xref: /netbsd-src/sys/netinet6/in6_pcb.c (revision 1ca5c1b28139779176bd5c13ad7c5f25c0bcd5f8)
1 /*	$NetBSD: in6_pcb.c,v 1.45 2001/12/21 08:54:54 itojun Exp $	*/
2 /*	$KAME: in6_pcb.c,v 1.84 2001/02/08 18:02:08 itojun 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  * Copyright (c) 1982, 1986, 1991, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. All advertising materials mentioning features or use of this software
46  *    must display the following acknowledgement:
47  *	This product includes software developed by the University of
48  *	California, Berkeley and its contributors.
49  * 4. Neither the name of the University nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63  * SUCH DAMAGE.
64  *
65  *	@(#)in_pcb.c	8.2 (Berkeley) 1/4/94
66  */
67 
68 #include <sys/cdefs.h>
69 __KERNEL_RCSID(0, "$NetBSD: in6_pcb.c,v 1.45 2001/12/21 08:54:54 itojun Exp $");
70 
71 #include "opt_ipsec.h"
72 
73 #include <sys/param.h>
74 #include <sys/systm.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/protosw.h>
78 #include <sys/socket.h>
79 #include <sys/socketvar.h>
80 #include <sys/ioctl.h>
81 #include <sys/errno.h>
82 #include <sys/time.h>
83 #include <sys/proc.h>
84 
85 #include <net/if.h>
86 #include <net/route.h>
87 
88 #include <netinet/in.h>
89 #include <netinet/in_var.h>
90 #include <netinet/in_systm.h>
91 #include <netinet/ip.h>
92 #include <netinet/in_pcb.h>
93 #include <netinet/ip6.h>
94 #include <netinet6/ip6_var.h>
95 #include <netinet6/in6_pcb.h>
96 #include <netinet6/nd6.h>
97 
98 #include "loop.h"
99 extern struct ifnet loif[NLOOP];
100 #include "faith.h"
101 
102 #ifdef IPSEC
103 #include <netinet6/ipsec.h>
104 #include <netkey/key.h>
105 #endif /* IPSEC */
106 
107 struct in6_addr zeroin6_addr;
108 
109 int ip6_anonportmin = IPV6PORT_ANONMIN;
110 int ip6_anonportmax = IPV6PORT_ANONMAX;
111 int ip6_lowportmin  = IPV6PORT_RESERVEDMIN;
112 int ip6_lowportmax  = IPV6PORT_RESERVEDMAX;
113 
114 int
115 in6_pcballoc(so, head)
116 	struct socket *so;
117 	struct in6pcb *head;
118 {
119 	struct in6pcb *in6p;
120 #ifdef IPSEC
121 	int error;
122 #endif
123 
124 	MALLOC(in6p, struct in6pcb *, sizeof(*in6p), M_PCB, M_NOWAIT);
125 	if (in6p == NULL)
126 		return(ENOBUFS);
127 	bzero((caddr_t)in6p, sizeof(*in6p));
128 	in6p->in6p_head = head;
129 	in6p->in6p_socket = so;
130 	in6p->in6p_hops = -1;	/* use kernel default */
131 	in6p->in6p_icmp6filt = NULL;
132 #ifdef IPSEC
133 	error = ipsec_init_policy(so, &in6p->in6p_sp);
134 	if (error != 0) {
135 		FREE(in6p, M_PCB);
136 		return error;
137 	}
138 #endif /* IPSEC */
139 	in6p->in6p_next = head->in6p_next;
140 	head->in6p_next = in6p;
141 	in6p->in6p_prev = head;
142 	in6p->in6p_next->in6p_prev = in6p;
143 	if (ip6_v6only)
144 		in6p->in6p_flags |= IN6P_IPV6_V6ONLY;
145 	so->so_pcb = (caddr_t)in6p;
146 	return(0);
147 }
148 
149 int
150 in6_pcbbind(in6p, nam, p)
151 	struct in6pcb *in6p;
152 	struct mbuf *nam;
153 	struct proc *p;
154 {
155 	struct socket *so = in6p->in6p_socket;
156 	struct in6pcb *head = in6p->in6p_head;
157 	struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)NULL;
158 	u_int16_t lport = 0;
159 	int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
160 
161 	if (in6p->in6p_lport || !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
162 		return(EINVAL);
163 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0 &&
164 	   ((so->so_proto->pr_flags & PR_CONNREQUIRED) == 0 ||
165 	    (so->so_options & SO_ACCEPTCONN) == 0))
166 		wild = IN6PLOOKUP_WILDCARD;
167 	if (nam) {
168 		sin6 = mtod(nam, struct sockaddr_in6 *);
169 		if (nam->m_len != sizeof(*sin6))
170 			return(EINVAL);
171 		/*
172 		 * We should check the family, but old programs
173 		 * incorrectly fail to intialize it.
174 		 */
175 		if (sin6->sin6_family != AF_INET6)
176 			return(EAFNOSUPPORT);
177 
178 		/*
179 		 * since we do not check port number duplicate with IPv4 space,
180 		 * we reject it at this moment.  If we leave it, we would
181 		 * mistakenly allow normal users to hijack tcp/udp ports from
182 		 * other users.
183 		 */
184 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr))
185 			return(EADDRNOTAVAIL);
186 
187 		/* KAME hack: embed scopeid */
188 		if (in6_embedscope(&sin6->sin6_addr, sin6, in6p, NULL) != 0)
189 			return EINVAL;
190 		/* this must be cleared for ifa_ifwithaddr() */
191 		sin6->sin6_scope_id = 0;
192 
193 		lport = sin6->sin6_port;
194 		if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
195 			/*
196 			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
197 			 * allow compepte duplication of binding if
198 			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
199 			 * and a multicast address is bound on both
200 			 * new and duplicated sockets.
201 			 */
202 			if (so->so_options & SO_REUSEADDR)
203 				reuseport = SO_REUSEADDR|SO_REUSEPORT;
204 		}
205 		else if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
206 			struct sockaddr_in sin;
207 
208 			bzero(&sin, sizeof(sin));
209 			sin.sin_len = sizeof(sin);
210 			sin.sin_family = AF_INET;
211 			bcopy(&sin6->sin6_addr.s6_addr32[3], &sin.sin_addr,
212 				sizeof(sin.sin_addr));
213 			if (ifa_ifwithaddr((struct sockaddr *)&sin) == 0)
214 				return EADDRNOTAVAIL;
215 		}
216 		else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
217 			struct ifaddr *ia = NULL;
218 
219 			sin6->sin6_port = 0;		/* yech... */
220 			if ((in6p->in6p_flags & IN6P_FAITH) == 0 &&
221 			    (ia = ifa_ifwithaddr((struct sockaddr *)sin6)) == 0)
222 				return(EADDRNOTAVAIL);
223 
224 			/*
225 			 * bind to an anycast address might accidentally
226 			 * cause sending a packet with an anycast source
227 			 * address, so we forbid it.
228 			 */
229 			if (ia &&
230 			    ((struct in6_ifaddr *)ia)->ia6_flags &
231 			    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|
232 			     IN6_IFF_DETACHED|IN6_IFF_DEPRECATED)) {
233 				return(EADDRNOTAVAIL);
234 			}
235 		}
236 		if (lport) {
237 #ifndef IPNOPRIVPORTS
238 			int priv;
239 
240 			/*
241 			 * NOTE: all operating systems use suser() for
242 			 * privilege check!  do not rewrite it into SS_PRIV.
243 			 */
244 			priv = (p && !suser(p->p_ucred, &p->p_acflag)) ? 1 : 0;
245 			/* GROSS */
246 			if (ntohs(lport) < IPV6PORT_RESERVED && !priv)
247 				return(EACCES);
248 #endif
249 
250 			if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
251 				/* should check this but we can't ... */
252 #if 0
253 				struct inpcb *t;
254 
255 				t = in_pcblookup_bind(&tcbtable,
256 					(struct in_addr *)&sin6->sin6_addr.s6_addr32[3],
257 					lport);
258 				if (t && (reuseport & t->inp_socket->so_options) == 0)
259 					return EADDRINUSE;
260 #endif
261 			} else
262 			{
263 				struct in6pcb *t;
264 
265 				t = in6_pcblookup(head, &zeroin6_addr, 0,
266 						  &sin6->sin6_addr, lport, wild);
267 				if (t && (reuseport & t->in6p_socket->so_options) == 0)
268 					return(EADDRINUSE);
269 			}
270 		}
271 		in6p->in6p_laddr = sin6->sin6_addr;
272 	}
273 
274 	if (lport == 0) {
275 		int e;
276 		if ((e = in6_pcbsetport(&in6p->in6p_laddr, in6p)) != 0)
277 			return(e);
278 	}
279 	else
280 		in6p->in6p_lport = lport;
281 
282 	in6p->in6p_flowinfo = sin6 ? sin6->sin6_flowinfo : 0;	/*XXX*/
283 	return(0);
284 }
285 
286 /*
287  * Connect from a socket to a specified address.
288  * Both address and port must be specified in argument sin6.
289  * If don't have a local address for this socket yet,
290  * then pick one.
291  */
292 int
293 in6_pcbconnect(in6p, nam)
294 	struct in6pcb *in6p;
295 	struct mbuf *nam;
296 {
297 	struct in6_addr *in6a = NULL;
298 	struct sockaddr_in6 *sin6 = mtod(nam, struct sockaddr_in6 *);
299 	struct ifnet *ifp = NULL;	/* outgoing interface */
300 	int error = 0;
301 #ifdef INET
302 	struct in6_addr mapped;
303 #endif
304 	struct sockaddr_in6 tmp;
305 
306 	(void)&in6a;				/* XXX fool gcc */
307 
308 	if (nam->m_len != sizeof(*sin6))
309 		return(EINVAL);
310 	if (sin6->sin6_family != AF_INET6)
311 		return(EAFNOSUPPORT);
312 	if (sin6->sin6_port == 0)
313 		return(EADDRNOTAVAIL);
314 
315 	/* sanity check for mapped address case */
316 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
317 		if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY) != 0)
318 			return EINVAL;
319 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
320 			in6p->in6p_laddr.s6_addr16[5] = htons(0xffff);
321 		if (!IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
322 			return EINVAL;
323 	} else
324 	{
325 		if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr))
326 			return EINVAL;
327 	}
328 
329 	/* protect *sin6 from overwrites */
330 	tmp = *sin6;
331 	sin6 = &tmp;
332 
333 	/* KAME hack: embed scopeid */
334 	if (in6_embedscope(&sin6->sin6_addr, sin6, in6p, &ifp) != 0)
335 		return EINVAL;
336 
337 	/* Source address selection. */
338 	if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr)
339 	 && in6p->in6p_laddr.s6_addr32[3] == 0) {
340 #ifdef INET
341 		struct sockaddr_in sin, *sinp;
342 
343 		bzero(&sin, sizeof(sin));
344 		sin.sin_len = sizeof(sin);
345 		sin.sin_family = AF_INET;
346 		bcopy(&sin6->sin6_addr.s6_addr32[3], &sin.sin_addr,
347 			sizeof(sin.sin_addr));
348 		sinp = in_selectsrc(&sin, (struct route *)&in6p->in6p_route,
349 			in6p->in6p_socket->so_options, NULL, &error);
350 		if (sinp == 0) {
351 			if (error == 0)
352 				error = EADDRNOTAVAIL;
353 			return(error);
354 		}
355 		bzero(&mapped, sizeof(mapped));
356 		mapped.s6_addr16[5] = htons(0xffff);
357 		bcopy(&sinp->sin_addr, &mapped.s6_addr32[3], sizeof(sinp->sin_addr));
358 		in6a = &mapped;
359 #else
360 		return EADDRNOTAVAIL;
361 #endif
362 	} else
363 	{
364 		/*
365 		 * XXX: in6_selectsrc might replace the bound local address
366 		 * with the address specified by setsockopt(IPV6_PKTINFO).
367 		 * Is it the intended behavior?
368 		 */
369 		in6a = in6_selectsrc(sin6, in6p->in6p_outputopts,
370 				     in6p->in6p_moptions,
371 				     &in6p->in6p_route,
372 				     &in6p->in6p_laddr, &error);
373 		if (in6a == 0) {
374 			if (error == 0)
375 				error = EADDRNOTAVAIL;
376 			return(error);
377 		}
378 	}
379 	if (in6p->in6p_route.ro_rt)
380 		ifp = in6p->in6p_route.ro_rt->rt_ifp;
381 
382 	in6p->in6p_ip6.ip6_hlim = (u_int8_t)in6_selecthlim(in6p, ifp);
383 
384 	if (in6_pcblookup(in6p->in6p_head,
385 			 &sin6->sin6_addr,
386 			 sin6->sin6_port,
387 			 IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) ?
388 			  in6a : &in6p->in6p_laddr,
389 			 in6p->in6p_lport,
390 			 0))
391 		return(EADDRINUSE);
392 	if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)
393 	 || (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr)
394 	  && in6p->in6p_laddr.s6_addr32[3] == 0))
395 	{
396 		/*
397 		 * XXX in IPv4 mapped address case, we should grab fresh
398 		 * local port number by in_pcbbind, not in6_pcbbind.
399 		 * if we are in bad luck, we may assign conflicting port number
400 		 * between IPv4 and IPv6 unwillingly.
401 		 */
402 		if (in6p->in6p_lport == 0) {
403 			(void)in6_pcbbind(in6p, (struct mbuf *)0,
404 			    (struct proc *)0);
405 		}
406 		in6p->in6p_laddr = *in6a;
407 	}
408 	in6p->in6p_faddr = sin6->sin6_addr;
409 	in6p->in6p_fport = sin6->sin6_port;
410 	/*
411 	 * xxx kazu flowlabel is necessary for connect?
412 	 * but if this line is missing, the garbage value remains.
413 	 */
414 	in6p->in6p_flowinfo = sin6->sin6_flowinfo;
415 #ifdef IPSEC
416 	if (in6p->in6p_socket->so_type == SOCK_STREAM)
417 		ipsec_pcbconn(in6p->in6p_sp);
418 #endif
419 	return(0);
420 }
421 
422 void
423 in6_pcbdisconnect(in6p)
424 	struct in6pcb *in6p;
425 {
426 	bzero((caddr_t)&in6p->in6p_faddr, sizeof(in6p->in6p_faddr));
427 	in6p->in6p_fport = 0;
428 	if (in6p->in6p_socket->so_state & SS_NOFDREF)
429 		in6_pcbdetach(in6p);
430 #ifdef IPSEC
431 	ipsec_pcbdisconn(in6p->in6p_sp);
432 #endif
433 }
434 
435 void
436 in6_pcbdetach(in6p)
437 	struct in6pcb *in6p;
438 {
439 	struct socket *so = in6p->in6p_socket;
440 
441 #ifdef IPSEC
442 	ipsec6_delete_pcbpolicy(in6p);
443 #endif /* IPSEC */
444 	sotoin6pcb(so) = 0;
445 	sofree(so);
446 	if (in6p->in6p_options)
447 		m_freem(in6p->in6p_options);
448 	if (in6p->in6p_outputopts) {
449 		if (in6p->in6p_outputopts->ip6po_rthdr &&
450 		    in6p->in6p_outputopts->ip6po_route.ro_rt)
451 			RTFREE(in6p->in6p_outputopts->ip6po_route.ro_rt);
452 		if (in6p->in6p_outputopts->ip6po_m)
453 			(void)m_free(in6p->in6p_outputopts->ip6po_m);
454 		free(in6p->in6p_outputopts, M_IP6OPT);
455 	}
456 	if (in6p->in6p_route.ro_rt)
457 		rtfree(in6p->in6p_route.ro_rt);
458 	ip6_freemoptions(in6p->in6p_moptions);
459 	in6p->in6p_next->in6p_prev = in6p->in6p_prev;
460 	in6p->in6p_prev->in6p_next = in6p->in6p_next;
461 	in6p->in6p_prev = NULL;
462 	FREE(in6p, M_PCB);
463 }
464 
465 void
466 in6_setsockaddr(in6p, nam)
467 	struct in6pcb *in6p;
468 	struct mbuf *nam;
469 {
470 	struct sockaddr_in6 *sin6;
471 
472 	nam->m_len = sizeof(*sin6);
473 	sin6 = mtod(nam, struct sockaddr_in6 *);
474 	bzero((caddr_t)sin6, sizeof(*sin6));
475 	sin6->sin6_family = AF_INET6;
476 	sin6->sin6_len = sizeof(struct sockaddr_in6);
477 	sin6->sin6_port = in6p->in6p_lport;
478 	/* KAME hack: recover scopeid */
479 	(void)in6_recoverscope(sin6, &in6p->in6p_laddr, NULL);
480 }
481 
482 void
483 in6_setpeeraddr(in6p, nam)
484 	struct in6pcb *in6p;
485 	struct mbuf *nam;
486 {
487 	struct sockaddr_in6 *sin6;
488 
489 	nam->m_len = sizeof(*sin6);
490 	sin6 = mtod(nam, struct sockaddr_in6 *);
491 	bzero((caddr_t)sin6, sizeof(*sin6));
492 	sin6->sin6_family = AF_INET6;
493 	sin6->sin6_len = sizeof(struct sockaddr_in6);
494 	sin6->sin6_port = in6p->in6p_fport;
495 	/* KAME hack: recover scopeid */
496 	(void)in6_recoverscope(sin6, &in6p->in6p_faddr, NULL);
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  * Must be called at splsoftnet.
509  *
510  * Note: src (4th arg) carries the flowlabel value on the original IPv6
511  * header, in sin6_flowinfo member.
512  */
513 int
514 in6_pcbnotify(head, dst, fport_arg, src, lport_arg, cmd, cmdarg, notify)
515 	struct in6pcb *head;
516 	struct sockaddr *dst, *src;
517 	u_int fport_arg, lport_arg;
518 	int cmd;
519 	void *cmdarg;
520 	void (*notify) __P((struct in6pcb *, int));
521 {
522 	struct in6pcb *in6p, *nin6p;
523 	struct sockaddr_in6 sa6_src, *sa6_dst;
524 	u_int16_t fport = fport_arg, lport = lport_arg;
525 	int errno;
526 	int nmatch = 0;
527 	u_int32_t flowinfo;
528 
529 	if ((unsigned)cmd > PRC_NCMDS || dst->sa_family != AF_INET6)
530 		return 0;
531 
532 	sa6_dst = (struct sockaddr_in6 *)dst;
533 	if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr))
534 		return 0;
535 
536 	/*
537 	 * note that src can be NULL when we get notify by local fragmentation.
538 	 */
539 	sa6_src = (src == NULL) ? sa6_any : *(struct sockaddr_in6 *)src;
540 	flowinfo = sa6_src.sin6_flowinfo;
541 
542 	/*
543 	 * Redirects go to all references to the destination,
544 	 * and use in6_rtchange to invalidate the route cache.
545 	 * Dead host indications: also use in6_rtchange to invalidate
546 	 * the cache, and deliver the error to all the sockets.
547 	 * Otherwise, if we have knowledge of the local port and address,
548 	 * deliver only to that socket.
549 	 */
550 	if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
551 		fport = 0;
552 		lport = 0;
553 		bzero((caddr_t)&sa6_src.sin6_addr, sizeof(sa6_src.sin6_addr));
554 
555 		if (cmd != PRC_HOSTDEAD)
556 			notify = in6_rtchange;
557 	}
558 
559 	errno = inet6ctlerrmap[cmd];
560 	for (in6p = head->in6p_next; in6p != head; in6p = nin6p) {
561 		nin6p = in6p->in6p_next;
562 
563 		/*
564 		 * Under the following condition, notify of redirects
565 		 * to the pcb, without making address matches against inpcb.
566 		 * - redirect notification is arrived.
567 		 * - the inpcb is unconnected.
568 		 * - the inpcb is caching !RTF_HOST routing entry.
569 		 * - the ICMPv6 notification is from the gateway cached in the
570 		 *   inpcb.  i.e. ICMPv6 notification is from nexthop gateway
571 		 *   the inpcb used very recently.
572 		 *
573 		 * This is to improve interaction between netbsd/openbsd
574 		 * redirect handling code, and inpcb route cache code.
575 		 * without the clause, !RTF_HOST routing entry (which carries
576 		 * gateway used by inpcb right before the ICMPv6 redirect)
577 		 * will be cached forever in unconnected inpcb.
578 		 *
579 		 * There still is a question regarding to what is TRT:
580 		 * - On bsdi/freebsd, RTF_HOST (cloned) routing entry will be
581 		 *   generated on packet output.  inpcb will always cache
582 		 *   RTF_HOST routing entry so there's no need for the clause
583 		 *   (ICMPv6 redirect will update RTF_HOST routing entry,
584 		 *   and inpcb is caching it already).
585 		 *   However, bsdi/freebsd are vulnerable to local DoS attacks
586 		 *   due to the cloned routing entries.
587 		 * - Specwise, "destination cache" is mentioned in RFC2461.
588 		 *   Jinmei says that it implies bsdi/freebsd behavior, itojun
589 		 *   is not really convinced.
590 		 * - Having hiwat/lowat on # of cloned host route (redirect/
591 		 *   pmtud) may be a good idea.  netbsd/openbsd has it.  see
592 		 *   icmp6_mtudisc_update().
593 		 */
594 		if ((PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) &&
595 		    IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
596 		    in6p->in6p_route.ro_rt &&
597 		    !(in6p->in6p_route.ro_rt->rt_flags & RTF_HOST)) {
598 			struct sockaddr_in6 *dst6;
599 
600 			dst6 = (struct sockaddr_in6 *)&in6p->in6p_route.ro_dst;
601 			if (IN6_ARE_ADDR_EQUAL(&dst6->sin6_addr,
602 			    &sa6_dst->sin6_addr))
603 				goto do_notify;
604 		}
605 
606 		/*
607 		 * Detect if we should notify the error. If no source and
608 		 * destination ports are specified, but non-zero flowinfo and
609 		 * local address match, notify the error. This is the case
610 		 * when the error is delivered with an encrypted buffer
611 		 * by ESP. Otherwise, just compare addresses and ports
612 		 * as usual.
613 		 */
614 		if (lport == 0 && fport == 0 && flowinfo &&
615 		    in6p->in6p_socket != NULL &&
616 		    flowinfo == (in6p->in6p_flowinfo & IPV6_FLOWLABEL_MASK) &&
617 		    IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &sa6_src.sin6_addr))
618 			goto do_notify;
619 		else if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr,
620 					     &sa6_dst->sin6_addr) ||
621 		    in6p->in6p_socket == 0 ||
622 		    (lport && in6p->in6p_lport != lport) ||
623 		    (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) &&
624 		     !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr,
625 					 &sa6_src.sin6_addr)) ||
626 		    (fport && in6p->in6p_fport != fport))
627 			continue;
628 
629 	  do_notify:
630 		if (notify)
631 			(*notify)(in6p, errno);
632 		nmatch++;
633 	}
634 	return nmatch;
635 }
636 
637 void
638 in6_pcbpurgeif0(head, ifp)
639 	struct in6pcb *head;
640 	struct ifnet *ifp;
641 {
642 	struct in6pcb *in6p, *nin6p;
643 	struct ip6_moptions *im6o;
644 	struct in6_multi_mship *imm, *nimm;
645 
646 	for (in6p = head->in6p_next; in6p != head; in6p = nin6p) {
647 		nin6p = in6p->in6p_next;
648 		im6o = in6p->in6p_moptions;
649 		if (im6o) {
650 			/*
651 			 * Unselect the outgoing interface if it is being
652 			 * detached.
653 			 */
654 			if (im6o->im6o_multicast_ifp == ifp)
655 				im6o->im6o_multicast_ifp = NULL;
656 
657 			/*
658 			 * Drop multicast group membership if we joined
659 			 * through the interface being detached.
660 			 * XXX controversial - is it really legal for kernel
661 			 * to force this?
662 			 */
663 			for (imm = im6o->im6o_memberships.lh_first;
664 			     imm != NULL; imm = nimm) {
665 				nimm = imm->i6mm_chain.le_next;
666 				if (imm->i6mm_maddr->in6m_ifp == ifp) {
667 					LIST_REMOVE(imm, i6mm_chain);
668 					in6_leavegroup(imm);
669 				}
670 			}
671 		}
672 	}
673 }
674 
675 void
676 in6_pcbpurgeif(head, ifp)
677 	struct in6pcb *head;
678 	struct ifnet *ifp;
679 {
680 	struct in6pcb *in6p, *nin6p;
681 
682 	for (in6p = head->in6p_next; in6p != head; in6p = nin6p) {
683 		nin6p = in6p->in6p_next;
684 		if (in6p->in6p_route.ro_rt != NULL &&
685 		    in6p->in6p_route.ro_rt->rt_ifp == ifp)
686 			in6_rtchange(in6p, 0);
687 	}
688 }
689 
690 /*
691  * Check for alternatives when higher level complains
692  * about service problems.  For now, invalidate cached
693  * routing information.  If the route was created dynamically
694  * (by a redirect), time to try a default gateway again.
695  */
696 void
697 in6_losing(in6p)
698 	struct in6pcb *in6p;
699 {
700 	struct rtentry *rt;
701 	struct rt_addrinfo info;
702 
703 	if ((rt = in6p->in6p_route.ro_rt) != NULL) {
704 		in6p->in6p_route.ro_rt = 0;
705 		bzero((caddr_t)&info, sizeof(info));
706 		info.rti_info[RTAX_DST] =
707 			(struct sockaddr *)&in6p->in6p_route.ro_dst;
708 		info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
709 		info.rti_info[RTAX_NETMASK] = rt_mask(rt);
710 		rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
711 		if (rt->rt_flags & RTF_DYNAMIC) {
712 			(void)rtrequest(RTM_DELETE, rt_key(rt),
713 					rt->rt_gateway, rt_mask(rt), rt->rt_flags,
714 					(struct rtentry **)0);
715 		} else {
716 			/*
717 			 * A new route can be allocated
718 			 * the next time output is attempted.
719 			 */
720 			rtfree(rt);
721 		}
722 	}
723 }
724 
725 /*
726  * After a routing change, flush old routing
727  * and allocate a (hopefully) better one.
728  */
729 void
730 in6_rtchange(in6p, errno)
731 	struct in6pcb *in6p;
732 	int errno;
733 {
734 	if (in6p->in6p_route.ro_rt) {
735 		rtfree(in6p->in6p_route.ro_rt);
736 		in6p->in6p_route.ro_rt = 0;
737 		/*
738 		 * A new route can be allocated the next time
739 		 * output is attempted.
740 		 */
741 	}
742 }
743 
744 struct in6pcb *
745 in6_pcblookup(head, faddr6, fport_arg, laddr6, lport_arg, flags)
746 	struct in6pcb *head;
747 	struct in6_addr *faddr6, *laddr6;
748 	u_int fport_arg, lport_arg;
749 	int flags;
750 {
751 	struct in6pcb *in6p, *match = 0;
752 	int matchwild = 3, wildcard;
753 	u_int16_t fport = fport_arg, lport = lport_arg;
754 
755 	for (in6p = head->in6p_next; in6p != head; in6p = in6p->in6p_next) {
756 		if (in6p->in6p_lport != lport)
757 			continue;
758 		wildcard = 0;
759 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
760 			if (IN6_IS_ADDR_UNSPECIFIED(laddr6))
761 				wildcard++;
762 			else if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, laddr6))
763 				continue;
764 		}
765 		else if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr)
766 			&& in6p->in6p_laddr.s6_addr32[3] == 0) {
767 			if (!IN6_IS_ADDR_V4MAPPED(laddr6))
768 				continue;
769 			if (laddr6->s6_addr32[3] == 0)
770 				;
771 			else
772 				wildcard++;
773 		}
774 		else {
775 			if (IN6_IS_ADDR_V4MAPPED(laddr6)) {
776 				if (in6p->in6p_flags & IN6P_IPV6_V6ONLY)
777 					continue;
778 				else
779 					wildcard++;
780 			} else if (!IN6_IS_ADDR_UNSPECIFIED(laddr6))
781 				wildcard++;
782 		}
783 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
784 			if (IN6_IS_ADDR_UNSPECIFIED(faddr6))
785 				wildcard++;
786 			else if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, faddr6)
787 			      || in6p->in6p_fport != fport)
788 				continue;
789 		}
790 		else if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_faddr)
791 			&& in6p->in6p_faddr.s6_addr32[3] == 0) {
792 			if (!IN6_IS_ADDR_V4MAPPED(faddr6))
793 				continue;
794 			if (faddr6->s6_addr32[3] == 0)
795 				;
796 			else
797 				wildcard++;
798 		}
799 		else {
800 			if (IN6_IS_ADDR_V4MAPPED(faddr6)) {
801 				if (in6p->in6p_flags & IN6P_IPV6_V6ONLY)
802 					continue;
803 				else
804 					wildcard++;
805 			} else if (!IN6_IS_ADDR_UNSPECIFIED(faddr6))
806 				wildcard++;
807 		}
808 
809 		if (wildcard && (flags & IN6PLOOKUP_WILDCARD) == 0)
810 			continue;
811 		if (wildcard < matchwild) {
812 			match = in6p;
813 			matchwild = wildcard;
814 			if (matchwild == 0)
815 				break;
816 		}
817 	}
818 	return(match);
819 }
820 
821 struct rtentry *
822 in6_pcbrtentry(in6p)
823 	struct in6pcb *in6p;
824 {
825 	struct route_in6 *ro;
826 	struct sockaddr_in6 *dst6;
827 
828 	ro = &in6p->in6p_route;
829 	dst6 = (struct sockaddr_in6 *)&ro->ro_dst;
830 
831 	if (ro->ro_rt == NULL) {
832 		/*
833 		 * No route yet, so try to acquire one.
834 		 */
835 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr)) {
836 			bzero(dst6, sizeof(*dst6));
837 			dst6->sin6_family = AF_INET6;
838 			dst6->sin6_len = sizeof(struct sockaddr_in6);
839 			dst6->sin6_addr = in6p->in6p_faddr;
840 			rtalloc((struct route *)ro);
841 		}
842 	}
843 	return (ro->ro_rt);
844 }
845 
846 struct in6pcb *
847 in6_pcblookup_connect(head, faddr6, fport_arg, laddr6, lport_arg, faith)
848 	struct in6pcb *head;
849 	struct in6_addr *faddr6, *laddr6;
850 	u_int fport_arg, lport_arg;
851 	int faith;
852 {
853 	struct in6pcb *in6p;
854 	u_int16_t fport = fport_arg, lport = lport_arg;
855 
856 	for (in6p = head->in6p_next; in6p != head; in6p = in6p->in6p_next) {
857 		/* find exact match on both source and dest */
858 		if (in6p->in6p_fport != fport)
859 			continue;
860 		if (in6p->in6p_lport != lport)
861 			continue;
862 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr))
863 			continue;
864 		if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, faddr6))
865 			continue;
866 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
867 			continue;
868 		if (!IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, laddr6))
869 			continue;
870 		if ((IN6_IS_ADDR_V4MAPPED(laddr6) ||
871 		     IN6_IS_ADDR_V4MAPPED(faddr6)) &&
872 		    (in6p->in6p_flags & IN6P_IPV6_V6ONLY))
873 			continue;
874 		return in6p;
875 	}
876 	return NULL;
877 }
878 
879 struct in6pcb *
880 in6_pcblookup_bind(head, laddr6, lport_arg, faith)
881 	struct in6pcb *head;
882 	struct in6_addr *laddr6;
883 	u_int lport_arg;
884 	int faith;
885 {
886 	struct in6pcb *in6p, *match;
887 	u_int16_t lport = lport_arg;
888 
889 	match = NULL;
890 	for (in6p = head->in6p_next; in6p != head; in6p = in6p->in6p_next) {
891 		/*
892 	 	 * find destination match.  exact match is preferred
893 		 * against wildcard match.
894 		 */
895 		if (faith && (in6p->in6p_flags & IN6P_FAITH) == 0)
896 			continue;
897 		if (in6p->in6p_fport != 0)
898 			continue;
899 		if (in6p->in6p_lport != lport)
900 			continue;
901 		if (IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
902 			if (IN6_IS_ADDR_V4MAPPED(laddr6)) {
903 				if ((in6p->in6p_flags & IN6P_IPV6_V6ONLY))
904 					continue;
905 				else
906 					match = in6p;
907 			} else
908 				match = in6p;
909 		}
910 		else if (IN6_IS_ADDR_V4MAPPED(&in6p->in6p_laddr) &&
911 			 !(in6p->in6p_flags & IN6P_IPV6_V6ONLY) &&
912 			 in6p->in6p_laddr.s6_addr32[3] == 0) {
913 			if (IN6_IS_ADDR_V4MAPPED(laddr6) &&
914 			    laddr6->s6_addr32[3] != 0)
915 				match = in6p;
916 		}
917 		else if (IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, laddr6))
918 			return in6p;
919 	}
920 	return match;
921 }
922