xref: /dragonfly/sys/netinet6/icmp6.c (revision e98bdfd3)
1 /*	$FreeBSD: src/sys/netinet6/icmp6.c,v 1.6.2.13 2003/05/06 06:46:58 suz Exp $	*/
2 /*	$KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1988, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
62  */
63 
64 #include "opt_inet.h"
65 #include "opt_inet6.h"
66 #include "opt_ipsec.h"
67 
68 #include <sys/param.h>
69 #include <sys/systm.h>
70 #include <sys/malloc.h>
71 #include <sys/mbuf.h>
72 #include <sys/protosw.h>
73 #include <sys/socket.h>
74 #include <sys/socketvar.h>
75 #include <sys/socketops.h>
76 #include <sys/time.h>
77 #include <sys/kernel.h>
78 #include <sys/syslog.h>
79 #include <sys/domain.h>
80 
81 #include <sys/thread2.h>
82 #include <sys/msgport2.h>
83 
84 #include <net/if.h>
85 #include <net/route.h>
86 #include <net/if_dl.h>
87 #include <net/if_types.h>
88 
89 #include <netinet/in.h>
90 #include <netinet/in_var.h>
91 #include <netinet/ip6.h>
92 #include <netinet6/ip6_var.h>
93 #include <netinet/icmp6.h>
94 #include <netinet6/mld6_var.h>
95 #include <netinet/in_pcb.h>
96 #include <netinet6/in6_pcb.h>
97 #include <netinet6/nd6.h>
98 #include <netinet6/in6_ifattach.h>
99 #include <netinet6/ip6protosw.h>
100 
101 #ifdef IPSEC
102 #include <netinet6/ipsec.h>
103 #include <netproto/key/key.h>
104 #endif
105 
106 #ifdef FAST_IPSEC
107 #include <netproto/ipsec/ipsec.h>
108 #include <netproto/ipsec/key.h>
109 #define	IPSEC
110 #endif
111 
112 #include <net/net_osdep.h>
113 
114 #ifdef HAVE_NRL_INPCB
115 /* inpcb members */
116 #define in6pcb		inpcb
117 #define in6p_laddr	inp_laddr6
118 #define in6p_faddr	inp_faddr6
119 #define in6p_icmp6filt	inp_icmp6filt
120 #define in6p_route	inp_route
121 #define in6p_socket	inp_socket
122 #define in6p_flags	inp_flags
123 #define in6p_moptions	inp_moptions6
124 #define in6p_outputopts	inp_outputopts6
125 #define in6p_ip6	inp_ipv6
126 #define in6p_flowinfo	inp_flowinfo
127 #define in6p_sp		inp_sp
128 #define in6p_next	inp_next
129 #define in6p_prev	inp_prev
130 /* function names */
131 #define in6_pcbdetach	in_pcbdetach
132 #define in6_rtchange	in_rtchange
133 
134 /*
135  * for KAME src sync over BSD*'s. XXX: FreeBSD (>=3) are VERY different from
136  * others...
137  */
138 #define in6p_ip6_nxt	inp_ipv6.ip6_nxt
139 #endif
140 
141 /*
142  * ppratecheck() is available, so define it here.
143  */
144 #define	HAVE_PPSRATECHECK
145 
146 extern struct domain inet6domain;
147 extern struct protosw inet6sw[];
148 extern u_char ip6_protox[];
149 
150 struct icmp6stat icmp6stat;
151 
152 extern struct inpcbinfo ripcbinfo;
153 extern int icmp6errppslim;
154 static int icmp6errpps_count = 0;
155 static struct timeval icmp6errppslim_last;
156 extern int icmp6_nodeinfo;
157 
158 static void icmp6_errcount (struct icmp6errstat *, int, int);
159 static int icmp6_rip6_input (struct mbuf **, int);
160 static int icmp6_ratelimit (const struct in6_addr *, const int, const int);
161 static const char *icmp6_redirect_diag (struct in6_addr *,
162 	struct in6_addr *, struct in6_addr *);
163 #ifndef HAVE_PPSRATECHECK
164 static int ppsratecheck (struct timeval *, int *, int);
165 #endif
166 static struct mbuf *ni6_input (struct mbuf *, int);
167 static struct mbuf *ni6_nametodns (const char *, int, int);
168 static int ni6_dnsmatch (const char *, int, const char *, int);
169 static int ni6_addrs (struct icmp6_nodeinfo *, struct mbuf *,
170 			  struct ifnet **, char *);
171 static int ni6_store_addrs (struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
172 				struct ifnet *, int);
173 static int icmp6_notify_error (struct mbuf *, int, int, int);
174 
175 #ifdef COMPAT_RFC1885
176 static struct route_in6 icmp6_reflect_rt;
177 #endif
178 
179 
180 void
181 icmp6_init(void)
182 {
183 	mld6_init();
184 }
185 
186 static void
187 icmp6_errcount(struct icmp6errstat *stat, int type, int code)
188 {
189 	switch (type) {
190 	case ICMP6_DST_UNREACH:
191 		switch (code) {
192 		case ICMP6_DST_UNREACH_NOROUTE:
193 			stat->icp6errs_dst_unreach_noroute++;
194 			return;
195 		case ICMP6_DST_UNREACH_ADMIN:
196 			stat->icp6errs_dst_unreach_admin++;
197 			return;
198 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
199 			stat->icp6errs_dst_unreach_beyondscope++;
200 			return;
201 		case ICMP6_DST_UNREACH_ADDR:
202 			stat->icp6errs_dst_unreach_addr++;
203 			return;
204 		case ICMP6_DST_UNREACH_NOPORT:
205 			stat->icp6errs_dst_unreach_noport++;
206 			return;
207 		}
208 		break;
209 	case ICMP6_PACKET_TOO_BIG:
210 		stat->icp6errs_packet_too_big++;
211 		return;
212 	case ICMP6_TIME_EXCEEDED:
213 		switch (code) {
214 		case ICMP6_TIME_EXCEED_TRANSIT:
215 			stat->icp6errs_time_exceed_transit++;
216 			return;
217 		case ICMP6_TIME_EXCEED_REASSEMBLY:
218 			stat->icp6errs_time_exceed_reassembly++;
219 			return;
220 		}
221 		break;
222 	case ICMP6_PARAM_PROB:
223 		switch (code) {
224 		case ICMP6_PARAMPROB_HEADER:
225 			stat->icp6errs_paramprob_header++;
226 			return;
227 		case ICMP6_PARAMPROB_NEXTHEADER:
228 			stat->icp6errs_paramprob_nextheader++;
229 			return;
230 		case ICMP6_PARAMPROB_OPTION:
231 			stat->icp6errs_paramprob_option++;
232 			return;
233 		}
234 		break;
235 	case ND_REDIRECT:
236 		stat->icp6errs_redirect++;
237 		return;
238 	}
239 	stat->icp6errs_unknown++;
240 }
241 
242 /*
243  * Generate an error packet of type error in response to bad IP6 packet.
244  */
245 void
246 icmp6_error(struct mbuf *m, int type, int code, int param)
247 {
248 	struct ip6_hdr *oip6, *nip6;
249 	struct icmp6_hdr *icmp6;
250 	u_int preplen;
251 	int off;
252 	int nxt;
253 
254 	icmp6stat.icp6s_error++;
255 
256 	/* count per-type-code statistics */
257 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
258 
259 #ifdef M_DECRYPTED	/*not openbsd*/
260 	if (m->m_flags & M_DECRYPTED) {
261 		icmp6stat.icp6s_canterror++;
262 		goto freeit;
263 	}
264 #endif
265 
266 #ifndef PULLDOWN_TEST
267 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
268 #else
269 	if (m->m_len < sizeof(struct ip6_hdr)) {
270 		m = m_pullup(m, sizeof(struct ip6_hdr));
271 		if (m == NULL)
272 			return;
273 	}
274 #endif
275 	oip6 = mtod(m, struct ip6_hdr *);
276 
277 	/*
278 	 * Multicast destination check. For unrecognized option errors,
279 	 * this check has already done in ip6_unknown_opt(), so we can
280 	 * check only for other errors.
281 	 */
282 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
283 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
284 	    (type != ICMP6_PACKET_TOO_BIG &&
285 	     (type != ICMP6_PARAM_PROB ||
286 	      code != ICMP6_PARAMPROB_OPTION)))
287 		goto freeit;
288 
289 	/* Source address check. XXX: the case of anycast source? */
290 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
291 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
292 		goto freeit;
293 
294 	/*
295 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
296 	 * don't do it.
297 	 */
298 	nxt = -1;
299 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
300 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
301 		struct icmp6_hdr *icp;
302 
303 #ifndef PULLDOWN_TEST
304 		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
305 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
306 #else
307 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
308 			sizeof(*icp));
309 		if (icp == NULL) {
310 			icmp6stat.icp6s_tooshort++;
311 			return;
312 		}
313 #endif
314 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
315 		    icp->icmp6_type == ND_REDIRECT) {
316 			/*
317 			 * ICMPv6 error
318 			 * Special case: for redirect (which is
319 			 * informational) we must not send icmp6 error.
320 			 */
321 			icmp6stat.icp6s_canterror++;
322 			goto freeit;
323 		} else {
324 			/* ICMPv6 informational - send the error */
325 		}
326 	} else {
327 		/* non-ICMPv6 - send the error */
328 	}
329 
330 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
331 
332 	/* Finally, do rate limitation check. */
333 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
334 		icmp6stat.icp6s_toofreq++;
335 		goto freeit;
336 	}
337 
338 	/*
339 	 * OK, ICMP6 can be generated.
340 	 */
341 
342 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
343 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
344 
345 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
346 	M_PREPEND(m, preplen, M_NOWAIT);
347 	if (m && m->m_len < preplen)
348 		m = m_pullup(m, preplen);
349 	if (m == NULL) {
350 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
351 		return;
352 	}
353 
354 	nip6 = mtod(m, struct ip6_hdr *);
355 	nip6->ip6_src  = oip6->ip6_src;
356 	nip6->ip6_dst  = oip6->ip6_dst;
357 
358 	in6_clearscope(&oip6->ip6_src);
359 	in6_clearscope(&oip6->ip6_dst);
360 
361 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
362 	icmp6->icmp6_type = type;
363 	icmp6->icmp6_code = code;
364 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
365 
366 	/*
367 	 * icmp6_reflect() is designed to be in the input path.
368 	 * icmp6_error() can be called from both input and outut path,
369 	 * and if we are in output path rcvif could contain bogus value.
370 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
371 	 * information in ip header (nip6).
372 	 */
373 	m->m_pkthdr.rcvif = NULL;
374 
375 	icmp6stat.icp6s_outhist[type]++;
376 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
377 
378 	return;
379 
380 freeit:
381 	/*
382 	 * If we can't tell wheter or not we can generate ICMP6, free it.
383 	 */
384 	m_freem(m);
385 }
386 
387 /*
388  * Process a received ICMP6 message.
389  */
390 int
391 icmp6_input(struct mbuf **mp, int *offp, int proto)
392 {
393 	struct mbuf *m = *mp, *n;
394 	struct ip6_hdr *ip6, *nip6;
395 	struct icmp6_hdr *icmp6, *nicmp6;
396 	int off = *offp;
397 	int icmp6len = m->m_pkthdr.len - *offp;
398 	int code, sum, noff;
399 
400 #ifndef PULLDOWN_TEST
401 	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
402 	/* m might change if M_LOOP.  So, call mtod after this */
403 #endif
404 
405 	/*
406 	 * Locate icmp6 structure in mbuf, and check
407 	 * that not corrupted and of at least minimum length
408 	 */
409 
410 	ip6 = mtod(m, struct ip6_hdr *);
411 	if (icmp6len < sizeof(struct icmp6_hdr)) {
412 		icmp6stat.icp6s_tooshort++;
413 		goto freeit;
414 	}
415 
416 	/*
417 	 * calculate the checksum
418 	 */
419 #ifndef PULLDOWN_TEST
420 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
421 #else
422 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
423 	if (icmp6 == NULL) {
424 		icmp6stat.icp6s_tooshort++;
425 		return IPPROTO_DONE;
426 	}
427 #endif
428 	code = icmp6->icmp6_code;
429 
430 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
431 		nd6log((LOG_ERR,
432 		    "ICMP6 checksum error(%d|%x) %s\n",
433 		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
434 		icmp6stat.icp6s_checksum++;
435 		goto freeit;
436 	}
437 
438 	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
439 		/*
440 		 * Deliver very specific ICMP6 type only.
441 		 * This is important to deilver TOOBIG.  Otherwise PMTUD
442 		 * will not work.
443 		 */
444 		switch (icmp6->icmp6_type) {
445 		case ICMP6_DST_UNREACH:
446 		case ICMP6_PACKET_TOO_BIG:
447 		case ICMP6_TIME_EXCEEDED:
448 			break;
449 		default:
450 			goto freeit;
451 		}
452 	}
453 
454 	icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
455 	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
456 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
457 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
458 
459 	switch (icmp6->icmp6_type) {
460 	case ICMP6_DST_UNREACH:
461 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
462 		switch (code) {
463 		case ICMP6_DST_UNREACH_NOROUTE:
464 			code = PRC_UNREACH_NET;
465 			break;
466 		case ICMP6_DST_UNREACH_ADMIN:
467 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
468 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
469 			break;
470 		case ICMP6_DST_UNREACH_ADDR:
471 			code = PRC_HOSTDEAD;
472 			break;
473 #ifdef COMPAT_RFC1885
474 		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
475 			code = PRC_UNREACH_SRCFAIL;
476 			break;
477 #else
478 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
479 			/* I mean "source address was incorrect." */
480 			code = PRC_PARAMPROB;
481 			break;
482 #endif
483 		case ICMP6_DST_UNREACH_NOPORT:
484 			code = PRC_UNREACH_PORT;
485 			break;
486 		default:
487 			goto badcode;
488 		}
489 		goto deliver;
490 		break;
491 
492 	case ICMP6_PACKET_TOO_BIG:
493 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
494 		if (code != 0)
495 			goto badcode;
496 
497 		code = PRC_MSGSIZE;
498 
499 		/*
500 		 * Updating the path MTU will be done after examining
501 		 * intermediate extension headers.
502 		 */
503 		goto deliver;
504 		break;
505 
506 	case ICMP6_TIME_EXCEEDED:
507 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
508 		switch (code) {
509 		case ICMP6_TIME_EXCEED_TRANSIT:
510 		case ICMP6_TIME_EXCEED_REASSEMBLY:
511 			code += PRC_TIMXCEED_INTRANS;
512 			break;
513 		default:
514 			goto badcode;
515 		}
516 		goto deliver;
517 		break;
518 
519 	case ICMP6_PARAM_PROB:
520 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
521 		switch (code) {
522 		case ICMP6_PARAMPROB_NEXTHEADER:
523 			code = PRC_UNREACH_PROTOCOL;
524 			break;
525 		case ICMP6_PARAMPROB_HEADER:
526 		case ICMP6_PARAMPROB_OPTION:
527 			code = PRC_PARAMPROB;
528 			break;
529 		default:
530 			goto badcode;
531 		}
532 		goto deliver;
533 		break;
534 
535 	case ICMP6_ECHO_REQUEST:
536 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
537 		if (code != 0)
538 			goto badcode;
539 		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
540 			/* Give up remote */
541 			break;
542 		}
543 		if ((n->m_flags & M_EXT) ||
544 		    n->m_len < off + sizeof(struct icmp6_hdr)) {
545 			struct mbuf *n0 = n;
546 			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
547 			int n0len;
548 
549 			/*
550 			 * Prepare an internal mbuf. m_pullup() doesn't
551 			 * always copy the length we specified.
552 			 */
553 			if (maxlen >= MCLBYTES) {
554 				/* Give up remote */
555 				m_freem(n0);
556 				break;
557 			}
558 			n = m_getb(maxlen, M_NOWAIT, n0->m_type, M_PKTHDR);
559 			if (n == NULL) {
560 				/* Give up remote */
561 				m_freem(n0);
562 				break;
563 			}
564 			n0len = n0->m_pkthdr.len;	/* save for use below */
565 			M_MOVE_PKTHDR(n, n0);
566 			/*
567 			 * Copy IPv6 and ICMPv6 only.
568 			 */
569 			nip6 = mtod(n, struct ip6_hdr *);
570 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
571 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
572 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
573 			noff = sizeof(struct ip6_hdr);
574 			/* new mbuf contains only ipv6+icmpv6 headers */
575 			n->m_len = noff + sizeof(struct icmp6_hdr);
576 			/*
577 			 * Adjust mbuf. ip6_plen will be adjusted in
578 			 * ip6_output().
579 			 */
580 			m_adj(n0, off + sizeof(struct icmp6_hdr));
581 			/* recalculate complete packet size */
582 			n->m_pkthdr.len = n0len + (noff - off);
583 			n->m_next = n0;
584 		} else {
585 			nip6 = mtod(n, struct ip6_hdr *);
586 			nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
587 			noff = off;
588 		}
589 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
590 		nicmp6->icmp6_code = 0;
591 		if (n) {
592 			icmp6stat.icp6s_reflect++;
593 			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
594 			icmp6_reflect(n, noff);
595 		}
596 		break;
597 
598 	case ICMP6_ECHO_REPLY:
599 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
600 		if (code != 0)
601 			goto badcode;
602 		break;
603 
604 	case MLD_LISTENER_QUERY:
605 	case MLD_LISTENER_REPORT:
606 		if (icmp6len < sizeof(struct mld_hdr))
607 			goto badlen;
608 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
609 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
610 		else
611 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
612 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
613 			/* give up local */
614 			mld6_input(m, off);
615 			m = NULL;
616 			goto freeit;
617 		}
618 		mld6_input(n, off);
619 		/* m stays. */
620 		break;
621 
622 	case MLD_LISTENER_DONE:
623 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
624 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
625 			goto badlen;
626 		break;		/* nothing to be done in kernel */
627 
628 	case MLD_MTRACE_RESP:
629 	case MLD_MTRACE:
630 		/* XXX: these two are experimental.  not officially defind. */
631 		/* XXX: per-interface statistics? */
632 		break;		/* just pass it to applications */
633 
634 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
635 	    {
636 		enum { WRU, FQDN } mode;
637 
638 		if (!icmp6_nodeinfo)
639 			break;
640 
641 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
642 			mode = WRU;
643 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
644 			mode = FQDN;
645 		else
646 			goto badlen;
647 
648 #define hostnamelen	strlen(hostname)
649 		if (mode == FQDN) {
650 #ifndef PULLDOWN_TEST
651 			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
652 					 IPPROTO_DONE);
653 #endif
654 			n = m_copy(m, 0, M_COPYALL);
655 			if (n)
656 				n = ni6_input(n, off);
657 			/* XXX meaningless if n == NULL */
658 			noff = sizeof(struct ip6_hdr);
659 		} else {
660 			u_char *p;
661 			int maxlen, maxhlen;
662 
663 			if ((icmp6_nodeinfo & 5) != 5)
664 				break;
665 
666 			if (code != 0)
667 				goto badcode;
668 			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
669 			if (maxlen >= MCLBYTES) {
670 				/* Give up remote */
671 				break;
672 			}
673 			n = m_getb(maxlen, M_NOWAIT, m->m_type, M_PKTHDR);
674 			if (n == NULL) {
675 				/* Give up remote */
676 				break;
677 			}
678 			if (!m_dup_pkthdr(n, m, M_NOWAIT)) {
679 				/*
680 				 * Previous code did a blind M_COPY_PKTHDR
681 				 * and said "just for rcvif".  If true, then
682 				 * we could tolerate the dup failing (due to
683 				 * the deep copy of the tag chain).  For now
684 				 * be conservative and just fail.
685 				 */
686 				m_free(n);
687 				break;
688 			}
689 			n->m_len = 0;
690 			maxhlen = M_TRAILINGSPACE(n) - maxlen;
691 			if (maxhlen > hostnamelen)
692 				maxhlen = hostnamelen;
693 			/*
694 			 * Copy IPv6 and ICMPv6 only.
695 			 */
696 			nip6 = mtod(n, struct ip6_hdr *);
697 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
698 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
699 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
700 			p = (u_char *)(nicmp6 + 1);
701 			bzero(p, 4);
702 			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
703 			noff = sizeof(struct ip6_hdr);
704 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
705 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
706 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
707 			nicmp6->icmp6_code = 0;
708 		}
709 #undef hostnamelen
710 		if (n) {
711 			icmp6stat.icp6s_reflect++;
712 			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
713 			icmp6_reflect(n, noff);
714 		}
715 		break;
716 	    }
717 
718 	case ICMP6_WRUREPLY:
719 		if (code != 0)
720 			goto badcode;
721 		break;
722 
723 	case ND_ROUTER_SOLICIT:
724 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
725 		if (code != 0)
726 			goto badcode;
727 		if (icmp6len < sizeof(struct nd_router_solicit))
728 			goto badlen;
729 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
730 			/* give up local */
731 			nd6_rs_input(m, off, icmp6len);
732 			m = NULL;
733 			goto freeit;
734 		}
735 		nd6_rs_input(n, off, icmp6len);
736 		/* m stays. */
737 		break;
738 
739 	case ND_ROUTER_ADVERT:
740 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
741 		if (code != 0)
742 			goto badcode;
743 		if (icmp6len < sizeof(struct nd_router_advert))
744 			goto badlen;
745 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
746 			/* give up local */
747 			nd6_ra_input(m, off, icmp6len);
748 			m = NULL;
749 			goto freeit;
750 		}
751 		nd6_ra_input(n, off, icmp6len);
752 		/* m stays. */
753 		break;
754 
755 	case ND_NEIGHBOR_SOLICIT:
756 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
757 		if (code != 0)
758 			goto badcode;
759 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
760 			goto badlen;
761 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
762 			/* give up local */
763 			nd6_ns_input(m, off, icmp6len);
764 			m = NULL;
765 			goto freeit;
766 		}
767 		nd6_ns_input(n, off, icmp6len);
768 		/* m stays. */
769 		break;
770 
771 	case ND_NEIGHBOR_ADVERT:
772 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
773 		if (code != 0)
774 			goto badcode;
775 		if (icmp6len < sizeof(struct nd_neighbor_advert))
776 			goto badlen;
777 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
778 			/* give up local */
779 			nd6_na_input(m, off, icmp6len);
780 			m = NULL;
781 			goto freeit;
782 		}
783 		nd6_na_input(n, off, icmp6len);
784 		/* m stays. */
785 		break;
786 
787 	case ND_REDIRECT:
788 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
789 		if (code != 0)
790 			goto badcode;
791 		if (icmp6len < sizeof(struct nd_redirect))
792 			goto badlen;
793 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
794 			/* give up local */
795 			icmp6_redirect_input(m, off);
796 			m = NULL;
797 			goto freeit;
798 		}
799 		icmp6_redirect_input(n, off);
800 		/* m stays. */
801 		break;
802 
803 	case ICMP6_ROUTER_RENUMBERING:
804 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
805 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
806 			goto badcode;
807 		if (icmp6len < sizeof(struct icmp6_router_renum))
808 			goto badlen;
809 		break;
810 
811 	default:
812 		nd6log((LOG_DEBUG,
813 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
814 		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
815 		    ip6_sprintf(&ip6->ip6_dst),
816 		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
817 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
818 			/* ICMPv6 error: MUST deliver it by spec... */
819 			code = PRC_NCMDS;
820 			/* deliver */
821 		} else {
822 			/* ICMPv6 informational: MUST not deliver */
823 			break;
824 		}
825 	deliver:
826 		if (icmp6_notify_error(m, off, icmp6len, code)) {
827 			/* In this case, m should've been freed. */
828 			return (IPPROTO_DONE);
829 		}
830 		break;
831 
832 	badcode:
833 		icmp6stat.icp6s_badcode++;
834 		break;
835 
836 	badlen:
837 		icmp6stat.icp6s_badlen++;
838 		break;
839 	}
840 
841 	/* deliver the packet to appropriate sockets */
842 	icmp6_rip6_input(&m, *offp);
843 
844 	return IPPROTO_DONE;
845 
846 freeit:
847 	m_freem(m);
848 	return IPPROTO_DONE;
849 }
850 
851 static int
852 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
853 {
854 	struct icmp6_hdr *icmp6;
855 	struct ip6_hdr *eip6;
856 	u_int32_t notifymtu;
857 	struct sockaddr_in6 icmp6src, icmp6dst;
858 
859 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
860 		icmp6stat.icp6s_tooshort++;
861 		goto freeit;
862 	}
863 #ifndef PULLDOWN_TEST
864 	IP6_EXTHDR_CHECK(m, off,
865 			 sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr),
866 			 -1);
867 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
868 #else
869 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
870 		       sizeof(*icmp6) + sizeof(struct ip6_hdr));
871 	if (icmp6 == NULL) {
872 		icmp6stat.icp6s_tooshort++;
873 		return (-1);
874 	}
875 #endif
876 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
877 
878 	/* Detect the upper level protocol */
879 	{
880 		u_int8_t nxt = eip6->ip6_nxt;
881 		int eoff = off + sizeof(struct icmp6_hdr) +
882 			sizeof(struct ip6_hdr);
883 		struct ip6ctlparam ip6cp;
884 		struct in6_addr *finaldst = NULL;
885 		int icmp6type = icmp6->icmp6_type;
886 		struct ip6_frag *fh;
887 		struct ip6_rthdr *rth;
888 		int rthlen;
889 
890 		while (1) { /* XXX: should avoid infinite loop explicitly? */
891 			struct ip6_ext *eh;
892 
893 			switch (nxt) {
894 			case IPPROTO_HOPOPTS:
895 			case IPPROTO_DSTOPTS:
896 			case IPPROTO_AH:
897 #ifndef PULLDOWN_TEST
898 				IP6_EXTHDR_CHECK(m, 0,
899 				    eoff + sizeof(struct ip6_ext), -1);
900 				eh = (struct ip6_ext *)
901 				    (mtod(m, caddr_t) + eoff);
902 #else
903 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
904 					       eoff, sizeof(*eh));
905 				if (eh == NULL) {
906 					icmp6stat.icp6s_tooshort++;
907 					return (-1);
908 				}
909 #endif
910 
911 				if (nxt == IPPROTO_AH)
912 					eoff += (eh->ip6e_len + 2) << 2;
913 				else
914 					eoff += (eh->ip6e_len + 1) << 3;
915 				nxt = eh->ip6e_nxt;
916 				break;
917 			case IPPROTO_ROUTING:
918 				/*
919 				 * When the erroneous packet contains a
920 				 * routing header, we should examine the
921 				 * header to determine the final destination.
922 				 * Otherwise, we can't properly update
923 				 * information that depends on the final
924 				 * destination (e.g. path MTU).
925 				 */
926 #ifndef PULLDOWN_TEST
927 				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth),
928 						 -1);
929 				rth = (struct ip6_rthdr *)(mtod(m, caddr_t)
930 							   + eoff);
931 #else
932 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
933 					       eoff, sizeof(*rth));
934 				if (rth == NULL) {
935 					icmp6stat.icp6s_tooshort++;
936 					return (-1);
937 				}
938 #endif
939 				rthlen = (rth->ip6r_len + 1) << 3;
940 				eoff += rthlen;
941 				nxt = rth->ip6r_nxt;
942 				break;
943 			case IPPROTO_FRAGMENT:
944 #ifndef PULLDOWN_TEST
945 				IP6_EXTHDR_CHECK(m, 0, eoff +
946 						 sizeof(struct ip6_frag),
947 						 -1);
948 				fh = (struct ip6_frag *)(mtod(m, caddr_t)
949 							 + eoff);
950 #else
951 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
952 					       eoff, sizeof(*fh));
953 				if (fh == NULL) {
954 					icmp6stat.icp6s_tooshort++;
955 					return (-1);
956 				}
957 #endif
958 				/*
959 				 * Data after a fragment header is meaningless
960 				 * unless it is the first fragment, but
961 				 * we'll go to the notify label for path MTU
962 				 * discovery.
963 				 */
964 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
965 					goto notify;
966 
967 				eoff += sizeof(struct ip6_frag);
968 				nxt = fh->ip6f_nxt;
969 				break;
970 			default:
971 				/*
972 				 * This case includes ESP and the No Next
973 				 * Header.  In such cases going to the notify
974 				 * label does not have any meaning
975 				 * (i.e. pr_ctlinput will be NULL), but we go
976 				 * anyway since we might have to update
977 				 * path MTU information.
978 				 */
979 				goto notify;
980 			}
981 		}
982 	  notify:
983 #ifndef PULLDOWN_TEST
984 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
985 #else
986 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
987 			       sizeof(*icmp6) + sizeof(struct ip6_hdr));
988 		if (icmp6 == NULL) {
989 			icmp6stat.icp6s_tooshort++;
990 			return (-1);
991 		}
992 #endif
993 
994 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
995 		bzero(&icmp6dst, sizeof(icmp6dst));
996 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
997 		icmp6dst.sin6_family = AF_INET6;
998 		if (finaldst == NULL)
999 			icmp6dst.sin6_addr = eip6->ip6_dst;
1000 		else
1001 			icmp6dst.sin6_addr = *finaldst;
1002 		if (in6_addr2zoneid(m->m_pkthdr.rcvif, &icmp6dst.sin6_addr,
1003 		    &icmp6dst.sin6_scope_id))
1004 			goto freeit;
1005 		if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1006 				   NULL, NULL)) {
1007 			/* should be impossbile */
1008 			nd6log((LOG_DEBUG,
1009 			    "icmp6_notify_error: in6_embedscope failed\n"));
1010 			goto freeit;
1011 		}
1012 
1013 		/*
1014 		 * retrieve parameters from the inner IPv6 header, and convert
1015 		 * them into sockaddr structures.
1016 		 */
1017 		bzero(&icmp6src, sizeof(icmp6src));
1018 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1019 		icmp6src.sin6_family = AF_INET6;
1020 		icmp6src.sin6_addr = eip6->ip6_src;
1021 		if (in6_addr2zoneid(m->m_pkthdr.rcvif, &icmp6src.sin6_addr,
1022 		    &icmp6src.sin6_scope_id))
1023 			goto freeit;
1024 		if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1025 				   NULL, NULL)) {
1026 			/* should be impossbile */
1027 			nd6log((LOG_DEBUG,
1028 			    "icmp6_notify_error: in6_embedscope failed\n"));
1029 			goto freeit;
1030 		}
1031 		icmp6src.sin6_flowinfo =
1032 			(eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1033 
1034 		if (finaldst == NULL)
1035 			finaldst = &eip6->ip6_dst;
1036 		ip6cp.ip6c_m = m;
1037 		ip6cp.ip6c_icmp6 = icmp6;
1038 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1039 		ip6cp.ip6c_off = eoff;
1040 		ip6cp.ip6c_finaldst = finaldst;
1041 		ip6cp.ip6c_src = &icmp6src;
1042 		ip6cp.ip6c_nxt = nxt;
1043 
1044 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1045 			notifymtu = ntohl(icmp6->icmp6_mtu);
1046 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1047 			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1048 		}
1049 
1050 		so_pr_ctlinput(
1051 			&inet6sw[ip6_protox[nxt]],
1052 			code, (struct sockaddr *)&icmp6dst, &ip6cp);
1053 	}
1054 	return (0);
1055 
1056 freeit:
1057 	m_freem(m);
1058 	return (-1);
1059 }
1060 
1061 void
1062 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1063 {
1064 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1065 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1066 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1067 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1068 	struct rtentry *rt = NULL;
1069 	struct sockaddr_in6 sin6;
1070 
1071 	if (!validated)
1072 		return;
1073 
1074 	bzero(&sin6, sizeof(sin6));
1075 	sin6.sin6_family = PF_INET6;
1076 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1077 	sin6.sin6_addr = *dst;
1078 	/* XXX normally, this won't happen */
1079 	if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1080 		sin6.sin6_addr.s6_addr16[1] =
1081 		    htons(m->m_pkthdr.rcvif->if_index);
1082 	}
1083 	/* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1084 	rt = rtpurelookup((struct sockaddr *)&sin6);
1085 	if (rt != NULL && (rt->rt_flags & RTF_HOST) &&
1086 	    !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1087 		if (mtu < IPV6_MMTU) {				/* XXX */
1088 			rt->rt_rmx.rmx_locks |= RTV_MTU;
1089 		} else if (mtu < rt->rt_ifp->if_mtu &&
1090 			   rt->rt_rmx.rmx_mtu > mtu) {
1091 			icmp6stat.icp6s_pmtuchg++;
1092 			rt->rt_rmx.rmx_mtu = mtu;
1093 		}
1094 	}
1095 	if (rt)
1096 		--rt->rt_refcnt;
1097 }
1098 
1099 /*
1100  * Process a Node Information Query packet, based on
1101  * draft-ietf-ipngwg-icmp-name-lookups-07.
1102  *
1103  * Spec incompatibilities:
1104  * - IPv6 Subject address handling
1105  * - IPv4 Subject address handling support missing
1106  * - Proxy reply (answer even if it's not for me)
1107  * - joins NI group address at in6_ifattach() time only, does not cope
1108  *   with hostname changes by sethostname(3)
1109  */
1110 #define hostnamelen	strlen(hostname)
1111 static struct mbuf *
1112 ni6_input(struct mbuf *m, int off)
1113 {
1114 	struct icmp6_nodeinfo *ni6, *nni6;
1115 	struct mbuf *n = NULL;
1116 	u_int16_t qtype;
1117 	int subjlen;
1118 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1119 	struct ni_reply_fqdn *fqdn;
1120 	int addrs;		/* for NI_QTYPE_NODEADDR */
1121 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1122 	struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1123 	struct sockaddr_in6 sin6_d; /* XXX: we should retrieve this from m_aux */
1124 	struct ip6_hdr *ip6;
1125 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1126 	char *subj = NULL;
1127 	struct in6_ifaddr *ia6 = NULL;
1128 
1129 	ip6 = mtod(m, struct ip6_hdr *);
1130 #ifndef PULLDOWN_TEST
1131 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1132 #else
1133 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1134 	if (ni6 == NULL) {
1135 		/* m is already reclaimed */
1136 		return NULL;
1137 	}
1138 #endif
1139 
1140 	/*
1141 	 * Validate IPv6 destination address.
1142 	 *
1143 	 * The Responder must discard the Query without further processing
1144 	 * unless it is one of the Responder's unicast or anycast addresses, or
1145 	 * a link-local scope multicast address which the Responder has joined.
1146 	 * [icmp-name-lookups-07, Section 4.]
1147 	 */
1148 	bzero(&sin6, sizeof(sin6));
1149 	sin6.sin6_family = AF_INET6;
1150 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1151 	bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1152 	/* XXX scopeid */
1153 	if ((ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)&sin6)) != NULL) {
1154 		/* unicast/anycast, fine */
1155 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1156 		    (icmp6_nodeinfo & 4) == 0) {
1157 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1158 				"a temporary address in %s:%d",
1159 			       __FILE__, __LINE__));
1160 			goto bad;
1161 		}
1162 	} else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1163 		; /* link-local multicast, fine */
1164 	else
1165 		goto bad;
1166 
1167 	/* validate query Subject field. */
1168 	qtype = ntohs(ni6->ni_qtype);
1169 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1170 	switch (qtype) {
1171 	case NI_QTYPE_NOOP:
1172 	case NI_QTYPE_SUPTYPES:
1173 		/* 07 draft */
1174 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1175 			break;
1176 		/* FALLTHROUGH */
1177 	case NI_QTYPE_FQDN:
1178 	case NI_QTYPE_NODEADDR:
1179 		switch (ni6->ni_code) {
1180 		case ICMP6_NI_SUBJ_IPV6:
1181 #if ICMP6_NI_SUBJ_IPV6 != 0
1182 		case 0:
1183 #endif
1184 			/*
1185 			 * backward compatibility - try to accept 03 draft
1186 			 * format, where no Subject is present.
1187 			 */
1188 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1189 			    subjlen == 0) {
1190 				oldfqdn++;
1191 				break;
1192 			}
1193 #if ICMP6_NI_SUBJ_IPV6 != 0
1194 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1195 				goto bad;
1196 #endif
1197 
1198 			if (subjlen != sizeof(sin6.sin6_addr))
1199 				goto bad;
1200 
1201 			/*
1202 			 * Validate Subject address.
1203 			 *
1204 			 * Not sure what exactly "address belongs to the node"
1205 			 * means in the spec, is it just unicast, or what?
1206 			 *
1207 			 * At this moment we consider Subject address as
1208 			 * "belong to the node" if the Subject address equals
1209 			 * to the IPv6 destination address; validation for
1210 			 * IPv6 destination address should have done enough
1211 			 * check for us.
1212 			 *
1213 			 * We do not do proxy at this moment.
1214 			 */
1215 			/* m_pulldown instead of copy? */
1216 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1217 			    subjlen, (caddr_t)&sin6.sin6_addr);
1218 			if (in6_addr2zoneid(m->m_pkthdr.rcvif,
1219 			    &sin6.sin6_addr, &sin6.sin6_scope_id))
1220 				goto bad;
1221 			in6_embedscope(&sin6.sin6_addr, &sin6, NULL, NULL);
1222 			bzero(&sin6_d, sizeof(sin6_d));
1223 			sin6_d.sin6_family = AF_INET6; /* not used, actually */
1224 			sin6_d.sin6_len = sizeof(sin6_d); /* ditto */
1225 			sin6_d.sin6_addr = ip6->ip6_dst;
1226 			if (in6_addr2zoneid(m->m_pkthdr.rcvif,
1227 			    &ip6->ip6_dst, &sin6_d.sin6_scope_id))
1228 				goto bad;
1229 			in6_embedscope(&sin6_d.sin6_addr, &sin6_d, NULL, NULL);
1230 			subj = (char *)&sin6;
1231 			if (SA6_ARE_ADDR_EQUAL(&sin6, &sin6_d))
1232 				break;
1233 
1234 			/*
1235 			 * XXX if we are to allow other cases, we should really
1236 			 * be careful about scope here.
1237 			 * basically, we should disallow queries toward IPv6
1238 			 * destination X with subject Y, if scope(X) > scope(Y).
1239 			 * if we allow scope(X) > scope(Y), it will result in
1240 			 * information leakage across scope boundary.
1241 			 */
1242 			goto bad;
1243 
1244 		case ICMP6_NI_SUBJ_FQDN:
1245 			/*
1246 			 * Validate Subject name with gethostname(3).
1247 			 *
1248 			 * The behavior may need some debate, since:
1249 			 * - we are not sure if the node has FQDN as
1250 			 *   hostname (returned by gethostname(3)).
1251 			 * - the code does wildcard match for truncated names.
1252 			 *   however, we are not sure if we want to perform
1253 			 *   wildcard match, if gethostname(3) side has
1254 			 *   truncated hostname.
1255 			 */
1256 			n = ni6_nametodns(hostname, hostnamelen, 0);
1257 			if (!n || n->m_next || n->m_len == 0)
1258 				goto bad;
1259 			IP6_EXTHDR_GET(subj, char *, m,
1260 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1261 			if (subj == NULL)
1262 				goto bad;
1263 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1264 					n->m_len)) {
1265 				goto bad;
1266 			}
1267 			m_freem(n);
1268 			n = NULL;
1269 			break;
1270 
1271 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1272 		default:
1273 			goto bad;
1274 		}
1275 		break;
1276 	}
1277 
1278 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1279 	switch (qtype) {
1280 	case NI_QTYPE_FQDN:
1281 		if ((icmp6_nodeinfo & 1) == 0)
1282 			goto bad;
1283 		break;
1284 	case NI_QTYPE_NODEADDR:
1285 		if ((icmp6_nodeinfo & 2) == 0)
1286 			goto bad;
1287 		break;
1288 	}
1289 
1290 	/* guess reply length */
1291 	switch (qtype) {
1292 	case NI_QTYPE_NOOP:
1293 		break;		/* no reply data */
1294 	case NI_QTYPE_SUPTYPES:
1295 		replylen += sizeof(u_int32_t);
1296 		break;
1297 	case NI_QTYPE_FQDN:
1298 		/* XXX will append an mbuf */
1299 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1300 		break;
1301 	case NI_QTYPE_NODEADDR:
1302 		addrs = ni6_addrs(ni6, m, &ifp, subj);
1303 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1304 					  sizeof(u_int32_t))) > MCLBYTES)
1305 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1306 		break;
1307 	default:
1308 		/*
1309 		 * XXX: We must return a reply with the ICMP6 code
1310 		 * `unknown Qtype' in this case. However we regard the case
1311 		 * as an FQDN query for backward compatibility.
1312 		 * Older versions set a random value to this field,
1313 		 * so it rarely varies in the defined qtypes.
1314 		 * But the mechanism is not reliable...
1315 		 * maybe we should obsolete older versions.
1316 		 */
1317 		qtype = NI_QTYPE_FQDN;
1318 		/* XXX will append an mbuf */
1319 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1320 		oldfqdn++;
1321 		break;
1322 	}
1323 
1324 	/* allocate an mbuf to reply. */
1325 	if (replylen > MCLBYTES) {
1326 		/*
1327 		 * XXX: should we try to allocate more? But MCLBYTES
1328 		 * is probably much larger than IPV6_MMTU...
1329 		 */
1330 		goto bad;
1331 	}
1332 	n = m_getb(replylen, M_NOWAIT, m->m_type, M_PKTHDR);
1333 	if (n == NULL) {
1334 		m_freem(m);
1335 		return (NULL);
1336 	}
1337 	M_MOVE_PKTHDR(n, m); /* just for recvif */
1338 	n->m_pkthdr.len = n->m_len = replylen;
1339 
1340 	/* copy mbuf header and IPv6 + Node Information base headers */
1341 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1342 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1343 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1344 
1345 	/* qtype dependent procedure */
1346 	switch (qtype) {
1347 	case NI_QTYPE_NOOP:
1348 		nni6->ni_code = ICMP6_NI_SUCCESS;
1349 		nni6->ni_flags = 0;
1350 		break;
1351 	case NI_QTYPE_SUPTYPES:
1352 	{
1353 		u_int32_t v;
1354 		nni6->ni_code = ICMP6_NI_SUCCESS;
1355 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1356 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1357 		v = (u_int32_t)htonl(0x0000000f);
1358 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1359 		break;
1360 	}
1361 	case NI_QTYPE_FQDN:
1362 		nni6->ni_code = ICMP6_NI_SUCCESS;
1363 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1364 						sizeof(struct ip6_hdr) +
1365 						sizeof(struct icmp6_nodeinfo));
1366 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1367 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1368 		/*
1369 		 * XXX do we really have FQDN in variable "hostname"?
1370 		 */
1371 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1372 		if (n->m_next == NULL)
1373 			goto bad;
1374 		/* XXX we assume that n->m_next is not a chain */
1375 		if (n->m_next->m_next != NULL)
1376 			goto bad;
1377 		n->m_pkthdr.len += n->m_next->m_len;
1378 		break;
1379 	case NI_QTYPE_NODEADDR:
1380 	{
1381 		int lenlim, copied;
1382 
1383 		nni6->ni_code = ICMP6_NI_SUCCESS;
1384 		n->m_pkthdr.len = n->m_len =
1385 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1386 		lenlim = M_TRAILINGSPACE(n);
1387 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1388 		/* XXX: reset mbuf length */
1389 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1390 			sizeof(struct icmp6_nodeinfo) + copied;
1391 		break;
1392 	}
1393 	default:
1394 		break;		/* XXX impossible! */
1395 	}
1396 
1397 	nni6->ni_type = ICMP6_NI_REPLY;
1398 	m_freem(m);
1399 	return (n);
1400 
1401 bad:
1402 	m_freem(m);
1403 	if (n)
1404 		m_freem(n);
1405 	return (NULL);
1406 }
1407 #undef hostnamelen
1408 
1409 /*
1410  * make a mbuf with DNS-encoded string.  no compression support.
1411  *
1412  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1413  * treated as truncated name (two \0 at the end).  this is a wild guess.
1414  */
1415 static struct mbuf *
1416 ni6_nametodns(const char *name, int namelen,
1417 	      int old) /* return pascal string if non-zero */
1418 {
1419 	struct mbuf *m;
1420 	char *cp, *ep;
1421 	const char *p, *q;
1422 	int i, len, nterm;
1423 
1424 	if (old)
1425 		len = namelen + 1;
1426 	else
1427 		len = MCLBYTES;
1428 
1429 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1430 	m = m_getb(len, M_NOWAIT, MT_DATA, 0);
1431 	if (!m)
1432 		goto fail;
1433 
1434 	if (old) {
1435 		m->m_len = len;
1436 		*mtod(m, char *) = namelen;
1437 		bcopy(name, mtod(m, char *) + 1, namelen);
1438 		return m;
1439 	} else {
1440 		m->m_len = 0;
1441 		cp = mtod(m, char *);
1442 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1443 
1444 		/* if not certain about my name, return empty buffer */
1445 		if (namelen == 0)
1446 			return m;
1447 
1448 		/*
1449 		 * guess if it looks like shortened hostname, or FQDN.
1450 		 * shortened hostname needs two trailing "\0".
1451 		 */
1452 		i = 0;
1453 		for (p = name; p < name + namelen; p++) {
1454 			if (*p && *p == '.')
1455 				i++;
1456 		}
1457 		if (i < 2)
1458 			nterm = 2;
1459 		else
1460 			nterm = 1;
1461 
1462 		p = name;
1463 		while (cp < ep && p < name + namelen) {
1464 			i = 0;
1465 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1466 				i++;
1467 			/* result does not fit into mbuf */
1468 			if (cp + i + 1 >= ep)
1469 				goto fail;
1470 			/*
1471 			 * DNS label length restriction, RFC1035 page 8.
1472 			 * "i == 0" case is included here to avoid returning
1473 			 * 0-length label on "foo..bar".
1474 			 */
1475 			if (i <= 0 || i >= 64)
1476 				goto fail;
1477 			*cp++ = i;
1478 			bcopy(p, cp, i);
1479 			cp += i;
1480 			p = q;
1481 			if (p < name + namelen && *p == '.')
1482 				p++;
1483 		}
1484 		/* termination */
1485 		if (cp + nterm >= ep)
1486 			goto fail;
1487 		while (nterm-- > 0)
1488 			*cp++ = '\0';
1489 		m->m_len = cp - mtod(m, char *);
1490 		return m;
1491 	}
1492 
1493 	panic("should not reach here");
1494 	/* NOTREACHED */
1495 
1496 fail:
1497 	if (m)
1498 		m_freem(m);
1499 	return NULL;
1500 }
1501 
1502 /*
1503  * check if two DNS-encoded string matches.  takes care of truncated
1504  * form (with \0\0 at the end).  no compression support.
1505  * XXX upper/lowercase match (see RFC2065)
1506  */
1507 static int
1508 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1509 {
1510 	const char *a0, *b0;
1511 	int l;
1512 
1513 	/* simplest case - need validation? */
1514 	if (alen == blen && bcmp(a, b, alen) == 0)
1515 		return 1;
1516 
1517 	a0 = a;
1518 	b0 = b;
1519 
1520 	/* termination is mandatory */
1521 	if (alen < 2 || blen < 2)
1522 		return 0;
1523 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1524 		return 0;
1525 	alen--;
1526 	blen--;
1527 
1528 	while (a - a0 < alen && b - b0 < blen) {
1529 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1530 			return 0;
1531 
1532 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1533 			return 0;
1534 		/* we don't support compression yet */
1535 		if (a[0] >= 64 || b[0] >= 64)
1536 			return 0;
1537 
1538 		/* truncated case */
1539 		if (a[0] == 0 && a - a0 == alen - 1)
1540 			return 1;
1541 		if (b[0] == 0 && b - b0 == blen - 1)
1542 			return 1;
1543 		if (a[0] == 0 || b[0] == 0)
1544 			return 0;
1545 
1546 		if (a[0] != b[0])
1547 			return 0;
1548 		l = a[0];
1549 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1550 			return 0;
1551 		if (bcmp(a + 1, b + 1, l) != 0)
1552 			return 0;
1553 
1554 		a += 1 + l;
1555 		b += 1 + l;
1556 	}
1557 
1558 	if (a - a0 == alen && b - b0 == blen)
1559 		return 1;
1560 	else
1561 		return 0;
1562 }
1563 
1564 /*
1565  * calculate the number of addresses to be returned in the node info reply.
1566  */
1567 static int
1568 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1569 	  char *subj)
1570 {
1571 	const struct ifnet_array *arr;
1572 	struct in6_ifaddr *ifa6;
1573 	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1574 	int addrs = 0, addrsofif, iffound = 0;
1575 	int niflags = ni6->ni_flags;
1576 	int i;
1577 
1578 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1579 		switch (ni6->ni_code) {
1580 		case ICMP6_NI_SUBJ_IPV6:
1581 			if (subj == NULL) /* must be impossible... */
1582 				return (0);
1583 			subj_ip6 = (struct sockaddr_in6 *)subj;
1584 			break;
1585 		default:
1586 			/*
1587 			 * XXX: we only support IPv6 subject address for
1588 			 * this Qtype.
1589 			 */
1590 			return (0);
1591 		}
1592 	}
1593 
1594 	arr = ifnet_array_get();
1595 	for (i = 0; i < arr->ifnet_count; ++i) {
1596 		struct ifnet *ifp = arr->ifnet_arr[i];
1597 		struct ifaddr_container *ifac;
1598 
1599 		addrsofif = 0;
1600 		TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link)
1601 		{
1602 			struct ifaddr *ifa = ifac->ifa;
1603 
1604 			if (ifa->ifa_addr->sa_family != AF_INET6)
1605 				continue;
1606 			ifa6 = (struct in6_ifaddr *)ifa;
1607 
1608 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1609 			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1610 					       &ifa6->ia_addr.sin6_addr))
1611 				iffound = 1;
1612 
1613 			/*
1614 			 * IPv4-mapped addresses can only be returned by a
1615 			 * Node Information proxy, since they represent
1616 			 * addresses of IPv4-only nodes, which perforce do
1617 			 * not implement this protocol.
1618 			 * [icmp-name-lookups-07, Section 5.4]
1619 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1620 			 * this function at this moment.
1621 			 */
1622 
1623 			/* What do we have to do about ::1? */
1624 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1625 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1626 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1627 					continue;
1628 				break;
1629 			case IPV6_ADDR_SCOPE_SITELOCAL:
1630 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1631 					continue;
1632 				break;
1633 			case IPV6_ADDR_SCOPE_GLOBAL:
1634 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1635 					continue;
1636 				break;
1637 			default:
1638 				continue;
1639 			}
1640 
1641 			/*
1642 			 * check if anycast is okay.
1643 			 * XXX: just experimental.  not in the spec.
1644 			 */
1645 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) &&
1646 			    !(niflags & NI_NODEADDR_FLAG_ANYCAST))
1647 				continue; /* we need only unicast addresses */
1648 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) &&
1649 			    (icmp6_nodeinfo & 4) == 0) {
1650 				continue;
1651 			}
1652 			addrsofif++; /* count the address */
1653 		}
1654 		if (iffound) {
1655 			*ifpp = ifp;
1656 			return (addrsofif);
1657 		}
1658 
1659 		addrs += addrsofif;
1660 	}
1661 
1662 	return (addrs);
1663 }
1664 
1665 static int
1666 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1667 		struct ifnet *ifp0, int resid)
1668 {
1669 	const struct ifnet_array *arr;
1670 	struct in6_ifaddr *ifa6;
1671 	int copied = 0, allow_deprecated = 0;
1672 	u_char *cp = (u_char *)(nni6 + 1);
1673 	int niflags = ni6->ni_flags;
1674 	u_int32_t ltime;
1675 	int idx, idx_dep = -1;
1676 
1677 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1678 		return (0);	/* needless to copy */
1679 
1680 	arr = ifnet_array_get();
1681 	if (ifp0 == NULL) {
1682 		idx = 0;
1683 	} else {
1684 		for (idx = 0; idx < arr->ifnet_count; ++idx) {
1685 			if (arr->ifnet_arr[idx] == ifp0)
1686 				break;
1687 		}
1688 	}
1689 
1690 again:
1691 
1692 	for (; idx < arr->ifnet_count; ++idx) {
1693 		struct ifnet *ifp = arr->ifnet_arr[idx];
1694 		struct ifaddr_container *ifac;
1695 
1696 		TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1697 			struct ifaddr *ifa = ifac->ifa;
1698 
1699 			if (ifa->ifa_addr->sa_family != AF_INET6)
1700 				continue;
1701 			ifa6 = (struct in6_ifaddr *)ifa;
1702 
1703 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) &&
1704 			    allow_deprecated == 0) {
1705 				/*
1706 				 * prefererred address should be put before
1707 				 * deprecated addresses.
1708 				 */
1709 
1710 				/* record the interface for later search */
1711 				if (idx_dep < 0)
1712 					idx_dep = idx;
1713 
1714 				continue;
1715 			}
1716 			else if (!(ifa6->ia6_flags & IN6_IFF_DEPRECATED) &&
1717 				 allow_deprecated != 0)
1718 				continue; /* we now collect deprecated addrs */
1719 
1720 			/* What do we have to do about ::1? */
1721 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1722 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1723 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1724 					continue;
1725 				break;
1726 			case IPV6_ADDR_SCOPE_SITELOCAL:
1727 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1728 					continue;
1729 				break;
1730 			case IPV6_ADDR_SCOPE_GLOBAL:
1731 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1732 					continue;
1733 				break;
1734 			default:
1735 				continue;
1736 			}
1737 
1738 			/*
1739 			 * check if anycast is okay.
1740 			 * XXX: just experimental. not in the spec.
1741 			 */
1742 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) &&
1743 			    !(niflags & NI_NODEADDR_FLAG_ANYCAST))
1744 				continue;
1745 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) &&
1746 			    (icmp6_nodeinfo & 4) == 0) {
1747 				continue;
1748 			}
1749 
1750 			/* now we can copy the address */
1751 			if (resid < sizeof(struct in6_addr) +
1752 			    sizeof(u_int32_t)) {
1753 				/*
1754 				 * We give up much more copy.
1755 				 * Set the truncate flag and return.
1756 				 */
1757 				nni6->ni_flags |=
1758 					NI_NODEADDR_FLAG_TRUNCATE;
1759 				return (copied);
1760 			}
1761 
1762 			/*
1763 			 * Set the TTL of the address.
1764 			 * The TTL value should be one of the following
1765 			 * according to the specification:
1766 			 *
1767 			 * 1. The remaining lifetime of a DHCP lease on the
1768 			 *    address, or
1769 			 * 2. The remaining Valid Lifetime of a prefix from
1770 			 *    which the address was derived through Stateless
1771 			 *    Autoconfiguration.
1772 			 *
1773 			 * Note that we currently do not support stateful
1774 			 * address configuration by DHCPv6, so the former
1775 			 * case can't happen.
1776 			 */
1777 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1778 				ltime = ND6_INFINITE_LIFETIME;
1779 			else {
1780 				if (ifa6->ia6_lifetime.ia6t_expire >
1781 				    time_uptime)
1782 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime);
1783 				else
1784 					ltime = 0;
1785 			}
1786 
1787 			bcopy(&ltime, cp, sizeof(u_int32_t));
1788 			cp += sizeof(u_int32_t);
1789 
1790 			/* copy the address itself */
1791 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1792 			      sizeof(struct in6_addr));
1793 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1794 			cp += sizeof(struct in6_addr);
1795 
1796 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1797 			copied += (sizeof(struct in6_addr) +
1798 				   sizeof(u_int32_t));
1799 		}
1800 		if (ifp0)	/* we need search only on the specified IF */
1801 			break;
1802 	}
1803 
1804 	if (allow_deprecated == 0 && idx_dep >= 0) {
1805 		idx = idx_dep;
1806 		allow_deprecated = 1;
1807 
1808 		goto again;
1809 	}
1810 
1811 	return (copied);
1812 }
1813 
1814 /*
1815  * XXX almost dup'ed code with rip6_input.
1816  */
1817 static int
1818 icmp6_rip6_input(struct	mbuf **mp, int	off)
1819 {
1820 	struct mbuf *m = *mp;
1821 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1822 	struct in6pcb *in6p;
1823 	struct in6pcb *last = NULL;
1824 	struct sockaddr_in6 fromsa;
1825 	struct icmp6_hdr *icmp6;
1826 	struct mbuf *opts = NULL;
1827 
1828 #ifndef PULLDOWN_TEST
1829 	/* this is assumed to be safe. */
1830 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1831 #else
1832 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1833 	if (icmp6 == NULL) {
1834 		/* m is already reclaimed */
1835 		return IPPROTO_DONE;
1836 	}
1837 #endif
1838 
1839 	bzero(&fromsa, sizeof(fromsa));
1840 	fromsa.sin6_len = sizeof(struct sockaddr_in6);
1841 	fromsa.sin6_family = AF_INET6;
1842 	/* KAME hack: recover scopeid */
1843 	in6_recoverscope(&fromsa, &ip6->ip6_src, m->m_pkthdr.rcvif);
1844 
1845 	LIST_FOREACH(in6p, &ripcbinfo.pcblisthead, inp_list)
1846 	{
1847 		if (in6p->inp_flags & INP_PLACEMARKER)
1848 			continue;
1849 		if (!INP_ISIPV6(in6p))
1850 			continue;
1851 		if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1852 			continue;
1853 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1854 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1855 			continue;
1856 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1857 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1858 			continue;
1859 		if (in6p->in6p_icmp6filt
1860 		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1861 				 in6p->in6p_icmp6filt))
1862 			continue;
1863 		if (last) {
1864 			struct socket *so;
1865 			struct mbuf *n;
1866 
1867 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1868 				if (last->in6p_flags & IN6P_CONTROLOPTS)
1869 					ip6_savecontrol(last, &opts, ip6, n);
1870 				/* strip intermediate headers */
1871 				m_adj(n, off);
1872 				so = last->in6p_socket;
1873 				lwkt_gettoken(&so->so_rcv.ssb_token);
1874 				if (ssb_appendaddr(&so->so_rcv,
1875 				    (struct sockaddr *)&fromsa, n, opts) == 0) {
1876 					/* should notify about lost packet */
1877 					m_freem(n);
1878 					if (opts)
1879 						m_freem(opts);
1880 				} else {
1881 					sorwakeup(so);
1882 				}
1883 				lwkt_reltoken(&so->so_rcv.ssb_token);
1884 				opts = NULL;
1885 			}
1886 		}
1887 		last = in6p;
1888 	}
1889 	if (last) {
1890 		struct socket *so;
1891 
1892 		if (last->in6p_flags & IN6P_CONTROLOPTS)
1893 			ip6_savecontrol(last, &opts, ip6, m);
1894 		/* strip intermediate headers */
1895 		m_adj(m, off);
1896 		so = last->in6p_socket;
1897 		lwkt_gettoken(&so->so_rcv.ssb_token);
1898 		if (ssb_appendaddr(&so->so_rcv, (struct sockaddr *)&fromsa,
1899 		    m, opts) == 0) {
1900 			m_freem(m);
1901 			if (opts)
1902 				m_freem(opts);
1903 		} else {
1904 			sorwakeup(so);
1905 		}
1906 		lwkt_reltoken(&so->so_rcv.ssb_token);
1907 	} else {
1908 		m_freem(m);
1909 		ip6stat.ip6s_delivered--;
1910 	}
1911 	return IPPROTO_DONE;
1912 }
1913 
1914 /*
1915  * Reflect the ip6 packet back to the source.
1916  * OFF points to the icmp6 header, counted from the top of the mbuf.
1917  */
1918 void
1919 icmp6_reflect(struct mbuf *m, size_t off)
1920 {
1921 	struct ip6_hdr *ip6;
1922 	struct icmp6_hdr *icmp6;
1923 	struct in6_ifaddr *ia;
1924 	struct in6_addr t, *src = NULL;
1925 	int plen;
1926 	int type, code;
1927 	struct ifnet *outif = NULL;
1928 	struct sockaddr_in6 sa6_src, sa6_dst;
1929 #ifdef COMPAT_RFC1885
1930 	int mtu = IPV6_MMTU;
1931 	struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
1932 #endif
1933 
1934 	/* too short to reflect */
1935 	if (off < sizeof(struct ip6_hdr)) {
1936 		nd6log((LOG_DEBUG,
1937 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
1938 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
1939 		    __FILE__, __LINE__));
1940 		goto bad;
1941 	}
1942 
1943 	/*
1944 	 * If there are extra headers between IPv6 and ICMPv6, strip
1945 	 * off that header first.
1946 	 */
1947 #ifdef DIAGNOSTIC
1948 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
1949 		panic("assumption failed in icmp6_reflect");
1950 #endif
1951 	if (off > sizeof(struct ip6_hdr)) {
1952 		size_t l;
1953 		struct ip6_hdr nip6;
1954 
1955 		l = off - sizeof(struct ip6_hdr);
1956 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
1957 		m_adj(m, l);
1958 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1959 		if (m->m_len < l) {
1960 			if ((m = m_pullup(m, l)) == NULL)
1961 				return;
1962 		}
1963 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
1964 	} else /* off == sizeof(struct ip6_hdr) */ {
1965 		size_t l;
1966 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1967 		if (m->m_len < l) {
1968 			if ((m = m_pullup(m, l)) == NULL)
1969 				return;
1970 		}
1971 	}
1972 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
1973 	ip6 = mtod(m, struct ip6_hdr *);
1974 	ip6->ip6_nxt = IPPROTO_ICMPV6;
1975 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
1976 	type = icmp6->icmp6_type; /* keep type for statistics */
1977 	code = icmp6->icmp6_code; /* ditto. */
1978 
1979 	t = ip6->ip6_dst;
1980 	/*
1981 	 * ip6_input() drops a packet if its src is multicast.
1982 	 * So, the src is never multicast.
1983 	 */
1984 	ip6->ip6_dst = ip6->ip6_src;
1985 
1986 	/*
1987 	 * XXX: make sure to embed scope zone information, using
1988 	 * already embedded IDs or the received interface (if any).
1989 	 * Note that rcvif may be NULL.
1990 	 * TODO: scoped routing case (XXX).
1991 	 */
1992 	bzero(&sa6_src, sizeof(sa6_src));
1993 	sa6_src.sin6_family = AF_INET6;
1994 	sa6_src.sin6_len = sizeof(sa6_src);
1995 	sa6_src.sin6_addr = ip6->ip6_dst;
1996 	in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
1997 	in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL);
1998 	bzero(&sa6_dst, sizeof(sa6_dst));
1999 	sa6_dst.sin6_family = AF_INET6;
2000 	sa6_dst.sin6_len = sizeof(sa6_dst);
2001 	sa6_dst.sin6_addr = t;
2002 	in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
2003 	in6_embedscope(&t, &sa6_dst, NULL, NULL);
2004 
2005 #ifdef COMPAT_RFC1885
2006 	/*
2007 	 * xxx guess MTU
2008 	 * RFC 1885 requires that echo reply should be truncated if it
2009 	 * does not fit in with (return) path MTU, but the description was
2010 	 * removed in the new spec.
2011 	 */
2012 	if (icmp6_reflect_rt.ro_rt == 0 ||
2013 	    ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
2014 		if (icmp6_reflect_rt.ro_rt) {
2015 			RTFREE(icmp6_reflect_rt.ro_rt);
2016 			icmp6_reflect_rt.ro_rt = 0;
2017 		}
2018 		bzero(sin6, sizeof(*sin6));
2019 		sin6->sin6_family = PF_INET6;
2020 		sin6->sin6_len = sizeof(struct sockaddr_in6);
2021 		sin6->sin6_addr = ip6->ip6_dst;
2022 
2023 		rtalloc_ign((struct route *)&icmp6_reflect_rt.ro_rt,
2024 			    RTF_PRCLONING);
2025 	}
2026 
2027 	if (icmp6_reflect_rt.ro_rt == 0)
2028 		goto bad;
2029 
2030 	if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
2031 	    && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
2032 		mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
2033 
2034 	if (mtu < m->m_pkthdr.len) {
2035 		plen -= (m->m_pkthdr.len - mtu);
2036 		m_adj(m, mtu - m->m_pkthdr.len);
2037 	}
2038 #endif
2039 	/*
2040 	 * If the incoming packet was addressed directly to us(i.e. unicast),
2041 	 * use dst as the src for the reply.
2042 	 * The IN6_IFF_NOTREADY case would be VERY rare, but is possible
2043 	 * (for example) when we encounter an error while forwarding procedure
2044 	 * destined to a duplicated address of ours.
2045 	 */
2046 	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2047 		if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2048 		    (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2049 			src = &t;
2050 			break;
2051 		}
2052 	if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2053 		/*
2054 		 * This is the case if the dst is our link-local address
2055 		 * and the sender is also ourselves.
2056 		 */
2057 		src = &t;
2058 	}
2059 
2060 	if (src == NULL) {
2061 		int e;
2062 		struct route_in6 ro;
2063 
2064 		/*
2065 		 * This case matches to multicasts, our anycast, or unicasts
2066 		 * that we do not own.  Select a source address based on the
2067 		 * source address of the erroneous packet.
2068 		 */
2069 		bzero(&ro, sizeof(ro));
2070 		src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e, NULL);
2071 		if (ro.ro_rt)
2072 			RTFREE(ro.ro_rt); /* XXX: we could use this */
2073 		if (src == NULL) {
2074 			nd6log((LOG_DEBUG,
2075 			    "icmp6_reflect: source can't be determined: "
2076 			    "dst=%s, error=%d\n",
2077 			    ip6_sprintf(&sa6_src.sin6_addr), e));
2078 			goto bad;
2079 		}
2080 	}
2081 
2082 	ip6->ip6_src = *src;
2083 
2084 	ip6->ip6_flow = 0;
2085 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2086 	ip6->ip6_vfc |= IPV6_VERSION;
2087 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2088 	if (m->m_pkthdr.rcvif) {
2089 		/* XXX: This may not be the outgoing interface */
2090 		ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2091 		if (ip6->ip6_hlim < ip6_minhlim)
2092 			ip6->ip6_hlim = ip6_minhlim;
2093 	} else {
2094 		ip6->ip6_hlim = ip6_defhlim;
2095 	}
2096 
2097 	icmp6->icmp6_cksum = 0;
2098 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2099 					sizeof(struct ip6_hdr), plen);
2100 
2101 	/*
2102 	 * XXX option handling
2103 	 */
2104 
2105 	m->m_flags &= ~(M_BCAST|M_MCAST);
2106 
2107 #ifdef COMPAT_RFC1885
2108 	ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif, NULL);
2109 #else
2110 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2111 #endif
2112 	if (outif)
2113 		icmp6_ifoutstat_inc(outif, type, code);
2114 
2115 	return;
2116 
2117 bad:
2118 	m_freem(m);
2119 	return;
2120 }
2121 
2122 void
2123 icmp6_fasttimo(void)
2124 {
2125 
2126 	mld6_fasttimeo();
2127 }
2128 
2129 static const char *
2130 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2131 		    struct in6_addr *tgt6)
2132 {
2133 	static char buf[1024];
2134 	ksnprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2135 		ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2136 	return buf;
2137 }
2138 
2139 void
2140 icmp6_redirect_input(struct mbuf *m, int off)
2141 {
2142 	struct ifnet *ifp = m->m_pkthdr.rcvif;
2143 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2144 	struct nd_redirect *nd_rd;
2145 	int icmp6len = ntohs(ip6->ip6_plen);
2146 	char *lladdr = NULL;
2147 	int lladdrlen = 0;
2148 #if 0
2149 	u_char *redirhdr = NULL;
2150 	int redirhdrlen = 0;
2151 #endif
2152 	struct rtentry *rt = NULL;
2153 	int is_router;
2154 	int is_onlink;
2155 	struct in6_addr src6 = ip6->ip6_src;
2156 	struct in6_addr redtgt6;
2157 	struct in6_addr reddst6;
2158 	union nd_opts ndopts;
2159 
2160 	if (!m || !ifp)
2161 		return;
2162 
2163 	/* XXX if we are router, we don't update route by icmp6 redirect */
2164 	if (ip6_forwarding)
2165 		goto freeit;
2166 	if (!icmp6_rediraccept)
2167 		goto freeit;
2168 
2169 #ifndef PULLDOWN_TEST
2170 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2171 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2172 #else
2173 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2174 	if (nd_rd == NULL) {
2175 		icmp6stat.icp6s_tooshort++;
2176 		return;
2177 	}
2178 #endif
2179 	redtgt6 = nd_rd->nd_rd_target;
2180 	reddst6 = nd_rd->nd_rd_dst;
2181 
2182 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2183 		redtgt6.s6_addr16[1] = htons(ifp->if_index);
2184 	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2185 		reddst6.s6_addr16[1] = htons(ifp->if_index);
2186 
2187 	/* validation */
2188 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2189 		nd6log((LOG_ERR,
2190 			"ICMP6 redirect sent from %s rejected; "
2191 			"must be from linklocal\n", ip6_sprintf(&src6)));
2192 		goto bad;
2193 	}
2194 	if (ip6->ip6_hlim != 255) {
2195 		nd6log((LOG_ERR,
2196 			"ICMP6 redirect sent from %s rejected; "
2197 			"hlim=%d (must be 255)\n",
2198 			ip6_sprintf(&src6), ip6->ip6_hlim));
2199 		goto bad;
2200 	}
2201     {
2202 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2203 	struct sockaddr_in6 sin6;
2204 	struct in6_addr *gw6;
2205 
2206 	bzero(&sin6, sizeof(sin6));
2207 	sin6.sin6_family = AF_INET6;
2208 	sin6.sin6_len = sizeof(struct sockaddr_in6);
2209 	bcopy(&reddst6, &sin6.sin6_addr, sizeof reddst6);
2210 	rt = rtpurelookup((struct sockaddr *)&sin6);
2211 	if (rt != NULL) {
2212 		if (rt->rt_gateway == NULL ||
2213 		    rt->rt_gateway->sa_family != AF_INET6) {
2214 			nd6log((LOG_ERR,
2215 			    "ICMP6 redirect rejected; no route "
2216 			    "with inet6 gateway found for redirect dst: %s\n",
2217 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2218 			--rt->rt_refcnt;
2219 			goto bad;
2220 		}
2221 
2222 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2223 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2224 			nd6log((LOG_ERR,
2225 				"ICMP6 redirect rejected; "
2226 				"not equal to gw-for-src=%s (must be same): "
2227 				"%s\n",
2228 				ip6_sprintf(gw6),
2229 				icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2230 			--rt->rt_refcnt;
2231 			goto bad;
2232 		}
2233 	} else {
2234 		nd6log((LOG_ERR,
2235 			"ICMP6 redirect rejected; "
2236 			"no route found for redirect dst: %s\n",
2237 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2238 		goto bad;
2239 	}
2240 	--rt->rt_refcnt;
2241 	rt = NULL;
2242     }
2243 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2244 		nd6log((LOG_ERR,
2245 			"ICMP6 redirect rejected; "
2246 			"redirect dst must be unicast: %s\n",
2247 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2248 		goto bad;
2249 	}
2250 
2251 	is_router = is_onlink = 0;
2252 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2253 		is_router = 1;	/* router case */
2254 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2255 		is_onlink = 1;	/* on-link destination case */
2256 	if (!is_router && !is_onlink) {
2257 		nd6log((LOG_ERR,
2258 			"ICMP6 redirect rejected; "
2259 			"neither router case nor onlink case: %s\n",
2260 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2261 		goto bad;
2262 	}
2263 	/* validation passed */
2264 
2265 	icmp6len -= sizeof(*nd_rd);
2266 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2267 	if (nd6_options(&ndopts) < 0) {
2268 		nd6log((LOG_INFO, "icmp6_redirect_input: "
2269 			"invalid ND option, rejected: %s\n",
2270 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2271 		/* nd6_options have incremented stats */
2272 		goto freeit;
2273 	}
2274 
2275 	if (ndopts.nd_opts_tgt_lladdr) {
2276 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2277 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2278 	}
2279 
2280 #if 0
2281 	if (ndopts.nd_opts_rh) {
2282 		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2283 		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2284 	}
2285 #endif
2286 
2287 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2288 		nd6log((LOG_INFO,
2289 			"icmp6_redirect_input: lladdrlen mismatch for %s "
2290 			"(if %d, icmp6 packet %d): %s\n",
2291 			ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2292 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2293 		goto bad;
2294 	}
2295 
2296 	/* RFC 2461 8.3 */
2297 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2298 			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2299 
2300 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2301 		/* perform rtredirect */
2302 		struct sockaddr_in6 sdst;
2303 		struct sockaddr_in6 sgw;
2304 		struct sockaddr_in6 ssrc;
2305 
2306 		bzero(&sdst, sizeof(sdst));
2307 		bzero(&sgw, sizeof(sgw));
2308 		bzero(&ssrc, sizeof(ssrc));
2309 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2310 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2311 			sizeof(struct sockaddr_in6);
2312 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2313 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2314 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2315 		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2316 			   NULL, RTF_GATEWAY | RTF_HOST,
2317 			   (struct sockaddr *)&ssrc);
2318 	}
2319 	/* finally update cached route in each socket via kpfctlinput */
2320     {
2321 	struct sockaddr_in6 sdst;
2322 
2323 	bzero(&sdst, sizeof(sdst));
2324 	sdst.sin6_family = AF_INET6;
2325 	sdst.sin6_len = sizeof(struct sockaddr_in6);
2326 	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2327 	kpfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2328 #ifdef IPSEC
2329 	key_sa_routechange((struct sockaddr *)&sdst);
2330 #endif
2331     }
2332 
2333 freeit:
2334 	m_freem(m);
2335 	return;
2336 
2337 bad:
2338 	icmp6stat.icp6s_badredirect++;
2339 	m_freem(m);
2340 }
2341 
2342 void
2343 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2344 {
2345 	struct ifnet *ifp;	/* my outgoing interface */
2346 	struct in6_addr *ifp_ll6;
2347 	struct in6_addr *router_ll6;
2348 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2349 	struct mbuf *m = NULL;	/* newly allocated one */
2350 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2351 	struct nd_redirect *nd_rd;
2352 	size_t maxlen;
2353 	u_char *p;
2354 	struct ifnet *outif = NULL;
2355 	struct sockaddr_in6 src_sa;
2356 
2357 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2358 
2359 	/* if we are not router, we don't send icmp6 redirect */
2360 	if (!ip6_forwarding || ip6_accept_rtadv)
2361 		goto fail;
2362 
2363 	/* sanity check */
2364 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2365 		goto fail;
2366 
2367 	/*
2368 	 * Address check:
2369 	 *  the source address must identify a neighbor, and
2370 	 *  the destination address must not be a multicast address
2371 	 *  [RFC 2461, sec 8.2]
2372 	 */
2373 	sip6 = mtod(m0, struct ip6_hdr *);
2374 	bzero(&src_sa, sizeof(src_sa));
2375 	src_sa.sin6_family = AF_INET6;
2376 	src_sa.sin6_len = sizeof(src_sa);
2377 	src_sa.sin6_addr = sip6->ip6_src;
2378 	/* we don't currently use sin6_scope_id, but eventually use it */
2379 	if (in6_addr2zoneid(ifp, &sip6->ip6_src, &src_sa.sin6_scope_id))
2380 		goto fail;
2381 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2382 		goto fail;
2383 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2384 		goto fail;	/* what should we do here? */
2385 
2386 	/* rate limit */
2387 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2388 		goto fail;
2389 
2390 	/*
2391 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2392 	 * we almost always ask for an mbuf cluster for simplicity.
2393 	 * (MHLEN < IPV6_MMTU is almost always true)
2394 	 */
2395 #if IPV6_MMTU >= MCLBYTES
2396 # error assumption failed about IPV6_MMTU and MCLBYTES
2397 #endif
2398 	m = m_getb(IPV6_MMTU, M_NOWAIT, MT_HEADER, M_PKTHDR);
2399 	if (!m)
2400 		goto fail;
2401 	m->m_len = 0;
2402 	maxlen = M_TRAILINGSPACE(m);
2403 	maxlen = min(IPV6_MMTU, maxlen);
2404 	/* just for safety */
2405 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2406 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2407 		goto fail;
2408 	}
2409 
2410 	{
2411 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2412 		struct in6_ifaddr *ia;
2413 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2414 						 IN6_IFF_NOTREADY|
2415 						 IN6_IFF_ANYCAST)) == NULL)
2416 			goto fail;
2417 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2418 	}
2419 
2420 	/* get ip6 linklocal address for the router. */
2421 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2422 		struct sockaddr_in6 *sin6;
2423 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2424 		router_ll6 = &sin6->sin6_addr;
2425 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2426 			router_ll6 = NULL;
2427 	} else
2428 		router_ll6 = NULL;
2429 
2430 	/* ip6 */
2431 	ip6 = mtod(m, struct ip6_hdr *);
2432 	ip6->ip6_flow = 0;
2433 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2434 	ip6->ip6_vfc |= IPV6_VERSION;
2435 	/* ip6->ip6_plen will be set later */
2436 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2437 	ip6->ip6_hlim = 255;
2438 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2439 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2440 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2441 
2442 	/* ND Redirect */
2443 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2444 	nd_rd->nd_rd_type = ND_REDIRECT;
2445 	nd_rd->nd_rd_code = 0;
2446 	nd_rd->nd_rd_reserved = 0;
2447 	if (rt->rt_flags & RTF_GATEWAY) {
2448 		/*
2449 		 * nd_rd->nd_rd_target must be a link-local address in
2450 		 * better router cases.
2451 		 */
2452 		if (!router_ll6)
2453 			goto fail;
2454 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2455 		      sizeof(nd_rd->nd_rd_target));
2456 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2457 		      sizeof(nd_rd->nd_rd_dst));
2458 	} else {
2459 		/* make sure redtgt == reddst */
2460 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2461 		      sizeof(nd_rd->nd_rd_target));
2462 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2463 		      sizeof(nd_rd->nd_rd_dst));
2464 	}
2465 
2466 	p = (u_char *)(nd_rd + 1);
2467 
2468 	if (!router_ll6)
2469 		goto nolladdropt;
2470 
2471 	{
2472 		/* target lladdr option */
2473 		struct rtentry *rt_router = NULL;
2474 		int len;
2475 		struct sockaddr_dl *sdl;
2476 		struct nd_opt_hdr *nd_opt;
2477 		char *lladdr;
2478 
2479 		rt_router = nd6_lookup(router_ll6, 0, ifp);
2480 		if (!rt_router)
2481 			goto nolladdropt;
2482 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2483 		len = (len + 7) & ~7;	/* round by 8 */
2484 		/* safety check */
2485 		if (len + (p - (u_char *)ip6) > maxlen)
2486 			goto nolladdropt;
2487 		if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2488 		    (rt_router->rt_flags & RTF_LLINFO) &&
2489 		    (rt_router->rt_gateway->sa_family == AF_LINK) &&
2490 		    (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2491 		    sdl->sdl_alen) {
2492 			nd_opt = (struct nd_opt_hdr *)p;
2493 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2494 			nd_opt->nd_opt_len = len >> 3;
2495 			lladdr = (char *)(nd_opt + 1);
2496 			bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2497 			p += len;
2498 		}
2499 	}
2500 nolladdropt:;
2501 
2502 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2503 
2504 	/* just to be safe */
2505 #ifdef M_DECRYPTED	/*not openbsd*/
2506 	if (m0->m_flags & M_DECRYPTED)
2507 		goto noredhdropt;
2508 #endif
2509 	if (p - (u_char *)ip6 > maxlen)
2510 		goto noredhdropt;
2511 
2512     {
2513 	/* redirected header option */
2514 	int len;
2515 	struct nd_opt_rd_hdr *nd_opt_rh;
2516 
2517 	/*
2518 	 * compute the maximum size for icmp6 redirect header option.
2519 	 * XXX room for auth header?
2520 	 */
2521 	len = maxlen - (p - (u_char *)ip6);
2522 	len &= ~7;
2523 
2524 	/* This is just for simplicity. */
2525 	if (m0->m_pkthdr.len != m0->m_len) {
2526 		if (m0->m_next) {
2527 			m_freem(m0->m_next);
2528 			m0->m_next = NULL;
2529 		}
2530 		m0->m_pkthdr.len = m0->m_len;
2531 	}
2532 
2533 	/*
2534 	 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2535 	 * about padding/truncate rule for the original IP packet.
2536 	 * From the discussion on IPv6imp in Feb 1999, the consensus was:
2537 	 * - "attach as much as possible" is the goal
2538 	 * - pad if not aligned (original size can be guessed by original
2539 	 *   ip6 header)
2540 	 * Following code adds the padding if it is simple enough,
2541 	 * and truncates if not.
2542 	 */
2543 	if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2544 		panic("assumption failed in %s:%d", __FILE__, __LINE__);
2545 
2546 	if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2547 		/* not enough room, truncate */
2548 		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2549 	} else {
2550 		/* enough room, pad or truncate */
2551 		size_t extra;
2552 
2553 		extra = m0->m_pkthdr.len % 8;
2554 		if (extra) {
2555 			/* pad if easy enough, truncate if not */
2556 			if (8 - extra <= M_TRAILINGSPACE(m0)) {
2557 				/* pad */
2558 				m0->m_len += (8 - extra);
2559 				m0->m_pkthdr.len += (8 - extra);
2560 			} else {
2561 				/* truncate */
2562 				m0->m_pkthdr.len -= extra;
2563 				m0->m_len -= extra;
2564 			}
2565 		}
2566 		len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2567 		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2568 	}
2569 
2570 	nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2571 	bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2572 	nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2573 	nd_opt_rh->nd_opt_rh_len = len >> 3;
2574 	p += sizeof(*nd_opt_rh);
2575 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2576 
2577 	/* connect m0 to m */
2578 	m->m_next = m0;
2579 	m->m_pkthdr.len = m->m_len + m0->m_len;
2580 	m0 = NULL;
2581     }
2582 noredhdropt:;
2583 	if (m0 != NULL) {
2584 		m_freem(m0);
2585 		m0 = NULL;
2586 	}
2587 
2588 	/* XXX: clear embedded link IDs in the inner header */
2589 	in6_clearscope(&sip6->ip6_src);
2590 	in6_clearscope(&sip6->ip6_dst);
2591 	in6_clearscope(&nd_rd->nd_rd_target);
2592 	in6_clearscope(&nd_rd->nd_rd_dst);
2593 
2594 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2595 
2596 	nd_rd->nd_rd_cksum = 0;
2597 	nd_rd->nd_rd_cksum
2598 		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2599 
2600 	/* send the packet to outside... */
2601 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2602 	if (outif) {
2603 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2604 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2605 	}
2606 	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2607 
2608 	return;
2609 
2610 fail:
2611 	if (m)
2612 		m_freem(m);
2613 	if (m0)
2614 		m_freem(m0);
2615 }
2616 
2617 #ifdef HAVE_NRL_INPCB
2618 #define in6pcb		inpcb
2619 #define in6p_icmp6filt	inp_icmp6filt
2620 #endif
2621 /*
2622  * ICMPv6 socket option processing.
2623  */
2624 void
2625 icmp6_ctloutput(netmsg_t msg)
2626 {
2627 	struct socket *so = msg->ctloutput.base.nm_so;
2628 	struct sockopt *sopt = msg->ctloutput.nm_sopt;
2629 	int error = 0;
2630 	int optlen;
2631 	struct inpcb *inp = so->so_pcb;
2632 	int level, op, optname;
2633 
2634 	if (sopt) {
2635 		level = sopt->sopt_level;
2636 		op = sopt->sopt_dir;
2637 		optname = sopt->sopt_name;
2638 		optlen = sopt->sopt_valsize;
2639 	} else
2640 		level = op = optname = optlen = 0;
2641 
2642 	if (level != IPPROTO_ICMPV6) {
2643 		error = EINVAL;
2644 		goto out;
2645 	}
2646 
2647 	switch (op) {
2648 	case PRCO_SETOPT:
2649 		switch (optname) {
2650 		case ICMP6_FILTER:
2651 		    {
2652 			if (optlen != sizeof(struct icmp6_filter)) {
2653 				error = EMSGSIZE;
2654 				break;
2655 			}
2656 			if (inp->in6p_icmp6filt == NULL) {
2657 				error = EINVAL;
2658 				break;
2659 			}
2660 			error = soopt_to_kbuf(sopt, inp->in6p_icmp6filt, optlen,
2661 				optlen);
2662 			break;
2663 		    }
2664 
2665 		default:
2666 			error = ENOPROTOOPT;
2667 			break;
2668 		}
2669 		break;
2670 
2671 	case PRCO_GETOPT:
2672 		switch (optname) {
2673 		case ICMP6_FILTER:
2674 		    {
2675 			if (inp->in6p_icmp6filt == NULL) {
2676 				error = EINVAL;
2677 				break;
2678 			}
2679 			soopt_from_kbuf(sopt, inp->in6p_icmp6filt,
2680 				sizeof(struct icmp6_filter));
2681 			break;
2682 		    }
2683 
2684 		default:
2685 			error = ENOPROTOOPT;
2686 			break;
2687 		}
2688 		break;
2689 	}
2690 out:
2691 	lwkt_replymsg(&msg->ctloutput.base.lmsg, error);
2692 }
2693 #ifdef HAVE_NRL_INPCB
2694 #undef in6pcb
2695 #undef in6p_icmp6filt
2696 #endif
2697 
2698 #ifndef HAVE_PPSRATECHECK
2699 #ifndef timersub
2700 #define	timersub(tvp, uvp, vvp)						\
2701 	do {								\
2702 		(vvp)->tv_sec = (tvp)->tv_sec - (uvp)->tv_sec;		\
2703 		(vvp)->tv_usec = (tvp)->tv_usec - (uvp)->tv_usec;	\
2704 		if ((vvp)->tv_usec < 0) {				\
2705 			(vvp)->tv_sec--;				\
2706 			(vvp)->tv_usec += 1000000;			\
2707 		}							\
2708 	} while (0)
2709 #endif
2710 
2711 /*
2712  * ppsratecheck(): packets (or events) per second limitation.
2713  */
2714 static int
2715 ppsratecheck(struct timeval *lasttime, int *curpps,
2716 	     int maxpps)	/* maximum pps allowed */
2717 {
2718 	struct timeval tv, delta;
2719 	int rv;
2720 
2721 	microtime(&tv);
2722 
2723 	timersub(&tv, lasttime, &delta);
2724 
2725 	/*
2726 	 * Check for 0,0 so that the message will be seen at least once.
2727 	 * If more than one second has passed since the last update of
2728 	 * lasttime, reset the counter.
2729 	 *
2730 	 * We do increment *curpps even in *curpps < maxpps case, as some may
2731 	 * try to use *curpps for stat purposes as well.
2732 	 */
2733 	if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) ||
2734 	    delta.tv_sec >= 1) {
2735 		*lasttime = tv;
2736 		*curpps = 0;
2737 		rv = 1;
2738 	} else if (maxpps < 0)
2739 		rv = 1;
2740 	else if (*curpps < maxpps)
2741 		rv = 1;
2742 	else
2743 		rv = 0;
2744 
2745 #if 1 /* DIAGNOSTIC? */
2746 	/* be careful about wrap-around */
2747 	if (*curpps + 1 > *curpps)
2748 		*curpps = *curpps + 1;
2749 #else
2750 	/*
2751 	 * assume that there's not too many calls to this function.
2752 	 * not sure if the assumption holds, as it depends on *caller's*
2753 	 * behavior, not the behavior of this function.
2754 	 * IMHO it is wrong to make assumption on the caller's behavior,
2755 	 * so the above #if is #if 1, not #ifdef DIAGNOSTIC.
2756 	 */
2757 	*curpps = *curpps + 1;
2758 #endif
2759 
2760 	return (rv);
2761 }
2762 #endif
2763 
2764 /*
2765  * Perform rate limit check.
2766  * Returns 0 if it is okay to send the icmp6 packet.
2767  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2768  * limitation.
2769  *
2770  * XXX per-destination/type check necessary?
2771  */
2772 static int
2773 icmp6_ratelimit(const struct in6_addr *dst,	/* not used at this moment */
2774 		const int type,			/* not used at this moment */
2775 		const int code)			/* not used at this moment */
2776 {
2777 	int ret;
2778 
2779 	ret = 0;	/* okay to send */
2780 
2781 	/* PPS limit */
2782 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2783 	    icmp6errppslim)) {
2784 		/* The packet is subject to rate limit */
2785 		ret++;
2786 	}
2787 
2788 	return ret;
2789 }
2790