
557 				ia6 = ifatoia6(ro->ro_rt->rt_ifa);
558 		}
559 		if (ia6 == 0) {
560 			*errorp = EHOSTUNREACH;	/* no route */
561 			return(0);
562 		}
563 		return(&satosin6(&ia6->ia_addr)->sin6_addr);
564 	}
565 
566 	*errorp = EADDRNOTAVAIL;
567 	return(0);
568 }
569 
570 /*
571  * Default hop limit selection. The precedence is as follows:
572  * 1. Hoplimit valued specified via ioctl.
573  * 2. (If the outgoing interface is detected) the current
574  *     hop limit of the interface specified by router advertisement.
575  * 3. The system default hoplimit.
576 */
577 int
578 in6_selecthlim(in6p, ifp)
579 	struct in6pcb *in6p;
580 	struct ifnet *ifp;
581 {
582 	if (in6p && in6p->in6p_hops >= 0)
583 		return(in6p->in6p_hops);
584 	else if (ifp)
585 		return(nd_ifinfo[ifp->if_index].chlim);
586 	else
587 		return(ip6_defhlim);
588 }
589 #endif
590 
591 void
592 in6_pcbdisconnect(inp)
593 	struct inpcb *inp;
594 {
595 	bzero((caddr_t)&inp->in6p_faddr, sizeof(inp->in6p_faddr));
596 	inp->inp_fport = 0;
597 	/* clear flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
598 	inp->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK;
599 	in_pcbrehash(inp);
600 	if (inp->inp_socket->so_state & SS_NOFDREF)
601 		in6_pcbdetach(inp);
602 }
603 
604 void
605 in6_pcbdetach(inp)
606 	struct inpcb *inp;
607 {
608 	struct socket *so = inp->inp_socket;
609 	struct inpcbinfo *ipi = inp->inp_pcbinfo;
610 
611 #ifdef IPSEC
612 	if (inp->in6p_sp != NULL)
613 		ipsec6_delete_pcbpolicy(inp);
614 #endif /* IPSEC */
615 	inp->inp_gencnt = ++ipi->ipi_gencnt;
616 	in_pcbremlists(inp);
617 	if (so) {
618 		so->so_pcb = NULL;
619 		sotryfree(so);
620 	}
621 	if (inp->in6p_options)
622 		m_freem(inp->in6p_options);
623  	ip6_freepcbopts(inp->in6p_outputopts);
624  	ip6_freemoptions(inp->in6p_moptions);
625 	if (inp->in6p_route.ro_rt)
626 		rtfree(inp->in6p_route.ro_rt);
627 	/* Check and free IPv4 related resources in case of mapped addr */
628 	if (inp->inp_options)
629 		(void)m_free(inp->inp_options);
630 	ip_freemoptions(inp->inp_moptions);
631 	inp->inp_vflag = 0;
632 	INP_LOCK_DESTROY(inp);
633 	uma_zfree(ipi->ipi_zone, inp);
634 }
635 
636 struct sockaddr *
637 in6_sockaddr(port, addr_p)
638 	in_port_t port;
639 	struct in6_addr *addr_p;
640 {
641 	struct sockaddr_in6 *sin6;
642 
643 	MALLOC(sin6, struct sockaddr_in6 *, sizeof *sin6, M_SONAME, M_WAITOK);
644 	bzero(sin6, sizeof *sin6);
645 	sin6->sin6_family = AF_INET6;
646 	sin6->sin6_len = sizeof(*sin6);
647 	sin6->sin6_port = port;
648 	sin6->sin6_addr = *addr_p;
649 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
650 		sin6->sin6_scope_id = ntohs(sin6->sin6_addr.s6_addr16[1]);
651 	else
652 		sin6->sin6_scope_id = 0;	/*XXX*/
653 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
654 		sin6->sin6_addr.s6_addr16[1] = 0;
655 
656 	return (struct sockaddr *)sin6;
657 }
658 
659 struct sockaddr *
660 in6_v4mapsin6_sockaddr(port, addr_p)
661 	in_port_t port;
662 	struct in_addr *addr_p;
663 {
664 	struct sockaddr_in sin;
665 	struct sockaddr_in6 *sin6_p;
666 
667 	bzero(&sin, sizeof sin);
668 	sin.sin_family = AF_INET;
669 	sin.sin_len = sizeof(sin);
670 	sin.sin_port = port;
671 	sin.sin_addr = *addr_p;
672 
673 	MALLOC(sin6_p, struct sockaddr_in6 *, sizeof *sin6_p, M_SONAME,
674 		M_WAITOK);
675 	in6_sin_2_v4mapsin6(&sin, sin6_p);
676 
677 	return (struct sockaddr *)sin6_p;
678 }
679 
680 /*
681  * The calling convention of in6_setsockaddr() and in6_setpeeraddr() was
682  * modified to match the pru_sockaddr() and pru_peeraddr() entry points
683  * in struct pr_usrreqs, so that protocols can just reference then directly
684  * without the need for a wrapper function.  The socket must have a valid
685  * (i.e., non-nil) PCB, but it should be impossible to get an invalid one
686  * except through a kernel programming error, so it is acceptable to panic
687  * (or in this case trap) if the PCB is invalid.  (Actually, we don't trap
688  * because there actually /is/ a programming error somewhere... XXX)
689  */
690 int
691 in6_setsockaddr(so, nam)
692 	struct socket *so;
693 	struct sockaddr **nam;
694 {
695 	int s;
696 	register struct inpcb *inp;
697 	struct in6_addr addr;
698 	in_port_t port;
699 
700 	s = splnet();
701 	inp = sotoinpcb(so);
702 	if (!inp) {
703 		splx(s);
704 		return EINVAL;
705 	}
706 	port = inp->inp_lport;
707 	addr = inp->in6p_laddr;
708 	splx(s);
709 
710 	*nam = in6_sockaddr(port, &addr);
711 	return 0;
712 }
713 
714 int
715 in6_setpeeraddr(so, nam)
716 	struct socket *so;
717 	struct sockaddr **nam;
718 {
719 	int s;
720 	struct inpcb *inp;
721 	struct in6_addr addr;
722 	in_port_t port;
723 
724 	s = splnet();
725 	inp = sotoinpcb(so);
726 	if (!inp) {
727 		splx(s);
728 		return EINVAL;
729 	}
730 	port = inp->inp_fport;
731 	addr = inp->in6p_faddr;
732 	splx(s);
733 
734 	*nam = in6_sockaddr(port, &addr);
735 	return 0;
736 }
737 
738 int
739 in6_mapped_sockaddr(struct socket *so, struct sockaddr **nam)
740 {
741 	struct	inpcb *inp = sotoinpcb(so);
742 	int	error;
743 
744 	if (inp == NULL)
745 		return EINVAL;
746 	if (inp->inp_vflag & INP_IPV4) {
747 		error = in_setsockaddr(so, nam, &tcbinfo);
748 		if (error == 0)
749 			in6_sin_2_v4mapsin6_in_sock(nam);
750 	} else
751 	/* scope issues will be handled in in6_setsockaddr(). */
752 	error = in6_setsockaddr(so, nam);
753 
754 	return error;
755 }
756 
757 int
758 in6_mapped_peeraddr(struct socket *so, struct sockaddr **nam)
759 {
760 	struct	inpcb *inp = sotoinpcb(so);
761 	int	error;
762 
763 	if (inp == NULL)
764 		return EINVAL;
765 	if (inp->inp_vflag & INP_IPV4) {
766 		error = in_setpeeraddr(so, nam, &tcbinfo);
767 		if (error == 0)
768 			in6_sin_2_v4mapsin6_in_sock(nam);
769 	} else
770 	/* scope issues will be handled in in6_setpeeraddr(). */
771 	error = in6_setpeeraddr(so, nam);
772 
773 	return error;
774 }
775 
776 /*
777  * Pass some notification to all connections of a protocol
778  * associated with address dst.  The local address and/or port numbers
779  * may be specified to limit the search.  The "usual action" will be
780  * taken, depending on the ctlinput cmd.  The caller must filter any
781  * cmds that are uninteresting (e.g., no error in the map).
782  * Call the protocol specific routine (if any) to report
783  * any errors for each matching socket.
784  *
785  * Must be called at splnet.
786  */
787 void
788 in6_pcbnotify(head, dst, fport_arg, src, lport_arg, cmd, notify)
789 	struct inpcbhead *head;
790 	struct sockaddr *dst;
791 	const struct sockaddr *src;
792 	u_int fport_arg, lport_arg;
793 	int cmd;
794 	struct inpcb *(*notify) __P((struct inpcb *, int));
795 {
796 	struct inpcb *inp, *ninp;
797 	struct sockaddr_in6 sa6_src, *sa6_dst;
798 	u_short	fport = fport_arg, lport = lport_arg;
799 	u_int32_t flowinfo;
800 	int errno, s;
801 
802 	if ((unsigned)cmd > PRC_NCMDS || dst->sa_family != AF_INET6)
803 		return;
804 
805 	sa6_dst = (struct sockaddr_in6 *)dst;
806 	if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr))
807 		return;
808 
809 	/*
810 	 * note that src can be NULL when we get notify by local fragmentation.
811 	 */
812 	sa6_src = (src == NULL) ? sa6_any : *(const struct sockaddr_in6 *)src;
813 	flowinfo = sa6_src.sin6_flowinfo;
814 
815 	/*
816 	 * Redirects go to all references to the destination,
817 	 * and use in6_rtchange to invalidate the route cache.
818 	 * Dead host indications: also use in6_rtchange to invalidate
819 	 * the cache, and deliver the error to all the sockets.
820 	 * Otherwise, if we have knowledge of the local port and address,
821 	 * deliver only to that socket.
822 	 */
823 	if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
824 		fport = 0;
825 		lport = 0;
826 		bzero((caddr_t)&sa6_src.sin6_addr, sizeof(sa6_src.sin6_addr));
827 
828 		if (cmd != PRC_HOSTDEAD)
829 			notify = in6_rtchange;
830 	}
831 	errno = inet6ctlerrmap[cmd];
832 	s = splnet();
833  	for (inp = LIST_FIRST(head); inp != NULL; inp = ninp) {
834  		ninp = LIST_NEXT(inp, inp_list);
835 
836  		if ((inp->inp_vflag & INP_IPV6) == 0)
837 			continue;
838 
839 		/*
840 		 * Detect if we should notify the error. If no source and
841 		 * destination ports are specifed, but non-zero flowinfo and
842 		 * local address match, notify the error. This is the case
843 		 * when the error is delivered with an encrypted buffer
844 		 * by ESP. Otherwise, just compare addresses and ports
845 		 * as usual.
846 		 */
847 		if (lport == 0 && fport == 0 && flowinfo &&
848 		    inp->inp_socket != NULL &&
849 		    flowinfo == (inp->in6p_flowinfo & IPV6_FLOWLABEL_MASK) &&
850 		    IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &sa6_src.sin6_addr))
851 			goto do_notify;
852 		else if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
853 					     &sa6_dst->sin6_addr) ||
854 			 inp->inp_socket == 0 ||
855 			 (lport && inp->inp_lport != lport) ||
856 			 (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) &&
857 			  !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
858 					      &sa6_src.sin6_addr)) ||
859 			 (fport && inp->inp_fport != fport))
860 			continue;
861 
862 	  do_notify:
863 		if (notify)
864 			(*notify)(inp, errno);
865 	}
866 	splx(s);
867 }
868 
869 /*
870  * Lookup a PCB based on the local address and port.
871  */
872 struct inpcb *
873 in6_pcblookup_local(pcbinfo, laddr, lport_arg, wild_okay)
874 	struct inpcbinfo *pcbinfo;
875 	struct in6_addr *laddr;
876 	u_int lport_arg;
877 	int wild_okay;
878 {
879 	register struct inpcb *inp;
880 	int matchwild = 3, wildcard;
881 	u_short lport = lport_arg;
882 
883 	if (!wild_okay) {
884 		struct inpcbhead *head;
885 		/*
886 		 * Look for an unconnected (wildcard foreign addr) PCB that
887 		 * matches the local address and port we're looking for.
888 		 */
889 		head = &pcbinfo->hashbase[INP_PCBHASH(INADDR_ANY, lport, 0,
890 						      pcbinfo->hashmask)];
891 		LIST_FOREACH(inp, head, inp_hash) {
892 			if ((inp->inp_vflag & INP_IPV6) == 0)
893 				continue;
894 			if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
895 			    IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
896 			    inp->inp_lport == lport) {
897 				/*
898 				 * Found.
899 				 */
900 				return (inp);
901 			}
902 		}
903 		/*
904 		 * Not found.
905 		 */
906 		return (NULL);
907 	} else {
908 		struct inpcbporthead *porthash;
909 		struct inpcbport *phd;
910 		struct inpcb *match = NULL;
911 		/*
912 		 * Best fit PCB lookup.
913 		 *
914 		 * First see if this local port is in use by looking on the
915 		 * port hash list.
916 		 */
917 		porthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(lport,
918 		    pcbinfo->porthashmask)];
919 		LIST_FOREACH(phd, porthash, phd_hash) {
920 			if (phd->phd_port == lport)
921 				break;
922 		}
923 		if (phd != NULL) {
924 			/*
925 			 * Port is in use by one or more PCBs. Look for best
926 			 * fit.
927 			 */
928 			LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
929 				wildcard = 0;
930 				if ((inp->inp_vflag & INP_IPV6) == 0)
931 					continue;
932 				if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr))
933 					wildcard++;
934 				if (!IN6_IS_ADDR_UNSPECIFIED(
935 					&inp->in6p_laddr)) {
936 					if (IN6_IS_ADDR_UNSPECIFIED(laddr))
937 						wildcard++;
938 					else if (!IN6_ARE_ADDR_EQUAL(
939 						&inp->in6p_laddr, laddr))
940 						continue;
941 				} else {
942 					if (!IN6_IS_ADDR_UNSPECIFIED(laddr))
943 						wildcard++;
944 				}
945 				if (wildcard < matchwild) {
946 					match = inp;
947 					matchwild = wildcard;
948 					if (matchwild == 0) {
949 						break;
950 					}
951 				}
952 			}
953 		}
954 		return (match);
955 	}
956 }
957 
958 void
959 in6_pcbpurgeif0(head, ifp)
960 	struct in6pcb *head;
961 	struct ifnet *ifp;
962 {
963 	struct in6pcb *in6p;
964 	struct ip6_moptions *im6o;
965 	struct in6_multi_mship *imm, *nimm;
966 
967 	for (in6p = head; in6p != NULL; in6p = LIST_NEXT(in6p, inp_list)) {
968 		im6o = in6p->in6p_moptions;
969 		if ((in6p->inp_vflag & INP_IPV6) &&
970 		    im6o) {
971 			/*
972 			 * Unselect the outgoing interface if it is being
973 			 * detached.
974 			 */
975 			if (im6o->im6o_multicast_ifp == ifp)
976 				im6o->im6o_multicast_ifp = NULL;
977 
978 			/*
979 			 * Drop multicast group membership if we joined
980 			 * through the interface being detached.
981 			 * XXX controversial - is it really legal for kernel
982 			 * to force this?
983 			 */
984 			for (imm = im6o->im6o_memberships.lh_first;
985 			     imm != NULL; imm = nimm) {
986 				nimm = imm->i6mm_chain.le_next;
987 				if (imm->i6mm_maddr->in6m_ifp == ifp) {
988 					LIST_REMOVE(imm, i6mm_chain);
989 					in6_delmulti(imm->i6mm_maddr);
990 					free(imm, M_IPMADDR);
991 				}
992 			}
993 		}
994 	}
995 }
996 
997 /*
998  * Check for alternatives when higher level complains
999  * about service problems.  For now, invalidate cached
1000  * routing information.  If the route was created dynamically
1001  * (by a redirect), time to try a default gateway again.
1002  */
1003 void
1004 in6_losing(in6p)
1005 	struct inpcb *in6p;
1006 {
1007 	struct rtentry *rt;
1008 	struct rt_addrinfo info;
1009 
1010 	if ((rt = in6p->in6p_route.ro_rt) != NULL) {
1011 		bzero((caddr_t)&info, sizeof(info));
1012 		info.rti_flags = rt->rt_flags;
1013 		info.rti_info[RTAX_DST] = rt_key(rt);
1014 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
1015 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
1016 		rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
1017 		if (rt->rt_flags & RTF_DYNAMIC)
1018 			(void)rtrequest1(RTM_DELETE, &info, NULL);
1019 		in6p->in6p_route.ro_rt = NULL;
1020 		rtfree(rt);
1021 		/*
1022 		 * A new route can be allocated
1023 		 * the next time output is attempted.
1024 		 */
1025 	}
1026 }
1027 
1028 /*
1029  * After a routing change, flush old routing
1030  * and allocate a (hopefully) better one.
1031  */
1032 struct inpcb *
1033 in6_rtchange(inp, errno)
1034 	struct inpcb *inp;
1035 	int errno;
1036 {
1037 	if (inp->in6p_route.ro_rt) {
1038 		rtfree(inp->in6p_route.ro_rt);
1039 		inp->in6p_route.ro_rt = 0;
1040 		/*
1041 		 * A new route can be allocated the next time
1042 		 * output is attempted.
1043 		 */
1044 	}
1045 	return inp;
1046 }
1047 
1048 /*
1049  * Lookup PCB in hash list.
1050  */
1051 struct inpcb *
1052 in6_pcblookup_hash(pcbinfo, faddr, fport_arg, laddr, lport_arg, wildcard, ifp)
1053 	struct inpcbinfo *pcbinfo;
1054 	struct in6_addr *faddr, *laddr;
1055 	u_int fport_arg, lport_arg;
1056 	int wildcard;
1057 	struct ifnet *ifp;
1058 {
1059 	struct inpcbhead *head;
1060 	register struct inpcb *inp;
1061 	u_short fport = fport_arg, lport = lport_arg;
1062 	int faith;
1063 
1064 	if (faithprefix_p != NULL)
1065 		faith = (*faithprefix_p)(laddr);
1066 	else
1067 		faith = 0;
1068 
1069 	/*
1070 	 * First look for an exact match.
1071 	 */
1072 	head = &pcbinfo->hashbase[INP_PCBHASH(faddr->s6_addr32[3] /* XXX */,
1073 					      lport, fport,
1074 					      pcbinfo->hashmask)];
1075 	LIST_FOREACH(inp, head, inp_hash) {
1076 		if ((inp->inp_vflag & INP_IPV6) == 0)
1077 			continue;
1078 		if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) &&
1079 		    IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
1080 		    inp->inp_fport == fport &&
1081 		    inp->inp_lport == lport) {
1082 			/*
1083 			 * Found.
1084 			 */
1085 			return (inp);
1086 		}
1087 	}
1088 	if (wildcard) {
1089 		struct inpcb *local_wild = NULL;
1090 
1091 		head = &pcbinfo->hashbase[INP_PCBHASH(INADDR_ANY, lport, 0,
1092 						      pcbinfo->hashmask)];
1093 		LIST_FOREACH(inp, head, inp_hash) {
1094 			if ((inp->inp_vflag & INP_IPV6) == 0)
1095 				continue;
1096 			if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1097 			    inp->inp_lport == lport) {
1098 				if (faith && (inp->inp_flags & INP_FAITH) == 0)
1099 					continue;
1100 				if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
1101 						       laddr))
1102 					return (inp);
1103 				else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
1104 					local_wild = inp;
1105 			}
1106 		}
1107 		return (local_wild);
1108 	}
1109 
1110 	/*
1111 	 * Not found.
1112 	 */
1113 	return (NULL);
1114 }
1115 
1116 void
1117 init_sin6(struct sockaddr_in6 *sin6, struct mbuf *m)
1118 {
1119 	struct ip6_hdr *ip;
1120 
1121 	ip = mtod(m, struct ip6_hdr *);
1122 	bzero(sin6, sizeof(*sin6));
1123 	sin6->sin6_len = sizeof(*sin6);
1124 	sin6->sin6_family = AF_INET6;
1125 	sin6->sin6_addr = ip->ip6_src;
1126 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
1127 		sin6->sin6_addr.s6_addr16[1] = 0;
1128 	sin6->sin6_scope_id =
1129 		(m->m_pkthdr.rcvif && IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
1130 		? m->m_pkthdr.rcvif->if_index : 0;
1131 
1132 	return;
1133 }
1134