xref: /dragonfly/sys/netinet6/in6_pcb.c (revision c37c9ab3)
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, &sin6->sin6_addr,
219 						lport, INPLOOKUP_WILDCARD,
220 						cred);
221 			if (t &&
222 			    (so->so_cred->cr_uid !=
223 			     t->inp_socket->so_cred->cr_uid)) {
224 				inp->in6p_laddr = kin6addr_any;
225 				error = EADDRINUSE;
226 				goto done;
227 			}
228 		}
229 		if (cred && cred->cr_prison &&
230 		    !prison_replace_wildcards(td, nam)) {
231 			inp->in6p_laddr = kin6addr_any;
232 			error = EADDRNOTAVAIL;
233 			goto done;
234 		}
235 		t = in6_pcblookup_local(porthash, &sin6->sin6_addr, lport,
236 					wild, cred);
237 		if (t && (reuseport & t->inp_socket->so_options) == 0) {
238 			inp->in6p_laddr = kin6addr_any;
239 			error = EADDRINUSE;
240 			goto done;
241 		}
242 
243 		inp->inp_lport = lport;
244 		in_pcbinsporthash(porthash, inp);
245 		error = 0;
246 done:
247 		REL_PORTHASH_TOKEN(porthash);
248 		return (error);
249 	} else {
250 auto_select:
251 		jsin6.sin6_addr = inp->in6p_laddr;
252 		jsin6.sin6_family = AF_INET6;
253 		if (!prison_replace_wildcards(td, (struct sockaddr*)&jsin6)) {
254 			inp->in6p_laddr = kin6addr_any;
255 			inp->inp_lport = 0;
256 			return (EINVAL);
257 		}
258 
259 		return in6_pcbsetlport(&inp->in6p_laddr, inp, td);
260 	}
261 }
262 
263 /*
264  *   Transform old in6_pcbconnect() into an inner subroutine for new
265  *   in6_pcbconnect(): Do some validity-checking on the remote
266  *   address (in mbuf 'nam') and then determine local host address
267  *   (i.e., which interface) to use to access that remote host.
268  *
269  *   This preserves definition of in6_pcbconnect(), while supporting a
270  *   slightly different version for T/TCP.  (This is more than
271  *   a bit of a kludge, but cleaning up the internal interfaces would
272  *   have forced minor changes in every protocol).
273  */
274 
275 int
276 in6_pcbladdr(struct inpcb *inp, struct sockaddr *nam,
277 	     struct in6_addr **plocal_addr6, struct thread *td)
278 {
279 	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
280 	struct ifnet *ifp = NULL;
281 	int error = 0;
282 
283 	if (nam->sa_len != sizeof (*sin6))
284 		return (EINVAL);
285 	if (sin6->sin6_family != AF_INET6)
286 		return (EAFNOSUPPORT);
287 	if (sin6->sin6_port == 0)
288 		return (EADDRNOTAVAIL);
289 
290 	/* KAME hack: embed scopeid */
291 	if (in6_embedscope(&sin6->sin6_addr, sin6, inp, &ifp) != 0)
292 		return EINVAL;
293 
294 	if (in6_ifaddr) {
295 		/*
296 		 * If the destination address is UNSPECIFIED addr,
297 		 * use the loopback addr, e.g ::1.
298 		 */
299 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr))
300 			sin6->sin6_addr = kin6addr_loopback;
301 	}
302 	{
303 		/*
304 		 * XXX: in6_selectsrc might replace the bound local address
305 		 * with the address specified by setsockopt(IPV6_PKTINFO).
306 		 * Is it the intended behavior?
307 		 */
308 		*plocal_addr6 = in6_selectsrc(sin6, inp->in6p_outputopts,
309 					      inp->in6p_moptions,
310 					      &inp->in6p_route,
311 					      &inp->in6p_laddr, &error, td);
312 		if (*plocal_addr6 == NULL) {
313 			if (error == 0)
314 				error = EADDRNOTAVAIL;
315 			return (error);
316 		}
317 		/*
318 		 * Don't do pcblookup call here; return interface in
319 		 * plocal_addr6
320 		 * and exit to caller, that will do the lookup.
321 		 */
322 	}
323 
324 	if (inp->in6p_route.ro_rt)
325 		ifp = inp->in6p_route.ro_rt->rt_ifp;
326 
327 	return (0);
328 }
329 
330 /*
331  * Outer subroutine:
332  * Connect from a socket to a specified address.
333  * Both address and port must be specified in argument sin.
334  * If don't have a local address for this socket yet,
335  * then pick one.
336  */
337 int
338 in6_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct thread *td)
339 {
340 	struct in6_addr *addr6;
341 	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)nam;
342 	int error;
343 
344 	/* Reject v4-mapped address */
345 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr))
346 		return EADDRNOTAVAIL;
347 
348 	/*
349 	 * Call inner routine, to assign local interface address.
350 	 * in6_pcbladdr() may automatically fill in sin6_scope_id.
351 	 */
352 	if ((error = in6_pcbladdr(inp, nam, &addr6, td)) != 0)
353 		return (error);
354 
355 	if (in6_pcblookup_hash(inp->inp_pcbinfo, &sin6->sin6_addr,
356 			       sin6->sin6_port,
357 			      IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)
358 			      ? addr6 : &inp->in6p_laddr,
359 			      inp->inp_lport, 0, NULL) != NULL) {
360 		return (EADDRINUSE);
361 	}
362 	if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
363 		if (inp->inp_lport == 0) {
364 			error = in6_pcbbind(inp, NULL, td);
365 			if (error)
366 				return (error);
367 		}
368 		inp->in6p_laddr = *addr6;
369 	}
370 	inp->in6p_faddr = sin6->sin6_addr;
371 	inp->inp_fport = sin6->sin6_port;
372 	/* update flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
373 	inp->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK;
374 	if (inp->in6p_flags & IN6P_AUTOFLOWLABEL)
375 		inp->in6p_flowinfo |=
376 		    (htonl(ip6_flow_seq++) & IPV6_FLOWLABEL_MASK);
377 
378 	in_pcbinsconnhash(inp);
379 	return (0);
380 }
381 
382 void
383 in6_pcbdisconnect(struct inpcb *inp)
384 {
385 	bzero((caddr_t)&inp->in6p_faddr, sizeof(inp->in6p_faddr));
386 	inp->inp_fport = 0;
387 	/* clear flowinfo - draft-itojun-ipv6-flowlabel-api-00 */
388 	inp->in6p_flowinfo &= ~IPV6_FLOWLABEL_MASK;
389 	in_pcbremconnhash(inp);
390 	if (inp->inp_socket->so_state & SS_NOFDREF)
391 		in6_pcbdetach(inp);
392 }
393 
394 void
395 in6_pcbdetach(struct inpcb *inp)
396 {
397 	struct socket *so = inp->inp_socket;
398 	struct inpcbinfo *ipi = inp->inp_pcbinfo;
399 
400 	inp->inp_gencnt = ++ipi->ipi_gencnt;
401 	in_pcbremlists(inp);
402 	so->so_pcb = NULL;
403 	KKASSERT((so->so_state & SS_ASSERTINPROG) == 0);
404 	sofree(so);		/* remove pcb ref */
405 
406 	if (inp->in6p_options)
407 		m_freem(inp->in6p_options);
408 	ip6_freepcbopts(inp->in6p_outputopts);
409 	ip6_freemoptions(inp->in6p_moptions);
410 	if (inp->in6p_route.ro_rt)
411 		rtfree(inp->in6p_route.ro_rt);
412 	/* Check and free IPv4 related resources in case of mapped addr */
413 	if (inp->inp_options)
414 		m_free(inp->inp_options);
415 	ip_freemoptions(inp->inp_moptions);
416 
417 	kfree(inp, M_PCB);
418 }
419 
420 /*
421  * The socket may have an invalid PCB, i.e. NULL.  For example, a TCP
422  * socket received RST.
423  */
424 static int
425 in6_setsockaddr(struct socket *so, struct sockaddr **nam)
426 {
427 	struct inpcb *inp;
428 	struct sockaddr_in6 *sin6;
429 
430 	KASSERT(curthread->td_type == TD_TYPE_NETISR, ("not in netisr"));
431 	inp = so->so_pcb;
432 	if (!inp)
433 		return EINVAL;
434 
435 	sin6 = kmalloc(sizeof *sin6, M_SONAME, M_WAITOK | M_ZERO);
436 	sin6->sin6_family = AF_INET6;
437 	sin6->sin6_len = sizeof(*sin6);
438 	sin6->sin6_port = inp->inp_lport;
439 	sin6->sin6_addr = inp->in6p_laddr;
440 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
441 		sin6->sin6_scope_id = ntohs(sin6->sin6_addr.s6_addr16[1]);
442 	else
443 		sin6->sin6_scope_id = 0;	/*XXX*/
444 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
445 		sin6->sin6_addr.s6_addr16[1] = 0;
446 
447 	*nam = (struct sockaddr *)sin6;
448 	return 0;
449 }
450 
451 void
452 in6_setsockaddr_dispatch(netmsg_t msg)
453 {
454 	int error;
455 
456 	error = in6_setsockaddr(msg->sockaddr.base.nm_so, msg->sockaddr.nm_nam);
457 	lwkt_replymsg(&msg->sockaddr.base.lmsg, error);
458 }
459 
460 void
461 in6_setpeeraddr_dispatch(netmsg_t msg)
462 {
463 	int error;
464 
465 	error = in6_setpeeraddr(msg->peeraddr.base.nm_so, msg->peeraddr.nm_nam);
466 	lwkt_replymsg(&msg->peeraddr.base.lmsg, error);
467 }
468 
469 /*
470  * The socket may have an invalid PCB, i.e. NULL.  For example, a TCP
471  * socket received RST.
472  */
473 int
474 in6_setpeeraddr(struct socket *so, struct sockaddr **nam)
475 {
476 	struct inpcb *inp;
477 	struct sockaddr_in6 *sin6;
478 
479 	KASSERT(curthread->td_type == TD_TYPE_NETISR, ("not in netisr"));
480 	inp = so->so_pcb;
481 	if (!inp)
482 		return EINVAL;
483 
484 	sin6 = kmalloc(sizeof(*sin6), M_SONAME, M_WAITOK | M_ZERO);
485 	sin6->sin6_family = AF_INET6;
486 	sin6->sin6_len = sizeof(struct sockaddr_in6);
487 	sin6->sin6_port = inp->inp_fport;
488 	sin6->sin6_addr = inp->in6p_faddr;
489 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
490 		sin6->sin6_scope_id = ntohs(sin6->sin6_addr.s6_addr16[1]);
491 	else
492 		sin6->sin6_scope_id = 0;	/*XXX*/
493 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
494 		sin6->sin6_addr.s6_addr16[1] = 0;
495 
496 	*nam = (struct sockaddr *)sin6;
497 	return 0;
498 }
499 
500 /*
501  * Pass some notification to all connections of a protocol
502  * associated with address dst.  The local address and/or port numbers
503  * may be specified to limit the search.  The "usual action" will be
504  * taken, depending on the ctlinput cmd.  The caller must filter any
505  * cmds that are uninteresting (e.g., no error in the map).
506  * Call the protocol specific routine (if any) to report
507  * any errors for each matching socket.
508  */
509 void
510 in6_pcbnotify(struct inpcbinfo *pcbinfo, struct sockaddr *dst, in_port_t fport,
511     const struct sockaddr *src, in_port_t lport, int cmd, int arg,
512     inp_notify_t notify)
513 {
514 	struct inpcb *inp, *marker;
515 	struct sockaddr_in6 sa6_src, *sa6_dst;
516 	u_int32_t flowinfo;
517 
518 	if ((unsigned)cmd >= PRC_NCMDS || dst->sa_family != AF_INET6)
519 		return;
520 
521 	sa6_dst = (struct sockaddr_in6 *)dst;
522 	if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr))
523 		return;
524 
525 	/*
526 	 * note that src can be NULL when we get notify by local fragmentation.
527 	 */
528 	sa6_src = (src == NULL) ? sa6_any : *(const struct sockaddr_in6 *)src;
529 	flowinfo = sa6_src.sin6_flowinfo;
530 
531 	/*
532 	 * Redirects go to all references to the destination,
533 	 * and use in6_rtchange to invalidate the route cache.
534 	 * Dead host indications: also use in6_rtchange to invalidate
535 	 * the cache, and deliver the error to all the sockets.
536 	 * Otherwise, if we have knowledge of the local port and address,
537 	 * deliver only to that socket.
538 	 */
539 	if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
540 		fport = 0;
541 		lport = 0;
542 		bzero((caddr_t)&sa6_src.sin6_addr, sizeof(sa6_src.sin6_addr));
543 
544 		if (cmd != PRC_HOSTDEAD)
545 			notify = in6_rtchange;
546 	}
547 	if (cmd != PRC_MSGSIZE)
548 		arg = inet6ctlerrmap[cmd];
549 
550 	marker = in_pcbmarker();
551 
552 	GET_PCBINFO_TOKEN(pcbinfo);
553 
554 	LIST_INSERT_HEAD(&pcbinfo->pcblisthead, marker, inp_list);
555 	while ((inp = LIST_NEXT(marker, inp_list)) != NULL) {
556 		LIST_REMOVE(marker, inp_list);
557 		LIST_INSERT_AFTER(inp, marker, inp_list);
558 
559 		if (inp->inp_flags & INP_PLACEMARKER)
560 			continue;
561 
562 		if (!INP_ISIPV6(inp))
563 			continue;
564 		/*
565 		 * If the error designates a new path MTU for a destination
566 		 * and the application (associated with this socket) wanted to
567 		 * know the value, notify. Note that we notify for all
568 		 * disconnected sockets if the corresponding application
569 		 * wanted. This is because some UDP applications keep sending
570 		 * sockets disconnected.
571 		 * XXX: should we avoid to notify the value to TCP sockets?
572 		 */
573 		if (cmd == PRC_MSGSIZE && (inp->inp_flags & IN6P_MTU) != 0 &&
574 		    (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) ||
575 		     IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &sa6_dst->sin6_addr))) {
576 			ip6_notify_pmtu(inp, (struct sockaddr_in6 *)dst, &arg);
577 		}
578 
579 		/*
580 		 * Detect if we should notify the error. If no source and
581 		 * destination ports are specifed, but non-zero flowinfo and
582 		 * local address match, notify the error. This is the case
583 		 * when the error is delivered with an encrypted buffer
584 		 * by ESP. Otherwise, just compare addresses and ports
585 		 * as usual.
586 		 */
587 		if (lport == 0 && fport == 0 && flowinfo &&
588 		    inp->inp_socket != NULL &&
589 		    flowinfo == (inp->in6p_flowinfo & IPV6_FLOWLABEL_MASK) &&
590 		    IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &sa6_src.sin6_addr))
591 			goto do_notify;
592 		else if (!IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr,
593 					     &sa6_dst->sin6_addr) ||
594 			 inp->inp_socket == 0 ||
595 			 (lport && inp->inp_lport != lport) ||
596 			 (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) &&
597 			  !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr,
598 					      &sa6_src.sin6_addr)) ||
599 			 (fport && inp->inp_fport != fport))
600 			continue;
601 
602 do_notify:
603 		if (notify)
604 			(*notify)(inp, arg);
605 	}
606 	LIST_REMOVE(marker, inp_list);
607 
608 	REL_PCBINFO_TOKEN(pcbinfo);
609 }
610 
611 /*
612  * Lookup a PCB based on the local address and port.
613  */
614 struct inpcb *
615 in6_pcblookup_local(struct inpcbporthead *porthash,
616     const struct in6_addr *laddr, u_int lport_arg, int wild_okay,
617     struct ucred *cred)
618 {
619 	struct prison *pscan;
620 	struct prison *pr;
621 	struct inpcb *inp;
622 	int matchwild = 3, wildcard;
623 	u_short lport = lport_arg;
624 	struct inpcbport *phd;
625 	struct inpcb *match = NULL;
626 
627 	/*
628 	 * If the porthashbase is shared across several cpus, it must
629 	 * have been locked.
630 	 */
631 	ASSERT_PORTHASH_TOKEN_HELD(porthash);
632 
633 	/*
634 	 * Best fit PCB lookup.
635 	 *
636 	 * First see if this local port is in use by looking on the
637 	 * port hash list.
638 	 */
639 	LIST_FOREACH(phd, porthash, phd_hash) {
640 		if (phd->phd_port == lport)
641 			break;
642 	}
643 
644 	if (phd != NULL) {
645 		pr = cred ? cred->cr_prison : NULL;
646 
647 		/*
648 		 * Port is in use by one or more PCBs. Look for best
649 		 * fit.
650 		 */
651 		LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
652 			wildcard = 0;
653 			if (!INP_ISIPV6(inp))
654 				continue;
655 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr))
656 				wildcard++;
657 
658 			if (inp->inp_socket && inp->inp_socket->so_cred)
659 				pscan = inp->inp_socket->so_cred->cr_prison;
660 			else
661 				pscan = NULL;
662 			if (pr != pscan)
663 				continue;
664 
665 			if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr)) {
666 				if (!IN6_IS_ADDR_UNSPECIFIED(laddr))
667 					wildcard++;
668 			} else {
669 				if (IN6_IS_ADDR_UNSPECIFIED(laddr))
670 					wildcard++;
671 				else if (!IN6_ARE_ADDR_EQUAL(
672 						    &inp->in6p_laddr, laddr))
673 					continue;
674 			}
675 			if (wildcard && !wild_okay)
676 				continue;
677 			if (wildcard < matchwild) {
678 				match = inp;
679 				matchwild = wildcard;
680 				if (wildcard == 0)
681 					break;
682 			}
683 		}
684 	}
685 	return (match);
686 }
687 
688 void
689 in6_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp)
690 {
691 	struct in6pcb *in6p, *marker;
692 	struct ip6_moptions *im6o;
693 	struct in6_multi_mship *imm, *nimm;
694 
695 	/*
696 	 * We only need to make sure that we are in netisr0, where all
697 	 * multicast operation happen.  We could check inpcbinfo which
698 	 * does not belong to netisr0 by holding the inpcbinfo's token.
699 	 * In this case, the pcbinfo must be able to be shared, i.e.
700 	 * pcbinfo->infotoken is not NULL.
701 	 */
702 	ASSERT_NETISR0;
703 	KASSERT(pcbinfo->cpu == 0 || pcbinfo->infotoken != NULL,
704 	    ("pcbinfo could not be shared"));
705 
706 	/*
707 	 * Get a marker for the current netisr (netisr0).
708 	 *
709 	 * It is possible that the multicast address deletion blocks,
710 	 * which could cause temporary token releasing.  So we use
711 	 * inpcb marker here to get a coherent view of the inpcb list.
712 	 *
713 	 * While, on the other hand, moptions are only added and deleted
714 	 * in netisr0, so we would not see staled moption or miss moption
715 	 * even if the token was released due to the blocking multicast
716 	 * address deletion.
717 	 */
718 	marker = in_pcbmarker();
719 
720 	GET_PCBINFO_TOKEN(pcbinfo);
721 
722 	LIST_INSERT_HEAD(&pcbinfo->pcblisthead, marker, inp_list);
723 	while ((in6p = LIST_NEXT(marker, inp_list)) != NULL) {
724 		LIST_REMOVE(marker, inp_list);
725 		LIST_INSERT_AFTER(in6p, marker, inp_list);
726 
727 		if (in6p->in6p_flags & INP_PLACEMARKER)
728 			continue;
729 		im6o = in6p->in6p_moptions;
730 		if (INP_ISIPV6(in6p) && im6o) {
731 			/*
732 			 * Unselect the outgoing interface if it is being
733 			 * detached.
734 			 */
735 			if (im6o->im6o_multicast_ifp == ifp)
736 				im6o->im6o_multicast_ifp = NULL;
737 
738 			/*
739 			 * Drop multicast group membership if we joined
740 			 * through the interface being detached.
741 			 * XXX controversial - is it really legal for kernel
742 			 * to force this?
743 			 */
744 			for (imm = im6o->im6o_memberships.lh_first;
745 			     imm != NULL; imm = nimm) {
746 				nimm = imm->i6mm_chain.le_next;
747 				if (imm->i6mm_maddr->in6m_ifp == ifp) {
748 					LIST_REMOVE(imm, i6mm_chain);
749 					in6_delmulti(imm->i6mm_maddr);
750 					kfree(imm, M_IPMADDR);
751 				}
752 			}
753 		}
754 	}
755 	LIST_REMOVE(marker, inp_list);
756 
757 	REL_PCBINFO_TOKEN(pcbinfo);
758 }
759 
760 /*
761  * Check for alternatives when higher level complains
762  * about service problems.  For now, invalidate cached
763  * routing information.  If the route was created dynamically
764  * (by a redirect), time to try a default gateway again.
765  */
766 void
767 in6_losing(struct inpcb *in6p)
768 {
769 	struct rtentry *rt;
770 	struct rt_addrinfo info;
771 
772 	if ((rt = in6p->in6p_route.ro_rt) != NULL) {
773 		bzero((caddr_t)&info, sizeof(info));
774 		info.rti_flags = rt->rt_flags;
775 		info.rti_info[RTAX_DST] = rt_key(rt);
776 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
777 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
778 		rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
779 		if (rt->rt_flags & RTF_DYNAMIC) {
780 			rtrequest(RTM_DELETE, rt_key(rt), rt->rt_gateway,
781 			    rt_mask(rt), rt->rt_flags, NULL);
782 		}
783 		in6p->in6p_route.ro_rt = NULL;
784 		rtfree(rt);
785 		/*
786 		 * A new route can be allocated
787 		 * the next time output is attempted.
788 		 */
789 	}
790 }
791 
792 /*
793  * After a routing change, flush old routing
794  * and allocate a (hopefully) better one.
795  */
796 void
797 in6_rtchange(struct inpcb *inp, int error)
798 {
799 	if (inp->in6p_route.ro_rt) {
800 		rtfree(inp->in6p_route.ro_rt);
801 		inp->in6p_route.ro_rt = 0;
802 		/*
803 		 * A new route can be allocated the next time
804 		 * output is attempted.
805 		 */
806 	}
807 }
808 
809 /*
810  * Lookup PCB in hash list.
811  */
812 struct inpcb *
813 in6_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in6_addr *faddr,
814 		   u_int fport_arg, struct in6_addr *laddr, u_int lport_arg,
815 		   int wildcard, struct ifnet *ifp)
816 {
817 	struct inpcbhead *head;
818 	struct inpcb *inp;
819 	struct inpcb *jinp = NULL;
820 	u_short fport = fport_arg, lport = lport_arg;
821 
822 	/*
823 	 * First look for an exact match.
824 	 */
825 	head = &pcbinfo->hashbase[INP_PCBCONNHASH(faddr->s6_addr32[3] /* XXX */,
826 					      fport,
827 					      laddr->s6_addr32[3], /* XXX JH */
828 					      lport,
829 					      pcbinfo->hashmask)];
830 	LIST_FOREACH(inp, head, inp_hash) {
831 		if (!INP_ISIPV6(inp))
832 			continue;
833 		if (IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, faddr) &&
834 		    IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr) &&
835 		    inp->inp_fport == fport &&
836 		    inp->inp_lport == lport) {
837 			/*
838 			 * Found.
839 			 */
840 			if (inp->inp_socket == NULL ||
841 			    inp->inp_socket->so_cred->cr_prison == NULL) {
842 				return (inp);
843 			}
844 			if  (jinp == NULL)
845 				jinp = inp;
846 		}
847 	}
848 	if (jinp != NULL)
849 		return(jinp);
850 
851 	if (wildcard) {
852 		struct inpcb *local_wild = NULL;
853 		struct inpcb *jinp_wild = NULL;
854 		struct inpcontainer *ic;
855 		struct inpcontainerhead *chead;
856 		struct sockaddr_in6 jsin6;
857 		struct ucred *cred;
858 		struct prison *pr;
859 		int net_listen_ov_local = 0;
860 		int net_listen_ov_wild = 0;
861 
862 		/*
863 		 * Order of socket selection:
864 		 * 1. non-jailed, non-wild.
865 		 * 2. non-jailed, wild.		(allow_listen_override on)
866 		 * 3. jailed, non-wild.
867 		 * 4. jailed, wild.
868 		 * 5. non-jailed, wild.		(allow_listen_override off)
869 		 *
870 		 * NOTE: jailed wildcards are still restricted to the jail
871 		 *	 IPs.
872 		 *
873 		 * NOTE: (1) and (3) already handled above.
874 		 */
875 		jsin6.sin6_family = AF_INET6;
876 		chead = &pcbinfo->wildcardhashbase[INP_PCBWILDCARDHASH(lport,
877 		    pcbinfo->wildcardhashmask)];
878 
879 		GET_PCBINFO_TOKEN(pcbinfo);
880 		LIST_FOREACH(ic, chead, ic_list) {
881 			inp = ic->ic_inp;
882 			if (inp->inp_flags & INP_PLACEMARKER)
883 				continue;
884 
885 			/*
886 			 * Basdic validation
887 			 */
888 			if (!INP_ISIPV6(inp))
889 				continue;
890 			if (inp->inp_lport != lport)
891 				continue;
892 
893 			/*
894 			 * Calculate prison, setup jsin for jailed_ip()
895 			 * check.
896 			 */
897 			jsin6.sin6_addr = *laddr;
898 			pr = NULL;
899 			cred = NULL;
900 			if (inp->inp_socket) {
901 				cred = inp->inp_socket->so_cred;
902 				if (cred)
903 					pr = cred->cr_prison;
904 			}
905 
906 			/*
907 			 * Assign jinp, jinp_wild, and local_wild as
908 			 * appropriate, track whether the jail supports
909 			 * listen overrides.
910 			 */
911 			if (pr) {
912 				if (!jailed_ip(pr, (struct sockaddr *)&jsin6))
913 					continue;
914 				if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)
915 				    && jinp == NULL) {
916 					jinp = inp;
917 					if (PRISON_CAP_ISSET(pr->pr_caps, PRISON_CAP_NET_LISTEN_OVERRIDE))
918 						net_listen_ov_local = 1;
919 				}
920 				if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
921 				    jinp_wild == NULL) {
922 					jinp_wild = inp;
923 					if (PRISON_CAP_ISSET(pr->pr_caps, PRISON_CAP_NET_LISTEN_OVERRIDE))
924 						net_listen_ov_wild = 1;
925 				}
926 			} else {
927 				if (IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, laddr)) {
928 					REL_PCBINFO_TOKEN(pcbinfo);
929 					return (inp);
930 				}
931 				if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr))
932 					local_wild = inp;
933 			}
934 		}
935 		REL_PCBINFO_TOKEN(pcbinfo);
936 
937 		if (net_listen_ov_local)
938 			return jinp;
939 		if (net_listen_ov_wild)
940 			return jinp_wild;
941 		if (local_wild)
942 			return local_wild;
943 		if (jinp)
944 			return jinp;
945 		return (jinp_wild);
946 	}
947 
948 	/*
949 	 * Not found.
950 	 */
951 	return (NULL);
952 }
953 
954 void
955 init_sin6(struct sockaddr_in6 *sin6, struct mbuf *m)
956 {
957 	struct ip6_hdr *ip;
958 
959 	ip = mtod(m, struct ip6_hdr *);
960 	bzero(sin6, sizeof(*sin6));
961 	sin6->sin6_len = sizeof(*sin6);
962 	sin6->sin6_family = AF_INET6;
963 	sin6->sin6_addr = ip->ip6_src;
964 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
965 		sin6->sin6_addr.s6_addr16[1] = 0;
966 	sin6->sin6_scope_id =
967 		(m->m_pkthdr.rcvif && IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
968 		? m->m_pkthdr.rcvif->if_index : 0;
969 
970 	return;
971 }
972 
973 void
974 in6_savefaddr(struct socket *so, const struct sockaddr *faddr)
975 {
976 	struct sockaddr_in6 *sin6;
977 
978 	KASSERT(faddr->sa_family == AF_INET6,
979 	    ("not AF_INET6 faddr %d", faddr->sa_family));
980 
981 	sin6 = kmalloc(sizeof(*sin6), M_SONAME, M_WAITOK | M_ZERO);
982 	sin6->sin6_family = AF_INET6;
983 	sin6->sin6_len = sizeof(*sin6);
984 
985 	sin6->sin6_port = ((const struct sockaddr_in6 *)faddr)->sin6_port;
986 	sin6->sin6_addr = ((const struct sockaddr_in6 *)faddr)->sin6_addr;
987 
988 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
989 		sin6->sin6_scope_id = ntohs(sin6->sin6_addr.s6_addr16[1]);
990 	else
991 		sin6->sin6_scope_id = 0;	/*XXX*/
992 	if (IN6_IS_SCOPE_LINKLOCAL(&sin6->sin6_addr))
993 		sin6->sin6_addr.s6_addr16[1] = 0;
994 
995 	so->so_faddr = (struct sockaddr *)sin6;
996 }
997