icmp6.c revision 120971
1/*	$FreeBSD: head/sys/netinet6/icmp6.c 120971 2003-10-10 16:04:00Z ume $	*/
2/*	$KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $	*/
3
4/*
5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 *    notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 *    notice, this list of conditions and the following disclaimer in the
15 *    documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the project nor the names of its contributors
17 *    may be used to endorse or promote products derived from this software
18 *    without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 */
32
33/*
34 * Copyright (c) 1982, 1986, 1988, 1993
35 *	The Regents of the University of California.  All rights reserved.
36 *
37 * Redistribution and use in source and binary forms, with or without
38 * modification, are permitted provided that the following conditions
39 * are met:
40 * 1. Redistributions of source code must retain the above copyright
41 *    notice, this list of conditions and the following disclaimer.
42 * 2. Redistributions in binary form must reproduce the above copyright
43 *    notice, this list of conditions and the following disclaimer in the
44 *    documentation and/or other materials provided with the distribution.
45 * 3. All advertising materials mentioning features or use of this software
46 *    must display the following acknowledgement:
47 *	This product includes software developed by the University of
48 *	California, Berkeley and its contributors.
49 * 4. Neither the name of the University nor the names of its contributors
50 *    may be used to endorse or promote products derived from this software
51 *    without specific prior written permission.
52 *
53 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63 * SUCH DAMAGE.
64 *
65 *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
66 */
67
68#include "opt_inet.h"
69#include "opt_inet6.h"
70#include "opt_ipsec.h"
71
72#include <sys/param.h>
73#include <sys/domain.h>
74#include <sys/kernel.h>
75#include <sys/lock.h>
76#include <sys/malloc.h>
77#include <sys/mbuf.h>
78#include <sys/protosw.h>
79#include <sys/signalvar.h>
80#include <sys/socket.h>
81#include <sys/socketvar.h>
82#include <sys/sx.h>
83#include <sys/syslog.h>
84#include <sys/systm.h>
85#include <sys/time.h>
86
87#include <net/if.h>
88#include <net/if_dl.h>
89#include <net/if_types.h>
90#include <net/route.h>
91
92#include <netinet/in.h>
93#include <netinet/in_pcb.h>
94#include <netinet/in_var.h>
95#include <netinet/ip6.h>
96#include <netinet/icmp6.h>
97#include <netinet6/in6_ifattach.h>
98#include <netinet6/in6_pcb.h>
99#include <netinet6/ip6protosw.h>
100#include <netinet6/ip6_var.h>
101#include <netinet6/mld6_var.h>
102#include <netinet6/nd6.h>
103
104#ifdef IPSEC
105#include <netinet6/ipsec.h>
106#include <netkey/key.h>
107#endif
108
109#ifdef FAST_IPSEC
110#include <netipsec/ipsec.h>
111#include <netipsec/key.h>
112#define	IPSEC
113#endif
114
115#include <net/net_osdep.h>
116
117#ifdef HAVE_NRL_INPCB
118/* inpcb members */
119#define in6pcb		inpcb
120#define in6p_laddr	inp_laddr6
121#define in6p_faddr	inp_faddr6
122#define in6p_icmp6filt	inp_icmp6filt
123#define in6p_route	inp_route
124#define in6p_socket	inp_socket
125#define in6p_flags	inp_flags
126#define in6p_moptions	inp_moptions6
127#define in6p_outputopts	inp_outputopts6
128#define in6p_ip6	inp_ipv6
129#define in6p_flowinfo	inp_flowinfo
130#define in6p_sp		inp_sp
131#define in6p_next	inp_next
132#define in6p_prev	inp_prev
133/* macro names */
134#define sotoin6pcb	sotoinpcb
135/* function names */
136#define in6_pcbdetach	in_pcbdetach
137#define in6_rtchange	in_rtchange
138
139/*
140 * for KAME src sync over BSD*'s. XXX: FreeBSD (>=3) are VERY different from
141 * others...
142 */
143#define in6p_ip6_nxt	inp_ipv6.ip6_nxt
144#endif
145
146extern struct domain inet6domain;
147
148struct icmp6stat icmp6stat;
149
150extern struct inpcbhead ripcb;
151extern int icmp6errppslim;
152static int icmp6errpps_count = 0;
153static struct timeval icmp6errppslim_last;
154extern int icmp6_nodeinfo;
155
156static void icmp6_errcount __P((struct icmp6errstat *, int, int));
157static int icmp6_rip6_input __P((struct mbuf **, int));
158static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int));
159static const char *icmp6_redirect_diag __P((struct in6_addr *,
160	struct in6_addr *, struct in6_addr *));
161#define	HAVE_PPSRATECHECK
162#ifndef HAVE_PPSRATECHECK
163static int ppsratecheck __P((struct timeval *, int *, int));
164#endif
165static struct mbuf *ni6_input __P((struct mbuf *, int));
166static struct mbuf *ni6_nametodns __P((const char *, int, int));
167static int ni6_dnsmatch __P((const char *, int, const char *, int));
168static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
169			  struct ifnet **, char *));
170static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
171				struct ifnet *, int));
172static int icmp6_notify_error __P((struct mbuf *, int, int, int));
173
174#ifdef COMPAT_RFC1885
175static struct route_in6 icmp6_reflect_rt;
176#endif
177
178
179void
180icmp6_init()
181{
182	mld6_init();
183}
184
185static void
186icmp6_errcount(stat, type, code)
187	struct icmp6errstat *stat;
188	int type, code;
189{
190	switch (type) {
191	case ICMP6_DST_UNREACH:
192		switch (code) {
193		case ICMP6_DST_UNREACH_NOROUTE:
194			stat->icp6errs_dst_unreach_noroute++;
195			return;
196		case ICMP6_DST_UNREACH_ADMIN:
197			stat->icp6errs_dst_unreach_admin++;
198			return;
199		case ICMP6_DST_UNREACH_BEYONDSCOPE:
200			stat->icp6errs_dst_unreach_beyondscope++;
201			return;
202		case ICMP6_DST_UNREACH_ADDR:
203			stat->icp6errs_dst_unreach_addr++;
204			return;
205		case ICMP6_DST_UNREACH_NOPORT:
206			stat->icp6errs_dst_unreach_noport++;
207			return;
208		}
209		break;
210	case ICMP6_PACKET_TOO_BIG:
211		stat->icp6errs_packet_too_big++;
212		return;
213	case ICMP6_TIME_EXCEEDED:
214		switch (code) {
215		case ICMP6_TIME_EXCEED_TRANSIT:
216			stat->icp6errs_time_exceed_transit++;
217			return;
218		case ICMP6_TIME_EXCEED_REASSEMBLY:
219			stat->icp6errs_time_exceed_reassembly++;
220			return;
221		}
222		break;
223	case ICMP6_PARAM_PROB:
224		switch (code) {
225		case ICMP6_PARAMPROB_HEADER:
226			stat->icp6errs_paramprob_header++;
227			return;
228		case ICMP6_PARAMPROB_NEXTHEADER:
229			stat->icp6errs_paramprob_nextheader++;
230			return;
231		case ICMP6_PARAMPROB_OPTION:
232			stat->icp6errs_paramprob_option++;
233			return;
234		}
235		break;
236	case ND_REDIRECT:
237		stat->icp6errs_redirect++;
238		return;
239	}
240	stat->icp6errs_unknown++;
241}
242
243/*
244 * Generate an error packet of type error in response to bad IP6 packet.
245 */
246void
247icmp6_error(m, type, code, param)
248	struct mbuf *m;
249	int type, code, param;
250{
251	struct ip6_hdr *oip6, *nip6;
252	struct icmp6_hdr *icmp6;
253	u_int preplen;
254	int off;
255	int nxt;
256
257	icmp6stat.icp6s_error++;
258
259	/* count per-type-code statistics */
260	icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
261
262#ifdef M_DECRYPTED	/*not openbsd*/
263	if (m->m_flags & M_DECRYPTED) {
264		icmp6stat.icp6s_canterror++;
265		goto freeit;
266	}
267#endif
268
269#ifndef PULLDOWN_TEST
270	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
271#else
272	if (m->m_len < sizeof(struct ip6_hdr)) {
273		m = m_pullup(m, sizeof(struct ip6_hdr));
274		if (m == NULL)
275			return;
276	}
277#endif
278	oip6 = mtod(m, struct ip6_hdr *);
279
280	/*
281	 * Multicast destination check. For unrecognized option errors,
282	 * this check has already done in ip6_unknown_opt(), so we can
283	 * check only for other errors.
284	 */
285	if ((m->m_flags & (M_BCAST|M_MCAST) ||
286	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
287	    (type != ICMP6_PACKET_TOO_BIG &&
288	     (type != ICMP6_PARAM_PROB ||
289	      code != ICMP6_PARAMPROB_OPTION)))
290		goto freeit;
291
292	/* Source address check. XXX: the case of anycast source? */
293	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
294	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
295		goto freeit;
296
297	/*
298	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
299	 * don't do it.
300	 */
301	nxt = -1;
302	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
303	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
304		struct icmp6_hdr *icp;
305
306#ifndef PULLDOWN_TEST
307		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
308		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
309#else
310		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
311			sizeof(*icp));
312		if (icp == NULL) {
313			icmp6stat.icp6s_tooshort++;
314			return;
315		}
316#endif
317		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
318		    icp->icmp6_type == ND_REDIRECT) {
319			/*
320			 * ICMPv6 error
321			 * Special case: for redirect (which is
322			 * informational) we must not send icmp6 error.
323			 */
324			icmp6stat.icp6s_canterror++;
325			goto freeit;
326		} else {
327			/* ICMPv6 informational - send the error */
328		}
329	} else {
330		/* non-ICMPv6 - send the error */
331	}
332
333	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
334
335	/* Finally, do rate limitation check. */
336	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
337		icmp6stat.icp6s_toofreq++;
338		goto freeit;
339	}
340
341	/*
342	 * OK, ICMP6 can be generated.
343	 */
344
345	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
346		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
347
348	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
349	M_PREPEND(m, preplen, M_DONTWAIT);
350	if (m && m->m_len < preplen)
351		m = m_pullup(m, preplen);
352	if (m == NULL) {
353		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
354		return;
355	}
356
357	nip6 = mtod(m, struct ip6_hdr *);
358	nip6->ip6_src  = oip6->ip6_src;
359	nip6->ip6_dst  = oip6->ip6_dst;
360
361	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src))
362		oip6->ip6_src.s6_addr16[1] = 0;
363	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst))
364		oip6->ip6_dst.s6_addr16[1] = 0;
365
366	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
367	icmp6->icmp6_type = type;
368	icmp6->icmp6_code = code;
369	icmp6->icmp6_pptr = htonl((u_int32_t)param);
370
371	/*
372	 * icmp6_reflect() is designed to be in the input path.
373	 * icmp6_error() can be called from both input and outut path,
374	 * and if we are in output path rcvif could contain bogus value.
375	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
376	 * information in ip header (nip6).
377	 */
378	m->m_pkthdr.rcvif = NULL;
379
380	icmp6stat.icp6s_outhist[type]++;
381	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
382
383	return;
384
385  freeit:
386	/*
387	 * If we can't tell wheter or not we can generate ICMP6, free it.
388	 */
389	m_freem(m);
390}
391
392/*
393 * Process a received ICMP6 message.
394 */
395int
396icmp6_input(mp, offp, proto)
397	struct mbuf **mp;
398	int *offp, proto;
399{
400	struct mbuf *m = *mp, *n;
401	struct ip6_hdr *ip6, *nip6;
402	struct icmp6_hdr *icmp6, *nicmp6;
403	int off = *offp;
404	int icmp6len = m->m_pkthdr.len - *offp;
405	int code, sum, noff;
406
407#ifndef PULLDOWN_TEST
408	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
409	/* m might change if M_LOOP.  So, call mtod after this */
410#endif
411
412	/*
413	 * Locate icmp6 structure in mbuf, and check
414	 * that not corrupted and of at least minimum length
415	 */
416
417	ip6 = mtod(m, struct ip6_hdr *);
418	if (icmp6len < sizeof(struct icmp6_hdr)) {
419		icmp6stat.icp6s_tooshort++;
420		goto freeit;
421	}
422
423	/*
424	 * calculate the checksum
425	 */
426#ifndef PULLDOWN_TEST
427	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
428#else
429	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
430	if (icmp6 == NULL) {
431		icmp6stat.icp6s_tooshort++;
432		return IPPROTO_DONE;
433	}
434#endif
435	code = icmp6->icmp6_code;
436
437	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
438		nd6log((LOG_ERR,
439		    "ICMP6 checksum error(%d|%x) %s\n",
440		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
441		icmp6stat.icp6s_checksum++;
442		goto freeit;
443	}
444
445	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
446		/*
447		 * Deliver very specific ICMP6 type only.
448		 * This is important to deilver TOOBIG.  Otherwise PMTUD
449		 * will not work.
450		 */
451		switch (icmp6->icmp6_type) {
452		case ICMP6_DST_UNREACH:
453		case ICMP6_PACKET_TOO_BIG:
454		case ICMP6_TIME_EXCEEDED:
455			break;
456		default:
457			goto freeit;
458		}
459	}
460
461	icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
462	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
463	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
464		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
465
466	switch (icmp6->icmp6_type) {
467	case ICMP6_DST_UNREACH:
468		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
469		switch (code) {
470		case ICMP6_DST_UNREACH_NOROUTE:
471			code = PRC_UNREACH_NET;
472			break;
473		case ICMP6_DST_UNREACH_ADMIN:
474			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
475			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
476			break;
477		case ICMP6_DST_UNREACH_ADDR:
478			code = PRC_HOSTDEAD;
479			break;
480#ifdef COMPAT_RFC1885
481		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
482			code = PRC_UNREACH_SRCFAIL;
483			break;
484#else
485		case ICMP6_DST_UNREACH_BEYONDSCOPE:
486			/* I mean "source address was incorrect." */
487			code = PRC_PARAMPROB;
488			break;
489#endif
490		case ICMP6_DST_UNREACH_NOPORT:
491			code = PRC_UNREACH_PORT;
492			break;
493		default:
494			goto badcode;
495		}
496		goto deliver;
497		break;
498
499	case ICMP6_PACKET_TOO_BIG:
500		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
501		if (code != 0)
502			goto badcode;
503
504		code = PRC_MSGSIZE;
505
506		/*
507		 * Updating the path MTU will be done after examining
508		 * intermediate extension headers.
509		 */
510		goto deliver;
511		break;
512
513	case ICMP6_TIME_EXCEEDED:
514		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
515		switch (code) {
516		case ICMP6_TIME_EXCEED_TRANSIT:
517		case ICMP6_TIME_EXCEED_REASSEMBLY:
518			code += PRC_TIMXCEED_INTRANS;
519			break;
520		default:
521			goto badcode;
522		}
523		goto deliver;
524		break;
525
526	case ICMP6_PARAM_PROB:
527		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
528		switch (code) {
529		case ICMP6_PARAMPROB_NEXTHEADER:
530			code = PRC_UNREACH_PROTOCOL;
531			break;
532		case ICMP6_PARAMPROB_HEADER:
533		case ICMP6_PARAMPROB_OPTION:
534			code = PRC_PARAMPROB;
535			break;
536		default:
537			goto badcode;
538		}
539		goto deliver;
540		break;
541
542	case ICMP6_ECHO_REQUEST:
543		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
544		if (code != 0)
545			goto badcode;
546		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
547			/* Give up remote */
548			break;
549		}
550		if ((n->m_flags & M_EXT) != 0
551		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
552			struct mbuf *n0 = n;
553			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
554			int n0len;
555
556			/*
557			 * Prepare an internal mbuf.  m_pullup() doesn't
558			 * always copy the length we specified.
559			 */
560			if (maxlen >= MCLBYTES) {
561				/* Give up remote */
562				m_freem(n0);
563				break;
564			}
565			MGETHDR(n, M_DONTWAIT, n0->m_type);
566			n0len = n0->m_pkthdr.len;	/* save for use below */
567			if (n)
568				M_MOVE_PKTHDR(n, n0);
569			if (n && maxlen >= MHLEN) {
570				MCLGET(n, M_DONTWAIT);
571				if ((n->m_flags & M_EXT) == 0) {
572					m_free(n);
573					n = NULL;
574				}
575			}
576			if (n == NULL) {
577				/* Give up remote */
578				m_freem(n0);
579				break;
580			}
581			/*
582			 * Copy IPv6 and ICMPv6 only.
583			 */
584			nip6 = mtod(n, struct ip6_hdr *);
585			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
586			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
587			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
588			noff = sizeof(struct ip6_hdr);
589			/* new mbuf contains only ipv6+icmpv6 headers */
590			n->m_len = noff + sizeof(struct icmp6_hdr);
591			/*
592			 * Adjust mbuf.  ip6_plen will be adjusted in
593			 * ip6_output().
594			 */
595			m_adj(n0, off + sizeof(struct icmp6_hdr));
596			/* recalculate complete packet size */
597			n->m_pkthdr.len = n0len + (noff - off);
598			n->m_next = n0;
599		} else {
600			nip6 = mtod(n, struct ip6_hdr *);
601			nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
602			noff = off;
603		}
604		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
605		nicmp6->icmp6_code = 0;
606		if (n) {
607			icmp6stat.icp6s_reflect++;
608			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
609			icmp6_reflect(n, noff);
610		}
611		break;
612
613	case ICMP6_ECHO_REPLY:
614		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
615		if (code != 0)
616			goto badcode;
617		break;
618
619	case MLD_LISTENER_QUERY:
620	case MLD_LISTENER_REPORT:
621		if (icmp6len < sizeof(struct mld_hdr))
622			goto badlen;
623		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
624			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
625		else
626			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
627		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
628			/* give up local */
629			mld6_input(m, off);
630			m = NULL;
631			goto freeit;
632		}
633		mld6_input(n, off);
634		/* m stays. */
635		break;
636
637	case MLD_LISTENER_DONE:
638		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
639		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
640			goto badlen;
641		break;		/* nothing to be done in kernel */
642
643	case MLD_MTRACE_RESP:
644	case MLD_MTRACE:
645		/* XXX: these two are experimental.  not officially defined. */
646		/* XXX: per-interface statistics? */
647		break;		/* just pass it to applications */
648
649	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
650	    {
651		enum { WRU, FQDN } mode;
652
653		if (!icmp6_nodeinfo)
654			break;
655
656		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
657			mode = WRU;
658		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
659			mode = FQDN;
660		else
661			goto badlen;
662
663#define hostnamelen	strlen(hostname)
664		if (mode == FQDN) {
665#ifndef PULLDOWN_TEST
666			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
667			    IPPROTO_DONE);
668#endif
669			n = m_copy(m, 0, M_COPYALL);
670			if (n)
671				n = ni6_input(n, off);
672			/* XXX meaningless if n == NULL */
673			noff = sizeof(struct ip6_hdr);
674		} else {
675			u_char *p;
676			int maxlen, maxhlen;
677
678			if ((icmp6_nodeinfo & 5) != 5)
679				break;
680
681			if (code != 0)
682				goto badcode;
683			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
684			if (maxlen >= MCLBYTES) {
685				/* Give up remote */
686				break;
687			}
688			MGETHDR(n, M_DONTWAIT, m->m_type);
689			if (n && maxlen > MHLEN) {
690				MCLGET(n, M_DONTWAIT);
691				if ((n->m_flags & M_EXT) == 0) {
692					m_free(n);
693					n = NULL;
694				}
695			}
696			if (!m_dup_pkthdr(n, m, M_DONTWAIT)) {
697				/*
698				 * Previous code did a blind M_COPY_PKTHDR
699				 * and said "just for rcvif".  If true, then
700				 * we could tolerate the dup failing (due to
701				 * the deep copy of the tag chain).  For now
702				 * be conservative and just fail.
703				 */
704				m_free(n);
705				n = NULL;
706			}
707			if (n == NULL) {
708				/* Give up remote */
709				break;
710			}
711			n->m_pkthdr.rcvif = NULL;
712			n->m_len = 0;
713			maxhlen = M_TRAILINGSPACE(n) - maxlen;
714			if (maxhlen > hostnamelen)
715				maxhlen = hostnamelen;
716			/*
717			 * Copy IPv6 and ICMPv6 only.
718			 */
719			nip6 = mtod(n, struct ip6_hdr *);
720			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
721			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
722			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
723			p = (u_char *)(nicmp6 + 1);
724			bzero(p, 4);
725			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
726			noff = sizeof(struct ip6_hdr);
727			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
728				sizeof(struct icmp6_hdr) + 4 + maxhlen;
729			nicmp6->icmp6_type = ICMP6_WRUREPLY;
730			nicmp6->icmp6_code = 0;
731		}
732#undef hostnamelen
733		if (n) {
734			icmp6stat.icp6s_reflect++;
735			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
736			icmp6_reflect(n, noff);
737		}
738		break;
739	    }
740
741	case ICMP6_WRUREPLY:
742		if (code != 0)
743			goto badcode;
744		break;
745
746	case ND_ROUTER_SOLICIT:
747		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
748		if (code != 0)
749			goto badcode;
750		if (icmp6len < sizeof(struct nd_router_solicit))
751			goto badlen;
752		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
753			/* give up local */
754			nd6_rs_input(m, off, icmp6len);
755			m = NULL;
756			goto freeit;
757		}
758		nd6_rs_input(n, off, icmp6len);
759		/* m stays. */
760		break;
761
762	case ND_ROUTER_ADVERT:
763		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
764		if (code != 0)
765			goto badcode;
766		if (icmp6len < sizeof(struct nd_router_advert))
767			goto badlen;
768		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
769			/* give up local */
770			nd6_ra_input(m, off, icmp6len);
771			m = NULL;
772			goto freeit;
773		}
774		nd6_ra_input(n, off, icmp6len);
775		/* m stays. */
776		break;
777
778	case ND_NEIGHBOR_SOLICIT:
779		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
780		if (code != 0)
781			goto badcode;
782		if (icmp6len < sizeof(struct nd_neighbor_solicit))
783			goto badlen;
784		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
785			/* give up local */
786			nd6_ns_input(m, off, icmp6len);
787			m = NULL;
788			goto freeit;
789		}
790		nd6_ns_input(n, off, icmp6len);
791		/* m stays. */
792		break;
793
794	case ND_NEIGHBOR_ADVERT:
795		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
796		if (code != 0)
797			goto badcode;
798		if (icmp6len < sizeof(struct nd_neighbor_advert))
799			goto badlen;
800		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
801			/* give up local */
802			nd6_na_input(m, off, icmp6len);
803			m = NULL;
804			goto freeit;
805		}
806		nd6_na_input(n, off, icmp6len);
807		/* m stays. */
808		break;
809
810	case ND_REDIRECT:
811		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
812		if (code != 0)
813			goto badcode;
814		if (icmp6len < sizeof(struct nd_redirect))
815			goto badlen;
816		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
817			/* give up local */
818			icmp6_redirect_input(m, off);
819			m = NULL;
820			goto freeit;
821		}
822		icmp6_redirect_input(n, off);
823		/* m stays. */
824		break;
825
826	case ICMP6_ROUTER_RENUMBERING:
827		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
828		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
829			goto badcode;
830		if (icmp6len < sizeof(struct icmp6_router_renum))
831			goto badlen;
832		break;
833
834	default:
835		nd6log((LOG_DEBUG,
836		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
837		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
838		    ip6_sprintf(&ip6->ip6_dst),
839		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
840		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
841			/* ICMPv6 error: MUST deliver it by spec... */
842			code = PRC_NCMDS;
843			/* deliver */
844		} else {
845			/* ICMPv6 informational: MUST not deliver */
846			break;
847		}
848	deliver:
849		if (icmp6_notify_error(m, off, icmp6len, code)) {
850			/* In this case, m should've been freed. */
851			return (IPPROTO_DONE);
852		}
853		break;
854
855	badcode:
856		icmp6stat.icp6s_badcode++;
857		break;
858
859	badlen:
860		icmp6stat.icp6s_badlen++;
861		break;
862	}
863
864	/* deliver the packet to appropriate sockets */
865	icmp6_rip6_input(&m, *offp);
866
867	return IPPROTO_DONE;
868
869 freeit:
870	m_freem(m);
871	return IPPROTO_DONE;
872}
873
874static int
875icmp6_notify_error(m, off, icmp6len, code)
876	struct mbuf *m;
877	int off, icmp6len, code;
878{
879	struct icmp6_hdr *icmp6;
880	struct ip6_hdr *eip6;
881	u_int32_t notifymtu;
882	struct sockaddr_in6 icmp6src, icmp6dst;
883
884	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
885		icmp6stat.icp6s_tooshort++;
886		goto freeit;
887	}
888#ifndef PULLDOWN_TEST
889	IP6_EXTHDR_CHECK(m, off,
890	    sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
891	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
892#else
893	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
894	    sizeof(*icmp6) + sizeof(struct ip6_hdr));
895	if (icmp6 == NULL) {
896		icmp6stat.icp6s_tooshort++;
897		return (-1);
898	}
899#endif
900	eip6 = (struct ip6_hdr *)(icmp6 + 1);
901
902	/* Detect the upper level protocol */
903	{
904		void (*ctlfunc) __P((int, struct sockaddr *, void *));
905		u_int8_t nxt = eip6->ip6_nxt;
906		int eoff = off + sizeof(struct icmp6_hdr) +
907		    sizeof(struct ip6_hdr);
908		struct ip6ctlparam ip6cp;
909		struct in6_addr *finaldst = NULL;
910		int icmp6type = icmp6->icmp6_type;
911		struct ip6_frag *fh;
912		struct ip6_rthdr *rth;
913		struct ip6_rthdr0 *rth0;
914		int rthlen;
915
916		while (1) { /* XXX: should avoid infinite loop explicitly? */
917			struct ip6_ext *eh;
918
919			switch (nxt) {
920			case IPPROTO_HOPOPTS:
921			case IPPROTO_DSTOPTS:
922			case IPPROTO_AH:
923#ifndef PULLDOWN_TEST
924				IP6_EXTHDR_CHECK(m, 0,
925				    eoff + sizeof(struct ip6_ext), -1);
926				eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
927#else
928				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
929				    eoff, sizeof(*eh));
930				if (eh == NULL) {
931					icmp6stat.icp6s_tooshort++;
932					return (-1);
933				}
934#endif
935
936				if (nxt == IPPROTO_AH)
937					eoff += (eh->ip6e_len + 2) << 2;
938				else
939					eoff += (eh->ip6e_len + 1) << 3;
940				nxt = eh->ip6e_nxt;
941				break;
942			case IPPROTO_ROUTING:
943				/*
944				 * When the erroneous packet contains a
945				 * routing header, we should examine the
946				 * header to determine the final destination.
947				 * Otherwise, we can't properly update
948				 * information that depends on the final
949				 * destination (e.g. path MTU).
950				 */
951#ifndef PULLDOWN_TEST
952				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
953				rth = (struct ip6_rthdr *)
954				    (mtod(m, caddr_t) + eoff);
955#else
956				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
957				    eoff, sizeof(*rth));
958				if (rth == NULL) {
959					icmp6stat.icp6s_tooshort++;
960					return (-1);
961				}
962#endif
963				rthlen = (rth->ip6r_len + 1) << 3;
964				/*
965				 * XXX: currently there is no
966				 * officially defined type other
967				 * than type-0.
968				 * Note that if the segment left field
969				 * is 0, all intermediate hops must
970				 * have been passed.
971				 */
972				if (rth->ip6r_segleft &&
973				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
974					int hops;
975
976#ifndef PULLDOWN_TEST
977					IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
978					rth0 = (struct ip6_rthdr0 *)
979					    (mtod(m, caddr_t) + eoff);
980#else
981					IP6_EXTHDR_GET(rth0,
982					    struct ip6_rthdr0 *, m,
983					    eoff, rthlen);
984					if (rth0 == NULL) {
985						icmp6stat.icp6s_tooshort++;
986						return (-1);
987					}
988#endif
989					/* just ignore a bogus header */
990					if ((rth0->ip6r0_len % 2) == 0 &&
991					    (hops = rth0->ip6r0_len/2))
992						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
993				}
994				eoff += rthlen;
995				nxt = rth->ip6r_nxt;
996				break;
997			case IPPROTO_FRAGMENT:
998#ifndef PULLDOWN_TEST
999				IP6_EXTHDR_CHECK(m, 0, eoff +
1000				    sizeof(struct ip6_frag), -1);
1001				fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1002				    eoff);
1003#else
1004				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1005				    eoff, sizeof(*fh));
1006				if (fh == NULL) {
1007					icmp6stat.icp6s_tooshort++;
1008					return (-1);
1009				}
1010#endif
1011				/*
1012				 * Data after a fragment header is meaningless
1013				 * unless it is the first fragment, but
1014				 * we'll go to the notify label for path MTU
1015				 * discovery.
1016				 */
1017				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1018					goto notify;
1019
1020				eoff += sizeof(struct ip6_frag);
1021				nxt = fh->ip6f_nxt;
1022				break;
1023			default:
1024				/*
1025				 * This case includes ESP and the No Next
1026				 * Header.  In such cases going to the notify
1027				 * label does not have any meaning
1028				 * (i.e. ctlfunc will be NULL), but we go
1029				 * anyway since we might have to update
1030				 * path MTU information.
1031				 */
1032				goto notify;
1033			}
1034		}
1035	  notify:
1036#ifndef PULLDOWN_TEST
1037		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1038#else
1039		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1040		    sizeof(*icmp6) + sizeof(struct ip6_hdr));
1041		if (icmp6 == NULL) {
1042			icmp6stat.icp6s_tooshort++;
1043			return (-1);
1044		}
1045#endif
1046
1047		/*
1048		 * retrieve parameters from the inner IPv6 header, and convert
1049		 * them into sockaddr structures.
1050		 * XXX: there is no guarantee that the source or destination
1051		 * addresses of the inner packet are in the same scope as
1052		 * the addresses of the icmp packet.  But there is no other
1053		 * way to determine the zone.
1054		 */
1055		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1056
1057		bzero(&icmp6dst, sizeof(icmp6dst));
1058		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1059		icmp6dst.sin6_family = AF_INET6;
1060		if (finaldst == NULL)
1061			icmp6dst.sin6_addr = eip6->ip6_dst;
1062		else
1063			icmp6dst.sin6_addr = *finaldst;
1064		icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1065							  &icmp6dst.sin6_addr);
1066		if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1067				   NULL, NULL)) {
1068			/* should be impossbile */
1069			nd6log((LOG_DEBUG,
1070			    "icmp6_notify_error: in6_embedscope failed\n"));
1071			goto freeit;
1072		}
1073
1074		/*
1075		 * retrieve parameters from the inner IPv6 header, and convert
1076		 * them into sockaddr structures.
1077		 */
1078		bzero(&icmp6src, sizeof(icmp6src));
1079		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1080		icmp6src.sin6_family = AF_INET6;
1081		icmp6src.sin6_addr = eip6->ip6_src;
1082		icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1083							  &icmp6src.sin6_addr);
1084		if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1085				   NULL, NULL)) {
1086			/* should be impossbile */
1087			nd6log((LOG_DEBUG,
1088			    "icmp6_notify_error: in6_embedscope failed\n"));
1089			goto freeit;
1090		}
1091		icmp6src.sin6_flowinfo = (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1092
1093		if (finaldst == NULL)
1094			finaldst = &eip6->ip6_dst;
1095		ip6cp.ip6c_m = m;
1096		ip6cp.ip6c_icmp6 = icmp6;
1097		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1098		ip6cp.ip6c_off = eoff;
1099		ip6cp.ip6c_finaldst = finaldst;
1100		ip6cp.ip6c_src = &icmp6src;
1101		ip6cp.ip6c_nxt = nxt;
1102
1103		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1104			notifymtu = ntohl(icmp6->icmp6_mtu);
1105			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1106			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1107		}
1108
1109		ctlfunc = (void (*) __P((int, struct sockaddr *, void *)))
1110		    (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1111		if (ctlfunc) {
1112			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1113			    &ip6cp);
1114		}
1115	}
1116	return (0);
1117
1118  freeit:
1119	m_freem(m);
1120	return (-1);
1121}
1122
1123void
1124icmp6_mtudisc_update(ip6cp, validated)
1125	struct ip6ctlparam *ip6cp;
1126	int validated;
1127{
1128	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1129	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1130	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1131	u_int mtu = ntohl(icmp6->icmp6_mtu);
1132	struct rtentry *rt = NULL;
1133	struct sockaddr_in6 sin6;
1134
1135	if (!validated)
1136		return;
1137
1138	bzero(&sin6, sizeof(sin6));
1139	sin6.sin6_family = PF_INET6;
1140	sin6.sin6_len = sizeof(struct sockaddr_in6);
1141	sin6.sin6_addr = *dst;
1142	/* XXX normally, this won't happen */
1143	if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1144		sin6.sin6_addr.s6_addr16[1] =
1145		    htons(m->m_pkthdr.rcvif->if_index);
1146	}
1147	/* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1148	rt = rtalloc1((struct sockaddr *)&sin6, 0,
1149		      RTF_CLONING | RTF_PRCLONING);
1150
1151	if (rt && (rt->rt_flags & RTF_HOST) &&
1152	    !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1153		if (mtu < IPV6_MMTU) {
1154				/* xxx */
1155			rt->rt_rmx.rmx_locks |= RTV_MTU;
1156		} else if (mtu < rt->rt_ifp->if_mtu &&
1157			   rt->rt_rmx.rmx_mtu > mtu) {
1158			icmp6stat.icp6s_pmtuchg++;
1159			rt->rt_rmx.rmx_mtu = mtu;
1160		}
1161	}
1162	if (rt)
1163		rtfree(rt);
1164}
1165
1166/*
1167 * Process a Node Information Query packet, based on
1168 * draft-ietf-ipngwg-icmp-name-lookups-07.
1169 *
1170 * Spec incompatibilities:
1171 * - IPv6 Subject address handling
1172 * - IPv4 Subject address handling support missing
1173 * - Proxy reply (answer even if it's not for me)
1174 * - joins NI group address at in6_ifattach() time only, does not cope
1175 *   with hostname changes by sethostname(3)
1176 */
1177#define hostnamelen	strlen(hostname)
1178static struct mbuf *
1179ni6_input(m, off)
1180	struct mbuf *m;
1181	int off;
1182{
1183	struct icmp6_nodeinfo *ni6, *nni6;
1184	struct mbuf *n = NULL;
1185	u_int16_t qtype;
1186	int subjlen;
1187	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1188	struct ni_reply_fqdn *fqdn;
1189	int addrs;		/* for NI_QTYPE_NODEADDR */
1190	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1191	struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1192	struct sockaddr_in6 sin6_d; /* XXX: we should retrieve this from m_aux */
1193	struct ip6_hdr *ip6;
1194	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1195	char *subj = NULL;
1196	struct in6_ifaddr *ia6 = NULL;
1197
1198	ip6 = mtod(m, struct ip6_hdr *);
1199#ifndef PULLDOWN_TEST
1200	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1201#else
1202	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1203	if (ni6 == NULL) {
1204		/* m is already reclaimed */
1205		return (NULL);
1206	}
1207#endif
1208
1209	/*
1210	 * Validate IPv6 destination address.
1211	 *
1212	 * The Responder must discard the Query without further processing
1213	 * unless it is one of the Responder's unicast or anycast addresses, or
1214	 * a link-local scope multicast address which the Responder has joined.
1215	 * [icmp-name-lookups-07, Section 4.]
1216	 */
1217	bzero(&sin6, sizeof(sin6));
1218	sin6.sin6_family = AF_INET6;
1219	sin6.sin6_len = sizeof(struct sockaddr_in6);
1220	bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1221	/* XXX scopeid */
1222	if ((ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)&sin6)) != NULL) {
1223		/* unicast/anycast, fine */
1224		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1225		    (icmp6_nodeinfo & 4) == 0) {
1226			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1227				"a temporary address in %s:%d",
1228			       __FILE__, __LINE__));
1229			goto bad;
1230		}
1231	} else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1232		; /* link-local multicast, fine */
1233	else
1234		goto bad;
1235
1236	/* validate query Subject field. */
1237	qtype = ntohs(ni6->ni_qtype);
1238	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1239	switch (qtype) {
1240	case NI_QTYPE_NOOP:
1241	case NI_QTYPE_SUPTYPES:
1242		/* 07 draft */
1243		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1244			break;
1245		/* FALLTHROUGH */
1246	case NI_QTYPE_FQDN:
1247	case NI_QTYPE_NODEADDR:
1248		switch (ni6->ni_code) {
1249		case ICMP6_NI_SUBJ_IPV6:
1250#if ICMP6_NI_SUBJ_IPV6 != 0
1251		case 0:
1252#endif
1253			/*
1254			 * backward compatibility - try to accept 03 draft
1255			 * format, where no Subject is present.
1256			 */
1257			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1258			    subjlen == 0) {
1259				oldfqdn++;
1260				break;
1261			}
1262#if ICMP6_NI_SUBJ_IPV6 != 0
1263			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1264				goto bad;
1265#endif
1266
1267			if (subjlen != sizeof(sin6.sin6_addr))
1268				goto bad;
1269
1270			/*
1271			 * Validate Subject address.
1272			 *
1273			 * Not sure what exactly "address belongs to the node"
1274			 * means in the spec, is it just unicast, or what?
1275			 *
1276			 * At this moment we consider Subject address as
1277			 * "belong to the node" if the Subject address equals
1278			 * to the IPv6 destination address; validation for
1279			 * IPv6 destination address should have done enough
1280			 * check for us.
1281			 *
1282			 * We do not do proxy at this moment.
1283			 */
1284			/* m_pulldown instead of copy? */
1285			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1286			    subjlen, (caddr_t)&sin6.sin6_addr);
1287			sin6.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1288							      &sin6.sin6_addr);
1289			in6_embedscope(&sin6.sin6_addr, &sin6, NULL, NULL);
1290			bzero(&sin6_d, sizeof(sin6_d));
1291			sin6_d.sin6_family = AF_INET6; /* not used, actually */
1292			sin6_d.sin6_len = sizeof(sin6_d); /* ditto */
1293			sin6_d.sin6_addr = ip6->ip6_dst;
1294			sin6_d.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1295								&ip6->ip6_dst);
1296			in6_embedscope(&sin6_d.sin6_addr, &sin6_d, NULL, NULL);
1297			subj = (char *)&sin6;
1298			if (SA6_ARE_ADDR_EQUAL(&sin6, &sin6_d))
1299				break;
1300
1301			/*
1302			 * XXX if we are to allow other cases, we should really
1303			 * be careful about scope here.
1304			 * basically, we should disallow queries toward IPv6
1305			 * destination X with subject Y,
1306			 * if scope(X) > scope(Y).
1307			 * if we allow scope(X) > scope(Y), it will result in
1308			 * information leakage across scope boundary.
1309			 */
1310			goto bad;
1311
1312		case ICMP6_NI_SUBJ_FQDN:
1313			/*
1314			 * Validate Subject name with gethostname(3).
1315			 *
1316			 * The behavior may need some debate, since:
1317			 * - we are not sure if the node has FQDN as
1318			 *   hostname (returned by gethostname(3)).
1319			 * - the code does wildcard match for truncated names.
1320			 *   however, we are not sure if we want to perform
1321			 *   wildcard match, if gethostname(3) side has
1322			 *   truncated hostname.
1323			 */
1324			n = ni6_nametodns(hostname, hostnamelen, 0);
1325			if (!n || n->m_next || n->m_len == 0)
1326				goto bad;
1327			IP6_EXTHDR_GET(subj, char *, m,
1328			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1329			if (subj == NULL)
1330				goto bad;
1331			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1332			    n->m_len)) {
1333				goto bad;
1334			}
1335			m_freem(n);
1336			n = NULL;
1337			break;
1338
1339		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1340		default:
1341			goto bad;
1342		}
1343		break;
1344	}
1345
1346	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1347	switch (qtype) {
1348	case NI_QTYPE_FQDN:
1349		if ((icmp6_nodeinfo & 1) == 0)
1350			goto bad;
1351		break;
1352	case NI_QTYPE_NODEADDR:
1353		if ((icmp6_nodeinfo & 2) == 0)
1354			goto bad;
1355		break;
1356	}
1357
1358	/* guess reply length */
1359	switch (qtype) {
1360	case NI_QTYPE_NOOP:
1361		break;		/* no reply data */
1362	case NI_QTYPE_SUPTYPES:
1363		replylen += sizeof(u_int32_t);
1364		break;
1365	case NI_QTYPE_FQDN:
1366		/* XXX will append an mbuf */
1367		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1368		break;
1369	case NI_QTYPE_NODEADDR:
1370		addrs = ni6_addrs(ni6, m, &ifp, subj);
1371		if ((replylen += addrs * (sizeof(struct in6_addr) +
1372		    sizeof(u_int32_t))) > MCLBYTES)
1373			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1374		break;
1375	default:
1376		/*
1377		 * XXX: We must return a reply with the ICMP6 code
1378		 * `unknown Qtype' in this case.  However we regard the case
1379		 * as an FQDN query for backward compatibility.
1380		 * Older versions set a random value to this field,
1381		 * so it rarely varies in the defined qtypes.
1382		 * But the mechanism is not reliable...
1383		 * maybe we should obsolete older versions.
1384		 */
1385		qtype = NI_QTYPE_FQDN;
1386		/* XXX will append an mbuf */
1387		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1388		oldfqdn++;
1389		break;
1390	}
1391
1392	/* allocate an mbuf to reply. */
1393	MGETHDR(n, M_DONTWAIT, m->m_type);
1394	if (n == NULL) {
1395		m_freem(m);
1396		return (NULL);
1397	}
1398	M_MOVE_PKTHDR(n, m); /* just for recvif */
1399	if (replylen > MHLEN) {
1400		if (replylen > MCLBYTES) {
1401			/*
1402			 * XXX: should we try to allocate more? But MCLBYTES
1403			 * is probably much larger than IPV6_MMTU...
1404			 */
1405			goto bad;
1406		}
1407		MCLGET(n, M_DONTWAIT);
1408		if ((n->m_flags & M_EXT) == 0) {
1409			goto bad;
1410		}
1411	}
1412	n->m_pkthdr.len = n->m_len = replylen;
1413
1414	/* copy mbuf header and IPv6 + Node Information base headers */
1415	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1416	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1417	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1418
1419	/* qtype dependent procedure */
1420	switch (qtype) {
1421	case NI_QTYPE_NOOP:
1422		nni6->ni_code = ICMP6_NI_SUCCESS;
1423		nni6->ni_flags = 0;
1424		break;
1425	case NI_QTYPE_SUPTYPES:
1426	{
1427		u_int32_t v;
1428		nni6->ni_code = ICMP6_NI_SUCCESS;
1429		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1430		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1431		v = (u_int32_t)htonl(0x0000000f);
1432		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1433		break;
1434	}
1435	case NI_QTYPE_FQDN:
1436		nni6->ni_code = ICMP6_NI_SUCCESS;
1437		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1438		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1439		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1440		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1441		/*
1442		 * XXX do we really have FQDN in variable "hostname"?
1443		 */
1444		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1445		if (n->m_next == NULL)
1446			goto bad;
1447		/* XXX we assume that n->m_next is not a chain */
1448		if (n->m_next->m_next != NULL)
1449			goto bad;
1450		n->m_pkthdr.len += n->m_next->m_len;
1451		break;
1452	case NI_QTYPE_NODEADDR:
1453	{
1454		int lenlim, copied;
1455
1456		nni6->ni_code = ICMP6_NI_SUCCESS;
1457		n->m_pkthdr.len = n->m_len =
1458		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1459		lenlim = M_TRAILINGSPACE(n);
1460		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1461		/* XXX: reset mbuf length */
1462		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1463		    sizeof(struct icmp6_nodeinfo) + copied;
1464		break;
1465	}
1466	default:
1467		break;		/* XXX impossible! */
1468	}
1469
1470	nni6->ni_type = ICMP6_NI_REPLY;
1471	m_freem(m);
1472	return (n);
1473
1474  bad:
1475	m_freem(m);
1476	if (n)
1477		m_freem(n);
1478	return (NULL);
1479}
1480#undef hostnamelen
1481
1482/*
1483 * make a mbuf with DNS-encoded string.  no compression support.
1484 *
1485 * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1486 * treated as truncated name (two \0 at the end).  this is a wild guess.
1487 */
1488static struct mbuf *
1489ni6_nametodns(name, namelen, old)
1490	const char *name;
1491	int namelen;
1492	int old;	/* return pascal string if non-zero */
1493{
1494	struct mbuf *m;
1495	char *cp, *ep;
1496	const char *p, *q;
1497	int i, len, nterm;
1498
1499	if (old)
1500		len = namelen + 1;
1501	else
1502		len = MCLBYTES;
1503
1504	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1505	MGET(m, M_DONTWAIT, MT_DATA);
1506	if (m && len > MLEN) {
1507		MCLGET(m, M_DONTWAIT);
1508		if ((m->m_flags & M_EXT) == 0)
1509			goto fail;
1510	}
1511	if (!m)
1512		goto fail;
1513	m->m_next = NULL;
1514
1515	if (old) {
1516		m->m_len = len;
1517		*mtod(m, char *) = namelen;
1518		bcopy(name, mtod(m, char *) + 1, namelen);
1519		return m;
1520	} else {
1521		m->m_len = 0;
1522		cp = mtod(m, char *);
1523		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1524
1525		/* if not certain about my name, return empty buffer */
1526		if (namelen == 0)
1527			return m;
1528
1529		/*
1530		 * guess if it looks like shortened hostname, or FQDN.
1531		 * shortened hostname needs two trailing "\0".
1532		 */
1533		i = 0;
1534		for (p = name; p < name + namelen; p++) {
1535			if (*p && *p == '.')
1536				i++;
1537		}
1538		if (i < 2)
1539			nterm = 2;
1540		else
1541			nterm = 1;
1542
1543		p = name;
1544		while (cp < ep && p < name + namelen) {
1545			i = 0;
1546			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1547				i++;
1548			/* result does not fit into mbuf */
1549			if (cp + i + 1 >= ep)
1550				goto fail;
1551			/*
1552			 * DNS label length restriction, RFC1035 page 8.
1553			 * "i == 0" case is included here to avoid returning
1554			 * 0-length label on "foo..bar".
1555			 */
1556			if (i <= 0 || i >= 64)
1557				goto fail;
1558			*cp++ = i;
1559			bcopy(p, cp, i);
1560			cp += i;
1561			p = q;
1562			if (p < name + namelen && *p == '.')
1563				p++;
1564		}
1565		/* termination */
1566		if (cp + nterm >= ep)
1567			goto fail;
1568		while (nterm-- > 0)
1569			*cp++ = '\0';
1570		m->m_len = cp - mtod(m, char *);
1571		return m;
1572	}
1573
1574	panic("should not reach here");
1575	/* NOTREACHED */
1576
1577 fail:
1578	if (m)
1579		m_freem(m);
1580	return NULL;
1581}
1582
1583/*
1584 * check if two DNS-encoded string matches.  takes care of truncated
1585 * form (with \0\0 at the end).  no compression support.
1586 * XXX upper/lowercase match (see RFC2065)
1587 */
1588static int
1589ni6_dnsmatch(a, alen, b, blen)
1590	const char *a;
1591	int alen;
1592	const char *b;
1593	int blen;
1594{
1595	const char *a0, *b0;
1596	int l;
1597
1598	/* simplest case - need validation? */
1599	if (alen == blen && bcmp(a, b, alen) == 0)
1600		return 1;
1601
1602	a0 = a;
1603	b0 = b;
1604
1605	/* termination is mandatory */
1606	if (alen < 2 || blen < 2)
1607		return 0;
1608	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1609		return 0;
1610	alen--;
1611	blen--;
1612
1613	while (a - a0 < alen && b - b0 < blen) {
1614		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1615			return 0;
1616
1617		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1618			return 0;
1619		/* we don't support compression yet */
1620		if (a[0] >= 64 || b[0] >= 64)
1621			return 0;
1622
1623		/* truncated case */
1624		if (a[0] == 0 && a - a0 == alen - 1)
1625			return 1;
1626		if (b[0] == 0 && b - b0 == blen - 1)
1627			return 1;
1628		if (a[0] == 0 || b[0] == 0)
1629			return 0;
1630
1631		if (a[0] != b[0])
1632			return 0;
1633		l = a[0];
1634		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1635			return 0;
1636		if (bcmp(a + 1, b + 1, l) != 0)
1637			return 0;
1638
1639		a += 1 + l;
1640		b += 1 + l;
1641	}
1642
1643	if (a - a0 == alen && b - b0 == blen)
1644		return 1;
1645	else
1646		return 0;
1647}
1648
1649/*
1650 * calculate the number of addresses to be returned in the node info reply.
1651 */
1652static int
1653ni6_addrs(ni6, m, ifpp, subj)
1654	struct icmp6_nodeinfo *ni6;
1655	struct mbuf *m;
1656	struct ifnet **ifpp;
1657	char *subj;
1658{
1659	struct ifnet *ifp;
1660	struct in6_ifaddr *ifa6;
1661	struct ifaddr *ifa;
1662	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1663	int addrs = 0, addrsofif, iffound = 0;
1664	int niflags = ni6->ni_flags;
1665
1666	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1667		switch (ni6->ni_code) {
1668		case ICMP6_NI_SUBJ_IPV6:
1669			if (subj == NULL) /* must be impossible... */
1670				return (0);
1671			subj_ip6 = (struct sockaddr_in6 *)subj;
1672			break;
1673		default:
1674			/*
1675			 * XXX: we only support IPv6 subject address for
1676			 * this Qtype.
1677			 */
1678			return (0);
1679		}
1680	}
1681
1682	IFNET_RLOCK();
1683	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1684		addrsofif = 0;
1685		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1686			if (ifa->ifa_addr->sa_family != AF_INET6)
1687				continue;
1688			ifa6 = (struct in6_ifaddr *)ifa;
1689
1690			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1691			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1692			    &ifa6->ia_addr.sin6_addr))
1693				iffound = 1;
1694
1695			/*
1696			 * IPv4-mapped addresses can only be returned by a
1697			 * Node Information proxy, since they represent
1698			 * addresses of IPv4-only nodes, which perforce do
1699			 * not implement this protocol.
1700			 * [icmp-name-lookups-07, Section 5.4]
1701			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1702			 * this function at this moment.
1703			 */
1704
1705			/* What do we have to do about ::1? */
1706			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1707			case IPV6_ADDR_SCOPE_LINKLOCAL:
1708				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1709					continue;
1710				break;
1711			case IPV6_ADDR_SCOPE_SITELOCAL:
1712				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1713					continue;
1714				break;
1715			case IPV6_ADDR_SCOPE_GLOBAL:
1716				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1717					continue;
1718				break;
1719			default:
1720				continue;
1721			}
1722
1723			/*
1724			 * check if anycast is okay.
1725			 * XXX: just experimental.  not in the spec.
1726			 */
1727			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1728			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1729				continue; /* we need only unicast addresses */
1730			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1731			    (icmp6_nodeinfo & 4) == 0) {
1732				continue;
1733			}
1734			addrsofif++; /* count the address */
1735		}
1736		if (iffound) {
1737			*ifpp = ifp;
1738			IFNET_RUNLOCK();
1739			return (addrsofif);
1740		}
1741
1742		addrs += addrsofif;
1743	}
1744	IFNET_RUNLOCK();
1745
1746	return (addrs);
1747}
1748
1749static int
1750ni6_store_addrs(ni6, nni6, ifp0, resid)
1751	struct icmp6_nodeinfo *ni6, *nni6;
1752	struct ifnet *ifp0;
1753	int resid;
1754{
1755	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1756	struct in6_ifaddr *ifa6;
1757	struct ifaddr *ifa;
1758	struct ifnet *ifp_dep = NULL;
1759	int copied = 0, allow_deprecated = 0;
1760	u_char *cp = (u_char *)(nni6 + 1);
1761	int niflags = ni6->ni_flags;
1762	u_int32_t ltime;
1763
1764	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1765		return (0);	/* needless to copy */
1766
1767	IFNET_RLOCK();
1768  again:
1769
1770	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1771		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1772		     ifa = ifa->ifa_list.tqe_next) {
1773			if (ifa->ifa_addr->sa_family != AF_INET6)
1774				continue;
1775			ifa6 = (struct in6_ifaddr *)ifa;
1776
1777			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1778			    allow_deprecated == 0) {
1779				/*
1780				 * prefererred address should be put before
1781				 * deprecated addresses.
1782				 */
1783
1784				/* record the interface for later search */
1785				if (ifp_dep == NULL)
1786					ifp_dep = ifp;
1787
1788				continue;
1789			} else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1790			    allow_deprecated != 0)
1791				continue; /* we now collect deprecated addrs */
1792
1793			/* What do we have to do about ::1? */
1794			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1795			case IPV6_ADDR_SCOPE_LINKLOCAL:
1796				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1797					continue;
1798				break;
1799			case IPV6_ADDR_SCOPE_SITELOCAL:
1800				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1801					continue;
1802				break;
1803			case IPV6_ADDR_SCOPE_GLOBAL:
1804				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1805					continue;
1806				break;
1807			default:
1808				continue;
1809			}
1810
1811			/*
1812			 * check if anycast is okay.
1813			 * XXX: just experimental.  not in the spec.
1814			 */
1815			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1816			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1817				continue;
1818			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1819			    (icmp6_nodeinfo & 4) == 0) {
1820				continue;
1821			}
1822
1823			/* now we can copy the address */
1824			if (resid < sizeof(struct in6_addr) +
1825			    sizeof(u_int32_t)) {
1826				/*
1827				 * We give up much more copy.
1828				 * Set the truncate flag and return.
1829				 */
1830				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1831				IFNET_RUNLOCK();
1832				return (copied);
1833			}
1834
1835			/*
1836			 * Set the TTL of the address.
1837			 * The TTL value should be one of the following
1838			 * according to the specification:
1839			 *
1840			 * 1. The remaining lifetime of a DHCP lease on the
1841			 *    address, or
1842			 * 2. The remaining Valid Lifetime of a prefix from
1843			 *    which the address was derived through Stateless
1844			 *    Autoconfiguration.
1845			 *
1846			 * Note that we currently do not support stateful
1847			 * address configuration by DHCPv6, so the former
1848			 * case can't happen.
1849			 */
1850			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1851				ltime = ND6_INFINITE_LIFETIME;
1852			else {
1853				if (ifa6->ia6_lifetime.ia6t_expire >
1854				    time_second)
1855					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1856				else
1857					ltime = 0;
1858			}
1859
1860			bcopy(&ltime, cp, sizeof(u_int32_t));
1861			cp += sizeof(u_int32_t);
1862
1863			/* copy the address itself */
1864			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1865			    sizeof(struct in6_addr));
1866			/* XXX: KAME link-local hack; remove ifindex */
1867			if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr))
1868				((struct in6_addr *)cp)->s6_addr16[1] = 0;
1869			cp += sizeof(struct in6_addr);
1870
1871			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1872			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1873		}
1874		if (ifp0)	/* we need search only on the specified IF */
1875			break;
1876	}
1877
1878	if (allow_deprecated == 0 && ifp_dep != NULL) {
1879		ifp = ifp_dep;
1880		allow_deprecated = 1;
1881
1882		goto again;
1883	}
1884
1885	IFNET_RUNLOCK();
1886
1887	return (copied);
1888}
1889
1890/*
1891 * XXX almost dup'ed code with rip6_input.
1892 */
1893static int
1894icmp6_rip6_input(mp, off)
1895	struct	mbuf **mp;
1896	int	off;
1897{
1898	struct mbuf *m = *mp;
1899	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1900	struct in6pcb *in6p;
1901	struct in6pcb *last = NULL;
1902	struct sockaddr_in6 rip6src;
1903	struct icmp6_hdr *icmp6;
1904	struct mbuf *opts = NULL;
1905
1906#ifndef PULLDOWN_TEST
1907	/* this is assumed to be safe. */
1908	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1909#else
1910	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1911	if (icmp6 == NULL) {
1912		/* m is already reclaimed */
1913		return (IPPROTO_DONE);
1914	}
1915#endif
1916
1917	bzero(&rip6src, sizeof(rip6src));
1918	rip6src.sin6_len = sizeof(struct sockaddr_in6);
1919	rip6src.sin6_family = AF_INET6;
1920	/* KAME hack: recover scopeid */
1921	(void)in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif);
1922
1923	LIST_FOREACH(in6p, &ripcb, inp_list) {
1924		if ((in6p->inp_vflag & INP_IPV6) == 0)
1925			continue;
1926#ifdef HAVE_NRL_INPCB
1927		if (!(in6p->in6p_flags & INP_IPV6))
1928			continue;
1929#endif
1930		if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1931			continue;
1932		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1933		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1934			continue;
1935		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1936		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1937			continue;
1938		if (in6p->in6p_icmp6filt
1939		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1940				 in6p->in6p_icmp6filt))
1941			continue;
1942		if (last) {
1943			struct	mbuf *n;
1944			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1945				if (last->in6p_flags & IN6P_CONTROLOPTS)
1946					ip6_savecontrol(last, &opts, ip6, n);
1947				/* strip intermediate headers */
1948				m_adj(n, off);
1949				if (sbappendaddr(&last->in6p_socket->so_rcv,
1950				    (struct sockaddr *)&rip6src, n, opts)
1951				    == 0) {
1952					/* should notify about lost packet */
1953					m_freem(n);
1954					if (opts) {
1955						m_freem(opts);
1956					}
1957				} else
1958					sorwakeup(last->in6p_socket);
1959				opts = NULL;
1960			}
1961		}
1962		last = in6p;
1963	}
1964	if (last) {
1965		if (last->in6p_flags & IN6P_CONTROLOPTS)
1966			ip6_savecontrol(last, &opts, ip6, m);
1967		/* strip intermediate headers */
1968		m_adj(m, off);
1969		if (sbappendaddr(&last->in6p_socket->so_rcv,
1970		    (struct sockaddr *)&rip6src, m, opts) == 0) {
1971			m_freem(m);
1972			if (opts)
1973				m_freem(opts);
1974		} else
1975			sorwakeup(last->in6p_socket);
1976	} else {
1977		m_freem(m);
1978		ip6stat.ip6s_delivered--;
1979	}
1980	return IPPROTO_DONE;
1981}
1982
1983/*
1984 * Reflect the ip6 packet back to the source.
1985 * OFF points to the icmp6 header, counted from the top of the mbuf.
1986 */
1987void
1988icmp6_reflect(m, off)
1989	struct	mbuf *m;
1990	size_t off;
1991{
1992	struct ip6_hdr *ip6;
1993	struct icmp6_hdr *icmp6;
1994	struct in6_ifaddr *ia;
1995	struct in6_addr t, *src = 0;
1996	int plen;
1997	int type, code;
1998	struct ifnet *outif = NULL;
1999	struct sockaddr_in6 sa6_src, sa6_dst;
2000#ifdef COMPAT_RFC1885
2001	int mtu = IPV6_MMTU;
2002	struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
2003#endif
2004
2005	/* too short to reflect */
2006	if (off < sizeof(struct ip6_hdr)) {
2007		nd6log((LOG_DEBUG,
2008		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2009		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2010		    __FILE__, __LINE__));
2011		goto bad;
2012	}
2013
2014	/*
2015	 * If there are extra headers between IPv6 and ICMPv6, strip
2016	 * off that header first.
2017	 */
2018#ifdef DIAGNOSTIC
2019	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2020		panic("assumption failed in icmp6_reflect");
2021#endif
2022	if (off > sizeof(struct ip6_hdr)) {
2023		size_t l;
2024		struct ip6_hdr nip6;
2025
2026		l = off - sizeof(struct ip6_hdr);
2027		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2028		m_adj(m, l);
2029		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2030		if (m->m_len < l) {
2031			if ((m = m_pullup(m, l)) == NULL)
2032				return;
2033		}
2034		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2035	} else /* off == sizeof(struct ip6_hdr) */ {
2036		size_t l;
2037		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2038		if (m->m_len < l) {
2039			if ((m = m_pullup(m, l)) == NULL)
2040				return;
2041		}
2042	}
2043	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2044	ip6 = mtod(m, struct ip6_hdr *);
2045	ip6->ip6_nxt = IPPROTO_ICMPV6;
2046	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2047	type = icmp6->icmp6_type; /* keep type for statistics */
2048	code = icmp6->icmp6_code; /* ditto. */
2049
2050	t = ip6->ip6_dst;
2051	/*
2052	 * ip6_input() drops a packet if its src is multicast.
2053	 * So, the src is never multicast.
2054	 */
2055	ip6->ip6_dst = ip6->ip6_src;
2056
2057	/*
2058	 * XXX: make sure to embed scope zone information, using
2059	 * already embedded IDs or the received interface (if any).
2060	 * Note that rcvif may be NULL.
2061	 * TODO: scoped routing case (XXX).
2062	 */
2063	bzero(&sa6_src, sizeof(sa6_src));
2064	sa6_src.sin6_family = AF_INET6;
2065	sa6_src.sin6_len = sizeof(sa6_src);
2066	sa6_src.sin6_addr = ip6->ip6_dst;
2067	in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
2068	in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL);
2069	bzero(&sa6_dst, sizeof(sa6_dst));
2070	sa6_dst.sin6_family = AF_INET6;
2071	sa6_dst.sin6_len = sizeof(sa6_dst);
2072	sa6_dst.sin6_addr = t;
2073	in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
2074	in6_embedscope(&t, &sa6_dst, NULL, NULL);
2075
2076#ifdef COMPAT_RFC1885
2077	/*
2078	 * xxx guess MTU
2079	 * RFC 1885 requires that echo reply should be truncated if it
2080	 * does not fit in with (return) path MTU, but the description was
2081	 * removed in the new spec.
2082	 */
2083	if (icmp6_reflect_rt.ro_rt == 0 ||
2084	    ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
2085		if (icmp6_reflect_rt.ro_rt) {
2086			RTFREE(icmp6_reflect_rt.ro_rt);
2087			icmp6_reflect_rt.ro_rt = 0;
2088		}
2089		bzero(sin6, sizeof(*sin6));
2090		sin6->sin6_family = PF_INET6;
2091		sin6->sin6_len = sizeof(struct sockaddr_in6);
2092		sin6->sin6_addr = ip6->ip6_dst;
2093
2094		rtalloc_ign((struct route *)&icmp6_reflect_rt.ro_rt,
2095			    RTF_PRCLONING);
2096	}
2097
2098	if (icmp6_reflect_rt.ro_rt == 0)
2099		goto bad;
2100
2101	if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
2102	    && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
2103		mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
2104
2105	if (mtu < m->m_pkthdr.len) {
2106		plen -= (m->m_pkthdr.len - mtu);
2107		m_adj(m, mtu - m->m_pkthdr.len);
2108	}
2109#endif
2110	/*
2111	 * If the incoming packet was addressed directly to us (i.e. unicast),
2112	 * use dst as the src for the reply.
2113	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2114	 * (for example) when we encounter an error while forwarding procedure
2115	 * destined to a duplicated address of ours.
2116	 */
2117	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2118		if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2119		    (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2120			src = &t;
2121			break;
2122		}
2123	if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2124		/*
2125		 * This is the case if the dst is our link-local address
2126		 * and the sender is also ourselves.
2127		 */
2128		src = &t;
2129	}
2130
2131	if (src == 0) {
2132		int e;
2133		struct route_in6 ro;
2134
2135		/*
2136		 * This case matches to multicasts, our anycast, or unicasts
2137		 * that we do not own.  Select a source address based on the
2138		 * source address of the erroneous packet.
2139		 */
2140		bzero(&ro, sizeof(ro));
2141		src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e);
2142		if (ro.ro_rt)
2143			RTFREE(ro.ro_rt); /* XXX: we could use this */
2144		if (src == NULL) {
2145			nd6log((LOG_DEBUG,
2146			    "icmp6_reflect: source can't be determined: "
2147			    "dst=%s, error=%d\n",
2148			    ip6_sprintf(&sa6_src.sin6_addr), e));
2149			goto bad;
2150		}
2151	}
2152
2153	ip6->ip6_src = *src;
2154
2155	ip6->ip6_flow = 0;
2156	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2157	ip6->ip6_vfc |= IPV6_VERSION;
2158	ip6->ip6_nxt = IPPROTO_ICMPV6;
2159	if (m->m_pkthdr.rcvif) {
2160		/* XXX: This may not be the outgoing interface */
2161		ip6->ip6_hlim = nd_ifinfo[m->m_pkthdr.rcvif->if_index].chlim;
2162	} else
2163		ip6->ip6_hlim = ip6_defhlim;
2164
2165	icmp6->icmp6_cksum = 0;
2166	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2167	    sizeof(struct ip6_hdr), plen);
2168
2169	/*
2170	 * XXX option handling
2171	 */
2172
2173	m->m_flags &= ~(M_BCAST|M_MCAST);
2174
2175#ifdef COMPAT_RFC1885
2176	ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif, NULL);
2177#else
2178	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2179#endif
2180	if (outif)
2181		icmp6_ifoutstat_inc(outif, type, code);
2182
2183	return;
2184
2185 bad:
2186	m_freem(m);
2187	return;
2188}
2189
2190void
2191icmp6_fasttimo()
2192{
2193
2194	mld6_fasttimeo();
2195}
2196
2197static const char *
2198icmp6_redirect_diag(src6, dst6, tgt6)
2199	struct in6_addr *src6;
2200	struct in6_addr *dst6;
2201	struct in6_addr *tgt6;
2202{
2203	static char buf[1024];
2204	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2205	    ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2206	return buf;
2207}
2208
2209void
2210icmp6_redirect_input(m, off)
2211	struct mbuf *m;
2212	int off;
2213{
2214	struct ifnet *ifp = m->m_pkthdr.rcvif;
2215	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2216	struct nd_redirect *nd_rd;
2217	int icmp6len = ntohs(ip6->ip6_plen);
2218	char *lladdr = NULL;
2219	int lladdrlen = 0;
2220	u_char *redirhdr = NULL;
2221	int redirhdrlen = 0;
2222	struct rtentry *rt = NULL;
2223	int is_router;
2224	int is_onlink;
2225	struct in6_addr src6 = ip6->ip6_src;
2226	struct in6_addr redtgt6;
2227	struct in6_addr reddst6;
2228	union nd_opts ndopts;
2229
2230	if (!m || !ifp)
2231		return;
2232
2233	/* XXX if we are router, we don't update route by icmp6 redirect */
2234	if (ip6_forwarding)
2235		goto freeit;
2236	if (!icmp6_rediraccept)
2237		goto freeit;
2238
2239#ifndef PULLDOWN_TEST
2240	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2241	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2242#else
2243	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2244	if (nd_rd == NULL) {
2245		icmp6stat.icp6s_tooshort++;
2246		return;
2247	}
2248#endif
2249	redtgt6 = nd_rd->nd_rd_target;
2250	reddst6 = nd_rd->nd_rd_dst;
2251
2252	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2253		redtgt6.s6_addr16[1] = htons(ifp->if_index);
2254	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2255		reddst6.s6_addr16[1] = htons(ifp->if_index);
2256
2257	/* validation */
2258	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2259		nd6log((LOG_ERR,
2260		    "ICMP6 redirect sent from %s rejected; "
2261		    "must be from linklocal\n",
2262		    ip6_sprintf(&src6)));
2263		goto bad;
2264	}
2265	if (ip6->ip6_hlim != 255) {
2266		nd6log((LOG_ERR,
2267		    "ICMP6 redirect sent from %s rejected; "
2268		    "hlim=%d (must be 255)\n",
2269		    ip6_sprintf(&src6), ip6->ip6_hlim));
2270		goto bad;
2271	}
2272    {
2273	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2274	struct sockaddr_in6 sin6;
2275	struct in6_addr *gw6;
2276
2277	bzero(&sin6, sizeof(sin6));
2278	sin6.sin6_family = AF_INET6;
2279	sin6.sin6_len = sizeof(struct sockaddr_in6);
2280	bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2281	rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL);
2282	if (rt) {
2283		if (rt->rt_gateway == NULL ||
2284		    rt->rt_gateway->sa_family != AF_INET6) {
2285			nd6log((LOG_ERR,
2286			    "ICMP6 redirect rejected; no route "
2287			    "with inet6 gateway found for redirect dst: %s\n",
2288			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2289			RTFREE_LOCKED(rt);
2290			goto bad;
2291		}
2292
2293		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2294		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2295			nd6log((LOG_ERR,
2296			    "ICMP6 redirect rejected; "
2297			    "not equal to gw-for-src=%s (must be same): "
2298			    "%s\n",
2299			    ip6_sprintf(gw6),
2300			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2301			RTFREE_LOCKED(rt);
2302			goto bad;
2303		}
2304	} else {
2305		nd6log((LOG_ERR,
2306		    "ICMP6 redirect rejected; "
2307		    "no route found for redirect dst: %s\n",
2308		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2309		goto bad;
2310	}
2311	RTFREE_LOCKED(rt);
2312	rt = NULL;
2313    }
2314	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2315		nd6log((LOG_ERR,
2316		    "ICMP6 redirect rejected; "
2317		    "redirect dst must be unicast: %s\n",
2318		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2319		goto bad;
2320	}
2321
2322	is_router = is_onlink = 0;
2323	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2324		is_router = 1;	/* router case */
2325	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2326		is_onlink = 1;	/* on-link destination case */
2327	if (!is_router && !is_onlink) {
2328		nd6log((LOG_ERR,
2329		    "ICMP6 redirect rejected; "
2330		    "neither router case nor onlink case: %s\n",
2331		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2332		goto bad;
2333	}
2334	/* validation passed */
2335
2336	icmp6len -= sizeof(*nd_rd);
2337	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2338	if (nd6_options(&ndopts) < 0) {
2339		nd6log((LOG_INFO, "icmp6_redirect_input: "
2340		    "invalid ND option, rejected: %s\n",
2341		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2342		/* nd6_options have incremented stats */
2343		goto freeit;
2344	}
2345
2346	if (ndopts.nd_opts_tgt_lladdr) {
2347		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2348		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2349	}
2350
2351	if (ndopts.nd_opts_rh) {
2352		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2353		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2354	}
2355
2356	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2357		nd6log((LOG_INFO,
2358		    "icmp6_redirect_input: lladdrlen mismatch for %s "
2359		    "(if %d, icmp6 packet %d): %s\n",
2360		    ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2361		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2362		goto bad;
2363	}
2364
2365	/* RFC 2461 8.3 */
2366	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2367	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2368
2369	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2370		/* perform rtredirect */
2371		struct sockaddr_in6 sdst;
2372		struct sockaddr_in6 sgw;
2373		struct sockaddr_in6 ssrc;
2374
2375		bzero(&sdst, sizeof(sdst));
2376		bzero(&sgw, sizeof(sgw));
2377		bzero(&ssrc, sizeof(ssrc));
2378		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2379		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2380			sizeof(struct sockaddr_in6);
2381		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2382		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2383		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2384		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2385		    (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2386		    (struct sockaddr *)&ssrc);
2387	}
2388	/* finally update cached route in each socket via pfctlinput */
2389    {
2390	struct sockaddr_in6 sdst;
2391
2392	bzero(&sdst, sizeof(sdst));
2393	sdst.sin6_family = AF_INET6;
2394	sdst.sin6_len = sizeof(struct sockaddr_in6);
2395	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2396	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2397#ifdef IPSEC
2398	key_sa_routechange((struct sockaddr *)&sdst);
2399#endif
2400    }
2401
2402 freeit:
2403	m_freem(m);
2404	return;
2405
2406 bad:
2407	icmp6stat.icp6s_badredirect++;
2408	m_freem(m);
2409}
2410
2411void
2412icmp6_redirect_output(m0, rt)
2413	struct mbuf *m0;
2414	struct rtentry *rt;
2415{
2416	struct ifnet *ifp;	/* my outgoing interface */
2417	struct in6_addr *ifp_ll6;
2418	struct in6_addr *router_ll6;
2419	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2420	struct mbuf *m = NULL;	/* newly allocated one */
2421	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2422	struct nd_redirect *nd_rd;
2423	size_t maxlen;
2424	u_char *p;
2425	struct ifnet *outif = NULL;
2426	struct sockaddr_in6 src_sa;
2427
2428	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2429
2430	/* if we are not router, we don't send icmp6 redirect */
2431	if (!ip6_forwarding || ip6_accept_rtadv)
2432		goto fail;
2433
2434	/* sanity check */
2435	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2436		goto fail;
2437
2438	/*
2439	 * Address check:
2440	 *  the source address must identify a neighbor, and
2441	 *  the destination address must not be a multicast address
2442	 *  [RFC 2461, sec 8.2]
2443	 */
2444	sip6 = mtod(m0, struct ip6_hdr *);
2445	bzero(&src_sa, sizeof(src_sa));
2446	src_sa.sin6_family = AF_INET6;
2447	src_sa.sin6_len = sizeof(src_sa);
2448	src_sa.sin6_addr = sip6->ip6_src;
2449	/* we don't currently use sin6_scope_id, but eventually use it */
2450	src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src);
2451	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2452		goto fail;
2453	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2454		goto fail;	/* what should we do here? */
2455
2456	/* rate limit */
2457	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2458		goto fail;
2459
2460	/*
2461	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2462	 * we almost always ask for an mbuf cluster for simplicity.
2463	 * (MHLEN < IPV6_MMTU is almost always true)
2464	 */
2465#if IPV6_MMTU >= MCLBYTES
2466# error assumption failed about IPV6_MMTU and MCLBYTES
2467#endif
2468	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2469	if (m && IPV6_MMTU >= MHLEN)
2470		MCLGET(m, M_DONTWAIT);
2471	if (!m)
2472		goto fail;
2473	m->m_pkthdr.rcvif = NULL;
2474	m->m_len = 0;
2475	maxlen = M_TRAILINGSPACE(m);
2476	maxlen = min(IPV6_MMTU, maxlen);
2477	/* just for safety */
2478	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2479	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2480		goto fail;
2481	}
2482
2483	{
2484		/* get ip6 linklocal address for ifp(my outgoing interface). */
2485		struct in6_ifaddr *ia;
2486		if ((ia = in6ifa_ifpforlinklocal(ifp,
2487						 IN6_IFF_NOTREADY|
2488						 IN6_IFF_ANYCAST)) == NULL)
2489			goto fail;
2490		ifp_ll6 = &ia->ia_addr.sin6_addr;
2491	}
2492
2493	/* get ip6 linklocal address for the router. */
2494	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2495		struct sockaddr_in6 *sin6;
2496		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2497		router_ll6 = &sin6->sin6_addr;
2498		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2499			router_ll6 = (struct in6_addr *)NULL;
2500	} else
2501		router_ll6 = (struct in6_addr *)NULL;
2502
2503	/* ip6 */
2504	ip6 = mtod(m, struct ip6_hdr *);
2505	ip6->ip6_flow = 0;
2506	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2507	ip6->ip6_vfc |= IPV6_VERSION;
2508	/* ip6->ip6_plen will be set later */
2509	ip6->ip6_nxt = IPPROTO_ICMPV6;
2510	ip6->ip6_hlim = 255;
2511	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2512	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2513	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2514
2515	/* ND Redirect */
2516	nd_rd = (struct nd_redirect *)(ip6 + 1);
2517	nd_rd->nd_rd_type = ND_REDIRECT;
2518	nd_rd->nd_rd_code = 0;
2519	nd_rd->nd_rd_reserved = 0;
2520	if (rt->rt_flags & RTF_GATEWAY) {
2521		/*
2522		 * nd_rd->nd_rd_target must be a link-local address in
2523		 * better router cases.
2524		 */
2525		if (!router_ll6)
2526			goto fail;
2527		bcopy(router_ll6, &nd_rd->nd_rd_target,
2528		    sizeof(nd_rd->nd_rd_target));
2529		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2530		    sizeof(nd_rd->nd_rd_dst));
2531	} else {
2532		/* make sure redtgt == reddst */
2533		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2534		    sizeof(nd_rd->nd_rd_target));
2535		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2536		    sizeof(nd_rd->nd_rd_dst));
2537	}
2538
2539	p = (u_char *)(nd_rd + 1);
2540
2541	if (!router_ll6)
2542		goto nolladdropt;
2543
2544	{
2545		/* target lladdr option */
2546		struct rtentry *rt_router = NULL;
2547		int len;
2548		struct sockaddr_dl *sdl;
2549		struct nd_opt_hdr *nd_opt;
2550		char *lladdr;
2551
2552		rt_router = nd6_lookup(router_ll6, 0, ifp);
2553		if (!rt_router)
2554			goto nolladdropt;
2555		len = sizeof(*nd_opt) + ifp->if_addrlen;
2556		len = (len + 7) & ~7;	/* round by 8 */
2557		/* safety check */
2558		if (len + (p - (u_char *)ip6) > maxlen)
2559			goto nolladdropt;
2560		if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2561		    (rt_router->rt_flags & RTF_LLINFO) &&
2562		    (rt_router->rt_gateway->sa_family == AF_LINK) &&
2563		    (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2564		    sdl->sdl_alen) {
2565			nd_opt = (struct nd_opt_hdr *)p;
2566			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2567			nd_opt->nd_opt_len = len >> 3;
2568			lladdr = (char *)(nd_opt + 1);
2569			bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2570			p += len;
2571		}
2572	}
2573nolladdropt:;
2574
2575	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2576
2577	/* just to be safe */
2578#ifdef M_DECRYPTED	/*not openbsd*/
2579	if (m0->m_flags & M_DECRYPTED)
2580		goto noredhdropt;
2581#endif
2582	if (p - (u_char *)ip6 > maxlen)
2583		goto noredhdropt;
2584
2585	{
2586		/* redirected header option */
2587		int len;
2588		struct nd_opt_rd_hdr *nd_opt_rh;
2589
2590		/*
2591		 * compute the maximum size for icmp6 redirect header option.
2592		 * XXX room for auth header?
2593		 */
2594		len = maxlen - (p - (u_char *)ip6);
2595		len &= ~7;
2596
2597		/* This is just for simplicity. */
2598		if (m0->m_pkthdr.len != m0->m_len) {
2599			if (m0->m_next) {
2600				m_freem(m0->m_next);
2601				m0->m_next = NULL;
2602			}
2603			m0->m_pkthdr.len = m0->m_len;
2604		}
2605
2606		/*
2607		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2608		 * about padding/truncate rule for the original IP packet.
2609		 * From the discussion on IPv6imp in Feb 1999,
2610		 * the consensus was:
2611		 * - "attach as much as possible" is the goal
2612		 * - pad if not aligned (original size can be guessed by
2613		 *   original ip6 header)
2614		 * Following code adds the padding if it is simple enough,
2615		 * and truncates if not.
2616		 */
2617		if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2618			panic("assumption failed in %s:%d", __FILE__,
2619			    __LINE__);
2620
2621		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2622			/* not enough room, truncate */
2623			m0->m_pkthdr.len = m0->m_len = len -
2624			    sizeof(*nd_opt_rh);
2625		} else {
2626			/* enough room, pad or truncate */
2627			size_t extra;
2628
2629			extra = m0->m_pkthdr.len % 8;
2630			if (extra) {
2631				/* pad if easy enough, truncate if not */
2632				if (8 - extra <= M_TRAILINGSPACE(m0)) {
2633					/* pad */
2634					m0->m_len += (8 - extra);
2635					m0->m_pkthdr.len += (8 - extra);
2636				} else {
2637					/* truncate */
2638					m0->m_pkthdr.len -= extra;
2639					m0->m_len -= extra;
2640				}
2641			}
2642			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2643			m0->m_pkthdr.len = m0->m_len = len -
2644			    sizeof(*nd_opt_rh);
2645		}
2646
2647		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2648		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2649		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2650		nd_opt_rh->nd_opt_rh_len = len >> 3;
2651		p += sizeof(*nd_opt_rh);
2652		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2653
2654		/* connect m0 to m */
2655		m_tag_delete_chain(m0, NULL);
2656		m0->m_flags &= ~M_PKTHDR;
2657		m->m_next = m0;
2658		m->m_pkthdr.len = m->m_len + m0->m_len;
2659		m0 = NULL;
2660	}
2661noredhdropt:;
2662	if (m0) {
2663		m_freem(m0);
2664		m0 = NULL;
2665	}
2666
2667	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
2668		sip6->ip6_src.s6_addr16[1] = 0;
2669	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
2670		sip6->ip6_dst.s6_addr16[1] = 0;
2671	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
2672		nd_rd->nd_rd_target.s6_addr16[1] = 0;
2673	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
2674		nd_rd->nd_rd_dst.s6_addr16[1] = 0;
2675
2676	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2677
2678	nd_rd->nd_rd_cksum = 0;
2679	nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2680	    sizeof(*ip6), ntohs(ip6->ip6_plen));
2681
2682	/* send the packet to outside... */
2683	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2684	if (outif) {
2685		icmp6_ifstat_inc(outif, ifs6_out_msg);
2686		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2687	}
2688	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2689
2690	return;
2691
2692fail:
2693	if (m)
2694		m_freem(m);
2695	if (m0)
2696		m_freem(m0);
2697}
2698
2699#ifdef HAVE_NRL_INPCB
2700#define sotoin6pcb	sotoinpcb
2701#define in6pcb		inpcb
2702#define in6p_icmp6filt	inp_icmp6filt
2703#endif
2704/*
2705 * ICMPv6 socket option processing.
2706 */
2707int
2708icmp6_ctloutput(so, sopt)
2709	struct socket *so;
2710	struct sockopt *sopt;
2711{
2712	int error = 0;
2713	int optlen;
2714	struct inpcb *inp = sotoinpcb(so);
2715	int level, op, optname;
2716
2717	if (sopt) {
2718		level = sopt->sopt_level;
2719		op = sopt->sopt_dir;
2720		optname = sopt->sopt_name;
2721		optlen = sopt->sopt_valsize;
2722	} else
2723		level = op = optname = optlen = 0;
2724
2725	if (level != IPPROTO_ICMPV6) {
2726		return EINVAL;
2727	}
2728
2729	switch (op) {
2730	case PRCO_SETOPT:
2731		switch (optname) {
2732		case ICMP6_FILTER:
2733		    {
2734			struct icmp6_filter *p;
2735
2736			if (optlen != sizeof(*p)) {
2737				error = EMSGSIZE;
2738				break;
2739			}
2740			if (inp->in6p_icmp6filt == NULL) {
2741				error = EINVAL;
2742				break;
2743			}
2744			error = sooptcopyin(sopt, inp->in6p_icmp6filt, optlen,
2745				optlen);
2746			break;
2747		    }
2748
2749		default:
2750			error = ENOPROTOOPT;
2751			break;
2752		}
2753		break;
2754
2755	case PRCO_GETOPT:
2756		switch (optname) {
2757		case ICMP6_FILTER:
2758		    {
2759			if (inp->in6p_icmp6filt == NULL) {
2760				error = EINVAL;
2761				break;
2762			}
2763			error = sooptcopyout(sopt, inp->in6p_icmp6filt,
2764				sizeof(struct icmp6_filter));
2765			break;
2766		    }
2767
2768		default:
2769			error = ENOPROTOOPT;
2770			break;
2771		}
2772		break;
2773	}
2774
2775	return (error);
2776}
2777#ifdef HAVE_NRL_INPCB
2778#undef sotoin6pcb
2779#undef in6pcb
2780#undef in6p_icmp6filt
2781#endif
2782
2783#ifndef HAVE_PPSRATECHECK
2784#ifndef timersub
2785#define	timersub(tvp, uvp, vvp)						\
2786	do {								\
2787		(vvp)->tv_sec = (tvp)->tv_sec - (uvp)->tv_sec;		\
2788		(vvp)->tv_usec = (tvp)->tv_usec - (uvp)->tv_usec;	\
2789		if ((vvp)->tv_usec < 0) {				\
2790			(vvp)->tv_sec--;				\
2791			(vvp)->tv_usec += 1000000;			\
2792		}							\
2793	} while (0)
2794#endif
2795
2796/*
2797 * ppsratecheck(): packets (or events) per second limitation.
2798 */
2799static int
2800ppsratecheck(lasttime, curpps, maxpps)
2801	struct timeval *lasttime;
2802	int *curpps;
2803	int maxpps;	/* maximum pps allowed */
2804{
2805	struct timeval tv, delta;
2806	int s, rv;
2807
2808	s = splclock();
2809	microtime(&tv);
2810	splx(s);
2811
2812	timersub(&tv, lasttime, &delta);
2813
2814	/*
2815	 * Check for 0,0 so that the message will be seen at least once.
2816	 * If more than one second has passed since the last update of
2817	 * lasttime, reset the counter.
2818	 *
2819	 * We do increment *curpps even in *curpps < maxpps case, as some may
2820	 * try to use *curpps for stat purposes as well.
2821	 */
2822	if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) ||
2823	    delta.tv_sec >= 1) {
2824		*lasttime = tv;
2825		*curpps = 0;
2826		rv = 1;
2827	} else if (maxpps < 0)
2828		rv = 1;
2829	else if (*curpps < maxpps)
2830		rv = 1;
2831	else
2832		rv = 0;
2833
2834#if 1 /* DIAGNOSTIC? */
2835	/* be careful about wrap-around */
2836	if (*curpps + 1 > *curpps)
2837		*curpps = *curpps + 1;
2838#else
2839	/*
2840	 * assume that there's not too many calls to this function.
2841	 * not sure if the assumption holds, as it depends on *caller's*
2842	 * behavior, not the behavior of this function.
2843	 * IMHO it is wrong to make assumption on the caller's behavior,
2844	 * so the above #if is #if 1, not #ifdef DIAGNOSTIC.
2845	 */
2846	*curpps = *curpps + 1;
2847#endif
2848
2849	return (rv);
2850}
2851#endif
2852
2853/*
2854 * Perform rate limit check.
2855 * Returns 0 if it is okay to send the icmp6 packet.
2856 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2857 * limitation.
2858 *
2859 * XXX per-destination/type check necessary?
2860 */
2861static int
2862icmp6_ratelimit(dst, type, code)
2863	const struct in6_addr *dst;	/* not used at this moment */
2864	const int type;			/* not used at this moment */
2865	const int code;			/* not used at this moment */
2866{
2867	int ret;
2868
2869	ret = 0;	/* okay to send */
2870
2871	/* PPS limit */
2872	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2873	    icmp6errppslim)) {
2874		/* The packet is subject to rate limit */
2875		ret++;
2876	}
2877
2878	return ret;
2879}
2880