xref: /freebsd/sys/netinet6/icmp6.c (revision a3e8fd0b7f663db7eafff527d5c3ca3bcfa8a537)
1 /*	$FreeBSD$	*/
2 /*	$KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1988, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. All advertising materials mentioning features or use of this software
46  *    must display the following acknowledgement:
47  *	This product includes software developed by the University of
48  *	California, Berkeley and its contributors.
49  * 4. Neither the name of the University nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63  * SUCH DAMAGE.
64  *
65  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
66  */
67 
68 #include "opt_inet.h"
69 #include "opt_inet6.h"
70 #include "opt_ipsec.h"
71 
72 #include <sys/param.h>
73 #include <sys/domain.h>
74 #include <sys/kernel.h>
75 #include <sys/lock.h>
76 #include <sys/malloc.h>
77 #include <sys/mbuf.h>
78 #include <sys/protosw.h>
79 #include <sys/signalvar.h>
80 #include <sys/socket.h>
81 #include <sys/socketvar.h>
82 #include <sys/sx.h>
83 #include <sys/syslog.h>
84 #include <sys/systm.h>
85 #include <sys/time.h>
86 
87 #include <net/if.h>
88 #include <net/if_dl.h>
89 #include <net/if_types.h>
90 #include <net/route.h>
91 
92 #include <netinet/in.h>
93 #include <netinet/in_pcb.h>
94 #include <netinet/in_var.h>
95 #include <netinet/ip6.h>
96 #include <netinet/icmp6.h>
97 #include <netinet6/in6_ifattach.h>
98 #include <netinet6/in6_pcb.h>
99 #include <netinet6/ip6protosw.h>
100 #include <netinet6/ip6_var.h>
101 #include <netinet6/mld6_var.h>
102 #include <netinet6/nd6.h>
103 
104 #ifdef IPSEC
105 #include <netinet6/ipsec.h>
106 #include <netkey/key.h>
107 #endif
108 
109 #ifdef FAST_IPSEC
110 #include <netipsec/ipsec.h>
111 #include <netipsec/key.h>
112 #define	IPSEC
113 #endif
114 
115 #include <net/net_osdep.h>
116 
117 #ifdef HAVE_NRL_INPCB
118 /* inpcb members */
119 #define in6pcb		inpcb
120 #define in6p_laddr	inp_laddr6
121 #define in6p_faddr	inp_faddr6
122 #define in6p_icmp6filt	inp_icmp6filt
123 #define in6p_route	inp_route
124 #define in6p_socket	inp_socket
125 #define in6p_flags	inp_flags
126 #define in6p_moptions	inp_moptions6
127 #define in6p_outputopts	inp_outputopts6
128 #define in6p_ip6	inp_ipv6
129 #define in6p_flowinfo	inp_flowinfo
130 #define in6p_sp		inp_sp
131 #define in6p_next	inp_next
132 #define in6p_prev	inp_prev
133 /* macro names */
134 #define sotoin6pcb	sotoinpcb
135 /* function names */
136 #define in6_pcbdetach	in_pcbdetach
137 #define in6_rtchange	in_rtchange
138 
139 /*
140  * for KAME src sync over BSD*'s. XXX: FreeBSD (>=3) are VERY different from
141  * others...
142  */
143 #define in6p_ip6_nxt	inp_ipv6.ip6_nxt
144 #endif
145 
146 extern struct domain inet6domain;
147 
148 struct icmp6stat icmp6stat;
149 
150 extern struct inpcbhead ripcb;
151 extern int icmp6errppslim;
152 static int icmp6errpps_count = 0;
153 static struct timeval icmp6errppslim_last;
154 extern int icmp6_nodeinfo;
155 
156 static void icmp6_errcount __P((struct icmp6errstat *, int, int));
157 static int icmp6_rip6_input __P((struct mbuf **, int));
158 static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int));
159 static const char *icmp6_redirect_diag __P((struct in6_addr *,
160 	struct in6_addr *, struct in6_addr *));
161 #ifndef HAVE_PPSRATECHECK
162 static int ppsratecheck __P((struct timeval *, int *, int));
163 #endif
164 static struct mbuf *ni6_input __P((struct mbuf *, int));
165 static struct mbuf *ni6_nametodns __P((const char *, int, int));
166 static int ni6_dnsmatch __P((const char *, int, const char *, int));
167 static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
168 			  struct ifnet **, char *));
169 static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
170 				struct ifnet *, int));
171 static int icmp6_notify_error __P((struct mbuf *, int, int, int));
172 
173 #ifdef COMPAT_RFC1885
174 static struct route_in6 icmp6_reflect_rt;
175 #endif
176 
177 
178 void
179 icmp6_init()
180 {
181 	mld6_init();
182 }
183 
184 static void
185 icmp6_errcount(stat, type, code)
186 	struct icmp6errstat *stat;
187 	int type, code;
188 {
189 	switch (type) {
190 	case ICMP6_DST_UNREACH:
191 		switch (code) {
192 		case ICMP6_DST_UNREACH_NOROUTE:
193 			stat->icp6errs_dst_unreach_noroute++;
194 			return;
195 		case ICMP6_DST_UNREACH_ADMIN:
196 			stat->icp6errs_dst_unreach_admin++;
197 			return;
198 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
199 			stat->icp6errs_dst_unreach_beyondscope++;
200 			return;
201 		case ICMP6_DST_UNREACH_ADDR:
202 			stat->icp6errs_dst_unreach_addr++;
203 			return;
204 		case ICMP6_DST_UNREACH_NOPORT:
205 			stat->icp6errs_dst_unreach_noport++;
206 			return;
207 		}
208 		break;
209 	case ICMP6_PACKET_TOO_BIG:
210 		stat->icp6errs_packet_too_big++;
211 		return;
212 	case ICMP6_TIME_EXCEEDED:
213 		switch (code) {
214 		case ICMP6_TIME_EXCEED_TRANSIT:
215 			stat->icp6errs_time_exceed_transit++;
216 			return;
217 		case ICMP6_TIME_EXCEED_REASSEMBLY:
218 			stat->icp6errs_time_exceed_reassembly++;
219 			return;
220 		}
221 		break;
222 	case ICMP6_PARAM_PROB:
223 		switch (code) {
224 		case ICMP6_PARAMPROB_HEADER:
225 			stat->icp6errs_paramprob_header++;
226 			return;
227 		case ICMP6_PARAMPROB_NEXTHEADER:
228 			stat->icp6errs_paramprob_nextheader++;
229 			return;
230 		case ICMP6_PARAMPROB_OPTION:
231 			stat->icp6errs_paramprob_option++;
232 			return;
233 		}
234 		break;
235 	case ND_REDIRECT:
236 		stat->icp6errs_redirect++;
237 		return;
238 	}
239 	stat->icp6errs_unknown++;
240 }
241 
242 /*
243  * Generate an error packet of type error in response to bad IP6 packet.
244  */
245 void
246 icmp6_error(m, type, code, param)
247 	struct mbuf *m;
248 	int type, code, param;
249 {
250 	struct ip6_hdr *oip6, *nip6;
251 	struct icmp6_hdr *icmp6;
252 	u_int preplen;
253 	int off;
254 	int nxt;
255 
256 	icmp6stat.icp6s_error++;
257 
258 	/* count per-type-code statistics */
259 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
260 
261 #ifdef M_DECRYPTED	/*not openbsd*/
262 	if (m->m_flags & M_DECRYPTED) {
263 		icmp6stat.icp6s_canterror++;
264 		goto freeit;
265 	}
266 #endif
267 
268 #ifndef PULLDOWN_TEST
269 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
270 #else
271 	if (m->m_len < sizeof(struct ip6_hdr)) {
272 		m = m_pullup(m, sizeof(struct ip6_hdr));
273 		if (m == NULL)
274 			return;
275 	}
276 #endif
277 	oip6 = mtod(m, struct ip6_hdr *);
278 
279 	/*
280 	 * Multicast destination check. For unrecognized option errors,
281 	 * this check has already done in ip6_unknown_opt(), so we can
282 	 * check only for other errors.
283 	 */
284 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
285 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
286 	    (type != ICMP6_PACKET_TOO_BIG &&
287 	     (type != ICMP6_PARAM_PROB ||
288 	      code != ICMP6_PARAMPROB_OPTION)))
289 		goto freeit;
290 
291 	/* Source address check. XXX: the case of anycast source? */
292 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
293 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
294 		goto freeit;
295 
296 	/*
297 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
298 	 * don't do it.
299 	 */
300 	nxt = -1;
301 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
302 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
303 		struct icmp6_hdr *icp;
304 
305 #ifndef PULLDOWN_TEST
306 		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
307 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
308 #else
309 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
310 			sizeof(*icp));
311 		if (icp == NULL) {
312 			icmp6stat.icp6s_tooshort++;
313 			return;
314 		}
315 #endif
316 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
317 		    icp->icmp6_type == ND_REDIRECT) {
318 			/*
319 			 * ICMPv6 error
320 			 * Special case: for redirect (which is
321 			 * informational) we must not send icmp6 error.
322 			 */
323 			icmp6stat.icp6s_canterror++;
324 			goto freeit;
325 		} else {
326 			/* ICMPv6 informational - send the error */
327 		}
328 	} else {
329 		/* non-ICMPv6 - send the error */
330 	}
331 
332 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
333 
334 	/* Finally, do rate limitation check. */
335 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
336 		icmp6stat.icp6s_toofreq++;
337 		goto freeit;
338 	}
339 
340 	/*
341 	 * OK, ICMP6 can be generated.
342 	 */
343 
344 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
345 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
346 
347 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
348 	M_PREPEND(m, preplen, M_DONTWAIT);
349 	if (m && m->m_len < preplen)
350 		m = m_pullup(m, preplen);
351 	if (m == NULL) {
352 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
353 		return;
354 	}
355 
356 	nip6 = mtod(m, struct ip6_hdr *);
357 	nip6->ip6_src  = oip6->ip6_src;
358 	nip6->ip6_dst  = oip6->ip6_dst;
359 
360 	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src))
361 		oip6->ip6_src.s6_addr16[1] = 0;
362 	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst))
363 		oip6->ip6_dst.s6_addr16[1] = 0;
364 
365 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
366 	icmp6->icmp6_type = type;
367 	icmp6->icmp6_code = code;
368 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
369 
370 	/*
371 	 * icmp6_reflect() is designed to be in the input path.
372 	 * icmp6_error() can be called from both input and outut path,
373 	 * and if we are in output path rcvif could contain bogus value.
374 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
375 	 * information in ip header (nip6).
376 	 */
377 	m->m_pkthdr.rcvif = NULL;
378 
379 	icmp6stat.icp6s_outhist[type]++;
380 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
381 
382 	return;
383 
384   freeit:
385 	/*
386 	 * If we can't tell wheter or not we can generate ICMP6, free it.
387 	 */
388 	m_freem(m);
389 }
390 
391 /*
392  * Process a received ICMP6 message.
393  */
394 int
395 icmp6_input(mp, offp, proto)
396 	struct mbuf **mp;
397 	int *offp, proto;
398 {
399 	struct mbuf *m = *mp, *n;
400 	struct ip6_hdr *ip6, *nip6;
401 	struct icmp6_hdr *icmp6, *nicmp6;
402 	int off = *offp;
403 	int icmp6len = m->m_pkthdr.len - *offp;
404 	int code, sum, noff;
405 
406 #ifndef PULLDOWN_TEST
407 	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
408 	/* m might change if M_LOOP.  So, call mtod after this */
409 #endif
410 
411 	/*
412 	 * Locate icmp6 structure in mbuf, and check
413 	 * that not corrupted and of at least minimum length
414 	 */
415 
416 	ip6 = mtod(m, struct ip6_hdr *);
417 	if (icmp6len < sizeof(struct icmp6_hdr)) {
418 		icmp6stat.icp6s_tooshort++;
419 		goto freeit;
420 	}
421 
422 	/*
423 	 * calculate the checksum
424 	 */
425 #ifndef PULLDOWN_TEST
426 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
427 #else
428 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
429 	if (icmp6 == NULL) {
430 		icmp6stat.icp6s_tooshort++;
431 		return IPPROTO_DONE;
432 	}
433 #endif
434 	code = icmp6->icmp6_code;
435 
436 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
437 		nd6log((LOG_ERR,
438 		    "ICMP6 checksum error(%d|%x) %s\n",
439 		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
440 		icmp6stat.icp6s_checksum++;
441 		goto freeit;
442 	}
443 
444 	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
445 		/*
446 		 * Deliver very specific ICMP6 type only.
447 		 * This is important to deilver TOOBIG.  Otherwise PMTUD
448 		 * will not work.
449 		 */
450 		switch (icmp6->icmp6_type) {
451 		case ICMP6_DST_UNREACH:
452 		case ICMP6_PACKET_TOO_BIG:
453 		case ICMP6_TIME_EXCEEDED:
454 			break;
455 		default:
456 			goto freeit;
457 		}
458 	}
459 
460 	icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
461 	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
462 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
463 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
464 
465 	switch (icmp6->icmp6_type) {
466 	case ICMP6_DST_UNREACH:
467 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
468 		switch (code) {
469 		case ICMP6_DST_UNREACH_NOROUTE:
470 			code = PRC_UNREACH_NET;
471 			break;
472 		case ICMP6_DST_UNREACH_ADMIN:
473 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
474 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
475 			break;
476 		case ICMP6_DST_UNREACH_ADDR:
477 			code = PRC_HOSTDEAD;
478 			break;
479 #ifdef COMPAT_RFC1885
480 		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
481 			code = PRC_UNREACH_SRCFAIL;
482 			break;
483 #else
484 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
485 			/* I mean "source address was incorrect." */
486 			code = PRC_PARAMPROB;
487 			break;
488 #endif
489 		case ICMP6_DST_UNREACH_NOPORT:
490 			code = PRC_UNREACH_PORT;
491 			break;
492 		default:
493 			goto badcode;
494 		}
495 		goto deliver;
496 		break;
497 
498 	case ICMP6_PACKET_TOO_BIG:
499 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
500 		if (code != 0)
501 			goto badcode;
502 
503 		code = PRC_MSGSIZE;
504 
505 		/*
506 		 * Updating the path MTU will be done after examining
507 		 * intermediate extension headers.
508 		 */
509 		goto deliver;
510 		break;
511 
512 	case ICMP6_TIME_EXCEEDED:
513 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
514 		switch (code) {
515 		case ICMP6_TIME_EXCEED_TRANSIT:
516 		case ICMP6_TIME_EXCEED_REASSEMBLY:
517 			code += PRC_TIMXCEED_INTRANS;
518 			break;
519 		default:
520 			goto badcode;
521 		}
522 		goto deliver;
523 		break;
524 
525 	case ICMP6_PARAM_PROB:
526 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
527 		switch (code) {
528 		case ICMP6_PARAMPROB_NEXTHEADER:
529 			code = PRC_UNREACH_PROTOCOL;
530 			break;
531 		case ICMP6_PARAMPROB_HEADER:
532 		case ICMP6_PARAMPROB_OPTION:
533 			code = PRC_PARAMPROB;
534 			break;
535 		default:
536 			goto badcode;
537 		}
538 		goto deliver;
539 		break;
540 
541 	case ICMP6_ECHO_REQUEST:
542 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
543 		if (code != 0)
544 			goto badcode;
545 		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
546 			/* Give up remote */
547 			break;
548 		}
549 		if ((n->m_flags & M_EXT) != 0
550 		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
551 			struct mbuf *n0 = n;
552 			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
553 
554 			/*
555 			 * Prepare an internal mbuf. m_pullup() doesn't
556 			 * always copy the length we specified.
557 			 */
558 			if (maxlen >= MCLBYTES) {
559 				/* Give up remote */
560 				m_freem(n0);
561 				break;
562 			}
563 			MGETHDR(n, M_DONTWAIT, n0->m_type);
564 			if (n && maxlen >= MHLEN) {
565 				MCLGET(n, M_DONTWAIT);
566 				if ((n->m_flags & M_EXT) == 0) {
567 					m_free(n);
568 					n = NULL;
569 				}
570 			}
571 			if (n == NULL) {
572 				/* Give up remote */
573 				m_freem(n0);
574 				break;
575 			}
576 			M_COPY_PKTHDR(n, n0);
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 			n->m_pkthdr.len = n->m_len =
586 				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 			n->m_pkthdr.len += n0->m_pkthdr.len;
593 			n->m_next = n0;
594 			n0->m_flags &= ~M_PKTHDR;
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, M_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, M_DONTWAIT, m->m_type);
685 			if (n && maxlen > MHLEN) {
686 				MCLGET(n, M_DONTWAIT);
687 				if ((n->m_flags & M_EXT) == 0) {
688 					m_free(n);
689 					n = NULL;
690 				}
691 			}
692 			if (n == NULL) {
693 				/* Give up remote */
694 				break;
695 			}
696 			n->m_pkthdr.rcvif = NULL;
697 			n->m_len = 0;
698 			maxhlen = M_TRAILINGSPACE(n) - maxlen;
699 			if (maxhlen > hostnamelen)
700 				maxhlen = hostnamelen;
701 			/*
702 			 * Copy IPv6 and ICMPv6 only.
703 			 */
704 			nip6 = mtod(n, struct ip6_hdr *);
705 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
706 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
707 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
708 			p = (u_char *)(nicmp6 + 1);
709 			bzero(p, 4);
710 			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
711 			noff = sizeof(struct ip6_hdr);
712 			M_COPY_PKTHDR(n, m); /* just for rcvif */
713 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
714 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
715 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
716 			nicmp6->icmp6_code = 0;
717 		}
718 #undef hostnamelen
719 		if (n) {
720 			icmp6stat.icp6s_reflect++;
721 			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
722 			icmp6_reflect(n, noff);
723 		}
724 		break;
725 	    }
726 
727 	case ICMP6_WRUREPLY:
728 		if (code != 0)
729 			goto badcode;
730 		break;
731 
732 	case ND_ROUTER_SOLICIT:
733 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
734 		if (code != 0)
735 			goto badcode;
736 		if (icmp6len < sizeof(struct nd_router_solicit))
737 			goto badlen;
738 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
739 			/* give up local */
740 			nd6_rs_input(m, off, icmp6len);
741 			m = NULL;
742 			goto freeit;
743 		}
744 		nd6_rs_input(n, off, icmp6len);
745 		/* m stays. */
746 		break;
747 
748 	case ND_ROUTER_ADVERT:
749 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
750 		if (code != 0)
751 			goto badcode;
752 		if (icmp6len < sizeof(struct nd_router_advert))
753 			goto badlen;
754 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
755 			/* give up local */
756 			nd6_ra_input(m, off, icmp6len);
757 			m = NULL;
758 			goto freeit;
759 		}
760 		nd6_ra_input(n, off, icmp6len);
761 		/* m stays. */
762 		break;
763 
764 	case ND_NEIGHBOR_SOLICIT:
765 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
766 		if (code != 0)
767 			goto badcode;
768 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
769 			goto badlen;
770 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
771 			/* give up local */
772 			nd6_ns_input(m, off, icmp6len);
773 			m = NULL;
774 			goto freeit;
775 		}
776 		nd6_ns_input(n, off, icmp6len);
777 		/* m stays. */
778 		break;
779 
780 	case ND_NEIGHBOR_ADVERT:
781 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
782 		if (code != 0)
783 			goto badcode;
784 		if (icmp6len < sizeof(struct nd_neighbor_advert))
785 			goto badlen;
786 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
787 			/* give up local */
788 			nd6_na_input(m, off, icmp6len);
789 			m = NULL;
790 			goto freeit;
791 		}
792 		nd6_na_input(n, off, icmp6len);
793 		/* m stays. */
794 		break;
795 
796 	case ND_REDIRECT:
797 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
798 		if (code != 0)
799 			goto badcode;
800 		if (icmp6len < sizeof(struct nd_redirect))
801 			goto badlen;
802 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
803 			/* give up local */
804 			icmp6_redirect_input(m, off);
805 			m = NULL;
806 			goto freeit;
807 		}
808 		icmp6_redirect_input(n, off);
809 		/* m stays. */
810 		break;
811 
812 	case ICMP6_ROUTER_RENUMBERING:
813 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
814 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
815 			goto badcode;
816 		if (icmp6len < sizeof(struct icmp6_router_renum))
817 			goto badlen;
818 		break;
819 
820 	default:
821 		nd6log((LOG_DEBUG,
822 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
823 		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
824 		    ip6_sprintf(&ip6->ip6_dst),
825 		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
826 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
827 			/* ICMPv6 error: MUST deliver it by spec... */
828 			code = PRC_NCMDS;
829 			/* deliver */
830 		} else {
831 			/* ICMPv6 informational: MUST not deliver */
832 			break;
833 		}
834 	deliver:
835 		if (icmp6_notify_error(m, off, icmp6len, code)) {
836 			/* In this case, m should've been freed. */
837 			return(IPPROTO_DONE);
838 		}
839 		break;
840 
841 	badcode:
842 		icmp6stat.icp6s_badcode++;
843 		break;
844 
845 	badlen:
846 		icmp6stat.icp6s_badlen++;
847 		break;
848 	}
849 
850 	/* deliver the packet to appropriate sockets */
851 	icmp6_rip6_input(&m, *offp);
852 
853 	return IPPROTO_DONE;
854 
855  freeit:
856 	m_freem(m);
857 	return IPPROTO_DONE;
858 }
859 
860 static int
861 icmp6_notify_error(m, off, icmp6len, code)
862 	struct mbuf *m;
863 	int off, icmp6len;
864 {
865 	struct icmp6_hdr *icmp6;
866 	struct ip6_hdr *eip6;
867 	u_int32_t notifymtu;
868 	struct sockaddr_in6 icmp6src, icmp6dst;
869 
870 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
871 		icmp6stat.icp6s_tooshort++;
872 		goto freeit;
873 	}
874 #ifndef PULLDOWN_TEST
875 	IP6_EXTHDR_CHECK(m, off,
876 			 sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr),
877 			 -1);
878 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
879 #else
880 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
881 		       sizeof(*icmp6) + sizeof(struct ip6_hdr));
882 	if (icmp6 == NULL) {
883 		icmp6stat.icp6s_tooshort++;
884 		return(-1);
885 	}
886 #endif
887 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
888 
889 	/* Detect the upper level protocol */
890 	{
891 		void (*ctlfunc) __P((int, struct sockaddr *, void *));
892 		u_int8_t nxt = eip6->ip6_nxt;
893 		int eoff = off + sizeof(struct icmp6_hdr) +
894 			sizeof(struct ip6_hdr);
895 		struct ip6ctlparam ip6cp;
896 		struct in6_addr *finaldst = NULL;
897 		int icmp6type = icmp6->icmp6_type;
898 		struct ip6_frag *fh;
899 		struct ip6_rthdr *rth;
900 		struct ip6_rthdr0 *rth0;
901 		int rthlen;
902 
903 		while (1) { /* XXX: should avoid infinite loop explicitly? */
904 			struct ip6_ext *eh;
905 
906 			switch (nxt) {
907 			case IPPROTO_HOPOPTS:
908 			case IPPROTO_DSTOPTS:
909 			case IPPROTO_AH:
910 #ifndef PULLDOWN_TEST
911 				IP6_EXTHDR_CHECK(m, 0, eoff +
912 						 sizeof(struct ip6_ext),
913 						 -1);
914 				eh = (struct ip6_ext *)(mtod(m, caddr_t)
915 							+ eoff);
916 #else
917 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
918 					       eoff, sizeof(*eh));
919 				if (eh == NULL) {
920 					icmp6stat.icp6s_tooshort++;
921 					return(-1);
922 				}
923 #endif
924 
925 				if (nxt == IPPROTO_AH)
926 					eoff += (eh->ip6e_len + 2) << 2;
927 				else
928 					eoff += (eh->ip6e_len + 1) << 3;
929 				nxt = eh->ip6e_nxt;
930 				break;
931 			case IPPROTO_ROUTING:
932 				/*
933 				 * When the erroneous packet contains a
934 				 * routing header, we should examine the
935 				 * header to determine the final destination.
936 				 * Otherwise, we can't properly update
937 				 * information that depends on the final
938 				 * destination (e.g. path MTU).
939 				 */
940 #ifndef PULLDOWN_TEST
941 				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth),
942 						 -1);
943 				rth = (struct ip6_rthdr *)(mtod(m, caddr_t)
944 							   + eoff);
945 #else
946 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
947 					       eoff, sizeof(*rth));
948 				if (rth == NULL) {
949 					icmp6stat.icp6s_tooshort++;
950 					return(-1);
951 				}
952 #endif
953 				rthlen = (rth->ip6r_len + 1) << 3;
954 				/*
955 				 * XXX: currently there is no
956 				 * officially defined type other
957 				 * than type-0.
958 				 * Note that if the segment left field
959 				 * is 0, all intermediate hops must
960 				 * have been passed.
961 				 */
962 				if (rth->ip6r_segleft &&
963 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
964 					int hops;
965 
966 #ifndef PULLDOWN_TEST
967 					IP6_EXTHDR_CHECK(m, 0, eoff + rthlen,
968 							 -1);
969 					rth0 = (struct ip6_rthdr0 *)(mtod(m, caddr_t) + eoff);
970 #else
971 					IP6_EXTHDR_GET(rth0,
972 						       struct ip6_rthdr0 *, m,
973 						       eoff, rthlen);
974 					if (rth0 == NULL) {
975 						icmp6stat.icp6s_tooshort++;
976 						return(-1);
977 					}
978 #endif
979 					/* just ignore a bogus header */
980 					if ((rth0->ip6r0_len % 2) == 0 &&
981 					    (hops = rth0->ip6r0_len/2))
982 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
983 				}
984 				eoff += rthlen;
985 				nxt = rth->ip6r_nxt;
986 				break;
987 			case IPPROTO_FRAGMENT:
988 #ifndef PULLDOWN_TEST
989 				IP6_EXTHDR_CHECK(m, 0, eoff +
990 						 sizeof(struct ip6_frag),
991 						 -1);
992 				fh = (struct ip6_frag *)(mtod(m, caddr_t)
993 							 + eoff);
994 #else
995 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
996 					       eoff, sizeof(*fh));
997 				if (fh == NULL) {
998 					icmp6stat.icp6s_tooshort++;
999 					return(-1);
1000 				}
1001 #endif
1002 				/*
1003 				 * Data after a fragment header is meaningless
1004 				 * unless it is the first fragment, but
1005 				 * we'll go to the notify label for path MTU
1006 				 * discovery.
1007 				 */
1008 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1009 					goto notify;
1010 
1011 				eoff += sizeof(struct ip6_frag);
1012 				nxt = fh->ip6f_nxt;
1013 				break;
1014 			default:
1015 				/*
1016 				 * This case includes ESP and the No Next
1017 				 * Header.  In such cases going to the notify
1018 				 * label does not have any meaning
1019 				 * (i.e. ctlfunc will be NULL), but we go
1020 				 * anyway since we might have to update
1021 				 * path MTU information.
1022 				 */
1023 				goto notify;
1024 			}
1025 		}
1026 	  notify:
1027 #ifndef PULLDOWN_TEST
1028 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1029 #else
1030 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
1031 			       sizeof(*icmp6) + sizeof(struct ip6_hdr));
1032 		if (icmp6 == NULL) {
1033 			icmp6stat.icp6s_tooshort++;
1034 			return(-1);
1035 		}
1036 #endif
1037 
1038 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1039 		bzero(&icmp6dst, sizeof(icmp6dst));
1040 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1041 		icmp6dst.sin6_family = AF_INET6;
1042 		if (finaldst == NULL)
1043 			icmp6dst.sin6_addr = eip6->ip6_dst;
1044 		else
1045 			icmp6dst.sin6_addr = *finaldst;
1046 		icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1047 							  &icmp6dst.sin6_addr);
1048 #ifndef SCOPEDROUTING
1049 		if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1050 				   NULL, NULL)) {
1051 			/* should be impossbile */
1052 			nd6log((LOG_DEBUG,
1053 			    "icmp6_notify_error: in6_embedscope failed\n"));
1054 			goto freeit;
1055 		}
1056 #endif
1057 
1058 		/*
1059 		 * retrieve parameters from the inner IPv6 header, and convert
1060 		 * them into sockaddr structures.
1061 		 */
1062 		bzero(&icmp6src, sizeof(icmp6src));
1063 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1064 		icmp6src.sin6_family = AF_INET6;
1065 		icmp6src.sin6_addr = eip6->ip6_src;
1066 		icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1067 							  &icmp6src.sin6_addr);
1068 #ifndef SCOPEDROUTING
1069 		if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1070 				   NULL, NULL)) {
1071 			/* should be impossbile */
1072 			nd6log((LOG_DEBUG,
1073 			    "icmp6_notify_error: in6_embedscope failed\n"));
1074 			goto freeit;
1075 		}
1076 #endif
1077 		icmp6src.sin6_flowinfo =
1078 			(eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1079 
1080 		if (finaldst == NULL)
1081 			finaldst = &eip6->ip6_dst;
1082 		ip6cp.ip6c_m = m;
1083 		ip6cp.ip6c_icmp6 = icmp6;
1084 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1085 		ip6cp.ip6c_off = eoff;
1086 		ip6cp.ip6c_finaldst = finaldst;
1087 		ip6cp.ip6c_src = &icmp6src;
1088 		ip6cp.ip6c_nxt = nxt;
1089 
1090 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1091 			notifymtu = ntohl(icmp6->icmp6_mtu);
1092 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1093 			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1094 		}
1095 
1096 		ctlfunc = (void (*) __P((int, struct sockaddr *, void *)))
1097 			(inet6sw[ip6_protox[nxt]].pr_ctlinput);
1098 		if (ctlfunc) {
1099 			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1100 					  &ip6cp);
1101 		}
1102 	}
1103 	return(0);
1104 
1105   freeit:
1106 	m_freem(m);
1107 	return(-1);
1108 }
1109 
1110 void
1111 icmp6_mtudisc_update(ip6cp, validated)
1112 	struct ip6ctlparam *ip6cp;
1113 	int validated;
1114 {
1115 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1116 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1117 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1118 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1119 	struct rtentry *rt = NULL;
1120 	struct sockaddr_in6 sin6;
1121 
1122 	if (!validated)
1123 		return;
1124 
1125 	bzero(&sin6, sizeof(sin6));
1126 	sin6.sin6_family = PF_INET6;
1127 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1128 	sin6.sin6_addr = *dst;
1129 	/* XXX normally, this won't happen */
1130 	if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1131 		sin6.sin6_addr.s6_addr16[1] =
1132 		    htons(m->m_pkthdr.rcvif->if_index);
1133 	}
1134 	/* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1135 	rt = rtalloc1((struct sockaddr *)&sin6, 0,
1136 		      RTF_CLONING | RTF_PRCLONING);
1137 
1138 	if (rt && (rt->rt_flags & RTF_HOST)
1139 	    && !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1140 		if (mtu < IPV6_MMTU) {
1141 				/* xxx */
1142 			rt->rt_rmx.rmx_locks |= RTV_MTU;
1143 		} else if (mtu < rt->rt_ifp->if_mtu &&
1144 			   rt->rt_rmx.rmx_mtu > mtu) {
1145 			icmp6stat.icp6s_pmtuchg++;
1146 			rt->rt_rmx.rmx_mtu = mtu;
1147 		}
1148 	}
1149 	if (rt) { /* XXX: need braces to avoid conflict with else in RTFREE. */
1150 		RTFREE(rt);
1151 	}
1152 }
1153 
1154 /*
1155  * Process a Node Information Query packet, based on
1156  * draft-ietf-ipngwg-icmp-name-lookups-07.
1157  *
1158  * Spec incompatibilities:
1159  * - IPv6 Subject address handling
1160  * - IPv4 Subject address handling support missing
1161  * - Proxy reply (answer even if it's not for me)
1162  * - joins NI group address at in6_ifattach() time only, does not cope
1163  *   with hostname changes by sethostname(3)
1164  */
1165 #define hostnamelen	strlen(hostname)
1166 static struct mbuf *
1167 ni6_input(m, off)
1168 	struct mbuf *m;
1169 	int off;
1170 {
1171 	struct icmp6_nodeinfo *ni6, *nni6;
1172 	struct mbuf *n = NULL;
1173 	u_int16_t qtype;
1174 	int subjlen;
1175 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1176 	struct ni_reply_fqdn *fqdn;
1177 	int addrs;		/* for NI_QTYPE_NODEADDR */
1178 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1179 	struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1180 	struct sockaddr_in6 sin6_d; /* XXX: we should retrieve this from m_aux */
1181 	struct ip6_hdr *ip6;
1182 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1183 	char *subj = NULL;
1184 	struct in6_ifaddr *ia6 = NULL;
1185 
1186 	ip6 = mtod(m, struct ip6_hdr *);
1187 #ifndef PULLDOWN_TEST
1188 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1189 #else
1190 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1191 	if (ni6 == NULL) {
1192 		/* m is already reclaimed */
1193 		return NULL;
1194 	}
1195 #endif
1196 
1197 	/*
1198 	 * Validate IPv6 destination address.
1199 	 *
1200 	 * The Responder must discard the Query without further processing
1201 	 * unless it is one of the Responder's unicast or anycast addresses, or
1202 	 * a link-local scope multicast address which the Responder has joined.
1203 	 * [icmp-name-lookups-07, Section 4.]
1204 	 */
1205 	bzero(&sin6, sizeof(sin6));
1206 	sin6.sin6_family = AF_INET6;
1207 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1208 	bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1209 	/* XXX scopeid */
1210 	if ((ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)&sin6)) != NULL) {
1211 		/* unicast/anycast, fine */
1212 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1213 		    (icmp6_nodeinfo & 4) == 0) {
1214 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1215 				"a temporary address in %s:%d",
1216 			       __FILE__, __LINE__));
1217 			goto bad;
1218 		}
1219 	} else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1220 		; /* link-local multicast, fine */
1221 	else
1222 		goto bad;
1223 
1224 	/* validate query Subject field. */
1225 	qtype = ntohs(ni6->ni_qtype);
1226 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1227 	switch (qtype) {
1228 	case NI_QTYPE_NOOP:
1229 	case NI_QTYPE_SUPTYPES:
1230 		/* 07 draft */
1231 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1232 			break;
1233 		/* FALLTHROUGH */
1234 	case NI_QTYPE_FQDN:
1235 	case NI_QTYPE_NODEADDR:
1236 		switch (ni6->ni_code) {
1237 		case ICMP6_NI_SUBJ_IPV6:
1238 #if ICMP6_NI_SUBJ_IPV6 != 0
1239 		case 0:
1240 #endif
1241 			/*
1242 			 * backward compatibility - try to accept 03 draft
1243 			 * format, where no Subject is present.
1244 			 */
1245 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1246 			    subjlen == 0) {
1247 				oldfqdn++;
1248 				break;
1249 			}
1250 #if ICMP6_NI_SUBJ_IPV6 != 0
1251 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1252 				goto bad;
1253 #endif
1254 
1255 			if (subjlen != sizeof(sin6.sin6_addr))
1256 				goto bad;
1257 
1258 			/*
1259 			 * Validate Subject address.
1260 			 *
1261 			 * Not sure what exactly "address belongs to the node"
1262 			 * means in the spec, is it just unicast, or what?
1263 			 *
1264 			 * At this moment we consider Subject address as
1265 			 * "belong to the node" if the Subject address equals
1266 			 * to the IPv6 destination address; validation for
1267 			 * IPv6 destination address should have done enough
1268 			 * check for us.
1269 			 *
1270 			 * We do not do proxy at this moment.
1271 			 */
1272 			/* m_pulldown instead of copy? */
1273 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1274 			    subjlen, (caddr_t)&sin6.sin6_addr);
1275 			sin6.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1276 							      &sin6.sin6_addr);
1277 #ifndef SCOPEDROUTING
1278 			in6_embedscope(&sin6.sin6_addr, &sin6, NULL, NULL);
1279 #endif
1280 			bzero(&sin6_d, sizeof(sin6_d));
1281 			sin6_d.sin6_family = AF_INET6; /* not used, actually */
1282 			sin6_d.sin6_len = sizeof(sin6_d); /* ditto */
1283 			sin6_d.sin6_addr = ip6->ip6_dst;
1284 			sin6_d.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1285 								&ip6->ip6_dst);
1286 #ifndef SCOPEDROUTING
1287 			in6_embedscope(&sin6_d.sin6_addr, &sin6_d, NULL, NULL);
1288 #endif
1289 			subj = (char *)&sin6;
1290 			if (SA6_ARE_ADDR_EQUAL(&sin6, &sin6_d))
1291 				break;
1292 
1293 			/*
1294 			 * XXX if we are to allow other cases, we should really
1295 			 * be careful about scope here.
1296 			 * basically, we should disallow queries toward IPv6
1297 			 * destination X with subject Y, if scope(X) > scope(Y).
1298 			 * if we allow scope(X) > scope(Y), it will result in
1299 			 * information leakage across scope boundary.
1300 			 */
1301 			goto bad;
1302 
1303 		case ICMP6_NI_SUBJ_FQDN:
1304 			/*
1305 			 * Validate Subject name with gethostname(3).
1306 			 *
1307 			 * The behavior may need some debate, since:
1308 			 * - we are not sure if the node has FQDN as
1309 			 *   hostname (returned by gethostname(3)).
1310 			 * - the code does wildcard match for truncated names.
1311 			 *   however, we are not sure if we want to perform
1312 			 *   wildcard match, if gethostname(3) side has
1313 			 *   truncated hostname.
1314 			 */
1315 			n = ni6_nametodns(hostname, hostnamelen, 0);
1316 			if (!n || n->m_next || n->m_len == 0)
1317 				goto bad;
1318 			IP6_EXTHDR_GET(subj, char *, m,
1319 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1320 			if (subj == NULL)
1321 				goto bad;
1322 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1323 					n->m_len)) {
1324 				goto bad;
1325 			}
1326 			m_freem(n);
1327 			n = NULL;
1328 			break;
1329 
1330 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1331 		default:
1332 			goto bad;
1333 		}
1334 		break;
1335 	}
1336 
1337 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1338 	switch (qtype) {
1339 	case NI_QTYPE_FQDN:
1340 		if ((icmp6_nodeinfo & 1) == 0)
1341 			goto bad;
1342 		break;
1343 	case NI_QTYPE_NODEADDR:
1344 		if ((icmp6_nodeinfo & 2) == 0)
1345 			goto bad;
1346 		break;
1347 	}
1348 
1349 	/* guess reply length */
1350 	switch (qtype) {
1351 	case NI_QTYPE_NOOP:
1352 		break;		/* no reply data */
1353 	case NI_QTYPE_SUPTYPES:
1354 		replylen += sizeof(u_int32_t);
1355 		break;
1356 	case NI_QTYPE_FQDN:
1357 		/* XXX will append an mbuf */
1358 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1359 		break;
1360 	case NI_QTYPE_NODEADDR:
1361 		addrs = ni6_addrs(ni6, m, &ifp, subj);
1362 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1363 					  sizeof(u_int32_t))) > MCLBYTES)
1364 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1365 		break;
1366 	default:
1367 		/*
1368 		 * XXX: We must return a reply with the ICMP6 code
1369 		 * `unknown Qtype' in this case. However we regard the case
1370 		 * as an FQDN query for backward compatibility.
1371 		 * Older versions set a random value to this field,
1372 		 * so it rarely varies in the defined qtypes.
1373 		 * But the mechanism is not reliable...
1374 		 * maybe we should obsolete older versions.
1375 		 */
1376 		qtype = NI_QTYPE_FQDN;
1377 		/* XXX will append an mbuf */
1378 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1379 		oldfqdn++;
1380 		break;
1381 	}
1382 
1383 	/* allocate an mbuf to reply. */
1384 	MGETHDR(n, M_DONTWAIT, m->m_type);
1385 	if (n == NULL) {
1386 		m_freem(m);
1387 		return(NULL);
1388 	}
1389 	M_COPY_PKTHDR(n, m); /* just for recvif */
1390 	if (replylen > MHLEN) {
1391 		if (replylen > MCLBYTES) {
1392 			/*
1393 			 * XXX: should we try to allocate more? But MCLBYTES
1394 			 * is probably much larger than IPV6_MMTU...
1395 			 */
1396 			goto bad;
1397 		}
1398 		MCLGET(n, M_DONTWAIT);
1399 		if ((n->m_flags & M_EXT) == 0) {
1400 			goto bad;
1401 		}
1402 	}
1403 	n->m_pkthdr.len = n->m_len = replylen;
1404 
1405 	/* copy mbuf header and IPv6 + Node Information base headers */
1406 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1407 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1408 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1409 
1410 	/* qtype dependent procedure */
1411 	switch (qtype) {
1412 	case NI_QTYPE_NOOP:
1413 		nni6->ni_code = ICMP6_NI_SUCCESS;
1414 		nni6->ni_flags = 0;
1415 		break;
1416 	case NI_QTYPE_SUPTYPES:
1417 	{
1418 		u_int32_t v;
1419 		nni6->ni_code = ICMP6_NI_SUCCESS;
1420 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1421 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1422 		v = (u_int32_t)htonl(0x0000000f);
1423 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1424 		break;
1425 	}
1426 	case NI_QTYPE_FQDN:
1427 		nni6->ni_code = ICMP6_NI_SUCCESS;
1428 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1429 						sizeof(struct ip6_hdr) +
1430 						sizeof(struct icmp6_nodeinfo));
1431 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1432 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1433 		/*
1434 		 * XXX do we really have FQDN in variable "hostname"?
1435 		 */
1436 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1437 		if (n->m_next == NULL)
1438 			goto bad;
1439 		/* XXX we assume that n->m_next is not a chain */
1440 		if (n->m_next->m_next != NULL)
1441 			goto bad;
1442 		n->m_pkthdr.len += n->m_next->m_len;
1443 		break;
1444 	case NI_QTYPE_NODEADDR:
1445 	{
1446 		int lenlim, copied;
1447 
1448 		nni6->ni_code = ICMP6_NI_SUCCESS;
1449 		n->m_pkthdr.len = n->m_len =
1450 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1451 		lenlim = M_TRAILINGSPACE(n);
1452 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1453 		/* XXX: reset mbuf length */
1454 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1455 			sizeof(struct icmp6_nodeinfo) + copied;
1456 		break;
1457 	}
1458 	default:
1459 		break;		/* XXX impossible! */
1460 	}
1461 
1462 	nni6->ni_type = ICMP6_NI_REPLY;
1463 	m_freem(m);
1464 	return(n);
1465 
1466   bad:
1467 	m_freem(m);
1468 	if (n)
1469 		m_freem(n);
1470 	return(NULL);
1471 }
1472 #undef hostnamelen
1473 
1474 /*
1475  * make a mbuf with DNS-encoded string.  no compression support.
1476  *
1477  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1478  * treated as truncated name (two \0 at the end).  this is a wild guess.
1479  */
1480 static struct mbuf *
1481 ni6_nametodns(name, namelen, old)
1482 	const char *name;
1483 	int namelen;
1484 	int old;	/* return pascal string if non-zero */
1485 {
1486 	struct mbuf *m;
1487 	char *cp, *ep;
1488 	const char *p, *q;
1489 	int i, len, nterm;
1490 
1491 	if (old)
1492 		len = namelen + 1;
1493 	else
1494 		len = MCLBYTES;
1495 
1496 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1497 	MGET(m, M_DONTWAIT, MT_DATA);
1498 	if (m && len > MLEN) {
1499 		MCLGET(m, M_DONTWAIT);
1500 		if ((m->m_flags & M_EXT) == 0)
1501 			goto fail;
1502 	}
1503 	if (!m)
1504 		goto fail;
1505 	m->m_next = NULL;
1506 
1507 	if (old) {
1508 		m->m_len = len;
1509 		*mtod(m, char *) = namelen;
1510 		bcopy(name, mtod(m, char *) + 1, namelen);
1511 		return m;
1512 	} else {
1513 		m->m_len = 0;
1514 		cp = mtod(m, char *);
1515 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1516 
1517 		/* if not certain about my name, return empty buffer */
1518 		if (namelen == 0)
1519 			return m;
1520 
1521 		/*
1522 		 * guess if it looks like shortened hostname, or FQDN.
1523 		 * shortened hostname needs two trailing "\0".
1524 		 */
1525 		i = 0;
1526 		for (p = name; p < name + namelen; p++) {
1527 			if (*p && *p == '.')
1528 				i++;
1529 		}
1530 		if (i < 2)
1531 			nterm = 2;
1532 		else
1533 			nterm = 1;
1534 
1535 		p = name;
1536 		while (cp < ep && p < name + namelen) {
1537 			i = 0;
1538 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1539 				i++;
1540 			/* result does not fit into mbuf */
1541 			if (cp + i + 1 >= ep)
1542 				goto fail;
1543 			/*
1544 			 * DNS label length restriction, RFC1035 page 8.
1545 			 * "i == 0" case is included here to avoid returning
1546 			 * 0-length label on "foo..bar".
1547 			 */
1548 			if (i <= 0 || i >= 64)
1549 				goto fail;
1550 			*cp++ = i;
1551 			bcopy(p, cp, i);
1552 			cp += i;
1553 			p = q;
1554 			if (p < name + namelen && *p == '.')
1555 				p++;
1556 		}
1557 		/* termination */
1558 		if (cp + nterm >= ep)
1559 			goto fail;
1560 		while (nterm-- > 0)
1561 			*cp++ = '\0';
1562 		m->m_len = cp - mtod(m, char *);
1563 		return m;
1564 	}
1565 
1566 	panic("should not reach here");
1567 	/* NOTREACHED */
1568 
1569  fail:
1570 	if (m)
1571 		m_freem(m);
1572 	return NULL;
1573 }
1574 
1575 /*
1576  * check if two DNS-encoded string matches.  takes care of truncated
1577  * form (with \0\0 at the end).  no compression support.
1578  * XXX upper/lowercase match (see RFC2065)
1579  */
1580 static int
1581 ni6_dnsmatch(a, alen, b, blen)
1582 	const char *a;
1583 	int alen;
1584 	const char *b;
1585 	int blen;
1586 {
1587 	const char *a0, *b0;
1588 	int l;
1589 
1590 	/* simplest case - need validation? */
1591 	if (alen == blen && bcmp(a, b, alen) == 0)
1592 		return 1;
1593 
1594 	a0 = a;
1595 	b0 = b;
1596 
1597 	/* termination is mandatory */
1598 	if (alen < 2 || blen < 2)
1599 		return 0;
1600 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1601 		return 0;
1602 	alen--;
1603 	blen--;
1604 
1605 	while (a - a0 < alen && b - b0 < blen) {
1606 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1607 			return 0;
1608 
1609 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1610 			return 0;
1611 		/* we don't support compression yet */
1612 		if (a[0] >= 64 || b[0] >= 64)
1613 			return 0;
1614 
1615 		/* truncated case */
1616 		if (a[0] == 0 && a - a0 == alen - 1)
1617 			return 1;
1618 		if (b[0] == 0 && b - b0 == blen - 1)
1619 			return 1;
1620 		if (a[0] == 0 || b[0] == 0)
1621 			return 0;
1622 
1623 		if (a[0] != b[0])
1624 			return 0;
1625 		l = a[0];
1626 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1627 			return 0;
1628 		if (bcmp(a + 1, b + 1, l) != 0)
1629 			return 0;
1630 
1631 		a += 1 + l;
1632 		b += 1 + l;
1633 	}
1634 
1635 	if (a - a0 == alen && b - b0 == blen)
1636 		return 1;
1637 	else
1638 		return 0;
1639 }
1640 
1641 /*
1642  * calculate the number of addresses to be returned in the node info reply.
1643  */
1644 static int
1645 ni6_addrs(ni6, m, ifpp, subj)
1646 	struct icmp6_nodeinfo *ni6;
1647 	struct mbuf *m;
1648 	struct ifnet **ifpp;
1649 	char *subj;
1650 {
1651 	struct ifnet *ifp;
1652 	struct in6_ifaddr *ifa6;
1653 	struct ifaddr *ifa;
1654 	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1655 	int addrs = 0, addrsofif, iffound = 0;
1656 	int niflags = ni6->ni_flags;
1657 
1658 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1659 		switch (ni6->ni_code) {
1660 		case ICMP6_NI_SUBJ_IPV6:
1661 			if (subj == NULL) /* must be impossible... */
1662 				return(0);
1663 			subj_ip6 = (struct sockaddr_in6 *)subj;
1664 			break;
1665 		default:
1666 			/*
1667 			 * XXX: we only support IPv6 subject address for
1668 			 * this Qtype.
1669 			 */
1670 			return(0);
1671 		}
1672 	}
1673 
1674 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
1675 	{
1676 		addrsofif = 0;
1677 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list)
1678 		{
1679 			if (ifa->ifa_addr->sa_family != AF_INET6)
1680 				continue;
1681 			ifa6 = (struct in6_ifaddr *)ifa;
1682 
1683 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1684 			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1685 					       &ifa6->ia_addr.sin6_addr))
1686 				iffound = 1;
1687 
1688 			/*
1689 			 * IPv4-mapped addresses can only be returned by a
1690 			 * Node Information proxy, since they represent
1691 			 * addresses of IPv4-only nodes, which perforce do
1692 			 * not implement this protocol.
1693 			 * [icmp-name-lookups-07, Section 5.4]
1694 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1695 			 * this function at this moment.
1696 			 */
1697 
1698 			/* What do we have to do about ::1? */
1699 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1700 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1701 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1702 					continue;
1703 				break;
1704 			case IPV6_ADDR_SCOPE_SITELOCAL:
1705 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1706 					continue;
1707 				break;
1708 			case IPV6_ADDR_SCOPE_GLOBAL:
1709 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1710 					continue;
1711 				break;
1712 			default:
1713 				continue;
1714 			}
1715 
1716 			/*
1717 			 * check if anycast is okay.
1718 			 * XXX: just experimental.  not in the spec.
1719 			 */
1720 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1721 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1722 				continue; /* we need only unicast addresses */
1723 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1724 			    (icmp6_nodeinfo & 4) == 0) {
1725 				continue;
1726 			}
1727 			addrsofif++; /* count the address */
1728 		}
1729 		if (iffound) {
1730 			*ifpp = ifp;
1731 			return(addrsofif);
1732 		}
1733 
1734 		addrs += addrsofif;
1735 	}
1736 
1737 	return(addrs);
1738 }
1739 
1740 static int
1741 ni6_store_addrs(ni6, nni6, ifp0, resid)
1742 	struct icmp6_nodeinfo *ni6, *nni6;
1743 	struct ifnet *ifp0;
1744 	int resid;
1745 {
1746 	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1747 	struct in6_ifaddr *ifa6;
1748 	struct ifaddr *ifa;
1749 	struct ifnet *ifp_dep = NULL;
1750 	int copied = 0, allow_deprecated = 0;
1751 	u_char *cp = (u_char *)(nni6 + 1);
1752 	int niflags = ni6->ni_flags;
1753 	u_int32_t ltime;
1754 
1755 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1756 		return(0);	/* needless to copy */
1757 
1758   again:
1759 
1760 	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list))
1761 	{
1762 		for (ifa = ifp->if_addrlist.tqh_first; ifa;
1763 		     ifa = ifa->ifa_list.tqe_next)
1764 		{
1765 			if (ifa->ifa_addr->sa_family != AF_INET6)
1766 				continue;
1767 			ifa6 = (struct in6_ifaddr *)ifa;
1768 
1769 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1770 			    allow_deprecated == 0) {
1771 				/*
1772 				 * prefererred address should be put before
1773 				 * deprecated addresses.
1774 				 */
1775 
1776 				/* record the interface for later search */
1777 				if (ifp_dep == NULL)
1778 					ifp_dep = ifp;
1779 
1780 				continue;
1781 			}
1782 			else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1783 				 allow_deprecated != 0)
1784 				continue; /* we now collect deprecated addrs */
1785 
1786 			/* What do we have to do about ::1? */
1787 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1788 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1789 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1790 					continue;
1791 				break;
1792 			case IPV6_ADDR_SCOPE_SITELOCAL:
1793 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1794 					continue;
1795 				break;
1796 			case IPV6_ADDR_SCOPE_GLOBAL:
1797 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1798 					continue;
1799 				break;
1800 			default:
1801 				continue;
1802 			}
1803 
1804 			/*
1805 			 * check if anycast is okay.
1806 			 * XXX: just experimental. not in the spec.
1807 			 */
1808 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1809 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1810 				continue;
1811 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1812 			    (icmp6_nodeinfo & 4) == 0) {
1813 				continue;
1814 			}
1815 
1816 			/* now we can copy the address */
1817 			if (resid < sizeof(struct in6_addr) +
1818 			    sizeof(u_int32_t)) {
1819 				/*
1820 				 * We give up much more copy.
1821 				 * Set the truncate flag and return.
1822 				 */
1823 				nni6->ni_flags |=
1824 					NI_NODEADDR_FLAG_TRUNCATE;
1825 				return(copied);
1826 			}
1827 
1828 			/*
1829 			 * Set the TTL of the address.
1830 			 * The TTL value should be one of the following
1831 			 * according to the specification:
1832 			 *
1833 			 * 1. The remaining lifetime of a DHCP lease on the
1834 			 *    address, or
1835 			 * 2. The remaining Valid Lifetime of a prefix from
1836 			 *    which the address was derived through Stateless
1837 			 *    Autoconfiguration.
1838 			 *
1839 			 * Note that we currently do not support stateful
1840 			 * address configuration by DHCPv6, so the former
1841 			 * case can't happen.
1842 			 */
1843 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1844 				ltime = ND6_INFINITE_LIFETIME;
1845 			else {
1846 				if (ifa6->ia6_lifetime.ia6t_expire >
1847 				    time_second)
1848 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1849 				else
1850 					ltime = 0;
1851 			}
1852 
1853 			bcopy(&ltime, cp, sizeof(u_int32_t));
1854 			cp += sizeof(u_int32_t);
1855 
1856 			/* copy the address itself */
1857 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1858 			      sizeof(struct in6_addr));
1859 			/* XXX: KAME link-local hack; remove ifindex */
1860 			if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr))
1861 				((struct in6_addr *)cp)->s6_addr16[1] = 0;
1862 			cp += sizeof(struct in6_addr);
1863 
1864 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1865 			copied += (sizeof(struct in6_addr) +
1866 				   sizeof(u_int32_t));
1867 		}
1868 		if (ifp0)	/* we need search only on the specified IF */
1869 			break;
1870 	}
1871 
1872 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1873 		ifp = ifp_dep;
1874 		allow_deprecated = 1;
1875 
1876 		goto again;
1877 	}
1878 
1879 	return(copied);
1880 }
1881 
1882 /*
1883  * XXX almost dup'ed code with rip6_input.
1884  */
1885 static int
1886 icmp6_rip6_input(mp, off)
1887 	struct	mbuf **mp;
1888 	int	off;
1889 {
1890 	struct mbuf *m = *mp;
1891 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1892 	struct in6pcb *in6p;
1893 	struct in6pcb *last = NULL;
1894 	struct sockaddr_in6 rip6src;
1895 	struct icmp6_hdr *icmp6;
1896 	struct mbuf *opts = NULL;
1897 
1898 #ifndef PULLDOWN_TEST
1899 	/* this is assumed to be safe. */
1900 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1901 #else
1902 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1903 	if (icmp6 == NULL) {
1904 		/* m is already reclaimed */
1905 		return IPPROTO_DONE;
1906 	}
1907 #endif
1908 
1909 	bzero(&rip6src, sizeof(rip6src));
1910 	rip6src.sin6_len = sizeof(struct sockaddr_in6);
1911 	rip6src.sin6_family = AF_INET6;
1912 	/* KAME hack: recover scopeid */
1913 	(void)in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif);
1914 
1915 	LIST_FOREACH(in6p, &ripcb, inp_list)
1916 	{
1917 		if ((in6p->inp_vflag & INP_IPV6) == 0)
1918 			continue;
1919 #ifdef HAVE_NRL_INPCB
1920 		if (!(in6p->in6p_flags & INP_IPV6))
1921 			continue;
1922 #endif
1923 		if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1924 			continue;
1925 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1926 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1927 			continue;
1928 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1929 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1930 			continue;
1931 		if (in6p->in6p_icmp6filt
1932 		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1933 				 in6p->in6p_icmp6filt))
1934 			continue;
1935 		if (last) {
1936 			struct	mbuf *n;
1937 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1938 				if (last->in6p_flags & IN6P_CONTROLOPTS)
1939 					ip6_savecontrol(last, &opts, ip6, n);
1940 				/* strip intermediate headers */
1941 				m_adj(n, off);
1942 				if (sbappendaddr(&last->in6p_socket->so_rcv,
1943 						 (struct sockaddr *)&rip6src,
1944 						 n, opts) == 0) {
1945 					/* should notify about lost packet */
1946 					m_freem(n);
1947 					if (opts) {
1948 						m_freem(opts);
1949 					}
1950 				} else
1951 					sorwakeup(last->in6p_socket);
1952 				opts = NULL;
1953 			}
1954 		}
1955 		last = in6p;
1956 	}
1957 	if (last) {
1958 		if (last->in6p_flags & IN6P_CONTROLOPTS)
1959 			ip6_savecontrol(last, &opts, ip6, m);
1960 		/* strip intermediate headers */
1961 		m_adj(m, off);
1962 		if (sbappendaddr(&last->in6p_socket->so_rcv,
1963 				 (struct sockaddr *)&rip6src, m, opts) == 0) {
1964 			m_freem(m);
1965 			if (opts)
1966 				m_freem(opts);
1967 		} else
1968 			sorwakeup(last->in6p_socket);
1969 	} else {
1970 		m_freem(m);
1971 		ip6stat.ip6s_delivered--;
1972 	}
1973 	return IPPROTO_DONE;
1974 }
1975 
1976 /*
1977  * Reflect the ip6 packet back to the source.
1978  * OFF points to the icmp6 header, counted from the top of the mbuf.
1979  */
1980 void
1981 icmp6_reflect(m, off)
1982 	struct	mbuf *m;
1983 	size_t off;
1984 {
1985 	struct ip6_hdr *ip6;
1986 	struct icmp6_hdr *icmp6;
1987 	struct in6_ifaddr *ia;
1988 	struct in6_addr t, *src = 0;
1989 	int plen;
1990 	int type, code;
1991 	struct ifnet *outif = NULL;
1992 	struct sockaddr_in6 sa6_src, sa6_dst;
1993 #ifdef COMPAT_RFC1885
1994 	int mtu = IPV6_MMTU;
1995 	struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
1996 #endif
1997 
1998 	/* too short to reflect */
1999 	if (off < sizeof(struct ip6_hdr)) {
2000 		nd6log((LOG_DEBUG,
2001 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2002 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2003 		    __FILE__, __LINE__));
2004 		goto bad;
2005 	}
2006 
2007 	/*
2008 	 * If there are extra headers between IPv6 and ICMPv6, strip
2009 	 * off that header first.
2010 	 */
2011 #ifdef DIAGNOSTIC
2012 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2013 		panic("assumption failed in icmp6_reflect");
2014 #endif
2015 	if (off > sizeof(struct ip6_hdr)) {
2016 		size_t l;
2017 		struct ip6_hdr nip6;
2018 
2019 		l = off - sizeof(struct ip6_hdr);
2020 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2021 		m_adj(m, l);
2022 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2023 		if (m->m_len < l) {
2024 			if ((m = m_pullup(m, l)) == NULL)
2025 				return;
2026 		}
2027 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2028 	} else /* off == sizeof(struct ip6_hdr) */ {
2029 		size_t l;
2030 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2031 		if (m->m_len < l) {
2032 			if ((m = m_pullup(m, l)) == NULL)
2033 				return;
2034 		}
2035 	}
2036 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2037 	ip6 = mtod(m, struct ip6_hdr *);
2038 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2039 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2040 	type = icmp6->icmp6_type; /* keep type for statistics */
2041 	code = icmp6->icmp6_code; /* ditto. */
2042 
2043 	t = ip6->ip6_dst;
2044 	/*
2045 	 * ip6_input() drops a packet if its src is multicast.
2046 	 * So, the src is never multicast.
2047 	 */
2048 	ip6->ip6_dst = ip6->ip6_src;
2049 
2050 	/*
2051 	 * XXX: make sure to embed scope zone information, using
2052 	 * already embedded IDs or the received interface (if any).
2053 	 * Note that rcvif may be NULL.
2054 	 * TODO: scoped routing case (XXX).
2055 	 */
2056 	bzero(&sa6_src, sizeof(sa6_src));
2057 	sa6_src.sin6_family = AF_INET6;
2058 	sa6_src.sin6_len = sizeof(sa6_src);
2059 	sa6_src.sin6_addr = ip6->ip6_dst;
2060 	in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
2061 	in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL);
2062 	bzero(&sa6_dst, sizeof(sa6_dst));
2063 	sa6_dst.sin6_family = AF_INET6;
2064 	sa6_dst.sin6_len = sizeof(sa6_dst);
2065 	sa6_dst.sin6_addr = t;
2066 	in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
2067 	in6_embedscope(&t, &sa6_dst, NULL, NULL);
2068 
2069 #ifdef COMPAT_RFC1885
2070 	/*
2071 	 * xxx guess MTU
2072 	 * RFC 1885 requires that echo reply should be truncated if it
2073 	 * does not fit in with (return) path MTU, but the description was
2074 	 * removed in the new spec.
2075 	 */
2076 	if (icmp6_reflect_rt.ro_rt == 0 ||
2077 	    ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
2078 		if (icmp6_reflect_rt.ro_rt) {
2079 			RTFREE(icmp6_reflect_rt.ro_rt);
2080 			icmp6_reflect_rt.ro_rt = 0;
2081 		}
2082 		bzero(sin6, sizeof(*sin6));
2083 		sin6->sin6_family = PF_INET6;
2084 		sin6->sin6_len = sizeof(struct sockaddr_in6);
2085 		sin6->sin6_addr = ip6->ip6_dst;
2086 
2087 		rtalloc_ign((struct route *)&icmp6_reflect_rt.ro_rt,
2088 			    RTF_PRCLONING);
2089 	}
2090 
2091 	if (icmp6_reflect_rt.ro_rt == 0)
2092 		goto bad;
2093 
2094 	if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
2095 	    && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
2096 		mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
2097 
2098 	if (mtu < m->m_pkthdr.len) {
2099 		plen -= (m->m_pkthdr.len - mtu);
2100 		m_adj(m, mtu - m->m_pkthdr.len);
2101 	}
2102 #endif
2103 	/*
2104 	 * If the incoming packet was addressed directly to us(i.e. unicast),
2105 	 * use dst as the src for the reply.
2106 	 * The IN6_IFF_NOTREADY case would be VERY rare, but is possible
2107 	 * (for example) when we encounter an error while forwarding procedure
2108 	 * destined to a duplicated address of ours.
2109 	 */
2110 	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2111 		if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2112 		    (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2113 			src = &t;
2114 			break;
2115 		}
2116 	if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2117 		/*
2118 		 * This is the case if the dst is our link-local address
2119 		 * and the sender is also ourselves.
2120 		 */
2121 		src = &t;
2122 	}
2123 
2124 	if (src == 0) {
2125 		int e;
2126 		struct route_in6 ro;
2127 
2128 		/*
2129 		 * This case matches to multicasts, our anycast, or unicasts
2130 		 * that we do not own.  Select a source address based on the
2131 		 * source address of the erroneous packet.
2132 		 */
2133 		bzero(&ro, sizeof(ro));
2134 		src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e);
2135 		if (ro.ro_rt)
2136 			RTFREE(ro.ro_rt); /* XXX: we could use this */
2137 		if (src == NULL) {
2138 			nd6log((LOG_DEBUG,
2139 			    "icmp6_reflect: source can't be determined: "
2140 			    "dst=%s, error=%d\n",
2141 			    ip6_sprintf(&sa6_src.sin6_addr), e));
2142 			goto bad;
2143 		}
2144 	}
2145 
2146 	ip6->ip6_src = *src;
2147 
2148 	ip6->ip6_flow = 0;
2149 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2150 	ip6->ip6_vfc |= IPV6_VERSION;
2151 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2152 	if (m->m_pkthdr.rcvif) {
2153 		/* XXX: This may not be the outgoing interface */
2154 		ip6->ip6_hlim = nd_ifinfo[m->m_pkthdr.rcvif->if_index].chlim;
2155 	} else
2156 		ip6->ip6_hlim = ip6_defhlim;
2157 
2158 	icmp6->icmp6_cksum = 0;
2159 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2160 					sizeof(struct ip6_hdr), plen);
2161 
2162 	/*
2163 	 * XXX option handling
2164 	 */
2165 
2166 	m->m_flags &= ~(M_BCAST|M_MCAST);
2167 
2168 #ifdef COMPAT_RFC1885
2169 	ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif, NULL);
2170 #else
2171 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2172 #endif
2173 	if (outif)
2174 		icmp6_ifoutstat_inc(outif, type, code);
2175 
2176 	return;
2177 
2178  bad:
2179 	m_freem(m);
2180 	return;
2181 }
2182 
2183 void
2184 icmp6_fasttimo()
2185 {
2186 
2187 	mld6_fasttimeo();
2188 }
2189 
2190 static const char *
2191 icmp6_redirect_diag(src6, dst6, tgt6)
2192 	struct in6_addr *src6;
2193 	struct in6_addr *dst6;
2194 	struct in6_addr *tgt6;
2195 {
2196 	static char buf[1024];
2197 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2198 		ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2199 	return buf;
2200 }
2201 
2202 void
2203 icmp6_redirect_input(m, off)
2204 	struct mbuf *m;
2205 	int off;
2206 {
2207 	struct ifnet *ifp = m->m_pkthdr.rcvif;
2208 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2209 	struct nd_redirect *nd_rd;
2210 	int icmp6len = ntohs(ip6->ip6_plen);
2211 	char *lladdr = NULL;
2212 	int lladdrlen = 0;
2213 	u_char *redirhdr = NULL;
2214 	int redirhdrlen = 0;
2215 	struct rtentry *rt = NULL;
2216 	int is_router;
2217 	int is_onlink;
2218 	struct in6_addr src6 = ip6->ip6_src;
2219 	struct in6_addr redtgt6;
2220 	struct in6_addr reddst6;
2221 	union nd_opts ndopts;
2222 
2223 	if (!m || !ifp)
2224 		return;
2225 
2226 	/* XXX if we are router, we don't update route by icmp6 redirect */
2227 	if (ip6_forwarding)
2228 		goto freeit;
2229 	if (!icmp6_rediraccept)
2230 		goto freeit;
2231 
2232 #ifndef PULLDOWN_TEST
2233 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2234 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2235 #else
2236 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2237 	if (nd_rd == NULL) {
2238 		icmp6stat.icp6s_tooshort++;
2239 		return;
2240 	}
2241 #endif
2242 	redtgt6 = nd_rd->nd_rd_target;
2243 	reddst6 = nd_rd->nd_rd_dst;
2244 
2245 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2246 		redtgt6.s6_addr16[1] = htons(ifp->if_index);
2247 	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2248 		reddst6.s6_addr16[1] = htons(ifp->if_index);
2249 
2250 	/* validation */
2251 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2252 		nd6log((LOG_ERR,
2253 			"ICMP6 redirect sent from %s rejected; "
2254 			"must be from linklocal\n", ip6_sprintf(&src6)));
2255 		goto bad;
2256 	}
2257 	if (ip6->ip6_hlim != 255) {
2258 		nd6log((LOG_ERR,
2259 			"ICMP6 redirect sent from %s rejected; "
2260 			"hlim=%d (must be 255)\n",
2261 			ip6_sprintf(&src6), ip6->ip6_hlim));
2262 		goto bad;
2263 	}
2264     {
2265 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2266 	struct sockaddr_in6 sin6;
2267 	struct in6_addr *gw6;
2268 
2269 	bzero(&sin6, sizeof(sin6));
2270 	sin6.sin6_family = AF_INET6;
2271 	sin6.sin6_len = sizeof(struct sockaddr_in6);
2272 	bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
2273 	rt = rtalloc1((struct sockaddr *)&sin6, 0, 0UL);
2274 	if (rt) {
2275 		if (rt->rt_gateway == NULL ||
2276 		    rt->rt_gateway->sa_family != AF_INET6) {
2277 			nd6log((LOG_ERR,
2278 			    "ICMP6 redirect rejected; no route "
2279 			    "with inet6 gateway found for redirect dst: %s\n",
2280 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2281 			RTFREE(rt);
2282 			goto bad;
2283 		}
2284 
2285 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2286 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2287 			nd6log((LOG_ERR,
2288 				"ICMP6 redirect rejected; "
2289 				"not equal to gw-for-src=%s (must be same): "
2290 				"%s\n",
2291 				ip6_sprintf(gw6),
2292 				icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2293 			RTFREE(rt);
2294 			goto bad;
2295 		}
2296 	} else {
2297 		nd6log((LOG_ERR,
2298 			"ICMP6 redirect rejected; "
2299 			"no route found for redirect dst: %s\n",
2300 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2301 		goto bad;
2302 	}
2303 	RTFREE(rt);
2304 	rt = NULL;
2305     }
2306 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2307 		nd6log((LOG_ERR,
2308 			"ICMP6 redirect rejected; "
2309 			"redirect dst must be unicast: %s\n",
2310 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2311 		goto bad;
2312 	}
2313 
2314 	is_router = is_onlink = 0;
2315 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2316 		is_router = 1;	/* router case */
2317 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2318 		is_onlink = 1;	/* on-link destination case */
2319 	if (!is_router && !is_onlink) {
2320 		nd6log((LOG_ERR,
2321 			"ICMP6 redirect rejected; "
2322 			"neither router case nor onlink case: %s\n",
2323 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2324 		goto bad;
2325 	}
2326 	/* validation passed */
2327 
2328 	icmp6len -= sizeof(*nd_rd);
2329 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2330 	if (nd6_options(&ndopts) < 0) {
2331 		nd6log((LOG_INFO, "icmp6_redirect_input: "
2332 			"invalid ND option, rejected: %s\n",
2333 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2334 		/* nd6_options have incremented stats */
2335 		goto freeit;
2336 	}
2337 
2338 	if (ndopts.nd_opts_tgt_lladdr) {
2339 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2340 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2341 	}
2342 
2343 	if (ndopts.nd_opts_rh) {
2344 		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2345 		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2346 	}
2347 
2348 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2349 		nd6log((LOG_INFO,
2350 			"icmp6_redirect_input: lladdrlen mismatch for %s "
2351 			"(if %d, icmp6 packet %d): %s\n",
2352 			ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2353 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2354 		goto bad;
2355 	}
2356 
2357 	/* RFC 2461 8.3 */
2358 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2359 			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2360 
2361 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2362 		/* perform rtredirect */
2363 		struct sockaddr_in6 sdst;
2364 		struct sockaddr_in6 sgw;
2365 		struct sockaddr_in6 ssrc;
2366 
2367 		bzero(&sdst, sizeof(sdst));
2368 		bzero(&sgw, sizeof(sgw));
2369 		bzero(&ssrc, sizeof(ssrc));
2370 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2371 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2372 			sizeof(struct sockaddr_in6);
2373 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2374 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2375 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2376 		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2377 			   (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
2378 			   (struct sockaddr *)&ssrc,
2379 			   (struct rtentry **)NULL);
2380 	}
2381 	/* finally update cached route in each socket via pfctlinput */
2382     {
2383 	struct sockaddr_in6 sdst;
2384 
2385 	bzero(&sdst, sizeof(sdst));
2386 	sdst.sin6_family = AF_INET6;
2387 	sdst.sin6_len = sizeof(struct sockaddr_in6);
2388 	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2389 	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2390 #ifdef IPSEC
2391 	key_sa_routechange((struct sockaddr *)&sdst);
2392 #endif
2393     }
2394 
2395  freeit:
2396 	m_freem(m);
2397 	return;
2398 
2399  bad:
2400 	icmp6stat.icp6s_badredirect++;
2401 	m_freem(m);
2402 }
2403 
2404 void
2405 icmp6_redirect_output(m0, rt)
2406 	struct mbuf *m0;
2407 	struct rtentry *rt;
2408 {
2409 	struct ifnet *ifp;	/* my outgoing interface */
2410 	struct in6_addr *ifp_ll6;
2411 	struct in6_addr *router_ll6;
2412 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2413 	struct mbuf *m = NULL;	/* newly allocated one */
2414 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2415 	struct nd_redirect *nd_rd;
2416 	size_t maxlen;
2417 	u_char *p;
2418 	struct ifnet *outif = NULL;
2419 	struct sockaddr_in6 src_sa;
2420 
2421 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2422 
2423 	/* if we are not router, we don't send icmp6 redirect */
2424 	if (!ip6_forwarding || ip6_accept_rtadv)
2425 		goto fail;
2426 
2427 	/* sanity check */
2428 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2429 		goto fail;
2430 
2431 	/*
2432 	 * Address check:
2433 	 *  the source address must identify a neighbor, and
2434 	 *  the destination address must not be a multicast address
2435 	 *  [RFC 2461, sec 8.2]
2436 	 */
2437 	sip6 = mtod(m0, struct ip6_hdr *);
2438 	bzero(&src_sa, sizeof(src_sa));
2439 	src_sa.sin6_family = AF_INET6;
2440 	src_sa.sin6_len = sizeof(src_sa);
2441 	src_sa.sin6_addr = sip6->ip6_src;
2442 	/* we don't currently use sin6_scope_id, but eventually use it */
2443 	src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src);
2444 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2445 		goto fail;
2446 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2447 		goto fail;	/* what should we do here? */
2448 
2449 	/* rate limit */
2450 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2451 		goto fail;
2452 
2453 	/*
2454 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2455 	 * we almost always ask for an mbuf cluster for simplicity.
2456 	 * (MHLEN < IPV6_MMTU is almost always true)
2457 	 */
2458 #if IPV6_MMTU >= MCLBYTES
2459 # error assumption failed about IPV6_MMTU and MCLBYTES
2460 #endif
2461 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
2462 	if (m && IPV6_MMTU >= MHLEN)
2463 		MCLGET(m, M_DONTWAIT);
2464 	if (!m)
2465 		goto fail;
2466 	m->m_pkthdr.rcvif = NULL;
2467 	m->m_len = 0;
2468 	maxlen = M_TRAILINGSPACE(m);
2469 	maxlen = min(IPV6_MMTU, maxlen);
2470 	/* just for safety */
2471 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2472 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2473 		goto fail;
2474 	}
2475 
2476 	{
2477 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2478 		struct in6_ifaddr *ia;
2479 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2480 						 IN6_IFF_NOTREADY|
2481 						 IN6_IFF_ANYCAST)) == NULL)
2482 			goto fail;
2483 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2484 	}
2485 
2486 	/* get ip6 linklocal address for the router. */
2487 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2488 		struct sockaddr_in6 *sin6;
2489 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2490 		router_ll6 = &sin6->sin6_addr;
2491 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2492 			router_ll6 = (struct in6_addr *)NULL;
2493 	} else
2494 		router_ll6 = (struct in6_addr *)NULL;
2495 
2496 	/* ip6 */
2497 	ip6 = mtod(m, struct ip6_hdr *);
2498 	ip6->ip6_flow = 0;
2499 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2500 	ip6->ip6_vfc |= IPV6_VERSION;
2501 	/* ip6->ip6_plen will be set later */
2502 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2503 	ip6->ip6_hlim = 255;
2504 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2505 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2506 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2507 
2508 	/* ND Redirect */
2509 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2510 	nd_rd->nd_rd_type = ND_REDIRECT;
2511 	nd_rd->nd_rd_code = 0;
2512 	nd_rd->nd_rd_reserved = 0;
2513 	if (rt->rt_flags & RTF_GATEWAY) {
2514 		/*
2515 		 * nd_rd->nd_rd_target must be a link-local address in
2516 		 * better router cases.
2517 		 */
2518 		if (!router_ll6)
2519 			goto fail;
2520 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2521 		      sizeof(nd_rd->nd_rd_target));
2522 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2523 		      sizeof(nd_rd->nd_rd_dst));
2524 	} else {
2525 		/* make sure redtgt == reddst */
2526 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2527 		      sizeof(nd_rd->nd_rd_target));
2528 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2529 		      sizeof(nd_rd->nd_rd_dst));
2530 	}
2531 
2532 	p = (u_char *)(nd_rd + 1);
2533 
2534 	if (!router_ll6)
2535 		goto nolladdropt;
2536 
2537     {
2538 	/* target lladdr option */
2539 	struct rtentry *rt_router = NULL;
2540 	int len;
2541 	struct sockaddr_dl *sdl;
2542 	struct nd_opt_hdr *nd_opt;
2543 	char *lladdr;
2544 
2545 	rt_router = nd6_lookup(router_ll6, 0, ifp);
2546 	if (!rt_router)
2547 		goto nolladdropt;
2548 	len = sizeof(*nd_opt) + ifp->if_addrlen;
2549 	len = (len + 7) & ~7;	/* round by 8 */
2550 	/* safety check */
2551 	if (len + (p - (u_char *)ip6) > maxlen)
2552 		goto nolladdropt;
2553 	if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2554 	    (rt_router->rt_flags & RTF_LLINFO) &&
2555 	    (rt_router->rt_gateway->sa_family == AF_LINK) &&
2556 	    (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2557 	    sdl->sdl_alen) {
2558 		nd_opt = (struct nd_opt_hdr *)p;
2559 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2560 		nd_opt->nd_opt_len = len >> 3;
2561 		lladdr = (char *)(nd_opt + 1);
2562 		bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2563 		p += len;
2564 	}
2565     }
2566 nolladdropt:;
2567 
2568 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2569 
2570 	/* just to be safe */
2571 #ifdef M_DECRYPTED	/*not openbsd*/
2572 	if (m0->m_flags & M_DECRYPTED)
2573 		goto noredhdropt;
2574 #endif
2575 	if (p - (u_char *)ip6 > maxlen)
2576 		goto noredhdropt;
2577 
2578     {
2579 	/* redirected header option */
2580 	int len;
2581 	struct nd_opt_rd_hdr *nd_opt_rh;
2582 
2583 	/*
2584 	 * compute the maximum size for icmp6 redirect header option.
2585 	 * XXX room for auth header?
2586 	 */
2587 	len = maxlen - (p - (u_char *)ip6);
2588 	len &= ~7;
2589 
2590 	/* This is just for simplicity. */
2591 	if (m0->m_pkthdr.len != m0->m_len) {
2592 		if (m0->m_next) {
2593 			m_freem(m0->m_next);
2594 			m0->m_next = NULL;
2595 		}
2596 		m0->m_pkthdr.len = m0->m_len;
2597 	}
2598 
2599 	/*
2600 	 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2601 	 * about padding/truncate rule for the original IP packet.
2602 	 * From the discussion on IPv6imp in Feb 1999, the consensus was:
2603 	 * - "attach as much as possible" is the goal
2604 	 * - pad if not aligned (original size can be guessed by original
2605 	 *   ip6 header)
2606 	 * Following code adds the padding if it is simple enough,
2607 	 * and truncates if not.
2608 	 */
2609 	if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2610 		panic("assumption failed in %s:%d\n", __FILE__, __LINE__);
2611 
2612 	if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2613 		/* not enough room, truncate */
2614 		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2615 	} else {
2616 		/* enough room, pad or truncate */
2617 		size_t extra;
2618 
2619 		extra = m0->m_pkthdr.len % 8;
2620 		if (extra) {
2621 			/* pad if easy enough, truncate if not */
2622 			if (8 - extra <= M_TRAILINGSPACE(m0)) {
2623 				/* pad */
2624 				m0->m_len += (8 - extra);
2625 				m0->m_pkthdr.len += (8 - extra);
2626 			} else {
2627 				/* truncate */
2628 				m0->m_pkthdr.len -= extra;
2629 				m0->m_len -= extra;
2630 			}
2631 		}
2632 		len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2633 		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2634 	}
2635 
2636 	nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2637 	bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2638 	nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2639 	nd_opt_rh->nd_opt_rh_len = len >> 3;
2640 	p += sizeof(*nd_opt_rh);
2641 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2642 
2643 	/* connect m0 to m */
2644 	m->m_next = m0;
2645 	m->m_pkthdr.len = m->m_len + m0->m_len;
2646     }
2647 noredhdropt:;
2648 
2649 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
2650 		sip6->ip6_src.s6_addr16[1] = 0;
2651 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
2652 		sip6->ip6_dst.s6_addr16[1] = 0;
2653 #if 0
2654 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
2655 		ip6->ip6_src.s6_addr16[1] = 0;
2656 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
2657 		ip6->ip6_dst.s6_addr16[1] = 0;
2658 #endif
2659 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
2660 		nd_rd->nd_rd_target.s6_addr16[1] = 0;
2661 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
2662 		nd_rd->nd_rd_dst.s6_addr16[1] = 0;
2663 
2664 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2665 
2666 	nd_rd->nd_rd_cksum = 0;
2667 	nd_rd->nd_rd_cksum
2668 		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2669 
2670 	/* send the packet to outside... */
2671 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2672 	if (outif) {
2673 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2674 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2675 	}
2676 	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2677 
2678 	return;
2679 
2680 fail:
2681 	if (m)
2682 		m_freem(m);
2683 	if (m0)
2684 		m_freem(m0);
2685 }
2686 
2687 #ifdef HAVE_NRL_INPCB
2688 #define sotoin6pcb	sotoinpcb
2689 #define in6pcb		inpcb
2690 #define in6p_icmp6filt	inp_icmp6filt
2691 #endif
2692 /*
2693  * ICMPv6 socket option processing.
2694  */
2695 int
2696 icmp6_ctloutput(so, sopt)
2697 	struct socket *so;
2698 	struct sockopt *sopt;
2699 {
2700 	int error = 0;
2701 	int optlen;
2702 	struct inpcb *inp = sotoinpcb(so);
2703 	int level, op, optname;
2704 
2705 	if (sopt) {
2706 		level = sopt->sopt_level;
2707 		op = sopt->sopt_dir;
2708 		optname = sopt->sopt_name;
2709 		optlen = sopt->sopt_valsize;
2710 	} else
2711 		level = op = optname = optlen = 0;
2712 
2713 	if (level != IPPROTO_ICMPV6) {
2714 		return EINVAL;
2715 	}
2716 
2717 	switch (op) {
2718 	case PRCO_SETOPT:
2719 		switch (optname) {
2720 		case ICMP6_FILTER:
2721 		    {
2722 			struct icmp6_filter *p;
2723 
2724 			if (optlen != sizeof(*p)) {
2725 				error = EMSGSIZE;
2726 				break;
2727 			}
2728 			if (inp->in6p_icmp6filt == NULL) {
2729 				error = EINVAL;
2730 				break;
2731 			}
2732 			error = sooptcopyin(sopt, inp->in6p_icmp6filt, optlen,
2733 				optlen);
2734 			break;
2735 		    }
2736 
2737 		default:
2738 			error = ENOPROTOOPT;
2739 			break;
2740 		}
2741 		break;
2742 
2743 	case PRCO_GETOPT:
2744 		switch (optname) {
2745 		case ICMP6_FILTER:
2746 		    {
2747 			if (inp->in6p_icmp6filt == NULL) {
2748 				error = EINVAL;
2749 				break;
2750 			}
2751 			error = sooptcopyout(sopt, inp->in6p_icmp6filt,
2752 				sizeof(struct icmp6_filter));
2753 			break;
2754 		    }
2755 
2756 		default:
2757 			error = ENOPROTOOPT;
2758 			break;
2759 		}
2760 		break;
2761 	}
2762 
2763 	return(error);
2764 }
2765 #ifdef HAVE_NRL_INPCB
2766 #undef sotoin6pcb
2767 #undef in6pcb
2768 #undef in6p_icmp6filt
2769 #endif
2770 
2771 #ifndef HAVE_PPSRATECHECK
2772 #ifndef timersub
2773 #define	timersub(tvp, uvp, vvp)						\
2774 	do {								\
2775 		(vvp)->tv_sec = (tvp)->tv_sec - (uvp)->tv_sec;		\
2776 		(vvp)->tv_usec = (tvp)->tv_usec - (uvp)->tv_usec;	\
2777 		if ((vvp)->tv_usec < 0) {				\
2778 			(vvp)->tv_sec--;				\
2779 			(vvp)->tv_usec += 1000000;			\
2780 		}							\
2781 	} while (0)
2782 #endif
2783 
2784 /*
2785  * ppsratecheck(): packets (or events) per second limitation.
2786  */
2787 static int
2788 ppsratecheck(lasttime, curpps, maxpps)
2789 	struct timeval *lasttime;
2790 	int *curpps;
2791 	int maxpps;	/* maximum pps allowed */
2792 {
2793 	struct timeval tv, delta;
2794 	int s, rv;
2795 
2796 	s = splclock();
2797 	microtime(&tv);
2798 	splx(s);
2799 
2800 	timersub(&tv, lasttime, &delta);
2801 
2802 	/*
2803 	 * Check for 0,0 so that the message will be seen at least once.
2804 	 * If more than one second has passed since the last update of
2805 	 * lasttime, reset the counter.
2806 	 *
2807 	 * We do increment *curpps even in *curpps < maxpps case, as some may
2808 	 * try to use *curpps for stat purposes as well.
2809 	 */
2810 	if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) ||
2811 	    delta.tv_sec >= 1) {
2812 		*lasttime = tv;
2813 		*curpps = 0;
2814 		rv = 1;
2815 	} else if (maxpps < 0)
2816 		rv = 1;
2817 	else if (*curpps < maxpps)
2818 		rv = 1;
2819 	else
2820 		rv = 0;
2821 
2822 #if 1 /* DIAGNOSTIC? */
2823 	/* be careful about wrap-around */
2824 	if (*curpps + 1 > *curpps)
2825 		*curpps = *curpps + 1;
2826 #else
2827 	/*
2828 	 * assume that there's not too many calls to this function.
2829 	 * not sure if the assumption holds, as it depends on *caller's*
2830 	 * behavior, not the behavior of this function.
2831 	 * IMHO it is wrong to make assumption on the caller's behavior,
2832 	 * so the above #if is #if 1, not #ifdef DIAGNOSTIC.
2833 	 */
2834 	*curpps = *curpps + 1;
2835 #endif
2836 
2837 	return (rv);
2838 }
2839 #endif
2840 
2841 /*
2842  * Perform rate limit check.
2843  * Returns 0 if it is okay to send the icmp6 packet.
2844  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2845  * limitation.
2846  *
2847  * XXX per-destination/type check necessary?
2848  */
2849 static int
2850 icmp6_ratelimit(dst, type, code)
2851 	const struct in6_addr *dst;	/* not used at this moment */
2852 	const int type;			/* not used at this moment */
2853 	const int code;			/* not used at this moment */
2854 {
2855 	int ret;
2856 
2857 	ret = 0;	/* okay to send */
2858 
2859 	/* PPS limit */
2860 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2861 	    icmp6errppslim)) {
2862 		/* The packet is subject to rate limit */
2863 		ret++;
2864 	}
2865 
2866 	return ret;
2867 }
2868