xref: /dragonfly/sys/netinet6/icmp6.c (revision c311ab13)
1 /*	$FreeBSD: src/sys/netinet6/icmp6.c,v 1.6.2.13 2003/05/06 06:46:58 suz Exp $	*/
2 /*	$KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1988, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 4. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
62  */
63 
64 #include "opt_inet.h"
65 #include "opt_inet6.h"
66 #include "opt_ipsec.h"
67 
68 #include <sys/param.h>
69 #include <sys/systm.h>
70 #include <sys/malloc.h>
71 #include <sys/mbuf.h>
72 #include <sys/protosw.h>
73 #include <sys/socket.h>
74 #include <sys/socketvar.h>
75 #include <sys/socketops.h>
76 #include <sys/time.h>
77 #include <sys/kernel.h>
78 #include <sys/syslog.h>
79 #include <sys/domain.h>
80 
81 #include <sys/thread2.h>
82 #include <sys/msgport2.h>
83 
84 #include <net/if.h>
85 #include <net/route.h>
86 #include <net/if_dl.h>
87 #include <net/if_types.h>
88 
89 #include <netinet/in.h>
90 #include <netinet/in_var.h>
91 #include <netinet/ip6.h>
92 #include <netinet6/ip6_var.h>
93 #include <netinet/icmp6.h>
94 #include <netinet6/mld6_var.h>
95 #include <netinet/in_pcb.h>
96 #include <netinet6/in6_pcb.h>
97 #include <netinet6/nd6.h>
98 #include <netinet6/in6_ifattach.h>
99 #include <netinet6/ip6protosw.h>
100 
101 #ifdef IPSEC
102 #include <netinet6/ipsec.h>
103 #include <netproto/key/key.h>
104 #endif
105 
106 #ifdef FAST_IPSEC
107 #include <netproto/ipsec/ipsec.h>
108 #include <netproto/ipsec/key.h>
109 #define	IPSEC
110 #endif
111 
112 #include <net/net_osdep.h>
113 
114 #ifdef HAVE_NRL_INPCB
115 /* inpcb members */
116 #define in6pcb		inpcb
117 #define in6p_laddr	inp_laddr6
118 #define in6p_faddr	inp_faddr6
119 #define in6p_icmp6filt	inp_icmp6filt
120 #define in6p_route	inp_route
121 #define in6p_socket	inp_socket
122 #define in6p_flags	inp_flags
123 #define in6p_moptions	inp_moptions6
124 #define in6p_outputopts	inp_outputopts6
125 #define in6p_ip6	inp_ipv6
126 #define in6p_flowinfo	inp_flowinfo
127 #define in6p_sp		inp_sp
128 #define in6p_next	inp_next
129 #define in6p_prev	inp_prev
130 /* function names */
131 #define in6_pcbdetach	in_pcbdetach
132 #define in6_rtchange	in_rtchange
133 
134 /*
135  * for KAME src sync over BSD*'s. XXX: FreeBSD (>=3) are VERY different from
136  * others...
137  */
138 #define in6p_ip6_nxt	inp_ipv6.ip6_nxt
139 #endif
140 
141 /*
142  * ppratecheck() is available, so define it here.
143  */
144 #define	HAVE_PPSRATECHECK
145 
146 extern struct domain inet6domain;
147 extern struct protosw inet6sw[];
148 extern u_char ip6_protox[];
149 
150 struct icmp6stat icmp6stat;
151 
152 extern struct inpcbinfo ripcbinfo;
153 extern int icmp6errppslim;
154 static int icmp6errpps_count = 0;
155 static struct timeval icmp6errppslim_last;
156 extern int icmp6_nodeinfo;
157 
158 static void icmp6_errcount (struct icmp6errstat *, int, int);
159 static int icmp6_rip6_input (struct mbuf **, int);
160 static int icmp6_ratelimit (const struct in6_addr *, const int, const int);
161 static const char *icmp6_redirect_diag (struct in6_addr *,
162 	struct in6_addr *, struct in6_addr *);
163 #ifndef HAVE_PPSRATECHECK
164 static int ppsratecheck (struct timeval *, int *, int);
165 #endif
166 static struct mbuf *ni6_input (struct mbuf *, int);
167 static struct mbuf *ni6_nametodns (const char *, int, int);
168 static int ni6_dnsmatch (const char *, int, const char *, int);
169 static int ni6_addrs (struct icmp6_nodeinfo *, struct mbuf *,
170 			  struct ifnet **, char *);
171 static int ni6_store_addrs (struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
172 				struct ifnet *, int);
173 static int icmp6_notify_error (struct mbuf *, int, int, int);
174 
175 #ifdef COMPAT_RFC1885
176 static struct route_in6 icmp6_reflect_rt;
177 #endif
178 
179 
180 void
181 icmp6_init(void)
182 {
183 	mld6_init();
184 }
185 
186 static void
187 icmp6_errcount(struct icmp6errstat *stat, int type, int code)
188 {
189 	switch (type) {
190 	case ICMP6_DST_UNREACH:
191 		switch (code) {
192 		case ICMP6_DST_UNREACH_NOROUTE:
193 			stat->icp6errs_dst_unreach_noroute++;
194 			return;
195 		case ICMP6_DST_UNREACH_ADMIN:
196 			stat->icp6errs_dst_unreach_admin++;
197 			return;
198 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
199 			stat->icp6errs_dst_unreach_beyondscope++;
200 			return;
201 		case ICMP6_DST_UNREACH_ADDR:
202 			stat->icp6errs_dst_unreach_addr++;
203 			return;
204 		case ICMP6_DST_UNREACH_NOPORT:
205 			stat->icp6errs_dst_unreach_noport++;
206 			return;
207 		}
208 		break;
209 	case ICMP6_PACKET_TOO_BIG:
210 		stat->icp6errs_packet_too_big++;
211 		return;
212 	case ICMP6_TIME_EXCEEDED:
213 		switch (code) {
214 		case ICMP6_TIME_EXCEED_TRANSIT:
215 			stat->icp6errs_time_exceed_transit++;
216 			return;
217 		case ICMP6_TIME_EXCEED_REASSEMBLY:
218 			stat->icp6errs_time_exceed_reassembly++;
219 			return;
220 		}
221 		break;
222 	case ICMP6_PARAM_PROB:
223 		switch (code) {
224 		case ICMP6_PARAMPROB_HEADER:
225 			stat->icp6errs_paramprob_header++;
226 			return;
227 		case ICMP6_PARAMPROB_NEXTHEADER:
228 			stat->icp6errs_paramprob_nextheader++;
229 			return;
230 		case ICMP6_PARAMPROB_OPTION:
231 			stat->icp6errs_paramprob_option++;
232 			return;
233 		}
234 		break;
235 	case ND_REDIRECT:
236 		stat->icp6errs_redirect++;
237 		return;
238 	}
239 	stat->icp6errs_unknown++;
240 }
241 
242 /*
243  * Generate an error packet of type error in response to bad IP6 packet.
244  */
245 void
246 icmp6_error(struct mbuf *m, int type, int code, int param)
247 {
248 	struct ip6_hdr *oip6, *nip6;
249 	struct icmp6_hdr *icmp6;
250 	u_int preplen;
251 	int off;
252 	int nxt;
253 
254 	icmp6stat.icp6s_error++;
255 
256 	/* count per-type-code statistics */
257 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
258 
259 #ifdef M_DECRYPTED	/*not openbsd*/
260 	if (m->m_flags & M_DECRYPTED) {
261 		icmp6stat.icp6s_canterror++;
262 		goto freeit;
263 	}
264 #endif
265 
266 #ifndef PULLDOWN_TEST
267 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
268 #else
269 	if (m->m_len < sizeof(struct ip6_hdr)) {
270 		m = m_pullup(m, sizeof(struct ip6_hdr));
271 		if (m == NULL)
272 			return;
273 	}
274 #endif
275 	oip6 = mtod(m, struct ip6_hdr *);
276 
277 	/*
278 	 * Multicast destination check. For unrecognized option errors,
279 	 * this check has already done in ip6_unknown_opt(), so we can
280 	 * check only for other errors.
281 	 */
282 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
283 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
284 	    (type != ICMP6_PACKET_TOO_BIG &&
285 	     (type != ICMP6_PARAM_PROB ||
286 	      code != ICMP6_PARAMPROB_OPTION)))
287 		goto freeit;
288 
289 	/* Source address check. XXX: the case of anycast source? */
290 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
291 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
292 		goto freeit;
293 
294 	/*
295 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
296 	 * don't do it.
297 	 */
298 	nxt = -1;
299 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
300 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
301 		struct icmp6_hdr *icp;
302 
303 #ifndef PULLDOWN_TEST
304 		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
305 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
306 #else
307 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
308 			sizeof(*icp));
309 		if (icp == NULL) {
310 			icmp6stat.icp6s_tooshort++;
311 			return;
312 		}
313 #endif
314 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
315 		    icp->icmp6_type == ND_REDIRECT) {
316 			/*
317 			 * ICMPv6 error
318 			 * Special case: for redirect (which is
319 			 * informational) we must not send icmp6 error.
320 			 */
321 			icmp6stat.icp6s_canterror++;
322 			goto freeit;
323 		} else {
324 			/* ICMPv6 informational - send the error */
325 		}
326 	} else {
327 		/* non-ICMPv6 - send the error */
328 	}
329 
330 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
331 
332 	/* Finally, do rate limitation check. */
333 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
334 		icmp6stat.icp6s_toofreq++;
335 		goto freeit;
336 	}
337 
338 	/*
339 	 * OK, ICMP6 can be generated.
340 	 */
341 
342 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
343 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
344 
345 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
346 	M_PREPEND(m, preplen, MB_DONTWAIT);
347 	if (m && m->m_len < preplen)
348 		m = m_pullup(m, preplen);
349 	if (m == NULL) {
350 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
351 		return;
352 	}
353 
354 	nip6 = mtod(m, struct ip6_hdr *);
355 	nip6->ip6_src  = oip6->ip6_src;
356 	nip6->ip6_dst  = oip6->ip6_dst;
357 
358 	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src))
359 		oip6->ip6_src.s6_addr16[1] = 0;
360 	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst))
361 		oip6->ip6_dst.s6_addr16[1] = 0;
362 
363 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
364 	icmp6->icmp6_type = type;
365 	icmp6->icmp6_code = code;
366 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
367 
368 	/*
369 	 * icmp6_reflect() is designed to be in the input path.
370 	 * icmp6_error() can be called from both input and outut path,
371 	 * and if we are in output path rcvif could contain bogus value.
372 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
373 	 * information in ip header (nip6).
374 	 */
375 	m->m_pkthdr.rcvif = NULL;
376 
377 	icmp6stat.icp6s_outhist[type]++;
378 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
379 
380 	return;
381 
382 freeit:
383 	/*
384 	 * If we can't tell wheter or not we can generate ICMP6, free it.
385 	 */
386 	m_freem(m);
387 }
388 
389 /*
390  * Process a received ICMP6 message.
391  */
392 int
393 icmp6_input(struct mbuf **mp, int *offp, int proto)
394 {
395 	struct mbuf *m = *mp, *n;
396 	struct ip6_hdr *ip6, *nip6;
397 	struct icmp6_hdr *icmp6, *nicmp6;
398 	int off = *offp;
399 	int icmp6len = m->m_pkthdr.len - *offp;
400 	int code, sum, noff;
401 
402 #ifndef PULLDOWN_TEST
403 	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
404 	/* m might change if M_LOOP.  So, call mtod after this */
405 #endif
406 
407 	/*
408 	 * Locate icmp6 structure in mbuf, and check
409 	 * that not corrupted and of at least minimum length
410 	 */
411 
412 	ip6 = mtod(m, struct ip6_hdr *);
413 	if (icmp6len < sizeof(struct icmp6_hdr)) {
414 		icmp6stat.icp6s_tooshort++;
415 		goto freeit;
416 	}
417 
418 	/*
419 	 * calculate the checksum
420 	 */
421 #ifndef PULLDOWN_TEST
422 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
423 #else
424 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
425 	if (icmp6 == NULL) {
426 		icmp6stat.icp6s_tooshort++;
427 		return IPPROTO_DONE;
428 	}
429 #endif
430 	code = icmp6->icmp6_code;
431 
432 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
433 		nd6log((LOG_ERR,
434 		    "ICMP6 checksum error(%d|%x) %s\n",
435 		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
436 		icmp6stat.icp6s_checksum++;
437 		goto freeit;
438 	}
439 
440 	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
441 		/*
442 		 * Deliver very specific ICMP6 type only.
443 		 * This is important to deilver TOOBIG.  Otherwise PMTUD
444 		 * will not work.
445 		 */
446 		switch (icmp6->icmp6_type) {
447 		case ICMP6_DST_UNREACH:
448 		case ICMP6_PACKET_TOO_BIG:
449 		case ICMP6_TIME_EXCEEDED:
450 			break;
451 		default:
452 			goto freeit;
453 		}
454 	}
455 
456 	icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
457 	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
458 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
459 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
460 
461 	switch (icmp6->icmp6_type) {
462 	case ICMP6_DST_UNREACH:
463 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
464 		switch (code) {
465 		case ICMP6_DST_UNREACH_NOROUTE:
466 			code = PRC_UNREACH_NET;
467 			break;
468 		case ICMP6_DST_UNREACH_ADMIN:
469 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
470 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
471 			break;
472 		case ICMP6_DST_UNREACH_ADDR:
473 			code = PRC_HOSTDEAD;
474 			break;
475 #ifdef COMPAT_RFC1885
476 		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
477 			code = PRC_UNREACH_SRCFAIL;
478 			break;
479 #else
480 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
481 			/* I mean "source address was incorrect." */
482 			code = PRC_PARAMPROB;
483 			break;
484 #endif
485 		case ICMP6_DST_UNREACH_NOPORT:
486 			code = PRC_UNREACH_PORT;
487 			break;
488 		default:
489 			goto badcode;
490 		}
491 		goto deliver;
492 		break;
493 
494 	case ICMP6_PACKET_TOO_BIG:
495 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
496 		if (code != 0)
497 			goto badcode;
498 
499 		code = PRC_MSGSIZE;
500 
501 		/*
502 		 * Updating the path MTU will be done after examining
503 		 * intermediate extension headers.
504 		 */
505 		goto deliver;
506 		break;
507 
508 	case ICMP6_TIME_EXCEEDED:
509 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
510 		switch (code) {
511 		case ICMP6_TIME_EXCEED_TRANSIT:
512 		case ICMP6_TIME_EXCEED_REASSEMBLY:
513 			code += PRC_TIMXCEED_INTRANS;
514 			break;
515 		default:
516 			goto badcode;
517 		}
518 		goto deliver;
519 		break;
520 
521 	case ICMP6_PARAM_PROB:
522 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
523 		switch (code) {
524 		case ICMP6_PARAMPROB_NEXTHEADER:
525 			code = PRC_UNREACH_PROTOCOL;
526 			break;
527 		case ICMP6_PARAMPROB_HEADER:
528 		case ICMP6_PARAMPROB_OPTION:
529 			code = PRC_PARAMPROB;
530 			break;
531 		default:
532 			goto badcode;
533 		}
534 		goto deliver;
535 		break;
536 
537 	case ICMP6_ECHO_REQUEST:
538 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
539 		if (code != 0)
540 			goto badcode;
541 		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
542 			/* Give up remote */
543 			break;
544 		}
545 		if ((n->m_flags & M_EXT) ||
546 		    n->m_len < off + sizeof(struct icmp6_hdr)) {
547 			struct mbuf *n0 = n;
548 			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
549 			int n0len;
550 
551 			/*
552 			 * Prepare an internal mbuf. m_pullup() doesn't
553 			 * always copy the length we specified.
554 			 */
555 			if (maxlen >= MCLBYTES) {
556 				/* Give up remote */
557 				m_freem(n0);
558 				break;
559 			}
560 			n = m_getb(maxlen, MB_DONTWAIT, n0->m_type, M_PKTHDR);
561 			if (n == NULL) {
562 				/* Give up remote */
563 				m_freem(n0);
564 				break;
565 			}
566 			n0len = n0->m_pkthdr.len;	/* save for use below */
567 			M_MOVE_PKTHDR(n, n0);
568 			/*
569 			 * Copy IPv6 and ICMPv6 only.
570 			 */
571 			nip6 = mtod(n, struct ip6_hdr *);
572 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
573 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
574 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
575 			noff = sizeof(struct ip6_hdr);
576 			/* new mbuf contains only ipv6+icmpv6 headers */
577 			n->m_len = noff + sizeof(struct icmp6_hdr);
578 			/*
579 			 * Adjust mbuf. ip6_plen will be adjusted in
580 			 * ip6_output().
581 			 */
582 			m_adj(n0, off + sizeof(struct icmp6_hdr));
583 			/* recalculate complete packet size */
584 			n->m_pkthdr.len = n0len + (noff - off);
585 			n->m_next = n0;
586 		} else {
587 			nip6 = mtod(n, struct ip6_hdr *);
588 			nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
589 			noff = off;
590 		}
591 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
592 		nicmp6->icmp6_code = 0;
593 		if (n) {
594 			icmp6stat.icp6s_reflect++;
595 			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
596 			icmp6_reflect(n, noff);
597 		}
598 		break;
599 
600 	case ICMP6_ECHO_REPLY:
601 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
602 		if (code != 0)
603 			goto badcode;
604 		break;
605 
606 	case MLD_LISTENER_QUERY:
607 	case MLD_LISTENER_REPORT:
608 		if (icmp6len < sizeof(struct mld_hdr))
609 			goto badlen;
610 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
611 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
612 		else
613 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
614 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
615 			/* give up local */
616 			mld6_input(m, off);
617 			m = NULL;
618 			goto freeit;
619 		}
620 		mld6_input(n, off);
621 		/* m stays. */
622 		break;
623 
624 	case MLD_LISTENER_DONE:
625 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
626 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
627 			goto badlen;
628 		break;		/* nothing to be done in kernel */
629 
630 	case MLD_MTRACE_RESP:
631 	case MLD_MTRACE:
632 		/* XXX: these two are experimental.  not officially defind. */
633 		/* XXX: per-interface statistics? */
634 		break;		/* just pass it to applications */
635 
636 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
637 	    {
638 		enum { WRU, FQDN } mode;
639 
640 		if (!icmp6_nodeinfo)
641 			break;
642 
643 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
644 			mode = WRU;
645 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
646 			mode = FQDN;
647 		else
648 			goto badlen;
649 
650 #define hostnamelen	strlen(hostname)
651 		if (mode == FQDN) {
652 #ifndef PULLDOWN_TEST
653 			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
654 					 IPPROTO_DONE);
655 #endif
656 			n = m_copy(m, 0, M_COPYALL);
657 			if (n)
658 				n = ni6_input(n, off);
659 			/* XXX meaningless if n == NULL */
660 			noff = sizeof(struct ip6_hdr);
661 		} else {
662 			u_char *p;
663 			int maxlen, maxhlen;
664 
665 			if ((icmp6_nodeinfo & 5) != 5)
666 				break;
667 
668 			if (code != 0)
669 				goto badcode;
670 			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
671 			if (maxlen >= MCLBYTES) {
672 				/* Give up remote */
673 				break;
674 			}
675 			n = m_getb(maxlen, MB_DONTWAIT, m->m_type, M_PKTHDR);
676 			if (n == NULL) {
677 				/* Give up remote */
678 				break;
679 			}
680 			if (!m_dup_pkthdr(n, m, MB_DONTWAIT)) {
681 				/*
682 				 * Previous code did a blind M_COPY_PKTHDR
683 				 * and said "just for rcvif".  If true, then
684 				 * we could tolerate the dup failing (due to
685 				 * the deep copy of the tag chain).  For now
686 				 * be conservative and just fail.
687 				 */
688 				m_free(n);
689 				break;
690 			}
691 			n->m_len = 0;
692 			maxhlen = M_TRAILINGSPACE(n) - maxlen;
693 			if (maxhlen > hostnamelen)
694 				maxhlen = hostnamelen;
695 			/*
696 			 * Copy IPv6 and ICMPv6 only.
697 			 */
698 			nip6 = mtod(n, struct ip6_hdr *);
699 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
700 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
701 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
702 			p = (u_char *)(nicmp6 + 1);
703 			bzero(p, 4);
704 			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
705 			noff = sizeof(struct ip6_hdr);
706 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
707 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
708 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
709 			nicmp6->icmp6_code = 0;
710 		}
711 #undef hostnamelen
712 		if (n) {
713 			icmp6stat.icp6s_reflect++;
714 			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
715 			icmp6_reflect(n, noff);
716 		}
717 		break;
718 	    }
719 
720 	case ICMP6_WRUREPLY:
721 		if (code != 0)
722 			goto badcode;
723 		break;
724 
725 	case ND_ROUTER_SOLICIT:
726 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
727 		if (code != 0)
728 			goto badcode;
729 		if (icmp6len < sizeof(struct nd_router_solicit))
730 			goto badlen;
731 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
732 			/* give up local */
733 			nd6_rs_input(m, off, icmp6len);
734 			m = NULL;
735 			goto freeit;
736 		}
737 		nd6_rs_input(n, off, icmp6len);
738 		/* m stays. */
739 		break;
740 
741 	case ND_ROUTER_ADVERT:
742 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
743 		if (code != 0)
744 			goto badcode;
745 		if (icmp6len < sizeof(struct nd_router_advert))
746 			goto badlen;
747 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
748 			/* give up local */
749 			nd6_ra_input(m, off, icmp6len);
750 			m = NULL;
751 			goto freeit;
752 		}
753 		nd6_ra_input(n, off, icmp6len);
754 		/* m stays. */
755 		break;
756 
757 	case ND_NEIGHBOR_SOLICIT:
758 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
759 		if (code != 0)
760 			goto badcode;
761 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
762 			goto badlen;
763 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
764 			/* give up local */
765 			nd6_ns_input(m, off, icmp6len);
766 			m = NULL;
767 			goto freeit;
768 		}
769 		nd6_ns_input(n, off, icmp6len);
770 		/* m stays. */
771 		break;
772 
773 	case ND_NEIGHBOR_ADVERT:
774 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
775 		if (code != 0)
776 			goto badcode;
777 		if (icmp6len < sizeof(struct nd_neighbor_advert))
778 			goto badlen;
779 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
780 			/* give up local */
781 			nd6_na_input(m, off, icmp6len);
782 			m = NULL;
783 			goto freeit;
784 		}
785 		nd6_na_input(n, off, icmp6len);
786 		/* m stays. */
787 		break;
788 
789 	case ND_REDIRECT:
790 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
791 		if (code != 0)
792 			goto badcode;
793 		if (icmp6len < sizeof(struct nd_redirect))
794 			goto badlen;
795 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
796 			/* give up local */
797 			icmp6_redirect_input(m, off);
798 			m = NULL;
799 			goto freeit;
800 		}
801 		icmp6_redirect_input(n, off);
802 		/* m stays. */
803 		break;
804 
805 	case ICMP6_ROUTER_RENUMBERING:
806 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
807 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
808 			goto badcode;
809 		if (icmp6len < sizeof(struct icmp6_router_renum))
810 			goto badlen;
811 		break;
812 
813 	default:
814 		nd6log((LOG_DEBUG,
815 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
816 		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
817 		    ip6_sprintf(&ip6->ip6_dst),
818 		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
819 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
820 			/* ICMPv6 error: MUST deliver it by spec... */
821 			code = PRC_NCMDS;
822 			/* deliver */
823 		} else {
824 			/* ICMPv6 informational: MUST not deliver */
825 			break;
826 		}
827 	deliver:
828 		if (icmp6_notify_error(m, off, icmp6len, code)) {
829 			/* In this case, m should've been freed. */
830 			return (IPPROTO_DONE);
831 		}
832 		break;
833 
834 	badcode:
835 		icmp6stat.icp6s_badcode++;
836 		break;
837 
838 	badlen:
839 		icmp6stat.icp6s_badlen++;
840 		break;
841 	}
842 
843 	/* deliver the packet to appropriate sockets */
844 	icmp6_rip6_input(&m, *offp);
845 
846 	return IPPROTO_DONE;
847 
848 freeit:
849 	m_freem(m);
850 	return IPPROTO_DONE;
851 }
852 
853 static int
854 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
855 {
856 	struct icmp6_hdr *icmp6;
857 	struct ip6_hdr *eip6;
858 	u_int32_t notifymtu;
859 	struct sockaddr_in6 icmp6src, icmp6dst;
860 
861 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
862 		icmp6stat.icp6s_tooshort++;
863 		goto freeit;
864 	}
865 #ifndef PULLDOWN_TEST
866 	IP6_EXTHDR_CHECK(m, off,
867 			 sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr),
868 			 -1);
869 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
870 #else
871 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
872 		       sizeof(*icmp6) + sizeof(struct ip6_hdr));
873 	if (icmp6 == NULL) {
874 		icmp6stat.icp6s_tooshort++;
875 		return (-1);
876 	}
877 #endif
878 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
879 
880 	/* Detect the upper level protocol */
881 	{
882 		u_int8_t nxt = eip6->ip6_nxt;
883 		int eoff = off + sizeof(struct icmp6_hdr) +
884 			sizeof(struct ip6_hdr);
885 		struct ip6ctlparam ip6cp;
886 		struct in6_addr *finaldst = NULL;
887 		int icmp6type = icmp6->icmp6_type;
888 		struct ip6_frag *fh;
889 		struct ip6_rthdr *rth;
890 		int rthlen;
891 
892 		while (1) { /* XXX: should avoid infinite loop explicitly? */
893 			struct ip6_ext *eh;
894 
895 			switch (nxt) {
896 			case IPPROTO_HOPOPTS:
897 			case IPPROTO_DSTOPTS:
898 			case IPPROTO_AH:
899 #ifndef PULLDOWN_TEST
900 				IP6_EXTHDR_CHECK(m, 0, eoff +
901 						 sizeof(struct ip6_ext),
902 						 -1);
903 				eh = (struct ip6_ext *)(mtod(m, caddr_t)
904 							+ eoff);
905 #else
906 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
907 					       eoff, sizeof(*eh));
908 				if (eh == NULL) {
909 					icmp6stat.icp6s_tooshort++;
910 					return (-1);
911 				}
912 #endif
913 
914 				if (nxt == IPPROTO_AH)
915 					eoff += (eh->ip6e_len + 2) << 2;
916 				else
917 					eoff += (eh->ip6e_len + 1) << 3;
918 				nxt = eh->ip6e_nxt;
919 				break;
920 			case IPPROTO_ROUTING:
921 				/*
922 				 * When the erroneous packet contains a
923 				 * routing header, we should examine the
924 				 * header to determine the final destination.
925 				 * Otherwise, we can't properly update
926 				 * information that depends on the final
927 				 * destination (e.g. path MTU).
928 				 */
929 #ifndef PULLDOWN_TEST
930 				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth),
931 						 -1);
932 				rth = (struct ip6_rthdr *)(mtod(m, caddr_t)
933 							   + eoff);
934 #else
935 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
936 					       eoff, sizeof(*rth));
937 				if (rth == NULL) {
938 					icmp6stat.icp6s_tooshort++;
939 					return (-1);
940 				}
941 #endif
942 				rthlen = (rth->ip6r_len + 1) << 3;
943 				eoff += rthlen;
944 				nxt = rth->ip6r_nxt;
945 				break;
946 			case IPPROTO_FRAGMENT:
947 #ifndef PULLDOWN_TEST
948 				IP6_EXTHDR_CHECK(m, 0, eoff +
949 						 sizeof(struct ip6_frag),
950 						 -1);
951 				fh = (struct ip6_frag *)(mtod(m, caddr_t)
952 							 + eoff);
953 #else
954 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
955 					       eoff, sizeof(*fh));
956 				if (fh == NULL) {
957 					icmp6stat.icp6s_tooshort++;
958 					return (-1);
959 				}
960 #endif
961 				/*
962 				 * Data after a fragment header is meaningless
963 				 * unless it is the first fragment, but
964 				 * we'll go to the notify label for path MTU
965 				 * discovery.
966 				 */
967 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
968 					goto notify;
969 
970 				eoff += sizeof(struct ip6_frag);
971 				nxt = fh->ip6f_nxt;
972 				break;
973 			default:
974 				/*
975 				 * This case includes ESP and the No Next
976 				 * Header.  In such cases going to the notify
977 				 * label does not have any meaning
978 				 * (i.e. pr_ctlinput will be NULL), but we go
979 				 * anyway since we might have to update
980 				 * path MTU information.
981 				 */
982 				goto notify;
983 			}
984 		}
985 	  notify:
986 #ifndef PULLDOWN_TEST
987 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
988 #else
989 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
990 			       sizeof(*icmp6) + sizeof(struct ip6_hdr));
991 		if (icmp6 == NULL) {
992 			icmp6stat.icp6s_tooshort++;
993 			return (-1);
994 		}
995 #endif
996 
997 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
998 		bzero(&icmp6dst, sizeof(icmp6dst));
999 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1000 		icmp6dst.sin6_family = AF_INET6;
1001 		if (finaldst == NULL)
1002 			icmp6dst.sin6_addr = eip6->ip6_dst;
1003 		else
1004 			icmp6dst.sin6_addr = *finaldst;
1005 		icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1006 							  &icmp6dst.sin6_addr);
1007 		if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1008 				   NULL, NULL)) {
1009 			/* should be impossbile */
1010 			nd6log((LOG_DEBUG,
1011 			    "icmp6_notify_error: in6_embedscope failed\n"));
1012 			goto freeit;
1013 		}
1014 
1015 		/*
1016 		 * retrieve parameters from the inner IPv6 header, and convert
1017 		 * them into sockaddr structures.
1018 		 */
1019 		bzero(&icmp6src, sizeof(icmp6src));
1020 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1021 		icmp6src.sin6_family = AF_INET6;
1022 		icmp6src.sin6_addr = eip6->ip6_src;
1023 		icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1024 							  &icmp6src.sin6_addr);
1025 		if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1026 				   NULL, NULL)) {
1027 			/* should be impossbile */
1028 			nd6log((LOG_DEBUG,
1029 			    "icmp6_notify_error: in6_embedscope failed\n"));
1030 			goto freeit;
1031 		}
1032 		icmp6src.sin6_flowinfo =
1033 			(eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1034 
1035 		if (finaldst == NULL)
1036 			finaldst = &eip6->ip6_dst;
1037 		ip6cp.ip6c_m = m;
1038 		ip6cp.ip6c_icmp6 = icmp6;
1039 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1040 		ip6cp.ip6c_off = eoff;
1041 		ip6cp.ip6c_finaldst = finaldst;
1042 		ip6cp.ip6c_src = &icmp6src;
1043 		ip6cp.ip6c_nxt = nxt;
1044 
1045 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1046 			notifymtu = ntohl(icmp6->icmp6_mtu);
1047 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1048 			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1049 		}
1050 
1051 		so_pru_ctlinput(
1052 			&inet6sw[ip6_protox[nxt]],
1053 			code, (struct sockaddr *)&icmp6dst, &ip6cp);
1054 	}
1055 	return (0);
1056 
1057 freeit:
1058 	m_freem(m);
1059 	return (-1);
1060 }
1061 
1062 void
1063 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1064 {
1065 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1066 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1067 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1068 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1069 	struct rtentry *rt = NULL;
1070 	struct sockaddr_in6 sin6;
1071 
1072 	if (!validated)
1073 		return;
1074 
1075 	bzero(&sin6, sizeof(sin6));
1076 	sin6.sin6_family = PF_INET6;
1077 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1078 	sin6.sin6_addr = *dst;
1079 	/* XXX normally, this won't happen */
1080 	if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1081 		sin6.sin6_addr.s6_addr16[1] =
1082 		    htons(m->m_pkthdr.rcvif->if_index);
1083 	}
1084 	/* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1085 	rt = rtpurelookup((struct sockaddr *)&sin6);
1086 	if (rt != NULL && (rt->rt_flags & RTF_HOST) &&
1087 	    !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1088 		if (mtu < IPV6_MMTU) {				/* XXX */
1089 			rt->rt_rmx.rmx_locks |= RTV_MTU;
1090 		} else if (mtu < rt->rt_ifp->if_mtu &&
1091 			   rt->rt_rmx.rmx_mtu > mtu) {
1092 			icmp6stat.icp6s_pmtuchg++;
1093 			rt->rt_rmx.rmx_mtu = mtu;
1094 		}
1095 	}
1096 	if (rt)
1097 		--rt->rt_refcnt;
1098 }
1099 
1100 /*
1101  * Process a Node Information Query packet, based on
1102  * draft-ietf-ipngwg-icmp-name-lookups-07.
1103  *
1104  * Spec incompatibilities:
1105  * - IPv6 Subject address handling
1106  * - IPv4 Subject address handling support missing
1107  * - Proxy reply (answer even if it's not for me)
1108  * - joins NI group address at in6_ifattach() time only, does not cope
1109  *   with hostname changes by sethostname(3)
1110  */
1111 #define hostnamelen	strlen(hostname)
1112 static struct mbuf *
1113 ni6_input(struct mbuf *m, int off)
1114 {
1115 	struct icmp6_nodeinfo *ni6, *nni6;
1116 	struct mbuf *n = NULL;
1117 	u_int16_t qtype;
1118 	int subjlen;
1119 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1120 	struct ni_reply_fqdn *fqdn;
1121 	int addrs;		/* for NI_QTYPE_NODEADDR */
1122 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1123 	struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1124 	struct sockaddr_in6 sin6_d; /* XXX: we should retrieve this from m_aux */
1125 	struct ip6_hdr *ip6;
1126 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1127 	char *subj = NULL;
1128 	struct in6_ifaddr *ia6 = NULL;
1129 
1130 	ip6 = mtod(m, struct ip6_hdr *);
1131 #ifndef PULLDOWN_TEST
1132 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1133 #else
1134 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1135 	if (ni6 == NULL) {
1136 		/* m is already reclaimed */
1137 		return NULL;
1138 	}
1139 #endif
1140 
1141 	/*
1142 	 * Validate IPv6 destination address.
1143 	 *
1144 	 * The Responder must discard the Query without further processing
1145 	 * unless it is one of the Responder's unicast or anycast addresses, or
1146 	 * a link-local scope multicast address which the Responder has joined.
1147 	 * [icmp-name-lookups-07, Section 4.]
1148 	 */
1149 	bzero(&sin6, sizeof(sin6));
1150 	sin6.sin6_family = AF_INET6;
1151 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1152 	bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1153 	/* XXX scopeid */
1154 	if ((ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)&sin6)) != NULL) {
1155 		/* unicast/anycast, fine */
1156 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1157 		    (icmp6_nodeinfo & 4) == 0) {
1158 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1159 				"a temporary address in %s:%d",
1160 			       __FILE__, __LINE__));
1161 			goto bad;
1162 		}
1163 	} else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1164 		; /* link-local multicast, fine */
1165 	else
1166 		goto bad;
1167 
1168 	/* validate query Subject field. */
1169 	qtype = ntohs(ni6->ni_qtype);
1170 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1171 	switch (qtype) {
1172 	case NI_QTYPE_NOOP:
1173 	case NI_QTYPE_SUPTYPES:
1174 		/* 07 draft */
1175 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1176 			break;
1177 		/* FALLTHROUGH */
1178 	case NI_QTYPE_FQDN:
1179 	case NI_QTYPE_NODEADDR:
1180 		switch (ni6->ni_code) {
1181 		case ICMP6_NI_SUBJ_IPV6:
1182 #if ICMP6_NI_SUBJ_IPV6 != 0
1183 		case 0:
1184 #endif
1185 			/*
1186 			 * backward compatibility - try to accept 03 draft
1187 			 * format, where no Subject is present.
1188 			 */
1189 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1190 			    subjlen == 0) {
1191 				oldfqdn++;
1192 				break;
1193 			}
1194 #if ICMP6_NI_SUBJ_IPV6 != 0
1195 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1196 				goto bad;
1197 #endif
1198 
1199 			if (subjlen != sizeof(sin6.sin6_addr))
1200 				goto bad;
1201 
1202 			/*
1203 			 * Validate Subject address.
1204 			 *
1205 			 * Not sure what exactly "address belongs to the node"
1206 			 * means in the spec, is it just unicast, or what?
1207 			 *
1208 			 * At this moment we consider Subject address as
1209 			 * "belong to the node" if the Subject address equals
1210 			 * to the IPv6 destination address; validation for
1211 			 * IPv6 destination address should have done enough
1212 			 * check for us.
1213 			 *
1214 			 * We do not do proxy at this moment.
1215 			 */
1216 			/* m_pulldown instead of copy? */
1217 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1218 			    subjlen, (caddr_t)&sin6.sin6_addr);
1219 			sin6.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1220 							      &sin6.sin6_addr);
1221 			in6_embedscope(&sin6.sin6_addr, &sin6, NULL, NULL);
1222 			bzero(&sin6_d, sizeof(sin6_d));
1223 			sin6_d.sin6_family = AF_INET6; /* not used, actually */
1224 			sin6_d.sin6_len = sizeof(sin6_d); /* ditto */
1225 			sin6_d.sin6_addr = ip6->ip6_dst;
1226 			sin6_d.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1227 								&ip6->ip6_dst);
1228 			in6_embedscope(&sin6_d.sin6_addr, &sin6_d, NULL, NULL);
1229 			subj = (char *)&sin6;
1230 			if (SA6_ARE_ADDR_EQUAL(&sin6, &sin6_d))
1231 				break;
1232 
1233 			/*
1234 			 * XXX if we are to allow other cases, we should really
1235 			 * be careful about scope here.
1236 			 * basically, we should disallow queries toward IPv6
1237 			 * destination X with subject Y, if scope(X) > scope(Y).
1238 			 * if we allow scope(X) > scope(Y), it will result in
1239 			 * information leakage across scope boundary.
1240 			 */
1241 			goto bad;
1242 
1243 		case ICMP6_NI_SUBJ_FQDN:
1244 			/*
1245 			 * Validate Subject name with gethostname(3).
1246 			 *
1247 			 * The behavior may need some debate, since:
1248 			 * - we are not sure if the node has FQDN as
1249 			 *   hostname (returned by gethostname(3)).
1250 			 * - the code does wildcard match for truncated names.
1251 			 *   however, we are not sure if we want to perform
1252 			 *   wildcard match, if gethostname(3) side has
1253 			 *   truncated hostname.
1254 			 */
1255 			n = ni6_nametodns(hostname, hostnamelen, 0);
1256 			if (!n || n->m_next || n->m_len == 0)
1257 				goto bad;
1258 			IP6_EXTHDR_GET(subj, char *, m,
1259 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1260 			if (subj == NULL)
1261 				goto bad;
1262 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1263 					n->m_len)) {
1264 				goto bad;
1265 			}
1266 			m_freem(n);
1267 			n = NULL;
1268 			break;
1269 
1270 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1271 		default:
1272 			goto bad;
1273 		}
1274 		break;
1275 	}
1276 
1277 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1278 	switch (qtype) {
1279 	case NI_QTYPE_FQDN:
1280 		if ((icmp6_nodeinfo & 1) == 0)
1281 			goto bad;
1282 		break;
1283 	case NI_QTYPE_NODEADDR:
1284 		if ((icmp6_nodeinfo & 2) == 0)
1285 			goto bad;
1286 		break;
1287 	}
1288 
1289 	/* guess reply length */
1290 	switch (qtype) {
1291 	case NI_QTYPE_NOOP:
1292 		break;		/* no reply data */
1293 	case NI_QTYPE_SUPTYPES:
1294 		replylen += sizeof(u_int32_t);
1295 		break;
1296 	case NI_QTYPE_FQDN:
1297 		/* XXX will append an mbuf */
1298 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1299 		break;
1300 	case NI_QTYPE_NODEADDR:
1301 		addrs = ni6_addrs(ni6, m, &ifp, subj);
1302 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1303 					  sizeof(u_int32_t))) > MCLBYTES)
1304 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1305 		break;
1306 	default:
1307 		/*
1308 		 * XXX: We must return a reply with the ICMP6 code
1309 		 * `unknown Qtype' in this case. However we regard the case
1310 		 * as an FQDN query for backward compatibility.
1311 		 * Older versions set a random value to this field,
1312 		 * so it rarely varies in the defined qtypes.
1313 		 * But the mechanism is not reliable...
1314 		 * maybe we should obsolete older versions.
1315 		 */
1316 		qtype = NI_QTYPE_FQDN;
1317 		/* XXX will append an mbuf */
1318 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1319 		oldfqdn++;
1320 		break;
1321 	}
1322 
1323 	/* allocate an mbuf to reply. */
1324 	if (replylen > MCLBYTES) {
1325 		/*
1326 		 * XXX: should we try to allocate more? But MCLBYTES
1327 		 * is probably much larger than IPV6_MMTU...
1328 		 */
1329 		goto bad;
1330 	}
1331 	n = m_getb(replylen, MB_DONTWAIT, m->m_type, M_PKTHDR);
1332 	if (n == NULL) {
1333 		m_freem(m);
1334 		return (NULL);
1335 	}
1336 	M_MOVE_PKTHDR(n, m); /* just for recvif */
1337 	n->m_pkthdr.len = n->m_len = replylen;
1338 
1339 	/* copy mbuf header and IPv6 + Node Information base headers */
1340 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1341 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1342 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1343 
1344 	/* qtype dependent procedure */
1345 	switch (qtype) {
1346 	case NI_QTYPE_NOOP:
1347 		nni6->ni_code = ICMP6_NI_SUCCESS;
1348 		nni6->ni_flags = 0;
1349 		break;
1350 	case NI_QTYPE_SUPTYPES:
1351 	{
1352 		u_int32_t v;
1353 		nni6->ni_code = ICMP6_NI_SUCCESS;
1354 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1355 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1356 		v = (u_int32_t)htonl(0x0000000f);
1357 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1358 		break;
1359 	}
1360 	case NI_QTYPE_FQDN:
1361 		nni6->ni_code = ICMP6_NI_SUCCESS;
1362 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1363 						sizeof(struct ip6_hdr) +
1364 						sizeof(struct icmp6_nodeinfo));
1365 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1366 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1367 		/*
1368 		 * XXX do we really have FQDN in variable "hostname"?
1369 		 */
1370 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1371 		if (n->m_next == NULL)
1372 			goto bad;
1373 		/* XXX we assume that n->m_next is not a chain */
1374 		if (n->m_next->m_next != NULL)
1375 			goto bad;
1376 		n->m_pkthdr.len += n->m_next->m_len;
1377 		break;
1378 	case NI_QTYPE_NODEADDR:
1379 	{
1380 		int lenlim, copied;
1381 
1382 		nni6->ni_code = ICMP6_NI_SUCCESS;
1383 		n->m_pkthdr.len = n->m_len =
1384 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1385 		lenlim = M_TRAILINGSPACE(n);
1386 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1387 		/* XXX: reset mbuf length */
1388 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1389 			sizeof(struct icmp6_nodeinfo) + copied;
1390 		break;
1391 	}
1392 	default:
1393 		break;		/* XXX impossible! */
1394 	}
1395 
1396 	nni6->ni_type = ICMP6_NI_REPLY;
1397 	m_freem(m);
1398 	return (n);
1399 
1400 bad:
1401 	m_freem(m);
1402 	if (n)
1403 		m_freem(n);
1404 	return (NULL);
1405 }
1406 #undef hostnamelen
1407 
1408 /*
1409  * make a mbuf with DNS-encoded string.  no compression support.
1410  *
1411  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1412  * treated as truncated name (two \0 at the end).  this is a wild guess.
1413  */
1414 static struct mbuf *
1415 ni6_nametodns(const char *name, int namelen,
1416 	      int old) /* return pascal string if non-zero */
1417 {
1418 	struct mbuf *m;
1419 	char *cp, *ep;
1420 	const char *p, *q;
1421 	int i, len, nterm;
1422 
1423 	if (old)
1424 		len = namelen + 1;
1425 	else
1426 		len = MCLBYTES;
1427 
1428 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1429 	m = m_getb(len, MB_DONTWAIT, MT_DATA, 0);
1430 	if (!m)
1431 		goto fail;
1432 
1433 	if (old) {
1434 		m->m_len = len;
1435 		*mtod(m, char *) = namelen;
1436 		bcopy(name, mtod(m, char *) + 1, namelen);
1437 		return m;
1438 	} else {
1439 		m->m_len = 0;
1440 		cp = mtod(m, char *);
1441 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1442 
1443 		/* if not certain about my name, return empty buffer */
1444 		if (namelen == 0)
1445 			return m;
1446 
1447 		/*
1448 		 * guess if it looks like shortened hostname, or FQDN.
1449 		 * shortened hostname needs two trailing "\0".
1450 		 */
1451 		i = 0;
1452 		for (p = name; p < name + namelen; p++) {
1453 			if (*p && *p == '.')
1454 				i++;
1455 		}
1456 		if (i < 2)
1457 			nterm = 2;
1458 		else
1459 			nterm = 1;
1460 
1461 		p = name;
1462 		while (cp < ep && p < name + namelen) {
1463 			i = 0;
1464 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1465 				i++;
1466 			/* result does not fit into mbuf */
1467 			if (cp + i + 1 >= ep)
1468 				goto fail;
1469 			/*
1470 			 * DNS label length restriction, RFC1035 page 8.
1471 			 * "i == 0" case is included here to avoid returning
1472 			 * 0-length label on "foo..bar".
1473 			 */
1474 			if (i <= 0 || i >= 64)
1475 				goto fail;
1476 			*cp++ = i;
1477 			bcopy(p, cp, i);
1478 			cp += i;
1479 			p = q;
1480 			if (p < name + namelen && *p == '.')
1481 				p++;
1482 		}
1483 		/* termination */
1484 		if (cp + nterm >= ep)
1485 			goto fail;
1486 		while (nterm-- > 0)
1487 			*cp++ = '\0';
1488 		m->m_len = cp - mtod(m, char *);
1489 		return m;
1490 	}
1491 
1492 	panic("should not reach here");
1493 	/* NOTREACHED */
1494 
1495 fail:
1496 	if (m)
1497 		m_freem(m);
1498 	return NULL;
1499 }
1500 
1501 /*
1502  * check if two DNS-encoded string matches.  takes care of truncated
1503  * form (with \0\0 at the end).  no compression support.
1504  * XXX upper/lowercase match (see RFC2065)
1505  */
1506 static int
1507 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1508 {
1509 	const char *a0, *b0;
1510 	int l;
1511 
1512 	/* simplest case - need validation? */
1513 	if (alen == blen && bcmp(a, b, alen) == 0)
1514 		return 1;
1515 
1516 	a0 = a;
1517 	b0 = b;
1518 
1519 	/* termination is mandatory */
1520 	if (alen < 2 || blen < 2)
1521 		return 0;
1522 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1523 		return 0;
1524 	alen--;
1525 	blen--;
1526 
1527 	while (a - a0 < alen && b - b0 < blen) {
1528 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1529 			return 0;
1530 
1531 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1532 			return 0;
1533 		/* we don't support compression yet */
1534 		if (a[0] >= 64 || b[0] >= 64)
1535 			return 0;
1536 
1537 		/* truncated case */
1538 		if (a[0] == 0 && a - a0 == alen - 1)
1539 			return 1;
1540 		if (b[0] == 0 && b - b0 == blen - 1)
1541 			return 1;
1542 		if (a[0] == 0 || b[0] == 0)
1543 			return 0;
1544 
1545 		if (a[0] != b[0])
1546 			return 0;
1547 		l = a[0];
1548 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1549 			return 0;
1550 		if (bcmp(a + 1, b + 1, l) != 0)
1551 			return 0;
1552 
1553 		a += 1 + l;
1554 		b += 1 + l;
1555 	}
1556 
1557 	if (a - a0 == alen && b - b0 == blen)
1558 		return 1;
1559 	else
1560 		return 0;
1561 }
1562 
1563 /*
1564  * calculate the number of addresses to be returned in the node info reply.
1565  */
1566 static int
1567 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1568 	  char *subj)
1569 {
1570 	struct ifnet *ifp;
1571 	struct in6_ifaddr *ifa6;
1572 	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1573 	int addrs = 0, addrsofif, iffound = 0;
1574 	int niflags = ni6->ni_flags;
1575 
1576 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1577 		switch (ni6->ni_code) {
1578 		case ICMP6_NI_SUBJ_IPV6:
1579 			if (subj == NULL) /* must be impossible... */
1580 				return (0);
1581 			subj_ip6 = (struct sockaddr_in6 *)subj;
1582 			break;
1583 		default:
1584 			/*
1585 			 * XXX: we only support IPv6 subject address for
1586 			 * this Qtype.
1587 			 */
1588 			return (0);
1589 		}
1590 	}
1591 
1592 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
1593 	{
1594 		struct ifaddr_container *ifac;
1595 
1596 		addrsofif = 0;
1597 		TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link)
1598 		{
1599 			struct ifaddr *ifa = ifac->ifa;
1600 
1601 			if (ifa->ifa_addr->sa_family != AF_INET6)
1602 				continue;
1603 			ifa6 = (struct in6_ifaddr *)ifa;
1604 
1605 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1606 			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1607 					       &ifa6->ia_addr.sin6_addr))
1608 				iffound = 1;
1609 
1610 			/*
1611 			 * IPv4-mapped addresses can only be returned by a
1612 			 * Node Information proxy, since they represent
1613 			 * addresses of IPv4-only nodes, which perforce do
1614 			 * not implement this protocol.
1615 			 * [icmp-name-lookups-07, Section 5.4]
1616 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1617 			 * this function at this moment.
1618 			 */
1619 
1620 			/* What do we have to do about ::1? */
1621 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1622 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1623 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1624 					continue;
1625 				break;
1626 			case IPV6_ADDR_SCOPE_SITELOCAL:
1627 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1628 					continue;
1629 				break;
1630 			case IPV6_ADDR_SCOPE_GLOBAL:
1631 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1632 					continue;
1633 				break;
1634 			default:
1635 				continue;
1636 			}
1637 
1638 			/*
1639 			 * check if anycast is okay.
1640 			 * XXX: just experimental.  not in the spec.
1641 			 */
1642 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) &&
1643 			    !(niflags & NI_NODEADDR_FLAG_ANYCAST))
1644 				continue; /* we need only unicast addresses */
1645 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) &&
1646 			    (icmp6_nodeinfo & 4) == 0) {
1647 				continue;
1648 			}
1649 			addrsofif++; /* count the address */
1650 		}
1651 		if (iffound) {
1652 			*ifpp = ifp;
1653 			return (addrsofif);
1654 		}
1655 
1656 		addrs += addrsofif;
1657 	}
1658 
1659 	return (addrs);
1660 }
1661 
1662 static int
1663 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1664 		struct ifnet *ifp0, int resid)
1665 {
1666 	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1667 	struct in6_ifaddr *ifa6;
1668 	struct ifnet *ifp_dep = NULL;
1669 	int copied = 0, allow_deprecated = 0;
1670 	u_char *cp = (u_char *)(nni6 + 1);
1671 	int niflags = ni6->ni_flags;
1672 	u_int32_t ltime;
1673 
1674 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1675 		return (0);	/* needless to copy */
1676 
1677 again:
1678 
1679 	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list))
1680 	{
1681 		struct ifaddr_container *ifac;
1682 
1683 		TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1684 			struct ifaddr *ifa = ifac->ifa;
1685 
1686 			if (ifa->ifa_addr->sa_family != AF_INET6)
1687 				continue;
1688 			ifa6 = (struct in6_ifaddr *)ifa;
1689 
1690 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) &&
1691 			    allow_deprecated == 0) {
1692 				/*
1693 				 * prefererred address should be put before
1694 				 * deprecated addresses.
1695 				 */
1696 
1697 				/* record the interface for later search */
1698 				if (ifp_dep == NULL)
1699 					ifp_dep = ifp;
1700 
1701 				continue;
1702 			}
1703 			else if (!(ifa6->ia6_flags & IN6_IFF_DEPRECATED) &&
1704 				 allow_deprecated != 0)
1705 				continue; /* we now collect deprecated addrs */
1706 
1707 			/* What do we have to do about ::1? */
1708 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1709 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1710 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1711 					continue;
1712 				break;
1713 			case IPV6_ADDR_SCOPE_SITELOCAL:
1714 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1715 					continue;
1716 				break;
1717 			case IPV6_ADDR_SCOPE_GLOBAL:
1718 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1719 					continue;
1720 				break;
1721 			default:
1722 				continue;
1723 			}
1724 
1725 			/*
1726 			 * check if anycast is okay.
1727 			 * XXX: just experimental. not in the spec.
1728 			 */
1729 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) &&
1730 			    !(niflags & NI_NODEADDR_FLAG_ANYCAST))
1731 				continue;
1732 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) &&
1733 			    (icmp6_nodeinfo & 4) == 0) {
1734 				continue;
1735 			}
1736 
1737 			/* now we can copy the address */
1738 			if (resid < sizeof(struct in6_addr) +
1739 			    sizeof(u_int32_t)) {
1740 				/*
1741 				 * We give up much more copy.
1742 				 * Set the truncate flag and return.
1743 				 */
1744 				nni6->ni_flags |=
1745 					NI_NODEADDR_FLAG_TRUNCATE;
1746 				return (copied);
1747 			}
1748 
1749 			/*
1750 			 * Set the TTL of the address.
1751 			 * The TTL value should be one of the following
1752 			 * according to the specification:
1753 			 *
1754 			 * 1. The remaining lifetime of a DHCP lease on the
1755 			 *    address, or
1756 			 * 2. The remaining Valid Lifetime of a prefix from
1757 			 *    which the address was derived through Stateless
1758 			 *    Autoconfiguration.
1759 			 *
1760 			 * Note that we currently do not support stateful
1761 			 * address configuration by DHCPv6, so the former
1762 			 * case can't happen.
1763 			 */
1764 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1765 				ltime = ND6_INFINITE_LIFETIME;
1766 			else {
1767 				if (ifa6->ia6_lifetime.ia6t_expire >
1768 				    time_second)
1769 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1770 				else
1771 					ltime = 0;
1772 			}
1773 
1774 			bcopy(&ltime, cp, sizeof(u_int32_t));
1775 			cp += sizeof(u_int32_t);
1776 
1777 			/* copy the address itself */
1778 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1779 			      sizeof(struct in6_addr));
1780 			/* XXX: KAME link-local hack; remove ifindex */
1781 			if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr))
1782 				((struct in6_addr *)cp)->s6_addr16[1] = 0;
1783 			cp += sizeof(struct in6_addr);
1784 
1785 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1786 			copied += (sizeof(struct in6_addr) +
1787 				   sizeof(u_int32_t));
1788 		}
1789 		if (ifp0)	/* we need search only on the specified IF */
1790 			break;
1791 	}
1792 
1793 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1794 		ifp = ifp_dep;
1795 		allow_deprecated = 1;
1796 
1797 		goto again;
1798 	}
1799 
1800 	return (copied);
1801 }
1802 
1803 /*
1804  * XXX almost dup'ed code with rip6_input.
1805  */
1806 static int
1807 icmp6_rip6_input(struct	mbuf **mp, int	off)
1808 {
1809 	struct mbuf *m = *mp;
1810 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1811 	struct in6pcb *in6p;
1812 	struct in6pcb *last = NULL;
1813 	struct sockaddr_in6 rip6src;
1814 	struct icmp6_hdr *icmp6;
1815 	struct mbuf *opts = NULL;
1816 
1817 #ifndef PULLDOWN_TEST
1818 	/* this is assumed to be safe. */
1819 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1820 #else
1821 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1822 	if (icmp6 == NULL) {
1823 		/* m is already reclaimed */
1824 		return IPPROTO_DONE;
1825 	}
1826 #endif
1827 
1828 	bzero(&rip6src, sizeof(rip6src));
1829 	rip6src.sin6_len = sizeof(struct sockaddr_in6);
1830 	rip6src.sin6_family = AF_INET6;
1831 	/* KAME hack: recover scopeid */
1832 	in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif);
1833 
1834 	LIST_FOREACH(in6p, &ripcbinfo.pcblisthead, inp_list)
1835 	{
1836 		if (in6p->inp_flags & INP_PLACEMARKER)
1837 			continue;
1838 		if ((in6p->inp_vflag & INP_IPV6) == 0)
1839 			continue;
1840 #ifdef HAVE_NRL_INPCB
1841 		if (!(in6p->in6p_flags & INP_IPV6))
1842 			continue;
1843 #endif
1844 		if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1845 			continue;
1846 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1847 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1848 			continue;
1849 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1850 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1851 			continue;
1852 		if (in6p->in6p_icmp6filt
1853 		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1854 				 in6p->in6p_icmp6filt))
1855 			continue;
1856 		if (last) {
1857 			struct socket *so;
1858 			struct mbuf *n;
1859 
1860 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1861 				if (last->in6p_flags & IN6P_CONTROLOPTS)
1862 					ip6_savecontrol(last, &opts, ip6, n);
1863 				/* strip intermediate headers */
1864 				m_adj(n, off);
1865 				so = last->in6p_socket;
1866 				lwkt_gettoken(&so->so_rcv.ssb_token);
1867 				if (ssb_appendaddr(&so->so_rcv,
1868 						 (struct sockaddr *)&rip6src,
1869 						 n, opts) == 0) {
1870 					/* should notify about lost packet */
1871 					m_freem(n);
1872 					if (opts)
1873 						m_freem(opts);
1874 				} else {
1875 					sorwakeup(so);
1876 				}
1877 				lwkt_reltoken(&so->so_rcv.ssb_token);
1878 				opts = NULL;
1879 			}
1880 		}
1881 		last = in6p;
1882 	}
1883 	if (last) {
1884 		struct socket *so;
1885 
1886 		if (last->in6p_flags & IN6P_CONTROLOPTS)
1887 			ip6_savecontrol(last, &opts, ip6, m);
1888 		/* strip intermediate headers */
1889 		m_adj(m, off);
1890 		so = last->in6p_socket;
1891 		lwkt_gettoken(&so->so_rcv.ssb_token);
1892 		if (ssb_appendaddr(&so->so_rcv, (struct sockaddr *)&rip6src,
1893 				   m, opts) == 0) {
1894 			m_freem(m);
1895 			if (opts)
1896 				m_freem(opts);
1897 		} else {
1898 			sorwakeup(so);
1899 		}
1900 		lwkt_reltoken(&so->so_rcv.ssb_token);
1901 	} else {
1902 		m_freem(m);
1903 		ip6stat.ip6s_delivered--;
1904 	}
1905 	return IPPROTO_DONE;
1906 }
1907 
1908 /*
1909  * Reflect the ip6 packet back to the source.
1910  * OFF points to the icmp6 header, counted from the top of the mbuf.
1911  */
1912 void
1913 icmp6_reflect(struct mbuf *m, size_t off)
1914 {
1915 	struct ip6_hdr *ip6;
1916 	struct icmp6_hdr *icmp6;
1917 	struct in6_ifaddr *ia;
1918 	struct in6_addr t, *src = NULL;
1919 	int plen;
1920 	int type, code;
1921 	struct ifnet *outif = NULL;
1922 	struct sockaddr_in6 sa6_src, sa6_dst;
1923 #ifdef COMPAT_RFC1885
1924 	int mtu = IPV6_MMTU;
1925 	struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
1926 #endif
1927 
1928 	/* too short to reflect */
1929 	if (off < sizeof(struct ip6_hdr)) {
1930 		nd6log((LOG_DEBUG,
1931 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
1932 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
1933 		    __FILE__, __LINE__));
1934 		goto bad;
1935 	}
1936 
1937 	/*
1938 	 * If there are extra headers between IPv6 and ICMPv6, strip
1939 	 * off that header first.
1940 	 */
1941 #ifdef DIAGNOSTIC
1942 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
1943 		panic("assumption failed in icmp6_reflect");
1944 #endif
1945 	if (off > sizeof(struct ip6_hdr)) {
1946 		size_t l;
1947 		struct ip6_hdr nip6;
1948 
1949 		l = off - sizeof(struct ip6_hdr);
1950 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
1951 		m_adj(m, l);
1952 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1953 		if (m->m_len < l) {
1954 			if ((m = m_pullup(m, l)) == NULL)
1955 				return;
1956 		}
1957 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
1958 	} else /* off == sizeof(struct ip6_hdr) */ {
1959 		size_t l;
1960 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1961 		if (m->m_len < l) {
1962 			if ((m = m_pullup(m, l)) == NULL)
1963 				return;
1964 		}
1965 	}
1966 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
1967 	ip6 = mtod(m, struct ip6_hdr *);
1968 	ip6->ip6_nxt = IPPROTO_ICMPV6;
1969 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
1970 	type = icmp6->icmp6_type; /* keep type for statistics */
1971 	code = icmp6->icmp6_code; /* ditto. */
1972 
1973 	t = ip6->ip6_dst;
1974 	/*
1975 	 * ip6_input() drops a packet if its src is multicast.
1976 	 * So, the src is never multicast.
1977 	 */
1978 	ip6->ip6_dst = ip6->ip6_src;
1979 
1980 	/*
1981 	 * XXX: make sure to embed scope zone information, using
1982 	 * already embedded IDs or the received interface (if any).
1983 	 * Note that rcvif may be NULL.
1984 	 * TODO: scoped routing case (XXX).
1985 	 */
1986 	bzero(&sa6_src, sizeof(sa6_src));
1987 	sa6_src.sin6_family = AF_INET6;
1988 	sa6_src.sin6_len = sizeof(sa6_src);
1989 	sa6_src.sin6_addr = ip6->ip6_dst;
1990 	in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
1991 	in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL);
1992 	bzero(&sa6_dst, sizeof(sa6_dst));
1993 	sa6_dst.sin6_family = AF_INET6;
1994 	sa6_dst.sin6_len = sizeof(sa6_dst);
1995 	sa6_dst.sin6_addr = t;
1996 	in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
1997 	in6_embedscope(&t, &sa6_dst, NULL, NULL);
1998 
1999 #ifdef COMPAT_RFC1885
2000 	/*
2001 	 * xxx guess MTU
2002 	 * RFC 1885 requires that echo reply should be truncated if it
2003 	 * does not fit in with (return) path MTU, but the description was
2004 	 * removed in the new spec.
2005 	 */
2006 	if (icmp6_reflect_rt.ro_rt == 0 ||
2007 	    ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
2008 		if (icmp6_reflect_rt.ro_rt) {
2009 			RTFREE(icmp6_reflect_rt.ro_rt);
2010 			icmp6_reflect_rt.ro_rt = 0;
2011 		}
2012 		bzero(sin6, sizeof(*sin6));
2013 		sin6->sin6_family = PF_INET6;
2014 		sin6->sin6_len = sizeof(struct sockaddr_in6);
2015 		sin6->sin6_addr = ip6->ip6_dst;
2016 
2017 		rtalloc_ign((struct route *)&icmp6_reflect_rt.ro_rt,
2018 			    RTF_PRCLONING);
2019 	}
2020 
2021 	if (icmp6_reflect_rt.ro_rt == 0)
2022 		goto bad;
2023 
2024 	if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
2025 	    && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
2026 		mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
2027 
2028 	if (mtu < m->m_pkthdr.len) {
2029 		plen -= (m->m_pkthdr.len - mtu);
2030 		m_adj(m, mtu - m->m_pkthdr.len);
2031 	}
2032 #endif
2033 	/*
2034 	 * If the incoming packet was addressed directly to us(i.e. unicast),
2035 	 * use dst as the src for the reply.
2036 	 * The IN6_IFF_NOTREADY case would be VERY rare, but is possible
2037 	 * (for example) when we encounter an error while forwarding procedure
2038 	 * destined to a duplicated address of ours.
2039 	 */
2040 	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2041 		if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2042 		    (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2043 			src = &t;
2044 			break;
2045 		}
2046 	if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2047 		/*
2048 		 * This is the case if the dst is our link-local address
2049 		 * and the sender is also ourselves.
2050 		 */
2051 		src = &t;
2052 	}
2053 
2054 	if (src == NULL) {
2055 		int e;
2056 		struct route_in6 ro;
2057 
2058 		/*
2059 		 * This case matches to multicasts, our anycast, or unicasts
2060 		 * that we do not own.  Select a source address based on the
2061 		 * source address of the erroneous packet.
2062 		 */
2063 		bzero(&ro, sizeof(ro));
2064 		src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e, NULL);
2065 		if (ro.ro_rt)
2066 			RTFREE(ro.ro_rt); /* XXX: we could use this */
2067 		if (src == NULL) {
2068 			nd6log((LOG_DEBUG,
2069 			    "icmp6_reflect: source can't be determined: "
2070 			    "dst=%s, error=%d\n",
2071 			    ip6_sprintf(&sa6_src.sin6_addr), e));
2072 			goto bad;
2073 		}
2074 	}
2075 
2076 	ip6->ip6_src = *src;
2077 
2078 	ip6->ip6_flow = 0;
2079 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2080 	ip6->ip6_vfc |= IPV6_VERSION;
2081 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2082 	if (m->m_pkthdr.rcvif) {
2083 		/* XXX: This may not be the outgoing interface */
2084 		ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2085 	} else
2086 		ip6->ip6_hlim = ip6_defhlim;
2087 
2088 	icmp6->icmp6_cksum = 0;
2089 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2090 					sizeof(struct ip6_hdr), plen);
2091 
2092 	/*
2093 	 * XXX option handling
2094 	 */
2095 
2096 	m->m_flags &= ~(M_BCAST|M_MCAST);
2097 
2098 #ifdef COMPAT_RFC1885
2099 	ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif, NULL);
2100 #else
2101 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2102 #endif
2103 	if (outif)
2104 		icmp6_ifoutstat_inc(outif, type, code);
2105 
2106 	return;
2107 
2108 bad:
2109 	m_freem(m);
2110 	return;
2111 }
2112 
2113 void
2114 icmp6_fasttimo(void)
2115 {
2116 
2117 	mld6_fasttimeo();
2118 }
2119 
2120 static const char *
2121 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2122 		    struct in6_addr *tgt6)
2123 {
2124 	static char buf[1024];
2125 	ksnprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2126 		ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2127 	return buf;
2128 }
2129 
2130 void
2131 icmp6_redirect_input(struct mbuf *m, int off)
2132 {
2133 	struct ifnet *ifp = m->m_pkthdr.rcvif;
2134 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2135 	struct nd_redirect *nd_rd;
2136 	int icmp6len = ntohs(ip6->ip6_plen);
2137 	char *lladdr = NULL;
2138 	int lladdrlen = 0;
2139 #if 0
2140 	u_char *redirhdr = NULL;
2141 	int redirhdrlen = 0;
2142 #endif
2143 	struct rtentry *rt = NULL;
2144 	int is_router;
2145 	int is_onlink;
2146 	struct in6_addr src6 = ip6->ip6_src;
2147 	struct in6_addr redtgt6;
2148 	struct in6_addr reddst6;
2149 	union nd_opts ndopts;
2150 
2151 	if (!m || !ifp)
2152 		return;
2153 
2154 	/* XXX if we are router, we don't update route by icmp6 redirect */
2155 	if (ip6_forwarding)
2156 		goto freeit;
2157 	if (!icmp6_rediraccept)
2158 		goto freeit;
2159 
2160 #ifndef PULLDOWN_TEST
2161 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2162 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2163 #else
2164 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2165 	if (nd_rd == NULL) {
2166 		icmp6stat.icp6s_tooshort++;
2167 		return;
2168 	}
2169 #endif
2170 	redtgt6 = nd_rd->nd_rd_target;
2171 	reddst6 = nd_rd->nd_rd_dst;
2172 
2173 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2174 		redtgt6.s6_addr16[1] = htons(ifp->if_index);
2175 	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2176 		reddst6.s6_addr16[1] = htons(ifp->if_index);
2177 
2178 	/* validation */
2179 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2180 		nd6log((LOG_ERR,
2181 			"ICMP6 redirect sent from %s rejected; "
2182 			"must be from linklocal\n", ip6_sprintf(&src6)));
2183 		goto bad;
2184 	}
2185 	if (ip6->ip6_hlim != 255) {
2186 		nd6log((LOG_ERR,
2187 			"ICMP6 redirect sent from %s rejected; "
2188 			"hlim=%d (must be 255)\n",
2189 			ip6_sprintf(&src6), ip6->ip6_hlim));
2190 		goto bad;
2191 	}
2192     {
2193 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2194 	struct sockaddr_in6 sin6;
2195 	struct in6_addr *gw6;
2196 
2197 	bzero(&sin6, sizeof(sin6));
2198 	sin6.sin6_family = AF_INET6;
2199 	sin6.sin6_len = sizeof(struct sockaddr_in6);
2200 	bcopy(&reddst6, &sin6.sin6_addr, sizeof reddst6);
2201 	rt = rtpurelookup((struct sockaddr *)&sin6);
2202 	if (rt != NULL) {
2203 		if (rt->rt_gateway == NULL ||
2204 		    rt->rt_gateway->sa_family != AF_INET6) {
2205 			nd6log((LOG_ERR,
2206 			    "ICMP6 redirect rejected; no route "
2207 			    "with inet6 gateway found for redirect dst: %s\n",
2208 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2209 			--rt->rt_refcnt;
2210 			goto bad;
2211 		}
2212 
2213 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2214 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2215 			nd6log((LOG_ERR,
2216 				"ICMP6 redirect rejected; "
2217 				"not equal to gw-for-src=%s (must be same): "
2218 				"%s\n",
2219 				ip6_sprintf(gw6),
2220 				icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2221 			--rt->rt_refcnt;
2222 			goto bad;
2223 		}
2224 	} else {
2225 		nd6log((LOG_ERR,
2226 			"ICMP6 redirect rejected; "
2227 			"no route found for redirect dst: %s\n",
2228 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2229 		goto bad;
2230 	}
2231 	--rt->rt_refcnt;
2232 	rt = NULL;
2233     }
2234 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2235 		nd6log((LOG_ERR,
2236 			"ICMP6 redirect rejected; "
2237 			"redirect dst must be unicast: %s\n",
2238 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2239 		goto bad;
2240 	}
2241 
2242 	is_router = is_onlink = 0;
2243 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2244 		is_router = 1;	/* router case */
2245 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2246 		is_onlink = 1;	/* on-link destination case */
2247 	if (!is_router && !is_onlink) {
2248 		nd6log((LOG_ERR,
2249 			"ICMP6 redirect rejected; "
2250 			"neither router case nor onlink case: %s\n",
2251 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2252 		goto bad;
2253 	}
2254 	/* validation passed */
2255 
2256 	icmp6len -= sizeof(*nd_rd);
2257 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2258 	if (nd6_options(&ndopts) < 0) {
2259 		nd6log((LOG_INFO, "icmp6_redirect_input: "
2260 			"invalid ND option, rejected: %s\n",
2261 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2262 		/* nd6_options have incremented stats */
2263 		goto freeit;
2264 	}
2265 
2266 	if (ndopts.nd_opts_tgt_lladdr) {
2267 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2268 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2269 	}
2270 
2271 #if 0
2272 	if (ndopts.nd_opts_rh) {
2273 		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2274 		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2275 	}
2276 #endif
2277 
2278 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2279 		nd6log((LOG_INFO,
2280 			"icmp6_redirect_input: lladdrlen mismatch for %s "
2281 			"(if %d, icmp6 packet %d): %s\n",
2282 			ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2283 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2284 		goto bad;
2285 	}
2286 
2287 	/* RFC 2461 8.3 */
2288 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2289 			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2290 
2291 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2292 		/* perform rtredirect */
2293 		struct sockaddr_in6 sdst;
2294 		struct sockaddr_in6 sgw;
2295 		struct sockaddr_in6 ssrc;
2296 
2297 		bzero(&sdst, sizeof(sdst));
2298 		bzero(&sgw, sizeof(sgw));
2299 		bzero(&ssrc, sizeof(ssrc));
2300 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2301 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2302 			sizeof(struct sockaddr_in6);
2303 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2304 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2305 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2306 		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2307 			   NULL, RTF_GATEWAY | RTF_HOST,
2308 			   (struct sockaddr *)&ssrc);
2309 	}
2310 	/* finally update cached route in each socket via kpfctlinput */
2311     {
2312 	struct sockaddr_in6 sdst;
2313 
2314 	bzero(&sdst, sizeof(sdst));
2315 	sdst.sin6_family = AF_INET6;
2316 	sdst.sin6_len = sizeof(struct sockaddr_in6);
2317 	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2318 	kpfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2319 #ifdef IPSEC
2320 	key_sa_routechange((struct sockaddr *)&sdst);
2321 #endif
2322     }
2323 
2324 freeit:
2325 	m_freem(m);
2326 	return;
2327 
2328 bad:
2329 	icmp6stat.icp6s_badredirect++;
2330 	m_freem(m);
2331 }
2332 
2333 void
2334 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2335 {
2336 	struct ifnet *ifp;	/* my outgoing interface */
2337 	struct in6_addr *ifp_ll6;
2338 	struct in6_addr *router_ll6;
2339 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2340 	struct mbuf *m = NULL;	/* newly allocated one */
2341 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2342 	struct nd_redirect *nd_rd;
2343 	size_t maxlen;
2344 	u_char *p;
2345 	struct ifnet *outif = NULL;
2346 	struct sockaddr_in6 src_sa;
2347 
2348 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2349 
2350 	/* if we are not router, we don't send icmp6 redirect */
2351 	if (!ip6_forwarding || ip6_accept_rtadv)
2352 		goto fail;
2353 
2354 	/* sanity check */
2355 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2356 		goto fail;
2357 
2358 	/*
2359 	 * Address check:
2360 	 *  the source address must identify a neighbor, and
2361 	 *  the destination address must not be a multicast address
2362 	 *  [RFC 2461, sec 8.2]
2363 	 */
2364 	sip6 = mtod(m0, struct ip6_hdr *);
2365 	bzero(&src_sa, sizeof(src_sa));
2366 	src_sa.sin6_family = AF_INET6;
2367 	src_sa.sin6_len = sizeof(src_sa);
2368 	src_sa.sin6_addr = sip6->ip6_src;
2369 	/* we don't currently use sin6_scope_id, but eventually use it */
2370 	src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src);
2371 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2372 		goto fail;
2373 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2374 		goto fail;	/* what should we do here? */
2375 
2376 	/* rate limit */
2377 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2378 		goto fail;
2379 
2380 	/*
2381 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2382 	 * we almost always ask for an mbuf cluster for simplicity.
2383 	 * (MHLEN < IPV6_MMTU is almost always true)
2384 	 */
2385 #if IPV6_MMTU >= MCLBYTES
2386 # error assumption failed about IPV6_MMTU and MCLBYTES
2387 #endif
2388 	m = m_getb(IPV6_MMTU, MB_DONTWAIT, MT_HEADER, M_PKTHDR);
2389 	if (!m)
2390 		goto fail;
2391 	m->m_len = 0;
2392 	maxlen = M_TRAILINGSPACE(m);
2393 	maxlen = min(IPV6_MMTU, maxlen);
2394 	/* just for safety */
2395 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2396 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2397 		goto fail;
2398 	}
2399 
2400 	{
2401 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2402 		struct in6_ifaddr *ia;
2403 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2404 						 IN6_IFF_NOTREADY|
2405 						 IN6_IFF_ANYCAST)) == NULL)
2406 			goto fail;
2407 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2408 	}
2409 
2410 	/* get ip6 linklocal address for the router. */
2411 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2412 		struct sockaddr_in6 *sin6;
2413 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2414 		router_ll6 = &sin6->sin6_addr;
2415 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2416 			router_ll6 = NULL;
2417 	} else
2418 		router_ll6 = NULL;
2419 
2420 	/* ip6 */
2421 	ip6 = mtod(m, struct ip6_hdr *);
2422 	ip6->ip6_flow = 0;
2423 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2424 	ip6->ip6_vfc |= IPV6_VERSION;
2425 	/* ip6->ip6_plen will be set later */
2426 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2427 	ip6->ip6_hlim = 255;
2428 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2429 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2430 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2431 
2432 	/* ND Redirect */
2433 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2434 	nd_rd->nd_rd_type = ND_REDIRECT;
2435 	nd_rd->nd_rd_code = 0;
2436 	nd_rd->nd_rd_reserved = 0;
2437 	if (rt->rt_flags & RTF_GATEWAY) {
2438 		/*
2439 		 * nd_rd->nd_rd_target must be a link-local address in
2440 		 * better router cases.
2441 		 */
2442 		if (!router_ll6)
2443 			goto fail;
2444 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2445 		      sizeof(nd_rd->nd_rd_target));
2446 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2447 		      sizeof(nd_rd->nd_rd_dst));
2448 	} else {
2449 		/* make sure redtgt == reddst */
2450 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2451 		      sizeof(nd_rd->nd_rd_target));
2452 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2453 		      sizeof(nd_rd->nd_rd_dst));
2454 	}
2455 
2456 	p = (u_char *)(nd_rd + 1);
2457 
2458 	if (!router_ll6)
2459 		goto nolladdropt;
2460 
2461     {
2462 	/* target lladdr option */
2463 	struct rtentry *rt_router = NULL;
2464 	int len;
2465 	struct sockaddr_dl *sdl;
2466 	struct nd_opt_hdr *nd_opt;
2467 	char *lladdr;
2468 
2469 	rt_router = nd6_lookup(router_ll6, 0, ifp);
2470 	if (!rt_router)
2471 		goto nolladdropt;
2472 	len = sizeof(*nd_opt) + ifp->if_addrlen;
2473 	len = (len + 7) & ~7;	/* round by 8 */
2474 	/* safety check */
2475 	if (len + (p - (u_char *)ip6) > maxlen)
2476 		goto nolladdropt;
2477 	if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2478 	    (rt_router->rt_flags & RTF_LLINFO) &&
2479 	    (rt_router->rt_gateway->sa_family == AF_LINK) &&
2480 	    (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2481 	    sdl->sdl_alen) {
2482 		nd_opt = (struct nd_opt_hdr *)p;
2483 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2484 		nd_opt->nd_opt_len = len >> 3;
2485 		lladdr = (char *)(nd_opt + 1);
2486 		bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2487 		p += len;
2488 	}
2489     }
2490 nolladdropt:;
2491 
2492 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2493 
2494 	/* just to be safe */
2495 #ifdef M_DECRYPTED	/*not openbsd*/
2496 	if (m0->m_flags & M_DECRYPTED)
2497 		goto noredhdropt;
2498 #endif
2499 	if (p - (u_char *)ip6 > maxlen)
2500 		goto noredhdropt;
2501 
2502     {
2503 	/* redirected header option */
2504 	int len;
2505 	struct nd_opt_rd_hdr *nd_opt_rh;
2506 
2507 	/*
2508 	 * compute the maximum size for icmp6 redirect header option.
2509 	 * XXX room for auth header?
2510 	 */
2511 	len = maxlen - (p - (u_char *)ip6);
2512 	len &= ~7;
2513 
2514 	/* This is just for simplicity. */
2515 	if (m0->m_pkthdr.len != m0->m_len) {
2516 		if (m0->m_next) {
2517 			m_freem(m0->m_next);
2518 			m0->m_next = NULL;
2519 		}
2520 		m0->m_pkthdr.len = m0->m_len;
2521 	}
2522 
2523 	/*
2524 	 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2525 	 * about padding/truncate rule for the original IP packet.
2526 	 * From the discussion on IPv6imp in Feb 1999, the consensus was:
2527 	 * - "attach as much as possible" is the goal
2528 	 * - pad if not aligned (original size can be guessed by original
2529 	 *   ip6 header)
2530 	 * Following code adds the padding if it is simple enough,
2531 	 * and truncates if not.
2532 	 */
2533 	if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2534 		panic("assumption failed in %s:%d", __FILE__, __LINE__);
2535 
2536 	if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2537 		/* not enough room, truncate */
2538 		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2539 	} else {
2540 		/* enough room, pad or truncate */
2541 		size_t extra;
2542 
2543 		extra = m0->m_pkthdr.len % 8;
2544 		if (extra) {
2545 			/* pad if easy enough, truncate if not */
2546 			if (8 - extra <= M_TRAILINGSPACE(m0)) {
2547 				/* pad */
2548 				m0->m_len += (8 - extra);
2549 				m0->m_pkthdr.len += (8 - extra);
2550 			} else {
2551 				/* truncate */
2552 				m0->m_pkthdr.len -= extra;
2553 				m0->m_len -= extra;
2554 			}
2555 		}
2556 		len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2557 		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2558 	}
2559 
2560 	nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2561 	bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2562 	nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2563 	nd_opt_rh->nd_opt_rh_len = len >> 3;
2564 	p += sizeof(*nd_opt_rh);
2565 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2566 
2567 	/* connect m0 to m */
2568 	m->m_next = m0;
2569 	m->m_pkthdr.len = m->m_len + m0->m_len;
2570     }
2571 noredhdropt:;
2572 
2573 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
2574 		sip6->ip6_src.s6_addr16[1] = 0;
2575 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
2576 		sip6->ip6_dst.s6_addr16[1] = 0;
2577 #if 0
2578 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
2579 		ip6->ip6_src.s6_addr16[1] = 0;
2580 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
2581 		ip6->ip6_dst.s6_addr16[1] = 0;
2582 #endif
2583 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
2584 		nd_rd->nd_rd_target.s6_addr16[1] = 0;
2585 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
2586 		nd_rd->nd_rd_dst.s6_addr16[1] = 0;
2587 
2588 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2589 
2590 	nd_rd->nd_rd_cksum = 0;
2591 	nd_rd->nd_rd_cksum
2592 		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2593 
2594 	/* send the packet to outside... */
2595 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2596 	if (outif) {
2597 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2598 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2599 	}
2600 	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2601 
2602 	return;
2603 
2604 fail:
2605 	if (m)
2606 		m_freem(m);
2607 	if (m0)
2608 		m_freem(m0);
2609 }
2610 
2611 #ifdef HAVE_NRL_INPCB
2612 #define in6pcb		inpcb
2613 #define in6p_icmp6filt	inp_icmp6filt
2614 #endif
2615 /*
2616  * ICMPv6 socket option processing.
2617  */
2618 void
2619 icmp6_ctloutput(netmsg_t msg)
2620 {
2621 	struct socket *so = msg->ctloutput.base.nm_so;
2622 	struct sockopt *sopt = msg->ctloutput.nm_sopt;
2623 	int error = 0;
2624 	int optlen;
2625 	struct inpcb *inp = so->so_pcb;
2626 	int level, op, optname;
2627 
2628 	if (sopt) {
2629 		level = sopt->sopt_level;
2630 		op = sopt->sopt_dir;
2631 		optname = sopt->sopt_name;
2632 		optlen = sopt->sopt_valsize;
2633 	} else
2634 		level = op = optname = optlen = 0;
2635 
2636 	if (level != IPPROTO_ICMPV6) {
2637 		error = EINVAL;
2638 		goto out;
2639 	}
2640 
2641 	switch (op) {
2642 	case PRCO_SETOPT:
2643 		switch (optname) {
2644 		case ICMP6_FILTER:
2645 		    {
2646 			if (optlen != sizeof(struct icmp6_filter)) {
2647 				error = EMSGSIZE;
2648 				break;
2649 			}
2650 			if (inp->in6p_icmp6filt == NULL) {
2651 				error = EINVAL;
2652 				break;
2653 			}
2654 			error = soopt_to_kbuf(sopt, inp->in6p_icmp6filt, optlen,
2655 				optlen);
2656 			break;
2657 		    }
2658 
2659 		default:
2660 			error = ENOPROTOOPT;
2661 			break;
2662 		}
2663 		break;
2664 
2665 	case PRCO_GETOPT:
2666 		switch (optname) {
2667 		case ICMP6_FILTER:
2668 		    {
2669 			if (inp->in6p_icmp6filt == NULL) {
2670 				error = EINVAL;
2671 				break;
2672 			}
2673 			soopt_from_kbuf(sopt, inp->in6p_icmp6filt,
2674 				sizeof(struct icmp6_filter));
2675 			break;
2676 		    }
2677 
2678 		default:
2679 			error = ENOPROTOOPT;
2680 			break;
2681 		}
2682 		break;
2683 	}
2684 out:
2685 	lwkt_replymsg(&msg->ctloutput.base.lmsg, error);
2686 }
2687 #ifdef HAVE_NRL_INPCB
2688 #undef in6pcb
2689 #undef in6p_icmp6filt
2690 #endif
2691 
2692 #ifndef HAVE_PPSRATECHECK
2693 #ifndef timersub
2694 #define	timersub(tvp, uvp, vvp)						\
2695 	do {								\
2696 		(vvp)->tv_sec = (tvp)->tv_sec - (uvp)->tv_sec;		\
2697 		(vvp)->tv_usec = (tvp)->tv_usec - (uvp)->tv_usec;	\
2698 		if ((vvp)->tv_usec < 0) {				\
2699 			(vvp)->tv_sec--;				\
2700 			(vvp)->tv_usec += 1000000;			\
2701 		}							\
2702 	} while (0)
2703 #endif
2704 
2705 /*
2706  * ppsratecheck(): packets (or events) per second limitation.
2707  */
2708 static int
2709 ppsratecheck(struct timeval *lasttime, int *curpps,
2710 	     int maxpps)	/* maximum pps allowed */
2711 {
2712 	struct timeval tv, delta;
2713 	int rv;
2714 
2715 	microtime(&tv);
2716 
2717 	timersub(&tv, lasttime, &delta);
2718 
2719 	/*
2720 	 * Check for 0,0 so that the message will be seen at least once.
2721 	 * If more than one second has passed since the last update of
2722 	 * lasttime, reset the counter.
2723 	 *
2724 	 * We do increment *curpps even in *curpps < maxpps case, as some may
2725 	 * try to use *curpps for stat purposes as well.
2726 	 */
2727 	if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) ||
2728 	    delta.tv_sec >= 1) {
2729 		*lasttime = tv;
2730 		*curpps = 0;
2731 		rv = 1;
2732 	} else if (maxpps < 0)
2733 		rv = 1;
2734 	else if (*curpps < maxpps)
2735 		rv = 1;
2736 	else
2737 		rv = 0;
2738 
2739 #if 1 /* DIAGNOSTIC? */
2740 	/* be careful about wrap-around */
2741 	if (*curpps + 1 > *curpps)
2742 		*curpps = *curpps + 1;
2743 #else
2744 	/*
2745 	 * assume that there's not too many calls to this function.
2746 	 * not sure if the assumption holds, as it depends on *caller's*
2747 	 * behavior, not the behavior of this function.
2748 	 * IMHO it is wrong to make assumption on the caller's behavior,
2749 	 * so the above #if is #if 1, not #ifdef DIAGNOSTIC.
2750 	 */
2751 	*curpps = *curpps + 1;
2752 #endif
2753 
2754 	return (rv);
2755 }
2756 #endif
2757 
2758 /*
2759  * Perform rate limit check.
2760  * Returns 0 if it is okay to send the icmp6 packet.
2761  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2762  * limitation.
2763  *
2764  * XXX per-destination/type check necessary?
2765  */
2766 static int
2767 icmp6_ratelimit(const struct in6_addr *dst,	/* not used at this moment */
2768 		const int type,			/* not used at this moment */
2769 		const int code)			/* not used at this moment */
2770 {
2771 	int ret;
2772 
2773 	ret = 0;	/* okay to send */
2774 
2775 	/* PPS limit */
2776 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2777 	    icmp6errppslim)) {
2778 		/* The packet is subject to rate limit */
2779 		ret++;
2780 	}
2781 
2782 	return ret;
2783 }
2784