xref: /dragonfly/sys/netinet6/icmp6.c (revision 6bd457ed)
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 1.20 2005/03/04 03:48:25 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 /* function names */
132 #define in6_pcbdetach	in_pcbdetach
133 #define in6_rtchange	in_rtchange
134 
135 /*
136  * for KAME src sync over BSD*'s. XXX: FreeBSD (>=3) are VERY different from
137  * others...
138  */
139 #define in6p_ip6_nxt	inp_ipv6.ip6_nxt
140 #endif
141 
142 /*
143  * ppratecheck() is available, so define it here.
144  */
145 #define	HAVE_PPSRATECHECK
146 
147 extern struct domain inet6domain;
148 extern struct ip6protosw inet6sw[];
149 extern u_char ip6_protox[];
150 
151 struct icmp6stat icmp6stat;
152 
153 extern struct inpcbinfo ripcbinfo;
154 extern int icmp6errppslim;
155 static int icmp6errpps_count = 0;
156 static struct timeval icmp6errppslim_last;
157 extern int icmp6_nodeinfo;
158 
159 static void icmp6_errcount (struct icmp6errstat *, int, int);
160 static int icmp6_rip6_input (struct mbuf **, int);
161 static int icmp6_ratelimit (const struct in6_addr *, const int, const int);
162 static const char *icmp6_redirect_diag (struct in6_addr *,
163 	struct in6_addr *, struct in6_addr *);
164 #ifndef HAVE_PPSRATECHECK
165 static int ppsratecheck (struct timeval *, int *, int);
166 #endif
167 static struct mbuf *ni6_input (struct mbuf *, int);
168 static struct mbuf *ni6_nametodns (const char *, int, int);
169 static int ni6_dnsmatch (const char *, int, const char *, int);
170 static int ni6_addrs (struct icmp6_nodeinfo *, struct mbuf *,
171 			  struct ifnet **, char *);
172 static int ni6_store_addrs (struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
173 				struct ifnet *, int);
174 static int icmp6_notify_error (struct mbuf *, int, int, int);
175 
176 #ifdef COMPAT_RFC1885
177 static struct route_in6 icmp6_reflect_rt;
178 #endif
179 
180 
181 void
182 icmp6_init(void)
183 {
184 	mld6_init();
185 }
186 
187 static void
188 icmp6_errcount(struct icmp6errstat *stat, int type, int code)
189 {
190 	switch (type) {
191 	case ICMP6_DST_UNREACH:
192 		switch (code) {
193 		case ICMP6_DST_UNREACH_NOROUTE:
194 			stat->icp6errs_dst_unreach_noroute++;
195 			return;
196 		case ICMP6_DST_UNREACH_ADMIN:
197 			stat->icp6errs_dst_unreach_admin++;
198 			return;
199 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
200 			stat->icp6errs_dst_unreach_beyondscope++;
201 			return;
202 		case ICMP6_DST_UNREACH_ADDR:
203 			stat->icp6errs_dst_unreach_addr++;
204 			return;
205 		case ICMP6_DST_UNREACH_NOPORT:
206 			stat->icp6errs_dst_unreach_noport++;
207 			return;
208 		}
209 		break;
210 	case ICMP6_PACKET_TOO_BIG:
211 		stat->icp6errs_packet_too_big++;
212 		return;
213 	case ICMP6_TIME_EXCEEDED:
214 		switch (code) {
215 		case ICMP6_TIME_EXCEED_TRANSIT:
216 			stat->icp6errs_time_exceed_transit++;
217 			return;
218 		case ICMP6_TIME_EXCEED_REASSEMBLY:
219 			stat->icp6errs_time_exceed_reassembly++;
220 			return;
221 		}
222 		break;
223 	case ICMP6_PARAM_PROB:
224 		switch (code) {
225 		case ICMP6_PARAMPROB_HEADER:
226 			stat->icp6errs_paramprob_header++;
227 			return;
228 		case ICMP6_PARAMPROB_NEXTHEADER:
229 			stat->icp6errs_paramprob_nextheader++;
230 			return;
231 		case ICMP6_PARAMPROB_OPTION:
232 			stat->icp6errs_paramprob_option++;
233 			return;
234 		}
235 		break;
236 	case ND_REDIRECT:
237 		stat->icp6errs_redirect++;
238 		return;
239 	}
240 	stat->icp6errs_unknown++;
241 }
242 
243 /*
244  * Generate an error packet of type error in response to bad IP6 packet.
245  */
246 void
247 icmp6_error(struct mbuf *m, int type, int code, int param)
248 {
249 	struct ip6_hdr *oip6, *nip6;
250 	struct icmp6_hdr *icmp6;
251 	u_int preplen;
252 	int off;
253 	int nxt;
254 
255 	icmp6stat.icp6s_error++;
256 
257 	/* count per-type-code statistics */
258 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
259 
260 #ifdef M_DECRYPTED	/*not openbsd*/
261 	if (m->m_flags & M_DECRYPTED) {
262 		icmp6stat.icp6s_canterror++;
263 		goto freeit;
264 	}
265 #endif
266 
267 #ifndef PULLDOWN_TEST
268 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
269 #else
270 	if (m->m_len < sizeof(struct ip6_hdr)) {
271 		m = m_pullup(m, sizeof(struct ip6_hdr));
272 		if (m == NULL)
273 			return;
274 	}
275 #endif
276 	oip6 = mtod(m, struct ip6_hdr *);
277 
278 	/*
279 	 * Multicast destination check. For unrecognized option errors,
280 	 * this check has already done in ip6_unknown_opt(), so we can
281 	 * check only for other errors.
282 	 */
283 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
284 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
285 	    (type != ICMP6_PACKET_TOO_BIG &&
286 	     (type != ICMP6_PARAM_PROB ||
287 	      code != ICMP6_PARAMPROB_OPTION)))
288 		goto freeit;
289 
290 	/* Source address check. XXX: the case of anycast source? */
291 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
292 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
293 		goto freeit;
294 
295 	/*
296 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
297 	 * don't do it.
298 	 */
299 	nxt = -1;
300 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
301 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
302 		struct icmp6_hdr *icp;
303 
304 #ifndef PULLDOWN_TEST
305 		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
306 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
307 #else
308 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
309 			sizeof(*icp));
310 		if (icp == NULL) {
311 			icmp6stat.icp6s_tooshort++;
312 			return;
313 		}
314 #endif
315 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
316 		    icp->icmp6_type == ND_REDIRECT) {
317 			/*
318 			 * ICMPv6 error
319 			 * Special case: for redirect (which is
320 			 * informational) we must not send icmp6 error.
321 			 */
322 			icmp6stat.icp6s_canterror++;
323 			goto freeit;
324 		} else {
325 			/* ICMPv6 informational - send the error */
326 		}
327 	} else {
328 		/* non-ICMPv6 - send the error */
329 	}
330 
331 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
332 
333 	/* Finally, do rate limitation check. */
334 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
335 		icmp6stat.icp6s_toofreq++;
336 		goto freeit;
337 	}
338 
339 	/*
340 	 * OK, ICMP6 can be generated.
341 	 */
342 
343 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
344 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
345 
346 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
347 	M_PREPEND(m, preplen, MB_DONTWAIT);
348 	if (m && m->m_len < preplen)
349 		m = m_pullup(m, preplen);
350 	if (m == NULL) {
351 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
352 		return;
353 	}
354 
355 	nip6 = mtod(m, struct ip6_hdr *);
356 	nip6->ip6_src  = oip6->ip6_src;
357 	nip6->ip6_dst  = oip6->ip6_dst;
358 
359 	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src))
360 		oip6->ip6_src.s6_addr16[1] = 0;
361 	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst))
362 		oip6->ip6_dst.s6_addr16[1] = 0;
363 
364 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
365 	icmp6->icmp6_type = type;
366 	icmp6->icmp6_code = code;
367 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
368 
369 	/*
370 	 * icmp6_reflect() is designed to be in the input path.
371 	 * icmp6_error() can be called from both input and outut path,
372 	 * and if we are in output path rcvif could contain bogus value.
373 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
374 	 * information in ip header (nip6).
375 	 */
376 	m->m_pkthdr.rcvif = NULL;
377 
378 	icmp6stat.icp6s_outhist[type]++;
379 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
380 
381 	return;
382 
383   freeit:
384 	/*
385 	 * If we can't tell wheter or not we can generate ICMP6, free it.
386 	 */
387 	m_freem(m);
388 }
389 
390 /*
391  * Process a received ICMP6 message.
392  */
393 int
394 icmp6_input(struct mbuf **mp, int *offp, int proto)
395 {
396 	struct mbuf *m = *mp, *n;
397 	struct ip6_hdr *ip6, *nip6;
398 	struct icmp6_hdr *icmp6, *nicmp6;
399 	int off = *offp;
400 	int icmp6len = m->m_pkthdr.len - *offp;
401 	int code, sum, noff;
402 
403 #ifndef PULLDOWN_TEST
404 	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
405 	/* m might change if M_LOOP.  So, call mtod after this */
406 #endif
407 
408 	/*
409 	 * Locate icmp6 structure in mbuf, and check
410 	 * that not corrupted and of at least minimum length
411 	 */
412 
413 	ip6 = mtod(m, struct ip6_hdr *);
414 	if (icmp6len < sizeof(struct icmp6_hdr)) {
415 		icmp6stat.icp6s_tooshort++;
416 		goto freeit;
417 	}
418 
419 	/*
420 	 * calculate the checksum
421 	 */
422 #ifndef PULLDOWN_TEST
423 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
424 #else
425 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
426 	if (icmp6 == NULL) {
427 		icmp6stat.icp6s_tooshort++;
428 		return IPPROTO_DONE;
429 	}
430 #endif
431 	code = icmp6->icmp6_code;
432 
433 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
434 		nd6log((LOG_ERR,
435 		    "ICMP6 checksum error(%d|%x) %s\n",
436 		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
437 		icmp6stat.icp6s_checksum++;
438 		goto freeit;
439 	}
440 
441 	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
442 		/*
443 		 * Deliver very specific ICMP6 type only.
444 		 * This is important to deilver TOOBIG.  Otherwise PMTUD
445 		 * will not work.
446 		 */
447 		switch (icmp6->icmp6_type) {
448 		case ICMP6_DST_UNREACH:
449 		case ICMP6_PACKET_TOO_BIG:
450 		case ICMP6_TIME_EXCEEDED:
451 			break;
452 		default:
453 			goto freeit;
454 		}
455 	}
456 
457 	icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
458 	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
459 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
460 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
461 
462 	switch (icmp6->icmp6_type) {
463 	case ICMP6_DST_UNREACH:
464 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
465 		switch (code) {
466 		case ICMP6_DST_UNREACH_NOROUTE:
467 			code = PRC_UNREACH_NET;
468 			break;
469 		case ICMP6_DST_UNREACH_ADMIN:
470 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
471 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
472 			break;
473 		case ICMP6_DST_UNREACH_ADDR:
474 			code = PRC_HOSTDEAD;
475 			break;
476 #ifdef COMPAT_RFC1885
477 		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
478 			code = PRC_UNREACH_SRCFAIL;
479 			break;
480 #else
481 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
482 			/* I mean "source address was incorrect." */
483 			code = PRC_PARAMPROB;
484 			break;
485 #endif
486 		case ICMP6_DST_UNREACH_NOPORT:
487 			code = PRC_UNREACH_PORT;
488 			break;
489 		default:
490 			goto badcode;
491 		}
492 		goto deliver;
493 		break;
494 
495 	case ICMP6_PACKET_TOO_BIG:
496 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
497 		if (code != 0)
498 			goto badcode;
499 
500 		code = PRC_MSGSIZE;
501 
502 		/*
503 		 * Updating the path MTU will be done after examining
504 		 * intermediate extension headers.
505 		 */
506 		goto deliver;
507 		break;
508 
509 	case ICMP6_TIME_EXCEEDED:
510 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
511 		switch (code) {
512 		case ICMP6_TIME_EXCEED_TRANSIT:
513 		case ICMP6_TIME_EXCEED_REASSEMBLY:
514 			code += PRC_TIMXCEED_INTRANS;
515 			break;
516 		default:
517 			goto badcode;
518 		}
519 		goto deliver;
520 		break;
521 
522 	case ICMP6_PARAM_PROB:
523 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
524 		switch (code) {
525 		case ICMP6_PARAMPROB_NEXTHEADER:
526 			code = PRC_UNREACH_PROTOCOL;
527 			break;
528 		case ICMP6_PARAMPROB_HEADER:
529 		case ICMP6_PARAMPROB_OPTION:
530 			code = PRC_PARAMPROB;
531 			break;
532 		default:
533 			goto badcode;
534 		}
535 		goto deliver;
536 		break;
537 
538 	case ICMP6_ECHO_REQUEST:
539 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
540 		if (code != 0)
541 			goto badcode;
542 		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
543 			/* Give up remote */
544 			break;
545 		}
546 		if ((n->m_flags & M_EXT) != 0
547 		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
548 			struct mbuf *n0 = n;
549 			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
550 			int n0len;
551 
552 			/*
553 			 * Prepare an internal mbuf. m_pullup() doesn't
554 			 * always copy the length we specified.
555 			 */
556 			if (maxlen >= MCLBYTES) {
557 				/* Give up remote */
558 				m_freem(n0);
559 				break;
560 			}
561 			MGETHDR(n, MB_DONTWAIT, n0->m_type);
562 			n0len = n0->m_pkthdr.len;	/* save for use below */
563 			if (n)
564 				M_MOVE_PKTHDR(n, n0);
565 			if (n && maxlen >= MHLEN) {
566 				MCLGET(n, MB_DONTWAIT);
567 				if ((n->m_flags & M_EXT) == 0) {
568 					m_free(n);
569 					n = NULL;
570 				}
571 			}
572 			if (n == NULL) {
573 				/* Give up remote */
574 				m_freem(n0);
575 				break;
576 			}
577 			/*
578 			 * Copy IPv6 and ICMPv6 only.
579 			 */
580 			nip6 = mtod(n, struct ip6_hdr *);
581 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
582 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
583 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
584 			noff = sizeof(struct ip6_hdr);
585 			/* new mbuf contains only ipv6+icmpv6 headers */
586 			n->m_len = noff + sizeof(struct icmp6_hdr);
587 			/*
588 			 * Adjust mbuf. ip6_plen will be adjusted in
589 			 * ip6_output().
590 			 */
591 			m_adj(n0, off + sizeof(struct icmp6_hdr));
592 			/* recalculate complete packet size */
593 			n->m_pkthdr.len = n0len + (noff - off);
594 			n->m_next = n0;
595 		} else {
596 			nip6 = mtod(n, struct ip6_hdr *);
597 			nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
598 			noff = off;
599 		}
600 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
601 		nicmp6->icmp6_code = 0;
602 		if (n) {
603 			icmp6stat.icp6s_reflect++;
604 			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
605 			icmp6_reflect(n, noff);
606 		}
607 		break;
608 
609 	case ICMP6_ECHO_REPLY:
610 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
611 		if (code != 0)
612 			goto badcode;
613 		break;
614 
615 	case MLD_LISTENER_QUERY:
616 	case MLD_LISTENER_REPORT:
617 		if (icmp6len < sizeof(struct mld_hdr))
618 			goto badlen;
619 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
620 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
621 		else
622 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
623 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
624 			/* give up local */
625 			mld6_input(m, off);
626 			m = NULL;
627 			goto freeit;
628 		}
629 		mld6_input(n, off);
630 		/* m stays. */
631 		break;
632 
633 	case MLD_LISTENER_DONE:
634 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
635 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
636 			goto badlen;
637 		break;		/* nothing to be done in kernel */
638 
639 	case MLD_MTRACE_RESP:
640 	case MLD_MTRACE:
641 		/* XXX: these two are experimental.  not officially defind. */
642 		/* XXX: per-interface statistics? */
643 		break;		/* just pass it to applications */
644 
645 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
646 	    {
647 		enum { WRU, FQDN } mode;
648 
649 		if (!icmp6_nodeinfo)
650 			break;
651 
652 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
653 			mode = WRU;
654 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
655 			mode = FQDN;
656 		else
657 			goto badlen;
658 
659 #define hostnamelen	strlen(hostname)
660 		if (mode == FQDN) {
661 #ifndef PULLDOWN_TEST
662 			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
663 					 IPPROTO_DONE);
664 #endif
665 			n = m_copy(m, 0, M_COPYALL);
666 			if (n)
667 				n = ni6_input(n, off);
668 			/* XXX meaningless if n == NULL */
669 			noff = sizeof(struct ip6_hdr);
670 		} else {
671 			u_char *p;
672 			int maxlen, maxhlen;
673 
674 			if ((icmp6_nodeinfo & 5) != 5)
675 				break;
676 
677 			if (code != 0)
678 				goto badcode;
679 			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
680 			if (maxlen >= MCLBYTES) {
681 				/* Give up remote */
682 				break;
683 			}
684 			MGETHDR(n, MB_DONTWAIT, m->m_type);
685 			if (n && maxlen > MHLEN) {
686 				MCLGET(n, MB_DONTWAIT);
687 				if ((n->m_flags & M_EXT) == 0) {
688 					m_free(n);
689 					n = NULL;
690 				}
691 			}
692 			if (!m_dup_pkthdr(n, m, MB_DONTWAIT)) {
693 				/*
694 				 * Previous code did a blind M_COPY_PKTHDR
695 				 * and said "just for rcvif".  If true, then
696 				 * we could tolerate the dup failing (due to
697 				 * the deep copy of the tag chain).  For now
698 				 * be conservative and just fail.
699 				 */
700 				m_free(n);
701 				n = NULL;
702 			}
703 			if (n == NULL) {
704 				/* Give up remote */
705 				break;
706 			}
707 			n->m_pkthdr.rcvif = NULL;
708 			n->m_len = 0;
709 			maxhlen = M_TRAILINGSPACE(n) - maxlen;
710 			if (maxhlen > hostnamelen)
711 				maxhlen = hostnamelen;
712 			/*
713 			 * Copy IPv6 and ICMPv6 only.
714 			 */
715 			nip6 = mtod(n, struct ip6_hdr *);
716 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
717 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
718 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
719 			p = (u_char *)(nicmp6 + 1);
720 			bzero(p, 4);
721 			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
722 			noff = sizeof(struct ip6_hdr);
723 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
724 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
725 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
726 			nicmp6->icmp6_code = 0;
727 		}
728 #undef hostnamelen
729 		if (n) {
730 			icmp6stat.icp6s_reflect++;
731 			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
732 			icmp6_reflect(n, noff);
733 		}
734 		break;
735 	    }
736 
737 	case ICMP6_WRUREPLY:
738 		if (code != 0)
739 			goto badcode;
740 		break;
741 
742 	case ND_ROUTER_SOLICIT:
743 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
744 		if (code != 0)
745 			goto badcode;
746 		if (icmp6len < sizeof(struct nd_router_solicit))
747 			goto badlen;
748 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
749 			/* give up local */
750 			nd6_rs_input(m, off, icmp6len);
751 			m = NULL;
752 			goto freeit;
753 		}
754 		nd6_rs_input(n, off, icmp6len);
755 		/* m stays. */
756 		break;
757 
758 	case ND_ROUTER_ADVERT:
759 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
760 		if (code != 0)
761 			goto badcode;
762 		if (icmp6len < sizeof(struct nd_router_advert))
763 			goto badlen;
764 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
765 			/* give up local */
766 			nd6_ra_input(m, off, icmp6len);
767 			m = NULL;
768 			goto freeit;
769 		}
770 		nd6_ra_input(n, off, icmp6len);
771 		/* m stays. */
772 		break;
773 
774 	case ND_NEIGHBOR_SOLICIT:
775 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
776 		if (code != 0)
777 			goto badcode;
778 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
779 			goto badlen;
780 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
781 			/* give up local */
782 			nd6_ns_input(m, off, icmp6len);
783 			m = NULL;
784 			goto freeit;
785 		}
786 		nd6_ns_input(n, off, icmp6len);
787 		/* m stays. */
788 		break;
789 
790 	case ND_NEIGHBOR_ADVERT:
791 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
792 		if (code != 0)
793 			goto badcode;
794 		if (icmp6len < sizeof(struct nd_neighbor_advert))
795 			goto badlen;
796 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
797 			/* give up local */
798 			nd6_na_input(m, off, icmp6len);
799 			m = NULL;
800 			goto freeit;
801 		}
802 		nd6_na_input(n, off, icmp6len);
803 		/* m stays. */
804 		break;
805 
806 	case ND_REDIRECT:
807 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
808 		if (code != 0)
809 			goto badcode;
810 		if (icmp6len < sizeof(struct nd_redirect))
811 			goto badlen;
812 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
813 			/* give up local */
814 			icmp6_redirect_input(m, off);
815 			m = NULL;
816 			goto freeit;
817 		}
818 		icmp6_redirect_input(n, off);
819 		/* m stays. */
820 		break;
821 
822 	case ICMP6_ROUTER_RENUMBERING:
823 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
824 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
825 			goto badcode;
826 		if (icmp6len < sizeof(struct icmp6_router_renum))
827 			goto badlen;
828 		break;
829 
830 	default:
831 		nd6log((LOG_DEBUG,
832 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
833 		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
834 		    ip6_sprintf(&ip6->ip6_dst),
835 		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
836 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
837 			/* ICMPv6 error: MUST deliver it by spec... */
838 			code = PRC_NCMDS;
839 			/* deliver */
840 		} else {
841 			/* ICMPv6 informational: MUST not deliver */
842 			break;
843 		}
844 	deliver:
845 		if (icmp6_notify_error(m, off, icmp6len, code)) {
846 			/* In this case, m should've been freed. */
847 			return(IPPROTO_DONE);
848 		}
849 		break;
850 
851 	badcode:
852 		icmp6stat.icp6s_badcode++;
853 		break;
854 
855 	badlen:
856 		icmp6stat.icp6s_badlen++;
857 		break;
858 	}
859 
860 	/* deliver the packet to appropriate sockets */
861 	icmp6_rip6_input(&m, *offp);
862 
863 	return IPPROTO_DONE;
864 
865  freeit:
866 	m_freem(m);
867 	return IPPROTO_DONE;
868 }
869 
870 static int
871 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
872 {
873 	struct icmp6_hdr *icmp6;
874 	struct ip6_hdr *eip6;
875 	u_int32_t notifymtu;
876 	struct sockaddr_in6 icmp6src, icmp6dst;
877 
878 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
879 		icmp6stat.icp6s_tooshort++;
880 		goto freeit;
881 	}
882 #ifndef PULLDOWN_TEST
883 	IP6_EXTHDR_CHECK(m, off,
884 			 sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr),
885 			 -1);
886 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
887 #else
888 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
889 		       sizeof(*icmp6) + sizeof(struct ip6_hdr));
890 	if (icmp6 == NULL) {
891 		icmp6stat.icp6s_tooshort++;
892 		return(-1);
893 	}
894 #endif
895 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
896 
897 	/* Detect the upper level protocol */
898 	{
899 		void (*ctlfunc) (int, struct sockaddr *, void *);
900 		u_int8_t nxt = eip6->ip6_nxt;
901 		int eoff = off + sizeof(struct icmp6_hdr) +
902 			sizeof(struct ip6_hdr);
903 		struct ip6ctlparam ip6cp;
904 		struct in6_addr *finaldst = NULL;
905 		int icmp6type = icmp6->icmp6_type;
906 		struct ip6_frag *fh;
907 		struct ip6_rthdr *rth;
908 		struct ip6_rthdr0 *rth0;
909 		int rthlen;
910 
911 		while (1) { /* XXX: should avoid infinite loop explicitly? */
912 			struct ip6_ext *eh;
913 
914 			switch (nxt) {
915 			case IPPROTO_HOPOPTS:
916 			case IPPROTO_DSTOPTS:
917 			case IPPROTO_AH:
918 #ifndef PULLDOWN_TEST
919 				IP6_EXTHDR_CHECK(m, 0, eoff +
920 						 sizeof(struct ip6_ext),
921 						 -1);
922 				eh = (struct ip6_ext *)(mtod(m, caddr_t)
923 							+ eoff);
924 #else
925 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
926 					       eoff, sizeof(*eh));
927 				if (eh == NULL) {
928 					icmp6stat.icp6s_tooshort++;
929 					return(-1);
930 				}
931 #endif
932 
933 				if (nxt == IPPROTO_AH)
934 					eoff += (eh->ip6e_len + 2) << 2;
935 				else
936 					eoff += (eh->ip6e_len + 1) << 3;
937 				nxt = eh->ip6e_nxt;
938 				break;
939 			case IPPROTO_ROUTING:
940 				/*
941 				 * When the erroneous packet contains a
942 				 * routing header, we should examine the
943 				 * header to determine the final destination.
944 				 * Otherwise, we can't properly update
945 				 * information that depends on the final
946 				 * destination (e.g. path MTU).
947 				 */
948 #ifndef PULLDOWN_TEST
949 				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth),
950 						 -1);
951 				rth = (struct ip6_rthdr *)(mtod(m, caddr_t)
952 							   + eoff);
953 #else
954 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
955 					       eoff, sizeof(*rth));
956 				if (rth == NULL) {
957 					icmp6stat.icp6s_tooshort++;
958 					return(-1);
959 				}
960 #endif
961 				rthlen = (rth->ip6r_len + 1) << 3;
962 				/*
963 				 * XXX: currently there is no
964 				 * officially defined type other
965 				 * than type-0.
966 				 * Note that if the segment left field
967 				 * is 0, all intermediate hops must
968 				 * have been passed.
969 				 */
970 				if (rth->ip6r_segleft &&
971 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
972 					int hops;
973 
974 #ifndef PULLDOWN_TEST
975 					IP6_EXTHDR_CHECK(m, 0, eoff + rthlen,
976 							 -1);
977 					rth0 = (struct ip6_rthdr0 *)(mtod(m, caddr_t) + eoff);
978 #else
979 					IP6_EXTHDR_GET(rth0,
980 						       struct ip6_rthdr0 *, m,
981 						       eoff, rthlen);
982 					if (rth0 == NULL) {
983 						icmp6stat.icp6s_tooshort++;
984 						return(-1);
985 					}
986 #endif
987 					/* just ignore a bogus header */
988 					if ((rth0->ip6r0_len % 2) == 0 &&
989 					    (hops = rth0->ip6r0_len/2))
990 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
991 				}
992 				eoff += rthlen;
993 				nxt = rth->ip6r_nxt;
994 				break;
995 			case IPPROTO_FRAGMENT:
996 #ifndef PULLDOWN_TEST
997 				IP6_EXTHDR_CHECK(m, 0, eoff +
998 						 sizeof(struct ip6_frag),
999 						 -1);
1000 				fh = (struct ip6_frag *)(mtod(m, caddr_t)
1001 							 + eoff);
1002 #else
1003 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
1004 					       eoff, sizeof(*fh));
1005 				if (fh == NULL) {
1006 					icmp6stat.icp6s_tooshort++;
1007 					return(-1);
1008 				}
1009 #endif
1010 				/*
1011 				 * Data after a fragment header is meaningless
1012 				 * unless it is the first fragment, but
1013 				 * we'll go to the notify label for path MTU
1014 				 * discovery.
1015 				 */
1016 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1017 					goto notify;
1018 
1019 				eoff += sizeof(struct ip6_frag);
1020 				nxt = fh->ip6f_nxt;
1021 				break;
1022 			default:
1023 				/*
1024 				 * This case includes ESP and the No Next
1025 				 * Header.  In such cases going to the notify
1026 				 * label does not have any meaning
1027 				 * (i.e. ctlfunc will be NULL), but we go
1028 				 * anyway since we might have to update
1029 				 * path MTU information.
1030 				 */
1031 				goto notify;
1032 			}
1033 		}
1034 	  notify:
1035 #ifndef PULLDOWN_TEST
1036 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1037 #else
1038 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1039 			       sizeof(*icmp6) + sizeof(struct ip6_hdr));
1040 		if (icmp6 == NULL) {
1041 			icmp6stat.icp6s_tooshort++;
1042 			return(-1);
1043 		}
1044 #endif
1045 
1046 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1047 		bzero(&icmp6dst, sizeof(icmp6dst));
1048 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1049 		icmp6dst.sin6_family = AF_INET6;
1050 		if (finaldst == NULL)
1051 			icmp6dst.sin6_addr = eip6->ip6_dst;
1052 		else
1053 			icmp6dst.sin6_addr = *finaldst;
1054 		icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1055 							  &icmp6dst.sin6_addr);
1056 		if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1057 				   NULL, NULL)) {
1058 			/* should be impossbile */
1059 			nd6log((LOG_DEBUG,
1060 			    "icmp6_notify_error: in6_embedscope failed\n"));
1061 			goto freeit;
1062 		}
1063 
1064 		/*
1065 		 * retrieve parameters from the inner IPv6 header, and convert
1066 		 * them into sockaddr structures.
1067 		 */
1068 		bzero(&icmp6src, sizeof(icmp6src));
1069 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1070 		icmp6src.sin6_family = AF_INET6;
1071 		icmp6src.sin6_addr = eip6->ip6_src;
1072 		icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1073 							  &icmp6src.sin6_addr);
1074 		if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1075 				   NULL, NULL)) {
1076 			/* should be impossbile */
1077 			nd6log((LOG_DEBUG,
1078 			    "icmp6_notify_error: in6_embedscope failed\n"));
1079 			goto freeit;
1080 		}
1081 		icmp6src.sin6_flowinfo =
1082 			(eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1083 
1084 		if (finaldst == NULL)
1085 			finaldst = &eip6->ip6_dst;
1086 		ip6cp.ip6c_m = m;
1087 		ip6cp.ip6c_icmp6 = icmp6;
1088 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1089 		ip6cp.ip6c_off = eoff;
1090 		ip6cp.ip6c_finaldst = finaldst;
1091 		ip6cp.ip6c_src = &icmp6src;
1092 		ip6cp.ip6c_nxt = nxt;
1093 
1094 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1095 			notifymtu = ntohl(icmp6->icmp6_mtu);
1096 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1097 			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1098 		}
1099 
1100 		ctlfunc = (void (*) (int, struct sockaddr *, void *))
1101 			(inet6sw[ip6_protox[nxt]].pr_ctlinput);
1102 		if (ctlfunc) {
1103 			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1104 					  &ip6cp);
1105 		}
1106 	}
1107 	return(0);
1108 
1109   freeit:
1110 	m_freem(m);
1111 	return(-1);
1112 }
1113 
1114 void
1115 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1116 {
1117 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1118 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1119 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1120 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1121 	struct rtentry *rt = NULL;
1122 	struct sockaddr_in6 sin6;
1123 
1124 	if (!validated)
1125 		return;
1126 
1127 	bzero(&sin6, sizeof(sin6));
1128 	sin6.sin6_family = PF_INET6;
1129 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1130 	sin6.sin6_addr = *dst;
1131 	/* XXX normally, this won't happen */
1132 	if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1133 		sin6.sin6_addr.s6_addr16[1] =
1134 		    htons(m->m_pkthdr.rcvif->if_index);
1135 	}
1136 	/* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1137 	rt = rtpurelookup((struct sockaddr *)&sin6);
1138 	if (rt != NULL && (rt->rt_flags & RTF_HOST) &&
1139 	    !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1140 		if (mtu < IPV6_MMTU) {				/* XXX */
1141 			rt->rt_rmx.rmx_locks |= RTV_MTU;
1142 		} else if (mtu < rt->rt_ifp->if_mtu &&
1143 			   rt->rt_rmx.rmx_mtu > mtu) {
1144 			icmp6stat.icp6s_pmtuchg++;
1145 			rt->rt_rmx.rmx_mtu = mtu;
1146 		}
1147 	}
1148 	if (rt)
1149 		--rt->rt_refcnt;
1150 }
1151 
1152 /*
1153  * Process a Node Information Query packet, based on
1154  * draft-ietf-ipngwg-icmp-name-lookups-07.
1155  *
1156  * Spec incompatibilities:
1157  * - IPv6 Subject address handling
1158  * - IPv4 Subject address handling support missing
1159  * - Proxy reply (answer even if it's not for me)
1160  * - joins NI group address at in6_ifattach() time only, does not cope
1161  *   with hostname changes by sethostname(3)
1162  */
1163 #define hostnamelen	strlen(hostname)
1164 static struct mbuf *
1165 ni6_input(struct mbuf *m, int off)
1166 {
1167 	struct icmp6_nodeinfo *ni6, *nni6;
1168 	struct mbuf *n = NULL;
1169 	u_int16_t qtype;
1170 	int subjlen;
1171 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1172 	struct ni_reply_fqdn *fqdn;
1173 	int addrs;		/* for NI_QTYPE_NODEADDR */
1174 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1175 	struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1176 	struct sockaddr_in6 sin6_d; /* XXX: we should retrieve this from m_aux */
1177 	struct ip6_hdr *ip6;
1178 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1179 	char *subj = NULL;
1180 	struct in6_ifaddr *ia6 = NULL;
1181 
1182 	ip6 = mtod(m, struct ip6_hdr *);
1183 #ifndef PULLDOWN_TEST
1184 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1185 #else
1186 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1187 	if (ni6 == NULL) {
1188 		/* m is already reclaimed */
1189 		return NULL;
1190 	}
1191 #endif
1192 
1193 	/*
1194 	 * Validate IPv6 destination address.
1195 	 *
1196 	 * The Responder must discard the Query without further processing
1197 	 * unless it is one of the Responder's unicast or anycast addresses, or
1198 	 * a link-local scope multicast address which the Responder has joined.
1199 	 * [icmp-name-lookups-07, Section 4.]
1200 	 */
1201 	bzero(&sin6, sizeof(sin6));
1202 	sin6.sin6_family = AF_INET6;
1203 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1204 	bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1205 	/* XXX scopeid */
1206 	if ((ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)&sin6)) != NULL) {
1207 		/* unicast/anycast, fine */
1208 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1209 		    (icmp6_nodeinfo & 4) == 0) {
1210 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1211 				"a temporary address in %s:%d",
1212 			       __FILE__, __LINE__));
1213 			goto bad;
1214 		}
1215 	} else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1216 		; /* link-local multicast, fine */
1217 	else
1218 		goto bad;
1219 
1220 	/* validate query Subject field. */
1221 	qtype = ntohs(ni6->ni_qtype);
1222 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1223 	switch (qtype) {
1224 	case NI_QTYPE_NOOP:
1225 	case NI_QTYPE_SUPTYPES:
1226 		/* 07 draft */
1227 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1228 			break;
1229 		/* FALLTHROUGH */
1230 	case NI_QTYPE_FQDN:
1231 	case NI_QTYPE_NODEADDR:
1232 		switch (ni6->ni_code) {
1233 		case ICMP6_NI_SUBJ_IPV6:
1234 #if ICMP6_NI_SUBJ_IPV6 != 0
1235 		case 0:
1236 #endif
1237 			/*
1238 			 * backward compatibility - try to accept 03 draft
1239 			 * format, where no Subject is present.
1240 			 */
1241 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1242 			    subjlen == 0) {
1243 				oldfqdn++;
1244 				break;
1245 			}
1246 #if ICMP6_NI_SUBJ_IPV6 != 0
1247 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1248 				goto bad;
1249 #endif
1250 
1251 			if (subjlen != sizeof(sin6.sin6_addr))
1252 				goto bad;
1253 
1254 			/*
1255 			 * Validate Subject address.
1256 			 *
1257 			 * Not sure what exactly "address belongs to the node"
1258 			 * means in the spec, is it just unicast, or what?
1259 			 *
1260 			 * At this moment we consider Subject address as
1261 			 * "belong to the node" if the Subject address equals
1262 			 * to the IPv6 destination address; validation for
1263 			 * IPv6 destination address should have done enough
1264 			 * check for us.
1265 			 *
1266 			 * We do not do proxy at this moment.
1267 			 */
1268 			/* m_pulldown instead of copy? */
1269 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1270 			    subjlen, (caddr_t)&sin6.sin6_addr);
1271 			sin6.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1272 							      &sin6.sin6_addr);
1273 			in6_embedscope(&sin6.sin6_addr, &sin6, NULL, NULL);
1274 			bzero(&sin6_d, sizeof(sin6_d));
1275 			sin6_d.sin6_family = AF_INET6; /* not used, actually */
1276 			sin6_d.sin6_len = sizeof(sin6_d); /* ditto */
1277 			sin6_d.sin6_addr = ip6->ip6_dst;
1278 			sin6_d.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1279 								&ip6->ip6_dst);
1280 			in6_embedscope(&sin6_d.sin6_addr, &sin6_d, NULL, NULL);
1281 			subj = (char *)&sin6;
1282 			if (SA6_ARE_ADDR_EQUAL(&sin6, &sin6_d))
1283 				break;
1284 
1285 			/*
1286 			 * XXX if we are to allow other cases, we should really
1287 			 * be careful about scope here.
1288 			 * basically, we should disallow queries toward IPv6
1289 			 * destination X with subject Y, if scope(X) > scope(Y).
1290 			 * if we allow scope(X) > scope(Y), it will result in
1291 			 * information leakage across scope boundary.
1292 			 */
1293 			goto bad;
1294 
1295 		case ICMP6_NI_SUBJ_FQDN:
1296 			/*
1297 			 * Validate Subject name with gethostname(3).
1298 			 *
1299 			 * The behavior may need some debate, since:
1300 			 * - we are not sure if the node has FQDN as
1301 			 *   hostname (returned by gethostname(3)).
1302 			 * - the code does wildcard match for truncated names.
1303 			 *   however, we are not sure if we want to perform
1304 			 *   wildcard match, if gethostname(3) side has
1305 			 *   truncated hostname.
1306 			 */
1307 			n = ni6_nametodns(hostname, hostnamelen, 0);
1308 			if (!n || n->m_next || n->m_len == 0)
1309 				goto bad;
1310 			IP6_EXTHDR_GET(subj, char *, m,
1311 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1312 			if (subj == NULL)
1313 				goto bad;
1314 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1315 					n->m_len)) {
1316 				goto bad;
1317 			}
1318 			m_freem(n);
1319 			n = NULL;
1320 			break;
1321 
1322 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1323 		default:
1324 			goto bad;
1325 		}
1326 		break;
1327 	}
1328 
1329 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1330 	switch (qtype) {
1331 	case NI_QTYPE_FQDN:
1332 		if ((icmp6_nodeinfo & 1) == 0)
1333 			goto bad;
1334 		break;
1335 	case NI_QTYPE_NODEADDR:
1336 		if ((icmp6_nodeinfo & 2) == 0)
1337 			goto bad;
1338 		break;
1339 	}
1340 
1341 	/* guess reply length */
1342 	switch (qtype) {
1343 	case NI_QTYPE_NOOP:
1344 		break;		/* no reply data */
1345 	case NI_QTYPE_SUPTYPES:
1346 		replylen += sizeof(u_int32_t);
1347 		break;
1348 	case NI_QTYPE_FQDN:
1349 		/* XXX will append an mbuf */
1350 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1351 		break;
1352 	case NI_QTYPE_NODEADDR:
1353 		addrs = ni6_addrs(ni6, m, &ifp, subj);
1354 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1355 					  sizeof(u_int32_t))) > MCLBYTES)
1356 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1357 		break;
1358 	default:
1359 		/*
1360 		 * XXX: We must return a reply with the ICMP6 code
1361 		 * `unknown Qtype' in this case. However we regard the case
1362 		 * as an FQDN query for backward compatibility.
1363 		 * Older versions set a random value to this field,
1364 		 * so it rarely varies in the defined qtypes.
1365 		 * But the mechanism is not reliable...
1366 		 * maybe we should obsolete older versions.
1367 		 */
1368 		qtype = NI_QTYPE_FQDN;
1369 		/* XXX will append an mbuf */
1370 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1371 		oldfqdn++;
1372 		break;
1373 	}
1374 
1375 	/* allocate an mbuf to reply. */
1376 	MGETHDR(n, MB_DONTWAIT, m->m_type);
1377 	if (n == NULL) {
1378 		m_freem(m);
1379 		return(NULL);
1380 	}
1381 	M_MOVE_PKTHDR(n, m); /* just for recvif */
1382 	if (replylen > MHLEN) {
1383 		if (replylen > MCLBYTES) {
1384 			/*
1385 			 * XXX: should we try to allocate more? But MCLBYTES
1386 			 * is probably much larger than IPV6_MMTU...
1387 			 */
1388 			goto bad;
1389 		}
1390 		MCLGET(n, MB_DONTWAIT);
1391 		if ((n->m_flags & M_EXT) == 0) {
1392 			goto bad;
1393 		}
1394 	}
1395 	n->m_pkthdr.len = n->m_len = replylen;
1396 
1397 	/* copy mbuf header and IPv6 + Node Information base headers */
1398 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1399 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1400 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1401 
1402 	/* qtype dependent procedure */
1403 	switch (qtype) {
1404 	case NI_QTYPE_NOOP:
1405 		nni6->ni_code = ICMP6_NI_SUCCESS;
1406 		nni6->ni_flags = 0;
1407 		break;
1408 	case NI_QTYPE_SUPTYPES:
1409 	{
1410 		u_int32_t v;
1411 		nni6->ni_code = ICMP6_NI_SUCCESS;
1412 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1413 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1414 		v = (u_int32_t)htonl(0x0000000f);
1415 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1416 		break;
1417 	}
1418 	case NI_QTYPE_FQDN:
1419 		nni6->ni_code = ICMP6_NI_SUCCESS;
1420 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1421 						sizeof(struct ip6_hdr) +
1422 						sizeof(struct icmp6_nodeinfo));
1423 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1424 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1425 		/*
1426 		 * XXX do we really have FQDN in variable "hostname"?
1427 		 */
1428 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1429 		if (n->m_next == NULL)
1430 			goto bad;
1431 		/* XXX we assume that n->m_next is not a chain */
1432 		if (n->m_next->m_next != NULL)
1433 			goto bad;
1434 		n->m_pkthdr.len += n->m_next->m_len;
1435 		break;
1436 	case NI_QTYPE_NODEADDR:
1437 	{
1438 		int lenlim, copied;
1439 
1440 		nni6->ni_code = ICMP6_NI_SUCCESS;
1441 		n->m_pkthdr.len = n->m_len =
1442 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1443 		lenlim = M_TRAILINGSPACE(n);
1444 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1445 		/* XXX: reset mbuf length */
1446 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1447 			sizeof(struct icmp6_nodeinfo) + copied;
1448 		break;
1449 	}
1450 	default:
1451 		break;		/* XXX impossible! */
1452 	}
1453 
1454 	nni6->ni_type = ICMP6_NI_REPLY;
1455 	m_freem(m);
1456 	return(n);
1457 
1458   bad:
1459 	m_freem(m);
1460 	if (n)
1461 		m_freem(n);
1462 	return(NULL);
1463 }
1464 #undef hostnamelen
1465 
1466 /*
1467  * make a mbuf with DNS-encoded string.  no compression support.
1468  *
1469  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1470  * treated as truncated name (two \0 at the end).  this is a wild guess.
1471  */
1472 static struct mbuf *
1473 ni6_nametodns(const char *name, int namelen,
1474 	      int old) /* return pascal string if non-zero */
1475 {
1476 	struct mbuf *m;
1477 	char *cp, *ep;
1478 	const char *p, *q;
1479 	int i, len, nterm;
1480 
1481 	if (old)
1482 		len = namelen + 1;
1483 	else
1484 		len = MCLBYTES;
1485 
1486 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1487 	MGET(m, MB_DONTWAIT, MT_DATA);
1488 	if (m && len > MLEN) {
1489 		MCLGET(m, MB_DONTWAIT);
1490 		if ((m->m_flags & M_EXT) == 0)
1491 			goto fail;
1492 	}
1493 	if (!m)
1494 		goto fail;
1495 	m->m_next = NULL;
1496 
1497 	if (old) {
1498 		m->m_len = len;
1499 		*mtod(m, char *) = namelen;
1500 		bcopy(name, mtod(m, char *) + 1, namelen);
1501 		return m;
1502 	} else {
1503 		m->m_len = 0;
1504 		cp = mtod(m, char *);
1505 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1506 
1507 		/* if not certain about my name, return empty buffer */
1508 		if (namelen == 0)
1509 			return m;
1510 
1511 		/*
1512 		 * guess if it looks like shortened hostname, or FQDN.
1513 		 * shortened hostname needs two trailing "\0".
1514 		 */
1515 		i = 0;
1516 		for (p = name; p < name + namelen; p++) {
1517 			if (*p && *p == '.')
1518 				i++;
1519 		}
1520 		if (i < 2)
1521 			nterm = 2;
1522 		else
1523 			nterm = 1;
1524 
1525 		p = name;
1526 		while (cp < ep && p < name + namelen) {
1527 			i = 0;
1528 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1529 				i++;
1530 			/* result does not fit into mbuf */
1531 			if (cp + i + 1 >= ep)
1532 				goto fail;
1533 			/*
1534 			 * DNS label length restriction, RFC1035 page 8.
1535 			 * "i == 0" case is included here to avoid returning
1536 			 * 0-length label on "foo..bar".
1537 			 */
1538 			if (i <= 0 || i >= 64)
1539 				goto fail;
1540 			*cp++ = i;
1541 			bcopy(p, cp, i);
1542 			cp += i;
1543 			p = q;
1544 			if (p < name + namelen && *p == '.')
1545 				p++;
1546 		}
1547 		/* termination */
1548 		if (cp + nterm >= ep)
1549 			goto fail;
1550 		while (nterm-- > 0)
1551 			*cp++ = '\0';
1552 		m->m_len = cp - mtod(m, char *);
1553 		return m;
1554 	}
1555 
1556 	panic("should not reach here");
1557 	/* NOTREACHED */
1558 
1559  fail:
1560 	if (m)
1561 		m_freem(m);
1562 	return NULL;
1563 }
1564 
1565 /*
1566  * check if two DNS-encoded string matches.  takes care of truncated
1567  * form (with \0\0 at the end).  no compression support.
1568  * XXX upper/lowercase match (see RFC2065)
1569  */
1570 static int
1571 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1572 {
1573 	const char *a0, *b0;
1574 	int l;
1575 
1576 	/* simplest case - need validation? */
1577 	if (alen == blen && bcmp(a, b, alen) == 0)
1578 		return 1;
1579 
1580 	a0 = a;
1581 	b0 = b;
1582 
1583 	/* termination is mandatory */
1584 	if (alen < 2 || blen < 2)
1585 		return 0;
1586 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1587 		return 0;
1588 	alen--;
1589 	blen--;
1590 
1591 	while (a - a0 < alen && b - b0 < blen) {
1592 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1593 			return 0;
1594 
1595 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1596 			return 0;
1597 		/* we don't support compression yet */
1598 		if (a[0] >= 64 || b[0] >= 64)
1599 			return 0;
1600 
1601 		/* truncated case */
1602 		if (a[0] == 0 && a - a0 == alen - 1)
1603 			return 1;
1604 		if (b[0] == 0 && b - b0 == blen - 1)
1605 			return 1;
1606 		if (a[0] == 0 || b[0] == 0)
1607 			return 0;
1608 
1609 		if (a[0] != b[0])
1610 			return 0;
1611 		l = a[0];
1612 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1613 			return 0;
1614 		if (bcmp(a + 1, b + 1, l) != 0)
1615 			return 0;
1616 
1617 		a += 1 + l;
1618 		b += 1 + l;
1619 	}
1620 
1621 	if (a - a0 == alen && b - b0 == blen)
1622 		return 1;
1623 	else
1624 		return 0;
1625 }
1626 
1627 /*
1628  * calculate the number of addresses to be returned in the node info reply.
1629  */
1630 static int
1631 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1632 	  char *subj)
1633 {
1634 	struct ifnet *ifp;
1635 	struct in6_ifaddr *ifa6;
1636 	struct ifaddr *ifa;
1637 	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1638 	int addrs = 0, addrsofif, iffound = 0;
1639 	int niflags = ni6->ni_flags;
1640 
1641 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1642 		switch (ni6->ni_code) {
1643 		case ICMP6_NI_SUBJ_IPV6:
1644 			if (subj == NULL) /* must be impossible... */
1645 				return(0);
1646 			subj_ip6 = (struct sockaddr_in6 *)subj;
1647 			break;
1648 		default:
1649 			/*
1650 			 * XXX: we only support IPv6 subject address for
1651 			 * this Qtype.
1652 			 */
1653 			return(0);
1654 		}
1655 	}
1656 
1657 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
1658 	{
1659 		addrsofif = 0;
1660 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list)
1661 		{
1662 			if (ifa->ifa_addr->sa_family != AF_INET6)
1663 				continue;
1664 			ifa6 = (struct in6_ifaddr *)ifa;
1665 
1666 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1667 			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1668 					       &ifa6->ia_addr.sin6_addr))
1669 				iffound = 1;
1670 
1671 			/*
1672 			 * IPv4-mapped addresses can only be returned by a
1673 			 * Node Information proxy, since they represent
1674 			 * addresses of IPv4-only nodes, which perforce do
1675 			 * not implement this protocol.
1676 			 * [icmp-name-lookups-07, Section 5.4]
1677 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1678 			 * this function at this moment.
1679 			 */
1680 
1681 			/* What do we have to do about ::1? */
1682 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1683 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1684 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1685 					continue;
1686 				break;
1687 			case IPV6_ADDR_SCOPE_SITELOCAL:
1688 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1689 					continue;
1690 				break;
1691 			case IPV6_ADDR_SCOPE_GLOBAL:
1692 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1693 					continue;
1694 				break;
1695 			default:
1696 				continue;
1697 			}
1698 
1699 			/*
1700 			 * check if anycast is okay.
1701 			 * XXX: just experimental.  not in the spec.
1702 			 */
1703 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1704 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1705 				continue; /* we need only unicast addresses */
1706 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1707 			    (icmp6_nodeinfo & 4) == 0) {
1708 				continue;
1709 			}
1710 			addrsofif++; /* count the address */
1711 		}
1712 		if (iffound) {
1713 			*ifpp = ifp;
1714 			return(addrsofif);
1715 		}
1716 
1717 		addrs += addrsofif;
1718 	}
1719 
1720 	return(addrs);
1721 }
1722 
1723 static int
1724 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1725 		struct ifnet *ifp0, int resid)
1726 {
1727 	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1728 	struct in6_ifaddr *ifa6;
1729 	struct ifaddr *ifa;
1730 	struct ifnet *ifp_dep = NULL;
1731 	int copied = 0, allow_deprecated = 0;
1732 	u_char *cp = (u_char *)(nni6 + 1);
1733 	int niflags = ni6->ni_flags;
1734 	u_int32_t ltime;
1735 
1736 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1737 		return(0);	/* needless to copy */
1738 
1739   again:
1740 
1741 	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list))
1742 	{
1743 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
1744 			if (ifa->ifa_addr->sa_family != AF_INET6)
1745 				continue;
1746 			ifa6 = (struct in6_ifaddr *)ifa;
1747 
1748 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1749 			    allow_deprecated == 0) {
1750 				/*
1751 				 * prefererred address should be put before
1752 				 * deprecated addresses.
1753 				 */
1754 
1755 				/* record the interface for later search */
1756 				if (ifp_dep == NULL)
1757 					ifp_dep = ifp;
1758 
1759 				continue;
1760 			}
1761 			else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1762 				 allow_deprecated != 0)
1763 				continue; /* we now collect deprecated addrs */
1764 
1765 			/* What do we have to do about ::1? */
1766 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1767 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1768 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1769 					continue;
1770 				break;
1771 			case IPV6_ADDR_SCOPE_SITELOCAL:
1772 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1773 					continue;
1774 				break;
1775 			case IPV6_ADDR_SCOPE_GLOBAL:
1776 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1777 					continue;
1778 				break;
1779 			default:
1780 				continue;
1781 			}
1782 
1783 			/*
1784 			 * check if anycast is okay.
1785 			 * XXX: just experimental. not in the spec.
1786 			 */
1787 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1788 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1789 				continue;
1790 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1791 			    (icmp6_nodeinfo & 4) == 0) {
1792 				continue;
1793 			}
1794 
1795 			/* now we can copy the address */
1796 			if (resid < sizeof(struct in6_addr) +
1797 			    sizeof(u_int32_t)) {
1798 				/*
1799 				 * We give up much more copy.
1800 				 * Set the truncate flag and return.
1801 				 */
1802 				nni6->ni_flags |=
1803 					NI_NODEADDR_FLAG_TRUNCATE;
1804 				return(copied);
1805 			}
1806 
1807 			/*
1808 			 * Set the TTL of the address.
1809 			 * The TTL value should be one of the following
1810 			 * according to the specification:
1811 			 *
1812 			 * 1. The remaining lifetime of a DHCP lease on the
1813 			 *    address, or
1814 			 * 2. The remaining Valid Lifetime of a prefix from
1815 			 *    which the address was derived through Stateless
1816 			 *    Autoconfiguration.
1817 			 *
1818 			 * Note that we currently do not support stateful
1819 			 * address configuration by DHCPv6, so the former
1820 			 * case can't happen.
1821 			 */
1822 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1823 				ltime = ND6_INFINITE_LIFETIME;
1824 			else {
1825 				if (ifa6->ia6_lifetime.ia6t_expire >
1826 				    time_second)
1827 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1828 				else
1829 					ltime = 0;
1830 			}
1831 
1832 			bcopy(&ltime, cp, sizeof(u_int32_t));
1833 			cp += sizeof(u_int32_t);
1834 
1835 			/* copy the address itself */
1836 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1837 			      sizeof(struct in6_addr));
1838 			/* XXX: KAME link-local hack; remove ifindex */
1839 			if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr))
1840 				((struct in6_addr *)cp)->s6_addr16[1] = 0;
1841 			cp += sizeof(struct in6_addr);
1842 
1843 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1844 			copied += (sizeof(struct in6_addr) +
1845 				   sizeof(u_int32_t));
1846 		}
1847 		if (ifp0)	/* we need search only on the specified IF */
1848 			break;
1849 	}
1850 
1851 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1852 		ifp = ifp_dep;
1853 		allow_deprecated = 1;
1854 
1855 		goto again;
1856 	}
1857 
1858 	return(copied);
1859 }
1860 
1861 /*
1862  * XXX almost dup'ed code with rip6_input.
1863  */
1864 static int
1865 icmp6_rip6_input(struct	mbuf **mp, int	off)
1866 {
1867 	struct mbuf *m = *mp;
1868 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1869 	struct in6pcb *in6p;
1870 	struct in6pcb *last = NULL;
1871 	struct sockaddr_in6 rip6src;
1872 	struct icmp6_hdr *icmp6;
1873 	struct mbuf *opts = NULL;
1874 
1875 #ifndef PULLDOWN_TEST
1876 	/* this is assumed to be safe. */
1877 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1878 #else
1879 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1880 	if (icmp6 == NULL) {
1881 		/* m is already reclaimed */
1882 		return IPPROTO_DONE;
1883 	}
1884 #endif
1885 
1886 	bzero(&rip6src, sizeof(rip6src));
1887 	rip6src.sin6_len = sizeof(struct sockaddr_in6);
1888 	rip6src.sin6_family = AF_INET6;
1889 	/* KAME hack: recover scopeid */
1890 	(void)in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif);
1891 
1892 	LIST_FOREACH(in6p, &ripcbinfo.pcblisthead, inp_list)
1893 	{
1894 		if (in6p->inp_flags & INP_PLACEMARKER)
1895 			continue;
1896 		if ((in6p->inp_vflag & INP_IPV6) == 0)
1897 			continue;
1898 #ifdef HAVE_NRL_INPCB
1899 		if (!(in6p->in6p_flags & INP_IPV6))
1900 			continue;
1901 #endif
1902 		if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1903 			continue;
1904 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1905 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1906 			continue;
1907 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1908 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1909 			continue;
1910 		if (in6p->in6p_icmp6filt
1911 		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1912 				 in6p->in6p_icmp6filt))
1913 			continue;
1914 		if (last) {
1915 			struct	mbuf *n;
1916 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1917 				if (last->in6p_flags & IN6P_CONTROLOPTS)
1918 					ip6_savecontrol(last, &opts, ip6, n);
1919 				/* strip intermediate headers */
1920 				m_adj(n, off);
1921 				if (sbappendaddr(&last->in6p_socket->so_rcv,
1922 						 (struct sockaddr *)&rip6src,
1923 						 n, opts) == 0) {
1924 					/* should notify about lost packet */
1925 					m_freem(n);
1926 					if (opts) {
1927 						m_freem(opts);
1928 					}
1929 				} else
1930 					sorwakeup(last->in6p_socket);
1931 				opts = NULL;
1932 			}
1933 		}
1934 		last = in6p;
1935 	}
1936 	if (last) {
1937 		if (last->in6p_flags & IN6P_CONTROLOPTS)
1938 			ip6_savecontrol(last, &opts, ip6, m);
1939 		/* strip intermediate headers */
1940 		m_adj(m, off);
1941 		if (sbappendaddr(&last->in6p_socket->so_rcv,
1942 				 (struct sockaddr *)&rip6src, m, opts) == 0) {
1943 			m_freem(m);
1944 			if (opts)
1945 				m_freem(opts);
1946 		} else
1947 			sorwakeup(last->in6p_socket);
1948 	} else {
1949 		m_freem(m);
1950 		ip6stat.ip6s_delivered--;
1951 	}
1952 	return IPPROTO_DONE;
1953 }
1954 
1955 /*
1956  * Reflect the ip6 packet back to the source.
1957  * OFF points to the icmp6 header, counted from the top of the mbuf.
1958  */
1959 void
1960 icmp6_reflect(struct mbuf *m, size_t off)
1961 {
1962 	struct ip6_hdr *ip6;
1963 	struct icmp6_hdr *icmp6;
1964 	struct in6_ifaddr *ia;
1965 	struct in6_addr t, *src = 0;
1966 	int plen;
1967 	int type, code;
1968 	struct ifnet *outif = NULL;
1969 	struct sockaddr_in6 sa6_src, sa6_dst;
1970 #ifdef COMPAT_RFC1885
1971 	int mtu = IPV6_MMTU;
1972 	struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
1973 #endif
1974 
1975 	/* too short to reflect */
1976 	if (off < sizeof(struct ip6_hdr)) {
1977 		nd6log((LOG_DEBUG,
1978 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
1979 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
1980 		    __FILE__, __LINE__));
1981 		goto bad;
1982 	}
1983 
1984 	/*
1985 	 * If there are extra headers between IPv6 and ICMPv6, strip
1986 	 * off that header first.
1987 	 */
1988 #ifdef DIAGNOSTIC
1989 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
1990 		panic("assumption failed in icmp6_reflect");
1991 #endif
1992 	if (off > sizeof(struct ip6_hdr)) {
1993 		size_t l;
1994 		struct ip6_hdr nip6;
1995 
1996 		l = off - sizeof(struct ip6_hdr);
1997 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
1998 		m_adj(m, l);
1999 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2000 		if (m->m_len < l) {
2001 			if ((m = m_pullup(m, l)) == NULL)
2002 				return;
2003 		}
2004 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2005 	} else /* off == sizeof(struct ip6_hdr) */ {
2006 		size_t l;
2007 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2008 		if (m->m_len < l) {
2009 			if ((m = m_pullup(m, l)) == NULL)
2010 				return;
2011 		}
2012 	}
2013 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2014 	ip6 = mtod(m, struct ip6_hdr *);
2015 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2016 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2017 	type = icmp6->icmp6_type; /* keep type for statistics */
2018 	code = icmp6->icmp6_code; /* ditto. */
2019 
2020 	t = ip6->ip6_dst;
2021 	/*
2022 	 * ip6_input() drops a packet if its src is multicast.
2023 	 * So, the src is never multicast.
2024 	 */
2025 	ip6->ip6_dst = ip6->ip6_src;
2026 
2027 	/*
2028 	 * XXX: make sure to embed scope zone information, using
2029 	 * already embedded IDs or the received interface (if any).
2030 	 * Note that rcvif may be NULL.
2031 	 * TODO: scoped routing case (XXX).
2032 	 */
2033 	bzero(&sa6_src, sizeof(sa6_src));
2034 	sa6_src.sin6_family = AF_INET6;
2035 	sa6_src.sin6_len = sizeof(sa6_src);
2036 	sa6_src.sin6_addr = ip6->ip6_dst;
2037 	in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
2038 	in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL);
2039 	bzero(&sa6_dst, sizeof(sa6_dst));
2040 	sa6_dst.sin6_family = AF_INET6;
2041 	sa6_dst.sin6_len = sizeof(sa6_dst);
2042 	sa6_dst.sin6_addr = t;
2043 	in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
2044 	in6_embedscope(&t, &sa6_dst, NULL, NULL);
2045 
2046 #ifdef COMPAT_RFC1885
2047 	/*
2048 	 * xxx guess MTU
2049 	 * RFC 1885 requires that echo reply should be truncated if it
2050 	 * does not fit in with (return) path MTU, but the description was
2051 	 * removed in the new spec.
2052 	 */
2053 	if (icmp6_reflect_rt.ro_rt == 0 ||
2054 	    ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
2055 		if (icmp6_reflect_rt.ro_rt) {
2056 			RTFREE(icmp6_reflect_rt.ro_rt);
2057 			icmp6_reflect_rt.ro_rt = 0;
2058 		}
2059 		bzero(sin6, sizeof(*sin6));
2060 		sin6->sin6_family = PF_INET6;
2061 		sin6->sin6_len = sizeof(struct sockaddr_in6);
2062 		sin6->sin6_addr = ip6->ip6_dst;
2063 
2064 		rtalloc_ign((struct route *)&icmp6_reflect_rt.ro_rt,
2065 			    RTF_PRCLONING);
2066 	}
2067 
2068 	if (icmp6_reflect_rt.ro_rt == 0)
2069 		goto bad;
2070 
2071 	if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
2072 	    && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
2073 		mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
2074 
2075 	if (mtu < m->m_pkthdr.len) {
2076 		plen -= (m->m_pkthdr.len - mtu);
2077 		m_adj(m, mtu - m->m_pkthdr.len);
2078 	}
2079 #endif
2080 	/*
2081 	 * If the incoming packet was addressed directly to us(i.e. unicast),
2082 	 * use dst as the src for the reply.
2083 	 * The IN6_IFF_NOTREADY case would be VERY rare, but is possible
2084 	 * (for example) when we encounter an error while forwarding procedure
2085 	 * destined to a duplicated address of ours.
2086 	 */
2087 	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2088 		if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2089 		    (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2090 			src = &t;
2091 			break;
2092 		}
2093 	if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2094 		/*
2095 		 * This is the case if the dst is our link-local address
2096 		 * and the sender is also ourselves.
2097 		 */
2098 		src = &t;
2099 	}
2100 
2101 	if (src == 0) {
2102 		int e;
2103 		struct route_in6 ro;
2104 
2105 		/*
2106 		 * This case matches to multicasts, our anycast, or unicasts
2107 		 * that we do not own.  Select a source address based on the
2108 		 * source address of the erroneous packet.
2109 		 */
2110 		bzero(&ro, sizeof(ro));
2111 		src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e);
2112 		if (ro.ro_rt)
2113 			RTFREE(ro.ro_rt); /* XXX: we could use this */
2114 		if (src == NULL) {
2115 			nd6log((LOG_DEBUG,
2116 			    "icmp6_reflect: source can't be determined: "
2117 			    "dst=%s, error=%d\n",
2118 			    ip6_sprintf(&sa6_src.sin6_addr), e));
2119 			goto bad;
2120 		}
2121 	}
2122 
2123 	ip6->ip6_src = *src;
2124 
2125 	ip6->ip6_flow = 0;
2126 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2127 	ip6->ip6_vfc |= IPV6_VERSION;
2128 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2129 	if (m->m_pkthdr.rcvif) {
2130 		/* XXX: This may not be the outgoing interface */
2131 		ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2132 	} else
2133 		ip6->ip6_hlim = ip6_defhlim;
2134 
2135 	icmp6->icmp6_cksum = 0;
2136 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2137 					sizeof(struct ip6_hdr), plen);
2138 
2139 	/*
2140 	 * XXX option handling
2141 	 */
2142 
2143 	m->m_flags &= ~(M_BCAST|M_MCAST);
2144 
2145 #ifdef COMPAT_RFC1885
2146 	ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif, NULL);
2147 #else
2148 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2149 #endif
2150 	if (outif)
2151 		icmp6_ifoutstat_inc(outif, type, code);
2152 
2153 	return;
2154 
2155  bad:
2156 	m_freem(m);
2157 	return;
2158 }
2159 
2160 void
2161 icmp6_fasttimo(void)
2162 {
2163 
2164 	mld6_fasttimeo();
2165 }
2166 
2167 static const char *
2168 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2169 		    struct in6_addr *tgt6)
2170 {
2171 	static char buf[1024];
2172 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2173 		ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2174 	return buf;
2175 }
2176 
2177 void
2178 icmp6_redirect_input(struct mbuf *m, int off)
2179 {
2180 	struct ifnet *ifp = m->m_pkthdr.rcvif;
2181 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2182 	struct nd_redirect *nd_rd;
2183 	int icmp6len = ntohs(ip6->ip6_plen);
2184 	char *lladdr = NULL;
2185 	int lladdrlen = 0;
2186 	u_char *redirhdr = NULL;
2187 	int redirhdrlen = 0;
2188 	struct rtentry *rt = NULL;
2189 	int is_router;
2190 	int is_onlink;
2191 	struct in6_addr src6 = ip6->ip6_src;
2192 	struct in6_addr redtgt6;
2193 	struct in6_addr reddst6;
2194 	union nd_opts ndopts;
2195 
2196 	if (!m || !ifp)
2197 		return;
2198 
2199 	/* XXX if we are router, we don't update route by icmp6 redirect */
2200 	if (ip6_forwarding)
2201 		goto freeit;
2202 	if (!icmp6_rediraccept)
2203 		goto freeit;
2204 
2205 #ifndef PULLDOWN_TEST
2206 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2207 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2208 #else
2209 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2210 	if (nd_rd == NULL) {
2211 		icmp6stat.icp6s_tooshort++;
2212 		return;
2213 	}
2214 #endif
2215 	redtgt6 = nd_rd->nd_rd_target;
2216 	reddst6 = nd_rd->nd_rd_dst;
2217 
2218 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2219 		redtgt6.s6_addr16[1] = htons(ifp->if_index);
2220 	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2221 		reddst6.s6_addr16[1] = htons(ifp->if_index);
2222 
2223 	/* validation */
2224 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2225 		nd6log((LOG_ERR,
2226 			"ICMP6 redirect sent from %s rejected; "
2227 			"must be from linklocal\n", ip6_sprintf(&src6)));
2228 		goto bad;
2229 	}
2230 	if (ip6->ip6_hlim != 255) {
2231 		nd6log((LOG_ERR,
2232 			"ICMP6 redirect sent from %s rejected; "
2233 			"hlim=%d (must be 255)\n",
2234 			ip6_sprintf(&src6), ip6->ip6_hlim));
2235 		goto bad;
2236 	}
2237     {
2238 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2239 	struct sockaddr_in6 sin6;
2240 	struct in6_addr *gw6;
2241 
2242 	bzero(&sin6, sizeof(sin6));
2243 	sin6.sin6_family = AF_INET6;
2244 	sin6.sin6_len = sizeof(struct sockaddr_in6);
2245 	bcopy(&reddst6, &sin6.sin6_addr, sizeof reddst6);
2246 	rt = rtpurelookup((struct sockaddr *)&sin6);
2247 	if (rt != NULL) {
2248 		if (rt->rt_gateway == NULL ||
2249 		    rt->rt_gateway->sa_family != AF_INET6) {
2250 			nd6log((LOG_ERR,
2251 			    "ICMP6 redirect rejected; no route "
2252 			    "with inet6 gateway found for redirect dst: %s\n",
2253 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2254 			--rt->rt_refcnt;
2255 			goto bad;
2256 		}
2257 
2258 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2259 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2260 			nd6log((LOG_ERR,
2261 				"ICMP6 redirect rejected; "
2262 				"not equal to gw-for-src=%s (must be same): "
2263 				"%s\n",
2264 				ip6_sprintf(gw6),
2265 				icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2266 			--rt->rt_refcnt;
2267 			goto bad;
2268 		}
2269 	} else {
2270 		nd6log((LOG_ERR,
2271 			"ICMP6 redirect rejected; "
2272 			"no route found for redirect dst: %s\n",
2273 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2274 		goto bad;
2275 	}
2276 	--rt->rt_refcnt;
2277 	rt = NULL;
2278     }
2279 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2280 		nd6log((LOG_ERR,
2281 			"ICMP6 redirect rejected; "
2282 			"redirect dst must be unicast: %s\n",
2283 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2284 		goto bad;
2285 	}
2286 
2287 	is_router = is_onlink = 0;
2288 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2289 		is_router = 1;	/* router case */
2290 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2291 		is_onlink = 1;	/* on-link destination case */
2292 	if (!is_router && !is_onlink) {
2293 		nd6log((LOG_ERR,
2294 			"ICMP6 redirect rejected; "
2295 			"neither router case nor onlink case: %s\n",
2296 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2297 		goto bad;
2298 	}
2299 	/* validation passed */
2300 
2301 	icmp6len -= sizeof(*nd_rd);
2302 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2303 	if (nd6_options(&ndopts) < 0) {
2304 		nd6log((LOG_INFO, "icmp6_redirect_input: "
2305 			"invalid ND option, rejected: %s\n",
2306 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2307 		/* nd6_options have incremented stats */
2308 		goto freeit;
2309 	}
2310 
2311 	if (ndopts.nd_opts_tgt_lladdr) {
2312 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2313 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2314 	}
2315 
2316 	if (ndopts.nd_opts_rh) {
2317 		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2318 		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2319 	}
2320 
2321 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2322 		nd6log((LOG_INFO,
2323 			"icmp6_redirect_input: lladdrlen mismatch for %s "
2324 			"(if %d, icmp6 packet %d): %s\n",
2325 			ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2326 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2327 		goto bad;
2328 	}
2329 
2330 	/* RFC 2461 8.3 */
2331 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2332 			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2333 
2334 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2335 		/* perform rtredirect */
2336 		struct sockaddr_in6 sdst;
2337 		struct sockaddr_in6 sgw;
2338 		struct sockaddr_in6 ssrc;
2339 
2340 		bzero(&sdst, sizeof(sdst));
2341 		bzero(&sgw, sizeof(sgw));
2342 		bzero(&ssrc, sizeof(ssrc));
2343 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2344 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2345 			sizeof(struct sockaddr_in6);
2346 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2347 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2348 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2349 		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2350 			   (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2351 			   (struct sockaddr *)&ssrc);
2352 	}
2353 	/* finally update cached route in each socket via pfctlinput */
2354     {
2355 	struct sockaddr_in6 sdst;
2356 
2357 	bzero(&sdst, sizeof(sdst));
2358 	sdst.sin6_family = AF_INET6;
2359 	sdst.sin6_len = sizeof(struct sockaddr_in6);
2360 	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2361 	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2362 #ifdef IPSEC
2363 	key_sa_routechange((struct sockaddr *)&sdst);
2364 #endif
2365     }
2366 
2367  freeit:
2368 	m_freem(m);
2369 	return;
2370 
2371  bad:
2372 	icmp6stat.icp6s_badredirect++;
2373 	m_freem(m);
2374 }
2375 
2376 void
2377 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2378 {
2379 	struct ifnet *ifp;	/* my outgoing interface */
2380 	struct in6_addr *ifp_ll6;
2381 	struct in6_addr *router_ll6;
2382 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2383 	struct mbuf *m = NULL;	/* newly allocated one */
2384 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2385 	struct nd_redirect *nd_rd;
2386 	size_t maxlen;
2387 	u_char *p;
2388 	struct ifnet *outif = NULL;
2389 	struct sockaddr_in6 src_sa;
2390 
2391 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2392 
2393 	/* if we are not router, we don't send icmp6 redirect */
2394 	if (!ip6_forwarding || ip6_accept_rtadv)
2395 		goto fail;
2396 
2397 	/* sanity check */
2398 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2399 		goto fail;
2400 
2401 	/*
2402 	 * Address check:
2403 	 *  the source address must identify a neighbor, and
2404 	 *  the destination address must not be a multicast address
2405 	 *  [RFC 2461, sec 8.2]
2406 	 */
2407 	sip6 = mtod(m0, struct ip6_hdr *);
2408 	bzero(&src_sa, sizeof(src_sa));
2409 	src_sa.sin6_family = AF_INET6;
2410 	src_sa.sin6_len = sizeof(src_sa);
2411 	src_sa.sin6_addr = sip6->ip6_src;
2412 	/* we don't currently use sin6_scope_id, but eventually use it */
2413 	src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src);
2414 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2415 		goto fail;
2416 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2417 		goto fail;	/* what should we do here? */
2418 
2419 	/* rate limit */
2420 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2421 		goto fail;
2422 
2423 	/*
2424 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2425 	 * we almost always ask for an mbuf cluster for simplicity.
2426 	 * (MHLEN < IPV6_MMTU is almost always true)
2427 	 */
2428 #if IPV6_MMTU >= MCLBYTES
2429 # error assumption failed about IPV6_MMTU and MCLBYTES
2430 #endif
2431 	MGETHDR(m, MB_DONTWAIT, MT_HEADER);
2432 	if (m && IPV6_MMTU >= MHLEN)
2433 		MCLGET(m, MB_DONTWAIT);
2434 	if (!m)
2435 		goto fail;
2436 	m->m_pkthdr.rcvif = NULL;
2437 	m->m_len = 0;
2438 	maxlen = M_TRAILINGSPACE(m);
2439 	maxlen = min(IPV6_MMTU, maxlen);
2440 	/* just for safety */
2441 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2442 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2443 		goto fail;
2444 	}
2445 
2446 	{
2447 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2448 		struct in6_ifaddr *ia;
2449 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2450 						 IN6_IFF_NOTREADY|
2451 						 IN6_IFF_ANYCAST)) == NULL)
2452 			goto fail;
2453 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2454 	}
2455 
2456 	/* get ip6 linklocal address for the router. */
2457 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2458 		struct sockaddr_in6 *sin6;
2459 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2460 		router_ll6 = &sin6->sin6_addr;
2461 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2462 			router_ll6 = (struct in6_addr *)NULL;
2463 	} else
2464 		router_ll6 = (struct in6_addr *)NULL;
2465 
2466 	/* ip6 */
2467 	ip6 = mtod(m, struct ip6_hdr *);
2468 	ip6->ip6_flow = 0;
2469 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2470 	ip6->ip6_vfc |= IPV6_VERSION;
2471 	/* ip6->ip6_plen will be set later */
2472 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2473 	ip6->ip6_hlim = 255;
2474 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2475 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2476 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2477 
2478 	/* ND Redirect */
2479 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2480 	nd_rd->nd_rd_type = ND_REDIRECT;
2481 	nd_rd->nd_rd_code = 0;
2482 	nd_rd->nd_rd_reserved = 0;
2483 	if (rt->rt_flags & RTF_GATEWAY) {
2484 		/*
2485 		 * nd_rd->nd_rd_target must be a link-local address in
2486 		 * better router cases.
2487 		 */
2488 		if (!router_ll6)
2489 			goto fail;
2490 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2491 		      sizeof(nd_rd->nd_rd_target));
2492 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2493 		      sizeof(nd_rd->nd_rd_dst));
2494 	} else {
2495 		/* make sure redtgt == reddst */
2496 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2497 		      sizeof(nd_rd->nd_rd_target));
2498 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2499 		      sizeof(nd_rd->nd_rd_dst));
2500 	}
2501 
2502 	p = (u_char *)(nd_rd + 1);
2503 
2504 	if (!router_ll6)
2505 		goto nolladdropt;
2506 
2507     {
2508 	/* target lladdr option */
2509 	struct rtentry *rt_router = NULL;
2510 	int len;
2511 	struct sockaddr_dl *sdl;
2512 	struct nd_opt_hdr *nd_opt;
2513 	char *lladdr;
2514 
2515 	rt_router = nd6_lookup(router_ll6, 0, ifp);
2516 	if (!rt_router)
2517 		goto nolladdropt;
2518 	len = sizeof(*nd_opt) + ifp->if_addrlen;
2519 	len = (len + 7) & ~7;	/* round by 8 */
2520 	/* safety check */
2521 	if (len + (p - (u_char *)ip6) > maxlen)
2522 		goto nolladdropt;
2523 	if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2524 	    (rt_router->rt_flags & RTF_LLINFO) &&
2525 	    (rt_router->rt_gateway->sa_family == AF_LINK) &&
2526 	    (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2527 	    sdl->sdl_alen) {
2528 		nd_opt = (struct nd_opt_hdr *)p;
2529 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2530 		nd_opt->nd_opt_len = len >> 3;
2531 		lladdr = (char *)(nd_opt + 1);
2532 		bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2533 		p += len;
2534 	}
2535     }
2536 nolladdropt:;
2537 
2538 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2539 
2540 	/* just to be safe */
2541 #ifdef M_DECRYPTED	/*not openbsd*/
2542 	if (m0->m_flags & M_DECRYPTED)
2543 		goto noredhdropt;
2544 #endif
2545 	if (p - (u_char *)ip6 > maxlen)
2546 		goto noredhdropt;
2547 
2548     {
2549 	/* redirected header option */
2550 	int len;
2551 	struct nd_opt_rd_hdr *nd_opt_rh;
2552 
2553 	/*
2554 	 * compute the maximum size for icmp6 redirect header option.
2555 	 * XXX room for auth header?
2556 	 */
2557 	len = maxlen - (p - (u_char *)ip6);
2558 	len &= ~7;
2559 
2560 	/* This is just for simplicity. */
2561 	if (m0->m_pkthdr.len != m0->m_len) {
2562 		if (m0->m_next) {
2563 			m_freem(m0->m_next);
2564 			m0->m_next = NULL;
2565 		}
2566 		m0->m_pkthdr.len = m0->m_len;
2567 	}
2568 
2569 	/*
2570 	 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2571 	 * about padding/truncate rule for the original IP packet.
2572 	 * From the discussion on IPv6imp in Feb 1999, the consensus was:
2573 	 * - "attach as much as possible" is the goal
2574 	 * - pad if not aligned (original size can be guessed by original
2575 	 *   ip6 header)
2576 	 * Following code adds the padding if it is simple enough,
2577 	 * and truncates if not.
2578 	 */
2579 	if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2580 		panic("assumption failed in %s:%d", __FILE__, __LINE__);
2581 
2582 	if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2583 		/* not enough room, truncate */
2584 		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2585 	} else {
2586 		/* enough room, pad or truncate */
2587 		size_t extra;
2588 
2589 		extra = m0->m_pkthdr.len % 8;
2590 		if (extra) {
2591 			/* pad if easy enough, truncate if not */
2592 			if (8 - extra <= M_TRAILINGSPACE(m0)) {
2593 				/* pad */
2594 				m0->m_len += (8 - extra);
2595 				m0->m_pkthdr.len += (8 - extra);
2596 			} else {
2597 				/* truncate */
2598 				m0->m_pkthdr.len -= extra;
2599 				m0->m_len -= extra;
2600 			}
2601 		}
2602 		len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2603 		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2604 	}
2605 
2606 	nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2607 	bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2608 	nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2609 	nd_opt_rh->nd_opt_rh_len = len >> 3;
2610 	p += sizeof(*nd_opt_rh);
2611 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2612 
2613 	/* connect m0 to m */
2614 	m->m_next = m0;
2615 	m->m_pkthdr.len = m->m_len + m0->m_len;
2616     }
2617 noredhdropt:;
2618 
2619 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
2620 		sip6->ip6_src.s6_addr16[1] = 0;
2621 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
2622 		sip6->ip6_dst.s6_addr16[1] = 0;
2623 #if 0
2624 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
2625 		ip6->ip6_src.s6_addr16[1] = 0;
2626 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
2627 		ip6->ip6_dst.s6_addr16[1] = 0;
2628 #endif
2629 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
2630 		nd_rd->nd_rd_target.s6_addr16[1] = 0;
2631 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
2632 		nd_rd->nd_rd_dst.s6_addr16[1] = 0;
2633 
2634 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2635 
2636 	nd_rd->nd_rd_cksum = 0;
2637 	nd_rd->nd_rd_cksum
2638 		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2639 
2640 	/* send the packet to outside... */
2641 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2642 	if (outif) {
2643 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2644 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2645 	}
2646 	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2647 
2648 	return;
2649 
2650 fail:
2651 	if (m)
2652 		m_freem(m);
2653 	if (m0)
2654 		m_freem(m0);
2655 }
2656 
2657 #ifdef HAVE_NRL_INPCB
2658 #define in6pcb		inpcb
2659 #define in6p_icmp6filt	inp_icmp6filt
2660 #endif
2661 /*
2662  * ICMPv6 socket option processing.
2663  */
2664 int
2665 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2666 {
2667 	int error = 0;
2668 	int optlen;
2669 	struct inpcb *inp = so->so_pcb;
2670 	int level, op, optname;
2671 
2672 	if (sopt) {
2673 		level = sopt->sopt_level;
2674 		op = sopt->sopt_dir;
2675 		optname = sopt->sopt_name;
2676 		optlen = sopt->sopt_valsize;
2677 	} else
2678 		level = op = optname = optlen = 0;
2679 
2680 	if (level != IPPROTO_ICMPV6) {
2681 		return EINVAL;
2682 	}
2683 
2684 	switch (op) {
2685 	case PRCO_SETOPT:
2686 		switch (optname) {
2687 		case ICMP6_FILTER:
2688 		    {
2689 			struct icmp6_filter *p;
2690 
2691 			if (optlen != sizeof(*p)) {
2692 				error = EMSGSIZE;
2693 				break;
2694 			}
2695 			if (inp->in6p_icmp6filt == NULL) {
2696 				error = EINVAL;
2697 				break;
2698 			}
2699 			error = sooptcopyin(sopt, inp->in6p_icmp6filt, optlen,
2700 				optlen);
2701 			break;
2702 		    }
2703 
2704 		default:
2705 			error = ENOPROTOOPT;
2706 			break;
2707 		}
2708 		break;
2709 
2710 	case PRCO_GETOPT:
2711 		switch (optname) {
2712 		case ICMP6_FILTER:
2713 		    {
2714 			if (inp->in6p_icmp6filt == NULL) {
2715 				error = EINVAL;
2716 				break;
2717 			}
2718 			error = sooptcopyout(sopt, inp->in6p_icmp6filt,
2719 				sizeof(struct icmp6_filter));
2720 			break;
2721 		    }
2722 
2723 		default:
2724 			error = ENOPROTOOPT;
2725 			break;
2726 		}
2727 		break;
2728 	}
2729 
2730 	return(error);
2731 }
2732 #ifdef HAVE_NRL_INPCB
2733 #undef in6pcb
2734 #undef in6p_icmp6filt
2735 #endif
2736 
2737 #ifndef HAVE_PPSRATECHECK
2738 #ifndef timersub
2739 #define	timersub(tvp, uvp, vvp)						\
2740 	do {								\
2741 		(vvp)->tv_sec = (tvp)->tv_sec - (uvp)->tv_sec;		\
2742 		(vvp)->tv_usec = (tvp)->tv_usec - (uvp)->tv_usec;	\
2743 		if ((vvp)->tv_usec < 0) {				\
2744 			(vvp)->tv_sec--;				\
2745 			(vvp)->tv_usec += 1000000;			\
2746 		}							\
2747 	} while (0)
2748 #endif
2749 
2750 /*
2751  * ppsratecheck(): packets (or events) per second limitation.
2752  */
2753 static int
2754 ppsratecheck(struct timeval *lasttime, int *curpps,
2755 	     int maxpps)	/* maximum pps allowed */
2756 {
2757 	struct timeval tv, delta;
2758 	int rv;
2759 
2760 	microtime(&tv);
2761 
2762 	timersub(&tv, lasttime, &delta);
2763 
2764 	/*
2765 	 * Check for 0,0 so that the message will be seen at least once.
2766 	 * If more than one second has passed since the last update of
2767 	 * lasttime, reset the counter.
2768 	 *
2769 	 * We do increment *curpps even in *curpps < maxpps case, as some may
2770 	 * try to use *curpps for stat purposes as well.
2771 	 */
2772 	if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) ||
2773 	    delta.tv_sec >= 1) {
2774 		*lasttime = tv;
2775 		*curpps = 0;
2776 		rv = 1;
2777 	} else if (maxpps < 0)
2778 		rv = 1;
2779 	else if (*curpps < maxpps)
2780 		rv = 1;
2781 	else
2782 		rv = 0;
2783 
2784 #if 1 /* DIAGNOSTIC? */
2785 	/* be careful about wrap-around */
2786 	if (*curpps + 1 > *curpps)
2787 		*curpps = *curpps + 1;
2788 #else
2789 	/*
2790 	 * assume that there's not too many calls to this function.
2791 	 * not sure if the assumption holds, as it depends on *caller's*
2792 	 * behavior, not the behavior of this function.
2793 	 * IMHO it is wrong to make assumption on the caller's behavior,
2794 	 * so the above #if is #if 1, not #ifdef DIAGNOSTIC.
2795 	 */
2796 	*curpps = *curpps + 1;
2797 #endif
2798 
2799 	return (rv);
2800 }
2801 #endif
2802 
2803 /*
2804  * Perform rate limit check.
2805  * Returns 0 if it is okay to send the icmp6 packet.
2806  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2807  * limitation.
2808  *
2809  * XXX per-destination/type check necessary?
2810  */
2811 static int
2812 icmp6_ratelimit(const struct in6_addr *dst,	/* not used at this moment */
2813 		const int type,			/* not used at this moment */
2814 		const int code)			/* not used at this moment */
2815 {
2816 	int ret;
2817 
2818 	ret = 0;	/* okay to send */
2819 
2820 	/* PPS limit */
2821 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2822 	    icmp6errppslim)) {
2823 		/* The packet is subject to rate limit */
2824 		ret++;
2825 	}
2826 
2827 	return ret;
2828 }
2829