xref: /freebsd/sys/netinet6/icmp6.c (revision c07d6445eb89d9dd3950361b065b7bd110e3a043)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	$KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $
32  */
33 
34 /*-
35  * Copyright (c) 1982, 1986, 1988, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. Neither the name of the University nor the names of its contributors
47  *    may be used to endorse or promote products derived from this software
48  *    without specific prior written permission.
49  *
50  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
51  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
52  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
53  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
54  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
55  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
56  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
57  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
58  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
59  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
60  * SUCH DAMAGE.
61  *
62  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
63  */
64 
65 #include <sys/cdefs.h>
66 __FBSDID("$FreeBSD$");
67 
68 #define	MBUF_PRIVATE	/* XXXRW: Optimisation tries to avoid M_EXT mbufs */
69 
70 #include "opt_inet.h"
71 #include "opt_inet6.h"
72 
73 #include <sys/param.h>
74 #include <sys/domain.h>
75 #include <sys/jail.h>
76 #include <sys/kernel.h>
77 #include <sys/lock.h>
78 #include <sys/malloc.h>
79 #include <sys/mbuf.h>
80 #include <sys/proc.h>
81 #include <sys/protosw.h>
82 #include <sys/signalvar.h>
83 #include <sys/socket.h>
84 #include <sys/socketvar.h>
85 #include <sys/sx.h>
86 #include <sys/syslog.h>
87 #include <sys/systm.h>
88 #include <sys/time.h>
89 
90 #include <net/if.h>
91 #include <net/if_var.h>
92 #include <net/if_dl.h>
93 #include <net/if_llatbl.h>
94 #include <net/if_types.h>
95 #include <net/route.h>
96 #include <net/route/route_ctl.h>
97 #include <net/route/nhop.h>
98 #include <net/vnet.h>
99 
100 #include <netinet/in.h>
101 #include <netinet/in_pcb.h>
102 #include <netinet/in_var.h>
103 #include <netinet/ip6.h>
104 #include <netinet/icmp6.h>
105 #include <netinet/tcp_var.h>
106 
107 #include <netinet6/in6_fib.h>
108 #include <netinet6/in6_ifattach.h>
109 #include <netinet6/in6_pcb.h>
110 #include <netinet6/ip6_var.h>
111 #include <netinet6/scope6_var.h>
112 #include <netinet6/mld6_var.h>
113 #include <netinet6/nd6.h>
114 #include <netinet6/send.h>
115 
116 extern ip6proto_ctlinput_t	*ip6_ctlprotox[];
117 
118 VNET_PCPUSTAT_DEFINE(struct icmp6stat, icmp6stat);
119 VNET_PCPUSTAT_SYSINIT(icmp6stat);
120 
121 #ifdef VIMAGE
122 VNET_PCPUSTAT_SYSUNINIT(icmp6stat);
123 #endif /* VIMAGE */
124 
125 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
126 VNET_DECLARE(int, icmp6errppslim);
127 VNET_DEFINE_STATIC(int, icmp6errpps_count) = 0;
128 VNET_DEFINE_STATIC(struct timeval, icmp6errppslim_last);
129 VNET_DECLARE(int, icmp6_nodeinfo);
130 
131 #define	V_ripcbinfo			VNET(ripcbinfo)
132 #define	V_icmp6errppslim		VNET(icmp6errppslim)
133 #define	V_icmp6errpps_count		VNET(icmp6errpps_count)
134 #define	V_icmp6errppslim_last		VNET(icmp6errppslim_last)
135 #define	V_icmp6_nodeinfo		VNET(icmp6_nodeinfo)
136 
137 static void icmp6_errcount(int, int);
138 static int icmp6_rip6_input(struct mbuf **, int);
139 static void icmp6_reflect(struct mbuf *, size_t);
140 static const char *icmp6_redirect_diag(struct in6_addr *,
141 	struct in6_addr *, struct in6_addr *);
142 static struct mbuf *ni6_input(struct mbuf *, int, struct prison *);
143 static struct mbuf *ni6_nametodns(const char *, int, int);
144 static int ni6_dnsmatch(const char *, int, const char *, int);
145 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
146 			  struct ifnet **, struct in6_addr *);
147 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
148 				struct ifnet *, int);
149 static int icmp6_notify_error(struct mbuf **, int, int);
150 
151 /*
152  * Kernel module interface for updating icmp6stat.  The argument is an index
153  * into icmp6stat treated as an array of u_quad_t.  While this encodes the
154  * general layout of icmp6stat into the caller, it doesn't encode its
155  * location, so that future changes to add, for example, per-CPU stats
156  * support won't cause binary compatibility problems for kernel modules.
157  */
158 void
159 kmod_icmp6stat_inc(int statnum)
160 {
161 
162 	counter_u64_add(VNET(icmp6stat)[statnum], 1);
163 }
164 
165 static void
166 icmp6_errcount(int type, int code)
167 {
168 	switch (type) {
169 	case ICMP6_DST_UNREACH:
170 		switch (code) {
171 		case ICMP6_DST_UNREACH_NOROUTE:
172 			ICMP6STAT_INC(icp6s_odst_unreach_noroute);
173 			return;
174 		case ICMP6_DST_UNREACH_ADMIN:
175 			ICMP6STAT_INC(icp6s_odst_unreach_admin);
176 			return;
177 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
178 			ICMP6STAT_INC(icp6s_odst_unreach_beyondscope);
179 			return;
180 		case ICMP6_DST_UNREACH_ADDR:
181 			ICMP6STAT_INC(icp6s_odst_unreach_addr);
182 			return;
183 		case ICMP6_DST_UNREACH_NOPORT:
184 			ICMP6STAT_INC(icp6s_odst_unreach_noport);
185 			return;
186 		}
187 		break;
188 	case ICMP6_PACKET_TOO_BIG:
189 		ICMP6STAT_INC(icp6s_opacket_too_big);
190 		return;
191 	case ICMP6_TIME_EXCEEDED:
192 		switch (code) {
193 		case ICMP6_TIME_EXCEED_TRANSIT:
194 			ICMP6STAT_INC(icp6s_otime_exceed_transit);
195 			return;
196 		case ICMP6_TIME_EXCEED_REASSEMBLY:
197 			ICMP6STAT_INC(icp6s_otime_exceed_reassembly);
198 			return;
199 		}
200 		break;
201 	case ICMP6_PARAM_PROB:
202 		switch (code) {
203 		case ICMP6_PARAMPROB_HEADER:
204 			ICMP6STAT_INC(icp6s_oparamprob_header);
205 			return;
206 		case ICMP6_PARAMPROB_NEXTHEADER:
207 			ICMP6STAT_INC(icp6s_oparamprob_nextheader);
208 			return;
209 		case ICMP6_PARAMPROB_OPTION:
210 			ICMP6STAT_INC(icp6s_oparamprob_option);
211 			return;
212 		}
213 		break;
214 	case ND_REDIRECT:
215 		ICMP6STAT_INC(icp6s_oredirect);
216 		return;
217 	}
218 	ICMP6STAT_INC(icp6s_ounknown);
219 }
220 
221 /*
222  * A wrapper function for icmp6_error() necessary when the erroneous packet
223  * may not contain enough scope zone information.
224  */
225 void
226 icmp6_error2(struct mbuf *m, int type, int code, int param,
227     struct ifnet *ifp)
228 {
229 	struct ip6_hdr *ip6;
230 
231 	if (ifp == NULL)
232 		return;
233 
234 	if (m->m_len < sizeof(struct ip6_hdr)) {
235 		m = m_pullup(m, sizeof(struct ip6_hdr));
236 		if (m == NULL) {
237 			IP6STAT_INC(ip6s_exthdrtoolong);
238 			return;
239 		}
240 	}
241 	ip6 = mtod(m, struct ip6_hdr *);
242 
243 	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
244 		return;
245 	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
246 		return;
247 
248 	icmp6_error(m, type, code, param);
249 }
250 
251 /*
252  * Generate an error packet of type error in response to bad IP6 packet.
253  */
254 void
255 icmp6_error(struct mbuf *m, int type, int code, int param)
256 {
257 	struct ip6_hdr *oip6, *nip6;
258 	struct icmp6_hdr *icmp6;
259 	struct epoch_tracker et;
260 	u_int preplen;
261 	int off;
262 	int nxt;
263 
264 	ICMP6STAT_INC(icp6s_error);
265 
266 	/* count per-type-code statistics */
267 	icmp6_errcount(type, code);
268 
269 #ifdef M_DECRYPTED	/*not openbsd*/
270 	if (m->m_flags & M_DECRYPTED) {
271 		ICMP6STAT_INC(icp6s_canterror);
272 		goto freeit;
273 	}
274 #endif
275 
276 	if (m->m_len < sizeof(struct ip6_hdr)) {
277 		m = m_pullup(m, sizeof(struct ip6_hdr));
278 		if (m == NULL) {
279 			IP6STAT_INC(ip6s_exthdrtoolong);
280 			return;
281 		}
282 	}
283 	oip6 = mtod(m, struct ip6_hdr *);
284 
285 	/*
286 	 * If the destination address of the erroneous packet is a multicast
287 	 * address, or the packet was sent using link-layer multicast,
288 	 * we should basically suppress sending an error (RFC 2463, Section
289 	 * 2.4).
290 	 * We have two exceptions (the item e.2 in that section):
291 	 * - the Packet Too Big message can be sent for path MTU discovery.
292 	 * - the Parameter Problem Message that can be allowed an icmp6 error
293 	 *   in the option type field.  This check has been done in
294 	 *   ip6_unknown_opt(), so we can just check the type and code.
295 	 */
296 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
297 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
298 	    (type != ICMP6_PACKET_TOO_BIG &&
299 	     (type != ICMP6_PARAM_PROB ||
300 	      code != ICMP6_PARAMPROB_OPTION)))
301 		goto freeit;
302 
303 	/*
304 	 * RFC 2463, 2.4 (e.5): source address check.
305 	 * XXX: the case of anycast source?
306 	 */
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 		if (m->m_len < off + sizeof(struct icmp6_hdr)) {
321 			m = m_pullup(m, off + sizeof(struct icmp6_hdr));
322 			if (m == NULL) {
323 				IP6STAT_INC(ip6s_exthdrtoolong);
324 				return;
325 			}
326 		}
327 		oip6 = mtod(m, struct ip6_hdr *);
328 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
329 
330 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
331 		    icp->icmp6_type == ND_REDIRECT) {
332 			/*
333 			 * ICMPv6 error
334 			 * Special case: for redirect (which is
335 			 * informational) we must not send icmp6 error.
336 			 */
337 			ICMP6STAT_INC(icp6s_canterror);
338 			goto freeit;
339 		} else {
340 			/* ICMPv6 informational - send the error */
341 		}
342 	} else {
343 		/* non-ICMPv6 - send the error */
344 	}
345 
346 	/* Finally, do rate limitation check. */
347 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
348 		ICMP6STAT_INC(icp6s_toofreq);
349 		goto freeit;
350 	}
351 
352 	/*
353 	 * OK, ICMP6 can be generated.
354 	 */
355 
356 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
357 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
358 
359 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
360 	M_PREPEND(m, preplen, M_NOWAIT);	/* FIB is also copied over. */
361 	if (m == NULL) {
362 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
363 		return;
364 	}
365 
366 	nip6 = mtod(m, struct ip6_hdr *);
367 	nip6->ip6_src  = oip6->ip6_src;
368 	nip6->ip6_dst  = oip6->ip6_dst;
369 
370 	in6_clearscope(&oip6->ip6_src);
371 	in6_clearscope(&oip6->ip6_dst);
372 
373 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
374 	icmp6->icmp6_type = type;
375 	icmp6->icmp6_code = code;
376 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
377 
378 	ICMP6STAT_INC(icp6s_outhist[type]);
379 	NET_EPOCH_ENTER(et);
380 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
381 	NET_EPOCH_EXIT(et);
382 
383 	return;
384 
385   freeit:
386 	/*
387 	 * If we can't tell whether or not we can generate ICMP6, free it.
388 	 */
389 	m_freem(m);
390 }
391 
392 int
393 icmp6_errmap(const struct icmp6_hdr *icmp6)
394 {
395 
396 	switch (icmp6->icmp6_type) {
397 	case ICMP6_DST_UNREACH:
398 		switch (icmp6->icmp6_code) {
399 		case ICMP6_DST_UNREACH_NOROUTE:
400 		case ICMP6_DST_UNREACH_ADDR:
401 			return (EHOSTUNREACH);
402 		case ICMP6_DST_UNREACH_NOPORT:
403 		case ICMP6_DST_UNREACH_ADMIN:
404 			return (ECONNREFUSED);
405 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
406 			return (ENOPROTOOPT);
407 		default:
408 			return (0);	/* Shouldn't happen. */
409 		}
410 	case ICMP6_PACKET_TOO_BIG:
411 		return (EMSGSIZE);
412 	case ICMP6_TIME_EXCEEDED:
413 		switch (icmp6->icmp6_code) {
414 		case ICMP6_TIME_EXCEED_TRANSIT:
415 			return (EHOSTUNREACH);
416 		case ICMP6_TIME_EXCEED_REASSEMBLY:
417 			return (0);
418 		default:
419 			return (0);	/* Shouldn't happen. */
420 		}
421 	case ICMP6_PARAM_PROB:
422 		switch (icmp6->icmp6_code) {
423 		case ICMP6_PARAMPROB_NEXTHEADER:
424 			return (ECONNREFUSED);
425 		case ICMP6_PARAMPROB_HEADER:
426 		case ICMP6_PARAMPROB_OPTION:
427 			return (ENOPROTOOPT);
428 		default:
429 			return (0);	/* Shouldn't happen. */
430 		}
431 	default:
432 		return (0);
433 	}
434 }
435 
436 /*
437  * Process a received ICMP6 message.
438  */
439 int
440 icmp6_input(struct mbuf **mp, int *offp, int proto)
441 {
442 	struct mbuf *m, *n;
443 	struct ifnet *ifp;
444 	struct ip6_hdr *ip6, *nip6;
445 	struct icmp6_hdr *icmp6, *nicmp6;
446 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
447 	int code, error, icmp6len, ip6len, noff, off, sum;
448 
449 	NET_EPOCH_ASSERT();
450 
451 	m = *mp;
452 	off = *offp;
453 
454 	if (m->m_len < off + sizeof(struct icmp6_hdr)) {
455 		m = m_pullup(m, off + sizeof(struct icmp6_hdr));
456 		if (m == NULL) {
457 			IP6STAT_INC(ip6s_exthdrtoolong);
458 			*mp = m;
459 			return (IPPROTO_DONE);
460 		}
461 	}
462 
463 	/*
464 	 * Locate icmp6 structure in mbuf, and check
465 	 * that not corrupted and of at least minimum length
466 	 */
467 
468 	icmp6len = m->m_pkthdr.len - off;
469 	if (icmp6len < sizeof(struct icmp6_hdr)) {
470 		ICMP6STAT_INC(icp6s_tooshort);
471 		goto freeit;
472 	}
473 
474 	ip6 = mtod(m, struct ip6_hdr *);
475 	ifp = m->m_pkthdr.rcvif;
476 	/*
477 	 * Check multicast group membership.
478 	 * Note: SSM filters are not applied for ICMPv6 traffic.
479 	 */
480 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
481 		struct in6_multi	*inm;
482 
483 		inm = in6m_lookup(ifp, &ip6->ip6_dst);
484 		if (inm == NULL) {
485 			IP6STAT_INC(ip6s_notmember);
486 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
487 			goto freeit;
488 		}
489 	}
490 
491 	/* Calculate the checksum. */
492 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
493 	code = icmp6->icmp6_code;
494 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
495 		nd6log((LOG_ERR,
496 		    "ICMP6 checksum error(%d|%x) %s\n",
497 		    icmp6->icmp6_type, sum,
498 		    ip6_sprintf(ip6bufs, &ip6->ip6_src)));
499 		ICMP6STAT_INC(icp6s_checksum);
500 		goto freeit;
501 	}
502 
503 	ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
504 	icmp6_ifstat_inc(ifp, ifs6_in_msg);
505 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
506 		icmp6_ifstat_inc(ifp, ifs6_in_error);
507 
508 	ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
509 	switch (icmp6->icmp6_type) {
510 	case ICMP6_DST_UNREACH:
511 		icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
512 		switch (code) {
513 		case ICMP6_DST_UNREACH_ADMIN:
514 			icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
515 		case ICMP6_DST_UNREACH_NOROUTE:
516 		case ICMP6_DST_UNREACH_ADDR:
517 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
518 		case ICMP6_DST_UNREACH_NOPORT:
519 			goto deliver;
520 		default:
521 			goto badcode;
522 		}
523 	case ICMP6_PACKET_TOO_BIG:
524 		icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
525 		/*
526 		 * Validation is made in icmp6_mtudisc_update.
527 		 * Updating the path MTU will be done after examining
528 		 * intermediate extension headers.
529 		 */
530 		goto deliver;
531 	case ICMP6_TIME_EXCEEDED:
532 		icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
533 		switch (code) {
534 		case ICMP6_TIME_EXCEED_TRANSIT:
535 		case ICMP6_TIME_EXCEED_REASSEMBLY:
536 			goto deliver;
537 		default:
538 			goto badcode;
539 		}
540 	case ICMP6_PARAM_PROB:
541 		icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
542 		switch (code) {
543 		case ICMP6_PARAMPROB_NEXTHEADER:
544 		case ICMP6_PARAMPROB_HEADER:
545 		case ICMP6_PARAMPROB_OPTION:
546 			goto deliver;
547 		default:
548 			goto badcode;
549 		}
550 	case ICMP6_ECHO_REQUEST:
551 		icmp6_ifstat_inc(ifp, ifs6_in_echo);
552 		if (code != 0)
553 			goto badcode;
554 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
555 			/* Give up remote */
556 			break;
557 		}
558 		if (!M_WRITABLE(n)
559 		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
560 			struct mbuf *n0 = n;
561 			int n0len;
562 
563 			CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) <= MHLEN);
564 			n = m_gethdr(M_NOWAIT, n0->m_type);
565 			if (n == NULL) {
566 				/* Give up remote */
567 				m_freem(n0);
568 				break;
569 			}
570 
571 			m_move_pkthdr(n, n0);	/* FIB copied. */
572 			n0len = n0->m_pkthdr.len;	/* save for use below */
573 			/*
574 			 * Copy IPv6 and ICMPv6 only.
575 			 */
576 			nip6 = mtod(n, struct ip6_hdr *);
577 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
578 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
579 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
580 			noff = sizeof(struct ip6_hdr);
581 			/* new mbuf contains only ipv6+icmpv6 headers */
582 			n->m_len = noff + sizeof(struct icmp6_hdr);
583 			/*
584 			 * Adjust mbuf.  ip6_plen will be adjusted in
585 			 * ip6_output().
586 			 */
587 			m_adj(n0, off + sizeof(struct icmp6_hdr));
588 			/* recalculate complete packet size */
589 			n->m_pkthdr.len = n0len + (noff - off);
590 			n->m_next = n0;
591 		} else {
592 			if (n->m_len < off + sizeof(*nicmp6)) {
593 				n = m_pullup(n, off + sizeof(*nicmp6));
594 				if (n == NULL) {
595 					IP6STAT_INC(ip6s_exthdrtoolong);
596 					break;
597 				}
598 			}
599 			nicmp6 = (struct icmp6_hdr *)(mtod(n, caddr_t) + off);
600 			noff = off;
601 		}
602 		if (n) {
603 			nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
604 			nicmp6->icmp6_code = 0;
605 			ICMP6STAT_INC(icp6s_reflect);
606 			ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
607 			icmp6_reflect(n, noff);
608 		}
609 		break;
610 
611 	case ICMP6_ECHO_REPLY:
612 		icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
613 		if (code != 0)
614 			goto badcode;
615 		break;
616 
617 	case MLD_LISTENER_QUERY:
618 	case MLD_LISTENER_REPORT:
619 	case MLD_LISTENER_DONE:
620 	case MLDV2_LISTENER_REPORT:
621 		/*
622 		 * Drop MLD traffic which is not link-local, has a hop limit
623 		 * of greater than 1 hop, or which does not have the
624 		 * IPv6 HBH Router Alert option.
625 		 * As IPv6 HBH options are stripped in ip6_input() we must
626 		 * check an mbuf header flag.
627 		 * XXX Should we also sanity check that these messages
628 		 * were directed to a link-local multicast prefix?
629 		 */
630 		if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
631 			goto freeit;
632 		if (mld_input(&m, off, icmp6len) != 0) {
633 			*mp = NULL;
634 			return (IPPROTO_DONE);
635 		}
636 		/* m stays. */
637 		break;
638 
639 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
640 	    {
641 		enum { WRU, FQDN } mode;
642 		struct prison *pr;
643 
644 		if (!V_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 		pr = NULL;
655 		sx_slock(&allprison_lock);
656 		TAILQ_FOREACH(pr, &allprison, pr_list)
657 			if (pr->pr_vnet == ifp->if_vnet)
658 				break;
659 		sx_sunlock(&allprison_lock);
660 		if (pr == NULL)
661 			pr = curthread->td_ucred->cr_prison;
662 		if (mode == FQDN) {
663 			if (m->m_len < off + sizeof(struct icmp6_nodeinfo)) {
664 				m = m_pullup(m, off +
665 				    sizeof(struct icmp6_nodeinfo));
666 				if (m == NULL) {
667 					IP6STAT_INC(ip6s_exthdrtoolong);
668 					*mp = m;
669 					return (IPPROTO_DONE);
670 				}
671 			}
672 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
673 			if (n)
674 				n = ni6_input(n, off, pr);
675 			/* XXX meaningless if n == NULL */
676 			noff = sizeof(struct ip6_hdr);
677 		} else {
678 			u_char *p;
679 			int maxhlen, hlen;
680 
681 			/*
682 			 * XXX: this combination of flags is pointless,
683 			 * but should we keep this for compatibility?
684 			 */
685 			if ((V_icmp6_nodeinfo & (ICMP6_NODEINFO_FQDNOK |
686 			    ICMP6_NODEINFO_TMPADDROK)) !=
687 			    (ICMP6_NODEINFO_FQDNOK | ICMP6_NODEINFO_TMPADDROK))
688 				break;
689 
690 			if (code != 0)
691 				goto badcode;
692 
693 			CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) + 4 <= MHLEN);
694 			n = m_gethdr(M_NOWAIT, m->m_type);
695 			if (n == NULL) {
696 				/* Give up remote */
697 				break;
698 			}
699 			if (!m_dup_pkthdr(n, m, M_NOWAIT)) {
700 				/*
701 				 * Previous code did a blind M_COPY_PKTHDR
702 				 * and said "just for rcvif".  If true, then
703 				 * we could tolerate the dup failing (due to
704 				 * the deep copy of the tag chain).  For now
705 				 * be conservative and just fail.
706 				 */
707 				m_free(n);
708 				n = NULL;
709 				break;
710 			}
711 			/*
712 			 * Copy IPv6 and ICMPv6 only.
713 			 */
714 			nip6 = mtod(n, struct ip6_hdr *);
715 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
716 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
717 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
718 			p = (u_char *)(nicmp6 + 1);
719 			bzero(p, 4);
720 
721 			maxhlen = M_TRAILINGSPACE(n) -
722 			    (sizeof(*nip6) + sizeof(*nicmp6) + 4);
723 			mtx_lock(&pr->pr_mtx);
724 			hlen = strlen(pr->pr_hostname);
725 			if (maxhlen > hlen)
726 				maxhlen = hlen;
727 			/* meaningless TTL */
728 			bcopy(pr->pr_hostname, p + 4, maxhlen);
729 			mtx_unlock(&pr->pr_mtx);
730 			noff = sizeof(struct ip6_hdr);
731 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
732 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
733 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
734 			nicmp6->icmp6_code = 0;
735 		}
736 		if (n) {
737 			ICMP6STAT_INC(icp6s_reflect);
738 			ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
739 			icmp6_reflect(n, noff);
740 		}
741 		break;
742 	    }
743 
744 	case ICMP6_WRUREPLY:
745 		if (code != 0)
746 			goto badcode;
747 		break;
748 
749 	case ND_ROUTER_SOLICIT:
750 		icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
751 		if (code != 0)
752 			goto badcode;
753 		if (icmp6len < sizeof(struct nd_router_solicit))
754 			goto badlen;
755 		if (send_sendso_input_hook != NULL) {
756 			if (m->m_len < off + icmp6len) {
757 				m = m_pullup(m, off + icmp6len);
758 				if (m == NULL) {
759 					IP6STAT_INC(ip6s_exthdrtoolong);
760 					*mp = NULL;
761 					return (IPPROTO_DONE);
762 				}
763 			}
764 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
765 			if (error == 0) {
766 				m = NULL;
767 				goto freeit;
768 			}
769 		}
770 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
771 		nd6_rs_input(m, off, icmp6len);
772 		m = n;
773 		if (m == NULL)
774 			goto freeit;
775 		break;
776 
777 	case ND_ROUTER_ADVERT:
778 		icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
779 		if (code != 0)
780 			goto badcode;
781 		if (icmp6len < sizeof(struct nd_router_advert))
782 			goto badlen;
783 		if (send_sendso_input_hook != NULL) {
784 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
785 			if (error == 0) {
786 				m = NULL;
787 				goto freeit;
788 			}
789 		}
790 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
791 		nd6_ra_input(m, off, icmp6len);
792 		m = n;
793 		if (m == NULL)
794 			goto freeit;
795 		break;
796 
797 	case ND_NEIGHBOR_SOLICIT:
798 		icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
799 		if (code != 0)
800 			goto badcode;
801 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
802 			goto badlen;
803 		if (send_sendso_input_hook != NULL) {
804 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
805 			if (error == 0) {
806 				m = NULL;
807 				goto freeit;
808 			}
809 		}
810 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
811 		nd6_ns_input(m, off, icmp6len);
812 		m = n;
813 		if (m == NULL)
814 			goto freeit;
815 		break;
816 
817 	case ND_NEIGHBOR_ADVERT:
818 		icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
819 		if (code != 0)
820 			goto badcode;
821 		if (icmp6len < sizeof(struct nd_neighbor_advert))
822 			goto badlen;
823 		if (send_sendso_input_hook != NULL) {
824 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
825 			if (error == 0) {
826 				m = NULL;
827 				goto freeit;
828 			}
829 		}
830 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
831 		nd6_na_input(m, off, icmp6len);
832 		m = n;
833 		if (m == NULL)
834 			goto freeit;
835 		break;
836 
837 	case ND_REDIRECT:
838 		icmp6_ifstat_inc(ifp, ifs6_in_redirect);
839 		if (code != 0)
840 			goto badcode;
841 		if (icmp6len < sizeof(struct nd_redirect))
842 			goto badlen;
843 		if (send_sendso_input_hook != NULL) {
844 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
845 			if (error == 0) {
846 				m = NULL;
847 				goto freeit;
848 			}
849 		}
850 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
851 		icmp6_redirect_input(m, off);
852 		m = n;
853 		if (m == NULL)
854 			goto freeit;
855 		break;
856 
857 	case ICMP6_ROUTER_RENUMBERING:
858 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
859 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
860 			goto badcode;
861 		if (icmp6len < sizeof(struct icmp6_router_renum))
862 			goto badlen;
863 		break;
864 
865 	default:
866 		nd6log((LOG_DEBUG,
867 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
868 		    icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
869 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
870 		    ifp ? ifp->if_index : 0));
871 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
872 			/* ICMPv6 error: MUST deliver it by spec... */
873 			goto deliver;
874 		} else {
875 			/* ICMPv6 informational: MUST not deliver */
876 			break;
877 		}
878 	deliver:
879 		if (icmp6_notify_error(&m, off, icmp6len) != 0) {
880 			/* In this case, m should've been freed. */
881 			*mp = NULL;
882 			return (IPPROTO_DONE);
883 		}
884 		break;
885 
886 	badcode:
887 		ICMP6STAT_INC(icp6s_badcode);
888 		break;
889 
890 	badlen:
891 		ICMP6STAT_INC(icp6s_badlen);
892 		break;
893 	}
894 
895 	/* deliver the packet to appropriate sockets */
896 	icmp6_rip6_input(&m, *offp);
897 
898 	*mp = m;
899 	return (IPPROTO_DONE);
900 
901  freeit:
902 	m_freem(m);
903 	*mp = NULL;
904 	return (IPPROTO_DONE);
905 }
906 
907 static int
908 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len)
909 {
910 	struct mbuf *m;
911 	struct icmp6_hdr *icmp6;
912 	struct ip6_hdr *eip6;
913 	u_int32_t notifymtu;
914 	struct sockaddr_in6 icmp6src, icmp6dst;
915 
916 	m = *mp;
917 
918 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
919 		ICMP6STAT_INC(icp6s_tooshort);
920 		goto freeit;
921 	}
922 
923 	if (m->m_len < off + sizeof(*icmp6) + sizeof(struct ip6_hdr)) {
924 		m = m_pullup(m, off + sizeof(*icmp6) + sizeof(struct ip6_hdr));
925 		if (m == NULL) {
926 			IP6STAT_INC(ip6s_exthdrtoolong);
927 			*mp = m;
928 			return (-1);
929 		}
930 	}
931 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
932 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
933 	bzero(&icmp6dst, sizeof(icmp6dst));
934 
935 	/* Detect the upper level protocol */
936 	{
937 		u_int8_t nxt = eip6->ip6_nxt;
938 		int eoff = off + sizeof(struct icmp6_hdr) +
939 		    sizeof(struct ip6_hdr);
940 		struct ip6ctlparam ip6cp;
941 		int icmp6type = icmp6->icmp6_type;
942 		struct ip6_frag *fh;
943 		struct ip6_rthdr *rth;
944 		struct ip6_rthdr0 *rth0;
945 		int rthlen;
946 
947 		while (1) { /* XXX: should avoid infinite loop explicitly? */
948 			struct ip6_ext *eh;
949 
950 			switch (nxt) {
951 			case IPPROTO_HOPOPTS:
952 			case IPPROTO_DSTOPTS:
953 			case IPPROTO_AH:
954 				if (m->m_len < eoff + sizeof(struct ip6_ext)) {
955 					m = m_pullup(m, eoff +
956 					    sizeof(struct ip6_ext));
957 					if (m == NULL) {
958 						IP6STAT_INC(ip6s_exthdrtoolong);
959 						*mp = m;
960 						return (-1);
961 					}
962 				}
963 				eh = (struct ip6_ext *)
964 				    (mtod(m, caddr_t) + eoff);
965 				if (nxt == IPPROTO_AH)
966 					eoff += (eh->ip6e_len + 2) << 2;
967 				else
968 					eoff += (eh->ip6e_len + 1) << 3;
969 				nxt = eh->ip6e_nxt;
970 				break;
971 			case IPPROTO_ROUTING:
972 				/*
973 				 * When the erroneous packet contains a
974 				 * routing header, we should examine the
975 				 * header to determine the final destination.
976 				 * Otherwise, we can't properly update
977 				 * information that depends on the final
978 				 * destination (e.g. path MTU).
979 				 */
980 				if (m->m_len < eoff + sizeof(*rth)) {
981 					m = m_pullup(m, eoff + sizeof(*rth));
982 					if (m == NULL) {
983 						IP6STAT_INC(ip6s_exthdrtoolong);
984 						*mp = m;
985 						return (-1);
986 					}
987 				}
988 				rth = (struct ip6_rthdr *)
989 				    (mtod(m, caddr_t) + eoff);
990 				rthlen = (rth->ip6r_len + 1) << 3;
991 				/*
992 				 * XXX: currently there is no
993 				 * officially defined type other
994 				 * than type-0.
995 				 * Note that if the segment left field
996 				 * is 0, all intermediate hops must
997 				 * have been passed.
998 				 */
999 				if (rth->ip6r_segleft &&
1000 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
1001 					int hops;
1002 
1003 					if (m->m_len < eoff + rthlen) {
1004 						m = m_pullup(m, eoff + rthlen);
1005 						if (m == NULL) {
1006 							IP6STAT_INC(
1007 							    ip6s_exthdrtoolong);
1008 							*mp = m;
1009 							return (-1);
1010 						}
1011 					}
1012 					rth0 = (struct ip6_rthdr0 *)
1013 					    (mtod(m, caddr_t) + eoff);
1014 
1015 					/* just ignore a bogus header */
1016 					if ((rth0->ip6r0_len % 2) == 0 &&
1017 					    (hops = rth0->ip6r0_len/2))
1018 						icmp6dst.sin6_addr = *((struct in6_addr *)(rth0 + 1) + (hops - 1));
1019 				}
1020 				eoff += rthlen;
1021 				nxt = rth->ip6r_nxt;
1022 				break;
1023 			case IPPROTO_FRAGMENT:
1024 				if (m->m_len < eoff + sizeof(struct ip6_frag)) {
1025 					m = m_pullup(m, eoff +
1026 					    sizeof(struct ip6_frag));
1027 					if (m == NULL) {
1028 						IP6STAT_INC(ip6s_exthdrtoolong);
1029 						*mp = m;
1030 						return (-1);
1031 					}
1032 				}
1033 				fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1034 				    eoff);
1035 				/*
1036 				 * Data after a fragment header is meaningless
1037 				 * unless it is the first fragment, but
1038 				 * we'll go to the notify label for path MTU
1039 				 * discovery.
1040 				 */
1041 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1042 					goto notify;
1043 
1044 				eoff += sizeof(struct ip6_frag);
1045 				nxt = fh->ip6f_nxt;
1046 				break;
1047 			default:
1048 				/*
1049 				 * This case includes ESP and the No Next
1050 				 * Header.  In such cases going to the notify
1051 				 * label does not have any meaning
1052 				 * (i.e. ctlfunc will be NULL), but we go
1053 				 * anyway since we might have to update
1054 				 * path MTU information.
1055 				 */
1056 				goto notify;
1057 			}
1058 		}
1059 	  notify:
1060 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1061 
1062 		/*
1063 		 * retrieve parameters from the inner IPv6 header, and convert
1064 		 * them into sockaddr structures.
1065 		 * XXX: there is no guarantee that the source or destination
1066 		 * addresses of the inner packet are in the same scope as
1067 		 * the addresses of the icmp packet.  But there is no other
1068 		 * way to determine the zone.
1069 		 */
1070 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1071 
1072 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1073 		icmp6dst.sin6_family = AF_INET6;
1074 		if (IN6_IS_ADDR_UNSPECIFIED(&icmp6dst.sin6_addr))
1075 			icmp6dst.sin6_addr = eip6->ip6_dst;
1076 		if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1077 			goto freeit;
1078 		bzero(&icmp6src, sizeof(icmp6src));
1079 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1080 		icmp6src.sin6_family = AF_INET6;
1081 		icmp6src.sin6_addr = eip6->ip6_src;
1082 		if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1083 			goto freeit;
1084 		icmp6src.sin6_flowinfo =
1085 		    (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1086 
1087 		ip6cp.ip6c_m = m;
1088 		ip6cp.ip6c_icmp6 = icmp6;
1089 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1090 		ip6cp.ip6c_off = eoff;
1091 		ip6cp.ip6c_finaldst = &icmp6dst;
1092 		ip6cp.ip6c_src = &icmp6src;
1093 		ip6cp.ip6c_nxt = nxt;
1094 
1095 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1096 			notifymtu = ntohl(icmp6->icmp6_mtu);
1097 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1098 			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1099 		}
1100 
1101 		if (ip6_ctlprotox[nxt] != NULL)
1102 			ip6_ctlprotox[nxt](&ip6cp);
1103 	}
1104 	*mp = m;
1105 	return (0);
1106 
1107   freeit:
1108 	m_freem(m);
1109 	*mp = NULL;
1110 	return (-1);
1111 }
1112 
1113 void
1114 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1115 {
1116 	struct in6_addr *dst = &ip6cp->ip6c_finaldst->sin6_addr;
1117 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1118 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1119 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1120 	struct in_conninfo inc;
1121 	uint32_t max_mtu;
1122 
1123 #if 0
1124 	/*
1125 	 * RFC2460 section 5, last paragraph.
1126 	 * even though minimum link MTU for IPv6 is IPV6_MMTU,
1127 	 * we may see ICMPv6 too big with mtu < IPV6_MMTU
1128 	 * due to packet translator in the middle.
1129 	 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1130 	 * special handling.
1131 	 */
1132 	if (mtu < IPV6_MMTU)
1133 		return;
1134 #endif
1135 
1136 	/*
1137 	 * we reject ICMPv6 too big with abnormally small value.
1138 	 * XXX what is the good definition of "abnormally small"?
1139 	 */
1140 	if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1141 		return;
1142 
1143 	if (!validated)
1144 		return;
1145 
1146 	/*
1147 	 * In case the suggested mtu is less than IPV6_MMTU, we
1148 	 * only need to remember that it was for above mentioned
1149 	 * "alwaysfrag" case.
1150 	 * Try to be as close to the spec as possible.
1151 	 */
1152 	if (mtu < IPV6_MMTU)
1153 		mtu = IPV6_MMTU - 8;
1154 
1155 	bzero(&inc, sizeof(inc));
1156 	inc.inc_fibnum = M_GETFIB(m);
1157 	inc.inc_flags |= INC_ISIPV6;
1158 	inc.inc6_faddr = *dst;
1159 	if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1160 		return;
1161 
1162 	max_mtu = tcp_hc_getmtu(&inc);
1163 	if (max_mtu == 0)
1164 		max_mtu = tcp_maxmtu6(&inc, NULL);
1165 
1166 	if (mtu < max_mtu) {
1167 		tcp_hc_updatemtu(&inc, mtu);
1168 		ICMP6STAT_INC(icp6s_pmtuchg);
1169 	}
1170 }
1171 
1172 /*
1173  * Process a Node Information Query packet, based on
1174  * draft-ietf-ipngwg-icmp-name-lookups-07.
1175  *
1176  * Spec incompatibilities:
1177  * - IPv6 Subject address handling
1178  * - IPv4 Subject address handling support missing
1179  * - Proxy reply (answer even if it's not for me)
1180  * - joins NI group address at in6_ifattach() time only, does not cope
1181  *   with hostname changes by sethostname(3)
1182  */
1183 static struct mbuf *
1184 ni6_input(struct mbuf *m, int off, struct prison *pr)
1185 {
1186 	struct icmp6_nodeinfo *ni6, *nni6;
1187 	struct mbuf *n = NULL;
1188 	u_int16_t qtype;
1189 	int subjlen;
1190 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1191 	struct ni_reply_fqdn *fqdn;
1192 	int addrs;		/* for NI_QTYPE_NODEADDR */
1193 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1194 	struct in6_addr in6_subj; /* subject address */
1195 	struct ip6_hdr *ip6;
1196 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1197 	char *subj = NULL;
1198 	struct in6_ifaddr *ia6 = NULL;
1199 
1200 	ip6 = mtod(m, struct ip6_hdr *);
1201 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1202 
1203 	/*
1204 	 * Validate IPv6 source address.
1205 	 * The default configuration MUST be to refuse answering queries from
1206 	 * global-scope addresses according to RFC4602.
1207 	 * Notes:
1208 	 *  - it's not very clear what "refuse" means; this implementation
1209 	 *    simply drops it.
1210 	 *  - it's not very easy to identify global-scope (unicast) addresses
1211 	 *    since there are many prefixes for them.  It should be safer
1212 	 *    and in practice sufficient to check "all" but loopback and
1213 	 *    link-local (note that site-local unicast was deprecated and
1214 	 *    ULA is defined as global scope-wise)
1215 	 */
1216 	if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1217 	    !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1218 	    !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1219 		goto bad;
1220 
1221 	/*
1222 	 * Validate IPv6 destination address.
1223 	 *
1224 	 * The Responder must discard the Query without further processing
1225 	 * unless it is one of the Responder's unicast or anycast addresses, or
1226 	 * a link-local scope multicast address which the Responder has joined.
1227 	 * [RFC4602, Section 5.]
1228 	 */
1229 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1230 		if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1231 			goto bad;
1232 		/* else it's a link-local multicast, fine */
1233 	} else {		/* unicast or anycast */
1234 		ia6 = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false);
1235 		if (ia6 == NULL)
1236 			goto bad; /* XXX impossible */
1237 
1238 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1239 		    !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1240 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1241 				"a temporary address in %s:%d",
1242 			       __FILE__, __LINE__));
1243 			goto bad;
1244 		}
1245 	}
1246 
1247 	/* validate query Subject field. */
1248 	qtype = ntohs(ni6->ni_qtype);
1249 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1250 	switch (qtype) {
1251 	case NI_QTYPE_NOOP:
1252 	case NI_QTYPE_SUPTYPES:
1253 		/* 07 draft */
1254 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1255 			break;
1256 		/* FALLTHROUGH */
1257 	case NI_QTYPE_FQDN:
1258 	case NI_QTYPE_NODEADDR:
1259 	case NI_QTYPE_IPV4ADDR:
1260 		switch (ni6->ni_code) {
1261 		case ICMP6_NI_SUBJ_IPV6:
1262 #if ICMP6_NI_SUBJ_IPV6 != 0
1263 		case 0:
1264 #endif
1265 			/*
1266 			 * backward compatibility - try to accept 03 draft
1267 			 * format, where no Subject is present.
1268 			 */
1269 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1270 			    subjlen == 0) {
1271 				oldfqdn++;
1272 				break;
1273 			}
1274 #if ICMP6_NI_SUBJ_IPV6 != 0
1275 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1276 				goto bad;
1277 #endif
1278 
1279 			if (subjlen != sizeof(struct in6_addr))
1280 				goto bad;
1281 
1282 			/*
1283 			 * Validate Subject address.
1284 			 *
1285 			 * Not sure what exactly "address belongs to the node"
1286 			 * means in the spec, is it just unicast, or what?
1287 			 *
1288 			 * At this moment we consider Subject address as
1289 			 * "belong to the node" if the Subject address equals
1290 			 * to the IPv6 destination address; validation for
1291 			 * IPv6 destination address should have done enough
1292 			 * check for us.
1293 			 *
1294 			 * We do not do proxy at this moment.
1295 			 */
1296 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1297 			    subjlen, (caddr_t)&in6_subj);
1298 			if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1299 				goto bad;
1300 
1301 			subj = (char *)&in6_subj;
1302 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1303 				break;
1304 
1305 			/*
1306 			 * XXX if we are to allow other cases, we should really
1307 			 * be careful about scope here.
1308 			 * basically, we should disallow queries toward IPv6
1309 			 * destination X with subject Y,
1310 			 * if scope(X) > scope(Y).
1311 			 * if we allow scope(X) > scope(Y), it will result in
1312 			 * information leakage across scope boundary.
1313 			 */
1314 			goto bad;
1315 
1316 		case ICMP6_NI_SUBJ_FQDN:
1317 			/*
1318 			 * Validate Subject name with gethostname(3).
1319 			 *
1320 			 * The behavior may need some debate, since:
1321 			 * - we are not sure if the node has FQDN as
1322 			 *   hostname (returned by gethostname(3)).
1323 			 * - the code does wildcard match for truncated names.
1324 			 *   however, we are not sure if we want to perform
1325 			 *   wildcard match, if gethostname(3) side has
1326 			 *   truncated hostname.
1327 			 */
1328 			mtx_lock(&pr->pr_mtx);
1329 			n = ni6_nametodns(pr->pr_hostname,
1330 			    strlen(pr->pr_hostname), 0);
1331 			mtx_unlock(&pr->pr_mtx);
1332 			if (!n || n->m_next || n->m_len == 0)
1333 				goto bad;
1334 			if (m->m_len < off + sizeof(struct icmp6_nodeinfo) +
1335 			    subjlen) {
1336 				m = m_pullup(m, off +
1337 				    sizeof(struct icmp6_nodeinfo) + subjlen);
1338 				if (m == NULL) {
1339 					IP6STAT_INC(ip6s_exthdrtoolong);
1340 					goto bad;
1341 				}
1342 			}
1343 			/* ip6 possibly invalid but not used after. */
1344 			ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1345 			subj = (char *)(mtod(m, caddr_t) + off +
1346 			    sizeof(struct icmp6_nodeinfo));
1347 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1348 			    n->m_len)) {
1349 				goto bad;
1350 			}
1351 			m_freem(n);
1352 			n = NULL;
1353 			break;
1354 
1355 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1356 		default:
1357 			goto bad;
1358 		}
1359 		break;
1360 	}
1361 
1362 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1363 	switch (qtype) {
1364 	case NI_QTYPE_FQDN:
1365 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1366 			goto bad;
1367 		break;
1368 	case NI_QTYPE_NODEADDR:
1369 	case NI_QTYPE_IPV4ADDR:
1370 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1371 			goto bad;
1372 		break;
1373 	}
1374 
1375 	/* guess reply length */
1376 	switch (qtype) {
1377 	case NI_QTYPE_NOOP:
1378 		break;		/* no reply data */
1379 	case NI_QTYPE_SUPTYPES:
1380 		replylen += sizeof(u_int32_t);
1381 		break;
1382 	case NI_QTYPE_FQDN:
1383 		/* XXX will append an mbuf */
1384 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1385 		break;
1386 	case NI_QTYPE_NODEADDR:
1387 		addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1388 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1389 		    sizeof(u_int32_t))) > MCLBYTES)
1390 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1391 		break;
1392 	case NI_QTYPE_IPV4ADDR:
1393 		/* unsupported - should respond with unknown Qtype? */
1394 		break;
1395 	default:
1396 		/*
1397 		 * XXX: We must return a reply with the ICMP6 code
1398 		 * `unknown Qtype' in this case.  However we regard the case
1399 		 * as an FQDN query for backward compatibility.
1400 		 * Older versions set a random value to this field,
1401 		 * so it rarely varies in the defined qtypes.
1402 		 * But the mechanism is not reliable...
1403 		 * maybe we should obsolete older versions.
1404 		 */
1405 		qtype = NI_QTYPE_FQDN;
1406 		/* XXX will append an mbuf */
1407 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1408 		oldfqdn++;
1409 		break;
1410 	}
1411 
1412 	/* Allocate an mbuf to reply. */
1413 	if (replylen > MCLBYTES) {
1414 		/*
1415 		 * XXX: should we try to allocate more? But MCLBYTES
1416 		 * is probably much larger than IPV6_MMTU...
1417 		 */
1418 		goto bad;
1419 	}
1420 	if (replylen > MHLEN)
1421 		n = m_getcl(M_NOWAIT, m->m_type, M_PKTHDR);
1422 	else
1423 		n = m_gethdr(M_NOWAIT, m->m_type);
1424 	if (n == NULL) {
1425 		m_freem(m);
1426 		return (NULL);
1427 	}
1428 	m_move_pkthdr(n, m); /* just for recvif and FIB */
1429 	n->m_pkthdr.len = n->m_len = replylen;
1430 
1431 	/* copy mbuf header and IPv6 + Node Information base headers */
1432 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1433 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1434 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1435 
1436 	/* qtype dependent procedure */
1437 	switch (qtype) {
1438 	case NI_QTYPE_NOOP:
1439 		nni6->ni_code = ICMP6_NI_SUCCESS;
1440 		nni6->ni_flags = 0;
1441 		break;
1442 	case NI_QTYPE_SUPTYPES:
1443 	{
1444 		u_int32_t v;
1445 		nni6->ni_code = ICMP6_NI_SUCCESS;
1446 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1447 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1448 		v = (u_int32_t)htonl(0x0000000f);
1449 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1450 		break;
1451 	}
1452 	case NI_QTYPE_FQDN:
1453 		nni6->ni_code = ICMP6_NI_SUCCESS;
1454 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1455 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1456 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1457 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1458 		/*
1459 		 * XXX do we really have FQDN in hostname?
1460 		 */
1461 		mtx_lock(&pr->pr_mtx);
1462 		n->m_next = ni6_nametodns(pr->pr_hostname,
1463 		    strlen(pr->pr_hostname), oldfqdn);
1464 		mtx_unlock(&pr->pr_mtx);
1465 		if (n->m_next == NULL)
1466 			goto bad;
1467 		/* XXX we assume that n->m_next is not a chain */
1468 		if (n->m_next->m_next != NULL)
1469 			goto bad;
1470 		n->m_pkthdr.len += n->m_next->m_len;
1471 		break;
1472 	case NI_QTYPE_NODEADDR:
1473 	{
1474 		int lenlim, copied;
1475 
1476 		nni6->ni_code = ICMP6_NI_SUCCESS;
1477 		n->m_pkthdr.len = n->m_len =
1478 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1479 		lenlim = M_TRAILINGSPACE(n);
1480 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1481 		/* XXX: reset mbuf length */
1482 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1483 		    sizeof(struct icmp6_nodeinfo) + copied;
1484 		break;
1485 	}
1486 	default:
1487 		break;		/* XXX impossible! */
1488 	}
1489 
1490 	nni6->ni_type = ICMP6_NI_REPLY;
1491 	m_freem(m);
1492 	return (n);
1493 
1494   bad:
1495 	m_freem(m);
1496 	if (n)
1497 		m_freem(n);
1498 	return (NULL);
1499 }
1500 
1501 /*
1502  * make a mbuf with DNS-encoded string.  no compression support.
1503  *
1504  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1505  * treated as truncated name (two \0 at the end).  this is a wild guess.
1506  *
1507  * old - return pascal string if non-zero
1508  */
1509 static struct mbuf *
1510 ni6_nametodns(const char *name, int namelen, int old)
1511 {
1512 	struct mbuf *m;
1513 	char *cp, *ep;
1514 	const char *p, *q;
1515 	int i, len, nterm;
1516 
1517 	if (old)
1518 		len = namelen + 1;
1519 	else
1520 		len = MCLBYTES;
1521 
1522 	/* Because MAXHOSTNAMELEN is usually 256, we use cluster mbuf. */
1523 	if (len > MLEN)
1524 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1525 	else
1526 		m = m_get(M_NOWAIT, MT_DATA);
1527 	if (m == NULL)
1528 		goto fail;
1529 
1530 	if (old) {
1531 		m->m_len = len;
1532 		*mtod(m, char *) = namelen;
1533 		bcopy(name, mtod(m, char *) + 1, namelen);
1534 		return m;
1535 	} else {
1536 		m->m_len = 0;
1537 		cp = mtod(m, char *);
1538 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1539 
1540 		/* if not certain about my name, return empty buffer */
1541 		if (namelen == 0)
1542 			return m;
1543 
1544 		/*
1545 		 * guess if it looks like shortened hostname, or FQDN.
1546 		 * shortened hostname needs two trailing "\0".
1547 		 */
1548 		i = 0;
1549 		for (p = name; p < name + namelen; p++) {
1550 			if (*p && *p == '.')
1551 				i++;
1552 		}
1553 		if (i < 2)
1554 			nterm = 2;
1555 		else
1556 			nterm = 1;
1557 
1558 		p = name;
1559 		while (cp < ep && p < name + namelen) {
1560 			i = 0;
1561 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1562 				i++;
1563 			/* result does not fit into mbuf */
1564 			if (cp + i + 1 >= ep)
1565 				goto fail;
1566 			/*
1567 			 * DNS label length restriction, RFC1035 page 8.
1568 			 * "i == 0" case is included here to avoid returning
1569 			 * 0-length label on "foo..bar".
1570 			 */
1571 			if (i <= 0 || i >= 64)
1572 				goto fail;
1573 			*cp++ = i;
1574 			bcopy(p, cp, i);
1575 			cp += i;
1576 			p = q;
1577 			if (p < name + namelen && *p == '.')
1578 				p++;
1579 		}
1580 		/* termination */
1581 		if (cp + nterm >= ep)
1582 			goto fail;
1583 		while (nterm-- > 0)
1584 			*cp++ = '\0';
1585 		m->m_len = cp - mtod(m, char *);
1586 		return m;
1587 	}
1588 
1589 	panic("should not reach here");
1590 	/* NOTREACHED */
1591 
1592  fail:
1593 	if (m)
1594 		m_freem(m);
1595 	return NULL;
1596 }
1597 
1598 /*
1599  * check if two DNS-encoded string matches.  takes care of truncated
1600  * form (with \0\0 at the end).  no compression support.
1601  * XXX upper/lowercase match (see RFC2065)
1602  */
1603 static int
1604 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1605 {
1606 	const char *a0, *b0;
1607 	int l;
1608 
1609 	/* simplest case - need validation? */
1610 	if (alen == blen && bcmp(a, b, alen) == 0)
1611 		return 1;
1612 
1613 	a0 = a;
1614 	b0 = b;
1615 
1616 	/* termination is mandatory */
1617 	if (alen < 2 || blen < 2)
1618 		return 0;
1619 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1620 		return 0;
1621 	alen--;
1622 	blen--;
1623 
1624 	while (a - a0 < alen && b - b0 < blen) {
1625 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1626 			return 0;
1627 
1628 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1629 			return 0;
1630 		/* we don't support compression yet */
1631 		if (a[0] >= 64 || b[0] >= 64)
1632 			return 0;
1633 
1634 		/* truncated case */
1635 		if (a[0] == 0 && a - a0 == alen - 1)
1636 			return 1;
1637 		if (b[0] == 0 && b - b0 == blen - 1)
1638 			return 1;
1639 		if (a[0] == 0 || b[0] == 0)
1640 			return 0;
1641 
1642 		if (a[0] != b[0])
1643 			return 0;
1644 		l = a[0];
1645 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1646 			return 0;
1647 		if (bcmp(a + 1, b + 1, l) != 0)
1648 			return 0;
1649 
1650 		a += 1 + l;
1651 		b += 1 + l;
1652 	}
1653 
1654 	if (a - a0 == alen && b - b0 == blen)
1655 		return 1;
1656 	else
1657 		return 0;
1658 }
1659 
1660 /*
1661  * calculate the number of addresses to be returned in the node info reply.
1662  */
1663 static int
1664 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1665     struct in6_addr *subj)
1666 {
1667 	struct ifnet *ifp;
1668 	struct in6_ifaddr *ifa6;
1669 	struct ifaddr *ifa;
1670 	int addrs = 0, addrsofif, iffound = 0;
1671 	int niflags = ni6->ni_flags;
1672 
1673 	NET_EPOCH_ASSERT();
1674 
1675 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1676 		switch (ni6->ni_code) {
1677 		case ICMP6_NI_SUBJ_IPV6:
1678 			if (subj == NULL) /* must be impossible... */
1679 				return (0);
1680 			break;
1681 		default:
1682 			/*
1683 			 * XXX: we only support IPv6 subject address for
1684 			 * this Qtype.
1685 			 */
1686 			return (0);
1687 		}
1688 	}
1689 
1690 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1691 		addrsofif = 0;
1692 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1693 			if (ifa->ifa_addr->sa_family != AF_INET6)
1694 				continue;
1695 			ifa6 = (struct in6_ifaddr *)ifa;
1696 
1697 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1698 			    IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1699 				iffound = 1;
1700 
1701 			/*
1702 			 * IPv4-mapped addresses can only be returned by a
1703 			 * Node Information proxy, since they represent
1704 			 * addresses of IPv4-only nodes, which perforce do
1705 			 * not implement this protocol.
1706 			 * [icmp-name-lookups-07, Section 5.4]
1707 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1708 			 * this function at this moment.
1709 			 */
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) != 0 &&
1734 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1735 				continue; /* we need only unicast addresses */
1736 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1737 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1738 				continue;
1739 			}
1740 			addrsofif++; /* count the address */
1741 		}
1742 		if (iffound) {
1743 			*ifpp = ifp;
1744 			return (addrsofif);
1745 		}
1746 
1747 		addrs += addrsofif;
1748 	}
1749 
1750 	return (addrs);
1751 }
1752 
1753 static int
1754 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1755     struct ifnet *ifp0, int resid)
1756 {
1757 	struct ifnet *ifp;
1758 	struct in6_ifaddr *ifa6;
1759 	struct ifaddr *ifa;
1760 	struct ifnet *ifp_dep = NULL;
1761 	int copied = 0, allow_deprecated = 0;
1762 	u_char *cp = (u_char *)(nni6 + 1);
1763 	int niflags = ni6->ni_flags;
1764 	u_int32_t ltime;
1765 
1766 	NET_EPOCH_ASSERT();
1767 
1768 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1769 		return (0);	/* needless to copy */
1770 
1771 	ifp = ifp0 ? ifp0 : CK_STAILQ_FIRST(&V_ifnet);
1772   again:
1773 
1774 	for (; ifp; ifp = CK_STAILQ_NEXT(ifp, if_link)) {
1775 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1776 			if (ifa->ifa_addr->sa_family != AF_INET6)
1777 				continue;
1778 			ifa6 = (struct in6_ifaddr *)ifa;
1779 
1780 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1781 			    allow_deprecated == 0) {
1782 				/*
1783 				 * prefererred address should be put before
1784 				 * deprecated addresses.
1785 				 */
1786 
1787 				/* record the interface for later search */
1788 				if (ifp_dep == NULL)
1789 					ifp_dep = ifp;
1790 
1791 				continue;
1792 			} else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1793 			    allow_deprecated != 0)
1794 				continue; /* we now collect deprecated addrs */
1795 
1796 			/* What do we have to do about ::1? */
1797 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1798 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1799 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1800 					continue;
1801 				break;
1802 			case IPV6_ADDR_SCOPE_SITELOCAL:
1803 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1804 					continue;
1805 				break;
1806 			case IPV6_ADDR_SCOPE_GLOBAL:
1807 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1808 					continue;
1809 				break;
1810 			default:
1811 				continue;
1812 			}
1813 
1814 			/*
1815 			 * check if anycast is okay.
1816 			 * XXX: just experimental.  not in the spec.
1817 			 */
1818 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1819 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1820 				continue;
1821 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1822 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1823 				continue;
1824 			}
1825 
1826 			/* now we can copy the address */
1827 			if (resid < sizeof(struct in6_addr) +
1828 			    sizeof(u_int32_t)) {
1829 				/*
1830 				 * We give up much more copy.
1831 				 * Set the truncate flag and return.
1832 				 */
1833 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1834 				return (copied);
1835 			}
1836 
1837 			/*
1838 			 * Set the TTL of the address.
1839 			 * The TTL value should be one of the following
1840 			 * according to the specification:
1841 			 *
1842 			 * 1. The remaining lifetime of a DHCP lease on the
1843 			 *    address, or
1844 			 * 2. The remaining Valid Lifetime of a prefix from
1845 			 *    which the address was derived through Stateless
1846 			 *    Autoconfiguration.
1847 			 *
1848 			 * Note that we currently do not support stateful
1849 			 * address configuration by DHCPv6, so the former
1850 			 * case can't happen.
1851 			 */
1852 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1853 				ltime = ND6_INFINITE_LIFETIME;
1854 			else {
1855 				if (ifa6->ia6_lifetime.ia6t_expire >
1856 				    time_uptime)
1857 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime);
1858 				else
1859 					ltime = 0;
1860 			}
1861 
1862 			bcopy(&ltime, cp, sizeof(u_int32_t));
1863 			cp += sizeof(u_int32_t);
1864 
1865 			/* copy the address itself */
1866 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1867 			    sizeof(struct in6_addr));
1868 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1869 			cp += sizeof(struct in6_addr);
1870 
1871 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1872 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1873 		}
1874 		if (ifp0)	/* we need search only on the specified IF */
1875 			break;
1876 	}
1877 
1878 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1879 		ifp = ifp_dep;
1880 		allow_deprecated = 1;
1881 
1882 		goto again;
1883 	}
1884 
1885 	return (copied);
1886 }
1887 
1888 static bool
1889 icmp6_rip6_match(const struct inpcb *inp, void *v)
1890 {
1891 	struct ip6_hdr *ip6 = v;
1892 
1893 	if ((inp->inp_vflag & INP_IPV6) == 0)
1894 		return (false);
1895 	if (inp->inp_ip_p != IPPROTO_ICMPV6)
1896 		return (false);
1897 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1898 	   !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &ip6->ip6_dst))
1899 		return (false);
1900 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1901 	   !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &ip6->ip6_src))
1902 		return (false);
1903 	return (true);
1904 }
1905 
1906 /*
1907  * XXX almost dup'ed code with rip6_input.
1908  */
1909 static int
1910 icmp6_rip6_input(struct mbuf **mp, int off)
1911 {
1912 	struct mbuf *n, *m = *mp;
1913 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1914 	struct inpcb_iterator inpi = INP_ITERATOR(&V_ripcbinfo,
1915 	    INPLOOKUP_RLOCKPCB, icmp6_rip6_match, ip6);
1916 	struct inpcb *inp;
1917 	struct sockaddr_in6 fromsa;
1918 	struct icmp6_hdr *icmp6;
1919 	struct mbuf *opts = NULL;
1920 	int delivered = 0;
1921 
1922 	/* This is assumed to be safe; icmp6_input() does a pullup. */
1923 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1924 
1925 	/*
1926 	 * XXX: the address may have embedded scope zone ID, which should be
1927 	 * hidden from applications.
1928 	 */
1929 	bzero(&fromsa, sizeof(fromsa));
1930 	fromsa.sin6_family = AF_INET6;
1931 	fromsa.sin6_len = sizeof(struct sockaddr_in6);
1932 	fromsa.sin6_addr = ip6->ip6_src;
1933 	if (sa6_recoverscope(&fromsa)) {
1934 		m_freem(m);
1935 		*mp = NULL;
1936 		return (IPPROTO_DONE);
1937 	}
1938 
1939 	while ((inp = inp_next(&inpi)) != NULL) {
1940 		if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1941 		    inp->in6p_icmp6filt))
1942 			continue;
1943 		/*
1944 		 * Recent network drivers tend to allocate a single
1945 		 * mbuf cluster, rather than to make a couple of
1946 		 * mbufs without clusters.  Also, since the IPv6 code
1947 		 * path tries to avoid m_pullup(), it is highly
1948 		 * probable that we still have an mbuf cluster here
1949 		 * even though the necessary length can be stored in an
1950 		 * mbuf's internal buffer.
1951 		 * Meanwhile, the default size of the receive socket
1952 		 * buffer for raw sockets is not so large.  This means
1953 		 * the possibility of packet loss is relatively higher
1954 		 * than before.  To avoid this scenario, we copy the
1955 		 * received data to a separate mbuf that does not use
1956 		 * a cluster, if possible.
1957 		 * XXX: it is better to copy the data after stripping
1958 		 * intermediate headers.
1959 		 */
1960 		if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1961 		    m->m_len <= MHLEN) {
1962 			n = m_get(M_NOWAIT, m->m_type);
1963 			if (n != NULL) {
1964 				if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1965 					bcopy(m->m_data, n->m_data, m->m_len);
1966 					n->m_len = m->m_len;
1967 				} else {
1968 					m_free(n);
1969 					n = NULL;
1970 				}
1971 			}
1972 		} else
1973 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
1974 		if (n == NULL)
1975 			continue;
1976 		if (inp->inp_flags & INP_CONTROLOPTS)
1977 			ip6_savecontrol(inp, n, &opts);
1978 		/* strip intermediate headers */
1979 		m_adj(n, off);
1980 		SOCKBUF_LOCK(&inp->inp_socket->so_rcv);
1981 		if (sbappendaddr_locked(&inp->inp_socket->so_rcv,
1982 		    (struct sockaddr *)&fromsa, n, opts) == 0) {
1983 			soroverflow_locked(inp->inp_socket);
1984 			m_freem(n);
1985 			if (opts)
1986 				m_freem(opts);
1987 		} else {
1988 			sorwakeup_locked(inp->inp_socket);
1989 			delivered++;
1990 		}
1991 		opts = NULL;
1992 	}
1993 	m_freem(m);
1994 	*mp = NULL;
1995 	if (delivered == 0)
1996 		IP6STAT_DEC(ip6s_delivered);
1997 	return (IPPROTO_DONE);
1998 }
1999 
2000 /*
2001  * Reflect the ip6 packet back to the source.
2002  * OFF points to the icmp6 header, counted from the top of the mbuf.
2003  */
2004 static void
2005 icmp6_reflect(struct mbuf *m, size_t off)
2006 {
2007 	struct in6_addr src6, *srcp;
2008 	struct ip6_hdr *ip6;
2009 	struct icmp6_hdr *icmp6;
2010 	struct in6_ifaddr *ia = NULL;
2011 	struct ifnet *outif = NULL;
2012 	int plen;
2013 	int type, code, hlim;
2014 
2015 	/* too short to reflect */
2016 	if (off < sizeof(struct ip6_hdr)) {
2017 		nd6log((LOG_DEBUG,
2018 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2019 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2020 		    __FILE__, __LINE__));
2021 		goto bad;
2022 	}
2023 
2024 	/*
2025 	 * If there are extra headers between IPv6 and ICMPv6, strip
2026 	 * off that header first.
2027 	 */
2028 #ifdef DIAGNOSTIC
2029 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2030 		panic("assumption failed in icmp6_reflect");
2031 #endif
2032 	if (off > sizeof(struct ip6_hdr)) {
2033 		size_t l;
2034 		struct ip6_hdr nip6;
2035 
2036 		l = off - sizeof(struct ip6_hdr);
2037 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2038 		m_adj(m, l);
2039 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2040 		if (m->m_len < l) {
2041 			if ((m = m_pullup(m, l)) == NULL)
2042 				return;
2043 		}
2044 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2045 	} else /* off == sizeof(struct ip6_hdr) */ {
2046 		size_t l;
2047 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2048 		if (m->m_len < l) {
2049 			if ((m = m_pullup(m, l)) == NULL)
2050 				return;
2051 		}
2052 	}
2053 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2054 	ip6 = mtod(m, struct ip6_hdr *);
2055 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2056 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2057 	type = icmp6->icmp6_type; /* keep type for statistics */
2058 	code = icmp6->icmp6_code; /* ditto. */
2059 	hlim = 0;
2060 	srcp = NULL;
2061 
2062 	/*
2063 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2064 	 * use dst as the src for the reply.
2065 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2066 	 * (for example) when we encounter an error while forwarding procedure
2067 	 * destined to a duplicated address of ours.
2068 	 */
2069 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
2070 		ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false);
2071 		if (ia != NULL && !(ia->ia6_flags &
2072 		    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2073 			src6 = ia->ia_addr.sin6_addr;
2074 			srcp = &src6;
2075 
2076 			if (m->m_pkthdr.rcvif != NULL) {
2077 				/* XXX: This may not be the outgoing interface */
2078 				hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2079 			} else
2080 				hlim = V_ip6_defhlim;
2081 		}
2082 	}
2083 
2084 	if (srcp == NULL) {
2085 		int error;
2086 		struct in6_addr dst6;
2087 		uint32_t scopeid;
2088 
2089 		/*
2090 		 * This case matches to multicasts, our anycast, or unicasts
2091 		 * that we do not own.  Select a source address based on the
2092 		 * source address of the erroneous packet.
2093 		 */
2094 		in6_splitscope(&ip6->ip6_src, &dst6, &scopeid);
2095 		error = in6_selectsrc_addr(M_GETFIB(m), &dst6,
2096 		    scopeid, NULL, &src6, &hlim);
2097 
2098 		if (error) {
2099 			char ip6buf[INET6_ADDRSTRLEN];
2100 			nd6log((LOG_DEBUG,
2101 			    "icmp6_reflect: source can't be determined: "
2102 			    "dst=%s, error=%d\n",
2103 			    ip6_sprintf(ip6buf, &ip6->ip6_dst), error));
2104 			goto bad;
2105 		}
2106 		srcp = &src6;
2107 	}
2108 	/*
2109 	 * ip6_input() drops a packet if its src is multicast.
2110 	 * So, the src is never multicast.
2111 	 */
2112 	ip6->ip6_dst = ip6->ip6_src;
2113 	ip6->ip6_src = *srcp;
2114 	ip6->ip6_flow = 0;
2115 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2116 	ip6->ip6_vfc |= IPV6_VERSION;
2117 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2118 	ip6->ip6_hlim = hlim;
2119 
2120 	icmp6->icmp6_cksum = 0;
2121 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2122 	    sizeof(struct ip6_hdr), plen);
2123 
2124 	/*
2125 	 * XXX option handling
2126 	 */
2127 
2128 	m->m_flags &= ~(M_BCAST|M_MCAST);
2129 	m->m_pkthdr.rcvif = NULL;
2130 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2131 	if (outif)
2132 		icmp6_ifoutstat_inc(outif, type, code);
2133 
2134 	return;
2135 
2136  bad:
2137 	m_freem(m);
2138 	return;
2139 }
2140 
2141 static const char *
2142 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2143     struct in6_addr *tgt6)
2144 {
2145 	static char buf[1024];
2146 	char ip6bufs[INET6_ADDRSTRLEN];
2147 	char ip6bufd[INET6_ADDRSTRLEN];
2148 	char ip6buft[INET6_ADDRSTRLEN];
2149 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2150 	    ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2151 	    ip6_sprintf(ip6buft, tgt6));
2152 	return buf;
2153 }
2154 
2155 void
2156 icmp6_redirect_input(struct mbuf *m, int off)
2157 {
2158 	struct ifnet *ifp;
2159 	struct ip6_hdr *ip6;
2160 	struct nd_redirect *nd_rd;
2161 	struct in6_addr src6, redtgt6, reddst6;
2162 	union nd_opts ndopts;
2163 	char ip6buf[INET6_ADDRSTRLEN];
2164 	char *lladdr;
2165 	int icmp6len, is_onlink, is_router, lladdrlen;
2166 
2167 	M_ASSERTPKTHDR(m);
2168 	KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2169 
2170 	/* XXX if we are router, we don't update route by icmp6 redirect */
2171 	if (V_ip6_forwarding)
2172 		goto freeit;
2173 	if (!V_icmp6_rediraccept)
2174 		goto freeit;
2175 
2176 	/* RFC 6980: Nodes MUST silently ignore fragments */
2177 	if(m->m_flags & M_FRAGMENTED)
2178 		goto freeit;
2179 
2180 	ip6 = mtod(m, struct ip6_hdr *);
2181 	icmp6len = ntohs(ip6->ip6_plen);
2182 	if (m->m_len < off + icmp6len) {
2183 		m = m_pullup(m, off + icmp6len);
2184 		if (m == NULL) {
2185 			IP6STAT_INC(ip6s_exthdrtoolong);
2186 			return;
2187 		}
2188 	}
2189 	ip6 = mtod(m, struct ip6_hdr *);
2190 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2191 
2192 	ifp = m->m_pkthdr.rcvif;
2193 	redtgt6 = nd_rd->nd_rd_target;
2194 	reddst6 = nd_rd->nd_rd_dst;
2195 
2196 	if (in6_setscope(&redtgt6, ifp, NULL) ||
2197 	    in6_setscope(&reddst6, ifp, NULL)) {
2198 		goto freeit;
2199 	}
2200 
2201 	/* validation */
2202 	src6 = ip6->ip6_src;
2203 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2204 		nd6log((LOG_ERR,
2205 		    "ICMP6 redirect sent from %s rejected; "
2206 		    "must be from linklocal\n",
2207 		    ip6_sprintf(ip6buf, &src6)));
2208 		goto bad;
2209 	}
2210 	if (__predict_false(ip6->ip6_hlim != 255)) {
2211 		ICMP6STAT_INC(icp6s_invlhlim);
2212 		nd6log((LOG_ERR,
2213 		    "ICMP6 redirect sent from %s rejected; "
2214 		    "hlim=%d (must be 255)\n",
2215 		    ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2216 		goto bad;
2217 	}
2218     {
2219 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2220 	struct nhop_object *nh;
2221 	struct in6_addr kdst;
2222 	uint32_t scopeid;
2223 
2224 	in6_splitscope(&reddst6, &kdst, &scopeid);
2225 	NET_EPOCH_ASSERT();
2226 	nh = fib6_lookup(ifp->if_fib, &kdst, scopeid, 0, 0);
2227 	if (nh != NULL) {
2228 		struct in6_addr nh_addr;
2229 		nh_addr = ifatoia6(nh->nh_ifa)->ia_addr.sin6_addr;
2230 		if ((nh->nh_flags & NHF_GATEWAY) == 0) {
2231 			nd6log((LOG_ERR,
2232 			    "ICMP6 redirect rejected; no route "
2233 			    "with inet6 gateway found for redirect dst: %s\n",
2234 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2235 			goto bad;
2236 		}
2237 
2238 		/*
2239 		 * Embed scope zone id into next hop address.
2240 		 */
2241 		nh_addr = nh->gw6_sa.sin6_addr;
2242 
2243 		if (IN6_ARE_ADDR_EQUAL(&src6, &nh_addr) == 0) {
2244 			nd6log((LOG_ERR,
2245 			    "ICMP6 redirect rejected; "
2246 			    "not equal to gw-for-src=%s (must be same): "
2247 			    "%s\n",
2248 			    ip6_sprintf(ip6buf, &nh_addr),
2249 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2250 			goto bad;
2251 		}
2252 	} else {
2253 		nd6log((LOG_ERR,
2254 		    "ICMP6 redirect rejected; "
2255 		    "no route found for redirect dst: %s\n",
2256 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2257 		goto bad;
2258 	}
2259     }
2260 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2261 		nd6log((LOG_ERR,
2262 		    "ICMP6 redirect rejected; "
2263 		    "redirect dst must be unicast: %s\n",
2264 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2265 		goto bad;
2266 	}
2267 
2268 	is_router = is_onlink = 0;
2269 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2270 		is_router = 1;	/* router case */
2271 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2272 		is_onlink = 1;	/* on-link destination case */
2273 	if (!is_router && !is_onlink) {
2274 		nd6log((LOG_ERR,
2275 		    "ICMP6 redirect rejected; "
2276 		    "neither router case nor onlink case: %s\n",
2277 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2278 		goto bad;
2279 	}
2280 
2281 	icmp6len -= sizeof(*nd_rd);
2282 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2283 	if (nd6_options(&ndopts) < 0) {
2284 		nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2285 		    __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2286 		/* nd6_options have incremented stats */
2287 		goto freeit;
2288 	}
2289 
2290 	lladdr = NULL;
2291 	lladdrlen = 0;
2292 	if (ndopts.nd_opts_tgt_lladdr) {
2293 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2294 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2295 	}
2296 
2297 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2298 		nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2299 		    "(if %d, icmp6 packet %d): %s\n",
2300 		    __func__, ip6_sprintf(ip6buf, &redtgt6),
2301 		    ifp->if_addrlen, lladdrlen - 2,
2302 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2303 		goto bad;
2304 	}
2305 
2306 	/* Validation passed. */
2307 
2308 	/* RFC 2461 8.3 */
2309 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2310 	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2311 
2312 	/*
2313 	 * Install a gateway route in the better-router case or an interface
2314 	 * route in the on-link-destination case.
2315 	 */
2316 	{
2317 		struct sockaddr_in6 sdst;
2318 		struct sockaddr_in6 sgw;
2319 		struct sockaddr_in6 ssrc;
2320 		struct sockaddr *gw;
2321 		int rt_flags;
2322 		u_int fibnum;
2323 
2324 		bzero(&sdst, sizeof(sdst));
2325 		bzero(&ssrc, sizeof(ssrc));
2326 		sdst.sin6_family = ssrc.sin6_family = AF_INET6;
2327 		sdst.sin6_len = ssrc.sin6_len = sizeof(struct sockaddr_in6);
2328 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2329 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2330 		rt_flags = 0;
2331 		if (is_router) {
2332 			bzero(&sgw, sizeof(sgw));
2333 			sgw.sin6_family = AF_INET6;
2334 			sgw.sin6_len = sizeof(struct sockaddr_in6);
2335 			bcopy(&redtgt6, &sgw.sin6_addr,
2336 				sizeof(struct in6_addr));
2337 			gw = (struct sockaddr *)&sgw;
2338 			rt_flags |= RTF_GATEWAY;
2339 		} else
2340 			gw = ifp->if_addr->ifa_addr;
2341 		for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2342 			rib_add_redirect(fibnum, (struct sockaddr *)&sdst, gw,
2343 			    (struct sockaddr *)&ssrc, ifp, rt_flags,
2344 			    V_icmp6_redirtimeout);
2345 	}
2346 
2347  freeit:
2348 	m_freem(m);
2349 	return;
2350 
2351  bad:
2352 	ICMP6STAT_INC(icp6s_badredirect);
2353 	m_freem(m);
2354 }
2355 
2356 void
2357 icmp6_redirect_output(struct mbuf *m0, struct nhop_object *nh)
2358 {
2359 	struct ifnet *ifp;	/* my outgoing interface */
2360 	struct in6_addr *ifp_ll6;
2361 	struct in6_addr *router_ll6;
2362 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2363 	struct mbuf *m = NULL;	/* newly allocated one */
2364 	struct m_tag *mtag;
2365 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2366 	struct nd_redirect *nd_rd;
2367 	struct llentry *ln = NULL;
2368 	size_t maxlen;
2369 	u_char *p;
2370 	struct ifnet *outif = NULL;
2371 	struct sockaddr_in6 src_sa;
2372 
2373 	icmp6_errcount(ND_REDIRECT, 0);
2374 
2375 	/* if we are not router, we don't send icmp6 redirect */
2376 	if (!V_ip6_forwarding)
2377 		goto fail;
2378 
2379 	/* sanity check */
2380 	if (!m0 || !nh || !(NH_IS_VALID(nh)) || !(ifp = nh->nh_ifp))
2381 		goto fail;
2382 
2383 	/*
2384 	 * Address check:
2385 	 *  the source address must identify a neighbor, and
2386 	 *  the destination address must not be a multicast address
2387 	 *  [RFC 2461, sec 8.2]
2388 	 */
2389 	sip6 = mtod(m0, struct ip6_hdr *);
2390 	bzero(&src_sa, sizeof(src_sa));
2391 	src_sa.sin6_family = AF_INET6;
2392 	src_sa.sin6_len = sizeof(src_sa);
2393 	src_sa.sin6_addr = sip6->ip6_src;
2394 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2395 		goto fail;
2396 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2397 		goto fail;	/* what should we do here? */
2398 
2399 	/* rate limit */
2400 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2401 		goto fail;
2402 
2403 	/*
2404 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2405 	 * we almost always ask for an mbuf cluster for simplicity.
2406 	 * (MHLEN < IPV6_MMTU is almost always true)
2407 	 */
2408 #if IPV6_MMTU >= MCLBYTES
2409 # error assumption failed about IPV6_MMTU and MCLBYTES
2410 #endif
2411 	m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2412 	if (m == NULL)
2413 		goto fail;
2414 	M_SETFIB(m, M_GETFIB(m0));
2415 	maxlen = M_TRAILINGSPACE(m);
2416 	maxlen = min(IPV6_MMTU, maxlen);
2417 	/* just for safety */
2418 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2419 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2420 		goto fail;
2421 	}
2422 
2423 	{
2424 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2425 		struct in6_ifaddr *ia;
2426 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2427 						 IN6_IFF_NOTREADY|
2428 						 IN6_IFF_ANYCAST)) == NULL)
2429 			goto fail;
2430 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2431 		/* XXXRW: reference released prematurely. */
2432 		ifa_free(&ia->ia_ifa);
2433 	}
2434 
2435 	/* get ip6 linklocal address for the router. */
2436 	if (nh->nh_flags & NHF_GATEWAY) {
2437 		struct sockaddr_in6 *sin6;
2438 		sin6 = &nh->gw6_sa;
2439 		router_ll6 = &sin6->sin6_addr;
2440 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2441 			router_ll6 = (struct in6_addr *)NULL;
2442 	} else
2443 		router_ll6 = (struct in6_addr *)NULL;
2444 
2445 	/* ip6 */
2446 	ip6 = mtod(m, struct ip6_hdr *);
2447 	ip6->ip6_flow = 0;
2448 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2449 	ip6->ip6_vfc |= IPV6_VERSION;
2450 	/* ip6->ip6_plen will be set later */
2451 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2452 	ip6->ip6_hlim = 255;
2453 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2454 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2455 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2456 
2457 	/* ND Redirect */
2458 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2459 	nd_rd->nd_rd_type = ND_REDIRECT;
2460 	nd_rd->nd_rd_code = 0;
2461 	nd_rd->nd_rd_reserved = 0;
2462 	if (nh->nh_flags & NHF_GATEWAY) {
2463 		/*
2464 		 * nd_rd->nd_rd_target must be a link-local address in
2465 		 * better router cases.
2466 		 */
2467 		if (!router_ll6)
2468 			goto fail;
2469 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2470 		    sizeof(nd_rd->nd_rd_target));
2471 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2472 		    sizeof(nd_rd->nd_rd_dst));
2473 	} else {
2474 		/* make sure redtgt == reddst */
2475 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2476 		    sizeof(nd_rd->nd_rd_target));
2477 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2478 		    sizeof(nd_rd->nd_rd_dst));
2479 	}
2480 
2481 	p = (u_char *)(nd_rd + 1);
2482 
2483 	if (!router_ll6)
2484 		goto nolladdropt;
2485 
2486 	{
2487 		/* target lladdr option */
2488 		int len;
2489 		struct nd_opt_hdr *nd_opt;
2490 		char *lladdr;
2491 
2492 		ln = nd6_lookup(router_ll6, LLE_SF(AF_INET6,  0), ifp);
2493 		if (ln == NULL)
2494 			goto nolladdropt;
2495 
2496 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2497 		len = (len + 7) & ~7;	/* round by 8 */
2498 		/* safety check */
2499 		if (len + (p - (u_char *)ip6) > maxlen)
2500 			goto nolladdropt;
2501 
2502 		if (ln->la_flags & LLE_VALID) {
2503 			nd_opt = (struct nd_opt_hdr *)p;
2504 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2505 			nd_opt->nd_opt_len = len >> 3;
2506 			lladdr = (char *)(nd_opt + 1);
2507 			bcopy(ln->ll_addr, lladdr, ifp->if_addrlen);
2508 			p += len;
2509 		}
2510 	}
2511 nolladdropt:
2512 	if (ln != NULL)
2513 		LLE_RUNLOCK(ln);
2514 
2515 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2516 
2517 	/* just to be safe */
2518 #ifdef M_DECRYPTED	/*not openbsd*/
2519 	if (m0->m_flags & M_DECRYPTED)
2520 		goto noredhdropt;
2521 #endif
2522 	if (p - (u_char *)ip6 > maxlen)
2523 		goto noredhdropt;
2524 
2525 	{
2526 		/* redirected header option */
2527 		int len;
2528 		struct nd_opt_rd_hdr *nd_opt_rh;
2529 
2530 		/*
2531 		 * compute the maximum size for icmp6 redirect header option.
2532 		 * XXX room for auth header?
2533 		 */
2534 		len = maxlen - (p - (u_char *)ip6);
2535 		len &= ~7;
2536 
2537 		/* This is just for simplicity. */
2538 		if (m0->m_pkthdr.len != m0->m_len) {
2539 			if (m0->m_next) {
2540 				m_freem(m0->m_next);
2541 				m0->m_next = NULL;
2542 			}
2543 			m0->m_pkthdr.len = m0->m_len;
2544 		}
2545 
2546 		/*
2547 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2548 		 * about padding/truncate rule for the original IP packet.
2549 		 * From the discussion on IPv6imp in Feb 1999,
2550 		 * the consensus was:
2551 		 * - "attach as much as possible" is the goal
2552 		 * - pad if not aligned (original size can be guessed by
2553 		 *   original ip6 header)
2554 		 * Following code adds the padding if it is simple enough,
2555 		 * and truncates if not.
2556 		 */
2557 		if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2558 			panic("assumption failed in %s:%d", __FILE__,
2559 			    __LINE__);
2560 
2561 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2562 			/* not enough room, truncate */
2563 			m0->m_pkthdr.len = m0->m_len = len -
2564 			    sizeof(*nd_opt_rh);
2565 		} else {
2566 			/* enough room, pad or truncate */
2567 			size_t extra;
2568 
2569 			extra = m0->m_pkthdr.len % 8;
2570 			if (extra) {
2571 				/* pad if easy enough, truncate if not */
2572 				if (8 - extra <= M_TRAILINGSPACE(m0)) {
2573 					/* pad */
2574 					m0->m_len += (8 - extra);
2575 					m0->m_pkthdr.len += (8 - extra);
2576 				} else {
2577 					/* truncate */
2578 					m0->m_pkthdr.len -= extra;
2579 					m0->m_len -= extra;
2580 				}
2581 			}
2582 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2583 			m0->m_pkthdr.len = m0->m_len = len -
2584 			    sizeof(*nd_opt_rh);
2585 		}
2586 
2587 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2588 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2589 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2590 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2591 		p += sizeof(*nd_opt_rh);
2592 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2593 
2594 		/* connect m0 to m */
2595 		m_tag_delete_chain(m0, NULL);
2596 		m0->m_flags &= ~M_PKTHDR;
2597 		m->m_next = m0;
2598 		m->m_pkthdr.len = m->m_len + m0->m_len;
2599 		m0 = NULL;
2600 	}
2601 noredhdropt:;
2602 	if (m0) {
2603 		m_freem(m0);
2604 		m0 = NULL;
2605 	}
2606 
2607 	/* XXX: clear embedded link IDs in the inner header */
2608 	in6_clearscope(&sip6->ip6_src);
2609 	in6_clearscope(&sip6->ip6_dst);
2610 	in6_clearscope(&nd_rd->nd_rd_target);
2611 	in6_clearscope(&nd_rd->nd_rd_dst);
2612 
2613 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2614 
2615 	nd_rd->nd_rd_cksum = 0;
2616 	nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2617 	    sizeof(*ip6), ntohs(ip6->ip6_plen));
2618 
2619         if (send_sendso_input_hook != NULL) {
2620 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2621 			M_NOWAIT);
2622 		if (mtag == NULL)
2623 			goto fail;
2624 		*(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2625 		m_tag_prepend(m, mtag);
2626 	}
2627 
2628 	/* send the packet to outside... */
2629 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2630 	if (outif) {
2631 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2632 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2633 	}
2634 	ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2635 
2636 	return;
2637 
2638 fail:
2639 	if (m)
2640 		m_freem(m);
2641 	if (m0)
2642 		m_freem(m0);
2643 }
2644 
2645 /*
2646  * ICMPv6 socket option processing.
2647  */
2648 int
2649 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2650 {
2651 	int error = 0;
2652 	int optlen;
2653 	struct inpcb *inp = sotoinpcb(so);
2654 	int level, op, optname;
2655 
2656 	if (sopt) {
2657 		level = sopt->sopt_level;
2658 		op = sopt->sopt_dir;
2659 		optname = sopt->sopt_name;
2660 		optlen = sopt->sopt_valsize;
2661 	} else
2662 		level = op = optname = optlen = 0;
2663 
2664 	if (level != IPPROTO_ICMPV6) {
2665 		return EINVAL;
2666 	}
2667 
2668 	switch (op) {
2669 	case PRCO_SETOPT:
2670 		switch (optname) {
2671 		case ICMP6_FILTER:
2672 		    {
2673 			struct icmp6_filter ic6f;
2674 
2675 			if (optlen != sizeof(ic6f)) {
2676 				error = EMSGSIZE;
2677 				break;
2678 			}
2679 			error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2680 			if (error == 0) {
2681 				INP_WLOCK(inp);
2682 				*inp->in6p_icmp6filt = ic6f;
2683 				INP_WUNLOCK(inp);
2684 			}
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 			struct icmp6_filter ic6f;
2699 
2700 			INP_RLOCK(inp);
2701 			ic6f = *inp->in6p_icmp6filt;
2702 			INP_RUNLOCK(inp);
2703 			error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2704 			break;
2705 		    }
2706 
2707 		default:
2708 			error = ENOPROTOOPT;
2709 			break;
2710 		}
2711 		break;
2712 	}
2713 
2714 	return (error);
2715 }
2716 
2717 /*
2718  * Perform rate limit check.
2719  * Returns 0 if it is okay to send the icmp6 packet.
2720  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2721  * limitation.
2722  *
2723  * XXX per-destination/type check necessary?
2724  *
2725  * dst - not used at this moment
2726  * type - not used at this moment
2727  * code - not used at this moment
2728  */
2729 int
2730 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2731     const int code)
2732 {
2733 	int ret;
2734 
2735 	ret = 0;	/* okay to send */
2736 
2737 	/* PPS limit */
2738 	if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2739 	    V_icmp6errppslim)) {
2740 		/* The packet is subject to rate limit */
2741 		ret++;
2742 	}
2743 
2744 	return ret;
2745 }
2746