xref: /dragonfly/sys/netinet6/in6_pcb.c (revision cad2e385)
1 /*	$FreeBSD: src/sys/netinet6/in6_pcb.c,v 1.10.2.9 2003/01/24 05:11:35 sam Exp $	*/
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. Neither the name of the University nor the names of its contributors
47  *    may be used to endorse or promote products derived from this software
48  *    without specific prior written permission.
49  *
50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60  * SUCH DAMAGE.
61  *
62  *	@(#)in_pcb.c	8.2 (Berkeley) 1/4/94
63  */
64 
65 #include "opt_inet.h"
66 #include "opt_inet6.h"
67 #include "opt_ipsec.h"
68 
69 #include <sys/param.h>
70 #include <sys/systm.h>
71 #include <sys/malloc.h>
72 #include <sys/mbuf.h>
73 #include <sys/domain.h>
74 #include <sys/protosw.h>
75 #include <sys/socket.h>
76 #include <sys/socketvar.h>
77 #include <sys/sockio.h>
78 #include <sys/errno.h>
79 #include <sys/time.h>
80 #include <sys/proc.h>
81 #include <sys/priv.h>
82 #include <sys/jail.h>
83 
84 #include <sys/thread2.h>
85 #include <sys/msgport2.h>
86 
87 #include <vm/vm_zone.h>
88 
89 #include <net/if.h>
90 #include <net/if_types.h>
91 #include <net/route.h>
92 #include <net/netisr2.h>
93 
94 #include <netinet/in.h>
95 #include <netinet/in_var.h>
96 #include <netinet/in_systm.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 <netproto/key/key.h>
114 #endif /* IPSEC */
115 
116 #ifdef FAST_IPSEC
117 #include <netproto/ipsec/ipsec.h>
118 #include <netproto/ipsec/ipsec6.h>
119 #include <netproto/ipsec/key.h>
120 #define	IPSEC
121 #endif /* FAST_IPSEC */
122 
123 struct	in6_addr zeroin6_addr;
124 
125 int
126 in6_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct thread *td)
127 {
128 	struct socket *so = inp->inp_socket;
129 	struct sockaddr_in6 jsin6;
130 	int error;
131 
132 	if (!in6_ifaddr) /* XXX broken! */
133 		return (EADDRNOTAVAIL);
134 	if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
135 		return (EINVAL);
136 
137 	if (nam) {
138 		struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
139 		struct inpcbinfo *pcbinfo;
140 		struct inpcbportinfo *portinfo;
141 		int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
142 		struct ucred *cred = NULL;
143 		struct inpcb *t;
144 		u_short	lport, lport_ho;
145 
146 		if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
147 			wild = 1;
148 		if (td->td_proc != NULL)
149 			cred = td->td_proc->p_ucred;
150 
151 		if (nam->sa_len != sizeof(*sin6))
152 			return (EINVAL);
153 		/*
154 		 * family check.
155 		 */
156 		if (nam->sa_family != AF_INET6)
157 			return (EAFNOSUPPORT);
158 
159 		/* Reject v4-mapped address */
160 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr))
161 			return (EADDRNOTAVAIL);
162 
163 		if (!prison_replace_wildcards(td, nam))
164 			return (EINVAL);
165 
166 		/* KAME hack: embed scopeid */
167 		if (in6_embedscope(&sin6->sin6_addr, sin6, inp, NULL) != 0)
168 			return (EINVAL);
169 		/* this must be cleared for ifa_ifwithaddr() */
170 		sin6->sin6_scope_id = 0;
171 
172 		lport = sin6->sin6_port;
173 		if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
174 			/*
175 			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
176 			 * allow compepte duplication of binding if
177 			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
178 			 * and a multicast address is bound on both
179 			 * new and duplicated sockets.
180 			 */
181 			if (so->so_options & SO_REUSEADDR)
182 				reuseport = SO_REUSEADDR|SO_REUSEPORT;
183 		} else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
184 			struct ifaddr *ia = NULL;
185 
186 			sin6->sin6_port = 0;		/* yech... */
187 			if (!prison_replace_wildcards(td, (struct sockaddr *)sin6)) {
188 				sin6->sin6_addr = kin6addr_any;
189 				return (EINVAL);
190 			}
191 			if ((ia = ifa_ifwithaddr((struct sockaddr *)sin6)) == NULL)
192 				return (EADDRNOTAVAIL);
193 
194 			/*
195 			 * XXX: bind to an anycast address might accidentally
196 			 * cause sending a packet with anycast source address.
197 			 * We should allow to bind to a deprecated address, since
198 			 * the application dares to use it.
199 			 */
200 			if (ia &&
201 			    ((struct in6_ifaddr *)ia)->ia6_flags &
202 			    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|IN6_IFF_DETACHED))
203 				return (EADDRNOTAVAIL);
204 		}
205 
206 		inp->in6p_laddr = sin6->sin6_addr;
207 
208 		if (lport == 0)
209 			goto auto_select;
210 		lport_ho = ntohs(lport);
211 
212 		/* GROSS */
213 		if (lport_ho < IPV6PORT_RESERVED && cred &&
214 		    priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0)) {
215 			inp->in6p_laddr = kin6addr_any;
216 			return (EACCES);
217 		}
218 
219 		/*
220 		 * Locate the proper portinfo based on lport
221 		 */
222 		pcbinfo = inp->inp_pcbinfo;
223 		portinfo =
224 		    &pcbinfo->portinfo[lport_ho & pcbinfo->portinfo_mask];
225 		KKASSERT((lport_ho & pcbinfo->portinfo_mask) ==
226 		    portinfo->offset);
227 
228 		/*
229 		 * This has to be atomic.  If the porthash is shared across
230 		 * multiple protocol threads (aka tcp) then the token must
231 		 * be held.
232 		 */
233 		GET_PORT_TOKEN(portinfo);
234 
235 		if (so->so_cred->cr_uid != 0 &&
236 		    !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
237 			t = in6_pcblookup_local(portinfo,
238 			    &sin6->sin6_addr, lport, INPLOOKUP_WILDCARD, cred);
239 			if (t &&
240 			    (so->so_cred->cr_uid !=
241 			     t->inp_socket->so_cred->cr_uid)) {
242 				inp->in6p_laddr = kin6addr_any;
243 				error = EADDRINUSE;
244 				goto done;
245 			}
246 		}
247 		if (cred && cred->cr_prison &&
248 		    !prison_replace_wildcards(td, nam)) {
249 			inp->in6p_laddr = kin6addr_any;
250 			error = EADDRNOTAVAIL;
251 			goto done;
252 		}
253 		t = in6_pcblookup_local(portinfo, &sin6->sin6_addr, lport,
254 		    wild, cred);
255 		if (t && (reuseport & t->inp_socket->so_options) == 0) {
256 			inp->in6p_laddr = kin6addr_any;
257 			error = EADDRINUSE;
258 			goto done;
259 		}
260 
261 		inp->inp_lport = lport;
262 		in_pcbinsporthash(portinfo, inp);
263 		error = 0;
264 done:
265 		REL_PORT_TOKEN(portinfo);
266 		return (error);
267 	} else {
268 auto_select:
269 		jsin6.sin6_addr = inp->in6p_laddr;
270 		jsin6.sin6_family = AF_INET6;
271 		if (!prison_replace_wildcards(td, (struct sockaddr*)&jsin6)) {
272 			inp->in6p_laddr = kin6addr_any;
273 			inp->inp_lport = 0;
274 			return (EINVAL);
275 		}
276 
277 		return in6_pcbsetlport(&inp->in6p_laddr, inp, td);
278 	}
279 }
280 
281 /*
282  *   Transform old in6_pcbconnect() into an inner subroutine for new
283  *   in6_pcbconnect(): Do some validity-checking on the remote
284  *   address (in mbuf 'nam') and then determine local host address
285  *   (i.e., which interface) to use to access that remote host.
286  *
287  *   This preserves definition of in6_pcbconnect(), while supporting a
288  *   slightly different version for T/TCP.  (This is more than
289  *   a bit of a kludge, but cleaning up the internal interfaces would
290  *   have forced minor changes in every protocol).
291  */
292 
293 int
294 in6_pcbladdr(struct inpcb *inp, struct sockaddr *nam,
295 	     struct in6_addr **plocal_addr6, struct thread *td)
296 {
297 	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
298 	struct ifnet *ifp = NULL;
299 	int error = 0;
300 
301 	if (nam->sa_len != sizeof (*sin6))
302 		return (EINVAL);
303 	if (sin6->sin6_family != AF_INET6)
304 		return (EAFNOSUPPORT);
305 	if (sin6->sin6_port == 0)
306 		return (EADDRNOTAVAIL);
307 
308 	/* KAME hack: embed scopeid */
309 	if (in6_embedscope(&sin6->sin6_addr, sin6, inp, &ifp) != 0)
310 		return EINVAL;
311 
312 	if (in6_ifaddr) {
313 		/*
314 		 * If the destination address is UNSPECIFIED addr,
315 		 * use the loopback addr, e.g ::1.
316 		 */
317 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
318 			sin6->sin6_addr = kin6addr_loopback;
319 	}
320 	{
321 		/*
322 		 * XXX: in6_selectsrc might replace the bound local address
323 		 * with the address specified by setsockopt(IPV6_PKTINFO).
324 		 * Is it the intended behavior?
325 		 */
326 		*plocal_addr6 = in6_selectsrc(sin6, inp->in6p_outputopts,
327 					      inp->in6p_moptions,
328 					      &inp->in6p_route,
329 					      &inp->in6p_laddr, &error, td);
330 		if (*plocal_addr6 == NULL) {
331 			if (error == 0)
332 				error = EADDRNOTAVAIL;
333 			return (error);
334 		}
335 		/*
336 		 * Don't do pcblookup call here; return interface in
337 		 * plocal_addr6
338 		 * and exit to caller, that will do the lookup.
339 		 */
340 	}
341 
342 	if (inp->in6p_route.ro_rt)
343 		ifp = inp->in6p_route.ro_rt->rt_ifp;
344 
345 	return (0);
346 }
347 
348 /*
349  * Outer subroutine:
350  * Connect from a socket to a specified address.
351  * Both address and port must be specified in argument sin.
352  * If don't have a local address for this socket yet,
353  * then pick one.
354  */
355 int
356 in6_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct thread *td)
357 {
358 	struct in6_addr *addr6;
359 	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
360 	int error;
361 
362 	/* Reject v4-mapped address */
363 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr))
364 		return EADDRNOTAVAIL;
365 
366 	/*
367 	 * Call inner routine, to assign local interface address.
368 	 * in6_pcbladdr() may automatically fill in sin6_scope_id.
369 	 */
370 	if ((error = in6_pcbladdr(inp, nam, &addr6, td)) != 0)
371 		return (error);
372 
373 	if (in6_pcblookup_hash(inp->inp_pcbinfo, &sin6->sin6_addr,
374 			       sin6->sin6_port,
375 			      IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
376 			      ? addr6 : &inp->in6p_laddr,
377 			      inp->inp_lport, 0, NULL) != NULL) {
378 		return (EADDRINUSE);
379 	}
380 	if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
381 		if (inp->inp_lport == 0) {
382 			error = in6_pcbbind(inp, NULL, td);
383 			if (error)
384 				return (error);
385 		}
386 		inp->in6p_laddr = *addr6;
387 	}
388 	inp->in6p_faddr = sin6->sin6_addr;
389 	inp->inp_fport = sin6->sin6_port;
390 	/* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
391 	inp->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK;
392 	if (inp->in6p_flags & IN6P_AUTOFLOWLABEL)
393 		inp->in6p_flowinfo |=
394 		    (htonl(ip6_flow_seq++) & IPV6_FLOWLABEL_MASK);
395 
396 	in_pcbinsconnhash(inp);
397 	return (0);
398 }
399 
400 void
401 in6_pcbdisconnect(struct inpcb *inp)
402 {
403 	bzero((caddr_t)&inp->in6p_faddr, sizeof(inp->in6p_faddr));
404 	inp->inp_fport = 0;
405 	/* clear flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
406 	inp->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK;
407 	in_pcbremconnhash(inp);
408 	if (inp->inp_socket->so_state & SS_NOFDREF)
409 		in6_pcbdetach(inp);
410 }
411 
412 void
413 in6_pcbdetach(struct inpcb *inp)
414 {
415 	struct socket *so = inp->inp_socket;
416 	struct inpcbinfo *ipi = inp->inp_pcbinfo;
417 
418 #ifdef IPSEC
419 	if (inp->in6p_sp != NULL)
420 		ipsec6_delete_pcbpolicy(inp);
421 #endif /* IPSEC */
422 	inp->inp_gencnt = ++ipi->ipi_gencnt;
423 	in_pcbremlists(inp);
424 	so->so_pcb = NULL;
425 	KKASSERT((so->so_state & SS_ASSERTINPROG) == 0);
426 	sofree(so);		/* remove pcb ref */
427 
428 	if (inp->in6p_options)
429 		m_freem(inp->in6p_options);
430 	ip6_freepcbopts(inp->in6p_outputopts);
431 	ip6_freemoptions(inp->in6p_moptions);
432 	if (inp->in6p_route.ro_rt)
433 		rtfree(inp->in6p_route.ro_rt);
434 	/* Check and free IPv4 related resources in case of mapped addr */
435 	if (inp->inp_options)
436 		m_free(inp->inp_options);
437 	ip_freemoptions(inp->inp_moptions);
438 
439 	kfree(inp, M_PCB);
440 }
441 
442 /*
443  * The socket may have an invalid PCB, i.e. NULL.  For example, a TCP
444  * socket received RST.
445  */
446 static int
447 in6_setsockaddr(struct socket *so, struct sockaddr **nam)
448 {
449 	struct inpcb *inp;
450 	struct sockaddr_in6 *sin6;
451 
452 	KASSERT(curthread->td_type == TD_TYPE_NETISR, ("not in netisr"));
453 	inp = so->so_pcb;
454 	if (!inp)
455 		return EINVAL;
456 
457 	sin6 = kmalloc(sizeof *sin6, M_SONAME, M_WAITOK | M_ZERO);
458 	sin6->sin6_family = AF_INET6;
459 	sin6->sin6_len = sizeof(*sin6);
460 	sin6->sin6_port = inp->inp_lport;
461 	sin6->sin6_addr = inp->in6p_laddr;
462 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
463 		sin6->sin6_scope_id = ntohs(sin6->sin6_addr.s6_addr16[1]);
464 	else
465 		sin6->sin6_scope_id = 0;	/*XXX*/
466 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
467 		sin6->sin6_addr.s6_addr16[1] = 0;
468 
469 	*nam = (struct sockaddr *)sin6;
470 	return 0;
471 }
472 
473 void
474 in6_setsockaddr_dispatch(netmsg_t msg)
475 {
476 	int error;
477 
478 	error = in6_setsockaddr(msg->sockaddr.base.nm_so, msg->sockaddr.nm_nam);
479 	lwkt_replymsg(&msg->sockaddr.base.lmsg, error);
480 }
481 
482 void
483 in6_setpeeraddr_dispatch(netmsg_t msg)
484 {
485 	int error;
486 
487 	error = in6_setpeeraddr(msg->peeraddr.base.nm_so, msg->peeraddr.nm_nam);
488 	lwkt_replymsg(&msg->peeraddr.base.lmsg, error);
489 }
490 
491 /*
492  * The socket may have an invalid PCB, i.e. NULL.  For example, a TCP
493  * socket received RST.
494  */
495 int
496 in6_setpeeraddr(struct socket *so, struct sockaddr **nam)
497 {
498 	struct inpcb *inp;
499 	struct sockaddr_in6 *sin6;
500 
501 	KASSERT(curthread->td_type == TD_TYPE_NETISR, ("not in netisr"));
502 	inp = so->so_pcb;
503 	if (!inp)
504 		return EINVAL;
505 
506 	sin6 = kmalloc(sizeof(*sin6), M_SONAME, M_WAITOK | M_ZERO);
507 	sin6->sin6_family = AF_INET6;
508 	sin6->sin6_len = sizeof(struct sockaddr_in6);
509 	sin6->sin6_port = inp->inp_fport;
510 	sin6->sin6_addr = inp->in6p_faddr;
511 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
512 		sin6->sin6_scope_id = ntohs(sin6->sin6_addr.s6_addr16[1]);
513 	else
514 		sin6->sin6_scope_id = 0;	/*XXX*/
515 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
516 		sin6->sin6_addr.s6_addr16[1] = 0;
517 
518 	*nam = (struct sockaddr *)sin6;
519 	return 0;
520 }
521 
522 /*
523  * Pass some notification to all connections of a protocol
524  * associated with address dst.  The local address and/or port numbers
525  * may be specified to limit the search.  The "usual action" will be
526  * taken, depending on the ctlinput cmd.  The caller must filter any
527  * cmds that are uninteresting (e.g., no error in the map).
528  * Call the protocol specific routine (if any) to report
529  * any errors for each matching socket.
530  */
531 void
532 in6_pcbnotify(struct inpcbinfo *pcbinfo, struct sockaddr *dst, in_port_t fport,
533     const struct sockaddr *src, in_port_t lport, int cmd, int arg,
534     inp_notify_t notify)
535 {
536 	struct inpcb *inp, *marker;
537 	struct sockaddr_in6 sa6_src, *sa6_dst;
538 	u_int32_t flowinfo;
539 
540 	if ((unsigned)cmd >= PRC_NCMDS || dst->sa_family != AF_INET6)
541 		return;
542 
543 	sa6_dst = (struct sockaddr_in6 *)dst;
544 	if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr))
545 		return;
546 
547 	/*
548 	 * note that src can be NULL when we get notify by local fragmentation.
549 	 */
550 	sa6_src = (src == NULL) ? sa6_any : *(const struct sockaddr_in6 *)src;
551 	flowinfo = sa6_src.sin6_flowinfo;
552 
553 	/*
554 	 * Redirects go to all references to the destination,
555 	 * and use in6_rtchange to invalidate the route cache.
556 	 * Dead host indications: also use in6_rtchange to invalidate
557 	 * the cache, and deliver the error to all the sockets.
558 	 * Otherwise, if we have knowledge of the local port and address,
559 	 * deliver only to that socket.
560 	 */
561 	if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
562 		fport = 0;
563 		lport = 0;
564 		bzero((caddr_t)&sa6_src.sin6_addr, sizeof(sa6_src.sin6_addr));
565 
566 		if (cmd != PRC_HOSTDEAD)
567 			notify = in6_rtchange;
568 	}
569 	if (cmd != PRC_MSGSIZE)
570 		arg = inet6ctlerrmap[cmd];
571 
572 	marker = in_pcbmarker(mycpuid);
573 
574 	GET_PCBINFO_TOKEN(pcbinfo);
575 
576 	LIST_INSERT_HEAD(&pcbinfo->pcblisthead, marker, inp_list);
577 	while ((inp = LIST_NEXT(marker, inp_list)) != NULL) {
578 		LIST_REMOVE(marker, inp_list);
579 		LIST_INSERT_AFTER(inp, marker, inp_list);
580 
581 		if (inp->inp_flags & INP_PLACEMARKER)
582 			continue;
583 
584 		if (!INP_ISIPV6(inp))
585 			continue;
586 		/*
587 		 * If the error designates a new path MTU for a destination
588 		 * and the application (associated with this socket) wanted to
589 		 * know the value, notify. Note that we notify for all
590 		 * disconnected sockets if the corresponding application
591 		 * wanted. This is because some UDP applications keep sending
592 		 * sockets disconnected.
593 		 * XXX: should we avoid to notify the value to TCP sockets?
594 		 */
595 		if (cmd == PRC_MSGSIZE && (inp->inp_flags & IN6P_MTU) != 0 &&
596 		    (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
597 		     IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &sa6_dst->sin6_addr))) {
598 			ip6_notify_pmtu(inp, (struct sockaddr_in6 *)dst, &arg);
599 		}
600 
601 		/*
602 		 * Detect if we should notify the error. If no source and
603 		 * destination ports are specifed, but non-zero flowinfo and
604 		 * local address match, notify the error. This is the case
605 		 * when the error is delivered with an encrypted buffer
606 		 * by ESP. Otherwise, just compare addresses and ports
607 		 * as usual.
608 		 */
609 		if (lport == 0 && fport == 0 && flowinfo &&
610 		    inp->inp_socket != NULL &&
611 		    flowinfo == (inp->in6p_flowinfo & IPV6_FLOWLABEL_MASK) &&
612 		    IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &sa6_src.sin6_addr))
613 			goto do_notify;
614 		else if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
615 					     &sa6_dst->sin6_addr) ||
616 			 inp->inp_socket == 0 ||
617 			 (lport && inp->inp_lport != lport) ||
618 			 (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) &&
619 			  !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
620 					      &sa6_src.sin6_addr)) ||
621 			 (fport && inp->inp_fport != fport))
622 			continue;
623 
624 do_notify:
625 		if (notify)
626 			(*notify)(inp, arg);
627 	}
628 	LIST_REMOVE(marker, inp_list);
629 
630 	REL_PCBINFO_TOKEN(pcbinfo);
631 }
632 
633 /*
634  * Lookup a PCB based on the local address and port.
635  */
636 struct inpcb *
637 in6_pcblookup_local(struct inpcbportinfo *portinfo,
638     const struct in6_addr *laddr, u_int lport_arg, int wild_okay,
639     struct ucred *cred)
640 {
641 	struct inpcb *inp;
642 	int matchwild = 3, wildcard;
643 	u_short lport = lport_arg;
644 	struct inpcbporthead *porthash;
645 	struct inpcbport *phd;
646 	struct inpcb *match = NULL;
647 
648 	/*
649 	 * If the porthashbase is shared across several cpus, it must
650 	 * have been locked.
651 	 */
652 	ASSERT_PORT_TOKEN_HELD(portinfo);
653 
654 	/*
655 	 * Best fit PCB lookup.
656 	 *
657 	 * First see if this local port is in use by looking on the
658 	 * port hash list.
659 	 */
660 	porthash = &portinfo->porthashbase[
661 				INP_PCBPORTHASH(lport, portinfo->porthashmask)];
662 	LIST_FOREACH(phd, porthash, phd_hash) {
663 		if (phd->phd_port == lport)
664 			break;
665 	}
666 
667 	if (phd != NULL) {
668 		/*
669 		 * Port is in use by one or more PCBs. Look for best
670 		 * fit.
671 		 */
672 		LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
673 			wildcard = 0;
674 			if (!INP_ISIPV6(inp))
675 				continue;
676 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr))
677 				wildcard++;
678 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
679 				if (IN6_IS_ADDR_UNSPECIFIED(laddr))
680 					wildcard++;
681 				else if (!IN6_ARE_ADDR_EQUAL(
682 					&inp->in6p_laddr, laddr))
683 					continue;
684 			} else {
685 				if (!IN6_IS_ADDR_UNSPECIFIED(laddr))
686 					wildcard++;
687 			}
688 			if (wildcard && !wild_okay)
689 				continue;
690 			if (wildcard < matchwild &&
691 			    (cred == NULL ||
692 			     cred->cr_prison ==
693 					inp->inp_socket->so_cred->cr_prison)) {
694 				match = inp;
695 				matchwild = wildcard;
696 				if (wildcard == 0)
697 					break;
698 				else
699 					matchwild = wildcard;
700 			}
701 		}
702 	}
703 	return (match);
704 }
705 
706 void
707 in6_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
708 {
709 	struct in6pcb *in6p, *marker;
710 	struct ip6_moptions *im6o;
711 	struct in6_multi_mship *imm, *nimm;
712 
713 	/*
714 	 * We only need to make sure that we are in netisr0, where all
715 	 * multicast operation happen.  We could check inpcbinfo which
716 	 * does not belong to netisr0 by holding the inpcbinfo's token.
717 	 * In this case, the pcbinfo must be able to be shared, i.e.
718 	 * pcbinfo->infotoken is not NULL.
719 	 */
720 	ASSERT_IN_NETISR(0);
721 	KASSERT(pcbinfo->cpu == 0 || pcbinfo->infotoken != NULL,
722 	    ("pcbinfo could not be shared"));
723 
724 	/*
725 	 * Get a marker for the current netisr (netisr0).
726 	 *
727 	 * It is possible that the multicast address deletion blocks,
728 	 * which could cause temporary token releasing.  So we use
729 	 * inpcb marker here to get a coherent view of the inpcb list.
730 	 *
731 	 * While, on the other hand, moptions are only added and deleted
732 	 * in netisr0, so we would not see staled moption or miss moption
733 	 * even if the token was released due to the blocking multicast
734 	 * address deletion.
735 	 */
736 	marker = in_pcbmarker(mycpuid);
737 
738 	GET_PCBINFO_TOKEN(pcbinfo);
739 
740 	LIST_INSERT_HEAD(&pcbinfo->pcblisthead, marker, inp_list);
741 	while ((in6p = LIST_NEXT(marker, inp_list)) != NULL) {
742 		LIST_REMOVE(marker, inp_list);
743 		LIST_INSERT_AFTER(in6p, marker, inp_list);
744 
745 		if (in6p->in6p_flags & INP_PLACEMARKER)
746 			continue;
747 		im6o = in6p->in6p_moptions;
748 		if (INP_ISIPV6(in6p) && im6o) {
749 			/*
750 			 * Unselect the outgoing interface if it is being
751 			 * detached.
752 			 */
753 			if (im6o->im6o_multicast_ifp == ifp)
754 				im6o->im6o_multicast_ifp = NULL;
755 
756 			/*
757 			 * Drop multicast group membership if we joined
758 			 * through the interface being detached.
759 			 * XXX controversial - is it really legal for kernel
760 			 * to force this?
761 			 */
762 			for (imm = im6o->im6o_memberships.lh_first;
763 			     imm != NULL; imm = nimm) {
764 				nimm = imm->i6mm_chain.le_next;
765 				if (imm->i6mm_maddr->in6m_ifp == ifp) {
766 					LIST_REMOVE(imm, i6mm_chain);
767 					in6_delmulti(imm->i6mm_maddr);
768 					kfree(imm, M_IPMADDR);
769 				}
770 			}
771 		}
772 	}
773 	LIST_REMOVE(marker, inp_list);
774 
775 	REL_PCBINFO_TOKEN(pcbinfo);
776 }
777 
778 /*
779  * Check for alternatives when higher level complains
780  * about service problems.  For now, invalidate cached
781  * routing information.  If the route was created dynamically
782  * (by a redirect), time to try a default gateway again.
783  */
784 void
785 in6_losing(struct inpcb *in6p)
786 {
787 	struct rtentry *rt;
788 	struct rt_addrinfo info;
789 
790 	if ((rt = in6p->in6p_route.ro_rt) != NULL) {
791 		bzero((caddr_t)&info, sizeof(info));
792 		info.rti_flags = rt->rt_flags;
793 		info.rti_info[RTAX_DST] = rt_key(rt);
794 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
795 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
796 		rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
797 		if (rt->rt_flags & RTF_DYNAMIC) {
798 			rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
799 			    rt_mask(rt), rt->rt_flags, NULL);
800 		}
801 		in6p->in6p_route.ro_rt = NULL;
802 		rtfree(rt);
803 		/*
804 		 * A new route can be allocated
805 		 * the next time output is attempted.
806 		 */
807 	}
808 }
809 
810 /*
811  * After a routing change, flush old routing
812  * and allocate a (hopefully) better one.
813  */
814 void
815 in6_rtchange(struct inpcb *inp, int error)
816 {
817 	if (inp->in6p_route.ro_rt) {
818 		rtfree(inp->in6p_route.ro_rt);
819 		inp->in6p_route.ro_rt = 0;
820 		/*
821 		 * A new route can be allocated the next time
822 		 * output is attempted.
823 		 */
824 	}
825 }
826 
827 /*
828  * Lookup PCB in hash list.
829  */
830 struct inpcb *
831 in6_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
832 		   u_int fport_arg, struct in6_addr *laddr, u_int lport_arg,
833 		   int wildcard, struct ifnet *ifp)
834 {
835 	struct inpcbhead *head;
836 	struct inpcb *inp;
837 	struct inpcb *jinp = NULL;
838 	u_short fport = fport_arg, lport = lport_arg;
839 	int faith;
840 
841 	if (faithprefix_p != NULL)
842 		faith = (*faithprefix_p)(laddr);
843 	else
844 		faith = 0;
845 
846 	/*
847 	 * First look for an exact match.
848 	 */
849 	head = &pcbinfo->hashbase[INP_PCBCONNHASH(faddr->s6_addr32[3] /* XXX */,
850 					      fport,
851 					      laddr->s6_addr32[3], /* XXX JH */
852 					      lport,
853 					      pcbinfo->hashmask)];
854 	LIST_FOREACH(inp, head, inp_hash) {
855 		if (!INP_ISIPV6(inp))
856 			continue;
857 		if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) &&
858 		    IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
859 		    inp->inp_fport == fport &&
860 		    inp->inp_lport == lport) {
861 			/*
862 			 * Found.
863 			 */
864 			if (inp->inp_socket == NULL ||
865 			inp->inp_socket->so_cred->cr_prison == NULL) {
866 				return (inp);
867 			} else {
868 				if  (jinp == NULL)
869 					jinp = inp;
870 			}
871 		}
872 	}
873 	if (jinp != NULL)
874 		return(jinp);
875 
876 	if (wildcard) {
877 		struct inpcontainerhead *chead;
878 		struct inpcontainer *ic;
879 		struct inpcb *local_wild = NULL;
880 		struct inpcb *jinp_wild = NULL;
881 		struct sockaddr_in6 jsin6;
882 		struct ucred *cred;
883 
884 		/*
885 		 * Order of socket selection:
886 		 * 1. non-jailed, non-wild.
887 		 * 2. non-jailed, wild.
888 		 * 3. jailed, non-wild.
889 		 * 4. jailed, wild.
890 		 */
891 		jsin6.sin6_family = AF_INET6;
892 		chead = &pcbinfo->wildcardhashbase[INP_PCBWILDCARDHASH(lport,
893 		    pcbinfo->wildcardhashmask)];
894 
895 		GET_PCBINFO_TOKEN(pcbinfo);
896 		LIST_FOREACH(ic, chead, ic_list) {
897 			inp = ic->ic_inp;
898 			if (inp->inp_flags & INP_PLACEMARKER)
899 				continue;
900 
901 			if (!INP_ISIPV6(inp))
902 				continue;
903 			if (inp->inp_socket != NULL)
904 				cred = inp->inp_socket->so_cred;
905 			else
906 				cred = NULL;
907 
908 			if (cred != NULL && jailed(cred)) {
909 				if (jinp != NULL) {
910 					continue;
911 				} else {
912 		                        jsin6.sin6_addr = *laddr;
913 					if (!jailed_ip(cred->cr_prison,
914 					    (struct sockaddr *)&jsin6))
915 						continue;
916 				}
917 			}
918 			if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
919 			    inp->inp_lport == lport) {
920 				if (faith && (inp->inp_flags & INP_FAITH) == 0)
921 					continue;
922 				if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
923 						       laddr)) {
924 					if (cred != NULL && jailed(cred)) {
925 						jinp = inp;
926 					} else {
927 						REL_PCBINFO_TOKEN(pcbinfo);
928 						return (inp);
929 					}
930 				} else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
931 					if (cred != NULL && jailed(cred))
932 						jinp_wild = inp;
933 					else
934 						local_wild = inp;
935 				}
936 			}
937 		}
938 		REL_PCBINFO_TOKEN(pcbinfo);
939 
940 		if (local_wild != NULL)
941 			return (local_wild);
942 		if (jinp != NULL)
943 			return (jinp);
944 		return (jinp_wild);
945 	}
946 
947 	/*
948 	 * Not found.
949 	 */
950 	return (NULL);
951 }
952 
953 void
954 init_sin6(struct sockaddr_in6 *sin6, struct mbuf *m)
955 {
956 	struct ip6_hdr *ip;
957 
958 	ip = mtod(m, struct ip6_hdr *);
959 	bzero(sin6, sizeof(*sin6));
960 	sin6->sin6_len = sizeof(*sin6);
961 	sin6->sin6_family = AF_INET6;
962 	sin6->sin6_addr = ip->ip6_src;
963 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
964 		sin6->sin6_addr.s6_addr16[1] = 0;
965 	sin6->sin6_scope_id =
966 		(m->m_pkthdr.rcvif && IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
967 		? m->m_pkthdr.rcvif->if_index : 0;
968 
969 	return;
970 }
971 
972 void
973 in6_savefaddr(struct socket *so, const struct sockaddr *faddr)
974 {
975 	struct sockaddr_in6 *sin6;
976 
977 	KASSERT(faddr->sa_family == AF_INET6,
978 	    ("not AF_INET6 faddr %d", faddr->sa_family));
979 
980 	sin6 = kmalloc(sizeof(*sin6), M_SONAME, M_WAITOK | M_ZERO);
981 	sin6->sin6_family = AF_INET6;
982 	sin6->sin6_len = sizeof(*sin6);
983 
984 	sin6->sin6_port = ((const struct sockaddr_in6 *)faddr)->sin6_port;
985 	sin6->sin6_addr = ((const struct sockaddr_in6 *)faddr)->sin6_addr;
986 
987 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
988 		sin6->sin6_scope_id = ntohs(sin6->sin6_addr.s6_addr16[1]);
989 	else
990 		sin6->sin6_scope_id = 0;	/*XXX*/
991 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
992 		sin6->sin6_addr.s6_addr16[1] = 0;
993 
994 	so->so_faddr = (struct sockaddr *)sin6;
995 }
996