xref: /openbsd-src/sys/netinet6/in6_pcb.c (revision db3296cf5c1dd9058ceecc3a29fe4aaa0bd26000)
1 /*	$OpenBSD: in6_pcb.c,v 1.34 2003/06/02 23:28:15 millert Exp $	*/
2 
3 /*
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 /*
33  *	@(#)COPYRIGHT	1.1 (NRL) 17 January 1995
34  *
35  * NRL grants permission for redistribution and use in source and binary
36  * forms, with or without modification, of the software and documentation
37  * created at NRL provided that the following conditions are met:
38  *
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 3. All advertising materials mentioning features or use of this software
45  *    must display the following acknowledgements:
46  * 	This product includes software developed by the University of
47  * 	California, Berkeley and its contributors.
48  * 	This product includes software developed at the Information
49  * 	Technology Division, US Naval Research Laboratory.
50  * 4. Neither the name of the NRL nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS
55  * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
56  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
57  * PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL NRL OR
58  * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
59  * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
60  * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
61  * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
62  * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
63  * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
64  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
65  *
66  * The views and conclusions contained in the software and documentation
67  * are those of the authors and should not be interpreted as representing
68  * official policies, either expressed or implied, of the US Naval
69  * Research Laboratory (NRL).
70  */
71 
72 /*
73  * Copyright (c) 1982, 1986, 1990, 1993, 1995
74  *	Regents of the University of California.  All rights reserved.
75  *
76  * Redistribution and use in source and binary forms, with or without
77  * modification, are permitted provided that the following conditions
78  * are met:
79  * 1. Redistributions of source code must retain the above copyright
80  *    notice, this list of conditions and the following disclaimer.
81  * 2. Redistributions in binary form must reproduce the above copyright
82  *    notice, this list of conditions and the following disclaimer in the
83  *    documentation and/or other materials provided with the distribution.
84  * 3. Neither the name of the University nor the names of its contributors
85  *    may be used to endorse or promote products derived from this software
86  *    without specific prior written permission.
87  *
88  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
89  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
90  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
91  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
92  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
93  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
94  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
95  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
96  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
97  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
98  * SUCH DAMAGE.
99  *
100  */
101 
102 #include <sys/param.h>
103 #include <sys/systm.h>
104 #include <sys/malloc.h>
105 #include <sys/mbuf.h>
106 #include <sys/domain.h>
107 #include <sys/protosw.h>
108 #include <sys/socket.h>
109 #include <sys/socketvar.h>
110 #include <sys/errno.h>
111 #include <sys/time.h>
112 #include <sys/proc.h>
113 
114 #include <net/if.h>
115 #include <net/route.h>
116 
117 #include <netinet/in.h>
118 #include <netinet/in_systm.h>
119 #include <netinet/ip.h>
120 #include <netinet/in_pcb.h>
121 
122 #include <netinet6/in6_var.h>
123 #include <netinet/ip6.h>
124 #include <netinet6/ip6_var.h>
125 
126 /*
127  * External globals
128  */
129 
130 #include <dev/rndvar.h>
131 
132 extern struct in6_ifaddr *in6_ifaddr;
133 extern struct in_ifaddr *in_ifaddr;
134 
135 /*
136  * Globals
137  */
138 
139 struct in6_addr zeroin6_addr;
140 
141 extern int ipport_firstauto;
142 extern int ipport_lastauto;
143 extern int ipport_hifirstauto;
144 extern int ipport_hilastauto;
145 
146 /*
147  * Keep separate inet6ctlerrmap, because I may remap some of these.
148  * I also put it here, because, quite frankly, it belongs here, not in
149  * ip{v6,}_input().
150  */
151 #if 0
152 u_char inet6ctlerrmap[PRC_NCMDS] = {
153 	0,		0,		0,		0,
154 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
155 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
156 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
157 	0,		0,		0,		0,
158 	ENOPROTOOPT
159 };
160 #endif
161 
162 /*
163  * Bind an address (or at least a port) to an PF_INET6 socket.
164  */
165 int
166 in6_pcbbind(inp, nam)
167 	struct inpcb *inp;
168 	struct mbuf *nam;
169 {
170 	struct socket *so = inp->inp_socket;
171 
172 	struct inpcbtable *head = inp->inp_table;
173 	struct sockaddr_in6 *sin6;
174 	struct proc *p = curproc;		/* XXX */
175 	u_short lport = 0;
176 	int wild = INPLOOKUP_IPV6, reuseport = (so->so_options & SO_REUSEPORT);
177 	int error;
178 
179 	/*
180 	 * REMINDER:  Once up to speed, flow label processing should go here,
181 	 * too.  (Same with in6_pcbconnect.)
182 	 */
183 	if (in6_ifaddr == 0 || in_ifaddr == 0)
184 		return EADDRNOTAVAIL;
185 
186 	if (inp->inp_lport != 0 || !IN6_IS_ADDR_UNSPECIFIED(&inp->inp_laddr6))
187 		return EINVAL;	/* If already bound, EINVAL! */
188 
189 	if ((so->so_options & (SO_REUSEADDR | SO_REUSEPORT)) == 0 &&
190 	    ((so->so_proto->pr_flags & PR_CONNREQUIRED) == 0 ||
191 	     (so->so_options & SO_ACCEPTCONN) == 0))
192 		wild |= INPLOOKUP_WILDCARD;
193 
194 	/*
195 	 * If I did get a sockaddr passed in...
196 	 */
197 	if (nam) {
198 		sin6 = mtod(nam, struct sockaddr_in6 *);
199 		if (nam->m_len != sizeof (*sin6))
200 			return EINVAL;
201 
202 		/*
203 		 * Unlike v4, I have no qualms about EAFNOSUPPORT if the
204 		 * wretched family is not filled in!
205 		 */
206 		if (sin6->sin6_family != AF_INET6)
207 			return EAFNOSUPPORT;
208 
209 		/* KAME hack: embed scopeid */
210 		if (in6_embedscope(&sin6->sin6_addr, sin6, inp, NULL) != 0)
211 			return EINVAL;
212 		/* this must be cleared for ifa_ifwithaddr() */
213 		sin6->sin6_scope_id = 0;
214 
215 		lport = sin6->sin6_port;
216 
217 		/* reject IPv4 mapped address, we have no support for it */
218 		if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr))
219 			return EADDRNOTAVAIL;
220 
221 		if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) {
222 			/*
223 			 * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
224 			 * allow complete duplication of binding if
225 			 * SO_REUSEPORT is set, or if SO_REUSEADDR is set
226 			 * and a multicast address is bound on both
227 			 * new and duplicated sockets.
228 			 */
229 			if (so->so_options & SO_REUSEADDR)
230 				reuseport = SO_REUSEADDR | SO_REUSEPORT;
231 		} else if (!IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
232 			struct ifaddr *ia = NULL;
233 
234 			sin6->sin6_port = 0;  /*
235 					       * Yechhhh, because of upcoming
236 					       * call to ifa_ifwithaddr(), which
237 					       * does bcmp's over the PORTS as
238 					       * well.  (What about flow?)
239 					       */
240 			sin6->sin6_flowinfo = 0;
241 			if ((ia = ifa_ifwithaddr((struct sockaddr *)sin6))
242 			    == NULL)
243 				return EADDRNOTAVAIL;
244 
245 			/*
246 			 * bind to an anycast address might accidentally
247 			 * cause sending a packet with an anycast source
248 			 * address, so we forbid it.
249 			 *
250 			 * We should allow to bind to a deprecated address,
251 			 * since the application dare to use it.
252 			 * But, can we assume that they are careful enough
253 			 * to check if the address is deprecated or not?
254 			 * Maybe, as a safeguard, we should have a setsockopt
255 			 * flag to control the bind(2) behavior against
256 			 * deprecated addresses (default: forbid bind(2)).
257 			 */
258 			if (ia &&
259 			    ((struct in6_ifaddr *)ia)->ia6_flags &
260 			    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY|IN6_IFF_DETACHED))
261 				return (EADDRNOTAVAIL);
262 		}
263 		if (lport) {
264 			struct inpcb *t;
265 
266 			/*
267 			 * Question:  Do we wish to continue the Berkeley
268 			 * tradition of ports < IPPORT_RESERVED be only for
269 			 * root?
270 			 * Answer: For now yes, but IMHO, it should be REMOVED!
271 			 * OUCH: One other thing, is there no better way of
272 			 * finding a process for a socket instead of using
273 			 * curproc?  (Marked with BSD's {in,}famous XXX ?
274 			 */
275 			if (ntohs(lport) < IPPORT_RESERVED &&
276 			    (error = suser(p->p_ucred, &p->p_acflag)))
277 				return error;
278 
279 			t = in_pcblookup(head,
280 			    (struct in_addr *)&zeroin6_addr, 0,
281 			    (struct in_addr *)&sin6->sin6_addr, lport,
282 			    wild);
283 
284 			if (t && (reuseport & t->inp_socket->so_options) == 0)
285 				return EADDRINUSE;
286 		}
287 		inp->inp_laddr6 = sin6->sin6_addr;
288 
289 		if (!IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
290 			inp->inp_ipv6.ip6_flow = htonl(0x60000000) |
291 			    (sin6->sin6_flowinfo & htonl(0x0fffffff));
292 		}
293 	}
294 
295 	if (lport == 0) {
296 		error = in6_pcbsetport(&inp->inp_laddr6, inp, p);
297 		if (error != 0)
298 			return error;
299 	} else
300 		inp->inp_lport = lport;
301 
302 	in_pcbrehash(inp);
303 
304 	return 0;
305 }
306 
307 int
308 in6_pcbsetport(laddr, inp, p)
309 	struct in6_addr *laddr;
310 	struct inpcb *inp;
311 	struct proc *p;
312 {
313 	struct socket *so = inp->inp_socket;
314 	struct inpcbtable *table = inp->inp_table;
315 	u_int16_t first, last, old = 0;
316 	u_int16_t *lastport = &inp->inp_table->inpt_lastport;
317 	u_int16_t lport = 0;
318 	int count;
319 	int loopcount = 0;
320 	int wild = INPLOOKUP_IPV6;
321 	int error;
322 
323 	/* XXX we no longer support IPv4 mapped address, so no tweaks here */
324 
325 	if ((so->so_options & (SO_REUSEADDR|SO_REUSEPORT)) == 0 &&
326 	    ((so->so_proto->pr_flags & PR_CONNREQUIRED) == 0 ||
327 	     (so->so_options & SO_ACCEPTCONN) == 0))
328 		wild |= INPLOOKUP_WILDCARD;
329 
330 	if (inp->inp_flags & INP_HIGHPORT) {
331 		first = ipport_hifirstauto;	/* sysctl */
332 		last = ipport_hilastauto;
333 	} else if (inp->inp_flags & INP_LOWPORT) {
334 		if ((error = suser(p->p_ucred, &p->p_acflag)))
335 			return (EACCES);
336 		first = IPPORT_RESERVED-1; /* 1023 */
337 		last = 600;		   /* not IPPORT_RESERVED/2 */
338 	} else {
339 		first = ipport_firstauto;	/* sysctl */
340 		last  = ipport_lastauto;
341 	}
342 
343 	/*
344 	 * Simple check to ensure all ports are not used up causing
345 	 * a deadlock here.
346 	 *
347 	 * We split the two cases (up and down) so that the direction
348 	 * is not being tested on each round of the loop.
349 	 */
350 
351 portloop:
352 	if (first > last) {
353 		/*
354 		 * counting down
355 		 */
356 		if (loopcount == 0) {	/* only do this once. */
357 			old = first;
358 			first -= (arc4random() % (first - last));
359 		}
360 		count = first - last;
361 		*lastport = first;		/* restart each time */
362 
363 		do {
364 			if (count-- <= 0) {	/* completely used? */
365 				if (loopcount == 0) {
366 					last = old;
367 					loopcount++;
368 					goto portloop;
369 				}
370 				return (EADDRNOTAVAIL);
371 			}
372 			--*lastport;
373 			if (*lastport > first || *lastport < last)
374 				*lastport = first;
375 			lport = htons(*lastport);
376 		} while (in_baddynamic(*lastport, so->so_proto->pr_protocol) ||
377 		    in_pcblookup(table, &zeroin6_addr, 0,
378 		    &inp->inp_laddr6, lport, wild));
379 	} else {
380 		/*
381 		 * counting up
382 		 */
383 		if (loopcount == 0) {	/* only do this once. */
384 			old = first;
385 			first += (arc4random() % (last - first));
386 		}
387 		count = last - first;
388 		*lastport = first;		/* restart each time */
389 
390 		do {
391 			if (count-- <= 0) {	/* completely used? */
392 				if (loopcount == 0) {
393 					first = old;
394 					loopcount++;
395 					goto portloop;
396 				}
397 				return (EADDRNOTAVAIL);
398 			}
399 			++*lastport;
400 			if (*lastport < first || *lastport > last)
401 				*lastport = first;
402 			lport = htons(*lastport);
403 		} while (in_baddynamic(*lastport, so->so_proto->pr_protocol) ||
404 		    in_pcblookup(table, &zeroin6_addr, 0,
405 		    &inp->inp_laddr6, lport, wild));
406 	}
407 
408 	inp->inp_lport = lport;
409 	in_pcbrehash(inp);
410 
411 	return 0;
412 }
413 
414 /*
415  * Connect from a socket to a specified address.
416  * Both address and port must be specified in argument sin6.
417  * Eventually, flow labels will have to be dealt with here, as well.
418  *
419  * If don't have a local address for this socket yet,
420  * then pick one.
421  *
422  * I believe this has to be called at splnet().
423  */
424 int
425 in6_pcbconnect(inp, nam)
426 	struct inpcb *inp;
427 	struct mbuf *nam;
428 {
429 	struct in6_addr *in6a = NULL;
430 	struct sockaddr_in6 *sin6 = mtod(nam, struct sockaddr_in6 *);
431 	struct ifnet *ifp = NULL;	/* outgoing interface */
432 	int error = 0;
433 	struct in6_addr mapped;
434 	struct sockaddr_in6 tmp;
435 
436 	(void)&in6a;				/* XXX fool gcc */
437 
438 	if (nam->m_len != sizeof(*sin6))
439 		return (EINVAL);
440 	if (sin6->sin6_family != AF_INET6)
441 		return (EAFNOSUPPORT);
442 	if (sin6->sin6_port == 0)
443 		return (EADDRNOTAVAIL);
444 
445 	/* reject IPv4 mapped address, we have no support for it */
446 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr))
447 		return EADDRNOTAVAIL;
448 
449 	/* sanity check for mapped address case */
450 	if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) {
451 		if (IN6_IS_ADDR_UNSPECIFIED(&inp->inp_laddr6))
452 			inp->inp_laddr6.s6_addr16[5] = htons(0xffff);
453 		if (!IN6_IS_ADDR_V4MAPPED(&inp->inp_laddr6))
454 			return EINVAL;
455 	} else {
456 		if (IN6_IS_ADDR_V4MAPPED(&inp->inp_laddr6))
457 			return EINVAL;
458 	}
459 
460 	/* protect *sin6 from overwrites */
461 	tmp = *sin6;
462 	sin6 = &tmp;
463 
464 	/* KAME hack: embed scopeid */
465 	if (in6_embedscope(&sin6->sin6_addr, sin6, inp, &ifp) != 0)
466 		return EINVAL;
467 	/* this must be cleared for ifa_ifwithaddr() */
468 	sin6->sin6_scope_id = 0;
469 
470 	/* Source address selection. */
471 	if (IN6_IS_ADDR_V4MAPPED(&inp->inp_laddr6) &&
472 	    inp->inp_laddr6.s6_addr32[3] == 0) {
473 		struct sockaddr_in sin, *sinp;
474 
475 		bzero(&sin, sizeof(sin));
476 		sin.sin_len = sizeof(sin);
477 		sin.sin_family = AF_INET;
478 		bcopy(&sin6->sin6_addr.s6_addr32[3], &sin.sin_addr,
479 		    sizeof(sin.sin_addr));
480 		sinp = in_selectsrc(&sin, (struct route *)&inp->inp_route6,
481 		    inp->inp_socket->so_options, NULL, &error);
482 		if (sinp == 0) {
483 			if (error == 0)
484 				error = EADDRNOTAVAIL;
485 			return (error);
486 		}
487 		bzero(&mapped, sizeof(mapped));
488 		mapped.s6_addr16[5] = htons(0xffff);
489 		bcopy(&sinp->sin_addr, &mapped.s6_addr32[3], sizeof(sinp->sin_addr));
490 		in6a = &mapped;
491 	} else {
492 		/*
493 		 * XXX: in6_selectsrc might replace the bound local address
494 		 * with the address specified by setsockopt(IPV6_PKTINFO).
495 		 * Is it the intended behavior?
496 		 */
497 		in6a = in6_selectsrc(sin6, inp->inp_outputopts6,
498 		    inp->inp_moptions6, &inp->inp_route6, &inp->inp_laddr6,
499 		    &error);
500 		if (in6a == 0) {
501 			if (error == 0)
502 				error = EADDRNOTAVAIL;
503 			return (error);
504 		}
505 	}
506 	if (inp->inp_route6.ro_rt)
507 		ifp = inp->inp_route6.ro_rt->rt_ifp;
508 
509 	inp->inp_ipv6.ip6_hlim = (u_int8_t)in6_selecthlim(inp, ifp);
510 
511 	if (in_pcblookup(inp->inp_table, &sin6->sin6_addr, sin6->sin6_port,
512 	    IN6_IS_ADDR_UNSPECIFIED(&inp->inp_laddr6) ? in6a : &inp->inp_laddr6,
513 	    inp->inp_lport, INPLOOKUP_IPV6)) {
514 		return (EADDRINUSE);
515 	}
516 	if (IN6_IS_ADDR_UNSPECIFIED(&inp->inp_laddr6) ||
517 	    (IN6_IS_ADDR_V4MAPPED(&inp->inp_laddr6) &&
518 	     inp->inp_laddr6.s6_addr32[3] == 0)) {
519 		if (inp->inp_lport == 0)
520 			(void)in6_pcbbind(inp, (struct mbuf *)0);
521 		inp->inp_laddr6 = *in6a;
522 	}
523 	inp->inp_faddr6 = sin6->sin6_addr;
524 	inp->inp_fport = sin6->sin6_port;
525 	/*
526 	 * xxx kazu flowlabel is necessary for connect?
527 	 * but if this line is missing, the garbage value remains.
528 	 */
529 	inp->inp_ipv6.ip6_flow = sin6->sin6_flowinfo;
530 	in_pcbrehash(inp);
531 	return (0);
532 }
533 
534 /*
535  * Pass some notification to all connections of a protocol
536  * associated with address dst.  The local address and/or port numbers
537  * may be specified to limit the search.  The "usual action" will be
538  * taken, depending on the ctlinput cmd.  The caller must filter any
539  * cmds that are uninteresting (e.g., no error in the map).
540  * Call the protocol specific routine (if any) to report
541  * any errors for each matching socket.
542  *
543  * Also perform input-side security policy check
544  *    once PCB to be notified has been located.
545  *
546  * Must be called at splnet.
547  */
548 int
549 in6_pcbnotify(head, dst, fport_arg, src, lport_arg, cmd, cmdarg, notify)
550 	struct inpcbtable *head;
551 	struct sockaddr *dst, *src;
552 	uint fport_arg;
553 	uint lport_arg;
554 	int cmd;
555 	void *cmdarg;
556 	void (*notify)(struct inpcb *, int);
557 {
558 	struct inpcb *inp, *ninp;
559 	u_short fport = fport_arg, lport = lport_arg;
560 	struct sockaddr_in6 sa6_src, *sa6_dst;
561 	int errno, nmatch = 0;
562 	u_int32_t flowinfo;
563 
564 	if ((unsigned)cmd > PRC_NCMDS || dst->sa_family != AF_INET6)
565 		return 1;
566 
567 	sa6_dst = (struct sockaddr_in6 *)dst;
568 	if (IN6_IS_ADDR_UNSPECIFIED(&sa6_dst->sin6_addr))
569 		return 1;
570 	if (IN6_IS_ADDR_V4MAPPED(&sa6_dst->sin6_addr))
571 		printf("Huh?  Thought in6_pcbnotify() never got "
572 		       "called with mapped!\n");
573 
574 	/*
575 	 * note that src can be NULL when we get notify by local fragmentation.
576 	 */
577 	sa6_src = (src == NULL) ? sa6_any : *(struct sockaddr_in6 *)src;
578 	flowinfo = sa6_src.sin6_flowinfo;
579 
580 	/*
581 	 * Redirects go to all references to the destination,
582 	 * and use in_rtchange to invalidate the route cache.
583 	 * Dead host indications: also use in_rtchange to invalidate
584 	 * the cache, and deliver the error to all the sockets.
585 	 * Otherwise, if we have knowledge of the local port and address,
586 	 * deliver only to that socket.
587 	 */
588 	if (PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) {
589 		fport = 0;
590 		lport = 0;
591 		sa6_src.sin6_addr = in6addr_any;
592 
593 		if (cmd != PRC_HOSTDEAD)
594 			notify = in_rtchange;
595 	}
596 	errno = inet6ctlerrmap[cmd];
597 
598 	for (inp = head->inpt_queue.cqh_first;
599 	     inp != (struct inpcb *)&head->inpt_queue; inp = ninp) {
600 		ninp = inp->inp_queue.cqe_next;
601 
602 		if ((inp->inp_flags & INP_IPV6) == 0)
603 			continue;
604 
605 		/*
606 		 * Under the following condition, notify of redirects
607 		 * to the pcb, without making address matches against inpcb.
608 		 * - redirect notification is arrived.
609 		 * - the inpcb is unconnected.
610 		 * - the inpcb is caching !RTF_HOST routing entry.
611 		 * - the ICMPv6 notification is from the gateway cached in the
612 		 *   inpcb.  i.e. ICMPv6 notification is from nexthop gateway
613 		 *   the inpcb used very recently.
614 		 *
615 		 * This is to improve interaction between netbsd/openbsd
616 		 * redirect handling code, and inpcb route cache code.
617 		 * without the clause, !RTF_HOST routing entry (which carries
618 		 * gateway used by inpcb right before the ICMPv6 redirect)
619 		 * will be cached forever in unconnected inpcb.
620 		 *
621 		 * There still is a question regarding to what is TRT:
622 		 * - On bsdi/freebsd, RTF_HOST (cloned) routing entry will be
623 		 *   generated on packet output.  inpcb will always cache
624 		 *   RTF_HOST routing entry so there's no need for the clause
625 		 *   (ICMPv6 redirect will update RTF_HOST routing entry,
626 		 *   and inpcb is caching it already).
627 		 *   However, bsdi/freebsd are vulnerable to local DoS attacks
628 		 *   due to the cloned routing entries.
629 		 * - Specwise, "destination cache" is mentioned in RFC2461.
630 		 *   Jinmei says that it implies bsdi/freebsd behavior, itojun
631 		 *   is not really convinced.
632 		 * - Having hiwat/lowat on # of cloned host route (redirect/
633 		 *   pmtud) may be a good idea.  netbsd/openbsd has it.  see
634 		 *   icmp6_mtudisc_update().
635 		 */
636 		if ((PRC_IS_REDIRECT(cmd) || cmd == PRC_HOSTDEAD) &&
637 		    IN6_IS_ADDR_UNSPECIFIED(&inp->inp_laddr6) &&
638 		    inp->inp_route.ro_rt &&
639 		    !(inp->inp_route.ro_rt->rt_flags & RTF_HOST)) {
640 			struct sockaddr_in6 *dst6;
641 
642 			dst6 = (struct sockaddr_in6 *)&inp->inp_route.ro_dst;
643 			if (IN6_ARE_ADDR_EQUAL(&dst6->sin6_addr,
644 			    &sa6_dst->sin6_addr))
645 				goto do_notify;
646 		}
647 
648 		/*
649 		 * Detect if we should notify the error. If no source and
650 		 * destination ports are specified, but non-zero flowinfo and
651 		 * local address match, notify the error. This is the case
652 		 * when the error is delivered with an encrypted buffer
653 		 * by ESP. Otherwise, just compare addresses and ports
654 		 * as usual.
655 		 */
656 		if (lport == 0 && fport == 0 && flowinfo &&
657 		    inp->inp_socket != NULL &&
658 		    flowinfo == (inp->inp_flowinfo & IPV6_FLOWLABEL_MASK) &&
659 		    IN6_ARE_ADDR_EQUAL(&inp->inp_laddr6, &sa6_src.sin6_addr))
660 			goto do_notify;
661 		else if (!IN6_ARE_ADDR_EQUAL(&inp->inp_faddr6,
662 					     &sa6_dst->sin6_addr) ||
663 			 inp->inp_socket == 0 ||
664 			 (lport && inp->inp_lport != lport) ||
665 			 (!IN6_IS_ADDR_UNSPECIFIED(&sa6_src.sin6_addr) &&
666 			  !IN6_ARE_ADDR_EQUAL(&inp->inp_laddr6,
667 					      &sa6_src.sin6_addr)) ||
668 			 (fport && inp->inp_fport != fport)) {
669 			continue;
670 		}
671 	  do_notify:
672 		nmatch++;
673 		if (notify)
674 			(*notify)(inp, errno);
675 	}
676 	return 0;
677 }
678 
679 /*
680  * Get the local address/port, and put it in a sockaddr_in6.
681  * This services the getsockname(2) call.
682  */
683 int
684 in6_setsockaddr(inp, nam)
685 	struct inpcb *inp;
686 	struct mbuf *nam;
687 {
688 	struct sockaddr_in6 *sin6;
689 
690 	nam->m_len = sizeof(struct sockaddr_in6);
691 	sin6 = mtod(nam,struct sockaddr_in6 *);
692 
693 	bzero ((caddr_t)sin6,sizeof(struct sockaddr_in6));
694 	sin6->sin6_family = AF_INET6;
695 	sin6->sin6_len = sizeof(struct sockaddr_in6);
696 	sin6->sin6_port = inp->inp_lport;
697 	sin6->sin6_addr = inp->inp_laddr6;
698 	/* KAME hack: recover scopeid */
699 	(void)in6_recoverscope(sin6, &inp->inp_laddr6, NULL);
700 
701 	return 0;
702 }
703 
704 /*
705  * Get the foreign address/port, and put it in a sockaddr_in6.
706  * This services the getpeername(2) call.
707  */
708 int
709 in6_setpeeraddr(inp, nam)
710 	struct inpcb *inp;
711 	struct mbuf *nam;
712 {
713 	struct sockaddr_in6 *sin6;
714 
715 	nam->m_len = sizeof(struct sockaddr_in6);
716 	sin6 = mtod(nam,struct sockaddr_in6 *);
717 
718 	bzero ((caddr_t)sin6,sizeof(struct sockaddr_in6));
719 	sin6->sin6_family = AF_INET6;
720 	sin6->sin6_len = sizeof(struct sockaddr_in6);
721 	sin6->sin6_port = inp->inp_fport;
722 	sin6->sin6_addr = inp->inp_faddr6;
723 	/* KAME hack: recover scopeid */
724 	(void)in6_recoverscope(sin6, &inp->inp_faddr6, NULL);
725 
726 	return 0;
727 }
728