icmp6.c revision 1.50
1/*	$NetBSD: icmp6.c,v 1.50 2001/01/08 06:12:46 itojun Exp $	*/
2/*	$KAME: icmp6.c,v 1.172 2000/12/11 19:27:06 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#ifdef ND6DEBUG
430		log(LOG_ERR,
431		    "ICMP6 checksum error(%d|%x) %s\n",
432		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src));
433#endif
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 = icmp6_mtudisc_clone((struct sockaddr *)&sin6);
1087
1088	if (rt && (rt->rt_flags & RTF_HOST)
1089	    && !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1090		if (mtu < IPV6_MMTU) {
1091				/* xxx */
1092			rt->rt_rmx.rmx_locks |= RTV_MTU;
1093		} else if (mtu < rt->rt_ifp->if_mtu &&
1094			   rt->rt_rmx.rmx_mtu > mtu) {
1095			icmp6stat.icp6s_pmtuchg++;
1096			rt->rt_rmx.rmx_mtu = mtu;
1097		}
1098	}
1099	if (rt)
1100		RTFREE(rt);
1101
1102	/*
1103	 * Notify protocols that the MTU for this destination
1104	 * has changed.
1105	 */
1106	for (mc = LIST_FIRST(&icmp6_mtudisc_callbacks); mc != NULL;
1107	     mc = LIST_NEXT(mc, mc_list))
1108		(*mc->mc_func)(&sin6.sin6_addr);
1109}
1110
1111/*
1112 * Process a Node Information Query packet, based on
1113 * draft-ietf-ipngwg-icmp-name-lookups-07.
1114 *
1115 * Spec incompatibilities:
1116 * - IPv6 Subject address handling
1117 * - IPv4 Subject address handling support missing
1118 * - Proxy reply (answer even if it's not for me)
1119 * - joins NI group address at in6_ifattach() time only, does not cope
1120 *   with hostname changes by sethostname(3)
1121 */
1122#ifndef offsetof		/* XXX */
1123#define	offsetof(type, member)	((size_t)(&((type *)0)->member))
1124#endif
1125static struct mbuf *
1126ni6_input(m, off)
1127	struct mbuf *m;
1128	int off;
1129{
1130	struct icmp6_nodeinfo *ni6, *nni6;
1131	struct mbuf *n = NULL;
1132	u_int16_t qtype;
1133	int subjlen;
1134	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1135	struct ni_reply_fqdn *fqdn;
1136	int addrs;		/* for NI_QTYPE_NODEADDR */
1137	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1138	struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1139	struct ip6_hdr *ip6;
1140	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1141	char *subj = NULL;
1142
1143	ip6 = mtod(m, struct ip6_hdr *);
1144#ifndef PULLDOWN_TEST
1145	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1146#else
1147	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1148	if (ni6 == NULL) {
1149		/* m is already reclaimed */
1150		return NULL;
1151	}
1152#endif
1153
1154	/*
1155	 * Validate IPv6 destination address.
1156	 *
1157	 * The Responder must discard the Query without further processing
1158	 * unless it is one of the Responder's unicast or anycast addresses, or
1159	 * a link-local scope multicast address which the Responder has joined.
1160	 * [icmp-name-lookups-07, Section 4.]
1161	 */
1162	bzero(&sin6, sizeof(sin6));
1163	sin6.sin6_family = AF_INET6;
1164	sin6.sin6_len = sizeof(struct sockaddr_in6);
1165	bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1166	/* XXX scopeid */
1167	if (ifa_ifwithaddr((struct sockaddr *)&sin6))
1168		; /* unicast/anycast, fine */
1169	else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1170		; /* link-local multicast, fine */
1171	else
1172		goto bad;
1173
1174	/* validate query Subject field. */
1175	qtype = ntohs(ni6->ni_qtype);
1176	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1177	switch (qtype) {
1178	case NI_QTYPE_NOOP:
1179	case NI_QTYPE_SUPTYPES:
1180		/* 07 draft */
1181		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1182			break;
1183		/*FALLTHROUGH*/
1184	case NI_QTYPE_FQDN:
1185	case NI_QTYPE_NODEADDR:
1186		switch (ni6->ni_code) {
1187		case ICMP6_NI_SUBJ_IPV6:
1188#if ICMP6_NI_SUBJ_IPV6 != 0
1189		case 0:
1190#endif
1191			/*
1192			 * backward compatibility - try to accept 03 draft
1193			 * format, where no Subject is present.
1194			 */
1195			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1196			    subjlen == 0) {
1197				oldfqdn++;
1198				break;
1199			}
1200#if ICMP6_NI_SUBJ_IPV6 != 0
1201			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1202				goto bad;
1203#endif
1204
1205			if (subjlen != sizeof(sin6.sin6_addr))
1206				goto bad;
1207
1208			/*
1209			 * Validate Subject address.
1210			 *
1211			 * Not sure what exactly does "address belongs to the
1212			 * node" mean in the spec, is it just unicast, or what?
1213			 *
1214			 * At this moment we consider Subject address as
1215			 * "belong to the node" if the Subject address equals
1216			 * to the IPv6 destination address; validation for
1217			 * IPv6 destination address should have done enough
1218			 * check for us.
1219			 *
1220			 * We do not do proxy at this moment.
1221			 */
1222			/* m_pulldown instead of copy? */
1223			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1224			    subjlen, (caddr_t)&sin6.sin6_addr);
1225			/* XXX kame scope hack */
1226			if (IN6_IS_SCOPE_LINKLOCAL(&sin6.sin6_addr)) {
1227#ifdef FAKE_LOOPBACK_IF
1228				if ((m->m_flags & M_PKTHDR) != 0 &&
1229				    m->m_pkthdr.rcvif) {
1230					sin6.sin6_addr.s6_addr16[1] =
1231					    htons(m->m_pkthdr.rcvif->if_index);
1232				}
1233#else
1234				if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) {
1235					sin6.sin6_addr.s6_addr16[1] =
1236					    ip6->ip6_dst.s6_addr16[1];
1237				}
1238#endif
1239			}
1240			subj = (char *)&sin6;
1241			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &sin6.sin6_addr))
1242				break;
1243
1244			/*
1245			 * XXX if we are to allow other cases, we should really
1246			 * be careful about scope here.
1247			 * basically, we should disallow queries toward IPv6
1248			 * destination X with subject Y, if scope(X) > scope(Y).
1249			 * if we allow scope(X) > scope(Y), it will result in
1250			 * information leakage across scope boundary.
1251			 */
1252			goto bad;
1253
1254		case ICMP6_NI_SUBJ_FQDN:
1255			/*
1256			 * Validate Subject name with gethostname(3).
1257			 *
1258			 * The behavior may need some debate, since:
1259			 * - we are not sure if the node has FQDN as
1260			 *   hostname (returned by gethostname(3)).
1261			 * - the code does wildcard match for truncated names.
1262			 *   however, we are not sure if we want to perform
1263			 *   wildcard match, if gethostname(3) side has
1264			 *   truncated hostname.
1265			 */
1266			n = ni6_nametodns(hostname, hostnamelen, 0);
1267			if (!n || n->m_next || n->m_len == 0)
1268				goto bad;
1269			IP6_EXTHDR_GET(subj, char *, m,
1270			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1271			if (subj == NULL)
1272				goto bad;
1273			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1274					n->m_len)) {
1275				goto bad;
1276			}
1277			m_freem(n);
1278			n = NULL;
1279			break;
1280
1281		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1282		default:
1283			goto bad;
1284		}
1285		break;
1286	}
1287
1288	/* guess reply length */
1289	switch (qtype) {
1290	case NI_QTYPE_NOOP:
1291		break;		/* no reply data */
1292	case NI_QTYPE_SUPTYPES:
1293		replylen += sizeof(u_int32_t);
1294		break;
1295	case NI_QTYPE_FQDN:
1296		/* XXX will append an mbuf */
1297		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1298		break;
1299	case NI_QTYPE_NODEADDR:
1300		addrs = ni6_addrs(ni6, m, &ifp, subj);
1301		if ((replylen += addrs * (sizeof(struct in6_addr) +
1302					  sizeof(u_int32_t))) > MCLBYTES)
1303			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1304		break;
1305	default:
1306		/*
1307		 * XXX: We must return a reply with the ICMP6 code
1308		 * `unknown Qtype' in this case. However we regard the case
1309		 * as an FQDN query for backward compatibility.
1310		 * Older versions set a random value to this field,
1311		 * so it rarely varies in the defined qtypes.
1312		 * But the mechanism is not reliable...
1313		 * maybe we should obsolete older versions.
1314		 */
1315		qtype = NI_QTYPE_FQDN;
1316		/* XXX will append an mbuf */
1317		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1318		oldfqdn++;
1319		break;
1320	}
1321
1322	/* allocate an mbuf to reply. */
1323	MGETHDR(n, M_DONTWAIT, m->m_type);
1324	if (n == NULL) {
1325		m_freem(m);
1326		return(NULL);
1327	}
1328	M_COPY_PKTHDR(n, m); /* just for recvif */
1329	if (replylen > MHLEN) {
1330		if (replylen > MCLBYTES) {
1331			 /*
1332			  * XXX: should we try to allocate more? But MCLBYTES
1333			  * is probably much larger than IPV6_MMTU...
1334			  */
1335			goto bad;
1336		}
1337		MCLGET(n, M_DONTWAIT);
1338		if ((n->m_flags & M_EXT) == 0) {
1339			goto bad;
1340		}
1341	}
1342	n->m_pkthdr.len = n->m_len = replylen;
1343
1344	/* copy mbuf header and IPv6 + Node Information base headers */
1345	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1346	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1347	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1348
1349	/* qtype dependent procedure */
1350	switch (qtype) {
1351	case NI_QTYPE_NOOP:
1352		nni6->ni_code = ICMP6_NI_SUCCESS;
1353		nni6->ni_flags = 0;
1354		break;
1355	case NI_QTYPE_SUPTYPES:
1356	{
1357		u_int32_t v;
1358		nni6->ni_code = ICMP6_NI_SUCCESS;
1359		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1360		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1361		v = (u_int32_t)htonl(0x0000000f);
1362		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1363		break;
1364	}
1365	case NI_QTYPE_FQDN:
1366		nni6->ni_code = ICMP6_NI_SUCCESS;
1367		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1368						sizeof(struct ip6_hdr) +
1369						sizeof(struct icmp6_nodeinfo));
1370		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1371		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1372		/*
1373		 * XXX do we really have FQDN in variable "hostname"?
1374		 */
1375		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1376		if (n->m_next == NULL)
1377			goto bad;
1378		/* XXX we assume that n->m_next is not a chain */
1379		if (n->m_next->m_next != NULL)
1380			goto bad;
1381		n->m_pkthdr.len += n->m_next->m_len;
1382		break;
1383	case NI_QTYPE_NODEADDR:
1384	{
1385		int lenlim, copied;
1386
1387		nni6->ni_code = ICMP6_NI_SUCCESS;
1388		n->m_pkthdr.len = n->m_len =
1389		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1390		lenlim = M_TRAILINGSPACE(n);
1391		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1392		/* XXX: reset mbuf length */
1393		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1394			sizeof(struct icmp6_nodeinfo) + copied;
1395		break;
1396	}
1397	default:
1398		break;		/* XXX impossible! */
1399	}
1400
1401	nni6->ni_type = ICMP6_NI_REPLY;
1402	m_freem(m);
1403	return(n);
1404
1405  bad:
1406	m_freem(m);
1407	if (n)
1408		m_freem(n);
1409	return(NULL);
1410}
1411#undef hostnamelen
1412
1413/*
1414 * make a mbuf with DNS-encoded string.  no compression support.
1415 *
1416 * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1417 * treated as truncated name (two \0 at the end).  this is a wild guess.
1418 */
1419static struct mbuf *
1420ni6_nametodns(name, namelen, old)
1421	const char *name;
1422	int namelen;
1423	int old;	/* return pascal string if non-zero */
1424{
1425	struct mbuf *m;
1426	char *cp, *ep;
1427	const char *p, *q;
1428	int i, len, nterm;
1429
1430	if (old)
1431		len = namelen + 1;
1432	else
1433		len = MCLBYTES;
1434
1435	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1436	MGET(m, M_DONTWAIT, MT_DATA);
1437	if (m && len > MLEN) {
1438		MCLGET(m, M_DONTWAIT);
1439		if ((m->m_flags & M_EXT) == 0)
1440			goto fail;
1441	}
1442	if (!m)
1443		goto fail;
1444	m->m_next = NULL;
1445
1446	if (old) {
1447		m->m_len = len;
1448		*mtod(m, char *) = namelen;
1449		bcopy(name, mtod(m, char *) + 1, namelen);
1450		return m;
1451	} else {
1452		m->m_len = 0;
1453		cp = mtod(m, char *);
1454		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1455
1456		/* if not certain about my name, return empty buffer */
1457		if (namelen == 0)
1458			return m;
1459
1460		/*
1461		 * guess if it looks like shortened hostname, or FQDN.
1462		 * shortened hostname needs two trailing "\0".
1463		 */
1464		i = 0;
1465		for (p = name; p < name + namelen; p++) {
1466			if (*p && *p == '.')
1467				i++;
1468		}
1469		if (i < 2)
1470			nterm = 2;
1471		else
1472			nterm = 1;
1473
1474		p = name;
1475		while (cp < ep && p < name + namelen) {
1476			i = 0;
1477			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1478				i++;
1479			/* result does not fit into mbuf */
1480			if (cp + i + 1 >= ep)
1481				goto fail;
1482			/* DNS label length restriction, RFC1035 page 8 */
1483			if (i >= 64)
1484				goto fail;
1485			*cp++ = i;
1486			bcopy(p, cp, i);
1487			cp += i;
1488			p = q;
1489			if (p < name + namelen && *p == '.')
1490				p++;
1491		}
1492		/* termination */
1493		if (cp + nterm >= ep)
1494			goto fail;
1495		while (nterm-- > 0)
1496			*cp++ = '\0';
1497		m->m_len = cp - mtod(m, char *);
1498		return m;
1499	}
1500
1501	panic("should not reach here");
1502	/*NOTREACHED*/
1503
1504 fail:
1505	if (m)
1506		m_freem(m);
1507	return NULL;
1508}
1509
1510/*
1511 * check if two DNS-encoded string matches.  takes care of truncated
1512 * form (with \0\0 at the end).  no compression support.
1513 * XXX upper/lowercase match (see RFC2065)
1514 */
1515static int
1516ni6_dnsmatch(a, alen, b, blen)
1517	const char *a;
1518	int alen;
1519	const char *b;
1520	int blen;
1521{
1522	const char *a0, *b0;
1523	int l;
1524
1525	/* simplest case - need validation? */
1526	if (alen == blen && bcmp(a, b, alen) == 0)
1527		return 1;
1528
1529	a0 = a;
1530	b0 = b;
1531
1532	/* termination is mandatory */
1533	if (alen < 2 || blen < 2)
1534		return 0;
1535	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1536		return 0;
1537	alen--;
1538	blen--;
1539
1540	while (a - a0 < alen && b - b0 < blen) {
1541		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1542			return 0;
1543
1544		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1545			return 0;
1546		/* we don't support compression yet */
1547		if (a[0] >= 64 || b[0] >= 64)
1548			return 0;
1549
1550		/* truncated case */
1551		if (a[0] == 0 && a - a0 == alen - 1)
1552			return 1;
1553		if (b[0] == 0 && b - b0 == blen - 1)
1554			return 1;
1555		if (a[0] == 0 || b[0] == 0)
1556			return 0;
1557
1558		if (a[0] != b[0])
1559			return 0;
1560		l = a[0];
1561		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1562			return 0;
1563		if (bcmp(a + 1, b + 1, l) != 0)
1564			return 0;
1565
1566		a += 1 + l;
1567		b += 1 + l;
1568	}
1569
1570	if (a - a0 == alen && b - b0 == blen)
1571		return 1;
1572	else
1573		return 0;
1574}
1575
1576/*
1577 * calculate the number of addresses to be returned in the node info reply.
1578 */
1579static int
1580ni6_addrs(ni6, m, ifpp, subj)
1581	struct icmp6_nodeinfo *ni6;
1582	struct mbuf *m;
1583	struct ifnet **ifpp;
1584	char *subj;
1585{
1586	register struct ifnet *ifp;
1587	register struct in6_ifaddr *ifa6;
1588	register struct ifaddr *ifa;
1589	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1590	int addrs = 0, addrsofif, iffound = 0;
1591	int niflags = ni6->ni_flags;
1592
1593	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1594		switch(ni6->ni_code) {
1595		case ICMP6_NI_SUBJ_IPV6:
1596			if (subj == NULL) /* must be impossible... */
1597				return(0);
1598			subj_ip6 = (struct sockaddr_in6 *)subj;
1599			break;
1600		default:
1601			/*
1602			 * XXX: we only support IPv6 subject address for
1603			 * this Qtype.
1604			 */
1605			return(0);
1606		}
1607	}
1608
1609	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
1610	{
1611		addrsofif = 0;
1612		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1613		     ifa = ifa->ifa_list.tqe_next)
1614		{
1615			if (ifa->ifa_addr->sa_family != AF_INET6)
1616				continue;
1617			ifa6 = (struct in6_ifaddr *)ifa;
1618
1619			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1620			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1621					       &ifa6->ia_addr.sin6_addr))
1622				iffound = 1;
1623
1624			/*
1625			 * IPv4-mapped addresses can only be returned by a
1626			 * Node Information proxy, since they represent
1627			 * addresses of IPv4-only nodes, which perforce do
1628			 * not implement this protocol.
1629			 * [icmp-name-lookups-07, Section 5.4]
1630			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1631			 * this function at this moment.
1632			 */
1633
1634			/* What do we have to do about ::1? */
1635			switch(in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1636			case IPV6_ADDR_SCOPE_LINKLOCAL:
1637				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL)
1638				    == 0)
1639					continue;
1640				break;
1641			case IPV6_ADDR_SCOPE_SITELOCAL:
1642				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL)
1643				    == 0)
1644					continue;
1645				break;
1646			case IPV6_ADDR_SCOPE_GLOBAL:
1647				if ((niflags & NI_NODEADDR_FLAG_GLOBAL)
1648				    == 0)
1649					continue;
1650				break;
1651			default:
1652				continue;
1653			}
1654
1655			/*
1656			 * check if anycast is okay.
1657			 * XXX: just experimental. not in the spec.
1658			 */
1659			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1660			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1661				continue; /* we need only unicast addresses */
1662
1663			addrsofif++; /* count the address */
1664		}
1665		if (iffound) {
1666			*ifpp = ifp;
1667			return(addrsofif);
1668		}
1669
1670		addrs += addrsofif;
1671	}
1672
1673	return(addrs);
1674}
1675
1676static int
1677ni6_store_addrs(ni6, nni6, ifp0, resid)
1678	struct icmp6_nodeinfo *ni6, *nni6;
1679	struct ifnet *ifp0;
1680	int resid;
1681{
1682	register struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1683	register struct in6_ifaddr *ifa6;
1684	register struct ifaddr *ifa;
1685	struct ifnet *ifp_dep = NULL;
1686	int copied = 0, allow_deprecated = 0;
1687	u_char *cp = (u_char *)(nni6 + 1);
1688	int niflags = ni6->ni_flags;
1689	u_int32_t ltime;
1690	long time_second = time.tv_sec;
1691
1692	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1693		return(0);	/* needless to copy */
1694
1695  again:
1696
1697	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list))
1698	{
1699		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1700		     ifa = ifa->ifa_list.tqe_next)
1701		{
1702			if (ifa->ifa_addr->sa_family != AF_INET6)
1703				continue;
1704			ifa6 = (struct in6_ifaddr *)ifa;
1705
1706			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1707			    allow_deprecated == 0) {
1708				/*
1709				 * prefererred address should be put before
1710				 * deprecated addresses.
1711				 */
1712
1713				/* record the interface for later search */
1714				if (ifp_dep == NULL)
1715					ifp_dep = ifp;
1716
1717				continue;
1718			}
1719			else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1720				 allow_deprecated != 0)
1721				continue; /* we now collect deprecated addrs */
1722
1723			/* What do we have to do about ::1? */
1724			switch(in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1725			case IPV6_ADDR_SCOPE_LINKLOCAL:
1726				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL)
1727				    == 0)
1728					continue;
1729				break;
1730			case IPV6_ADDR_SCOPE_SITELOCAL:
1731				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL)
1732				    == 0)
1733					continue;
1734				break;
1735			case IPV6_ADDR_SCOPE_GLOBAL:
1736				if ((niflags & NI_NODEADDR_FLAG_GLOBAL)
1737				    == 0)
1738					continue;
1739				break;
1740			default:
1741				continue;
1742			}
1743
1744			/*
1745			 * check if anycast is okay.
1746			 * XXX: just experimental. not in the spec.
1747			 */
1748			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1749			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1750				continue;
1751
1752			/* now we can copy the address */
1753			if (resid < sizeof(struct in6_addr) +
1754			    sizeof(u_int32_t)) {
1755				/*
1756				 * We give up much more copy.
1757				 * Set the truncate flag and return.
1758				 */
1759				nni6->ni_flags |=
1760					NI_NODEADDR_FLAG_TRUNCATE;
1761				return(copied);
1762			}
1763
1764			/*
1765			 * Set the TTL of the address.
1766			 * The TTL value should be one of the following
1767			 * according to the specification:
1768			 *
1769			 * 1. The remaining lifetime of a DHCP lease on the
1770			 *    address, or
1771			 * 2. The remaining Valid Lifetime of a prefix from
1772			 *    which the address was derived through Stateless
1773			 *    Autoconfiguration.
1774			 *
1775			 * Note that we currently do not support stateful
1776			 * address configuration by DHCPv6, so the former
1777			 * case can't happen.
1778			 */
1779			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1780				ltime = ND6_INFINITE_LIFETIME;
1781			else {
1782				if (ifa6->ia6_lifetime.ia6t_expire >
1783				    time_second)
1784					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1785				else
1786					ltime = 0;
1787			}
1788
1789			bcopy(&ltime, cp, sizeof(u_int32_t));
1790			cp += sizeof(u_int32_t);
1791
1792			/* copy the address itself */
1793			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1794			      sizeof(struct in6_addr));
1795			/* XXX: KAME link-local hack; remove ifindex */
1796			if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr))
1797				((struct in6_addr *)cp)->s6_addr16[1] = 0;
1798			cp += sizeof(struct in6_addr);
1799
1800			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1801			copied += (sizeof(struct in6_addr) +
1802				   sizeof(u_int32_t));
1803		}
1804		if (ifp0)	/* we need search only on the specified IF */
1805			break;
1806	}
1807
1808	if (allow_deprecated == 0 && ifp_dep != NULL) {
1809		ifp = ifp_dep;
1810		allow_deprecated = 1;
1811
1812		goto again;
1813	}
1814
1815	return(copied);
1816}
1817
1818/*
1819 * XXX almost dup'ed code with rip6_input.
1820 */
1821static int
1822icmp6_rip6_input(mp, off)
1823	struct	mbuf **mp;
1824	int	off;
1825{
1826	struct mbuf *m = *mp;
1827	register struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1828	register struct in6pcb *in6p;
1829	struct in6pcb *last = NULL;
1830	struct sockaddr_in6 rip6src;
1831	struct icmp6_hdr *icmp6;
1832	struct mbuf *opts = NULL;
1833
1834#ifndef PULLDOWN_TEST
1835	/* this is assumed to be safe. */
1836	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1837#else
1838	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1839	if (icmp6 == NULL) {
1840		/* m is already reclaimed */
1841		return IPPROTO_DONE;
1842	}
1843#endif
1844
1845	bzero(&rip6src, sizeof(rip6src));
1846	rip6src.sin6_len = sizeof(struct sockaddr_in6);
1847	rip6src.sin6_family = AF_INET6;
1848	/* KAME hack: recover scopeid */
1849	(void)in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif);
1850
1851	for (in6p = rawin6pcb.in6p_next;
1852	     in6p != &rawin6pcb; in6p = in6p->in6p_next)
1853	{
1854		if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1855			continue;
1856		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1857		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1858			continue;
1859		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1860		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1861			continue;
1862		if (in6p->in6p_icmp6filt
1863		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1864				 in6p->in6p_icmp6filt))
1865			continue;
1866		if (last) {
1867			struct	mbuf *n;
1868			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1869				if (last->in6p_flags & IN6P_CONTROLOPTS)
1870					ip6_savecontrol(last, &opts, ip6, n);
1871				/* strip intermediate headers */
1872				m_adj(n, off);
1873				if (sbappendaddr(&last->in6p_socket->so_rcv,
1874						 (struct sockaddr *)&rip6src,
1875						 n, opts) == 0) {
1876					/* should notify about lost packet */
1877					m_freem(n);
1878					if (opts)
1879						m_freem(opts);
1880				} else
1881					sorwakeup(last->in6p_socket);
1882				opts = NULL;
1883			}
1884		}
1885		last = in6p;
1886	}
1887	if (last) {
1888		if (last->in6p_flags & IN6P_CONTROLOPTS)
1889			ip6_savecontrol(last, &opts, ip6, m);
1890		/* strip intermediate headers */
1891		m_adj(m, off);
1892		if (sbappendaddr(&last->in6p_socket->so_rcv,
1893				(struct sockaddr *)&rip6src, m, opts) == 0) {
1894			m_freem(m);
1895			if (opts)
1896				m_freem(opts);
1897		} else
1898			sorwakeup(last->in6p_socket);
1899	} else {
1900		m_freem(m);
1901		ip6stat.ip6s_delivered--;
1902	}
1903	return IPPROTO_DONE;
1904}
1905
1906/*
1907 * Reflect the ip6 packet back to the source.
1908 * OFF points to the icmp6 header, counted from the top of the mbuf.
1909 */
1910void
1911icmp6_reflect(m, off)
1912	struct	mbuf *m;
1913	size_t off;
1914{
1915	struct ip6_hdr *ip6;
1916	struct icmp6_hdr *icmp6;
1917	struct in6_ifaddr *ia;
1918	struct in6_addr t, *src = 0;
1919	int plen;
1920	int type, code;
1921	struct ifnet *outif = NULL;
1922#ifdef COMPAT_RFC1885
1923	int mtu = IPV6_MMTU;
1924	struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
1925#endif
1926
1927	/* too short to reflect */
1928	if (off < sizeof(struct ip6_hdr)) {
1929		printf("sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
1930		       (u_long)off, (u_long)sizeof(struct ip6_hdr),
1931		       __FILE__, __LINE__);
1932		goto bad;
1933	}
1934
1935	/*
1936	 * If there are extra headers between IPv6 and ICMPv6, strip
1937	 * off that header first.
1938	 */
1939#ifdef DIAGNOSTIC
1940	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
1941		panic("assumption failed in icmp6_reflect");
1942#endif
1943	if (off > sizeof(struct ip6_hdr)) {
1944		size_t l;
1945		struct ip6_hdr nip6;
1946
1947		l = off - sizeof(struct ip6_hdr);
1948		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
1949		m_adj(m, l);
1950		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1951		if (m->m_len < l) {
1952			if ((m = m_pullup(m, l)) == NULL)
1953				return;
1954		}
1955		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
1956	} else /* off == sizeof(struct ip6_hdr) */ {
1957		size_t l;
1958		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1959		if (m->m_len < l) {
1960			if ((m = m_pullup(m, l)) == NULL)
1961				return;
1962		}
1963	}
1964	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
1965	ip6 = mtod(m, struct ip6_hdr *);
1966	ip6->ip6_nxt = IPPROTO_ICMPV6;
1967	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
1968	type = icmp6->icmp6_type; /* keep type for statistics */
1969	code = icmp6->icmp6_code; /* ditto. */
1970
1971	t = ip6->ip6_dst;
1972	/*
1973	 * ip6_input() drops a packet if its src is multicast.
1974	 * So, the src is never multicast.
1975	 */
1976	ip6->ip6_dst = ip6->ip6_src;
1977
1978	/* XXX hack for link-local addresses */
1979	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
1980		ip6->ip6_dst.s6_addr16[1] =
1981			htons(m->m_pkthdr.rcvif->if_index);
1982	if (IN6_IS_ADDR_LINKLOCAL(&t))
1983		t.s6_addr16[1] = htons(m->m_pkthdr.rcvif->if_index);
1984
1985#ifdef COMPAT_RFC1885
1986	/*
1987	 * xxx guess MTU
1988	 * RFC 1885 requires that echo reply should be truncated if it
1989	 * does not fit in with (return) path MTU, but the description was
1990	 * removed in the new spec.
1991	 */
1992	if (icmp6_reflect_rt.ro_rt == 0 ||
1993	    ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
1994		if (icmp6_reflect_rt.ro_rt) {
1995			icmp6_reflect_rt.ro_rt = 0;
1996		}
1997		bzero(sin6, sizeof(*sin6));
1998		sin6->sin6_family = PF_INET6;
1999		sin6->sin6_len = sizeof(struct sockaddr_in6);
2000		sin6->sin6_addr = ip6->ip6_dst;
2001
2002		rtalloc((struct route *)&icmp6_reflect_rt.ro_rt);
2003	}
2004
2005	if (icmp6_reflect_rt.ro_rt == 0)
2006		goto bad;
2007
2008	if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
2009	    && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
2010		mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
2011
2012	if (mtu < m->m_pkthdr.len) {
2013		plen -= (m->m_pkthdr.len - mtu);
2014		m_adj(m, mtu - m->m_pkthdr.len);
2015	}
2016#endif
2017	/*
2018	 * If the incoming packet was addressed directly to us(i.e. unicast),
2019	 * use dst as the src for the reply.
2020	 * The IN6_IFF_NOTREADY case would be VERY rare, but is possible
2021	 * (for example) when we encounter an error while forwarding procedure
2022	 * destined to a duplicated address of ours.
2023	 */
2024	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2025		if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2026		    (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2027			src = &t;
2028			break;
2029		}
2030	if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2031		/*
2032		 * This is the case if the dst is our link-local address
2033		 * and the sender is also ourseleves.
2034		 */
2035		src = &t;
2036	}
2037
2038	if (src == 0)
2039		/*
2040		 * This case matches to multicasts, our anycast, or unicasts
2041		 * that we do not own. Select a source address which has the
2042		 * same scope.
2043		 * XXX: for (non link-local) multicast addresses, this might
2044		 * not be a good choice.
2045		 */
2046		if ((ia = in6_ifawithscope(m->m_pkthdr.rcvif, &t)) != 0)
2047			src = &IA6_SIN6(ia)->sin6_addr;
2048
2049	if (src == 0)
2050		goto bad;
2051
2052	ip6->ip6_src = *src;
2053
2054	ip6->ip6_flow = 0;
2055	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2056	ip6->ip6_vfc |= IPV6_VERSION;
2057	ip6->ip6_nxt = IPPROTO_ICMPV6;
2058	if (m->m_pkthdr.rcvif) {
2059		/* XXX: This may not be the outgoing interface */
2060		ip6->ip6_hlim = nd_ifinfo[m->m_pkthdr.rcvif->if_index].chlim;
2061	}
2062
2063	icmp6->icmp6_cksum = 0;
2064	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2065					sizeof(struct ip6_hdr), plen);
2066
2067	/*
2068	 * xxx option handling
2069	 */
2070
2071	m->m_flags &= ~(M_BCAST|M_MCAST);
2072#ifdef IPSEC
2073	/* Don't lookup socket */
2074	ipsec_setsocket(m, NULL);
2075#endif /*IPSEC*/
2076
2077#ifdef COMPAT_RFC1885
2078	ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif);
2079#else
2080	ip6_output(m, NULL, NULL, 0, NULL, &outif);
2081#endif
2082	if (outif)
2083		icmp6_ifoutstat_inc(outif, type, code);
2084
2085	return;
2086
2087 bad:
2088	m_freem(m);
2089	return;
2090}
2091
2092void
2093icmp6_fasttimo()
2094{
2095
2096	mld6_fasttimeo();
2097}
2098
2099static const char *
2100icmp6_redirect_diag(src6, dst6, tgt6)
2101	struct in6_addr *src6;
2102	struct in6_addr *dst6;
2103	struct in6_addr *tgt6;
2104{
2105	static char buf[1024];
2106	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2107		ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2108	return buf;
2109}
2110
2111void
2112icmp6_redirect_input(m, off)
2113	register struct mbuf *m;
2114	int off;
2115{
2116	struct ifnet *ifp = m->m_pkthdr.rcvif;
2117	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2118	struct nd_redirect *nd_rd;
2119	int icmp6len = ntohs(ip6->ip6_plen);
2120	char *lladdr = NULL;
2121	int lladdrlen = 0;
2122	u_char *redirhdr = NULL;
2123	int redirhdrlen = 0;
2124	struct rtentry *rt = NULL;
2125	int is_router;
2126	int is_onlink;
2127	struct in6_addr src6 = ip6->ip6_src;
2128	struct in6_addr redtgt6;
2129	struct in6_addr reddst6;
2130	union nd_opts ndopts;
2131
2132	if (!m || !ifp)
2133		return;
2134
2135	/* XXX if we are router, we don't update route by icmp6 redirect */
2136	if (ip6_forwarding)
2137		goto freeit;
2138	if (!icmp6_rediraccept)
2139		goto freeit;
2140
2141#ifndef PULLDOWN_TEST
2142	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2143	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2144#else
2145	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2146	if (nd_rd == NULL) {
2147		icmp6stat.icp6s_tooshort++;
2148		return;
2149	}
2150#endif
2151	redtgt6 = nd_rd->nd_rd_target;
2152	reddst6 = nd_rd->nd_rd_dst;
2153
2154	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2155		redtgt6.s6_addr16[1] = htons(ifp->if_index);
2156	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2157		reddst6.s6_addr16[1] = htons(ifp->if_index);
2158
2159	/* validation */
2160	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2161		log(LOG_ERR,
2162			"ICMP6 redirect sent from %s rejected; "
2163			"must be from linklocal\n", ip6_sprintf(&src6));
2164		goto freeit;
2165	}
2166	if (ip6->ip6_hlim != 255) {
2167		log(LOG_ERR,
2168			"ICMP6 redirect sent from %s rejected; "
2169			"hlim=%d (must be 255)\n",
2170			ip6_sprintf(&src6), ip6->ip6_hlim);
2171		goto freeit;
2172	}
2173    {
2174	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2175	struct sockaddr_in6 sin6;
2176	struct in6_addr *gw6;
2177
2178	bzero(&sin6, sizeof(sin6));
2179	sin6.sin6_family = AF_INET6;
2180	sin6.sin6_len = sizeof(struct sockaddr_in6);
2181	bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2182	rt = rtalloc1((struct sockaddr *)&sin6, 0);
2183	if (rt) {
2184		if (rt->rt_gateway == NULL ||
2185		    rt->rt_gateway->sa_family != AF_INET6) {
2186			log(LOG_ERR,
2187			    "ICMP6 redirect rejected; no route "
2188			    "with inet6 gateway found for redirect dst: %s\n",
2189			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2190			RTFREE(rt);
2191			goto freeit;
2192		}
2193
2194		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2195		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2196			log(LOG_ERR,
2197				"ICMP6 redirect rejected; "
2198				"not equal to gw-for-src=%s (must be same): "
2199				"%s\n",
2200				ip6_sprintf(gw6),
2201				icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2202			RTFREE(rt);
2203			goto freeit;
2204		}
2205	} else {
2206		log(LOG_ERR,
2207			"ICMP6 redirect rejected; "
2208			"no route found for redirect dst: %s\n",
2209			icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2210		goto freeit;
2211	}
2212	RTFREE(rt);
2213	rt = NULL;
2214    }
2215	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2216		log(LOG_ERR,
2217			"ICMP6 redirect rejected; "
2218			"redirect dst must be unicast: %s\n",
2219			icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2220		goto freeit;
2221	}
2222
2223	is_router = is_onlink = 0;
2224	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2225		is_router = 1;	/* router case */
2226	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2227		is_onlink = 1;	/* on-link destination case */
2228	if (!is_router && !is_onlink) {
2229		log(LOG_ERR,
2230			"ICMP6 redirect rejected; "
2231			"neither router case nor onlink case: %s\n",
2232			icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2233		goto freeit;
2234	}
2235	/* validation passed */
2236
2237	icmp6len -= sizeof(*nd_rd);
2238	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2239	if (nd6_options(&ndopts) < 0) {
2240		log(LOG_INFO, "icmp6_redirect_input: "
2241			"invalid ND option, rejected: %s\n",
2242			icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2243		goto freeit;
2244	}
2245
2246	if (ndopts.nd_opts_tgt_lladdr) {
2247		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2248		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2249	}
2250
2251	if (ndopts.nd_opts_rh) {
2252		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2253		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2254	}
2255
2256	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2257		log(LOG_INFO,
2258			"icmp6_redirect_input: lladdrlen mismatch for %s "
2259			"(if %d, icmp6 packet %d): %s\n",
2260			ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2261			icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
2262	}
2263
2264	/* RFC 2461 8.3 */
2265	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2266			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2267
2268	if (!is_onlink) {	/* better router case. perform rtredirect. */
2269		/* perform rtredirect */
2270		struct sockaddr_in6 sdst;
2271		struct sockaddr_in6 sgw;
2272		struct sockaddr_in6 ssrc;
2273
2274		bzero(&sdst, sizeof(sdst));
2275		bzero(&sgw, sizeof(sgw));
2276		bzero(&ssrc, sizeof(ssrc));
2277		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2278		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2279			sizeof(struct sockaddr_in6);
2280		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2281		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2282		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2283		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2284			   (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2285			   (struct sockaddr *)&ssrc,
2286			   (struct rtentry **)NULL
2287			   );
2288	}
2289	/* finally update cached route in each socket via pfctlinput */
2290    {
2291	struct sockaddr_in6 sdst;
2292#if 1
2293#else
2294	struct ip6protosw *pr;
2295#endif
2296
2297	bzero(&sdst, sizeof(sdst));
2298	sdst.sin6_family = AF_INET6;
2299	sdst.sin6_len = sizeof(struct sockaddr_in6);
2300	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2301#if 1
2302	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2303#else
2304	/*
2305	 * do not use pfctlinput() here, we have different prototype for
2306	 * xx_ctlinput() in ip6proto.
2307	 */
2308	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
2309	     pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW;
2310	     pr++) {
2311		if (pr->pr_ctlinput) {
2312			(*pr->pr_ctlinput)(PRC_REDIRECT_HOST,
2313				(struct sockaddr *)&sdst, NULL, NULL, 0);
2314		}
2315	}
2316#endif
2317#ifdef IPSEC
2318	key_sa_routechange((struct sockaddr *)&sdst);
2319#endif
2320    }
2321
2322 freeit:
2323	m_freem(m);
2324}
2325
2326void
2327icmp6_redirect_output(m0, rt)
2328	struct mbuf *m0;
2329	struct rtentry *rt;
2330{
2331	struct ifnet *ifp;	/* my outgoing interface */
2332	struct in6_addr *ifp_ll6;
2333	struct in6_addr *router_ll6;
2334	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2335	struct mbuf *m = NULL;	/* newly allocated one */
2336	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2337	struct nd_redirect *nd_rd;
2338	size_t maxlen;
2339	u_char *p;
2340	struct ifnet *outif = NULL;
2341	struct sockaddr_in6 src_sa;
2342
2343	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2344
2345	/* if we are not router, we don't send icmp6 redirect */
2346	if (!ip6_forwarding || ip6_accept_rtadv)
2347		goto fail;
2348
2349	/* sanity check */
2350	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2351		goto fail;
2352
2353	/*
2354	 * Address check:
2355	 *  the source address must identify a neighbor, and
2356	 *  the destination address must not be a multicast address
2357	 *  [RFC 2461, sec 8.2]
2358	 */
2359	sip6 = mtod(m0, struct ip6_hdr *);
2360	bzero(&src_sa, sizeof(src_sa));
2361	src_sa.sin6_family = AF_INET6;
2362	src_sa.sin6_len = sizeof(src_sa);
2363	src_sa.sin6_addr = sip6->ip6_src;
2364	/* we don't currently use sin6_scope_id, but eventually use it */
2365	src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src);
2366	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2367		goto fail;
2368	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2369		goto fail;	/* what should we do here? */
2370
2371	/* rate limit */
2372	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2373		goto fail;
2374
2375	/*
2376	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2377	 * we almost always ask for an mbuf cluster for simplicity.
2378	 * (MHLEN < IPV6_MMTU is almost always true)
2379	 */
2380#if IPV6_MMTU >= MCLBYTES
2381# error assumption failed about IPV6_MMTU and MCLBYTES
2382#endif
2383	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2384	if (m && IPV6_MMTU >= MHLEN)
2385		MCLGET(m, M_DONTWAIT);
2386	if (!m)
2387		goto fail;
2388	m->m_len = 0;
2389	maxlen = M_TRAILINGSPACE(m);
2390	maxlen = min(IPV6_MMTU, maxlen);
2391	/* just for safety */
2392	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2393	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2394		goto fail;
2395	}
2396
2397	{
2398		/* get ip6 linklocal address for ifp(my outgoing interface). */
2399		struct in6_ifaddr *ia;
2400		if ((ia = in6ifa_ifpforlinklocal(ifp,
2401						 IN6_IFF_NOTREADY|
2402						 IN6_IFF_ANYCAST)) == NULL)
2403			goto fail;
2404		ifp_ll6 = &ia->ia_addr.sin6_addr;
2405	}
2406
2407	/* get ip6 linklocal address for the router. */
2408	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2409		struct sockaddr_in6 *sin6;
2410		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2411		router_ll6 = &sin6->sin6_addr;
2412		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2413			router_ll6 = (struct in6_addr *)NULL;
2414	} else
2415		router_ll6 = (struct in6_addr *)NULL;
2416
2417	/* ip6 */
2418	ip6 = mtod(m, struct ip6_hdr *);
2419	ip6->ip6_flow = 0;
2420	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2421	ip6->ip6_vfc |= IPV6_VERSION;
2422	/* ip6->ip6_plen will be set later */
2423	ip6->ip6_nxt = IPPROTO_ICMPV6;
2424	ip6->ip6_hlim = 255;
2425	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2426	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2427	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2428
2429	/* ND Redirect */
2430	nd_rd = (struct nd_redirect *)(ip6 + 1);
2431	nd_rd->nd_rd_type = ND_REDIRECT;
2432	nd_rd->nd_rd_code = 0;
2433	nd_rd->nd_rd_reserved = 0;
2434	if (rt->rt_flags & RTF_GATEWAY) {
2435		/*
2436		 * nd_rd->nd_rd_target must be a link-local address in
2437		 * better router cases.
2438		 */
2439		if (!router_ll6)
2440			goto fail;
2441		bcopy(router_ll6, &nd_rd->nd_rd_target,
2442		      sizeof(nd_rd->nd_rd_target));
2443		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2444		      sizeof(nd_rd->nd_rd_dst));
2445	} else {
2446		/* make sure redtgt == reddst */
2447		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2448		      sizeof(nd_rd->nd_rd_target));
2449		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2450		      sizeof(nd_rd->nd_rd_dst));
2451	}
2452
2453	p = (u_char *)(nd_rd + 1);
2454
2455	if (!router_ll6)
2456		goto nolladdropt;
2457
2458    {
2459	/* target lladdr option */
2460	struct rtentry *rt_router = NULL;
2461	int len;
2462	struct sockaddr_dl *sdl;
2463	struct nd_opt_hdr *nd_opt;
2464	char *lladdr;
2465
2466	rt_router = nd6_lookup(router_ll6, 0, ifp);
2467	if (!rt_router)
2468		goto nolladdropt;
2469	len = sizeof(*nd_opt) + ifp->if_addrlen;
2470	len = (len + 7) & ~7;	/*round by 8*/
2471	/* safety check */
2472	if (len + (p - (u_char *)ip6) > maxlen)
2473		goto nolladdropt;
2474	if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2475	    (rt_router->rt_flags & RTF_LLINFO) &&
2476	    (rt_router->rt_gateway->sa_family == AF_LINK) &&
2477	    (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2478	    sdl->sdl_alen) {
2479		nd_opt = (struct nd_opt_hdr *)p;
2480		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2481		nd_opt->nd_opt_len = len >> 3;
2482		lladdr = (char *)(nd_opt + 1);
2483		bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2484		p += len;
2485	}
2486    }
2487nolladdropt:;
2488
2489	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2490
2491	/* just to be safe */
2492	if (m0->m_flags & M_DECRYPTED)
2493		goto noredhdropt;
2494	if (p - (u_char *)ip6 > maxlen)
2495		goto noredhdropt;
2496
2497    {
2498	/* redirected header option */
2499	int len;
2500	struct nd_opt_rd_hdr *nd_opt_rh;
2501
2502	/*
2503	 * compute the maximum size for icmp6 redirect header option.
2504	 * XXX room for auth header?
2505	 */
2506	len = maxlen - (p - (u_char *)ip6);
2507	len &= ~7;
2508
2509	/* This is just for simplicity. */
2510	if (m0->m_pkthdr.len != m0->m_len) {
2511		if (m0->m_next) {
2512			m_freem(m0->m_next);
2513			m0->m_next = NULL;
2514		}
2515		m0->m_pkthdr.len = m0->m_len;
2516	}
2517
2518	/*
2519	 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2520	 * about padding/truncate rule for the original IP packet.
2521	 * From the discussion on IPv6imp in Feb 1999, the consensus was:
2522	 * - "attach as much as possible" is the goal
2523	 * - pad if not aligned (original size can be guessed by original
2524	 *   ip6 header)
2525	 * Following code adds the padding if it is simple enough,
2526	 * and truncates if not.
2527	 */
2528	if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2529		panic("assumption failed in %s:%d\n", __FILE__, __LINE__);
2530
2531	if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2532		/* not enough room, truncate */
2533		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2534	} else {
2535		/* enough room, pad or truncate */
2536		size_t extra;
2537
2538		extra = m0->m_pkthdr.len % 8;
2539		if (extra) {
2540			/* pad if easy enough, truncate if not */
2541			if (8 - extra <= M_TRAILINGSPACE(m0)) {
2542				/* pad */
2543				m0->m_len += (8 - extra);
2544				m0->m_pkthdr.len += (8 - extra);
2545			} else {
2546				/* truncate */
2547				m0->m_pkthdr.len -= extra;
2548				m0->m_len -= extra;
2549			}
2550		}
2551		len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2552		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2553	}
2554
2555	nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2556	bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2557	nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2558	nd_opt_rh->nd_opt_rh_len = len >> 3;
2559	p += sizeof(*nd_opt_rh);
2560	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2561
2562	/* connect m0 to m */
2563	m->m_next = m0;
2564	m->m_pkthdr.len = m->m_len + m0->m_len;
2565    }
2566noredhdropt:;
2567
2568	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
2569		sip6->ip6_src.s6_addr16[1] = 0;
2570	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
2571		sip6->ip6_dst.s6_addr16[1] = 0;
2572#if 0
2573	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
2574		ip6->ip6_src.s6_addr16[1] = 0;
2575	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
2576		ip6->ip6_dst.s6_addr16[1] = 0;
2577#endif
2578	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
2579		nd_rd->nd_rd_target.s6_addr16[1] = 0;
2580	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
2581		nd_rd->nd_rd_dst.s6_addr16[1] = 0;
2582
2583	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2584
2585	nd_rd->nd_rd_cksum = 0;
2586	nd_rd->nd_rd_cksum
2587		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2588
2589	/* send the packet to outside... */
2590#ifdef IPSEC
2591	/* Don't lookup socket */
2592	ipsec_setsocket(m, NULL);
2593#endif /*IPSEC*/
2594	ip6_output(m, NULL, NULL, 0, NULL, &outif);
2595	if (outif) {
2596		icmp6_ifstat_inc(outif, ifs6_out_msg);
2597		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2598	}
2599	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2600
2601	return;
2602
2603fail:
2604	if (m)
2605		m_freem(m);
2606	if (m0)
2607		m_freem(m0);
2608}
2609
2610/*
2611 * ICMPv6 socket option processing.
2612 */
2613int
2614icmp6_ctloutput(op, so, level, optname, mp)
2615	int op;
2616	struct socket *so;
2617	int level, optname;
2618	struct mbuf **mp;
2619{
2620	register struct in6pcb *in6p = sotoin6pcb(so);
2621	register struct mbuf *m = *mp;
2622	int error = 0;
2623
2624	if (level != IPPROTO_ICMPV6) {
2625		error = EINVAL;
2626		if (op == PRCO_SETOPT && m)
2627			(void)m_free(m);
2628	} else switch(op) {
2629	 case PRCO_SETOPT:
2630		 switch (optname) {
2631		  case ICMP6_FILTER:
2632		  {
2633			  struct icmp6_filter *p;
2634
2635			  p = mtod(m, struct icmp6_filter *);
2636			  if (!p || !in6p->in6p_icmp6filt) {
2637				  error = EINVAL;
2638				  break;
2639			  }
2640			  bcopy(p, in6p->in6p_icmp6filt,
2641				sizeof(struct icmp6_filter));
2642			  error = 0;
2643			  break;
2644		  }
2645
2646		  default:
2647			  error = ENOPROTOOPT;
2648			  break;
2649		 }
2650		 if (m)
2651			 (void)m_free(m);
2652		 break;
2653
2654	 case PRCO_GETOPT:
2655		 switch (optname) {
2656		  case ICMP6_FILTER:
2657		  {
2658			  struct icmp6_filter *p;
2659
2660			  if (!in6p->in6p_icmp6filt) {
2661				  error = EINVAL;
2662				  break;
2663			  }
2664			  *mp = m = m_get(M_WAIT, MT_SOOPTS);
2665			  m->m_len = sizeof(struct icmp6_filter);
2666			  p = mtod(m, struct icmp6_filter *);
2667			  bcopy(in6p->in6p_icmp6filt, p,
2668				sizeof(struct icmp6_filter));
2669			  error = 0;
2670			  break;
2671		  }
2672
2673		  default:
2674			  error = ENOPROTOOPT;
2675			  break;
2676		 }
2677		 break;
2678	}
2679
2680	return(error);
2681}
2682
2683/*
2684 * Perform rate limit check.
2685 * Returns 0 if it is okay to send the icmp6 packet.
2686 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2687 * limitation.
2688 *
2689 * XXX per-destination/type check necessary?
2690 */
2691static int
2692icmp6_ratelimit(dst, type, code)
2693	const struct in6_addr *dst;	/* not used at this moment */
2694	const int type;			/* not used at this moment */
2695	const int code;			/* not used at this moment */
2696{
2697	int ret;
2698
2699	ret = 0;	/*okay to send*/
2700
2701	/* PPS limit */
2702	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2703	    icmp6errppslim)) {
2704		/* The packet is subject to rate limit */
2705		ret++;
2706	}
2707
2708	return ret;
2709}
2710
2711static struct rtentry *
2712icmp6_mtudisc_clone(dst)
2713	struct sockaddr *dst;
2714{
2715	struct rtentry *rt;
2716	int    error;
2717
2718	rt = rtalloc1(dst, 1);
2719	if (rt == 0)
2720		return NULL;
2721
2722	/* If we didn't get a host route, allocate one */
2723	if ((rt->rt_flags & RTF_HOST) == 0) {
2724		struct rtentry *nrt;
2725
2726		error = rtrequest((int) RTM_ADD, dst,
2727		    (struct sockaddr *) rt->rt_gateway,
2728		    (struct sockaddr *) 0,
2729		    RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt);
2730		if (error) {
2731			rtfree(rt);
2732			rtfree(nrt);
2733			return NULL;
2734		}
2735		nrt->rt_rmx = rt->rt_rmx;
2736		rtfree(rt);
2737		rt = nrt;
2738	}
2739	error = rt_timer_add(rt, icmp6_mtudisc_timeout,
2740			icmp6_mtudisc_timeout_q);
2741	if (error) {
2742		rtfree(rt);
2743		return NULL;
2744	}
2745
2746	return rt;	/* caller need to call rtfree() */
2747}
2748
2749static void
2750icmp6_mtudisc_timeout(rt, r)
2751	struct rtentry *rt;
2752	struct rttimer *r;
2753{
2754	if (rt == NULL)
2755		panic("icmp6_mtudisc_timeout: bad route to timeout");
2756	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
2757	    (RTF_DYNAMIC | RTF_HOST)) {
2758		rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt),
2759		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
2760	} else {
2761		if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) {
2762			rt->rt_rmx.rmx_mtu = 0;
2763		}
2764	}
2765}
2766
2767#include <uvm/uvm_extern.h>
2768#include <sys/sysctl.h>
2769
2770int
2771icmp6_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
2772	int *name;
2773	u_int namelen;
2774	void *oldp;
2775	size_t *oldlenp;
2776	void *newp;
2777	size_t newlen;
2778{
2779
2780	/* All sysctl names at this level are terminal. */
2781	if (namelen != 1)
2782		return ENOTDIR;
2783
2784	switch (name[0]) {
2785
2786	case ICMPV6CTL_REDIRACCEPT:
2787		return sysctl_int(oldp, oldlenp, newp, newlen,
2788				&icmp6_rediraccept);
2789	case ICMPV6CTL_REDIRTIMEOUT:
2790		return sysctl_int(oldp, oldlenp, newp, newlen,
2791				&icmp6_redirtimeout);
2792	case ICMPV6CTL_STATS:
2793		return sysctl_rdstruct(oldp, oldlenp, newp,
2794				&icmp6stat, sizeof(icmp6stat));
2795	case ICMPV6CTL_ND6_PRUNE:
2796		return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_prune);
2797	case ICMPV6CTL_ND6_DELAY:
2798		return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_delay);
2799	case ICMPV6CTL_ND6_UMAXTRIES:
2800		return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_umaxtries);
2801	case ICMPV6CTL_ND6_MMAXTRIES:
2802		return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_mmaxtries);
2803	case ICMPV6CTL_ND6_USELOOPBACK:
2804		return sysctl_int(oldp, oldlenp, newp, newlen,
2805				&nd6_useloopback);
2806	case ICMPV6CTL_NODEINFO:
2807		return sysctl_int(oldp, oldlenp, newp, newlen, &icmp6_nodeinfo);
2808	case ICMPV6CTL_ERRPPSLIMIT:
2809		return sysctl_int(oldp, oldlenp, newp, newlen, &icmp6errppslim);
2810	case ICMPV6CTL_ND6_MAXNUDHINT:
2811		return sysctl_int(oldp, oldlenp, newp, newlen,
2812				&nd6_maxnudhint);
2813	case ICMPV6CTL_MTUDISC_HIWAT:
2814		return sysctl_int(oldp, oldlenp, newp, newlen,
2815				&icmp6_mtudisc_hiwat);
2816	case ICMPV6CTL_MTUDISC_LOWAT:
2817		return sysctl_int(oldp, oldlenp, newp, newlen,
2818				&icmp6_mtudisc_lowat);
2819	default:
2820		return ENOPROTOOPT;
2821	}
2822	/* NOTREACHED */
2823}
2824