xref: /netbsd-src/sys/netinet6/icmp6.c (revision bf1e9b32e27832f0c493206710fb8b58a980838a)
1 /*	$NetBSD: icmp6.c,v 1.109 2005/05/29 21:43:51 christos 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 1.109 2005/05/29 21:43:51 christos 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 <netinet/icmp6.h>
93 #include <netinet6/mld6_var.h>
94 #include <netinet6/in6_pcb.h>
95 #include <netinet6/nd6.h>
96 #include <netinet6/in6_ifattach.h>
97 #include <netinet6/ip6protosw.h>
98 
99 #ifdef IPSEC
100 #include <netinet6/ipsec.h>
101 #include <netkey/key.h>
102 #endif
103 
104 #include "faith.h"
105 #if defined(NFAITH) && 0 < NFAITH
106 #include <net/if_faith.h>
107 #endif
108 
109 #include <net/net_osdep.h>
110 
111 extern struct domain inet6domain;
112 
113 struct icmp6stat icmp6stat;
114 
115 extern struct inpcbtable raw6cbtable;
116 extern int icmp6errppslim;
117 static int icmp6errpps_count = 0;
118 static struct timeval icmp6errppslim_last;
119 extern int icmp6_nodeinfo;
120 
121 /*
122  * List of callbacks to notify when Path MTU changes are made.
123  */
124 struct icmp6_mtudisc_callback {
125 	LIST_ENTRY(icmp6_mtudisc_callback) mc_list;
126 	void (*mc_func) __P((struct in6_addr *));
127 };
128 
129 LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks =
130     LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks);
131 
132 static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL;
133 extern int pmtu_expire;
134 
135 /* XXX do these values make any sense? */
136 static int icmp6_mtudisc_hiwat = 1280;
137 static int icmp6_mtudisc_lowat = 256;
138 
139 /*
140  * keep track of # of redirect routes.
141  */
142 static struct rttimer_queue *icmp6_redirect_timeout_q = NULL;
143 
144 /* XXX experimental, turned off */
145 static int icmp6_redirect_hiwat = -1;
146 static int icmp6_redirect_lowat = -1;
147 
148 static void icmp6_errcount __P((struct icmp6errstat *, int, int));
149 static int icmp6_rip6_input __P((struct mbuf **, int));
150 static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int));
151 static const char *icmp6_redirect_diag __P((struct in6_addr *,
152 	struct in6_addr *, struct in6_addr *));
153 static struct mbuf *ni6_input __P((struct mbuf *, int));
154 static struct mbuf *ni6_nametodns __P((const char *, int, int));
155 static int ni6_dnsmatch __P((const char *, int, const char *, int));
156 static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
157 			  struct ifnet **, char *));
158 static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
159 				struct ifnet *, int));
160 static int icmp6_notify_error __P((struct mbuf *, int, int, int));
161 static struct rtentry *icmp6_mtudisc_clone __P((struct sockaddr *));
162 static void icmp6_mtudisc_timeout __P((struct rtentry *, struct rttimer *));
163 static void icmp6_redirect_timeout __P((struct rtentry *, struct rttimer *));
164 
165 void
166 icmp6_init()
167 {
168 	mld6_init();
169 	icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire);
170 	icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout);
171 }
172 
173 static void
174 icmp6_errcount(stat, type, code)
175 	struct icmp6errstat *stat;
176 	int type, code;
177 {
178 	switch (type) {
179 	case ICMP6_DST_UNREACH:
180 		switch (code) {
181 		case ICMP6_DST_UNREACH_NOROUTE:
182 			stat->icp6errs_dst_unreach_noroute++;
183 			return;
184 		case ICMP6_DST_UNREACH_ADMIN:
185 			stat->icp6errs_dst_unreach_admin++;
186 			return;
187 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
188 			stat->icp6errs_dst_unreach_beyondscope++;
189 			return;
190 		case ICMP6_DST_UNREACH_ADDR:
191 			stat->icp6errs_dst_unreach_addr++;
192 			return;
193 		case ICMP6_DST_UNREACH_NOPORT:
194 			stat->icp6errs_dst_unreach_noport++;
195 			return;
196 		}
197 		break;
198 	case ICMP6_PACKET_TOO_BIG:
199 		stat->icp6errs_packet_too_big++;
200 		return;
201 	case ICMP6_TIME_EXCEEDED:
202 		switch (code) {
203 		case ICMP6_TIME_EXCEED_TRANSIT:
204 			stat->icp6errs_time_exceed_transit++;
205 			return;
206 		case ICMP6_TIME_EXCEED_REASSEMBLY:
207 			stat->icp6errs_time_exceed_reassembly++;
208 			return;
209 		}
210 		break;
211 	case ICMP6_PARAM_PROB:
212 		switch (code) {
213 		case ICMP6_PARAMPROB_HEADER:
214 			stat->icp6errs_paramprob_header++;
215 			return;
216 		case ICMP6_PARAMPROB_NEXTHEADER:
217 			stat->icp6errs_paramprob_nextheader++;
218 			return;
219 		case ICMP6_PARAMPROB_OPTION:
220 			stat->icp6errs_paramprob_option++;
221 			return;
222 		}
223 		break;
224 	case ND_REDIRECT:
225 		stat->icp6errs_redirect++;
226 		return;
227 	}
228 	stat->icp6errs_unknown++;
229 }
230 
231 /*
232  * Register a Path MTU Discovery callback.
233  */
234 void
235 icmp6_mtudisc_callback_register(func)
236 	void (*func) __P((struct in6_addr *));
237 {
238 	struct icmp6_mtudisc_callback *mc;
239 
240 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
241 	     mc = LIST_NEXT(mc, mc_list)) {
242 		if (mc->mc_func == func)
243 			return;
244 	}
245 
246 	mc = malloc(sizeof(*mc), M_PCB, M_NOWAIT);
247 	if (mc == NULL)
248 		panic("icmp6_mtudisc_callback_register");
249 
250 	mc->mc_func = func;
251 	LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list);
252 }
253 
254 /*
255  * Generate an error packet of type error in response to bad IP6 packet.
256  */
257 void
258 icmp6_error(m, type, code, param)
259 	struct mbuf *m;
260 	int type, code, param;
261 {
262 	struct ip6_hdr *oip6, *nip6;
263 	struct icmp6_hdr *icmp6;
264 	u_int preplen;
265 	int off;
266 	int nxt;
267 
268 	icmp6stat.icp6s_error++;
269 
270 	/* count per-type-code statistics */
271 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
272 
273 	if (m->m_flags & M_DECRYPTED) {
274 		icmp6stat.icp6s_canterror++;
275 		goto freeit;
276 	}
277 
278 	if (m->m_len < sizeof(struct ip6_hdr)) {
279 		m = m_pullup(m, sizeof(struct ip6_hdr));
280 		if (m == NULL)
281 			return;
282 	}
283 	oip6 = mtod(m, struct ip6_hdr *);
284 
285 	/*
286 	 * If the destination address of the erroneous packet is a multicast
287 	 * address, or the packet was sent using link-layer multicast,
288 	 * we should basically suppress sending an error (RFC 2463, Section
289 	 * 2.4).
290 	 * We have two exceptions (the item e.2 in that section):
291 	 * - the Pakcet Too Big message can be sent for path MTU discovery.
292 	 * - the Parameter Problem Message that can be allowed an icmp6 error
293 	 *   in the option type field.  This check has been done in
294 	 *   ip6_unknown_opt(), so we can just check the type and code.
295 	 */
296 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
297 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
298 	    (type != ICMP6_PACKET_TOO_BIG &&
299 	     (type != ICMP6_PARAM_PROB ||
300 	      code != ICMP6_PARAMPROB_OPTION)))
301 		goto freeit;
302 
303 	/*
304 	 * RFC 2463, 2.4 (e.5): source address check.
305 	 * XXX: the case of anycast source?
306 	 */
307 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
308 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
309 		goto freeit;
310 
311 	/*
312 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
313 	 * don't do it.
314 	 */
315 	nxt = -1;
316 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
317 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
318 		struct icmp6_hdr *icp;
319 
320 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
321 			sizeof(*icp));
322 		if (icp == NULL) {
323 			icmp6stat.icp6s_tooshort++;
324 			return;
325 		}
326 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
327 		    icp->icmp6_type == ND_REDIRECT) {
328 			/*
329 			 * ICMPv6 error
330 			 * Special case: for redirect (which is
331 			 * informational) we must not send icmp6 error.
332 			 */
333 			icmp6stat.icp6s_canterror++;
334 			goto freeit;
335 		} else {
336 			/* ICMPv6 informational - send the error */
337 		}
338 	}
339 #if 0 /* controversial */
340 	else if (off >= 0 && nxt == IPPROTO_ESP) {
341 		/*
342 		 * It could be ICMPv6 error inside ESP.  Take a safer side,
343 		 * don't respond.
344 		 */
345 		icmp6stat.icp6s_canterror++;
346 		goto freeit;
347 	}
348 #endif
349 	else {
350 		/* non-ICMPv6 - send the error */
351 	}
352 
353 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
354 
355 	/* Finally, do rate limitation check. */
356 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
357 		icmp6stat.icp6s_toofreq++;
358 		goto freeit;
359 	}
360 
361 	/*
362 	 * OK, ICMP6 can be generated.
363 	 */
364 
365 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
366 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
367 
368 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
369 	M_PREPEND(m, preplen, M_DONTWAIT);
370 	if (m && m->m_len < preplen)
371 		m = m_pullup(m, preplen);
372 	if (m == NULL) {
373 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
374 		return;
375 	}
376 
377 	nip6 = mtod(m, struct ip6_hdr *);
378 	nip6->ip6_src  = oip6->ip6_src;
379 	nip6->ip6_dst  = oip6->ip6_dst;
380 
381 	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src))
382 		oip6->ip6_src.s6_addr16[1] = 0;
383 	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst))
384 		oip6->ip6_dst.s6_addr16[1] = 0;
385 
386 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
387 	icmp6->icmp6_type = type;
388 	icmp6->icmp6_code = code;
389 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
390 
391 	/*
392 	 * icmp6_reflect() is designed to be in the input path.
393 	 * icmp6_error() can be called from both input and outut path,
394 	 * and if we are in output path rcvif could contain bogus value.
395 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
396 	 * information in ip header (nip6).
397 	 */
398 	m->m_pkthdr.rcvif = NULL;
399 
400 	icmp6stat.icp6s_outhist[type]++;
401 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
402 
403 	return;
404 
405   freeit:
406 	/*
407 	 * If we can't tell wheter or not we can generate ICMP6, free it.
408 	 */
409 	m_freem(m);
410 }
411 
412 /*
413  * Process a received ICMP6 message.
414  */
415 int
416 icmp6_input(mp, offp, proto)
417 	struct mbuf **mp;
418 	int *offp, proto;
419 {
420 	struct mbuf *m = *mp, *n;
421 	struct ip6_hdr *ip6, *nip6;
422 	struct icmp6_hdr *icmp6, *nicmp6;
423 	int off = *offp;
424 	int icmp6len = m->m_pkthdr.len - *offp;
425 	int code, sum, noff;
426 
427 	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
428 
429 	/*
430 	 * Locate icmp6 structure in mbuf, and check
431 	 * that not corrupted and of at least minimum length
432 	 */
433 
434 	ip6 = mtod(m, struct ip6_hdr *);
435 	if (icmp6len < sizeof(struct icmp6_hdr)) {
436 		icmp6stat.icp6s_tooshort++;
437 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
438 		goto freeit;
439 	}
440 
441 	/*
442 	 * calculate the checksum
443 	 */
444 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
445 	if (icmp6 == NULL) {
446 		icmp6stat.icp6s_tooshort++;
447 		/* m is invalid */
448 		/*icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);*/
449 		return IPPROTO_DONE;
450 	}
451 	KASSERT(IP6_HDR_ALIGNED_P(icmp6));
452 	code = icmp6->icmp6_code;
453 
454 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
455 		nd6log((LOG_ERR,
456 		    "ICMP6 checksum error(%d|%x) %s\n",
457 		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
458 		icmp6stat.icp6s_checksum++;
459 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
460 		goto freeit;
461 	}
462 
463 #if defined(NFAITH) && 0 < NFAITH
464 	if (faithprefix(&ip6->ip6_dst)) {
465 		/*
466 		 * Deliver very specific ICMP6 type only.
467 		 * This is important to deilver TOOBIG.  Otherwise PMTUD
468 		 * will not work.
469 		 */
470 		switch (icmp6->icmp6_type) {
471 		case ICMP6_DST_UNREACH:
472 		case ICMP6_PACKET_TOO_BIG:
473 		case ICMP6_TIME_EXCEEDED:
474 			break;
475 		default:
476 			goto freeit;
477 		}
478 	}
479 #endif
480 
481 	icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
482 
483 	switch (icmp6->icmp6_type) {
484 	case ICMP6_DST_UNREACH:
485 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
486 		switch (code) {
487 		case ICMP6_DST_UNREACH_NOROUTE:
488 			code = PRC_UNREACH_NET;
489 			break;
490 		case ICMP6_DST_UNREACH_ADMIN:
491 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
492 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
493 			break;
494 		case ICMP6_DST_UNREACH_ADDR:
495 			code = PRC_HOSTDEAD;
496 			break;
497 #ifdef COMPAT_RFC1885
498 		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
499 			code = PRC_UNREACH_SRCFAIL;
500 			break;
501 #else
502 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
503 			/* I mean "source address was incorrect." */
504 			code = PRC_UNREACH_NET;
505 			break;
506 #endif
507 		case ICMP6_DST_UNREACH_NOPORT:
508 			code = PRC_UNREACH_PORT;
509 			break;
510 		default:
511 			goto badcode;
512 		}
513 		goto deliver;
514 
515 	case ICMP6_PACKET_TOO_BIG:
516 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
517 
518 		code = PRC_MSGSIZE;
519 
520 		/*
521 		 * Updating the path MTU will be done after examining
522 		 * intermediate extension headers.
523 		 */
524 		goto deliver;
525 
526 	case ICMP6_TIME_EXCEEDED:
527 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
528 		switch (code) {
529 		case ICMP6_TIME_EXCEED_TRANSIT:
530 			code = PRC_TIMXCEED_INTRANS;
531 			break;
532 		case ICMP6_TIME_EXCEED_REASSEMBLY:
533 			code = PRC_TIMXCEED_REASS;
534 			break;
535 		default:
536 			goto badcode;
537 		}
538 		goto deliver;
539 
540 	case ICMP6_PARAM_PROB:
541 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
542 		switch (code) {
543 		case ICMP6_PARAMPROB_NEXTHEADER:
544 			code = PRC_UNREACH_PROTOCOL;
545 			break;
546 		case ICMP6_PARAMPROB_HEADER:
547 		case ICMP6_PARAMPROB_OPTION:
548 			code = PRC_PARAMPROB;
549 			break;
550 		default:
551 			goto badcode;
552 		}
553 		goto deliver;
554 
555 	case ICMP6_ECHO_REQUEST:
556 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
557 		if (code != 0)
558 			goto badcode;
559 		/*
560 		 * Copy mbuf to send to two data paths: userland socket(s),
561 		 * and to the querier (echo reply).
562 		 * m: a copy for socket, n: a copy for querier
563 		 */
564 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
565 			/* Give up local */
566 			n = m;
567 			m = NULL;
568 			goto deliverecho;
569 		}
570 		/*
571 		 * If the first mbuf is shared, or the first mbuf is too short,
572 		 * copy the first part of the data into a fresh mbuf.
573 		 * Otherwise, we will wrongly overwrite both copies.
574 		 */
575 		if ((n->m_flags & M_EXT) != 0 ||
576 		    n->m_len < off + sizeof(struct icmp6_hdr)) {
577 			struct mbuf *n0 = n;
578 			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
579 
580 			/*
581 			 * Prepare an internal mbuf.  m_pullup() doesn't
582 			 * always copy the length we specified.
583 			 */
584 			if (maxlen >= MCLBYTES) {
585 				/* Give up remote */
586 				m_freem(n0);
587 				break;
588 			}
589 			MGETHDR(n, M_DONTWAIT, n0->m_type);
590 			if (n && maxlen >= MHLEN) {
591 				MCLGET(n, M_DONTWAIT);
592 				if ((n->m_flags & M_EXT) == 0) {
593 					m_free(n);
594 					n = NULL;
595 				}
596 			}
597 			if (n == NULL) {
598 				/* Give up local */
599 				m_freem(n0);
600 				n = m;
601 				m = NULL;
602 				goto deliverecho;
603 			}
604 			M_COPY_PKTHDR(n, n0);
605 			/*
606 			 * Copy IPv6 and ICMPv6 only.
607 			 */
608 			nip6 = mtod(n, struct ip6_hdr *);
609 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
610 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
611 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
612 			noff = sizeof(struct ip6_hdr);
613 			n->m_len = noff + sizeof(struct icmp6_hdr);
614 			/*
615 			 * Adjust mbuf.  ip6_plen will be adjusted in
616 			 * ip6_output().
617 			 * n->m_pkthdr.len == n0->m_pkthdr.len at this point.
618 			 */
619 			n->m_pkthdr.len += noff + sizeof(struct icmp6_hdr);
620 			n->m_pkthdr.len -= (off + sizeof(struct icmp6_hdr));
621 			m_adj(n0, off + sizeof(struct icmp6_hdr));
622 			n->m_next = n0;
623 			n0->m_flags &= ~M_PKTHDR;
624 		} else {
625 	 deliverecho:
626 			nip6 = mtod(n, struct ip6_hdr *);
627 			nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
628 			noff = off;
629 		}
630 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
631 		nicmp6->icmp6_code = 0;
632 		if (n) {
633 			icmp6stat.icp6s_reflect++;
634 			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
635 			icmp6_reflect(n, noff);
636 		}
637 		if (!m)
638 			goto freeit;
639 		break;
640 
641 	case ICMP6_ECHO_REPLY:
642 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
643 		if (code != 0)
644 			goto badcode;
645 		break;
646 
647 	case MLD_LISTENER_QUERY:
648 	case MLD_LISTENER_REPORT:
649 		if (icmp6len < sizeof(struct mld_hdr))
650 			goto badlen;
651 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
652 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
653 		else
654 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
655 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
656 			/* give up local */
657 			mld6_input(m, off);
658 			m = NULL;
659 			goto freeit;
660 		}
661 		mld6_input(n, off);
662 		/* m stays. */
663 		break;
664 
665 	case MLD_LISTENER_DONE:
666 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
667 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
668 			goto badlen;
669 		break;		/* nothing to be done in kernel */
670 
671 	case MLD_MTRACE_RESP:
672 	case MLD_MTRACE:
673 		/* XXX: these two are experimental.  not officially defined. */
674 		/* XXX: per-interface statistics? */
675 		break;		/* just pass it to applications */
676 
677 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
678 	    {
679 		enum { WRU, FQDN } mode;
680 
681 		if (!icmp6_nodeinfo)
682 			break;
683 
684 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
685 			mode = WRU;
686 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
687 			mode = FQDN;
688 		else
689 			goto badlen;
690 
691 		if (mode == FQDN) {
692 			n = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
693 			if (n)
694 				n = ni6_input(n, off);
695 			/* XXX meaningless if n == NULL */
696 			noff = sizeof(struct ip6_hdr);
697 		} else {
698 			u_char *p;
699 			int maxlen, maxhlen;
700 
701 			if ((icmp6_nodeinfo & 5) != 5)
702 				break;
703 
704 			if (code != 0)
705 				goto badcode;
706 			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
707 			if (maxlen >= MCLBYTES) {
708 				/* Give up remote */
709 				break;
710 			}
711 			MGETHDR(n, M_DONTWAIT, m->m_type);
712 			if (n && 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) - 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 			icmp6stat.icp6s_reflect++;
748 			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
749 			icmp6_reflect(n, noff);
750 		}
751 		break;
752 	    }
753 
754 	case ICMP6_WRUREPLY:
755 		if (code != 0)
756 			goto badcode;
757 		break;
758 
759 	case ND_ROUTER_SOLICIT:
760 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
761 		if (code != 0)
762 			goto badcode;
763 		if (icmp6len < sizeof(struct nd_router_solicit))
764 			goto badlen;
765 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
766 			/* give up local */
767 			nd6_rs_input(m, off, icmp6len);
768 			m = NULL;
769 			goto freeit;
770 		}
771 		nd6_rs_input(n, off, icmp6len);
772 		/* m stays. */
773 		break;
774 
775 	case ND_ROUTER_ADVERT:
776 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
777 		if (code != 0)
778 			goto badcode;
779 		if (icmp6len < sizeof(struct nd_router_advert))
780 			goto badlen;
781 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
782 			/* give up local */
783 			nd6_ra_input(m, off, icmp6len);
784 			m = NULL;
785 			goto freeit;
786 		}
787 		nd6_ra_input(n, off, icmp6len);
788 		/* m stays. */
789 		break;
790 
791 	case ND_NEIGHBOR_SOLICIT:
792 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
793 		if (code != 0)
794 			goto badcode;
795 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
796 			goto badlen;
797 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
798 			/* give up local */
799 			nd6_ns_input(m, off, icmp6len);
800 			m = NULL;
801 			goto freeit;
802 		}
803 		nd6_ns_input(n, off, icmp6len);
804 		/* m stays. */
805 		break;
806 
807 	case ND_NEIGHBOR_ADVERT:
808 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
809 		if (code != 0)
810 			goto badcode;
811 		if (icmp6len < sizeof(struct nd_neighbor_advert))
812 			goto badlen;
813 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
814 			/* give up local */
815 			nd6_na_input(m, off, icmp6len);
816 			m = NULL;
817 			goto freeit;
818 		}
819 		nd6_na_input(n, off, icmp6len);
820 		/* m stays. */
821 		break;
822 
823 	case ND_REDIRECT:
824 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
825 		if (code != 0)
826 			goto badcode;
827 		if (icmp6len < sizeof(struct nd_redirect))
828 			goto badlen;
829 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
830 			/* give up local */
831 			icmp6_redirect_input(m, off);
832 			m = NULL;
833 			goto freeit;
834 		}
835 		icmp6_redirect_input(n, off);
836 		/* m stays. */
837 		break;
838 
839 	case ICMP6_ROUTER_RENUMBERING:
840 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
841 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
842 			goto badcode;
843 		if (icmp6len < sizeof(struct icmp6_router_renum))
844 			goto badlen;
845 		break;
846 
847 	default:
848 		nd6log((LOG_DEBUG,
849 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
850 		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
851 		    ip6_sprintf(&ip6->ip6_dst),
852 		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
853 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
854 			/* ICMPv6 error: MUST deliver it by spec... */
855 			code = PRC_NCMDS;
856 			/* deliver */
857 		} else {
858 			/* ICMPv6 informational: MUST not deliver */
859 			break;
860 		}
861 	deliver:
862 		if (icmp6_notify_error(m, off, icmp6len, code)) {
863 			/* In this case, m should've been freed. */
864 			return (IPPROTO_DONE);
865 		}
866 		break;
867 
868 	badcode:
869 		icmp6stat.icp6s_badcode++;
870 		break;
871 
872 	badlen:
873 		icmp6stat.icp6s_badlen++;
874 		break;
875 	}
876 
877 	/* deliver the packet to appropriate sockets */
878 	icmp6_rip6_input(&m, *offp);
879 
880 	return IPPROTO_DONE;
881 
882  freeit:
883 	m_freem(m);
884 	return IPPROTO_DONE;
885 }
886 
887 static int
888 icmp6_notify_error(m, off, icmp6len, code)
889 	struct mbuf *m;
890 	int off, icmp6len;
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 		icmp6stat.icp6s_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 		icmp6stat.icp6s_tooshort++;
905 		return (-1);
906 	}
907 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
908 
909 	/* Detect the upper level protocol */
910 	{
911 		void (*ctlfunc) __P((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 					icmp6stat.icp6s_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 					icmp6stat.icp6s_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 						icmp6stat.icp6s_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 					icmp6stat.icp6s_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 			icmp6stat.icp6s_tooshort++;
1022 			return (-1);
1023 		}
1024 
1025 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1026 		bzero(&icmp6dst, sizeof(icmp6dst));
1027 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1028 		icmp6dst.sin6_family = AF_INET6;
1029 		if (finaldst == NULL)
1030 			icmp6dst.sin6_addr = eip6->ip6_dst;
1031 		else
1032 			icmp6dst.sin6_addr = *finaldst;
1033 		icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1034 							  &icmp6dst.sin6_addr);
1035 #ifndef SCOPEDROUTING
1036 		if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1037 				   NULL, NULL)) {
1038 			/* should be impossbile */
1039 			nd6log((LOG_DEBUG,
1040 			    "icmp6_notify_error: in6_embedscope failed\n"));
1041 			goto freeit;
1042 		}
1043 #endif
1044 
1045 		/*
1046 		 * retrieve parameters from the inner IPv6 header, and convert
1047 		 * them into sockaddr structures.
1048 		 */
1049 		bzero(&icmp6src, sizeof(icmp6src));
1050 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1051 		icmp6src.sin6_family = AF_INET6;
1052 		icmp6src.sin6_addr = eip6->ip6_src;
1053 		icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1054 							  &icmp6src.sin6_addr);
1055 #ifndef SCOPEDROUTING
1056 		if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1057 				   NULL, NULL)) {
1058 			/* should be impossbile */
1059 			nd6log((LOG_DEBUG,
1060 			    "icmp6_notify_error: in6_embedscope failed\n"));
1061 			goto freeit;
1062 		}
1063 #endif
1064 		icmp6src.sin6_flowinfo =
1065 			(eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1066 
1067 		if (finaldst == NULL)
1068 			finaldst = &eip6->ip6_dst;
1069 		ip6cp.ip6c_m = m;
1070 		ip6cp.ip6c_icmp6 = icmp6;
1071 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1072 		ip6cp.ip6c_off = eoff;
1073 		ip6cp.ip6c_finaldst = finaldst;
1074 		ip6cp.ip6c_src = &icmp6src;
1075 		ip6cp.ip6c_nxt = nxt;
1076 
1077 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1078 			notifymtu = ntohl(icmp6->icmp6_mtu);
1079 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1080 		}
1081 
1082 		ctlfunc = (void (*) __P((int, struct sockaddr *, void *)))
1083 			(inet6sw[ip6_protox[nxt]].pr_ctlinput);
1084 		if (ctlfunc) {
1085 			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1086 					  &ip6cp);
1087 		}
1088 	}
1089 	return (0);
1090 
1091   freeit:
1092 	m_freem(m);
1093 	return (-1);
1094 }
1095 
1096 void
1097 icmp6_mtudisc_update(ip6cp, validated)
1098 	struct ip6ctlparam *ip6cp;
1099 	int validated;
1100 {
1101 	unsigned long rtcount;
1102 	struct icmp6_mtudisc_callback *mc;
1103 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1104 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1105 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1106 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1107 	struct rtentry *rt = NULL;
1108 	struct sockaddr_in6 sin6;
1109 
1110 	/*
1111 	 * allow non-validated cases if memory is plenty, to make traffic
1112 	 * from non-connected pcb happy.
1113 	 */
1114 	rtcount = rt_timer_count(icmp6_mtudisc_timeout_q);
1115 	if (validated) {
1116 		if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat)
1117 			return;
1118 		else if (0 <= icmp6_mtudisc_lowat &&
1119 		    rtcount > icmp6_mtudisc_lowat) {
1120 			/*
1121 			 * XXX nuke a victim, install the new one.
1122 			 */
1123 		}
1124 	} else {
1125 		if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat)
1126 			return;
1127 	}
1128 
1129 	bzero(&sin6, sizeof(sin6));
1130 	sin6.sin6_family = PF_INET6;
1131 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1132 	sin6.sin6_addr = *dst;
1133 	/* XXX normally, this won't happen */
1134 	if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1135 		sin6.sin6_addr.s6_addr16[1] =
1136 		    htons(m->m_pkthdr.rcvif->if_index);
1137 	}
1138 	/* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1139 	rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6);
1140 
1141 	if (rt && (rt->rt_flags & RTF_HOST) &&
1142 	    !(rt->rt_rmx.rmx_locks & RTV_MTU) &&
1143 	    (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) {
1144 		if (mtu < IN6_LINKMTU(rt->rt_ifp)) {
1145 			icmp6stat.icp6s_pmtuchg++;
1146 			rt->rt_rmx.rmx_mtu = mtu;
1147 		}
1148 	}
1149 	if (rt) { /* XXX: need braces to avoid conflict with else in RTFREE. */
1150 		RTFREE(rt);
1151 	}
1152 
1153 	/*
1154 	 * Notify protocols that the MTU for this destination
1155 	 * has changed.
1156 	 */
1157 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
1158 	     mc = LIST_NEXT(mc, mc_list))
1159 		(*mc->mc_func)(&sin6.sin6_addr);
1160 }
1161 
1162 /*
1163  * Process a Node Information Query packet, based on
1164  * draft-ietf-ipngwg-icmp-name-lookups-07.
1165  *
1166  * Spec incompatibilities:
1167  * - IPv6 Subject address handling
1168  * - IPv4 Subject address handling support missing
1169  * - Proxy reply (answer even if it's not for me)
1170  * - joins NI group address at in6_ifattach() time only, does not cope
1171  *   with hostname changes by sethostname(3)
1172  */
1173 #ifndef offsetof		/* XXX */
1174 #define	offsetof(type, member)	((size_t)(&((type *)0)->member))
1175 #endif
1176 static struct mbuf *
1177 ni6_input(m, off)
1178 	struct mbuf *m;
1179 	int off;
1180 {
1181 	struct icmp6_nodeinfo *ni6, *nni6;
1182 	struct mbuf *n = NULL;
1183 	u_int16_t qtype;
1184 	int subjlen;
1185 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1186 	struct ni_reply_fqdn *fqdn;
1187 	int addrs;		/* for NI_QTYPE_NODEADDR */
1188 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1189 	struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1190 	struct ip6_hdr *ip6;
1191 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1192 	char *subj = NULL;
1193 
1194 	ip6 = mtod(m, struct ip6_hdr *);
1195 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1196 	if (ni6 == NULL) {
1197 		/* m is already reclaimed */
1198 		return NULL;
1199 	}
1200 
1201 	/*
1202 	 * Validate IPv6 destination address.
1203 	 *
1204 	 * The Responder must discard the Query without further processing
1205 	 * unless it is one of the Responder's unicast or anycast addresses, or
1206 	 * a link-local scope multicast address which the Responder has joined.
1207 	 * [icmp-name-lookups-07, Section 4.]
1208 	 */
1209 	bzero(&sin6, sizeof(sin6));
1210 	sin6.sin6_family = AF_INET6;
1211 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1212 	bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1213 	/* XXX scopeid */
1214 	if (ifa_ifwithaddr((struct sockaddr *)&sin6))
1215 		; /* unicast/anycast, fine */
1216 	else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1217 		; /* link-local multicast, fine */
1218 	else
1219 		goto bad;
1220 
1221 	/* validate query Subject field. */
1222 	qtype = ntohs(ni6->ni_qtype);
1223 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1224 	switch (qtype) {
1225 	case NI_QTYPE_NOOP:
1226 	case NI_QTYPE_SUPTYPES:
1227 		/* 07 draft */
1228 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1229 			break;
1230 		/* FALLTHROUGH */
1231 	case NI_QTYPE_FQDN:
1232 	case NI_QTYPE_NODEADDR:
1233 		switch (ni6->ni_code) {
1234 		case ICMP6_NI_SUBJ_IPV6:
1235 #if ICMP6_NI_SUBJ_IPV6 != 0
1236 		case 0:
1237 #endif
1238 			/*
1239 			 * backward compatibility - try to accept 03 draft
1240 			 * format, where no Subject is present.
1241 			 */
1242 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1243 			    subjlen == 0) {
1244 				oldfqdn++;
1245 				break;
1246 			}
1247 #if ICMP6_NI_SUBJ_IPV6 != 0
1248 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1249 				goto bad;
1250 #endif
1251 
1252 			if (subjlen != sizeof(sin6.sin6_addr))
1253 				goto bad;
1254 
1255 			/*
1256 			 * Validate Subject address.
1257 			 *
1258 			 * Not sure what exactly "address belongs to the node"
1259 			 * means in the spec, is it just unicast, or what?
1260 			 *
1261 			 * At this moment we consider Subject address as
1262 			 * "belong to the node" if the Subject address equals
1263 			 * to the IPv6 destination address; validation for
1264 			 * IPv6 destination address should have done enough
1265 			 * check for us.
1266 			 *
1267 			 * We do not do proxy at this moment.
1268 			 */
1269 			/* m_pulldown instead of copy? */
1270 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1271 			    subjlen, (caddr_t)&sin6.sin6_addr);
1272 			/* XXX kame scope hack */
1273 			if (IN6_IS_SCOPE_LINKLOCAL(&sin6.sin6_addr)) {
1274 				if ((m->m_flags & M_PKTHDR) != 0 &&
1275 				    m->m_pkthdr.rcvif) {
1276 					sin6.sin6_addr.s6_addr16[1] =
1277 					    htons(m->m_pkthdr.rcvif->if_index);
1278 				}
1279 			}
1280 			subj = (char *)&sin6;
1281 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &sin6.sin6_addr))
1282 				break;
1283 
1284 			/*
1285 			 * XXX if we are to allow other cases, we should really
1286 			 * be careful about scope here.
1287 			 * basically, we should disallow queries toward IPv6
1288 			 * destination X with subject Y, if scope(X) > scope(Y).
1289 			 * if we allow scope(X) > scope(Y), it will result in
1290 			 * information leakage across scope boundary.
1291 			 */
1292 			goto bad;
1293 
1294 		case ICMP6_NI_SUBJ_FQDN:
1295 			/*
1296 			 * Validate Subject name with gethostname(3).
1297 			 *
1298 			 * The behavior may need some debate, since:
1299 			 * - we are not sure if the node has FQDN as
1300 			 *   hostname (returned by gethostname(3)).
1301 			 * - the code does wildcard match for truncated names.
1302 			 *   however, we are not sure if we want to perform
1303 			 *   wildcard match, if gethostname(3) side has
1304 			 *   truncated hostname.
1305 			 */
1306 			n = ni6_nametodns(hostname, hostnamelen, 0);
1307 			if (!n || n->m_next || n->m_len == 0)
1308 				goto bad;
1309 			IP6_EXTHDR_GET(subj, char *, m,
1310 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1311 			if (subj == NULL)
1312 				goto bad;
1313 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1314 					n->m_len)) {
1315 				goto bad;
1316 			}
1317 			m_freem(n);
1318 			n = NULL;
1319 			break;
1320 
1321 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1322 		default:
1323 			goto bad;
1324 		}
1325 		break;
1326 	}
1327 
1328 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1329 	switch (qtype) {
1330 	case NI_QTYPE_FQDN:
1331 		if ((icmp6_nodeinfo & 1) == 0)
1332 			goto bad;
1333 		break;
1334 	case NI_QTYPE_NODEADDR:
1335 		if ((icmp6_nodeinfo & 2) == 0)
1336 			goto bad;
1337 		break;
1338 	}
1339 
1340 	/* guess reply length */
1341 	switch (qtype) {
1342 	case NI_QTYPE_NOOP:
1343 		break;		/* no reply data */
1344 	case NI_QTYPE_SUPTYPES:
1345 		replylen += sizeof(u_int32_t);
1346 		break;
1347 	case NI_QTYPE_FQDN:
1348 		/* XXX will append an mbuf */
1349 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1350 		break;
1351 	case NI_QTYPE_NODEADDR:
1352 		addrs = ni6_addrs(ni6, m, &ifp, subj);
1353 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1354 					  sizeof(u_int32_t))) > MCLBYTES)
1355 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1356 		break;
1357 	default:
1358 		/*
1359 		 * XXX: We must return a reply with the ICMP6 code
1360 		 * `unknown Qtype' in this case.  However we regard the case
1361 		 * as an FQDN query for backward compatibility.
1362 		 * Older versions set a random value to this field,
1363 		 * so it rarely varies in the defined qtypes.
1364 		 * But the mechanism is not reliable...
1365 		 * maybe we should obsolete older versions.
1366 		 */
1367 		qtype = NI_QTYPE_FQDN;
1368 		/* XXX will append an mbuf */
1369 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1370 		oldfqdn++;
1371 		break;
1372 	}
1373 
1374 	/* allocate an mbuf to reply. */
1375 	MGETHDR(n, M_DONTWAIT, m->m_type);
1376 	if (n == NULL) {
1377 		m_freem(m);
1378 		return (NULL);
1379 	}
1380 	M_COPY_PKTHDR(n, m); /* just for rcvif */
1381 	if (replylen > MHLEN) {
1382 		if (replylen > MCLBYTES) {
1383 			/*
1384 			 * XXX: should we try to allocate more? But MCLBYTES
1385 			 * is probably much larger than IPV6_MMTU...
1386 			 */
1387 			goto bad;
1388 		}
1389 		MCLGET(n, M_DONTWAIT);
1390 		if ((n->m_flags & M_EXT) == 0) {
1391 			goto bad;
1392 		}
1393 	}
1394 	n->m_pkthdr.len = n->m_len = replylen;
1395 
1396 	/* copy mbuf header and IPv6 + Node Information base headers */
1397 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1398 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1399 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1400 
1401 	/* qtype dependent procedure */
1402 	switch (qtype) {
1403 	case NI_QTYPE_NOOP:
1404 		nni6->ni_code = ICMP6_NI_SUCCESS;
1405 		nni6->ni_flags = 0;
1406 		break;
1407 	case NI_QTYPE_SUPTYPES:
1408 	{
1409 		u_int32_t v;
1410 		nni6->ni_code = ICMP6_NI_SUCCESS;
1411 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1412 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1413 		v = (u_int32_t)htonl(0x0000000f);
1414 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1415 		break;
1416 	}
1417 	case NI_QTYPE_FQDN:
1418 		nni6->ni_code = ICMP6_NI_SUCCESS;
1419 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1420 						sizeof(struct ip6_hdr) +
1421 						sizeof(struct icmp6_nodeinfo));
1422 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1423 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1424 		/*
1425 		 * XXX do we really have FQDN in variable "hostname"?
1426 		 */
1427 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1428 		if (n->m_next == NULL)
1429 			goto bad;
1430 		/* XXX we assume that n->m_next is not a chain */
1431 		if (n->m_next->m_next != NULL)
1432 			goto bad;
1433 		n->m_pkthdr.len += n->m_next->m_len;
1434 		break;
1435 	case NI_QTYPE_NODEADDR:
1436 	{
1437 		int lenlim, copied;
1438 
1439 		nni6->ni_code = ICMP6_NI_SUCCESS;
1440 		n->m_pkthdr.len = n->m_len =
1441 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1442 		lenlim = M_TRAILINGSPACE(n);
1443 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1444 		/* XXX: reset mbuf length */
1445 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1446 			sizeof(struct icmp6_nodeinfo) + copied;
1447 		break;
1448 	}
1449 	default:
1450 		break;		/* XXX impossible! */
1451 	}
1452 
1453 	nni6->ni_type = ICMP6_NI_REPLY;
1454 	m_freem(m);
1455 	return (n);
1456 
1457   bad:
1458 	m_freem(m);
1459 	if (n)
1460 		m_freem(n);
1461 	return (NULL);
1462 }
1463 #undef hostnamelen
1464 
1465 #define isupper(x) ('A' <= (x) && (x) <= 'Z')
1466 #define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z'))
1467 #define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9'))
1468 #define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x))
1469 
1470 /*
1471  * make a mbuf with DNS-encoded string.  no compression support.
1472  *
1473  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1474  * treated as truncated name (two \0 at the end).  this is a wild guess.
1475  */
1476 static struct mbuf *
1477 ni6_nametodns(name, namelen, old)
1478 	const char *name;
1479 	int namelen;
1480 	int old;	/* return pascal string if non-zero */
1481 {
1482 	struct mbuf *m;
1483 	char *cp, *ep;
1484 	const char *p, *q;
1485 	int i, len, nterm;
1486 
1487 	if (old)
1488 		len = namelen + 1;
1489 	else
1490 		len = MCLBYTES;
1491 
1492 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1493 	MGET(m, M_DONTWAIT, MT_DATA);
1494 	if (m && len > MLEN) {
1495 		MCLGET(m, M_DONTWAIT);
1496 		if ((m->m_flags & M_EXT) == 0)
1497 			goto fail;
1498 	}
1499 	if (!m)
1500 		goto fail;
1501 	m->m_next = NULL;
1502 
1503 	if (old) {
1504 		m->m_len = len;
1505 		*mtod(m, char *) = namelen;
1506 		bcopy(name, mtod(m, char *) + 1, namelen);
1507 		return m;
1508 	} else {
1509 		m->m_len = 0;
1510 		cp = mtod(m, char *);
1511 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1512 
1513 		/* if not certain about my name, return empty buffer */
1514 		if (namelen == 0)
1515 			return m;
1516 
1517 		/*
1518 		 * guess if it looks like shortened hostname, or FQDN.
1519 		 * shortened hostname needs two trailing "\0".
1520 		 */
1521 		i = 0;
1522 		for (p = name; p < name + namelen; p++) {
1523 			if (*p && *p == '.')
1524 				i++;
1525 		}
1526 		if (i < 2)
1527 			nterm = 2;
1528 		else
1529 			nterm = 1;
1530 
1531 		p = name;
1532 		while (cp < ep && p < name + namelen) {
1533 			i = 0;
1534 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1535 				i++;
1536 			/* result does not fit into mbuf */
1537 			if (cp + i + 1 >= ep)
1538 				goto fail;
1539 			/*
1540 			 * DNS label length restriction, RFC1035 page 8.
1541 			 * "i == 0" case is included here to avoid returning
1542 			 * 0-length label on "foo..bar".
1543 			 */
1544 			if (i <= 0 || i >= 64)
1545 				goto fail;
1546 			*cp++ = i;
1547 			if (!isalpha(p[0]) || !isalnum(p[i - 1]))
1548 				goto fail;
1549 			while (i > 0) {
1550 				if (!isalnum(*p) && *p != '-')
1551 					goto fail;
1552 				if (isupper(*p)) {
1553 					*cp++ = tolower(*p);
1554 					p++;
1555 				} else
1556 					*cp++ = *p++;
1557 				i--;
1558 			}
1559 			p = q;
1560 			if (p < name + namelen && *p == '.')
1561 				p++;
1562 		}
1563 		/* termination */
1564 		if (cp + nterm >= ep)
1565 			goto fail;
1566 		while (nterm-- > 0)
1567 			*cp++ = '\0';
1568 		m->m_len = cp - mtod(m, char *);
1569 		return m;
1570 	}
1571 
1572 	panic("should not reach here");
1573 	/* NOTREACHED */
1574 
1575  fail:
1576 	if (m)
1577 		m_freem(m);
1578 	return NULL;
1579 }
1580 
1581 /*
1582  * check if two DNS-encoded string matches.  takes care of truncated
1583  * form (with \0\0 at the end).  no compression support.
1584  * XXX upper/lowercase match (see RFC2065)
1585  */
1586 static int
1587 ni6_dnsmatch(a, alen, b, blen)
1588 	const char *a;
1589 	int alen;
1590 	const char *b;
1591 	int blen;
1592 {
1593 	const char *a0, *b0;
1594 	int l;
1595 
1596 	/* simplest case - need validation? */
1597 	if (alen == blen && bcmp(a, b, alen) == 0)
1598 		return 1;
1599 
1600 	a0 = a;
1601 	b0 = b;
1602 
1603 	/* termination is mandatory */
1604 	if (alen < 2 || blen < 2)
1605 		return 0;
1606 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1607 		return 0;
1608 	alen--;
1609 	blen--;
1610 
1611 	while (a - a0 < alen && b - b0 < blen) {
1612 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1613 			return 0;
1614 
1615 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1616 			return 0;
1617 		/* we don't support compression yet */
1618 		if (a[0] >= 64 || b[0] >= 64)
1619 			return 0;
1620 
1621 		/* truncated case */
1622 		if (a[0] == 0 && a - a0 == alen - 1)
1623 			return 1;
1624 		if (b[0] == 0 && b - b0 == blen - 1)
1625 			return 1;
1626 		if (a[0] == 0 || b[0] == 0)
1627 			return 0;
1628 
1629 		if (a[0] != b[0])
1630 			return 0;
1631 		l = a[0];
1632 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1633 			return 0;
1634 		if (bcmp(a + 1, b + 1, l) != 0)
1635 			return 0;
1636 
1637 		a += 1 + l;
1638 		b += 1 + l;
1639 	}
1640 
1641 	if (a - a0 == alen && b - b0 == blen)
1642 		return 1;
1643 	else
1644 		return 0;
1645 }
1646 
1647 /*
1648  * calculate the number of addresses to be returned in the node info reply.
1649  */
1650 static int
1651 ni6_addrs(ni6, m, ifpp, subj)
1652 	struct icmp6_nodeinfo *ni6;
1653 	struct mbuf *m;
1654 	struct ifnet **ifpp;
1655 	char *subj;
1656 {
1657 	struct ifnet *ifp;
1658 	struct in6_ifaddr *ifa6;
1659 	struct ifaddr *ifa;
1660 	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1661 	int addrs = 0, addrsofif, iffound = 0;
1662 	int niflags = ni6->ni_flags;
1663 
1664 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1665 		switch (ni6->ni_code) {
1666 		case ICMP6_NI_SUBJ_IPV6:
1667 			if (subj == NULL) /* must be impossible... */
1668 				return (0);
1669 			subj_ip6 = (struct sockaddr_in6 *)subj;
1670 			break;
1671 		default:
1672 			/*
1673 			 * XXX: we only support IPv6 subject address for
1674 			 * this Qtype.
1675 			 */
1676 			return (0);
1677 		}
1678 	}
1679 
1680 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
1681 	{
1682 		addrsofif = 0;
1683 		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1684 		     ifa = ifa->ifa_list.tqe_next)
1685 		{
1686 			if (ifa->ifa_addr->sa_family != AF_INET6)
1687 				continue;
1688 			ifa6 = (struct in6_ifaddr *)ifa;
1689 
1690 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1691 			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1692 					       &ifa6->ia_addr.sin6_addr))
1693 				iffound = 1;
1694 
1695 			/*
1696 			 * IPv4-mapped addresses can only be returned by a
1697 			 * Node Information proxy, since they represent
1698 			 * addresses of IPv4-only nodes, which perforce do
1699 			 * not implement this protocol.
1700 			 * [icmp-name-lookups-07, Section 5.4]
1701 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1702 			 * this function at this moment.
1703 			 */
1704 
1705 			/* What do we have to do about ::1? */
1706 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1707 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1708 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1709 					continue;
1710 				break;
1711 			case IPV6_ADDR_SCOPE_SITELOCAL:
1712 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1713 					continue;
1714 				break;
1715 			case IPV6_ADDR_SCOPE_GLOBAL:
1716 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1717 					continue;
1718 				break;
1719 			default:
1720 				continue;
1721 			}
1722 
1723 			/*
1724 			 * check if anycast is okay.
1725 			 * XXX: just experimental.  not in the spec.
1726 			 */
1727 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1728 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1729 				continue; /* we need only unicast addresses */
1730 
1731 			addrsofif++; /* count the address */
1732 		}
1733 		if (iffound) {
1734 			*ifpp = ifp;
1735 			return (addrsofif);
1736 		}
1737 
1738 		addrs += addrsofif;
1739 	}
1740 
1741 	return (addrs);
1742 }
1743 
1744 static int
1745 ni6_store_addrs(ni6, nni6, ifp0, resid)
1746 	struct icmp6_nodeinfo *ni6, *nni6;
1747 	struct ifnet *ifp0;
1748 	int resid;
1749 {
1750 	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1751 	struct in6_ifaddr *ifa6;
1752 	struct ifaddr *ifa;
1753 	struct ifnet *ifp_dep = NULL;
1754 	int copied = 0, allow_deprecated = 0;
1755 	u_char *cp = (u_char *)(nni6 + 1);
1756 	int niflags = ni6->ni_flags;
1757 	u_int32_t ltime;
1758 
1759 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1760 		return (0);	/* needless to copy */
1761 
1762   again:
1763 
1764 	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list))
1765 	{
1766 		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1767 		     ifa = ifa->ifa_list.tqe_next)
1768 		{
1769 			if (ifa->ifa_addr->sa_family != AF_INET6)
1770 				continue;
1771 			ifa6 = (struct in6_ifaddr *)ifa;
1772 
1773 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1774 			    allow_deprecated == 0) {
1775 				/*
1776 				 * prefererred address should be put before
1777 				 * deprecated addresses.
1778 				 */
1779 
1780 				/* record the interface for later search */
1781 				if (ifp_dep == NULL)
1782 					ifp_dep = ifp;
1783 
1784 				continue;
1785 			}
1786 			else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1787 				 allow_deprecated != 0)
1788 				continue; /* we now collect deprecated addrs */
1789 
1790 			/* What do we have to do about ::1? */
1791 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1792 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1793 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1794 					continue;
1795 				break;
1796 			case IPV6_ADDR_SCOPE_SITELOCAL:
1797 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1798 					continue;
1799 				break;
1800 			case IPV6_ADDR_SCOPE_GLOBAL:
1801 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1802 					continue;
1803 				break;
1804 			default:
1805 				continue;
1806 			}
1807 
1808 			/*
1809 			 * check if anycast is okay.
1810 			 * XXX: just experimental.  not in the spec.
1811 			 */
1812 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1813 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1814 				continue;
1815 
1816 			/* now we can copy the address */
1817 			if (resid < sizeof(struct in6_addr) +
1818 			    sizeof(u_int32_t)) {
1819 				/*
1820 				 * We give up much more copy.
1821 				 * Set the truncate flag and return.
1822 				 */
1823 				nni6->ni_flags |=
1824 					NI_NODEADDR_FLAG_TRUNCATE;
1825 				return (copied);
1826 			}
1827 
1828 			/*
1829 			 * Set the TTL of the address.
1830 			 * The TTL value should be one of the following
1831 			 * according to the specification:
1832 			 *
1833 			 * 1. The remaining lifetime of a DHCP lease on the
1834 			 *    address, or
1835 			 * 2. The remaining Valid Lifetime of a prefix from
1836 			 *    which the address was derived through Stateless
1837 			 *    Autoconfiguration.
1838 			 *
1839 			 * Note that we currently do not support stateful
1840 			 * address configuration by DHCPv6, so the former
1841 			 * case can't happen.
1842 			 *
1843 			 * TTL must be 2^31 > TTL >= 0.
1844 			 */
1845 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1846 				ltime = ND6_INFINITE_LIFETIME;
1847 			else {
1848 				if (ifa6->ia6_lifetime.ia6t_expire >
1849 				    time.tv_sec)
1850 					ltime = ifa6->ia6_lifetime.ia6t_expire - time.tv_sec;
1851 				else
1852 					ltime = 0;
1853 			}
1854 			if (ltime > 0x7fffffff)
1855 				ltime = 0x7fffffff;
1856 			ltime = htonl(ltime);
1857 
1858 			bcopy(&ltime, cp, sizeof(u_int32_t));
1859 			cp += sizeof(u_int32_t);
1860 
1861 			/* copy the address itself */
1862 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1863 			      sizeof(struct in6_addr));
1864 			/* XXX: KAME link-local hack; remove ifindex */
1865 			if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr))
1866 				((struct in6_addr *)cp)->s6_addr16[1] = 0;
1867 			cp += sizeof(struct in6_addr);
1868 
1869 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1870 			copied += (sizeof(struct in6_addr) +
1871 				   sizeof(u_int32_t));
1872 		}
1873 		if (ifp0)	/* we need search only on the specified IF */
1874 			break;
1875 	}
1876 
1877 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1878 		ifp = ifp_dep;
1879 		allow_deprecated = 1;
1880 
1881 		goto again;
1882 	}
1883 
1884 	return (copied);
1885 }
1886 
1887 /*
1888  * XXX almost dup'ed code with rip6_input.
1889  */
1890 static int
1891 icmp6_rip6_input(mp, off)
1892 	struct	mbuf **mp;
1893 	int	off;
1894 {
1895 	struct mbuf *m = *mp;
1896 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1897 	struct inpcb_hdr *inph;
1898 	struct in6pcb *in6p;
1899 	struct in6pcb *last = NULL;
1900 	struct sockaddr_in6 rip6src;
1901 	struct icmp6_hdr *icmp6;
1902 	struct mbuf *opts = NULL;
1903 
1904 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1905 	if (icmp6 == NULL) {
1906 		/* m is already reclaimed */
1907 		return IPPROTO_DONE;
1908 	}
1909 
1910 	bzero(&rip6src, sizeof(rip6src));
1911 	rip6src.sin6_len = sizeof(struct sockaddr_in6);
1912 	rip6src.sin6_family = AF_INET6;
1913 	/* KAME hack: recover scopeid */
1914 	(void)in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif);
1915 
1916 	CIRCLEQ_FOREACH(inph, &raw6cbtable.inpt_queue, inph_queue) {
1917 		in6p = (struct in6pcb *)inph;
1918 		if (in6p->in6p_af != AF_INET6)
1919 			continue;
1920 		if (in6p->in6p_ip6.ip6_nxt != IPPROTO_ICMPV6)
1921 			continue;
1922 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1923 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1924 			continue;
1925 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1926 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1927 			continue;
1928 		if (in6p->in6p_icmp6filt
1929 		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1930 				 in6p->in6p_icmp6filt))
1931 			continue;
1932 		if (last) {
1933 			struct	mbuf *n;
1934 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1935 				if (last->in6p_flags & IN6P_CONTROLOPTS)
1936 					ip6_savecontrol(last, &opts, ip6, n);
1937 				/* strip intermediate headers */
1938 				m_adj(n, off);
1939 				if (sbappendaddr(&last->in6p_socket->so_rcv,
1940 						 (struct sockaddr *)&rip6src,
1941 						 n, opts) == 0) {
1942 					/* should notify about lost packet */
1943 					m_freem(n);
1944 					if (opts)
1945 						m_freem(opts);
1946 				} else
1947 					sorwakeup(last->in6p_socket);
1948 				opts = NULL;
1949 			}
1950 		}
1951 		last = in6p;
1952 	}
1953 	if (last) {
1954 		if (last->in6p_flags & IN6P_CONTROLOPTS)
1955 			ip6_savecontrol(last, &opts, ip6, m);
1956 		/* strip intermediate headers */
1957 		m_adj(m, off);
1958 		if (sbappendaddr(&last->in6p_socket->so_rcv,
1959 				(struct sockaddr *)&rip6src, m, opts) == 0) {
1960 			m_freem(m);
1961 			if (opts)
1962 				m_freem(opts);
1963 		} else
1964 			sorwakeup(last->in6p_socket);
1965 	} else {
1966 		m_freem(m);
1967 		ip6stat.ip6s_delivered--;
1968 	}
1969 	return IPPROTO_DONE;
1970 }
1971 
1972 /*
1973  * Reflect the ip6 packet back to the source.
1974  * OFF points to the icmp6 header, counted from the top of the mbuf.
1975  *
1976  * Note: RFC 1885 required that an echo reply should be truncated if it
1977  * did not fit in with (return) path MTU, and KAME code supported the
1978  * behavior.  However, as a clarification after the RFC, this limitation
1979  * was removed in a revised version of the spec, RFC 2463.  We had kept the
1980  * old behavior, with a (non-default) ifdef block, while the new version of
1981  * the spec was an internet-draft status, and even after the new RFC was
1982  * published.  But it would rather make sense to clean the obsoleted part
1983  * up, and to make the code simpler at this stage.
1984  */
1985 void
1986 icmp6_reflect(m, off)
1987 	struct	mbuf *m;
1988 	size_t off;
1989 {
1990 	struct ip6_hdr *ip6;
1991 	struct icmp6_hdr *icmp6;
1992 	struct in6_ifaddr *ia;
1993 	struct in6_addr t, *src = 0;
1994 	int plen;
1995 	int type, code;
1996 	struct ifnet *outif = NULL;
1997 	struct sockaddr_in6 sa6_src, sa6_dst;
1998 
1999 	/* too short to reflect */
2000 	if (off < sizeof(struct ip6_hdr)) {
2001 		nd6log((LOG_DEBUG,
2002 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2003 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2004 		    __FILE__, __LINE__));
2005 		goto bad;
2006 	}
2007 
2008 	/*
2009 	 * If there are extra headers between IPv6 and ICMPv6, strip
2010 	 * off that header first.
2011 	 */
2012 #ifdef DIAGNOSTIC
2013 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2014 		panic("assumption failed in icmp6_reflect");
2015 #endif
2016 	if (off > sizeof(struct ip6_hdr)) {
2017 		size_t l;
2018 		struct ip6_hdr nip6;
2019 
2020 		l = off - sizeof(struct ip6_hdr);
2021 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2022 		m_adj(m, l);
2023 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2024 		if (m->m_len < l) {
2025 			if ((m = m_pullup(m, l)) == NULL)
2026 				return;
2027 		}
2028 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2029 	} else /* off == sizeof(struct ip6_hdr) */ {
2030 		size_t l;
2031 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2032 		if (m->m_len < l) {
2033 			if ((m = m_pullup(m, l)) == NULL)
2034 				return;
2035 		}
2036 	}
2037 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2038 	ip6 = mtod(m, struct ip6_hdr *);
2039 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2040 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2041 	type = icmp6->icmp6_type; /* keep type for statistics */
2042 	code = icmp6->icmp6_code; /* ditto. */
2043 
2044 	t = ip6->ip6_dst;
2045 	/*
2046 	 * ip6_input() drops a packet if its src is multicast.
2047 	 * So, the src is never multicast.
2048 	 */
2049 	ip6->ip6_dst = ip6->ip6_src;
2050 
2051 	/*
2052 	 * XXX: make sure to embed scope zone information, using
2053 	 * already embedded IDs or the received interface (if any).
2054 	 * Note that rcvif may be NULL.
2055 	 * TODO: scoped routing case (XXX).
2056 	 */
2057 	bzero(&sa6_src, sizeof(sa6_src));
2058 	sa6_src.sin6_family = AF_INET6;
2059 	sa6_src.sin6_len = sizeof(sa6_src);
2060 	sa6_src.sin6_addr = ip6->ip6_dst;
2061 	in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
2062 	in6_embedscope(&sa6_src.sin6_addr, &sa6_src, NULL, NULL);
2063 	ip6->ip6_dst = sa6_src.sin6_addr;
2064 
2065 	bzero(&sa6_dst, sizeof(sa6_dst));
2066 	sa6_dst.sin6_family = AF_INET6;
2067 	sa6_dst.sin6_len = sizeof(sa6_dst);
2068 	sa6_dst.sin6_addr = t;
2069 	in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
2070 	in6_embedscope(&t, &sa6_dst, NULL, NULL);
2071 
2072 	/*
2073 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2074 	 * use dst as the src for the reply.
2075 	 * The IN6_IFF_NOTREADY case would be VERY rare, but is possible
2076 	 * (for example) when we encounter an error while forwarding procedure
2077 	 * destined to a duplicated address of ours.
2078 	 */
2079 	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2080 		if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2081 		    (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2082 			src = &t;
2083 			break;
2084 		}
2085 	if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2086 		/*
2087 		 * This is the case if the dst is our link-local address
2088 		 * and the sender is also ourselves.
2089 		 */
2090 		src = &t;
2091 	}
2092 
2093 	if (src == 0) {
2094 		int e;
2095 		struct route_in6 ro;
2096 
2097 		/*
2098 		 * This case matches to multicasts, our anycast, or unicasts
2099 		 * that we do not own.  Select a source address based on the
2100 		 * source address of the erroneous packet.
2101 		 */
2102 		bzero(&ro, sizeof(ro));
2103 		src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e);
2104 		if (ro.ro_rt) { /* XXX: see comments in icmp6_mtudisc_update */
2105 			RTFREE(ro.ro_rt); /* XXX: we could use this */
2106 		}
2107 		if (src == NULL) {
2108 			nd6log((LOG_DEBUG,
2109 			    "icmp6_reflect: source can't be determined: "
2110 			    "dst=%s, error=%d\n",
2111 			    ip6_sprintf(&sa6_src.sin6_addr), e));
2112 			goto bad;
2113 		}
2114 	}
2115 
2116 	ip6->ip6_src = *src;
2117 
2118 	ip6->ip6_flow = 0;
2119 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2120 	ip6->ip6_vfc |= IPV6_VERSION;
2121 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2122 	if (m->m_pkthdr.rcvif) {
2123 		/* XXX: This may not be the outgoing interface */
2124 		ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2125 	} else
2126 		ip6->ip6_hlim = ip6_defhlim;
2127 
2128 	icmp6->icmp6_cksum = 0;
2129 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2130 					sizeof(struct ip6_hdr), plen);
2131 
2132 	/*
2133 	 * XXX option handling
2134 	 */
2135 
2136 	m->m_flags &= ~(M_BCAST|M_MCAST);
2137 
2138 	/*
2139 	 * To avoid a "too big" situation at an intermediate router
2140 	 * and the path MTU discovery process, specify the IPV6_MINMTU flag.
2141 	 * Note that only echo and node information replies are affected,
2142 	 * since the length of ICMP6 errors is limited to the minimum MTU.
2143 	 */
2144 	if (ip6_output(m, NULL, NULL, IPV6_MINMTU,
2145 		(struct ip6_moptions *)NULL, (struct socket *)NULL, &outif) != 0
2146 	    && outif)
2147 		icmp6_ifstat_inc(outif, ifs6_out_error);
2148 
2149 	if (outif)
2150 		icmp6_ifoutstat_inc(outif, type, code);
2151 
2152 	return;
2153 
2154  bad:
2155 	m_freem(m);
2156 	return;
2157 }
2158 
2159 void
2160 icmp6_fasttimo()
2161 {
2162 
2163 	mld6_fasttimeo();
2164 }
2165 
2166 static const char *
2167 icmp6_redirect_diag(src6, dst6, tgt6)
2168 	struct in6_addr *src6;
2169 	struct in6_addr *dst6;
2170 	struct in6_addr *tgt6;
2171 {
2172 	static char buf[1024];
2173 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2174 		ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2175 	return buf;
2176 }
2177 
2178 void
2179 icmp6_redirect_input(m, off)
2180 	struct mbuf *m;
2181 	int off;
2182 {
2183 	struct ifnet *ifp = m->m_pkthdr.rcvif;
2184 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2185 	struct nd_redirect *nd_rd;
2186 	int icmp6len = ntohs(ip6->ip6_plen);
2187 	char *lladdr = NULL;
2188 	int lladdrlen = 0;
2189 	struct rtentry *rt = NULL;
2190 	int is_router;
2191 	int is_onlink;
2192 	struct in6_addr src6 = ip6->ip6_src;
2193 	struct in6_addr redtgt6;
2194 	struct in6_addr reddst6;
2195 	union nd_opts ndopts;
2196 
2197 	if (!ifp)
2198 		return;
2199 
2200 	/* XXX if we are router, we don't update route by icmp6 redirect */
2201 	if (ip6_forwarding)
2202 		goto freeit;
2203 	if (!icmp6_rediraccept)
2204 		goto freeit;
2205 
2206 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2207 	if (nd_rd == NULL) {
2208 		icmp6stat.icp6s_tooshort++;
2209 		return;
2210 	}
2211 	redtgt6 = nd_rd->nd_rd_target;
2212 	reddst6 = nd_rd->nd_rd_dst;
2213 
2214 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2215 		redtgt6.s6_addr16[1] = htons(ifp->if_index);
2216 	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2217 		reddst6.s6_addr16[1] = htons(ifp->if_index);
2218 
2219 	/* validation */
2220 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2221 		nd6log((LOG_ERR,
2222 			"ICMP6 redirect sent from %s rejected; "
2223 			"must be from linklocal\n", ip6_sprintf(&src6)));
2224 		goto bad;
2225 	}
2226 	if (ip6->ip6_hlim != 255) {
2227 		nd6log((LOG_ERR,
2228 			"ICMP6 redirect sent from %s rejected; "
2229 			"hlim=%d (must be 255)\n",
2230 			ip6_sprintf(&src6), ip6->ip6_hlim));
2231 		goto bad;
2232 	}
2233     {
2234 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2235 	struct sockaddr_in6 sin6;
2236 	struct in6_addr *gw6;
2237 
2238 	bzero(&sin6, sizeof(sin6));
2239 	sin6.sin6_family = AF_INET6;
2240 	sin6.sin6_len = sizeof(struct sockaddr_in6);
2241 	bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2242 	rt = rtalloc1((struct sockaddr *)&sin6, 0);
2243 	if (rt) {
2244 		if (rt->rt_gateway == NULL ||
2245 		    rt->rt_gateway->sa_family != AF_INET6) {
2246 			nd6log((LOG_ERR,
2247 			    "ICMP6 redirect rejected; no route "
2248 			    "with inet6 gateway found for redirect dst: %s\n",
2249 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2250 			RTFREE(rt);
2251 			goto bad;
2252 		}
2253 
2254 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2255 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2256 			nd6log((LOG_ERR,
2257 				"ICMP6 redirect rejected; "
2258 				"not equal to gw-for-src=%s (must be same): "
2259 				"%s\n",
2260 				ip6_sprintf(gw6),
2261 				icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2262 			RTFREE(rt);
2263 			goto bad;
2264 		}
2265 	} else {
2266 		nd6log((LOG_ERR,
2267 			"ICMP6 redirect rejected; "
2268 			"no route found for redirect dst: %s\n",
2269 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2270 		goto bad;
2271 	}
2272 	RTFREE(rt);
2273 	rt = NULL;
2274     }
2275 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2276 		nd6log((LOG_ERR,
2277 			"ICMP6 redirect rejected; "
2278 			"redirect dst must be unicast: %s\n",
2279 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2280 		goto bad;
2281 	}
2282 
2283 	is_router = is_onlink = 0;
2284 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2285 		is_router = 1;	/* router case */
2286 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2287 		is_onlink = 1;	/* on-link destination case */
2288 	if (!is_router && !is_onlink) {
2289 		nd6log((LOG_ERR,
2290 			"ICMP6 redirect rejected; "
2291 			"neither router case nor onlink case: %s\n",
2292 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2293 		goto bad;
2294 	}
2295 	/* validation passed */
2296 
2297 	icmp6len -= sizeof(*nd_rd);
2298 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2299 	if (nd6_options(&ndopts) < 0) {
2300 		nd6log((LOG_INFO, "icmp6_redirect_input: "
2301 			"invalid ND option, rejected: %s\n",
2302 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2303 		/* nd6_options have incremented stats */
2304 		goto freeit;
2305 	}
2306 
2307 	if (ndopts.nd_opts_tgt_lladdr) {
2308 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2309 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2310 	}
2311 
2312 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2313 		nd6log((LOG_INFO,
2314 			"icmp6_redirect_input: lladdrlen mismatch for %s "
2315 			"(if %d, icmp6 packet %d): %s\n",
2316 			ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2317 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2318 		goto bad;
2319 	}
2320 
2321 	/* RFC 2461 8.3 */
2322 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2323 			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2324 
2325 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2326 		/* perform rtredirect */
2327 		struct sockaddr_in6 sdst;
2328 		struct sockaddr_in6 sgw;
2329 		struct sockaddr_in6 ssrc;
2330 		unsigned long rtcount;
2331 		struct rtentry *newrt = NULL;
2332 
2333 		/*
2334 		 * do not install redirect route, if the number of entries
2335 		 * is too much (> hiwat).  note that, the node (= host) will
2336 		 * work just fine even if we do not install redirect route
2337 		 * (there will be additional hops, though).
2338 		 */
2339 		rtcount = rt_timer_count(icmp6_redirect_timeout_q);
2340 		if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat)
2341 			return;
2342 		else if (0 <= icmp6_redirect_lowat &&
2343 		    rtcount > icmp6_redirect_lowat) {
2344 			/*
2345 			 * XXX nuke a victim, install the new one.
2346 			 */
2347 		}
2348 
2349 		bzero(&sdst, sizeof(sdst));
2350 		bzero(&sgw, sizeof(sgw));
2351 		bzero(&ssrc, sizeof(ssrc));
2352 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2353 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2354 			sizeof(struct sockaddr_in6);
2355 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2356 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2357 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2358 		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2359 			   (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2360 			   (struct sockaddr *)&ssrc,
2361 			   &newrt);
2362 
2363 		if (newrt) {
2364 			(void)rt_timer_add(newrt, icmp6_redirect_timeout,
2365 			    icmp6_redirect_timeout_q);
2366 			rtfree(newrt);
2367 		}
2368 	}
2369 	/* finally update cached route in each socket via pfctlinput */
2370 	{
2371 		struct sockaddr_in6 sdst;
2372 
2373 		bzero(&sdst, sizeof(sdst));
2374 		sdst.sin6_family = AF_INET6;
2375 		sdst.sin6_len = sizeof(struct sockaddr_in6);
2376 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2377 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2378 #ifdef IPSEC
2379 		key_sa_routechange((struct sockaddr *)&sdst);
2380 #endif
2381 	}
2382 
2383  freeit:
2384 	m_freem(m);
2385 	return;
2386 
2387  bad:
2388 	icmp6stat.icp6s_badredirect++;
2389 	m_freem(m);
2390 }
2391 
2392 void
2393 icmp6_redirect_output(m0, rt)
2394 	struct mbuf *m0;
2395 	struct rtentry *rt;
2396 {
2397 	struct ifnet *ifp;	/* my outgoing interface */
2398 	struct in6_addr *ifp_ll6;
2399 	struct in6_addr *nexthop;
2400 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2401 	struct mbuf *m = NULL;	/* newly allocated one */
2402 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2403 	struct nd_redirect *nd_rd;
2404 	size_t maxlen;
2405 	u_char *p;
2406 	struct sockaddr_in6 src_sa;
2407 
2408 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2409 
2410 	/* if we are not router, we don't send icmp6 redirect */
2411 	if (!ip6_forwarding)
2412 		goto fail;
2413 
2414 	/* sanity check */
2415 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2416 		goto fail;
2417 
2418 	/*
2419 	 * Address check:
2420 	 *  the source address must identify a neighbor, and
2421 	 *  the destination address must not be a multicast address
2422 	 *  [RFC 2461, sec 8.2]
2423 	 */
2424 	sip6 = mtod(m0, struct ip6_hdr *);
2425 	bzero(&src_sa, sizeof(src_sa));
2426 	src_sa.sin6_family = AF_INET6;
2427 	src_sa.sin6_len = sizeof(src_sa);
2428 	src_sa.sin6_addr = sip6->ip6_src;
2429 	/* we don't currently use sin6_scope_id, but eventually use it */
2430 	src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src);
2431 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2432 		goto fail;
2433 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2434 		goto fail;	/* what should we do here? */
2435 
2436 	/* rate limit */
2437 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2438 		goto fail;
2439 
2440 	/*
2441 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2442 	 * we almost always ask for an mbuf cluster for simplicity.
2443 	 * (MHLEN < IPV6_MMTU is almost always true)
2444 	 */
2445 #if IPV6_MMTU >= MCLBYTES
2446 # error assumption failed about IPV6_MMTU and MCLBYTES
2447 #endif
2448 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2449 	if (m && IPV6_MMTU >= MHLEN)
2450 		MCLGET(m, M_DONTWAIT);
2451 	if (!m)
2452 		goto fail;
2453 	m->m_pkthdr.rcvif = NULL;
2454 	m->m_len = 0;
2455 	maxlen = M_TRAILINGSPACE(m);
2456 	maxlen = min(IPV6_MMTU, maxlen);
2457 	/* just for safety */
2458 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2459 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2460 		goto fail;
2461 	}
2462 
2463 	{
2464 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2465 		struct in6_ifaddr *ia;
2466 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2467 						 IN6_IFF_NOTREADY|
2468 						 IN6_IFF_ANYCAST)) == NULL)
2469 			goto fail;
2470 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2471 	}
2472 
2473 	/* get ip6 linklocal address for the router. */
2474 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2475 		struct sockaddr_in6 *sin6;
2476 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2477 		nexthop = &sin6->sin6_addr;
2478 		if (!IN6_IS_ADDR_LINKLOCAL(nexthop))
2479 			nexthop = NULL;
2480 	} else
2481 		nexthop = NULL;
2482 
2483 	/* ip6 */
2484 	ip6 = mtod(m, struct ip6_hdr *);
2485 	ip6->ip6_flow = 0;
2486 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2487 	ip6->ip6_vfc |= IPV6_VERSION;
2488 	/* ip6->ip6_plen will be set later */
2489 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2490 	ip6->ip6_hlim = 255;
2491 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2492 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2493 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2494 
2495 	/* ND Redirect */
2496 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2497 	nd_rd->nd_rd_type = ND_REDIRECT;
2498 	nd_rd->nd_rd_code = 0;
2499 	nd_rd->nd_rd_reserved = 0;
2500 	if (rt->rt_flags & RTF_GATEWAY) {
2501 		/*
2502 		 * nd_rd->nd_rd_target must be a link-local address in
2503 		 * better router cases.
2504 		 */
2505 		if (!nexthop)
2506 			goto fail;
2507 		bcopy(nexthop, &nd_rd->nd_rd_target,
2508 		      sizeof(nd_rd->nd_rd_target));
2509 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2510 		      sizeof(nd_rd->nd_rd_dst));
2511 	} else {
2512 		/* make sure redtgt == reddst */
2513 		nexthop = &sip6->ip6_dst;
2514 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2515 		      sizeof(nd_rd->nd_rd_target));
2516 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2517 		      sizeof(nd_rd->nd_rd_dst));
2518 	}
2519 
2520 	p = (u_char *)(nd_rd + 1);
2521 
2522 	{
2523 		/* target lladdr option */
2524 		struct rtentry *rt_nexthop = NULL;
2525 		int len;
2526 		struct sockaddr_dl *sdl;
2527 		struct nd_opt_hdr *nd_opt;
2528 		char *lladdr;
2529 
2530 		rt_nexthop = nd6_lookup(nexthop, 0, ifp);
2531 		if (!rt_nexthop)
2532 			goto nolladdropt;
2533 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2534 		len = (len + 7) & ~7;	/* round by 8 */
2535 		/* safety check */
2536 		if (len + (p - (u_char *)ip6) > maxlen)
2537 			goto nolladdropt;
2538 		if (!(rt_nexthop->rt_flags & RTF_GATEWAY) &&
2539 		    (rt_nexthop->rt_flags & RTF_LLINFO) &&
2540 		    (rt_nexthop->rt_gateway->sa_family == AF_LINK) &&
2541 		    (sdl = (struct sockaddr_dl *)rt_nexthop->rt_gateway) &&
2542 		    sdl->sdl_alen) {
2543 			nd_opt = (struct nd_opt_hdr *)p;
2544 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2545 			nd_opt->nd_opt_len = len >> 3;
2546 			lladdr = (char *)(nd_opt + 1);
2547 			bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2548 			p += len;
2549 		}
2550 	}
2551   nolladdropt:;
2552 
2553 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2554 
2555 	/* just to be safe */
2556 	if (m0->m_flags & M_DECRYPTED)
2557 		goto noredhdropt;
2558 	if (p - (u_char *)ip6 > maxlen)
2559 		goto noredhdropt;
2560 
2561 	{
2562 		/* redirected header option */
2563 		int len;
2564 		struct nd_opt_rd_hdr *nd_opt_rh;
2565 
2566 		/*
2567 		 * compute the maximum size for icmp6 redirect header option.
2568 		 * XXX room for auth header?
2569 		 */
2570 		len = maxlen - (p - (u_char *)ip6);
2571 		len &= ~7;
2572 
2573 		/*
2574 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2575 		 * about padding/truncate rule for the original IP packet.
2576 		 * From the discussion on IPv6imp in Feb 1999,
2577 		 * the consensus was:
2578 		 * - "attach as much as possible" is the goal
2579 		 * - pad if not aligned (original size can be guessed by
2580 		 *   original ip6 header)
2581 		 * Following code adds the padding if it is simple enough,
2582 		 * and truncates if not.
2583 		 */
2584 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2585 			/* not enough room, truncate */
2586 			m_adj(m0, (len - sizeof(*nd_opt_rh)) -
2587 			    m0->m_pkthdr.len);
2588 		} else {
2589 			/*
2590                          * enough room, truncate if not aligned.
2591 			 * we don't pad here for simplicity.
2592 			 */
2593 			size_t extra;
2594 
2595 			extra = m0->m_pkthdr.len % 8;
2596 			if (extra) {
2597 				/* truncate */
2598 				m_adj(m0, -extra);
2599 			}
2600 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2601 		}
2602 
2603 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2604 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2605 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2606 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2607 		p += sizeof(*nd_opt_rh);
2608 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2609 
2610 		/* connect m0 to m */
2611 		m->m_pkthdr.len += m0->m_pkthdr.len;
2612 		m_cat(m, m0);
2613 		m0 = NULL;
2614 	}
2615 noredhdropt:
2616 	if (m0) {
2617 		m_freem(m0);
2618 		m0 = NULL;
2619 	}
2620 
2621 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
2622 		sip6->ip6_src.s6_addr16[1] = 0;
2623 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
2624 		sip6->ip6_dst.s6_addr16[1] = 0;
2625 #if 0
2626 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
2627 		ip6->ip6_src.s6_addr16[1] = 0;
2628 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
2629 		ip6->ip6_dst.s6_addr16[1] = 0;
2630 #endif
2631 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
2632 		nd_rd->nd_rd_target.s6_addr16[1] = 0;
2633 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
2634 		nd_rd->nd_rd_dst.s6_addr16[1] = 0;
2635 
2636 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2637 
2638 	nd_rd->nd_rd_cksum = 0;
2639 	nd_rd->nd_rd_cksum
2640 		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2641 
2642 	/* send the packet to outside... */
2643 	if (ip6_output(m, NULL, NULL, 0,
2644 		(struct ip6_moptions *)NULL, (struct socket *)NULL, NULL) != 0)
2645 		icmp6_ifstat_inc(ifp, ifs6_out_error);
2646 
2647 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
2648 	icmp6_ifstat_inc(ifp, ifs6_out_redirect);
2649 	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2650 
2651 	return;
2652 
2653 fail:
2654 	if (m)
2655 		m_freem(m);
2656 	if (m0)
2657 		m_freem(m0);
2658 }
2659 
2660 /*
2661  * ICMPv6 socket option processing.
2662  */
2663 int
2664 icmp6_ctloutput(op, so, level, optname, mp)
2665 	int op;
2666 	struct socket *so;
2667 	int level, optname;
2668 	struct mbuf **mp;
2669 {
2670 	int error = 0;
2671 	int optlen;
2672 	struct in6pcb *in6p = sotoin6pcb(so);
2673 	struct mbuf *m = *mp;
2674 
2675 	optlen = m ? m->m_len : 0;
2676 
2677 	if (level != IPPROTO_ICMPV6) {
2678 		if (op == PRCO_SETOPT && m)
2679 			(void)m_free(m);
2680 		return EINVAL;
2681 	}
2682 
2683 	switch (op) {
2684 	case PRCO_SETOPT:
2685 		switch (optname) {
2686 		case ICMP6_FILTER:
2687 		    {
2688 			struct icmp6_filter *p;
2689 
2690 			if (optlen != sizeof(*p)) {
2691 				error = EMSGSIZE;
2692 				break;
2693 			}
2694 			p = mtod(m, struct icmp6_filter *);
2695 			if (!p || !in6p->in6p_icmp6filt) {
2696 				error = EINVAL;
2697 				break;
2698 			}
2699 			bcopy(p, in6p->in6p_icmp6filt,
2700 				sizeof(struct icmp6_filter));
2701 			error = 0;
2702 			break;
2703 		    }
2704 
2705 		default:
2706 			error = ENOPROTOOPT;
2707 			break;
2708 		}
2709 		if (m)
2710 			(void)m_freem(m);
2711 		break;
2712 
2713 	case PRCO_GETOPT:
2714 		switch (optname) {
2715 		case ICMP6_FILTER:
2716 		    {
2717 			struct icmp6_filter *p;
2718 
2719 			if (!in6p->in6p_icmp6filt) {
2720 				error = EINVAL;
2721 				break;
2722 			}
2723 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
2724 			m->m_len = sizeof(struct icmp6_filter);
2725 			p = mtod(m, struct icmp6_filter *);
2726 			bcopy(in6p->in6p_icmp6filt, p,
2727 				sizeof(struct icmp6_filter));
2728 			error = 0;
2729 			break;
2730 		    }
2731 
2732 		default:
2733 			error = ENOPROTOOPT;
2734 			break;
2735 		}
2736 		break;
2737 	}
2738 
2739 	return (error);
2740 }
2741 
2742 /*
2743  * Perform rate limit check.
2744  * Returns 0 if it is okay to send the icmp6 packet.
2745  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2746  * limitation.
2747  *
2748  * XXX per-destination/type check necessary?
2749  */
2750 static int
2751 icmp6_ratelimit(dst, type, code)
2752 	const struct in6_addr *dst;	/* not used at this moment */
2753 	const int type;			/* not used at this moment */
2754 	const int code;			/* not used at this moment */
2755 {
2756 	int ret;
2757 
2758 	ret = 0;	/* okay to send */
2759 
2760 	/* PPS limit */
2761 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2762 	    icmp6errppslim)) {
2763 		/* The packet is subject to rate limit */
2764 		ret++;
2765 	}
2766 
2767 	return ret;
2768 }
2769 
2770 static struct rtentry *
2771 icmp6_mtudisc_clone(dst)
2772 	struct sockaddr *dst;
2773 {
2774 	struct rtentry *rt;
2775 	int    error;
2776 
2777 	rt = rtalloc1(dst, 1);
2778 	if (rt == 0)
2779 		return NULL;
2780 
2781 	/* If we didn't get a host route, allocate one */
2782 	if ((rt->rt_flags & RTF_HOST) == 0) {
2783 		struct rtentry *nrt;
2784 
2785 		error = rtrequest((int) RTM_ADD, dst,
2786 		    (struct sockaddr *) rt->rt_gateway,
2787 		    (struct sockaddr *) 0,
2788 		    RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt);
2789 		if (error) {
2790 			rtfree(rt);
2791 			return NULL;
2792 		}
2793 		nrt->rt_rmx = rt->rt_rmx;
2794 		rtfree(rt);
2795 		rt = nrt;
2796 	}
2797 	error = rt_timer_add(rt, icmp6_mtudisc_timeout,
2798 			icmp6_mtudisc_timeout_q);
2799 	if (error) {
2800 		rtfree(rt);
2801 		return NULL;
2802 	}
2803 
2804 	return rt;	/* caller need to call rtfree() */
2805 }
2806 
2807 static void
2808 icmp6_mtudisc_timeout(rt, r)
2809 	struct rtentry *rt;
2810 	struct rttimer *r;
2811 {
2812 	if (rt == NULL)
2813 		panic("icmp6_mtudisc_timeout: bad route to timeout");
2814 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
2815 	    (RTF_DYNAMIC | RTF_HOST)) {
2816 		rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt),
2817 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
2818 	} else {
2819 		if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
2820 			rt->rt_rmx.rmx_mtu = 0;
2821 	}
2822 }
2823 
2824 static void
2825 icmp6_redirect_timeout(rt, r)
2826 	struct rtentry *rt;
2827 	struct rttimer *r;
2828 {
2829 	if (rt == NULL)
2830 		panic("icmp6_redirect_timeout: bad route to timeout");
2831 	if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) ==
2832 	    (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) {
2833 		rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt),
2834 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
2835 	}
2836 }
2837 
2838 /*
2839  * sysctl helper routine for the net.inet6.icmp6.nd6 nodes.  silly?
2840  */
2841 static int
2842 sysctl_net_inet6_icmp6_nd6(SYSCTLFN_ARGS)
2843 {
2844 
2845 	if (namelen != 0)
2846 		return (EINVAL);
2847 
2848 	return (nd6_sysctl(rnode->sysctl_num, oldp, oldlenp,
2849 	    /*XXXUNCONST*/
2850 	    __UNCONST(newp), newlen));
2851 }
2852 
2853 SYSCTL_SETUP(sysctl_net_inet6_icmp6_setup,
2854 	     "sysctl net.inet6.icmp6 subtree setup")
2855 {
2856 
2857 	sysctl_createv(clog, 0, NULL, NULL,
2858 		       CTLFLAG_PERMANENT,
2859 		       CTLTYPE_NODE, "net", NULL,
2860 		       NULL, 0, NULL, 0,
2861 		       CTL_NET, CTL_EOL);
2862 	sysctl_createv(clog, 0, NULL, NULL,
2863 		       CTLFLAG_PERMANENT,
2864 		       CTLTYPE_NODE, "inet6", NULL,
2865 		       NULL, 0, NULL, 0,
2866 		       CTL_NET, PF_INET6, CTL_EOL);
2867 	sysctl_createv(clog, 0, NULL, NULL,
2868 		       CTLFLAG_PERMANENT,
2869 		       CTLTYPE_NODE, "icmp6",
2870 		       SYSCTL_DESCR("ICMPv6 related settings"),
2871 		       NULL, 0, NULL, 0,
2872 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6, CTL_EOL);
2873 
2874 	sysctl_createv(clog, 0, NULL, NULL,
2875 		       CTLFLAG_PERMANENT,
2876 		       CTLTYPE_STRUCT, "stats",
2877 		       SYSCTL_DESCR("ICMPv6 transmission statistics"),
2878 		       NULL, 0, &icmp6stat, sizeof(icmp6stat),
2879 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2880 		       ICMPV6CTL_STATS, CTL_EOL);
2881 	sysctl_createv(clog, 0, NULL, NULL,
2882 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2883 		       CTLTYPE_INT, "rediraccept",
2884 		       SYSCTL_DESCR("Accept and process redirect messages"),
2885 		       NULL, 0, &icmp6_rediraccept, 0,
2886 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2887 		       ICMPV6CTL_REDIRACCEPT, CTL_EOL);
2888 	sysctl_createv(clog, 0, NULL, NULL,
2889 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2890 		       CTLTYPE_INT, "redirtimeout",
2891 		       SYSCTL_DESCR("Redirect generated route lifetime"),
2892 		       NULL, 0, &icmp6_redirtimeout, 0,
2893 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2894 		       ICMPV6CTL_REDIRTIMEOUT, CTL_EOL);
2895 #if 0 /* obsoleted */
2896 	sysctl_createv(clog, 0, NULL, NULL,
2897 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2898 		       CTLTYPE_INT, "errratelimit", NULL,
2899 		       NULL, 0, &icmp6_errratelimit, 0,
2900 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2901 		       ICMPV6CTL_ERRRATELIMIT, CTL_EOL);
2902 #endif
2903 	sysctl_createv(clog, 0, NULL, NULL,
2904 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2905 		       CTLTYPE_INT, "nd6_prune",
2906 		       SYSCTL_DESCR("Neighbor discovery prune interval"),
2907 		       NULL, 0, &nd6_prune, 0,
2908 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2909 		       ICMPV6CTL_ND6_PRUNE, CTL_EOL);
2910 	sysctl_createv(clog, 0, NULL, NULL,
2911 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2912 		       CTLTYPE_INT, "nd6_delay",
2913 		       SYSCTL_DESCR("First probe delay time"),
2914 		       NULL, 0, &nd6_delay, 0,
2915 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2916 		       ICMPV6CTL_ND6_DELAY, CTL_EOL);
2917 	sysctl_createv(clog, 0, NULL, NULL,
2918 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2919 		       CTLTYPE_INT, "nd6_umaxtries",
2920 		       SYSCTL_DESCR("Number of unicast discovery attempts"),
2921 		       NULL, 0, &nd6_umaxtries, 0,
2922 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2923 		       ICMPV6CTL_ND6_UMAXTRIES, CTL_EOL);
2924 	sysctl_createv(clog, 0, NULL, NULL,
2925 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2926 		       CTLTYPE_INT, "nd6_mmaxtries",
2927 		       SYSCTL_DESCR("Number of multicast discovery attempts"),
2928 		       NULL, 0, &nd6_mmaxtries, 0,
2929 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2930 		       ICMPV6CTL_ND6_MMAXTRIES, CTL_EOL);
2931 	sysctl_createv(clog, 0, NULL, NULL,
2932 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2933 		       CTLTYPE_INT, "nd6_useloopback",
2934 		       SYSCTL_DESCR("Use loopback interface for local traffic"),
2935 		       NULL, 0, &nd6_useloopback, 0,
2936 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2937 		       ICMPV6CTL_ND6_USELOOPBACK, CTL_EOL);
2938 #if 0 /* obsoleted */
2939 	sysctl_createv(clog, 0, NULL, NULL,
2940 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2941 		       CTLTYPE_INT, "nd6_proxyall", NULL,
2942 		       NULL, 0, &nd6_proxyall, 0,
2943 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2944 		       ICMPV6CTL_ND6_PROXYALL, CTL_EOL);
2945 #endif
2946 	sysctl_createv(clog, 0, NULL, NULL,
2947 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2948 		       CTLTYPE_INT, "nodeinfo",
2949 		       SYSCTL_DESCR("Respond to node information requests"),
2950 		       NULL, 0, &icmp6_nodeinfo, 0,
2951 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2952 		       ICMPV6CTL_NODEINFO, CTL_EOL);
2953 	sysctl_createv(clog, 0, NULL, NULL,
2954 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2955 		       CTLTYPE_INT, "errppslimit",
2956 		       SYSCTL_DESCR("Maximum ICMP errors sent per second"),
2957 		       NULL, 0, &icmp6errppslim, 0,
2958 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2959 		       ICMPV6CTL_ERRPPSLIMIT, CTL_EOL);
2960 	sysctl_createv(clog, 0, NULL, NULL,
2961 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2962 		       CTLTYPE_INT, "nd6_maxnudhint",
2963 		       SYSCTL_DESCR("Maximum neighbor unreachable hint count"),
2964 		       NULL, 0, &nd6_maxnudhint, 0,
2965 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2966 		       ICMPV6CTL_ND6_MAXNUDHINT, CTL_EOL);
2967 	sysctl_createv(clog, 0, NULL, NULL,
2968 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2969 		       CTLTYPE_INT, "mtudisc_hiwat",
2970 		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
2971 		       NULL, 0, &icmp6_mtudisc_hiwat, 0,
2972 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2973 		       ICMPV6CTL_MTUDISC_HIWAT, CTL_EOL);
2974 	sysctl_createv(clog, 0, NULL, NULL,
2975 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2976 		       CTLTYPE_INT, "mtudisc_lowat",
2977 		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
2978 		       NULL, 0, &icmp6_mtudisc_lowat, 0,
2979 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2980 		       ICMPV6CTL_MTUDISC_LOWAT, CTL_EOL);
2981 	sysctl_createv(clog, 0, NULL, NULL,
2982 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2983 		       CTLTYPE_INT, "nd6_debug",
2984 		       SYSCTL_DESCR("Enable neighbor discovery debug output"),
2985 		       NULL, 0, &nd6_debug, 0,
2986 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2987 		       ICMPV6CTL_ND6_DEBUG, CTL_EOL);
2988 	sysctl_createv(clog, 0, NULL, NULL,
2989 		       CTLFLAG_PERMANENT,
2990 		       CTLTYPE_STRUCT, "nd6_drlist",
2991 		       SYSCTL_DESCR("Default router list"),
2992 		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
2993 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2994 		       ICMPV6CTL_ND6_DRLIST, CTL_EOL);
2995 	sysctl_createv(clog, 0, NULL, NULL,
2996 		       CTLFLAG_PERMANENT,
2997 		       CTLTYPE_STRUCT, "nd6_prlist",
2998 		       SYSCTL_DESCR("Prefix list"),
2999 		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
3000 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3001 		       ICMPV6CTL_ND6_PRLIST, CTL_EOL);
3002 }
3003