xref: /freebsd/sys/netinet6/icmp6.c (revision 3cbb4cc200f8a0ad7ed08233425ea54524a21f1c)
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/ip6protosw.h>
111 #include <netinet6/ip6_var.h>
112 #include <netinet6/scope6_var.h>
113 #include <netinet6/mld6_var.h>
114 #include <netinet6/nd6.h>
115 #include <netinet6/send.h>
116 
117 extern struct domain inet6domain;
118 
119 VNET_PCPUSTAT_DEFINE(struct icmp6stat, icmp6stat);
120 VNET_PCPUSTAT_SYSINIT(icmp6stat);
121 
122 #ifdef VIMAGE
123 VNET_PCPUSTAT_SYSUNINIT(icmp6stat);
124 #endif /* VIMAGE */
125 
126 VNET_DECLARE(struct inpcbinfo, ripcbinfo);
127 VNET_DECLARE(struct inpcbhead, ripcb);
128 VNET_DECLARE(int, icmp6errppslim);
129 VNET_DEFINE_STATIC(int, icmp6errpps_count) = 0;
130 VNET_DEFINE_STATIC(struct timeval, icmp6errppslim_last);
131 VNET_DECLARE(int, icmp6_nodeinfo);
132 
133 #define	V_ripcbinfo			VNET(ripcbinfo)
134 #define	V_ripcb				VNET(ripcb)
135 #define	V_icmp6errppslim		VNET(icmp6errppslim)
136 #define	V_icmp6errpps_count		VNET(icmp6errpps_count)
137 #define	V_icmp6errppslim_last		VNET(icmp6errppslim_last)
138 #define	V_icmp6_nodeinfo		VNET(icmp6_nodeinfo)
139 
140 static void icmp6_errcount(int, int);
141 static int icmp6_rip6_input(struct mbuf **, int);
142 static void icmp6_reflect(struct mbuf *, size_t);
143 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
144 static const char *icmp6_redirect_diag(struct in6_addr *,
145 	struct in6_addr *, struct in6_addr *);
146 static struct mbuf *ni6_input(struct mbuf *, int, struct prison *);
147 static struct mbuf *ni6_nametodns(const char *, int, int);
148 static int ni6_dnsmatch(const char *, int, const char *, int);
149 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
150 			  struct ifnet **, struct in6_addr *);
151 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
152 				struct ifnet *, int);
153 static int icmp6_notify_error(struct mbuf **, int, int, int);
154 
155 /*
156  * Kernel module interface for updating icmp6stat.  The argument is an index
157  * into icmp6stat treated as an array of u_quad_t.  While this encodes the
158  * general layout of icmp6stat into the caller, it doesn't encode its
159  * location, so that future changes to add, for example, per-CPU stats
160  * support won't cause binary compatibility problems for kernel modules.
161  */
162 void
163 kmod_icmp6stat_inc(int statnum)
164 {
165 
166 	counter_u64_add(VNET(icmp6stat)[statnum], 1);
167 }
168 
169 static void
170 icmp6_errcount(int type, int code)
171 {
172 	switch (type) {
173 	case ICMP6_DST_UNREACH:
174 		switch (code) {
175 		case ICMP6_DST_UNREACH_NOROUTE:
176 			ICMP6STAT_INC(icp6s_odst_unreach_noroute);
177 			return;
178 		case ICMP6_DST_UNREACH_ADMIN:
179 			ICMP6STAT_INC(icp6s_odst_unreach_admin);
180 			return;
181 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
182 			ICMP6STAT_INC(icp6s_odst_unreach_beyondscope);
183 			return;
184 		case ICMP6_DST_UNREACH_ADDR:
185 			ICMP6STAT_INC(icp6s_odst_unreach_addr);
186 			return;
187 		case ICMP6_DST_UNREACH_NOPORT:
188 			ICMP6STAT_INC(icp6s_odst_unreach_noport);
189 			return;
190 		}
191 		break;
192 	case ICMP6_PACKET_TOO_BIG:
193 		ICMP6STAT_INC(icp6s_opacket_too_big);
194 		return;
195 	case ICMP6_TIME_EXCEEDED:
196 		switch (code) {
197 		case ICMP6_TIME_EXCEED_TRANSIT:
198 			ICMP6STAT_INC(icp6s_otime_exceed_transit);
199 			return;
200 		case ICMP6_TIME_EXCEED_REASSEMBLY:
201 			ICMP6STAT_INC(icp6s_otime_exceed_reassembly);
202 			return;
203 		}
204 		break;
205 	case ICMP6_PARAM_PROB:
206 		switch (code) {
207 		case ICMP6_PARAMPROB_HEADER:
208 			ICMP6STAT_INC(icp6s_oparamprob_header);
209 			return;
210 		case ICMP6_PARAMPROB_NEXTHEADER:
211 			ICMP6STAT_INC(icp6s_oparamprob_nextheader);
212 			return;
213 		case ICMP6_PARAMPROB_OPTION:
214 			ICMP6STAT_INC(icp6s_oparamprob_option);
215 			return;
216 		}
217 		break;
218 	case ND_REDIRECT:
219 		ICMP6STAT_INC(icp6s_oredirect);
220 		return;
221 	}
222 	ICMP6STAT_INC(icp6s_ounknown);
223 }
224 
225 /*
226  * A wrapper function for icmp6_error() necessary when the erroneous packet
227  * may not contain enough scope zone information.
228  */
229 void
230 icmp6_error2(struct mbuf *m, int type, int code, int param,
231     struct ifnet *ifp)
232 {
233 	struct ip6_hdr *ip6;
234 
235 	if (ifp == NULL)
236 		return;
237 
238 	if (m->m_len < sizeof(struct ip6_hdr)) {
239 		m = m_pullup(m, sizeof(struct ip6_hdr));
240 		if (m == NULL) {
241 			IP6STAT_INC(ip6s_exthdrtoolong);
242 			return;
243 		}
244 	}
245 	ip6 = mtod(m, struct ip6_hdr *);
246 
247 	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
248 		return;
249 	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
250 		return;
251 
252 	icmp6_error(m, type, code, param);
253 }
254 
255 /*
256  * Generate an error packet of type error in response to bad IP6 packet.
257  */
258 void
259 icmp6_error(struct mbuf *m, int type, int code, int param)
260 {
261 	struct ip6_hdr *oip6, *nip6;
262 	struct icmp6_hdr *icmp6;
263 	u_int preplen;
264 	int off;
265 	int nxt;
266 
267 	ICMP6STAT_INC(icp6s_error);
268 
269 	/* count per-type-code statistics */
270 	icmp6_errcount(type, code);
271 
272 #ifdef M_DECRYPTED	/*not openbsd*/
273 	if (m->m_flags & M_DECRYPTED) {
274 		ICMP6STAT_INC(icp6s_canterror);
275 		goto freeit;
276 	}
277 #endif
278 
279 	if (m->m_len < sizeof(struct ip6_hdr)) {
280 		m = m_pullup(m, sizeof(struct ip6_hdr));
281 		if (m == NULL) {
282 			IP6STAT_INC(ip6s_exthdrtoolong);
283 			return;
284 		}
285 	}
286 	oip6 = mtod(m, struct ip6_hdr *);
287 
288 	/*
289 	 * If the destination address of the erroneous packet is a multicast
290 	 * address, or the packet was sent using link-layer multicast,
291 	 * we should basically suppress sending an error (RFC 2463, Section
292 	 * 2.4).
293 	 * We have two exceptions (the item e.2 in that section):
294 	 * - the Packet Too Big message can be sent for path MTU discovery.
295 	 * - the Parameter Problem Message that can be allowed an icmp6 error
296 	 *   in the option type field.  This check has been done in
297 	 *   ip6_unknown_opt(), so we can just check the type and code.
298 	 */
299 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
300 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
301 	    (type != ICMP6_PACKET_TOO_BIG &&
302 	     (type != ICMP6_PARAM_PROB ||
303 	      code != ICMP6_PARAMPROB_OPTION)))
304 		goto freeit;
305 
306 	/*
307 	 * RFC 2463, 2.4 (e.5): source address check.
308 	 * XXX: the case of anycast source?
309 	 */
310 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
311 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
312 		goto freeit;
313 
314 	/*
315 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
316 	 * don't do it.
317 	 */
318 	nxt = -1;
319 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
320 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
321 		struct icmp6_hdr *icp;
322 
323 		if (m->m_len < off + sizeof(struct icmp6_hdr)) {
324 			m = m_pullup(m, off + sizeof(struct icmp6_hdr));
325 			if (m == NULL) {
326 				IP6STAT_INC(ip6s_exthdrtoolong);
327 				return;
328 			}
329 		}
330 		oip6 = mtod(m, struct ip6_hdr *);
331 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
332 
333 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
334 		    icp->icmp6_type == ND_REDIRECT) {
335 			/*
336 			 * ICMPv6 error
337 			 * Special case: for redirect (which is
338 			 * informational) we must not send icmp6 error.
339 			 */
340 			ICMP6STAT_INC(icp6s_canterror);
341 			goto freeit;
342 		} else {
343 			/* ICMPv6 informational - send the error */
344 		}
345 	} else {
346 		/* non-ICMPv6 - send the error */
347 	}
348 
349 	/* Finally, do rate limitation check. */
350 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
351 		ICMP6STAT_INC(icp6s_toofreq);
352 		goto freeit;
353 	}
354 
355 	/*
356 	 * OK, ICMP6 can be generated.
357 	 */
358 
359 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
360 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
361 
362 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
363 	M_PREPEND(m, preplen, M_NOWAIT);	/* FIB is also copied over. */
364 	if (m == NULL) {
365 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
366 		return;
367 	}
368 
369 	nip6 = mtod(m, struct ip6_hdr *);
370 	nip6->ip6_src  = oip6->ip6_src;
371 	nip6->ip6_dst  = oip6->ip6_dst;
372 
373 	in6_clearscope(&oip6->ip6_src);
374 	in6_clearscope(&oip6->ip6_dst);
375 
376 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
377 	icmp6->icmp6_type = type;
378 	icmp6->icmp6_code = code;
379 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
380 
381 	ICMP6STAT_INC(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 whether 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, *n;
400 	struct ifnet *ifp;
401 	struct ip6_hdr *ip6, *nip6;
402 	struct icmp6_hdr *icmp6, *nicmp6;
403 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
404 	int code, error, icmp6len, ip6len, noff, off, sum;
405 
406 	NET_EPOCH_ASSERT();
407 
408 	m = *mp;
409 	off = *offp;
410 
411 	if (m->m_len < off + sizeof(struct icmp6_hdr)) {
412 		m = m_pullup(m, off + sizeof(struct icmp6_hdr));
413 		if (m == NULL) {
414 			IP6STAT_INC(ip6s_exthdrtoolong);
415 			*mp = m;
416 			return (IPPROTO_DONE);
417 		}
418 	}
419 
420 	/*
421 	 * Locate icmp6 structure in mbuf, and check
422 	 * that not corrupted and of at least minimum length
423 	 */
424 
425 	icmp6len = m->m_pkthdr.len - off;
426 	if (icmp6len < sizeof(struct icmp6_hdr)) {
427 		ICMP6STAT_INC(icp6s_tooshort);
428 		goto freeit;
429 	}
430 
431 	ip6 = mtod(m, struct ip6_hdr *);
432 	ifp = m->m_pkthdr.rcvif;
433 	/*
434 	 * Check multicast group membership.
435 	 * Note: SSM filters are not applied for ICMPv6 traffic.
436 	 */
437 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
438 		struct in6_multi	*inm;
439 
440 		inm = in6m_lookup(ifp, &ip6->ip6_dst);
441 		if (inm == NULL) {
442 			IP6STAT_INC(ip6s_notmember);
443 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
444 			goto freeit;
445 		}
446 	}
447 
448 	/* Calculate the checksum. */
449 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
450 	code = icmp6->icmp6_code;
451 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
452 		nd6log((LOG_ERR,
453 		    "ICMP6 checksum error(%d|%x) %s\n",
454 		    icmp6->icmp6_type, sum,
455 		    ip6_sprintf(ip6bufs, &ip6->ip6_src)));
456 		ICMP6STAT_INC(icp6s_checksum);
457 		goto freeit;
458 	}
459 
460 	ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
461 	icmp6_ifstat_inc(ifp, ifs6_in_msg);
462 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
463 		icmp6_ifstat_inc(ifp, ifs6_in_error);
464 
465 	ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
466 	switch (icmp6->icmp6_type) {
467 	case ICMP6_DST_UNREACH:
468 		icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
469 		switch (code) {
470 		case ICMP6_DST_UNREACH_NOROUTE:
471 		case ICMP6_DST_UNREACH_ADDR:	/* PRC_HOSTDEAD is a DOS */
472 			code = PRC_UNREACH_NET;
473 			break;
474 		case ICMP6_DST_UNREACH_ADMIN:
475 			icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
476 			code = PRC_UNREACH_ADMIN_PROHIB;
477 			break;
478 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
479 			/* I mean "source address was incorrect." */
480 			code = PRC_PARAMPROB;
481 			break;
482 		case ICMP6_DST_UNREACH_NOPORT:
483 			code = PRC_UNREACH_PORT;
484 			break;
485 		default:
486 			goto badcode;
487 		}
488 		goto deliver;
489 		break;
490 
491 	case ICMP6_PACKET_TOO_BIG:
492 		icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
493 
494 		/* validation is made in icmp6_mtudisc_update */
495 
496 		code = PRC_MSGSIZE;
497 
498 		/*
499 		 * Updating the path MTU will be done after examining
500 		 * intermediate extension headers.
501 		 */
502 		goto deliver;
503 		break;
504 
505 	case ICMP6_TIME_EXCEEDED:
506 		icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
507 		switch (code) {
508 		case ICMP6_TIME_EXCEED_TRANSIT:
509 			code = PRC_TIMXCEED_INTRANS;
510 			break;
511 		case ICMP6_TIME_EXCEED_REASSEMBLY:
512 			code = PRC_TIMXCEED_REASS;
513 			break;
514 		default:
515 			goto badcode;
516 		}
517 		goto deliver;
518 		break;
519 
520 	case ICMP6_PARAM_PROB:
521 		icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
522 		switch (code) {
523 		case ICMP6_PARAMPROB_NEXTHEADER:
524 			code = PRC_UNREACH_PROTOCOL;
525 			break;
526 		case ICMP6_PARAMPROB_HEADER:
527 		case ICMP6_PARAMPROB_OPTION:
528 			code = PRC_PARAMPROB;
529 			break;
530 		default:
531 			goto badcode;
532 		}
533 		goto deliver;
534 		break;
535 
536 	case ICMP6_ECHO_REQUEST:
537 		icmp6_ifstat_inc(ifp, ifs6_in_echo);
538 		if (code != 0)
539 			goto badcode;
540 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
541 			/* Give up remote */
542 			break;
543 		}
544 		if (!M_WRITABLE(n)
545 		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
546 			struct mbuf *n0 = n;
547 			int n0len;
548 
549 			CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) <= MHLEN);
550 			n = m_gethdr(M_NOWAIT, n0->m_type);
551 			if (n == NULL) {
552 				/* Give up remote */
553 				m_freem(n0);
554 				break;
555 			}
556 
557 			m_move_pkthdr(n, n0);	/* FIB copied. */
558 			n0len = n0->m_pkthdr.len;	/* save for use below */
559 			/*
560 			 * Copy IPv6 and ICMPv6 only.
561 			 */
562 			nip6 = mtod(n, struct ip6_hdr *);
563 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
564 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
565 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
566 			noff = sizeof(struct ip6_hdr);
567 			/* new mbuf contains only ipv6+icmpv6 headers */
568 			n->m_len = noff + sizeof(struct icmp6_hdr);
569 			/*
570 			 * Adjust mbuf.  ip6_plen will be adjusted in
571 			 * ip6_output().
572 			 */
573 			m_adj(n0, off + sizeof(struct icmp6_hdr));
574 			/* recalculate complete packet size */
575 			n->m_pkthdr.len = n0len + (noff - off);
576 			n->m_next = n0;
577 		} else {
578 			if (n->m_len < off + sizeof(*nicmp6)) {
579 				n = m_pullup(n, off + sizeof(*nicmp6));
580 				if (n == NULL) {
581 					IP6STAT_INC(ip6s_exthdrtoolong);
582 					break;
583 				}
584 			}
585 			nicmp6 = (struct icmp6_hdr *)(mtod(n, caddr_t) + off);
586 			noff = off;
587 		}
588 		if (n) {
589 			nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
590 			nicmp6->icmp6_code = 0;
591 			ICMP6STAT_INC(icp6s_reflect);
592 			ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
593 			icmp6_reflect(n, noff);
594 		}
595 		break;
596 
597 	case ICMP6_ECHO_REPLY:
598 		icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
599 		if (code != 0)
600 			goto badcode;
601 		break;
602 
603 	case MLD_LISTENER_QUERY:
604 	case MLD_LISTENER_REPORT:
605 	case MLD_LISTENER_DONE:
606 	case MLDV2_LISTENER_REPORT:
607 		/*
608 		 * Drop MLD traffic which is not link-local, has a hop limit
609 		 * of greater than 1 hop, or which does not have the
610 		 * IPv6 HBH Router Alert option.
611 		 * As IPv6 HBH options are stripped in ip6_input() we must
612 		 * check an mbuf header flag.
613 		 * XXX Should we also sanity check that these messages
614 		 * were directed to a link-local multicast prefix?
615 		 */
616 		if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
617 			goto freeit;
618 		if (mld_input(&m, off, icmp6len) != 0) {
619 			*mp = NULL;
620 			return (IPPROTO_DONE);
621 		}
622 		/* m stays. */
623 		break;
624 
625 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
626 	    {
627 		enum { WRU, FQDN } mode;
628 		struct prison *pr;
629 
630 		if (!V_icmp6_nodeinfo)
631 			break;
632 
633 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
634 			mode = WRU;
635 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
636 			mode = FQDN;
637 		else
638 			goto badlen;
639 
640 		pr = NULL;
641 		sx_slock(&allprison_lock);
642 		TAILQ_FOREACH(pr, &allprison, pr_list)
643 			if (pr->pr_vnet == ifp->if_vnet)
644 				break;
645 		sx_sunlock(&allprison_lock);
646 		if (pr == NULL)
647 			pr = curthread->td_ucred->cr_prison;
648 		if (mode == FQDN) {
649 			if (m->m_len < off + sizeof(struct icmp6_nodeinfo)) {
650 				m = m_pullup(m, off +
651 				    sizeof(struct icmp6_nodeinfo));
652 				if (m == NULL) {
653 					IP6STAT_INC(ip6s_exthdrtoolong);
654 					*mp = m;
655 					return (IPPROTO_DONE);
656 				}
657 			}
658 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
659 			if (n)
660 				n = ni6_input(n, off, pr);
661 			/* XXX meaningless if n == NULL */
662 			noff = sizeof(struct ip6_hdr);
663 		} else {
664 			u_char *p;
665 			int maxhlen, hlen;
666 
667 			/*
668 			 * XXX: this combination of flags is pointless,
669 			 * but should we keep this for compatibility?
670 			 */
671 			if ((V_icmp6_nodeinfo & (ICMP6_NODEINFO_FQDNOK |
672 			    ICMP6_NODEINFO_TMPADDROK)) !=
673 			    (ICMP6_NODEINFO_FQDNOK | ICMP6_NODEINFO_TMPADDROK))
674 				break;
675 
676 			if (code != 0)
677 				goto badcode;
678 
679 			CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) + 4 <= MHLEN);
680 			n = m_gethdr(M_NOWAIT, m->m_type);
681 			if (n == NULL) {
682 				/* Give up remote */
683 				break;
684 			}
685 			if (!m_dup_pkthdr(n, m, M_NOWAIT)) {
686 				/*
687 				 * Previous code did a blind M_COPY_PKTHDR
688 				 * and said "just for rcvif".  If true, then
689 				 * we could tolerate the dup failing (due to
690 				 * the deep copy of the tag chain).  For now
691 				 * be conservative and just fail.
692 				 */
693 				m_free(n);
694 				n = NULL;
695 				break;
696 			}
697 			/*
698 			 * Copy IPv6 and ICMPv6 only.
699 			 */
700 			nip6 = mtod(n, struct ip6_hdr *);
701 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
702 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
703 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
704 			p = (u_char *)(nicmp6 + 1);
705 			bzero(p, 4);
706 
707 			maxhlen = M_TRAILINGSPACE(n) -
708 			    (sizeof(*nip6) + sizeof(*nicmp6) + 4);
709 			mtx_lock(&pr->pr_mtx);
710 			hlen = strlen(pr->pr_hostname);
711 			if (maxhlen > hlen)
712 				maxhlen = hlen;
713 			/* meaningless TTL */
714 			bcopy(pr->pr_hostname, p + 4, maxhlen);
715 			mtx_unlock(&pr->pr_mtx);
716 			noff = sizeof(struct ip6_hdr);
717 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
718 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
719 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
720 			nicmp6->icmp6_code = 0;
721 		}
722 		if (n) {
723 			ICMP6STAT_INC(icp6s_reflect);
724 			ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
725 			icmp6_reflect(n, noff);
726 		}
727 		break;
728 	    }
729 
730 	case ICMP6_WRUREPLY:
731 		if (code != 0)
732 			goto badcode;
733 		break;
734 
735 	case ND_ROUTER_SOLICIT:
736 		icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
737 		if (code != 0)
738 			goto badcode;
739 		if (icmp6len < sizeof(struct nd_router_solicit))
740 			goto badlen;
741 		if (send_sendso_input_hook != NULL) {
742 			if (m->m_len < off + icmp6len) {
743 				m = m_pullup(m, off + icmp6len);
744 				if (m == NULL) {
745 					IP6STAT_INC(ip6s_exthdrtoolong);
746 					*mp = NULL;
747 					return (IPPROTO_DONE);
748 				}
749 			}
750 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
751 			if (error == 0) {
752 				m = NULL;
753 				goto freeit;
754 			}
755 		}
756 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
757 		nd6_rs_input(m, off, icmp6len);
758 		m = n;
759 		if (m == NULL)
760 			goto freeit;
761 		break;
762 
763 	case ND_ROUTER_ADVERT:
764 		icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
765 		if (code != 0)
766 			goto badcode;
767 		if (icmp6len < sizeof(struct nd_router_advert))
768 			goto badlen;
769 		if (send_sendso_input_hook != NULL) {
770 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
771 			if (error == 0) {
772 				m = NULL;
773 				goto freeit;
774 			}
775 		}
776 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
777 		nd6_ra_input(m, off, icmp6len);
778 		m = n;
779 		if (m == NULL)
780 			goto freeit;
781 		break;
782 
783 	case ND_NEIGHBOR_SOLICIT:
784 		icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
785 		if (code != 0)
786 			goto badcode;
787 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
788 			goto badlen;
789 		if (send_sendso_input_hook != NULL) {
790 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
791 			if (error == 0) {
792 				m = NULL;
793 				goto freeit;
794 			}
795 		}
796 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
797 		nd6_ns_input(m, off, icmp6len);
798 		m = n;
799 		if (m == NULL)
800 			goto freeit;
801 		break;
802 
803 	case ND_NEIGHBOR_ADVERT:
804 		icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
805 		if (code != 0)
806 			goto badcode;
807 		if (icmp6len < sizeof(struct nd_neighbor_advert))
808 			goto badlen;
809 		if (send_sendso_input_hook != NULL) {
810 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
811 			if (error == 0) {
812 				m = NULL;
813 				goto freeit;
814 			}
815 		}
816 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
817 		nd6_na_input(m, off, icmp6len);
818 		m = n;
819 		if (m == NULL)
820 			goto freeit;
821 		break;
822 
823 	case ND_REDIRECT:
824 		icmp6_ifstat_inc(ifp, ifs6_in_redirect);
825 		if (code != 0)
826 			goto badcode;
827 		if (icmp6len < sizeof(struct nd_redirect))
828 			goto badlen;
829 		if (send_sendso_input_hook != NULL) {
830 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
831 			if (error == 0) {
832 				m = NULL;
833 				goto freeit;
834 			}
835 		}
836 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
837 		icmp6_redirect_input(m, off);
838 		m = n;
839 		if (m == NULL)
840 			goto freeit;
841 		break;
842 
843 	case ICMP6_ROUTER_RENUMBERING:
844 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
845 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
846 			goto badcode;
847 		if (icmp6len < sizeof(struct icmp6_router_renum))
848 			goto badlen;
849 		break;
850 
851 	default:
852 		nd6log((LOG_DEBUG,
853 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
854 		    icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
855 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
856 		    ifp ? ifp->if_index : 0));
857 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
858 			/* ICMPv6 error: MUST deliver it by spec... */
859 			code = PRC_NCMDS;
860 			/* deliver */
861 		} else {
862 			/* ICMPv6 informational: MUST not deliver */
863 			break;
864 		}
865 	deliver:
866 		if (icmp6_notify_error(&m, off, icmp6len, code) != 0) {
867 			/* In this case, m should've been freed. */
868 			*mp = NULL;
869 			return (IPPROTO_DONE);
870 		}
871 		break;
872 
873 	badcode:
874 		ICMP6STAT_INC(icp6s_badcode);
875 		break;
876 
877 	badlen:
878 		ICMP6STAT_INC(icp6s_badlen);
879 		break;
880 	}
881 
882 	/* deliver the packet to appropriate sockets */
883 	icmp6_rip6_input(&m, *offp);
884 
885 	*mp = m;
886 	return (IPPROTO_DONE);
887 
888  freeit:
889 	m_freem(m);
890 	*mp = NULL;
891 	return (IPPROTO_DONE);
892 }
893 
894 static int
895 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
896 {
897 	struct mbuf *m;
898 	struct icmp6_hdr *icmp6;
899 	struct ip6_hdr *eip6;
900 	u_int32_t notifymtu;
901 	struct sockaddr_in6 icmp6src, icmp6dst;
902 
903 	m = *mp;
904 
905 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
906 		ICMP6STAT_INC(icp6s_tooshort);
907 		goto freeit;
908 	}
909 
910 	if (m->m_len < off + sizeof(*icmp6) + sizeof(struct ip6_hdr)) {
911 		m = m_pullup(m, off + sizeof(*icmp6) + sizeof(struct ip6_hdr));
912 		if (m == NULL) {
913 			IP6STAT_INC(ip6s_exthdrtoolong);
914 			*mp = m;
915 			return (-1);
916 		}
917 	}
918 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
919 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
920 	bzero(&icmp6dst, sizeof(icmp6dst));
921 
922 	/* Detect the upper level protocol */
923 	{
924 		void (*ctlfunc)(int, struct sockaddr *, void *);
925 		u_int8_t nxt = eip6->ip6_nxt;
926 		int eoff = off + sizeof(struct icmp6_hdr) +
927 		    sizeof(struct ip6_hdr);
928 		struct ip6ctlparam ip6cp;
929 		int icmp6type = icmp6->icmp6_type;
930 		struct ip6_frag *fh;
931 		struct ip6_rthdr *rth;
932 		struct ip6_rthdr0 *rth0;
933 		int rthlen;
934 
935 		while (1) { /* XXX: should avoid infinite loop explicitly? */
936 			struct ip6_ext *eh;
937 
938 			switch (nxt) {
939 			case IPPROTO_HOPOPTS:
940 			case IPPROTO_DSTOPTS:
941 			case IPPROTO_AH:
942 				if (m->m_len < eoff + sizeof(struct ip6_ext)) {
943 					m = m_pullup(m, eoff +
944 					    sizeof(struct ip6_ext));
945 					if (m == NULL) {
946 						IP6STAT_INC(ip6s_exthdrtoolong);
947 						*mp = m;
948 						return (-1);
949 					}
950 				}
951 				eh = (struct ip6_ext *)
952 				    (mtod(m, caddr_t) + eoff);
953 				if (nxt == IPPROTO_AH)
954 					eoff += (eh->ip6e_len + 2) << 2;
955 				else
956 					eoff += (eh->ip6e_len + 1) << 3;
957 				nxt = eh->ip6e_nxt;
958 				break;
959 			case IPPROTO_ROUTING:
960 				/*
961 				 * When the erroneous packet contains a
962 				 * routing header, we should examine the
963 				 * header to determine the final destination.
964 				 * Otherwise, we can't properly update
965 				 * information that depends on the final
966 				 * destination (e.g. path MTU).
967 				 */
968 				if (m->m_len < eoff + sizeof(*rth)) {
969 					m = m_pullup(m, eoff + sizeof(*rth));
970 					if (m == NULL) {
971 						IP6STAT_INC(ip6s_exthdrtoolong);
972 						*mp = m;
973 						return (-1);
974 					}
975 				}
976 				rth = (struct ip6_rthdr *)
977 				    (mtod(m, caddr_t) + eoff);
978 				rthlen = (rth->ip6r_len + 1) << 3;
979 				/*
980 				 * XXX: currently there is no
981 				 * officially defined type other
982 				 * than type-0.
983 				 * Note that if the segment left field
984 				 * is 0, all intermediate hops must
985 				 * have been passed.
986 				 */
987 				if (rth->ip6r_segleft &&
988 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
989 					int hops;
990 
991 					if (m->m_len < eoff + rthlen) {
992 						m = m_pullup(m, eoff + rthlen);
993 						if (m == NULL) {
994 							IP6STAT_INC(
995 							    ip6s_exthdrtoolong);
996 							*mp = m;
997 							return (-1);
998 						}
999 					}
1000 					rth0 = (struct ip6_rthdr0 *)
1001 					    (mtod(m, caddr_t) + eoff);
1002 
1003 					/* just ignore a bogus header */
1004 					if ((rth0->ip6r0_len % 2) == 0 &&
1005 					    (hops = rth0->ip6r0_len/2))
1006 						icmp6dst.sin6_addr = *((struct in6_addr *)(rth0 + 1) + (hops - 1));
1007 				}
1008 				eoff += rthlen;
1009 				nxt = rth->ip6r_nxt;
1010 				break;
1011 			case IPPROTO_FRAGMENT:
1012 				if (m->m_len < eoff + sizeof(struct ip6_frag)) {
1013 					m = m_pullup(m, eoff +
1014 					    sizeof(struct ip6_frag));
1015 					if (m == NULL) {
1016 						IP6STAT_INC(ip6s_exthdrtoolong);
1017 						*mp = m;
1018 						return (-1);
1019 					}
1020 				}
1021 				fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1022 				    eoff);
1023 				/*
1024 				 * Data after a fragment header is meaningless
1025 				 * unless it is the first fragment, but
1026 				 * we'll go to the notify label for path MTU
1027 				 * discovery.
1028 				 */
1029 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1030 					goto notify;
1031 
1032 				eoff += sizeof(struct ip6_frag);
1033 				nxt = fh->ip6f_nxt;
1034 				break;
1035 			default:
1036 				/*
1037 				 * This case includes ESP and the No Next
1038 				 * Header.  In such cases going to the notify
1039 				 * label does not have any meaning
1040 				 * (i.e. ctlfunc will be NULL), but we go
1041 				 * anyway since we might have to update
1042 				 * path MTU information.
1043 				 */
1044 				goto notify;
1045 			}
1046 		}
1047 	  notify:
1048 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1049 
1050 		/*
1051 		 * retrieve parameters from the inner IPv6 header, and convert
1052 		 * them into sockaddr structures.
1053 		 * XXX: there is no guarantee that the source or destination
1054 		 * addresses of the inner packet are in the same scope as
1055 		 * the addresses of the icmp packet.  But there is no other
1056 		 * way to determine the zone.
1057 		 */
1058 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1059 
1060 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1061 		icmp6dst.sin6_family = AF_INET6;
1062 		if (IN6_IS_ADDR_UNSPECIFIED(&icmp6dst.sin6_addr))
1063 			icmp6dst.sin6_addr = eip6->ip6_dst;
1064 		if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1065 			goto freeit;
1066 		bzero(&icmp6src, sizeof(icmp6src));
1067 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1068 		icmp6src.sin6_family = AF_INET6;
1069 		icmp6src.sin6_addr = eip6->ip6_src;
1070 		if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1071 			goto freeit;
1072 		icmp6src.sin6_flowinfo =
1073 		    (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1074 
1075 		ip6cp.ip6c_m = m;
1076 		ip6cp.ip6c_icmp6 = icmp6;
1077 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1078 		ip6cp.ip6c_off = eoff;
1079 		ip6cp.ip6c_finaldst = &icmp6dst.sin6_addr;
1080 		ip6cp.ip6c_src = &icmp6src;
1081 		ip6cp.ip6c_nxt = nxt;
1082 
1083 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1084 			notifymtu = ntohl(icmp6->icmp6_mtu);
1085 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1086 			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1087 		}
1088 
1089 		ctlfunc = (void (*)(int, struct sockaddr *, void *))
1090 		    (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1091 		if (ctlfunc) {
1092 			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1093 			    &ip6cp);
1094 		}
1095 	}
1096 	*mp = m;
1097 	return (0);
1098 
1099   freeit:
1100 	m_freem(m);
1101 	*mp = NULL;
1102 	return (-1);
1103 }
1104 
1105 void
1106 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1107 {
1108 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1109 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1110 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1111 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1112 	struct in_conninfo inc;
1113 
1114 #if 0
1115 	/*
1116 	 * RFC2460 section 5, last paragraph.
1117 	 * even though minimum link MTU for IPv6 is IPV6_MMTU,
1118 	 * we may see ICMPv6 too big with mtu < IPV6_MMTU
1119 	 * due to packet translator in the middle.
1120 	 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1121 	 * special handling.
1122 	 */
1123 	if (mtu < IPV6_MMTU)
1124 		return;
1125 #endif
1126 
1127 	/*
1128 	 * we reject ICMPv6 too big with abnormally small value.
1129 	 * XXX what is the good definition of "abnormally small"?
1130 	 */
1131 	if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1132 		return;
1133 
1134 	if (!validated)
1135 		return;
1136 
1137 	/*
1138 	 * In case the suggested mtu is less than IPV6_MMTU, we
1139 	 * only need to remember that it was for above mentioned
1140 	 * "alwaysfrag" case.
1141 	 * Try to be as close to the spec as possible.
1142 	 */
1143 	if (mtu < IPV6_MMTU)
1144 		mtu = IPV6_MMTU - 8;
1145 
1146 	bzero(&inc, sizeof(inc));
1147 	inc.inc_fibnum = M_GETFIB(m);
1148 	inc.inc_flags |= INC_ISIPV6;
1149 	inc.inc6_faddr = *dst;
1150 	if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1151 		return;
1152 
1153 	if (mtu < tcp_maxmtu6(&inc, NULL)) {
1154 		tcp_hc_updatemtu(&inc, mtu);
1155 		ICMP6STAT_INC(icp6s_pmtuchg);
1156 	}
1157 }
1158 
1159 /*
1160  * Process a Node Information Query packet, based on
1161  * draft-ietf-ipngwg-icmp-name-lookups-07.
1162  *
1163  * Spec incompatibilities:
1164  * - IPv6 Subject address handling
1165  * - IPv4 Subject address handling support missing
1166  * - Proxy reply (answer even if it's not for me)
1167  * - joins NI group address at in6_ifattach() time only, does not cope
1168  *   with hostname changes by sethostname(3)
1169  */
1170 static struct mbuf *
1171 ni6_input(struct mbuf *m, int off, struct prison *pr)
1172 {
1173 	struct icmp6_nodeinfo *ni6, *nni6;
1174 	struct mbuf *n = NULL;
1175 	u_int16_t qtype;
1176 	int subjlen;
1177 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1178 	struct ni_reply_fqdn *fqdn;
1179 	int addrs;		/* for NI_QTYPE_NODEADDR */
1180 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1181 	struct in6_addr in6_subj; /* subject address */
1182 	struct ip6_hdr *ip6;
1183 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1184 	char *subj = NULL;
1185 	struct in6_ifaddr *ia6 = NULL;
1186 
1187 	ip6 = mtod(m, struct ip6_hdr *);
1188 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1189 
1190 	/*
1191 	 * Validate IPv6 source address.
1192 	 * The default configuration MUST be to refuse answering queries from
1193 	 * global-scope addresses according to RFC4602.
1194 	 * Notes:
1195 	 *  - it's not very clear what "refuse" means; this implementation
1196 	 *    simply drops it.
1197 	 *  - it's not very easy to identify global-scope (unicast) addresses
1198 	 *    since there are many prefixes for them.  It should be safer
1199 	 *    and in practice sufficient to check "all" but loopback and
1200 	 *    link-local (note that site-local unicast was deprecated and
1201 	 *    ULA is defined as global scope-wise)
1202 	 */
1203 	if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1204 	    !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1205 	    !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1206 		goto bad;
1207 
1208 	/*
1209 	 * Validate IPv6 destination address.
1210 	 *
1211 	 * The Responder must discard the Query without further processing
1212 	 * unless it is one of the Responder's unicast or anycast addresses, or
1213 	 * a link-local scope multicast address which the Responder has joined.
1214 	 * [RFC4602, Section 5.]
1215 	 */
1216 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1217 		if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1218 			goto bad;
1219 		/* else it's a link-local multicast, fine */
1220 	} else {		/* unicast or anycast */
1221 		ia6 = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */);
1222 		if (ia6 == NULL)
1223 			goto bad; /* XXX impossible */
1224 
1225 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1226 		    !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1227 			ifa_free(&ia6->ia_ifa);
1228 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1229 				"a temporary address in %s:%d",
1230 			       __FILE__, __LINE__));
1231 			goto bad;
1232 		}
1233 		ifa_free(&ia6->ia_ifa);
1234 	}
1235 
1236 	/* validate query Subject field. */
1237 	qtype = ntohs(ni6->ni_qtype);
1238 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1239 	switch (qtype) {
1240 	case NI_QTYPE_NOOP:
1241 	case NI_QTYPE_SUPTYPES:
1242 		/* 07 draft */
1243 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1244 			break;
1245 		/* FALLTHROUGH */
1246 	case NI_QTYPE_FQDN:
1247 	case NI_QTYPE_NODEADDR:
1248 	case NI_QTYPE_IPV4ADDR:
1249 		switch (ni6->ni_code) {
1250 		case ICMP6_NI_SUBJ_IPV6:
1251 #if ICMP6_NI_SUBJ_IPV6 != 0
1252 		case 0:
1253 #endif
1254 			/*
1255 			 * backward compatibility - try to accept 03 draft
1256 			 * format, where no Subject is present.
1257 			 */
1258 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1259 			    subjlen == 0) {
1260 				oldfqdn++;
1261 				break;
1262 			}
1263 #if ICMP6_NI_SUBJ_IPV6 != 0
1264 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1265 				goto bad;
1266 #endif
1267 
1268 			if (subjlen != sizeof(struct in6_addr))
1269 				goto bad;
1270 
1271 			/*
1272 			 * Validate Subject address.
1273 			 *
1274 			 * Not sure what exactly "address belongs to the node"
1275 			 * means in the spec, is it just unicast, or what?
1276 			 *
1277 			 * At this moment we consider Subject address as
1278 			 * "belong to the node" if the Subject address equals
1279 			 * to the IPv6 destination address; validation for
1280 			 * IPv6 destination address should have done enough
1281 			 * check for us.
1282 			 *
1283 			 * We do not do proxy at this moment.
1284 			 */
1285 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1286 			    subjlen, (caddr_t)&in6_subj);
1287 			if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1288 				goto bad;
1289 
1290 			subj = (char *)&in6_subj;
1291 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1292 				break;
1293 
1294 			/*
1295 			 * XXX if we are to allow other cases, we should really
1296 			 * be careful about scope here.
1297 			 * basically, we should disallow queries toward IPv6
1298 			 * destination X with subject Y,
1299 			 * if scope(X) > scope(Y).
1300 			 * if we allow scope(X) > scope(Y), it will result in
1301 			 * information leakage across scope boundary.
1302 			 */
1303 			goto bad;
1304 
1305 		case ICMP6_NI_SUBJ_FQDN:
1306 			/*
1307 			 * Validate Subject name with gethostname(3).
1308 			 *
1309 			 * The behavior may need some debate, since:
1310 			 * - we are not sure if the node has FQDN as
1311 			 *   hostname (returned by gethostname(3)).
1312 			 * - the code does wildcard match for truncated names.
1313 			 *   however, we are not sure if we want to perform
1314 			 *   wildcard match, if gethostname(3) side has
1315 			 *   truncated hostname.
1316 			 */
1317 			mtx_lock(&pr->pr_mtx);
1318 			n = ni6_nametodns(pr->pr_hostname,
1319 			    strlen(pr->pr_hostname), 0);
1320 			mtx_unlock(&pr->pr_mtx);
1321 			if (!n || n->m_next || n->m_len == 0)
1322 				goto bad;
1323 			if (m->m_len < off + sizeof(struct icmp6_nodeinfo) +
1324 			    subjlen) {
1325 				m = m_pullup(m, off +
1326 				    sizeof(struct icmp6_nodeinfo) + subjlen);
1327 				if (m == NULL) {
1328 					IP6STAT_INC(ip6s_exthdrtoolong);
1329 					goto bad;
1330 				}
1331 			}
1332 			/* ip6 possibly invalid but not used after. */
1333 			ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1334 			subj = (char *)(mtod(m, caddr_t) + off +
1335 			    sizeof(struct icmp6_nodeinfo));
1336 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1337 			    n->m_len)) {
1338 				goto bad;
1339 			}
1340 			m_freem(n);
1341 			n = NULL;
1342 			break;
1343 
1344 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1345 		default:
1346 			goto bad;
1347 		}
1348 		break;
1349 	}
1350 
1351 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1352 	switch (qtype) {
1353 	case NI_QTYPE_FQDN:
1354 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1355 			goto bad;
1356 		break;
1357 	case NI_QTYPE_NODEADDR:
1358 	case NI_QTYPE_IPV4ADDR:
1359 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1360 			goto bad;
1361 		break;
1362 	}
1363 
1364 	/* guess reply length */
1365 	switch (qtype) {
1366 	case NI_QTYPE_NOOP:
1367 		break;		/* no reply data */
1368 	case NI_QTYPE_SUPTYPES:
1369 		replylen += sizeof(u_int32_t);
1370 		break;
1371 	case NI_QTYPE_FQDN:
1372 		/* XXX will append an mbuf */
1373 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1374 		break;
1375 	case NI_QTYPE_NODEADDR:
1376 		addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1377 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1378 		    sizeof(u_int32_t))) > MCLBYTES)
1379 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1380 		break;
1381 	case NI_QTYPE_IPV4ADDR:
1382 		/* unsupported - should respond with unknown Qtype? */
1383 		break;
1384 	default:
1385 		/*
1386 		 * XXX: We must return a reply with the ICMP6 code
1387 		 * `unknown Qtype' in this case.  However we regard the case
1388 		 * as an FQDN query for backward compatibility.
1389 		 * Older versions set a random value to this field,
1390 		 * so it rarely varies in the defined qtypes.
1391 		 * But the mechanism is not reliable...
1392 		 * maybe we should obsolete older versions.
1393 		 */
1394 		qtype = NI_QTYPE_FQDN;
1395 		/* XXX will append an mbuf */
1396 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1397 		oldfqdn++;
1398 		break;
1399 	}
1400 
1401 	/* Allocate an mbuf to reply. */
1402 	if (replylen > MCLBYTES) {
1403 		/*
1404 		 * XXX: should we try to allocate more? But MCLBYTES
1405 		 * is probably much larger than IPV6_MMTU...
1406 		 */
1407 		goto bad;
1408 	}
1409 	if (replylen > MHLEN)
1410 		n = m_getcl(M_NOWAIT, m->m_type, M_PKTHDR);
1411 	else
1412 		n = m_gethdr(M_NOWAIT, m->m_type);
1413 	if (n == NULL) {
1414 		m_freem(m);
1415 		return (NULL);
1416 	}
1417 	m_move_pkthdr(n, m); /* just for recvif and FIB */
1418 	n->m_pkthdr.len = n->m_len = replylen;
1419 
1420 	/* copy mbuf header and IPv6 + Node Information base headers */
1421 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1422 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1423 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1424 
1425 	/* qtype dependent procedure */
1426 	switch (qtype) {
1427 	case NI_QTYPE_NOOP:
1428 		nni6->ni_code = ICMP6_NI_SUCCESS;
1429 		nni6->ni_flags = 0;
1430 		break;
1431 	case NI_QTYPE_SUPTYPES:
1432 	{
1433 		u_int32_t v;
1434 		nni6->ni_code = ICMP6_NI_SUCCESS;
1435 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1436 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1437 		v = (u_int32_t)htonl(0x0000000f);
1438 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1439 		break;
1440 	}
1441 	case NI_QTYPE_FQDN:
1442 		nni6->ni_code = ICMP6_NI_SUCCESS;
1443 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1444 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1445 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1446 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1447 		/*
1448 		 * XXX do we really have FQDN in hostname?
1449 		 */
1450 		mtx_lock(&pr->pr_mtx);
1451 		n->m_next = ni6_nametodns(pr->pr_hostname,
1452 		    strlen(pr->pr_hostname), oldfqdn);
1453 		mtx_unlock(&pr->pr_mtx);
1454 		if (n->m_next == NULL)
1455 			goto bad;
1456 		/* XXX we assume that n->m_next is not a chain */
1457 		if (n->m_next->m_next != NULL)
1458 			goto bad;
1459 		n->m_pkthdr.len += n->m_next->m_len;
1460 		break;
1461 	case NI_QTYPE_NODEADDR:
1462 	{
1463 		int lenlim, copied;
1464 
1465 		nni6->ni_code = ICMP6_NI_SUCCESS;
1466 		n->m_pkthdr.len = n->m_len =
1467 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1468 		lenlim = M_TRAILINGSPACE(n);
1469 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1470 		/* XXX: reset mbuf length */
1471 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1472 		    sizeof(struct icmp6_nodeinfo) + copied;
1473 		break;
1474 	}
1475 	default:
1476 		break;		/* XXX impossible! */
1477 	}
1478 
1479 	nni6->ni_type = ICMP6_NI_REPLY;
1480 	m_freem(m);
1481 	return (n);
1482 
1483   bad:
1484 	m_freem(m);
1485 	if (n)
1486 		m_freem(n);
1487 	return (NULL);
1488 }
1489 
1490 /*
1491  * make a mbuf with DNS-encoded string.  no compression support.
1492  *
1493  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1494  * treated as truncated name (two \0 at the end).  this is a wild guess.
1495  *
1496  * old - return pascal string if non-zero
1497  */
1498 static struct mbuf *
1499 ni6_nametodns(const char *name, int namelen, int old)
1500 {
1501 	struct mbuf *m;
1502 	char *cp, *ep;
1503 	const char *p, *q;
1504 	int i, len, nterm;
1505 
1506 	if (old)
1507 		len = namelen + 1;
1508 	else
1509 		len = MCLBYTES;
1510 
1511 	/* Because MAXHOSTNAMELEN is usually 256, we use cluster mbuf. */
1512 	if (len > MLEN)
1513 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1514 	else
1515 		m = m_get(M_NOWAIT, MT_DATA);
1516 	if (m == NULL)
1517 		goto fail;
1518 
1519 	if (old) {
1520 		m->m_len = len;
1521 		*mtod(m, char *) = namelen;
1522 		bcopy(name, mtod(m, char *) + 1, namelen);
1523 		return m;
1524 	} else {
1525 		m->m_len = 0;
1526 		cp = mtod(m, char *);
1527 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1528 
1529 		/* if not certain about my name, return empty buffer */
1530 		if (namelen == 0)
1531 			return m;
1532 
1533 		/*
1534 		 * guess if it looks like shortened hostname, or FQDN.
1535 		 * shortened hostname needs two trailing "\0".
1536 		 */
1537 		i = 0;
1538 		for (p = name; p < name + namelen; p++) {
1539 			if (*p && *p == '.')
1540 				i++;
1541 		}
1542 		if (i < 2)
1543 			nterm = 2;
1544 		else
1545 			nterm = 1;
1546 
1547 		p = name;
1548 		while (cp < ep && p < name + namelen) {
1549 			i = 0;
1550 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1551 				i++;
1552 			/* result does not fit into mbuf */
1553 			if (cp + i + 1 >= ep)
1554 				goto fail;
1555 			/*
1556 			 * DNS label length restriction, RFC1035 page 8.
1557 			 * "i == 0" case is included here to avoid returning
1558 			 * 0-length label on "foo..bar".
1559 			 */
1560 			if (i <= 0 || i >= 64)
1561 				goto fail;
1562 			*cp++ = i;
1563 			bcopy(p, cp, i);
1564 			cp += i;
1565 			p = q;
1566 			if (p < name + namelen && *p == '.')
1567 				p++;
1568 		}
1569 		/* termination */
1570 		if (cp + nterm >= ep)
1571 			goto fail;
1572 		while (nterm-- > 0)
1573 			*cp++ = '\0';
1574 		m->m_len = cp - mtod(m, char *);
1575 		return m;
1576 	}
1577 
1578 	panic("should not reach here");
1579 	/* NOTREACHED */
1580 
1581  fail:
1582 	if (m)
1583 		m_freem(m);
1584 	return NULL;
1585 }
1586 
1587 /*
1588  * check if two DNS-encoded string matches.  takes care of truncated
1589  * form (with \0\0 at the end).  no compression support.
1590  * XXX upper/lowercase match (see RFC2065)
1591  */
1592 static int
1593 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1594 {
1595 	const char *a0, *b0;
1596 	int l;
1597 
1598 	/* simplest case - need validation? */
1599 	if (alen == blen && bcmp(a, b, alen) == 0)
1600 		return 1;
1601 
1602 	a0 = a;
1603 	b0 = b;
1604 
1605 	/* termination is mandatory */
1606 	if (alen < 2 || blen < 2)
1607 		return 0;
1608 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1609 		return 0;
1610 	alen--;
1611 	blen--;
1612 
1613 	while (a - a0 < alen && b - b0 < blen) {
1614 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1615 			return 0;
1616 
1617 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1618 			return 0;
1619 		/* we don't support compression yet */
1620 		if (a[0] >= 64 || b[0] >= 64)
1621 			return 0;
1622 
1623 		/* truncated case */
1624 		if (a[0] == 0 && a - a0 == alen - 1)
1625 			return 1;
1626 		if (b[0] == 0 && b - b0 == blen - 1)
1627 			return 1;
1628 		if (a[0] == 0 || b[0] == 0)
1629 			return 0;
1630 
1631 		if (a[0] != b[0])
1632 			return 0;
1633 		l = a[0];
1634 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1635 			return 0;
1636 		if (bcmp(a + 1, b + 1, l) != 0)
1637 			return 0;
1638 
1639 		a += 1 + l;
1640 		b += 1 + l;
1641 	}
1642 
1643 	if (a - a0 == alen && b - b0 == blen)
1644 		return 1;
1645 	else
1646 		return 0;
1647 }
1648 
1649 /*
1650  * calculate the number of addresses to be returned in the node info reply.
1651  */
1652 static int
1653 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1654     struct in6_addr *subj)
1655 {
1656 	struct ifnet *ifp;
1657 	struct in6_ifaddr *ifa6;
1658 	struct ifaddr *ifa;
1659 	int addrs = 0, addrsofif, iffound = 0;
1660 	int niflags = ni6->ni_flags;
1661 
1662 	NET_EPOCH_ASSERT();
1663 
1664 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1665 		switch (ni6->ni_code) {
1666 		case ICMP6_NI_SUBJ_IPV6:
1667 			if (subj == NULL) /* must be impossible... */
1668 				return (0);
1669 			break;
1670 		default:
1671 			/*
1672 			 * XXX: we only support IPv6 subject address for
1673 			 * this Qtype.
1674 			 */
1675 			return (0);
1676 		}
1677 	}
1678 
1679 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1680 		addrsofif = 0;
1681 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1682 			if (ifa->ifa_addr->sa_family != AF_INET6)
1683 				continue;
1684 			ifa6 = (struct in6_ifaddr *)ifa;
1685 
1686 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1687 			    IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1688 				iffound = 1;
1689 
1690 			/*
1691 			 * IPv4-mapped addresses can only be returned by a
1692 			 * Node Information proxy, since they represent
1693 			 * addresses of IPv4-only nodes, which perforce do
1694 			 * not implement this protocol.
1695 			 * [icmp-name-lookups-07, Section 5.4]
1696 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1697 			 * this function at this moment.
1698 			 */
1699 
1700 			/* What do we have to do about ::1? */
1701 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1702 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1703 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1704 					continue;
1705 				break;
1706 			case IPV6_ADDR_SCOPE_SITELOCAL:
1707 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1708 					continue;
1709 				break;
1710 			case IPV6_ADDR_SCOPE_GLOBAL:
1711 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1712 					continue;
1713 				break;
1714 			default:
1715 				continue;
1716 			}
1717 
1718 			/*
1719 			 * check if anycast is okay.
1720 			 * XXX: just experimental.  not in the spec.
1721 			 */
1722 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1723 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1724 				continue; /* we need only unicast addresses */
1725 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1726 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1727 				continue;
1728 			}
1729 			addrsofif++; /* count the address */
1730 		}
1731 		if (iffound) {
1732 			*ifpp = ifp;
1733 			return (addrsofif);
1734 		}
1735 
1736 		addrs += addrsofif;
1737 	}
1738 
1739 	return (addrs);
1740 }
1741 
1742 static int
1743 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1744     struct ifnet *ifp0, int resid)
1745 {
1746 	struct ifnet *ifp;
1747 	struct in6_ifaddr *ifa6;
1748 	struct ifaddr *ifa;
1749 	struct ifnet *ifp_dep = NULL;
1750 	int copied = 0, allow_deprecated = 0;
1751 	u_char *cp = (u_char *)(nni6 + 1);
1752 	int niflags = ni6->ni_flags;
1753 	u_int32_t ltime;
1754 
1755 	NET_EPOCH_ASSERT();
1756 
1757 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1758 		return (0);	/* needless to copy */
1759 
1760 	ifp = ifp0 ? ifp0 : CK_STAILQ_FIRST(&V_ifnet);
1761   again:
1762 
1763 	for (; ifp; ifp = CK_STAILQ_NEXT(ifp, if_link)) {
1764 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1765 			if (ifa->ifa_addr->sa_family != AF_INET6)
1766 				continue;
1767 			ifa6 = (struct in6_ifaddr *)ifa;
1768 
1769 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1770 			    allow_deprecated == 0) {
1771 				/*
1772 				 * prefererred address should be put before
1773 				 * deprecated addresses.
1774 				 */
1775 
1776 				/* record the interface for later search */
1777 				if (ifp_dep == NULL)
1778 					ifp_dep = ifp;
1779 
1780 				continue;
1781 			} else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1782 			    allow_deprecated != 0)
1783 				continue; /* we now collect deprecated addrs */
1784 
1785 			/* What do we have to do about ::1? */
1786 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1787 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1788 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1789 					continue;
1790 				break;
1791 			case IPV6_ADDR_SCOPE_SITELOCAL:
1792 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1793 					continue;
1794 				break;
1795 			case IPV6_ADDR_SCOPE_GLOBAL:
1796 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1797 					continue;
1798 				break;
1799 			default:
1800 				continue;
1801 			}
1802 
1803 			/*
1804 			 * check if anycast is okay.
1805 			 * XXX: just experimental.  not in the spec.
1806 			 */
1807 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1808 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1809 				continue;
1810 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1811 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1812 				continue;
1813 			}
1814 
1815 			/* now we can copy the address */
1816 			if (resid < sizeof(struct in6_addr) +
1817 			    sizeof(u_int32_t)) {
1818 				/*
1819 				 * We give up much more copy.
1820 				 * Set the truncate flag and return.
1821 				 */
1822 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1823 				return (copied);
1824 			}
1825 
1826 			/*
1827 			 * Set the TTL of the address.
1828 			 * The TTL value should be one of the following
1829 			 * according to the specification:
1830 			 *
1831 			 * 1. The remaining lifetime of a DHCP lease on the
1832 			 *    address, or
1833 			 * 2. The remaining Valid Lifetime of a prefix from
1834 			 *    which the address was derived through Stateless
1835 			 *    Autoconfiguration.
1836 			 *
1837 			 * Note that we currently do not support stateful
1838 			 * address configuration by DHCPv6, so the former
1839 			 * case can't happen.
1840 			 */
1841 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1842 				ltime = ND6_INFINITE_LIFETIME;
1843 			else {
1844 				if (ifa6->ia6_lifetime.ia6t_expire >
1845 				    time_uptime)
1846 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime);
1847 				else
1848 					ltime = 0;
1849 			}
1850 
1851 			bcopy(&ltime, cp, sizeof(u_int32_t));
1852 			cp += sizeof(u_int32_t);
1853 
1854 			/* copy the address itself */
1855 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1856 			    sizeof(struct in6_addr));
1857 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1858 			cp += sizeof(struct in6_addr);
1859 
1860 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1861 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1862 		}
1863 		if (ifp0)	/* we need search only on the specified IF */
1864 			break;
1865 	}
1866 
1867 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1868 		ifp = ifp_dep;
1869 		allow_deprecated = 1;
1870 
1871 		goto again;
1872 	}
1873 
1874 	return (copied);
1875 }
1876 
1877 /*
1878  * XXX almost dup'ed code with rip6_input.
1879  */
1880 static int
1881 icmp6_rip6_input(struct mbuf **mp, int off)
1882 {
1883 	struct mbuf *m = *mp;
1884 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1885 	struct inpcb *inp;
1886 	struct inpcb *last = NULL;
1887 	struct sockaddr_in6 fromsa;
1888 	struct icmp6_hdr *icmp6;
1889 	struct mbuf *opts = NULL;
1890 
1891 	NET_EPOCH_ASSERT();
1892 
1893 	/* This is assumed to be safe; icmp6_input() does a pullup. */
1894 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1895 
1896 	/*
1897 	 * XXX: the address may have embedded scope zone ID, which should be
1898 	 * hidden from applications.
1899 	 */
1900 	bzero(&fromsa, sizeof(fromsa));
1901 	fromsa.sin6_family = AF_INET6;
1902 	fromsa.sin6_len = sizeof(struct sockaddr_in6);
1903 	fromsa.sin6_addr = ip6->ip6_src;
1904 	if (sa6_recoverscope(&fromsa)) {
1905 		m_freem(m);
1906 		*mp = NULL;
1907 		return (IPPROTO_DONE);
1908 	}
1909 
1910 	CK_LIST_FOREACH(inp, &V_ripcb, inp_list) {
1911 		if ((inp->inp_vflag & INP_IPV6) == 0)
1912 			continue;
1913 		if (inp->inp_ip_p != IPPROTO_ICMPV6)
1914 			continue;
1915 		if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1916 		   !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &ip6->ip6_dst))
1917 			continue;
1918 		if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1919 		   !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &ip6->ip6_src))
1920 			continue;
1921 		INP_RLOCK(inp);
1922 		if (__predict_false(inp->inp_flags2 & INP_FREED)) {
1923 			INP_RUNLOCK(inp);
1924 			continue;
1925 		}
1926 		if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1927 		    inp->in6p_icmp6filt)) {
1928 			INP_RUNLOCK(inp);
1929 			continue;
1930 		}
1931 		if (last != NULL) {
1932 			struct	mbuf *n = NULL;
1933 
1934 			/*
1935 			 * Recent network drivers tend to allocate a single
1936 			 * mbuf cluster, rather than to make a couple of
1937 			 * mbufs without clusters.  Also, since the IPv6 code
1938 			 * path tries to avoid m_pullup(), it is highly
1939 			 * probable that we still have an mbuf cluster here
1940 			 * even though the necessary length can be stored in an
1941 			 * mbuf's internal buffer.
1942 			 * Meanwhile, the default size of the receive socket
1943 			 * buffer for raw sockets is not so large.  This means
1944 			 * the possibility of packet loss is relatively higher
1945 			 * than before.  To avoid this scenario, we copy the
1946 			 * received data to a separate mbuf that does not use
1947 			 * a cluster, if possible.
1948 			 * XXX: it is better to copy the data after stripping
1949 			 * intermediate headers.
1950 			 */
1951 			if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1952 			    m->m_len <= MHLEN) {
1953 				n = m_get(M_NOWAIT, m->m_type);
1954 				if (n != NULL) {
1955 					if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1956 						bcopy(m->m_data, n->m_data,
1957 						      m->m_len);
1958 						n->m_len = m->m_len;
1959 					} else {
1960 						m_free(n);
1961 						n = NULL;
1962 					}
1963 				}
1964 			}
1965 			if (n != NULL ||
1966 			    (n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) != NULL) {
1967 				if (last->inp_flags & INP_CONTROLOPTS)
1968 					ip6_savecontrol(last, n, &opts);
1969 				/* strip intermediate headers */
1970 				m_adj(n, off);
1971 				SOCKBUF_LOCK(&last->inp_socket->so_rcv);
1972 				if (sbappendaddr_locked(
1973 				    &last->inp_socket->so_rcv,
1974 				    (struct sockaddr *)&fromsa, n, opts)
1975 				    == 0) {
1976 					/* should notify about lost packet */
1977 					m_freem(n);
1978 					if (opts) {
1979 						m_freem(opts);
1980 					}
1981 					SOCKBUF_UNLOCK(
1982 					    &last->inp_socket->so_rcv);
1983 				} else
1984 					sorwakeup_locked(last->inp_socket);
1985 				opts = NULL;
1986 			}
1987 			INP_RUNLOCK(last);
1988 		}
1989 		last = inp;
1990 	}
1991 	if (last != NULL) {
1992 		if (last->inp_flags & INP_CONTROLOPTS)
1993 			ip6_savecontrol(last, m, &opts);
1994 		/* strip intermediate headers */
1995 		m_adj(m, off);
1996 
1997 		/* avoid using mbuf clusters if possible (see above) */
1998 		if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1999 		    m->m_len <= MHLEN) {
2000 			struct mbuf *n;
2001 
2002 			n = m_get(M_NOWAIT, m->m_type);
2003 			if (n != NULL) {
2004 				if (m_dup_pkthdr(n, m, M_NOWAIT)) {
2005 					bcopy(m->m_data, n->m_data, m->m_len);
2006 					n->m_len = m->m_len;
2007 
2008 					m_freem(m);
2009 					m = n;
2010 				} else {
2011 					m_freem(n);
2012 					n = NULL;
2013 				}
2014 			}
2015 		}
2016 		SOCKBUF_LOCK(&last->inp_socket->so_rcv);
2017 		if (sbappendaddr_locked(&last->inp_socket->so_rcv,
2018 		    (struct sockaddr *)&fromsa, m, opts) == 0) {
2019 			m_freem(m);
2020 			if (opts)
2021 				m_freem(opts);
2022 			SOCKBUF_UNLOCK(&last->inp_socket->so_rcv);
2023 		} else
2024 			sorwakeup_locked(last->inp_socket);
2025 		INP_RUNLOCK(last);
2026 	} else {
2027 		m_freem(m);
2028 		IP6STAT_DEC(ip6s_delivered);
2029 	}
2030 	*mp = NULL;
2031 	return (IPPROTO_DONE);
2032 }
2033 
2034 /*
2035  * Reflect the ip6 packet back to the source.
2036  * OFF points to the icmp6 header, counted from the top of the mbuf.
2037  */
2038 static void
2039 icmp6_reflect(struct mbuf *m, size_t off)
2040 {
2041 	struct in6_addr src6, *srcp;
2042 	struct ip6_hdr *ip6;
2043 	struct icmp6_hdr *icmp6;
2044 	struct in6_ifaddr *ia = NULL;
2045 	struct ifnet *outif = NULL;
2046 	int plen;
2047 	int type, code, hlim;
2048 
2049 	/* too short to reflect */
2050 	if (off < sizeof(struct ip6_hdr)) {
2051 		nd6log((LOG_DEBUG,
2052 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2053 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2054 		    __FILE__, __LINE__));
2055 		goto bad;
2056 	}
2057 
2058 	/*
2059 	 * If there are extra headers between IPv6 and ICMPv6, strip
2060 	 * off that header first.
2061 	 */
2062 #ifdef DIAGNOSTIC
2063 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2064 		panic("assumption failed in icmp6_reflect");
2065 #endif
2066 	if (off > sizeof(struct ip6_hdr)) {
2067 		size_t l;
2068 		struct ip6_hdr nip6;
2069 
2070 		l = off - sizeof(struct ip6_hdr);
2071 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2072 		m_adj(m, l);
2073 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2074 		if (m->m_len < l) {
2075 			if ((m = m_pullup(m, l)) == NULL)
2076 				return;
2077 		}
2078 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2079 	} else /* off == sizeof(struct ip6_hdr) */ {
2080 		size_t l;
2081 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2082 		if (m->m_len < l) {
2083 			if ((m = m_pullup(m, l)) == NULL)
2084 				return;
2085 		}
2086 	}
2087 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2088 	ip6 = mtod(m, struct ip6_hdr *);
2089 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2090 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2091 	type = icmp6->icmp6_type; /* keep type for statistics */
2092 	code = icmp6->icmp6_code; /* ditto. */
2093 	hlim = 0;
2094 	srcp = NULL;
2095 
2096 	/*
2097 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2098 	 * use dst as the src for the reply.
2099 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2100 	 * (for example) when we encounter an error while forwarding procedure
2101 	 * destined to a duplicated address of ours.
2102 	 */
2103 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
2104 		ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */);
2105 		if (ia != NULL && !(ia->ia6_flags &
2106 		    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2107 			src6 = ia->ia_addr.sin6_addr;
2108 			srcp = &src6;
2109 
2110 			if (m->m_pkthdr.rcvif != NULL) {
2111 				/* XXX: This may not be the outgoing interface */
2112 				hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2113 			} else
2114 				hlim = V_ip6_defhlim;
2115 		}
2116 		if (ia != NULL)
2117 			ifa_free(&ia->ia_ifa);
2118 	}
2119 
2120 	if (srcp == NULL) {
2121 		int error;
2122 		struct in6_addr dst6;
2123 		uint32_t scopeid;
2124 
2125 		/*
2126 		 * This case matches to multicasts, our anycast, or unicasts
2127 		 * that we do not own.  Select a source address based on the
2128 		 * source address of the erroneous packet.
2129 		 */
2130 		in6_splitscope(&ip6->ip6_src, &dst6, &scopeid);
2131 		error = in6_selectsrc_addr(M_GETFIB(m), &dst6,
2132 		    scopeid, NULL, &src6, &hlim);
2133 
2134 		if (error) {
2135 			char ip6buf[INET6_ADDRSTRLEN];
2136 			nd6log((LOG_DEBUG,
2137 			    "icmp6_reflect: source can't be determined: "
2138 			    "dst=%s, error=%d\n",
2139 			    ip6_sprintf(ip6buf, &ip6->ip6_dst), error));
2140 			goto bad;
2141 		}
2142 		srcp = &src6;
2143 	}
2144 	/*
2145 	 * ip6_input() drops a packet if its src is multicast.
2146 	 * So, the src is never multicast.
2147 	 */
2148 	ip6->ip6_dst = ip6->ip6_src;
2149 	ip6->ip6_src = *srcp;
2150 	ip6->ip6_flow = 0;
2151 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2152 	ip6->ip6_vfc |= IPV6_VERSION;
2153 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2154 	ip6->ip6_hlim = hlim;
2155 
2156 	icmp6->icmp6_cksum = 0;
2157 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2158 	    sizeof(struct ip6_hdr), plen);
2159 
2160 	/*
2161 	 * XXX option handling
2162 	 */
2163 
2164 	m->m_flags &= ~(M_BCAST|M_MCAST);
2165 	m->m_pkthdr.rcvif = NULL;
2166 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2167 	if (outif)
2168 		icmp6_ifoutstat_inc(outif, type, code);
2169 
2170 	return;
2171 
2172  bad:
2173 	m_freem(m);
2174 	return;
2175 }
2176 
2177 void
2178 icmp6_fasttimo(void)
2179 {
2180 
2181 	mld_fasttimo();
2182 }
2183 
2184 void
2185 icmp6_slowtimo(void)
2186 {
2187 
2188 	mld_slowtimo();
2189 }
2190 
2191 static const char *
2192 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2193     struct in6_addr *tgt6)
2194 {
2195 	static char buf[1024];
2196 	char ip6bufs[INET6_ADDRSTRLEN];
2197 	char ip6bufd[INET6_ADDRSTRLEN];
2198 	char ip6buft[INET6_ADDRSTRLEN];
2199 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2200 	    ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2201 	    ip6_sprintf(ip6buft, tgt6));
2202 	return buf;
2203 }
2204 
2205 void
2206 icmp6_redirect_input(struct mbuf *m, int off)
2207 {
2208 	struct ifnet *ifp;
2209 	struct ip6_hdr *ip6;
2210 	struct nd_redirect *nd_rd;
2211 	struct in6_addr src6, redtgt6, reddst6;
2212 	union nd_opts ndopts;
2213 	char ip6buf[INET6_ADDRSTRLEN];
2214 	char *lladdr;
2215 	int icmp6len, is_onlink, is_router, lladdrlen;
2216 
2217 	M_ASSERTPKTHDR(m);
2218 	KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2219 
2220 	/* XXX if we are router, we don't update route by icmp6 redirect */
2221 	if (V_ip6_forwarding)
2222 		goto freeit;
2223 	if (!V_icmp6_rediraccept)
2224 		goto freeit;
2225 
2226 	/* RFC 6980: Nodes MUST silently ignore fragments */
2227 	if(m->m_flags & M_FRAGMENTED)
2228 		goto freeit;
2229 
2230 	ip6 = mtod(m, struct ip6_hdr *);
2231 	icmp6len = ntohs(ip6->ip6_plen);
2232 	if (m->m_len < off + icmp6len) {
2233 		m = m_pullup(m, off + icmp6len);
2234 		if (m == NULL) {
2235 			IP6STAT_INC(ip6s_exthdrtoolong);
2236 			return;
2237 		}
2238 	}
2239 	ip6 = mtod(m, struct ip6_hdr *);
2240 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2241 
2242 	ifp = m->m_pkthdr.rcvif;
2243 	redtgt6 = nd_rd->nd_rd_target;
2244 	reddst6 = nd_rd->nd_rd_dst;
2245 
2246 	if (in6_setscope(&redtgt6, ifp, NULL) ||
2247 	    in6_setscope(&reddst6, ifp, NULL)) {
2248 		goto freeit;
2249 	}
2250 
2251 	/* validation */
2252 	src6 = ip6->ip6_src;
2253 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2254 		nd6log((LOG_ERR,
2255 		    "ICMP6 redirect sent from %s rejected; "
2256 		    "must be from linklocal\n",
2257 		    ip6_sprintf(ip6buf, &src6)));
2258 		goto bad;
2259 	}
2260 	if (__predict_false(ip6->ip6_hlim != 255)) {
2261 		ICMP6STAT_INC(icp6s_invlhlim);
2262 		nd6log((LOG_ERR,
2263 		    "ICMP6 redirect sent from %s rejected; "
2264 		    "hlim=%d (must be 255)\n",
2265 		    ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2266 		goto bad;
2267 	}
2268     {
2269 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2270 	struct nhop_object *nh;
2271 	struct in6_addr kdst;
2272 	uint32_t scopeid;
2273 
2274 	in6_splitscope(&reddst6, &kdst, &scopeid);
2275 	NET_EPOCH_ASSERT();
2276 	nh = fib6_lookup(ifp->if_fib, &kdst, scopeid, 0, 0);
2277 	if (nh != NULL) {
2278 		struct in6_addr nh_addr;
2279 		nh_addr = ifatoia6(nh->nh_ifa)->ia_addr.sin6_addr;
2280 		if ((nh->nh_flags & NHF_GATEWAY) == 0) {
2281 			nd6log((LOG_ERR,
2282 			    "ICMP6 redirect rejected; no route "
2283 			    "with inet6 gateway found for redirect dst: %s\n",
2284 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2285 			goto bad;
2286 		}
2287 
2288 		/*
2289 		 * Embed scope zone id into next hop address.
2290 		 */
2291 		nh_addr = nh->gw6_sa.sin6_addr;
2292 
2293 		if (IN6_ARE_ADDR_EQUAL(&src6, &nh_addr) == 0) {
2294 			nd6log((LOG_ERR,
2295 			    "ICMP6 redirect rejected; "
2296 			    "not equal to gw-for-src=%s (must be same): "
2297 			    "%s\n",
2298 			    ip6_sprintf(ip6buf, &nh_addr),
2299 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2300 			goto bad;
2301 		}
2302 	} else {
2303 		nd6log((LOG_ERR,
2304 		    "ICMP6 redirect rejected; "
2305 		    "no route found for redirect dst: %s\n",
2306 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2307 		goto bad;
2308 	}
2309     }
2310 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2311 		nd6log((LOG_ERR,
2312 		    "ICMP6 redirect rejected; "
2313 		    "redirect dst must be unicast: %s\n",
2314 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2315 		goto bad;
2316 	}
2317 
2318 	is_router = is_onlink = 0;
2319 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2320 		is_router = 1;	/* router case */
2321 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2322 		is_onlink = 1;	/* on-link destination case */
2323 	if (!is_router && !is_onlink) {
2324 		nd6log((LOG_ERR,
2325 		    "ICMP6 redirect rejected; "
2326 		    "neither router case nor onlink case: %s\n",
2327 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2328 		goto bad;
2329 	}
2330 
2331 	icmp6len -= sizeof(*nd_rd);
2332 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2333 	if (nd6_options(&ndopts) < 0) {
2334 		nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2335 		    __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2336 		/* nd6_options have incremented stats */
2337 		goto freeit;
2338 	}
2339 
2340 	lladdr = NULL;
2341 	lladdrlen = 0;
2342 	if (ndopts.nd_opts_tgt_lladdr) {
2343 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2344 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2345 	}
2346 
2347 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2348 		nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2349 		    "(if %d, icmp6 packet %d): %s\n",
2350 		    __func__, ip6_sprintf(ip6buf, &redtgt6),
2351 		    ifp->if_addrlen, lladdrlen - 2,
2352 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2353 		goto bad;
2354 	}
2355 
2356 	/* Validation passed. */
2357 
2358 	/* RFC 2461 8.3 */
2359 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2360 	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2361 
2362 	/*
2363 	 * Install a gateway route in the better-router case or an interface
2364 	 * route in the on-link-destination case.
2365 	 */
2366 	{
2367 		struct sockaddr_in6 sdst;
2368 		struct sockaddr_in6 sgw;
2369 		struct sockaddr_in6 ssrc;
2370 		struct sockaddr *gw;
2371 		int rt_flags;
2372 		u_int fibnum;
2373 
2374 		bzero(&sdst, sizeof(sdst));
2375 		bzero(&ssrc, sizeof(ssrc));
2376 		sdst.sin6_family = ssrc.sin6_family = AF_INET6;
2377 		sdst.sin6_len = ssrc.sin6_len = sizeof(struct sockaddr_in6);
2378 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2379 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2380 		rt_flags = 0;
2381 		if (is_router) {
2382 			bzero(&sgw, sizeof(sgw));
2383 			sgw.sin6_family = AF_INET6;
2384 			sgw.sin6_len = sizeof(struct sockaddr_in6);
2385 			bcopy(&redtgt6, &sgw.sin6_addr,
2386 				sizeof(struct in6_addr));
2387 			gw = (struct sockaddr *)&sgw;
2388 			rt_flags |= RTF_GATEWAY;
2389 		} else
2390 			gw = ifp->if_addr->ifa_addr;
2391 		for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2392 			rib_add_redirect(fibnum, (struct sockaddr *)&sdst, gw,
2393 			    (struct sockaddr *)&ssrc, ifp, rt_flags,
2394 			    V_icmp6_redirtimeout);
2395 	}
2396 	/* finally update cached route in each socket via pfctlinput */
2397     {
2398 	struct sockaddr_in6 sdst;
2399 
2400 	bzero(&sdst, sizeof(sdst));
2401 	sdst.sin6_family = AF_INET6;
2402 	sdst.sin6_len = sizeof(struct sockaddr_in6);
2403 	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2404 	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2405     }
2406 
2407  freeit:
2408 	m_freem(m);
2409 	return;
2410 
2411  bad:
2412 	ICMP6STAT_INC(icp6s_badredirect);
2413 	m_freem(m);
2414 }
2415 
2416 void
2417 icmp6_redirect_output(struct mbuf *m0, struct nhop_object *nh)
2418 {
2419 	struct ifnet *ifp;	/* my outgoing interface */
2420 	struct in6_addr *ifp_ll6;
2421 	struct in6_addr *router_ll6;
2422 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2423 	struct mbuf *m = NULL;	/* newly allocated one */
2424 	struct m_tag *mtag;
2425 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2426 	struct nd_redirect *nd_rd;
2427 	struct llentry *ln = NULL;
2428 	size_t maxlen;
2429 	u_char *p;
2430 	struct ifnet *outif = NULL;
2431 	struct sockaddr_in6 src_sa;
2432 
2433 	icmp6_errcount(ND_REDIRECT, 0);
2434 
2435 	/* if we are not router, we don't send icmp6 redirect */
2436 	if (!V_ip6_forwarding)
2437 		goto fail;
2438 
2439 	/* sanity check */
2440 	if (!m0 || !nh || !(NH_IS_VALID(nh)) || !(ifp = nh->nh_ifp))
2441 		goto fail;
2442 
2443 	/*
2444 	 * Address check:
2445 	 *  the source address must identify a neighbor, and
2446 	 *  the destination address must not be a multicast address
2447 	 *  [RFC 2461, sec 8.2]
2448 	 */
2449 	sip6 = mtod(m0, struct ip6_hdr *);
2450 	bzero(&src_sa, sizeof(src_sa));
2451 	src_sa.sin6_family = AF_INET6;
2452 	src_sa.sin6_len = sizeof(src_sa);
2453 	src_sa.sin6_addr = sip6->ip6_src;
2454 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2455 		goto fail;
2456 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2457 		goto fail;	/* what should we do here? */
2458 
2459 	/* rate limit */
2460 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2461 		goto fail;
2462 
2463 	/*
2464 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2465 	 * we almost always ask for an mbuf cluster for simplicity.
2466 	 * (MHLEN < IPV6_MMTU is almost always true)
2467 	 */
2468 #if IPV6_MMTU >= MCLBYTES
2469 # error assumption failed about IPV6_MMTU and MCLBYTES
2470 #endif
2471 	m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2472 	if (m == NULL)
2473 		goto fail;
2474 	M_SETFIB(m, M_GETFIB(m0));
2475 	maxlen = M_TRAILINGSPACE(m);
2476 	maxlen = min(IPV6_MMTU, maxlen);
2477 	/* just for safety */
2478 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2479 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2480 		goto fail;
2481 	}
2482 
2483 	{
2484 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2485 		struct in6_ifaddr *ia;
2486 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2487 						 IN6_IFF_NOTREADY|
2488 						 IN6_IFF_ANYCAST)) == NULL)
2489 			goto fail;
2490 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2491 		/* XXXRW: reference released prematurely. */
2492 		ifa_free(&ia->ia_ifa);
2493 	}
2494 
2495 	/* get ip6 linklocal address for the router. */
2496 	if (nh->nh_flags & NHF_GATEWAY) {
2497 		struct sockaddr_in6 *sin6;
2498 		sin6 = &nh->gw6_sa;
2499 		router_ll6 = &sin6->sin6_addr;
2500 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2501 			router_ll6 = (struct in6_addr *)NULL;
2502 	} else
2503 		router_ll6 = (struct in6_addr *)NULL;
2504 
2505 	/* ip6 */
2506 	ip6 = mtod(m, struct ip6_hdr *);
2507 	ip6->ip6_flow = 0;
2508 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2509 	ip6->ip6_vfc |= IPV6_VERSION;
2510 	/* ip6->ip6_plen will be set later */
2511 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2512 	ip6->ip6_hlim = 255;
2513 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2514 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2515 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2516 
2517 	/* ND Redirect */
2518 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2519 	nd_rd->nd_rd_type = ND_REDIRECT;
2520 	nd_rd->nd_rd_code = 0;
2521 	nd_rd->nd_rd_reserved = 0;
2522 	if (nh->nh_flags & NHF_GATEWAY) {
2523 		/*
2524 		 * nd_rd->nd_rd_target must be a link-local address in
2525 		 * better router cases.
2526 		 */
2527 		if (!router_ll6)
2528 			goto fail;
2529 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2530 		    sizeof(nd_rd->nd_rd_target));
2531 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2532 		    sizeof(nd_rd->nd_rd_dst));
2533 	} else {
2534 		/* make sure redtgt == reddst */
2535 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2536 		    sizeof(nd_rd->nd_rd_target));
2537 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2538 		    sizeof(nd_rd->nd_rd_dst));
2539 	}
2540 
2541 	p = (u_char *)(nd_rd + 1);
2542 
2543 	if (!router_ll6)
2544 		goto nolladdropt;
2545 
2546 	{
2547 		/* target lladdr option */
2548 		int len;
2549 		struct nd_opt_hdr *nd_opt;
2550 		char *lladdr;
2551 
2552 		ln = nd6_lookup(router_ll6, 0, ifp);
2553 		if (ln == NULL)
2554 			goto nolladdropt;
2555 
2556 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2557 		len = (len + 7) & ~7;	/* round by 8 */
2558 		/* safety check */
2559 		if (len + (p - (u_char *)ip6) > maxlen)
2560 			goto nolladdropt;
2561 
2562 		if (ln->la_flags & LLE_VALID) {
2563 			nd_opt = (struct nd_opt_hdr *)p;
2564 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2565 			nd_opt->nd_opt_len = len >> 3;
2566 			lladdr = (char *)(nd_opt + 1);
2567 			bcopy(ln->ll_addr, lladdr, ifp->if_addrlen);
2568 			p += len;
2569 		}
2570 	}
2571 nolladdropt:
2572 	if (ln != NULL)
2573 		LLE_RUNLOCK(ln);
2574 
2575 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2576 
2577 	/* just to be safe */
2578 #ifdef M_DECRYPTED	/*not openbsd*/
2579 	if (m0->m_flags & M_DECRYPTED)
2580 		goto noredhdropt;
2581 #endif
2582 	if (p - (u_char *)ip6 > maxlen)
2583 		goto noredhdropt;
2584 
2585 	{
2586 		/* redirected header option */
2587 		int len;
2588 		struct nd_opt_rd_hdr *nd_opt_rh;
2589 
2590 		/*
2591 		 * compute the maximum size for icmp6 redirect header option.
2592 		 * XXX room for auth header?
2593 		 */
2594 		len = maxlen - (p - (u_char *)ip6);
2595 		len &= ~7;
2596 
2597 		/* This is just for simplicity. */
2598 		if (m0->m_pkthdr.len != m0->m_len) {
2599 			if (m0->m_next) {
2600 				m_freem(m0->m_next);
2601 				m0->m_next = NULL;
2602 			}
2603 			m0->m_pkthdr.len = m0->m_len;
2604 		}
2605 
2606 		/*
2607 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2608 		 * about padding/truncate rule for the original IP packet.
2609 		 * From the discussion on IPv6imp in Feb 1999,
2610 		 * the consensus was:
2611 		 * - "attach as much as possible" is the goal
2612 		 * - pad if not aligned (original size can be guessed by
2613 		 *   original ip6 header)
2614 		 * Following code adds the padding if it is simple enough,
2615 		 * and truncates if not.
2616 		 */
2617 		if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2618 			panic("assumption failed in %s:%d", __FILE__,
2619 			    __LINE__);
2620 
2621 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2622 			/* not enough room, truncate */
2623 			m0->m_pkthdr.len = m0->m_len = len -
2624 			    sizeof(*nd_opt_rh);
2625 		} else {
2626 			/* enough room, pad or truncate */
2627 			size_t extra;
2628 
2629 			extra = m0->m_pkthdr.len % 8;
2630 			if (extra) {
2631 				/* pad if easy enough, truncate if not */
2632 				if (8 - extra <= M_TRAILINGSPACE(m0)) {
2633 					/* pad */
2634 					m0->m_len += (8 - extra);
2635 					m0->m_pkthdr.len += (8 - extra);
2636 				} else {
2637 					/* truncate */
2638 					m0->m_pkthdr.len -= extra;
2639 					m0->m_len -= extra;
2640 				}
2641 			}
2642 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2643 			m0->m_pkthdr.len = m0->m_len = len -
2644 			    sizeof(*nd_opt_rh);
2645 		}
2646 
2647 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2648 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2649 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2650 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2651 		p += sizeof(*nd_opt_rh);
2652 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2653 
2654 		/* connect m0 to m */
2655 		m_tag_delete_chain(m0, NULL);
2656 		m0->m_flags &= ~M_PKTHDR;
2657 		m->m_next = m0;
2658 		m->m_pkthdr.len = m->m_len + m0->m_len;
2659 		m0 = NULL;
2660 	}
2661 noredhdropt:;
2662 	if (m0) {
2663 		m_freem(m0);
2664 		m0 = NULL;
2665 	}
2666 
2667 	/* XXX: clear embedded link IDs in the inner header */
2668 	in6_clearscope(&sip6->ip6_src);
2669 	in6_clearscope(&sip6->ip6_dst);
2670 	in6_clearscope(&nd_rd->nd_rd_target);
2671 	in6_clearscope(&nd_rd->nd_rd_dst);
2672 
2673 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2674 
2675 	nd_rd->nd_rd_cksum = 0;
2676 	nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2677 	    sizeof(*ip6), ntohs(ip6->ip6_plen));
2678 
2679         if (send_sendso_input_hook != NULL) {
2680 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2681 			M_NOWAIT);
2682 		if (mtag == NULL)
2683 			goto fail;
2684 		*(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2685 		m_tag_prepend(m, mtag);
2686 	}
2687 
2688 	/* send the packet to outside... */
2689 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2690 	if (outif) {
2691 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2692 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2693 	}
2694 	ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2695 
2696 	return;
2697 
2698 fail:
2699 	if (m)
2700 		m_freem(m);
2701 	if (m0)
2702 		m_freem(m0);
2703 }
2704 
2705 /*
2706  * ICMPv6 socket option processing.
2707  */
2708 int
2709 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2710 {
2711 	int error = 0;
2712 	int optlen;
2713 	struct inpcb *inp = sotoinpcb(so);
2714 	int level, op, optname;
2715 
2716 	if (sopt) {
2717 		level = sopt->sopt_level;
2718 		op = sopt->sopt_dir;
2719 		optname = sopt->sopt_name;
2720 		optlen = sopt->sopt_valsize;
2721 	} else
2722 		level = op = optname = optlen = 0;
2723 
2724 	if (level != IPPROTO_ICMPV6) {
2725 		return EINVAL;
2726 	}
2727 
2728 	switch (op) {
2729 	case PRCO_SETOPT:
2730 		switch (optname) {
2731 		case ICMP6_FILTER:
2732 		    {
2733 			struct icmp6_filter ic6f;
2734 
2735 			if (optlen != sizeof(ic6f)) {
2736 				error = EMSGSIZE;
2737 				break;
2738 			}
2739 			error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2740 			if (error == 0) {
2741 				INP_WLOCK(inp);
2742 				*inp->in6p_icmp6filt = ic6f;
2743 				INP_WUNLOCK(inp);
2744 			}
2745 			break;
2746 		    }
2747 
2748 		default:
2749 			error = ENOPROTOOPT;
2750 			break;
2751 		}
2752 		break;
2753 
2754 	case PRCO_GETOPT:
2755 		switch (optname) {
2756 		case ICMP6_FILTER:
2757 		    {
2758 			struct icmp6_filter ic6f;
2759 
2760 			INP_RLOCK(inp);
2761 			ic6f = *inp->in6p_icmp6filt;
2762 			INP_RUNLOCK(inp);
2763 			error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2764 			break;
2765 		    }
2766 
2767 		default:
2768 			error = ENOPROTOOPT;
2769 			break;
2770 		}
2771 		break;
2772 	}
2773 
2774 	return (error);
2775 }
2776 
2777 /*
2778  * Perform rate limit check.
2779  * Returns 0 if it is okay to send the icmp6 packet.
2780  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2781  * limitation.
2782  *
2783  * XXX per-destination/type check necessary?
2784  *
2785  * dst - not used at this moment
2786  * type - not used at this moment
2787  * code - not used at this moment
2788  */
2789 static int
2790 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2791     const int code)
2792 {
2793 	int ret;
2794 
2795 	ret = 0;	/* okay to send */
2796 
2797 	/* PPS limit */
2798 	if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2799 	    V_icmp6errppslim)) {
2800 		/* The packet is subject to rate limit */
2801 		ret++;
2802 	}
2803 
2804 	return ret;
2805 }
2806