xref: /dragonfly/sys/netinet6/icmp6.c (revision 611395e5)
1 /*	$FreeBSD: src/sys/netinet6/icmp6.c,v 1.6.2.13 2003/05/06 06:46:58 suz Exp $	*/
2 /*	$DragonFly: src/sys/netinet6/icmp6.c,v 1.13 2004/10/15 22:59:10 hsu Exp $	*/
3 /*	$KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $	*/
4 
5 /*
6  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. Neither the name of the project nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 /*
35  * Copyright (c) 1982, 1986, 1988, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. All advertising materials mentioning features or use of this software
47  *    must display the following acknowledgement:
48  *	This product includes software developed by the University of
49  *	California, Berkeley and its contributors.
50  * 4. Neither the name of the University nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64  * SUCH DAMAGE.
65  *
66  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
67  */
68 
69 #include "opt_inet.h"
70 #include "opt_inet6.h"
71 #include "opt_ipsec.h"
72 
73 #include <sys/param.h>
74 #include <sys/systm.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/protosw.h>
78 #include <sys/socket.h>
79 #include <sys/socketvar.h>
80 #include <sys/time.h>
81 #include <sys/kernel.h>
82 #include <sys/syslog.h>
83 #include <sys/domain.h>
84 
85 #include <net/if.h>
86 #include <net/route.h>
87 #include <net/if_dl.h>
88 #include <net/if_types.h>
89 
90 #include <netinet/in.h>
91 #include <netinet/in_var.h>
92 #include <netinet/ip6.h>
93 #include <netinet6/ip6_var.h>
94 #include <netinet/icmp6.h>
95 #include <netinet6/mld6_var.h>
96 #include <netinet/in_pcb.h>
97 #include <netinet6/in6_pcb.h>
98 #include <netinet6/nd6.h>
99 #include <netinet6/in6_ifattach.h>
100 #include <netinet6/ip6protosw.h>
101 
102 #ifdef IPSEC
103 #include <netinet6/ipsec.h>
104 #include <netproto/key/key.h>
105 #endif
106 
107 #ifdef FAST_IPSEC
108 #include <netproto/ipsec/ipsec.h>
109 #include <netproto/ipsec/key.h>
110 #define	IPSEC
111 #endif
112 
113 #include <net/net_osdep.h>
114 
115 #ifdef HAVE_NRL_INPCB
116 /* inpcb members */
117 #define in6pcb		inpcb
118 #define in6p_laddr	inp_laddr6
119 #define in6p_faddr	inp_faddr6
120 #define in6p_icmp6filt	inp_icmp6filt
121 #define in6p_route	inp_route
122 #define in6p_socket	inp_socket
123 #define in6p_flags	inp_flags
124 #define in6p_moptions	inp_moptions6
125 #define in6p_outputopts	inp_outputopts6
126 #define in6p_ip6	inp_ipv6
127 #define in6p_flowinfo	inp_flowinfo
128 #define in6p_sp		inp_sp
129 #define in6p_next	inp_next
130 #define in6p_prev	inp_prev
131 /* macro names */
132 #define sotoin6pcb	sotoinpcb
133 /* function names */
134 #define in6_pcbdetach	in_pcbdetach
135 #define in6_rtchange	in_rtchange
136 
137 /*
138  * for KAME src sync over BSD*'s. XXX: FreeBSD (>=3) are VERY different from
139  * others...
140  */
141 #define in6p_ip6_nxt	inp_ipv6.ip6_nxt
142 #endif
143 
144 /*
145  * ppratecheck() is available, so define it here.
146  */
147 #define	HAVE_PPSRATECHECK
148 
149 extern struct domain inet6domain;
150 extern struct ip6protosw inet6sw[];
151 extern u_char ip6_protox[];
152 
153 struct icmp6stat icmp6stat;
154 
155 extern struct inpcbinfo ripcbinfo;
156 extern int icmp6errppslim;
157 static int icmp6errpps_count = 0;
158 static struct timeval icmp6errppslim_last;
159 extern int icmp6_nodeinfo;
160 
161 static void icmp6_errcount (struct icmp6errstat *, int, int);
162 static int icmp6_rip6_input (struct mbuf **, int);
163 static int icmp6_ratelimit (const struct in6_addr *, const int, const int);
164 static const char *icmp6_redirect_diag (struct in6_addr *,
165 	struct in6_addr *, struct in6_addr *);
166 #ifndef HAVE_PPSRATECHECK
167 static int ppsratecheck (struct timeval *, int *, int);
168 #endif
169 static struct mbuf *ni6_input (struct mbuf *, int);
170 static struct mbuf *ni6_nametodns (const char *, int, int);
171 static int ni6_dnsmatch (const char *, int, const char *, int);
172 static int ni6_addrs (struct icmp6_nodeinfo *, struct mbuf *,
173 			  struct ifnet **, char *);
174 static int ni6_store_addrs (struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
175 				struct ifnet *, int);
176 static int icmp6_notify_error (struct mbuf *, int, int, int);
177 
178 #ifdef COMPAT_RFC1885
179 static struct route_in6 icmp6_reflect_rt;
180 #endif
181 
182 
183 void
184 icmp6_init(void)
185 {
186 	mld6_init();
187 }
188 
189 static void
190 icmp6_errcount(struct icmp6errstat *stat, int type, int code)
191 {
192 	switch (type) {
193 	case ICMP6_DST_UNREACH:
194 		switch (code) {
195 		case ICMP6_DST_UNREACH_NOROUTE:
196 			stat->icp6errs_dst_unreach_noroute++;
197 			return;
198 		case ICMP6_DST_UNREACH_ADMIN:
199 			stat->icp6errs_dst_unreach_admin++;
200 			return;
201 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
202 			stat->icp6errs_dst_unreach_beyondscope++;
203 			return;
204 		case ICMP6_DST_UNREACH_ADDR:
205 			stat->icp6errs_dst_unreach_addr++;
206 			return;
207 		case ICMP6_DST_UNREACH_NOPORT:
208 			stat->icp6errs_dst_unreach_noport++;
209 			return;
210 		}
211 		break;
212 	case ICMP6_PACKET_TOO_BIG:
213 		stat->icp6errs_packet_too_big++;
214 		return;
215 	case ICMP6_TIME_EXCEEDED:
216 		switch (code) {
217 		case ICMP6_TIME_EXCEED_TRANSIT:
218 			stat->icp6errs_time_exceed_transit++;
219 			return;
220 		case ICMP6_TIME_EXCEED_REASSEMBLY:
221 			stat->icp6errs_time_exceed_reassembly++;
222 			return;
223 		}
224 		break;
225 	case ICMP6_PARAM_PROB:
226 		switch (code) {
227 		case ICMP6_PARAMPROB_HEADER:
228 			stat->icp6errs_paramprob_header++;
229 			return;
230 		case ICMP6_PARAMPROB_NEXTHEADER:
231 			stat->icp6errs_paramprob_nextheader++;
232 			return;
233 		case ICMP6_PARAMPROB_OPTION:
234 			stat->icp6errs_paramprob_option++;
235 			return;
236 		}
237 		break;
238 	case ND_REDIRECT:
239 		stat->icp6errs_redirect++;
240 		return;
241 	}
242 	stat->icp6errs_unknown++;
243 }
244 
245 /*
246  * Generate an error packet of type error in response to bad IP6 packet.
247  */
248 void
249 icmp6_error(struct mbuf *m, int type, int code, int param)
250 {
251 	struct ip6_hdr *oip6, *nip6;
252 	struct icmp6_hdr *icmp6;
253 	u_int preplen;
254 	int off;
255 	int nxt;
256 
257 	icmp6stat.icp6s_error++;
258 
259 	/* count per-type-code statistics */
260 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
261 
262 #ifdef M_DECRYPTED	/*not openbsd*/
263 	if (m->m_flags & M_DECRYPTED) {
264 		icmp6stat.icp6s_canterror++;
265 		goto freeit;
266 	}
267 #endif
268 
269 #ifndef PULLDOWN_TEST
270 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
271 #else
272 	if (m->m_len < sizeof(struct ip6_hdr)) {
273 		m = m_pullup(m, sizeof(struct ip6_hdr));
274 		if (m == NULL)
275 			return;
276 	}
277 #endif
278 	oip6 = mtod(m, struct ip6_hdr *);
279 
280 	/*
281 	 * Multicast destination check. For unrecognized option errors,
282 	 * this check has already done in ip6_unknown_opt(), so we can
283 	 * check only for other errors.
284 	 */
285 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
286 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
287 	    (type != ICMP6_PACKET_TOO_BIG &&
288 	     (type != ICMP6_PARAM_PROB ||
289 	      code != ICMP6_PARAMPROB_OPTION)))
290 		goto freeit;
291 
292 	/* Source address check. XXX: the case of anycast source? */
293 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
294 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
295 		goto freeit;
296 
297 	/*
298 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
299 	 * don't do it.
300 	 */
301 	nxt = -1;
302 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
303 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
304 		struct icmp6_hdr *icp;
305 
306 #ifndef PULLDOWN_TEST
307 		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
308 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
309 #else
310 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
311 			sizeof(*icp));
312 		if (icp == NULL) {
313 			icmp6stat.icp6s_tooshort++;
314 			return;
315 		}
316 #endif
317 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
318 		    icp->icmp6_type == ND_REDIRECT) {
319 			/*
320 			 * ICMPv6 error
321 			 * Special case: for redirect (which is
322 			 * informational) we must not send icmp6 error.
323 			 */
324 			icmp6stat.icp6s_canterror++;
325 			goto freeit;
326 		} else {
327 			/* ICMPv6 informational - send the error */
328 		}
329 	} else {
330 		/* non-ICMPv6 - send the error */
331 	}
332 
333 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
334 
335 	/* Finally, do rate limitation check. */
336 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
337 		icmp6stat.icp6s_toofreq++;
338 		goto freeit;
339 	}
340 
341 	/*
342 	 * OK, ICMP6 can be generated.
343 	 */
344 
345 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
346 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
347 
348 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
349 	M_PREPEND(m, preplen, MB_DONTWAIT);
350 	if (m && m->m_len < preplen)
351 		m = m_pullup(m, preplen);
352 	if (m == NULL) {
353 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
354 		return;
355 	}
356 
357 	nip6 = mtod(m, struct ip6_hdr *);
358 	nip6->ip6_src  = oip6->ip6_src;
359 	nip6->ip6_dst  = oip6->ip6_dst;
360 
361 	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src))
362 		oip6->ip6_src.s6_addr16[1] = 0;
363 	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst))
364 		oip6->ip6_dst.s6_addr16[1] = 0;
365 
366 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
367 	icmp6->icmp6_type = type;
368 	icmp6->icmp6_code = code;
369 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
370 
371 	/*
372 	 * icmp6_reflect() is designed to be in the input path.
373 	 * icmp6_error() can be called from both input and outut path,
374 	 * and if we are in output path rcvif could contain bogus value.
375 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
376 	 * information in ip header (nip6).
377 	 */
378 	m->m_pkthdr.rcvif = NULL;
379 
380 	icmp6stat.icp6s_outhist[type]++;
381 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
382 
383 	return;
384 
385   freeit:
386 	/*
387 	 * If we can't tell wheter or not we can generate ICMP6, free it.
388 	 */
389 	m_freem(m);
390 }
391 
392 /*
393  * Process a received ICMP6 message.
394  */
395 int
396 icmp6_input(struct mbuf **mp, int *offp, int proto)
397 {
398 	struct mbuf *m = *mp, *n;
399 	struct ip6_hdr *ip6, *nip6;
400 	struct icmp6_hdr *icmp6, *nicmp6;
401 	int off = *offp;
402 	int icmp6len = m->m_pkthdr.len - *offp;
403 	int code, sum, noff;
404 
405 #ifndef PULLDOWN_TEST
406 	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
407 	/* m might change if M_LOOP.  So, call mtod after this */
408 #endif
409 
410 	/*
411 	 * Locate icmp6 structure in mbuf, and check
412 	 * that not corrupted and of at least minimum length
413 	 */
414 
415 	ip6 = mtod(m, struct ip6_hdr *);
416 	if (icmp6len < sizeof(struct icmp6_hdr)) {
417 		icmp6stat.icp6s_tooshort++;
418 		goto freeit;
419 	}
420 
421 	/*
422 	 * calculate the checksum
423 	 */
424 #ifndef PULLDOWN_TEST
425 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
426 #else
427 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
428 	if (icmp6 == NULL) {
429 		icmp6stat.icp6s_tooshort++;
430 		return IPPROTO_DONE;
431 	}
432 #endif
433 	code = icmp6->icmp6_code;
434 
435 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
436 		nd6log((LOG_ERR,
437 		    "ICMP6 checksum error(%d|%x) %s\n",
438 		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
439 		icmp6stat.icp6s_checksum++;
440 		goto freeit;
441 	}
442 
443 	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
444 		/*
445 		 * Deliver very specific ICMP6 type only.
446 		 * This is important to deilver TOOBIG.  Otherwise PMTUD
447 		 * will not work.
448 		 */
449 		switch (icmp6->icmp6_type) {
450 		case ICMP6_DST_UNREACH:
451 		case ICMP6_PACKET_TOO_BIG:
452 		case ICMP6_TIME_EXCEEDED:
453 			break;
454 		default:
455 			goto freeit;
456 		}
457 	}
458 
459 	icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
460 	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
461 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
462 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
463 
464 	switch (icmp6->icmp6_type) {
465 	case ICMP6_DST_UNREACH:
466 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
467 		switch (code) {
468 		case ICMP6_DST_UNREACH_NOROUTE:
469 			code = PRC_UNREACH_NET;
470 			break;
471 		case ICMP6_DST_UNREACH_ADMIN:
472 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
473 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
474 			break;
475 		case ICMP6_DST_UNREACH_ADDR:
476 			code = PRC_HOSTDEAD;
477 			break;
478 #ifdef COMPAT_RFC1885
479 		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
480 			code = PRC_UNREACH_SRCFAIL;
481 			break;
482 #else
483 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
484 			/* I mean "source address was incorrect." */
485 			code = PRC_PARAMPROB;
486 			break;
487 #endif
488 		case ICMP6_DST_UNREACH_NOPORT:
489 			code = PRC_UNREACH_PORT;
490 			break;
491 		default:
492 			goto badcode;
493 		}
494 		goto deliver;
495 		break;
496 
497 	case ICMP6_PACKET_TOO_BIG:
498 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
499 		if (code != 0)
500 			goto badcode;
501 
502 		code = PRC_MSGSIZE;
503 
504 		/*
505 		 * Updating the path MTU will be done after examining
506 		 * intermediate extension headers.
507 		 */
508 		goto deliver;
509 		break;
510 
511 	case ICMP6_TIME_EXCEEDED:
512 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
513 		switch (code) {
514 		case ICMP6_TIME_EXCEED_TRANSIT:
515 		case ICMP6_TIME_EXCEED_REASSEMBLY:
516 			code += PRC_TIMXCEED_INTRANS;
517 			break;
518 		default:
519 			goto badcode;
520 		}
521 		goto deliver;
522 		break;
523 
524 	case ICMP6_PARAM_PROB:
525 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
526 		switch (code) {
527 		case ICMP6_PARAMPROB_NEXTHEADER:
528 			code = PRC_UNREACH_PROTOCOL;
529 			break;
530 		case ICMP6_PARAMPROB_HEADER:
531 		case ICMP6_PARAMPROB_OPTION:
532 			code = PRC_PARAMPROB;
533 			break;
534 		default:
535 			goto badcode;
536 		}
537 		goto deliver;
538 		break;
539 
540 	case ICMP6_ECHO_REQUEST:
541 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
542 		if (code != 0)
543 			goto badcode;
544 		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
545 			/* Give up remote */
546 			break;
547 		}
548 		if ((n->m_flags & M_EXT) != 0
549 		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
550 			struct mbuf *n0 = n;
551 			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
552 			int n0len;
553 
554 			/*
555 			 * Prepare an internal mbuf. m_pullup() doesn't
556 			 * always copy the length we specified.
557 			 */
558 			if (maxlen >= MCLBYTES) {
559 				/* Give up remote */
560 				m_freem(n0);
561 				break;
562 			}
563 			MGETHDR(n, MB_DONTWAIT, n0->m_type);
564 			n0len = n0->m_pkthdr.len;	/* save for use below */
565 			if (n)
566 				M_MOVE_PKTHDR(n, n0);
567 			if (n && maxlen >= MHLEN) {
568 				MCLGET(n, MB_DONTWAIT);
569 				if ((n->m_flags & M_EXT) == 0) {
570 					m_free(n);
571 					n = NULL;
572 				}
573 			}
574 			if (n == NULL) {
575 				/* Give up remote */
576 				m_freem(n0);
577 				break;
578 			}
579 			/*
580 			 * Copy IPv6 and ICMPv6 only.
581 			 */
582 			nip6 = mtod(n, struct ip6_hdr *);
583 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
584 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
585 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
586 			noff = sizeof(struct ip6_hdr);
587 			/* new mbuf contains only ipv6+icmpv6 headers */
588 			n->m_len = noff + sizeof(struct icmp6_hdr);
589 			/*
590 			 * Adjust mbuf. ip6_plen will be adjusted in
591 			 * ip6_output().
592 			 */
593 			m_adj(n0, off + sizeof(struct icmp6_hdr));
594 			/* recalculate complete packet size */
595 			n->m_pkthdr.len = n0len + (noff - off);
596 			n->m_next = n0;
597 		} else {
598 			nip6 = mtod(n, struct ip6_hdr *);
599 			nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
600 			noff = off;
601 		}
602 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
603 		nicmp6->icmp6_code = 0;
604 		if (n) {
605 			icmp6stat.icp6s_reflect++;
606 			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
607 			icmp6_reflect(n, noff);
608 		}
609 		break;
610 
611 	case ICMP6_ECHO_REPLY:
612 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
613 		if (code != 0)
614 			goto badcode;
615 		break;
616 
617 	case MLD_LISTENER_QUERY:
618 	case MLD_LISTENER_REPORT:
619 		if (icmp6len < sizeof(struct mld_hdr))
620 			goto badlen;
621 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
622 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
623 		else
624 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
625 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
626 			/* give up local */
627 			mld6_input(m, off);
628 			m = NULL;
629 			goto freeit;
630 		}
631 		mld6_input(n, off);
632 		/* m stays. */
633 		break;
634 
635 	case MLD_LISTENER_DONE:
636 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
637 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
638 			goto badlen;
639 		break;		/* nothing to be done in kernel */
640 
641 	case MLD_MTRACE_RESP:
642 	case MLD_MTRACE:
643 		/* XXX: these two are experimental.  not officially defind. */
644 		/* XXX: per-interface statistics? */
645 		break;		/* just pass it to applications */
646 
647 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
648 	    {
649 		enum { WRU, FQDN } mode;
650 
651 		if (!icmp6_nodeinfo)
652 			break;
653 
654 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
655 			mode = WRU;
656 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
657 			mode = FQDN;
658 		else
659 			goto badlen;
660 
661 #define hostnamelen	strlen(hostname)
662 		if (mode == FQDN) {
663 #ifndef PULLDOWN_TEST
664 			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
665 					 IPPROTO_DONE);
666 #endif
667 			n = m_copy(m, 0, M_COPYALL);
668 			if (n)
669 				n = ni6_input(n, off);
670 			/* XXX meaningless if n == NULL */
671 			noff = sizeof(struct ip6_hdr);
672 		} else {
673 			u_char *p;
674 			int maxlen, maxhlen;
675 
676 			if ((icmp6_nodeinfo & 5) != 5)
677 				break;
678 
679 			if (code != 0)
680 				goto badcode;
681 			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
682 			if (maxlen >= MCLBYTES) {
683 				/* Give up remote */
684 				break;
685 			}
686 			MGETHDR(n, MB_DONTWAIT, m->m_type);
687 			if (n && maxlen > MHLEN) {
688 				MCLGET(n, MB_DONTWAIT);
689 				if ((n->m_flags & M_EXT) == 0) {
690 					m_free(n);
691 					n = NULL;
692 				}
693 			}
694 			if (!m_dup_pkthdr(n, m, MB_DONTWAIT)) {
695 				/*
696 				 * Previous code did a blind M_COPY_PKTHDR
697 				 * and said "just for rcvif".  If true, then
698 				 * we could tolerate the dup failing (due to
699 				 * the deep copy of the tag chain).  For now
700 				 * be conservative and just fail.
701 				 */
702 				m_free(n);
703 				n = NULL;
704 			}
705 			if (n == NULL) {
706 				/* Give up remote */
707 				break;
708 			}
709 			n->m_pkthdr.rcvif = NULL;
710 			n->m_len = 0;
711 			maxhlen = M_TRAILINGSPACE(n) - maxlen;
712 			if (maxhlen > hostnamelen)
713 				maxhlen = hostnamelen;
714 			/*
715 			 * Copy IPv6 and ICMPv6 only.
716 			 */
717 			nip6 = mtod(n, struct ip6_hdr *);
718 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
719 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
720 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
721 			p = (u_char *)(nicmp6 + 1);
722 			bzero(p, 4);
723 			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
724 			noff = sizeof(struct ip6_hdr);
725 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
726 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
727 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
728 			nicmp6->icmp6_code = 0;
729 		}
730 #undef hostnamelen
731 		if (n) {
732 			icmp6stat.icp6s_reflect++;
733 			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
734 			icmp6_reflect(n, noff);
735 		}
736 		break;
737 	    }
738 
739 	case ICMP6_WRUREPLY:
740 		if (code != 0)
741 			goto badcode;
742 		break;
743 
744 	case ND_ROUTER_SOLICIT:
745 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
746 		if (code != 0)
747 			goto badcode;
748 		if (icmp6len < sizeof(struct nd_router_solicit))
749 			goto badlen;
750 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
751 			/* give up local */
752 			nd6_rs_input(m, off, icmp6len);
753 			m = NULL;
754 			goto freeit;
755 		}
756 		nd6_rs_input(n, off, icmp6len);
757 		/* m stays. */
758 		break;
759 
760 	case ND_ROUTER_ADVERT:
761 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
762 		if (code != 0)
763 			goto badcode;
764 		if (icmp6len < sizeof(struct nd_router_advert))
765 			goto badlen;
766 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
767 			/* give up local */
768 			nd6_ra_input(m, off, icmp6len);
769 			m = NULL;
770 			goto freeit;
771 		}
772 		nd6_ra_input(n, off, icmp6len);
773 		/* m stays. */
774 		break;
775 
776 	case ND_NEIGHBOR_SOLICIT:
777 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
778 		if (code != 0)
779 			goto badcode;
780 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
781 			goto badlen;
782 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
783 			/* give up local */
784 			nd6_ns_input(m, off, icmp6len);
785 			m = NULL;
786 			goto freeit;
787 		}
788 		nd6_ns_input(n, off, icmp6len);
789 		/* m stays. */
790 		break;
791 
792 	case ND_NEIGHBOR_ADVERT:
793 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
794 		if (code != 0)
795 			goto badcode;
796 		if (icmp6len < sizeof(struct nd_neighbor_advert))
797 			goto badlen;
798 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
799 			/* give up local */
800 			nd6_na_input(m, off, icmp6len);
801 			m = NULL;
802 			goto freeit;
803 		}
804 		nd6_na_input(n, off, icmp6len);
805 		/* m stays. */
806 		break;
807 
808 	case ND_REDIRECT:
809 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
810 		if (code != 0)
811 			goto badcode;
812 		if (icmp6len < sizeof(struct nd_redirect))
813 			goto badlen;
814 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
815 			/* give up local */
816 			icmp6_redirect_input(m, off);
817 			m = NULL;
818 			goto freeit;
819 		}
820 		icmp6_redirect_input(n, off);
821 		/* m stays. */
822 		break;
823 
824 	case ICMP6_ROUTER_RENUMBERING:
825 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
826 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
827 			goto badcode;
828 		if (icmp6len < sizeof(struct icmp6_router_renum))
829 			goto badlen;
830 		break;
831 
832 	default:
833 		nd6log((LOG_DEBUG,
834 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
835 		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
836 		    ip6_sprintf(&ip6->ip6_dst),
837 		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
838 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
839 			/* ICMPv6 error: MUST deliver it by spec... */
840 			code = PRC_NCMDS;
841 			/* deliver */
842 		} else {
843 			/* ICMPv6 informational: MUST not deliver */
844 			break;
845 		}
846 	deliver:
847 		if (icmp6_notify_error(m, off, icmp6len, code)) {
848 			/* In this case, m should've been freed. */
849 			return(IPPROTO_DONE);
850 		}
851 		break;
852 
853 	badcode:
854 		icmp6stat.icp6s_badcode++;
855 		break;
856 
857 	badlen:
858 		icmp6stat.icp6s_badlen++;
859 		break;
860 	}
861 
862 	/* deliver the packet to appropriate sockets */
863 	icmp6_rip6_input(&m, *offp);
864 
865 	return IPPROTO_DONE;
866 
867  freeit:
868 	m_freem(m);
869 	return IPPROTO_DONE;
870 }
871 
872 static int
873 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
874 {
875 	struct icmp6_hdr *icmp6;
876 	struct ip6_hdr *eip6;
877 	u_int32_t notifymtu;
878 	struct sockaddr_in6 icmp6src, icmp6dst;
879 
880 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
881 		icmp6stat.icp6s_tooshort++;
882 		goto freeit;
883 	}
884 #ifndef PULLDOWN_TEST
885 	IP6_EXTHDR_CHECK(m, off,
886 			 sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr),
887 			 -1);
888 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
889 #else
890 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
891 		       sizeof(*icmp6) + sizeof(struct ip6_hdr));
892 	if (icmp6 == NULL) {
893 		icmp6stat.icp6s_tooshort++;
894 		return(-1);
895 	}
896 #endif
897 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
898 
899 	/* Detect the upper level protocol */
900 	{
901 		void (*ctlfunc) (int, struct sockaddr *, void *);
902 		u_int8_t nxt = eip6->ip6_nxt;
903 		int eoff = off + sizeof(struct icmp6_hdr) +
904 			sizeof(struct ip6_hdr);
905 		struct ip6ctlparam ip6cp;
906 		struct in6_addr *finaldst = NULL;
907 		int icmp6type = icmp6->icmp6_type;
908 		struct ip6_frag *fh;
909 		struct ip6_rthdr *rth;
910 		struct ip6_rthdr0 *rth0;
911 		int rthlen;
912 
913 		while (1) { /* XXX: should avoid infinite loop explicitly? */
914 			struct ip6_ext *eh;
915 
916 			switch (nxt) {
917 			case IPPROTO_HOPOPTS:
918 			case IPPROTO_DSTOPTS:
919 			case IPPROTO_AH:
920 #ifndef PULLDOWN_TEST
921 				IP6_EXTHDR_CHECK(m, 0, eoff +
922 						 sizeof(struct ip6_ext),
923 						 -1);
924 				eh = (struct ip6_ext *)(mtod(m, caddr_t)
925 							+ eoff);
926 #else
927 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
928 					       eoff, sizeof(*eh));
929 				if (eh == NULL) {
930 					icmp6stat.icp6s_tooshort++;
931 					return(-1);
932 				}
933 #endif
934 
935 				if (nxt == IPPROTO_AH)
936 					eoff += (eh->ip6e_len + 2) << 2;
937 				else
938 					eoff += (eh->ip6e_len + 1) << 3;
939 				nxt = eh->ip6e_nxt;
940 				break;
941 			case IPPROTO_ROUTING:
942 				/*
943 				 * When the erroneous packet contains a
944 				 * routing header, we should examine the
945 				 * header to determine the final destination.
946 				 * Otherwise, we can't properly update
947 				 * information that depends on the final
948 				 * destination (e.g. path MTU).
949 				 */
950 #ifndef PULLDOWN_TEST
951 				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth),
952 						 -1);
953 				rth = (struct ip6_rthdr *)(mtod(m, caddr_t)
954 							   + eoff);
955 #else
956 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
957 					       eoff, sizeof(*rth));
958 				if (rth == NULL) {
959 					icmp6stat.icp6s_tooshort++;
960 					return(-1);
961 				}
962 #endif
963 				rthlen = (rth->ip6r_len + 1) << 3;
964 				/*
965 				 * XXX: currently there is no
966 				 * officially defined type other
967 				 * than type-0.
968 				 * Note that if the segment left field
969 				 * is 0, all intermediate hops must
970 				 * have been passed.
971 				 */
972 				if (rth->ip6r_segleft &&
973 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
974 					int hops;
975 
976 #ifndef PULLDOWN_TEST
977 					IP6_EXTHDR_CHECK(m, 0, eoff + rthlen,
978 							 -1);
979 					rth0 = (struct ip6_rthdr0 *)(mtod(m, caddr_t) + eoff);
980 #else
981 					IP6_EXTHDR_GET(rth0,
982 						       struct ip6_rthdr0 *, m,
983 						       eoff, rthlen);
984 					if (rth0 == NULL) {
985 						icmp6stat.icp6s_tooshort++;
986 						return(-1);
987 					}
988 #endif
989 					/* just ignore a bogus header */
990 					if ((rth0->ip6r0_len % 2) == 0 &&
991 					    (hops = rth0->ip6r0_len/2))
992 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
993 				}
994 				eoff += rthlen;
995 				nxt = rth->ip6r_nxt;
996 				break;
997 			case IPPROTO_FRAGMENT:
998 #ifndef PULLDOWN_TEST
999 				IP6_EXTHDR_CHECK(m, 0, eoff +
1000 						 sizeof(struct ip6_frag),
1001 						 -1);
1002 				fh = (struct ip6_frag *)(mtod(m, caddr_t)
1003 							 + eoff);
1004 #else
1005 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1006 					       eoff, sizeof(*fh));
1007 				if (fh == NULL) {
1008 					icmp6stat.icp6s_tooshort++;
1009 					return(-1);
1010 				}
1011 #endif
1012 				/*
1013 				 * Data after a fragment header is meaningless
1014 				 * unless it is the first fragment, but
1015 				 * we'll go to the notify label for path MTU
1016 				 * discovery.
1017 				 */
1018 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1019 					goto notify;
1020 
1021 				eoff += sizeof(struct ip6_frag);
1022 				nxt = fh->ip6f_nxt;
1023 				break;
1024 			default:
1025 				/*
1026 				 * This case includes ESP and the No Next
1027 				 * Header.  In such cases going to the notify
1028 				 * label does not have any meaning
1029 				 * (i.e. ctlfunc will be NULL), but we go
1030 				 * anyway since we might have to update
1031 				 * path MTU information.
1032 				 */
1033 				goto notify;
1034 			}
1035 		}
1036 	  notify:
1037 #ifndef PULLDOWN_TEST
1038 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1039 #else
1040 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1041 			       sizeof(*icmp6) + sizeof(struct ip6_hdr));
1042 		if (icmp6 == NULL) {
1043 			icmp6stat.icp6s_tooshort++;
1044 			return(-1);
1045 		}
1046 #endif
1047 
1048 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1049 		bzero(&icmp6dst, sizeof(icmp6dst));
1050 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1051 		icmp6dst.sin6_family = AF_INET6;
1052 		if (finaldst == NULL)
1053 			icmp6dst.sin6_addr = eip6->ip6_dst;
1054 		else
1055 			icmp6dst.sin6_addr = *finaldst;
1056 		icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1057 							  &icmp6dst.sin6_addr);
1058 #ifndef SCOPEDROUTING
1059 		if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1060 				   NULL, NULL)) {
1061 			/* should be impossbile */
1062 			nd6log((LOG_DEBUG,
1063 			    "icmp6_notify_error: in6_embedscope failed\n"));
1064 			goto freeit;
1065 		}
1066 #endif
1067 
1068 		/*
1069 		 * retrieve parameters from the inner IPv6 header, and convert
1070 		 * them into sockaddr structures.
1071 		 */
1072 		bzero(&icmp6src, sizeof(icmp6src));
1073 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1074 		icmp6src.sin6_family = AF_INET6;
1075 		icmp6src.sin6_addr = eip6->ip6_src;
1076 		icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1077 							  &icmp6src.sin6_addr);
1078 #ifndef SCOPEDROUTING
1079 		if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1080 				   NULL, NULL)) {
1081 			/* should be impossbile */
1082 			nd6log((LOG_DEBUG,
1083 			    "icmp6_notify_error: in6_embedscope failed\n"));
1084 			goto freeit;
1085 		}
1086 #endif
1087 		icmp6src.sin6_flowinfo =
1088 			(eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1089 
1090 		if (finaldst == NULL)
1091 			finaldst = &eip6->ip6_dst;
1092 		ip6cp.ip6c_m = m;
1093 		ip6cp.ip6c_icmp6 = icmp6;
1094 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1095 		ip6cp.ip6c_off = eoff;
1096 		ip6cp.ip6c_finaldst = finaldst;
1097 		ip6cp.ip6c_src = &icmp6src;
1098 		ip6cp.ip6c_nxt = nxt;
1099 
1100 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1101 			notifymtu = ntohl(icmp6->icmp6_mtu);
1102 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1103 			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1104 		}
1105 
1106 		ctlfunc = (void (*) (int, struct sockaddr *, void *))
1107 			(inet6sw[ip6_protox[nxt]].pr_ctlinput);
1108 		if (ctlfunc) {
1109 			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1110 					  &ip6cp);
1111 		}
1112 	}
1113 	return(0);
1114 
1115   freeit:
1116 	m_freem(m);
1117 	return(-1);
1118 }
1119 
1120 void
1121 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1122 {
1123 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1124 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1125 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1126 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1127 	struct rtentry *rt = NULL;
1128 	struct sockaddr_in6 sin6;
1129 
1130 	if (!validated)
1131 		return;
1132 
1133 	bzero(&sin6, sizeof(sin6));
1134 	sin6.sin6_family = PF_INET6;
1135 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1136 	sin6.sin6_addr = *dst;
1137 	/* XXX normally, this won't happen */
1138 	if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1139 		sin6.sin6_addr.s6_addr16[1] =
1140 		    htons(m->m_pkthdr.rcvif->if_index);
1141 	}
1142 	/* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1143 	rt = rtalloc1((struct sockaddr *)&sin6, 0,
1144 		      RTF_CLONING | RTF_PRCLONING);
1145 
1146 	if (rt && (rt->rt_flags & RTF_HOST)
1147 	    && !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1148 		if (mtu < IPV6_MMTU) {
1149 				/* xxx */
1150 			rt->rt_rmx.rmx_locks |= RTV_MTU;
1151 		} else if (mtu < rt->rt_ifp->if_mtu &&
1152 			   rt->rt_rmx.rmx_mtu > mtu) {
1153 			icmp6stat.icp6s_pmtuchg++;
1154 			rt->rt_rmx.rmx_mtu = mtu;
1155 		}
1156 	}
1157 	if (rt) { /* XXX: need braces to avoid conflict with else in RTFREE. */
1158 		RTFREE(rt);
1159 	}
1160 }
1161 
1162 /*
1163  * Process a Node Information Query packet, based on
1164  * draft-ietf-ipngwg-icmp-name-lookups-07.
1165  *
1166  * Spec incompatibilities:
1167  * - IPv6 Subject address handling
1168  * - IPv4 Subject address handling support missing
1169  * - Proxy reply (answer even if it's not for me)
1170  * - joins NI group address at in6_ifattach() time only, does not cope
1171  *   with hostname changes by sethostname(3)
1172  */
1173 #define hostnamelen	strlen(hostname)
1174 static struct mbuf *
1175 ni6_input(struct mbuf *m, int off)
1176 {
1177 	struct icmp6_nodeinfo *ni6, *nni6;
1178 	struct mbuf *n = NULL;
1179 	u_int16_t qtype;
1180 	int subjlen;
1181 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1182 	struct ni_reply_fqdn *fqdn;
1183 	int addrs;		/* for NI_QTYPE_NODEADDR */
1184 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1185 	struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1186 	struct sockaddr_in6 sin6_d; /* XXX: we should retrieve this from m_aux */
1187 	struct ip6_hdr *ip6;
1188 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1189 	char *subj = NULL;
1190 	struct in6_ifaddr *ia6 = NULL;
1191 
1192 	ip6 = mtod(m, struct ip6_hdr *);
1193 #ifndef PULLDOWN_TEST
1194 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1195 #else
1196 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1197 	if (ni6 == NULL) {
1198 		/* m is already reclaimed */
1199 		return NULL;
1200 	}
1201 #endif
1202 
1203 	/*
1204 	 * Validate IPv6 destination address.
1205 	 *
1206 	 * The Responder must discard the Query without further processing
1207 	 * unless it is one of the Responder's unicast or anycast addresses, or
1208 	 * a link-local scope multicast address which the Responder has joined.
1209 	 * [icmp-name-lookups-07, Section 4.]
1210 	 */
1211 	bzero(&sin6, sizeof(sin6));
1212 	sin6.sin6_family = AF_INET6;
1213 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1214 	bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1215 	/* XXX scopeid */
1216 	if ((ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)&sin6)) != NULL) {
1217 		/* unicast/anycast, fine */
1218 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1219 		    (icmp6_nodeinfo & 4) == 0) {
1220 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1221 				"a temporary address in %s:%d",
1222 			       __FILE__, __LINE__));
1223 			goto bad;
1224 		}
1225 	} else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1226 		; /* link-local multicast, fine */
1227 	else
1228 		goto bad;
1229 
1230 	/* validate query Subject field. */
1231 	qtype = ntohs(ni6->ni_qtype);
1232 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1233 	switch (qtype) {
1234 	case NI_QTYPE_NOOP:
1235 	case NI_QTYPE_SUPTYPES:
1236 		/* 07 draft */
1237 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1238 			break;
1239 		/* FALLTHROUGH */
1240 	case NI_QTYPE_FQDN:
1241 	case NI_QTYPE_NODEADDR:
1242 		switch (ni6->ni_code) {
1243 		case ICMP6_NI_SUBJ_IPV6:
1244 #if ICMP6_NI_SUBJ_IPV6 != 0
1245 		case 0:
1246 #endif
1247 			/*
1248 			 * backward compatibility - try to accept 03 draft
1249 			 * format, where no Subject is present.
1250 			 */
1251 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1252 			    subjlen == 0) {
1253 				oldfqdn++;
1254 				break;
1255 			}
1256 #if ICMP6_NI_SUBJ_IPV6 != 0
1257 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1258 				goto bad;
1259 #endif
1260 
1261 			if (subjlen != sizeof(sin6.sin6_addr))
1262 				goto bad;
1263 
1264 			/*
1265 			 * Validate Subject address.
1266 			 *
1267 			 * Not sure what exactly "address belongs to the node"
1268 			 * means in the spec, is it just unicast, or what?
1269 			 *
1270 			 * At this moment we consider Subject address as
1271 			 * "belong to the node" if the Subject address equals
1272 			 * to the IPv6 destination address; validation for
1273 			 * IPv6 destination address should have done enough
1274 			 * check for us.
1275 			 *
1276 			 * We do not do proxy at this moment.
1277 			 */
1278 			/* m_pulldown instead of copy? */
1279 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1280 			    subjlen, (caddr_t)&sin6.sin6_addr);
1281 			sin6.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1282 							      &sin6.sin6_addr);
1283 #ifndef SCOPEDROUTING
1284 			in6_embedscope(&sin6.sin6_addr, &sin6, NULL, NULL);
1285 #endif
1286 			bzero(&sin6_d, sizeof(sin6_d));
1287 			sin6_d.sin6_family = AF_INET6; /* not used, actually */
1288 			sin6_d.sin6_len = sizeof(sin6_d); /* ditto */
1289 			sin6_d.sin6_addr = ip6->ip6_dst;
1290 			sin6_d.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1291 								&ip6->ip6_dst);
1292 #ifndef SCOPEDROUTING
1293 			in6_embedscope(&sin6_d.sin6_addr, &sin6_d, NULL, NULL);
1294 #endif
1295 			subj = (char *)&sin6;
1296 			if (SA6_ARE_ADDR_EQUAL(&sin6, &sin6_d))
1297 				break;
1298 
1299 			/*
1300 			 * XXX if we are to allow other cases, we should really
1301 			 * be careful about scope here.
1302 			 * basically, we should disallow queries toward IPv6
1303 			 * destination X with subject Y, if scope(X) > scope(Y).
1304 			 * if we allow scope(X) > scope(Y), it will result in
1305 			 * information leakage across scope boundary.
1306 			 */
1307 			goto bad;
1308 
1309 		case ICMP6_NI_SUBJ_FQDN:
1310 			/*
1311 			 * Validate Subject name with gethostname(3).
1312 			 *
1313 			 * The behavior may need some debate, since:
1314 			 * - we are not sure if the node has FQDN as
1315 			 *   hostname (returned by gethostname(3)).
1316 			 * - the code does wildcard match for truncated names.
1317 			 *   however, we are not sure if we want to perform
1318 			 *   wildcard match, if gethostname(3) side has
1319 			 *   truncated hostname.
1320 			 */
1321 			n = ni6_nametodns(hostname, hostnamelen, 0);
1322 			if (!n || n->m_next || n->m_len == 0)
1323 				goto bad;
1324 			IP6_EXTHDR_GET(subj, char *, m,
1325 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1326 			if (subj == NULL)
1327 				goto bad;
1328 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1329 					n->m_len)) {
1330 				goto bad;
1331 			}
1332 			m_freem(n);
1333 			n = NULL;
1334 			break;
1335 
1336 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1337 		default:
1338 			goto bad;
1339 		}
1340 		break;
1341 	}
1342 
1343 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1344 	switch (qtype) {
1345 	case NI_QTYPE_FQDN:
1346 		if ((icmp6_nodeinfo & 1) == 0)
1347 			goto bad;
1348 		break;
1349 	case NI_QTYPE_NODEADDR:
1350 		if ((icmp6_nodeinfo & 2) == 0)
1351 			goto bad;
1352 		break;
1353 	}
1354 
1355 	/* guess reply length */
1356 	switch (qtype) {
1357 	case NI_QTYPE_NOOP:
1358 		break;		/* no reply data */
1359 	case NI_QTYPE_SUPTYPES:
1360 		replylen += sizeof(u_int32_t);
1361 		break;
1362 	case NI_QTYPE_FQDN:
1363 		/* XXX will append an mbuf */
1364 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1365 		break;
1366 	case NI_QTYPE_NODEADDR:
1367 		addrs = ni6_addrs(ni6, m, &ifp, subj);
1368 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1369 					  sizeof(u_int32_t))) > MCLBYTES)
1370 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1371 		break;
1372 	default:
1373 		/*
1374 		 * XXX: We must return a reply with the ICMP6 code
1375 		 * `unknown Qtype' in this case. However we regard the case
1376 		 * as an FQDN query for backward compatibility.
1377 		 * Older versions set a random value to this field,
1378 		 * so it rarely varies in the defined qtypes.
1379 		 * But the mechanism is not reliable...
1380 		 * maybe we should obsolete older versions.
1381 		 */
1382 		qtype = NI_QTYPE_FQDN;
1383 		/* XXX will append an mbuf */
1384 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1385 		oldfqdn++;
1386 		break;
1387 	}
1388 
1389 	/* allocate an mbuf to reply. */
1390 	MGETHDR(n, MB_DONTWAIT, m->m_type);
1391 	if (n == NULL) {
1392 		m_freem(m);
1393 		return(NULL);
1394 	}
1395 	M_MOVE_PKTHDR(n, m); /* just for recvif */
1396 	if (replylen > MHLEN) {
1397 		if (replylen > MCLBYTES) {
1398 			/*
1399 			 * XXX: should we try to allocate more? But MCLBYTES
1400 			 * is probably much larger than IPV6_MMTU...
1401 			 */
1402 			goto bad;
1403 		}
1404 		MCLGET(n, MB_DONTWAIT);
1405 		if ((n->m_flags & M_EXT) == 0) {
1406 			goto bad;
1407 		}
1408 	}
1409 	n->m_pkthdr.len = n->m_len = replylen;
1410 
1411 	/* copy mbuf header and IPv6 + Node Information base headers */
1412 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1413 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1414 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1415 
1416 	/* qtype dependent procedure */
1417 	switch (qtype) {
1418 	case NI_QTYPE_NOOP:
1419 		nni6->ni_code = ICMP6_NI_SUCCESS;
1420 		nni6->ni_flags = 0;
1421 		break;
1422 	case NI_QTYPE_SUPTYPES:
1423 	{
1424 		u_int32_t v;
1425 		nni6->ni_code = ICMP6_NI_SUCCESS;
1426 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1427 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1428 		v = (u_int32_t)htonl(0x0000000f);
1429 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1430 		break;
1431 	}
1432 	case NI_QTYPE_FQDN:
1433 		nni6->ni_code = ICMP6_NI_SUCCESS;
1434 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1435 						sizeof(struct ip6_hdr) +
1436 						sizeof(struct icmp6_nodeinfo));
1437 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1438 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1439 		/*
1440 		 * XXX do we really have FQDN in variable "hostname"?
1441 		 */
1442 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1443 		if (n->m_next == NULL)
1444 			goto bad;
1445 		/* XXX we assume that n->m_next is not a chain */
1446 		if (n->m_next->m_next != NULL)
1447 			goto bad;
1448 		n->m_pkthdr.len += n->m_next->m_len;
1449 		break;
1450 	case NI_QTYPE_NODEADDR:
1451 	{
1452 		int lenlim, copied;
1453 
1454 		nni6->ni_code = ICMP6_NI_SUCCESS;
1455 		n->m_pkthdr.len = n->m_len =
1456 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1457 		lenlim = M_TRAILINGSPACE(n);
1458 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1459 		/* XXX: reset mbuf length */
1460 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1461 			sizeof(struct icmp6_nodeinfo) + copied;
1462 		break;
1463 	}
1464 	default:
1465 		break;		/* XXX impossible! */
1466 	}
1467 
1468 	nni6->ni_type = ICMP6_NI_REPLY;
1469 	m_freem(m);
1470 	return(n);
1471 
1472   bad:
1473 	m_freem(m);
1474 	if (n)
1475 		m_freem(n);
1476 	return(NULL);
1477 }
1478 #undef hostnamelen
1479 
1480 /*
1481  * make a mbuf with DNS-encoded string.  no compression support.
1482  *
1483  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1484  * treated as truncated name (two \0 at the end).  this is a wild guess.
1485  */
1486 static struct mbuf *
1487 ni6_nametodns(const char *name, int namelen,
1488 	      int old) /* return pascal string if non-zero */
1489 {
1490 	struct mbuf *m;
1491 	char *cp, *ep;
1492 	const char *p, *q;
1493 	int i, len, nterm;
1494 
1495 	if (old)
1496 		len = namelen + 1;
1497 	else
1498 		len = MCLBYTES;
1499 
1500 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1501 	MGET(m, MB_DONTWAIT, MT_DATA);
1502 	if (m && len > MLEN) {
1503 		MCLGET(m, MB_DONTWAIT);
1504 		if ((m->m_flags & M_EXT) == 0)
1505 			goto fail;
1506 	}
1507 	if (!m)
1508 		goto fail;
1509 	m->m_next = NULL;
1510 
1511 	if (old) {
1512 		m->m_len = len;
1513 		*mtod(m, char *) = namelen;
1514 		bcopy(name, mtod(m, char *) + 1, namelen);
1515 		return m;
1516 	} else {
1517 		m->m_len = 0;
1518 		cp = mtod(m, char *);
1519 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1520 
1521 		/* if not certain about my name, return empty buffer */
1522 		if (namelen == 0)
1523 			return m;
1524 
1525 		/*
1526 		 * guess if it looks like shortened hostname, or FQDN.
1527 		 * shortened hostname needs two trailing "\0".
1528 		 */
1529 		i = 0;
1530 		for (p = name; p < name + namelen; p++) {
1531 			if (*p && *p == '.')
1532 				i++;
1533 		}
1534 		if (i < 2)
1535 			nterm = 2;
1536 		else
1537 			nterm = 1;
1538 
1539 		p = name;
1540 		while (cp < ep && p < name + namelen) {
1541 			i = 0;
1542 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1543 				i++;
1544 			/* result does not fit into mbuf */
1545 			if (cp + i + 1 >= ep)
1546 				goto fail;
1547 			/*
1548 			 * DNS label length restriction, RFC1035 page 8.
1549 			 * "i == 0" case is included here to avoid returning
1550 			 * 0-length label on "foo..bar".
1551 			 */
1552 			if (i <= 0 || i >= 64)
1553 				goto fail;
1554 			*cp++ = i;
1555 			bcopy(p, cp, i);
1556 			cp += i;
1557 			p = q;
1558 			if (p < name + namelen && *p == '.')
1559 				p++;
1560 		}
1561 		/* termination */
1562 		if (cp + nterm >= ep)
1563 			goto fail;
1564 		while (nterm-- > 0)
1565 			*cp++ = '\0';
1566 		m->m_len = cp - mtod(m, char *);
1567 		return m;
1568 	}
1569 
1570 	panic("should not reach here");
1571 	/* NOTREACHED */
1572 
1573  fail:
1574 	if (m)
1575 		m_freem(m);
1576 	return NULL;
1577 }
1578 
1579 /*
1580  * check if two DNS-encoded string matches.  takes care of truncated
1581  * form (with \0\0 at the end).  no compression support.
1582  * XXX upper/lowercase match (see RFC2065)
1583  */
1584 static int
1585 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1586 {
1587 	const char *a0, *b0;
1588 	int l;
1589 
1590 	/* simplest case - need validation? */
1591 	if (alen == blen && bcmp(a, b, alen) == 0)
1592 		return 1;
1593 
1594 	a0 = a;
1595 	b0 = b;
1596 
1597 	/* termination is mandatory */
1598 	if (alen < 2 || blen < 2)
1599 		return 0;
1600 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1601 		return 0;
1602 	alen--;
1603 	blen--;
1604 
1605 	while (a - a0 < alen && b - b0 < blen) {
1606 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1607 			return 0;
1608 
1609 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1610 			return 0;
1611 		/* we don't support compression yet */
1612 		if (a[0] >= 64 || b[0] >= 64)
1613 			return 0;
1614 
1615 		/* truncated case */
1616 		if (a[0] == 0 && a - a0 == alen - 1)
1617 			return 1;
1618 		if (b[0] == 0 && b - b0 == blen - 1)
1619 			return 1;
1620 		if (a[0] == 0 || b[0] == 0)
1621 			return 0;
1622 
1623 		if (a[0] != b[0])
1624 			return 0;
1625 		l = a[0];
1626 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1627 			return 0;
1628 		if (bcmp(a + 1, b + 1, l) != 0)
1629 			return 0;
1630 
1631 		a += 1 + l;
1632 		b += 1 + l;
1633 	}
1634 
1635 	if (a - a0 == alen && b - b0 == blen)
1636 		return 1;
1637 	else
1638 		return 0;
1639 }
1640 
1641 /*
1642  * calculate the number of addresses to be returned in the node info reply.
1643  */
1644 static int
1645 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1646 	  char *subj)
1647 {
1648 	struct ifnet *ifp;
1649 	struct in6_ifaddr *ifa6;
1650 	struct ifaddr *ifa;
1651 	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1652 	int addrs = 0, addrsofif, iffound = 0;
1653 	int niflags = ni6->ni_flags;
1654 
1655 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1656 		switch (ni6->ni_code) {
1657 		case ICMP6_NI_SUBJ_IPV6:
1658 			if (subj == NULL) /* must be impossible... */
1659 				return(0);
1660 			subj_ip6 = (struct sockaddr_in6 *)subj;
1661 			break;
1662 		default:
1663 			/*
1664 			 * XXX: we only support IPv6 subject address for
1665 			 * this Qtype.
1666 			 */
1667 			return(0);
1668 		}
1669 	}
1670 
1671 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
1672 	{
1673 		addrsofif = 0;
1674 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list)
1675 		{
1676 			if (ifa->ifa_addr->sa_family != AF_INET6)
1677 				continue;
1678 			ifa6 = (struct in6_ifaddr *)ifa;
1679 
1680 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1681 			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1682 					       &ifa6->ia_addr.sin6_addr))
1683 				iffound = 1;
1684 
1685 			/*
1686 			 * IPv4-mapped addresses can only be returned by a
1687 			 * Node Information proxy, since they represent
1688 			 * addresses of IPv4-only nodes, which perforce do
1689 			 * not implement this protocol.
1690 			 * [icmp-name-lookups-07, Section 5.4]
1691 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1692 			 * this function at this moment.
1693 			 */
1694 
1695 			/* What do we have to do about ::1? */
1696 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1697 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1698 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1699 					continue;
1700 				break;
1701 			case IPV6_ADDR_SCOPE_SITELOCAL:
1702 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1703 					continue;
1704 				break;
1705 			case IPV6_ADDR_SCOPE_GLOBAL:
1706 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1707 					continue;
1708 				break;
1709 			default:
1710 				continue;
1711 			}
1712 
1713 			/*
1714 			 * check if anycast is okay.
1715 			 * XXX: just experimental.  not in the spec.
1716 			 */
1717 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1718 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1719 				continue; /* we need only unicast addresses */
1720 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1721 			    (icmp6_nodeinfo & 4) == 0) {
1722 				continue;
1723 			}
1724 			addrsofif++; /* count the address */
1725 		}
1726 		if (iffound) {
1727 			*ifpp = ifp;
1728 			return(addrsofif);
1729 		}
1730 
1731 		addrs += addrsofif;
1732 	}
1733 
1734 	return(addrs);
1735 }
1736 
1737 static int
1738 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1739 		struct ifnet *ifp0, int resid)
1740 {
1741 	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1742 	struct in6_ifaddr *ifa6;
1743 	struct ifaddr *ifa;
1744 	struct ifnet *ifp_dep = NULL;
1745 	int copied = 0, allow_deprecated = 0;
1746 	u_char *cp = (u_char *)(nni6 + 1);
1747 	int niflags = ni6->ni_flags;
1748 	u_int32_t ltime;
1749 
1750 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1751 		return(0);	/* needless to copy */
1752 
1753   again:
1754 
1755 	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list))
1756 	{
1757 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1758 			if (ifa->ifa_addr->sa_family != AF_INET6)
1759 				continue;
1760 			ifa6 = (struct in6_ifaddr *)ifa;
1761 
1762 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1763 			    allow_deprecated == 0) {
1764 				/*
1765 				 * prefererred address should be put before
1766 				 * deprecated addresses.
1767 				 */
1768 
1769 				/* record the interface for later search */
1770 				if (ifp_dep == NULL)
1771 					ifp_dep = ifp;
1772 
1773 				continue;
1774 			}
1775 			else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1776 				 allow_deprecated != 0)
1777 				continue; /* we now collect deprecated addrs */
1778 
1779 			/* What do we have to do about ::1? */
1780 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1781 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1782 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1783 					continue;
1784 				break;
1785 			case IPV6_ADDR_SCOPE_SITELOCAL:
1786 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1787 					continue;
1788 				break;
1789 			case IPV6_ADDR_SCOPE_GLOBAL:
1790 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1791 					continue;
1792 				break;
1793 			default:
1794 				continue;
1795 			}
1796 
1797 			/*
1798 			 * check if anycast is okay.
1799 			 * XXX: just experimental. not in the spec.
1800 			 */
1801 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1802 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1803 				continue;
1804 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1805 			    (icmp6_nodeinfo & 4) == 0) {
1806 				continue;
1807 			}
1808 
1809 			/* now we can copy the address */
1810 			if (resid < sizeof(struct in6_addr) +
1811 			    sizeof(u_int32_t)) {
1812 				/*
1813 				 * We give up much more copy.
1814 				 * Set the truncate flag and return.
1815 				 */
1816 				nni6->ni_flags |=
1817 					NI_NODEADDR_FLAG_TRUNCATE;
1818 				return(copied);
1819 			}
1820 
1821 			/*
1822 			 * Set the TTL of the address.
1823 			 * The TTL value should be one of the following
1824 			 * according to the specification:
1825 			 *
1826 			 * 1. The remaining lifetime of a DHCP lease on the
1827 			 *    address, or
1828 			 * 2. The remaining Valid Lifetime of a prefix from
1829 			 *    which the address was derived through Stateless
1830 			 *    Autoconfiguration.
1831 			 *
1832 			 * Note that we currently do not support stateful
1833 			 * address configuration by DHCPv6, so the former
1834 			 * case can't happen.
1835 			 */
1836 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1837 				ltime = ND6_INFINITE_LIFETIME;
1838 			else {
1839 				if (ifa6->ia6_lifetime.ia6t_expire >
1840 				    time_second)
1841 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1842 				else
1843 					ltime = 0;
1844 			}
1845 
1846 			bcopy(&ltime, cp, sizeof(u_int32_t));
1847 			cp += sizeof(u_int32_t);
1848 
1849 			/* copy the address itself */
1850 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1851 			      sizeof(struct in6_addr));
1852 			/* XXX: KAME link-local hack; remove ifindex */
1853 			if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr))
1854 				((struct in6_addr *)cp)->s6_addr16[1] = 0;
1855 			cp += sizeof(struct in6_addr);
1856 
1857 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1858 			copied += (sizeof(struct in6_addr) +
1859 				   sizeof(u_int32_t));
1860 		}
1861 		if (ifp0)	/* we need search only on the specified IF */
1862 			break;
1863 	}
1864 
1865 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1866 		ifp = ifp_dep;
1867 		allow_deprecated = 1;
1868 
1869 		goto again;
1870 	}
1871 
1872 	return(copied);
1873 }
1874 
1875 /*
1876  * XXX almost dup'ed code with rip6_input.
1877  */
1878 static int
1879 icmp6_rip6_input(struct	mbuf **mp, int	off)
1880 {
1881 	struct mbuf *m = *mp;
1882 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1883 	struct in6pcb *in6p;
1884 	struct in6pcb *last = NULL;
1885 	struct sockaddr_in6 rip6src;
1886 	struct icmp6_hdr *icmp6;
1887 	struct mbuf *opts = NULL;
1888 
1889 #ifndef PULLDOWN_TEST
1890 	/* this is assumed to be safe. */
1891 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1892 #else
1893 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1894 	if (icmp6 == NULL) {
1895 		/* m is already reclaimed */
1896 		return IPPROTO_DONE;
1897 	}
1898 #endif
1899 
1900 	bzero(&rip6src, sizeof(rip6src));
1901 	rip6src.sin6_len = sizeof(struct sockaddr_in6);
1902 	rip6src.sin6_family = AF_INET6;
1903 	/* KAME hack: recover scopeid */
1904 	(void)in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif);
1905 
1906 	LIST_FOREACH(in6p, &ripcbinfo.pcblisthead, inp_list)
1907 	{
1908 		if (in6p->inp_flags & INP_PLACEMARKER)
1909 			continue;
1910 		if ((in6p->inp_vflag & INP_IPV6) == 0)
1911 			continue;
1912 #ifdef HAVE_NRL_INPCB
1913 		if (!(in6p->in6p_flags & INP_IPV6))
1914 			continue;
1915 #endif
1916 		if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1917 			continue;
1918 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1919 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1920 			continue;
1921 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1922 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1923 			continue;
1924 		if (in6p->in6p_icmp6filt
1925 		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1926 				 in6p->in6p_icmp6filt))
1927 			continue;
1928 		if (last) {
1929 			struct	mbuf *n;
1930 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1931 				if (last->in6p_flags & IN6P_CONTROLOPTS)
1932 					ip6_savecontrol(last, &opts, ip6, n);
1933 				/* strip intermediate headers */
1934 				m_adj(n, off);
1935 				if (sbappendaddr(&last->in6p_socket->so_rcv,
1936 						 (struct sockaddr *)&rip6src,
1937 						 n, opts) == 0) {
1938 					/* should notify about lost packet */
1939 					m_freem(n);
1940 					if (opts) {
1941 						m_freem(opts);
1942 					}
1943 				} else
1944 					sorwakeup(last->in6p_socket);
1945 				opts = NULL;
1946 			}
1947 		}
1948 		last = in6p;
1949 	}
1950 	if (last) {
1951 		if (last->in6p_flags & IN6P_CONTROLOPTS)
1952 			ip6_savecontrol(last, &opts, ip6, m);
1953 		/* strip intermediate headers */
1954 		m_adj(m, off);
1955 		if (sbappendaddr(&last->in6p_socket->so_rcv,
1956 				 (struct sockaddr *)&rip6src, m, opts) == 0) {
1957 			m_freem(m);
1958 			if (opts)
1959 				m_freem(opts);
1960 		} else
1961 			sorwakeup(last->in6p_socket);
1962 	} else {
1963 		m_freem(m);
1964 		ip6stat.ip6s_delivered--;
1965 	}
1966 	return IPPROTO_DONE;
1967 }
1968 
1969 /*
1970  * Reflect the ip6 packet back to the source.
1971  * OFF points to the icmp6 header, counted from the top of the mbuf.
1972  */
1973 void
1974 icmp6_reflect(struct mbuf *m, size_t off)
1975 {
1976 	struct ip6_hdr *ip6;
1977 	struct icmp6_hdr *icmp6;
1978 	struct in6_ifaddr *ia;
1979 	struct in6_addr t, *src = 0;
1980 	int plen;
1981 	int type, code;
1982 	struct ifnet *outif = NULL;
1983 	struct sockaddr_in6 sa6_src, sa6_dst;
1984 #ifdef COMPAT_RFC1885
1985 	int mtu = IPV6_MMTU;
1986 	struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
1987 #endif
1988 
1989 	/* too short to reflect */
1990 	if (off < sizeof(struct ip6_hdr)) {
1991 		nd6log((LOG_DEBUG,
1992 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
1993 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
1994 		    __FILE__, __LINE__));
1995 		goto bad;
1996 	}
1997 
1998 	/*
1999 	 * If there are extra headers between IPv6 and ICMPv6, strip
2000 	 * off that header first.
2001 	 */
2002 #ifdef DIAGNOSTIC
2003 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2004 		panic("assumption failed in icmp6_reflect");
2005 #endif
2006 	if (off > sizeof(struct ip6_hdr)) {
2007 		size_t l;
2008 		struct ip6_hdr nip6;
2009 
2010 		l = off - sizeof(struct ip6_hdr);
2011 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2012 		m_adj(m, l);
2013 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2014 		if (m->m_len < l) {
2015 			if ((m = m_pullup(m, l)) == NULL)
2016 				return;
2017 		}
2018 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2019 	} else /* off == sizeof(struct ip6_hdr) */ {
2020 		size_t l;
2021 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2022 		if (m->m_len < l) {
2023 			if ((m = m_pullup(m, l)) == NULL)
2024 				return;
2025 		}
2026 	}
2027 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2028 	ip6 = mtod(m, struct ip6_hdr *);
2029 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2030 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2031 	type = icmp6->icmp6_type; /* keep type for statistics */
2032 	code = icmp6->icmp6_code; /* ditto. */
2033 
2034 	t = ip6->ip6_dst;
2035 	/*
2036 	 * ip6_input() drops a packet if its src is multicast.
2037 	 * So, the src is never multicast.
2038 	 */
2039 	ip6->ip6_dst = ip6->ip6_src;
2040 
2041 	/*
2042 	 * XXX: make sure to embed scope zone information, using
2043 	 * already embedded IDs or the received interface (if any).
2044 	 * Note that rcvif may be NULL.
2045 	 * TODO: scoped routing case (XXX).
2046 	 */
2047 	bzero(&sa6_src, sizeof(sa6_src));
2048 	sa6_src.sin6_family = AF_INET6;
2049 	sa6_src.sin6_len = sizeof(sa6_src);
2050 	sa6_src.sin6_addr = ip6->ip6_dst;
2051 	in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
2052 	in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL);
2053 	bzero(&sa6_dst, sizeof(sa6_dst));
2054 	sa6_dst.sin6_family = AF_INET6;
2055 	sa6_dst.sin6_len = sizeof(sa6_dst);
2056 	sa6_dst.sin6_addr = t;
2057 	in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
2058 	in6_embedscope(&t, &sa6_dst, NULL, NULL);
2059 
2060 #ifdef COMPAT_RFC1885
2061 	/*
2062 	 * xxx guess MTU
2063 	 * RFC 1885 requires that echo reply should be truncated if it
2064 	 * does not fit in with (return) path MTU, but the description was
2065 	 * removed in the new spec.
2066 	 */
2067 	if (icmp6_reflect_rt.ro_rt == 0 ||
2068 	    ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
2069 		if (icmp6_reflect_rt.ro_rt) {
2070 			RTFREE(icmp6_reflect_rt.ro_rt);
2071 			icmp6_reflect_rt.ro_rt = 0;
2072 		}
2073 		bzero(sin6, sizeof(*sin6));
2074 		sin6->sin6_family = PF_INET6;
2075 		sin6->sin6_len = sizeof(struct sockaddr_in6);
2076 		sin6->sin6_addr = ip6->ip6_dst;
2077 
2078 		rtalloc_ign((struct route *)&icmp6_reflect_rt.ro_rt,
2079 			    RTF_PRCLONING);
2080 	}
2081 
2082 	if (icmp6_reflect_rt.ro_rt == 0)
2083 		goto bad;
2084 
2085 	if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
2086 	    && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
2087 		mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
2088 
2089 	if (mtu < m->m_pkthdr.len) {
2090 		plen -= (m->m_pkthdr.len - mtu);
2091 		m_adj(m, mtu - m->m_pkthdr.len);
2092 	}
2093 #endif
2094 	/*
2095 	 * If the incoming packet was addressed directly to us(i.e. unicast),
2096 	 * use dst as the src for the reply.
2097 	 * The IN6_IFF_NOTREADY case would be VERY rare, but is possible
2098 	 * (for example) when we encounter an error while forwarding procedure
2099 	 * destined to a duplicated address of ours.
2100 	 */
2101 	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2102 		if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2103 		    (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2104 			src = &t;
2105 			break;
2106 		}
2107 	if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2108 		/*
2109 		 * This is the case if the dst is our link-local address
2110 		 * and the sender is also ourselves.
2111 		 */
2112 		src = &t;
2113 	}
2114 
2115 	if (src == 0) {
2116 		int e;
2117 		struct route_in6 ro;
2118 
2119 		/*
2120 		 * This case matches to multicasts, our anycast, or unicasts
2121 		 * that we do not own.  Select a source address based on the
2122 		 * source address of the erroneous packet.
2123 		 */
2124 		bzero(&ro, sizeof(ro));
2125 		src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e);
2126 		if (ro.ro_rt)
2127 			RTFREE(ro.ro_rt); /* XXX: we could use this */
2128 		if (src == NULL) {
2129 			nd6log((LOG_DEBUG,
2130 			    "icmp6_reflect: source can't be determined: "
2131 			    "dst=%s, error=%d\n",
2132 			    ip6_sprintf(&sa6_src.sin6_addr), e));
2133 			goto bad;
2134 		}
2135 	}
2136 
2137 	ip6->ip6_src = *src;
2138 
2139 	ip6->ip6_flow = 0;
2140 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2141 	ip6->ip6_vfc |= IPV6_VERSION;
2142 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2143 	if (m->m_pkthdr.rcvif) {
2144 		/* XXX: This may not be the outgoing interface */
2145 		ip6->ip6_hlim = nd_ifinfo[m->m_pkthdr.rcvif->if_index].chlim;
2146 	} else
2147 		ip6->ip6_hlim = ip6_defhlim;
2148 
2149 	icmp6->icmp6_cksum = 0;
2150 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2151 					sizeof(struct ip6_hdr), plen);
2152 
2153 	/*
2154 	 * XXX option handling
2155 	 */
2156 
2157 	m->m_flags &= ~(M_BCAST|M_MCAST);
2158 
2159 #ifdef COMPAT_RFC1885
2160 	ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif, NULL);
2161 #else
2162 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2163 #endif
2164 	if (outif)
2165 		icmp6_ifoutstat_inc(outif, type, code);
2166 
2167 	return;
2168 
2169  bad:
2170 	m_freem(m);
2171 	return;
2172 }
2173 
2174 void
2175 icmp6_fasttimo(void)
2176 {
2177 
2178 	mld6_fasttimeo();
2179 }
2180 
2181 static const char *
2182 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2183 		    struct in6_addr *tgt6)
2184 {
2185 	static char buf[1024];
2186 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2187 		ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2188 	return buf;
2189 }
2190 
2191 void
2192 icmp6_redirect_input(struct mbuf *m, int off)
2193 {
2194 	struct ifnet *ifp = m->m_pkthdr.rcvif;
2195 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2196 	struct nd_redirect *nd_rd;
2197 	int icmp6len = ntohs(ip6->ip6_plen);
2198 	char *lladdr = NULL;
2199 	int lladdrlen = 0;
2200 	u_char *redirhdr = NULL;
2201 	int redirhdrlen = 0;
2202 	struct rtentry *rt = NULL;
2203 	int is_router;
2204 	int is_onlink;
2205 	struct in6_addr src6 = ip6->ip6_src;
2206 	struct in6_addr redtgt6;
2207 	struct in6_addr reddst6;
2208 	union nd_opts ndopts;
2209 
2210 	if (!m || !ifp)
2211 		return;
2212 
2213 	/* XXX if we are router, we don't update route by icmp6 redirect */
2214 	if (ip6_forwarding)
2215 		goto freeit;
2216 	if (!icmp6_rediraccept)
2217 		goto freeit;
2218 
2219 #ifndef PULLDOWN_TEST
2220 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2221 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2222 #else
2223 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2224 	if (nd_rd == NULL) {
2225 		icmp6stat.icp6s_tooshort++;
2226 		return;
2227 	}
2228 #endif
2229 	redtgt6 = nd_rd->nd_rd_target;
2230 	reddst6 = nd_rd->nd_rd_dst;
2231 
2232 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2233 		redtgt6.s6_addr16[1] = htons(ifp->if_index);
2234 	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2235 		reddst6.s6_addr16[1] = htons(ifp->if_index);
2236 
2237 	/* validation */
2238 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2239 		nd6log((LOG_ERR,
2240 			"ICMP6 redirect sent from %s rejected; "
2241 			"must be from linklocal\n", ip6_sprintf(&src6)));
2242 		goto bad;
2243 	}
2244 	if (ip6->ip6_hlim != 255) {
2245 		nd6log((LOG_ERR,
2246 			"ICMP6 redirect sent from %s rejected; "
2247 			"hlim=%d (must be 255)\n",
2248 			ip6_sprintf(&src6), ip6->ip6_hlim));
2249 		goto bad;
2250 	}
2251     {
2252 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2253 	struct sockaddr_in6 sin6;
2254 	struct in6_addr *gw6;
2255 
2256 	bzero(&sin6, sizeof(sin6));
2257 	sin6.sin6_family = AF_INET6;
2258 	sin6.sin6_len = sizeof(struct sockaddr_in6);
2259 	bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2260 	rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL);
2261 	if (rt) {
2262 		if (rt->rt_gateway == NULL ||
2263 		    rt->rt_gateway->sa_family != AF_INET6) {
2264 			nd6log((LOG_ERR,
2265 			    "ICMP6 redirect rejected; no route "
2266 			    "with inet6 gateway found for redirect dst: %s\n",
2267 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2268 			RTFREE(rt);
2269 			goto bad;
2270 		}
2271 
2272 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2273 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2274 			nd6log((LOG_ERR,
2275 				"ICMP6 redirect rejected; "
2276 				"not equal to gw-for-src=%s (must be same): "
2277 				"%s\n",
2278 				ip6_sprintf(gw6),
2279 				icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2280 			RTFREE(rt);
2281 			goto bad;
2282 		}
2283 	} else {
2284 		nd6log((LOG_ERR,
2285 			"ICMP6 redirect rejected; "
2286 			"no route found for redirect dst: %s\n",
2287 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2288 		goto bad;
2289 	}
2290 	RTFREE(rt);
2291 	rt = NULL;
2292     }
2293 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2294 		nd6log((LOG_ERR,
2295 			"ICMP6 redirect rejected; "
2296 			"redirect dst must be unicast: %s\n",
2297 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2298 		goto bad;
2299 	}
2300 
2301 	is_router = is_onlink = 0;
2302 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2303 		is_router = 1;	/* router case */
2304 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2305 		is_onlink = 1;	/* on-link destination case */
2306 	if (!is_router && !is_onlink) {
2307 		nd6log((LOG_ERR,
2308 			"ICMP6 redirect rejected; "
2309 			"neither router case nor onlink case: %s\n",
2310 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2311 		goto bad;
2312 	}
2313 	/* validation passed */
2314 
2315 	icmp6len -= sizeof(*nd_rd);
2316 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2317 	if (nd6_options(&ndopts) < 0) {
2318 		nd6log((LOG_INFO, "icmp6_redirect_input: "
2319 			"invalid ND option, rejected: %s\n",
2320 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2321 		/* nd6_options have incremented stats */
2322 		goto freeit;
2323 	}
2324 
2325 	if (ndopts.nd_opts_tgt_lladdr) {
2326 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2327 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2328 	}
2329 
2330 	if (ndopts.nd_opts_rh) {
2331 		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2332 		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2333 	}
2334 
2335 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2336 		nd6log((LOG_INFO,
2337 			"icmp6_redirect_input: lladdrlen mismatch for %s "
2338 			"(if %d, icmp6 packet %d): %s\n",
2339 			ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2340 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2341 		goto bad;
2342 	}
2343 
2344 	/* RFC 2461 8.3 */
2345 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2346 			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2347 
2348 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2349 		/* perform rtredirect */
2350 		struct sockaddr_in6 sdst;
2351 		struct sockaddr_in6 sgw;
2352 		struct sockaddr_in6 ssrc;
2353 
2354 		bzero(&sdst, sizeof(sdst));
2355 		bzero(&sgw, sizeof(sgw));
2356 		bzero(&ssrc, sizeof(ssrc));
2357 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2358 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2359 			sizeof(struct sockaddr_in6);
2360 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2361 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2362 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2363 		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2364 			   (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2365 			   (struct sockaddr *)&ssrc,
2366 			   (struct rtentry **)NULL);
2367 	}
2368 	/* finally update cached route in each socket via pfctlinput */
2369     {
2370 	struct sockaddr_in6 sdst;
2371 
2372 	bzero(&sdst, sizeof(sdst));
2373 	sdst.sin6_family = AF_INET6;
2374 	sdst.sin6_len = sizeof(struct sockaddr_in6);
2375 	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2376 	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2377 #ifdef IPSEC
2378 	key_sa_routechange((struct sockaddr *)&sdst);
2379 #endif
2380     }
2381 
2382  freeit:
2383 	m_freem(m);
2384 	return;
2385 
2386  bad:
2387 	icmp6stat.icp6s_badredirect++;
2388 	m_freem(m);
2389 }
2390 
2391 void
2392 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2393 {
2394 	struct ifnet *ifp;	/* my outgoing interface */
2395 	struct in6_addr *ifp_ll6;
2396 	struct in6_addr *router_ll6;
2397 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2398 	struct mbuf *m = NULL;	/* newly allocated one */
2399 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2400 	struct nd_redirect *nd_rd;
2401 	size_t maxlen;
2402 	u_char *p;
2403 	struct ifnet *outif = NULL;
2404 	struct sockaddr_in6 src_sa;
2405 
2406 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2407 
2408 	/* if we are not router, we don't send icmp6 redirect */
2409 	if (!ip6_forwarding || ip6_accept_rtadv)
2410 		goto fail;
2411 
2412 	/* sanity check */
2413 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2414 		goto fail;
2415 
2416 	/*
2417 	 * Address check:
2418 	 *  the source address must identify a neighbor, and
2419 	 *  the destination address must not be a multicast address
2420 	 *  [RFC 2461, sec 8.2]
2421 	 */
2422 	sip6 = mtod(m0, struct ip6_hdr *);
2423 	bzero(&src_sa, sizeof(src_sa));
2424 	src_sa.sin6_family = AF_INET6;
2425 	src_sa.sin6_len = sizeof(src_sa);
2426 	src_sa.sin6_addr = sip6->ip6_src;
2427 	/* we don't currently use sin6_scope_id, but eventually use it */
2428 	src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src);
2429 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2430 		goto fail;
2431 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2432 		goto fail;	/* what should we do here? */
2433 
2434 	/* rate limit */
2435 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2436 		goto fail;
2437 
2438 	/*
2439 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2440 	 * we almost always ask for an mbuf cluster for simplicity.
2441 	 * (MHLEN < IPV6_MMTU is almost always true)
2442 	 */
2443 #if IPV6_MMTU >= MCLBYTES
2444 # error assumption failed about IPV6_MMTU and MCLBYTES
2445 #endif
2446 	MGETHDR(m, MB_DONTWAIT, MT_HEADER);
2447 	if (m && IPV6_MMTU >= MHLEN)
2448 		MCLGET(m, MB_DONTWAIT);
2449 	if (!m)
2450 		goto fail;
2451 	m->m_pkthdr.rcvif = NULL;
2452 	m->m_len = 0;
2453 	maxlen = M_TRAILINGSPACE(m);
2454 	maxlen = min(IPV6_MMTU, maxlen);
2455 	/* just for safety */
2456 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2457 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2458 		goto fail;
2459 	}
2460 
2461 	{
2462 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2463 		struct in6_ifaddr *ia;
2464 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2465 						 IN6_IFF_NOTREADY|
2466 						 IN6_IFF_ANYCAST)) == NULL)
2467 			goto fail;
2468 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2469 	}
2470 
2471 	/* get ip6 linklocal address for the router. */
2472 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2473 		struct sockaddr_in6 *sin6;
2474 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2475 		router_ll6 = &sin6->sin6_addr;
2476 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2477 			router_ll6 = (struct in6_addr *)NULL;
2478 	} else
2479 		router_ll6 = (struct in6_addr *)NULL;
2480 
2481 	/* ip6 */
2482 	ip6 = mtod(m, struct ip6_hdr *);
2483 	ip6->ip6_flow = 0;
2484 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2485 	ip6->ip6_vfc |= IPV6_VERSION;
2486 	/* ip6->ip6_plen will be set later */
2487 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2488 	ip6->ip6_hlim = 255;
2489 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2490 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2491 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2492 
2493 	/* ND Redirect */
2494 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2495 	nd_rd->nd_rd_type = ND_REDIRECT;
2496 	nd_rd->nd_rd_code = 0;
2497 	nd_rd->nd_rd_reserved = 0;
2498 	if (rt->rt_flags & RTF_GATEWAY) {
2499 		/*
2500 		 * nd_rd->nd_rd_target must be a link-local address in
2501 		 * better router cases.
2502 		 */
2503 		if (!router_ll6)
2504 			goto fail;
2505 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2506 		      sizeof(nd_rd->nd_rd_target));
2507 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2508 		      sizeof(nd_rd->nd_rd_dst));
2509 	} else {
2510 		/* make sure redtgt == reddst */
2511 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2512 		      sizeof(nd_rd->nd_rd_target));
2513 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2514 		      sizeof(nd_rd->nd_rd_dst));
2515 	}
2516 
2517 	p = (u_char *)(nd_rd + 1);
2518 
2519 	if (!router_ll6)
2520 		goto nolladdropt;
2521 
2522     {
2523 	/* target lladdr option */
2524 	struct rtentry *rt_router = NULL;
2525 	int len;
2526 	struct sockaddr_dl *sdl;
2527 	struct nd_opt_hdr *nd_opt;
2528 	char *lladdr;
2529 
2530 	rt_router = nd6_lookup(router_ll6, 0, ifp);
2531 	if (!rt_router)
2532 		goto nolladdropt;
2533 	len = sizeof(*nd_opt) + ifp->if_addrlen;
2534 	len = (len + 7) & ~7;	/* round by 8 */
2535 	/* safety check */
2536 	if (len + (p - (u_char *)ip6) > maxlen)
2537 		goto nolladdropt;
2538 	if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2539 	    (rt_router->rt_flags & RTF_LLINFO) &&
2540 	    (rt_router->rt_gateway->sa_family == AF_LINK) &&
2541 	    (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2542 	    sdl->sdl_alen) {
2543 		nd_opt = (struct nd_opt_hdr *)p;
2544 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2545 		nd_opt->nd_opt_len = len >> 3;
2546 		lladdr = (char *)(nd_opt + 1);
2547 		bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2548 		p += len;
2549 	}
2550     }
2551 nolladdropt:;
2552 
2553 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2554 
2555 	/* just to be safe */
2556 #ifdef M_DECRYPTED	/*not openbsd*/
2557 	if (m0->m_flags & M_DECRYPTED)
2558 		goto noredhdropt;
2559 #endif
2560 	if (p - (u_char *)ip6 > maxlen)
2561 		goto noredhdropt;
2562 
2563     {
2564 	/* redirected header option */
2565 	int len;
2566 	struct nd_opt_rd_hdr *nd_opt_rh;
2567 
2568 	/*
2569 	 * compute the maximum size for icmp6 redirect header option.
2570 	 * XXX room for auth header?
2571 	 */
2572 	len = maxlen - (p - (u_char *)ip6);
2573 	len &= ~7;
2574 
2575 	/* This is just for simplicity. */
2576 	if (m0->m_pkthdr.len != m0->m_len) {
2577 		if (m0->m_next) {
2578 			m_freem(m0->m_next);
2579 			m0->m_next = NULL;
2580 		}
2581 		m0->m_pkthdr.len = m0->m_len;
2582 	}
2583 
2584 	/*
2585 	 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2586 	 * about padding/truncate rule for the original IP packet.
2587 	 * From the discussion on IPv6imp in Feb 1999, the consensus was:
2588 	 * - "attach as much as possible" is the goal
2589 	 * - pad if not aligned (original size can be guessed by original
2590 	 *   ip6 header)
2591 	 * Following code adds the padding if it is simple enough,
2592 	 * and truncates if not.
2593 	 */
2594 	if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2595 		panic("assumption failed in %s:%d", __FILE__, __LINE__);
2596 
2597 	if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2598 		/* not enough room, truncate */
2599 		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2600 	} else {
2601 		/* enough room, pad or truncate */
2602 		size_t extra;
2603 
2604 		extra = m0->m_pkthdr.len % 8;
2605 		if (extra) {
2606 			/* pad if easy enough, truncate if not */
2607 			if (8 - extra <= M_TRAILINGSPACE(m0)) {
2608 				/* pad */
2609 				m0->m_len += (8 - extra);
2610 				m0->m_pkthdr.len += (8 - extra);
2611 			} else {
2612 				/* truncate */
2613 				m0->m_pkthdr.len -= extra;
2614 				m0->m_len -= extra;
2615 			}
2616 		}
2617 		len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2618 		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2619 	}
2620 
2621 	nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2622 	bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2623 	nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2624 	nd_opt_rh->nd_opt_rh_len = len >> 3;
2625 	p += sizeof(*nd_opt_rh);
2626 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2627 
2628 	/* connect m0 to m */
2629 	m->m_next = m0;
2630 	m->m_pkthdr.len = m->m_len + m0->m_len;
2631     }
2632 noredhdropt:;
2633 
2634 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
2635 		sip6->ip6_src.s6_addr16[1] = 0;
2636 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
2637 		sip6->ip6_dst.s6_addr16[1] = 0;
2638 #if 0
2639 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
2640 		ip6->ip6_src.s6_addr16[1] = 0;
2641 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
2642 		ip6->ip6_dst.s6_addr16[1] = 0;
2643 #endif
2644 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
2645 		nd_rd->nd_rd_target.s6_addr16[1] = 0;
2646 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
2647 		nd_rd->nd_rd_dst.s6_addr16[1] = 0;
2648 
2649 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2650 
2651 	nd_rd->nd_rd_cksum = 0;
2652 	nd_rd->nd_rd_cksum
2653 		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2654 
2655 	/* send the packet to outside... */
2656 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2657 	if (outif) {
2658 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2659 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2660 	}
2661 	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2662 
2663 	return;
2664 
2665 fail:
2666 	if (m)
2667 		m_freem(m);
2668 	if (m0)
2669 		m_freem(m0);
2670 }
2671 
2672 #ifdef HAVE_NRL_INPCB
2673 #define sotoin6pcb	sotoinpcb
2674 #define in6pcb		inpcb
2675 #define in6p_icmp6filt	inp_icmp6filt
2676 #endif
2677 /*
2678  * ICMPv6 socket option processing.
2679  */
2680 int
2681 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2682 {
2683 	int error = 0;
2684 	int optlen;
2685 	struct inpcb *inp = sotoinpcb(so);
2686 	int level, op, optname;
2687 
2688 	if (sopt) {
2689 		level = sopt->sopt_level;
2690 		op = sopt->sopt_dir;
2691 		optname = sopt->sopt_name;
2692 		optlen = sopt->sopt_valsize;
2693 	} else
2694 		level = op = optname = optlen = 0;
2695 
2696 	if (level != IPPROTO_ICMPV6) {
2697 		return EINVAL;
2698 	}
2699 
2700 	switch (op) {
2701 	case PRCO_SETOPT:
2702 		switch (optname) {
2703 		case ICMP6_FILTER:
2704 		    {
2705 			struct icmp6_filter *p;
2706 
2707 			if (optlen != sizeof(*p)) {
2708 				error = EMSGSIZE;
2709 				break;
2710 			}
2711 			if (inp->in6p_icmp6filt == NULL) {
2712 				error = EINVAL;
2713 				break;
2714 			}
2715 			error = sooptcopyin(sopt, inp->in6p_icmp6filt, optlen,
2716 				optlen);
2717 			break;
2718 		    }
2719 
2720 		default:
2721 			error = ENOPROTOOPT;
2722 			break;
2723 		}
2724 		break;
2725 
2726 	case PRCO_GETOPT:
2727 		switch (optname) {
2728 		case ICMP6_FILTER:
2729 		    {
2730 			if (inp->in6p_icmp6filt == NULL) {
2731 				error = EINVAL;
2732 				break;
2733 			}
2734 			error = sooptcopyout(sopt, inp->in6p_icmp6filt,
2735 				sizeof(struct icmp6_filter));
2736 			break;
2737 		    }
2738 
2739 		default:
2740 			error = ENOPROTOOPT;
2741 			break;
2742 		}
2743 		break;
2744 	}
2745 
2746 	return(error);
2747 }
2748 #ifdef HAVE_NRL_INPCB
2749 #undef sotoin6pcb
2750 #undef in6pcb
2751 #undef in6p_icmp6filt
2752 #endif
2753 
2754 #ifndef HAVE_PPSRATECHECK
2755 #ifndef timersub
2756 #define	timersub(tvp, uvp, vvp)						\
2757 	do {								\
2758 		(vvp)->tv_sec = (tvp)->tv_sec - (uvp)->tv_sec;		\
2759 		(vvp)->tv_usec = (tvp)->tv_usec - (uvp)->tv_usec;	\
2760 		if ((vvp)->tv_usec < 0) {				\
2761 			(vvp)->tv_sec--;				\
2762 			(vvp)->tv_usec += 1000000;			\
2763 		}							\
2764 	} while (0)
2765 #endif
2766 
2767 /*
2768  * ppsratecheck(): packets (or events) per second limitation.
2769  */
2770 static int
2771 ppsratecheck(struct timeval *lasttime, int *curpps,
2772 	     int maxpps)	/* maximum pps allowed */
2773 {
2774 	struct timeval tv, delta;
2775 	int rv;
2776 
2777 	microtime(&tv);
2778 
2779 	timersub(&tv, lasttime, &delta);
2780 
2781 	/*
2782 	 * Check for 0,0 so that the message will be seen at least once.
2783 	 * If more than one second has passed since the last update of
2784 	 * lasttime, reset the counter.
2785 	 *
2786 	 * We do increment *curpps even in *curpps < maxpps case, as some may
2787 	 * try to use *curpps for stat purposes as well.
2788 	 */
2789 	if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) ||
2790 	    delta.tv_sec >= 1) {
2791 		*lasttime = tv;
2792 		*curpps = 0;
2793 		rv = 1;
2794 	} else if (maxpps < 0)
2795 		rv = 1;
2796 	else if (*curpps < maxpps)
2797 		rv = 1;
2798 	else
2799 		rv = 0;
2800 
2801 #if 1 /* DIAGNOSTIC? */
2802 	/* be careful about wrap-around */
2803 	if (*curpps + 1 > *curpps)
2804 		*curpps = *curpps + 1;
2805 #else
2806 	/*
2807 	 * assume that there's not too many calls to this function.
2808 	 * not sure if the assumption holds, as it depends on *caller's*
2809 	 * behavior, not the behavior of this function.
2810 	 * IMHO it is wrong to make assumption on the caller's behavior,
2811 	 * so the above #if is #if 1, not #ifdef DIAGNOSTIC.
2812 	 */
2813 	*curpps = *curpps + 1;
2814 #endif
2815 
2816 	return (rv);
2817 }
2818 #endif
2819 
2820 /*
2821  * Perform rate limit check.
2822  * Returns 0 if it is okay to send the icmp6 packet.
2823  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2824  * limitation.
2825  *
2826  * XXX per-destination/type check necessary?
2827  */
2828 static int
2829 icmp6_ratelimit(const struct in6_addr *dst,	/* not used at this moment */
2830 		const int type,			/* not used at this moment */
2831 		const int code)			/* not used at this moment */
2832 {
2833 	int ret;
2834 
2835 	ret = 0;	/* okay to send */
2836 
2837 	/* PPS limit */
2838 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2839 	    icmp6errppslim)) {
2840 		/* The packet is subject to rate limit */
2841 		ret++;
2842 	}
2843 
2844 	return ret;
2845 }
2846