xref: /dragonfly/sys/netinet6/icmp6.c (revision 5503793e)
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 #include <net/netisr2.h>
89 #include <net/netmsg2.h>
90 
91 #include <netinet/in.h>
92 #include <netinet/in_var.h>
93 #include <netinet/ip6.h>
94 #include <netinet6/ip6_var.h>
95 #include <netinet/icmp6.h>
96 #include <netinet6/mld6_var.h>
97 #include <netinet/in_pcb.h>
98 #include <netinet6/in6_pcb.h>
99 #include <netinet6/nd6.h>
100 #include <netinet6/in6_ifattach.h>
101 #include <netinet6/ip6protosw.h>
102 
103 #ifdef IPSEC
104 #include <netinet6/ipsec.h>
105 #include <netproto/key/key.h>
106 #endif
107 
108 #ifdef FAST_IPSEC
109 #include <netproto/ipsec/ipsec.h>
110 #include <netproto/ipsec/key.h>
111 #define	IPSEC
112 #endif
113 
114 #include <net/net_osdep.h>
115 
116 #ifdef HAVE_NRL_INPCB
117 /* inpcb members */
118 #define in6pcb		inpcb
119 #define in6p_laddr	inp_laddr6
120 #define in6p_faddr	inp_faddr6
121 #define in6p_icmp6filt	inp_icmp6filt
122 #define in6p_route	inp_route
123 #define in6p_socket	inp_socket
124 #define in6p_flags	inp_flags
125 #define in6p_moptions	inp_moptions6
126 #define in6p_outputopts	inp_outputopts6
127 #define in6p_ip6	inp_ipv6
128 #define in6p_flowinfo	inp_flowinfo
129 #define in6p_sp		inp_sp
130 #define in6p_next	inp_next
131 #define in6p_prev	inp_prev
132 /* function names */
133 #define in6_pcbdetach	in_pcbdetach
134 #define in6_rtchange	in_rtchange
135 
136 /*
137  * for KAME src sync over BSD*'s. XXX: FreeBSD (>=3) are VERY different from
138  * others...
139  */
140 #define in6p_ip6_nxt	inp_ipv6.ip6_nxt
141 #endif
142 
143 #define ICMP6_FASTTIMO	(hz / PR_FASTHZ)
144 
145 /*
146  * ppratecheck() is available, so define it here.
147  */
148 #define	HAVE_PPSRATECHECK
149 
150 extern struct domain inet6domain;
151 extern struct protosw inet6sw[];
152 extern u_char ip6_protox[];
153 
154 struct icmp6stat icmp6stat;
155 
156 extern struct inpcbinfo ripcbinfo;
157 extern int icmp6errppslim;
158 static int icmp6errpps_count = 0;
159 static struct timeval icmp6errppslim_last;
160 extern int icmp6_nodeinfo;
161 
162 static void icmp6_errcount (struct icmp6errstat *, int, int);
163 static int icmp6_rip6_input (struct mbuf **, int);
164 static int icmp6_ratelimit (const struct in6_addr *, const int, const int);
165 static const char *icmp6_redirect_diag (struct in6_addr *,
166 	struct in6_addr *, struct in6_addr *);
167 #ifndef HAVE_PPSRATECHECK
168 static int ppsratecheck (struct timeval *, int *, int);
169 #endif
170 static struct mbuf *ni6_input (struct mbuf *, int);
171 static struct mbuf *ni6_nametodns (const char *, int, int);
172 static int ni6_dnsmatch (const char *, int, const char *, int);
173 static int ni6_addrs (struct icmp6_nodeinfo *, struct mbuf *,
174 			  struct ifnet **, char *);
175 static int ni6_store_addrs (struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
176 				struct ifnet *, int);
177 static int icmp6_notify_error (struct mbuf *, int, int, int);
178 
179 static void			icmp6_fasttimo(void *);
180 static void			icmp6_fasttimo_dispatch(netmsg_t);
181 
182 #ifdef COMPAT_RFC1885
183 static struct route_in6 icmp6_reflect_rt;
184 #endif
185 
186 static struct callout		icmp6_fasttimo_ch;
187 static struct netmsg_base	icmp6_fasttimo_nmsg;
188 
189 void
190 icmp6_init(void)
191 {
192 
193 	mld6_init();
194 
195 	callout_init_mp(&icmp6_fasttimo_ch);
196 	netmsg_init(&icmp6_fasttimo_nmsg, NULL, &netisr_adone_rport,
197 	    MSGF_PRIORITY, icmp6_fasttimo_dispatch);
198 
199 	callout_reset_bycpu(&icmp6_fasttimo_ch, ICMP6_FASTTIMO,
200 	    icmp6_fasttimo, NULL, 0);
201 }
202 
203 static void
204 icmp6_errcount(struct icmp6errstat *stat, int type, int code)
205 {
206 	switch (type) {
207 	case ICMP6_DST_UNREACH:
208 		switch (code) {
209 		case ICMP6_DST_UNREACH_NOROUTE:
210 			stat->icp6errs_dst_unreach_noroute++;
211 			return;
212 		case ICMP6_DST_UNREACH_ADMIN:
213 			stat->icp6errs_dst_unreach_admin++;
214 			return;
215 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
216 			stat->icp6errs_dst_unreach_beyondscope++;
217 			return;
218 		case ICMP6_DST_UNREACH_ADDR:
219 			stat->icp6errs_dst_unreach_addr++;
220 			return;
221 		case ICMP6_DST_UNREACH_NOPORT:
222 			stat->icp6errs_dst_unreach_noport++;
223 			return;
224 		}
225 		break;
226 	case ICMP6_PACKET_TOO_BIG:
227 		stat->icp6errs_packet_too_big++;
228 		return;
229 	case ICMP6_TIME_EXCEEDED:
230 		switch (code) {
231 		case ICMP6_TIME_EXCEED_TRANSIT:
232 			stat->icp6errs_time_exceed_transit++;
233 			return;
234 		case ICMP6_TIME_EXCEED_REASSEMBLY:
235 			stat->icp6errs_time_exceed_reassembly++;
236 			return;
237 		}
238 		break;
239 	case ICMP6_PARAM_PROB:
240 		switch (code) {
241 		case ICMP6_PARAMPROB_HEADER:
242 			stat->icp6errs_paramprob_header++;
243 			return;
244 		case ICMP6_PARAMPROB_NEXTHEADER:
245 			stat->icp6errs_paramprob_nextheader++;
246 			return;
247 		case ICMP6_PARAMPROB_OPTION:
248 			stat->icp6errs_paramprob_option++;
249 			return;
250 		}
251 		break;
252 	case ND_REDIRECT:
253 		stat->icp6errs_redirect++;
254 		return;
255 	}
256 	stat->icp6errs_unknown++;
257 }
258 
259 /*
260  * Generate an error packet of type error in response to bad IP6 packet.
261  */
262 void
263 icmp6_error(struct mbuf *m, int type, int code, int param)
264 {
265 	struct ip6_hdr *oip6, *nip6;
266 	struct icmp6_hdr *icmp6;
267 	u_int preplen;
268 	int off;
269 	int nxt;
270 
271 	icmp6stat.icp6s_error++;
272 
273 	/* count per-type-code statistics */
274 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
275 
276 #ifdef M_DECRYPTED	/*not openbsd*/
277 	if (m->m_flags & M_DECRYPTED) {
278 		icmp6stat.icp6s_canterror++;
279 		goto freeit;
280 	}
281 #endif
282 
283 #ifndef PULLDOWN_TEST
284 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
285 #else
286 	if (m->m_len < sizeof(struct ip6_hdr)) {
287 		m = m_pullup(m, sizeof(struct ip6_hdr));
288 		if (m == NULL)
289 			return;
290 	}
291 #endif
292 	oip6 = mtod(m, struct ip6_hdr *);
293 
294 	/*
295 	 * Multicast destination check. For unrecognized option errors,
296 	 * this check has already done in ip6_unknown_opt(), so we can
297 	 * check only for other errors.
298 	 */
299 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
300 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
301 	    (type != ICMP6_PACKET_TOO_BIG &&
302 	     (type != ICMP6_PARAM_PROB ||
303 	      code != ICMP6_PARAMPROB_OPTION)))
304 		goto freeit;
305 
306 	/* Source address check. XXX: the case of anycast source? */
307 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
308 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
309 		goto freeit;
310 
311 	/*
312 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
313 	 * don't do it.
314 	 */
315 	nxt = -1;
316 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
317 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
318 		struct icmp6_hdr *icp;
319 
320 #ifndef PULLDOWN_TEST
321 		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
322 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
323 #else
324 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
325 			sizeof(*icp));
326 		if (icp == NULL) {
327 			icmp6stat.icp6s_tooshort++;
328 			return;
329 		}
330 #endif
331 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
332 		    icp->icmp6_type == ND_REDIRECT) {
333 			/*
334 			 * ICMPv6 error
335 			 * Special case: for redirect (which is
336 			 * informational) we must not send icmp6 error.
337 			 */
338 			icmp6stat.icp6s_canterror++;
339 			goto freeit;
340 		} else {
341 			/* ICMPv6 informational - send the error */
342 		}
343 	} else {
344 		/* non-ICMPv6 - send the error */
345 	}
346 
347 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
348 
349 	/* Finally, do rate limitation check. */
350 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
351 		icmp6stat.icp6s_toofreq++;
352 		goto freeit;
353 	}
354 
355 	/*
356 	 * OK, ICMP6 can be generated.
357 	 */
358 
359 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
360 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
361 
362 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
363 	M_PREPEND(m, preplen, M_NOWAIT);
364 	if (m && m->m_len < preplen)
365 		m = m_pullup(m, preplen);
366 	if (m == NULL) {
367 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
368 		return;
369 	}
370 
371 	nip6 = mtod(m, struct ip6_hdr *);
372 	nip6->ip6_src  = oip6->ip6_src;
373 	nip6->ip6_dst  = oip6->ip6_dst;
374 
375 	in6_clearscope(&oip6->ip6_src);
376 	in6_clearscope(&oip6->ip6_dst);
377 
378 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
379 	icmp6->icmp6_type = type;
380 	icmp6->icmp6_code = code;
381 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
382 
383 	/*
384 	 * icmp6_reflect() is designed to be in the input path.
385 	 * icmp6_error() can be called from both input and outut path,
386 	 * and if we are in output path rcvif could contain bogus value.
387 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
388 	 * information in ip header (nip6).
389 	 */
390 	m->m_pkthdr.rcvif = NULL;
391 
392 	icmp6stat.icp6s_outhist[type]++;
393 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
394 
395 	return;
396 
397 freeit:
398 	/*
399 	 * If we can't tell wheter or not we can generate ICMP6, free it.
400 	 */
401 	m_freem(m);
402 }
403 
404 /*
405  * Process a received ICMP6 message.
406  */
407 int
408 icmp6_input(struct mbuf **mp, int *offp, int proto)
409 {
410 	struct mbuf *m = *mp, *n;
411 	struct ip6_hdr *ip6, *nip6;
412 	struct icmp6_hdr *icmp6, *nicmp6;
413 	int off = *offp;
414 	int icmp6len = m->m_pkthdr.len - *offp;
415 	int code, sum, noff;
416 
417 #ifndef PULLDOWN_TEST
418 	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
419 	/* m might change if M_LOOP.  So, call mtod after this */
420 #endif
421 
422 	/*
423 	 * Locate icmp6 structure in mbuf, and check
424 	 * that not corrupted and of at least minimum length
425 	 */
426 
427 	ip6 = mtod(m, struct ip6_hdr *);
428 	if (icmp6len < sizeof(struct icmp6_hdr)) {
429 		icmp6stat.icp6s_tooshort++;
430 		goto freeit;
431 	}
432 
433 	/*
434 	 * calculate the checksum
435 	 */
436 #ifndef PULLDOWN_TEST
437 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
438 #else
439 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
440 	if (icmp6 == NULL) {
441 		icmp6stat.icp6s_tooshort++;
442 		return IPPROTO_DONE;
443 	}
444 #endif
445 	code = icmp6->icmp6_code;
446 
447 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
448 		nd6log((LOG_ERR,
449 		    "ICMP6 checksum error(%d|%x) %s\n",
450 		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
451 		icmp6stat.icp6s_checksum++;
452 		goto freeit;
453 	}
454 
455 	icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
456 	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
457 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
458 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
459 
460 	switch (icmp6->icmp6_type) {
461 	case ICMP6_DST_UNREACH:
462 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
463 		switch (code) {
464 		case ICMP6_DST_UNREACH_NOROUTE:
465 			code = PRC_UNREACH_NET;
466 			break;
467 		case ICMP6_DST_UNREACH_ADMIN:
468 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
469 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
470 			break;
471 		case ICMP6_DST_UNREACH_ADDR:
472 			code = PRC_HOSTDEAD;
473 			break;
474 #ifdef COMPAT_RFC1885
475 		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
476 			code = PRC_UNREACH_SRCFAIL;
477 			break;
478 #else
479 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
480 			/* I mean "source address was incorrect." */
481 			code = PRC_PARAMPROB;
482 			break;
483 #endif
484 		case ICMP6_DST_UNREACH_NOPORT:
485 			code = PRC_UNREACH_PORT;
486 			break;
487 		default:
488 			goto badcode;
489 		}
490 		goto deliver;
491 		break;
492 
493 	case ICMP6_PACKET_TOO_BIG:
494 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
495 		if (code != 0)
496 			goto badcode;
497 
498 		code = PRC_MSGSIZE;
499 
500 		/*
501 		 * Updating the path MTU will be done after examining
502 		 * intermediate extension headers.
503 		 */
504 		goto deliver;
505 		break;
506 
507 	case ICMP6_TIME_EXCEEDED:
508 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
509 		switch (code) {
510 		case ICMP6_TIME_EXCEED_TRANSIT:
511 		case ICMP6_TIME_EXCEED_REASSEMBLY:
512 			code += PRC_TIMXCEED_INTRANS;
513 			break;
514 		default:
515 			goto badcode;
516 		}
517 		goto deliver;
518 		break;
519 
520 	case ICMP6_PARAM_PROB:
521 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
522 		switch (code) {
523 		case ICMP6_PARAMPROB_NEXTHEADER:
524 			code = PRC_UNREACH_PROTOCOL;
525 			break;
526 		case ICMP6_PARAMPROB_HEADER:
527 		case ICMP6_PARAMPROB_OPTION:
528 			code = PRC_PARAMPROB;
529 			break;
530 		default:
531 			goto badcode;
532 		}
533 		goto deliver;
534 		break;
535 
536 	case ICMP6_ECHO_REQUEST:
537 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
538 		if (code != 0)
539 			goto badcode;
540 		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
541 			/* Give up remote */
542 			break;
543 		}
544 		if ((n->m_flags & M_EXT) ||
545 		    n->m_len < off + sizeof(struct icmp6_hdr)) {
546 			struct mbuf *n0 = n;
547 			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
548 			int n0len;
549 
550 			/*
551 			 * Prepare an internal mbuf. m_pullup() doesn't
552 			 * always copy the length we specified.
553 			 */
554 			if (maxlen >= MCLBYTES) {
555 				/* Give up remote */
556 				m_freem(n0);
557 				break;
558 			}
559 			n = m_getb(maxlen, M_NOWAIT, n0->m_type, M_PKTHDR);
560 			if (n == NULL) {
561 				/* Give up remote */
562 				m_freem(n0);
563 				break;
564 			}
565 			n0len = n0->m_pkthdr.len;	/* save for use below */
566 			M_MOVE_PKTHDR(n, n0);
567 			/*
568 			 * Copy IPv6 and ICMPv6 only.
569 			 */
570 			nip6 = mtod(n, struct ip6_hdr *);
571 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
572 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
573 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
574 			noff = sizeof(struct ip6_hdr);
575 			/* new mbuf contains only ipv6+icmpv6 headers */
576 			n->m_len = noff + sizeof(struct icmp6_hdr);
577 			/*
578 			 * Adjust mbuf. ip6_plen will be adjusted in
579 			 * ip6_output().
580 			 */
581 			m_adj(n0, off + sizeof(struct icmp6_hdr));
582 			/* recalculate complete packet size */
583 			n->m_pkthdr.len = n0len + (noff - off);
584 			n->m_next = n0;
585 		} else {
586 			nip6 = mtod(n, struct ip6_hdr *);
587 			nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
588 			noff = off;
589 		}
590 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
591 		nicmp6->icmp6_code = 0;
592 		if (n) {
593 			icmp6stat.icp6s_reflect++;
594 			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
595 			icmp6_reflect(n, noff);
596 		}
597 		break;
598 
599 	case ICMP6_ECHO_REPLY:
600 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
601 		if (code != 0)
602 			goto badcode;
603 		break;
604 
605 	case MLD_LISTENER_QUERY:
606 	case MLD_LISTENER_REPORT:
607 		if (icmp6len < sizeof(struct mld_hdr))
608 			goto badlen;
609 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
610 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
611 		else
612 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
613 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
614 			/* give up local */
615 			mld6_input(m, off);
616 			m = NULL;
617 			goto freeit;
618 		}
619 		mld6_input(n, off);
620 		/* m stays. */
621 		break;
622 
623 	case MLD_LISTENER_DONE:
624 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
625 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
626 			goto badlen;
627 		break;		/* nothing to be done in kernel */
628 
629 	case MLD_MTRACE_RESP:
630 	case MLD_MTRACE:
631 		/* XXX: these two are experimental.  not officially defind. */
632 		/* XXX: per-interface statistics? */
633 		break;		/* just pass it to applications */
634 
635 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
636 	    {
637 		enum { WRU, FQDN } mode;
638 
639 		if (!icmp6_nodeinfo)
640 			break;
641 
642 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
643 			mode = WRU;
644 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
645 			mode = FQDN;
646 		else
647 			goto badlen;
648 
649 #define hostnamelen	strlen(hostname)
650 		if (mode == FQDN) {
651 #ifndef PULLDOWN_TEST
652 			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
653 					 IPPROTO_DONE);
654 #endif
655 			n = m_copy(m, 0, M_COPYALL);
656 			if (n)
657 				n = ni6_input(n, off);
658 			/* XXX meaningless if n == NULL */
659 			noff = sizeof(struct ip6_hdr);
660 		} else {
661 			u_char *p;
662 			int maxlen, maxhlen;
663 
664 			if ((icmp6_nodeinfo & 5) != 5)
665 				break;
666 
667 			if (code != 0)
668 				goto badcode;
669 			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
670 			if (maxlen >= MCLBYTES) {
671 				/* Give up remote */
672 				break;
673 			}
674 			n = m_getb(maxlen, M_NOWAIT, m->m_type, M_PKTHDR);
675 			if (n == NULL) {
676 				/* Give up remote */
677 				break;
678 			}
679 			if (!m_dup_pkthdr(n, m, M_NOWAIT)) {
680 				/*
681 				 * Previous code did a blind M_COPY_PKTHDR
682 				 * and said "just for rcvif".  If true, then
683 				 * we could tolerate the dup failing (due to
684 				 * the deep copy of the tag chain).  For now
685 				 * be conservative and just fail.
686 				 */
687 				m_free(n);
688 				break;
689 			}
690 			n->m_len = 0;
691 			maxhlen = M_TRAILINGSPACE(n) - maxlen;
692 			if (maxhlen > hostnamelen)
693 				maxhlen = hostnamelen;
694 			/*
695 			 * Copy IPv6 and ICMPv6 only.
696 			 */
697 			nip6 = mtod(n, struct ip6_hdr *);
698 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
699 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
700 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
701 			p = (u_char *)(nicmp6 + 1);
702 			bzero(p, 4);
703 			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
704 			noff = sizeof(struct ip6_hdr);
705 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
706 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
707 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
708 			nicmp6->icmp6_code = 0;
709 		}
710 #undef hostnamelen
711 		if (n) {
712 			icmp6stat.icp6s_reflect++;
713 			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
714 			icmp6_reflect(n, noff);
715 		}
716 		break;
717 	    }
718 
719 	case ICMP6_WRUREPLY:
720 		if (code != 0)
721 			goto badcode;
722 		break;
723 
724 	case ND_ROUTER_SOLICIT:
725 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
726 		if (code != 0)
727 			goto badcode;
728 		if (icmp6len < sizeof(struct nd_router_solicit))
729 			goto badlen;
730 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
731 			/* give up local */
732 			nd6_rs_input(m, off, icmp6len);
733 			m = NULL;
734 			goto freeit;
735 		}
736 		nd6_rs_input(n, off, icmp6len);
737 		/* m stays. */
738 		break;
739 
740 	case ND_ROUTER_ADVERT:
741 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
742 		if (code != 0)
743 			goto badcode;
744 		if (icmp6len < sizeof(struct nd_router_advert))
745 			goto badlen;
746 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
747 			/* give up local */
748 			nd6_ra_input(m, off, icmp6len);
749 			m = NULL;
750 			goto freeit;
751 		}
752 		nd6_ra_input(n, off, icmp6len);
753 		/* m stays. */
754 		break;
755 
756 	case ND_NEIGHBOR_SOLICIT:
757 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
758 		if (code != 0)
759 			goto badcode;
760 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
761 			goto badlen;
762 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
763 			/* give up local */
764 			nd6_ns_input(m, off, icmp6len);
765 			m = NULL;
766 			goto freeit;
767 		}
768 		nd6_ns_input(n, off, icmp6len);
769 		/* m stays. */
770 		break;
771 
772 	case ND_NEIGHBOR_ADVERT:
773 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
774 		if (code != 0)
775 			goto badcode;
776 		if (icmp6len < sizeof(struct nd_neighbor_advert))
777 			goto badlen;
778 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
779 			/* give up local */
780 			nd6_na_input(m, off, icmp6len);
781 			m = NULL;
782 			goto freeit;
783 		}
784 		nd6_na_input(n, off, icmp6len);
785 		/* m stays. */
786 		break;
787 
788 	case ND_REDIRECT:
789 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
790 		if (code != 0)
791 			goto badcode;
792 		if (icmp6len < sizeof(struct nd_redirect))
793 			goto badlen;
794 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
795 			/* give up local */
796 			icmp6_redirect_input(m, off);
797 			m = NULL;
798 			goto freeit;
799 		}
800 		icmp6_redirect_input(n, off);
801 		/* m stays. */
802 		break;
803 
804 	case ICMP6_ROUTER_RENUMBERING:
805 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
806 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
807 			goto badcode;
808 		if (icmp6len < sizeof(struct icmp6_router_renum))
809 			goto badlen;
810 		break;
811 
812 	default:
813 		nd6log((LOG_DEBUG,
814 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
815 		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
816 		    ip6_sprintf(&ip6->ip6_dst),
817 		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
818 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
819 			/* ICMPv6 error: MUST deliver it by spec... */
820 			code = PRC_NCMDS;
821 			/* deliver */
822 		} else {
823 			/* ICMPv6 informational: MUST not deliver */
824 			break;
825 		}
826 	deliver:
827 		if (icmp6_notify_error(m, off, icmp6len, code)) {
828 			/* In this case, m should've been freed. */
829 			return (IPPROTO_DONE);
830 		}
831 		break;
832 
833 	badcode:
834 		icmp6stat.icp6s_badcode++;
835 		break;
836 
837 	badlen:
838 		icmp6stat.icp6s_badlen++;
839 		break;
840 	}
841 
842 	/* deliver the packet to appropriate sockets */
843 	icmp6_rip6_input(&m, *offp);
844 
845 	return IPPROTO_DONE;
846 
847 freeit:
848 	m_freem(m);
849 	return IPPROTO_DONE;
850 }
851 
852 static int
853 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
854 {
855 	struct icmp6_hdr *icmp6;
856 	struct ip6_hdr *eip6;
857 	u_int32_t notifymtu;
858 	struct sockaddr_in6 icmp6src, icmp6dst;
859 
860 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
861 		icmp6stat.icp6s_tooshort++;
862 		goto freeit;
863 	}
864 #ifndef PULLDOWN_TEST
865 	IP6_EXTHDR_CHECK(m, off,
866 			 sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr),
867 			 -1);
868 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
869 #else
870 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
871 		       sizeof(*icmp6) + sizeof(struct ip6_hdr));
872 	if (icmp6 == NULL) {
873 		icmp6stat.icp6s_tooshort++;
874 		return (-1);
875 	}
876 #endif
877 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
878 
879 	/* Detect the upper level protocol */
880 	{
881 		u_int8_t nxt = eip6->ip6_nxt;
882 		int eoff = off + sizeof(struct icmp6_hdr) +
883 			sizeof(struct ip6_hdr);
884 		struct ip6ctlparam ip6cp;
885 		struct in6_addr *finaldst = NULL;
886 		int icmp6type = icmp6->icmp6_type;
887 		struct ip6_frag *fh;
888 		struct ip6_rthdr *rth;
889 		int rthlen;
890 
891 		while (1) { /* XXX: should avoid infinite loop explicitly? */
892 			struct ip6_ext *eh;
893 
894 			switch (nxt) {
895 			case IPPROTO_HOPOPTS:
896 			case IPPROTO_DSTOPTS:
897 			case IPPROTO_AH:
898 #ifndef PULLDOWN_TEST
899 				IP6_EXTHDR_CHECK(m, 0,
900 				    eoff + sizeof(struct ip6_ext), -1);
901 				eh = (struct ip6_ext *)
902 				    (mtod(m, caddr_t) + eoff);
903 #else
904 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
905 					       eoff, sizeof(*eh));
906 				if (eh == NULL) {
907 					icmp6stat.icp6s_tooshort++;
908 					return (-1);
909 				}
910 #endif
911 
912 				if (nxt == IPPROTO_AH)
913 					eoff += (eh->ip6e_len + 2) << 2;
914 				else
915 					eoff += (eh->ip6e_len + 1) << 3;
916 				nxt = eh->ip6e_nxt;
917 				break;
918 			case IPPROTO_ROUTING:
919 				/*
920 				 * When the erroneous packet contains a
921 				 * routing header, we should examine the
922 				 * header to determine the final destination.
923 				 * Otherwise, we can't properly update
924 				 * information that depends on the final
925 				 * destination (e.g. path MTU).
926 				 */
927 #ifndef PULLDOWN_TEST
928 				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth),
929 						 -1);
930 				rth = (struct ip6_rthdr *)(mtod(m, caddr_t)
931 							   + eoff);
932 #else
933 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
934 					       eoff, sizeof(*rth));
935 				if (rth == NULL) {
936 					icmp6stat.icp6s_tooshort++;
937 					return (-1);
938 				}
939 #endif
940 				rthlen = (rth->ip6r_len + 1) << 3;
941 				eoff += rthlen;
942 				nxt = rth->ip6r_nxt;
943 				break;
944 			case IPPROTO_FRAGMENT:
945 #ifndef PULLDOWN_TEST
946 				IP6_EXTHDR_CHECK(m, 0, eoff +
947 						 sizeof(struct ip6_frag),
948 						 -1);
949 				fh = (struct ip6_frag *)(mtod(m, caddr_t)
950 							 + eoff);
951 #else
952 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
953 					       eoff, sizeof(*fh));
954 				if (fh == NULL) {
955 					icmp6stat.icp6s_tooshort++;
956 					return (-1);
957 				}
958 #endif
959 				/*
960 				 * Data after a fragment header is meaningless
961 				 * unless it is the first fragment, but
962 				 * we'll go to the notify label for path MTU
963 				 * discovery.
964 				 */
965 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
966 					goto notify;
967 
968 				eoff += sizeof(struct ip6_frag);
969 				nxt = fh->ip6f_nxt;
970 				break;
971 			default:
972 				/*
973 				 * This case includes ESP and the No Next
974 				 * Header.  In such cases going to the notify
975 				 * label does not have any meaning
976 				 * (i.e. pr_ctlinput will be NULL), but we go
977 				 * anyway since we might have to update
978 				 * path MTU information.
979 				 */
980 				goto notify;
981 			}
982 		}
983 	  notify:
984 #ifndef PULLDOWN_TEST
985 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
986 #else
987 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
988 			       sizeof(*icmp6) + sizeof(struct ip6_hdr));
989 		if (icmp6 == NULL) {
990 			icmp6stat.icp6s_tooshort++;
991 			return (-1);
992 		}
993 #endif
994 
995 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
996 		bzero(&icmp6dst, sizeof(icmp6dst));
997 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
998 		icmp6dst.sin6_family = AF_INET6;
999 		if (finaldst == NULL)
1000 			icmp6dst.sin6_addr = eip6->ip6_dst;
1001 		else
1002 			icmp6dst.sin6_addr = *finaldst;
1003 		if (in6_addr2zoneid(m->m_pkthdr.rcvif, &icmp6dst.sin6_addr,
1004 		    &icmp6dst.sin6_scope_id))
1005 			goto freeit;
1006 		if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1007 				   NULL, NULL)) {
1008 			/* should be impossbile */
1009 			nd6log((LOG_DEBUG,
1010 			    "icmp6_notify_error: in6_embedscope failed\n"));
1011 			goto freeit;
1012 		}
1013 
1014 		/*
1015 		 * retrieve parameters from the inner IPv6 header, and convert
1016 		 * them into sockaddr structures.
1017 		 */
1018 		bzero(&icmp6src, sizeof(icmp6src));
1019 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1020 		icmp6src.sin6_family = AF_INET6;
1021 		icmp6src.sin6_addr = eip6->ip6_src;
1022 		if (in6_addr2zoneid(m->m_pkthdr.rcvif, &icmp6src.sin6_addr,
1023 		    &icmp6src.sin6_scope_id))
1024 			goto freeit;
1025 		if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1026 				   NULL, NULL)) {
1027 			/* should be impossbile */
1028 			nd6log((LOG_DEBUG,
1029 			    "icmp6_notify_error: in6_embedscope failed\n"));
1030 			goto freeit;
1031 		}
1032 		icmp6src.sin6_flowinfo =
1033 			(eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1034 
1035 		if (finaldst == NULL)
1036 			finaldst = &eip6->ip6_dst;
1037 		ip6cp.ip6c_m = m;
1038 		ip6cp.ip6c_icmp6 = icmp6;
1039 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1040 		ip6cp.ip6c_off = eoff;
1041 		ip6cp.ip6c_finaldst = finaldst;
1042 		ip6cp.ip6c_src = &icmp6src;
1043 		ip6cp.ip6c_nxt = nxt;
1044 
1045 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1046 			notifymtu = ntohl(icmp6->icmp6_mtu);
1047 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1048 			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1049 		}
1050 
1051 		so_pr_ctlinput(
1052 			&inet6sw[ip6_protox[nxt]],
1053 			code, (struct sockaddr *)&icmp6dst, &ip6cp);
1054 	}
1055 	return (0);
1056 
1057 freeit:
1058 	m_freem(m);
1059 	return (-1);
1060 }
1061 
1062 void
1063 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1064 {
1065 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1066 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1067 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1068 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1069 	struct rtentry *rt = NULL;
1070 	struct sockaddr_in6 sin6;
1071 
1072 	if (!validated)
1073 		return;
1074 
1075 	bzero(&sin6, sizeof(sin6));
1076 	sin6.sin6_family = PF_INET6;
1077 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1078 	sin6.sin6_addr = *dst;
1079 	/* XXX normally, this won't happen */
1080 	if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1081 		sin6.sin6_addr.s6_addr16[1] =
1082 		    htons(m->m_pkthdr.rcvif->if_index);
1083 	}
1084 	/* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1085 	rt = rtpurelookup((struct sockaddr *)&sin6);
1086 	if (rt != NULL && (rt->rt_flags & RTF_HOST) &&
1087 	    !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1088 		if (mtu < IPV6_MMTU) {				/* XXX */
1089 			rt->rt_rmx.rmx_locks |= RTV_MTU;
1090 		} else if (mtu < rt->rt_ifp->if_mtu &&
1091 			   rt->rt_rmx.rmx_mtu > mtu) {
1092 			icmp6stat.icp6s_pmtuchg++;
1093 			rt->rt_rmx.rmx_mtu = mtu;
1094 		}
1095 	}
1096 	if (rt)
1097 		--rt->rt_refcnt;
1098 }
1099 
1100 /*
1101  * Process a Node Information Query packet, based on
1102  * draft-ietf-ipngwg-icmp-name-lookups-07.
1103  *
1104  * Spec incompatibilities:
1105  * - IPv6 Subject address handling
1106  * - IPv4 Subject address handling support missing
1107  * - Proxy reply (answer even if it's not for me)
1108  * - joins NI group address at in6_ifattach() time only, does not cope
1109  *   with hostname changes by sethostname(3)
1110  */
1111 #define hostnamelen	strlen(hostname)
1112 static struct mbuf *
1113 ni6_input(struct mbuf *m, int off)
1114 {
1115 	struct icmp6_nodeinfo *ni6, *nni6;
1116 	struct mbuf *n = NULL;
1117 	u_int16_t qtype;
1118 	int subjlen;
1119 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1120 	struct ni_reply_fqdn *fqdn;
1121 	int addrs;		/* for NI_QTYPE_NODEADDR */
1122 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1123 	struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1124 	struct sockaddr_in6 sin6_d; /* XXX: we should retrieve this from m_aux */
1125 	struct ip6_hdr *ip6;
1126 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1127 	char *subj = NULL;
1128 	struct in6_ifaddr *ia6 = NULL;
1129 
1130 	ip6 = mtod(m, struct ip6_hdr *);
1131 #ifndef PULLDOWN_TEST
1132 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1133 #else
1134 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1135 	if (ni6 == NULL) {
1136 		/* m is already reclaimed */
1137 		return NULL;
1138 	}
1139 #endif
1140 
1141 	/*
1142 	 * Validate IPv6 destination address.
1143 	 *
1144 	 * The Responder must discard the Query without further processing
1145 	 * unless it is one of the Responder's unicast or anycast addresses, or
1146 	 * a link-local scope multicast address which the Responder has joined.
1147 	 * [icmp-name-lookups-07, Section 4.]
1148 	 */
1149 	bzero(&sin6, sizeof(sin6));
1150 	sin6.sin6_family = AF_INET6;
1151 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1152 	bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1153 	/* XXX scopeid */
1154 	if ((ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)&sin6)) != NULL) {
1155 		/* unicast/anycast, fine */
1156 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1157 		    (icmp6_nodeinfo & 4) == 0) {
1158 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1159 				"a temporary address in %s:%d",
1160 			       __FILE__, __LINE__));
1161 			goto bad;
1162 		}
1163 	} else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1164 		; /* link-local multicast, fine */
1165 	else
1166 		goto bad;
1167 
1168 	/* validate query Subject field. */
1169 	qtype = ntohs(ni6->ni_qtype);
1170 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1171 	switch (qtype) {
1172 	case NI_QTYPE_NOOP:
1173 	case NI_QTYPE_SUPTYPES:
1174 		/* 07 draft */
1175 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1176 			break;
1177 		/* FALLTHROUGH */
1178 	case NI_QTYPE_FQDN:
1179 	case NI_QTYPE_NODEADDR:
1180 		switch (ni6->ni_code) {
1181 		case ICMP6_NI_SUBJ_IPV6:
1182 #if ICMP6_NI_SUBJ_IPV6 != 0
1183 		case 0:
1184 #endif
1185 			/*
1186 			 * backward compatibility - try to accept 03 draft
1187 			 * format, where no Subject is present.
1188 			 */
1189 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1190 			    subjlen == 0) {
1191 				oldfqdn++;
1192 				break;
1193 			}
1194 #if ICMP6_NI_SUBJ_IPV6 != 0
1195 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1196 				goto bad;
1197 #endif
1198 
1199 			if (subjlen != sizeof(sin6.sin6_addr))
1200 				goto bad;
1201 
1202 			/*
1203 			 * Validate Subject address.
1204 			 *
1205 			 * Not sure what exactly "address belongs to the node"
1206 			 * means in the spec, is it just unicast, or what?
1207 			 *
1208 			 * At this moment we consider Subject address as
1209 			 * "belong to the node" if the Subject address equals
1210 			 * to the IPv6 destination address; validation for
1211 			 * IPv6 destination address should have done enough
1212 			 * check for us.
1213 			 *
1214 			 * We do not do proxy at this moment.
1215 			 */
1216 			/* m_pulldown instead of copy? */
1217 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1218 			    subjlen, (caddr_t)&sin6.sin6_addr);
1219 			if (in6_addr2zoneid(m->m_pkthdr.rcvif,
1220 			    &sin6.sin6_addr, &sin6.sin6_scope_id))
1221 				goto bad;
1222 			in6_embedscope(&sin6.sin6_addr, &sin6, NULL, NULL);
1223 			bzero(&sin6_d, sizeof(sin6_d));
1224 			sin6_d.sin6_family = AF_INET6; /* not used, actually */
1225 			sin6_d.sin6_len = sizeof(sin6_d); /* ditto */
1226 			sin6_d.sin6_addr = ip6->ip6_dst;
1227 			if (in6_addr2zoneid(m->m_pkthdr.rcvif,
1228 			    &ip6->ip6_dst, &sin6_d.sin6_scope_id))
1229 				goto bad;
1230 			in6_embedscope(&sin6_d.sin6_addr, &sin6_d, NULL, NULL);
1231 			subj = (char *)&sin6;
1232 			if (SA6_ARE_ADDR_EQUAL(&sin6, &sin6_d))
1233 				break;
1234 
1235 			/*
1236 			 * XXX if we are to allow other cases, we should really
1237 			 * be careful about scope here.
1238 			 * basically, we should disallow queries toward IPv6
1239 			 * destination X with subject Y, if scope(X) > scope(Y).
1240 			 * if we allow scope(X) > scope(Y), it will result in
1241 			 * information leakage across scope boundary.
1242 			 */
1243 			goto bad;
1244 
1245 		case ICMP6_NI_SUBJ_FQDN:
1246 			/*
1247 			 * Validate Subject name with gethostname(3).
1248 			 *
1249 			 * The behavior may need some debate, since:
1250 			 * - we are not sure if the node has FQDN as
1251 			 *   hostname (returned by gethostname(3)).
1252 			 * - the code does wildcard match for truncated names.
1253 			 *   however, we are not sure if we want to perform
1254 			 *   wildcard match, if gethostname(3) side has
1255 			 *   truncated hostname.
1256 			 */
1257 			n = ni6_nametodns(hostname, hostnamelen, 0);
1258 			if (!n || n->m_next || n->m_len == 0)
1259 				goto bad;
1260 			IP6_EXTHDR_GET(subj, char *, m,
1261 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1262 			if (subj == NULL)
1263 				goto bad;
1264 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1265 					n->m_len)) {
1266 				goto bad;
1267 			}
1268 			m_freem(n);
1269 			n = NULL;
1270 			break;
1271 
1272 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1273 		default:
1274 			goto bad;
1275 		}
1276 		break;
1277 	}
1278 
1279 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1280 	switch (qtype) {
1281 	case NI_QTYPE_FQDN:
1282 		if ((icmp6_nodeinfo & 1) == 0)
1283 			goto bad;
1284 		break;
1285 	case NI_QTYPE_NODEADDR:
1286 		if ((icmp6_nodeinfo & 2) == 0)
1287 			goto bad;
1288 		break;
1289 	}
1290 
1291 	/* guess reply length */
1292 	switch (qtype) {
1293 	case NI_QTYPE_NOOP:
1294 		break;		/* no reply data */
1295 	case NI_QTYPE_SUPTYPES:
1296 		replylen += sizeof(u_int32_t);
1297 		break;
1298 	case NI_QTYPE_FQDN:
1299 		/* XXX will append an mbuf */
1300 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1301 		break;
1302 	case NI_QTYPE_NODEADDR:
1303 		addrs = ni6_addrs(ni6, m, &ifp, subj);
1304 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1305 					  sizeof(u_int32_t))) > MCLBYTES)
1306 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1307 		break;
1308 	default:
1309 		/*
1310 		 * XXX: We must return a reply with the ICMP6 code
1311 		 * `unknown Qtype' in this case. However we regard the case
1312 		 * as an FQDN query for backward compatibility.
1313 		 * Older versions set a random value to this field,
1314 		 * so it rarely varies in the defined qtypes.
1315 		 * But the mechanism is not reliable...
1316 		 * maybe we should obsolete older versions.
1317 		 */
1318 		qtype = NI_QTYPE_FQDN;
1319 		/* XXX will append an mbuf */
1320 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1321 		oldfqdn++;
1322 		break;
1323 	}
1324 
1325 	/* allocate an mbuf to reply. */
1326 	if (replylen > MCLBYTES) {
1327 		/*
1328 		 * XXX: should we try to allocate more? But MCLBYTES
1329 		 * is probably much larger than IPV6_MMTU...
1330 		 */
1331 		goto bad;
1332 	}
1333 	n = m_getb(replylen, M_NOWAIT, m->m_type, M_PKTHDR);
1334 	if (n == NULL) {
1335 		m_freem(m);
1336 		return (NULL);
1337 	}
1338 	M_MOVE_PKTHDR(n, m); /* just for recvif */
1339 	n->m_pkthdr.len = n->m_len = replylen;
1340 
1341 	/* copy mbuf header and IPv6 + Node Information base headers */
1342 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1343 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1344 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1345 
1346 	/* qtype dependent procedure */
1347 	switch (qtype) {
1348 	case NI_QTYPE_NOOP:
1349 		nni6->ni_code = ICMP6_NI_SUCCESS;
1350 		nni6->ni_flags = 0;
1351 		break;
1352 	case NI_QTYPE_SUPTYPES:
1353 	{
1354 		u_int32_t v;
1355 		nni6->ni_code = ICMP6_NI_SUCCESS;
1356 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1357 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1358 		v = (u_int32_t)htonl(0x0000000f);
1359 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1360 		break;
1361 	}
1362 	case NI_QTYPE_FQDN:
1363 		nni6->ni_code = ICMP6_NI_SUCCESS;
1364 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1365 						sizeof(struct ip6_hdr) +
1366 						sizeof(struct icmp6_nodeinfo));
1367 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1368 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1369 		/*
1370 		 * XXX do we really have FQDN in variable "hostname"?
1371 		 */
1372 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1373 		if (n->m_next == NULL)
1374 			goto bad;
1375 		/* XXX we assume that n->m_next is not a chain */
1376 		if (n->m_next->m_next != NULL)
1377 			goto bad;
1378 		n->m_pkthdr.len += n->m_next->m_len;
1379 		break;
1380 	case NI_QTYPE_NODEADDR:
1381 	{
1382 		int lenlim, copied;
1383 
1384 		nni6->ni_code = ICMP6_NI_SUCCESS;
1385 		n->m_pkthdr.len = n->m_len =
1386 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1387 		lenlim = M_TRAILINGSPACE(n);
1388 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1389 		/* XXX: reset mbuf length */
1390 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1391 			sizeof(struct icmp6_nodeinfo) + copied;
1392 		break;
1393 	}
1394 	default:
1395 		break;		/* XXX impossible! */
1396 	}
1397 
1398 	nni6->ni_type = ICMP6_NI_REPLY;
1399 	m_freem(m);
1400 	return (n);
1401 
1402 bad:
1403 	m_freem(m);
1404 	if (n)
1405 		m_freem(n);
1406 	return (NULL);
1407 }
1408 #undef hostnamelen
1409 
1410 /*
1411  * make a mbuf with DNS-encoded string.  no compression support.
1412  *
1413  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1414  * treated as truncated name (two \0 at the end).  this is a wild guess.
1415  */
1416 static struct mbuf *
1417 ni6_nametodns(const char *name, int namelen,
1418 	      int old) /* return pascal string if non-zero */
1419 {
1420 	struct mbuf *m;
1421 	char *cp, *ep;
1422 	const char *p, *q;
1423 	int i, len, nterm;
1424 
1425 	if (old)
1426 		len = namelen + 1;
1427 	else
1428 		len = MCLBYTES;
1429 
1430 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1431 	m = m_getb(len, M_NOWAIT, MT_DATA, 0);
1432 	if (!m)
1433 		goto fail;
1434 
1435 	if (old) {
1436 		m->m_len = len;
1437 		*mtod(m, char *) = namelen;
1438 		bcopy(name, mtod(m, char *) + 1, namelen);
1439 		return m;
1440 	} else {
1441 		m->m_len = 0;
1442 		cp = mtod(m, char *);
1443 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1444 
1445 		/* if not certain about my name, return empty buffer */
1446 		if (namelen == 0)
1447 			return m;
1448 
1449 		/*
1450 		 * guess if it looks like shortened hostname, or FQDN.
1451 		 * shortened hostname needs two trailing "\0".
1452 		 */
1453 		i = 0;
1454 		for (p = name; p < name + namelen; p++) {
1455 			if (*p && *p == '.')
1456 				i++;
1457 		}
1458 		if (i < 2)
1459 			nterm = 2;
1460 		else
1461 			nterm = 1;
1462 
1463 		p = name;
1464 		while (cp < ep && p < name + namelen) {
1465 			i = 0;
1466 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1467 				i++;
1468 			/* result does not fit into mbuf */
1469 			if (cp + i + 1 >= ep)
1470 				goto fail;
1471 			/*
1472 			 * DNS label length restriction, RFC1035 page 8.
1473 			 * "i == 0" case is included here to avoid returning
1474 			 * 0-length label on "foo..bar".
1475 			 */
1476 			if (i <= 0 || i >= 64)
1477 				goto fail;
1478 			*cp++ = i;
1479 			bcopy(p, cp, i);
1480 			cp += i;
1481 			p = q;
1482 			if (p < name + namelen && *p == '.')
1483 				p++;
1484 		}
1485 		/* termination */
1486 		if (cp + nterm >= ep)
1487 			goto fail;
1488 		while (nterm-- > 0)
1489 			*cp++ = '\0';
1490 		m->m_len = cp - mtod(m, char *);
1491 		return m;
1492 	}
1493 
1494 	panic("should not reach here");
1495 	/* NOTREACHED */
1496 
1497 fail:
1498 	if (m)
1499 		m_freem(m);
1500 	return NULL;
1501 }
1502 
1503 /*
1504  * check if two DNS-encoded string matches.  takes care of truncated
1505  * form (with \0\0 at the end).  no compression support.
1506  * XXX upper/lowercase match (see RFC2065)
1507  */
1508 static int
1509 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1510 {
1511 	const char *a0, *b0;
1512 	int l;
1513 
1514 	/* simplest case - need validation? */
1515 	if (alen == blen && bcmp(a, b, alen) == 0)
1516 		return 1;
1517 
1518 	a0 = a;
1519 	b0 = b;
1520 
1521 	/* termination is mandatory */
1522 	if (alen < 2 || blen < 2)
1523 		return 0;
1524 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1525 		return 0;
1526 	alen--;
1527 	blen--;
1528 
1529 	while (a - a0 < alen && b - b0 < blen) {
1530 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1531 			return 0;
1532 
1533 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1534 			return 0;
1535 		/* we don't support compression yet */
1536 		if (a[0] >= 64 || b[0] >= 64)
1537 			return 0;
1538 
1539 		/* truncated case */
1540 		if (a[0] == 0 && a - a0 == alen - 1)
1541 			return 1;
1542 		if (b[0] == 0 && b - b0 == blen - 1)
1543 			return 1;
1544 		if (a[0] == 0 || b[0] == 0)
1545 			return 0;
1546 
1547 		if (a[0] != b[0])
1548 			return 0;
1549 		l = a[0];
1550 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1551 			return 0;
1552 		if (bcmp(a + 1, b + 1, l) != 0)
1553 			return 0;
1554 
1555 		a += 1 + l;
1556 		b += 1 + l;
1557 	}
1558 
1559 	if (a - a0 == alen && b - b0 == blen)
1560 		return 1;
1561 	else
1562 		return 0;
1563 }
1564 
1565 /*
1566  * calculate the number of addresses to be returned in the node info reply.
1567  */
1568 static int
1569 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1570 	  char *subj)
1571 {
1572 	const struct ifnet_array *arr;
1573 	struct in6_ifaddr *ifa6;
1574 	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1575 	int addrs = 0, addrsofif, iffound = 0;
1576 	int niflags = ni6->ni_flags;
1577 	int i;
1578 
1579 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1580 		switch (ni6->ni_code) {
1581 		case ICMP6_NI_SUBJ_IPV6:
1582 			if (subj == NULL) /* must be impossible... */
1583 				return (0);
1584 			subj_ip6 = (struct sockaddr_in6 *)subj;
1585 			break;
1586 		default:
1587 			/*
1588 			 * XXX: we only support IPv6 subject address for
1589 			 * this Qtype.
1590 			 */
1591 			return (0);
1592 		}
1593 	}
1594 
1595 	arr = ifnet_array_get();
1596 	for (i = 0; i < arr->ifnet_count; ++i) {
1597 		struct ifnet *ifp = arr->ifnet_arr[i];
1598 		struct ifaddr_container *ifac;
1599 
1600 		addrsofif = 0;
1601 		TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link)
1602 		{
1603 			struct ifaddr *ifa = ifac->ifa;
1604 
1605 			if (ifa->ifa_addr->sa_family != AF_INET6)
1606 				continue;
1607 			ifa6 = (struct in6_ifaddr *)ifa;
1608 
1609 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1610 			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1611 					       &ifa6->ia_addr.sin6_addr))
1612 				iffound = 1;
1613 
1614 			/*
1615 			 * IPv4-mapped addresses can only be returned by a
1616 			 * Node Information proxy, since they represent
1617 			 * addresses of IPv4-only nodes, which perforce do
1618 			 * not implement this protocol.
1619 			 * [icmp-name-lookups-07, Section 5.4]
1620 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1621 			 * this function at this moment.
1622 			 */
1623 
1624 			/* What do we have to do about ::1? */
1625 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1626 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1627 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1628 					continue;
1629 				break;
1630 			case IPV6_ADDR_SCOPE_SITELOCAL:
1631 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1632 					continue;
1633 				break;
1634 			case IPV6_ADDR_SCOPE_GLOBAL:
1635 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1636 					continue;
1637 				break;
1638 			default:
1639 				continue;
1640 			}
1641 
1642 			/*
1643 			 * check if anycast is okay.
1644 			 * XXX: just experimental.  not in the spec.
1645 			 */
1646 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) &&
1647 			    !(niflags & NI_NODEADDR_FLAG_ANYCAST))
1648 				continue; /* we need only unicast addresses */
1649 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) &&
1650 			    (icmp6_nodeinfo & 4) == 0) {
1651 				continue;
1652 			}
1653 			addrsofif++; /* count the address */
1654 		}
1655 		if (iffound) {
1656 			*ifpp = ifp;
1657 			return (addrsofif);
1658 		}
1659 
1660 		addrs += addrsofif;
1661 	}
1662 
1663 	return (addrs);
1664 }
1665 
1666 static int
1667 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1668 		struct ifnet *ifp0, int resid)
1669 {
1670 	const struct ifnet_array *arr;
1671 	struct in6_ifaddr *ifa6;
1672 	int copied = 0, allow_deprecated = 0;
1673 	u_char *cp = (u_char *)(nni6 + 1);
1674 	int niflags = ni6->ni_flags;
1675 	u_int32_t ltime;
1676 	int idx, idx_dep = -1;
1677 
1678 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1679 		return (0);	/* needless to copy */
1680 
1681 	arr = ifnet_array_get();
1682 	if (ifp0 == NULL) {
1683 		idx = 0;
1684 	} else {
1685 		for (idx = 0; idx < arr->ifnet_count; ++idx) {
1686 			if (arr->ifnet_arr[idx] == ifp0)
1687 				break;
1688 		}
1689 	}
1690 
1691 again:
1692 
1693 	for (; idx < arr->ifnet_count; ++idx) {
1694 		struct ifnet *ifp = arr->ifnet_arr[idx];
1695 		struct ifaddr_container *ifac;
1696 
1697 		TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1698 			struct ifaddr *ifa = ifac->ifa;
1699 
1700 			if (ifa->ifa_addr->sa_family != AF_INET6)
1701 				continue;
1702 			ifa6 = (struct in6_ifaddr *)ifa;
1703 
1704 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) &&
1705 			    allow_deprecated == 0) {
1706 				/*
1707 				 * prefererred address should be put before
1708 				 * deprecated addresses.
1709 				 */
1710 
1711 				/* record the interface for later search */
1712 				if (idx_dep < 0)
1713 					idx_dep = idx;
1714 
1715 				continue;
1716 			}
1717 			else if (!(ifa6->ia6_flags & IN6_IFF_DEPRECATED) &&
1718 				 allow_deprecated != 0)
1719 				continue; /* we now collect deprecated addrs */
1720 
1721 			/* What do we have to do about ::1? */
1722 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1723 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1724 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1725 					continue;
1726 				break;
1727 			case IPV6_ADDR_SCOPE_SITELOCAL:
1728 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1729 					continue;
1730 				break;
1731 			case IPV6_ADDR_SCOPE_GLOBAL:
1732 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1733 					continue;
1734 				break;
1735 			default:
1736 				continue;
1737 			}
1738 
1739 			/*
1740 			 * check if anycast is okay.
1741 			 * XXX: just experimental. not in the spec.
1742 			 */
1743 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) &&
1744 			    !(niflags & NI_NODEADDR_FLAG_ANYCAST))
1745 				continue;
1746 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) &&
1747 			    (icmp6_nodeinfo & 4) == 0) {
1748 				continue;
1749 			}
1750 
1751 			/* now we can copy the address */
1752 			if (resid < sizeof(struct in6_addr) +
1753 			    sizeof(u_int32_t)) {
1754 				/*
1755 				 * We give up much more copy.
1756 				 * Set the truncate flag and return.
1757 				 */
1758 				nni6->ni_flags |=
1759 					NI_NODEADDR_FLAG_TRUNCATE;
1760 				return (copied);
1761 			}
1762 
1763 			/*
1764 			 * Set the TTL of the address.
1765 			 * The TTL value should be one of the following
1766 			 * according to the specification:
1767 			 *
1768 			 * 1. The remaining lifetime of a DHCP lease on the
1769 			 *    address, or
1770 			 * 2. The remaining Valid Lifetime of a prefix from
1771 			 *    which the address was derived through Stateless
1772 			 *    Autoconfiguration.
1773 			 *
1774 			 * Note that we currently do not support stateful
1775 			 * address configuration by DHCPv6, so the former
1776 			 * case can't happen.
1777 			 */
1778 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1779 				ltime = ND6_INFINITE_LIFETIME;
1780 			else {
1781 				if (ifa6->ia6_lifetime.ia6t_expire >
1782 				    time_uptime)
1783 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime);
1784 				else
1785 					ltime = 0;
1786 			}
1787 
1788 			bcopy(&ltime, cp, sizeof(u_int32_t));
1789 			cp += sizeof(u_int32_t);
1790 
1791 			/* copy the address itself */
1792 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1793 			      sizeof(struct in6_addr));
1794 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1795 			cp += sizeof(struct in6_addr);
1796 
1797 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1798 			copied += (sizeof(struct in6_addr) +
1799 				   sizeof(u_int32_t));
1800 		}
1801 		if (ifp0)	/* we need search only on the specified IF */
1802 			break;
1803 	}
1804 
1805 	if (allow_deprecated == 0 && idx_dep >= 0) {
1806 		idx = idx_dep;
1807 		allow_deprecated = 1;
1808 
1809 		goto again;
1810 	}
1811 
1812 	return (copied);
1813 }
1814 
1815 /*
1816  * XXX almost dup'ed code with rip6_input.
1817  */
1818 static int
1819 icmp6_rip6_input(struct	mbuf **mp, int	off)
1820 {
1821 	struct mbuf *m = *mp;
1822 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1823 	struct in6pcb *in6p;
1824 	struct in6pcb *last = NULL;
1825 	struct sockaddr_in6 fromsa;
1826 	struct icmp6_hdr *icmp6;
1827 	struct mbuf *opts = NULL;
1828 
1829 #ifndef PULLDOWN_TEST
1830 	/* this is assumed to be safe. */
1831 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1832 #else
1833 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1834 	if (icmp6 == NULL) {
1835 		/* m is already reclaimed */
1836 		return IPPROTO_DONE;
1837 	}
1838 #endif
1839 
1840 	bzero(&fromsa, sizeof(fromsa));
1841 	fromsa.sin6_len = sizeof(struct sockaddr_in6);
1842 	fromsa.sin6_family = AF_INET6;
1843 	/* KAME hack: recover scopeid */
1844 	in6_recoverscope(&fromsa, &ip6->ip6_src, m->m_pkthdr.rcvif);
1845 
1846 	LIST_FOREACH(in6p, &ripcbinfo.pcblisthead, inp_list)
1847 	{
1848 		if (in6p->inp_flags & INP_PLACEMARKER)
1849 			continue;
1850 		if (!INP_ISIPV6(in6p))
1851 			continue;
1852 		if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1853 			continue;
1854 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1855 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1856 			continue;
1857 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1858 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1859 			continue;
1860 		if (in6p->in6p_icmp6filt
1861 		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1862 				 in6p->in6p_icmp6filt))
1863 			continue;
1864 		if (last) {
1865 			struct socket *so;
1866 			struct mbuf *n;
1867 
1868 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1869 				if (last->in6p_flags & IN6P_CONTROLOPTS)
1870 					ip6_savecontrol(last, &opts, ip6, n);
1871 				/* strip intermediate headers */
1872 				m_adj(n, off);
1873 				so = last->in6p_socket;
1874 				lwkt_gettoken(&so->so_rcv.ssb_token);
1875 				if (ssb_appendaddr(&so->so_rcv,
1876 				    (struct sockaddr *)&fromsa, n, opts) == 0) {
1877 					/* should notify about lost packet */
1878 					m_freem(n);
1879 					if (opts)
1880 						m_freem(opts);
1881 				} else {
1882 					sorwakeup(so);
1883 				}
1884 				lwkt_reltoken(&so->so_rcv.ssb_token);
1885 				opts = NULL;
1886 			}
1887 		}
1888 		last = in6p;
1889 	}
1890 	if (last) {
1891 		struct socket *so;
1892 
1893 		if (last->in6p_flags & IN6P_CONTROLOPTS)
1894 			ip6_savecontrol(last, &opts, ip6, m);
1895 		/* strip intermediate headers */
1896 		m_adj(m, off);
1897 		so = last->in6p_socket;
1898 		lwkt_gettoken(&so->so_rcv.ssb_token);
1899 		if (ssb_appendaddr(&so->so_rcv, (struct sockaddr *)&fromsa,
1900 		    m, opts) == 0) {
1901 			m_freem(m);
1902 			if (opts)
1903 				m_freem(opts);
1904 		} else {
1905 			sorwakeup(so);
1906 		}
1907 		lwkt_reltoken(&so->so_rcv.ssb_token);
1908 	} else {
1909 		m_freem(m);
1910 		ip6stat.ip6s_delivered--;
1911 	}
1912 	return IPPROTO_DONE;
1913 }
1914 
1915 /*
1916  * Reflect the ip6 packet back to the source.
1917  * OFF points to the icmp6 header, counted from the top of the mbuf.
1918  */
1919 void
1920 icmp6_reflect(struct mbuf *m, size_t off)
1921 {
1922 	struct ip6_hdr *ip6;
1923 	struct icmp6_hdr *icmp6;
1924 	struct in6_ifaddr *ia;
1925 	struct in6_addr t, *src = NULL;
1926 	int plen;
1927 	int type, code;
1928 	struct ifnet *outif = NULL;
1929 	struct sockaddr_in6 sa6_src, sa6_dst;
1930 #ifdef COMPAT_RFC1885
1931 	int mtu = IPV6_MMTU;
1932 	struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
1933 #endif
1934 
1935 	/* too short to reflect */
1936 	if (off < sizeof(struct ip6_hdr)) {
1937 		nd6log((LOG_DEBUG,
1938 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
1939 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
1940 		    __FILE__, __LINE__));
1941 		goto bad;
1942 	}
1943 
1944 	/*
1945 	 * If there are extra headers between IPv6 and ICMPv6, strip
1946 	 * off that header first.
1947 	 */
1948 #ifdef DIAGNOSTIC
1949 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
1950 		panic("assumption failed in icmp6_reflect");
1951 #endif
1952 	if (off > sizeof(struct ip6_hdr)) {
1953 		size_t l;
1954 		struct ip6_hdr nip6;
1955 
1956 		l = off - sizeof(struct ip6_hdr);
1957 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
1958 		m_adj(m, l);
1959 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1960 		if (m->m_len < l) {
1961 			if ((m = m_pullup(m, l)) == NULL)
1962 				return;
1963 		}
1964 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
1965 	} else /* off == sizeof(struct ip6_hdr) */ {
1966 		size_t l;
1967 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1968 		if (m->m_len < l) {
1969 			if ((m = m_pullup(m, l)) == NULL)
1970 				return;
1971 		}
1972 	}
1973 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
1974 	ip6 = mtod(m, struct ip6_hdr *);
1975 	ip6->ip6_nxt = IPPROTO_ICMPV6;
1976 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
1977 	type = icmp6->icmp6_type; /* keep type for statistics */
1978 	code = icmp6->icmp6_code; /* ditto. */
1979 
1980 	t = ip6->ip6_dst;
1981 	/*
1982 	 * ip6_input() drops a packet if its src is multicast.
1983 	 * So, the src is never multicast.
1984 	 */
1985 	ip6->ip6_dst = ip6->ip6_src;
1986 
1987 	/*
1988 	 * XXX: make sure to embed scope zone information, using
1989 	 * already embedded IDs or the received interface (if any).
1990 	 * Note that rcvif may be NULL.
1991 	 * TODO: scoped routing case (XXX).
1992 	 */
1993 	bzero(&sa6_src, sizeof(sa6_src));
1994 	sa6_src.sin6_family = AF_INET6;
1995 	sa6_src.sin6_len = sizeof(sa6_src);
1996 	sa6_src.sin6_addr = ip6->ip6_dst;
1997 	in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
1998 	in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL);
1999 	bzero(&sa6_dst, sizeof(sa6_dst));
2000 	sa6_dst.sin6_family = AF_INET6;
2001 	sa6_dst.sin6_len = sizeof(sa6_dst);
2002 	sa6_dst.sin6_addr = t;
2003 	in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
2004 	in6_embedscope(&t, &sa6_dst, NULL, NULL);
2005 
2006 #ifdef COMPAT_RFC1885
2007 	/*
2008 	 * xxx guess MTU
2009 	 * RFC 1885 requires that echo reply should be truncated if it
2010 	 * does not fit in with (return) path MTU, but the description was
2011 	 * removed in the new spec.
2012 	 */
2013 	if (icmp6_reflect_rt.ro_rt == 0 ||
2014 	    ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
2015 		if (icmp6_reflect_rt.ro_rt) {
2016 			RTFREE(icmp6_reflect_rt.ro_rt);
2017 			icmp6_reflect_rt.ro_rt = 0;
2018 		}
2019 		bzero(sin6, sizeof(*sin6));
2020 		sin6->sin6_family = PF_INET6;
2021 		sin6->sin6_len = sizeof(struct sockaddr_in6);
2022 		sin6->sin6_addr = ip6->ip6_dst;
2023 
2024 		rtalloc_ign((struct route *)&icmp6_reflect_rt.ro_rt,
2025 			    RTF_PRCLONING);
2026 	}
2027 
2028 	if (icmp6_reflect_rt.ro_rt == 0)
2029 		goto bad;
2030 
2031 	if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
2032 	    && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
2033 		mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
2034 
2035 	if (mtu < m->m_pkthdr.len) {
2036 		plen -= (m->m_pkthdr.len - mtu);
2037 		m_adj(m, mtu - m->m_pkthdr.len);
2038 	}
2039 #endif
2040 	/*
2041 	 * If the incoming packet was addressed directly to us(i.e. unicast),
2042 	 * use dst as the src for the reply.
2043 	 * The IN6_IFF_NOTREADY case would be VERY rare, but is possible
2044 	 * (for example) when we encounter an error while forwarding procedure
2045 	 * destined to a duplicated address of ours.
2046 	 */
2047 	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2048 		if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2049 		    (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2050 			src = &t;
2051 			break;
2052 		}
2053 	if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2054 		/*
2055 		 * This is the case if the dst is our link-local address
2056 		 * and the sender is also ourselves.
2057 		 */
2058 		src = &t;
2059 	}
2060 
2061 	if (src == NULL) {
2062 		int e;
2063 		struct route_in6 ro;
2064 
2065 		/*
2066 		 * This case matches to multicasts, our anycast, or unicasts
2067 		 * that we do not own.  Select a source address based on the
2068 		 * source address of the erroneous packet.
2069 		 */
2070 		bzero(&ro, sizeof(ro));
2071 		src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e, NULL);
2072 		if (ro.ro_rt)
2073 			RTFREE(ro.ro_rt); /* XXX: we could use this */
2074 		if (src == NULL) {
2075 			nd6log((LOG_DEBUG,
2076 			    "icmp6_reflect: source can't be determined: "
2077 			    "dst=%s, error=%d\n",
2078 			    ip6_sprintf(&sa6_src.sin6_addr), e));
2079 			goto bad;
2080 		}
2081 	}
2082 
2083 	ip6->ip6_src = *src;
2084 
2085 	ip6->ip6_flow = 0;
2086 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2087 	ip6->ip6_vfc |= IPV6_VERSION;
2088 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2089 	if (m->m_pkthdr.rcvif) {
2090 		/* XXX: This may not be the outgoing interface */
2091 		ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2092 		if (ip6->ip6_hlim < ip6_minhlim)
2093 			ip6->ip6_hlim = ip6_minhlim;
2094 	} else {
2095 		ip6->ip6_hlim = ip6_defhlim;
2096 	}
2097 
2098 	icmp6->icmp6_cksum = 0;
2099 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2100 					sizeof(struct ip6_hdr), plen);
2101 
2102 	/*
2103 	 * XXX option handling
2104 	 */
2105 
2106 	m->m_flags &= ~(M_BCAST|M_MCAST);
2107 
2108 #ifdef COMPAT_RFC1885
2109 	ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif, NULL);
2110 #else
2111 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2112 #endif
2113 	if (outif)
2114 		icmp6_ifoutstat_inc(outif, type, code);
2115 
2116 	return;
2117 
2118 bad:
2119 	m_freem(m);
2120 	return;
2121 }
2122 
2123 static void
2124 icmp6_fasttimo_dispatch(netmsg_t nmsg)
2125 {
2126 
2127 	ASSERT_NETISR0;
2128 
2129 	/* Reply ASAP. */
2130 	crit_enter();
2131 	netisr_replymsg(&nmsg->base, 0);
2132 	crit_exit();
2133 
2134 	mld6_fasttimeo();
2135 
2136 	callout_reset(&icmp6_fasttimo_ch, ICMP6_FASTTIMO, icmp6_fasttimo, NULL);
2137 }
2138 
2139 static void
2140 icmp6_fasttimo(void *dummy __unused)
2141 {
2142 	struct netmsg_base *nmsg = &icmp6_fasttimo_nmsg;
2143 
2144 	KKASSERT(mycpuid == 0);
2145 
2146 	crit_enter();
2147 	if (nmsg->lmsg.ms_flags & MSGF_DONE)
2148 		netisr_sendmsg_oncpu(nmsg);
2149 	crit_exit();
2150 }
2151 
2152 static const char *
2153 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2154 		    struct in6_addr *tgt6)
2155 {
2156 	static char buf[1024];
2157 	ksnprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2158 		ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2159 	return buf;
2160 }
2161 
2162 void
2163 icmp6_redirect_input(struct mbuf *m, int off)
2164 {
2165 	struct ifnet *ifp = m->m_pkthdr.rcvif;
2166 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2167 	struct nd_redirect *nd_rd;
2168 	int icmp6len = ntohs(ip6->ip6_plen);
2169 	char *lladdr = NULL;
2170 	int lladdrlen = 0;
2171 #if 0
2172 	u_char *redirhdr = NULL;
2173 	int redirhdrlen = 0;
2174 #endif
2175 	struct rtentry *rt = NULL;
2176 	int is_router;
2177 	int is_onlink;
2178 	struct in6_addr src6 = ip6->ip6_src;
2179 	struct in6_addr redtgt6;
2180 	struct in6_addr reddst6;
2181 	union nd_opts ndopts;
2182 
2183 	if (!m || !ifp)
2184 		return;
2185 
2186 	/* XXX if we are router, we don't update route by icmp6 redirect */
2187 	if (ip6_forwarding)
2188 		goto freeit;
2189 	if (!icmp6_rediraccept)
2190 		goto freeit;
2191 
2192 #ifndef PULLDOWN_TEST
2193 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2194 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2195 #else
2196 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2197 	if (nd_rd == NULL) {
2198 		icmp6stat.icp6s_tooshort++;
2199 		return;
2200 	}
2201 #endif
2202 	redtgt6 = nd_rd->nd_rd_target;
2203 	reddst6 = nd_rd->nd_rd_dst;
2204 
2205 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2206 		redtgt6.s6_addr16[1] = htons(ifp->if_index);
2207 	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2208 		reddst6.s6_addr16[1] = htons(ifp->if_index);
2209 
2210 	/* validation */
2211 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2212 		nd6log((LOG_ERR,
2213 			"ICMP6 redirect sent from %s rejected; "
2214 			"must be from linklocal\n", ip6_sprintf(&src6)));
2215 		goto bad;
2216 	}
2217 	if (ip6->ip6_hlim != 255) {
2218 		nd6log((LOG_ERR,
2219 			"ICMP6 redirect sent from %s rejected; "
2220 			"hlim=%d (must be 255)\n",
2221 			ip6_sprintf(&src6), ip6->ip6_hlim));
2222 		goto bad;
2223 	}
2224     {
2225 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2226 	struct sockaddr_in6 sin6;
2227 	struct in6_addr *gw6;
2228 
2229 	bzero(&sin6, sizeof(sin6));
2230 	sin6.sin6_family = AF_INET6;
2231 	sin6.sin6_len = sizeof(struct sockaddr_in6);
2232 	bcopy(&reddst6, &sin6.sin6_addr, sizeof reddst6);
2233 	rt = rtpurelookup((struct sockaddr *)&sin6);
2234 	if (rt != NULL) {
2235 		if (rt->rt_gateway == NULL ||
2236 		    rt->rt_gateway->sa_family != AF_INET6) {
2237 			nd6log((LOG_ERR,
2238 			    "ICMP6 redirect rejected; no route "
2239 			    "with inet6 gateway found for redirect dst: %s\n",
2240 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2241 			--rt->rt_refcnt;
2242 			goto bad;
2243 		}
2244 
2245 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2246 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2247 			nd6log((LOG_ERR,
2248 				"ICMP6 redirect rejected; "
2249 				"not equal to gw-for-src=%s (must be same): "
2250 				"%s\n",
2251 				ip6_sprintf(gw6),
2252 				icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2253 			--rt->rt_refcnt;
2254 			goto bad;
2255 		}
2256 	} else {
2257 		nd6log((LOG_ERR,
2258 			"ICMP6 redirect rejected; "
2259 			"no route found for redirect dst: %s\n",
2260 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2261 		goto bad;
2262 	}
2263 	--rt->rt_refcnt;
2264 	rt = NULL;
2265     }
2266 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2267 		nd6log((LOG_ERR,
2268 			"ICMP6 redirect rejected; "
2269 			"redirect dst must be unicast: %s\n",
2270 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2271 		goto bad;
2272 	}
2273 
2274 	is_router = is_onlink = 0;
2275 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2276 		is_router = 1;	/* router case */
2277 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2278 		is_onlink = 1;	/* on-link destination case */
2279 	if (!is_router && !is_onlink) {
2280 		nd6log((LOG_ERR,
2281 			"ICMP6 redirect rejected; "
2282 			"neither router case nor onlink case: %s\n",
2283 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2284 		goto bad;
2285 	}
2286 	/* validation passed */
2287 
2288 	icmp6len -= sizeof(*nd_rd);
2289 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2290 	if (nd6_options(&ndopts) < 0) {
2291 		nd6log((LOG_INFO, "icmp6_redirect_input: "
2292 			"invalid ND option, rejected: %s\n",
2293 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2294 		/* nd6_options have incremented stats */
2295 		goto freeit;
2296 	}
2297 
2298 	if (ndopts.nd_opts_tgt_lladdr) {
2299 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2300 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2301 	}
2302 
2303 #if 0
2304 	if (ndopts.nd_opts_rh) {
2305 		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2306 		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2307 	}
2308 #endif
2309 
2310 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2311 		nd6log((LOG_INFO,
2312 			"icmp6_redirect_input: lladdrlen mismatch for %s "
2313 			"(if %d, icmp6 packet %d): %s\n",
2314 			ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2315 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2316 		goto bad;
2317 	}
2318 
2319 	/* RFC 2461 8.3 */
2320 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2321 			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2322 
2323 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2324 		/* perform rtredirect */
2325 		struct sockaddr_in6 sdst;
2326 		struct sockaddr_in6 sgw;
2327 		struct sockaddr_in6 ssrc;
2328 
2329 		bzero(&sdst, sizeof(sdst));
2330 		bzero(&sgw, sizeof(sgw));
2331 		bzero(&ssrc, sizeof(ssrc));
2332 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2333 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2334 			sizeof(struct sockaddr_in6);
2335 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2336 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2337 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2338 		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2339 			   NULL, RTF_GATEWAY | RTF_HOST,
2340 			   (struct sockaddr *)&ssrc);
2341 	}
2342 	/* finally update cached route in each socket via kpfctlinput */
2343     {
2344 	struct sockaddr_in6 sdst;
2345 
2346 	bzero(&sdst, sizeof(sdst));
2347 	sdst.sin6_family = AF_INET6;
2348 	sdst.sin6_len = sizeof(struct sockaddr_in6);
2349 	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2350 	kpfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2351 #ifdef IPSEC
2352 	key_sa_routechange((struct sockaddr *)&sdst);
2353 #endif
2354     }
2355 
2356 freeit:
2357 	m_freem(m);
2358 	return;
2359 
2360 bad:
2361 	icmp6stat.icp6s_badredirect++;
2362 	m_freem(m);
2363 }
2364 
2365 void
2366 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2367 {
2368 	struct ifnet *ifp;	/* my outgoing interface */
2369 	struct in6_addr *ifp_ll6;
2370 	struct in6_addr *router_ll6;
2371 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2372 	struct mbuf *m = NULL;	/* newly allocated one */
2373 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2374 	struct nd_redirect *nd_rd;
2375 	size_t maxlen;
2376 	u_char *p;
2377 	struct ifnet *outif = NULL;
2378 	struct sockaddr_in6 src_sa;
2379 
2380 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2381 
2382 	/* if we are not router, we don't send icmp6 redirect */
2383 	if (!ip6_forwarding || ip6_accept_rtadv)
2384 		goto fail;
2385 
2386 	/* sanity check */
2387 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2388 		goto fail;
2389 
2390 	/*
2391 	 * Address check:
2392 	 *  the source address must identify a neighbor, and
2393 	 *  the destination address must not be a multicast address
2394 	 *  [RFC 2461, sec 8.2]
2395 	 */
2396 	sip6 = mtod(m0, struct ip6_hdr *);
2397 	bzero(&src_sa, sizeof(src_sa));
2398 	src_sa.sin6_family = AF_INET6;
2399 	src_sa.sin6_len = sizeof(src_sa);
2400 	src_sa.sin6_addr = sip6->ip6_src;
2401 	/* we don't currently use sin6_scope_id, but eventually use it */
2402 	if (in6_addr2zoneid(ifp, &sip6->ip6_src, &src_sa.sin6_scope_id))
2403 		goto fail;
2404 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2405 		goto fail;
2406 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2407 		goto fail;	/* what should we do here? */
2408 
2409 	/* rate limit */
2410 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2411 		goto fail;
2412 
2413 	/*
2414 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2415 	 * we almost always ask for an mbuf cluster for simplicity.
2416 	 * (MHLEN < IPV6_MMTU is almost always true)
2417 	 */
2418 #if IPV6_MMTU >= MCLBYTES
2419 # error assumption failed about IPV6_MMTU and MCLBYTES
2420 #endif
2421 	m = m_getb(IPV6_MMTU, M_NOWAIT, MT_HEADER, M_PKTHDR);
2422 	if (!m)
2423 		goto fail;
2424 	m->m_len = 0;
2425 	maxlen = M_TRAILINGSPACE(m);
2426 	maxlen = min(IPV6_MMTU, maxlen);
2427 	/* just for safety */
2428 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2429 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2430 		goto fail;
2431 	}
2432 
2433 	{
2434 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2435 		struct in6_ifaddr *ia;
2436 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2437 						 IN6_IFF_NOTREADY|
2438 						 IN6_IFF_ANYCAST)) == NULL)
2439 			goto fail;
2440 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2441 	}
2442 
2443 	/* get ip6 linklocal address for the router. */
2444 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2445 		struct sockaddr_in6 *sin6;
2446 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2447 		router_ll6 = &sin6->sin6_addr;
2448 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2449 			router_ll6 = NULL;
2450 	} else
2451 		router_ll6 = NULL;
2452 
2453 	/* ip6 */
2454 	ip6 = mtod(m, struct ip6_hdr *);
2455 	ip6->ip6_flow = 0;
2456 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2457 	ip6->ip6_vfc |= IPV6_VERSION;
2458 	/* ip6->ip6_plen will be set later */
2459 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2460 	ip6->ip6_hlim = 255;
2461 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2462 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2463 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2464 
2465 	/* ND Redirect */
2466 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2467 	nd_rd->nd_rd_type = ND_REDIRECT;
2468 	nd_rd->nd_rd_code = 0;
2469 	nd_rd->nd_rd_reserved = 0;
2470 	if (rt->rt_flags & RTF_GATEWAY) {
2471 		/*
2472 		 * nd_rd->nd_rd_target must be a link-local address in
2473 		 * better router cases.
2474 		 */
2475 		if (!router_ll6)
2476 			goto fail;
2477 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2478 		      sizeof(nd_rd->nd_rd_target));
2479 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2480 		      sizeof(nd_rd->nd_rd_dst));
2481 	} else {
2482 		/* make sure redtgt == reddst */
2483 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2484 		      sizeof(nd_rd->nd_rd_target));
2485 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2486 		      sizeof(nd_rd->nd_rd_dst));
2487 	}
2488 
2489 	p = (u_char *)(nd_rd + 1);
2490 
2491 	if (!router_ll6)
2492 		goto nolladdropt;
2493 
2494 	{
2495 		/* target lladdr option */
2496 		struct rtentry *rt_router = NULL;
2497 		int len;
2498 		struct sockaddr_dl *sdl;
2499 		struct nd_opt_hdr *nd_opt;
2500 		char *lladdr;
2501 
2502 		rt_router = nd6_lookup(router_ll6, 0, ifp);
2503 		if (!rt_router)
2504 			goto nolladdropt;
2505 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2506 		len = (len + 7) & ~7;	/* round by 8 */
2507 		/* safety check */
2508 		if (len + (p - (u_char *)ip6) > maxlen)
2509 			goto nolladdropt;
2510 		if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2511 		    (rt_router->rt_flags & RTF_LLINFO) &&
2512 		    (rt_router->rt_gateway->sa_family == AF_LINK) &&
2513 		    (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2514 		    sdl->sdl_alen) {
2515 			nd_opt = (struct nd_opt_hdr *)p;
2516 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2517 			nd_opt->nd_opt_len = len >> 3;
2518 			lladdr = (char *)(nd_opt + 1);
2519 			bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2520 			p += len;
2521 		}
2522 	}
2523 nolladdropt:;
2524 
2525 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2526 
2527 	/* just to be safe */
2528 #ifdef M_DECRYPTED	/*not openbsd*/
2529 	if (m0->m_flags & M_DECRYPTED)
2530 		goto noredhdropt;
2531 #endif
2532 	if (p - (u_char *)ip6 > maxlen)
2533 		goto noredhdropt;
2534 
2535     {
2536 	/* redirected header option */
2537 	int len;
2538 	struct nd_opt_rd_hdr *nd_opt_rh;
2539 
2540 	/*
2541 	 * compute the maximum size for icmp6 redirect header option.
2542 	 * XXX room for auth header?
2543 	 */
2544 	len = maxlen - (p - (u_char *)ip6);
2545 	len &= ~7;
2546 
2547 	/* This is just for simplicity. */
2548 	if (m0->m_pkthdr.len != m0->m_len) {
2549 		if (m0->m_next) {
2550 			m_freem(m0->m_next);
2551 			m0->m_next = NULL;
2552 		}
2553 		m0->m_pkthdr.len = m0->m_len;
2554 	}
2555 
2556 	/*
2557 	 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2558 	 * about padding/truncate rule for the original IP packet.
2559 	 * From the discussion on IPv6imp in Feb 1999, the consensus was:
2560 	 * - "attach as much as possible" is the goal
2561 	 * - pad if not aligned (original size can be guessed by original
2562 	 *   ip6 header)
2563 	 * Following code adds the padding if it is simple enough,
2564 	 * and truncates if not.
2565 	 */
2566 	if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2567 		panic("assumption failed in %s:%d", __FILE__, __LINE__);
2568 
2569 	if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2570 		/* not enough room, truncate */
2571 		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2572 	} else {
2573 		/* enough room, pad or truncate */
2574 		size_t extra;
2575 
2576 		extra = m0->m_pkthdr.len % 8;
2577 		if (extra) {
2578 			/* pad if easy enough, truncate if not */
2579 			if (8 - extra <= M_TRAILINGSPACE(m0)) {
2580 				/* pad */
2581 				m0->m_len += (8 - extra);
2582 				m0->m_pkthdr.len += (8 - extra);
2583 			} else {
2584 				/* truncate */
2585 				m0->m_pkthdr.len -= extra;
2586 				m0->m_len -= extra;
2587 			}
2588 		}
2589 		len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2590 		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2591 	}
2592 
2593 	nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2594 	bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2595 	nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2596 	nd_opt_rh->nd_opt_rh_len = len >> 3;
2597 	p += sizeof(*nd_opt_rh);
2598 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2599 
2600 	/* connect m0 to m */
2601 	m->m_next = m0;
2602 	m->m_pkthdr.len = m->m_len + m0->m_len;
2603 	m0 = NULL;
2604     }
2605 noredhdropt:;
2606 	if (m0 != NULL) {
2607 		m_freem(m0);
2608 		m0 = NULL;
2609 	}
2610 
2611 	/* XXX: clear embedded link IDs in the inner header */
2612 	in6_clearscope(&sip6->ip6_src);
2613 	in6_clearscope(&sip6->ip6_dst);
2614 	in6_clearscope(&nd_rd->nd_rd_target);
2615 	in6_clearscope(&nd_rd->nd_rd_dst);
2616 
2617 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2618 
2619 	nd_rd->nd_rd_cksum = 0;
2620 	nd_rd->nd_rd_cksum
2621 		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2622 
2623 	/* send the packet to outside... */
2624 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2625 	if (outif) {
2626 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2627 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2628 	}
2629 	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2630 
2631 	return;
2632 
2633 fail:
2634 	if (m)
2635 		m_freem(m);
2636 	if (m0)
2637 		m_freem(m0);
2638 }
2639 
2640 #ifdef HAVE_NRL_INPCB
2641 #define in6pcb		inpcb
2642 #define in6p_icmp6filt	inp_icmp6filt
2643 #endif
2644 /*
2645  * ICMPv6 socket option processing.
2646  */
2647 void
2648 icmp6_ctloutput(netmsg_t msg)
2649 {
2650 	struct socket *so = msg->ctloutput.base.nm_so;
2651 	struct sockopt *sopt = msg->ctloutput.nm_sopt;
2652 	int error = 0;
2653 	int optlen;
2654 	struct inpcb *inp = so->so_pcb;
2655 	int level, op, optname;
2656 
2657 	if (sopt) {
2658 		level = sopt->sopt_level;
2659 		op = sopt->sopt_dir;
2660 		optname = sopt->sopt_name;
2661 		optlen = sopt->sopt_valsize;
2662 	} else
2663 		level = op = optname = optlen = 0;
2664 
2665 	if (level != IPPROTO_ICMPV6) {
2666 		error = EINVAL;
2667 		goto out;
2668 	}
2669 
2670 	switch (op) {
2671 	case PRCO_SETOPT:
2672 		switch (optname) {
2673 		case ICMP6_FILTER:
2674 		    {
2675 			if (optlen != sizeof(struct icmp6_filter)) {
2676 				error = EMSGSIZE;
2677 				break;
2678 			}
2679 			if (inp->in6p_icmp6filt == NULL) {
2680 				error = EINVAL;
2681 				break;
2682 			}
2683 			error = soopt_to_kbuf(sopt, inp->in6p_icmp6filt, optlen,
2684 				optlen);
2685 			break;
2686 		    }
2687 
2688 		default:
2689 			error = ENOPROTOOPT;
2690 			break;
2691 		}
2692 		break;
2693 
2694 	case PRCO_GETOPT:
2695 		switch (optname) {
2696 		case ICMP6_FILTER:
2697 		    {
2698 			if (inp->in6p_icmp6filt == NULL) {
2699 				error = EINVAL;
2700 				break;
2701 			}
2702 			soopt_from_kbuf(sopt, inp->in6p_icmp6filt,
2703 				sizeof(struct icmp6_filter));
2704 			break;
2705 		    }
2706 
2707 		default:
2708 			error = ENOPROTOOPT;
2709 			break;
2710 		}
2711 		break;
2712 	}
2713 out:
2714 	lwkt_replymsg(&msg->ctloutput.base.lmsg, error);
2715 }
2716 #ifdef HAVE_NRL_INPCB
2717 #undef in6pcb
2718 #undef in6p_icmp6filt
2719 #endif
2720 
2721 #ifndef HAVE_PPSRATECHECK
2722 #ifndef timersub
2723 #define	timersub(tvp, uvp, vvp)						\
2724 	do {								\
2725 		(vvp)->tv_sec = (tvp)->tv_sec - (uvp)->tv_sec;		\
2726 		(vvp)->tv_usec = (tvp)->tv_usec - (uvp)->tv_usec;	\
2727 		if ((vvp)->tv_usec < 0) {				\
2728 			(vvp)->tv_sec--;				\
2729 			(vvp)->tv_usec += 1000000;			\
2730 		}							\
2731 	} while (0)
2732 #endif
2733 
2734 /*
2735  * ppsratecheck(): packets (or events) per second limitation.
2736  */
2737 static int
2738 ppsratecheck(struct timeval *lasttime, int *curpps,
2739 	     int maxpps)	/* maximum pps allowed */
2740 {
2741 	struct timeval tv, delta;
2742 	int rv;
2743 
2744 	microtime(&tv);
2745 
2746 	timersub(&tv, lasttime, &delta);
2747 
2748 	/*
2749 	 * Check for 0,0 so that the message will be seen at least once.
2750 	 * If more than one second has passed since the last update of
2751 	 * lasttime, reset the counter.
2752 	 *
2753 	 * We do increment *curpps even in *curpps < maxpps case, as some may
2754 	 * try to use *curpps for stat purposes as well.
2755 	 */
2756 	if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) ||
2757 	    delta.tv_sec >= 1) {
2758 		*lasttime = tv;
2759 		*curpps = 0;
2760 		rv = 1;
2761 	} else if (maxpps < 0)
2762 		rv = 1;
2763 	else if (*curpps < maxpps)
2764 		rv = 1;
2765 	else
2766 		rv = 0;
2767 
2768 #if 1 /* DIAGNOSTIC? */
2769 	/* be careful about wrap-around */
2770 	if (*curpps + 1 > *curpps)
2771 		*curpps = *curpps + 1;
2772 #else
2773 	/*
2774 	 * assume that there's not too many calls to this function.
2775 	 * not sure if the assumption holds, as it depends on *caller's*
2776 	 * behavior, not the behavior of this function.
2777 	 * IMHO it is wrong to make assumption on the caller's behavior,
2778 	 * so the above #if is #if 1, not #ifdef DIAGNOSTIC.
2779 	 */
2780 	*curpps = *curpps + 1;
2781 #endif
2782 
2783 	return (rv);
2784 }
2785 #endif
2786 
2787 /*
2788  * Perform rate limit check.
2789  * Returns 0 if it is okay to send the icmp6 packet.
2790  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2791  * limitation.
2792  *
2793  * XXX per-destination/type check necessary?
2794  */
2795 static int
2796 icmp6_ratelimit(const struct in6_addr *dst,	/* not used at this moment */
2797 		const int type,			/* not used at this moment */
2798 		const int code)			/* not used at this moment */
2799 {
2800 	int ret;
2801 
2802 	ret = 0;	/* okay to send */
2803 
2804 	/* PPS limit */
2805 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2806 	    icmp6errppslim)) {
2807 		/* The packet is subject to rate limit */
2808 		ret++;
2809 	}
2810 
2811 	return ret;
2812 }
2813