icmp6.c revision 1.35
1/*	$NetBSD: icmp6.c,v 1.35 2000/07/06 12:36:18 itojun Exp $	*/
2/*	$KAME: icmp6.c,v 1.120 2000/07/06 11:47:20 itojun 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. All advertising materials mentioning features or use of this software
46 *    must display the following acknowledgement:
47 *	This product includes software developed by the University of
48 *	California, Berkeley and its contributors.
49 * 4. Neither the name of the University nor the names of its contributors
50 *    may be used to endorse or promote products derived from this software
51 *    without specific prior written permission.
52 *
53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63 * SUCH DAMAGE.
64 *
65 *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
66 */
67
68#include "opt_inet.h"
69#include "opt_ipsec.h"
70
71#include <sys/param.h>
72#include <sys/systm.h>
73#include <sys/malloc.h>
74#include <sys/mbuf.h>
75#include <sys/protosw.h>
76#include <sys/socket.h>
77#include <sys/socketvar.h>
78#include <sys/time.h>
79#include <sys/kernel.h>
80#include <sys/syslog.h>
81#include <sys/domain.h>
82
83#include <net/if.h>
84#include <net/route.h>
85#include <net/if_dl.h>
86#include <net/if_types.h>
87
88#include <netinet/in.h>
89#include <netinet/in_var.h>
90#include <netinet/ip6.h>
91#include <netinet6/ip6_var.h>
92#include <netinet/icmp6.h>
93#include <netinet6/mld6_var.h>
94#include <netinet6/in6_pcb.h>
95#include <netinet6/nd6.h>
96#include <netinet6/in6_ifattach.h>
97#include <netinet6/ip6protosw.h>
98
99
100#ifdef IPSEC
101#include <netinet6/ipsec.h>
102#include <netkey/key.h>
103#endif
104
105#include "faith.h"
106
107#include <net/net_osdep.h>
108
109extern struct domain inet6domain;
110
111struct icmp6stat icmp6stat;
112
113extern struct in6pcb rawin6pcb;
114extern struct timeval icmp6errratelim;
115static struct timeval icmp6errratelim_last;
116extern int icmp6errppslim;
117static int icmp6errpps_count = 0;
118extern int icmp6_nodeinfo;
119static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL;
120extern int pmtu_expire;
121
122static void icmp6_errcount __P((struct icmp6errstat *, int, int));
123static int icmp6_rip6_input __P((struct mbuf **, int));
124static void icmp6_mtudisc_update __P((struct in6_addr *, struct icmp6_hdr *,
125				      struct mbuf *));
126static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int));
127static const char *icmp6_redirect_diag __P((struct in6_addr *,
128	struct in6_addr *, struct in6_addr *));
129static struct mbuf *ni6_input __P((struct mbuf *, int));
130static struct mbuf *ni6_nametodns __P((const char *, int, int));
131static int ni6_dnsmatch __P((const char *, int, const char *, int));
132static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
133			  struct ifnet **));
134static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
135				struct ifnet *, int));
136static struct rtentry *icmp6_mtudisc_clone __P((struct sockaddr *));
137static void icmp6_mtudisc_timeout __P((struct rtentry *, struct rttimer *));
138
139#ifdef COMPAT_RFC1885
140static struct route_in6 icmp6_reflect_rt;
141#endif
142
143void
144icmp6_init()
145{
146	mld6_init();
147	icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire);
148}
149
150static void
151icmp6_errcount(stat, type, code)
152	struct icmp6errstat *stat;
153	int type, code;
154{
155	switch(type) {
156	case ICMP6_DST_UNREACH:
157		switch (code) {
158		case ICMP6_DST_UNREACH_NOROUTE:
159			stat->icp6errs_dst_unreach_noroute++;
160			return;
161		case ICMP6_DST_UNREACH_ADMIN:
162			stat->icp6errs_dst_unreach_admin++;
163			return;
164		case ICMP6_DST_UNREACH_BEYONDSCOPE:
165			stat->icp6errs_dst_unreach_beyondscope++;
166			return;
167		case ICMP6_DST_UNREACH_ADDR:
168			stat->icp6errs_dst_unreach_addr++;
169			return;
170		case ICMP6_DST_UNREACH_NOPORT:
171			stat->icp6errs_dst_unreach_noport++;
172			return;
173		}
174		break;
175	case ICMP6_PACKET_TOO_BIG:
176		stat->icp6errs_packet_too_big++;
177		return;
178	case ICMP6_TIME_EXCEEDED:
179		switch(code) {
180		case ICMP6_TIME_EXCEED_TRANSIT:
181			stat->icp6errs_time_exceed_transit++;
182			return;
183		case ICMP6_TIME_EXCEED_REASSEMBLY:
184			stat->icp6errs_time_exceed_reassembly++;
185			return;
186		}
187		break;
188	case ICMP6_PARAM_PROB:
189		switch(code) {
190		case ICMP6_PARAMPROB_HEADER:
191			stat->icp6errs_paramprob_header++;
192			return;
193		case ICMP6_PARAMPROB_NEXTHEADER:
194			stat->icp6errs_paramprob_nextheader++;
195			return;
196		case ICMP6_PARAMPROB_OPTION:
197			stat->icp6errs_paramprob_option++;
198			return;
199		}
200		break;
201	case ND_REDIRECT:
202		stat->icp6errs_redirect++;
203		return;
204	}
205	stat->icp6errs_unknown++;
206}
207
208/*
209 * Generate an error packet of type error in response to bad IP6 packet.
210 */
211void
212icmp6_error(m, type, code, param)
213	struct mbuf *m;
214	int type, code, param;
215{
216	struct ip6_hdr *oip6, *nip6;
217	struct icmp6_hdr *icmp6;
218	u_int preplen;
219	int off;
220	int nxt;
221
222	icmp6stat.icp6s_error++;
223
224	/* count per-type-code statistics */
225	icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
226
227	if (m->m_flags & M_DECRYPTED) {
228		icmp6stat.icp6s_canterror++;
229		goto freeit;
230	}
231
232#ifndef PULLDOWN_TEST
233	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
234#else
235	if (m->m_len < sizeof(struct ip6_hdr)) {
236		m = m_pullup(m, sizeof(struct ip6_hdr));
237		if (m == NULL)
238			return;
239	}
240#endif
241	oip6 = mtod(m, struct ip6_hdr *);
242
243	/*
244	 * Multicast destination check. For unrecognized option errors,
245	 * this check has already done in ip6_unknown_opt(), so we can
246	 * check only for other errors.
247	 */
248	if ((m->m_flags & (M_BCAST|M_MCAST) ||
249	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
250	    (type != ICMP6_PACKET_TOO_BIG &&
251	     (type != ICMP6_PARAM_PROB ||
252	      code != ICMP6_PARAMPROB_OPTION)))
253		goto freeit;
254
255	/* Source address check. XXX: the case of anycast source? */
256	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
257	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
258		goto freeit;
259
260	/*
261	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
262	 * don't do it.
263	 */
264	nxt = -1;
265	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
266	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
267		struct icmp6_hdr *icp;
268
269#ifndef PULLDOWN_TEST
270		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
271		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
272#else
273		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
274			sizeof(*icp));
275		if (icp == NULL) {
276			icmp6stat.icp6s_tooshort++;
277			return;
278		}
279#endif
280		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
281		    icp->icmp6_type == ND_REDIRECT) {
282			/*
283			 * ICMPv6 error
284			 * Special case: for redirect (which is
285			 * informational) we must not send icmp6 error.
286			 */
287			icmp6stat.icp6s_canterror++;
288			goto freeit;
289		} else {
290			/* ICMPv6 informational - send the error */
291		}
292	} else {
293		/* non-ICMPv6 - send the error */
294	}
295
296	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
297
298	/* Finally, do rate limitation check. */
299	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
300		icmp6stat.icp6s_toofreq++;
301		goto freeit;
302	}
303
304	/*
305	 * OK, ICMP6 can be generated.
306	 */
307
308	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
309		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
310
311	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
312	M_PREPEND(m, preplen, M_DONTWAIT);
313	if (m && m->m_len < preplen)
314		m = m_pullup(m, preplen);
315	if (m == NULL) {
316		printf("ENOBUFS in icmp6_error %d\n", __LINE__);
317		return;
318	}
319
320	nip6 = mtod(m, struct ip6_hdr *);
321	nip6->ip6_src  = oip6->ip6_src;
322	nip6->ip6_dst  = oip6->ip6_dst;
323
324	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src))
325		oip6->ip6_src.s6_addr16[1] = 0;
326	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst))
327		oip6->ip6_dst.s6_addr16[1] = 0;
328
329	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
330	icmp6->icmp6_type = type;
331	icmp6->icmp6_code = code;
332	icmp6->icmp6_pptr = htonl((u_int32_t)param);
333
334	icmp6stat.icp6s_outhist[type]++;
335	icmp6_reflect(m, sizeof(struct ip6_hdr)); /*header order: IPv6 - ICMPv6*/
336
337	return;
338
339  freeit:
340	/*
341	 * If we can't tell wheter or not we can generate ICMP6, free it.
342	 */
343	m_freem(m);
344}
345
346/*
347 * Process a received ICMP6 message.
348 */
349int
350icmp6_input(mp, offp, proto)
351	struct mbuf **mp;
352	int *offp, proto;
353{
354	struct mbuf *m = *mp, *n;
355	struct ip6_hdr *ip6, *nip6;
356	struct icmp6_hdr *icmp6, *nicmp6;
357	int off = *offp;
358	int icmp6len = m->m_pkthdr.len - *offp;
359	int code, sum, noff;
360	struct sockaddr_in6 icmp6src;
361
362#ifndef PULLDOWN_TEST
363	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
364	/* m might change if M_LOOP. So, call mtod after this */
365#endif
366
367	/*
368	 * Locate icmp6 structure in mbuf, and check
369	 * that not corrupted and of at least minimum length
370	 */
371
372	ip6 = mtod(m, struct ip6_hdr *);
373	if (icmp6len < sizeof(struct icmp6_hdr)) {
374		icmp6stat.icp6s_tooshort++;
375		goto freeit;
376	}
377
378	/*
379	 * calculate the checksum
380	 */
381#ifndef PULLDOWN_TEST
382	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
383#else
384	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
385	if (icmp6 == NULL) {
386		icmp6stat.icp6s_tooshort++;
387		return IPPROTO_DONE;
388	}
389#endif
390	code = icmp6->icmp6_code;
391
392	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
393		log(LOG_ERR,
394		    "ICMP6 checksum error(%d|%x) %s\n",
395		    icmp6->icmp6_type,
396		    sum,
397		    ip6_sprintf(&ip6->ip6_src));
398		icmp6stat.icp6s_checksum++;
399		goto freeit;
400	}
401
402#if defined(NFAITH) && 0 < NFAITH
403	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
404		/*
405		 * Deliver very specific ICMP6 type only.
406		 * This is important to deilver TOOBIG.  Otherwise PMTUD
407		 * will not work.
408		 */
409		switch (icmp6->icmp6_type) {
410		case ICMP6_DST_UNREACH:
411		case ICMP6_PACKET_TOO_BIG:
412		case ICMP6_TIME_EXCEEDED:
413			break;
414		default:
415			goto freeit;
416		}
417	}
418#endif
419
420#ifdef IPSEC
421	/* drop it if it does not match the default policy */
422	if (ipsec6_in_reject(m, NULL)) {
423		ipsecstat.in_polvio++;
424		goto freeit;
425	}
426#endif
427
428	icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
429	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
430	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
431		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
432
433	switch (icmp6->icmp6_type) {
434
435	case ICMP6_DST_UNREACH:
436		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
437		switch (code) {
438		case ICMP6_DST_UNREACH_NOROUTE:
439			code = PRC_UNREACH_NET;
440			break;
441		case ICMP6_DST_UNREACH_ADMIN:
442			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
443			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
444			break;
445		case ICMP6_DST_UNREACH_ADDR:
446			code = PRC_HOSTDEAD;
447			break;
448#ifdef COMPAT_RFC1885
449		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
450			code = PRC_UNREACH_SRCFAIL;
451			break;
452#else
453		case ICMP6_DST_UNREACH_BEYONDSCOPE:
454			/* I mean "source address was incorrect." */
455			code = PRC_PARAMPROB;
456			break;
457#endif
458		case ICMP6_DST_UNREACH_NOPORT:
459			code = PRC_UNREACH_PORT;
460			break;
461		default:
462			goto badcode;
463		}
464		goto deliver;
465		break;
466
467	case ICMP6_PACKET_TOO_BIG:
468		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
469		if (code != 0)
470			goto badcode;
471
472		code = PRC_MSGSIZE;
473
474		/*
475		 * Updating the path MTU will be done after examining
476		 * intermediate extension headers.
477		 */
478		goto deliver;
479		break;
480
481	case ICMP6_TIME_EXCEEDED:
482		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
483		switch (code) {
484		case ICMP6_TIME_EXCEED_TRANSIT:
485		case ICMP6_TIME_EXCEED_REASSEMBLY:
486			code += PRC_TIMXCEED_INTRANS;
487			break;
488		default:
489			goto badcode;
490		}
491		goto deliver;
492		break;
493
494	case ICMP6_PARAM_PROB:
495		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
496		switch (code) {
497		case ICMP6_PARAMPROB_NEXTHEADER:
498			code = PRC_UNREACH_PROTOCOL;
499			break;
500		case ICMP6_PARAMPROB_HEADER:
501		case ICMP6_PARAMPROB_OPTION:
502			code = PRC_PARAMPROB;
503			break;
504		default:
505			goto badcode;
506		}
507		goto deliver;
508		break;
509
510	case ICMP6_ECHO_REQUEST:
511		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
512		if (code != 0)
513			goto badcode;
514		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
515			/* Give up remote */
516			break;
517		}
518		if ((n->m_flags & M_EXT) != 0
519		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
520			struct mbuf *n0 = n;
521			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
522
523			/*
524			 * Prepare an internal mbuf. m_pullup() doesn't
525			 * always copy the length we specified.
526			 */
527			if (maxlen >= MCLBYTES) {
528#ifdef DIAGNOSTIC
529				printf("MCLBYTES too small\n");
530#endif
531				/* Give up remote */
532				m_freem(n0);
533				break;
534			}
535			MGETHDR(n, M_DONTWAIT, n0->m_type);
536			if (n && maxlen >= MHLEN) {
537				MCLGET(n, M_DONTWAIT);
538				if ((n->m_flags & M_EXT) == 0) {
539					m_free(n);
540					n = NULL;
541				}
542			}
543			if (n == NULL) {
544				/* Give up remote */
545				m_freem(n0);
546				break;
547			}
548			M_COPY_PKTHDR(n, n0);
549			/*
550			 * Copy IPv6 and ICMPv6 only.
551			 */
552			nip6 = mtod(n, struct ip6_hdr *);
553			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
554			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
555			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
556			noff = sizeof(struct ip6_hdr);
557			n->m_pkthdr.len = n->m_len =
558				noff + sizeof(struct icmp6_hdr);
559			/*
560			 * Adjust mbuf. ip6_plen will be adjusted in
561			 * ip6_output().
562			 */
563			m_adj(n0, off + sizeof(struct icmp6_hdr));
564			n->m_pkthdr.len += n0->m_pkthdr.len;
565			n->m_next = n0;
566			n0->m_flags &= ~M_PKTHDR;
567		} else {
568			nip6 = mtod(n, struct ip6_hdr *);
569			nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
570			noff = off;
571		}
572		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
573		nicmp6->icmp6_code = 0;
574		if (n) {
575			icmp6stat.icp6s_reflect++;
576			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
577			icmp6_reflect(n, noff);
578		}
579		break;
580
581	case ICMP6_ECHO_REPLY:
582		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
583		if (code != 0)
584			goto badcode;
585		break;
586
587	case MLD6_LISTENER_QUERY:
588	case MLD6_LISTENER_REPORT:
589		if (icmp6len < sizeof(struct mld6_hdr))
590			goto badlen;
591		if (icmp6->icmp6_type == MLD6_LISTENER_QUERY) /* XXX: ugly... */
592			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
593		else
594			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
595		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
596			/* give up local */
597			mld6_input(m, off);
598			m = NULL;
599			goto freeit;
600		}
601		mld6_input(n, off);
602		/* m stays. */
603		break;
604
605	case MLD6_LISTENER_DONE:
606		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
607		if (icmp6len < sizeof(struct mld6_hdr))	/* necessary? */
608			goto badlen;
609		break;		/* nothing to be done in kernel */
610
611	case MLD6_MTRACE_RESP:
612	case MLD6_MTRACE:
613		/* XXX: these two are experimental. not officially defind. */
614		/* XXX: per-interface statistics? */
615		break;		/* just pass it to applications */
616
617	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
618	    {
619		enum { WRU, FQDN } mode;
620
621		if (!icmp6_nodeinfo)
622			break;
623
624		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
625			mode = WRU;
626		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
627			mode = FQDN;
628		else
629			goto badlen;
630
631		if (mode == FQDN) {
632#ifndef PULLDOWN_TEST
633			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
634					 IPPROTO_DONE);
635#endif
636			n = m_copy(m, 0, M_COPYALL);
637			if (n)
638				n = ni6_input(n, off);
639			/* XXX meaningless if n == NULL */
640			noff = sizeof(struct ip6_hdr);
641		} else {
642			u_char *p;
643			int maxlen, maxhlen;
644
645			if (code != 0)
646				goto badcode;
647			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
648			if (maxlen >= MCLBYTES) {
649#ifdef DIAGNOSTIC
650				printf("MCLBYTES too small\n");
651#endif
652				/* Give up remote */
653				break;
654			}
655			MGETHDR(n, M_DONTWAIT, m->m_type);
656			if (n && maxlen > MHLEN) {
657				MCLGET(n, M_DONTWAIT);
658				if ((n->m_flags & M_EXT) == 0) {
659					m_free(n);
660					n = NULL;
661				}
662			}
663			if (n == NULL) {
664				/* Give up remote */
665				break;
666			}
667			n->m_len = 0;
668			maxhlen = M_TRAILINGSPACE(n) - maxlen;
669			if (maxhlen > hostnamelen)
670				maxhlen = hostnamelen;
671			/*
672			 * Copy IPv6 and ICMPv6 only.
673			 */
674			nip6 = mtod(n, struct ip6_hdr *);
675			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
676			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
677			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
678			p = (u_char *)(nicmp6 + 1);
679			bzero(p, 4);
680			bcopy(hostname, p + 4, maxhlen); /*meaningless TTL*/
681			noff = sizeof(struct ip6_hdr);
682			M_COPY_PKTHDR(n, m); /* just for recvif */
683			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
684				sizeof(struct icmp6_hdr) + 4 + maxhlen;
685			nicmp6->icmp6_type = ICMP6_WRUREPLY;
686			nicmp6->icmp6_code = 0;
687		}
688#undef hostnamelen
689		if (n) {
690			icmp6stat.icp6s_reflect++;
691			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
692			icmp6_reflect(n, noff);
693		}
694		break;
695	    }
696
697	case ICMP6_WRUREPLY:
698		if (code != 0)
699			goto badcode;
700		break;
701
702	case ND_ROUTER_SOLICIT:
703		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
704		if (code != 0)
705			goto badcode;
706		if (icmp6len < sizeof(struct nd_router_solicit))
707			goto badlen;
708		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
709			/* give up local */
710			nd6_rs_input(m, off, icmp6len);
711			m = NULL;
712			goto freeit;
713		}
714		nd6_rs_input(n, off, icmp6len);
715		/* m stays. */
716		break;
717
718	case ND_ROUTER_ADVERT:
719		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
720		if (code != 0)
721			goto badcode;
722		if (icmp6len < sizeof(struct nd_router_advert))
723			goto badlen;
724		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
725			/* give up local */
726			nd6_ra_input(m, off, icmp6len);
727			m = NULL;
728			goto freeit;
729		}
730		nd6_ra_input(n, off, icmp6len);
731		/* m stays. */
732		break;
733
734	case ND_NEIGHBOR_SOLICIT:
735		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
736		if (code != 0)
737			goto badcode;
738		if (icmp6len < sizeof(struct nd_neighbor_solicit))
739			goto badlen;
740		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
741			/* give up local */
742			nd6_ns_input(m, off, icmp6len);
743			m = NULL;
744			goto freeit;
745		}
746		nd6_ns_input(n, off, icmp6len);
747		/* m stays. */
748		break;
749
750	case ND_NEIGHBOR_ADVERT:
751		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
752		if (code != 0)
753			goto badcode;
754		if (icmp6len < sizeof(struct nd_neighbor_advert))
755			goto badlen;
756		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
757			/* give up local */
758			nd6_na_input(m, off, icmp6len);
759			m = NULL;
760			goto freeit;
761		}
762		nd6_na_input(n, off, icmp6len);
763		/* m stays. */
764		break;
765
766	case ND_REDIRECT:
767		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
768		if (code != 0)
769			goto badcode;
770		if (icmp6len < sizeof(struct nd_redirect))
771			goto badlen;
772		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
773			/* give up local */
774			icmp6_redirect_input(m, off);
775			m = NULL;
776			goto freeit;
777		}
778		icmp6_redirect_input(n, off);
779		/* m stays. */
780		break;
781
782	case ICMP6_ROUTER_RENUMBERING:
783		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
784		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
785			goto badcode;
786		if (icmp6len < sizeof(struct icmp6_router_renum))
787			goto badlen;
788		break;
789
790	default:
791		printf("icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
792		       icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
793		       ip6_sprintf(&ip6->ip6_dst),
794		       m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0);
795		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
796			/* ICMPv6 error: MUST deliver it by spec... */
797			code = PRC_NCMDS;
798			/* deliver */
799		} else {
800			/* ICMPv6 informational: MUST not deliver */
801			break;
802		}
803	deliver:
804		if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
805			icmp6stat.icp6s_tooshort++;
806			goto freeit;
807		}
808#ifndef PULLDOWN_TEST
809		IP6_EXTHDR_CHECK(m, off,
810			sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr),
811			IPPROTO_DONE);
812		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
813#else
814		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
815			sizeof(*icmp6) + sizeof(struct ip6_hdr));
816		if (icmp6 == NULL) {
817			icmp6stat.icp6s_tooshort++;
818			return IPPROTO_DONE;
819		}
820#endif
821		bzero(&icmp6src, sizeof(icmp6src));
822		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
823		icmp6src.sin6_family = AF_INET6;
824		icmp6src.sin6_addr = ((struct ip6_hdr *)(icmp6 + 1))->ip6_dst;
825
826		/* Detect the upper level protocol */
827	    {
828		void (*ctlfunc) __P((int, struct sockaddr *, void *));
829		struct ip6_hdr *eip6 = (struct ip6_hdr *)(icmp6 + 1);
830		u_int8_t nxt = eip6->ip6_nxt;
831		int eoff = off + sizeof(struct icmp6_hdr) +
832			sizeof(struct ip6_hdr);
833		struct ip6ctlparam ip6cp;
834		struct in6_addr *finaldst = NULL;
835		int icmp6type = icmp6->icmp6_type;
836		struct ip6_frag *fh;
837		struct ip6_rthdr *rth;
838		struct ip6_rthdr0 *rth0;
839		int rthlen;
840
841		while (1) { /* XXX: should avoid inf. loop explicitly? */
842			struct ip6_ext *eh;
843
844			switch(nxt) {
845			case IPPROTO_HOPOPTS:
846			case IPPROTO_DSTOPTS:
847			case IPPROTO_AH:
848#ifndef PULLDOWN_TEST
849				IP6_EXTHDR_CHECK(m, 0, eoff +
850						 sizeof(struct ip6_ext),
851						 IPPROTO_DONE);
852				eh = (struct ip6_ext *)(mtod(m, caddr_t)
853							+ eoff);
854#else
855				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
856					eoff, sizeof(*eh));
857				if (eh == NULL) {
858					icmp6stat.icp6s_tooshort++;
859					return IPPROTO_DONE;
860				}
861#endif
862
863				if (nxt == IPPROTO_AH)
864					eoff += (eh->ip6e_len + 2) << 2;
865				else
866					eoff += (eh->ip6e_len + 1) << 3;
867				nxt = eh->ip6e_nxt;
868				break;
869			case IPPROTO_ROUTING:
870				/*
871				 * When the erroneous packet contains a
872				 * routing header, we should examine the
873				 * header to determine the final destination.
874				 * Otherwise, we can't properly update
875				 * information that depends on the final
876				 * destination (e.g. path MTU).
877				 */
878#ifndef PULLDOWN_TEST
879				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth),
880						 IPPROTO_DONE);
881				rth = (struct ip6_rthdr *)(mtod(m, caddr_t)
882							   + eoff);
883#else
884				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
885					eoff, sizeof(*rth));
886				if (rth == NULL) {
887					icmp6stat.icp6s_tooshort++;
888					return IPPROTO_DONE;
889				}
890#endif
891				rthlen = (rth->ip6r_len + 1) << 3;
892				/*
893				 * XXX: currently there is no
894				 * officially defined type other
895				 * than type-0.
896				 * Note that if the segment left field
897				 * is 0, all intermediate hops must
898				 * have been passed.
899				 */
900				if (rth->ip6r_segleft &&
901				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
902					int hops;
903
904#ifndef PULLDOWN_TEST
905					IP6_EXTHDR_CHECK(m, 0, eoff + rthlen,
906							 IPPROTO_DONE);
907					rth0 = (struct ip6_rthdr0 *)(mtod(m, caddr_t) + eoff);
908#else
909					IP6_EXTHDR_GET(rth0,
910						       struct ip6_rthdr0 *, m,
911						       eoff, rthlen);
912					if (rth0 == NULL) {
913						icmp6stat.icp6s_tooshort++;
914						return IPPROTO_DONE;
915					}
916#endif
917					/* just ignore a bogus header */
918					if ((rth0->ip6r0_len % 2) == 0 &&
919					    (hops = rth0->ip6r0_len/2))
920						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
921				}
922				eoff += rthlen;
923				nxt = rth->ip6r_nxt;
924				break;
925			case IPPROTO_FRAGMENT:
926#ifndef PULLDOWN_TEST
927				IP6_EXTHDR_CHECK(m, 0, eoff +
928						 sizeof(struct ip6_frag),
929						 IPPROTO_DONE);
930				fh = (struct ip6_frag *)(mtod(m, caddr_t)
931							 + eoff);
932#else
933				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
934					eoff, sizeof(*fh));
935				if (fh == NULL) {
936					icmp6stat.icp6s_tooshort++;
937					return IPPROTO_DONE;
938				}
939#endif
940				/*
941				 * Data after a fragment header is meaningless
942				 * unless it is the first fragment, but
943				 * we'll go to the notify label for path MTU
944				 * discovery.
945				 */
946				if (fh->ip6f_offlg & IP6F_OFF_MASK)
947					goto notify;
948
949				eoff += sizeof(struct ip6_frag);
950				nxt = fh->ip6f_nxt;
951				break;
952			default:
953				/*
954				 * This case includes ESP and the No Next
955				 * Header. In such cases going to the notify
956				 * label does not have any meaning
957				 * (i.e. ctlfunc will be NULL), but we go
958				 * anyway since we might have to update
959				 * path MTU information.
960				 */
961				goto notify;
962			}
963		}
964	    notify:
965#ifndef PULLDOWN_TEST
966		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
967#else
968		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
969			sizeof(*icmp6) + sizeof(struct ip6_hdr));
970		if (icmp6 == NULL) {
971			icmp6stat.icp6s_tooshort++;
972			return IPPROTO_DONE;
973		}
974#endif
975		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
976			if (finaldst == NULL)
977				finaldst = &((struct ip6_hdr *)(icmp6 + 1))->ip6_dst;
978			icmp6_mtudisc_update(finaldst, icmp6, m);
979		}
980
981		ctlfunc = (void (*) __P((int, struct sockaddr *, void *)))
982			(inet6sw[ip6_protox[nxt]].pr_ctlinput);
983		if (ctlfunc) {
984			ip6cp.ip6c_m = m;
985			ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
986			ip6cp.ip6c_off = eoff;
987			(*ctlfunc)(code, (struct sockaddr *)&icmp6src, &ip6cp);
988		}
989	    }
990		break;
991
992	badcode:
993		icmp6stat.icp6s_badcode++;
994		break;
995
996	badlen:
997		icmp6stat.icp6s_badlen++;
998		break;
999	}
1000
1001	icmp6_rip6_input(&m, *offp);
1002	return IPPROTO_DONE;
1003
1004 freeit:
1005	m_freem(m);
1006	return IPPROTO_DONE;
1007}
1008
1009static void
1010icmp6_mtudisc_update(dst, icmp6, m)
1011	struct in6_addr *dst;
1012	struct icmp6_hdr *icmp6;/* we can assume the validity of the pointer */
1013	struct mbuf *m;	/* currently unused but added for scoped addrs */
1014{
1015	u_int mtu = ntohl(icmp6->icmp6_mtu);
1016	struct rtentry *rt = NULL;
1017	struct sockaddr_in6 sin6;
1018
1019	bzero(&sin6, sizeof(sin6));
1020	sin6.sin6_family = PF_INET6;
1021	sin6.sin6_len = sizeof(struct sockaddr_in6);
1022	sin6.sin6_addr = *dst;
1023	/* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1024	rt = rtalloc1((struct sockaddr *)&sin6, 1);	/*clone*/
1025	if (!rt || (rt->rt_flags & RTF_HOST) == 0) {
1026		if (rt)
1027			RTFREE(rt);
1028		rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6);
1029	}
1030
1031	if (rt && (rt->rt_flags & RTF_HOST)
1032	    && !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1033		if (mtu < IPV6_MMTU) {
1034				/* xxx */
1035			rt->rt_rmx.rmx_locks |= RTV_MTU;
1036		} else if (mtu < rt->rt_ifp->if_mtu &&
1037			   rt->rt_rmx.rmx_mtu > mtu) {
1038			rt->rt_rmx.rmx_mtu = mtu;
1039		}
1040	}
1041	if (rt)
1042		RTFREE(rt);
1043}
1044
1045/*
1046 * Process a Node Information Query packet, (roughly) based on
1047 * draft-ietf-ipngwg-icmp-name-lookups-05.
1048 *
1049 * Spec incompatibilities:
1050 * - IPv6 Subject address handling
1051 * - IPv4 Subject address handling support missing
1052 * - Proxy reply (answer even if it's not for me)
1053 * - "Supported Qtypes" support missing
1054 * - joins NI group address at in6_ifattach() time only, does not cope
1055 *   with hostname changes by sethostname(3)
1056 */
1057#ifndef offsetof		/* XXX */
1058#define	offsetof(type, member)	((size_t)(&((type *)0)->member))
1059#endif
1060static struct mbuf *
1061ni6_input(m, off)
1062	struct mbuf *m;
1063	int off;
1064{
1065	struct icmp6_nodeinfo *ni6, *nni6;
1066	struct mbuf *n = NULL;
1067	u_int16_t qtype;
1068	int subjlen;
1069	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1070	struct ni_reply_fqdn *fqdn;
1071	int addrs;		/* for NI_QTYPE_NODEADDR */
1072	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1073	struct sockaddr_in6 sin6;
1074	struct ip6_hdr *ip6;
1075	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1076	char *subj;
1077
1078	ip6 = mtod(m, struct ip6_hdr *);
1079#ifndef PULLDOWN_TEST
1080	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1081#else
1082	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1083	if (ni6 == NULL) {
1084		/* m is already reclaimed */
1085		return NULL;
1086	}
1087#endif
1088
1089	/*
1090	 * Validate IPv6 destination address.
1091	 *
1092	 * We accept packets with the following IPv6 destination address:
1093	 * - Responder's unicast/anycast address,
1094	 * - link-local multicast address
1095	 * This is a violation to last paragraph in icmp-name-lookups-05
1096	 * page 4, which restricts IPv6 destination address of a query to:
1097	 * - Responder's unicast/anycast address,
1098	 * - NI group address for a name belongs to the Responder, or
1099	 * - NI group address for a name for which the Responder is providing
1100	 *   proxy service.
1101	 * (note: NI group address is a link-local multicast address)
1102	 *
1103	 * We allow any link-local multicast address, since "ping6 -w ff02::1"
1104	 * has been really useful for us debugging our network.  Also this is
1105	 * still questionable if the restriction in spec buy us security at all,
1106	 * since RFC2463 permits echo packet to multicast destination.
1107	 * Even if we forbid NI query to ff02::1, we can effectively get the
1108	 * same result as "ping6 -w ff02::1" by the following steps:
1109	 * - run "ping6 ff02::1", then
1110	 * - run "ping6 -w" for all addresses replied.
1111	 */
1112	bzero(&sin6, sizeof(sin6));
1113	sin6.sin6_family = AF_INET6;
1114	sin6.sin6_len = sizeof(struct sockaddr_in6);
1115	bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1116	/* XXX scopeid */
1117	if (ifa_ifwithaddr((struct sockaddr *)&sin6))
1118		; /*unicast/anycast, fine*/
1119	else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1120		; /*violates spec slightly, see above*/
1121	else
1122		goto bad;
1123
1124	/* guess reply length */
1125	qtype = ntohs(ni6->ni_qtype);
1126	switch (qtype) {
1127	case NI_QTYPE_NOOP:
1128		break;		/* no reply data */
1129	case NI_QTYPE_SUPTYPES:
1130		goto bad;	/* xxx: to be implemented */
1131		break;
1132	case NI_QTYPE_FQDN:
1133		/* XXX will append a mbuf */
1134		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1135		break;
1136	case NI_QTYPE_NODEADDR:
1137		addrs = ni6_addrs(ni6, m, &ifp);
1138		if ((replylen += addrs * sizeof(struct in6_addr)) > MCLBYTES)
1139			replylen = MCLBYTES; /* XXX: we'll truncate later */
1140		break;
1141	default:
1142		/*
1143		 * XXX: We must return a reply with the ICMP6 code
1144		 * `unknown Qtype' in this case. However we regard the case
1145		 * as an FQDN query for backward compatibility.
1146		 * Older versions set a random value to this field,
1147		 * so it rarely varies in the defined qtypes.
1148		 * But the mechanism is not reliable...
1149		 * maybe we should obsolete older versions.
1150		 */
1151		qtype = NI_QTYPE_FQDN;
1152		/* XXX will append a mbuf */
1153		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1154		oldfqdn++;
1155		break;
1156	}
1157
1158	/* validate query Subject field. */
1159	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1160	switch (qtype) {
1161	case NI_QTYPE_NOOP:
1162	case NI_QTYPE_SUPTYPES:
1163		if (subjlen != 0)
1164			goto bad;
1165		break;
1166
1167	case NI_QTYPE_FQDN:
1168	case NI_QTYPE_NODEADDR:
1169		switch (ni6->ni_code) {
1170		case ICMP6_NI_SUBJ_IPV6:
1171#if ICMP6_NI_SUBJ_IPV6 != 0
1172		case 0:
1173#endif
1174			/*
1175			 * backward compatibility - try to accept 03 draft
1176			 * format, where no Subject is present.
1177			 */
1178			if (subjlen == 0) {
1179				oldfqdn++;
1180				break;
1181			}
1182
1183			if (subjlen != sizeof(sin6.sin6_addr))
1184				goto bad;
1185
1186			/*
1187			 * Validate Subject address.
1188			 *
1189			 * Not sure what exactly does "address belongs to the
1190			 * node" mean in the spec, is it just unicast, or what?
1191			 *
1192			 * At this moment we consider Subject address as
1193			 * "belong to the node" if the Subject address equals
1194			 * to the IPv6 destination address; validation for
1195			 * IPv6 destination address should have done enough
1196			 * check for us.
1197			 *
1198			 * We do not do proxy at this moment.
1199			 */
1200			/* m_pulldown instead of copy? */
1201			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1202			    subjlen, (caddr_t)&sin6.sin6_addr);
1203			/* XXX kame scope hack */
1204			if (IN6_IS_SCOPE_LINKLOCAL(&sin6.sin6_addr)) {
1205#ifdef FAKE_LOOPBACK_IF
1206				if ((m->m_flags & M_PKTHDR) != 0 &&
1207				    m->m_pkthdr.rcvif) {
1208					sin6.sin6_addr.s6_addr16[1] =
1209					    htons(m->m_pkthdr.rcvif->if_index);
1210				}
1211#else
1212				if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) {
1213					sin6.sin6_addr.s6_addr16[1] =
1214					    ip6->ip6_dst.s6_addr16[1];
1215				}
1216#endif
1217			}
1218			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &sin6.sin6_addr))
1219				break;
1220			/*
1221			 * XXX if we are to allow other cases, we should really
1222			 * be careful about scope here.
1223			 * basically, we should disallow queries toward IPv6
1224			 * destination X with subject Y, if scope(X) > scope(Y).
1225			 * if we allow scope(X) > scope(Y), it will result in
1226			 * information leakage across scope boundary.
1227			 */
1228			goto bad;
1229
1230		case ICMP6_NI_SUBJ_FQDN:
1231			/*
1232			 * Validate Subject name with gethostname(3).
1233			 *
1234			 * The behavior may need some debate, since:
1235			 * - we are not sure if the node has FQDN as
1236			 *   hostname (returned by gethostname(3)).
1237			 * - the code does wildcard match for truncated names.
1238			 *   however, we are not sure if we want to perform
1239			 *   wildcard match, if gethostname(3) side has
1240			 *   truncated hostname.
1241			 */
1242			n = ni6_nametodns(hostname, hostnamelen, 0);
1243			if (!n || n->m_next || n->m_len == 0)
1244				goto bad;
1245			IP6_EXTHDR_GET(subj, char *, m,
1246			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1247			if (subj == NULL)
1248				goto bad;
1249			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1250					n->m_len)) {
1251				goto bad;
1252			}
1253			m_freem(n);
1254			n = NULL;
1255			break;
1256
1257		case ICMP6_NI_SUBJ_IPV4:	/* xxx: to be implemented? */
1258		default:
1259			goto bad;
1260		}
1261		break;
1262
1263	default:
1264		/* should never be here due to "switch (qtype)" above */
1265		goto bad;
1266	}
1267
1268	/* allocate a mbuf to reply. */
1269	MGETHDR(n, M_DONTWAIT, m->m_type);
1270	if (n == NULL) {
1271		m_freem(m);
1272		return(NULL);
1273	}
1274	M_COPY_PKTHDR(n, m); /* just for recvif */
1275	if (replylen > MHLEN) {
1276		if (replylen > MCLBYTES) {
1277			 /*
1278			  * XXX: should we try to allocate more? But MCLBYTES is
1279			  * probably much larger than IPV6_MMTU...
1280			  */
1281			goto bad;
1282		}
1283		MCLGET(n, M_DONTWAIT);
1284		if ((n->m_flags & M_EXT) == 0) {
1285			goto bad;
1286		}
1287	}
1288	n->m_pkthdr.len = n->m_len = replylen;
1289
1290	/* copy mbuf header and IPv6 + Node Information base headers */
1291	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1292	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1293	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1294
1295	/* qtype dependent procedure */
1296	switch (qtype) {
1297	case NI_QTYPE_NOOP:
1298		nni6->ni_flags = 0;
1299		break;
1300	case NI_QTYPE_SUPTYPES:
1301		goto bad;	/* xxx: to be implemented */
1302		break;
1303	case NI_QTYPE_FQDN:
1304		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1305						sizeof(struct ip6_hdr) +
1306						sizeof(struct icmp6_nodeinfo));
1307		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1308		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1309		/*
1310		 * XXX do we really have FQDN in variable "hostname"?
1311		 */
1312		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1313		if (n->m_next == NULL)
1314			goto bad;
1315		/* XXX we assume that n->m_next is not a chain */
1316		if (n->m_next->m_next != NULL)
1317			goto bad;
1318		n->m_pkthdr.len += n->m_next->m_len;
1319		break;
1320	case NI_QTYPE_NODEADDR:
1321	{
1322		int lenlim, copied;
1323
1324		if (n->m_flags & M_EXT)
1325			lenlim = MCLBYTES - sizeof(struct ip6_hdr) -
1326				sizeof(struct icmp6_nodeinfo);
1327		else
1328			lenlim = MHLEN - sizeof(struct ip6_hdr) -
1329				sizeof(struct icmp6_nodeinfo);
1330		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1331		/* XXX: reset mbuf length */
1332		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1333			sizeof(struct icmp6_nodeinfo) + copied;
1334		break;
1335	}
1336	default:
1337		break;		/* XXX impossible! */
1338	}
1339
1340	nni6->ni_type = ICMP6_NI_REPLY;
1341	nni6->ni_code = ICMP6_NI_SUCESS;
1342	m_freem(m);
1343	return(n);
1344
1345  bad:
1346	m_freem(m);
1347	if (n)
1348		m_freem(n);
1349	return(NULL);
1350}
1351#undef hostnamelen
1352
1353/*
1354 * make a mbuf with DNS-encoded string.  no compression support.
1355 *
1356 * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1357 * treated as truncated name (two \0 at the end).  this is a wild guess.
1358 */
1359static struct mbuf *
1360ni6_nametodns(name, namelen, old)
1361	const char *name;
1362	int namelen;
1363	int old;	/* return pascal string if non-zero */
1364{
1365	struct mbuf *m;
1366	char *cp, *ep;
1367	const char *p, *q;
1368	int i, len, nterm;
1369
1370	if (old)
1371		len = namelen + 1;
1372	else
1373		len = MCLBYTES;
1374
1375	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1376	MGET(m, M_DONTWAIT, MT_DATA);
1377	if (m && len > MLEN) {
1378		MCLGET(m, M_DONTWAIT);
1379		if ((m->m_flags & M_EXT) == 0)
1380			goto fail;
1381	}
1382	if (!m)
1383		goto fail;
1384	m->m_next = NULL;
1385
1386	if (old) {
1387		m->m_len = len;
1388		*mtod(m, char *) = namelen;
1389		bcopy(name, mtod(m, char *) + 1, namelen);
1390		return m;
1391	} else {
1392		m->m_len = 0;
1393		cp = mtod(m, char *);
1394		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1395
1396		/* if not certain about my name, return empty buffer */
1397		if (namelen == 0)
1398			return m;
1399
1400		/*
1401		 * guess if it looks like shortened hostname, or FQDN.
1402		 * shortened hostname needs two trailing "\0".
1403		 */
1404		i = 0;
1405		for (p = name; p < name + namelen; p++) {
1406			if (*p && *p == '.')
1407				i++;
1408		}
1409		if (i < 2)
1410			nterm = 2;
1411		else
1412			nterm = 1;
1413
1414		p = name;
1415		while (cp < ep && p < name + namelen) {
1416			i = 0;
1417			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1418				i++;
1419			/* result does not fit into mbuf */
1420			if (cp + i + 1 >= ep)
1421				goto fail;
1422			/* DNS label length restriction, RFC1035 page 8 */
1423			if (i >= 64)
1424				goto fail;
1425			*cp++ = i;
1426			bcopy(p, cp, i);
1427			cp += i;
1428			p = q;
1429			if (p < name + namelen && *p == '.')
1430				p++;
1431		}
1432		/* termination */
1433		if (cp + nterm >= ep)
1434			goto fail;
1435		while (nterm-- > 0)
1436			*cp++ = '\0';
1437		m->m_len = cp - mtod(m, char *);
1438		return m;
1439	}
1440
1441	panic("should not reach here");
1442	/*NOTREACHED*/
1443
1444 fail:
1445	if (m)
1446		m_freem(m);
1447	return NULL;
1448}
1449
1450/*
1451 * check if two DNS-encoded string matches.  takes care of truncated
1452 * form (with \0\0 at the end).  no compression support.
1453 */
1454static int
1455ni6_dnsmatch(a, alen, b, blen)
1456	const char *a;
1457	int alen;
1458	const char *b;
1459	int blen;
1460{
1461	const char *a0, *b0;
1462	int l;
1463
1464	/* simplest case - need validation? */
1465	if (alen == blen && bcmp(a, b, alen) == 0)
1466		return 1;
1467
1468	a0 = a;
1469	b0 = b;
1470
1471	/* termination is mandatory */
1472	if (alen < 2 || blen < 2)
1473		return 0;
1474	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1475		return 0;
1476	alen--;
1477	blen--;
1478
1479	while (a - a0 < alen && b - b0 < blen) {
1480		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1481			return 0;
1482
1483		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1484			return 0;
1485		/* we don't support compression yet */
1486		if (a[0] >= 64 || b[0] >= 64)
1487			return 0;
1488
1489		/* truncated case */
1490		if (a[0] == 0 && a - a0 == alen - 1)
1491			return 1;
1492		if (b[0] == 0 && b - b0 == blen - 1)
1493			return 1;
1494		if (a[0] == 0 || b[0] == 0)
1495			return 0;
1496
1497		if (a[0] != b[0])
1498			return 0;
1499		l = a[0];
1500		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1501			return 0;
1502		if (bcmp(a + 1, b + 1, l) != 0)
1503			return 0;
1504
1505		a += 1 + l;
1506		b += 1 + l;
1507	}
1508
1509	if (a - a0 == alen && b - b0 == blen)
1510		return 1;
1511	else
1512		return 0;
1513}
1514
1515/*
1516 * calculate the number of addresses to be returned in the node info reply.
1517 */
1518static int
1519ni6_addrs(ni6, m, ifpp)
1520	struct icmp6_nodeinfo *ni6;
1521	struct mbuf *m;
1522	struct ifnet **ifpp;
1523{
1524	register struct ifnet *ifp;
1525	register struct in6_ifaddr *ifa6;
1526	register struct ifaddr *ifa;
1527	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1528	int addrs = 0, addrsofif, iffound = 0;
1529
1530	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
1531	{
1532		addrsofif = 0;
1533		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1534		     ifa = ifa->ifa_list.tqe_next)
1535		{
1536			if (ifa->ifa_addr->sa_family != AF_INET6)
1537				continue;
1538			ifa6 = (struct in6_ifaddr *)ifa;
1539
1540			if (!(ni6->ni_flags & NI_NODEADDR_FLAG_ALL) &&
1541			    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1542					       &ifa6->ia_addr.sin6_addr))
1543				iffound = 1;
1544
1545			/*
1546			 * IPv4-mapped addresses can only be returned by a
1547			 * Node Information proxy, since they represent
1548			 * addresses of IPv4-only nodes, which perforce do
1549			 * not implement this protocol.
1550			 * [icmp-name-lookups-05]
1551			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1552			 * this function at this moment.
1553			 */
1554
1555			if (ifa6->ia6_flags & IN6_IFF_ANYCAST)
1556				continue; /* we need only unicast addresses */
1557
1558			if ((ni6->ni_flags & (NI_NODEADDR_FLAG_LINKLOCAL |
1559					      NI_NODEADDR_FLAG_SITELOCAL |
1560					      NI_NODEADDR_FLAG_GLOBAL)) == 0)
1561				continue;
1562
1563			/* What do we have to do about ::1? */
1564			switch(in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1565			 case IPV6_ADDR_SCOPE_LINKLOCAL:
1566				if (ni6->ni_flags & NI_NODEADDR_FLAG_LINKLOCAL)
1567					addrsofif++;
1568				break;
1569			 case IPV6_ADDR_SCOPE_SITELOCAL:
1570				if (ni6->ni_flags & NI_NODEADDR_FLAG_SITELOCAL)
1571					addrsofif++;
1572				break;
1573			 case IPV6_ADDR_SCOPE_GLOBAL:
1574				 if (ni6->ni_flags & NI_NODEADDR_FLAG_GLOBAL)
1575					 addrsofif++;
1576				 break;
1577			 default:
1578				 continue;
1579			}
1580		}
1581		if (iffound) {
1582			*ifpp = ifp;
1583			return(addrsofif);
1584		}
1585
1586		addrs += addrsofif;
1587	}
1588
1589	return(addrs);
1590}
1591
1592static int
1593ni6_store_addrs(ni6, nni6, ifp0, resid)
1594	struct icmp6_nodeinfo *ni6, *nni6;
1595	struct ifnet *ifp0;
1596	int resid;
1597{
1598	register struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1599	register struct in6_ifaddr *ifa6;
1600	register struct ifaddr *ifa;
1601	int docopy, copied = 0;
1602	u_char *cp = (u_char *)(nni6 + 1);
1603
1604	if (ifp0 == NULL && !(ni6->ni_flags & NI_NODEADDR_FLAG_ALL))
1605		return(0);	/* needless to copy */
1606
1607	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list))
1608	{
1609		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1610		     ifa = ifa->ifa_list.tqe_next)
1611		{
1612			docopy = 0;
1613
1614			if (ifa->ifa_addr->sa_family != AF_INET6)
1615				continue;
1616			ifa6 = (struct in6_ifaddr *)ifa;
1617
1618			if (ifa6->ia6_flags & IN6_IFF_ANYCAST) {
1619				/* just experimental. not in the spec. */
1620				if (ni6->ni_flags & NI_NODEADDR_FLAG_ANYCAST)
1621					docopy = 1;
1622				else
1623					continue;
1624			}
1625			else {	/* unicast address */
1626				if (ni6->ni_flags & NI_NODEADDR_FLAG_ANYCAST)
1627					continue;
1628				else
1629					docopy = 1;
1630			}
1631
1632			/* What do we have to do about ::1? */
1633			switch(in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1634			 case IPV6_ADDR_SCOPE_LINKLOCAL:
1635				if (ni6->ni_flags & NI_NODEADDR_FLAG_LINKLOCAL)
1636					docopy = 1;
1637				break;
1638			 case IPV6_ADDR_SCOPE_SITELOCAL:
1639				if (ni6->ni_flags & NI_NODEADDR_FLAG_SITELOCAL)
1640					docopy = 1;
1641				break;
1642			 case IPV6_ADDR_SCOPE_GLOBAL:
1643				 if (ni6->ni_flags & NI_NODEADDR_FLAG_GLOBAL)
1644					 docopy = 1;
1645				 break;
1646			 default:
1647				 continue;
1648			}
1649
1650			if (docopy) {
1651				if (resid < sizeof(struct in6_addr)) {
1652					/*
1653					 * We give up much more copy.
1654					 * Set the truncate flag and return.
1655					 */
1656					nni6->ni_flags |=
1657						NI_NODEADDR_FLAG_TRUNCATE;
1658					return(copied);
1659				}
1660				bcopy(&ifa6->ia_addr.sin6_addr, cp,
1661				      sizeof(struct in6_addr));
1662				/* XXX: KAME link-local hack; remove ifindex */
1663				if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr))
1664					((struct in6_addr *)cp)->s6_addr16[1] = 0;
1665				cp += sizeof(struct in6_addr);
1666				resid -= sizeof(struct in6_addr);
1667				copied += sizeof(struct in6_addr);
1668			}
1669		}
1670		if (ifp0)	/* we need search only on the specified IF */
1671			break;
1672	}
1673
1674	return(copied);
1675}
1676
1677/*
1678 * XXX almost dup'ed code with rip6_input.
1679 */
1680static int
1681icmp6_rip6_input(mp, off)
1682	struct	mbuf **mp;
1683	int	off;
1684{
1685	struct mbuf *m = *mp;
1686	register struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1687	register struct in6pcb *in6p;
1688	struct in6pcb *last = NULL;
1689	struct sockaddr_in6 rip6src;
1690	struct icmp6_hdr *icmp6;
1691	struct mbuf *opts = NULL;
1692
1693#ifndef PULLDOWN_TEST
1694	/* this is assumed to be safe. */
1695	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1696#else
1697	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1698	if (icmp6 == NULL) {
1699		/* m is already reclaimed */
1700		return IPPROTO_DONE;
1701	}
1702#endif
1703
1704	bzero(&rip6src, sizeof(rip6src));
1705	rip6src.sin6_len = sizeof(struct sockaddr_in6);
1706	rip6src.sin6_family = AF_INET6;
1707	rip6src.sin6_addr = ip6->ip6_src;
1708	if (IN6_IS_SCOPE_LINKLOCAL(&rip6src.sin6_addr))
1709		rip6src.sin6_addr.s6_addr16[1] = 0;
1710	if (m->m_pkthdr.rcvif) {
1711		if (IN6_IS_SCOPE_LINKLOCAL(&rip6src.sin6_addr))
1712			rip6src.sin6_scope_id = m->m_pkthdr.rcvif->if_index;
1713		else
1714			rip6src.sin6_scope_id = 0;
1715	} else
1716		rip6src.sin6_scope_id = 0;
1717
1718	for (in6p = rawin6pcb.in6p_next;
1719	     in6p != &rawin6pcb; in6p = in6p->in6p_next)
1720	{
1721		if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1722			continue;
1723		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1724		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1725			continue;
1726		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1727		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1728			continue;
1729		if (in6p->in6p_icmp6filt
1730		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1731				 in6p->in6p_icmp6filt))
1732			continue;
1733		if (last) {
1734			struct	mbuf *n;
1735			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1736				if (last->in6p_flags & IN6P_CONTROLOPTS)
1737					ip6_savecontrol(last, &opts, ip6, n);
1738				/* strip intermediate headers */
1739				m_adj(n, off);
1740				if (sbappendaddr(&last->in6p_socket->so_rcv,
1741						 (struct sockaddr *)&rip6src,
1742						 n, opts) == 0) {
1743					/* should notify about lost packet */
1744					m_freem(n);
1745					if (opts)
1746						m_freem(opts);
1747				} else
1748					sorwakeup(last->in6p_socket);
1749				opts = NULL;
1750			}
1751		}
1752		last = in6p;
1753	}
1754	if (last) {
1755		if (last->in6p_flags & IN6P_CONTROLOPTS)
1756			ip6_savecontrol(last, &opts, ip6, m);
1757		/* strip intermediate headers */
1758		m_adj(m, off);
1759		if (sbappendaddr(&last->in6p_socket->so_rcv,
1760				(struct sockaddr *)&rip6src, m, opts) == 0) {
1761			m_freem(m);
1762			if (opts)
1763				m_freem(opts);
1764		} else
1765			sorwakeup(last->in6p_socket);
1766	} else {
1767		m_freem(m);
1768		ip6stat.ip6s_delivered--;
1769	}
1770	return IPPROTO_DONE;
1771}
1772
1773/*
1774 * Reflect the ip6 packet back to the source.
1775 * OFF points to the icmp6 header, counted from the top of the mbuf.
1776 */
1777void
1778icmp6_reflect(m, off)
1779	struct	mbuf *m;
1780	size_t off;
1781{
1782	struct ip6_hdr *ip6;
1783	struct icmp6_hdr *icmp6;
1784	struct in6_ifaddr *ia;
1785	struct in6_addr t, *src = 0;
1786	int plen;
1787	int type, code;
1788	struct ifnet *outif = NULL;
1789#ifdef COMPAT_RFC1885
1790	int mtu = IPV6_MMTU;
1791	struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
1792#endif
1793
1794	/* too short to reflect */
1795	if (off < sizeof(struct ip6_hdr)) {
1796		printf("sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
1797		       (u_long)off, (u_long)sizeof(struct ip6_hdr),
1798		       __FILE__, __LINE__);
1799		goto bad;
1800	}
1801
1802	/*
1803	 * If there are extra headers between IPv6 and ICMPv6, strip
1804	 * off that header first.
1805	 */
1806#ifdef DIAGNOSTIC
1807	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
1808		panic("assumption failed in icmp6_reflect");
1809#endif
1810	if (off > sizeof(struct ip6_hdr)) {
1811		size_t l;
1812		struct ip6_hdr nip6;
1813
1814		l = off - sizeof(struct ip6_hdr);
1815		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
1816		m_adj(m, l);
1817		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1818		if (m->m_len < l) {
1819			if ((m = m_pullup(m, l)) == NULL)
1820				return;
1821		}
1822		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
1823	} else /* off == sizeof(struct ip6_hdr) */ {
1824		size_t l;
1825		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1826		if (m->m_len < l) {
1827			if ((m = m_pullup(m, l)) == NULL)
1828				return;
1829		}
1830	}
1831	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
1832	ip6 = mtod(m, struct ip6_hdr *);
1833	ip6->ip6_nxt = IPPROTO_ICMPV6;
1834	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
1835	type = icmp6->icmp6_type; /* keep type for statistics */
1836	code = icmp6->icmp6_code; /* ditto. */
1837
1838	t = ip6->ip6_dst;
1839	/*
1840	 * ip6_input() drops a packet if its src is multicast.
1841	 * So, the src is never multicast.
1842	 */
1843	ip6->ip6_dst = ip6->ip6_src;
1844
1845	/* XXX hack for link-local addresses */
1846	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
1847		ip6->ip6_dst.s6_addr16[1] =
1848			htons(m->m_pkthdr.rcvif->if_index);
1849	if (IN6_IS_ADDR_LINKLOCAL(&t))
1850		t.s6_addr16[1] = htons(m->m_pkthdr.rcvif->if_index);
1851
1852#ifdef COMPAT_RFC1885
1853	/*
1854	 * xxx guess MTU
1855	 * RFC 1885 requires that echo reply should be truncated if it
1856	 * does not fit in with (return) path MTU, but the description was
1857	 * removed in the new spec.
1858	 */
1859	if (icmp6_reflect_rt.ro_rt == 0 ||
1860	    ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
1861		if (icmp6_reflect_rt.ro_rt) {
1862			icmp6_reflect_rt.ro_rt = 0;
1863		}
1864		bzero(sin6, sizeof(*sin6));
1865		sin6->sin6_family = PF_INET6;
1866		sin6->sin6_len = sizeof(struct sockaddr_in6);
1867		sin6->sin6_addr = ip6->ip6_dst;
1868
1869		rtalloc((struct route *)&icmp6_reflect_rt.ro_rt);
1870	}
1871
1872	if (icmp6_reflect_rt.ro_rt == 0)
1873		goto bad;
1874
1875	if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
1876	    && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
1877		mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
1878
1879	if (mtu < m->m_pkthdr.len) {
1880		plen -= (m->m_pkthdr.len - mtu);
1881		m_adj(m, mtu - m->m_pkthdr.len);
1882	}
1883#endif
1884	/*
1885	 * If the incoming packet was addressed directly to us(i.e. unicast),
1886	 * use dst as the src for the reply.
1887	 * The IN6_IFF_NOTREADY case would be VERY rare, but is possible
1888	 * (for example) when we encounter an error while forwarding procedure
1889	 * destined to a duplicated address of ours.
1890	 */
1891	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
1892		if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
1893		    (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
1894			src = &t;
1895			break;
1896		}
1897	if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
1898		/*
1899		 * This is the case if the dst is our link-local address
1900		 * and the sender is also ourseleves.
1901		 */
1902		src = &t;
1903	}
1904
1905	if (src == 0)
1906		/*
1907		 * This case matches to multicasts, our anycast, or unicasts
1908		 * that we do not own. Select a source address which has the
1909		 * same scope.
1910		 * XXX: for (non link-local) multicast addresses, this might
1911		 * not be a good choice.
1912		 */
1913		if ((ia = in6_ifawithscope(m->m_pkthdr.rcvif, &t)) != 0)
1914			src = &IA6_SIN6(ia)->sin6_addr;
1915
1916	if (src == 0)
1917		goto bad;
1918
1919	ip6->ip6_src = *src;
1920
1921	ip6->ip6_flow = 0;
1922	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
1923	ip6->ip6_vfc |= IPV6_VERSION;
1924	ip6->ip6_nxt = IPPROTO_ICMPV6;
1925	if (m->m_pkthdr.rcvif) {
1926		/* XXX: This may not be the outgoing interface */
1927		ip6->ip6_hlim = nd_ifinfo[m->m_pkthdr.rcvif->if_index].chlim;
1928	}
1929
1930	icmp6->icmp6_cksum = 0;
1931	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
1932					sizeof(struct ip6_hdr), plen);
1933
1934	/*
1935	 * xxx option handling
1936	 */
1937
1938	m->m_flags &= ~(M_BCAST|M_MCAST);
1939#ifdef IPSEC
1940	/* Don't lookup socket */
1941	ipsec_setsocket(m, NULL);
1942#endif /*IPSEC*/
1943
1944#ifdef COMPAT_RFC1885
1945	ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif);
1946#else
1947	ip6_output(m, NULL, NULL, 0, NULL, &outif);
1948#endif
1949	if (outif)
1950		icmp6_ifoutstat_inc(outif, type, code);
1951
1952	return;
1953
1954 bad:
1955	m_freem(m);
1956	return;
1957}
1958
1959void
1960icmp6_fasttimo()
1961{
1962
1963	mld6_fasttimeo();
1964
1965	/* reset ICMPv6 pps limit */
1966	icmp6errpps_count = 0;
1967}
1968
1969static const char *
1970icmp6_redirect_diag(src6, dst6, tgt6)
1971	struct in6_addr *src6;
1972	struct in6_addr *dst6;
1973	struct in6_addr *tgt6;
1974{
1975	static char buf[1024];
1976	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
1977		ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
1978	return buf;
1979}
1980
1981void
1982icmp6_redirect_input(m, off)
1983	register struct mbuf *m;
1984	int off;
1985{
1986	struct ifnet *ifp = m->m_pkthdr.rcvif;
1987	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1988	struct nd_redirect *nd_rd;
1989	int icmp6len = ntohs(ip6->ip6_plen);
1990	char *lladdr = NULL;
1991	int lladdrlen = 0;
1992	u_char *redirhdr = NULL;
1993	int redirhdrlen = 0;
1994	struct rtentry *rt = NULL;
1995	int is_router;
1996	int is_onlink;
1997	struct in6_addr src6 = ip6->ip6_src;
1998	struct in6_addr redtgt6;
1999	struct in6_addr reddst6;
2000	union nd_opts ndopts;
2001
2002	if (!m || !ifp)
2003		return;
2004
2005	/* XXX if we are router, we don't update route by icmp6 redirect */
2006	if (ip6_forwarding)
2007		goto freeit;
2008	if (!icmp6_rediraccept)
2009		goto freeit;
2010
2011#ifndef PULLDOWN_TEST
2012	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2013	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2014#else
2015	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2016	if (nd_rd == NULL) {
2017		icmp6stat.icp6s_tooshort++;
2018		return;
2019	}
2020#endif
2021	redtgt6 = nd_rd->nd_rd_target;
2022	reddst6 = nd_rd->nd_rd_dst;
2023
2024	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2025		redtgt6.s6_addr16[1] = htons(ifp->if_index);
2026	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2027		reddst6.s6_addr16[1] = htons(ifp->if_index);
2028
2029	/* validation */
2030	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2031		log(LOG_ERR,
2032			"ICMP6 redirect sent from %s rejected; "
2033			"must be from linklocal\n", ip6_sprintf(&src6));
2034		goto freeit;
2035	}
2036	if (ip6->ip6_hlim != 255) {
2037		log(LOG_ERR,
2038			"ICMP6 redirect sent from %s rejected; "
2039			"hlim=%d (must be 255)\n",
2040			ip6_sprintf(&src6), ip6->ip6_hlim);
2041		goto freeit;
2042	}
2043    {
2044	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2045	struct sockaddr_in6 sin6;
2046	struct in6_addr *gw6;
2047
2048	bzero(&sin6, sizeof(sin6));
2049	sin6.sin6_family = AF_INET6;
2050	sin6.sin6_len = sizeof(struct sockaddr_in6);
2051	bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2052	rt = rtalloc1((struct sockaddr *)&sin6, 0);
2053	if (rt) {
2054		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2055		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2056			log(LOG_ERR,
2057				"ICMP6 redirect rejected; "
2058				"not equal to gw-for-src=%s (must be same): "
2059				"%s\n",
2060				ip6_sprintf(gw6),
2061				icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2062			RTFREE(rt);
2063			goto freeit;
2064		}
2065	} else {
2066		log(LOG_ERR,
2067			"ICMP6 redirect rejected; "
2068			"no route found for redirect dst: %s\n",
2069			icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2070		goto freeit;
2071	}
2072	RTFREE(rt);
2073	rt = NULL;
2074    }
2075	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2076		log(LOG_ERR,
2077			"ICMP6 redirect rejected; "
2078			"redirect dst must be unicast: %s\n",
2079			icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2080		goto freeit;
2081	}
2082
2083	is_router = is_onlink = 0;
2084	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2085		is_router = 1;	/* router case */
2086	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2087		is_onlink = 1;	/* on-link destination case */
2088	if (!is_router && !is_onlink) {
2089		log(LOG_ERR,
2090			"ICMP6 redirect rejected; "
2091			"neither router case nor onlink case: %s\n",
2092			icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2093		goto freeit;
2094	}
2095	/* validation passed */
2096
2097	icmp6len -= sizeof(*nd_rd);
2098	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2099	if (nd6_options(&ndopts) < 0) {
2100		log(LOG_INFO, "icmp6_redirect_input: "
2101			"invalid ND option, rejected: %s\n",
2102			icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2103		goto freeit;
2104	}
2105
2106	if (ndopts.nd_opts_tgt_lladdr) {
2107		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2108		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2109	}
2110
2111	if (ndopts.nd_opts_rh) {
2112		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2113		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2114	}
2115
2116	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2117		log(LOG_INFO,
2118			"icmp6_redirect_input: lladdrlen mismatch for %s "
2119			"(if %d, icmp6 packet %d): %s\n",
2120			ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2121			icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2122	}
2123
2124	/* RFC 2461 8.3 */
2125	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2126			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2127
2128	if (!is_onlink) {	/* better router case. perform rtredirect. */
2129		/* perform rtredirect */
2130		struct sockaddr_in6 sdst;
2131		struct sockaddr_in6 sgw;
2132		struct sockaddr_in6 ssrc;
2133
2134		bzero(&sdst, sizeof(sdst));
2135		bzero(&sgw, sizeof(sgw));
2136		bzero(&ssrc, sizeof(ssrc));
2137		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2138		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2139			sizeof(struct sockaddr_in6);
2140		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2141		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2142		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2143		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2144			   (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2145			   (struct sockaddr *)&ssrc,
2146			   (struct rtentry **)NULL
2147			   );
2148	}
2149	/* finally update cached route in each socket via pfctlinput */
2150    {
2151	struct sockaddr_in6 sdst;
2152#if 1
2153#else
2154	struct ip6protosw *pr;
2155#endif
2156
2157	bzero(&sdst, sizeof(sdst));
2158	sdst.sin6_family = AF_INET6;
2159	sdst.sin6_len = sizeof(struct sockaddr_in6);
2160	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2161#if 1
2162	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2163#else
2164	/*
2165	 * do not use pfctlinput() here, we have different prototype for
2166	 * xx_ctlinput() in ip6proto.
2167	 */
2168	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
2169	     pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW;
2170	     pr++) {
2171		if (pr->pr_ctlinput) {
2172			(*pr->pr_ctlinput)(PRC_REDIRECT_HOST,
2173				(struct sockaddr *)&sdst, NULL, NULL, 0);
2174		}
2175	}
2176#endif
2177#ifdef IPSEC
2178	key_sa_routechange((struct sockaddr *)&sdst);
2179#endif
2180    }
2181
2182 freeit:
2183	m_freem(m);
2184}
2185
2186void
2187icmp6_redirect_output(m0, rt)
2188	struct mbuf *m0;
2189	struct rtentry *rt;
2190{
2191	struct ifnet *ifp;	/* my outgoing interface */
2192	struct in6_addr *ifp_ll6;
2193	struct in6_addr *router_ll6;
2194	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2195	struct mbuf *m = NULL;	/* newly allocated one */
2196	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2197	struct nd_redirect *nd_rd;
2198	size_t maxlen;
2199	u_char *p;
2200	struct ifnet *outif = NULL;
2201	struct sockaddr_in6 src_sa;
2202
2203	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2204
2205	/* if we are not router, we don't send icmp6 redirect */
2206	if (!ip6_forwarding || ip6_accept_rtadv)
2207		goto fail;
2208
2209	/* sanity check */
2210	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2211		goto fail;
2212
2213	/*
2214	 * Address check:
2215	 *  the source address must identify a neighbor, and
2216	 *  the destination address must not be a multicast address
2217	 *  [RFC 2461, sec 8.2]
2218	 */
2219	sip6 = mtod(m0, struct ip6_hdr *);
2220	bzero(&src_sa, sizeof(src_sa));
2221	src_sa.sin6_family = AF_INET6;
2222	src_sa.sin6_len = sizeof(src_sa);
2223	src_sa.sin6_addr = sip6->ip6_src;
2224	/* we don't currently use sin6_scope_id, but eventually use it */
2225	src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src);
2226	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2227		goto fail;
2228	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2229		goto fail;	/* what should we do here? */
2230
2231	/* rate limit */
2232	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2233		goto fail;
2234
2235	/*
2236	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2237	 * we almost always ask for an mbuf cluster for simplicity.
2238	 * (MHLEN < IPV6_MMTU is almost always true)
2239	 */
2240#if IPV6_MMTU >= MCLBYTES
2241# error assumption failed about IPV6_MMTU and MCLBYTES
2242#endif
2243	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2244	if (m && IPV6_MMTU >= MHLEN)
2245		MCLGET(m, M_DONTWAIT);
2246	if (!m)
2247		goto fail;
2248	maxlen = (m->m_flags & M_EXT) ? MCLBYTES : MHLEN;
2249	maxlen = min(IPV6_MMTU, maxlen);
2250	/* just for safety */
2251	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2252	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2253		goto fail;
2254	}
2255
2256	{
2257		/* get ip6 linklocal address for ifp(my outgoing interface). */
2258		struct in6_ifaddr *ia;
2259		if ((ia = in6ifa_ifpforlinklocal(ifp,
2260						 IN6_IFF_NOTREADY|
2261						 IN6_IFF_ANYCAST)) == NULL)
2262			goto fail;
2263		ifp_ll6 = &ia->ia_addr.sin6_addr;
2264	}
2265
2266	/* get ip6 linklocal address for the router. */
2267	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2268		struct sockaddr_in6 *sin6;
2269		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2270		router_ll6 = &sin6->sin6_addr;
2271		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2272			router_ll6 = (struct in6_addr *)NULL;
2273	} else
2274		router_ll6 = (struct in6_addr *)NULL;
2275
2276	/* ip6 */
2277	ip6 = mtod(m, struct ip6_hdr *);
2278	ip6->ip6_flow = 0;
2279	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2280	ip6->ip6_vfc |= IPV6_VERSION;
2281	/* ip6->ip6_plen will be set later */
2282	ip6->ip6_nxt = IPPROTO_ICMPV6;
2283	ip6->ip6_hlim = 255;
2284	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2285	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2286	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2287
2288	/* ND Redirect */
2289	nd_rd = (struct nd_redirect *)(ip6 + 1);
2290	nd_rd->nd_rd_type = ND_REDIRECT;
2291	nd_rd->nd_rd_code = 0;
2292	nd_rd->nd_rd_reserved = 0;
2293	if (rt->rt_flags & RTF_GATEWAY) {
2294		/*
2295		 * nd_rd->nd_rd_target must be a link-local address in
2296		 * better router cases.
2297		 */
2298		if (!router_ll6)
2299			goto fail;
2300		bcopy(router_ll6, &nd_rd->nd_rd_target,
2301		      sizeof(nd_rd->nd_rd_target));
2302		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2303		      sizeof(nd_rd->nd_rd_dst));
2304	} else {
2305		/* make sure redtgt == reddst */
2306		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2307		      sizeof(nd_rd->nd_rd_target));
2308		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2309		      sizeof(nd_rd->nd_rd_dst));
2310	}
2311
2312	p = (u_char *)(nd_rd + 1);
2313
2314	if (!router_ll6)
2315		goto nolladdropt;
2316
2317    {
2318	/* target lladdr option */
2319	struct rtentry *rt_router = NULL;
2320	int len;
2321	struct sockaddr_dl *sdl;
2322	struct nd_opt_hdr *nd_opt;
2323	char *lladdr;
2324
2325	rt_router = nd6_lookup(router_ll6, 0, ifp);
2326	if (!rt_router)
2327		goto nolladdropt;
2328	len = sizeof(*nd_opt) + ifp->if_addrlen;
2329	len = (len + 7) & ~7;	/*round by 8*/
2330	/* safety check */
2331	if (len + (p - (u_char *)ip6) > maxlen)
2332		goto nolladdropt;
2333	if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2334	    (rt_router->rt_flags & RTF_LLINFO) &&
2335	    (rt_router->rt_gateway->sa_family == AF_LINK) &&
2336	    (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2337	    sdl->sdl_alen) {
2338		nd_opt = (struct nd_opt_hdr *)p;
2339		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2340		nd_opt->nd_opt_len = len >> 3;
2341		lladdr = (char *)(nd_opt + 1);
2342		bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2343		p += len;
2344	}
2345    }
2346nolladdropt:;
2347
2348	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2349
2350	/* just to be safe */
2351	if (m0->m_flags & M_DECRYPTED)
2352		goto noredhdropt;
2353	if (p - (u_char *)ip6 > maxlen)
2354		goto noredhdropt;
2355
2356    {
2357	/* redirected header option */
2358	int len;
2359	struct nd_opt_rd_hdr *nd_opt_rh;
2360
2361	/*
2362	 * compute the maximum size for icmp6 redirect header option.
2363	 * XXX room for auth header?
2364	 */
2365	len = maxlen - (p - (u_char *)ip6);
2366	len &= ~7;
2367
2368	/* This is just for simplicity. */
2369	if (m0->m_pkthdr.len != m0->m_len) {
2370		if (m0->m_next) {
2371			m_freem(m0->m_next);
2372			m0->m_next = NULL;
2373		}
2374		m0->m_pkthdr.len = m0->m_len;
2375	}
2376
2377	/*
2378	 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2379	 * about padding/truncate rule for the original IP packet.
2380	 * From the discussion on IPv6imp in Feb 1999, the consensus was:
2381	 * - "attach as much as possible" is the goal
2382	 * - pad if not aligned (original size can be guessed by original
2383	 *   ip6 header)
2384	 * Following code adds the padding if it is simple enough,
2385	 * and truncates if not.
2386	 */
2387	if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2388		panic("assumption failed in %s:%d\n", __FILE__, __LINE__);
2389
2390	if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2391		/* not enough room, truncate */
2392		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2393	} else {
2394		/* enough room, pad or truncate */
2395		size_t extra;
2396
2397		extra = m0->m_pkthdr.len % 8;
2398		if (extra) {
2399			/* pad if easy enough, truncate if not */
2400			if (8 - extra <= M_TRAILINGSPACE(m0)) {
2401				/* pad */
2402				m0->m_len += (8 - extra);
2403				m0->m_pkthdr.len += (8 - extra);
2404			} else {
2405				/* truncate */
2406				m0->m_pkthdr.len -= extra;
2407				m0->m_len -= extra;
2408			}
2409		}
2410		len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2411		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2412	}
2413
2414	nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2415	bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2416	nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2417	nd_opt_rh->nd_opt_rh_len = len >> 3;
2418	p += sizeof(*nd_opt_rh);
2419	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2420
2421	/* connect m0 to m */
2422	m->m_next = m0;
2423	m->m_pkthdr.len = m->m_len + m0->m_len;
2424    }
2425noredhdropt:;
2426
2427	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
2428		sip6->ip6_src.s6_addr16[1] = 0;
2429	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
2430		sip6->ip6_dst.s6_addr16[1] = 0;
2431#if 0
2432	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
2433		ip6->ip6_src.s6_addr16[1] = 0;
2434	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
2435		ip6->ip6_dst.s6_addr16[1] = 0;
2436#endif
2437	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
2438		nd_rd->nd_rd_target.s6_addr16[1] = 0;
2439	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
2440		nd_rd->nd_rd_dst.s6_addr16[1] = 0;
2441
2442	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2443
2444	nd_rd->nd_rd_cksum = 0;
2445	nd_rd->nd_rd_cksum
2446		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2447
2448	/* send the packet to outside... */
2449#ifdef IPSEC
2450	/* Don't lookup socket */
2451	ipsec_setsocket(m, NULL);
2452#endif /*IPSEC*/
2453	ip6_output(m, NULL, NULL, 0, NULL, &outif);
2454	if (outif) {
2455		icmp6_ifstat_inc(outif, ifs6_out_msg);
2456		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2457	}
2458	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2459
2460	return;
2461
2462fail:
2463	if (m)
2464		m_freem(m);
2465	if (m0)
2466		m_freem(m0);
2467}
2468
2469/*
2470 * ICMPv6 socket option processing.
2471 */
2472int
2473icmp6_ctloutput(op, so, level, optname, mp)
2474	int op;
2475	struct socket *so;
2476	int level, optname;
2477	struct mbuf **mp;
2478{
2479	register struct in6pcb *in6p = sotoin6pcb(so);
2480	register struct mbuf *m = *mp;
2481	int error = 0;
2482
2483	if (level != IPPROTO_ICMPV6) {
2484		error = EINVAL;
2485		if (op == PRCO_SETOPT && m)
2486			(void)m_free(m);
2487	} else switch(op) {
2488	 case PRCO_SETOPT:
2489		 switch (optname) {
2490		  case ICMP6_FILTER:
2491		  {
2492			  struct icmp6_filter *p;
2493
2494			  p = mtod(m, struct icmp6_filter *);
2495			  if (!p || !in6p->in6p_icmp6filt) {
2496				  error = EINVAL;
2497				  break;
2498			  }
2499			  bcopy(p, in6p->in6p_icmp6filt,
2500				sizeof(struct icmp6_filter));
2501			  error = 0;
2502			  break;
2503		  }
2504
2505		  default:
2506			  error = ENOPROTOOPT;
2507			  break;
2508		 }
2509		 if (m)
2510			 (void)m_free(m);
2511		 break;
2512
2513	 case PRCO_GETOPT:
2514		 switch (optname) {
2515		  case ICMP6_FILTER:
2516		  {
2517			  struct icmp6_filter *p;
2518
2519			  if (!in6p->in6p_icmp6filt) {
2520				  error = EINVAL;
2521				  break;
2522			  }
2523			  *mp = m = m_get(M_WAIT, MT_SOOPTS);
2524			  m->m_len = sizeof(struct icmp6_filter);
2525			  p = mtod(m, struct icmp6_filter *);
2526			  bcopy(in6p->in6p_icmp6filt, p,
2527				sizeof(struct icmp6_filter));
2528			  error = 0;
2529			  break;
2530		  }
2531
2532		  default:
2533			  error = ENOPROTOOPT;
2534			  break;
2535		 }
2536		 break;
2537	}
2538
2539	return(error);
2540}
2541
2542/*
2543 * Perform rate limit check.
2544 * Returns 0 if it is okay to send the icmp6 packet.
2545 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2546 * limitation.
2547 *
2548 * There are two limitations defined:
2549 * - pps limit: ICMPv6 error packet cannot exceed defined packet-per-second.
2550 *   we measure it every 0.2 second, since fasttimo works every 0.2 second.
2551 * - rate limit: ICMPv6 error packet cannot appear more than once per
2552 *   defined interval.
2553 * In any case, if we perform rate limitation, we'll see jitter in the ICMPv6
2554 * error packets.
2555 *
2556 * XXX per-destination/type check necessary?
2557 */
2558static int
2559icmp6_ratelimit(dst, type, code)
2560	const struct in6_addr *dst;	/* not used at this moment */
2561	const int type;			/* not used at this moment */
2562	const int code;			/* not used at this moment */
2563{
2564	int ret;
2565
2566	ret = 0;	/*okay to send*/
2567
2568	/* PPS limit */
2569	icmp6errpps_count++;
2570	if (icmp6errppslim && icmp6errpps_count > icmp6errppslim / 5) {
2571		/* The packet is subject to pps limit */
2572		ret++;
2573	}
2574
2575	if (!ratecheck(&icmp6errratelim_last, &icmp6errratelim)) {
2576		/* The packet is subject to rate limit */
2577		ret++;
2578	}
2579
2580	return ret;
2581}
2582
2583static struct rtentry *
2584icmp6_mtudisc_clone(dst)
2585	struct sockaddr *dst;
2586{
2587	struct rtentry *rt;
2588	int    error;
2589
2590	rt = rtalloc1(dst, 1);
2591	if (rt == 0)
2592		return NULL;
2593
2594	/* If we didn't get a host route, allocate one */
2595	if ((rt->rt_flags & RTF_HOST) == 0) {
2596		struct rtentry *nrt;
2597
2598		error = rtrequest((int) RTM_ADD, dst,
2599		    (struct sockaddr *) rt->rt_gateway,
2600		    (struct sockaddr *) 0,
2601		    RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt);
2602		if (error) {
2603			rtfree(rt);
2604			rtfree(nrt);
2605			return NULL;
2606		}
2607		nrt->rt_rmx = rt->rt_rmx;
2608		rtfree(rt);
2609		rt = nrt;
2610	}
2611	error = rt_timer_add(rt, icmp6_mtudisc_timeout,
2612			icmp6_mtudisc_timeout_q);
2613	if (error) {
2614		rtfree(rt);
2615		return NULL;
2616	}
2617
2618	return rt;	/* caller need to call rtfree() */
2619}
2620
2621static void
2622icmp6_mtudisc_timeout(rt, r)
2623	struct rtentry *rt;
2624	struct rttimer *r;
2625{
2626	if (rt == NULL)
2627		panic("icmp6_mtudisc_timeout: bad route to timeout");
2628	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
2629	    (RTF_DYNAMIC | RTF_HOST)) {
2630		rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt),
2631		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
2632	} else {
2633		if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) {
2634			rt->rt_rmx.rmx_mtu = 0;
2635		}
2636	}
2637}
2638
2639#include <uvm/uvm_extern.h>
2640#include <sys/sysctl.h>
2641
2642int
2643icmp6_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
2644	int *name;
2645	u_int namelen;
2646	void *oldp;
2647	size_t *oldlenp;
2648	void *newp;
2649	size_t newlen;
2650{
2651
2652	/* All sysctl names at this level are terminal. */
2653	if (namelen != 1)
2654		return ENOTDIR;
2655
2656	switch (name[0]) {
2657
2658	case ICMPV6CTL_REDIRACCEPT:
2659		return sysctl_int(oldp, oldlenp, newp, newlen,
2660				&icmp6_rediraccept);
2661	case ICMPV6CTL_REDIRTIMEOUT:
2662		return sysctl_int(oldp, oldlenp, newp, newlen,
2663				&icmp6_redirtimeout);
2664	case ICMPV6CTL_STATS:
2665		return sysctl_rdstruct(oldp, oldlenp, newp,
2666				&icmp6stat, sizeof(icmp6stat));
2667	case ICMPV6CTL_ERRRATELIMIT:
2668	    {
2669		int rate_usec, error, s;
2670
2671		/*
2672		 * The sysctl specifies the rate in usec-between-icmp,
2673		 * so we must convert from/to a timeval.
2674		 */
2675		rate_usec = (icmp6errratelim.tv_sec * 1000000) +
2676		    icmp6errratelim.tv_usec;
2677		error = sysctl_int(oldp, oldlenp, newp, newlen, &rate_usec);
2678		if (error)
2679			return (error);
2680		if (rate_usec < 0)
2681			return (EINVAL);
2682		s = splsoftnet();
2683		icmp6errratelim.tv_sec = rate_usec / 1000000;
2684		icmp6errratelim.tv_usec = rate_usec % 1000000;
2685		splx(s);
2686
2687		return (0);
2688	    }
2689	case ICMPV6CTL_ND6_PRUNE:
2690		return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_prune);
2691	case ICMPV6CTL_ND6_DELAY:
2692		return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_delay);
2693	case ICMPV6CTL_ND6_UMAXTRIES:
2694		return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_umaxtries);
2695	case ICMPV6CTL_ND6_MMAXTRIES:
2696		return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_mmaxtries);
2697	case ICMPV6CTL_ND6_USELOOPBACK:
2698		return sysctl_int(oldp, oldlenp, newp, newlen,
2699				&nd6_useloopback);
2700	case ICMPV6CTL_NODEINFO:
2701		return sysctl_int(oldp, oldlenp, newp, newlen, &icmp6_nodeinfo);
2702	case ICMPV6CTL_ERRPPSLIMIT:
2703		return sysctl_int(oldp, oldlenp, newp, newlen, &icmp6errppslim);
2704	case ICMPV6CTL_ND6_MAXNUDHINT:
2705		return sysctl_int(oldp, oldlenp, newp, newlen,
2706				&nd6_maxnudhint);
2707	default:
2708		return ENOPROTOOPT;
2709	}
2710	/* NOTREACHED */
2711}
2712