icmp6.c revision 1.202
1/*	$NetBSD: icmp6.c,v 1.202 2016/12/11 07:35:42 ozaki-r 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 1.202 2016/12/11 07:35:42 ozaki-r Exp $");
66
67#ifdef _KERNEL_OPT
68#include "opt_inet.h"
69#include "opt_ipsec.h"
70#endif
71
72#include <sys/param.h>
73#include <sys/systm.h>
74#include <sys/kmem.h>
75#include <sys/mbuf.h>
76#include <sys/protosw.h>
77#include <sys/socket.h>
78#include <sys/socketvar.h>
79#include <sys/time.h>
80#include <sys/kernel.h>
81#include <sys/syslog.h>
82#include <sys/domain.h>
83#include <sys/sysctl.h>
84
85#include <net/if.h>
86#include <net/route.h>
87#include <net/if_dl.h>
88#include <net/if_types.h>
89
90#include <netinet/in.h>
91#include <netinet/in_var.h>
92#include <netinet/ip6.h>
93#include <netinet6/ip6_var.h>
94#include <netinet6/ip6_private.h>
95#include <netinet/icmp6.h>
96#include <netinet6/icmp6_private.h>
97#include <netinet6/mld6_var.h>
98#include <netinet6/in6_pcb.h>
99#include <netinet6/nd6.h>
100#include <netinet6/in6_ifattach.h>
101#include <netinet6/ip6protosw.h>
102#include <netinet6/scope6_var.h>
103
104#ifdef IPSEC
105#include <netipsec/ipsec.h>
106#include <netipsec/key.h>
107#endif
108
109
110#include "faith.h"
111#if defined(NFAITH) && 0 < NFAITH
112#include <net/if_faith.h>
113#endif
114
115#include <net/net_osdep.h>
116
117extern struct domain inet6domain;
118
119percpu_t *icmp6stat_percpu;
120
121extern struct inpcbtable raw6cbtable;
122extern int icmp6errppslim;
123static int icmp6errpps_count = 0;
124static struct timeval icmp6errppslim_last;
125extern int icmp6_nodeinfo;
126
127/*
128 * List of callbacks to notify when Path MTU changes are made.
129 */
130struct icmp6_mtudisc_callback {
131	LIST_ENTRY(icmp6_mtudisc_callback) mc_list;
132	void (*mc_func)(struct in6_addr *);
133};
134
135LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks =
136    LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks);
137
138static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL;
139extern int pmtu_expire;
140
141/* XXX do these values make any sense? */
142static int icmp6_mtudisc_hiwat = 1280;
143static int icmp6_mtudisc_lowat = 256;
144
145/*
146 * keep track of # of redirect routes.
147 */
148static struct rttimer_queue *icmp6_redirect_timeout_q = NULL;
149
150/* XXX experimental, turned off */
151static int icmp6_redirect_hiwat = -1;
152static int icmp6_redirect_lowat = -1;
153
154static void icmp6_errcount(u_int, int, int);
155static int icmp6_rip6_input(struct mbuf **, int);
156static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
157static const char *icmp6_redirect_diag(struct in6_addr *,
158	struct in6_addr *, struct in6_addr *);
159static struct mbuf *ni6_input(struct mbuf *, int);
160static struct mbuf *ni6_nametodns(const char *, int, int);
161static int ni6_dnsmatch(const char *, int, const char *, int);
162static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
163			  struct ifnet **, char *, struct psref *);
164static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
165				struct ifnet *, int);
166static int icmp6_notify_error(struct mbuf *, int, int, int);
167static struct rtentry *icmp6_mtudisc_clone(struct sockaddr *);
168static void icmp6_mtudisc_timeout(struct rtentry *, struct rttimer *);
169static void icmp6_redirect_timeout(struct rtentry *, struct rttimer *);
170static void sysctl_net_inet6_icmp6_setup(struct sysctllog **);
171
172
173void
174icmp6_init(void)
175{
176
177	sysctl_net_inet6_icmp6_setup(NULL);
178	mld_init();
179	icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire);
180	icmp6_redirect_timeout_q = rt_timer_queue_create(icmp6_redirtimeout);
181
182	icmp6stat_percpu = percpu_alloc(sizeof(uint64_t) * ICMP6_NSTATS);
183}
184
185static void
186icmp6_errcount(u_int base, int type, int code)
187{
188	switch (type) {
189	case ICMP6_DST_UNREACH:
190		switch (code) {
191		case ICMP6_DST_UNREACH_NOROUTE:
192			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOROUTE);
193			return;
194		case ICMP6_DST_UNREACH_ADMIN:
195			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADMIN);
196			return;
197		case ICMP6_DST_UNREACH_BEYONDSCOPE:
198			ICMP6_STATINC(base +
199				      ICMP6_ERRSTAT_DST_UNREACH_BEYONDSCOPE);
200			return;
201		case ICMP6_DST_UNREACH_ADDR:
202			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_ADDR);
203			return;
204		case ICMP6_DST_UNREACH_NOPORT:
205			ICMP6_STATINC(base + ICMP6_ERRSTAT_DST_UNREACH_NOPORT);
206			return;
207		}
208		break;
209	case ICMP6_PACKET_TOO_BIG:
210		ICMP6_STATINC(base + ICMP6_ERRSTAT_PACKET_TOO_BIG);
211		return;
212	case ICMP6_TIME_EXCEEDED:
213		switch (code) {
214		case ICMP6_TIME_EXCEED_TRANSIT:
215			ICMP6_STATINC(base + ICMP6_ERRSTAT_TIME_EXCEED_TRANSIT);
216			return;
217		case ICMP6_TIME_EXCEED_REASSEMBLY:
218			ICMP6_STATINC(base +
219				      ICMP6_ERRSTAT_TIME_EXCEED_REASSEMBLY);
220			return;
221		}
222		break;
223	case ICMP6_PARAM_PROB:
224		switch (code) {
225		case ICMP6_PARAMPROB_HEADER:
226			ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_HEADER);
227			return;
228		case ICMP6_PARAMPROB_NEXTHEADER:
229			ICMP6_STATINC(base +
230				      ICMP6_ERRSTAT_PARAMPROB_NEXTHEADER);
231			return;
232		case ICMP6_PARAMPROB_OPTION:
233			ICMP6_STATINC(base + ICMP6_ERRSTAT_PARAMPROB_OPTION);
234			return;
235		}
236		break;
237	case ND_REDIRECT:
238		ICMP6_STATINC(base + ICMP6_ERRSTAT_REDIRECT);
239		return;
240	}
241	ICMP6_STATINC(base + ICMP6_ERRSTAT_UNKNOWN);
242}
243
244/*
245 * Register a Path MTU Discovery callback.
246 */
247void
248icmp6_mtudisc_callback_register(void (*func)(struct in6_addr *))
249{
250	struct icmp6_mtudisc_callback *mc;
251
252	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
253	     mc = LIST_NEXT(mc, mc_list)) {
254		if (mc->mc_func == func)
255			return;
256	}
257
258	mc = kmem_alloc(sizeof(*mc), KM_SLEEP);
259	mc->mc_func = func;
260	LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list);
261}
262
263/*
264 * A wrapper function for icmp6_error() necessary when the erroneous packet
265 * may not contain enough scope zone information.
266 */
267void
268icmp6_error2(struct mbuf *m, int type, int code, int param,
269	struct ifnet *ifp)
270{
271	struct ip6_hdr *ip6;
272
273	if (ifp == NULL)
274		return;
275
276	if (m->m_len < sizeof(struct ip6_hdr)) {
277		m = m_pullup(m, sizeof(struct ip6_hdr));
278		if (m == NULL)
279			return;
280	}
281
282	ip6 = mtod(m, struct ip6_hdr *);
283
284	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
285		return;
286	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
287		return;
288
289	icmp6_error(m, type, code, param);
290}
291
292/*
293 * Generate an error packet of type error in response to bad IP6 packet.
294 */
295void
296icmp6_error(struct mbuf *m, int type, int code, int param)
297{
298	struct ip6_hdr *oip6, *nip6;
299	struct icmp6_hdr *icmp6;
300	u_int preplen;
301	int off;
302	int nxt;
303
304	ICMP6_STATINC(ICMP6_STAT_ERROR);
305
306	/* count per-type-code statistics */
307	icmp6_errcount(ICMP6_STAT_OUTERRHIST, type, code);
308
309	if (m->m_flags & M_DECRYPTED) {
310		ICMP6_STATINC(ICMP6_STAT_CANTERROR);
311		goto freeit;
312	}
313
314	if (M_UNWRITABLE(m, sizeof(struct ip6_hdr)) &&
315	    (m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL)
316		return;
317	oip6 = mtod(m, struct ip6_hdr *);
318
319	/*
320	 * If the destination address of the erroneous packet is a multicast
321	 * address, or the packet was sent using link-layer multicast,
322	 * we should basically suppress sending an error (RFC 2463, Section
323	 * 2.4).
324	 * We have two exceptions (the item e.2 in that section):
325	 * - the Pakcet Too Big message can be sent for path MTU discovery.
326	 * - the Parameter Problem Message that can be allowed an icmp6 error
327	 *   in the option type field.  This check has been done in
328	 *   ip6_unknown_opt(), so we can just check the type and code.
329	 */
330	if ((m->m_flags & (M_BCAST|M_MCAST) ||
331	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
332	    (type != ICMP6_PACKET_TOO_BIG &&
333	     (type != ICMP6_PARAM_PROB ||
334	      code != ICMP6_PARAMPROB_OPTION)))
335		goto freeit;
336
337	/*
338	 * RFC 2463, 2.4 (e.5): source address check.
339	 * XXX: the case of anycast source?
340	 */
341	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
342	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
343		goto freeit;
344
345	/*
346	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
347	 * don't do it.
348	 */
349	nxt = -1;
350	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
351	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
352		struct icmp6_hdr *icp;
353
354		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
355			sizeof(*icp));
356		if (icp == NULL) {
357			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
358			return;
359		}
360		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
361		    icp->icmp6_type == ND_REDIRECT) {
362			/*
363			 * ICMPv6 error
364			 * Special case: for redirect (which is
365			 * informational) we must not send icmp6 error.
366			 */
367			ICMP6_STATINC(ICMP6_STAT_CANTERROR);
368			goto freeit;
369		} else {
370			/* ICMPv6 informational - send the error */
371		}
372	}
373#if 0 /* controversial */
374	else if (off >= 0 && nxt == IPPROTO_ESP) {
375		/*
376		 * It could be ICMPv6 error inside ESP.  Take a safer side,
377		 * don't respond.
378		 */
379		ICMP6_STATINC(ICMP6_STAT_CANTERROR);
380		goto freeit;
381	}
382#endif
383	else {
384		/* non-ICMPv6 - send the error */
385	}
386
387	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
388
389	/* Finally, do rate limitation check. */
390	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
391		ICMP6_STATINC(ICMP6_STAT_TOOFREQ);
392		goto freeit;
393	}
394
395	/*
396	 * OK, ICMP6 can be generated.
397	 */
398
399	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
400		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
401
402	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
403	M_PREPEND(m, preplen, M_DONTWAIT);
404	if (m && M_UNWRITABLE(m, preplen))
405		m = m_pullup(m, preplen);
406	if (m == NULL) {
407		nd6log(LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__);
408		return;
409	}
410
411	nip6 = mtod(m, struct ip6_hdr *);
412	nip6->ip6_src  = oip6->ip6_src;
413	nip6->ip6_dst  = oip6->ip6_dst;
414
415	in6_clearscope(&oip6->ip6_src);
416	in6_clearscope(&oip6->ip6_dst);
417
418	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
419	icmp6->icmp6_type = type;
420	icmp6->icmp6_code = code;
421	icmp6->icmp6_pptr = htonl((u_int32_t)param);
422
423	/*
424	 * icmp6_reflect() is designed to be in the input path.
425	 * icmp6_error() can be called from both input and output path,
426	 * and if we are in output path rcvif could contain bogus value.
427	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
428	 * information in ip header (nip6).
429	 */
430	m_reset_rcvif(m);
431
432	ICMP6_STATINC(ICMP6_STAT_OUTHIST + type);
433	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
434
435	return;
436
437  freeit:
438	/*
439	 * If we can't tell whether or not we can generate ICMP6, free it.
440	 */
441	m_freem(m);
442}
443
444/*
445 * Process a received ICMP6 message.
446 */
447int
448icmp6_input(struct mbuf **mp, int *offp, int proto)
449{
450	struct mbuf *m = *mp, *n;
451	struct ip6_hdr *ip6, *nip6;
452	struct icmp6_hdr *icmp6, *nicmp6;
453	int off = *offp;
454	int icmp6len = m->m_pkthdr.len - *offp;
455	int code, sum, noff;
456	struct ifnet *rcvif;
457	struct psref psref;
458
459	rcvif = m_get_rcvif_psref(m, &psref);
460	if (__predict_false(rcvif == NULL))
461		goto freeit;
462
463#define ICMP6_MAXLEN (sizeof(*nip6) + sizeof(*nicmp6) + 4)
464	KASSERT(ICMP6_MAXLEN < MCLBYTES);
465	icmp6_ifstat_inc(rcvif, ifs6_in_msg);
466
467	/*
468	 * Locate icmp6 structure in mbuf, and check
469	 * that not corrupted and of at least minimum length
470	 */
471
472	if (icmp6len < sizeof(struct icmp6_hdr)) {
473		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
474		icmp6_ifstat_inc(rcvif, ifs6_in_error);
475		goto freeit;
476	}
477
478	ip6 = mtod(m, struct ip6_hdr *);
479	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
480	if (icmp6 == NULL) {
481		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
482		icmp6_ifstat_inc(rcvif, ifs6_in_error);
483		goto freeit;
484	}
485	/*
486	 * Enforce alignment requirements that are violated in
487	 * some cases, see kern/50766 for details.
488	 */
489	if (IP6_HDR_ALIGNED_P(icmp6) == 0) {
490		m = m_copyup(m, off + sizeof(struct icmp6_hdr), 0);
491		if (m == NULL) {
492			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
493			icmp6_ifstat_inc(rcvif, ifs6_in_error);
494			goto freeit;
495		}
496		ip6 = mtod(m, struct ip6_hdr *);
497		icmp6 = (struct icmp6_hdr *)(ip6 + 1);
498	}
499	KASSERT(IP6_HDR_ALIGNED_P(icmp6));
500
501	/*
502	 * calculate the checksum
503	 */
504	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
505		nd6log(LOG_ERR, "ICMP6 checksum error(%d|%x) %s\n",
506		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src));
507		ICMP6_STATINC(ICMP6_STAT_CHECKSUM);
508		icmp6_ifstat_inc(rcvif, ifs6_in_error);
509		goto freeit;
510	}
511
512#if defined(NFAITH) && 0 < NFAITH
513	if (faithprefix(&ip6->ip6_dst)) {
514		/*
515		 * Deliver very specific ICMP6 type only.
516		 * This is important to deliver TOOBIG.  Otherwise PMTUD
517		 * will not work.
518		 */
519		switch (icmp6->icmp6_type) {
520		case ICMP6_DST_UNREACH:
521		case ICMP6_PACKET_TOO_BIG:
522		case ICMP6_TIME_EXCEEDED:
523			break;
524		default:
525			goto freeit;
526		}
527	}
528#endif
529
530	code = icmp6->icmp6_code;
531	ICMP6_STATINC(ICMP6_STAT_INHIST + icmp6->icmp6_type);
532
533	switch (icmp6->icmp6_type) {
534	case ICMP6_DST_UNREACH:
535		icmp6_ifstat_inc(rcvif, ifs6_in_dstunreach);
536		switch (code) {
537		case ICMP6_DST_UNREACH_NOROUTE:
538			code = PRC_UNREACH_NET;
539			break;
540		case ICMP6_DST_UNREACH_ADMIN:
541			icmp6_ifstat_inc(rcvif, ifs6_in_adminprohib);
542			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
543			break;
544		case ICMP6_DST_UNREACH_ADDR:
545			code = PRC_HOSTDEAD;
546			break;
547#ifdef COMPAT_RFC1885
548		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
549			code = PRC_UNREACH_SRCFAIL;
550			break;
551#else
552		case ICMP6_DST_UNREACH_BEYONDSCOPE:
553			/* I mean "source address was incorrect." */
554			code = PRC_UNREACH_NET;
555			break;
556#endif
557		case ICMP6_DST_UNREACH_NOPORT:
558			code = PRC_UNREACH_PORT;
559			break;
560		default:
561			goto badcode;
562		}
563		goto deliver;
564
565	case ICMP6_PACKET_TOO_BIG:
566		icmp6_ifstat_inc(rcvif, ifs6_in_pkttoobig);
567
568		/*
569		 * MTU is checked in icmp6_mtudisc.
570		 */
571		code = PRC_MSGSIZE;
572
573		/*
574		 * Updating the path MTU will be done after examining
575		 * intermediate extension headers.
576		 */
577		goto deliver;
578
579	case ICMP6_TIME_EXCEEDED:
580		icmp6_ifstat_inc(rcvif, ifs6_in_timeexceed);
581		switch (code) {
582		case ICMP6_TIME_EXCEED_TRANSIT:
583			code = PRC_TIMXCEED_INTRANS;
584			break;
585		case ICMP6_TIME_EXCEED_REASSEMBLY:
586			code = PRC_TIMXCEED_REASS;
587			break;
588		default:
589			goto badcode;
590		}
591		goto deliver;
592
593	case ICMP6_PARAM_PROB:
594		icmp6_ifstat_inc(rcvif, ifs6_in_paramprob);
595		switch (code) {
596		case ICMP6_PARAMPROB_NEXTHEADER:
597			code = PRC_UNREACH_PROTOCOL;
598			break;
599		case ICMP6_PARAMPROB_HEADER:
600		case ICMP6_PARAMPROB_OPTION:
601			code = PRC_PARAMPROB;
602			break;
603		default:
604			goto badcode;
605		}
606		goto deliver;
607
608	case ICMP6_ECHO_REQUEST:
609		icmp6_ifstat_inc(rcvif, ifs6_in_echo);
610		if (code != 0)
611			goto badcode;
612		/*
613		 * Copy mbuf to send to two data paths: userland socket(s),
614		 * and to the querier (echo reply).
615		 * m: a copy for socket, n: a copy for querier
616		 *
617		 * If the first mbuf is shared, or the first mbuf is too short,
618		 * copy the first part of the data into a fresh mbuf.
619		 * Otherwise, we will wrongly overwrite both copies.
620		 */
621		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
622			/* Give up local */
623			n = m;
624			m = NULL;
625		} else if (M_READONLY(n) ||
626		    n->m_len < off + sizeof(struct icmp6_hdr)) {
627			struct mbuf *n0 = n;
628
629			/*
630			 * Prepare an internal mbuf.  m_pullup() doesn't
631			 * always copy the length we specified.
632			 */
633			if ((n = m_dup(n0, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
634				/* Give up local */
635				n = m;
636				m = NULL;
637			}
638			m_freem(n0);
639		}
640		IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
641		    sizeof(*nicmp6));
642		if (nicmp6 == NULL)
643			goto freeit;
644		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
645		nicmp6->icmp6_code = 0;
646		if (n) {
647			uint64_t *icmp6s = ICMP6_STAT_GETREF();
648			icmp6s[ICMP6_STAT_REFLECT]++;
649			icmp6s[ICMP6_STAT_OUTHIST + ICMP6_ECHO_REPLY]++;
650			ICMP6_STAT_PUTREF();
651			icmp6_reflect(n, off);
652		}
653		if (!m)
654			goto freeit;
655		break;
656
657	case ICMP6_ECHO_REPLY:
658		icmp6_ifstat_inc(rcvif, ifs6_in_echoreply);
659		if (code != 0)
660			goto badcode;
661		break;
662
663	case MLD_LISTENER_QUERY:
664	case MLD_LISTENER_REPORT:
665		if (icmp6len < sizeof(struct mld_hdr))
666			goto badlen;
667		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
668			icmp6_ifstat_inc(rcvif, ifs6_in_mldquery);
669		else
670			icmp6_ifstat_inc(rcvif, ifs6_in_mldreport);
671		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
672			/* give up local */
673			mld_input(m, off);
674			m = NULL;
675			goto freeit;
676		}
677		mld_input(n, off);
678		/* m stays. */
679		break;
680
681	case MLD_LISTENER_DONE:
682		icmp6_ifstat_inc(rcvif, ifs6_in_mlddone);
683		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
684			goto badlen;
685		break;		/* nothing to be done in kernel */
686
687	case MLD_MTRACE_RESP:
688	case MLD_MTRACE:
689		/* XXX: these two are experimental.  not officially defined. */
690		/* XXX: per-interface statistics? */
691		break;		/* just pass it to applications */
692
693	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
694	    {
695		enum { WRU, FQDN } mode;
696
697		if (!icmp6_nodeinfo)
698			break;
699
700		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
701			mode = WRU;
702		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
703			mode = FQDN;
704		else
705			goto badlen;
706
707		if (mode == FQDN) {
708			n = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
709			if (n)
710				n = ni6_input(n, off);
711			/* XXX meaningless if n == NULL */
712			noff = sizeof(struct ip6_hdr);
713		} else {
714			u_char *p;
715			int maxhlen;
716
717			if ((icmp6_nodeinfo & 5) != 5)
718				break;
719
720			if (code != 0)
721				goto badcode;
722			MGETHDR(n, M_DONTWAIT, m->m_type);
723			if (n && ICMP6_MAXLEN > MHLEN) {
724				MCLGET(n, M_DONTWAIT);
725				if ((n->m_flags & M_EXT) == 0) {
726					m_free(n);
727					n = NULL;
728				}
729			}
730			if (n == NULL) {
731				/* Give up remote */
732				break;
733			}
734			m_reset_rcvif(n);
735			n->m_len = 0;
736			maxhlen = M_TRAILINGSPACE(n) - ICMP6_MAXLEN;
737			if (maxhlen < 0)
738				break;
739			if (maxhlen > hostnamelen)
740				maxhlen = hostnamelen;
741			/*
742			 * Copy IPv6 and ICMPv6 only.
743			 */
744			nip6 = mtod(n, struct ip6_hdr *);
745			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
746			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
747			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
748			p = (u_char *)(nicmp6 + 1);
749			memset(p, 0, 4);
750			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
751			noff = sizeof(struct ip6_hdr);
752			M_COPY_PKTHDR(n, m); /* just for rcvif */
753			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
754				sizeof(struct icmp6_hdr) + 4 + maxhlen;
755			nicmp6->icmp6_type = ICMP6_WRUREPLY;
756			nicmp6->icmp6_code = 0;
757		}
758#undef hostnamelen
759		if (n) {
760			uint64_t *icmp6s = ICMP6_STAT_GETREF();
761			icmp6s[ICMP6_STAT_REFLECT]++;
762			icmp6s[ICMP6_STAT_OUTHIST + ICMP6_WRUREPLY]++;
763			ICMP6_STAT_PUTREF();
764			icmp6_reflect(n, noff);
765		}
766		break;
767	    }
768
769	case ICMP6_WRUREPLY:
770		if (code != 0)
771			goto badcode;
772		break;
773
774	case ND_ROUTER_SOLICIT:
775		icmp6_ifstat_inc(rcvif, ifs6_in_routersolicit);
776		if (code != 0)
777			goto badcode;
778		if (icmp6len < sizeof(struct nd_router_solicit))
779			goto badlen;
780		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
781			/* give up local */
782			nd6_rs_input(m, off, icmp6len);
783			m = NULL;
784			goto freeit;
785		}
786		nd6_rs_input(n, off, icmp6len);
787		/* m stays. */
788		break;
789
790	case ND_ROUTER_ADVERT:
791		icmp6_ifstat_inc(rcvif, ifs6_in_routeradvert);
792		if (code != 0)
793			goto badcode;
794		if (icmp6len < sizeof(struct nd_router_advert))
795			goto badlen;
796		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
797			/* give up local */
798			nd6_ra_input(m, off, icmp6len);
799			m = NULL;
800			goto freeit;
801		}
802		nd6_ra_input(n, off, icmp6len);
803		/* m stays. */
804		break;
805
806	case ND_NEIGHBOR_SOLICIT:
807		icmp6_ifstat_inc(rcvif, ifs6_in_neighborsolicit);
808		if (code != 0)
809			goto badcode;
810		if (icmp6len < sizeof(struct nd_neighbor_solicit))
811			goto badlen;
812		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
813			/* give up local */
814			nd6_ns_input(m, off, icmp6len);
815			m = NULL;
816			goto freeit;
817		}
818		nd6_ns_input(n, off, icmp6len);
819		/* m stays. */
820		break;
821
822	case ND_NEIGHBOR_ADVERT:
823		icmp6_ifstat_inc(rcvif, ifs6_in_neighboradvert);
824		if (code != 0)
825			goto badcode;
826		if (icmp6len < sizeof(struct nd_neighbor_advert))
827			goto badlen;
828		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
829			/* give up local */
830			nd6_na_input(m, off, icmp6len);
831			m = NULL;
832			goto freeit;
833		}
834		nd6_na_input(n, off, icmp6len);
835		/* m stays. */
836		break;
837
838	case ND_REDIRECT:
839		icmp6_ifstat_inc(rcvif, ifs6_in_redirect);
840		if (code != 0)
841			goto badcode;
842		if (icmp6len < sizeof(struct nd_redirect))
843			goto badlen;
844		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
845			/* give up local */
846			icmp6_redirect_input(m, off);
847			m = NULL;
848			goto freeit;
849		}
850		icmp6_redirect_input(n, off);
851		/* m stays. */
852		break;
853
854	case ICMP6_ROUTER_RENUMBERING:
855		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
856		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
857			goto badcode;
858		if (icmp6len < sizeof(struct icmp6_router_renum))
859			goto badlen;
860		break;
861
862	default:
863		nd6log(LOG_DEBUG, "unknown type %d(src=%s, dst=%s, ifid=%d)\n",
864		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
865		    ip6_sprintf(&ip6->ip6_dst),
866		    rcvif ? rcvif->if_index : 0);
867		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
868			/* ICMPv6 error: MUST deliver it by spec... */
869			code = PRC_NCMDS;
870			/* deliver */
871		} else {
872			/* ICMPv6 informational: MUST not deliver */
873			break;
874		}
875	deliver:
876		if (icmp6_notify_error(m, off, icmp6len, code)) {
877			/* In this case, m should've been freed. */
878			m_put_rcvif_psref(rcvif, &psref);
879			return (IPPROTO_DONE);
880		}
881		break;
882
883	badcode:
884		ICMP6_STATINC(ICMP6_STAT_BADCODE);
885		break;
886
887	badlen:
888		ICMP6_STATINC(ICMP6_STAT_BADLEN);
889		break;
890	}
891	m_put_rcvif_psref(rcvif, &psref);
892
893	/* deliver the packet to appropriate sockets */
894	icmp6_rip6_input(&m, *offp);
895
896	return IPPROTO_DONE;
897
898 freeit:
899	m_put_rcvif_psref(rcvif, &psref);
900	m_freem(m);
901	return IPPROTO_DONE;
902}
903
904static int
905icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
906{
907	struct icmp6_hdr *icmp6;
908	struct ip6_hdr *eip6;
909	u_int32_t notifymtu;
910	struct sockaddr_in6 icmp6src, icmp6dst;
911
912	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
913		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
914		goto freeit;
915	}
916	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
917		       sizeof(*icmp6) + sizeof(struct ip6_hdr));
918	if (icmp6 == NULL) {
919		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
920		return (-1);
921	}
922	eip6 = (struct ip6_hdr *)(icmp6 + 1);
923
924	/* Detect the upper level protocol */
925	{
926		void (*ctlfunc)(int, struct sockaddr *, void *);
927		u_int8_t nxt = eip6->ip6_nxt;
928		int eoff = off + sizeof(struct icmp6_hdr) +
929			sizeof(struct ip6_hdr);
930		struct ip6ctlparam ip6cp;
931		struct in6_addr *finaldst = NULL;
932		int icmp6type = icmp6->icmp6_type;
933		struct ip6_frag *fh;
934		struct ip6_rthdr *rth;
935		struct ip6_rthdr0 *rth0;
936		int rthlen;
937		struct ifnet *rcvif;
938		int s;
939
940		while (1) { /* XXX: should avoid infinite loop explicitly? */
941			struct ip6_ext *eh;
942
943			switch (nxt) {
944			case IPPROTO_HOPOPTS:
945			case IPPROTO_DSTOPTS:
946			case IPPROTO_AH:
947				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
948					       eoff, sizeof(*eh));
949				if (eh == NULL) {
950					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
951					return (-1);
952				}
953
954				if (nxt == IPPROTO_AH)
955					eoff += (eh->ip6e_len + 2) << 2;
956				else
957					eoff += (eh->ip6e_len + 1) << 3;
958				nxt = eh->ip6e_nxt;
959				break;
960			case IPPROTO_ROUTING:
961				/*
962				 * When the erroneous packet contains a
963				 * routing header, we should examine the
964				 * header to determine the final destination.
965				 * Otherwise, we can't properly update
966				 * information that depends on the final
967				 * destination (e.g. path MTU).
968				 */
969				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
970					       eoff, sizeof(*rth));
971				if (rth == NULL) {
972					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
973					return (-1);
974				}
975				rthlen = (rth->ip6r_len + 1) << 3;
976				/*
977				 * XXX: currently there is no
978				 * officially defined type other
979				 * than type-0.
980				 * Note that if the segment left field
981				 * is 0, all intermediate hops must
982				 * have been passed.
983				 */
984				if (rth->ip6r_segleft &&
985				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
986					int hops;
987
988					IP6_EXTHDR_GET(rth0,
989						       struct ip6_rthdr0 *, m,
990						       eoff, rthlen);
991					if (rth0 == NULL) {
992						ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
993						return (-1);
994					}
995					/* just ignore a bogus header */
996					if ((rth0->ip6r0_len % 2) == 0 &&
997					    (hops = rth0->ip6r0_len/2))
998						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
999				}
1000				eoff += rthlen;
1001				nxt = rth->ip6r_nxt;
1002				break;
1003			case IPPROTO_FRAGMENT:
1004				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1005					       eoff, sizeof(*fh));
1006				if (fh == NULL) {
1007					ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
1008					return (-1);
1009				}
1010				/*
1011				 * Data after a fragment header is meaningless
1012				 * unless it is the first fragment, but
1013				 * we'll go to the notify label for path MTU
1014				 * discovery.
1015				 */
1016				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1017					goto notify;
1018
1019				eoff += sizeof(struct ip6_frag);
1020				nxt = fh->ip6f_nxt;
1021				break;
1022			default:
1023				/*
1024				 * This case includes ESP and the No Next
1025				 * Header.  In such cases going to the notify
1026				 * label does not have any meaning
1027				 * (i.e. ctlfunc will be NULL), but we go
1028				 * anyway since we might have to update
1029				 * path MTU information.
1030				 */
1031				goto notify;
1032			}
1033		}
1034	  notify:
1035		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1036			       sizeof(*icmp6) + sizeof(struct ip6_hdr));
1037		if (icmp6 == NULL) {
1038			ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
1039			return (-1);
1040		}
1041
1042		/*
1043		 * retrieve parameters from the inner IPv6 header, and convert
1044		 * them into sockaddr structures.
1045		 * XXX: there is no guarantee that the source or destination
1046		 * addresses of the inner packet are in the same scope zone as
1047		 * the addresses of the icmp packet.  But there is no other
1048		 * way to determine the zone.
1049		 */
1050		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1051
1052		rcvif = m_get_rcvif(m, &s);
1053		sockaddr_in6_init(&icmp6dst,
1054		    (finaldst == NULL) ? &eip6->ip6_dst : finaldst, 0, 0, 0);
1055		if (in6_setscope(&icmp6dst.sin6_addr, rcvif, NULL)) {
1056			m_put_rcvif(rcvif, &s);
1057			goto freeit;
1058		}
1059		sockaddr_in6_init(&icmp6src, &eip6->ip6_src, 0, 0, 0);
1060		if (in6_setscope(&icmp6src.sin6_addr, rcvif, NULL)) {
1061			m_put_rcvif(rcvif, &s);
1062			goto freeit;
1063		}
1064		m_put_rcvif(rcvif, &s);
1065
1066		icmp6src.sin6_flowinfo =
1067			(eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1068
1069		if (finaldst == NULL)
1070			finaldst = &eip6->ip6_dst;
1071		ip6cp.ip6c_m = m;
1072		ip6cp.ip6c_icmp6 = icmp6;
1073		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1074		ip6cp.ip6c_off = eoff;
1075		ip6cp.ip6c_finaldst = finaldst;
1076		ip6cp.ip6c_src = &icmp6src;
1077		ip6cp.ip6c_nxt = nxt;
1078
1079		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1080			notifymtu = ntohl(icmp6->icmp6_mtu);
1081			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1082		}
1083
1084		ctlfunc = (void (*)(int, struct sockaddr *, void *))
1085			(inet6sw[ip6_protox[nxt]].pr_ctlinput);
1086		if (ctlfunc) {
1087			(void) (*ctlfunc)(code, sin6tosa(&icmp6dst),
1088					  &ip6cp);
1089		}
1090	}
1091	return (0);
1092
1093  freeit:
1094	m_freem(m);
1095	return (-1);
1096}
1097
1098void
1099icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1100{
1101	unsigned long rtcount;
1102	struct icmp6_mtudisc_callback *mc;
1103	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1104	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1105	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1106	u_int mtu = ntohl(icmp6->icmp6_mtu);
1107	struct rtentry *rt = NULL;
1108	struct sockaddr_in6 sin6;
1109	struct ifnet *rcvif;
1110	int s;
1111
1112	/*
1113	 * The MTU should not be less than the minimal IPv6 MTU except for the
1114	 * hack in ip6_output/ip6_setpmtu where we always include a frag header.
1115	 * In that one case, the MTU might be less than 1280.
1116	 */
1117	if (__predict_false(mtu < IPV6_MMTU - sizeof(struct ip6_frag))) {
1118		/* is the mtu even sane? */
1119		if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1120			return;
1121		if (!validated)
1122			return;
1123		mtu = IPV6_MMTU - sizeof(struct ip6_frag);
1124	}
1125
1126	/*
1127	 * allow non-validated cases if memory is plenty, to make traffic
1128	 * from non-connected pcb happy.
1129	 */
1130	rtcount = rt_timer_count(icmp6_mtudisc_timeout_q);
1131	if (validated) {
1132		if (0 <= icmp6_mtudisc_hiwat && rtcount > icmp6_mtudisc_hiwat)
1133			return;
1134		else if (0 <= icmp6_mtudisc_lowat &&
1135		    rtcount > icmp6_mtudisc_lowat) {
1136			/*
1137			 * XXX nuke a victim, install the new one.
1138			 */
1139		}
1140	} else {
1141		if (0 <= icmp6_mtudisc_lowat && rtcount > icmp6_mtudisc_lowat)
1142			return;
1143	}
1144
1145	memset(&sin6, 0, sizeof(sin6));
1146	sin6.sin6_family = PF_INET6;
1147	sin6.sin6_len = sizeof(struct sockaddr_in6);
1148	sin6.sin6_addr = *dst;
1149	rcvif = m_get_rcvif(m, &s);
1150	if (in6_setscope(&sin6.sin6_addr, rcvif, NULL)) {
1151		m_put_rcvif(rcvif, &s);
1152		return;
1153	}
1154	m_put_rcvif(rcvif, &s);
1155
1156	rt = icmp6_mtudisc_clone(sin6tosa(&sin6));
1157
1158	if (rt && (rt->rt_flags & RTF_HOST) &&
1159	    !(rt->rt_rmx.rmx_locks & RTV_MTU) &&
1160	    (rt->rt_rmx.rmx_mtu > mtu || rt->rt_rmx.rmx_mtu == 0)) {
1161		if (mtu < IN6_LINKMTU(rt->rt_ifp)) {
1162			ICMP6_STATINC(ICMP6_STAT_PMTUCHG);
1163			rt->rt_rmx.rmx_mtu = mtu;
1164		}
1165	}
1166	if (rt) {
1167		rtfree(rt);
1168	}
1169
1170	/*
1171	 * Notify protocols that the MTU for this destination
1172	 * has changed.
1173	 */
1174	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
1175	     mc = LIST_NEXT(mc, mc_list))
1176		(*mc->mc_func)(&sin6.sin6_addr);
1177}
1178
1179/*
1180 * Process a Node Information Query packet, based on
1181 * draft-ietf-ipngwg-icmp-name-lookups-07.
1182 *
1183 * Spec incompatibilities:
1184 * - IPv6 Subject address handling
1185 * - IPv4 Subject address handling support missing
1186 * - Proxy reply (answer even if it's not for me)
1187 * - joins NI group address at in6_ifattach() time only, does not cope
1188 *   with hostname changes by sethostname(3)
1189 */
1190static struct mbuf *
1191ni6_input(struct mbuf *m, int off)
1192{
1193	struct icmp6_nodeinfo *ni6, *nni6;
1194	struct mbuf *n = NULL;
1195	u_int16_t qtype;
1196	int subjlen;
1197	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1198	struct ni_reply_fqdn *fqdn;
1199	int addrs;		/* for NI_QTYPE_NODEADDR */
1200	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1201	struct sockaddr_in6 sin6; /* ip6_dst */
1202	struct in6_addr in6_subj; /* subject address */
1203	struct ip6_hdr *ip6;
1204	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1205	char *subj = NULL;
1206	struct ifnet *rcvif;
1207	int s, ss;
1208	struct ifaddr *ifa;
1209	struct psref psref;
1210
1211	ip6 = mtod(m, struct ip6_hdr *);
1212	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1213	if (ni6 == NULL) {
1214		/* m is already reclaimed */
1215		return NULL;
1216	}
1217
1218	/*
1219	 * Validate IPv6 destination address.
1220	 *
1221	 * The Responder must discard the Query without further processing
1222	 * unless it is one of the Responder's unicast or anycast addresses, or
1223	 * a link-local scope multicast address which the Responder has joined.
1224	 * [icmp-name-lookups-07, Section 4.]
1225	 */
1226	sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0);
1227	/* XXX scopeid */
1228	ss = pserialize_read_enter();
1229	ifa = ifa_ifwithaddr(sin6tosa(&sin6));
1230	if (ifa != NULL)
1231		; /* unicast/anycast, fine */
1232	else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1233		; /* link-local multicast, fine */
1234	else {
1235		pserialize_read_exit(ss);
1236		goto bad;
1237	}
1238	pserialize_read_exit(ss);
1239
1240	/* validate query Subject field. */
1241	qtype = ntohs(ni6->ni_qtype);
1242	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1243	switch (qtype) {
1244	case NI_QTYPE_NOOP:
1245	case NI_QTYPE_SUPTYPES:
1246		/* 07 draft */
1247		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1248			break;
1249		/* FALLTHROUGH */
1250	case NI_QTYPE_FQDN:
1251	case NI_QTYPE_NODEADDR:
1252	case NI_QTYPE_IPV4ADDR:
1253		switch (ni6->ni_code) {
1254		case ICMP6_NI_SUBJ_IPV6:
1255#if ICMP6_NI_SUBJ_IPV6 != 0
1256		case 0:
1257#endif
1258			/*
1259			 * backward compatibility - try to accept 03 draft
1260			 * format, where no Subject is present.
1261			 */
1262			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1263			    subjlen == 0) {
1264				oldfqdn++;
1265				break;
1266			}
1267#if ICMP6_NI_SUBJ_IPV6 != 0
1268			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1269				goto bad;
1270#endif
1271
1272			if (subjlen != sizeof(sin6.sin6_addr))
1273				goto bad;
1274
1275			/*
1276			 * Validate Subject address.
1277			 *
1278			 * Not sure what exactly "address belongs to the node"
1279			 * means in the spec, is it just unicast, or what?
1280			 *
1281			 * At this moment we consider Subject address as
1282			 * "belong to the node" if the Subject address equals
1283			 * to the IPv6 destination address; validation for
1284			 * IPv6 destination address should have done enough
1285			 * check for us.
1286			 *
1287			 * We do not do proxy at this moment.
1288			 */
1289			/* m_pulldown instead of copy? */
1290			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1291			    subjlen, (void *)&in6_subj);
1292			rcvif = m_get_rcvif(m, &s);
1293			if (in6_setscope(&in6_subj, rcvif, NULL)) {
1294				m_put_rcvif(rcvif, &s);
1295				goto bad;
1296			}
1297			m_put_rcvif(rcvif, &s);
1298
1299			subj = (char *)&in6_subj;
1300			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1301				break;
1302
1303			/*
1304			 * XXX if we are to allow other cases, we should really
1305			 * be careful about scope here.
1306			 * basically, we should disallow queries toward IPv6
1307			 * destination X with subject Y, if scope(X) > scope(Y).
1308			 * if we allow scope(X) > scope(Y), it will result in
1309			 * information leakage across scope boundary.
1310			 */
1311			goto bad;
1312
1313		case ICMP6_NI_SUBJ_FQDN:
1314			/*
1315			 * Validate Subject name with gethostname(3).
1316			 *
1317			 * The behavior may need some debate, since:
1318			 * - we are not sure if the node has FQDN as
1319			 *   hostname (returned by gethostname(3)).
1320			 * - the code does wildcard match for truncated names.
1321			 *   however, we are not sure if we want to perform
1322			 *   wildcard match, if gethostname(3) side has
1323			 *   truncated hostname.
1324			 */
1325			n = ni6_nametodns(hostname, hostnamelen, 0);
1326			if (!n || n->m_next || n->m_len == 0)
1327				goto bad;
1328			IP6_EXTHDR_GET(subj, char *, m,
1329			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1330			if (subj == NULL)
1331				goto bad;
1332			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1333					n->m_len)) {
1334				goto bad;
1335			}
1336			m_freem(n);
1337			n = NULL;
1338			break;
1339
1340		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1341		default:
1342			goto bad;
1343		}
1344		break;
1345	}
1346
1347	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1348	switch (qtype) {
1349	case NI_QTYPE_FQDN:
1350		if ((icmp6_nodeinfo & 1) == 0)
1351			goto bad;
1352		break;
1353	case NI_QTYPE_NODEADDR:
1354	case NI_QTYPE_IPV4ADDR:
1355		if ((icmp6_nodeinfo & 2) == 0)
1356			goto bad;
1357		break;
1358	}
1359
1360	/* guess reply length */
1361	switch (qtype) {
1362	case NI_QTYPE_NOOP:
1363		break;		/* no reply data */
1364	case NI_QTYPE_SUPTYPES:
1365		replylen += sizeof(u_int32_t);
1366		break;
1367	case NI_QTYPE_FQDN:
1368		/* XXX will append an mbuf */
1369		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1370		break;
1371	case NI_QTYPE_NODEADDR:
1372		addrs = ni6_addrs(ni6, m, &ifp, subj, &psref);
1373		if ((replylen += addrs * (sizeof(struct in6_addr) +
1374					  sizeof(u_int32_t))) > MCLBYTES)
1375			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1376		break;
1377	case NI_QTYPE_IPV4ADDR:
1378		/* unsupported - should respond with unknown Qtype? */
1379		goto bad;
1380	default:
1381		/*
1382		 * XXX: We must return a reply with the ICMP6 code
1383		 * `unknown Qtype' in this case.  However we regard the case
1384		 * as an FQDN query for backward compatibility.
1385		 * Older versions set a random value to this field,
1386		 * so it rarely varies in the defined qtypes.
1387		 * But the mechanism is not reliable...
1388		 * maybe we should obsolete older versions.
1389		 */
1390		qtype = NI_QTYPE_FQDN;
1391		/* XXX will append an mbuf */
1392		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1393		oldfqdn++;
1394		break;
1395	}
1396
1397	/* allocate an mbuf to reply. */
1398	MGETHDR(n, M_DONTWAIT, m->m_type);
1399	if (n == NULL) {
1400		if_put(ifp, &psref);
1401		m_freem(m);
1402		return (NULL);
1403	}
1404	M_MOVE_PKTHDR(n, m); /* just for rcvif */
1405	if (replylen > MHLEN) {
1406		if (replylen > MCLBYTES) {
1407			/*
1408			 * XXX: should we try to allocate more? But MCLBYTES
1409			 * is probably much larger than IPV6_MMTU...
1410			 */
1411			goto bad;
1412		}
1413		MCLGET(n, M_DONTWAIT);
1414		if ((n->m_flags & M_EXT) == 0) {
1415			goto bad;
1416		}
1417	}
1418	n->m_pkthdr.len = n->m_len = replylen;
1419
1420	/* copy mbuf header and IPv6 + Node Information base headers */
1421	bcopy(mtod(m, void *), mtod(n, void *), sizeof(struct ip6_hdr));
1422	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1423	bcopy((void *)ni6, (void *)nni6, sizeof(struct icmp6_nodeinfo));
1424
1425	/* qtype dependent procedure */
1426	switch (qtype) {
1427	case NI_QTYPE_NOOP:
1428		nni6->ni_code = ICMP6_NI_SUCCESS;
1429		nni6->ni_flags = 0;
1430		break;
1431	case NI_QTYPE_SUPTYPES:
1432	{
1433		u_int32_t v;
1434		nni6->ni_code = ICMP6_NI_SUCCESS;
1435		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1436		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1437		v = (u_int32_t)htonl(0x0000000f);
1438		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1439		break;
1440	}
1441	case NI_QTYPE_FQDN:
1442		nni6->ni_code = ICMP6_NI_SUCCESS;
1443		fqdn = (struct ni_reply_fqdn *)(mtod(n, char *) +
1444						sizeof(struct ip6_hdr) +
1445						sizeof(struct icmp6_nodeinfo));
1446		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1447		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1448		/*
1449		 * XXX do we really have FQDN in variable "hostname"?
1450		 */
1451		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1452		if (n->m_next == NULL)
1453			goto bad;
1454		/* XXX we assume that n->m_next is not a chain */
1455		if (n->m_next->m_next != NULL)
1456			goto bad;
1457		n->m_pkthdr.len += n->m_next->m_len;
1458		break;
1459	case NI_QTYPE_NODEADDR:
1460	{
1461		int lenlim, copied;
1462
1463		nni6->ni_code = ICMP6_NI_SUCCESS;
1464		n->m_pkthdr.len = n->m_len =
1465		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1466		lenlim = M_TRAILINGSPACE(n);
1467		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1468		if_put(ifp, &psref);
1469		ifp = NULL;
1470		/* XXX: reset mbuf length */
1471		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1472			sizeof(struct icmp6_nodeinfo) + copied;
1473		break;
1474	}
1475	default:
1476		break;		/* XXX impossible! */
1477	}
1478
1479	nni6->ni_type = ICMP6_NI_REPLY;
1480	m_freem(m);
1481	return (n);
1482
1483  bad:
1484	if_put(ifp, &psref);
1485	m_freem(m);
1486	if (n)
1487		m_freem(n);
1488	return (NULL);
1489}
1490#undef hostnamelen
1491
1492#define isupper(x) ('A' <= (x) && (x) <= 'Z')
1493#define isalpha(x) (('A' <= (x) && (x) <= 'Z') || ('a' <= (x) && (x) <= 'z'))
1494#define isalnum(x) (isalpha(x) || ('0' <= (x) && (x) <= '9'))
1495#define tolower(x) (isupper(x) ? (x) + 'a' - 'A' : (x))
1496
1497/*
1498 * make a mbuf with DNS-encoded string.  no compression support.
1499 *
1500 * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1501 * treated as truncated name (two \0 at the end).  this is a wild guess.
1502 *
1503 * old - return pascal string if non-zero
1504 */
1505static struct mbuf *
1506ni6_nametodns(const char *name, int namelen, int old)
1507{
1508	struct mbuf *m;
1509	char *cp, *ep;
1510	const char *p, *q;
1511	int i, len, nterm;
1512
1513	if (old)
1514		len = namelen + 1;
1515	else
1516		len = MCLBYTES;
1517
1518	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1519	MGET(m, M_DONTWAIT, MT_DATA);
1520	if (m && len > MLEN) {
1521		MCLGET(m, M_DONTWAIT);
1522		if ((m->m_flags & M_EXT) == 0)
1523			goto fail;
1524	}
1525	if (!m)
1526		goto fail;
1527	m->m_next = NULL;
1528
1529	if (old) {
1530		m->m_len = len;
1531		*mtod(m, char *) = namelen;
1532		bcopy(name, mtod(m, char *) + 1, namelen);
1533		return m;
1534	} else {
1535		m->m_len = 0;
1536		cp = mtod(m, char *);
1537		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1538
1539		/* if not certain about my name, return empty buffer */
1540		if (namelen == 0)
1541			return m;
1542
1543		/*
1544		 * guess if it looks like shortened hostname, or FQDN.
1545		 * shortened hostname needs two trailing "\0".
1546		 */
1547		i = 0;
1548		for (p = name; p < name + namelen; p++) {
1549			if (*p == '.')
1550				i++;
1551		}
1552		if (i < 2)
1553			nterm = 2;
1554		else
1555			nterm = 1;
1556
1557		p = name;
1558		while (cp < ep && p < name + namelen) {
1559			i = 0;
1560			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1561				i++;
1562			/* result does not fit into mbuf */
1563			if (cp + i + 1 >= ep)
1564				goto fail;
1565			/*
1566			 * DNS label length restriction, RFC1035 page 8.
1567			 * "i == 0" case is included here to avoid returning
1568			 * 0-length label on "foo..bar".
1569			 */
1570			if (i <= 0 || i >= 64)
1571				goto fail;
1572			*cp++ = i;
1573			if (!isalpha(p[0]) || !isalnum(p[i - 1]))
1574				goto fail;
1575			while (i > 0) {
1576				if (!isalnum(*p) && *p != '-')
1577					goto fail;
1578				if (isupper(*p)) {
1579					*cp++ = tolower(*p);
1580					p++;
1581				} else
1582					*cp++ = *p++;
1583				i--;
1584			}
1585			p = q;
1586			if (p < name + namelen && *p == '.')
1587				p++;
1588		}
1589		/* termination */
1590		if (cp + nterm >= ep)
1591			goto fail;
1592		while (nterm-- > 0)
1593			*cp++ = '\0';
1594		m->m_len = cp - mtod(m, char *);
1595		return m;
1596	}
1597
1598	panic("should not reach here");
1599	/* NOTREACHED */
1600
1601 fail:
1602	if (m)
1603		m_freem(m);
1604	return NULL;
1605}
1606
1607/*
1608 * check if two DNS-encoded string matches.  takes care of truncated
1609 * form (with \0\0 at the end).  no compression support.
1610 * XXX upper/lowercase match (see RFC2065)
1611 */
1612static int
1613ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1614{
1615	const char *a0, *b0;
1616	int l;
1617
1618	/* simplest case - need validation? */
1619	if (alen == blen && memcmp(a, b, alen) == 0)
1620		return 1;
1621
1622	a0 = a;
1623	b0 = b;
1624
1625	/* termination is mandatory */
1626	if (alen < 2 || blen < 2)
1627		return 0;
1628	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1629		return 0;
1630	alen--;
1631	blen--;
1632
1633	while (a - a0 < alen && b - b0 < blen) {
1634		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1635			return 0;
1636
1637		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1638			return 0;
1639		/* we don't support compression yet */
1640		if (a[0] >= 64 || b[0] >= 64)
1641			return 0;
1642
1643		/* truncated case */
1644		if (a[0] == 0 && a - a0 == alen - 1)
1645			return 1;
1646		if (b[0] == 0 && b - b0 == blen - 1)
1647			return 1;
1648		if (a[0] == 0 || b[0] == 0)
1649			return 0;
1650
1651		if (a[0] != b[0])
1652			return 0;
1653		l = a[0];
1654		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1655			return 0;
1656		if (memcmp(a + 1, b + 1, l) != 0)
1657			return 0;
1658
1659		a += 1 + l;
1660		b += 1 + l;
1661	}
1662
1663	if (a - a0 == alen && b - b0 == blen)
1664		return 1;
1665	else
1666		return 0;
1667}
1668
1669/*
1670 * calculate the number of addresses to be returned in the node info reply.
1671 */
1672static int
1673ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m,
1674    struct ifnet **ifpp, char *subj, struct psref *psref)
1675{
1676	struct ifnet *ifp;
1677	struct in6_ifaddr *ia6;
1678	struct ifaddr *ifa;
1679	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1680	int addrs = 0, addrsofif, iffound = 0;
1681	int niflags = ni6->ni_flags;
1682	int s;
1683
1684	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1685		switch (ni6->ni_code) {
1686		case ICMP6_NI_SUBJ_IPV6:
1687			if (subj == NULL) /* must be impossible... */
1688				return (0);
1689			subj_ip6 = (struct sockaddr_in6 *)subj;
1690			break;
1691		default:
1692			/*
1693			 * XXX: we only support IPv6 subject address for
1694			 * this Qtype.
1695			 */
1696			return (0);
1697		}
1698	}
1699
1700	s = pserialize_read_enter();
1701	IFNET_READER_FOREACH(ifp) {
1702		addrsofif = 0;
1703		IFADDR_READER_FOREACH(ifa, ifp) {
1704			if (ifa->ifa_addr->sa_family != AF_INET6)
1705				continue;
1706			ia6 = (struct in6_ifaddr *)ifa;
1707
1708			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1709			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1710					       &ia6->ia_addr.sin6_addr))
1711				iffound = 1;
1712
1713			/*
1714			 * IPv4-mapped addresses can only be returned by a
1715			 * Node Information proxy, since they represent
1716			 * addresses of IPv4-only nodes, which perforce do
1717			 * not implement this protocol.
1718			 * [icmp-name-lookups-07, Section 5.4]
1719			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1720			 * this function at this moment.
1721			 */
1722
1723			/* What do we have to do about ::1? */
1724			switch (in6_addrscope(&ia6->ia_addr.sin6_addr)) {
1725			case IPV6_ADDR_SCOPE_LINKLOCAL:
1726				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1727					continue;
1728				break;
1729			case IPV6_ADDR_SCOPE_SITELOCAL:
1730				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1731					continue;
1732				break;
1733			case IPV6_ADDR_SCOPE_GLOBAL:
1734				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1735					continue;
1736				break;
1737			default:
1738				continue;
1739			}
1740
1741			/*
1742			 * check if anycast is okay.
1743			 * XXX: just experimental.  not in the spec.
1744			 */
1745			if ((ia6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1746			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1747				continue; /* we need only unicast addresses */
1748
1749			addrsofif++; /* count the address */
1750		}
1751		if (iffound) {
1752			if_acquire_NOMPSAFE(ifp, psref);
1753			pserialize_read_exit(s);
1754			*ifpp = ifp;
1755			return (addrsofif);
1756		}
1757
1758		addrs += addrsofif;
1759	}
1760	pserialize_read_exit(s);
1761
1762	return (addrs);
1763}
1764
1765static int
1766ni6_store_addrs(struct icmp6_nodeinfo *ni6,
1767	struct icmp6_nodeinfo *nni6, struct ifnet *ifp0,
1768	int resid)
1769{
1770	struct ifnet *ifp;
1771	struct in6_ifaddr *ia6;
1772	struct ifaddr *ifa;
1773	struct ifnet *ifp_dep = NULL;
1774	int copied = 0, allow_deprecated = 0;
1775	u_char *cp = (u_char *)(nni6 + 1);
1776	int niflags = ni6->ni_flags;
1777	u_int32_t ltime;
1778	int s;
1779
1780	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1781		return (0);	/* needless to copy */
1782
1783	s = pserialize_read_enter();
1784	ifp = ifp0 ? ifp0 : IFNET_READER_FIRST();
1785  again:
1786
1787	for (; ifp; ifp = IFNET_READER_NEXT(ifp))
1788	{
1789		IFADDR_READER_FOREACH(ifa, ifp) {
1790			if (ifa->ifa_addr->sa_family != AF_INET6)
1791				continue;
1792			ia6 = (struct in6_ifaddr *)ifa;
1793
1794			if ((ia6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1795			    allow_deprecated == 0) {
1796				/*
1797				 * prefererred address should be put before
1798				 * deprecated addresses.
1799				 */
1800
1801				/* record the interface for later search */
1802				if (ifp_dep == NULL)
1803					ifp_dep = ifp;
1804
1805				continue;
1806			}
1807			else if ((ia6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1808				 allow_deprecated != 0)
1809				continue; /* we now collect deprecated addrs */
1810
1811			/* What do we have to do about ::1? */
1812			switch (in6_addrscope(&ia6->ia_addr.sin6_addr)) {
1813			case IPV6_ADDR_SCOPE_LINKLOCAL:
1814				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1815					continue;
1816				break;
1817			case IPV6_ADDR_SCOPE_SITELOCAL:
1818				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1819					continue;
1820				break;
1821			case IPV6_ADDR_SCOPE_GLOBAL:
1822				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1823					continue;
1824				break;
1825			default:
1826				continue;
1827			}
1828
1829			/*
1830			 * check if anycast is okay.
1831			 * XXX: just experimental.  not in the spec.
1832			 */
1833			if ((ia6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1834			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1835				continue;
1836
1837			/* now we can copy the address */
1838			if (resid < sizeof(struct in6_addr) +
1839			    sizeof(u_int32_t)) {
1840				/*
1841				 * We give up much more copy.
1842				 * Set the truncate flag and return.
1843				 */
1844				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1845				goto out;
1846			}
1847
1848			/*
1849			 * Set the TTL of the address.
1850			 * The TTL value should be one of the following
1851			 * according to the specification:
1852			 *
1853			 * 1. The remaining lifetime of a DHCP lease on the
1854			 *    address, or
1855			 * 2. The remaining Valid Lifetime of a prefix from
1856			 *    which the address was derived through Stateless
1857			 *    Autoconfiguration.
1858			 *
1859			 * Note that we currently do not support stateful
1860			 * address configuration by DHCPv6, so the former
1861			 * case can't happen.
1862			 *
1863			 * TTL must be 2^31 > TTL >= 0.
1864			 */
1865			if (ia6->ia6_lifetime.ia6t_expire == 0)
1866				ltime = ND6_INFINITE_LIFETIME;
1867			else {
1868				if (ia6->ia6_lifetime.ia6t_expire >
1869				    time_uptime)
1870					ltime = ia6->ia6_lifetime.ia6t_expire -
1871					    time_uptime;
1872				else
1873					ltime = 0;
1874			}
1875			if (ltime > 0x7fffffff)
1876				ltime = 0x7fffffff;
1877			ltime = htonl(ltime);
1878
1879			bcopy(&ltime, cp, sizeof(u_int32_t));
1880			cp += sizeof(u_int32_t);
1881
1882			/* copy the address itself */
1883			bcopy(&ia6->ia_addr.sin6_addr, cp,
1884			      sizeof(struct in6_addr));
1885			in6_clearscope((struct in6_addr *)cp); /* XXX */
1886			cp += sizeof(struct in6_addr);
1887
1888			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1889			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1890		}
1891		if (ifp0)	/* we need search only on the specified IF */
1892			break;
1893	}
1894
1895	if (allow_deprecated == 0 && ifp_dep != NULL) {
1896		ifp = ifp_dep;
1897		allow_deprecated = 1;
1898
1899		goto again;
1900	}
1901out:
1902	pserialize_read_exit(s);
1903	return (copied);
1904}
1905
1906/*
1907 * XXX almost dup'ed code with rip6_input.
1908 */
1909static int
1910icmp6_rip6_input(struct mbuf **mp, int off)
1911{
1912	struct mbuf *m = *mp;
1913	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1914	struct inpcb_hdr *inph;
1915	struct in6pcb *in6p;
1916	struct in6pcb *last = NULL;
1917	struct sockaddr_in6 rip6src;
1918	struct icmp6_hdr *icmp6;
1919	struct mbuf *opts = NULL;
1920
1921	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1922	if (icmp6 == NULL) {
1923		/* m is already reclaimed */
1924		return IPPROTO_DONE;
1925	}
1926
1927	/*
1928	 * XXX: the address may have embedded scope zone ID, which should be
1929	 * hidden from applications.
1930	 */
1931	sockaddr_in6_init(&rip6src, &ip6->ip6_src, 0, 0, 0);
1932	if (sa6_recoverscope(&rip6src)) {
1933		m_freem(m);
1934		return (IPPROTO_DONE);
1935	}
1936
1937	TAILQ_FOREACH(inph, &raw6cbtable.inpt_queue, inph_queue) {
1938		in6p = (struct in6pcb *)inph;
1939		if (in6p->in6p_af != AF_INET6)
1940			continue;
1941		if (in6p->in6p_ip6.ip6_nxt != IPPROTO_ICMPV6)
1942			continue;
1943		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1944		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1945			continue;
1946		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1947		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1948			continue;
1949		if (in6p->in6p_icmp6filt
1950		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1951				 in6p->in6p_icmp6filt))
1952			continue;
1953		if (last) {
1954			struct	mbuf *n;
1955			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1956				if (last->in6p_flags & IN6P_CONTROLOPTS)
1957					ip6_savecontrol(last, &opts, ip6, n);
1958				/* strip intermediate headers */
1959				m_adj(n, off);
1960				if (sbappendaddr(&last->in6p_socket->so_rcv,
1961				    sin6tosa(&rip6src), n, opts) == 0) {
1962					/* should notify about lost packet */
1963					m_freem(n);
1964					if (opts)
1965						m_freem(opts);
1966				} else
1967					sorwakeup(last->in6p_socket);
1968				opts = NULL;
1969			}
1970		}
1971		last = in6p;
1972	}
1973	if (last) {
1974		if (last->in6p_flags & IN6P_CONTROLOPTS)
1975			ip6_savecontrol(last, &opts, ip6, m);
1976		/* strip intermediate headers */
1977		m_adj(m, off);
1978		if (sbappendaddr(&last->in6p_socket->so_rcv,
1979		    sin6tosa(&rip6src), m, opts) == 0) {
1980			m_freem(m);
1981			if (opts)
1982				m_freem(opts);
1983		} else
1984			sorwakeup(last->in6p_socket);
1985	} else {
1986		m_freem(m);
1987		IP6_STATDEC(IP6_STAT_DELIVERED);
1988	}
1989	return IPPROTO_DONE;
1990}
1991
1992/*
1993 * Reflect the ip6 packet back to the source.
1994 * OFF points to the icmp6 header, counted from the top of the mbuf.
1995 *
1996 * Note: RFC 1885 required that an echo reply should be truncated if it
1997 * did not fit in with (return) path MTU, and KAME code supported the
1998 * behavior.  However, as a clarification after the RFC, this limitation
1999 * was removed in a revised version of the spec, RFC 2463.  We had kept the
2000 * old behavior, with a (non-default) ifdef block, while the new version of
2001 * the spec was an internet-draft status, and even after the new RFC was
2002 * published.  But it would rather make sense to clean the obsoleted part
2003 * up, and to make the code simpler at this stage.
2004 */
2005void
2006icmp6_reflect(struct mbuf *m, size_t off)
2007{
2008	struct ip6_hdr *ip6;
2009	struct icmp6_hdr *icmp6;
2010	const struct in6_ifaddr *ia;
2011	const struct ip6aux *ip6a;
2012	int plen;
2013	int type, code;
2014	struct ifnet *outif = NULL;
2015	struct in6_addr origdst;
2016	struct ifnet *rcvif;
2017	int s;
2018	bool ip6_src_filled = false;
2019
2020	/* too short to reflect */
2021	if (off < sizeof(struct ip6_hdr)) {
2022		nd6log(LOG_DEBUG,
2023		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2024		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2025		    __FILE__, __LINE__);
2026		goto bad;
2027	}
2028
2029	/*
2030	 * If there are extra headers between IPv6 and ICMPv6, strip
2031	 * off that header first.
2032	 */
2033#ifdef DIAGNOSTIC
2034	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2035		panic("assumption failed in icmp6_reflect");
2036#endif
2037	if (off > sizeof(struct ip6_hdr)) {
2038		size_t l;
2039		struct ip6_hdr nip6;
2040
2041		l = off - sizeof(struct ip6_hdr);
2042		m_copydata(m, 0, sizeof(nip6), (void *)&nip6);
2043		m_adj(m, l);
2044		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2045		if (m->m_len < l) {
2046			if ((m = m_pullup(m, l)) == NULL)
2047				return;
2048		}
2049		bcopy((void *)&nip6, mtod(m, void *), sizeof(nip6));
2050	} else /* off == sizeof(struct ip6_hdr) */ {
2051		size_t l;
2052		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2053		if (m->m_len < l) {
2054			if ((m = m_pullup(m, l)) == NULL)
2055				return;
2056		}
2057	}
2058	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2059	ip6 = mtod(m, struct ip6_hdr *);
2060	ip6->ip6_nxt = IPPROTO_ICMPV6;
2061	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2062	type = icmp6->icmp6_type; /* keep type for statistics */
2063	code = icmp6->icmp6_code; /* ditto. */
2064
2065	origdst = ip6->ip6_dst;
2066	/*
2067	 * ip6_input() drops a packet if its src is multicast.
2068	 * So, the src is never multicast.
2069	 */
2070	ip6->ip6_dst = ip6->ip6_src;
2071
2072	/*
2073	 * If the incoming packet was addressed directly to us (i.e. unicast),
2074	 * use dst as the src for the reply.
2075	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2076	 * (for example) when we encounter an error while forwarding procedure
2077	 * destined to a duplicated address of ours.
2078	 * Note that ip6_getdstifaddr() may fail if we are in an error handling
2079	 * procedure of an outgoing packet of our own, in which case we need
2080	 * to search in the ifaddr list.
2081	 */
2082	if (IN6_IS_ADDR_MULTICAST(&origdst))
2083		;
2084	else if ((ip6a = ip6_getdstifaddr(m)) != NULL) {
2085		if ((ip6a->ip6a_flags &
2086		     (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2087			ip6->ip6_src = ip6a->ip6a_src;
2088			ip6_src_filled = true;
2089		}
2090	} else {
2091		union {
2092			struct sockaddr_in6 sin6;
2093			struct sockaddr sa;
2094		} u;
2095		int _s;
2096		struct ifaddr *ifa;
2097
2098		sockaddr_in6_init(&u.sin6, &origdst, 0, 0, 0);
2099
2100		_s = pserialize_read_enter();
2101		ifa = ifa_ifwithaddr(&u.sa);
2102
2103		if (ifa != NULL) {
2104			ia = ifatoia6(ifa);
2105			if ((ia->ia6_flags &
2106				 (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2107				ip6->ip6_src = ia->ia_addr.sin6_addr;
2108				ip6_src_filled = true;
2109			}
2110		}
2111		pserialize_read_exit(_s);
2112	}
2113
2114	if (!ip6_src_filled) {
2115		int e;
2116		struct sockaddr_in6 sin6;
2117		struct route ro;
2118
2119		/*
2120		 * This case matches to multicasts, our anycast, or unicasts
2121		 * that we do not own.  Select a source address based on the
2122		 * source address of the erroneous packet.
2123		 */
2124		/* zone ID should be embedded */
2125		sockaddr_in6_init(&sin6, &ip6->ip6_dst, 0, 0, 0);
2126
2127		memset(&ro, 0, sizeof(ro));
2128		e = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, NULL, NULL,
2129		    &ip6->ip6_src);
2130		rtcache_free(&ro);
2131		if (e != 0) {
2132			nd6log(LOG_DEBUG,
2133			    "source can't be determined: "
2134			    "dst=%s, error=%d\n",
2135			    ip6_sprintf(&sin6.sin6_addr), e);
2136			goto bad;
2137		}
2138	}
2139
2140	ip6->ip6_flow = 0;
2141	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2142	ip6->ip6_vfc |= IPV6_VERSION;
2143	ip6->ip6_nxt = IPPROTO_ICMPV6;
2144	rcvif = m_get_rcvif(m, &s);
2145	if (rcvif) {
2146		/* XXX: This may not be the outgoing interface */
2147		ip6->ip6_hlim = ND_IFINFO(rcvif)->chlim;
2148	} else
2149		ip6->ip6_hlim = ip6_defhlim;
2150	m_put_rcvif(rcvif, &s);
2151
2152	m->m_pkthdr.csum_flags = 0;
2153	icmp6->icmp6_cksum = 0;
2154	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2155					sizeof(struct ip6_hdr), plen);
2156
2157	/*
2158	 * XXX option handling
2159	 */
2160
2161	m->m_flags &= ~(M_BCAST|M_MCAST);
2162
2163	/*
2164	 * To avoid a "too big" situation at an intermediate router
2165	 * and the path MTU discovery process, specify the IPV6_MINMTU flag.
2166	 * Note that only echo and node information replies are affected,
2167	 * since the length of ICMP6 errors is limited to the minimum MTU.
2168	 */
2169	if (ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL, &outif)
2170	    != 0 && outif)
2171		icmp6_ifstat_inc(outif, ifs6_out_error);
2172	if (outif)
2173		icmp6_ifoutstat_inc(outif, type, code);
2174
2175	return;
2176
2177 bad:
2178	m_freem(m);
2179	return;
2180}
2181
2182static const char *
2183icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2184	struct in6_addr *tgt6)
2185{
2186	static char buf[1024];
2187	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2188		ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2189	return buf;
2190}
2191
2192void
2193icmp6_redirect_input(struct mbuf *m, int off)
2194{
2195	struct ifnet *ifp;
2196	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2197	struct nd_redirect *nd_rd;
2198	int icmp6len = ntohs(ip6->ip6_plen);
2199	char *lladdr = NULL;
2200	int lladdrlen = 0;
2201	struct rtentry *rt = NULL;
2202	int is_router;
2203	int is_onlink;
2204	struct in6_addr src6 = ip6->ip6_src;
2205	struct in6_addr redtgt6;
2206	struct in6_addr reddst6;
2207	union nd_opts ndopts;
2208	struct psref psref;
2209
2210	ifp = m_get_rcvif_psref(m, &psref);
2211	if (ifp == NULL)
2212		goto freeit;
2213
2214	/* XXX if we are router, we don't update route by icmp6 redirect */
2215	if (ip6_forwarding)
2216		goto freeit;
2217	if (!icmp6_rediraccept)
2218		goto freeit;
2219
2220	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2221	if (nd_rd == NULL) {
2222		ICMP6_STATINC(ICMP6_STAT_TOOSHORT);
2223		m_put_rcvif_psref(ifp, &psref);
2224		return;
2225	}
2226	redtgt6 = nd_rd->nd_rd_target;
2227	reddst6 = nd_rd->nd_rd_dst;
2228
2229	if (in6_setscope(&redtgt6, ifp, NULL) ||
2230	    in6_setscope(&reddst6, ifp, NULL)) {
2231		goto freeit;
2232	}
2233
2234	/* validation */
2235	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2236		nd6log(LOG_ERR,
2237		    "ICMP6 redirect sent from %s rejected; "
2238		    "must be from linklocal\n", ip6_sprintf(&src6));
2239		goto bad;
2240	}
2241	if (ip6->ip6_hlim != 255) {
2242		nd6log(LOG_ERR,
2243		    "ICMP6 redirect sent from %s rejected; "
2244		    "hlim=%d (must be 255)\n",
2245		    ip6_sprintf(&src6), ip6->ip6_hlim);
2246		goto bad;
2247	}
2248    {
2249	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2250	struct sockaddr_in6 sin6;
2251	struct in6_addr *gw6;
2252
2253	sockaddr_in6_init(&sin6, &reddst6, 0, 0, 0);
2254	rt = rtalloc1(sin6tosa(&sin6), 0);
2255	if (rt) {
2256		if (rt->rt_gateway == NULL ||
2257		    rt->rt_gateway->sa_family != AF_INET6) {
2258			nd6log(LOG_ERR,
2259			    "ICMP6 redirect rejected; no route "
2260			    "with inet6 gateway found for redirect dst: %s\n",
2261			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2262			rtfree(rt);
2263			goto bad;
2264		}
2265
2266		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2267		if (memcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2268			nd6log(LOG_ERR,
2269			    "ICMP6 redirect rejected; "
2270			    "not equal to gw-for-src=%s (must be same): %s\n",
2271			    ip6_sprintf(gw6),
2272			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2273			rtfree(rt);
2274			goto bad;
2275		}
2276	} else {
2277		nd6log(LOG_ERR, "ICMP6 redirect rejected; "
2278		    "no route found for redirect dst: %s\n",
2279		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2280		goto bad;
2281	}
2282	rtfree(rt);
2283	rt = NULL;
2284    }
2285	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2286		nd6log(LOG_ERR, "ICMP6 redirect rejected; "
2287		    "redirect dst must be unicast: %s\n",
2288		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2289		goto bad;
2290	}
2291
2292	is_router = is_onlink = 0;
2293	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2294		is_router = 1;	/* router case */
2295	if (memcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2296		is_onlink = 1;	/* on-link destination case */
2297	if (!is_router && !is_onlink) {
2298		nd6log(LOG_ERR, "ICMP6 redirect rejected; "
2299		    "neither router case nor onlink case: %s\n",
2300		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2301		goto bad;
2302	}
2303	/* validation passed */
2304
2305	icmp6len -= sizeof(*nd_rd);
2306	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2307	if (nd6_options(&ndopts) < 0) {
2308		nd6log(LOG_INFO, "invalid ND option, rejected: %s\n",
2309		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2310		/* nd6_options have incremented stats */
2311		goto freeit;
2312	}
2313
2314	if (ndopts.nd_opts_tgt_lladdr) {
2315		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2316		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2317	}
2318
2319	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2320		nd6log(LOG_INFO, "lladdrlen mismatch for %s "
2321		    "(if %d, icmp6 packet %d): %s\n",
2322		    ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2323		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2324		goto bad;
2325	}
2326
2327	/* RFC 2461 8.3 */
2328	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2329			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2330
2331	m_put_rcvif_psref(ifp, &psref);
2332	ifp = NULL;
2333
2334	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2335		/* perform rtredirect */
2336		struct sockaddr_in6 sdst;
2337		struct sockaddr_in6 sgw;
2338		struct sockaddr_in6 ssrc;
2339		unsigned long rtcount;
2340		struct rtentry *newrt = NULL;
2341
2342		/*
2343		 * do not install redirect route, if the number of entries
2344		 * is too much (> hiwat).  note that, the node (= host) will
2345		 * work just fine even if we do not install redirect route
2346		 * (there will be additional hops, though).
2347		 */
2348		rtcount = rt_timer_count(icmp6_redirect_timeout_q);
2349		if (0 <= ip6_maxdynroutes && rtcount >= ip6_maxdynroutes)
2350			goto freeit;
2351		if (0 <= icmp6_redirect_hiwat && rtcount > icmp6_redirect_hiwat)
2352			goto freeit;
2353		else if (0 <= icmp6_redirect_lowat &&
2354		    rtcount > icmp6_redirect_lowat) {
2355			/*
2356			 * XXX nuke a victim, install the new one.
2357			 */
2358		}
2359
2360		memset(&sdst, 0, sizeof(sdst));
2361		memset(&sgw, 0, sizeof(sgw));
2362		memset(&ssrc, 0, sizeof(ssrc));
2363		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2364		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2365			sizeof(struct sockaddr_in6);
2366		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2367		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2368		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2369		rtredirect(sin6tosa(&sdst), sin6tosa(&sgw), NULL,
2370		           RTF_GATEWAY | RTF_HOST, sin6tosa(&ssrc),
2371			   &newrt);
2372
2373		if (newrt) {
2374			(void)rt_timer_add(newrt, icmp6_redirect_timeout,
2375			    icmp6_redirect_timeout_q);
2376			rtfree(newrt);
2377		}
2378	}
2379	/* finally update cached route in each socket via pfctlinput */
2380	{
2381		struct sockaddr_in6 sdst;
2382
2383		sockaddr_in6_init(&sdst, &reddst6, 0, 0, 0);
2384		pfctlinput(PRC_REDIRECT_HOST, sin6tosa(&sdst));
2385#if defined(IPSEC)
2386		if (ipsec_used)
2387			key_sa_routechange(sin6tosa(&sdst));
2388#endif
2389	}
2390
2391 freeit:
2392	if (ifp != NULL)
2393		m_put_rcvif_psref(ifp, &psref);
2394	m_freem(m);
2395	return;
2396
2397 bad:
2398	m_put_rcvif_psref(ifp, &psref);
2399	ICMP6_STATINC(ICMP6_STAT_BADREDIRECT);
2400	m_freem(m);
2401}
2402
2403void
2404icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2405{
2406	struct ifnet *ifp;	/* my outgoing interface */
2407	struct in6_addr *ifp_ll6;
2408	struct in6_addr *nexthop;
2409	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2410	struct mbuf *m = NULL;	/* newly allocated one */
2411	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2412	struct nd_redirect *nd_rd;
2413	size_t maxlen;
2414	u_char *p;
2415	struct sockaddr_in6 src_sa;
2416
2417	icmp6_errcount(ICMP6_STAT_OUTERRHIST, ND_REDIRECT, 0);
2418
2419	/* if we are not router, we don't send icmp6 redirect */
2420	if (!ip6_forwarding)
2421		goto fail;
2422
2423	/* sanity check */
2424	KASSERT(m0 != NULL);
2425	KASSERT(rt != NULL);
2426
2427	ifp = rt->rt_ifp;
2428
2429	/*
2430	 * Address check:
2431	 *  the source address must identify a neighbor, and
2432	 *  the destination address must not be a multicast address
2433	 *  [RFC 2461, sec 8.2]
2434	 */
2435	sip6 = mtod(m0, struct ip6_hdr *);
2436	sockaddr_in6_init(&src_sa, &sip6->ip6_src, 0, 0, 0);
2437	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2438		goto fail;
2439	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2440		goto fail;	/* what should we do here? */
2441
2442	/* rate limit */
2443	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2444		goto fail;
2445
2446	/*
2447	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2448	 * we almost always ask for an mbuf cluster for simplicity.
2449	 * (MHLEN < IPV6_MMTU is almost always true)
2450	 */
2451	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2452	if (m && IPV6_MMTU >= MHLEN) {
2453#if IPV6_MMTU >= MCLBYTES
2454		_MCLGET(m, mcl_cache, IPV6_MMTU, M_DONTWAIT);
2455#else
2456		MCLGET(m, M_DONTWAIT);
2457#endif
2458	}
2459
2460	if (!m)
2461		goto fail;
2462	m_reset_rcvif(m);
2463	m->m_len = 0;
2464	maxlen = M_TRAILINGSPACE(m);
2465	maxlen = min(IPV6_MMTU, maxlen);
2466	/* just for safety */
2467	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2468	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2469		goto fail;
2470	}
2471
2472	{
2473		/* get ip6 linklocal address for ifp(my outgoing interface). */
2474		struct in6_ifaddr *ia;
2475		int s = pserialize_read_enter();
2476		if ((ia = in6ifa_ifpforlinklocal(ifp,
2477						 IN6_IFF_NOTREADY|
2478						 IN6_IFF_ANYCAST)) == NULL) {
2479			pserialize_read_exit(s);
2480			goto fail;
2481		}
2482		ifp_ll6 = &ia->ia_addr.sin6_addr;
2483		pserialize_read_exit(s);
2484	}
2485
2486	/* get ip6 linklocal address for the router. */
2487	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2488		struct sockaddr_in6 *sin6;
2489		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2490		nexthop = &sin6->sin6_addr;
2491		if (!IN6_IS_ADDR_LINKLOCAL(nexthop))
2492			nexthop = NULL;
2493	} else
2494		nexthop = NULL;
2495
2496	/* ip6 */
2497	ip6 = mtod(m, struct ip6_hdr *);
2498	ip6->ip6_flow = 0;
2499	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2500	ip6->ip6_vfc |= IPV6_VERSION;
2501	/* ip6->ip6_plen will be set later */
2502	ip6->ip6_nxt = IPPROTO_ICMPV6;
2503	ip6->ip6_hlim = 255;
2504	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2505	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2506	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2507
2508	/* ND Redirect */
2509	nd_rd = (struct nd_redirect *)(ip6 + 1);
2510	nd_rd->nd_rd_type = ND_REDIRECT;
2511	nd_rd->nd_rd_code = 0;
2512	nd_rd->nd_rd_reserved = 0;
2513	if (rt->rt_flags & RTF_GATEWAY) {
2514		/*
2515		 * nd_rd->nd_rd_target must be a link-local address in
2516		 * better router cases.
2517		 */
2518		if (!nexthop)
2519			goto fail;
2520		bcopy(nexthop, &nd_rd->nd_rd_target,
2521		      sizeof(nd_rd->nd_rd_target));
2522		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2523		      sizeof(nd_rd->nd_rd_dst));
2524	} else {
2525		/* make sure redtgt == reddst */
2526		nexthop = &sip6->ip6_dst;
2527		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2528		      sizeof(nd_rd->nd_rd_target));
2529		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2530		      sizeof(nd_rd->nd_rd_dst));
2531	}
2532
2533	p = (u_char *)(nd_rd + 1);
2534
2535	{
2536		/* target lladdr option */
2537		struct llentry *ln = NULL;
2538		int len;
2539		struct nd_opt_hdr *nd_opt;
2540		char *lladdr;
2541
2542		ln = nd6_lookup(nexthop, ifp, false);
2543		if (ln == NULL)
2544			goto nolladdropt;
2545		len = sizeof(*nd_opt) + ifp->if_addrlen;
2546		len = (len + 7) & ~7;	/* round by 8 */
2547		/* safety check */
2548		if (len + (p - (u_char *)ip6) > maxlen) {
2549			LLE_RUNLOCK(ln);
2550			goto nolladdropt;
2551		}
2552		if (ln->la_flags & LLE_VALID) {
2553			nd_opt = (struct nd_opt_hdr *)p;
2554			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2555			nd_opt->nd_opt_len = len >> 3;
2556			lladdr = (char *)(nd_opt + 1);
2557			memcpy(lladdr, &ln->ll_addr, ifp->if_addrlen);
2558			p += len;
2559		}
2560		LLE_RUNLOCK(ln);
2561	}
2562  nolladdropt:;
2563
2564	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2565
2566	/* just to be safe */
2567	if (m0->m_flags & M_DECRYPTED)
2568		goto noredhdropt;
2569	if (p - (u_char *)ip6 > maxlen)
2570		goto noredhdropt;
2571
2572	{
2573		/* redirected header option */
2574		int len;
2575		struct nd_opt_rd_hdr *nd_opt_rh;
2576
2577		/*
2578		 * compute the maximum size for icmp6 redirect header option.
2579		 * XXX room for auth header?
2580		 */
2581		len = maxlen - (p - (u_char *)ip6);
2582		len &= ~7;
2583
2584		/*
2585		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2586		 * about padding/truncate rule for the original IP packet.
2587		 * From the discussion on IPv6imp in Feb 1999,
2588		 * the consensus was:
2589		 * - "attach as much as possible" is the goal
2590		 * - pad if not aligned (original size can be guessed by
2591		 *   original ip6 header)
2592		 * Following code adds the padding if it is simple enough,
2593		 * and truncates if not.
2594		 */
2595		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2596			/* not enough room, truncate */
2597			m_adj(m0, (len - sizeof(*nd_opt_rh)) -
2598			    m0->m_pkthdr.len);
2599		} else {
2600			/*
2601			 * enough room, truncate if not aligned.
2602			 * we don't pad here for simplicity.
2603			 */
2604			size_t extra;
2605
2606			extra = m0->m_pkthdr.len % 8;
2607			if (extra) {
2608				/* truncate */
2609				m_adj(m0, -extra);
2610			}
2611			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2612		}
2613
2614		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2615		memset(nd_opt_rh, 0, sizeof(*nd_opt_rh));
2616		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2617		nd_opt_rh->nd_opt_rh_len = len >> 3;
2618		p += sizeof(*nd_opt_rh);
2619		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2620
2621		/* connect m0 to m */
2622		m->m_pkthdr.len += m0->m_pkthdr.len;
2623		m_cat(m, m0);
2624		m0 = NULL;
2625	}
2626noredhdropt:
2627	if (m0) {
2628		m_freem(m0);
2629		m0 = NULL;
2630	}
2631
2632	/* XXX: clear embedded link IDs in the inner header */
2633	in6_clearscope(&sip6->ip6_src);
2634	in6_clearscope(&sip6->ip6_dst);
2635	in6_clearscope(&nd_rd->nd_rd_target);
2636	in6_clearscope(&nd_rd->nd_rd_dst);
2637
2638	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2639
2640	nd_rd->nd_rd_cksum = 0;
2641	nd_rd->nd_rd_cksum
2642		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2643
2644	/* send the packet to outside... */
2645	if (ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL) != 0)
2646		icmp6_ifstat_inc(ifp, ifs6_out_error);
2647
2648	icmp6_ifstat_inc(ifp, ifs6_out_msg);
2649	icmp6_ifstat_inc(ifp, ifs6_out_redirect);
2650	ICMP6_STATINC(ICMP6_STAT_OUTHIST + ND_REDIRECT);
2651
2652	return;
2653
2654fail:
2655	if (m)
2656		m_freem(m);
2657	if (m0)
2658		m_freem(m0);
2659}
2660
2661/*
2662 * ICMPv6 socket option processing.
2663 */
2664int
2665icmp6_ctloutput(int op, struct socket *so, struct sockopt *sopt)
2666{
2667	int error = 0;
2668	struct in6pcb *in6p = sotoin6pcb(so);
2669
2670	if (sopt->sopt_level != IPPROTO_ICMPV6)
2671		return rip6_ctloutput(op, so, sopt);
2672
2673	switch (op) {
2674	case PRCO_SETOPT:
2675		switch (sopt->sopt_name) {
2676		case ICMP6_FILTER:
2677		    {
2678			struct icmp6_filter fil;
2679
2680			error = sockopt_get(sopt, &fil, sizeof(fil));
2681			if (error)
2682				break;
2683			memcpy(in6p->in6p_icmp6filt, &fil,
2684			    sizeof(struct icmp6_filter));
2685			error = 0;
2686			break;
2687		    }
2688
2689		default:
2690			error = ENOPROTOOPT;
2691			break;
2692		}
2693		break;
2694
2695	case PRCO_GETOPT:
2696		switch (sopt->sopt_name) {
2697		case ICMP6_FILTER:
2698		    {
2699			if (in6p->in6p_icmp6filt == NULL) {
2700				error = EINVAL;
2701				break;
2702			}
2703			error = sockopt_set(sopt, in6p->in6p_icmp6filt,
2704			    sizeof(struct icmp6_filter));
2705			break;
2706		    }
2707
2708		default:
2709			error = ENOPROTOOPT;
2710			break;
2711		}
2712		break;
2713	}
2714
2715	return (error);
2716}
2717
2718/*
2719 * Perform rate limit check.
2720 * Returns 0 if it is okay to send the icmp6 packet.
2721 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2722 * limitation.
2723 *
2724 * XXX per-destination/type check necessary?
2725 */
2726static int
2727icmp6_ratelimit(
2728	const struct in6_addr *dst,	/* not used at this moment */
2729	const int type,		/* not used at this moment */
2730	const int code)		/* not used at this moment */
2731{
2732	int ret;
2733
2734	ret = 0;	/* okay to send */
2735
2736	/* PPS limit */
2737	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2738	    icmp6errppslim)) {
2739		/* The packet is subject to rate limit */
2740		ret++;
2741	}
2742
2743	return ret;
2744}
2745
2746static struct rtentry *
2747icmp6_mtudisc_clone(struct sockaddr *dst)
2748{
2749	struct rtentry *rt;
2750	int    error;
2751
2752	rt = rtalloc1(dst, 1);
2753	if (rt == 0)
2754		return NULL;
2755
2756	/* If we didn't get a host route, allocate one */
2757	if ((rt->rt_flags & RTF_HOST) == 0) {
2758		struct rtentry *nrt;
2759
2760		error = rtrequest(RTM_ADD, dst, rt->rt_gateway, NULL,
2761		    RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt);
2762		if (error) {
2763			rtfree(rt);
2764			return NULL;
2765		}
2766		nrt->rt_rmx = rt->rt_rmx;
2767		rtfree(rt);
2768		rt = nrt;
2769	}
2770	error = rt_timer_add(rt, icmp6_mtudisc_timeout,
2771			icmp6_mtudisc_timeout_q);
2772	if (error) {
2773		rtfree(rt);
2774		return NULL;
2775	}
2776
2777	return rt;	/* caller need to call rtfree() */
2778}
2779
2780static void
2781icmp6_mtudisc_timeout(struct rtentry *rt, struct rttimer *r)
2782{
2783
2784	KASSERT(rt != NULL);
2785	rt_assert_referenced(rt);
2786
2787	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
2788	    (RTF_DYNAMIC | RTF_HOST)) {
2789		rtrequest(RTM_DELETE, rt_getkey(rt),
2790		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL);
2791	} else {
2792		if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
2793			rt->rt_rmx.rmx_mtu = 0;
2794	}
2795}
2796
2797static void
2798icmp6_redirect_timeout(struct rtentry *rt, struct rttimer *r)
2799{
2800
2801	KASSERT(rt != NULL);
2802	rt_assert_referenced(rt);
2803
2804	if ((rt->rt_flags & (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) ==
2805	    (RTF_GATEWAY | RTF_DYNAMIC | RTF_HOST)) {
2806		rtrequest(RTM_DELETE, rt_getkey(rt),
2807		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, NULL);
2808	}
2809}
2810
2811/*
2812 * sysctl helper routine for the net.inet6.icmp6.nd6 nodes.  silly?
2813 */
2814static int
2815sysctl_net_inet6_icmp6_nd6(SYSCTLFN_ARGS)
2816{
2817	(void)&name;
2818	(void)&l;
2819	(void)&oname;
2820
2821	if (namelen != 0)
2822		return (EINVAL);
2823
2824	return (nd6_sysctl(rnode->sysctl_num, oldp, oldlenp,
2825	    /*XXXUNCONST*/
2826	    __UNCONST(newp), newlen));
2827}
2828
2829static int
2830sysctl_net_inet6_icmp6_stats(SYSCTLFN_ARGS)
2831{
2832
2833	return (NETSTAT_SYSCTL(icmp6stat_percpu, ICMP6_NSTATS));
2834}
2835
2836static int
2837sysctl_net_inet6_icmp6_redirtimeout(SYSCTLFN_ARGS)
2838{
2839	int error, tmp;
2840	struct sysctlnode node;
2841
2842	node = *rnode;
2843	node.sysctl_data = &tmp;
2844	tmp = icmp6_redirtimeout;
2845	error = sysctl_lookup(SYSCTLFN_CALL(&node));
2846	if (error || newp == NULL)
2847		return error;
2848	if (tmp < 0)
2849		return EINVAL;
2850	icmp6_redirtimeout = tmp;
2851
2852	if (icmp6_redirect_timeout_q != NULL) {
2853		if (icmp6_redirtimeout == 0) {
2854			rt_timer_queue_destroy(icmp6_redirect_timeout_q);
2855		} else {
2856			rt_timer_queue_change(icmp6_redirect_timeout_q,
2857			    icmp6_redirtimeout);
2858		}
2859	} else if (icmp6_redirtimeout > 0) {
2860		icmp6_redirect_timeout_q =
2861		    rt_timer_queue_create(icmp6_redirtimeout);
2862	}
2863
2864	return 0;
2865}
2866
2867static void
2868sysctl_net_inet6_icmp6_setup(struct sysctllog **clog)
2869{
2870	extern int nd6_maxqueuelen; /* defined in nd6.c */
2871
2872	sysctl_createv(clog, 0, NULL, NULL,
2873		       CTLFLAG_PERMANENT,
2874		       CTLTYPE_NODE, "inet6", NULL,
2875		       NULL, 0, NULL, 0,
2876		       CTL_NET, PF_INET6, CTL_EOL);
2877	sysctl_createv(clog, 0, NULL, NULL,
2878		       CTLFLAG_PERMANENT,
2879		       CTLTYPE_NODE, "icmp6",
2880		       SYSCTL_DESCR("ICMPv6 related settings"),
2881		       NULL, 0, NULL, 0,
2882		       CTL_NET, PF_INET6, IPPROTO_ICMPV6, CTL_EOL);
2883
2884	sysctl_createv(clog, 0, NULL, NULL,
2885		       CTLFLAG_PERMANENT,
2886		       CTLTYPE_STRUCT, "stats",
2887		       SYSCTL_DESCR("ICMPv6 transmission statistics"),
2888		       sysctl_net_inet6_icmp6_stats, 0, NULL, 0,
2889		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2890		       ICMPV6CTL_STATS, CTL_EOL);
2891	sysctl_createv(clog, 0, NULL, NULL,
2892		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2893		       CTLTYPE_INT, "rediraccept",
2894		       SYSCTL_DESCR("Accept and process redirect messages"),
2895		       NULL, 0, &icmp6_rediraccept, 0,
2896		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2897		       ICMPV6CTL_REDIRACCEPT, CTL_EOL);
2898	sysctl_createv(clog, 0, NULL, NULL,
2899		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2900		       CTLTYPE_INT, "redirtimeout",
2901		       SYSCTL_DESCR("Redirect generated route lifetime"),
2902		       sysctl_net_inet6_icmp6_redirtimeout, 0,
2903		       &icmp6_redirtimeout, 0,
2904		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2905		       ICMPV6CTL_REDIRTIMEOUT, CTL_EOL);
2906#if 0 /* obsoleted */
2907	sysctl_createv(clog, 0, NULL, NULL,
2908		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2909		       CTLTYPE_INT, "errratelimit", NULL,
2910		       NULL, 0, &icmp6_errratelimit, 0,
2911		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2912		       ICMPV6CTL_ERRRATELIMIT, CTL_EOL);
2913#endif
2914	sysctl_createv(clog, 0, NULL, NULL,
2915		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2916		       CTLTYPE_INT, "nd6_prune",
2917		       SYSCTL_DESCR("Neighbor discovery prune interval"),
2918		       NULL, 0, &nd6_prune, 0,
2919		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2920		       ICMPV6CTL_ND6_PRUNE, CTL_EOL);
2921	sysctl_createv(clog, 0, NULL, NULL,
2922		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2923		       CTLTYPE_INT, "nd6_delay",
2924		       SYSCTL_DESCR("First probe delay time"),
2925		       NULL, 0, &nd6_delay, 0,
2926		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2927		       ICMPV6CTL_ND6_DELAY, CTL_EOL);
2928	sysctl_createv(clog, 0, NULL, NULL,
2929		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2930		       CTLTYPE_INT, "nd6_umaxtries",
2931		       SYSCTL_DESCR("Number of unicast discovery attempts"),
2932		       NULL, 0, &nd6_umaxtries, 0,
2933		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2934		       ICMPV6CTL_ND6_UMAXTRIES, CTL_EOL);
2935	sysctl_createv(clog, 0, NULL, NULL,
2936		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2937		       CTLTYPE_INT, "nd6_mmaxtries",
2938		       SYSCTL_DESCR("Number of multicast discovery attempts"),
2939		       NULL, 0, &nd6_mmaxtries, 0,
2940		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2941		       ICMPV6CTL_ND6_MMAXTRIES, CTL_EOL);
2942	sysctl_createv(clog, 0, NULL, NULL,
2943		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2944		       CTLTYPE_INT, "nd6_useloopback",
2945		       SYSCTL_DESCR("Use loopback interface for local traffic"),
2946		       NULL, 0, &nd6_useloopback, 0,
2947		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2948		       ICMPV6CTL_ND6_USELOOPBACK, CTL_EOL);
2949#if 0 /* obsoleted */
2950	sysctl_createv(clog, 0, NULL, NULL,
2951		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2952		       CTLTYPE_INT, "nd6_proxyall", NULL,
2953		       NULL, 0, &nd6_proxyall, 0,
2954		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2955		       ICMPV6CTL_ND6_PROXYALL, CTL_EOL);
2956#endif
2957	sysctl_createv(clog, 0, NULL, NULL,
2958		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2959		       CTLTYPE_INT, "nodeinfo",
2960		       SYSCTL_DESCR("Respond to node information requests"),
2961		       NULL, 0, &icmp6_nodeinfo, 0,
2962		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2963		       ICMPV6CTL_NODEINFO, CTL_EOL);
2964	sysctl_createv(clog, 0, NULL, NULL,
2965		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2966		       CTLTYPE_INT, "errppslimit",
2967		       SYSCTL_DESCR("Maximum ICMP errors sent per second"),
2968		       NULL, 0, &icmp6errppslim, 0,
2969		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2970		       ICMPV6CTL_ERRPPSLIMIT, CTL_EOL);
2971	sysctl_createv(clog, 0, NULL, NULL,
2972		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2973		       CTLTYPE_INT, "nd6_maxnudhint",
2974		       SYSCTL_DESCR("Maximum neighbor unreachable hint count"),
2975		       NULL, 0, &nd6_maxnudhint, 0,
2976		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2977		       ICMPV6CTL_ND6_MAXNUDHINT, CTL_EOL);
2978	sysctl_createv(clog, 0, NULL, NULL,
2979		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2980		       CTLTYPE_INT, "mtudisc_hiwat",
2981		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
2982		       NULL, 0, &icmp6_mtudisc_hiwat, 0,
2983		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2984		       ICMPV6CTL_MTUDISC_HIWAT, CTL_EOL);
2985	sysctl_createv(clog, 0, NULL, NULL,
2986		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2987		       CTLTYPE_INT, "mtudisc_lowat",
2988		       SYSCTL_DESCR("Low mark on MTU Discovery route timers"),
2989		       NULL, 0, &icmp6_mtudisc_lowat, 0,
2990		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2991		       ICMPV6CTL_MTUDISC_LOWAT, CTL_EOL);
2992	sysctl_createv(clog, 0, NULL, NULL,
2993		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
2994		       CTLTYPE_INT, "nd6_debug",
2995		       SYSCTL_DESCR("Enable neighbor discovery debug output"),
2996		       NULL, 0, &nd6_debug, 0,
2997		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
2998		       ICMPV6CTL_ND6_DEBUG, CTL_EOL);
2999	sysctl_createv(clog, 0, NULL, NULL,
3000		       CTLFLAG_PERMANENT,
3001		       CTLTYPE_STRUCT, "nd6_drlist",
3002		       SYSCTL_DESCR("Default router list"),
3003		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
3004		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3005		       ICMPV6CTL_ND6_DRLIST, CTL_EOL);
3006	sysctl_createv(clog, 0, NULL, NULL,
3007		       CTLFLAG_PERMANENT,
3008		       CTLTYPE_STRUCT, "nd6_prlist",
3009		       SYSCTL_DESCR("Prefix list"),
3010		       sysctl_net_inet6_icmp6_nd6, 0, NULL, 0,
3011		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3012		       ICMPV6CTL_ND6_PRLIST, CTL_EOL);
3013	sysctl_createv(clog, 0, NULL, NULL,
3014		       CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
3015		       CTLTYPE_INT, "maxqueuelen",
3016		       SYSCTL_DESCR("max packet queue len for a unresolved ND"),
3017		       NULL, 1, &nd6_maxqueuelen, 0,
3018		       CTL_NET, PF_INET6, IPPROTO_ICMPV6,
3019		       ICMPV6CTL_ND6_MAXQLEN, CTL_EOL);
3020}
3021
3022void
3023icmp6_statinc(u_int stat)
3024{
3025
3026	KASSERT(stat < ICMP6_NSTATS);
3027	ICMP6_STATINC(stat);
3028}
3029