xref: /netbsd-src/sys/netinet6/icmp6.c (revision ce2c90c7c172d95d2402a5b3d96d8f8e6d138a21)
1 /*	$NetBSD: icmp6.c,v 1.122 2006/10/12 01:32:38 christos Exp $	*/
2 /*	$KAME: icmp6.c,v 1.217 2001/06/20 15:03:29 jinmei Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1988, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
62  */
63 
64 #include <sys/cdefs.h>
65 __KERNEL_RCSID(0, "$NetBSD: icmp6.c,v 1.122 2006/10/12 01:32:38 christos Exp $");
66 
67 #include "opt_inet.h"
68 #include "opt_ipsec.h"
69 
70 #include <sys/param.h>
71 #include <sys/systm.h>
72 #include <sys/malloc.h>
73 #include <sys/mbuf.h>
74 #include <sys/protosw.h>
75 #include <sys/socket.h>
76 #include <sys/socketvar.h>
77 #include <sys/time.h>
78 #include <sys/kernel.h>
79 #include <sys/syslog.h>
80 #include <sys/domain.h>
81 #include <sys/sysctl.h>
82 
83 #include <net/if.h>
84 #include <net/route.h>
85 #include <net/if_dl.h>
86 #include <net/if_types.h>
87 
88 #include <netinet/in.h>
89 #include <netinet/in_var.h>
90 #include <netinet/ip6.h>
91 #include <netinet6/ip6_var.h>
92 #include <netinet/icmp6.h>
93 #include <netinet6/mld6_var.h>
94 #include <netinet6/in6_pcb.h>
95 #include <netinet6/nd6.h>
96 #include <netinet6/in6_ifattach.h>
97 #include <netinet6/ip6protosw.h>
98 #include <netinet6/scope6_var.h>
99 
100 #ifdef IPSEC
101 #include <netinet6/ipsec.h>
102 #include <netkey/key.h>
103 #endif
104 
105 #include "faith.h"
106 #if defined(NFAITH) && 0 < NFAITH
107 #include <net/if_faith.h>
108 #endif
109 
110 #include <net/net_osdep.h>
111 
112 extern struct domain inet6domain;
113 
114 struct icmp6stat icmp6stat;
115 
116 extern struct inpcbtable raw6cbtable;
117 extern int icmp6errppslim;
118 static int icmp6errpps_count = 0;
119 static struct timeval icmp6errppslim_last;
120 extern int icmp6_nodeinfo;
121 
122 /*
123  * List of callbacks to notify when Path MTU changes are made.
124  */
125 struct icmp6_mtudisc_callback {
126 	LIST_ENTRY(icmp6_mtudisc_callback) mc_list;
127 	void (*mc_func) __P((struct in6_addr *));
128 };
129 
130 LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks =
131     LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks);
132 
133 static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL;
134 extern int pmtu_expire;
135 
136 /* XXX do these values make any sense? */
137 static int icmp6_mtudisc_hiwat = 1280;
138 static int icmp6_mtudisc_lowat = 256;
139 
140 /*
141  * keep track of # of redirect routes.
142  */
143 static struct rttimer_queue *icmp6_redirect_timeout_q = NULL;
144 
145 /* XXX experimental, turned off */
146 static int icmp6_redirect_hiwat = -1;
147 static int icmp6_redirect_lowat = -1;
148 
149 static void icmp6_errcount __P((struct icmp6errstat *, int, int));
150 static int icmp6_rip6_input __P((struct mbuf **, int));
151 static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int));
152 static const char *icmp6_redirect_diag __P((struct in6_addr *,
153 	struct in6_addr *, struct in6_addr *));
154 static struct mbuf *ni6_input __P((struct mbuf *, int));
155 static struct mbuf *ni6_nametodns __P((const char *, int, int));
156 static int ni6_dnsmatch __P((const char *, int, const char *, int));
157 static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
158 			  struct ifnet **, char *));
159 static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
160 				struct ifnet *, int));
161 static int icmp6_notify_error __P((struct mbuf *, int, int, int));
162 static struct rtentry *icmp6_mtudisc_clone __P((struct sockaddr *));
163 static void icmp6_mtudisc_timeout __P((struct rtentry *, struct rttimer *));
164 static void icmp6_redirect_timeout __P((struct rtentry *, struct rttimer *));
165 
166 
167 void
168 icmp6_init()
169 {
170 	mld_init();
171 	icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire);
172 	icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout);
173 }
174 
175 static void
176 icmp6_errcount(stat, type, code)
177 	struct icmp6errstat *stat;
178 	int type, code;
179 {
180 	switch (type) {
181 	case ICMP6_DST_UNREACH:
182 		switch (code) {
183 		case ICMP6_DST_UNREACH_NOROUTE:
184 			stat->icp6errs_dst_unreach_noroute++;
185 			return;
186 		case ICMP6_DST_UNREACH_ADMIN:
187 			stat->icp6errs_dst_unreach_admin++;
188 			return;
189 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
190 			stat->icp6errs_dst_unreach_beyondscope++;
191 			return;
192 		case ICMP6_DST_UNREACH_ADDR:
193 			stat->icp6errs_dst_unreach_addr++;
194 			return;
195 		case ICMP6_DST_UNREACH_NOPORT:
196 			stat->icp6errs_dst_unreach_noport++;
197 			return;
198 		}
199 		break;
200 	case ICMP6_PACKET_TOO_BIG:
201 		stat->icp6errs_packet_too_big++;
202 		return;
203 	case ICMP6_TIME_EXCEEDED:
204 		switch (code) {
205 		case ICMP6_TIME_EXCEED_TRANSIT:
206 			stat->icp6errs_time_exceed_transit++;
207 			return;
208 		case ICMP6_TIME_EXCEED_REASSEMBLY:
209 			stat->icp6errs_time_exceed_reassembly++;
210 			return;
211 		}
212 		break;
213 	case ICMP6_PARAM_PROB:
214 		switch (code) {
215 		case ICMP6_PARAMPROB_HEADER:
216 			stat->icp6errs_paramprob_header++;
217 			return;
218 		case ICMP6_PARAMPROB_NEXTHEADER:
219 			stat->icp6errs_paramprob_nextheader++;
220 			return;
221 		case ICMP6_PARAMPROB_OPTION:
222 			stat->icp6errs_paramprob_option++;
223 			return;
224 		}
225 		break;
226 	case ND_REDIRECT:
227 		stat->icp6errs_redirect++;
228 		return;
229 	}
230 	stat->icp6errs_unknown++;
231 }
232 
233 /*
234  * Register a Path MTU Discovery callback.
235  */
236 void
237 icmp6_mtudisc_callback_register(func)
238 	void (*func) __P((struct in6_addr *));
239 {
240 	struct icmp6_mtudisc_callback *mc;
241 
242 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
243 	     mc = LIST_NEXT(mc, mc_list)) {
244 		if (mc->mc_func == func)
245 			return;
246 	}
247 
248 	mc = malloc(sizeof(*mc), M_PCB, M_NOWAIT);
249 	if (mc == NULL)
250 		panic("icmp6_mtudisc_callback_register");
251 
252 	mc->mc_func = func;
253 	LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list);
254 }
255 
256 /*
257  * A wrapper function for icmp6_error() necessary when the erroneous packet
258  * may not contain enough scope zone information.
259  */
260 void
261 icmp6_error2(m, type, code, param, ifp)
262 	struct mbuf *m;
263 	int type, code, param;
264 	struct ifnet *ifp;
265 {
266 	struct ip6_hdr *ip6;
267 
268 	if (ifp == NULL)
269 		return;
270 
271 	if (m->m_len < sizeof(struct ip6_hdr)) {
272 		m = m_pullup(m, sizeof(struct ip6_hdr));
273 		if (m == NULL)
274 			return;
275 	}
276 
277 	ip6 = mtod(m, struct ip6_hdr *);
278 
279 	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
280 		return;
281 	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
282 		return;
283 
284 	icmp6_error(m, type, code, param);
285 }
286 
287 /*
288  * Generate an error packet of type error in response to bad IP6 packet.
289  */
290 void
291 icmp6_error(m, type, code, param)
292 	struct mbuf *m;
293 	int type, code, param;
294 {
295 	struct ip6_hdr *oip6, *nip6;
296 	struct icmp6_hdr *icmp6;
297 	u_int preplen;
298 	int off;
299 	int nxt;
300 
301 	icmp6stat.icp6s_error++;
302 
303 	/* count per-type-code statistics */
304 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
305 
306 	if (m->m_flags & M_DECRYPTED) {
307 		icmp6stat.icp6s_canterror++;
308 		goto freeit;
309 	}
310 
311 	if (m->m_len < sizeof(struct ip6_hdr)) {
312 		m = m_pullup(m, sizeof(struct ip6_hdr));
313 		if (m == NULL)
314 			return;
315 	}
316 	oip6 = mtod(m, struct ip6_hdr *);
317 
318 	/*
319 	 * If the destination address of the erroneous packet is a multicast
320 	 * address, or the packet was sent using link-layer multicast,
321 	 * we should basically suppress sending an error (RFC 2463, Section
322 	 * 2.4).
323 	 * We have two exceptions (the item e.2 in that section):
324 	 * - the Pakcet Too Big message can be sent for path MTU discovery.
325 	 * - the Parameter Problem Message that can be allowed an icmp6 error
326 	 *   in the option type field.  This check has been done in
327 	 *   ip6_unknown_opt(), so we can just check the type and code.
328 	 */
329 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
330 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
331 	    (type != ICMP6_PACKET_TOO_BIG &&
332 	     (type != ICMP6_PARAM_PROB ||
333 	      code != ICMP6_PARAMPROB_OPTION)))
334 		goto freeit;
335 
336 	/*
337 	 * RFC 2463, 2.4 (e.5): source address check.
338 	 * XXX: the case of anycast source?
339 	 */
340 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
341 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
342 		goto freeit;
343 
344 	/*
345 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
346 	 * don't do it.
347 	 */
348 	nxt = -1;
349 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
350 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
351 		struct icmp6_hdr *icp;
352 
353 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
354 			sizeof(*icp));
355 		if (icp == NULL) {
356 			icmp6stat.icp6s_tooshort++;
357 			return;
358 		}
359 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
360 		    icp->icmp6_type == ND_REDIRECT) {
361 			/*
362 			 * ICMPv6 error
363 			 * Special case: for redirect (which is
364 			 * informational) we must not send icmp6 error.
365 			 */
366 			icmp6stat.icp6s_canterror++;
367 			goto freeit;
368 		} else {
369 			/* ICMPv6 informational - send the error */
370 		}
371 	}
372 #if 0 /* controversial */
373 	else if (off >= 0 && nxt == IPPROTO_ESP) {
374 		/*
375 		 * It could be ICMPv6 error inside ESP.  Take a safer side,
376 		 * don't respond.
377 		 */
378 		icmp6stat.icp6s_canterror++;
379 		goto freeit;
380 	}
381 #endif
382 	else {
383 		/* non-ICMPv6 - send the error */
384 	}
385 
386 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
387 
388 	/* Finally, do rate limitation check. */
389 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
390 		icmp6stat.icp6s_toofreq++;
391 		goto freeit;
392 	}
393 
394 	/*
395 	 * OK, ICMP6 can be generated.
396 	 */
397 
398 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
399 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
400 
401 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
402 	M_PREPEND(m, preplen, M_DONTWAIT);
403 	if (m && m->m_len < preplen)
404 		m = m_pullup(m, preplen);
405 	if (m == NULL) {
406 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
407 		return;
408 	}
409 
410 	nip6 = mtod(m, struct ip6_hdr *);
411 	nip6->ip6_src  = oip6->ip6_src;
412 	nip6->ip6_dst  = oip6->ip6_dst;
413 
414 	in6_clearscope(&oip6->ip6_src);
415 	in6_clearscope(&oip6->ip6_dst);
416 
417 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
418 	icmp6->icmp6_type = type;
419 	icmp6->icmp6_code = code;
420 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
421 
422 	/*
423 	 * icmp6_reflect() is designed to be in the input path.
424 	 * icmp6_error() can be called from both input and output path,
425 	 * and if we are in output path rcvif could contain bogus value.
426 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
427 	 * information in ip header (nip6).
428 	 */
429 	m->m_pkthdr.rcvif = NULL;
430 
431 	icmp6stat.icp6s_outhist[type]++;
432 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
433 
434 	return;
435 
436   freeit:
437 	/*
438 	 * If we can't tell whether or not we can generate ICMP6, free it.
439 	 */
440 	m_freem(m);
441 }
442 
443 /*
444  * Process a received ICMP6 message.
445  */
446 int
447 icmp6_input(struct mbuf **mp, int *offp, int proto __unused)
448 {
449 	struct mbuf *m = *mp, *n;
450 	struct ip6_hdr *ip6, *nip6;
451 	struct icmp6_hdr *icmp6, *nicmp6;
452 	int off = *offp;
453 	int icmp6len = m->m_pkthdr.len - *offp;
454 	int code, sum, noff;
455 
456 #define ICMP6_MAXLEN (sizeof(*nip6) + sizeof(*nicmp6) + 4)
457 	KASSERT(ICMP6_MAXLEN < MCLBYTES);
458 	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
459 
460 	/*
461 	 * Locate icmp6 structure in mbuf, and check
462 	 * that not corrupted and of at least minimum length
463 	 */
464 
465 	ip6 = mtod(m, struct ip6_hdr *);
466 	if (icmp6len < sizeof(struct icmp6_hdr)) {
467 		icmp6stat.icp6s_tooshort++;
468 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
469 		goto freeit;
470 	}
471 
472 	/*
473 	 * calculate the checksum
474 	 */
475 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
476 	if (icmp6 == NULL) {
477 		icmp6stat.icp6s_tooshort++;
478 		/* m is invalid */
479 		/*icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);*/
480 		return IPPROTO_DONE;
481 	}
482 	KASSERT(IP6_HDR_ALIGNED_P(icmp6));
483 	code = icmp6->icmp6_code;
484 
485 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
486 		nd6log((LOG_ERR,
487 		    "ICMP6 checksum error(%d|%x) %s\n",
488 		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
489 		icmp6stat.icp6s_checksum++;
490 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
491 		goto freeit;
492 	}
493 
494 #if defined(NFAITH) && 0 < NFAITH
495 	if (faithprefix(&ip6->ip6_dst)) {
496 		/*
497 		 * Deliver very specific ICMP6 type only.
498 		 * This is important to deliver TOOBIG.  Otherwise PMTUD
499 		 * will not work.
500 		 */
501 		switch (icmp6->icmp6_type) {
502 		case ICMP6_DST_UNREACH:
503 		case ICMP6_PACKET_TOO_BIG:
504 		case ICMP6_TIME_EXCEEDED:
505 			break;
506 		default:
507 			goto freeit;
508 		}
509 	}
510 #endif
511 
512 	icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
513 
514 	switch (icmp6->icmp6_type) {
515 	case ICMP6_DST_UNREACH:
516 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
517 		switch (code) {
518 		case ICMP6_DST_UNREACH_NOROUTE:
519 			code = PRC_UNREACH_NET;
520 			break;
521 		case ICMP6_DST_UNREACH_ADMIN:
522 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
523 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
524 			break;
525 		case ICMP6_DST_UNREACH_ADDR:
526 			code = PRC_HOSTDEAD;
527 			break;
528 #ifdef COMPAT_RFC1885
529 		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
530 			code = PRC_UNREACH_SRCFAIL;
531 			break;
532 #else
533 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
534 			/* I mean "source address was incorrect." */
535 			code = PRC_UNREACH_NET;
536 			break;
537 #endif
538 		case ICMP6_DST_UNREACH_NOPORT:
539 			code = PRC_UNREACH_PORT;
540 			break;
541 		default:
542 			goto badcode;
543 		}
544 		goto deliver;
545 
546 	case ICMP6_PACKET_TOO_BIG:
547 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
548 
549 		code = PRC_MSGSIZE;
550 
551 		/*
552 		 * Updating the path MTU will be done after examining
553 		 * intermediate extension headers.
554 		 */
555 		goto deliver;
556 
557 	case ICMP6_TIME_EXCEEDED:
558 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
559 		switch (code) {
560 		case ICMP6_TIME_EXCEED_TRANSIT:
561 			code = PRC_TIMXCEED_INTRANS;
562 			break;
563 		case ICMP6_TIME_EXCEED_REASSEMBLY:
564 			code = PRC_TIMXCEED_REASS;
565 			break;
566 		default:
567 			goto badcode;
568 		}
569 		goto deliver;
570 
571 	case ICMP6_PARAM_PROB:
572 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
573 		switch (code) {
574 		case ICMP6_PARAMPROB_NEXTHEADER:
575 			code = PRC_UNREACH_PROTOCOL;
576 			break;
577 		case ICMP6_PARAMPROB_HEADER:
578 		case ICMP6_PARAMPROB_OPTION:
579 			code = PRC_PARAMPROB;
580 			break;
581 		default:
582 			goto badcode;
583 		}
584 		goto deliver;
585 
586 	case ICMP6_ECHO_REQUEST:
587 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
588 		if (code != 0)
589 			goto badcode;
590 		/*
591 		 * Copy mbuf to send to two data paths: userland socket(s),
592 		 * and to the querier (echo reply).
593 		 * m: a copy for socket, n: a copy for querier
594 		 *
595 		 * If the first mbuf is shared, or the first mbuf is too short,
596 		 * copy the first part of the data into a fresh mbuf.
597 		 * Otherwise, we will wrongly overwrite both copies.
598 		 */
599 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
600 			/* Give up local */
601 			n = m;
602 			m = NULL;
603 		} else if (M_READONLY(n) ||
604 		    n->m_len < off + sizeof(struct icmp6_hdr)) {
605 			struct mbuf *n0 = n;
606 
607 			/*
608 			 * Prepare an internal mbuf.  m_pullup() doesn't
609 			 * always copy the length we specified.
610 			 */
611 			if ((n = m_dup(n0, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
612 				/* Give up local */
613 				n = m;
614 				m = NULL;
615 			}
616 			m_freem(n0);
617 		}
618 		nip6 = mtod(n, struct ip6_hdr *);
619 		nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
620 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
621 		nicmp6->icmp6_code = 0;
622 		if (n) {
623 			icmp6stat.icp6s_reflect++;
624 			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
625 			icmp6_reflect(n, off);
626 		}
627 		if (!m)
628 			goto freeit;
629 		break;
630 
631 	case ICMP6_ECHO_REPLY:
632 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
633 		if (code != 0)
634 			goto badcode;
635 		break;
636 
637 	case MLD_LISTENER_QUERY:
638 	case MLD_LISTENER_REPORT:
639 		if (icmp6len < sizeof(struct mld_hdr))
640 			goto badlen;
641 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
642 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
643 		else
644 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
645 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
646 			/* give up local */
647 			mld_input(m, off);
648 			m = NULL;
649 			goto freeit;
650 		}
651 		mld_input(n, off);
652 		/* m stays. */
653 		break;
654 
655 	case MLD_LISTENER_DONE:
656 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
657 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
658 			goto badlen;
659 		break;		/* nothing to be done in kernel */
660 
661 	case MLD_MTRACE_RESP:
662 	case MLD_MTRACE:
663 		/* XXX: these two are experimental.  not officially defined. */
664 		/* XXX: per-interface statistics? */
665 		break;		/* just pass it to applications */
666 
667 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
668 	    {
669 		enum { WRU, FQDN } mode;
670 
671 		if (!icmp6_nodeinfo)
672 			break;
673 
674 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
675 			mode = WRU;
676 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
677 			mode = FQDN;
678 		else
679 			goto badlen;
680 
681 		if (mode == FQDN) {
682 			n = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
683 			if (n)
684 				n = ni6_input(n, off);
685 			/* XXX meaningless if n == NULL */
686 			noff = sizeof(struct ip6_hdr);
687 		} else {
688 			u_char *p;
689 			int maxhlen;
690 
691 			if ((icmp6_nodeinfo & 5) != 5)
692 				break;
693 
694 			if (code != 0)
695 				goto badcode;
696 			MGETHDR(n, M_DONTWAIT, m->m_type);
697 			if (n && ICMP6_MAXLEN > MHLEN) {
698 				MCLGET(n, M_DONTWAIT);
699 				if ((n->m_flags & M_EXT) == 0) {
700 					m_free(n);
701 					n = NULL;
702 				}
703 			}
704 			if (n == NULL) {
705 				/* Give up remote */
706 				break;
707 			}
708 			n->m_pkthdr.rcvif = NULL;
709 			n->m_len = 0;
710 			maxhlen = M_TRAILINGSPACE(n) - ICMP6_MAXLEN;
711 			if (maxhlen > hostnamelen)
712 				maxhlen = hostnamelen;
713 			/*
714 			 * Copy IPv6 and ICMPv6 only.
715 			 */
716 			nip6 = mtod(n, struct ip6_hdr *);
717 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
718 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
719 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
720 			p = (u_char *)(nicmp6 + 1);
721 			bzero(p, 4);
722 			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
723 			noff = sizeof(struct ip6_hdr);
724 			M_COPY_PKTHDR(n, m); /* just for rcvif */
725 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
726 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
727 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
728 			nicmp6->icmp6_code = 0;
729 		}
730 #undef hostnamelen
731 		if (n) {
732 			icmp6stat.icp6s_reflect++;
733 			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
734 			icmp6_reflect(n, noff);
735 		}
736 		break;
737 	    }
738 
739 	case ICMP6_WRUREPLY:
740 		if (code != 0)
741 			goto badcode;
742 		break;
743 
744 	case ND_ROUTER_SOLICIT:
745 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
746 		if (code != 0)
747 			goto badcode;
748 		if (icmp6len < sizeof(struct nd_router_solicit))
749 			goto badlen;
750 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
751 			/* give up local */
752 			nd6_rs_input(m, off, icmp6len);
753 			m = NULL;
754 			goto freeit;
755 		}
756 		nd6_rs_input(n, off, icmp6len);
757 		/* m stays. */
758 		break;
759 
760 	case ND_ROUTER_ADVERT:
761 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
762 		if (code != 0)
763 			goto badcode;
764 		if (icmp6len < sizeof(struct nd_router_advert))
765 			goto badlen;
766 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
767 			/* give up local */
768 			nd6_ra_input(m, off, icmp6len);
769 			m = NULL;
770 			goto freeit;
771 		}
772 		nd6_ra_input(n, off, icmp6len);
773 		/* m stays. */
774 		break;
775 
776 	case ND_NEIGHBOR_SOLICIT:
777 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
778 		if (code != 0)
779 			goto badcode;
780 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
781 			goto badlen;
782 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
783 			/* give up local */
784 			nd6_ns_input(m, off, icmp6len);
785 			m = NULL;
786 			goto freeit;
787 		}
788 		nd6_ns_input(n, off, icmp6len);
789 		/* m stays. */
790 		break;
791 
792 	case ND_NEIGHBOR_ADVERT:
793 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
794 		if (code != 0)
795 			goto badcode;
796 		if (icmp6len < sizeof(struct nd_neighbor_advert))
797 			goto badlen;
798 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
799 			/* give up local */
800 			nd6_na_input(m, off, icmp6len);
801 			m = NULL;
802 			goto freeit;
803 		}
804 		nd6_na_input(n, off, icmp6len);
805 		/* m stays. */
806 		break;
807 
808 	case ND_REDIRECT:
809 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
810 		if (code != 0)
811 			goto badcode;
812 		if (icmp6len < sizeof(struct nd_redirect))
813 			goto badlen;
814 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
815 			/* give up local */
816 			icmp6_redirect_input(m, off);
817 			m = NULL;
818 			goto freeit;
819 		}
820 		icmp6_redirect_input(n, off);
821 		/* m stays. */
822 		break;
823 
824 	case ICMP6_ROUTER_RENUMBERING:
825 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
826 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
827 			goto badcode;
828 		if (icmp6len < sizeof(struct icmp6_router_renum))
829 			goto badlen;
830 		break;
831 
832 	default:
833 		nd6log((LOG_DEBUG,
834 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
835 		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
836 		    ip6_sprintf(&ip6->ip6_dst),
837 		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
838 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
839 			/* ICMPv6 error: MUST deliver it by spec... */
840 			code = PRC_NCMDS;
841 			/* deliver */
842 		} else {
843 			/* ICMPv6 informational: MUST not deliver */
844 			break;
845 		}
846 	deliver:
847 		if (icmp6_notify_error(m, off, icmp6len, code)) {
848 			/* In this case, m should've been freed. */
849 			return (IPPROTO_DONE);
850 		}
851 		break;
852 
853 	badcode:
854 		icmp6stat.icp6s_badcode++;
855 		break;
856 
857 	badlen:
858 		icmp6stat.icp6s_badlen++;
859 		break;
860 	}
861 
862 	/* deliver the packet to appropriate sockets */
863 	icmp6_rip6_input(&m, *offp);
864 
865 	return IPPROTO_DONE;
866 
867  freeit:
868 	m_freem(m);
869 	return IPPROTO_DONE;
870 }
871 
872 static int
873 icmp6_notify_error(m, off, icmp6len, code)
874 	struct mbuf *m;
875 	int off, icmp6len, code;
876 {
877 	struct icmp6_hdr *icmp6;
878 	struct ip6_hdr *eip6;
879 	u_int32_t notifymtu;
880 	struct sockaddr_in6 icmp6src, icmp6dst;
881 
882 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
883 		icmp6stat.icp6s_tooshort++;
884 		goto freeit;
885 	}
886 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
887 		       sizeof(*icmp6) + sizeof(struct ip6_hdr));
888 	if (icmp6 == NULL) {
889 		icmp6stat.icp6s_tooshort++;
890 		return (-1);
891 	}
892 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
893 
894 	/* Detect the upper level protocol */
895 	{
896 		void (*ctlfunc) __P((int, struct sockaddr *, void *));
897 		u_int8_t nxt = eip6->ip6_nxt;
898 		int eoff = off + sizeof(struct icmp6_hdr) +
899 			sizeof(struct ip6_hdr);
900 		struct ip6ctlparam ip6cp;
901 		struct in6_addr *finaldst = NULL;
902 		int icmp6type = icmp6->icmp6_type;
903 		struct ip6_frag *fh;
904 		struct ip6_rthdr *rth;
905 		struct ip6_rthdr0 *rth0;
906 		int rthlen;
907 
908 		while (1) { /* XXX: should avoid infinite loop explicitly? */
909 			struct ip6_ext *eh;
910 
911 			switch (nxt) {
912 			case IPPROTO_HOPOPTS:
913 			case IPPROTO_DSTOPTS:
914 			case IPPROTO_AH:
915 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
916 					       eoff, sizeof(*eh));
917 				if (eh == NULL) {
918 					icmp6stat.icp6s_tooshort++;
919 					return (-1);
920 				}
921 
922 				if (nxt == IPPROTO_AH)
923 					eoff += (eh->ip6e_len + 2) << 2;
924 				else
925 					eoff += (eh->ip6e_len + 1) << 3;
926 				nxt = eh->ip6e_nxt;
927 				break;
928 			case IPPROTO_ROUTING:
929 				/*
930 				 * When the erroneous packet contains a
931 				 * routing header, we should examine the
932 				 * header to determine the final destination.
933 				 * Otherwise, we can't properly update
934 				 * information that depends on the final
935 				 * destination (e.g. path MTU).
936 				 */
937 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
938 					       eoff, sizeof(*rth));
939 				if (rth == NULL) {
940 					icmp6stat.icp6s_tooshort++;
941 					return (-1);
942 				}
943 				rthlen = (rth->ip6r_len + 1) << 3;
944 				/*
945 				 * XXX: currently there is no
946 				 * officially defined type other
947 				 * than type-0.
948 				 * Note that if the segment left field
949 				 * is 0, all intermediate hops must
950 				 * have been passed.
951 				 */
952 				if (rth->ip6r_segleft &&
953 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
954 					int hops;
955 
956 					IP6_EXTHDR_GET(rth0,
957 						       struct ip6_rthdr0 *, m,
958 						       eoff, rthlen);
959 					if (rth0 == NULL) {
960 						icmp6stat.icp6s_tooshort++;
961 						return (-1);
962 					}
963 					/* just ignore a bogus header */
964 					if ((rth0->ip6r0_len % 2) == 0 &&
965 					    (hops = rth0->ip6r0_len/2))
966 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
967 				}
968 				eoff += rthlen;
969 				nxt = rth->ip6r_nxt;
970 				break;
971 			case IPPROTO_FRAGMENT:
972 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
973 					       eoff, sizeof(*fh));
974 				if (fh == NULL) {
975 					icmp6stat.icp6s_tooshort++;
976 					return (-1);
977 				}
978 				/*
979 				 * Data after a fragment header is meaningless
980 				 * unless it is the first fragment, but
981 				 * we'll go to the notify label for path MTU
982 				 * discovery.
983 				 */
984 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
985 					goto notify;
986 
987 				eoff += sizeof(struct ip6_frag);
988 				nxt = fh->ip6f_nxt;
989 				break;
990 			default:
991 				/*
992 				 * This case includes ESP and the No Next
993 				 * Header.  In such cases going to the notify
994 				 * label does not have any meaning
995 				 * (i.e. ctlfunc will be NULL), but we go
996 				 * anyway since we might have to update
997 				 * path MTU information.
998 				 */
999 				goto notify;
1000 			}
1001 		}
1002 	  notify:
1003 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1004 			       sizeof(*icmp6) + sizeof(struct ip6_hdr));
1005 		if (icmp6 == NULL) {
1006 			icmp6stat.icp6s_tooshort++;
1007 			return (-1);
1008 		}
1009 
1010 		/*
1011 		 * retrieve parameters from the inner IPv6 header, and convert
1012 		 * them into sockaddr structures.
1013 		 * XXX: there is no guarantee that the source or destination
1014 		 * addresses of the inner packet are in the same scope zone as
1015 		 * the addresses of the icmp packet.  But there is no other
1016 		 * way to determine the zone.
1017 		 */
1018 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1019 
1020 		bzero(&icmp6dst, sizeof(icmp6dst));
1021 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1022 		icmp6dst.sin6_family = AF_INET6;
1023 		if (finaldst == NULL)
1024 			icmp6dst.sin6_addr = eip6->ip6_dst;
1025 		else
1026 			icmp6dst.sin6_addr = *finaldst;
1027 		if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1028 			goto freeit;
1029 		bzero(&icmp6src, sizeof(icmp6src));
1030 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1031 		icmp6src.sin6_family = AF_INET6;
1032 		icmp6src.sin6_addr = eip6->ip6_src;
1033 		if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1034 			goto freeit;
1035 		icmp6src.sin6_flowinfo =
1036 			(eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1037 
1038 		if (finaldst == NULL)
1039 			finaldst = &eip6->ip6_dst;
1040 		ip6cp.ip6c_m = m;
1041 		ip6cp.ip6c_icmp6 = icmp6;
1042 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1043 		ip6cp.ip6c_off = eoff;
1044 		ip6cp.ip6c_finaldst = finaldst;
1045 		ip6cp.ip6c_src = &icmp6src;
1046 		ip6cp.ip6c_nxt = nxt;
1047 
1048 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1049 			notifymtu = ntohl(icmp6->icmp6_mtu);
1050 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1051 		}
1052 
1053 		ctlfunc = (void (*) __P((int, struct sockaddr *, void *)))
1054 			(inet6sw[ip6_protox[nxt]].pr_ctlinput);
1055 		if (ctlfunc) {
1056 			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1057 					  &ip6cp);
1058 		}
1059 	}
1060 	return (0);
1061 
1062   freeit:
1063 	m_freem(m);
1064 	return (-1);
1065 }
1066 
1067 void
1068 icmp6_mtudisc_update(ip6cp, validated)
1069 	struct ip6ctlparam *ip6cp;
1070 	int validated;
1071 {
1072 	unsigned long rtcount;
1073 	struct icmp6_mtudisc_callback *mc;
1074 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1075 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1076 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1077 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1078 	struct rtentry *rt = NULL;
1079 	struct sockaddr_in6 sin6;
1080 
1081 	/*
1082 	 * allow non-validated cases if memory is plenty, to make traffic
1083 	 * from non-connected pcb happy.
1084 	 */
1085 	rtcount = rt_timer_count(icmp6_mtudisc_timeout_q);
1086 	if (validated) {
1087 		if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat)
1088 			return;
1089 		else if (0 <= icmp6_mtudisc_lowat &&
1090 		    rtcount > icmp6_mtudisc_lowat) {
1091 			/*
1092 			 * XXX nuke a victim, install the new one.
1093 			 */
1094 		}
1095 	} else {
1096 		if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat)
1097 			return;
1098 	}
1099 
1100 	bzero(&sin6, sizeof(sin6));
1101 	sin6.sin6_family = PF_INET6;
1102 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1103 	sin6.sin6_addr = *dst;
1104 	if (in6_setscope(&sin6.sin6_addr, m->m_pkthdr.rcvif, NULL))
1105 		return;
1106 
1107 	rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6);
1108 
1109 	if (rt && (rt->rt_flags & RTF_HOST) &&
1110 	    !(rt->rt_rmx.rmx_locks & RTV_MTU) &&
1111 	    (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) {
1112 		if (mtu < IN6_LINKMTU(rt->rt_ifp)) {
1113 			icmp6stat.icp6s_pmtuchg++;
1114 			rt->rt_rmx.rmx_mtu = mtu;
1115 		}
1116 	}
1117 	if (rt) { /* XXX: need braces to avoid conflict with else in RTFREE. */
1118 		RTFREE(rt);
1119 	}
1120 
1121 	/*
1122 	 * Notify protocols that the MTU for this destination
1123 	 * has changed.
1124 	 */
1125 	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
1126 	     mc = LIST_NEXT(mc, mc_list))
1127 		(*mc->mc_func)(&sin6.sin6_addr);
1128 }
1129 
1130 /*
1131  * Process a Node Information Query packet, based on
1132  * draft-ietf-ipngwg-icmp-name-lookups-07.
1133  *
1134  * Spec incompatibilities:
1135  * - IPv6 Subject address handling
1136  * - IPv4 Subject address handling support missing
1137  * - Proxy reply (answer even if it's not for me)
1138  * - joins NI group address at in6_ifattach() time only, does not cope
1139  *   with hostname changes by sethostname(3)
1140  */
1141 #ifndef offsetof		/* XXX */
1142 #define	offsetof(type, member)	((size_t)(&((type *)0)->member))
1143 #endif
1144 static struct mbuf *
1145 ni6_input(m, off)
1146 	struct mbuf *m;
1147 	int off;
1148 {
1149 	struct icmp6_nodeinfo *ni6, *nni6;
1150 	struct mbuf *n = NULL;
1151 	u_int16_t qtype;
1152 	int subjlen;
1153 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1154 	struct ni_reply_fqdn *fqdn;
1155 	int addrs;		/* for NI_QTYPE_NODEADDR */
1156 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1157 	struct sockaddr_in6 sin6; /* ip6_dst */
1158 	struct in6_addr in6_subj; /* subject address */
1159 	struct ip6_hdr *ip6;
1160 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1161 	char *subj = NULL;
1162 
1163 	ip6 = mtod(m, struct ip6_hdr *);
1164 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1165 	if (ni6 == NULL) {
1166 		/* m is already reclaimed */
1167 		return NULL;
1168 	}
1169 
1170 	/*
1171 	 * Validate IPv6 destination address.
1172 	 *
1173 	 * The Responder must discard the Query without further processing
1174 	 * unless it is one of the Responder's unicast or anycast addresses, or
1175 	 * a link-local scope multicast address which the Responder has joined.
1176 	 * [icmp-name-lookups-07, Section 4.]
1177 	 */
1178 	bzero(&sin6, sizeof(sin6));
1179 	sin6.sin6_family = AF_INET6;
1180 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1181 	bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1182 	/* XXX scopeid */
1183 	if (ifa_ifwithaddr((struct sockaddr *)&sin6))
1184 		; /* unicast/anycast, fine */
1185 	else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1186 		; /* link-local multicast, fine */
1187 	else
1188 		goto bad;
1189 
1190 	/* validate query Subject field. */
1191 	qtype = ntohs(ni6->ni_qtype);
1192 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1193 	switch (qtype) {
1194 	case NI_QTYPE_NOOP:
1195 	case NI_QTYPE_SUPTYPES:
1196 		/* 07 draft */
1197 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1198 			break;
1199 		/* FALLTHROUGH */
1200 	case NI_QTYPE_FQDN:
1201 	case NI_QTYPE_NODEADDR:
1202 	case NI_QTYPE_IPV4ADDR:
1203 		switch (ni6->ni_code) {
1204 		case ICMP6_NI_SUBJ_IPV6:
1205 #if ICMP6_NI_SUBJ_IPV6 != 0
1206 		case 0:
1207 #endif
1208 			/*
1209 			 * backward compatibility - try to accept 03 draft
1210 			 * format, where no Subject is present.
1211 			 */
1212 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1213 			    subjlen == 0) {
1214 				oldfqdn++;
1215 				break;
1216 			}
1217 #if ICMP6_NI_SUBJ_IPV6 != 0
1218 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1219 				goto bad;
1220 #endif
1221 
1222 			if (subjlen != sizeof(sin6.sin6_addr))
1223 				goto bad;
1224 
1225 			/*
1226 			 * Validate Subject address.
1227 			 *
1228 			 * Not sure what exactly "address belongs to the node"
1229 			 * means in the spec, is it just unicast, or what?
1230 			 *
1231 			 * At this moment we consider Subject address as
1232 			 * "belong to the node" if the Subject address equals
1233 			 * to the IPv6 destination address; validation for
1234 			 * IPv6 destination address should have done enough
1235 			 * check for us.
1236 			 *
1237 			 * We do not do proxy at this moment.
1238 			 */
1239 			/* m_pulldown instead of copy? */
1240 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1241 			    subjlen, (caddr_t)&in6_subj);
1242 			if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1243 				goto bad;
1244 
1245 			subj = (char *)&in6_subj;
1246 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1247 				break;
1248 
1249 			/*
1250 			 * XXX if we are to allow other cases, we should really
1251 			 * be careful about scope here.
1252 			 * basically, we should disallow queries toward IPv6
1253 			 * destination X with subject Y, if scope(X) > scope(Y).
1254 			 * if we allow scope(X) > scope(Y), it will result in
1255 			 * information leakage across scope boundary.
1256 			 */
1257 			goto bad;
1258 
1259 		case ICMP6_NI_SUBJ_FQDN:
1260 			/*
1261 			 * Validate Subject name with gethostname(3).
1262 			 *
1263 			 * The behavior may need some debate, since:
1264 			 * - we are not sure if the node has FQDN as
1265 			 *   hostname (returned by gethostname(3)).
1266 			 * - the code does wildcard match for truncated names.
1267 			 *   however, we are not sure if we want to perform
1268 			 *   wildcard match, if gethostname(3) side has
1269 			 *   truncated hostname.
1270 			 */
1271 			n = ni6_nametodns(hostname, hostnamelen, 0);
1272 			if (!n || n->m_next || n->m_len == 0)
1273 				goto bad;
1274 			IP6_EXTHDR_GET(subj, char *, m,
1275 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1276 			if (subj == NULL)
1277 				goto bad;
1278 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1279 					n->m_len)) {
1280 				goto bad;
1281 			}
1282 			m_freem(n);
1283 			n = NULL;
1284 			break;
1285 
1286 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1287 		default:
1288 			goto bad;
1289 		}
1290 		break;
1291 	}
1292 
1293 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1294 	switch (qtype) {
1295 	case NI_QTYPE_FQDN:
1296 		if ((icmp6_nodeinfo & 1) == 0)
1297 			goto bad;
1298 		break;
1299 	case NI_QTYPE_NODEADDR:
1300 	case NI_QTYPE_IPV4ADDR:
1301 		if ((icmp6_nodeinfo & 2) == 0)
1302 			goto bad;
1303 		break;
1304 	}
1305 
1306 	/* guess reply length */
1307 	switch (qtype) {
1308 	case NI_QTYPE_NOOP:
1309 		break;		/* no reply data */
1310 	case NI_QTYPE_SUPTYPES:
1311 		replylen += sizeof(u_int32_t);
1312 		break;
1313 	case NI_QTYPE_FQDN:
1314 		/* XXX will append an mbuf */
1315 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1316 		break;
1317 	case NI_QTYPE_NODEADDR:
1318 		addrs = ni6_addrs(ni6, m, &ifp, subj);
1319 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1320 					  sizeof(u_int32_t))) > MCLBYTES)
1321 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1322 		break;
1323 	case NI_QTYPE_IPV4ADDR:
1324 		/* unsupported - should respond with unknown Qtype? */
1325 		goto bad;
1326 	default:
1327 		/*
1328 		 * XXX: We must return a reply with the ICMP6 code
1329 		 * `unknown Qtype' in this case.  However we regard the case
1330 		 * as an FQDN query for backward compatibility.
1331 		 * Older versions set a random value to this field,
1332 		 * so it rarely varies in the defined qtypes.
1333 		 * But the mechanism is not reliable...
1334 		 * maybe we should obsolete older versions.
1335 		 */
1336 		qtype = NI_QTYPE_FQDN;
1337 		/* XXX will append an mbuf */
1338 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1339 		oldfqdn++;
1340 		break;
1341 	}
1342 
1343 	/* allocate an mbuf to reply. */
1344 	MGETHDR(n, M_DONTWAIT, m->m_type);
1345 	if (n == NULL) {
1346 		m_freem(m);
1347 		return (NULL);
1348 	}
1349 	M_MOVE_PKTHDR(n, m); /* just for rcvif */
1350 	if (replylen > MHLEN) {
1351 		if (replylen > MCLBYTES) {
1352 			/*
1353 			 * XXX: should we try to allocate more? But MCLBYTES
1354 			 * is probably much larger than IPV6_MMTU...
1355 			 */
1356 			goto bad;
1357 		}
1358 		MCLGET(n, M_DONTWAIT);
1359 		if ((n->m_flags & M_EXT) == 0) {
1360 			goto bad;
1361 		}
1362 	}
1363 	n->m_pkthdr.len = n->m_len = replylen;
1364 
1365 	/* copy mbuf header and IPv6 + Node Information base headers */
1366 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1367 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1368 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1369 
1370 	/* qtype dependent procedure */
1371 	switch (qtype) {
1372 	case NI_QTYPE_NOOP:
1373 		nni6->ni_code = ICMP6_NI_SUCCESS;
1374 		nni6->ni_flags = 0;
1375 		break;
1376 	case NI_QTYPE_SUPTYPES:
1377 	{
1378 		u_int32_t v;
1379 		nni6->ni_code = ICMP6_NI_SUCCESS;
1380 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1381 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1382 		v = (u_int32_t)htonl(0x0000000f);
1383 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1384 		break;
1385 	}
1386 	case NI_QTYPE_FQDN:
1387 		nni6->ni_code = ICMP6_NI_SUCCESS;
1388 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1389 						sizeof(struct ip6_hdr) +
1390 						sizeof(struct icmp6_nodeinfo));
1391 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1392 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1393 		/*
1394 		 * XXX do we really have FQDN in variable "hostname"?
1395 		 */
1396 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1397 		if (n->m_next == NULL)
1398 			goto bad;
1399 		/* XXX we assume that n->m_next is not a chain */
1400 		if (n->m_next->m_next != NULL)
1401 			goto bad;
1402 		n->m_pkthdr.len += n->m_next->m_len;
1403 		break;
1404 	case NI_QTYPE_NODEADDR:
1405 	{
1406 		int lenlim, copied;
1407 
1408 		nni6->ni_code = ICMP6_NI_SUCCESS;
1409 		n->m_pkthdr.len = n->m_len =
1410 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1411 		lenlim = M_TRAILINGSPACE(n);
1412 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1413 		/* XXX: reset mbuf length */
1414 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1415 			sizeof(struct icmp6_nodeinfo) + copied;
1416 		break;
1417 	}
1418 	default:
1419 		break;		/* XXX impossible! */
1420 	}
1421 
1422 	nni6->ni_type = ICMP6_NI_REPLY;
1423 	m_freem(m);
1424 	return (n);
1425 
1426   bad:
1427 	m_freem(m);
1428 	if (n)
1429 		m_freem(n);
1430 	return (NULL);
1431 }
1432 #undef hostnamelen
1433 
1434 #define isupper(x) ('A' <= (x) && (x) <= 'Z')
1435 #define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z'))
1436 #define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9'))
1437 #define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x))
1438 
1439 /*
1440  * make a mbuf with DNS-encoded string.  no compression support.
1441  *
1442  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1443  * treated as truncated name (two \0 at the end).  this is a wild guess.
1444  */
1445 static struct mbuf *
1446 ni6_nametodns(name, namelen, old)
1447 	const char *name;
1448 	int namelen;
1449 	int old;	/* return pascal string if non-zero */
1450 {
1451 	struct mbuf *m;
1452 	char *cp, *ep;
1453 	const char *p, *q;
1454 	int i, len, nterm;
1455 
1456 	if (old)
1457 		len = namelen + 1;
1458 	else
1459 		len = MCLBYTES;
1460 
1461 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1462 	MGET(m, M_DONTWAIT, MT_DATA);
1463 	if (m && len > MLEN) {
1464 		MCLGET(m, M_DONTWAIT);
1465 		if ((m->m_flags & M_EXT) == 0)
1466 			goto fail;
1467 	}
1468 	if (!m)
1469 		goto fail;
1470 	m->m_next = NULL;
1471 
1472 	if (old) {
1473 		m->m_len = len;
1474 		*mtod(m, char *) = namelen;
1475 		bcopy(name, mtod(m, char *) + 1, namelen);
1476 		return m;
1477 	} else {
1478 		m->m_len = 0;
1479 		cp = mtod(m, char *);
1480 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1481 
1482 		/* if not certain about my name, return empty buffer */
1483 		if (namelen == 0)
1484 			return m;
1485 
1486 		/*
1487 		 * guess if it looks like shortened hostname, or FQDN.
1488 		 * shortened hostname needs two trailing "\0".
1489 		 */
1490 		i = 0;
1491 		for (p = name; p < name + namelen; p++) {
1492 			if (*p && *p == '.')
1493 				i++;
1494 		}
1495 		if (i < 2)
1496 			nterm = 2;
1497 		else
1498 			nterm = 1;
1499 
1500 		p = name;
1501 		while (cp < ep && p < name + namelen) {
1502 			i = 0;
1503 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1504 				i++;
1505 			/* result does not fit into mbuf */
1506 			if (cp + i + 1 >= ep)
1507 				goto fail;
1508 			/*
1509 			 * DNS label length restriction, RFC1035 page 8.
1510 			 * "i == 0" case is included here to avoid returning
1511 			 * 0-length label on "foo..bar".
1512 			 */
1513 			if (i <= 0 || i >= 64)
1514 				goto fail;
1515 			*cp++ = i;
1516 			if (!isalpha(p[0]) || !isalnum(p[i - 1]))
1517 				goto fail;
1518 			while (i > 0) {
1519 				if (!isalnum(*p) && *p != '-')
1520 					goto fail;
1521 				if (isupper(*p)) {
1522 					*cp++ = tolower(*p);
1523 					p++;
1524 				} else
1525 					*cp++ = *p++;
1526 				i--;
1527 			}
1528 			p = q;
1529 			if (p < name + namelen && *p == '.')
1530 				p++;
1531 		}
1532 		/* termination */
1533 		if (cp + nterm >= ep)
1534 			goto fail;
1535 		while (nterm-- > 0)
1536 			*cp++ = '\0';
1537 		m->m_len = cp - mtod(m, char *);
1538 		return m;
1539 	}
1540 
1541 	panic("should not reach here");
1542 	/* NOTREACHED */
1543 
1544  fail:
1545 	if (m)
1546 		m_freem(m);
1547 	return NULL;
1548 }
1549 
1550 /*
1551  * check if two DNS-encoded string matches.  takes care of truncated
1552  * form (with \0\0 at the end).  no compression support.
1553  * XXX upper/lowercase match (see RFC2065)
1554  */
1555 static int
1556 ni6_dnsmatch(a, alen, b, blen)
1557 	const char *a;
1558 	int alen;
1559 	const char *b;
1560 	int blen;
1561 {
1562 	const char *a0, *b0;
1563 	int l;
1564 
1565 	/* simplest case - need validation? */
1566 	if (alen == blen && bcmp(a, b, alen) == 0)
1567 		return 1;
1568 
1569 	a0 = a;
1570 	b0 = b;
1571 
1572 	/* termination is mandatory */
1573 	if (alen < 2 || blen < 2)
1574 		return 0;
1575 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1576 		return 0;
1577 	alen--;
1578 	blen--;
1579 
1580 	while (a - a0 < alen && b - b0 < blen) {
1581 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1582 			return 0;
1583 
1584 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1585 			return 0;
1586 		/* we don't support compression yet */
1587 		if (a[0] >= 64 || b[0] >= 64)
1588 			return 0;
1589 
1590 		/* truncated case */
1591 		if (a[0] == 0 && a - a0 == alen - 1)
1592 			return 1;
1593 		if (b[0] == 0 && b - b0 == blen - 1)
1594 			return 1;
1595 		if (a[0] == 0 || b[0] == 0)
1596 			return 0;
1597 
1598 		if (a[0] != b[0])
1599 			return 0;
1600 		l = a[0];
1601 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1602 			return 0;
1603 		if (bcmp(a + 1, b + 1, l) != 0)
1604 			return 0;
1605 
1606 		a += 1 + l;
1607 		b += 1 + l;
1608 	}
1609 
1610 	if (a - a0 == alen && b - b0 == blen)
1611 		return 1;
1612 	else
1613 		return 0;
1614 }
1615 
1616 /*
1617  * calculate the number of addresses to be returned in the node info reply.
1618  */
1619 static int
1620 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m __unused,
1621     struct ifnet **ifpp, char *subj)
1622 {
1623 	struct ifnet *ifp;
1624 	struct in6_ifaddr *ifa6;
1625 	struct ifaddr *ifa;
1626 	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1627 	int addrs = 0, addrsofif, iffound = 0;
1628 	int niflags = ni6->ni_flags;
1629 
1630 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1631 		switch (ni6->ni_code) {
1632 		case ICMP6_NI_SUBJ_IPV6:
1633 			if (subj == NULL) /* must be impossible... */
1634 				return (0);
1635 			subj_ip6 = (struct sockaddr_in6 *)subj;
1636 			break;
1637 		default:
1638 			/*
1639 			 * XXX: we only support IPv6 subject address for
1640 			 * this Qtype.
1641 			 */
1642 			return (0);
1643 		}
1644 	}
1645 
1646 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
1647 	{
1648 		addrsofif = 0;
1649 		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1650 		     ifa = ifa->ifa_list.tqe_next)
1651 		{
1652 			if (ifa->ifa_addr->sa_family != AF_INET6)
1653 				continue;
1654 			ifa6 = (struct in6_ifaddr *)ifa;
1655 
1656 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1657 			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1658 					       &ifa6->ia_addr.sin6_addr))
1659 				iffound = 1;
1660 
1661 			/*
1662 			 * IPv4-mapped addresses can only be returned by a
1663 			 * Node Information proxy, since they represent
1664 			 * addresses of IPv4-only nodes, which perforce do
1665 			 * not implement this protocol.
1666 			 * [icmp-name-lookups-07, Section 5.4]
1667 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1668 			 * this function at this moment.
1669 			 */
1670 
1671 			/* What do we have to do about ::1? */
1672 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1673 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1674 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1675 					continue;
1676 				break;
1677 			case IPV6_ADDR_SCOPE_SITELOCAL:
1678 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1679 					continue;
1680 				break;
1681 			case IPV6_ADDR_SCOPE_GLOBAL:
1682 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1683 					continue;
1684 				break;
1685 			default:
1686 				continue;
1687 			}
1688 
1689 			/*
1690 			 * check if anycast is okay.
1691 			 * XXX: just experimental.  not in the spec.
1692 			 */
1693 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1694 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1695 				continue; /* we need only unicast addresses */
1696 
1697 			addrsofif++; /* count the address */
1698 		}
1699 		if (iffound) {
1700 			*ifpp = ifp;
1701 			return (addrsofif);
1702 		}
1703 
1704 		addrs += addrsofif;
1705 	}
1706 
1707 	return (addrs);
1708 }
1709 
1710 static int
1711 ni6_store_addrs(ni6, nni6, ifp0, resid)
1712 	struct icmp6_nodeinfo *ni6, *nni6;
1713 	struct ifnet *ifp0;
1714 	int resid;
1715 {
1716 	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1717 	struct in6_ifaddr *ifa6;
1718 	struct ifaddr *ifa;
1719 	struct ifnet *ifp_dep = NULL;
1720 	int copied = 0, allow_deprecated = 0;
1721 	u_char *cp = (u_char *)(nni6 + 1);
1722 	int niflags = ni6->ni_flags;
1723 	u_int32_t ltime;
1724 
1725 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1726 		return (0);	/* needless to copy */
1727 
1728   again:
1729 
1730 	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list))
1731 	{
1732 		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1733 		     ifa = ifa->ifa_list.tqe_next)
1734 		{
1735 			if (ifa->ifa_addr->sa_family != AF_INET6)
1736 				continue;
1737 			ifa6 = (struct in6_ifaddr *)ifa;
1738 
1739 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1740 			    allow_deprecated == 0) {
1741 				/*
1742 				 * prefererred address should be put before
1743 				 * deprecated addresses.
1744 				 */
1745 
1746 				/* record the interface for later search */
1747 				if (ifp_dep == NULL)
1748 					ifp_dep = ifp;
1749 
1750 				continue;
1751 			}
1752 			else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1753 				 allow_deprecated != 0)
1754 				continue; /* we now collect deprecated addrs */
1755 
1756 			/* What do we have to do about ::1? */
1757 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1758 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1759 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1760 					continue;
1761 				break;
1762 			case IPV6_ADDR_SCOPE_SITELOCAL:
1763 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1764 					continue;
1765 				break;
1766 			case IPV6_ADDR_SCOPE_GLOBAL:
1767 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1768 					continue;
1769 				break;
1770 			default:
1771 				continue;
1772 			}
1773 
1774 			/*
1775 			 * check if anycast is okay.
1776 			 * XXX: just experimental.  not in the spec.
1777 			 */
1778 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1779 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1780 				continue;
1781 
1782 			/* now we can copy the address */
1783 			if (resid < sizeof(struct in6_addr) +
1784 			    sizeof(u_int32_t)) {
1785 				/*
1786 				 * We give up much more copy.
1787 				 * Set the truncate flag and return.
1788 				 */
1789 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1790 				return (copied);
1791 			}
1792 
1793 			/*
1794 			 * Set the TTL of the address.
1795 			 * The TTL value should be one of the following
1796 			 * according to the specification:
1797 			 *
1798 			 * 1. The remaining lifetime of a DHCP lease on the
1799 			 *    address, or
1800 			 * 2. The remaining Valid Lifetime of a prefix from
1801 			 *    which the address was derived through Stateless
1802 			 *    Autoconfiguration.
1803 			 *
1804 			 * Note that we currently do not support stateful
1805 			 * address configuration by DHCPv6, so the former
1806 			 * case can't happen.
1807 			 *
1808 			 * TTL must be 2^31 > TTL >= 0.
1809 			 */
1810 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1811 				ltime = ND6_INFINITE_LIFETIME;
1812 			else {
1813 				if (ifa6->ia6_lifetime.ia6t_expire >
1814 				    time_second)
1815 					ltime = ifa6->ia6_lifetime.ia6t_expire -
1816 					    time_second;
1817 				else
1818 					ltime = 0;
1819 			}
1820 			if (ltime > 0x7fffffff)
1821 				ltime = 0x7fffffff;
1822 			ltime = htonl(ltime);
1823 
1824 			bcopy(&ltime, cp, sizeof(u_int32_t));
1825 			cp += sizeof(u_int32_t);
1826 
1827 			/* copy the address itself */
1828 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1829 			      sizeof(struct in6_addr));
1830 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1831 			cp += sizeof(struct in6_addr);
1832 
1833 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1834 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1835 		}
1836 		if (ifp0)	/* we need search only on the specified IF */
1837 			break;
1838 	}
1839 
1840 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1841 		ifp = ifp_dep;
1842 		allow_deprecated = 1;
1843 
1844 		goto again;
1845 	}
1846 
1847 	return (copied);
1848 }
1849 
1850 /*
1851  * XXX almost dup'ed code with rip6_input.
1852  */
1853 static int
1854 icmp6_rip6_input(mp, off)
1855 	struct	mbuf **mp;
1856 	int	off;
1857 {
1858 	struct mbuf *m = *mp;
1859 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1860 	struct inpcb_hdr *inph;
1861 	struct in6pcb *in6p;
1862 	struct in6pcb *last = NULL;
1863 	struct sockaddr_in6 rip6src;
1864 	struct icmp6_hdr *icmp6;
1865 	struct mbuf *opts = NULL;
1866 
1867 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1868 	if (icmp6 == NULL) {
1869 		/* m is already reclaimed */
1870 		return IPPROTO_DONE;
1871 	}
1872 
1873 	/*
1874 	 * XXX: the address may have embedded scope zone ID, which should be
1875 	 * hidden from applications.
1876 	 */
1877 	bzero(&rip6src, sizeof(rip6src));
1878 	rip6src.sin6_len = sizeof(struct sockaddr_in6);
1879 	rip6src.sin6_family = AF_INET6;
1880 	rip6src.sin6_addr = ip6->ip6_src;
1881 	if (sa6_recoverscope(&rip6src)) {
1882 		m_freem(m);
1883 		return (IPPROTO_DONE);
1884 	}
1885 
1886 	CIRCLEQ_FOREACH(inph, &raw6cbtable.inpt_queue, inph_queue) {
1887 		in6p = (struct in6pcb *)inph;
1888 		if (in6p->in6p_af != AF_INET6)
1889 			continue;
1890 		if (in6p->in6p_ip6.ip6_nxt != IPPROTO_ICMPV6)
1891 			continue;
1892 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1893 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1894 			continue;
1895 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1896 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1897 			continue;
1898 		if (in6p->in6p_icmp6filt
1899 		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1900 				 in6p->in6p_icmp6filt))
1901 			continue;
1902 		if (last) {
1903 			struct	mbuf *n;
1904 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1905 				if (last->in6p_flags & IN6P_CONTROLOPTS)
1906 					ip6_savecontrol(last, &opts, ip6, n);
1907 				/* strip intermediate headers */
1908 				m_adj(n, off);
1909 				if (sbappendaddr(&last->in6p_socket->so_rcv,
1910 						 (struct sockaddr *)&rip6src,
1911 						 n, opts) == 0) {
1912 					/* should notify about lost packet */
1913 					m_freem(n);
1914 					if (opts)
1915 						m_freem(opts);
1916 				} else
1917 					sorwakeup(last->in6p_socket);
1918 				opts = NULL;
1919 			}
1920 		}
1921 		last = in6p;
1922 	}
1923 	if (last) {
1924 		if (last->in6p_flags & IN6P_CONTROLOPTS)
1925 			ip6_savecontrol(last, &opts, ip6, m);
1926 		/* strip intermediate headers */
1927 		m_adj(m, off);
1928 		if (sbappendaddr(&last->in6p_socket->so_rcv,
1929 				(struct sockaddr *)&rip6src, m, opts) == 0) {
1930 			m_freem(m);
1931 			if (opts)
1932 				m_freem(opts);
1933 		} else
1934 			sorwakeup(last->in6p_socket);
1935 	} else {
1936 		m_freem(m);
1937 		ip6stat.ip6s_delivered--;
1938 	}
1939 	return IPPROTO_DONE;
1940 }
1941 
1942 /*
1943  * Reflect the ip6 packet back to the source.
1944  * OFF points to the icmp6 header, counted from the top of the mbuf.
1945  *
1946  * Note: RFC 1885 required that an echo reply should be truncated if it
1947  * did not fit in with (return) path MTU, and KAME code supported the
1948  * behavior.  However, as a clarification after the RFC, this limitation
1949  * was removed in a revised version of the spec, RFC 2463.  We had kept the
1950  * old behavior, with a (non-default) ifdef block, while the new version of
1951  * the spec was an internet-draft status, and even after the new RFC was
1952  * published.  But it would rather make sense to clean the obsoleted part
1953  * up, and to make the code simpler at this stage.
1954  */
1955 void
1956 icmp6_reflect(m, off)
1957 	struct	mbuf *m;
1958 	size_t off;
1959 {
1960 	struct ip6_hdr *ip6;
1961 	struct icmp6_hdr *icmp6;
1962 	struct in6_ifaddr *ia;
1963 	int plen;
1964 	int type, code;
1965 	struct ifnet *outif = NULL;
1966 	struct in6_addr origdst, *src = NULL;
1967 
1968 	/* too short to reflect */
1969 	if (off < sizeof(struct ip6_hdr)) {
1970 		nd6log((LOG_DEBUG,
1971 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
1972 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
1973 		    __FILE__, __LINE__));
1974 		goto bad;
1975 	}
1976 
1977 	/*
1978 	 * If there are extra headers between IPv6 and ICMPv6, strip
1979 	 * off that header first.
1980 	 */
1981 #ifdef DIAGNOSTIC
1982 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
1983 		panic("assumption failed in icmp6_reflect");
1984 #endif
1985 	if (off > sizeof(struct ip6_hdr)) {
1986 		size_t l;
1987 		struct ip6_hdr nip6;
1988 
1989 		l = off - sizeof(struct ip6_hdr);
1990 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
1991 		m_adj(m, l);
1992 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1993 		if (m->m_len < l) {
1994 			if ((m = m_pullup(m, l)) == NULL)
1995 				return;
1996 		}
1997 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
1998 	} else /* off == sizeof(struct ip6_hdr) */ {
1999 		size_t l;
2000 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2001 		if (m->m_len < l) {
2002 			if ((m = m_pullup(m, l)) == NULL)
2003 				return;
2004 		}
2005 	}
2006 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2007 	ip6 = mtod(m, struct ip6_hdr *);
2008 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2009 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2010 	type = icmp6->icmp6_type; /* keep type for statistics */
2011 	code = icmp6->icmp6_code; /* ditto. */
2012 
2013 	origdst = ip6->ip6_dst;
2014 	/*
2015 	 * ip6_input() drops a packet if its src is multicast.
2016 	 * So, the src is never multicast.
2017 	 */
2018 	ip6->ip6_dst = ip6->ip6_src;
2019 
2020 	/*
2021 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2022 	 * use dst as the src for the reply.
2023 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2024 	 * (for example) when we encounter an error while forwarding procedure
2025 	 * destined to a duplicated address of ours.
2026 	 * Note that ip6_getdstifaddr() may fail if we are in an error handling
2027 	 * procedure of an outgoing packet of our own, in which case we need
2028 	 * to search in the ifaddr list.
2029 	 */
2030 	if (!IN6_IS_ADDR_MULTICAST(&origdst)) {
2031 		if ((ia = ip6_getdstifaddr(m))) {
2032 			if (!(ia->ia6_flags &
2033 			    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)))
2034 				src = &ia->ia_addr.sin6_addr;
2035 		} else {
2036 			struct sockaddr_in6 d;
2037 
2038 			bzero(&d, sizeof(d));
2039 			d.sin6_family = AF_INET6;
2040 			d.sin6_len = sizeof(d);
2041 			d.sin6_addr = origdst;
2042 			ia = (struct in6_ifaddr *)
2043 			    ifa_ifwithaddr((struct sockaddr *)&d);
2044 			if (ia &&
2045 			    !(ia->ia6_flags &
2046 			    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2047 				src = &ia->ia_addr.sin6_addr;
2048 			}
2049 		}
2050 	}
2051 
2052 	if (src == NULL) {
2053 		int e;
2054 		struct sockaddr_in6 sin6;
2055 		struct route_in6 ro;
2056 
2057 		/*
2058 		 * This case matches to multicasts, our anycast, or unicasts
2059 		 * that we do not own.  Select a source address based on the
2060 		 * source address of the erroneous packet.
2061 		 */
2062 		bzero(&sin6, sizeof(sin6));
2063 		sin6.sin6_family = AF_INET6;
2064 		sin6.sin6_len = sizeof(sin6);
2065 		sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */
2066 
2067 		bzero(&ro, sizeof(ro));
2068 		src = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &e);
2069 		if (ro.ro_rt) { /* XXX: see comments in icmp6_mtudisc_update */
2070 			RTFREE(ro.ro_rt); /* XXX: we could use this */
2071 		}
2072 		if (src == NULL) {
2073 			nd6log((LOG_DEBUG,
2074 			    "icmp6_reflect: source can't be determined: "
2075 			    "dst=%s, error=%d\n",
2076 			    ip6_sprintf(&sin6.sin6_addr), e));
2077 			goto bad;
2078 		}
2079 	}
2080 
2081 	ip6->ip6_src = *src;
2082 	ip6->ip6_flow = 0;
2083 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2084 	ip6->ip6_vfc |= IPV6_VERSION;
2085 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2086 	if (m->m_pkthdr.rcvif) {
2087 		/* XXX: This may not be the outgoing interface */
2088 		ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2089 	} else
2090 		ip6->ip6_hlim = ip6_defhlim;
2091 
2092 	m->m_pkthdr.csum_flags = 0;
2093 	icmp6->icmp6_cksum = 0;
2094 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2095 					sizeof(struct ip6_hdr), plen);
2096 
2097 	/*
2098 	 * XXX option handling
2099 	 */
2100 
2101 	m->m_flags &= ~(M_BCAST|M_MCAST);
2102 
2103 	/*
2104 	 * To avoid a "too big" situation at an intermediate router
2105 	 * and the path MTU discovery process, specify the IPV6_MINMTU flag.
2106 	 * Note that only echo and node information replies are affected,
2107 	 * since the length of ICMP6 errors is limited to the minimum MTU.
2108 	 */
2109 	if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL, &outif) != 0 &&
2110 	    outif)
2111 		icmp6_ifstat_inc(outif, ifs6_out_error);
2112 
2113 	if (outif)
2114 		icmp6_ifoutstat_inc(outif, type, code);
2115 
2116 	return;
2117 
2118  bad:
2119 	m_freem(m);
2120 	return;
2121 }
2122 
2123 static const char *
2124 icmp6_redirect_diag(src6, dst6, tgt6)
2125 	struct in6_addr *src6;
2126 	struct in6_addr *dst6;
2127 	struct in6_addr *tgt6;
2128 {
2129 	static char buf[1024];
2130 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2131 		ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2132 	return buf;
2133 }
2134 
2135 void
2136 icmp6_redirect_input(m, off)
2137 	struct mbuf *m;
2138 	int off;
2139 {
2140 	struct ifnet *ifp = m->m_pkthdr.rcvif;
2141 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2142 	struct nd_redirect *nd_rd;
2143 	int icmp6len = ntohs(ip6->ip6_plen);
2144 	char *lladdr = NULL;
2145 	int lladdrlen = 0;
2146 	struct rtentry *rt = NULL;
2147 	int is_router;
2148 	int is_onlink;
2149 	struct in6_addr src6 = ip6->ip6_src;
2150 	struct in6_addr redtgt6;
2151 	struct in6_addr reddst6;
2152 	union nd_opts ndopts;
2153 
2154 	if (!ifp)
2155 		return;
2156 
2157 	/* XXX if we are router, we don't update route by icmp6 redirect */
2158 	if (ip6_forwarding)
2159 		goto freeit;
2160 	if (!icmp6_rediraccept)
2161 		goto freeit;
2162 
2163 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2164 	if (nd_rd == NULL) {
2165 		icmp6stat.icp6s_tooshort++;
2166 		return;
2167 	}
2168 	redtgt6 = nd_rd->nd_rd_target;
2169 	reddst6 = nd_rd->nd_rd_dst;
2170 
2171 	if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2172 	    in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2173 		goto freeit;
2174 	}
2175 
2176 	/* validation */
2177 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2178 		nd6log((LOG_ERR,
2179 			"ICMP6 redirect sent from %s rejected; "
2180 			"must be from linklocal\n", ip6_sprintf(&src6)));
2181 		goto bad;
2182 	}
2183 	if (ip6->ip6_hlim != 255) {
2184 		nd6log((LOG_ERR,
2185 			"ICMP6 redirect sent from %s rejected; "
2186 			"hlim=%d (must be 255)\n",
2187 			ip6_sprintf(&src6), ip6->ip6_hlim));
2188 		goto bad;
2189 	}
2190     {
2191 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2192 	struct sockaddr_in6 sin6;
2193 	struct in6_addr *gw6;
2194 
2195 	bzero(&sin6, sizeof(sin6));
2196 	sin6.sin6_family = AF_INET6;
2197 	sin6.sin6_len = sizeof(struct sockaddr_in6);
2198 	bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2199 	rt = rtalloc1((struct sockaddr *)&sin6, 0);
2200 	if (rt) {
2201 		if (rt->rt_gateway == NULL ||
2202 		    rt->rt_gateway->sa_family != AF_INET6) {
2203 			nd6log((LOG_ERR,
2204 			    "ICMP6 redirect rejected; no route "
2205 			    "with inet6 gateway found for redirect dst: %s\n",
2206 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2207 			RTFREE(rt);
2208 			goto bad;
2209 		}
2210 
2211 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2212 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2213 			nd6log((LOG_ERR,
2214 				"ICMP6 redirect rejected; "
2215 				"not equal to gw-for-src=%s (must be same): "
2216 				"%s\n",
2217 				ip6_sprintf(gw6),
2218 				icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2219 			RTFREE(rt);
2220 			goto bad;
2221 		}
2222 	} else {
2223 		nd6log((LOG_ERR,
2224 			"ICMP6 redirect rejected; "
2225 			"no route found for redirect dst: %s\n",
2226 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2227 		goto bad;
2228 	}
2229 	RTFREE(rt);
2230 	rt = NULL;
2231     }
2232 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2233 		nd6log((LOG_ERR,
2234 			"ICMP6 redirect rejected; "
2235 			"redirect dst must be unicast: %s\n",
2236 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2237 		goto bad;
2238 	}
2239 
2240 	is_router = is_onlink = 0;
2241 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2242 		is_router = 1;	/* router case */
2243 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2244 		is_onlink = 1;	/* on-link destination case */
2245 	if (!is_router && !is_onlink) {
2246 		nd6log((LOG_ERR,
2247 			"ICMP6 redirect rejected; "
2248 			"neither router case nor onlink case: %s\n",
2249 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2250 		goto bad;
2251 	}
2252 	/* validation passed */
2253 
2254 	icmp6len -= sizeof(*nd_rd);
2255 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2256 	if (nd6_options(&ndopts) < 0) {
2257 		nd6log((LOG_INFO, "icmp6_redirect_input: "
2258 			"invalid ND option, rejected: %s\n",
2259 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2260 		/* nd6_options have incremented stats */
2261 		goto freeit;
2262 	}
2263 
2264 	if (ndopts.nd_opts_tgt_lladdr) {
2265 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2266 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2267 	}
2268 
2269 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2270 		nd6log((LOG_INFO,
2271 			"icmp6_redirect_input: lladdrlen mismatch for %s "
2272 			"(if %d, icmp6 packet %d): %s\n",
2273 			ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2274 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2275 		goto bad;
2276 	}
2277 
2278 	/* RFC 2461 8.3 */
2279 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2280 			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2281 
2282 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2283 		/* perform rtredirect */
2284 		struct sockaddr_in6 sdst;
2285 		struct sockaddr_in6 sgw;
2286 		struct sockaddr_in6 ssrc;
2287 		unsigned long rtcount;
2288 		struct rtentry *newrt = NULL;
2289 
2290 		/*
2291 		 * do not install redirect route, if the number of entries
2292 		 * is too much (> hiwat).  note that, the node (= host) will
2293 		 * work just fine even if we do not install redirect route
2294 		 * (there will be additional hops, though).
2295 		 */
2296 		rtcount = rt_timer_count(icmp6_redirect_timeout_q);
2297 		if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat)
2298 			return;
2299 		else if (0 <= icmp6_redirect_lowat &&
2300 		    rtcount > icmp6_redirect_lowat) {
2301 			/*
2302 			 * XXX nuke a victim, install the new one.
2303 			 */
2304 		}
2305 
2306 		bzero(&sdst, sizeof(sdst));
2307 		bzero(&sgw, sizeof(sgw));
2308 		bzero(&ssrc, sizeof(ssrc));
2309 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2310 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2311 			sizeof(struct sockaddr_in6);
2312 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2313 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2314 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2315 		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2316 			   (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2317 			   (struct sockaddr *)&ssrc,
2318 			   &newrt);
2319 
2320 		if (newrt) {
2321 			(void)rt_timer_add(newrt, icmp6_redirect_timeout,
2322 			    icmp6_redirect_timeout_q);
2323 			rtfree(newrt);
2324 		}
2325 	}
2326 	/* finally update cached route in each socket via pfctlinput */
2327 	{
2328 		struct sockaddr_in6 sdst;
2329 
2330 		bzero(&sdst, sizeof(sdst));
2331 		sdst.sin6_family = AF_INET6;
2332 		sdst.sin6_len = sizeof(struct sockaddr_in6);
2333 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2334 		pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2335 #ifdef IPSEC
2336 		key_sa_routechange((struct sockaddr *)&sdst);
2337 #endif
2338 	}
2339 
2340  freeit:
2341 	m_freem(m);
2342 	return;
2343 
2344  bad:
2345 	icmp6stat.icp6s_badredirect++;
2346 	m_freem(m);
2347 }
2348 
2349 void
2350 icmp6_redirect_output(m0, rt)
2351 	struct mbuf *m0;
2352 	struct rtentry *rt;
2353 {
2354 	struct ifnet *ifp;	/* my outgoing interface */
2355 	struct in6_addr *ifp_ll6;
2356 	struct in6_addr *nexthop;
2357 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2358 	struct mbuf *m = NULL;	/* newly allocated one */
2359 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2360 	struct nd_redirect *nd_rd;
2361 	size_t maxlen;
2362 	u_char *p;
2363 	struct sockaddr_in6 src_sa;
2364 
2365 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2366 
2367 	/* if we are not router, we don't send icmp6 redirect */
2368 	if (!ip6_forwarding)
2369 		goto fail;
2370 
2371 	/* sanity check */
2372 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2373 		goto fail;
2374 
2375 	/*
2376 	 * Address check:
2377 	 *  the source address must identify a neighbor, and
2378 	 *  the destination address must not be a multicast address
2379 	 *  [RFC 2461, sec 8.2]
2380 	 */
2381 	sip6 = mtod(m0, struct ip6_hdr *);
2382 	bzero(&src_sa, sizeof(src_sa));
2383 	src_sa.sin6_family = AF_INET6;
2384 	src_sa.sin6_len = sizeof(src_sa);
2385 	src_sa.sin6_addr = sip6->ip6_src;
2386 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2387 		goto fail;
2388 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2389 		goto fail;	/* what should we do here? */
2390 
2391 	/* rate limit */
2392 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2393 		goto fail;
2394 
2395 	/*
2396 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2397 	 * we almost always ask for an mbuf cluster for simplicity.
2398 	 * (MHLEN < IPV6_MMTU is almost always true)
2399 	 */
2400 #if IPV6_MMTU >= MCLBYTES
2401 # error assumption failed about IPV6_MMTU and MCLBYTES
2402 #endif
2403 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2404 	if (m && IPV6_MMTU >= MHLEN)
2405 		MCLGET(m, M_DONTWAIT);
2406 	if (!m)
2407 		goto fail;
2408 	m->m_pkthdr.rcvif = NULL;
2409 	m->m_len = 0;
2410 	maxlen = M_TRAILINGSPACE(m);
2411 	maxlen = min(IPV6_MMTU, maxlen);
2412 	/* just for safety */
2413 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2414 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2415 		goto fail;
2416 	}
2417 
2418 	{
2419 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2420 		struct in6_ifaddr *ia;
2421 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2422 						 IN6_IFF_NOTREADY|
2423 						 IN6_IFF_ANYCAST)) == NULL)
2424 			goto fail;
2425 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2426 	}
2427 
2428 	/* get ip6 linklocal address for the router. */
2429 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2430 		struct sockaddr_in6 *sin6;
2431 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2432 		nexthop = &sin6->sin6_addr;
2433 		if (!IN6_IS_ADDR_LINKLOCAL(nexthop))
2434 			nexthop = NULL;
2435 	} else
2436 		nexthop = NULL;
2437 
2438 	/* ip6 */
2439 	ip6 = mtod(m, struct ip6_hdr *);
2440 	ip6->ip6_flow = 0;
2441 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2442 	ip6->ip6_vfc |= IPV6_VERSION;
2443 	/* ip6->ip6_plen will be set later */
2444 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2445 	ip6->ip6_hlim = 255;
2446 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2447 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2448 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2449 
2450 	/* ND Redirect */
2451 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2452 	nd_rd->nd_rd_type = ND_REDIRECT;
2453 	nd_rd->nd_rd_code = 0;
2454 	nd_rd->nd_rd_reserved = 0;
2455 	if (rt->rt_flags & RTF_GATEWAY) {
2456 		/*
2457 		 * nd_rd->nd_rd_target must be a link-local address in
2458 		 * better router cases.
2459 		 */
2460 		if (!nexthop)
2461 			goto fail;
2462 		bcopy(nexthop, &nd_rd->nd_rd_target,
2463 		      sizeof(nd_rd->nd_rd_target));
2464 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2465 		      sizeof(nd_rd->nd_rd_dst));
2466 	} else {
2467 		/* make sure redtgt == reddst */
2468 		nexthop = &sip6->ip6_dst;
2469 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2470 		      sizeof(nd_rd->nd_rd_target));
2471 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2472 		      sizeof(nd_rd->nd_rd_dst));
2473 	}
2474 
2475 	p = (u_char *)(nd_rd + 1);
2476 
2477 	{
2478 		/* target lladdr option */
2479 		struct rtentry *rt_nexthop = NULL;
2480 		int len;
2481 		struct sockaddr_dl *sdl;
2482 		struct nd_opt_hdr *nd_opt;
2483 		char *lladdr;
2484 
2485 		rt_nexthop = nd6_lookup(nexthop, 0, ifp);
2486 		if (!rt_nexthop)
2487 			goto nolladdropt;
2488 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2489 		len = (len + 7) & ~7;	/* round by 8 */
2490 		/* safety check */
2491 		if (len + (p - (u_char *)ip6) > maxlen)
2492 			goto nolladdropt;
2493 		if (!(rt_nexthop->rt_flags & RTF_GATEWAY) &&
2494 		    (rt_nexthop->rt_flags & RTF_LLINFO) &&
2495 		    (rt_nexthop->rt_gateway->sa_family == AF_LINK) &&
2496 		    (sdl = (struct sockaddr_dl *)rt_nexthop->rt_gateway) &&
2497 		    sdl->sdl_alen) {
2498 			nd_opt = (struct nd_opt_hdr *)p;
2499 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2500 			nd_opt->nd_opt_len = len >> 3;
2501 			lladdr = (char *)(nd_opt + 1);
2502 			bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2503 			p += len;
2504 		}
2505 	}
2506   nolladdropt:;
2507 
2508 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2509 
2510 	/* just to be safe */
2511 	if (m0->m_flags & M_DECRYPTED)
2512 		goto noredhdropt;
2513 	if (p - (u_char *)ip6 > maxlen)
2514 		goto noredhdropt;
2515 
2516 	{
2517 		/* redirected header option */
2518 		int len;
2519 		struct nd_opt_rd_hdr *nd_opt_rh;
2520 
2521 		/*
2522 		 * compute the maximum size for icmp6 redirect header option.
2523 		 * XXX room for auth header?
2524 		 */
2525 		len = maxlen - (p - (u_char *)ip6);
2526 		len &= ~7;
2527 
2528 		/*
2529 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2530 		 * about padding/truncate rule for the original IP packet.
2531 		 * From the discussion on IPv6imp in Feb 1999,
2532 		 * the consensus was:
2533 		 * - "attach as much as possible" is the goal
2534 		 * - pad if not aligned (original size can be guessed by
2535 		 *   original ip6 header)
2536 		 * Following code adds the padding if it is simple enough,
2537 		 * and truncates if not.
2538 		 */
2539 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2540 			/* not enough room, truncate */
2541 			m_adj(m0, (len - sizeof(*nd_opt_rh)) -
2542 			    m0->m_pkthdr.len);
2543 		} else {
2544 			/*
2545 			 * enough room, truncate if not aligned.
2546 			 * we don't pad here for simplicity.
2547 			 */
2548 			size_t extra;
2549 
2550 			extra = m0->m_pkthdr.len % 8;
2551 			if (extra) {
2552 				/* truncate */
2553 				m_adj(m0, -extra);
2554 			}
2555 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2556 		}
2557 
2558 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2559 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2560 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2561 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2562 		p += sizeof(*nd_opt_rh);
2563 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2564 
2565 		/* connect m0 to m */
2566 		m->m_pkthdr.len += m0->m_pkthdr.len;
2567 		m_cat(m, m0);
2568 		m0 = NULL;
2569 	}
2570 noredhdropt:
2571 	if (m0) {
2572 		m_freem(m0);
2573 		m0 = NULL;
2574 	}
2575 
2576 	/* XXX: clear embedded link IDs in the inner header */
2577 	in6_clearscope(&sip6->ip6_src);
2578 	in6_clearscope(&sip6->ip6_dst);
2579 	in6_clearscope(&nd_rd->nd_rd_target);
2580 	in6_clearscope(&nd_rd->nd_rd_dst);
2581 
2582 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2583 
2584 	nd_rd->nd_rd_cksum = 0;
2585 	nd_rd->nd_rd_cksum
2586 		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2587 
2588 	/* send the packet to outside... */
2589 	if (ip6_output(m, NULL, NULL, 0,
2590 		(struct ip6_moptions *)NULL, (struct socket *)NULL, NULL) != 0)
2591 		icmp6_ifstat_inc(ifp, ifs6_out_error);
2592 
2593 	icmp6_ifstat_inc(ifp, ifs6_out_msg);
2594 	icmp6_ifstat_inc(ifp, ifs6_out_redirect);
2595 	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2596 
2597 	return;
2598 
2599 fail:
2600 	if (m)
2601 		m_freem(m);
2602 	if (m0)
2603 		m_freem(m0);
2604 }
2605 
2606 /*
2607  * ICMPv6 socket option processing.
2608  */
2609 int
2610 icmp6_ctloutput(op, so, level, optname, mp)
2611 	int op;
2612 	struct socket *so;
2613 	int level, optname;
2614 	struct mbuf **mp;
2615 {
2616 	int error = 0;
2617 	int optlen;
2618 	struct in6pcb *in6p = sotoin6pcb(so);
2619 	struct mbuf *m = *mp;
2620 
2621 	optlen = m ? m->m_len : 0;
2622 
2623 	if (level != IPPROTO_ICMPV6) {
2624 		if (op == PRCO_SETOPT && m)
2625 			(void)m_free(m);
2626 		return EINVAL;
2627 	}
2628 
2629 	switch (op) {
2630 	case PRCO_SETOPT:
2631 		switch (optname) {
2632 		case ICMP6_FILTER:
2633 		    {
2634 			struct icmp6_filter *p;
2635 
2636 			if (optlen != sizeof(*p)) {
2637 				error = EMSGSIZE;
2638 				break;
2639 			}
2640 			p = mtod(m, struct icmp6_filter *);
2641 			if (!p || !in6p->in6p_icmp6filt) {
2642 				error = EINVAL;
2643 				break;
2644 			}
2645 			bcopy(p, in6p->in6p_icmp6filt,
2646 				sizeof(struct icmp6_filter));
2647 			error = 0;
2648 			break;
2649 		    }
2650 
2651 		default:
2652 			error = ENOPROTOOPT;
2653 			break;
2654 		}
2655 		if (m)
2656 			(void)m_freem(m);
2657 		break;
2658 
2659 	case PRCO_GETOPT:
2660 		switch (optname) {
2661 		case ICMP6_FILTER:
2662 		    {
2663 			struct icmp6_filter *p;
2664 
2665 			if (!in6p->in6p_icmp6filt) {
2666 				error = EINVAL;
2667 				break;
2668 			}
2669 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
2670 			m->m_len = sizeof(struct icmp6_filter);
2671 			p = mtod(m, struct icmp6_filter *);
2672 			bcopy(in6p->in6p_icmp6filt, p,
2673 				sizeof(struct icmp6_filter));
2674 			error = 0;
2675 			break;
2676 		    }
2677 
2678 		default:
2679 			error = ENOPROTOOPT;
2680 			break;
2681 		}
2682 		break;
2683 	}
2684 
2685 	return (error);
2686 }
2687 
2688 /*
2689  * Perform rate limit check.
2690  * Returns 0 if it is okay to send the icmp6 packet.
2691  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2692  * limitation.
2693  *
2694  * XXX per-destination/type check necessary?
2695  */
2696 static int
2697 icmp6_ratelimit(
2698 	const struct in6_addr *dst __unused,	/* not used at this moment */
2699 	const int type __unused,		/* not used at this moment */
2700 	const int code __unused)		/* not used at this moment */
2701 {
2702 	int ret;
2703 
2704 	ret = 0;	/* okay to send */
2705 
2706 	/* PPS limit */
2707 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2708 	    icmp6errppslim)) {
2709 		/* The packet is subject to rate limit */
2710 		ret++;
2711 	}
2712 
2713 	return ret;
2714 }
2715 
2716 static struct rtentry *
2717 icmp6_mtudisc_clone(dst)
2718 	struct sockaddr *dst;
2719 {
2720 	struct rtentry *rt;
2721 	int    error;
2722 
2723 	rt = rtalloc1(dst, 1);
2724 	if (rt == 0)
2725 		return NULL;
2726 
2727 	/* If we didn't get a host route, allocate one */
2728 	if ((rt->rt_flags & RTF_HOST) == 0) {
2729 		struct rtentry *nrt;
2730 
2731 		error = rtrequest((int) RTM_ADD, dst,
2732 		    (struct sockaddr *) rt->rt_gateway,
2733 		    (struct sockaddr *) 0,
2734 		    RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt);
2735 		if (error) {
2736 			rtfree(rt);
2737 			return NULL;
2738 		}
2739 		nrt->rt_rmx = rt->rt_rmx;
2740 		rtfree(rt);
2741 		rt = nrt;
2742 	}
2743 	error = rt_timer_add(rt, icmp6_mtudisc_timeout,
2744 			icmp6_mtudisc_timeout_q);
2745 	if (error) {
2746 		rtfree(rt);
2747 		return NULL;
2748 	}
2749 
2750 	return rt;	/* caller need to call rtfree() */
2751 }
2752 
2753 static void
2754 icmp6_mtudisc_timeout(struct rtentry *rt, struct rttimer *r __unused)
2755 {
2756 	if (rt == NULL)
2757 		panic("icmp6_mtudisc_timeout: bad route to timeout");
2758 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
2759 	    (RTF_DYNAMIC | RTF_HOST)) {
2760 		rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt),
2761 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
2762 	} else {
2763 		if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
2764 			rt->rt_rmx.rmx_mtu = 0;
2765 	}
2766 }
2767 
2768 static void
2769 icmp6_redirect_timeout(struct rtentry *rt, struct rttimer *r __unused)
2770 {
2771 	if (rt == NULL)
2772 		panic("icmp6_redirect_timeout: bad route to timeout");
2773 	if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) ==
2774 	    (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) {
2775 		rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt),
2776 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
2777 	}
2778 }
2779 
2780 /*
2781  * sysctl helper routine for the net.inet6.icmp6.nd6 nodes.  silly?
2782  */
2783 static int
2784 sysctl_net_inet6_icmp6_nd6(SYSCTLFN_ARGS)
2785 {
2786 	(void)&name;
2787 	(void)&l;
2788 	(void)&oname;
2789 
2790 	if (namelen != 0)
2791 		return (EINVAL);
2792 
2793 	return (nd6_sysctl(rnode->sysctl_num, oldp, oldlenp,
2794 	    /*XXXUNCONST*/
2795 	    __UNCONST(newp), newlen));
2796 }
2797 
2798 SYSCTL_SETUP(sysctl_net_inet6_icmp6_setup,
2799 	     "sysctl net.inet6.icmp6 subtree setup")
2800 {
2801 	extern int nd6_maxqueuelen; /* defined in nd6.c */
2802 
2803 	sysctl_createv(clog, 0, NULL, NULL,
2804 		       CTLFLAG_PERMANENT,
2805 		       CTLTYPE_NODE, "net", NULL,
2806 		       NULL, 0, NULL, 0,
2807 		       CTL_NET, CTL_EOL);
2808 	sysctl_createv(clog, 0, NULL, NULL,
2809 		       CTLFLAG_PERMANENT,
2810 		       CTLTYPE_NODE, "inet6", NULL,
2811 		       NULL, 0, NULL, 0,
2812 		       CTL_NET, PF_INET6, CTL_EOL);
2813 	sysctl_createv(clog, 0, NULL, NULL,
2814 		       CTLFLAG_PERMANENT,
2815 		       CTLTYPE_NODE, "icmp6",
2816 		       SYSCTL_DESCR("ICMPv6 related settings"),
2817 		       NULL, 0, NULL, 0,
2818 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6, CTL_EOL);
2819 
2820 	sysctl_createv(clog, 0, NULL, NULL,
2821 		       CTLFLAG_PERMANENT,
2822 		       CTLTYPE_STRUCT, "stats",
2823 		       SYSCTL_DESCR("ICMPv6 transmission statistics"),
2824 		       NULL, 0, &icmp6stat, sizeof(icmp6stat),
2825 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2826 		       ICMPV6CTL_STATS, CTL_EOL);
2827 	sysctl_createv(clog, 0, NULL, NULL,
2828 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2829 		       CTLTYPE_INT, "rediraccept",
2830 		       SYSCTL_DESCR("Accept and process redirect messages"),
2831 		       NULL, 0, &icmp6_rediraccept, 0,
2832 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2833 		       ICMPV6CTL_REDIRACCEPT, CTL_EOL);
2834 	sysctl_createv(clog, 0, NULL, NULL,
2835 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2836 		       CTLTYPE_INT, "redirtimeout",
2837 		       SYSCTL_DESCR("Redirect generated route lifetime"),
2838 		       NULL, 0, &icmp6_redirtimeout, 0,
2839 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2840 		       ICMPV6CTL_REDIRTIMEOUT, CTL_EOL);
2841 #if 0 /* obsoleted */
2842 	sysctl_createv(clog, 0, NULL, NULL,
2843 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2844 		       CTLTYPE_INT, "errratelimit", NULL,
2845 		       NULL, 0, &icmp6_errratelimit, 0,
2846 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2847 		       ICMPV6CTL_ERRRATELIMIT, CTL_EOL);
2848 #endif
2849 	sysctl_createv(clog, 0, NULL, NULL,
2850 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2851 		       CTLTYPE_INT, "nd6_prune",
2852 		       SYSCTL_DESCR("Neighbor discovery prune interval"),
2853 		       NULL, 0, &nd6_prune, 0,
2854 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2855 		       ICMPV6CTL_ND6_PRUNE, CTL_EOL);
2856 	sysctl_createv(clog, 0, NULL, NULL,
2857 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2858 		       CTLTYPE_INT, "nd6_delay",
2859 		       SYSCTL_DESCR("First probe delay time"),
2860 		       NULL, 0, &nd6_delay, 0,
2861 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2862 		       ICMPV6CTL_ND6_DELAY, CTL_EOL);
2863 	sysctl_createv(clog, 0, NULL, NULL,
2864 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2865 		       CTLTYPE_INT, "nd6_umaxtries",
2866 		       SYSCTL_DESCR("Number of unicast discovery attempts"),
2867 		       NULL, 0, &nd6_umaxtries, 0,
2868 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2869 		       ICMPV6CTL_ND6_UMAXTRIES, CTL_EOL);
2870 	sysctl_createv(clog, 0, NULL, NULL,
2871 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2872 		       CTLTYPE_INT, "nd6_mmaxtries",
2873 		       SYSCTL_DESCR("Number of multicast discovery attempts"),
2874 		       NULL, 0, &nd6_mmaxtries, 0,
2875 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2876 		       ICMPV6CTL_ND6_MMAXTRIES, CTL_EOL);
2877 	sysctl_createv(clog, 0, NULL, NULL,
2878 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2879 		       CTLTYPE_INT, "nd6_useloopback",
2880 		       SYSCTL_DESCR("Use loopback interface for local traffic"),
2881 		       NULL, 0, &nd6_useloopback, 0,
2882 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2883 		       ICMPV6CTL_ND6_USELOOPBACK, CTL_EOL);
2884 #if 0 /* obsoleted */
2885 	sysctl_createv(clog, 0, NULL, NULL,
2886 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2887 		       CTLTYPE_INT, "nd6_proxyall", NULL,
2888 		       NULL, 0, &nd6_proxyall, 0,
2889 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2890 		       ICMPV6CTL_ND6_PROXYALL, CTL_EOL);
2891 #endif
2892 	sysctl_createv(clog, 0, NULL, NULL,
2893 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2894 		       CTLTYPE_INT, "nodeinfo",
2895 		       SYSCTL_DESCR("Respond to node information requests"),
2896 		       NULL, 0, &icmp6_nodeinfo, 0,
2897 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2898 		       ICMPV6CTL_NODEINFO, CTL_EOL);
2899 	sysctl_createv(clog, 0, NULL, NULL,
2900 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2901 		       CTLTYPE_INT, "errppslimit",
2902 		       SYSCTL_DESCR("Maximum ICMP errors sent per second"),
2903 		       NULL, 0, &icmp6errppslim, 0,
2904 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2905 		       ICMPV6CTL_ERRPPSLIMIT, CTL_EOL);
2906 	sysctl_createv(clog, 0, NULL, NULL,
2907 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2908 		       CTLTYPE_INT, "nd6_maxnudhint",
2909 		       SYSCTL_DESCR("Maximum neighbor unreachable hint count"),
2910 		       NULL, 0, &nd6_maxnudhint, 0,
2911 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2912 		       ICMPV6CTL_ND6_MAXNUDHINT, CTL_EOL);
2913 	sysctl_createv(clog, 0, NULL, NULL,
2914 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2915 		       CTLTYPE_INT, "mtudisc_hiwat",
2916 		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
2917 		       NULL, 0, &icmp6_mtudisc_hiwat, 0,
2918 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2919 		       ICMPV6CTL_MTUDISC_HIWAT, CTL_EOL);
2920 	sysctl_createv(clog, 0, NULL, NULL,
2921 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2922 		       CTLTYPE_INT, "mtudisc_lowat",
2923 		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
2924 		       NULL, 0, &icmp6_mtudisc_lowat, 0,
2925 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2926 		       ICMPV6CTL_MTUDISC_LOWAT, CTL_EOL);
2927 	sysctl_createv(clog, 0, NULL, NULL,
2928 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2929 		       CTLTYPE_INT, "nd6_debug",
2930 		       SYSCTL_DESCR("Enable neighbor discovery debug output"),
2931 		       NULL, 0, &nd6_debug, 0,
2932 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2933 		       ICMPV6CTL_ND6_DEBUG, CTL_EOL);
2934 	sysctl_createv(clog, 0, NULL, NULL,
2935 		       CTLFLAG_PERMANENT,
2936 		       CTLTYPE_STRUCT, "nd6_drlist",
2937 		       SYSCTL_DESCR("Default router list"),
2938 		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
2939 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2940 		       ICMPV6CTL_ND6_DRLIST, CTL_EOL);
2941 	sysctl_createv(clog, 0, NULL, NULL,
2942 		       CTLFLAG_PERMANENT,
2943 		       CTLTYPE_STRUCT, "nd6_prlist",
2944 		       SYSCTL_DESCR("Prefix list"),
2945 		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
2946 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2947 		       ICMPV6CTL_ND6_PRLIST, CTL_EOL);
2948 	sysctl_createv(clog, 0, NULL, NULL,
2949 		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2950 		       CTLTYPE_INT, "maxqueuelen",
2951 		       SYSCTL_DESCR("max packet queue len for a unresolved ND"),
2952 		       NULL, 1, &nd6_maxqueuelen, 0,
2953 		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2954 		       ICMPV6CTL_ND6_MAXQLEN, CTL_EOL);
2955 }
2956