icmp6.c revision 148940
1/*	$FreeBSD: head/sys/netinet6/icmp6.c 148940 2005-08-10 17:30:10Z ume $	*/
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 * 4. Neither the name of the University nor the names of its contributors
46 *    may be used to endorse or promote products derived from this software
47 *    without specific prior written permission.
48 *
49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59 * SUCH DAMAGE.
60 *
61 *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
62 */
63
64#include "opt_inet.h"
65#include "opt_inet6.h"
66#include "opt_ipsec.h"
67
68#include <sys/param.h>
69#include <sys/domain.h>
70#include <sys/kernel.h>
71#include <sys/lock.h>
72#include <sys/malloc.h>
73#include <sys/mbuf.h>
74#include <sys/protosw.h>
75#include <sys/signalvar.h>
76#include <sys/socket.h>
77#include <sys/socketvar.h>
78#include <sys/sx.h>
79#include <sys/syslog.h>
80#include <sys/systm.h>
81#include <sys/time.h>
82
83#include <net/if.h>
84#include <net/if_dl.h>
85#include <net/if_types.h>
86#include <net/route.h>
87
88#include <netinet/in.h>
89#include <netinet/in_pcb.h>
90#include <netinet/in_var.h>
91#include <netinet/ip6.h>
92#include <netinet/icmp6.h>
93#include <netinet/tcp_var.h>
94#include <netinet6/in6_ifattach.h>
95#include <netinet6/in6_pcb.h>
96#include <netinet6/ip6protosw.h>
97#include <netinet6/ip6_var.h>
98#include <netinet6/scope6_var.h>
99#include <netinet6/mld6_var.h>
100#include <netinet6/nd6.h>
101
102#ifdef IPSEC
103#include <netinet6/ipsec.h>
104#include <netkey/key.h>
105#endif
106
107#ifdef FAST_IPSEC
108#include <netipsec/ipsec.h>
109#include <netipsec/key.h>
110#endif
111
112#include <net/net_osdep.h>
113
114extern struct domain inet6domain;
115
116struct icmp6stat icmp6stat;
117
118extern struct inpcbinfo ripcbinfo;
119extern struct inpcbhead ripcb;
120extern int icmp6errppslim;
121static int icmp6errpps_count = 0;
122static struct timeval icmp6errppslim_last;
123extern int icmp6_nodeinfo;
124
125static void icmp6_errcount __P((struct icmp6errstat *, int, int));
126static int icmp6_rip6_input __P((struct mbuf **, int));
127static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int));
128static const char *icmp6_redirect_diag __P((struct in6_addr *,
129	struct in6_addr *, struct in6_addr *));
130static struct mbuf *ni6_input __P((struct mbuf *, int));
131static struct mbuf *ni6_nametodns __P((const char *, int, int));
132static int ni6_dnsmatch __P((const char *, int, const char *, int));
133static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
134			  struct ifnet **, struct in6_addr *));
135static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
136				struct ifnet *, int));
137static int icmp6_notify_error __P((struct mbuf **, int, int, int));
138
139
140void
141icmp6_init()
142{
143	mld6_init();
144}
145
146static void
147icmp6_errcount(stat, type, code)
148	struct icmp6errstat *stat;
149	int type, code;
150{
151	switch (type) {
152	case ICMP6_DST_UNREACH:
153		switch (code) {
154		case ICMP6_DST_UNREACH_NOROUTE:
155			stat->icp6errs_dst_unreach_noroute++;
156			return;
157		case ICMP6_DST_UNREACH_ADMIN:
158			stat->icp6errs_dst_unreach_admin++;
159			return;
160		case ICMP6_DST_UNREACH_BEYONDSCOPE:
161			stat->icp6errs_dst_unreach_beyondscope++;
162			return;
163		case ICMP6_DST_UNREACH_ADDR:
164			stat->icp6errs_dst_unreach_addr++;
165			return;
166		case ICMP6_DST_UNREACH_NOPORT:
167			stat->icp6errs_dst_unreach_noport++;
168			return;
169		}
170		break;
171	case ICMP6_PACKET_TOO_BIG:
172		stat->icp6errs_packet_too_big++;
173		return;
174	case ICMP6_TIME_EXCEEDED:
175		switch (code) {
176		case ICMP6_TIME_EXCEED_TRANSIT:
177			stat->icp6errs_time_exceed_transit++;
178			return;
179		case ICMP6_TIME_EXCEED_REASSEMBLY:
180			stat->icp6errs_time_exceed_reassembly++;
181			return;
182		}
183		break;
184	case ICMP6_PARAM_PROB:
185		switch (code) {
186		case ICMP6_PARAMPROB_HEADER:
187			stat->icp6errs_paramprob_header++;
188			return;
189		case ICMP6_PARAMPROB_NEXTHEADER:
190			stat->icp6errs_paramprob_nextheader++;
191			return;
192		case ICMP6_PARAMPROB_OPTION:
193			stat->icp6errs_paramprob_option++;
194			return;
195		}
196		break;
197	case ND_REDIRECT:
198		stat->icp6errs_redirect++;
199		return;
200	}
201	stat->icp6errs_unknown++;
202}
203
204/*
205 * A wrapper function for icmp6_error() necessary when the erroneous packet
206 * may not contain enough scope zone information.
207 */
208void
209icmp6_error2(m, type, code, param, ifp)
210	struct mbuf *m;
211	int type, code, param;
212	struct ifnet *ifp;
213{
214	struct ip6_hdr *ip6;
215
216	if (ifp == NULL)
217		return;
218
219#ifndef PULLDOWN_TEST
220	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
221#else
222	if (m->m_len < sizeof(struct ip6_hdr)) {
223		m = m_pullup(m, sizeof(struct ip6_hdr));
224		if (m == NULL)
225			return;
226	}
227#endif
228
229	ip6 = mtod(m, struct ip6_hdr *);
230
231	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
232		return;
233	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
234		return;
235
236	icmp6_error(m, type, code, param);
237}
238
239/*
240 * Generate an error packet of type error in response to bad IP6 packet.
241 */
242void
243icmp6_error(m, type, code, param)
244	struct mbuf *m;
245	int type, code, param;
246{
247	struct ip6_hdr *oip6, *nip6;
248	struct icmp6_hdr *icmp6;
249	u_int preplen;
250	int off;
251	int nxt;
252
253	icmp6stat.icp6s_error++;
254
255	/* count per-type-code statistics */
256	icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
257
258#ifdef M_DECRYPTED	/*not openbsd*/
259	if (m->m_flags & M_DECRYPTED) {
260		icmp6stat.icp6s_canterror++;
261		goto freeit;
262	}
263#endif
264
265#ifndef PULLDOWN_TEST
266	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
267#else
268	if (m->m_len < sizeof(struct ip6_hdr)) {
269		m = m_pullup(m, sizeof(struct ip6_hdr));
270		if (m == NULL)
271			return;
272	}
273#endif
274	oip6 = mtod(m, struct ip6_hdr *);
275
276	/*
277	 * If the destination address of the erroneous packet is a multicast
278	 * address, or the packet was sent using link-layer multicast,
279	 * we should basically suppress sending an error (RFC 2463, Section
280	 * 2.4).
281	 * We have two exceptions (the item e.2 in that section):
282	 * - the Pakcet Too Big message can be sent for path MTU discovery.
283	 * - the Parameter Problem Message that can be allowed an icmp6 error
284	 *   in the option type field.  This check has been done in
285	 *   ip6_unknown_opt(), so we can just check the type and code.
286	 */
287	if ((m->m_flags & (M_BCAST|M_MCAST) ||
288	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
289	    (type != ICMP6_PACKET_TOO_BIG &&
290	     (type != ICMP6_PARAM_PROB ||
291	      code != ICMP6_PARAMPROB_OPTION)))
292		goto freeit;
293
294	/*
295	 * RFC 2463, 2.4 (e.5): source address check.
296	 * XXX: the case of anycast source?
297	 */
298	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
299	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
300		goto freeit;
301
302	/*
303	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
304	 * don't do it.
305	 */
306	nxt = -1;
307	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
308	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
309		struct icmp6_hdr *icp;
310
311#ifndef PULLDOWN_TEST
312		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
313		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
314#else
315		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
316			sizeof(*icp));
317		if (icp == NULL) {
318			icmp6stat.icp6s_tooshort++;
319			return;
320		}
321#endif
322		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
323		    icp->icmp6_type == ND_REDIRECT) {
324			/*
325			 * ICMPv6 error
326			 * Special case: for redirect (which is
327			 * informational) we must not send icmp6 error.
328			 */
329			icmp6stat.icp6s_canterror++;
330			goto freeit;
331		} else {
332			/* ICMPv6 informational - send the error */
333		}
334	} else {
335		/* non-ICMPv6 - send the error */
336	}
337
338	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
339
340	/* Finally, do rate limitation check. */
341	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
342		icmp6stat.icp6s_toofreq++;
343		goto freeit;
344	}
345
346	/*
347	 * OK, ICMP6 can be generated.
348	 */
349
350	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
351		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
352
353	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
354	M_PREPEND(m, preplen, M_DONTWAIT);
355	if (m && m->m_len < preplen)
356		m = m_pullup(m, preplen);
357	if (m == NULL) {
358		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
359		return;
360	}
361
362	nip6 = mtod(m, struct ip6_hdr *);
363	nip6->ip6_src  = oip6->ip6_src;
364	nip6->ip6_dst  = oip6->ip6_dst;
365
366	in6_clearscope(&oip6->ip6_src);
367	in6_clearscope(&oip6->ip6_dst);
368
369	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
370	icmp6->icmp6_type = type;
371	icmp6->icmp6_code = code;
372	icmp6->icmp6_pptr = htonl((u_int32_t)param);
373
374	/*
375	 * icmp6_reflect() is designed to be in the input path.
376	 * icmp6_error() can be called from both input and outut path,
377	 * and if we are in output path rcvif could contain bogus value.
378	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
379	 * information in ip header (nip6).
380	 */
381	m->m_pkthdr.rcvif = NULL;
382
383	icmp6stat.icp6s_outhist[type]++;
384	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
385
386	return;
387
388  freeit:
389	/*
390	 * If we can't tell wheter or not we can generate ICMP6, free it.
391	 */
392	m_freem(m);
393}
394
395/*
396 * Process a received ICMP6 message.
397 */
398int
399icmp6_input(mp, offp, proto)
400	struct mbuf **mp;
401	int *offp, proto;
402{
403	struct mbuf *m = *mp, *n;
404	struct ip6_hdr *ip6, *nip6;
405	struct icmp6_hdr *icmp6, *nicmp6;
406	int off = *offp;
407	int icmp6len = m->m_pkthdr.len - *offp;
408	int code, sum, noff;
409
410#ifndef PULLDOWN_TEST
411	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
412	/* m might change if M_LOOP.  So, call mtod after this */
413#endif
414
415	/*
416	 * Locate icmp6 structure in mbuf, and check
417	 * that not corrupted and of at least minimum length
418	 */
419
420	ip6 = mtod(m, struct ip6_hdr *);
421	if (icmp6len < sizeof(struct icmp6_hdr)) {
422		icmp6stat.icp6s_tooshort++;
423		goto freeit;
424	}
425
426	/*
427	 * calculate the checksum
428	 */
429#ifndef PULLDOWN_TEST
430	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
431#else
432	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
433	if (icmp6 == NULL) {
434		icmp6stat.icp6s_tooshort++;
435		return IPPROTO_DONE;
436	}
437#endif
438	code = icmp6->icmp6_code;
439
440	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
441		nd6log((LOG_ERR,
442		    "ICMP6 checksum error(%d|%x) %s\n",
443		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
444		icmp6stat.icp6s_checksum++;
445		goto freeit;
446	}
447
448	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
449		/*
450		 * Deliver very specific ICMP6 type only.
451		 * This is important to deilver TOOBIG.  Otherwise PMTUD
452		 * will not work.
453		 */
454		switch (icmp6->icmp6_type) {
455		case ICMP6_DST_UNREACH:
456		case ICMP6_PACKET_TOO_BIG:
457		case ICMP6_TIME_EXCEEDED:
458			break;
459		default:
460			goto freeit;
461		}
462	}
463
464	icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
465	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
466	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
467		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
468
469	switch (icmp6->icmp6_type) {
470	case ICMP6_DST_UNREACH:
471		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
472		switch (code) {
473		case ICMP6_DST_UNREACH_NOROUTE:
474			code = PRC_UNREACH_NET;
475			break;
476		case ICMP6_DST_UNREACH_ADMIN:
477			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
478			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
479			break;
480		case ICMP6_DST_UNREACH_ADDR:
481			code = PRC_HOSTDEAD;
482			break;
483		case ICMP6_DST_UNREACH_BEYONDSCOPE:
484			/* I mean "source address was incorrect." */
485			code = PRC_PARAMPROB;
486			break;
487		case ICMP6_DST_UNREACH_NOPORT:
488			code = PRC_UNREACH_PORT;
489			break;
490		default:
491			goto badcode;
492		}
493		goto deliver;
494		break;
495
496	case ICMP6_PACKET_TOO_BIG:
497		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
498
499		/* validation is made in icmp6_mtudisc_update */
500
501		code = PRC_MSGSIZE;
502
503		/*
504		 * Updating the path MTU will be done after examining
505		 * intermediate extension headers.
506		 */
507		goto deliver;
508		break;
509
510	case ICMP6_TIME_EXCEEDED:
511		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
512		switch (code) {
513		case ICMP6_TIME_EXCEED_TRANSIT:
514		case ICMP6_TIME_EXCEED_REASSEMBLY:
515			code += PRC_TIMXCEED_INTRANS;
516			break;
517		default:
518			goto badcode;
519		}
520		goto deliver;
521		break;
522
523	case ICMP6_PARAM_PROB:
524		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
525		switch (code) {
526		case ICMP6_PARAMPROB_NEXTHEADER:
527			code = PRC_UNREACH_PROTOCOL;
528			break;
529		case ICMP6_PARAMPROB_HEADER:
530		case ICMP6_PARAMPROB_OPTION:
531			code = PRC_PARAMPROB;
532			break;
533		default:
534			goto badcode;
535		}
536		goto deliver;
537		break;
538
539	case ICMP6_ECHO_REQUEST:
540		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
541		if (code != 0)
542			goto badcode;
543		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
544			/* Give up remote */
545			break;
546		}
547		if ((n->m_flags & M_EXT) != 0
548		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
549			struct mbuf *n0 = n;
550			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
551			int n0len;
552
553			MGETHDR(n, M_DONTWAIT, n0->m_type);
554			n0len = n0->m_pkthdr.len;	/* save for use below */
555			if (n)
556				M_MOVE_PKTHDR(n, n0);
557			if (n && maxlen >= MHLEN) {
558				MCLGET(n, M_DONTWAIT);
559				if ((n->m_flags & M_EXT) == 0) {
560					m_free(n);
561					n = NULL;
562				}
563			}
564			if (n == NULL) {
565				/* Give up remote */
566				m_freem(n0);
567				break;
568			}
569			/*
570			 * Copy IPv6 and ICMPv6 only.
571			 */
572			nip6 = mtod(n, struct ip6_hdr *);
573			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
574			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
575			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
576			noff = sizeof(struct ip6_hdr);
577			/* new mbuf contains only ipv6+icmpv6 headers */
578			n->m_len = noff + sizeof(struct icmp6_hdr);
579			/*
580			 * Adjust mbuf.  ip6_plen will be adjusted in
581			 * ip6_output().
582			 */
583			m_adj(n0, off + sizeof(struct icmp6_hdr));
584			/* recalculate complete packet size */
585			n->m_pkthdr.len = n0len + (noff - off);
586			n->m_next = n0;
587		} else {
588			nip6 = mtod(n, struct ip6_hdr *);
589			nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
590			noff = off;
591		}
592		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
593		nicmp6->icmp6_code = 0;
594		if (n) {
595			icmp6stat.icp6s_reflect++;
596			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
597			icmp6_reflect(n, noff);
598		}
599		break;
600
601	case ICMP6_ECHO_REPLY:
602		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
603		if (code != 0)
604			goto badcode;
605		break;
606
607	case MLD_LISTENER_QUERY:
608	case MLD_LISTENER_REPORT:
609		if (icmp6len < sizeof(struct mld_hdr))
610			goto badlen;
611		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
612			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
613		else
614			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
615		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
616			/* give up local */
617			mld6_input(m, off);
618			m = NULL;
619			goto freeit;
620		}
621		mld6_input(n, off);
622		/* m stays. */
623		break;
624
625	case MLD_LISTENER_DONE:
626		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
627		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
628			goto badlen;
629		break;		/* nothing to be done in kernel */
630
631	case MLD_MTRACE_RESP:
632	case MLD_MTRACE:
633		/* XXX: these two are experimental.  not officially defined. */
634		/* XXX: per-interface statistics? */
635		break;		/* just pass it to applications */
636
637	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
638	    {
639		enum { WRU, FQDN } mode;
640
641		if (!icmp6_nodeinfo)
642			break;
643
644		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
645			mode = WRU;
646		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
647			mode = FQDN;
648		else
649			goto badlen;
650
651#define hostnamelen	strlen(hostname)
652		if (mode == FQDN) {
653#ifndef PULLDOWN_TEST
654			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
655			    IPPROTO_DONE);
656#endif
657			n = m_copy(m, 0, M_COPYALL);
658			if (n)
659				n = ni6_input(n, off);
660			/* XXX meaningless if n == NULL */
661			noff = sizeof(struct ip6_hdr);
662		} else {
663			u_char *p;
664			int maxlen, maxhlen;
665
666			if ((icmp6_nodeinfo & 5) != 5)
667				break;
668
669			if (code != 0)
670				goto badcode;
671			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
672			if (maxlen >= MCLBYTES) {
673				/* Give up remote */
674				break;
675			}
676			MGETHDR(n, M_DONTWAIT, m->m_type);
677			if (n && maxlen > MHLEN) {
678				MCLGET(n, M_DONTWAIT);
679				if ((n->m_flags & M_EXT) == 0) {
680					m_free(n);
681					n = NULL;
682				}
683			}
684			if (!m_dup_pkthdr(n, m, M_DONTWAIT)) {
685				/*
686				 * Previous code did a blind M_COPY_PKTHDR
687				 * and said "just for rcvif".  If true, then
688				 * we could tolerate the dup failing (due to
689				 * the deep copy of the tag chain).  For now
690				 * be conservative and just fail.
691				 */
692				m_free(n);
693				n = NULL;
694			}
695			if (n == NULL) {
696				/* Give up remote */
697				break;
698			}
699			n->m_pkthdr.rcvif = NULL;
700			n->m_len = 0;
701			maxhlen = M_TRAILINGSPACE(n) - maxlen;
702			if (maxhlen > hostnamelen)
703				maxhlen = hostnamelen;
704			/*
705			 * Copy IPv6 and ICMPv6 only.
706			 */
707			nip6 = mtod(n, struct ip6_hdr *);
708			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
709			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
710			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
711			p = (u_char *)(nicmp6 + 1);
712			bzero(p, 4);
713			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
714			noff = sizeof(struct ip6_hdr);
715			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
716				sizeof(struct icmp6_hdr) + 4 + maxhlen;
717			nicmp6->icmp6_type = ICMP6_WRUREPLY;
718			nicmp6->icmp6_code = 0;
719		}
720#undef hostnamelen
721		if (n) {
722			icmp6stat.icp6s_reflect++;
723			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
724			icmp6_reflect(n, noff);
725		}
726		break;
727	    }
728
729	case ICMP6_WRUREPLY:
730		if (code != 0)
731			goto badcode;
732		break;
733
734	case ND_ROUTER_SOLICIT:
735		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
736		if (code != 0)
737			goto badcode;
738		if (icmp6len < sizeof(struct nd_router_solicit))
739			goto badlen;
740		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
741			/* give up local */
742			nd6_rs_input(m, off, icmp6len);
743			m = NULL;
744			goto freeit;
745		}
746		nd6_rs_input(n, off, icmp6len);
747		/* m stays. */
748		break;
749
750	case ND_ROUTER_ADVERT:
751		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
752		if (code != 0)
753			goto badcode;
754		if (icmp6len < sizeof(struct nd_router_advert))
755			goto badlen;
756		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
757			/* give up local */
758			nd6_ra_input(m, off, icmp6len);
759			m = NULL;
760			goto freeit;
761		}
762		nd6_ra_input(n, off, icmp6len);
763		/* m stays. */
764		break;
765
766	case ND_NEIGHBOR_SOLICIT:
767		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
768		if (code != 0)
769			goto badcode;
770		if (icmp6len < sizeof(struct nd_neighbor_solicit))
771			goto badlen;
772		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
773			/* give up local */
774			nd6_ns_input(m, off, icmp6len);
775			m = NULL;
776			goto freeit;
777		}
778		nd6_ns_input(n, off, icmp6len);
779		/* m stays. */
780		break;
781
782	case ND_NEIGHBOR_ADVERT:
783		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
784		if (code != 0)
785			goto badcode;
786		if (icmp6len < sizeof(struct nd_neighbor_advert))
787			goto badlen;
788		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
789			/* give up local */
790			nd6_na_input(m, off, icmp6len);
791			m = NULL;
792			goto freeit;
793		}
794		nd6_na_input(n, off, icmp6len);
795		/* m stays. */
796		break;
797
798	case ND_REDIRECT:
799		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
800		if (code != 0)
801			goto badcode;
802		if (icmp6len < sizeof(struct nd_redirect))
803			goto badlen;
804		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
805			/* give up local */
806			icmp6_redirect_input(m, off);
807			m = NULL;
808			goto freeit;
809		}
810		icmp6_redirect_input(n, off);
811		/* m stays. */
812		break;
813
814	case ICMP6_ROUTER_RENUMBERING:
815		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
816		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
817			goto badcode;
818		if (icmp6len < sizeof(struct icmp6_router_renum))
819			goto badlen;
820		break;
821
822	default:
823		nd6log((LOG_DEBUG,
824		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
825		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
826		    ip6_sprintf(&ip6->ip6_dst),
827		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
828		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
829			/* ICMPv6 error: MUST deliver it by spec... */
830			code = PRC_NCMDS;
831			/* deliver */
832		} else {
833			/* ICMPv6 informational: MUST not deliver */
834			break;
835		}
836	deliver:
837		if (icmp6_notify_error(&m, off, icmp6len, code)) {
838			/* In this case, m should've been freed. */
839			return (IPPROTO_DONE);
840		}
841		break;
842
843	badcode:
844		icmp6stat.icp6s_badcode++;
845		break;
846
847	badlen:
848		icmp6stat.icp6s_badlen++;
849		break;
850	}
851
852	/* deliver the packet to appropriate sockets */
853	icmp6_rip6_input(&m, *offp);
854
855	return IPPROTO_DONE;
856
857 freeit:
858	m_freem(m);
859	return IPPROTO_DONE;
860}
861
862static int
863icmp6_notify_error(mp, off, icmp6len, code)
864	struct mbuf **mp;
865	int off, icmp6len, code;
866{
867	struct mbuf *m = *mp;
868	struct icmp6_hdr *icmp6;
869	struct ip6_hdr *eip6;
870	u_int32_t notifymtu;
871	struct sockaddr_in6 icmp6src, icmp6dst;
872
873	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
874		icmp6stat.icp6s_tooshort++;
875		goto freeit;
876	}
877#ifndef PULLDOWN_TEST
878	IP6_EXTHDR_CHECK(m, off,
879	    sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
880	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
881#else
882	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
883	    sizeof(*icmp6) + sizeof(struct ip6_hdr));
884	if (icmp6 == NULL) {
885		icmp6stat.icp6s_tooshort++;
886		return (-1);
887	}
888#endif
889	eip6 = (struct ip6_hdr *)(icmp6 + 1);
890
891	/* Detect the upper level protocol */
892	{
893		void (*ctlfunc) __P((int, struct sockaddr *, void *));
894		u_int8_t nxt = eip6->ip6_nxt;
895		int eoff = off + sizeof(struct icmp6_hdr) +
896		    sizeof(struct ip6_hdr);
897		struct ip6ctlparam ip6cp;
898		struct in6_addr *finaldst = NULL;
899		int icmp6type = icmp6->icmp6_type;
900		struct ip6_frag *fh;
901		struct ip6_rthdr *rth;
902		struct ip6_rthdr0 *rth0;
903		int rthlen;
904
905		while (1) { /* XXX: should avoid infinite loop explicitly? */
906			struct ip6_ext *eh;
907
908			switch (nxt) {
909			case IPPROTO_HOPOPTS:
910			case IPPROTO_DSTOPTS:
911			case IPPROTO_AH:
912#ifndef PULLDOWN_TEST
913				IP6_EXTHDR_CHECK(m, 0,
914				    eoff + sizeof(struct ip6_ext), -1);
915				eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
916#else
917				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
918				    eoff, sizeof(*eh));
919				if (eh == NULL) {
920					icmp6stat.icp6s_tooshort++;
921					return (-1);
922				}
923#endif
924
925				if (nxt == IPPROTO_AH)
926					eoff += (eh->ip6e_len + 2) << 2;
927				else
928					eoff += (eh->ip6e_len + 1) << 3;
929				nxt = eh->ip6e_nxt;
930				break;
931			case IPPROTO_ROUTING:
932				/*
933				 * When the erroneous packet contains a
934				 * routing header, we should examine the
935				 * header to determine the final destination.
936				 * Otherwise, we can't properly update
937				 * information that depends on the final
938				 * destination (e.g. path MTU).
939				 */
940#ifndef PULLDOWN_TEST
941				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
942				rth = (struct ip6_rthdr *)
943				    (mtod(m, caddr_t) + eoff);
944#else
945				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
946				    eoff, sizeof(*rth));
947				if (rth == NULL) {
948					icmp6stat.icp6s_tooshort++;
949					return (-1);
950				}
951#endif
952				rthlen = (rth->ip6r_len + 1) << 3;
953				/*
954				 * XXX: currently there is no
955				 * officially defined type other
956				 * than type-0.
957				 * Note that if the segment left field
958				 * is 0, all intermediate hops must
959				 * have been passed.
960				 */
961				if (rth->ip6r_segleft &&
962				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
963					int hops;
964
965#ifndef PULLDOWN_TEST
966					IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
967					rth0 = (struct ip6_rthdr0 *)
968					    (mtod(m, caddr_t) + eoff);
969#else
970					IP6_EXTHDR_GET(rth0,
971					    struct ip6_rthdr0 *, m,
972					    eoff, rthlen);
973					if (rth0 == NULL) {
974						icmp6stat.icp6s_tooshort++;
975						return (-1);
976					}
977#endif
978					/* just ignore a bogus header */
979					if ((rth0->ip6r0_len % 2) == 0 &&
980					    (hops = rth0->ip6r0_len/2))
981						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
982				}
983				eoff += rthlen;
984				nxt = rth->ip6r_nxt;
985				break;
986			case IPPROTO_FRAGMENT:
987#ifndef PULLDOWN_TEST
988				IP6_EXTHDR_CHECK(m, 0, eoff +
989				    sizeof(struct ip6_frag), -1);
990				fh = (struct ip6_frag *)(mtod(m, caddr_t) +
991				    eoff);
992#else
993				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
994				    eoff, sizeof(*fh));
995				if (fh == NULL) {
996					icmp6stat.icp6s_tooshort++;
997					return (-1);
998				}
999#endif
1000				/*
1001				 * Data after a fragment header is meaningless
1002				 * unless it is the first fragment, but
1003				 * we'll go to the notify label for path MTU
1004				 * discovery.
1005				 */
1006				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1007					goto notify;
1008
1009				eoff += sizeof(struct ip6_frag);
1010				nxt = fh->ip6f_nxt;
1011				break;
1012			default:
1013				/*
1014				 * This case includes ESP and the No Next
1015				 * Header.  In such cases going to the notify
1016				 * label does not have any meaning
1017				 * (i.e. ctlfunc will be NULL), but we go
1018				 * anyway since we might have to update
1019				 * path MTU information.
1020				 */
1021				goto notify;
1022			}
1023		}
1024	  notify:
1025#ifndef PULLDOWN_TEST
1026		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1027#else
1028		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1029		    sizeof(*icmp6) + sizeof(struct ip6_hdr));
1030		if (icmp6 == NULL) {
1031			icmp6stat.icp6s_tooshort++;
1032			return (-1);
1033		}
1034#endif
1035
1036		/*
1037		 * retrieve parameters from the inner IPv6 header, and convert
1038		 * them into sockaddr structures.
1039		 * XXX: there is no guarantee that the source or destination
1040		 * addresses of the inner packet are in the same scope as
1041		 * the addresses of the icmp packet.  But there is no other
1042		 * way to determine the zone.
1043		 */
1044		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1045
1046		bzero(&icmp6dst, sizeof(icmp6dst));
1047		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1048		icmp6dst.sin6_family = AF_INET6;
1049		if (finaldst == NULL)
1050			icmp6dst.sin6_addr = eip6->ip6_dst;
1051		else
1052			icmp6dst.sin6_addr = *finaldst;
1053		if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1054			goto freeit;
1055		bzero(&icmp6src, sizeof(icmp6src));
1056		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1057		icmp6src.sin6_family = AF_INET6;
1058		icmp6src.sin6_addr = eip6->ip6_src;
1059		if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1060			goto freeit;
1061		icmp6src.sin6_flowinfo =
1062		    (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1063
1064		if (finaldst == NULL)
1065			finaldst = &eip6->ip6_dst;
1066		ip6cp.ip6c_m = m;
1067		ip6cp.ip6c_icmp6 = icmp6;
1068		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1069		ip6cp.ip6c_off = eoff;
1070		ip6cp.ip6c_finaldst = finaldst;
1071		ip6cp.ip6c_src = &icmp6src;
1072		ip6cp.ip6c_nxt = nxt;
1073
1074		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1075			notifymtu = ntohl(icmp6->icmp6_mtu);
1076			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1077			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1078		}
1079
1080		ctlfunc = (void (*) __P((int, struct sockaddr *, void *)))
1081		    (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1082		if (ctlfunc) {
1083			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1084			    &ip6cp);
1085		}
1086	}
1087	*mp = m;
1088	return (0);
1089
1090  freeit:
1091	m_freem(m);
1092	return (-1);
1093}
1094
1095void
1096icmp6_mtudisc_update(ip6cp, validated)
1097	struct ip6ctlparam *ip6cp;
1098	int validated;
1099{
1100	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1101	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1102	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1103	u_int mtu = ntohl(icmp6->icmp6_mtu);
1104	struct in_conninfo inc;
1105
1106#if 0
1107	/*
1108	 * RFC2460 section 5, last paragraph.
1109	 * even though minimum link MTU for IPv6 is IPV6_MMTU,
1110	 * we may see ICMPv6 too big with mtu < IPV6_MMTU
1111	 * due to packet translator in the middle.
1112	 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1113	 * special handling.
1114	 */
1115	if (mtu < IPV6_MMTU)
1116		return;
1117#endif
1118
1119	/*
1120	 * we reject ICMPv6 too big with abnormally small value.
1121	 * XXX what is the good definition of "abnormally small"?
1122	 */
1123	if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1124		return;
1125
1126	if (!validated)
1127		return;
1128
1129	bzero(&inc, sizeof(inc));
1130	inc.inc_flags = 1; /* IPv6 */
1131	inc.inc6_faddr = *dst;
1132	if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1133		return;
1134
1135	if (mtu < tcp_maxmtu6(&inc)) {
1136		tcp_hc_updatemtu(&inc, mtu);
1137		icmp6stat.icp6s_pmtuchg++;
1138	}
1139}
1140
1141/*
1142 * Process a Node Information Query packet, based on
1143 * draft-ietf-ipngwg-icmp-name-lookups-07.
1144 *
1145 * Spec incompatibilities:
1146 * - IPv6 Subject address handling
1147 * - IPv4 Subject address handling support missing
1148 * - Proxy reply (answer even if it's not for me)
1149 * - joins NI group address at in6_ifattach() time only, does not cope
1150 *   with hostname changes by sethostname(3)
1151 */
1152#define hostnamelen	strlen(hostname)
1153static struct mbuf *
1154ni6_input(m, off)
1155	struct mbuf *m;
1156	int off;
1157{
1158	struct icmp6_nodeinfo *ni6, *nni6;
1159	struct mbuf *n = NULL;
1160	u_int16_t qtype;
1161	int subjlen;
1162	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1163	struct ni_reply_fqdn *fqdn;
1164	int addrs;		/* for NI_QTYPE_NODEADDR */
1165	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1166	struct in6_addr in6_subj; /* subject address */
1167	struct ip6_hdr *ip6;
1168	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1169	char *subj = NULL;
1170	struct in6_ifaddr *ia6 = NULL;
1171
1172	ip6 = mtod(m, struct ip6_hdr *);
1173#ifndef PULLDOWN_TEST
1174	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1175#else
1176	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1177	if (ni6 == NULL) {
1178		/* m is already reclaimed */
1179		return (NULL);
1180	}
1181#endif
1182
1183	/*
1184	 * Validate IPv6 destination address.
1185	 *
1186	 * The Responder must discard the Query without further processing
1187	 * unless it is one of the Responder's unicast or anycast addresses, or
1188	 * a link-local scope multicast address which the Responder has joined.
1189	 * [icmp-name-lookups-08, Section 4.]
1190	 */
1191	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1192		if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1193			goto bad;
1194		/* else it's a link-local multicast, fine */
1195	} else {		/* unicast or anycast */
1196		if ((ia6 = ip6_getdstifaddr(m)) == NULL)
1197			goto bad; /* XXX impossible */
1198
1199		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1200		    !(icmp6_nodeinfo & 4)) {
1201			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1202				"a temporary address in %s:%d",
1203			       __FILE__, __LINE__));
1204			goto bad;
1205		}
1206	}
1207
1208	/* validate query Subject field. */
1209	qtype = ntohs(ni6->ni_qtype);
1210	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1211	switch (qtype) {
1212	case NI_QTYPE_NOOP:
1213	case NI_QTYPE_SUPTYPES:
1214		/* 07 draft */
1215		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1216			break;
1217		/* FALLTHROUGH */
1218	case NI_QTYPE_FQDN:
1219	case NI_QTYPE_NODEADDR:
1220		switch (ni6->ni_code) {
1221		case ICMP6_NI_SUBJ_IPV6:
1222#if ICMP6_NI_SUBJ_IPV6 != 0
1223		case 0:
1224#endif
1225			/*
1226			 * backward compatibility - try to accept 03 draft
1227			 * format, where no Subject is present.
1228			 */
1229			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1230			    subjlen == 0) {
1231				oldfqdn++;
1232				break;
1233			}
1234#if ICMP6_NI_SUBJ_IPV6 != 0
1235			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1236				goto bad;
1237#endif
1238
1239			if (subjlen != sizeof(struct in6_addr))
1240				goto bad;
1241
1242			/*
1243			 * Validate Subject address.
1244			 *
1245			 * Not sure what exactly "address belongs to the node"
1246			 * means in the spec, is it just unicast, or what?
1247			 *
1248			 * At this moment we consider Subject address as
1249			 * "belong to the node" if the Subject address equals
1250			 * to the IPv6 destination address; validation for
1251			 * IPv6 destination address should have done enough
1252			 * check for us.
1253			 *
1254			 * We do not do proxy at this moment.
1255			 */
1256			/* m_pulldown instead of copy? */
1257			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1258			    subjlen, (caddr_t)&in6_subj);
1259			if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1260				goto bad;
1261
1262			subj = (char *)&in6_subj;
1263			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1264				break;
1265
1266			/*
1267			 * XXX if we are to allow other cases, we should really
1268			 * be careful about scope here.
1269			 * basically, we should disallow queries toward IPv6
1270			 * destination X with subject Y,
1271			 * if scope(X) > scope(Y).
1272			 * if we allow scope(X) > scope(Y), it will result in
1273			 * information leakage across scope boundary.
1274			 */
1275			goto bad;
1276
1277		case ICMP6_NI_SUBJ_FQDN:
1278			/*
1279			 * Validate Subject name with gethostname(3).
1280			 *
1281			 * The behavior may need some debate, since:
1282			 * - we are not sure if the node has FQDN as
1283			 *   hostname (returned by gethostname(3)).
1284			 * - the code does wildcard match for truncated names.
1285			 *   however, we are not sure if we want to perform
1286			 *   wildcard match, if gethostname(3) side has
1287			 *   truncated hostname.
1288			 */
1289			n = ni6_nametodns(hostname, hostnamelen, 0);
1290			if (!n || n->m_next || n->m_len == 0)
1291				goto bad;
1292			IP6_EXTHDR_GET(subj, char *, m,
1293			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1294			if (subj == NULL)
1295				goto bad;
1296			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1297			    n->m_len)) {
1298				goto bad;
1299			}
1300			m_freem(n);
1301			n = NULL;
1302			break;
1303
1304		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1305		default:
1306			goto bad;
1307		}
1308		break;
1309	}
1310
1311	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1312	switch (qtype) {
1313	case NI_QTYPE_FQDN:
1314		if ((icmp6_nodeinfo & 1) == 0)
1315			goto bad;
1316		break;
1317	case NI_QTYPE_NODEADDR:
1318		if ((icmp6_nodeinfo & 2) == 0)
1319			goto bad;
1320		break;
1321	}
1322
1323	/* guess reply length */
1324	switch (qtype) {
1325	case NI_QTYPE_NOOP:
1326		break;		/* no reply data */
1327	case NI_QTYPE_SUPTYPES:
1328		replylen += sizeof(u_int32_t);
1329		break;
1330	case NI_QTYPE_FQDN:
1331		/* XXX will append an mbuf */
1332		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1333		break;
1334	case NI_QTYPE_NODEADDR:
1335		addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1336		if ((replylen += addrs * (sizeof(struct in6_addr) +
1337		    sizeof(u_int32_t))) > MCLBYTES)
1338			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1339		break;
1340	default:
1341		/*
1342		 * XXX: We must return a reply with the ICMP6 code
1343		 * `unknown Qtype' in this case.  However we regard the case
1344		 * as an FQDN query for backward compatibility.
1345		 * Older versions set a random value to this field,
1346		 * so it rarely varies in the defined qtypes.
1347		 * But the mechanism is not reliable...
1348		 * maybe we should obsolete older versions.
1349		 */
1350		qtype = NI_QTYPE_FQDN;
1351		/* XXX will append an mbuf */
1352		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1353		oldfqdn++;
1354		break;
1355	}
1356
1357	/* allocate an mbuf to reply. */
1358	MGETHDR(n, M_DONTWAIT, m->m_type);
1359	if (n == NULL) {
1360		m_freem(m);
1361		return (NULL);
1362	}
1363	M_MOVE_PKTHDR(n, m); /* just for recvif */
1364	if (replylen > MHLEN) {
1365		if (replylen > MCLBYTES) {
1366			/*
1367			 * XXX: should we try to allocate more? But MCLBYTES
1368			 * is probably much larger than IPV6_MMTU...
1369			 */
1370			goto bad;
1371		}
1372		MCLGET(n, M_DONTWAIT);
1373		if ((n->m_flags & M_EXT) == 0) {
1374			goto bad;
1375		}
1376	}
1377	n->m_pkthdr.len = n->m_len = replylen;
1378
1379	/* copy mbuf header and IPv6 + Node Information base headers */
1380	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1381	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1382	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1383
1384	/* qtype dependent procedure */
1385	switch (qtype) {
1386	case NI_QTYPE_NOOP:
1387		nni6->ni_code = ICMP6_NI_SUCCESS;
1388		nni6->ni_flags = 0;
1389		break;
1390	case NI_QTYPE_SUPTYPES:
1391	{
1392		u_int32_t v;
1393		nni6->ni_code = ICMP6_NI_SUCCESS;
1394		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1395		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1396		v = (u_int32_t)htonl(0x0000000f);
1397		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1398		break;
1399	}
1400	case NI_QTYPE_FQDN:
1401		nni6->ni_code = ICMP6_NI_SUCCESS;
1402		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1403		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1404		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1405		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1406		/*
1407		 * XXX do we really have FQDN in variable "hostname"?
1408		 */
1409		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1410		if (n->m_next == NULL)
1411			goto bad;
1412		/* XXX we assume that n->m_next is not a chain */
1413		if (n->m_next->m_next != NULL)
1414			goto bad;
1415		n->m_pkthdr.len += n->m_next->m_len;
1416		break;
1417	case NI_QTYPE_NODEADDR:
1418	{
1419		int lenlim, copied;
1420
1421		nni6->ni_code = ICMP6_NI_SUCCESS;
1422		n->m_pkthdr.len = n->m_len =
1423		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1424		lenlim = M_TRAILINGSPACE(n);
1425		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1426		/* XXX: reset mbuf length */
1427		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1428		    sizeof(struct icmp6_nodeinfo) + copied;
1429		break;
1430	}
1431	default:
1432		break;		/* XXX impossible! */
1433	}
1434
1435	nni6->ni_type = ICMP6_NI_REPLY;
1436	m_freem(m);
1437	return (n);
1438
1439  bad:
1440	m_freem(m);
1441	if (n)
1442		m_freem(n);
1443	return (NULL);
1444}
1445#undef hostnamelen
1446
1447/*
1448 * make a mbuf with DNS-encoded string.  no compression support.
1449 *
1450 * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1451 * treated as truncated name (two \0 at the end).  this is a wild guess.
1452 */
1453static struct mbuf *
1454ni6_nametodns(name, namelen, old)
1455	const char *name;
1456	int namelen;
1457	int old;	/* return pascal string if non-zero */
1458{
1459	struct mbuf *m;
1460	char *cp, *ep;
1461	const char *p, *q;
1462	int i, len, nterm;
1463
1464	if (old)
1465		len = namelen + 1;
1466	else
1467		len = MCLBYTES;
1468
1469	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1470	MGET(m, M_DONTWAIT, MT_DATA);
1471	if (m && len > MLEN) {
1472		MCLGET(m, M_DONTWAIT);
1473		if ((m->m_flags & M_EXT) == 0)
1474			goto fail;
1475	}
1476	if (!m)
1477		goto fail;
1478	m->m_next = NULL;
1479
1480	if (old) {
1481		m->m_len = len;
1482		*mtod(m, char *) = namelen;
1483		bcopy(name, mtod(m, char *) + 1, namelen);
1484		return m;
1485	} else {
1486		m->m_len = 0;
1487		cp = mtod(m, char *);
1488		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1489
1490		/* if not certain about my name, return empty buffer */
1491		if (namelen == 0)
1492			return m;
1493
1494		/*
1495		 * guess if it looks like shortened hostname, or FQDN.
1496		 * shortened hostname needs two trailing "\0".
1497		 */
1498		i = 0;
1499		for (p = name; p < name + namelen; p++) {
1500			if (*p && *p == '.')
1501				i++;
1502		}
1503		if (i < 2)
1504			nterm = 2;
1505		else
1506			nterm = 1;
1507
1508		p = name;
1509		while (cp < ep && p < name + namelen) {
1510			i = 0;
1511			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1512				i++;
1513			/* result does not fit into mbuf */
1514			if (cp + i + 1 >= ep)
1515				goto fail;
1516			/*
1517			 * DNS label length restriction, RFC1035 page 8.
1518			 * "i == 0" case is included here to avoid returning
1519			 * 0-length label on "foo..bar".
1520			 */
1521			if (i <= 0 || i >= 64)
1522				goto fail;
1523			*cp++ = i;
1524			bcopy(p, cp, i);
1525			cp += i;
1526			p = q;
1527			if (p < name + namelen && *p == '.')
1528				p++;
1529		}
1530		/* termination */
1531		if (cp + nterm >= ep)
1532			goto fail;
1533		while (nterm-- > 0)
1534			*cp++ = '\0';
1535		m->m_len = cp - mtod(m, char *);
1536		return m;
1537	}
1538
1539	panic("should not reach here");
1540	/* NOTREACHED */
1541
1542 fail:
1543	if (m)
1544		m_freem(m);
1545	return NULL;
1546}
1547
1548/*
1549 * check if two DNS-encoded string matches.  takes care of truncated
1550 * form (with \0\0 at the end).  no compression support.
1551 * XXX upper/lowercase match (see RFC2065)
1552 */
1553static int
1554ni6_dnsmatch(a, alen, b, blen)
1555	const char *a;
1556	int alen;
1557	const char *b;
1558	int blen;
1559{
1560	const char *a0, *b0;
1561	int l;
1562
1563	/* simplest case - need validation? */
1564	if (alen == blen && bcmp(a, b, alen) == 0)
1565		return 1;
1566
1567	a0 = a;
1568	b0 = b;
1569
1570	/* termination is mandatory */
1571	if (alen < 2 || blen < 2)
1572		return 0;
1573	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1574		return 0;
1575	alen--;
1576	blen--;
1577
1578	while (a - a0 < alen && b - b0 < blen) {
1579		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1580			return 0;
1581
1582		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1583			return 0;
1584		/* we don't support compression yet */
1585		if (a[0] >= 64 || b[0] >= 64)
1586			return 0;
1587
1588		/* truncated case */
1589		if (a[0] == 0 && a - a0 == alen - 1)
1590			return 1;
1591		if (b[0] == 0 && b - b0 == blen - 1)
1592			return 1;
1593		if (a[0] == 0 || b[0] == 0)
1594			return 0;
1595
1596		if (a[0] != b[0])
1597			return 0;
1598		l = a[0];
1599		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1600			return 0;
1601		if (bcmp(a + 1, b + 1, l) != 0)
1602			return 0;
1603
1604		a += 1 + l;
1605		b += 1 + l;
1606	}
1607
1608	if (a - a0 == alen && b - b0 == blen)
1609		return 1;
1610	else
1611		return 0;
1612}
1613
1614/*
1615 * calculate the number of addresses to be returned in the node info reply.
1616 */
1617static int
1618ni6_addrs(ni6, m, ifpp, subj)
1619	struct icmp6_nodeinfo *ni6;
1620	struct mbuf *m;
1621	struct ifnet **ifpp;
1622	struct in6_addr *subj;
1623{
1624	struct ifnet *ifp;
1625	struct in6_ifaddr *ifa6;
1626	struct ifaddr *ifa;
1627	int addrs = 0, addrsofif, iffound = 0;
1628	int niflags = ni6->ni_flags;
1629
1630	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1631		switch (ni6->ni_code) {
1632		case ICMP6_NI_SUBJ_IPV6:
1633			if (subj == NULL) /* must be impossible... */
1634				return (0);
1635			break;
1636		default:
1637			/*
1638			 * XXX: we only support IPv6 subject address for
1639			 * this Qtype.
1640			 */
1641			return (0);
1642		}
1643	}
1644
1645	IFNET_RLOCK();
1646	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1647		addrsofif = 0;
1648		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1649			if (ifa->ifa_addr->sa_family != AF_INET6)
1650				continue;
1651			ifa6 = (struct in6_ifaddr *)ifa;
1652
1653			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1654			    IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1655				iffound = 1;
1656
1657			/*
1658			 * IPv4-mapped addresses can only be returned by a
1659			 * Node Information proxy, since they represent
1660			 * addresses of IPv4-only nodes, which perforce do
1661			 * not implement this protocol.
1662			 * [icmp-name-lookups-07, Section 5.4]
1663			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1664			 * this function at this moment.
1665			 */
1666
1667			/* What do we have to do about ::1? */
1668			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1669			case IPV6_ADDR_SCOPE_LINKLOCAL:
1670				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1671					continue;
1672				break;
1673			case IPV6_ADDR_SCOPE_SITELOCAL:
1674				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1675					continue;
1676				break;
1677			case IPV6_ADDR_SCOPE_GLOBAL:
1678				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1679					continue;
1680				break;
1681			default:
1682				continue;
1683			}
1684
1685			/*
1686			 * check if anycast is okay.
1687			 * XXX: just experimental.  not in the spec.
1688			 */
1689			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1690			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1691				continue; /* we need only unicast addresses */
1692			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1693			    (icmp6_nodeinfo & 4) == 0) {
1694				continue;
1695			}
1696			addrsofif++; /* count the address */
1697		}
1698		if (iffound) {
1699			*ifpp = ifp;
1700			IFNET_RUNLOCK();
1701			return (addrsofif);
1702		}
1703
1704		addrs += addrsofif;
1705	}
1706	IFNET_RUNLOCK();
1707
1708	return (addrs);
1709}
1710
1711static int
1712ni6_store_addrs(ni6, nni6, ifp0, resid)
1713	struct icmp6_nodeinfo *ni6, *nni6;
1714	struct ifnet *ifp0;
1715	int resid;
1716{
1717	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1718	struct in6_ifaddr *ifa6;
1719	struct ifaddr *ifa;
1720	struct ifnet *ifp_dep = NULL;
1721	int copied = 0, allow_deprecated = 0;
1722	u_char *cp = (u_char *)(nni6 + 1);
1723	int niflags = ni6->ni_flags;
1724	u_int32_t ltime;
1725
1726	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1727		return (0);	/* needless to copy */
1728
1729	IFNET_RLOCK();
1730  again:
1731
1732	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1733		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1734		     ifa = ifa->ifa_list.tqe_next) {
1735			if (ifa->ifa_addr->sa_family != AF_INET6)
1736				continue;
1737			ifa6 = (struct in6_ifaddr *)ifa;
1738
1739			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1740			    allow_deprecated == 0) {
1741				/*
1742				 * prefererred address should be put before
1743				 * deprecated addresses.
1744				 */
1745
1746				/* record the interface for later search */
1747				if (ifp_dep == NULL)
1748					ifp_dep = ifp;
1749
1750				continue;
1751			} else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1752			    allow_deprecated != 0)
1753				continue; /* we now collect deprecated addrs */
1754
1755			/* What do we have to do about ::1? */
1756			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1757			case IPV6_ADDR_SCOPE_LINKLOCAL:
1758				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1759					continue;
1760				break;
1761			case IPV6_ADDR_SCOPE_SITELOCAL:
1762				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1763					continue;
1764				break;
1765			case IPV6_ADDR_SCOPE_GLOBAL:
1766				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1767					continue;
1768				break;
1769			default:
1770				continue;
1771			}
1772
1773			/*
1774			 * check if anycast is okay.
1775			 * XXX: just experimental.  not in the spec.
1776			 */
1777			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1778			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1779				continue;
1780			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1781			    (icmp6_nodeinfo & 4) == 0) {
1782				continue;
1783			}
1784
1785			/* now we can copy the address */
1786			if (resid < sizeof(struct in6_addr) +
1787			    sizeof(u_int32_t)) {
1788				/*
1789				 * We give up much more copy.
1790				 * Set the truncate flag and return.
1791				 */
1792				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1793				IFNET_RUNLOCK();
1794				return (copied);
1795			}
1796
1797			/*
1798			 * Set the TTL of the address.
1799			 * The TTL value should be one of the following
1800			 * according to the specification:
1801			 *
1802			 * 1. The remaining lifetime of a DHCP lease on the
1803			 *    address, or
1804			 * 2. The remaining Valid Lifetime of a prefix from
1805			 *    which the address was derived through Stateless
1806			 *    Autoconfiguration.
1807			 *
1808			 * Note that we currently do not support stateful
1809			 * address configuration by DHCPv6, so the former
1810			 * case can't happen.
1811			 */
1812			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1813				ltime = ND6_INFINITE_LIFETIME;
1814			else {
1815				if (ifa6->ia6_lifetime.ia6t_expire >
1816				    time_second)
1817					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1818				else
1819					ltime = 0;
1820			}
1821
1822			bcopy(&ltime, cp, sizeof(u_int32_t));
1823			cp += sizeof(u_int32_t);
1824
1825			/* copy the address itself */
1826			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1827			    sizeof(struct in6_addr));
1828			in6_clearscope((struct in6_addr *)cp); /* XXX */
1829			cp += sizeof(struct in6_addr);
1830
1831			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1832			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1833		}
1834		if (ifp0)	/* we need search only on the specified IF */
1835			break;
1836	}
1837
1838	if (allow_deprecated == 0 && ifp_dep != NULL) {
1839		ifp = ifp_dep;
1840		allow_deprecated = 1;
1841
1842		goto again;
1843	}
1844
1845	IFNET_RUNLOCK();
1846
1847	return (copied);
1848}
1849
1850/*
1851 * XXX almost dup'ed code with rip6_input.
1852 */
1853static int
1854icmp6_rip6_input(mp, off)
1855	struct	mbuf **mp;
1856	int	off;
1857{
1858	struct mbuf *m = *mp;
1859	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1860	struct in6pcb *in6p;
1861	struct in6pcb *last = NULL;
1862	struct sockaddr_in6 fromsa;
1863	struct icmp6_hdr *icmp6;
1864	struct mbuf *opts = NULL;
1865
1866#ifndef PULLDOWN_TEST
1867	/* this is assumed to be safe. */
1868	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1869#else
1870	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1871	if (icmp6 == NULL) {
1872		/* m is already reclaimed */
1873		return (IPPROTO_DONE);
1874	}
1875#endif
1876
1877	/*
1878	 * XXX: the address may have embedded scope zone ID, which should be
1879	 * hidden from applications.
1880	 */
1881	bzero(&fromsa, sizeof(fromsa));
1882	fromsa.sin6_family = AF_INET6;
1883	fromsa.sin6_len = sizeof(struct sockaddr_in6);
1884	fromsa.sin6_addr = ip6->ip6_src;
1885	if (sa6_recoverscope(&fromsa)) {
1886		m_freem(m);
1887		return (IPPROTO_DONE);
1888	}
1889
1890	INP_INFO_RLOCK(&ripcbinfo);
1891	LIST_FOREACH(in6p, &ripcb, inp_list) {
1892		INP_LOCK(in6p);
1893		if ((in6p->inp_vflag & INP_IPV6) == 0) {
1894	docontinue:
1895			INP_UNLOCK(in6p);
1896			continue;
1897		}
1898		if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1899			goto docontinue;
1900		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1901		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1902			goto docontinue;
1903		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1904		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1905			goto docontinue;
1906		if (in6p->in6p_icmp6filt
1907		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1908				 in6p->in6p_icmp6filt))
1909			goto docontinue;
1910		if (last) {
1911			struct	mbuf *n = NULL;
1912
1913			/*
1914			 * Recent network drivers tend to allocate a single
1915			 * mbuf cluster, rather than to make a couple of
1916			 * mbufs without clusters.  Also, since the IPv6 code
1917			 * path tries to avoid m_pullup(), it is highly
1918			 * probable that we still have an mbuf cluster here
1919			 * even though the necessary length can be stored in an
1920			 * mbuf's internal buffer.
1921			 * Meanwhile, the default size of the receive socket
1922			 * buffer for raw sockets is not so large.  This means
1923			 * the possibility of packet loss is relatively higher
1924			 * than before.  To avoid this scenario, we copy the
1925			 * received data to a separate mbuf that does not use
1926			 * a cluster, if possible.
1927			 * XXX: it is better to copy the data after stripping
1928			 * intermediate headers.
1929			 */
1930			if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1931			    m->m_len <= MHLEN) {
1932				MGET(n, M_DONTWAIT, m->m_type);
1933				if (n != NULL) {
1934					if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1935						bcopy(m->m_data, n->m_data,
1936						      m->m_len);
1937						n->m_len = m->m_len;
1938					} else {
1939						m_free(n);
1940						n = NULL;
1941					}
1942				}
1943			}
1944			if (n != NULL ||
1945			    (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1946				if (last->in6p_flags & IN6P_CONTROLOPTS)
1947					ip6_savecontrol(last, n, &opts);
1948				/* strip intermediate headers */
1949				m_adj(n, off);
1950				if (sbappendaddr(&last->in6p_socket->so_rcv,
1951				    (struct sockaddr *)&fromsa, n, opts)
1952				    == 0) {
1953					/* should notify about lost packet */
1954					m_freem(n);
1955					if (opts) {
1956						m_freem(opts);
1957					}
1958				} else
1959					sorwakeup(last->in6p_socket);
1960				opts = NULL;
1961			}
1962			INP_UNLOCK(last);
1963		}
1964		last = in6p;
1965	}
1966	if (last) {
1967		if (last->in6p_flags & IN6P_CONTROLOPTS)
1968			ip6_savecontrol(last, m, &opts);
1969		/* strip intermediate headers */
1970		m_adj(m, off);
1971
1972		/* avoid using mbuf clusters if possible (see above) */
1973		if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1974		    m->m_len <= MHLEN) {
1975			struct mbuf *n;
1976
1977			MGET(n, M_DONTWAIT, m->m_type);
1978			if (n != NULL) {
1979				if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1980					bcopy(m->m_data, n->m_data, m->m_len);
1981					n->m_len = m->m_len;
1982
1983					m_freem(m);
1984					m = n;
1985				} else {
1986					m_freem(n);
1987					n = NULL;
1988				}
1989			}
1990		}
1991		if (sbappendaddr(&last->in6p_socket->so_rcv,
1992		    (struct sockaddr *)&fromsa, m, opts) == 0) {
1993			m_freem(m);
1994			if (opts)
1995				m_freem(opts);
1996		} else
1997			sorwakeup(last->in6p_socket);
1998		INP_UNLOCK(last);
1999	} else {
2000		m_freem(m);
2001		ip6stat.ip6s_delivered--;
2002	}
2003	INP_INFO_RUNLOCK(&ripcbinfo);
2004	return IPPROTO_DONE;
2005}
2006
2007/*
2008 * Reflect the ip6 packet back to the source.
2009 * OFF points to the icmp6 header, counted from the top of the mbuf.
2010 */
2011void
2012icmp6_reflect(m, off)
2013	struct	mbuf *m;
2014	size_t off;
2015{
2016	struct ip6_hdr *ip6;
2017	struct icmp6_hdr *icmp6;
2018	struct in6_ifaddr *ia;
2019	int plen;
2020	int type, code;
2021	struct ifnet *outif = NULL;
2022	struct in6_addr origdst, *src = NULL;
2023
2024	/* too short to reflect */
2025	if (off < sizeof(struct ip6_hdr)) {
2026		nd6log((LOG_DEBUG,
2027		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2028		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2029		    __FILE__, __LINE__));
2030		goto bad;
2031	}
2032
2033	/*
2034	 * If there are extra headers between IPv6 and ICMPv6, strip
2035	 * off that header first.
2036	 */
2037#ifdef DIAGNOSTIC
2038	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2039		panic("assumption failed in icmp6_reflect");
2040#endif
2041	if (off > sizeof(struct ip6_hdr)) {
2042		size_t l;
2043		struct ip6_hdr nip6;
2044
2045		l = off - sizeof(struct ip6_hdr);
2046		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2047		m_adj(m, l);
2048		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2049		if (m->m_len < l) {
2050			if ((m = m_pullup(m, l)) == NULL)
2051				return;
2052		}
2053		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2054	} else /* off == sizeof(struct ip6_hdr) */ {
2055		size_t l;
2056		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2057		if (m->m_len < l) {
2058			if ((m = m_pullup(m, l)) == NULL)
2059				return;
2060		}
2061	}
2062	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2063	ip6 = mtod(m, struct ip6_hdr *);
2064	ip6->ip6_nxt = IPPROTO_ICMPV6;
2065	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2066	type = icmp6->icmp6_type; /* keep type for statistics */
2067	code = icmp6->icmp6_code; /* ditto. */
2068
2069	origdst = ip6->ip6_dst;
2070	/*
2071	 * ip6_input() drops a packet if its src is multicast.
2072	 * So, the src is never multicast.
2073	 */
2074	ip6->ip6_dst = ip6->ip6_src;
2075
2076	/*
2077	 * If the incoming packet was addressed directly to us (i.e. unicast),
2078	 * use dst as the src for the reply.
2079	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2080	 * (for example) when we encounter an error while forwarding procedure
2081	 * destined to a duplicated address of ours.
2082	 * Note that ip6_getdstifaddr() may fail if we are in an error handling
2083	 * procedure of an outgoing packet of our own, in which case we need
2084	 * to search in the ifaddr list.
2085	 */
2086	if (!IN6_IS_ADDR_MULTICAST(&origdst)) {
2087		if ((ia = ip6_getdstifaddr(m))) {
2088			if (!(ia->ia6_flags &
2089			    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)))
2090				src = &ia->ia_addr.sin6_addr;
2091		} else {
2092			struct sockaddr_in6 d;
2093
2094			bzero(&d, sizeof(d));
2095			d.sin6_family = AF_INET6;
2096			d.sin6_len = sizeof(d);
2097			d.sin6_addr = origdst;
2098			ia = (struct in6_ifaddr *)
2099			    ifa_ifwithaddr((struct sockaddr *)&d);
2100			if (ia &&
2101			    !(ia->ia6_flags &
2102			    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2103				src = &ia->ia_addr.sin6_addr;
2104			}
2105		}
2106	}
2107
2108	if (src == NULL) {
2109		int e;
2110		struct sockaddr_in6 sin6;
2111		struct route_in6 ro;
2112
2113		/*
2114		 * This case matches to multicasts, our anycast, or unicasts
2115		 * that we do not own.  Select a source address based on the
2116		 * source address of the erroneous packet.
2117		 */
2118		bzero(&sin6, sizeof(sin6));
2119		sin6.sin6_family = AF_INET6;
2120		sin6.sin6_len = sizeof(sin6);
2121		sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */
2122
2123		bzero(&ro, sizeof(ro));
2124		src = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &e);
2125		if (ro.ro_rt)
2126			RTFREE(ro.ro_rt); /* XXX: we could use this */
2127		if (src == NULL) {
2128			nd6log((LOG_DEBUG,
2129			    "icmp6_reflect: source can't be determined: "
2130			    "dst=%s, error=%d\n",
2131			    ip6_sprintf(&sin6.sin6_addr), e));
2132			goto bad;
2133		}
2134	}
2135
2136	ip6->ip6_src = *src;
2137	ip6->ip6_flow = 0;
2138	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2139	ip6->ip6_vfc |= IPV6_VERSION;
2140	ip6->ip6_nxt = IPPROTO_ICMPV6;
2141	if (outif)
2142		ip6->ip6_hlim = ND_IFINFO(outif)->chlim;
2143	else if (m->m_pkthdr.rcvif) {
2144		/* XXX: This may not be the outgoing interface */
2145		ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2146	} else
2147		ip6->ip6_hlim = ip6_defhlim;
2148
2149	icmp6->icmp6_cksum = 0;
2150	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2151	    sizeof(struct ip6_hdr), plen);
2152
2153	/*
2154	 * XXX option handling
2155	 */
2156
2157	m->m_flags &= ~(M_BCAST|M_MCAST);
2158
2159	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2160	if (outif)
2161		icmp6_ifoutstat_inc(outif, type, code);
2162
2163	return;
2164
2165 bad:
2166	m_freem(m);
2167	return;
2168}
2169
2170void
2171icmp6_fasttimo()
2172{
2173
2174	mld6_fasttimeo();
2175}
2176
2177static const char *
2178icmp6_redirect_diag(src6, dst6, tgt6)
2179	struct in6_addr *src6;
2180	struct in6_addr *dst6;
2181	struct in6_addr *tgt6;
2182{
2183	static char buf[1024];
2184	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2185	    ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2186	return buf;
2187}
2188
2189void
2190icmp6_redirect_input(m, off)
2191	struct mbuf *m;
2192	int off;
2193{
2194	struct ifnet *ifp = m->m_pkthdr.rcvif;
2195	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2196	struct nd_redirect *nd_rd;
2197	int icmp6len = ntohs(ip6->ip6_plen);
2198	char *lladdr = NULL;
2199	int lladdrlen = 0;
2200	u_char *redirhdr = NULL;
2201	int redirhdrlen = 0;
2202	struct rtentry *rt = NULL;
2203	int is_router;
2204	int is_onlink;
2205	struct in6_addr src6 = ip6->ip6_src;
2206	struct in6_addr redtgt6;
2207	struct in6_addr reddst6;
2208	union nd_opts ndopts;
2209
2210	if (!m || !ifp)
2211		return;
2212
2213	/* XXX if we are router, we don't update route by icmp6 redirect */
2214	if (ip6_forwarding)
2215		goto freeit;
2216	if (!icmp6_rediraccept)
2217		goto freeit;
2218
2219#ifndef PULLDOWN_TEST
2220	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2221	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2222#else
2223	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2224	if (nd_rd == NULL) {
2225		icmp6stat.icp6s_tooshort++;
2226		return;
2227	}
2228#endif
2229	redtgt6 = nd_rd->nd_rd_target;
2230	reddst6 = nd_rd->nd_rd_dst;
2231
2232	if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2233	    in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2234		goto freeit;
2235	}
2236
2237	/* validation */
2238	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2239		nd6log((LOG_ERR,
2240		    "ICMP6 redirect sent from %s rejected; "
2241		    "must be from linklocal\n",
2242		    ip6_sprintf(&src6)));
2243		goto bad;
2244	}
2245	if (ip6->ip6_hlim != 255) {
2246		nd6log((LOG_ERR,
2247		    "ICMP6 redirect sent from %s rejected; "
2248		    "hlim=%d (must be 255)\n",
2249		    ip6_sprintf(&src6), ip6->ip6_hlim));
2250		goto bad;
2251	}
2252    {
2253	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2254	struct sockaddr_in6 sin6;
2255	struct in6_addr *gw6;
2256
2257	bzero(&sin6, sizeof(sin6));
2258	sin6.sin6_family = AF_INET6;
2259	sin6.sin6_len = sizeof(struct sockaddr_in6);
2260	bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2261	rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL);
2262	if (rt) {
2263		if (rt->rt_gateway == NULL ||
2264		    rt->rt_gateway->sa_family != AF_INET6) {
2265			nd6log((LOG_ERR,
2266			    "ICMP6 redirect rejected; no route "
2267			    "with inet6 gateway found for redirect dst: %s\n",
2268			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2269			RTFREE_LOCKED(rt);
2270			goto bad;
2271		}
2272
2273		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2274		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2275			nd6log((LOG_ERR,
2276			    "ICMP6 redirect rejected; "
2277			    "not equal to gw-for-src=%s (must be same): "
2278			    "%s\n",
2279			    ip6_sprintf(gw6),
2280			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2281			RTFREE_LOCKED(rt);
2282			goto bad;
2283		}
2284	} else {
2285		nd6log((LOG_ERR,
2286		    "ICMP6 redirect rejected; "
2287		    "no route found for redirect dst: %s\n",
2288		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2289		goto bad;
2290	}
2291	RTFREE_LOCKED(rt);
2292	rt = NULL;
2293    }
2294	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2295		nd6log((LOG_ERR,
2296		    "ICMP6 redirect rejected; "
2297		    "redirect dst must be unicast: %s\n",
2298		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2299		goto bad;
2300	}
2301
2302	is_router = is_onlink = 0;
2303	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2304		is_router = 1;	/* router case */
2305	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2306		is_onlink = 1;	/* on-link destination case */
2307	if (!is_router && !is_onlink) {
2308		nd6log((LOG_ERR,
2309		    "ICMP6 redirect rejected; "
2310		    "neither router case nor onlink case: %s\n",
2311		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2312		goto bad;
2313	}
2314	/* validation passed */
2315
2316	icmp6len -= sizeof(*nd_rd);
2317	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2318	if (nd6_options(&ndopts) < 0) {
2319		nd6log((LOG_INFO, "icmp6_redirect_input: "
2320		    "invalid ND option, rejected: %s\n",
2321		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2322		/* nd6_options have incremented stats */
2323		goto freeit;
2324	}
2325
2326	if (ndopts.nd_opts_tgt_lladdr) {
2327		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2328		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2329	}
2330
2331	if (ndopts.nd_opts_rh) {
2332		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2333		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2334	}
2335
2336	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2337		nd6log((LOG_INFO,
2338		    "icmp6_redirect_input: lladdrlen mismatch for %s "
2339		    "(if %d, icmp6 packet %d): %s\n",
2340		    ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2341		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2342		goto bad;
2343	}
2344
2345	/* RFC 2461 8.3 */
2346	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2347	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2348
2349	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2350		/* perform rtredirect */
2351		struct sockaddr_in6 sdst;
2352		struct sockaddr_in6 sgw;
2353		struct sockaddr_in6 ssrc;
2354
2355		bzero(&sdst, sizeof(sdst));
2356		bzero(&sgw, sizeof(sgw));
2357		bzero(&ssrc, sizeof(ssrc));
2358		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2359		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2360			sizeof(struct sockaddr_in6);
2361		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2362		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2363		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2364		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2365		    (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2366		    (struct sockaddr *)&ssrc);
2367	}
2368	/* finally update cached route in each socket via pfctlinput */
2369    {
2370	struct sockaddr_in6 sdst;
2371
2372	bzero(&sdst, sizeof(sdst));
2373	sdst.sin6_family = AF_INET6;
2374	sdst.sin6_len = sizeof(struct sockaddr_in6);
2375	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2376	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2377#if defined(IPSEC) || defined(FAST_IPSEC)
2378	key_sa_routechange((struct sockaddr *)&sdst);
2379#endif
2380    }
2381
2382 freeit:
2383	m_freem(m);
2384	return;
2385
2386 bad:
2387	icmp6stat.icp6s_badredirect++;
2388	m_freem(m);
2389}
2390
2391void
2392icmp6_redirect_output(m0, rt)
2393	struct mbuf *m0;
2394	struct rtentry *rt;
2395{
2396	struct ifnet *ifp;	/* my outgoing interface */
2397	struct in6_addr *ifp_ll6;
2398	struct in6_addr *router_ll6;
2399	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2400	struct mbuf *m = NULL;	/* newly allocated one */
2401	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2402	struct nd_redirect *nd_rd;
2403	size_t maxlen;
2404	u_char *p;
2405	struct ifnet *outif = NULL;
2406	struct sockaddr_in6 src_sa;
2407
2408	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2409
2410	/* if we are not router, we don't send icmp6 redirect */
2411	if (!ip6_forwarding || ip6_accept_rtadv)
2412		goto fail;
2413
2414	/* sanity check */
2415	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2416		goto fail;
2417
2418	/*
2419	 * Address check:
2420	 *  the source address must identify a neighbor, and
2421	 *  the destination address must not be a multicast address
2422	 *  [RFC 2461, sec 8.2]
2423	 */
2424	sip6 = mtod(m0, struct ip6_hdr *);
2425	bzero(&src_sa, sizeof(src_sa));
2426	src_sa.sin6_family = AF_INET6;
2427	src_sa.sin6_len = sizeof(src_sa);
2428	src_sa.sin6_addr = sip6->ip6_src;
2429	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2430		goto fail;
2431	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2432		goto fail;	/* what should we do here? */
2433
2434	/* rate limit */
2435	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2436		goto fail;
2437
2438	/*
2439	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2440	 * we almost always ask for an mbuf cluster for simplicity.
2441	 * (MHLEN < IPV6_MMTU is almost always true)
2442	 */
2443#if IPV6_MMTU >= MCLBYTES
2444# error assumption failed about IPV6_MMTU and MCLBYTES
2445#endif
2446	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2447	if (m && IPV6_MMTU >= MHLEN)
2448		MCLGET(m, M_DONTWAIT);
2449	if (!m)
2450		goto fail;
2451	m->m_pkthdr.rcvif = NULL;
2452	m->m_len = 0;
2453	maxlen = M_TRAILINGSPACE(m);
2454	maxlen = min(IPV6_MMTU, maxlen);
2455	/* just for safety */
2456	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2457	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2458		goto fail;
2459	}
2460
2461	{
2462		/* get ip6 linklocal address for ifp(my outgoing interface). */
2463		struct in6_ifaddr *ia;
2464		if ((ia = in6ifa_ifpforlinklocal(ifp,
2465						 IN6_IFF_NOTREADY|
2466						 IN6_IFF_ANYCAST)) == NULL)
2467			goto fail;
2468		ifp_ll6 = &ia->ia_addr.sin6_addr;
2469	}
2470
2471	/* get ip6 linklocal address for the router. */
2472	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2473		struct sockaddr_in6 *sin6;
2474		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2475		router_ll6 = &sin6->sin6_addr;
2476		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2477			router_ll6 = (struct in6_addr *)NULL;
2478	} else
2479		router_ll6 = (struct in6_addr *)NULL;
2480
2481	/* ip6 */
2482	ip6 = mtod(m, struct ip6_hdr *);
2483	ip6->ip6_flow = 0;
2484	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2485	ip6->ip6_vfc |= IPV6_VERSION;
2486	/* ip6->ip6_plen will be set later */
2487	ip6->ip6_nxt = IPPROTO_ICMPV6;
2488	ip6->ip6_hlim = 255;
2489	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2490	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2491	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2492
2493	/* ND Redirect */
2494	nd_rd = (struct nd_redirect *)(ip6 + 1);
2495	nd_rd->nd_rd_type = ND_REDIRECT;
2496	nd_rd->nd_rd_code = 0;
2497	nd_rd->nd_rd_reserved = 0;
2498	if (rt->rt_flags & RTF_GATEWAY) {
2499		/*
2500		 * nd_rd->nd_rd_target must be a link-local address in
2501		 * better router cases.
2502		 */
2503		if (!router_ll6)
2504			goto fail;
2505		bcopy(router_ll6, &nd_rd->nd_rd_target,
2506		    sizeof(nd_rd->nd_rd_target));
2507		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2508		    sizeof(nd_rd->nd_rd_dst));
2509	} else {
2510		/* make sure redtgt == reddst */
2511		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2512		    sizeof(nd_rd->nd_rd_target));
2513		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2514		    sizeof(nd_rd->nd_rd_dst));
2515	}
2516
2517	p = (u_char *)(nd_rd + 1);
2518
2519	if (!router_ll6)
2520		goto nolladdropt;
2521
2522	{
2523		/* target lladdr option */
2524		struct rtentry *rt_router = NULL;
2525		int len;
2526		struct sockaddr_dl *sdl;
2527		struct nd_opt_hdr *nd_opt;
2528		char *lladdr;
2529
2530		rt_router = nd6_lookup(router_ll6, 0, ifp);
2531		if (!rt_router)
2532			goto nolladdropt;
2533		len = sizeof(*nd_opt) + ifp->if_addrlen;
2534		len = (len + 7) & ~7;	/* round by 8 */
2535		/* safety check */
2536		if (len + (p - (u_char *)ip6) > maxlen)
2537			goto nolladdropt;
2538		if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2539		    (rt_router->rt_flags & RTF_LLINFO) &&
2540		    (rt_router->rt_gateway->sa_family == AF_LINK) &&
2541		    (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2542		    sdl->sdl_alen) {
2543			nd_opt = (struct nd_opt_hdr *)p;
2544			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2545			nd_opt->nd_opt_len = len >> 3;
2546			lladdr = (char *)(nd_opt + 1);
2547			bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2548			p += len;
2549		}
2550	}
2551nolladdropt:;
2552
2553	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2554
2555	/* just to be safe */
2556#ifdef M_DECRYPTED	/*not openbsd*/
2557	if (m0->m_flags & M_DECRYPTED)
2558		goto noredhdropt;
2559#endif
2560	if (p - (u_char *)ip6 > maxlen)
2561		goto noredhdropt;
2562
2563	{
2564		/* redirected header option */
2565		int len;
2566		struct nd_opt_rd_hdr *nd_opt_rh;
2567
2568		/*
2569		 * compute the maximum size for icmp6 redirect header option.
2570		 * XXX room for auth header?
2571		 */
2572		len = maxlen - (p - (u_char *)ip6);
2573		len &= ~7;
2574
2575		/* This is just for simplicity. */
2576		if (m0->m_pkthdr.len != m0->m_len) {
2577			if (m0->m_next) {
2578				m_freem(m0->m_next);
2579				m0->m_next = NULL;
2580			}
2581			m0->m_pkthdr.len = m0->m_len;
2582		}
2583
2584		/*
2585		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2586		 * about padding/truncate rule for the original IP packet.
2587		 * From the discussion on IPv6imp in Feb 1999,
2588		 * the consensus was:
2589		 * - "attach as much as possible" is the goal
2590		 * - pad if not aligned (original size can be guessed by
2591		 *   original ip6 header)
2592		 * Following code adds the padding if it is simple enough,
2593		 * and truncates if not.
2594		 */
2595		if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2596			panic("assumption failed in %s:%d", __FILE__,
2597			    __LINE__);
2598
2599		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2600			/* not enough room, truncate */
2601			m0->m_pkthdr.len = m0->m_len = len -
2602			    sizeof(*nd_opt_rh);
2603		} else {
2604			/* enough room, pad or truncate */
2605			size_t extra;
2606
2607			extra = m0->m_pkthdr.len % 8;
2608			if (extra) {
2609				/* pad if easy enough, truncate if not */
2610				if (8 - extra <= M_TRAILINGSPACE(m0)) {
2611					/* pad */
2612					m0->m_len += (8 - extra);
2613					m0->m_pkthdr.len += (8 - extra);
2614				} else {
2615					/* truncate */
2616					m0->m_pkthdr.len -= extra;
2617					m0->m_len -= extra;
2618				}
2619			}
2620			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2621			m0->m_pkthdr.len = m0->m_len = len -
2622			    sizeof(*nd_opt_rh);
2623		}
2624
2625		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2626		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2627		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2628		nd_opt_rh->nd_opt_rh_len = len >> 3;
2629		p += sizeof(*nd_opt_rh);
2630		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2631
2632		/* connect m0 to m */
2633		m_tag_delete_chain(m0, NULL);
2634		m0->m_flags &= ~M_PKTHDR;
2635		m->m_next = m0;
2636		m->m_pkthdr.len = m->m_len + m0->m_len;
2637		m0 = NULL;
2638	}
2639noredhdropt:;
2640	if (m0) {
2641		m_freem(m0);
2642		m0 = NULL;
2643	}
2644
2645	/* XXX: clear embedded link IDs in the inner header */
2646	in6_clearscope(&sip6->ip6_src);
2647	in6_clearscope(&sip6->ip6_dst);
2648	in6_clearscope(&nd_rd->nd_rd_target);
2649	in6_clearscope(&nd_rd->nd_rd_dst);
2650
2651	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2652
2653	nd_rd->nd_rd_cksum = 0;
2654	nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2655	    sizeof(*ip6), ntohs(ip6->ip6_plen));
2656
2657	/* send the packet to outside... */
2658	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2659	if (outif) {
2660		icmp6_ifstat_inc(outif, ifs6_out_msg);
2661		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2662	}
2663	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2664
2665	return;
2666
2667fail:
2668	if (m)
2669		m_freem(m);
2670	if (m0)
2671		m_freem(m0);
2672}
2673
2674/*
2675 * ICMPv6 socket option processing.
2676 */
2677int
2678icmp6_ctloutput(so, sopt)
2679	struct socket *so;
2680	struct sockopt *sopt;
2681{
2682	int error = 0;
2683	int optlen;
2684	struct inpcb *inp = sotoinpcb(so);
2685	int level, op, optname;
2686
2687	if (sopt) {
2688		level = sopt->sopt_level;
2689		op = sopt->sopt_dir;
2690		optname = sopt->sopt_name;
2691		optlen = sopt->sopt_valsize;
2692	} else
2693		level = op = optname = optlen = 0;
2694
2695	if (level != IPPROTO_ICMPV6) {
2696		return EINVAL;
2697	}
2698
2699	switch (op) {
2700	case PRCO_SETOPT:
2701		switch (optname) {
2702		case ICMP6_FILTER:
2703		    {
2704			struct icmp6_filter *p;
2705
2706			if (optlen != sizeof(*p)) {
2707				error = EMSGSIZE;
2708				break;
2709			}
2710			if (inp->in6p_icmp6filt == NULL) {
2711				error = EINVAL;
2712				break;
2713			}
2714			error = sooptcopyin(sopt, inp->in6p_icmp6filt, optlen,
2715				optlen);
2716			break;
2717		    }
2718
2719		default:
2720			error = ENOPROTOOPT;
2721			break;
2722		}
2723		break;
2724
2725	case PRCO_GETOPT:
2726		switch (optname) {
2727		case ICMP6_FILTER:
2728		    {
2729			if (inp->in6p_icmp6filt == NULL) {
2730				error = EINVAL;
2731				break;
2732			}
2733			error = sooptcopyout(sopt, inp->in6p_icmp6filt,
2734				sizeof(struct icmp6_filter));
2735			break;
2736		    }
2737
2738		default:
2739			error = ENOPROTOOPT;
2740			break;
2741		}
2742		break;
2743	}
2744
2745	return (error);
2746}
2747
2748/*
2749 * Perform rate limit check.
2750 * Returns 0 if it is okay to send the icmp6 packet.
2751 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2752 * limitation.
2753 *
2754 * XXX per-destination/type check necessary?
2755 */
2756static int
2757icmp6_ratelimit(dst, type, code)
2758	const struct in6_addr *dst;	/* not used at this moment */
2759	const int type;			/* not used at this moment */
2760	const int code;			/* not used at this moment */
2761{
2762	int ret;
2763
2764	ret = 0;	/* okay to send */
2765
2766	/* PPS limit */
2767	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2768	    icmp6errppslim)) {
2769		/* The packet is subject to rate limit */
2770		ret++;
2771	}
2772
2773	return ret;
2774}
2775