icmp6.c revision 142987
1/*	$FreeBSD: head/sys/netinet6/icmp6.c 142987 2005-03-02 05:14:15Z suz $	*/
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 * 4. Neither the name of the University nor the names of its contributors
46 *    may be used to endorse or promote products derived from this software
47 *    without specific prior written permission.
48 *
49 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59 * SUCH DAMAGE.
60 *
61 *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
62 */
63
64#include "opt_inet.h"
65#include "opt_inet6.h"
66#include "opt_ipsec.h"
67
68#include <sys/param.h>
69#include <sys/domain.h>
70#include <sys/kernel.h>
71#include <sys/lock.h>
72#include <sys/malloc.h>
73#include <sys/mbuf.h>
74#include <sys/protosw.h>
75#include <sys/signalvar.h>
76#include <sys/socket.h>
77#include <sys/socketvar.h>
78#include <sys/sx.h>
79#include <sys/syslog.h>
80#include <sys/systm.h>
81#include <sys/time.h>
82
83#include <net/if.h>
84#include <net/if_dl.h>
85#include <net/if_types.h>
86#include <net/route.h>
87
88#include <netinet/in.h>
89#include <netinet/in_pcb.h>
90#include <netinet/in_var.h>
91#include <netinet/ip6.h>
92#include <netinet/icmp6.h>
93#include <netinet/tcp_var.h>
94#include <netinet6/in6_ifattach.h>
95#include <netinet6/in6_pcb.h>
96#include <netinet6/ip6protosw.h>
97#include <netinet6/ip6_var.h>
98#include <netinet6/mld6_var.h>
99#include <netinet6/nd6.h>
100
101#ifdef IPSEC
102#include <netinet6/ipsec.h>
103#include <netkey/key.h>
104#endif
105
106#ifdef FAST_IPSEC
107#include <netipsec/ipsec.h>
108#include <netipsec/key.h>
109#endif
110
111#include <net/net_osdep.h>
112
113extern struct domain inet6domain;
114
115struct icmp6stat icmp6stat;
116
117extern struct inpcbinfo ripcbinfo;
118extern struct inpcbhead ripcb;
119extern int icmp6errppslim;
120static int icmp6errpps_count = 0;
121static struct timeval icmp6errppslim_last;
122extern int icmp6_nodeinfo;
123
124static void icmp6_errcount __P((struct icmp6errstat *, int, int));
125static int icmp6_rip6_input __P((struct mbuf **, int));
126static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int));
127static const char *icmp6_redirect_diag __P((struct in6_addr *,
128	struct in6_addr *, struct in6_addr *));
129static struct mbuf *ni6_input __P((struct mbuf *, int));
130static struct mbuf *ni6_nametodns __P((const char *, int, int));
131static int ni6_dnsmatch __P((const char *, int, const char *, int));
132static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
133			  struct ifnet **, char *));
134static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
135				struct ifnet *, int));
136static int icmp6_notify_error __P((struct mbuf **, int, int, int));
137
138#ifdef COMPAT_RFC1885
139static struct route_in6 icmp6_reflect_rt;
140#endif
141
142
143void
144icmp6_init()
145{
146	mld6_init();
147}
148
149static void
150icmp6_errcount(stat, type, code)
151	struct icmp6errstat *stat;
152	int type, code;
153{
154	switch (type) {
155	case ICMP6_DST_UNREACH:
156		switch (code) {
157		case ICMP6_DST_UNREACH_NOROUTE:
158			stat->icp6errs_dst_unreach_noroute++;
159			return;
160		case ICMP6_DST_UNREACH_ADMIN:
161			stat->icp6errs_dst_unreach_admin++;
162			return;
163		case ICMP6_DST_UNREACH_BEYONDSCOPE:
164			stat->icp6errs_dst_unreach_beyondscope++;
165			return;
166		case ICMP6_DST_UNREACH_ADDR:
167			stat->icp6errs_dst_unreach_addr++;
168			return;
169		case ICMP6_DST_UNREACH_NOPORT:
170			stat->icp6errs_dst_unreach_noport++;
171			return;
172		}
173		break;
174	case ICMP6_PACKET_TOO_BIG:
175		stat->icp6errs_packet_too_big++;
176		return;
177	case ICMP6_TIME_EXCEEDED:
178		switch (code) {
179		case ICMP6_TIME_EXCEED_TRANSIT:
180			stat->icp6errs_time_exceed_transit++;
181			return;
182		case ICMP6_TIME_EXCEED_REASSEMBLY:
183			stat->icp6errs_time_exceed_reassembly++;
184			return;
185		}
186		break;
187	case ICMP6_PARAM_PROB:
188		switch (code) {
189		case ICMP6_PARAMPROB_HEADER:
190			stat->icp6errs_paramprob_header++;
191			return;
192		case ICMP6_PARAMPROB_NEXTHEADER:
193			stat->icp6errs_paramprob_nextheader++;
194			return;
195		case ICMP6_PARAMPROB_OPTION:
196			stat->icp6errs_paramprob_option++;
197			return;
198		}
199		break;
200	case ND_REDIRECT:
201		stat->icp6errs_redirect++;
202		return;
203	}
204	stat->icp6errs_unknown++;
205}
206
207/*
208 * Generate an error packet of type error in response to bad IP6 packet.
209 */
210void
211icmp6_error(m, type, code, param)
212	struct mbuf *m;
213	int type, code, param;
214{
215	struct ip6_hdr *oip6, *nip6;
216	struct icmp6_hdr *icmp6;
217	u_int preplen;
218	int off;
219	int nxt;
220
221	icmp6stat.icp6s_error++;
222
223	/* count per-type-code statistics */
224	icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
225
226#ifdef M_DECRYPTED	/*not openbsd*/
227	if (m->m_flags & M_DECRYPTED) {
228		icmp6stat.icp6s_canterror++;
229		goto freeit;
230	}
231#endif
232
233#ifndef PULLDOWN_TEST
234	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
235#else
236	if (m->m_len < sizeof(struct ip6_hdr)) {
237		m = m_pullup(m, sizeof(struct ip6_hdr));
238		if (m == NULL)
239			return;
240	}
241#endif
242	oip6 = mtod(m, struct ip6_hdr *);
243
244	/*
245	 * If the destination address of the erroneous packet is a multicast
246	 * address, or the packet was sent using link-layer multicast,
247	 * we should basically suppress sending an error (RFC 2463, Section
248	 * 2.4).
249	 * We have two exceptions (the item e.2 in that section):
250	 * - the Pakcet Too Big message can be sent for path MTU discovery.
251	 * - the Parameter Problem Message that can be allowed an icmp6 error
252	 *   in the option type field.  This check has been done in
253	 *   ip6_unknown_opt(), so we can just check the type and code.
254	 */
255	if ((m->m_flags & (M_BCAST|M_MCAST) ||
256	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
257	    (type != ICMP6_PACKET_TOO_BIG &&
258	     (type != ICMP6_PARAM_PROB ||
259	      code != ICMP6_PARAMPROB_OPTION)))
260		goto freeit;
261
262	/*
263	 * RFC 2463, 2.4 (e.5): source address check.
264	 * XXX: the case of anycast source?
265	 */
266	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
267	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
268		goto freeit;
269
270	/*
271	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
272	 * don't do it.
273	 */
274	nxt = -1;
275	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
276	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
277		struct icmp6_hdr *icp;
278
279#ifndef PULLDOWN_TEST
280		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
281		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
282#else
283		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
284			sizeof(*icp));
285		if (icp == NULL) {
286			icmp6stat.icp6s_tooshort++;
287			return;
288		}
289#endif
290		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
291		    icp->icmp6_type == ND_REDIRECT) {
292			/*
293			 * ICMPv6 error
294			 * Special case: for redirect (which is
295			 * informational) we must not send icmp6 error.
296			 */
297			icmp6stat.icp6s_canterror++;
298			goto freeit;
299		} else {
300			/* ICMPv6 informational - send the error */
301		}
302	} else {
303		/* non-ICMPv6 - send the error */
304	}
305
306	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
307
308	/* Finally, do rate limitation check. */
309	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
310		icmp6stat.icp6s_toofreq++;
311		goto freeit;
312	}
313
314	/*
315	 * OK, ICMP6 can be generated.
316	 */
317
318	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
319		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
320
321	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
322	M_PREPEND(m, preplen, M_DONTWAIT);
323	if (m && m->m_len < preplen)
324		m = m_pullup(m, preplen);
325	if (m == NULL) {
326		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
327		return;
328	}
329
330	nip6 = mtod(m, struct ip6_hdr *);
331	nip6->ip6_src  = oip6->ip6_src;
332	nip6->ip6_dst  = oip6->ip6_dst;
333
334	in6_clearscope(&oip6->ip6_src);
335	in6_clearscope(&oip6->ip6_dst);
336
337	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
338	icmp6->icmp6_type = type;
339	icmp6->icmp6_code = code;
340	icmp6->icmp6_pptr = htonl((u_int32_t)param);
341
342	/*
343	 * icmp6_reflect() is designed to be in the input path.
344	 * icmp6_error() can be called from both input and outut path,
345	 * and if we are in output path rcvif could contain bogus value.
346	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
347	 * information in ip header (nip6).
348	 */
349	m->m_pkthdr.rcvif = NULL;
350
351	icmp6stat.icp6s_outhist[type]++;
352	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
353
354	return;
355
356  freeit:
357	/*
358	 * If we can't tell wheter or not we can generate ICMP6, free it.
359	 */
360	m_freem(m);
361}
362
363/*
364 * Process a received ICMP6 message.
365 */
366int
367icmp6_input(mp, offp, proto)
368	struct mbuf **mp;
369	int *offp, proto;
370{
371	struct mbuf *m = *mp, *n;
372	struct ip6_hdr *ip6, *nip6;
373	struct icmp6_hdr *icmp6, *nicmp6;
374	int off = *offp;
375	int icmp6len = m->m_pkthdr.len - *offp;
376	int code, sum, noff;
377
378#ifndef PULLDOWN_TEST
379	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
380	/* m might change if M_LOOP.  So, call mtod after this */
381#endif
382
383	/*
384	 * Locate icmp6 structure in mbuf, and check
385	 * that not corrupted and of at least minimum length
386	 */
387
388	ip6 = mtod(m, struct ip6_hdr *);
389	if (icmp6len < sizeof(struct icmp6_hdr)) {
390		icmp6stat.icp6s_tooshort++;
391		goto freeit;
392	}
393
394	/*
395	 * calculate the checksum
396	 */
397#ifndef PULLDOWN_TEST
398	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
399#else
400	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
401	if (icmp6 == NULL) {
402		icmp6stat.icp6s_tooshort++;
403		return IPPROTO_DONE;
404	}
405#endif
406	code = icmp6->icmp6_code;
407
408	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
409		nd6log((LOG_ERR,
410		    "ICMP6 checksum error(%d|%x) %s\n",
411		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
412		icmp6stat.icp6s_checksum++;
413		goto freeit;
414	}
415
416	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
417		/*
418		 * Deliver very specific ICMP6 type only.
419		 * This is important to deilver TOOBIG.  Otherwise PMTUD
420		 * will not work.
421		 */
422		switch (icmp6->icmp6_type) {
423		case ICMP6_DST_UNREACH:
424		case ICMP6_PACKET_TOO_BIG:
425		case ICMP6_TIME_EXCEEDED:
426			break;
427		default:
428			goto freeit;
429		}
430	}
431
432	icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
433	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
434	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
435		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
436
437	switch (icmp6->icmp6_type) {
438	case ICMP6_DST_UNREACH:
439		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
440		switch (code) {
441		case ICMP6_DST_UNREACH_NOROUTE:
442			code = PRC_UNREACH_NET;
443			break;
444		case ICMP6_DST_UNREACH_ADMIN:
445			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
446			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
447			break;
448		case ICMP6_DST_UNREACH_ADDR:
449			code = PRC_HOSTDEAD;
450			break;
451#ifdef COMPAT_RFC1885
452		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
453			code = PRC_UNREACH_SRCFAIL;
454			break;
455#else
456		case ICMP6_DST_UNREACH_BEYONDSCOPE:
457			/* I mean "source address was incorrect." */
458			code = PRC_PARAMPROB;
459			break;
460#endif
461		case ICMP6_DST_UNREACH_NOPORT:
462			code = PRC_UNREACH_PORT;
463			break;
464		default:
465			goto badcode;
466		}
467		goto deliver;
468		break;
469
470	case ICMP6_PACKET_TOO_BIG:
471		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
472
473		/* validation is made in icmp6_mtudisc_update */
474
475		code = PRC_MSGSIZE;
476
477		/*
478		 * Updating the path MTU will be done after examining
479		 * intermediate extension headers.
480		 */
481		goto deliver;
482		break;
483
484	case ICMP6_TIME_EXCEEDED:
485		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
486		switch (code) {
487		case ICMP6_TIME_EXCEED_TRANSIT:
488		case ICMP6_TIME_EXCEED_REASSEMBLY:
489			code += PRC_TIMXCEED_INTRANS;
490			break;
491		default:
492			goto badcode;
493		}
494		goto deliver;
495		break;
496
497	case ICMP6_PARAM_PROB:
498		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
499		switch (code) {
500		case ICMP6_PARAMPROB_NEXTHEADER:
501			code = PRC_UNREACH_PROTOCOL;
502			break;
503		case ICMP6_PARAMPROB_HEADER:
504		case ICMP6_PARAMPROB_OPTION:
505			code = PRC_PARAMPROB;
506			break;
507		default:
508			goto badcode;
509		}
510		goto deliver;
511		break;
512
513	case ICMP6_ECHO_REQUEST:
514		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
515		if (code != 0)
516			goto badcode;
517		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
518			/* Give up remote */
519			break;
520		}
521		if ((n->m_flags & M_EXT) != 0
522		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
523			struct mbuf *n0 = n;
524			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
525			int n0len;
526
527			/*
528			 * Prepare an internal mbuf.  m_pullup() doesn't
529			 * always copy the length we specified.
530			 */
531			if (maxlen >= MCLBYTES) {
532				/* Give up remote */
533				m_freem(n0);
534				break;
535			}
536			MGETHDR(n, M_DONTWAIT, n0->m_type);
537			n0len = n0->m_pkthdr.len;	/* save for use below */
538			if (n)
539				M_MOVE_PKTHDR(n, n0);
540			if (n && maxlen >= MHLEN) {
541				MCLGET(n, M_DONTWAIT);
542				if ((n->m_flags & M_EXT) == 0) {
543					m_free(n);
544					n = NULL;
545				}
546			}
547			if (n == NULL) {
548				/* Give up remote */
549				m_freem(n0);
550				break;
551			}
552			/*
553			 * Copy IPv6 and ICMPv6 only.
554			 */
555			nip6 = mtod(n, struct ip6_hdr *);
556			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
557			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
558			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
559			noff = sizeof(struct ip6_hdr);
560			/* new mbuf contains only ipv6+icmpv6 headers */
561			n->m_len = noff + sizeof(struct icmp6_hdr);
562			/*
563			 * Adjust mbuf.  ip6_plen will be adjusted in
564			 * ip6_output().
565			 */
566			m_adj(n0, off + sizeof(struct icmp6_hdr));
567			/* recalculate complete packet size */
568			n->m_pkthdr.len = n0len + (noff - off);
569			n->m_next = n0;
570		} else {
571			nip6 = mtod(n, struct ip6_hdr *);
572			nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
573			noff = off;
574		}
575		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
576		nicmp6->icmp6_code = 0;
577		if (n) {
578			icmp6stat.icp6s_reflect++;
579			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
580			icmp6_reflect(n, noff);
581		}
582		break;
583
584	case ICMP6_ECHO_REPLY:
585		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
586		if (code != 0)
587			goto badcode;
588		break;
589
590	case MLD_LISTENER_QUERY:
591	case MLD_LISTENER_REPORT:
592		if (icmp6len < sizeof(struct mld_hdr))
593			goto badlen;
594		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
595			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
596		else
597			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
598		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
599			/* give up local */
600			mld6_input(m, off);
601			m = NULL;
602			goto freeit;
603		}
604		mld6_input(n, off);
605		/* m stays. */
606		break;
607
608	case MLD_LISTENER_DONE:
609		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
610		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
611			goto badlen;
612		break;		/* nothing to be done in kernel */
613
614	case MLD_MTRACE_RESP:
615	case MLD_MTRACE:
616		/* XXX: these two are experimental.  not officially defined. */
617		/* XXX: per-interface statistics? */
618		break;		/* just pass it to applications */
619
620	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
621	    {
622		enum { WRU, FQDN } mode;
623
624		if (!icmp6_nodeinfo)
625			break;
626
627		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
628			mode = WRU;
629		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
630			mode = FQDN;
631		else
632			goto badlen;
633
634#define hostnamelen	strlen(hostname)
635		if (mode == FQDN) {
636#ifndef PULLDOWN_TEST
637			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
638			    IPPROTO_DONE);
639#endif
640			n = m_copy(m, 0, M_COPYALL);
641			if (n)
642				n = ni6_input(n, off);
643			/* XXX meaningless if n == NULL */
644			noff = sizeof(struct ip6_hdr);
645		} else {
646			u_char *p;
647			int maxlen, maxhlen;
648
649			if ((icmp6_nodeinfo & 5) != 5)
650				break;
651
652			if (code != 0)
653				goto badcode;
654			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
655			if (maxlen >= MCLBYTES) {
656				/* Give up remote */
657				break;
658			}
659			MGETHDR(n, M_DONTWAIT, m->m_type);
660			if (n && maxlen > MHLEN) {
661				MCLGET(n, M_DONTWAIT);
662				if ((n->m_flags & M_EXT) == 0) {
663					m_free(n);
664					n = NULL;
665				}
666			}
667			if (!m_dup_pkthdr(n, m, M_DONTWAIT)) {
668				/*
669				 * Previous code did a blind M_COPY_PKTHDR
670				 * and said "just for rcvif".  If true, then
671				 * we could tolerate the dup failing (due to
672				 * the deep copy of the tag chain).  For now
673				 * be conservative and just fail.
674				 */
675				m_free(n);
676				n = NULL;
677			}
678			if (n == NULL) {
679				/* Give up remote */
680				break;
681			}
682			n->m_pkthdr.rcvif = NULL;
683			n->m_len = 0;
684			maxhlen = M_TRAILINGSPACE(n) - maxlen;
685			if (maxhlen > hostnamelen)
686				maxhlen = hostnamelen;
687			/*
688			 * Copy IPv6 and ICMPv6 only.
689			 */
690			nip6 = mtod(n, struct ip6_hdr *);
691			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
692			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
693			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
694			p = (u_char *)(nicmp6 + 1);
695			bzero(p, 4);
696			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
697			noff = sizeof(struct ip6_hdr);
698			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
699				sizeof(struct icmp6_hdr) + 4 + maxhlen;
700			nicmp6->icmp6_type = ICMP6_WRUREPLY;
701			nicmp6->icmp6_code = 0;
702		}
703#undef hostnamelen
704		if (n) {
705			icmp6stat.icp6s_reflect++;
706			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
707			icmp6_reflect(n, noff);
708		}
709		break;
710	    }
711
712	case ICMP6_WRUREPLY:
713		if (code != 0)
714			goto badcode;
715		break;
716
717	case ND_ROUTER_SOLICIT:
718		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
719		if (code != 0)
720			goto badcode;
721		if (icmp6len < sizeof(struct nd_router_solicit))
722			goto badlen;
723		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
724			/* give up local */
725			nd6_rs_input(m, off, icmp6len);
726			m = NULL;
727			goto freeit;
728		}
729		nd6_rs_input(n, off, icmp6len);
730		/* m stays. */
731		break;
732
733	case ND_ROUTER_ADVERT:
734		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
735		if (code != 0)
736			goto badcode;
737		if (icmp6len < sizeof(struct nd_router_advert))
738			goto badlen;
739		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
740			/* give up local */
741			nd6_ra_input(m, off, icmp6len);
742			m = NULL;
743			goto freeit;
744		}
745		nd6_ra_input(n, off, icmp6len);
746		/* m stays. */
747		break;
748
749	case ND_NEIGHBOR_SOLICIT:
750		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
751		if (code != 0)
752			goto badcode;
753		if (icmp6len < sizeof(struct nd_neighbor_solicit))
754			goto badlen;
755		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
756			/* give up local */
757			nd6_ns_input(m, off, icmp6len);
758			m = NULL;
759			goto freeit;
760		}
761		nd6_ns_input(n, off, icmp6len);
762		/* m stays. */
763		break;
764
765	case ND_NEIGHBOR_ADVERT:
766		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
767		if (code != 0)
768			goto badcode;
769		if (icmp6len < sizeof(struct nd_neighbor_advert))
770			goto badlen;
771		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
772			/* give up local */
773			nd6_na_input(m, off, icmp6len);
774			m = NULL;
775			goto freeit;
776		}
777		nd6_na_input(n, off, icmp6len);
778		/* m stays. */
779		break;
780
781	case ND_REDIRECT:
782		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
783		if (code != 0)
784			goto badcode;
785		if (icmp6len < sizeof(struct nd_redirect))
786			goto badlen;
787		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
788			/* give up local */
789			icmp6_redirect_input(m, off);
790			m = NULL;
791			goto freeit;
792		}
793		icmp6_redirect_input(n, off);
794		/* m stays. */
795		break;
796
797	case ICMP6_ROUTER_RENUMBERING:
798		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
799		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
800			goto badcode;
801		if (icmp6len < sizeof(struct icmp6_router_renum))
802			goto badlen;
803		break;
804
805	default:
806		nd6log((LOG_DEBUG,
807		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
808		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
809		    ip6_sprintf(&ip6->ip6_dst),
810		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
811		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
812			/* ICMPv6 error: MUST deliver it by spec... */
813			code = PRC_NCMDS;
814			/* deliver */
815		} else {
816			/* ICMPv6 informational: MUST not deliver */
817			break;
818		}
819	deliver:
820		if (icmp6_notify_error(&m, off, icmp6len, code)) {
821			/* In this case, m should've been freed. */
822			return (IPPROTO_DONE);
823		}
824		break;
825
826	badcode:
827		icmp6stat.icp6s_badcode++;
828		break;
829
830	badlen:
831		icmp6stat.icp6s_badlen++;
832		break;
833	}
834
835	/* deliver the packet to appropriate sockets */
836	icmp6_rip6_input(&m, *offp);
837
838	return IPPROTO_DONE;
839
840 freeit:
841	m_freem(m);
842	return IPPROTO_DONE;
843}
844
845static int
846icmp6_notify_error(mp, off, icmp6len, code)
847	struct mbuf **mp;
848	int off, icmp6len, code;
849{
850	struct mbuf *m = *mp;
851	struct icmp6_hdr *icmp6;
852	struct ip6_hdr *eip6;
853	u_int32_t notifymtu;
854	struct sockaddr_in6 icmp6src, icmp6dst;
855
856	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
857		icmp6stat.icp6s_tooshort++;
858		goto freeit;
859	}
860#ifndef PULLDOWN_TEST
861	IP6_EXTHDR_CHECK(m, off,
862	    sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
863	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
864#else
865	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
866	    sizeof(*icmp6) + sizeof(struct ip6_hdr));
867	if (icmp6 == NULL) {
868		icmp6stat.icp6s_tooshort++;
869		return (-1);
870	}
871#endif
872	eip6 = (struct ip6_hdr *)(icmp6 + 1);
873
874	/* Detect the upper level protocol */
875	{
876		void (*ctlfunc) __P((int, struct sockaddr *, void *));
877		u_int8_t nxt = eip6->ip6_nxt;
878		int eoff = off + sizeof(struct icmp6_hdr) +
879		    sizeof(struct ip6_hdr);
880		struct ip6ctlparam ip6cp;
881		struct in6_addr *finaldst = NULL;
882		int icmp6type = icmp6->icmp6_type;
883		struct ip6_frag *fh;
884		struct ip6_rthdr *rth;
885		struct ip6_rthdr0 *rth0;
886		int rthlen;
887
888		while (1) { /* XXX: should avoid infinite loop explicitly? */
889			struct ip6_ext *eh;
890
891			switch (nxt) {
892			case IPPROTO_HOPOPTS:
893			case IPPROTO_DSTOPTS:
894			case IPPROTO_AH:
895#ifndef PULLDOWN_TEST
896				IP6_EXTHDR_CHECK(m, 0,
897				    eoff + sizeof(struct ip6_ext), -1);
898				eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
899#else
900				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
901				    eoff, sizeof(*eh));
902				if (eh == NULL) {
903					icmp6stat.icp6s_tooshort++;
904					return (-1);
905				}
906#endif
907
908				if (nxt == IPPROTO_AH)
909					eoff += (eh->ip6e_len + 2) << 2;
910				else
911					eoff += (eh->ip6e_len + 1) << 3;
912				nxt = eh->ip6e_nxt;
913				break;
914			case IPPROTO_ROUTING:
915				/*
916				 * When the erroneous packet contains a
917				 * routing header, we should examine the
918				 * header to determine the final destination.
919				 * Otherwise, we can't properly update
920				 * information that depends on the final
921				 * destination (e.g. path MTU).
922				 */
923#ifndef PULLDOWN_TEST
924				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
925				rth = (struct ip6_rthdr *)
926				    (mtod(m, caddr_t) + eoff);
927#else
928				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
929				    eoff, sizeof(*rth));
930				if (rth == NULL) {
931					icmp6stat.icp6s_tooshort++;
932					return (-1);
933				}
934#endif
935				rthlen = (rth->ip6r_len + 1) << 3;
936				/*
937				 * XXX: currently there is no
938				 * officially defined type other
939				 * than type-0.
940				 * Note that if the segment left field
941				 * is 0, all intermediate hops must
942				 * have been passed.
943				 */
944				if (rth->ip6r_segleft &&
945				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
946					int hops;
947
948#ifndef PULLDOWN_TEST
949					IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
950					rth0 = (struct ip6_rthdr0 *)
951					    (mtod(m, caddr_t) + eoff);
952#else
953					IP6_EXTHDR_GET(rth0,
954					    struct ip6_rthdr0 *, m,
955					    eoff, rthlen);
956					if (rth0 == NULL) {
957						icmp6stat.icp6s_tooshort++;
958						return (-1);
959					}
960#endif
961					/* just ignore a bogus header */
962					if ((rth0->ip6r0_len % 2) == 0 &&
963					    (hops = rth0->ip6r0_len/2))
964						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
965				}
966				eoff += rthlen;
967				nxt = rth->ip6r_nxt;
968				break;
969			case IPPROTO_FRAGMENT:
970#ifndef PULLDOWN_TEST
971				IP6_EXTHDR_CHECK(m, 0, eoff +
972				    sizeof(struct ip6_frag), -1);
973				fh = (struct ip6_frag *)(mtod(m, caddr_t) +
974				    eoff);
975#else
976				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
977				    eoff, sizeof(*fh));
978				if (fh == NULL) {
979					icmp6stat.icp6s_tooshort++;
980					return (-1);
981				}
982#endif
983				/*
984				 * Data after a fragment header is meaningless
985				 * unless it is the first fragment, but
986				 * we'll go to the notify label for path MTU
987				 * discovery.
988				 */
989				if (fh->ip6f_offlg & IP6F_OFF_MASK)
990					goto notify;
991
992				eoff += sizeof(struct ip6_frag);
993				nxt = fh->ip6f_nxt;
994				break;
995			default:
996				/*
997				 * This case includes ESP and the No Next
998				 * Header.  In such cases going to the notify
999				 * label does not have any meaning
1000				 * (i.e. ctlfunc will be NULL), but we go
1001				 * anyway since we might have to update
1002				 * path MTU information.
1003				 */
1004				goto notify;
1005			}
1006		}
1007	  notify:
1008#ifndef PULLDOWN_TEST
1009		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1010#else
1011		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1012		    sizeof(*icmp6) + sizeof(struct ip6_hdr));
1013		if (icmp6 == NULL) {
1014			icmp6stat.icp6s_tooshort++;
1015			return (-1);
1016		}
1017#endif
1018
1019		/*
1020		 * retrieve parameters from the inner IPv6 header, and convert
1021		 * them into sockaddr structures.
1022		 * XXX: there is no guarantee that the source or destination
1023		 * addresses of the inner packet are in the same scope as
1024		 * the addresses of the icmp packet.  But there is no other
1025		 * way to determine the zone.
1026		 */
1027		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1028
1029		bzero(&icmp6dst, sizeof(icmp6dst));
1030		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1031		icmp6dst.sin6_family = AF_INET6;
1032		if (finaldst == NULL)
1033			icmp6dst.sin6_addr = eip6->ip6_dst;
1034		else
1035			icmp6dst.sin6_addr = *finaldst;
1036		if (in6_addr2zoneid(m->m_pkthdr.rcvif, &icmp6dst.sin6_addr,
1037		    &icmp6dst.sin6_scope_id))
1038			goto freeit;
1039		if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1040				   NULL, NULL)) {
1041			/* should be impossible */
1042			nd6log((LOG_DEBUG,
1043			    "icmp6_notify_error: in6_embedscope failed\n"));
1044			goto freeit;
1045		}
1046
1047		/*
1048		 * retrieve parameters from the inner IPv6 header, and convert
1049		 * them into sockaddr structures.
1050		 */
1051		bzero(&icmp6src, sizeof(icmp6src));
1052		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1053		icmp6src.sin6_family = AF_INET6;
1054		icmp6src.sin6_addr = eip6->ip6_src;
1055		if (in6_addr2zoneid(m->m_pkthdr.rcvif, &icmp6src.sin6_addr,
1056		    &icmp6src.sin6_scope_id)) {
1057			goto freeit;
1058		}
1059		if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1060				   NULL, NULL)) {
1061			/* should be impossible */
1062			nd6log((LOG_DEBUG,
1063			    "icmp6_notify_error: in6_embedscope failed\n"));
1064			goto freeit;
1065		}
1066		icmp6src.sin6_flowinfo = (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1067
1068		if (finaldst == NULL)
1069			finaldst = &eip6->ip6_dst;
1070		ip6cp.ip6c_m = m;
1071		ip6cp.ip6c_icmp6 = icmp6;
1072		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1073		ip6cp.ip6c_off = eoff;
1074		ip6cp.ip6c_finaldst = finaldst;
1075		ip6cp.ip6c_src = &icmp6src;
1076		ip6cp.ip6c_nxt = nxt;
1077
1078		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1079			notifymtu = ntohl(icmp6->icmp6_mtu);
1080			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1081			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1082		}
1083
1084		ctlfunc = (void (*) __P((int, struct sockaddr *, void *)))
1085		    (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1086		if (ctlfunc) {
1087			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1088			    &ip6cp);
1089		}
1090	}
1091	*mp = m;
1092	return (0);
1093
1094  freeit:
1095	m_freem(m);
1096	return (-1);
1097}
1098
1099void
1100icmp6_mtudisc_update(ip6cp, validated)
1101	struct ip6ctlparam *ip6cp;
1102	int validated;
1103{
1104	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1105	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1106	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1107	u_int mtu = ntohl(icmp6->icmp6_mtu);
1108	struct in_conninfo inc;
1109
1110#if 0
1111	/*
1112	 * RFC2460 section 5, last paragraph.
1113	 * even though minimum link MTU for IPv6 is IPV6_MMTU,
1114	 * we may see ICMPv6 too big with mtu < IPV6_MMTU
1115	 * due to packet translator in the middle.
1116	 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1117	 * special handling.
1118	 */
1119	if (mtu < IPV6_MMTU)
1120		return;
1121#endif
1122
1123	/*
1124	 * we reject ICMPv6 too big with abnormally small value.
1125	 * XXX what is the good definition of "abnormally small"?
1126	 */
1127	if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1128		return;
1129
1130	if (!validated)
1131		return;
1132
1133	bzero(&inc, sizeof(inc));
1134	inc.inc_flags = 1; /* IPv6 */
1135	inc.inc6_faddr = *dst;
1136	/* XXX normally, this won't happen */
1137	if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1138		inc.inc6_faddr.s6_addr16[1] =
1139		    htons(m->m_pkthdr.rcvif->if_index);
1140	}
1141
1142	if (mtu < tcp_maxmtu6(&inc)) {
1143		tcp_hc_updatemtu(&inc, mtu);
1144		icmp6stat.icp6s_pmtuchg++;
1145	}
1146}
1147
1148/*
1149 * Process a Node Information Query packet, based on
1150 * draft-ietf-ipngwg-icmp-name-lookups-07.
1151 *
1152 * Spec incompatibilities:
1153 * - IPv6 Subject address handling
1154 * - IPv4 Subject address handling support missing
1155 * - Proxy reply (answer even if it's not for me)
1156 * - joins NI group address at in6_ifattach() time only, does not cope
1157 *   with hostname changes by sethostname(3)
1158 */
1159#define hostnamelen	strlen(hostname)
1160static struct mbuf *
1161ni6_input(m, off)
1162	struct mbuf *m;
1163	int off;
1164{
1165	struct icmp6_nodeinfo *ni6, *nni6;
1166	struct mbuf *n = NULL;
1167	u_int16_t qtype;
1168	int subjlen;
1169	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1170	struct ni_reply_fqdn *fqdn;
1171	int addrs;		/* for NI_QTYPE_NODEADDR */
1172	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1173	struct sockaddr_in6 sin6_sbj; /* subject address */
1174	struct ip6_hdr *ip6;
1175	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1176	char *subj = NULL;
1177	struct in6_ifaddr *ia6 = NULL;
1178
1179	ip6 = mtod(m, struct ip6_hdr *);
1180#ifndef PULLDOWN_TEST
1181	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1182#else
1183	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1184	if (ni6 == NULL) {
1185		/* m is already reclaimed */
1186		return (NULL);
1187	}
1188#endif
1189
1190	/*
1191	 * Validate IPv6 destination address.
1192	 *
1193	 * The Responder must discard the Query without further processing
1194	 * unless it is one of the Responder's unicast or anycast addresses, or
1195	 * a link-local scope multicast address which the Responder has joined.
1196	 * [icmp-name-lookups-08, Section 4.]
1197	 */
1198	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1199		if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1200			goto bad;
1201		/* else it's a link-local multicast, fine */
1202	} else {		/* unicast or anycast */
1203		if ((ia6 = ip6_getdstifaddr(m)) == NULL)
1204			goto bad; /* XXX impossible */
1205
1206		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1207		    !(icmp6_nodeinfo & 4)) {
1208			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1209				"a temporary address in %s:%d",
1210			       __FILE__, __LINE__));
1211			goto bad;
1212		}
1213	}
1214
1215	/* validate query Subject field. */
1216	qtype = ntohs(ni6->ni_qtype);
1217	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1218	switch (qtype) {
1219	case NI_QTYPE_NOOP:
1220	case NI_QTYPE_SUPTYPES:
1221		/* 07 draft */
1222		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1223			break;
1224		/* FALLTHROUGH */
1225	case NI_QTYPE_FQDN:
1226	case NI_QTYPE_NODEADDR:
1227		switch (ni6->ni_code) {
1228		case ICMP6_NI_SUBJ_IPV6:
1229#if ICMP6_NI_SUBJ_IPV6 != 0
1230		case 0:
1231#endif
1232			/*
1233			 * backward compatibility - try to accept 03 draft
1234			 * format, where no Subject is present.
1235			 */
1236			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1237			    subjlen == 0) {
1238				oldfqdn++;
1239				break;
1240			}
1241#if ICMP6_NI_SUBJ_IPV6 != 0
1242			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1243				goto bad;
1244#endif
1245
1246			if (subjlen != sizeof(struct in6_addr))
1247				goto bad;
1248
1249			/*
1250			 * Validate Subject address.
1251			 *
1252			 * Not sure what exactly "address belongs to the node"
1253			 * means in the spec, is it just unicast, or what?
1254			 *
1255			 * At this moment we consider Subject address as
1256			 * "belong to the node" if the Subject address equals
1257			 * to the IPv6 destination address; validation for
1258			 * IPv6 destination address should have done enough
1259			 * check for us.
1260			 *
1261			 * We do not do proxy at this moment.
1262			 */
1263			/* m_pulldown instead of copy? */
1264			bzero(&sin6_sbj, sizeof(sin6_sbj));
1265			sin6_sbj.sin6_family = AF_INET6;
1266			sin6_sbj.sin6_len = sizeof(sin6_sbj);
1267			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1268			    subjlen, (caddr_t)&sin6_sbj.sin6_addr);
1269			if (in6_addr2zoneid(m->m_pkthdr.rcvif,
1270			    &sin6_sbj.sin6_addr, &sin6_sbj.sin6_scope_id)) {
1271				goto bad;
1272			}
1273			if (in6_embedscope(&sin6_sbj.sin6_addr, &sin6_sbj,
1274			    NULL, NULL))
1275				goto bad; /* XXX should not happen */
1276
1277			subj = (char *)&sin6_sbj;
1278			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1279			    &sin6_sbj.sin6_addr))
1280				break;
1281
1282			/*
1283			 * XXX if we are to allow other cases, we should really
1284			 * be careful about scope here.
1285			 * basically, we should disallow queries toward IPv6
1286			 * destination X with subject Y,
1287			 * if scope(X) > scope(Y).
1288			 * if we allow scope(X) > scope(Y), it will result in
1289			 * information leakage across scope boundary.
1290			 */
1291			goto bad;
1292
1293		case ICMP6_NI_SUBJ_FQDN:
1294			/*
1295			 * Validate Subject name with gethostname(3).
1296			 *
1297			 * The behavior may need some debate, since:
1298			 * - we are not sure if the node has FQDN as
1299			 *   hostname (returned by gethostname(3)).
1300			 * - the code does wildcard match for truncated names.
1301			 *   however, we are not sure if we want to perform
1302			 *   wildcard match, if gethostname(3) side has
1303			 *   truncated hostname.
1304			 */
1305			n = ni6_nametodns(hostname, hostnamelen, 0);
1306			if (!n || n->m_next || n->m_len == 0)
1307				goto bad;
1308			IP6_EXTHDR_GET(subj, char *, m,
1309			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1310			if (subj == NULL)
1311				goto bad;
1312			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1313			    n->m_len)) {
1314				goto bad;
1315			}
1316			m_freem(n);
1317			n = NULL;
1318			break;
1319
1320		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1321		default:
1322			goto bad;
1323		}
1324		break;
1325	}
1326
1327	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1328	switch (qtype) {
1329	case NI_QTYPE_FQDN:
1330		if ((icmp6_nodeinfo & 1) == 0)
1331			goto bad;
1332		break;
1333	case NI_QTYPE_NODEADDR:
1334		if ((icmp6_nodeinfo & 2) == 0)
1335			goto bad;
1336		break;
1337	}
1338
1339	/* guess reply length */
1340	switch (qtype) {
1341	case NI_QTYPE_NOOP:
1342		break;		/* no reply data */
1343	case NI_QTYPE_SUPTYPES:
1344		replylen += sizeof(u_int32_t);
1345		break;
1346	case NI_QTYPE_FQDN:
1347		/* XXX will append an mbuf */
1348		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1349		break;
1350	case NI_QTYPE_NODEADDR:
1351		addrs = ni6_addrs(ni6, m, &ifp, subj);
1352		if ((replylen += addrs * (sizeof(struct in6_addr) +
1353		    sizeof(u_int32_t))) > MCLBYTES)
1354			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1355		break;
1356	default:
1357		/*
1358		 * XXX: We must return a reply with the ICMP6 code
1359		 * `unknown Qtype' in this case.  However we regard the case
1360		 * as an FQDN query for backward compatibility.
1361		 * Older versions set a random value to this field,
1362		 * so it rarely varies in the defined qtypes.
1363		 * But the mechanism is not reliable...
1364		 * maybe we should obsolete older versions.
1365		 */
1366		qtype = NI_QTYPE_FQDN;
1367		/* XXX will append an mbuf */
1368		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1369		oldfqdn++;
1370		break;
1371	}
1372
1373	/* allocate an mbuf to reply. */
1374	MGETHDR(n, M_DONTWAIT, m->m_type);
1375	if (n == NULL) {
1376		m_freem(m);
1377		return (NULL);
1378	}
1379	M_MOVE_PKTHDR(n, m); /* just for recvif */
1380	if (replylen > MHLEN) {
1381		if (replylen > MCLBYTES) {
1382			/*
1383			 * XXX: should we try to allocate more? But MCLBYTES
1384			 * is probably much larger than IPV6_MMTU...
1385			 */
1386			goto bad;
1387		}
1388		MCLGET(n, M_DONTWAIT);
1389		if ((n->m_flags & M_EXT) == 0) {
1390			goto bad;
1391		}
1392	}
1393	n->m_pkthdr.len = n->m_len = replylen;
1394
1395	/* copy mbuf header and IPv6 + Node Information base headers */
1396	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1397	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1398	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1399
1400	/* qtype dependent procedure */
1401	switch (qtype) {
1402	case NI_QTYPE_NOOP:
1403		nni6->ni_code = ICMP6_NI_SUCCESS;
1404		nni6->ni_flags = 0;
1405		break;
1406	case NI_QTYPE_SUPTYPES:
1407	{
1408		u_int32_t v;
1409		nni6->ni_code = ICMP6_NI_SUCCESS;
1410		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1411		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1412		v = (u_int32_t)htonl(0x0000000f);
1413		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1414		break;
1415	}
1416	case NI_QTYPE_FQDN:
1417		nni6->ni_code = ICMP6_NI_SUCCESS;
1418		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1419		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1420		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1421		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1422		/*
1423		 * XXX do we really have FQDN in variable "hostname"?
1424		 */
1425		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1426		if (n->m_next == NULL)
1427			goto bad;
1428		/* XXX we assume that n->m_next is not a chain */
1429		if (n->m_next->m_next != NULL)
1430			goto bad;
1431		n->m_pkthdr.len += n->m_next->m_len;
1432		break;
1433	case NI_QTYPE_NODEADDR:
1434	{
1435		int lenlim, copied;
1436
1437		nni6->ni_code = ICMP6_NI_SUCCESS;
1438		n->m_pkthdr.len = n->m_len =
1439		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1440		lenlim = M_TRAILINGSPACE(n);
1441		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1442		/* XXX: reset mbuf length */
1443		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1444		    sizeof(struct icmp6_nodeinfo) + copied;
1445		break;
1446	}
1447	default:
1448		break;		/* XXX impossible! */
1449	}
1450
1451	nni6->ni_type = ICMP6_NI_REPLY;
1452	m_freem(m);
1453	return (n);
1454
1455  bad:
1456	m_freem(m);
1457	if (n)
1458		m_freem(n);
1459	return (NULL);
1460}
1461#undef hostnamelen
1462
1463/*
1464 * make a mbuf with DNS-encoded string.  no compression support.
1465 *
1466 * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1467 * treated as truncated name (two \0 at the end).  this is a wild guess.
1468 */
1469static struct mbuf *
1470ni6_nametodns(name, namelen, old)
1471	const char *name;
1472	int namelen;
1473	int old;	/* return pascal string if non-zero */
1474{
1475	struct mbuf *m;
1476	char *cp, *ep;
1477	const char *p, *q;
1478	int i, len, nterm;
1479
1480	if (old)
1481		len = namelen + 1;
1482	else
1483		len = MCLBYTES;
1484
1485	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1486	MGET(m, M_DONTWAIT, MT_DATA);
1487	if (m && len > MLEN) {
1488		MCLGET(m, M_DONTWAIT);
1489		if ((m->m_flags & M_EXT) == 0)
1490			goto fail;
1491	}
1492	if (!m)
1493		goto fail;
1494	m->m_next = NULL;
1495
1496	if (old) {
1497		m->m_len = len;
1498		*mtod(m, char *) = namelen;
1499		bcopy(name, mtod(m, char *) + 1, namelen);
1500		return m;
1501	} else {
1502		m->m_len = 0;
1503		cp = mtod(m, char *);
1504		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1505
1506		/* if not certain about my name, return empty buffer */
1507		if (namelen == 0)
1508			return m;
1509
1510		/*
1511		 * guess if it looks like shortened hostname, or FQDN.
1512		 * shortened hostname needs two trailing "\0".
1513		 */
1514		i = 0;
1515		for (p = name; p < name + namelen; p++) {
1516			if (*p && *p == '.')
1517				i++;
1518		}
1519		if (i < 2)
1520			nterm = 2;
1521		else
1522			nterm = 1;
1523
1524		p = name;
1525		while (cp < ep && p < name + namelen) {
1526			i = 0;
1527			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1528				i++;
1529			/* result does not fit into mbuf */
1530			if (cp + i + 1 >= ep)
1531				goto fail;
1532			/*
1533			 * DNS label length restriction, RFC1035 page 8.
1534			 * "i == 0" case is included here to avoid returning
1535			 * 0-length label on "foo..bar".
1536			 */
1537			if (i <= 0 || i >= 64)
1538				goto fail;
1539			*cp++ = i;
1540			bcopy(p, cp, i);
1541			cp += i;
1542			p = q;
1543			if (p < name + namelen && *p == '.')
1544				p++;
1545		}
1546		/* termination */
1547		if (cp + nterm >= ep)
1548			goto fail;
1549		while (nterm-- > 0)
1550			*cp++ = '\0';
1551		m->m_len = cp - mtod(m, char *);
1552		return m;
1553	}
1554
1555	panic("should not reach here");
1556	/* NOTREACHED */
1557
1558 fail:
1559	if (m)
1560		m_freem(m);
1561	return NULL;
1562}
1563
1564/*
1565 * check if two DNS-encoded string matches.  takes care of truncated
1566 * form (with \0\0 at the end).  no compression support.
1567 * XXX upper/lowercase match (see RFC2065)
1568 */
1569static int
1570ni6_dnsmatch(a, alen, b, blen)
1571	const char *a;
1572	int alen;
1573	const char *b;
1574	int blen;
1575{
1576	const char *a0, *b0;
1577	int l;
1578
1579	/* simplest case - need validation? */
1580	if (alen == blen && bcmp(a, b, alen) == 0)
1581		return 1;
1582
1583	a0 = a;
1584	b0 = b;
1585
1586	/* termination is mandatory */
1587	if (alen < 2 || blen < 2)
1588		return 0;
1589	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1590		return 0;
1591	alen--;
1592	blen--;
1593
1594	while (a - a0 < alen && b - b0 < blen) {
1595		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1596			return 0;
1597
1598		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1599			return 0;
1600		/* we don't support compression yet */
1601		if (a[0] >= 64 || b[0] >= 64)
1602			return 0;
1603
1604		/* truncated case */
1605		if (a[0] == 0 && a - a0 == alen - 1)
1606			return 1;
1607		if (b[0] == 0 && b - b0 == blen - 1)
1608			return 1;
1609		if (a[0] == 0 || b[0] == 0)
1610			return 0;
1611
1612		if (a[0] != b[0])
1613			return 0;
1614		l = a[0];
1615		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1616			return 0;
1617		if (bcmp(a + 1, b + 1, l) != 0)
1618			return 0;
1619
1620		a += 1 + l;
1621		b += 1 + l;
1622	}
1623
1624	if (a - a0 == alen && b - b0 == blen)
1625		return 1;
1626	else
1627		return 0;
1628}
1629
1630/*
1631 * calculate the number of addresses to be returned in the node info reply.
1632 */
1633static int
1634ni6_addrs(ni6, m, ifpp, subj)
1635	struct icmp6_nodeinfo *ni6;
1636	struct mbuf *m;
1637	struct ifnet **ifpp;
1638	char *subj;
1639{
1640	struct ifnet *ifp;
1641	struct in6_ifaddr *ifa6;
1642	struct ifaddr *ifa;
1643	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1644	int addrs = 0, addrsofif, iffound = 0;
1645	int niflags = ni6->ni_flags;
1646
1647	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1648		switch (ni6->ni_code) {
1649		case ICMP6_NI_SUBJ_IPV6:
1650			if (subj == NULL) /* must be impossible... */
1651				return (0);
1652			subj_ip6 = (struct sockaddr_in6 *)subj;
1653			break;
1654		default:
1655			/*
1656			 * XXX: we only support IPv6 subject address for
1657			 * this Qtype.
1658			 */
1659			return (0);
1660		}
1661	}
1662
1663	IFNET_RLOCK();
1664	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1665		addrsofif = 0;
1666		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1667			if (ifa->ifa_addr->sa_family != AF_INET6)
1668				continue;
1669			ifa6 = (struct in6_ifaddr *)ifa;
1670
1671			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1672			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1673			    &ifa6->ia_addr.sin6_addr))
1674				iffound = 1;
1675
1676			/*
1677			 * IPv4-mapped addresses can only be returned by a
1678			 * Node Information proxy, since they represent
1679			 * addresses of IPv4-only nodes, which perforce do
1680			 * not implement this protocol.
1681			 * [icmp-name-lookups-07, Section 5.4]
1682			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1683			 * this function at this moment.
1684			 */
1685
1686			/* What do we have to do about ::1? */
1687			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1688			case IPV6_ADDR_SCOPE_LINKLOCAL:
1689				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1690					continue;
1691				break;
1692			case IPV6_ADDR_SCOPE_SITELOCAL:
1693				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1694					continue;
1695				break;
1696			case IPV6_ADDR_SCOPE_GLOBAL:
1697				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1698					continue;
1699				break;
1700			default:
1701				continue;
1702			}
1703
1704			/*
1705			 * check if anycast is okay.
1706			 * XXX: just experimental.  not in the spec.
1707			 */
1708			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1709			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1710				continue; /* we need only unicast addresses */
1711			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1712			    (icmp6_nodeinfo & 4) == 0) {
1713				continue;
1714			}
1715			addrsofif++; /* count the address */
1716		}
1717		if (iffound) {
1718			*ifpp = ifp;
1719			IFNET_RUNLOCK();
1720			return (addrsofif);
1721		}
1722
1723		addrs += addrsofif;
1724	}
1725	IFNET_RUNLOCK();
1726
1727	return (addrs);
1728}
1729
1730static int
1731ni6_store_addrs(ni6, nni6, ifp0, resid)
1732	struct icmp6_nodeinfo *ni6, *nni6;
1733	struct ifnet *ifp0;
1734	int resid;
1735{
1736	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1737	struct in6_ifaddr *ifa6;
1738	struct ifaddr *ifa;
1739	struct ifnet *ifp_dep = NULL;
1740	int copied = 0, allow_deprecated = 0;
1741	u_char *cp = (u_char *)(nni6 + 1);
1742	int niflags = ni6->ni_flags;
1743	u_int32_t ltime;
1744
1745	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1746		return (0);	/* needless to copy */
1747
1748	IFNET_RLOCK();
1749  again:
1750
1751	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
1752		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1753		     ifa = ifa->ifa_list.tqe_next) {
1754			if (ifa->ifa_addr->sa_family != AF_INET6)
1755				continue;
1756			ifa6 = (struct in6_ifaddr *)ifa;
1757
1758			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1759			    allow_deprecated == 0) {
1760				/*
1761				 * prefererred address should be put before
1762				 * deprecated addresses.
1763				 */
1764
1765				/* record the interface for later search */
1766				if (ifp_dep == NULL)
1767					ifp_dep = ifp;
1768
1769				continue;
1770			} else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1771			    allow_deprecated != 0)
1772				continue; /* we now collect deprecated addrs */
1773
1774			/* What do we have to do about ::1? */
1775			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1776			case IPV6_ADDR_SCOPE_LINKLOCAL:
1777				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1778					continue;
1779				break;
1780			case IPV6_ADDR_SCOPE_SITELOCAL:
1781				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1782					continue;
1783				break;
1784			case IPV6_ADDR_SCOPE_GLOBAL:
1785				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1786					continue;
1787				break;
1788			default:
1789				continue;
1790			}
1791
1792			/*
1793			 * check if anycast is okay.
1794			 * XXX: just experimental.  not in the spec.
1795			 */
1796			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1797			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1798				continue;
1799			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1800			    (icmp6_nodeinfo & 4) == 0) {
1801				continue;
1802			}
1803
1804			/* now we can copy the address */
1805			if (resid < sizeof(struct in6_addr) +
1806			    sizeof(u_int32_t)) {
1807				/*
1808				 * We give up much more copy.
1809				 * Set the truncate flag and return.
1810				 */
1811				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1812				IFNET_RUNLOCK();
1813				return (copied);
1814			}
1815
1816			/*
1817			 * Set the TTL of the address.
1818			 * The TTL value should be one of the following
1819			 * according to the specification:
1820			 *
1821			 * 1. The remaining lifetime of a DHCP lease on the
1822			 *    address, or
1823			 * 2. The remaining Valid Lifetime of a prefix from
1824			 *    which the address was derived through Stateless
1825			 *    Autoconfiguration.
1826			 *
1827			 * Note that we currently do not support stateful
1828			 * address configuration by DHCPv6, so the former
1829			 * case can't happen.
1830			 */
1831			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1832				ltime = ND6_INFINITE_LIFETIME;
1833			else {
1834				if (ifa6->ia6_lifetime.ia6t_expire >
1835				    time_second)
1836					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1837				else
1838					ltime = 0;
1839			}
1840
1841			bcopy(&ltime, cp, sizeof(u_int32_t));
1842			cp += sizeof(u_int32_t);
1843
1844			/* copy the address itself */
1845			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1846			    sizeof(struct in6_addr));
1847			in6_clearscope((struct in6_addr *)cp); /* XXX */
1848			cp += sizeof(struct in6_addr);
1849
1850			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1851			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1852		}
1853		if (ifp0)	/* we need search only on the specified IF */
1854			break;
1855	}
1856
1857	if (allow_deprecated == 0 && ifp_dep != NULL) {
1858		ifp = ifp_dep;
1859		allow_deprecated = 1;
1860
1861		goto again;
1862	}
1863
1864	IFNET_RUNLOCK();
1865
1866	return (copied);
1867}
1868
1869/*
1870 * XXX almost dup'ed code with rip6_input.
1871 */
1872static int
1873icmp6_rip6_input(mp, off)
1874	struct	mbuf **mp;
1875	int	off;
1876{
1877	struct mbuf *m = *mp;
1878	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1879	struct in6pcb *in6p;
1880	struct in6pcb *last = NULL;
1881	struct sockaddr_in6 fromsa;
1882	struct icmp6_hdr *icmp6;
1883	struct mbuf *opts = NULL;
1884
1885#ifndef PULLDOWN_TEST
1886	/* this is assumed to be safe. */
1887	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1888#else
1889	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1890	if (icmp6 == NULL) {
1891		/* m is already reclaimed */
1892		return (IPPROTO_DONE);
1893	}
1894#endif
1895
1896	bzero(&fromsa, sizeof(fromsa));
1897	fromsa.sin6_len = sizeof(struct sockaddr_in6);
1898	fromsa.sin6_family = AF_INET6;
1899	/* KAME hack: recover scopeid */
1900	(void)in6_recoverscope(&fromsa, &ip6->ip6_src, m->m_pkthdr.rcvif);
1901
1902	INP_INFO_RLOCK(&ripcbinfo);
1903	LIST_FOREACH(in6p, &ripcb, inp_list) {
1904		INP_LOCK(in6p);
1905		if ((in6p->inp_vflag & INP_IPV6) == 0) {
1906	docontinue:
1907			INP_UNLOCK(in6p);
1908			continue;
1909		}
1910		if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1911			goto docontinue;
1912		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1913		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1914			goto docontinue;
1915		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1916		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1917			goto docontinue;
1918		if (in6p->in6p_icmp6filt
1919		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1920				 in6p->in6p_icmp6filt))
1921			goto docontinue;
1922		if (last) {
1923			struct	mbuf *n = NULL;
1924
1925			/*
1926			 * Recent network drivers tend to allocate a single
1927			 * mbuf cluster, rather than to make a couple of
1928			 * mbufs without clusters.  Also, since the IPv6 code
1929			 * path tries to avoid m_pullup(), it is highly
1930			 * probable that we still have an mbuf cluster here
1931			 * even though the necessary length can be stored in an
1932			 * mbuf's internal buffer.
1933			 * Meanwhile, the default size of the receive socket
1934			 * buffer for raw sockets is not so large.  This means
1935			 * the possibility of packet loss is relatively higher
1936			 * than before.  To avoid this scenario, we copy the
1937			 * received data to a separate mbuf that does not use
1938			 * a cluster, if possible.
1939			 * XXX: it is better to copy the data after stripping
1940			 * intermediate headers.
1941			 */
1942			if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1943			    m->m_len <= MHLEN) {
1944				MGET(n, M_DONTWAIT, m->m_type);
1945				if (n != NULL) {
1946					m_dup_pkthdr(n, m, M_NOWAIT);
1947					bcopy(m->m_data, n->m_data, m->m_len);
1948					n->m_len = m->m_len;
1949				}
1950			}
1951			if (n != NULL ||
1952			    (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1953				if (last->in6p_flags & IN6P_CONTROLOPTS)
1954					ip6_savecontrol(last, n, &opts);
1955				/* strip intermediate headers */
1956				m_adj(n, off);
1957				if (sbappendaddr(&last->in6p_socket->so_rcv,
1958				    (struct sockaddr *)&fromsa, n, opts)
1959				    == 0) {
1960					/* should notify about lost packet */
1961					m_freem(n);
1962					if (opts) {
1963						m_freem(opts);
1964					}
1965				} else
1966					sorwakeup(last->in6p_socket);
1967				opts = NULL;
1968			}
1969			INP_UNLOCK(last);
1970		}
1971		last = in6p;
1972	}
1973	if (last) {
1974		if (last->in6p_flags & IN6P_CONTROLOPTS)
1975			ip6_savecontrol(last, m, &opts);
1976		/* strip intermediate headers */
1977		m_adj(m, off);
1978
1979		/* avoid using mbuf clusters if possible (see above) */
1980		if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1981		    m->m_len <= MHLEN) {
1982			struct mbuf *n;
1983
1984			MGET(n, M_DONTWAIT, m->m_type);
1985			if (n != NULL) {
1986				m_dup_pkthdr(n, m, M_NOWAIT);
1987				bcopy(m->m_data, n->m_data, m->m_len);
1988				n->m_len = m->m_len;
1989
1990				m_freem(m);
1991				m = n;
1992			}
1993		}
1994		if (sbappendaddr(&last->in6p_socket->so_rcv,
1995		    (struct sockaddr *)&fromsa, m, opts) == 0) {
1996			m_freem(m);
1997			if (opts)
1998				m_freem(opts);
1999		} else
2000			sorwakeup(last->in6p_socket);
2001		INP_UNLOCK(last);
2002	} else {
2003		m_freem(m);
2004		ip6stat.ip6s_delivered--;
2005	}
2006	INP_INFO_RUNLOCK(&ripcbinfo);
2007	return IPPROTO_DONE;
2008}
2009
2010/*
2011 * Reflect the ip6 packet back to the source.
2012 * OFF points to the icmp6 header, counted from the top of the mbuf.
2013 */
2014void
2015icmp6_reflect(m, off)
2016	struct	mbuf *m;
2017	size_t off;
2018{
2019	struct ip6_hdr *ip6;
2020	struct icmp6_hdr *icmp6;
2021	struct in6_ifaddr *ia;
2022	struct in6_addr t, *src = 0;
2023	int plen;
2024	int type, code;
2025	struct ifnet *outif = NULL;
2026	struct sockaddr_in6 sa6_src, sa6_dst;
2027#ifdef COMPAT_RFC1885
2028	int mtu = IPV6_MMTU;
2029	struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
2030#endif
2031
2032	/* too short to reflect */
2033	if (off < sizeof(struct ip6_hdr)) {
2034		nd6log((LOG_DEBUG,
2035		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2036		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2037		    __FILE__, __LINE__));
2038		goto bad;
2039	}
2040
2041	/*
2042	 * If there are extra headers between IPv6 and ICMPv6, strip
2043	 * off that header first.
2044	 */
2045#ifdef DIAGNOSTIC
2046	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2047		panic("assumption failed in icmp6_reflect");
2048#endif
2049	if (off > sizeof(struct ip6_hdr)) {
2050		size_t l;
2051		struct ip6_hdr nip6;
2052
2053		l = off - sizeof(struct ip6_hdr);
2054		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2055		m_adj(m, l);
2056		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2057		if (m->m_len < l) {
2058			if ((m = m_pullup(m, l)) == NULL)
2059				return;
2060		}
2061		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2062	} else /* off == sizeof(struct ip6_hdr) */ {
2063		size_t l;
2064		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2065		if (m->m_len < l) {
2066			if ((m = m_pullup(m, l)) == NULL)
2067				return;
2068		}
2069	}
2070	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2071	ip6 = mtod(m, struct ip6_hdr *);
2072	ip6->ip6_nxt = IPPROTO_ICMPV6;
2073	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2074	type = icmp6->icmp6_type; /* keep type for statistics */
2075	code = icmp6->icmp6_code; /* ditto. */
2076
2077	t = ip6->ip6_dst;
2078	/*
2079	 * ip6_input() drops a packet if its src is multicast.
2080	 * So, the src is never multicast.
2081	 */
2082	ip6->ip6_dst = ip6->ip6_src;
2083
2084	/*
2085	 * XXX: make sure to embed scope zone information, using
2086	 * already embedded IDs or the received interface (if any).
2087	 * Note that rcvif may be NULL.
2088	 * TODO: scoped routing case (XXX).
2089	 */
2090	bzero(&sa6_src, sizeof(sa6_src));
2091	sa6_src.sin6_family = AF_INET6;
2092	sa6_src.sin6_len = sizeof(sa6_src);
2093	sa6_src.sin6_addr = ip6->ip6_dst;
2094	in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
2095	in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL);
2096	bzero(&sa6_dst, sizeof(sa6_dst));
2097	sa6_dst.sin6_family = AF_INET6;
2098	sa6_dst.sin6_len = sizeof(sa6_dst);
2099	sa6_dst.sin6_addr = t;
2100	in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
2101	in6_embedscope(&t, &sa6_dst, NULL, NULL);
2102
2103#ifdef COMPAT_RFC1885
2104	/*
2105	 * xxx guess MTU
2106	 * RFC 1885 requires that echo reply should be truncated if it
2107	 * does not fit in with (return) path MTU, but the description was
2108	 * removed in the new spec.
2109	 */
2110	if (icmp6_reflect_rt.ro_rt == 0 ||
2111	    ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
2112		if (icmp6_reflect_rt.ro_rt) {
2113			RTFREE(icmp6_reflect_rt.ro_rt);
2114			icmp6_reflect_rt.ro_rt = 0;
2115		}
2116		bzero(sin6, sizeof(*sin6));
2117		sin6->sin6_family = PF_INET6;
2118		sin6->sin6_len = sizeof(struct sockaddr_in6);
2119		sin6->sin6_addr = ip6->ip6_dst;
2120
2121		rtalloc((struct route *)&icmp6_reflect_rt.ro_rt);
2122	}
2123
2124	if (icmp6_reflect_rt.ro_rt == 0)
2125		goto bad;
2126
2127	if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
2128	    && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
2129		mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
2130
2131	if (mtu < m->m_pkthdr.len) {
2132		plen -= (m->m_pkthdr.len - mtu);
2133		m_adj(m, mtu - m->m_pkthdr.len);
2134	}
2135#endif
2136	/*
2137	 * If the incoming packet was addressed directly to us (i.e. unicast),
2138	 * use dst as the src for the reply.
2139	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2140	 * (for example) when we encounter an error while forwarding procedure
2141	 * destined to a duplicated address of ours.
2142	 */
2143	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2144		if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2145		    (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2146			src = &t;
2147			break;
2148		}
2149	if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2150		/*
2151		 * This is the case if the dst is our link-local address
2152		 * and the sender is also ourselves.
2153		 */
2154		src = &t;
2155	}
2156
2157	if (src == 0) {
2158		int e;
2159		struct route_in6 ro;
2160
2161		/*
2162		 * This case matches to multicasts, our anycast, or unicasts
2163		 * that we do not own.  Select a source address based on the
2164		 * source address of the erroneous packet.
2165		 */
2166		bzero(&ro, sizeof(ro));
2167		src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e);
2168		if (ro.ro_rt)
2169			RTFREE(ro.ro_rt); /* XXX: we could use this */
2170		if (src == NULL) {
2171			nd6log((LOG_DEBUG,
2172			    "icmp6_reflect: source can't be determined: "
2173			    "dst=%s, error=%d\n",
2174			    ip6_sprintf(&sa6_src.sin6_addr), e));
2175			goto bad;
2176		}
2177	}
2178
2179	ip6->ip6_src = *src;
2180
2181	ip6->ip6_flow = 0;
2182	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2183	ip6->ip6_vfc |= IPV6_VERSION;
2184	ip6->ip6_nxt = IPPROTO_ICMPV6;
2185	if (outif)
2186		ip6->ip6_hlim = ND_IFINFO(outif)->chlim;
2187	else if (m->m_pkthdr.rcvif) {
2188		/* XXX: This may not be the outgoing interface */
2189		ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2190	} else
2191		ip6->ip6_hlim = ip6_defhlim;
2192
2193	icmp6->icmp6_cksum = 0;
2194	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2195	    sizeof(struct ip6_hdr), plen);
2196
2197	/*
2198	 * XXX option handling
2199	 */
2200
2201	m->m_flags &= ~(M_BCAST|M_MCAST);
2202
2203#ifdef COMPAT_RFC1885
2204	ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif, NULL);
2205#else
2206	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2207#endif
2208	if (outif)
2209		icmp6_ifoutstat_inc(outif, type, code);
2210
2211	return;
2212
2213 bad:
2214	m_freem(m);
2215	return;
2216}
2217
2218void
2219icmp6_fasttimo()
2220{
2221
2222	mld6_fasttimeo();
2223}
2224
2225static const char *
2226icmp6_redirect_diag(src6, dst6, tgt6)
2227	struct in6_addr *src6;
2228	struct in6_addr *dst6;
2229	struct in6_addr *tgt6;
2230{
2231	static char buf[1024];
2232	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2233	    ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2234	return buf;
2235}
2236
2237void
2238icmp6_redirect_input(m, off)
2239	struct mbuf *m;
2240	int off;
2241{
2242	struct ifnet *ifp = m->m_pkthdr.rcvif;
2243	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2244	struct nd_redirect *nd_rd;
2245	int icmp6len = ntohs(ip6->ip6_plen);
2246	char *lladdr = NULL;
2247	int lladdrlen = 0;
2248	u_char *redirhdr = NULL;
2249	int redirhdrlen = 0;
2250	struct rtentry *rt = NULL;
2251	int is_router;
2252	int is_onlink;
2253	struct in6_addr src6 = ip6->ip6_src;
2254	struct in6_addr redtgt6;
2255	struct in6_addr reddst6;
2256	union nd_opts ndopts;
2257
2258	if (!m || !ifp)
2259		return;
2260
2261	/* XXX if we are router, we don't update route by icmp6 redirect */
2262	if (ip6_forwarding)
2263		goto freeit;
2264	if (!icmp6_rediraccept)
2265		goto freeit;
2266
2267#ifndef PULLDOWN_TEST
2268	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2269	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2270#else
2271	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2272	if (nd_rd == NULL) {
2273		icmp6stat.icp6s_tooshort++;
2274		return;
2275	}
2276#endif
2277	redtgt6 = nd_rd->nd_rd_target;
2278	reddst6 = nd_rd->nd_rd_dst;
2279
2280	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2281		redtgt6.s6_addr16[1] = htons(ifp->if_index);
2282	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2283		reddst6.s6_addr16[1] = htons(ifp->if_index);
2284
2285	/* validation */
2286	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2287		nd6log((LOG_ERR,
2288		    "ICMP6 redirect sent from %s rejected; "
2289		    "must be from linklocal\n",
2290		    ip6_sprintf(&src6)));
2291		goto bad;
2292	}
2293	if (ip6->ip6_hlim != 255) {
2294		nd6log((LOG_ERR,
2295		    "ICMP6 redirect sent from %s rejected; "
2296		    "hlim=%d (must be 255)\n",
2297		    ip6_sprintf(&src6), ip6->ip6_hlim));
2298		goto bad;
2299	}
2300    {
2301	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2302	struct sockaddr_in6 sin6;
2303	struct in6_addr *gw6;
2304
2305	bzero(&sin6, sizeof(sin6));
2306	sin6.sin6_family = AF_INET6;
2307	sin6.sin6_len = sizeof(struct sockaddr_in6);
2308	bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2309	rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL);
2310	if (rt) {
2311		if (rt->rt_gateway == NULL ||
2312		    rt->rt_gateway->sa_family != AF_INET6) {
2313			nd6log((LOG_ERR,
2314			    "ICMP6 redirect rejected; no route "
2315			    "with inet6 gateway found for redirect dst: %s\n",
2316			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2317			RTFREE_LOCKED(rt);
2318			goto bad;
2319		}
2320
2321		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2322		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2323			nd6log((LOG_ERR,
2324			    "ICMP6 redirect rejected; "
2325			    "not equal to gw-for-src=%s (must be same): "
2326			    "%s\n",
2327			    ip6_sprintf(gw6),
2328			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2329			RTFREE_LOCKED(rt);
2330			goto bad;
2331		}
2332	} else {
2333		nd6log((LOG_ERR,
2334		    "ICMP6 redirect rejected; "
2335		    "no route found for redirect dst: %s\n",
2336		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2337		goto bad;
2338	}
2339	RTFREE_LOCKED(rt);
2340	rt = NULL;
2341    }
2342	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2343		nd6log((LOG_ERR,
2344		    "ICMP6 redirect rejected; "
2345		    "redirect dst must be unicast: %s\n",
2346		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2347		goto bad;
2348	}
2349
2350	is_router = is_onlink = 0;
2351	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2352		is_router = 1;	/* router case */
2353	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2354		is_onlink = 1;	/* on-link destination case */
2355	if (!is_router && !is_onlink) {
2356		nd6log((LOG_ERR,
2357		    "ICMP6 redirect rejected; "
2358		    "neither router case nor onlink case: %s\n",
2359		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2360		goto bad;
2361	}
2362	/* validation passed */
2363
2364	icmp6len -= sizeof(*nd_rd);
2365	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2366	if (nd6_options(&ndopts) < 0) {
2367		nd6log((LOG_INFO, "icmp6_redirect_input: "
2368		    "invalid ND option, rejected: %s\n",
2369		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2370		/* nd6_options have incremented stats */
2371		goto freeit;
2372	}
2373
2374	if (ndopts.nd_opts_tgt_lladdr) {
2375		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2376		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2377	}
2378
2379	if (ndopts.nd_opts_rh) {
2380		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2381		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2382	}
2383
2384	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2385		nd6log((LOG_INFO,
2386		    "icmp6_redirect_input: lladdrlen mismatch for %s "
2387		    "(if %d, icmp6 packet %d): %s\n",
2388		    ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2389		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2390		goto bad;
2391	}
2392
2393	/* RFC 2461 8.3 */
2394	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2395	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2396
2397	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2398		/* perform rtredirect */
2399		struct sockaddr_in6 sdst;
2400		struct sockaddr_in6 sgw;
2401		struct sockaddr_in6 ssrc;
2402
2403		bzero(&sdst, sizeof(sdst));
2404		bzero(&sgw, sizeof(sgw));
2405		bzero(&ssrc, sizeof(ssrc));
2406		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2407		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2408			sizeof(struct sockaddr_in6);
2409		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2410		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2411		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2412		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2413		    (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2414		    (struct sockaddr *)&ssrc);
2415	}
2416	/* finally update cached route in each socket via pfctlinput */
2417    {
2418	struct sockaddr_in6 sdst;
2419
2420	bzero(&sdst, sizeof(sdst));
2421	sdst.sin6_family = AF_INET6;
2422	sdst.sin6_len = sizeof(struct sockaddr_in6);
2423	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2424	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2425#if defined(IPSEC) || defined(FAST_IPSEC)
2426	key_sa_routechange((struct sockaddr *)&sdst);
2427#endif
2428    }
2429
2430 freeit:
2431	m_freem(m);
2432	return;
2433
2434 bad:
2435	icmp6stat.icp6s_badredirect++;
2436	m_freem(m);
2437}
2438
2439void
2440icmp6_redirect_output(m0, rt)
2441	struct mbuf *m0;
2442	struct rtentry *rt;
2443{
2444	struct ifnet *ifp;	/* my outgoing interface */
2445	struct in6_addr *ifp_ll6;
2446	struct in6_addr *router_ll6;
2447	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2448	struct mbuf *m = NULL;	/* newly allocated one */
2449	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2450	struct nd_redirect *nd_rd;
2451	size_t maxlen;
2452	u_char *p;
2453	struct ifnet *outif = NULL;
2454	struct sockaddr_in6 src_sa;
2455
2456	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2457
2458	/* if we are not router, we don't send icmp6 redirect */
2459	if (!ip6_forwarding || ip6_accept_rtadv)
2460		goto fail;
2461
2462	/* sanity check */
2463	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2464		goto fail;
2465
2466	/*
2467	 * Address check:
2468	 *  the source address must identify a neighbor, and
2469	 *  the destination address must not be a multicast address
2470	 *  [RFC 2461, sec 8.2]
2471	 */
2472	sip6 = mtod(m0, struct ip6_hdr *);
2473	bzero(&src_sa, sizeof(src_sa));
2474	src_sa.sin6_family = AF_INET6;
2475	src_sa.sin6_len = sizeof(src_sa);
2476	src_sa.sin6_addr = sip6->ip6_src;
2477	/* we don't currently use sin6_scope_id, but eventually use it */
2478	if (in6_addr2zoneid(ifp, &sip6->ip6_src, &src_sa.sin6_scope_id))
2479		goto fail;
2480	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2481		goto fail;
2482	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2483		goto fail;	/* what should we do here? */
2484
2485	/* rate limit */
2486	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2487		goto fail;
2488
2489	/*
2490	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2491	 * we almost always ask for an mbuf cluster for simplicity.
2492	 * (MHLEN < IPV6_MMTU is almost always true)
2493	 */
2494#if IPV6_MMTU >= MCLBYTES
2495# error assumption failed about IPV6_MMTU and MCLBYTES
2496#endif
2497	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2498	if (m && IPV6_MMTU >= MHLEN)
2499		MCLGET(m, M_DONTWAIT);
2500	if (!m)
2501		goto fail;
2502	m->m_pkthdr.rcvif = NULL;
2503	m->m_len = 0;
2504	maxlen = M_TRAILINGSPACE(m);
2505	maxlen = min(IPV6_MMTU, maxlen);
2506	/* just for safety */
2507	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2508	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2509		goto fail;
2510	}
2511
2512	{
2513		/* get ip6 linklocal address for ifp(my outgoing interface). */
2514		struct in6_ifaddr *ia;
2515		if ((ia = in6ifa_ifpforlinklocal(ifp,
2516						 IN6_IFF_NOTREADY|
2517						 IN6_IFF_ANYCAST)) == NULL)
2518			goto fail;
2519		ifp_ll6 = &ia->ia_addr.sin6_addr;
2520	}
2521
2522	/* get ip6 linklocal address for the router. */
2523	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2524		struct sockaddr_in6 *sin6;
2525		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2526		router_ll6 = &sin6->sin6_addr;
2527		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2528			router_ll6 = (struct in6_addr *)NULL;
2529	} else
2530		router_ll6 = (struct in6_addr *)NULL;
2531
2532	/* ip6 */
2533	ip6 = mtod(m, struct ip6_hdr *);
2534	ip6->ip6_flow = 0;
2535	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2536	ip6->ip6_vfc |= IPV6_VERSION;
2537	/* ip6->ip6_plen will be set later */
2538	ip6->ip6_nxt = IPPROTO_ICMPV6;
2539	ip6->ip6_hlim = 255;
2540	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2541	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2542	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2543
2544	/* ND Redirect */
2545	nd_rd = (struct nd_redirect *)(ip6 + 1);
2546	nd_rd->nd_rd_type = ND_REDIRECT;
2547	nd_rd->nd_rd_code = 0;
2548	nd_rd->nd_rd_reserved = 0;
2549	if (rt->rt_flags & RTF_GATEWAY) {
2550		/*
2551		 * nd_rd->nd_rd_target must be a link-local address in
2552		 * better router cases.
2553		 */
2554		if (!router_ll6)
2555			goto fail;
2556		bcopy(router_ll6, &nd_rd->nd_rd_target,
2557		    sizeof(nd_rd->nd_rd_target));
2558		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2559		    sizeof(nd_rd->nd_rd_dst));
2560	} else {
2561		/* make sure redtgt == reddst */
2562		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2563		    sizeof(nd_rd->nd_rd_target));
2564		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2565		    sizeof(nd_rd->nd_rd_dst));
2566	}
2567
2568	p = (u_char *)(nd_rd + 1);
2569
2570	if (!router_ll6)
2571		goto nolladdropt;
2572
2573	{
2574		/* target lladdr option */
2575		struct rtentry *rt_router = NULL;
2576		int len;
2577		struct sockaddr_dl *sdl;
2578		struct nd_opt_hdr *nd_opt;
2579		char *lladdr;
2580
2581		rt_router = nd6_lookup(router_ll6, 0, ifp);
2582		if (!rt_router)
2583			goto nolladdropt;
2584		len = sizeof(*nd_opt) + ifp->if_addrlen;
2585		len = (len + 7) & ~7;	/* round by 8 */
2586		/* safety check */
2587		if (len + (p - (u_char *)ip6) > maxlen)
2588			goto nolladdropt;
2589		if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2590		    (rt_router->rt_flags & RTF_LLINFO) &&
2591		    (rt_router->rt_gateway->sa_family == AF_LINK) &&
2592		    (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2593		    sdl->sdl_alen) {
2594			nd_opt = (struct nd_opt_hdr *)p;
2595			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2596			nd_opt->nd_opt_len = len >> 3;
2597			lladdr = (char *)(nd_opt + 1);
2598			bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2599			p += len;
2600		}
2601	}
2602nolladdropt:;
2603
2604	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2605
2606	/* just to be safe */
2607#ifdef M_DECRYPTED	/*not openbsd*/
2608	if (m0->m_flags & M_DECRYPTED)
2609		goto noredhdropt;
2610#endif
2611	if (p - (u_char *)ip6 > maxlen)
2612		goto noredhdropt;
2613
2614	{
2615		/* redirected header option */
2616		int len;
2617		struct nd_opt_rd_hdr *nd_opt_rh;
2618
2619		/*
2620		 * compute the maximum size for icmp6 redirect header option.
2621		 * XXX room for auth header?
2622		 */
2623		len = maxlen - (p - (u_char *)ip6);
2624		len &= ~7;
2625
2626		/* This is just for simplicity. */
2627		if (m0->m_pkthdr.len != m0->m_len) {
2628			if (m0->m_next) {
2629				m_freem(m0->m_next);
2630				m0->m_next = NULL;
2631			}
2632			m0->m_pkthdr.len = m0->m_len;
2633		}
2634
2635		/*
2636		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2637		 * about padding/truncate rule for the original IP packet.
2638		 * From the discussion on IPv6imp in Feb 1999,
2639		 * the consensus was:
2640		 * - "attach as much as possible" is the goal
2641		 * - pad if not aligned (original size can be guessed by
2642		 *   original ip6 header)
2643		 * Following code adds the padding if it is simple enough,
2644		 * and truncates if not.
2645		 */
2646		if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2647			panic("assumption failed in %s:%d", __FILE__,
2648			    __LINE__);
2649
2650		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2651			/* not enough room, truncate */
2652			m0->m_pkthdr.len = m0->m_len = len -
2653			    sizeof(*nd_opt_rh);
2654		} else {
2655			/* enough room, pad or truncate */
2656			size_t extra;
2657
2658			extra = m0->m_pkthdr.len % 8;
2659			if (extra) {
2660				/* pad if easy enough, truncate if not */
2661				if (8 - extra <= M_TRAILINGSPACE(m0)) {
2662					/* pad */
2663					m0->m_len += (8 - extra);
2664					m0->m_pkthdr.len += (8 - extra);
2665				} else {
2666					/* truncate */
2667					m0->m_pkthdr.len -= extra;
2668					m0->m_len -= extra;
2669				}
2670			}
2671			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2672			m0->m_pkthdr.len = m0->m_len = len -
2673			    sizeof(*nd_opt_rh);
2674		}
2675
2676		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2677		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2678		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2679		nd_opt_rh->nd_opt_rh_len = len >> 3;
2680		p += sizeof(*nd_opt_rh);
2681		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2682
2683		/* connect m0 to m */
2684		m_tag_delete_chain(m0, NULL);
2685		m0->m_flags &= ~M_PKTHDR;
2686		m->m_next = m0;
2687		m->m_pkthdr.len = m->m_len + m0->m_len;
2688		m0 = NULL;
2689	}
2690noredhdropt:;
2691	if (m0) {
2692		m_freem(m0);
2693		m0 = NULL;
2694	}
2695
2696	/* XXX: clear embedded link IDs in the inner header */
2697	in6_clearscope(&sip6->ip6_src);
2698	in6_clearscope(&sip6->ip6_dst);
2699	in6_clearscope(&nd_rd->nd_rd_target);
2700	in6_clearscope(&nd_rd->nd_rd_dst);
2701
2702	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2703
2704	nd_rd->nd_rd_cksum = 0;
2705	nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2706	    sizeof(*ip6), ntohs(ip6->ip6_plen));
2707
2708	/* send the packet to outside... */
2709	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2710	if (outif) {
2711		icmp6_ifstat_inc(outif, ifs6_out_msg);
2712		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2713	}
2714	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2715
2716	return;
2717
2718fail:
2719	if (m)
2720		m_freem(m);
2721	if (m0)
2722		m_freem(m0);
2723}
2724
2725/*
2726 * ICMPv6 socket option processing.
2727 */
2728int
2729icmp6_ctloutput(so, sopt)
2730	struct socket *so;
2731	struct sockopt *sopt;
2732{
2733	int error = 0;
2734	int optlen;
2735	struct inpcb *inp = sotoinpcb(so);
2736	int level, op, optname;
2737
2738	if (sopt) {
2739		level = sopt->sopt_level;
2740		op = sopt->sopt_dir;
2741		optname = sopt->sopt_name;
2742		optlen = sopt->sopt_valsize;
2743	} else
2744		level = op = optname = optlen = 0;
2745
2746	if (level != IPPROTO_ICMPV6) {
2747		return EINVAL;
2748	}
2749
2750	switch (op) {
2751	case PRCO_SETOPT:
2752		switch (optname) {
2753		case ICMP6_FILTER:
2754		    {
2755			struct icmp6_filter *p;
2756
2757			if (optlen != sizeof(*p)) {
2758				error = EMSGSIZE;
2759				break;
2760			}
2761			if (inp->in6p_icmp6filt == NULL) {
2762				error = EINVAL;
2763				break;
2764			}
2765			error = sooptcopyin(sopt, inp->in6p_icmp6filt, optlen,
2766				optlen);
2767			break;
2768		    }
2769
2770		default:
2771			error = ENOPROTOOPT;
2772			break;
2773		}
2774		break;
2775
2776	case PRCO_GETOPT:
2777		switch (optname) {
2778		case ICMP6_FILTER:
2779		    {
2780			if (inp->in6p_icmp6filt == NULL) {
2781				error = EINVAL;
2782				break;
2783			}
2784			error = sooptcopyout(sopt, inp->in6p_icmp6filt,
2785				sizeof(struct icmp6_filter));
2786			break;
2787		    }
2788
2789		default:
2790			error = ENOPROTOOPT;
2791			break;
2792		}
2793		break;
2794	}
2795
2796	return (error);
2797}
2798
2799/*
2800 * Perform rate limit check.
2801 * Returns 0 if it is okay to send the icmp6 packet.
2802 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2803 * limitation.
2804 *
2805 * XXX per-destination/type check necessary?
2806 */
2807static int
2808icmp6_ratelimit(dst, type, code)
2809	const struct in6_addr *dst;	/* not used at this moment */
2810	const int type;			/* not used at this moment */
2811	const int code;			/* not used at this moment */
2812{
2813	int ret;
2814
2815	ret = 0;	/* okay to send */
2816
2817	/* PPS limit */
2818	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2819	    icmp6errppslim)) {
2820		/* The packet is subject to rate limit */
2821		ret++;
2822	}
2823
2824	return ret;
2825}
2826