icmp6.c revision 112781
1/*	$FreeBSD: head/sys/netinet6/icmp6.c 112781 2003-03-29 08:31:28Z suz $	*/
2/*	$KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 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_inet6.h"
70#include "opt_ipsec.h"
71
72#include <sys/param.h>
73#include <sys/domain.h>
74#include <sys/kernel.h>
75#include <sys/lock.h>
76#include <sys/malloc.h>
77#include <sys/mbuf.h>
78#include <sys/protosw.h>
79#include <sys/signalvar.h>
80#include <sys/socket.h>
81#include <sys/socketvar.h>
82#include <sys/sx.h>
83#include <sys/syslog.h>
84#include <sys/systm.h>
85#include <sys/time.h>
86
87#include <net/if.h>
88#include <net/if_dl.h>
89#include <net/if_types.h>
90#include <net/route.h>
91
92#include <netinet/in.h>
93#include <netinet/in_pcb.h>
94#include <netinet/in_var.h>
95#include <netinet/ip6.h>
96#include <netinet/icmp6.h>
97#include <netinet6/in6_ifattach.h>
98#include <netinet6/in6_pcb.h>
99#include <netinet6/ip6protosw.h>
100#include <netinet6/ip6_var.h>
101#include <netinet6/mld6_var.h>
102#include <netinet6/nd6.h>
103
104#ifdef IPSEC
105#include <netinet6/ipsec.h>
106#include <netkey/key.h>
107#endif
108
109#ifdef FAST_IPSEC
110#include <netipsec/ipsec.h>
111#include <netipsec/key.h>
112#define	IPSEC
113#endif
114
115#include <net/net_osdep.h>
116
117#ifdef HAVE_NRL_INPCB
118/* inpcb members */
119#define in6pcb		inpcb
120#define in6p_laddr	inp_laddr6
121#define in6p_faddr	inp_faddr6
122#define in6p_icmp6filt	inp_icmp6filt
123#define in6p_route	inp_route
124#define in6p_socket	inp_socket
125#define in6p_flags	inp_flags
126#define in6p_moptions	inp_moptions6
127#define in6p_outputopts	inp_outputopts6
128#define in6p_ip6	inp_ipv6
129#define in6p_flowinfo	inp_flowinfo
130#define in6p_sp		inp_sp
131#define in6p_next	inp_next
132#define in6p_prev	inp_prev
133/* macro names */
134#define sotoin6pcb	sotoinpcb
135/* function names */
136#define in6_pcbdetach	in_pcbdetach
137#define in6_rtchange	in_rtchange
138
139/*
140 * for KAME src sync over BSD*'s. XXX: FreeBSD (>=3) are VERY different from
141 * others...
142 */
143#define in6p_ip6_nxt	inp_ipv6.ip6_nxt
144#endif
145
146extern struct domain inet6domain;
147
148struct icmp6stat icmp6stat;
149
150extern struct inpcbhead ripcb;
151extern int icmp6errppslim;
152static int icmp6errpps_count = 0;
153static struct timeval icmp6errppslim_last;
154extern int icmp6_nodeinfo;
155
156static void icmp6_errcount __P((struct icmp6errstat *, int, int));
157static int icmp6_rip6_input __P((struct mbuf **, int));
158static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int));
159static const char *icmp6_redirect_diag __P((struct in6_addr *,
160	struct in6_addr *, struct in6_addr *));
161#define	HAVE_PPSRATECHECK
162#ifndef HAVE_PPSRATECHECK
163static int ppsratecheck __P((struct timeval *, int *, int));
164#endif
165static struct mbuf *ni6_input __P((struct mbuf *, int));
166static struct mbuf *ni6_nametodns __P((const char *, int, int));
167static int ni6_dnsmatch __P((const char *, int, const char *, int));
168static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
169			  struct ifnet **, char *));
170static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
171				struct ifnet *, int));
172static int icmp6_notify_error __P((struct mbuf *, int, int, int));
173
174#ifdef COMPAT_RFC1885
175static struct route_in6 icmp6_reflect_rt;
176#endif
177
178
179void
180icmp6_init()
181{
182	mld6_init();
183}
184
185static void
186icmp6_errcount(stat, type, code)
187	struct icmp6errstat *stat;
188	int type, code;
189{
190	switch (type) {
191	case ICMP6_DST_UNREACH:
192		switch (code) {
193		case ICMP6_DST_UNREACH_NOROUTE:
194			stat->icp6errs_dst_unreach_noroute++;
195			return;
196		case ICMP6_DST_UNREACH_ADMIN:
197			stat->icp6errs_dst_unreach_admin++;
198			return;
199		case ICMP6_DST_UNREACH_BEYONDSCOPE:
200			stat->icp6errs_dst_unreach_beyondscope++;
201			return;
202		case ICMP6_DST_UNREACH_ADDR:
203			stat->icp6errs_dst_unreach_addr++;
204			return;
205		case ICMP6_DST_UNREACH_NOPORT:
206			stat->icp6errs_dst_unreach_noport++;
207			return;
208		}
209		break;
210	case ICMP6_PACKET_TOO_BIG:
211		stat->icp6errs_packet_too_big++;
212		return;
213	case ICMP6_TIME_EXCEEDED:
214		switch (code) {
215		case ICMP6_TIME_EXCEED_TRANSIT:
216			stat->icp6errs_time_exceed_transit++;
217			return;
218		case ICMP6_TIME_EXCEED_REASSEMBLY:
219			stat->icp6errs_time_exceed_reassembly++;
220			return;
221		}
222		break;
223	case ICMP6_PARAM_PROB:
224		switch (code) {
225		case ICMP6_PARAMPROB_HEADER:
226			stat->icp6errs_paramprob_header++;
227			return;
228		case ICMP6_PARAMPROB_NEXTHEADER:
229			stat->icp6errs_paramprob_nextheader++;
230			return;
231		case ICMP6_PARAMPROB_OPTION:
232			stat->icp6errs_paramprob_option++;
233			return;
234		}
235		break;
236	case ND_REDIRECT:
237		stat->icp6errs_redirect++;
238		return;
239	}
240	stat->icp6errs_unknown++;
241}
242
243/*
244 * Generate an error packet of type error in response to bad IP6 packet.
245 */
246void
247icmp6_error(m, type, code, param)
248	struct mbuf *m;
249	int type, code, param;
250{
251	struct ip6_hdr *oip6, *nip6;
252	struct icmp6_hdr *icmp6;
253	u_int preplen;
254	int off;
255	int nxt;
256
257	icmp6stat.icp6s_error++;
258
259	/* count per-type-code statistics */
260	icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
261
262#ifdef M_DECRYPTED	/*not openbsd*/
263	if (m->m_flags & M_DECRYPTED) {
264		icmp6stat.icp6s_canterror++;
265		goto freeit;
266	}
267#endif
268
269#ifndef PULLDOWN_TEST
270	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
271#else
272	if (m->m_len < sizeof(struct ip6_hdr)) {
273		m = m_pullup(m, sizeof(struct ip6_hdr));
274		if (m == NULL)
275			return;
276	}
277#endif
278	oip6 = mtod(m, struct ip6_hdr *);
279
280	/*
281	 * Multicast destination check. For unrecognized option errors,
282	 * this check has already done in ip6_unknown_opt(), so we can
283	 * check only for other errors.
284	 */
285	if ((m->m_flags & (M_BCAST|M_MCAST) ||
286	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
287	    (type != ICMP6_PACKET_TOO_BIG &&
288	     (type != ICMP6_PARAM_PROB ||
289	      code != ICMP6_PARAMPROB_OPTION)))
290		goto freeit;
291
292	/* Source address check. XXX: the case of anycast source? */
293	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
294	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
295		goto freeit;
296
297	/*
298	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
299	 * don't do it.
300	 */
301	nxt = -1;
302	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
303	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
304		struct icmp6_hdr *icp;
305
306#ifndef PULLDOWN_TEST
307		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
308		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
309#else
310		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
311			sizeof(*icp));
312		if (icp == NULL) {
313			icmp6stat.icp6s_tooshort++;
314			return;
315		}
316#endif
317		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
318		    icp->icmp6_type == ND_REDIRECT) {
319			/*
320			 * ICMPv6 error
321			 * Special case: for redirect (which is
322			 * informational) we must not send icmp6 error.
323			 */
324			icmp6stat.icp6s_canterror++;
325			goto freeit;
326		} else {
327			/* ICMPv6 informational - send the error */
328		}
329	} else {
330		/* non-ICMPv6 - send the error */
331	}
332
333	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
334
335	/* Finally, do rate limitation check. */
336	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
337		icmp6stat.icp6s_toofreq++;
338		goto freeit;
339	}
340
341	/*
342	 * OK, ICMP6 can be generated.
343	 */
344
345	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
346		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
347
348	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
349	M_PREPEND(m, preplen, M_DONTWAIT);
350	if (m && m->m_len < preplen)
351		m = m_pullup(m, preplen);
352	if (m == NULL) {
353		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
354		return;
355	}
356
357	nip6 = mtod(m, struct ip6_hdr *);
358	nip6->ip6_src  = oip6->ip6_src;
359	nip6->ip6_dst  = oip6->ip6_dst;
360
361	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src))
362		oip6->ip6_src.s6_addr16[1] = 0;
363	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst))
364		oip6->ip6_dst.s6_addr16[1] = 0;
365
366	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
367	icmp6->icmp6_type = type;
368	icmp6->icmp6_code = code;
369	icmp6->icmp6_pptr = htonl((u_int32_t)param);
370
371	/*
372	 * icmp6_reflect() is designed to be in the input path.
373	 * icmp6_error() can be called from both input and outut path,
374	 * and if we are in output path rcvif could contain bogus value.
375	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
376	 * information in ip header (nip6).
377	 */
378	m->m_pkthdr.rcvif = NULL;
379
380	icmp6stat.icp6s_outhist[type]++;
381	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
382
383	return;
384
385  freeit:
386	/*
387	 * If we can't tell wheter or not we can generate ICMP6, free it.
388	 */
389	m_freem(m);
390}
391
392/*
393 * Process a received ICMP6 message.
394 */
395int
396icmp6_input(mp, offp, proto)
397	struct mbuf **mp;
398	int *offp, proto;
399{
400	struct mbuf *m = *mp, *n;
401	struct ip6_hdr *ip6, *nip6;
402	struct icmp6_hdr *icmp6, *nicmp6;
403	int off = *offp;
404	int icmp6len = m->m_pkthdr.len - *offp;
405	int code, sum, noff;
406
407#ifndef PULLDOWN_TEST
408	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
409	/* m might change if M_LOOP.  So, call mtod after this */
410#endif
411
412	/*
413	 * Locate icmp6 structure in mbuf, and check
414	 * that not corrupted and of at least minimum length
415	 */
416
417	ip6 = mtod(m, struct ip6_hdr *);
418	if (icmp6len < sizeof(struct icmp6_hdr)) {
419		icmp6stat.icp6s_tooshort++;
420		goto freeit;
421	}
422
423	/*
424	 * calculate the checksum
425	 */
426#ifndef PULLDOWN_TEST
427	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
428#else
429	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
430	if (icmp6 == NULL) {
431		icmp6stat.icp6s_tooshort++;
432		return IPPROTO_DONE;
433	}
434#endif
435	code = icmp6->icmp6_code;
436
437	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
438		nd6log((LOG_ERR,
439		    "ICMP6 checksum error(%d|%x) %s\n",
440		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
441		icmp6stat.icp6s_checksum++;
442		goto freeit;
443	}
444
445	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
446		/*
447		 * Deliver very specific ICMP6 type only.
448		 * This is important to deilver TOOBIG.  Otherwise PMTUD
449		 * will not work.
450		 */
451		switch (icmp6->icmp6_type) {
452		case ICMP6_DST_UNREACH:
453		case ICMP6_PACKET_TOO_BIG:
454		case ICMP6_TIME_EXCEEDED:
455			break;
456		default:
457			goto freeit;
458		}
459	}
460
461	icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
462	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
463	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
464		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
465
466	switch (icmp6->icmp6_type) {
467	case ICMP6_DST_UNREACH:
468		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
469		switch (code) {
470		case ICMP6_DST_UNREACH_NOROUTE:
471			code = PRC_UNREACH_NET;
472			break;
473		case ICMP6_DST_UNREACH_ADMIN:
474			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
475			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
476			break;
477		case ICMP6_DST_UNREACH_ADDR:
478			code = PRC_HOSTDEAD;
479			break;
480#ifdef COMPAT_RFC1885
481		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
482			code = PRC_UNREACH_SRCFAIL;
483			break;
484#else
485		case ICMP6_DST_UNREACH_BEYONDSCOPE:
486			/* I mean "source address was incorrect." */
487			code = PRC_PARAMPROB;
488			break;
489#endif
490		case ICMP6_DST_UNREACH_NOPORT:
491			code = PRC_UNREACH_PORT;
492			break;
493		default:
494			goto badcode;
495		}
496		goto deliver;
497		break;
498
499	case ICMP6_PACKET_TOO_BIG:
500		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
501		if (code != 0)
502			goto badcode;
503
504		code = PRC_MSGSIZE;
505
506		/*
507		 * Updating the path MTU will be done after examining
508		 * intermediate extension headers.
509		 */
510		goto deliver;
511		break;
512
513	case ICMP6_TIME_EXCEEDED:
514		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
515		switch (code) {
516		case ICMP6_TIME_EXCEED_TRANSIT:
517		case ICMP6_TIME_EXCEED_REASSEMBLY:
518			code += PRC_TIMXCEED_INTRANS;
519			break;
520		default:
521			goto badcode;
522		}
523		goto deliver;
524		break;
525
526	case ICMP6_PARAM_PROB:
527		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
528		switch (code) {
529		case ICMP6_PARAMPROB_NEXTHEADER:
530			code = PRC_UNREACH_PROTOCOL;
531			break;
532		case ICMP6_PARAMPROB_HEADER:
533		case ICMP6_PARAMPROB_OPTION:
534			code = PRC_PARAMPROB;
535			break;
536		default:
537			goto badcode;
538		}
539		goto deliver;
540		break;
541
542	case ICMP6_ECHO_REQUEST:
543		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
544		if (code != 0)
545			goto badcode;
546		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
547			/* Give up remote */
548			break;
549		}
550		if ((n->m_flags & M_EXT) != 0
551		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
552			struct mbuf *n0 = n;
553			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
554			int n0len;
555
556			/*
557			 * Prepare an internal mbuf. m_pullup() doesn't
558			 * always copy the length we specified.
559			 */
560			if (maxlen >= MCLBYTES) {
561				/* Give up remote */
562				m_freem(n0);
563				break;
564			}
565			MGETHDR(n, M_DONTWAIT, n0->m_type);
566			n0len = n0->m_pkthdr.len;	/* save for use below */
567			if (n)
568				M_MOVE_PKTHDR(n, n0);
569			if (n && maxlen >= MHLEN) {
570				MCLGET(n, M_DONTWAIT);
571				if ((n->m_flags & M_EXT) == 0) {
572					m_free(n);
573					n = NULL;
574				}
575			}
576			if (n == NULL) {
577				/* Give up remote */
578				m_freem(n0);
579				break;
580			}
581			/*
582			 * Copy IPv6 and ICMPv6 only.
583			 */
584			nip6 = mtod(n, struct ip6_hdr *);
585			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
586			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
587			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
588			noff = sizeof(struct ip6_hdr);
589			/* new mbuf contains only ipv6+icmpv6 headers */
590			n->m_len = noff + sizeof(struct icmp6_hdr);
591			/*
592			 * Adjust mbuf. ip6_plen will be adjusted in
593			 * ip6_output().
594			 */
595			m_adj(n0, off + sizeof(struct icmp6_hdr));
596			/* recalculate complete packet size */
597			n->m_pkthdr.len = n0len + (noff - off);
598			n->m_next = n0;
599		} else {
600			nip6 = mtod(n, struct ip6_hdr *);
601			nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
602			noff = off;
603		}
604		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
605		nicmp6->icmp6_code = 0;
606		if (n) {
607			icmp6stat.icp6s_reflect++;
608			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
609			icmp6_reflect(n, noff);
610		}
611		break;
612
613	case ICMP6_ECHO_REPLY:
614		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
615		if (code != 0)
616			goto badcode;
617		break;
618
619	case MLD_LISTENER_QUERY:
620	case MLD_LISTENER_REPORT:
621		if (icmp6len < sizeof(struct mld_hdr))
622			goto badlen;
623		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
624			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
625		else
626			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
627		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
628			/* give up local */
629			mld6_input(m, off);
630			m = NULL;
631			goto freeit;
632		}
633		mld6_input(n, off);
634		/* m stays. */
635		break;
636
637	case MLD_LISTENER_DONE:
638		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
639		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
640			goto badlen;
641		break;		/* nothing to be done in kernel */
642
643	case MLD_MTRACE_RESP:
644	case MLD_MTRACE:
645		/* XXX: these two are experimental.  not officially defind. */
646		/* XXX: per-interface statistics? */
647		break;		/* just pass it to applications */
648
649	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
650	    {
651		enum { WRU, FQDN } mode;
652
653		if (!icmp6_nodeinfo)
654			break;
655
656		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
657			mode = WRU;
658		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
659			mode = FQDN;
660		else
661			goto badlen;
662
663#define hostnamelen	strlen(hostname)
664		if (mode == FQDN) {
665#ifndef PULLDOWN_TEST
666			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
667					 IPPROTO_DONE);
668#endif
669			n = m_copy(m, 0, M_COPYALL);
670			if (n)
671				n = ni6_input(n, off);
672			/* XXX meaningless if n == NULL */
673			noff = sizeof(struct ip6_hdr);
674		} else {
675			u_char *p;
676			int maxlen, maxhlen;
677
678			if ((icmp6_nodeinfo & 5) != 5)
679				break;
680
681			if (code != 0)
682				goto badcode;
683			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
684			if (maxlen >= MCLBYTES) {
685				/* Give up remote */
686				break;
687			}
688			MGETHDR(n, M_DONTWAIT, m->m_type);
689			if (n && maxlen > MHLEN) {
690				MCLGET(n, M_DONTWAIT);
691				if ((n->m_flags & M_EXT) == 0) {
692					m_free(n);
693					n = NULL;
694				}
695			}
696			if (!m_dup_pkthdr(n, m, M_DONTWAIT)) {
697				/*
698				 * Previous code did a blind M_COPY_PKTHDR
699				 * and said "just for rcvif".  If true, then
700				 * we could tolerate the dup failing (due to
701				 * the deep copy of the tag chain).  For now
702				 * be conservative and just fail.
703				 */
704				m_free(n);
705				n = NULL;
706			}
707			if (n == NULL) {
708				/* Give up remote */
709				break;
710			}
711			n->m_pkthdr.rcvif = NULL;
712			n->m_len = 0;
713			maxhlen = M_TRAILINGSPACE(n) - maxlen;
714			if (maxhlen > hostnamelen)
715				maxhlen = hostnamelen;
716			/*
717			 * Copy IPv6 and ICMPv6 only.
718			 */
719			nip6 = mtod(n, struct ip6_hdr *);
720			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
721			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
722			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
723			p = (u_char *)(nicmp6 + 1);
724			bzero(p, 4);
725			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
726			noff = sizeof(struct ip6_hdr);
727			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
728				sizeof(struct icmp6_hdr) + 4 + maxhlen;
729			nicmp6->icmp6_type = ICMP6_WRUREPLY;
730			nicmp6->icmp6_code = 0;
731		}
732#undef hostnamelen
733		if (n) {
734			icmp6stat.icp6s_reflect++;
735			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
736			icmp6_reflect(n, noff);
737		}
738		break;
739	    }
740
741	case ICMP6_WRUREPLY:
742		if (code != 0)
743			goto badcode;
744		break;
745
746	case ND_ROUTER_SOLICIT:
747		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
748		if (code != 0)
749			goto badcode;
750		if (icmp6len < sizeof(struct nd_router_solicit))
751			goto badlen;
752		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
753			/* give up local */
754			nd6_rs_input(m, off, icmp6len);
755			m = NULL;
756			goto freeit;
757		}
758		nd6_rs_input(n, off, icmp6len);
759		/* m stays. */
760		break;
761
762	case ND_ROUTER_ADVERT:
763		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
764		if (code != 0)
765			goto badcode;
766		if (icmp6len < sizeof(struct nd_router_advert))
767			goto badlen;
768		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
769			/* give up local */
770			nd6_ra_input(m, off, icmp6len);
771			m = NULL;
772			goto freeit;
773		}
774		nd6_ra_input(n, off, icmp6len);
775		/* m stays. */
776		break;
777
778	case ND_NEIGHBOR_SOLICIT:
779		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
780		if (code != 0)
781			goto badcode;
782		if (icmp6len < sizeof(struct nd_neighbor_solicit))
783			goto badlen;
784		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
785			/* give up local */
786			nd6_ns_input(m, off, icmp6len);
787			m = NULL;
788			goto freeit;
789		}
790		nd6_ns_input(n, off, icmp6len);
791		/* m stays. */
792		break;
793
794	case ND_NEIGHBOR_ADVERT:
795		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
796		if (code != 0)
797			goto badcode;
798		if (icmp6len < sizeof(struct nd_neighbor_advert))
799			goto badlen;
800		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
801			/* give up local */
802			nd6_na_input(m, off, icmp6len);
803			m = NULL;
804			goto freeit;
805		}
806		nd6_na_input(n, off, icmp6len);
807		/* m stays. */
808		break;
809
810	case ND_REDIRECT:
811		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
812		if (code != 0)
813			goto badcode;
814		if (icmp6len < sizeof(struct nd_redirect))
815			goto badlen;
816		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
817			/* give up local */
818			icmp6_redirect_input(m, off);
819			m = NULL;
820			goto freeit;
821		}
822		icmp6_redirect_input(n, off);
823		/* m stays. */
824		break;
825
826	case ICMP6_ROUTER_RENUMBERING:
827		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
828		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
829			goto badcode;
830		if (icmp6len < sizeof(struct icmp6_router_renum))
831			goto badlen;
832		break;
833
834	default:
835		nd6log((LOG_DEBUG,
836		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
837		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
838		    ip6_sprintf(&ip6->ip6_dst),
839		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
840		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
841			/* ICMPv6 error: MUST deliver it by spec... */
842			code = PRC_NCMDS;
843			/* deliver */
844		} else {
845			/* ICMPv6 informational: MUST not deliver */
846			break;
847		}
848	deliver:
849		if (icmp6_notify_error(m, off, icmp6len, code)) {
850			/* In this case, m should've been freed. */
851			return(IPPROTO_DONE);
852		}
853		break;
854
855	badcode:
856		icmp6stat.icp6s_badcode++;
857		break;
858
859	badlen:
860		icmp6stat.icp6s_badlen++;
861		break;
862	}
863
864	/* deliver the packet to appropriate sockets */
865	icmp6_rip6_input(&m, *offp);
866
867	return IPPROTO_DONE;
868
869 freeit:
870	m_freem(m);
871	return IPPROTO_DONE;
872}
873
874static int
875icmp6_notify_error(m, off, icmp6len, code)
876	struct mbuf *m;
877	int off, icmp6len;
878{
879	struct icmp6_hdr *icmp6;
880	struct ip6_hdr *eip6;
881	u_int32_t notifymtu;
882	struct sockaddr_in6 icmp6src, icmp6dst;
883
884	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
885		icmp6stat.icp6s_tooshort++;
886		goto freeit;
887	}
888#ifndef PULLDOWN_TEST
889	IP6_EXTHDR_CHECK(m, off,
890			 sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr),
891			 -1);
892	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
893#else
894	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
895		       sizeof(*icmp6) + sizeof(struct ip6_hdr));
896	if (icmp6 == NULL) {
897		icmp6stat.icp6s_tooshort++;
898		return(-1);
899	}
900#endif
901	eip6 = (struct ip6_hdr *)(icmp6 + 1);
902
903	/* Detect the upper level protocol */
904	{
905		void (*ctlfunc) __P((int, struct sockaddr *, void *));
906		u_int8_t nxt = eip6->ip6_nxt;
907		int eoff = off + sizeof(struct icmp6_hdr) +
908			sizeof(struct ip6_hdr);
909		struct ip6ctlparam ip6cp;
910		struct in6_addr *finaldst = NULL;
911		int icmp6type = icmp6->icmp6_type;
912		struct ip6_frag *fh;
913		struct ip6_rthdr *rth;
914		struct ip6_rthdr0 *rth0;
915		int rthlen;
916
917		while (1) { /* XXX: should avoid infinite loop explicitly? */
918			struct ip6_ext *eh;
919
920			switch (nxt) {
921			case IPPROTO_HOPOPTS:
922			case IPPROTO_DSTOPTS:
923			case IPPROTO_AH:
924#ifndef PULLDOWN_TEST
925				IP6_EXTHDR_CHECK(m, 0, eoff +
926						 sizeof(struct ip6_ext),
927						 -1);
928				eh = (struct ip6_ext *)(mtod(m, caddr_t)
929							+ eoff);
930#else
931				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
932					       eoff, sizeof(*eh));
933				if (eh == NULL) {
934					icmp6stat.icp6s_tooshort++;
935					return(-1);
936				}
937#endif
938
939				if (nxt == IPPROTO_AH)
940					eoff += (eh->ip6e_len + 2) << 2;
941				else
942					eoff += (eh->ip6e_len + 1) << 3;
943				nxt = eh->ip6e_nxt;
944				break;
945			case IPPROTO_ROUTING:
946				/*
947				 * When the erroneous packet contains a
948				 * routing header, we should examine the
949				 * header to determine the final destination.
950				 * Otherwise, we can't properly update
951				 * information that depends on the final
952				 * destination (e.g. path MTU).
953				 */
954#ifndef PULLDOWN_TEST
955				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth),
956						 -1);
957				rth = (struct ip6_rthdr *)(mtod(m, caddr_t)
958							   + eoff);
959#else
960				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
961					       eoff, sizeof(*rth));
962				if (rth == NULL) {
963					icmp6stat.icp6s_tooshort++;
964					return(-1);
965				}
966#endif
967				rthlen = (rth->ip6r_len + 1) << 3;
968				/*
969				 * XXX: currently there is no
970				 * officially defined type other
971				 * than type-0.
972				 * Note that if the segment left field
973				 * is 0, all intermediate hops must
974				 * have been passed.
975				 */
976				if (rth->ip6r_segleft &&
977				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
978					int hops;
979
980#ifndef PULLDOWN_TEST
981					IP6_EXTHDR_CHECK(m, 0, eoff + rthlen,
982							 -1);
983					rth0 = (struct ip6_rthdr0 *)(mtod(m, caddr_t) + eoff);
984#else
985					IP6_EXTHDR_GET(rth0,
986						       struct ip6_rthdr0 *, m,
987						       eoff, rthlen);
988					if (rth0 == NULL) {
989						icmp6stat.icp6s_tooshort++;
990						return(-1);
991					}
992#endif
993					/* just ignore a bogus header */
994					if ((rth0->ip6r0_len % 2) == 0 &&
995					    (hops = rth0->ip6r0_len/2))
996						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
997				}
998				eoff += rthlen;
999				nxt = rth->ip6r_nxt;
1000				break;
1001			case IPPROTO_FRAGMENT:
1002#ifndef PULLDOWN_TEST
1003				IP6_EXTHDR_CHECK(m, 0, eoff +
1004						 sizeof(struct ip6_frag),
1005						 -1);
1006				fh = (struct ip6_frag *)(mtod(m, caddr_t)
1007							 + eoff);
1008#else
1009				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1010					       eoff, sizeof(*fh));
1011				if (fh == NULL) {
1012					icmp6stat.icp6s_tooshort++;
1013					return(-1);
1014				}
1015#endif
1016				/*
1017				 * Data after a fragment header is meaningless
1018				 * unless it is the first fragment, but
1019				 * we'll go to the notify label for path MTU
1020				 * discovery.
1021				 */
1022				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1023					goto notify;
1024
1025				eoff += sizeof(struct ip6_frag);
1026				nxt = fh->ip6f_nxt;
1027				break;
1028			default:
1029				/*
1030				 * This case includes ESP and the No Next
1031				 * Header.  In such cases going to the notify
1032				 * label does not have any meaning
1033				 * (i.e. ctlfunc will be NULL), but we go
1034				 * anyway since we might have to update
1035				 * path MTU information.
1036				 */
1037				goto notify;
1038			}
1039		}
1040	  notify:
1041#ifndef PULLDOWN_TEST
1042		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1043#else
1044		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1045			       sizeof(*icmp6) + sizeof(struct ip6_hdr));
1046		if (icmp6 == NULL) {
1047			icmp6stat.icp6s_tooshort++;
1048			return(-1);
1049		}
1050#endif
1051
1052		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1053		bzero(&icmp6dst, sizeof(icmp6dst));
1054		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1055		icmp6dst.sin6_family = AF_INET6;
1056		if (finaldst == NULL)
1057			icmp6dst.sin6_addr = eip6->ip6_dst;
1058		else
1059			icmp6dst.sin6_addr = *finaldst;
1060		icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1061							  &icmp6dst.sin6_addr);
1062#ifndef SCOPEDROUTING
1063		if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1064				   NULL, NULL)) {
1065			/* should be impossbile */
1066			nd6log((LOG_DEBUG,
1067			    "icmp6_notify_error: in6_embedscope failed\n"));
1068			goto freeit;
1069		}
1070#endif
1071
1072		/*
1073		 * retrieve parameters from the inner IPv6 header, and convert
1074		 * them into sockaddr structures.
1075		 */
1076		bzero(&icmp6src, sizeof(icmp6src));
1077		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1078		icmp6src.sin6_family = AF_INET6;
1079		icmp6src.sin6_addr = eip6->ip6_src;
1080		icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1081							  &icmp6src.sin6_addr);
1082#ifndef SCOPEDROUTING
1083		if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1084				   NULL, NULL)) {
1085			/* should be impossbile */
1086			nd6log((LOG_DEBUG,
1087			    "icmp6_notify_error: in6_embedscope failed\n"));
1088			goto freeit;
1089		}
1090#endif
1091		icmp6src.sin6_flowinfo =
1092			(eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1093
1094		if (finaldst == NULL)
1095			finaldst = &eip6->ip6_dst;
1096		ip6cp.ip6c_m = m;
1097		ip6cp.ip6c_icmp6 = icmp6;
1098		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1099		ip6cp.ip6c_off = eoff;
1100		ip6cp.ip6c_finaldst = finaldst;
1101		ip6cp.ip6c_src = &icmp6src;
1102		ip6cp.ip6c_nxt = nxt;
1103
1104		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1105			notifymtu = ntohl(icmp6->icmp6_mtu);
1106			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1107			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1108		}
1109
1110		ctlfunc = (void (*) __P((int, struct sockaddr *, void *)))
1111			(inet6sw[ip6_protox[nxt]].pr_ctlinput);
1112		if (ctlfunc) {
1113			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1114					  &ip6cp);
1115		}
1116	}
1117	return(0);
1118
1119  freeit:
1120	m_freem(m);
1121	return(-1);
1122}
1123
1124void
1125icmp6_mtudisc_update(ip6cp, validated)
1126	struct ip6ctlparam *ip6cp;
1127	int validated;
1128{
1129	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1130	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1131	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1132	u_int mtu = ntohl(icmp6->icmp6_mtu);
1133	struct rtentry *rt = NULL;
1134	struct sockaddr_in6 sin6;
1135
1136	if (!validated)
1137		return;
1138
1139	bzero(&sin6, sizeof(sin6));
1140	sin6.sin6_family = PF_INET6;
1141	sin6.sin6_len = sizeof(struct sockaddr_in6);
1142	sin6.sin6_addr = *dst;
1143	/* XXX normally, this won't happen */
1144	if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1145		sin6.sin6_addr.s6_addr16[1] =
1146		    htons(m->m_pkthdr.rcvif->if_index);
1147	}
1148	/* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1149	rt = rtalloc1((struct sockaddr *)&sin6, 0,
1150		      RTF_CLONING | RTF_PRCLONING);
1151
1152	if (rt && (rt->rt_flags & RTF_HOST)
1153	    && !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1154		if (mtu < IPV6_MMTU) {
1155				/* xxx */
1156			rt->rt_rmx.rmx_locks |= RTV_MTU;
1157		} else if (mtu < rt->rt_ifp->if_mtu &&
1158			   rt->rt_rmx.rmx_mtu > mtu) {
1159			icmp6stat.icp6s_pmtuchg++;
1160			rt->rt_rmx.rmx_mtu = mtu;
1161		}
1162	}
1163	if (rt) { /* XXX: need braces to avoid conflict with else in RTFREE. */
1164		RTFREE(rt);
1165	}
1166}
1167
1168/*
1169 * Process a Node Information Query packet, based on
1170 * draft-ietf-ipngwg-icmp-name-lookups-07.
1171 *
1172 * Spec incompatibilities:
1173 * - IPv6 Subject address handling
1174 * - IPv4 Subject address handling support missing
1175 * - Proxy reply (answer even if it's not for me)
1176 * - joins NI group address at in6_ifattach() time only, does not cope
1177 *   with hostname changes by sethostname(3)
1178 */
1179#define hostnamelen	strlen(hostname)
1180static struct mbuf *
1181ni6_input(m, off)
1182	struct mbuf *m;
1183	int off;
1184{
1185	struct icmp6_nodeinfo *ni6, *nni6;
1186	struct mbuf *n = NULL;
1187	u_int16_t qtype;
1188	int subjlen;
1189	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1190	struct ni_reply_fqdn *fqdn;
1191	int addrs;		/* for NI_QTYPE_NODEADDR */
1192	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1193	struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1194	struct sockaddr_in6 sin6_d; /* XXX: we should retrieve this from m_aux */
1195	struct ip6_hdr *ip6;
1196	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1197	char *subj = NULL;
1198	struct in6_ifaddr *ia6 = NULL;
1199
1200	ip6 = mtod(m, struct ip6_hdr *);
1201#ifndef PULLDOWN_TEST
1202	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1203#else
1204	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1205	if (ni6 == NULL) {
1206		/* m is already reclaimed */
1207		return NULL;
1208	}
1209#endif
1210
1211	/*
1212	 * Validate IPv6 destination address.
1213	 *
1214	 * The Responder must discard the Query without further processing
1215	 * unless it is one of the Responder's unicast or anycast addresses, or
1216	 * a link-local scope multicast address which the Responder has joined.
1217	 * [icmp-name-lookups-07, Section 4.]
1218	 */
1219	bzero(&sin6, sizeof(sin6));
1220	sin6.sin6_family = AF_INET6;
1221	sin6.sin6_len = sizeof(struct sockaddr_in6);
1222	bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1223	/* XXX scopeid */
1224	if ((ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)&sin6)) != NULL) {
1225		/* unicast/anycast, fine */
1226		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1227		    (icmp6_nodeinfo & 4) == 0) {
1228			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1229				"a temporary address in %s:%d",
1230			       __FILE__, __LINE__));
1231			goto bad;
1232		}
1233	} else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1234		; /* link-local multicast, fine */
1235	else
1236		goto bad;
1237
1238	/* validate query Subject field. */
1239	qtype = ntohs(ni6->ni_qtype);
1240	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1241	switch (qtype) {
1242	case NI_QTYPE_NOOP:
1243	case NI_QTYPE_SUPTYPES:
1244		/* 07 draft */
1245		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1246			break;
1247		/* FALLTHROUGH */
1248	case NI_QTYPE_FQDN:
1249	case NI_QTYPE_NODEADDR:
1250		switch (ni6->ni_code) {
1251		case ICMP6_NI_SUBJ_IPV6:
1252#if ICMP6_NI_SUBJ_IPV6 != 0
1253		case 0:
1254#endif
1255			/*
1256			 * backward compatibility - try to accept 03 draft
1257			 * format, where no Subject is present.
1258			 */
1259			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1260			    subjlen == 0) {
1261				oldfqdn++;
1262				break;
1263			}
1264#if ICMP6_NI_SUBJ_IPV6 != 0
1265			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1266				goto bad;
1267#endif
1268
1269			if (subjlen != sizeof(sin6.sin6_addr))
1270				goto bad;
1271
1272			/*
1273			 * Validate Subject address.
1274			 *
1275			 * Not sure what exactly "address belongs to the node"
1276			 * means in the spec, is it just unicast, or what?
1277			 *
1278			 * At this moment we consider Subject address as
1279			 * "belong to the node" if the Subject address equals
1280			 * to the IPv6 destination address; validation for
1281			 * IPv6 destination address should have done enough
1282			 * check for us.
1283			 *
1284			 * We do not do proxy at this moment.
1285			 */
1286			/* m_pulldown instead of copy? */
1287			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1288			    subjlen, (caddr_t)&sin6.sin6_addr);
1289			sin6.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1290							      &sin6.sin6_addr);
1291#ifndef SCOPEDROUTING
1292			in6_embedscope(&sin6.sin6_addr, &sin6, NULL, NULL);
1293#endif
1294			bzero(&sin6_d, sizeof(sin6_d));
1295			sin6_d.sin6_family = AF_INET6; /* not used, actually */
1296			sin6_d.sin6_len = sizeof(sin6_d); /* ditto */
1297			sin6_d.sin6_addr = ip6->ip6_dst;
1298			sin6_d.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1299								&ip6->ip6_dst);
1300#ifndef SCOPEDROUTING
1301			in6_embedscope(&sin6_d.sin6_addr, &sin6_d, NULL, NULL);
1302#endif
1303			subj = (char *)&sin6;
1304			if (SA6_ARE_ADDR_EQUAL(&sin6, &sin6_d))
1305				break;
1306
1307			/*
1308			 * XXX if we are to allow other cases, we should really
1309			 * be careful about scope here.
1310			 * basically, we should disallow queries toward IPv6
1311			 * destination X with subject Y, if scope(X) > scope(Y).
1312			 * if we allow scope(X) > scope(Y), it will result in
1313			 * information leakage across scope boundary.
1314			 */
1315			goto bad;
1316
1317		case ICMP6_NI_SUBJ_FQDN:
1318			/*
1319			 * Validate Subject name with gethostname(3).
1320			 *
1321			 * The behavior may need some debate, since:
1322			 * - we are not sure if the node has FQDN as
1323			 *   hostname (returned by gethostname(3)).
1324			 * - the code does wildcard match for truncated names.
1325			 *   however, we are not sure if we want to perform
1326			 *   wildcard match, if gethostname(3) side has
1327			 *   truncated hostname.
1328			 */
1329			n = ni6_nametodns(hostname, hostnamelen, 0);
1330			if (!n || n->m_next || n->m_len == 0)
1331				goto bad;
1332			IP6_EXTHDR_GET(subj, char *, m,
1333			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1334			if (subj == NULL)
1335				goto bad;
1336			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1337					n->m_len)) {
1338				goto bad;
1339			}
1340			m_freem(n);
1341			n = NULL;
1342			break;
1343
1344		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1345		default:
1346			goto bad;
1347		}
1348		break;
1349	}
1350
1351	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1352	switch (qtype) {
1353	case NI_QTYPE_FQDN:
1354		if ((icmp6_nodeinfo & 1) == 0)
1355			goto bad;
1356		break;
1357	case NI_QTYPE_NODEADDR:
1358		if ((icmp6_nodeinfo & 2) == 0)
1359			goto bad;
1360		break;
1361	}
1362
1363	/* guess reply length */
1364	switch (qtype) {
1365	case NI_QTYPE_NOOP:
1366		break;		/* no reply data */
1367	case NI_QTYPE_SUPTYPES:
1368		replylen += sizeof(u_int32_t);
1369		break;
1370	case NI_QTYPE_FQDN:
1371		/* XXX will append an mbuf */
1372		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1373		break;
1374	case NI_QTYPE_NODEADDR:
1375		addrs = ni6_addrs(ni6, m, &ifp, subj);
1376		if ((replylen += addrs * (sizeof(struct in6_addr) +
1377					  sizeof(u_int32_t))) > MCLBYTES)
1378			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1379		break;
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		m_freem(m);
1401		return(NULL);
1402	}
1403	M_MOVE_PKTHDR(n, m); /* just for recvif */
1404	if (replylen > MHLEN) {
1405		if (replylen > MCLBYTES) {
1406			/*
1407			 * XXX: should we try to allocate more? But MCLBYTES
1408			 * is probably much larger than IPV6_MMTU...
1409			 */
1410			goto bad;
1411		}
1412		MCLGET(n, M_DONTWAIT);
1413		if ((n->m_flags & M_EXT) == 0) {
1414			goto bad;
1415		}
1416	}
1417	n->m_pkthdr.len = n->m_len = replylen;
1418
1419	/* copy mbuf header and IPv6 + Node Information base headers */
1420	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1421	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1422	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1423
1424	/* qtype dependent procedure */
1425	switch (qtype) {
1426	case NI_QTYPE_NOOP:
1427		nni6->ni_code = ICMP6_NI_SUCCESS;
1428		nni6->ni_flags = 0;
1429		break;
1430	case NI_QTYPE_SUPTYPES:
1431	{
1432		u_int32_t v;
1433		nni6->ni_code = ICMP6_NI_SUCCESS;
1434		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1435		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1436		v = (u_int32_t)htonl(0x0000000f);
1437		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1438		break;
1439	}
1440	case NI_QTYPE_FQDN:
1441		nni6->ni_code = ICMP6_NI_SUCCESS;
1442		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1443						sizeof(struct ip6_hdr) +
1444						sizeof(struct icmp6_nodeinfo));
1445		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1446		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1447		/*
1448		 * XXX do we really have FQDN in variable "hostname"?
1449		 */
1450		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1451		if (n->m_next == NULL)
1452			goto bad;
1453		/* XXX we assume that n->m_next is not a chain */
1454		if (n->m_next->m_next != NULL)
1455			goto bad;
1456		n->m_pkthdr.len += n->m_next->m_len;
1457		break;
1458	case NI_QTYPE_NODEADDR:
1459	{
1460		int lenlim, copied;
1461
1462		nni6->ni_code = ICMP6_NI_SUCCESS;
1463		n->m_pkthdr.len = n->m_len =
1464		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1465		lenlim = M_TRAILINGSPACE(n);
1466		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1467		/* XXX: reset mbuf length */
1468		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1469			sizeof(struct icmp6_nodeinfo) + copied;
1470		break;
1471	}
1472	default:
1473		break;		/* XXX impossible! */
1474	}
1475
1476	nni6->ni_type = ICMP6_NI_REPLY;
1477	m_freem(m);
1478	return(n);
1479
1480  bad:
1481	m_freem(m);
1482	if (n)
1483		m_freem(n);
1484	return(NULL);
1485}
1486#undef hostnamelen
1487
1488/*
1489 * make a mbuf with DNS-encoded string.  no compression support.
1490 *
1491 * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1492 * treated as truncated name (two \0 at the end).  this is a wild guess.
1493 */
1494static struct mbuf *
1495ni6_nametodns(name, namelen, old)
1496	const char *name;
1497	int namelen;
1498	int old;	/* return pascal string if non-zero */
1499{
1500	struct mbuf *m;
1501	char *cp, *ep;
1502	const char *p, *q;
1503	int i, len, nterm;
1504
1505	if (old)
1506		len = namelen + 1;
1507	else
1508		len = MCLBYTES;
1509
1510	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1511	MGET(m, M_DONTWAIT, MT_DATA);
1512	if (m && len > MLEN) {
1513		MCLGET(m, M_DONTWAIT);
1514		if ((m->m_flags & M_EXT) == 0)
1515			goto fail;
1516	}
1517	if (!m)
1518		goto fail;
1519	m->m_next = NULL;
1520
1521	if (old) {
1522		m->m_len = len;
1523		*mtod(m, char *) = namelen;
1524		bcopy(name, mtod(m, char *) + 1, namelen);
1525		return m;
1526	} else {
1527		m->m_len = 0;
1528		cp = mtod(m, char *);
1529		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1530
1531		/* if not certain about my name, return empty buffer */
1532		if (namelen == 0)
1533			return m;
1534
1535		/*
1536		 * guess if it looks like shortened hostname, or FQDN.
1537		 * shortened hostname needs two trailing "\0".
1538		 */
1539		i = 0;
1540		for (p = name; p < name + namelen; p++) {
1541			if (*p && *p == '.')
1542				i++;
1543		}
1544		if (i < 2)
1545			nterm = 2;
1546		else
1547			nterm = 1;
1548
1549		p = name;
1550		while (cp < ep && p < name + namelen) {
1551			i = 0;
1552			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1553				i++;
1554			/* result does not fit into mbuf */
1555			if (cp + i + 1 >= ep)
1556				goto fail;
1557			/*
1558			 * DNS label length restriction, RFC1035 page 8.
1559			 * "i == 0" case is included here to avoid returning
1560			 * 0-length label on "foo..bar".
1561			 */
1562			if (i <= 0 || i >= 64)
1563				goto fail;
1564			*cp++ = i;
1565			bcopy(p, cp, i);
1566			cp += i;
1567			p = q;
1568			if (p < name + namelen && *p == '.')
1569				p++;
1570		}
1571		/* termination */
1572		if (cp + nterm >= ep)
1573			goto fail;
1574		while (nterm-- > 0)
1575			*cp++ = '\0';
1576		m->m_len = cp - mtod(m, char *);
1577		return m;
1578	}
1579
1580	panic("should not reach here");
1581	/* NOTREACHED */
1582
1583 fail:
1584	if (m)
1585		m_freem(m);
1586	return NULL;
1587}
1588
1589/*
1590 * check if two DNS-encoded string matches.  takes care of truncated
1591 * form (with \0\0 at the end).  no compression support.
1592 * XXX upper/lowercase match (see RFC2065)
1593 */
1594static int
1595ni6_dnsmatch(a, alen, b, blen)
1596	const char *a;
1597	int alen;
1598	const char *b;
1599	int blen;
1600{
1601	const char *a0, *b0;
1602	int l;
1603
1604	/* simplest case - need validation? */
1605	if (alen == blen && bcmp(a, b, alen) == 0)
1606		return 1;
1607
1608	a0 = a;
1609	b0 = b;
1610
1611	/* termination is mandatory */
1612	if (alen < 2 || blen < 2)
1613		return 0;
1614	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1615		return 0;
1616	alen--;
1617	blen--;
1618
1619	while (a - a0 < alen && b - b0 < blen) {
1620		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1621			return 0;
1622
1623		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1624			return 0;
1625		/* we don't support compression yet */
1626		if (a[0] >= 64 || b[0] >= 64)
1627			return 0;
1628
1629		/* truncated case */
1630		if (a[0] == 0 && a - a0 == alen - 1)
1631			return 1;
1632		if (b[0] == 0 && b - b0 == blen - 1)
1633			return 1;
1634		if (a[0] == 0 || b[0] == 0)
1635			return 0;
1636
1637		if (a[0] != b[0])
1638			return 0;
1639		l = a[0];
1640		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1641			return 0;
1642		if (bcmp(a + 1, b + 1, l) != 0)
1643			return 0;
1644
1645		a += 1 + l;
1646		b += 1 + l;
1647	}
1648
1649	if (a - a0 == alen && b - b0 == blen)
1650		return 1;
1651	else
1652		return 0;
1653}
1654
1655/*
1656 * calculate the number of addresses to be returned in the node info reply.
1657 */
1658static int
1659ni6_addrs(ni6, m, ifpp, subj)
1660	struct icmp6_nodeinfo *ni6;
1661	struct mbuf *m;
1662	struct ifnet **ifpp;
1663	char *subj;
1664{
1665	struct ifnet *ifp;
1666	struct in6_ifaddr *ifa6;
1667	struct ifaddr *ifa;
1668	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1669	int addrs = 0, addrsofif, iffound = 0;
1670	int niflags = ni6->ni_flags;
1671
1672	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1673		switch (ni6->ni_code) {
1674		case ICMP6_NI_SUBJ_IPV6:
1675			if (subj == NULL) /* must be impossible... */
1676				return(0);
1677			subj_ip6 = (struct sockaddr_in6 *)subj;
1678			break;
1679		default:
1680			/*
1681			 * XXX: we only support IPv6 subject address for
1682			 * this Qtype.
1683			 */
1684			return(0);
1685		}
1686	}
1687
1688	IFNET_RLOCK();
1689	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
1690	{
1691		addrsofif = 0;
1692		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list)
1693		{
1694			if (ifa->ifa_addr->sa_family != AF_INET6)
1695				continue;
1696			ifa6 = (struct in6_ifaddr *)ifa;
1697
1698			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1699			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1700					       &ifa6->ia_addr.sin6_addr))
1701				iffound = 1;
1702
1703			/*
1704			 * IPv4-mapped addresses can only be returned by a
1705			 * Node Information proxy, since they represent
1706			 * addresses of IPv4-only nodes, which perforce do
1707			 * not implement this protocol.
1708			 * [icmp-name-lookups-07, Section 5.4]
1709			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1710			 * this function at this moment.
1711			 */
1712
1713			/* What do we have to do about ::1? */
1714			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1715			case IPV6_ADDR_SCOPE_LINKLOCAL:
1716				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1717					continue;
1718				break;
1719			case IPV6_ADDR_SCOPE_SITELOCAL:
1720				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1721					continue;
1722				break;
1723			case IPV6_ADDR_SCOPE_GLOBAL:
1724				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1725					continue;
1726				break;
1727			default:
1728				continue;
1729			}
1730
1731			/*
1732			 * check if anycast is okay.
1733			 * XXX: just experimental.  not in the spec.
1734			 */
1735			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1736			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1737				continue; /* we need only unicast addresses */
1738			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1739			    (icmp6_nodeinfo & 4) == 0) {
1740				continue;
1741			}
1742			addrsofif++; /* count the address */
1743		}
1744		if (iffound) {
1745			*ifpp = ifp;
1746			IFNET_RUNLOCK();
1747			return(addrsofif);
1748		}
1749
1750		addrs += addrsofif;
1751	}
1752	IFNET_RUNLOCK();
1753
1754	return(addrs);
1755}
1756
1757static int
1758ni6_store_addrs(ni6, nni6, ifp0, resid)
1759	struct icmp6_nodeinfo *ni6, *nni6;
1760	struct ifnet *ifp0;
1761	int resid;
1762{
1763	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1764	struct in6_ifaddr *ifa6;
1765	struct ifaddr *ifa;
1766	struct ifnet *ifp_dep = NULL;
1767	int copied = 0, allow_deprecated = 0;
1768	u_char *cp = (u_char *)(nni6 + 1);
1769	int niflags = ni6->ni_flags;
1770	u_int32_t ltime;
1771
1772	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1773		return(0);	/* needless to copy */
1774
1775	IFNET_RLOCK();
1776  again:
1777	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list))
1778	{
1779		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1780		     ifa = ifa->ifa_list.tqe_next)
1781		{
1782			if (ifa->ifa_addr->sa_family != AF_INET6)
1783				continue;
1784			ifa6 = (struct in6_ifaddr *)ifa;
1785
1786			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1787			    allow_deprecated == 0) {
1788				/*
1789				 * prefererred address should be put before
1790				 * deprecated addresses.
1791				 */
1792
1793				/* record the interface for later search */
1794				if (ifp_dep == NULL)
1795					ifp_dep = ifp;
1796
1797				continue;
1798			}
1799			else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1800				 allow_deprecated != 0)
1801				continue; /* we now collect deprecated addrs */
1802
1803			/* What do we have to do about ::1? */
1804			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1805			case IPV6_ADDR_SCOPE_LINKLOCAL:
1806				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1807					continue;
1808				break;
1809			case IPV6_ADDR_SCOPE_SITELOCAL:
1810				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1811					continue;
1812				break;
1813			case IPV6_ADDR_SCOPE_GLOBAL:
1814				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1815					continue;
1816				break;
1817			default:
1818				continue;
1819			}
1820
1821			/*
1822			 * check if anycast is okay.
1823			 * XXX: just experimental. not in the spec.
1824			 */
1825			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1826			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1827				continue;
1828			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1829			    (icmp6_nodeinfo & 4) == 0) {
1830				continue;
1831			}
1832
1833			/* now we can copy the address */
1834			if (resid < sizeof(struct in6_addr) +
1835			    sizeof(u_int32_t)) {
1836				/*
1837				 * We give up much more copy.
1838				 * Set the truncate flag and return.
1839				 */
1840				nni6->ni_flags |=
1841					NI_NODEADDR_FLAG_TRUNCATE;
1842				IFNET_RUNLOCK();
1843				return(copied);
1844			}
1845
1846			/*
1847			 * Set the TTL of the address.
1848			 * The TTL value should be one of the following
1849			 * according to the specification:
1850			 *
1851			 * 1. The remaining lifetime of a DHCP lease on the
1852			 *    address, or
1853			 * 2. The remaining Valid Lifetime of a prefix from
1854			 *    which the address was derived through Stateless
1855			 *    Autoconfiguration.
1856			 *
1857			 * Note that we currently do not support stateful
1858			 * address configuration by DHCPv6, so the former
1859			 * case can't happen.
1860			 */
1861			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1862				ltime = ND6_INFINITE_LIFETIME;
1863			else {
1864				if (ifa6->ia6_lifetime.ia6t_expire >
1865				    time_second)
1866					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1867				else
1868					ltime = 0;
1869			}
1870
1871			bcopy(&ltime, cp, sizeof(u_int32_t));
1872			cp += sizeof(u_int32_t);
1873
1874			/* copy the address itself */
1875			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1876			      sizeof(struct in6_addr));
1877			/* XXX: KAME link-local hack; remove ifindex */
1878			if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr))
1879				((struct in6_addr *)cp)->s6_addr16[1] = 0;
1880			cp += sizeof(struct in6_addr);
1881
1882			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1883			copied += (sizeof(struct in6_addr) +
1884				   sizeof(u_int32_t));
1885		}
1886		if (ifp0)	/* we need search only on the specified IF */
1887			break;
1888	}
1889
1890	if (allow_deprecated == 0 && ifp_dep != NULL) {
1891		ifp = ifp_dep;
1892		allow_deprecated = 1;
1893
1894		goto again;
1895	}
1896
1897	IFNET_RUNLOCK();
1898
1899	return(copied);
1900}
1901
1902/*
1903 * XXX almost dup'ed code with rip6_input.
1904 */
1905static int
1906icmp6_rip6_input(mp, off)
1907	struct	mbuf **mp;
1908	int	off;
1909{
1910	struct mbuf *m = *mp;
1911	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1912	struct in6pcb *in6p;
1913	struct in6pcb *last = NULL;
1914	struct sockaddr_in6 rip6src;
1915	struct icmp6_hdr *icmp6;
1916	struct mbuf *opts = NULL;
1917
1918#ifndef PULLDOWN_TEST
1919	/* this is assumed to be safe. */
1920	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1921#else
1922	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1923	if (icmp6 == NULL) {
1924		/* m is already reclaimed */
1925		return IPPROTO_DONE;
1926	}
1927#endif
1928
1929	bzero(&rip6src, sizeof(rip6src));
1930	rip6src.sin6_len = sizeof(struct sockaddr_in6);
1931	rip6src.sin6_family = AF_INET6;
1932	/* KAME hack: recover scopeid */
1933	(void)in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif);
1934
1935	LIST_FOREACH(in6p, &ripcb, inp_list)
1936	{
1937		if ((in6p->inp_vflag & INP_IPV6) == 0)
1938			continue;
1939#ifdef HAVE_NRL_INPCB
1940		if (!(in6p->in6p_flags & INP_IPV6))
1941			continue;
1942#endif
1943		if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1944			continue;
1945		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1946		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1947			continue;
1948		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1949		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1950			continue;
1951		if (in6p->in6p_icmp6filt
1952		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1953				 in6p->in6p_icmp6filt))
1954			continue;
1955		if (last) {
1956			struct	mbuf *n;
1957			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1958				if (last->in6p_flags & IN6P_CONTROLOPTS)
1959					ip6_savecontrol(last, &opts, ip6, n);
1960				/* strip intermediate headers */
1961				m_adj(n, off);
1962				if (sbappendaddr(&last->in6p_socket->so_rcv,
1963						 (struct sockaddr *)&rip6src,
1964						 n, opts) == 0) {
1965					/* should notify about lost packet */
1966					m_freem(n);
1967					if (opts) {
1968						m_freem(opts);
1969					}
1970				} else
1971					sorwakeup(last->in6p_socket);
1972				opts = NULL;
1973			}
1974		}
1975		last = in6p;
1976	}
1977	if (last) {
1978		if (last->in6p_flags & IN6P_CONTROLOPTS)
1979			ip6_savecontrol(last, &opts, ip6, m);
1980		/* strip intermediate headers */
1981		m_adj(m, off);
1982		if (sbappendaddr(&last->in6p_socket->so_rcv,
1983				 (struct sockaddr *)&rip6src, m, opts) == 0) {
1984			m_freem(m);
1985			if (opts)
1986				m_freem(opts);
1987		} else
1988			sorwakeup(last->in6p_socket);
1989	} else {
1990		m_freem(m);
1991		ip6stat.ip6s_delivered--;
1992	}
1993	return IPPROTO_DONE;
1994}
1995
1996/*
1997 * Reflect the ip6 packet back to the source.
1998 * OFF points to the icmp6 header, counted from the top of the mbuf.
1999 */
2000void
2001icmp6_reflect(m, off)
2002	struct	mbuf *m;
2003	size_t off;
2004{
2005	struct ip6_hdr *ip6;
2006	struct icmp6_hdr *icmp6;
2007	struct in6_ifaddr *ia;
2008	struct in6_addr t, *src = 0;
2009	int plen;
2010	int type, code;
2011	struct ifnet *outif = NULL;
2012	struct sockaddr_in6 sa6_src, sa6_dst;
2013#ifdef COMPAT_RFC1885
2014	int mtu = IPV6_MMTU;
2015	struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
2016#endif
2017
2018	/* too short to reflect */
2019	if (off < sizeof(struct ip6_hdr)) {
2020		nd6log((LOG_DEBUG,
2021		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2022		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2023		    __FILE__, __LINE__));
2024		goto bad;
2025	}
2026
2027	/*
2028	 * If there are extra headers between IPv6 and ICMPv6, strip
2029	 * off that header first.
2030	 */
2031#ifdef DIAGNOSTIC
2032	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2033		panic("assumption failed in icmp6_reflect");
2034#endif
2035	if (off > sizeof(struct ip6_hdr)) {
2036		size_t l;
2037		struct ip6_hdr nip6;
2038
2039		l = off - sizeof(struct ip6_hdr);
2040		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2041		m_adj(m, l);
2042		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2043		if (m->m_len < l) {
2044			if ((m = m_pullup(m, l)) == NULL)
2045				return;
2046		}
2047		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2048	} else /* off == sizeof(struct ip6_hdr) */ {
2049		size_t l;
2050		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2051		if (m->m_len < l) {
2052			if ((m = m_pullup(m, l)) == NULL)
2053				return;
2054		}
2055	}
2056	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2057	ip6 = mtod(m, struct ip6_hdr *);
2058	ip6->ip6_nxt = IPPROTO_ICMPV6;
2059	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2060	type = icmp6->icmp6_type; /* keep type for statistics */
2061	code = icmp6->icmp6_code; /* ditto. */
2062
2063	t = ip6->ip6_dst;
2064	/*
2065	 * ip6_input() drops a packet if its src is multicast.
2066	 * So, the src is never multicast.
2067	 */
2068	ip6->ip6_dst = ip6->ip6_src;
2069
2070	/*
2071	 * XXX: make sure to embed scope zone information, using
2072	 * already embedded IDs or the received interface (if any).
2073	 * Note that rcvif may be NULL.
2074	 * TODO: scoped routing case (XXX).
2075	 */
2076	bzero(&sa6_src, sizeof(sa6_src));
2077	sa6_src.sin6_family = AF_INET6;
2078	sa6_src.sin6_len = sizeof(sa6_src);
2079	sa6_src.sin6_addr = ip6->ip6_dst;
2080	in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
2081	in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL);
2082	bzero(&sa6_dst, sizeof(sa6_dst));
2083	sa6_dst.sin6_family = AF_INET6;
2084	sa6_dst.sin6_len = sizeof(sa6_dst);
2085	sa6_dst.sin6_addr = t;
2086	in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
2087	in6_embedscope(&t, &sa6_dst, NULL, NULL);
2088
2089#ifdef COMPAT_RFC1885
2090	/*
2091	 * xxx guess MTU
2092	 * RFC 1885 requires that echo reply should be truncated if it
2093	 * does not fit in with (return) path MTU, but the description was
2094	 * removed in the new spec.
2095	 */
2096	if (icmp6_reflect_rt.ro_rt == 0 ||
2097	    ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
2098		if (icmp6_reflect_rt.ro_rt) {
2099			RTFREE(icmp6_reflect_rt.ro_rt);
2100			icmp6_reflect_rt.ro_rt = 0;
2101		}
2102		bzero(sin6, sizeof(*sin6));
2103		sin6->sin6_family = PF_INET6;
2104		sin6->sin6_len = sizeof(struct sockaddr_in6);
2105		sin6->sin6_addr = ip6->ip6_dst;
2106
2107		rtalloc_ign((struct route *)&icmp6_reflect_rt.ro_rt,
2108			    RTF_PRCLONING);
2109	}
2110
2111	if (icmp6_reflect_rt.ro_rt == 0)
2112		goto bad;
2113
2114	if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
2115	    && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
2116		mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
2117
2118	if (mtu < m->m_pkthdr.len) {
2119		plen -= (m->m_pkthdr.len - mtu);
2120		m_adj(m, mtu - m->m_pkthdr.len);
2121	}
2122#endif
2123	/*
2124	 * If the incoming packet was addressed directly to us(i.e. unicast),
2125	 * use dst as the src for the reply.
2126	 * The IN6_IFF_NOTREADY case would be VERY rare, but is possible
2127	 * (for example) when we encounter an error while forwarding procedure
2128	 * destined to a duplicated address of ours.
2129	 */
2130	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2131		if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2132		    (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2133			src = &t;
2134			break;
2135		}
2136	if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2137		/*
2138		 * This is the case if the dst is our link-local address
2139		 * and the sender is also ourselves.
2140		 */
2141		src = &t;
2142	}
2143
2144	if (src == 0) {
2145		int e;
2146		struct route_in6 ro;
2147
2148		/*
2149		 * This case matches to multicasts, our anycast, or unicasts
2150		 * that we do not own.  Select a source address based on the
2151		 * source address of the erroneous packet.
2152		 */
2153		bzero(&ro, sizeof(ro));
2154		src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e);
2155		if (ro.ro_rt)
2156			RTFREE(ro.ro_rt); /* XXX: we could use this */
2157		if (src == NULL) {
2158			nd6log((LOG_DEBUG,
2159			    "icmp6_reflect: source can't be determined: "
2160			    "dst=%s, error=%d\n",
2161			    ip6_sprintf(&sa6_src.sin6_addr), e));
2162			goto bad;
2163		}
2164	}
2165
2166	ip6->ip6_src = *src;
2167
2168	ip6->ip6_flow = 0;
2169	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2170	ip6->ip6_vfc |= IPV6_VERSION;
2171	ip6->ip6_nxt = IPPROTO_ICMPV6;
2172	if (m->m_pkthdr.rcvif) {
2173		/* XXX: This may not be the outgoing interface */
2174		ip6->ip6_hlim = nd_ifinfo[m->m_pkthdr.rcvif->if_index].chlim;
2175	} else
2176		ip6->ip6_hlim = ip6_defhlim;
2177
2178	icmp6->icmp6_cksum = 0;
2179	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2180					sizeof(struct ip6_hdr), plen);
2181
2182	/*
2183	 * XXX option handling
2184	 */
2185
2186	m->m_flags &= ~(M_BCAST|M_MCAST);
2187
2188#ifdef COMPAT_RFC1885
2189	ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif, NULL);
2190#else
2191	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2192#endif
2193	if (outif)
2194		icmp6_ifoutstat_inc(outif, type, code);
2195
2196	return;
2197
2198 bad:
2199	m_freem(m);
2200	return;
2201}
2202
2203void
2204icmp6_fasttimo()
2205{
2206
2207	mld6_fasttimeo();
2208}
2209
2210static const char *
2211icmp6_redirect_diag(src6, dst6, tgt6)
2212	struct in6_addr *src6;
2213	struct in6_addr *dst6;
2214	struct in6_addr *tgt6;
2215{
2216	static char buf[1024];
2217	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2218		ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2219	return buf;
2220}
2221
2222void
2223icmp6_redirect_input(m, off)
2224	struct mbuf *m;
2225	int off;
2226{
2227	struct ifnet *ifp = m->m_pkthdr.rcvif;
2228	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2229	struct nd_redirect *nd_rd;
2230	int icmp6len = ntohs(ip6->ip6_plen);
2231	char *lladdr = NULL;
2232	int lladdrlen = 0;
2233	u_char *redirhdr = NULL;
2234	int redirhdrlen = 0;
2235	struct rtentry *rt = NULL;
2236	int is_router;
2237	int is_onlink;
2238	struct in6_addr src6 = ip6->ip6_src;
2239	struct in6_addr redtgt6;
2240	struct in6_addr reddst6;
2241	union nd_opts ndopts;
2242
2243	if (!m || !ifp)
2244		return;
2245
2246	/* XXX if we are router, we don't update route by icmp6 redirect */
2247	if (ip6_forwarding)
2248		goto freeit;
2249	if (!icmp6_rediraccept)
2250		goto freeit;
2251
2252#ifndef PULLDOWN_TEST
2253	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2254	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2255#else
2256	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2257	if (nd_rd == NULL) {
2258		icmp6stat.icp6s_tooshort++;
2259		return;
2260	}
2261#endif
2262	redtgt6 = nd_rd->nd_rd_target;
2263	reddst6 = nd_rd->nd_rd_dst;
2264
2265	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2266		redtgt6.s6_addr16[1] = htons(ifp->if_index);
2267	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2268		reddst6.s6_addr16[1] = htons(ifp->if_index);
2269
2270	/* validation */
2271	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2272		nd6log((LOG_ERR,
2273			"ICMP6 redirect sent from %s rejected; "
2274			"must be from linklocal\n", ip6_sprintf(&src6)));
2275		goto bad;
2276	}
2277	if (ip6->ip6_hlim != 255) {
2278		nd6log((LOG_ERR,
2279			"ICMP6 redirect sent from %s rejected; "
2280			"hlim=%d (must be 255)\n",
2281			ip6_sprintf(&src6), ip6->ip6_hlim));
2282		goto bad;
2283	}
2284    {
2285	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2286	struct sockaddr_in6 sin6;
2287	struct in6_addr *gw6;
2288
2289	bzero(&sin6, sizeof(sin6));
2290	sin6.sin6_family = AF_INET6;
2291	sin6.sin6_len = sizeof(struct sockaddr_in6);
2292	bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2293	rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL);
2294	if (rt) {
2295		if (rt->rt_gateway == NULL ||
2296		    rt->rt_gateway->sa_family != AF_INET6) {
2297			nd6log((LOG_ERR,
2298			    "ICMP6 redirect rejected; no route "
2299			    "with inet6 gateway found for redirect dst: %s\n",
2300			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2301			RTFREE(rt);
2302			goto bad;
2303		}
2304
2305		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2306		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2307			nd6log((LOG_ERR,
2308				"ICMP6 redirect rejected; "
2309				"not equal to gw-for-src=%s (must be same): "
2310				"%s\n",
2311				ip6_sprintf(gw6),
2312				icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2313			RTFREE(rt);
2314			goto bad;
2315		}
2316	} else {
2317		nd6log((LOG_ERR,
2318			"ICMP6 redirect rejected; "
2319			"no route found for redirect dst: %s\n",
2320			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2321		goto bad;
2322	}
2323	RTFREE(rt);
2324	rt = NULL;
2325    }
2326	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2327		nd6log((LOG_ERR,
2328			"ICMP6 redirect rejected; "
2329			"redirect dst must be unicast: %s\n",
2330			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2331		goto bad;
2332	}
2333
2334	is_router = is_onlink = 0;
2335	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2336		is_router = 1;	/* router case */
2337	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2338		is_onlink = 1;	/* on-link destination case */
2339	if (!is_router && !is_onlink) {
2340		nd6log((LOG_ERR,
2341			"ICMP6 redirect rejected; "
2342			"neither router case nor onlink case: %s\n",
2343			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2344		goto bad;
2345	}
2346	/* validation passed */
2347
2348	icmp6len -= sizeof(*nd_rd);
2349	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2350	if (nd6_options(&ndopts) < 0) {
2351		nd6log((LOG_INFO, "icmp6_redirect_input: "
2352			"invalid ND option, rejected: %s\n",
2353			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2354		/* nd6_options have incremented stats */
2355		goto freeit;
2356	}
2357
2358	if (ndopts.nd_opts_tgt_lladdr) {
2359		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2360		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2361	}
2362
2363	if (ndopts.nd_opts_rh) {
2364		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2365		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2366	}
2367
2368	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2369		nd6log((LOG_INFO,
2370			"icmp6_redirect_input: lladdrlen mismatch for %s "
2371			"(if %d, icmp6 packet %d): %s\n",
2372			ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2373			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2374		goto bad;
2375	}
2376
2377	/* RFC 2461 8.3 */
2378	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2379			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2380
2381	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2382		/* perform rtredirect */
2383		struct sockaddr_in6 sdst;
2384		struct sockaddr_in6 sgw;
2385		struct sockaddr_in6 ssrc;
2386
2387		bzero(&sdst, sizeof(sdst));
2388		bzero(&sgw, sizeof(sgw));
2389		bzero(&ssrc, sizeof(ssrc));
2390		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2391		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2392			sizeof(struct sockaddr_in6);
2393		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2394		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2395		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2396		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2397			   (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2398			   (struct sockaddr *)&ssrc,
2399			   (struct rtentry **)NULL);
2400	}
2401	/* finally update cached route in each socket via pfctlinput */
2402    {
2403	struct sockaddr_in6 sdst;
2404
2405	bzero(&sdst, sizeof(sdst));
2406	sdst.sin6_family = AF_INET6;
2407	sdst.sin6_len = sizeof(struct sockaddr_in6);
2408	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2409	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2410#ifdef IPSEC
2411	key_sa_routechange((struct sockaddr *)&sdst);
2412#endif
2413    }
2414
2415 freeit:
2416	m_freem(m);
2417	return;
2418
2419 bad:
2420	icmp6stat.icp6s_badredirect++;
2421	m_freem(m);
2422}
2423
2424void
2425icmp6_redirect_output(m0, rt)
2426	struct mbuf *m0;
2427	struct rtentry *rt;
2428{
2429	struct ifnet *ifp;	/* my outgoing interface */
2430	struct in6_addr *ifp_ll6;
2431	struct in6_addr *router_ll6;
2432	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2433	struct mbuf *m = NULL;	/* newly allocated one */
2434	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2435	struct nd_redirect *nd_rd;
2436	size_t maxlen;
2437	u_char *p;
2438	struct ifnet *outif = NULL;
2439	struct sockaddr_in6 src_sa;
2440
2441	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2442
2443	/* if we are not router, we don't send icmp6 redirect */
2444	if (!ip6_forwarding || ip6_accept_rtadv)
2445		goto fail;
2446
2447	/* sanity check */
2448	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2449		goto fail;
2450
2451	/*
2452	 * Address check:
2453	 *  the source address must identify a neighbor, and
2454	 *  the destination address must not be a multicast address
2455	 *  [RFC 2461, sec 8.2]
2456	 */
2457	sip6 = mtod(m0, struct ip6_hdr *);
2458	bzero(&src_sa, sizeof(src_sa));
2459	src_sa.sin6_family = AF_INET6;
2460	src_sa.sin6_len = sizeof(src_sa);
2461	src_sa.sin6_addr = sip6->ip6_src;
2462	/* we don't currently use sin6_scope_id, but eventually use it */
2463	src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src);
2464	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2465		goto fail;
2466	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2467		goto fail;	/* what should we do here? */
2468
2469	/* rate limit */
2470	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2471		goto fail;
2472
2473	/*
2474	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2475	 * we almost always ask for an mbuf cluster for simplicity.
2476	 * (MHLEN < IPV6_MMTU is almost always true)
2477	 */
2478#if IPV6_MMTU >= MCLBYTES
2479# error assumption failed about IPV6_MMTU and MCLBYTES
2480#endif
2481	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2482	if (m && IPV6_MMTU >= MHLEN)
2483		MCLGET(m, M_DONTWAIT);
2484	if (!m)
2485		goto fail;
2486	m->m_pkthdr.rcvif = NULL;
2487	m->m_len = 0;
2488	maxlen = M_TRAILINGSPACE(m);
2489	maxlen = min(IPV6_MMTU, maxlen);
2490	/* just for safety */
2491	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2492	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2493		goto fail;
2494	}
2495
2496	{
2497		/* get ip6 linklocal address for ifp(my outgoing interface). */
2498		struct in6_ifaddr *ia;
2499		if ((ia = in6ifa_ifpforlinklocal(ifp,
2500						 IN6_IFF_NOTREADY|
2501						 IN6_IFF_ANYCAST)) == NULL)
2502			goto fail;
2503		ifp_ll6 = &ia->ia_addr.sin6_addr;
2504	}
2505
2506	/* get ip6 linklocal address for the router. */
2507	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2508		struct sockaddr_in6 *sin6;
2509		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2510		router_ll6 = &sin6->sin6_addr;
2511		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2512			router_ll6 = (struct in6_addr *)NULL;
2513	} else
2514		router_ll6 = (struct in6_addr *)NULL;
2515
2516	/* ip6 */
2517	ip6 = mtod(m, struct ip6_hdr *);
2518	ip6->ip6_flow = 0;
2519	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2520	ip6->ip6_vfc |= IPV6_VERSION;
2521	/* ip6->ip6_plen will be set later */
2522	ip6->ip6_nxt = IPPROTO_ICMPV6;
2523	ip6->ip6_hlim = 255;
2524	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2525	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2526	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2527
2528	/* ND Redirect */
2529	nd_rd = (struct nd_redirect *)(ip6 + 1);
2530	nd_rd->nd_rd_type = ND_REDIRECT;
2531	nd_rd->nd_rd_code = 0;
2532	nd_rd->nd_rd_reserved = 0;
2533	if (rt->rt_flags & RTF_GATEWAY) {
2534		/*
2535		 * nd_rd->nd_rd_target must be a link-local address in
2536		 * better router cases.
2537		 */
2538		if (!router_ll6)
2539			goto fail;
2540		bcopy(router_ll6, &nd_rd->nd_rd_target,
2541		      sizeof(nd_rd->nd_rd_target));
2542		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2543		      sizeof(nd_rd->nd_rd_dst));
2544	} else {
2545		/* make sure redtgt == reddst */
2546		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2547		      sizeof(nd_rd->nd_rd_target));
2548		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2549		      sizeof(nd_rd->nd_rd_dst));
2550	}
2551
2552	p = (u_char *)(nd_rd + 1);
2553
2554	if (!router_ll6)
2555		goto nolladdropt;
2556
2557    {
2558	/* target lladdr option */
2559	struct rtentry *rt_router = NULL;
2560	int len;
2561	struct sockaddr_dl *sdl;
2562	struct nd_opt_hdr *nd_opt;
2563	char *lladdr;
2564
2565	rt_router = nd6_lookup(router_ll6, 0, ifp);
2566	if (!rt_router)
2567		goto nolladdropt;
2568	len = sizeof(*nd_opt) + ifp->if_addrlen;
2569	len = (len + 7) & ~7;	/* round by 8 */
2570	/* safety check */
2571	if (len + (p - (u_char *)ip6) > maxlen)
2572		goto nolladdropt;
2573	if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2574	    (rt_router->rt_flags & RTF_LLINFO) &&
2575	    (rt_router->rt_gateway->sa_family == AF_LINK) &&
2576	    (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2577	    sdl->sdl_alen) {
2578		nd_opt = (struct nd_opt_hdr *)p;
2579		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2580		nd_opt->nd_opt_len = len >> 3;
2581		lladdr = (char *)(nd_opt + 1);
2582		bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2583		p += len;
2584	}
2585    }
2586nolladdropt:;
2587
2588	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2589
2590	/* just to be safe */
2591#ifdef M_DECRYPTED	/*not openbsd*/
2592	if (m0->m_flags & M_DECRYPTED)
2593		goto noredhdropt;
2594#endif
2595	if (p - (u_char *)ip6 > maxlen)
2596		goto noredhdropt;
2597
2598    {
2599	/* redirected header option */
2600	int len;
2601	struct nd_opt_rd_hdr *nd_opt_rh;
2602
2603	/*
2604	 * compute the maximum size for icmp6 redirect header option.
2605	 * XXX room for auth header?
2606	 */
2607	len = maxlen - (p - (u_char *)ip6);
2608	len &= ~7;
2609
2610	/* This is just for simplicity. */
2611	if (m0->m_pkthdr.len != m0->m_len) {
2612		if (m0->m_next) {
2613			m_freem(m0->m_next);
2614			m0->m_next = NULL;
2615		}
2616		m0->m_pkthdr.len = m0->m_len;
2617	}
2618
2619	/*
2620	 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2621	 * about padding/truncate rule for the original IP packet.
2622	 * From the discussion on IPv6imp in Feb 1999, the consensus was:
2623	 * - "attach as much as possible" is the goal
2624	 * - pad if not aligned (original size can be guessed by original
2625	 *   ip6 header)
2626	 * Following code adds the padding if it is simple enough,
2627	 * and truncates if not.
2628	 */
2629	if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2630		panic("assumption failed in %s:%d\n", __FILE__, __LINE__);
2631
2632	if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2633		/* not enough room, truncate */
2634		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2635	} else {
2636		/* enough room, pad or truncate */
2637		size_t extra;
2638
2639		extra = m0->m_pkthdr.len % 8;
2640		if (extra) {
2641			/* pad if easy enough, truncate if not */
2642			if (8 - extra <= M_TRAILINGSPACE(m0)) {
2643				/* pad */
2644				m0->m_len += (8 - extra);
2645				m0->m_pkthdr.len += (8 - extra);
2646			} else {
2647				/* truncate */
2648				m0->m_pkthdr.len -= extra;
2649				m0->m_len -= extra;
2650			}
2651		}
2652		len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2653		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2654	}
2655
2656	nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2657	bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2658	nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2659	nd_opt_rh->nd_opt_rh_len = len >> 3;
2660	p += sizeof(*nd_opt_rh);
2661	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2662
2663	/* connect m0 to m */
2664	m_tag_delete_chain(m0, NULL);
2665	m0->m_flags &= ~M_PKTHDR;
2666	m->m_next = m0;
2667	m->m_pkthdr.len = m->m_len + m0->m_len;
2668	m0 = NULL;
2669    }
2670noredhdropt:;
2671	if (m0) {
2672		m_freem(m0);
2673		m0 = NULL;
2674	}
2675
2676	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
2677		sip6->ip6_src.s6_addr16[1] = 0;
2678	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
2679		sip6->ip6_dst.s6_addr16[1] = 0;
2680#if 0
2681	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
2682		ip6->ip6_src.s6_addr16[1] = 0;
2683	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
2684		ip6->ip6_dst.s6_addr16[1] = 0;
2685#endif
2686	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
2687		nd_rd->nd_rd_target.s6_addr16[1] = 0;
2688	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
2689		nd_rd->nd_rd_dst.s6_addr16[1] = 0;
2690
2691	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2692
2693	nd_rd->nd_rd_cksum = 0;
2694	nd_rd->nd_rd_cksum
2695		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2696
2697	/* send the packet to outside... */
2698	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2699	if (outif) {
2700		icmp6_ifstat_inc(outif, ifs6_out_msg);
2701		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2702	}
2703	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2704
2705	return;
2706
2707fail:
2708	if (m)
2709		m_freem(m);
2710	if (m0)
2711		m_freem(m0);
2712}
2713
2714#ifdef HAVE_NRL_INPCB
2715#define sotoin6pcb	sotoinpcb
2716#define in6pcb		inpcb
2717#define in6p_icmp6filt	inp_icmp6filt
2718#endif
2719/*
2720 * ICMPv6 socket option processing.
2721 */
2722int
2723icmp6_ctloutput(so, sopt)
2724	struct socket *so;
2725	struct sockopt *sopt;
2726{
2727	int error = 0;
2728	int optlen;
2729	struct inpcb *inp = sotoinpcb(so);
2730	int level, op, optname;
2731
2732	if (sopt) {
2733		level = sopt->sopt_level;
2734		op = sopt->sopt_dir;
2735		optname = sopt->sopt_name;
2736		optlen = sopt->sopt_valsize;
2737	} else
2738		level = op = optname = optlen = 0;
2739
2740	if (level != IPPROTO_ICMPV6) {
2741		return EINVAL;
2742	}
2743
2744	switch (op) {
2745	case PRCO_SETOPT:
2746		switch (optname) {
2747		case ICMP6_FILTER:
2748		    {
2749			struct icmp6_filter *p;
2750
2751			if (optlen != sizeof(*p)) {
2752				error = EMSGSIZE;
2753				break;
2754			}
2755			if (inp->in6p_icmp6filt == NULL) {
2756				error = EINVAL;
2757				break;
2758			}
2759			error = sooptcopyin(sopt, inp->in6p_icmp6filt, optlen,
2760				optlen);
2761			break;
2762		    }
2763
2764		default:
2765			error = ENOPROTOOPT;
2766			break;
2767		}
2768		break;
2769
2770	case PRCO_GETOPT:
2771		switch (optname) {
2772		case ICMP6_FILTER:
2773		    {
2774			if (inp->in6p_icmp6filt == NULL) {
2775				error = EINVAL;
2776				break;
2777			}
2778			error = sooptcopyout(sopt, inp->in6p_icmp6filt,
2779				sizeof(struct icmp6_filter));
2780			break;
2781		    }
2782
2783		default:
2784			error = ENOPROTOOPT;
2785			break;
2786		}
2787		break;
2788	}
2789
2790	return(error);
2791}
2792#ifdef HAVE_NRL_INPCB
2793#undef sotoin6pcb
2794#undef in6pcb
2795#undef in6p_icmp6filt
2796#endif
2797
2798#ifndef HAVE_PPSRATECHECK
2799#ifndef timersub
2800#define	timersub(tvp, uvp, vvp)						\
2801	do {								\
2802		(vvp)->tv_sec = (tvp)->tv_sec - (uvp)->tv_sec;		\
2803		(vvp)->tv_usec = (tvp)->tv_usec - (uvp)->tv_usec;	\
2804		if ((vvp)->tv_usec < 0) {				\
2805			(vvp)->tv_sec--;				\
2806			(vvp)->tv_usec += 1000000;			\
2807		}							\
2808	} while (0)
2809#endif
2810
2811/*
2812 * ppsratecheck(): packets (or events) per second limitation.
2813 */
2814static int
2815ppsratecheck(lasttime, curpps, maxpps)
2816	struct timeval *lasttime;
2817	int *curpps;
2818	int maxpps;	/* maximum pps allowed */
2819{
2820	struct timeval tv, delta;
2821	int s, rv;
2822
2823	s = splclock();
2824	microtime(&tv);
2825	splx(s);
2826
2827	timersub(&tv, lasttime, &delta);
2828
2829	/*
2830	 * Check for 0,0 so that the message will be seen at least once.
2831	 * If more than one second has passed since the last update of
2832	 * lasttime, reset the counter.
2833	 *
2834	 * We do increment *curpps even in *curpps < maxpps case, as some may
2835	 * try to use *curpps for stat purposes as well.
2836	 */
2837	if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) ||
2838	    delta.tv_sec >= 1) {
2839		*lasttime = tv;
2840		*curpps = 0;
2841		rv = 1;
2842	} else if (maxpps < 0)
2843		rv = 1;
2844	else if (*curpps < maxpps)
2845		rv = 1;
2846	else
2847		rv = 0;
2848
2849#if 1 /* DIAGNOSTIC? */
2850	/* be careful about wrap-around */
2851	if (*curpps + 1 > *curpps)
2852		*curpps = *curpps + 1;
2853#else
2854	/*
2855	 * assume that there's not too many calls to this function.
2856	 * not sure if the assumption holds, as it depends on *caller's*
2857	 * behavior, not the behavior of this function.
2858	 * IMHO it is wrong to make assumption on the caller's behavior,
2859	 * so the above #if is #if 1, not #ifdef DIAGNOSTIC.
2860	 */
2861	*curpps = *curpps + 1;
2862#endif
2863
2864	return (rv);
2865}
2866#endif
2867
2868/*
2869 * Perform rate limit check.
2870 * Returns 0 if it is okay to send the icmp6 packet.
2871 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2872 * limitation.
2873 *
2874 * XXX per-destination/type check necessary?
2875 */
2876static int
2877icmp6_ratelimit(dst, type, code)
2878	const struct in6_addr *dst;	/* not used at this moment */
2879	const int type;			/* not used at this moment */
2880	const int code;			/* not used at this moment */
2881{
2882	int ret;
2883
2884	ret = 0;	/* okay to send */
2885
2886	/* PPS limit */
2887	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2888	    icmp6errppslim)) {
2889		/* The packet is subject to rate limit */
2890		ret++;
2891	}
2892
2893	return ret;
2894}
2895