xref: /dragonfly/sys/netinet6/icmp6.c (revision 64355093)
1 /*	$FreeBSD: src/sys/netinet6/icmp6.c,v 1.6.2.13 2003/05/06 06:46:58 suz Exp $	*/
2 /*	$KAME: icmp6.c,v 1.211 2001/04/04 05:56:20 itojun Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1988, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
62  */
63 
64 #include "opt_inet.h"
65 #include "opt_inet6.h"
66 
67 #include <sys/param.h>
68 #include <sys/systm.h>
69 #include <sys/malloc.h>
70 #include <sys/mbuf.h>
71 #include <sys/protosw.h>
72 #include <sys/socket.h>
73 #include <sys/socketvar.h>
74 #include <sys/socketops.h>
75 #include <sys/time.h>
76 #include <sys/kernel.h>
77 #include <sys/syslog.h>
78 #include <sys/domain.h>
79 
80 #include <sys/thread2.h>
81 #include <sys/msgport2.h>
82 
83 #include <net/if.h>
84 #include <net/route.h>
85 #include <net/if_dl.h>
86 #include <net/if_types.h>
87 #include <net/netisr2.h>
88 #include <net/netmsg2.h>
89 
90 #include <netinet/in.h>
91 #include <netinet/in_var.h>
92 #include <netinet/ip6.h>
93 #include <netinet6/ip6_var.h>
94 #include <netinet/icmp6.h>
95 #include <netinet6/mld6_var.h>
96 #include <netinet/in_pcb.h>
97 #include <netinet6/in6_pcb.h>
98 #include <netinet6/nd6.h>
99 #include <netinet6/in6_ifattach.h>
100 #include <netinet6/ip6protosw.h>
101 
102 #include <net/net_osdep.h>
103 
104 #ifdef HAVE_NRL_INPCB
105 /* inpcb members */
106 #define in6pcb		inpcb
107 #define in6p_laddr	inp_laddr6
108 #define in6p_faddr	inp_faddr6
109 #define in6p_icmp6filt	inp_icmp6filt
110 #define in6p_route	inp_route
111 #define in6p_socket	inp_socket
112 #define in6p_flags	inp_flags
113 #define in6p_moptions	inp_moptions6
114 #define in6p_outputopts	inp_outputopts6
115 #define in6p_ip6	inp_ipv6
116 #define in6p_flowinfo	inp_flowinfo
117 #define in6p_next	inp_next
118 #define in6p_prev	inp_prev
119 /* function names */
120 #define in6_pcbdetach	in_pcbdetach
121 #define in6_rtchange	in_rtchange
122 
123 /*
124  * for KAME src sync over BSD*'s. XXX: FreeBSD (>=3) are VERY different from
125  * others...
126  */
127 #define in6p_ip6_nxt	inp_ipv6.ip6_nxt
128 #endif
129 
130 #define ICMP6_FASTTIMO	(hz / PR_FASTHZ)
131 
132 /*
133  * ppratecheck() is available, so define it here.
134  */
135 #define	HAVE_PPSRATECHECK
136 
137 extern struct domain inet6domain;
138 extern struct protosw inet6sw[];
139 extern u_char ip6_protox[];
140 
141 struct icmp6stat icmp6stat;
142 
143 extern struct inpcbinfo ripcbinfo;
144 extern int icmp6errppslim;
145 static int icmp6errpps_count = 0;
146 static struct timeval icmp6errppslim_last;
147 extern int icmp6_nodeinfo;
148 
149 static void icmp6_errcount (struct icmp6errstat *, int, int);
150 static int icmp6_rip6_input (struct mbuf **, int);
151 static int icmp6_ratelimit (const struct in6_addr *, const int, const int);
152 static const char *icmp6_redirect_diag (struct in6_addr *,
153 	struct in6_addr *, struct in6_addr *);
154 #ifndef HAVE_PPSRATECHECK
155 static int ppsratecheck (struct timeval *, int *, int);
156 #endif
157 static struct mbuf *ni6_input (struct mbuf *, int);
158 static struct mbuf *ni6_nametodns (const char *, int, int);
159 static int ni6_dnsmatch (const char *, int, const char *, int);
160 static int ni6_addrs (struct icmp6_nodeinfo *, struct mbuf *,
161 			  struct ifnet **, char *);
162 static int ni6_store_addrs (struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
163 				struct ifnet *, int);
164 static int icmp6_notify_error (struct mbuf *, int, int, int);
165 
166 static void			icmp6_fasttimo(void *);
167 static void			icmp6_fasttimo_dispatch(netmsg_t);
168 
169 #ifdef COMPAT_RFC1885
170 static struct route_in6 icmp6_reflect_rt;
171 #endif
172 
173 static struct callout		icmp6_fasttimo_ch;
174 static struct netmsg_base	icmp6_fasttimo_nmsg;
175 
176 void
177 icmp6_init(void)
178 {
179 
180 	mld6_init();
181 
182 	callout_init_mp(&icmp6_fasttimo_ch);
183 	netmsg_init(&icmp6_fasttimo_nmsg, NULL, &netisr_adone_rport,
184 	    MSGF_PRIORITY, icmp6_fasttimo_dispatch);
185 
186 	callout_reset_bycpu(&icmp6_fasttimo_ch, ICMP6_FASTTIMO,
187 	    icmp6_fasttimo, NULL, 0);
188 }
189 
190 static void
191 icmp6_errcount(struct icmp6errstat *stat, int type, int code)
192 {
193 	switch (type) {
194 	case ICMP6_DST_UNREACH:
195 		switch (code) {
196 		case ICMP6_DST_UNREACH_NOROUTE:
197 			stat->icp6errs_dst_unreach_noroute++;
198 			return;
199 		case ICMP6_DST_UNREACH_ADMIN:
200 			stat->icp6errs_dst_unreach_admin++;
201 			return;
202 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
203 			stat->icp6errs_dst_unreach_beyondscope++;
204 			return;
205 		case ICMP6_DST_UNREACH_ADDR:
206 			stat->icp6errs_dst_unreach_addr++;
207 			return;
208 		case ICMP6_DST_UNREACH_NOPORT:
209 			stat->icp6errs_dst_unreach_noport++;
210 			return;
211 		}
212 		break;
213 	case ICMP6_PACKET_TOO_BIG:
214 		stat->icp6errs_packet_too_big++;
215 		return;
216 	case ICMP6_TIME_EXCEEDED:
217 		switch (code) {
218 		case ICMP6_TIME_EXCEED_TRANSIT:
219 			stat->icp6errs_time_exceed_transit++;
220 			return;
221 		case ICMP6_TIME_EXCEED_REASSEMBLY:
222 			stat->icp6errs_time_exceed_reassembly++;
223 			return;
224 		}
225 		break;
226 	case ICMP6_PARAM_PROB:
227 		switch (code) {
228 		case ICMP6_PARAMPROB_HEADER:
229 			stat->icp6errs_paramprob_header++;
230 			return;
231 		case ICMP6_PARAMPROB_NEXTHEADER:
232 			stat->icp6errs_paramprob_nextheader++;
233 			return;
234 		case ICMP6_PARAMPROB_OPTION:
235 			stat->icp6errs_paramprob_option++;
236 			return;
237 		}
238 		break;
239 	case ND_REDIRECT:
240 		stat->icp6errs_redirect++;
241 		return;
242 	}
243 	stat->icp6errs_unknown++;
244 }
245 
246 /*
247  * Generate an error packet of type error in response to bad IP6 packet.
248  */
249 void
250 icmp6_error(struct mbuf *m, int type, int code, int param)
251 {
252 	struct ip6_hdr *oip6, *nip6;
253 	struct icmp6_hdr *icmp6;
254 	u_int preplen;
255 	int off;
256 	int nxt;
257 
258 	icmp6stat.icp6s_error++;
259 
260 	/* count per-type-code statistics */
261 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
262 
263 #ifdef M_DECRYPTED	/*not openbsd*/
264 	if (m->m_flags & M_DECRYPTED) {
265 		icmp6stat.icp6s_canterror++;
266 		goto freeit;
267 	}
268 #endif
269 
270 #ifndef PULLDOWN_TEST
271 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
272 #else
273 	if (m->m_len < sizeof(struct ip6_hdr)) {
274 		m = m_pullup(m, sizeof(struct ip6_hdr));
275 		if (m == NULL)
276 			return;
277 	}
278 #endif
279 	oip6 = mtod(m, struct ip6_hdr *);
280 
281 	/*
282 	 * Multicast destination check. For unrecognized option errors,
283 	 * this check has already done in ip6_unknown_opt(), so we can
284 	 * check only for other errors.
285 	 */
286 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
287 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
288 	    (type != ICMP6_PACKET_TOO_BIG &&
289 	     (type != ICMP6_PARAM_PROB ||
290 	      code != ICMP6_PARAMPROB_OPTION)))
291 		goto freeit;
292 
293 	/* Source address check. XXX: the case of anycast source? */
294 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
295 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
296 		goto freeit;
297 
298 	/*
299 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
300 	 * don't do it.
301 	 */
302 	nxt = -1;
303 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
304 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
305 		struct icmp6_hdr *icp;
306 
307 #ifndef PULLDOWN_TEST
308 		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
309 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
310 #else
311 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
312 			sizeof(*icp));
313 		if (icp == NULL) {
314 			icmp6stat.icp6s_tooshort++;
315 			return;
316 		}
317 #endif
318 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
319 		    icp->icmp6_type == ND_REDIRECT) {
320 			/*
321 			 * ICMPv6 error
322 			 * Special case: for redirect (which is
323 			 * informational) we must not send icmp6 error.
324 			 */
325 			icmp6stat.icp6s_canterror++;
326 			goto freeit;
327 		} else {
328 			/* ICMPv6 informational - send the error */
329 		}
330 	} else {
331 		/* non-ICMPv6 - send the error */
332 	}
333 
334 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
335 
336 	/* Finally, do rate limitation check. */
337 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
338 		icmp6stat.icp6s_toofreq++;
339 		goto freeit;
340 	}
341 
342 	/*
343 	 * OK, ICMP6 can be generated.
344 	 */
345 
346 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
347 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
348 
349 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
350 	M_PREPEND(m, preplen, M_NOWAIT);
351 	if (m && m->m_len < preplen)
352 		m = m_pullup(m, preplen);
353 	if (m == NULL) {
354 		nd6log((LOG_DEBUG, "ENOBUFS in icmp6_error %d\n", __LINE__));
355 		return;
356 	}
357 
358 	nip6 = mtod(m, struct ip6_hdr *);
359 	nip6->ip6_src  = oip6->ip6_src;
360 	nip6->ip6_dst  = oip6->ip6_dst;
361 
362 	in6_clearscope(&oip6->ip6_src);
363 	in6_clearscope(&oip6->ip6_dst);
364 
365 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
366 	icmp6->icmp6_type = type;
367 	icmp6->icmp6_code = code;
368 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
369 
370 	/*
371 	 * icmp6_reflect() is designed to be in the input path.
372 	 * icmp6_error() can be called from both input and outut path,
373 	 * and if we are in output path rcvif could contain bogus value.
374 	 * clear m->m_pkthdr.rcvif for safety, we should have enough scope
375 	 * information in ip header (nip6).
376 	 */
377 	m->m_pkthdr.rcvif = NULL;
378 
379 	icmp6stat.icp6s_outhist[type]++;
380 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /* header order: IPv6 - ICMPv6 */
381 
382 	return;
383 
384 freeit:
385 	/*
386 	 * If we can't tell wheter or not we can generate ICMP6, free it.
387 	 */
388 	m_freem(m);
389 }
390 
391 /*
392  * Process a received ICMP6 message.
393  */
394 int
395 icmp6_input(struct mbuf **mp, int *offp, int proto)
396 {
397 	struct mbuf *m = *mp, *n;
398 	struct ip6_hdr *ip6, *nip6;
399 	struct icmp6_hdr *icmp6, *nicmp6;
400 	int off = *offp;
401 	int icmp6len = m->m_pkthdr.len - *offp;
402 	int code, sum, noff;
403 
404 #ifndef PULLDOWN_TEST
405 	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
406 	/* m might change if M_LOOP.  So, call mtod after this */
407 #endif
408 
409 	/*
410 	 * Locate icmp6 structure in mbuf, and check
411 	 * that not corrupted and of at least minimum length
412 	 */
413 
414 	ip6 = mtod(m, struct ip6_hdr *);
415 	if (icmp6len < sizeof(struct icmp6_hdr)) {
416 		icmp6stat.icp6s_tooshort++;
417 		goto freeit;
418 	}
419 
420 	/*
421 	 * calculate the checksum
422 	 */
423 #ifndef PULLDOWN_TEST
424 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
425 #else
426 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
427 	if (icmp6 == NULL) {
428 		icmp6stat.icp6s_tooshort++;
429 		return IPPROTO_DONE;
430 	}
431 #endif
432 	code = icmp6->icmp6_code;
433 
434 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
435 		nd6log((LOG_ERR,
436 		    "ICMP6 checksum error(%d|%x) %s\n",
437 		    icmp6->icmp6_type, sum, ip6_sprintf(&ip6->ip6_src)));
438 		icmp6stat.icp6s_checksum++;
439 		goto freeit;
440 	}
441 
442 	icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
443 	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
444 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
445 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
446 
447 	switch (icmp6->icmp6_type) {
448 	case ICMP6_DST_UNREACH:
449 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
450 		switch (code) {
451 		case ICMP6_DST_UNREACH_NOROUTE:
452 			code = PRC_UNREACH_NET;
453 			break;
454 		case ICMP6_DST_UNREACH_ADMIN:
455 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
456 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
457 			break;
458 		case ICMP6_DST_UNREACH_ADDR:
459 			code = PRC_HOSTDEAD;
460 			break;
461 #ifdef COMPAT_RFC1885
462 		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
463 			code = PRC_UNREACH_SRCFAIL;
464 			break;
465 #else
466 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
467 			/* I mean "source address was incorrect." */
468 			code = PRC_PARAMPROB;
469 			break;
470 #endif
471 		case ICMP6_DST_UNREACH_NOPORT:
472 			code = PRC_UNREACH_PORT;
473 			break;
474 		default:
475 			goto badcode;
476 		}
477 		goto deliver;
478 		break;
479 
480 	case ICMP6_PACKET_TOO_BIG:
481 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
482 		if (code != 0)
483 			goto badcode;
484 
485 		code = PRC_MSGSIZE;
486 
487 		/*
488 		 * Updating the path MTU will be done after examining
489 		 * intermediate extension headers.
490 		 */
491 		goto deliver;
492 		break;
493 
494 	case ICMP6_TIME_EXCEEDED:
495 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
496 		switch (code) {
497 		case ICMP6_TIME_EXCEED_TRANSIT:
498 		case ICMP6_TIME_EXCEED_REASSEMBLY:
499 			code += PRC_TIMXCEED_INTRANS;
500 			break;
501 		default:
502 			goto badcode;
503 		}
504 		goto deliver;
505 		break;
506 
507 	case ICMP6_PARAM_PROB:
508 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
509 		switch (code) {
510 		case ICMP6_PARAMPROB_NEXTHEADER:
511 			code = PRC_UNREACH_PROTOCOL;
512 			break;
513 		case ICMP6_PARAMPROB_HEADER:
514 		case ICMP6_PARAMPROB_OPTION:
515 			code = PRC_PARAMPROB;
516 			break;
517 		default:
518 			goto badcode;
519 		}
520 		goto deliver;
521 		break;
522 
523 	case ICMP6_ECHO_REQUEST:
524 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
525 		if (code != 0)
526 			goto badcode;
527 		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
528 			/* Give up remote */
529 			break;
530 		}
531 		if ((n->m_flags & M_EXT) ||
532 		    n->m_len < off + sizeof(struct icmp6_hdr)) {
533 			struct mbuf *n0 = n;
534 			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
535 			int n0len;
536 
537 			/*
538 			 * Prepare an internal mbuf. m_pullup() doesn't
539 			 * always copy the length we specified.
540 			 */
541 			if (maxlen >= MCLBYTES) {
542 				/* Give up remote */
543 				m_freem(n0);
544 				break;
545 			}
546 			n = m_getb(maxlen, M_NOWAIT, n0->m_type, M_PKTHDR);
547 			if (n == NULL) {
548 				/* Give up remote */
549 				m_freem(n0);
550 				break;
551 			}
552 			n0len = n0->m_pkthdr.len;	/* save for use below */
553 			M_MOVE_PKTHDR(n, n0);
554 			/*
555 			 * Copy IPv6 and ICMPv6 only.
556 			 */
557 			nip6 = mtod(n, struct ip6_hdr *);
558 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
559 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
560 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
561 			noff = sizeof(struct ip6_hdr);
562 			/* new mbuf contains only ipv6+icmpv6 headers */
563 			n->m_len = noff + sizeof(struct icmp6_hdr);
564 			/*
565 			 * Adjust mbuf. ip6_plen will be adjusted in
566 			 * ip6_output().
567 			 */
568 			m_adj(n0, off + sizeof(struct icmp6_hdr));
569 			/* recalculate complete packet size */
570 			n->m_pkthdr.len = n0len + (noff - off);
571 			n->m_next = n0;
572 		} else {
573 			nip6 = mtod(n, struct ip6_hdr *);
574 			nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
575 			noff = off;
576 		}
577 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
578 		nicmp6->icmp6_code = 0;
579 		if (n) {
580 			icmp6stat.icp6s_reflect++;
581 			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
582 			icmp6_reflect(n, noff);
583 		}
584 		break;
585 
586 	case ICMP6_ECHO_REPLY:
587 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
588 		if (code != 0)
589 			goto badcode;
590 		break;
591 
592 	case MLD_LISTENER_QUERY:
593 	case MLD_LISTENER_REPORT:
594 		if (icmp6len < sizeof(struct mld_hdr))
595 			goto badlen;
596 		if (icmp6->icmp6_type == MLD_LISTENER_QUERY) /* XXX: ugly... */
597 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
598 		else
599 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
600 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
601 			/* give up local */
602 			mld6_input(m, off);
603 			m = NULL;
604 			goto freeit;
605 		}
606 		mld6_input(n, off);
607 		/* m stays. */
608 		break;
609 
610 	case MLD_LISTENER_DONE:
611 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
612 		if (icmp6len < sizeof(struct mld_hdr))	/* necessary? */
613 			goto badlen;
614 		break;		/* nothing to be done in kernel */
615 
616 	case MLD_MTRACE_RESP:
617 	case MLD_MTRACE:
618 		/* XXX: these two are experimental.  not officially defind. */
619 		/* XXX: per-interface statistics? */
620 		break;		/* just pass it to applications */
621 
622 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
623 	    {
624 		enum { WRU, FQDN } mode;
625 
626 		if (!icmp6_nodeinfo)
627 			break;
628 
629 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
630 			mode = WRU;
631 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
632 			mode = FQDN;
633 		else
634 			goto badlen;
635 
636 #define hostnamelen	strlen(hostname)
637 		if (mode == FQDN) {
638 #ifndef PULLDOWN_TEST
639 			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
640 					 IPPROTO_DONE);
641 #endif
642 			n = m_copy(m, 0, M_COPYALL);
643 			if (n)
644 				n = ni6_input(n, off);
645 			/* XXX meaningless if n == NULL */
646 			noff = sizeof(struct ip6_hdr);
647 		} else {
648 			u_char *p;
649 			int maxlen, maxhlen;
650 
651 			if ((icmp6_nodeinfo & 5) != 5)
652 				break;
653 
654 			if (code != 0)
655 				goto badcode;
656 			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
657 			if (maxlen >= MCLBYTES) {
658 				/* Give up remote */
659 				break;
660 			}
661 			n = m_getb(maxlen, M_NOWAIT, m->m_type, M_PKTHDR);
662 			if (n == NULL) {
663 				/* Give up remote */
664 				break;
665 			}
666 			if (!m_dup_pkthdr(n, m, M_NOWAIT)) {
667 				/*
668 				 * Previous code did a blind M_COPY_PKTHDR
669 				 * and said "just for rcvif".  If true, then
670 				 * we could tolerate the dup failing (due to
671 				 * the deep copy of the tag chain).  For now
672 				 * be conservative and just fail.
673 				 */
674 				m_free(n);
675 				break;
676 			}
677 			n->m_len = 0;
678 			maxhlen = M_TRAILINGSPACE(n) - maxlen;
679 			if (maxhlen > hostnamelen)
680 				maxhlen = hostnamelen;
681 			/*
682 			 * Copy IPv6 and ICMPv6 only.
683 			 */
684 			nip6 = mtod(n, struct ip6_hdr *);
685 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
686 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
687 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
688 			p = (u_char *)(nicmp6 + 1);
689 			bzero(p, 4);
690 			bcopy(hostname, p + 4, maxhlen); /* meaningless TTL */
691 			noff = sizeof(struct ip6_hdr);
692 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
693 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
694 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
695 			nicmp6->icmp6_code = 0;
696 		}
697 #undef hostnamelen
698 		if (n) {
699 			icmp6stat.icp6s_reflect++;
700 			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
701 			icmp6_reflect(n, noff);
702 		}
703 		break;
704 	    }
705 
706 	case ICMP6_WRUREPLY:
707 		if (code != 0)
708 			goto badcode;
709 		break;
710 
711 	case ND_ROUTER_SOLICIT:
712 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
713 		if (code != 0)
714 			goto badcode;
715 		if (icmp6len < sizeof(struct nd_router_solicit))
716 			goto badlen;
717 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
718 			/* give up local */
719 			nd6_rs_input(m, off, icmp6len);
720 			m = NULL;
721 			goto freeit;
722 		}
723 		nd6_rs_input(n, off, icmp6len);
724 		/* m stays. */
725 		break;
726 
727 	case ND_ROUTER_ADVERT:
728 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
729 		if (code != 0)
730 			goto badcode;
731 		if (icmp6len < sizeof(struct nd_router_advert))
732 			goto badlen;
733 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
734 			/* give up local */
735 			nd6_ra_input(m, off, icmp6len);
736 			m = NULL;
737 			goto freeit;
738 		}
739 		nd6_ra_input(n, off, icmp6len);
740 		/* m stays. */
741 		break;
742 
743 	case ND_NEIGHBOR_SOLICIT:
744 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
745 		if (code != 0)
746 			goto badcode;
747 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
748 			goto badlen;
749 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
750 			/* give up local */
751 			nd6_ns_input(m, off, icmp6len);
752 			m = NULL;
753 			goto freeit;
754 		}
755 		nd6_ns_input(n, off, icmp6len);
756 		/* m stays. */
757 		break;
758 
759 	case ND_NEIGHBOR_ADVERT:
760 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
761 		if (code != 0)
762 			goto badcode;
763 		if (icmp6len < sizeof(struct nd_neighbor_advert))
764 			goto badlen;
765 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
766 			/* give up local */
767 			nd6_na_input(m, off, icmp6len);
768 			m = NULL;
769 			goto freeit;
770 		}
771 		nd6_na_input(n, off, icmp6len);
772 		/* m stays. */
773 		break;
774 
775 	case ND_REDIRECT:
776 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
777 		if (code != 0)
778 			goto badcode;
779 		if (icmp6len < sizeof(struct nd_redirect))
780 			goto badlen;
781 		if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
782 			/* give up local */
783 			icmp6_redirect_input(m, off);
784 			m = NULL;
785 			goto freeit;
786 		}
787 		icmp6_redirect_input(n, off);
788 		/* m stays. */
789 		break;
790 
791 	case ICMP6_ROUTER_RENUMBERING:
792 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
793 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
794 			goto badcode;
795 		if (icmp6len < sizeof(struct icmp6_router_renum))
796 			goto badlen;
797 		break;
798 
799 	default:
800 		nd6log((LOG_DEBUG,
801 		    "icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
802 		    icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
803 		    ip6_sprintf(&ip6->ip6_dst),
804 		    m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0));
805 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
806 			/* ICMPv6 error: MUST deliver it by spec... */
807 			code = PRC_NCMDS;
808 			/* deliver */
809 		} else {
810 			/* ICMPv6 informational: MUST not deliver */
811 			break;
812 		}
813 	deliver:
814 		if (icmp6_notify_error(m, off, icmp6len, code)) {
815 			/* In this case, m should've been freed. */
816 			return (IPPROTO_DONE);
817 		}
818 		break;
819 
820 	badcode:
821 		icmp6stat.icp6s_badcode++;
822 		break;
823 
824 	badlen:
825 		icmp6stat.icp6s_badlen++;
826 		break;
827 	}
828 
829 	/* deliver the packet to appropriate sockets */
830 	icmp6_rip6_input(&m, *offp);
831 
832 	return IPPROTO_DONE;
833 
834 freeit:
835 	m_freem(m);
836 	return IPPROTO_DONE;
837 }
838 
839 static int
840 icmp6_notify_error(struct mbuf *m, int off, int icmp6len, int code)
841 {
842 	struct icmp6_hdr *icmp6;
843 	struct ip6_hdr *eip6;
844 	u_int32_t notifymtu;
845 	struct sockaddr_in6 icmp6src, icmp6dst;
846 
847 	if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
848 		icmp6stat.icp6s_tooshort++;
849 		goto freeit;
850 	}
851 #ifndef PULLDOWN_TEST
852 	IP6_EXTHDR_CHECK(m, off,
853 			 sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr),
854 			 -1);
855 	icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
856 #else
857 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
858 		       sizeof(*icmp6) + sizeof(struct ip6_hdr));
859 	if (icmp6 == NULL) {
860 		icmp6stat.icp6s_tooshort++;
861 		return (-1);
862 	}
863 #endif
864 	eip6 = (struct ip6_hdr *)(icmp6 + 1);
865 
866 	/* Detect the upper level protocol */
867 	{
868 		u_int8_t nxt = eip6->ip6_nxt;
869 		int eoff = off + sizeof(struct icmp6_hdr) +
870 			sizeof(struct ip6_hdr);
871 		struct ip6ctlparam ip6cp;
872 		struct in6_addr *finaldst = NULL;
873 		int icmp6type = icmp6->icmp6_type;
874 		struct ip6_frag *fh;
875 		struct ip6_rthdr *rth;
876 		int rthlen;
877 
878 		while (1) { /* XXX: should avoid infinite loop explicitly? */
879 			struct ip6_ext *eh;
880 
881 			switch (nxt) {
882 			case IPPROTO_HOPOPTS:
883 			case IPPROTO_DSTOPTS:
884 			case IPPROTO_AH:
885 #ifndef PULLDOWN_TEST
886 				IP6_EXTHDR_CHECK(m, 0,
887 				    eoff + sizeof(struct ip6_ext), -1);
888 				eh = (struct ip6_ext *)
889 				    (mtod(m, caddr_t) + eoff);
890 #else
891 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
892 					       eoff, sizeof(*eh));
893 				if (eh == NULL) {
894 					icmp6stat.icp6s_tooshort++;
895 					return (-1);
896 				}
897 #endif
898 
899 				if (nxt == IPPROTO_AH)
900 					eoff += (eh->ip6e_len + 2) << 2;
901 				else
902 					eoff += (eh->ip6e_len + 1) << 3;
903 				nxt = eh->ip6e_nxt;
904 				break;
905 			case IPPROTO_ROUTING:
906 				/*
907 				 * When the erroneous packet contains a
908 				 * routing header, we should examine the
909 				 * header to determine the final destination.
910 				 * Otherwise, we can't properly update
911 				 * information that depends on the final
912 				 * destination (e.g. path MTU).
913 				 */
914 #ifndef PULLDOWN_TEST
915 				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth),
916 						 -1);
917 				rth = (struct ip6_rthdr *)(mtod(m, caddr_t)
918 							   + eoff);
919 #else
920 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
921 					       eoff, sizeof(*rth));
922 				if (rth == NULL) {
923 					icmp6stat.icp6s_tooshort++;
924 					return (-1);
925 				}
926 #endif
927 				rthlen = (rth->ip6r_len + 1) << 3;
928 				eoff += rthlen;
929 				nxt = rth->ip6r_nxt;
930 				break;
931 			case IPPROTO_FRAGMENT:
932 #ifndef PULLDOWN_TEST
933 				IP6_EXTHDR_CHECK(m, 0, eoff +
934 						 sizeof(struct ip6_frag),
935 						 -1);
936 				fh = (struct ip6_frag *)(mtod(m, caddr_t)
937 							 + eoff);
938 #else
939 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
940 					       eoff, sizeof(*fh));
941 				if (fh == NULL) {
942 					icmp6stat.icp6s_tooshort++;
943 					return (-1);
944 				}
945 #endif
946 				/*
947 				 * Data after a fragment header is meaningless
948 				 * unless it is the first fragment, but
949 				 * we'll go to the notify label for path MTU
950 				 * discovery.
951 				 */
952 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
953 					goto notify;
954 
955 				eoff += sizeof(struct ip6_frag);
956 				nxt = fh->ip6f_nxt;
957 				break;
958 			default:
959 				/*
960 				 * This case includes ESP and the No Next
961 				 * Header.  In such cases going to the notify
962 				 * label does not have any meaning
963 				 * (i.e. pr_ctlinput will be NULL), but we go
964 				 * anyway since we might have to update
965 				 * path MTU information.
966 				 */
967 				goto notify;
968 			}
969 		}
970 	  notify:
971 #ifndef PULLDOWN_TEST
972 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
973 #else
974 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
975 			       sizeof(*icmp6) + sizeof(struct ip6_hdr));
976 		if (icmp6 == NULL) {
977 			icmp6stat.icp6s_tooshort++;
978 			return (-1);
979 		}
980 #endif
981 
982 		eip6 = (struct ip6_hdr *)(icmp6 + 1);
983 		bzero(&icmp6dst, sizeof(icmp6dst));
984 		icmp6dst.sin6_len = sizeof(struct sockaddr_in6);
985 		icmp6dst.sin6_family = AF_INET6;
986 		if (finaldst == NULL)
987 			icmp6dst.sin6_addr = eip6->ip6_dst;
988 		else
989 			icmp6dst.sin6_addr = *finaldst;
990 		if (in6_addr2zoneid(m->m_pkthdr.rcvif, &icmp6dst.sin6_addr,
991 		    &icmp6dst.sin6_scope_id))
992 			goto freeit;
993 		if (in6_embedscope(&icmp6dst.sin6_addr, &icmp6dst,
994 				   NULL, NULL)) {
995 			/* should be impossbile */
996 			nd6log((LOG_DEBUG,
997 			    "icmp6_notify_error: in6_embedscope failed\n"));
998 			goto freeit;
999 		}
1000 
1001 		/*
1002 		 * retrieve parameters from the inner IPv6 header, and convert
1003 		 * them into sockaddr structures.
1004 		 */
1005 		bzero(&icmp6src, sizeof(icmp6src));
1006 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
1007 		icmp6src.sin6_family = AF_INET6;
1008 		icmp6src.sin6_addr = eip6->ip6_src;
1009 		if (in6_addr2zoneid(m->m_pkthdr.rcvif, &icmp6src.sin6_addr,
1010 		    &icmp6src.sin6_scope_id))
1011 			goto freeit;
1012 		if (in6_embedscope(&icmp6src.sin6_addr, &icmp6src,
1013 				   NULL, NULL)) {
1014 			/* should be impossbile */
1015 			nd6log((LOG_DEBUG,
1016 			    "icmp6_notify_error: in6_embedscope failed\n"));
1017 			goto freeit;
1018 		}
1019 		icmp6src.sin6_flowinfo =
1020 			(eip6->ip6_flow & IPV6_FLOWLABEL_MASK);
1021 
1022 		if (finaldst == NULL)
1023 			finaldst = &eip6->ip6_dst;
1024 		ip6cp.ip6c_m = m;
1025 		ip6cp.ip6c_icmp6 = icmp6;
1026 		ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
1027 		ip6cp.ip6c_off = eoff;
1028 		ip6cp.ip6c_finaldst = finaldst;
1029 		ip6cp.ip6c_src = &icmp6src;
1030 		ip6cp.ip6c_nxt = nxt;
1031 
1032 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
1033 			notifymtu = ntohl(icmp6->icmp6_mtu);
1034 			ip6cp.ip6c_cmdarg = (void *)&notifymtu;
1035 			icmp6_mtudisc_update(&ip6cp, 1);	/*XXX*/
1036 		}
1037 
1038 		so_pr_ctlinput(
1039 			&inet6sw[ip6_protox[nxt]],
1040 			code, (struct sockaddr *)&icmp6dst, &ip6cp);
1041 	}
1042 	return (0);
1043 
1044 freeit:
1045 	m_freem(m);
1046 	return (-1);
1047 }
1048 
1049 void
1050 icmp6_mtudisc_update(struct ip6ctlparam *ip6cp, int validated)
1051 {
1052 	struct in6_addr *dst = ip6cp->ip6c_finaldst;
1053 	struct icmp6_hdr *icmp6 = ip6cp->ip6c_icmp6;
1054 	struct mbuf *m = ip6cp->ip6c_m;	/* will be necessary for scope issue */
1055 	u_int mtu = ntohl(icmp6->icmp6_mtu);
1056 	struct rtentry *rt = NULL;
1057 	struct sockaddr_in6 sin6;
1058 
1059 	if (!validated)
1060 		return;
1061 
1062 	bzero(&sin6, sizeof(sin6));
1063 	sin6.sin6_family = PF_INET6;
1064 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1065 	sin6.sin6_addr = *dst;
1066 	/* XXX normally, this won't happen */
1067 	if (IN6_IS_ADDR_LINKLOCAL(dst)) {
1068 		sin6.sin6_addr.s6_addr16[1] =
1069 		    htons(m->m_pkthdr.rcvif->if_index);
1070 	}
1071 	/* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
1072 	rt = rtpurelookup((struct sockaddr *)&sin6);
1073 	if (rt != NULL && (rt->rt_flags & RTF_HOST) &&
1074 	    !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
1075 		if (mtu < IPV6_MMTU) {				/* XXX */
1076 			rt->rt_rmx.rmx_locks |= RTV_MTU;
1077 		} else if (mtu < rt->rt_ifp->if_mtu &&
1078 			   rt->rt_rmx.rmx_mtu > mtu) {
1079 			icmp6stat.icp6s_pmtuchg++;
1080 			rt->rt_rmx.rmx_mtu = mtu;
1081 		}
1082 	}
1083 	if (rt)
1084 		--rt->rt_refcnt;
1085 }
1086 
1087 /*
1088  * Process a Node Information Query packet, based on
1089  * draft-ietf-ipngwg-icmp-name-lookups-07.
1090  *
1091  * Spec incompatibilities:
1092  * - IPv6 Subject address handling
1093  * - IPv4 Subject address handling support missing
1094  * - Proxy reply (answer even if it's not for me)
1095  * - joins NI group address at in6_ifattach() time only, does not cope
1096  *   with hostname changes by sethostname(3)
1097  */
1098 #define hostnamelen	strlen(hostname)
1099 static struct mbuf *
1100 ni6_input(struct mbuf *m, int off)
1101 {
1102 	struct icmp6_nodeinfo *ni6, *nni6;
1103 	struct mbuf *n = NULL;
1104 	u_int16_t qtype;
1105 	int subjlen;
1106 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1107 	struct ni_reply_fqdn *fqdn;
1108 	int addrs;		/* for NI_QTYPE_NODEADDR */
1109 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
1110 	struct sockaddr_in6 sin6; /* double meaning; ip6_dst and subjectaddr */
1111 	struct sockaddr_in6 sin6_d; /* XXX: we should retrieve this from m_aux */
1112 	struct ip6_hdr *ip6;
1113 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
1114 	char *subj = NULL;
1115 	struct in6_ifaddr *ia6 = NULL;
1116 
1117 	ip6 = mtod(m, struct ip6_hdr *);
1118 #ifndef PULLDOWN_TEST
1119 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
1120 #else
1121 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
1122 	if (ni6 == NULL) {
1123 		/* m is already reclaimed */
1124 		return NULL;
1125 	}
1126 #endif
1127 
1128 	/*
1129 	 * Validate IPv6 destination address.
1130 	 *
1131 	 * The Responder must discard the Query without further processing
1132 	 * unless it is one of the Responder's unicast or anycast addresses, or
1133 	 * a link-local scope multicast address which the Responder has joined.
1134 	 * [icmp-name-lookups-07, Section 4.]
1135 	 */
1136 	bzero(&sin6, sizeof(sin6));
1137 	sin6.sin6_family = AF_INET6;
1138 	sin6.sin6_len = sizeof(struct sockaddr_in6);
1139 	bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
1140 	/* XXX scopeid */
1141 	if ((ia6 = (struct in6_ifaddr *)ifa_ifwithaddr((struct sockaddr *)&sin6)) != NULL) {
1142 		/* unicast/anycast, fine */
1143 		if ((ia6->ia6_flags & IN6_IFF_TEMPORARY) &&
1144 		    (icmp6_nodeinfo & 4) == 0) {
1145 			nd6log((LOG_DEBUG, "ni6_input: ignore node info to "
1146 				"a temporary address in %s:%d",
1147 			       __FILE__, __LINE__));
1148 			goto bad;
1149 		}
1150 	} else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
1151 		; /* link-local multicast, fine */
1152 	else
1153 		goto bad;
1154 
1155 	/* validate query Subject field. */
1156 	qtype = ntohs(ni6->ni_qtype);
1157 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
1158 	switch (qtype) {
1159 	case NI_QTYPE_NOOP:
1160 	case NI_QTYPE_SUPTYPES:
1161 		/* 07 draft */
1162 		if (ni6->ni_code == ICMP6_NI_SUBJ_FQDN && subjlen == 0)
1163 			break;
1164 		/* FALLTHROUGH */
1165 	case NI_QTYPE_FQDN:
1166 	case NI_QTYPE_NODEADDR:
1167 		switch (ni6->ni_code) {
1168 		case ICMP6_NI_SUBJ_IPV6:
1169 #if ICMP6_NI_SUBJ_IPV6 != 0
1170 		case 0:
1171 #endif
1172 			/*
1173 			 * backward compatibility - try to accept 03 draft
1174 			 * format, where no Subject is present.
1175 			 */
1176 			if (qtype == NI_QTYPE_FQDN && ni6->ni_code == 0 &&
1177 			    subjlen == 0) {
1178 				oldfqdn++;
1179 				break;
1180 			}
1181 #if ICMP6_NI_SUBJ_IPV6 != 0
1182 			if (ni6->ni_code != ICMP6_NI_SUBJ_IPV6)
1183 				goto bad;
1184 #endif
1185 
1186 			if (subjlen != sizeof(sin6.sin6_addr))
1187 				goto bad;
1188 
1189 			/*
1190 			 * Validate Subject address.
1191 			 *
1192 			 * Not sure what exactly "address belongs to the node"
1193 			 * means in the spec, is it just unicast, or what?
1194 			 *
1195 			 * At this moment we consider Subject address as
1196 			 * "belong to the node" if the Subject address equals
1197 			 * to the IPv6 destination address; validation for
1198 			 * IPv6 destination address should have done enough
1199 			 * check for us.
1200 			 *
1201 			 * We do not do proxy at this moment.
1202 			 */
1203 			/* m_pulldown instead of copy? */
1204 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
1205 			    subjlen, (caddr_t)&sin6.sin6_addr);
1206 			if (in6_addr2zoneid(m->m_pkthdr.rcvif,
1207 			    &sin6.sin6_addr, &sin6.sin6_scope_id))
1208 				goto bad;
1209 			in6_embedscope(&sin6.sin6_addr, &sin6, NULL, NULL);
1210 			bzero(&sin6_d, sizeof(sin6_d));
1211 			sin6_d.sin6_family = AF_INET6; /* not used, actually */
1212 			sin6_d.sin6_len = sizeof(sin6_d); /* ditto */
1213 			sin6_d.sin6_addr = ip6->ip6_dst;
1214 			if (in6_addr2zoneid(m->m_pkthdr.rcvif,
1215 			    &ip6->ip6_dst, &sin6_d.sin6_scope_id))
1216 				goto bad;
1217 			in6_embedscope(&sin6_d.sin6_addr, &sin6_d, NULL, NULL);
1218 			subj = (char *)&sin6;
1219 			if (SA6_ARE_ADDR_EQUAL(&sin6, &sin6_d))
1220 				break;
1221 
1222 			/*
1223 			 * XXX if we are to allow other cases, we should really
1224 			 * be careful about scope here.
1225 			 * basically, we should disallow queries toward IPv6
1226 			 * destination X with subject Y, if scope(X) > scope(Y).
1227 			 * if we allow scope(X) > scope(Y), it will result in
1228 			 * information leakage across scope boundary.
1229 			 */
1230 			goto bad;
1231 
1232 		case ICMP6_NI_SUBJ_FQDN:
1233 			/*
1234 			 * Validate Subject name with gethostname(3).
1235 			 *
1236 			 * The behavior may need some debate, since:
1237 			 * - we are not sure if the node has FQDN as
1238 			 *   hostname (returned by gethostname(3)).
1239 			 * - the code does wildcard match for truncated names.
1240 			 *   however, we are not sure if we want to perform
1241 			 *   wildcard match, if gethostname(3) side has
1242 			 *   truncated hostname.
1243 			 */
1244 			n = ni6_nametodns(hostname, hostnamelen, 0);
1245 			if (!n || n->m_next || n->m_len == 0)
1246 				goto bad;
1247 			IP6_EXTHDR_GET(subj, char *, m,
1248 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
1249 			if (subj == NULL)
1250 				goto bad;
1251 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
1252 					n->m_len)) {
1253 				goto bad;
1254 			}
1255 			m_freem(n);
1256 			n = NULL;
1257 			break;
1258 
1259 		case ICMP6_NI_SUBJ_IPV4:	/* XXX: to be implemented? */
1260 		default:
1261 			goto bad;
1262 		}
1263 		break;
1264 	}
1265 
1266 	/* refuse based on configuration.  XXX ICMP6_NI_REFUSED? */
1267 	switch (qtype) {
1268 	case NI_QTYPE_FQDN:
1269 		if ((icmp6_nodeinfo & 1) == 0)
1270 			goto bad;
1271 		break;
1272 	case NI_QTYPE_NODEADDR:
1273 		if ((icmp6_nodeinfo & 2) == 0)
1274 			goto bad;
1275 		break;
1276 	}
1277 
1278 	/* guess reply length */
1279 	switch (qtype) {
1280 	case NI_QTYPE_NOOP:
1281 		break;		/* no reply data */
1282 	case NI_QTYPE_SUPTYPES:
1283 		replylen += sizeof(u_int32_t);
1284 		break;
1285 	case NI_QTYPE_FQDN:
1286 		/* XXX will append an mbuf */
1287 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1288 		break;
1289 	case NI_QTYPE_NODEADDR:
1290 		addrs = ni6_addrs(ni6, m, &ifp, subj);
1291 		if ((replylen += addrs * (sizeof(struct in6_addr) +
1292 					  sizeof(u_int32_t))) > MCLBYTES)
1293 			replylen = MCLBYTES; /* XXX: will truncate pkt later */
1294 		break;
1295 	default:
1296 		/*
1297 		 * XXX: We must return a reply with the ICMP6 code
1298 		 * `unknown Qtype' in this case. However we regard the case
1299 		 * as an FQDN query for backward compatibility.
1300 		 * Older versions set a random value to this field,
1301 		 * so it rarely varies in the defined qtypes.
1302 		 * But the mechanism is not reliable...
1303 		 * maybe we should obsolete older versions.
1304 		 */
1305 		qtype = NI_QTYPE_FQDN;
1306 		/* XXX will append an mbuf */
1307 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
1308 		oldfqdn++;
1309 		break;
1310 	}
1311 
1312 	/* allocate an mbuf to reply. */
1313 	if (replylen > MCLBYTES) {
1314 		/*
1315 		 * XXX: should we try to allocate more? But MCLBYTES
1316 		 * is probably much larger than IPV6_MMTU...
1317 		 */
1318 		goto bad;
1319 	}
1320 	n = m_getb(replylen, M_NOWAIT, m->m_type, M_PKTHDR);
1321 	if (n == NULL) {
1322 		m_freem(m);
1323 		return (NULL);
1324 	}
1325 	M_MOVE_PKTHDR(n, m); /* just for recvif */
1326 	n->m_pkthdr.len = n->m_len = replylen;
1327 
1328 	/* copy mbuf header and IPv6 + Node Information base headers */
1329 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
1330 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
1331 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
1332 
1333 	/* qtype dependent procedure */
1334 	switch (qtype) {
1335 	case NI_QTYPE_NOOP:
1336 		nni6->ni_code = ICMP6_NI_SUCCESS;
1337 		nni6->ni_flags = 0;
1338 		break;
1339 	case NI_QTYPE_SUPTYPES:
1340 	{
1341 		u_int32_t v;
1342 		nni6->ni_code = ICMP6_NI_SUCCESS;
1343 		nni6->ni_flags = htons(0x0000);	/* raw bitmap */
1344 		/* supports NOOP, SUPTYPES, FQDN, and NODEADDR */
1345 		v = (u_int32_t)htonl(0x0000000f);
1346 		bcopy(&v, nni6 + 1, sizeof(u_int32_t));
1347 		break;
1348 	}
1349 	case NI_QTYPE_FQDN:
1350 		nni6->ni_code = ICMP6_NI_SUCCESS;
1351 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
1352 						sizeof(struct ip6_hdr) +
1353 						sizeof(struct icmp6_nodeinfo));
1354 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
1355 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
1356 		/*
1357 		 * XXX do we really have FQDN in variable "hostname"?
1358 		 */
1359 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
1360 		if (n->m_next == NULL)
1361 			goto bad;
1362 		/* XXX we assume that n->m_next is not a chain */
1363 		if (n->m_next->m_next != NULL)
1364 			goto bad;
1365 		n->m_pkthdr.len += n->m_next->m_len;
1366 		break;
1367 	case NI_QTYPE_NODEADDR:
1368 	{
1369 		int lenlim, copied;
1370 
1371 		nni6->ni_code = ICMP6_NI_SUCCESS;
1372 		n->m_pkthdr.len = n->m_len =
1373 		    sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
1374 		lenlim = M_TRAILINGSPACE(n);
1375 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
1376 		/* XXX: reset mbuf length */
1377 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
1378 			sizeof(struct icmp6_nodeinfo) + copied;
1379 		break;
1380 	}
1381 	default:
1382 		break;		/* XXX impossible! */
1383 	}
1384 
1385 	nni6->ni_type = ICMP6_NI_REPLY;
1386 	m_freem(m);
1387 	return (n);
1388 
1389 bad:
1390 	m_freem(m);
1391 	if (n)
1392 		m_freem(n);
1393 	return (NULL);
1394 }
1395 #undef hostnamelen
1396 
1397 /*
1398  * make a mbuf with DNS-encoded string.  no compression support.
1399  *
1400  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
1401  * treated as truncated name (two \0 at the end).  this is a wild guess.
1402  */
1403 static struct mbuf *
1404 ni6_nametodns(const char *name, int namelen,
1405 	      int old) /* return pascal string if non-zero */
1406 {
1407 	struct mbuf *m;
1408 	char *cp, *ep;
1409 	const char *p, *q;
1410 	int i, len, nterm;
1411 
1412 	if (old)
1413 		len = namelen + 1;
1414 	else
1415 		len = MCLBYTES;
1416 
1417 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
1418 	m = m_getb(len, M_NOWAIT, MT_DATA, 0);
1419 	if (!m)
1420 		goto fail;
1421 
1422 	if (old) {
1423 		m->m_len = len;
1424 		*mtod(m, char *) = namelen;
1425 		bcopy(name, mtod(m, char *) + 1, namelen);
1426 		return m;
1427 	} else {
1428 		m->m_len = 0;
1429 		cp = mtod(m, char *);
1430 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
1431 
1432 		/* if not certain about my name, return empty buffer */
1433 		if (namelen == 0)
1434 			return m;
1435 
1436 		/*
1437 		 * guess if it looks like shortened hostname, or FQDN.
1438 		 * shortened hostname needs two trailing "\0".
1439 		 */
1440 		i = 0;
1441 		for (p = name; p < name + namelen; p++) {
1442 			if (*p && *p == '.')
1443 				i++;
1444 		}
1445 		if (i < 2)
1446 			nterm = 2;
1447 		else
1448 			nterm = 1;
1449 
1450 		p = name;
1451 		while (cp < ep && p < name + namelen) {
1452 			i = 0;
1453 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
1454 				i++;
1455 			/* result does not fit into mbuf */
1456 			if (cp + i + 1 >= ep)
1457 				goto fail;
1458 			/*
1459 			 * DNS label length restriction, RFC1035 page 8.
1460 			 * "i == 0" case is included here to avoid returning
1461 			 * 0-length label on "foo..bar".
1462 			 */
1463 			if (i <= 0 || i >= 64)
1464 				goto fail;
1465 			*cp++ = i;
1466 			bcopy(p, cp, i);
1467 			cp += i;
1468 			p = q;
1469 			if (p < name + namelen && *p == '.')
1470 				p++;
1471 		}
1472 		/* termination */
1473 		if (cp + nterm >= ep)
1474 			goto fail;
1475 		while (nterm-- > 0)
1476 			*cp++ = '\0';
1477 		m->m_len = cp - mtod(m, char *);
1478 		return m;
1479 	}
1480 
1481 	panic("should not reach here");
1482 	/* NOTREACHED */
1483 
1484 fail:
1485 	if (m)
1486 		m_freem(m);
1487 	return NULL;
1488 }
1489 
1490 /*
1491  * check if two DNS-encoded string matches.  takes care of truncated
1492  * form (with \0\0 at the end).  no compression support.
1493  * XXX upper/lowercase match (see RFC2065)
1494  */
1495 static int
1496 ni6_dnsmatch(const char *a, int alen, const char *b, int blen)
1497 {
1498 	const char *a0, *b0;
1499 	int l;
1500 
1501 	/* simplest case - need validation? */
1502 	if (alen == blen && bcmp(a, b, alen) == 0)
1503 		return 1;
1504 
1505 	a0 = a;
1506 	b0 = b;
1507 
1508 	/* termination is mandatory */
1509 	if (alen < 2 || blen < 2)
1510 		return 0;
1511 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
1512 		return 0;
1513 	alen--;
1514 	blen--;
1515 
1516 	while (a - a0 < alen && b - b0 < blen) {
1517 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
1518 			return 0;
1519 
1520 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
1521 			return 0;
1522 		/* we don't support compression yet */
1523 		if (a[0] >= 64 || b[0] >= 64)
1524 			return 0;
1525 
1526 		/* truncated case */
1527 		if (a[0] == 0 && a - a0 == alen - 1)
1528 			return 1;
1529 		if (b[0] == 0 && b - b0 == blen - 1)
1530 			return 1;
1531 		if (a[0] == 0 || b[0] == 0)
1532 			return 0;
1533 
1534 		if (a[0] != b[0])
1535 			return 0;
1536 		l = a[0];
1537 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
1538 			return 0;
1539 		if (bcmp(a + 1, b + 1, l) != 0)
1540 			return 0;
1541 
1542 		a += 1 + l;
1543 		b += 1 + l;
1544 	}
1545 
1546 	if (a - a0 == alen && b - b0 == blen)
1547 		return 1;
1548 	else
1549 		return 0;
1550 }
1551 
1552 /*
1553  * calculate the number of addresses to be returned in the node info reply.
1554  */
1555 static int
1556 ni6_addrs(struct icmp6_nodeinfo *ni6, struct mbuf *m, struct ifnet **ifpp,
1557 	  char *subj)
1558 {
1559 	const struct ifnet_array *arr;
1560 	struct in6_ifaddr *ifa6;
1561 	struct sockaddr_in6 *subj_ip6 = NULL; /* XXX pedant */
1562 	int addrs = 0, addrsofif, iffound = 0;
1563 	int niflags = ni6->ni_flags;
1564 	int i;
1565 
1566 	if ((niflags & NI_NODEADDR_FLAG_ALL) == 0) {
1567 		switch (ni6->ni_code) {
1568 		case ICMP6_NI_SUBJ_IPV6:
1569 			if (subj == NULL) /* must be impossible... */
1570 				return (0);
1571 			subj_ip6 = (struct sockaddr_in6 *)subj;
1572 			break;
1573 		default:
1574 			/*
1575 			 * XXX: we only support IPv6 subject address for
1576 			 * this Qtype.
1577 			 */
1578 			return (0);
1579 		}
1580 	}
1581 
1582 	arr = ifnet_array_get();
1583 	for (i = 0; i < arr->ifnet_count; ++i) {
1584 		struct ifnet *ifp = arr->ifnet_arr[i];
1585 		struct ifaddr_container *ifac;
1586 
1587 		addrsofif = 0;
1588 		TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link)
1589 		{
1590 			struct ifaddr *ifa = ifac->ifa;
1591 
1592 			if (ifa->ifa_addr->sa_family != AF_INET6)
1593 				continue;
1594 			ifa6 = (struct in6_ifaddr *)ifa;
1595 
1596 			if ((niflags & NI_NODEADDR_FLAG_ALL) == 0 &&
1597 			    IN6_ARE_ADDR_EQUAL(&subj_ip6->sin6_addr,
1598 					       &ifa6->ia_addr.sin6_addr))
1599 				iffound = 1;
1600 
1601 			/*
1602 			 * IPv4-mapped addresses can only be returned by a
1603 			 * Node Information proxy, since they represent
1604 			 * addresses of IPv4-only nodes, which perforce do
1605 			 * not implement this protocol.
1606 			 * [icmp-name-lookups-07, Section 5.4]
1607 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
1608 			 * this function at this moment.
1609 			 */
1610 
1611 			/* What do we have to do about ::1? */
1612 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1613 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1614 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1615 					continue;
1616 				break;
1617 			case IPV6_ADDR_SCOPE_SITELOCAL:
1618 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1619 					continue;
1620 				break;
1621 			case IPV6_ADDR_SCOPE_GLOBAL:
1622 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1623 					continue;
1624 				break;
1625 			default:
1626 				continue;
1627 			}
1628 
1629 			/*
1630 			 * check if anycast is okay.
1631 			 * XXX: just experimental.  not in the spec.
1632 			 */
1633 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) &&
1634 			    !(niflags & NI_NODEADDR_FLAG_ANYCAST))
1635 				continue; /* we need only unicast addresses */
1636 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) &&
1637 			    (icmp6_nodeinfo & 4) == 0) {
1638 				continue;
1639 			}
1640 			addrsofif++; /* count the address */
1641 		}
1642 		if (iffound) {
1643 			*ifpp = ifp;
1644 			return (addrsofif);
1645 		}
1646 
1647 		addrs += addrsofif;
1648 	}
1649 
1650 	return (addrs);
1651 }
1652 
1653 static int
1654 ni6_store_addrs(struct icmp6_nodeinfo *ni6, struct icmp6_nodeinfo *nni6,
1655 		struct ifnet *ifp0, int resid)
1656 {
1657 	const struct ifnet_array *arr;
1658 	struct in6_ifaddr *ifa6;
1659 	int copied = 0, allow_deprecated = 0;
1660 	u_char *cp = (u_char *)(nni6 + 1);
1661 	int niflags = ni6->ni_flags;
1662 	u_int32_t ltime;
1663 	int idx, idx_dep = -1;
1664 
1665 	if (ifp0 == NULL && !(niflags & NI_NODEADDR_FLAG_ALL))
1666 		return (0);	/* needless to copy */
1667 
1668 	arr = ifnet_array_get();
1669 	if (ifp0 == NULL) {
1670 		idx = 0;
1671 	} else {
1672 		for (idx = 0; idx < arr->ifnet_count; ++idx) {
1673 			if (arr->ifnet_arr[idx] == ifp0)
1674 				break;
1675 		}
1676 	}
1677 
1678 again:
1679 
1680 	for (; idx < arr->ifnet_count; ++idx) {
1681 		struct ifnet *ifp = arr->ifnet_arr[idx];
1682 		struct ifaddr_container *ifac;
1683 
1684 		TAILQ_FOREACH(ifac, &ifp->if_addrheads[mycpuid], ifa_link) {
1685 			struct ifaddr *ifa = ifac->ifa;
1686 
1687 			if (ifa->ifa_addr->sa_family != AF_INET6)
1688 				continue;
1689 			ifa6 = (struct in6_ifaddr *)ifa;
1690 
1691 			if ((ifa6->ia6_flags & IN6_IFF_DEPRECATED) &&
1692 			    allow_deprecated == 0) {
1693 				/*
1694 				 * prefererred address should be put before
1695 				 * deprecated addresses.
1696 				 */
1697 
1698 				/* record the interface for later search */
1699 				if (idx_dep < 0)
1700 					idx_dep = idx;
1701 
1702 				continue;
1703 			}
1704 			else if (!(ifa6->ia6_flags & IN6_IFF_DEPRECATED) &&
1705 				 allow_deprecated != 0)
1706 				continue; /* we now collect deprecated addrs */
1707 
1708 			/* What do we have to do about ::1? */
1709 			switch (in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
1710 			case IPV6_ADDR_SCOPE_LINKLOCAL:
1711 				if ((niflags & NI_NODEADDR_FLAG_LINKLOCAL) == 0)
1712 					continue;
1713 				break;
1714 			case IPV6_ADDR_SCOPE_SITELOCAL:
1715 				if ((niflags & NI_NODEADDR_FLAG_SITELOCAL) == 0)
1716 					continue;
1717 				break;
1718 			case IPV6_ADDR_SCOPE_GLOBAL:
1719 				if ((niflags & NI_NODEADDR_FLAG_GLOBAL) == 0)
1720 					continue;
1721 				break;
1722 			default:
1723 				continue;
1724 			}
1725 
1726 			/*
1727 			 * check if anycast is okay.
1728 			 * XXX: just experimental. not in the spec.
1729 			 */
1730 			if ((ifa6->ia6_flags & IN6_IFF_ANYCAST) &&
1731 			    !(niflags & NI_NODEADDR_FLAG_ANYCAST))
1732 				continue;
1733 			if ((ifa6->ia6_flags & IN6_IFF_TEMPORARY) &&
1734 			    (icmp6_nodeinfo & 4) == 0) {
1735 				continue;
1736 			}
1737 
1738 			/* now we can copy the address */
1739 			if (resid < sizeof(struct in6_addr) +
1740 			    sizeof(u_int32_t)) {
1741 				/*
1742 				 * We give up much more copy.
1743 				 * Set the truncate flag and return.
1744 				 */
1745 				nni6->ni_flags |=
1746 					NI_NODEADDR_FLAG_TRUNCATE;
1747 				return (copied);
1748 			}
1749 
1750 			/*
1751 			 * Set the TTL of the address.
1752 			 * The TTL value should be one of the following
1753 			 * according to the specification:
1754 			 *
1755 			 * 1. The remaining lifetime of a DHCP lease on the
1756 			 *    address, or
1757 			 * 2. The remaining Valid Lifetime of a prefix from
1758 			 *    which the address was derived through Stateless
1759 			 *    Autoconfiguration.
1760 			 *
1761 			 * Note that we currently do not support stateful
1762 			 * address configuration by DHCPv6, so the former
1763 			 * case can't happen.
1764 			 */
1765 			if (ifa6->ia6_lifetime.ia6t_expire == 0)
1766 				ltime = ND6_INFINITE_LIFETIME;
1767 			else {
1768 				if (ifa6->ia6_lifetime.ia6t_expire >
1769 				    time_uptime)
1770 					ltime = htonl(ifa6->ia6_lifetime.ia6t_expire - time_uptime);
1771 				else
1772 					ltime = 0;
1773 			}
1774 
1775 			bcopy(&ltime, cp, sizeof(u_int32_t));
1776 			cp += sizeof(u_int32_t);
1777 
1778 			/* copy the address itself */
1779 			bcopy(&ifa6->ia_addr.sin6_addr, cp,
1780 			      sizeof(struct in6_addr));
1781 			in6_clearscope((struct in6_addr *)cp); /* XXX */
1782 			cp += sizeof(struct in6_addr);
1783 
1784 			resid -= (sizeof(struct in6_addr) + sizeof(u_int32_t));
1785 			copied += (sizeof(struct in6_addr) +
1786 				   sizeof(u_int32_t));
1787 		}
1788 		if (ifp0)	/* we need search only on the specified IF */
1789 			break;
1790 	}
1791 
1792 	if (allow_deprecated == 0 && idx_dep >= 0) {
1793 		idx = idx_dep;
1794 		allow_deprecated = 1;
1795 
1796 		goto again;
1797 	}
1798 
1799 	return (copied);
1800 }
1801 
1802 /*
1803  * XXX almost dup'ed code with rip6_input.
1804  */
1805 static int
1806 icmp6_rip6_input(struct	mbuf **mp, int	off)
1807 {
1808 	struct mbuf *m = *mp;
1809 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1810 	struct in6pcb *in6p;
1811 	struct in6pcb *last = NULL;
1812 	struct sockaddr_in6 fromsa;
1813 	struct icmp6_hdr *icmp6;
1814 	struct mbuf *opts = NULL;
1815 
1816 #ifndef PULLDOWN_TEST
1817 	/* this is assumed to be safe. */
1818 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
1819 #else
1820 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
1821 	if (icmp6 == NULL) {
1822 		/* m is already reclaimed */
1823 		return IPPROTO_DONE;
1824 	}
1825 #endif
1826 
1827 	bzero(&fromsa, sizeof(fromsa));
1828 	fromsa.sin6_len = sizeof(struct sockaddr_in6);
1829 	fromsa.sin6_family = AF_INET6;
1830 	/* KAME hack: recover scopeid */
1831 	in6_recoverscope(&fromsa, &ip6->ip6_src, m->m_pkthdr.rcvif);
1832 
1833 	LIST_FOREACH(in6p, &ripcbinfo.pcblisthead, inp_list)
1834 	{
1835 		if (in6p->inp_flags & INP_PLACEMARKER)
1836 			continue;
1837 		if (!INP_ISIPV6(in6p))
1838 			continue;
1839 		if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
1840 			continue;
1841 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
1842 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
1843 			continue;
1844 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
1845 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
1846 			continue;
1847 		if (in6p->in6p_icmp6filt
1848 		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
1849 				 in6p->in6p_icmp6filt))
1850 			continue;
1851 		if (last) {
1852 			struct socket *so;
1853 			struct mbuf *n;
1854 
1855 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
1856 				if (last->in6p_flags & IN6P_CONTROLOPTS)
1857 					ip6_savecontrol(last, &opts, ip6, n);
1858 				/* strip intermediate headers */
1859 				m_adj(n, off);
1860 				so = last->in6p_socket;
1861 				lwkt_gettoken(&so->so_rcv.ssb_token);
1862 				if (ssb_appendaddr(&so->so_rcv,
1863 				    (struct sockaddr *)&fromsa, n, opts) == 0) {
1864 					m_freem(n);
1865 					if (opts)
1866 						m_freem(opts);
1867 					soroverflow(so);
1868 				} else {
1869 					sorwakeup(so);
1870 				}
1871 				lwkt_reltoken(&so->so_rcv.ssb_token);
1872 				opts = NULL;
1873 			}
1874 		}
1875 		last = in6p;
1876 	}
1877 	if (last) {
1878 		struct socket *so;
1879 
1880 		if (last->in6p_flags & IN6P_CONTROLOPTS)
1881 			ip6_savecontrol(last, &opts, ip6, m);
1882 		/* strip intermediate headers */
1883 		m_adj(m, off);
1884 		so = last->in6p_socket;
1885 		lwkt_gettoken(&so->so_rcv.ssb_token);
1886 		if (ssb_appendaddr(&so->so_rcv, (struct sockaddr *)&fromsa,
1887 		    m, opts) == 0) {
1888 			m_freem(m);
1889 			if (opts)
1890 				m_freem(opts);
1891 			soroverflow(so);
1892 		} else {
1893 			sorwakeup(so);
1894 		}
1895 		lwkt_reltoken(&so->so_rcv.ssb_token);
1896 	} else {
1897 		m_freem(m);
1898 		ip6stat.ip6s_delivered--;
1899 	}
1900 	return IPPROTO_DONE;
1901 }
1902 
1903 /*
1904  * Reflect the ip6 packet back to the source.
1905  * OFF points to the icmp6 header, counted from the top of the mbuf.
1906  */
1907 void
1908 icmp6_reflect(struct mbuf *m, size_t off)
1909 {
1910 	struct ip6_hdr *ip6;
1911 	struct icmp6_hdr *icmp6;
1912 	struct in6_ifaddr *ia;
1913 	struct in6_addr t, *src = NULL;
1914 	int plen;
1915 	int type, code;
1916 	struct ifnet *outif = NULL;
1917 	struct sockaddr_in6 sa6_src, sa6_dst;
1918 #ifdef COMPAT_RFC1885
1919 	int mtu = IPV6_MMTU;
1920 	struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
1921 #endif
1922 
1923 	/* too short to reflect */
1924 	if (off < sizeof(struct ip6_hdr)) {
1925 		nd6log((LOG_DEBUG,
1926 		    "sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
1927 		    (u_long)off, (u_long)sizeof(struct ip6_hdr),
1928 		    __FILE__, __LINE__));
1929 		goto bad;
1930 	}
1931 
1932 	/*
1933 	 * If there are extra headers between IPv6 and ICMPv6, strip
1934 	 * off that header first.
1935 	 */
1936 #ifdef DIAGNOSTIC
1937 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
1938 		panic("assumption failed in icmp6_reflect");
1939 #endif
1940 	if (off > sizeof(struct ip6_hdr)) {
1941 		size_t l;
1942 		struct ip6_hdr nip6;
1943 
1944 		l = off - sizeof(struct ip6_hdr);
1945 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
1946 		m_adj(m, l);
1947 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1948 		if (m->m_len < l) {
1949 			if ((m = m_pullup(m, l)) == NULL)
1950 				return;
1951 		}
1952 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
1953 	} else /* off == sizeof(struct ip6_hdr) */ {
1954 		size_t l;
1955 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
1956 		if (m->m_len < l) {
1957 			if ((m = m_pullup(m, l)) == NULL)
1958 				return;
1959 		}
1960 	}
1961 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
1962 	ip6 = mtod(m, struct ip6_hdr *);
1963 	ip6->ip6_nxt = IPPROTO_ICMPV6;
1964 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
1965 	type = icmp6->icmp6_type; /* keep type for statistics */
1966 	code = icmp6->icmp6_code; /* ditto. */
1967 
1968 	t = ip6->ip6_dst;
1969 	/*
1970 	 * ip6_input() drops a packet if its src is multicast.
1971 	 * So, the src is never multicast.
1972 	 */
1973 	ip6->ip6_dst = ip6->ip6_src;
1974 
1975 	/*
1976 	 * XXX: make sure to embed scope zone information, using
1977 	 * already embedded IDs or the received interface (if any).
1978 	 * Note that rcvif may be NULL.
1979 	 * TODO: scoped routing case (XXX).
1980 	 */
1981 	bzero(&sa6_src, sizeof(sa6_src));
1982 	sa6_src.sin6_family = AF_INET6;
1983 	sa6_src.sin6_len = sizeof(sa6_src);
1984 	sa6_src.sin6_addr = ip6->ip6_dst;
1985 	in6_recoverscope(&sa6_src, &ip6->ip6_dst, m->m_pkthdr.rcvif);
1986 	in6_embedscope(&ip6->ip6_dst, &sa6_src, NULL, NULL);
1987 	bzero(&sa6_dst, sizeof(sa6_dst));
1988 	sa6_dst.sin6_family = AF_INET6;
1989 	sa6_dst.sin6_len = sizeof(sa6_dst);
1990 	sa6_dst.sin6_addr = t;
1991 	in6_recoverscope(&sa6_dst, &t, m->m_pkthdr.rcvif);
1992 	in6_embedscope(&t, &sa6_dst, NULL, NULL);
1993 
1994 #ifdef COMPAT_RFC1885
1995 	/*
1996 	 * xxx guess MTU
1997 	 * RFC 1885 requires that echo reply should be truncated if it
1998 	 * does not fit in with (return) path MTU, but the description was
1999 	 * removed in the new spec.
2000 	 */
2001 	if (icmp6_reflect_rt.ro_rt == 0 ||
2002 	    ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
2003 		if (icmp6_reflect_rt.ro_rt) {
2004 			RTFREE(icmp6_reflect_rt.ro_rt);
2005 			icmp6_reflect_rt.ro_rt = 0;
2006 		}
2007 		bzero(sin6, sizeof(*sin6));
2008 		sin6->sin6_family = PF_INET6;
2009 		sin6->sin6_len = sizeof(struct sockaddr_in6);
2010 		sin6->sin6_addr = ip6->ip6_dst;
2011 
2012 		rtalloc_ign((struct route *)&icmp6_reflect_rt.ro_rt,
2013 			    RTF_PRCLONING);
2014 	}
2015 
2016 	if (icmp6_reflect_rt.ro_rt == 0)
2017 		goto bad;
2018 
2019 	if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
2020 	    && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
2021 		mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
2022 
2023 	if (mtu < m->m_pkthdr.len) {
2024 		plen -= (m->m_pkthdr.len - mtu);
2025 		m_adj(m, mtu - m->m_pkthdr.len);
2026 	}
2027 #endif
2028 	/*
2029 	 * If the incoming packet was addressed directly to us(i.e. unicast),
2030 	 * use dst as the src for the reply.
2031 	 * The IN6_IFF_NOTREADY case would be VERY rare, but is possible
2032 	 * (for example) when we encounter an error while forwarding procedure
2033 	 * destined to a duplicated address of ours.
2034 	 */
2035 	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
2036 		if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
2037 		    (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
2038 			src = &t;
2039 			break;
2040 		}
2041 	if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
2042 		/*
2043 		 * This is the case if the dst is our link-local address
2044 		 * and the sender is also ourselves.
2045 		 */
2046 		src = &t;
2047 	}
2048 
2049 	if (src == NULL) {
2050 		int e;
2051 		struct route_in6 ro;
2052 
2053 		/*
2054 		 * This case matches to multicasts, our anycast, or unicasts
2055 		 * that we do not own.  Select a source address based on the
2056 		 * source address of the erroneous packet.
2057 		 */
2058 		bzero(&ro, sizeof(ro));
2059 		src = in6_selectsrc(&sa6_src, NULL, NULL, &ro, NULL, &e, NULL);
2060 		if (ro.ro_rt)
2061 			RTFREE(ro.ro_rt); /* XXX: we could use this */
2062 		if (src == NULL) {
2063 			nd6log((LOG_DEBUG,
2064 			    "icmp6_reflect: source can't be determined: "
2065 			    "dst=%s, error=%d\n",
2066 			    ip6_sprintf(&sa6_src.sin6_addr), e));
2067 			goto bad;
2068 		}
2069 	}
2070 
2071 	ip6->ip6_src = *src;
2072 
2073 	ip6->ip6_flow = 0;
2074 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2075 	ip6->ip6_vfc |= IPV6_VERSION;
2076 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2077 	if (m->m_pkthdr.rcvif) {
2078 		/* XXX: This may not be the outgoing interface */
2079 		ip6->ip6_hlim = ND_IFINFO(m->m_pkthdr.rcvif)->chlim;
2080 		if (ip6->ip6_hlim < ip6_minhlim)
2081 			ip6->ip6_hlim = ip6_minhlim;
2082 	} else {
2083 		ip6->ip6_hlim = ip6_defhlim;
2084 	}
2085 
2086 	icmp6->icmp6_cksum = 0;
2087 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
2088 					sizeof(struct ip6_hdr), plen);
2089 
2090 	/*
2091 	 * XXX option handling
2092 	 */
2093 
2094 	m->m_flags &= ~(M_BCAST|M_MCAST);
2095 
2096 #ifdef COMPAT_RFC1885
2097 	ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif, NULL);
2098 #else
2099 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2100 #endif
2101 	if (outif)
2102 		icmp6_ifoutstat_inc(outif, type, code);
2103 
2104 	return;
2105 
2106 bad:
2107 	m_freem(m);
2108 	return;
2109 }
2110 
2111 static void
2112 icmp6_fasttimo_dispatch(netmsg_t nmsg)
2113 {
2114 
2115 	ASSERT_NETISR0;
2116 
2117 	/* Reply ASAP. */
2118 	crit_enter();
2119 	netisr_replymsg(&nmsg->base, 0);
2120 	crit_exit();
2121 
2122 	mld6_fasttimeo();
2123 
2124 	callout_reset(&icmp6_fasttimo_ch, ICMP6_FASTTIMO, icmp6_fasttimo, NULL);
2125 }
2126 
2127 static void
2128 icmp6_fasttimo(void *dummy __unused)
2129 {
2130 	struct netmsg_base *nmsg = &icmp6_fasttimo_nmsg;
2131 
2132 	KKASSERT(mycpuid == 0);
2133 
2134 	crit_enter();
2135 	if (nmsg->lmsg.ms_flags & MSGF_DONE)
2136 		netisr_sendmsg_oncpu(nmsg);
2137 	crit_exit();
2138 }
2139 
2140 static const char *
2141 icmp6_redirect_diag(struct in6_addr *src6, struct in6_addr *dst6,
2142 		    struct in6_addr *tgt6)
2143 {
2144 	static char buf[1024];
2145 	ksnprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
2146 		ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
2147 	return buf;
2148 }
2149 
2150 void
2151 icmp6_redirect_input(struct mbuf *m, int off)
2152 {
2153 	struct ifnet *ifp = m->m_pkthdr.rcvif;
2154 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2155 	struct nd_redirect *nd_rd;
2156 	int icmp6len = ntohs(ip6->ip6_plen);
2157 	char *lladdr = NULL;
2158 	int lladdrlen = 0;
2159 #if 0
2160 	u_char *redirhdr = NULL;
2161 	int redirhdrlen = 0;
2162 #endif
2163 	struct rtentry *rt = NULL;
2164 	int is_router;
2165 	int is_onlink;
2166 	struct in6_addr src6 = ip6->ip6_src;
2167 	struct in6_addr redtgt6;
2168 	struct in6_addr reddst6;
2169 	union nd_opts ndopts;
2170 
2171 	if (!m || !ifp)
2172 		return;
2173 
2174 	/* XXX if we are router, we don't update route by icmp6 redirect */
2175 	if (ip6_forwarding)
2176 		goto freeit;
2177 	if (!icmp6_rediraccept)
2178 		goto freeit;
2179 
2180 #ifndef PULLDOWN_TEST
2181 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
2182 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
2183 #else
2184 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
2185 	if (nd_rd == NULL) {
2186 		icmp6stat.icp6s_tooshort++;
2187 		return;
2188 	}
2189 #endif
2190 	redtgt6 = nd_rd->nd_rd_target;
2191 	reddst6 = nd_rd->nd_rd_dst;
2192 
2193 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2194 		redtgt6.s6_addr16[1] = htons(ifp->if_index);
2195 	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
2196 		reddst6.s6_addr16[1] = htons(ifp->if_index);
2197 
2198 	/* validation */
2199 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
2200 		nd6log((LOG_ERR,
2201 			"ICMP6 redirect sent from %s rejected; "
2202 			"must be from linklocal\n", ip6_sprintf(&src6)));
2203 		goto bad;
2204 	}
2205 	if (ip6->ip6_hlim != 255) {
2206 		nd6log((LOG_ERR,
2207 			"ICMP6 redirect sent from %s rejected; "
2208 			"hlim=%d (must be 255)\n",
2209 			ip6_sprintf(&src6), ip6->ip6_hlim));
2210 		goto bad;
2211 	}
2212     {
2213 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
2214 	struct sockaddr_in6 sin6;
2215 	struct in6_addr *gw6;
2216 
2217 	bzero(&sin6, sizeof(sin6));
2218 	sin6.sin6_family = AF_INET6;
2219 	sin6.sin6_len = sizeof(struct sockaddr_in6);
2220 	bcopy(&reddst6, &sin6.sin6_addr, sizeof reddst6);
2221 	rt = rtpurelookup((struct sockaddr *)&sin6);
2222 	if (rt != NULL) {
2223 		if (rt->rt_gateway == NULL ||
2224 		    rt->rt_gateway->sa_family != AF_INET6) {
2225 			nd6log((LOG_ERR,
2226 			    "ICMP6 redirect rejected; no route "
2227 			    "with inet6 gateway found for redirect dst: %s\n",
2228 			    icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2229 			--rt->rt_refcnt;
2230 			goto bad;
2231 		}
2232 
2233 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
2234 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
2235 			nd6log((LOG_ERR,
2236 				"ICMP6 redirect rejected; "
2237 				"not equal to gw-for-src=%s (must be same): "
2238 				"%s\n",
2239 				ip6_sprintf(gw6),
2240 				icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2241 			--rt->rt_refcnt;
2242 			goto bad;
2243 		}
2244 	} else {
2245 		nd6log((LOG_ERR,
2246 			"ICMP6 redirect rejected; "
2247 			"no route found for redirect dst: %s\n",
2248 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2249 		goto bad;
2250 	}
2251 	--rt->rt_refcnt;
2252 	rt = NULL;
2253     }
2254 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
2255 		nd6log((LOG_ERR,
2256 			"ICMP6 redirect rejected; "
2257 			"redirect dst must be unicast: %s\n",
2258 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2259 		goto bad;
2260 	}
2261 
2262 	is_router = is_onlink = 0;
2263 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
2264 		is_router = 1;	/* router case */
2265 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
2266 		is_onlink = 1;	/* on-link destination case */
2267 	if (!is_router && !is_onlink) {
2268 		nd6log((LOG_ERR,
2269 			"ICMP6 redirect rejected; "
2270 			"neither router case nor onlink case: %s\n",
2271 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2272 		goto bad;
2273 	}
2274 	/* validation passed */
2275 
2276 	icmp6len -= sizeof(*nd_rd);
2277 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
2278 	if (nd6_options(&ndopts) < 0) {
2279 		nd6log((LOG_INFO, "icmp6_redirect_input: "
2280 			"invalid ND option, rejected: %s\n",
2281 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2282 		/* nd6_options have incremented stats */
2283 		goto freeit;
2284 	}
2285 
2286 	if (ndopts.nd_opts_tgt_lladdr) {
2287 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
2288 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
2289 	}
2290 
2291 #if 0
2292 	if (ndopts.nd_opts_rh) {
2293 		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
2294 		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
2295 	}
2296 #endif
2297 
2298 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
2299 		nd6log((LOG_INFO,
2300 			"icmp6_redirect_input: lladdrlen mismatch for %s "
2301 			"(if %d, icmp6 packet %d): %s\n",
2302 			ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
2303 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6)));
2304 		goto bad;
2305 	}
2306 
2307 	/* RFC 2461 8.3 */
2308 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
2309 			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
2310 
2311 	if (!is_onlink) {	/* better router case.  perform rtredirect. */
2312 		/* perform rtredirect */
2313 		struct sockaddr_in6 sdst;
2314 		struct sockaddr_in6 sgw;
2315 		struct sockaddr_in6 ssrc;
2316 
2317 		bzero(&sdst, sizeof(sdst));
2318 		bzero(&sgw, sizeof(sgw));
2319 		bzero(&ssrc, sizeof(ssrc));
2320 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
2321 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
2322 			sizeof(struct sockaddr_in6);
2323 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
2324 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2325 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
2326 		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
2327 			   NULL, RTF_GATEWAY | RTF_HOST,
2328 			   (struct sockaddr *)&ssrc);
2329 	}
2330 	/* finally update cached route in each socket via kpfctlinput */
2331     {
2332 	struct sockaddr_in6 sdst;
2333 
2334 	bzero(&sdst, sizeof(sdst));
2335 	sdst.sin6_family = AF_INET6;
2336 	sdst.sin6_len = sizeof(struct sockaddr_in6);
2337 	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
2338 	kpfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
2339     }
2340 
2341 freeit:
2342 	m_freem(m);
2343 	return;
2344 
2345 bad:
2346 	icmp6stat.icp6s_badredirect++;
2347 	m_freem(m);
2348 }
2349 
2350 void
2351 icmp6_redirect_output(struct mbuf *m0, struct rtentry *rt)
2352 {
2353 	struct ifnet *ifp;	/* my outgoing interface */
2354 	struct in6_addr *ifp_ll6;
2355 	struct in6_addr *router_ll6;
2356 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
2357 	struct mbuf *m = NULL;	/* newly allocated one */
2358 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
2359 	struct nd_redirect *nd_rd;
2360 	size_t maxlen;
2361 	u_char *p;
2362 	struct ifnet *outif = NULL;
2363 	struct sockaddr_in6 src_sa;
2364 
2365 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
2366 
2367 	/* if we are not router, we don't send icmp6 redirect */
2368 	if (!ip6_forwarding || ip6_accept_rtadv)
2369 		goto fail;
2370 
2371 	/* sanity check */
2372 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
2373 		goto fail;
2374 
2375 	/*
2376 	 * Address check:
2377 	 *  the source address must identify a neighbor, and
2378 	 *  the destination address must not be a multicast address
2379 	 *  [RFC 2461, sec 8.2]
2380 	 */
2381 	sip6 = mtod(m0, struct ip6_hdr *);
2382 	bzero(&src_sa, sizeof(src_sa));
2383 	src_sa.sin6_family = AF_INET6;
2384 	src_sa.sin6_len = sizeof(src_sa);
2385 	src_sa.sin6_addr = sip6->ip6_src;
2386 	/* we don't currently use sin6_scope_id, but eventually use it */
2387 	if (in6_addr2zoneid(ifp, &sip6->ip6_src, &src_sa.sin6_scope_id))
2388 		goto fail;
2389 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
2390 		goto fail;
2391 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
2392 		goto fail;	/* what should we do here? */
2393 
2394 	/* rate limit */
2395 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
2396 		goto fail;
2397 
2398 	/*
2399 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
2400 	 * we almost always ask for an mbuf cluster for simplicity.
2401 	 * (MHLEN < IPV6_MMTU is almost always true)
2402 	 */
2403 #if IPV6_MMTU >= MCLBYTES
2404 # error assumption failed about IPV6_MMTU and MCLBYTES
2405 #endif
2406 	m = m_getb(IPV6_MMTU, M_NOWAIT, MT_HEADER, M_PKTHDR);
2407 	if (!m)
2408 		goto fail;
2409 	m->m_len = 0;
2410 	maxlen = M_TRAILINGSPACE(m);
2411 	maxlen = min(IPV6_MMTU, maxlen);
2412 	/* just for safety */
2413 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
2414 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
2415 		goto fail;
2416 	}
2417 
2418 	{
2419 		/* get ip6 linklocal address for ifp(my outgoing interface). */
2420 		struct in6_ifaddr *ia;
2421 		if ((ia = in6ifa_ifpforlinklocal(ifp,
2422 						 IN6_IFF_NOTREADY|
2423 						 IN6_IFF_ANYCAST)) == NULL)
2424 			goto fail;
2425 		ifp_ll6 = &ia->ia_addr.sin6_addr;
2426 	}
2427 
2428 	/* get ip6 linklocal address for the router. */
2429 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
2430 		struct sockaddr_in6 *sin6;
2431 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
2432 		router_ll6 = &sin6->sin6_addr;
2433 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
2434 			router_ll6 = NULL;
2435 	} else
2436 		router_ll6 = NULL;
2437 
2438 	/* ip6 */
2439 	ip6 = mtod(m, struct ip6_hdr *);
2440 	ip6->ip6_flow = 0;
2441 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
2442 	ip6->ip6_vfc |= IPV6_VERSION;
2443 	/* ip6->ip6_plen will be set later */
2444 	ip6->ip6_nxt = IPPROTO_ICMPV6;
2445 	ip6->ip6_hlim = 255;
2446 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
2447 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
2448 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
2449 
2450 	/* ND Redirect */
2451 	nd_rd = (struct nd_redirect *)(ip6 + 1);
2452 	nd_rd->nd_rd_type = ND_REDIRECT;
2453 	nd_rd->nd_rd_code = 0;
2454 	nd_rd->nd_rd_reserved = 0;
2455 	if (rt->rt_flags & RTF_GATEWAY) {
2456 		/*
2457 		 * nd_rd->nd_rd_target must be a link-local address in
2458 		 * better router cases.
2459 		 */
2460 		if (!router_ll6)
2461 			goto fail;
2462 		bcopy(router_ll6, &nd_rd->nd_rd_target,
2463 		      sizeof(nd_rd->nd_rd_target));
2464 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2465 		      sizeof(nd_rd->nd_rd_dst));
2466 	} else {
2467 		/* make sure redtgt == reddst */
2468 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
2469 		      sizeof(nd_rd->nd_rd_target));
2470 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
2471 		      sizeof(nd_rd->nd_rd_dst));
2472 	}
2473 
2474 	p = (u_char *)(nd_rd + 1);
2475 
2476 	if (!router_ll6)
2477 		goto nolladdropt;
2478 
2479 	{
2480 		/* target lladdr option */
2481 		struct rtentry *rt_router = NULL;
2482 		int len;
2483 		struct sockaddr_dl *sdl;
2484 		struct nd_opt_hdr *nd_opt;
2485 		char *lladdr;
2486 
2487 		rt_router = nd6_lookup(router_ll6, 0, ifp);
2488 		if (!rt_router)
2489 			goto nolladdropt;
2490 		len = sizeof(*nd_opt) + ifp->if_addrlen;
2491 		len = (len + 7) & ~7;	/* round by 8 */
2492 		/* safety check */
2493 		if (len + (p - (u_char *)ip6) > maxlen)
2494 			goto nolladdropt;
2495 		if (!(rt_router->rt_flags & RTF_GATEWAY) &&
2496 		    (rt_router->rt_flags & RTF_LLINFO) &&
2497 		    (rt_router->rt_gateway->sa_family == AF_LINK) &&
2498 		    (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
2499 		    sdl->sdl_alen) {
2500 			nd_opt = (struct nd_opt_hdr *)p;
2501 			nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
2502 			nd_opt->nd_opt_len = len >> 3;
2503 			lladdr = (char *)(nd_opt + 1);
2504 			bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
2505 			p += len;
2506 		}
2507 	}
2508 nolladdropt:;
2509 
2510 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2511 
2512 	/* just to be safe */
2513 #ifdef M_DECRYPTED	/*not openbsd*/
2514 	if (m0->m_flags & M_DECRYPTED)
2515 		goto noredhdropt;
2516 #endif
2517 	if (p - (u_char *)ip6 > maxlen)
2518 		goto noredhdropt;
2519 
2520     {
2521 	/* redirected header option */
2522 	int len;
2523 	struct nd_opt_rd_hdr *nd_opt_rh;
2524 
2525 	/*
2526 	 * compute the maximum size for icmp6 redirect header option.
2527 	 * XXX room for auth header?
2528 	 */
2529 	len = maxlen - (p - (u_char *)ip6);
2530 	len &= ~7;
2531 
2532 	/* This is just for simplicity. */
2533 	if (m0->m_pkthdr.len != m0->m_len) {
2534 		if (m0->m_next) {
2535 			m_freem(m0->m_next);
2536 			m0->m_next = NULL;
2537 		}
2538 		m0->m_pkthdr.len = m0->m_len;
2539 	}
2540 
2541 	/*
2542 	 * Redirected header option spec (RFC2461 4.6.3) talks nothing
2543 	 * about padding/truncate rule for the original IP packet.
2544 	 * From the discussion on IPv6imp in Feb 1999, the consensus was:
2545 	 * - "attach as much as possible" is the goal
2546 	 * - pad if not aligned (original size can be guessed by original
2547 	 *   ip6 header)
2548 	 * Following code adds the padding if it is simple enough,
2549 	 * and truncates if not.
2550 	 */
2551 	if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
2552 		panic("assumption failed in %s:%d", __FILE__, __LINE__);
2553 
2554 	if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
2555 		/* not enough room, truncate */
2556 		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2557 	} else {
2558 		/* enough room, pad or truncate */
2559 		size_t extra;
2560 
2561 		extra = m0->m_pkthdr.len % 8;
2562 		if (extra) {
2563 			/* pad if easy enough, truncate if not */
2564 			if (8 - extra <= M_TRAILINGSPACE(m0)) {
2565 				/* pad */
2566 				m0->m_len += (8 - extra);
2567 				m0->m_pkthdr.len += (8 - extra);
2568 			} else {
2569 				/* truncate */
2570 				m0->m_pkthdr.len -= extra;
2571 				m0->m_len -= extra;
2572 			}
2573 		}
2574 		len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
2575 		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
2576 	}
2577 
2578 	nd_opt_rh = (struct nd_opt_rd_hdr *)p;
2579 	bzero(nd_opt_rh, sizeof(*nd_opt_rh));
2580 	nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
2581 	nd_opt_rh->nd_opt_rh_len = len >> 3;
2582 	p += sizeof(*nd_opt_rh);
2583 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
2584 
2585 	/* connect m0 to m */
2586 	m->m_next = m0;
2587 	m->m_pkthdr.len = m->m_len + m0->m_len;
2588 	m0 = NULL;
2589     }
2590 noredhdropt:;
2591 	if (m0 != NULL) {
2592 		m_freem(m0);
2593 		m0 = NULL;
2594 	}
2595 
2596 	/* XXX: clear embedded link IDs in the inner header */
2597 	in6_clearscope(&sip6->ip6_src);
2598 	in6_clearscope(&sip6->ip6_dst);
2599 	in6_clearscope(&nd_rd->nd_rd_target);
2600 	in6_clearscope(&nd_rd->nd_rd_dst);
2601 
2602 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
2603 
2604 	nd_rd->nd_rd_cksum = 0;
2605 	nd_rd->nd_rd_cksum
2606 		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
2607 
2608 	/* send the packet to outside... */
2609 	ip6_output(m, NULL, NULL, 0, NULL, &outif, NULL);
2610 	if (outif) {
2611 		icmp6_ifstat_inc(outif, ifs6_out_msg);
2612 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
2613 	}
2614 	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
2615 
2616 	return;
2617 
2618 fail:
2619 	if (m)
2620 		m_freem(m);
2621 	if (m0)
2622 		m_freem(m0);
2623 }
2624 
2625 #ifdef HAVE_NRL_INPCB
2626 #define in6pcb		inpcb
2627 #define in6p_icmp6filt	inp_icmp6filt
2628 #endif
2629 /*
2630  * ICMPv6 socket option processing.
2631  */
2632 void
2633 icmp6_ctloutput(netmsg_t msg)
2634 {
2635 	struct socket *so = msg->ctloutput.base.nm_so;
2636 	struct sockopt *sopt = msg->ctloutput.nm_sopt;
2637 	int error = 0;
2638 	int optlen;
2639 	struct inpcb *inp = so->so_pcb;
2640 	int level, op, optname;
2641 
2642 	if (sopt) {
2643 		level = sopt->sopt_level;
2644 		op = sopt->sopt_dir;
2645 		optname = sopt->sopt_name;
2646 		optlen = sopt->sopt_valsize;
2647 	} else
2648 		level = op = optname = optlen = 0;
2649 
2650 	if (level != IPPROTO_ICMPV6) {
2651 		error = EINVAL;
2652 		goto out;
2653 	}
2654 
2655 	switch (op) {
2656 	case PRCO_SETOPT:
2657 		switch (optname) {
2658 		case ICMP6_FILTER:
2659 		    {
2660 			if (optlen != sizeof(struct icmp6_filter)) {
2661 				error = EMSGSIZE;
2662 				break;
2663 			}
2664 			if (inp->in6p_icmp6filt == NULL) {
2665 				error = EINVAL;
2666 				break;
2667 			}
2668 			error = soopt_to_kbuf(sopt, inp->in6p_icmp6filt, optlen,
2669 				optlen);
2670 			break;
2671 		    }
2672 
2673 		default:
2674 			error = ENOPROTOOPT;
2675 			break;
2676 		}
2677 		break;
2678 
2679 	case PRCO_GETOPT:
2680 		switch (optname) {
2681 		case ICMP6_FILTER:
2682 		    {
2683 			if (inp->in6p_icmp6filt == NULL) {
2684 				error = EINVAL;
2685 				break;
2686 			}
2687 			soopt_from_kbuf(sopt, inp->in6p_icmp6filt,
2688 				sizeof(struct icmp6_filter));
2689 			break;
2690 		    }
2691 
2692 		default:
2693 			error = ENOPROTOOPT;
2694 			break;
2695 		}
2696 		break;
2697 	}
2698 out:
2699 	lwkt_replymsg(&msg->ctloutput.base.lmsg, error);
2700 }
2701 #ifdef HAVE_NRL_INPCB
2702 #undef in6pcb
2703 #undef in6p_icmp6filt
2704 #endif
2705 
2706 #ifndef HAVE_PPSRATECHECK
2707 #ifndef timersub
2708 #define	timersub(tvp, uvp, vvp)						\
2709 	do {								\
2710 		(vvp)->tv_sec = (tvp)->tv_sec - (uvp)->tv_sec;		\
2711 		(vvp)->tv_usec = (tvp)->tv_usec - (uvp)->tv_usec;	\
2712 		if ((vvp)->tv_usec < 0) {				\
2713 			(vvp)->tv_sec--;				\
2714 			(vvp)->tv_usec += 1000000;			\
2715 		}							\
2716 	} while (0)
2717 #endif
2718 
2719 /*
2720  * ppsratecheck(): packets (or events) per second limitation.
2721  */
2722 static int
2723 ppsratecheck(struct timeval *lasttime, int *curpps,
2724 	     int maxpps)	/* maximum pps allowed */
2725 {
2726 	struct timeval tv, delta;
2727 	int rv;
2728 
2729 	microtime(&tv);
2730 
2731 	timersub(&tv, lasttime, &delta);
2732 
2733 	/*
2734 	 * Check for 0,0 so that the message will be seen at least once.
2735 	 * If more than one second has passed since the last update of
2736 	 * lasttime, reset the counter.
2737 	 *
2738 	 * We do increment *curpps even in *curpps < maxpps case, as some may
2739 	 * try to use *curpps for stat purposes as well.
2740 	 */
2741 	if ((lasttime->tv_sec == 0 && lasttime->tv_usec == 0) ||
2742 	    delta.tv_sec >= 1) {
2743 		*lasttime = tv;
2744 		*curpps = 0;
2745 		rv = 1;
2746 	} else if (maxpps < 0)
2747 		rv = 1;
2748 	else if (*curpps < maxpps)
2749 		rv = 1;
2750 	else
2751 		rv = 0;
2752 
2753 #if 1 /* DIAGNOSTIC? */
2754 	/* be careful about wrap-around */
2755 	if (*curpps + 1 > *curpps)
2756 		*curpps = *curpps + 1;
2757 #else
2758 	/*
2759 	 * assume that there's not too many calls to this function.
2760 	 * not sure if the assumption holds, as it depends on *caller's*
2761 	 * behavior, not the behavior of this function.
2762 	 * IMHO it is wrong to make assumption on the caller's behavior,
2763 	 * so the above #if is #if 1, not #ifdef DIAGNOSTIC.
2764 	 */
2765 	*curpps = *curpps + 1;
2766 #endif
2767 
2768 	return (rv);
2769 }
2770 #endif
2771 
2772 /*
2773  * Perform rate limit check.
2774  * Returns 0 if it is okay to send the icmp6 packet.
2775  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
2776  * limitation.
2777  *
2778  * XXX per-destination/type check necessary?
2779  */
2780 static int
2781 icmp6_ratelimit(const struct in6_addr *dst,	/* not used at this moment */
2782 		const int type,			/* not used at this moment */
2783 		const int code)			/* not used at this moment */
2784 {
2785 	int ret;
2786 
2787 	ret = 0;	/* okay to send */
2788 
2789 	/* PPS limit */
2790 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
2791 	    icmp6errppslim)) {
2792 		/* The packet is subject to rate limit */
2793 		ret++;
2794 	}
2795 
2796 	return ret;
2797 }
2798