xref: /netbsd-src/sys/netinet6/icmp6.c (revision 23c8222edbfb0f0932d88a8351d3a0cf817dfb9e)
1 /*	$NetBSD: icmp6.c,v 1.107 2004/05/25 04:34:00 atatat 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 1.107 2004/05/25 04:34:00 atatat 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 		if (code != 0)
518 			goto badcode;
519 
520 		code = PRC_MSGSIZE;
521 
522 		/*
523 		 * Updating the path MTU will be done after examining
524 		 * intermediate extension headers.
525 		 */
526 		goto deliver;
527 
528 	case ICMP6_TIME_EXCEEDED:
529 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
530 		switch (code) {
531 		case ICMP6_TIME_EXCEED_TRANSIT:
532 			code = PRC_TIMXCEED_INTRANS;
533 			break;
534 		case ICMP6_TIME_EXCEED_REASSEMBLY:
535 			code = PRC_TIMXCEED_REASS;
536 			break;
537 		default:
538 			goto badcode;
539 		}
540 		goto deliver;
541 
542 	case ICMP6_PARAM_PROB:
543 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
544 		switch (code) {
545 		case ICMP6_PARAMPROB_NEXTHEADER:
546 			code = PRC_UNREACH_PROTOCOL;
547 			break;
548 		case ICMP6_PARAMPROB_HEADER:
549 		case ICMP6_PARAMPROB_OPTION:
550 			code = PRC_PARAMPROB;
551 			break;
552 		default:
553 			goto badcode;
554 		}
555 		goto deliver;
556 
557 	case ICMP6_ECHO_REQUEST:
558 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
559 		if (code != 0)
560 			goto badcode;
561 		/*
562 		 * Copy mbuf to send to two data paths: userland socket(s),
563 		 * and to the querier (echo reply).
564 		 * m: a copy for socket, n: a copy for querier
565 		 */
566 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
567 			/* Give up local */
568 			n = m;
569 			m = NULL;
570 			goto deliverecho;
571 		}
572 		/*
573 		 * If the first mbuf is shared, or the first mbuf is too short,
574 		 * copy the first part of the data into a fresh mbuf.
575 		 * Otherwise, we will wrongly overwrite both copies.
576 		 */
577 		if ((n->m_flags & M_EXT) != 0 ||
578 		    n->m_len < off + sizeof(struct icmp6_hdr)) {
579 			struct mbuf *n0 = n;
580 			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
581 
582 			/*
583 			 * Prepare an internal mbuf.  m_pullup() doesn't
584 			 * always copy the length we specified.
585 			 */
586 			if (maxlen >= MCLBYTES) {
587 				/* Give up remote */
588 				m_freem(n0);
589 				break;
590 			}
591 			MGETHDR(n, M_DONTWAIT, n0->m_type);
592 			if (n && maxlen >= MHLEN) {
593 				MCLGET(n, M_DONTWAIT);
594 				if ((n->m_flags & M_EXT) == 0) {
595 					m_free(n);
596 					n = NULL;
597 				}
598 			}
599 			if (n == NULL) {
600 				/* Give up local */
601 				m_freem(n0);
602 				n = m;
603 				m = NULL;
604 				goto deliverecho;
605 			}
606 			M_COPY_PKTHDR(n, n0);
607 			/*
608 			 * Copy IPv6 and ICMPv6 only.
609 			 */
610 			nip6 = mtod(n, struct ip6_hdr *);
611 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
612 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
613 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
614 			noff = sizeof(struct ip6_hdr);
615 			n->m_len = noff + sizeof(struct icmp6_hdr);
616 			/*
617 			 * Adjust mbuf.  ip6_plen will be adjusted in
618 			 * ip6_output().
619 			 * n->m_pkthdr.len == n0->m_pkthdr.len at this point.
620 			 */
621 			n->m_pkthdr.len += noff + sizeof(struct icmp6_hdr);
622 			n->m_pkthdr.len -= (off + sizeof(struct icmp6_hdr));
623 			m_adj(n0, off + sizeof(struct icmp6_hdr));
624 			n->m_next = n0;
625 			n0->m_flags &= ~M_PKTHDR;
626 		} else {
627 	 deliverecho:
628 			nip6 = mtod(n, struct ip6_hdr *);
629 			nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
630 			noff = off;
631 		}
632 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
633 		nicmp6->icmp6_code = 0;
634 		if (n) {
635 			icmp6stat.icp6s_reflect++;
636 			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
637 			icmp6_reflect(n, noff);
638 		}
639 		if (!m)
640 			goto freeit;
641 		break;
642 
643 	case ICMP6_ECHO_REPLY:
644 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
645 		if (code != 0)
646 			goto badcode;
647 		break;
648 
649 	case MLD_LISTENER_QUERY:
650 	case MLD_LISTENER_REPORT:
651 		if (icmp6len < sizeof(struct mld_hdr))
652 			goto badlen;
653 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
654 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
655 		else
656 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
657 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
658 			/* give up local */
659 			mld6_input(m, off);
660 			m = NULL;
661 			goto freeit;
662 		}
663 		mld6_input(n, off);
664 		/* m stays. */
665 		break;
666 
667 	case MLD_LISTENER_DONE:
668 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
669 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
670 			goto badlen;
671 		break;		/* nothing to be done in kernel */
672 
673 	case MLD_MTRACE_RESP:
674 	case MLD_MTRACE:
675 		/* XXX: these two are experimental.  not officially defined. */
676 		/* XXX: per-interface statistics? */
677 		break;		/* just pass it to applications */
678 
679 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
680 	    {
681 		enum { WRU, FQDN } mode;
682 
683 		if (!icmp6_nodeinfo)
684 			break;
685 
686 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
687 			mode = WRU;
688 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
689 			mode = FQDN;
690 		else
691 			goto badlen;
692 
693 		if (mode == FQDN) {
694 			n = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
695 			if (n)
696 				n = ni6_input(n, off);
697 			/* XXX meaningless if n == NULL */
698 			noff = sizeof(struct ip6_hdr);
699 		} else {
700 			u_char *p;
701 			int maxlen, maxhlen;
702 
703 			if ((icmp6_nodeinfo & 5) != 5)
704 				break;
705 
706 			if (code != 0)
707 				goto badcode;
708 			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
709 			if (maxlen >= MCLBYTES) {
710 				/* Give up remote */
711 				break;
712 			}
713 			MGETHDR(n, M_DONTWAIT, m->m_type);
714 			if (n && maxlen > MHLEN) {
715 				MCLGET(n, M_DONTWAIT);
716 				if ((n->m_flags & M_EXT) == 0) {
717 					m_free(n);
718 					n = NULL;
719 				}
720 			}
721 			if (n == NULL) {
722 				/* Give up remote */
723 				break;
724 			}
725 			n->m_pkthdr.rcvif = NULL;
726 			n->m_len = 0;
727 			maxhlen = M_TRAILINGSPACE(n) - maxlen;
728 			if (maxhlen > hostnamelen)
729 				maxhlen = hostnamelen;
730 			/*
731 			 * Copy IPv6 and ICMPv6 only.
732 			 */
733 			nip6 = mtod(n, struct ip6_hdr *);
734 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
735 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
736 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
737 			p = (u_char *)(nicmp6 + 1);
738 			bzero(p, 4);
739 			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
740 			noff = sizeof(struct ip6_hdr);
741 			M_COPY_PKTHDR(n, m); /* just for rcvif */
742 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
743 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
744 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
745 			nicmp6->icmp6_code = 0;
746 		}
747 #undef hostnamelen
748 		if (n) {
749 			icmp6stat.icp6s_reflect++;
750 			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
751 			icmp6_reflect(n, noff);
752 		}
753 		break;
754 	    }
755 
756 	case ICMP6_WRUREPLY:
757 		if (code != 0)
758 			goto badcode;
759 		break;
760 
761 	case ND_ROUTER_SOLICIT:
762 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
763 		if (code != 0)
764 			goto badcode;
765 		if (icmp6len < sizeof(struct nd_router_solicit))
766 			goto badlen;
767 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
768 			/* give up local */
769 			nd6_rs_input(m, off, icmp6len);
770 			m = NULL;
771 			goto freeit;
772 		}
773 		nd6_rs_input(n, off, icmp6len);
774 		/* m stays. */
775 		break;
776 
777 	case ND_ROUTER_ADVERT:
778 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
779 		if (code != 0)
780 			goto badcode;
781 		if (icmp6len < sizeof(struct nd_router_advert))
782 			goto badlen;
783 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
784 			/* give up local */
785 			nd6_ra_input(m, off, icmp6len);
786 			m = NULL;
787 			goto freeit;
788 		}
789 		nd6_ra_input(n, off, icmp6len);
790 		/* m stays. */
791 		break;
792 
793 	case ND_NEIGHBOR_SOLICIT:
794 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
795 		if (code != 0)
796 			goto badcode;
797 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
798 			goto badlen;
799 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
800 			/* give up local */
801 			nd6_ns_input(m, off, icmp6len);
802 			m = NULL;
803 			goto freeit;
804 		}
805 		nd6_ns_input(n, off, icmp6len);
806 		/* m stays. */
807 		break;
808 
809 	case ND_NEIGHBOR_ADVERT:
810 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
811 		if (code != 0)
812 			goto badcode;
813 		if (icmp6len < sizeof(struct nd_neighbor_advert))
814 			goto badlen;
815 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
816 			/* give up local */
817 			nd6_na_input(m, off, icmp6len);
818 			m = NULL;
819 			goto freeit;
820 		}
821 		nd6_na_input(n, off, icmp6len);
822 		/* m stays. */
823 		break;
824 
825 	case ND_REDIRECT:
826 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
827 		if (code != 0)
828 			goto badcode;
829 		if (icmp6len < sizeof(struct nd_redirect))
830 			goto badlen;
831 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
832 			/* give up local */
833 			icmp6_redirect_input(m, off);
834 			m = NULL;
835 			goto freeit;
836 		}
837 		icmp6_redirect_input(n, off);
838 		/* m stays. */
839 		break;
840 
841 	case ICMP6_ROUTER_RENUMBERING:
842 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
843 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
844 			goto badcode;
845 		if (icmp6len < sizeof(struct icmp6_router_renum))
846 			goto badlen;
847 		break;
848 
849 	default:
850 		nd6log((LOG_DEBUG,
851 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
852 		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
853 		    ip6_sprintf(&ip6->ip6_dst),
854 		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
855 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
856 			/* ICMPv6 error: MUST deliver it by spec... */
857 			code = PRC_NCMDS;
858 			/* deliver */
859 		} else {
860 			/* ICMPv6 informational: MUST not deliver */
861 			break;
862 		}
863 	deliver:
864 		if (icmp6_notify_error(m, off, icmp6len, code)) {
865 			/* In this case, m should've been freed. */
866 			return (IPPROTO_DONE);
867 		}
868 		break;
869 
870 	badcode:
871 		icmp6stat.icp6s_badcode++;
872 		break;
873 
874 	badlen:
875 		icmp6stat.icp6s_badlen++;
876 		break;
877 	}
878 
879 	/* deliver the packet to appropriate sockets */
880 	icmp6_rip6_input(&m, *offp);
881 
882 	return IPPROTO_DONE;
883 
884  freeit:
885 	m_freem(m);
886 	return IPPROTO_DONE;
887 }
888 
889 static int
890 icmp6_notify_error(m, off, icmp6len, code)
891 	struct mbuf *m;
892 	int off, icmp6len;
893 {
894 	struct icmp6_hdr *icmp6;
895 	struct ip6_hdr *eip6;
896 	u_int32_t notifymtu;
897 	struct sockaddr_in6 icmp6src, icmp6dst;
898 
899 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
900 		icmp6stat.icp6s_tooshort++;
901 		goto freeit;
902 	}
903 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
904 		       sizeof(*icmp6) + sizeof(struct ip6_hdr));
905 	if (icmp6 == NULL) {
906 		icmp6stat.icp6s_tooshort++;
907 		return (-1);
908 	}
909 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
910 
911 	/* Detect the upper level protocol */
912 	{
913 		void (*ctlfunc) __P((int, struct sockaddr *, void *));
914 		u_int8_t nxt = eip6->ip6_nxt;
915 		int eoff = off + sizeof(struct icmp6_hdr) +
916 			sizeof(struct ip6_hdr);
917 		struct ip6ctlparam ip6cp;
918 		struct in6_addr *finaldst = NULL;
919 		int icmp6type = icmp6->icmp6_type;
920 		struct ip6_frag *fh;
921 		struct ip6_rthdr *rth;
922 		struct ip6_rthdr0 *rth0;
923 		int rthlen;
924 
925 		while (1) { /* XXX: should avoid infinite loop explicitly? */
926 			struct ip6_ext *eh;
927 
928 			switch (nxt) {
929 			case IPPROTO_HOPOPTS:
930 			case IPPROTO_DSTOPTS:
931 			case IPPROTO_AH:
932 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
933 					       eoff, sizeof(*eh));
934 				if (eh == NULL) {
935 					icmp6stat.icp6s_tooshort++;
936 					return (-1);
937 				}
938 
939 				if (nxt == IPPROTO_AH)
940 					eoff += (eh->ip6e_len + 2) << 2;
941 				else
942 					eoff += (eh->ip6e_len + 1) << 3;
943 				nxt = eh->ip6e_nxt;
944 				break;
945 			case IPPROTO_ROUTING:
946 				/*
947 				 * When the erroneous packet contains a
948 				 * routing header, we should examine the
949 				 * header to determine the final destination.
950 				 * Otherwise, we can't properly update
951 				 * information that depends on the final
952 				 * destination (e.g. path MTU).
953 				 */
954 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
955 					       eoff, sizeof(*rth));
956 				if (rth == NULL) {
957 					icmp6stat.icp6s_tooshort++;
958 					return (-1);
959 				}
960 				rthlen = (rth->ip6r_len + 1) << 3;
961 				/*
962 				 * XXX: currently there is no
963 				 * officially defined type other
964 				 * than type-0.
965 				 * Note that if the segment left field
966 				 * is 0, all intermediate hops must
967 				 * have been passed.
968 				 */
969 				if (rth->ip6r_segleft &&
970 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
971 					int hops;
972 
973 					IP6_EXTHDR_GET(rth0,
974 						       struct ip6_rthdr0 *, m,
975 						       eoff, rthlen);
976 					if (rth0 == NULL) {
977 						icmp6stat.icp6s_tooshort++;
978 						return (-1);
979 					}
980 					/* just ignore a bogus header */
981 					if ((rth0->ip6r0_len % 2) == 0 &&
982 					    (hops = rth0->ip6r0_len/2))
983 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
984 				}
985 				eoff += rthlen;
986 				nxt = rth->ip6r_nxt;
987 				break;
988 			case IPPROTO_FRAGMENT:
989 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
990 					       eoff, sizeof(*fh));
991 				if (fh == NULL) {
992 					icmp6stat.icp6s_tooshort++;
993 					return (-1);
994 				}
995 				/*
996 				 * Data after a fragment header is meaningless
997 				 * unless it is the first fragment, but
998 				 * we'll go to the notify label for path MTU
999 				 * discovery.
1000 				 */
1001 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1002 					goto notify;
1003 
1004 				eoff += sizeof(struct ip6_frag);
1005 				nxt = fh->ip6f_nxt;
1006 				break;
1007 			default:
1008 				/*
1009 				 * This case includes ESP and the No Next
1010 				 * Header.  In such cases going to the notify
1011 				 * label does not have any meaning
1012 				 * (i.e. ctlfunc will be NULL), but we go
1013 				 * anyway since we might have to update
1014 				 * path MTU information.
1015 				 */
1016 				goto notify;
1017 			}
1018 		}
1019 	  notify:
1020 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1021 			       sizeof(*icmp6) + sizeof(struct ip6_hdr));
1022 		if (icmp6 == NULL) {
1023 			icmp6stat.icp6s_tooshort++;
1024 			return (-1);
1025 		}
1026 
1027 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1028 		bzero(&icmp6dst, sizeof(icmp6dst));
1029 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1030 		icmp6dst.sin6_family = AF_INET6;
1031 		if (finaldst == NULL)
1032 			icmp6dst.sin6_addr = eip6->ip6_dst;
1033 		else
1034 			icmp6dst.sin6_addr = *finaldst;
1035 		icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1036 							  &icmp6dst.sin6_addr);
1037 #ifndef SCOPEDROUTING
1038 		if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1039 				   NULL, NULL)) {
1040 			/* should be impossbile */
1041 			nd6log((LOG_DEBUG,
1042 			    "icmp6_notify_error: in6_embedscope failed\n"));
1043 			goto freeit;
1044 		}
1045 #endif
1046 
1047 		/*
1048 		 * retrieve parameters from the inner IPv6 header, and convert
1049 		 * them into sockaddr structures.
1050 		 */
1051 		bzero(&icmp6src, sizeof(icmp6src));
1052 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1053 		icmp6src.sin6_family = AF_INET6;
1054 		icmp6src.sin6_addr = eip6->ip6_src;
1055 		icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1056 							  &icmp6src.sin6_addr);
1057 #ifndef SCOPEDROUTING
1058 		if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1059 				   NULL, NULL)) {
1060 			/* should be impossbile */
1061 			nd6log((LOG_DEBUG,
1062 			    "icmp6_notify_error: in6_embedscope failed\n"));
1063 			goto freeit;
1064 		}
1065 #endif
1066 		icmp6src.sin6_flowinfo =
1067 			(eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1068 
1069 		if (finaldst == NULL)
1070 			finaldst = &eip6->ip6_dst;
1071 		ip6cp.ip6c_m = m;
1072 		ip6cp.ip6c_icmp6 = icmp6;
1073 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1074 		ip6cp.ip6c_off = eoff;
1075 		ip6cp.ip6c_finaldst = finaldst;
1076 		ip6cp.ip6c_src = &icmp6src;
1077 		ip6cp.ip6c_nxt = nxt;
1078 
1079 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1080 			notifymtu = ntohl(icmp6->icmp6_mtu);
1081 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1082 		}
1083 
1084 		ctlfunc = (void (*) __P((int, struct sockaddr *, void *)))
1085 			(inet6sw[ip6_protox[nxt]].pr_ctlinput);
1086 		if (ctlfunc) {
1087 			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1088 					  &ip6cp);
1089 		}
1090 	}
1091 	return (0);
1092 
1093   freeit:
1094 	m_freem(m);
1095 	return (-1);
1096 }
1097 
1098 void
1099 icmp6_mtudisc_update(ip6cp, validated)
1100 	struct ip6ctlparam *ip6cp;
1101 	int validated;
1102 {
1103 	unsigned long rtcount;
1104 	struct icmp6_mtudisc_callback *mc;
1105 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1106 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1107 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1108 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1109 	struct rtentry *rt = NULL;
1110 	struct sockaddr_in6 sin6;
1111 
1112 	/*
1113 	 * allow non-validated cases if memory is plenty, to make traffic
1114 	 * from non-connected pcb happy.
1115 	 */
1116 	rtcount = rt_timer_count(icmp6_mtudisc_timeout_q);
1117 	if (validated) {
1118 		if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat)
1119 			return;
1120 		else if (0 <= icmp6_mtudisc_lowat &&
1121 		    rtcount > icmp6_mtudisc_lowat) {
1122 			/*
1123 			 * XXX nuke a victim, install the new one.
1124 			 */
1125 		}
1126 	} else {
1127 		if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat)
1128 			return;
1129 	}
1130 
1131 	bzero(&sin6, sizeof(sin6));
1132 	sin6.sin6_family = PF_INET6;
1133 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1134 	sin6.sin6_addr = *dst;
1135 	/* XXX normally, this won't happen */
1136 	if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1137 		sin6.sin6_addr.s6_addr16[1] =
1138 		    htons(m->m_pkthdr.rcvif->if_index);
1139 	}
1140 	/* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1141 	rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6);
1142 
1143 	if (rt && (rt->rt_flags & RTF_HOST) &&
1144 	    !(rt->rt_rmx.rmx_locks & RTV_MTU) &&
1145 	    (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) {
1146 		if (mtu < IN6_LINKMTU(rt->rt_ifp)) {
1147 			icmp6stat.icp6s_pmtuchg++;
1148 			rt->rt_rmx.rmx_mtu = mtu;
1149 		}
1150 	}
1151 	if (rt) { /* XXX: need braces to avoid conflict with else in RTFREE. */
1152 		RTFREE(rt);
1153 	}
1154 
1155 	/*
1156 	 * Notify protocols that the MTU for this destination
1157 	 * has changed.
1158 	 */
1159 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
1160 	     mc = LIST_NEXT(mc, mc_list))
1161 		(*mc->mc_func)(&sin6.sin6_addr);
1162 }
1163 
1164 /*
1165  * Process a Node Information Query packet, based on
1166  * draft-ietf-ipngwg-icmp-name-lookups-07.
1167  *
1168  * Spec incompatibilities:
1169  * - IPv6 Subject address handling
1170  * - IPv4 Subject address handling support missing
1171  * - Proxy reply (answer even if it's not for me)
1172  * - joins NI group address at in6_ifattach() time only, does not cope
1173  *   with hostname changes by sethostname(3)
1174  */
1175 #ifndef offsetof		/* XXX */
1176 #define	offsetof(type, member)	((size_t)(&((type *)0)->member))
1177 #endif
1178 static struct mbuf *
1179 ni6_input(m, off)
1180 	struct mbuf *m;
1181 	int off;
1182 {
1183 	struct icmp6_nodeinfo *ni6, *nni6;
1184 	struct mbuf *n = NULL;
1185 	u_int16_t qtype;
1186 	int subjlen;
1187 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1188 	struct ni_reply_fqdn *fqdn;
1189 	int addrs;		/* for NI_QTYPE_NODEADDR */
1190 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1191 	struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1192 	struct ip6_hdr *ip6;
1193 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1194 	char *subj = NULL;
1195 
1196 	ip6 = mtod(m, struct ip6_hdr *);
1197 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1198 	if (ni6 == NULL) {
1199 		/* m is already reclaimed */
1200 		return NULL;
1201 	}
1202 
1203 	/*
1204 	 * Validate IPv6 destination address.
1205 	 *
1206 	 * The Responder must discard the Query without further processing
1207 	 * unless it is one of the Responder's unicast or anycast addresses, or
1208 	 * a link-local scope multicast address which the Responder has joined.
1209 	 * [icmp-name-lookups-07, Section 4.]
1210 	 */
1211 	bzero(&sin6, sizeof(sin6));
1212 	sin6.sin6_family = AF_INET6;
1213 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1214 	bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1215 	/* XXX scopeid */
1216 	if (ifa_ifwithaddr((struct sockaddr *)&sin6))
1217 		; /* unicast/anycast, fine */
1218 	else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1219 		; /* link-local multicast, fine */
1220 	else
1221 		goto bad;
1222 
1223 	/* validate query Subject field. */
1224 	qtype = ntohs(ni6->ni_qtype);
1225 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1226 	switch (qtype) {
1227 	case NI_QTYPE_NOOP:
1228 	case NI_QTYPE_SUPTYPES:
1229 		/* 07 draft */
1230 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1231 			break;
1232 		/* FALLTHROUGH */
1233 	case NI_QTYPE_FQDN:
1234 	case NI_QTYPE_NODEADDR:
1235 		switch (ni6->ni_code) {
1236 		case ICMP6_NI_SUBJ_IPV6:
1237 #if ICMP6_NI_SUBJ_IPV6 != 0
1238 		case 0:
1239 #endif
1240 			/*
1241 			 * backward compatibility - try to accept 03 draft
1242 			 * format, where no Subject is present.
1243 			 */
1244 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1245 			    subjlen == 0) {
1246 				oldfqdn++;
1247 				break;
1248 			}
1249 #if ICMP6_NI_SUBJ_IPV6 != 0
1250 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1251 				goto bad;
1252 #endif
1253 
1254 			if (subjlen != sizeof(sin6.sin6_addr))
1255 				goto bad;
1256 
1257 			/*
1258 			 * Validate Subject address.
1259 			 *
1260 			 * Not sure what exactly "address belongs to the node"
1261 			 * means in the spec, is it just unicast, or what?
1262 			 *
1263 			 * At this moment we consider Subject address as
1264 			 * "belong to the node" if the Subject address equals
1265 			 * to the IPv6 destination address; validation for
1266 			 * IPv6 destination address should have done enough
1267 			 * check for us.
1268 			 *
1269 			 * We do not do proxy at this moment.
1270 			 */
1271 			/* m_pulldown instead of copy? */
1272 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1273 			    subjlen, (caddr_t)&sin6.sin6_addr);
1274 			/* XXX kame scope hack */
1275 			if (IN6_IS_SCOPE_LINKLOCAL(&sin6.sin6_addr)) {
1276 				if ((m->m_flags & M_PKTHDR) != 0 &&
1277 				    m->m_pkthdr.rcvif) {
1278 					sin6.sin6_addr.s6_addr16[1] =
1279 					    htons(m->m_pkthdr.rcvif->if_index);
1280 				}
1281 			}
1282 			subj = (char *)&sin6;
1283 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &sin6.sin6_addr))
1284 				break;
1285 
1286 			/*
1287 			 * XXX if we are to allow other cases, we should really
1288 			 * be careful about scope here.
1289 			 * basically, we should disallow queries toward IPv6
1290 			 * destination X with subject Y, if scope(X) > scope(Y).
1291 			 * if we allow scope(X) > scope(Y), it will result in
1292 			 * information leakage across scope boundary.
1293 			 */
1294 			goto bad;
1295 
1296 		case ICMP6_NI_SUBJ_FQDN:
1297 			/*
1298 			 * Validate Subject name with gethostname(3).
1299 			 *
1300 			 * The behavior may need some debate, since:
1301 			 * - we are not sure if the node has FQDN as
1302 			 *   hostname (returned by gethostname(3)).
1303 			 * - the code does wildcard match for truncated names.
1304 			 *   however, we are not sure if we want to perform
1305 			 *   wildcard match, if gethostname(3) side has
1306 			 *   truncated hostname.
1307 			 */
1308 			n = ni6_nametodns(hostname, hostnamelen, 0);
1309 			if (!n || n->m_next || n->m_len == 0)
1310 				goto bad;
1311 			IP6_EXTHDR_GET(subj, char *, m,
1312 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1313 			if (subj == NULL)
1314 				goto bad;
1315 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1316 					n->m_len)) {
1317 				goto bad;
1318 			}
1319 			m_freem(n);
1320 			n = NULL;
1321 			break;
1322 
1323 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1324 		default:
1325 			goto bad;
1326 		}
1327 		break;
1328 	}
1329 
1330 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1331 	switch (qtype) {
1332 	case NI_QTYPE_FQDN:
1333 		if ((icmp6_nodeinfo & 1) == 0)
1334 			goto bad;
1335 		break;
1336 	case NI_QTYPE_NODEADDR:
1337 		if ((icmp6_nodeinfo & 2) == 0)
1338 			goto bad;
1339 		break;
1340 	}
1341 
1342 	/* guess reply length */
1343 	switch (qtype) {
1344 	case NI_QTYPE_NOOP:
1345 		break;		/* no reply data */
1346 	case NI_QTYPE_SUPTYPES:
1347 		replylen += sizeof(u_int32_t);
1348 		break;
1349 	case NI_QTYPE_FQDN:
1350 		/* XXX will append an mbuf */
1351 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1352 		break;
1353 	case NI_QTYPE_NODEADDR:
1354 		addrs = ni6_addrs(ni6, m, &ifp, subj);
1355 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1356 					  sizeof(u_int32_t))) > MCLBYTES)
1357 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1358 		break;
1359 	default:
1360 		/*
1361 		 * XXX: We must return a reply with the ICMP6 code
1362 		 * `unknown Qtype' in this case.  However we regard the case
1363 		 * as an FQDN query for backward compatibility.
1364 		 * Older versions set a random value to this field,
1365 		 * so it rarely varies in the defined qtypes.
1366 		 * But the mechanism is not reliable...
1367 		 * maybe we should obsolete older versions.
1368 		 */
1369 		qtype = NI_QTYPE_FQDN;
1370 		/* XXX will append an mbuf */
1371 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1372 		oldfqdn++;
1373 		break;
1374 	}
1375 
1376 	/* allocate an mbuf to reply. */
1377 	MGETHDR(n, M_DONTWAIT, m->m_type);
1378 	if (n == NULL) {
1379 		m_freem(m);
1380 		return (NULL);
1381 	}
1382 	M_COPY_PKTHDR(n, m); /* just for rcvif */
1383 	if (replylen > MHLEN) {
1384 		if (replylen > MCLBYTES) {
1385 			/*
1386 			 * XXX: should we try to allocate more? But MCLBYTES
1387 			 * is probably much larger than IPV6_MMTU...
1388 			 */
1389 			goto bad;
1390 		}
1391 		MCLGET(n, M_DONTWAIT);
1392 		if ((n->m_flags & M_EXT) == 0) {
1393 			goto bad;
1394 		}
1395 	}
1396 	n->m_pkthdr.len = n->m_len = replylen;
1397 
1398 	/* copy mbuf header and IPv6 + Node Information base headers */
1399 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1400 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1401 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1402 
1403 	/* qtype dependent procedure */
1404 	switch (qtype) {
1405 	case NI_QTYPE_NOOP:
1406 		nni6->ni_code = ICMP6_NI_SUCCESS;
1407 		nni6->ni_flags = 0;
1408 		break;
1409 	case NI_QTYPE_SUPTYPES:
1410 	{
1411 		u_int32_t v;
1412 		nni6->ni_code = ICMP6_NI_SUCCESS;
1413 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1414 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1415 		v = (u_int32_t)htonl(0x0000000f);
1416 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1417 		break;
1418 	}
1419 	case NI_QTYPE_FQDN:
1420 		nni6->ni_code = ICMP6_NI_SUCCESS;
1421 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1422 						sizeof(struct ip6_hdr) +
1423 						sizeof(struct icmp6_nodeinfo));
1424 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1425 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1426 		/*
1427 		 * XXX do we really have FQDN in variable "hostname"?
1428 		 */
1429 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1430 		if (n->m_next == NULL)
1431 			goto bad;
1432 		/* XXX we assume that n->m_next is not a chain */
1433 		if (n->m_next->m_next != NULL)
1434 			goto bad;
1435 		n->m_pkthdr.len += n->m_next->m_len;
1436 		break;
1437 	case NI_QTYPE_NODEADDR:
1438 	{
1439 		int lenlim, copied;
1440 
1441 		nni6->ni_code = ICMP6_NI_SUCCESS;
1442 		n->m_pkthdr.len = n->m_len =
1443 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1444 		lenlim = M_TRAILINGSPACE(n);
1445 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1446 		/* XXX: reset mbuf length */
1447 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1448 			sizeof(struct icmp6_nodeinfo) + copied;
1449 		break;
1450 	}
1451 	default:
1452 		break;		/* XXX impossible! */
1453 	}
1454 
1455 	nni6->ni_type = ICMP6_NI_REPLY;
1456 	m_freem(m);
1457 	return (n);
1458 
1459   bad:
1460 	m_freem(m);
1461 	if (n)
1462 		m_freem(n);
1463 	return (NULL);
1464 }
1465 #undef hostnamelen
1466 
1467 #define isupper(x) ('A' <= (x) && (x) <= 'Z')
1468 #define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z'))
1469 #define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9'))
1470 #define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x))
1471 
1472 /*
1473  * make a mbuf with DNS-encoded string.  no compression support.
1474  *
1475  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1476  * treated as truncated name (two \0 at the end).  this is a wild guess.
1477  */
1478 static struct mbuf *
1479 ni6_nametodns(name, namelen, old)
1480 	const char *name;
1481 	int namelen;
1482 	int old;	/* return pascal string if non-zero */
1483 {
1484 	struct mbuf *m;
1485 	char *cp, *ep;
1486 	const char *p, *q;
1487 	int i, len, nterm;
1488 
1489 	if (old)
1490 		len = namelen + 1;
1491 	else
1492 		len = MCLBYTES;
1493 
1494 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1495 	MGET(m, M_DONTWAIT, MT_DATA);
1496 	if (m && len > MLEN) {
1497 		MCLGET(m, M_DONTWAIT);
1498 		if ((m->m_flags & M_EXT) == 0)
1499 			goto fail;
1500 	}
1501 	if (!m)
1502 		goto fail;
1503 	m->m_next = NULL;
1504 
1505 	if (old) {
1506 		m->m_len = len;
1507 		*mtod(m, char *) = namelen;
1508 		bcopy(name, mtod(m, char *) + 1, namelen);
1509 		return m;
1510 	} else {
1511 		m->m_len = 0;
1512 		cp = mtod(m, char *);
1513 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1514 
1515 		/* if not certain about my name, return empty buffer */
1516 		if (namelen == 0)
1517 			return m;
1518 
1519 		/*
1520 		 * guess if it looks like shortened hostname, or FQDN.
1521 		 * shortened hostname needs two trailing "\0".
1522 		 */
1523 		i = 0;
1524 		for (p = name; p < name + namelen; p++) {
1525 			if (*p && *p == '.')
1526 				i++;
1527 		}
1528 		if (i < 2)
1529 			nterm = 2;
1530 		else
1531 			nterm = 1;
1532 
1533 		p = name;
1534 		while (cp < ep && p < name + namelen) {
1535 			i = 0;
1536 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1537 				i++;
1538 			/* result does not fit into mbuf */
1539 			if (cp + i + 1 >= ep)
1540 				goto fail;
1541 			/*
1542 			 * DNS label length restriction, RFC1035 page 8.
1543 			 * "i == 0" case is included here to avoid returning
1544 			 * 0-length label on "foo..bar".
1545 			 */
1546 			if (i <= 0 || i >= 64)
1547 				goto fail;
1548 			*cp++ = i;
1549 			if (!isalpha(p[0]) || !isalnum(p[i - 1]))
1550 				goto fail;
1551 			while (i > 0) {
1552 				if (!isalnum(*p) && *p != '-')
1553 					goto fail;
1554 				if (isupper(*p)) {
1555 					*cp++ = tolower(*p);
1556 					p++;
1557 				} else
1558 					*cp++ = *p++;
1559 				i--;
1560 			}
1561 			p = q;
1562 			if (p < name + namelen && *p == '.')
1563 				p++;
1564 		}
1565 		/* termination */
1566 		if (cp + nterm >= ep)
1567 			goto fail;
1568 		while (nterm-- > 0)
1569 			*cp++ = '\0';
1570 		m->m_len = cp - mtod(m, char *);
1571 		return m;
1572 	}
1573 
1574 	panic("should not reach here");
1575 	/* NOTREACHED */
1576 
1577  fail:
1578 	if (m)
1579 		m_freem(m);
1580 	return NULL;
1581 }
1582 
1583 /*
1584  * check if two DNS-encoded string matches.  takes care of truncated
1585  * form (with \0\0 at the end).  no compression support.
1586  * XXX upper/lowercase match (see RFC2065)
1587  */
1588 static int
1589 ni6_dnsmatch(a, alen, b, blen)
1590 	const char *a;
1591 	int alen;
1592 	const char *b;
1593 	int blen;
1594 {
1595 	const char *a0, *b0;
1596 	int l;
1597 
1598 	/* simplest case - need validation? */
1599 	if (alen == blen && bcmp(a, b, alen) == 0)
1600 		return 1;
1601 
1602 	a0 = a;
1603 	b0 = b;
1604 
1605 	/* termination is mandatory */
1606 	if (alen < 2 || blen < 2)
1607 		return 0;
1608 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1609 		return 0;
1610 	alen--;
1611 	blen--;
1612 
1613 	while (a - a0 < alen && b - b0 < blen) {
1614 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1615 			return 0;
1616 
1617 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1618 			return 0;
1619 		/* we don't support compression yet */
1620 		if (a[0] >= 64 || b[0] >= 64)
1621 			return 0;
1622 
1623 		/* truncated case */
1624 		if (a[0] == 0 && a - a0 == alen - 1)
1625 			return 1;
1626 		if (b[0] == 0 && b - b0 == blen - 1)
1627 			return 1;
1628 		if (a[0] == 0 || b[0] == 0)
1629 			return 0;
1630 
1631 		if (a[0] != b[0])
1632 			return 0;
1633 		l = a[0];
1634 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1635 			return 0;
1636 		if (bcmp(a + 1, b + 1, l) != 0)
1637 			return 0;
1638 
1639 		a += 1 + l;
1640 		b += 1 + l;
1641 	}
1642 
1643 	if (a - a0 == alen && b - b0 == blen)
1644 		return 1;
1645 	else
1646 		return 0;
1647 }
1648 
1649 /*
1650  * calculate the number of addresses to be returned in the node info reply.
1651  */
1652 static int
1653 ni6_addrs(ni6, m, ifpp, subj)
1654 	struct icmp6_nodeinfo *ni6;
1655 	struct mbuf *m;
1656 	struct ifnet **ifpp;
1657 	char *subj;
1658 {
1659 	struct ifnet *ifp;
1660 	struct in6_ifaddr *ifa6;
1661 	struct ifaddr *ifa;
1662 	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1663 	int addrs = 0, addrsofif, iffound = 0;
1664 	int niflags = ni6->ni_flags;
1665 
1666 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1667 		switch (ni6->ni_code) {
1668 		case ICMP6_NI_SUBJ_IPV6:
1669 			if (subj == NULL) /* must be impossible... */
1670 				return (0);
1671 			subj_ip6 = (struct sockaddr_in6 *)subj;
1672 			break;
1673 		default:
1674 			/*
1675 			 * XXX: we only support IPv6 subject address for
1676 			 * this Qtype.
1677 			 */
1678 			return (0);
1679 		}
1680 	}
1681 
1682 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
1683 	{
1684 		addrsofif = 0;
1685 		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1686 		     ifa = ifa->ifa_list.tqe_next)
1687 		{
1688 			if (ifa->ifa_addr->sa_family != AF_INET6)
1689 				continue;
1690 			ifa6 = (struct in6_ifaddr *)ifa;
1691 
1692 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1693 			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1694 					       &ifa6->ia_addr.sin6_addr))
1695 				iffound = 1;
1696 
1697 			/*
1698 			 * IPv4-mapped addresses can only be returned by a
1699 			 * Node Information proxy, since they represent
1700 			 * addresses of IPv4-only nodes, which perforce do
1701 			 * not implement this protocol.
1702 			 * [icmp-name-lookups-07, Section 5.4]
1703 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1704 			 * this function at this moment.
1705 			 */
1706 
1707 			/* What do we have to do about ::1? */
1708 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1709 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1710 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1711 					continue;
1712 				break;
1713 			case IPV6_ADDR_SCOPE_SITELOCAL:
1714 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1715 					continue;
1716 				break;
1717 			case IPV6_ADDR_SCOPE_GLOBAL:
1718 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1719 					continue;
1720 				break;
1721 			default:
1722 				continue;
1723 			}
1724 
1725 			/*
1726 			 * check if anycast is okay.
1727 			 * XXX: just experimental.  not in the spec.
1728 			 */
1729 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1730 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1731 				continue; /* we need only unicast addresses */
1732 
1733 			addrsofif++; /* count the address */
1734 		}
1735 		if (iffound) {
1736 			*ifpp = ifp;
1737 			return (addrsofif);
1738 		}
1739 
1740 		addrs += addrsofif;
1741 	}
1742 
1743 	return (addrs);
1744 }
1745 
1746 static int
1747 ni6_store_addrs(ni6, nni6, ifp0, resid)
1748 	struct icmp6_nodeinfo *ni6, *nni6;
1749 	struct ifnet *ifp0;
1750 	int resid;
1751 {
1752 	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1753 	struct in6_ifaddr *ifa6;
1754 	struct ifaddr *ifa;
1755 	struct ifnet *ifp_dep = NULL;
1756 	int copied = 0, allow_deprecated = 0;
1757 	u_char *cp = (u_char *)(nni6 + 1);
1758 	int niflags = ni6->ni_flags;
1759 	u_int32_t ltime;
1760 
1761 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1762 		return (0);	/* needless to copy */
1763 
1764   again:
1765 
1766 	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list))
1767 	{
1768 		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1769 		     ifa = ifa->ifa_list.tqe_next)
1770 		{
1771 			if (ifa->ifa_addr->sa_family != AF_INET6)
1772 				continue;
1773 			ifa6 = (struct in6_ifaddr *)ifa;
1774 
1775 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1776 			    allow_deprecated == 0) {
1777 				/*
1778 				 * prefererred address should be put before
1779 				 * deprecated addresses.
1780 				 */
1781 
1782 				/* record the interface for later search */
1783 				if (ifp_dep == NULL)
1784 					ifp_dep = ifp;
1785 
1786 				continue;
1787 			}
1788 			else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1789 				 allow_deprecated != 0)
1790 				continue; /* we now collect deprecated addrs */
1791 
1792 			/* What do we have to do about ::1? */
1793 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1794 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1795 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1796 					continue;
1797 				break;
1798 			case IPV6_ADDR_SCOPE_SITELOCAL:
1799 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1800 					continue;
1801 				break;
1802 			case IPV6_ADDR_SCOPE_GLOBAL:
1803 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1804 					continue;
1805 				break;
1806 			default:
1807 				continue;
1808 			}
1809 
1810 			/*
1811 			 * check if anycast is okay.
1812 			 * XXX: just experimental.  not in the spec.
1813 			 */
1814 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1815 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1816 				continue;
1817 
1818 			/* now we can copy the address */
1819 			if (resid < sizeof(struct in6_addr) +
1820 			    sizeof(u_int32_t)) {
1821 				/*
1822 				 * We give up much more copy.
1823 				 * Set the truncate flag and return.
1824 				 */
1825 				nni6->ni_flags |=
1826 					NI_NODEADDR_FLAG_TRUNCATE;
1827 				return (copied);
1828 			}
1829 
1830 			/*
1831 			 * Set the TTL of the address.
1832 			 * The TTL value should be one of the following
1833 			 * according to the specification:
1834 			 *
1835 			 * 1. The remaining lifetime of a DHCP lease on the
1836 			 *    address, or
1837 			 * 2. The remaining Valid Lifetime of a prefix from
1838 			 *    which the address was derived through Stateless
1839 			 *    Autoconfiguration.
1840 			 *
1841 			 * Note that we currently do not support stateful
1842 			 * address configuration by DHCPv6, so the former
1843 			 * case can't happen.
1844 			 *
1845 			 * TTL must be 2^31 > TTL >= 0.
1846 			 */
1847 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1848 				ltime = ND6_INFINITE_LIFETIME;
1849 			else {
1850 				if (ifa6->ia6_lifetime.ia6t_expire >
1851 				    time.tv_sec)
1852 					ltime = ifa6->ia6_lifetime.ia6t_expire - time.tv_sec;
1853 				else
1854 					ltime = 0;
1855 			}
1856 			if (ltime > 0x7fffffff)
1857 				ltime = 0x7fffffff;
1858 			ltime = htonl(ltime);
1859 
1860 			bcopy(&ltime, cp, sizeof(u_int32_t));
1861 			cp += sizeof(u_int32_t);
1862 
1863 			/* copy the address itself */
1864 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1865 			      sizeof(struct in6_addr));
1866 			/* XXX: KAME link-local hack; remove ifindex */
1867 			if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr))
1868 				((struct in6_addr *)cp)->s6_addr16[1] = 0;
1869 			cp += sizeof(struct in6_addr);
1870 
1871 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1872 			copied += (sizeof(struct in6_addr) +
1873 				   sizeof(u_int32_t));
1874 		}
1875 		if (ifp0)	/* we need search only on the specified IF */
1876 			break;
1877 	}
1878 
1879 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1880 		ifp = ifp_dep;
1881 		allow_deprecated = 1;
1882 
1883 		goto again;
1884 	}
1885 
1886 	return (copied);
1887 }
1888 
1889 /*
1890  * XXX almost dup'ed code with rip6_input.
1891  */
1892 static int
1893 icmp6_rip6_input(mp, off)
1894 	struct	mbuf **mp;
1895 	int	off;
1896 {
1897 	struct mbuf *m = *mp;
1898 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1899 	struct inpcb_hdr *inph;
1900 	struct in6pcb *in6p;
1901 	struct in6pcb *last = NULL;
1902 	struct sockaddr_in6 rip6src;
1903 	struct icmp6_hdr *icmp6;
1904 	struct mbuf *opts = NULL;
1905 
1906 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1907 	if (icmp6 == NULL) {
1908 		/* m is already reclaimed */
1909 		return IPPROTO_DONE;
1910 	}
1911 
1912 	bzero(&rip6src, sizeof(rip6src));
1913 	rip6src.sin6_len = sizeof(struct sockaddr_in6);
1914 	rip6src.sin6_family = AF_INET6;
1915 	/* KAME hack: recover scopeid */
1916 	(void)in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif);
1917 
1918 	CIRCLEQ_FOREACH(inph, &raw6cbtable.inpt_queue, inph_queue) {
1919 		in6p = (struct in6pcb *)inph;
1920 		if (in6p->in6p_af != AF_INET6)
1921 			continue;
1922 		if (in6p->in6p_ip6.ip6_nxt != IPPROTO_ICMPV6)
1923 			continue;
1924 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1925 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1926 			continue;
1927 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1928 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1929 			continue;
1930 		if (in6p->in6p_icmp6filt
1931 		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1932 				 in6p->in6p_icmp6filt))
1933 			continue;
1934 		if (last) {
1935 			struct	mbuf *n;
1936 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1937 				if (last->in6p_flags & IN6P_CONTROLOPTS)
1938 					ip6_savecontrol(last, &opts, ip6, n);
1939 				/* strip intermediate headers */
1940 				m_adj(n, off);
1941 				if (sbappendaddr(&last->in6p_socket->so_rcv,
1942 						 (struct sockaddr *)&rip6src,
1943 						 n, opts) == 0) {
1944 					/* should notify about lost packet */
1945 					m_freem(n);
1946 					if (opts)
1947 						m_freem(opts);
1948 				} else
1949 					sorwakeup(last->in6p_socket);
1950 				opts = NULL;
1951 			}
1952 		}
1953 		last = in6p;
1954 	}
1955 	if (last) {
1956 		if (last->in6p_flags & IN6P_CONTROLOPTS)
1957 			ip6_savecontrol(last, &opts, ip6, m);
1958 		/* strip intermediate headers */
1959 		m_adj(m, off);
1960 		if (sbappendaddr(&last->in6p_socket->so_rcv,
1961 				(struct sockaddr *)&rip6src, m, opts) == 0) {
1962 			m_freem(m);
1963 			if (opts)
1964 				m_freem(opts);
1965 		} else
1966 			sorwakeup(last->in6p_socket);
1967 	} else {
1968 		m_freem(m);
1969 		ip6stat.ip6s_delivered--;
1970 	}
1971 	return IPPROTO_DONE;
1972 }
1973 
1974 /*
1975  * Reflect the ip6 packet back to the source.
1976  * OFF points to the icmp6 header, counted from the top of the mbuf.
1977  *
1978  * Note: RFC 1885 required that an echo reply should be truncated if it
1979  * did not fit in with (return) path MTU, and KAME code supported the
1980  * behavior.  However, as a clarification after the RFC, this limitation
1981  * was removed in a revised version of the spec, RFC 2463.  We had kept the
1982  * old behavior, with a (non-default) ifdef block, while the new version of
1983  * the spec was an internet-draft status, and even after the new RFC was
1984  * published.  But it would rather make sense to clean the obsoleted part
1985  * up, and to make the code simpler at this stage.
1986  */
1987 void
1988 icmp6_reflect(m, off)
1989 	struct	mbuf *m;
1990 	size_t off;
1991 {
1992 	struct ip6_hdr *ip6;
1993 	struct icmp6_hdr *icmp6;
1994 	struct in6_ifaddr *ia;
1995 	struct in6_addr t, *src = 0;
1996 	int plen;
1997 	int type, code;
1998 	struct ifnet *outif = NULL;
1999 	struct sockaddr_in6 sa6_src, sa6_dst;
2000 
2001 	/* too short to reflect */
2002 	if (off < sizeof(struct ip6_hdr)) {
2003 		nd6log((LOG_DEBUG,
2004 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2005 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2006 		    __FILE__, __LINE__));
2007 		goto bad;
2008 	}
2009 
2010 	/*
2011 	 * If there are extra headers between IPv6 and ICMPv6, strip
2012 	 * off that header first.
2013 	 */
2014 #ifdef DIAGNOSTIC
2015 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2016 		panic("assumption failed in icmp6_reflect");
2017 #endif
2018 	if (off > sizeof(struct ip6_hdr)) {
2019 		size_t l;
2020 		struct ip6_hdr nip6;
2021 
2022 		l = off - sizeof(struct ip6_hdr);
2023 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2024 		m_adj(m, l);
2025 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2026 		if (m->m_len < l) {
2027 			if ((m = m_pullup(m, l)) == NULL)
2028 				return;
2029 		}
2030 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2031 	} else /* off == sizeof(struct ip6_hdr) */ {
2032 		size_t l;
2033 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2034 		if (m->m_len < l) {
2035 			if ((m = m_pullup(m, l)) == NULL)
2036 				return;
2037 		}
2038 	}
2039 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2040 	ip6 = mtod(m, struct ip6_hdr *);
2041 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2042 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2043 	type = icmp6->icmp6_type; /* keep type for statistics */
2044 	code = icmp6->icmp6_code; /* ditto. */
2045 
2046 	t = ip6->ip6_dst;
2047 	/*
2048 	 * ip6_input() drops a packet if its src is multicast.
2049 	 * So, the src is never multicast.
2050 	 */
2051 	ip6->ip6_dst = ip6->ip6_src;
2052 
2053 	/*
2054 	 * XXX: make sure to embed scope zone information, using
2055 	 * already embedded IDs or the received interface (if any).
2056 	 * Note that rcvif may be NULL.
2057 	 * TODO: scoped routing case (XXX).
2058 	 */
2059 	bzero(&sa6_src, sizeof(sa6_src));
2060 	sa6_src.sin6_family = AF_INET6;
2061 	sa6_src.sin6_len = sizeof(sa6_src);
2062 	sa6_src.sin6_addr = ip6->ip6_dst;
2063 	in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
2064 	in6_embedscope(&sa6_src.sin6_addr, &sa6_src, NULL, NULL);
2065 	ip6->ip6_dst = sa6_src.sin6_addr;
2066 
2067 	bzero(&sa6_dst, sizeof(sa6_dst));
2068 	sa6_dst.sin6_family = AF_INET6;
2069 	sa6_dst.sin6_len = sizeof(sa6_dst);
2070 	sa6_dst.sin6_addr = t;
2071 	in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
2072 	in6_embedscope(&t, &sa6_dst, NULL, NULL);
2073 
2074 	/*
2075 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2076 	 * use dst as the src for the reply.
2077 	 * The IN6_IFF_NOTREADY case would be VERY rare, but is possible
2078 	 * (for example) when we encounter an error while forwarding procedure
2079 	 * destined to a duplicated address of ours.
2080 	 */
2081 	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2082 		if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2083 		    (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2084 			src = &t;
2085 			break;
2086 		}
2087 	if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2088 		/*
2089 		 * This is the case if the dst is our link-local address
2090 		 * and the sender is also ourselves.
2091 		 */
2092 		src = &t;
2093 	}
2094 
2095 	if (src == 0) {
2096 		int e;
2097 		struct route_in6 ro;
2098 
2099 		/*
2100 		 * This case matches to multicasts, our anycast, or unicasts
2101 		 * that we do not own.  Select a source address based on the
2102 		 * source address of the erroneous packet.
2103 		 */
2104 		bzero(&ro, sizeof(ro));
2105 		src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e);
2106 		if (ro.ro_rt) { /* XXX: see comments in icmp6_mtudisc_update */
2107 			RTFREE(ro.ro_rt); /* XXX: we could use this */
2108 		}
2109 		if (src == NULL) {
2110 			nd6log((LOG_DEBUG,
2111 			    "icmp6_reflect: source can't be determined: "
2112 			    "dst=%s, error=%d\n",
2113 			    ip6_sprintf(&sa6_src.sin6_addr), e));
2114 			goto bad;
2115 		}
2116 	}
2117 
2118 	ip6->ip6_src = *src;
2119 
2120 	ip6->ip6_flow = 0;
2121 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2122 	ip6->ip6_vfc |= IPV6_VERSION;
2123 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2124 	if (m->m_pkthdr.rcvif) {
2125 		/* XXX: This may not be the outgoing interface */
2126 		ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2127 	} else
2128 		ip6->ip6_hlim = ip6_defhlim;
2129 
2130 	icmp6->icmp6_cksum = 0;
2131 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2132 					sizeof(struct ip6_hdr), plen);
2133 
2134 	/*
2135 	 * XXX option handling
2136 	 */
2137 
2138 	m->m_flags &= ~(M_BCAST|M_MCAST);
2139 
2140 	/*
2141 	 * To avoid a "too big" situation at an intermediate router
2142 	 * and the path MTU discovery process, specify the IPV6_MINMTU flag.
2143 	 * Note that only echo and node information replies are affected,
2144 	 * since the length of ICMP6 errors is limited to the minimum MTU.
2145 	 */
2146 	if (ip6_output(m, NULL, NULL, IPV6_MINMTU,
2147 		(struct ip6_moptions *)NULL, (struct socket *)NULL, &outif) != 0
2148 	    && outif)
2149 		icmp6_ifstat_inc(outif, ifs6_out_error);
2150 
2151 	if (outif)
2152 		icmp6_ifoutstat_inc(outif, type, code);
2153 
2154 	return;
2155 
2156  bad:
2157 	m_freem(m);
2158 	return;
2159 }
2160 
2161 void
2162 icmp6_fasttimo()
2163 {
2164 
2165 	mld6_fasttimeo();
2166 }
2167 
2168 static const char *
2169 icmp6_redirect_diag(src6, dst6, tgt6)
2170 	struct in6_addr *src6;
2171 	struct in6_addr *dst6;
2172 	struct in6_addr *tgt6;
2173 {
2174 	static char buf[1024];
2175 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2176 		ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2177 	return buf;
2178 }
2179 
2180 void
2181 icmp6_redirect_input(m, off)
2182 	struct mbuf *m;
2183 	int off;
2184 {
2185 	struct ifnet *ifp = m->m_pkthdr.rcvif;
2186 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2187 	struct nd_redirect *nd_rd;
2188 	int icmp6len = ntohs(ip6->ip6_plen);
2189 	char *lladdr = NULL;
2190 	int lladdrlen = 0;
2191 	struct rtentry *rt = NULL;
2192 	int is_router;
2193 	int is_onlink;
2194 	struct in6_addr src6 = ip6->ip6_src;
2195 	struct in6_addr redtgt6;
2196 	struct in6_addr reddst6;
2197 	union nd_opts ndopts;
2198 
2199 	if (!ifp)
2200 		return;
2201 
2202 	/* XXX if we are router, we don't update route by icmp6 redirect */
2203 	if (ip6_forwarding)
2204 		goto freeit;
2205 	if (!icmp6_rediraccept)
2206 		goto freeit;
2207 
2208 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2209 	if (nd_rd == NULL) {
2210 		icmp6stat.icp6s_tooshort++;
2211 		return;
2212 	}
2213 	redtgt6 = nd_rd->nd_rd_target;
2214 	reddst6 = nd_rd->nd_rd_dst;
2215 
2216 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2217 		redtgt6.s6_addr16[1] = htons(ifp->if_index);
2218 	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2219 		reddst6.s6_addr16[1] = htons(ifp->if_index);
2220 
2221 	/* validation */
2222 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2223 		nd6log((LOG_ERR,
2224 			"ICMP6 redirect sent from %s rejected; "
2225 			"must be from linklocal\n", ip6_sprintf(&src6)));
2226 		goto bad;
2227 	}
2228 	if (ip6->ip6_hlim != 255) {
2229 		nd6log((LOG_ERR,
2230 			"ICMP6 redirect sent from %s rejected; "
2231 			"hlim=%d (must be 255)\n",
2232 			ip6_sprintf(&src6), ip6->ip6_hlim));
2233 		goto bad;
2234 	}
2235     {
2236 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2237 	struct sockaddr_in6 sin6;
2238 	struct in6_addr *gw6;
2239 
2240 	bzero(&sin6, sizeof(sin6));
2241 	sin6.sin6_family = AF_INET6;
2242 	sin6.sin6_len = sizeof(struct sockaddr_in6);
2243 	bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2244 	rt = rtalloc1((struct sockaddr *)&sin6, 0);
2245 	if (rt) {
2246 		if (rt->rt_gateway == NULL ||
2247 		    rt->rt_gateway->sa_family != AF_INET6) {
2248 			nd6log((LOG_ERR,
2249 			    "ICMP6 redirect rejected; no route "
2250 			    "with inet6 gateway found for redirect dst: %s\n",
2251 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2252 			RTFREE(rt);
2253 			goto bad;
2254 		}
2255 
2256 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2257 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2258 			nd6log((LOG_ERR,
2259 				"ICMP6 redirect rejected; "
2260 				"not equal to gw-for-src=%s (must be same): "
2261 				"%s\n",
2262 				ip6_sprintf(gw6),
2263 				icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2264 			RTFREE(rt);
2265 			goto bad;
2266 		}
2267 	} else {
2268 		nd6log((LOG_ERR,
2269 			"ICMP6 redirect rejected; "
2270 			"no route found for redirect dst: %s\n",
2271 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2272 		goto bad;
2273 	}
2274 	RTFREE(rt);
2275 	rt = NULL;
2276     }
2277 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2278 		nd6log((LOG_ERR,
2279 			"ICMP6 redirect rejected; "
2280 			"redirect dst must be unicast: %s\n",
2281 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2282 		goto bad;
2283 	}
2284 
2285 	is_router = is_onlink = 0;
2286 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2287 		is_router = 1;	/* router case */
2288 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2289 		is_onlink = 1;	/* on-link destination case */
2290 	if (!is_router && !is_onlink) {
2291 		nd6log((LOG_ERR,
2292 			"ICMP6 redirect rejected; "
2293 			"neither router case nor onlink case: %s\n",
2294 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2295 		goto bad;
2296 	}
2297 	/* validation passed */
2298 
2299 	icmp6len -= sizeof(*nd_rd);
2300 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2301 	if (nd6_options(&ndopts) < 0) {
2302 		nd6log((LOG_INFO, "icmp6_redirect_input: "
2303 			"invalid ND option, rejected: %s\n",
2304 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2305 		/* nd6_options have incremented stats */
2306 		goto freeit;
2307 	}
2308 
2309 	if (ndopts.nd_opts_tgt_lladdr) {
2310 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2311 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2312 	}
2313 
2314 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2315 		nd6log((LOG_INFO,
2316 			"icmp6_redirect_input: lladdrlen mismatch for %s "
2317 			"(if %d, icmp6 packet %d): %s\n",
2318 			ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2319 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2320 		goto bad;
2321 	}
2322 
2323 	/* RFC 2461 8.3 */
2324 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2325 			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2326 
2327 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2328 		/* perform rtredirect */
2329 		struct sockaddr_in6 sdst;
2330 		struct sockaddr_in6 sgw;
2331 		struct sockaddr_in6 ssrc;
2332 		unsigned long rtcount;
2333 		struct rtentry *newrt = NULL;
2334 
2335 		/*
2336 		 * do not install redirect route, if the number of entries
2337 		 * is too much (> hiwat).  note that, the node (= host) will
2338 		 * work just fine even if we do not install redirect route
2339 		 * (there will be additional hops, though).
2340 		 */
2341 		rtcount = rt_timer_count(icmp6_redirect_timeout_q);
2342 		if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat)
2343 			return;
2344 		else if (0 <= icmp6_redirect_lowat &&
2345 		    rtcount > icmp6_redirect_lowat) {
2346 			/*
2347 			 * XXX nuke a victim, install the new one.
2348 			 */
2349 		}
2350 
2351 		bzero(&sdst, sizeof(sdst));
2352 		bzero(&sgw, sizeof(sgw));
2353 		bzero(&ssrc, sizeof(ssrc));
2354 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2355 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2356 			sizeof(struct sockaddr_in6);
2357 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2358 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2359 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2360 		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2361 			   (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2362 			   (struct sockaddr *)&ssrc,
2363 			   &newrt);
2364 
2365 		if (newrt) {
2366 			(void)rt_timer_add(newrt, icmp6_redirect_timeout,
2367 			    icmp6_redirect_timeout_q);
2368 			rtfree(newrt);
2369 		}
2370 	}
2371 	/* finally update cached route in each socket via pfctlinput */
2372 	{
2373 		struct sockaddr_in6 sdst;
2374 
2375 		bzero(&sdst, sizeof(sdst));
2376 		sdst.sin6_family = AF_INET6;
2377 		sdst.sin6_len = sizeof(struct sockaddr_in6);
2378 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2379 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2380 #ifdef IPSEC
2381 		key_sa_routechange((struct sockaddr *)&sdst);
2382 #endif
2383 	}
2384 
2385  freeit:
2386 	m_freem(m);
2387 	return;
2388 
2389  bad:
2390 	icmp6stat.icp6s_badredirect++;
2391 	m_freem(m);
2392 }
2393 
2394 void
2395 icmp6_redirect_output(m0, rt)
2396 	struct mbuf *m0;
2397 	struct rtentry *rt;
2398 {
2399 	struct ifnet *ifp;	/* my outgoing interface */
2400 	struct in6_addr *ifp_ll6;
2401 	struct in6_addr *nexthop;
2402 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2403 	struct mbuf *m = NULL;	/* newly allocated one */
2404 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2405 	struct nd_redirect *nd_rd;
2406 	size_t maxlen;
2407 	u_char *p;
2408 	struct sockaddr_in6 src_sa;
2409 
2410 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2411 
2412 	/* if we are not router, we don't send icmp6 redirect */
2413 	if (!ip6_forwarding)
2414 		goto fail;
2415 
2416 	/* sanity check */
2417 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2418 		goto fail;
2419 
2420 	/*
2421 	 * Address check:
2422 	 *  the source address must identify a neighbor, and
2423 	 *  the destination address must not be a multicast address
2424 	 *  [RFC 2461, sec 8.2]
2425 	 */
2426 	sip6 = mtod(m0, struct ip6_hdr *);
2427 	bzero(&src_sa, sizeof(src_sa));
2428 	src_sa.sin6_family = AF_INET6;
2429 	src_sa.sin6_len = sizeof(src_sa);
2430 	src_sa.sin6_addr = sip6->ip6_src;
2431 	/* we don't currently use sin6_scope_id, but eventually use it */
2432 	src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src);
2433 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2434 		goto fail;
2435 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2436 		goto fail;	/* what should we do here? */
2437 
2438 	/* rate limit */
2439 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2440 		goto fail;
2441 
2442 	/*
2443 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2444 	 * we almost always ask for an mbuf cluster for simplicity.
2445 	 * (MHLEN < IPV6_MMTU is almost always true)
2446 	 */
2447 #if IPV6_MMTU >= MCLBYTES
2448 # error assumption failed about IPV6_MMTU and MCLBYTES
2449 #endif
2450 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2451 	if (m && IPV6_MMTU >= MHLEN)
2452 		MCLGET(m, M_DONTWAIT);
2453 	if (!m)
2454 		goto fail;
2455 	m->m_pkthdr.rcvif = NULL;
2456 	m->m_len = 0;
2457 	maxlen = M_TRAILINGSPACE(m);
2458 	maxlen = min(IPV6_MMTU, maxlen);
2459 	/* just for safety */
2460 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2461 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2462 		goto fail;
2463 	}
2464 
2465 	{
2466 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2467 		struct in6_ifaddr *ia;
2468 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2469 						 IN6_IFF_NOTREADY|
2470 						 IN6_IFF_ANYCAST)) == NULL)
2471 			goto fail;
2472 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2473 	}
2474 
2475 	/* get ip6 linklocal address for the router. */
2476 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2477 		struct sockaddr_in6 *sin6;
2478 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2479 		nexthop = &sin6->sin6_addr;
2480 		if (!IN6_IS_ADDR_LINKLOCAL(nexthop))
2481 			nexthop = NULL;
2482 	} else
2483 		nexthop = NULL;
2484 
2485 	/* ip6 */
2486 	ip6 = mtod(m, struct ip6_hdr *);
2487 	ip6->ip6_flow = 0;
2488 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2489 	ip6->ip6_vfc |= IPV6_VERSION;
2490 	/* ip6->ip6_plen will be set later */
2491 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2492 	ip6->ip6_hlim = 255;
2493 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2494 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2495 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2496 
2497 	/* ND Redirect */
2498 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2499 	nd_rd->nd_rd_type = ND_REDIRECT;
2500 	nd_rd->nd_rd_code = 0;
2501 	nd_rd->nd_rd_reserved = 0;
2502 	if (rt->rt_flags & RTF_GATEWAY) {
2503 		/*
2504 		 * nd_rd->nd_rd_target must be a link-local address in
2505 		 * better router cases.
2506 		 */
2507 		if (!nexthop)
2508 			goto fail;
2509 		bcopy(nexthop, &nd_rd->nd_rd_target,
2510 		      sizeof(nd_rd->nd_rd_target));
2511 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2512 		      sizeof(nd_rd->nd_rd_dst));
2513 	} else {
2514 		/* make sure redtgt == reddst */
2515 		nexthop = &sip6->ip6_dst;
2516 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2517 		      sizeof(nd_rd->nd_rd_target));
2518 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2519 		      sizeof(nd_rd->nd_rd_dst));
2520 	}
2521 
2522 	p = (u_char *)(nd_rd + 1);
2523 
2524 	{
2525 		/* target lladdr option */
2526 		struct rtentry *rt_nexthop = NULL;
2527 		int len;
2528 		struct sockaddr_dl *sdl;
2529 		struct nd_opt_hdr *nd_opt;
2530 		char *lladdr;
2531 
2532 		rt_nexthop = nd6_lookup(nexthop, 0, ifp);
2533 		if (!rt_nexthop)
2534 			goto nolladdropt;
2535 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2536 		len = (len + 7) & ~7;	/* round by 8 */
2537 		/* safety check */
2538 		if (len + (p - (u_char *)ip6) > maxlen)
2539 			goto nolladdropt;
2540 		if (!(rt_nexthop->rt_flags & RTF_GATEWAY) &&
2541 		    (rt_nexthop->rt_flags & RTF_LLINFO) &&
2542 		    (rt_nexthop->rt_gateway->sa_family == AF_LINK) &&
2543 		    (sdl = (struct sockaddr_dl *)rt_nexthop->rt_gateway) &&
2544 		    sdl->sdl_alen) {
2545 			nd_opt = (struct nd_opt_hdr *)p;
2546 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2547 			nd_opt->nd_opt_len = len >> 3;
2548 			lladdr = (char *)(nd_opt + 1);
2549 			bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2550 			p += len;
2551 		}
2552 	}
2553   nolladdropt:;
2554 
2555 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2556 
2557 	/* just to be safe */
2558 	if (m0->m_flags & M_DECRYPTED)
2559 		goto noredhdropt;
2560 	if (p - (u_char *)ip6 > maxlen)
2561 		goto noredhdropt;
2562 
2563 	{
2564 		/* redirected header option */
2565 		int len;
2566 		struct nd_opt_rd_hdr *nd_opt_rh;
2567 
2568 		/*
2569 		 * compute the maximum size for icmp6 redirect header option.
2570 		 * XXX room for auth header?
2571 		 */
2572 		len = maxlen - (p - (u_char *)ip6);
2573 		len &= ~7;
2574 
2575 		/*
2576 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2577 		 * about padding/truncate rule for the original IP packet.
2578 		 * From the discussion on IPv6imp in Feb 1999,
2579 		 * the consensus was:
2580 		 * - "attach as much as possible" is the goal
2581 		 * - pad if not aligned (original size can be guessed by
2582 		 *   original ip6 header)
2583 		 * Following code adds the padding if it is simple enough,
2584 		 * and truncates if not.
2585 		 */
2586 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2587 			/* not enough room, truncate */
2588 			m_adj(m0, (len - sizeof(*nd_opt_rh)) -
2589 			    m0->m_pkthdr.len);
2590 		} else {
2591 			/*
2592                          * enough room, truncate if not aligned.
2593 			 * we don't pad here for simplicity.
2594 			 */
2595 			size_t extra;
2596 
2597 			extra = m0->m_pkthdr.len % 8;
2598 			if (extra) {
2599 				/* truncate */
2600 				m_adj(m0, -extra);
2601 			}
2602 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2603 		}
2604 
2605 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2606 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2607 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2608 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2609 		p += sizeof(*nd_opt_rh);
2610 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2611 
2612 		/* connect m0 to m */
2613 		m->m_pkthdr.len += m0->m_pkthdr.len;
2614 		m_cat(m, m0);
2615 		m0 = NULL;
2616 	}
2617 noredhdropt:
2618 	if (m0) {
2619 		m_freem(m0);
2620 		m0 = NULL;
2621 	}
2622 
2623 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
2624 		sip6->ip6_src.s6_addr16[1] = 0;
2625 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
2626 		sip6->ip6_dst.s6_addr16[1] = 0;
2627 #if 0
2628 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
2629 		ip6->ip6_src.s6_addr16[1] = 0;
2630 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
2631 		ip6->ip6_dst.s6_addr16[1] = 0;
2632 #endif
2633 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
2634 		nd_rd->nd_rd_target.s6_addr16[1] = 0;
2635 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
2636 		nd_rd->nd_rd_dst.s6_addr16[1] = 0;
2637 
2638 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2639 
2640 	nd_rd->nd_rd_cksum = 0;
2641 	nd_rd->nd_rd_cksum
2642 		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2643 
2644 	/* send the packet to outside... */
2645 	if (ip6_output(m, NULL, NULL, 0,
2646 		(struct ip6_moptions *)NULL, (struct socket *)NULL, NULL) != 0)
2647 		icmp6_ifstat_inc(ifp, ifs6_out_error);
2648 
2649 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
2650 	icmp6_ifstat_inc(ifp, ifs6_out_redirect);
2651 	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2652 
2653 	return;
2654 
2655 fail:
2656 	if (m)
2657 		m_freem(m);
2658 	if (m0)
2659 		m_freem(m0);
2660 }
2661 
2662 /*
2663  * ICMPv6 socket option processing.
2664  */
2665 int
2666 icmp6_ctloutput(op, so, level, optname, mp)
2667 	int op;
2668 	struct socket *so;
2669 	int level, optname;
2670 	struct mbuf **mp;
2671 {
2672 	int error = 0;
2673 	int optlen;
2674 	struct in6pcb *in6p = sotoin6pcb(so);
2675 	struct mbuf *m = *mp;
2676 
2677 	optlen = m ? m->m_len : 0;
2678 
2679 	if (level != IPPROTO_ICMPV6) {
2680 		if (op == PRCO_SETOPT && m)
2681 			(void)m_free(m);
2682 		return EINVAL;
2683 	}
2684 
2685 	switch (op) {
2686 	case PRCO_SETOPT:
2687 		switch (optname) {
2688 		case ICMP6_FILTER:
2689 		    {
2690 			struct icmp6_filter *p;
2691 
2692 			if (optlen != sizeof(*p)) {
2693 				error = EMSGSIZE;
2694 				break;
2695 			}
2696 			p = mtod(m, struct icmp6_filter *);
2697 			if (!p || !in6p->in6p_icmp6filt) {
2698 				error = EINVAL;
2699 				break;
2700 			}
2701 			bcopy(p, in6p->in6p_icmp6filt,
2702 				sizeof(struct icmp6_filter));
2703 			error = 0;
2704 			break;
2705 		    }
2706 
2707 		default:
2708 			error = ENOPROTOOPT;
2709 			break;
2710 		}
2711 		if (m)
2712 			(void)m_freem(m);
2713 		break;
2714 
2715 	case PRCO_GETOPT:
2716 		switch (optname) {
2717 		case ICMP6_FILTER:
2718 		    {
2719 			struct icmp6_filter *p;
2720 
2721 			if (!in6p->in6p_icmp6filt) {
2722 				error = EINVAL;
2723 				break;
2724 			}
2725 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
2726 			m->m_len = sizeof(struct icmp6_filter);
2727 			p = mtod(m, struct icmp6_filter *);
2728 			bcopy(in6p->in6p_icmp6filt, p,
2729 				sizeof(struct icmp6_filter));
2730 			error = 0;
2731 			break;
2732 		    }
2733 
2734 		default:
2735 			error = ENOPROTOOPT;
2736 			break;
2737 		}
2738 		break;
2739 	}
2740 
2741 	return (error);
2742 }
2743 
2744 /*
2745  * Perform rate limit check.
2746  * Returns 0 if it is okay to send the icmp6 packet.
2747  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2748  * limitation.
2749  *
2750  * XXX per-destination/type check necessary?
2751  */
2752 static int
2753 icmp6_ratelimit(dst, type, code)
2754 	const struct in6_addr *dst;	/* not used at this moment */
2755 	const int type;			/* not used at this moment */
2756 	const int code;			/* not used at this moment */
2757 {
2758 	int ret;
2759 
2760 	ret = 0;	/* okay to send */
2761 
2762 	/* PPS limit */
2763 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2764 	    icmp6errppslim)) {
2765 		/* The packet is subject to rate limit */
2766 		ret++;
2767 	}
2768 
2769 	return ret;
2770 }
2771 
2772 static struct rtentry *
2773 icmp6_mtudisc_clone(dst)
2774 	struct sockaddr *dst;
2775 {
2776 	struct rtentry *rt;
2777 	int    error;
2778 
2779 	rt = rtalloc1(dst, 1);
2780 	if (rt == 0)
2781 		return NULL;
2782 
2783 	/* If we didn't get a host route, allocate one */
2784 	if ((rt->rt_flags & RTF_HOST) == 0) {
2785 		struct rtentry *nrt;
2786 
2787 		error = rtrequest((int) RTM_ADD, dst,
2788 		    (struct sockaddr *) rt->rt_gateway,
2789 		    (struct sockaddr *) 0,
2790 		    RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt);
2791 		if (error) {
2792 			rtfree(rt);
2793 			return NULL;
2794 		}
2795 		nrt->rt_rmx = rt->rt_rmx;
2796 		rtfree(rt);
2797 		rt = nrt;
2798 	}
2799 	error = rt_timer_add(rt, icmp6_mtudisc_timeout,
2800 			icmp6_mtudisc_timeout_q);
2801 	if (error) {
2802 		rtfree(rt);
2803 		return NULL;
2804 	}
2805 
2806 	return rt;	/* caller need to call rtfree() */
2807 }
2808 
2809 static void
2810 icmp6_mtudisc_timeout(rt, r)
2811 	struct rtentry *rt;
2812 	struct rttimer *r;
2813 {
2814 	if (rt == NULL)
2815 		panic("icmp6_mtudisc_timeout: bad route to timeout");
2816 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
2817 	    (RTF_DYNAMIC | RTF_HOST)) {
2818 		rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt),
2819 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
2820 	} else {
2821 		if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
2822 			rt->rt_rmx.rmx_mtu = 0;
2823 	}
2824 }
2825 
2826 static void
2827 icmp6_redirect_timeout(rt, r)
2828 	struct rtentry *rt;
2829 	struct rttimer *r;
2830 {
2831 	if (rt == NULL)
2832 		panic("icmp6_redirect_timeout: bad route to timeout");
2833 	if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) ==
2834 	    (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) {
2835 		rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt),
2836 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
2837 	}
2838 }
2839 
2840 /*
2841  * sysctl helper routine for the net.inet6.icmp6.nd6 nodes.  silly?
2842  */
2843 static int
2844 sysctl_net_inet6_icmp6_nd6(SYSCTLFN_ARGS)
2845 {
2846 
2847 	if (namelen != 0)
2848 		return (EINVAL);
2849 
2850 	return (nd6_sysctl(rnode->sysctl_num, oldp, oldlenp,
2851 	    (void*)newp, newlen));
2852 }
2853 
2854 SYSCTL_SETUP(sysctl_net_inet6_icmp6_setup,
2855 	     "sysctl net.inet6.icmp6 subtree setup")
2856 {
2857 
2858 	sysctl_createv(clog, 0, NULL, NULL,
2859 		       CTLFLAG_PERMANENT,
2860 		       CTLTYPE_NODE, "net", NULL,
2861 		       NULL, 0, NULL, 0,
2862 		       CTL_NET, CTL_EOL);
2863 	sysctl_createv(clog, 0, NULL, NULL,
2864 		       CTLFLAG_PERMANENT,
2865 		       CTLTYPE_NODE, "inet6", NULL,
2866 		       NULL, 0, NULL, 0,
2867 		       CTL_NET, PF_INET6, CTL_EOL);
2868 	sysctl_createv(clog, 0, NULL, NULL,
2869 		       CTLFLAG_PERMANENT,
2870 		       CTLTYPE_NODE, "icmp6",
2871 		       SYSCTL_DESCR("ICMPv6 related settings"),
2872 		       NULL, 0, NULL, 0,
2873 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6, CTL_EOL);
2874 
2875 	sysctl_createv(clog, 0, NULL, NULL,
2876 		       CTLFLAG_PERMANENT,
2877 		       CTLTYPE_STRUCT, "stats",
2878 		       SYSCTL_DESCR("ICMPv6 transmission statistics"),
2879 		       NULL, 0, &icmp6stat, sizeof(icmp6stat),
2880 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2881 		       ICMPV6CTL_STATS, CTL_EOL);
2882 	sysctl_createv(clog, 0, NULL, NULL,
2883 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2884 		       CTLTYPE_INT, "rediraccept",
2885 		       SYSCTL_DESCR("Accept and process redirect messages"),
2886 		       NULL, 0, &icmp6_rediraccept, 0,
2887 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2888 		       ICMPV6CTL_REDIRACCEPT, CTL_EOL);
2889 	sysctl_createv(clog, 0, NULL, NULL,
2890 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2891 		       CTLTYPE_INT, "redirtimeout",
2892 		       SYSCTL_DESCR("Redirect generated route lifetime"),
2893 		       NULL, 0, &icmp6_redirtimeout, 0,
2894 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2895 		       ICMPV6CTL_REDIRTIMEOUT, CTL_EOL);
2896 #if 0 /* obsoleted */
2897 	sysctl_createv(clog, 0, NULL, NULL,
2898 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2899 		       CTLTYPE_INT, "errratelimit", NULL,
2900 		       NULL, 0, &icmp6_errratelimit, 0,
2901 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2902 		       ICMPV6CTL_ERRRATELIMIT, CTL_EOL);
2903 #endif
2904 	sysctl_createv(clog, 0, NULL, NULL,
2905 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2906 		       CTLTYPE_INT, "nd6_prune",
2907 		       SYSCTL_DESCR("Neighbor discovery prune interval"),
2908 		       NULL, 0, &nd6_prune, 0,
2909 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2910 		       ICMPV6CTL_ND6_PRUNE, CTL_EOL);
2911 	sysctl_createv(clog, 0, NULL, NULL,
2912 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2913 		       CTLTYPE_INT, "nd6_delay",
2914 		       SYSCTL_DESCR("First probe delay time"),
2915 		       NULL, 0, &nd6_delay, 0,
2916 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2917 		       ICMPV6CTL_ND6_DELAY, CTL_EOL);
2918 	sysctl_createv(clog, 0, NULL, NULL,
2919 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2920 		       CTLTYPE_INT, "nd6_umaxtries",
2921 		       SYSCTL_DESCR("Number of unicast discovery attempts"),
2922 		       NULL, 0, &nd6_umaxtries, 0,
2923 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2924 		       ICMPV6CTL_ND6_UMAXTRIES, CTL_EOL);
2925 	sysctl_createv(clog, 0, NULL, NULL,
2926 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2927 		       CTLTYPE_INT, "nd6_mmaxtries",
2928 		       SYSCTL_DESCR("Number of multicast discovery attempts"),
2929 		       NULL, 0, &nd6_mmaxtries, 0,
2930 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2931 		       ICMPV6CTL_ND6_MMAXTRIES, CTL_EOL);
2932 	sysctl_createv(clog, 0, NULL, NULL,
2933 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2934 		       CTLTYPE_INT, "nd6_useloopback",
2935 		       SYSCTL_DESCR("Use loopback interface for local traffic"),
2936 		       NULL, 0, &nd6_useloopback, 0,
2937 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2938 		       ICMPV6CTL_ND6_USELOOPBACK, CTL_EOL);
2939 #if 0 /* obsoleted */
2940 	sysctl_createv(clog, 0, NULL, NULL,
2941 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2942 		       CTLTYPE_INT, "nd6_proxyall", NULL,
2943 		       NULL, 0, &nd6_proxyall, 0,
2944 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2945 		       ICMPV6CTL_ND6_PROXYALL, CTL_EOL);
2946 #endif
2947 	sysctl_createv(clog, 0, NULL, NULL,
2948 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2949 		       CTLTYPE_INT, "nodeinfo",
2950 		       SYSCTL_DESCR("Respond to node information requests"),
2951 		       NULL, 0, &icmp6_nodeinfo, 0,
2952 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2953 		       ICMPV6CTL_NODEINFO, CTL_EOL);
2954 	sysctl_createv(clog, 0, NULL, NULL,
2955 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2956 		       CTLTYPE_INT, "errppslimit",
2957 		       SYSCTL_DESCR("Maximum ICMP errors sent per second"),
2958 		       NULL, 0, &icmp6errppslim, 0,
2959 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2960 		       ICMPV6CTL_ERRPPSLIMIT, CTL_EOL);
2961 	sysctl_createv(clog, 0, NULL, NULL,
2962 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2963 		       CTLTYPE_INT, "nd6_maxnudhint",
2964 		       SYSCTL_DESCR("Maximum neighbor unreachable hint count"),
2965 		       NULL, 0, &nd6_maxnudhint, 0,
2966 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2967 		       ICMPV6CTL_ND6_MAXNUDHINT, CTL_EOL);
2968 	sysctl_createv(clog, 0, NULL, NULL,
2969 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2970 		       CTLTYPE_INT, "mtudisc_hiwat",
2971 		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
2972 		       NULL, 0, &icmp6_mtudisc_hiwat, 0,
2973 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2974 		       ICMPV6CTL_MTUDISC_HIWAT, CTL_EOL);
2975 	sysctl_createv(clog, 0, NULL, NULL,
2976 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2977 		       CTLTYPE_INT, "mtudisc_lowat",
2978 		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
2979 		       NULL, 0, &icmp6_mtudisc_lowat, 0,
2980 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2981 		       ICMPV6CTL_MTUDISC_LOWAT, CTL_EOL);
2982 	sysctl_createv(clog, 0, NULL, NULL,
2983 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2984 		       CTLTYPE_INT, "nd6_debug",
2985 		       SYSCTL_DESCR("Enable neighbor discovery debug output"),
2986 		       NULL, 0, &nd6_debug, 0,
2987 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2988 		       ICMPV6CTL_ND6_DEBUG, CTL_EOL);
2989 	sysctl_createv(clog, 0, NULL, NULL,
2990 		       CTLFLAG_PERMANENT,
2991 		       CTLTYPE_STRUCT, "nd6_drlist",
2992 		       SYSCTL_DESCR("Default router list"),
2993 		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
2994 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2995 		       ICMPV6CTL_ND6_DRLIST, CTL_EOL);
2996 	sysctl_createv(clog, 0, NULL, NULL,
2997 		       CTLFLAG_PERMANENT,
2998 		       CTLTYPE_STRUCT, "nd6_prlist",
2999 		       SYSCTL_DESCR("Prefix list"),
3000 		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
3001 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3002 		       ICMPV6CTL_ND6_PRLIST, CTL_EOL);
3003 }
3004