icmp6.c revision 83934
1/*	$FreeBSD: head/sys/netinet6/icmp6.c 83934 2001-09-25 18:40:52Z brooks $	*/
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/systm.h>
74#include <sys/malloc.h>
75#include <sys/mbuf.h>
76#include <sys/protosw.h>
77#include <sys/socket.h>
78#include <sys/socketvar.h>
79#include <sys/time.h>
80#include <sys/kernel.h>
81#include <sys/syslog.h>
82#include <sys/domain.h>
83
84#include <net/if.h>
85#include <net/route.h>
86#include <net/if_dl.h>
87#include <net/if_types.h>
88
89#include <netinet/in.h>
90#include <netinet/in_var.h>
91#include <netinet/ip6.h>
92#include <netinet6/ip6_var.h>
93#include <netinet/icmp6.h>
94#include <netinet6/mld6_var.h>
95#include <netinet/in_pcb.h>
96#include <netinet6/in6_pcb.h>
97#include <netinet6/nd6.h>
98#include <netinet6/in6_ifattach.h>
99#include <netinet6/ip6protosw.h>
100
101#ifdef IPSEC
102#include <netinet6/ipsec.h>
103#include <netkey/key.h>
104#endif
105
106#include <net/net_osdep.h>
107
108#ifdef HAVE_NRL_INPCB
109/* inpcb members */
110#define in6pcb		inpcb
111#define in6p_laddr	inp_laddr6
112#define in6p_faddr	inp_faddr6
113#define in6p_icmp6filt	inp_icmp6filt
114#define in6p_route	inp_route
115#define in6p_socket	inp_socket
116#define in6p_flags	inp_flags
117#define in6p_moptions	inp_moptions6
118#define in6p_outputopts	inp_outputopts6
119#define in6p_ip6	inp_ipv6
120#define in6p_flowinfo	inp_flowinfo
121#define in6p_sp		inp_sp
122#define in6p_next	inp_next
123#define in6p_prev	inp_prev
124/* macro names */
125#define sotoin6pcb	sotoinpcb
126/* function names */
127#define in6_pcbdetach	in_pcbdetach
128#define in6_rtchange	in_rtchange
129
130/*
131 * for KAME src sync over BSD*'s. XXX: FreeBSD (>=3) are VERY different from
132 * others...
133 */
134#define in6p_ip6_nxt	inp_ipv6.ip6_nxt
135#endif
136
137extern struct domain inet6domain;
138extern struct ip6protosw inet6sw[];
139extern u_char ip6_protox[];
140
141struct icmp6stat icmp6stat;
142
143extern struct inpcbhead ripcb;
144extern int icmp6errppslim;
145static int icmp6errpps_count = 0;
146static struct timeval icmp6errppslim_last;
147extern int icmp6_nodeinfo;
148
149static void icmp6_errcount __P((struct icmp6errstat *, int, int));
150static int icmp6_rip6_input __P((struct mbuf **, int));
151static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int));
152static const char *icmp6_redirect_diag __P((struct in6_addr *,
153	struct in6_addr *, struct in6_addr *));
154#ifndef HAVE_PPSRATECHECK
155static int ppsratecheck __P((struct timeval *, int *, int));
156#endif
157static struct mbuf *ni6_input __P((struct mbuf *, int));
158static struct mbuf *ni6_nametodns __P((const char *, int, int));
159static int ni6_dnsmatch __P((const char *, int, const char *, int));
160static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
161			  struct ifnet **, char *));
162static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
163				struct ifnet *, int));
164static int icmp6_notify_error __P((struct mbuf *, int, int, int));
165
166#ifdef COMPAT_RFC1885
167static struct route_in6 icmp6_reflect_rt;
168#endif
169
170
171void
172icmp6_init()
173{
174	mld6_init();
175}
176
177static void
178icmp6_errcount(stat, type, code)
179	struct icmp6errstat *stat;
180	int type, code;
181{
182	switch (type) {
183	case ICMP6_DST_UNREACH:
184		switch (code) {
185		case ICMP6_DST_UNREACH_NOROUTE:
186			stat->icp6errs_dst_unreach_noroute++;
187			return;
188		case ICMP6_DST_UNREACH_ADMIN:
189			stat->icp6errs_dst_unreach_admin++;
190			return;
191		case ICMP6_DST_UNREACH_BEYONDSCOPE:
192			stat->icp6errs_dst_unreach_beyondscope++;
193			return;
194		case ICMP6_DST_UNREACH_ADDR:
195			stat->icp6errs_dst_unreach_addr++;
196			return;
197		case ICMP6_DST_UNREACH_NOPORT:
198			stat->icp6errs_dst_unreach_noport++;
199			return;
200		}
201		break;
202	case ICMP6_PACKET_TOO_BIG:
203		stat->icp6errs_packet_too_big++;
204		return;
205	case ICMP6_TIME_EXCEEDED:
206		switch (code) {
207		case ICMP6_TIME_EXCEED_TRANSIT:
208			stat->icp6errs_time_exceed_transit++;
209			return;
210		case ICMP6_TIME_EXCEED_REASSEMBLY:
211			stat->icp6errs_time_exceed_reassembly++;
212			return;
213		}
214		break;
215	case ICMP6_PARAM_PROB:
216		switch (code) {
217		case ICMP6_PARAMPROB_HEADER:
218			stat->icp6errs_paramprob_header++;
219			return;
220		case ICMP6_PARAMPROB_NEXTHEADER:
221			stat->icp6errs_paramprob_nextheader++;
222			return;
223		case ICMP6_PARAMPROB_OPTION:
224			stat->icp6errs_paramprob_option++;
225			return;
226		}
227		break;
228	case ND_REDIRECT:
229		stat->icp6errs_redirect++;
230		return;
231	}
232	stat->icp6errs_unknown++;
233}
234
235/*
236 * Generate an error packet of type error in response to bad IP6 packet.
237 */
238void
239icmp6_error(m, type, code, param)
240	struct mbuf *m;
241	int type, code, param;
242{
243	struct ip6_hdr *oip6, *nip6;
244	struct icmp6_hdr *icmp6;
245	u_int preplen;
246	int off;
247	int nxt;
248
249	icmp6stat.icp6s_error++;
250
251	/* count per-type-code statistics */
252	icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
253
254#ifdef M_DECRYPTED	/*not openbsd*/
255	if (m->m_flags & M_DECRYPTED) {
256		icmp6stat.icp6s_canterror++;
257		goto freeit;
258	}
259#endif
260
261#ifndef PULLDOWN_TEST
262	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
263#else
264	if (m->m_len < sizeof(struct ip6_hdr)) {
265		m = m_pullup(m, sizeof(struct ip6_hdr));
266		if (m == NULL)
267			return;
268	}
269#endif
270	oip6 = mtod(m, struct ip6_hdr *);
271
272	/*
273	 * Multicast destination check. For unrecognized option errors,
274	 * this check has already done in ip6_unknown_opt(), so we can
275	 * check only for other errors.
276	 */
277	if ((m->m_flags & (M_BCAST|M_MCAST) ||
278	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
279	    (type != ICMP6_PACKET_TOO_BIG &&
280	     (type != ICMP6_PARAM_PROB ||
281	      code != ICMP6_PARAMPROB_OPTION)))
282		goto freeit;
283
284	/* Source address check. XXX: the case of anycast source? */
285	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
286	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
287		goto freeit;
288
289	/*
290	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
291	 * don't do it.
292	 */
293	nxt = -1;
294	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
295	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
296		struct icmp6_hdr *icp;
297
298#ifndef PULLDOWN_TEST
299		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
300		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
301#else
302		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
303			sizeof(*icp));
304		if (icp == NULL) {
305			icmp6stat.icp6s_tooshort++;
306			return;
307		}
308#endif
309		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
310		    icp->icmp6_type == ND_REDIRECT) {
311			/*
312			 * ICMPv6 error
313			 * Special case: for redirect (which is
314			 * informational) we must not send icmp6 error.
315			 */
316			icmp6stat.icp6s_canterror++;
317			goto freeit;
318		} else {
319			/* ICMPv6 informational - send the error */
320		}
321	} else {
322		/* non-ICMPv6 - send the error */
323	}
324
325	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
326
327	/* Finally, do rate limitation check. */
328	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
329		icmp6stat.icp6s_toofreq++;
330		goto freeit;
331	}
332
333	/*
334	 * OK, ICMP6 can be generated.
335	 */
336
337	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
338		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
339
340	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
341	M_PREPEND(m, preplen, M_DONTWAIT);
342	if (m && m->m_len < preplen)
343		m = m_pullup(m, preplen);
344	if (m == NULL) {
345		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
346		return;
347	}
348
349	nip6 = mtod(m, struct ip6_hdr *);
350	nip6->ip6_src  = oip6->ip6_src;
351	nip6->ip6_dst  = oip6->ip6_dst;
352
353	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src))
354		oip6->ip6_src.s6_addr16[1] = 0;
355	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst))
356		oip6->ip6_dst.s6_addr16[1] = 0;
357
358	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
359	icmp6->icmp6_type = type;
360	icmp6->icmp6_code = code;
361	icmp6->icmp6_pptr = htonl((u_int32_t)param);
362
363	/*
364	 * icmp6_reflect() is designed to be in the input path.
365	 * icmp6_error() can be called from both input and outut path,
366	 * and if we are in output path rcvif could contain bogus value.
367	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
368	 * information in ip header (nip6).
369	 */
370	m->m_pkthdr.rcvif = NULL;
371
372	icmp6stat.icp6s_outhist[type]++;
373	icmp6_reflect(m, sizeof(struct ip6_hdr)); /*header order: IPv6 - ICMPv6*/
374
375	return;
376
377  freeit:
378	/*
379	 * If we can't tell wheter or not we can generate ICMP6, free it.
380	 */
381	m_freem(m);
382}
383
384/*
385 * Process a received ICMP6 message.
386 */
387int
388icmp6_input(mp, offp, proto)
389	struct mbuf **mp;
390	int *offp, proto;
391{
392	struct mbuf *m = *mp, *n;
393	struct ip6_hdr *ip6, *nip6;
394	struct icmp6_hdr *icmp6, *nicmp6;
395	int off = *offp;
396	int icmp6len = m->m_pkthdr.len - *offp;
397	int code, sum, noff;
398
399#ifndef PULLDOWN_TEST
400	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
401	/* m might change if M_LOOP. So, call mtod after this */
402#endif
403
404	/*
405	 * Locate icmp6 structure in mbuf, and check
406	 * that not corrupted and of at least minimum length
407	 */
408
409	ip6 = mtod(m, struct ip6_hdr *);
410	if (icmp6len < sizeof(struct icmp6_hdr)) {
411		icmp6stat.icp6s_tooshort++;
412		goto freeit;
413	}
414
415	/*
416	 * calculate the checksum
417	 */
418#ifndef PULLDOWN_TEST
419	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
420#else
421	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
422	if (icmp6 == NULL) {
423		icmp6stat.icp6s_tooshort++;
424		return IPPROTO_DONE;
425	}
426#endif
427	code = icmp6->icmp6_code;
428
429	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
430		nd6log((LOG_ERR,
431		    "ICMP6 checksum error(%d|%x) %s\n",
432		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
433		icmp6stat.icp6s_checksum++;
434		goto freeit;
435	}
436
437	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
438		/*
439		 * Deliver very specific ICMP6 type only.
440		 * This is important to deilver TOOBIG.  Otherwise PMTUD
441		 * will not work.
442		 */
443		switch (icmp6->icmp6_type) {
444		case ICMP6_DST_UNREACH:
445		case ICMP6_PACKET_TOO_BIG:
446		case ICMP6_TIME_EXCEEDED:
447			break;
448		default:
449			goto freeit;
450		}
451	}
452
453	icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
454	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
455	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
456		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
457
458	switch (icmp6->icmp6_type) {
459	case ICMP6_DST_UNREACH:
460		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
461		switch (code) {
462		case ICMP6_DST_UNREACH_NOROUTE:
463			code = PRC_UNREACH_NET;
464			break;
465		case ICMP6_DST_UNREACH_ADMIN:
466			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
467			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
468			break;
469		case ICMP6_DST_UNREACH_ADDR:
470			code = PRC_HOSTDEAD;
471			break;
472#ifdef COMPAT_RFC1885
473		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
474			code = PRC_UNREACH_SRCFAIL;
475			break;
476#else
477		case ICMP6_DST_UNREACH_BEYONDSCOPE:
478			/* I mean "source address was incorrect." */
479			code = PRC_PARAMPROB;
480			break;
481#endif
482		case ICMP6_DST_UNREACH_NOPORT:
483			code = PRC_UNREACH_PORT;
484			break;
485		default:
486			goto badcode;
487		}
488		goto deliver;
489		break;
490
491	case ICMP6_PACKET_TOO_BIG:
492		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
493		if (code != 0)
494			goto badcode;
495
496		code = PRC_MSGSIZE;
497
498		/*
499		 * Updating the path MTU will be done after examining
500		 * intermediate extension headers.
501		 */
502		goto deliver;
503		break;
504
505	case ICMP6_TIME_EXCEEDED:
506		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
507		switch (code) {
508		case ICMP6_TIME_EXCEED_TRANSIT:
509		case ICMP6_TIME_EXCEED_REASSEMBLY:
510			code += PRC_TIMXCEED_INTRANS;
511			break;
512		default:
513			goto badcode;
514		}
515		goto deliver;
516		break;
517
518	case ICMP6_PARAM_PROB:
519		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
520		switch (code) {
521		case ICMP6_PARAMPROB_NEXTHEADER:
522			code = PRC_UNREACH_PROTOCOL;
523			break;
524		case ICMP6_PARAMPROB_HEADER:
525		case ICMP6_PARAMPROB_OPTION:
526			code = PRC_PARAMPROB;
527			break;
528		default:
529			goto badcode;
530		}
531		goto deliver;
532		break;
533
534	case ICMP6_ECHO_REQUEST:
535		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
536		if (code != 0)
537			goto badcode;
538		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
539			/* Give up remote */
540			break;
541		}
542		if ((n->m_flags & M_EXT) != 0
543		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
544			struct mbuf *n0 = n;
545			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
546
547			/*
548			 * Prepare an internal mbuf. m_pullup() doesn't
549			 * always copy the length we specified.
550			 */
551			if (maxlen >= MCLBYTES) {
552				/* Give up remote */
553				m_freem(n0);
554				break;
555			}
556			MGETHDR(n, M_DONTWAIT, n0->m_type);
557			if (n && maxlen >= MHLEN) {
558				MCLGET(n, M_DONTWAIT);
559				if ((n->m_flags & M_EXT) == 0) {
560					m_free(n);
561					n = NULL;
562				}
563			}
564			if (n == NULL) {
565				/* Give up remote */
566				m_freem(n0);
567				break;
568			}
569			M_COPY_PKTHDR(n, n0);
570			/*
571			 * Copy IPv6 and ICMPv6 only.
572			 */
573			nip6 = mtod(n, struct ip6_hdr *);
574			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
575			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
576			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
577			noff = sizeof(struct ip6_hdr);
578			n->m_pkthdr.len = n->m_len =
579				noff + sizeof(struct icmp6_hdr);
580			/*
581			 * Adjust mbuf. ip6_plen will be adjusted in
582			 * ip6_output().
583			 */
584			m_adj(n0, off + sizeof(struct icmp6_hdr));
585			n->m_pkthdr.len += n0->m_pkthdr.len;
586			n->m_next = n0;
587			n0->m_flags &= ~M_PKTHDR;
588		} else {
589			nip6 = mtod(n, struct ip6_hdr *);
590			nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
591			noff = off;
592		}
593		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
594		nicmp6->icmp6_code = 0;
595		if (n) {
596			icmp6stat.icp6s_reflect++;
597			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
598			icmp6_reflect(n, noff);
599		}
600		break;
601
602	case ICMP6_ECHO_REPLY:
603		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
604		if (code != 0)
605			goto badcode;
606		break;
607
608	case MLD6_LISTENER_QUERY:
609	case MLD6_LISTENER_REPORT:
610		if (icmp6len < sizeof(struct mld6_hdr))
611			goto badlen;
612		if (icmp6->icmp6_type == MLD6_LISTENER_QUERY) /* XXX: ugly... */
613			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
614		else
615			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
616		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
617			/* give up local */
618			mld6_input(m, off);
619			m = NULL;
620			goto freeit;
621		}
622		mld6_input(n, off);
623		/* m stays. */
624		break;
625
626	case MLD6_LISTENER_DONE:
627		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
628		if (icmp6len < sizeof(struct mld6_hdr))	/* necessary? */
629			goto badlen;
630		break;		/* nothing to be done in kernel */
631
632	case MLD6_MTRACE_RESP:
633	case MLD6_MTRACE:
634		/* XXX: these two are experimental. not officially defind. */
635		/* XXX: per-interface statistics? */
636		break;		/* just pass it to applications */
637
638	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
639	    {
640		enum { WRU, FQDN } mode;
641
642		if (!icmp6_nodeinfo)
643			break;
644
645		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
646			mode = WRU;
647		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
648			mode = FQDN;
649		else
650			goto badlen;
651
652#define hostnamelen	strlen(hostname)
653		if (mode == FQDN) {
654#ifndef PULLDOWN_TEST
655			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
656					 IPPROTO_DONE);
657#endif
658			n = m_copy(m, 0, M_COPYALL);
659			if (n)
660				n = ni6_input(n, off);
661			/* XXX meaningless if n == NULL */
662			noff = sizeof(struct ip6_hdr);
663		} else {
664			u_char *p;
665			int maxlen, maxhlen;
666
667			if ((icmp6_nodeinfo & 5) != 5)
668				break;
669
670			if (code != 0)
671				goto badcode;
672			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
673			if (maxlen >= MCLBYTES) {
674				/* Give up remote */
675				break;
676			}
677			MGETHDR(n, M_DONTWAIT, m->m_type);
678			if (n && maxlen > MHLEN) {
679				MCLGET(n, M_DONTWAIT);
680				if ((n->m_flags & M_EXT) == 0) {
681					m_free(n);
682					n = NULL;
683				}
684			}
685			if (n == NULL) {
686				/* Give up remote */
687				break;
688			}
689			n->m_pkthdr.rcvif = NULL;
690			n->m_len = 0;
691			maxhlen = M_TRAILINGSPACE(n) - maxlen;
692			if (maxhlen > hostnamelen)
693				maxhlen = hostnamelen;
694			/*
695			 * Copy IPv6 and ICMPv6 only.
696			 */
697			nip6 = mtod(n, struct ip6_hdr *);
698			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
699			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
700			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
701			p = (u_char *)(nicmp6 + 1);
702			bzero(p, 4);
703			bcopy(hostname, p + 4, maxhlen); /*meaningless TTL*/
704			noff = sizeof(struct ip6_hdr);
705			M_COPY_PKTHDR(n, m); /* just for recvif */
706			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
707				sizeof(struct icmp6_hdr) + 4 + maxhlen;
708			nicmp6->icmp6_type = ICMP6_WRUREPLY;
709			nicmp6->icmp6_code = 0;
710		}
711#undef hostnamelen
712		if (n) {
713			icmp6stat.icp6s_reflect++;
714			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
715			icmp6_reflect(n, noff);
716		}
717		break;
718	    }
719
720	case ICMP6_WRUREPLY:
721		if (code != 0)
722			goto badcode;
723		break;
724
725	case ND_ROUTER_SOLICIT:
726		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
727		if (code != 0)
728			goto badcode;
729		if (icmp6len < sizeof(struct nd_router_solicit))
730			goto badlen;
731		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
732			/* give up local */
733			nd6_rs_input(m, off, icmp6len);
734			m = NULL;
735			goto freeit;
736		}
737		nd6_rs_input(n, off, icmp6len);
738		/* m stays. */
739		break;
740
741	case ND_ROUTER_ADVERT:
742		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
743		if (code != 0)
744			goto badcode;
745		if (icmp6len < sizeof(struct nd_router_advert))
746			goto badlen;
747		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
748			/* give up local */
749			nd6_ra_input(m, off, icmp6len);
750			m = NULL;
751			goto freeit;
752		}
753		nd6_ra_input(n, off, icmp6len);
754		/* m stays. */
755		break;
756
757	case ND_NEIGHBOR_SOLICIT:
758		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
759		if (code != 0)
760			goto badcode;
761		if (icmp6len < sizeof(struct nd_neighbor_solicit))
762			goto badlen;
763		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
764			/* give up local */
765			nd6_ns_input(m, off, icmp6len);
766			m = NULL;
767			goto freeit;
768		}
769		nd6_ns_input(n, off, icmp6len);
770		/* m stays. */
771		break;
772
773	case ND_NEIGHBOR_ADVERT:
774		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
775		if (code != 0)
776			goto badcode;
777		if (icmp6len < sizeof(struct nd_neighbor_advert))
778			goto badlen;
779		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
780			/* give up local */
781			nd6_na_input(m, off, icmp6len);
782			m = NULL;
783			goto freeit;
784		}
785		nd6_na_input(n, off, icmp6len);
786		/* m stays. */
787		break;
788
789	case ND_REDIRECT:
790		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
791		if (code != 0)
792			goto badcode;
793		if (icmp6len < sizeof(struct nd_redirect))
794			goto badlen;
795		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
796			/* give up local */
797			icmp6_redirect_input(m, off);
798			m = NULL;
799			goto freeit;
800		}
801		icmp6_redirect_input(n, off);
802		/* m stays. */
803		break;
804
805	case ICMP6_ROUTER_RENUMBERING:
806		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
807		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
808			goto badcode;
809		if (icmp6len < sizeof(struct icmp6_router_renum))
810			goto badlen;
811		break;
812
813	default:
814		nd6log((LOG_DEBUG,
815		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
816		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
817		    ip6_sprintf(&ip6->ip6_dst),
818		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
819		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
820			/* ICMPv6 error: MUST deliver it by spec... */
821			code = PRC_NCMDS;
822			/* deliver */
823		} else {
824			/* ICMPv6 informational: MUST not deliver */
825			break;
826		}
827	deliver:
828		if (icmp6_notify_error(m, off, icmp6len, code)) {
829			/* In this case, m should've been freed. */
830			return(IPPROTO_DONE);
831		}
832		break;
833
834	badcode:
835		icmp6stat.icp6s_badcode++;
836		break;
837
838	badlen:
839		icmp6stat.icp6s_badlen++;
840		break;
841	}
842
843	/* deliver the packet to appropriate sockets */
844	icmp6_rip6_input(&m, *offp);
845
846	return IPPROTO_DONE;
847
848 freeit:
849	m_freem(m);
850	return IPPROTO_DONE;
851}
852
853static int
854icmp6_notify_error(m, off, icmp6len, code)
855	struct mbuf *m;
856	int off, icmp6len;
857{
858	struct icmp6_hdr *icmp6;
859	struct ip6_hdr *eip6;
860	u_int32_t notifymtu;
861	struct sockaddr_in6 icmp6src, icmp6dst;
862
863	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
864		icmp6stat.icp6s_tooshort++;
865		goto freeit;
866	}
867#ifndef PULLDOWN_TEST
868	IP6_EXTHDR_CHECK(m, off,
869			 sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr),
870			 -1);
871	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
872#else
873	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
874		       sizeof(*icmp6) + sizeof(struct ip6_hdr));
875	if (icmp6 == NULL) {
876		icmp6stat.icp6s_tooshort++;
877		return(-1);
878	}
879#endif
880	eip6 = (struct ip6_hdr *)(icmp6 + 1);
881
882	/* Detect the upper level protocol */
883	{
884		void (*ctlfunc) __P((int, struct sockaddr *, void *));
885		u_int8_t nxt = eip6->ip6_nxt;
886		int eoff = off + sizeof(struct icmp6_hdr) +
887			sizeof(struct ip6_hdr);
888		struct ip6ctlparam ip6cp;
889		struct in6_addr *finaldst = NULL;
890		int icmp6type = icmp6->icmp6_type;
891		struct ip6_frag *fh;
892		struct ip6_rthdr *rth;
893		struct ip6_rthdr0 *rth0;
894		int rthlen;
895
896		while (1) { /* XXX: should avoid inf. loop explicitly? */
897			struct ip6_ext *eh;
898
899			switch (nxt) {
900			case IPPROTO_HOPOPTS:
901			case IPPROTO_DSTOPTS:
902			case IPPROTO_AH:
903#ifndef PULLDOWN_TEST
904				IP6_EXTHDR_CHECK(m, 0, eoff +
905						 sizeof(struct ip6_ext),
906						 -1);
907				eh = (struct ip6_ext *)(mtod(m, caddr_t)
908							+ 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),
935						 -1);
936				rth = (struct ip6_rthdr *)(mtod(m, caddr_t)
937							   + eoff);
938#else
939				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
940					       eoff, sizeof(*rth));
941				if (rth == NULL) {
942					icmp6stat.icp6s_tooshort++;
943					return(-1);
944				}
945#endif
946				rthlen = (rth->ip6r_len + 1) << 3;
947				/*
948				 * XXX: currently there is no
949				 * officially defined type other
950				 * than type-0.
951				 * Note that if the segment left field
952				 * is 0, all intermediate hops must
953				 * have been passed.
954				 */
955				if (rth->ip6r_segleft &&
956				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
957					int hops;
958
959#ifndef PULLDOWN_TEST
960					IP6_EXTHDR_CHECK(m, 0, eoff + rthlen,
961							 -1);
962					rth0 = (struct ip6_rthdr0 *)(mtod(m, caddr_t) + eoff);
963#else
964					IP6_EXTHDR_GET(rth0,
965						       struct ip6_rthdr0 *, m,
966						       eoff, rthlen);
967					if (rth0 == NULL) {
968						icmp6stat.icp6s_tooshort++;
969						return(-1);
970					}
971#endif
972					/* just ignore a bogus header */
973					if ((rth0->ip6r0_len % 2) == 0 &&
974					    (hops = rth0->ip6r0_len/2))
975						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
976				}
977				eoff += rthlen;
978				nxt = rth->ip6r_nxt;
979				break;
980			case IPPROTO_FRAGMENT:
981#ifndef PULLDOWN_TEST
982				IP6_EXTHDR_CHECK(m, 0, eoff +
983						 sizeof(struct ip6_frag),
984						 -1);
985				fh = (struct ip6_frag *)(mtod(m, caddr_t)
986							 + eoff);
987#else
988				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
989					       eoff, sizeof(*fh));
990				if (fh == NULL) {
991					icmp6stat.icp6s_tooshort++;
992					return(-1);
993				}
994#endif
995				/*
996				 * Data after a fragment header is meaningless
997				 * unless it is the first fragment, but
998				 * we'll go to the notify label for path MTU
999				 * discovery.
1000				 */
1001				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1002					goto notify;
1003
1004				eoff += sizeof(struct ip6_frag);
1005				nxt = fh->ip6f_nxt;
1006				break;
1007			default:
1008				/*
1009				 * This case includes ESP and the No Next
1010				 * Header. In such cases going to the notify
1011				 * label does not have any meaning
1012				 * (i.e. ctlfunc will be NULL), but we go
1013				 * anyway since we might have to update
1014				 * path MTU information.
1015				 */
1016				goto notify;
1017			}
1018		}
1019	  notify:
1020#ifndef PULLDOWN_TEST
1021		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1022#else
1023		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1024			       sizeof(*icmp6) + sizeof(struct ip6_hdr));
1025		if (icmp6 == NULL) {
1026			icmp6stat.icp6s_tooshort++;
1027			return(-1);
1028		}
1029#endif
1030
1031		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1032		bzero(&icmp6dst, sizeof(icmp6dst));
1033		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1034		icmp6dst.sin6_family = AF_INET6;
1035		if (finaldst == NULL)
1036			icmp6dst.sin6_addr = eip6->ip6_dst;
1037		else
1038			icmp6dst.sin6_addr = *finaldst;
1039		icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1040							  &icmp6dst.sin6_addr);
1041#ifndef SCOPEDROUTING
1042		if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1043				   NULL, NULL)) {
1044			/* should be impossbile */
1045			nd6log((LOG_DEBUG,
1046			    "icmp6_notify_error: in6_embedscope failed\n"));
1047			goto freeit;
1048		}
1049#endif
1050
1051		/*
1052		 * retrieve parameters from the inner IPv6 header, and convert
1053		 * them into sockaddr structures.
1054		 */
1055		bzero(&icmp6src, sizeof(icmp6src));
1056		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1057		icmp6src.sin6_family = AF_INET6;
1058		icmp6src.sin6_addr = eip6->ip6_src;
1059		icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1060							  &icmp6src.sin6_addr);
1061#ifndef SCOPEDROUTING
1062		if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1063				   NULL, NULL)) {
1064			/* should be impossbile */
1065			nd6log((LOG_DEBUG,
1066			    "icmp6_notify_error: in6_embedscope failed\n"));
1067			goto freeit;
1068		}
1069#endif
1070		icmp6src.sin6_flowinfo =
1071			(eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1072
1073		if (finaldst == NULL)
1074			finaldst = &eip6->ip6_dst;
1075		ip6cp.ip6c_m = m;
1076		ip6cp.ip6c_icmp6 = icmp6;
1077		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1078		ip6cp.ip6c_off = eoff;
1079		ip6cp.ip6c_finaldst = finaldst;
1080		ip6cp.ip6c_src = &icmp6src;
1081		ip6cp.ip6c_nxt = nxt;
1082
1083		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1084			notifymtu = ntohl(icmp6->icmp6_mtu);
1085			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1086			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1087		}
1088
1089		ctlfunc = (void (*) __P((int, struct sockaddr *, void *)))
1090			(inet6sw[ip6_protox[nxt]].pr_ctlinput);
1091		if (ctlfunc) {
1092			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1093					  &ip6cp);
1094		}
1095	}
1096	return(0);
1097
1098  freeit:
1099	m_freem(m);
1100	return(-1);
1101}
1102
1103void
1104icmp6_mtudisc_update(ip6cp, validated)
1105	struct ip6ctlparam *ip6cp;
1106	int validated;
1107{
1108	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1109	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1110	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1111	u_int mtu = ntohl(icmp6->icmp6_mtu);
1112	struct rtentry *rt = NULL;
1113	struct sockaddr_in6 sin6;
1114
1115	if (!validated)
1116		return;
1117
1118	bzero(&sin6, sizeof(sin6));
1119	sin6.sin6_family = PF_INET6;
1120	sin6.sin6_len = sizeof(struct sockaddr_in6);
1121	sin6.sin6_addr = *dst;
1122	/* XXX normally, this won't happen */
1123	if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1124		sin6.sin6_addr.s6_addr16[1] =
1125		    htons(m->m_pkthdr.rcvif->if_index);
1126	}
1127	/* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1128	rt = rtalloc1((struct sockaddr *)&sin6, 0,
1129		      RTF_CLONING | RTF_PRCLONING);
1130
1131	if (rt && (rt->rt_flags & RTF_HOST)
1132	    && !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1133		if (mtu < IPV6_MMTU) {
1134				/* xxx */
1135			rt->rt_rmx.rmx_locks |= RTV_MTU;
1136		} else if (mtu < rt->rt_ifp->if_mtu &&
1137			   rt->rt_rmx.rmx_mtu > mtu) {
1138			icmp6stat.icp6s_pmtuchg++;
1139			rt->rt_rmx.rmx_mtu = mtu;
1140		}
1141	}
1142	if (rt)
1143		RTFREE(rt);
1144}
1145
1146/*
1147 * Process a Node Information Query packet, based on
1148 * draft-ietf-ipngwg-icmp-name-lookups-07.
1149 *
1150 * Spec incompatibilities:
1151 * - IPv6 Subject address handling
1152 * - IPv4 Subject address handling support missing
1153 * - Proxy reply (answer even if it's not for me)
1154 * - joins NI group address at in6_ifattach() time only, does not cope
1155 *   with hostname changes by sethostname(3)
1156 */
1157#define hostnamelen	strlen(hostname)
1158static struct mbuf *
1159ni6_input(m, off)
1160	struct mbuf *m;
1161	int off;
1162{
1163	struct icmp6_nodeinfo *ni6, *nni6;
1164	struct mbuf *n = NULL;
1165	u_int16_t qtype;
1166	int subjlen;
1167	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1168	struct ni_reply_fqdn *fqdn;
1169	int addrs;		/* for NI_QTYPE_NODEADDR */
1170	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1171	struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1172	struct sockaddr_in6 sin6_d; /* XXX: we should retrieve this from m_aux */
1173	struct ip6_hdr *ip6;
1174	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1175	char *subj = NULL;
1176	struct in6_ifaddr *ia6 = NULL;
1177
1178	ip6 = mtod(m, struct ip6_hdr *);
1179#ifndef PULLDOWN_TEST
1180	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1181#else
1182	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1183	if (ni6 == NULL) {
1184		/* m is already reclaimed */
1185		return NULL;
1186	}
1187#endif
1188
1189	/*
1190	 * Validate IPv6 destination address.
1191	 *
1192	 * The Responder must discard the Query without further processing
1193	 * unless it is one of the Responder's unicast or anycast addresses, or
1194	 * a link-local scope multicast address which the Responder has joined.
1195	 * [icmp-name-lookups-07, Section 4.]
1196	 */
1197	bzero(&sin6, sizeof(sin6));
1198	sin6.sin6_family = AF_INET6;
1199	sin6.sin6_len = sizeof(struct sockaddr_in6);
1200	bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1201	/* XXX scopeid */
1202	if ((ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)&sin6)) != NULL) {
1203		/* unicast/anycast, fine */
1204		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1205		    (icmp6_nodeinfo & 4) == 0) {
1206			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1207				"a temporary address in %s:%d",
1208			       __FILE__, __LINE__));
1209			goto bad;
1210		}
1211	} else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1212		; /* link-local multicast, fine */
1213	else
1214		goto bad;
1215
1216	/* validate query Subject field. */
1217	qtype = ntohs(ni6->ni_qtype);
1218	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1219	switch (qtype) {
1220	case NI_QTYPE_NOOP:
1221	case NI_QTYPE_SUPTYPES:
1222		/* 07 draft */
1223		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1224			break;
1225		/* FALLTHROUGH */
1226	case NI_QTYPE_FQDN:
1227	case NI_QTYPE_NODEADDR:
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(sin6.sin6_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)&sin6.sin6_addr);
1267			sin6.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1268							      &sin6.sin6_addr);
1269#ifndef SCOPEDROUTING
1270			in6_embedscope(&sin6.sin6_addr, &sin6, NULL, NULL);
1271#endif
1272			bzero(&sin6_d, sizeof(sin6_d));
1273			sin6_d.sin6_family = AF_INET6; /* not used, actually */
1274			sin6_d.sin6_len = sizeof(sin6_d); /* ditto */
1275			sin6_d.sin6_addr = ip6->ip6_dst;
1276			sin6_d.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1277								&ip6->ip6_dst);
1278#ifndef SCOPEDROUTING
1279			in6_embedscope(&sin6_d.sin6_addr, &sin6_d, NULL, NULL);
1280#endif
1281			subj = (char *)&sin6;
1282			if (SA6_ARE_ADDR_EQUAL(&sin6, &sin6_d))
1283				break;
1284
1285			/*
1286			 * XXX if we are to allow other cases, we should really
1287			 * be careful about scope here.
1288			 * basically, we should disallow queries toward IPv6
1289			 * destination X with subject Y, if scope(X) > scope(Y).
1290			 * if we allow scope(X) > scope(Y), it will result in
1291			 * information leakage across scope boundary.
1292			 */
1293			goto bad;
1294
1295		case ICMP6_NI_SUBJ_FQDN:
1296			/*
1297			 * Validate Subject name with gethostname(3).
1298			 *
1299			 * The behavior may need some debate, since:
1300			 * - we are not sure if the node has FQDN as
1301			 *   hostname (returned by gethostname(3)).
1302			 * - the code does wildcard match for truncated names.
1303			 *   however, we are not sure if we want to perform
1304			 *   wildcard match, if gethostname(3) side has
1305			 *   truncated hostname.
1306			 */
1307			n = ni6_nametodns(hostname, hostnamelen, 0);
1308			if (!n || n->m_next || n->m_len == 0)
1309				goto bad;
1310			IP6_EXTHDR_GET(subj, char *, m,
1311			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1312			if (subj == NULL)
1313				goto bad;
1314			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1315					n->m_len)) {
1316				goto bad;
1317			}
1318			m_freem(n);
1319			n = NULL;
1320			break;
1321
1322		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1323		default:
1324			goto bad;
1325		}
1326		break;
1327	}
1328
1329	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1330	switch (qtype) {
1331	case NI_QTYPE_FQDN:
1332		if ((icmp6_nodeinfo & 1) == 0)
1333			goto bad;
1334		break;
1335	case NI_QTYPE_NODEADDR:
1336		if ((icmp6_nodeinfo & 2) == 0)
1337			goto bad;
1338		break;
1339	}
1340
1341	/* guess reply length */
1342	switch (qtype) {
1343	case NI_QTYPE_NOOP:
1344		break;		/* no reply data */
1345	case NI_QTYPE_SUPTYPES:
1346		replylen += sizeof(u_int32_t);
1347		break;
1348	case NI_QTYPE_FQDN:
1349		/* XXX will append an mbuf */
1350		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1351		break;
1352	case NI_QTYPE_NODEADDR:
1353		addrs = ni6_addrs(ni6, m, &ifp, subj);
1354		if ((replylen += addrs * (sizeof(struct in6_addr) +
1355					  sizeof(u_int32_t))) > MCLBYTES)
1356			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1357		break;
1358	default:
1359		/*
1360		 * XXX: We must return a reply with the ICMP6 code
1361		 * `unknown Qtype' in this case. However we regard the case
1362		 * as an FQDN query for backward compatibility.
1363		 * Older versions set a random value to this field,
1364		 * so it rarely varies in the defined qtypes.
1365		 * But the mechanism is not reliable...
1366		 * maybe we should obsolete older versions.
1367		 */
1368		qtype = NI_QTYPE_FQDN;
1369		/* XXX will append an mbuf */
1370		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1371		oldfqdn++;
1372		break;
1373	}
1374
1375	/* allocate an mbuf to reply. */
1376	MGETHDR(n, M_DONTWAIT, m->m_type);
1377	if (n == NULL) {
1378		m_freem(m);
1379		return(NULL);
1380	}
1381	M_COPY_PKTHDR(n, m); /* just for recvif */
1382	if (replylen > MHLEN) {
1383		if (replylen > MCLBYTES) {
1384			/*
1385			 * XXX: should we try to allocate more? But MCLBYTES
1386			 * is probably much larger than IPV6_MMTU...
1387			 */
1388			goto bad;
1389		}
1390		MCLGET(n, M_DONTWAIT);
1391		if ((n->m_flags & M_EXT) == 0) {
1392			goto bad;
1393		}
1394	}
1395	n->m_pkthdr.len = n->m_len = replylen;
1396
1397	/* copy mbuf header and IPv6 + Node Information base headers */
1398	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1399	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1400	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1401
1402	/* qtype dependent procedure */
1403	switch (qtype) {
1404	case NI_QTYPE_NOOP:
1405		nni6->ni_code = ICMP6_NI_SUCCESS;
1406		nni6->ni_flags = 0;
1407		break;
1408	case NI_QTYPE_SUPTYPES:
1409	{
1410		u_int32_t v;
1411		nni6->ni_code = ICMP6_NI_SUCCESS;
1412		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1413		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1414		v = (u_int32_t)htonl(0x0000000f);
1415		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1416		break;
1417	}
1418	case NI_QTYPE_FQDN:
1419		nni6->ni_code = ICMP6_NI_SUCCESS;
1420		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1421						sizeof(struct ip6_hdr) +
1422						sizeof(struct icmp6_nodeinfo));
1423		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1424		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1425		/*
1426		 * XXX do we really have FQDN in variable "hostname"?
1427		 */
1428		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1429		if (n->m_next == NULL)
1430			goto bad;
1431		/* XXX we assume that n->m_next is not a chain */
1432		if (n->m_next->m_next != NULL)
1433			goto bad;
1434		n->m_pkthdr.len += n->m_next->m_len;
1435		break;
1436	case NI_QTYPE_NODEADDR:
1437	{
1438		int lenlim, copied;
1439
1440		nni6->ni_code = ICMP6_NI_SUCCESS;
1441		n->m_pkthdr.len = n->m_len =
1442		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1443		lenlim = M_TRAILINGSPACE(n);
1444		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1445		/* XXX: reset mbuf length */
1446		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1447			sizeof(struct icmp6_nodeinfo) + copied;
1448		break;
1449	}
1450	default:
1451		break;		/* XXX impossible! */
1452	}
1453
1454	nni6->ni_type = ICMP6_NI_REPLY;
1455	m_freem(m);
1456	return(n);
1457
1458  bad:
1459	m_freem(m);
1460	if (n)
1461		m_freem(n);
1462	return(NULL);
1463}
1464#undef hostnamelen
1465
1466/*
1467 * make a mbuf with DNS-encoded string.  no compression support.
1468 *
1469 * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1470 * treated as truncated name (two \0 at the end).  this is a wild guess.
1471 */
1472static struct mbuf *
1473ni6_nametodns(name, namelen, old)
1474	const char *name;
1475	int namelen;
1476	int old;	/* return pascal string if non-zero */
1477{
1478	struct mbuf *m;
1479	char *cp, *ep;
1480	const char *p, *q;
1481	int i, len, nterm;
1482
1483	if (old)
1484		len = namelen + 1;
1485	else
1486		len = MCLBYTES;
1487
1488	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1489	MGET(m, M_DONTWAIT, MT_DATA);
1490	if (m && len > MLEN) {
1491		MCLGET(m, M_DONTWAIT);
1492		if ((m->m_flags & M_EXT) == 0)
1493			goto fail;
1494	}
1495	if (!m)
1496		goto fail;
1497	m->m_next = NULL;
1498
1499	if (old) {
1500		m->m_len = len;
1501		*mtod(m, char *) = namelen;
1502		bcopy(name, mtod(m, char *) + 1, namelen);
1503		return m;
1504	} else {
1505		m->m_len = 0;
1506		cp = mtod(m, char *);
1507		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1508
1509		/* if not certain about my name, return empty buffer */
1510		if (namelen == 0)
1511			return m;
1512
1513		/*
1514		 * guess if it looks like shortened hostname, or FQDN.
1515		 * shortened hostname needs two trailing "\0".
1516		 */
1517		i = 0;
1518		for (p = name; p < name + namelen; p++) {
1519			if (*p && *p == '.')
1520				i++;
1521		}
1522		if (i < 2)
1523			nterm = 2;
1524		else
1525			nterm = 1;
1526
1527		p = name;
1528		while (cp < ep && p < name + namelen) {
1529			i = 0;
1530			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1531				i++;
1532			/* result does not fit into mbuf */
1533			if (cp + i + 1 >= ep)
1534				goto fail;
1535			/*
1536			 * DNS label length restriction, RFC1035 page 8.
1537			 * "i == 0" case is included here to avoid returning
1538			 * 0-length label on "foo..bar".
1539			 */
1540			if (i <= 0 || i >= 64)
1541				goto fail;
1542			*cp++ = i;
1543			bcopy(p, cp, i);
1544			cp += i;
1545			p = q;
1546			if (p < name + namelen && *p == '.')
1547				p++;
1548		}
1549		/* termination */
1550		if (cp + nterm >= ep)
1551			goto fail;
1552		while (nterm-- > 0)
1553			*cp++ = '\0';
1554		m->m_len = cp - mtod(m, char *);
1555		return m;
1556	}
1557
1558	panic("should not reach here");
1559	/*NOTREACHED*/
1560
1561 fail:
1562	if (m)
1563		m_freem(m);
1564	return NULL;
1565}
1566
1567/*
1568 * check if two DNS-encoded string matches.  takes care of truncated
1569 * form (with \0\0 at the end).  no compression support.
1570 * XXX upper/lowercase match (see RFC2065)
1571 */
1572static int
1573ni6_dnsmatch(a, alen, b, blen)
1574	const char *a;
1575	int alen;
1576	const char *b;
1577	int blen;
1578{
1579	const char *a0, *b0;
1580	int l;
1581
1582	/* simplest case - need validation? */
1583	if (alen == blen && bcmp(a, b, alen) == 0)
1584		return 1;
1585
1586	a0 = a;
1587	b0 = b;
1588
1589	/* termination is mandatory */
1590	if (alen < 2 || blen < 2)
1591		return 0;
1592	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1593		return 0;
1594	alen--;
1595	blen--;
1596
1597	while (a - a0 < alen && b - b0 < blen) {
1598		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1599			return 0;
1600
1601		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1602			return 0;
1603		/* we don't support compression yet */
1604		if (a[0] >= 64 || b[0] >= 64)
1605			return 0;
1606
1607		/* truncated case */
1608		if (a[0] == 0 && a - a0 == alen - 1)
1609			return 1;
1610		if (b[0] == 0 && b - b0 == blen - 1)
1611			return 1;
1612		if (a[0] == 0 || b[0] == 0)
1613			return 0;
1614
1615		if (a[0] != b[0])
1616			return 0;
1617		l = a[0];
1618		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1619			return 0;
1620		if (bcmp(a + 1, b + 1, l) != 0)
1621			return 0;
1622
1623		a += 1 + l;
1624		b += 1 + l;
1625	}
1626
1627	if (a - a0 == alen && b - b0 == blen)
1628		return 1;
1629	else
1630		return 0;
1631}
1632
1633/*
1634 * calculate the number of addresses to be returned in the node info reply.
1635 */
1636static int
1637ni6_addrs(ni6, m, ifpp, subj)
1638	struct icmp6_nodeinfo *ni6;
1639	struct mbuf *m;
1640	struct ifnet **ifpp;
1641	char *subj;
1642{
1643	struct ifnet *ifp;
1644	struct in6_ifaddr *ifa6;
1645	struct ifaddr *ifa;
1646	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1647	int addrs = 0, addrsofif, iffound = 0;
1648	int niflags = ni6->ni_flags;
1649
1650	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1651		switch (ni6->ni_code) {
1652		case ICMP6_NI_SUBJ_IPV6:
1653			if (subj == NULL) /* must be impossible... */
1654				return(0);
1655			subj_ip6 = (struct sockaddr_in6 *)subj;
1656			break;
1657		default:
1658			/*
1659			 * XXX: we only support IPv6 subject address for
1660			 * this Qtype.
1661			 */
1662			return(0);
1663		}
1664	}
1665
1666	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
1667	{
1668		addrsofif = 0;
1669		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list)
1670		{
1671			if (ifa->ifa_addr->sa_family != AF_INET6)
1672				continue;
1673			ifa6 = (struct in6_ifaddr *)ifa;
1674
1675			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1676			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1677					       &ifa6->ia_addr.sin6_addr))
1678				iffound = 1;
1679
1680			/*
1681			 * IPv4-mapped addresses can only be returned by a
1682			 * Node Information proxy, since they represent
1683			 * addresses of IPv4-only nodes, which perforce do
1684			 * not implement this protocol.
1685			 * [icmp-name-lookups-07, Section 5.4]
1686			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1687			 * this function at this moment.
1688			 */
1689
1690			/* What do we have to do about ::1? */
1691			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1692			case IPV6_ADDR_SCOPE_LINKLOCAL:
1693				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1694					continue;
1695				break;
1696			case IPV6_ADDR_SCOPE_SITELOCAL:
1697				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1698					continue;
1699				break;
1700			case IPV6_ADDR_SCOPE_GLOBAL:
1701				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1702					continue;
1703				break;
1704			default:
1705				continue;
1706			}
1707
1708			/*
1709			 * check if anycast is okay.
1710			 * XXX: just experimental. not in the spec.
1711			 */
1712			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1713			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1714				continue; /* we need only unicast addresses */
1715			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1716			    (icmp6_nodeinfo & 4) == 0) {
1717				continue;
1718			}
1719			addrsofif++; /* count the address */
1720		}
1721		if (iffound) {
1722			*ifpp = ifp;
1723			return(addrsofif);
1724		}
1725
1726		addrs += addrsofif;
1727	}
1728
1729	return(addrs);
1730}
1731
1732static int
1733ni6_store_addrs(ni6, nni6, ifp0, resid)
1734	struct icmp6_nodeinfo *ni6, *nni6;
1735	struct ifnet *ifp0;
1736	int resid;
1737{
1738	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1739	struct in6_ifaddr *ifa6;
1740	struct ifaddr *ifa;
1741	struct ifnet *ifp_dep = NULL;
1742	int copied = 0, allow_deprecated = 0;
1743	u_char *cp = (u_char *)(nni6 + 1);
1744	int niflags = ni6->ni_flags;
1745	u_int32_t ltime;
1746
1747	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1748		return(0);	/* needless to copy */
1749
1750  again:
1751
1752	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list))
1753	{
1754		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1755		     ifa = ifa->ifa_list.tqe_next)
1756		{
1757			if (ifa->ifa_addr->sa_family != AF_INET6)
1758				continue;
1759			ifa6 = (struct in6_ifaddr *)ifa;
1760
1761			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1762			    allow_deprecated == 0) {
1763				/*
1764				 * prefererred address should be put before
1765				 * deprecated addresses.
1766				 */
1767
1768				/* record the interface for later search */
1769				if (ifp_dep == NULL)
1770					ifp_dep = ifp;
1771
1772				continue;
1773			}
1774			else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1775				 allow_deprecated != 0)
1776				continue; /* we now collect deprecated addrs */
1777
1778			/* What do we have to do about ::1? */
1779			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1780			case IPV6_ADDR_SCOPE_LINKLOCAL:
1781				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1782					continue;
1783				break;
1784			case IPV6_ADDR_SCOPE_SITELOCAL:
1785				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1786					continue;
1787				break;
1788			case IPV6_ADDR_SCOPE_GLOBAL:
1789				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1790					continue;
1791				break;
1792			default:
1793				continue;
1794			}
1795
1796			/*
1797			 * check if anycast is okay.
1798			 * XXX: just experimental. not in the spec.
1799			 */
1800			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1801			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1802				continue;
1803			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1804			    (icmp6_nodeinfo & 4) == 0) {
1805				continue;
1806			}
1807
1808			/* now we can copy the address */
1809			if (resid < sizeof(struct in6_addr) +
1810			    sizeof(u_int32_t)) {
1811				/*
1812				 * We give up much more copy.
1813				 * Set the truncate flag and return.
1814				 */
1815				nni6->ni_flags |=
1816					NI_NODEADDR_FLAG_TRUNCATE;
1817				return(copied);
1818			}
1819
1820			/*
1821			 * Set the TTL of the address.
1822			 * The TTL value should be one of the following
1823			 * according to the specification:
1824			 *
1825			 * 1. The remaining lifetime of a DHCP lease on the
1826			 *    address, or
1827			 * 2. The remaining Valid Lifetime of a prefix from
1828			 *    which the address was derived through Stateless
1829			 *    Autoconfiguration.
1830			 *
1831			 * Note that we currently do not support stateful
1832			 * address configuration by DHCPv6, so the former
1833			 * case can't happen.
1834			 */
1835			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1836				ltime = ND6_INFINITE_LIFETIME;
1837			else {
1838				if (ifa6->ia6_lifetime.ia6t_expire >
1839				    time_second)
1840					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1841				else
1842					ltime = 0;
1843			}
1844
1845			bcopy(&ltime, cp, sizeof(u_int32_t));
1846			cp += sizeof(u_int32_t);
1847
1848			/* copy the address itself */
1849			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1850			      sizeof(struct in6_addr));
1851			/* XXX: KAME link-local hack; remove ifindex */
1852			if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr))
1853				((struct in6_addr *)cp)->s6_addr16[1] = 0;
1854			cp += sizeof(struct in6_addr);
1855
1856			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1857			copied += (sizeof(struct in6_addr) +
1858				   sizeof(u_int32_t));
1859		}
1860		if (ifp0)	/* we need search only on the specified IF */
1861			break;
1862	}
1863
1864	if (allow_deprecated == 0 && ifp_dep != NULL) {
1865		ifp = ifp_dep;
1866		allow_deprecated = 1;
1867
1868		goto again;
1869	}
1870
1871	return(copied);
1872}
1873
1874/*
1875 * XXX almost dup'ed code with rip6_input.
1876 */
1877static int
1878icmp6_rip6_input(mp, off)
1879	struct	mbuf **mp;
1880	int	off;
1881{
1882	struct mbuf *m = *mp;
1883	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1884	struct in6pcb *in6p;
1885	struct in6pcb *last = NULL;
1886	struct sockaddr_in6 rip6src;
1887	struct icmp6_hdr *icmp6;
1888	struct mbuf *opts = NULL;
1889
1890#ifndef PULLDOWN_TEST
1891	/* this is assumed to be safe. */
1892	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1893#else
1894	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1895	if (icmp6 == NULL) {
1896		/* m is already reclaimed */
1897		return IPPROTO_DONE;
1898	}
1899#endif
1900
1901	bzero(&rip6src, sizeof(rip6src));
1902	rip6src.sin6_len = sizeof(struct sockaddr_in6);
1903	rip6src.sin6_family = AF_INET6;
1904	/* KAME hack: recover scopeid */
1905	(void)in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif);
1906
1907	LIST_FOREACH(in6p, &ripcb, inp_list)
1908	{
1909		if ((in6p->inp_vflag & INP_IPV6) == 0)
1910			continue;
1911#ifdef HAVE_NRL_INPCB
1912		if (!(in6p->in6p_flags & INP_IPV6))
1913			continue;
1914#endif
1915		if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1916			continue;
1917		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1918		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1919			continue;
1920		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1921		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1922			continue;
1923		if (in6p->in6p_icmp6filt
1924		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1925				 in6p->in6p_icmp6filt))
1926			continue;
1927		if (last) {
1928			struct	mbuf *n;
1929			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1930				if (last->in6p_flags & IN6P_CONTROLOPTS)
1931					ip6_savecontrol(last, &opts, ip6, n);
1932				/* strip intermediate headers */
1933				m_adj(n, off);
1934				if (sbappendaddr(&last->in6p_socket->so_rcv,
1935						 (struct sockaddr *)&rip6src,
1936						 n, opts) == 0) {
1937					/* should notify about lost packet */
1938					m_freem(n);
1939					if (opts) {
1940						m_freem(opts);
1941					}
1942				} else
1943					sorwakeup(last->in6p_socket);
1944				opts = NULL;
1945			}
1946		}
1947		last = in6p;
1948	}
1949	if (last) {
1950		if (last->in6p_flags & IN6P_CONTROLOPTS)
1951			ip6_savecontrol(last, &opts, ip6, m);
1952		/* strip intermediate headers */
1953		m_adj(m, off);
1954		if (sbappendaddr(&last->in6p_socket->so_rcv,
1955				 (struct sockaddr *)&rip6src, m, opts) == 0) {
1956			m_freem(m);
1957			if (opts)
1958				m_freem(opts);
1959		} else
1960			sorwakeup(last->in6p_socket);
1961	} else {
1962		m_freem(m);
1963		ip6stat.ip6s_delivered--;
1964	}
1965	return IPPROTO_DONE;
1966}
1967
1968/*
1969 * Reflect the ip6 packet back to the source.
1970 * OFF points to the icmp6 header, counted from the top of the mbuf.
1971 */
1972void
1973icmp6_reflect(m, off)
1974	struct	mbuf *m;
1975	size_t off;
1976{
1977	struct ip6_hdr *ip6;
1978	struct icmp6_hdr *icmp6;
1979	struct in6_ifaddr *ia;
1980	struct in6_addr t, *src = 0;
1981	int plen;
1982	int type, code;
1983	struct ifnet *outif = NULL;
1984	struct sockaddr_in6 sa6_src, sa6_dst;
1985#ifdef COMPAT_RFC1885
1986	int mtu = IPV6_MMTU;
1987	struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
1988#endif
1989
1990	/* too short to reflect */
1991	if (off < sizeof(struct ip6_hdr)) {
1992		nd6log((LOG_DEBUG,
1993		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
1994		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
1995		    __FILE__, __LINE__));
1996		goto bad;
1997	}
1998
1999	/*
2000	 * If there are extra headers between IPv6 and ICMPv6, strip
2001	 * off that header first.
2002	 */
2003#ifdef DIAGNOSTIC
2004	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2005		panic("assumption failed in icmp6_reflect");
2006#endif
2007	if (off > sizeof(struct ip6_hdr)) {
2008		size_t l;
2009		struct ip6_hdr nip6;
2010
2011		l = off - sizeof(struct ip6_hdr);
2012		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2013		m_adj(m, l);
2014		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2015		if (m->m_len < l) {
2016			if ((m = m_pullup(m, l)) == NULL)
2017				return;
2018		}
2019		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2020	} else /* off == sizeof(struct ip6_hdr) */ {
2021		size_t l;
2022		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2023		if (m->m_len < l) {
2024			if ((m = m_pullup(m, l)) == NULL)
2025				return;
2026		}
2027	}
2028	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2029	ip6 = mtod(m, struct ip6_hdr *);
2030	ip6->ip6_nxt = IPPROTO_ICMPV6;
2031	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2032	type = icmp6->icmp6_type; /* keep type for statistics */
2033	code = icmp6->icmp6_code; /* ditto. */
2034
2035	t = ip6->ip6_dst;
2036	/*
2037	 * ip6_input() drops a packet if its src is multicast.
2038	 * So, the src is never multicast.
2039	 */
2040	ip6->ip6_dst = ip6->ip6_src;
2041
2042	/*
2043	 * XXX: make sure to embed scope zone information, using
2044	 * already embedded IDs or the received interface (if any).
2045	 * Note that rcvif may be NULL.
2046	 * TODO: scoped routing case (XXX).
2047	 */
2048	bzero(&sa6_src, sizeof(sa6_src));
2049	sa6_src.sin6_family = AF_INET6;
2050	sa6_src.sin6_len = sizeof(sa6_src);
2051	sa6_src.sin6_addr = ip6->ip6_dst;
2052	in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
2053	in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL);
2054	bzero(&sa6_dst, sizeof(sa6_dst));
2055	sa6_dst.sin6_family = AF_INET6;
2056	sa6_dst.sin6_len = sizeof(sa6_dst);
2057	sa6_dst.sin6_addr = t;
2058	in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
2059	in6_embedscope(&t, &sa6_dst, NULL, NULL);
2060
2061#ifdef COMPAT_RFC1885
2062	/*
2063	 * xxx guess MTU
2064	 * RFC 1885 requires that echo reply should be truncated if it
2065	 * does not fit in with (return) path MTU, but the description was
2066	 * removed in the new spec.
2067	 */
2068	if (icmp6_reflect_rt.ro_rt == 0 ||
2069	    ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
2070		if (icmp6_reflect_rt.ro_rt) {
2071			RTFREE(icmp6_reflect_rt.ro_rt);
2072			icmp6_reflect_rt.ro_rt = 0;
2073		}
2074		bzero(sin6, sizeof(*sin6));
2075		sin6->sin6_family = PF_INET6;
2076		sin6->sin6_len = sizeof(struct sockaddr_in6);
2077		sin6->sin6_addr = ip6->ip6_dst;
2078
2079		rtalloc_ign((struct route *)&icmp6_reflect_rt.ro_rt,
2080			    RTF_PRCLONING);
2081	}
2082
2083	if (icmp6_reflect_rt.ro_rt == 0)
2084		goto bad;
2085
2086	if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
2087	    && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
2088		mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
2089
2090	if (mtu < m->m_pkthdr.len) {
2091		plen -= (m->m_pkthdr.len - mtu);
2092		m_adj(m, mtu - m->m_pkthdr.len);
2093	}
2094#endif
2095	/*
2096	 * If the incoming packet was addressed directly to us(i.e. unicast),
2097	 * use dst as the src for the reply.
2098	 * The IN6_IFF_NOTREADY case would be VERY rare, but is possible
2099	 * (for example) when we encounter an error while forwarding procedure
2100	 * destined to a duplicated address of ours.
2101	 */
2102	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2103		if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2104		    (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2105			src = &t;
2106			break;
2107		}
2108	if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2109		/*
2110		 * This is the case if the dst is our link-local address
2111		 * and the sender is also ourseleves.
2112		 */
2113		src = &t;
2114	}
2115
2116	if (src == 0) {
2117		int e;
2118		struct route_in6 ro;
2119
2120		/*
2121		 * This case matches to multicasts, our anycast, or unicasts
2122		 * that we do not own. Select a source address based on the
2123		 * source address of the erroneous packet.
2124		 */
2125		bzero(&ro, sizeof(ro));
2126		src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e);
2127		if (ro.ro_rt)
2128			RTFREE(ro.ro_rt); /* XXX: we could use this */
2129		if (src == NULL) {
2130			nd6log((LOG_DEBUG,
2131			    "icmp6_reflect: source can't be determined: "
2132			    "dst=%s, error=%d\n",
2133			    ip6_sprintf(&sa6_src.sin6_addr), e));
2134			goto bad;
2135		}
2136	}
2137
2138	ip6->ip6_src = *src;
2139
2140	ip6->ip6_flow = 0;
2141	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2142	ip6->ip6_vfc |= IPV6_VERSION;
2143	ip6->ip6_nxt = IPPROTO_ICMPV6;
2144	if (m->m_pkthdr.rcvif) {
2145		/* XXX: This may not be the outgoing interface */
2146		ip6->ip6_hlim = nd_ifinfo[m->m_pkthdr.rcvif->if_index].chlim;
2147	} else
2148		ip6->ip6_hlim = ip6_defhlim;
2149
2150	icmp6->icmp6_cksum = 0;
2151	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2152					sizeof(struct ip6_hdr), plen);
2153
2154	/*
2155	 * XXX option handling
2156	 */
2157
2158	m->m_flags &= ~(M_BCAST|M_MCAST);
2159#ifdef IPSEC
2160	/* Don't lookup socket */
2161	(void)ipsec_setsocket(m, NULL);
2162#endif /*IPSEC*/
2163
2164#ifdef COMPAT_RFC1885
2165	ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif);
2166#else
2167	ip6_output(m, NULL, NULL, 0, NULL, &outif);
2168#endif
2169	if (outif)
2170		icmp6_ifoutstat_inc(outif, type, code);
2171
2172	return;
2173
2174 bad:
2175	m_freem(m);
2176	return;
2177}
2178
2179void
2180icmp6_fasttimo()
2181{
2182
2183	mld6_fasttimeo();
2184}
2185
2186static const char *
2187icmp6_redirect_diag(src6, dst6, tgt6)
2188	struct in6_addr *src6;
2189	struct in6_addr *dst6;
2190	struct in6_addr *tgt6;
2191{
2192	static char buf[1024];
2193	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2194		ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2195	return buf;
2196}
2197
2198void
2199icmp6_redirect_input(m, off)
2200	struct mbuf *m;
2201	int off;
2202{
2203	struct ifnet *ifp = m->m_pkthdr.rcvif;
2204	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2205	struct nd_redirect *nd_rd;
2206	int icmp6len = ntohs(ip6->ip6_plen);
2207	char *lladdr = NULL;
2208	int lladdrlen = 0;
2209	u_char *redirhdr = NULL;
2210	int redirhdrlen = 0;
2211	struct rtentry *rt = NULL;
2212	int is_router;
2213	int is_onlink;
2214	struct in6_addr src6 = ip6->ip6_src;
2215	struct in6_addr redtgt6;
2216	struct in6_addr reddst6;
2217	union nd_opts ndopts;
2218
2219	if (!m || !ifp)
2220		return;
2221
2222	/* XXX if we are router, we don't update route by icmp6 redirect */
2223	if (ip6_forwarding)
2224		goto freeit;
2225	if (!icmp6_rediraccept)
2226		goto freeit;
2227
2228#ifndef PULLDOWN_TEST
2229	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2230	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2231#else
2232	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2233	if (nd_rd == NULL) {
2234		icmp6stat.icp6s_tooshort++;
2235		return;
2236	}
2237#endif
2238	redtgt6 = nd_rd->nd_rd_target;
2239	reddst6 = nd_rd->nd_rd_dst;
2240
2241	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2242		redtgt6.s6_addr16[1] = htons(ifp->if_index);
2243	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2244		reddst6.s6_addr16[1] = htons(ifp->if_index);
2245
2246	/* validation */
2247	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2248		nd6log((LOG_ERR,
2249			"ICMP6 redirect sent from %s rejected; "
2250			"must be from linklocal\n", ip6_sprintf(&src6)));
2251		goto bad;
2252	}
2253	if (ip6->ip6_hlim != 255) {
2254		nd6log((LOG_ERR,
2255			"ICMP6 redirect sent from %s rejected; "
2256			"hlim=%d (must be 255)\n",
2257			ip6_sprintf(&src6), ip6->ip6_hlim));
2258		goto bad;
2259	}
2260    {
2261	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2262	struct sockaddr_in6 sin6;
2263	struct in6_addr *gw6;
2264
2265	bzero(&sin6, sizeof(sin6));
2266	sin6.sin6_family = AF_INET6;
2267	sin6.sin6_len = sizeof(struct sockaddr_in6);
2268	bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2269	rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL);
2270	if (rt) {
2271		if (rt->rt_gateway == NULL ||
2272		    rt->rt_gateway->sa_family != AF_INET6) {
2273			nd6log((LOG_ERR,
2274			    "ICMP6 redirect rejected; no route "
2275			    "with inet6 gateway found for redirect dst: %s\n",
2276			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2277			RTFREE(rt);
2278			goto bad;
2279		}
2280
2281		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2282		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2283			nd6log((LOG_ERR,
2284				"ICMP6 redirect rejected; "
2285				"not equal to gw-for-src=%s (must be same): "
2286				"%s\n",
2287				ip6_sprintf(gw6),
2288				icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2289			RTFREE(rt);
2290			goto bad;
2291		}
2292	} else {
2293		nd6log((LOG_ERR,
2294			"ICMP6 redirect rejected; "
2295			"no route found for redirect dst: %s\n",
2296			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2297		goto bad;
2298	}
2299	RTFREE(rt);
2300	rt = NULL;
2301    }
2302	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2303		nd6log((LOG_ERR,
2304			"ICMP6 redirect rejected; "
2305			"redirect dst must be unicast: %s\n",
2306			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2307		goto bad;
2308	}
2309
2310	is_router = is_onlink = 0;
2311	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2312		is_router = 1;	/* router case */
2313	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2314		is_onlink = 1;	/* on-link destination case */
2315	if (!is_router && !is_onlink) {
2316		nd6log((LOG_ERR,
2317			"ICMP6 redirect rejected; "
2318			"neither router case nor onlink case: %s\n",
2319			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2320		goto bad;
2321	}
2322	/* validation passed */
2323
2324	icmp6len -= sizeof(*nd_rd);
2325	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2326	if (nd6_options(&ndopts) < 0) {
2327		nd6log((LOG_INFO, "icmp6_redirect_input: "
2328			"invalid ND option, rejected: %s\n",
2329			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2330		/* nd6_options have incremented stats */
2331		goto freeit;
2332	}
2333
2334	if (ndopts.nd_opts_tgt_lladdr) {
2335		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2336		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2337	}
2338
2339	if (ndopts.nd_opts_rh) {
2340		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2341		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2342	}
2343
2344	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2345		nd6log((LOG_INFO,
2346			"icmp6_redirect_input: lladdrlen mismatch for %s "
2347			"(if %d, icmp6 packet %d): %s\n",
2348			ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2349			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2350		goto bad;
2351	}
2352
2353	/* RFC 2461 8.3 */
2354	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2355			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2356
2357	if (!is_onlink) {	/* better router case. perform rtredirect. */
2358		/* perform rtredirect */
2359		struct sockaddr_in6 sdst;
2360		struct sockaddr_in6 sgw;
2361		struct sockaddr_in6 ssrc;
2362
2363		bzero(&sdst, sizeof(sdst));
2364		bzero(&sgw, sizeof(sgw));
2365		bzero(&ssrc, sizeof(ssrc));
2366		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2367		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2368			sizeof(struct sockaddr_in6);
2369		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2370		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2371		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2372		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2373			   (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2374			   (struct sockaddr *)&ssrc,
2375			   (struct rtentry **)NULL);
2376	}
2377	/* finally update cached route in each socket via pfctlinput */
2378    {
2379	struct sockaddr_in6 sdst;
2380
2381	bzero(&sdst, sizeof(sdst));
2382	sdst.sin6_family = AF_INET6;
2383	sdst.sin6_len = sizeof(struct sockaddr_in6);
2384	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2385	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2386#ifdef IPSEC
2387	key_sa_routechange((struct sockaddr *)&sdst);
2388#endif
2389    }
2390
2391 freeit:
2392	m_freem(m);
2393	return;
2394
2395 bad:
2396	icmp6stat.icp6s_badredirect++;
2397	m_freem(m);
2398}
2399
2400void
2401icmp6_redirect_output(m0, rt)
2402	struct mbuf *m0;
2403	struct rtentry *rt;
2404{
2405	struct ifnet *ifp;	/* my outgoing interface */
2406	struct in6_addr *ifp_ll6;
2407	struct in6_addr *router_ll6;
2408	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2409	struct mbuf *m = NULL;	/* newly allocated one */
2410	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2411	struct nd_redirect *nd_rd;
2412	size_t maxlen;
2413	u_char *p;
2414	struct ifnet *outif = NULL;
2415	struct sockaddr_in6 src_sa;
2416
2417	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2418
2419	/* if we are not router, we don't send icmp6 redirect */
2420	if (!ip6_forwarding || ip6_accept_rtadv)
2421		goto fail;
2422
2423	/* sanity check */
2424	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2425		goto fail;
2426
2427	/*
2428	 * Address check:
2429	 *  the source address must identify a neighbor, and
2430	 *  the destination address must not be a multicast address
2431	 *  [RFC 2461, sec 8.2]
2432	 */
2433	sip6 = mtod(m0, struct ip6_hdr *);
2434	bzero(&src_sa, sizeof(src_sa));
2435	src_sa.sin6_family = AF_INET6;
2436	src_sa.sin6_len = sizeof(src_sa);
2437	src_sa.sin6_addr = sip6->ip6_src;
2438	/* we don't currently use sin6_scope_id, but eventually use it */
2439	src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src);
2440	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2441		goto fail;
2442	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2443		goto fail;	/* what should we do here? */
2444
2445	/* rate limit */
2446	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2447		goto fail;
2448
2449	/*
2450	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2451	 * we almost always ask for an mbuf cluster for simplicity.
2452	 * (MHLEN < IPV6_MMTU is almost always true)
2453	 */
2454#if IPV6_MMTU >= MCLBYTES
2455# error assumption failed about IPV6_MMTU and MCLBYTES
2456#endif
2457	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2458	if (m && IPV6_MMTU >= MHLEN)
2459		MCLGET(m, M_DONTWAIT);
2460	if (!m)
2461		goto fail;
2462	m->m_pkthdr.rcvif = NULL;
2463	m->m_len = 0;
2464	maxlen = M_TRAILINGSPACE(m);
2465	maxlen = min(IPV6_MMTU, maxlen);
2466	/* just for safety */
2467	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2468	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2469		goto fail;
2470	}
2471
2472	{
2473		/* get ip6 linklocal address for ifp(my outgoing interface). */
2474		struct in6_ifaddr *ia;
2475		if ((ia = in6ifa_ifpforlinklocal(ifp,
2476						 IN6_IFF_NOTREADY|
2477						 IN6_IFF_ANYCAST)) == NULL)
2478			goto fail;
2479		ifp_ll6 = &ia->ia_addr.sin6_addr;
2480	}
2481
2482	/* get ip6 linklocal address for the router. */
2483	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2484		struct sockaddr_in6 *sin6;
2485		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2486		router_ll6 = &sin6->sin6_addr;
2487		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2488			router_ll6 = (struct in6_addr *)NULL;
2489	} else
2490		router_ll6 = (struct in6_addr *)NULL;
2491
2492	/* ip6 */
2493	ip6 = mtod(m, struct ip6_hdr *);
2494	ip6->ip6_flow = 0;
2495	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2496	ip6->ip6_vfc |= IPV6_VERSION;
2497	/* ip6->ip6_plen will be set later */
2498	ip6->ip6_nxt = IPPROTO_ICMPV6;
2499	ip6->ip6_hlim = 255;
2500	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2501	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2502	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2503
2504	/* ND Redirect */
2505	nd_rd = (struct nd_redirect *)(ip6 + 1);
2506	nd_rd->nd_rd_type = ND_REDIRECT;
2507	nd_rd->nd_rd_code = 0;
2508	nd_rd->nd_rd_reserved = 0;
2509	if (rt->rt_flags & RTF_GATEWAY) {
2510		/*
2511		 * nd_rd->nd_rd_target must be a link-local address in
2512		 * better router cases.
2513		 */
2514		if (!router_ll6)
2515			goto fail;
2516		bcopy(router_ll6, &nd_rd->nd_rd_target,
2517		      sizeof(nd_rd->nd_rd_target));
2518		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2519		      sizeof(nd_rd->nd_rd_dst));
2520	} else {
2521		/* make sure redtgt == reddst */
2522		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2523		      sizeof(nd_rd->nd_rd_target));
2524		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2525		      sizeof(nd_rd->nd_rd_dst));
2526	}
2527
2528	p = (u_char *)(nd_rd + 1);
2529
2530	if (!router_ll6)
2531		goto nolladdropt;
2532
2533    {
2534	/* target lladdr option */
2535	struct rtentry *rt_router = NULL;
2536	int len;
2537	struct sockaddr_dl *sdl;
2538	struct nd_opt_hdr *nd_opt;
2539	char *lladdr;
2540
2541	rt_router = nd6_lookup(router_ll6, 0, ifp);
2542	if (!rt_router)
2543		goto nolladdropt;
2544	len = sizeof(*nd_opt) + ifp->if_addrlen;
2545	len = (len + 7) & ~7;	/*round by 8*/
2546	/* safety check */
2547	if (len + (p - (u_char *)ip6) > maxlen)
2548		goto nolladdropt;
2549	if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2550	    (rt_router->rt_flags & RTF_LLINFO) &&
2551	    (rt_router->rt_gateway->sa_family == AF_LINK) &&
2552	    (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2553	    sdl->sdl_alen) {
2554		nd_opt = (struct nd_opt_hdr *)p;
2555		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2556		nd_opt->nd_opt_len = len >> 3;
2557		lladdr = (char *)(nd_opt + 1);
2558		bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2559		p += len;
2560	}
2561    }
2562nolladdropt:;
2563
2564	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2565
2566	/* just to be safe */
2567#ifdef M_DECRYPTED	/*not openbsd*/
2568	if (m0->m_flags & M_DECRYPTED)
2569		goto noredhdropt;
2570#endif
2571	if (p - (u_char *)ip6 > maxlen)
2572		goto noredhdropt;
2573
2574    {
2575	/* redirected header option */
2576	int len;
2577	struct nd_opt_rd_hdr *nd_opt_rh;
2578
2579	/*
2580	 * compute the maximum size for icmp6 redirect header option.
2581	 * XXX room for auth header?
2582	 */
2583	len = maxlen - (p - (u_char *)ip6);
2584	len &= ~7;
2585
2586	/* This is just for simplicity. */
2587	if (m0->m_pkthdr.len != m0->m_len) {
2588		if (m0->m_next) {
2589			m_freem(m0->m_next);
2590			m0->m_next = NULL;
2591		}
2592		m0->m_pkthdr.len = m0->m_len;
2593	}
2594
2595	/*
2596	 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2597	 * about padding/truncate rule for the original IP packet.
2598	 * From the discussion on IPv6imp in Feb 1999, the consensus was:
2599	 * - "attach as much as possible" is the goal
2600	 * - pad if not aligned (original size can be guessed by original
2601	 *   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\n", __FILE__, __LINE__);
2607
2608	if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2609		/* not enough room, truncate */
2610		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2611	} else {
2612		/* enough room, pad or truncate */
2613		size_t extra;
2614
2615		extra = m0->m_pkthdr.len % 8;
2616		if (extra) {
2617			/* pad if easy enough, truncate if not */
2618			if (8 - extra <= M_TRAILINGSPACE(m0)) {
2619				/* pad */
2620				m0->m_len += (8 - extra);
2621				m0->m_pkthdr.len += (8 - extra);
2622			} else {
2623				/* truncate */
2624				m0->m_pkthdr.len -= extra;
2625				m0->m_len -= extra;
2626			}
2627		}
2628		len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2629		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2630	}
2631
2632	nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2633	bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2634	nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2635	nd_opt_rh->nd_opt_rh_len = len >> 3;
2636	p += sizeof(*nd_opt_rh);
2637	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2638
2639	/* connect m0 to m */
2640	m->m_next = m0;
2641	m->m_pkthdr.len = m->m_len + m0->m_len;
2642    }
2643noredhdropt:;
2644
2645	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
2646		sip6->ip6_src.s6_addr16[1] = 0;
2647	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
2648		sip6->ip6_dst.s6_addr16[1] = 0;
2649#if 0
2650	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
2651		ip6->ip6_src.s6_addr16[1] = 0;
2652	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
2653		ip6->ip6_dst.s6_addr16[1] = 0;
2654#endif
2655	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
2656		nd_rd->nd_rd_target.s6_addr16[1] = 0;
2657	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
2658		nd_rd->nd_rd_dst.s6_addr16[1] = 0;
2659
2660	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2661
2662	nd_rd->nd_rd_cksum = 0;
2663	nd_rd->nd_rd_cksum
2664		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2665
2666	/* send the packet to outside... */
2667#ifdef IPSEC
2668	/* Don't lookup socket */
2669	(void)ipsec_setsocket(m, NULL);
2670#endif /*IPSEC*/
2671	ip6_output(m, NULL, NULL, 0, NULL, &outif);
2672	if (outif) {
2673		icmp6_ifstat_inc(outif, ifs6_out_msg);
2674		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2675	}
2676	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2677
2678	return;
2679
2680fail:
2681	if (m)
2682		m_freem(m);
2683	if (m0)
2684		m_freem(m0);
2685}
2686
2687#ifdef HAVE_NRL_INPCB
2688#define sotoin6pcb	sotoinpcb
2689#define in6pcb		inpcb
2690#define in6p_icmp6filt	inp_icmp6filt
2691#endif
2692/*
2693 * ICMPv6 socket option processing.
2694 */
2695int
2696icmp6_ctloutput(so, sopt)
2697	struct socket *so;
2698	struct sockopt *sopt;
2699{
2700	int error = 0;
2701	int optlen;
2702	struct inpcb *inp = sotoinpcb(so);
2703	int level, op, optname;
2704
2705	if (sopt) {
2706		level = sopt->sopt_level;
2707		op = sopt->sopt_dir;
2708		optname = sopt->sopt_name;
2709		optlen = sopt->sopt_valsize;
2710	} else
2711		level = op = optname = optlen = 0;
2712
2713	if (level != IPPROTO_ICMPV6) {
2714		return EINVAL;
2715	}
2716
2717	switch (op) {
2718	case PRCO_SETOPT:
2719		switch (optname) {
2720		case ICMP6_FILTER:
2721		    {
2722			struct icmp6_filter *p;
2723
2724			if (optlen != sizeof(*p)) {
2725				error = EMSGSIZE;
2726				break;
2727			}
2728			if (inp->in6p_icmp6filt == NULL) {
2729				error = EINVAL;
2730				break;
2731			}
2732			error = sooptcopyin(sopt, inp->in6p_icmp6filt, optlen,
2733				optlen);
2734			break;
2735		    }
2736
2737		default:
2738			error = ENOPROTOOPT;
2739			break;
2740		}
2741		break;
2742
2743	case PRCO_GETOPT:
2744		switch (optname) {
2745		case ICMP6_FILTER:
2746		    {
2747			if (inp->in6p_icmp6filt == NULL) {
2748				error = EINVAL;
2749				break;
2750			}
2751			error = sooptcopyout(sopt, inp->in6p_icmp6filt,
2752				sizeof(struct icmp6_filter));
2753			break;
2754		    }
2755
2756		default:
2757			error = ENOPROTOOPT;
2758			break;
2759		}
2760		break;
2761	}
2762
2763	return(error);
2764}
2765#ifdef HAVE_NRL_INPCB
2766#undef sotoin6pcb
2767#undef in6pcb
2768#undef in6p_icmp6filt
2769#endif
2770
2771#ifndef HAVE_PPSRATECHECK
2772#ifndef timersub
2773#define	timersub(tvp, uvp, vvp)						\
2774	do {								\
2775		(vvp)->tv_sec = (tvp)->tv_sec - (uvp)->tv_sec;		\
2776		(vvp)->tv_usec = (tvp)->tv_usec - (uvp)->tv_usec;	\
2777		if ((vvp)->tv_usec < 0) {				\
2778			(vvp)->tv_sec--;				\
2779			(vvp)->tv_usec += 1000000;			\
2780		}							\
2781	} while (0)
2782#endif
2783
2784/*
2785 * ppsratecheck(): packets (or events) per second limitation.
2786 */
2787static int
2788ppsratecheck(lasttime, curpps, maxpps)
2789	struct timeval *lasttime;
2790	int *curpps;
2791	int maxpps;	/* maximum pps allowed */
2792{
2793	struct timeval tv, delta;
2794	int s, rv;
2795
2796	s = splclock();
2797	microtime(&tv);
2798	splx(s);
2799
2800	timersub(&tv, lasttime, &delta);
2801
2802	/*
2803	 * check for 0,0 is so that the message will be seen at least once.
2804	 * if more than one second have passed since the last update of
2805	 * lasttime, reset the counter.
2806	 *
2807	 * we do increment *curpps even in *curpps < maxpps case, as some may
2808	 * try to use *curpps for stat purposes as well.
2809	 */
2810	if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) ||
2811	    delta.tv_sec >= 1) {
2812		*lasttime = tv;
2813		*curpps = 0;
2814		rv = 1;
2815	} else if (maxpps < 0)
2816		rv = 1;
2817	else if (*curpps < maxpps)
2818		rv = 1;
2819	else
2820		rv = 0;
2821
2822#if 1 /*DIAGNOSTIC?*/
2823	/* be careful about wrap-around */
2824	if (*curpps + 1 > *curpps)
2825		*curpps = *curpps + 1;
2826#else
2827	/*
2828	 * assume that there's not too many calls to this function.
2829	 * not sure if the assumption holds, as it depends on *caller's*
2830	 * behavior, not the behavior of this function.
2831	 * IMHO it is wrong to make assumption on the caller's behavior,
2832	 * so the above #if is #if 1, not #ifdef DIAGNOSTIC.
2833	 */
2834	*curpps = *curpps + 1;
2835#endif
2836
2837	return (rv);
2838}
2839#endif
2840
2841/*
2842 * Perform rate limit check.
2843 * Returns 0 if it is okay to send the icmp6 packet.
2844 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2845 * limitation.
2846 *
2847 * XXX per-destination/type check necessary?
2848 */
2849static int
2850icmp6_ratelimit(dst, type, code)
2851	const struct in6_addr *dst;	/* not used at this moment */
2852	const int type;			/* not used at this moment */
2853	const int code;			/* not used at this moment */
2854{
2855	int ret;
2856
2857	ret = 0;	/*okay to send*/
2858
2859	/* PPS limit */
2860	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2861	    icmp6errppslim)) {
2862		/* The packet is subject to rate limit */
2863		ret++;
2864	}
2865
2866	return ret;
2867}
2868