xref: /netbsd-src/sys/netinet6/icmp6.c (revision 404fbe5fb94ca1e054339640cabb2801ce52dd30)
1 /*	$NetBSD: icmp6.c,v 1.150 2008/10/03 08:23:06 adrianp 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. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
62  */
63 
64 #include <sys/cdefs.h>
65 __KERNEL_RCSID(0, "$NetBSD: icmp6.c,v 1.150 2008/10/03 08:23:06 adrianp Exp $");
66 
67 #include "opt_inet.h"
68 #include "opt_ipsec.h"
69 
70 #include <sys/param.h>
71 #include <sys/systm.h>
72 #include <sys/malloc.h>
73 #include <sys/mbuf.h>
74 #include <sys/protosw.h>
75 #include <sys/socket.h>
76 #include <sys/socketvar.h>
77 #include <sys/time.h>
78 #include <sys/kernel.h>
79 #include <sys/syslog.h>
80 #include <sys/domain.h>
81 #include <sys/sysctl.h>
82 
83 #include <net/if.h>
84 #include <net/route.h>
85 #include <net/if_dl.h>
86 #include <net/if_types.h>
87 
88 #include <netinet/in.h>
89 #include <netinet/in_var.h>
90 #include <netinet/ip6.h>
91 #include <netinet6/ip6_var.h>
92 #include <netinet6/ip6_private.h>
93 #include <netinet/icmp6.h>
94 #include <netinet6/icmp6_private.h>
95 #include <netinet6/mld6_var.h>
96 #include <netinet6/in6_pcb.h>
97 #include <netinet6/nd6.h>
98 #include <netinet6/in6_ifattach.h>
99 #include <netinet6/ip6protosw.h>
100 #include <netinet6/scope6_var.h>
101 
102 #ifdef IPSEC
103 #include <netinet6/ipsec.h>
104 #include <netkey/key.h>
105 #endif
106 
107 #ifdef FAST_IPSEC
108 #include <netipsec/ipsec.h>
109 #include <netipsec/key.h>
110 #endif
111 
112 
113 #include "faith.h"
114 #if defined(NFAITH) && 0 < NFAITH
115 #include <net/if_faith.h>
116 #endif
117 
118 #include <net/net_osdep.h>
119 
120 extern struct domain inet6domain;
121 
122 percpu_t *icmp6stat_percpu;
123 
124 extern struct inpcbtable raw6cbtable;
125 extern int icmp6errppslim;
126 static int icmp6errpps_count = 0;
127 static struct timeval icmp6errppslim_last;
128 extern int icmp6_nodeinfo;
129 
130 /*
131  * List of callbacks to notify when Path MTU changes are made.
132  */
133 struct icmp6_mtudisc_callback {
134 	LIST_ENTRY(icmp6_mtudisc_callback) mc_list;
135 	void (*mc_func)(struct in6_addr *);
136 };
137 
138 LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks =
139     LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks);
140 
141 static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL;
142 extern int pmtu_expire;
143 
144 /* XXX do these values make any sense? */
145 static int icmp6_mtudisc_hiwat = 1280;
146 static int icmp6_mtudisc_lowat = 256;
147 
148 /*
149  * keep track of # of redirect routes.
150  */
151 static struct rttimer_queue *icmp6_redirect_timeout_q = NULL;
152 
153 /* XXX experimental, turned off */
154 static int icmp6_redirect_hiwat = -1;
155 static int icmp6_redirect_lowat = -1;
156 
157 static void icmp6_errcount(u_int, int, int);
158 static int icmp6_rip6_input(struct mbuf **, int);
159 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
160 static const char *icmp6_redirect_diag(struct in6_addr *,
161 	struct in6_addr *, struct in6_addr *);
162 static struct mbuf *ni6_input(struct mbuf *, int);
163 static struct mbuf *ni6_nametodns(const char *, int, int);
164 static int ni6_dnsmatch(const char *, int, const char *, int);
165 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
166 			  struct ifnet **, char *);
167 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
168 				struct ifnet *, int);
169 static int icmp6_notify_error(struct mbuf *, int, int, int);
170 static struct rtentry *icmp6_mtudisc_clone(struct sockaddr *);
171 static void icmp6_mtudisc_timeout(struct rtentry *, struct rttimer *);
172 static void icmp6_redirect_timeout(struct rtentry *, struct rttimer *);
173 
174 
175 void
176 icmp6_init(void)
177 {
178 	mld_init();
179 	icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire);
180 	icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout);
181 
182 	icmp6stat_percpu = percpu_alloc(sizeof(uint64_t) * ICMP6_NSTATS);
183 }
184 
185 static void
186 icmp6_errcount(u_int base, int type, int code)
187 {
188 	switch (type) {
189 	case ICMP6_DST_UNREACH:
190 		switch (code) {
191 		case ICMP6_DST_UNREACH_NOROUTE:
192 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOROUTE);
193 			return;
194 		case ICMP6_DST_UNREACH_ADMIN:
195 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADMIN);
196 			return;
197 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
198 			ICMP6_STATINC(base +
199 				      ICMP6_ERRSTAT_DST_UNREACH_BEYONDSCOPE);
200 			return;
201 		case ICMP6_DST_UNREACH_ADDR:
202 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADDR);
203 			return;
204 		case ICMP6_DST_UNREACH_NOPORT:
205 			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOPORT);
206 			return;
207 		}
208 		break;
209 	case ICMP6_PACKET_TOO_BIG:
210 		ICMP6_STATINC(base + ICMP6_ERRSTAT_PACKET_TOO_BIG);
211 		return;
212 	case ICMP6_TIME_EXCEEDED:
213 		switch (code) {
214 		case ICMP6_TIME_EXCEED_TRANSIT:
215 			ICMP6_STATINC(base + ICMP6_ERRSTAT_TIME_EXCEED_TRANSIT);
216 			return;
217 		case ICMP6_TIME_EXCEED_REASSEMBLY:
218 			ICMP6_STATINC(base +
219 				      ICMP6_ERRSTAT_TIME_EXCEED_REASSEMBLY);
220 			return;
221 		}
222 		break;
223 	case ICMP6_PARAM_PROB:
224 		switch (code) {
225 		case ICMP6_PARAMPROB_HEADER:
226 			ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_HEADER);
227 			return;
228 		case ICMP6_PARAMPROB_NEXTHEADER:
229 			ICMP6_STATINC(base +
230 				      ICMP6_ERRSTAT_PARAMPROB_NEXTHEADER);
231 			return;
232 		case ICMP6_PARAMPROB_OPTION:
233 			ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_OPTION);
234 			return;
235 		}
236 		break;
237 	case ND_REDIRECT:
238 		ICMP6_STATINC(base + ICMP6_ERRSTAT_REDIRECT);
239 		return;
240 	}
241 	ICMP6_STATINC(base + ICMP6_ERRSTAT_UNKNOWN);
242 }
243 
244 /*
245  * Register a Path MTU Discovery callback.
246  */
247 void
248 icmp6_mtudisc_callback_register(void (*func)(struct in6_addr *))
249 {
250 	struct icmp6_mtudisc_callback *mc;
251 
252 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
253 	     mc = LIST_NEXT(mc, mc_list)) {
254 		if (mc->mc_func == func)
255 			return;
256 	}
257 
258 	mc = malloc(sizeof(*mc), M_PCB, M_NOWAIT);
259 	if (mc == NULL)
260 		panic("icmp6_mtudisc_callback_register");
261 
262 	mc->mc_func = func;
263 	LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list);
264 }
265 
266 /*
267  * A wrapper function for icmp6_error() necessary when the erroneous packet
268  * may not contain enough scope zone information.
269  */
270 void
271 icmp6_error2(struct mbuf *m, int type, int code, int param,
272 	struct ifnet *ifp)
273 {
274 	struct ip6_hdr *ip6;
275 
276 	if (ifp == NULL)
277 		return;
278 
279 	if (m->m_len < sizeof(struct ip6_hdr)) {
280 		m = m_pullup(m, sizeof(struct ip6_hdr));
281 		if (m == NULL)
282 			return;
283 	}
284 
285 	ip6 = mtod(m, struct ip6_hdr *);
286 
287 	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
288 		return;
289 	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
290 		return;
291 
292 	icmp6_error(m, type, code, param);
293 }
294 
295 /*
296  * Generate an error packet of type error in response to bad IP6 packet.
297  */
298 void
299 icmp6_error(struct mbuf *m, int type, int code, int param)
300 {
301 	struct ip6_hdr *oip6, *nip6;
302 	struct icmp6_hdr *icmp6;
303 	u_int preplen;
304 	int off;
305 	int nxt;
306 
307 	ICMP6_STATINC(ICMP6_STAT_ERROR);
308 
309 	/* count per-type-code statistics */
310 	icmp6_errcount(ICMP6_STAT_OUTERRHIST, type, code);
311 
312 	if (m->m_flags & M_DECRYPTED) {
313 		ICMP6_STATINC(ICMP6_STAT_CANTERROR);
314 		goto freeit;
315 	}
316 
317 	if (M_UNWRITABLE(m, sizeof(struct ip6_hdr)) &&
318 	    (m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL)
319 		return;
320 	oip6 = mtod(m, struct ip6_hdr *);
321 
322 	/*
323 	 * If the destination address of the erroneous packet is a multicast
324 	 * address, or the packet was sent using link-layer multicast,
325 	 * we should basically suppress sending an error (RFC 2463, Section
326 	 * 2.4).
327 	 * We have two exceptions (the item e.2 in that section):
328 	 * - the Pakcet Too Big message can be sent for path MTU discovery.
329 	 * - the Parameter Problem Message that can be allowed an icmp6 error
330 	 *   in the option type field.  This check has been done in
331 	 *   ip6_unknown_opt(), so we can just check the type and code.
332 	 */
333 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
334 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
335 	    (type != ICMP6_PACKET_TOO_BIG &&
336 	     (type != ICMP6_PARAM_PROB ||
337 	      code != ICMP6_PARAMPROB_OPTION)))
338 		goto freeit;
339 
340 	/*
341 	 * RFC 2463, 2.4 (e.5): source address check.
342 	 * XXX: the case of anycast source?
343 	 */
344 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
345 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
346 		goto freeit;
347 
348 	/*
349 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
350 	 * don't do it.
351 	 */
352 	nxt = -1;
353 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
354 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
355 		struct icmp6_hdr *icp;
356 
357 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
358 			sizeof(*icp));
359 		if (icp == NULL) {
360 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
361 			return;
362 		}
363 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
364 		    icp->icmp6_type == ND_REDIRECT) {
365 			/*
366 			 * ICMPv6 error
367 			 * Special case: for redirect (which is
368 			 * informational) we must not send icmp6 error.
369 			 */
370 			ICMP6_STATINC(ICMP6_STAT_CANTERROR);
371 			goto freeit;
372 		} else {
373 			/* ICMPv6 informational - send the error */
374 		}
375 	}
376 #if 0 /* controversial */
377 	else if (off >= 0 && nxt == IPPROTO_ESP) {
378 		/*
379 		 * It could be ICMPv6 error inside ESP.  Take a safer side,
380 		 * don't respond.
381 		 */
382 		ICMP6_STATINC(ICMP6_STAT_CANTERROR);
383 		goto freeit;
384 	}
385 #endif
386 	else {
387 		/* non-ICMPv6 - send the error */
388 	}
389 
390 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
391 
392 	/* Finally, do rate limitation check. */
393 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
394 		ICMP6_STATINC(ICMP6_STAT_TOOFREQ);
395 		goto freeit;
396 	}
397 
398 	/*
399 	 * OK, ICMP6 can be generated.
400 	 */
401 
402 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
403 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
404 
405 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
406 	M_PREPEND(m, preplen, M_DONTWAIT);
407 	if (m && M_UNWRITABLE(m, preplen))
408 		m = m_pullup(m, preplen);
409 	if (m == NULL) {
410 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
411 		return;
412 	}
413 
414 	nip6 = mtod(m, struct ip6_hdr *);
415 	nip6->ip6_src  = oip6->ip6_src;
416 	nip6->ip6_dst  = oip6->ip6_dst;
417 
418 	in6_clearscope(&oip6->ip6_src);
419 	in6_clearscope(&oip6->ip6_dst);
420 
421 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
422 	icmp6->icmp6_type = type;
423 	icmp6->icmp6_code = code;
424 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
425 
426 	/*
427 	 * icmp6_reflect() is designed to be in the input path.
428 	 * icmp6_error() can be called from both input and output path,
429 	 * and if we are in output path rcvif could contain bogus value.
430 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
431 	 * information in ip header (nip6).
432 	 */
433 	m->m_pkthdr.rcvif = NULL;
434 
435 	ICMP6_STATINC(ICMP6_STAT_OUTHIST + type);
436 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
437 
438 	return;
439 
440   freeit:
441 	/*
442 	 * If we can't tell whether or not we can generate ICMP6, free it.
443 	 */
444 	m_freem(m);
445 }
446 
447 /*
448  * Process a received ICMP6 message.
449  */
450 int
451 icmp6_input(struct mbuf **mp, int *offp, int proto)
452 {
453 	struct mbuf *m = *mp, *n;
454 	struct ip6_hdr *ip6, *nip6;
455 	struct icmp6_hdr *icmp6, *nicmp6;
456 	int off = *offp;
457 	int icmp6len = m->m_pkthdr.len - *offp;
458 	int code, sum, noff, i;
459 
460 #define ICMP6_MAXLEN (sizeof(*nip6) + sizeof(*nicmp6) + 4)
461 	KASSERT(ICMP6_MAXLEN < MCLBYTES);
462 	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
463 
464 	/*
465 	 * Locate icmp6 structure in mbuf, and check
466 	 * that not corrupted and of at least minimum length
467 	 */
468 
469 	if (icmp6len < sizeof(struct icmp6_hdr)) {
470 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
471 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
472 		goto freeit;
473 	}
474 
475 	i = off + sizeof(*icmp6);
476 	if ((m->m_len < i || M_READONLY(m)) && (m = m_pullup(m, i)) == 0) {
477 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
478 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
479 		goto freeit;
480 	}
481 	ip6 = mtod(m, struct ip6_hdr *);
482 	/*
483 	 * calculate the checksum
484 	 */
485 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
486 	if (icmp6 == NULL) {
487 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
488 		/* m is invalid */
489 		/*icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);*/
490 		return IPPROTO_DONE;
491 	}
492 	KASSERT(IP6_HDR_ALIGNED_P(icmp6));
493 	code = icmp6->icmp6_code;
494 
495 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
496 		nd6log((LOG_ERR,
497 		    "ICMP6 checksum error(%d|%x) %s\n",
498 		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
499 		ICMP6_STATINC(ICMP6_STAT_CHECKSUM);
500 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
501 		goto freeit;
502 	}
503 
504 #if defined(NFAITH) && 0 < NFAITH
505 	if (faithprefix(&ip6->ip6_dst)) {
506 		/*
507 		 * Deliver very specific ICMP6 type only.
508 		 * This is important to deliver TOOBIG.  Otherwise PMTUD
509 		 * will not work.
510 		 */
511 		switch (icmp6->icmp6_type) {
512 		case ICMP6_DST_UNREACH:
513 		case ICMP6_PACKET_TOO_BIG:
514 		case ICMP6_TIME_EXCEEDED:
515 			break;
516 		default:
517 			goto freeit;
518 		}
519 	}
520 #endif
521 
522 	ICMP6_STATINC(ICMP6_STAT_INHIST + icmp6->icmp6_type);
523 
524 	switch (icmp6->icmp6_type) {
525 	case ICMP6_DST_UNREACH:
526 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
527 		switch (code) {
528 		case ICMP6_DST_UNREACH_NOROUTE:
529 			code = PRC_UNREACH_NET;
530 			break;
531 		case ICMP6_DST_UNREACH_ADMIN:
532 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
533 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
534 			break;
535 		case ICMP6_DST_UNREACH_ADDR:
536 			code = PRC_HOSTDEAD;
537 			break;
538 #ifdef COMPAT_RFC1885
539 		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
540 			code = PRC_UNREACH_SRCFAIL;
541 			break;
542 #else
543 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
544 			/* I mean "source address was incorrect." */
545 			code = PRC_UNREACH_NET;
546 			break;
547 #endif
548 		case ICMP6_DST_UNREACH_NOPORT:
549 			code = PRC_UNREACH_PORT;
550 			break;
551 		default:
552 			goto badcode;
553 		}
554 		goto deliver;
555 
556 	case ICMP6_PACKET_TOO_BIG:
557 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
558 
559 		/*
560 		 * MTU is checked in icmp6_mtudisc.
561 		 */
562 		code = PRC_MSGSIZE;
563 
564 		/*
565 		 * Updating the path MTU will be done after examining
566 		 * intermediate extension headers.
567 		 */
568 		goto deliver;
569 
570 	case ICMP6_TIME_EXCEEDED:
571 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
572 		switch (code) {
573 		case ICMP6_TIME_EXCEED_TRANSIT:
574 			code = PRC_TIMXCEED_INTRANS;
575 			break;
576 		case ICMP6_TIME_EXCEED_REASSEMBLY:
577 			code = PRC_TIMXCEED_REASS;
578 			break;
579 		default:
580 			goto badcode;
581 		}
582 		goto deliver;
583 
584 	case ICMP6_PARAM_PROB:
585 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
586 		switch (code) {
587 		case ICMP6_PARAMPROB_NEXTHEADER:
588 			code = PRC_UNREACH_PROTOCOL;
589 			break;
590 		case ICMP6_PARAMPROB_HEADER:
591 		case ICMP6_PARAMPROB_OPTION:
592 			code = PRC_PARAMPROB;
593 			break;
594 		default:
595 			goto badcode;
596 		}
597 		goto deliver;
598 
599 	case ICMP6_ECHO_REQUEST:
600 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
601 		if (code != 0)
602 			goto badcode;
603 		/*
604 		 * Copy mbuf to send to two data paths: userland socket(s),
605 		 * and to the querier (echo reply).
606 		 * m: a copy for socket, n: a copy for querier
607 		 *
608 		 * If the first mbuf is shared, or the first mbuf is too short,
609 		 * copy the first part of the data into a fresh mbuf.
610 		 * Otherwise, we will wrongly overwrite both copies.
611 		 */
612 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
613 			/* Give up local */
614 			n = m;
615 			m = NULL;
616 		} else if (M_READONLY(n) ||
617 		    n->m_len < off + sizeof(struct icmp6_hdr)) {
618 			struct mbuf *n0 = n;
619 
620 			/*
621 			 * Prepare an internal mbuf.  m_pullup() doesn't
622 			 * always copy the length we specified.
623 			 */
624 			if ((n = m_dup(n0, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
625 				/* Give up local */
626 				n = m;
627 				m = NULL;
628 			}
629 			m_freem(n0);
630 		}
631 		IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
632 		    sizeof(*nicmp6));
633 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
634 		nicmp6->icmp6_code = 0;
635 		if (n) {
636 			uint64_t *icmp6s = ICMP6_STAT_GETREF();
637 			icmp6s[ICMP6_STAT_REFLECT]++;
638 			icmp6s[ICMP6_STAT_OUTHIST + ICMP6_ECHO_REPLY]++;
639 			ICMP6_STAT_PUTREF();
640 			icmp6_reflect(n, off);
641 		}
642 		if (!m)
643 			goto freeit;
644 		break;
645 
646 	case ICMP6_ECHO_REPLY:
647 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
648 		if (code != 0)
649 			goto badcode;
650 		break;
651 
652 	case MLD_LISTENER_QUERY:
653 	case MLD_LISTENER_REPORT:
654 		if (icmp6len < sizeof(struct mld_hdr))
655 			goto badlen;
656 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
657 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
658 		else
659 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
660 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
661 			/* give up local */
662 			mld_input(m, off);
663 			m = NULL;
664 			goto freeit;
665 		}
666 		mld_input(n, off);
667 		/* m stays. */
668 		break;
669 
670 	case MLD_LISTENER_DONE:
671 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
672 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
673 			goto badlen;
674 		break;		/* nothing to be done in kernel */
675 
676 	case MLD_MTRACE_RESP:
677 	case MLD_MTRACE:
678 		/* XXX: these two are experimental.  not officially defined. */
679 		/* XXX: per-interface statistics? */
680 		break;		/* just pass it to applications */
681 
682 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
683 	    {
684 		enum { WRU, FQDN } mode;
685 
686 		if (!icmp6_nodeinfo)
687 			break;
688 
689 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
690 			mode = WRU;
691 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
692 			mode = FQDN;
693 		else
694 			goto badlen;
695 
696 		if (mode == FQDN) {
697 			n = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
698 			if (n)
699 				n = ni6_input(n, off);
700 			/* XXX meaningless if n == NULL */
701 			noff = sizeof(struct ip6_hdr);
702 		} else {
703 			u_char *p;
704 			int maxhlen;
705 
706 			if ((icmp6_nodeinfo & 5) != 5)
707 				break;
708 
709 			if (code != 0)
710 				goto badcode;
711 			MGETHDR(n, M_DONTWAIT, m->m_type);
712 			if (n && ICMP6_MAXLEN > MHLEN) {
713 				MCLGET(n, M_DONTWAIT);
714 				if ((n->m_flags & M_EXT) == 0) {
715 					m_free(n);
716 					n = NULL;
717 				}
718 			}
719 			if (n == NULL) {
720 				/* Give up remote */
721 				break;
722 			}
723 			n->m_pkthdr.rcvif = NULL;
724 			n->m_len = 0;
725 			maxhlen = M_TRAILINGSPACE(n) - ICMP6_MAXLEN;
726 			if (maxhlen > hostnamelen)
727 				maxhlen = hostnamelen;
728 			/*
729 			 * Copy IPv6 and ICMPv6 only.
730 			 */
731 			nip6 = mtod(n, struct ip6_hdr *);
732 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
733 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
734 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
735 			p = (u_char *)(nicmp6 + 1);
736 			bzero(p, 4);
737 			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
738 			noff = sizeof(struct ip6_hdr);
739 			M_COPY_PKTHDR(n, m); /* just for rcvif */
740 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
741 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
742 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
743 			nicmp6->icmp6_code = 0;
744 		}
745 #undef hostnamelen
746 		if (n) {
747 			uint64_t *icmp6s = ICMP6_STAT_GETREF();
748 			icmp6s[ICMP6_STAT_REFLECT]++;
749 			icmp6s[ICMP6_STAT_OUTHIST + ICMP6_WRUREPLY]++;
750 			ICMP6_STAT_PUTREF();
751 			icmp6_reflect(n, noff);
752 		}
753 		break;
754 	    }
755 
756 	case ICMP6_WRUREPLY:
757 		if (code != 0)
758 			goto badcode;
759 		break;
760 
761 	case ND_ROUTER_SOLICIT:
762 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
763 		if (code != 0)
764 			goto badcode;
765 		if (icmp6len < sizeof(struct nd_router_solicit))
766 			goto badlen;
767 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
768 			/* give up local */
769 			nd6_rs_input(m, off, icmp6len);
770 			m = NULL;
771 			goto freeit;
772 		}
773 		nd6_rs_input(n, off, icmp6len);
774 		/* m stays. */
775 		break;
776 
777 	case ND_ROUTER_ADVERT:
778 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
779 		if (code != 0)
780 			goto badcode;
781 		if (icmp6len < sizeof(struct nd_router_advert))
782 			goto badlen;
783 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
784 			/* give up local */
785 			nd6_ra_input(m, off, icmp6len);
786 			m = NULL;
787 			goto freeit;
788 		}
789 		nd6_ra_input(n, off, icmp6len);
790 		/* m stays. */
791 		break;
792 
793 	case ND_NEIGHBOR_SOLICIT:
794 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
795 		if (code != 0)
796 			goto badcode;
797 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
798 			goto badlen;
799 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
800 			/* give up local */
801 			nd6_ns_input(m, off, icmp6len);
802 			m = NULL;
803 			goto freeit;
804 		}
805 		nd6_ns_input(n, off, icmp6len);
806 		/* m stays. */
807 		break;
808 
809 	case ND_NEIGHBOR_ADVERT:
810 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
811 		if (code != 0)
812 			goto badcode;
813 		if (icmp6len < sizeof(struct nd_neighbor_advert))
814 			goto badlen;
815 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
816 			/* give up local */
817 			nd6_na_input(m, off, icmp6len);
818 			m = NULL;
819 			goto freeit;
820 		}
821 		nd6_na_input(n, off, icmp6len);
822 		/* m stays. */
823 		break;
824 
825 	case ND_REDIRECT:
826 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
827 		if (code != 0)
828 			goto badcode;
829 		if (icmp6len < sizeof(struct nd_redirect))
830 			goto badlen;
831 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
832 			/* give up local */
833 			icmp6_redirect_input(m, off);
834 			m = NULL;
835 			goto freeit;
836 		}
837 		icmp6_redirect_input(n, off);
838 		/* m stays. */
839 		break;
840 
841 	case ICMP6_ROUTER_RENUMBERING:
842 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
843 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
844 			goto badcode;
845 		if (icmp6len < sizeof(struct icmp6_router_renum))
846 			goto badlen;
847 		break;
848 
849 	default:
850 		nd6log((LOG_DEBUG,
851 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
852 		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
853 		    ip6_sprintf(&ip6->ip6_dst),
854 		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
855 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
856 			/* ICMPv6 error: MUST deliver it by spec... */
857 			code = PRC_NCMDS;
858 			/* deliver */
859 		} else {
860 			/* ICMPv6 informational: MUST not deliver */
861 			break;
862 		}
863 	deliver:
864 		if (icmp6_notify_error(m, off, icmp6len, code)) {
865 			/* In this case, m should've been freed. */
866 			return (IPPROTO_DONE);
867 		}
868 		break;
869 
870 	badcode:
871 		ICMP6_STATINC(ICMP6_STAT_BADCODE);
872 		break;
873 
874 	badlen:
875 		ICMP6_STATINC(ICMP6_STAT_BADLEN);
876 		break;
877 	}
878 
879 	/* deliver the packet to appropriate sockets */
880 	icmp6_rip6_input(&m, *offp);
881 
882 	return IPPROTO_DONE;
883 
884  freeit:
885 	m_freem(m);
886 	return IPPROTO_DONE;
887 }
888 
889 static int
890 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
891 {
892 	struct icmp6_hdr *icmp6;
893 	struct ip6_hdr *eip6;
894 	u_int32_t notifymtu;
895 	struct sockaddr_in6 icmp6src, icmp6dst;
896 
897 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
898 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
899 		goto freeit;
900 	}
901 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
902 		       sizeof(*icmp6) + sizeof(struct ip6_hdr));
903 	if (icmp6 == NULL) {
904 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
905 		return (-1);
906 	}
907 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
908 
909 	/* Detect the upper level protocol */
910 	{
911 		void (*ctlfunc)(int, struct sockaddr *, void *);
912 		u_int8_t nxt = eip6->ip6_nxt;
913 		int eoff = off + sizeof(struct icmp6_hdr) +
914 			sizeof(struct ip6_hdr);
915 		struct ip6ctlparam ip6cp;
916 		struct in6_addr *finaldst = NULL;
917 		int icmp6type = icmp6->icmp6_type;
918 		struct ip6_frag *fh;
919 		struct ip6_rthdr *rth;
920 		struct ip6_rthdr0 *rth0;
921 		int rthlen;
922 
923 		while (1) { /* XXX: should avoid infinite loop explicitly? */
924 			struct ip6_ext *eh;
925 
926 			switch (nxt) {
927 			case IPPROTO_HOPOPTS:
928 			case IPPROTO_DSTOPTS:
929 			case IPPROTO_AH:
930 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
931 					       eoff, sizeof(*eh));
932 				if (eh == NULL) {
933 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
934 					return (-1);
935 				}
936 
937 				if (nxt == IPPROTO_AH)
938 					eoff += (eh->ip6e_len + 2) << 2;
939 				else
940 					eoff += (eh->ip6e_len + 1) << 3;
941 				nxt = eh->ip6e_nxt;
942 				break;
943 			case IPPROTO_ROUTING:
944 				/*
945 				 * When the erroneous packet contains a
946 				 * routing header, we should examine the
947 				 * header to determine the final destination.
948 				 * Otherwise, we can't properly update
949 				 * information that depends on the final
950 				 * destination (e.g. path MTU).
951 				 */
952 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
953 					       eoff, sizeof(*rth));
954 				if (rth == NULL) {
955 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
956 					return (-1);
957 				}
958 				rthlen = (rth->ip6r_len + 1) << 3;
959 				/*
960 				 * XXX: currently there is no
961 				 * officially defined type other
962 				 * than type-0.
963 				 * Note that if the segment left field
964 				 * is 0, all intermediate hops must
965 				 * have been passed.
966 				 */
967 				if (rth->ip6r_segleft &&
968 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
969 					int hops;
970 
971 					IP6_EXTHDR_GET(rth0,
972 						       struct ip6_rthdr0 *, m,
973 						       eoff, rthlen);
974 					if (rth0 == NULL) {
975 						ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
976 						return (-1);
977 					}
978 					/* just ignore a bogus header */
979 					if ((rth0->ip6r0_len % 2) == 0 &&
980 					    (hops = rth0->ip6r0_len/2))
981 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
982 				}
983 				eoff += rthlen;
984 				nxt = rth->ip6r_nxt;
985 				break;
986 			case IPPROTO_FRAGMENT:
987 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
988 					       eoff, sizeof(*fh));
989 				if (fh == NULL) {
990 					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
991 					return (-1);
992 				}
993 				/*
994 				 * Data after a fragment header is meaningless
995 				 * unless it is the first fragment, but
996 				 * we'll go to the notify label for path MTU
997 				 * discovery.
998 				 */
999 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1000 					goto notify;
1001 
1002 				eoff += sizeof(struct ip6_frag);
1003 				nxt = fh->ip6f_nxt;
1004 				break;
1005 			default:
1006 				/*
1007 				 * This case includes ESP and the No Next
1008 				 * Header.  In such cases going to the notify
1009 				 * label does not have any meaning
1010 				 * (i.e. ctlfunc will be NULL), but we go
1011 				 * anyway since we might have to update
1012 				 * path MTU information.
1013 				 */
1014 				goto notify;
1015 			}
1016 		}
1017 	  notify:
1018 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1019 			       sizeof(*icmp6) + sizeof(struct ip6_hdr));
1020 		if (icmp6 == NULL) {
1021 			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
1022 			return (-1);
1023 		}
1024 
1025 		/*
1026 		 * retrieve parameters from the inner IPv6 header, and convert
1027 		 * them into sockaddr structures.
1028 		 * XXX: there is no guarantee that the source or destination
1029 		 * addresses of the inner packet are in the same scope zone as
1030 		 * the addresses of the icmp packet.  But there is no other
1031 		 * way to determine the zone.
1032 		 */
1033 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1034 
1035 		sockaddr_in6_init(&icmp6dst,
1036 		    (finaldst == NULL) ? &eip6->ip6_dst : finaldst, 0, 0, 0);
1037 		if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1038 			goto freeit;
1039 		sockaddr_in6_init(&icmp6src, &eip6->ip6_src, 0, 0, 0);
1040 		if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1041 			goto freeit;
1042 		icmp6src.sin6_flowinfo =
1043 			(eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1044 
1045 		if (finaldst == NULL)
1046 			finaldst = &eip6->ip6_dst;
1047 		ip6cp.ip6c_m = m;
1048 		ip6cp.ip6c_icmp6 = icmp6;
1049 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1050 		ip6cp.ip6c_off = eoff;
1051 		ip6cp.ip6c_finaldst = finaldst;
1052 		ip6cp.ip6c_src = &icmp6src;
1053 		ip6cp.ip6c_nxt = nxt;
1054 
1055 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1056 			notifymtu = ntohl(icmp6->icmp6_mtu);
1057 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1058 		}
1059 
1060 		ctlfunc = (void (*)(int, struct sockaddr *, void *))
1061 			(inet6sw[ip6_protox[nxt]].pr_ctlinput);
1062 		if (ctlfunc) {
1063 			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1064 					  &ip6cp);
1065 		}
1066 	}
1067 	return (0);
1068 
1069   freeit:
1070 	m_freem(m);
1071 	return (-1);
1072 }
1073 
1074 void
1075 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1076 {
1077 	unsigned long rtcount;
1078 	struct icmp6_mtudisc_callback *mc;
1079 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1080 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1081 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1082 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1083 	struct rtentry *rt = NULL;
1084 	struct sockaddr_in6 sin6;
1085 
1086 	/*
1087 	 * The MTU should not be less than the minimal IPv6 MTU except for the
1088 	 * hack in ip6_output/ip6_setpmtu where we always include a frag header.
1089 	 * In that one case, the MTU might be less than 1280.
1090 	 */
1091 	if (__predict_false(mtu < IPV6_MMTU - sizeof(struct ip6_frag))) {
1092 		/* is the mtu even sane? */
1093 		if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1094 			return;
1095 		if (!validated)
1096 			return;
1097 		mtu = IPV6_MMTU - sizeof(struct ip6_frag);
1098 	}
1099 
1100 	/*
1101 	 * allow non-validated cases if memory is plenty, to make traffic
1102 	 * from non-connected pcb happy.
1103 	 */
1104 	rtcount = rt_timer_count(icmp6_mtudisc_timeout_q);
1105 	if (validated) {
1106 		if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat)
1107 			return;
1108 		else if (0 <= icmp6_mtudisc_lowat &&
1109 		    rtcount > icmp6_mtudisc_lowat) {
1110 			/*
1111 			 * XXX nuke a victim, install the new one.
1112 			 */
1113 		}
1114 	} else {
1115 		if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat)
1116 			return;
1117 	}
1118 
1119 	bzero(&sin6, sizeof(sin6));
1120 	sin6.sin6_family = PF_INET6;
1121 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1122 	sin6.sin6_addr = *dst;
1123 	if (in6_setscope(&sin6.sin6_addr, m->m_pkthdr.rcvif, NULL))
1124 		return;
1125 
1126 	rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6);
1127 
1128 	if (rt && (rt->rt_flags & RTF_HOST) &&
1129 	    !(rt->rt_rmx.rmx_locks & RTV_MTU) &&
1130 	    (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) {
1131 		if (mtu < IN6_LINKMTU(rt->rt_ifp)) {
1132 			ICMP6_STATINC(ICMP6_STAT_PMTUCHG);
1133 			rt->rt_rmx.rmx_mtu = mtu;
1134 		}
1135 	}
1136 	if (rt) { /* XXX: need braces to avoid conflict with else in RTFREE. */
1137 		RTFREE(rt);
1138 	}
1139 
1140 	/*
1141 	 * Notify protocols that the MTU for this destination
1142 	 * has changed.
1143 	 */
1144 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
1145 	     mc = LIST_NEXT(mc, mc_list))
1146 		(*mc->mc_func)(&sin6.sin6_addr);
1147 }
1148 
1149 /*
1150  * Process a Node Information Query packet, based on
1151  * draft-ietf-ipngwg-icmp-name-lookups-07.
1152  *
1153  * Spec incompatibilities:
1154  * - IPv6 Subject address handling
1155  * - IPv4 Subject address handling support missing
1156  * - Proxy reply (answer even if it's not for me)
1157  * - joins NI group address at in6_ifattach() time only, does not cope
1158  *   with hostname changes by sethostname(3)
1159  */
1160 #ifndef offsetof		/* XXX */
1161 #define	offsetof(type, member)	((size_t)(&((type *)0)->member))
1162 #endif
1163 static struct mbuf *
1164 ni6_input(struct mbuf *m, int off)
1165 {
1166 	struct icmp6_nodeinfo *ni6, *nni6;
1167 	struct mbuf *n = NULL;
1168 	u_int16_t qtype;
1169 	int subjlen;
1170 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1171 	struct ni_reply_fqdn *fqdn;
1172 	int addrs;		/* for NI_QTYPE_NODEADDR */
1173 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1174 	struct sockaddr_in6 sin6; /* ip6_dst */
1175 	struct in6_addr in6_subj; /* subject address */
1176 	struct ip6_hdr *ip6;
1177 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1178 	char *subj = NULL;
1179 
1180 	ip6 = mtod(m, struct ip6_hdr *);
1181 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1182 	if (ni6 == NULL) {
1183 		/* m is already reclaimed */
1184 		return NULL;
1185 	}
1186 
1187 	/*
1188 	 * Validate IPv6 destination address.
1189 	 *
1190 	 * The Responder must discard the Query without further processing
1191 	 * unless it is one of the Responder's unicast or anycast addresses, or
1192 	 * a link-local scope multicast address which the Responder has joined.
1193 	 * [icmp-name-lookups-07, Section 4.]
1194 	 */
1195 	sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0);
1196 	/* XXX scopeid */
1197 	if (ifa_ifwithaddr((struct sockaddr *)&sin6))
1198 		; /* unicast/anycast, fine */
1199 	else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1200 		; /* link-local multicast, fine */
1201 	else
1202 		goto bad;
1203 
1204 	/* validate query Subject field. */
1205 	qtype = ntohs(ni6->ni_qtype);
1206 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1207 	switch (qtype) {
1208 	case NI_QTYPE_NOOP:
1209 	case NI_QTYPE_SUPTYPES:
1210 		/* 07 draft */
1211 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1212 			break;
1213 		/* FALLTHROUGH */
1214 	case NI_QTYPE_FQDN:
1215 	case NI_QTYPE_NODEADDR:
1216 	case NI_QTYPE_IPV4ADDR:
1217 		switch (ni6->ni_code) {
1218 		case ICMP6_NI_SUBJ_IPV6:
1219 #if ICMP6_NI_SUBJ_IPV6 != 0
1220 		case 0:
1221 #endif
1222 			/*
1223 			 * backward compatibility - try to accept 03 draft
1224 			 * format, where no Subject is present.
1225 			 */
1226 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1227 			    subjlen == 0) {
1228 				oldfqdn++;
1229 				break;
1230 			}
1231 #if ICMP6_NI_SUBJ_IPV6 != 0
1232 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1233 				goto bad;
1234 #endif
1235 
1236 			if (subjlen != sizeof(sin6.sin6_addr))
1237 				goto bad;
1238 
1239 			/*
1240 			 * Validate Subject address.
1241 			 *
1242 			 * Not sure what exactly "address belongs to the node"
1243 			 * means in the spec, is it just unicast, or what?
1244 			 *
1245 			 * At this moment we consider Subject address as
1246 			 * "belong to the node" if the Subject address equals
1247 			 * to the IPv6 destination address; validation for
1248 			 * IPv6 destination address should have done enough
1249 			 * check for us.
1250 			 *
1251 			 * We do not do proxy at this moment.
1252 			 */
1253 			/* m_pulldown instead of copy? */
1254 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1255 			    subjlen, (void *)&in6_subj);
1256 			if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1257 				goto bad;
1258 
1259 			subj = (char *)&in6_subj;
1260 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1261 				break;
1262 
1263 			/*
1264 			 * XXX if we are to allow other cases, we should really
1265 			 * be careful about scope here.
1266 			 * basically, we should disallow queries toward IPv6
1267 			 * destination X with subject Y, if scope(X) > scope(Y).
1268 			 * if we allow scope(X) > scope(Y), it will result in
1269 			 * information leakage across scope boundary.
1270 			 */
1271 			goto bad;
1272 
1273 		case ICMP6_NI_SUBJ_FQDN:
1274 			/*
1275 			 * Validate Subject name with gethostname(3).
1276 			 *
1277 			 * The behavior may need some debate, since:
1278 			 * - we are not sure if the node has FQDN as
1279 			 *   hostname (returned by gethostname(3)).
1280 			 * - the code does wildcard match for truncated names.
1281 			 *   however, we are not sure if we want to perform
1282 			 *   wildcard match, if gethostname(3) side has
1283 			 *   truncated hostname.
1284 			 */
1285 			n = ni6_nametodns(hostname, hostnamelen, 0);
1286 			if (!n || n->m_next || n->m_len == 0)
1287 				goto bad;
1288 			IP6_EXTHDR_GET(subj, char *, m,
1289 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1290 			if (subj == NULL)
1291 				goto bad;
1292 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1293 					n->m_len)) {
1294 				goto bad;
1295 			}
1296 			m_freem(n);
1297 			n = NULL;
1298 			break;
1299 
1300 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1301 		default:
1302 			goto bad;
1303 		}
1304 		break;
1305 	}
1306 
1307 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1308 	switch (qtype) {
1309 	case NI_QTYPE_FQDN:
1310 		if ((icmp6_nodeinfo & 1) == 0)
1311 			goto bad;
1312 		break;
1313 	case NI_QTYPE_NODEADDR:
1314 	case NI_QTYPE_IPV4ADDR:
1315 		if ((icmp6_nodeinfo & 2) == 0)
1316 			goto bad;
1317 		break;
1318 	}
1319 
1320 	/* guess reply length */
1321 	switch (qtype) {
1322 	case NI_QTYPE_NOOP:
1323 		break;		/* no reply data */
1324 	case NI_QTYPE_SUPTYPES:
1325 		replylen += sizeof(u_int32_t);
1326 		break;
1327 	case NI_QTYPE_FQDN:
1328 		/* XXX will append an mbuf */
1329 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1330 		break;
1331 	case NI_QTYPE_NODEADDR:
1332 		addrs = ni6_addrs(ni6, m, &ifp, subj);
1333 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1334 					  sizeof(u_int32_t))) > MCLBYTES)
1335 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1336 		break;
1337 	case NI_QTYPE_IPV4ADDR:
1338 		/* unsupported - should respond with unknown Qtype? */
1339 		goto bad;
1340 	default:
1341 		/*
1342 		 * XXX: We must return a reply with the ICMP6 code
1343 		 * `unknown Qtype' in this case.  However we regard the case
1344 		 * as an FQDN query for backward compatibility.
1345 		 * Older versions set a random value to this field,
1346 		 * so it rarely varies in the defined qtypes.
1347 		 * But the mechanism is not reliable...
1348 		 * maybe we should obsolete older versions.
1349 		 */
1350 		qtype = NI_QTYPE_FQDN;
1351 		/* XXX will append an mbuf */
1352 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1353 		oldfqdn++;
1354 		break;
1355 	}
1356 
1357 	/* allocate an mbuf to reply. */
1358 	MGETHDR(n, M_DONTWAIT, m->m_type);
1359 	if (n == NULL) {
1360 		m_freem(m);
1361 		return (NULL);
1362 	}
1363 	M_MOVE_PKTHDR(n, m); /* just for rcvif */
1364 	if (replylen > MHLEN) {
1365 		if (replylen > MCLBYTES) {
1366 			/*
1367 			 * XXX: should we try to allocate more? But MCLBYTES
1368 			 * is probably much larger than IPV6_MMTU...
1369 			 */
1370 			goto bad;
1371 		}
1372 		MCLGET(n, M_DONTWAIT);
1373 		if ((n->m_flags & M_EXT) == 0) {
1374 			goto bad;
1375 		}
1376 	}
1377 	n->m_pkthdr.len = n->m_len = replylen;
1378 
1379 	/* copy mbuf header and IPv6 + Node Information base headers */
1380 	bcopy(mtod(m, void *), mtod(n, void *), sizeof(struct ip6_hdr));
1381 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1382 	bcopy((void *)ni6, (void *)nni6, sizeof(struct icmp6_nodeinfo));
1383 
1384 	/* qtype dependent procedure */
1385 	switch (qtype) {
1386 	case NI_QTYPE_NOOP:
1387 		nni6->ni_code = ICMP6_NI_SUCCESS;
1388 		nni6->ni_flags = 0;
1389 		break;
1390 	case NI_QTYPE_SUPTYPES:
1391 	{
1392 		u_int32_t v;
1393 		nni6->ni_code = ICMP6_NI_SUCCESS;
1394 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1395 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1396 		v = (u_int32_t)htonl(0x0000000f);
1397 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1398 		break;
1399 	}
1400 	case NI_QTYPE_FQDN:
1401 		nni6->ni_code = ICMP6_NI_SUCCESS;
1402 		fqdn = (struct ni_reply_fqdn *)(mtod(n, char *) +
1403 						sizeof(struct ip6_hdr) +
1404 						sizeof(struct icmp6_nodeinfo));
1405 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1406 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1407 		/*
1408 		 * XXX do we really have FQDN in variable "hostname"?
1409 		 */
1410 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1411 		if (n->m_next == NULL)
1412 			goto bad;
1413 		/* XXX we assume that n->m_next is not a chain */
1414 		if (n->m_next->m_next != NULL)
1415 			goto bad;
1416 		n->m_pkthdr.len += n->m_next->m_len;
1417 		break;
1418 	case NI_QTYPE_NODEADDR:
1419 	{
1420 		int lenlim, copied;
1421 
1422 		nni6->ni_code = ICMP6_NI_SUCCESS;
1423 		n->m_pkthdr.len = n->m_len =
1424 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1425 		lenlim = M_TRAILINGSPACE(n);
1426 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1427 		/* XXX: reset mbuf length */
1428 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1429 			sizeof(struct icmp6_nodeinfo) + copied;
1430 		break;
1431 	}
1432 	default:
1433 		break;		/* XXX impossible! */
1434 	}
1435 
1436 	nni6->ni_type = ICMP6_NI_REPLY;
1437 	m_freem(m);
1438 	return (n);
1439 
1440   bad:
1441 	m_freem(m);
1442 	if (n)
1443 		m_freem(n);
1444 	return (NULL);
1445 }
1446 #undef hostnamelen
1447 
1448 #define isupper(x) ('A' <= (x) && (x) <= 'Z')
1449 #define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z'))
1450 #define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9'))
1451 #define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x))
1452 
1453 /*
1454  * make a mbuf with DNS-encoded string.  no compression support.
1455  *
1456  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1457  * treated as truncated name (two \0 at the end).  this is a wild guess.
1458  *
1459  * old - return pascal string if non-zero
1460  */
1461 static struct mbuf *
1462 ni6_nametodns(const char *name, int namelen, int old)
1463 {
1464 	struct mbuf *m;
1465 	char *cp, *ep;
1466 	const char *p, *q;
1467 	int i, len, nterm;
1468 
1469 	if (old)
1470 		len = namelen + 1;
1471 	else
1472 		len = MCLBYTES;
1473 
1474 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1475 	MGET(m, M_DONTWAIT, MT_DATA);
1476 	if (m && len > MLEN) {
1477 		MCLGET(m, M_DONTWAIT);
1478 		if ((m->m_flags & M_EXT) == 0)
1479 			goto fail;
1480 	}
1481 	if (!m)
1482 		goto fail;
1483 	m->m_next = NULL;
1484 
1485 	if (old) {
1486 		m->m_len = len;
1487 		*mtod(m, char *) = namelen;
1488 		bcopy(name, mtod(m, char *) + 1, namelen);
1489 		return m;
1490 	} else {
1491 		m->m_len = 0;
1492 		cp = mtod(m, char *);
1493 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1494 
1495 		/* if not certain about my name, return empty buffer */
1496 		if (namelen == 0)
1497 			return m;
1498 
1499 		/*
1500 		 * guess if it looks like shortened hostname, or FQDN.
1501 		 * shortened hostname needs two trailing "\0".
1502 		 */
1503 		i = 0;
1504 		for (p = name; p < name + namelen; p++) {
1505 			if (*p && *p == '.')
1506 				i++;
1507 		}
1508 		if (i < 2)
1509 			nterm = 2;
1510 		else
1511 			nterm = 1;
1512 
1513 		p = name;
1514 		while (cp < ep && p < name + namelen) {
1515 			i = 0;
1516 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1517 				i++;
1518 			/* result does not fit into mbuf */
1519 			if (cp + i + 1 >= ep)
1520 				goto fail;
1521 			/*
1522 			 * DNS label length restriction, RFC1035 page 8.
1523 			 * "i == 0" case is included here to avoid returning
1524 			 * 0-length label on "foo..bar".
1525 			 */
1526 			if (i <= 0 || i >= 64)
1527 				goto fail;
1528 			*cp++ = i;
1529 			if (!isalpha(p[0]) || !isalnum(p[i - 1]))
1530 				goto fail;
1531 			while (i > 0) {
1532 				if (!isalnum(*p) && *p != '-')
1533 					goto fail;
1534 				if (isupper(*p)) {
1535 					*cp++ = tolower(*p);
1536 					p++;
1537 				} else
1538 					*cp++ = *p++;
1539 				i--;
1540 			}
1541 			p = q;
1542 			if (p < name + namelen && *p == '.')
1543 				p++;
1544 		}
1545 		/* termination */
1546 		if (cp + nterm >= ep)
1547 			goto fail;
1548 		while (nterm-- > 0)
1549 			*cp++ = '\0';
1550 		m->m_len = cp - mtod(m, char *);
1551 		return m;
1552 	}
1553 
1554 	panic("should not reach here");
1555 	/* NOTREACHED */
1556 
1557  fail:
1558 	if (m)
1559 		m_freem(m);
1560 	return NULL;
1561 }
1562 
1563 /*
1564  * check if two DNS-encoded string matches.  takes care of truncated
1565  * form (with \0\0 at the end).  no compression support.
1566  * XXX upper/lowercase match (see RFC2065)
1567  */
1568 static int
1569 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1570 {
1571 	const char *a0, *b0;
1572 	int l;
1573 
1574 	/* simplest case - need validation? */
1575 	if (alen == blen && bcmp(a, b, alen) == 0)
1576 		return 1;
1577 
1578 	a0 = a;
1579 	b0 = b;
1580 
1581 	/* termination is mandatory */
1582 	if (alen < 2 || blen < 2)
1583 		return 0;
1584 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1585 		return 0;
1586 	alen--;
1587 	blen--;
1588 
1589 	while (a - a0 < alen && b - b0 < blen) {
1590 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1591 			return 0;
1592 
1593 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1594 			return 0;
1595 		/* we don't support compression yet */
1596 		if (a[0] >= 64 || b[0] >= 64)
1597 			return 0;
1598 
1599 		/* truncated case */
1600 		if (a[0] == 0 && a - a0 == alen - 1)
1601 			return 1;
1602 		if (b[0] == 0 && b - b0 == blen - 1)
1603 			return 1;
1604 		if (a[0] == 0 || b[0] == 0)
1605 			return 0;
1606 
1607 		if (a[0] != b[0])
1608 			return 0;
1609 		l = a[0];
1610 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1611 			return 0;
1612 		if (bcmp(a + 1, b + 1, l) != 0)
1613 			return 0;
1614 
1615 		a += 1 + l;
1616 		b += 1 + l;
1617 	}
1618 
1619 	if (a - a0 == alen && b - b0 == blen)
1620 		return 1;
1621 	else
1622 		return 0;
1623 }
1624 
1625 /*
1626  * calculate the number of addresses to be returned in the node info reply.
1627  */
1628 static int
1629 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m,
1630     struct ifnet **ifpp, char *subj)
1631 {
1632 	struct ifnet *ifp;
1633 	struct in6_ifaddr *ifa6;
1634 	struct ifaddr *ifa;
1635 	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1636 	int addrs = 0, addrsofif, iffound = 0;
1637 	int niflags = ni6->ni_flags;
1638 
1639 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1640 		switch (ni6->ni_code) {
1641 		case ICMP6_NI_SUBJ_IPV6:
1642 			if (subj == NULL) /* must be impossible... */
1643 				return (0);
1644 			subj_ip6 = (struct sockaddr_in6 *)subj;
1645 			break;
1646 		default:
1647 			/*
1648 			 * XXX: we only support IPv6 subject address for
1649 			 * this Qtype.
1650 			 */
1651 			return (0);
1652 		}
1653 	}
1654 
1655 	IFNET_FOREACH(ifp) {
1656 		addrsofif = 0;
1657 		IFADDR_FOREACH(ifa, ifp) {
1658 			if (ifa->ifa_addr->sa_family != AF_INET6)
1659 				continue;
1660 			ifa6 = (struct in6_ifaddr *)ifa;
1661 
1662 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1663 			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1664 					       &ifa6->ia_addr.sin6_addr))
1665 				iffound = 1;
1666 
1667 			/*
1668 			 * IPv4-mapped addresses can only be returned by a
1669 			 * Node Information proxy, since they represent
1670 			 * addresses of IPv4-only nodes, which perforce do
1671 			 * not implement this protocol.
1672 			 * [icmp-name-lookups-07, Section 5.4]
1673 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1674 			 * this function at this moment.
1675 			 */
1676 
1677 			/* What do we have to do about ::1? */
1678 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1679 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1680 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1681 					continue;
1682 				break;
1683 			case IPV6_ADDR_SCOPE_SITELOCAL:
1684 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1685 					continue;
1686 				break;
1687 			case IPV6_ADDR_SCOPE_GLOBAL:
1688 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1689 					continue;
1690 				break;
1691 			default:
1692 				continue;
1693 			}
1694 
1695 			/*
1696 			 * check if anycast is okay.
1697 			 * XXX: just experimental.  not in the spec.
1698 			 */
1699 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1700 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1701 				continue; /* we need only unicast addresses */
1702 
1703 			addrsofif++; /* count the address */
1704 		}
1705 		if (iffound) {
1706 			*ifpp = ifp;
1707 			return (addrsofif);
1708 		}
1709 
1710 		addrs += addrsofif;
1711 	}
1712 
1713 	return (addrs);
1714 }
1715 
1716 static int
1717 ni6_store_addrs(struct icmp6_nodeinfo *ni6,
1718 	struct icmp6_nodeinfo *nni6, struct ifnet *ifp0,
1719 	int resid)
1720 {
1721 	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1722 	struct in6_ifaddr *ifa6;
1723 	struct ifaddr *ifa;
1724 	struct ifnet *ifp_dep = NULL;
1725 	int copied = 0, allow_deprecated = 0;
1726 	u_char *cp = (u_char *)(nni6 + 1);
1727 	int niflags = ni6->ni_flags;
1728 	u_int32_t ltime;
1729 
1730 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1731 		return (0);	/* needless to copy */
1732 
1733   again:
1734 
1735 	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list))
1736 	{
1737 		IFADDR_FOREACH(ifa, ifp) {
1738 			if (ifa->ifa_addr->sa_family != AF_INET6)
1739 				continue;
1740 			ifa6 = (struct in6_ifaddr *)ifa;
1741 
1742 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1743 			    allow_deprecated == 0) {
1744 				/*
1745 				 * prefererred address should be put before
1746 				 * deprecated addresses.
1747 				 */
1748 
1749 				/* record the interface for later search */
1750 				if (ifp_dep == NULL)
1751 					ifp_dep = ifp;
1752 
1753 				continue;
1754 			}
1755 			else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1756 				 allow_deprecated != 0)
1757 				continue; /* we now collect deprecated addrs */
1758 
1759 			/* What do we have to do about ::1? */
1760 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1761 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1762 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1763 					continue;
1764 				break;
1765 			case IPV6_ADDR_SCOPE_SITELOCAL:
1766 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1767 					continue;
1768 				break;
1769 			case IPV6_ADDR_SCOPE_GLOBAL:
1770 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1771 					continue;
1772 				break;
1773 			default:
1774 				continue;
1775 			}
1776 
1777 			/*
1778 			 * check if anycast is okay.
1779 			 * XXX: just experimental.  not in the spec.
1780 			 */
1781 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1782 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1783 				continue;
1784 
1785 			/* now we can copy the address */
1786 			if (resid < sizeof(struct in6_addr) +
1787 			    sizeof(u_int32_t)) {
1788 				/*
1789 				 * We give up much more copy.
1790 				 * Set the truncate flag and return.
1791 				 */
1792 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1793 				return (copied);
1794 			}
1795 
1796 			/*
1797 			 * Set the TTL of the address.
1798 			 * The TTL value should be one of the following
1799 			 * according to the specification:
1800 			 *
1801 			 * 1. The remaining lifetime of a DHCP lease on the
1802 			 *    address, or
1803 			 * 2. The remaining Valid Lifetime of a prefix from
1804 			 *    which the address was derived through Stateless
1805 			 *    Autoconfiguration.
1806 			 *
1807 			 * Note that we currently do not support stateful
1808 			 * address configuration by DHCPv6, so the former
1809 			 * case can't happen.
1810 			 *
1811 			 * TTL must be 2^31 > TTL >= 0.
1812 			 */
1813 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1814 				ltime = ND6_INFINITE_LIFETIME;
1815 			else {
1816 				if (ifa6->ia6_lifetime.ia6t_expire >
1817 				    time_second)
1818 					ltime = ifa6->ia6_lifetime.ia6t_expire -
1819 					    time_second;
1820 				else
1821 					ltime = 0;
1822 			}
1823 			if (ltime > 0x7fffffff)
1824 				ltime = 0x7fffffff;
1825 			ltime = htonl(ltime);
1826 
1827 			bcopy(&ltime, cp, sizeof(u_int32_t));
1828 			cp += sizeof(u_int32_t);
1829 
1830 			/* copy the address itself */
1831 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1832 			      sizeof(struct in6_addr));
1833 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1834 			cp += sizeof(struct in6_addr);
1835 
1836 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1837 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1838 		}
1839 		if (ifp0)	/* we need search only on the specified IF */
1840 			break;
1841 	}
1842 
1843 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1844 		ifp = ifp_dep;
1845 		allow_deprecated = 1;
1846 
1847 		goto again;
1848 	}
1849 
1850 	return (copied);
1851 }
1852 
1853 /*
1854  * XXX almost dup'ed code with rip6_input.
1855  */
1856 static int
1857 icmp6_rip6_input(struct mbuf **mp, int off)
1858 {
1859 	struct mbuf *m = *mp;
1860 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1861 	struct inpcb_hdr *inph;
1862 	struct in6pcb *in6p;
1863 	struct in6pcb *last = NULL;
1864 	struct sockaddr_in6 rip6src;
1865 	struct icmp6_hdr *icmp6;
1866 	struct mbuf *opts = NULL;
1867 
1868 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1869 	if (icmp6 == NULL) {
1870 		/* m is already reclaimed */
1871 		return IPPROTO_DONE;
1872 	}
1873 
1874 	/*
1875 	 * XXX: the address may have embedded scope zone ID, which should be
1876 	 * hidden from applications.
1877 	 */
1878 	sockaddr_in6_init(&rip6src, &ip6->ip6_src, 0, 0, 0);
1879 	if (sa6_recoverscope(&rip6src)) {
1880 		m_freem(m);
1881 		return (IPPROTO_DONE);
1882 	}
1883 
1884 	CIRCLEQ_FOREACH(inph, &raw6cbtable.inpt_queue, inph_queue) {
1885 		in6p = (struct in6pcb *)inph;
1886 		if (in6p->in6p_af != AF_INET6)
1887 			continue;
1888 		if (in6p->in6p_ip6.ip6_nxt != IPPROTO_ICMPV6)
1889 			continue;
1890 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1891 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1892 			continue;
1893 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1894 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1895 			continue;
1896 		if (in6p->in6p_icmp6filt
1897 		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1898 				 in6p->in6p_icmp6filt))
1899 			continue;
1900 		if (last) {
1901 			struct	mbuf *n;
1902 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1903 				if (last->in6p_flags & IN6P_CONTROLOPTS)
1904 					ip6_savecontrol(last, &opts, ip6, n);
1905 				/* strip intermediate headers */
1906 				m_adj(n, off);
1907 				if (sbappendaddr(&last->in6p_socket->so_rcv,
1908 						 (struct sockaddr *)&rip6src,
1909 						 n, opts) == 0) {
1910 					/* should notify about lost packet */
1911 					m_freem(n);
1912 					if (opts)
1913 						m_freem(opts);
1914 				} else
1915 					sorwakeup(last->in6p_socket);
1916 				opts = NULL;
1917 			}
1918 		}
1919 		last = in6p;
1920 	}
1921 	if (last) {
1922 		if (last->in6p_flags & IN6P_CONTROLOPTS)
1923 			ip6_savecontrol(last, &opts, ip6, m);
1924 		/* strip intermediate headers */
1925 		m_adj(m, off);
1926 		if (sbappendaddr(&last->in6p_socket->so_rcv,
1927 				(struct sockaddr *)&rip6src, m, opts) == 0) {
1928 			m_freem(m);
1929 			if (opts)
1930 				m_freem(opts);
1931 		} else
1932 			sorwakeup(last->in6p_socket);
1933 	} else {
1934 		m_freem(m);
1935 		IP6_STATDEC(IP6_STAT_DELIVERED);
1936 	}
1937 	return IPPROTO_DONE;
1938 }
1939 
1940 /*
1941  * Reflect the ip6 packet back to the source.
1942  * OFF points to the icmp6 header, counted from the top of the mbuf.
1943  *
1944  * Note: RFC 1885 required that an echo reply should be truncated if it
1945  * did not fit in with (return) path MTU, and KAME code supported the
1946  * behavior.  However, as a clarification after the RFC, this limitation
1947  * was removed in a revised version of the spec, RFC 2463.  We had kept the
1948  * old behavior, with a (non-default) ifdef block, while the new version of
1949  * the spec was an internet-draft status, and even after the new RFC was
1950  * published.  But it would rather make sense to clean the obsoleted part
1951  * up, and to make the code simpler at this stage.
1952  */
1953 void
1954 icmp6_reflect(struct mbuf *m, size_t off)
1955 {
1956 	struct ip6_hdr *ip6;
1957 	struct icmp6_hdr *icmp6;
1958 	const struct in6_ifaddr *ia;
1959 	const struct ip6aux *ip6a;
1960 	int plen;
1961 	int type, code;
1962 	struct ifnet *outif = NULL;
1963 	struct in6_addr origdst;
1964 	const struct in6_addr *src = NULL;
1965 
1966 	/* too short to reflect */
1967 	if (off < sizeof(struct ip6_hdr)) {
1968 		nd6log((LOG_DEBUG,
1969 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
1970 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
1971 		    __FILE__, __LINE__));
1972 		goto bad;
1973 	}
1974 
1975 	/*
1976 	 * If there are extra headers between IPv6 and ICMPv6, strip
1977 	 * off that header first.
1978 	 */
1979 #ifdef DIAGNOSTIC
1980 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
1981 		panic("assumption failed in icmp6_reflect");
1982 #endif
1983 	if (off > sizeof(struct ip6_hdr)) {
1984 		size_t l;
1985 		struct ip6_hdr nip6;
1986 
1987 		l = off - sizeof(struct ip6_hdr);
1988 		m_copydata(m, 0, sizeof(nip6), (void *)&nip6);
1989 		m_adj(m, l);
1990 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1991 		if (m->m_len < l) {
1992 			if ((m = m_pullup(m, l)) == NULL)
1993 				return;
1994 		}
1995 		bcopy((void *)&nip6, mtod(m, void *), sizeof(nip6));
1996 	} else /* off == sizeof(struct ip6_hdr) */ {
1997 		size_t l;
1998 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1999 		if (m->m_len < l) {
2000 			if ((m = m_pullup(m, l)) == NULL)
2001 				return;
2002 		}
2003 	}
2004 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2005 	ip6 = mtod(m, struct ip6_hdr *);
2006 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2007 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2008 	type = icmp6->icmp6_type; /* keep type for statistics */
2009 	code = icmp6->icmp6_code; /* ditto. */
2010 
2011 	origdst = ip6->ip6_dst;
2012 	/*
2013 	 * ip6_input() drops a packet if its src is multicast.
2014 	 * So, the src is never multicast.
2015 	 */
2016 	ip6->ip6_dst = ip6->ip6_src;
2017 
2018 	/*
2019 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2020 	 * use dst as the src for the reply.
2021 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2022 	 * (for example) when we encounter an error while forwarding procedure
2023 	 * destined to a duplicated address of ours.
2024 	 * Note that ip6_getdstifaddr() may fail if we are in an error handling
2025 	 * procedure of an outgoing packet of our own, in which case we need
2026 	 * to search in the ifaddr list.
2027 	 */
2028 	if (IN6_IS_ADDR_MULTICAST(&origdst))
2029 		;
2030 	else if ((ip6a = ip6_getdstifaddr(m)) != NULL) {
2031 		if ((ip6a->ip6a_flags &
2032 		     (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0)
2033 			src = &ip6a->ip6a_src;
2034 	} else {
2035 		union {
2036 			struct sockaddr_in6 sin6;
2037 			struct sockaddr sa;
2038 		} u;
2039 
2040 		sockaddr_in6_init(&u.sin6, &origdst, 0, 0, 0);
2041 
2042 		ia = (struct in6_ifaddr *)ifa_ifwithaddr(&u.sa);
2043 
2044 		if (ia == NULL)
2045 			;
2046 		else if ((ia->ia6_flags &
2047 			 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0)
2048 			src = &ia->ia_addr.sin6_addr;
2049 	}
2050 
2051 	if (src == NULL) {
2052 		int e;
2053 		struct sockaddr_in6 sin6;
2054 		struct route ro;
2055 
2056 		/*
2057 		 * This case matches to multicasts, our anycast, or unicasts
2058 		 * that we do not own.  Select a source address based on the
2059 		 * source address of the erroneous packet.
2060 		 */
2061 		/* zone ID should be embedded */
2062 		sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0);
2063 
2064 		memset(&ro, 0, sizeof(ro));
2065 		src = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &e);
2066 		rtcache_free(&ro);
2067 		if (src == NULL) {
2068 			nd6log((LOG_DEBUG,
2069 			    "icmp6_reflect: source can't be determined: "
2070 			    "dst=%s, error=%d\n",
2071 			    ip6_sprintf(&sin6.sin6_addr), e));
2072 			goto bad;
2073 		}
2074 	}
2075 
2076 	ip6->ip6_src = *src;
2077 	ip6->ip6_flow = 0;
2078 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2079 	ip6->ip6_vfc |= IPV6_VERSION;
2080 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2081 	if (m->m_pkthdr.rcvif) {
2082 		/* XXX: This may not be the outgoing interface */
2083 		ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2084 	} else
2085 		ip6->ip6_hlim = ip6_defhlim;
2086 
2087 	m->m_pkthdr.csum_flags = 0;
2088 	icmp6->icmp6_cksum = 0;
2089 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2090 					sizeof(struct ip6_hdr), plen);
2091 
2092 	/*
2093 	 * XXX option handling
2094 	 */
2095 
2096 	m->m_flags &= ~(M_BCAST|M_MCAST);
2097 
2098 	/*
2099 	 * To avoid a "too big" situation at an intermediate router
2100 	 * and the path MTU discovery process, specify the IPV6_MINMTU flag.
2101 	 * Note that only echo and node information replies are affected,
2102 	 * since the length of ICMP6 errors is limited to the minimum MTU.
2103 	 */
2104 	if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL, &outif) != 0 &&
2105 	    outif)
2106 		icmp6_ifstat_inc(outif, ifs6_out_error);
2107 
2108 	if (outif)
2109 		icmp6_ifoutstat_inc(outif, type, code);
2110 
2111 	return;
2112 
2113  bad:
2114 	m_freem(m);
2115 	return;
2116 }
2117 
2118 static const char *
2119 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2120 	struct in6_addr *tgt6)
2121 {
2122 	static char buf[1024];
2123 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2124 		ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2125 	return buf;
2126 }
2127 
2128 void
2129 icmp6_redirect_input(struct mbuf *m, int off)
2130 {
2131 	struct ifnet *ifp = m->m_pkthdr.rcvif;
2132 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2133 	struct nd_redirect *nd_rd;
2134 	int icmp6len = ntohs(ip6->ip6_plen);
2135 	char *lladdr = NULL;
2136 	int lladdrlen = 0;
2137 	struct rtentry *rt = NULL;
2138 	int is_router;
2139 	int is_onlink;
2140 	struct in6_addr src6 = ip6->ip6_src;
2141 	struct in6_addr redtgt6;
2142 	struct in6_addr reddst6;
2143 	union nd_opts ndopts;
2144 
2145 	if (!ifp)
2146 		return;
2147 
2148 	/* XXX if we are router, we don't update route by icmp6 redirect */
2149 	if (ip6_forwarding)
2150 		goto freeit;
2151 	if (!icmp6_rediraccept)
2152 		goto freeit;
2153 
2154 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2155 	if (nd_rd == NULL) {
2156 		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
2157 		return;
2158 	}
2159 	redtgt6 = nd_rd->nd_rd_target;
2160 	reddst6 = nd_rd->nd_rd_dst;
2161 
2162 	if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2163 	    in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2164 		goto freeit;
2165 	}
2166 
2167 	/* validation */
2168 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2169 		nd6log((LOG_ERR,
2170 			"ICMP6 redirect sent from %s rejected; "
2171 			"must be from linklocal\n", ip6_sprintf(&src6)));
2172 		goto bad;
2173 	}
2174 	if (ip6->ip6_hlim != 255) {
2175 		nd6log((LOG_ERR,
2176 			"ICMP6 redirect sent from %s rejected; "
2177 			"hlim=%d (must be 255)\n",
2178 			ip6_sprintf(&src6), ip6->ip6_hlim));
2179 		goto bad;
2180 	}
2181     {
2182 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2183 	struct sockaddr_in6 sin6;
2184 	struct in6_addr *gw6;
2185 
2186 	sockaddr_in6_init(&sin6, &reddst6, 0, 0, 0);
2187 	rt = rtalloc1((struct sockaddr *)&sin6, 0);
2188 	if (rt) {
2189 		if (rt->rt_gateway == NULL ||
2190 		    rt->rt_gateway->sa_family != AF_INET6) {
2191 			nd6log((LOG_ERR,
2192 			    "ICMP6 redirect rejected; no route "
2193 			    "with inet6 gateway found for redirect dst: %s\n",
2194 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2195 			RTFREE(rt);
2196 			goto bad;
2197 		}
2198 
2199 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2200 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2201 			nd6log((LOG_ERR,
2202 				"ICMP6 redirect rejected; "
2203 				"not equal to gw-for-src=%s (must be same): "
2204 				"%s\n",
2205 				ip6_sprintf(gw6),
2206 				icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2207 			RTFREE(rt);
2208 			goto bad;
2209 		}
2210 	} else {
2211 		nd6log((LOG_ERR,
2212 			"ICMP6 redirect rejected; "
2213 			"no route found for redirect dst: %s\n",
2214 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2215 		goto bad;
2216 	}
2217 	RTFREE(rt);
2218 	rt = NULL;
2219     }
2220 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2221 		nd6log((LOG_ERR,
2222 			"ICMP6 redirect rejected; "
2223 			"redirect dst must be unicast: %s\n",
2224 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2225 		goto bad;
2226 	}
2227 
2228 	is_router = is_onlink = 0;
2229 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2230 		is_router = 1;	/* router case */
2231 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2232 		is_onlink = 1;	/* on-link destination case */
2233 	if (!is_router && !is_onlink) {
2234 		nd6log((LOG_ERR,
2235 			"ICMP6 redirect rejected; "
2236 			"neither router case nor onlink case: %s\n",
2237 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2238 		goto bad;
2239 	}
2240 	/* validation passed */
2241 
2242 	icmp6len -= sizeof(*nd_rd);
2243 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2244 	if (nd6_options(&ndopts) < 0) {
2245 		nd6log((LOG_INFO, "icmp6_redirect_input: "
2246 			"invalid ND option, rejected: %s\n",
2247 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2248 		/* nd6_options have incremented stats */
2249 		goto freeit;
2250 	}
2251 
2252 	if (ndopts.nd_opts_tgt_lladdr) {
2253 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2254 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2255 	}
2256 
2257 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2258 		nd6log((LOG_INFO,
2259 			"icmp6_redirect_input: lladdrlen mismatch for %s "
2260 			"(if %d, icmp6 packet %d): %s\n",
2261 			ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2262 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2263 		goto bad;
2264 	}
2265 
2266 	/* RFC 2461 8.3 */
2267 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2268 			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2269 
2270 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2271 		/* perform rtredirect */
2272 		struct sockaddr_in6 sdst;
2273 		struct sockaddr_in6 sgw;
2274 		struct sockaddr_in6 ssrc;
2275 		unsigned long rtcount;
2276 		struct rtentry *newrt = NULL;
2277 
2278 		/*
2279 		 * do not install redirect route, if the number of entries
2280 		 * is too much (> hiwat).  note that, the node (= host) will
2281 		 * work just fine even if we do not install redirect route
2282 		 * (there will be additional hops, though).
2283 		 */
2284 		rtcount = rt_timer_count(icmp6_redirect_timeout_q);
2285 		if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat)
2286 			return;
2287 		else if (0 <= icmp6_redirect_lowat &&
2288 		    rtcount > icmp6_redirect_lowat) {
2289 			/*
2290 			 * XXX nuke a victim, install the new one.
2291 			 */
2292 		}
2293 
2294 		bzero(&sdst, sizeof(sdst));
2295 		bzero(&sgw, sizeof(sgw));
2296 		bzero(&ssrc, sizeof(ssrc));
2297 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2298 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2299 			sizeof(struct sockaddr_in6);
2300 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2301 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2302 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2303 		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2304 			   (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2305 			   (struct sockaddr *)&ssrc,
2306 			   &newrt);
2307 
2308 		if (newrt) {
2309 			(void)rt_timer_add(newrt, icmp6_redirect_timeout,
2310 			    icmp6_redirect_timeout_q);
2311 			rtfree(newrt);
2312 		}
2313 	}
2314 	/* finally update cached route in each socket via pfctlinput */
2315 	{
2316 		struct sockaddr_in6 sdst;
2317 
2318 		sockaddr_in6_init(&sdst, &reddst6, 0, 0, 0);
2319 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2320 #if defined(IPSEC) || defined(FAST_IPSEC)
2321 		key_sa_routechange((struct sockaddr *)&sdst);
2322 #endif
2323 	}
2324 
2325  freeit:
2326 	m_freem(m);
2327 	return;
2328 
2329  bad:
2330 	ICMP6_STATINC(ICMP6_STAT_BADREDIRECT);
2331 	m_freem(m);
2332 }
2333 
2334 void
2335 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2336 {
2337 	struct ifnet *ifp;	/* my outgoing interface */
2338 	struct in6_addr *ifp_ll6;
2339 	struct in6_addr *nexthop;
2340 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2341 	struct mbuf *m = NULL;	/* newly allocated one */
2342 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2343 	struct nd_redirect *nd_rd;
2344 	size_t maxlen;
2345 	u_char *p;
2346 	struct sockaddr_in6 src_sa;
2347 
2348 	icmp6_errcount(ICMP6_STAT_OUTERRHIST, ND_REDIRECT, 0);
2349 
2350 	/* if we are not router, we don't send icmp6 redirect */
2351 	if (!ip6_forwarding)
2352 		goto fail;
2353 
2354 	/* sanity check */
2355 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2356 		goto fail;
2357 
2358 	/*
2359 	 * Address check:
2360 	 *  the source address must identify a neighbor, and
2361 	 *  the destination address must not be a multicast address
2362 	 *  [RFC 2461, sec 8.2]
2363 	 */
2364 	sip6 = mtod(m0, struct ip6_hdr *);
2365 	sockaddr_in6_init(&src_sa, &sip6->ip6_src, 0, 0, 0);
2366 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2367 		goto fail;
2368 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2369 		goto fail;	/* what should we do here? */
2370 
2371 	/* rate limit */
2372 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2373 		goto fail;
2374 
2375 	/*
2376 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2377 	 * we almost always ask for an mbuf cluster for simplicity.
2378 	 * (MHLEN < IPV6_MMTU is almost always true)
2379 	 */
2380 #if IPV6_MMTU >= MCLBYTES
2381 # error assumption failed about IPV6_MMTU and MCLBYTES
2382 #endif
2383 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2384 	if (m && IPV6_MMTU >= MHLEN)
2385 		MCLGET(m, M_DONTWAIT);
2386 	if (!m)
2387 		goto fail;
2388 	m->m_pkthdr.rcvif = NULL;
2389 	m->m_len = 0;
2390 	maxlen = M_TRAILINGSPACE(m);
2391 	maxlen = min(IPV6_MMTU, maxlen);
2392 	/* just for safety */
2393 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2394 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2395 		goto fail;
2396 	}
2397 
2398 	{
2399 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2400 		struct in6_ifaddr *ia;
2401 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2402 						 IN6_IFF_NOTREADY|
2403 						 IN6_IFF_ANYCAST)) == NULL)
2404 			goto fail;
2405 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2406 	}
2407 
2408 	/* get ip6 linklocal address for the router. */
2409 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2410 		struct sockaddr_in6 *sin6;
2411 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2412 		nexthop = &sin6->sin6_addr;
2413 		if (!IN6_IS_ADDR_LINKLOCAL(nexthop))
2414 			nexthop = NULL;
2415 	} else
2416 		nexthop = NULL;
2417 
2418 	/* ip6 */
2419 	ip6 = mtod(m, struct ip6_hdr *);
2420 	ip6->ip6_flow = 0;
2421 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2422 	ip6->ip6_vfc |= IPV6_VERSION;
2423 	/* ip6->ip6_plen will be set later */
2424 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2425 	ip6->ip6_hlim = 255;
2426 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2427 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2428 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2429 
2430 	/* ND Redirect */
2431 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2432 	nd_rd->nd_rd_type = ND_REDIRECT;
2433 	nd_rd->nd_rd_code = 0;
2434 	nd_rd->nd_rd_reserved = 0;
2435 	if (rt->rt_flags & RTF_GATEWAY) {
2436 		/*
2437 		 * nd_rd->nd_rd_target must be a link-local address in
2438 		 * better router cases.
2439 		 */
2440 		if (!nexthop)
2441 			goto fail;
2442 		bcopy(nexthop, &nd_rd->nd_rd_target,
2443 		      sizeof(nd_rd->nd_rd_target));
2444 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2445 		      sizeof(nd_rd->nd_rd_dst));
2446 	} else {
2447 		/* make sure redtgt == reddst */
2448 		nexthop = &sip6->ip6_dst;
2449 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2450 		      sizeof(nd_rd->nd_rd_target));
2451 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2452 		      sizeof(nd_rd->nd_rd_dst));
2453 	}
2454 
2455 	p = (u_char *)(nd_rd + 1);
2456 
2457 	{
2458 		/* target lladdr option */
2459 		struct rtentry *rt_nexthop = NULL;
2460 		int len;
2461 		const struct sockaddr_dl *sdl;
2462 		struct nd_opt_hdr *nd_opt;
2463 		char *lladdr;
2464 
2465 		rt_nexthop = nd6_lookup(nexthop, 0, ifp);
2466 		if (!rt_nexthop)
2467 			goto nolladdropt;
2468 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2469 		len = (len + 7) & ~7;	/* round by 8 */
2470 		/* safety check */
2471 		if (len + (p - (u_char *)ip6) > maxlen)
2472 			goto nolladdropt;
2473 		if (!(rt_nexthop->rt_flags & RTF_GATEWAY) &&
2474 		    (rt_nexthop->rt_flags & RTF_LLINFO) &&
2475 		    (rt_nexthop->rt_gateway->sa_family == AF_LINK) &&
2476 		    (sdl = satocsdl(rt_nexthop->rt_gateway)) &&
2477 		    sdl->sdl_alen) {
2478 			nd_opt = (struct nd_opt_hdr *)p;
2479 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2480 			nd_opt->nd_opt_len = len >> 3;
2481 			lladdr = (char *)(nd_opt + 1);
2482 			memcpy(lladdr, CLLADDR(sdl), ifp->if_addrlen);
2483 			p += len;
2484 		}
2485 	}
2486   nolladdropt:;
2487 
2488 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2489 
2490 	/* just to be safe */
2491 	if (m0->m_flags & M_DECRYPTED)
2492 		goto noredhdropt;
2493 	if (p - (u_char *)ip6 > maxlen)
2494 		goto noredhdropt;
2495 
2496 	{
2497 		/* redirected header option */
2498 		int len;
2499 		struct nd_opt_rd_hdr *nd_opt_rh;
2500 
2501 		/*
2502 		 * compute the maximum size for icmp6 redirect header option.
2503 		 * XXX room for auth header?
2504 		 */
2505 		len = maxlen - (p - (u_char *)ip6);
2506 		len &= ~7;
2507 
2508 		/*
2509 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2510 		 * about padding/truncate rule for the original IP packet.
2511 		 * From the discussion on IPv6imp in Feb 1999,
2512 		 * the consensus was:
2513 		 * - "attach as much as possible" is the goal
2514 		 * - pad if not aligned (original size can be guessed by
2515 		 *   original ip6 header)
2516 		 * Following code adds the padding if it is simple enough,
2517 		 * and truncates if not.
2518 		 */
2519 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2520 			/* not enough room, truncate */
2521 			m_adj(m0, (len - sizeof(*nd_opt_rh)) -
2522 			    m0->m_pkthdr.len);
2523 		} else {
2524 			/*
2525 			 * enough room, truncate if not aligned.
2526 			 * we don't pad here for simplicity.
2527 			 */
2528 			size_t extra;
2529 
2530 			extra = m0->m_pkthdr.len % 8;
2531 			if (extra) {
2532 				/* truncate */
2533 				m_adj(m0, -extra);
2534 			}
2535 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2536 		}
2537 
2538 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2539 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2540 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2541 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2542 		p += sizeof(*nd_opt_rh);
2543 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2544 
2545 		/* connect m0 to m */
2546 		m->m_pkthdr.len += m0->m_pkthdr.len;
2547 		m_cat(m, m0);
2548 		m0 = NULL;
2549 	}
2550 noredhdropt:
2551 	if (m0) {
2552 		m_freem(m0);
2553 		m0 = NULL;
2554 	}
2555 
2556 	/* XXX: clear embedded link IDs in the inner header */
2557 	in6_clearscope(&sip6->ip6_src);
2558 	in6_clearscope(&sip6->ip6_dst);
2559 	in6_clearscope(&nd_rd->nd_rd_target);
2560 	in6_clearscope(&nd_rd->nd_rd_dst);
2561 
2562 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2563 
2564 	nd_rd->nd_rd_cksum = 0;
2565 	nd_rd->nd_rd_cksum
2566 		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2567 
2568 	/* send the packet to outside... */
2569 	if (ip6_output(m, NULL, NULL, 0,
2570 		(struct ip6_moptions *)NULL, (struct socket *)NULL, NULL) != 0)
2571 		icmp6_ifstat_inc(ifp, ifs6_out_error);
2572 
2573 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
2574 	icmp6_ifstat_inc(ifp, ifs6_out_redirect);
2575 	ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_REDIRECT);
2576 
2577 	return;
2578 
2579 fail:
2580 	if (m)
2581 		m_freem(m);
2582 	if (m0)
2583 		m_freem(m0);
2584 }
2585 
2586 /*
2587  * ICMPv6 socket option processing.
2588  */
2589 int
2590 icmp6_ctloutput(int op, struct socket *so, struct sockopt *sopt)
2591 {
2592 	int error = 0;
2593 	struct in6pcb *in6p = sotoin6pcb(so);
2594 
2595 	if (sopt->sopt_level != IPPROTO_ICMPV6)
2596 		return rip6_ctloutput(op, so, sopt);
2597 
2598 	switch (op) {
2599 	case PRCO_SETOPT:
2600 		switch (sopt->sopt_name) {
2601 		case ICMP6_FILTER:
2602 		    {
2603 			struct icmp6_filter fil;
2604 
2605 			error = sockopt_get(sopt, &fil, sizeof(fil));
2606 			if (error)
2607 				break;
2608 			memcpy(in6p->in6p_icmp6filt, &fil,
2609 			    sizeof(struct icmp6_filter));
2610 			error = 0;
2611 			break;
2612 		    }
2613 
2614 		default:
2615 			error = ENOPROTOOPT;
2616 			break;
2617 		}
2618 		break;
2619 
2620 	case PRCO_GETOPT:
2621 		switch (sopt->sopt_name) {
2622 		case ICMP6_FILTER:
2623 		    {
2624 			if (in6p->in6p_icmp6filt == NULL) {
2625 				error = EINVAL;
2626 				break;
2627 			}
2628 			error = sockopt_set(sopt, in6p->in6p_icmp6filt,
2629 			    sizeof(struct icmp6_filter));
2630 			break;
2631 		    }
2632 
2633 		default:
2634 			error = ENOPROTOOPT;
2635 			break;
2636 		}
2637 		break;
2638 	}
2639 
2640 	return (error);
2641 }
2642 
2643 /*
2644  * Perform rate limit check.
2645  * Returns 0 if it is okay to send the icmp6 packet.
2646  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2647  * limitation.
2648  *
2649  * XXX per-destination/type check necessary?
2650  */
2651 static int
2652 icmp6_ratelimit(
2653 	const struct in6_addr *dst,	/* not used at this moment */
2654 	const int type,		/* not used at this moment */
2655 	const int code)		/* not used at this moment */
2656 {
2657 	int ret;
2658 
2659 	ret = 0;	/* okay to send */
2660 
2661 	/* PPS limit */
2662 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2663 	    icmp6errppslim)) {
2664 		/* The packet is subject to rate limit */
2665 		ret++;
2666 	}
2667 
2668 	return ret;
2669 }
2670 
2671 static struct rtentry *
2672 icmp6_mtudisc_clone(struct sockaddr *dst)
2673 {
2674 	struct rtentry *rt;
2675 	int    error;
2676 
2677 	rt = rtalloc1(dst, 1);
2678 	if (rt == 0)
2679 		return NULL;
2680 
2681 	/* If we didn't get a host route, allocate one */
2682 	if ((rt->rt_flags & RTF_HOST) == 0) {
2683 		struct rtentry *nrt;
2684 
2685 		error = rtrequest((int) RTM_ADD, dst,
2686 		    (struct sockaddr *) rt->rt_gateway,
2687 		    (struct sockaddr *) 0,
2688 		    RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt);
2689 		if (error) {
2690 			rtfree(rt);
2691 			return NULL;
2692 		}
2693 		nrt->rt_rmx = rt->rt_rmx;
2694 		rtfree(rt);
2695 		rt = nrt;
2696 	}
2697 	error = rt_timer_add(rt, icmp6_mtudisc_timeout,
2698 			icmp6_mtudisc_timeout_q);
2699 	if (error) {
2700 		rtfree(rt);
2701 		return NULL;
2702 	}
2703 
2704 	return rt;	/* caller need to call rtfree() */
2705 }
2706 
2707 static void
2708 icmp6_mtudisc_timeout(struct rtentry *rt, struct rttimer *r)
2709 {
2710 	if (rt == NULL)
2711 		panic("icmp6_mtudisc_timeout: bad route to timeout");
2712 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
2713 	    (RTF_DYNAMIC | RTF_HOST)) {
2714 		rtrequest((int) RTM_DELETE, rt_getkey(rt),
2715 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
2716 	} else {
2717 		if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
2718 			rt->rt_rmx.rmx_mtu = 0;
2719 	}
2720 }
2721 
2722 static void
2723 icmp6_redirect_timeout(struct rtentry *rt, struct rttimer *r)
2724 {
2725 	if (rt == NULL)
2726 		panic("icmp6_redirect_timeout: bad route to timeout");
2727 	if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) ==
2728 	    (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) {
2729 		rtrequest((int) RTM_DELETE, rt_getkey(rt),
2730 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
2731 	}
2732 }
2733 
2734 /*
2735  * sysctl helper routine for the net.inet6.icmp6.nd6 nodes.  silly?
2736  */
2737 static int
2738 sysctl_net_inet6_icmp6_nd6(SYSCTLFN_ARGS)
2739 {
2740 	(void)&name;
2741 	(void)&l;
2742 	(void)&oname;
2743 
2744 	if (namelen != 0)
2745 		return (EINVAL);
2746 
2747 	return (nd6_sysctl(rnode->sysctl_num, oldp, oldlenp,
2748 	    /*XXXUNCONST*/
2749 	    __UNCONST(newp), newlen));
2750 }
2751 
2752 static int
2753 sysctl_net_inet6_icmp6_stats(SYSCTLFN_ARGS)
2754 {
2755 
2756 	return (NETSTAT_SYSCTL(icmp6stat_percpu, ICMP6_NSTATS));
2757 }
2758 
2759 SYSCTL_SETUP(sysctl_net_inet6_icmp6_setup,
2760 	     "sysctl net.inet6.icmp6 subtree setup")
2761 {
2762 	extern int nd6_maxqueuelen; /* defined in nd6.c */
2763 
2764 	sysctl_createv(clog, 0, NULL, NULL,
2765 		       CTLFLAG_PERMANENT,
2766 		       CTLTYPE_NODE, "net", NULL,
2767 		       NULL, 0, NULL, 0,
2768 		       CTL_NET, CTL_EOL);
2769 	sysctl_createv(clog, 0, NULL, NULL,
2770 		       CTLFLAG_PERMANENT,
2771 		       CTLTYPE_NODE, "inet6", NULL,
2772 		       NULL, 0, NULL, 0,
2773 		       CTL_NET, PF_INET6, CTL_EOL);
2774 	sysctl_createv(clog, 0, NULL, NULL,
2775 		       CTLFLAG_PERMANENT,
2776 		       CTLTYPE_NODE, "icmp6",
2777 		       SYSCTL_DESCR("ICMPv6 related settings"),
2778 		       NULL, 0, NULL, 0,
2779 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6, CTL_EOL);
2780 
2781 	sysctl_createv(clog, 0, NULL, NULL,
2782 		       CTLFLAG_PERMANENT,
2783 		       CTLTYPE_STRUCT, "stats",
2784 		       SYSCTL_DESCR("ICMPv6 transmission statistics"),
2785 		       sysctl_net_inet6_icmp6_stats, 0, NULL, 0,
2786 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2787 		       ICMPV6CTL_STATS, CTL_EOL);
2788 	sysctl_createv(clog, 0, NULL, NULL,
2789 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2790 		       CTLTYPE_INT, "rediraccept",
2791 		       SYSCTL_DESCR("Accept and process redirect messages"),
2792 		       NULL, 0, &icmp6_rediraccept, 0,
2793 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2794 		       ICMPV6CTL_REDIRACCEPT, CTL_EOL);
2795 	sysctl_createv(clog, 0, NULL, NULL,
2796 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2797 		       CTLTYPE_INT, "redirtimeout",
2798 		       SYSCTL_DESCR("Redirect generated route lifetime"),
2799 		       NULL, 0, &icmp6_redirtimeout, 0,
2800 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2801 		       ICMPV6CTL_REDIRTIMEOUT, CTL_EOL);
2802 #if 0 /* obsoleted */
2803 	sysctl_createv(clog, 0, NULL, NULL,
2804 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2805 		       CTLTYPE_INT, "errratelimit", NULL,
2806 		       NULL, 0, &icmp6_errratelimit, 0,
2807 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2808 		       ICMPV6CTL_ERRRATELIMIT, CTL_EOL);
2809 #endif
2810 	sysctl_createv(clog, 0, NULL, NULL,
2811 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2812 		       CTLTYPE_INT, "nd6_prune",
2813 		       SYSCTL_DESCR("Neighbor discovery prune interval"),
2814 		       NULL, 0, &nd6_prune, 0,
2815 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2816 		       ICMPV6CTL_ND6_PRUNE, CTL_EOL);
2817 	sysctl_createv(clog, 0, NULL, NULL,
2818 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2819 		       CTLTYPE_INT, "nd6_delay",
2820 		       SYSCTL_DESCR("First probe delay time"),
2821 		       NULL, 0, &nd6_delay, 0,
2822 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2823 		       ICMPV6CTL_ND6_DELAY, CTL_EOL);
2824 	sysctl_createv(clog, 0, NULL, NULL,
2825 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2826 		       CTLTYPE_INT, "nd6_umaxtries",
2827 		       SYSCTL_DESCR("Number of unicast discovery attempts"),
2828 		       NULL, 0, &nd6_umaxtries, 0,
2829 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2830 		       ICMPV6CTL_ND6_UMAXTRIES, CTL_EOL);
2831 	sysctl_createv(clog, 0, NULL, NULL,
2832 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2833 		       CTLTYPE_INT, "nd6_mmaxtries",
2834 		       SYSCTL_DESCR("Number of multicast discovery attempts"),
2835 		       NULL, 0, &nd6_mmaxtries, 0,
2836 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2837 		       ICMPV6CTL_ND6_MMAXTRIES, CTL_EOL);
2838 	sysctl_createv(clog, 0, NULL, NULL,
2839 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2840 		       CTLTYPE_INT, "nd6_useloopback",
2841 		       SYSCTL_DESCR("Use loopback interface for local traffic"),
2842 		       NULL, 0, &nd6_useloopback, 0,
2843 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2844 		       ICMPV6CTL_ND6_USELOOPBACK, CTL_EOL);
2845 #if 0 /* obsoleted */
2846 	sysctl_createv(clog, 0, NULL, NULL,
2847 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2848 		       CTLTYPE_INT, "nd6_proxyall", NULL,
2849 		       NULL, 0, &nd6_proxyall, 0,
2850 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2851 		       ICMPV6CTL_ND6_PROXYALL, CTL_EOL);
2852 #endif
2853 	sysctl_createv(clog, 0, NULL, NULL,
2854 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2855 		       CTLTYPE_INT, "nodeinfo",
2856 		       SYSCTL_DESCR("Respond to node information requests"),
2857 		       NULL, 0, &icmp6_nodeinfo, 0,
2858 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2859 		       ICMPV6CTL_NODEINFO, CTL_EOL);
2860 	sysctl_createv(clog, 0, NULL, NULL,
2861 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2862 		       CTLTYPE_INT, "errppslimit",
2863 		       SYSCTL_DESCR("Maximum ICMP errors sent per second"),
2864 		       NULL, 0, &icmp6errppslim, 0,
2865 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2866 		       ICMPV6CTL_ERRPPSLIMIT, CTL_EOL);
2867 	sysctl_createv(clog, 0, NULL, NULL,
2868 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2869 		       CTLTYPE_INT, "nd6_maxnudhint",
2870 		       SYSCTL_DESCR("Maximum neighbor unreachable hint count"),
2871 		       NULL, 0, &nd6_maxnudhint, 0,
2872 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2873 		       ICMPV6CTL_ND6_MAXNUDHINT, CTL_EOL);
2874 	sysctl_createv(clog, 0, NULL, NULL,
2875 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2876 		       CTLTYPE_INT, "mtudisc_hiwat",
2877 		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
2878 		       NULL, 0, &icmp6_mtudisc_hiwat, 0,
2879 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2880 		       ICMPV6CTL_MTUDISC_HIWAT, CTL_EOL);
2881 	sysctl_createv(clog, 0, NULL, NULL,
2882 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2883 		       CTLTYPE_INT, "mtudisc_lowat",
2884 		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
2885 		       NULL, 0, &icmp6_mtudisc_lowat, 0,
2886 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2887 		       ICMPV6CTL_MTUDISC_LOWAT, CTL_EOL);
2888 	sysctl_createv(clog, 0, NULL, NULL,
2889 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2890 		       CTLTYPE_INT, "nd6_debug",
2891 		       SYSCTL_DESCR("Enable neighbor discovery debug output"),
2892 		       NULL, 0, &nd6_debug, 0,
2893 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2894 		       ICMPV6CTL_ND6_DEBUG, CTL_EOL);
2895 	sysctl_createv(clog, 0, NULL, NULL,
2896 		       CTLFLAG_PERMANENT,
2897 		       CTLTYPE_STRUCT, "nd6_drlist",
2898 		       SYSCTL_DESCR("Default router list"),
2899 		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
2900 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2901 		       ICMPV6CTL_ND6_DRLIST, CTL_EOL);
2902 	sysctl_createv(clog, 0, NULL, NULL,
2903 		       CTLFLAG_PERMANENT,
2904 		       CTLTYPE_STRUCT, "nd6_prlist",
2905 		       SYSCTL_DESCR("Prefix list"),
2906 		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
2907 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2908 		       ICMPV6CTL_ND6_PRLIST, CTL_EOL);
2909 	sysctl_createv(clog, 0, NULL, NULL,
2910 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2911 		       CTLTYPE_INT, "maxqueuelen",
2912 		       SYSCTL_DESCR("max packet queue len for a unresolved ND"),
2913 		       NULL, 1, &nd6_maxqueuelen, 0,
2914 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2915 		       ICMPV6CTL_ND6_MAXQLEN, CTL_EOL);
2916 }
2917 
2918 void
2919 icmp6_statinc(u_int stat)
2920 {
2921 
2922 	KASSERT(stat < ICMP6_NSTATS);
2923 	ICMP6_STATINC(stat);
2924 }
2925