icmp6.c revision 306683
1/*-
2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 *    notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 *    notice, this list of conditions and the following disclaimer in the
12 *    documentation and/or other materials provided with the distribution.
13 * 3. Neither the name of the project nor the names of its contributors
14 *    may be used to endorse or promote products derived from this software
15 *    without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 *	$KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $
30 */
31
32/*-
33 * Copyright (c) 1982, 1986, 1988, 1993
34 *	The Regents of the University of California.  All rights reserved.
35 *
36 * Redistribution and use in source and binary forms, with or without
37 * modification, are permitted provided that the following conditions
38 * are met:
39 * 1. Redistributions of source code must retain the above copyright
40 *    notice, this list of conditions and the following disclaimer.
41 * 2. Redistributions in binary form must reproduce the above copyright
42 *    notice, this list of conditions and the following disclaimer in the
43 *    documentation and/or other materials provided with the distribution.
44 * 4. Neither the name of the University nor the names of its contributors
45 *    may be used to endorse or promote products derived from this software
46 *    without specific prior written permission.
47 *
48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58 * SUCH DAMAGE.
59 *
60 *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
61 */
62
63#include <sys/cdefs.h>
64__FBSDID("$FreeBSD: stable/11/sys/netinet6/icmp6.c 306683 2016-10-04 18:07:33Z markj $");
65
66#define	MBUF_PRIVATE	/* XXXRW: Optimisation tries to avoid M_EXT mbufs */
67
68#include "opt_inet.h"
69#include "opt_inet6.h"
70
71#include <sys/param.h>
72#include <sys/domain.h>
73#include <sys/jail.h>
74#include <sys/kernel.h>
75#include <sys/lock.h>
76#include <sys/malloc.h>
77#include <sys/mbuf.h>
78#include <sys/proc.h>
79#include <sys/protosw.h>
80#include <sys/signalvar.h>
81#include <sys/socket.h>
82#include <sys/socketvar.h>
83#include <sys/sx.h>
84#include <sys/syslog.h>
85#include <sys/systm.h>
86#include <sys/time.h>
87
88#include <net/if.h>
89#include <net/if_var.h>
90#include <net/if_dl.h>
91#include <net/if_llatbl.h>
92#include <net/if_types.h>
93#include <net/route.h>
94#include <net/vnet.h>
95
96#include <netinet/in.h>
97#include <netinet/in_pcb.h>
98#include <netinet/in_var.h>
99#include <netinet/ip6.h>
100#include <netinet/icmp6.h>
101#include <netinet/tcp_var.h>
102
103#include <netinet6/in6_fib.h>
104#include <netinet6/in6_ifattach.h>
105#include <netinet6/in6_pcb.h>
106#include <netinet6/ip6protosw.h>
107#include <netinet6/ip6_var.h>
108#include <netinet6/scope6_var.h>
109#include <netinet6/mld6_var.h>
110#include <netinet6/nd6.h>
111#include <netinet6/send.h>
112
113extern struct domain inet6domain;
114
115VNET_PCPUSTAT_DEFINE(struct icmp6stat, icmp6stat);
116VNET_PCPUSTAT_SYSINIT(icmp6stat);
117
118#ifdef VIMAGE
119VNET_PCPUSTAT_SYSUNINIT(icmp6stat);
120#endif /* VIMAGE */
121
122VNET_DECLARE(struct inpcbinfo, ripcbinfo);
123VNET_DECLARE(struct inpcbhead, ripcb);
124VNET_DECLARE(int, icmp6errppslim);
125static VNET_DEFINE(int, icmp6errpps_count) = 0;
126static VNET_DEFINE(struct timeval, icmp6errppslim_last);
127VNET_DECLARE(int, icmp6_nodeinfo);
128
129#define	V_ripcbinfo			VNET(ripcbinfo)
130#define	V_ripcb				VNET(ripcb)
131#define	V_icmp6errppslim		VNET(icmp6errppslim)
132#define	V_icmp6errpps_count		VNET(icmp6errpps_count)
133#define	V_icmp6errppslim_last		VNET(icmp6errppslim_last)
134#define	V_icmp6_nodeinfo		VNET(icmp6_nodeinfo)
135
136static void icmp6_errcount(int, int);
137static int icmp6_rip6_input(struct mbuf **, int);
138static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
139static const char *icmp6_redirect_diag(struct in6_addr *,
140	struct in6_addr *, struct in6_addr *);
141static struct mbuf *ni6_input(struct mbuf *, int);
142static struct mbuf *ni6_nametodns(const char *, int, int);
143static int ni6_dnsmatch(const char *, int, const char *, int);
144static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
145			  struct ifnet **, struct in6_addr *);
146static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
147				struct ifnet *, int);
148static int icmp6_notify_error(struct mbuf **, int, int, int);
149
150/*
151 * Kernel module interface for updating icmp6stat.  The argument is an index
152 * into icmp6stat treated as an array of u_quad_t.  While this encodes the
153 * general layout of icmp6stat into the caller, it doesn't encode its
154 * location, so that future changes to add, for example, per-CPU stats
155 * support won't cause binary compatibility problems for kernel modules.
156 */
157void
158kmod_icmp6stat_inc(int statnum)
159{
160
161	counter_u64_add(VNET(icmp6stat)[statnum], 1);
162}
163
164static void
165icmp6_errcount(int type, int code)
166{
167	switch (type) {
168	case ICMP6_DST_UNREACH:
169		switch (code) {
170		case ICMP6_DST_UNREACH_NOROUTE:
171			ICMP6STAT_INC(icp6s_odst_unreach_noroute);
172			return;
173		case ICMP6_DST_UNREACH_ADMIN:
174			ICMP6STAT_INC(icp6s_odst_unreach_admin);
175			return;
176		case ICMP6_DST_UNREACH_BEYONDSCOPE:
177			ICMP6STAT_INC(icp6s_odst_unreach_beyondscope);
178			return;
179		case ICMP6_DST_UNREACH_ADDR:
180			ICMP6STAT_INC(icp6s_odst_unreach_addr);
181			return;
182		case ICMP6_DST_UNREACH_NOPORT:
183			ICMP6STAT_INC(icp6s_odst_unreach_noport);
184			return;
185		}
186		break;
187	case ICMP6_PACKET_TOO_BIG:
188		ICMP6STAT_INC(icp6s_opacket_too_big);
189		return;
190	case ICMP6_TIME_EXCEEDED:
191		switch (code) {
192		case ICMP6_TIME_EXCEED_TRANSIT:
193			ICMP6STAT_INC(icp6s_otime_exceed_transit);
194			return;
195		case ICMP6_TIME_EXCEED_REASSEMBLY:
196			ICMP6STAT_INC(icp6s_otime_exceed_reassembly);
197			return;
198		}
199		break;
200	case ICMP6_PARAM_PROB:
201		switch (code) {
202		case ICMP6_PARAMPROB_HEADER:
203			ICMP6STAT_INC(icp6s_oparamprob_header);
204			return;
205		case ICMP6_PARAMPROB_NEXTHEADER:
206			ICMP6STAT_INC(icp6s_oparamprob_nextheader);
207			return;
208		case ICMP6_PARAMPROB_OPTION:
209			ICMP6STAT_INC(icp6s_oparamprob_option);
210			return;
211		}
212		break;
213	case ND_REDIRECT:
214		ICMP6STAT_INC(icp6s_oredirect);
215		return;
216	}
217	ICMP6STAT_INC(icp6s_ounknown);
218}
219
220/*
221 * A wrapper function for icmp6_error() necessary when the erroneous packet
222 * may not contain enough scope zone information.
223 */
224void
225icmp6_error2(struct mbuf *m, int type, int code, int param,
226    struct ifnet *ifp)
227{
228	struct ip6_hdr *ip6;
229
230	if (ifp == NULL)
231		return;
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
243	ip6 = mtod(m, struct ip6_hdr *);
244
245	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
246		return;
247	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
248		return;
249
250	icmp6_error(m, type, code, param);
251}
252
253/*
254 * Generate an error packet of type error in response to bad IP6 packet.
255 */
256void
257icmp6_error(struct mbuf *m, int type, int code, int param)
258{
259	struct ip6_hdr *oip6, *nip6;
260	struct icmp6_hdr *icmp6;
261	u_int preplen;
262	int off;
263	int nxt;
264
265	ICMP6STAT_INC(icp6s_error);
266
267	/* count per-type-code statistics */
268	icmp6_errcount(type, code);
269
270#ifdef M_DECRYPTED	/*not openbsd*/
271	if (m->m_flags & M_DECRYPTED) {
272		ICMP6STAT_INC(icp6s_canterror);
273		goto freeit;
274	}
275#endif
276
277#ifndef PULLDOWN_TEST
278	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
279#else
280	if (m->m_len < sizeof(struct ip6_hdr)) {
281		m = m_pullup(m, sizeof(struct ip6_hdr));
282		if (m == NULL)
283			return;
284	}
285#endif
286	oip6 = mtod(m, struct ip6_hdr *);
287
288	/*
289	 * If the destination address of the erroneous packet is a multicast
290	 * address, or the packet was sent using link-layer multicast,
291	 * we should basically suppress sending an error (RFC 2463, Section
292	 * 2.4).
293	 * We have two exceptions (the item e.2 in that section):
294	 * - the Packet Too Big message can be sent for path MTU discovery.
295	 * - the Parameter Problem Message that can be allowed an icmp6 error
296	 *   in the option type field.  This check has been done in
297	 *   ip6_unknown_opt(), so we can just check the type and code.
298	 */
299	if ((m->m_flags & (M_BCAST|M_MCAST) ||
300	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
301	    (type != ICMP6_PACKET_TOO_BIG &&
302	     (type != ICMP6_PARAM_PROB ||
303	      code != ICMP6_PARAMPROB_OPTION)))
304		goto freeit;
305
306	/*
307	 * RFC 2463, 2.4 (e.5): source address check.
308	 * XXX: the case of anycast source?
309	 */
310	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
311	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
312		goto freeit;
313
314	/*
315	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
316	 * don't do it.
317	 */
318	nxt = -1;
319	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
320	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
321		struct icmp6_hdr *icp;
322
323#ifndef PULLDOWN_TEST
324		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
325		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
326#else
327		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
328			sizeof(*icp));
329		if (icp == NULL) {
330			ICMP6STAT_INC(icp6s_tooshort);
331			return;
332		}
333#endif
334		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
335		    icp->icmp6_type == ND_REDIRECT) {
336			/*
337			 * ICMPv6 error
338			 * Special case: for redirect (which is
339			 * informational) we must not send icmp6 error.
340			 */
341			ICMP6STAT_INC(icp6s_canterror);
342			goto freeit;
343		} else {
344			/* ICMPv6 informational - send the error */
345		}
346	} else {
347		/* non-ICMPv6 - send the error */
348	}
349
350	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
351
352	/* Finally, do rate limitation check. */
353	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
354		ICMP6STAT_INC(icp6s_toofreq);
355		goto freeit;
356	}
357
358	/*
359	 * OK, ICMP6 can be generated.
360	 */
361
362	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
363		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
364
365	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
366	M_PREPEND(m, preplen, M_NOWAIT);	/* FIB is also copied over. */
367	if (m == NULL) {
368		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
369		return;
370	}
371
372	nip6 = mtod(m, struct ip6_hdr *);
373	nip6->ip6_src  = oip6->ip6_src;
374	nip6->ip6_dst  = oip6->ip6_dst;
375
376	in6_clearscope(&oip6->ip6_src);
377	in6_clearscope(&oip6->ip6_dst);
378
379	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
380	icmp6->icmp6_type = type;
381	icmp6->icmp6_code = code;
382	icmp6->icmp6_pptr = htonl((u_int32_t)param);
383
384	/*
385	 * icmp6_reflect() is designed to be in the input path.
386	 * icmp6_error() can be called from both input and output path,
387	 * and if we are in output path rcvif could contain bogus value.
388	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
389	 * information in ip header (nip6).
390	 */
391	m->m_pkthdr.rcvif = NULL;
392
393	ICMP6STAT_INC(icp6s_outhist[type]);
394	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
395
396	return;
397
398  freeit:
399	/*
400	 * If we can't tell whether or not we can generate ICMP6, free it.
401	 */
402	m_freem(m);
403}
404
405/*
406 * Process a received ICMP6 message.
407 */
408int
409icmp6_input(struct mbuf **mp, int *offp, int proto)
410{
411	struct mbuf *m = *mp, *n;
412	struct ifnet *ifp;
413	struct ip6_hdr *ip6, *nip6;
414	struct icmp6_hdr *icmp6, *nicmp6;
415	int off = *offp;
416	int icmp6len = m->m_pkthdr.len - *offp;
417	int code, sum, noff;
418	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
419	int ip6len, error;
420
421	ifp = m->m_pkthdr.rcvif;
422
423#ifndef PULLDOWN_TEST
424	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
425	/* m might change if M_LOOP.  So, call mtod after this */
426#endif
427
428	/*
429	 * Locate icmp6 structure in mbuf, and check
430	 * that not corrupted and of at least minimum length
431	 */
432
433	ip6 = mtod(m, struct ip6_hdr *);
434	ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
435	if (icmp6len < sizeof(struct icmp6_hdr)) {
436		ICMP6STAT_INC(icp6s_tooshort);
437		goto freeit;
438	}
439
440	/*
441	 * Check multicast group membership.
442	 * Note: SSM filters are not applied for ICMPv6 traffic.
443	 */
444	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
445		struct in6_multi	*inm;
446
447		inm = in6m_lookup(ifp, &ip6->ip6_dst);
448		if (inm == NULL) {
449			IP6STAT_INC(ip6s_notmember);
450			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
451			goto freeit;
452		}
453	}
454
455	/*
456	 * calculate the checksum
457	 */
458#ifndef PULLDOWN_TEST
459	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
460#else
461	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
462	if (icmp6 == NULL) {
463		ICMP6STAT_INC(icp6s_tooshort);
464		return IPPROTO_DONE;
465	}
466#endif
467	code = icmp6->icmp6_code;
468
469	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
470		nd6log((LOG_ERR,
471		    "ICMP6 checksum error(%d|%x) %s\n",
472		    icmp6->icmp6_type, sum,
473		    ip6_sprintf(ip6bufs, &ip6->ip6_src)));
474		ICMP6STAT_INC(icp6s_checksum);
475		goto freeit;
476	}
477
478	ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
479	icmp6_ifstat_inc(ifp, ifs6_in_msg);
480	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
481		icmp6_ifstat_inc(ifp, ifs6_in_error);
482
483	switch (icmp6->icmp6_type) {
484	case ICMP6_DST_UNREACH:
485		icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
486		switch (code) {
487		case ICMP6_DST_UNREACH_NOROUTE:
488		case ICMP6_DST_UNREACH_ADDR:	/* PRC_HOSTDEAD is a DOS */
489			code = PRC_UNREACH_NET;
490			break;
491		case ICMP6_DST_UNREACH_ADMIN:
492			icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
493			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
494			break;
495		case ICMP6_DST_UNREACH_BEYONDSCOPE:
496			/* I mean "source address was incorrect." */
497			code = PRC_PARAMPROB;
498			break;
499		case ICMP6_DST_UNREACH_NOPORT:
500			code = PRC_UNREACH_PORT;
501			break;
502		default:
503			goto badcode;
504		}
505		goto deliver;
506		break;
507
508	case ICMP6_PACKET_TOO_BIG:
509		icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
510
511		/* validation is made in icmp6_mtudisc_update */
512
513		code = PRC_MSGSIZE;
514
515		/*
516		 * Updating the path MTU will be done after examining
517		 * intermediate extension headers.
518		 */
519		goto deliver;
520		break;
521
522	case ICMP6_TIME_EXCEEDED:
523		icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
524		switch (code) {
525		case ICMP6_TIME_EXCEED_TRANSIT:
526			code = PRC_TIMXCEED_INTRANS;
527			break;
528		case ICMP6_TIME_EXCEED_REASSEMBLY:
529			code = PRC_TIMXCEED_REASS;
530			break;
531		default:
532			goto badcode;
533		}
534		goto deliver;
535		break;
536
537	case ICMP6_PARAM_PROB:
538		icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
539		switch (code) {
540		case ICMP6_PARAMPROB_NEXTHEADER:
541			code = PRC_UNREACH_PROTOCOL;
542			break;
543		case ICMP6_PARAMPROB_HEADER:
544		case ICMP6_PARAMPROB_OPTION:
545			code = PRC_PARAMPROB;
546			break;
547		default:
548			goto badcode;
549		}
550		goto deliver;
551		break;
552
553	case ICMP6_ECHO_REQUEST:
554		icmp6_ifstat_inc(ifp, ifs6_in_echo);
555		if (code != 0)
556			goto badcode;
557		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
558			/* Give up remote */
559			break;
560		}
561		if (!M_WRITABLE(n)
562		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
563			struct mbuf *n0 = n;
564			int n0len;
565
566			CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) <= MHLEN);
567			n = m_gethdr(M_NOWAIT, n0->m_type);
568			if (n == NULL) {
569				/* Give up remote */
570				m_freem(n0);
571				break;
572			}
573
574			m_move_pkthdr(n, n0);	/* FIB copied. */
575			n0len = n0->m_pkthdr.len;	/* save for use below */
576			/*
577			 * Copy IPv6 and ICMPv6 only.
578			 */
579			nip6 = mtod(n, struct ip6_hdr *);
580			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
581			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
582			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
583			noff = sizeof(struct ip6_hdr);
584			/* new mbuf contains only ipv6+icmpv6 headers */
585			n->m_len = noff + sizeof(struct icmp6_hdr);
586			/*
587			 * Adjust mbuf.  ip6_plen will be adjusted in
588			 * ip6_output().
589			 */
590			m_adj(n0, off + sizeof(struct icmp6_hdr));
591			/* recalculate complete packet size */
592			n->m_pkthdr.len = n0len + (noff - off);
593			n->m_next = n0;
594		} else {
595			nip6 = mtod(n, struct ip6_hdr *);
596			IP6_EXTHDR_GET(nicmp6, struct icmp6_hdr *, n, off,
597			    sizeof(*nicmp6));
598			noff = off;
599		}
600		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
601		nicmp6->icmp6_code = 0;
602		if (n) {
603			ICMP6STAT_INC(icp6s_reflect);
604			ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
605			icmp6_reflect(n, noff);
606		}
607		break;
608
609	case ICMP6_ECHO_REPLY:
610		icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
611		if (code != 0)
612			goto badcode;
613		break;
614
615	case MLD_LISTENER_QUERY:
616	case MLD_LISTENER_REPORT:
617	case MLD_LISTENER_DONE:
618	case MLDV2_LISTENER_REPORT:
619		/*
620		 * Drop MLD traffic which is not link-local, has a hop limit
621		 * of greater than 1 hop, or which does not have the
622		 * IPv6 HBH Router Alert option.
623		 * As IPv6 HBH options are stripped in ip6_input() we must
624		 * check an mbuf header flag.
625		 * XXX Should we also sanity check that these messages
626		 * were directed to a link-local multicast prefix?
627		 */
628		if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
629			goto freeit;
630		if (mld_input(m, off, icmp6len) != 0)
631			return (IPPROTO_DONE);
632		/* m stays. */
633		break;
634
635	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
636	    {
637		enum { WRU, FQDN } mode;
638
639		if (!V_icmp6_nodeinfo)
640			break;
641
642		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
643			mode = WRU;
644		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
645			mode = FQDN;
646		else
647			goto badlen;
648
649		if (mode == FQDN) {
650#ifndef PULLDOWN_TEST
651			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
652			    IPPROTO_DONE);
653#endif
654			n = m_copy(m, 0, M_COPYALL);
655			if (n)
656				n = ni6_input(n, off);
657			/* XXX meaningless if n == NULL */
658			noff = sizeof(struct ip6_hdr);
659		} else {
660			struct prison *pr;
661			u_char *p;
662			int maxhlen, hlen;
663
664			/*
665			 * XXX: this combination of flags is pointless,
666			 * but should we keep this for compatibility?
667			 */
668			if ((V_icmp6_nodeinfo & (ICMP6_NODEINFO_FQDNOK |
669			    ICMP6_NODEINFO_TMPADDROK)) !=
670			    (ICMP6_NODEINFO_FQDNOK | ICMP6_NODEINFO_TMPADDROK))
671				break;
672
673			if (code != 0)
674				goto badcode;
675
676			CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) + 4 <= MHLEN);
677			n = m_gethdr(M_NOWAIT, m->m_type);
678			if (n == NULL) {
679				/* Give up remote */
680				break;
681			}
682			if (!m_dup_pkthdr(n, m, M_NOWAIT)) {
683				/*
684				 * Previous code did a blind M_COPY_PKTHDR
685				 * and said "just for rcvif".  If true, then
686				 * we could tolerate the dup failing (due to
687				 * the deep copy of the tag chain).  For now
688				 * be conservative and just fail.
689				 */
690				m_free(n);
691				n = NULL;
692			}
693			maxhlen = M_TRAILINGSPACE(n) -
694			    (sizeof(*nip6) + sizeof(*nicmp6) + 4);
695			pr = curthread->td_ucred->cr_prison;
696			mtx_lock(&pr->pr_mtx);
697			hlen = strlen(pr->pr_hostname);
698			if (maxhlen > hlen)
699				maxhlen = hlen;
700			/*
701			 * Copy IPv6 and ICMPv6 only.
702			 */
703			nip6 = mtod(n, struct ip6_hdr *);
704			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
705			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
706			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
707			p = (u_char *)(nicmp6 + 1);
708			bzero(p, 4);
709			/* meaningless TTL */
710			bcopy(pr->pr_hostname, p + 4, maxhlen);
711			mtx_unlock(&pr->pr_mtx);
712			noff = sizeof(struct ip6_hdr);
713			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
714				sizeof(struct icmp6_hdr) + 4 + maxhlen;
715			nicmp6->icmp6_type = ICMP6_WRUREPLY;
716			nicmp6->icmp6_code = 0;
717		}
718		if (n) {
719			ICMP6STAT_INC(icp6s_reflect);
720			ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
721			icmp6_reflect(n, noff);
722		}
723		break;
724	    }
725
726	case ICMP6_WRUREPLY:
727		if (code != 0)
728			goto badcode;
729		break;
730
731	case ND_ROUTER_SOLICIT:
732		icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
733		if (code != 0)
734			goto badcode;
735		if (icmp6len < sizeof(struct nd_router_solicit))
736			goto badlen;
737		if (send_sendso_input_hook != NULL) {
738			IP6_EXTHDR_CHECK(m, off, icmp6len, IPPROTO_DONE);
739			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
740			if (error == 0) {
741				m = NULL;
742				goto freeit;
743			}
744		}
745		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
746		nd6_rs_input(m, off, icmp6len);
747		m = n;
748		if (m == NULL)
749			goto freeit;
750		break;
751
752	case ND_ROUTER_ADVERT:
753		icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
754		if (code != 0)
755			goto badcode;
756		if (icmp6len < sizeof(struct nd_router_advert))
757			goto badlen;
758		if (send_sendso_input_hook != NULL) {
759			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
760			if (error == 0) {
761				m = NULL;
762				goto freeit;
763			}
764		}
765		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
766		nd6_ra_input(m, off, icmp6len);
767		m = n;
768		if (m == NULL)
769			goto freeit;
770		break;
771
772	case ND_NEIGHBOR_SOLICIT:
773		icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
774		if (code != 0)
775			goto badcode;
776		if (icmp6len < sizeof(struct nd_neighbor_solicit))
777			goto badlen;
778		if (send_sendso_input_hook != NULL) {
779			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
780			if (error == 0) {
781				m = NULL;
782				goto freeit;
783			}
784		}
785		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
786		nd6_ns_input(m, off, icmp6len);
787		m = n;
788		if (m == NULL)
789			goto freeit;
790		break;
791
792	case ND_NEIGHBOR_ADVERT:
793		icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
794		if (code != 0)
795			goto badcode;
796		if (icmp6len < sizeof(struct nd_neighbor_advert))
797			goto badlen;
798		if (send_sendso_input_hook != NULL) {
799			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
800			if (error == 0) {
801				m = NULL;
802				goto freeit;
803			}
804		}
805		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
806		nd6_na_input(m, off, icmp6len);
807		m = n;
808		if (m == NULL)
809			goto freeit;
810		break;
811
812	case ND_REDIRECT:
813		icmp6_ifstat_inc(ifp, ifs6_in_redirect);
814		if (code != 0)
815			goto badcode;
816		if (icmp6len < sizeof(struct nd_redirect))
817			goto badlen;
818		if (send_sendso_input_hook != NULL) {
819			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
820			if (error == 0) {
821				m = NULL;
822				goto freeit;
823			}
824		}
825		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
826		icmp6_redirect_input(m, off);
827		m = n;
828		if (m == NULL)
829			goto freeit;
830		break;
831
832	case ICMP6_ROUTER_RENUMBERING:
833		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
834		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
835			goto badcode;
836		if (icmp6len < sizeof(struct icmp6_router_renum))
837			goto badlen;
838		break;
839
840	default:
841		nd6log((LOG_DEBUG,
842		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
843		    icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
844		    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
845		    ifp ? ifp->if_index : 0));
846		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
847			/* ICMPv6 error: MUST deliver it by spec... */
848			code = PRC_NCMDS;
849			/* deliver */
850		} else {
851			/* ICMPv6 informational: MUST not deliver */
852			break;
853		}
854	deliver:
855		if (icmp6_notify_error(&m, off, icmp6len, code) != 0) {
856			/* In this case, m should've been freed. */
857			return (IPPROTO_DONE);
858		}
859		break;
860
861	badcode:
862		ICMP6STAT_INC(icp6s_badcode);
863		break;
864
865	badlen:
866		ICMP6STAT_INC(icp6s_badlen);
867		break;
868	}
869
870	/* deliver the packet to appropriate sockets */
871	icmp6_rip6_input(&m, *offp);
872
873	return IPPROTO_DONE;
874
875 freeit:
876	m_freem(m);
877	return IPPROTO_DONE;
878}
879
880static int
881icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
882{
883	struct mbuf *m = *mp;
884	struct icmp6_hdr *icmp6;
885	struct ip6_hdr *eip6;
886	u_int32_t notifymtu;
887	struct sockaddr_in6 icmp6src, icmp6dst;
888
889	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
890		ICMP6STAT_INC(icp6s_tooshort);
891		goto freeit;
892	}
893#ifndef PULLDOWN_TEST
894	IP6_EXTHDR_CHECK(m, off,
895	    sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr), -1);
896	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
897#else
898	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
899	    sizeof(*icmp6) + sizeof(struct ip6_hdr));
900	if (icmp6 == NULL) {
901		ICMP6STAT_INC(icp6s_tooshort);
902		return (-1);
903	}
904#endif
905	eip6 = (struct ip6_hdr *)(icmp6 + 1);
906
907	/* Detect the upper level protocol */
908	{
909		void (*ctlfunc)(int, struct sockaddr *, void *);
910		u_int8_t nxt = eip6->ip6_nxt;
911		int eoff = off + sizeof(struct icmp6_hdr) +
912		    sizeof(struct ip6_hdr);
913		struct ip6ctlparam ip6cp;
914		struct in6_addr *finaldst = NULL;
915		int icmp6type = icmp6->icmp6_type;
916		struct ip6_frag *fh;
917		struct ip6_rthdr *rth;
918		struct ip6_rthdr0 *rth0;
919		int rthlen;
920
921		while (1) { /* XXX: should avoid infinite loop explicitly? */
922			struct ip6_ext *eh;
923
924			switch (nxt) {
925			case IPPROTO_HOPOPTS:
926			case IPPROTO_DSTOPTS:
927			case IPPROTO_AH:
928#ifndef PULLDOWN_TEST
929				IP6_EXTHDR_CHECK(m, 0,
930				    eoff + sizeof(struct ip6_ext), -1);
931				eh = (struct ip6_ext *)(mtod(m, caddr_t) + eoff);
932#else
933				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
934				    eoff, sizeof(*eh));
935				if (eh == NULL) {
936					ICMP6STAT_INC(icp6s_tooshort);
937					return (-1);
938				}
939#endif
940
941				if (nxt == IPPROTO_AH)
942					eoff += (eh->ip6e_len + 2) << 2;
943				else
944					eoff += (eh->ip6e_len + 1) << 3;
945				nxt = eh->ip6e_nxt;
946				break;
947			case IPPROTO_ROUTING:
948				/*
949				 * When the erroneous packet contains a
950				 * routing header, we should examine the
951				 * header to determine the final destination.
952				 * Otherwise, we can't properly update
953				 * information that depends on the final
954				 * destination (e.g. path MTU).
955				 */
956#ifndef PULLDOWN_TEST
957				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth), -1);
958				rth = (struct ip6_rthdr *)
959				    (mtod(m, caddr_t) + eoff);
960#else
961				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
962				    eoff, sizeof(*rth));
963				if (rth == NULL) {
964					ICMP6STAT_INC(icp6s_tooshort);
965					return (-1);
966				}
967#endif
968				rthlen = (rth->ip6r_len + 1) << 3;
969				/*
970				 * XXX: currently there is no
971				 * officially defined type other
972				 * than type-0.
973				 * Note that if the segment left field
974				 * is 0, all intermediate hops must
975				 * have been passed.
976				 */
977				if (rth->ip6r_segleft &&
978				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
979					int hops;
980
981#ifndef PULLDOWN_TEST
982					IP6_EXTHDR_CHECK(m, 0, eoff + rthlen, -1);
983					rth0 = (struct ip6_rthdr0 *)
984					    (mtod(m, caddr_t) + eoff);
985#else
986					IP6_EXTHDR_GET(rth0,
987					    struct ip6_rthdr0 *, m,
988					    eoff, rthlen);
989					if (rth0 == NULL) {
990						ICMP6STAT_INC(icp6s_tooshort);
991						return (-1);
992					}
993#endif
994					/* just ignore a bogus header */
995					if ((rth0->ip6r0_len % 2) == 0 &&
996					    (hops = rth0->ip6r0_len/2))
997						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
998				}
999				eoff += rthlen;
1000				nxt = rth->ip6r_nxt;
1001				break;
1002			case IPPROTO_FRAGMENT:
1003#ifndef PULLDOWN_TEST
1004				IP6_EXTHDR_CHECK(m, 0, eoff +
1005				    sizeof(struct ip6_frag), -1);
1006				fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1007				    eoff);
1008#else
1009				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1010				    eoff, sizeof(*fh));
1011				if (fh == NULL) {
1012					ICMP6STAT_INC(icp6s_tooshort);
1013					return (-1);
1014				}
1015#endif
1016				/*
1017				 * Data after a fragment header is meaningless
1018				 * unless it is the first fragment, but
1019				 * we'll go to the notify label for path MTU
1020				 * discovery.
1021				 */
1022				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1023					goto notify;
1024
1025				eoff += sizeof(struct ip6_frag);
1026				nxt = fh->ip6f_nxt;
1027				break;
1028			default:
1029				/*
1030				 * This case includes ESP and the No Next
1031				 * Header.  In such cases going to the notify
1032				 * label does not have any meaning
1033				 * (i.e. ctlfunc will be NULL), but we go
1034				 * anyway since we might have to update
1035				 * path MTU information.
1036				 */
1037				goto notify;
1038			}
1039		}
1040	  notify:
1041#ifndef PULLDOWN_TEST
1042		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1043#else
1044		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1045		    sizeof(*icmp6) + sizeof(struct ip6_hdr));
1046		if (icmp6 == NULL) {
1047			ICMP6STAT_INC(icp6s_tooshort);
1048			return (-1);
1049		}
1050#endif
1051
1052		/*
1053		 * retrieve parameters from the inner IPv6 header, and convert
1054		 * them into sockaddr structures.
1055		 * XXX: there is no guarantee that the source or destination
1056		 * addresses of the inner packet are in the same scope as
1057		 * the addresses of the icmp packet.  But there is no other
1058		 * way to determine the zone.
1059		 */
1060		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1061
1062		bzero(&icmp6dst, sizeof(icmp6dst));
1063		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1064		icmp6dst.sin6_family = AF_INET6;
1065		if (finaldst == NULL)
1066			icmp6dst.sin6_addr = eip6->ip6_dst;
1067		else
1068			icmp6dst.sin6_addr = *finaldst;
1069		if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1070			goto freeit;
1071		bzero(&icmp6src, sizeof(icmp6src));
1072		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1073		icmp6src.sin6_family = AF_INET6;
1074		icmp6src.sin6_addr = eip6->ip6_src;
1075		if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1076			goto freeit;
1077		icmp6src.sin6_flowinfo =
1078		    (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1079
1080		if (finaldst == NULL)
1081			finaldst = &eip6->ip6_dst;
1082		ip6cp.ip6c_m = m;
1083		ip6cp.ip6c_icmp6 = icmp6;
1084		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1085		ip6cp.ip6c_off = eoff;
1086		ip6cp.ip6c_finaldst = finaldst;
1087		ip6cp.ip6c_src = &icmp6src;
1088		ip6cp.ip6c_nxt = nxt;
1089
1090		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1091			notifymtu = ntohl(icmp6->icmp6_mtu);
1092			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1093			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1094		}
1095
1096		ctlfunc = (void (*)(int, struct sockaddr *, void *))
1097		    (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1098		if (ctlfunc) {
1099			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1100			    &ip6cp);
1101		}
1102	}
1103	*mp = m;
1104	return (0);
1105
1106  freeit:
1107	m_freem(m);
1108	return (-1);
1109}
1110
1111void
1112icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1113{
1114	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1115	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1116	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1117	u_int mtu = ntohl(icmp6->icmp6_mtu);
1118	struct in_conninfo inc;
1119
1120#if 0
1121	/*
1122	 * RFC2460 section 5, last paragraph.
1123	 * even though minimum link MTU for IPv6 is IPV6_MMTU,
1124	 * we may see ICMPv6 too big with mtu < IPV6_MMTU
1125	 * due to packet translator in the middle.
1126	 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1127	 * special handling.
1128	 */
1129	if (mtu < IPV6_MMTU)
1130		return;
1131#endif
1132
1133	/*
1134	 * we reject ICMPv6 too big with abnormally small value.
1135	 * XXX what is the good definition of "abnormally small"?
1136	 */
1137	if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1138		return;
1139
1140	if (!validated)
1141		return;
1142
1143	/*
1144	 * In case the suggested mtu is less than IPV6_MMTU, we
1145	 * only need to remember that it was for above mentioned
1146	 * "alwaysfrag" case.
1147	 * Try to be as close to the spec as possible.
1148	 */
1149	if (mtu < IPV6_MMTU)
1150		mtu = IPV6_MMTU - 8;
1151
1152	bzero(&inc, sizeof(inc));
1153	inc.inc_fibnum = M_GETFIB(m);
1154	inc.inc_flags |= INC_ISIPV6;
1155	inc.inc6_faddr = *dst;
1156	if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1157		return;
1158
1159	if (mtu < tcp_maxmtu6(&inc, NULL)) {
1160		tcp_hc_updatemtu(&inc, mtu);
1161		ICMP6STAT_INC(icp6s_pmtuchg);
1162	}
1163}
1164
1165/*
1166 * Process a Node Information Query packet, based on
1167 * draft-ietf-ipngwg-icmp-name-lookups-07.
1168 *
1169 * Spec incompatibilities:
1170 * - IPv6 Subject address handling
1171 * - IPv4 Subject address handling support missing
1172 * - Proxy reply (answer even if it's not for me)
1173 * - joins NI group address at in6_ifattach() time only, does not cope
1174 *   with hostname changes by sethostname(3)
1175 */
1176static struct mbuf *
1177ni6_input(struct mbuf *m, int off)
1178{
1179	struct icmp6_nodeinfo *ni6, *nni6;
1180	struct mbuf *n = NULL;
1181	struct prison *pr;
1182	u_int16_t qtype;
1183	int subjlen;
1184	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1185	struct ni_reply_fqdn *fqdn;
1186	int addrs;		/* for NI_QTYPE_NODEADDR */
1187	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1188	struct in6_addr in6_subj; /* subject address */
1189	struct ip6_hdr *ip6;
1190	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1191	char *subj = NULL;
1192	struct in6_ifaddr *ia6 = NULL;
1193
1194	ip6 = mtod(m, struct ip6_hdr *);
1195#ifndef PULLDOWN_TEST
1196	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1197#else
1198	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1199	if (ni6 == NULL) {
1200		/* m is already reclaimed */
1201		return (NULL);
1202	}
1203#endif
1204
1205	/*
1206	 * Validate IPv6 source address.
1207	 * The default configuration MUST be to refuse answering queries from
1208	 * global-scope addresses according to RFC4602.
1209	 * Notes:
1210	 *  - it's not very clear what "refuse" means; this implementation
1211	 *    simply drops it.
1212	 *  - it's not very easy to identify global-scope (unicast) addresses
1213	 *    since there are many prefixes for them.  It should be safer
1214	 *    and in practice sufficient to check "all" but loopback and
1215	 *    link-local (note that site-local unicast was deprecated and
1216	 *    ULA is defined as global scope-wise)
1217	 */
1218	if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1219	    !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1220	    !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1221		goto bad;
1222
1223	/*
1224	 * Validate IPv6 destination address.
1225	 *
1226	 * The Responder must discard the Query without further processing
1227	 * unless it is one of the Responder's unicast or anycast addresses, or
1228	 * a link-local scope multicast address which the Responder has joined.
1229	 * [RFC4602, Section 5.]
1230	 */
1231	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1232		if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1233			goto bad;
1234		/* else it's a link-local multicast, fine */
1235	} else {		/* unicast or anycast */
1236		ia6 = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */);
1237		if (ia6 == NULL)
1238			goto bad; /* XXX impossible */
1239
1240		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1241		    !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1242			ifa_free(&ia6->ia_ifa);
1243			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1244				"a temporary address in %s:%d",
1245			       __FILE__, __LINE__));
1246			goto bad;
1247		}
1248		ifa_free(&ia6->ia_ifa);
1249	}
1250
1251	/* validate query Subject field. */
1252	qtype = ntohs(ni6->ni_qtype);
1253	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1254	switch (qtype) {
1255	case NI_QTYPE_NOOP:
1256	case NI_QTYPE_SUPTYPES:
1257		/* 07 draft */
1258		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1259			break;
1260		/* FALLTHROUGH */
1261	case NI_QTYPE_FQDN:
1262	case NI_QTYPE_NODEADDR:
1263	case NI_QTYPE_IPV4ADDR:
1264		switch (ni6->ni_code) {
1265		case ICMP6_NI_SUBJ_IPV6:
1266#if ICMP6_NI_SUBJ_IPV6 != 0
1267		case 0:
1268#endif
1269			/*
1270			 * backward compatibility - try to accept 03 draft
1271			 * format, where no Subject is present.
1272			 */
1273			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1274			    subjlen == 0) {
1275				oldfqdn++;
1276				break;
1277			}
1278#if ICMP6_NI_SUBJ_IPV6 != 0
1279			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1280				goto bad;
1281#endif
1282
1283			if (subjlen != sizeof(struct in6_addr))
1284				goto bad;
1285
1286			/*
1287			 * Validate Subject address.
1288			 *
1289			 * Not sure what exactly "address belongs to the node"
1290			 * means in the spec, is it just unicast, or what?
1291			 *
1292			 * At this moment we consider Subject address as
1293			 * "belong to the node" if the Subject address equals
1294			 * to the IPv6 destination address; validation for
1295			 * IPv6 destination address should have done enough
1296			 * check for us.
1297			 *
1298			 * We do not do proxy at this moment.
1299			 */
1300			/* m_pulldown instead of copy? */
1301			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1302			    subjlen, (caddr_t)&in6_subj);
1303			if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1304				goto bad;
1305
1306			subj = (char *)&in6_subj;
1307			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1308				break;
1309
1310			/*
1311			 * XXX if we are to allow other cases, we should really
1312			 * be careful about scope here.
1313			 * basically, we should disallow queries toward IPv6
1314			 * destination X with subject Y,
1315			 * if scope(X) > scope(Y).
1316			 * if we allow scope(X) > scope(Y), it will result in
1317			 * information leakage across scope boundary.
1318			 */
1319			goto bad;
1320
1321		case ICMP6_NI_SUBJ_FQDN:
1322			/*
1323			 * Validate Subject name with gethostname(3).
1324			 *
1325			 * The behavior may need some debate, since:
1326			 * - we are not sure if the node has FQDN as
1327			 *   hostname (returned by gethostname(3)).
1328			 * - the code does wildcard match for truncated names.
1329			 *   however, we are not sure if we want to perform
1330			 *   wildcard match, if gethostname(3) side has
1331			 *   truncated hostname.
1332			 */
1333			pr = curthread->td_ucred->cr_prison;
1334			mtx_lock(&pr->pr_mtx);
1335			n = ni6_nametodns(pr->pr_hostname,
1336			    strlen(pr->pr_hostname), 0);
1337			mtx_unlock(&pr->pr_mtx);
1338			if (!n || n->m_next || n->m_len == 0)
1339				goto bad;
1340			IP6_EXTHDR_GET(subj, char *, m,
1341			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1342			if (subj == NULL)
1343				goto bad;
1344			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1345			    n->m_len)) {
1346				goto bad;
1347			}
1348			m_freem(n);
1349			n = NULL;
1350			break;
1351
1352		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1353		default:
1354			goto bad;
1355		}
1356		break;
1357	}
1358
1359	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1360	switch (qtype) {
1361	case NI_QTYPE_FQDN:
1362		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1363			goto bad;
1364		break;
1365	case NI_QTYPE_NODEADDR:
1366	case NI_QTYPE_IPV4ADDR:
1367		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1368			goto bad;
1369		break;
1370	}
1371
1372	/* guess reply length */
1373	switch (qtype) {
1374	case NI_QTYPE_NOOP:
1375		break;		/* no reply data */
1376	case NI_QTYPE_SUPTYPES:
1377		replylen += sizeof(u_int32_t);
1378		break;
1379	case NI_QTYPE_FQDN:
1380		/* XXX will append an mbuf */
1381		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1382		break;
1383	case NI_QTYPE_NODEADDR:
1384		addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1385		if ((replylen += addrs * (sizeof(struct in6_addr) +
1386		    sizeof(u_int32_t))) > MCLBYTES)
1387			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1388		break;
1389	case NI_QTYPE_IPV4ADDR:
1390		/* unsupported - should respond with unknown Qtype? */
1391		break;
1392	default:
1393		/*
1394		 * XXX: We must return a reply with the ICMP6 code
1395		 * `unknown Qtype' in this case.  However we regard the case
1396		 * as an FQDN query for backward compatibility.
1397		 * Older versions set a random value to this field,
1398		 * so it rarely varies in the defined qtypes.
1399		 * But the mechanism is not reliable...
1400		 * maybe we should obsolete older versions.
1401		 */
1402		qtype = NI_QTYPE_FQDN;
1403		/* XXX will append an mbuf */
1404		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1405		oldfqdn++;
1406		break;
1407	}
1408
1409	/* Allocate an mbuf to reply. */
1410	if (replylen > MCLBYTES) {
1411		/*
1412		 * XXX: should we try to allocate more? But MCLBYTES
1413		 * is probably much larger than IPV6_MMTU...
1414		 */
1415		goto bad;
1416	}
1417	if (replylen > MHLEN)
1418		n = m_getcl(M_NOWAIT, m->m_type, M_PKTHDR);
1419	else
1420		n = m_gethdr(M_NOWAIT, m->m_type);
1421	if (n == NULL) {
1422		m_freem(m);
1423		return (NULL);
1424	}
1425	m_move_pkthdr(n, m); /* just for recvif and FIB */
1426	n->m_pkthdr.len = n->m_len = replylen;
1427
1428	/* copy mbuf header and IPv6 + Node Information base headers */
1429	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1430	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1431	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1432
1433	/* qtype dependent procedure */
1434	switch (qtype) {
1435	case NI_QTYPE_NOOP:
1436		nni6->ni_code = ICMP6_NI_SUCCESS;
1437		nni6->ni_flags = 0;
1438		break;
1439	case NI_QTYPE_SUPTYPES:
1440	{
1441		u_int32_t v;
1442		nni6->ni_code = ICMP6_NI_SUCCESS;
1443		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1444		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1445		v = (u_int32_t)htonl(0x0000000f);
1446		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1447		break;
1448	}
1449	case NI_QTYPE_FQDN:
1450		nni6->ni_code = ICMP6_NI_SUCCESS;
1451		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1452		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1453		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1454		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1455		/*
1456		 * XXX do we really have FQDN in hostname?
1457		 */
1458		pr = curthread->td_ucred->cr_prison;
1459		mtx_lock(&pr->pr_mtx);
1460		n->m_next = ni6_nametodns(pr->pr_hostname,
1461		    strlen(pr->pr_hostname), oldfqdn);
1462		mtx_unlock(&pr->pr_mtx);
1463		if (n->m_next == NULL)
1464			goto bad;
1465		/* XXX we assume that n->m_next is not a chain */
1466		if (n->m_next->m_next != NULL)
1467			goto bad;
1468		n->m_pkthdr.len += n->m_next->m_len;
1469		break;
1470	case NI_QTYPE_NODEADDR:
1471	{
1472		int lenlim, copied;
1473
1474		nni6->ni_code = ICMP6_NI_SUCCESS;
1475		n->m_pkthdr.len = n->m_len =
1476		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1477		lenlim = M_TRAILINGSPACE(n);
1478		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1479		/* XXX: reset mbuf length */
1480		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1481		    sizeof(struct icmp6_nodeinfo) + copied;
1482		break;
1483	}
1484	default:
1485		break;		/* XXX impossible! */
1486	}
1487
1488	nni6->ni_type = ICMP6_NI_REPLY;
1489	m_freem(m);
1490	return (n);
1491
1492  bad:
1493	m_freem(m);
1494	if (n)
1495		m_freem(n);
1496	return (NULL);
1497}
1498
1499/*
1500 * make a mbuf with DNS-encoded string.  no compression support.
1501 *
1502 * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1503 * treated as truncated name (two \0 at the end).  this is a wild guess.
1504 *
1505 * old - return pascal string if non-zero
1506 */
1507static struct mbuf *
1508ni6_nametodns(const char *name, int namelen, int old)
1509{
1510	struct mbuf *m;
1511	char *cp, *ep;
1512	const char *p, *q;
1513	int i, len, nterm;
1514
1515	if (old)
1516		len = namelen + 1;
1517	else
1518		len = MCLBYTES;
1519
1520	/* Because MAXHOSTNAMELEN is usually 256, we use cluster mbuf. */
1521	if (len > MLEN)
1522		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1523	else
1524		m = m_get(M_NOWAIT, MT_DATA);
1525	if (m == NULL)
1526		goto fail;
1527
1528	if (old) {
1529		m->m_len = len;
1530		*mtod(m, char *) = namelen;
1531		bcopy(name, mtod(m, char *) + 1, namelen);
1532		return m;
1533	} else {
1534		m->m_len = 0;
1535		cp = mtod(m, char *);
1536		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1537
1538		/* if not certain about my name, return empty buffer */
1539		if (namelen == 0)
1540			return m;
1541
1542		/*
1543		 * guess if it looks like shortened hostname, or FQDN.
1544		 * shortened hostname needs two trailing "\0".
1545		 */
1546		i = 0;
1547		for (p = name; p < name + namelen; p++) {
1548			if (*p && *p == '.')
1549				i++;
1550		}
1551		if (i < 2)
1552			nterm = 2;
1553		else
1554			nterm = 1;
1555
1556		p = name;
1557		while (cp < ep && p < name + namelen) {
1558			i = 0;
1559			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1560				i++;
1561			/* result does not fit into mbuf */
1562			if (cp + i + 1 >= ep)
1563				goto fail;
1564			/*
1565			 * DNS label length restriction, RFC1035 page 8.
1566			 * "i == 0" case is included here to avoid returning
1567			 * 0-length label on "foo..bar".
1568			 */
1569			if (i <= 0 || i >= 64)
1570				goto fail;
1571			*cp++ = i;
1572			bcopy(p, cp, i);
1573			cp += i;
1574			p = q;
1575			if (p < name + namelen && *p == '.')
1576				p++;
1577		}
1578		/* termination */
1579		if (cp + nterm >= ep)
1580			goto fail;
1581		while (nterm-- > 0)
1582			*cp++ = '\0';
1583		m->m_len = cp - mtod(m, char *);
1584		return m;
1585	}
1586
1587	panic("should not reach here");
1588	/* NOTREACHED */
1589
1590 fail:
1591	if (m)
1592		m_freem(m);
1593	return NULL;
1594}
1595
1596/*
1597 * check if two DNS-encoded string matches.  takes care of truncated
1598 * form (with \0\0 at the end).  no compression support.
1599 * XXX upper/lowercase match (see RFC2065)
1600 */
1601static int
1602ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1603{
1604	const char *a0, *b0;
1605	int l;
1606
1607	/* simplest case - need validation? */
1608	if (alen == blen && bcmp(a, b, alen) == 0)
1609		return 1;
1610
1611	a0 = a;
1612	b0 = b;
1613
1614	/* termination is mandatory */
1615	if (alen < 2 || blen < 2)
1616		return 0;
1617	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1618		return 0;
1619	alen--;
1620	blen--;
1621
1622	while (a - a0 < alen && b - b0 < blen) {
1623		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1624			return 0;
1625
1626		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1627			return 0;
1628		/* we don't support compression yet */
1629		if (a[0] >= 64 || b[0] >= 64)
1630			return 0;
1631
1632		/* truncated case */
1633		if (a[0] == 0 && a - a0 == alen - 1)
1634			return 1;
1635		if (b[0] == 0 && b - b0 == blen - 1)
1636			return 1;
1637		if (a[0] == 0 || b[0] == 0)
1638			return 0;
1639
1640		if (a[0] != b[0])
1641			return 0;
1642		l = a[0];
1643		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1644			return 0;
1645		if (bcmp(a + 1, b + 1, l) != 0)
1646			return 0;
1647
1648		a += 1 + l;
1649		b += 1 + l;
1650	}
1651
1652	if (a - a0 == alen && b - b0 == blen)
1653		return 1;
1654	else
1655		return 0;
1656}
1657
1658/*
1659 * calculate the number of addresses to be returned in the node info reply.
1660 */
1661static int
1662ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1663    struct in6_addr *subj)
1664{
1665	struct ifnet *ifp;
1666	struct in6_ifaddr *ifa6;
1667	struct ifaddr *ifa;
1668	int addrs = 0, addrsofif, iffound = 0;
1669	int niflags = ni6->ni_flags;
1670
1671	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1672		switch (ni6->ni_code) {
1673		case ICMP6_NI_SUBJ_IPV6:
1674			if (subj == NULL) /* must be impossible... */
1675				return (0);
1676			break;
1677		default:
1678			/*
1679			 * XXX: we only support IPv6 subject address for
1680			 * this Qtype.
1681			 */
1682			return (0);
1683		}
1684	}
1685
1686	IFNET_RLOCK_NOSLEEP();
1687	TAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1688		addrsofif = 0;
1689		IF_ADDR_RLOCK(ifp);
1690		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1691			if (ifa->ifa_addr->sa_family != AF_INET6)
1692				continue;
1693			ifa6 = (struct in6_ifaddr *)ifa;
1694
1695			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1696			    IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1697				iffound = 1;
1698
1699			/*
1700			 * IPv4-mapped addresses can only be returned by a
1701			 * Node Information proxy, since they represent
1702			 * addresses of IPv4-only nodes, which perforce do
1703			 * not implement this protocol.
1704			 * [icmp-name-lookups-07, Section 5.4]
1705			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1706			 * this function at this moment.
1707			 */
1708
1709			/* What do we have to do about ::1? */
1710			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1711			case IPV6_ADDR_SCOPE_LINKLOCAL:
1712				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1713					continue;
1714				break;
1715			case IPV6_ADDR_SCOPE_SITELOCAL:
1716				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1717					continue;
1718				break;
1719			case IPV6_ADDR_SCOPE_GLOBAL:
1720				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1721					continue;
1722				break;
1723			default:
1724				continue;
1725			}
1726
1727			/*
1728			 * check if anycast is okay.
1729			 * XXX: just experimental.  not in the spec.
1730			 */
1731			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1732			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1733				continue; /* we need only unicast addresses */
1734			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1735			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1736				continue;
1737			}
1738			addrsofif++; /* count the address */
1739		}
1740		IF_ADDR_RUNLOCK(ifp);
1741		if (iffound) {
1742			*ifpp = ifp;
1743			IFNET_RUNLOCK_NOSLEEP();
1744			return (addrsofif);
1745		}
1746
1747		addrs += addrsofif;
1748	}
1749	IFNET_RUNLOCK_NOSLEEP();
1750
1751	return (addrs);
1752}
1753
1754static int
1755ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1756    struct ifnet *ifp0, int resid)
1757{
1758	struct ifnet *ifp;
1759	struct in6_ifaddr *ifa6;
1760	struct ifaddr *ifa;
1761	struct ifnet *ifp_dep = NULL;
1762	int copied = 0, allow_deprecated = 0;
1763	u_char *cp = (u_char *)(nni6 + 1);
1764	int niflags = ni6->ni_flags;
1765	u_int32_t ltime;
1766
1767	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1768		return (0);	/* needless to copy */
1769
1770	IFNET_RLOCK_NOSLEEP();
1771	ifp = ifp0 ? ifp0 : TAILQ_FIRST(&V_ifnet);
1772  again:
1773
1774	for (; ifp; ifp = TAILQ_NEXT(ifp, if_link)) {
1775		IF_ADDR_RLOCK(ifp);
1776		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1777			if (ifa->ifa_addr->sa_family != AF_INET6)
1778				continue;
1779			ifa6 = (struct in6_ifaddr *)ifa;
1780
1781			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1782			    allow_deprecated == 0) {
1783				/*
1784				 * prefererred address should be put before
1785				 * deprecated addresses.
1786				 */
1787
1788				/* record the interface for later search */
1789				if (ifp_dep == NULL)
1790					ifp_dep = ifp;
1791
1792				continue;
1793			} else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1794			    allow_deprecated != 0)
1795				continue; /* we now collect deprecated addrs */
1796
1797			/* What do we have to do about ::1? */
1798			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1799			case IPV6_ADDR_SCOPE_LINKLOCAL:
1800				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1801					continue;
1802				break;
1803			case IPV6_ADDR_SCOPE_SITELOCAL:
1804				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1805					continue;
1806				break;
1807			case IPV6_ADDR_SCOPE_GLOBAL:
1808				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1809					continue;
1810				break;
1811			default:
1812				continue;
1813			}
1814
1815			/*
1816			 * check if anycast is okay.
1817			 * XXX: just experimental.  not in the spec.
1818			 */
1819			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1820			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1821				continue;
1822			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1823			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1824				continue;
1825			}
1826
1827			/* now we can copy the address */
1828			if (resid < sizeof(struct in6_addr) +
1829			    sizeof(u_int32_t)) {
1830				IF_ADDR_RUNLOCK(ifp);
1831				/*
1832				 * We give up much more copy.
1833				 * Set the truncate flag and return.
1834				 */
1835				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1836				IFNET_RUNLOCK_NOSLEEP();
1837				return (copied);
1838			}
1839
1840			/*
1841			 * Set the TTL of the address.
1842			 * The TTL value should be one of the following
1843			 * according to the specification:
1844			 *
1845			 * 1. The remaining lifetime of a DHCP lease on the
1846			 *    address, or
1847			 * 2. The remaining Valid Lifetime of a prefix from
1848			 *    which the address was derived through Stateless
1849			 *    Autoconfiguration.
1850			 *
1851			 * Note that we currently do not support stateful
1852			 * address configuration by DHCPv6, so the former
1853			 * case can't happen.
1854			 */
1855			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1856				ltime = ND6_INFINITE_LIFETIME;
1857			else {
1858				if (ifa6->ia6_lifetime.ia6t_expire >
1859				    time_uptime)
1860					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime);
1861				else
1862					ltime = 0;
1863			}
1864
1865			bcopy(&ltime, cp, sizeof(u_int32_t));
1866			cp += sizeof(u_int32_t);
1867
1868			/* copy the address itself */
1869			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1870			    sizeof(struct in6_addr));
1871			in6_clearscope((struct in6_addr *)cp); /* XXX */
1872			cp += sizeof(struct in6_addr);
1873
1874			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1875			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1876		}
1877		IF_ADDR_RUNLOCK(ifp);
1878		if (ifp0)	/* we need search only on the specified IF */
1879			break;
1880	}
1881
1882	if (allow_deprecated == 0 && ifp_dep != NULL) {
1883		ifp = ifp_dep;
1884		allow_deprecated = 1;
1885
1886		goto again;
1887	}
1888
1889	IFNET_RUNLOCK_NOSLEEP();
1890
1891	return (copied);
1892}
1893
1894/*
1895 * XXX almost dup'ed code with rip6_input.
1896 */
1897static int
1898icmp6_rip6_input(struct mbuf **mp, int off)
1899{
1900	struct mbuf *m = *mp;
1901	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1902	struct inpcb *in6p;
1903	struct inpcb *last = NULL;
1904	struct sockaddr_in6 fromsa;
1905	struct icmp6_hdr *icmp6;
1906	struct mbuf *opts = NULL;
1907
1908#ifndef PULLDOWN_TEST
1909	/* this is assumed to be safe. */
1910	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1911#else
1912	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1913	if (icmp6 == NULL) {
1914		/* m is already reclaimed */
1915		return (IPPROTO_DONE);
1916	}
1917#endif
1918
1919	/*
1920	 * XXX: the address may have embedded scope zone ID, which should be
1921	 * hidden from applications.
1922	 */
1923	bzero(&fromsa, sizeof(fromsa));
1924	fromsa.sin6_family = AF_INET6;
1925	fromsa.sin6_len = sizeof(struct sockaddr_in6);
1926	fromsa.sin6_addr = ip6->ip6_src;
1927	if (sa6_recoverscope(&fromsa)) {
1928		m_freem(m);
1929		return (IPPROTO_DONE);
1930	}
1931
1932	INP_INFO_RLOCK(&V_ripcbinfo);
1933	LIST_FOREACH(in6p, &V_ripcb, inp_list) {
1934		if ((in6p->inp_vflag & INP_IPV6) == 0)
1935			continue;
1936		if (in6p->inp_ip_p != IPPROTO_ICMPV6)
1937			continue;
1938		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1939		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1940			continue;
1941		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1942		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1943			continue;
1944		INP_RLOCK(in6p);
1945		if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1946		    in6p->in6p_icmp6filt)) {
1947			INP_RUNLOCK(in6p);
1948			continue;
1949		}
1950		if (last != NULL) {
1951			struct	mbuf *n = NULL;
1952
1953			/*
1954			 * Recent network drivers tend to allocate a single
1955			 * mbuf cluster, rather than to make a couple of
1956			 * mbufs without clusters.  Also, since the IPv6 code
1957			 * path tries to avoid m_pullup(), it is highly
1958			 * probable that we still have an mbuf cluster here
1959			 * even though the necessary length can be stored in an
1960			 * mbuf's internal buffer.
1961			 * Meanwhile, the default size of the receive socket
1962			 * buffer for raw sockets is not so large.  This means
1963			 * the possibility of packet loss is relatively higher
1964			 * than before.  To avoid this scenario, we copy the
1965			 * received data to a separate mbuf that does not use
1966			 * a cluster, if possible.
1967			 * XXX: it is better to copy the data after stripping
1968			 * intermediate headers.
1969			 */
1970			if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1971			    m->m_len <= MHLEN) {
1972				n = m_get(M_NOWAIT, m->m_type);
1973				if (n != NULL) {
1974					if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1975						bcopy(m->m_data, n->m_data,
1976						      m->m_len);
1977						n->m_len = m->m_len;
1978					} else {
1979						m_free(n);
1980						n = NULL;
1981					}
1982				}
1983			}
1984			if (n != NULL ||
1985			    (n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1986				if (last->inp_flags & INP_CONTROLOPTS)
1987					ip6_savecontrol(last, n, &opts);
1988				/* strip intermediate headers */
1989				m_adj(n, off);
1990				SOCKBUF_LOCK(&last->inp_socket->so_rcv);
1991				if (sbappendaddr_locked(
1992				    &last->inp_socket->so_rcv,
1993				    (struct sockaddr *)&fromsa, n, opts)
1994				    == 0) {
1995					/* should notify about lost packet */
1996					m_freem(n);
1997					if (opts) {
1998						m_freem(opts);
1999					}
2000					SOCKBUF_UNLOCK(
2001					    &last->inp_socket->so_rcv);
2002				} else
2003					sorwakeup_locked(last->inp_socket);
2004				opts = NULL;
2005			}
2006			INP_RUNLOCK(last);
2007		}
2008		last = in6p;
2009	}
2010	INP_INFO_RUNLOCK(&V_ripcbinfo);
2011	if (last != NULL) {
2012		if (last->inp_flags & INP_CONTROLOPTS)
2013			ip6_savecontrol(last, m, &opts);
2014		/* strip intermediate headers */
2015		m_adj(m, off);
2016
2017		/* avoid using mbuf clusters if possible (see above) */
2018		if ((m->m_flags & M_EXT) && m->m_next == NULL &&
2019		    m->m_len <= MHLEN) {
2020			struct mbuf *n;
2021
2022			n = m_get(M_NOWAIT, m->m_type);
2023			if (n != NULL) {
2024				if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2025					bcopy(m->m_data, n->m_data, m->m_len);
2026					n->m_len = m->m_len;
2027
2028					m_freem(m);
2029					m = n;
2030				} else {
2031					m_freem(n);
2032					n = NULL;
2033				}
2034			}
2035		}
2036		SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2037		if (sbappendaddr_locked(&last->inp_socket->so_rcv,
2038		    (struct sockaddr *)&fromsa, m, opts) == 0) {
2039			m_freem(m);
2040			if (opts)
2041				m_freem(opts);
2042			SOCKBUF_UNLOCK(&last->inp_socket->so_rcv);
2043		} else
2044			sorwakeup_locked(last->inp_socket);
2045		INP_RUNLOCK(last);
2046	} else {
2047		m_freem(m);
2048		IP6STAT_DEC(ip6s_delivered);
2049	}
2050	return IPPROTO_DONE;
2051}
2052
2053/*
2054 * Reflect the ip6 packet back to the source.
2055 * OFF points to the icmp6 header, counted from the top of the mbuf.
2056 */
2057void
2058icmp6_reflect(struct mbuf *m, size_t off)
2059{
2060	struct in6_addr src6, *srcp;
2061	struct ip6_hdr *ip6;
2062	struct icmp6_hdr *icmp6;
2063	struct in6_ifaddr *ia = NULL;
2064	struct ifnet *outif = NULL;
2065	int plen;
2066	int type, code, hlim;
2067
2068	/* too short to reflect */
2069	if (off < sizeof(struct ip6_hdr)) {
2070		nd6log((LOG_DEBUG,
2071		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2072		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2073		    __FILE__, __LINE__));
2074		goto bad;
2075	}
2076
2077	/*
2078	 * If there are extra headers between IPv6 and ICMPv6, strip
2079	 * off that header first.
2080	 */
2081#ifdef DIAGNOSTIC
2082	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2083		panic("assumption failed in icmp6_reflect");
2084#endif
2085	if (off > sizeof(struct ip6_hdr)) {
2086		size_t l;
2087		struct ip6_hdr nip6;
2088
2089		l = off - sizeof(struct ip6_hdr);
2090		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2091		m_adj(m, l);
2092		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2093		if (m->m_len < l) {
2094			if ((m = m_pullup(m, l)) == NULL)
2095				return;
2096		}
2097		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2098	} else /* off == sizeof(struct ip6_hdr) */ {
2099		size_t l;
2100		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2101		if (m->m_len < l) {
2102			if ((m = m_pullup(m, l)) == NULL)
2103				return;
2104		}
2105	}
2106	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2107	ip6 = mtod(m, struct ip6_hdr *);
2108	ip6->ip6_nxt = IPPROTO_ICMPV6;
2109	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2110	type = icmp6->icmp6_type; /* keep type for statistics */
2111	code = icmp6->icmp6_code; /* ditto. */
2112	hlim = 0;
2113	srcp = NULL;
2114
2115	/*
2116	 * If the incoming packet was addressed directly to us (i.e. unicast),
2117	 * use dst as the src for the reply.
2118	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2119	 * (for example) when we encounter an error while forwarding procedure
2120	 * destined to a duplicated address of ours.
2121	 */
2122	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
2123		ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */);
2124		if (ia != NULL && !(ia->ia6_flags &
2125		    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2126			src6 = ia->ia_addr.sin6_addr;
2127			srcp = &src6;
2128
2129			if (m->m_pkthdr.rcvif != NULL) {
2130				/* XXX: This may not be the outgoing interface */
2131				hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2132			} else
2133				hlim = V_ip6_defhlim;
2134		}
2135		if (ia != NULL)
2136			ifa_free(&ia->ia_ifa);
2137	}
2138
2139	if (srcp == NULL) {
2140		int error;
2141		struct in6_addr dst6;
2142		uint32_t scopeid;
2143
2144		/*
2145		 * This case matches to multicasts, our anycast, or unicasts
2146		 * that we do not own.  Select a source address based on the
2147		 * source address of the erroneous packet.
2148		 */
2149		in6_splitscope(&ip6->ip6_dst, &dst6, &scopeid);
2150		error = in6_selectsrc_addr(RT_DEFAULT_FIB, &dst6,
2151		    scopeid, NULL, &src6, &hlim);
2152
2153		if (error) {
2154			char ip6buf[INET6_ADDRSTRLEN];
2155			nd6log((LOG_DEBUG,
2156			    "icmp6_reflect: source can't be determined: "
2157			    "dst=%s, error=%d\n",
2158			    ip6_sprintf(ip6buf, &ip6->ip6_dst), error));
2159			goto bad;
2160		}
2161		srcp = &src6;
2162	}
2163	/*
2164	 * ip6_input() drops a packet if its src is multicast.
2165	 * So, the src is never multicast.
2166	 */
2167	ip6->ip6_dst = ip6->ip6_src;
2168	ip6->ip6_src = *srcp;
2169	ip6->ip6_flow = 0;
2170	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2171	ip6->ip6_vfc |= IPV6_VERSION;
2172	ip6->ip6_nxt = IPPROTO_ICMPV6;
2173	ip6->ip6_hlim = hlim;
2174
2175	icmp6->icmp6_cksum = 0;
2176	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2177	    sizeof(struct ip6_hdr), plen);
2178
2179	/*
2180	 * XXX option handling
2181	 */
2182
2183	m->m_flags &= ~(M_BCAST|M_MCAST);
2184
2185	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2186	if (outif)
2187		icmp6_ifoutstat_inc(outif, type, code);
2188
2189	return;
2190
2191 bad:
2192	m_freem(m);
2193	return;
2194}
2195
2196void
2197icmp6_fasttimo(void)
2198{
2199
2200	mld_fasttimo();
2201}
2202
2203void
2204icmp6_slowtimo(void)
2205{
2206
2207	mld_slowtimo();
2208}
2209
2210static const char *
2211icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2212    struct in6_addr *tgt6)
2213{
2214	static char buf[1024];
2215	char ip6bufs[INET6_ADDRSTRLEN];
2216	char ip6bufd[INET6_ADDRSTRLEN];
2217	char ip6buft[INET6_ADDRSTRLEN];
2218	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2219	    ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2220	    ip6_sprintf(ip6buft, tgt6));
2221	return buf;
2222}
2223
2224void
2225icmp6_redirect_input(struct mbuf *m, int off)
2226{
2227	struct ifnet *ifp;
2228	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2229	struct nd_redirect *nd_rd;
2230	int icmp6len = ntohs(ip6->ip6_plen);
2231	char *lladdr = NULL;
2232	int lladdrlen = 0;
2233	int is_router;
2234	int is_onlink;
2235	struct in6_addr src6 = ip6->ip6_src;
2236	struct in6_addr redtgt6;
2237	struct in6_addr reddst6;
2238	union nd_opts ndopts;
2239	char ip6buf[INET6_ADDRSTRLEN];
2240
2241	M_ASSERTPKTHDR(m);
2242	KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2243
2244	ifp = m->m_pkthdr.rcvif;
2245
2246	/* XXX if we are router, we don't update route by icmp6 redirect */
2247	if (V_ip6_forwarding)
2248		goto freeit;
2249	if (!V_icmp6_rediraccept)
2250		goto freeit;
2251
2252#ifndef PULLDOWN_TEST
2253	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2254	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2255#else
2256	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2257	if (nd_rd == NULL) {
2258		ICMP6STAT_INC(icp6s_tooshort);
2259		return;
2260	}
2261#endif
2262	redtgt6 = nd_rd->nd_rd_target;
2263	reddst6 = nd_rd->nd_rd_dst;
2264
2265	if (in6_setscope(&redtgt6, m->m_pkthdr.rcvif, NULL) ||
2266	    in6_setscope(&reddst6, m->m_pkthdr.rcvif, NULL)) {
2267		goto freeit;
2268	}
2269
2270	/* validation */
2271	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2272		nd6log((LOG_ERR,
2273		    "ICMP6 redirect sent from %s rejected; "
2274		    "must be from linklocal\n",
2275		    ip6_sprintf(ip6buf, &src6)));
2276		goto bad;
2277	}
2278	if (ip6->ip6_hlim != 255) {
2279		nd6log((LOG_ERR,
2280		    "ICMP6 redirect sent from %s rejected; "
2281		    "hlim=%d (must be 255)\n",
2282		    ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2283		goto bad;
2284	}
2285    {
2286	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2287	struct nhop6_basic nh6;
2288	struct in6_addr kdst;
2289	uint32_t scopeid;
2290
2291	in6_splitscope(&reddst6, &kdst, &scopeid);
2292	if (fib6_lookup_nh_basic(RT_DEFAULT_FIB, &kdst, scopeid, 0, 0,&nh6)==0){
2293		if ((nh6.nh_flags & NHF_GATEWAY) == 0) {
2294			nd6log((LOG_ERR,
2295			    "ICMP6 redirect rejected; no route "
2296			    "with inet6 gateway found for redirect dst: %s\n",
2297			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2298			goto bad;
2299		}
2300
2301		if (IN6_ARE_ADDR_EQUAL(&src6, &nh6.nh_addr) == 0) {
2302			nd6log((LOG_ERR,
2303			    "ICMP6 redirect rejected; "
2304			    "not equal to gw-for-src=%s (must be same): "
2305			    "%s\n",
2306			    ip6_sprintf(ip6buf, &nh6.nh_addr),
2307			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2308			goto bad;
2309		}
2310	} else {
2311		nd6log((LOG_ERR,
2312		    "ICMP6 redirect rejected; "
2313		    "no route found for redirect dst: %s\n",
2314		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2315		goto bad;
2316	}
2317    }
2318	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2319		nd6log((LOG_ERR,
2320		    "ICMP6 redirect rejected; "
2321		    "redirect dst must be unicast: %s\n",
2322		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2323		goto bad;
2324	}
2325
2326	is_router = is_onlink = 0;
2327	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2328		is_router = 1;	/* router case */
2329	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2330		is_onlink = 1;	/* on-link destination case */
2331	if (!is_router && !is_onlink) {
2332		nd6log((LOG_ERR,
2333		    "ICMP6 redirect rejected; "
2334		    "neither router case nor onlink case: %s\n",
2335		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2336		goto bad;
2337	}
2338
2339	icmp6len -= sizeof(*nd_rd);
2340	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2341	if (nd6_options(&ndopts) < 0) {
2342		nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2343		    __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2344		/* nd6_options have incremented stats */
2345		goto freeit;
2346	}
2347
2348	if (ndopts.nd_opts_tgt_lladdr) {
2349		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2350		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2351	}
2352
2353	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2354		nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2355		    "(if %d, icmp6 packet %d): %s\n",
2356		    __func__, ip6_sprintf(ip6buf, &redtgt6),
2357		    ifp->if_addrlen, lladdrlen - 2,
2358		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2359		goto bad;
2360	}
2361
2362	/* Validation passed. */
2363
2364	/* RFC 2461 8.3 */
2365	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2366	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2367
2368	/*
2369	 * Install a gateway route in the better-router case or an interface
2370	 * route in the on-link-destination case.
2371	 */
2372	{
2373		struct sockaddr_in6 sdst;
2374		struct sockaddr_in6 sgw;
2375		struct sockaddr_in6 ssrc;
2376		struct sockaddr *gw;
2377		int rt_flags;
2378		u_int fibnum;
2379
2380		bzero(&sdst, sizeof(sdst));
2381		bzero(&ssrc, sizeof(ssrc));
2382		sdst.sin6_family = ssrc.sin6_family = AF_INET6;
2383		sdst.sin6_len = ssrc.sin6_len = sizeof(struct sockaddr_in6);
2384		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2385		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2386		rt_flags = RTF_HOST;
2387		if (is_router) {
2388			bzero(&sgw, sizeof(sgw));
2389			sgw.sin6_family = AF_INET6;
2390			sgw.sin6_len = sizeof(struct sockaddr_in6);
2391			bcopy(&redtgt6, &sgw.sin6_addr,
2392				sizeof(struct in6_addr));
2393			gw = (struct sockaddr *)&sgw;
2394			rt_flags |= RTF_GATEWAY;
2395		} else
2396			gw = ifp->if_addr->ifa_addr;
2397		for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2398			in6_rtredirect((struct sockaddr *)&sdst, gw,
2399			    (struct sockaddr *)NULL, rt_flags,
2400			    (struct sockaddr *)&ssrc, fibnum);
2401	}
2402	/* finally update cached route in each socket via pfctlinput */
2403    {
2404	struct sockaddr_in6 sdst;
2405
2406	bzero(&sdst, sizeof(sdst));
2407	sdst.sin6_family = AF_INET6;
2408	sdst.sin6_len = sizeof(struct sockaddr_in6);
2409	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2410	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2411    }
2412
2413 freeit:
2414	m_freem(m);
2415	return;
2416
2417 bad:
2418	ICMP6STAT_INC(icp6s_badredirect);
2419	m_freem(m);
2420}
2421
2422void
2423icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2424{
2425	struct ifnet *ifp;	/* my outgoing interface */
2426	struct in6_addr *ifp_ll6;
2427	struct in6_addr *router_ll6;
2428	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2429	struct mbuf *m = NULL;	/* newly allocated one */
2430	struct m_tag *mtag;
2431	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2432	struct nd_redirect *nd_rd;
2433	struct llentry *ln = NULL;
2434	size_t maxlen;
2435	u_char *p;
2436	struct ifnet *outif = NULL;
2437	struct sockaddr_in6 src_sa;
2438
2439	icmp6_errcount(ND_REDIRECT, 0);
2440
2441	/* if we are not router, we don't send icmp6 redirect */
2442	if (!V_ip6_forwarding)
2443		goto fail;
2444
2445	/* sanity check */
2446	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2447		goto fail;
2448
2449	/*
2450	 * Address check:
2451	 *  the source address must identify a neighbor, and
2452	 *  the destination address must not be a multicast address
2453	 *  [RFC 2461, sec 8.2]
2454	 */
2455	sip6 = mtod(m0, struct ip6_hdr *);
2456	bzero(&src_sa, sizeof(src_sa));
2457	src_sa.sin6_family = AF_INET6;
2458	src_sa.sin6_len = sizeof(src_sa);
2459	src_sa.sin6_addr = sip6->ip6_src;
2460	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2461		goto fail;
2462	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2463		goto fail;	/* what should we do here? */
2464
2465	/* rate limit */
2466	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2467		goto fail;
2468
2469	/*
2470	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2471	 * we almost always ask for an mbuf cluster for simplicity.
2472	 * (MHLEN < IPV6_MMTU is almost always true)
2473	 */
2474#if IPV6_MMTU >= MCLBYTES
2475# error assumption failed about IPV6_MMTU and MCLBYTES
2476#endif
2477	m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2478	if (m == NULL)
2479		goto fail;
2480	M_SETFIB(m, rt->rt_fibnum);
2481	maxlen = M_TRAILINGSPACE(m);
2482	maxlen = min(IPV6_MMTU, maxlen);
2483	/* just for safety */
2484	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2485	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2486		goto fail;
2487	}
2488
2489	{
2490		/* get ip6 linklocal address for ifp(my outgoing interface). */
2491		struct in6_ifaddr *ia;
2492		if ((ia = in6ifa_ifpforlinklocal(ifp,
2493						 IN6_IFF_NOTREADY|
2494						 IN6_IFF_ANYCAST)) == NULL)
2495			goto fail;
2496		ifp_ll6 = &ia->ia_addr.sin6_addr;
2497		/* XXXRW: reference released prematurely. */
2498		ifa_free(&ia->ia_ifa);
2499	}
2500
2501	/* get ip6 linklocal address for the router. */
2502	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2503		struct sockaddr_in6 *sin6;
2504		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2505		router_ll6 = &sin6->sin6_addr;
2506		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2507			router_ll6 = (struct in6_addr *)NULL;
2508	} else
2509		router_ll6 = (struct in6_addr *)NULL;
2510
2511	/* ip6 */
2512	ip6 = mtod(m, struct ip6_hdr *);
2513	ip6->ip6_flow = 0;
2514	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2515	ip6->ip6_vfc |= IPV6_VERSION;
2516	/* ip6->ip6_plen will be set later */
2517	ip6->ip6_nxt = IPPROTO_ICMPV6;
2518	ip6->ip6_hlim = 255;
2519	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2520	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2521	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2522
2523	/* ND Redirect */
2524	nd_rd = (struct nd_redirect *)(ip6 + 1);
2525	nd_rd->nd_rd_type = ND_REDIRECT;
2526	nd_rd->nd_rd_code = 0;
2527	nd_rd->nd_rd_reserved = 0;
2528	if (rt->rt_flags & RTF_GATEWAY) {
2529		/*
2530		 * nd_rd->nd_rd_target must be a link-local address in
2531		 * better router cases.
2532		 */
2533		if (!router_ll6)
2534			goto fail;
2535		bcopy(router_ll6, &nd_rd->nd_rd_target,
2536		    sizeof(nd_rd->nd_rd_target));
2537		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2538		    sizeof(nd_rd->nd_rd_dst));
2539	} else {
2540		/* make sure redtgt == reddst */
2541		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2542		    sizeof(nd_rd->nd_rd_target));
2543		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2544		    sizeof(nd_rd->nd_rd_dst));
2545	}
2546
2547	p = (u_char *)(nd_rd + 1);
2548
2549	if (!router_ll6)
2550		goto nolladdropt;
2551
2552	{
2553		/* target lladdr option */
2554		int len;
2555		struct nd_opt_hdr *nd_opt;
2556		char *lladdr;
2557
2558		IF_AFDATA_RLOCK(ifp);
2559		ln = nd6_lookup(router_ll6, 0, ifp);
2560		IF_AFDATA_RUNLOCK(ifp);
2561		if (ln == NULL)
2562			goto nolladdropt;
2563
2564		len = sizeof(*nd_opt) + ifp->if_addrlen;
2565		len = (len + 7) & ~7;	/* round by 8 */
2566		/* safety check */
2567		if (len + (p - (u_char *)ip6) > maxlen)
2568			goto nolladdropt;
2569
2570		if (ln->la_flags & LLE_VALID) {
2571			nd_opt = (struct nd_opt_hdr *)p;
2572			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2573			nd_opt->nd_opt_len = len >> 3;
2574			lladdr = (char *)(nd_opt + 1);
2575			bcopy(ln->ll_addr, lladdr, ifp->if_addrlen);
2576			p += len;
2577		}
2578	}
2579nolladdropt:
2580	if (ln != NULL)
2581		LLE_RUNLOCK(ln);
2582
2583	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2584
2585	/* just to be safe */
2586#ifdef M_DECRYPTED	/*not openbsd*/
2587	if (m0->m_flags & M_DECRYPTED)
2588		goto noredhdropt;
2589#endif
2590	if (p - (u_char *)ip6 > maxlen)
2591		goto noredhdropt;
2592
2593	{
2594		/* redirected header option */
2595		int len;
2596		struct nd_opt_rd_hdr *nd_opt_rh;
2597
2598		/*
2599		 * compute the maximum size for icmp6 redirect header option.
2600		 * XXX room for auth header?
2601		 */
2602		len = maxlen - (p - (u_char *)ip6);
2603		len &= ~7;
2604
2605		/* This is just for simplicity. */
2606		if (m0->m_pkthdr.len != m0->m_len) {
2607			if (m0->m_next) {
2608				m_freem(m0->m_next);
2609				m0->m_next = NULL;
2610			}
2611			m0->m_pkthdr.len = m0->m_len;
2612		}
2613
2614		/*
2615		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2616		 * about padding/truncate rule for the original IP packet.
2617		 * From the discussion on IPv6imp in Feb 1999,
2618		 * the consensus was:
2619		 * - "attach as much as possible" is the goal
2620		 * - pad if not aligned (original size can be guessed by
2621		 *   original ip6 header)
2622		 * Following code adds the padding if it is simple enough,
2623		 * and truncates if not.
2624		 */
2625		if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2626			panic("assumption failed in %s:%d", __FILE__,
2627			    __LINE__);
2628
2629		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2630			/* not enough room, truncate */
2631			m0->m_pkthdr.len = m0->m_len = len -
2632			    sizeof(*nd_opt_rh);
2633		} else {
2634			/* enough room, pad or truncate */
2635			size_t extra;
2636
2637			extra = m0->m_pkthdr.len % 8;
2638			if (extra) {
2639				/* pad if easy enough, truncate if not */
2640				if (8 - extra <= M_TRAILINGSPACE(m0)) {
2641					/* pad */
2642					m0->m_len += (8 - extra);
2643					m0->m_pkthdr.len += (8 - extra);
2644				} else {
2645					/* truncate */
2646					m0->m_pkthdr.len -= extra;
2647					m0->m_len -= extra;
2648				}
2649			}
2650			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2651			m0->m_pkthdr.len = m0->m_len = len -
2652			    sizeof(*nd_opt_rh);
2653		}
2654
2655		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2656		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2657		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2658		nd_opt_rh->nd_opt_rh_len = len >> 3;
2659		p += sizeof(*nd_opt_rh);
2660		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2661
2662		/* connect m0 to m */
2663		m_tag_delete_chain(m0, NULL);
2664		m0->m_flags &= ~M_PKTHDR;
2665		m->m_next = m0;
2666		m->m_pkthdr.len = m->m_len + m0->m_len;
2667		m0 = NULL;
2668	}
2669noredhdropt:;
2670	if (m0) {
2671		m_freem(m0);
2672		m0 = NULL;
2673	}
2674
2675	/* XXX: clear embedded link IDs in the inner header */
2676	in6_clearscope(&sip6->ip6_src);
2677	in6_clearscope(&sip6->ip6_dst);
2678	in6_clearscope(&nd_rd->nd_rd_target);
2679	in6_clearscope(&nd_rd->nd_rd_dst);
2680
2681	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2682
2683	nd_rd->nd_rd_cksum = 0;
2684	nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2685	    sizeof(*ip6), ntohs(ip6->ip6_plen));
2686
2687        if (send_sendso_input_hook != NULL) {
2688		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2689			M_NOWAIT);
2690		if (mtag == NULL)
2691			goto fail;
2692		*(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2693		m_tag_prepend(m, mtag);
2694	}
2695
2696	/* send the packet to outside... */
2697	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2698	if (outif) {
2699		icmp6_ifstat_inc(outif, ifs6_out_msg);
2700		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2701	}
2702	ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2703
2704	return;
2705
2706fail:
2707	if (m)
2708		m_freem(m);
2709	if (m0)
2710		m_freem(m0);
2711}
2712
2713/*
2714 * ICMPv6 socket option processing.
2715 */
2716int
2717icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2718{
2719	int error = 0;
2720	int optlen;
2721	struct inpcb *inp = sotoinpcb(so);
2722	int level, op, optname;
2723
2724	if (sopt) {
2725		level = sopt->sopt_level;
2726		op = sopt->sopt_dir;
2727		optname = sopt->sopt_name;
2728		optlen = sopt->sopt_valsize;
2729	} else
2730		level = op = optname = optlen = 0;
2731
2732	if (level != IPPROTO_ICMPV6) {
2733		return EINVAL;
2734	}
2735
2736	switch (op) {
2737	case PRCO_SETOPT:
2738		switch (optname) {
2739		case ICMP6_FILTER:
2740		    {
2741			struct icmp6_filter ic6f;
2742
2743			if (optlen != sizeof(ic6f)) {
2744				error = EMSGSIZE;
2745				break;
2746			}
2747			error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2748			if (error == 0) {
2749				INP_WLOCK(inp);
2750				*inp->in6p_icmp6filt = ic6f;
2751				INP_WUNLOCK(inp);
2752			}
2753			break;
2754		    }
2755
2756		default:
2757			error = ENOPROTOOPT;
2758			break;
2759		}
2760		break;
2761
2762	case PRCO_GETOPT:
2763		switch (optname) {
2764		case ICMP6_FILTER:
2765		    {
2766			struct icmp6_filter ic6f;
2767
2768			INP_RLOCK(inp);
2769			ic6f = *inp->in6p_icmp6filt;
2770			INP_RUNLOCK(inp);
2771			error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2772			break;
2773		    }
2774
2775		default:
2776			error = ENOPROTOOPT;
2777			break;
2778		}
2779		break;
2780	}
2781
2782	return (error);
2783}
2784
2785/*
2786 * Perform rate limit check.
2787 * Returns 0 if it is okay to send the icmp6 packet.
2788 * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2789 * limitation.
2790 *
2791 * XXX per-destination/type check necessary?
2792 *
2793 * dst - not used at this moment
2794 * type - not used at this moment
2795 * code - not used at this moment
2796 */
2797static int
2798icmp6_ratelimit(const struct in6_addr *dst, const int type,
2799    const int code)
2800{
2801	int ret;
2802
2803	ret = 0;	/* okay to send */
2804
2805	/* PPS limit */
2806	if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2807	    V_icmp6errppslim)) {
2808		/* The packet is subject to rate limit */
2809		ret++;
2810	}
2811
2812	return ret;
2813}
2814