xref: /freebsd/sys/netinet6/icmp6.c (revision 5956d97f4b3204318ceb6aa9c77bd0bc6ea87a41)
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(int, icmp6errppslim);
128 VNET_DEFINE_STATIC(int, icmp6errpps_count) = 0;
129 VNET_DEFINE_STATIC(struct timeval, icmp6errppslim_last);
130 VNET_DECLARE(int, icmp6_nodeinfo);
131 
132 #define	V_ripcbinfo			VNET(ripcbinfo)
133 #define	V_icmp6errppslim		VNET(icmp6errppslim)
134 #define	V_icmp6errpps_count		VNET(icmp6errpps_count)
135 #define	V_icmp6errppslim_last		VNET(icmp6errppslim_last)
136 #define	V_icmp6_nodeinfo		VNET(icmp6_nodeinfo)
137 
138 static void icmp6_errcount(int, int);
139 static int icmp6_rip6_input(struct mbuf **, int);
140 static void icmp6_reflect(struct mbuf *, size_t);
141 static int icmp6_ratelimit(const struct in6_addr *, const int, const int);
142 static const char *icmp6_redirect_diag(struct in6_addr *,
143 	struct in6_addr *, struct in6_addr *);
144 static struct mbuf *ni6_input(struct mbuf *, int, struct prison *);
145 static struct mbuf *ni6_nametodns(const char *, int, int);
146 static int ni6_dnsmatch(const char *, int, const char *, int);
147 static int ni6_addrs(struct icmp6_nodeinfo *, struct mbuf *,
148 			  struct ifnet **, struct in6_addr *);
149 static int ni6_store_addrs(struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
150 				struct ifnet *, int);
151 static int icmp6_notify_error(struct mbuf **, int, int, int);
152 
153 /*
154  * Kernel module interface for updating icmp6stat.  The argument is an index
155  * into icmp6stat treated as an array of u_quad_t.  While this encodes the
156  * general layout of icmp6stat into the caller, it doesn't encode its
157  * location, so that future changes to add, for example, per-CPU stats
158  * support won't cause binary compatibility problems for kernel modules.
159  */
160 void
161 kmod_icmp6stat_inc(int statnum)
162 {
163 
164 	counter_u64_add(VNET(icmp6stat)[statnum], 1);
165 }
166 
167 static void
168 icmp6_errcount(int type, int code)
169 {
170 	switch (type) {
171 	case ICMP6_DST_UNREACH:
172 		switch (code) {
173 		case ICMP6_DST_UNREACH_NOROUTE:
174 			ICMP6STAT_INC(icp6s_odst_unreach_noroute);
175 			return;
176 		case ICMP6_DST_UNREACH_ADMIN:
177 			ICMP6STAT_INC(icp6s_odst_unreach_admin);
178 			return;
179 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
180 			ICMP6STAT_INC(icp6s_odst_unreach_beyondscope);
181 			return;
182 		case ICMP6_DST_UNREACH_ADDR:
183 			ICMP6STAT_INC(icp6s_odst_unreach_addr);
184 			return;
185 		case ICMP6_DST_UNREACH_NOPORT:
186 			ICMP6STAT_INC(icp6s_odst_unreach_noport);
187 			return;
188 		}
189 		break;
190 	case ICMP6_PACKET_TOO_BIG:
191 		ICMP6STAT_INC(icp6s_opacket_too_big);
192 		return;
193 	case ICMP6_TIME_EXCEEDED:
194 		switch (code) {
195 		case ICMP6_TIME_EXCEED_TRANSIT:
196 			ICMP6STAT_INC(icp6s_otime_exceed_transit);
197 			return;
198 		case ICMP6_TIME_EXCEED_REASSEMBLY:
199 			ICMP6STAT_INC(icp6s_otime_exceed_reassembly);
200 			return;
201 		}
202 		break;
203 	case ICMP6_PARAM_PROB:
204 		switch (code) {
205 		case ICMP6_PARAMPROB_HEADER:
206 			ICMP6STAT_INC(icp6s_oparamprob_header);
207 			return;
208 		case ICMP6_PARAMPROB_NEXTHEADER:
209 			ICMP6STAT_INC(icp6s_oparamprob_nextheader);
210 			return;
211 		case ICMP6_PARAMPROB_OPTION:
212 			ICMP6STAT_INC(icp6s_oparamprob_option);
213 			return;
214 		}
215 		break;
216 	case ND_REDIRECT:
217 		ICMP6STAT_INC(icp6s_oredirect);
218 		return;
219 	}
220 	ICMP6STAT_INC(icp6s_ounknown);
221 }
222 
223 /*
224  * A wrapper function for icmp6_error() necessary when the erroneous packet
225  * may not contain enough scope zone information.
226  */
227 void
228 icmp6_error2(struct mbuf *m, int type, int code, int param,
229     struct ifnet *ifp)
230 {
231 	struct ip6_hdr *ip6;
232 
233 	if (ifp == NULL)
234 		return;
235 
236 	if (m->m_len < sizeof(struct ip6_hdr)) {
237 		m = m_pullup(m, sizeof(struct ip6_hdr));
238 		if (m == NULL) {
239 			IP6STAT_INC(ip6s_exthdrtoolong);
240 			return;
241 		}
242 	}
243 	ip6 = mtod(m, struct ip6_hdr *);
244 
245 	if (in6_setscope(&ip6->ip6_src, ifp, NULL) != 0)
246 		return;
247 	if (in6_setscope(&ip6->ip6_dst, ifp, NULL) != 0)
248 		return;
249 
250 	icmp6_error(m, type, code, param);
251 }
252 
253 /*
254  * Generate an error packet of type error in response to bad IP6 packet.
255  */
256 void
257 icmp6_error(struct mbuf *m, int type, int code, int param)
258 {
259 	struct ip6_hdr *oip6, *nip6;
260 	struct icmp6_hdr *icmp6;
261 	struct epoch_tracker et;
262 	u_int preplen;
263 	int off;
264 	int nxt;
265 
266 	ICMP6STAT_INC(icp6s_error);
267 
268 	/* count per-type-code statistics */
269 	icmp6_errcount(type, code);
270 
271 #ifdef M_DECRYPTED	/*not openbsd*/
272 	if (m->m_flags & M_DECRYPTED) {
273 		ICMP6STAT_INC(icp6s_canterror);
274 		goto freeit;
275 	}
276 #endif
277 
278 	if (m->m_len < sizeof(struct ip6_hdr)) {
279 		m = m_pullup(m, sizeof(struct ip6_hdr));
280 		if (m == NULL) {
281 			IP6STAT_INC(ip6s_exthdrtoolong);
282 			return;
283 		}
284 	}
285 	oip6 = mtod(m, struct ip6_hdr *);
286 
287 	/*
288 	 * If the destination address of the erroneous packet is a multicast
289 	 * address, or the packet was sent using link-layer multicast,
290 	 * we should basically suppress sending an error (RFC 2463, Section
291 	 * 2.4).
292 	 * We have two exceptions (the item e.2 in that section):
293 	 * - the Packet Too Big message can be sent for path MTU discovery.
294 	 * - the Parameter Problem Message that can be allowed an icmp6 error
295 	 *   in the option type field.  This check has been done in
296 	 *   ip6_unknown_opt(), so we can just check the type and code.
297 	 */
298 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
299 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
300 	    (type != ICMP6_PACKET_TOO_BIG &&
301 	     (type != ICMP6_PARAM_PROB ||
302 	      code != ICMP6_PARAMPROB_OPTION)))
303 		goto freeit;
304 
305 	/*
306 	 * RFC 2463, 2.4 (e.5): source address check.
307 	 * XXX: the case of anycast source?
308 	 */
309 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
310 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
311 		goto freeit;
312 
313 	/*
314 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
315 	 * don't do it.
316 	 */
317 	nxt = -1;
318 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
319 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
320 		struct icmp6_hdr *icp;
321 
322 		if (m->m_len < off + sizeof(struct icmp6_hdr)) {
323 			m = m_pullup(m, off + sizeof(struct icmp6_hdr));
324 			if (m == NULL) {
325 				IP6STAT_INC(ip6s_exthdrtoolong);
326 				return;
327 			}
328 		}
329 		oip6 = mtod(m, struct ip6_hdr *);
330 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
331 
332 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
333 		    icp->icmp6_type == ND_REDIRECT) {
334 			/*
335 			 * ICMPv6 error
336 			 * Special case: for redirect (which is
337 			 * informational) we must not send icmp6 error.
338 			 */
339 			ICMP6STAT_INC(icp6s_canterror);
340 			goto freeit;
341 		} else {
342 			/* ICMPv6 informational - send the error */
343 		}
344 	} else {
345 		/* non-ICMPv6 - send the error */
346 	}
347 
348 	/* Finally, do rate limitation check. */
349 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
350 		ICMP6STAT_INC(icp6s_toofreq);
351 		goto freeit;
352 	}
353 
354 	/*
355 	 * OK, ICMP6 can be generated.
356 	 */
357 
358 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
359 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
360 
361 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
362 	M_PREPEND(m, preplen, M_NOWAIT);	/* FIB is also copied over. */
363 	if (m == NULL) {
364 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
365 		return;
366 	}
367 
368 	nip6 = mtod(m, struct ip6_hdr *);
369 	nip6->ip6_src  = oip6->ip6_src;
370 	nip6->ip6_dst  = oip6->ip6_dst;
371 
372 	in6_clearscope(&oip6->ip6_src);
373 	in6_clearscope(&oip6->ip6_dst);
374 
375 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
376 	icmp6->icmp6_type = type;
377 	icmp6->icmp6_code = code;
378 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
379 
380 	ICMP6STAT_INC(icp6s_outhist[type]);
381 	NET_EPOCH_ENTER(et);
382 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
383 	NET_EPOCH_EXIT(et);
384 
385 	return;
386 
387   freeit:
388 	/*
389 	 * If we can't tell whether or not we can generate ICMP6, free it.
390 	 */
391 	m_freem(m);
392 }
393 
394 /*
395  * Process a received ICMP6 message.
396  */
397 int
398 icmp6_input(struct mbuf **mp, int *offp, int proto)
399 {
400 	struct mbuf *m, *n;
401 	struct ifnet *ifp;
402 	struct ip6_hdr *ip6, *nip6;
403 	struct icmp6_hdr *icmp6, *nicmp6;
404 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
405 	int code, error, icmp6len, ip6len, noff, off, sum;
406 
407 	NET_EPOCH_ASSERT();
408 
409 	m = *mp;
410 	off = *offp;
411 
412 	if (m->m_len < off + sizeof(struct icmp6_hdr)) {
413 		m = m_pullup(m, off + sizeof(struct icmp6_hdr));
414 		if (m == NULL) {
415 			IP6STAT_INC(ip6s_exthdrtoolong);
416 			*mp = m;
417 			return (IPPROTO_DONE);
418 		}
419 	}
420 
421 	/*
422 	 * Locate icmp6 structure in mbuf, and check
423 	 * that not corrupted and of at least minimum length
424 	 */
425 
426 	icmp6len = m->m_pkthdr.len - off;
427 	if (icmp6len < sizeof(struct icmp6_hdr)) {
428 		ICMP6STAT_INC(icp6s_tooshort);
429 		goto freeit;
430 	}
431 
432 	ip6 = mtod(m, struct ip6_hdr *);
433 	ifp = m->m_pkthdr.rcvif;
434 	/*
435 	 * Check multicast group membership.
436 	 * Note: SSM filters are not applied for ICMPv6 traffic.
437 	 */
438 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
439 		struct in6_multi	*inm;
440 
441 		inm = in6m_lookup(ifp, &ip6->ip6_dst);
442 		if (inm == NULL) {
443 			IP6STAT_INC(ip6s_notmember);
444 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
445 			goto freeit;
446 		}
447 	}
448 
449 	/* Calculate the checksum. */
450 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
451 	code = icmp6->icmp6_code;
452 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
453 		nd6log((LOG_ERR,
454 		    "ICMP6 checksum error(%d|%x) %s\n",
455 		    icmp6->icmp6_type, sum,
456 		    ip6_sprintf(ip6bufs, &ip6->ip6_src)));
457 		ICMP6STAT_INC(icp6s_checksum);
458 		goto freeit;
459 	}
460 
461 	ICMP6STAT_INC(icp6s_inhist[icmp6->icmp6_type]);
462 	icmp6_ifstat_inc(ifp, ifs6_in_msg);
463 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
464 		icmp6_ifstat_inc(ifp, ifs6_in_error);
465 
466 	ip6len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
467 	switch (icmp6->icmp6_type) {
468 	case ICMP6_DST_UNREACH:
469 		icmp6_ifstat_inc(ifp, ifs6_in_dstunreach);
470 		switch (code) {
471 		case ICMP6_DST_UNREACH_NOROUTE:
472 		case ICMP6_DST_UNREACH_ADDR:	/* PRC_HOSTDEAD is a DOS */
473 			code = PRC_UNREACH_NET;
474 			break;
475 		case ICMP6_DST_UNREACH_ADMIN:
476 			icmp6_ifstat_inc(ifp, ifs6_in_adminprohib);
477 			code = PRC_UNREACH_ADMIN_PROHIB;
478 			break;
479 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
480 			/* I mean "source address was incorrect." */
481 			code = PRC_PARAMPROB;
482 			break;
483 		case ICMP6_DST_UNREACH_NOPORT:
484 			code = PRC_UNREACH_PORT;
485 			break;
486 		default:
487 			goto badcode;
488 		}
489 		goto deliver;
490 		break;
491 
492 	case ICMP6_PACKET_TOO_BIG:
493 		icmp6_ifstat_inc(ifp, ifs6_in_pkttoobig);
494 
495 		/* validation is made in icmp6_mtudisc_update */
496 
497 		code = PRC_MSGSIZE;
498 
499 		/*
500 		 * Updating the path MTU will be done after examining
501 		 * intermediate extension headers.
502 		 */
503 		goto deliver;
504 		break;
505 
506 	case ICMP6_TIME_EXCEEDED:
507 		icmp6_ifstat_inc(ifp, ifs6_in_timeexceed);
508 		switch (code) {
509 		case ICMP6_TIME_EXCEED_TRANSIT:
510 			code = PRC_TIMXCEED_INTRANS;
511 			break;
512 		case ICMP6_TIME_EXCEED_REASSEMBLY:
513 			code = PRC_TIMXCEED_REASS;
514 			break;
515 		default:
516 			goto badcode;
517 		}
518 		goto deliver;
519 		break;
520 
521 	case ICMP6_PARAM_PROB:
522 		icmp6_ifstat_inc(ifp, ifs6_in_paramprob);
523 		switch (code) {
524 		case ICMP6_PARAMPROB_NEXTHEADER:
525 			code = PRC_UNREACH_PROTOCOL;
526 			break;
527 		case ICMP6_PARAMPROB_HEADER:
528 		case ICMP6_PARAMPROB_OPTION:
529 			code = PRC_PARAMPROB;
530 			break;
531 		default:
532 			goto badcode;
533 		}
534 		goto deliver;
535 		break;
536 
537 	case ICMP6_ECHO_REQUEST:
538 		icmp6_ifstat_inc(ifp, ifs6_in_echo);
539 		if (code != 0)
540 			goto badcode;
541 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
542 			/* Give up remote */
543 			break;
544 		}
545 		if (!M_WRITABLE(n)
546 		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
547 			struct mbuf *n0 = n;
548 			int n0len;
549 
550 			CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) <= MHLEN);
551 			n = m_gethdr(M_NOWAIT, n0->m_type);
552 			if (n == NULL) {
553 				/* Give up remote */
554 				m_freem(n0);
555 				break;
556 			}
557 
558 			m_move_pkthdr(n, n0);	/* FIB copied. */
559 			n0len = n0->m_pkthdr.len;	/* save for use below */
560 			/*
561 			 * Copy IPv6 and ICMPv6 only.
562 			 */
563 			nip6 = mtod(n, struct ip6_hdr *);
564 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
565 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
566 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
567 			noff = sizeof(struct ip6_hdr);
568 			/* new mbuf contains only ipv6+icmpv6 headers */
569 			n->m_len = noff + sizeof(struct icmp6_hdr);
570 			/*
571 			 * Adjust mbuf.  ip6_plen will be adjusted in
572 			 * ip6_output().
573 			 */
574 			m_adj(n0, off + sizeof(struct icmp6_hdr));
575 			/* recalculate complete packet size */
576 			n->m_pkthdr.len = n0len + (noff - off);
577 			n->m_next = n0;
578 		} else {
579 			if (n->m_len < off + sizeof(*nicmp6)) {
580 				n = m_pullup(n, off + sizeof(*nicmp6));
581 				if (n == NULL) {
582 					IP6STAT_INC(ip6s_exthdrtoolong);
583 					break;
584 				}
585 			}
586 			nicmp6 = (struct icmp6_hdr *)(mtod(n, caddr_t) + off);
587 			noff = off;
588 		}
589 		if (n) {
590 			nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
591 			nicmp6->icmp6_code = 0;
592 			ICMP6STAT_INC(icp6s_reflect);
593 			ICMP6STAT_INC(icp6s_outhist[ICMP6_ECHO_REPLY]);
594 			icmp6_reflect(n, noff);
595 		}
596 		break;
597 
598 	case ICMP6_ECHO_REPLY:
599 		icmp6_ifstat_inc(ifp, ifs6_in_echoreply);
600 		if (code != 0)
601 			goto badcode;
602 		break;
603 
604 	case MLD_LISTENER_QUERY:
605 	case MLD_LISTENER_REPORT:
606 	case MLD_LISTENER_DONE:
607 	case MLDV2_LISTENER_REPORT:
608 		/*
609 		 * Drop MLD traffic which is not link-local, has a hop limit
610 		 * of greater than 1 hop, or which does not have the
611 		 * IPv6 HBH Router Alert option.
612 		 * As IPv6 HBH options are stripped in ip6_input() we must
613 		 * check an mbuf header flag.
614 		 * XXX Should we also sanity check that these messages
615 		 * were directed to a link-local multicast prefix?
616 		 */
617 		if ((ip6->ip6_hlim != 1) || (m->m_flags & M_RTALERT_MLD) == 0)
618 			goto freeit;
619 		if (mld_input(&m, off, icmp6len) != 0) {
620 			*mp = NULL;
621 			return (IPPROTO_DONE);
622 		}
623 		/* m stays. */
624 		break;
625 
626 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
627 	    {
628 		enum { WRU, FQDN } mode;
629 		struct prison *pr;
630 
631 		if (!V_icmp6_nodeinfo)
632 			break;
633 
634 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
635 			mode = WRU;
636 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
637 			mode = FQDN;
638 		else
639 			goto badlen;
640 
641 		pr = NULL;
642 		sx_slock(&allprison_lock);
643 		TAILQ_FOREACH(pr, &allprison, pr_list)
644 			if (pr->pr_vnet == ifp->if_vnet)
645 				break;
646 		sx_sunlock(&allprison_lock);
647 		if (pr == NULL)
648 			pr = curthread->td_ucred->cr_prison;
649 		if (mode == FQDN) {
650 			if (m->m_len < off + sizeof(struct icmp6_nodeinfo)) {
651 				m = m_pullup(m, off +
652 				    sizeof(struct icmp6_nodeinfo));
653 				if (m == NULL) {
654 					IP6STAT_INC(ip6s_exthdrtoolong);
655 					*mp = m;
656 					return (IPPROTO_DONE);
657 				}
658 			}
659 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
660 			if (n)
661 				n = ni6_input(n, off, pr);
662 			/* XXX meaningless if n == NULL */
663 			noff = sizeof(struct ip6_hdr);
664 		} else {
665 			u_char *p;
666 			int maxhlen, hlen;
667 
668 			/*
669 			 * XXX: this combination of flags is pointless,
670 			 * but should we keep this for compatibility?
671 			 */
672 			if ((V_icmp6_nodeinfo & (ICMP6_NODEINFO_FQDNOK |
673 			    ICMP6_NODEINFO_TMPADDROK)) !=
674 			    (ICMP6_NODEINFO_FQDNOK | ICMP6_NODEINFO_TMPADDROK))
675 				break;
676 
677 			if (code != 0)
678 				goto badcode;
679 
680 			CTASSERT(sizeof(*nip6) + sizeof(*nicmp6) + 4 <= MHLEN);
681 			n = m_gethdr(M_NOWAIT, m->m_type);
682 			if (n == NULL) {
683 				/* Give up remote */
684 				break;
685 			}
686 			if (!m_dup_pkthdr(n, m, M_NOWAIT)) {
687 				/*
688 				 * Previous code did a blind M_COPY_PKTHDR
689 				 * and said "just for rcvif".  If true, then
690 				 * we could tolerate the dup failing (due to
691 				 * the deep copy of the tag chain).  For now
692 				 * be conservative and just fail.
693 				 */
694 				m_free(n);
695 				n = NULL;
696 				break;
697 			}
698 			/*
699 			 * Copy IPv6 and ICMPv6 only.
700 			 */
701 			nip6 = mtod(n, struct ip6_hdr *);
702 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
703 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
704 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
705 			p = (u_char *)(nicmp6 + 1);
706 			bzero(p, 4);
707 
708 			maxhlen = M_TRAILINGSPACE(n) -
709 			    (sizeof(*nip6) + sizeof(*nicmp6) + 4);
710 			mtx_lock(&pr->pr_mtx);
711 			hlen = strlen(pr->pr_hostname);
712 			if (maxhlen > hlen)
713 				maxhlen = hlen;
714 			/* meaningless TTL */
715 			bcopy(pr->pr_hostname, p + 4, maxhlen);
716 			mtx_unlock(&pr->pr_mtx);
717 			noff = sizeof(struct ip6_hdr);
718 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
719 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
720 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
721 			nicmp6->icmp6_code = 0;
722 		}
723 		if (n) {
724 			ICMP6STAT_INC(icp6s_reflect);
725 			ICMP6STAT_INC(icp6s_outhist[ICMP6_WRUREPLY]);
726 			icmp6_reflect(n, noff);
727 		}
728 		break;
729 	    }
730 
731 	case ICMP6_WRUREPLY:
732 		if (code != 0)
733 			goto badcode;
734 		break;
735 
736 	case ND_ROUTER_SOLICIT:
737 		icmp6_ifstat_inc(ifp, ifs6_in_routersolicit);
738 		if (code != 0)
739 			goto badcode;
740 		if (icmp6len < sizeof(struct nd_router_solicit))
741 			goto badlen;
742 		if (send_sendso_input_hook != NULL) {
743 			if (m->m_len < off + icmp6len) {
744 				m = m_pullup(m, off + icmp6len);
745 				if (m == NULL) {
746 					IP6STAT_INC(ip6s_exthdrtoolong);
747 					*mp = NULL;
748 					return (IPPROTO_DONE);
749 				}
750 			}
751 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
752 			if (error == 0) {
753 				m = NULL;
754 				goto freeit;
755 			}
756 		}
757 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
758 		nd6_rs_input(m, off, icmp6len);
759 		m = n;
760 		if (m == NULL)
761 			goto freeit;
762 		break;
763 
764 	case ND_ROUTER_ADVERT:
765 		icmp6_ifstat_inc(ifp, ifs6_in_routeradvert);
766 		if (code != 0)
767 			goto badcode;
768 		if (icmp6len < sizeof(struct nd_router_advert))
769 			goto badlen;
770 		if (send_sendso_input_hook != NULL) {
771 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
772 			if (error == 0) {
773 				m = NULL;
774 				goto freeit;
775 			}
776 		}
777 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
778 		nd6_ra_input(m, off, icmp6len);
779 		m = n;
780 		if (m == NULL)
781 			goto freeit;
782 		break;
783 
784 	case ND_NEIGHBOR_SOLICIT:
785 		icmp6_ifstat_inc(ifp, ifs6_in_neighborsolicit);
786 		if (code != 0)
787 			goto badcode;
788 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
789 			goto badlen;
790 		if (send_sendso_input_hook != NULL) {
791 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
792 			if (error == 0) {
793 				m = NULL;
794 				goto freeit;
795 			}
796 		}
797 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
798 		nd6_ns_input(m, off, icmp6len);
799 		m = n;
800 		if (m == NULL)
801 			goto freeit;
802 		break;
803 
804 	case ND_NEIGHBOR_ADVERT:
805 		icmp6_ifstat_inc(ifp, ifs6_in_neighboradvert);
806 		if (code != 0)
807 			goto badcode;
808 		if (icmp6len < sizeof(struct nd_neighbor_advert))
809 			goto badlen;
810 		if (send_sendso_input_hook != NULL) {
811 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
812 			if (error == 0) {
813 				m = NULL;
814 				goto freeit;
815 			}
816 		}
817 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
818 		nd6_na_input(m, off, icmp6len);
819 		m = n;
820 		if (m == NULL)
821 			goto freeit;
822 		break;
823 
824 	case ND_REDIRECT:
825 		icmp6_ifstat_inc(ifp, ifs6_in_redirect);
826 		if (code != 0)
827 			goto badcode;
828 		if (icmp6len < sizeof(struct nd_redirect))
829 			goto badlen;
830 		if (send_sendso_input_hook != NULL) {
831 			error = send_sendso_input_hook(m, ifp, SND_IN, ip6len);
832 			if (error == 0) {
833 				m = NULL;
834 				goto freeit;
835 			}
836 		}
837 		n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
838 		icmp6_redirect_input(m, off);
839 		m = n;
840 		if (m == NULL)
841 			goto freeit;
842 		break;
843 
844 	case ICMP6_ROUTER_RENUMBERING:
845 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
846 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
847 			goto badcode;
848 		if (icmp6len < sizeof(struct icmp6_router_renum))
849 			goto badlen;
850 		break;
851 
852 	default:
853 		nd6log((LOG_DEBUG,
854 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
855 		    icmp6->icmp6_type, ip6_sprintf(ip6bufs, &ip6->ip6_src),
856 		    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
857 		    ifp ? ifp->if_index : 0));
858 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
859 			/* ICMPv6 error: MUST deliver it by spec... */
860 			code = PRC_NCMDS;
861 			/* deliver */
862 		} else {
863 			/* ICMPv6 informational: MUST not deliver */
864 			break;
865 		}
866 	deliver:
867 		if (icmp6_notify_error(&m, off, icmp6len, code) != 0) {
868 			/* In this case, m should've been freed. */
869 			*mp = NULL;
870 			return (IPPROTO_DONE);
871 		}
872 		break;
873 
874 	badcode:
875 		ICMP6STAT_INC(icp6s_badcode);
876 		break;
877 
878 	badlen:
879 		ICMP6STAT_INC(icp6s_badlen);
880 		break;
881 	}
882 
883 	/* deliver the packet to appropriate sockets */
884 	icmp6_rip6_input(&m, *offp);
885 
886 	*mp = m;
887 	return (IPPROTO_DONE);
888 
889  freeit:
890 	m_freem(m);
891 	*mp = NULL;
892 	return (IPPROTO_DONE);
893 }
894 
895 static int
896 icmp6_notify_error(struct mbuf **mp, int off, int icmp6len, int code)
897 {
898 	struct mbuf *m;
899 	struct icmp6_hdr *icmp6;
900 	struct ip6_hdr *eip6;
901 	u_int32_t notifymtu;
902 	struct sockaddr_in6 icmp6src, icmp6dst;
903 
904 	m = *mp;
905 
906 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
907 		ICMP6STAT_INC(icp6s_tooshort);
908 		goto freeit;
909 	}
910 
911 	if (m->m_len < off + sizeof(*icmp6) + sizeof(struct ip6_hdr)) {
912 		m = m_pullup(m, off + sizeof(*icmp6) + sizeof(struct ip6_hdr));
913 		if (m == NULL) {
914 			IP6STAT_INC(ip6s_exthdrtoolong);
915 			*mp = m;
916 			return (-1);
917 		}
918 	}
919 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
920 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
921 	bzero(&icmp6dst, sizeof(icmp6dst));
922 
923 	/* Detect the upper level protocol */
924 	{
925 		void (*ctlfunc)(int, struct sockaddr *, void *);
926 		u_int8_t nxt = eip6->ip6_nxt;
927 		int eoff = off + sizeof(struct icmp6_hdr) +
928 		    sizeof(struct ip6_hdr);
929 		struct ip6ctlparam ip6cp;
930 		int icmp6type = icmp6->icmp6_type;
931 		struct ip6_frag *fh;
932 		struct ip6_rthdr *rth;
933 		struct ip6_rthdr0 *rth0;
934 		int rthlen;
935 
936 		while (1) { /* XXX: should avoid infinite loop explicitly? */
937 			struct ip6_ext *eh;
938 
939 			switch (nxt) {
940 			case IPPROTO_HOPOPTS:
941 			case IPPROTO_DSTOPTS:
942 			case IPPROTO_AH:
943 				if (m->m_len < eoff + sizeof(struct ip6_ext)) {
944 					m = m_pullup(m, eoff +
945 					    sizeof(struct ip6_ext));
946 					if (m == NULL) {
947 						IP6STAT_INC(ip6s_exthdrtoolong);
948 						*mp = m;
949 						return (-1);
950 					}
951 				}
952 				eh = (struct ip6_ext *)
953 				    (mtod(m, caddr_t) + eoff);
954 				if (nxt == IPPROTO_AH)
955 					eoff += (eh->ip6e_len + 2) << 2;
956 				else
957 					eoff += (eh->ip6e_len + 1) << 3;
958 				nxt = eh->ip6e_nxt;
959 				break;
960 			case IPPROTO_ROUTING:
961 				/*
962 				 * When the erroneous packet contains a
963 				 * routing header, we should examine the
964 				 * header to determine the final destination.
965 				 * Otherwise, we can't properly update
966 				 * information that depends on the final
967 				 * destination (e.g. path MTU).
968 				 */
969 				if (m->m_len < eoff + sizeof(*rth)) {
970 					m = m_pullup(m, eoff + sizeof(*rth));
971 					if (m == NULL) {
972 						IP6STAT_INC(ip6s_exthdrtoolong);
973 						*mp = m;
974 						return (-1);
975 					}
976 				}
977 				rth = (struct ip6_rthdr *)
978 				    (mtod(m, caddr_t) + eoff);
979 				rthlen = (rth->ip6r_len + 1) << 3;
980 				/*
981 				 * XXX: currently there is no
982 				 * officially defined type other
983 				 * than type-0.
984 				 * Note that if the segment left field
985 				 * is 0, all intermediate hops must
986 				 * have been passed.
987 				 */
988 				if (rth->ip6r_segleft &&
989 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
990 					int hops;
991 
992 					if (m->m_len < eoff + rthlen) {
993 						m = m_pullup(m, eoff + rthlen);
994 						if (m == NULL) {
995 							IP6STAT_INC(
996 							    ip6s_exthdrtoolong);
997 							*mp = m;
998 							return (-1);
999 						}
1000 					}
1001 					rth0 = (struct ip6_rthdr0 *)
1002 					    (mtod(m, caddr_t) + eoff);
1003 
1004 					/* just ignore a bogus header */
1005 					if ((rth0->ip6r0_len % 2) == 0 &&
1006 					    (hops = rth0->ip6r0_len/2))
1007 						icmp6dst.sin6_addr = *((struct in6_addr *)(rth0 + 1) + (hops - 1));
1008 				}
1009 				eoff += rthlen;
1010 				nxt = rth->ip6r_nxt;
1011 				break;
1012 			case IPPROTO_FRAGMENT:
1013 				if (m->m_len < eoff + sizeof(struct ip6_frag)) {
1014 					m = m_pullup(m, eoff +
1015 					    sizeof(struct ip6_frag));
1016 					if (m == NULL) {
1017 						IP6STAT_INC(ip6s_exthdrtoolong);
1018 						*mp = m;
1019 						return (-1);
1020 					}
1021 				}
1022 				fh = (struct ip6_frag *)(mtod(m, caddr_t) +
1023 				    eoff);
1024 				/*
1025 				 * Data after a fragment header is meaningless
1026 				 * unless it is the first fragment, but
1027 				 * we'll go to the notify label for path MTU
1028 				 * discovery.
1029 				 */
1030 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
1031 					goto notify;
1032 
1033 				eoff += sizeof(struct ip6_frag);
1034 				nxt = fh->ip6f_nxt;
1035 				break;
1036 			default:
1037 				/*
1038 				 * This case includes ESP and the No Next
1039 				 * Header.  In such cases going to the notify
1040 				 * label does not have any meaning
1041 				 * (i.e. ctlfunc will be NULL), but we go
1042 				 * anyway since we might have to update
1043 				 * path MTU information.
1044 				 */
1045 				goto notify;
1046 			}
1047 		}
1048 	  notify:
1049 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
1050 
1051 		/*
1052 		 * retrieve parameters from the inner IPv6 header, and convert
1053 		 * them into sockaddr structures.
1054 		 * XXX: there is no guarantee that the source or destination
1055 		 * addresses of the inner packet are in the same scope as
1056 		 * the addresses of the icmp packet.  But there is no other
1057 		 * way to determine the zone.
1058 		 */
1059 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
1060 
1061 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
1062 		icmp6dst.sin6_family = AF_INET6;
1063 		if (IN6_IS_ADDR_UNSPECIFIED(&icmp6dst.sin6_addr))
1064 			icmp6dst.sin6_addr = eip6->ip6_dst;
1065 		if (in6_setscope(&icmp6dst.sin6_addr, m->m_pkthdr.rcvif, NULL))
1066 			goto freeit;
1067 		bzero(&icmp6src, sizeof(icmp6src));
1068 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1069 		icmp6src.sin6_family = AF_INET6;
1070 		icmp6src.sin6_addr = eip6->ip6_src;
1071 		if (in6_setscope(&icmp6src.sin6_addr, m->m_pkthdr.rcvif, NULL))
1072 			goto freeit;
1073 		icmp6src.sin6_flowinfo =
1074 		    (eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1075 
1076 		ip6cp.ip6c_m = m;
1077 		ip6cp.ip6c_icmp6 = icmp6;
1078 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1079 		ip6cp.ip6c_off = eoff;
1080 		ip6cp.ip6c_finaldst = &icmp6dst.sin6_addr;
1081 		ip6cp.ip6c_src = &icmp6src;
1082 		ip6cp.ip6c_nxt = nxt;
1083 
1084 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1085 			notifymtu = ntohl(icmp6->icmp6_mtu);
1086 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1087 			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1088 		}
1089 
1090 		ctlfunc = (void (*)(int, struct sockaddr *, void *))
1091 		    (inet6sw[ip6_protox[nxt]].pr_ctlinput);
1092 		if (ctlfunc) {
1093 			(void) (*ctlfunc)(code, (struct sockaddr *)&icmp6dst,
1094 			    &ip6cp);
1095 		}
1096 	}
1097 	*mp = m;
1098 	return (0);
1099 
1100   freeit:
1101 	m_freem(m);
1102 	*mp = NULL;
1103 	return (-1);
1104 }
1105 
1106 void
1107 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1108 {
1109 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1110 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1111 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1112 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1113 	struct in_conninfo inc;
1114 	uint32_t max_mtu;
1115 
1116 #if 0
1117 	/*
1118 	 * RFC2460 section 5, last paragraph.
1119 	 * even though minimum link MTU for IPv6 is IPV6_MMTU,
1120 	 * we may see ICMPv6 too big with mtu < IPV6_MMTU
1121 	 * due to packet translator in the middle.
1122 	 * see ip6_output() and ip6_getpmtu() "alwaysfrag" case for
1123 	 * special handling.
1124 	 */
1125 	if (mtu < IPV6_MMTU)
1126 		return;
1127 #endif
1128 
1129 	/*
1130 	 * we reject ICMPv6 too big with abnormally small value.
1131 	 * XXX what is the good definition of "abnormally small"?
1132 	 */
1133 	if (mtu < sizeof(struct ip6_hdr) + sizeof(struct ip6_frag) + 8)
1134 		return;
1135 
1136 	if (!validated)
1137 		return;
1138 
1139 	/*
1140 	 * In case the suggested mtu is less than IPV6_MMTU, we
1141 	 * only need to remember that it was for above mentioned
1142 	 * "alwaysfrag" case.
1143 	 * Try to be as close to the spec as possible.
1144 	 */
1145 	if (mtu < IPV6_MMTU)
1146 		mtu = IPV6_MMTU - 8;
1147 
1148 	bzero(&inc, sizeof(inc));
1149 	inc.inc_fibnum = M_GETFIB(m);
1150 	inc.inc_flags |= INC_ISIPV6;
1151 	inc.inc6_faddr = *dst;
1152 	if (in6_setscope(&inc.inc6_faddr, m->m_pkthdr.rcvif, NULL))
1153 		return;
1154 
1155 	max_mtu = tcp_hc_getmtu(&inc);
1156 	if (max_mtu == 0)
1157 		max_mtu = tcp_maxmtu6(&inc, NULL);
1158 
1159 	if (mtu < max_mtu) {
1160 		tcp_hc_updatemtu(&inc, mtu);
1161 		ICMP6STAT_INC(icp6s_pmtuchg);
1162 	}
1163 }
1164 
1165 /*
1166  * Process a Node Information Query packet, based on
1167  * draft-ietf-ipngwg-icmp-name-lookups-07.
1168  *
1169  * Spec incompatibilities:
1170  * - IPv6 Subject address handling
1171  * - IPv4 Subject address handling support missing
1172  * - Proxy reply (answer even if it's not for me)
1173  * - joins NI group address at in6_ifattach() time only, does not cope
1174  *   with hostname changes by sethostname(3)
1175  */
1176 static struct mbuf *
1177 ni6_input(struct mbuf *m, int off, struct prison *pr)
1178 {
1179 	struct icmp6_nodeinfo *ni6, *nni6;
1180 	struct mbuf *n = NULL;
1181 	u_int16_t qtype;
1182 	int subjlen;
1183 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1184 	struct ni_reply_fqdn *fqdn;
1185 	int addrs;		/* for NI_QTYPE_NODEADDR */
1186 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1187 	struct in6_addr in6_subj; /* subject address */
1188 	struct ip6_hdr *ip6;
1189 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1190 	char *subj = NULL;
1191 	struct in6_ifaddr *ia6 = NULL;
1192 
1193 	ip6 = mtod(m, struct ip6_hdr *);
1194 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1195 
1196 	/*
1197 	 * Validate IPv6 source address.
1198 	 * The default configuration MUST be to refuse answering queries from
1199 	 * global-scope addresses according to RFC4602.
1200 	 * Notes:
1201 	 *  - it's not very clear what "refuse" means; this implementation
1202 	 *    simply drops it.
1203 	 *  - it's not very easy to identify global-scope (unicast) addresses
1204 	 *    since there are many prefixes for them.  It should be safer
1205 	 *    and in practice sufficient to check "all" but loopback and
1206 	 *    link-local (note that site-local unicast was deprecated and
1207 	 *    ULA is defined as global scope-wise)
1208 	 */
1209 	if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_GLOBALOK) == 0 &&
1210 	    !IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) &&
1211 	    !IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
1212 		goto bad;
1213 
1214 	/*
1215 	 * Validate IPv6 destination address.
1216 	 *
1217 	 * The Responder must discard the Query without further processing
1218 	 * unless it is one of the Responder's unicast or anycast addresses, or
1219 	 * a link-local scope multicast address which the Responder has joined.
1220 	 * [RFC4602, Section 5.]
1221 	 */
1222 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
1223 		if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1224 			goto bad;
1225 		/* else it's a link-local multicast, fine */
1226 	} else {		/* unicast or anycast */
1227 		ia6 = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false);
1228 		if (ia6 == NULL)
1229 			goto bad; /* XXX impossible */
1230 
1231 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1232 		    !(V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK)) {
1233 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1234 				"a temporary address in %s:%d",
1235 			       __FILE__, __LINE__));
1236 			goto bad;
1237 		}
1238 	}
1239 
1240 	/* validate query Subject field. */
1241 	qtype = ntohs(ni6->ni_qtype);
1242 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1243 	switch (qtype) {
1244 	case NI_QTYPE_NOOP:
1245 	case NI_QTYPE_SUPTYPES:
1246 		/* 07 draft */
1247 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1248 			break;
1249 		/* FALLTHROUGH */
1250 	case NI_QTYPE_FQDN:
1251 	case NI_QTYPE_NODEADDR:
1252 	case NI_QTYPE_IPV4ADDR:
1253 		switch (ni6->ni_code) {
1254 		case ICMP6_NI_SUBJ_IPV6:
1255 #if ICMP6_NI_SUBJ_IPV6 != 0
1256 		case 0:
1257 #endif
1258 			/*
1259 			 * backward compatibility - try to accept 03 draft
1260 			 * format, where no Subject is present.
1261 			 */
1262 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1263 			    subjlen == 0) {
1264 				oldfqdn++;
1265 				break;
1266 			}
1267 #if ICMP6_NI_SUBJ_IPV6 != 0
1268 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1269 				goto bad;
1270 #endif
1271 
1272 			if (subjlen != sizeof(struct in6_addr))
1273 				goto bad;
1274 
1275 			/*
1276 			 * Validate Subject address.
1277 			 *
1278 			 * Not sure what exactly "address belongs to the node"
1279 			 * means in the spec, is it just unicast, or what?
1280 			 *
1281 			 * At this moment we consider Subject address as
1282 			 * "belong to the node" if the Subject address equals
1283 			 * to the IPv6 destination address; validation for
1284 			 * IPv6 destination address should have done enough
1285 			 * check for us.
1286 			 *
1287 			 * We do not do proxy at this moment.
1288 			 */
1289 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1290 			    subjlen, (caddr_t)&in6_subj);
1291 			if (in6_setscope(&in6_subj, m->m_pkthdr.rcvif, NULL))
1292 				goto bad;
1293 
1294 			subj = (char *)&in6_subj;
1295 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &in6_subj))
1296 				break;
1297 
1298 			/*
1299 			 * XXX if we are to allow other cases, we should really
1300 			 * be careful about scope here.
1301 			 * basically, we should disallow queries toward IPv6
1302 			 * destination X with subject Y,
1303 			 * if scope(X) > scope(Y).
1304 			 * if we allow scope(X) > scope(Y), it will result in
1305 			 * information leakage across scope boundary.
1306 			 */
1307 			goto bad;
1308 
1309 		case ICMP6_NI_SUBJ_FQDN:
1310 			/*
1311 			 * Validate Subject name with gethostname(3).
1312 			 *
1313 			 * The behavior may need some debate, since:
1314 			 * - we are not sure if the node has FQDN as
1315 			 *   hostname (returned by gethostname(3)).
1316 			 * - the code does wildcard match for truncated names.
1317 			 *   however, we are not sure if we want to perform
1318 			 *   wildcard match, if gethostname(3) side has
1319 			 *   truncated hostname.
1320 			 */
1321 			mtx_lock(&pr->pr_mtx);
1322 			n = ni6_nametodns(pr->pr_hostname,
1323 			    strlen(pr->pr_hostname), 0);
1324 			mtx_unlock(&pr->pr_mtx);
1325 			if (!n || n->m_next || n->m_len == 0)
1326 				goto bad;
1327 			if (m->m_len < off + sizeof(struct icmp6_nodeinfo) +
1328 			    subjlen) {
1329 				m = m_pullup(m, off +
1330 				    sizeof(struct icmp6_nodeinfo) + subjlen);
1331 				if (m == NULL) {
1332 					IP6STAT_INC(ip6s_exthdrtoolong);
1333 					goto bad;
1334 				}
1335 			}
1336 			/* ip6 possibly invalid but not used after. */
1337 			ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1338 			subj = (char *)(mtod(m, caddr_t) + off +
1339 			    sizeof(struct icmp6_nodeinfo));
1340 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1341 			    n->m_len)) {
1342 				goto bad;
1343 			}
1344 			m_freem(n);
1345 			n = NULL;
1346 			break;
1347 
1348 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1349 		default:
1350 			goto bad;
1351 		}
1352 		break;
1353 	}
1354 
1355 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1356 	switch (qtype) {
1357 	case NI_QTYPE_FQDN:
1358 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_FQDNOK) == 0)
1359 			goto bad;
1360 		break;
1361 	case NI_QTYPE_NODEADDR:
1362 	case NI_QTYPE_IPV4ADDR:
1363 		if ((V_icmp6_nodeinfo & ICMP6_NODEINFO_NODEADDROK) == 0)
1364 			goto bad;
1365 		break;
1366 	}
1367 
1368 	/* guess reply length */
1369 	switch (qtype) {
1370 	case NI_QTYPE_NOOP:
1371 		break;		/* no reply data */
1372 	case NI_QTYPE_SUPTYPES:
1373 		replylen += sizeof(u_int32_t);
1374 		break;
1375 	case NI_QTYPE_FQDN:
1376 		/* XXX will append an mbuf */
1377 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1378 		break;
1379 	case NI_QTYPE_NODEADDR:
1380 		addrs = ni6_addrs(ni6, m, &ifp, (struct in6_addr *)subj);
1381 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1382 		    sizeof(u_int32_t))) > MCLBYTES)
1383 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1384 		break;
1385 	case NI_QTYPE_IPV4ADDR:
1386 		/* unsupported - should respond with unknown Qtype? */
1387 		break;
1388 	default:
1389 		/*
1390 		 * XXX: We must return a reply with the ICMP6 code
1391 		 * `unknown Qtype' in this case.  However we regard the case
1392 		 * as an FQDN query for backward compatibility.
1393 		 * Older versions set a random value to this field,
1394 		 * so it rarely varies in the defined qtypes.
1395 		 * But the mechanism is not reliable...
1396 		 * maybe we should obsolete older versions.
1397 		 */
1398 		qtype = NI_QTYPE_FQDN;
1399 		/* XXX will append an mbuf */
1400 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1401 		oldfqdn++;
1402 		break;
1403 	}
1404 
1405 	/* Allocate an mbuf to reply. */
1406 	if (replylen > MCLBYTES) {
1407 		/*
1408 		 * XXX: should we try to allocate more? But MCLBYTES
1409 		 * is probably much larger than IPV6_MMTU...
1410 		 */
1411 		goto bad;
1412 	}
1413 	if (replylen > MHLEN)
1414 		n = m_getcl(M_NOWAIT, m->m_type, M_PKTHDR);
1415 	else
1416 		n = m_gethdr(M_NOWAIT, m->m_type);
1417 	if (n == NULL) {
1418 		m_freem(m);
1419 		return (NULL);
1420 	}
1421 	m_move_pkthdr(n, m); /* just for recvif and FIB */
1422 	n->m_pkthdr.len = n->m_len = replylen;
1423 
1424 	/* copy mbuf header and IPv6 + Node Information base headers */
1425 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1426 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1427 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1428 
1429 	/* qtype dependent procedure */
1430 	switch (qtype) {
1431 	case NI_QTYPE_NOOP:
1432 		nni6->ni_code = ICMP6_NI_SUCCESS;
1433 		nni6->ni_flags = 0;
1434 		break;
1435 	case NI_QTYPE_SUPTYPES:
1436 	{
1437 		u_int32_t v;
1438 		nni6->ni_code = ICMP6_NI_SUCCESS;
1439 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1440 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1441 		v = (u_int32_t)htonl(0x0000000f);
1442 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1443 		break;
1444 	}
1445 	case NI_QTYPE_FQDN:
1446 		nni6->ni_code = ICMP6_NI_SUCCESS;
1447 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1448 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo));
1449 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1450 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1451 		/*
1452 		 * XXX do we really have FQDN in hostname?
1453 		 */
1454 		mtx_lock(&pr->pr_mtx);
1455 		n->m_next = ni6_nametodns(pr->pr_hostname,
1456 		    strlen(pr->pr_hostname), oldfqdn);
1457 		mtx_unlock(&pr->pr_mtx);
1458 		if (n->m_next == NULL)
1459 			goto bad;
1460 		/* XXX we assume that n->m_next is not a chain */
1461 		if (n->m_next->m_next != NULL)
1462 			goto bad;
1463 		n->m_pkthdr.len += n->m_next->m_len;
1464 		break;
1465 	case NI_QTYPE_NODEADDR:
1466 	{
1467 		int lenlim, copied;
1468 
1469 		nni6->ni_code = ICMP6_NI_SUCCESS;
1470 		n->m_pkthdr.len = n->m_len =
1471 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1472 		lenlim = M_TRAILINGSPACE(n);
1473 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1474 		/* XXX: reset mbuf length */
1475 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1476 		    sizeof(struct icmp6_nodeinfo) + copied;
1477 		break;
1478 	}
1479 	default:
1480 		break;		/* XXX impossible! */
1481 	}
1482 
1483 	nni6->ni_type = ICMP6_NI_REPLY;
1484 	m_freem(m);
1485 	return (n);
1486 
1487   bad:
1488 	m_freem(m);
1489 	if (n)
1490 		m_freem(n);
1491 	return (NULL);
1492 }
1493 
1494 /*
1495  * make a mbuf with DNS-encoded string.  no compression support.
1496  *
1497  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1498  * treated as truncated name (two \0 at the end).  this is a wild guess.
1499  *
1500  * old - return pascal string if non-zero
1501  */
1502 static struct mbuf *
1503 ni6_nametodns(const char *name, int namelen, int old)
1504 {
1505 	struct mbuf *m;
1506 	char *cp, *ep;
1507 	const char *p, *q;
1508 	int i, len, nterm;
1509 
1510 	if (old)
1511 		len = namelen + 1;
1512 	else
1513 		len = MCLBYTES;
1514 
1515 	/* Because MAXHOSTNAMELEN is usually 256, we use cluster mbuf. */
1516 	if (len > MLEN)
1517 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1518 	else
1519 		m = m_get(M_NOWAIT, MT_DATA);
1520 	if (m == NULL)
1521 		goto fail;
1522 
1523 	if (old) {
1524 		m->m_len = len;
1525 		*mtod(m, char *) = namelen;
1526 		bcopy(name, mtod(m, char *) + 1, namelen);
1527 		return m;
1528 	} else {
1529 		m->m_len = 0;
1530 		cp = mtod(m, char *);
1531 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1532 
1533 		/* if not certain about my name, return empty buffer */
1534 		if (namelen == 0)
1535 			return m;
1536 
1537 		/*
1538 		 * guess if it looks like shortened hostname, or FQDN.
1539 		 * shortened hostname needs two trailing "\0".
1540 		 */
1541 		i = 0;
1542 		for (p = name; p < name + namelen; p++) {
1543 			if (*p && *p == '.')
1544 				i++;
1545 		}
1546 		if (i < 2)
1547 			nterm = 2;
1548 		else
1549 			nterm = 1;
1550 
1551 		p = name;
1552 		while (cp < ep && p < name + namelen) {
1553 			i = 0;
1554 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1555 				i++;
1556 			/* result does not fit into mbuf */
1557 			if (cp + i + 1 >= ep)
1558 				goto fail;
1559 			/*
1560 			 * DNS label length restriction, RFC1035 page 8.
1561 			 * "i == 0" case is included here to avoid returning
1562 			 * 0-length label on "foo..bar".
1563 			 */
1564 			if (i <= 0 || i >= 64)
1565 				goto fail;
1566 			*cp++ = i;
1567 			bcopy(p, cp, i);
1568 			cp += i;
1569 			p = q;
1570 			if (p < name + namelen && *p == '.')
1571 				p++;
1572 		}
1573 		/* termination */
1574 		if (cp + nterm >= ep)
1575 			goto fail;
1576 		while (nterm-- > 0)
1577 			*cp++ = '\0';
1578 		m->m_len = cp - mtod(m, char *);
1579 		return m;
1580 	}
1581 
1582 	panic("should not reach here");
1583 	/* NOTREACHED */
1584 
1585  fail:
1586 	if (m)
1587 		m_freem(m);
1588 	return NULL;
1589 }
1590 
1591 /*
1592  * check if two DNS-encoded string matches.  takes care of truncated
1593  * form (with \0\0 at the end).  no compression support.
1594  * XXX upper/lowercase match (see RFC2065)
1595  */
1596 static int
1597 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1598 {
1599 	const char *a0, *b0;
1600 	int l;
1601 
1602 	/* simplest case - need validation? */
1603 	if (alen == blen && bcmp(a, b, alen) == 0)
1604 		return 1;
1605 
1606 	a0 = a;
1607 	b0 = b;
1608 
1609 	/* termination is mandatory */
1610 	if (alen < 2 || blen < 2)
1611 		return 0;
1612 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1613 		return 0;
1614 	alen--;
1615 	blen--;
1616 
1617 	while (a - a0 < alen && b - b0 < blen) {
1618 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1619 			return 0;
1620 
1621 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1622 			return 0;
1623 		/* we don't support compression yet */
1624 		if (a[0] >= 64 || b[0] >= 64)
1625 			return 0;
1626 
1627 		/* truncated case */
1628 		if (a[0] == 0 && a - a0 == alen - 1)
1629 			return 1;
1630 		if (b[0] == 0 && b - b0 == blen - 1)
1631 			return 1;
1632 		if (a[0] == 0 || b[0] == 0)
1633 			return 0;
1634 
1635 		if (a[0] != b[0])
1636 			return 0;
1637 		l = a[0];
1638 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1639 			return 0;
1640 		if (bcmp(a + 1, b + 1, l) != 0)
1641 			return 0;
1642 
1643 		a += 1 + l;
1644 		b += 1 + l;
1645 	}
1646 
1647 	if (a - a0 == alen && b - b0 == blen)
1648 		return 1;
1649 	else
1650 		return 0;
1651 }
1652 
1653 /*
1654  * calculate the number of addresses to be returned in the node info reply.
1655  */
1656 static int
1657 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1658     struct in6_addr *subj)
1659 {
1660 	struct ifnet *ifp;
1661 	struct in6_ifaddr *ifa6;
1662 	struct ifaddr *ifa;
1663 	int addrs = 0, addrsofif, iffound = 0;
1664 	int niflags = ni6->ni_flags;
1665 
1666 	NET_EPOCH_ASSERT();
1667 
1668 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1669 		switch (ni6->ni_code) {
1670 		case ICMP6_NI_SUBJ_IPV6:
1671 			if (subj == NULL) /* must be impossible... */
1672 				return (0);
1673 			break;
1674 		default:
1675 			/*
1676 			 * XXX: we only support IPv6 subject address for
1677 			 * this Qtype.
1678 			 */
1679 			return (0);
1680 		}
1681 	}
1682 
1683 	CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) {
1684 		addrsofif = 0;
1685 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1686 			if (ifa->ifa_addr->sa_family != AF_INET6)
1687 				continue;
1688 			ifa6 = (struct in6_ifaddr *)ifa;
1689 
1690 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1691 			    IN6_ARE_ADDR_EQUAL(subj, &ifa6->ia_addr.sin6_addr))
1692 				iffound = 1;
1693 
1694 			/*
1695 			 * IPv4-mapped addresses can only be returned by a
1696 			 * Node Information proxy, since they represent
1697 			 * addresses of IPv4-only nodes, which perforce do
1698 			 * not implement this protocol.
1699 			 * [icmp-name-lookups-07, Section 5.4]
1700 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1701 			 * this function at this moment.
1702 			 */
1703 
1704 			/* What do we have to do about ::1? */
1705 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1706 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1707 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1708 					continue;
1709 				break;
1710 			case IPV6_ADDR_SCOPE_SITELOCAL:
1711 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1712 					continue;
1713 				break;
1714 			case IPV6_ADDR_SCOPE_GLOBAL:
1715 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1716 					continue;
1717 				break;
1718 			default:
1719 				continue;
1720 			}
1721 
1722 			/*
1723 			 * check if anycast is okay.
1724 			 * XXX: just experimental.  not in the spec.
1725 			 */
1726 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1727 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1728 				continue; /* we need only unicast addresses */
1729 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1730 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1731 				continue;
1732 			}
1733 			addrsofif++; /* count the address */
1734 		}
1735 		if (iffound) {
1736 			*ifpp = ifp;
1737 			return (addrsofif);
1738 		}
1739 
1740 		addrs += addrsofif;
1741 	}
1742 
1743 	return (addrs);
1744 }
1745 
1746 static int
1747 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1748     struct ifnet *ifp0, int resid)
1749 {
1750 	struct ifnet *ifp;
1751 	struct in6_ifaddr *ifa6;
1752 	struct ifaddr *ifa;
1753 	struct ifnet *ifp_dep = NULL;
1754 	int copied = 0, allow_deprecated = 0;
1755 	u_char *cp = (u_char *)(nni6 + 1);
1756 	int niflags = ni6->ni_flags;
1757 	u_int32_t ltime;
1758 
1759 	NET_EPOCH_ASSERT();
1760 
1761 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1762 		return (0);	/* needless to copy */
1763 
1764 	ifp = ifp0 ? ifp0 : CK_STAILQ_FIRST(&V_ifnet);
1765   again:
1766 
1767 	for (; ifp; ifp = CK_STAILQ_NEXT(ifp, if_link)) {
1768 		CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1769 			if (ifa->ifa_addr->sa_family != AF_INET6)
1770 				continue;
1771 			ifa6 = (struct in6_ifaddr *)ifa;
1772 
1773 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) != 0 &&
1774 			    allow_deprecated == 0) {
1775 				/*
1776 				 * prefererred address should be put before
1777 				 * deprecated addresses.
1778 				 */
1779 
1780 				/* record the interface for later search */
1781 				if (ifp_dep == NULL)
1782 					ifp_dep = ifp;
1783 
1784 				continue;
1785 			} else if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) == 0 &&
1786 			    allow_deprecated != 0)
1787 				continue; /* we now collect deprecated addrs */
1788 
1789 			/* What do we have to do about ::1? */
1790 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1791 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1792 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1793 					continue;
1794 				break;
1795 			case IPV6_ADDR_SCOPE_SITELOCAL:
1796 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1797 					continue;
1798 				break;
1799 			case IPV6_ADDR_SCOPE_GLOBAL:
1800 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1801 					continue;
1802 				break;
1803 			default:
1804 				continue;
1805 			}
1806 
1807 			/*
1808 			 * check if anycast is okay.
1809 			 * XXX: just experimental.  not in the spec.
1810 			 */
1811 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) != 0 &&
1812 			    (niflags & NI_NODEADDR_FLAG_ANYCAST) == 0)
1813 				continue;
1814 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) != 0 &&
1815 			    (V_icmp6_nodeinfo & ICMP6_NODEINFO_TMPADDROK) == 0) {
1816 				continue;
1817 			}
1818 
1819 			/* now we can copy the address */
1820 			if (resid < sizeof(struct in6_addr) +
1821 			    sizeof(u_int32_t)) {
1822 				/*
1823 				 * We give up much more copy.
1824 				 * Set the truncate flag and return.
1825 				 */
1826 				nni6->ni_flags |= NI_NODEADDR_FLAG_TRUNCATE;
1827 				return (copied);
1828 			}
1829 
1830 			/*
1831 			 * Set the TTL of the address.
1832 			 * The TTL value should be one of the following
1833 			 * according to the specification:
1834 			 *
1835 			 * 1. The remaining lifetime of a DHCP lease on the
1836 			 *    address, or
1837 			 * 2. The remaining Valid Lifetime of a prefix from
1838 			 *    which the address was derived through Stateless
1839 			 *    Autoconfiguration.
1840 			 *
1841 			 * Note that we currently do not support stateful
1842 			 * address configuration by DHCPv6, so the former
1843 			 * case can't happen.
1844 			 */
1845 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1846 				ltime = ND6_INFINITE_LIFETIME;
1847 			else {
1848 				if (ifa6->ia6_lifetime.ia6t_expire >
1849 				    time_uptime)
1850 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime);
1851 				else
1852 					ltime = 0;
1853 			}
1854 
1855 			bcopy(&ltime, cp, sizeof(u_int32_t));
1856 			cp += sizeof(u_int32_t);
1857 
1858 			/* copy the address itself */
1859 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1860 			    sizeof(struct in6_addr));
1861 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1862 			cp += sizeof(struct in6_addr);
1863 
1864 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1865 			copied += (sizeof(struct in6_addr) + sizeof(u_int32_t));
1866 		}
1867 		if (ifp0)	/* we need search only on the specified IF */
1868 			break;
1869 	}
1870 
1871 	if (allow_deprecated == 0 && ifp_dep != NULL) {
1872 		ifp = ifp_dep;
1873 		allow_deprecated = 1;
1874 
1875 		goto again;
1876 	}
1877 
1878 	return (copied);
1879 }
1880 
1881 static bool
1882 icmp6_rip6_match(const struct inpcb *inp, void *v)
1883 {
1884 	struct ip6_hdr *ip6 = v;
1885 
1886 	if ((inp->inp_vflag & INP_IPV6) == 0)
1887 		return (false);
1888 	if (inp->inp_ip_p != IPPROTO_ICMPV6)
1889 		return (false);
1890 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) &&
1891 	   !IN6_ARE_ADDR_EQUAL(&inp->in6p_laddr, &ip6->ip6_dst))
1892 		return (false);
1893 	if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr) &&
1894 	   !IN6_ARE_ADDR_EQUAL(&inp->in6p_faddr, &ip6->ip6_src))
1895 		return (false);
1896 	return (true);
1897 }
1898 
1899 /*
1900  * XXX almost dup'ed code with rip6_input.
1901  */
1902 static int
1903 icmp6_rip6_input(struct mbuf **mp, int off)
1904 {
1905 	struct mbuf *n, *m = *mp;
1906 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1907 	struct inpcb_iterator inpi = INP_ITERATOR(&V_ripcbinfo,
1908 	    INPLOOKUP_RLOCKPCB, icmp6_rip6_match, ip6);
1909 	struct inpcb *inp;
1910 	struct sockaddr_in6 fromsa;
1911 	struct icmp6_hdr *icmp6;
1912 	struct mbuf *opts = NULL;
1913 	int delivered = 0;
1914 
1915 	/* This is assumed to be safe; icmp6_input() does a pullup. */
1916 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1917 
1918 	/*
1919 	 * XXX: the address may have embedded scope zone ID, which should be
1920 	 * hidden from applications.
1921 	 */
1922 	bzero(&fromsa, sizeof(fromsa));
1923 	fromsa.sin6_family = AF_INET6;
1924 	fromsa.sin6_len = sizeof(struct sockaddr_in6);
1925 	fromsa.sin6_addr = ip6->ip6_src;
1926 	if (sa6_recoverscope(&fromsa)) {
1927 		m_freem(m);
1928 		*mp = NULL;
1929 		return (IPPROTO_DONE);
1930 	}
1931 
1932 	while ((inp = inp_next(&inpi)) != NULL) {
1933 		if (ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1934 		    inp->in6p_icmp6filt))
1935 			continue;
1936 		/*
1937 		 * Recent network drivers tend to allocate a single
1938 		 * mbuf cluster, rather than to make a couple of
1939 		 * mbufs without clusters.  Also, since the IPv6 code
1940 		 * path tries to avoid m_pullup(), it is highly
1941 		 * probable that we still have an mbuf cluster here
1942 		 * even though the necessary length can be stored in an
1943 		 * mbuf's internal buffer.
1944 		 * Meanwhile, the default size of the receive socket
1945 		 * buffer for raw sockets is not so large.  This means
1946 		 * the possibility of packet loss is relatively higher
1947 		 * than before.  To avoid this scenario, we copy the
1948 		 * received data to a separate mbuf that does not use
1949 		 * a cluster, if possible.
1950 		 * XXX: it is better to copy the data after stripping
1951 		 * intermediate headers.
1952 		 */
1953 		if ((m->m_flags & M_EXT) && m->m_next == NULL &&
1954 		    m->m_len <= MHLEN) {
1955 			n = m_get(M_NOWAIT, m->m_type);
1956 			if (n != NULL) {
1957 				if (m_dup_pkthdr(n, m, M_NOWAIT)) {
1958 					bcopy(m->m_data, n->m_data, m->m_len);
1959 					n->m_len = m->m_len;
1960 				} else {
1961 					m_free(n);
1962 					n = NULL;
1963 				}
1964 			}
1965 		} else
1966 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
1967 		if (n == NULL)
1968 			continue;
1969 		if (inp->inp_flags & INP_CONTROLOPTS)
1970 			ip6_savecontrol(inp, n, &opts);
1971 		/* strip intermediate headers */
1972 		m_adj(n, off);
1973 		SOCKBUF_LOCK(&inp->inp_socket->so_rcv);
1974 		if (sbappendaddr_locked(&inp->inp_socket->so_rcv,
1975 		    (struct sockaddr *)&fromsa, n, opts) == 0) {
1976 			soroverflow_locked(inp->inp_socket);
1977 			m_freem(n);
1978 			if (opts)
1979 				m_freem(opts);
1980 		} else {
1981 			sorwakeup_locked(inp->inp_socket);
1982 			delivered++;
1983 		}
1984 		opts = NULL;
1985 	}
1986 	m_freem(m);
1987 	*mp = NULL;
1988 	if (delivered == 0)
1989 		IP6STAT_DEC(ip6s_delivered);
1990 	return (IPPROTO_DONE);
1991 }
1992 
1993 /*
1994  * Reflect the ip6 packet back to the source.
1995  * OFF points to the icmp6 header, counted from the top of the mbuf.
1996  */
1997 static void
1998 icmp6_reflect(struct mbuf *m, size_t off)
1999 {
2000 	struct in6_addr src6, *srcp;
2001 	struct ip6_hdr *ip6;
2002 	struct icmp6_hdr *icmp6;
2003 	struct in6_ifaddr *ia = NULL;
2004 	struct ifnet *outif = NULL;
2005 	int plen;
2006 	int type, code, hlim;
2007 
2008 	/* too short to reflect */
2009 	if (off < sizeof(struct ip6_hdr)) {
2010 		nd6log((LOG_DEBUG,
2011 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
2012 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
2013 		    __FILE__, __LINE__));
2014 		goto bad;
2015 	}
2016 
2017 	/*
2018 	 * If there are extra headers between IPv6 and ICMPv6, strip
2019 	 * off that header first.
2020 	 */
2021 #ifdef DIAGNOSTIC
2022 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
2023 		panic("assumption failed in icmp6_reflect");
2024 #endif
2025 	if (off > sizeof(struct ip6_hdr)) {
2026 		size_t l;
2027 		struct ip6_hdr nip6;
2028 
2029 		l = off - sizeof(struct ip6_hdr);
2030 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
2031 		m_adj(m, l);
2032 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2033 		if (m->m_len < l) {
2034 			if ((m = m_pullup(m, l)) == NULL)
2035 				return;
2036 		}
2037 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
2038 	} else /* off == sizeof(struct ip6_hdr) */ {
2039 		size_t l;
2040 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
2041 		if (m->m_len < l) {
2042 			if ((m = m_pullup(m, l)) == NULL)
2043 				return;
2044 		}
2045 	}
2046 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
2047 	ip6 = mtod(m, struct ip6_hdr *);
2048 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2049 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
2050 	type = icmp6->icmp6_type; /* keep type for statistics */
2051 	code = icmp6->icmp6_code; /* ditto. */
2052 	hlim = 0;
2053 	srcp = NULL;
2054 
2055 	/*
2056 	 * If the incoming packet was addressed directly to us (i.e. unicast),
2057 	 * use dst as the src for the reply.
2058 	 * The IN6_IFF_NOTREADY case should be VERY rare, but is possible
2059 	 * (for example) when we encounter an error while forwarding procedure
2060 	 * destined to a duplicated address of ours.
2061 	 */
2062 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
2063 		ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false);
2064 		if (ia != NULL && !(ia->ia6_flags &
2065 		    (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY))) {
2066 			src6 = ia->ia_addr.sin6_addr;
2067 			srcp = &src6;
2068 
2069 			if (m->m_pkthdr.rcvif != NULL) {
2070 				/* XXX: This may not be the outgoing interface */
2071 				hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2072 			} else
2073 				hlim = V_ip6_defhlim;
2074 		}
2075 	}
2076 
2077 	if (srcp == NULL) {
2078 		int error;
2079 		struct in6_addr dst6;
2080 		uint32_t scopeid;
2081 
2082 		/*
2083 		 * This case matches to multicasts, our anycast, or unicasts
2084 		 * that we do not own.  Select a source address based on the
2085 		 * source address of the erroneous packet.
2086 		 */
2087 		in6_splitscope(&ip6->ip6_src, &dst6, &scopeid);
2088 		error = in6_selectsrc_addr(M_GETFIB(m), &dst6,
2089 		    scopeid, NULL, &src6, &hlim);
2090 
2091 		if (error) {
2092 			char ip6buf[INET6_ADDRSTRLEN];
2093 			nd6log((LOG_DEBUG,
2094 			    "icmp6_reflect: source can't be determined: "
2095 			    "dst=%s, error=%d\n",
2096 			    ip6_sprintf(ip6buf, &ip6->ip6_dst), error));
2097 			goto bad;
2098 		}
2099 		srcp = &src6;
2100 	}
2101 	/*
2102 	 * ip6_input() drops a packet if its src is multicast.
2103 	 * So, the src is never multicast.
2104 	 */
2105 	ip6->ip6_dst = ip6->ip6_src;
2106 	ip6->ip6_src = *srcp;
2107 	ip6->ip6_flow = 0;
2108 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2109 	ip6->ip6_vfc |= IPV6_VERSION;
2110 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2111 	ip6->ip6_hlim = hlim;
2112 
2113 	icmp6->icmp6_cksum = 0;
2114 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2115 	    sizeof(struct ip6_hdr), plen);
2116 
2117 	/*
2118 	 * XXX option handling
2119 	 */
2120 
2121 	m->m_flags &= ~(M_BCAST|M_MCAST);
2122 	m->m_pkthdr.rcvif = NULL;
2123 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2124 	if (outif)
2125 		icmp6_ifoutstat_inc(outif, type, code);
2126 
2127 	return;
2128 
2129  bad:
2130 	m_freem(m);
2131 	return;
2132 }
2133 
2134 void
2135 icmp6_fasttimo(void)
2136 {
2137 
2138 	mld_fasttimo();
2139 }
2140 
2141 void
2142 icmp6_slowtimo(void)
2143 {
2144 
2145 	mld_slowtimo();
2146 }
2147 
2148 static const char *
2149 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2150     struct in6_addr *tgt6)
2151 {
2152 	static char buf[1024];
2153 	char ip6bufs[INET6_ADDRSTRLEN];
2154 	char ip6bufd[INET6_ADDRSTRLEN];
2155 	char ip6buft[INET6_ADDRSTRLEN];
2156 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2157 	    ip6_sprintf(ip6bufs, src6), ip6_sprintf(ip6bufd, dst6),
2158 	    ip6_sprintf(ip6buft, tgt6));
2159 	return buf;
2160 }
2161 
2162 void
2163 icmp6_redirect_input(struct mbuf *m, int off)
2164 {
2165 	struct ifnet *ifp;
2166 	struct ip6_hdr *ip6;
2167 	struct nd_redirect *nd_rd;
2168 	struct in6_addr src6, redtgt6, reddst6;
2169 	union nd_opts ndopts;
2170 	char ip6buf[INET6_ADDRSTRLEN];
2171 	char *lladdr;
2172 	int icmp6len, is_onlink, is_router, lladdrlen;
2173 
2174 	M_ASSERTPKTHDR(m);
2175 	KASSERT(m->m_pkthdr.rcvif != NULL, ("%s: no rcvif", __func__));
2176 
2177 	/* XXX if we are router, we don't update route by icmp6 redirect */
2178 	if (V_ip6_forwarding)
2179 		goto freeit;
2180 	if (!V_icmp6_rediraccept)
2181 		goto freeit;
2182 
2183 	/* RFC 6980: Nodes MUST silently ignore fragments */
2184 	if(m->m_flags & M_FRAGMENTED)
2185 		goto freeit;
2186 
2187 	ip6 = mtod(m, struct ip6_hdr *);
2188 	icmp6len = ntohs(ip6->ip6_plen);
2189 	if (m->m_len < off + icmp6len) {
2190 		m = m_pullup(m, off + icmp6len);
2191 		if (m == NULL) {
2192 			IP6STAT_INC(ip6s_exthdrtoolong);
2193 			return;
2194 		}
2195 	}
2196 	ip6 = mtod(m, struct ip6_hdr *);
2197 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2198 
2199 	ifp = m->m_pkthdr.rcvif;
2200 	redtgt6 = nd_rd->nd_rd_target;
2201 	reddst6 = nd_rd->nd_rd_dst;
2202 
2203 	if (in6_setscope(&redtgt6, ifp, NULL) ||
2204 	    in6_setscope(&reddst6, ifp, NULL)) {
2205 		goto freeit;
2206 	}
2207 
2208 	/* validation */
2209 	src6 = ip6->ip6_src;
2210 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2211 		nd6log((LOG_ERR,
2212 		    "ICMP6 redirect sent from %s rejected; "
2213 		    "must be from linklocal\n",
2214 		    ip6_sprintf(ip6buf, &src6)));
2215 		goto bad;
2216 	}
2217 	if (__predict_false(ip6->ip6_hlim != 255)) {
2218 		ICMP6STAT_INC(icp6s_invlhlim);
2219 		nd6log((LOG_ERR,
2220 		    "ICMP6 redirect sent from %s rejected; "
2221 		    "hlim=%d (must be 255)\n",
2222 		    ip6_sprintf(ip6buf, &src6), ip6->ip6_hlim));
2223 		goto bad;
2224 	}
2225     {
2226 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2227 	struct nhop_object *nh;
2228 	struct in6_addr kdst;
2229 	uint32_t scopeid;
2230 
2231 	in6_splitscope(&reddst6, &kdst, &scopeid);
2232 	NET_EPOCH_ASSERT();
2233 	nh = fib6_lookup(ifp->if_fib, &kdst, scopeid, 0, 0);
2234 	if (nh != NULL) {
2235 		struct in6_addr nh_addr;
2236 		nh_addr = ifatoia6(nh->nh_ifa)->ia_addr.sin6_addr;
2237 		if ((nh->nh_flags & NHF_GATEWAY) == 0) {
2238 			nd6log((LOG_ERR,
2239 			    "ICMP6 redirect rejected; no route "
2240 			    "with inet6 gateway found for redirect dst: %s\n",
2241 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2242 			goto bad;
2243 		}
2244 
2245 		/*
2246 		 * Embed scope zone id into next hop address.
2247 		 */
2248 		nh_addr = nh->gw6_sa.sin6_addr;
2249 
2250 		if (IN6_ARE_ADDR_EQUAL(&src6, &nh_addr) == 0) {
2251 			nd6log((LOG_ERR,
2252 			    "ICMP6 redirect rejected; "
2253 			    "not equal to gw-for-src=%s (must be same): "
2254 			    "%s\n",
2255 			    ip6_sprintf(ip6buf, &nh_addr),
2256 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2257 			goto bad;
2258 		}
2259 	} else {
2260 		nd6log((LOG_ERR,
2261 		    "ICMP6 redirect rejected; "
2262 		    "no route found for redirect dst: %s\n",
2263 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2264 		goto bad;
2265 	}
2266     }
2267 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2268 		nd6log((LOG_ERR,
2269 		    "ICMP6 redirect rejected; "
2270 		    "redirect dst must be unicast: %s\n",
2271 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2272 		goto bad;
2273 	}
2274 
2275 	is_router = is_onlink = 0;
2276 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2277 		is_router = 1;	/* router case */
2278 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2279 		is_onlink = 1;	/* on-link destination case */
2280 	if (!is_router && !is_onlink) {
2281 		nd6log((LOG_ERR,
2282 		    "ICMP6 redirect rejected; "
2283 		    "neither router case nor onlink case: %s\n",
2284 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2285 		goto bad;
2286 	}
2287 
2288 	icmp6len -= sizeof(*nd_rd);
2289 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2290 	if (nd6_options(&ndopts) < 0) {
2291 		nd6log((LOG_INFO, "%s: invalid ND option, rejected: %s\n",
2292 		    __func__, icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2293 		/* nd6_options have incremented stats */
2294 		goto freeit;
2295 	}
2296 
2297 	lladdr = NULL;
2298 	lladdrlen = 0;
2299 	if (ndopts.nd_opts_tgt_lladdr) {
2300 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2301 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2302 	}
2303 
2304 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2305 		nd6log((LOG_INFO, "%s: lladdrlen mismatch for %s "
2306 		    "(if %d, icmp6 packet %d): %s\n",
2307 		    __func__, ip6_sprintf(ip6buf, &redtgt6),
2308 		    ifp->if_addrlen, lladdrlen - 2,
2309 		    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2310 		goto bad;
2311 	}
2312 
2313 	/* Validation passed. */
2314 
2315 	/* RFC 2461 8.3 */
2316 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2317 	    is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2318 
2319 	/*
2320 	 * Install a gateway route in the better-router case or an interface
2321 	 * route in the on-link-destination case.
2322 	 */
2323 	{
2324 		struct sockaddr_in6 sdst;
2325 		struct sockaddr_in6 sgw;
2326 		struct sockaddr_in6 ssrc;
2327 		struct sockaddr *gw;
2328 		int rt_flags;
2329 		u_int fibnum;
2330 
2331 		bzero(&sdst, sizeof(sdst));
2332 		bzero(&ssrc, sizeof(ssrc));
2333 		sdst.sin6_family = ssrc.sin6_family = AF_INET6;
2334 		sdst.sin6_len = ssrc.sin6_len = sizeof(struct sockaddr_in6);
2335 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2336 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2337 		rt_flags = 0;
2338 		if (is_router) {
2339 			bzero(&sgw, sizeof(sgw));
2340 			sgw.sin6_family = AF_INET6;
2341 			sgw.sin6_len = sizeof(struct sockaddr_in6);
2342 			bcopy(&redtgt6, &sgw.sin6_addr,
2343 				sizeof(struct in6_addr));
2344 			gw = (struct sockaddr *)&sgw;
2345 			rt_flags |= RTF_GATEWAY;
2346 		} else
2347 			gw = ifp->if_addr->ifa_addr;
2348 		for (fibnum = 0; fibnum < rt_numfibs; fibnum++)
2349 			rib_add_redirect(fibnum, (struct sockaddr *)&sdst, gw,
2350 			    (struct sockaddr *)&ssrc, ifp, rt_flags,
2351 			    V_icmp6_redirtimeout);
2352 	}
2353 	/* finally update cached route in each socket via pfctlinput */
2354     {
2355 	struct sockaddr_in6 sdst;
2356 
2357 	bzero(&sdst, sizeof(sdst));
2358 	sdst.sin6_family = AF_INET6;
2359 	sdst.sin6_len = sizeof(struct sockaddr_in6);
2360 	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2361 	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2362     }
2363 
2364  freeit:
2365 	m_freem(m);
2366 	return;
2367 
2368  bad:
2369 	ICMP6STAT_INC(icp6s_badredirect);
2370 	m_freem(m);
2371 }
2372 
2373 void
2374 icmp6_redirect_output(struct mbuf *m0, struct nhop_object *nh)
2375 {
2376 	struct ifnet *ifp;	/* my outgoing interface */
2377 	struct in6_addr *ifp_ll6;
2378 	struct in6_addr *router_ll6;
2379 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2380 	struct mbuf *m = NULL;	/* newly allocated one */
2381 	struct m_tag *mtag;
2382 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2383 	struct nd_redirect *nd_rd;
2384 	struct llentry *ln = NULL;
2385 	size_t maxlen;
2386 	u_char *p;
2387 	struct ifnet *outif = NULL;
2388 	struct sockaddr_in6 src_sa;
2389 
2390 	icmp6_errcount(ND_REDIRECT, 0);
2391 
2392 	/* if we are not router, we don't send icmp6 redirect */
2393 	if (!V_ip6_forwarding)
2394 		goto fail;
2395 
2396 	/* sanity check */
2397 	if (!m0 || !nh || !(NH_IS_VALID(nh)) || !(ifp = nh->nh_ifp))
2398 		goto fail;
2399 
2400 	/*
2401 	 * Address check:
2402 	 *  the source address must identify a neighbor, and
2403 	 *  the destination address must not be a multicast address
2404 	 *  [RFC 2461, sec 8.2]
2405 	 */
2406 	sip6 = mtod(m0, struct ip6_hdr *);
2407 	bzero(&src_sa, sizeof(src_sa));
2408 	src_sa.sin6_family = AF_INET6;
2409 	src_sa.sin6_len = sizeof(src_sa);
2410 	src_sa.sin6_addr = sip6->ip6_src;
2411 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2412 		goto fail;
2413 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2414 		goto fail;	/* what should we do here? */
2415 
2416 	/* rate limit */
2417 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2418 		goto fail;
2419 
2420 	/*
2421 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2422 	 * we almost always ask for an mbuf cluster for simplicity.
2423 	 * (MHLEN < IPV6_MMTU is almost always true)
2424 	 */
2425 #if IPV6_MMTU >= MCLBYTES
2426 # error assumption failed about IPV6_MMTU and MCLBYTES
2427 #endif
2428 	m = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
2429 	if (m == NULL)
2430 		goto fail;
2431 	M_SETFIB(m, M_GETFIB(m0));
2432 	maxlen = M_TRAILINGSPACE(m);
2433 	maxlen = min(IPV6_MMTU, maxlen);
2434 	/* just for safety */
2435 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2436 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2437 		goto fail;
2438 	}
2439 
2440 	{
2441 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2442 		struct in6_ifaddr *ia;
2443 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2444 						 IN6_IFF_NOTREADY|
2445 						 IN6_IFF_ANYCAST)) == NULL)
2446 			goto fail;
2447 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2448 		/* XXXRW: reference released prematurely. */
2449 		ifa_free(&ia->ia_ifa);
2450 	}
2451 
2452 	/* get ip6 linklocal address for the router. */
2453 	if (nh->nh_flags & NHF_GATEWAY) {
2454 		struct sockaddr_in6 *sin6;
2455 		sin6 = &nh->gw6_sa;
2456 		router_ll6 = &sin6->sin6_addr;
2457 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2458 			router_ll6 = (struct in6_addr *)NULL;
2459 	} else
2460 		router_ll6 = (struct in6_addr *)NULL;
2461 
2462 	/* ip6 */
2463 	ip6 = mtod(m, struct ip6_hdr *);
2464 	ip6->ip6_flow = 0;
2465 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2466 	ip6->ip6_vfc |= IPV6_VERSION;
2467 	/* ip6->ip6_plen will be set later */
2468 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2469 	ip6->ip6_hlim = 255;
2470 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2471 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2472 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2473 
2474 	/* ND Redirect */
2475 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2476 	nd_rd->nd_rd_type = ND_REDIRECT;
2477 	nd_rd->nd_rd_code = 0;
2478 	nd_rd->nd_rd_reserved = 0;
2479 	if (nh->nh_flags & NHF_GATEWAY) {
2480 		/*
2481 		 * nd_rd->nd_rd_target must be a link-local address in
2482 		 * better router cases.
2483 		 */
2484 		if (!router_ll6)
2485 			goto fail;
2486 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2487 		    sizeof(nd_rd->nd_rd_target));
2488 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2489 		    sizeof(nd_rd->nd_rd_dst));
2490 	} else {
2491 		/* make sure redtgt == reddst */
2492 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2493 		    sizeof(nd_rd->nd_rd_target));
2494 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2495 		    sizeof(nd_rd->nd_rd_dst));
2496 	}
2497 
2498 	p = (u_char *)(nd_rd + 1);
2499 
2500 	if (!router_ll6)
2501 		goto nolladdropt;
2502 
2503 	{
2504 		/* target lladdr option */
2505 		int len;
2506 		struct nd_opt_hdr *nd_opt;
2507 		char *lladdr;
2508 
2509 		ln = nd6_lookup(router_ll6, LLE_SF(AF_INET6,  0), ifp);
2510 		if (ln == NULL)
2511 			goto nolladdropt;
2512 
2513 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2514 		len = (len + 7) & ~7;	/* round by 8 */
2515 		/* safety check */
2516 		if (len + (p - (u_char *)ip6) > maxlen)
2517 			goto nolladdropt;
2518 
2519 		if (ln->la_flags & LLE_VALID) {
2520 			nd_opt = (struct nd_opt_hdr *)p;
2521 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2522 			nd_opt->nd_opt_len = len >> 3;
2523 			lladdr = (char *)(nd_opt + 1);
2524 			bcopy(ln->ll_addr, lladdr, ifp->if_addrlen);
2525 			p += len;
2526 		}
2527 	}
2528 nolladdropt:
2529 	if (ln != NULL)
2530 		LLE_RUNLOCK(ln);
2531 
2532 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2533 
2534 	/* just to be safe */
2535 #ifdef M_DECRYPTED	/*not openbsd*/
2536 	if (m0->m_flags & M_DECRYPTED)
2537 		goto noredhdropt;
2538 #endif
2539 	if (p - (u_char *)ip6 > maxlen)
2540 		goto noredhdropt;
2541 
2542 	{
2543 		/* redirected header option */
2544 		int len;
2545 		struct nd_opt_rd_hdr *nd_opt_rh;
2546 
2547 		/*
2548 		 * compute the maximum size for icmp6 redirect header option.
2549 		 * XXX room for auth header?
2550 		 */
2551 		len = maxlen - (p - (u_char *)ip6);
2552 		len &= ~7;
2553 
2554 		/* This is just for simplicity. */
2555 		if (m0->m_pkthdr.len != m0->m_len) {
2556 			if (m0->m_next) {
2557 				m_freem(m0->m_next);
2558 				m0->m_next = NULL;
2559 			}
2560 			m0->m_pkthdr.len = m0->m_len;
2561 		}
2562 
2563 		/*
2564 		 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2565 		 * about padding/truncate rule for the original IP packet.
2566 		 * From the discussion on IPv6imp in Feb 1999,
2567 		 * the consensus was:
2568 		 * - "attach as much as possible" is the goal
2569 		 * - pad if not aligned (original size can be guessed by
2570 		 *   original ip6 header)
2571 		 * Following code adds the padding if it is simple enough,
2572 		 * and truncates if not.
2573 		 */
2574 		if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2575 			panic("assumption failed in %s:%d", __FILE__,
2576 			    __LINE__);
2577 
2578 		if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2579 			/* not enough room, truncate */
2580 			m0->m_pkthdr.len = m0->m_len = len -
2581 			    sizeof(*nd_opt_rh);
2582 		} else {
2583 			/* enough room, pad or truncate */
2584 			size_t extra;
2585 
2586 			extra = m0->m_pkthdr.len % 8;
2587 			if (extra) {
2588 				/* pad if easy enough, truncate if not */
2589 				if (8 - extra <= M_TRAILINGSPACE(m0)) {
2590 					/* pad */
2591 					m0->m_len += (8 - extra);
2592 					m0->m_pkthdr.len += (8 - extra);
2593 				} else {
2594 					/* truncate */
2595 					m0->m_pkthdr.len -= extra;
2596 					m0->m_len -= extra;
2597 				}
2598 			}
2599 			len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2600 			m0->m_pkthdr.len = m0->m_len = len -
2601 			    sizeof(*nd_opt_rh);
2602 		}
2603 
2604 		nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2605 		bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2606 		nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2607 		nd_opt_rh->nd_opt_rh_len = len >> 3;
2608 		p += sizeof(*nd_opt_rh);
2609 		m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2610 
2611 		/* connect m0 to m */
2612 		m_tag_delete_chain(m0, NULL);
2613 		m0->m_flags &= ~M_PKTHDR;
2614 		m->m_next = m0;
2615 		m->m_pkthdr.len = m->m_len + m0->m_len;
2616 		m0 = NULL;
2617 	}
2618 noredhdropt:;
2619 	if (m0) {
2620 		m_freem(m0);
2621 		m0 = NULL;
2622 	}
2623 
2624 	/* XXX: clear embedded link IDs in the inner header */
2625 	in6_clearscope(&sip6->ip6_src);
2626 	in6_clearscope(&sip6->ip6_dst);
2627 	in6_clearscope(&nd_rd->nd_rd_target);
2628 	in6_clearscope(&nd_rd->nd_rd_dst);
2629 
2630 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2631 
2632 	nd_rd->nd_rd_cksum = 0;
2633 	nd_rd->nd_rd_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2634 	    sizeof(*ip6), ntohs(ip6->ip6_plen));
2635 
2636         if (send_sendso_input_hook != NULL) {
2637 		mtag = m_tag_get(PACKET_TAG_ND_OUTGOING, sizeof(unsigned short),
2638 			M_NOWAIT);
2639 		if (mtag == NULL)
2640 			goto fail;
2641 		*(unsigned short *)(mtag + 1) = nd_rd->nd_rd_type;
2642 		m_tag_prepend(m, mtag);
2643 	}
2644 
2645 	/* send the packet to outside... */
2646 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2647 	if (outif) {
2648 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2649 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2650 	}
2651 	ICMP6STAT_INC(icp6s_outhist[ND_REDIRECT]);
2652 
2653 	return;
2654 
2655 fail:
2656 	if (m)
2657 		m_freem(m);
2658 	if (m0)
2659 		m_freem(m0);
2660 }
2661 
2662 /*
2663  * ICMPv6 socket option processing.
2664  */
2665 int
2666 icmp6_ctloutput(struct socket *so, struct sockopt *sopt)
2667 {
2668 	int error = 0;
2669 	int optlen;
2670 	struct inpcb *inp = sotoinpcb(so);
2671 	int level, op, optname;
2672 
2673 	if (sopt) {
2674 		level = sopt->sopt_level;
2675 		op = sopt->sopt_dir;
2676 		optname = sopt->sopt_name;
2677 		optlen = sopt->sopt_valsize;
2678 	} else
2679 		level = op = optname = optlen = 0;
2680 
2681 	if (level != IPPROTO_ICMPV6) {
2682 		return EINVAL;
2683 	}
2684 
2685 	switch (op) {
2686 	case PRCO_SETOPT:
2687 		switch (optname) {
2688 		case ICMP6_FILTER:
2689 		    {
2690 			struct icmp6_filter ic6f;
2691 
2692 			if (optlen != sizeof(ic6f)) {
2693 				error = EMSGSIZE;
2694 				break;
2695 			}
2696 			error = sooptcopyin(sopt, &ic6f, optlen, optlen);
2697 			if (error == 0) {
2698 				INP_WLOCK(inp);
2699 				*inp->in6p_icmp6filt = ic6f;
2700 				INP_WUNLOCK(inp);
2701 			}
2702 			break;
2703 		    }
2704 
2705 		default:
2706 			error = ENOPROTOOPT;
2707 			break;
2708 		}
2709 		break;
2710 
2711 	case PRCO_GETOPT:
2712 		switch (optname) {
2713 		case ICMP6_FILTER:
2714 		    {
2715 			struct icmp6_filter ic6f;
2716 
2717 			INP_RLOCK(inp);
2718 			ic6f = *inp->in6p_icmp6filt;
2719 			INP_RUNLOCK(inp);
2720 			error = sooptcopyout(sopt, &ic6f, sizeof(ic6f));
2721 			break;
2722 		    }
2723 
2724 		default:
2725 			error = ENOPROTOOPT;
2726 			break;
2727 		}
2728 		break;
2729 	}
2730 
2731 	return (error);
2732 }
2733 
2734 /*
2735  * Perform rate limit check.
2736  * Returns 0 if it is okay to send the icmp6 packet.
2737  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2738  * limitation.
2739  *
2740  * XXX per-destination/type check necessary?
2741  *
2742  * dst - not used at this moment
2743  * type - not used at this moment
2744  * code - not used at this moment
2745  */
2746 static int
2747 icmp6_ratelimit(const struct in6_addr *dst, const int type,
2748     const int code)
2749 {
2750 	int ret;
2751 
2752 	ret = 0;	/* okay to send */
2753 
2754 	/* PPS limit */
2755 	if (!ppsratecheck(&V_icmp6errppslim_last, &V_icmp6errpps_count,
2756 	    V_icmp6errppslim)) {
2757 		/* The packet is subject to rate limit */
2758 		ret++;
2759 	}
2760 
2761 	return ret;
2762 }
2763