in6_pcb.c revision 263198
1/*-
2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3 * Copyright (c) 2010-2011 Juniper Networks, Inc.
4 * All rights reserved.
5 *
6 * Portions of this software were developed by Robert N. M. Watson under
7 * contract to Juniper Networks, Inc.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 * 1. Redistributions of source code must retain the above copyright
13 *    notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 *    notice, this list of conditions and the following disclaimer in the
16 *    documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the project nor the names of its contributors
18 *    may be used to endorse or promote products derived from this software
19 *    without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 *	$KAME: in6_pcb.c,v 1.31 2001/05/21 05:45:10 jinmei Exp $
34 */
35
36/*-
37 * Copyright (c) 1982, 1986, 1991, 1993
38 *	The Regents of the University of California.  All rights reserved.
39 *
40 * Redistribution and use in source and binary forms, with or without
41 * modification, are permitted provided that the following conditions
42 * are met:
43 * 1. Redistributions of source code must retain the above copyright
44 *    notice, this list of conditions and the following disclaimer.
45 * 2. Redistributions in binary form must reproduce the above copyright
46 *    notice, this list of conditions and the following disclaimer in the
47 *    documentation and/or other materials provided with the distribution.
48 * 4. Neither the name of the University nor the names of its contributors
49 *    may be used to endorse or promote products derived from this software
50 *    without specific prior written permission.
51 *
52 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
53 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
54 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
55 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
56 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
57 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
58 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
59 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
60 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
61 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
62 * SUCH DAMAGE.
63 *
64 *	@(#)in_pcb.c	8.2 (Berkeley) 1/4/94
65 */
66
67#include <sys/cdefs.h>
68__FBSDID("$FreeBSD: head/sys/netinet6/in6_pcb.c 263198 2014-03-15 00:57:50Z rwatson $");
69
70#include "opt_inet.h"
71#include "opt_inet6.h"
72#include "opt_ipsec.h"
73#include "opt_pcbgroup.h"
74#include "opt_rss.h"
75
76#include <sys/param.h>
77#include <sys/systm.h>
78#include <sys/malloc.h>
79#include <sys/mbuf.h>
80#include <sys/domain.h>
81#include <sys/protosw.h>
82#include <sys/socket.h>
83#include <sys/socketvar.h>
84#include <sys/sockio.h>
85#include <sys/errno.h>
86#include <sys/time.h>
87#include <sys/priv.h>
88#include <sys/proc.h>
89#include <sys/jail.h>
90
91#include <vm/uma.h>
92
93#include <net/if.h>
94#include <net/if_var.h>
95#include <net/if_types.h>
96#include <net/route.h>
97
98#include <netinet/in.h>
99#include <netinet/in_var.h>
100#include <netinet/in_systm.h>
101#include <netinet/tcp_var.h>
102#include <netinet/ip6.h>
103#include <netinet/ip_var.h>
104
105#include <netinet6/ip6_var.h>
106#include <netinet6/nd6.h>
107#include <netinet/in_pcb.h>
108#include <netinet6/in6_pcb.h>
109#include <netinet6/scope6_var.h>
110
111int
112in6_pcbbind(register struct inpcb *inp, struct sockaddr *nam,
113    struct ucred *cred)
114{
115	struct socket *so = inp->inp_socket;
116	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)NULL;
117	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
118	u_short	lport = 0;
119	int error, lookupflags = 0;
120	int reuseport = (so->so_options & SO_REUSEPORT);
121
122	INP_WLOCK_ASSERT(inp);
123	INP_HASH_WLOCK_ASSERT(pcbinfo);
124
125	if (TAILQ_EMPTY(&V_in6_ifaddrhead))	/* XXX broken! */
126		return (EADDRNOTAVAIL);
127	if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
128		return (EINVAL);
129	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
130		lookupflags = INPLOOKUP_WILDCARD;
131	if (nam == NULL) {
132		if ((error = prison_local_ip6(cred, &inp->in6p_laddr,
133		    ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
134			return (error);
135	} else {
136		sin6 = (struct sockaddr_in6 *)nam;
137		if (nam->sa_len != sizeof(*sin6))
138			return (EINVAL);
139		/*
140		 * family check.
141		 */
142		if (nam->sa_family != AF_INET6)
143			return (EAFNOSUPPORT);
144
145		if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
146			return(error);
147
148		if ((error = prison_local_ip6(cred, &sin6->sin6_addr,
149		    ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0))) != 0)
150			return (error);
151
152		lport = sin6->sin6_port;
153		if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
154			/*
155			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
156			 * allow compepte duplication of binding if
157			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
158			 * and a multicast address is bound on both
159			 * new and duplicated sockets.
160			 */
161			if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) != 0)
162				reuseport = SO_REUSEADDR|SO_REUSEPORT;
163		} else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
164			struct ifaddr *ifa;
165
166			sin6->sin6_port = 0;		/* yech... */
167			if ((ifa = ifa_ifwithaddr((struct sockaddr *)sin6)) ==
168			    NULL &&
169			    (inp->inp_flags & INP_BINDANY) == 0) {
170				return (EADDRNOTAVAIL);
171			}
172
173			/*
174			 * XXX: bind to an anycast address might accidentally
175			 * cause sending a packet with anycast source address.
176			 * We should allow to bind to a deprecated address, since
177			 * the application dares to use it.
178			 */
179			if (ifa != NULL &&
180			    ((struct in6_ifaddr *)ifa)->ia6_flags &
181			    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|IN6_IFF_DETACHED)) {
182				ifa_free(ifa);
183				return (EADDRNOTAVAIL);
184			}
185			if (ifa != NULL)
186				ifa_free(ifa);
187		}
188		if (lport) {
189			struct inpcb *t;
190			struct tcptw *tw;
191
192			/* GROSS */
193			if (ntohs(lport) <= V_ipport_reservedhigh &&
194			    ntohs(lport) >= V_ipport_reservedlow &&
195			    priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT,
196			    0))
197				return (EACCES);
198			if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr) &&
199			    priv_check_cred(inp->inp_cred,
200			    PRIV_NETINET_REUSEPORT, 0) != 0) {
201				t = in6_pcblookup_local(pcbinfo,
202				    &sin6->sin6_addr, lport,
203				    INPLOOKUP_WILDCARD, cred);
204				if (t &&
205				    ((t->inp_flags & INP_TIMEWAIT) == 0) &&
206				    (so->so_type != SOCK_STREAM ||
207				     IN6_IS_ADDR_UNSPECIFIED(&t->in6p_faddr)) &&
208				    (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr) ||
209				     !IN6_IS_ADDR_UNSPECIFIED(&t->in6p_laddr) ||
210				     (t->inp_flags2 & INP_REUSEPORT) == 0) &&
211				    (inp->inp_cred->cr_uid !=
212				     t->inp_cred->cr_uid))
213					return (EADDRINUSE);
214#ifdef INET
215				if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
216				    IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
217					struct sockaddr_in sin;
218
219					in6_sin6_2_sin(&sin, sin6);
220					t = in_pcblookup_local(pcbinfo,
221					    sin.sin_addr, lport,
222					    INPLOOKUP_WILDCARD, cred);
223					if (t &&
224					    ((t->inp_flags &
225					      INP_TIMEWAIT) == 0) &&
226					    (so->so_type != SOCK_STREAM ||
227					     ntohl(t->inp_faddr.s_addr) ==
228					      INADDR_ANY) &&
229					    (inp->inp_cred->cr_uid !=
230					     t->inp_cred->cr_uid))
231						return (EADDRINUSE);
232				}
233#endif
234			}
235			t = in6_pcblookup_local(pcbinfo, &sin6->sin6_addr,
236			    lport, lookupflags, cred);
237			if (t && (t->inp_flags & INP_TIMEWAIT)) {
238				/*
239				 * XXXRW: If an incpb has had its timewait
240				 * state recycled, we treat the address as
241				 * being in use (for now).  This is better
242				 * than a panic, but not desirable.
243				 */
244				tw = intotw(t);
245				if (tw == NULL ||
246				    (reuseport & tw->tw_so_options) == 0)
247					return (EADDRINUSE);
248			} else if (t && (reuseport & inp_so_options(t)) == 0) {
249				return (EADDRINUSE);
250			}
251#ifdef INET
252			if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0 &&
253			    IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
254				struct sockaddr_in sin;
255
256				in6_sin6_2_sin(&sin, sin6);
257				t = in_pcblookup_local(pcbinfo, sin.sin_addr,
258				    lport, lookupflags, cred);
259				if (t && t->inp_flags & INP_TIMEWAIT) {
260					tw = intotw(t);
261					if (tw == NULL)
262						return (EADDRINUSE);
263					if ((reuseport & tw->tw_so_options) == 0
264					    && (ntohl(t->inp_laddr.s_addr) !=
265					     INADDR_ANY || ((inp->inp_vflag &
266					     INP_IPV6PROTO) ==
267					     (t->inp_vflag & INP_IPV6PROTO))))
268						return (EADDRINUSE);
269				} else if (t &&
270				    (reuseport & inp_so_options(t)) == 0 &&
271				    (ntohl(t->inp_laddr.s_addr) != INADDR_ANY ||
272				    (t->inp_vflag & INP_IPV6PROTO) != 0))
273					return (EADDRINUSE);
274			}
275#endif
276		}
277		inp->in6p_laddr = sin6->sin6_addr;
278	}
279	if (lport == 0) {
280		if ((error = in6_pcbsetport(&inp->in6p_laddr, inp, cred)) != 0) {
281			/* Undo an address bind that may have occurred. */
282			inp->in6p_laddr = in6addr_any;
283			return (error);
284		}
285	} else {
286		inp->inp_lport = lport;
287		if (in_pcbinshash(inp) != 0) {
288			inp->in6p_laddr = in6addr_any;
289			inp->inp_lport = 0;
290			return (EAGAIN);
291		}
292	}
293	return (0);
294}
295
296/*
297 *   Transform old in6_pcbconnect() into an inner subroutine for new
298 *   in6_pcbconnect(): Do some validity-checking on the remote
299 *   address (in mbuf 'nam') and then determine local host address
300 *   (i.e., which interface) to use to access that remote host.
301 *
302 *   This preserves definition of in6_pcbconnect(), while supporting a
303 *   slightly different version for T/TCP.  (This is more than
304 *   a bit of a kludge, but cleaning up the internal interfaces would
305 *   have forced minor changes in every protocol).
306 */
307int
308in6_pcbladdr(register struct inpcb *inp, struct sockaddr *nam,
309    struct in6_addr *plocal_addr6)
310{
311	register struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
312	int error = 0;
313	struct ifnet *ifp = NULL;
314	int scope_ambiguous = 0;
315	struct in6_addr in6a;
316
317	INP_WLOCK_ASSERT(inp);
318	INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);	/* XXXRW: why? */
319
320	if (nam->sa_len != sizeof (*sin6))
321		return (EINVAL);
322	if (sin6->sin6_family != AF_INET6)
323		return (EAFNOSUPPORT);
324	if (sin6->sin6_port == 0)
325		return (EADDRNOTAVAIL);
326
327	if (sin6->sin6_scope_id == 0 && !V_ip6_use_defzone)
328		scope_ambiguous = 1;
329	if ((error = sa6_embedscope(sin6, V_ip6_use_defzone)) != 0)
330		return(error);
331
332	if (!TAILQ_EMPTY(&V_in6_ifaddrhead)) {
333		/*
334		 * If the destination address is UNSPECIFIED addr,
335		 * use the loopback addr, e.g ::1.
336		 */
337		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
338			sin6->sin6_addr = in6addr_loopback;
339	}
340	if ((error = prison_remote_ip6(inp->inp_cred, &sin6->sin6_addr)) != 0)
341		return (error);
342
343	error = in6_selectsrc(sin6, inp->in6p_outputopts,
344	    inp, NULL, inp->inp_cred, &ifp, &in6a);
345	if (error)
346		return (error);
347
348	if (ifp && scope_ambiguous &&
349	    (error = in6_setscope(&sin6->sin6_addr, ifp, NULL)) != 0) {
350		return(error);
351	}
352
353	/*
354	 * Do not update this earlier, in case we return with an error.
355	 *
356	 * XXX: this in6_selectsrc result might replace the bound local
357	 * address with the address specified by setsockopt(IPV6_PKTINFO).
358	 * Is it the intended behavior?
359	 */
360	*plocal_addr6 = in6a;
361
362	/*
363	 * Don't do pcblookup call here; return interface in
364	 * plocal_addr6
365	 * and exit to caller, that will do the lookup.
366	 */
367
368	return (0);
369}
370
371/*
372 * Outer subroutine:
373 * Connect from a socket to a specified address.
374 * Both address and port must be specified in argument sin.
375 * If don't have a local address for this socket yet,
376 * then pick one.
377 */
378int
379in6_pcbconnect_mbuf(register struct inpcb *inp, struct sockaddr *nam,
380    struct ucred *cred, struct mbuf *m)
381{
382	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
383	register struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
384	struct in6_addr addr6;
385	int error;
386
387	INP_WLOCK_ASSERT(inp);
388	INP_HASH_WLOCK_ASSERT(pcbinfo);
389
390	/*
391	 * Call inner routine, to assign local interface address.
392	 * in6_pcbladdr() may automatically fill in sin6_scope_id.
393	 */
394	if ((error = in6_pcbladdr(inp, nam, &addr6)) != 0)
395		return (error);
396
397	if (in6_pcblookup_hash_locked(pcbinfo, &sin6->sin6_addr,
398			       sin6->sin6_port,
399			      IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
400			      ? &addr6 : &inp->in6p_laddr,
401			      inp->inp_lport, 0, NULL) != NULL) {
402		return (EADDRINUSE);
403	}
404	if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
405		if (inp->inp_lport == 0) {
406			error = in6_pcbbind(inp, (struct sockaddr *)0, cred);
407			if (error)
408				return (error);
409		}
410		inp->in6p_laddr = addr6;
411	}
412	inp->in6p_faddr = sin6->sin6_addr;
413	inp->inp_fport = sin6->sin6_port;
414	/* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
415	inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
416	if (inp->inp_flags & IN6P_AUTOFLOWLABEL)
417		inp->inp_flow |=
418		    (htonl(ip6_randomflowlabel()) & IPV6_FLOWLABEL_MASK);
419
420	in_pcbrehash_mbuf(inp, m);
421
422	return (0);
423}
424
425int
426in6_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct ucred *cred)
427{
428
429	return (in6_pcbconnect_mbuf(inp, nam, cred, NULL));
430}
431
432void
433in6_pcbdisconnect(struct inpcb *inp)
434{
435
436	INP_WLOCK_ASSERT(inp);
437	INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
438
439	bzero((caddr_t)&inp->in6p_faddr, sizeof(inp->in6p_faddr));
440	inp->inp_fport = 0;
441	/* clear flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
442	inp->inp_flow &= ~IPV6_FLOWLABEL_MASK;
443	in_pcbrehash(inp);
444}
445
446struct sockaddr *
447in6_sockaddr(in_port_t port, struct in6_addr *addr_p)
448{
449	struct sockaddr_in6 *sin6;
450
451	sin6 = malloc(sizeof *sin6, M_SONAME, M_WAITOK);
452	bzero(sin6, sizeof *sin6);
453	sin6->sin6_family = AF_INET6;
454	sin6->sin6_len = sizeof(*sin6);
455	sin6->sin6_port = port;
456	sin6->sin6_addr = *addr_p;
457	(void)sa6_recoverscope(sin6); /* XXX: should catch errors */
458
459	return (struct sockaddr *)sin6;
460}
461
462struct sockaddr *
463in6_v4mapsin6_sockaddr(in_port_t port, struct in_addr *addr_p)
464{
465	struct sockaddr_in sin;
466	struct sockaddr_in6 *sin6_p;
467
468	bzero(&sin, sizeof sin);
469	sin.sin_family = AF_INET;
470	sin.sin_len = sizeof(sin);
471	sin.sin_port = port;
472	sin.sin_addr = *addr_p;
473
474	sin6_p = malloc(sizeof *sin6_p, M_SONAME,
475		M_WAITOK);
476	in6_sin_2_v4mapsin6(&sin, sin6_p);
477
478	return (struct sockaddr *)sin6_p;
479}
480
481int
482in6_getsockaddr(struct socket *so, struct sockaddr **nam)
483{
484	register struct inpcb *inp;
485	struct in6_addr addr;
486	in_port_t port;
487
488	inp = sotoinpcb(so);
489	KASSERT(inp != NULL, ("in6_getsockaddr: inp == NULL"));
490
491	INP_RLOCK(inp);
492	port = inp->inp_lport;
493	addr = inp->in6p_laddr;
494	INP_RUNLOCK(inp);
495
496	*nam = in6_sockaddr(port, &addr);
497	return 0;
498}
499
500int
501in6_getpeeraddr(struct socket *so, struct sockaddr **nam)
502{
503	struct inpcb *inp;
504	struct in6_addr addr;
505	in_port_t port;
506
507	inp = sotoinpcb(so);
508	KASSERT(inp != NULL, ("in6_getpeeraddr: inp == NULL"));
509
510	INP_RLOCK(inp);
511	port = inp->inp_fport;
512	addr = inp->in6p_faddr;
513	INP_RUNLOCK(inp);
514
515	*nam = in6_sockaddr(port, &addr);
516	return 0;
517}
518
519int
520in6_mapped_sockaddr(struct socket *so, struct sockaddr **nam)
521{
522	struct	inpcb *inp;
523	int	error;
524
525	inp = sotoinpcb(so);
526	KASSERT(inp != NULL, ("in6_mapped_sockaddr: inp == NULL"));
527
528#ifdef INET
529	if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) {
530		error = in_getsockaddr(so, nam);
531		if (error == 0)
532			in6_sin_2_v4mapsin6_in_sock(nam);
533	} else
534#endif
535	{
536		/* scope issues will be handled in in6_getsockaddr(). */
537		error = in6_getsockaddr(so, nam);
538	}
539
540	return error;
541}
542
543int
544in6_mapped_peeraddr(struct socket *so, struct sockaddr **nam)
545{
546	struct	inpcb *inp;
547	int	error;
548
549	inp = sotoinpcb(so);
550	KASSERT(inp != NULL, ("in6_mapped_peeraddr: inp == NULL"));
551
552#ifdef INET
553	if ((inp->inp_vflag & (INP_IPV4 | INP_IPV6)) == INP_IPV4) {
554		error = in_getpeeraddr(so, nam);
555		if (error == 0)
556			in6_sin_2_v4mapsin6_in_sock(nam);
557	} else
558#endif
559	/* scope issues will be handled in in6_getpeeraddr(). */
560	error = in6_getpeeraddr(so, nam);
561
562	return error;
563}
564
565/*
566 * Pass some notification to all connections of a protocol
567 * associated with address dst.  The local address and/or port numbers
568 * may be specified to limit the search.  The "usual action" will be
569 * taken, depending on the ctlinput cmd.  The caller must filter any
570 * cmds that are uninteresting (e.g., no error in the map).
571 * Call the protocol specific routine (if any) to report
572 * any errors for each matching socket.
573 */
574void
575in6_pcbnotify(struct inpcbinfo *pcbinfo, struct sockaddr *dst,
576    u_int fport_arg, const struct sockaddr *src, u_int lport_arg,
577    int cmd, void *cmdarg,
578    struct inpcb *(*notify)(struct inpcb *, int))
579{
580	struct inpcb *inp, *inp_temp;
581	struct sockaddr_in6 sa6_src, *sa6_dst;
582	u_short	fport = fport_arg, lport = lport_arg;
583	u_int32_t flowinfo;
584	int errno;
585
586	if ((unsigned)cmd >= PRC_NCMDS || dst->sa_family != AF_INET6)
587		return;
588
589	sa6_dst = (struct sockaddr_in6 *)dst;
590	if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr))
591		return;
592
593	/*
594	 * note that src can be NULL when we get notify by local fragmentation.
595	 */
596	sa6_src = (src == NULL) ? sa6_any : *(const struct sockaddr_in6 *)src;
597	flowinfo = sa6_src.sin6_flowinfo;
598
599	/*
600	 * Redirects go to all references to the destination,
601	 * and use in6_rtchange to invalidate the route cache.
602	 * Dead host indications: also use in6_rtchange to invalidate
603	 * the cache, and deliver the error to all the sockets.
604	 * Otherwise, if we have knowledge of the local port and address,
605	 * deliver only to that socket.
606	 */
607	if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
608		fport = 0;
609		lport = 0;
610		bzero((caddr_t)&sa6_src.sin6_addr, sizeof(sa6_src.sin6_addr));
611
612		if (cmd != PRC_HOSTDEAD)
613			notify = in6_rtchange;
614	}
615	errno = inet6ctlerrmap[cmd];
616	INP_INFO_WLOCK(pcbinfo);
617	LIST_FOREACH_SAFE(inp, pcbinfo->ipi_listhead, inp_list, inp_temp) {
618		INP_WLOCK(inp);
619		if ((inp->inp_vflag & INP_IPV6) == 0) {
620			INP_WUNLOCK(inp);
621			continue;
622		}
623
624		/*
625		 * If the error designates a new path MTU for a destination
626		 * and the application (associated with this socket) wanted to
627		 * know the value, notify. Note that we notify for all
628		 * disconnected sockets if the corresponding application
629		 * wanted. This is because some UDP applications keep sending
630		 * sockets disconnected.
631		 * XXX: should we avoid to notify the value to TCP sockets?
632		 */
633		if (cmd == PRC_MSGSIZE && (inp->inp_flags & IN6P_MTU) != 0 &&
634		    (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
635		     IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &sa6_dst->sin6_addr))) {
636			ip6_notify_pmtu(inp, (struct sockaddr_in6 *)dst,
637					(u_int32_t *)cmdarg);
638		}
639
640		/*
641		 * Detect if we should notify the error. If no source and
642		 * destination ports are specifed, but non-zero flowinfo and
643		 * local address match, notify the error. This is the case
644		 * when the error is delivered with an encrypted buffer
645		 * by ESP. Otherwise, just compare addresses and ports
646		 * as usual.
647		 */
648		if (lport == 0 && fport == 0 && flowinfo &&
649		    inp->inp_socket != NULL &&
650		    flowinfo == (inp->inp_flow & IPV6_FLOWLABEL_MASK) &&
651		    IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &sa6_src.sin6_addr))
652			goto do_notify;
653		else if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
654					     &sa6_dst->sin6_addr) ||
655			 inp->inp_socket == 0 ||
656			 (lport && inp->inp_lport != lport) ||
657			 (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) &&
658			  !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
659					      &sa6_src.sin6_addr)) ||
660			 (fport && inp->inp_fport != fport)) {
661			INP_WUNLOCK(inp);
662			continue;
663		}
664
665	  do_notify:
666		if (notify) {
667			if ((*notify)(inp, errno))
668				INP_WUNLOCK(inp);
669		} else
670			INP_WUNLOCK(inp);
671	}
672	INP_INFO_WUNLOCK(pcbinfo);
673}
674
675/*
676 * Lookup a PCB based on the local address and port.  Caller must hold the
677 * hash lock.  No inpcb locks or references are acquired.
678 */
679struct inpcb *
680in6_pcblookup_local(struct inpcbinfo *pcbinfo, struct in6_addr *laddr,
681    u_short lport, int lookupflags, struct ucred *cred)
682{
683	register struct inpcb *inp;
684	int matchwild = 3, wildcard;
685
686	KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
687	    ("%s: invalid lookup flags %d", __func__, lookupflags));
688
689	INP_HASH_WLOCK_ASSERT(pcbinfo);
690
691	if ((lookupflags & INPLOOKUP_WILDCARD) == 0) {
692		struct inpcbhead *head;
693		/*
694		 * Look for an unconnected (wildcard foreign addr) PCB that
695		 * matches the local address and port we're looking for.
696		 */
697		head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
698		    0, pcbinfo->ipi_hashmask)];
699		LIST_FOREACH(inp, head, inp_hash) {
700			/* XXX inp locking */
701			if ((inp->inp_vflag & INP_IPV6) == 0)
702				continue;
703			if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
704			    IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
705			    inp->inp_lport == lport) {
706				/* Found. */
707				if (cred == NULL ||
708				    prison_equal_ip6(cred->cr_prison,
709					inp->inp_cred->cr_prison))
710					return (inp);
711			}
712		}
713		/*
714		 * Not found.
715		 */
716		return (NULL);
717	} else {
718		struct inpcbporthead *porthash;
719		struct inpcbport *phd;
720		struct inpcb *match = NULL;
721		/*
722		 * Best fit PCB lookup.
723		 *
724		 * First see if this local port is in use by looking on the
725		 * port hash list.
726		 */
727		porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport,
728		    pcbinfo->ipi_porthashmask)];
729		LIST_FOREACH(phd, porthash, phd_hash) {
730			if (phd->phd_port == lport)
731				break;
732		}
733		if (phd != NULL) {
734			/*
735			 * Port is in use by one or more PCBs. Look for best
736			 * fit.
737			 */
738			LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
739				wildcard = 0;
740				if (cred != NULL &&
741				    !prison_equal_ip6(cred->cr_prison,
742					inp->inp_cred->cr_prison))
743					continue;
744				/* XXX inp locking */
745				if ((inp->inp_vflag & INP_IPV6) == 0)
746					continue;
747				if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr))
748					wildcard++;
749				if (!IN6_IS_ADDR_UNSPECIFIED(
750					&inp->in6p_laddr)) {
751					if (IN6_IS_ADDR_UNSPECIFIED(laddr))
752						wildcard++;
753					else if (!IN6_ARE_ADDR_EQUAL(
754					    &inp->in6p_laddr, laddr))
755						continue;
756				} else {
757					if (!IN6_IS_ADDR_UNSPECIFIED(laddr))
758						wildcard++;
759				}
760				if (wildcard < matchwild) {
761					match = inp;
762					matchwild = wildcard;
763					if (matchwild == 0)
764						break;
765				}
766			}
767		}
768		return (match);
769	}
770}
771
772void
773in6_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
774{
775	struct inpcb *in6p;
776	struct ip6_moptions *im6o;
777	int i, gap;
778
779	INP_INFO_RLOCK(pcbinfo);
780	LIST_FOREACH(in6p, pcbinfo->ipi_listhead, inp_list) {
781		INP_WLOCK(in6p);
782		im6o = in6p->in6p_moptions;
783		if ((in6p->inp_vflag & INP_IPV6) && im6o != NULL) {
784			/*
785			 * Unselect the outgoing ifp for multicast if it
786			 * is being detached.
787			 */
788			if (im6o->im6o_multicast_ifp == ifp)
789				im6o->im6o_multicast_ifp = NULL;
790			/*
791			 * Drop multicast group membership if we joined
792			 * through the interface being detached.
793			 */
794			gap = 0;
795			for (i = 0; i < im6o->im6o_num_memberships; i++) {
796				if (im6o->im6o_membership[i]->in6m_ifp ==
797				    ifp) {
798					in6_mc_leave(im6o->im6o_membership[i],
799					    NULL);
800					gap++;
801				} else if (gap != 0) {
802					im6o->im6o_membership[i - gap] =
803					    im6o->im6o_membership[i];
804				}
805			}
806			im6o->im6o_num_memberships -= gap;
807		}
808		INP_WUNLOCK(in6p);
809	}
810	INP_INFO_RUNLOCK(pcbinfo);
811}
812
813/*
814 * Check for alternatives when higher level complains
815 * about service problems.  For now, invalidate cached
816 * routing information.  If the route was created dynamically
817 * (by a redirect), time to try a default gateway again.
818 */
819void
820in6_losing(struct inpcb *in6p)
821{
822
823	/*
824	 * We don't store route pointers in the routing table anymore
825	 */
826	return;
827}
828
829/*
830 * After a routing change, flush old routing
831 * and allocate a (hopefully) better one.
832 */
833struct inpcb *
834in6_rtchange(struct inpcb *inp, int errno)
835{
836	/*
837	 * We don't store route pointers in the routing table anymore
838	 */
839	return inp;
840}
841
842#ifdef PCBGROUP
843/*
844 * Lookup PCB in hash list, using pcbgroup tables.
845 */
846static struct inpcb *
847in6_pcblookup_group(struct inpcbinfo *pcbinfo, struct inpcbgroup *pcbgroup,
848    struct in6_addr *faddr, u_int fport_arg, struct in6_addr *laddr,
849    u_int lport_arg, int lookupflags, struct ifnet *ifp)
850{
851	struct inpcbhead *head;
852	struct inpcb *inp, *tmpinp;
853	u_short fport = fport_arg, lport = lport_arg;
854	int faith;
855
856	if (faithprefix_p != NULL)
857		faith = (*faithprefix_p)(laddr);
858	else
859		faith = 0;
860
861	/*
862	 * First look for an exact match.
863	 */
864	tmpinp = NULL;
865	INP_GROUP_LOCK(pcbgroup);
866	head = &pcbgroup->ipg_hashbase[
867	    INP_PCBHASH(faddr->s6_addr32[3] /* XXX */, lport, fport,
868	    pcbgroup->ipg_hashmask)];
869	LIST_FOREACH(inp, head, inp_pcbgrouphash) {
870		/* XXX inp locking */
871		if ((inp->inp_vflag & INP_IPV6) == 0)
872			continue;
873		if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) &&
874		    IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
875		    inp->inp_fport == fport &&
876		    inp->inp_lport == lport) {
877			/*
878			 * XXX We should be able to directly return
879			 * the inp here, without any checks.
880			 * Well unless both bound with SO_REUSEPORT?
881			 */
882			if (prison_flag(inp->inp_cred, PR_IP6))
883				goto found;
884			if (tmpinp == NULL)
885				tmpinp = inp;
886		}
887	}
888	if (tmpinp != NULL) {
889		inp = tmpinp;
890		goto found;
891	}
892
893	/*
894	 * Then look for a wildcard match, if requested.
895	 */
896	if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
897		struct inpcb *local_wild = NULL, *local_exact = NULL;
898		struct inpcb *jail_wild = NULL;
899		int injail;
900
901		/*
902		 * Order of socket selection - we always prefer jails.
903		 *      1. jailed, non-wild.
904		 *      2. jailed, wild.
905		 *      3. non-jailed, non-wild.
906		 *      4. non-jailed, wild.
907		 */
908		head = &pcbinfo->ipi_wildbase[INP_PCBHASH(INADDR_ANY, lport,
909		    0, pcbinfo->ipi_wildmask)];
910		LIST_FOREACH(inp, head, inp_pcbgroup_wild) {
911			/* XXX inp locking */
912			if ((inp->inp_vflag & INP_IPV6) == 0)
913				continue;
914
915			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
916			    inp->inp_lport != lport) {
917				continue;
918			}
919
920			/* XXX inp locking */
921			if (faith && (inp->inp_flags & INP_FAITH) == 0)
922				continue;
923
924			injail = prison_flag(inp->inp_cred, PR_IP6);
925			if (injail) {
926				if (prison_check_ip6(inp->inp_cred,
927				    laddr) != 0)
928					continue;
929			} else {
930				if (local_exact != NULL)
931					continue;
932			}
933
934			if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) {
935				if (injail)
936					goto found;
937				else
938					local_exact = inp;
939			} else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
940				if (injail)
941					jail_wild = inp;
942				else
943					local_wild = inp;
944			}
945		} /* LIST_FOREACH */
946
947		inp = jail_wild;
948		if (inp == NULL)
949			inp = jail_wild;
950		if (inp == NULL)
951			inp = local_exact;
952		if (inp == NULL)
953			inp = local_wild;
954		if (inp != NULL)
955			goto found;
956	} /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */
957	INP_GROUP_UNLOCK(pcbgroup);
958	return (NULL);
959
960found:
961	in_pcbref(inp);
962	INP_GROUP_UNLOCK(pcbgroup);
963	if (lookupflags & INPLOOKUP_WLOCKPCB) {
964		INP_WLOCK(inp);
965		if (in_pcbrele_wlocked(inp))
966			return (NULL);
967	} else if (lookupflags & INPLOOKUP_RLOCKPCB) {
968		INP_RLOCK(inp);
969		if (in_pcbrele_rlocked(inp))
970			return (NULL);
971	} else
972		panic("%s: locking buf", __func__);
973	return (inp);
974}
975#endif /* PCBGROUP */
976
977/*
978 * Lookup PCB in hash list.
979 */
980struct inpcb *
981in6_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
982    u_int fport_arg, struct in6_addr *laddr, u_int lport_arg,
983    int lookupflags, struct ifnet *ifp)
984{
985	struct inpcbhead *head;
986	struct inpcb *inp, *tmpinp;
987	u_short fport = fport_arg, lport = lport_arg;
988	int faith;
989
990	KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0,
991	    ("%s: invalid lookup flags %d", __func__, lookupflags));
992
993	INP_HASH_LOCK_ASSERT(pcbinfo);
994
995	if (faithprefix_p != NULL)
996		faith = (*faithprefix_p)(laddr);
997	else
998		faith = 0;
999
1000	/*
1001	 * First look for an exact match.
1002	 */
1003	tmpinp = NULL;
1004	head = &pcbinfo->ipi_hashbase[
1005	    INP_PCBHASH(faddr->s6_addr32[3] /* XXX */, lport, fport,
1006	    pcbinfo->ipi_hashmask)];
1007	LIST_FOREACH(inp, head, inp_hash) {
1008		/* XXX inp locking */
1009		if ((inp->inp_vflag & INP_IPV6) == 0)
1010			continue;
1011		if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) &&
1012		    IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
1013		    inp->inp_fport == fport &&
1014		    inp->inp_lport == lport) {
1015			/*
1016			 * XXX We should be able to directly return
1017			 * the inp here, without any checks.
1018			 * Well unless both bound with SO_REUSEPORT?
1019			 */
1020			if (prison_flag(inp->inp_cred, PR_IP6))
1021				return (inp);
1022			if (tmpinp == NULL)
1023				tmpinp = inp;
1024		}
1025	}
1026	if (tmpinp != NULL)
1027		return (tmpinp);
1028
1029	/*
1030	 * Then look for a wildcard match, if requested.
1031	 */
1032	if ((lookupflags & INPLOOKUP_WILDCARD) != 0) {
1033		struct inpcb *local_wild = NULL, *local_exact = NULL;
1034		struct inpcb *jail_wild = NULL;
1035		int injail;
1036
1037		/*
1038		 * Order of socket selection - we always prefer jails.
1039		 *      1. jailed, non-wild.
1040		 *      2. jailed, wild.
1041		 *      3. non-jailed, non-wild.
1042		 *      4. non-jailed, wild.
1043		 */
1044		head = &pcbinfo->ipi_hashbase[INP_PCBHASH(INADDR_ANY, lport,
1045		    0, pcbinfo->ipi_hashmask)];
1046		LIST_FOREACH(inp, head, inp_hash) {
1047			/* XXX inp locking */
1048			if ((inp->inp_vflag & INP_IPV6) == 0)
1049				continue;
1050
1051			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
1052			    inp->inp_lport != lport) {
1053				continue;
1054			}
1055
1056			/* XXX inp locking */
1057			if (faith && (inp->inp_flags & INP_FAITH) == 0)
1058				continue;
1059
1060			injail = prison_flag(inp->inp_cred, PR_IP6);
1061			if (injail) {
1062				if (prison_check_ip6(inp->inp_cred,
1063				    laddr) != 0)
1064					continue;
1065			} else {
1066				if (local_exact != NULL)
1067					continue;
1068			}
1069
1070			if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) {
1071				if (injail)
1072					return (inp);
1073				else
1074					local_exact = inp;
1075			} else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
1076				if (injail)
1077					jail_wild = inp;
1078				else
1079					local_wild = inp;
1080			}
1081		} /* LIST_FOREACH */
1082
1083		if (jail_wild != NULL)
1084			return (jail_wild);
1085		if (local_exact != NULL)
1086			return (local_exact);
1087		if (local_wild != NULL)
1088			return (local_wild);
1089	} /* if ((lookupflags & INPLOOKUP_WILDCARD) != 0) */
1090
1091	/*
1092	 * Not found.
1093	 */
1094	return (NULL);
1095}
1096
1097/*
1098 * Lookup PCB in hash list, using pcbinfo tables.  This variation locks the
1099 * hash list lock, and will return the inpcb locked (i.e., requires
1100 * INPLOOKUP_LOCKPCB).
1101 */
1102static struct inpcb *
1103in6_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
1104    u_int fport, struct in6_addr *laddr, u_int lport, int lookupflags,
1105    struct ifnet *ifp)
1106{
1107	struct inpcb *inp;
1108
1109	INP_HASH_RLOCK(pcbinfo);
1110	inp = in6_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport,
1111	    (lookupflags & ~(INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)), ifp);
1112	if (inp != NULL) {
1113		in_pcbref(inp);
1114		INP_HASH_RUNLOCK(pcbinfo);
1115		if (lookupflags & INPLOOKUP_WLOCKPCB) {
1116			INP_WLOCK(inp);
1117			if (in_pcbrele_wlocked(inp))
1118				return (NULL);
1119		} else if (lookupflags & INPLOOKUP_RLOCKPCB) {
1120			INP_RLOCK(inp);
1121			if (in_pcbrele_rlocked(inp))
1122				return (NULL);
1123		} else
1124			panic("%s: locking bug", __func__);
1125	} else
1126		INP_HASH_RUNLOCK(pcbinfo);
1127	return (inp);
1128}
1129
1130/*
1131 * Public inpcb lookup routines, accepting a 4-tuple, and optionally, an mbuf
1132 * from which a pre-calculated hash value may be extracted.
1133 *
1134 * Possibly more of this logic should be in in6_pcbgroup.c.
1135 */
1136struct inpcb *
1137in6_pcblookup(struct inpcbinfo *pcbinfo, struct in6_addr *faddr, u_int fport,
1138    struct in6_addr *laddr, u_int lport, int lookupflags, struct ifnet *ifp)
1139{
1140#if defined(PCBGROUP) && !defined(RSS)
1141	struct inpcbgroup *pcbgroup;
1142#endif
1143
1144	KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1145	    ("%s: invalid lookup flags %d", __func__, lookupflags));
1146	KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1147	    ("%s: LOCKPCB not set", __func__));
1148
1149	/*
1150	 * When not using RSS, use connection groups in preference to the
1151	 * reservation table when looking up 4-tuples.  When using RSS, just
1152	 * use the reservation table, due to the cost of the Toeplitz hash
1153	 * in software.
1154	 *
1155	 * XXXRW: This policy belongs in the pcbgroup code, as in principle
1156	 * we could be doing RSS with a non-Toeplitz hash that is affordable
1157	 * in software.
1158	 */
1159#if defined(PCBGROUP) && !defined(RSS)
1160	if (in_pcbgroup_enabled(pcbinfo)) {
1161		pcbgroup = in6_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr,
1162		    fport);
1163		return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr, fport,
1164		    laddr, lport, lookupflags, ifp));
1165	}
1166#endif
1167	return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
1168	    lookupflags, ifp));
1169}
1170
1171struct inpcb *
1172in6_pcblookup_mbuf(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
1173    u_int fport, struct in6_addr *laddr, u_int lport, int lookupflags,
1174    struct ifnet *ifp, struct mbuf *m)
1175{
1176#ifdef PCBGROUP
1177	struct inpcbgroup *pcbgroup;
1178#endif
1179
1180	KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0,
1181	    ("%s: invalid lookup flags %d", __func__, lookupflags));
1182	KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0,
1183	    ("%s: LOCKPCB not set", __func__));
1184
1185#ifdef PCBGROUP
1186	/*
1187	 * If we can use a hardware-generated hash to look up the connection
1188	 * group, use that connection group to find the inpcb.  Otherwise
1189	 * fall back on a software hash -- or the reservation table if we're
1190	 * using RSS.
1191	 *
1192	 * XXXRW: As above, that policy belongs in the pcbgroup code.
1193	 */
1194	if (in_pcbgroup_enabled(pcbinfo) &&
1195	    !(M_HASHTYPE_TEST(m, M_HASHTYPE_NONE))) {
1196		pcbgroup = in6_pcbgroup_byhash(pcbinfo, M_HASHTYPE_GET(m),
1197		    m->m_pkthdr.flowid);
1198		if (pcbgroup != NULL)
1199			return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr,
1200			    fport, laddr, lport, lookupflags, ifp));
1201#ifndef RSS
1202		pcbgroup = in6_pcbgroup_bytuple(pcbinfo, laddr, lport, faddr,
1203		    fport);
1204		return (in6_pcblookup_group(pcbinfo, pcbgroup, faddr, fport,
1205		    laddr, lport, lookupflags, ifp));
1206#endif
1207	}
1208#endif
1209	return (in6_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport,
1210	    lookupflags, ifp));
1211}
1212
1213void
1214init_sin6(struct sockaddr_in6 *sin6, struct mbuf *m)
1215{
1216	struct ip6_hdr *ip;
1217
1218	ip = mtod(m, struct ip6_hdr *);
1219	bzero(sin6, sizeof(*sin6));
1220	sin6->sin6_len = sizeof(*sin6);
1221	sin6->sin6_family = AF_INET6;
1222	sin6->sin6_addr = ip->ip6_src;
1223
1224	(void)sa6_recoverscope(sin6); /* XXX: should catch errors... */
1225
1226	return;
1227}
1228