icmp6.c revision 171259
1/*	$FreeBSD: head/sys/netinet6/icmp6.c 171259 2007-07-05 16:23:49Z delphij $	*/
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 <netipsec/ipsec.h>
104#include <netipsec/key.h>
105#endif
106
107extern struct domain inet6domain;
108
109struct icmp6stat icmp6stat;
110
111extern struct inpcbinfo ripcbinfo;
112extern struct inpcbhead ripcb;
113extern int icmp6errppslim;
114static int icmp6errpps_count = 0;
115static struct timeval icmp6errppslim_last;
116extern int icmp6_nodeinfo;
117
118static void icmp6_errcount __P((struct icmp6errstat *, int, int));
119static int icmp6_rip6_input __P((struct mbuf **, int));
120static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int));
121static const char *icmp6_redirect_diag __P((struct in6_addr *,
122	struct in6_addr *, struct in6_addr *));
123static struct mbuf *ni6_input __P((struct mbuf *, int));
124static struct mbuf *ni6_nametodns __P((const char *, int, int));
125static int ni6_dnsmatch __P((const char *, int, const char *, int));
126static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
127			  struct ifnet **, struct in6_addr *));
128static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
129				struct ifnet *, int));
130static int icmp6_notify_error __P((struct mbuf **, int, int, int));
131
132
133void
134icmp6_init(void)
135{
136
137	mld6_init();
138}
139
140static void
141icmp6_errcount(struct icmp6errstat *stat, int type, int code)
142{
143	switch (type) {
144	case ICMP6_DST_UNREACH:
145		switch (code) {
146		case ICMP6_DST_UNREACH_NOROUTE:
147			stat->icp6errs_dst_unreach_noroute++;
148			return;
149		case ICMP6_DST_UNREACH_ADMIN:
150			stat->icp6errs_dst_unreach_admin++;
151			return;
152		case ICMP6_DST_UNREACH_BEYONDSCOPE:
153			stat->icp6errs_dst_unreach_beyondscope++;
154			return;
155		case ICMP6_DST_UNREACH_ADDR:
156			stat->icp6errs_dst_unreach_addr++;
157			return;
158		case ICMP6_DST_UNREACH_NOPORT:
159			stat->icp6errs_dst_unreach_noport++;
160			return;
161		}
162		break;
163	case ICMP6_PACKET_TOO_BIG:
164		stat->icp6errs_packet_too_big++;
165		return;
166	case ICMP6_TIME_EXCEEDED:
167		switch (code) {
168		case ICMP6_TIME_EXCEED_TRANSIT:
169			stat->icp6errs_time_exceed_transit++;
170			return;
171		case ICMP6_TIME_EXCEED_REASSEMBLY:
172			stat->icp6errs_time_exceed_reassembly++;
173			return;
174		}
175		break;
176	case ICMP6_PARAM_PROB:
177		switch (code) {
178		case ICMP6_PARAMPROB_HEADER:
179			stat->icp6errs_paramprob_header++;
180			return;
181		case ICMP6_PARAMPROB_NEXTHEADER:
182			stat->icp6errs_paramprob_nextheader++;
183			return;
184		case ICMP6_PARAMPROB_OPTION:
185			stat->icp6errs_paramprob_option++;
186			return;
187		}
188		break;
189	case ND_REDIRECT:
190		stat->icp6errs_redirect++;
191		return;
192	}
193	stat->icp6errs_unknown++;
194}
195
196/*
197 * A wrapper function for icmp6_error() necessary when the erroneous packet
198 * may not contain enough scope zone information.
199 */
200void
201icmp6_error2(struct mbuf *m, int type, int code, int param,
202    struct ifnet *ifp)
203{
204	struct ip6_hdr *ip6;
205
206	if (ifp == NULL)
207		return;
208
209#ifndef PULLDOWN_TEST
210	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
211#else
212	if (m->m_len < sizeof(struct ip6_hdr)) {
213		m = m_pullup(m, sizeof(struct ip6_hdr));
214		if (m == NULL)
215			return;
216	}
217#endif
218
219	ip6 = mtod(m, struct ip6_hdr *);
220
221	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
222		return;
223	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
224		return;
225
226	icmp6_error(m, type, code, param);
227}
228
229/*
230 * Generate an error packet of type error in response to bad IP6 packet.
231 */
232void
233icmp6_error(struct mbuf *m, int type, int code, int param)
234{
235	struct ip6_hdr *oip6, *nip6;
236	struct icmp6_hdr *icmp6;
237	u_int preplen;
238	int off;
239	int nxt;
240
241	icmp6stat.icp6s_error++;
242
243	/* count per-type-code statistics */
244	icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
245
246#ifdef M_DECRYPTED	/*not openbsd*/
247	if (m->m_flags & M_DECRYPTED) {
248		icmp6stat.icp6s_canterror++;
249		goto freeit;
250	}
251#endif
252
253#ifndef PULLDOWN_TEST
254	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
255#else
256	if (m->m_len < sizeof(struct ip6_hdr)) {
257		m = m_pullup(m, sizeof(struct ip6_hdr));
258		if (m == NULL)
259			return;
260	}
261#endif
262	oip6 = mtod(m, struct ip6_hdr *);
263
264	/*
265	 * If the destination address of the erroneous packet is a multicast
266	 * address, or the packet was sent using link-layer multicast,
267	 * we should basically suppress sending an error (RFC 2463, Section
268	 * 2.4).
269	 * We have two exceptions (the item e.2 in that section):
270	 * - the Pakcet Too Big message can be sent for path MTU discovery.
271	 * - the Parameter Problem Message that can be allowed an icmp6 error
272	 *   in the option type field.  This check has been done in
273	 *   ip6_unknown_opt(), so we can just check the type and code.
274	 */
275	if ((m->m_flags & (M_BCAST|M_MCAST) ||
276	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
277	    (type != ICMP6_PACKET_TOO_BIG &&
278	     (type != ICMP6_PARAM_PROB ||
279	      code != ICMP6_PARAMPROB_OPTION)))
280		goto freeit;
281
282	/*
283	 * RFC 2463, 2.4 (e.5): source address check.
284	 * XXX: the case of anycast source?
285	 */
286	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
287	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
288		goto freeit;
289
290	/*
291	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
292	 * don't do it.
293	 */
294	nxt = -1;
295	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
296	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
297		struct icmp6_hdr *icp;
298
299#ifndef PULLDOWN_TEST
300		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
301		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
302#else
303		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
304			sizeof(*icp));
305		if (icp == NULL) {
306			icmp6stat.icp6s_tooshort++;
307			return;
308		}
309#endif
310		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
311		    icp->icmp6_type == ND_REDIRECT) {
312			/*
313			 * ICMPv6 error
314			 * Special case: for redirect (which is
315			 * informational) we must not send icmp6 error.
316			 */
317			icmp6stat.icp6s_canterror++;
318			goto freeit;
319		} else {
320			/* ICMPv6 informational - send the error */
321		}
322	} else {
323		/* non-ICMPv6 - send the error */
324	}
325
326	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
327
328	/* Finally, do rate limitation check. */
329	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
330		icmp6stat.icp6s_toofreq++;
331		goto freeit;
332	}
333
334	/*
335	 * OK, ICMP6 can be generated.
336	 */
337
338	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
339		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
340
341	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
342	M_PREPEND(m, preplen, M_DONTWAIT);
343	if (m && m->m_len < preplen)
344		m = m_pullup(m, preplen);
345	if (m == NULL) {
346		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
347		return;
348	}
349
350	nip6 = mtod(m, struct ip6_hdr *);
351	nip6->ip6_src  = oip6->ip6_src;
352	nip6->ip6_dst  = oip6->ip6_dst;
353
354	in6_clearscope(&oip6->ip6_src);
355	in6_clearscope(&oip6->ip6_dst);
356
357	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
358	icmp6->icmp6_type = type;
359	icmp6->icmp6_code = code;
360	icmp6->icmp6_pptr = htonl((u_int32_t)param);
361
362	/*
363	 * icmp6_reflect() is designed to be in the input path.
364	 * icmp6_error() can be called from both input and output path,
365	 * and if we are in output path rcvif could contain bogus value.
366	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
367	 * information in ip header (nip6).
368	 */
369	m->m_pkthdr.rcvif = NULL;
370
371	icmp6stat.icp6s_outhist[type]++;
372	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
373
374	return;
375
376  freeit:
377	/*
378	 * If we can't tell whether or not we can generate ICMP6, free it.
379	 */
380	m_freem(m);
381}
382
383/*
384 * Process a received ICMP6 message.
385 */
386int
387icmp6_input(struct mbuf **mp, int *offp, int proto)
388{
389	struct mbuf *m = *mp, *n;
390	struct ip6_hdr *ip6, *nip6;
391	struct icmp6_hdr *icmp6, *nicmp6;
392	int off = *offp;
393	int icmp6len = m->m_pkthdr.len - *offp;
394	int code, sum, noff;
395	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
396
397#ifndef PULLDOWN_TEST
398	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
399	/* m might change if M_LOOP.  So, call mtod after this */
400#endif
401
402	/*
403	 * Locate icmp6 structure in mbuf, and check
404	 * that not corrupted and of at least minimum length
405	 */
406
407	ip6 = mtod(m, struct ip6_hdr *);
408	if (icmp6len < sizeof(struct icmp6_hdr)) {
409		icmp6stat.icp6s_tooshort++;
410		goto freeit;
411	}
412
413	/*
414	 * calculate the checksum
415	 */
416#ifndef PULLDOWN_TEST
417	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
418#else
419	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
420	if (icmp6 == NULL) {
421		icmp6stat.icp6s_tooshort++;
422		return IPPROTO_DONE;
423	}
424#endif
425	code = icmp6->icmp6_code;
426
427	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
428		nd6log((LOG_ERR,
429		    "ICMP6 checksum error(%d|%x) %s\n",
430		    icmp6->icmp6_type, sum,
431		    ip6_sprintf(ip6bufs, &ip6->ip6_src)));
432		icmp6stat.icp6s_checksum++;
433		goto freeit;
434	}
435
436	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
437		/*
438		 * Deliver very specific ICMP6 type only.
439		 * This is important to deliver TOOBIG.  Otherwise PMTUD
440		 * will not work.
441		 */
442		switch (icmp6->icmp6_type) {
443		case ICMP6_DST_UNREACH:
444		case ICMP6_PACKET_TOO_BIG:
445		case ICMP6_TIME_EXCEEDED:
446			break;
447		default:
448			goto freeit;
449		}
450	}
451
452	icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
453	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
454	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
455		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
456
457	switch (icmp6->icmp6_type) {
458	case ICMP6_DST_UNREACH:
459		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
460		switch (code) {
461		case ICMP6_DST_UNREACH_NOROUTE:
462			code = PRC_UNREACH_NET;
463			break;
464		case ICMP6_DST_UNREACH_ADMIN:
465			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
466			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
467			break;
468		case ICMP6_DST_UNREACH_ADDR:
469			code = PRC_HOSTDEAD;
470			break;
471		case ICMP6_DST_UNREACH_BEYONDSCOPE:
472			/* I mean "source address was incorrect." */
473			code = PRC_PARAMPROB;
474			break;
475		case ICMP6_DST_UNREACH_NOPORT:
476			code = PRC_UNREACH_PORT;
477			break;
478		default:
479			goto badcode;
480		}
481		goto deliver;
482		break;
483
484	case ICMP6_PACKET_TOO_BIG:
485		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
486
487		/* validation is made in icmp6_mtudisc_update */
488
489		code = PRC_MSGSIZE;
490
491		/*
492		 * Updating the path MTU will be done after examining
493		 * intermediate extension headers.
494		 */
495		goto deliver;
496		break;
497
498	case ICMP6_TIME_EXCEEDED:
499		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
500		switch (code) {
501		case ICMP6_TIME_EXCEED_TRANSIT:
502			code = PRC_TIMXCEED_INTRANS;
503			break;
504		case ICMP6_TIME_EXCEED_REASSEMBLY:
505			code = PRC_TIMXCEED_REASS;
506			break;
507		default:
508			goto badcode;
509		}
510		goto deliver;
511		break;
512
513	case ICMP6_PARAM_PROB:
514		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
515		switch (code) {
516		case ICMP6_PARAMPROB_NEXTHEADER:
517			code = PRC_UNREACH_PROTOCOL;
518			break;
519		case ICMP6_PARAMPROB_HEADER:
520		case ICMP6_PARAMPROB_OPTION:
521			code = PRC_PARAMPROB;
522			break;
523		default:
524			goto badcode;
525		}
526		goto deliver;
527		break;
528
529	case ICMP6_ECHO_REQUEST:
530		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
531		if (code != 0)
532			goto badcode;
533		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
534			/* Give up remote */
535			break;
536		}
537		if ((n->m_flags & M_EXT) != 0
538		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
539			struct mbuf *n0 = n;
540			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
541			int n0len;
542
543			MGETHDR(n, M_DONTWAIT, n0->m_type);
544			n0len = n0->m_pkthdr.len;	/* save for use below */
545			if (n)
546				M_MOVE_PKTHDR(n, n0);
547			if (n && maxlen >= MHLEN) {
548				MCLGET(n, M_DONTWAIT);
549				if ((n->m_flags & M_EXT) == 0) {
550					m_free(n);
551					n = NULL;
552				}
553			}
554			if (n == NULL) {
555				/* Give up remote */
556				m_freem(n0);
557				break;
558			}
559			/*
560			 * Copy IPv6 and ICMPv6 only.
561			 */
562			nip6 = mtod(n, struct ip6_hdr *);
563			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
564			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
565			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
566			noff = sizeof(struct ip6_hdr);
567			/* new mbuf contains only ipv6+icmpv6 headers */
568			n->m_len = noff + sizeof(struct icmp6_hdr);
569			/*
570			 * Adjust mbuf.  ip6_plen will be adjusted in
571			 * ip6_output().
572			 */
573			m_adj(n0, off + sizeof(struct icmp6_hdr));
574			/* recalculate complete packet size */
575			n->m_pkthdr.len = n0len + (noff - off);
576			n->m_next = n0;
577		} else {
578			nip6 = mtod(n, struct ip6_hdr *);
579			IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
580			    sizeof(*nicmp6));
581			noff = off;
582		}
583		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
584		nicmp6->icmp6_code = 0;
585		if (n) {
586			icmp6stat.icp6s_reflect++;
587			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
588			icmp6_reflect(n, noff);
589		}
590		break;
591
592	case ICMP6_ECHO_REPLY:
593		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
594		if (code != 0)
595			goto badcode;
596		break;
597
598	case MLD_LISTENER_QUERY:
599	case MLD_LISTENER_REPORT:
600		if (icmp6len < sizeof(struct mld_hdr))
601			goto badlen;
602		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
603			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
604		else
605			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
606		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
607			/* give up local */
608			mld6_input(m, off);
609			m = NULL;
610			goto freeit;
611		}
612		mld6_input(n, off);
613		/* m stays. */
614		break;
615
616	case MLD_LISTENER_DONE:
617		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
618		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
619			goto badlen;
620		break;		/* nothing to be done in kernel */
621
622	case MLD_MTRACE_RESP:
623	case MLD_MTRACE:
624		/* XXX: these two are experimental.  not officially defined. */
625		/* XXX: per-interface statistics? */
626		break;		/* just pass it to applications */
627
628	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
629	    {
630		enum { WRU, FQDN } mode;
631
632		if (!icmp6_nodeinfo)
633			break;
634
635		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
636			mode = WRU;
637		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
638			mode = FQDN;
639		else
640			goto badlen;
641
642#define hostnamelen	strlen(hostname)
643		if (mode == FQDN) {
644#ifndef PULLDOWN_TEST
645			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
646			    IPPROTO_DONE);
647#endif
648			n = m_copy(m, 0, M_COPYALL);
649			if (n)
650				n = ni6_input(n, off);
651			/* XXX meaningless if n == NULL */
652			noff = sizeof(struct ip6_hdr);
653		} else {
654			u_char *p;
655			int maxlen, maxhlen;
656
657			/*
658			 * XXX: this combination of flags is pointless,
659			 * but should we keep this for compatibility?
660			 */
661			if ((icmp6_nodeinfo & 5) != 5)
662				break;
663
664			if (code != 0)
665				goto badcode;
666			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
667			if (maxlen >= MCLBYTES) {
668				/* Give up remote */
669				break;
670			}
671			MGETHDR(n, M_DONTWAIT, m->m_type);
672			if (n && maxlen > MHLEN) {
673				MCLGET(n, M_DONTWAIT);
674				if ((n->m_flags & M_EXT) == 0) {
675					m_free(n);
676					n = NULL;
677				}
678			}
679			if (n && !m_dup_pkthdr(n, m, M_DONTWAIT)) {
680				/*
681				 * Previous code did a blind M_COPY_PKTHDR
682				 * and said "just for rcvif".  If true, then
683				 * we could tolerate the dup failing (due to
684				 * the deep copy of the tag chain).  For now
685				 * be conservative and just fail.
686				 */
687				m_free(n);
688				n = NULL;
689			}
690			if (n == NULL) {
691				/* Give up remote */
692				break;
693			}
694			n->m_pkthdr.rcvif = NULL;
695			n->m_len = 0;
696			maxhlen = M_TRAILINGSPACE(n) - maxlen;
697			if (maxhlen > hostnamelen)
698				maxhlen = hostnamelen;
699			/*
700			 * Copy IPv6 and ICMPv6 only.
701			 */
702			nip6 = mtod(n, struct ip6_hdr *);
703			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
704			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
705			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
706			p = (u_char *)(nicmp6 + 1);
707			bzero(p, 4);
708			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
709			noff = sizeof(struct ip6_hdr);
710			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
711				sizeof(struct icmp6_hdr) + 4 + maxhlen;
712			nicmp6->icmp6_type = ICMP6_WRUREPLY;
713			nicmp6->icmp6_code = 0;
714		}
715#undef hostnamelen
716		if (n) {
717			icmp6stat.icp6s_reflect++;
718			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
719			icmp6_reflect(n, noff);
720		}
721		break;
722	    }
723
724	case ICMP6_WRUREPLY:
725		if (code != 0)
726			goto badcode;
727		break;
728
729	case ND_ROUTER_SOLICIT:
730		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
731		if (code != 0)
732			goto badcode;
733		if (icmp6len < sizeof(struct nd_router_solicit))
734			goto badlen;
735		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
736			/* give up local */
737			nd6_rs_input(m, off, icmp6len);
738			m = NULL;
739			goto freeit;
740		}
741		nd6_rs_input(n, off, icmp6len);
742		/* m stays. */
743		break;
744
745	case ND_ROUTER_ADVERT:
746		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
747		if (code != 0)
748			goto badcode;
749		if (icmp6len < sizeof(struct nd_router_advert))
750			goto badlen;
751		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
752			/* give up local */
753			nd6_ra_input(m, off, icmp6len);
754			m = NULL;
755			goto freeit;
756		}
757		nd6_ra_input(n, off, icmp6len);
758		/* m stays. */
759		break;
760
761	case ND_NEIGHBOR_SOLICIT:
762		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
763		if (code != 0)
764			goto badcode;
765		if (icmp6len < sizeof(struct nd_neighbor_solicit))
766			goto badlen;
767		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
768			/* give up local */
769			nd6_ns_input(m, off, icmp6len);
770			m = NULL;
771			goto freeit;
772		}
773		nd6_ns_input(n, off, icmp6len);
774		/* m stays. */
775		break;
776
777	case ND_NEIGHBOR_ADVERT:
778		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
779		if (code != 0)
780			goto badcode;
781		if (icmp6len < sizeof(struct nd_neighbor_advert))
782			goto badlen;
783		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
784			/* give up local */
785			nd6_na_input(m, off, icmp6len);
786			m = NULL;
787			goto freeit;
788		}
789		nd6_na_input(n, off, icmp6len);
790		/* m stays. */
791		break;
792
793	case ND_REDIRECT:
794		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
795		if (code != 0)
796			goto badcode;
797		if (icmp6len < sizeof(struct nd_redirect))
798			goto badlen;
799		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
800			/* give up local */
801			icmp6_redirect_input(m, off);
802			m = NULL;
803			goto freeit;
804		}
805		icmp6_redirect_input(n, off);
806		/* m stays. */
807		break;
808
809	case ICMP6_ROUTER_RENUMBERING:
810		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
811		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
812			goto badcode;
813		if (icmp6len < sizeof(struct icmp6_router_renum))
814			goto badlen;
815		break;
816
817	default:
818		nd6log((LOG_DEBUG,
819		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
820		    icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
821		    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
822		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
823		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
824			/* ICMPv6 error: MUST deliver it by spec... */
825			code = PRC_NCMDS;
826			/* deliver */
827		} else {
828			/* ICMPv6 informational: MUST not deliver */
829			break;
830		}
831	deliver:
832		if (icmp6_notify_error(&m, off, icmp6len, code)) {
833			/* In this case, m should've been freed. */
834			return (IPPROTO_DONE);
835		}
836		break;
837
838	badcode:
839		icmp6stat.icp6s_badcode++;
840		break;
841
842	badlen:
843		icmp6stat.icp6s_badlen++;
844		break;
845	}
846
847	/* deliver the packet to appropriate sockets */
848	icmp6_rip6_input(&m, *offp);
849
850	return IPPROTO_DONE;
851
852 freeit:
853	m_freem(m);
854	return IPPROTO_DONE;
855}
856
857static int
858icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
859{
860	struct mbuf *m = *mp;
861	struct icmp6_hdr *icmp6;
862	struct ip6_hdr *eip6;
863	u_int32_t notifymtu;
864	struct sockaddr_in6 icmp6src, icmp6dst;
865
866	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
867		icmp6stat.icp6s_tooshort++;
868		goto freeit;
869	}
870#ifndef PULLDOWN_TEST
871	IP6_EXTHDR_CHECK(m, off,
872	    sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
873	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
874#else
875	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
876	    sizeof(*icmp6) + sizeof(struct ip6_hdr));
877	if (icmp6 == NULL) {
878		icmp6stat.icp6s_tooshort++;
879		return (-1);
880	}
881#endif
882	eip6 = (struct ip6_hdr *)(icmp6 + 1);
883
884	/* Detect the upper level protocol */
885	{
886		void (*ctlfunc) __P((int, struct sockaddr *, void *));
887		u_int8_t nxt = eip6->ip6_nxt;
888		int eoff = off + sizeof(struct icmp6_hdr) +
889		    sizeof(struct ip6_hdr);
890		struct ip6ctlparam ip6cp;
891		struct in6_addr *finaldst = NULL;
892		int icmp6type = icmp6->icmp6_type;
893		struct ip6_frag *fh;
894		struct ip6_rthdr *rth;
895		struct ip6_rthdr0 *rth0;
896		int rthlen;
897
898		while (1) { /* XXX: should avoid infinite loop explicitly? */
899			struct ip6_ext *eh;
900
901			switch (nxt) {
902			case IPPROTO_HOPOPTS:
903			case IPPROTO_DSTOPTS:
904			case IPPROTO_AH:
905#ifndef PULLDOWN_TEST
906				IP6_EXTHDR_CHECK(m, 0,
907				    eoff + sizeof(struct ip6_ext), -1);
908				eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
909#else
910				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
911				    eoff, sizeof(*eh));
912				if (eh == NULL) {
913					icmp6stat.icp6s_tooshort++;
914					return (-1);
915				}
916#endif
917
918				if (nxt == IPPROTO_AH)
919					eoff += (eh->ip6e_len + 2) << 2;
920				else
921					eoff += (eh->ip6e_len + 1) << 3;
922				nxt = eh->ip6e_nxt;
923				break;
924			case IPPROTO_ROUTING:
925				/*
926				 * When the erroneous packet contains a
927				 * routing header, we should examine the
928				 * header to determine the final destination.
929				 * Otherwise, we can't properly update
930				 * information that depends on the final
931				 * destination (e.g. path MTU).
932				 */
933#ifndef PULLDOWN_TEST
934				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
935				rth = (struct ip6_rthdr *)
936				    (mtod(m, caddr_t) + eoff);
937#else
938				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
939				    eoff, sizeof(*rth));
940				if (rth == NULL) {
941					icmp6stat.icp6s_tooshort++;
942					return (-1);
943				}
944#endif
945				rthlen = (rth->ip6r_len + 1) << 3;
946				/*
947				 * XXX: currently there is no
948				 * officially defined type other
949				 * than type-0.
950				 * Note that if the segment left field
951				 * is 0, all intermediate hops must
952				 * have been passed.
953				 */
954				if (rth->ip6r_segleft &&
955				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
956					int hops;
957
958#ifndef PULLDOWN_TEST
959					IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
960					rth0 = (struct ip6_rthdr0 *)
961					    (mtod(m, caddr_t) + eoff);
962#else
963					IP6_EXTHDR_GET(rth0,
964					    struct ip6_rthdr0 *, m,
965					    eoff, rthlen);
966					if (rth0 == NULL) {
967						icmp6stat.icp6s_tooshort++;
968						return (-1);
969					}
970#endif
971					/* just ignore a bogus header */
972					if ((rth0->ip6r0_len % 2) == 0 &&
973					    (hops = rth0->ip6r0_len/2))
974						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
975				}
976				eoff += rthlen;
977				nxt = rth->ip6r_nxt;
978				break;
979			case IPPROTO_FRAGMENT:
980#ifndef PULLDOWN_TEST
981				IP6_EXTHDR_CHECK(m, 0, eoff +
982				    sizeof(struct ip6_frag), -1);
983				fh = (struct ip6_frag *)(mtod(m, caddr_t) +
984				    eoff);
985#else
986				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
987				    eoff, sizeof(*fh));
988				if (fh == NULL) {
989					icmp6stat.icp6s_tooshort++;
990					return (-1);
991				}
992#endif
993				/*
994				 * Data after a fragment header is meaningless
995				 * unless it is the first fragment, but
996				 * we'll go to the notify label for path MTU
997				 * discovery.
998				 */
999				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1000					goto notify;
1001
1002				eoff += sizeof(struct ip6_frag);
1003				nxt = fh->ip6f_nxt;
1004				break;
1005			default:
1006				/*
1007				 * This case includes ESP and the No Next
1008				 * Header.  In such cases going to the notify
1009				 * label does not have any meaning
1010				 * (i.e. ctlfunc will be NULL), but we go
1011				 * anyway since we might have to update
1012				 * path MTU information.
1013				 */
1014				goto notify;
1015			}
1016		}
1017	  notify:
1018#ifndef PULLDOWN_TEST
1019		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1020#else
1021		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1022		    sizeof(*icmp6) + sizeof(struct ip6_hdr));
1023		if (icmp6 == NULL) {
1024			icmp6stat.icp6s_tooshort++;
1025			return (-1);
1026		}
1027#endif
1028
1029		/*
1030		 * retrieve parameters from the inner IPv6 header, and convert
1031		 * them into sockaddr structures.
1032		 * XXX: there is no guarantee that the source or destination
1033		 * addresses of the inner packet are in the same scope as
1034		 * the addresses of the icmp packet.  But there is no other
1035		 * way to determine the zone.
1036		 */
1037		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1038
1039		bzero(&icmp6dst, sizeof(icmp6dst));
1040		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1041		icmp6dst.sin6_family = AF_INET6;
1042		if (finaldst == NULL)
1043			icmp6dst.sin6_addr = eip6->ip6_dst;
1044		else
1045			icmp6dst.sin6_addr = *finaldst;
1046		if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1047			goto freeit;
1048		bzero(&icmp6src, sizeof(icmp6src));
1049		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1050		icmp6src.sin6_family = AF_INET6;
1051		icmp6src.sin6_addr = eip6->ip6_src;
1052		if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1053			goto freeit;
1054		icmp6src.sin6_flowinfo =
1055		    (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1056
1057		if (finaldst == NULL)
1058			finaldst = &eip6->ip6_dst;
1059		ip6cp.ip6c_m = m;
1060		ip6cp.ip6c_icmp6 = icmp6;
1061		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1062		ip6cp.ip6c_off = eoff;
1063		ip6cp.ip6c_finaldst = finaldst;
1064		ip6cp.ip6c_src = &icmp6src;
1065		ip6cp.ip6c_nxt = nxt;
1066
1067		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1068			notifymtu = ntohl(icmp6->icmp6_mtu);
1069			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1070			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1071		}
1072
1073		ctlfunc = (void (*) __P((int, struct sockaddr *, void *)))
1074		    (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1075		if (ctlfunc) {
1076			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1077			    &ip6cp);
1078		}
1079	}
1080	*mp = m;
1081	return (0);
1082
1083  freeit:
1084	m_freem(m);
1085	return (-1);
1086}
1087
1088void
1089icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1090{
1091	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1092	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1093	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1094	u_int mtu = ntohl(icmp6->icmp6_mtu);
1095	struct in_conninfo inc;
1096
1097#if 0
1098	/*
1099	 * RFC2460 section 5, last paragraph.
1100	 * even though minimum link MTU for IPv6 is IPV6_MMTU,
1101	 * we may see ICMPv6 too big with mtu < IPV6_MMTU
1102	 * due to packet translator in the middle.
1103	 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1104	 * special handling.
1105	 */
1106	if (mtu < IPV6_MMTU)
1107		return;
1108#endif
1109
1110	/*
1111	 * we reject ICMPv6 too big with abnormally small value.
1112	 * XXX what is the good definition of "abnormally small"?
1113	 */
1114	if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1115		return;
1116
1117	if (!validated)
1118		return;
1119
1120	bzero(&inc, sizeof(inc));
1121	inc.inc_flags = 1; /* IPv6 */
1122	inc.inc6_faddr = *dst;
1123	if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1124		return;
1125
1126	if (mtu < tcp_maxmtu6(&inc, NULL)) {
1127		tcp_hc_updatemtu(&inc, mtu);
1128		icmp6stat.icp6s_pmtuchg++;
1129	}
1130}
1131
1132/*
1133 * Process a Node Information Query packet, based on
1134 * draft-ietf-ipngwg-icmp-name-lookups-07.
1135 *
1136 * Spec incompatibilities:
1137 * - IPv6 Subject address handling
1138 * - IPv4 Subject address handling support missing
1139 * - Proxy reply (answer even if it's not for me)
1140 * - joins NI group address at in6_ifattach() time only, does not cope
1141 *   with hostname changes by sethostname(3)
1142 */
1143#define hostnamelen	strlen(hostname)
1144static struct mbuf *
1145ni6_input(struct mbuf *m, int off)
1146{
1147	struct icmp6_nodeinfo *ni6, *nni6;
1148	struct mbuf *n = NULL;
1149	u_int16_t qtype;
1150	int subjlen;
1151	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1152	struct ni_reply_fqdn *fqdn;
1153	int addrs;		/* for NI_QTYPE_NODEADDR */
1154	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1155	struct in6_addr in6_subj; /* subject address */
1156	struct ip6_hdr *ip6;
1157	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1158	char *subj = NULL;
1159	struct in6_ifaddr *ia6 = NULL;
1160
1161	ip6 = mtod(m, struct ip6_hdr *);
1162#ifndef PULLDOWN_TEST
1163	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1164#else
1165	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1166	if (ni6 == NULL) {
1167		/* m is already reclaimed */
1168		return (NULL);
1169	}
1170#endif
1171
1172	/*
1173	 * Validate IPv6 source address.
1174	 * The default configuration MUST be to refuse answering queries from
1175	 * global-scope addresses according to RFC4602.
1176	 * Notes:
1177	 *  - it's not very clear what "refuse" means; this implementation
1178	 *    simply drops it.
1179	 *  - it's not very easy to identify global-scope (unicast) addresses
1180	 *    since there are many prefixes for them.  It should be safer
1181	 *    and in practice sufficient to check "all" but loopback and
1182	 *    link-local (note that site-local unicast was deprecated and
1183	 *    ULA is defined as global scope-wise)
1184	 */
1185	if ((icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1186	    !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1187	    !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1188		goto bad;
1189
1190	/*
1191	 * Validate IPv6 destination address.
1192	 *
1193	 * The Responder must discard the Query without further processing
1194	 * unless it is one of the Responder's unicast or anycast addresses, or
1195	 * a link-local scope multicast address which the Responder has joined.
1196	 * [RFC4602, Section 5.]
1197	 */
1198	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1199		if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1200			goto bad;
1201		/* else it's a link-local multicast, fine */
1202	} else {		/* unicast or anycast */
1203		if ((ia6 = ip6_getdstifaddr(m)) == NULL)
1204			goto bad; /* XXX impossible */
1205
1206		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1207		    !(icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1208			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1209				"a temporary address in %s:%d",
1210			       __FILE__, __LINE__));
1211			goto bad;
1212		}
1213	}
1214
1215	/* validate query Subject field. */
1216	qtype = ntohs(ni6->ni_qtype);
1217	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1218	switch (qtype) {
1219	case NI_QTYPE_NOOP:
1220	case NI_QTYPE_SUPTYPES:
1221		/* 07 draft */
1222		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1223			break;
1224		/* FALLTHROUGH */
1225	case NI_QTYPE_FQDN:
1226	case NI_QTYPE_NODEADDR:
1227	case NI_QTYPE_IPV4ADDR:
1228		switch (ni6->ni_code) {
1229		case ICMP6_NI_SUBJ_IPV6:
1230#if ICMP6_NI_SUBJ_IPV6 != 0
1231		case 0:
1232#endif
1233			/*
1234			 * backward compatibility - try to accept 03 draft
1235			 * format, where no Subject is present.
1236			 */
1237			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1238			    subjlen == 0) {
1239				oldfqdn++;
1240				break;
1241			}
1242#if ICMP6_NI_SUBJ_IPV6 != 0
1243			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1244				goto bad;
1245#endif
1246
1247			if (subjlen != sizeof(struct in6_addr))
1248				goto bad;
1249
1250			/*
1251			 * Validate Subject address.
1252			 *
1253			 * Not sure what exactly "address belongs to the node"
1254			 * means in the spec, is it just unicast, or what?
1255			 *
1256			 * At this moment we consider Subject address as
1257			 * "belong to the node" if the Subject address equals
1258			 * to the IPv6 destination address; validation for
1259			 * IPv6 destination address should have done enough
1260			 * check for us.
1261			 *
1262			 * We do not do proxy at this moment.
1263			 */
1264			/* m_pulldown instead of copy? */
1265			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1266			    subjlen, (caddr_t)&in6_subj);
1267			if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1268				goto bad;
1269
1270			subj = (char *)&in6_subj;
1271			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1272				break;
1273
1274			/*
1275			 * XXX if we are to allow other cases, we should really
1276			 * be careful about scope here.
1277			 * basically, we should disallow queries toward IPv6
1278			 * destination X with subject Y,
1279			 * if scope(X) > scope(Y).
1280			 * if we allow scope(X) > scope(Y), it will result in
1281			 * information leakage across scope boundary.
1282			 */
1283			goto bad;
1284
1285		case ICMP6_NI_SUBJ_FQDN:
1286			/*
1287			 * Validate Subject name with gethostname(3).
1288			 *
1289			 * The behavior may need some debate, since:
1290			 * - we are not sure if the node has FQDN as
1291			 *   hostname (returned by gethostname(3)).
1292			 * - the code does wildcard match for truncated names.
1293			 *   however, we are not sure if we want to perform
1294			 *   wildcard match, if gethostname(3) side has
1295			 *   truncated hostname.
1296			 */
1297			n = ni6_nametodns(hostname, hostnamelen, 0);
1298			if (!n || n->m_next || n->m_len == 0)
1299				goto bad;
1300			IP6_EXTHDR_GET(subj, char *, m,
1301			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1302			if (subj == NULL)
1303				goto bad;
1304			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1305			    n->m_len)) {
1306				goto bad;
1307			}
1308			m_freem(n);
1309			n = NULL;
1310			break;
1311
1312		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1313		default:
1314			goto bad;
1315		}
1316		break;
1317	}
1318
1319	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1320	switch (qtype) {
1321	case NI_QTYPE_FQDN:
1322		if ((icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1323			goto bad;
1324		break;
1325	case NI_QTYPE_NODEADDR:
1326	case NI_QTYPE_IPV4ADDR:
1327		if ((icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1328			goto bad;
1329		break;
1330	}
1331
1332	/* guess reply length */
1333	switch (qtype) {
1334	case NI_QTYPE_NOOP:
1335		break;		/* no reply data */
1336	case NI_QTYPE_SUPTYPES:
1337		replylen += sizeof(u_int32_t);
1338		break;
1339	case NI_QTYPE_FQDN:
1340		/* XXX will append an mbuf */
1341		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1342		break;
1343	case NI_QTYPE_NODEADDR:
1344		addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1345		if ((replylen += addrs * (sizeof(struct in6_addr) +
1346		    sizeof(u_int32_t))) > MCLBYTES)
1347			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1348		break;
1349	case NI_QTYPE_IPV4ADDR:
1350		/* unsupported - should respond with unknown Qtype? */
1351		break;
1352	default:
1353		/*
1354		 * XXX: We must return a reply with the ICMP6 code
1355		 * `unknown Qtype' in this case.  However we regard the case
1356		 * as an FQDN query for backward compatibility.
1357		 * Older versions set a random value to this field,
1358		 * so it rarely varies in the defined qtypes.
1359		 * But the mechanism is not reliable...
1360		 * maybe we should obsolete older versions.
1361		 */
1362		qtype = NI_QTYPE_FQDN;
1363		/* XXX will append an mbuf */
1364		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1365		oldfqdn++;
1366		break;
1367	}
1368
1369	/* allocate an mbuf to reply. */
1370	MGETHDR(n, M_DONTWAIT, m->m_type);
1371	if (n == NULL) {
1372		m_freem(m);
1373		return (NULL);
1374	}
1375	M_MOVE_PKTHDR(n, m); /* just for recvif */
1376	if (replylen > MHLEN) {
1377		if (replylen > MCLBYTES) {
1378			/*
1379			 * XXX: should we try to allocate more? But MCLBYTES
1380			 * is probably much larger than IPV6_MMTU...
1381			 */
1382			goto bad;
1383		}
1384		MCLGET(n, M_DONTWAIT);
1385		if ((n->m_flags & M_EXT) == 0) {
1386			goto bad;
1387		}
1388	}
1389	n->m_pkthdr.len = n->m_len = replylen;
1390
1391	/* copy mbuf header and IPv6 + Node Information base headers */
1392	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1393	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1394	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1395
1396	/* qtype dependent procedure */
1397	switch (qtype) {
1398	case NI_QTYPE_NOOP:
1399		nni6->ni_code = ICMP6_NI_SUCCESS;
1400		nni6->ni_flags = 0;
1401		break;
1402	case NI_QTYPE_SUPTYPES:
1403	{
1404		u_int32_t v;
1405		nni6->ni_code = ICMP6_NI_SUCCESS;
1406		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1407		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1408		v = (u_int32_t)htonl(0x0000000f);
1409		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1410		break;
1411	}
1412	case NI_QTYPE_FQDN:
1413		nni6->ni_code = ICMP6_NI_SUCCESS;
1414		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1415		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1416		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1417		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1418		/*
1419		 * XXX do we really have FQDN in variable "hostname"?
1420		 */
1421		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1422		if (n->m_next == NULL)
1423			goto bad;
1424		/* XXX we assume that n->m_next is not a chain */
1425		if (n->m_next->m_next != NULL)
1426			goto bad;
1427		n->m_pkthdr.len += n->m_next->m_len;
1428		break;
1429	case NI_QTYPE_NODEADDR:
1430	{
1431		int lenlim, copied;
1432
1433		nni6->ni_code = ICMP6_NI_SUCCESS;
1434		n->m_pkthdr.len = n->m_len =
1435		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1436		lenlim = M_TRAILINGSPACE(n);
1437		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1438		/* XXX: reset mbuf length */
1439		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1440		    sizeof(struct icmp6_nodeinfo) + copied;
1441		break;
1442	}
1443	default:
1444		break;		/* XXX impossible! */
1445	}
1446
1447	nni6->ni_type = ICMP6_NI_REPLY;
1448	m_freem(m);
1449	return (n);
1450
1451  bad:
1452	m_freem(m);
1453	if (n)
1454		m_freem(n);
1455	return (NULL);
1456}
1457#undef hostnamelen
1458
1459/*
1460 * make a mbuf with DNS-encoded string.  no compression support.
1461 *
1462 * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1463 * treated as truncated name (two \0 at the end).  this is a wild guess.
1464 *
1465 * old - return pascal string if non-zero
1466 */
1467static struct mbuf *
1468ni6_nametodns(const char *name, int namelen, int old)
1469{
1470	struct mbuf *m;
1471	char *cp, *ep;
1472	const char *p, *q;
1473	int i, len, nterm;
1474
1475	if (old)
1476		len = namelen + 1;
1477	else
1478		len = MCLBYTES;
1479
1480	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1481	MGET(m, M_DONTWAIT, MT_DATA);
1482	if (m && len > MLEN) {
1483		MCLGET(m, M_DONTWAIT);
1484		if ((m->m_flags & M_EXT) == 0)
1485			goto fail;
1486	}
1487	if (!m)
1488		goto fail;
1489	m->m_next = NULL;
1490
1491	if (old) {
1492		m->m_len = len;
1493		*mtod(m, char *) = namelen;
1494		bcopy(name, mtod(m, char *) + 1, namelen);
1495		return m;
1496	} else {
1497		m->m_len = 0;
1498		cp = mtod(m, char *);
1499		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1500
1501		/* if not certain about my name, return empty buffer */
1502		if (namelen == 0)
1503			return m;
1504
1505		/*
1506		 * guess if it looks like shortened hostname, or FQDN.
1507		 * shortened hostname needs two trailing "\0".
1508		 */
1509		i = 0;
1510		for (p = name; p < name + namelen; p++) {
1511			if (*p && *p == '.')
1512				i++;
1513		}
1514		if (i < 2)
1515			nterm = 2;
1516		else
1517			nterm = 1;
1518
1519		p = name;
1520		while (cp < ep && p < name + namelen) {
1521			i = 0;
1522			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1523				i++;
1524			/* result does not fit into mbuf */
1525			if (cp + i + 1 >= ep)
1526				goto fail;
1527			/*
1528			 * DNS label length restriction, RFC1035 page 8.
1529			 * "i == 0" case is included here to avoid returning
1530			 * 0-length label on "foo..bar".
1531			 */
1532			if (i <= 0 || i >= 64)
1533				goto fail;
1534			*cp++ = i;
1535			bcopy(p, cp, i);
1536			cp += i;
1537			p = q;
1538			if (p < name + namelen && *p == '.')
1539				p++;
1540		}
1541		/* termination */
1542		if (cp + nterm >= ep)
1543			goto fail;
1544		while (nterm-- > 0)
1545			*cp++ = '\0';
1546		m->m_len = cp - mtod(m, char *);
1547		return m;
1548	}
1549
1550	panic("should not reach here");
1551	/* NOTREACHED */
1552
1553 fail:
1554	if (m)
1555		m_freem(m);
1556	return NULL;
1557}
1558
1559/*
1560 * check if two DNS-encoded string matches.  takes care of truncated
1561 * form (with \0\0 at the end).  no compression support.
1562 * XXX upper/lowercase match (see RFC2065)
1563 */
1564static int
1565ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1566{
1567	const char *a0, *b0;
1568	int l;
1569
1570	/* simplest case - need validation? */
1571	if (alen == blen && bcmp(a, b, alen) == 0)
1572		return 1;
1573
1574	a0 = a;
1575	b0 = b;
1576
1577	/* termination is mandatory */
1578	if (alen < 2 || blen < 2)
1579		return 0;
1580	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1581		return 0;
1582	alen--;
1583	blen--;
1584
1585	while (a - a0 < alen && b - b0 < blen) {
1586		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1587			return 0;
1588
1589		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1590			return 0;
1591		/* we don't support compression yet */
1592		if (a[0] >= 64 || b[0] >= 64)
1593			return 0;
1594
1595		/* truncated case */
1596		if (a[0] == 0 && a - a0 == alen - 1)
1597			return 1;
1598		if (b[0] == 0 && b - b0 == blen - 1)
1599			return 1;
1600		if (a[0] == 0 || b[0] == 0)
1601			return 0;
1602
1603		if (a[0] != b[0])
1604			return 0;
1605		l = a[0];
1606		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1607			return 0;
1608		if (bcmp(a + 1, b + 1, l) != 0)
1609			return 0;
1610
1611		a += 1 + l;
1612		b += 1 + l;
1613	}
1614
1615	if (a - a0 == alen && b - b0 == blen)
1616		return 1;
1617	else
1618		return 0;
1619}
1620
1621/*
1622 * calculate the number of addresses to be returned in the node info reply.
1623 */
1624static int
1625ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1626    struct in6_addr *subj)
1627{
1628	struct ifnet *ifp;
1629	struct in6_ifaddr *ifa6;
1630	struct ifaddr *ifa;
1631	int addrs = 0, addrsofif, iffound = 0;
1632	int niflags = ni6->ni_flags;
1633
1634	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1635		switch (ni6->ni_code) {
1636		case ICMP6_NI_SUBJ_IPV6:
1637			if (subj == NULL) /* must be impossible... */
1638				return (0);
1639			break;
1640		default:
1641			/*
1642			 * XXX: we only support IPv6 subject address for
1643			 * this Qtype.
1644			 */
1645			return (0);
1646		}
1647	}
1648
1649	IFNET_RLOCK();
1650	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1651		addrsofif = 0;
1652		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1653			if (ifa->ifa_addr->sa_family != AF_INET6)
1654				continue;
1655			ifa6 = (struct in6_ifaddr *)ifa;
1656
1657			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1658			    IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1659				iffound = 1;
1660
1661			/*
1662			 * IPv4-mapped addresses can only be returned by a
1663			 * Node Information proxy, since they represent
1664			 * addresses of IPv4-only nodes, which perforce do
1665			 * not implement this protocol.
1666			 * [icmp-name-lookups-07, Section 5.4]
1667			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1668			 * this function at this moment.
1669			 */
1670
1671			/* What do we have to do about ::1? */
1672			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1673			case IPV6_ADDR_SCOPE_LINKLOCAL:
1674				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1675					continue;
1676				break;
1677			case IPV6_ADDR_SCOPE_SITELOCAL:
1678				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1679					continue;
1680				break;
1681			case IPV6_ADDR_SCOPE_GLOBAL:
1682				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1683					continue;
1684				break;
1685			default:
1686				continue;
1687			}
1688
1689			/*
1690			 * check if anycast is okay.
1691			 * XXX: just experimental.  not in the spec.
1692			 */
1693			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1694			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1695				continue; /* we need only unicast addresses */
1696			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1697			    (icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1698				continue;
1699			}
1700			addrsofif++; /* count the address */
1701		}
1702		if (iffound) {
1703			*ifpp = ifp;
1704			IFNET_RUNLOCK();
1705			return (addrsofif);
1706		}
1707
1708		addrs += addrsofif;
1709	}
1710	IFNET_RUNLOCK();
1711
1712	return (addrs);
1713}
1714
1715static int
1716ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1717    struct ifnet *ifp0, int resid)
1718{
1719	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1720	struct in6_ifaddr *ifa6;
1721	struct ifaddr *ifa;
1722	struct ifnet *ifp_dep = NULL;
1723	int copied = 0, allow_deprecated = 0;
1724	u_char *cp = (u_char *)(nni6 + 1);
1725	int niflags = ni6->ni_flags;
1726	u_int32_t ltime;
1727
1728	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1729		return (0);	/* needless to copy */
1730
1731	IFNET_RLOCK();
1732  again:
1733
1734	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1735		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1736		     ifa = ifa->ifa_list.tqe_next) {
1737			if (ifa->ifa_addr->sa_family != AF_INET6)
1738				continue;
1739			ifa6 = (struct in6_ifaddr *)ifa;
1740
1741			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1742			    allow_deprecated == 0) {
1743				/*
1744				 * prefererred address should be put before
1745				 * deprecated addresses.
1746				 */
1747
1748				/* record the interface for later search */
1749				if (ifp_dep == NULL)
1750					ifp_dep = ifp;
1751
1752				continue;
1753			} else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1754			    allow_deprecated != 0)
1755				continue; /* we now collect deprecated addrs */
1756
1757			/* What do we have to do about ::1? */
1758			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1759			case IPV6_ADDR_SCOPE_LINKLOCAL:
1760				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1761					continue;
1762				break;
1763			case IPV6_ADDR_SCOPE_SITELOCAL:
1764				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1765					continue;
1766				break;
1767			case IPV6_ADDR_SCOPE_GLOBAL:
1768				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1769					continue;
1770				break;
1771			default:
1772				continue;
1773			}
1774
1775			/*
1776			 * check if anycast is okay.
1777			 * XXX: just experimental.  not in the spec.
1778			 */
1779			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1780			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1781				continue;
1782			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1783			    (icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1784				continue;
1785			}
1786
1787			/* now we can copy the address */
1788			if (resid < sizeof(struct in6_addr) +
1789			    sizeof(u_int32_t)) {
1790				/*
1791				 * We give up much more copy.
1792				 * Set the truncate flag and return.
1793				 */
1794				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1795				IFNET_RUNLOCK();
1796				return (copied);
1797			}
1798
1799			/*
1800			 * Set the TTL of the address.
1801			 * The TTL value should be one of the following
1802			 * according to the specification:
1803			 *
1804			 * 1. The remaining lifetime of a DHCP lease on the
1805			 *    address, or
1806			 * 2. The remaining Valid Lifetime of a prefix from
1807			 *    which the address was derived through Stateless
1808			 *    Autoconfiguration.
1809			 *
1810			 * Note that we currently do not support stateful
1811			 * address configuration by DHCPv6, so the former
1812			 * case can't happen.
1813			 */
1814			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1815				ltime = ND6_INFINITE_LIFETIME;
1816			else {
1817				if (ifa6->ia6_lifetime.ia6t_expire >
1818				    time_second)
1819					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1820				else
1821					ltime = 0;
1822			}
1823
1824			bcopy(&ltime, cp, sizeof(u_int32_t));
1825			cp += sizeof(u_int32_t);
1826
1827			/* copy the address itself */
1828			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1829			    sizeof(struct in6_addr));
1830			in6_clearscope((struct in6_addr *)cp); /* XXX */
1831			cp += sizeof(struct in6_addr);
1832
1833			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1834			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1835		}
1836		if (ifp0)	/* we need search only on the specified IF */
1837			break;
1838	}
1839
1840	if (allow_deprecated == 0 && ifp_dep != NULL) {
1841		ifp = ifp_dep;
1842		allow_deprecated = 1;
1843
1844		goto again;
1845	}
1846
1847	IFNET_RUNLOCK();
1848
1849	return (copied);
1850}
1851
1852/*
1853 * XXX almost dup'ed code with rip6_input.
1854 */
1855static int
1856icmp6_rip6_input(struct mbuf **mp, 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				SOCKBUF_LOCK(&last->in6p_socket->so_rcv);
1951				if (sbappendaddr_locked(
1952				    &last->in6p_socket->so_rcv,
1953				    (struct sockaddr *)&fromsa, n, opts)
1954				    == 0) {
1955					/* should notify about lost packet */
1956					m_freem(n);
1957					if (opts) {
1958						m_freem(opts);
1959					}
1960					SOCKBUF_UNLOCK(
1961					    &last->in6p_socket->so_rcv);
1962				} else
1963					sorwakeup_locked(last->in6p_socket);
1964				opts = NULL;
1965			}
1966			INP_UNLOCK(last);
1967		}
1968		last = in6p;
1969	}
1970	if (last) {
1971		if (last->in6p_flags & IN6P_CONTROLOPTS)
1972			ip6_savecontrol(last, m, &opts);
1973		/* strip intermediate headers */
1974		m_adj(m, off);
1975
1976		/* avoid using mbuf clusters if possible (see above) */
1977		if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1978		    m->m_len <= MHLEN) {
1979			struct mbuf *n;
1980
1981			MGET(n, M_DONTWAIT, m->m_type);
1982			if (n != NULL) {
1983				if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1984					bcopy(m->m_data, n->m_data, m->m_len);
1985					n->m_len = m->m_len;
1986
1987					m_freem(m);
1988					m = n;
1989				} else {
1990					m_freem(n);
1991					n = NULL;
1992				}
1993			}
1994		}
1995		SOCKBUF_LOCK(&last->in6p_socket->so_rcv);
1996		if (sbappendaddr_locked(&last->in6p_socket->so_rcv,
1997		    (struct sockaddr *)&fromsa, m, opts) == 0) {
1998			m_freem(m);
1999			if (opts)
2000				m_freem(opts);
2001			SOCKBUF_UNLOCK(&last->in6p_socket->so_rcv);
2002		} else
2003			sorwakeup_locked(last->in6p_socket);
2004		INP_UNLOCK(last);
2005	} else {
2006		m_freem(m);
2007		ip6stat.ip6s_delivered--;
2008	}
2009	INP_INFO_RUNLOCK(&ripcbinfo);
2010	return IPPROTO_DONE;
2011}
2012
2013/*
2014 * Reflect the ip6 packet back to the source.
2015 * OFF points to the icmp6 header, counted from the top of the mbuf.
2016 */
2017void
2018icmp6_reflect(struct mbuf *m, size_t off)
2019{
2020	struct ip6_hdr *ip6;
2021	struct icmp6_hdr *icmp6;
2022	struct in6_ifaddr *ia;
2023	int plen;
2024	int type, code;
2025	struct ifnet *outif = NULL;
2026	struct in6_addr origdst, *src = NULL;
2027
2028	/* too short to reflect */
2029	if (off < sizeof(struct ip6_hdr)) {
2030		nd6log((LOG_DEBUG,
2031		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2032		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2033		    __FILE__, __LINE__));
2034		goto bad;
2035	}
2036
2037	/*
2038	 * If there are extra headers between IPv6 and ICMPv6, strip
2039	 * off that header first.
2040	 */
2041#ifdef DIAGNOSTIC
2042	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2043		panic("assumption failed in icmp6_reflect");
2044#endif
2045	if (off > sizeof(struct ip6_hdr)) {
2046		size_t l;
2047		struct ip6_hdr nip6;
2048
2049		l = off - sizeof(struct ip6_hdr);
2050		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2051		m_adj(m, l);
2052		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2053		if (m->m_len < l) {
2054			if ((m = m_pullup(m, l)) == NULL)
2055				return;
2056		}
2057		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2058	} else /* off == sizeof(struct ip6_hdr) */ {
2059		size_t l;
2060		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2061		if (m->m_len < l) {
2062			if ((m = m_pullup(m, l)) == NULL)
2063				return;
2064		}
2065	}
2066	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2067	ip6 = mtod(m, struct ip6_hdr *);
2068	ip6->ip6_nxt = IPPROTO_ICMPV6;
2069	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2070	type = icmp6->icmp6_type; /* keep type for statistics */
2071	code = icmp6->icmp6_code; /* ditto. */
2072
2073	origdst = ip6->ip6_dst;
2074	/*
2075	 * ip6_input() drops a packet if its src is multicast.
2076	 * So, the src is never multicast.
2077	 */
2078	ip6->ip6_dst = ip6->ip6_src;
2079
2080	/*
2081	 * If the incoming packet was addressed directly to us (i.e. unicast),
2082	 * use dst as the src for the reply.
2083	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2084	 * (for example) when we encounter an error while forwarding procedure
2085	 * destined to a duplicated address of ours.
2086	 * Note that ip6_getdstifaddr() may fail if we are in an error handling
2087	 * procedure of an outgoing packet of our own, in which case we need
2088	 * to search in the ifaddr list.
2089	 */
2090	if (!IN6_IS_ADDR_MULTICAST(&origdst)) {
2091		if ((ia = ip6_getdstifaddr(m))) {
2092			if (!(ia->ia6_flags &
2093			    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)))
2094				src = &ia->ia_addr.sin6_addr;
2095		} else {
2096			struct sockaddr_in6 d;
2097
2098			bzero(&d, sizeof(d));
2099			d.sin6_family = AF_INET6;
2100			d.sin6_len = sizeof(d);
2101			d.sin6_addr = origdst;
2102			ia = (struct in6_ifaddr *)
2103			    ifa_ifwithaddr((struct sockaddr *)&d);
2104			if (ia &&
2105			    !(ia->ia6_flags &
2106			    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2107				src = &ia->ia_addr.sin6_addr;
2108			}
2109		}
2110	}
2111
2112	if (src == NULL) {
2113		int e;
2114		struct sockaddr_in6 sin6;
2115		struct route_in6 ro;
2116
2117		/*
2118		 * This case matches to multicasts, our anycast, or unicasts
2119		 * that we do not own.  Select a source address based on the
2120		 * source address of the erroneous packet.
2121		 */
2122		bzero(&sin6, sizeof(sin6));
2123		sin6.sin6_family = AF_INET6;
2124		sin6.sin6_len = sizeof(sin6);
2125		sin6.sin6_addr = ip6->ip6_dst; /* zone ID should be embedded */
2126
2127		bzero(&ro, sizeof(ro));
2128		src = in6_selectsrc(&sin6, NULL, NULL, &ro, NULL, &outif, &e);
2129		if (ro.ro_rt)
2130			RTFREE(ro.ro_rt); /* XXX: we could use this */
2131		if (src == NULL) {
2132			char ip6buf[INET6_ADDRSTRLEN];
2133			nd6log((LOG_DEBUG,
2134			    "icmp6_reflect: source can't be determined: "
2135			    "dst=%s, error=%d\n",
2136			    ip6_sprintf(ip6buf, &sin6.sin6_addr), e));
2137			goto bad;
2138		}
2139	}
2140
2141	ip6->ip6_src = *src;
2142	ip6->ip6_flow = 0;
2143	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2144	ip6->ip6_vfc |= IPV6_VERSION;
2145	ip6->ip6_nxt = IPPROTO_ICMPV6;
2146	if (outif)
2147		ip6->ip6_hlim = ND_IFINFO(outif)->chlim;
2148	else if (m->m_pkthdr.rcvif) {
2149		/* XXX: This may not be the outgoing interface */
2150		ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2151	} else
2152		ip6->ip6_hlim = ip6_defhlim;
2153
2154	icmp6->icmp6_cksum = 0;
2155	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2156	    sizeof(struct ip6_hdr), plen);
2157
2158	/*
2159	 * XXX option handling
2160	 */
2161
2162	m->m_flags &= ~(M_BCAST|M_MCAST);
2163
2164	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2165	if (outif)
2166		icmp6_ifoutstat_inc(outif, type, code);
2167
2168	return;
2169
2170 bad:
2171	m_freem(m);
2172	return;
2173}
2174
2175void
2176icmp6_fasttimo(void)
2177{
2178
2179	return;
2180}
2181
2182static const char *
2183icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2184    struct in6_addr *tgt6)
2185{
2186	static char buf[1024];
2187	char ip6bufs[INET6_ADDRSTRLEN];
2188	char ip6bufd[INET6_ADDRSTRLEN];
2189	char ip6buft[INET6_ADDRSTRLEN];
2190	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2191	    ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2192	    ip6_sprintf(ip6buft, tgt6));
2193	return buf;
2194}
2195
2196void
2197icmp6_redirect_input(struct mbuf *m, int off)
2198{
2199	struct ifnet *ifp;
2200	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2201	struct nd_redirect *nd_rd;
2202	int icmp6len = ntohs(ip6->ip6_plen);
2203	char *lladdr = NULL;
2204	int lladdrlen = 0;
2205	u_char *redirhdr = NULL;
2206	int redirhdrlen = 0;
2207	struct rtentry *rt = NULL;
2208	int is_router;
2209	int is_onlink;
2210	struct in6_addr src6 = ip6->ip6_src;
2211	struct in6_addr redtgt6;
2212	struct in6_addr reddst6;
2213	union nd_opts ndopts;
2214	char ip6buf[INET6_ADDRSTRLEN];
2215
2216	if (!m)
2217		return;
2218
2219	ifp = m->m_pkthdr.rcvif;
2220
2221	if (!ifp)
2222		return;
2223
2224	/* XXX if we are router, we don't update route by icmp6 redirect */
2225	if (ip6_forwarding)
2226		goto freeit;
2227	if (!icmp6_rediraccept)
2228		goto freeit;
2229
2230#ifndef PULLDOWN_TEST
2231	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2232	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2233#else
2234	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2235	if (nd_rd == NULL) {
2236		icmp6stat.icp6s_tooshort++;
2237		return;
2238	}
2239#endif
2240	redtgt6 = nd_rd->nd_rd_target;
2241	reddst6 = nd_rd->nd_rd_dst;
2242
2243	if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2244	    in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2245		goto freeit;
2246	}
2247
2248	/* validation */
2249	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2250		nd6log((LOG_ERR,
2251		    "ICMP6 redirect sent from %s rejected; "
2252		    "must be from linklocal\n",
2253		    ip6_sprintf(ip6buf, &src6)));
2254		goto bad;
2255	}
2256	if (ip6->ip6_hlim != 255) {
2257		nd6log((LOG_ERR,
2258		    "ICMP6 redirect sent from %s rejected; "
2259		    "hlim=%d (must be 255)\n",
2260		    ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2261		goto bad;
2262	}
2263    {
2264	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2265	struct sockaddr_in6 sin6;
2266	struct in6_addr *gw6;
2267
2268	bzero(&sin6, sizeof(sin6));
2269	sin6.sin6_family = AF_INET6;
2270	sin6.sin6_len = sizeof(struct sockaddr_in6);
2271	bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2272	rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL);
2273	if (rt) {
2274		if (rt->rt_gateway == NULL ||
2275		    rt->rt_gateway->sa_family != AF_INET6) {
2276			nd6log((LOG_ERR,
2277			    "ICMP6 redirect rejected; no route "
2278			    "with inet6 gateway found for redirect dst: %s\n",
2279			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2280			RTFREE_LOCKED(rt);
2281			goto bad;
2282		}
2283
2284		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2285		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2286			nd6log((LOG_ERR,
2287			    "ICMP6 redirect rejected; "
2288			    "not equal to gw-for-src=%s (must be same): "
2289			    "%s\n",
2290			    ip6_sprintf(ip6buf, gw6),
2291			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2292			RTFREE_LOCKED(rt);
2293			goto bad;
2294		}
2295	} else {
2296		nd6log((LOG_ERR,
2297		    "ICMP6 redirect rejected; "
2298		    "no route found for redirect dst: %s\n",
2299		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2300		goto bad;
2301	}
2302	RTFREE_LOCKED(rt);
2303	rt = NULL;
2304    }
2305	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2306		nd6log((LOG_ERR,
2307		    "ICMP6 redirect rejected; "
2308		    "redirect dst must be unicast: %s\n",
2309		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2310		goto bad;
2311	}
2312
2313	is_router = is_onlink = 0;
2314	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2315		is_router = 1;	/* router case */
2316	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2317		is_onlink = 1;	/* on-link destination case */
2318	if (!is_router && !is_onlink) {
2319		nd6log((LOG_ERR,
2320		    "ICMP6 redirect rejected; "
2321		    "neither router case nor onlink case: %s\n",
2322		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2323		goto bad;
2324	}
2325	/* validation passed */
2326
2327	icmp6len -= sizeof(*nd_rd);
2328	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2329	if (nd6_options(&ndopts) < 0) {
2330		nd6log((LOG_INFO, "icmp6_redirect_input: "
2331		    "invalid ND option, rejected: %s\n",
2332		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2333		/* nd6_options have incremented stats */
2334		goto freeit;
2335	}
2336
2337	if (ndopts.nd_opts_tgt_lladdr) {
2338		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2339		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2340	}
2341
2342	if (ndopts.nd_opts_rh) {
2343		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2344		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2345	}
2346
2347	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2348		nd6log((LOG_INFO,
2349		    "icmp6_redirect_input: lladdrlen mismatch for %s "
2350		    "(if %d, icmp6 packet %d): %s\n",
2351		    ip6_sprintf(ip6buf, &redtgt6),
2352		    ifp->if_addrlen, lladdrlen - 2,
2353		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2354		goto bad;
2355	}
2356
2357	/* RFC 2461 8.3 */
2358	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2359	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2360
2361	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2362		/* perform rtredirect */
2363		struct sockaddr_in6 sdst;
2364		struct sockaddr_in6 sgw;
2365		struct sockaddr_in6 ssrc;
2366
2367		bzero(&sdst, sizeof(sdst));
2368		bzero(&sgw, sizeof(sgw));
2369		bzero(&ssrc, sizeof(ssrc));
2370		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2371		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2372			sizeof(struct sockaddr_in6);
2373		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2374		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2375		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2376		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2377		    (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2378		    (struct sockaddr *)&ssrc);
2379	}
2380	/* finally update cached route in each socket via pfctlinput */
2381    {
2382	struct sockaddr_in6 sdst;
2383
2384	bzero(&sdst, sizeof(sdst));
2385	sdst.sin6_family = AF_INET6;
2386	sdst.sin6_len = sizeof(struct sockaddr_in6);
2387	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2388	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2389#ifdef IPSEC
2390	key_sa_routechange((struct sockaddr *)&sdst);
2391#endif /* IPSEC */
2392    }
2393
2394 freeit:
2395	m_freem(m);
2396	return;
2397
2398 bad:
2399	icmp6stat.icp6s_badredirect++;
2400	m_freem(m);
2401}
2402
2403void
2404icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2405{
2406	struct ifnet *ifp;	/* my outgoing interface */
2407	struct in6_addr *ifp_ll6;
2408	struct in6_addr *router_ll6;
2409	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2410	struct mbuf *m = NULL;	/* newly allocated one */
2411	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2412	struct nd_redirect *nd_rd;
2413	size_t maxlen;
2414	u_char *p;
2415	struct ifnet *outif = NULL;
2416	struct sockaddr_in6 src_sa;
2417
2418	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2419
2420	/* if we are not router, we don't send icmp6 redirect */
2421	if (!ip6_forwarding)
2422		goto fail;
2423
2424	/* sanity check */
2425	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2426		goto fail;
2427
2428	/*
2429	 * Address check:
2430	 *  the source address must identify a neighbor, and
2431	 *  the destination address must not be a multicast address
2432	 *  [RFC 2461, sec 8.2]
2433	 */
2434	sip6 = mtod(m0, struct ip6_hdr *);
2435	bzero(&src_sa, sizeof(src_sa));
2436	src_sa.sin6_family = AF_INET6;
2437	src_sa.sin6_len = sizeof(src_sa);
2438	src_sa.sin6_addr = sip6->ip6_src;
2439	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2440		goto fail;
2441	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2442		goto fail;	/* what should we do here? */
2443
2444	/* rate limit */
2445	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2446		goto fail;
2447
2448	/*
2449	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2450	 * we almost always ask for an mbuf cluster for simplicity.
2451	 * (MHLEN < IPV6_MMTU is almost always true)
2452	 */
2453#if IPV6_MMTU >= MCLBYTES
2454# error assumption failed about IPV6_MMTU and MCLBYTES
2455#endif
2456	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2457	if (m && IPV6_MMTU >= MHLEN)
2458		MCLGET(m, M_DONTWAIT);
2459	if (!m)
2460		goto fail;
2461	m->m_pkthdr.rcvif = NULL;
2462	m->m_len = 0;
2463	maxlen = M_TRAILINGSPACE(m);
2464	maxlen = min(IPV6_MMTU, maxlen);
2465	/* just for safety */
2466	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2467	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2468		goto fail;
2469	}
2470
2471	{
2472		/* get ip6 linklocal address for ifp(my outgoing interface). */
2473		struct in6_ifaddr *ia;
2474		if ((ia = in6ifa_ifpforlinklocal(ifp,
2475						 IN6_IFF_NOTREADY|
2476						 IN6_IFF_ANYCAST)) == NULL)
2477			goto fail;
2478		ifp_ll6 = &ia->ia_addr.sin6_addr;
2479	}
2480
2481	/* get ip6 linklocal address for the router. */
2482	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2483		struct sockaddr_in6 *sin6;
2484		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2485		router_ll6 = &sin6->sin6_addr;
2486		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2487			router_ll6 = (struct in6_addr *)NULL;
2488	} else
2489		router_ll6 = (struct in6_addr *)NULL;
2490
2491	/* ip6 */
2492	ip6 = mtod(m, struct ip6_hdr *);
2493	ip6->ip6_flow = 0;
2494	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2495	ip6->ip6_vfc |= IPV6_VERSION;
2496	/* ip6->ip6_plen will be set later */
2497	ip6->ip6_nxt = IPPROTO_ICMPV6;
2498	ip6->ip6_hlim = 255;
2499	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2500	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2501	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2502
2503	/* ND Redirect */
2504	nd_rd = (struct nd_redirect *)(ip6 + 1);
2505	nd_rd->nd_rd_type = ND_REDIRECT;
2506	nd_rd->nd_rd_code = 0;
2507	nd_rd->nd_rd_reserved = 0;
2508	if (rt->rt_flags & RTF_GATEWAY) {
2509		/*
2510		 * nd_rd->nd_rd_target must be a link-local address in
2511		 * better router cases.
2512		 */
2513		if (!router_ll6)
2514			goto fail;
2515		bcopy(router_ll6, &nd_rd->nd_rd_target,
2516		    sizeof(nd_rd->nd_rd_target));
2517		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2518		    sizeof(nd_rd->nd_rd_dst));
2519	} else {
2520		/* make sure redtgt == reddst */
2521		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2522		    sizeof(nd_rd->nd_rd_target));
2523		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2524		    sizeof(nd_rd->nd_rd_dst));
2525	}
2526
2527	p = (u_char *)(nd_rd + 1);
2528
2529	if (!router_ll6)
2530		goto nolladdropt;
2531
2532	{
2533		/* target lladdr option */
2534		struct rtentry *rt_router = NULL;
2535		int len;
2536		struct sockaddr_dl *sdl;
2537		struct nd_opt_hdr *nd_opt;
2538		char *lladdr;
2539
2540		rt_router = nd6_lookup(router_ll6, 0, ifp);
2541		if (!rt_router)
2542			goto nolladdropt;
2543		len = sizeof(*nd_opt) + ifp->if_addrlen;
2544		len = (len + 7) & ~7;	/* round by 8 */
2545		/* safety check */
2546		if (len + (p - (u_char *)ip6) > maxlen)
2547			goto nolladdropt;
2548		if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2549		    (rt_router->rt_flags & RTF_LLINFO) &&
2550		    (rt_router->rt_gateway->sa_family == AF_LINK) &&
2551		    (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2552		    sdl->sdl_alen) {
2553			nd_opt = (struct nd_opt_hdr *)p;
2554			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2555			nd_opt->nd_opt_len = len >> 3;
2556			lladdr = (char *)(nd_opt + 1);
2557			bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2558			p += len;
2559		}
2560	}
2561nolladdropt:;
2562
2563	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2564
2565	/* just to be safe */
2566#ifdef M_DECRYPTED	/*not openbsd*/
2567	if (m0->m_flags & M_DECRYPTED)
2568		goto noredhdropt;
2569#endif
2570	if (p - (u_char *)ip6 > maxlen)
2571		goto noredhdropt;
2572
2573	{
2574		/* redirected header option */
2575		int len;
2576		struct nd_opt_rd_hdr *nd_opt_rh;
2577
2578		/*
2579		 * compute the maximum size for icmp6 redirect header option.
2580		 * XXX room for auth header?
2581		 */
2582		len = maxlen - (p - (u_char *)ip6);
2583		len &= ~7;
2584
2585		/* This is just for simplicity. */
2586		if (m0->m_pkthdr.len != m0->m_len) {
2587			if (m0->m_next) {
2588				m_freem(m0->m_next);
2589				m0->m_next = NULL;
2590			}
2591			m0->m_pkthdr.len = m0->m_len;
2592		}
2593
2594		/*
2595		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2596		 * about padding/truncate rule for the original IP packet.
2597		 * From the discussion on IPv6imp in Feb 1999,
2598		 * the consensus was:
2599		 * - "attach as much as possible" is the goal
2600		 * - pad if not aligned (original size can be guessed by
2601		 *   original ip6 header)
2602		 * Following code adds the padding if it is simple enough,
2603		 * and truncates if not.
2604		 */
2605		if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2606			panic("assumption failed in %s:%d", __FILE__,
2607			    __LINE__);
2608
2609		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2610			/* not enough room, truncate */
2611			m0->m_pkthdr.len = m0->m_len = len -
2612			    sizeof(*nd_opt_rh);
2613		} else {
2614			/* enough room, pad or truncate */
2615			size_t extra;
2616
2617			extra = m0->m_pkthdr.len % 8;
2618			if (extra) {
2619				/* pad if easy enough, truncate if not */
2620				if (8 - extra <= M_TRAILINGSPACE(m0)) {
2621					/* pad */
2622					m0->m_len += (8 - extra);
2623					m0->m_pkthdr.len += (8 - extra);
2624				} else {
2625					/* truncate */
2626					m0->m_pkthdr.len -= extra;
2627					m0->m_len -= extra;
2628				}
2629			}
2630			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2631			m0->m_pkthdr.len = m0->m_len = len -
2632			    sizeof(*nd_opt_rh);
2633		}
2634
2635		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2636		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2637		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2638		nd_opt_rh->nd_opt_rh_len = len >> 3;
2639		p += sizeof(*nd_opt_rh);
2640		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2641
2642		/* connect m0 to m */
2643		m_tag_delete_chain(m0, NULL);
2644		m0->m_flags &= ~M_PKTHDR;
2645		m->m_next = m0;
2646		m->m_pkthdr.len = m->m_len + m0->m_len;
2647		m0 = NULL;
2648	}
2649noredhdropt:;
2650	if (m0) {
2651		m_freem(m0);
2652		m0 = NULL;
2653	}
2654
2655	/* XXX: clear embedded link IDs in the inner header */
2656	in6_clearscope(&sip6->ip6_src);
2657	in6_clearscope(&sip6->ip6_dst);
2658	in6_clearscope(&nd_rd->nd_rd_target);
2659	in6_clearscope(&nd_rd->nd_rd_dst);
2660
2661	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2662
2663	nd_rd->nd_rd_cksum = 0;
2664	nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2665	    sizeof(*ip6), ntohs(ip6->ip6_plen));
2666
2667	/* send the packet to outside... */
2668	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2669	if (outif) {
2670		icmp6_ifstat_inc(outif, ifs6_out_msg);
2671		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2672	}
2673	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2674
2675	return;
2676
2677fail:
2678	if (m)
2679		m_freem(m);
2680	if (m0)
2681		m_freem(m0);
2682}
2683
2684/*
2685 * ICMPv6 socket option processing.
2686 */
2687int
2688icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2689{
2690	int error = 0;
2691	int optlen;
2692	struct inpcb *inp = sotoinpcb(so);
2693	int level, op, optname;
2694
2695	if (sopt) {
2696		level = sopt->sopt_level;
2697		op = sopt->sopt_dir;
2698		optname = sopt->sopt_name;
2699		optlen = sopt->sopt_valsize;
2700	} else
2701		level = op = optname = optlen = 0;
2702
2703	if (level != IPPROTO_ICMPV6) {
2704		return EINVAL;
2705	}
2706
2707	switch (op) {
2708	case PRCO_SETOPT:
2709		switch (optname) {
2710		case ICMP6_FILTER:
2711		    {
2712			struct icmp6_filter *p;
2713
2714			if (optlen != sizeof(*p)) {
2715				error = EMSGSIZE;
2716				break;
2717			}
2718			if (inp->in6p_icmp6filt == NULL) {
2719				error = EINVAL;
2720				break;
2721			}
2722			error = sooptcopyin(sopt, inp->in6p_icmp6filt, optlen,
2723				optlen);
2724			break;
2725		    }
2726
2727		default:
2728			error = ENOPROTOOPT;
2729			break;
2730		}
2731		break;
2732
2733	case PRCO_GETOPT:
2734		switch (optname) {
2735		case ICMP6_FILTER:
2736		    {
2737			if (inp->in6p_icmp6filt == NULL) {
2738				error = EINVAL;
2739				break;
2740			}
2741			error = sooptcopyout(sopt, inp->in6p_icmp6filt,
2742				sizeof(struct icmp6_filter));
2743			break;
2744		    }
2745
2746		default:
2747			error = ENOPROTOOPT;
2748			break;
2749		}
2750		break;
2751	}
2752
2753	return (error);
2754}
2755
2756/*
2757 * Perform rate limit check.
2758 * Returns 0 if it is okay to send the icmp6 packet.
2759 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2760 * limitation.
2761 *
2762 * XXX per-destination/type check necessary?
2763 *
2764 * dst - not used at this moment
2765 * type - not used at this moment
2766 * code - not used at this moment
2767 */
2768static int
2769icmp6_ratelimit(const struct in6_addr *dst, const int type,
2770    const int code)
2771{
2772	int ret;
2773
2774	ret = 0;	/* okay to send */
2775
2776	/* PPS limit */
2777	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2778	    icmp6errppslim)) {
2779		/* The packet is subject to rate limit */
2780		ret++;
2781	}
2782
2783	return ret;
2784}
2785