icmp6.c revision 1.48
1/*	$NetBSD: icmp6.c,v 1.48 2000/12/09 01:29:50 itojun Exp $	*/
2/*	$KAME: icmp6.c,v 1.156 2000/10/19 19:21:07 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_ipsec.h"
70
71#include <sys/param.h>
72#include <sys/systm.h>
73#include <sys/malloc.h>
74#include <sys/mbuf.h>
75#include <sys/protosw.h>
76#include <sys/socket.h>
77#include <sys/socketvar.h>
78#include <sys/time.h>
79#include <sys/kernel.h>
80#include <sys/syslog.h>
81#include <sys/domain.h>
82
83#include <net/if.h>
84#include <net/route.h>
85#include <net/if_dl.h>
86#include <net/if_types.h>
87
88#include <netinet/in.h>
89#include <netinet/in_var.h>
90#include <netinet/ip6.h>
91#include <netinet6/ip6_var.h>
92#include <netinet/icmp6.h>
93#include <netinet6/mld6_var.h>
94#include <netinet6/in6_pcb.h>
95#include <netinet6/nd6.h>
96#include <netinet6/in6_ifattach.h>
97#include <netinet6/ip6protosw.h>
98
99
100#ifdef IPSEC
101#include <netinet6/ipsec.h>
102#include <netkey/key.h>
103#endif
104
105#include "faith.h"
106
107#include <net/net_osdep.h>
108
109extern struct domain inet6domain;
110
111struct icmp6stat icmp6stat;
112
113extern struct in6pcb rawin6pcb;
114extern int icmp6errppslim;
115static int icmp6errpps_count = 0;
116static struct timeval icmp6errppslim_last;
117extern int icmp6_nodeinfo;
118
119/*
120 * List of callbacks to notify when Path MTU changes are made.
121 */
122struct icmp6_mtudisc_callback {
123	LIST_ENTRY(icmp6_mtudisc_callback) mc_list;
124	void (*mc_func) __P((struct in6_addr *));
125};
126
127LIST_HEAD(, icmp6_mtudisc_callback) icmp6_mtudisc_callbacks =
128    LIST_HEAD_INITIALIZER(&icmp6_mtudisc_callbacks);
129
130static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL;
131extern int pmtu_expire;
132
133/* XXX do these values make any sense? */
134int icmp6_mtudisc_hiwat = 1280;
135int icmp6_mtudisc_lowat = 256;
136
137static void icmp6_errcount __P((struct icmp6errstat *, int, int));
138static int icmp6_rip6_input __P((struct mbuf **, int));
139static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int));
140static const char *icmp6_redirect_diag __P((struct in6_addr *,
141	struct in6_addr *, struct in6_addr *));
142static struct mbuf *ni6_input __P((struct mbuf *, int));
143static struct mbuf *ni6_nametodns __P((const char *, int, int));
144static int ni6_dnsmatch __P((const char *, int, const char *, int));
145static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
146			  struct ifnet **, char *));
147static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
148				struct ifnet *, int));
149static struct rtentry *icmp6_mtudisc_clone __P((struct sockaddr *));
150static void icmp6_mtudisc_timeout __P((struct rtentry *, struct rttimer *));
151
152#ifdef COMPAT_RFC1885
153static struct route_in6 icmp6_reflect_rt;
154#endif
155
156void
157icmp6_init()
158{
159	mld6_init();
160	icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire);
161}
162
163static void
164icmp6_errcount(stat, type, code)
165	struct icmp6errstat *stat;
166	int type, code;
167{
168	switch(type) {
169	case ICMP6_DST_UNREACH:
170		switch (code) {
171		case ICMP6_DST_UNREACH_NOROUTE:
172			stat->icp6errs_dst_unreach_noroute++;
173			return;
174		case ICMP6_DST_UNREACH_ADMIN:
175			stat->icp6errs_dst_unreach_admin++;
176			return;
177		case ICMP6_DST_UNREACH_BEYONDSCOPE:
178			stat->icp6errs_dst_unreach_beyondscope++;
179			return;
180		case ICMP6_DST_UNREACH_ADDR:
181			stat->icp6errs_dst_unreach_addr++;
182			return;
183		case ICMP6_DST_UNREACH_NOPORT:
184			stat->icp6errs_dst_unreach_noport++;
185			return;
186		}
187		break;
188	case ICMP6_PACKET_TOO_BIG:
189		stat->icp6errs_packet_too_big++;
190		return;
191	case ICMP6_TIME_EXCEEDED:
192		switch(code) {
193		case ICMP6_TIME_EXCEED_TRANSIT:
194			stat->icp6errs_time_exceed_transit++;
195			return;
196		case ICMP6_TIME_EXCEED_REASSEMBLY:
197			stat->icp6errs_time_exceed_reassembly++;
198			return;
199		}
200		break;
201	case ICMP6_PARAM_PROB:
202		switch(code) {
203		case ICMP6_PARAMPROB_HEADER:
204			stat->icp6errs_paramprob_header++;
205			return;
206		case ICMP6_PARAMPROB_NEXTHEADER:
207			stat->icp6errs_paramprob_nextheader++;
208			return;
209		case ICMP6_PARAMPROB_OPTION:
210			stat->icp6errs_paramprob_option++;
211			return;
212		}
213		break;
214	case ND_REDIRECT:
215		stat->icp6errs_redirect++;
216		return;
217	}
218	stat->icp6errs_unknown++;
219}
220
221/*
222 * Register a Path MTU Discovery callback.
223 */
224void
225icmp6_mtudisc_callback_register(func)
226	void (*func) __P((struct in6_addr *));
227{
228	struct icmp6_mtudisc_callback *mc;
229
230	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
231	     mc = LIST_NEXT(mc, mc_list)) {
232		if (mc->mc_func == func)
233			return;
234	}
235
236	mc = malloc(sizeof(*mc), M_PCB, M_NOWAIT);
237	if (mc == NULL)
238		panic("icmp6_mtudisc_callback_register");
239
240	mc->mc_func = func;
241	LIST_INSERT_HEAD(&icmp6_mtudisc_callbacks, mc, mc_list);
242}
243
244/*
245 * Generate an error packet of type error in response to bad IP6 packet.
246 */
247void
248icmp6_error(m, type, code, param)
249	struct mbuf *m;
250	int type, code, param;
251{
252	struct ip6_hdr *oip6, *nip6;
253	struct icmp6_hdr *icmp6;
254	u_int preplen;
255	int off;
256	int nxt;
257
258	icmp6stat.icp6s_error++;
259
260	/* count per-type-code statistics */
261	icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
262
263	if (m->m_flags & M_DECRYPTED) {
264		icmp6stat.icp6s_canterror++;
265		goto freeit;
266	}
267
268#ifndef PULLDOWN_TEST
269	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
270#else
271	if (m->m_len < sizeof(struct ip6_hdr)) {
272		m = m_pullup(m, sizeof(struct ip6_hdr));
273		if (m == NULL)
274			return;
275	}
276#endif
277	oip6 = mtod(m, struct ip6_hdr *);
278
279	/*
280	 * Multicast destination check. For unrecognized option errors,
281	 * this check has already done in ip6_unknown_opt(), so we can
282	 * check only for other errors.
283	 */
284	if ((m->m_flags & (M_BCAST|M_MCAST) ||
285	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
286	    (type != ICMP6_PACKET_TOO_BIG &&
287	     (type != ICMP6_PARAM_PROB ||
288	      code != ICMP6_PARAMPROB_OPTION)))
289		goto freeit;
290
291	/* Source address check. XXX: the case of anycast source? */
292	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
293	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
294		goto freeit;
295
296	/*
297	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
298	 * don't do it.
299	 */
300	nxt = -1;
301	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
302	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
303		struct icmp6_hdr *icp;
304
305#ifndef PULLDOWN_TEST
306		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
307		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
308#else
309		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
310			sizeof(*icp));
311		if (icp == NULL) {
312			icmp6stat.icp6s_tooshort++;
313			return;
314		}
315#endif
316		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
317		    icp->icmp6_type == ND_REDIRECT) {
318			/*
319			 * ICMPv6 error
320			 * Special case: for redirect (which is
321			 * informational) we must not send icmp6 error.
322			 */
323			icmp6stat.icp6s_canterror++;
324			goto freeit;
325		} else {
326			/* ICMPv6 informational - send the error */
327		}
328	} else {
329		/* non-ICMPv6 - send the error */
330	}
331
332	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
333
334	/* Finally, do rate limitation check. */
335	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
336		icmp6stat.icp6s_toofreq++;
337		goto freeit;
338	}
339
340	/*
341	 * OK, ICMP6 can be generated.
342	 */
343
344	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
345		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
346
347	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
348	M_PREPEND(m, preplen, M_DONTWAIT);
349	if (m && m->m_len < preplen)
350		m = m_pullup(m, preplen);
351	if (m == NULL) {
352		printf("ENOBUFS in icmp6_error %d\n", __LINE__);
353		return;
354	}
355
356	nip6 = mtod(m, struct ip6_hdr *);
357	nip6->ip6_src  = oip6->ip6_src;
358	nip6->ip6_dst  = oip6->ip6_dst;
359
360	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src))
361		oip6->ip6_src.s6_addr16[1] = 0;
362	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst))
363		oip6->ip6_dst.s6_addr16[1] = 0;
364
365	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
366	icmp6->icmp6_type = type;
367	icmp6->icmp6_code = code;
368	icmp6->icmp6_pptr = htonl((u_int32_t)param);
369
370	icmp6stat.icp6s_outhist[type]++;
371	icmp6_reflect(m, sizeof(struct ip6_hdr)); /*header order: IPv6 - ICMPv6*/
372
373	return;
374
375  freeit:
376	/*
377	 * If we can't tell wheter or not we can generate ICMP6, free it.
378	 */
379	m_freem(m);
380}
381
382/*
383 * Process a received ICMP6 message.
384 */
385int
386icmp6_input(mp, offp, proto)
387	struct mbuf **mp;
388	int *offp, proto;
389{
390	struct mbuf *m = *mp, *n;
391	struct ip6_hdr *ip6, *nip6;
392	struct icmp6_hdr *icmp6, *nicmp6;
393	int off = *offp;
394	int icmp6len = m->m_pkthdr.len - *offp;
395	int code, sum, noff;
396	struct sockaddr_in6 icmp6src;
397
398#ifndef PULLDOWN_TEST
399	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
400	/* m might change if M_LOOP. So, call mtod after this */
401#endif
402
403	/*
404	 * Locate icmp6 structure in mbuf, and check
405	 * that not corrupted and of at least minimum length
406	 */
407
408	ip6 = mtod(m, struct ip6_hdr *);
409	if (icmp6len < sizeof(struct icmp6_hdr)) {
410		icmp6stat.icp6s_tooshort++;
411		goto freeit;
412	}
413
414	/*
415	 * calculate the checksum
416	 */
417#ifndef PULLDOWN_TEST
418	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
419#else
420	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
421	if (icmp6 == NULL) {
422		icmp6stat.icp6s_tooshort++;
423		return IPPROTO_DONE;
424	}
425#endif
426	code = icmp6->icmp6_code;
427
428	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
429		log(LOG_ERR,
430		    "ICMP6 checksum error(%d|%x) %s\n",
431		    icmp6->icmp6_type,
432		    sum,
433		    ip6_sprintf(&ip6->ip6_src));
434		icmp6stat.icp6s_checksum++;
435		goto freeit;
436	}
437
438#if defined(NFAITH) && 0 < NFAITH
439	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
440		/*
441		 * Deliver very specific ICMP6 type only.
442		 * This is important to deilver TOOBIG.  Otherwise PMTUD
443		 * will not work.
444		 */
445		switch (icmp6->icmp6_type) {
446		case ICMP6_DST_UNREACH:
447		case ICMP6_PACKET_TOO_BIG:
448		case ICMP6_TIME_EXCEEDED:
449			break;
450		default:
451			goto freeit;
452		}
453	}
454#endif
455
456#ifdef IPSEC
457	/* drop it if it does not match the default policy */
458	if (ipsec6_in_reject(m, NULL)) {
459		ipsec6stat.in_polvio++;
460		goto freeit;
461	}
462#endif
463
464	icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
465	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
466	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
467		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
468
469	switch (icmp6->icmp6_type) {
470
471	case ICMP6_DST_UNREACH:
472		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
473		switch (code) {
474		case ICMP6_DST_UNREACH_NOROUTE:
475			code = PRC_UNREACH_NET;
476			break;
477		case ICMP6_DST_UNREACH_ADMIN:
478			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
479			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
480			break;
481		case ICMP6_DST_UNREACH_ADDR:
482			code = PRC_HOSTDEAD;
483			break;
484#ifdef COMPAT_RFC1885
485		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
486			code = PRC_UNREACH_SRCFAIL;
487			break;
488#else
489		case ICMP6_DST_UNREACH_BEYONDSCOPE:
490			/* I mean "source address was incorrect." */
491			code = PRC_PARAMPROB;
492			break;
493#endif
494		case ICMP6_DST_UNREACH_NOPORT:
495			code = PRC_UNREACH_PORT;
496			break;
497		default:
498			goto badcode;
499		}
500		goto deliver;
501		break;
502
503	case ICMP6_PACKET_TOO_BIG:
504		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
505		if (code != 0)
506			goto badcode;
507
508		code = PRC_MSGSIZE;
509
510		/*
511		 * Updating the path MTU will be done after examining
512		 * intermediate extension headers.
513		 */
514		goto deliver;
515		break;
516
517	case ICMP6_TIME_EXCEEDED:
518		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
519		switch (code) {
520		case ICMP6_TIME_EXCEED_TRANSIT:
521		case ICMP6_TIME_EXCEED_REASSEMBLY:
522			code += PRC_TIMXCEED_INTRANS;
523			break;
524		default:
525			goto badcode;
526		}
527		goto deliver;
528		break;
529
530	case ICMP6_PARAM_PROB:
531		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
532		switch (code) {
533		case ICMP6_PARAMPROB_NEXTHEADER:
534			code = PRC_UNREACH_PROTOCOL;
535			break;
536		case ICMP6_PARAMPROB_HEADER:
537		case ICMP6_PARAMPROB_OPTION:
538			code = PRC_PARAMPROB;
539			break;
540		default:
541			goto badcode;
542		}
543		goto deliver;
544		break;
545
546	case ICMP6_ECHO_REQUEST:
547		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
548		if (code != 0)
549			goto badcode;
550		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
551			/* Give up remote */
552			break;
553		}
554		if ((n->m_flags & M_EXT) != 0
555		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
556			struct mbuf *n0 = n;
557			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
558
559			/*
560			 * Prepare an internal mbuf. m_pullup() doesn't
561			 * always copy the length we specified.
562			 */
563			if (maxlen >= MCLBYTES) {
564#ifdef DIAGNOSTIC
565				printf("MCLBYTES too small\n");
566#endif
567				/* Give up remote */
568				m_freem(n0);
569				break;
570			}
571			MGETHDR(n, M_DONTWAIT, n0->m_type);
572			if (n && maxlen >= MHLEN) {
573				MCLGET(n, M_DONTWAIT);
574				if ((n->m_flags & M_EXT) == 0) {
575					m_free(n);
576					n = NULL;
577				}
578			}
579			if (n == NULL) {
580				/* Give up remote */
581				m_freem(n0);
582				break;
583			}
584			M_COPY_PKTHDR(n, n0);
585			/*
586			 * Copy IPv6 and ICMPv6 only.
587			 */
588			nip6 = mtod(n, struct ip6_hdr *);
589			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
590			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
591			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
592			noff = sizeof(struct ip6_hdr);
593			n->m_pkthdr.len = n->m_len =
594				noff + sizeof(struct icmp6_hdr);
595			/*
596			 * Adjust mbuf. ip6_plen will be adjusted in
597			 * ip6_output().
598			 */
599			m_adj(n0, off + sizeof(struct icmp6_hdr));
600			n->m_pkthdr.len += n0->m_pkthdr.len;
601			n->m_next = n0;
602			n0->m_flags &= ~M_PKTHDR;
603		} else {
604			nip6 = mtod(n, struct ip6_hdr *);
605			nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
606			noff = off;
607		}
608		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
609		nicmp6->icmp6_code = 0;
610		if (n) {
611			icmp6stat.icp6s_reflect++;
612			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
613			icmp6_reflect(n, noff);
614		}
615		break;
616
617	case ICMP6_ECHO_REPLY:
618		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
619		if (code != 0)
620			goto badcode;
621		break;
622
623	case MLD6_LISTENER_QUERY:
624	case MLD6_LISTENER_REPORT:
625		if (icmp6len < sizeof(struct mld6_hdr))
626			goto badlen;
627		if (icmp6->icmp6_type == MLD6_LISTENER_QUERY) /* XXX: ugly... */
628			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
629		else
630			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
631		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
632			/* give up local */
633			mld6_input(m, off);
634			m = NULL;
635			goto freeit;
636		}
637		mld6_input(n, off);
638		/* m stays. */
639		break;
640
641	case MLD6_LISTENER_DONE:
642		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
643		if (icmp6len < sizeof(struct mld6_hdr))	/* necessary? */
644			goto badlen;
645		break;		/* nothing to be done in kernel */
646
647	case MLD6_MTRACE_RESP:
648	case MLD6_MTRACE:
649		/* XXX: these two are experimental. not officially defind. */
650		/* XXX: per-interface statistics? */
651		break;		/* just pass it to applications */
652
653	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
654	    {
655		enum { WRU, FQDN } mode;
656
657		if (!icmp6_nodeinfo)
658			break;
659
660		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
661			mode = WRU;
662		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
663			mode = FQDN;
664		else
665			goto badlen;
666
667		if (mode == FQDN) {
668#ifndef PULLDOWN_TEST
669			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
670					 IPPROTO_DONE);
671#endif
672			n = m_copy(m, 0, M_COPYALL);
673			if (n)
674				n = ni6_input(n, off);
675			/* XXX meaningless if n == NULL */
676			noff = sizeof(struct ip6_hdr);
677		} else {
678			u_char *p;
679			int maxlen, maxhlen;
680
681			if (code != 0)
682				goto badcode;
683			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
684			if (maxlen >= MCLBYTES) {
685#ifdef DIAGNOSTIC
686				printf("MCLBYTES too small\n");
687#endif
688				/* Give up remote */
689				break;
690			}
691			MGETHDR(n, M_DONTWAIT, m->m_type);
692			if (n && maxlen > MHLEN) {
693				MCLGET(n, M_DONTWAIT);
694				if ((n->m_flags & M_EXT) == 0) {
695					m_free(n);
696					n = NULL;
697				}
698			}
699			if (n == NULL) {
700				/* Give up remote */
701				break;
702			}
703			n->m_len = 0;
704			maxhlen = M_TRAILINGSPACE(n) - maxlen;
705			if (maxhlen > hostnamelen)
706				maxhlen = hostnamelen;
707			/*
708			 * Copy IPv6 and ICMPv6 only.
709			 */
710			nip6 = mtod(n, struct ip6_hdr *);
711			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
712			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
713			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
714			p = (u_char *)(nicmp6 + 1);
715			bzero(p, 4);
716			bcopy(hostname, p + 4, maxhlen); /*meaningless TTL*/
717			noff = sizeof(struct ip6_hdr);
718			M_COPY_PKTHDR(n, m); /* just for recvif */
719			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
720				sizeof(struct icmp6_hdr) + 4 + maxhlen;
721			nicmp6->icmp6_type = ICMP6_WRUREPLY;
722			nicmp6->icmp6_code = 0;
723		}
724#undef hostnamelen
725		if (n) {
726			icmp6stat.icp6s_reflect++;
727			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
728			icmp6_reflect(n, noff);
729		}
730		break;
731	    }
732
733	case ICMP6_WRUREPLY:
734		if (code != 0)
735			goto badcode;
736		break;
737
738	case ND_ROUTER_SOLICIT:
739		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
740		if (code != 0)
741			goto badcode;
742		if (icmp6len < sizeof(struct nd_router_solicit))
743			goto badlen;
744		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
745			/* give up local */
746			nd6_rs_input(m, off, icmp6len);
747			m = NULL;
748			goto freeit;
749		}
750		nd6_rs_input(n, off, icmp6len);
751		/* m stays. */
752		break;
753
754	case ND_ROUTER_ADVERT:
755		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
756		if (code != 0)
757			goto badcode;
758		if (icmp6len < sizeof(struct nd_router_advert))
759			goto badlen;
760		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
761			/* give up local */
762			nd6_ra_input(m, off, icmp6len);
763			m = NULL;
764			goto freeit;
765		}
766		nd6_ra_input(n, off, icmp6len);
767		/* m stays. */
768		break;
769
770	case ND_NEIGHBOR_SOLICIT:
771		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
772		if (code != 0)
773			goto badcode;
774		if (icmp6len < sizeof(struct nd_neighbor_solicit))
775			goto badlen;
776		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
777			/* give up local */
778			nd6_ns_input(m, off, icmp6len);
779			m = NULL;
780			goto freeit;
781		}
782		nd6_ns_input(n, off, icmp6len);
783		/* m stays. */
784		break;
785
786	case ND_NEIGHBOR_ADVERT:
787		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
788		if (code != 0)
789			goto badcode;
790		if (icmp6len < sizeof(struct nd_neighbor_advert))
791			goto badlen;
792		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
793			/* give up local */
794			nd6_na_input(m, off, icmp6len);
795			m = NULL;
796			goto freeit;
797		}
798		nd6_na_input(n, off, icmp6len);
799		/* m stays. */
800		break;
801
802	case ND_REDIRECT:
803		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
804		if (code != 0)
805			goto badcode;
806		if (icmp6len < sizeof(struct nd_redirect))
807			goto badlen;
808		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
809			/* give up local */
810			icmp6_redirect_input(m, off);
811			m = NULL;
812			goto freeit;
813		}
814		icmp6_redirect_input(n, off);
815		/* m stays. */
816		break;
817
818	case ICMP6_ROUTER_RENUMBERING:
819		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
820		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
821			goto badcode;
822		if (icmp6len < sizeof(struct icmp6_router_renum))
823			goto badlen;
824		break;
825
826	default:
827		printf("icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
828		       icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
829		       ip6_sprintf(&ip6->ip6_dst),
830		       m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0);
831		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
832			/* ICMPv6 error: MUST deliver it by spec... */
833			code = PRC_NCMDS;
834			/* deliver */
835		} else {
836			/* ICMPv6 informational: MUST not deliver */
837			break;
838		}
839	deliver:
840		if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
841			icmp6stat.icp6s_tooshort++;
842			goto freeit;
843		}
844#ifndef PULLDOWN_TEST
845		IP6_EXTHDR_CHECK(m, off,
846			sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr),
847			IPPROTO_DONE);
848		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
849#else
850		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
851			sizeof(*icmp6) + sizeof(struct ip6_hdr));
852		if (icmp6 == NULL) {
853			icmp6stat.icp6s_tooshort++;
854			return IPPROTO_DONE;
855		}
856#endif
857		bzero(&icmp6src, sizeof(icmp6src));
858		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
859		icmp6src.sin6_family = AF_INET6;
860		icmp6src.sin6_addr = ((struct ip6_hdr *)(icmp6 + 1))->ip6_dst;
861
862		/* Detect the upper level protocol */
863	    {
864		void (*ctlfunc) __P((int, struct sockaddr *, void *));
865		struct ip6_hdr *eip6 = (struct ip6_hdr *)(icmp6 + 1);
866		u_int8_t nxt = eip6->ip6_nxt;
867		int eoff = off + sizeof(struct icmp6_hdr) +
868			sizeof(struct ip6_hdr);
869		struct ip6ctlparam ip6cp;
870		struct in6_addr *finaldst = NULL;
871		struct ip6_frag *fh;
872		struct ip6_rthdr *rth;
873		struct ip6_rthdr0 *rth0;
874		int rthlen;
875
876		while (1) { /* XXX: should avoid inf. loop explicitly? */
877			struct ip6_ext *eh;
878
879			switch(nxt) {
880			case IPPROTO_HOPOPTS:
881			case IPPROTO_DSTOPTS:
882			case IPPROTO_AH:
883#ifndef PULLDOWN_TEST
884				IP6_EXTHDR_CHECK(m, 0, eoff +
885						 sizeof(struct ip6_ext),
886						 IPPROTO_DONE);
887				eh = (struct ip6_ext *)(mtod(m, caddr_t)
888							+ eoff);
889#else
890				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
891					eoff, sizeof(*eh));
892				if (eh == NULL) {
893					icmp6stat.icp6s_tooshort++;
894					return IPPROTO_DONE;
895				}
896#endif
897
898				if (nxt == IPPROTO_AH)
899					eoff += (eh->ip6e_len + 2) << 2;
900				else
901					eoff += (eh->ip6e_len + 1) << 3;
902				nxt = eh->ip6e_nxt;
903				break;
904			case IPPROTO_ROUTING:
905				/*
906				 * When the erroneous packet contains a
907				 * routing header, we should examine the
908				 * header to determine the final destination.
909				 * Otherwise, we can't properly update
910				 * information that depends on the final
911				 * destination (e.g. path MTU).
912				 */
913#ifndef PULLDOWN_TEST
914				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth),
915						 IPPROTO_DONE);
916				rth = (struct ip6_rthdr *)(mtod(m, caddr_t)
917							   + eoff);
918#else
919				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
920					eoff, sizeof(*rth));
921				if (rth == NULL) {
922					icmp6stat.icp6s_tooshort++;
923					return IPPROTO_DONE;
924				}
925#endif
926				rthlen = (rth->ip6r_len + 1) << 3;
927				/*
928				 * XXX: currently there is no
929				 * officially defined type other
930				 * than type-0.
931				 * Note that if the segment left field
932				 * is 0, all intermediate hops must
933				 * have been passed.
934				 */
935				if (rth->ip6r_segleft &&
936				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
937					int hops;
938
939#ifndef PULLDOWN_TEST
940					IP6_EXTHDR_CHECK(m, 0, eoff + rthlen,
941							 IPPROTO_DONE);
942					rth0 = (struct ip6_rthdr0 *)(mtod(m, caddr_t) + eoff);
943#else
944					IP6_EXTHDR_GET(rth0,
945						       struct ip6_rthdr0 *, m,
946						       eoff, rthlen);
947					if (rth0 == NULL) {
948						icmp6stat.icp6s_tooshort++;
949						return IPPROTO_DONE;
950					}
951#endif
952					/* just ignore a bogus header */
953					if ((rth0->ip6r0_len % 2) == 0 &&
954					    (hops = rth0->ip6r0_len/2))
955						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
956				}
957				eoff += rthlen;
958				nxt = rth->ip6r_nxt;
959				break;
960			case IPPROTO_FRAGMENT:
961#ifndef PULLDOWN_TEST
962				IP6_EXTHDR_CHECK(m, 0, eoff +
963						 sizeof(struct ip6_frag),
964						 IPPROTO_DONE);
965				fh = (struct ip6_frag *)(mtod(m, caddr_t)
966							 + eoff);
967#else
968				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
969					eoff, sizeof(*fh));
970				if (fh == NULL) {
971					icmp6stat.icp6s_tooshort++;
972					return IPPROTO_DONE;
973				}
974#endif
975				/*
976				 * Data after a fragment header is meaningless
977				 * unless it is the first fragment, but
978				 * we'll go to the notify label for path MTU
979				 * discovery.
980				 */
981				if (fh->ip6f_offlg & IP6F_OFF_MASK)
982					goto notify;
983
984				eoff += sizeof(struct ip6_frag);
985				nxt = fh->ip6f_nxt;
986				break;
987			default:
988				/*
989				 * This case includes ESP and the No Next
990				 * Header. In such cases going to the notify
991				 * label does not have any meaning
992				 * (i.e. ctlfunc will be NULL), but we go
993				 * anyway since we might have to update
994				 * path MTU information.
995				 */
996				goto notify;
997			}
998		}
999	    notify:
1000#ifndef PULLDOWN_TEST
1001		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1002#else
1003		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1004			sizeof(*icmp6) + sizeof(struct ip6_hdr));
1005		if (icmp6 == NULL) {
1006			icmp6stat.icp6s_tooshort++;
1007			return IPPROTO_DONE;
1008		}
1009#endif
1010		if (finaldst == NULL)
1011			finaldst = &((struct ip6_hdr *)(icmp6 + 1))->ip6_dst;
1012		ip6cp.ip6c_m = m;
1013		ip6cp.ip6c_icmp6 = icmp6;
1014		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1015		ip6cp.ip6c_off = eoff;
1016		ip6cp.ip6c_finaldst = finaldst;
1017
1018		ctlfunc = (void (*) __P((int, struct sockaddr *, void *)))
1019			(inet6sw[ip6_protox[nxt]].pr_ctlinput);
1020		if (ctlfunc) {
1021			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6src,
1022			    &ip6cp);
1023		}
1024	    }
1025		break;
1026
1027	badcode:
1028		icmp6stat.icp6s_badcode++;
1029		break;
1030
1031	badlen:
1032		icmp6stat.icp6s_badlen++;
1033		break;
1034	}
1035
1036	icmp6_rip6_input(&m, *offp);
1037	return IPPROTO_DONE;
1038
1039 freeit:
1040	m_freem(m);
1041	return IPPROTO_DONE;
1042}
1043
1044void
1045icmp6_mtudisc_update(ip6cp, validated)
1046	struct ip6ctlparam *ip6cp;
1047	int validated;
1048{
1049	unsigned long rtcount;
1050	struct icmp6_mtudisc_callback *mc;
1051	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1052	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1053	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1054	u_int mtu = ntohl(icmp6->icmp6_mtu);
1055	struct rtentry *rt = NULL;
1056	struct sockaddr_in6 sin6;
1057
1058	/*
1059	 * allow non-validated cases if memory is plenty, to make traffic
1060	 * from non-connected pcb happy.
1061	 */
1062	rtcount = rt_timer_count(icmp6_mtudisc_timeout_q);
1063	if (validated) {
1064		if (rtcount > icmp6_mtudisc_hiwat)
1065			return;
1066		else if (rtcount > icmp6_mtudisc_lowat) {
1067			/*
1068			 * XXX nuke a victim, install the new one.
1069			 */
1070		}
1071	} else {
1072		if (rtcount > icmp6_mtudisc_lowat)
1073			return;
1074	}
1075
1076	bzero(&sin6, sizeof(sin6));
1077	sin6.sin6_family = PF_INET6;
1078	sin6.sin6_len = sizeof(struct sockaddr_in6);
1079	sin6.sin6_addr = *dst;
1080	/* XXX normally, this won't happen */
1081	if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1082		sin6.sin6_addr.s6_addr16[1] =
1083		    htons(m->m_pkthdr.rcvif->if_index);
1084	}
1085	/* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1086	rt = rtalloc1((struct sockaddr *)&sin6, 1);	/*clone*/
1087	if (!rt || (rt->rt_flags & RTF_HOST) == 0) {
1088		if (rt)
1089			RTFREE(rt);
1090		rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6);
1091	}
1092
1093	if (rt && (rt->rt_flags & RTF_HOST)
1094	    && !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1095		if (mtu < IPV6_MMTU) {
1096				/* xxx */
1097			rt->rt_rmx.rmx_locks |= RTV_MTU;
1098		} else if (mtu < rt->rt_ifp->if_mtu &&
1099			   rt->rt_rmx.rmx_mtu > mtu) {
1100			icmp6stat.icp6s_pmtuchg++;
1101			rt->rt_rmx.rmx_mtu = mtu;
1102		}
1103	}
1104	if (rt)
1105		RTFREE(rt);
1106
1107	/*
1108	 * Notify protocols that the MTU for this destination
1109	 * has changed.
1110	 */
1111	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
1112	     mc = LIST_NEXT(mc, mc_list))
1113		(*mc->mc_func)(&sin6.sin6_addr);
1114}
1115
1116/*
1117 * Process a Node Information Query packet, based on
1118 * draft-ietf-ipngwg-icmp-name-lookups-07.
1119 *
1120 * Spec incompatibilities:
1121 * - IPv6 Subject address handling
1122 * - IPv4 Subject address handling support missing
1123 * - Proxy reply (answer even if it's not for me)
1124 * - joins NI group address at in6_ifattach() time only, does not cope
1125 *   with hostname changes by sethostname(3)
1126 */
1127#ifndef offsetof		/* XXX */
1128#define	offsetof(type, member)	((size_t)(&((type *)0)->member))
1129#endif
1130static struct mbuf *
1131ni6_input(m, off)
1132	struct mbuf *m;
1133	int off;
1134{
1135	struct icmp6_nodeinfo *ni6, *nni6;
1136	struct mbuf *n = NULL;
1137	u_int16_t qtype;
1138	int subjlen;
1139	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1140	struct ni_reply_fqdn *fqdn;
1141	int addrs;		/* for NI_QTYPE_NODEADDR */
1142	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1143	struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1144	struct ip6_hdr *ip6;
1145	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1146	char *subj = NULL;
1147
1148	ip6 = mtod(m, struct ip6_hdr *);
1149#ifndef PULLDOWN_TEST
1150	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1151#else
1152	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1153	if (ni6 == NULL) {
1154		/* m is already reclaimed */
1155		return NULL;
1156	}
1157#endif
1158
1159	/*
1160	 * Validate IPv6 destination address.
1161	 *
1162	 * The Responder must discard the Query without further processing
1163	 * unless it is one of the Responder's unicast or anycast addresses, or
1164	 * a link-local scope multicast address which the Responder has joined.
1165	 * [icmp-name-lookups-07, Section 4.]
1166	 */
1167	bzero(&sin6, sizeof(sin6));
1168	sin6.sin6_family = AF_INET6;
1169	sin6.sin6_len = sizeof(struct sockaddr_in6);
1170	bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1171	/* XXX scopeid */
1172	if (ifa_ifwithaddr((struct sockaddr *)&sin6))
1173		; /* unicast/anycast, fine */
1174	else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1175		; /* link-local multicast, fine */
1176	else
1177		goto bad;
1178
1179	/* validate query Subject field. */
1180	qtype = ntohs(ni6->ni_qtype);
1181	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1182	switch (qtype) {
1183	case NI_QTYPE_NOOP:
1184	case NI_QTYPE_SUPTYPES:
1185		/* 07 draft */
1186		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1187			break;
1188		/*FALLTHROUGH*/
1189	case NI_QTYPE_FQDN:
1190	case NI_QTYPE_NODEADDR:
1191		switch (ni6->ni_code) {
1192		case ICMP6_NI_SUBJ_IPV6:
1193#if ICMP6_NI_SUBJ_IPV6 != 0
1194		case 0:
1195#endif
1196			/*
1197			 * backward compatibility - try to accept 03 draft
1198			 * format, where no Subject is present.
1199			 */
1200			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1201			    subjlen == 0) {
1202				oldfqdn++;
1203				break;
1204			}
1205#if ICMP6_NI_SUBJ_IPV6 != 0
1206			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1207				goto bad;
1208#endif
1209
1210			if (subjlen != sizeof(sin6.sin6_addr))
1211				goto bad;
1212
1213			/*
1214			 * Validate Subject address.
1215			 *
1216			 * Not sure what exactly does "address belongs to the
1217			 * node" mean in the spec, is it just unicast, or what?
1218			 *
1219			 * At this moment we consider Subject address as
1220			 * "belong to the node" if the Subject address equals
1221			 * to the IPv6 destination address; validation for
1222			 * IPv6 destination address should have done enough
1223			 * check for us.
1224			 *
1225			 * We do not do proxy at this moment.
1226			 */
1227			/* m_pulldown instead of copy? */
1228			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1229			    subjlen, (caddr_t)&sin6.sin6_addr);
1230			/* XXX kame scope hack */
1231			if (IN6_IS_SCOPE_LINKLOCAL(&sin6.sin6_addr)) {
1232#ifdef FAKE_LOOPBACK_IF
1233				if ((m->m_flags & M_PKTHDR) != 0 &&
1234				    m->m_pkthdr.rcvif) {
1235					sin6.sin6_addr.s6_addr16[1] =
1236					    htons(m->m_pkthdr.rcvif->if_index);
1237				}
1238#else
1239				if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) {
1240					sin6.sin6_addr.s6_addr16[1] =
1241					    ip6->ip6_dst.s6_addr16[1];
1242				}
1243#endif
1244			}
1245			subj = (char *)&sin6;
1246			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &sin6.sin6_addr))
1247				break;
1248
1249			/*
1250			 * XXX if we are to allow other cases, we should really
1251			 * be careful about scope here.
1252			 * basically, we should disallow queries toward IPv6
1253			 * destination X with subject Y, if scope(X) > scope(Y).
1254			 * if we allow scope(X) > scope(Y), it will result in
1255			 * information leakage across scope boundary.
1256			 */
1257			goto bad;
1258
1259		case ICMP6_NI_SUBJ_FQDN:
1260			/*
1261			 * Validate Subject name with gethostname(3).
1262			 *
1263			 * The behavior may need some debate, since:
1264			 * - we are not sure if the node has FQDN as
1265			 *   hostname (returned by gethostname(3)).
1266			 * - the code does wildcard match for truncated names.
1267			 *   however, we are not sure if we want to perform
1268			 *   wildcard match, if gethostname(3) side has
1269			 *   truncated hostname.
1270			 */
1271			n = ni6_nametodns(hostname, hostnamelen, 0);
1272			if (!n || n->m_next || n->m_len == 0)
1273				goto bad;
1274			IP6_EXTHDR_GET(subj, char *, m,
1275			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1276			if (subj == NULL)
1277				goto bad;
1278			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1279					n->m_len)) {
1280				goto bad;
1281			}
1282			m_freem(n);
1283			n = NULL;
1284			break;
1285
1286		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1287		default:
1288			goto bad;
1289		}
1290		break;
1291	}
1292
1293	/* guess reply length */
1294	switch (qtype) {
1295	case NI_QTYPE_NOOP:
1296		break;		/* no reply data */
1297	case NI_QTYPE_SUPTYPES:
1298		replylen += sizeof(u_int32_t);
1299		break;
1300	case NI_QTYPE_FQDN:
1301		/* XXX will append an mbuf */
1302		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1303		break;
1304	case NI_QTYPE_NODEADDR:
1305		addrs = ni6_addrs(ni6, m, &ifp, subj);
1306		if ((replylen += addrs * (sizeof(struct in6_addr) +
1307					  sizeof(u_int32_t))) > MCLBYTES)
1308			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1309		break;
1310	default:
1311		/*
1312		 * XXX: We must return a reply with the ICMP6 code
1313		 * `unknown Qtype' in this case. However we regard the case
1314		 * as an FQDN query for backward compatibility.
1315		 * Older versions set a random value to this field,
1316		 * so it rarely varies in the defined qtypes.
1317		 * But the mechanism is not reliable...
1318		 * maybe we should obsolete older versions.
1319		 */
1320		qtype = NI_QTYPE_FQDN;
1321		/* XXX will append an mbuf */
1322		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1323		oldfqdn++;
1324		break;
1325	}
1326
1327	/* allocate an mbuf to reply. */
1328	MGETHDR(n, M_DONTWAIT, m->m_type);
1329	if (n == NULL) {
1330		m_freem(m);
1331		return(NULL);
1332	}
1333	M_COPY_PKTHDR(n, m); /* just for recvif */
1334	if (replylen > MHLEN) {
1335		if (replylen > MCLBYTES) {
1336			 /*
1337			  * XXX: should we try to allocate more? But MCLBYTES
1338			  * is probably much larger than IPV6_MMTU...
1339			  */
1340			goto bad;
1341		}
1342		MCLGET(n, M_DONTWAIT);
1343		if ((n->m_flags & M_EXT) == 0) {
1344			goto bad;
1345		}
1346	}
1347	n->m_pkthdr.len = n->m_len = replylen;
1348
1349	/* copy mbuf header and IPv6 + Node Information base headers */
1350	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1351	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1352	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1353
1354	/* qtype dependent procedure */
1355	switch (qtype) {
1356	case NI_QTYPE_NOOP:
1357		nni6->ni_code = ICMP6_NI_SUCCESS;
1358		nni6->ni_flags = 0;
1359		break;
1360	case NI_QTYPE_SUPTYPES:
1361	{
1362		u_int32_t v;
1363		nni6->ni_code = ICMP6_NI_SUCCESS;
1364		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1365		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1366		v = (u_int32_t)htonl(0x0000000f);
1367		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1368		break;
1369	}
1370	case NI_QTYPE_FQDN:
1371		nni6->ni_code = ICMP6_NI_SUCCESS;
1372		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1373						sizeof(struct ip6_hdr) +
1374						sizeof(struct icmp6_nodeinfo));
1375		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1376		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1377		/*
1378		 * XXX do we really have FQDN in variable "hostname"?
1379		 */
1380		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1381		if (n->m_next == NULL)
1382			goto bad;
1383		/* XXX we assume that n->m_next is not a chain */
1384		if (n->m_next->m_next != NULL)
1385			goto bad;
1386		n->m_pkthdr.len += n->m_next->m_len;
1387		break;
1388	case NI_QTYPE_NODEADDR:
1389	{
1390		int lenlim, copied;
1391
1392		nni6->ni_code = ICMP6_NI_SUCCESS;
1393		n->m_pkthdr.len = n->m_len =
1394		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1395		lenlim = M_TRAILINGSPACE(n);
1396		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1397		/* XXX: reset mbuf length */
1398		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1399			sizeof(struct icmp6_nodeinfo) + copied;
1400		break;
1401	}
1402	default:
1403		break;		/* XXX impossible! */
1404	}
1405
1406	nni6->ni_type = ICMP6_NI_REPLY;
1407	m_freem(m);
1408	return(n);
1409
1410  bad:
1411	m_freem(m);
1412	if (n)
1413		m_freem(n);
1414	return(NULL);
1415}
1416#undef hostnamelen
1417
1418/*
1419 * make a mbuf with DNS-encoded string.  no compression support.
1420 *
1421 * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1422 * treated as truncated name (two \0 at the end).  this is a wild guess.
1423 */
1424static struct mbuf *
1425ni6_nametodns(name, namelen, old)
1426	const char *name;
1427	int namelen;
1428	int old;	/* return pascal string if non-zero */
1429{
1430	struct mbuf *m;
1431	char *cp, *ep;
1432	const char *p, *q;
1433	int i, len, nterm;
1434
1435	if (old)
1436		len = namelen + 1;
1437	else
1438		len = MCLBYTES;
1439
1440	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1441	MGET(m, M_DONTWAIT, MT_DATA);
1442	if (m && len > MLEN) {
1443		MCLGET(m, M_DONTWAIT);
1444		if ((m->m_flags & M_EXT) == 0)
1445			goto fail;
1446	}
1447	if (!m)
1448		goto fail;
1449	m->m_next = NULL;
1450
1451	if (old) {
1452		m->m_len = len;
1453		*mtod(m, char *) = namelen;
1454		bcopy(name, mtod(m, char *) + 1, namelen);
1455		return m;
1456	} else {
1457		m->m_len = 0;
1458		cp = mtod(m, char *);
1459		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1460
1461		/* if not certain about my name, return empty buffer */
1462		if (namelen == 0)
1463			return m;
1464
1465		/*
1466		 * guess if it looks like shortened hostname, or FQDN.
1467		 * shortened hostname needs two trailing "\0".
1468		 */
1469		i = 0;
1470		for (p = name; p < name + namelen; p++) {
1471			if (*p && *p == '.')
1472				i++;
1473		}
1474		if (i < 2)
1475			nterm = 2;
1476		else
1477			nterm = 1;
1478
1479		p = name;
1480		while (cp < ep && p < name + namelen) {
1481			i = 0;
1482			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1483				i++;
1484			/* result does not fit into mbuf */
1485			if (cp + i + 1 >= ep)
1486				goto fail;
1487			/* DNS label length restriction, RFC1035 page 8 */
1488			if (i >= 64)
1489				goto fail;
1490			*cp++ = i;
1491			bcopy(p, cp, i);
1492			cp += i;
1493			p = q;
1494			if (p < name + namelen && *p == '.')
1495				p++;
1496		}
1497		/* termination */
1498		if (cp + nterm >= ep)
1499			goto fail;
1500		while (nterm-- > 0)
1501			*cp++ = '\0';
1502		m->m_len = cp - mtod(m, char *);
1503		return m;
1504	}
1505
1506	panic("should not reach here");
1507	/*NOTREACHED*/
1508
1509 fail:
1510	if (m)
1511		m_freem(m);
1512	return NULL;
1513}
1514
1515/*
1516 * check if two DNS-encoded string matches.  takes care of truncated
1517 * form (with \0\0 at the end).  no compression support.
1518 * XXX upper/lowercase match (see RFC2065)
1519 */
1520static int
1521ni6_dnsmatch(a, alen, b, blen)
1522	const char *a;
1523	int alen;
1524	const char *b;
1525	int blen;
1526{
1527	const char *a0, *b0;
1528	int l;
1529
1530	/* simplest case - need validation? */
1531	if (alen == blen && bcmp(a, b, alen) == 0)
1532		return 1;
1533
1534	a0 = a;
1535	b0 = b;
1536
1537	/* termination is mandatory */
1538	if (alen < 2 || blen < 2)
1539		return 0;
1540	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1541		return 0;
1542	alen--;
1543	blen--;
1544
1545	while (a - a0 < alen && b - b0 < blen) {
1546		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1547			return 0;
1548
1549		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1550			return 0;
1551		/* we don't support compression yet */
1552		if (a[0] >= 64 || b[0] >= 64)
1553			return 0;
1554
1555		/* truncated case */
1556		if (a[0] == 0 && a - a0 == alen - 1)
1557			return 1;
1558		if (b[0] == 0 && b - b0 == blen - 1)
1559			return 1;
1560		if (a[0] == 0 || b[0] == 0)
1561			return 0;
1562
1563		if (a[0] != b[0])
1564			return 0;
1565		l = a[0];
1566		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1567			return 0;
1568		if (bcmp(a + 1, b + 1, l) != 0)
1569			return 0;
1570
1571		a += 1 + l;
1572		b += 1 + l;
1573	}
1574
1575	if (a - a0 == alen && b - b0 == blen)
1576		return 1;
1577	else
1578		return 0;
1579}
1580
1581/*
1582 * calculate the number of addresses to be returned in the node info reply.
1583 */
1584static int
1585ni6_addrs(ni6, m, ifpp, subj)
1586	struct icmp6_nodeinfo *ni6;
1587	struct mbuf *m;
1588	struct ifnet **ifpp;
1589	char *subj;
1590{
1591	register struct ifnet *ifp;
1592	register struct in6_ifaddr *ifa6;
1593	register struct ifaddr *ifa;
1594	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1595	int addrs = 0, addrsofif, iffound = 0;
1596	int niflags = ni6->ni_flags;
1597
1598	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1599		switch(ni6->ni_code) {
1600		case ICMP6_NI_SUBJ_IPV6:
1601			if (subj == NULL) /* must be impossible... */
1602				return(0);
1603			subj_ip6 = (struct sockaddr_in6 *)subj;
1604			break;
1605		default:
1606			/*
1607			 * XXX: we only support IPv6 subject address for
1608			 * this Qtype.
1609			 */
1610			return(0);
1611		}
1612	}
1613
1614	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
1615	{
1616		addrsofif = 0;
1617		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1618		     ifa = ifa->ifa_list.tqe_next)
1619		{
1620			if (ifa->ifa_addr->sa_family != AF_INET6)
1621				continue;
1622			ifa6 = (struct in6_ifaddr *)ifa;
1623
1624			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1625			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1626					       &ifa6->ia_addr.sin6_addr))
1627				iffound = 1;
1628
1629			/*
1630			 * IPv4-mapped addresses can only be returned by a
1631			 * Node Information proxy, since they represent
1632			 * addresses of IPv4-only nodes, which perforce do
1633			 * not implement this protocol.
1634			 * [icmp-name-lookups-07, Section 5.4]
1635			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1636			 * this function at this moment.
1637			 */
1638
1639			/* What do we have to do about ::1? */
1640			switch(in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1641			case IPV6_ADDR_SCOPE_LINKLOCAL:
1642				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL)
1643				    == 0)
1644					continue;
1645				break;
1646			case IPV6_ADDR_SCOPE_SITELOCAL:
1647				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL)
1648				    == 0)
1649					continue;
1650				break;
1651			case IPV6_ADDR_SCOPE_GLOBAL:
1652				if ((niflags & NI_NODEADDR_FLAG_GLOBAL)
1653				    == 0)
1654					continue;
1655				break;
1656			default:
1657				continue;
1658			}
1659
1660			/*
1661			 * check if anycast is okay.
1662			 * XXX: just experimental. not in the spec.
1663			 */
1664			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1665			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1666				continue; /* we need only unicast addresses */
1667
1668			addrsofif++; /* count the address */
1669		}
1670		if (iffound) {
1671			*ifpp = ifp;
1672			return(addrsofif);
1673		}
1674
1675		addrs += addrsofif;
1676	}
1677
1678	return(addrs);
1679}
1680
1681static int
1682ni6_store_addrs(ni6, nni6, ifp0, resid)
1683	struct icmp6_nodeinfo *ni6, *nni6;
1684	struct ifnet *ifp0;
1685	int resid;
1686{
1687	register struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1688	register struct in6_ifaddr *ifa6;
1689	register struct ifaddr *ifa;
1690	struct ifnet *ifp_dep = NULL;
1691	int copied = 0, allow_deprecated = 0;
1692	u_char *cp = (u_char *)(nni6 + 1);
1693	int niflags = ni6->ni_flags;
1694	u_int32_t ltime;
1695	long time_second = time.tv_sec;
1696
1697	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1698		return(0);	/* needless to copy */
1699
1700  again:
1701
1702	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list))
1703	{
1704		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1705		     ifa = ifa->ifa_list.tqe_next)
1706		{
1707			if (ifa->ifa_addr->sa_family != AF_INET6)
1708				continue;
1709			ifa6 = (struct in6_ifaddr *)ifa;
1710
1711			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1712			    allow_deprecated == 0) {
1713				/*
1714				 * prefererred address should be put before
1715				 * deprecated addresses.
1716				 */
1717
1718				/* record the interface for later search */
1719				if (ifp_dep == NULL)
1720					ifp_dep = ifp;
1721
1722				continue;
1723			}
1724			else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1725				 allow_deprecated != 0)
1726				continue; /* we now collect deprecated addrs */
1727
1728			/* What do we have to do about ::1? */
1729			switch(in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1730			case IPV6_ADDR_SCOPE_LINKLOCAL:
1731				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL)
1732				    == 0)
1733					continue;
1734				break;
1735			case IPV6_ADDR_SCOPE_SITELOCAL:
1736				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL)
1737				    == 0)
1738					continue;
1739				break;
1740			case IPV6_ADDR_SCOPE_GLOBAL:
1741				if ((niflags & NI_NODEADDR_FLAG_GLOBAL)
1742				    == 0)
1743					continue;
1744				break;
1745			default:
1746				continue;
1747			}
1748
1749			/*
1750			 * check if anycast is okay.
1751			 * XXX: just experimental. not in the spec.
1752			 */
1753			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1754			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1755				continue;
1756
1757			/* now we can copy the address */
1758			if (resid < sizeof(struct in6_addr) +
1759			    sizeof(u_int32_t)) {
1760				/*
1761				 * We give up much more copy.
1762				 * Set the truncate flag and return.
1763				 */
1764				nni6->ni_flags |=
1765					NI_NODEADDR_FLAG_TRUNCATE;
1766				return(copied);
1767			}
1768
1769			/*
1770			 * Set the TTL of the address.
1771			 * The TTL value should be one of the following
1772			 * according to the specification:
1773			 *
1774			 * 1. The remaining lifetime of a DHCP lease on the
1775			 *    address, or
1776			 * 2. The remaining Valid Lifetime of a prefix from
1777			 *    which the address was derived through Stateless
1778			 *    Autoconfiguration.
1779			 *
1780			 * Note that we currently do not support stateful
1781			 * address configuration by DHCPv6, so the former
1782			 * case can't happen.
1783			 */
1784			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1785				ltime = ND6_INFINITE_LIFETIME;
1786			else {
1787				if (ifa6->ia6_lifetime.ia6t_expire >
1788				    time_second)
1789					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1790				else
1791					ltime = 0;
1792			}
1793
1794			bcopy(&ltime, cp, sizeof(u_int32_t));
1795			cp += sizeof(u_int32_t);
1796
1797			/* copy the address itself */
1798			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1799			      sizeof(struct in6_addr));
1800			/* XXX: KAME link-local hack; remove ifindex */
1801			if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr))
1802				((struct in6_addr *)cp)->s6_addr16[1] = 0;
1803			cp += sizeof(struct in6_addr);
1804
1805			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1806			copied += (sizeof(struct in6_addr) +
1807				   sizeof(u_int32_t));
1808		}
1809		if (ifp0)	/* we need search only on the specified IF */
1810			break;
1811	}
1812
1813	if (allow_deprecated == 0 && ifp_dep != NULL) {
1814		ifp = ifp_dep;
1815		allow_deprecated = 1;
1816
1817		goto again;
1818	}
1819
1820	return(copied);
1821}
1822
1823/*
1824 * XXX almost dup'ed code with rip6_input.
1825 */
1826static int
1827icmp6_rip6_input(mp, off)
1828	struct	mbuf **mp;
1829	int	off;
1830{
1831	struct mbuf *m = *mp;
1832	register struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1833	register struct in6pcb *in6p;
1834	struct in6pcb *last = NULL;
1835	struct sockaddr_in6 rip6src;
1836	struct icmp6_hdr *icmp6;
1837	struct mbuf *opts = NULL;
1838
1839#ifndef PULLDOWN_TEST
1840	/* this is assumed to be safe. */
1841	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1842#else
1843	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1844	if (icmp6 == NULL) {
1845		/* m is already reclaimed */
1846		return IPPROTO_DONE;
1847	}
1848#endif
1849
1850	bzero(&rip6src, sizeof(rip6src));
1851	rip6src.sin6_len = sizeof(struct sockaddr_in6);
1852	rip6src.sin6_family = AF_INET6;
1853	/* KAME hack: recover scopeid */
1854	(void)in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif);
1855
1856	for (in6p = rawin6pcb.in6p_next;
1857	     in6p != &rawin6pcb; in6p = in6p->in6p_next)
1858	{
1859		if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1860			continue;
1861		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1862		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1863			continue;
1864		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1865		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1866			continue;
1867		if (in6p->in6p_icmp6filt
1868		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1869				 in6p->in6p_icmp6filt))
1870			continue;
1871		if (last) {
1872			struct	mbuf *n;
1873			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1874				if (last->in6p_flags & IN6P_CONTROLOPTS)
1875					ip6_savecontrol(last, &opts, ip6, n);
1876				/* strip intermediate headers */
1877				m_adj(n, off);
1878				if (sbappendaddr(&last->in6p_socket->so_rcv,
1879						 (struct sockaddr *)&rip6src,
1880						 n, opts) == 0) {
1881					/* should notify about lost packet */
1882					m_freem(n);
1883					if (opts)
1884						m_freem(opts);
1885				} else
1886					sorwakeup(last->in6p_socket);
1887				opts = NULL;
1888			}
1889		}
1890		last = in6p;
1891	}
1892	if (last) {
1893		if (last->in6p_flags & IN6P_CONTROLOPTS)
1894			ip6_savecontrol(last, &opts, ip6, m);
1895		/* strip intermediate headers */
1896		m_adj(m, off);
1897		if (sbappendaddr(&last->in6p_socket->so_rcv,
1898				(struct sockaddr *)&rip6src, m, opts) == 0) {
1899			m_freem(m);
1900			if (opts)
1901				m_freem(opts);
1902		} else
1903			sorwakeup(last->in6p_socket);
1904	} else {
1905		m_freem(m);
1906		ip6stat.ip6s_delivered--;
1907	}
1908	return IPPROTO_DONE;
1909}
1910
1911/*
1912 * Reflect the ip6 packet back to the source.
1913 * OFF points to the icmp6 header, counted from the top of the mbuf.
1914 */
1915void
1916icmp6_reflect(m, off)
1917	struct	mbuf *m;
1918	size_t off;
1919{
1920	struct ip6_hdr *ip6;
1921	struct icmp6_hdr *icmp6;
1922	struct in6_ifaddr *ia;
1923	struct in6_addr t, *src = 0;
1924	int plen;
1925	int type, code;
1926	struct ifnet *outif = NULL;
1927#ifdef COMPAT_RFC1885
1928	int mtu = IPV6_MMTU;
1929	struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
1930#endif
1931
1932	/* too short to reflect */
1933	if (off < sizeof(struct ip6_hdr)) {
1934		printf("sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
1935		       (u_long)off, (u_long)sizeof(struct ip6_hdr),
1936		       __FILE__, __LINE__);
1937		goto bad;
1938	}
1939
1940	/*
1941	 * If there are extra headers between IPv6 and ICMPv6, strip
1942	 * off that header first.
1943	 */
1944#ifdef DIAGNOSTIC
1945	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
1946		panic("assumption failed in icmp6_reflect");
1947#endif
1948	if (off > sizeof(struct ip6_hdr)) {
1949		size_t l;
1950		struct ip6_hdr nip6;
1951
1952		l = off - sizeof(struct ip6_hdr);
1953		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
1954		m_adj(m, l);
1955		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1956		if (m->m_len < l) {
1957			if ((m = m_pullup(m, l)) == NULL)
1958				return;
1959		}
1960		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
1961	} else /* off == sizeof(struct ip6_hdr) */ {
1962		size_t l;
1963		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1964		if (m->m_len < l) {
1965			if ((m = m_pullup(m, l)) == NULL)
1966				return;
1967		}
1968	}
1969	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
1970	ip6 = mtod(m, struct ip6_hdr *);
1971	ip6->ip6_nxt = IPPROTO_ICMPV6;
1972	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
1973	type = icmp6->icmp6_type; /* keep type for statistics */
1974	code = icmp6->icmp6_code; /* ditto. */
1975
1976	t = ip6->ip6_dst;
1977	/*
1978	 * ip6_input() drops a packet if its src is multicast.
1979	 * So, the src is never multicast.
1980	 */
1981	ip6->ip6_dst = ip6->ip6_src;
1982
1983	/* XXX hack for link-local addresses */
1984	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
1985		ip6->ip6_dst.s6_addr16[1] =
1986			htons(m->m_pkthdr.rcvif->if_index);
1987	if (IN6_IS_ADDR_LINKLOCAL(&t))
1988		t.s6_addr16[1] = htons(m->m_pkthdr.rcvif->if_index);
1989
1990#ifdef COMPAT_RFC1885
1991	/*
1992	 * xxx guess MTU
1993	 * RFC 1885 requires that echo reply should be truncated if it
1994	 * does not fit in with (return) path MTU, but the description was
1995	 * removed in the new spec.
1996	 */
1997	if (icmp6_reflect_rt.ro_rt == 0 ||
1998	    ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
1999		if (icmp6_reflect_rt.ro_rt) {
2000			icmp6_reflect_rt.ro_rt = 0;
2001		}
2002		bzero(sin6, sizeof(*sin6));
2003		sin6->sin6_family = PF_INET6;
2004		sin6->sin6_len = sizeof(struct sockaddr_in6);
2005		sin6->sin6_addr = ip6->ip6_dst;
2006
2007		rtalloc((struct route *)&icmp6_reflect_rt.ro_rt);
2008	}
2009
2010	if (icmp6_reflect_rt.ro_rt == 0)
2011		goto bad;
2012
2013	if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
2014	    && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
2015		mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
2016
2017	if (mtu < m->m_pkthdr.len) {
2018		plen -= (m->m_pkthdr.len - mtu);
2019		m_adj(m, mtu - m->m_pkthdr.len);
2020	}
2021#endif
2022	/*
2023	 * If the incoming packet was addressed directly to us(i.e. unicast),
2024	 * use dst as the src for the reply.
2025	 * The IN6_IFF_NOTREADY case would be VERY rare, but is possible
2026	 * (for example) when we encounter an error while forwarding procedure
2027	 * destined to a duplicated address of ours.
2028	 */
2029	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2030		if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2031		    (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2032			src = &t;
2033			break;
2034		}
2035	if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2036		/*
2037		 * This is the case if the dst is our link-local address
2038		 * and the sender is also ourseleves.
2039		 */
2040		src = &t;
2041	}
2042
2043	if (src == 0)
2044		/*
2045		 * This case matches to multicasts, our anycast, or unicasts
2046		 * that we do not own. Select a source address which has the
2047		 * same scope.
2048		 * XXX: for (non link-local) multicast addresses, this might
2049		 * not be a good choice.
2050		 */
2051		if ((ia = in6_ifawithscope(m->m_pkthdr.rcvif, &t)) != 0)
2052			src = &IA6_SIN6(ia)->sin6_addr;
2053
2054	if (src == 0)
2055		goto bad;
2056
2057	ip6->ip6_src = *src;
2058
2059	ip6->ip6_flow = 0;
2060	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2061	ip6->ip6_vfc |= IPV6_VERSION;
2062	ip6->ip6_nxt = IPPROTO_ICMPV6;
2063	if (m->m_pkthdr.rcvif) {
2064		/* XXX: This may not be the outgoing interface */
2065		ip6->ip6_hlim = nd_ifinfo[m->m_pkthdr.rcvif->if_index].chlim;
2066	}
2067
2068	icmp6->icmp6_cksum = 0;
2069	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2070					sizeof(struct ip6_hdr), plen);
2071
2072	/*
2073	 * xxx option handling
2074	 */
2075
2076	m->m_flags &= ~(M_BCAST|M_MCAST);
2077#ifdef IPSEC
2078	/* Don't lookup socket */
2079	ipsec_setsocket(m, NULL);
2080#endif /*IPSEC*/
2081
2082#ifdef COMPAT_RFC1885
2083	ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif);
2084#else
2085	ip6_output(m, NULL, NULL, 0, NULL, &outif);
2086#endif
2087	if (outif)
2088		icmp6_ifoutstat_inc(outif, type, code);
2089
2090	return;
2091
2092 bad:
2093	m_freem(m);
2094	return;
2095}
2096
2097void
2098icmp6_fasttimo()
2099{
2100
2101	mld6_fasttimeo();
2102}
2103
2104static const char *
2105icmp6_redirect_diag(src6, dst6, tgt6)
2106	struct in6_addr *src6;
2107	struct in6_addr *dst6;
2108	struct in6_addr *tgt6;
2109{
2110	static char buf[1024];
2111	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2112		ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2113	return buf;
2114}
2115
2116void
2117icmp6_redirect_input(m, off)
2118	register struct mbuf *m;
2119	int off;
2120{
2121	struct ifnet *ifp = m->m_pkthdr.rcvif;
2122	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2123	struct nd_redirect *nd_rd;
2124	int icmp6len = ntohs(ip6->ip6_plen);
2125	char *lladdr = NULL;
2126	int lladdrlen = 0;
2127	u_char *redirhdr = NULL;
2128	int redirhdrlen = 0;
2129	struct rtentry *rt = NULL;
2130	int is_router;
2131	int is_onlink;
2132	struct in6_addr src6 = ip6->ip6_src;
2133	struct in6_addr redtgt6;
2134	struct in6_addr reddst6;
2135	union nd_opts ndopts;
2136
2137	if (!m || !ifp)
2138		return;
2139
2140	/* XXX if we are router, we don't update route by icmp6 redirect */
2141	if (ip6_forwarding)
2142		goto freeit;
2143	if (!icmp6_rediraccept)
2144		goto freeit;
2145
2146#ifndef PULLDOWN_TEST
2147	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2148	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2149#else
2150	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2151	if (nd_rd == NULL) {
2152		icmp6stat.icp6s_tooshort++;
2153		return;
2154	}
2155#endif
2156	redtgt6 = nd_rd->nd_rd_target;
2157	reddst6 = nd_rd->nd_rd_dst;
2158
2159	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2160		redtgt6.s6_addr16[1] = htons(ifp->if_index);
2161	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2162		reddst6.s6_addr16[1] = htons(ifp->if_index);
2163
2164	/* validation */
2165	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2166		log(LOG_ERR,
2167			"ICMP6 redirect sent from %s rejected; "
2168			"must be from linklocal\n", ip6_sprintf(&src6));
2169		goto freeit;
2170	}
2171	if (ip6->ip6_hlim != 255) {
2172		log(LOG_ERR,
2173			"ICMP6 redirect sent from %s rejected; "
2174			"hlim=%d (must be 255)\n",
2175			ip6_sprintf(&src6), ip6->ip6_hlim);
2176		goto freeit;
2177	}
2178    {
2179	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2180	struct sockaddr_in6 sin6;
2181	struct in6_addr *gw6;
2182
2183	bzero(&sin6, sizeof(sin6));
2184	sin6.sin6_family = AF_INET6;
2185	sin6.sin6_len = sizeof(struct sockaddr_in6);
2186	bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2187	rt = rtalloc1((struct sockaddr *)&sin6, 0);
2188	if (rt) {
2189		if (rt->rt_gateway == NULL ||
2190		    rt->rt_gateway->sa_family != AF_INET6) {
2191			log(LOG_ERR,
2192			    "ICMP6 redirect rejected; no route "
2193			    "with inet6 gateway found for redirect dst: %s\n",
2194			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2195			RTFREE(rt);
2196			goto freeit;
2197		}
2198
2199		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2200		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2201			log(LOG_ERR,
2202				"ICMP6 redirect rejected; "
2203				"not equal to gw-for-src=%s (must be same): "
2204				"%s\n",
2205				ip6_sprintf(gw6),
2206				icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2207			RTFREE(rt);
2208			goto freeit;
2209		}
2210	} else {
2211		log(LOG_ERR,
2212			"ICMP6 redirect rejected; "
2213			"no route found for redirect dst: %s\n",
2214			icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2215		goto freeit;
2216	}
2217	RTFREE(rt);
2218	rt = NULL;
2219    }
2220	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2221		log(LOG_ERR,
2222			"ICMP6 redirect rejected; "
2223			"redirect dst must be unicast: %s\n",
2224			icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2225		goto freeit;
2226	}
2227
2228	is_router = is_onlink = 0;
2229	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2230		is_router = 1;	/* router case */
2231	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2232		is_onlink = 1;	/* on-link destination case */
2233	if (!is_router && !is_onlink) {
2234		log(LOG_ERR,
2235			"ICMP6 redirect rejected; "
2236			"neither router case nor onlink case: %s\n",
2237			icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2238		goto freeit;
2239	}
2240	/* validation passed */
2241
2242	icmp6len -= sizeof(*nd_rd);
2243	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2244	if (nd6_options(&ndopts) < 0) {
2245		log(LOG_INFO, "icmp6_redirect_input: "
2246			"invalid ND option, rejected: %s\n",
2247			icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2248		goto freeit;
2249	}
2250
2251	if (ndopts.nd_opts_tgt_lladdr) {
2252		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2253		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2254	}
2255
2256	if (ndopts.nd_opts_rh) {
2257		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2258		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2259	}
2260
2261	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2262		log(LOG_INFO,
2263			"icmp6_redirect_input: lladdrlen mismatch for %s "
2264			"(if %d, icmp6 packet %d): %s\n",
2265			ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2266			icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2267	}
2268
2269	/* RFC 2461 8.3 */
2270	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2271			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2272
2273	if (!is_onlink) {	/* better router case. perform rtredirect. */
2274		/* perform rtredirect */
2275		struct sockaddr_in6 sdst;
2276		struct sockaddr_in6 sgw;
2277		struct sockaddr_in6 ssrc;
2278
2279		bzero(&sdst, sizeof(sdst));
2280		bzero(&sgw, sizeof(sgw));
2281		bzero(&ssrc, sizeof(ssrc));
2282		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2283		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2284			sizeof(struct sockaddr_in6);
2285		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2286		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2287		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2288		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2289			   (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2290			   (struct sockaddr *)&ssrc,
2291			   (struct rtentry **)NULL
2292			   );
2293	}
2294	/* finally update cached route in each socket via pfctlinput */
2295    {
2296	struct sockaddr_in6 sdst;
2297#if 1
2298#else
2299	struct ip6protosw *pr;
2300#endif
2301
2302	bzero(&sdst, sizeof(sdst));
2303	sdst.sin6_family = AF_INET6;
2304	sdst.sin6_len = sizeof(struct sockaddr_in6);
2305	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2306#if 1
2307	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2308#else
2309	/*
2310	 * do not use pfctlinput() here, we have different prototype for
2311	 * xx_ctlinput() in ip6proto.
2312	 */
2313	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
2314	     pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW;
2315	     pr++) {
2316		if (pr->pr_ctlinput) {
2317			(*pr->pr_ctlinput)(PRC_REDIRECT_HOST,
2318				(struct sockaddr *)&sdst, NULL, NULL, 0);
2319		}
2320	}
2321#endif
2322#ifdef IPSEC
2323	key_sa_routechange((struct sockaddr *)&sdst);
2324#endif
2325    }
2326
2327 freeit:
2328	m_freem(m);
2329}
2330
2331void
2332icmp6_redirect_output(m0, rt)
2333	struct mbuf *m0;
2334	struct rtentry *rt;
2335{
2336	struct ifnet *ifp;	/* my outgoing interface */
2337	struct in6_addr *ifp_ll6;
2338	struct in6_addr *router_ll6;
2339	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2340	struct mbuf *m = NULL;	/* newly allocated one */
2341	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2342	struct nd_redirect *nd_rd;
2343	size_t maxlen;
2344	u_char *p;
2345	struct ifnet *outif = NULL;
2346	struct sockaddr_in6 src_sa;
2347
2348	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2349
2350	/* if we are not router, we don't send icmp6 redirect */
2351	if (!ip6_forwarding || ip6_accept_rtadv)
2352		goto fail;
2353
2354	/* sanity check */
2355	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2356		goto fail;
2357
2358	/*
2359	 * Address check:
2360	 *  the source address must identify a neighbor, and
2361	 *  the destination address must not be a multicast address
2362	 *  [RFC 2461, sec 8.2]
2363	 */
2364	sip6 = mtod(m0, struct ip6_hdr *);
2365	bzero(&src_sa, sizeof(src_sa));
2366	src_sa.sin6_family = AF_INET6;
2367	src_sa.sin6_len = sizeof(src_sa);
2368	src_sa.sin6_addr = sip6->ip6_src;
2369	/* we don't currently use sin6_scope_id, but eventually use it */
2370	src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src);
2371	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2372		goto fail;
2373	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2374		goto fail;	/* what should we do here? */
2375
2376	/* rate limit */
2377	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2378		goto fail;
2379
2380	/*
2381	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2382	 * we almost always ask for an mbuf cluster for simplicity.
2383	 * (MHLEN < IPV6_MMTU is almost always true)
2384	 */
2385#if IPV6_MMTU >= MCLBYTES
2386# error assumption failed about IPV6_MMTU and MCLBYTES
2387#endif
2388	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2389	if (m && IPV6_MMTU >= MHLEN)
2390		MCLGET(m, M_DONTWAIT);
2391	if (!m)
2392		goto fail;
2393	m->m_len = 0;
2394	maxlen = M_TRAILINGSPACE(m);
2395	maxlen = min(IPV6_MMTU, maxlen);
2396	/* just for safety */
2397	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2398	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2399		goto fail;
2400	}
2401
2402	{
2403		/* get ip6 linklocal address for ifp(my outgoing interface). */
2404		struct in6_ifaddr *ia;
2405		if ((ia = in6ifa_ifpforlinklocal(ifp,
2406						 IN6_IFF_NOTREADY|
2407						 IN6_IFF_ANYCAST)) == NULL)
2408			goto fail;
2409		ifp_ll6 = &ia->ia_addr.sin6_addr;
2410	}
2411
2412	/* get ip6 linklocal address for the router. */
2413	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2414		struct sockaddr_in6 *sin6;
2415		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2416		router_ll6 = &sin6->sin6_addr;
2417		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2418			router_ll6 = (struct in6_addr *)NULL;
2419	} else
2420		router_ll6 = (struct in6_addr *)NULL;
2421
2422	/* ip6 */
2423	ip6 = mtod(m, struct ip6_hdr *);
2424	ip6->ip6_flow = 0;
2425	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2426	ip6->ip6_vfc |= IPV6_VERSION;
2427	/* ip6->ip6_plen will be set later */
2428	ip6->ip6_nxt = IPPROTO_ICMPV6;
2429	ip6->ip6_hlim = 255;
2430	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2431	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2432	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2433
2434	/* ND Redirect */
2435	nd_rd = (struct nd_redirect *)(ip6 + 1);
2436	nd_rd->nd_rd_type = ND_REDIRECT;
2437	nd_rd->nd_rd_code = 0;
2438	nd_rd->nd_rd_reserved = 0;
2439	if (rt->rt_flags & RTF_GATEWAY) {
2440		/*
2441		 * nd_rd->nd_rd_target must be a link-local address in
2442		 * better router cases.
2443		 */
2444		if (!router_ll6)
2445			goto fail;
2446		bcopy(router_ll6, &nd_rd->nd_rd_target,
2447		      sizeof(nd_rd->nd_rd_target));
2448		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2449		      sizeof(nd_rd->nd_rd_dst));
2450	} else {
2451		/* make sure redtgt == reddst */
2452		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2453		      sizeof(nd_rd->nd_rd_target));
2454		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2455		      sizeof(nd_rd->nd_rd_dst));
2456	}
2457
2458	p = (u_char *)(nd_rd + 1);
2459
2460	if (!router_ll6)
2461		goto nolladdropt;
2462
2463    {
2464	/* target lladdr option */
2465	struct rtentry *rt_router = NULL;
2466	int len;
2467	struct sockaddr_dl *sdl;
2468	struct nd_opt_hdr *nd_opt;
2469	char *lladdr;
2470
2471	rt_router = nd6_lookup(router_ll6, 0, ifp);
2472	if (!rt_router)
2473		goto nolladdropt;
2474	len = sizeof(*nd_opt) + ifp->if_addrlen;
2475	len = (len + 7) & ~7;	/*round by 8*/
2476	/* safety check */
2477	if (len + (p - (u_char *)ip6) > maxlen)
2478		goto nolladdropt;
2479	if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2480	    (rt_router->rt_flags & RTF_LLINFO) &&
2481	    (rt_router->rt_gateway->sa_family == AF_LINK) &&
2482	    (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2483	    sdl->sdl_alen) {
2484		nd_opt = (struct nd_opt_hdr *)p;
2485		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2486		nd_opt->nd_opt_len = len >> 3;
2487		lladdr = (char *)(nd_opt + 1);
2488		bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2489		p += len;
2490	}
2491    }
2492nolladdropt:;
2493
2494	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2495
2496	/* just to be safe */
2497	if (m0->m_flags & M_DECRYPTED)
2498		goto noredhdropt;
2499	if (p - (u_char *)ip6 > maxlen)
2500		goto noredhdropt;
2501
2502    {
2503	/* redirected header option */
2504	int len;
2505	struct nd_opt_rd_hdr *nd_opt_rh;
2506
2507	/*
2508	 * compute the maximum size for icmp6 redirect header option.
2509	 * XXX room for auth header?
2510	 */
2511	len = maxlen - (p - (u_char *)ip6);
2512	len &= ~7;
2513
2514	/* This is just for simplicity. */
2515	if (m0->m_pkthdr.len != m0->m_len) {
2516		if (m0->m_next) {
2517			m_freem(m0->m_next);
2518			m0->m_next = NULL;
2519		}
2520		m0->m_pkthdr.len = m0->m_len;
2521	}
2522
2523	/*
2524	 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2525	 * about padding/truncate rule for the original IP packet.
2526	 * From the discussion on IPv6imp in Feb 1999, the consensus was:
2527	 * - "attach as much as possible" is the goal
2528	 * - pad if not aligned (original size can be guessed by original
2529	 *   ip6 header)
2530	 * Following code adds the padding if it is simple enough,
2531	 * and truncates if not.
2532	 */
2533	if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2534		panic("assumption failed in %s:%d\n", __FILE__, __LINE__);
2535
2536	if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2537		/* not enough room, truncate */
2538		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2539	} else {
2540		/* enough room, pad or truncate */
2541		size_t extra;
2542
2543		extra = m0->m_pkthdr.len % 8;
2544		if (extra) {
2545			/* pad if easy enough, truncate if not */
2546			if (8 - extra <= M_TRAILINGSPACE(m0)) {
2547				/* pad */
2548				m0->m_len += (8 - extra);
2549				m0->m_pkthdr.len += (8 - extra);
2550			} else {
2551				/* truncate */
2552				m0->m_pkthdr.len -= extra;
2553				m0->m_len -= extra;
2554			}
2555		}
2556		len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2557		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2558	}
2559
2560	nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2561	bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2562	nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2563	nd_opt_rh->nd_opt_rh_len = len >> 3;
2564	p += sizeof(*nd_opt_rh);
2565	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2566
2567	/* connect m0 to m */
2568	m->m_next = m0;
2569	m->m_pkthdr.len = m->m_len + m0->m_len;
2570    }
2571noredhdropt:;
2572
2573	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
2574		sip6->ip6_src.s6_addr16[1] = 0;
2575	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
2576		sip6->ip6_dst.s6_addr16[1] = 0;
2577#if 0
2578	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
2579		ip6->ip6_src.s6_addr16[1] = 0;
2580	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
2581		ip6->ip6_dst.s6_addr16[1] = 0;
2582#endif
2583	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
2584		nd_rd->nd_rd_target.s6_addr16[1] = 0;
2585	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
2586		nd_rd->nd_rd_dst.s6_addr16[1] = 0;
2587
2588	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2589
2590	nd_rd->nd_rd_cksum = 0;
2591	nd_rd->nd_rd_cksum
2592		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2593
2594	/* send the packet to outside... */
2595#ifdef IPSEC
2596	/* Don't lookup socket */
2597	ipsec_setsocket(m, NULL);
2598#endif /*IPSEC*/
2599	ip6_output(m, NULL, NULL, 0, NULL, &outif);
2600	if (outif) {
2601		icmp6_ifstat_inc(outif, ifs6_out_msg);
2602		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2603	}
2604	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2605
2606	return;
2607
2608fail:
2609	if (m)
2610		m_freem(m);
2611	if (m0)
2612		m_freem(m0);
2613}
2614
2615/*
2616 * ICMPv6 socket option processing.
2617 */
2618int
2619icmp6_ctloutput(op, so, level, optname, mp)
2620	int op;
2621	struct socket *so;
2622	int level, optname;
2623	struct mbuf **mp;
2624{
2625	register struct in6pcb *in6p = sotoin6pcb(so);
2626	register struct mbuf *m = *mp;
2627	int error = 0;
2628
2629	if (level != IPPROTO_ICMPV6) {
2630		error = EINVAL;
2631		if (op == PRCO_SETOPT && m)
2632			(void)m_free(m);
2633	} else switch(op) {
2634	 case PRCO_SETOPT:
2635		 switch (optname) {
2636		  case ICMP6_FILTER:
2637		  {
2638			  struct icmp6_filter *p;
2639
2640			  p = mtod(m, struct icmp6_filter *);
2641			  if (!p || !in6p->in6p_icmp6filt) {
2642				  error = EINVAL;
2643				  break;
2644			  }
2645			  bcopy(p, in6p->in6p_icmp6filt,
2646				sizeof(struct icmp6_filter));
2647			  error = 0;
2648			  break;
2649		  }
2650
2651		  default:
2652			  error = ENOPROTOOPT;
2653			  break;
2654		 }
2655		 if (m)
2656			 (void)m_free(m);
2657		 break;
2658
2659	 case PRCO_GETOPT:
2660		 switch (optname) {
2661		  case ICMP6_FILTER:
2662		  {
2663			  struct icmp6_filter *p;
2664
2665			  if (!in6p->in6p_icmp6filt) {
2666				  error = EINVAL;
2667				  break;
2668			  }
2669			  *mp = m = m_get(M_WAIT, MT_SOOPTS);
2670			  m->m_len = sizeof(struct icmp6_filter);
2671			  p = mtod(m, struct icmp6_filter *);
2672			  bcopy(in6p->in6p_icmp6filt, p,
2673				sizeof(struct icmp6_filter));
2674			  error = 0;
2675			  break;
2676		  }
2677
2678		  default:
2679			  error = ENOPROTOOPT;
2680			  break;
2681		 }
2682		 break;
2683	}
2684
2685	return(error);
2686}
2687
2688/*
2689 * Perform rate limit check.
2690 * Returns 0 if it is okay to send the icmp6 packet.
2691 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2692 * limitation.
2693 *
2694 * XXX per-destination/type check necessary?
2695 */
2696static int
2697icmp6_ratelimit(dst, type, code)
2698	const struct in6_addr *dst;	/* not used at this moment */
2699	const int type;			/* not used at this moment */
2700	const int code;			/* not used at this moment */
2701{
2702	int ret;
2703
2704	ret = 0;	/*okay to send*/
2705
2706	/* PPS limit */
2707	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2708	    icmp6errppslim)) {
2709		/* The packet is subject to rate limit */
2710		ret++;
2711	}
2712
2713	return ret;
2714}
2715
2716static struct rtentry *
2717icmp6_mtudisc_clone(dst)
2718	struct sockaddr *dst;
2719{
2720	struct rtentry *rt;
2721	int    error;
2722
2723	rt = rtalloc1(dst, 1);
2724	if (rt == 0)
2725		return NULL;
2726
2727	/* If we didn't get a host route, allocate one */
2728	if ((rt->rt_flags & RTF_HOST) == 0) {
2729		struct rtentry *nrt;
2730
2731		error = rtrequest((int) RTM_ADD, dst,
2732		    (struct sockaddr *) rt->rt_gateway,
2733		    (struct sockaddr *) 0,
2734		    RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt);
2735		if (error) {
2736			rtfree(rt);
2737			rtfree(nrt);
2738			return NULL;
2739		}
2740		nrt->rt_rmx = rt->rt_rmx;
2741		rtfree(rt);
2742		rt = nrt;
2743	}
2744	error = rt_timer_add(rt, icmp6_mtudisc_timeout,
2745			icmp6_mtudisc_timeout_q);
2746	if (error) {
2747		rtfree(rt);
2748		return NULL;
2749	}
2750
2751	return rt;	/* caller need to call rtfree() */
2752}
2753
2754static void
2755icmp6_mtudisc_timeout(rt, r)
2756	struct rtentry *rt;
2757	struct rttimer *r;
2758{
2759	if (rt == NULL)
2760		panic("icmp6_mtudisc_timeout: bad route to timeout");
2761	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
2762	    (RTF_DYNAMIC | RTF_HOST)) {
2763		rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt),
2764		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
2765	} else {
2766		if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) {
2767			rt->rt_rmx.rmx_mtu = 0;
2768		}
2769	}
2770}
2771
2772#include <uvm/uvm_extern.h>
2773#include <sys/sysctl.h>
2774
2775int
2776icmp6_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
2777	int *name;
2778	u_int namelen;
2779	void *oldp;
2780	size_t *oldlenp;
2781	void *newp;
2782	size_t newlen;
2783{
2784
2785	/* All sysctl names at this level are terminal. */
2786	if (namelen != 1)
2787		return ENOTDIR;
2788
2789	switch (name[0]) {
2790
2791	case ICMPV6CTL_REDIRACCEPT:
2792		return sysctl_int(oldp, oldlenp, newp, newlen,
2793				&icmp6_rediraccept);
2794	case ICMPV6CTL_REDIRTIMEOUT:
2795		return sysctl_int(oldp, oldlenp, newp, newlen,
2796				&icmp6_redirtimeout);
2797	case ICMPV6CTL_STATS:
2798		return sysctl_rdstruct(oldp, oldlenp, newp,
2799				&icmp6stat, sizeof(icmp6stat));
2800	case ICMPV6CTL_ND6_PRUNE:
2801		return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_prune);
2802	case ICMPV6CTL_ND6_DELAY:
2803		return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_delay);
2804	case ICMPV6CTL_ND6_UMAXTRIES:
2805		return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_umaxtries);
2806	case ICMPV6CTL_ND6_MMAXTRIES:
2807		return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_mmaxtries);
2808	case ICMPV6CTL_ND6_USELOOPBACK:
2809		return sysctl_int(oldp, oldlenp, newp, newlen,
2810				&nd6_useloopback);
2811	case ICMPV6CTL_NODEINFO:
2812		return sysctl_int(oldp, oldlenp, newp, newlen, &icmp6_nodeinfo);
2813	case ICMPV6CTL_ERRPPSLIMIT:
2814		return sysctl_int(oldp, oldlenp, newp, newlen, &icmp6errppslim);
2815	case ICMPV6CTL_ND6_MAXNUDHINT:
2816		return sysctl_int(oldp, oldlenp, newp, newlen,
2817				&nd6_maxnudhint);
2818	case ICMPV6CTL_MTUDISC_HIWAT:
2819		return sysctl_int(oldp, oldlenp, newp, newlen,
2820				&icmp6_mtudisc_hiwat);
2821	case ICMPV6CTL_MTUDISC_LOWAT:
2822		return sysctl_int(oldp, oldlenp, newp, newlen,
2823				&icmp6_mtudisc_lowat);
2824	default:
2825		return ENOPROTOOPT;
2826	}
2827	/* NOTREACHED */
2828}
2829