icmp6.c revision 121358
1/*	$FreeBSD: head/sys/netinet6/icmp6.c 121358 2003-10-22 19:23:51Z 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 * 3. All advertising materials mentioning features or use of this software
46 *    must display the following acknowledgement:
47 *	This product includes software developed by the University of
48 *	California, Berkeley and its contributors.
49 * 4. Neither the name of the University nor the names of its contributors
50 *    may be used to endorse or promote products derived from this software
51 *    without specific prior written permission.
52 *
53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63 * SUCH DAMAGE.
64 *
65 *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
66 */
67
68#include "opt_inet.h"
69#include "opt_inet6.h"
70#include "opt_ipsec.h"
71
72#include <sys/param.h>
73#include <sys/domain.h>
74#include <sys/kernel.h>
75#include <sys/lock.h>
76#include <sys/malloc.h>
77#include <sys/mbuf.h>
78#include <sys/protosw.h>
79#include <sys/signalvar.h>
80#include <sys/socket.h>
81#include <sys/socketvar.h>
82#include <sys/sx.h>
83#include <sys/syslog.h>
84#include <sys/systm.h>
85#include <sys/time.h>
86
87#include <net/if.h>
88#include <net/if_dl.h>
89#include <net/if_types.h>
90#include <net/route.h>
91
92#include <netinet/in.h>
93#include <netinet/in_pcb.h>
94#include <netinet/in_var.h>
95#include <netinet/ip6.h>
96#include <netinet/icmp6.h>
97#include <netinet6/in6_ifattach.h>
98#include <netinet6/in6_pcb.h>
99#include <netinet6/ip6protosw.h>
100#include <netinet6/ip6_var.h>
101#include <netinet6/mld6_var.h>
102#include <netinet6/nd6.h>
103
104#ifdef IPSEC
105#include <netinet6/ipsec.h>
106#include <netkey/key.h>
107#endif
108
109#ifdef FAST_IPSEC
110#include <netipsec/ipsec.h>
111#include <netipsec/key.h>
112#define	IPSEC
113#endif
114
115#include <net/net_osdep.h>
116
117extern struct domain inet6domain;
118
119struct icmp6stat icmp6stat;
120
121extern struct inpcbhead ripcb;
122extern int icmp6errppslim;
123static int icmp6errpps_count = 0;
124static struct timeval icmp6errppslim_last;
125extern int icmp6_nodeinfo;
126
127static void icmp6_errcount __P((struct icmp6errstat *, int, int));
128static int icmp6_rip6_input __P((struct mbuf **, int));
129static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int));
130static const char *icmp6_redirect_diag __P((struct in6_addr *,
131	struct in6_addr *, struct in6_addr *));
132static struct mbuf *ni6_input __P((struct mbuf *, int));
133static struct mbuf *ni6_nametodns __P((const char *, int, int));
134static int ni6_dnsmatch __P((const char *, int, const char *, int));
135static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
136			  struct ifnet **, char *));
137static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
138				struct ifnet *, int));
139static int icmp6_notify_error __P((struct mbuf *, int, int, int));
140
141#ifdef COMPAT_RFC1885
142static struct route_in6 icmp6_reflect_rt;
143#endif
144
145
146void
147icmp6_init()
148{
149	mld6_init();
150}
151
152static void
153icmp6_errcount(stat, type, code)
154	struct icmp6errstat *stat;
155	int type, code;
156{
157	switch (type) {
158	case ICMP6_DST_UNREACH:
159		switch (code) {
160		case ICMP6_DST_UNREACH_NOROUTE:
161			stat->icp6errs_dst_unreach_noroute++;
162			return;
163		case ICMP6_DST_UNREACH_ADMIN:
164			stat->icp6errs_dst_unreach_admin++;
165			return;
166		case ICMP6_DST_UNREACH_BEYONDSCOPE:
167			stat->icp6errs_dst_unreach_beyondscope++;
168			return;
169		case ICMP6_DST_UNREACH_ADDR:
170			stat->icp6errs_dst_unreach_addr++;
171			return;
172		case ICMP6_DST_UNREACH_NOPORT:
173			stat->icp6errs_dst_unreach_noport++;
174			return;
175		}
176		break;
177	case ICMP6_PACKET_TOO_BIG:
178		stat->icp6errs_packet_too_big++;
179		return;
180	case ICMP6_TIME_EXCEEDED:
181		switch (code) {
182		case ICMP6_TIME_EXCEED_TRANSIT:
183			stat->icp6errs_time_exceed_transit++;
184			return;
185		case ICMP6_TIME_EXCEED_REASSEMBLY:
186			stat->icp6errs_time_exceed_reassembly++;
187			return;
188		}
189		break;
190	case ICMP6_PARAM_PROB:
191		switch (code) {
192		case ICMP6_PARAMPROB_HEADER:
193			stat->icp6errs_paramprob_header++;
194			return;
195		case ICMP6_PARAMPROB_NEXTHEADER:
196			stat->icp6errs_paramprob_nextheader++;
197			return;
198		case ICMP6_PARAMPROB_OPTION:
199			stat->icp6errs_paramprob_option++;
200			return;
201		}
202		break;
203	case ND_REDIRECT:
204		stat->icp6errs_redirect++;
205		return;
206	}
207	stat->icp6errs_unknown++;
208}
209
210/*
211 * Generate an error packet of type error in response to bad IP6 packet.
212 */
213void
214icmp6_error(m, type, code, param)
215	struct mbuf *m;
216	int type, code, param;
217{
218	struct ip6_hdr *oip6, *nip6;
219	struct icmp6_hdr *icmp6;
220	u_int preplen;
221	int off;
222	int nxt;
223
224	icmp6stat.icp6s_error++;
225
226	/* count per-type-code statistics */
227	icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
228
229#ifdef M_DECRYPTED	/*not openbsd*/
230	if (m->m_flags & M_DECRYPTED) {
231		icmp6stat.icp6s_canterror++;
232		goto freeit;
233	}
234#endif
235
236#ifndef PULLDOWN_TEST
237	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
238#else
239	if (m->m_len < sizeof(struct ip6_hdr)) {
240		m = m_pullup(m, sizeof(struct ip6_hdr));
241		if (m == NULL)
242			return;
243	}
244#endif
245	oip6 = mtod(m, struct ip6_hdr *);
246
247	/*
248	 * Multicast destination check. For unrecognized option errors,
249	 * this check has already done in ip6_unknown_opt(), so we can
250	 * check only for other errors.
251	 */
252	if ((m->m_flags & (M_BCAST|M_MCAST) ||
253	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
254	    (type != ICMP6_PACKET_TOO_BIG &&
255	     (type != ICMP6_PARAM_PROB ||
256	      code != ICMP6_PARAMPROB_OPTION)))
257		goto freeit;
258
259	/* Source address check. XXX: the case of anycast source? */
260	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
261	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
262		goto freeit;
263
264	/*
265	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
266	 * don't do it.
267	 */
268	nxt = -1;
269	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
270	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
271		struct icmp6_hdr *icp;
272
273#ifndef PULLDOWN_TEST
274		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
275		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
276#else
277		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
278			sizeof(*icp));
279		if (icp == NULL) {
280			icmp6stat.icp6s_tooshort++;
281			return;
282		}
283#endif
284		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
285		    icp->icmp6_type == ND_REDIRECT) {
286			/*
287			 * ICMPv6 error
288			 * Special case: for redirect (which is
289			 * informational) we must not send icmp6 error.
290			 */
291			icmp6stat.icp6s_canterror++;
292			goto freeit;
293		} else {
294			/* ICMPv6 informational - send the error */
295		}
296	} else {
297		/* non-ICMPv6 - send the error */
298	}
299
300	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
301
302	/* Finally, do rate limitation check. */
303	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
304		icmp6stat.icp6s_toofreq++;
305		goto freeit;
306	}
307
308	/*
309	 * OK, ICMP6 can be generated.
310	 */
311
312	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
313		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
314
315	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
316	M_PREPEND(m, preplen, M_DONTWAIT);
317	if (m && m->m_len < preplen)
318		m = m_pullup(m, preplen);
319	if (m == NULL) {
320		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
321		return;
322	}
323
324	nip6 = mtod(m, struct ip6_hdr *);
325	nip6->ip6_src  = oip6->ip6_src;
326	nip6->ip6_dst  = oip6->ip6_dst;
327
328	in6_clearscope(&oip6->ip6_src);
329	in6_clearscope(&oip6->ip6_dst);
330
331	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
332	icmp6->icmp6_type = type;
333	icmp6->icmp6_code = code;
334	icmp6->icmp6_pptr = htonl((u_int32_t)param);
335
336	/*
337	 * icmp6_reflect() is designed to be in the input path.
338	 * icmp6_error() can be called from both input and outut path,
339	 * and if we are in output path rcvif could contain bogus value.
340	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
341	 * information in ip header (nip6).
342	 */
343	m->m_pkthdr.rcvif = NULL;
344
345	icmp6stat.icp6s_outhist[type]++;
346	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
347
348	return;
349
350  freeit:
351	/*
352	 * If we can't tell wheter or not we can generate ICMP6, free it.
353	 */
354	m_freem(m);
355}
356
357/*
358 * Process a received ICMP6 message.
359 */
360int
361icmp6_input(mp, offp, proto)
362	struct mbuf **mp;
363	int *offp, proto;
364{
365	struct mbuf *m = *mp, *n;
366	struct ip6_hdr *ip6, *nip6;
367	struct icmp6_hdr *icmp6, *nicmp6;
368	int off = *offp;
369	int icmp6len = m->m_pkthdr.len - *offp;
370	int code, sum, noff;
371
372#ifndef PULLDOWN_TEST
373	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
374	/* m might change if M_LOOP.  So, call mtod after this */
375#endif
376
377	/*
378	 * Locate icmp6 structure in mbuf, and check
379	 * that not corrupted and of at least minimum length
380	 */
381
382	ip6 = mtod(m, struct ip6_hdr *);
383	if (icmp6len < sizeof(struct icmp6_hdr)) {
384		icmp6stat.icp6s_tooshort++;
385		goto freeit;
386	}
387
388	/*
389	 * calculate the checksum
390	 */
391#ifndef PULLDOWN_TEST
392	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
393#else
394	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
395	if (icmp6 == NULL) {
396		icmp6stat.icp6s_tooshort++;
397		return IPPROTO_DONE;
398	}
399#endif
400	code = icmp6->icmp6_code;
401
402	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
403		nd6log((LOG_ERR,
404		    "ICMP6 checksum error(%d|%x) %s\n",
405		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
406		icmp6stat.icp6s_checksum++;
407		goto freeit;
408	}
409
410	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
411		/*
412		 * Deliver very specific ICMP6 type only.
413		 * This is important to deilver TOOBIG.  Otherwise PMTUD
414		 * will not work.
415		 */
416		switch (icmp6->icmp6_type) {
417		case ICMP6_DST_UNREACH:
418		case ICMP6_PACKET_TOO_BIG:
419		case ICMP6_TIME_EXCEEDED:
420			break;
421		default:
422			goto freeit;
423		}
424	}
425
426	icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
427	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
428	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
429		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
430
431	switch (icmp6->icmp6_type) {
432	case ICMP6_DST_UNREACH:
433		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
434		switch (code) {
435		case ICMP6_DST_UNREACH_NOROUTE:
436			code = PRC_UNREACH_NET;
437			break;
438		case ICMP6_DST_UNREACH_ADMIN:
439			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
440			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
441			break;
442		case ICMP6_DST_UNREACH_ADDR:
443			code = PRC_HOSTDEAD;
444			break;
445#ifdef COMPAT_RFC1885
446		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
447			code = PRC_UNREACH_SRCFAIL;
448			break;
449#else
450		case ICMP6_DST_UNREACH_BEYONDSCOPE:
451			/* I mean "source address was incorrect." */
452			code = PRC_PARAMPROB;
453			break;
454#endif
455		case ICMP6_DST_UNREACH_NOPORT:
456			code = PRC_UNREACH_PORT;
457			break;
458		default:
459			goto badcode;
460		}
461		goto deliver;
462		break;
463
464	case ICMP6_PACKET_TOO_BIG:
465		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
466		if (code != 0)
467			goto badcode;
468
469		code = PRC_MSGSIZE;
470
471		/*
472		 * Updating the path MTU will be done after examining
473		 * intermediate extension headers.
474		 */
475		goto deliver;
476		break;
477
478	case ICMP6_TIME_EXCEEDED:
479		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
480		switch (code) {
481		case ICMP6_TIME_EXCEED_TRANSIT:
482		case ICMP6_TIME_EXCEED_REASSEMBLY:
483			code += PRC_TIMXCEED_INTRANS;
484			break;
485		default:
486			goto badcode;
487		}
488		goto deliver;
489		break;
490
491	case ICMP6_PARAM_PROB:
492		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
493		switch (code) {
494		case ICMP6_PARAMPROB_NEXTHEADER:
495			code = PRC_UNREACH_PROTOCOL;
496			break;
497		case ICMP6_PARAMPROB_HEADER:
498		case ICMP6_PARAMPROB_OPTION:
499			code = PRC_PARAMPROB;
500			break;
501		default:
502			goto badcode;
503		}
504		goto deliver;
505		break;
506
507	case ICMP6_ECHO_REQUEST:
508		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
509		if (code != 0)
510			goto badcode;
511		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
512			/* Give up remote */
513			break;
514		}
515		if ((n->m_flags & M_EXT) != 0
516		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
517			struct mbuf *n0 = n;
518			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
519			int n0len;
520
521			/*
522			 * Prepare an internal mbuf.  m_pullup() doesn't
523			 * always copy the length we specified.
524			 */
525			if (maxlen >= MCLBYTES) {
526				/* Give up remote */
527				m_freem(n0);
528				break;
529			}
530			MGETHDR(n, M_DONTWAIT, n0->m_type);
531			n0len = n0->m_pkthdr.len;	/* save for use below */
532			if (n)
533				M_MOVE_PKTHDR(n, n0);
534			if (n && maxlen >= MHLEN) {
535				MCLGET(n, M_DONTWAIT);
536				if ((n->m_flags & M_EXT) == 0) {
537					m_free(n);
538					n = NULL;
539				}
540			}
541			if (n == NULL) {
542				/* Give up remote */
543				m_freem(n0);
544				break;
545			}
546			/*
547			 * Copy IPv6 and ICMPv6 only.
548			 */
549			nip6 = mtod(n, struct ip6_hdr *);
550			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
551			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
552			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
553			noff = sizeof(struct ip6_hdr);
554			/* new mbuf contains only ipv6+icmpv6 headers */
555			n->m_len = noff + sizeof(struct icmp6_hdr);
556			/*
557			 * Adjust mbuf.  ip6_plen will be adjusted in
558			 * ip6_output().
559			 */
560			m_adj(n0, off + sizeof(struct icmp6_hdr));
561			/* recalculate complete packet size */
562			n->m_pkthdr.len = n0len + (noff - off);
563			n->m_next = n0;
564		} else {
565			nip6 = mtod(n, struct ip6_hdr *);
566			nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
567			noff = off;
568		}
569		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
570		nicmp6->icmp6_code = 0;
571		if (n) {
572			icmp6stat.icp6s_reflect++;
573			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
574			icmp6_reflect(n, noff);
575		}
576		break;
577
578	case ICMP6_ECHO_REPLY:
579		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
580		if (code != 0)
581			goto badcode;
582		break;
583
584	case MLD_LISTENER_QUERY:
585	case MLD_LISTENER_REPORT:
586		if (icmp6len < sizeof(struct mld_hdr))
587			goto badlen;
588		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
589			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
590		else
591			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
592		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
593			/* give up local */
594			mld6_input(m, off);
595			m = NULL;
596			goto freeit;
597		}
598		mld6_input(n, off);
599		/* m stays. */
600		break;
601
602	case MLD_LISTENER_DONE:
603		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
604		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
605			goto badlen;
606		break;		/* nothing to be done in kernel */
607
608	case MLD_MTRACE_RESP:
609	case MLD_MTRACE:
610		/* XXX: these two are experimental.  not officially defined. */
611		/* XXX: per-interface statistics? */
612		break;		/* just pass it to applications */
613
614	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
615	    {
616		enum { WRU, FQDN } mode;
617
618		if (!icmp6_nodeinfo)
619			break;
620
621		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
622			mode = WRU;
623		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
624			mode = FQDN;
625		else
626			goto badlen;
627
628#define hostnamelen	strlen(hostname)
629		if (mode == FQDN) {
630#ifndef PULLDOWN_TEST
631			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
632			    IPPROTO_DONE);
633#endif
634			n = m_copy(m, 0, M_COPYALL);
635			if (n)
636				n = ni6_input(n, off);
637			/* XXX meaningless if n == NULL */
638			noff = sizeof(struct ip6_hdr);
639		} else {
640			u_char *p;
641			int maxlen, maxhlen;
642
643			if ((icmp6_nodeinfo & 5) != 5)
644				break;
645
646			if (code != 0)
647				goto badcode;
648			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
649			if (maxlen >= MCLBYTES) {
650				/* Give up remote */
651				break;
652			}
653			MGETHDR(n, M_DONTWAIT, m->m_type);
654			if (n && maxlen > MHLEN) {
655				MCLGET(n, M_DONTWAIT);
656				if ((n->m_flags & M_EXT) == 0) {
657					m_free(n);
658					n = NULL;
659				}
660			}
661			if (!m_dup_pkthdr(n, m, M_DONTWAIT)) {
662				/*
663				 * Previous code did a blind M_COPY_PKTHDR
664				 * and said "just for rcvif".  If true, then
665				 * we could tolerate the dup failing (due to
666				 * the deep copy of the tag chain).  For now
667				 * be conservative and just fail.
668				 */
669				m_free(n);
670				n = NULL;
671			}
672			if (n == NULL) {
673				/* Give up remote */
674				break;
675			}
676			n->m_pkthdr.rcvif = NULL;
677			n->m_len = 0;
678			maxhlen = M_TRAILINGSPACE(n) - maxlen;
679			if (maxhlen > hostnamelen)
680				maxhlen = hostnamelen;
681			/*
682			 * Copy IPv6 and ICMPv6 only.
683			 */
684			nip6 = mtod(n, struct ip6_hdr *);
685			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
686			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
687			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
688			p = (u_char *)(nicmp6 + 1);
689			bzero(p, 4);
690			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
691			noff = sizeof(struct ip6_hdr);
692			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
693				sizeof(struct icmp6_hdr) + 4 + maxhlen;
694			nicmp6->icmp6_type = ICMP6_WRUREPLY;
695			nicmp6->icmp6_code = 0;
696		}
697#undef hostnamelen
698		if (n) {
699			icmp6stat.icp6s_reflect++;
700			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
701			icmp6_reflect(n, noff);
702		}
703		break;
704	    }
705
706	case ICMP6_WRUREPLY:
707		if (code != 0)
708			goto badcode;
709		break;
710
711	case ND_ROUTER_SOLICIT:
712		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
713		if (code != 0)
714			goto badcode;
715		if (icmp6len < sizeof(struct nd_router_solicit))
716			goto badlen;
717		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
718			/* give up local */
719			nd6_rs_input(m, off, icmp6len);
720			m = NULL;
721			goto freeit;
722		}
723		nd6_rs_input(n, off, icmp6len);
724		/* m stays. */
725		break;
726
727	case ND_ROUTER_ADVERT:
728		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
729		if (code != 0)
730			goto badcode;
731		if (icmp6len < sizeof(struct nd_router_advert))
732			goto badlen;
733		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
734			/* give up local */
735			nd6_ra_input(m, off, icmp6len);
736			m = NULL;
737			goto freeit;
738		}
739		nd6_ra_input(n, off, icmp6len);
740		/* m stays. */
741		break;
742
743	case ND_NEIGHBOR_SOLICIT:
744		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
745		if (code != 0)
746			goto badcode;
747		if (icmp6len < sizeof(struct nd_neighbor_solicit))
748			goto badlen;
749		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
750			/* give up local */
751			nd6_ns_input(m, off, icmp6len);
752			m = NULL;
753			goto freeit;
754		}
755		nd6_ns_input(n, off, icmp6len);
756		/* m stays. */
757		break;
758
759	case ND_NEIGHBOR_ADVERT:
760		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
761		if (code != 0)
762			goto badcode;
763		if (icmp6len < sizeof(struct nd_neighbor_advert))
764			goto badlen;
765		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
766			/* give up local */
767			nd6_na_input(m, off, icmp6len);
768			m = NULL;
769			goto freeit;
770		}
771		nd6_na_input(n, off, icmp6len);
772		/* m stays. */
773		break;
774
775	case ND_REDIRECT:
776		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
777		if (code != 0)
778			goto badcode;
779		if (icmp6len < sizeof(struct nd_redirect))
780			goto badlen;
781		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
782			/* give up local */
783			icmp6_redirect_input(m, off);
784			m = NULL;
785			goto freeit;
786		}
787		icmp6_redirect_input(n, off);
788		/* m stays. */
789		break;
790
791	case ICMP6_ROUTER_RENUMBERING:
792		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
793		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
794			goto badcode;
795		if (icmp6len < sizeof(struct icmp6_router_renum))
796			goto badlen;
797		break;
798
799	default:
800		nd6log((LOG_DEBUG,
801		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
802		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
803		    ip6_sprintf(&ip6->ip6_dst),
804		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
805		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
806			/* ICMPv6 error: MUST deliver it by spec... */
807			code = PRC_NCMDS;
808			/* deliver */
809		} else {
810			/* ICMPv6 informational: MUST not deliver */
811			break;
812		}
813	deliver:
814		if (icmp6_notify_error(m, off, icmp6len, code)) {
815			/* In this case, m should've been freed. */
816			return (IPPROTO_DONE);
817		}
818		break;
819
820	badcode:
821		icmp6stat.icp6s_badcode++;
822		break;
823
824	badlen:
825		icmp6stat.icp6s_badlen++;
826		break;
827	}
828
829	/* deliver the packet to appropriate sockets */
830	icmp6_rip6_input(&m, *offp);
831
832	return IPPROTO_DONE;
833
834 freeit:
835	m_freem(m);
836	return IPPROTO_DONE;
837}
838
839static int
840icmp6_notify_error(m, off, icmp6len, code)
841	struct mbuf *m;
842	int off, icmp6len, code;
843{
844	struct icmp6_hdr *icmp6;
845	struct ip6_hdr *eip6;
846	u_int32_t notifymtu;
847	struct sockaddr_in6 icmp6src, icmp6dst;
848
849	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
850		icmp6stat.icp6s_tooshort++;
851		goto freeit;
852	}
853#ifndef PULLDOWN_TEST
854	IP6_EXTHDR_CHECK(m, off,
855	    sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
856	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
857#else
858	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
859	    sizeof(*icmp6) + sizeof(struct ip6_hdr));
860	if (icmp6 == NULL) {
861		icmp6stat.icp6s_tooshort++;
862		return (-1);
863	}
864#endif
865	eip6 = (struct ip6_hdr *)(icmp6 + 1);
866
867	/* Detect the upper level protocol */
868	{
869		void (*ctlfunc) __P((int, struct sockaddr *, void *));
870		u_int8_t nxt = eip6->ip6_nxt;
871		int eoff = off + sizeof(struct icmp6_hdr) +
872		    sizeof(struct ip6_hdr);
873		struct ip6ctlparam ip6cp;
874		struct in6_addr *finaldst = NULL;
875		int icmp6type = icmp6->icmp6_type;
876		struct ip6_frag *fh;
877		struct ip6_rthdr *rth;
878		struct ip6_rthdr0 *rth0;
879		int rthlen;
880
881		while (1) { /* XXX: should avoid infinite loop explicitly? */
882			struct ip6_ext *eh;
883
884			switch (nxt) {
885			case IPPROTO_HOPOPTS:
886			case IPPROTO_DSTOPTS:
887			case IPPROTO_AH:
888#ifndef PULLDOWN_TEST
889				IP6_EXTHDR_CHECK(m, 0,
890				    eoff + sizeof(struct ip6_ext), -1);
891				eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
892#else
893				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
894				    eoff, sizeof(*eh));
895				if (eh == NULL) {
896					icmp6stat.icp6s_tooshort++;
897					return (-1);
898				}
899#endif
900
901				if (nxt == IPPROTO_AH)
902					eoff += (eh->ip6e_len + 2) << 2;
903				else
904					eoff += (eh->ip6e_len + 1) << 3;
905				nxt = eh->ip6e_nxt;
906				break;
907			case IPPROTO_ROUTING:
908				/*
909				 * When the erroneous packet contains a
910				 * routing header, we should examine the
911				 * header to determine the final destination.
912				 * Otherwise, we can't properly update
913				 * information that depends on the final
914				 * destination (e.g. path MTU).
915				 */
916#ifndef PULLDOWN_TEST
917				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
918				rth = (struct ip6_rthdr *)
919				    (mtod(m, caddr_t) + eoff);
920#else
921				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
922				    eoff, sizeof(*rth));
923				if (rth == NULL) {
924					icmp6stat.icp6s_tooshort++;
925					return (-1);
926				}
927#endif
928				rthlen = (rth->ip6r_len + 1) << 3;
929				/*
930				 * XXX: currently there is no
931				 * officially defined type other
932				 * than type-0.
933				 * Note that if the segment left field
934				 * is 0, all intermediate hops must
935				 * have been passed.
936				 */
937				if (rth->ip6r_segleft &&
938				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
939					int hops;
940
941#ifndef PULLDOWN_TEST
942					IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
943					rth0 = (struct ip6_rthdr0 *)
944					    (mtod(m, caddr_t) + eoff);
945#else
946					IP6_EXTHDR_GET(rth0,
947					    struct ip6_rthdr0 *, m,
948					    eoff, rthlen);
949					if (rth0 == NULL) {
950						icmp6stat.icp6s_tooshort++;
951						return (-1);
952					}
953#endif
954					/* just ignore a bogus header */
955					if ((rth0->ip6r0_len % 2) == 0 &&
956					    (hops = rth0->ip6r0_len/2))
957						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
958				}
959				eoff += rthlen;
960				nxt = rth->ip6r_nxt;
961				break;
962			case IPPROTO_FRAGMENT:
963#ifndef PULLDOWN_TEST
964				IP6_EXTHDR_CHECK(m, 0, eoff +
965				    sizeof(struct ip6_frag), -1);
966				fh = (struct ip6_frag *)(mtod(m, caddr_t) +
967				    eoff);
968#else
969				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
970				    eoff, sizeof(*fh));
971				if (fh == NULL) {
972					icmp6stat.icp6s_tooshort++;
973					return (-1);
974				}
975#endif
976				/*
977				 * Data after a fragment header is meaningless
978				 * unless it is the first fragment, but
979				 * we'll go to the notify label for path MTU
980				 * discovery.
981				 */
982				if (fh->ip6f_offlg & IP6F_OFF_MASK)
983					goto notify;
984
985				eoff += sizeof(struct ip6_frag);
986				nxt = fh->ip6f_nxt;
987				break;
988			default:
989				/*
990				 * This case includes ESP and the No Next
991				 * Header.  In such cases going to the notify
992				 * label does not have any meaning
993				 * (i.e. ctlfunc will be NULL), but we go
994				 * anyway since we might have to update
995				 * path MTU information.
996				 */
997				goto notify;
998			}
999		}
1000	  notify:
1001#ifndef PULLDOWN_TEST
1002		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1003#else
1004		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1005		    sizeof(*icmp6) + sizeof(struct ip6_hdr));
1006		if (icmp6 == NULL) {
1007			icmp6stat.icp6s_tooshort++;
1008			return (-1);
1009		}
1010#endif
1011
1012		/*
1013		 * retrieve parameters from the inner IPv6 header, and convert
1014		 * them into sockaddr structures.
1015		 * XXX: there is no guarantee that the source or destination
1016		 * addresses of the inner packet are in the same scope as
1017		 * the addresses of the icmp packet.  But there is no other
1018		 * way to determine the zone.
1019		 */
1020		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1021
1022		bzero(&icmp6dst, sizeof(icmp6dst));
1023		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1024		icmp6dst.sin6_family = AF_INET6;
1025		if (finaldst == NULL)
1026			icmp6dst.sin6_addr = eip6->ip6_dst;
1027		else
1028			icmp6dst.sin6_addr = *finaldst;
1029		if (in6_addr2zoneid(m->m_pkthdr.rcvif, &icmp6dst.sin6_addr,
1030		    &icmp6dst.sin6_scope_id))
1031			goto freeit;
1032		if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1033				   NULL, NULL)) {
1034			/* should be impossbile */
1035			nd6log((LOG_DEBUG,
1036			    "icmp6_notify_error: in6_embedscope failed\n"));
1037			goto freeit;
1038		}
1039
1040		/*
1041		 * retrieve parameters from the inner IPv6 header, and convert
1042		 * them into sockaddr structures.
1043		 */
1044		bzero(&icmp6src, sizeof(icmp6src));
1045		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1046		icmp6src.sin6_family = AF_INET6;
1047		icmp6src.sin6_addr = eip6->ip6_src;
1048		if (in6_addr2zoneid(m->m_pkthdr.rcvif, &icmp6src.sin6_addr,
1049		    &icmp6src.sin6_scope_id)) {
1050			goto freeit;
1051		}
1052		if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1053				   NULL, NULL)) {
1054			/* should be impossbile */
1055			nd6log((LOG_DEBUG,
1056			    "icmp6_notify_error: in6_embedscope failed\n"));
1057			goto freeit;
1058		}
1059		icmp6src.sin6_flowinfo = (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1060
1061		if (finaldst == NULL)
1062			finaldst = &eip6->ip6_dst;
1063		ip6cp.ip6c_m = m;
1064		ip6cp.ip6c_icmp6 = icmp6;
1065		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1066		ip6cp.ip6c_off = eoff;
1067		ip6cp.ip6c_finaldst = finaldst;
1068		ip6cp.ip6c_src = &icmp6src;
1069		ip6cp.ip6c_nxt = nxt;
1070
1071		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1072			notifymtu = ntohl(icmp6->icmp6_mtu);
1073			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1074			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1075		}
1076
1077		ctlfunc = (void (*) __P((int, struct sockaddr *, void *)))
1078		    (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1079		if (ctlfunc) {
1080			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1081			    &ip6cp);
1082		}
1083	}
1084	return (0);
1085
1086  freeit:
1087	m_freem(m);
1088	return (-1);
1089}
1090
1091void
1092icmp6_mtudisc_update(ip6cp, validated)
1093	struct ip6ctlparam *ip6cp;
1094	int validated;
1095{
1096	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1097	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1098	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1099	u_int mtu = ntohl(icmp6->icmp6_mtu);
1100	struct rtentry *rt = NULL;
1101	struct sockaddr_in6 sin6;
1102
1103	if (!validated)
1104		return;
1105
1106	bzero(&sin6, sizeof(sin6));
1107	sin6.sin6_family = PF_INET6;
1108	sin6.sin6_len = sizeof(struct sockaddr_in6);
1109	sin6.sin6_addr = *dst;
1110	/* XXX normally, this won't happen */
1111	if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1112		sin6.sin6_addr.s6_addr16[1] =
1113		    htons(m->m_pkthdr.rcvif->if_index);
1114	}
1115	/* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1116	rt = rtalloc1((struct sockaddr *)&sin6, 0,
1117		      RTF_CLONING | RTF_PRCLONING);
1118
1119	if (rt && (rt->rt_flags & RTF_HOST) &&
1120	    !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1121		if (mtu < IPV6_MMTU) {
1122				/* xxx */
1123			rt->rt_rmx.rmx_locks |= RTV_MTU;
1124		} else if (mtu < rt->rt_ifp->if_mtu &&
1125			   rt->rt_rmx.rmx_mtu > mtu) {
1126			icmp6stat.icp6s_pmtuchg++;
1127			rt->rt_rmx.rmx_mtu = mtu;
1128		}
1129	}
1130	if (rt)
1131		rtfree(rt);
1132}
1133
1134/*
1135 * Process a Node Information Query packet, based on
1136 * draft-ietf-ipngwg-icmp-name-lookups-07.
1137 *
1138 * Spec incompatibilities:
1139 * - IPv6 Subject address handling
1140 * - IPv4 Subject address handling support missing
1141 * - Proxy reply (answer even if it's not for me)
1142 * - joins NI group address at in6_ifattach() time only, does not cope
1143 *   with hostname changes by sethostname(3)
1144 */
1145#define hostnamelen	strlen(hostname)
1146static struct mbuf *
1147ni6_input(m, off)
1148	struct mbuf *m;
1149	int off;
1150{
1151	struct icmp6_nodeinfo *ni6, *nni6;
1152	struct mbuf *n = NULL;
1153	u_int16_t qtype;
1154	int subjlen;
1155	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1156	struct ni_reply_fqdn *fqdn;
1157	int addrs;		/* for NI_QTYPE_NODEADDR */
1158	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1159	struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1160	struct sockaddr_in6 sin6_d; /* XXX: we should retrieve this from m_aux */
1161	struct ip6_hdr *ip6;
1162	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1163	char *subj = NULL;
1164	struct in6_ifaddr *ia6 = NULL;
1165
1166	ip6 = mtod(m, struct ip6_hdr *);
1167#ifndef PULLDOWN_TEST
1168	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1169#else
1170	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1171	if (ni6 == NULL) {
1172		/* m is already reclaimed */
1173		return (NULL);
1174	}
1175#endif
1176
1177	/*
1178	 * Validate IPv6 destination address.
1179	 *
1180	 * The Responder must discard the Query without further processing
1181	 * unless it is one of the Responder's unicast or anycast addresses, or
1182	 * a link-local scope multicast address which the Responder has joined.
1183	 * [icmp-name-lookups-07, Section 4.]
1184	 */
1185	bzero(&sin6, sizeof(sin6));
1186	sin6.sin6_family = AF_INET6;
1187	sin6.sin6_len = sizeof(struct sockaddr_in6);
1188	bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1189	/* XXX scopeid */
1190	if ((ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)&sin6)) != NULL) {
1191		/* unicast/anycast, fine */
1192		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1193		    (icmp6_nodeinfo & 4) == 0) {
1194			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1195				"a temporary address in %s:%d",
1196			       __FILE__, __LINE__));
1197			goto bad;
1198		}
1199	} else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1200		; /* link-local multicast, fine */
1201	else
1202		goto bad;
1203
1204	/* validate query Subject field. */
1205	qtype = ntohs(ni6->ni_qtype);
1206	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1207	switch (qtype) {
1208	case NI_QTYPE_NOOP:
1209	case NI_QTYPE_SUPTYPES:
1210		/* 07 draft */
1211		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1212			break;
1213		/* FALLTHROUGH */
1214	case NI_QTYPE_FQDN:
1215	case NI_QTYPE_NODEADDR:
1216		switch (ni6->ni_code) {
1217		case ICMP6_NI_SUBJ_IPV6:
1218#if ICMP6_NI_SUBJ_IPV6 != 0
1219		case 0:
1220#endif
1221			/*
1222			 * backward compatibility - try to accept 03 draft
1223			 * format, where no Subject is present.
1224			 */
1225			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1226			    subjlen == 0) {
1227				oldfqdn++;
1228				break;
1229			}
1230#if ICMP6_NI_SUBJ_IPV6 != 0
1231			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1232				goto bad;
1233#endif
1234
1235			if (subjlen != sizeof(sin6.sin6_addr))
1236				goto bad;
1237
1238			/*
1239			 * Validate Subject address.
1240			 *
1241			 * Not sure what exactly "address belongs to the node"
1242			 * means in the spec, is it just unicast, or what?
1243			 *
1244			 * At this moment we consider Subject address as
1245			 * "belong to the node" if the Subject address equals
1246			 * to the IPv6 destination address; validation for
1247			 * IPv6 destination address should have done enough
1248			 * check for us.
1249			 *
1250			 * We do not do proxy at this moment.
1251			 */
1252			/* m_pulldown instead of copy? */
1253			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1254			    subjlen, (caddr_t)&sin6.sin6_addr);
1255			if (in6_addr2zoneid(m->m_pkthdr.rcvif,
1256			    &sin6.sin6_addr, &sin6.sin6_scope_id)) {
1257				goto bad;
1258			}
1259			in6_embedscope(&sin6.sin6_addr, &sin6, NULL, NULL);
1260			bzero(&sin6_d, sizeof(sin6_d));
1261			sin6_d.sin6_family = AF_INET6; /* not used, actually */
1262			sin6_d.sin6_len = sizeof(sin6_d); /* ditto */
1263			sin6_d.sin6_addr = ip6->ip6_dst;
1264			if (in6_addr2zoneid(m->m_pkthdr.rcvif,
1265			    &ip6->ip6_dst, &sin6_d.sin6_scope_id)) {
1266				goto bad;
1267			}
1268			in6_embedscope(&sin6_d.sin6_addr, &sin6_d, NULL, NULL);
1269			subj = (char *)&sin6;
1270			if (SA6_ARE_ADDR_EQUAL(&sin6, &sin6_d))
1271				break;
1272
1273			/*
1274			 * XXX if we are to allow other cases, we should really
1275			 * be careful about scope here.
1276			 * basically, we should disallow queries toward IPv6
1277			 * destination X with subject Y,
1278			 * if scope(X) > scope(Y).
1279			 * if we allow scope(X) > scope(Y), it will result in
1280			 * information leakage across scope boundary.
1281			 */
1282			goto bad;
1283
1284		case ICMP6_NI_SUBJ_FQDN:
1285			/*
1286			 * Validate Subject name with gethostname(3).
1287			 *
1288			 * The behavior may need some debate, since:
1289			 * - we are not sure if the node has FQDN as
1290			 *   hostname (returned by gethostname(3)).
1291			 * - the code does wildcard match for truncated names.
1292			 *   however, we are not sure if we want to perform
1293			 *   wildcard match, if gethostname(3) side has
1294			 *   truncated hostname.
1295			 */
1296			n = ni6_nametodns(hostname, hostnamelen, 0);
1297			if (!n || n->m_next || n->m_len == 0)
1298				goto bad;
1299			IP6_EXTHDR_GET(subj, char *, m,
1300			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1301			if (subj == NULL)
1302				goto bad;
1303			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1304			    n->m_len)) {
1305				goto bad;
1306			}
1307			m_freem(n);
1308			n = NULL;
1309			break;
1310
1311		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1312		default:
1313			goto bad;
1314		}
1315		break;
1316	}
1317
1318	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1319	switch (qtype) {
1320	case NI_QTYPE_FQDN:
1321		if ((icmp6_nodeinfo & 1) == 0)
1322			goto bad;
1323		break;
1324	case NI_QTYPE_NODEADDR:
1325		if ((icmp6_nodeinfo & 2) == 0)
1326			goto bad;
1327		break;
1328	}
1329
1330	/* guess reply length */
1331	switch (qtype) {
1332	case NI_QTYPE_NOOP:
1333		break;		/* no reply data */
1334	case NI_QTYPE_SUPTYPES:
1335		replylen += sizeof(u_int32_t);
1336		break;
1337	case NI_QTYPE_FQDN:
1338		/* XXX will append an mbuf */
1339		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1340		break;
1341	case NI_QTYPE_NODEADDR:
1342		addrs = ni6_addrs(ni6, m, &ifp, subj);
1343		if ((replylen += addrs * (sizeof(struct in6_addr) +
1344		    sizeof(u_int32_t))) > MCLBYTES)
1345			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1346		break;
1347	default:
1348		/*
1349		 * XXX: We must return a reply with the ICMP6 code
1350		 * `unknown Qtype' in this case.  However we regard the case
1351		 * as an FQDN query for backward compatibility.
1352		 * Older versions set a random value to this field,
1353		 * so it rarely varies in the defined qtypes.
1354		 * But the mechanism is not reliable...
1355		 * maybe we should obsolete older versions.
1356		 */
1357		qtype = NI_QTYPE_FQDN;
1358		/* XXX will append an mbuf */
1359		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1360		oldfqdn++;
1361		break;
1362	}
1363
1364	/* allocate an mbuf to reply. */
1365	MGETHDR(n, M_DONTWAIT, m->m_type);
1366	if (n == NULL) {
1367		m_freem(m);
1368		return (NULL);
1369	}
1370	M_MOVE_PKTHDR(n, m); /* just for recvif */
1371	if (replylen > MHLEN) {
1372		if (replylen > MCLBYTES) {
1373			/*
1374			 * XXX: should we try to allocate more? But MCLBYTES
1375			 * is probably much larger than IPV6_MMTU...
1376			 */
1377			goto bad;
1378		}
1379		MCLGET(n, M_DONTWAIT);
1380		if ((n->m_flags & M_EXT) == 0) {
1381			goto bad;
1382		}
1383	}
1384	n->m_pkthdr.len = n->m_len = replylen;
1385
1386	/* copy mbuf header and IPv6 + Node Information base headers */
1387	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1388	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1389	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1390
1391	/* qtype dependent procedure */
1392	switch (qtype) {
1393	case NI_QTYPE_NOOP:
1394		nni6->ni_code = ICMP6_NI_SUCCESS;
1395		nni6->ni_flags = 0;
1396		break;
1397	case NI_QTYPE_SUPTYPES:
1398	{
1399		u_int32_t v;
1400		nni6->ni_code = ICMP6_NI_SUCCESS;
1401		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1402		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1403		v = (u_int32_t)htonl(0x0000000f);
1404		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1405		break;
1406	}
1407	case NI_QTYPE_FQDN:
1408		nni6->ni_code = ICMP6_NI_SUCCESS;
1409		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1410		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1411		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1412		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1413		/*
1414		 * XXX do we really have FQDN in variable "hostname"?
1415		 */
1416		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1417		if (n->m_next == NULL)
1418			goto bad;
1419		/* XXX we assume that n->m_next is not a chain */
1420		if (n->m_next->m_next != NULL)
1421			goto bad;
1422		n->m_pkthdr.len += n->m_next->m_len;
1423		break;
1424	case NI_QTYPE_NODEADDR:
1425	{
1426		int lenlim, copied;
1427
1428		nni6->ni_code = ICMP6_NI_SUCCESS;
1429		n->m_pkthdr.len = n->m_len =
1430		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1431		lenlim = M_TRAILINGSPACE(n);
1432		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1433		/* XXX: reset mbuf length */
1434		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1435		    sizeof(struct icmp6_nodeinfo) + copied;
1436		break;
1437	}
1438	default:
1439		break;		/* XXX impossible! */
1440	}
1441
1442	nni6->ni_type = ICMP6_NI_REPLY;
1443	m_freem(m);
1444	return (n);
1445
1446  bad:
1447	m_freem(m);
1448	if (n)
1449		m_freem(n);
1450	return (NULL);
1451}
1452#undef hostnamelen
1453
1454/*
1455 * make a mbuf with DNS-encoded string.  no compression support.
1456 *
1457 * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1458 * treated as truncated name (two \0 at the end).  this is a wild guess.
1459 */
1460static struct mbuf *
1461ni6_nametodns(name, namelen, old)
1462	const char *name;
1463	int namelen;
1464	int old;	/* return pascal string if non-zero */
1465{
1466	struct mbuf *m;
1467	char *cp, *ep;
1468	const char *p, *q;
1469	int i, len, nterm;
1470
1471	if (old)
1472		len = namelen + 1;
1473	else
1474		len = MCLBYTES;
1475
1476	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1477	MGET(m, M_DONTWAIT, MT_DATA);
1478	if (m && len > MLEN) {
1479		MCLGET(m, M_DONTWAIT);
1480		if ((m->m_flags & M_EXT) == 0)
1481			goto fail;
1482	}
1483	if (!m)
1484		goto fail;
1485	m->m_next = NULL;
1486
1487	if (old) {
1488		m->m_len = len;
1489		*mtod(m, char *) = namelen;
1490		bcopy(name, mtod(m, char *) + 1, namelen);
1491		return m;
1492	} else {
1493		m->m_len = 0;
1494		cp = mtod(m, char *);
1495		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1496
1497		/* if not certain about my name, return empty buffer */
1498		if (namelen == 0)
1499			return m;
1500
1501		/*
1502		 * guess if it looks like shortened hostname, or FQDN.
1503		 * shortened hostname needs two trailing "\0".
1504		 */
1505		i = 0;
1506		for (p = name; p < name + namelen; p++) {
1507			if (*p && *p == '.')
1508				i++;
1509		}
1510		if (i < 2)
1511			nterm = 2;
1512		else
1513			nterm = 1;
1514
1515		p = name;
1516		while (cp < ep && p < name + namelen) {
1517			i = 0;
1518			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1519				i++;
1520			/* result does not fit into mbuf */
1521			if (cp + i + 1 >= ep)
1522				goto fail;
1523			/*
1524			 * DNS label length restriction, RFC1035 page 8.
1525			 * "i == 0" case is included here to avoid returning
1526			 * 0-length label on "foo..bar".
1527			 */
1528			if (i <= 0 || i >= 64)
1529				goto fail;
1530			*cp++ = i;
1531			bcopy(p, cp, i);
1532			cp += i;
1533			p = q;
1534			if (p < name + namelen && *p == '.')
1535				p++;
1536		}
1537		/* termination */
1538		if (cp + nterm >= ep)
1539			goto fail;
1540		while (nterm-- > 0)
1541			*cp++ = '\0';
1542		m->m_len = cp - mtod(m, char *);
1543		return m;
1544	}
1545
1546	panic("should not reach here");
1547	/* NOTREACHED */
1548
1549 fail:
1550	if (m)
1551		m_freem(m);
1552	return NULL;
1553}
1554
1555/*
1556 * check if two DNS-encoded string matches.  takes care of truncated
1557 * form (with \0\0 at the end).  no compression support.
1558 * XXX upper/lowercase match (see RFC2065)
1559 */
1560static int
1561ni6_dnsmatch(a, alen, b, blen)
1562	const char *a;
1563	int alen;
1564	const char *b;
1565	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(ni6, m, ifpp, subj)
1626	struct icmp6_nodeinfo *ni6;
1627	struct mbuf *m;
1628	struct ifnet **ifpp;
1629	char *subj;
1630{
1631	struct ifnet *ifp;
1632	struct in6_ifaddr *ifa6;
1633	struct ifaddr *ifa;
1634	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1635	int addrs = 0, addrsofif, iffound = 0;
1636	int niflags = ni6->ni_flags;
1637
1638	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1639		switch (ni6->ni_code) {
1640		case ICMP6_NI_SUBJ_IPV6:
1641			if (subj == NULL) /* must be impossible... */
1642				return (0);
1643			subj_ip6 = (struct sockaddr_in6 *)subj;
1644			break;
1645		default:
1646			/*
1647			 * XXX: we only support IPv6 subject address for
1648			 * this Qtype.
1649			 */
1650			return (0);
1651		}
1652	}
1653
1654	IFNET_RLOCK();
1655	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1656		addrsofif = 0;
1657		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1658			if (ifa->ifa_addr->sa_family != AF_INET6)
1659				continue;
1660			ifa6 = (struct in6_ifaddr *)ifa;
1661
1662			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1663			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1664			    &ifa6->ia_addr.sin6_addr))
1665				iffound = 1;
1666
1667			/*
1668			 * IPv4-mapped addresses can only be returned by a
1669			 * Node Information proxy, since they represent
1670			 * addresses of IPv4-only nodes, which perforce do
1671			 * not implement this protocol.
1672			 * [icmp-name-lookups-07, Section 5.4]
1673			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1674			 * this function at this moment.
1675			 */
1676
1677			/* What do we have to do about ::1? */
1678			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1679			case IPV6_ADDR_SCOPE_LINKLOCAL:
1680				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1681					continue;
1682				break;
1683			case IPV6_ADDR_SCOPE_SITELOCAL:
1684				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1685					continue;
1686				break;
1687			case IPV6_ADDR_SCOPE_GLOBAL:
1688				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1689					continue;
1690				break;
1691			default:
1692				continue;
1693			}
1694
1695			/*
1696			 * check if anycast is okay.
1697			 * XXX: just experimental.  not in the spec.
1698			 */
1699			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1700			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1701				continue; /* we need only unicast addresses */
1702			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1703			    (icmp6_nodeinfo & 4) == 0) {
1704				continue;
1705			}
1706			addrsofif++; /* count the address */
1707		}
1708		if (iffound) {
1709			*ifpp = ifp;
1710			IFNET_RUNLOCK();
1711			return (addrsofif);
1712		}
1713
1714		addrs += addrsofif;
1715	}
1716	IFNET_RUNLOCK();
1717
1718	return (addrs);
1719}
1720
1721static int
1722ni6_store_addrs(ni6, nni6, ifp0, resid)
1723	struct icmp6_nodeinfo *ni6, *nni6;
1724	struct ifnet *ifp0;
1725	int resid;
1726{
1727	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1728	struct in6_ifaddr *ifa6;
1729	struct ifaddr *ifa;
1730	struct ifnet *ifp_dep = NULL;
1731	int copied = 0, allow_deprecated = 0;
1732	u_char *cp = (u_char *)(nni6 + 1);
1733	int niflags = ni6->ni_flags;
1734	u_int32_t ltime;
1735
1736	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1737		return (0);	/* needless to copy */
1738
1739	IFNET_RLOCK();
1740  again:
1741
1742	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1743		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1744		     ifa = ifa->ifa_list.tqe_next) {
1745			if (ifa->ifa_addr->sa_family != AF_INET6)
1746				continue;
1747			ifa6 = (struct in6_ifaddr *)ifa;
1748
1749			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1750			    allow_deprecated == 0) {
1751				/*
1752				 * prefererred address should be put before
1753				 * deprecated addresses.
1754				 */
1755
1756				/* record the interface for later search */
1757				if (ifp_dep == NULL)
1758					ifp_dep = ifp;
1759
1760				continue;
1761			} else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1762			    allow_deprecated != 0)
1763				continue; /* we now collect deprecated addrs */
1764
1765			/* What do we have to do about ::1? */
1766			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1767			case IPV6_ADDR_SCOPE_LINKLOCAL:
1768				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1769					continue;
1770				break;
1771			case IPV6_ADDR_SCOPE_SITELOCAL:
1772				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1773					continue;
1774				break;
1775			case IPV6_ADDR_SCOPE_GLOBAL:
1776				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1777					continue;
1778				break;
1779			default:
1780				continue;
1781			}
1782
1783			/*
1784			 * check if anycast is okay.
1785			 * XXX: just experimental.  not in the spec.
1786			 */
1787			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1788			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1789				continue;
1790			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1791			    (icmp6_nodeinfo & 4) == 0) {
1792				continue;
1793			}
1794
1795			/* now we can copy the address */
1796			if (resid < sizeof(struct in6_addr) +
1797			    sizeof(u_int32_t)) {
1798				/*
1799				 * We give up much more copy.
1800				 * Set the truncate flag and return.
1801				 */
1802				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1803				IFNET_RUNLOCK();
1804				return (copied);
1805			}
1806
1807			/*
1808			 * Set the TTL of the address.
1809			 * The TTL value should be one of the following
1810			 * according to the specification:
1811			 *
1812			 * 1. The remaining lifetime of a DHCP lease on the
1813			 *    address, or
1814			 * 2. The remaining Valid Lifetime of a prefix from
1815			 *    which the address was derived through Stateless
1816			 *    Autoconfiguration.
1817			 *
1818			 * Note that we currently do not support stateful
1819			 * address configuration by DHCPv6, so the former
1820			 * case can't happen.
1821			 */
1822			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1823				ltime = ND6_INFINITE_LIFETIME;
1824			else {
1825				if (ifa6->ia6_lifetime.ia6t_expire >
1826				    time_second)
1827					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1828				else
1829					ltime = 0;
1830			}
1831
1832			bcopy(&ltime, cp, sizeof(u_int32_t));
1833			cp += sizeof(u_int32_t);
1834
1835			/* copy the address itself */
1836			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1837			    sizeof(struct in6_addr));
1838			in6_clearscope((struct in6_addr *)cp); /* XXX */
1839			cp += sizeof(struct in6_addr);
1840
1841			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1842			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1843		}
1844		if (ifp0)	/* we need search only on the specified IF */
1845			break;
1846	}
1847
1848	if (allow_deprecated == 0 && ifp_dep != NULL) {
1849		ifp = ifp_dep;
1850		allow_deprecated = 1;
1851
1852		goto again;
1853	}
1854
1855	IFNET_RUNLOCK();
1856
1857	return (copied);
1858}
1859
1860/*
1861 * XXX almost dup'ed code with rip6_input.
1862 */
1863static int
1864icmp6_rip6_input(mp, off)
1865	struct	mbuf **mp;
1866	int	off;
1867{
1868	struct mbuf *m = *mp;
1869	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1870	struct in6pcb *in6p;
1871	struct in6pcb *last = NULL;
1872	struct sockaddr_in6 fromsa;
1873	struct icmp6_hdr *icmp6;
1874	struct mbuf *opts = NULL;
1875
1876#ifndef PULLDOWN_TEST
1877	/* this is assumed to be safe. */
1878	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1879#else
1880	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1881	if (icmp6 == NULL) {
1882		/* m is already reclaimed */
1883		return (IPPROTO_DONE);
1884	}
1885#endif
1886
1887	bzero(&fromsa, sizeof(fromsa));
1888	fromsa.sin6_len = sizeof(struct sockaddr_in6);
1889	fromsa.sin6_family = AF_INET6;
1890	/* KAME hack: recover scopeid */
1891	(void)in6_recoverscope(&fromsa, &ip6->ip6_src, m->m_pkthdr.rcvif);
1892
1893	LIST_FOREACH(in6p, &ripcb, inp_list) {
1894		if ((in6p->inp_vflag & INP_IPV6) == 0)
1895			continue;
1896		if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1897			continue;
1898		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1899		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1900			continue;
1901		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1902		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1903			continue;
1904		if (in6p->in6p_icmp6filt
1905		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1906				 in6p->in6p_icmp6filt))
1907			continue;
1908		if (last) {
1909			struct	mbuf *n;
1910			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1911				if (last->in6p_flags & IN6P_CONTROLOPTS)
1912					ip6_savecontrol(last, &opts, ip6, n);
1913				/* strip intermediate headers */
1914				m_adj(n, off);
1915				if (sbappendaddr(&last->in6p_socket->so_rcv,
1916				    (struct sockaddr *)&fromsa, n, opts)
1917				    == 0) {
1918					/* should notify about lost packet */
1919					m_freem(n);
1920					if (opts) {
1921						m_freem(opts);
1922					}
1923				} else
1924					sorwakeup(last->in6p_socket);
1925				opts = NULL;
1926			}
1927		}
1928		last = in6p;
1929	}
1930	if (last) {
1931		if (last->in6p_flags & IN6P_CONTROLOPTS)
1932			ip6_savecontrol(last, &opts, ip6, m);
1933		/* strip intermediate headers */
1934		m_adj(m, off);
1935		if (sbappendaddr(&last->in6p_socket->so_rcv,
1936		    (struct sockaddr *)&fromsa, m, opts) == 0) {
1937			m_freem(m);
1938			if (opts)
1939				m_freem(opts);
1940		} else
1941			sorwakeup(last->in6p_socket);
1942	} else {
1943		m_freem(m);
1944		ip6stat.ip6s_delivered--;
1945	}
1946	return IPPROTO_DONE;
1947}
1948
1949/*
1950 * Reflect the ip6 packet back to the source.
1951 * OFF points to the icmp6 header, counted from the top of the mbuf.
1952 */
1953void
1954icmp6_reflect(m, off)
1955	struct	mbuf *m;
1956	size_t off;
1957{
1958	struct ip6_hdr *ip6;
1959	struct icmp6_hdr *icmp6;
1960	struct in6_ifaddr *ia;
1961	struct in6_addr t, *src = 0;
1962	int plen;
1963	int type, code;
1964	struct ifnet *outif = NULL;
1965	struct sockaddr_in6 sa6_src, sa6_dst;
1966#ifdef COMPAT_RFC1885
1967	int mtu = IPV6_MMTU;
1968	struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
1969#endif
1970
1971	/* too short to reflect */
1972	if (off < sizeof(struct ip6_hdr)) {
1973		nd6log((LOG_DEBUG,
1974		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
1975		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
1976		    __FILE__, __LINE__));
1977		goto bad;
1978	}
1979
1980	/*
1981	 * If there are extra headers between IPv6 and ICMPv6, strip
1982	 * off that header first.
1983	 */
1984#ifdef DIAGNOSTIC
1985	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
1986		panic("assumption failed in icmp6_reflect");
1987#endif
1988	if (off > sizeof(struct ip6_hdr)) {
1989		size_t l;
1990		struct ip6_hdr nip6;
1991
1992		l = off - sizeof(struct ip6_hdr);
1993		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
1994		m_adj(m, l);
1995		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1996		if (m->m_len < l) {
1997			if ((m = m_pullup(m, l)) == NULL)
1998				return;
1999		}
2000		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2001	} else /* off == sizeof(struct ip6_hdr) */ {
2002		size_t l;
2003		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2004		if (m->m_len < l) {
2005			if ((m = m_pullup(m, l)) == NULL)
2006				return;
2007		}
2008	}
2009	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2010	ip6 = mtod(m, struct ip6_hdr *);
2011	ip6->ip6_nxt = IPPROTO_ICMPV6;
2012	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2013	type = icmp6->icmp6_type; /* keep type for statistics */
2014	code = icmp6->icmp6_code; /* ditto. */
2015
2016	t = ip6->ip6_dst;
2017	/*
2018	 * ip6_input() drops a packet if its src is multicast.
2019	 * So, the src is never multicast.
2020	 */
2021	ip6->ip6_dst = ip6->ip6_src;
2022
2023	/*
2024	 * XXX: make sure to embed scope zone information, using
2025	 * already embedded IDs or the received interface (if any).
2026	 * Note that rcvif may be NULL.
2027	 * TODO: scoped routing case (XXX).
2028	 */
2029	bzero(&sa6_src, sizeof(sa6_src));
2030	sa6_src.sin6_family = AF_INET6;
2031	sa6_src.sin6_len = sizeof(sa6_src);
2032	sa6_src.sin6_addr = ip6->ip6_dst;
2033	in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
2034	in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL);
2035	bzero(&sa6_dst, sizeof(sa6_dst));
2036	sa6_dst.sin6_family = AF_INET6;
2037	sa6_dst.sin6_len = sizeof(sa6_dst);
2038	sa6_dst.sin6_addr = t;
2039	in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
2040	in6_embedscope(&t, &sa6_dst, NULL, NULL);
2041
2042#ifdef COMPAT_RFC1885
2043	/*
2044	 * xxx guess MTU
2045	 * RFC 1885 requires that echo reply should be truncated if it
2046	 * does not fit in with (return) path MTU, but the description was
2047	 * removed in the new spec.
2048	 */
2049	if (icmp6_reflect_rt.ro_rt == 0 ||
2050	    ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
2051		if (icmp6_reflect_rt.ro_rt) {
2052			RTFREE(icmp6_reflect_rt.ro_rt);
2053			icmp6_reflect_rt.ro_rt = 0;
2054		}
2055		bzero(sin6, sizeof(*sin6));
2056		sin6->sin6_family = PF_INET6;
2057		sin6->sin6_len = sizeof(struct sockaddr_in6);
2058		sin6->sin6_addr = ip6->ip6_dst;
2059
2060		rtalloc_ign((struct route *)&icmp6_reflect_rt.ro_rt,
2061			    RTF_PRCLONING);
2062	}
2063
2064	if (icmp6_reflect_rt.ro_rt == 0)
2065		goto bad;
2066
2067	if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
2068	    && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
2069		mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
2070
2071	if (mtu < m->m_pkthdr.len) {
2072		plen -= (m->m_pkthdr.len - mtu);
2073		m_adj(m, mtu - m->m_pkthdr.len);
2074	}
2075#endif
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	 */
2083	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2084		if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2085		    (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2086			src = &t;
2087			break;
2088		}
2089	if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2090		/*
2091		 * This is the case if the dst is our link-local address
2092		 * and the sender is also ourselves.
2093		 */
2094		src = &t;
2095	}
2096
2097	if (src == 0) {
2098		int e;
2099		struct route_in6 ro;
2100
2101		/*
2102		 * This case matches to multicasts, our anycast, or unicasts
2103		 * that we do not own.  Select a source address based on the
2104		 * source address of the erroneous packet.
2105		 */
2106		bzero(&ro, sizeof(ro));
2107		src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e);
2108		if (ro.ro_rt)
2109			RTFREE(ro.ro_rt); /* XXX: we could use this */
2110		if (src == NULL) {
2111			nd6log((LOG_DEBUG,
2112			    "icmp6_reflect: source can't be determined: "
2113			    "dst=%s, error=%d\n",
2114			    ip6_sprintf(&sa6_src.sin6_addr), e));
2115			goto bad;
2116		}
2117	}
2118
2119	ip6->ip6_src = *src;
2120
2121	ip6->ip6_flow = 0;
2122	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2123	ip6->ip6_vfc |= IPV6_VERSION;
2124	ip6->ip6_nxt = IPPROTO_ICMPV6;
2125	if (m->m_pkthdr.rcvif) {
2126		/* XXX: This may not be the outgoing interface */
2127		ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2128	} else
2129		ip6->ip6_hlim = ip6_defhlim;
2130
2131	icmp6->icmp6_cksum = 0;
2132	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2133	    sizeof(struct ip6_hdr), plen);
2134
2135	/*
2136	 * XXX option handling
2137	 */
2138
2139	m->m_flags &= ~(M_BCAST|M_MCAST);
2140
2141#ifdef COMPAT_RFC1885
2142	ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif, NULL);
2143#else
2144	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2145#endif
2146	if (outif)
2147		icmp6_ifoutstat_inc(outif, type, code);
2148
2149	return;
2150
2151 bad:
2152	m_freem(m);
2153	return;
2154}
2155
2156void
2157icmp6_fasttimo()
2158{
2159
2160	mld6_fasttimeo();
2161}
2162
2163static const char *
2164icmp6_redirect_diag(src6, dst6, tgt6)
2165	struct in6_addr *src6;
2166	struct in6_addr *dst6;
2167	struct in6_addr *tgt6;
2168{
2169	static char buf[1024];
2170	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2171	    ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2172	return buf;
2173}
2174
2175void
2176icmp6_redirect_input(m, off)
2177	struct mbuf *m;
2178	int off;
2179{
2180	struct ifnet *ifp = m->m_pkthdr.rcvif;
2181	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2182	struct nd_redirect *nd_rd;
2183	int icmp6len = ntohs(ip6->ip6_plen);
2184	char *lladdr = NULL;
2185	int lladdrlen = 0;
2186	u_char *redirhdr = NULL;
2187	int redirhdrlen = 0;
2188	struct rtentry *rt = NULL;
2189	int is_router;
2190	int is_onlink;
2191	struct in6_addr src6 = ip6->ip6_src;
2192	struct in6_addr redtgt6;
2193	struct in6_addr reddst6;
2194	union nd_opts ndopts;
2195
2196	if (!m || !ifp)
2197		return;
2198
2199	/* XXX if we are router, we don't update route by icmp6 redirect */
2200	if (ip6_forwarding)
2201		goto freeit;
2202	if (!icmp6_rediraccept)
2203		goto freeit;
2204
2205#ifndef PULLDOWN_TEST
2206	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2207	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2208#else
2209	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2210	if (nd_rd == NULL) {
2211		icmp6stat.icp6s_tooshort++;
2212		return;
2213	}
2214#endif
2215	redtgt6 = nd_rd->nd_rd_target;
2216	reddst6 = nd_rd->nd_rd_dst;
2217
2218	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2219		redtgt6.s6_addr16[1] = htons(ifp->if_index);
2220	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2221		reddst6.s6_addr16[1] = htons(ifp->if_index);
2222
2223	/* validation */
2224	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2225		nd6log((LOG_ERR,
2226		    "ICMP6 redirect sent from %s rejected; "
2227		    "must be from linklocal\n",
2228		    ip6_sprintf(&src6)));
2229		goto bad;
2230	}
2231	if (ip6->ip6_hlim != 255) {
2232		nd6log((LOG_ERR,
2233		    "ICMP6 redirect sent from %s rejected; "
2234		    "hlim=%d (must be 255)\n",
2235		    ip6_sprintf(&src6), ip6->ip6_hlim));
2236		goto bad;
2237	}
2238    {
2239	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2240	struct sockaddr_in6 sin6;
2241	struct in6_addr *gw6;
2242
2243	bzero(&sin6, sizeof(sin6));
2244	sin6.sin6_family = AF_INET6;
2245	sin6.sin6_len = sizeof(struct sockaddr_in6);
2246	bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2247	rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL);
2248	if (rt) {
2249		if (rt->rt_gateway == NULL ||
2250		    rt->rt_gateway->sa_family != AF_INET6) {
2251			nd6log((LOG_ERR,
2252			    "ICMP6 redirect rejected; no route "
2253			    "with inet6 gateway found for redirect dst: %s\n",
2254			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2255			RTFREE_LOCKED(rt);
2256			goto bad;
2257		}
2258
2259		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2260		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2261			nd6log((LOG_ERR,
2262			    "ICMP6 redirect rejected; "
2263			    "not equal to gw-for-src=%s (must be same): "
2264			    "%s\n",
2265			    ip6_sprintf(gw6),
2266			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2267			RTFREE_LOCKED(rt);
2268			goto bad;
2269		}
2270	} else {
2271		nd6log((LOG_ERR,
2272		    "ICMP6 redirect rejected; "
2273		    "no route found for redirect dst: %s\n",
2274		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2275		goto bad;
2276	}
2277	RTFREE_LOCKED(rt);
2278	rt = NULL;
2279    }
2280	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2281		nd6log((LOG_ERR,
2282		    "ICMP6 redirect rejected; "
2283		    "redirect dst must be unicast: %s\n",
2284		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2285		goto bad;
2286	}
2287
2288	is_router = is_onlink = 0;
2289	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2290		is_router = 1;	/* router case */
2291	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2292		is_onlink = 1;	/* on-link destination case */
2293	if (!is_router && !is_onlink) {
2294		nd6log((LOG_ERR,
2295		    "ICMP6 redirect rejected; "
2296		    "neither router case nor onlink case: %s\n",
2297		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2298		goto bad;
2299	}
2300	/* validation passed */
2301
2302	icmp6len -= sizeof(*nd_rd);
2303	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2304	if (nd6_options(&ndopts) < 0) {
2305		nd6log((LOG_INFO, "icmp6_redirect_input: "
2306		    "invalid ND option, rejected: %s\n",
2307		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2308		/* nd6_options have incremented stats */
2309		goto freeit;
2310	}
2311
2312	if (ndopts.nd_opts_tgt_lladdr) {
2313		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2314		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2315	}
2316
2317	if (ndopts.nd_opts_rh) {
2318		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2319		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2320	}
2321
2322	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2323		nd6log((LOG_INFO,
2324		    "icmp6_redirect_input: lladdrlen mismatch for %s "
2325		    "(if %d, icmp6 packet %d): %s\n",
2326		    ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2327		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2328		goto bad;
2329	}
2330
2331	/* RFC 2461 8.3 */
2332	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2333	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2334
2335	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2336		/* perform rtredirect */
2337		struct sockaddr_in6 sdst;
2338		struct sockaddr_in6 sgw;
2339		struct sockaddr_in6 ssrc;
2340
2341		bzero(&sdst, sizeof(sdst));
2342		bzero(&sgw, sizeof(sgw));
2343		bzero(&ssrc, sizeof(ssrc));
2344		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2345		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2346			sizeof(struct sockaddr_in6);
2347		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2348		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2349		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2350		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2351		    (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2352		    (struct sockaddr *)&ssrc);
2353	}
2354	/* finally update cached route in each socket via pfctlinput */
2355    {
2356	struct sockaddr_in6 sdst;
2357
2358	bzero(&sdst, sizeof(sdst));
2359	sdst.sin6_family = AF_INET6;
2360	sdst.sin6_len = sizeof(struct sockaddr_in6);
2361	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2362	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2363#ifdef IPSEC
2364	key_sa_routechange((struct sockaddr *)&sdst);
2365#endif
2366    }
2367
2368 freeit:
2369	m_freem(m);
2370	return;
2371
2372 bad:
2373	icmp6stat.icp6s_badredirect++;
2374	m_freem(m);
2375}
2376
2377void
2378icmp6_redirect_output(m0, rt)
2379	struct mbuf *m0;
2380	struct rtentry *rt;
2381{
2382	struct ifnet *ifp;	/* my outgoing interface */
2383	struct in6_addr *ifp_ll6;
2384	struct in6_addr *router_ll6;
2385	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2386	struct mbuf *m = NULL;	/* newly allocated one */
2387	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2388	struct nd_redirect *nd_rd;
2389	size_t maxlen;
2390	u_char *p;
2391	struct ifnet *outif = NULL;
2392	struct sockaddr_in6 src_sa;
2393
2394	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2395
2396	/* if we are not router, we don't send icmp6 redirect */
2397	if (!ip6_forwarding || ip6_accept_rtadv)
2398		goto fail;
2399
2400	/* sanity check */
2401	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2402		goto fail;
2403
2404	/*
2405	 * Address check:
2406	 *  the source address must identify a neighbor, and
2407	 *  the destination address must not be a multicast address
2408	 *  [RFC 2461, sec 8.2]
2409	 */
2410	sip6 = mtod(m0, struct ip6_hdr *);
2411	bzero(&src_sa, sizeof(src_sa));
2412	src_sa.sin6_family = AF_INET6;
2413	src_sa.sin6_len = sizeof(src_sa);
2414	src_sa.sin6_addr = sip6->ip6_src;
2415	/* we don't currently use sin6_scope_id, but eventually use it */
2416	if (in6_addr2zoneid(ifp, &sip6->ip6_src, &src_sa.sin6_scope_id))
2417		goto fail;
2418	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2419		goto fail;
2420	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2421		goto fail;	/* what should we do here? */
2422
2423	/* rate limit */
2424	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2425		goto fail;
2426
2427	/*
2428	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2429	 * we almost always ask for an mbuf cluster for simplicity.
2430	 * (MHLEN < IPV6_MMTU is almost always true)
2431	 */
2432#if IPV6_MMTU >= MCLBYTES
2433# error assumption failed about IPV6_MMTU and MCLBYTES
2434#endif
2435	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2436	if (m && IPV6_MMTU >= MHLEN)
2437		MCLGET(m, M_DONTWAIT);
2438	if (!m)
2439		goto fail;
2440	m->m_pkthdr.rcvif = NULL;
2441	m->m_len = 0;
2442	maxlen = M_TRAILINGSPACE(m);
2443	maxlen = min(IPV6_MMTU, maxlen);
2444	/* just for safety */
2445	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2446	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2447		goto fail;
2448	}
2449
2450	{
2451		/* get ip6 linklocal address for ifp(my outgoing interface). */
2452		struct in6_ifaddr *ia;
2453		if ((ia = in6ifa_ifpforlinklocal(ifp,
2454						 IN6_IFF_NOTREADY|
2455						 IN6_IFF_ANYCAST)) == NULL)
2456			goto fail;
2457		ifp_ll6 = &ia->ia_addr.sin6_addr;
2458	}
2459
2460	/* get ip6 linklocal address for the router. */
2461	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2462		struct sockaddr_in6 *sin6;
2463		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2464		router_ll6 = &sin6->sin6_addr;
2465		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2466			router_ll6 = (struct in6_addr *)NULL;
2467	} else
2468		router_ll6 = (struct in6_addr *)NULL;
2469
2470	/* ip6 */
2471	ip6 = mtod(m, struct ip6_hdr *);
2472	ip6->ip6_flow = 0;
2473	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2474	ip6->ip6_vfc |= IPV6_VERSION;
2475	/* ip6->ip6_plen will be set later */
2476	ip6->ip6_nxt = IPPROTO_ICMPV6;
2477	ip6->ip6_hlim = 255;
2478	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2479	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2480	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2481
2482	/* ND Redirect */
2483	nd_rd = (struct nd_redirect *)(ip6 + 1);
2484	nd_rd->nd_rd_type = ND_REDIRECT;
2485	nd_rd->nd_rd_code = 0;
2486	nd_rd->nd_rd_reserved = 0;
2487	if (rt->rt_flags & RTF_GATEWAY) {
2488		/*
2489		 * nd_rd->nd_rd_target must be a link-local address in
2490		 * better router cases.
2491		 */
2492		if (!router_ll6)
2493			goto fail;
2494		bcopy(router_ll6, &nd_rd->nd_rd_target,
2495		    sizeof(nd_rd->nd_rd_target));
2496		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2497		    sizeof(nd_rd->nd_rd_dst));
2498	} else {
2499		/* make sure redtgt == reddst */
2500		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2501		    sizeof(nd_rd->nd_rd_target));
2502		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2503		    sizeof(nd_rd->nd_rd_dst));
2504	}
2505
2506	p = (u_char *)(nd_rd + 1);
2507
2508	if (!router_ll6)
2509		goto nolladdropt;
2510
2511	{
2512		/* target lladdr option */
2513		struct rtentry *rt_router = NULL;
2514		int len;
2515		struct sockaddr_dl *sdl;
2516		struct nd_opt_hdr *nd_opt;
2517		char *lladdr;
2518
2519		rt_router = nd6_lookup(router_ll6, 0, ifp);
2520		if (!rt_router)
2521			goto nolladdropt;
2522		len = sizeof(*nd_opt) + ifp->if_addrlen;
2523		len = (len + 7) & ~7;	/* round by 8 */
2524		/* safety check */
2525		if (len + (p - (u_char *)ip6) > maxlen)
2526			goto nolladdropt;
2527		if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2528		    (rt_router->rt_flags & RTF_LLINFO) &&
2529		    (rt_router->rt_gateway->sa_family == AF_LINK) &&
2530		    (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2531		    sdl->sdl_alen) {
2532			nd_opt = (struct nd_opt_hdr *)p;
2533			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2534			nd_opt->nd_opt_len = len >> 3;
2535			lladdr = (char *)(nd_opt + 1);
2536			bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2537			p += len;
2538		}
2539	}
2540nolladdropt:;
2541
2542	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2543
2544	/* just to be safe */
2545#ifdef M_DECRYPTED	/*not openbsd*/
2546	if (m0->m_flags & M_DECRYPTED)
2547		goto noredhdropt;
2548#endif
2549	if (p - (u_char *)ip6 > maxlen)
2550		goto noredhdropt;
2551
2552	{
2553		/* redirected header option */
2554		int len;
2555		struct nd_opt_rd_hdr *nd_opt_rh;
2556
2557		/*
2558		 * compute the maximum size for icmp6 redirect header option.
2559		 * XXX room for auth header?
2560		 */
2561		len = maxlen - (p - (u_char *)ip6);
2562		len &= ~7;
2563
2564		/* This is just for simplicity. */
2565		if (m0->m_pkthdr.len != m0->m_len) {
2566			if (m0->m_next) {
2567				m_freem(m0->m_next);
2568				m0->m_next = NULL;
2569			}
2570			m0->m_pkthdr.len = m0->m_len;
2571		}
2572
2573		/*
2574		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2575		 * about padding/truncate rule for the original IP packet.
2576		 * From the discussion on IPv6imp in Feb 1999,
2577		 * the consensus was:
2578		 * - "attach as much as possible" is the goal
2579		 * - pad if not aligned (original size can be guessed by
2580		 *   original ip6 header)
2581		 * Following code adds the padding if it is simple enough,
2582		 * and truncates if not.
2583		 */
2584		if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2585			panic("assumption failed in %s:%d", __FILE__,
2586			    __LINE__);
2587
2588		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2589			/* not enough room, truncate */
2590			m0->m_pkthdr.len = m0->m_len = len -
2591			    sizeof(*nd_opt_rh);
2592		} else {
2593			/* enough room, pad or truncate */
2594			size_t extra;
2595
2596			extra = m0->m_pkthdr.len % 8;
2597			if (extra) {
2598				/* pad if easy enough, truncate if not */
2599				if (8 - extra <= M_TRAILINGSPACE(m0)) {
2600					/* pad */
2601					m0->m_len += (8 - extra);
2602					m0->m_pkthdr.len += (8 - extra);
2603				} else {
2604					/* truncate */
2605					m0->m_pkthdr.len -= extra;
2606					m0->m_len -= extra;
2607				}
2608			}
2609			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2610			m0->m_pkthdr.len = m0->m_len = len -
2611			    sizeof(*nd_opt_rh);
2612		}
2613
2614		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2615		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2616		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2617		nd_opt_rh->nd_opt_rh_len = len >> 3;
2618		p += sizeof(*nd_opt_rh);
2619		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2620
2621		/* connect m0 to m */
2622		m_tag_delete_chain(m0, NULL);
2623		m0->m_flags &= ~M_PKTHDR;
2624		m->m_next = m0;
2625		m->m_pkthdr.len = m->m_len + m0->m_len;
2626		m0 = NULL;
2627	}
2628noredhdropt:;
2629	if (m0) {
2630		m_freem(m0);
2631		m0 = NULL;
2632	}
2633
2634	/* XXX: clear embedded link IDs in the inner header */
2635	in6_clearscope(&sip6->ip6_src);
2636	in6_clearscope(&sip6->ip6_dst);
2637	in6_clearscope(&nd_rd->nd_rd_target);
2638	in6_clearscope(&nd_rd->nd_rd_dst);
2639
2640	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2641
2642	nd_rd->nd_rd_cksum = 0;
2643	nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2644	    sizeof(*ip6), ntohs(ip6->ip6_plen));
2645
2646	/* send the packet to outside... */
2647	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2648	if (outif) {
2649		icmp6_ifstat_inc(outif, ifs6_out_msg);
2650		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2651	}
2652	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2653
2654	return;
2655
2656fail:
2657	if (m)
2658		m_freem(m);
2659	if (m0)
2660		m_freem(m0);
2661}
2662
2663/*
2664 * ICMPv6 socket option processing.
2665 */
2666int
2667icmp6_ctloutput(so, sopt)
2668	struct socket *so;
2669	struct sockopt *sopt;
2670{
2671	int error = 0;
2672	int optlen;
2673	struct inpcb *inp = sotoinpcb(so);
2674	int level, op, optname;
2675
2676	if (sopt) {
2677		level = sopt->sopt_level;
2678		op = sopt->sopt_dir;
2679		optname = sopt->sopt_name;
2680		optlen = sopt->sopt_valsize;
2681	} else
2682		level = op = optname = optlen = 0;
2683
2684	if (level != IPPROTO_ICMPV6) {
2685		return EINVAL;
2686	}
2687
2688	switch (op) {
2689	case PRCO_SETOPT:
2690		switch (optname) {
2691		case ICMP6_FILTER:
2692		    {
2693			struct icmp6_filter *p;
2694
2695			if (optlen != sizeof(*p)) {
2696				error = EMSGSIZE;
2697				break;
2698			}
2699			if (inp->in6p_icmp6filt == NULL) {
2700				error = EINVAL;
2701				break;
2702			}
2703			error = sooptcopyin(sopt, inp->in6p_icmp6filt, optlen,
2704				optlen);
2705			break;
2706		    }
2707
2708		default:
2709			error = ENOPROTOOPT;
2710			break;
2711		}
2712		break;
2713
2714	case PRCO_GETOPT:
2715		switch (optname) {
2716		case ICMP6_FILTER:
2717		    {
2718			if (inp->in6p_icmp6filt == NULL) {
2719				error = EINVAL;
2720				break;
2721			}
2722			error = sooptcopyout(sopt, inp->in6p_icmp6filt,
2723				sizeof(struct icmp6_filter));
2724			break;
2725		    }
2726
2727		default:
2728			error = ENOPROTOOPT;
2729			break;
2730		}
2731		break;
2732	}
2733
2734	return (error);
2735}
2736
2737/*
2738 * Perform rate limit check.
2739 * Returns 0 if it is okay to send the icmp6 packet.
2740 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2741 * limitation.
2742 *
2743 * XXX per-destination/type check necessary?
2744 */
2745static int
2746icmp6_ratelimit(dst, type, code)
2747	const struct in6_addr *dst;	/* not used at this moment */
2748	const int type;			/* not used at this moment */
2749	const int code;			/* not used at this moment */
2750{
2751	int ret;
2752
2753	ret = 0;	/* okay to send */
2754
2755	/* PPS limit */
2756	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2757	    icmp6errppslim)) {
2758		/* The packet is subject to rate limit */
2759		ret++;
2760	}
2761
2762	return ret;
2763}
2764