xref: /netbsd-src/sys/netinet6/icmp6.c (revision 08c81a9c2dc8c7300e893321eb65c0925d60871c)
1 /*	$NetBSD: icmp6.c,v 1.86 2002/09/11 02:46:44 itojun Exp $	*/
2 /*	$KAME: icmp6.c,v 1.217 2001/06/20 15:03:29 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  * Copyright (c) 1982, 1986, 1988, 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  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
66  */
67 
68 #include <sys/cdefs.h>
69 __KERNEL_RCSID(0, "$NetBSD: icmp6.c,v 1.86 2002/09/11 02:46:44 itojun Exp $");
70 
71 #include "opt_inet.h"
72 #include "opt_ipsec.h"
73 
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/malloc.h>
77 #include <sys/mbuf.h>
78 #include <sys/protosw.h>
79 #include <sys/socket.h>
80 #include <sys/socketvar.h>
81 #include <sys/time.h>
82 #include <sys/kernel.h>
83 #include <sys/syslog.h>
84 #include <sys/domain.h>
85 #include <sys/sysctl.h>
86 
87 #include <net/if.h>
88 #include <net/route.h>
89 #include <net/if_dl.h>
90 #include <net/if_types.h>
91 
92 #include <netinet/in.h>
93 #include <netinet/in_var.h>
94 #include <netinet/ip6.h>
95 #include <netinet6/ip6_var.h>
96 #include <netinet/icmp6.h>
97 #include <netinet6/mld6_var.h>
98 #include <netinet6/in6_pcb.h>
99 #include <netinet6/nd6.h>
100 #include <netinet6/in6_ifattach.h>
101 #include <netinet6/ip6protosw.h>
102 
103 #ifdef IPSEC
104 #include <netinet6/ipsec.h>
105 #include <netkey/key.h>
106 #endif
107 
108 #include "faith.h"
109 #if defined(NFAITH) && 0 < NFAITH
110 #include <net/if_faith.h>
111 #endif
112 
113 #include <net/net_osdep.h>
114 
115 extern struct domain inet6domain;
116 
117 struct icmp6stat icmp6stat;
118 
119 extern struct in6pcb rawin6pcb;
120 extern int icmp6errppslim;
121 static int icmp6errpps_count = 0;
122 static struct timeval icmp6errppslim_last;
123 extern int icmp6_nodeinfo;
124 
125 /*
126  * List of callbacks to notify when Path MTU changes are made.
127  */
128 struct icmp6_mtudisc_callback {
129 	LIST_ENTRY(icmp6_mtudisc_callback) mc_list;
130 	void (*mc_func) __P((struct in6_addr *));
131 };
132 
133 LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks =
134     LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks);
135 
136 static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL;
137 extern int pmtu_expire;
138 
139 /* XXX do these values make any sense? */
140 static int icmp6_mtudisc_hiwat = 1280;
141 static int icmp6_mtudisc_lowat = 256;
142 
143 /*
144  * keep track of # of redirect routes.
145  */
146 static struct rttimer_queue *icmp6_redirect_timeout_q = NULL;
147 
148 /* XXX experimental, turned off */
149 static int icmp6_redirect_hiwat = -1;
150 static int icmp6_redirect_lowat = -1;
151 
152 static void icmp6_errcount __P((struct icmp6errstat *, int, int));
153 static int icmp6_rip6_input __P((struct mbuf **, int));
154 static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int));
155 static const char *icmp6_redirect_diag __P((struct in6_addr *,
156 	struct in6_addr *, struct in6_addr *));
157 static struct mbuf *ni6_input __P((struct mbuf *, int));
158 static struct mbuf *ni6_nametodns __P((const char *, int, int));
159 static int ni6_dnsmatch __P((const char *, int, const char *, int));
160 static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
161 			  struct ifnet **, char *));
162 static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
163 				struct ifnet *, int));
164 static int icmp6_notify_error __P((struct mbuf *, int, int, int));
165 static struct rtentry *icmp6_mtudisc_clone __P((struct sockaddr *));
166 static void icmp6_mtudisc_timeout __P((struct rtentry *, struct rttimer *));
167 static void icmp6_redirect_timeout __P((struct rtentry *, struct rttimer *));
168 
169 void
170 icmp6_init()
171 {
172 	mld6_init();
173 	icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire);
174 	icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout);
175 }
176 
177 static void
178 icmp6_errcount(stat, type, code)
179 	struct icmp6errstat *stat;
180 	int type, code;
181 {
182 	switch (type) {
183 	case ICMP6_DST_UNREACH:
184 		switch (code) {
185 		case ICMP6_DST_UNREACH_NOROUTE:
186 			stat->icp6errs_dst_unreach_noroute++;
187 			return;
188 		case ICMP6_DST_UNREACH_ADMIN:
189 			stat->icp6errs_dst_unreach_admin++;
190 			return;
191 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
192 			stat->icp6errs_dst_unreach_beyondscope++;
193 			return;
194 		case ICMP6_DST_UNREACH_ADDR:
195 			stat->icp6errs_dst_unreach_addr++;
196 			return;
197 		case ICMP6_DST_UNREACH_NOPORT:
198 			stat->icp6errs_dst_unreach_noport++;
199 			return;
200 		}
201 		break;
202 	case ICMP6_PACKET_TOO_BIG:
203 		stat->icp6errs_packet_too_big++;
204 		return;
205 	case ICMP6_TIME_EXCEEDED:
206 		switch (code) {
207 		case ICMP6_TIME_EXCEED_TRANSIT:
208 			stat->icp6errs_time_exceed_transit++;
209 			return;
210 		case ICMP6_TIME_EXCEED_REASSEMBLY:
211 			stat->icp6errs_time_exceed_reassembly++;
212 			return;
213 		}
214 		break;
215 	case ICMP6_PARAM_PROB:
216 		switch (code) {
217 		case ICMP6_PARAMPROB_HEADER:
218 			stat->icp6errs_paramprob_header++;
219 			return;
220 		case ICMP6_PARAMPROB_NEXTHEADER:
221 			stat->icp6errs_paramprob_nextheader++;
222 			return;
223 		case ICMP6_PARAMPROB_OPTION:
224 			stat->icp6errs_paramprob_option++;
225 			return;
226 		}
227 		break;
228 	case ND_REDIRECT:
229 		stat->icp6errs_redirect++;
230 		return;
231 	}
232 	stat->icp6errs_unknown++;
233 }
234 
235 /*
236  * Register a Path MTU Discovery callback.
237  */
238 void
239 icmp6_mtudisc_callback_register(func)
240 	void (*func) __P((struct in6_addr *));
241 {
242 	struct icmp6_mtudisc_callback *mc;
243 
244 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
245 	     mc = LIST_NEXT(mc, mc_list)) {
246 		if (mc->mc_func == func)
247 			return;
248 	}
249 
250 	mc = malloc(sizeof(*mc), M_PCB, M_NOWAIT);
251 	if (mc == NULL)
252 		panic("icmp6_mtudisc_callback_register");
253 
254 	mc->mc_func = func;
255 	LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list);
256 }
257 
258 /*
259  * Generate an error packet of type error in response to bad IP6 packet.
260  */
261 void
262 icmp6_error(m, type, code, param)
263 	struct mbuf *m;
264 	int type, code, param;
265 {
266 	struct ip6_hdr *oip6, *nip6;
267 	struct icmp6_hdr *icmp6;
268 	u_int preplen;
269 	int off;
270 	int nxt;
271 
272 	icmp6stat.icp6s_error++;
273 
274 	/* count per-type-code statistics */
275 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
276 
277 	if (m->m_flags & M_DECRYPTED) {
278 		icmp6stat.icp6s_canterror++;
279 		goto freeit;
280 	}
281 
282 #ifndef PULLDOWN_TEST
283 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
284 #else
285 	if (m->m_len < sizeof(struct ip6_hdr)) {
286 		m = m_pullup(m, sizeof(struct ip6_hdr));
287 		if (m == NULL)
288 			return;
289 	}
290 #endif
291 	oip6 = mtod(m, struct ip6_hdr *);
292 
293 	/*
294 	 * If the destination address of the erroneous packet is a multicast
295 	 * address, or the packet was sent using link-layer multicast,
296 	 * we should basically suppress sending an error (RFC 2463, Section
297 	 * 2.4).
298 	 * We have two exceptions (the item e.2 in that section):
299 	 * - the Pakcet Too Big message can be sent for path MTU discovery.
300 	 * - the Parameter Problem Message that can be allowed an icmp6 error
301 	 *   in the option type field.  This check has been done in
302 	 *   ip6_unknown_opt(), so we can just check the type and code.
303 	 */
304 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
305 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
306 	    (type != ICMP6_PACKET_TOO_BIG &&
307 	     (type != ICMP6_PARAM_PROB ||
308 	      code != ICMP6_PARAMPROB_OPTION)))
309 		goto freeit;
310 
311 	/*
312 	 * RFC 2463, 2.4 (e.5): source address check.
313 	 * XXX: the case of anycast source?
314 	 */
315 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
316 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
317 		goto freeit;
318 
319 	/*
320 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
321 	 * don't do it.
322 	 */
323 	nxt = -1;
324 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
325 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
326 		struct icmp6_hdr *icp;
327 
328 #ifndef PULLDOWN_TEST
329 		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
330 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
331 #else
332 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
333 			sizeof(*icp));
334 		if (icp == NULL) {
335 			icmp6stat.icp6s_tooshort++;
336 			return;
337 		}
338 #endif
339 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
340 		    icp->icmp6_type == ND_REDIRECT) {
341 			/*
342 			 * ICMPv6 error
343 			 * Special case: for redirect (which is
344 			 * informational) we must not send icmp6 error.
345 			 */
346 			icmp6stat.icp6s_canterror++;
347 			goto freeit;
348 		} else {
349 			/* ICMPv6 informational - send the error */
350 		}
351 	}
352 #if 0 /* controversial */
353 	else if (off >= 0 && nxt == IPPROTO_ESP) {
354 		/*
355 		 * It could be ICMPv6 error inside ESP.  Take a safer side,
356 		 * don't respond.
357 		 */
358 		icmp6stat.icp6s_canterror++;
359 		goto freeit;
360 	}
361 #endif
362 	else {
363 		/* non-ICMPv6 - send the error */
364 	}
365 
366 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
367 
368 	/* Finally, do rate limitation check. */
369 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
370 		icmp6stat.icp6s_toofreq++;
371 		goto freeit;
372 	}
373 
374 	/*
375 	 * OK, ICMP6 can be generated.
376 	 */
377 
378 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
379 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
380 
381 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
382 	M_PREPEND(m, preplen, M_DONTWAIT);
383 	if (m && m->m_len < preplen)
384 		m = m_pullup(m, preplen);
385 	if (m == NULL) {
386 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
387 		return;
388 	}
389 
390 	nip6 = mtod(m, struct ip6_hdr *);
391 	nip6->ip6_src  = oip6->ip6_src;
392 	nip6->ip6_dst  = oip6->ip6_dst;
393 
394 	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src))
395 		oip6->ip6_src.s6_addr16[1] = 0;
396 	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst))
397 		oip6->ip6_dst.s6_addr16[1] = 0;
398 
399 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
400 	icmp6->icmp6_type = type;
401 	icmp6->icmp6_code = code;
402 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
403 
404 	/*
405 	 * icmp6_reflect() is designed to be in the input path.
406 	 * icmp6_error() can be called from both input and outut path,
407 	 * and if we are in output path rcvif could contain bogus value.
408 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
409 	 * information in ip header (nip6).
410 	 */
411 	m->m_pkthdr.rcvif = NULL;
412 
413 	icmp6stat.icp6s_outhist[type]++;
414 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
415 
416 	return;
417 
418   freeit:
419 	/*
420 	 * If we can't tell wheter or not we can generate ICMP6, free it.
421 	 */
422 	m_freem(m);
423 }
424 
425 /*
426  * Process a received ICMP6 message.
427  */
428 int
429 icmp6_input(mp, offp, proto)
430 	struct mbuf **mp;
431 	int *offp, proto;
432 {
433 	struct mbuf *m = *mp, *n;
434 	struct ip6_hdr *ip6, *nip6;
435 	struct icmp6_hdr *icmp6, *nicmp6;
436 	int off = *offp;
437 	int icmp6len = m->m_pkthdr.len - *offp;
438 	int code, sum, noff;
439 
440 	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
441 
442 #ifndef PULLDOWN_TEST
443 	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
444 	/* m might change if M_LOOP.  So, call mtod after this */
445 #endif
446 
447 	/*
448 	 * Locate icmp6 structure in mbuf, and check
449 	 * that not corrupted and of at least minimum length
450 	 */
451 
452 	ip6 = mtod(m, struct ip6_hdr *);
453 	if (icmp6len < sizeof(struct icmp6_hdr)) {
454 		icmp6stat.icp6s_tooshort++;
455 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
456 		goto freeit;
457 	}
458 
459 	/*
460 	 * calculate the checksum
461 	 */
462 #ifndef PULLDOWN_TEST
463 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
464 #else
465 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
466 	if (icmp6 == NULL) {
467 		icmp6stat.icp6s_tooshort++;
468 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
469 		return IPPROTO_DONE;
470 	}
471 #endif
472 	KASSERT(IP6_HDR_ALIGNED_P(icmp6));
473 	code = icmp6->icmp6_code;
474 
475 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
476 		nd6log((LOG_ERR,
477 		    "ICMP6 checksum error(%d|%x) %s\n",
478 		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
479 		icmp6stat.icp6s_checksum++;
480 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
481 		goto freeit;
482 	}
483 
484 #if defined(NFAITH) && 0 < NFAITH
485 	if (faithprefix(&ip6->ip6_dst)) {
486 		/*
487 		 * Deliver very specific ICMP6 type only.
488 		 * This is important to deilver TOOBIG.  Otherwise PMTUD
489 		 * will not work.
490 		 */
491 		switch (icmp6->icmp6_type) {
492 		case ICMP6_DST_UNREACH:
493 		case ICMP6_PACKET_TOO_BIG:
494 		case ICMP6_TIME_EXCEEDED:
495 			break;
496 		default:
497 			goto freeit;
498 		}
499 	}
500 #endif
501 
502 	icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
503 
504 	switch (icmp6->icmp6_type) {
505 	case ICMP6_DST_UNREACH:
506 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
507 		switch (code) {
508 		case ICMP6_DST_UNREACH_NOROUTE:
509 			code = PRC_UNREACH_NET;
510 			break;
511 		case ICMP6_DST_UNREACH_ADMIN:
512 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
513 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
514 			break;
515 		case ICMP6_DST_UNREACH_ADDR:
516 			code = PRC_HOSTDEAD;
517 			break;
518 #ifdef COMPAT_RFC1885
519 		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
520 			code = PRC_UNREACH_SRCFAIL;
521 			break;
522 #else
523 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
524 			/* I mean "source address was incorrect." */
525 			code = PRC_UNREACH_NET;
526 			break;
527 #endif
528 		case ICMP6_DST_UNREACH_NOPORT:
529 			code = PRC_UNREACH_PORT;
530 			break;
531 		default:
532 			goto badcode;
533 		}
534 		goto deliver;
535 		break;
536 
537 	case ICMP6_PACKET_TOO_BIG:
538 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
539 		if (code != 0)
540 			goto badcode;
541 
542 		code = PRC_MSGSIZE;
543 
544 		/*
545 		 * Updating the path MTU will be done after examining
546 		 * intermediate extension headers.
547 		 */
548 		goto deliver;
549 		break;
550 
551 	case ICMP6_TIME_EXCEEDED:
552 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
553 		switch (code) {
554 		case ICMP6_TIME_EXCEED_TRANSIT:
555 			code = PRC_TIMXCEED_INTRANS;
556 			break;
557 		case ICMP6_TIME_EXCEED_REASSEMBLY:
558 			code = PRC_TIMXCEED_REASS;
559 			break;
560 		default:
561 			goto badcode;
562 		}
563 		goto deliver;
564 		break;
565 
566 	case ICMP6_PARAM_PROB:
567 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
568 		switch (code) {
569 		case ICMP6_PARAMPROB_NEXTHEADER:
570 			code = PRC_UNREACH_PROTOCOL;
571 			break;
572 		case ICMP6_PARAMPROB_HEADER:
573 		case ICMP6_PARAMPROB_OPTION:
574 			code = PRC_PARAMPROB;
575 			break;
576 		default:
577 			goto badcode;
578 		}
579 		goto deliver;
580 		break;
581 
582 	case ICMP6_ECHO_REQUEST:
583 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
584 		if (code != 0)
585 			goto badcode;
586 		/*
587 		 * Copy mbuf to send to two data paths: userland socket(s),
588 		 * and to the querier (echo reply).
589 		 * m: a copy for socket, n: a copy for querier
590 		 */
591 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
592 			/* Give up local */
593 			n = m;
594 			m = NULL;
595 			goto deliverecho;
596 		}
597 		/*
598 		 * If the first mbuf is shared, or the first mbuf is too short,
599 		 * copy the first part of the data into a fresh mbuf.
600 		 * Otherwise, we will wrongly overwrite both copies.
601 		 */
602 		if ((n->m_flags & M_EXT) != 0 ||
603 		    n->m_len < off + sizeof(struct icmp6_hdr)) {
604 			struct mbuf *n0 = n;
605 			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
606 
607 			/*
608 			 * Prepare an internal mbuf.  m_pullup() doesn't
609 			 * always copy the length we specified.
610 			 */
611 			if (maxlen >= MCLBYTES) {
612 				/* Give up remote */
613 				m_freem(n0);
614 				break;
615 			}
616 			MGETHDR(n, M_DONTWAIT, n0->m_type);
617 			if (n && maxlen >= MHLEN) {
618 				MCLGET(n, M_DONTWAIT);
619 				if ((n->m_flags & M_EXT) == 0) {
620 					m_free(n);
621 					n = NULL;
622 				}
623 			}
624 			if (n == NULL) {
625 				/* Give up local */
626 				m_freem(n0);
627 				n = m;
628 				m = NULL;
629 				goto deliverecho;
630 			}
631 			M_COPY_PKTHDR(n, n0);
632 			/*
633 			 * Copy IPv6 and ICMPv6 only.
634 			 */
635 			nip6 = mtod(n, struct ip6_hdr *);
636 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
637 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
638 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
639 			noff = sizeof(struct ip6_hdr);
640 			n->m_len = noff + sizeof(struct icmp6_hdr);
641 			/*
642 			 * Adjust mbuf.  ip6_plen will be adjusted in
643 			 * ip6_output().
644 			 * n->m_pkthdr.len == n0->m_pkthdr.len at this point.
645 			 */
646 			n->m_pkthdr.len += noff + sizeof(struct icmp6_hdr);
647 			n->m_pkthdr.len -= (off + sizeof(struct icmp6_hdr));
648 			m_adj(n0, off + sizeof(struct icmp6_hdr));
649 			n->m_next = n0;
650 			n0->m_flags &= ~M_PKTHDR;
651 		} else {
652 	 deliverecho:
653 			nip6 = mtod(n, struct ip6_hdr *);
654 			nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
655 			noff = off;
656 		}
657 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
658 		nicmp6->icmp6_code = 0;
659 		if (n) {
660 			icmp6stat.icp6s_reflect++;
661 			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
662 			icmp6_reflect(n, noff);
663 		}
664 		if (!m)
665 			goto freeit;
666 		break;
667 
668 	case ICMP6_ECHO_REPLY:
669 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
670 		if (code != 0)
671 			goto badcode;
672 		break;
673 
674 	case MLD6_LISTENER_QUERY:
675 	case MLD6_LISTENER_REPORT:
676 		if (icmp6len < sizeof(struct mld6_hdr))
677 			goto badlen;
678 		if (icmp6->icmp6_type == MLD6_LISTENER_QUERY) /* XXX: ugly... */
679 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
680 		else
681 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
682 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
683 			/* give up local */
684 			mld6_input(m, off);
685 			m = NULL;
686 			goto freeit;
687 		}
688 		mld6_input(n, off);
689 		/* m stays. */
690 		break;
691 
692 	case MLD6_LISTENER_DONE:
693 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
694 		if (icmp6len < sizeof(struct mld6_hdr))	/* necessary? */
695 			goto badlen;
696 		break;		/* nothing to be done in kernel */
697 
698 	case MLD6_MTRACE_RESP:
699 	case MLD6_MTRACE:
700 		/* XXX: these two are experimental.  not officially defined. */
701 		/* XXX: per-interface statistics? */
702 		break;		/* just pass it to applications */
703 
704 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
705 	    {
706 		enum { WRU, FQDN } mode;
707 
708 		if (!icmp6_nodeinfo)
709 			break;
710 
711 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
712 			mode = WRU;
713 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
714 			mode = FQDN;
715 		else
716 			goto badlen;
717 
718 		if (mode == FQDN) {
719 #ifndef PULLDOWN_TEST
720 			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
721 					 IPPROTO_DONE);
722 #endif
723 			n = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
724 			if (n)
725 				n = ni6_input(n, off);
726 			/* XXX meaningless if n == NULL */
727 			noff = sizeof(struct ip6_hdr);
728 		} else {
729 			u_char *p;
730 			int maxlen, maxhlen;
731 
732 			if ((icmp6_nodeinfo & 5) != 5)
733 				break;
734 
735 			if (code != 0)
736 				goto badcode;
737 			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
738 			if (maxlen >= MCLBYTES) {
739 				/* Give up remote */
740 				break;
741 			}
742 			MGETHDR(n, M_DONTWAIT, m->m_type);
743 			if (n && maxlen > MHLEN) {
744 				MCLGET(n, M_DONTWAIT);
745 				if ((n->m_flags & M_EXT) == 0) {
746 					m_free(n);
747 					n = NULL;
748 				}
749 			}
750 			if (n == NULL) {
751 				/* Give up remote */
752 				break;
753 			}
754 			n->m_pkthdr.rcvif = NULL;
755 			n->m_len = 0;
756 			maxhlen = M_TRAILINGSPACE(n) - maxlen;
757 			if (maxhlen > hostnamelen)
758 				maxhlen = hostnamelen;
759 			/*
760 			 * Copy IPv6 and ICMPv6 only.
761 			 */
762 			nip6 = mtod(n, struct ip6_hdr *);
763 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
764 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
765 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
766 			p = (u_char *)(nicmp6 + 1);
767 			bzero(p, 4);
768 			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
769 			noff = sizeof(struct ip6_hdr);
770 			M_COPY_PKTHDR(n, m); /* just for rcvif */
771 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
772 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
773 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
774 			nicmp6->icmp6_code = 0;
775 		}
776 #undef hostnamelen
777 		if (n) {
778 			icmp6stat.icp6s_reflect++;
779 			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
780 			icmp6_reflect(n, noff);
781 		}
782 		break;
783 	    }
784 
785 	case ICMP6_WRUREPLY:
786 		if (code != 0)
787 			goto badcode;
788 		break;
789 
790 	case ND_ROUTER_SOLICIT:
791 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
792 		if (code != 0)
793 			goto badcode;
794 		if (icmp6len < sizeof(struct nd_router_solicit))
795 			goto badlen;
796 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
797 			/* give up local */
798 			nd6_rs_input(m, off, icmp6len);
799 			m = NULL;
800 			goto freeit;
801 		}
802 		nd6_rs_input(n, off, icmp6len);
803 		/* m stays. */
804 		break;
805 
806 	case ND_ROUTER_ADVERT:
807 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
808 		if (code != 0)
809 			goto badcode;
810 		if (icmp6len < sizeof(struct nd_router_advert))
811 			goto badlen;
812 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
813 			/* give up local */
814 			nd6_ra_input(m, off, icmp6len);
815 			m = NULL;
816 			goto freeit;
817 		}
818 		nd6_ra_input(n, off, icmp6len);
819 		/* m stays. */
820 		break;
821 
822 	case ND_NEIGHBOR_SOLICIT:
823 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
824 		if (code != 0)
825 			goto badcode;
826 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
827 			goto badlen;
828 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
829 			/* give up local */
830 			nd6_ns_input(m, off, icmp6len);
831 			m = NULL;
832 			goto freeit;
833 		}
834 		nd6_ns_input(n, off, icmp6len);
835 		/* m stays. */
836 		break;
837 
838 	case ND_NEIGHBOR_ADVERT:
839 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
840 		if (code != 0)
841 			goto badcode;
842 		if (icmp6len < sizeof(struct nd_neighbor_advert))
843 			goto badlen;
844 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
845 			/* give up local */
846 			nd6_na_input(m, off, icmp6len);
847 			m = NULL;
848 			goto freeit;
849 		}
850 		nd6_na_input(n, off, icmp6len);
851 		/* m stays. */
852 		break;
853 
854 	case ND_REDIRECT:
855 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
856 		if (code != 0)
857 			goto badcode;
858 		if (icmp6len < sizeof(struct nd_redirect))
859 			goto badlen;
860 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
861 			/* give up local */
862 			icmp6_redirect_input(m, off);
863 			m = NULL;
864 			goto freeit;
865 		}
866 		icmp6_redirect_input(n, off);
867 		/* m stays. */
868 		break;
869 
870 	case ICMP6_ROUTER_RENUMBERING:
871 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
872 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
873 			goto badcode;
874 		if (icmp6len < sizeof(struct icmp6_router_renum))
875 			goto badlen;
876 		break;
877 
878 	default:
879 		nd6log((LOG_DEBUG,
880 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
881 		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
882 		    ip6_sprintf(&ip6->ip6_dst),
883 		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
884 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
885 			/* ICMPv6 error: MUST deliver it by spec... */
886 			code = PRC_NCMDS;
887 			/* deliver */
888 		} else {
889 			/* ICMPv6 informational: MUST not deliver */
890 			break;
891 		}
892 	deliver:
893 		if (icmp6_notify_error(m, off, icmp6len, code)) {
894 			/* In this case, m should've been freed. */
895 			return (IPPROTO_DONE);
896 		}
897 		break;
898 
899 	badcode:
900 		icmp6stat.icp6s_badcode++;
901 		break;
902 
903 	badlen:
904 		icmp6stat.icp6s_badlen++;
905 		break;
906 	}
907 
908 	/* deliver the packet to appropriate sockets */
909 	icmp6_rip6_input(&m, *offp);
910 
911 	return IPPROTO_DONE;
912 
913  freeit:
914 	m_freem(m);
915 	return IPPROTO_DONE;
916 }
917 
918 static int
919 icmp6_notify_error(m, off, icmp6len, code)
920 	struct mbuf *m;
921 	int off, icmp6len;
922 {
923 	struct icmp6_hdr *icmp6;
924 	struct ip6_hdr *eip6;
925 	u_int32_t notifymtu;
926 	struct sockaddr_in6 icmp6src, icmp6dst;
927 
928 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
929 		icmp6stat.icp6s_tooshort++;
930 		goto freeit;
931 	}
932 #ifndef PULLDOWN_TEST
933 	IP6_EXTHDR_CHECK(m, off,
934 			 sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr),
935 			 -1);
936 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
937 #else
938 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
939 		       sizeof(*icmp6) + sizeof(struct ip6_hdr));
940 	if (icmp6 == NULL) {
941 		icmp6stat.icp6s_tooshort++;
942 		return (-1);
943 	}
944 #endif
945 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
946 
947 	/* Detect the upper level protocol */
948 	{
949 		void (*ctlfunc) __P((int, struct sockaddr *, void *));
950 		u_int8_t nxt = eip6->ip6_nxt;
951 		int eoff = off + sizeof(struct icmp6_hdr) +
952 			sizeof(struct ip6_hdr);
953 		struct ip6ctlparam ip6cp;
954 		struct in6_addr *finaldst = NULL;
955 		int icmp6type = icmp6->icmp6_type;
956 		struct ip6_frag *fh;
957 		struct ip6_rthdr *rth;
958 		struct ip6_rthdr0 *rth0;
959 		int rthlen;
960 
961 		while (1) { /* XXX: should avoid infinite loop explicitly? */
962 			struct ip6_ext *eh;
963 
964 			switch (nxt) {
965 			case IPPROTO_HOPOPTS:
966 			case IPPROTO_DSTOPTS:
967 			case IPPROTO_AH:
968 #ifndef PULLDOWN_TEST
969 				IP6_EXTHDR_CHECK(m, 0, eoff +
970 						 sizeof(struct ip6_ext),
971 						 -1);
972 				eh = (struct ip6_ext *)(mtod(m, caddr_t)
973 							+ eoff);
974 #else
975 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
976 					       eoff, sizeof(*eh));
977 				if (eh == NULL) {
978 					icmp6stat.icp6s_tooshort++;
979 					return (-1);
980 				}
981 #endif
982 
983 				if (nxt == IPPROTO_AH)
984 					eoff += (eh->ip6e_len + 2) << 2;
985 				else
986 					eoff += (eh->ip6e_len + 1) << 3;
987 				nxt = eh->ip6e_nxt;
988 				break;
989 			case IPPROTO_ROUTING:
990 				/*
991 				 * When the erroneous packet contains a
992 				 * routing header, we should examine the
993 				 * header to determine the final destination.
994 				 * Otherwise, we can't properly update
995 				 * information that depends on the final
996 				 * destination (e.g. path MTU).
997 				 */
998 #ifndef PULLDOWN_TEST
999 				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth),
1000 						 -1);
1001 				rth = (struct ip6_rthdr *)(mtod(m, caddr_t)
1002 							   + eoff);
1003 #else
1004 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
1005 					       eoff, sizeof(*rth));
1006 				if (rth == NULL) {
1007 					icmp6stat.icp6s_tooshort++;
1008 					return (-1);
1009 				}
1010 #endif
1011 				rthlen = (rth->ip6r_len + 1) << 3;
1012 				/*
1013 				 * XXX: currently there is no
1014 				 * officially defined type other
1015 				 * than type-0.
1016 				 * Note that if the segment left field
1017 				 * is 0, all intermediate hops must
1018 				 * have been passed.
1019 				 */
1020 				if (rth->ip6r_segleft &&
1021 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
1022 					int hops;
1023 
1024 #ifndef PULLDOWN_TEST
1025 					IP6_EXTHDR_CHECK(m, 0, eoff + rthlen,
1026 							 -1);
1027 					rth0 = (struct ip6_rthdr0 *)(mtod(m, caddr_t) + eoff);
1028 #else
1029 					IP6_EXTHDR_GET(rth0,
1030 						       struct ip6_rthdr0 *, m,
1031 						       eoff, rthlen);
1032 					if (rth0 == NULL) {
1033 						icmp6stat.icp6s_tooshort++;
1034 						return (-1);
1035 					}
1036 #endif
1037 					/* just ignore a bogus header */
1038 					if ((rth0->ip6r0_len % 2) == 0 &&
1039 					    (hops = rth0->ip6r0_len/2))
1040 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
1041 				}
1042 				eoff += rthlen;
1043 				nxt = rth->ip6r_nxt;
1044 				break;
1045 			case IPPROTO_FRAGMENT:
1046 #ifndef PULLDOWN_TEST
1047 				IP6_EXTHDR_CHECK(m, 0, eoff +
1048 						 sizeof(struct ip6_frag),
1049 						 -1);
1050 				fh = (struct ip6_frag *)(mtod(m, caddr_t)
1051 							 + eoff);
1052 #else
1053 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1054 					       eoff, sizeof(*fh));
1055 				if (fh == NULL) {
1056 					icmp6stat.icp6s_tooshort++;
1057 					return (-1);
1058 				}
1059 #endif
1060 				/*
1061 				 * Data after a fragment header is meaningless
1062 				 * unless it is the first fragment, but
1063 				 * we'll go to the notify label for path MTU
1064 				 * discovery.
1065 				 */
1066 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1067 					goto notify;
1068 
1069 				eoff += sizeof(struct ip6_frag);
1070 				nxt = fh->ip6f_nxt;
1071 				break;
1072 			default:
1073 				/*
1074 				 * This case includes ESP and the No Next
1075 				 * Header.  In such cases going to the notify
1076 				 * label does not have any meaning
1077 				 * (i.e. ctlfunc will be NULL), but we go
1078 				 * anyway since we might have to update
1079 				 * path MTU information.
1080 				 */
1081 				goto notify;
1082 			}
1083 		}
1084 	  notify:
1085 #ifndef PULLDOWN_TEST
1086 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1087 #else
1088 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1089 			       sizeof(*icmp6) + sizeof(struct ip6_hdr));
1090 		if (icmp6 == NULL) {
1091 			icmp6stat.icp6s_tooshort++;
1092 			return (-1);
1093 		}
1094 #endif
1095 
1096 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1097 		bzero(&icmp6dst, sizeof(icmp6dst));
1098 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1099 		icmp6dst.sin6_family = AF_INET6;
1100 		if (finaldst == NULL)
1101 			icmp6dst.sin6_addr = eip6->ip6_dst;
1102 		else
1103 			icmp6dst.sin6_addr = *finaldst;
1104 		icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1105 							  &icmp6dst.sin6_addr);
1106 #ifndef SCOPEDROUTING
1107 		if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1108 				   NULL, NULL)) {
1109 			/* should be impossbile */
1110 			nd6log((LOG_DEBUG,
1111 			    "icmp6_notify_error: in6_embedscope failed\n"));
1112 			goto freeit;
1113 		}
1114 #endif
1115 
1116 		/*
1117 		 * retrieve parameters from the inner IPv6 header, and convert
1118 		 * them into sockaddr structures.
1119 		 */
1120 		bzero(&icmp6src, sizeof(icmp6src));
1121 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1122 		icmp6src.sin6_family = AF_INET6;
1123 		icmp6src.sin6_addr = eip6->ip6_src;
1124 		icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1125 							  &icmp6src.sin6_addr);
1126 #ifndef SCOPEDROUTING
1127 		if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1128 				   NULL, NULL)) {
1129 			/* should be impossbile */
1130 			nd6log((LOG_DEBUG,
1131 			    "icmp6_notify_error: in6_embedscope failed\n"));
1132 			goto freeit;
1133 		}
1134 #endif
1135 		icmp6src.sin6_flowinfo =
1136 			(eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1137 
1138 		if (finaldst == NULL)
1139 			finaldst = &eip6->ip6_dst;
1140 		ip6cp.ip6c_m = m;
1141 		ip6cp.ip6c_icmp6 = icmp6;
1142 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1143 		ip6cp.ip6c_off = eoff;
1144 		ip6cp.ip6c_finaldst = finaldst;
1145 		ip6cp.ip6c_src = &icmp6src;
1146 		ip6cp.ip6c_nxt = nxt;
1147 
1148 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1149 			notifymtu = ntohl(icmp6->icmp6_mtu);
1150 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1151 		}
1152 
1153 		ctlfunc = (void (*) __P((int, struct sockaddr *, void *)))
1154 			(inet6sw[ip6_protox[nxt]].pr_ctlinput);
1155 		if (ctlfunc) {
1156 			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1157 					  &ip6cp);
1158 		}
1159 	}
1160 	return (0);
1161 
1162   freeit:
1163 	m_freem(m);
1164 	return (-1);
1165 }
1166 
1167 void
1168 icmp6_mtudisc_update(ip6cp, validated)
1169 	struct ip6ctlparam *ip6cp;
1170 	int validated;
1171 {
1172 	unsigned long rtcount;
1173 	struct icmp6_mtudisc_callback *mc;
1174 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1175 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1176 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1177 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1178 	struct rtentry *rt = NULL;
1179 	struct sockaddr_in6 sin6;
1180 
1181 	/*
1182 	 * allow non-validated cases if memory is plenty, to make traffic
1183 	 * from non-connected pcb happy.
1184 	 */
1185 	rtcount = rt_timer_count(icmp6_mtudisc_timeout_q);
1186 	if (validated) {
1187 		if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat)
1188 			return;
1189 		else if (0 <= icmp6_mtudisc_lowat &&
1190 		    rtcount > icmp6_mtudisc_lowat) {
1191 			/*
1192 			 * XXX nuke a victim, install the new one.
1193 			 */
1194 		}
1195 	} else {
1196 		if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat)
1197 			return;
1198 	}
1199 
1200 	bzero(&sin6, sizeof(sin6));
1201 	sin6.sin6_family = PF_INET6;
1202 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1203 	sin6.sin6_addr = *dst;
1204 	/* XXX normally, this won't happen */
1205 	if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1206 		sin6.sin6_addr.s6_addr16[1] =
1207 		    htons(m->m_pkthdr.rcvif->if_index);
1208 	}
1209 	/* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1210 	rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6);
1211 
1212 	if (rt && (rt->rt_flags & RTF_HOST) &&
1213 	    !(rt->rt_rmx.rmx_locks & RTV_MTU) &&
1214 	    (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) {
1215 		if (mtu < IN6_LINKMTU(rt->rt_ifp)) {
1216 			icmp6stat.icp6s_pmtuchg++;
1217 			rt->rt_rmx.rmx_mtu = mtu;
1218 		}
1219 	}
1220 	if (rt) { /* XXX: need braces to avoid conflict with else in RTFREE. */
1221 		RTFREE(rt);
1222 	}
1223 
1224 	/*
1225 	 * Notify protocols that the MTU for this destination
1226 	 * has changed.
1227 	 */
1228 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
1229 	     mc = LIST_NEXT(mc, mc_list))
1230 		(*mc->mc_func)(&sin6.sin6_addr);
1231 }
1232 
1233 /*
1234  * Process a Node Information Query packet, based on
1235  * draft-ietf-ipngwg-icmp-name-lookups-07.
1236  *
1237  * Spec incompatibilities:
1238  * - IPv6 Subject address handling
1239  * - IPv4 Subject address handling support missing
1240  * - Proxy reply (answer even if it's not for me)
1241  * - joins NI group address at in6_ifattach() time only, does not cope
1242  *   with hostname changes by sethostname(3)
1243  */
1244 #ifndef offsetof		/* XXX */
1245 #define	offsetof(type, member)	((size_t)(&((type *)0)->member))
1246 #endif
1247 static struct mbuf *
1248 ni6_input(m, off)
1249 	struct mbuf *m;
1250 	int off;
1251 {
1252 	struct icmp6_nodeinfo *ni6, *nni6;
1253 	struct mbuf *n = NULL;
1254 	u_int16_t qtype;
1255 	int subjlen;
1256 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1257 	struct ni_reply_fqdn *fqdn;
1258 	int addrs;		/* for NI_QTYPE_NODEADDR */
1259 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1260 	struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1261 	struct ip6_hdr *ip6;
1262 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1263 	char *subj = NULL;
1264 
1265 	ip6 = mtod(m, struct ip6_hdr *);
1266 #ifndef PULLDOWN_TEST
1267 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1268 #else
1269 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1270 	if (ni6 == NULL) {
1271 		/* m is already reclaimed */
1272 		return NULL;
1273 	}
1274 #endif
1275 
1276 	/*
1277 	 * Validate IPv6 destination address.
1278 	 *
1279 	 * The Responder must discard the Query without further processing
1280 	 * unless it is one of the Responder's unicast or anycast addresses, or
1281 	 * a link-local scope multicast address which the Responder has joined.
1282 	 * [icmp-name-lookups-07, Section 4.]
1283 	 */
1284 	bzero(&sin6, sizeof(sin6));
1285 	sin6.sin6_family = AF_INET6;
1286 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1287 	bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1288 	/* XXX scopeid */
1289 	if (ifa_ifwithaddr((struct sockaddr *)&sin6))
1290 		; /* unicast/anycast, fine */
1291 	else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1292 		; /* link-local multicast, fine */
1293 	else
1294 		goto bad;
1295 
1296 	/* validate query Subject field. */
1297 	qtype = ntohs(ni6->ni_qtype);
1298 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1299 	switch (qtype) {
1300 	case NI_QTYPE_NOOP:
1301 	case NI_QTYPE_SUPTYPES:
1302 		/* 07 draft */
1303 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1304 			break;
1305 		/* FALLTHROUGH */
1306 	case NI_QTYPE_FQDN:
1307 	case NI_QTYPE_NODEADDR:
1308 		switch (ni6->ni_code) {
1309 		case ICMP6_NI_SUBJ_IPV6:
1310 #if ICMP6_NI_SUBJ_IPV6 != 0
1311 		case 0:
1312 #endif
1313 			/*
1314 			 * backward compatibility - try to accept 03 draft
1315 			 * format, where no Subject is present.
1316 			 */
1317 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1318 			    subjlen == 0) {
1319 				oldfqdn++;
1320 				break;
1321 			}
1322 #if ICMP6_NI_SUBJ_IPV6 != 0
1323 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1324 				goto bad;
1325 #endif
1326 
1327 			if (subjlen != sizeof(sin6.sin6_addr))
1328 				goto bad;
1329 
1330 			/*
1331 			 * Validate Subject address.
1332 			 *
1333 			 * Not sure what exactly "address belongs to the node"
1334 			 * means in the spec, is it just unicast, or what?
1335 			 *
1336 			 * At this moment we consider Subject address as
1337 			 * "belong to the node" if the Subject address equals
1338 			 * to the IPv6 destination address; validation for
1339 			 * IPv6 destination address should have done enough
1340 			 * check for us.
1341 			 *
1342 			 * We do not do proxy at this moment.
1343 			 */
1344 			/* m_pulldown instead of copy? */
1345 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1346 			    subjlen, (caddr_t)&sin6.sin6_addr);
1347 			/* XXX kame scope hack */
1348 			if (IN6_IS_SCOPE_LINKLOCAL(&sin6.sin6_addr)) {
1349 				if ((m->m_flags & M_PKTHDR) != 0 &&
1350 				    m->m_pkthdr.rcvif) {
1351 					sin6.sin6_addr.s6_addr16[1] =
1352 					    htons(m->m_pkthdr.rcvif->if_index);
1353 				}
1354 			}
1355 			subj = (char *)&sin6;
1356 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &sin6.sin6_addr))
1357 				break;
1358 
1359 			/*
1360 			 * XXX if we are to allow other cases, we should really
1361 			 * be careful about scope here.
1362 			 * basically, we should disallow queries toward IPv6
1363 			 * destination X with subject Y, if scope(X) > scope(Y).
1364 			 * if we allow scope(X) > scope(Y), it will result in
1365 			 * information leakage across scope boundary.
1366 			 */
1367 			goto bad;
1368 
1369 		case ICMP6_NI_SUBJ_FQDN:
1370 			/*
1371 			 * Validate Subject name with gethostname(3).
1372 			 *
1373 			 * The behavior may need some debate, since:
1374 			 * - we are not sure if the node has FQDN as
1375 			 *   hostname (returned by gethostname(3)).
1376 			 * - the code does wildcard match for truncated names.
1377 			 *   however, we are not sure if we want to perform
1378 			 *   wildcard match, if gethostname(3) side has
1379 			 *   truncated hostname.
1380 			 */
1381 			n = ni6_nametodns(hostname, hostnamelen, 0);
1382 			if (!n || n->m_next || n->m_len == 0)
1383 				goto bad;
1384 			IP6_EXTHDR_GET(subj, char *, m,
1385 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1386 			if (subj == NULL)
1387 				goto bad;
1388 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1389 					n->m_len)) {
1390 				goto bad;
1391 			}
1392 			m_freem(n);
1393 			n = NULL;
1394 			break;
1395 
1396 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1397 		default:
1398 			goto bad;
1399 		}
1400 		break;
1401 	}
1402 
1403 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1404 	switch (qtype) {
1405 	case NI_QTYPE_FQDN:
1406 		if ((icmp6_nodeinfo & 1) == 0)
1407 			goto bad;
1408 		break;
1409 	case NI_QTYPE_NODEADDR:
1410 		if ((icmp6_nodeinfo & 2) == 0)
1411 			goto bad;
1412 		break;
1413 	}
1414 
1415 	/* guess reply length */
1416 	switch (qtype) {
1417 	case NI_QTYPE_NOOP:
1418 		break;		/* no reply data */
1419 	case NI_QTYPE_SUPTYPES:
1420 		replylen += sizeof(u_int32_t);
1421 		break;
1422 	case NI_QTYPE_FQDN:
1423 		/* XXX will append an mbuf */
1424 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1425 		break;
1426 	case NI_QTYPE_NODEADDR:
1427 		addrs = ni6_addrs(ni6, m, &ifp, subj);
1428 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1429 					  sizeof(u_int32_t))) > MCLBYTES)
1430 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1431 		break;
1432 	default:
1433 		/*
1434 		 * XXX: We must return a reply with the ICMP6 code
1435 		 * `unknown Qtype' in this case.  However we regard the case
1436 		 * as an FQDN query for backward compatibility.
1437 		 * Older versions set a random value to this field,
1438 		 * so it rarely varies in the defined qtypes.
1439 		 * But the mechanism is not reliable...
1440 		 * maybe we should obsolete older versions.
1441 		 */
1442 		qtype = NI_QTYPE_FQDN;
1443 		/* XXX will append an mbuf */
1444 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1445 		oldfqdn++;
1446 		break;
1447 	}
1448 
1449 	/* allocate an mbuf to reply. */
1450 	MGETHDR(n, M_DONTWAIT, m->m_type);
1451 	if (n == NULL) {
1452 		m_freem(m);
1453 		return (NULL);
1454 	}
1455 	M_COPY_PKTHDR(n, m); /* just for rcvif */
1456 	if (replylen > MHLEN) {
1457 		if (replylen > MCLBYTES) {
1458 			/*
1459 			 * XXX: should we try to allocate more? But MCLBYTES
1460 			 * is probably much larger than IPV6_MMTU...
1461 			 */
1462 			goto bad;
1463 		}
1464 		MCLGET(n, M_DONTWAIT);
1465 		if ((n->m_flags & M_EXT) == 0) {
1466 			goto bad;
1467 		}
1468 	}
1469 	n->m_pkthdr.len = n->m_len = replylen;
1470 
1471 	/* copy mbuf header and IPv6 + Node Information base headers */
1472 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1473 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1474 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1475 
1476 	/* qtype dependent procedure */
1477 	switch (qtype) {
1478 	case NI_QTYPE_NOOP:
1479 		nni6->ni_code = ICMP6_NI_SUCCESS;
1480 		nni6->ni_flags = 0;
1481 		break;
1482 	case NI_QTYPE_SUPTYPES:
1483 	{
1484 		u_int32_t v;
1485 		nni6->ni_code = ICMP6_NI_SUCCESS;
1486 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1487 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1488 		v = (u_int32_t)htonl(0x0000000f);
1489 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1490 		break;
1491 	}
1492 	case NI_QTYPE_FQDN:
1493 		nni6->ni_code = ICMP6_NI_SUCCESS;
1494 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1495 						sizeof(struct ip6_hdr) +
1496 						sizeof(struct icmp6_nodeinfo));
1497 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1498 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1499 		/*
1500 		 * XXX do we really have FQDN in variable "hostname"?
1501 		 */
1502 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1503 		if (n->m_next == NULL)
1504 			goto bad;
1505 		/* XXX we assume that n->m_next is not a chain */
1506 		if (n->m_next->m_next != NULL)
1507 			goto bad;
1508 		n->m_pkthdr.len += n->m_next->m_len;
1509 		break;
1510 	case NI_QTYPE_NODEADDR:
1511 	{
1512 		int lenlim, copied;
1513 
1514 		nni6->ni_code = ICMP6_NI_SUCCESS;
1515 		n->m_pkthdr.len = n->m_len =
1516 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1517 		lenlim = M_TRAILINGSPACE(n);
1518 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1519 		/* XXX: reset mbuf length */
1520 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1521 			sizeof(struct icmp6_nodeinfo) + copied;
1522 		break;
1523 	}
1524 	default:
1525 		break;		/* XXX impossible! */
1526 	}
1527 
1528 	nni6->ni_type = ICMP6_NI_REPLY;
1529 	m_freem(m);
1530 	return (n);
1531 
1532   bad:
1533 	m_freem(m);
1534 	if (n)
1535 		m_freem(n);
1536 	return (NULL);
1537 }
1538 #undef hostnamelen
1539 
1540 #define isupper(x) ('A' <= (x) && (x) <= 'Z')
1541 #define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z'))
1542 #define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9'))
1543 #define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x))
1544 
1545 /*
1546  * make a mbuf with DNS-encoded string.  no compression support.
1547  *
1548  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1549  * treated as truncated name (two \0 at the end).  this is a wild guess.
1550  */
1551 static struct mbuf *
1552 ni6_nametodns(name, namelen, old)
1553 	const char *name;
1554 	int namelen;
1555 	int old;	/* return pascal string if non-zero */
1556 {
1557 	struct mbuf *m;
1558 	char *cp, *ep;
1559 	const char *p, *q;
1560 	int i, len, nterm;
1561 
1562 	if (old)
1563 		len = namelen + 1;
1564 	else
1565 		len = MCLBYTES;
1566 
1567 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1568 	MGET(m, M_DONTWAIT, MT_DATA);
1569 	if (m && len > MLEN) {
1570 		MCLGET(m, M_DONTWAIT);
1571 		if ((m->m_flags & M_EXT) == 0)
1572 			goto fail;
1573 	}
1574 	if (!m)
1575 		goto fail;
1576 	m->m_next = NULL;
1577 
1578 	if (old) {
1579 		m->m_len = len;
1580 		*mtod(m, char *) = namelen;
1581 		bcopy(name, mtod(m, char *) + 1, namelen);
1582 		return m;
1583 	} else {
1584 		m->m_len = 0;
1585 		cp = mtod(m, char *);
1586 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1587 
1588 		/* if not certain about my name, return empty buffer */
1589 		if (namelen == 0)
1590 			return m;
1591 
1592 		/*
1593 		 * guess if it looks like shortened hostname, or FQDN.
1594 		 * shortened hostname needs two trailing "\0".
1595 		 */
1596 		i = 0;
1597 		for (p = name; p < name + namelen; p++) {
1598 			if (*p && *p == '.')
1599 				i++;
1600 		}
1601 		if (i < 2)
1602 			nterm = 2;
1603 		else
1604 			nterm = 1;
1605 
1606 		p = name;
1607 		while (cp < ep && p < name + namelen) {
1608 			i = 0;
1609 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1610 				i++;
1611 			/* result does not fit into mbuf */
1612 			if (cp + i + 1 >= ep)
1613 				goto fail;
1614 			/*
1615 			 * DNS label length restriction, RFC1035 page 8.
1616 			 * "i == 0" case is included here to avoid returning
1617 			 * 0-length label on "foo..bar".
1618 			 */
1619 			if (i <= 0 || i >= 64)
1620 				goto fail;
1621 			*cp++ = i;
1622 			if (!isalpha(p[0]) || !isalnum(p[i - 1]))
1623 				goto fail;
1624 			while (i > 0) {
1625 				if (!isalnum(*p) && *p != '-')
1626 					goto fail;
1627 				if (isupper(*p)) {
1628 					*cp++ = tolower(*p);
1629 					p++;
1630 				} else
1631 					*cp++ = *p++;
1632 				i--;
1633 			}
1634 			p = q;
1635 			if (p < name + namelen && *p == '.')
1636 				p++;
1637 		}
1638 		/* termination */
1639 		if (cp + nterm >= ep)
1640 			goto fail;
1641 		while (nterm-- > 0)
1642 			*cp++ = '\0';
1643 		m->m_len = cp - mtod(m, char *);
1644 		return m;
1645 	}
1646 
1647 	panic("should not reach here");
1648 	/* NOTREACHED */
1649 
1650  fail:
1651 	if (m)
1652 		m_freem(m);
1653 	return NULL;
1654 }
1655 
1656 /*
1657  * check if two DNS-encoded string matches.  takes care of truncated
1658  * form (with \0\0 at the end).  no compression support.
1659  * XXX upper/lowercase match (see RFC2065)
1660  */
1661 static int
1662 ni6_dnsmatch(a, alen, b, blen)
1663 	const char *a;
1664 	int alen;
1665 	const char *b;
1666 	int blen;
1667 {
1668 	const char *a0, *b0;
1669 	int l;
1670 
1671 	/* simplest case - need validation? */
1672 	if (alen == blen && bcmp(a, b, alen) == 0)
1673 		return 1;
1674 
1675 	a0 = a;
1676 	b0 = b;
1677 
1678 	/* termination is mandatory */
1679 	if (alen < 2 || blen < 2)
1680 		return 0;
1681 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1682 		return 0;
1683 	alen--;
1684 	blen--;
1685 
1686 	while (a - a0 < alen && b - b0 < blen) {
1687 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1688 			return 0;
1689 
1690 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1691 			return 0;
1692 		/* we don't support compression yet */
1693 		if (a[0] >= 64 || b[0] >= 64)
1694 			return 0;
1695 
1696 		/* truncated case */
1697 		if (a[0] == 0 && a - a0 == alen - 1)
1698 			return 1;
1699 		if (b[0] == 0 && b - b0 == blen - 1)
1700 			return 1;
1701 		if (a[0] == 0 || b[0] == 0)
1702 			return 0;
1703 
1704 		if (a[0] != b[0])
1705 			return 0;
1706 		l = a[0];
1707 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1708 			return 0;
1709 		if (bcmp(a + 1, b + 1, l) != 0)
1710 			return 0;
1711 
1712 		a += 1 + l;
1713 		b += 1 + l;
1714 	}
1715 
1716 	if (a - a0 == alen && b - b0 == blen)
1717 		return 1;
1718 	else
1719 		return 0;
1720 }
1721 
1722 /*
1723  * calculate the number of addresses to be returned in the node info reply.
1724  */
1725 static int
1726 ni6_addrs(ni6, m, ifpp, subj)
1727 	struct icmp6_nodeinfo *ni6;
1728 	struct mbuf *m;
1729 	struct ifnet **ifpp;
1730 	char *subj;
1731 {
1732 	struct ifnet *ifp;
1733 	struct in6_ifaddr *ifa6;
1734 	struct ifaddr *ifa;
1735 	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1736 	int addrs = 0, addrsofif, iffound = 0;
1737 	int niflags = ni6->ni_flags;
1738 
1739 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1740 		switch (ni6->ni_code) {
1741 		case ICMP6_NI_SUBJ_IPV6:
1742 			if (subj == NULL) /* must be impossible... */
1743 				return (0);
1744 			subj_ip6 = (struct sockaddr_in6 *)subj;
1745 			break;
1746 		default:
1747 			/*
1748 			 * XXX: we only support IPv6 subject address for
1749 			 * this Qtype.
1750 			 */
1751 			return (0);
1752 		}
1753 	}
1754 
1755 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
1756 	{
1757 		addrsofif = 0;
1758 		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1759 		     ifa = ifa->ifa_list.tqe_next)
1760 		{
1761 			if (ifa->ifa_addr->sa_family != AF_INET6)
1762 				continue;
1763 			ifa6 = (struct in6_ifaddr *)ifa;
1764 
1765 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1766 			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1767 					       &ifa6->ia_addr.sin6_addr))
1768 				iffound = 1;
1769 
1770 			/*
1771 			 * IPv4-mapped addresses can only be returned by a
1772 			 * Node Information proxy, since they represent
1773 			 * addresses of IPv4-only nodes, which perforce do
1774 			 * not implement this protocol.
1775 			 * [icmp-name-lookups-07, Section 5.4]
1776 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1777 			 * this function at this moment.
1778 			 */
1779 
1780 			/* What do we have to do about ::1? */
1781 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1782 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1783 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1784 					continue;
1785 				break;
1786 			case IPV6_ADDR_SCOPE_SITELOCAL:
1787 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1788 					continue;
1789 				break;
1790 			case IPV6_ADDR_SCOPE_GLOBAL:
1791 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1792 					continue;
1793 				break;
1794 			default:
1795 				continue;
1796 			}
1797 
1798 			/*
1799 			 * check if anycast is okay.
1800 			 * XXX: just experimental.  not in the spec.
1801 			 */
1802 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1803 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1804 				continue; /* we need only unicast addresses */
1805 
1806 			addrsofif++; /* count the address */
1807 		}
1808 		if (iffound) {
1809 			*ifpp = ifp;
1810 			return (addrsofif);
1811 		}
1812 
1813 		addrs += addrsofif;
1814 	}
1815 
1816 	return (addrs);
1817 }
1818 
1819 static int
1820 ni6_store_addrs(ni6, nni6, ifp0, resid)
1821 	struct icmp6_nodeinfo *ni6, *nni6;
1822 	struct ifnet *ifp0;
1823 	int resid;
1824 {
1825 	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1826 	struct in6_ifaddr *ifa6;
1827 	struct ifaddr *ifa;
1828 	struct ifnet *ifp_dep = NULL;
1829 	int copied = 0, allow_deprecated = 0;
1830 	u_char *cp = (u_char *)(nni6 + 1);
1831 	int niflags = ni6->ni_flags;
1832 	u_int32_t ltime;
1833 	long time_second = time.tv_sec;
1834 
1835 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1836 		return (0);	/* needless to copy */
1837 
1838   again:
1839 
1840 	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list))
1841 	{
1842 		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1843 		     ifa = ifa->ifa_list.tqe_next)
1844 		{
1845 			if (ifa->ifa_addr->sa_family != AF_INET6)
1846 				continue;
1847 			ifa6 = (struct in6_ifaddr *)ifa;
1848 
1849 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1850 			    allow_deprecated == 0) {
1851 				/*
1852 				 * prefererred address should be put before
1853 				 * deprecated addresses.
1854 				 */
1855 
1856 				/* record the interface for later search */
1857 				if (ifp_dep == NULL)
1858 					ifp_dep = ifp;
1859 
1860 				continue;
1861 			}
1862 			else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1863 				 allow_deprecated != 0)
1864 				continue; /* we now collect deprecated addrs */
1865 
1866 			/* What do we have to do about ::1? */
1867 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1868 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1869 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1870 					continue;
1871 				break;
1872 			case IPV6_ADDR_SCOPE_SITELOCAL:
1873 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1874 					continue;
1875 				break;
1876 			case IPV6_ADDR_SCOPE_GLOBAL:
1877 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1878 					continue;
1879 				break;
1880 			default:
1881 				continue;
1882 			}
1883 
1884 			/*
1885 			 * check if anycast is okay.
1886 			 * XXX: just experimental.  not in the spec.
1887 			 */
1888 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1889 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1890 				continue;
1891 
1892 			/* now we can copy the address */
1893 			if (resid < sizeof(struct in6_addr) +
1894 			    sizeof(u_int32_t)) {
1895 				/*
1896 				 * We give up much more copy.
1897 				 * Set the truncate flag and return.
1898 				 */
1899 				nni6->ni_flags |=
1900 					NI_NODEADDR_FLAG_TRUNCATE;
1901 				return (copied);
1902 			}
1903 
1904 			/*
1905 			 * Set the TTL of the address.
1906 			 * The TTL value should be one of the following
1907 			 * according to the specification:
1908 			 *
1909 			 * 1. The remaining lifetime of a DHCP lease on the
1910 			 *    address, or
1911 			 * 2. The remaining Valid Lifetime of a prefix from
1912 			 *    which the address was derived through Stateless
1913 			 *    Autoconfiguration.
1914 			 *
1915 			 * Note that we currently do not support stateful
1916 			 * address configuration by DHCPv6, so the former
1917 			 * case can't happen.
1918 			 *
1919 			 * TTL must be 2^31 > TTL >= 0.
1920 			 */
1921 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1922 				ltime = ND6_INFINITE_LIFETIME;
1923 			else {
1924 				if (ifa6->ia6_lifetime.ia6t_expire >
1925 				    time_second)
1926 					ltime = ifa6->ia6_lifetime.ia6t_expire - time_second;
1927 				else
1928 					ltime = 0;
1929 			}
1930 			if (ltime > 0x7fffffff)
1931 				ltime = 0x7fffffff;
1932 			ltime = htonl(ltime);
1933 
1934 			bcopy(&ltime, cp, sizeof(u_int32_t));
1935 			cp += sizeof(u_int32_t);
1936 
1937 			/* copy the address itself */
1938 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1939 			      sizeof(struct in6_addr));
1940 			/* XXX: KAME link-local hack; remove ifindex */
1941 			if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr))
1942 				((struct in6_addr *)cp)->s6_addr16[1] = 0;
1943 			cp += sizeof(struct in6_addr);
1944 
1945 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1946 			copied += (sizeof(struct in6_addr) +
1947 				   sizeof(u_int32_t));
1948 		}
1949 		if (ifp0)	/* we need search only on the specified IF */
1950 			break;
1951 	}
1952 
1953 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1954 		ifp = ifp_dep;
1955 		allow_deprecated = 1;
1956 
1957 		goto again;
1958 	}
1959 
1960 	return (copied);
1961 }
1962 
1963 /*
1964  * XXX almost dup'ed code with rip6_input.
1965  */
1966 static int
1967 icmp6_rip6_input(mp, off)
1968 	struct	mbuf **mp;
1969 	int	off;
1970 {
1971 	struct mbuf *m = *mp;
1972 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1973 	struct in6pcb *in6p;
1974 	struct in6pcb *last = NULL;
1975 	struct sockaddr_in6 rip6src;
1976 	struct icmp6_hdr *icmp6;
1977 	struct mbuf *opts = NULL;
1978 
1979 #ifndef PULLDOWN_TEST
1980 	/* this is assumed to be safe. */
1981 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1982 #else
1983 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1984 	if (icmp6 == NULL) {
1985 		/* m is already reclaimed */
1986 		return IPPROTO_DONE;
1987 	}
1988 #endif
1989 
1990 	bzero(&rip6src, sizeof(rip6src));
1991 	rip6src.sin6_len = sizeof(struct sockaddr_in6);
1992 	rip6src.sin6_family = AF_INET6;
1993 	/* KAME hack: recover scopeid */
1994 	(void)in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif);
1995 
1996 	for (in6p = rawin6pcb.in6p_next;
1997 	     in6p != &rawin6pcb; in6p = in6p->in6p_next)
1998 	{
1999 		if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
2000 			continue;
2001 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
2002 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
2003 			continue;
2004 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
2005 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
2006 			continue;
2007 		if (in6p->in6p_icmp6filt
2008 		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
2009 				 in6p->in6p_icmp6filt))
2010 			continue;
2011 		if (last) {
2012 			struct	mbuf *n;
2013 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
2014 				if (last->in6p_flags & IN6P_CONTROLOPTS)
2015 					ip6_savecontrol(last, &opts, ip6, n);
2016 				/* strip intermediate headers */
2017 				m_adj(n, off);
2018 				if (sbappendaddr(&last->in6p_socket->so_rcv,
2019 						 (struct sockaddr *)&rip6src,
2020 						 n, opts) == 0) {
2021 					/* should notify about lost packet */
2022 					m_freem(n);
2023 					if (opts)
2024 						m_freem(opts);
2025 				} else
2026 					sorwakeup(last->in6p_socket);
2027 				opts = NULL;
2028 			}
2029 		}
2030 		last = in6p;
2031 	}
2032 	if (last) {
2033 		if (last->in6p_flags & IN6P_CONTROLOPTS)
2034 			ip6_savecontrol(last, &opts, ip6, m);
2035 		/* strip intermediate headers */
2036 		m_adj(m, off);
2037 		if (sbappendaddr(&last->in6p_socket->so_rcv,
2038 				(struct sockaddr *)&rip6src, m, opts) == 0) {
2039 			m_freem(m);
2040 			if (opts)
2041 				m_freem(opts);
2042 		} else
2043 			sorwakeup(last->in6p_socket);
2044 	} else {
2045 		m_freem(m);
2046 		ip6stat.ip6s_delivered--;
2047 	}
2048 	return IPPROTO_DONE;
2049 }
2050 
2051 /*
2052  * Reflect the ip6 packet back to the source.
2053  * OFF points to the icmp6 header, counted from the top of the mbuf.
2054  *
2055  * Note: RFC 1885 required that an echo reply should be truncated if it
2056  * did not fit in with (return) path MTU, and KAME code supported the
2057  * behavior.  However, as a clarification after the RFC, this limitation
2058  * was removed in a revised version of the spec, RFC 2463.  We had kept the
2059  * old behavior, with a (non-default) ifdef block, while the new version of
2060  * the spec was an internet-draft status, and even after the new RFC was
2061  * published.  But it would rather make sense to clean the obsoleted part
2062  * up, and to make the code simpler at this stage.
2063  */
2064 void
2065 icmp6_reflect(m, off)
2066 	struct	mbuf *m;
2067 	size_t off;
2068 {
2069 	struct ip6_hdr *ip6;
2070 	struct icmp6_hdr *icmp6;
2071 	struct in6_ifaddr *ia;
2072 	struct in6_addr t, *src = 0;
2073 	int plen;
2074 	int type, code;
2075 	struct ifnet *outif = NULL;
2076 	struct sockaddr_in6 sa6_src, sa6_dst;
2077 
2078 	/* too short to reflect */
2079 	if (off < sizeof(struct ip6_hdr)) {
2080 		nd6log((LOG_DEBUG,
2081 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2082 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2083 		    __FILE__, __LINE__));
2084 		goto bad;
2085 	}
2086 
2087 	/*
2088 	 * If there are extra headers between IPv6 and ICMPv6, strip
2089 	 * off that header first.
2090 	 */
2091 #ifdef DIAGNOSTIC
2092 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2093 		panic("assumption failed in icmp6_reflect");
2094 #endif
2095 	if (off > sizeof(struct ip6_hdr)) {
2096 		size_t l;
2097 		struct ip6_hdr nip6;
2098 
2099 		l = off - sizeof(struct ip6_hdr);
2100 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2101 		m_adj(m, l);
2102 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2103 		if (m->m_len < l) {
2104 			if ((m = m_pullup(m, l)) == NULL)
2105 				return;
2106 		}
2107 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2108 	} else /* off == sizeof(struct ip6_hdr) */ {
2109 		size_t l;
2110 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2111 		if (m->m_len < l) {
2112 			if ((m = m_pullup(m, l)) == NULL)
2113 				return;
2114 		}
2115 	}
2116 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2117 	ip6 = mtod(m, struct ip6_hdr *);
2118 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2119 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2120 	type = icmp6->icmp6_type; /* keep type for statistics */
2121 	code = icmp6->icmp6_code; /* ditto. */
2122 
2123 	t = ip6->ip6_dst;
2124 	/*
2125 	 * ip6_input() drops a packet if its src is multicast.
2126 	 * So, the src is never multicast.
2127 	 */
2128 	ip6->ip6_dst = ip6->ip6_src;
2129 
2130 	/*
2131 	 * XXX: make sure to embed scope zone information, using
2132 	 * already embedded IDs or the received interface (if any).
2133 	 * Note that rcvif may be NULL.
2134 	 * TODO: scoped routing case (XXX).
2135 	 */
2136 	bzero(&sa6_src, sizeof(sa6_src));
2137 	sa6_src.sin6_family = AF_INET6;
2138 	sa6_src.sin6_len = sizeof(sa6_src);
2139 	sa6_src.sin6_addr = ip6->ip6_dst;
2140 	in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
2141 	in6_embedscope(&sa6_src.sin6_addr, &sa6_src, NULL, NULL);
2142 	ip6->ip6_dst = sa6_src.sin6_addr;
2143 
2144 	bzero(&sa6_dst, sizeof(sa6_dst));
2145 	sa6_dst.sin6_family = AF_INET6;
2146 	sa6_dst.sin6_len = sizeof(sa6_dst);
2147 	sa6_dst.sin6_addr = t;
2148 	in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
2149 	in6_embedscope(&t, &sa6_dst, NULL, NULL);
2150 
2151 	/*
2152 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2153 	 * use dst as the src for the reply.
2154 	 * The IN6_IFF_NOTREADY case would be VERY rare, but is possible
2155 	 * (for example) when we encounter an error while forwarding procedure
2156 	 * destined to a duplicated address of ours.
2157 	 */
2158 	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2159 		if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2160 		    (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2161 			src = &t;
2162 			break;
2163 		}
2164 	if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2165 		/*
2166 		 * This is the case if the dst is our link-local address
2167 		 * and the sender is also ourselves.
2168 		 */
2169 		src = &t;
2170 	}
2171 
2172 	if (src == 0) {
2173 		int e;
2174 		struct route_in6 ro;
2175 
2176 		/*
2177 		 * This case matches to multicasts, our anycast, or unicasts
2178 		 * that we do not own.  Select a source address based on the
2179 		 * source address of the erroneous packet.
2180 		 */
2181 		bzero(&ro, sizeof(ro));
2182 		src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e);
2183 		if (ro.ro_rt) { /* XXX: see comments in icmp6_mtudisc_update */
2184 			RTFREE(ro.ro_rt); /* XXX: we could use this */
2185 		}
2186 		if (src == NULL) {
2187 			nd6log((LOG_DEBUG,
2188 			    "icmp6_reflect: source can't be determined: "
2189 			    "dst=%s, error=%d\n",
2190 			    ip6_sprintf(&sa6_src.sin6_addr), e));
2191 			goto bad;
2192 		}
2193 	}
2194 
2195 	ip6->ip6_src = *src;
2196 
2197 	ip6->ip6_flow = 0;
2198 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2199 	ip6->ip6_vfc |= IPV6_VERSION;
2200 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2201 	if (m->m_pkthdr.rcvif) {
2202 		/* XXX: This may not be the outgoing interface */
2203 		ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2204 	} else
2205 		ip6->ip6_hlim = ip6_defhlim;
2206 
2207 	icmp6->icmp6_cksum = 0;
2208 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2209 					sizeof(struct ip6_hdr), plen);
2210 
2211 	/*
2212 	 * XXX option handling
2213 	 */
2214 
2215 	m->m_flags &= ~(M_BCAST|M_MCAST);
2216 #ifdef IPSEC
2217 	/* Don't lookup socket */
2218 	(void)ipsec_setsocket(m, NULL);
2219 #endif /* IPSEC */
2220 
2221 	/*
2222 	 * To avoid a "too big" situation at an intermediate router
2223 	 * and the path MTU discovery process, specify the IPV6_MINMTU flag.
2224 	 * Note that only echo and node information replies are affected,
2225 	 * since the length of ICMP6 errors is limited to the minimum MTU.
2226 	 */
2227 	if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, &outif) != 0 && outif)
2228 		icmp6_ifstat_inc(outif, ifs6_out_error);
2229 
2230 	if (outif)
2231 		icmp6_ifoutstat_inc(outif, type, code);
2232 
2233 	return;
2234 
2235  bad:
2236 	m_freem(m);
2237 	return;
2238 }
2239 
2240 void
2241 icmp6_fasttimo()
2242 {
2243 
2244 	mld6_fasttimeo();
2245 }
2246 
2247 static const char *
2248 icmp6_redirect_diag(src6, dst6, tgt6)
2249 	struct in6_addr *src6;
2250 	struct in6_addr *dst6;
2251 	struct in6_addr *tgt6;
2252 {
2253 	static char buf[1024];
2254 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2255 		ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2256 	return buf;
2257 }
2258 
2259 void
2260 icmp6_redirect_input(m, off)
2261 	struct mbuf *m;
2262 	int off;
2263 {
2264 	struct ifnet *ifp = m->m_pkthdr.rcvif;
2265 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2266 	struct nd_redirect *nd_rd;
2267 	int icmp6len = ntohs(ip6->ip6_plen);
2268 	char *lladdr = NULL;
2269 	int lladdrlen = 0;
2270 	u_char *redirhdr = NULL;
2271 	int redirhdrlen = 0;
2272 	struct rtentry *rt = NULL;
2273 	int is_router;
2274 	int is_onlink;
2275 	struct in6_addr src6 = ip6->ip6_src;
2276 	struct in6_addr redtgt6;
2277 	struct in6_addr reddst6;
2278 	union nd_opts ndopts;
2279 
2280 	if (!ifp)
2281 		return;
2282 
2283 	/* XXX if we are router, we don't update route by icmp6 redirect */
2284 	if (ip6_forwarding)
2285 		goto freeit;
2286 	if (!icmp6_rediraccept)
2287 		goto freeit;
2288 
2289 #ifndef PULLDOWN_TEST
2290 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2291 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2292 #else
2293 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2294 	if (nd_rd == NULL) {
2295 		icmp6stat.icp6s_tooshort++;
2296 		return;
2297 	}
2298 #endif
2299 	redtgt6 = nd_rd->nd_rd_target;
2300 	reddst6 = nd_rd->nd_rd_dst;
2301 
2302 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2303 		redtgt6.s6_addr16[1] = htons(ifp->if_index);
2304 	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2305 		reddst6.s6_addr16[1] = htons(ifp->if_index);
2306 
2307 	/* validation */
2308 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2309 		nd6log((LOG_ERR,
2310 			"ICMP6 redirect sent from %s rejected; "
2311 			"must be from linklocal\n", ip6_sprintf(&src6)));
2312 		goto bad;
2313 	}
2314 	if (ip6->ip6_hlim != 255) {
2315 		nd6log((LOG_ERR,
2316 			"ICMP6 redirect sent from %s rejected; "
2317 			"hlim=%d (must be 255)\n",
2318 			ip6_sprintf(&src6), ip6->ip6_hlim));
2319 		goto bad;
2320 	}
2321     {
2322 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2323 	struct sockaddr_in6 sin6;
2324 	struct in6_addr *gw6;
2325 
2326 	bzero(&sin6, sizeof(sin6));
2327 	sin6.sin6_family = AF_INET6;
2328 	sin6.sin6_len = sizeof(struct sockaddr_in6);
2329 	bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2330 	rt = rtalloc1((struct sockaddr *)&sin6, 0);
2331 	if (rt) {
2332 		if (rt->rt_gateway == NULL ||
2333 		    rt->rt_gateway->sa_family != AF_INET6) {
2334 			nd6log((LOG_ERR,
2335 			    "ICMP6 redirect rejected; no route "
2336 			    "with inet6 gateway found for redirect dst: %s\n",
2337 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2338 			RTFREE(rt);
2339 			goto bad;
2340 		}
2341 
2342 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2343 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2344 			nd6log((LOG_ERR,
2345 				"ICMP6 redirect rejected; "
2346 				"not equal to gw-for-src=%s (must be same): "
2347 				"%s\n",
2348 				ip6_sprintf(gw6),
2349 				icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2350 			RTFREE(rt);
2351 			goto bad;
2352 		}
2353 	} else {
2354 		nd6log((LOG_ERR,
2355 			"ICMP6 redirect rejected; "
2356 			"no route found for redirect dst: %s\n",
2357 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2358 		goto bad;
2359 	}
2360 	RTFREE(rt);
2361 	rt = NULL;
2362     }
2363 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2364 		nd6log((LOG_ERR,
2365 			"ICMP6 redirect rejected; "
2366 			"redirect dst must be unicast: %s\n",
2367 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2368 		goto bad;
2369 	}
2370 
2371 	is_router = is_onlink = 0;
2372 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2373 		is_router = 1;	/* router case */
2374 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2375 		is_onlink = 1;	/* on-link destination case */
2376 	if (!is_router && !is_onlink) {
2377 		nd6log((LOG_ERR,
2378 			"ICMP6 redirect rejected; "
2379 			"neither router case nor onlink case: %s\n",
2380 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2381 		goto bad;
2382 	}
2383 	/* validation passed */
2384 
2385 	icmp6len -= sizeof(*nd_rd);
2386 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2387 	if (nd6_options(&ndopts) < 0) {
2388 		nd6log((LOG_INFO, "icmp6_redirect_input: "
2389 			"invalid ND option, rejected: %s\n",
2390 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2391 		/* nd6_options have incremented stats */
2392 		goto freeit;
2393 	}
2394 
2395 	if (ndopts.nd_opts_tgt_lladdr) {
2396 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2397 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2398 	}
2399 
2400 	if (ndopts.nd_opts_rh) {
2401 		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2402 		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2403 	}
2404 
2405 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2406 		nd6log((LOG_INFO,
2407 			"icmp6_redirect_input: lladdrlen mismatch for %s "
2408 			"(if %d, icmp6 packet %d): %s\n",
2409 			ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2410 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2411 		goto bad;
2412 	}
2413 
2414 	/* RFC 2461 8.3 */
2415 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2416 			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2417 
2418 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2419 		/* perform rtredirect */
2420 		struct sockaddr_in6 sdst;
2421 		struct sockaddr_in6 sgw;
2422 		struct sockaddr_in6 ssrc;
2423 		unsigned long rtcount;
2424 		struct rtentry *newrt = NULL;
2425 
2426 		/*
2427 		 * do not install redirect route, if the number of entries
2428 		 * is too much (> hiwat).  note that, the node (= host) will
2429 		 * work just fine even if we do not install redirect route
2430 		 * (there will be additional hops, though).
2431 		 */
2432 		rtcount = rt_timer_count(icmp6_redirect_timeout_q);
2433 		if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat)
2434 			return;
2435 		else if (0 <= icmp6_redirect_lowat &&
2436 		    rtcount > icmp6_redirect_lowat) {
2437 			/*
2438 			 * XXX nuke a victim, install the new one.
2439 			 */
2440 		}
2441 
2442 		bzero(&sdst, sizeof(sdst));
2443 		bzero(&sgw, sizeof(sgw));
2444 		bzero(&ssrc, sizeof(ssrc));
2445 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2446 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2447 			sizeof(struct sockaddr_in6);
2448 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2449 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2450 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2451 		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2452 			   (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2453 			   (struct sockaddr *)&ssrc,
2454 			   &newrt);
2455 
2456 		if (newrt) {
2457 			(void)rt_timer_add(newrt, icmp6_redirect_timeout,
2458 			    icmp6_redirect_timeout_q);
2459 			rtfree(newrt);
2460 		}
2461 	}
2462 	/* finally update cached route in each socket via pfctlinput */
2463 	{
2464 		struct sockaddr_in6 sdst;
2465 
2466 		bzero(&sdst, sizeof(sdst));
2467 		sdst.sin6_family = AF_INET6;
2468 		sdst.sin6_len = sizeof(struct sockaddr_in6);
2469 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2470 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2471 #ifdef IPSEC
2472 		key_sa_routechange((struct sockaddr *)&sdst);
2473 #endif
2474 	}
2475 
2476  freeit:
2477 	m_freem(m);
2478 	return;
2479 
2480  bad:
2481 	icmp6stat.icp6s_badredirect++;
2482 	m_freem(m);
2483 }
2484 
2485 void
2486 icmp6_redirect_output(m0, rt)
2487 	struct mbuf *m0;
2488 	struct rtentry *rt;
2489 {
2490 	struct ifnet *ifp;	/* my outgoing interface */
2491 	struct in6_addr *ifp_ll6;
2492 	struct in6_addr *nexthop;
2493 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2494 	struct mbuf *m = NULL;	/* newly allocated one */
2495 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2496 	struct nd_redirect *nd_rd;
2497 	size_t maxlen;
2498 	u_char *p;
2499 	struct sockaddr_in6 src_sa;
2500 
2501 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2502 
2503 	/* if we are not router, we don't send icmp6 redirect */
2504 	if (!ip6_forwarding || ip6_accept_rtadv)
2505 		goto fail;
2506 
2507 	/* sanity check */
2508 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2509 		goto fail;
2510 
2511 	/*
2512 	 * Address check:
2513 	 *  the source address must identify a neighbor, and
2514 	 *  the destination address must not be a multicast address
2515 	 *  [RFC 2461, sec 8.2]
2516 	 */
2517 	sip6 = mtod(m0, struct ip6_hdr *);
2518 	bzero(&src_sa, sizeof(src_sa));
2519 	src_sa.sin6_family = AF_INET6;
2520 	src_sa.sin6_len = sizeof(src_sa);
2521 	src_sa.sin6_addr = sip6->ip6_src;
2522 	/* we don't currently use sin6_scope_id, but eventually use it */
2523 	src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src);
2524 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2525 		goto fail;
2526 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2527 		goto fail;	/* what should we do here? */
2528 
2529 	/* rate limit */
2530 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2531 		goto fail;
2532 
2533 	/*
2534 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2535 	 * we almost always ask for an mbuf cluster for simplicity.
2536 	 * (MHLEN < IPV6_MMTU is almost always true)
2537 	 */
2538 #if IPV6_MMTU >= MCLBYTES
2539 # error assumption failed about IPV6_MMTU and MCLBYTES
2540 #endif
2541 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2542 	if (m && IPV6_MMTU >= MHLEN)
2543 		MCLGET(m, M_DONTWAIT);
2544 	if (!m)
2545 		goto fail;
2546 	m->m_pkthdr.rcvif = NULL;
2547 	m->m_len = 0;
2548 	maxlen = M_TRAILINGSPACE(m);
2549 	maxlen = min(IPV6_MMTU, maxlen);
2550 	/* just for safety */
2551 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2552 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2553 		goto fail;
2554 	}
2555 
2556 	{
2557 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2558 		struct in6_ifaddr *ia;
2559 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2560 						 IN6_IFF_NOTREADY|
2561 						 IN6_IFF_ANYCAST)) == NULL)
2562 			goto fail;
2563 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2564 	}
2565 
2566 	/* get ip6 linklocal address for the router. */
2567 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2568 		struct sockaddr_in6 *sin6;
2569 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2570 		nexthop = &sin6->sin6_addr;
2571 		if (!IN6_IS_ADDR_LINKLOCAL(nexthop))
2572 			nexthop = NULL;
2573 	} else
2574 		nexthop = NULL;
2575 
2576 	/* ip6 */
2577 	ip6 = mtod(m, struct ip6_hdr *);
2578 	ip6->ip6_flow = 0;
2579 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2580 	ip6->ip6_vfc |= IPV6_VERSION;
2581 	/* ip6->ip6_plen will be set later */
2582 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2583 	ip6->ip6_hlim = 255;
2584 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2585 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2586 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2587 
2588 	/* ND Redirect */
2589 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2590 	nd_rd->nd_rd_type = ND_REDIRECT;
2591 	nd_rd->nd_rd_code = 0;
2592 	nd_rd->nd_rd_reserved = 0;
2593 	if (rt->rt_flags & RTF_GATEWAY) {
2594 		/*
2595 		 * nd_rd->nd_rd_target must be a link-local address in
2596 		 * better router cases.
2597 		 */
2598 		if (!nexthop)
2599 			goto fail;
2600 		bcopy(nexthop, &nd_rd->nd_rd_target,
2601 		      sizeof(nd_rd->nd_rd_target));
2602 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2603 		      sizeof(nd_rd->nd_rd_dst));
2604 	} else {
2605 		/* make sure redtgt == reddst */
2606 		nexthop = &sip6->ip6_dst;
2607 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2608 		      sizeof(nd_rd->nd_rd_target));
2609 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2610 		      sizeof(nd_rd->nd_rd_dst));
2611 	}
2612 
2613 	p = (u_char *)(nd_rd + 1);
2614 
2615 	{
2616 		/* target lladdr option */
2617 		struct rtentry *rt_nexthop = NULL;
2618 		int len;
2619 		struct sockaddr_dl *sdl;
2620 		struct nd_opt_hdr *nd_opt;
2621 		char *lladdr;
2622 
2623 		rt_nexthop = nd6_lookup(nexthop, 0, ifp);
2624 		if (!rt_nexthop)
2625 			goto nolladdropt;
2626 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2627 		len = (len + 7) & ~7;	/* round by 8 */
2628 		/* safety check */
2629 		if (len + (p - (u_char *)ip6) > maxlen)
2630 			goto nolladdropt;
2631 		if (!(rt_nexthop->rt_flags & RTF_GATEWAY) &&
2632 		    (rt_nexthop->rt_flags & RTF_LLINFO) &&
2633 		    (rt_nexthop->rt_gateway->sa_family == AF_LINK) &&
2634 		    (sdl = (struct sockaddr_dl *)rt_nexthop->rt_gateway) &&
2635 		    sdl->sdl_alen) {
2636 			nd_opt = (struct nd_opt_hdr *)p;
2637 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2638 			nd_opt->nd_opt_len = len >> 3;
2639 			lladdr = (char *)(nd_opt + 1);
2640 			bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2641 			p += len;
2642 		}
2643 	}
2644   nolladdropt:;
2645 
2646 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2647 
2648 	/* just to be safe */
2649 	if (m0->m_flags & M_DECRYPTED)
2650 		goto noredhdropt;
2651 	if (p - (u_char *)ip6 > maxlen)
2652 		goto noredhdropt;
2653 
2654     {
2655 	/* redirected header option */
2656 	int len;
2657 	struct nd_opt_rd_hdr *nd_opt_rh;
2658 
2659 	/*
2660 	 * compute the maximum size for icmp6 redirect header option.
2661 	 * XXX room for auth header?
2662 	 */
2663 	len = maxlen - (p - (u_char *)ip6);
2664 	len &= ~7;
2665 
2666 	/* This is just for simplicity. */
2667 	if (m0->m_pkthdr.len != m0->m_len) {
2668 		if (m0->m_next) {
2669 			m_freem(m0->m_next);
2670 			m0->m_next = NULL;
2671 		}
2672 		m0->m_pkthdr.len = m0->m_len;
2673 	}
2674 
2675 	/*
2676 	 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2677 	 * about padding/truncate rule for the original IP packet.
2678 	 * From the discussion on IPv6imp in Feb 1999, the consensus was:
2679 	 * - "attach as much as possible" is the goal
2680 	 * - pad if not aligned (original size can be guessed by original
2681 	 *   ip6 header)
2682 	 * Following code adds the padding if it is simple enough,
2683 	 * and truncates if not.
2684 	 */
2685 	if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2686 		panic("assumption failed in %s:%d\n", __FILE__, __LINE__);
2687 
2688 	if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2689 		/* not enough room, truncate */
2690 		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2691 	} else {
2692 		/* enough room, pad or truncate */
2693 		size_t extra;
2694 
2695 		extra = m0->m_pkthdr.len % 8;
2696 		if (extra) {
2697 			/* pad if easy enough, truncate if not */
2698 			if (8 - extra <= M_TRAILINGSPACE(m0)) {
2699 				/* pad */
2700 				m0->m_len += (8 - extra);
2701 				m0->m_pkthdr.len += (8 - extra);
2702 			} else {
2703 				/* truncate */
2704 				m0->m_pkthdr.len -= extra;
2705 				m0->m_len -= extra;
2706 			}
2707 		}
2708 		len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2709 		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2710 	}
2711 
2712 	nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2713 	bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2714 	nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2715 	nd_opt_rh->nd_opt_rh_len = len >> 3;
2716 	p += sizeof(*nd_opt_rh);
2717 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2718 
2719 	/* connect m0 to m */
2720 	m->m_next = m0;
2721 	m->m_pkthdr.len = m->m_len + m0->m_len;
2722     }
2723 noredhdropt:;
2724 
2725 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
2726 		sip6->ip6_src.s6_addr16[1] = 0;
2727 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
2728 		sip6->ip6_dst.s6_addr16[1] = 0;
2729 #if 0
2730 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
2731 		ip6->ip6_src.s6_addr16[1] = 0;
2732 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
2733 		ip6->ip6_dst.s6_addr16[1] = 0;
2734 #endif
2735 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
2736 		nd_rd->nd_rd_target.s6_addr16[1] = 0;
2737 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
2738 		nd_rd->nd_rd_dst.s6_addr16[1] = 0;
2739 
2740 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2741 
2742 	nd_rd->nd_rd_cksum = 0;
2743 	nd_rd->nd_rd_cksum
2744 		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2745 
2746 	/* send the packet to outside... */
2747 #ifdef IPSEC
2748 	/* Don't lookup socket */
2749 	(void)ipsec_setsocket(m, NULL);
2750 #endif /* IPSEC */
2751 	if (ip6_output(m, NULL, NULL, 0, NULL, NULL) != 0)
2752 		icmp6_ifstat_inc(ifp, ifs6_out_error);
2753 
2754 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
2755 	icmp6_ifstat_inc(ifp, ifs6_out_redirect);
2756 	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2757 
2758 	return;
2759 
2760 fail:
2761 	if (m)
2762 		m_freem(m);
2763 	if (m0)
2764 		m_freem(m0);
2765 }
2766 
2767 /*
2768  * ICMPv6 socket option processing.
2769  */
2770 int
2771 icmp6_ctloutput(op, so, level, optname, mp)
2772 	int op;
2773 	struct socket *so;
2774 	int level, optname;
2775 	struct mbuf **mp;
2776 {
2777 	int error = 0;
2778 	int optlen;
2779 	struct in6pcb *in6p = sotoin6pcb(so);
2780 	struct mbuf *m = *mp;
2781 
2782 	optlen = m ? m->m_len : 0;
2783 
2784 	if (level != IPPROTO_ICMPV6) {
2785 		if (op == PRCO_SETOPT && m)
2786 			(void)m_free(m);
2787 		return EINVAL;
2788 	}
2789 
2790 	switch (op) {
2791 	case PRCO_SETOPT:
2792 		switch (optname) {
2793 		case ICMP6_FILTER:
2794 		    {
2795 			struct icmp6_filter *p;
2796 
2797 			if (optlen != sizeof(*p)) {
2798 				error = EMSGSIZE;
2799 				break;
2800 			}
2801 			p = mtod(m, struct icmp6_filter *);
2802 			if (!p || !in6p->in6p_icmp6filt) {
2803 				error = EINVAL;
2804 				break;
2805 			}
2806 			bcopy(p, in6p->in6p_icmp6filt,
2807 				sizeof(struct icmp6_filter));
2808 			error = 0;
2809 			break;
2810 		    }
2811 
2812 		default:
2813 			error = ENOPROTOOPT;
2814 			break;
2815 		}
2816 		if (m)
2817 			(void)m_freem(m);
2818 		break;
2819 
2820 	case PRCO_GETOPT:
2821 		switch (optname) {
2822 		case ICMP6_FILTER:
2823 		    {
2824 			struct icmp6_filter *p;
2825 
2826 			if (!in6p->in6p_icmp6filt) {
2827 				error = EINVAL;
2828 				break;
2829 			}
2830 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
2831 			m->m_len = sizeof(struct icmp6_filter);
2832 			p = mtod(m, struct icmp6_filter *);
2833 			bcopy(in6p->in6p_icmp6filt, p,
2834 				sizeof(struct icmp6_filter));
2835 			error = 0;
2836 			break;
2837 		    }
2838 
2839 		default:
2840 			error = ENOPROTOOPT;
2841 			break;
2842 		}
2843 		break;
2844 	}
2845 
2846 	return (error);
2847 }
2848 
2849 /*
2850  * Perform rate limit check.
2851  * Returns 0 if it is okay to send the icmp6 packet.
2852  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2853  * limitation.
2854  *
2855  * XXX per-destination/type check necessary?
2856  */
2857 static int
2858 icmp6_ratelimit(dst, type, code)
2859 	const struct in6_addr *dst;	/* not used at this moment */
2860 	const int type;			/* not used at this moment */
2861 	const int code;			/* not used at this moment */
2862 {
2863 	int ret;
2864 
2865 	ret = 0;	/* okay to send */
2866 
2867 	/* PPS limit */
2868 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2869 	    icmp6errppslim)) {
2870 		/* The packet is subject to rate limit */
2871 		ret++;
2872 	}
2873 
2874 	return ret;
2875 }
2876 
2877 static struct rtentry *
2878 icmp6_mtudisc_clone(dst)
2879 	struct sockaddr *dst;
2880 {
2881 	struct rtentry *rt;
2882 	int    error;
2883 
2884 	rt = rtalloc1(dst, 1);
2885 	if (rt == 0)
2886 		return NULL;
2887 
2888 	/* If we didn't get a host route, allocate one */
2889 	if ((rt->rt_flags & RTF_HOST) == 0) {
2890 		struct rtentry *nrt;
2891 
2892 		error = rtrequest((int) RTM_ADD, dst,
2893 		    (struct sockaddr *) rt->rt_gateway,
2894 		    (struct sockaddr *) 0,
2895 		    RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt);
2896 		if (error) {
2897 			rtfree(rt);
2898 			return NULL;
2899 		}
2900 		nrt->rt_rmx = rt->rt_rmx;
2901 		rtfree(rt);
2902 		rt = nrt;
2903 	}
2904 	error = rt_timer_add(rt, icmp6_mtudisc_timeout,
2905 			icmp6_mtudisc_timeout_q);
2906 	if (error) {
2907 		rtfree(rt);
2908 		return NULL;
2909 	}
2910 
2911 	return rt;	/* caller need to call rtfree() */
2912 }
2913 
2914 static void
2915 icmp6_mtudisc_timeout(rt, r)
2916 	struct rtentry *rt;
2917 	struct rttimer *r;
2918 {
2919 	if (rt == NULL)
2920 		panic("icmp6_mtudisc_timeout: bad route to timeout");
2921 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
2922 	    (RTF_DYNAMIC | RTF_HOST)) {
2923 		rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt),
2924 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
2925 	} else {
2926 		if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
2927 			rt->rt_rmx.rmx_mtu = 0;
2928 	}
2929 }
2930 
2931 static void
2932 icmp6_redirect_timeout(rt, r)
2933 	struct rtentry *rt;
2934 	struct rttimer *r;
2935 {
2936 	if (rt == NULL)
2937 		panic("icmp6_redirect_timeout: bad route to timeout");
2938 	if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) ==
2939 	    (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) {
2940 		rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt),
2941 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
2942 	}
2943 }
2944 
2945 int
2946 icmp6_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
2947 	int *name;
2948 	u_int namelen;
2949 	void *oldp;
2950 	size_t *oldlenp;
2951 	void *newp;
2952 	size_t newlen;
2953 {
2954 
2955 	/* All sysctl names at this level are terminal. */
2956 	if (namelen != 1)
2957 		return ENOTDIR;
2958 
2959 	switch (name[0]) {
2960 
2961 	case ICMPV6CTL_REDIRACCEPT:
2962 		return sysctl_int(oldp, oldlenp, newp, newlen,
2963 				&icmp6_rediraccept);
2964 	case ICMPV6CTL_REDIRTIMEOUT:
2965 		return sysctl_int(oldp, oldlenp, newp, newlen,
2966 				&icmp6_redirtimeout);
2967 	case ICMPV6CTL_STATS:
2968 		return sysctl_rdstruct(oldp, oldlenp, newp,
2969 				&icmp6stat, sizeof(icmp6stat));
2970 	case ICMPV6CTL_ND6_PRUNE:
2971 		return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_prune);
2972 	case ICMPV6CTL_ND6_DELAY:
2973 		return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_delay);
2974 	case ICMPV6CTL_ND6_UMAXTRIES:
2975 		return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_umaxtries);
2976 	case ICMPV6CTL_ND6_MMAXTRIES:
2977 		return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_mmaxtries);
2978 	case ICMPV6CTL_ND6_USELOOPBACK:
2979 		return sysctl_int(oldp, oldlenp, newp, newlen,
2980 				&nd6_useloopback);
2981 	case ICMPV6CTL_NODEINFO:
2982 		return sysctl_int(oldp, oldlenp, newp, newlen, &icmp6_nodeinfo);
2983 	case ICMPV6CTL_ERRPPSLIMIT:
2984 		return sysctl_int(oldp, oldlenp, newp, newlen, &icmp6errppslim);
2985 	case ICMPV6CTL_ND6_MAXNUDHINT:
2986 		return sysctl_int(oldp, oldlenp, newp, newlen,
2987 				&nd6_maxnudhint);
2988 	case ICMPV6CTL_MTUDISC_HIWAT:
2989 		return sysctl_int(oldp, oldlenp, newp, newlen,
2990 				&icmp6_mtudisc_hiwat);
2991 	case ICMPV6CTL_MTUDISC_LOWAT:
2992 		return sysctl_int(oldp, oldlenp, newp, newlen,
2993 				&icmp6_mtudisc_lowat);
2994 	case ICMPV6CTL_ND6_DEBUG:
2995 		return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_debug);
2996 	case ICMPV6CTL_ND6_DRLIST:
2997 	case ICMPV6CTL_ND6_PRLIST:
2998 		return nd6_sysctl(name[0], oldp, oldlenp, newp, newlen);
2999 	default:
3000 		return ENOPROTOOPT;
3001 	}
3002 	/* NOTREACHED */
3003 }
3004