Home | History | Annotate | Line # | Download | only in netinet6
icmp6.c revision 1.37
      1 /*	$NetBSD: icmp6.c,v 1.37 2000/07/09 06:44:58 itojun Exp $	*/
      2 /*	$KAME: icmp6.c,v 1.120 2000/07/06 11:47:20 itojun Exp $	*/
      3 
      4 /*
      5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
      6  * All rights reserved.
      7  *
      8  * Redistribution and use in source and binary forms, with or without
      9  * modification, are permitted provided that the following conditions
     10  * are met:
     11  * 1. Redistributions of source code must retain the above copyright
     12  *    notice, this list of conditions and the following disclaimer.
     13  * 2. Redistributions in binary form must reproduce the above copyright
     14  *    notice, this list of conditions and the following disclaimer in the
     15  *    documentation and/or other materials provided with the distribution.
     16  * 3. Neither the name of the project nor the names of its contributors
     17  *    may be used to endorse or promote products derived from this software
     18  *    without specific prior written permission.
     19  *
     20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
     21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
     24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     30  * SUCH DAMAGE.
     31  */
     32 
     33 /*
     34  * Copyright (c) 1982, 1986, 1988, 1993
     35  *	The Regents of the University of California.  All rights reserved.
     36  *
     37  * Redistribution and use in source and binary forms, with or without
     38  * modification, are permitted provided that the following conditions
     39  * are met:
     40  * 1. Redistributions of source code must retain the above copyright
     41  *    notice, this list of conditions and the following disclaimer.
     42  * 2. Redistributions in binary form must reproduce the above copyright
     43  *    notice, this list of conditions and the following disclaimer in the
     44  *    documentation and/or other materials provided with the distribution.
     45  * 3. All advertising materials mentioning features or use of this software
     46  *    must display the following acknowledgement:
     47  *	This product includes software developed by the University of
     48  *	California, Berkeley and its contributors.
     49  * 4. Neither the name of the University nor the names of its contributors
     50  *    may be used to endorse or promote products derived from this software
     51  *    without specific prior written permission.
     52  *
     53  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     54  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     55  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     56  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     57  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     58  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     59  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     61  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     62  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     63  * SUCH DAMAGE.
     64  *
     65  *	@(#)ip_icmp.c	8.2 (Berkeley) 1/4/94
     66  */
     67 
     68 #include "opt_inet.h"
     69 #include "opt_ipsec.h"
     70 
     71 #include <sys/param.h>
     72 #include <sys/systm.h>
     73 #include <sys/malloc.h>
     74 #include <sys/mbuf.h>
     75 #include <sys/protosw.h>
     76 #include <sys/socket.h>
     77 #include <sys/socketvar.h>
     78 #include <sys/time.h>
     79 #include <sys/kernel.h>
     80 #include <sys/syslog.h>
     81 #include <sys/domain.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 
     88 #include <netinet/in.h>
     89 #include <netinet/in_var.h>
     90 #include <netinet/ip6.h>
     91 #include <netinet6/ip6_var.h>
     92 #include <netinet/icmp6.h>
     93 #include <netinet6/mld6_var.h>
     94 #include <netinet6/in6_pcb.h>
     95 #include <netinet6/nd6.h>
     96 #include <netinet6/in6_ifattach.h>
     97 #include <netinet6/ip6protosw.h>
     98 
     99 
    100 #ifdef IPSEC
    101 #include <netinet6/ipsec.h>
    102 #include <netkey/key.h>
    103 #endif
    104 
    105 #include "faith.h"
    106 
    107 #include <net/net_osdep.h>
    108 
    109 extern struct domain inet6domain;
    110 
    111 struct icmp6stat icmp6stat;
    112 
    113 extern struct in6pcb rawin6pcb;
    114 extern struct timeval icmp6errratelim;
    115 static struct timeval icmp6errratelim_last;
    116 extern int icmp6errppslim;
    117 static int icmp6errpps_count = 0;
    118 static struct timeval icmp6errppslim_last;
    119 extern int icmp6_nodeinfo;
    120 static struct rttimer_queue *icmp6_mtudisc_timeout_q = NULL;
    121 extern int pmtu_expire;
    122 
    123 static void icmp6_errcount __P((struct icmp6errstat *, int, int));
    124 static int icmp6_rip6_input __P((struct mbuf **, int));
    125 static void icmp6_mtudisc_update __P((struct in6_addr *, struct icmp6_hdr *,
    126 				      struct mbuf *));
    127 static int icmp6_ratelimit __P((const struct in6_addr *, const int, const int));
    128 static const char *icmp6_redirect_diag __P((struct in6_addr *,
    129 	struct in6_addr *, struct in6_addr *));
    130 static struct mbuf *ni6_input __P((struct mbuf *, int));
    131 static struct mbuf *ni6_nametodns __P((const char *, int, int));
    132 static int ni6_dnsmatch __P((const char *, int, const char *, int));
    133 static int ni6_addrs __P((struct icmp6_nodeinfo *, struct mbuf *,
    134 			  struct ifnet **));
    135 static int ni6_store_addrs __P((struct icmp6_nodeinfo *, struct icmp6_nodeinfo *,
    136 				struct ifnet *, int));
    137 static struct rtentry *icmp6_mtudisc_clone __P((struct sockaddr *));
    138 static void icmp6_mtudisc_timeout __P((struct rtentry *, struct rttimer *));
    139 
    140 #ifdef COMPAT_RFC1885
    141 static struct route_in6 icmp6_reflect_rt;
    142 #endif
    143 
    144 void
    145 icmp6_init()
    146 {
    147 	mld6_init();
    148 	icmp6_mtudisc_timeout_q = rt_timer_queue_create(pmtu_expire);
    149 }
    150 
    151 static void
    152 icmp6_errcount(stat, type, code)
    153 	struct icmp6errstat *stat;
    154 	int type, code;
    155 {
    156 	switch(type) {
    157 	case ICMP6_DST_UNREACH:
    158 		switch (code) {
    159 		case ICMP6_DST_UNREACH_NOROUTE:
    160 			stat->icp6errs_dst_unreach_noroute++;
    161 			return;
    162 		case ICMP6_DST_UNREACH_ADMIN:
    163 			stat->icp6errs_dst_unreach_admin++;
    164 			return;
    165 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
    166 			stat->icp6errs_dst_unreach_beyondscope++;
    167 			return;
    168 		case ICMP6_DST_UNREACH_ADDR:
    169 			stat->icp6errs_dst_unreach_addr++;
    170 			return;
    171 		case ICMP6_DST_UNREACH_NOPORT:
    172 			stat->icp6errs_dst_unreach_noport++;
    173 			return;
    174 		}
    175 		break;
    176 	case ICMP6_PACKET_TOO_BIG:
    177 		stat->icp6errs_packet_too_big++;
    178 		return;
    179 	case ICMP6_TIME_EXCEEDED:
    180 		switch(code) {
    181 		case ICMP6_TIME_EXCEED_TRANSIT:
    182 			stat->icp6errs_time_exceed_transit++;
    183 			return;
    184 		case ICMP6_TIME_EXCEED_REASSEMBLY:
    185 			stat->icp6errs_time_exceed_reassembly++;
    186 			return;
    187 		}
    188 		break;
    189 	case ICMP6_PARAM_PROB:
    190 		switch(code) {
    191 		case ICMP6_PARAMPROB_HEADER:
    192 			stat->icp6errs_paramprob_header++;
    193 			return;
    194 		case ICMP6_PARAMPROB_NEXTHEADER:
    195 			stat->icp6errs_paramprob_nextheader++;
    196 			return;
    197 		case ICMP6_PARAMPROB_OPTION:
    198 			stat->icp6errs_paramprob_option++;
    199 			return;
    200 		}
    201 		break;
    202 	case ND_REDIRECT:
    203 		stat->icp6errs_redirect++;
    204 		return;
    205 	}
    206 	stat->icp6errs_unknown++;
    207 }
    208 
    209 /*
    210  * Generate an error packet of type error in response to bad IP6 packet.
    211  */
    212 void
    213 icmp6_error(m, type, code, param)
    214 	struct mbuf *m;
    215 	int type, code, param;
    216 {
    217 	struct ip6_hdr *oip6, *nip6;
    218 	struct icmp6_hdr *icmp6;
    219 	u_int preplen;
    220 	int off;
    221 	int nxt;
    222 
    223 	icmp6stat.icp6s_error++;
    224 
    225 	/* count per-type-code statistics */
    226 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, type, code);
    227 
    228 	if (m->m_flags & M_DECRYPTED) {
    229 		icmp6stat.icp6s_canterror++;
    230 		goto freeit;
    231 	}
    232 
    233 #ifndef PULLDOWN_TEST
    234 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), );
    235 #else
    236 	if (m->m_len < sizeof(struct ip6_hdr)) {
    237 		m = m_pullup(m, sizeof(struct ip6_hdr));
    238 		if (m == NULL)
    239 			return;
    240 	}
    241 #endif
    242 	oip6 = mtod(m, struct ip6_hdr *);
    243 
    244 	/*
    245 	 * Multicast destination check. For unrecognized option errors,
    246 	 * this check has already done in ip6_unknown_opt(), so we can
    247 	 * check only for other errors.
    248 	 */
    249 	if ((m->m_flags & (M_BCAST|M_MCAST) ||
    250 	     IN6_IS_ADDR_MULTICAST(&oip6->ip6_dst)) &&
    251 	    (type != ICMP6_PACKET_TOO_BIG &&
    252 	     (type != ICMP6_PARAM_PROB ||
    253 	      code != ICMP6_PARAMPROB_OPTION)))
    254 		goto freeit;
    255 
    256 	/* Source address check. XXX: the case of anycast source? */
    257 	if (IN6_IS_ADDR_UNSPECIFIED(&oip6->ip6_src) ||
    258 	    IN6_IS_ADDR_MULTICAST(&oip6->ip6_src))
    259 		goto freeit;
    260 
    261 	/*
    262 	 * If we are about to send ICMPv6 against ICMPv6 error/redirect,
    263 	 * don't do it.
    264 	 */
    265 	nxt = -1;
    266 	off = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
    267 	if (off >= 0 && nxt == IPPROTO_ICMPV6) {
    268 		struct icmp6_hdr *icp;
    269 
    270 #ifndef PULLDOWN_TEST
    271 		IP6_EXTHDR_CHECK(m, 0, off + sizeof(struct icmp6_hdr), );
    272 		icp = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
    273 #else
    274 		IP6_EXTHDR_GET(icp, struct icmp6_hdr *, m, off,
    275 			sizeof(*icp));
    276 		if (icp == NULL) {
    277 			icmp6stat.icp6s_tooshort++;
    278 			return;
    279 		}
    280 #endif
    281 		if (icp->icmp6_type < ICMP6_ECHO_REQUEST ||
    282 		    icp->icmp6_type == ND_REDIRECT) {
    283 			/*
    284 			 * ICMPv6 error
    285 			 * Special case: for redirect (which is
    286 			 * informational) we must not send icmp6 error.
    287 			 */
    288 			icmp6stat.icp6s_canterror++;
    289 			goto freeit;
    290 		} else {
    291 			/* ICMPv6 informational - send the error */
    292 		}
    293 	} else {
    294 		/* non-ICMPv6 - send the error */
    295 	}
    296 
    297 	oip6 = mtod(m, struct ip6_hdr *); /* adjust pointer */
    298 
    299 	/* Finally, do rate limitation check. */
    300 	if (icmp6_ratelimit(&oip6->ip6_src, type, code)) {
    301 		icmp6stat.icp6s_toofreq++;
    302 		goto freeit;
    303 	}
    304 
    305 	/*
    306 	 * OK, ICMP6 can be generated.
    307 	 */
    308 
    309 	if (m->m_pkthdr.len >= ICMPV6_PLD_MAXLEN)
    310 		m_adj(m, ICMPV6_PLD_MAXLEN - m->m_pkthdr.len);
    311 
    312 	preplen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
    313 	M_PREPEND(m, preplen, M_DONTWAIT);
    314 	if (m && m->m_len < preplen)
    315 		m = m_pullup(m, preplen);
    316 	if (m == NULL) {
    317 		printf("ENOBUFS in icmp6_error %d\n", __LINE__);
    318 		return;
    319 	}
    320 
    321 	nip6 = mtod(m, struct ip6_hdr *);
    322 	nip6->ip6_src  = oip6->ip6_src;
    323 	nip6->ip6_dst  = oip6->ip6_dst;
    324 
    325 	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src))
    326 		oip6->ip6_src.s6_addr16[1] = 0;
    327 	if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst))
    328 		oip6->ip6_dst.s6_addr16[1] = 0;
    329 
    330 	icmp6 = (struct icmp6_hdr *)(nip6 + 1);
    331 	icmp6->icmp6_type = type;
    332 	icmp6->icmp6_code = code;
    333 	icmp6->icmp6_pptr = htonl((u_int32_t)param);
    334 
    335 	icmp6stat.icp6s_outhist[type]++;
    336 	icmp6_reflect(m, sizeof(struct ip6_hdr)); /*header order: IPv6 - ICMPv6*/
    337 
    338 	return;
    339 
    340   freeit:
    341 	/*
    342 	 * If we can't tell wheter or not we can generate ICMP6, free it.
    343 	 */
    344 	m_freem(m);
    345 }
    346 
    347 /*
    348  * Process a received ICMP6 message.
    349  */
    350 int
    351 icmp6_input(mp, offp, proto)
    352 	struct mbuf **mp;
    353 	int *offp, proto;
    354 {
    355 	struct mbuf *m = *mp, *n;
    356 	struct ip6_hdr *ip6, *nip6;
    357 	struct icmp6_hdr *icmp6, *nicmp6;
    358 	int off = *offp;
    359 	int icmp6len = m->m_pkthdr.len - *offp;
    360 	int code, sum, noff;
    361 	struct sockaddr_in6 icmp6src;
    362 
    363 #ifndef PULLDOWN_TEST
    364 	IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_hdr), IPPROTO_DONE);
    365 	/* m might change if M_LOOP. So, call mtod after this */
    366 #endif
    367 
    368 	/*
    369 	 * Locate icmp6 structure in mbuf, and check
    370 	 * that not corrupted and of at least minimum length
    371 	 */
    372 
    373 	ip6 = mtod(m, struct ip6_hdr *);
    374 	if (icmp6len < sizeof(struct icmp6_hdr)) {
    375 		icmp6stat.icp6s_tooshort++;
    376 		goto freeit;
    377 	}
    378 
    379 	/*
    380 	 * calculate the checksum
    381 	 */
    382 #ifndef PULLDOWN_TEST
    383 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
    384 #else
    385 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
    386 	if (icmp6 == NULL) {
    387 		icmp6stat.icp6s_tooshort++;
    388 		return IPPROTO_DONE;
    389 	}
    390 #endif
    391 	code = icmp6->icmp6_code;
    392 
    393 	if ((sum = in6_cksum(m, IPPROTO_ICMPV6, off, icmp6len)) != 0) {
    394 		log(LOG_ERR,
    395 		    "ICMP6 checksum error(%d|%x) %s\n",
    396 		    icmp6->icmp6_type,
    397 		    sum,
    398 		    ip6_sprintf(&ip6->ip6_src));
    399 		icmp6stat.icp6s_checksum++;
    400 		goto freeit;
    401 	}
    402 
    403 #if defined(NFAITH) && 0 < NFAITH
    404 	if (m->m_pkthdr.rcvif && m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
    405 		/*
    406 		 * Deliver very specific ICMP6 type only.
    407 		 * This is important to deilver TOOBIG.  Otherwise PMTUD
    408 		 * will not work.
    409 		 */
    410 		switch (icmp6->icmp6_type) {
    411 		case ICMP6_DST_UNREACH:
    412 		case ICMP6_PACKET_TOO_BIG:
    413 		case ICMP6_TIME_EXCEEDED:
    414 			break;
    415 		default:
    416 			goto freeit;
    417 		}
    418 	}
    419 #endif
    420 
    421 #ifdef IPSEC
    422 	/* drop it if it does not match the default policy */
    423 	if (ipsec6_in_reject(m, NULL)) {
    424 		ipsecstat.in_polvio++;
    425 		goto freeit;
    426 	}
    427 #endif
    428 
    429 	icmp6stat.icp6s_inhist[icmp6->icmp6_type]++;
    430 	icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_msg);
    431 	if (icmp6->icmp6_type < ICMP6_INFOMSG_MASK)
    432 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_error);
    433 
    434 	switch (icmp6->icmp6_type) {
    435 
    436 	case ICMP6_DST_UNREACH:
    437 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_dstunreach);
    438 		switch (code) {
    439 		case ICMP6_DST_UNREACH_NOROUTE:
    440 			code = PRC_UNREACH_NET;
    441 			break;
    442 		case ICMP6_DST_UNREACH_ADMIN:
    443 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_adminprohib);
    444 			code = PRC_UNREACH_PROTOCOL; /* is this a good code? */
    445 			break;
    446 		case ICMP6_DST_UNREACH_ADDR:
    447 			code = PRC_HOSTDEAD;
    448 			break;
    449 #ifdef COMPAT_RFC1885
    450 		case ICMP6_DST_UNREACH_NOTNEIGHBOR:
    451 			code = PRC_UNREACH_SRCFAIL;
    452 			break;
    453 #else
    454 		case ICMP6_DST_UNREACH_BEYONDSCOPE:
    455 			/* I mean "source address was incorrect." */
    456 			code = PRC_PARAMPROB;
    457 			break;
    458 #endif
    459 		case ICMP6_DST_UNREACH_NOPORT:
    460 			code = PRC_UNREACH_PORT;
    461 			break;
    462 		default:
    463 			goto badcode;
    464 		}
    465 		goto deliver;
    466 		break;
    467 
    468 	case ICMP6_PACKET_TOO_BIG:
    469 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_pkttoobig);
    470 		if (code != 0)
    471 			goto badcode;
    472 
    473 		code = PRC_MSGSIZE;
    474 
    475 		/*
    476 		 * Updating the path MTU will be done after examining
    477 		 * intermediate extension headers.
    478 		 */
    479 		goto deliver;
    480 		break;
    481 
    482 	case ICMP6_TIME_EXCEEDED:
    483 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_timeexceed);
    484 		switch (code) {
    485 		case ICMP6_TIME_EXCEED_TRANSIT:
    486 		case ICMP6_TIME_EXCEED_REASSEMBLY:
    487 			code += PRC_TIMXCEED_INTRANS;
    488 			break;
    489 		default:
    490 			goto badcode;
    491 		}
    492 		goto deliver;
    493 		break;
    494 
    495 	case ICMP6_PARAM_PROB:
    496 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_paramprob);
    497 		switch (code) {
    498 		case ICMP6_PARAMPROB_NEXTHEADER:
    499 			code = PRC_UNREACH_PROTOCOL;
    500 			break;
    501 		case ICMP6_PARAMPROB_HEADER:
    502 		case ICMP6_PARAMPROB_OPTION:
    503 			code = PRC_PARAMPROB;
    504 			break;
    505 		default:
    506 			goto badcode;
    507 		}
    508 		goto deliver;
    509 		break;
    510 
    511 	case ICMP6_ECHO_REQUEST:
    512 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echo);
    513 		if (code != 0)
    514 			goto badcode;
    515 		if ((n = m_copy(m, 0, M_COPYALL)) == NULL) {
    516 			/* Give up remote */
    517 			break;
    518 		}
    519 		if ((n->m_flags & M_EXT) != 0
    520 		 || n->m_len < off + sizeof(struct icmp6_hdr)) {
    521 			struct mbuf *n0 = n;
    522 			const int maxlen = sizeof(*nip6) + sizeof(*nicmp6);
    523 
    524 			/*
    525 			 * Prepare an internal mbuf. m_pullup() doesn't
    526 			 * always copy the length we specified.
    527 			 */
    528 			if (maxlen >= MCLBYTES) {
    529 #ifdef DIAGNOSTIC
    530 				printf("MCLBYTES too small\n");
    531 #endif
    532 				/* Give up remote */
    533 				m_freem(n0);
    534 				break;
    535 			}
    536 			MGETHDR(n, M_DONTWAIT, n0->m_type);
    537 			if (n && maxlen >= MHLEN) {
    538 				MCLGET(n, M_DONTWAIT);
    539 				if ((n->m_flags & M_EXT) == 0) {
    540 					m_free(n);
    541 					n = NULL;
    542 				}
    543 			}
    544 			if (n == NULL) {
    545 				/* Give up remote */
    546 				m_freem(n0);
    547 				break;
    548 			}
    549 			M_COPY_PKTHDR(n, n0);
    550 			/*
    551 			 * Copy IPv6 and ICMPv6 only.
    552 			 */
    553 			nip6 = mtod(n, struct ip6_hdr *);
    554 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
    555 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
    556 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
    557 			noff = sizeof(struct ip6_hdr);
    558 			n->m_pkthdr.len = n->m_len =
    559 				noff + sizeof(struct icmp6_hdr);
    560 			/*
    561 			 * Adjust mbuf. ip6_plen will be adjusted in
    562 			 * ip6_output().
    563 			 */
    564 			m_adj(n0, off + sizeof(struct icmp6_hdr));
    565 			n->m_pkthdr.len += n0->m_pkthdr.len;
    566 			n->m_next = n0;
    567 			n0->m_flags &= ~M_PKTHDR;
    568 		} else {
    569 			nip6 = mtod(n, struct ip6_hdr *);
    570 			nicmp6 = (struct icmp6_hdr *)((caddr_t)nip6 + off);
    571 			noff = off;
    572 		}
    573 		nicmp6->icmp6_type = ICMP6_ECHO_REPLY;
    574 		nicmp6->icmp6_code = 0;
    575 		if (n) {
    576 			icmp6stat.icp6s_reflect++;
    577 			icmp6stat.icp6s_outhist[ICMP6_ECHO_REPLY]++;
    578 			icmp6_reflect(n, noff);
    579 		}
    580 		break;
    581 
    582 	case ICMP6_ECHO_REPLY:
    583 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_echoreply);
    584 		if (code != 0)
    585 			goto badcode;
    586 		break;
    587 
    588 	case MLD6_LISTENER_QUERY:
    589 	case MLD6_LISTENER_REPORT:
    590 		if (icmp6len < sizeof(struct mld6_hdr))
    591 			goto badlen;
    592 		if (icmp6->icmp6_type == MLD6_LISTENER_QUERY) /* XXX: ugly... */
    593 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldquery);
    594 		else
    595 			icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mldreport);
    596 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
    597 			/* give up local */
    598 			mld6_input(m, off);
    599 			m = NULL;
    600 			goto freeit;
    601 		}
    602 		mld6_input(n, off);
    603 		/* m stays. */
    604 		break;
    605 
    606 	case MLD6_LISTENER_DONE:
    607 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mlddone);
    608 		if (icmp6len < sizeof(struct mld6_hdr))	/* necessary? */
    609 			goto badlen;
    610 		break;		/* nothing to be done in kernel */
    611 
    612 	case MLD6_MTRACE_RESP:
    613 	case MLD6_MTRACE:
    614 		/* XXX: these two are experimental. not officially defind. */
    615 		/* XXX: per-interface statistics? */
    616 		break;		/* just pass it to applications */
    617 
    618 	case ICMP6_WRUREQUEST:	/* ICMP6_FQDN_QUERY */
    619 	    {
    620 		enum { WRU, FQDN } mode;
    621 
    622 		if (!icmp6_nodeinfo)
    623 			break;
    624 
    625 		if (icmp6len == sizeof(struct icmp6_hdr) + 4)
    626 			mode = WRU;
    627 		else if (icmp6len >= sizeof(struct icmp6_nodeinfo))
    628 			mode = FQDN;
    629 		else
    630 			goto badlen;
    631 
    632 		if (mode == FQDN) {
    633 #ifndef PULLDOWN_TEST
    634 			IP6_EXTHDR_CHECK(m, off, sizeof(struct icmp6_nodeinfo),
    635 					 IPPROTO_DONE);
    636 #endif
    637 			n = m_copy(m, 0, M_COPYALL);
    638 			if (n)
    639 				n = ni6_input(n, off);
    640 			/* XXX meaningless if n == NULL */
    641 			noff = sizeof(struct ip6_hdr);
    642 		} else {
    643 			u_char *p;
    644 			int maxlen, maxhlen;
    645 
    646 			if (code != 0)
    647 				goto badcode;
    648 			maxlen = sizeof(*nip6) + sizeof(*nicmp6) + 4;
    649 			if (maxlen >= MCLBYTES) {
    650 #ifdef DIAGNOSTIC
    651 				printf("MCLBYTES too small\n");
    652 #endif
    653 				/* Give up remote */
    654 				break;
    655 			}
    656 			MGETHDR(n, M_DONTWAIT, m->m_type);
    657 			if (n && maxlen > MHLEN) {
    658 				MCLGET(n, M_DONTWAIT);
    659 				if ((n->m_flags & M_EXT) == 0) {
    660 					m_free(n);
    661 					n = NULL;
    662 				}
    663 			}
    664 			if (n == NULL) {
    665 				/* Give up remote */
    666 				break;
    667 			}
    668 			n->m_len = 0;
    669 			maxhlen = M_TRAILINGSPACE(n) - maxlen;
    670 			if (maxhlen > hostnamelen)
    671 				maxhlen = hostnamelen;
    672 			/*
    673 			 * Copy IPv6 and ICMPv6 only.
    674 			 */
    675 			nip6 = mtod(n, struct ip6_hdr *);
    676 			bcopy(ip6, nip6, sizeof(struct ip6_hdr));
    677 			nicmp6 = (struct icmp6_hdr *)(nip6 + 1);
    678 			bcopy(icmp6, nicmp6, sizeof(struct icmp6_hdr));
    679 			p = (u_char *)(nicmp6 + 1);
    680 			bzero(p, 4);
    681 			bcopy(hostname, p + 4, maxhlen); /*meaningless TTL*/
    682 			noff = sizeof(struct ip6_hdr);
    683 			M_COPY_PKTHDR(n, m); /* just for recvif */
    684 			n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
    685 				sizeof(struct icmp6_hdr) + 4 + maxhlen;
    686 			nicmp6->icmp6_type = ICMP6_WRUREPLY;
    687 			nicmp6->icmp6_code = 0;
    688 		}
    689 #undef hostnamelen
    690 		if (n) {
    691 			icmp6stat.icp6s_reflect++;
    692 			icmp6stat.icp6s_outhist[ICMP6_WRUREPLY]++;
    693 			icmp6_reflect(n, noff);
    694 		}
    695 		break;
    696 	    }
    697 
    698 	case ICMP6_WRUREPLY:
    699 		if (code != 0)
    700 			goto badcode;
    701 		break;
    702 
    703 	case ND_ROUTER_SOLICIT:
    704 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routersolicit);
    705 		if (code != 0)
    706 			goto badcode;
    707 		if (icmp6len < sizeof(struct nd_router_solicit))
    708 			goto badlen;
    709 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
    710 			/* give up local */
    711 			nd6_rs_input(m, off, icmp6len);
    712 			m = NULL;
    713 			goto freeit;
    714 		}
    715 		nd6_rs_input(n, off, icmp6len);
    716 		/* m stays. */
    717 		break;
    718 
    719 	case ND_ROUTER_ADVERT:
    720 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_routeradvert);
    721 		if (code != 0)
    722 			goto badcode;
    723 		if (icmp6len < sizeof(struct nd_router_advert))
    724 			goto badlen;
    725 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
    726 			/* give up local */
    727 			nd6_ra_input(m, off, icmp6len);
    728 			m = NULL;
    729 			goto freeit;
    730 		}
    731 		nd6_ra_input(n, off, icmp6len);
    732 		/* m stays. */
    733 		break;
    734 
    735 	case ND_NEIGHBOR_SOLICIT:
    736 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighborsolicit);
    737 		if (code != 0)
    738 			goto badcode;
    739 		if (icmp6len < sizeof(struct nd_neighbor_solicit))
    740 			goto badlen;
    741 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
    742 			/* give up local */
    743 			nd6_ns_input(m, off, icmp6len);
    744 			m = NULL;
    745 			goto freeit;
    746 		}
    747 		nd6_ns_input(n, off, icmp6len);
    748 		/* m stays. */
    749 		break;
    750 
    751 	case ND_NEIGHBOR_ADVERT:
    752 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_neighboradvert);
    753 		if (code != 0)
    754 			goto badcode;
    755 		if (icmp6len < sizeof(struct nd_neighbor_advert))
    756 			goto badlen;
    757 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
    758 			/* give up local */
    759 			nd6_na_input(m, off, icmp6len);
    760 			m = NULL;
    761 			goto freeit;
    762 		}
    763 		nd6_na_input(n, off, icmp6len);
    764 		/* m stays. */
    765 		break;
    766 
    767 	case ND_REDIRECT:
    768 		icmp6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_redirect);
    769 		if (code != 0)
    770 			goto badcode;
    771 		if (icmp6len < sizeof(struct nd_redirect))
    772 			goto badlen;
    773 		if ((n = m_copym(m, 0, M_COPYALL, M_DONTWAIT)) == NULL) {
    774 			/* give up local */
    775 			icmp6_redirect_input(m, off);
    776 			m = NULL;
    777 			goto freeit;
    778 		}
    779 		icmp6_redirect_input(n, off);
    780 		/* m stays. */
    781 		break;
    782 
    783 	case ICMP6_ROUTER_RENUMBERING:
    784 		if (code != ICMP6_ROUTER_RENUMBERING_COMMAND &&
    785 		    code != ICMP6_ROUTER_RENUMBERING_RESULT)
    786 			goto badcode;
    787 		if (icmp6len < sizeof(struct icmp6_router_renum))
    788 			goto badlen;
    789 		break;
    790 
    791 	default:
    792 		printf("icmp6_input: unknown type %d(src=%s, dst=%s, ifid=%d)\n",
    793 		       icmp6->icmp6_type, ip6_sprintf(&ip6->ip6_src),
    794 		       ip6_sprintf(&ip6->ip6_dst),
    795 		       m->m_pkthdr.rcvif ? m->m_pkthdr.rcvif->if_index : 0);
    796 		if (icmp6->icmp6_type < ICMP6_ECHO_REQUEST) {
    797 			/* ICMPv6 error: MUST deliver it by spec... */
    798 			code = PRC_NCMDS;
    799 			/* deliver */
    800 		} else {
    801 			/* ICMPv6 informational: MUST not deliver */
    802 			break;
    803 		}
    804 	deliver:
    805 		if (icmp6len < sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr)) {
    806 			icmp6stat.icp6s_tooshort++;
    807 			goto freeit;
    808 		}
    809 #ifndef PULLDOWN_TEST
    810 		IP6_EXTHDR_CHECK(m, off,
    811 			sizeof(struct icmp6_hdr) + sizeof(struct ip6_hdr),
    812 			IPPROTO_DONE);
    813 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
    814 #else
    815 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
    816 			sizeof(*icmp6) + sizeof(struct ip6_hdr));
    817 		if (icmp6 == NULL) {
    818 			icmp6stat.icp6s_tooshort++;
    819 			return IPPROTO_DONE;
    820 		}
    821 #endif
    822 		bzero(&icmp6src, sizeof(icmp6src));
    823 		icmp6src.sin6_len = sizeof(struct sockaddr_in6);
    824 		icmp6src.sin6_family = AF_INET6;
    825 		icmp6src.sin6_addr = ((struct ip6_hdr *)(icmp6 + 1))->ip6_dst;
    826 
    827 		/* Detect the upper level protocol */
    828 	    {
    829 		void (*ctlfunc) __P((int, struct sockaddr *, void *));
    830 		struct ip6_hdr *eip6 = (struct ip6_hdr *)(icmp6 + 1);
    831 		u_int8_t nxt = eip6->ip6_nxt;
    832 		int eoff = off + sizeof(struct icmp6_hdr) +
    833 			sizeof(struct ip6_hdr);
    834 		struct ip6ctlparam ip6cp;
    835 		struct in6_addr *finaldst = NULL;
    836 		int icmp6type = icmp6->icmp6_type;
    837 		struct ip6_frag *fh;
    838 		struct ip6_rthdr *rth;
    839 		struct ip6_rthdr0 *rth0;
    840 		int rthlen;
    841 
    842 		while (1) { /* XXX: should avoid inf. loop explicitly? */
    843 			struct ip6_ext *eh;
    844 
    845 			switch(nxt) {
    846 			case IPPROTO_HOPOPTS:
    847 			case IPPROTO_DSTOPTS:
    848 			case IPPROTO_AH:
    849 #ifndef PULLDOWN_TEST
    850 				IP6_EXTHDR_CHECK(m, 0, eoff +
    851 						 sizeof(struct ip6_ext),
    852 						 IPPROTO_DONE);
    853 				eh = (struct ip6_ext *)(mtod(m, caddr_t)
    854 							+ eoff);
    855 #else
    856 				IP6_EXTHDR_GET(eh, struct ip6_ext *, m,
    857 					eoff, sizeof(*eh));
    858 				if (eh == NULL) {
    859 					icmp6stat.icp6s_tooshort++;
    860 					return IPPROTO_DONE;
    861 				}
    862 #endif
    863 
    864 				if (nxt == IPPROTO_AH)
    865 					eoff += (eh->ip6e_len + 2) << 2;
    866 				else
    867 					eoff += (eh->ip6e_len + 1) << 3;
    868 				nxt = eh->ip6e_nxt;
    869 				break;
    870 			case IPPROTO_ROUTING:
    871 				/*
    872 				 * When the erroneous packet contains a
    873 				 * routing header, we should examine the
    874 				 * header to determine the final destination.
    875 				 * Otherwise, we can't properly update
    876 				 * information that depends on the final
    877 				 * destination (e.g. path MTU).
    878 				 */
    879 #ifndef PULLDOWN_TEST
    880 				IP6_EXTHDR_CHECK(m, 0, eoff + sizeof(*rth),
    881 						 IPPROTO_DONE);
    882 				rth = (struct ip6_rthdr *)(mtod(m, caddr_t)
    883 							   + eoff);
    884 #else
    885 				IP6_EXTHDR_GET(rth, struct ip6_rthdr *, m,
    886 					eoff, sizeof(*rth));
    887 				if (rth == NULL) {
    888 					icmp6stat.icp6s_tooshort++;
    889 					return IPPROTO_DONE;
    890 				}
    891 #endif
    892 				rthlen = (rth->ip6r_len + 1) << 3;
    893 				/*
    894 				 * XXX: currently there is no
    895 				 * officially defined type other
    896 				 * than type-0.
    897 				 * Note that if the segment left field
    898 				 * is 0, all intermediate hops must
    899 				 * have been passed.
    900 				 */
    901 				if (rth->ip6r_segleft &&
    902 				    rth->ip6r_type == IPV6_RTHDR_TYPE_0) {
    903 					int hops;
    904 
    905 #ifndef PULLDOWN_TEST
    906 					IP6_EXTHDR_CHECK(m, 0, eoff + rthlen,
    907 							 IPPROTO_DONE);
    908 					rth0 = (struct ip6_rthdr0 *)(mtod(m, caddr_t) + eoff);
    909 #else
    910 					IP6_EXTHDR_GET(rth0,
    911 						       struct ip6_rthdr0 *, m,
    912 						       eoff, rthlen);
    913 					if (rth0 == NULL) {
    914 						icmp6stat.icp6s_tooshort++;
    915 						return IPPROTO_DONE;
    916 					}
    917 #endif
    918 					/* just ignore a bogus header */
    919 					if ((rth0->ip6r0_len % 2) == 0 &&
    920 					    (hops = rth0->ip6r0_len/2))
    921 						finaldst = (struct in6_addr *)(rth0 + 1) + (hops - 1);
    922 				}
    923 				eoff += rthlen;
    924 				nxt = rth->ip6r_nxt;
    925 				break;
    926 			case IPPROTO_FRAGMENT:
    927 #ifndef PULLDOWN_TEST
    928 				IP6_EXTHDR_CHECK(m, 0, eoff +
    929 						 sizeof(struct ip6_frag),
    930 						 IPPROTO_DONE);
    931 				fh = (struct ip6_frag *)(mtod(m, caddr_t)
    932 							 + eoff);
    933 #else
    934 				IP6_EXTHDR_GET(fh, struct ip6_frag *, m,
    935 					eoff, sizeof(*fh));
    936 				if (fh == NULL) {
    937 					icmp6stat.icp6s_tooshort++;
    938 					return IPPROTO_DONE;
    939 				}
    940 #endif
    941 				/*
    942 				 * Data after a fragment header is meaningless
    943 				 * unless it is the first fragment, but
    944 				 * we'll go to the notify label for path MTU
    945 				 * discovery.
    946 				 */
    947 				if (fh->ip6f_offlg & IP6F_OFF_MASK)
    948 					goto notify;
    949 
    950 				eoff += sizeof(struct ip6_frag);
    951 				nxt = fh->ip6f_nxt;
    952 				break;
    953 			default:
    954 				/*
    955 				 * This case includes ESP and the No Next
    956 				 * Header. In such cases going to the notify
    957 				 * label does not have any meaning
    958 				 * (i.e. ctlfunc will be NULL), but we go
    959 				 * anyway since we might have to update
    960 				 * path MTU information.
    961 				 */
    962 				goto notify;
    963 			}
    964 		}
    965 	    notify:
    966 #ifndef PULLDOWN_TEST
    967 		icmp6 = (struct icmp6_hdr *)(mtod(m, caddr_t) + off);
    968 #else
    969 		IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off,
    970 			sizeof(*icmp6) + sizeof(struct ip6_hdr));
    971 		if (icmp6 == NULL) {
    972 			icmp6stat.icp6s_tooshort++;
    973 			return IPPROTO_DONE;
    974 		}
    975 #endif
    976 		if (icmp6type == ICMP6_PACKET_TOO_BIG) {
    977 			if (finaldst == NULL)
    978 				finaldst = &((struct ip6_hdr *)(icmp6 + 1))->ip6_dst;
    979 			icmp6_mtudisc_update(finaldst, icmp6, m);
    980 		}
    981 
    982 		ctlfunc = (void (*) __P((int, struct sockaddr *, void *)))
    983 			(inet6sw[ip6_protox[nxt]].pr_ctlinput);
    984 		if (ctlfunc) {
    985 			ip6cp.ip6c_m = m;
    986 			ip6cp.ip6c_ip6 = (struct ip6_hdr *)(icmp6 + 1);
    987 			ip6cp.ip6c_off = eoff;
    988 			(*ctlfunc)(code, (struct sockaddr *)&icmp6src, &ip6cp);
    989 		}
    990 	    }
    991 		break;
    992 
    993 	badcode:
    994 		icmp6stat.icp6s_badcode++;
    995 		break;
    996 
    997 	badlen:
    998 		icmp6stat.icp6s_badlen++;
    999 		break;
   1000 	}
   1001 
   1002 	icmp6_rip6_input(&m, *offp);
   1003 	return IPPROTO_DONE;
   1004 
   1005  freeit:
   1006 	m_freem(m);
   1007 	return IPPROTO_DONE;
   1008 }
   1009 
   1010 static void
   1011 icmp6_mtudisc_update(dst, icmp6, m)
   1012 	struct in6_addr *dst;
   1013 	struct icmp6_hdr *icmp6;/* we can assume the validity of the pointer */
   1014 	struct mbuf *m;	/* currently unused but added for scoped addrs */
   1015 {
   1016 	u_int mtu = ntohl(icmp6->icmp6_mtu);
   1017 	struct rtentry *rt = NULL;
   1018 	struct sockaddr_in6 sin6;
   1019 
   1020 	bzero(&sin6, sizeof(sin6));
   1021 	sin6.sin6_family = PF_INET6;
   1022 	sin6.sin6_len = sizeof(struct sockaddr_in6);
   1023 	sin6.sin6_addr = *dst;
   1024 	/* sin6.sin6_scope_id = XXX: should be set if DST is a scoped addr */
   1025 	rt = rtalloc1((struct sockaddr *)&sin6, 1);	/*clone*/
   1026 	if (!rt || (rt->rt_flags & RTF_HOST) == 0) {
   1027 		if (rt)
   1028 			RTFREE(rt);
   1029 		rt = icmp6_mtudisc_clone((struct sockaddr *)&sin6);
   1030 	}
   1031 
   1032 	if (rt && (rt->rt_flags & RTF_HOST)
   1033 	    && !(rt->rt_rmx.rmx_locks & RTV_MTU)) {
   1034 		if (mtu < IPV6_MMTU) {
   1035 				/* xxx */
   1036 			rt->rt_rmx.rmx_locks |= RTV_MTU;
   1037 		} else if (mtu < rt->rt_ifp->if_mtu &&
   1038 			   rt->rt_rmx.rmx_mtu > mtu) {
   1039 			rt->rt_rmx.rmx_mtu = mtu;
   1040 		}
   1041 	}
   1042 	if (rt)
   1043 		RTFREE(rt);
   1044 }
   1045 
   1046 /*
   1047  * Process a Node Information Query packet, (roughly) based on
   1048  * draft-ietf-ipngwg-icmp-name-lookups-05.
   1049  *
   1050  * Spec incompatibilities:
   1051  * - IPv6 Subject address handling
   1052  * - IPv4 Subject address handling support missing
   1053  * - Proxy reply (answer even if it's not for me)
   1054  * - "Supported Qtypes" support missing
   1055  * - joins NI group address at in6_ifattach() time only, does not cope
   1056  *   with hostname changes by sethostname(3)
   1057  */
   1058 #ifndef offsetof		/* XXX */
   1059 #define	offsetof(type, member)	((size_t)(&((type *)0)->member))
   1060 #endif
   1061 static struct mbuf *
   1062 ni6_input(m, off)
   1063 	struct mbuf *m;
   1064 	int off;
   1065 {
   1066 	struct icmp6_nodeinfo *ni6, *nni6;
   1067 	struct mbuf *n = NULL;
   1068 	u_int16_t qtype;
   1069 	int subjlen;
   1070 	int replylen = sizeof(struct ip6_hdr) + sizeof(struct icmp6_nodeinfo);
   1071 	struct ni_reply_fqdn *fqdn;
   1072 	int addrs;		/* for NI_QTYPE_NODEADDR */
   1073 	struct ifnet *ifp = NULL; /* for NI_QTYPE_NODEADDR */
   1074 	struct sockaddr_in6 sin6;
   1075 	struct ip6_hdr *ip6;
   1076 	int oldfqdn = 0;	/* if 1, return pascal string (03 draft) */
   1077 	char *subj;
   1078 
   1079 	ip6 = mtod(m, struct ip6_hdr *);
   1080 #ifndef PULLDOWN_TEST
   1081 	ni6 = (struct icmp6_nodeinfo *)(mtod(m, caddr_t) + off);
   1082 #else
   1083 	IP6_EXTHDR_GET(ni6, struct icmp6_nodeinfo *, m, off, sizeof(*ni6));
   1084 	if (ni6 == NULL) {
   1085 		/* m is already reclaimed */
   1086 		return NULL;
   1087 	}
   1088 #endif
   1089 
   1090 	/*
   1091 	 * Validate IPv6 destination address.
   1092 	 *
   1093 	 * We accept packets with the following IPv6 destination address:
   1094 	 * - Responder's unicast/anycast address,
   1095 	 * - link-local multicast address
   1096 	 * This is a violation to last paragraph in icmp-name-lookups-05
   1097 	 * page 4, which restricts IPv6 destination address of a query to:
   1098 	 * - Responder's unicast/anycast address,
   1099 	 * - NI group address for a name belongs to the Responder, or
   1100 	 * - NI group address for a name for which the Responder is providing
   1101 	 *   proxy service.
   1102 	 * (note: NI group address is a link-local multicast address)
   1103 	 *
   1104 	 * We allow any link-local multicast address, since "ping6 -w ff02::1"
   1105 	 * has been really useful for us debugging our network.  Also this is
   1106 	 * still questionable if the restriction in spec buy us security at all,
   1107 	 * since RFC2463 permits echo packet to multicast destination.
   1108 	 * Even if we forbid NI query to ff02::1, we can effectively get the
   1109 	 * same result as "ping6 -w ff02::1" by the following steps:
   1110 	 * - run "ping6 ff02::1", then
   1111 	 * - run "ping6 -w" for all addresses replied.
   1112 	 */
   1113 	bzero(&sin6, sizeof(sin6));
   1114 	sin6.sin6_family = AF_INET6;
   1115 	sin6.sin6_len = sizeof(struct sockaddr_in6);
   1116 	bcopy(&ip6->ip6_dst, &sin6.sin6_addr, sizeof(sin6.sin6_addr));
   1117 	/* XXX scopeid */
   1118 	if (ifa_ifwithaddr((struct sockaddr *)&sin6))
   1119 		; /*unicast/anycast, fine*/
   1120 	else if (IN6_IS_ADDR_MC_LINKLOCAL(&sin6.sin6_addr))
   1121 		; /*violates spec slightly, see above*/
   1122 	else
   1123 		goto bad;
   1124 
   1125 	/* guess reply length */
   1126 	qtype = ntohs(ni6->ni_qtype);
   1127 	switch (qtype) {
   1128 	case NI_QTYPE_NOOP:
   1129 		break;		/* no reply data */
   1130 	case NI_QTYPE_SUPTYPES:
   1131 		goto bad;	/* xxx: to be implemented */
   1132 		break;
   1133 	case NI_QTYPE_FQDN:
   1134 		/* XXX will append a mbuf */
   1135 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
   1136 		break;
   1137 	case NI_QTYPE_NODEADDR:
   1138 		addrs = ni6_addrs(ni6, m, &ifp);
   1139 		if ((replylen += addrs * sizeof(struct in6_addr)) > MCLBYTES)
   1140 			replylen = MCLBYTES; /* XXX: we'll truncate later */
   1141 		break;
   1142 	default:
   1143 		/*
   1144 		 * XXX: We must return a reply with the ICMP6 code
   1145 		 * `unknown Qtype' in this case. However we regard the case
   1146 		 * as an FQDN query for backward compatibility.
   1147 		 * Older versions set a random value to this field,
   1148 		 * so it rarely varies in the defined qtypes.
   1149 		 * But the mechanism is not reliable...
   1150 		 * maybe we should obsolete older versions.
   1151 		 */
   1152 		qtype = NI_QTYPE_FQDN;
   1153 		/* XXX will append a mbuf */
   1154 		replylen += offsetof(struct ni_reply_fqdn, ni_fqdn_namelen);
   1155 		oldfqdn++;
   1156 		break;
   1157 	}
   1158 
   1159 	/* validate query Subject field. */
   1160 	subjlen = m->m_pkthdr.len - off - sizeof(struct icmp6_nodeinfo);
   1161 	switch (qtype) {
   1162 	case NI_QTYPE_NOOP:
   1163 	case NI_QTYPE_SUPTYPES:
   1164 		if (subjlen != 0)
   1165 			goto bad;
   1166 		break;
   1167 
   1168 	case NI_QTYPE_FQDN:
   1169 	case NI_QTYPE_NODEADDR:
   1170 		switch (ni6->ni_code) {
   1171 		case ICMP6_NI_SUBJ_IPV6:
   1172 #if ICMP6_NI_SUBJ_IPV6 != 0
   1173 		case 0:
   1174 #endif
   1175 			/*
   1176 			 * backward compatibility - try to accept 03 draft
   1177 			 * format, where no Subject is present.
   1178 			 */
   1179 			if (subjlen == 0) {
   1180 				oldfqdn++;
   1181 				break;
   1182 			}
   1183 
   1184 			if (subjlen != sizeof(sin6.sin6_addr))
   1185 				goto bad;
   1186 
   1187 			/*
   1188 			 * Validate Subject address.
   1189 			 *
   1190 			 * Not sure what exactly does "address belongs to the
   1191 			 * node" mean in the spec, is it just unicast, or what?
   1192 			 *
   1193 			 * At this moment we consider Subject address as
   1194 			 * "belong to the node" if the Subject address equals
   1195 			 * to the IPv6 destination address; validation for
   1196 			 * IPv6 destination address should have done enough
   1197 			 * check for us.
   1198 			 *
   1199 			 * We do not do proxy at this moment.
   1200 			 */
   1201 			/* m_pulldown instead of copy? */
   1202 			m_copydata(m, off + sizeof(struct icmp6_nodeinfo),
   1203 			    subjlen, (caddr_t)&sin6.sin6_addr);
   1204 			/* XXX kame scope hack */
   1205 			if (IN6_IS_SCOPE_LINKLOCAL(&sin6.sin6_addr)) {
   1206 #ifdef FAKE_LOOPBACK_IF
   1207 				if ((m->m_flags & M_PKTHDR) != 0 &&
   1208 				    m->m_pkthdr.rcvif) {
   1209 					sin6.sin6_addr.s6_addr16[1] =
   1210 					    htons(m->m_pkthdr.rcvif->if_index);
   1211 				}
   1212 #else
   1213 				if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) {
   1214 					sin6.sin6_addr.s6_addr16[1] =
   1215 					    ip6->ip6_dst.s6_addr16[1];
   1216 				}
   1217 #endif
   1218 			}
   1219 			if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &sin6.sin6_addr))
   1220 				break;
   1221 			/*
   1222 			 * XXX if we are to allow other cases, we should really
   1223 			 * be careful about scope here.
   1224 			 * basically, we should disallow queries toward IPv6
   1225 			 * destination X with subject Y, if scope(X) > scope(Y).
   1226 			 * if we allow scope(X) > scope(Y), it will result in
   1227 			 * information leakage across scope boundary.
   1228 			 */
   1229 			goto bad;
   1230 
   1231 		case ICMP6_NI_SUBJ_FQDN:
   1232 			/*
   1233 			 * Validate Subject name with gethostname(3).
   1234 			 *
   1235 			 * The behavior may need some debate, since:
   1236 			 * - we are not sure if the node has FQDN as
   1237 			 *   hostname (returned by gethostname(3)).
   1238 			 * - the code does wildcard match for truncated names.
   1239 			 *   however, we are not sure if we want to perform
   1240 			 *   wildcard match, if gethostname(3) side has
   1241 			 *   truncated hostname.
   1242 			 */
   1243 			n = ni6_nametodns(hostname, hostnamelen, 0);
   1244 			if (!n || n->m_next || n->m_len == 0)
   1245 				goto bad;
   1246 			IP6_EXTHDR_GET(subj, char *, m,
   1247 			    off + sizeof(struct icmp6_nodeinfo), subjlen);
   1248 			if (subj == NULL)
   1249 				goto bad;
   1250 			if (!ni6_dnsmatch(subj, subjlen, mtod(n, const char *),
   1251 					n->m_len)) {
   1252 				goto bad;
   1253 			}
   1254 			m_freem(n);
   1255 			n = NULL;
   1256 			break;
   1257 
   1258 		case ICMP6_NI_SUBJ_IPV4:	/* xxx: to be implemented? */
   1259 		default:
   1260 			goto bad;
   1261 		}
   1262 		break;
   1263 
   1264 	default:
   1265 		/* should never be here due to "switch (qtype)" above */
   1266 		goto bad;
   1267 	}
   1268 
   1269 	/* allocate a mbuf to reply. */
   1270 	MGETHDR(n, M_DONTWAIT, m->m_type);
   1271 	if (n == NULL) {
   1272 		m_freem(m);
   1273 		return(NULL);
   1274 	}
   1275 	M_COPY_PKTHDR(n, m); /* just for recvif */
   1276 	if (replylen > MHLEN) {
   1277 		if (replylen > MCLBYTES) {
   1278 			 /*
   1279 			  * XXX: should we try to allocate more? But MCLBYTES is
   1280 			  * probably much larger than IPV6_MMTU...
   1281 			  */
   1282 			goto bad;
   1283 		}
   1284 		MCLGET(n, M_DONTWAIT);
   1285 		if ((n->m_flags & M_EXT) == 0) {
   1286 			goto bad;
   1287 		}
   1288 	}
   1289 	n->m_pkthdr.len = n->m_len = replylen;
   1290 
   1291 	/* copy mbuf header and IPv6 + Node Information base headers */
   1292 	bcopy(mtod(m, caddr_t), mtod(n, caddr_t), sizeof(struct ip6_hdr));
   1293 	nni6 = (struct icmp6_nodeinfo *)(mtod(n, struct ip6_hdr *) + 1);
   1294 	bcopy((caddr_t)ni6, (caddr_t)nni6, sizeof(struct icmp6_nodeinfo));
   1295 
   1296 	/* qtype dependent procedure */
   1297 	switch (qtype) {
   1298 	case NI_QTYPE_NOOP:
   1299 		nni6->ni_flags = 0;
   1300 		break;
   1301 	case NI_QTYPE_SUPTYPES:
   1302 		goto bad;	/* xxx: to be implemented */
   1303 		break;
   1304 	case NI_QTYPE_FQDN:
   1305 		fqdn = (struct ni_reply_fqdn *)(mtod(n, caddr_t) +
   1306 						sizeof(struct ip6_hdr) +
   1307 						sizeof(struct icmp6_nodeinfo));
   1308 		nni6->ni_flags = 0; /* XXX: meaningless TTL */
   1309 		fqdn->ni_fqdn_ttl = 0;	/* ditto. */
   1310 		/*
   1311 		 * XXX do we really have FQDN in variable "hostname"?
   1312 		 */
   1313 		n->m_next = ni6_nametodns(hostname, hostnamelen, oldfqdn);
   1314 		if (n->m_next == NULL)
   1315 			goto bad;
   1316 		/* XXX we assume that n->m_next is not a chain */
   1317 		if (n->m_next->m_next != NULL)
   1318 			goto bad;
   1319 		n->m_pkthdr.len += n->m_next->m_len;
   1320 		break;
   1321 	case NI_QTYPE_NODEADDR:
   1322 	{
   1323 		int lenlim, copied;
   1324 
   1325 		if (n->m_flags & M_EXT)
   1326 			lenlim = MCLBYTES - sizeof(struct ip6_hdr) -
   1327 				sizeof(struct icmp6_nodeinfo);
   1328 		else
   1329 			lenlim = MHLEN - sizeof(struct ip6_hdr) -
   1330 				sizeof(struct icmp6_nodeinfo);
   1331 		copied = ni6_store_addrs(ni6, nni6, ifp, lenlim);
   1332 		/* XXX: reset mbuf length */
   1333 		n->m_pkthdr.len = n->m_len = sizeof(struct ip6_hdr) +
   1334 			sizeof(struct icmp6_nodeinfo) + copied;
   1335 		break;
   1336 	}
   1337 	default:
   1338 		break;		/* XXX impossible! */
   1339 	}
   1340 
   1341 	nni6->ni_type = ICMP6_NI_REPLY;
   1342 	nni6->ni_code = ICMP6_NI_SUCESS;
   1343 	m_freem(m);
   1344 	return(n);
   1345 
   1346   bad:
   1347 	m_freem(m);
   1348 	if (n)
   1349 		m_freem(n);
   1350 	return(NULL);
   1351 }
   1352 #undef hostnamelen
   1353 
   1354 /*
   1355  * make a mbuf with DNS-encoded string.  no compression support.
   1356  *
   1357  * XXX names with less than 2 dots (like "foo" or "foo.section") will be
   1358  * treated as truncated name (two \0 at the end).  this is a wild guess.
   1359  */
   1360 static struct mbuf *
   1361 ni6_nametodns(name, namelen, old)
   1362 	const char *name;
   1363 	int namelen;
   1364 	int old;	/* return pascal string if non-zero */
   1365 {
   1366 	struct mbuf *m;
   1367 	char *cp, *ep;
   1368 	const char *p, *q;
   1369 	int i, len, nterm;
   1370 
   1371 	if (old)
   1372 		len = namelen + 1;
   1373 	else
   1374 		len = MCLBYTES;
   1375 
   1376 	/* because MAXHOSTNAMELEN is usually 256, we use cluster mbuf */
   1377 	MGET(m, M_DONTWAIT, MT_DATA);
   1378 	if (m && len > MLEN) {
   1379 		MCLGET(m, M_DONTWAIT);
   1380 		if ((m->m_flags & M_EXT) == 0)
   1381 			goto fail;
   1382 	}
   1383 	if (!m)
   1384 		goto fail;
   1385 	m->m_next = NULL;
   1386 
   1387 	if (old) {
   1388 		m->m_len = len;
   1389 		*mtod(m, char *) = namelen;
   1390 		bcopy(name, mtod(m, char *) + 1, namelen);
   1391 		return m;
   1392 	} else {
   1393 		m->m_len = 0;
   1394 		cp = mtod(m, char *);
   1395 		ep = mtod(m, char *) + M_TRAILINGSPACE(m);
   1396 
   1397 		/* if not certain about my name, return empty buffer */
   1398 		if (namelen == 0)
   1399 			return m;
   1400 
   1401 		/*
   1402 		 * guess if it looks like shortened hostname, or FQDN.
   1403 		 * shortened hostname needs two trailing "\0".
   1404 		 */
   1405 		i = 0;
   1406 		for (p = name; p < name + namelen; p++) {
   1407 			if (*p && *p == '.')
   1408 				i++;
   1409 		}
   1410 		if (i < 2)
   1411 			nterm = 2;
   1412 		else
   1413 			nterm = 1;
   1414 
   1415 		p = name;
   1416 		while (cp < ep && p < name + namelen) {
   1417 			i = 0;
   1418 			for (q = p; q < name + namelen && *q && *q != '.'; q++)
   1419 				i++;
   1420 			/* result does not fit into mbuf */
   1421 			if (cp + i + 1 >= ep)
   1422 				goto fail;
   1423 			/* DNS label length restriction, RFC1035 page 8 */
   1424 			if (i >= 64)
   1425 				goto fail;
   1426 			*cp++ = i;
   1427 			bcopy(p, cp, i);
   1428 			cp += i;
   1429 			p = q;
   1430 			if (p < name + namelen && *p == '.')
   1431 				p++;
   1432 		}
   1433 		/* termination */
   1434 		if (cp + nterm >= ep)
   1435 			goto fail;
   1436 		while (nterm-- > 0)
   1437 			*cp++ = '\0';
   1438 		m->m_len = cp - mtod(m, char *);
   1439 		return m;
   1440 	}
   1441 
   1442 	panic("should not reach here");
   1443 	/*NOTREACHED*/
   1444 
   1445  fail:
   1446 	if (m)
   1447 		m_freem(m);
   1448 	return NULL;
   1449 }
   1450 
   1451 /*
   1452  * check if two DNS-encoded string matches.  takes care of truncated
   1453  * form (with \0\0 at the end).  no compression support.
   1454  */
   1455 static int
   1456 ni6_dnsmatch(a, alen, b, blen)
   1457 	const char *a;
   1458 	int alen;
   1459 	const char *b;
   1460 	int blen;
   1461 {
   1462 	const char *a0, *b0;
   1463 	int l;
   1464 
   1465 	/* simplest case - need validation? */
   1466 	if (alen == blen && bcmp(a, b, alen) == 0)
   1467 		return 1;
   1468 
   1469 	a0 = a;
   1470 	b0 = b;
   1471 
   1472 	/* termination is mandatory */
   1473 	if (alen < 2 || blen < 2)
   1474 		return 0;
   1475 	if (a0[alen - 1] != '\0' || b0[blen - 1] != '\0')
   1476 		return 0;
   1477 	alen--;
   1478 	blen--;
   1479 
   1480 	while (a - a0 < alen && b - b0 < blen) {
   1481 		if (a - a0 + 1 > alen || b - b0 + 1 > blen)
   1482 			return 0;
   1483 
   1484 		if ((signed char)a[0] < 0 || (signed char)b[0] < 0)
   1485 			return 0;
   1486 		/* we don't support compression yet */
   1487 		if (a[0] >= 64 || b[0] >= 64)
   1488 			return 0;
   1489 
   1490 		/* truncated case */
   1491 		if (a[0] == 0 && a - a0 == alen - 1)
   1492 			return 1;
   1493 		if (b[0] == 0 && b - b0 == blen - 1)
   1494 			return 1;
   1495 		if (a[0] == 0 || b[0] == 0)
   1496 			return 0;
   1497 
   1498 		if (a[0] != b[0])
   1499 			return 0;
   1500 		l = a[0];
   1501 		if (a - a0 + 1 + l > alen || b - b0 + 1 + l > blen)
   1502 			return 0;
   1503 		if (bcmp(a + 1, b + 1, l) != 0)
   1504 			return 0;
   1505 
   1506 		a += 1 + l;
   1507 		b += 1 + l;
   1508 	}
   1509 
   1510 	if (a - a0 == alen && b - b0 == blen)
   1511 		return 1;
   1512 	else
   1513 		return 0;
   1514 }
   1515 
   1516 /*
   1517  * calculate the number of addresses to be returned in the node info reply.
   1518  */
   1519 static int
   1520 ni6_addrs(ni6, m, ifpp)
   1521 	struct icmp6_nodeinfo *ni6;
   1522 	struct mbuf *m;
   1523 	struct ifnet **ifpp;
   1524 {
   1525 	register struct ifnet *ifp;
   1526 	register struct in6_ifaddr *ifa6;
   1527 	register struct ifaddr *ifa;
   1528 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
   1529 	int addrs = 0, addrsofif, iffound = 0;
   1530 
   1531 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list))
   1532 	{
   1533 		addrsofif = 0;
   1534 		for (ifa = ifp->if_addrlist.tqh_first; ifa;
   1535 		     ifa = ifa->ifa_list.tqe_next)
   1536 		{
   1537 			if (ifa->ifa_addr->sa_family != AF_INET6)
   1538 				continue;
   1539 			ifa6 = (struct in6_ifaddr *)ifa;
   1540 
   1541 			if (!(ni6->ni_flags & NI_NODEADDR_FLAG_ALL) &&
   1542 			    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
   1543 					       &ifa6->ia_addr.sin6_addr))
   1544 				iffound = 1;
   1545 
   1546 			/*
   1547 			 * IPv4-mapped addresses can only be returned by a
   1548 			 * Node Information proxy, since they represent
   1549 			 * addresses of IPv4-only nodes, which perforce do
   1550 			 * not implement this protocol.
   1551 			 * [icmp-name-lookups-05]
   1552 			 * So we don't support NI_NODEADDR_FLAG_COMPAT in
   1553 			 * this function at this moment.
   1554 			 */
   1555 
   1556 			if (ifa6->ia6_flags & IN6_IFF_ANYCAST)
   1557 				continue; /* we need only unicast addresses */
   1558 
   1559 			if ((ni6->ni_flags & (NI_NODEADDR_FLAG_LINKLOCAL |
   1560 					      NI_NODEADDR_FLAG_SITELOCAL |
   1561 					      NI_NODEADDR_FLAG_GLOBAL)) == 0)
   1562 				continue;
   1563 
   1564 			/* What do we have to do about ::1? */
   1565 			switch(in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
   1566 			 case IPV6_ADDR_SCOPE_LINKLOCAL:
   1567 				if (ni6->ni_flags & NI_NODEADDR_FLAG_LINKLOCAL)
   1568 					addrsofif++;
   1569 				break;
   1570 			 case IPV6_ADDR_SCOPE_SITELOCAL:
   1571 				if (ni6->ni_flags & NI_NODEADDR_FLAG_SITELOCAL)
   1572 					addrsofif++;
   1573 				break;
   1574 			 case IPV6_ADDR_SCOPE_GLOBAL:
   1575 				 if (ni6->ni_flags & NI_NODEADDR_FLAG_GLOBAL)
   1576 					 addrsofif++;
   1577 				 break;
   1578 			 default:
   1579 				 continue;
   1580 			}
   1581 		}
   1582 		if (iffound) {
   1583 			*ifpp = ifp;
   1584 			return(addrsofif);
   1585 		}
   1586 
   1587 		addrs += addrsofif;
   1588 	}
   1589 
   1590 	return(addrs);
   1591 }
   1592 
   1593 static int
   1594 ni6_store_addrs(ni6, nni6, ifp0, resid)
   1595 	struct icmp6_nodeinfo *ni6, *nni6;
   1596 	struct ifnet *ifp0;
   1597 	int resid;
   1598 {
   1599 	register struct ifnet *ifp = ifp0 ? ifp0 : TAILQ_FIRST(&ifnet);
   1600 	register struct in6_ifaddr *ifa6;
   1601 	register struct ifaddr *ifa;
   1602 	int docopy, copied = 0;
   1603 	u_char *cp = (u_char *)(nni6 + 1);
   1604 
   1605 	if (ifp0 == NULL && !(ni6->ni_flags & NI_NODEADDR_FLAG_ALL))
   1606 		return(0);	/* needless to copy */
   1607 
   1608 	for (; ifp; ifp = TAILQ_NEXT(ifp, if_list))
   1609 	{
   1610 		for (ifa = ifp->if_addrlist.tqh_first; ifa;
   1611 		     ifa = ifa->ifa_list.tqe_next)
   1612 		{
   1613 			docopy = 0;
   1614 
   1615 			if (ifa->ifa_addr->sa_family != AF_INET6)
   1616 				continue;
   1617 			ifa6 = (struct in6_ifaddr *)ifa;
   1618 
   1619 			if (ifa6->ia6_flags & IN6_IFF_ANYCAST) {
   1620 				/* just experimental. not in the spec. */
   1621 				if (ni6->ni_flags & NI_NODEADDR_FLAG_ANYCAST)
   1622 					docopy = 1;
   1623 				else
   1624 					continue;
   1625 			}
   1626 			else {	/* unicast address */
   1627 				if (ni6->ni_flags & NI_NODEADDR_FLAG_ANYCAST)
   1628 					continue;
   1629 				else
   1630 					docopy = 1;
   1631 			}
   1632 
   1633 			/* What do we have to do about ::1? */
   1634 			switch(in6_addrscope(&ifa6->ia_addr.sin6_addr)) {
   1635 			 case IPV6_ADDR_SCOPE_LINKLOCAL:
   1636 				if (ni6->ni_flags & NI_NODEADDR_FLAG_LINKLOCAL)
   1637 					docopy = 1;
   1638 				break;
   1639 			 case IPV6_ADDR_SCOPE_SITELOCAL:
   1640 				if (ni6->ni_flags & NI_NODEADDR_FLAG_SITELOCAL)
   1641 					docopy = 1;
   1642 				break;
   1643 			 case IPV6_ADDR_SCOPE_GLOBAL:
   1644 				 if (ni6->ni_flags & NI_NODEADDR_FLAG_GLOBAL)
   1645 					 docopy = 1;
   1646 				 break;
   1647 			 default:
   1648 				 continue;
   1649 			}
   1650 
   1651 			if (docopy) {
   1652 				if (resid < sizeof(struct in6_addr)) {
   1653 					/*
   1654 					 * We give up much more copy.
   1655 					 * Set the truncate flag and return.
   1656 					 */
   1657 					nni6->ni_flags |=
   1658 						NI_NODEADDR_FLAG_TRUNCATE;
   1659 					return(copied);
   1660 				}
   1661 				bcopy(&ifa6->ia_addr.sin6_addr, cp,
   1662 				      sizeof(struct in6_addr));
   1663 				/* XXX: KAME link-local hack; remove ifindex */
   1664 				if (IN6_IS_ADDR_LINKLOCAL(&ifa6->ia_addr.sin6_addr))
   1665 					((struct in6_addr *)cp)->s6_addr16[1] = 0;
   1666 				cp += sizeof(struct in6_addr);
   1667 				resid -= sizeof(struct in6_addr);
   1668 				copied += sizeof(struct in6_addr);
   1669 			}
   1670 		}
   1671 		if (ifp0)	/* we need search only on the specified IF */
   1672 			break;
   1673 	}
   1674 
   1675 	return(copied);
   1676 }
   1677 
   1678 /*
   1679  * XXX almost dup'ed code with rip6_input.
   1680  */
   1681 static int
   1682 icmp6_rip6_input(mp, off)
   1683 	struct	mbuf **mp;
   1684 	int	off;
   1685 {
   1686 	struct mbuf *m = *mp;
   1687 	register struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
   1688 	register struct in6pcb *in6p;
   1689 	struct in6pcb *last = NULL;
   1690 	struct sockaddr_in6 rip6src;
   1691 	struct icmp6_hdr *icmp6;
   1692 	struct mbuf *opts = NULL;
   1693 
   1694 #ifndef PULLDOWN_TEST
   1695 	/* this is assumed to be safe. */
   1696 	icmp6 = (struct icmp6_hdr *)((caddr_t)ip6 + off);
   1697 #else
   1698 	IP6_EXTHDR_GET(icmp6, struct icmp6_hdr *, m, off, sizeof(*icmp6));
   1699 	if (icmp6 == NULL) {
   1700 		/* m is already reclaimed */
   1701 		return IPPROTO_DONE;
   1702 	}
   1703 #endif
   1704 
   1705 	bzero(&rip6src, sizeof(rip6src));
   1706 	rip6src.sin6_len = sizeof(struct sockaddr_in6);
   1707 	rip6src.sin6_family = AF_INET6;
   1708 	/* KAME hack: recover scopeid */
   1709 	(void)in6_recoverscope(&rip6src, &ip6->ip6_src, m->m_pkthdr.rcvif);
   1710 
   1711 	for (in6p = rawin6pcb.in6p_next;
   1712 	     in6p != &rawin6pcb; in6p = in6p->in6p_next)
   1713 	{
   1714 		if (in6p->in6p_ip6_nxt != IPPROTO_ICMPV6)
   1715 			continue;
   1716 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr) &&
   1717 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_laddr, &ip6->ip6_dst))
   1718 			continue;
   1719 		if (!IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_faddr) &&
   1720 		   !IN6_ARE_ADDR_EQUAL(&in6p->in6p_faddr, &ip6->ip6_src))
   1721 			continue;
   1722 		if (in6p->in6p_icmp6filt
   1723 		    && ICMP6_FILTER_WILLBLOCK(icmp6->icmp6_type,
   1724 				 in6p->in6p_icmp6filt))
   1725 			continue;
   1726 		if (last) {
   1727 			struct	mbuf *n;
   1728 			if ((n = m_copy(m, 0, (int)M_COPYALL)) != NULL) {
   1729 				if (last->in6p_flags & IN6P_CONTROLOPTS)
   1730 					ip6_savecontrol(last, &opts, ip6, n);
   1731 				/* strip intermediate headers */
   1732 				m_adj(n, off);
   1733 				if (sbappendaddr(&last->in6p_socket->so_rcv,
   1734 						 (struct sockaddr *)&rip6src,
   1735 						 n, opts) == 0) {
   1736 					/* should notify about lost packet */
   1737 					m_freem(n);
   1738 					if (opts)
   1739 						m_freem(opts);
   1740 				} else
   1741 					sorwakeup(last->in6p_socket);
   1742 				opts = NULL;
   1743 			}
   1744 		}
   1745 		last = in6p;
   1746 	}
   1747 	if (last) {
   1748 		if (last->in6p_flags & IN6P_CONTROLOPTS)
   1749 			ip6_savecontrol(last, &opts, ip6, m);
   1750 		/* strip intermediate headers */
   1751 		m_adj(m, off);
   1752 		if (sbappendaddr(&last->in6p_socket->so_rcv,
   1753 				(struct sockaddr *)&rip6src, m, opts) == 0) {
   1754 			m_freem(m);
   1755 			if (opts)
   1756 				m_freem(opts);
   1757 		} else
   1758 			sorwakeup(last->in6p_socket);
   1759 	} else {
   1760 		m_freem(m);
   1761 		ip6stat.ip6s_delivered--;
   1762 	}
   1763 	return IPPROTO_DONE;
   1764 }
   1765 
   1766 /*
   1767  * Reflect the ip6 packet back to the source.
   1768  * OFF points to the icmp6 header, counted from the top of the mbuf.
   1769  */
   1770 void
   1771 icmp6_reflect(m, off)
   1772 	struct	mbuf *m;
   1773 	size_t off;
   1774 {
   1775 	struct ip6_hdr *ip6;
   1776 	struct icmp6_hdr *icmp6;
   1777 	struct in6_ifaddr *ia;
   1778 	struct in6_addr t, *src = 0;
   1779 	int plen;
   1780 	int type, code;
   1781 	struct ifnet *outif = NULL;
   1782 #ifdef COMPAT_RFC1885
   1783 	int mtu = IPV6_MMTU;
   1784 	struct sockaddr_in6 *sin6 = &icmp6_reflect_rt.ro_dst;
   1785 #endif
   1786 
   1787 	/* too short to reflect */
   1788 	if (off < sizeof(struct ip6_hdr)) {
   1789 		printf("sanity fail: off=%lx, sizeof(ip6)=%lx in %s:%d\n",
   1790 		       (u_long)off, (u_long)sizeof(struct ip6_hdr),
   1791 		       __FILE__, __LINE__);
   1792 		goto bad;
   1793 	}
   1794 
   1795 	/*
   1796 	 * If there are extra headers between IPv6 and ICMPv6, strip
   1797 	 * off that header first.
   1798 	 */
   1799 #ifdef DIAGNOSTIC
   1800 	if (sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) > MHLEN)
   1801 		panic("assumption failed in icmp6_reflect");
   1802 #endif
   1803 	if (off > sizeof(struct ip6_hdr)) {
   1804 		size_t l;
   1805 		struct ip6_hdr nip6;
   1806 
   1807 		l = off - sizeof(struct ip6_hdr);
   1808 		m_copydata(m, 0, sizeof(nip6), (caddr_t)&nip6);
   1809 		m_adj(m, l);
   1810 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
   1811 		if (m->m_len < l) {
   1812 			if ((m = m_pullup(m, l)) == NULL)
   1813 				return;
   1814 		}
   1815 		bcopy((caddr_t)&nip6, mtod(m, caddr_t), sizeof(nip6));
   1816 	} else /* off == sizeof(struct ip6_hdr) */ {
   1817 		size_t l;
   1818 		l = sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr);
   1819 		if (m->m_len < l) {
   1820 			if ((m = m_pullup(m, l)) == NULL)
   1821 				return;
   1822 		}
   1823 	}
   1824 	plen = m->m_pkthdr.len - sizeof(struct ip6_hdr);
   1825 	ip6 = mtod(m, struct ip6_hdr *);
   1826 	ip6->ip6_nxt = IPPROTO_ICMPV6;
   1827 	icmp6 = (struct icmp6_hdr *)(ip6 + 1);
   1828 	type = icmp6->icmp6_type; /* keep type for statistics */
   1829 	code = icmp6->icmp6_code; /* ditto. */
   1830 
   1831 	t = ip6->ip6_dst;
   1832 	/*
   1833 	 * ip6_input() drops a packet if its src is multicast.
   1834 	 * So, the src is never multicast.
   1835 	 */
   1836 	ip6->ip6_dst = ip6->ip6_src;
   1837 
   1838 	/* XXX hack for link-local addresses */
   1839 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
   1840 		ip6->ip6_dst.s6_addr16[1] =
   1841 			htons(m->m_pkthdr.rcvif->if_index);
   1842 	if (IN6_IS_ADDR_LINKLOCAL(&t))
   1843 		t.s6_addr16[1] = htons(m->m_pkthdr.rcvif->if_index);
   1844 
   1845 #ifdef COMPAT_RFC1885
   1846 	/*
   1847 	 * xxx guess MTU
   1848 	 * RFC 1885 requires that echo reply should be truncated if it
   1849 	 * does not fit in with (return) path MTU, but the description was
   1850 	 * removed in the new spec.
   1851 	 */
   1852 	if (icmp6_reflect_rt.ro_rt == 0 ||
   1853 	    ! (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, &ip6->ip6_dst))) {
   1854 		if (icmp6_reflect_rt.ro_rt) {
   1855 			icmp6_reflect_rt.ro_rt = 0;
   1856 		}
   1857 		bzero(sin6, sizeof(*sin6));
   1858 		sin6->sin6_family = PF_INET6;
   1859 		sin6->sin6_len = sizeof(struct sockaddr_in6);
   1860 		sin6->sin6_addr = ip6->ip6_dst;
   1861 
   1862 		rtalloc((struct route *)&icmp6_reflect_rt.ro_rt);
   1863 	}
   1864 
   1865 	if (icmp6_reflect_rt.ro_rt == 0)
   1866 		goto bad;
   1867 
   1868 	if ((icmp6_reflect_rt.ro_rt->rt_flags & RTF_HOST)
   1869 	    && mtu < icmp6_reflect_rt.ro_rt->rt_ifp->if_mtu)
   1870 		mtu = icmp6_reflect_rt.ro_rt->rt_rmx.rmx_mtu;
   1871 
   1872 	if (mtu < m->m_pkthdr.len) {
   1873 		plen -= (m->m_pkthdr.len - mtu);
   1874 		m_adj(m, mtu - m->m_pkthdr.len);
   1875 	}
   1876 #endif
   1877 	/*
   1878 	 * If the incoming packet was addressed directly to us(i.e. unicast),
   1879 	 * use dst as the src for the reply.
   1880 	 * The IN6_IFF_NOTREADY case would be VERY rare, but is possible
   1881 	 * (for example) when we encounter an error while forwarding procedure
   1882 	 * destined to a duplicated address of ours.
   1883 	 */
   1884 	for (ia = in6_ifaddr; ia; ia = ia->ia_next)
   1885 		if (IN6_ARE_ADDR_EQUAL(&t, &ia->ia_addr.sin6_addr) &&
   1886 		    (ia->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) == 0) {
   1887 			src = &t;
   1888 			break;
   1889 		}
   1890 	if (ia == NULL && IN6_IS_ADDR_LINKLOCAL(&t) && (m->m_flags & M_LOOP)) {
   1891 		/*
   1892 		 * This is the case if the dst is our link-local address
   1893 		 * and the sender is also ourseleves.
   1894 		 */
   1895 		src = &t;
   1896 	}
   1897 
   1898 	if (src == 0)
   1899 		/*
   1900 		 * This case matches to multicasts, our anycast, or unicasts
   1901 		 * that we do not own. Select a source address which has the
   1902 		 * same scope.
   1903 		 * XXX: for (non link-local) multicast addresses, this might
   1904 		 * not be a good choice.
   1905 		 */
   1906 		if ((ia = in6_ifawithscope(m->m_pkthdr.rcvif, &t)) != 0)
   1907 			src = &IA6_SIN6(ia)->sin6_addr;
   1908 
   1909 	if (src == 0)
   1910 		goto bad;
   1911 
   1912 	ip6->ip6_src = *src;
   1913 
   1914 	ip6->ip6_flow = 0;
   1915 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
   1916 	ip6->ip6_vfc |= IPV6_VERSION;
   1917 	ip6->ip6_nxt = IPPROTO_ICMPV6;
   1918 	if (m->m_pkthdr.rcvif) {
   1919 		/* XXX: This may not be the outgoing interface */
   1920 		ip6->ip6_hlim = nd_ifinfo[m->m_pkthdr.rcvif->if_index].chlim;
   1921 	}
   1922 
   1923 	icmp6->icmp6_cksum = 0;
   1924 	icmp6->icmp6_cksum = in6_cksum(m, IPPROTO_ICMPV6,
   1925 					sizeof(struct ip6_hdr), plen);
   1926 
   1927 	/*
   1928 	 * xxx option handling
   1929 	 */
   1930 
   1931 	m->m_flags &= ~(M_BCAST|M_MCAST);
   1932 #ifdef IPSEC
   1933 	/* Don't lookup socket */
   1934 	ipsec_setsocket(m, NULL);
   1935 #endif /*IPSEC*/
   1936 
   1937 #ifdef COMPAT_RFC1885
   1938 	ip6_output(m, NULL, &icmp6_reflect_rt, 0, NULL, &outif);
   1939 #else
   1940 	ip6_output(m, NULL, NULL, 0, NULL, &outif);
   1941 #endif
   1942 	if (outif)
   1943 		icmp6_ifoutstat_inc(outif, type, code);
   1944 
   1945 	return;
   1946 
   1947  bad:
   1948 	m_freem(m);
   1949 	return;
   1950 }
   1951 
   1952 void
   1953 icmp6_fasttimo()
   1954 {
   1955 
   1956 	mld6_fasttimeo();
   1957 }
   1958 
   1959 static const char *
   1960 icmp6_redirect_diag(src6, dst6, tgt6)
   1961 	struct in6_addr *src6;
   1962 	struct in6_addr *dst6;
   1963 	struct in6_addr *tgt6;
   1964 {
   1965 	static char buf[1024];
   1966 	snprintf(buf, sizeof(buf), "(src=%s dst=%s tgt=%s)",
   1967 		ip6_sprintf(src6), ip6_sprintf(dst6), ip6_sprintf(tgt6));
   1968 	return buf;
   1969 }
   1970 
   1971 void
   1972 icmp6_redirect_input(m, off)
   1973 	register struct mbuf *m;
   1974 	int off;
   1975 {
   1976 	struct ifnet *ifp = m->m_pkthdr.rcvif;
   1977 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
   1978 	struct nd_redirect *nd_rd;
   1979 	int icmp6len = ntohs(ip6->ip6_plen);
   1980 	char *lladdr = NULL;
   1981 	int lladdrlen = 0;
   1982 	u_char *redirhdr = NULL;
   1983 	int redirhdrlen = 0;
   1984 	struct rtentry *rt = NULL;
   1985 	int is_router;
   1986 	int is_onlink;
   1987 	struct in6_addr src6 = ip6->ip6_src;
   1988 	struct in6_addr redtgt6;
   1989 	struct in6_addr reddst6;
   1990 	union nd_opts ndopts;
   1991 
   1992 	if (!m || !ifp)
   1993 		return;
   1994 
   1995 	/* XXX if we are router, we don't update route by icmp6 redirect */
   1996 	if (ip6_forwarding)
   1997 		goto freeit;
   1998 	if (!icmp6_rediraccept)
   1999 		goto freeit;
   2000 
   2001 #ifndef PULLDOWN_TEST
   2002 	IP6_EXTHDR_CHECK(m, off, icmp6len,);
   2003 	nd_rd = (struct nd_redirect *)((caddr_t)ip6 + off);
   2004 #else
   2005 	IP6_EXTHDR_GET(nd_rd, struct nd_redirect *, m, off, icmp6len);
   2006 	if (nd_rd == NULL) {
   2007 		icmp6stat.icp6s_tooshort++;
   2008 		return;
   2009 	}
   2010 #endif
   2011 	redtgt6 = nd_rd->nd_rd_target;
   2012 	reddst6 = nd_rd->nd_rd_dst;
   2013 
   2014 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
   2015 		redtgt6.s6_addr16[1] = htons(ifp->if_index);
   2016 	if (IN6_IS_ADDR_LINKLOCAL(&reddst6))
   2017 		reddst6.s6_addr16[1] = htons(ifp->if_index);
   2018 
   2019 	/* validation */
   2020 	if (!IN6_IS_ADDR_LINKLOCAL(&src6)) {
   2021 		log(LOG_ERR,
   2022 			"ICMP6 redirect sent from %s rejected; "
   2023 			"must be from linklocal\n", ip6_sprintf(&src6));
   2024 		goto freeit;
   2025 	}
   2026 	if (ip6->ip6_hlim != 255) {
   2027 		log(LOG_ERR,
   2028 			"ICMP6 redirect sent from %s rejected; "
   2029 			"hlim=%d (must be 255)\n",
   2030 			ip6_sprintf(&src6), ip6->ip6_hlim);
   2031 		goto freeit;
   2032 	}
   2033     {
   2034 	/* ip6->ip6_src must be equal to gw for icmp6->icmp6_reddst */
   2035 	struct sockaddr_in6 sin6;
   2036 	struct in6_addr *gw6;
   2037 
   2038 	bzero(&sin6, sizeof(sin6));
   2039 	sin6.sin6_family = AF_INET6;
   2040 	sin6.sin6_len = sizeof(struct sockaddr_in6);
   2041 	bcopy(&reddst6, &sin6.sin6_addr, sizeof(reddst6));
   2042 	rt = rtalloc1((struct sockaddr *)&sin6, 0);
   2043 	if (rt) {
   2044 		gw6 = &(((struct sockaddr_in6 *)rt->rt_gateway)->sin6_addr);
   2045 		if (bcmp(&src6, gw6, sizeof(struct in6_addr)) != 0) {
   2046 			log(LOG_ERR,
   2047 				"ICMP6 redirect rejected; "
   2048 				"not equal to gw-for-src=%s (must be same): "
   2049 				"%s\n",
   2050 				ip6_sprintf(gw6),
   2051 				icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
   2052 			RTFREE(rt);
   2053 			goto freeit;
   2054 		}
   2055 	} else {
   2056 		log(LOG_ERR,
   2057 			"ICMP6 redirect rejected; "
   2058 			"no route found for redirect dst: %s\n",
   2059 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
   2060 		goto freeit;
   2061 	}
   2062 	RTFREE(rt);
   2063 	rt = NULL;
   2064     }
   2065 	if (IN6_IS_ADDR_MULTICAST(&reddst6)) {
   2066 		log(LOG_ERR,
   2067 			"ICMP6 redirect rejected; "
   2068 			"redirect dst must be unicast: %s\n",
   2069 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
   2070 		goto freeit;
   2071 	}
   2072 
   2073 	is_router = is_onlink = 0;
   2074 	if (IN6_IS_ADDR_LINKLOCAL(&redtgt6))
   2075 		is_router = 1;	/* router case */
   2076 	if (bcmp(&redtgt6, &reddst6, sizeof(redtgt6)) == 0)
   2077 		is_onlink = 1;	/* on-link destination case */
   2078 	if (!is_router && !is_onlink) {
   2079 		log(LOG_ERR,
   2080 			"ICMP6 redirect rejected; "
   2081 			"neither router case nor onlink case: %s\n",
   2082 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
   2083 		goto freeit;
   2084 	}
   2085 	/* validation passed */
   2086 
   2087 	icmp6len -= sizeof(*nd_rd);
   2088 	nd6_option_init(nd_rd + 1, icmp6len, &ndopts);
   2089 	if (nd6_options(&ndopts) < 0) {
   2090 		log(LOG_INFO, "icmp6_redirect_input: "
   2091 			"invalid ND option, rejected: %s\n",
   2092 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
   2093 		goto freeit;
   2094 	}
   2095 
   2096 	if (ndopts.nd_opts_tgt_lladdr) {
   2097 		lladdr = (char *)(ndopts.nd_opts_tgt_lladdr + 1);
   2098 		lladdrlen = ndopts.nd_opts_tgt_lladdr->nd_opt_len << 3;
   2099 	}
   2100 
   2101 	if (ndopts.nd_opts_rh) {
   2102 		redirhdrlen = ndopts.nd_opts_rh->nd_opt_rh_len;
   2103 		redirhdr = (u_char *)(ndopts.nd_opts_rh + 1); /* xxx */
   2104 	}
   2105 
   2106 	if (lladdr && ((ifp->if_addrlen + 2 + 7) & ~7) != lladdrlen) {
   2107 		log(LOG_INFO,
   2108 			"icmp6_redirect_input: lladdrlen mismatch for %s "
   2109 			"(if %d, icmp6 packet %d): %s\n",
   2110 			ip6_sprintf(&redtgt6), ifp->if_addrlen, lladdrlen - 2,
   2111 			icmp6_redirect_diag(&src6, &reddst6, &redtgt6));
   2112 	}
   2113 
   2114 	/* RFC 2461 8.3 */
   2115 	nd6_cache_lladdr(ifp, &redtgt6, lladdr, lladdrlen, ND_REDIRECT,
   2116 			 is_onlink ? ND_REDIRECT_ONLINK : ND_REDIRECT_ROUTER);
   2117 
   2118 	if (!is_onlink) {	/* better router case. perform rtredirect. */
   2119 		/* perform rtredirect */
   2120 		struct sockaddr_in6 sdst;
   2121 		struct sockaddr_in6 sgw;
   2122 		struct sockaddr_in6 ssrc;
   2123 
   2124 		bzero(&sdst, sizeof(sdst));
   2125 		bzero(&sgw, sizeof(sgw));
   2126 		bzero(&ssrc, sizeof(ssrc));
   2127 		sdst.sin6_family = sgw.sin6_family = ssrc.sin6_family = AF_INET6;
   2128 		sdst.sin6_len = sgw.sin6_len = ssrc.sin6_len =
   2129 			sizeof(struct sockaddr_in6);
   2130 		bcopy(&redtgt6, &sgw.sin6_addr, sizeof(struct in6_addr));
   2131 		bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
   2132 		bcopy(&src6, &ssrc.sin6_addr, sizeof(struct in6_addr));
   2133 		rtredirect((struct sockaddr *)&sdst, (struct sockaddr *)&sgw,
   2134 			   (struct sockaddr *)NULL, RTF_GATEWAY | RTF_HOST,
   2135 			   (struct sockaddr *)&ssrc,
   2136 			   (struct rtentry **)NULL
   2137 			   );
   2138 	}
   2139 	/* finally update cached route in each socket via pfctlinput */
   2140     {
   2141 	struct sockaddr_in6 sdst;
   2142 #if 1
   2143 #else
   2144 	struct ip6protosw *pr;
   2145 #endif
   2146 
   2147 	bzero(&sdst, sizeof(sdst));
   2148 	sdst.sin6_family = AF_INET6;
   2149 	sdst.sin6_len = sizeof(struct sockaddr_in6);
   2150 	bcopy(&reddst6, &sdst.sin6_addr, sizeof(struct in6_addr));
   2151 #if 1
   2152 	pfctlinput(PRC_REDIRECT_HOST, (struct sockaddr *)&sdst);
   2153 #else
   2154 	/*
   2155 	 * do not use pfctlinput() here, we have different prototype for
   2156 	 * xx_ctlinput() in ip6proto.
   2157 	 */
   2158 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
   2159 	     pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW;
   2160 	     pr++) {
   2161 		if (pr->pr_ctlinput) {
   2162 			(*pr->pr_ctlinput)(PRC_REDIRECT_HOST,
   2163 				(struct sockaddr *)&sdst, NULL, NULL, 0);
   2164 		}
   2165 	}
   2166 #endif
   2167 #ifdef IPSEC
   2168 	key_sa_routechange((struct sockaddr *)&sdst);
   2169 #endif
   2170     }
   2171 
   2172  freeit:
   2173 	m_freem(m);
   2174 }
   2175 
   2176 void
   2177 icmp6_redirect_output(m0, rt)
   2178 	struct mbuf *m0;
   2179 	struct rtentry *rt;
   2180 {
   2181 	struct ifnet *ifp;	/* my outgoing interface */
   2182 	struct in6_addr *ifp_ll6;
   2183 	struct in6_addr *router_ll6;
   2184 	struct ip6_hdr *sip6;	/* m0 as struct ip6_hdr */
   2185 	struct mbuf *m = NULL;	/* newly allocated one */
   2186 	struct ip6_hdr *ip6;	/* m as struct ip6_hdr */
   2187 	struct nd_redirect *nd_rd;
   2188 	size_t maxlen;
   2189 	u_char *p;
   2190 	struct ifnet *outif = NULL;
   2191 	struct sockaddr_in6 src_sa;
   2192 
   2193 	icmp6_errcount(&icmp6stat.icp6s_outerrhist, ND_REDIRECT, 0);
   2194 
   2195 	/* if we are not router, we don't send icmp6 redirect */
   2196 	if (!ip6_forwarding || ip6_accept_rtadv)
   2197 		goto fail;
   2198 
   2199 	/* sanity check */
   2200 	if (!m0 || !rt || !(rt->rt_flags & RTF_UP) || !(ifp = rt->rt_ifp))
   2201 		goto fail;
   2202 
   2203 	/*
   2204 	 * Address check:
   2205 	 *  the source address must identify a neighbor, and
   2206 	 *  the destination address must not be a multicast address
   2207 	 *  [RFC 2461, sec 8.2]
   2208 	 */
   2209 	sip6 = mtod(m0, struct ip6_hdr *);
   2210 	bzero(&src_sa, sizeof(src_sa));
   2211 	src_sa.sin6_family = AF_INET6;
   2212 	src_sa.sin6_len = sizeof(src_sa);
   2213 	src_sa.sin6_addr = sip6->ip6_src;
   2214 	/* we don't currently use sin6_scope_id, but eventually use it */
   2215 	src_sa.sin6_scope_id = in6_addr2scopeid(ifp, &sip6->ip6_src);
   2216 	if (nd6_is_addr_neighbor(&src_sa, ifp) == 0)
   2217 		goto fail;
   2218 	if (IN6_IS_ADDR_MULTICAST(&sip6->ip6_dst))
   2219 		goto fail;	/* what should we do here? */
   2220 
   2221 	/* rate limit */
   2222 	if (icmp6_ratelimit(&sip6->ip6_src, ND_REDIRECT, 0))
   2223 		goto fail;
   2224 
   2225 	/*
   2226 	 * Since we are going to append up to 1280 bytes (= IPV6_MMTU),
   2227 	 * we almost always ask for an mbuf cluster for simplicity.
   2228 	 * (MHLEN < IPV6_MMTU is almost always true)
   2229 	 */
   2230 #if IPV6_MMTU >= MCLBYTES
   2231 # error assumption failed about IPV6_MMTU and MCLBYTES
   2232 #endif
   2233 	MGETHDR(m, M_DONTWAIT, MT_HEADER);
   2234 	if (m && IPV6_MMTU >= MHLEN)
   2235 		MCLGET(m, M_DONTWAIT);
   2236 	if (!m)
   2237 		goto fail;
   2238 	maxlen = (m->m_flags & M_EXT) ? MCLBYTES : MHLEN;
   2239 	maxlen = min(IPV6_MMTU, maxlen);
   2240 	/* just for safety */
   2241 	if (maxlen < sizeof(struct ip6_hdr) + sizeof(struct icmp6_hdr) +
   2242 	    ((sizeof(struct nd_opt_hdr) + ifp->if_addrlen + 7) & ~7)) {
   2243 		goto fail;
   2244 	}
   2245 
   2246 	{
   2247 		/* get ip6 linklocal address for ifp(my outgoing interface). */
   2248 		struct in6_ifaddr *ia;
   2249 		if ((ia = in6ifa_ifpforlinklocal(ifp,
   2250 						 IN6_IFF_NOTREADY|
   2251 						 IN6_IFF_ANYCAST)) == NULL)
   2252 			goto fail;
   2253 		ifp_ll6 = &ia->ia_addr.sin6_addr;
   2254 	}
   2255 
   2256 	/* get ip6 linklocal address for the router. */
   2257 	if (rt->rt_gateway && (rt->rt_flags & RTF_GATEWAY)) {
   2258 		struct sockaddr_in6 *sin6;
   2259 		sin6 = (struct sockaddr_in6 *)rt->rt_gateway;
   2260 		router_ll6 = &sin6->sin6_addr;
   2261 		if (!IN6_IS_ADDR_LINKLOCAL(router_ll6))
   2262 			router_ll6 = (struct in6_addr *)NULL;
   2263 	} else
   2264 		router_ll6 = (struct in6_addr *)NULL;
   2265 
   2266 	/* ip6 */
   2267 	ip6 = mtod(m, struct ip6_hdr *);
   2268 	ip6->ip6_flow = 0;
   2269 	ip6->ip6_vfc &= ~IPV6_VERSION_MASK;
   2270 	ip6->ip6_vfc |= IPV6_VERSION;
   2271 	/* ip6->ip6_plen will be set later */
   2272 	ip6->ip6_nxt = IPPROTO_ICMPV6;
   2273 	ip6->ip6_hlim = 255;
   2274 	/* ip6->ip6_src must be linklocal addr for my outgoing if. */
   2275 	bcopy(ifp_ll6, &ip6->ip6_src, sizeof(struct in6_addr));
   2276 	bcopy(&sip6->ip6_src, &ip6->ip6_dst, sizeof(struct in6_addr));
   2277 
   2278 	/* ND Redirect */
   2279 	nd_rd = (struct nd_redirect *)(ip6 + 1);
   2280 	nd_rd->nd_rd_type = ND_REDIRECT;
   2281 	nd_rd->nd_rd_code = 0;
   2282 	nd_rd->nd_rd_reserved = 0;
   2283 	if (rt->rt_flags & RTF_GATEWAY) {
   2284 		/*
   2285 		 * nd_rd->nd_rd_target must be a link-local address in
   2286 		 * better router cases.
   2287 		 */
   2288 		if (!router_ll6)
   2289 			goto fail;
   2290 		bcopy(router_ll6, &nd_rd->nd_rd_target,
   2291 		      sizeof(nd_rd->nd_rd_target));
   2292 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
   2293 		      sizeof(nd_rd->nd_rd_dst));
   2294 	} else {
   2295 		/* make sure redtgt == reddst */
   2296 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_target,
   2297 		      sizeof(nd_rd->nd_rd_target));
   2298 		bcopy(&sip6->ip6_dst, &nd_rd->nd_rd_dst,
   2299 		      sizeof(nd_rd->nd_rd_dst));
   2300 	}
   2301 
   2302 	p = (u_char *)(nd_rd + 1);
   2303 
   2304 	if (!router_ll6)
   2305 		goto nolladdropt;
   2306 
   2307     {
   2308 	/* target lladdr option */
   2309 	struct rtentry *rt_router = NULL;
   2310 	int len;
   2311 	struct sockaddr_dl *sdl;
   2312 	struct nd_opt_hdr *nd_opt;
   2313 	char *lladdr;
   2314 
   2315 	rt_router = nd6_lookup(router_ll6, 0, ifp);
   2316 	if (!rt_router)
   2317 		goto nolladdropt;
   2318 	len = sizeof(*nd_opt) + ifp->if_addrlen;
   2319 	len = (len + 7) & ~7;	/*round by 8*/
   2320 	/* safety check */
   2321 	if (len + (p - (u_char *)ip6) > maxlen)
   2322 		goto nolladdropt;
   2323 	if (!(rt_router->rt_flags & RTF_GATEWAY) &&
   2324 	    (rt_router->rt_flags & RTF_LLINFO) &&
   2325 	    (rt_router->rt_gateway->sa_family == AF_LINK) &&
   2326 	    (sdl = (struct sockaddr_dl *)rt_router->rt_gateway) &&
   2327 	    sdl->sdl_alen) {
   2328 		nd_opt = (struct nd_opt_hdr *)p;
   2329 		nd_opt->nd_opt_type = ND_OPT_TARGET_LINKADDR;
   2330 		nd_opt->nd_opt_len = len >> 3;
   2331 		lladdr = (char *)(nd_opt + 1);
   2332 		bcopy(LLADDR(sdl), lladdr, ifp->if_addrlen);
   2333 		p += len;
   2334 	}
   2335     }
   2336 nolladdropt:;
   2337 
   2338 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
   2339 
   2340 	/* just to be safe */
   2341 	if (m0->m_flags & M_DECRYPTED)
   2342 		goto noredhdropt;
   2343 	if (p - (u_char *)ip6 > maxlen)
   2344 		goto noredhdropt;
   2345 
   2346     {
   2347 	/* redirected header option */
   2348 	int len;
   2349 	struct nd_opt_rd_hdr *nd_opt_rh;
   2350 
   2351 	/*
   2352 	 * compute the maximum size for icmp6 redirect header option.
   2353 	 * XXX room for auth header?
   2354 	 */
   2355 	len = maxlen - (p - (u_char *)ip6);
   2356 	len &= ~7;
   2357 
   2358 	/* This is just for simplicity. */
   2359 	if (m0->m_pkthdr.len != m0->m_len) {
   2360 		if (m0->m_next) {
   2361 			m_freem(m0->m_next);
   2362 			m0->m_next = NULL;
   2363 		}
   2364 		m0->m_pkthdr.len = m0->m_len;
   2365 	}
   2366 
   2367 	/*
   2368 	 * Redirected header option spec (RFC2461 4.6.3) talks nothing
   2369 	 * about padding/truncate rule for the original IP packet.
   2370 	 * From the discussion on IPv6imp in Feb 1999, the consensus was:
   2371 	 * - "attach as much as possible" is the goal
   2372 	 * - pad if not aligned (original size can be guessed by original
   2373 	 *   ip6 header)
   2374 	 * Following code adds the padding if it is simple enough,
   2375 	 * and truncates if not.
   2376 	 */
   2377 	if (m0->m_next || m0->m_pkthdr.len != m0->m_len)
   2378 		panic("assumption failed in %s:%d\n", __FILE__, __LINE__);
   2379 
   2380 	if (len - sizeof(*nd_opt_rh) < m0->m_pkthdr.len) {
   2381 		/* not enough room, truncate */
   2382 		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
   2383 	} else {
   2384 		/* enough room, pad or truncate */
   2385 		size_t extra;
   2386 
   2387 		extra = m0->m_pkthdr.len % 8;
   2388 		if (extra) {
   2389 			/* pad if easy enough, truncate if not */
   2390 			if (8 - extra <= M_TRAILINGSPACE(m0)) {
   2391 				/* pad */
   2392 				m0->m_len += (8 - extra);
   2393 				m0->m_pkthdr.len += (8 - extra);
   2394 			} else {
   2395 				/* truncate */
   2396 				m0->m_pkthdr.len -= extra;
   2397 				m0->m_len -= extra;
   2398 			}
   2399 		}
   2400 		len = m0->m_pkthdr.len + sizeof(*nd_opt_rh);
   2401 		m0->m_pkthdr.len = m0->m_len = len - sizeof(*nd_opt_rh);
   2402 	}
   2403 
   2404 	nd_opt_rh = (struct nd_opt_rd_hdr *)p;
   2405 	bzero(nd_opt_rh, sizeof(*nd_opt_rh));
   2406 	nd_opt_rh->nd_opt_rh_type = ND_OPT_REDIRECTED_HEADER;
   2407 	nd_opt_rh->nd_opt_rh_len = len >> 3;
   2408 	p += sizeof(*nd_opt_rh);
   2409 	m->m_pkthdr.len = m->m_len = p - (u_char *)ip6;
   2410 
   2411 	/* connect m0 to m */
   2412 	m->m_next = m0;
   2413 	m->m_pkthdr.len = m->m_len + m0->m_len;
   2414     }
   2415 noredhdropt:;
   2416 
   2417 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_src))
   2418 		sip6->ip6_src.s6_addr16[1] = 0;
   2419 	if (IN6_IS_ADDR_LINKLOCAL(&sip6->ip6_dst))
   2420 		sip6->ip6_dst.s6_addr16[1] = 0;
   2421 #if 0
   2422 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_src))
   2423 		ip6->ip6_src.s6_addr16[1] = 0;
   2424 	if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst))
   2425 		ip6->ip6_dst.s6_addr16[1] = 0;
   2426 #endif
   2427 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_target))
   2428 		nd_rd->nd_rd_target.s6_addr16[1] = 0;
   2429 	if (IN6_IS_ADDR_LINKLOCAL(&nd_rd->nd_rd_dst))
   2430 		nd_rd->nd_rd_dst.s6_addr16[1] = 0;
   2431 
   2432 	ip6->ip6_plen = htons(m->m_pkthdr.len - sizeof(struct ip6_hdr));
   2433 
   2434 	nd_rd->nd_rd_cksum = 0;
   2435 	nd_rd->nd_rd_cksum
   2436 		= in6_cksum(m, IPPROTO_ICMPV6, sizeof(*ip6), ntohs(ip6->ip6_plen));
   2437 
   2438 	/* send the packet to outside... */
   2439 #ifdef IPSEC
   2440 	/* Don't lookup socket */
   2441 	ipsec_setsocket(m, NULL);
   2442 #endif /*IPSEC*/
   2443 	ip6_output(m, NULL, NULL, 0, NULL, &outif);
   2444 	if (outif) {
   2445 		icmp6_ifstat_inc(outif, ifs6_out_msg);
   2446 		icmp6_ifstat_inc(outif, ifs6_out_redirect);
   2447 	}
   2448 	icmp6stat.icp6s_outhist[ND_REDIRECT]++;
   2449 
   2450 	return;
   2451 
   2452 fail:
   2453 	if (m)
   2454 		m_freem(m);
   2455 	if (m0)
   2456 		m_freem(m0);
   2457 }
   2458 
   2459 /*
   2460  * ICMPv6 socket option processing.
   2461  */
   2462 int
   2463 icmp6_ctloutput(op, so, level, optname, mp)
   2464 	int op;
   2465 	struct socket *so;
   2466 	int level, optname;
   2467 	struct mbuf **mp;
   2468 {
   2469 	register struct in6pcb *in6p = sotoin6pcb(so);
   2470 	register struct mbuf *m = *mp;
   2471 	int error = 0;
   2472 
   2473 	if (level != IPPROTO_ICMPV6) {
   2474 		error = EINVAL;
   2475 		if (op == PRCO_SETOPT && m)
   2476 			(void)m_free(m);
   2477 	} else switch(op) {
   2478 	 case PRCO_SETOPT:
   2479 		 switch (optname) {
   2480 		  case ICMP6_FILTER:
   2481 		  {
   2482 			  struct icmp6_filter *p;
   2483 
   2484 			  p = mtod(m, struct icmp6_filter *);
   2485 			  if (!p || !in6p->in6p_icmp6filt) {
   2486 				  error = EINVAL;
   2487 				  break;
   2488 			  }
   2489 			  bcopy(p, in6p->in6p_icmp6filt,
   2490 				sizeof(struct icmp6_filter));
   2491 			  error = 0;
   2492 			  break;
   2493 		  }
   2494 
   2495 		  default:
   2496 			  error = ENOPROTOOPT;
   2497 			  break;
   2498 		 }
   2499 		 if (m)
   2500 			 (void)m_free(m);
   2501 		 break;
   2502 
   2503 	 case PRCO_GETOPT:
   2504 		 switch (optname) {
   2505 		  case ICMP6_FILTER:
   2506 		  {
   2507 			  struct icmp6_filter *p;
   2508 
   2509 			  if (!in6p->in6p_icmp6filt) {
   2510 				  error = EINVAL;
   2511 				  break;
   2512 			  }
   2513 			  *mp = m = m_get(M_WAIT, MT_SOOPTS);
   2514 			  m->m_len = sizeof(struct icmp6_filter);
   2515 			  p = mtod(m, struct icmp6_filter *);
   2516 			  bcopy(in6p->in6p_icmp6filt, p,
   2517 				sizeof(struct icmp6_filter));
   2518 			  error = 0;
   2519 			  break;
   2520 		  }
   2521 
   2522 		  default:
   2523 			  error = ENOPROTOOPT;
   2524 			  break;
   2525 		 }
   2526 		 break;
   2527 	}
   2528 
   2529 	return(error);
   2530 }
   2531 
   2532 /*
   2533  * Perform rate limit check.
   2534  * Returns 0 if it is okay to send the icmp6 packet.
   2535  * Returns 1 if the router SHOULD NOT send this icmp6 packet due to rate
   2536  * limitation.
   2537  *
   2538  * There are two limitations defined:
   2539  * - pps limit: ICMPv6 error packet cannot exceed defined packet-per-second.
   2540  *   we measure it every 0.2 second, since fasttimo works every 0.2 second.
   2541  * - rate limit: ICMPv6 error packet cannot appear more than once per
   2542  *   defined interval.
   2543  * In any case, if we perform rate limitation, we'll see jitter in the ICMPv6
   2544  * error packets.
   2545  *
   2546  * XXX per-destination/type check necessary?
   2547  */
   2548 static int
   2549 icmp6_ratelimit(dst, type, code)
   2550 	const struct in6_addr *dst;	/* not used at this moment */
   2551 	const int type;			/* not used at this moment */
   2552 	const int code;			/* not used at this moment */
   2553 {
   2554 	int ret;
   2555 
   2556 	ret = 0;	/*okay to send*/
   2557 
   2558 	/* PPS limit */
   2559 	if (!ppsratecheck(&icmp6errppslim_last, &icmp6errpps_count,
   2560 	    icmp6errppslim)) {
   2561 		/* The packet is subject to rate limit */
   2562 		ret++;
   2563 	}
   2564 
   2565 	if (!ratecheck(&icmp6errratelim_last, &icmp6errratelim)) {
   2566 		/* The packet is subject to rate limit */
   2567 		ret++;
   2568 	}
   2569 
   2570 	return ret;
   2571 }
   2572 
   2573 static struct rtentry *
   2574 icmp6_mtudisc_clone(dst)
   2575 	struct sockaddr *dst;
   2576 {
   2577 	struct rtentry *rt;
   2578 	int    error;
   2579 
   2580 	rt = rtalloc1(dst, 1);
   2581 	if (rt == 0)
   2582 		return NULL;
   2583 
   2584 	/* If we didn't get a host route, allocate one */
   2585 	if ((rt->rt_flags & RTF_HOST) == 0) {
   2586 		struct rtentry *nrt;
   2587 
   2588 		error = rtrequest((int) RTM_ADD, dst,
   2589 		    (struct sockaddr *) rt->rt_gateway,
   2590 		    (struct sockaddr *) 0,
   2591 		    RTF_GATEWAY | RTF_HOST | RTF_DYNAMIC, &nrt);
   2592 		if (error) {
   2593 			rtfree(rt);
   2594 			rtfree(nrt);
   2595 			return NULL;
   2596 		}
   2597 		nrt->rt_rmx = rt->rt_rmx;
   2598 		rtfree(rt);
   2599 		rt = nrt;
   2600 	}
   2601 	error = rt_timer_add(rt, icmp6_mtudisc_timeout,
   2602 			icmp6_mtudisc_timeout_q);
   2603 	if (error) {
   2604 		rtfree(rt);
   2605 		return NULL;
   2606 	}
   2607 
   2608 	return rt;	/* caller need to call rtfree() */
   2609 }
   2610 
   2611 static void
   2612 icmp6_mtudisc_timeout(rt, r)
   2613 	struct rtentry *rt;
   2614 	struct rttimer *r;
   2615 {
   2616 	if (rt == NULL)
   2617 		panic("icmp6_mtudisc_timeout: bad route to timeout");
   2618 	if ((rt->rt_flags & (RTF_DYNAMIC | RTF_HOST)) ==
   2619 	    (RTF_DYNAMIC | RTF_HOST)) {
   2620 		rtrequest((int) RTM_DELETE, (struct sockaddr *)rt_key(rt),
   2621 		    rt->rt_gateway, rt_mask(rt), rt->rt_flags, 0);
   2622 	} else {
   2623 		if ((rt->rt_rmx.rmx_locks & RTV_MTU) == 0) {
   2624 			rt->rt_rmx.rmx_mtu = 0;
   2625 		}
   2626 	}
   2627 }
   2628 
   2629 #include <uvm/uvm_extern.h>
   2630 #include <sys/sysctl.h>
   2631 
   2632 int
   2633 icmp6_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
   2634 	int *name;
   2635 	u_int namelen;
   2636 	void *oldp;
   2637 	size_t *oldlenp;
   2638 	void *newp;
   2639 	size_t newlen;
   2640 {
   2641 
   2642 	/* All sysctl names at this level are terminal. */
   2643 	if (namelen != 1)
   2644 		return ENOTDIR;
   2645 
   2646 	switch (name[0]) {
   2647 
   2648 	case ICMPV6CTL_REDIRACCEPT:
   2649 		return sysctl_int(oldp, oldlenp, newp, newlen,
   2650 				&icmp6_rediraccept);
   2651 	case ICMPV6CTL_REDIRTIMEOUT:
   2652 		return sysctl_int(oldp, oldlenp, newp, newlen,
   2653 				&icmp6_redirtimeout);
   2654 	case ICMPV6CTL_STATS:
   2655 		return sysctl_rdstruct(oldp, oldlenp, newp,
   2656 				&icmp6stat, sizeof(icmp6stat));
   2657 	case ICMPV6CTL_ERRRATELIMIT:
   2658 	    {
   2659 		int rate_usec, error, s;
   2660 
   2661 		/*
   2662 		 * The sysctl specifies the rate in usec-between-icmp,
   2663 		 * so we must convert from/to a timeval.
   2664 		 */
   2665 		rate_usec = (icmp6errratelim.tv_sec * 1000000) +
   2666 		    icmp6errratelim.tv_usec;
   2667 		error = sysctl_int(oldp, oldlenp, newp, newlen, &rate_usec);
   2668 		if (error)
   2669 			return (error);
   2670 		if (rate_usec < 0)
   2671 			return (EINVAL);
   2672 		s = splsoftnet();
   2673 		icmp6errratelim.tv_sec = rate_usec / 1000000;
   2674 		icmp6errratelim.tv_usec = rate_usec % 1000000;
   2675 		splx(s);
   2676 
   2677 		return (0);
   2678 	    }
   2679 	case ICMPV6CTL_ND6_PRUNE:
   2680 		return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_prune);
   2681 	case ICMPV6CTL_ND6_DELAY:
   2682 		return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_delay);
   2683 	case ICMPV6CTL_ND6_UMAXTRIES:
   2684 		return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_umaxtries);
   2685 	case ICMPV6CTL_ND6_MMAXTRIES:
   2686 		return sysctl_int(oldp, oldlenp, newp, newlen, &nd6_mmaxtries);
   2687 	case ICMPV6CTL_ND6_USELOOPBACK:
   2688 		return sysctl_int(oldp, oldlenp, newp, newlen,
   2689 				&nd6_useloopback);
   2690 	case ICMPV6CTL_NODEINFO:
   2691 		return sysctl_int(oldp, oldlenp, newp, newlen, &icmp6_nodeinfo);
   2692 	case ICMPV6CTL_ERRPPSLIMIT:
   2693 		return sysctl_int(oldp, oldlenp, newp, newlen, &icmp6errppslim);
   2694 	case ICMPV6CTL_ND6_MAXNUDHINT:
   2695 		return sysctl_int(oldp, oldlenp, newp, newlen,
   2696 				&nd6_maxnudhint);
   2697 	default:
   2698 		return ENOPROTOOPT;
   2699 	}
   2700 	/* NOTREACHED */
   2701 }
   2702