xref: /dragonfly/sys/netinet6/in6_pcb.c (revision 7b1120e5)
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 (!in6_ifaddr) /* XXX broken! */
112 		return (EADDRNOTAVAIL);
113 	if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
114 		return (EINVAL);
115 
116 	if (nam) {
117 		struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
118 		struct inpcbinfo *pcbinfo;
119 		struct inpcbportinfo *portinfo;
120 		struct inpcbporthead *porthash;
121 		int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
122 		struct ucred *cred = NULL;
123 		struct inpcb *t;
124 		u_short	lport, lport_ho;
125 
126 		if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0)
127 			wild = 1;
128 		if (td->td_proc != NULL)
129 			cred = td->td_proc->p_ucred;
130 
131 		if (nam->sa_len != sizeof(*sin6))
132 			return (EINVAL);
133 		/*
134 		 * family check.
135 		 */
136 		if (nam->sa_family != AF_INET6)
137 			return (EAFNOSUPPORT);
138 
139 		/* Reject v4-mapped address */
140 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr))
141 			return (EADDRNOTAVAIL);
142 
143 		if (!prison_replace_wildcards(td, nam))
144 			return (EINVAL);
145 
146 		/* KAME hack: embed scopeid */
147 		if (in6_embedscope(&sin6->sin6_addr, sin6, inp, NULL) != 0)
148 			return (EINVAL);
149 		/* this must be cleared for ifa_ifwithaddr() */
150 		sin6->sin6_scope_id = 0;
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)
162 				reuseport = SO_REUSEADDR|SO_REUSEPORT;
163 		} else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
164 			struct ifaddr *ia = NULL;
165 
166 			sin6->sin6_port = 0;		/* yech... */
167 			if (!prison_replace_wildcards(td, (struct sockaddr *)sin6)) {
168 				sin6->sin6_addr = kin6addr_any;
169 				return (EINVAL);
170 			}
171 			if ((ia = ifa_ifwithaddr((struct sockaddr *)sin6)) == NULL)
172 				return (EADDRNOTAVAIL);
173 
174 			/*
175 			 * XXX: bind to an anycast address might accidentally
176 			 * cause sending a packet with anycast source address.
177 			 * We should allow to bind to a deprecated address, since
178 			 * the application dares to use it.
179 			 */
180 			if (ia &&
181 			    ((struct in6_ifaddr *)ia)->ia6_flags &
182 			    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|IN6_IFF_DETACHED))
183 				return (EADDRNOTAVAIL);
184 		}
185 
186 		inp->in6p_laddr = sin6->sin6_addr;
187 
188 		if (lport == 0)
189 			goto auto_select;
190 		lport_ho = ntohs(lport);
191 
192 		/* GROSS */
193 		if (lport_ho < IPV6PORT_RESERVED && cred &&
194 		    priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT, 0)) {
195 			inp->in6p_laddr = kin6addr_any;
196 			return (EACCES);
197 		}
198 
199 		/*
200 		 * Locate the proper portinfo based on lport
201 		 */
202 		pcbinfo = inp->inp_pcbinfo;
203 		portinfo =
204 		    &pcbinfo->portinfo[lport_ho % pcbinfo->portinfo_cnt];
205 		KKASSERT((lport_ho % pcbinfo->portinfo_cnt) ==
206 		    portinfo->offset);
207 
208 		/*
209 		 * This has to be atomic.  If the porthash is shared across
210 		 * multiple protocol threads (aka tcp) then the token must
211 		 * be held.
212 		 */
213 		porthash = in_pcbporthash_head(portinfo, lport);
214 		GET_PORTHASH_TOKEN(porthash);
215 
216 		if (so->so_cred->cr_uid != 0 &&
217 		    !IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
218 			t = in6_pcblookup_local(porthash,
219 			    &sin6->sin6_addr, lport, INPLOOKUP_WILDCARD, cred);
220 			if (t &&
221 			    (so->so_cred->cr_uid !=
222 			     t->inp_socket->so_cred->cr_uid)) {
223 				inp->in6p_laddr = kin6addr_any;
224 				error = EADDRINUSE;
225 				goto done;
226 			}
227 		}
228 		if (cred && cred->cr_prison &&
229 		    !prison_replace_wildcards(td, nam)) {
230 			inp->in6p_laddr = kin6addr_any;
231 			error = EADDRNOTAVAIL;
232 			goto done;
233 		}
234 		t = in6_pcblookup_local(porthash, &sin6->sin6_addr, lport,
235 		    wild, cred);
236 		if (t && (reuseport & t->inp_socket->so_options) == 0) {
237 			inp->in6p_laddr = kin6addr_any;
238 			error = EADDRINUSE;
239 			goto done;
240 		}
241 
242 		inp->inp_lport = lport;
243 		in_pcbinsporthash(porthash, inp);
244 		error = 0;
245 done:
246 		REL_PORTHASH_TOKEN(porthash);
247 		return (error);
248 	} else {
249 auto_select:
250 		jsin6.sin6_addr = inp->in6p_laddr;
251 		jsin6.sin6_family = AF_INET6;
252 		if (!prison_replace_wildcards(td, (struct sockaddr*)&jsin6)) {
253 			inp->in6p_laddr = kin6addr_any;
254 			inp->inp_lport = 0;
255 			return (EINVAL);
256 		}
257 
258 		return in6_pcbsetlport(&inp->in6p_laddr, inp, td);
259 	}
260 }
261 
262 /*
263  *   Transform old in6_pcbconnect() into an inner subroutine for new
264  *   in6_pcbconnect(): Do some validity-checking on the remote
265  *   address (in mbuf 'nam') and then determine local host address
266  *   (i.e., which interface) to use to access that remote host.
267  *
268  *   This preserves definition of in6_pcbconnect(), while supporting a
269  *   slightly different version for T/TCP.  (This is more than
270  *   a bit of a kludge, but cleaning up the internal interfaces would
271  *   have forced minor changes in every protocol).
272  */
273 
274 int
275 in6_pcbladdr(struct inpcb *inp, struct sockaddr *nam,
276 	     struct in6_addr **plocal_addr6, struct thread *td)
277 {
278 	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
279 	struct ifnet *ifp = NULL;
280 	int error = 0;
281 
282 	if (nam->sa_len != sizeof (*sin6))
283 		return (EINVAL);
284 	if (sin6->sin6_family != AF_INET6)
285 		return (EAFNOSUPPORT);
286 	if (sin6->sin6_port == 0)
287 		return (EADDRNOTAVAIL);
288 
289 	/* KAME hack: embed scopeid */
290 	if (in6_embedscope(&sin6->sin6_addr, sin6, inp, &ifp) != 0)
291 		return EINVAL;
292 
293 	if (in6_ifaddr) {
294 		/*
295 		 * If the destination address is UNSPECIFIED addr,
296 		 * use the loopback addr, e.g ::1.
297 		 */
298 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
299 			sin6->sin6_addr = kin6addr_loopback;
300 	}
301 	{
302 		/*
303 		 * XXX: in6_selectsrc might replace the bound local address
304 		 * with the address specified by setsockopt(IPV6_PKTINFO).
305 		 * Is it the intended behavior?
306 		 */
307 		*plocal_addr6 = in6_selectsrc(sin6, inp->in6p_outputopts,
308 					      inp->in6p_moptions,
309 					      &inp->in6p_route,
310 					      &inp->in6p_laddr, &error, td);
311 		if (*plocal_addr6 == NULL) {
312 			if (error == 0)
313 				error = EADDRNOTAVAIL;
314 			return (error);
315 		}
316 		/*
317 		 * Don't do pcblookup call here; return interface in
318 		 * plocal_addr6
319 		 * and exit to caller, that will do the lookup.
320 		 */
321 	}
322 
323 	if (inp->in6p_route.ro_rt)
324 		ifp = inp->in6p_route.ro_rt->rt_ifp;
325 
326 	return (0);
327 }
328 
329 /*
330  * Outer subroutine:
331  * Connect from a socket to a specified address.
332  * Both address and port must be specified in argument sin.
333  * If don't have a local address for this socket yet,
334  * then pick one.
335  */
336 int
337 in6_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct thread *td)
338 {
339 	struct in6_addr *addr6;
340 	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
341 	int error;
342 
343 	/* Reject v4-mapped address */
344 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr))
345 		return EADDRNOTAVAIL;
346 
347 	/*
348 	 * Call inner routine, to assign local interface address.
349 	 * in6_pcbladdr() may automatically fill in sin6_scope_id.
350 	 */
351 	if ((error = in6_pcbladdr(inp, nam, &addr6, td)) != 0)
352 		return (error);
353 
354 	if (in6_pcblookup_hash(inp->inp_pcbinfo, &sin6->sin6_addr,
355 			       sin6->sin6_port,
356 			      IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
357 			      ? addr6 : &inp->in6p_laddr,
358 			      inp->inp_lport, 0, NULL) != NULL) {
359 		return (EADDRINUSE);
360 	}
361 	if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
362 		if (inp->inp_lport == 0) {
363 			error = in6_pcbbind(inp, NULL, td);
364 			if (error)
365 				return (error);
366 		}
367 		inp->in6p_laddr = *addr6;
368 	}
369 	inp->in6p_faddr = sin6->sin6_addr;
370 	inp->inp_fport = sin6->sin6_port;
371 	/* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
372 	inp->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK;
373 	if (inp->in6p_flags & IN6P_AUTOFLOWLABEL)
374 		inp->in6p_flowinfo |=
375 		    (htonl(ip6_flow_seq++) & IPV6_FLOWLABEL_MASK);
376 
377 	in_pcbinsconnhash(inp);
378 	return (0);
379 }
380 
381 void
382 in6_pcbdisconnect(struct inpcb *inp)
383 {
384 	bzero((caddr_t)&inp->in6p_faddr, sizeof(inp->in6p_faddr));
385 	inp->inp_fport = 0;
386 	/* clear flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
387 	inp->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK;
388 	in_pcbremconnhash(inp);
389 	if (inp->inp_socket->so_state & SS_NOFDREF)
390 		in6_pcbdetach(inp);
391 }
392 
393 void
394 in6_pcbdetach(struct inpcb *inp)
395 {
396 	struct socket *so = inp->inp_socket;
397 	struct inpcbinfo *ipi = inp->inp_pcbinfo;
398 
399 	inp->inp_gencnt = ++ipi->ipi_gencnt;
400 	in_pcbremlists(inp);
401 	so->so_pcb = NULL;
402 	KKASSERT((so->so_state & SS_ASSERTINPROG) == 0);
403 	sofree(so);		/* remove pcb ref */
404 
405 	if (inp->in6p_options)
406 		m_freem(inp->in6p_options);
407 	ip6_freepcbopts(inp->in6p_outputopts);
408 	ip6_freemoptions(inp->in6p_moptions);
409 	if (inp->in6p_route.ro_rt)
410 		rtfree(inp->in6p_route.ro_rt);
411 	/* Check and free IPv4 related resources in case of mapped addr */
412 	if (inp->inp_options)
413 		m_free(inp->inp_options);
414 	ip_freemoptions(inp->inp_moptions);
415 
416 	kfree(inp, M_PCB);
417 }
418 
419 /*
420  * The socket may have an invalid PCB, i.e. NULL.  For example, a TCP
421  * socket received RST.
422  */
423 static int
424 in6_setsockaddr(struct socket *so, struct sockaddr **nam)
425 {
426 	struct inpcb *inp;
427 	struct sockaddr_in6 *sin6;
428 
429 	KASSERT(curthread->td_type == TD_TYPE_NETISR, ("not in netisr"));
430 	inp = so->so_pcb;
431 	if (!inp)
432 		return EINVAL;
433 
434 	sin6 = kmalloc(sizeof *sin6, M_SONAME, M_WAITOK | M_ZERO);
435 	sin6->sin6_family = AF_INET6;
436 	sin6->sin6_len = sizeof(*sin6);
437 	sin6->sin6_port = inp->inp_lport;
438 	sin6->sin6_addr = inp->in6p_laddr;
439 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
440 		sin6->sin6_scope_id = ntohs(sin6->sin6_addr.s6_addr16[1]);
441 	else
442 		sin6->sin6_scope_id = 0;	/*XXX*/
443 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
444 		sin6->sin6_addr.s6_addr16[1] = 0;
445 
446 	*nam = (struct sockaddr *)sin6;
447 	return 0;
448 }
449 
450 void
451 in6_setsockaddr_dispatch(netmsg_t msg)
452 {
453 	int error;
454 
455 	error = in6_setsockaddr(msg->sockaddr.base.nm_so, msg->sockaddr.nm_nam);
456 	lwkt_replymsg(&msg->sockaddr.base.lmsg, error);
457 }
458 
459 void
460 in6_setpeeraddr_dispatch(netmsg_t msg)
461 {
462 	int error;
463 
464 	error = in6_setpeeraddr(msg->peeraddr.base.nm_so, msg->peeraddr.nm_nam);
465 	lwkt_replymsg(&msg->peeraddr.base.lmsg, error);
466 }
467 
468 /*
469  * The socket may have an invalid PCB, i.e. NULL.  For example, a TCP
470  * socket received RST.
471  */
472 int
473 in6_setpeeraddr(struct socket *so, struct sockaddr **nam)
474 {
475 	struct inpcb *inp;
476 	struct sockaddr_in6 *sin6;
477 
478 	KASSERT(curthread->td_type == TD_TYPE_NETISR, ("not in netisr"));
479 	inp = so->so_pcb;
480 	if (!inp)
481 		return EINVAL;
482 
483 	sin6 = kmalloc(sizeof(*sin6), M_SONAME, M_WAITOK | M_ZERO);
484 	sin6->sin6_family = AF_INET6;
485 	sin6->sin6_len = sizeof(struct sockaddr_in6);
486 	sin6->sin6_port = inp->inp_fport;
487 	sin6->sin6_addr = inp->in6p_faddr;
488 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
489 		sin6->sin6_scope_id = ntohs(sin6->sin6_addr.s6_addr16[1]);
490 	else
491 		sin6->sin6_scope_id = 0;	/*XXX*/
492 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
493 		sin6->sin6_addr.s6_addr16[1] = 0;
494 
495 	*nam = (struct sockaddr *)sin6;
496 	return 0;
497 }
498 
499 /*
500  * Pass some notification to all connections of a protocol
501  * associated with address dst.  The local address and/or port numbers
502  * may be specified to limit the search.  The "usual action" will be
503  * taken, depending on the ctlinput cmd.  The caller must filter any
504  * cmds that are uninteresting (e.g., no error in the map).
505  * Call the protocol specific routine (if any) to report
506  * any errors for each matching socket.
507  */
508 void
509 in6_pcbnotify(struct inpcbinfo *pcbinfo, struct sockaddr *dst, in_port_t fport,
510     const struct sockaddr *src, in_port_t lport, int cmd, int arg,
511     inp_notify_t notify)
512 {
513 	struct inpcb *inp, *marker;
514 	struct sockaddr_in6 sa6_src, *sa6_dst;
515 	u_int32_t flowinfo;
516 
517 	if ((unsigned)cmd >= PRC_NCMDS || dst->sa_family != AF_INET6)
518 		return;
519 
520 	sa6_dst = (struct sockaddr_in6 *)dst;
521 	if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr))
522 		return;
523 
524 	/*
525 	 * note that src can be NULL when we get notify by local fragmentation.
526 	 */
527 	sa6_src = (src == NULL) ? sa6_any : *(const struct sockaddr_in6 *)src;
528 	flowinfo = sa6_src.sin6_flowinfo;
529 
530 	/*
531 	 * Redirects go to all references to the destination,
532 	 * and use in6_rtchange to invalidate the route cache.
533 	 * Dead host indications: also use in6_rtchange to invalidate
534 	 * the cache, and deliver the error to all the sockets.
535 	 * Otherwise, if we have knowledge of the local port and address,
536 	 * deliver only to that socket.
537 	 */
538 	if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
539 		fport = 0;
540 		lport = 0;
541 		bzero((caddr_t)&sa6_src.sin6_addr, sizeof(sa6_src.sin6_addr));
542 
543 		if (cmd != PRC_HOSTDEAD)
544 			notify = in6_rtchange;
545 	}
546 	if (cmd != PRC_MSGSIZE)
547 		arg = inet6ctlerrmap[cmd];
548 
549 	marker = in_pcbmarker();
550 
551 	GET_PCBINFO_TOKEN(pcbinfo);
552 
553 	LIST_INSERT_HEAD(&pcbinfo->pcblisthead, marker, inp_list);
554 	while ((inp = LIST_NEXT(marker, inp_list)) != NULL) {
555 		LIST_REMOVE(marker, inp_list);
556 		LIST_INSERT_AFTER(inp, marker, inp_list);
557 
558 		if (inp->inp_flags & INP_PLACEMARKER)
559 			continue;
560 
561 		if (!INP_ISIPV6(inp))
562 			continue;
563 		/*
564 		 * If the error designates a new path MTU for a destination
565 		 * and the application (associated with this socket) wanted to
566 		 * know the value, notify. Note that we notify for all
567 		 * disconnected sockets if the corresponding application
568 		 * wanted. This is because some UDP applications keep sending
569 		 * sockets disconnected.
570 		 * XXX: should we avoid to notify the value to TCP sockets?
571 		 */
572 		if (cmd == PRC_MSGSIZE && (inp->inp_flags & IN6P_MTU) != 0 &&
573 		    (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
574 		     IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &sa6_dst->sin6_addr))) {
575 			ip6_notify_pmtu(inp, (struct sockaddr_in6 *)dst, &arg);
576 		}
577 
578 		/*
579 		 * Detect if we should notify the error. If no source and
580 		 * destination ports are specifed, but non-zero flowinfo and
581 		 * local address match, notify the error. This is the case
582 		 * when the error is delivered with an encrypted buffer
583 		 * by ESP. Otherwise, just compare addresses and ports
584 		 * as usual.
585 		 */
586 		if (lport == 0 && fport == 0 && flowinfo &&
587 		    inp->inp_socket != NULL &&
588 		    flowinfo == (inp->in6p_flowinfo & IPV6_FLOWLABEL_MASK) &&
589 		    IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &sa6_src.sin6_addr))
590 			goto do_notify;
591 		else if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
592 					     &sa6_dst->sin6_addr) ||
593 			 inp->inp_socket == 0 ||
594 			 (lport && inp->inp_lport != lport) ||
595 			 (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) &&
596 			  !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
597 					      &sa6_src.sin6_addr)) ||
598 			 (fport && inp->inp_fport != fport))
599 			continue;
600 
601 do_notify:
602 		if (notify)
603 			(*notify)(inp, arg);
604 	}
605 	LIST_REMOVE(marker, inp_list);
606 
607 	REL_PCBINFO_TOKEN(pcbinfo);
608 }
609 
610 /*
611  * Lookup a PCB based on the local address and port.
612  */
613 struct inpcb *
614 in6_pcblookup_local(struct inpcbporthead *porthash,
615     const struct in6_addr *laddr, u_int lport_arg, int wild_okay,
616     struct ucred *cred)
617 {
618 	struct inpcb *inp;
619 	int matchwild = 3, wildcard;
620 	u_short lport = lport_arg;
621 	struct inpcbport *phd;
622 	struct inpcb *match = NULL;
623 
624 	/*
625 	 * If the porthashbase is shared across several cpus, it must
626 	 * have been locked.
627 	 */
628 	ASSERT_PORTHASH_TOKEN_HELD(porthash);
629 
630 	/*
631 	 * Best fit PCB lookup.
632 	 *
633 	 * First see if this local port is in use by looking on the
634 	 * port hash list.
635 	 */
636 	LIST_FOREACH(phd, porthash, phd_hash) {
637 		if (phd->phd_port == lport)
638 			break;
639 	}
640 
641 	if (phd != NULL) {
642 		/*
643 		 * Port is in use by one or more PCBs. Look for best
644 		 * fit.
645 		 */
646 		LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
647 			wildcard = 0;
648 			if (!INP_ISIPV6(inp))
649 				continue;
650 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr))
651 				wildcard++;
652 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
653 				if (IN6_IS_ADDR_UNSPECIFIED(laddr))
654 					wildcard++;
655 				else if (!IN6_ARE_ADDR_EQUAL(
656 					&inp->in6p_laddr, laddr))
657 					continue;
658 			} else {
659 				if (!IN6_IS_ADDR_UNSPECIFIED(laddr))
660 					wildcard++;
661 			}
662 			if (wildcard && !wild_okay)
663 				continue;
664 			if (wildcard < matchwild &&
665 			    (cred == NULL ||
666 			     cred->cr_prison ==
667 					inp->inp_socket->so_cred->cr_prison)) {
668 				match = inp;
669 				matchwild = wildcard;
670 				if (wildcard == 0)
671 					break;
672 				else
673 					matchwild = wildcard;
674 			}
675 		}
676 	}
677 	return (match);
678 }
679 
680 void
681 in6_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
682 {
683 	struct in6pcb *in6p, *marker;
684 	struct ip6_moptions *im6o;
685 	struct in6_multi_mship *imm, *nimm;
686 
687 	/*
688 	 * We only need to make sure that we are in netisr0, where all
689 	 * multicast operation happen.  We could check inpcbinfo which
690 	 * does not belong to netisr0 by holding the inpcbinfo's token.
691 	 * In this case, the pcbinfo must be able to be shared, i.e.
692 	 * pcbinfo->infotoken is not NULL.
693 	 */
694 	ASSERT_NETISR0;
695 	KASSERT(pcbinfo->cpu == 0 || pcbinfo->infotoken != NULL,
696 	    ("pcbinfo could not be shared"));
697 
698 	/*
699 	 * Get a marker for the current netisr (netisr0).
700 	 *
701 	 * It is possible that the multicast address deletion blocks,
702 	 * which could cause temporary token releasing.  So we use
703 	 * inpcb marker here to get a coherent view of the inpcb list.
704 	 *
705 	 * While, on the other hand, moptions are only added and deleted
706 	 * in netisr0, so we would not see staled moption or miss moption
707 	 * even if the token was released due to the blocking multicast
708 	 * address deletion.
709 	 */
710 	marker = in_pcbmarker();
711 
712 	GET_PCBINFO_TOKEN(pcbinfo);
713 
714 	LIST_INSERT_HEAD(&pcbinfo->pcblisthead, marker, inp_list);
715 	while ((in6p = LIST_NEXT(marker, inp_list)) != NULL) {
716 		LIST_REMOVE(marker, inp_list);
717 		LIST_INSERT_AFTER(in6p, marker, inp_list);
718 
719 		if (in6p->in6p_flags & INP_PLACEMARKER)
720 			continue;
721 		im6o = in6p->in6p_moptions;
722 		if (INP_ISIPV6(in6p) && im6o) {
723 			/*
724 			 * Unselect the outgoing interface if it is being
725 			 * detached.
726 			 */
727 			if (im6o->im6o_multicast_ifp == ifp)
728 				im6o->im6o_multicast_ifp = NULL;
729 
730 			/*
731 			 * Drop multicast group membership if we joined
732 			 * through the interface being detached.
733 			 * XXX controversial - is it really legal for kernel
734 			 * to force this?
735 			 */
736 			for (imm = im6o->im6o_memberships.lh_first;
737 			     imm != NULL; imm = nimm) {
738 				nimm = imm->i6mm_chain.le_next;
739 				if (imm->i6mm_maddr->in6m_ifp == ifp) {
740 					LIST_REMOVE(imm, i6mm_chain);
741 					in6_delmulti(imm->i6mm_maddr);
742 					kfree(imm, M_IPMADDR);
743 				}
744 			}
745 		}
746 	}
747 	LIST_REMOVE(marker, inp_list);
748 
749 	REL_PCBINFO_TOKEN(pcbinfo);
750 }
751 
752 /*
753  * Check for alternatives when higher level complains
754  * about service problems.  For now, invalidate cached
755  * routing information.  If the route was created dynamically
756  * (by a redirect), time to try a default gateway again.
757  */
758 void
759 in6_losing(struct inpcb *in6p)
760 {
761 	struct rtentry *rt;
762 	struct rt_addrinfo info;
763 
764 	if ((rt = in6p->in6p_route.ro_rt) != NULL) {
765 		bzero((caddr_t)&info, sizeof(info));
766 		info.rti_flags = rt->rt_flags;
767 		info.rti_info[RTAX_DST] = rt_key(rt);
768 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
769 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
770 		rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
771 		if (rt->rt_flags & RTF_DYNAMIC) {
772 			rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
773 			    rt_mask(rt), rt->rt_flags, NULL);
774 		}
775 		in6p->in6p_route.ro_rt = NULL;
776 		rtfree(rt);
777 		/*
778 		 * A new route can be allocated
779 		 * the next time output is attempted.
780 		 */
781 	}
782 }
783 
784 /*
785  * After a routing change, flush old routing
786  * and allocate a (hopefully) better one.
787  */
788 void
789 in6_rtchange(struct inpcb *inp, int error)
790 {
791 	if (inp->in6p_route.ro_rt) {
792 		rtfree(inp->in6p_route.ro_rt);
793 		inp->in6p_route.ro_rt = 0;
794 		/*
795 		 * A new route can be allocated the next time
796 		 * output is attempted.
797 		 */
798 	}
799 }
800 
801 /*
802  * Lookup PCB in hash list.
803  */
804 struct inpcb *
805 in6_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
806 		   u_int fport_arg, struct in6_addr *laddr, u_int lport_arg,
807 		   int wildcard, struct ifnet *ifp)
808 {
809 	struct inpcbhead *head;
810 	struct inpcb *inp;
811 	struct inpcb *jinp = NULL;
812 	u_short fport = fport_arg, lport = lport_arg;
813 
814 	/*
815 	 * First look for an exact match.
816 	 */
817 	head = &pcbinfo->hashbase[INP_PCBCONNHASH(faddr->s6_addr32[3] /* XXX */,
818 					      fport,
819 					      laddr->s6_addr32[3], /* XXX JH */
820 					      lport,
821 					      pcbinfo->hashmask)];
822 	LIST_FOREACH(inp, head, inp_hash) {
823 		if (!INP_ISIPV6(inp))
824 			continue;
825 		if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) &&
826 		    IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
827 		    inp->inp_fport == fport &&
828 		    inp->inp_lport == lport) {
829 			/*
830 			 * Found.
831 			 */
832 			if (inp->inp_socket == NULL ||
833 			inp->inp_socket->so_cred->cr_prison == NULL) {
834 				return (inp);
835 			} else {
836 				if  (jinp == NULL)
837 					jinp = inp;
838 			}
839 		}
840 	}
841 	if (jinp != NULL)
842 		return(jinp);
843 
844 	if (wildcard) {
845 		struct inpcontainerhead *chead;
846 		struct inpcontainer *ic;
847 		struct inpcb *local_wild = NULL;
848 		struct inpcb *jinp_wild = NULL;
849 		struct sockaddr_in6 jsin6;
850 		struct ucred *cred;
851 
852 		/*
853 		 * Order of socket selection:
854 		 * 1. non-jailed, non-wild.
855 		 * 2. non-jailed, wild.
856 		 * 3. jailed, non-wild.
857 		 * 4. jailed, wild.
858 		 */
859 		jsin6.sin6_family = AF_INET6;
860 		chead = &pcbinfo->wildcardhashbase[INP_PCBWILDCARDHASH(lport,
861 		    pcbinfo->wildcardhashmask)];
862 
863 		GET_PCBINFO_TOKEN(pcbinfo);
864 		LIST_FOREACH(ic, chead, ic_list) {
865 			inp = ic->ic_inp;
866 			if (inp->inp_flags & INP_PLACEMARKER)
867 				continue;
868 
869 			if (!INP_ISIPV6(inp))
870 				continue;
871 			if (inp->inp_socket != NULL)
872 				cred = inp->inp_socket->so_cred;
873 			else
874 				cred = NULL;
875 
876 			if (cred != NULL && jailed(cred)) {
877 				if (jinp != NULL) {
878 					continue;
879 				} else {
880 		                        jsin6.sin6_addr = *laddr;
881 					if (!jailed_ip(cred->cr_prison,
882 					    (struct sockaddr *)&jsin6))
883 						continue;
884 				}
885 			}
886 			if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
887 			    inp->inp_lport == lport) {
888 				if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
889 						       laddr)) {
890 					if (cred != NULL && jailed(cred)) {
891 						jinp = inp;
892 					} else {
893 						REL_PCBINFO_TOKEN(pcbinfo);
894 						return (inp);
895 					}
896 				} else if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
897 					if (cred != NULL && jailed(cred))
898 						jinp_wild = inp;
899 					else
900 						local_wild = inp;
901 				}
902 			}
903 		}
904 		REL_PCBINFO_TOKEN(pcbinfo);
905 
906 		if (local_wild != NULL)
907 			return (local_wild);
908 		if (jinp != NULL)
909 			return (jinp);
910 		return (jinp_wild);
911 	}
912 
913 	/*
914 	 * Not found.
915 	 */
916 	return (NULL);
917 }
918 
919 void
920 init_sin6(struct sockaddr_in6 *sin6, struct mbuf *m)
921 {
922 	struct ip6_hdr *ip;
923 
924 	ip = mtod(m, struct ip6_hdr *);
925 	bzero(sin6, sizeof(*sin6));
926 	sin6->sin6_len = sizeof(*sin6);
927 	sin6->sin6_family = AF_INET6;
928 	sin6->sin6_addr = ip->ip6_src;
929 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
930 		sin6->sin6_addr.s6_addr16[1] = 0;
931 	sin6->sin6_scope_id =
932 		(m->m_pkthdr.rcvif && IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
933 		? m->m_pkthdr.rcvif->if_index : 0;
934 
935 	return;
936 }
937 
938 void
939 in6_savefaddr(struct socket *so, const struct sockaddr *faddr)
940 {
941 	struct sockaddr_in6 *sin6;
942 
943 	KASSERT(faddr->sa_family == AF_INET6,
944 	    ("not AF_INET6 faddr %d", faddr->sa_family));
945 
946 	sin6 = kmalloc(sizeof(*sin6), M_SONAME, M_WAITOK | M_ZERO);
947 	sin6->sin6_family = AF_INET6;
948 	sin6->sin6_len = sizeof(*sin6);
949 
950 	sin6->sin6_port = ((const struct sockaddr_in6 *)faddr)->sin6_port;
951 	sin6->sin6_addr = ((const struct sockaddr_in6 *)faddr)->sin6_addr;
952 
953 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
954 		sin6->sin6_scope_id = ntohs(sin6->sin6_addr.s6_addr16[1]);
955 	else
956 		sin6->sin6_scope_id = 0;	/*XXX*/
957 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
958 		sin6->sin6_addr.s6_addr16[1] = 0;
959 
960 	so->so_faddr = (struct sockaddr *)sin6;
961 }
962