xref: /dragonfly/sys/netinet6/icmp6.c (revision 678e8cc6)
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. All advertising materials mentioning features or use of this software
46  *    must display the following acknowledgement:
47  *	This product includes software developed by the University of
48  *	California, Berkeley and its contributors.
49  * 4. Neither the name of the University nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63  * SUCH DAMAGE.
64  *
65  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
66  */
67 
68 #include "opt_inet.h"
69 #include "opt_inet6.h"
70 #include "opt_ipsec.h"
71 
72 #include <sys/param.h>
73 #include <sys/systm.h>
74 #include <sys/malloc.h>
75 #include <sys/mbuf.h>
76 #include <sys/protosw.h>
77 #include <sys/socket.h>
78 #include <sys/socketvar.h>
79 #include <sys/socketops.h>
80 #include <sys/time.h>
81 #include <sys/kernel.h>
82 #include <sys/syslog.h>
83 #include <sys/domain.h>
84 
85 #include <sys/thread2.h>
86 #include <sys/msgport2.h>
87 
88 #include <net/if.h>
89 #include <net/route.h>
90 #include <net/if_dl.h>
91 #include <net/if_types.h>
92 
93 #include <netinet/in.h>
94 #include <netinet/in_var.h>
95 #include <netinet/ip6.h>
96 #include <netinet6/ip6_var.h>
97 #include <netinet/icmp6.h>
98 #include <netinet6/mld6_var.h>
99 #include <netinet/in_pcb.h>
100 #include <netinet6/in6_pcb.h>
101 #include <netinet6/nd6.h>
102 #include <netinet6/in6_ifattach.h>
103 #include <netinet6/ip6protosw.h>
104 
105 #ifdef IPSEC
106 #include <netinet6/ipsec.h>
107 #include <netproto/key/key.h>
108 #endif
109 
110 #ifdef FAST_IPSEC
111 #include <netproto/ipsec/ipsec.h>
112 #include <netproto/ipsec/key.h>
113 #define	IPSEC
114 #endif
115 
116 #include <net/net_osdep.h>
117 
118 #ifdef HAVE_NRL_INPCB
119 /* inpcb members */
120 #define in6pcb		inpcb
121 #define in6p_laddr	inp_laddr6
122 #define in6p_faddr	inp_faddr6
123 #define in6p_icmp6filt	inp_icmp6filt
124 #define in6p_route	inp_route
125 #define in6p_socket	inp_socket
126 #define in6p_flags	inp_flags
127 #define in6p_moptions	inp_moptions6
128 #define in6p_outputopts	inp_outputopts6
129 #define in6p_ip6	inp_ipv6
130 #define in6p_flowinfo	inp_flowinfo
131 #define in6p_sp		inp_sp
132 #define in6p_next	inp_next
133 #define in6p_prev	inp_prev
134 /* function names */
135 #define in6_pcbdetach	in_pcbdetach
136 #define in6_rtchange	in_rtchange
137 
138 /*
139  * for KAME src sync over BSD*'s. XXX: FreeBSD (>=3) are VERY different from
140  * others...
141  */
142 #define in6p_ip6_nxt	inp_ipv6.ip6_nxt
143 #endif
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 #ifdef COMPAT_RFC1885
180 static struct route_in6 icmp6_reflect_rt;
181 #endif
182 
183 
184 void
185 icmp6_init(void)
186 {
187 	mld6_init();
188 }
189 
190 static void
191 icmp6_errcount(struct icmp6errstat *stat, int type, int code)
192 {
193 	switch (type) {
194 	case ICMP6_DST_UNREACH:
195 		switch (code) {
196 		case ICMP6_DST_UNREACH_NOROUTE:
197 			stat->icp6errs_dst_unreach_noroute++;
198 			return;
199 		case ICMP6_DST_UNREACH_ADMIN:
200 			stat->icp6errs_dst_unreach_admin++;
201 			return;
202 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
203 			stat->icp6errs_dst_unreach_beyondscope++;
204 			return;
205 		case ICMP6_DST_UNREACH_ADDR:
206 			stat->icp6errs_dst_unreach_addr++;
207 			return;
208 		case ICMP6_DST_UNREACH_NOPORT:
209 			stat->icp6errs_dst_unreach_noport++;
210 			return;
211 		}
212 		break;
213 	case ICMP6_PACKET_TOO_BIG:
214 		stat->icp6errs_packet_too_big++;
215 		return;
216 	case ICMP6_TIME_EXCEEDED:
217 		switch (code) {
218 		case ICMP6_TIME_EXCEED_TRANSIT:
219 			stat->icp6errs_time_exceed_transit++;
220 			return;
221 		case ICMP6_TIME_EXCEED_REASSEMBLY:
222 			stat->icp6errs_time_exceed_reassembly++;
223 			return;
224 		}
225 		break;
226 	case ICMP6_PARAM_PROB:
227 		switch (code) {
228 		case ICMP6_PARAMPROB_HEADER:
229 			stat->icp6errs_paramprob_header++;
230 			return;
231 		case ICMP6_PARAMPROB_NEXTHEADER:
232 			stat->icp6errs_paramprob_nextheader++;
233 			return;
234 		case ICMP6_PARAMPROB_OPTION:
235 			stat->icp6errs_paramprob_option++;
236 			return;
237 		}
238 		break;
239 	case ND_REDIRECT:
240 		stat->icp6errs_redirect++;
241 		return;
242 	}
243 	stat->icp6errs_unknown++;
244 }
245 
246 /*
247  * Generate an error packet of type error in response to bad IP6 packet.
248  */
249 void
250 icmp6_error(struct mbuf *m, int type, int code, int param)
251 {
252 	struct ip6_hdr *oip6, *nip6;
253 	struct icmp6_hdr *icmp6;
254 	u_int preplen;
255 	int off;
256 	int nxt;
257 
258 	icmp6stat.icp6s_error++;
259 
260 	/* count per-type-code statistics */
261 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
262 
263 #ifdef M_DECRYPTED	/*not openbsd*/
264 	if (m->m_flags & M_DECRYPTED) {
265 		icmp6stat.icp6s_canterror++;
266 		goto freeit;
267 	}
268 #endif
269 
270 #ifndef PULLDOWN_TEST
271 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
272 #else
273 	if (m->m_len < sizeof(struct ip6_hdr)) {
274 		m = m_pullup(m, sizeof(struct ip6_hdr));
275 		if (m == NULL)
276 			return;
277 	}
278 #endif
279 	oip6 = mtod(m, struct ip6_hdr *);
280 
281 	/*
282 	 * Multicast destination check. For unrecognized option errors,
283 	 * this check has already done in ip6_unknown_opt(), so we can
284 	 * check only for other errors.
285 	 */
286 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
287 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
288 	    (type != ICMP6_PACKET_TOO_BIG &&
289 	     (type != ICMP6_PARAM_PROB ||
290 	      code != ICMP6_PARAMPROB_OPTION)))
291 		goto freeit;
292 
293 	/* Source address check. XXX: the case of anycast source? */
294 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
295 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
296 		goto freeit;
297 
298 	/*
299 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
300 	 * don't do it.
301 	 */
302 	nxt = -1;
303 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
304 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
305 		struct icmp6_hdr *icp;
306 
307 #ifndef PULLDOWN_TEST
308 		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
309 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
310 #else
311 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
312 			sizeof(*icp));
313 		if (icp == NULL) {
314 			icmp6stat.icp6s_tooshort++;
315 			return;
316 		}
317 #endif
318 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
319 		    icp->icmp6_type == ND_REDIRECT) {
320 			/*
321 			 * ICMPv6 error
322 			 * Special case: for redirect (which is
323 			 * informational) we must not send icmp6 error.
324 			 */
325 			icmp6stat.icp6s_canterror++;
326 			goto freeit;
327 		} else {
328 			/* ICMPv6 informational - send the error */
329 		}
330 	} else {
331 		/* non-ICMPv6 - send the error */
332 	}
333 
334 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
335 
336 	/* Finally, do rate limitation check. */
337 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
338 		icmp6stat.icp6s_toofreq++;
339 		goto freeit;
340 	}
341 
342 	/*
343 	 * OK, ICMP6 can be generated.
344 	 */
345 
346 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
347 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
348 
349 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
350 	M_PREPEND(m, preplen, MB_DONTWAIT);
351 	if (m && m->m_len < preplen)
352 		m = m_pullup(m, preplen);
353 	if (m == NULL) {
354 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
355 		return;
356 	}
357 
358 	nip6 = mtod(m, struct ip6_hdr *);
359 	nip6->ip6_src  = oip6->ip6_src;
360 	nip6->ip6_dst  = oip6->ip6_dst;
361 
362 	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src))
363 		oip6->ip6_src.s6_addr16[1] = 0;
364 	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst))
365 		oip6->ip6_dst.s6_addr16[1] = 0;
366 
367 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
368 	icmp6->icmp6_type = type;
369 	icmp6->icmp6_code = code;
370 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
371 
372 	/*
373 	 * icmp6_reflect() is designed to be in the input path.
374 	 * icmp6_error() can be called from both input and outut path,
375 	 * and if we are in output path rcvif could contain bogus value.
376 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
377 	 * information in ip header (nip6).
378 	 */
379 	m->m_pkthdr.rcvif = NULL;
380 
381 	icmp6stat.icp6s_outhist[type]++;
382 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
383 
384 	return;
385 
386 freeit:
387 	/*
388 	 * If we can't tell wheter or not we can generate ICMP6, free it.
389 	 */
390 	m_freem(m);
391 }
392 
393 /*
394  * Process a received ICMP6 message.
395  */
396 int
397 icmp6_input(struct mbuf **mp, int *offp, int proto)
398 {
399 	struct mbuf *m = *mp, *n;
400 	struct ip6_hdr *ip6, *nip6;
401 	struct icmp6_hdr *icmp6, *nicmp6;
402 	int off = *offp;
403 	int icmp6len = m->m_pkthdr.len - *offp;
404 	int code, sum, noff;
405 
406 #ifndef PULLDOWN_TEST
407 	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
408 	/* m might change if M_LOOP.  So, call mtod after this */
409 #endif
410 
411 	/*
412 	 * Locate icmp6 structure in mbuf, and check
413 	 * that not corrupted and of at least minimum length
414 	 */
415 
416 	ip6 = mtod(m, struct ip6_hdr *);
417 	if (icmp6len < sizeof(struct icmp6_hdr)) {
418 		icmp6stat.icp6s_tooshort++;
419 		goto freeit;
420 	}
421 
422 	/*
423 	 * calculate the checksum
424 	 */
425 #ifndef PULLDOWN_TEST
426 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
427 #else
428 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
429 	if (icmp6 == NULL) {
430 		icmp6stat.icp6s_tooshort++;
431 		return IPPROTO_DONE;
432 	}
433 #endif
434 	code = icmp6->icmp6_code;
435 
436 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
437 		nd6log((LOG_ERR,
438 		    "ICMP6 checksum error(%d|%x) %s\n",
439 		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
440 		icmp6stat.icp6s_checksum++;
441 		goto freeit;
442 	}
443 
444 	if (faithprefix_p != NULL && (*faithprefix_p)(&ip6->ip6_dst)) {
445 		/*
446 		 * Deliver very specific ICMP6 type only.
447 		 * This is important to deilver TOOBIG.  Otherwise PMTUD
448 		 * will not work.
449 		 */
450 		switch (icmp6->icmp6_type) {
451 		case ICMP6_DST_UNREACH:
452 		case ICMP6_PACKET_TOO_BIG:
453 		case ICMP6_TIME_EXCEEDED:
454 			break;
455 		default:
456 			goto freeit;
457 		}
458 	}
459 
460 	icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
461 	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
462 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
463 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
464 
465 	switch (icmp6->icmp6_type) {
466 	case ICMP6_DST_UNREACH:
467 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
468 		switch (code) {
469 		case ICMP6_DST_UNREACH_NOROUTE:
470 			code = PRC_UNREACH_NET;
471 			break;
472 		case ICMP6_DST_UNREACH_ADMIN:
473 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
474 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
475 			break;
476 		case ICMP6_DST_UNREACH_ADDR:
477 			code = PRC_HOSTDEAD;
478 			break;
479 #ifdef COMPAT_RFC1885
480 		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
481 			code = PRC_UNREACH_SRCFAIL;
482 			break;
483 #else
484 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
485 			/* I mean "source address was incorrect." */
486 			code = PRC_PARAMPROB;
487 			break;
488 #endif
489 		case ICMP6_DST_UNREACH_NOPORT:
490 			code = PRC_UNREACH_PORT;
491 			break;
492 		default:
493 			goto badcode;
494 		}
495 		goto deliver;
496 		break;
497 
498 	case ICMP6_PACKET_TOO_BIG:
499 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
500 		if (code != 0)
501 			goto badcode;
502 
503 		code = PRC_MSGSIZE;
504 
505 		/*
506 		 * Updating the path MTU will be done after examining
507 		 * intermediate extension headers.
508 		 */
509 		goto deliver;
510 		break;
511 
512 	case ICMP6_TIME_EXCEEDED:
513 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
514 		switch (code) {
515 		case ICMP6_TIME_EXCEED_TRANSIT:
516 		case ICMP6_TIME_EXCEED_REASSEMBLY:
517 			code += PRC_TIMXCEED_INTRANS;
518 			break;
519 		default:
520 			goto badcode;
521 		}
522 		goto deliver;
523 		break;
524 
525 	case ICMP6_PARAM_PROB:
526 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
527 		switch (code) {
528 		case ICMP6_PARAMPROB_NEXTHEADER:
529 			code = PRC_UNREACH_PROTOCOL;
530 			break;
531 		case ICMP6_PARAMPROB_HEADER:
532 		case ICMP6_PARAMPROB_OPTION:
533 			code = PRC_PARAMPROB;
534 			break;
535 		default:
536 			goto badcode;
537 		}
538 		goto deliver;
539 		break;
540 
541 	case ICMP6_ECHO_REQUEST:
542 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
543 		if (code != 0)
544 			goto badcode;
545 		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
546 			/* Give up remote */
547 			break;
548 		}
549 		if ((n->m_flags & M_EXT) ||
550 		    n->m_len < off + sizeof(struct icmp6_hdr)) {
551 			struct mbuf *n0 = n;
552 			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
553 			int n0len;
554 
555 			/*
556 			 * Prepare an internal mbuf. m_pullup() doesn't
557 			 * always copy the length we specified.
558 			 */
559 			if (maxlen >= MCLBYTES) {
560 				/* Give up remote */
561 				m_freem(n0);
562 				break;
563 			}
564 			n = m_getb(maxlen, MB_DONTWAIT, n0->m_type, M_PKTHDR);
565 			if (n == NULL) {
566 				/* Give up remote */
567 				m_freem(n0);
568 				break;
569 			}
570 			n0len = n0->m_pkthdr.len;	/* save for use below */
571 			M_MOVE_PKTHDR(n, n0);
572 			/*
573 			 * Copy IPv6 and ICMPv6 only.
574 			 */
575 			nip6 = mtod(n, struct ip6_hdr *);
576 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
577 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
578 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
579 			noff = sizeof(struct ip6_hdr);
580 			/* new mbuf contains only ipv6+icmpv6 headers */
581 			n->m_len = noff + sizeof(struct icmp6_hdr);
582 			/*
583 			 * Adjust mbuf. ip6_plen will be adjusted in
584 			 * ip6_output().
585 			 */
586 			m_adj(n0, off + sizeof(struct icmp6_hdr));
587 			/* recalculate complete packet size */
588 			n->m_pkthdr.len = n0len + (noff - off);
589 			n->m_next = n0;
590 		} else {
591 			nip6 = mtod(n, struct ip6_hdr *);
592 			nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
593 			noff = off;
594 		}
595 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
596 		nicmp6->icmp6_code = 0;
597 		if (n) {
598 			icmp6stat.icp6s_reflect++;
599 			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
600 			icmp6_reflect(n, noff);
601 		}
602 		break;
603 
604 	case ICMP6_ECHO_REPLY:
605 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
606 		if (code != 0)
607 			goto badcode;
608 		break;
609 
610 	case MLD_LISTENER_QUERY:
611 	case MLD_LISTENER_REPORT:
612 		if (icmp6len < sizeof(struct mld_hdr))
613 			goto badlen;
614 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
615 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
616 		else
617 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
618 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
619 			/* give up local */
620 			mld6_input(m, off);
621 			m = NULL;
622 			goto freeit;
623 		}
624 		mld6_input(n, off);
625 		/* m stays. */
626 		break;
627 
628 	case MLD_LISTENER_DONE:
629 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
630 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
631 			goto badlen;
632 		break;		/* nothing to be done in kernel */
633 
634 	case MLD_MTRACE_RESP:
635 	case MLD_MTRACE:
636 		/* XXX: these two are experimental.  not officially defind. */
637 		/* XXX: per-interface statistics? */
638 		break;		/* just pass it to applications */
639 
640 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
641 	    {
642 		enum { WRU, FQDN } mode;
643 
644 		if (!icmp6_nodeinfo)
645 			break;
646 
647 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
648 			mode = WRU;
649 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
650 			mode = FQDN;
651 		else
652 			goto badlen;
653 
654 #define hostnamelen	strlen(hostname)
655 		if (mode == FQDN) {
656 #ifndef PULLDOWN_TEST
657 			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
658 					 IPPROTO_DONE);
659 #endif
660 			n = m_copy(m, 0, M_COPYALL);
661 			if (n)
662 				n = ni6_input(n, off);
663 			/* XXX meaningless if n == NULL */
664 			noff = sizeof(struct ip6_hdr);
665 		} else {
666 			u_char *p;
667 			int maxlen, maxhlen;
668 
669 			if ((icmp6_nodeinfo & 5) != 5)
670 				break;
671 
672 			if (code != 0)
673 				goto badcode;
674 			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
675 			if (maxlen >= MCLBYTES) {
676 				/* Give up remote */
677 				break;
678 			}
679 			n = m_getb(maxlen, MB_DONTWAIT, m->m_type, M_PKTHDR);
680 			if (n == NULL) {
681 				/* Give up remote */
682 				break;
683 			}
684 			if (!m_dup_pkthdr(n, m, MB_DONTWAIT)) {
685 				/*
686 				 * Previous code did a blind M_COPY_PKTHDR
687 				 * and said "just for rcvif".  If true, then
688 				 * we could tolerate the dup failing (due to
689 				 * the deep copy of the tag chain).  For now
690 				 * be conservative and just fail.
691 				 */
692 				m_free(n);
693 				break;
694 			}
695 			n->m_len = 0;
696 			maxhlen = M_TRAILINGSPACE(n) - maxlen;
697 			if (maxhlen > hostnamelen)
698 				maxhlen = hostnamelen;
699 			/*
700 			 * Copy IPv6 and ICMPv6 only.
701 			 */
702 			nip6 = mtod(n, struct ip6_hdr *);
703 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
704 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
705 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
706 			p = (u_char *)(nicmp6 + 1);
707 			bzero(p, 4);
708 			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
709 			noff = sizeof(struct ip6_hdr);
710 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
711 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
712 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
713 			nicmp6->icmp6_code = 0;
714 		}
715 #undef hostnamelen
716 		if (n) {
717 			icmp6stat.icp6s_reflect++;
718 			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
719 			icmp6_reflect(n, noff);
720 		}
721 		break;
722 	    }
723 
724 	case ICMP6_WRUREPLY:
725 		if (code != 0)
726 			goto badcode;
727 		break;
728 
729 	case ND_ROUTER_SOLICIT:
730 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
731 		if (code != 0)
732 			goto badcode;
733 		if (icmp6len < sizeof(struct nd_router_solicit))
734 			goto badlen;
735 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
736 			/* give up local */
737 			nd6_rs_input(m, off, icmp6len);
738 			m = NULL;
739 			goto freeit;
740 		}
741 		nd6_rs_input(n, off, icmp6len);
742 		/* m stays. */
743 		break;
744 
745 	case ND_ROUTER_ADVERT:
746 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
747 		if (code != 0)
748 			goto badcode;
749 		if (icmp6len < sizeof(struct nd_router_advert))
750 			goto badlen;
751 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
752 			/* give up local */
753 			nd6_ra_input(m, off, icmp6len);
754 			m = NULL;
755 			goto freeit;
756 		}
757 		nd6_ra_input(n, off, icmp6len);
758 		/* m stays. */
759 		break;
760 
761 	case ND_NEIGHBOR_SOLICIT:
762 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
763 		if (code != 0)
764 			goto badcode;
765 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
766 			goto badlen;
767 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
768 			/* give up local */
769 			nd6_ns_input(m, off, icmp6len);
770 			m = NULL;
771 			goto freeit;
772 		}
773 		nd6_ns_input(n, off, icmp6len);
774 		/* m stays. */
775 		break;
776 
777 	case ND_NEIGHBOR_ADVERT:
778 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
779 		if (code != 0)
780 			goto badcode;
781 		if (icmp6len < sizeof(struct nd_neighbor_advert))
782 			goto badlen;
783 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
784 			/* give up local */
785 			nd6_na_input(m, off, icmp6len);
786 			m = NULL;
787 			goto freeit;
788 		}
789 		nd6_na_input(n, off, icmp6len);
790 		/* m stays. */
791 		break;
792 
793 	case ND_REDIRECT:
794 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
795 		if (code != 0)
796 			goto badcode;
797 		if (icmp6len < sizeof(struct nd_redirect))
798 			goto badlen;
799 		if ((n = m_copym(m, 0, M_COPYALL, MB_DONTWAIT)) == NULL) {
800 			/* give up local */
801 			icmp6_redirect_input(m, off);
802 			m = NULL;
803 			goto freeit;
804 		}
805 		icmp6_redirect_input(n, off);
806 		/* m stays. */
807 		break;
808 
809 	case ICMP6_ROUTER_RENUMBERING:
810 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
811 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
812 			goto badcode;
813 		if (icmp6len < sizeof(struct icmp6_router_renum))
814 			goto badlen;
815 		break;
816 
817 	default:
818 		nd6log((LOG_DEBUG,
819 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
820 		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
821 		    ip6_sprintf(&ip6->ip6_dst),
822 		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
823 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
824 			/* ICMPv6 error: MUST deliver it by spec... */
825 			code = PRC_NCMDS;
826 			/* deliver */
827 		} else {
828 			/* ICMPv6 informational: MUST not deliver */
829 			break;
830 		}
831 	deliver:
832 		if (icmp6_notify_error(m, off, icmp6len, code)) {
833 			/* In this case, m should've been freed. */
834 			return (IPPROTO_DONE);
835 		}
836 		break;
837 
838 	badcode:
839 		icmp6stat.icp6s_badcode++;
840 		break;
841 
842 	badlen:
843 		icmp6stat.icp6s_badlen++;
844 		break;
845 	}
846 
847 	/* deliver the packet to appropriate sockets */
848 	icmp6_rip6_input(&m, *offp);
849 
850 	return IPPROTO_DONE;
851 
852 freeit:
853 	m_freem(m);
854 	return IPPROTO_DONE;
855 }
856 
857 static int
858 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
859 {
860 	struct icmp6_hdr *icmp6;
861 	struct ip6_hdr *eip6;
862 	u_int32_t notifymtu;
863 	struct sockaddr_in6 icmp6src, icmp6dst;
864 
865 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
866 		icmp6stat.icp6s_tooshort++;
867 		goto freeit;
868 	}
869 #ifndef PULLDOWN_TEST
870 	IP6_EXTHDR_CHECK(m, off,
871 			 sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr),
872 			 -1);
873 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
874 #else
875 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
876 		       sizeof(*icmp6) + sizeof(struct ip6_hdr));
877 	if (icmp6 == NULL) {
878 		icmp6stat.icp6s_tooshort++;
879 		return (-1);
880 	}
881 #endif
882 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
883 
884 	/* Detect the upper level protocol */
885 	{
886 		u_int8_t nxt = eip6->ip6_nxt;
887 		int eoff = off + sizeof(struct icmp6_hdr) +
888 			sizeof(struct ip6_hdr);
889 		struct ip6ctlparam ip6cp;
890 		struct in6_addr *finaldst = NULL;
891 		int icmp6type = icmp6->icmp6_type;
892 		struct ip6_frag *fh;
893 		struct ip6_rthdr *rth;
894 		int rthlen;
895 
896 		while (1) { /* XXX: should avoid infinite loop explicitly? */
897 			struct ip6_ext *eh;
898 
899 			switch (nxt) {
900 			case IPPROTO_HOPOPTS:
901 			case IPPROTO_DSTOPTS:
902 			case IPPROTO_AH:
903 #ifndef PULLDOWN_TEST
904 				IP6_EXTHDR_CHECK(m, 0, eoff +
905 						 sizeof(struct ip6_ext),
906 						 -1);
907 				eh = (struct ip6_ext *)(mtod(m, caddr_t)
908 							+ eoff);
909 #else
910 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
911 					       eoff, sizeof(*eh));
912 				if (eh == NULL) {
913 					icmp6stat.icp6s_tooshort++;
914 					return (-1);
915 				}
916 #endif
917 
918 				if (nxt == IPPROTO_AH)
919 					eoff += (eh->ip6e_len + 2) << 2;
920 				else
921 					eoff += (eh->ip6e_len + 1) << 3;
922 				nxt = eh->ip6e_nxt;
923 				break;
924 			case IPPROTO_ROUTING:
925 				/*
926 				 * When the erroneous packet contains a
927 				 * routing header, we should examine the
928 				 * header to determine the final destination.
929 				 * Otherwise, we can't properly update
930 				 * information that depends on the final
931 				 * destination (e.g. path MTU).
932 				 */
933 #ifndef PULLDOWN_TEST
934 				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth),
935 						 -1);
936 				rth = (struct ip6_rthdr *)(mtod(m, caddr_t)
937 							   + eoff);
938 #else
939 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
940 					       eoff, sizeof(*rth));
941 				if (rth == NULL) {
942 					icmp6stat.icp6s_tooshort++;
943 					return (-1);
944 				}
945 #endif
946 				rthlen = (rth->ip6r_len + 1) << 3;
947 				eoff += rthlen;
948 				nxt = rth->ip6r_nxt;
949 				break;
950 			case IPPROTO_FRAGMENT:
951 #ifndef PULLDOWN_TEST
952 				IP6_EXTHDR_CHECK(m, 0, eoff +
953 						 sizeof(struct ip6_frag),
954 						 -1);
955 				fh = (struct ip6_frag *)(mtod(m, caddr_t)
956 							 + eoff);
957 #else
958 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
959 					       eoff, sizeof(*fh));
960 				if (fh == NULL) {
961 					icmp6stat.icp6s_tooshort++;
962 					return (-1);
963 				}
964 #endif
965 				/*
966 				 * Data after a fragment header is meaningless
967 				 * unless it is the first fragment, but
968 				 * we'll go to the notify label for path MTU
969 				 * discovery.
970 				 */
971 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
972 					goto notify;
973 
974 				eoff += sizeof(struct ip6_frag);
975 				nxt = fh->ip6f_nxt;
976 				break;
977 			default:
978 				/*
979 				 * This case includes ESP and the No Next
980 				 * Header.  In such cases going to the notify
981 				 * label does not have any meaning
982 				 * (i.e. pr_ctlinput will be NULL), but we go
983 				 * anyway since we might have to update
984 				 * path MTU information.
985 				 */
986 				goto notify;
987 			}
988 		}
989 	  notify:
990 #ifndef PULLDOWN_TEST
991 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
992 #else
993 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
994 			       sizeof(*icmp6) + sizeof(struct ip6_hdr));
995 		if (icmp6 == NULL) {
996 			icmp6stat.icp6s_tooshort++;
997 			return (-1);
998 		}
999 #endif
1000 
1001 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1002 		bzero(&icmp6dst, sizeof(icmp6dst));
1003 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1004 		icmp6dst.sin6_family = AF_INET6;
1005 		if (finaldst == NULL)
1006 			icmp6dst.sin6_addr = eip6->ip6_dst;
1007 		else
1008 			icmp6dst.sin6_addr = *finaldst;
1009 		icmp6dst.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1010 							  &icmp6dst.sin6_addr);
1011 		if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
1012 				   NULL, NULL)) {
1013 			/* should be impossbile */
1014 			nd6log((LOG_DEBUG,
1015 			    "icmp6_notify_error: in6_embedscope failed\n"));
1016 			goto freeit;
1017 		}
1018 
1019 		/*
1020 		 * retrieve parameters from the inner IPv6 header, and convert
1021 		 * them into sockaddr structures.
1022 		 */
1023 		bzero(&icmp6src, sizeof(icmp6src));
1024 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1025 		icmp6src.sin6_family = AF_INET6;
1026 		icmp6src.sin6_addr = eip6->ip6_src;
1027 		icmp6src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1028 							  &icmp6src.sin6_addr);
1029 		if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1030 				   NULL, NULL)) {
1031 			/* should be impossbile */
1032 			nd6log((LOG_DEBUG,
1033 			    "icmp6_notify_error: in6_embedscope failed\n"));
1034 			goto freeit;
1035 		}
1036 		icmp6src.sin6_flowinfo =
1037 			(eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1038 
1039 		if (finaldst == NULL)
1040 			finaldst = &eip6->ip6_dst;
1041 		ip6cp.ip6c_m = m;
1042 		ip6cp.ip6c_icmp6 = icmp6;
1043 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1044 		ip6cp.ip6c_off = eoff;
1045 		ip6cp.ip6c_finaldst = finaldst;
1046 		ip6cp.ip6c_src = &icmp6src;
1047 		ip6cp.ip6c_nxt = nxt;
1048 
1049 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1050 			notifymtu = ntohl(icmp6->icmp6_mtu);
1051 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1052 			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1053 		}
1054 
1055 		so_pru_ctlinput(
1056 			(struct protosw *)&inet6sw[ip6_protox[nxt]],
1057 			code, (struct sockaddr *)&icmp6dst, &ip6cp);
1058 	}
1059 	return (0);
1060 
1061 freeit:
1062 	m_freem(m);
1063 	return (-1);
1064 }
1065 
1066 void
1067 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1068 {
1069 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1070 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1071 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1072 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1073 	struct rtentry *rt = NULL;
1074 	struct sockaddr_in6 sin6;
1075 
1076 	if (!validated)
1077 		return;
1078 
1079 	bzero(&sin6, sizeof(sin6));
1080 	sin6.sin6_family = PF_INET6;
1081 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1082 	sin6.sin6_addr = *dst;
1083 	/* XXX normally, this won't happen */
1084 	if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1085 		sin6.sin6_addr.s6_addr16[1] =
1086 		    htons(m->m_pkthdr.rcvif->if_index);
1087 	}
1088 	/* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1089 	rt = rtpurelookup((struct sockaddr *)&sin6);
1090 	if (rt != NULL && (rt->rt_flags & RTF_HOST) &&
1091 	    !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1092 		if (mtu < IPV6_MMTU) {				/* XXX */
1093 			rt->rt_rmx.rmx_locks |= RTV_MTU;
1094 		} else if (mtu < rt->rt_ifp->if_mtu &&
1095 			   rt->rt_rmx.rmx_mtu > mtu) {
1096 			icmp6stat.icp6s_pmtuchg++;
1097 			rt->rt_rmx.rmx_mtu = mtu;
1098 		}
1099 	}
1100 	if (rt)
1101 		--rt->rt_refcnt;
1102 }
1103 
1104 /*
1105  * Process a Node Information Query packet, based on
1106  * draft-ietf-ipngwg-icmp-name-lookups-07.
1107  *
1108  * Spec incompatibilities:
1109  * - IPv6 Subject address handling
1110  * - IPv4 Subject address handling support missing
1111  * - Proxy reply (answer even if it's not for me)
1112  * - joins NI group address at in6_ifattach() time only, does not cope
1113  *   with hostname changes by sethostname(3)
1114  */
1115 #define hostnamelen	strlen(hostname)
1116 static struct mbuf *
1117 ni6_input(struct mbuf *m, int off)
1118 {
1119 	struct icmp6_nodeinfo *ni6, *nni6;
1120 	struct mbuf *n = NULL;
1121 	u_int16_t qtype;
1122 	int subjlen;
1123 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1124 	struct ni_reply_fqdn *fqdn;
1125 	int addrs;		/* for NI_QTYPE_NODEADDR */
1126 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1127 	struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1128 	struct sockaddr_in6 sin6_d; /* XXX: we should retrieve this from m_aux */
1129 	struct ip6_hdr *ip6;
1130 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1131 	char *subj = NULL;
1132 	struct in6_ifaddr *ia6 = NULL;
1133 
1134 	ip6 = mtod(m, struct ip6_hdr *);
1135 #ifndef PULLDOWN_TEST
1136 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1137 #else
1138 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1139 	if (ni6 == NULL) {
1140 		/* m is already reclaimed */
1141 		return NULL;
1142 	}
1143 #endif
1144 
1145 	/*
1146 	 * Validate IPv6 destination address.
1147 	 *
1148 	 * The Responder must discard the Query without further processing
1149 	 * unless it is one of the Responder's unicast or anycast addresses, or
1150 	 * a link-local scope multicast address which the Responder has joined.
1151 	 * [icmp-name-lookups-07, Section 4.]
1152 	 */
1153 	bzero(&sin6, sizeof(sin6));
1154 	sin6.sin6_family = AF_INET6;
1155 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1156 	bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1157 	/* XXX scopeid */
1158 	if ((ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)&sin6)) != NULL) {
1159 		/* unicast/anycast, fine */
1160 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1161 		    (icmp6_nodeinfo & 4) == 0) {
1162 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1163 				"a temporary address in %s:%d",
1164 			       __FILE__, __LINE__));
1165 			goto bad;
1166 		}
1167 	} else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1168 		; /* link-local multicast, fine */
1169 	else
1170 		goto bad;
1171 
1172 	/* validate query Subject field. */
1173 	qtype = ntohs(ni6->ni_qtype);
1174 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1175 	switch (qtype) {
1176 	case NI_QTYPE_NOOP:
1177 	case NI_QTYPE_SUPTYPES:
1178 		/* 07 draft */
1179 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1180 			break;
1181 		/* FALLTHROUGH */
1182 	case NI_QTYPE_FQDN:
1183 	case NI_QTYPE_NODEADDR:
1184 		switch (ni6->ni_code) {
1185 		case ICMP6_NI_SUBJ_IPV6:
1186 #if ICMP6_NI_SUBJ_IPV6 != 0
1187 		case 0:
1188 #endif
1189 			/*
1190 			 * backward compatibility - try to accept 03 draft
1191 			 * format, where no Subject is present.
1192 			 */
1193 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1194 			    subjlen == 0) {
1195 				oldfqdn++;
1196 				break;
1197 			}
1198 #if ICMP6_NI_SUBJ_IPV6 != 0
1199 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1200 				goto bad;
1201 #endif
1202 
1203 			if (subjlen != sizeof(sin6.sin6_addr))
1204 				goto bad;
1205 
1206 			/*
1207 			 * Validate Subject address.
1208 			 *
1209 			 * Not sure what exactly "address belongs to the node"
1210 			 * means in the spec, is it just unicast, or what?
1211 			 *
1212 			 * At this moment we consider Subject address as
1213 			 * "belong to the node" if the Subject address equals
1214 			 * to the IPv6 destination address; validation for
1215 			 * IPv6 destination address should have done enough
1216 			 * check for us.
1217 			 *
1218 			 * We do not do proxy at this moment.
1219 			 */
1220 			/* m_pulldown instead of copy? */
1221 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1222 			    subjlen, (caddr_t)&sin6.sin6_addr);
1223 			sin6.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1224 							      &sin6.sin6_addr);
1225 			in6_embedscope(&sin6.sin6_addr, &sin6, NULL, NULL);
1226 			bzero(&sin6_d, sizeof(sin6_d));
1227 			sin6_d.sin6_family = AF_INET6; /* not used, actually */
1228 			sin6_d.sin6_len = sizeof(sin6_d); /* ditto */
1229 			sin6_d.sin6_addr = ip6->ip6_dst;
1230 			sin6_d.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
1231 								&ip6->ip6_dst);
1232 			in6_embedscope(&sin6_d.sin6_addr, &sin6_d, NULL, NULL);
1233 			subj = (char *)&sin6;
1234 			if (SA6_ARE_ADDR_EQUAL(&sin6, &sin6_d))
1235 				break;
1236 
1237 			/*
1238 			 * XXX if we are to allow other cases, we should really
1239 			 * be careful about scope here.
1240 			 * basically, we should disallow queries toward IPv6
1241 			 * destination X with subject Y, if scope(X) > scope(Y).
1242 			 * if we allow scope(X) > scope(Y), it will result in
1243 			 * information leakage across scope boundary.
1244 			 */
1245 			goto bad;
1246 
1247 		case ICMP6_NI_SUBJ_FQDN:
1248 			/*
1249 			 * Validate Subject name with gethostname(3).
1250 			 *
1251 			 * The behavior may need some debate, since:
1252 			 * - we are not sure if the node has FQDN as
1253 			 *   hostname (returned by gethostname(3)).
1254 			 * - the code does wildcard match for truncated names.
1255 			 *   however, we are not sure if we want to perform
1256 			 *   wildcard match, if gethostname(3) side has
1257 			 *   truncated hostname.
1258 			 */
1259 			n = ni6_nametodns(hostname, hostnamelen, 0);
1260 			if (!n || n->m_next || n->m_len == 0)
1261 				goto bad;
1262 			IP6_EXTHDR_GET(subj, char *, m,
1263 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1264 			if (subj == NULL)
1265 				goto bad;
1266 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1267 					n->m_len)) {
1268 				goto bad;
1269 			}
1270 			m_freem(n);
1271 			n = NULL;
1272 			break;
1273 
1274 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1275 		default:
1276 			goto bad;
1277 		}
1278 		break;
1279 	}
1280 
1281 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1282 	switch (qtype) {
1283 	case NI_QTYPE_FQDN:
1284 		if ((icmp6_nodeinfo & 1) == 0)
1285 			goto bad;
1286 		break;
1287 	case NI_QTYPE_NODEADDR:
1288 		if ((icmp6_nodeinfo & 2) == 0)
1289 			goto bad;
1290 		break;
1291 	}
1292 
1293 	/* guess reply length */
1294 	switch (qtype) {
1295 	case NI_QTYPE_NOOP:
1296 		break;		/* no reply data */
1297 	case NI_QTYPE_SUPTYPES:
1298 		replylen += sizeof(u_int32_t);
1299 		break;
1300 	case NI_QTYPE_FQDN:
1301 		/* XXX will append an mbuf */
1302 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1303 		break;
1304 	case NI_QTYPE_NODEADDR:
1305 		addrs = ni6_addrs(ni6, m, &ifp, subj);
1306 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1307 					  sizeof(u_int32_t))) > MCLBYTES)
1308 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1309 		break;
1310 	default:
1311 		/*
1312 		 * XXX: We must return a reply with the ICMP6 code
1313 		 * `unknown Qtype' in this case. However we regard the case
1314 		 * as an FQDN query for backward compatibility.
1315 		 * Older versions set a random value to this field,
1316 		 * so it rarely varies in the defined qtypes.
1317 		 * But the mechanism is not reliable...
1318 		 * maybe we should obsolete older versions.
1319 		 */
1320 		qtype = NI_QTYPE_FQDN;
1321 		/* XXX will append an mbuf */
1322 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1323 		oldfqdn++;
1324 		break;
1325 	}
1326 
1327 	/* allocate an mbuf to reply. */
1328 	if (replylen > MCLBYTES) {
1329 		/*
1330 		 * XXX: should we try to allocate more? But MCLBYTES
1331 		 * is probably much larger than IPV6_MMTU...
1332 		 */
1333 		goto bad;
1334 	}
1335 	n = m_getb(replylen, MB_DONTWAIT, m->m_type, M_PKTHDR);
1336 	if (n == NULL) {
1337 		m_freem(m);
1338 		return (NULL);
1339 	}
1340 	M_MOVE_PKTHDR(n, m); /* just for recvif */
1341 	n->m_pkthdr.len = n->m_len = replylen;
1342 
1343 	/* copy mbuf header and IPv6 + Node Information base headers */
1344 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1345 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1346 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1347 
1348 	/* qtype dependent procedure */
1349 	switch (qtype) {
1350 	case NI_QTYPE_NOOP:
1351 		nni6->ni_code = ICMP6_NI_SUCCESS;
1352 		nni6->ni_flags = 0;
1353 		break;
1354 	case NI_QTYPE_SUPTYPES:
1355 	{
1356 		u_int32_t v;
1357 		nni6->ni_code = ICMP6_NI_SUCCESS;
1358 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1359 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1360 		v = (u_int32_t)htonl(0x0000000f);
1361 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1362 		break;
1363 	}
1364 	case NI_QTYPE_FQDN:
1365 		nni6->ni_code = ICMP6_NI_SUCCESS;
1366 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1367 						sizeof(struct ip6_hdr) +
1368 						sizeof(struct icmp6_nodeinfo));
1369 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1370 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1371 		/*
1372 		 * XXX do we really have FQDN in variable "hostname"?
1373 		 */
1374 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1375 		if (n->m_next == NULL)
1376 			goto bad;
1377 		/* XXX we assume that n->m_next is not a chain */
1378 		if (n->m_next->m_next != NULL)
1379 			goto bad;
1380 		n->m_pkthdr.len += n->m_next->m_len;
1381 		break;
1382 	case NI_QTYPE_NODEADDR:
1383 	{
1384 		int lenlim, copied;
1385 
1386 		nni6->ni_code = ICMP6_NI_SUCCESS;
1387 		n->m_pkthdr.len = n->m_len =
1388 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1389 		lenlim = M_TRAILINGSPACE(n);
1390 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1391 		/* XXX: reset mbuf length */
1392 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1393 			sizeof(struct icmp6_nodeinfo) + copied;
1394 		break;
1395 	}
1396 	default:
1397 		break;		/* XXX impossible! */
1398 	}
1399 
1400 	nni6->ni_type = ICMP6_NI_REPLY;
1401 	m_freem(m);
1402 	return (n);
1403 
1404 bad:
1405 	m_freem(m);
1406 	if (n)
1407 		m_freem(n);
1408 	return (NULL);
1409 }
1410 #undef hostnamelen
1411 
1412 /*
1413  * make a mbuf with DNS-encoded string.  no compression support.
1414  *
1415  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1416  * treated as truncated name (two \0 at the end).  this is a wild guess.
1417  */
1418 static struct mbuf *
1419 ni6_nametodns(const char *name, int namelen,
1420 	      int old) /* return pascal string if non-zero */
1421 {
1422 	struct mbuf *m;
1423 	char *cp, *ep;
1424 	const char *p, *q;
1425 	int i, len, nterm;
1426 
1427 	if (old)
1428 		len = namelen + 1;
1429 	else
1430 		len = MCLBYTES;
1431 
1432 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1433 	m = m_getb(len, MB_DONTWAIT, MT_DATA, 0);
1434 	if (!m)
1435 		goto fail;
1436 
1437 	if (old) {
1438 		m->m_len = len;
1439 		*mtod(m, char *) = namelen;
1440 		bcopy(name, mtod(m, char *) + 1, namelen);
1441 		return m;
1442 	} else {
1443 		m->m_len = 0;
1444 		cp = mtod(m, char *);
1445 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1446 
1447 		/* if not certain about my name, return empty buffer */
1448 		if (namelen == 0)
1449 			return m;
1450 
1451 		/*
1452 		 * guess if it looks like shortened hostname, or FQDN.
1453 		 * shortened hostname needs two trailing "\0".
1454 		 */
1455 		i = 0;
1456 		for (p = name; p < name + namelen; p++) {
1457 			if (*p && *p == '.')
1458 				i++;
1459 		}
1460 		if (i < 2)
1461 			nterm = 2;
1462 		else
1463 			nterm = 1;
1464 
1465 		p = name;
1466 		while (cp < ep && p < name + namelen) {
1467 			i = 0;
1468 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1469 				i++;
1470 			/* result does not fit into mbuf */
1471 			if (cp + i + 1 >= ep)
1472 				goto fail;
1473 			/*
1474 			 * DNS label length restriction, RFC1035 page 8.
1475 			 * "i == 0" case is included here to avoid returning
1476 			 * 0-length label on "foo..bar".
1477 			 */
1478 			if (i <= 0 || i >= 64)
1479 				goto fail;
1480 			*cp++ = i;
1481 			bcopy(p, cp, i);
1482 			cp += i;
1483 			p = q;
1484 			if (p < name + namelen && *p == '.')
1485 				p++;
1486 		}
1487 		/* termination */
1488 		if (cp + nterm >= ep)
1489 			goto fail;
1490 		while (nterm-- > 0)
1491 			*cp++ = '\0';
1492 		m->m_len = cp - mtod(m, char *);
1493 		return m;
1494 	}
1495 
1496 	panic("should not reach here");
1497 	/* NOTREACHED */
1498 
1499 fail:
1500 	if (m)
1501 		m_freem(m);
1502 	return NULL;
1503 }
1504 
1505 /*
1506  * check if two DNS-encoded string matches.  takes care of truncated
1507  * form (with \0\0 at the end).  no compression support.
1508  * XXX upper/lowercase match (see RFC2065)
1509  */
1510 static int
1511 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1512 {
1513 	const char *a0, *b0;
1514 	int l;
1515 
1516 	/* simplest case - need validation? */
1517 	if (alen == blen && bcmp(a, b, alen) == 0)
1518 		return 1;
1519 
1520 	a0 = a;
1521 	b0 = b;
1522 
1523 	/* termination is mandatory */
1524 	if (alen < 2 || blen < 2)
1525 		return 0;
1526 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1527 		return 0;
1528 	alen--;
1529 	blen--;
1530 
1531 	while (a - a0 < alen && b - b0 < blen) {
1532 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1533 			return 0;
1534 
1535 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1536 			return 0;
1537 		/* we don't support compression yet */
1538 		if (a[0] >= 64 || b[0] >= 64)
1539 			return 0;
1540 
1541 		/* truncated case */
1542 		if (a[0] == 0 && a - a0 == alen - 1)
1543 			return 1;
1544 		if (b[0] == 0 && b - b0 == blen - 1)
1545 			return 1;
1546 		if (a[0] == 0 || b[0] == 0)
1547 			return 0;
1548 
1549 		if (a[0] != b[0])
1550 			return 0;
1551 		l = a[0];
1552 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1553 			return 0;
1554 		if (bcmp(a + 1, b + 1, l) != 0)
1555 			return 0;
1556 
1557 		a += 1 + l;
1558 		b += 1 + l;
1559 	}
1560 
1561 	if (a - a0 == alen && b - b0 == blen)
1562 		return 1;
1563 	else
1564 		return 0;
1565 }
1566 
1567 /*
1568  * calculate the number of addresses to be returned in the node info reply.
1569  */
1570 static int
1571 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1572 	  char *subj)
1573 {
1574 	struct ifnet *ifp;
1575 	struct in6_ifaddr *ifa6;
1576 	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1577 	int addrs = 0, addrsofif, iffound = 0;
1578 	int niflags = ni6->ni_flags;
1579 
1580 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1581 		switch (ni6->ni_code) {
1582 		case ICMP6_NI_SUBJ_IPV6:
1583 			if (subj == NULL) /* must be impossible... */
1584 				return (0);
1585 			subj_ip6 = (struct sockaddr_in6 *)subj;
1586 			break;
1587 		default:
1588 			/*
1589 			 * XXX: we only support IPv6 subject address for
1590 			 * this Qtype.
1591 			 */
1592 			return (0);
1593 		}
1594 	}
1595 
1596 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
1597 	{
1598 		struct ifaddr_container *ifac;
1599 
1600 		addrsofif = 0;
1601 		TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link)
1602 		{
1603 			struct ifaddr *ifa = ifac->ifa;
1604 
1605 			if (ifa->ifa_addr->sa_family != AF_INET6)
1606 				continue;
1607 			ifa6 = (struct in6_ifaddr *)ifa;
1608 
1609 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1610 			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1611 					       &ifa6->ia_addr.sin6_addr))
1612 				iffound = 1;
1613 
1614 			/*
1615 			 * IPv4-mapped addresses can only be returned by a
1616 			 * Node Information proxy, since they represent
1617 			 * addresses of IPv4-only nodes, which perforce do
1618 			 * not implement this protocol.
1619 			 * [icmp-name-lookups-07, Section 5.4]
1620 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1621 			 * this function at this moment.
1622 			 */
1623 
1624 			/* What do we have to do about ::1? */
1625 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1626 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1627 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1628 					continue;
1629 				break;
1630 			case IPV6_ADDR_SCOPE_SITELOCAL:
1631 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1632 					continue;
1633 				break;
1634 			case IPV6_ADDR_SCOPE_GLOBAL:
1635 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1636 					continue;
1637 				break;
1638 			default:
1639 				continue;
1640 			}
1641 
1642 			/*
1643 			 * check if anycast is okay.
1644 			 * XXX: just experimental.  not in the spec.
1645 			 */
1646 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) &&
1647 			    !(niflags & NI_NODEADDR_FLAG_ANYCAST))
1648 				continue; /* we need only unicast addresses */
1649 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) &&
1650 			    (icmp6_nodeinfo & 4) == 0) {
1651 				continue;
1652 			}
1653 			addrsofif++; /* count the address */
1654 		}
1655 		if (iffound) {
1656 			*ifpp = ifp;
1657 			return (addrsofif);
1658 		}
1659 
1660 		addrs += addrsofif;
1661 	}
1662 
1663 	return (addrs);
1664 }
1665 
1666 static int
1667 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1668 		struct ifnet *ifp0, int resid)
1669 {
1670 	struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
1671 	struct in6_ifaddr *ifa6;
1672 	struct ifnet *ifp_dep = NULL;
1673 	int copied = 0, allow_deprecated = 0;
1674 	u_char *cp = (u_char *)(nni6 + 1);
1675 	int niflags = ni6->ni_flags;
1676 	u_int32_t ltime;
1677 
1678 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1679 		return (0);	/* needless to copy */
1680 
1681 again:
1682 
1683 	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list))
1684 	{
1685 		struct ifaddr_container *ifac;
1686 
1687 		TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1688 			struct ifaddr *ifa = ifac->ifa;
1689 
1690 			if (ifa->ifa_addr->sa_family != AF_INET6)
1691 				continue;
1692 			ifa6 = (struct in6_ifaddr *)ifa;
1693 
1694 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) &&
1695 			    allow_deprecated == 0) {
1696 				/*
1697 				 * prefererred address should be put before
1698 				 * deprecated addresses.
1699 				 */
1700 
1701 				/* record the interface for later search */
1702 				if (ifp_dep == NULL)
1703 					ifp_dep = ifp;
1704 
1705 				continue;
1706 			}
1707 			else if (!(ifa6->ia6_flags & IN6_IFF_DEPRECATED) &&
1708 				 allow_deprecated != 0)
1709 				continue; /* we now collect deprecated addrs */
1710 
1711 			/* What do we have to do about ::1? */
1712 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1713 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1714 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1715 					continue;
1716 				break;
1717 			case IPV6_ADDR_SCOPE_SITELOCAL:
1718 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1719 					continue;
1720 				break;
1721 			case IPV6_ADDR_SCOPE_GLOBAL:
1722 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1723 					continue;
1724 				break;
1725 			default:
1726 				continue;
1727 			}
1728 
1729 			/*
1730 			 * check if anycast is okay.
1731 			 * XXX: just experimental. not in the spec.
1732 			 */
1733 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) &&
1734 			    !(niflags & NI_NODEADDR_FLAG_ANYCAST))
1735 				continue;
1736 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) &&
1737 			    (icmp6_nodeinfo & 4) == 0) {
1738 				continue;
1739 			}
1740 
1741 			/* now we can copy the address */
1742 			if (resid < sizeof(struct in6_addr) +
1743 			    sizeof(u_int32_t)) {
1744 				/*
1745 				 * We give up much more copy.
1746 				 * Set the truncate flag and return.
1747 				 */
1748 				nni6->ni_flags |=
1749 					NI_NODEADDR_FLAG_TRUNCATE;
1750 				return (copied);
1751 			}
1752 
1753 			/*
1754 			 * Set the TTL of the address.
1755 			 * The TTL value should be one of the following
1756 			 * according to the specification:
1757 			 *
1758 			 * 1. The remaining lifetime of a DHCP lease on the
1759 			 *    address, or
1760 			 * 2. The remaining Valid Lifetime of a prefix from
1761 			 *    which the address was derived through Stateless
1762 			 *    Autoconfiguration.
1763 			 *
1764 			 * Note that we currently do not support stateful
1765 			 * address configuration by DHCPv6, so the former
1766 			 * case can't happen.
1767 			 */
1768 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1769 				ltime = ND6_INFINITE_LIFETIME;
1770 			else {
1771 				if (ifa6->ia6_lifetime.ia6t_expire >
1772 				    time_second)
1773 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_second);
1774 				else
1775 					ltime = 0;
1776 			}
1777 
1778 			bcopy(&ltime, cp, sizeof(u_int32_t));
1779 			cp += sizeof(u_int32_t);
1780 
1781 			/* copy the address itself */
1782 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1783 			      sizeof(struct in6_addr));
1784 			/* XXX: KAME link-local hack; remove ifindex */
1785 			if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr))
1786 				((struct in6_addr *)cp)->s6_addr16[1] = 0;
1787 			cp += sizeof(struct in6_addr);
1788 
1789 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1790 			copied += (sizeof(struct in6_addr) +
1791 				   sizeof(u_int32_t));
1792 		}
1793 		if (ifp0)	/* we need search only on the specified IF */
1794 			break;
1795 	}
1796 
1797 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1798 		ifp = ifp_dep;
1799 		allow_deprecated = 1;
1800 
1801 		goto again;
1802 	}
1803 
1804 	return (copied);
1805 }
1806 
1807 /*
1808  * XXX almost dup'ed code with rip6_input.
1809  */
1810 static int
1811 icmp6_rip6_input(struct	mbuf **mp, int	off)
1812 {
1813 	struct mbuf *m = *mp;
1814 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1815 	struct in6pcb *in6p;
1816 	struct in6pcb *last = NULL;
1817 	struct sockaddr_in6 rip6src;
1818 	struct icmp6_hdr *icmp6;
1819 	struct mbuf *opts = NULL;
1820 
1821 #ifndef PULLDOWN_TEST
1822 	/* this is assumed to be safe. */
1823 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1824 #else
1825 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1826 	if (icmp6 == NULL) {
1827 		/* m is already reclaimed */
1828 		return IPPROTO_DONE;
1829 	}
1830 #endif
1831 
1832 	bzero(&rip6src, sizeof(rip6src));
1833 	rip6src.sin6_len = sizeof(struct sockaddr_in6);
1834 	rip6src.sin6_family = AF_INET6;
1835 	/* KAME hack: recover scopeid */
1836 	in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif);
1837 
1838 	LIST_FOREACH(in6p, &ripcbinfo.pcblisthead, inp_list)
1839 	{
1840 		if (in6p->inp_flags & INP_PLACEMARKER)
1841 			continue;
1842 		if ((in6p->inp_vflag & INP_IPV6) == 0)
1843 			continue;
1844 #ifdef HAVE_NRL_INPCB
1845 		if (!(in6p->in6p_flags & INP_IPV6))
1846 			continue;
1847 #endif
1848 		if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1849 			continue;
1850 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1851 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1852 			continue;
1853 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1854 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1855 			continue;
1856 		if (in6p->in6p_icmp6filt
1857 		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1858 				 in6p->in6p_icmp6filt))
1859 			continue;
1860 		if (last) {
1861 			struct socket *so;
1862 			struct mbuf *n;
1863 
1864 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1865 				if (last->in6p_flags & IN6P_CONTROLOPTS)
1866 					ip6_savecontrol(last, &opts, ip6, n);
1867 				/* strip intermediate headers */
1868 				m_adj(n, off);
1869 				so = last->in6p_socket;
1870 				lwkt_gettoken(&so->so_rcv.ssb_token);
1871 				if (ssb_appendaddr(&so->so_rcv,
1872 						 (struct sockaddr *)&rip6src,
1873 						 n, opts) == 0) {
1874 					/* should notify about lost packet */
1875 					m_freem(n);
1876 					if (opts)
1877 						m_freem(opts);
1878 				} else {
1879 					sorwakeup(so);
1880 				}
1881 				lwkt_reltoken(&so->so_rcv.ssb_token);
1882 				opts = NULL;
1883 			}
1884 		}
1885 		last = in6p;
1886 	}
1887 	if (last) {
1888 		struct socket *so;
1889 
1890 		if (last->in6p_flags & IN6P_CONTROLOPTS)
1891 			ip6_savecontrol(last, &opts, ip6, m);
1892 		/* strip intermediate headers */
1893 		m_adj(m, off);
1894 		so = last->in6p_socket;
1895 		lwkt_gettoken(&so->so_rcv.ssb_token);
1896 		if (ssb_appendaddr(&so->so_rcv, (struct sockaddr *)&rip6src,
1897 				   m, opts) == 0) {
1898 			m_freem(m);
1899 			if (opts)
1900 				m_freem(opts);
1901 		} else {
1902 			sorwakeup(so);
1903 		}
1904 		lwkt_reltoken(&so->so_rcv.ssb_token);
1905 	} else {
1906 		m_freem(m);
1907 		ip6stat.ip6s_delivered--;
1908 	}
1909 	return IPPROTO_DONE;
1910 }
1911 
1912 /*
1913  * Reflect the ip6 packet back to the source.
1914  * OFF points to the icmp6 header, counted from the top of the mbuf.
1915  */
1916 void
1917 icmp6_reflect(struct mbuf *m, size_t off)
1918 {
1919 	struct ip6_hdr *ip6;
1920 	struct icmp6_hdr *icmp6;
1921 	struct in6_ifaddr *ia;
1922 	struct in6_addr t, *src = NULL;
1923 	int plen;
1924 	int type, code;
1925 	struct ifnet *outif = NULL;
1926 	struct sockaddr_in6 sa6_src, sa6_dst;
1927 #ifdef COMPAT_RFC1885
1928 	int mtu = IPV6_MMTU;
1929 	struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
1930 #endif
1931 
1932 	/* too short to reflect */
1933 	if (off < sizeof(struct ip6_hdr)) {
1934 		nd6log((LOG_DEBUG,
1935 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
1936 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
1937 		    __FILE__, __LINE__));
1938 		goto bad;
1939 	}
1940 
1941 	/*
1942 	 * If there are extra headers between IPv6 and ICMPv6, strip
1943 	 * off that header first.
1944 	 */
1945 #ifdef DIAGNOSTIC
1946 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
1947 		panic("assumption failed in icmp6_reflect");
1948 #endif
1949 	if (off > sizeof(struct ip6_hdr)) {
1950 		size_t l;
1951 		struct ip6_hdr nip6;
1952 
1953 		l = off - sizeof(struct ip6_hdr);
1954 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
1955 		m_adj(m, l);
1956 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1957 		if (m->m_len < l) {
1958 			if ((m = m_pullup(m, l)) == NULL)
1959 				return;
1960 		}
1961 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
1962 	} else /* off == sizeof(struct ip6_hdr) */ {
1963 		size_t l;
1964 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1965 		if (m->m_len < l) {
1966 			if ((m = m_pullup(m, l)) == NULL)
1967 				return;
1968 		}
1969 	}
1970 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
1971 	ip6 = mtod(m, struct ip6_hdr *);
1972 	ip6->ip6_nxt = IPPROTO_ICMPV6;
1973 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
1974 	type = icmp6->icmp6_type; /* keep type for statistics */
1975 	code = icmp6->icmp6_code; /* ditto. */
1976 
1977 	t = ip6->ip6_dst;
1978 	/*
1979 	 * ip6_input() drops a packet if its src is multicast.
1980 	 * So, the src is never multicast.
1981 	 */
1982 	ip6->ip6_dst = ip6->ip6_src;
1983 
1984 	/*
1985 	 * XXX: make sure to embed scope zone information, using
1986 	 * already embedded IDs or the received interface (if any).
1987 	 * Note that rcvif may be NULL.
1988 	 * TODO: scoped routing case (XXX).
1989 	 */
1990 	bzero(&sa6_src, sizeof(sa6_src));
1991 	sa6_src.sin6_family = AF_INET6;
1992 	sa6_src.sin6_len = sizeof(sa6_src);
1993 	sa6_src.sin6_addr = ip6->ip6_dst;
1994 	in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
1995 	in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL);
1996 	bzero(&sa6_dst, sizeof(sa6_dst));
1997 	sa6_dst.sin6_family = AF_INET6;
1998 	sa6_dst.sin6_len = sizeof(sa6_dst);
1999 	sa6_dst.sin6_addr = t;
2000 	in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
2001 	in6_embedscope(&t, &sa6_dst, NULL, NULL);
2002 
2003 #ifdef COMPAT_RFC1885
2004 	/*
2005 	 * xxx guess MTU
2006 	 * RFC 1885 requires that echo reply should be truncated if it
2007 	 * does not fit in with (return) path MTU, but the description was
2008 	 * removed in the new spec.
2009 	 */
2010 	if (icmp6_reflect_rt.ro_rt == 0 ||
2011 	    ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
2012 		if (icmp6_reflect_rt.ro_rt) {
2013 			RTFREE(icmp6_reflect_rt.ro_rt);
2014 			icmp6_reflect_rt.ro_rt = 0;
2015 		}
2016 		bzero(sin6, sizeof(*sin6));
2017 		sin6->sin6_family = PF_INET6;
2018 		sin6->sin6_len = sizeof(struct sockaddr_in6);
2019 		sin6->sin6_addr = ip6->ip6_dst;
2020 
2021 		rtalloc_ign((struct route *)&icmp6_reflect_rt.ro_rt,
2022 			    RTF_PRCLONING);
2023 	}
2024 
2025 	if (icmp6_reflect_rt.ro_rt == 0)
2026 		goto bad;
2027 
2028 	if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
2029 	    && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
2030 		mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
2031 
2032 	if (mtu < m->m_pkthdr.len) {
2033 		plen -= (m->m_pkthdr.len - mtu);
2034 		m_adj(m, mtu - m->m_pkthdr.len);
2035 	}
2036 #endif
2037 	/*
2038 	 * If the incoming packet was addressed directly to us(i.e. unicast),
2039 	 * use dst as the src for the reply.
2040 	 * The IN6_IFF_NOTREADY case would be VERY rare, but is possible
2041 	 * (for example) when we encounter an error while forwarding procedure
2042 	 * destined to a duplicated address of ours.
2043 	 */
2044 	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2045 		if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2046 		    (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2047 			src = &t;
2048 			break;
2049 		}
2050 	if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2051 		/*
2052 		 * This is the case if the dst is our link-local address
2053 		 * and the sender is also ourselves.
2054 		 */
2055 		src = &t;
2056 	}
2057 
2058 	if (src == NULL) {
2059 		int e;
2060 		struct route_in6 ro;
2061 
2062 		/*
2063 		 * This case matches to multicasts, our anycast, or unicasts
2064 		 * that we do not own.  Select a source address based on the
2065 		 * source address of the erroneous packet.
2066 		 */
2067 		bzero(&ro, sizeof(ro));
2068 		src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e, NULL);
2069 		if (ro.ro_rt)
2070 			RTFREE(ro.ro_rt); /* XXX: we could use this */
2071 		if (src == NULL) {
2072 			nd6log((LOG_DEBUG,
2073 			    "icmp6_reflect: source can't be determined: "
2074 			    "dst=%s, error=%d\n",
2075 			    ip6_sprintf(&sa6_src.sin6_addr), e));
2076 			goto bad;
2077 		}
2078 	}
2079 
2080 	ip6->ip6_src = *src;
2081 
2082 	ip6->ip6_flow = 0;
2083 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2084 	ip6->ip6_vfc |= IPV6_VERSION;
2085 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2086 	if (m->m_pkthdr.rcvif) {
2087 		/* XXX: This may not be the outgoing interface */
2088 		ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2089 	} else
2090 		ip6->ip6_hlim = ip6_defhlim;
2091 
2092 	icmp6->icmp6_cksum = 0;
2093 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2094 					sizeof(struct ip6_hdr), plen);
2095 
2096 	/*
2097 	 * XXX option handling
2098 	 */
2099 
2100 	m->m_flags &= ~(M_BCAST|M_MCAST);
2101 
2102 #ifdef COMPAT_RFC1885
2103 	ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif, NULL);
2104 #else
2105 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2106 #endif
2107 	if (outif)
2108 		icmp6_ifoutstat_inc(outif, type, code);
2109 
2110 	return;
2111 
2112 bad:
2113 	m_freem(m);
2114 	return;
2115 }
2116 
2117 void
2118 icmp6_fasttimo(void)
2119 {
2120 
2121 	mld6_fasttimeo();
2122 }
2123 
2124 static const char *
2125 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2126 		    struct in6_addr *tgt6)
2127 {
2128 	static char buf[1024];
2129 	ksnprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2130 		ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2131 	return buf;
2132 }
2133 
2134 void
2135 icmp6_redirect_input(struct mbuf *m, int off)
2136 {
2137 	struct ifnet *ifp = m->m_pkthdr.rcvif;
2138 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2139 	struct nd_redirect *nd_rd;
2140 	int icmp6len = ntohs(ip6->ip6_plen);
2141 	char *lladdr = NULL;
2142 	int lladdrlen = 0;
2143 	u_char *redirhdr = NULL;
2144 	int redirhdrlen = 0;
2145 	struct rtentry *rt = NULL;
2146 	int is_router;
2147 	int is_onlink;
2148 	struct in6_addr src6 = ip6->ip6_src;
2149 	struct in6_addr redtgt6;
2150 	struct in6_addr reddst6;
2151 	union nd_opts ndopts;
2152 
2153 	if (!m || !ifp)
2154 		return;
2155 
2156 	/* XXX if we are router, we don't update route by icmp6 redirect */
2157 	if (ip6_forwarding)
2158 		goto freeit;
2159 	if (!icmp6_rediraccept)
2160 		goto freeit;
2161 
2162 #ifndef PULLDOWN_TEST
2163 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2164 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2165 #else
2166 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2167 	if (nd_rd == NULL) {
2168 		icmp6stat.icp6s_tooshort++;
2169 		return;
2170 	}
2171 #endif
2172 	redtgt6 = nd_rd->nd_rd_target;
2173 	reddst6 = nd_rd->nd_rd_dst;
2174 
2175 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2176 		redtgt6.s6_addr16[1] = htons(ifp->if_index);
2177 	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2178 		reddst6.s6_addr16[1] = htons(ifp->if_index);
2179 
2180 	/* validation */
2181 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2182 		nd6log((LOG_ERR,
2183 			"ICMP6 redirect sent from %s rejected; "
2184 			"must be from linklocal\n", ip6_sprintf(&src6)));
2185 		goto bad;
2186 	}
2187 	if (ip6->ip6_hlim != 255) {
2188 		nd6log((LOG_ERR,
2189 			"ICMP6 redirect sent from %s rejected; "
2190 			"hlim=%d (must be 255)\n",
2191 			ip6_sprintf(&src6), ip6->ip6_hlim));
2192 		goto bad;
2193 	}
2194     {
2195 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2196 	struct sockaddr_in6 sin6;
2197 	struct in6_addr *gw6;
2198 
2199 	bzero(&sin6, sizeof(sin6));
2200 	sin6.sin6_family = AF_INET6;
2201 	sin6.sin6_len = sizeof(struct sockaddr_in6);
2202 	bcopy(&reddst6, &sin6.sin6_addr, sizeof reddst6);
2203 	rt = rtpurelookup((struct sockaddr *)&sin6);
2204 	if (rt != NULL) {
2205 		if (rt->rt_gateway == NULL ||
2206 		    rt->rt_gateway->sa_family != AF_INET6) {
2207 			nd6log((LOG_ERR,
2208 			    "ICMP6 redirect rejected; no route "
2209 			    "with inet6 gateway found for redirect dst: %s\n",
2210 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2211 			--rt->rt_refcnt;
2212 			goto bad;
2213 		}
2214 
2215 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2216 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2217 			nd6log((LOG_ERR,
2218 				"ICMP6 redirect rejected; "
2219 				"not equal to gw-for-src=%s (must be same): "
2220 				"%s\n",
2221 				ip6_sprintf(gw6),
2222 				icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2223 			--rt->rt_refcnt;
2224 			goto bad;
2225 		}
2226 	} else {
2227 		nd6log((LOG_ERR,
2228 			"ICMP6 redirect rejected; "
2229 			"no route found for redirect dst: %s\n",
2230 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2231 		goto bad;
2232 	}
2233 	--rt->rt_refcnt;
2234 	rt = NULL;
2235     }
2236 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2237 		nd6log((LOG_ERR,
2238 			"ICMP6 redirect rejected; "
2239 			"redirect dst must be unicast: %s\n",
2240 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2241 		goto bad;
2242 	}
2243 
2244 	is_router = is_onlink = 0;
2245 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2246 		is_router = 1;	/* router case */
2247 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2248 		is_onlink = 1;	/* on-link destination case */
2249 	if (!is_router && !is_onlink) {
2250 		nd6log((LOG_ERR,
2251 			"ICMP6 redirect rejected; "
2252 			"neither router case nor onlink case: %s\n",
2253 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2254 		goto bad;
2255 	}
2256 	/* validation passed */
2257 
2258 	icmp6len -= sizeof(*nd_rd);
2259 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2260 	if (nd6_options(&ndopts) < 0) {
2261 		nd6log((LOG_INFO, "icmp6_redirect_input: "
2262 			"invalid ND option, rejected: %s\n",
2263 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2264 		/* nd6_options have incremented stats */
2265 		goto freeit;
2266 	}
2267 
2268 	if (ndopts.nd_opts_tgt_lladdr) {
2269 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2270 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2271 	}
2272 
2273 	if (ndopts.nd_opts_rh) {
2274 		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2275 		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2276 	}
2277 
2278 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2279 		nd6log((LOG_INFO,
2280 			"icmp6_redirect_input: lladdrlen mismatch for %s "
2281 			"(if %d, icmp6 packet %d): %s\n",
2282 			ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2283 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2284 		goto bad;
2285 	}
2286 
2287 	/* RFC 2461 8.3 */
2288 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2289 			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2290 
2291 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2292 		/* perform rtredirect */
2293 		struct sockaddr_in6 sdst;
2294 		struct sockaddr_in6 sgw;
2295 		struct sockaddr_in6 ssrc;
2296 
2297 		bzero(&sdst, sizeof(sdst));
2298 		bzero(&sgw, sizeof(sgw));
2299 		bzero(&ssrc, sizeof(ssrc));
2300 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2301 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2302 			sizeof(struct sockaddr_in6);
2303 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2304 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2305 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2306 		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2307 			   NULL, RTF_GATEWAY | RTF_HOST,
2308 			   (struct sockaddr *)&ssrc);
2309 	}
2310 	/* finally update cached route in each socket via kpfctlinput */
2311     {
2312 	struct sockaddr_in6 sdst;
2313 
2314 	bzero(&sdst, sizeof(sdst));
2315 	sdst.sin6_family = AF_INET6;
2316 	sdst.sin6_len = sizeof(struct sockaddr_in6);
2317 	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2318 	kpfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2319 #ifdef IPSEC
2320 	key_sa_routechange((struct sockaddr *)&sdst);
2321 #endif
2322     }
2323 
2324 freeit:
2325 	m_freem(m);
2326 	return;
2327 
2328 bad:
2329 	icmp6stat.icp6s_badredirect++;
2330 	m_freem(m);
2331 }
2332 
2333 void
2334 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2335 {
2336 	struct ifnet *ifp;	/* my outgoing interface */
2337 	struct in6_addr *ifp_ll6;
2338 	struct in6_addr *router_ll6;
2339 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2340 	struct mbuf *m = NULL;	/* newly allocated one */
2341 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2342 	struct nd_redirect *nd_rd;
2343 	size_t maxlen;
2344 	u_char *p;
2345 	struct ifnet *outif = NULL;
2346 	struct sockaddr_in6 src_sa;
2347 
2348 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2349 
2350 	/* if we are not router, we don't send icmp6 redirect */
2351 	if (!ip6_forwarding || ip6_accept_rtadv)
2352 		goto fail;
2353 
2354 	/* sanity check */
2355 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2356 		goto fail;
2357 
2358 	/*
2359 	 * Address check:
2360 	 *  the source address must identify a neighbor, and
2361 	 *  the destination address must not be a multicast address
2362 	 *  [RFC 2461, sec 8.2]
2363 	 */
2364 	sip6 = mtod(m0, struct ip6_hdr *);
2365 	bzero(&src_sa, sizeof(src_sa));
2366 	src_sa.sin6_family = AF_INET6;
2367 	src_sa.sin6_len = sizeof(src_sa);
2368 	src_sa.sin6_addr = sip6->ip6_src;
2369 	/* we don't currently use sin6_scope_id, but eventually use it */
2370 	src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src);
2371 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2372 		goto fail;
2373 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2374 		goto fail;	/* what should we do here? */
2375 
2376 	/* rate limit */
2377 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2378 		goto fail;
2379 
2380 	/*
2381 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2382 	 * we almost always ask for an mbuf cluster for simplicity.
2383 	 * (MHLEN < IPV6_MMTU is almost always true)
2384 	 */
2385 #if IPV6_MMTU >= MCLBYTES
2386 # error assumption failed about IPV6_MMTU and MCLBYTES
2387 #endif
2388 	m = m_getb(IPV6_MMTU, MB_DONTWAIT, MT_HEADER, M_PKTHDR);
2389 	if (!m)
2390 		goto fail;
2391 	m->m_len = 0;
2392 	maxlen = M_TRAILINGSPACE(m);
2393 	maxlen = min(IPV6_MMTU, maxlen);
2394 	/* just for safety */
2395 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2396 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2397 		goto fail;
2398 	}
2399 
2400 	{
2401 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2402 		struct in6_ifaddr *ia;
2403 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2404 						 IN6_IFF_NOTREADY|
2405 						 IN6_IFF_ANYCAST)) == NULL)
2406 			goto fail;
2407 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2408 	}
2409 
2410 	/* get ip6 linklocal address for the router. */
2411 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2412 		struct sockaddr_in6 *sin6;
2413 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2414 		router_ll6 = &sin6->sin6_addr;
2415 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2416 			router_ll6 = NULL;
2417 	} else
2418 		router_ll6 = NULL;
2419 
2420 	/* ip6 */
2421 	ip6 = mtod(m, struct ip6_hdr *);
2422 	ip6->ip6_flow = 0;
2423 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2424 	ip6->ip6_vfc |= IPV6_VERSION;
2425 	/* ip6->ip6_plen will be set later */
2426 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2427 	ip6->ip6_hlim = 255;
2428 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2429 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2430 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2431 
2432 	/* ND Redirect */
2433 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2434 	nd_rd->nd_rd_type = ND_REDIRECT;
2435 	nd_rd->nd_rd_code = 0;
2436 	nd_rd->nd_rd_reserved = 0;
2437 	if (rt->rt_flags & RTF_GATEWAY) {
2438 		/*
2439 		 * nd_rd->nd_rd_target must be a link-local address in
2440 		 * better router cases.
2441 		 */
2442 		if (!router_ll6)
2443 			goto fail;
2444 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2445 		      sizeof(nd_rd->nd_rd_target));
2446 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2447 		      sizeof(nd_rd->nd_rd_dst));
2448 	} else {
2449 		/* make sure redtgt == reddst */
2450 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2451 		      sizeof(nd_rd->nd_rd_target));
2452 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2453 		      sizeof(nd_rd->nd_rd_dst));
2454 	}
2455 
2456 	p = (u_char *)(nd_rd + 1);
2457 
2458 	if (!router_ll6)
2459 		goto nolladdropt;
2460 
2461     {
2462 	/* target lladdr option */
2463 	struct rtentry *rt_router = NULL;
2464 	int len;
2465 	struct sockaddr_dl *sdl;
2466 	struct nd_opt_hdr *nd_opt;
2467 	char *lladdr;
2468 
2469 	rt_router = nd6_lookup(router_ll6, 0, ifp);
2470 	if (!rt_router)
2471 		goto nolladdropt;
2472 	len = sizeof(*nd_opt) + ifp->if_addrlen;
2473 	len = (len + 7) & ~7;	/* round by 8 */
2474 	/* safety check */
2475 	if (len + (p - (u_char *)ip6) > maxlen)
2476 		goto nolladdropt;
2477 	if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2478 	    (rt_router->rt_flags & RTF_LLINFO) &&
2479 	    (rt_router->rt_gateway->sa_family == AF_LINK) &&
2480 	    (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2481 	    sdl->sdl_alen) {
2482 		nd_opt = (struct nd_opt_hdr *)p;
2483 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2484 		nd_opt->nd_opt_len = len >> 3;
2485 		lladdr = (char *)(nd_opt + 1);
2486 		bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2487 		p += len;
2488 	}
2489     }
2490 nolladdropt:;
2491 
2492 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2493 
2494 	/* just to be safe */
2495 #ifdef M_DECRYPTED	/*not openbsd*/
2496 	if (m0->m_flags & M_DECRYPTED)
2497 		goto noredhdropt;
2498 #endif
2499 	if (p - (u_char *)ip6 > maxlen)
2500 		goto noredhdropt;
2501 
2502     {
2503 	/* redirected header option */
2504 	int len;
2505 	struct nd_opt_rd_hdr *nd_opt_rh;
2506 
2507 	/*
2508 	 * compute the maximum size for icmp6 redirect header option.
2509 	 * XXX room for auth header?
2510 	 */
2511 	len = maxlen - (p - (u_char *)ip6);
2512 	len &= ~7;
2513 
2514 	/* This is just for simplicity. */
2515 	if (m0->m_pkthdr.len != m0->m_len) {
2516 		if (m0->m_next) {
2517 			m_freem(m0->m_next);
2518 			m0->m_next = NULL;
2519 		}
2520 		m0->m_pkthdr.len = m0->m_len;
2521 	}
2522 
2523 	/*
2524 	 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2525 	 * about padding/truncate rule for the original IP packet.
2526 	 * From the discussion on IPv6imp in Feb 1999, the consensus was:
2527 	 * - "attach as much as possible" is the goal
2528 	 * - pad if not aligned (original size can be guessed by original
2529 	 *   ip6 header)
2530 	 * Following code adds the padding if it is simple enough,
2531 	 * and truncates if not.
2532 	 */
2533 	if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2534 		panic("assumption failed in %s:%d", __FILE__, __LINE__);
2535 
2536 	if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2537 		/* not enough room, truncate */
2538 		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2539 	} else {
2540 		/* enough room, pad or truncate */
2541 		size_t extra;
2542 
2543 		extra = m0->m_pkthdr.len % 8;
2544 		if (extra) {
2545 			/* pad if easy enough, truncate if not */
2546 			if (8 - extra <= M_TRAILINGSPACE(m0)) {
2547 				/* pad */
2548 				m0->m_len += (8 - extra);
2549 				m0->m_pkthdr.len += (8 - extra);
2550 			} else {
2551 				/* truncate */
2552 				m0->m_pkthdr.len -= extra;
2553 				m0->m_len -= extra;
2554 			}
2555 		}
2556 		len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2557 		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2558 	}
2559 
2560 	nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2561 	bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2562 	nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2563 	nd_opt_rh->nd_opt_rh_len = len >> 3;
2564 	p += sizeof(*nd_opt_rh);
2565 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2566 
2567 	/* connect m0 to m */
2568 	m->m_next = m0;
2569 	m->m_pkthdr.len = m->m_len + m0->m_len;
2570     }
2571 noredhdropt:;
2572 
2573 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
2574 		sip6->ip6_src.s6_addr16[1] = 0;
2575 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
2576 		sip6->ip6_dst.s6_addr16[1] = 0;
2577 #if 0
2578 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
2579 		ip6->ip6_src.s6_addr16[1] = 0;
2580 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
2581 		ip6->ip6_dst.s6_addr16[1] = 0;
2582 #endif
2583 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
2584 		nd_rd->nd_rd_target.s6_addr16[1] = 0;
2585 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
2586 		nd_rd->nd_rd_dst.s6_addr16[1] = 0;
2587 
2588 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2589 
2590 	nd_rd->nd_rd_cksum = 0;
2591 	nd_rd->nd_rd_cksum
2592 		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2593 
2594 	/* send the packet to outside... */
2595 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2596 	if (outif) {
2597 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2598 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2599 	}
2600 	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2601 
2602 	return;
2603 
2604 fail:
2605 	if (m)
2606 		m_freem(m);
2607 	if (m0)
2608 		m_freem(m0);
2609 }
2610 
2611 #ifdef HAVE_NRL_INPCB
2612 #define in6pcb		inpcb
2613 #define in6p_icmp6filt	inp_icmp6filt
2614 #endif
2615 /*
2616  * ICMPv6 socket option processing.
2617  */
2618 void
2619 icmp6_ctloutput(netmsg_t msg)
2620 {
2621 	struct socket *so = msg->ctloutput.base.nm_so;
2622 	struct sockopt *sopt = msg->ctloutput.nm_sopt;
2623 	int error = 0;
2624 	int optlen;
2625 	struct inpcb *inp = so->so_pcb;
2626 	int level, op, optname;
2627 
2628 	if (sopt) {
2629 		level = sopt->sopt_level;
2630 		op = sopt->sopt_dir;
2631 		optname = sopt->sopt_name;
2632 		optlen = sopt->sopt_valsize;
2633 	} else
2634 		level = op = optname = optlen = 0;
2635 
2636 	if (level != IPPROTO_ICMPV6) {
2637 		error = EINVAL;
2638 		goto out;
2639 	}
2640 
2641 	switch (op) {
2642 	case PRCO_SETOPT:
2643 		switch (optname) {
2644 		case ICMP6_FILTER:
2645 		    {
2646 			if (optlen != sizeof(struct icmp6_filter)) {
2647 				error = EMSGSIZE;
2648 				break;
2649 			}
2650 			if (inp->in6p_icmp6filt == NULL) {
2651 				error = EINVAL;
2652 				break;
2653 			}
2654 			error = soopt_to_kbuf(sopt, inp->in6p_icmp6filt, optlen,
2655 				optlen);
2656 			break;
2657 		    }
2658 
2659 		default:
2660 			error = ENOPROTOOPT;
2661 			break;
2662 		}
2663 		break;
2664 
2665 	case PRCO_GETOPT:
2666 		switch (optname) {
2667 		case ICMP6_FILTER:
2668 		    {
2669 			if (inp->in6p_icmp6filt == NULL) {
2670 				error = EINVAL;
2671 				break;
2672 			}
2673 			soopt_from_kbuf(sopt, inp->in6p_icmp6filt,
2674 				sizeof(struct icmp6_filter));
2675 			break;
2676 		    }
2677 
2678 		default:
2679 			error = ENOPROTOOPT;
2680 			break;
2681 		}
2682 		break;
2683 	}
2684 out:
2685 	lwkt_replymsg(&msg->ctloutput.base.lmsg, error);
2686 }
2687 #ifdef HAVE_NRL_INPCB
2688 #undef in6pcb
2689 #undef in6p_icmp6filt
2690 #endif
2691 
2692 #ifndef HAVE_PPSRATECHECK
2693 #ifndef timersub
2694 #define	timersub(tvp, uvp, vvp)						\
2695 	do {								\
2696 		(vvp)->tv_sec = (tvp)->tv_sec - (uvp)->tv_sec;		\
2697 		(vvp)->tv_usec = (tvp)->tv_usec - (uvp)->tv_usec;	\
2698 		if ((vvp)->tv_usec < 0) {				\
2699 			(vvp)->tv_sec--;				\
2700 			(vvp)->tv_usec += 1000000;			\
2701 		}							\
2702 	} while (0)
2703 #endif
2704 
2705 /*
2706  * ppsratecheck(): packets (or events) per second limitation.
2707  */
2708 static int
2709 ppsratecheck(struct timeval *lasttime, int *curpps,
2710 	     int maxpps)	/* maximum pps allowed */
2711 {
2712 	struct timeval tv, delta;
2713 	int rv;
2714 
2715 	microtime(&tv);
2716 
2717 	timersub(&tv, lasttime, &delta);
2718 
2719 	/*
2720 	 * Check for 0,0 so that the message will be seen at least once.
2721 	 * If more than one second has passed since the last update of
2722 	 * lasttime, reset the counter.
2723 	 *
2724 	 * We do increment *curpps even in *curpps < maxpps case, as some may
2725 	 * try to use *curpps for stat purposes as well.
2726 	 */
2727 	if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) ||
2728 	    delta.tv_sec >= 1) {
2729 		*lasttime = tv;
2730 		*curpps = 0;
2731 		rv = 1;
2732 	} else if (maxpps < 0)
2733 		rv = 1;
2734 	else if (*curpps < maxpps)
2735 		rv = 1;
2736 	else
2737 		rv = 0;
2738 
2739 #if 1 /* DIAGNOSTIC? */
2740 	/* be careful about wrap-around */
2741 	if (*curpps + 1 > *curpps)
2742 		*curpps = *curpps + 1;
2743 #else
2744 	/*
2745 	 * assume that there's not too many calls to this function.
2746 	 * not sure if the assumption holds, as it depends on *caller's*
2747 	 * behavior, not the behavior of this function.
2748 	 * IMHO it is wrong to make assumption on the caller's behavior,
2749 	 * so the above #if is #if 1, not #ifdef DIAGNOSTIC.
2750 	 */
2751 	*curpps = *curpps + 1;
2752 #endif
2753 
2754 	return (rv);
2755 }
2756 #endif
2757 
2758 /*
2759  * Perform rate limit check.
2760  * Returns 0 if it is okay to send the icmp6 packet.
2761  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2762  * limitation.
2763  *
2764  * XXX per-destination/type check necessary?
2765  */
2766 static int
2767 icmp6_ratelimit(const struct in6_addr *dst,	/* not used at this moment */
2768 		const int type,			/* not used at this moment */
2769 		const int code)			/* not used at this moment */
2770 {
2771 	int ret;
2772 
2773 	ret = 0;	/* okay to send */
2774 
2775 	/* PPS limit */
2776 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2777 	    icmp6errppslim)) {
2778 		/* The packet is subject to rate limit */
2779 		ret++;
2780 	}
2781 
2782 	return ret;
2783 }
2784