xref: /dragonfly/sys/netinet6/in6_pcb.c (revision 7d3e9a5b)
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 
68 #include <sys/param.h>
69 #include <sys/systm.h>
70 #include <sys/malloc.h>
71 #include <sys/mbuf.h>
72 #include <sys/domain.h>
73 #include <sys/protosw.h>
74 #include <sys/socket.h>
75 #include <sys/socketvar.h>
76 #include <sys/sockio.h>
77 #include <sys/errno.h>
78 #include <sys/time.h>
79 #include <sys/proc.h>
80 #include <sys/priv.h>
81 #include <sys/jail.h>
82 
83 #include <sys/msgport2.h>
84 
85 #include <vm/vm_zone.h>
86 
87 #include <net/if.h>
88 #include <net/if_types.h>
89 #include <net/route.h>
90 #include <net/netisr2.h>
91 
92 #include <netinet/in.h>
93 #include <netinet/in_var.h>
94 #include <netinet/in_systm.h>
95 #include <netinet/ip6.h>
96 #include <netinet/ip_var.h>
97 #include <netinet6/ip6_var.h>
98 #include <netinet6/nd6.h>
99 #include <netinet/in_pcb.h>
100 #include <netinet6/in6_pcb.h>
101 
102 struct	in6_addr zeroin6_addr;
103 
104 int
105 in6_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct thread *td)
106 {
107 	struct socket *so = inp->inp_socket;
108 	struct sockaddr_in6 jsin6;
109 	int error;
110 
111 	if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
112 		return (EINVAL);
113 
114 	if (nam) {
115 		struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
116 		struct inpcbinfo *pcbinfo;
117 		struct inpcbportinfo *portinfo;
118 		struct inpcbporthead *porthash;
119 		int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
120 		struct ucred *cred = NULL;
121 		struct inpcb *t;
122 		u_short	lport, lport_ho;
123 
124 		if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
125 			wild = 1;
126 		if (td->td_proc != NULL)
127 			cred = td->td_proc->p_ucred;
128 
129 		if (nam->sa_len != sizeof(*sin6))
130 			return (EINVAL);
131 		/*
132 		 * family check.
133 		 */
134 		if (nam->sa_family != AF_INET6)
135 			return (EAFNOSUPPORT);
136 
137 		/* Reject v4-mapped address */
138 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr))
139 			return (EADDRNOTAVAIL);
140 
141 		if (!prison_replace_wildcards(td, nam))
142 			return (EINVAL);
143 
144 		/* KAME hack: embed scopeid */
145 		if (in6_embedscope(&sin6->sin6_addr, sin6, inp, NULL) != 0)
146 			return (EINVAL);
147 		/* this must be cleared for ifa_ifwithaddr() */
148 		sin6->sin6_scope_id = 0;
149 
150 		lport = sin6->sin6_port;
151 		if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
152 			/*
153 			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
154 			 * allow compepte duplication of binding if
155 			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
156 			 * and a multicast address is bound on both
157 			 * new and duplicated sockets.
158 			 */
159 			if (so->so_options & SO_REUSEADDR)
160 				reuseport = SO_REUSEADDR|SO_REUSEPORT;
161 		} else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
162 			struct ifaddr *ia = NULL;
163 
164 			sin6->sin6_port = 0;		/* yech... */
165 			if (!prison_replace_wildcards(td, (struct sockaddr *)sin6)) {
166 				sin6->sin6_addr = kin6addr_any;
167 				return (EINVAL);
168 			}
169 			if ((ia = ifa_ifwithaddr((struct sockaddr *)sin6)) == NULL)
170 				return (EADDRNOTAVAIL);
171 
172 			/*
173 			 * XXX: bind to an anycast address might accidentally
174 			 * cause sending a packet with anycast source address.
175 			 * We should allow to bind to a deprecated address, since
176 			 * the application dares to use it.
177 			 */
178 			if (ia &&
179 			    ((struct in6_ifaddr *)ia)->ia6_flags &
180 			    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|IN6_IFF_DETACHED))
181 				return (EADDRNOTAVAIL);
182 		}
183 
184 		inp->in6p_laddr = sin6->sin6_addr;
185 
186 		if (lport == 0)
187 			goto auto_select;
188 		lport_ho = ntohs(lport);
189 
190 		/* GROSS */
191 		if (lport_ho < IPV6PORT_RESERVED && cred &&
192 		    priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0)) {
193 			inp->in6p_laddr = kin6addr_any;
194 			return (EACCES);
195 		}
196 
197 		/*
198 		 * Locate the proper portinfo based on lport
199 		 */
200 		pcbinfo = inp->inp_pcbinfo;
201 		portinfo =
202 		    &pcbinfo->portinfo[lport_ho % pcbinfo->portinfo_cnt];
203 		KKASSERT((lport_ho % pcbinfo->portinfo_cnt) ==
204 		    portinfo->offset);
205 
206 		/*
207 		 * This has to be atomic.  If the porthash is shared across
208 		 * multiple protocol threads (aka tcp) then the token must
209 		 * be held.
210 		 */
211 		porthash = in_pcbporthash_head(portinfo, lport);
212 		GET_PORTHASH_TOKEN(porthash);
213 
214 		if (so->so_cred->cr_uid != 0 &&
215 		    !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
216 			t = in6_pcblookup_local(porthash, &sin6->sin6_addr,
217 						lport, INPLOOKUP_WILDCARD,
218 						cred);
219 			if (t &&
220 			    (so->so_cred->cr_uid !=
221 			     t->inp_socket->so_cred->cr_uid)) {
222 				inp->in6p_laddr = kin6addr_any;
223 				error = EADDRINUSE;
224 				goto done;
225 			}
226 		}
227 		if (cred && cred->cr_prison &&
228 		    !prison_replace_wildcards(td, nam)) {
229 			inp->in6p_laddr = kin6addr_any;
230 			error = EADDRNOTAVAIL;
231 			goto done;
232 		}
233 		t = in6_pcblookup_local(porthash, &sin6->sin6_addr, lport,
234 					wild, cred);
235 		if (t && (reuseport & t->inp_socket->so_options) == 0) {
236 			inp->in6p_laddr = kin6addr_any;
237 			error = EADDRINUSE;
238 			goto done;
239 		}
240 
241 		inp->inp_lport = lport;
242 		in_pcbinsporthash(porthash, inp);
243 		error = 0;
244 done:
245 		REL_PORTHASH_TOKEN(porthash);
246 		return (error);
247 	} else {
248 auto_select:
249 		jsin6.sin6_addr = inp->in6p_laddr;
250 		jsin6.sin6_family = AF_INET6;
251 		if (!prison_replace_wildcards(td, (struct sockaddr*)&jsin6)) {
252 			inp->in6p_laddr = kin6addr_any;
253 			inp->inp_lport = 0;
254 			return (EINVAL);
255 		}
256 
257 		return in6_pcbsetlport(&inp->in6p_laddr, inp, td);
258 	}
259 }
260 
261 /*
262  *   Transform old in6_pcbconnect() into an inner subroutine for new
263  *   in6_pcbconnect(): Do some validity-checking on the remote
264  *   address (in mbuf 'nam') and then determine local host address
265  *   (i.e., which interface) to use to access that remote host.
266  *
267  *   This preserves definition of in6_pcbconnect(), while supporting a
268  *   slightly different version for T/TCP.  (This is more than
269  *   a bit of a kludge, but cleaning up the internal interfaces would
270  *   have forced minor changes in every protocol).
271  */
272 
273 int
274 in6_pcbladdr(struct inpcb *inp, struct sockaddr *nam,
275 	     struct in6_addr **plocal_addr6, struct thread *td)
276 {
277 	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
278 	struct ifnet *ifp = NULL;
279 	int error = 0;
280 
281 	if (nam->sa_len != sizeof (*sin6))
282 		return (EINVAL);
283 	if (sin6->sin6_family != AF_INET6)
284 		return (EAFNOSUPPORT);
285 	if (sin6->sin6_port == 0)
286 		return (EADDRNOTAVAIL);
287 
288 	/* KAME hack: embed scopeid */
289 	if (in6_embedscope(&sin6->sin6_addr, sin6, inp, &ifp) != 0)
290 		return EINVAL;
291 
292 	if (in6_ifaddr) {
293 		/*
294 		 * If the destination address is UNSPECIFIED addr,
295 		 * use the loopback addr, e.g ::1.
296 		 */
297 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
298 			sin6->sin6_addr = kin6addr_loopback;
299 	}
300 	{
301 		/*
302 		 * XXX: in6_selectsrc might replace the bound local address
303 		 * with the address specified by setsockopt(IPV6_PKTINFO).
304 		 * Is it the intended behavior?
305 		 */
306 		*plocal_addr6 = in6_selectsrc(sin6, inp->in6p_outputopts,
307 					      inp->in6p_moptions,
308 					      &inp->in6p_route,
309 					      &inp->in6p_laddr, &error, td);
310 		if (*plocal_addr6 == NULL) {
311 			if (error == 0)
312 				error = EADDRNOTAVAIL;
313 			return (error);
314 		}
315 		/*
316 		 * Don't do pcblookup call here; return interface in
317 		 * plocal_addr6
318 		 * and exit to caller, that will do the lookup.
319 		 */
320 	}
321 
322 	if (inp->in6p_route.ro_rt)
323 		ifp = inp->in6p_route.ro_rt->rt_ifp;
324 
325 	return (0);
326 }
327 
328 /*
329  * Outer subroutine:
330  * Connect from a socket to a specified address.
331  * Both address and port must be specified in argument sin.
332  * If don't have a local address for this socket yet,
333  * then pick one.
334  */
335 int
336 in6_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct thread *td)
337 {
338 	struct in6_addr *addr6;
339 	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
340 	int error;
341 
342 	/* Reject v4-mapped address */
343 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr))
344 		return EADDRNOTAVAIL;
345 
346 	/*
347 	 * Call inner routine, to assign local interface address.
348 	 * in6_pcbladdr() may automatically fill in sin6_scope_id.
349 	 */
350 	if ((error = in6_pcbladdr(inp, nam, &addr6, td)) != 0)
351 		return (error);
352 
353 	if (in6_pcblookup_hash(inp->inp_pcbinfo, &sin6->sin6_addr,
354 			       sin6->sin6_port,
355 			      IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
356 			      ? addr6 : &inp->in6p_laddr,
357 			      inp->inp_lport, 0, NULL) != NULL) {
358 		return (EADDRINUSE);
359 	}
360 	if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
361 		if (inp->inp_lport == 0) {
362 			error = in6_pcbbind(inp, NULL, td);
363 			if (error)
364 				return (error);
365 		}
366 		inp->in6p_laddr = *addr6;
367 	}
368 	inp->in6p_faddr = sin6->sin6_addr;
369 	inp->inp_fport = sin6->sin6_port;
370 	/* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
371 	inp->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK;
372 	if (inp->in6p_flags & IN6P_AUTOFLOWLABEL)
373 		inp->in6p_flowinfo |=
374 		    (htonl(ip6_flow_seq++) & IPV6_FLOWLABEL_MASK);
375 
376 	in_pcbinsconnhash(inp);
377 	return (0);
378 }
379 
380 void
381 in6_pcbdisconnect(struct inpcb *inp)
382 {
383 	bzero((caddr_t)&inp->in6p_faddr, sizeof(inp->in6p_faddr));
384 	inp->inp_fport = 0;
385 	/* clear flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
386 	inp->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK;
387 	in_pcbremconnhash(inp);
388 	if (inp->inp_socket->so_state & SS_NOFDREF)
389 		in6_pcbdetach(inp);
390 }
391 
392 void
393 in6_pcbdetach(struct inpcb *inp)
394 {
395 	struct socket *so = inp->inp_socket;
396 	struct inpcbinfo *ipi = inp->inp_pcbinfo;
397 
398 	inp->inp_gencnt = ++ipi->ipi_gencnt;
399 	in_pcbremlists(inp);
400 	so->so_pcb = NULL;
401 	KKASSERT((so->so_state & SS_ASSERTINPROG) == 0);
402 	sofree(so);		/* remove pcb ref */
403 
404 	if (inp->in6p_options)
405 		m_freem(inp->in6p_options);
406 	ip6_freepcbopts(inp->in6p_outputopts);
407 	ip6_freemoptions(inp->in6p_moptions);
408 	if (inp->in6p_route.ro_rt)
409 		rtfree(inp->in6p_route.ro_rt);
410 	/* Check and free IPv4 related resources in case of mapped addr */
411 	if (inp->inp_options)
412 		m_free(inp->inp_options);
413 	ip_freemoptions(inp->inp_moptions);
414 
415 	kfree(inp, M_PCB);
416 }
417 
418 /*
419  * The socket may have an invalid PCB, i.e. NULL.  For example, a TCP
420  * socket received RST.
421  */
422 static int
423 in6_setsockaddr(struct socket *so, struct sockaddr **nam)
424 {
425 	struct inpcb *inp;
426 	struct sockaddr_in6 *sin6;
427 
428 	KASSERT(curthread->td_type == TD_TYPE_NETISR, ("not in netisr"));
429 	inp = so->so_pcb;
430 	if (!inp)
431 		return EINVAL;
432 
433 	sin6 = kmalloc(sizeof *sin6, M_SONAME, M_WAITOK | M_ZERO);
434 	sin6->sin6_family = AF_INET6;
435 	sin6->sin6_len = sizeof(*sin6);
436 	sin6->sin6_port = inp->inp_lport;
437 	sin6->sin6_addr = inp->in6p_laddr;
438 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
439 		sin6->sin6_scope_id = ntohs(sin6->sin6_addr.s6_addr16[1]);
440 	else
441 		sin6->sin6_scope_id = 0;	/*XXX*/
442 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
443 		sin6->sin6_addr.s6_addr16[1] = 0;
444 
445 	*nam = (struct sockaddr *)sin6;
446 	return 0;
447 }
448 
449 void
450 in6_setsockaddr_dispatch(netmsg_t msg)
451 {
452 	int error;
453 
454 	error = in6_setsockaddr(msg->sockaddr.base.nm_so, msg->sockaddr.nm_nam);
455 	lwkt_replymsg(&msg->sockaddr.base.lmsg, error);
456 }
457 
458 void
459 in6_setpeeraddr_dispatch(netmsg_t msg)
460 {
461 	int error;
462 
463 	error = in6_setpeeraddr(msg->peeraddr.base.nm_so, msg->peeraddr.nm_nam);
464 	lwkt_replymsg(&msg->peeraddr.base.lmsg, error);
465 }
466 
467 /*
468  * The socket may have an invalid PCB, i.e. NULL.  For example, a TCP
469  * socket received RST.
470  */
471 int
472 in6_setpeeraddr(struct socket *so, struct sockaddr **nam)
473 {
474 	struct inpcb *inp;
475 	struct sockaddr_in6 *sin6;
476 
477 	KASSERT(curthread->td_type == TD_TYPE_NETISR, ("not in netisr"));
478 	inp = so->so_pcb;
479 	if (!inp)
480 		return EINVAL;
481 
482 	sin6 = kmalloc(sizeof(*sin6), M_SONAME, M_WAITOK | M_ZERO);
483 	sin6->sin6_family = AF_INET6;
484 	sin6->sin6_len = sizeof(struct sockaddr_in6);
485 	sin6->sin6_port = inp->inp_fport;
486 	sin6->sin6_addr = inp->in6p_faddr;
487 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
488 		sin6->sin6_scope_id = ntohs(sin6->sin6_addr.s6_addr16[1]);
489 	else
490 		sin6->sin6_scope_id = 0;	/*XXX*/
491 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
492 		sin6->sin6_addr.s6_addr16[1] = 0;
493 
494 	*nam = (struct sockaddr *)sin6;
495 	return 0;
496 }
497 
498 /*
499  * Pass some notification to all connections of a protocol
500  * associated with address dst.  The local address and/or port numbers
501  * may be specified to limit the search.  The "usual action" will be
502  * taken, depending on the ctlinput cmd.  The caller must filter any
503  * cmds that are uninteresting (e.g., no error in the map).
504  * Call the protocol specific routine (if any) to report
505  * any errors for each matching socket.
506  */
507 void
508 in6_pcbnotify(struct inpcbinfo *pcbinfo, struct sockaddr *dst, in_port_t fport,
509     const struct sockaddr *src, in_port_t lport, int cmd, int arg,
510     inp_notify_t notify)
511 {
512 	struct inpcb *inp, *marker;
513 	struct sockaddr_in6 sa6_src, *sa6_dst;
514 	u_int32_t flowinfo;
515 
516 	if ((unsigned)cmd >= PRC_NCMDS || dst->sa_family != AF_INET6)
517 		return;
518 
519 	sa6_dst = (struct sockaddr_in6 *)dst;
520 	if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr))
521 		return;
522 
523 	/*
524 	 * note that src can be NULL when we get notify by local fragmentation.
525 	 */
526 	sa6_src = (src == NULL) ? sa6_any : *(const struct sockaddr_in6 *)src;
527 	flowinfo = sa6_src.sin6_flowinfo;
528 
529 	/*
530 	 * Redirects go to all references to the destination,
531 	 * and use in6_rtchange to invalidate the route cache.
532 	 * Dead host indications: also use in6_rtchange to invalidate
533 	 * the cache, and deliver the error to all the sockets.
534 	 * Otherwise, if we have knowledge of the local port and address,
535 	 * deliver only to that socket.
536 	 */
537 	if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
538 		fport = 0;
539 		lport = 0;
540 		bzero((caddr_t)&sa6_src.sin6_addr, sizeof(sa6_src.sin6_addr));
541 
542 		if (cmd != PRC_HOSTDEAD)
543 			notify = in6_rtchange;
544 	}
545 	if (cmd != PRC_MSGSIZE)
546 		arg = inet6ctlerrmap[cmd];
547 
548 	marker = in_pcbmarker();
549 
550 	GET_PCBINFO_TOKEN(pcbinfo);
551 
552 	LIST_INSERT_HEAD(&pcbinfo->pcblisthead, marker, inp_list);
553 	while ((inp = LIST_NEXT(marker, inp_list)) != NULL) {
554 		LIST_REMOVE(marker, inp_list);
555 		LIST_INSERT_AFTER(inp, marker, inp_list);
556 
557 		if (inp->inp_flags & INP_PLACEMARKER)
558 			continue;
559 
560 		if (!INP_ISIPV6(inp))
561 			continue;
562 		/*
563 		 * If the error designates a new path MTU for a destination
564 		 * and the application (associated with this socket) wanted to
565 		 * know the value, notify. Note that we notify for all
566 		 * disconnected sockets if the corresponding application
567 		 * wanted. This is because some UDP applications keep sending
568 		 * sockets disconnected.
569 		 * XXX: should we avoid to notify the value to TCP sockets?
570 		 */
571 		if (cmd == PRC_MSGSIZE && (inp->inp_flags & IN6P_MTU) != 0 &&
572 		    (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
573 		     IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &sa6_dst->sin6_addr))) {
574 			ip6_notify_pmtu(inp, (struct sockaddr_in6 *)dst, &arg);
575 		}
576 
577 		/*
578 		 * Detect if we should notify the error. If no source and
579 		 * destination ports are specifed, but non-zero flowinfo and
580 		 * local address match, notify the error. This is the case
581 		 * when the error is delivered with an encrypted buffer
582 		 * by ESP. Otherwise, just compare addresses and ports
583 		 * as usual.
584 		 */
585 		if (lport == 0 && fport == 0 && flowinfo &&
586 		    inp->inp_socket != NULL &&
587 		    flowinfo == (inp->in6p_flowinfo & IPV6_FLOWLABEL_MASK) &&
588 		    IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &sa6_src.sin6_addr))
589 			goto do_notify;
590 		else if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
591 					     &sa6_dst->sin6_addr) ||
592 			 inp->inp_socket == 0 ||
593 			 (lport && inp->inp_lport != lport) ||
594 			 (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) &&
595 			  !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
596 					      &sa6_src.sin6_addr)) ||
597 			 (fport && inp->inp_fport != fport))
598 			continue;
599 
600 do_notify:
601 		if (notify)
602 			(*notify)(inp, arg);
603 	}
604 	LIST_REMOVE(marker, inp_list);
605 
606 	REL_PCBINFO_TOKEN(pcbinfo);
607 }
608 
609 /*
610  * Lookup a PCB based on the local address and port.
611  */
612 struct inpcb *
613 in6_pcblookup_local(struct inpcbporthead *porthash,
614     const struct in6_addr *laddr, u_int lport_arg, int wild_okay,
615     struct ucred *cred)
616 {
617 	struct prison *pscan;
618 	struct prison *pr;
619 	struct inpcb *inp;
620 	int matchwild = 3, wildcard;
621 	u_short lport = lport_arg;
622 	struct inpcbport *phd;
623 	struct inpcb *match = NULL;
624 
625 	/*
626 	 * If the porthashbase is shared across several cpus, it must
627 	 * have been locked.
628 	 */
629 	ASSERT_PORTHASH_TOKEN_HELD(porthash);
630 
631 	/*
632 	 * Best fit PCB lookup.
633 	 *
634 	 * First see if this local port is in use by looking on the
635 	 * port hash list.
636 	 */
637 	LIST_FOREACH(phd, porthash, phd_hash) {
638 		if (phd->phd_port == lport)
639 			break;
640 	}
641 
642 	if (phd != NULL) {
643 		pr = cred ? cred->cr_prison : NULL;
644 
645 		/*
646 		 * Port is in use by one or more PCBs. Look for best
647 		 * fit.
648 		 */
649 		LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
650 			wildcard = 0;
651 			if (!INP_ISIPV6(inp))
652 				continue;
653 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr))
654 				wildcard++;
655 
656 			if (inp->inp_socket && inp->inp_socket->so_cred)
657 				pscan = inp->inp_socket->so_cred->cr_prison;
658 			else
659 				pscan = NULL;
660 			if (pr != pscan)
661 				continue;
662 
663 			if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
664 				if (!IN6_IS_ADDR_UNSPECIFIED(laddr))
665 					wildcard++;
666 			} else {
667 				if (IN6_IS_ADDR_UNSPECIFIED(laddr))
668 					wildcard++;
669 				else if (!IN6_ARE_ADDR_EQUAL(
670 						    &inp->in6p_laddr, laddr))
671 					continue;
672 			}
673 			if (wildcard && !wild_okay)
674 				continue;
675 			if (wildcard < matchwild) {
676 				match = inp;
677 				matchwild = wildcard;
678 				if (wildcard == 0)
679 					break;
680 			}
681 		}
682 	}
683 	return (match);
684 }
685 
686 void
687 in6_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
688 {
689 	struct in6pcb *in6p, *marker;
690 	struct ip6_moptions *im6o;
691 	struct in6_multi_mship *imm, *nimm;
692 
693 	/*
694 	 * We only need to make sure that we are in netisr0, where all
695 	 * multicast operation happen.  We could check inpcbinfo which
696 	 * does not belong to netisr0 by holding the inpcbinfo's token.
697 	 * In this case, the pcbinfo must be able to be shared, i.e.
698 	 * pcbinfo->infotoken is not NULL.
699 	 */
700 	ASSERT_NETISR0;
701 	KASSERT(pcbinfo->cpu == 0 || pcbinfo->infotoken != NULL,
702 	    ("pcbinfo could not be shared"));
703 
704 	/*
705 	 * Get a marker for the current netisr (netisr0).
706 	 *
707 	 * It is possible that the multicast address deletion blocks,
708 	 * which could cause temporary token releasing.  So we use
709 	 * inpcb marker here to get a coherent view of the inpcb list.
710 	 *
711 	 * While, on the other hand, moptions are only added and deleted
712 	 * in netisr0, so we would not see staled moption or miss moption
713 	 * even if the token was released due to the blocking multicast
714 	 * address deletion.
715 	 */
716 	marker = in_pcbmarker();
717 
718 	GET_PCBINFO_TOKEN(pcbinfo);
719 
720 	LIST_INSERT_HEAD(&pcbinfo->pcblisthead, marker, inp_list);
721 	while ((in6p = LIST_NEXT(marker, inp_list)) != NULL) {
722 		LIST_REMOVE(marker, inp_list);
723 		LIST_INSERT_AFTER(in6p, marker, inp_list);
724 
725 		if (in6p->in6p_flags & INP_PLACEMARKER)
726 			continue;
727 		im6o = in6p->in6p_moptions;
728 		if (INP_ISIPV6(in6p) && im6o) {
729 			/*
730 			 * Unselect the outgoing interface if it is being
731 			 * detached.
732 			 */
733 			if (im6o->im6o_multicast_ifp == ifp)
734 				im6o->im6o_multicast_ifp = NULL;
735 
736 			/*
737 			 * Drop multicast group membership if we joined
738 			 * through the interface being detached.
739 			 * XXX controversial - is it really legal for kernel
740 			 * to force this?
741 			 */
742 			for (imm = im6o->im6o_memberships.lh_first;
743 			     imm != NULL; imm = nimm) {
744 				nimm = imm->i6mm_chain.le_next;
745 				if (imm->i6mm_maddr->in6m_ifp == ifp) {
746 					LIST_REMOVE(imm, i6mm_chain);
747 					in6_delmulti(imm->i6mm_maddr);
748 					kfree(imm, M_IPMADDR);
749 				}
750 			}
751 		}
752 	}
753 	LIST_REMOVE(marker, inp_list);
754 
755 	REL_PCBINFO_TOKEN(pcbinfo);
756 }
757 
758 /*
759  * Check for alternatives when higher level complains
760  * about service problems.  For now, invalidate cached
761  * routing information.  If the route was created dynamically
762  * (by a redirect), time to try a default gateway again.
763  */
764 void
765 in6_losing(struct inpcb *in6p)
766 {
767 	struct rtentry *rt;
768 	struct rt_addrinfo info;
769 
770 	if ((rt = in6p->in6p_route.ro_rt) != NULL) {
771 		bzero((caddr_t)&info, sizeof(info));
772 		info.rti_flags = rt->rt_flags;
773 		info.rti_info[RTAX_DST] = rt_key(rt);
774 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
775 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
776 		rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
777 		if (rt->rt_flags & RTF_DYNAMIC) {
778 			rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
779 			    rt_mask(rt), rt->rt_flags, NULL);
780 		}
781 		in6p->in6p_route.ro_rt = NULL;
782 		rtfree(rt);
783 		/*
784 		 * A new route can be allocated
785 		 * the next time output is attempted.
786 		 */
787 	}
788 }
789 
790 /*
791  * After a routing change, flush old routing
792  * and allocate a (hopefully) better one.
793  */
794 void
795 in6_rtchange(struct inpcb *inp, int error)
796 {
797 	if (inp->in6p_route.ro_rt) {
798 		rtfree(inp->in6p_route.ro_rt);
799 		inp->in6p_route.ro_rt = 0;
800 		/*
801 		 * A new route can be allocated the next time
802 		 * output is attempted.
803 		 */
804 	}
805 }
806 
807 /*
808  * Lookup PCB in hash list.
809  */
810 struct inpcb *
811 in6_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
812 		   u_int fport_arg, struct in6_addr *laddr, u_int lport_arg,
813 		   int wildcard, struct ifnet *ifp)
814 {
815 	struct inpcbhead *head;
816 	struct inpcb *inp;
817 	struct inpcb *jinp = NULL;
818 	u_short fport = fport_arg, lport = lport_arg;
819 
820 	/*
821 	 * First look for an exact match.
822 	 */
823 	head = &pcbinfo->hashbase[INP_PCBCONNHASH(faddr->s6_addr32[3] /* XXX */,
824 					      fport,
825 					      laddr->s6_addr32[3], /* XXX JH */
826 					      lport,
827 					      pcbinfo->hashmask)];
828 	LIST_FOREACH(inp, head, inp_hash) {
829 		if (!INP_ISIPV6(inp))
830 			continue;
831 		if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) &&
832 		    IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
833 		    inp->inp_fport == fport &&
834 		    inp->inp_lport == lport) {
835 			/*
836 			 * Found.
837 			 */
838 			if (inp->inp_socket == NULL ||
839 			    inp->inp_socket->so_cred->cr_prison == NULL) {
840 				return (inp);
841 			}
842 			if  (jinp == NULL)
843 				jinp = inp;
844 		}
845 	}
846 	if (jinp != NULL)
847 		return(jinp);
848 
849 	if (wildcard) {
850 		struct inpcb *local_wild = NULL;
851 		struct inpcb *jinp_wild = NULL;
852 		struct inpcontainer *ic;
853 		struct inpcontainerhead *chead;
854 		struct sockaddr_in6 jsin6;
855 		struct ucred *cred;
856 		struct prison *pr;
857 		int net_listen_ov_local = 0;
858 		int net_listen_ov_wild = 0;
859 
860 		/*
861 		 * Order of socket selection:
862 		 * 1. non-jailed, non-wild.
863 		 * 2. non-jailed, wild.		(allow_listen_override on)
864 		 * 3. jailed, non-wild.
865 		 * 4. jailed, wild.
866 		 * 5. non-jailed, wild.		(allow_listen_override off)
867 		 *
868 		 * NOTE: jailed wildcards are still restricted to the jail
869 		 *	 IPs.
870 		 *
871 		 * NOTE: (1) and (3) already handled above.
872 		 */
873 		jsin6.sin6_family = AF_INET6;
874 		chead = &pcbinfo->wildcardhashbase[INP_PCBWILDCARDHASH(lport,
875 		    pcbinfo->wildcardhashmask)];
876 
877 		GET_PCBINFO_TOKEN(pcbinfo);
878 		LIST_FOREACH(ic, chead, ic_list) {
879 			inp = ic->ic_inp;
880 			if (inp->inp_flags & INP_PLACEMARKER)
881 				continue;
882 
883 			/*
884 			 * Basdic validation
885 			 */
886 			if (!INP_ISIPV6(inp))
887 				continue;
888 			if (inp->inp_lport != lport)
889 				continue;
890 
891 			/*
892 			 * Calculate prison, setup jsin for jailed_ip()
893 			 * check.
894 			 */
895 			jsin6.sin6_addr = *laddr;
896 			pr = NULL;
897 			cred = NULL;
898 			if (inp->inp_socket) {
899 				cred = inp->inp_socket->so_cred;
900 				if (cred)
901 					pr = cred->cr_prison;
902 			}
903 
904 			/*
905 			 * Assign jinp, jinp_wild, and local_wild as
906 			 * appropriate, track whether the jail supports
907 			 * listen overrides.
908 			 */
909 			if (pr) {
910 				if (!jailed_ip(pr, (struct sockaddr *)&jsin6))
911 					continue;
912 				if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)
913 				    && jinp == NULL) {
914 					jinp = inp;
915 					if (PRISON_CAP_ISSET(pr->pr_caps, PRISON_CAP_NET_LISTEN_OVERRIDE))
916 						net_listen_ov_local = 1;
917 				}
918 				if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
919 				    jinp_wild == NULL) {
920 					jinp_wild = inp;
921 					if (PRISON_CAP_ISSET(pr->pr_caps, PRISON_CAP_NET_LISTEN_OVERRIDE))
922 						net_listen_ov_wild = 1;
923 				}
924 			} else {
925 				if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) {
926 					REL_PCBINFO_TOKEN(pcbinfo);
927 					return (inp);
928 				}
929 				if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr))
930 					local_wild = inp;
931 			}
932 		}
933 		REL_PCBINFO_TOKEN(pcbinfo);
934 
935 		if (net_listen_ov_local)
936 			return jinp;
937 		if (net_listen_ov_wild)
938 			return jinp_wild;
939 		if (local_wild)
940 			return local_wild;
941 		if (jinp)
942 			return jinp;
943 		return (jinp_wild);
944 	}
945 
946 	/*
947 	 * Not found.
948 	 */
949 	return (NULL);
950 }
951 
952 void
953 init_sin6(struct sockaddr_in6 *sin6, struct mbuf *m)
954 {
955 	struct ip6_hdr *ip;
956 
957 	ip = mtod(m, struct ip6_hdr *);
958 	bzero(sin6, sizeof(*sin6));
959 	sin6->sin6_len = sizeof(*sin6);
960 	sin6->sin6_family = AF_INET6;
961 	sin6->sin6_addr = ip->ip6_src;
962 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
963 		sin6->sin6_addr.s6_addr16[1] = 0;
964 	sin6->sin6_scope_id =
965 		(m->m_pkthdr.rcvif && IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
966 		? m->m_pkthdr.rcvif->if_index : 0;
967 
968 	return;
969 }
970 
971 void
972 in6_savefaddr(struct socket *so, const struct sockaddr *faddr)
973 {
974 	struct sockaddr_in6 *sin6;
975 
976 	KASSERT(faddr->sa_family == AF_INET6,
977 	    ("not AF_INET6 faddr %d", faddr->sa_family));
978 
979 	sin6 = kmalloc(sizeof(*sin6), M_SONAME, M_WAITOK | M_ZERO);
980 	sin6->sin6_family = AF_INET6;
981 	sin6->sin6_len = sizeof(*sin6);
982 
983 	sin6->sin6_port = ((const struct sockaddr_in6 *)faddr)->sin6_port;
984 	sin6->sin6_addr = ((const struct sockaddr_in6 *)faddr)->sin6_addr;
985 
986 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
987 		sin6->sin6_scope_id = ntohs(sin6->sin6_addr.s6_addr16[1]);
988 	else
989 		sin6->sin6_scope_id = 0;	/*XXX*/
990 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
991 		sin6->sin6_addr.s6_addr16[1] = 0;
992 
993 	so->so_faddr = (struct sockaddr *)sin6;
994 }
995