xref: /dragonfly/sys/netinet6/icmp6.c (revision 927da715)
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 1.30 2008/09/04 09:08:22 hasso 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) ||
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 			n = m_getb(maxlen, MB_DONTWAIT, n0->m_type, M_PKTHDR);
562 			if (n == NULL) {
563 				/* Give up remote */
564 				m_freem(n0);
565 				break;
566 			}
567 			n0len = n0->m_pkthdr.len;	/* save for use below */
568 			M_MOVE_PKTHDR(n, n0);
569 			/*
570 			 * Copy IPv6 and ICMPv6 only.
571 			 */
572 			nip6 = mtod(n, struct ip6_hdr *);
573 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
574 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
575 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
576 			noff = sizeof(struct ip6_hdr);
577 			/* new mbuf contains only ipv6+icmpv6 headers */
578 			n->m_len = noff + sizeof(struct icmp6_hdr);
579 			/*
580 			 * Adjust mbuf. ip6_plen will be adjusted in
581 			 * ip6_output().
582 			 */
583 			m_adj(n0, off + sizeof(struct icmp6_hdr));
584 			/* recalculate complete packet size */
585 			n->m_pkthdr.len = n0len + (noff - off);
586 			n->m_next = n0;
587 		} else {
588 			nip6 = mtod(n, struct ip6_hdr *);
589 			nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
590 			noff = off;
591 		}
592 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
593 		nicmp6->icmp6_code = 0;
594 		if (n) {
595 			icmp6stat.icp6s_reflect++;
596 			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
597 			icmp6_reflect(n, noff);
598 		}
599 		break;
600 
601 	case ICMP6_ECHO_REPLY:
602 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
603 		if (code != 0)
604 			goto badcode;
605 		break;
606 
607 	case MLD_LISTENER_QUERY:
608 	case MLD_LISTENER_REPORT:
609 		if (icmp6len < sizeof(struct mld_hdr))
610 			goto badlen;
611 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
612 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
613 		else
614 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
615 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
616 			/* give up local */
617 			mld6_input(m, off);
618 			m = NULL;
619 			goto freeit;
620 		}
621 		mld6_input(n, off);
622 		/* m stays. */
623 		break;
624 
625 	case MLD_LISTENER_DONE:
626 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
627 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
628 			goto badlen;
629 		break;		/* nothing to be done in kernel */
630 
631 	case MLD_MTRACE_RESP:
632 	case MLD_MTRACE:
633 		/* XXX: these two are experimental.  not officially defind. */
634 		/* XXX: per-interface statistics? */
635 		break;		/* just pass it to applications */
636 
637 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
638 	    {
639 		enum { WRU, FQDN } mode;
640 
641 		if (!icmp6_nodeinfo)
642 			break;
643 
644 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
645 			mode = WRU;
646 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
647 			mode = FQDN;
648 		else
649 			goto badlen;
650 
651 #define hostnamelen	strlen(hostname)
652 		if (mode == FQDN) {
653 #ifndef PULLDOWN_TEST
654 			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
655 					 IPPROTO_DONE);
656 #endif
657 			n = m_copy(m, 0, M_COPYALL);
658 			if (n)
659 				n = ni6_input(n, off);
660 			/* XXX meaningless if n == NULL */
661 			noff = sizeof(struct ip6_hdr);
662 		} else {
663 			u_char *p;
664 			int maxlen, maxhlen;
665 
666 			if ((icmp6_nodeinfo & 5) != 5)
667 				break;
668 
669 			if (code != 0)
670 				goto badcode;
671 			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
672 			if (maxlen >= MCLBYTES) {
673 				/* Give up remote */
674 				break;
675 			}
676 			n = m_getb(maxlen, MB_DONTWAIT, m->m_type, M_PKTHDR);
677 			if (n == NULL) {
678 				/* Give up remote */
679 				break;
680 			}
681 			if (!m_dup_pkthdr(n, m, MB_DONTWAIT)) {
682 				/*
683 				 * Previous code did a blind M_COPY_PKTHDR
684 				 * and said "just for rcvif".  If true, then
685 				 * we could tolerate the dup failing (due to
686 				 * the deep copy of the tag chain).  For now
687 				 * be conservative and just fail.
688 				 */
689 				m_free(n);
690 				break;
691 			}
692 			n->m_len = 0;
693 			maxhlen = M_TRAILINGSPACE(n) - maxlen;
694 			if (maxhlen > hostnamelen)
695 				maxhlen = hostnamelen;
696 			/*
697 			 * Copy IPv6 and ICMPv6 only.
698 			 */
699 			nip6 = mtod(n, struct ip6_hdr *);
700 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
701 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
702 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
703 			p = (u_char *)(nicmp6 + 1);
704 			bzero(p, 4);
705 			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
706 			noff = sizeof(struct ip6_hdr);
707 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
708 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
709 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
710 			nicmp6->icmp6_code = 0;
711 		}
712 #undef hostnamelen
713 		if (n) {
714 			icmp6stat.icp6s_reflect++;
715 			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
716 			icmp6_reflect(n, noff);
717 		}
718 		break;
719 	    }
720 
721 	case ICMP6_WRUREPLY:
722 		if (code != 0)
723 			goto badcode;
724 		break;
725 
726 	case ND_ROUTER_SOLICIT:
727 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
728 		if (code != 0)
729 			goto badcode;
730 		if (icmp6len < sizeof(struct nd_router_solicit))
731 			goto badlen;
732 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
733 			/* give up local */
734 			nd6_rs_input(m, off, icmp6len);
735 			m = NULL;
736 			goto freeit;
737 		}
738 		nd6_rs_input(n, off, icmp6len);
739 		/* m stays. */
740 		break;
741 
742 	case ND_ROUTER_ADVERT:
743 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
744 		if (code != 0)
745 			goto badcode;
746 		if (icmp6len < sizeof(struct nd_router_advert))
747 			goto badlen;
748 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
749 			/* give up local */
750 			nd6_ra_input(m, off, icmp6len);
751 			m = NULL;
752 			goto freeit;
753 		}
754 		nd6_ra_input(n, off, icmp6len);
755 		/* m stays. */
756 		break;
757 
758 	case ND_NEIGHBOR_SOLICIT:
759 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
760 		if (code != 0)
761 			goto badcode;
762 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
763 			goto badlen;
764 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
765 			/* give up local */
766 			nd6_ns_input(m, off, icmp6len);
767 			m = NULL;
768 			goto freeit;
769 		}
770 		nd6_ns_input(n, off, icmp6len);
771 		/* m stays. */
772 		break;
773 
774 	case ND_NEIGHBOR_ADVERT:
775 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
776 		if (code != 0)
777 			goto badcode;
778 		if (icmp6len < sizeof(struct nd_neighbor_advert))
779 			goto badlen;
780 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
781 			/* give up local */
782 			nd6_na_input(m, off, icmp6len);
783 			m = NULL;
784 			goto freeit;
785 		}
786 		nd6_na_input(n, off, icmp6len);
787 		/* m stays. */
788 		break;
789 
790 	case ND_REDIRECT:
791 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
792 		if (code != 0)
793 			goto badcode;
794 		if (icmp6len < sizeof(struct nd_redirect))
795 			goto badlen;
796 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
797 			/* give up local */
798 			icmp6_redirect_input(m, off);
799 			m = NULL;
800 			goto freeit;
801 		}
802 		icmp6_redirect_input(n, off);
803 		/* m stays. */
804 		break;
805 
806 	case ICMP6_ROUTER_RENUMBERING:
807 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
808 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
809 			goto badcode;
810 		if (icmp6len < sizeof(struct icmp6_router_renum))
811 			goto badlen;
812 		break;
813 
814 	default:
815 		nd6log((LOG_DEBUG,
816 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
817 		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
818 		    ip6_sprintf(&ip6->ip6_dst),
819 		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
820 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
821 			/* ICMPv6 error: MUST deliver it by spec... */
822 			code = PRC_NCMDS;
823 			/* deliver */
824 		} else {
825 			/* ICMPv6 informational: MUST not deliver */
826 			break;
827 		}
828 	deliver:
829 		if (icmp6_notify_error(m, off, icmp6len, code)) {
830 			/* In this case, m should've been freed. */
831 			return (IPPROTO_DONE);
832 		}
833 		break;
834 
835 	badcode:
836 		icmp6stat.icp6s_badcode++;
837 		break;
838 
839 	badlen:
840 		icmp6stat.icp6s_badlen++;
841 		break;
842 	}
843 
844 	/* deliver the packet to appropriate sockets */
845 	icmp6_rip6_input(&m, *offp);
846 
847 	return IPPROTO_DONE;
848 
849 freeit:
850 	m_freem(m);
851 	return IPPROTO_DONE;
852 }
853 
854 static int
855 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
856 {
857 	struct icmp6_hdr *icmp6;
858 	struct ip6_hdr *eip6;
859 	u_int32_t notifymtu;
860 	struct sockaddr_in6 icmp6src, icmp6dst;
861 
862 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
863 		icmp6stat.icp6s_tooshort++;
864 		goto freeit;
865 	}
866 #ifndef PULLDOWN_TEST
867 	IP6_EXTHDR_CHECK(m, off,
868 			 sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr),
869 			 -1);
870 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
871 #else
872 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
873 		       sizeof(*icmp6) + sizeof(struct ip6_hdr));
874 	if (icmp6 == NULL) {
875 		icmp6stat.icp6s_tooshort++;
876 		return (-1);
877 	}
878 #endif
879 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
880 
881 	/* Detect the upper level protocol */
882 	{
883 		void (*ctlfunc) (int, struct sockaddr *, void *);
884 		u_int8_t nxt = eip6->ip6_nxt;
885 		int eoff = off + sizeof(struct icmp6_hdr) +
886 			sizeof(struct ip6_hdr);
887 		struct ip6ctlparam ip6cp;
888 		struct in6_addr *finaldst = NULL;
889 		int icmp6type = icmp6->icmp6_type;
890 		struct ip6_frag *fh;
891 		struct ip6_rthdr *rth;
892 		int rthlen;
893 
894 		while (1) { /* XXX: should avoid infinite loop explicitly? */
895 			struct ip6_ext *eh;
896 
897 			switch (nxt) {
898 			case IPPROTO_HOPOPTS:
899 			case IPPROTO_DSTOPTS:
900 			case IPPROTO_AH:
901 #ifndef PULLDOWN_TEST
902 				IP6_EXTHDR_CHECK(m, 0, eoff +
903 						 sizeof(struct ip6_ext),
904 						 -1);
905 				eh = (struct ip6_ext *)(mtod(m, caddr_t)
906 							+ eoff);
907 #else
908 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
909 					       eoff, sizeof(*eh));
910 				if (eh == NULL) {
911 					icmp6stat.icp6s_tooshort++;
912 					return (-1);
913 				}
914 #endif
915 
916 				if (nxt == IPPROTO_AH)
917 					eoff += (eh->ip6e_len + 2) << 2;
918 				else
919 					eoff += (eh->ip6e_len + 1) << 3;
920 				nxt = eh->ip6e_nxt;
921 				break;
922 			case IPPROTO_ROUTING:
923 				/*
924 				 * When the erroneous packet contains a
925 				 * routing header, we should examine the
926 				 * header to determine the final destination.
927 				 * Otherwise, we can't properly update
928 				 * information that depends on the final
929 				 * destination (e.g. path MTU).
930 				 */
931 #ifndef PULLDOWN_TEST
932 				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth),
933 						 -1);
934 				rth = (struct ip6_rthdr *)(mtod(m, caddr_t)
935 							   + eoff);
936 #else
937 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
938 					       eoff, sizeof(*rth));
939 				if (rth == NULL) {
940 					icmp6stat.icp6s_tooshort++;
941 					return (-1);
942 				}
943 #endif
944 				rthlen = (rth->ip6r_len + 1) << 3;
945 				eoff += rthlen;
946 				nxt = rth->ip6r_nxt;
947 				break;
948 			case IPPROTO_FRAGMENT:
949 #ifndef PULLDOWN_TEST
950 				IP6_EXTHDR_CHECK(m, 0, eoff +
951 						 sizeof(struct ip6_frag),
952 						 -1);
953 				fh = (struct ip6_frag *)(mtod(m, caddr_t)
954 							 + eoff);
955 #else
956 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
957 					       eoff, sizeof(*fh));
958 				if (fh == NULL) {
959 					icmp6stat.icp6s_tooshort++;
960 					return (-1);
961 				}
962 #endif
963 				/*
964 				 * Data after a fragment header is meaningless
965 				 * unless it is the first fragment, but
966 				 * we'll go to the notify label for path MTU
967 				 * discovery.
968 				 */
969 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
970 					goto notify;
971 
972 				eoff += sizeof(struct ip6_frag);
973 				nxt = fh->ip6f_nxt;
974 				break;
975 			default:
976 				/*
977 				 * This case includes ESP and the No Next
978 				 * Header.  In such cases going to the notify
979 				 * label does not have any meaning
980 				 * (i.e. ctlfunc will be NULL), but we go
981 				 * anyway since we might have to update
982 				 * path MTU information.
983 				 */
984 				goto notify;
985 			}
986 		}
987 	  notify:
988 #ifndef PULLDOWN_TEST
989 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
990 #else
991 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
992 			       sizeof(*icmp6) + sizeof(struct ip6_hdr));
993 		if (icmp6 == NULL) {
994 			icmp6stat.icp6s_tooshort++;
995 			return (-1);
996 		}
997 #endif
998 
999 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1000 		bzero(&icmp6dst, sizeof(icmp6dst));
1001 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1002 		icmp6dst.sin6_family = AF_INET6;
1003 		if (finaldst == NULL)
1004 			icmp6dst.sin6_addr = eip6->ip6_dst;
1005 		else
1006 			icmp6dst.sin6_addr = *finaldst;
1007 		icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1008 							  &icmp6dst.sin6_addr);
1009 		if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1010 				   NULL, NULL)) {
1011 			/* should be impossbile */
1012 			nd6log((LOG_DEBUG,
1013 			    "icmp6_notify_error: in6_embedscope failed\n"));
1014 			goto freeit;
1015 		}
1016 
1017 		/*
1018 		 * retrieve parameters from the inner IPv6 header, and convert
1019 		 * them into sockaddr structures.
1020 		 */
1021 		bzero(&icmp6src, sizeof(icmp6src));
1022 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1023 		icmp6src.sin6_family = AF_INET6;
1024 		icmp6src.sin6_addr = eip6->ip6_src;
1025 		icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1026 							  &icmp6src.sin6_addr);
1027 		if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1028 				   NULL, NULL)) {
1029 			/* should be impossbile */
1030 			nd6log((LOG_DEBUG,
1031 			    "icmp6_notify_error: in6_embedscope failed\n"));
1032 			goto freeit;
1033 		}
1034 		icmp6src.sin6_flowinfo =
1035 			(eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1036 
1037 		if (finaldst == NULL)
1038 			finaldst = &eip6->ip6_dst;
1039 		ip6cp.ip6c_m = m;
1040 		ip6cp.ip6c_icmp6 = icmp6;
1041 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1042 		ip6cp.ip6c_off = eoff;
1043 		ip6cp.ip6c_finaldst = finaldst;
1044 		ip6cp.ip6c_src = &icmp6src;
1045 		ip6cp.ip6c_nxt = nxt;
1046 
1047 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1048 			notifymtu = ntohl(icmp6->icmp6_mtu);
1049 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1050 			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1051 		}
1052 
1053 		ctlfunc = (void (*) (int, struct sockaddr *, void *))
1054 			(inet6sw[ip6_protox[nxt]].pr_ctlinput);
1055 		if (ctlfunc) {
1056 			(*ctlfunc)(code, (struct sockaddr *)&icmp6dst, &ip6cp);
1057 		}
1058 	}
1059 	return (0);
1060 
1061 freeit:
1062 	m_freem(m);
1063 	return (-1);
1064 }
1065 
1066 void
1067 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1068 {
1069 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1070 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1071 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1072 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1073 	struct rtentry *rt = NULL;
1074 	struct sockaddr_in6 sin6;
1075 
1076 	if (!validated)
1077 		return;
1078 
1079 	bzero(&sin6, sizeof(sin6));
1080 	sin6.sin6_family = PF_INET6;
1081 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1082 	sin6.sin6_addr = *dst;
1083 	/* XXX normally, this won't happen */
1084 	if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1085 		sin6.sin6_addr.s6_addr16[1] =
1086 		    htons(m->m_pkthdr.rcvif->if_index);
1087 	}
1088 	/* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1089 	rt = rtpurelookup((struct sockaddr *)&sin6);
1090 	if (rt != NULL && (rt->rt_flags & RTF_HOST) &&
1091 	    !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1092 		if (mtu < IPV6_MMTU) {				/* XXX */
1093 			rt->rt_rmx.rmx_locks |= RTV_MTU;
1094 		} else if (mtu < rt->rt_ifp->if_mtu &&
1095 			   rt->rt_rmx.rmx_mtu > mtu) {
1096 			icmp6stat.icp6s_pmtuchg++;
1097 			rt->rt_rmx.rmx_mtu = mtu;
1098 		}
1099 	}
1100 	if (rt)
1101 		--rt->rt_refcnt;
1102 }
1103 
1104 /*
1105  * Process a Node Information Query packet, based on
1106  * draft-ietf-ipngwg-icmp-name-lookups-07.
1107  *
1108  * Spec incompatibilities:
1109  * - IPv6 Subject address handling
1110  * - IPv4 Subject address handling support missing
1111  * - Proxy reply (answer even if it's not for me)
1112  * - joins NI group address at in6_ifattach() time only, does not cope
1113  *   with hostname changes by sethostname(3)
1114  */
1115 #define hostnamelen	strlen(hostname)
1116 static struct mbuf *
1117 ni6_input(struct mbuf *m, int off)
1118 {
1119 	struct icmp6_nodeinfo *ni6, *nni6;
1120 	struct mbuf *n = NULL;
1121 	u_int16_t qtype;
1122 	int subjlen;
1123 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1124 	struct ni_reply_fqdn *fqdn;
1125 	int addrs;		/* for NI_QTYPE_NODEADDR */
1126 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1127 	struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1128 	struct sockaddr_in6 sin6_d; /* XXX: we should retrieve this from m_aux */
1129 	struct ip6_hdr *ip6;
1130 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1131 	char *subj = NULL;
1132 	struct in6_ifaddr *ia6 = NULL;
1133 
1134 	ip6 = mtod(m, struct ip6_hdr *);
1135 #ifndef PULLDOWN_TEST
1136 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1137 #else
1138 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1139 	if (ni6 == NULL) {
1140 		/* m is already reclaimed */
1141 		return NULL;
1142 	}
1143 #endif
1144 
1145 	/*
1146 	 * Validate IPv6 destination address.
1147 	 *
1148 	 * The Responder must discard the Query without further processing
1149 	 * unless it is one of the Responder's unicast or anycast addresses, or
1150 	 * a link-local scope multicast address which the Responder has joined.
1151 	 * [icmp-name-lookups-07, Section 4.]
1152 	 */
1153 	bzero(&sin6, sizeof(sin6));
1154 	sin6.sin6_family = AF_INET6;
1155 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1156 	bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1157 	/* XXX scopeid */
1158 	if ((ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)&sin6)) != NULL) {
1159 		/* unicast/anycast, fine */
1160 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1161 		    (icmp6_nodeinfo & 4) == 0) {
1162 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1163 				"a temporary address in %s:%d",
1164 			       __FILE__, __LINE__));
1165 			goto bad;
1166 		}
1167 	} else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1168 		; /* link-local multicast, fine */
1169 	else
1170 		goto bad;
1171 
1172 	/* validate query Subject field. */
1173 	qtype = ntohs(ni6->ni_qtype);
1174 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1175 	switch (qtype) {
1176 	case NI_QTYPE_NOOP:
1177 	case NI_QTYPE_SUPTYPES:
1178 		/* 07 draft */
1179 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1180 			break;
1181 		/* FALLTHROUGH */
1182 	case NI_QTYPE_FQDN:
1183 	case NI_QTYPE_NODEADDR:
1184 		switch (ni6->ni_code) {
1185 		case ICMP6_NI_SUBJ_IPV6:
1186 #if ICMP6_NI_SUBJ_IPV6 != 0
1187 		case 0:
1188 #endif
1189 			/*
1190 			 * backward compatibility - try to accept 03 draft
1191 			 * format, where no Subject is present.
1192 			 */
1193 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1194 			    subjlen == 0) {
1195 				oldfqdn++;
1196 				break;
1197 			}
1198 #if ICMP6_NI_SUBJ_IPV6 != 0
1199 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1200 				goto bad;
1201 #endif
1202 
1203 			if (subjlen != sizeof(sin6.sin6_addr))
1204 				goto bad;
1205 
1206 			/*
1207 			 * Validate Subject address.
1208 			 *
1209 			 * Not sure what exactly "address belongs to the node"
1210 			 * means in the spec, is it just unicast, or what?
1211 			 *
1212 			 * At this moment we consider Subject address as
1213 			 * "belong to the node" if the Subject address equals
1214 			 * to the IPv6 destination address; validation for
1215 			 * IPv6 destination address should have done enough
1216 			 * check for us.
1217 			 *
1218 			 * We do not do proxy at this moment.
1219 			 */
1220 			/* m_pulldown instead of copy? */
1221 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1222 			    subjlen, (caddr_t)&sin6.sin6_addr);
1223 			sin6.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1224 							      &sin6.sin6_addr);
1225 			in6_embedscope(&sin6.sin6_addr, &sin6, NULL, NULL);
1226 			bzero(&sin6_d, sizeof(sin6_d));
1227 			sin6_d.sin6_family = AF_INET6; /* not used, actually */
1228 			sin6_d.sin6_len = sizeof(sin6_d); /* ditto */
1229 			sin6_d.sin6_addr = ip6->ip6_dst;
1230 			sin6_d.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1231 								&ip6->ip6_dst);
1232 			in6_embedscope(&sin6_d.sin6_addr, &sin6_d, NULL, NULL);
1233 			subj = (char *)&sin6;
1234 			if (SA6_ARE_ADDR_EQUAL(&sin6, &sin6_d))
1235 				break;
1236 
1237 			/*
1238 			 * XXX if we are to allow other cases, we should really
1239 			 * be careful about scope here.
1240 			 * basically, we should disallow queries toward IPv6
1241 			 * destination X with subject Y, if scope(X) > scope(Y).
1242 			 * if we allow scope(X) > scope(Y), it will result in
1243 			 * information leakage across scope boundary.
1244 			 */
1245 			goto bad;
1246 
1247 		case ICMP6_NI_SUBJ_FQDN:
1248 			/*
1249 			 * Validate Subject name with gethostname(3).
1250 			 *
1251 			 * The behavior may need some debate, since:
1252 			 * - we are not sure if the node has FQDN as
1253 			 *   hostname (returned by gethostname(3)).
1254 			 * - the code does wildcard match for truncated names.
1255 			 *   however, we are not sure if we want to perform
1256 			 *   wildcard match, if gethostname(3) side has
1257 			 *   truncated hostname.
1258 			 */
1259 			n = ni6_nametodns(hostname, hostnamelen, 0);
1260 			if (!n || n->m_next || n->m_len == 0)
1261 				goto bad;
1262 			IP6_EXTHDR_GET(subj, char *, m,
1263 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1264 			if (subj == NULL)
1265 				goto bad;
1266 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1267 					n->m_len)) {
1268 				goto bad;
1269 			}
1270 			m_freem(n);
1271 			n = NULL;
1272 			break;
1273 
1274 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1275 		default:
1276 			goto bad;
1277 		}
1278 		break;
1279 	}
1280 
1281 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1282 	switch (qtype) {
1283 	case NI_QTYPE_FQDN:
1284 		if ((icmp6_nodeinfo & 1) == 0)
1285 			goto bad;
1286 		break;
1287 	case NI_QTYPE_NODEADDR:
1288 		if ((icmp6_nodeinfo & 2) == 0)
1289 			goto bad;
1290 		break;
1291 	}
1292 
1293 	/* guess reply length */
1294 	switch (qtype) {
1295 	case NI_QTYPE_NOOP:
1296 		break;		/* no reply data */
1297 	case NI_QTYPE_SUPTYPES:
1298 		replylen += sizeof(u_int32_t);
1299 		break;
1300 	case NI_QTYPE_FQDN:
1301 		/* XXX will append an mbuf */
1302 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1303 		break;
1304 	case NI_QTYPE_NODEADDR:
1305 		addrs = ni6_addrs(ni6, m, &ifp, subj);
1306 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1307 					  sizeof(u_int32_t))) > MCLBYTES)
1308 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1309 		break;
1310 	default:
1311 		/*
1312 		 * XXX: We must return a reply with the ICMP6 code
1313 		 * `unknown Qtype' in this case. However we regard the case
1314 		 * as an FQDN query for backward compatibility.
1315 		 * Older versions set a random value to this field,
1316 		 * so it rarely varies in the defined qtypes.
1317 		 * But the mechanism is not reliable...
1318 		 * maybe we should obsolete older versions.
1319 		 */
1320 		qtype = NI_QTYPE_FQDN;
1321 		/* XXX will append an mbuf */
1322 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1323 		oldfqdn++;
1324 		break;
1325 	}
1326 
1327 	/* allocate an mbuf to reply. */
1328 	if (replylen > MCLBYTES) {
1329 		/*
1330 		 * XXX: should we try to allocate more? But MCLBYTES
1331 		 * is probably much larger than IPV6_MMTU...
1332 		 */
1333 		goto bad;
1334 	}
1335 	n = m_getb(replylen, MB_DONTWAIT, m->m_type, M_PKTHDR);
1336 	if (n == NULL) {
1337 		m_freem(m);
1338 		return (NULL);
1339 	}
1340 	M_MOVE_PKTHDR(n, m); /* just for recvif */
1341 	n->m_pkthdr.len = n->m_len = replylen;
1342 
1343 	/* copy mbuf header and IPv6 + Node Information base headers */
1344 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1345 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1346 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1347 
1348 	/* qtype dependent procedure */
1349 	switch (qtype) {
1350 	case NI_QTYPE_NOOP:
1351 		nni6->ni_code = ICMP6_NI_SUCCESS;
1352 		nni6->ni_flags = 0;
1353 		break;
1354 	case NI_QTYPE_SUPTYPES:
1355 	{
1356 		u_int32_t v;
1357 		nni6->ni_code = ICMP6_NI_SUCCESS;
1358 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1359 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1360 		v = (u_int32_t)htonl(0x0000000f);
1361 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1362 		break;
1363 	}
1364 	case NI_QTYPE_FQDN:
1365 		nni6->ni_code = ICMP6_NI_SUCCESS;
1366 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1367 						sizeof(struct ip6_hdr) +
1368 						sizeof(struct icmp6_nodeinfo));
1369 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1370 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1371 		/*
1372 		 * XXX do we really have FQDN in variable "hostname"?
1373 		 */
1374 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1375 		if (n->m_next == NULL)
1376 			goto bad;
1377 		/* XXX we assume that n->m_next is not a chain */
1378 		if (n->m_next->m_next != NULL)
1379 			goto bad;
1380 		n->m_pkthdr.len += n->m_next->m_len;
1381 		break;
1382 	case NI_QTYPE_NODEADDR:
1383 	{
1384 		int lenlim, copied;
1385 
1386 		nni6->ni_code = ICMP6_NI_SUCCESS;
1387 		n->m_pkthdr.len = n->m_len =
1388 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1389 		lenlim = M_TRAILINGSPACE(n);
1390 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1391 		/* XXX: reset mbuf length */
1392 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1393 			sizeof(struct icmp6_nodeinfo) + copied;
1394 		break;
1395 	}
1396 	default:
1397 		break;		/* XXX impossible! */
1398 	}
1399 
1400 	nni6->ni_type = ICMP6_NI_REPLY;
1401 	m_freem(m);
1402 	return (n);
1403 
1404 bad:
1405 	m_freem(m);
1406 	if (n)
1407 		m_freem(n);
1408 	return (NULL);
1409 }
1410 #undef hostnamelen
1411 
1412 /*
1413  * make a mbuf with DNS-encoded string.  no compression support.
1414  *
1415  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1416  * treated as truncated name (two \0 at the end).  this is a wild guess.
1417  */
1418 static struct mbuf *
1419 ni6_nametodns(const char *name, int namelen,
1420 	      int old) /* return pascal string if non-zero */
1421 {
1422 	struct mbuf *m;
1423 	char *cp, *ep;
1424 	const char *p, *q;
1425 	int i, len, nterm;
1426 
1427 	if (old)
1428 		len = namelen + 1;
1429 	else
1430 		len = MCLBYTES;
1431 
1432 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1433 	m = m_getb(len, MB_DONTWAIT, MT_DATA, 0);
1434 	if (!m)
1435 		goto fail;
1436 
1437 	if (old) {
1438 		m->m_len = len;
1439 		*mtod(m, char *) = namelen;
1440 		bcopy(name, mtod(m, char *) + 1, namelen);
1441 		return m;
1442 	} else {
1443 		m->m_len = 0;
1444 		cp = mtod(m, char *);
1445 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1446 
1447 		/* if not certain about my name, return empty buffer */
1448 		if (namelen == 0)
1449 			return m;
1450 
1451 		/*
1452 		 * guess if it looks like shortened hostname, or FQDN.
1453 		 * shortened hostname needs two trailing "\0".
1454 		 */
1455 		i = 0;
1456 		for (p = name; p < name + namelen; p++) {
1457 			if (*p && *p == '.')
1458 				i++;
1459 		}
1460 		if (i < 2)
1461 			nterm = 2;
1462 		else
1463 			nterm = 1;
1464 
1465 		p = name;
1466 		while (cp < ep && p < name + namelen) {
1467 			i = 0;
1468 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1469 				i++;
1470 			/* result does not fit into mbuf */
1471 			if (cp + i + 1 >= ep)
1472 				goto fail;
1473 			/*
1474 			 * DNS label length restriction, RFC1035 page 8.
1475 			 * "i == 0" case is included here to avoid returning
1476 			 * 0-length label on "foo..bar".
1477 			 */
1478 			if (i <= 0 || i >= 64)
1479 				goto fail;
1480 			*cp++ = i;
1481 			bcopy(p, cp, i);
1482 			cp += i;
1483 			p = q;
1484 			if (p < name + namelen && *p == '.')
1485 				p++;
1486 		}
1487 		/* termination */
1488 		if (cp + nterm >= ep)
1489 			goto fail;
1490 		while (nterm-- > 0)
1491 			*cp++ = '\0';
1492 		m->m_len = cp - mtod(m, char *);
1493 		return m;
1494 	}
1495 
1496 	panic("should not reach here");
1497 	/* NOTREACHED */
1498 
1499 fail:
1500 	if (m)
1501 		m_freem(m);
1502 	return NULL;
1503 }
1504 
1505 /*
1506  * check if two DNS-encoded string matches.  takes care of truncated
1507  * form (with \0\0 at the end).  no compression support.
1508  * XXX upper/lowercase match (see RFC2065)
1509  */
1510 static int
1511 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1512 {
1513 	const char *a0, *b0;
1514 	int l;
1515 
1516 	/* simplest case - need validation? */
1517 	if (alen == blen && bcmp(a, b, alen) == 0)
1518 		return 1;
1519 
1520 	a0 = a;
1521 	b0 = b;
1522 
1523 	/* termination is mandatory */
1524 	if (alen < 2 || blen < 2)
1525 		return 0;
1526 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1527 		return 0;
1528 	alen--;
1529 	blen--;
1530 
1531 	while (a - a0 < alen && b - b0 < blen) {
1532 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1533 			return 0;
1534 
1535 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1536 			return 0;
1537 		/* we don't support compression yet */
1538 		if (a[0] >= 64 || b[0] >= 64)
1539 			return 0;
1540 
1541 		/* truncated case */
1542 		if (a[0] == 0 && a - a0 == alen - 1)
1543 			return 1;
1544 		if (b[0] == 0 && b - b0 == blen - 1)
1545 			return 1;
1546 		if (a[0] == 0 || b[0] == 0)
1547 			return 0;
1548 
1549 		if (a[0] != b[0])
1550 			return 0;
1551 		l = a[0];
1552 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1553 			return 0;
1554 		if (bcmp(a + 1, b + 1, l) != 0)
1555 			return 0;
1556 
1557 		a += 1 + l;
1558 		b += 1 + l;
1559 	}
1560 
1561 	if (a - a0 == alen && b - b0 == blen)
1562 		return 1;
1563 	else
1564 		return 0;
1565 }
1566 
1567 /*
1568  * calculate the number of addresses to be returned in the node info reply.
1569  */
1570 static int
1571 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1572 	  char *subj)
1573 {
1574 	struct ifnet *ifp;
1575 	struct in6_ifaddr *ifa6;
1576 	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1577 	int addrs = 0, addrsofif, iffound = 0;
1578 	int niflags = ni6->ni_flags;
1579 
1580 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1581 		switch (ni6->ni_code) {
1582 		case ICMP6_NI_SUBJ_IPV6:
1583 			if (subj == NULL) /* must be impossible... */
1584 				return (0);
1585 			subj_ip6 = (struct sockaddr_in6 *)subj;
1586 			break;
1587 		default:
1588 			/*
1589 			 * XXX: we only support IPv6 subject address for
1590 			 * this Qtype.
1591 			 */
1592 			return (0);
1593 		}
1594 	}
1595 
1596 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
1597 	{
1598 		struct ifaddr_container *ifac;
1599 
1600 		addrsofif = 0;
1601 		TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link)
1602 		{
1603 			struct ifaddr *ifa = ifac->ifa;
1604 
1605 			if (ifa->ifa_addr->sa_family != AF_INET6)
1606 				continue;
1607 			ifa6 = (struct in6_ifaddr *)ifa;
1608 
1609 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1610 			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1611 					       &ifa6->ia_addr.sin6_addr))
1612 				iffound = 1;
1613 
1614 			/*
1615 			 * IPv4-mapped addresses can only be returned by a
1616 			 * Node Information proxy, since they represent
1617 			 * addresses of IPv4-only nodes, which perforce do
1618 			 * not implement this protocol.
1619 			 * [icmp-name-lookups-07, Section 5.4]
1620 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1621 			 * this function at this moment.
1622 			 */
1623 
1624 			/* What do we have to do about ::1? */
1625 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1626 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1627 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1628 					continue;
1629 				break;
1630 			case IPV6_ADDR_SCOPE_SITELOCAL:
1631 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1632 					continue;
1633 				break;
1634 			case IPV6_ADDR_SCOPE_GLOBAL:
1635 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1636 					continue;
1637 				break;
1638 			default:
1639 				continue;
1640 			}
1641 
1642 			/*
1643 			 * check if anycast is okay.
1644 			 * XXX: just experimental.  not in the spec.
1645 			 */
1646 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) &&
1647 			    !(niflags & NI_NODEADDR_FLAG_ANYCAST))
1648 				continue; /* we need only unicast addresses */
1649 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) &&
1650 			    (icmp6_nodeinfo & 4) == 0) {
1651 				continue;
1652 			}
1653 			addrsofif++; /* count the address */
1654 		}
1655 		if (iffound) {
1656 			*ifpp = ifp;
1657 			return (addrsofif);
1658 		}
1659 
1660 		addrs += addrsofif;
1661 	}
1662 
1663 	return (addrs);
1664 }
1665 
1666 static int
1667 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1668 		struct ifnet *ifp0, int resid)
1669 {
1670 	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1671 	struct in6_ifaddr *ifa6;
1672 	struct ifnet *ifp_dep = NULL;
1673 	int copied = 0, allow_deprecated = 0;
1674 	u_char *cp = (u_char *)(nni6 + 1);
1675 	int niflags = ni6->ni_flags;
1676 	u_int32_t ltime;
1677 
1678 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1679 		return (0);	/* needless to copy */
1680 
1681 again:
1682 
1683 	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list))
1684 	{
1685 		struct ifaddr_container *ifac;
1686 
1687 		TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1688 			struct ifaddr *ifa = ifac->ifa;
1689 
1690 			if (ifa->ifa_addr->sa_family != AF_INET6)
1691 				continue;
1692 			ifa6 = (struct in6_ifaddr *)ifa;
1693 
1694 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) &&
1695 			    allow_deprecated == 0) {
1696 				/*
1697 				 * prefererred address should be put before
1698 				 * deprecated addresses.
1699 				 */
1700 
1701 				/* record the interface for later search */
1702 				if (ifp_dep == NULL)
1703 					ifp_dep = ifp;
1704 
1705 				continue;
1706 			}
1707 			else if (!(ifa6->ia6_flags & IN6_IFF_DEPRECATED) &&
1708 				 allow_deprecated != 0)
1709 				continue; /* we now collect deprecated addrs */
1710 
1711 			/* What do we have to do about ::1? */
1712 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1713 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1714 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1715 					continue;
1716 				break;
1717 			case IPV6_ADDR_SCOPE_SITELOCAL:
1718 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1719 					continue;
1720 				break;
1721 			case IPV6_ADDR_SCOPE_GLOBAL:
1722 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1723 					continue;
1724 				break;
1725 			default:
1726 				continue;
1727 			}
1728 
1729 			/*
1730 			 * check if anycast is okay.
1731 			 * XXX: just experimental. not in the spec.
1732 			 */
1733 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) &&
1734 			    !(niflags & NI_NODEADDR_FLAG_ANYCAST))
1735 				continue;
1736 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) &&
1737 			    (icmp6_nodeinfo & 4) == 0) {
1738 				continue;
1739 			}
1740 
1741 			/* now we can copy the address */
1742 			if (resid < sizeof(struct in6_addr) +
1743 			    sizeof(u_int32_t)) {
1744 				/*
1745 				 * We give up much more copy.
1746 				 * Set the truncate flag and return.
1747 				 */
1748 				nni6->ni_flags |=
1749 					NI_NODEADDR_FLAG_TRUNCATE;
1750 				return (copied);
1751 			}
1752 
1753 			/*
1754 			 * Set the TTL of the address.
1755 			 * The TTL value should be one of the following
1756 			 * according to the specification:
1757 			 *
1758 			 * 1. The remaining lifetime of a DHCP lease on the
1759 			 *    address, or
1760 			 * 2. The remaining Valid Lifetime of a prefix from
1761 			 *    which the address was derived through Stateless
1762 			 *    Autoconfiguration.
1763 			 *
1764 			 * Note that we currently do not support stateful
1765 			 * address configuration by DHCPv6, so the former
1766 			 * case can't happen.
1767 			 */
1768 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1769 				ltime = ND6_INFINITE_LIFETIME;
1770 			else {
1771 				if (ifa6->ia6_lifetime.ia6t_expire >
1772 				    time_second)
1773 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1774 				else
1775 					ltime = 0;
1776 			}
1777 
1778 			bcopy(&ltime, cp, sizeof(u_int32_t));
1779 			cp += sizeof(u_int32_t);
1780 
1781 			/* copy the address itself */
1782 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1783 			      sizeof(struct in6_addr));
1784 			/* XXX: KAME link-local hack; remove ifindex */
1785 			if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr))
1786 				((struct in6_addr *)cp)->s6_addr16[1] = 0;
1787 			cp += sizeof(struct in6_addr);
1788 
1789 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1790 			copied += (sizeof(struct in6_addr) +
1791 				   sizeof(u_int32_t));
1792 		}
1793 		if (ifp0)	/* we need search only on the specified IF */
1794 			break;
1795 	}
1796 
1797 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1798 		ifp = ifp_dep;
1799 		allow_deprecated = 1;
1800 
1801 		goto again;
1802 	}
1803 
1804 	return (copied);
1805 }
1806 
1807 /*
1808  * XXX almost dup'ed code with rip6_input.
1809  */
1810 static int
1811 icmp6_rip6_input(struct	mbuf **mp, int	off)
1812 {
1813 	struct mbuf *m = *mp;
1814 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1815 	struct in6pcb *in6p;
1816 	struct in6pcb *last = NULL;
1817 	struct sockaddr_in6 rip6src;
1818 	struct icmp6_hdr *icmp6;
1819 	struct mbuf *opts = NULL;
1820 
1821 #ifndef PULLDOWN_TEST
1822 	/* this is assumed to be safe. */
1823 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1824 #else
1825 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1826 	if (icmp6 == NULL) {
1827 		/* m is already reclaimed */
1828 		return IPPROTO_DONE;
1829 	}
1830 #endif
1831 
1832 	bzero(&rip6src, sizeof(rip6src));
1833 	rip6src.sin6_len = sizeof(struct sockaddr_in6);
1834 	rip6src.sin6_family = AF_INET6;
1835 	/* KAME hack: recover scopeid */
1836 	in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif);
1837 
1838 	LIST_FOREACH(in6p, &ripcbinfo.pcblisthead, inp_list)
1839 	{
1840 		if (in6p->inp_flags & INP_PLACEMARKER)
1841 			continue;
1842 		if ((in6p->inp_vflag & INP_IPV6) == 0)
1843 			continue;
1844 #ifdef HAVE_NRL_INPCB
1845 		if (!(in6p->in6p_flags & INP_IPV6))
1846 			continue;
1847 #endif
1848 		if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1849 			continue;
1850 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1851 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1852 			continue;
1853 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1854 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1855 			continue;
1856 		if (in6p->in6p_icmp6filt
1857 		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1858 				 in6p->in6p_icmp6filt))
1859 			continue;
1860 		if (last) {
1861 			struct	mbuf *n;
1862 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1863 				if (last->in6p_flags & IN6P_CONTROLOPTS)
1864 					ip6_savecontrol(last, &opts, ip6, n);
1865 				/* strip intermediate headers */
1866 				m_adj(n, off);
1867 				if (ssb_appendaddr(&last->in6p_socket->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 					}
1875 				} else
1876 					sorwakeup(last->in6p_socket);
1877 				opts = NULL;
1878 			}
1879 		}
1880 		last = in6p;
1881 	}
1882 	if (last) {
1883 		if (last->in6p_flags & IN6P_CONTROLOPTS)
1884 			ip6_savecontrol(last, &opts, ip6, m);
1885 		/* strip intermediate headers */
1886 		m_adj(m, off);
1887 		if (ssb_appendaddr(&last->in6p_socket->so_rcv,
1888 				 (struct sockaddr *)&rip6src, m, opts) == 0) {
1889 			m_freem(m);
1890 			if (opts)
1891 				m_freem(opts);
1892 		} else
1893 			sorwakeup(last->in6p_socket);
1894 	} else {
1895 		m_freem(m);
1896 		ip6stat.ip6s_delivered--;
1897 	}
1898 	return IPPROTO_DONE;
1899 }
1900 
1901 /*
1902  * Reflect the ip6 packet back to the source.
1903  * OFF points to the icmp6 header, counted from the top of the mbuf.
1904  */
1905 void
1906 icmp6_reflect(struct mbuf *m, size_t off)
1907 {
1908 	struct ip6_hdr *ip6;
1909 	struct icmp6_hdr *icmp6;
1910 	struct in6_ifaddr *ia;
1911 	struct in6_addr t, *src = 0;
1912 	int plen;
1913 	int type, code;
1914 	struct ifnet *outif = NULL;
1915 	struct sockaddr_in6 sa6_src, sa6_dst;
1916 #ifdef COMPAT_RFC1885
1917 	int mtu = IPV6_MMTU;
1918 	struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
1919 #endif
1920 
1921 	/* too short to reflect */
1922 	if (off < sizeof(struct ip6_hdr)) {
1923 		nd6log((LOG_DEBUG,
1924 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
1925 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
1926 		    __FILE__, __LINE__));
1927 		goto bad;
1928 	}
1929 
1930 	/*
1931 	 * If there are extra headers between IPv6 and ICMPv6, strip
1932 	 * off that header first.
1933 	 */
1934 #ifdef DIAGNOSTIC
1935 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
1936 		panic("assumption failed in icmp6_reflect");
1937 #endif
1938 	if (off > sizeof(struct ip6_hdr)) {
1939 		size_t l;
1940 		struct ip6_hdr nip6;
1941 
1942 		l = off - sizeof(struct ip6_hdr);
1943 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
1944 		m_adj(m, l);
1945 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1946 		if (m->m_len < l) {
1947 			if ((m = m_pullup(m, l)) == NULL)
1948 				return;
1949 		}
1950 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
1951 	} else /* off == sizeof(struct ip6_hdr) */ {
1952 		size_t l;
1953 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1954 		if (m->m_len < l) {
1955 			if ((m = m_pullup(m, l)) == NULL)
1956 				return;
1957 		}
1958 	}
1959 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
1960 	ip6 = mtod(m, struct ip6_hdr *);
1961 	ip6->ip6_nxt = IPPROTO_ICMPV6;
1962 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
1963 	type = icmp6->icmp6_type; /* keep type for statistics */
1964 	code = icmp6->icmp6_code; /* ditto. */
1965 
1966 	t = ip6->ip6_dst;
1967 	/*
1968 	 * ip6_input() drops a packet if its src is multicast.
1969 	 * So, the src is never multicast.
1970 	 */
1971 	ip6->ip6_dst = ip6->ip6_src;
1972 
1973 	/*
1974 	 * XXX: make sure to embed scope zone information, using
1975 	 * already embedded IDs or the received interface (if any).
1976 	 * Note that rcvif may be NULL.
1977 	 * TODO: scoped routing case (XXX).
1978 	 */
1979 	bzero(&sa6_src, sizeof(sa6_src));
1980 	sa6_src.sin6_family = AF_INET6;
1981 	sa6_src.sin6_len = sizeof(sa6_src);
1982 	sa6_src.sin6_addr = ip6->ip6_dst;
1983 	in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
1984 	in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL);
1985 	bzero(&sa6_dst, sizeof(sa6_dst));
1986 	sa6_dst.sin6_family = AF_INET6;
1987 	sa6_dst.sin6_len = sizeof(sa6_dst);
1988 	sa6_dst.sin6_addr = t;
1989 	in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
1990 	in6_embedscope(&t, &sa6_dst, NULL, NULL);
1991 
1992 #ifdef COMPAT_RFC1885
1993 	/*
1994 	 * xxx guess MTU
1995 	 * RFC 1885 requires that echo reply should be truncated if it
1996 	 * does not fit in with (return) path MTU, but the description was
1997 	 * removed in the new spec.
1998 	 */
1999 	if (icmp6_reflect_rt.ro_rt == 0 ||
2000 	    ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
2001 		if (icmp6_reflect_rt.ro_rt) {
2002 			RTFREE(icmp6_reflect_rt.ro_rt);
2003 			icmp6_reflect_rt.ro_rt = 0;
2004 		}
2005 		bzero(sin6, sizeof(*sin6));
2006 		sin6->sin6_family = PF_INET6;
2007 		sin6->sin6_len = sizeof(struct sockaddr_in6);
2008 		sin6->sin6_addr = ip6->ip6_dst;
2009 
2010 		rtalloc_ign((struct route *)&icmp6_reflect_rt.ro_rt,
2011 			    RTF_PRCLONING);
2012 	}
2013 
2014 	if (icmp6_reflect_rt.ro_rt == 0)
2015 		goto bad;
2016 
2017 	if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
2018 	    && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
2019 		mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
2020 
2021 	if (mtu < m->m_pkthdr.len) {
2022 		plen -= (m->m_pkthdr.len - mtu);
2023 		m_adj(m, mtu - m->m_pkthdr.len);
2024 	}
2025 #endif
2026 	/*
2027 	 * If the incoming packet was addressed directly to us(i.e. unicast),
2028 	 * use dst as the src for the reply.
2029 	 * The IN6_IFF_NOTREADY case would be VERY rare, but is possible
2030 	 * (for example) when we encounter an error while forwarding procedure
2031 	 * destined to a duplicated address of ours.
2032 	 */
2033 	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2034 		if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2035 		    (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2036 			src = &t;
2037 			break;
2038 		}
2039 	if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2040 		/*
2041 		 * This is the case if the dst is our link-local address
2042 		 * and the sender is also ourselves.
2043 		 */
2044 		src = &t;
2045 	}
2046 
2047 	if (src == 0) {
2048 		int e;
2049 		struct route_in6 ro;
2050 
2051 		/*
2052 		 * This case matches to multicasts, our anycast, or unicasts
2053 		 * that we do not own.  Select a source address based on the
2054 		 * source address of the erroneous packet.
2055 		 */
2056 		bzero(&ro, sizeof(ro));
2057 		src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e, NULL);
2058 		if (ro.ro_rt)
2059 			RTFREE(ro.ro_rt); /* XXX: we could use this */
2060 		if (src == NULL) {
2061 			nd6log((LOG_DEBUG,
2062 			    "icmp6_reflect: source can't be determined: "
2063 			    "dst=%s, error=%d\n",
2064 			    ip6_sprintf(&sa6_src.sin6_addr), e));
2065 			goto bad;
2066 		}
2067 	}
2068 
2069 	ip6->ip6_src = *src;
2070 
2071 	ip6->ip6_flow = 0;
2072 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2073 	ip6->ip6_vfc |= IPV6_VERSION;
2074 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2075 	if (m->m_pkthdr.rcvif) {
2076 		/* XXX: This may not be the outgoing interface */
2077 		ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2078 	} else
2079 		ip6->ip6_hlim = ip6_defhlim;
2080 
2081 	icmp6->icmp6_cksum = 0;
2082 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2083 					sizeof(struct ip6_hdr), plen);
2084 
2085 	/*
2086 	 * XXX option handling
2087 	 */
2088 
2089 	m->m_flags &= ~(M_BCAST|M_MCAST);
2090 
2091 #ifdef COMPAT_RFC1885
2092 	ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif, NULL);
2093 #else
2094 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2095 #endif
2096 	if (outif)
2097 		icmp6_ifoutstat_inc(outif, type, code);
2098 
2099 	return;
2100 
2101 bad:
2102 	m_freem(m);
2103 	return;
2104 }
2105 
2106 void
2107 icmp6_fasttimo(void)
2108 {
2109 
2110 	mld6_fasttimeo();
2111 }
2112 
2113 static const char *
2114 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2115 		    struct in6_addr *tgt6)
2116 {
2117 	static char buf[1024];
2118 	ksnprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2119 		ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2120 	return buf;
2121 }
2122 
2123 void
2124 icmp6_redirect_input(struct mbuf *m, int off)
2125 {
2126 	struct ifnet *ifp = m->m_pkthdr.rcvif;
2127 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2128 	struct nd_redirect *nd_rd;
2129 	int icmp6len = ntohs(ip6->ip6_plen);
2130 	char *lladdr = NULL;
2131 	int lladdrlen = 0;
2132 	u_char *redirhdr = NULL;
2133 	int redirhdrlen = 0;
2134 	struct rtentry *rt = NULL;
2135 	int is_router;
2136 	int is_onlink;
2137 	struct in6_addr src6 = ip6->ip6_src;
2138 	struct in6_addr redtgt6;
2139 	struct in6_addr reddst6;
2140 	union nd_opts ndopts;
2141 
2142 	if (!m || !ifp)
2143 		return;
2144 
2145 	/* XXX if we are router, we don't update route by icmp6 redirect */
2146 	if (ip6_forwarding)
2147 		goto freeit;
2148 	if (!icmp6_rediraccept)
2149 		goto freeit;
2150 
2151 #ifndef PULLDOWN_TEST
2152 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2153 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2154 #else
2155 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2156 	if (nd_rd == NULL) {
2157 		icmp6stat.icp6s_tooshort++;
2158 		return;
2159 	}
2160 #endif
2161 	redtgt6 = nd_rd->nd_rd_target;
2162 	reddst6 = nd_rd->nd_rd_dst;
2163 
2164 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2165 		redtgt6.s6_addr16[1] = htons(ifp->if_index);
2166 	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2167 		reddst6.s6_addr16[1] = htons(ifp->if_index);
2168 
2169 	/* validation */
2170 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2171 		nd6log((LOG_ERR,
2172 			"ICMP6 redirect sent from %s rejected; "
2173 			"must be from linklocal\n", ip6_sprintf(&src6)));
2174 		goto bad;
2175 	}
2176 	if (ip6->ip6_hlim != 255) {
2177 		nd6log((LOG_ERR,
2178 			"ICMP6 redirect sent from %s rejected; "
2179 			"hlim=%d (must be 255)\n",
2180 			ip6_sprintf(&src6), ip6->ip6_hlim));
2181 		goto bad;
2182 	}
2183     {
2184 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2185 	struct sockaddr_in6 sin6;
2186 	struct in6_addr *gw6;
2187 
2188 	bzero(&sin6, sizeof(sin6));
2189 	sin6.sin6_family = AF_INET6;
2190 	sin6.sin6_len = sizeof(struct sockaddr_in6);
2191 	bcopy(&reddst6, &sin6.sin6_addr, sizeof reddst6);
2192 	rt = rtpurelookup((struct sockaddr *)&sin6);
2193 	if (rt != NULL) {
2194 		if (rt->rt_gateway == NULL ||
2195 		    rt->rt_gateway->sa_family != AF_INET6) {
2196 			nd6log((LOG_ERR,
2197 			    "ICMP6 redirect rejected; no route "
2198 			    "with inet6 gateway found for redirect dst: %s\n",
2199 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2200 			--rt->rt_refcnt;
2201 			goto bad;
2202 		}
2203 
2204 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2205 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2206 			nd6log((LOG_ERR,
2207 				"ICMP6 redirect rejected; "
2208 				"not equal to gw-for-src=%s (must be same): "
2209 				"%s\n",
2210 				ip6_sprintf(gw6),
2211 				icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2212 			--rt->rt_refcnt;
2213 			goto bad;
2214 		}
2215 	} else {
2216 		nd6log((LOG_ERR,
2217 			"ICMP6 redirect rejected; "
2218 			"no route found for redirect dst: %s\n",
2219 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2220 		goto bad;
2221 	}
2222 	--rt->rt_refcnt;
2223 	rt = NULL;
2224     }
2225 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2226 		nd6log((LOG_ERR,
2227 			"ICMP6 redirect rejected; "
2228 			"redirect dst must be unicast: %s\n",
2229 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2230 		goto bad;
2231 	}
2232 
2233 	is_router = is_onlink = 0;
2234 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2235 		is_router = 1;	/* router case */
2236 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2237 		is_onlink = 1;	/* on-link destination case */
2238 	if (!is_router && !is_onlink) {
2239 		nd6log((LOG_ERR,
2240 			"ICMP6 redirect rejected; "
2241 			"neither router case nor onlink case: %s\n",
2242 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2243 		goto bad;
2244 	}
2245 	/* validation passed */
2246 
2247 	icmp6len -= sizeof(*nd_rd);
2248 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2249 	if (nd6_options(&ndopts) < 0) {
2250 		nd6log((LOG_INFO, "icmp6_redirect_input: "
2251 			"invalid ND option, rejected: %s\n",
2252 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2253 		/* nd6_options have incremented stats */
2254 		goto freeit;
2255 	}
2256 
2257 	if (ndopts.nd_opts_tgt_lladdr) {
2258 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2259 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2260 	}
2261 
2262 	if (ndopts.nd_opts_rh) {
2263 		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2264 		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2265 	}
2266 
2267 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2268 		nd6log((LOG_INFO,
2269 			"icmp6_redirect_input: lladdrlen mismatch for %s "
2270 			"(if %d, icmp6 packet %d): %s\n",
2271 			ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2272 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2273 		goto bad;
2274 	}
2275 
2276 	/* RFC 2461 8.3 */
2277 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2278 			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2279 
2280 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2281 		/* perform rtredirect */
2282 		struct sockaddr_in6 sdst;
2283 		struct sockaddr_in6 sgw;
2284 		struct sockaddr_in6 ssrc;
2285 
2286 		bzero(&sdst, sizeof(sdst));
2287 		bzero(&sgw, sizeof(sgw));
2288 		bzero(&ssrc, sizeof(ssrc));
2289 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2290 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2291 			sizeof(struct sockaddr_in6);
2292 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2293 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2294 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2295 		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2296 			   (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2297 			   (struct sockaddr *)&ssrc);
2298 	}
2299 	/* finally update cached route in each socket via kpfctlinput */
2300     {
2301 	struct sockaddr_in6 sdst;
2302 
2303 	bzero(&sdst, sizeof(sdst));
2304 	sdst.sin6_family = AF_INET6;
2305 	sdst.sin6_len = sizeof(struct sockaddr_in6);
2306 	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2307 	kpfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2308 #ifdef IPSEC
2309 	key_sa_routechange((struct sockaddr *)&sdst);
2310 #endif
2311     }
2312 
2313 freeit:
2314 	m_freem(m);
2315 	return;
2316 
2317 bad:
2318 	icmp6stat.icp6s_badredirect++;
2319 	m_freem(m);
2320 }
2321 
2322 void
2323 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2324 {
2325 	struct ifnet *ifp;	/* my outgoing interface */
2326 	struct in6_addr *ifp_ll6;
2327 	struct in6_addr *router_ll6;
2328 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2329 	struct mbuf *m = NULL;	/* newly allocated one */
2330 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2331 	struct nd_redirect *nd_rd;
2332 	size_t maxlen;
2333 	u_char *p;
2334 	struct ifnet *outif = NULL;
2335 	struct sockaddr_in6 src_sa;
2336 
2337 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2338 
2339 	/* if we are not router, we don't send icmp6 redirect */
2340 	if (!ip6_forwarding || ip6_accept_rtadv)
2341 		goto fail;
2342 
2343 	/* sanity check */
2344 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2345 		goto fail;
2346 
2347 	/*
2348 	 * Address check:
2349 	 *  the source address must identify a neighbor, and
2350 	 *  the destination address must not be a multicast address
2351 	 *  [RFC 2461, sec 8.2]
2352 	 */
2353 	sip6 = mtod(m0, struct ip6_hdr *);
2354 	bzero(&src_sa, sizeof(src_sa));
2355 	src_sa.sin6_family = AF_INET6;
2356 	src_sa.sin6_len = sizeof(src_sa);
2357 	src_sa.sin6_addr = sip6->ip6_src;
2358 	/* we don't currently use sin6_scope_id, but eventually use it */
2359 	src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src);
2360 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2361 		goto fail;
2362 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2363 		goto fail;	/* what should we do here? */
2364 
2365 	/* rate limit */
2366 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2367 		goto fail;
2368 
2369 	/*
2370 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2371 	 * we almost always ask for an mbuf cluster for simplicity.
2372 	 * (MHLEN < IPV6_MMTU is almost always true)
2373 	 */
2374 #if IPV6_MMTU >= MCLBYTES
2375 # error assumption failed about IPV6_MMTU and MCLBYTES
2376 #endif
2377 	m = m_getb(IPV6_MMTU, MB_DONTWAIT, MT_HEADER, M_PKTHDR);
2378 	if (!m)
2379 		goto fail;
2380 	m->m_len = 0;
2381 	maxlen = M_TRAILINGSPACE(m);
2382 	maxlen = min(IPV6_MMTU, maxlen);
2383 	/* just for safety */
2384 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2385 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2386 		goto fail;
2387 	}
2388 
2389 	{
2390 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2391 		struct in6_ifaddr *ia;
2392 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2393 						 IN6_IFF_NOTREADY|
2394 						 IN6_IFF_ANYCAST)) == NULL)
2395 			goto fail;
2396 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2397 	}
2398 
2399 	/* get ip6 linklocal address for the router. */
2400 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2401 		struct sockaddr_in6 *sin6;
2402 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2403 		router_ll6 = &sin6->sin6_addr;
2404 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2405 			router_ll6 = (struct in6_addr *)NULL;
2406 	} else
2407 		router_ll6 = (struct in6_addr *)NULL;
2408 
2409 	/* ip6 */
2410 	ip6 = mtod(m, struct ip6_hdr *);
2411 	ip6->ip6_flow = 0;
2412 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2413 	ip6->ip6_vfc |= IPV6_VERSION;
2414 	/* ip6->ip6_plen will be set later */
2415 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2416 	ip6->ip6_hlim = 255;
2417 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2418 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2419 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2420 
2421 	/* ND Redirect */
2422 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2423 	nd_rd->nd_rd_type = ND_REDIRECT;
2424 	nd_rd->nd_rd_code = 0;
2425 	nd_rd->nd_rd_reserved = 0;
2426 	if (rt->rt_flags & RTF_GATEWAY) {
2427 		/*
2428 		 * nd_rd->nd_rd_target must be a link-local address in
2429 		 * better router cases.
2430 		 */
2431 		if (!router_ll6)
2432 			goto fail;
2433 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2434 		      sizeof(nd_rd->nd_rd_target));
2435 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2436 		      sizeof(nd_rd->nd_rd_dst));
2437 	} else {
2438 		/* make sure redtgt == reddst */
2439 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2440 		      sizeof(nd_rd->nd_rd_target));
2441 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2442 		      sizeof(nd_rd->nd_rd_dst));
2443 	}
2444 
2445 	p = (u_char *)(nd_rd + 1);
2446 
2447 	if (!router_ll6)
2448 		goto nolladdropt;
2449 
2450     {
2451 	/* target lladdr option */
2452 	struct rtentry *rt_router = NULL;
2453 	int len;
2454 	struct sockaddr_dl *sdl;
2455 	struct nd_opt_hdr *nd_opt;
2456 	char *lladdr;
2457 
2458 	rt_router = nd6_lookup(router_ll6, 0, ifp);
2459 	if (!rt_router)
2460 		goto nolladdropt;
2461 	len = sizeof(*nd_opt) + ifp->if_addrlen;
2462 	len = (len + 7) & ~7;	/* round by 8 */
2463 	/* safety check */
2464 	if (len + (p - (u_char *)ip6) > maxlen)
2465 		goto nolladdropt;
2466 	if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2467 	    (rt_router->rt_flags & RTF_LLINFO) &&
2468 	    (rt_router->rt_gateway->sa_family == AF_LINK) &&
2469 	    (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2470 	    sdl->sdl_alen) {
2471 		nd_opt = (struct nd_opt_hdr *)p;
2472 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2473 		nd_opt->nd_opt_len = len >> 3;
2474 		lladdr = (char *)(nd_opt + 1);
2475 		bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2476 		p += len;
2477 	}
2478     }
2479 nolladdropt:;
2480 
2481 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2482 
2483 	/* just to be safe */
2484 #ifdef M_DECRYPTED	/*not openbsd*/
2485 	if (m0->m_flags & M_DECRYPTED)
2486 		goto noredhdropt;
2487 #endif
2488 	if (p - (u_char *)ip6 > maxlen)
2489 		goto noredhdropt;
2490 
2491     {
2492 	/* redirected header option */
2493 	int len;
2494 	struct nd_opt_rd_hdr *nd_opt_rh;
2495 
2496 	/*
2497 	 * compute the maximum size for icmp6 redirect header option.
2498 	 * XXX room for auth header?
2499 	 */
2500 	len = maxlen - (p - (u_char *)ip6);
2501 	len &= ~7;
2502 
2503 	/* This is just for simplicity. */
2504 	if (m0->m_pkthdr.len != m0->m_len) {
2505 		if (m0->m_next) {
2506 			m_freem(m0->m_next);
2507 			m0->m_next = NULL;
2508 		}
2509 		m0->m_pkthdr.len = m0->m_len;
2510 	}
2511 
2512 	/*
2513 	 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2514 	 * about padding/truncate rule for the original IP packet.
2515 	 * From the discussion on IPv6imp in Feb 1999, the consensus was:
2516 	 * - "attach as much as possible" is the goal
2517 	 * - pad if not aligned (original size can be guessed by original
2518 	 *   ip6 header)
2519 	 * Following code adds the padding if it is simple enough,
2520 	 * and truncates if not.
2521 	 */
2522 	if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2523 		panic("assumption failed in %s:%d", __FILE__, __LINE__);
2524 
2525 	if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2526 		/* not enough room, truncate */
2527 		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2528 	} else {
2529 		/* enough room, pad or truncate */
2530 		size_t extra;
2531 
2532 		extra = m0->m_pkthdr.len % 8;
2533 		if (extra) {
2534 			/* pad if easy enough, truncate if not */
2535 			if (8 - extra <= M_TRAILINGSPACE(m0)) {
2536 				/* pad */
2537 				m0->m_len += (8 - extra);
2538 				m0->m_pkthdr.len += (8 - extra);
2539 			} else {
2540 				/* truncate */
2541 				m0->m_pkthdr.len -= extra;
2542 				m0->m_len -= extra;
2543 			}
2544 		}
2545 		len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2546 		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2547 	}
2548 
2549 	nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2550 	bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2551 	nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2552 	nd_opt_rh->nd_opt_rh_len = len >> 3;
2553 	p += sizeof(*nd_opt_rh);
2554 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2555 
2556 	/* connect m0 to m */
2557 	m->m_next = m0;
2558 	m->m_pkthdr.len = m->m_len + m0->m_len;
2559     }
2560 noredhdropt:;
2561 
2562 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
2563 		sip6->ip6_src.s6_addr16[1] = 0;
2564 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
2565 		sip6->ip6_dst.s6_addr16[1] = 0;
2566 #if 0
2567 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
2568 		ip6->ip6_src.s6_addr16[1] = 0;
2569 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
2570 		ip6->ip6_dst.s6_addr16[1] = 0;
2571 #endif
2572 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
2573 		nd_rd->nd_rd_target.s6_addr16[1] = 0;
2574 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
2575 		nd_rd->nd_rd_dst.s6_addr16[1] = 0;
2576 
2577 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2578 
2579 	nd_rd->nd_rd_cksum = 0;
2580 	nd_rd->nd_rd_cksum
2581 		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2582 
2583 	/* send the packet to outside... */
2584 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2585 	if (outif) {
2586 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2587 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2588 	}
2589 	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2590 
2591 	return;
2592 
2593 fail:
2594 	if (m)
2595 		m_freem(m);
2596 	if (m0)
2597 		m_freem(m0);
2598 }
2599 
2600 #ifdef HAVE_NRL_INPCB
2601 #define in6pcb		inpcb
2602 #define in6p_icmp6filt	inp_icmp6filt
2603 #endif
2604 /*
2605  * ICMPv6 socket option processing.
2606  */
2607 int
2608 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2609 {
2610 	int error = 0;
2611 	int optlen;
2612 	struct inpcb *inp = so->so_pcb;
2613 	int level, op, optname;
2614 
2615 	if (sopt) {
2616 		level = sopt->sopt_level;
2617 		op = sopt->sopt_dir;
2618 		optname = sopt->sopt_name;
2619 		optlen = sopt->sopt_valsize;
2620 	} else
2621 		level = op = optname = optlen = 0;
2622 
2623 	if (level != IPPROTO_ICMPV6) {
2624 		return EINVAL;
2625 	}
2626 
2627 	switch (op) {
2628 	case PRCO_SETOPT:
2629 		switch (optname) {
2630 		case ICMP6_FILTER:
2631 		    {
2632 			struct icmp6_filter *p;
2633 
2634 			if (optlen != sizeof(*p)) {
2635 				error = EMSGSIZE;
2636 				break;
2637 			}
2638 			if (inp->in6p_icmp6filt == NULL) {
2639 				error = EINVAL;
2640 				break;
2641 			}
2642 			error = soopt_to_kbuf(sopt, inp->in6p_icmp6filt, optlen,
2643 				optlen);
2644 			break;
2645 		    }
2646 
2647 		default:
2648 			error = ENOPROTOOPT;
2649 			break;
2650 		}
2651 		break;
2652 
2653 	case PRCO_GETOPT:
2654 		switch (optname) {
2655 		case ICMP6_FILTER:
2656 		    {
2657 			if (inp->in6p_icmp6filt == NULL) {
2658 				error = EINVAL;
2659 				break;
2660 			}
2661 			soopt_from_kbuf(sopt, inp->in6p_icmp6filt,
2662 				sizeof(struct icmp6_filter));
2663 			break;
2664 		    }
2665 
2666 		default:
2667 			error = ENOPROTOOPT;
2668 			break;
2669 		}
2670 		break;
2671 	}
2672 
2673 	return (error);
2674 }
2675 #ifdef HAVE_NRL_INPCB
2676 #undef in6pcb
2677 #undef in6p_icmp6filt
2678 #endif
2679 
2680 #ifndef HAVE_PPSRATECHECK
2681 #ifndef timersub
2682 #define	timersub(tvp, uvp, vvp)						\
2683 	do {								\
2684 		(vvp)->tv_sec = (tvp)->tv_sec - (uvp)->tv_sec;		\
2685 		(vvp)->tv_usec = (tvp)->tv_usec - (uvp)->tv_usec;	\
2686 		if ((vvp)->tv_usec < 0) {				\
2687 			(vvp)->tv_sec--;				\
2688 			(vvp)->tv_usec += 1000000;			\
2689 		}							\
2690 	} while (0)
2691 #endif
2692 
2693 /*
2694  * ppsratecheck(): packets (or events) per second limitation.
2695  */
2696 static int
2697 ppsratecheck(struct timeval *lasttime, int *curpps,
2698 	     int maxpps)	/* maximum pps allowed */
2699 {
2700 	struct timeval tv, delta;
2701 	int rv;
2702 
2703 	microtime(&tv);
2704 
2705 	timersub(&tv, lasttime, &delta);
2706 
2707 	/*
2708 	 * Check for 0,0 so that the message will be seen at least once.
2709 	 * If more than one second has passed since the last update of
2710 	 * lasttime, reset the counter.
2711 	 *
2712 	 * We do increment *curpps even in *curpps < maxpps case, as some may
2713 	 * try to use *curpps for stat purposes as well.
2714 	 */
2715 	if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) ||
2716 	    delta.tv_sec >= 1) {
2717 		*lasttime = tv;
2718 		*curpps = 0;
2719 		rv = 1;
2720 	} else if (maxpps < 0)
2721 		rv = 1;
2722 	else if (*curpps < maxpps)
2723 		rv = 1;
2724 	else
2725 		rv = 0;
2726 
2727 #if 1 /* DIAGNOSTIC? */
2728 	/* be careful about wrap-around */
2729 	if (*curpps + 1 > *curpps)
2730 		*curpps = *curpps + 1;
2731 #else
2732 	/*
2733 	 * assume that there's not too many calls to this function.
2734 	 * not sure if the assumption holds, as it depends on *caller's*
2735 	 * behavior, not the behavior of this function.
2736 	 * IMHO it is wrong to make assumption on the caller's behavior,
2737 	 * so the above #if is #if 1, not #ifdef DIAGNOSTIC.
2738 	 */
2739 	*curpps = *curpps + 1;
2740 #endif
2741 
2742 	return (rv);
2743 }
2744 #endif
2745 
2746 /*
2747  * Perform rate limit check.
2748  * Returns 0 if it is okay to send the icmp6 packet.
2749  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2750  * limitation.
2751  *
2752  * XXX per-destination/type check necessary?
2753  */
2754 static int
2755 icmp6_ratelimit(const struct in6_addr *dst,	/* not used at this moment */
2756 		const int type,			/* not used at this moment */
2757 		const int code)			/* not used at this moment */
2758 {
2759 	int ret;
2760 
2761 	ret = 0;	/* okay to send */
2762 
2763 	/* PPS limit */
2764 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2765 	    icmp6errppslim)) {
2766 		/* The packet is subject to rate limit */
2767 		ret++;
2768 	}
2769 
2770 	return ret;
2771 }
2772