Home | History | Annotate | Line # | Download | only in netinet6
ip6_output.c revision 1.23.2.7
      1 /*	$NetBSD: ip6_output.c,v 1.23.2.7 2002/09/23 13:48:36 itojun Exp $	*/
      2 /*	$KAME: ip6_output.c,v 1.109 2000/05/31 05:03:09 jinmei 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, 1990, 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_output.c	8.3 (Berkeley) 1/21/94
     66  */
     67 
     68 #include "opt_inet.h"
     69 #include "opt_ipsec.h"
     70 #include "opt_pfil_hooks.h"
     71 
     72 #include <sys/param.h>
     73 #include <sys/malloc.h>
     74 #include <sys/mbuf.h>
     75 #include <sys/errno.h>
     76 #include <sys/protosw.h>
     77 #include <sys/socket.h>
     78 #include <sys/socketvar.h>
     79 #include <sys/systm.h>
     80 #include <sys/proc.h>
     81 
     82 #include <net/if.h>
     83 #include <net/route.h>
     84 #ifdef PFIL_HOOKS
     85 #include <net/pfil.h>
     86 #endif
     87 
     88 #include <netinet/in.h>
     89 #include <netinet/in_var.h>
     90 #include <netinet/ip6.h>
     91 #include <netinet/icmp6.h>
     92 #include <netinet6/ip6_var.h>
     93 #include <netinet6/in6_pcb.h>
     94 #include <netinet6/nd6.h>
     95 #include <netinet6/ip6protosw.h>
     96 
     97 #ifdef IPSEC
     98 #include <netinet6/ipsec.h>
     99 #include <netkey/key.h>
    100 #include <netkey/key_debug.h>
    101 #endif /* IPSEC */
    102 
    103 #include "loop.h"
    104 
    105 #include <net/net_osdep.h>
    106 
    107 #ifdef IPV6FIREWALL
    108 #include <netinet6/ip6_fw.h>
    109 #endif
    110 
    111 struct ip6_exthdrs {
    112 	struct mbuf *ip6e_ip6;
    113 	struct mbuf *ip6e_hbh;
    114 	struct mbuf *ip6e_dest1;
    115 	struct mbuf *ip6e_rthdr;
    116 	struct mbuf *ip6e_dest2;
    117 };
    118 
    119 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
    120 			    struct socket *));
    121 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *));
    122 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **));
    123 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int));
    124 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
    125 				  struct ip6_frag **));
    126 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t));
    127 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *));
    128 
    129 extern struct ifnet loif[NLOOP];
    130 
    131 /*
    132  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
    133  * header (with pri, len, nxt, hlim, src, dst).
    134  * This function may modify ver and hlim only.
    135  * The mbuf chain containing the packet will be freed.
    136  * The mbuf opt, if present, will not be freed.
    137  */
    138 int
    139 ip6_output(m0, opt, ro, flags, im6o, ifpp)
    140 	struct mbuf *m0;
    141 	struct ip6_pktopts *opt;
    142 	struct route_in6 *ro;
    143 	int flags;
    144 	struct ip6_moptions *im6o;
    145 	struct ifnet **ifpp;		/* XXX: just for statistics */
    146 {
    147 	struct ip6_hdr *ip6, *mhip6;
    148 	struct ifnet *ifp, *origifp;
    149 	struct mbuf *m = m0;
    150 	int hlen, tlen, len, off;
    151 	struct route_in6 ip6route;
    152 	struct sockaddr_in6 *dst;
    153 	int error = 0;
    154 	struct in6_ifaddr *ia;
    155 	u_long mtu;
    156 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
    157 	struct ip6_exthdrs exthdrs;
    158 	struct in6_addr finaldst;
    159 	struct route_in6 *ro_pmtu = NULL;
    160 	int hdrsplit = 0;
    161 	int needipsec = 0;
    162 #ifdef PFIL_HOOKS
    163 	struct packet_filter_hook *pfh;
    164 	struct mbuf *m1;
    165 	int rv;
    166 #endif /* PFIL_HOOKS */
    167 #ifdef IPSEC
    168 	int needipsectun = 0;
    169 	struct socket *so;
    170 	struct secpolicy *sp = NULL;
    171 
    172 	/* for AH processing. stupid to have "socket" variable in IP layer... */
    173 	so = ipsec_getsocket(m);
    174 	(void)ipsec_setsocket(m, NULL);
    175 	ip6 = mtod(m, struct ip6_hdr *);
    176 #endif /* IPSEC */
    177 
    178 #define MAKE_EXTHDR(hp, mp)						\
    179     do {								\
    180 	if (hp) {							\
    181 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
    182 		error = ip6_copyexthdr((mp), (caddr_t)(hp), 		\
    183 				       ((eh)->ip6e_len + 1) << 3);	\
    184 		if (error)						\
    185 			goto freehdrs;					\
    186 	}								\
    187     } while (0)
    188 
    189 	bzero(&exthdrs, sizeof(exthdrs));
    190 	if (opt) {
    191 		/* Hop-by-Hop options header */
    192 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
    193 		/* Destination options header(1st part) */
    194 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
    195 		/* Routing header */
    196 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
    197 		/* Destination options header(2nd part) */
    198 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
    199 	}
    200 
    201 #ifdef IPSEC
    202 	/* get a security policy for this packet */
    203 	if (so == NULL)
    204 		sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
    205 	else
    206 		sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
    207 
    208 	if (sp == NULL) {
    209 		ipsec6stat.out_inval++;
    210 		goto freehdrs;
    211 	}
    212 
    213 	error = 0;
    214 
    215 	/* check policy */
    216 	switch (sp->policy) {
    217 	case IPSEC_POLICY_DISCARD:
    218 		/*
    219 		 * This packet is just discarded.
    220 		 */
    221 		ipsec6stat.out_polvio++;
    222 		goto freehdrs;
    223 
    224 	case IPSEC_POLICY_BYPASS:
    225 	case IPSEC_POLICY_NONE:
    226 		/* no need to do IPsec. */
    227 		needipsec = 0;
    228 		break;
    229 
    230 	case IPSEC_POLICY_IPSEC:
    231 		if (sp->req == NULL) {
    232 			/* XXX should be panic ? */
    233 			printf("ip6_output: No IPsec request specified.\n");
    234 			error = EINVAL;
    235 			goto freehdrs;
    236 		}
    237 		needipsec = 1;
    238 		break;
    239 
    240 	case IPSEC_POLICY_ENTRUST:
    241 	default:
    242 		printf("ip6_output: Invalid policy found. %d\n", sp->policy);
    243 	}
    244 #endif /* IPSEC */
    245 
    246 	/*
    247 	 * Calculate the total length of the extension header chain.
    248 	 * Keep the length of the unfragmentable part for fragmentation.
    249 	 */
    250 	optlen = 0;
    251 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
    252 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
    253 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
    254 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
    255 	/* NOTE: we don't add AH/ESP length here. do that later. */
    256 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
    257 
    258 	/*
    259 	 * If we need IPsec, or there is at least one extension header,
    260 	 * separate IP6 header from the payload.
    261 	 */
    262 	if ((needipsec || optlen) && !hdrsplit) {
    263 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
    264 			m = NULL;
    265 			goto freehdrs;
    266 		}
    267 		m = exthdrs.ip6e_ip6;
    268 		hdrsplit++;
    269 	}
    270 
    271 	/* adjust pointer */
    272 	ip6 = mtod(m, struct ip6_hdr *);
    273 
    274 	/* adjust mbuf packet header length */
    275 	m->m_pkthdr.len += optlen;
    276 	plen = m->m_pkthdr.len - sizeof(*ip6);
    277 
    278 	/* If this is a jumbo payload, insert a jumbo payload option. */
    279 	if (plen > IPV6_MAXPACKET) {
    280 		if (!hdrsplit) {
    281 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
    282 				m = NULL;
    283 				goto freehdrs;
    284 			}
    285 			m = exthdrs.ip6e_ip6;
    286 			hdrsplit++;
    287 		}
    288 		/* adjust pointer */
    289 		ip6 = mtod(m, struct ip6_hdr *);
    290 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
    291 			goto freehdrs;
    292 		ip6->ip6_plen = 0;
    293 	} else
    294 		ip6->ip6_plen = htons(plen);
    295 
    296 	/*
    297 	 * Concatenate headers and fill in next header fields.
    298 	 * Here we have, on "m"
    299 	 *	IPv6 payload
    300 	 * and we insert headers accordingly.  Finally, we should be getting:
    301 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
    302 	 *
    303 	 * during the header composing process, "m" points to IPv6 header.
    304 	 * "mprev" points to an extension header prior to esp.
    305 	 */
    306 	{
    307 		u_char *nexthdrp = &ip6->ip6_nxt;
    308 		struct mbuf *mprev = m;
    309 
    310 		/*
    311 		 * we treat dest2 specially.  this makes IPsec processing
    312 		 * much easier.
    313 		 *
    314 		 * result: IPv6 dest2 payload
    315 		 * m and mprev will point to IPv6 header.
    316 		 */
    317 		if (exthdrs.ip6e_dest2) {
    318 			if (!hdrsplit)
    319 				panic("assumption failed: hdr not split");
    320 			exthdrs.ip6e_dest2->m_next = m->m_next;
    321 			m->m_next = exthdrs.ip6e_dest2;
    322 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
    323 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
    324 		}
    325 
    326 #define MAKE_CHAIN(m, mp, p, i)\
    327     do {\
    328 	if (m) {\
    329 		if (!hdrsplit) \
    330 			panic("assumption failed: hdr not split"); \
    331 		*mtod((m), u_char *) = *(p);\
    332 		*(p) = (i);\
    333 		p = mtod((m), u_char *);\
    334 		(m)->m_next = (mp)->m_next;\
    335 		(mp)->m_next = (m);\
    336 		(mp) = (m);\
    337 	}\
    338     } while (0)
    339 		/*
    340 		 * result: IPv6 hbh dest1 rthdr dest2 payload
    341 		 * m will point to IPv6 header.  mprev will point to the
    342 		 * extension header prior to dest2 (rthdr in the above case).
    343 		 */
    344 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev,
    345 			   nexthdrp, IPPROTO_HOPOPTS);
    346 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev,
    347 			   nexthdrp, IPPROTO_DSTOPTS);
    348 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev,
    349 			   nexthdrp, IPPROTO_ROUTING);
    350 
    351 #ifdef IPSEC
    352 		if (!needipsec)
    353 			goto skip_ipsec2;
    354 
    355 		/*
    356 		 * pointers after IPsec headers are not valid any more.
    357 		 * other pointers need a great care too.
    358 		 * (IPsec routines should not mangle mbufs prior to AH/ESP)
    359 		 */
    360 		exthdrs.ip6e_dest2 = NULL;
    361 
    362 	    {
    363 		struct ip6_rthdr *rh = NULL;
    364 		int segleft_org = 0;
    365 		struct ipsec_output_state state;
    366 
    367 		if (exthdrs.ip6e_rthdr) {
    368 			rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
    369 			segleft_org = rh->ip6r_segleft;
    370 			rh->ip6r_segleft = 0;
    371 		}
    372 
    373 		bzero(&state, sizeof(state));
    374 		state.m = m;
    375 		error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
    376 			&needipsectun);
    377 		m = state.m;
    378 		if (error) {
    379 			/* mbuf is already reclaimed in ipsec6_output_trans. */
    380 			m = NULL;
    381 			switch (error) {
    382 			case EHOSTUNREACH:
    383 			case ENETUNREACH:
    384 			case EMSGSIZE:
    385 			case ENOBUFS:
    386 			case ENOMEM:
    387 				break;
    388 			default:
    389 				printf("ip6_output (ipsec): error code %d\n", error);
    390 				/*fall through*/
    391 			case ENOENT:
    392 				/* don't show these error codes to the user */
    393 				error = 0;
    394 				break;
    395 			}
    396 			goto bad;
    397 		}
    398 		if (exthdrs.ip6e_rthdr) {
    399 			/* ah6_output doesn't modify mbuf chain */
    400 			rh->ip6r_segleft = segleft_org;
    401 		}
    402 	    }
    403 skip_ipsec2:;
    404 #endif
    405 	}
    406 
    407 	/*
    408 	 * If there is a routing header, replace destination address field
    409 	 * with the first hop of the routing header.
    410 	 */
    411 	if (exthdrs.ip6e_rthdr) {
    412 		struct ip6_rthdr *rh =
    413 			(struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
    414 						  struct ip6_rthdr *));
    415 		struct ip6_rthdr0 *rh0;
    416 
    417 		finaldst = ip6->ip6_dst;
    418 		switch(rh->ip6r_type) {
    419 		case IPV6_RTHDR_TYPE_0:
    420 			 rh0 = (struct ip6_rthdr0 *)rh;
    421 			 ip6->ip6_dst = rh0->ip6r0_addr[0];
    422 			 bcopy((caddr_t)&rh0->ip6r0_addr[1],
    423 				 (caddr_t)&rh0->ip6r0_addr[0],
    424 				 sizeof(struct in6_addr)*(rh0->ip6r0_segleft - 1)
    425 				 );
    426 			 rh0->ip6r0_addr[rh0->ip6r0_segleft - 1] = finaldst;
    427 			 break;
    428 		default:	/* is it possible? */
    429 			 error = EINVAL;
    430 			 goto bad;
    431 		}
    432 	}
    433 
    434 	/* Source address validation */
    435 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
    436 	    (flags & IPV6_DADOUTPUT) == 0) {
    437 		error = EOPNOTSUPP;
    438 		ip6stat.ip6s_badscope++;
    439 		goto bad;
    440 	}
    441 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
    442 		error = EOPNOTSUPP;
    443 		ip6stat.ip6s_badscope++;
    444 		goto bad;
    445 	}
    446 
    447 	ip6stat.ip6s_localout++;
    448 
    449 	/*
    450 	 * Route packet.
    451 	 */
    452 	if (ro == 0) {
    453 		ro = &ip6route;
    454 		bzero((caddr_t)ro, sizeof(*ro));
    455 	}
    456 	ro_pmtu = ro;
    457 	if (opt && opt->ip6po_rthdr)
    458 		ro = &opt->ip6po_route;
    459 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
    460 	/*
    461 	 * If there is a cached route,
    462 	 * check that it is to the same destination
    463 	 * and is still up. If not, free it and try again.
    464 	 */
    465 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
    466 			 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
    467 		RTFREE(ro->ro_rt);
    468 		ro->ro_rt = (struct rtentry *)0;
    469 	}
    470 	if (ro->ro_rt == 0) {
    471 		bzero(dst, sizeof(*dst));
    472 		dst->sin6_family = AF_INET6;
    473 		dst->sin6_len = sizeof(struct sockaddr_in6);
    474 		dst->sin6_addr = ip6->ip6_dst;
    475 	}
    476 #ifdef IPSEC
    477 	if (needipsec && needipsectun) {
    478 		struct ipsec_output_state state;
    479 
    480 		/*
    481 		 * All the extension headers will become inaccessible
    482 		 * (since they can be encrypted).
    483 		 * Don't panic, we need no more updates to extension headers
    484 		 * on inner IPv6 packet (since they are now encapsulated).
    485 		 *
    486 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
    487 		 */
    488 		bzero(&exthdrs, sizeof(exthdrs));
    489 		exthdrs.ip6e_ip6 = m;
    490 
    491 		bzero(&state, sizeof(state));
    492 		state.m = m;
    493 		state.ro = (struct route *)ro;
    494 		state.dst = (struct sockaddr *)dst;
    495 
    496 		error = ipsec6_output_tunnel(&state, sp, flags);
    497 
    498 		m = state.m;
    499 		ro = (struct route_in6 *)state.ro;
    500 		dst = (struct sockaddr_in6 *)state.dst;
    501 		if (error) {
    502 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
    503 			m0 = m = NULL;
    504 			m = NULL;
    505 			switch (error) {
    506 			case EHOSTUNREACH:
    507 			case ENETUNREACH:
    508 			case EMSGSIZE:
    509 			case ENOBUFS:
    510 			case ENOMEM:
    511 				break;
    512 			default:
    513 				printf("ip6_output (ipsec): error code %d\n", error);
    514 				/*fall through*/
    515 			case ENOENT:
    516 				/* don't show these error codes to the user */
    517 				error = 0;
    518 				break;
    519 			}
    520 			goto bad;
    521 		}
    522 
    523 		exthdrs.ip6e_ip6 = m;
    524 	}
    525 #endif /*IPSEC*/
    526 
    527 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
    528 		/* Unicast */
    529 
    530 #define ifatoia6(ifa)	((struct in6_ifaddr *)(ifa))
    531 #define sin6tosa(sin6)	((struct sockaddr *)(sin6))
    532 		/* xxx
    533 		 * interface selection comes here
    534 		 * if an interface is specified from an upper layer,
    535 		 * ifp must point it.
    536 		 */
    537 		if (ro->ro_rt == 0) {
    538 			/*
    539 			 * non-bsdi always clone routes, if parent is
    540 			 * PRF_CLONING.
    541 			 */
    542 			rtalloc((struct route *)ro);
    543 		}
    544 		if (ro->ro_rt == 0) {
    545 			ip6stat.ip6s_noroute++;
    546 			error = EHOSTUNREACH;
    547 			/* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
    548 			goto bad;
    549 		}
    550 		ia = ifatoia6(ro->ro_rt->rt_ifa);
    551 		ifp = ro->ro_rt->rt_ifp;
    552 		ro->ro_rt->rt_use++;
    553 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
    554 			dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
    555 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
    556 
    557 		in6_ifstat_inc(ifp, ifs6_out_request);
    558 
    559 		/*
    560 		 * Check if the outgoing interface conflicts with
    561 		 * the interface specified by ifi6_ifindex (if specified).
    562 		 * Note that loopback interface is always okay.
    563 		 * (this may happen when we are sending a packet to one of
    564 		 *  our own addresses.)
    565 		 */
    566 		if (opt && opt->ip6po_pktinfo
    567 		 && opt->ip6po_pktinfo->ipi6_ifindex) {
    568 			if (!(ifp->if_flags & IFF_LOOPBACK)
    569 			 && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
    570 				ip6stat.ip6s_noroute++;
    571 				in6_ifstat_inc(ifp, ifs6_out_discard);
    572 				error = EHOSTUNREACH;
    573 				goto bad;
    574 			}
    575 		}
    576 
    577 		if (opt && opt->ip6po_hlim != -1)
    578 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
    579 	} else {
    580 		/* Multicast */
    581 		struct	in6_multi *in6m;
    582 
    583 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
    584 
    585 		/*
    586 		 * See if the caller provided any multicast options
    587 		 */
    588 		ifp = NULL;
    589 		if (im6o != NULL) {
    590 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
    591 			if (im6o->im6o_multicast_ifp != NULL)
    592 				ifp = im6o->im6o_multicast_ifp;
    593 		} else
    594 			ip6->ip6_hlim = ip6_defmcasthlim;
    595 
    596 		/*
    597 		 * See if the caller provided the outgoing interface
    598 		 * as an ancillary data.
    599 		 * Boundary check for ifindex is assumed to be already done.
    600 		 */
    601 		if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
    602 			ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex];
    603 
    604 		/*
    605 		 * If the destination is a node-local scope multicast,
    606 		 * the packet should be loop-backed only.
    607 		 */
    608 		if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
    609 			/*
    610 			 * If the outgoing interface is already specified,
    611 			 * it should be a loopback interface.
    612 			 */
    613 			if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) {
    614 				ip6stat.ip6s_badscope++;
    615 				error = ENETUNREACH; /* XXX: better error? */
    616 				/* XXX correct ifp? */
    617 				in6_ifstat_inc(ifp, ifs6_out_discard);
    618 				goto bad;
    619 			} else {
    620 				ifp = &loif[0];
    621 			}
    622 		}
    623 
    624 		if (opt && opt->ip6po_hlim != -1)
    625 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
    626 
    627 		/*
    628 		 * If caller did not provide an interface lookup a
    629 		 * default in the routing table.  This is either a
    630 		 * default for the speicfied group (i.e. a host
    631 		 * route), or a multicast default (a route for the
    632 		 * ``net'' ff00::/8).
    633 		 */
    634 		if (ifp == NULL) {
    635 			if (ro->ro_rt == 0) {
    636 				ro->ro_rt = rtalloc1((struct sockaddr *)
    637 						&ro->ro_dst, 0
    638 						);
    639 			}
    640 			if (ro->ro_rt == 0) {
    641 				ip6stat.ip6s_noroute++;
    642 				error = EHOSTUNREACH;
    643 				/* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
    644 				goto bad;
    645 			}
    646 			ia = ifatoia6(ro->ro_rt->rt_ifa);
    647 			ifp = ro->ro_rt->rt_ifp;
    648 			ro->ro_rt->rt_use++;
    649 		}
    650 
    651 		if ((flags & IPV6_FORWARDING) == 0)
    652 			in6_ifstat_inc(ifp, ifs6_out_request);
    653 		in6_ifstat_inc(ifp, ifs6_out_mcast);
    654 
    655 		/*
    656 		 * Confirm that the outgoing interface supports multicast.
    657 		 */
    658 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
    659 			ip6stat.ip6s_noroute++;
    660 			in6_ifstat_inc(ifp, ifs6_out_discard);
    661 			error = ENETUNREACH;
    662 			goto bad;
    663 		}
    664 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
    665 		if (in6m != NULL &&
    666 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
    667 			/*
    668 			 * If we belong to the destination multicast group
    669 			 * on the outgoing interface, and the caller did not
    670 			 * forbid loopback, loop back a copy.
    671 			 */
    672 			ip6_mloopback(ifp, m, dst);
    673 		} else {
    674 			/*
    675 			 * If we are acting as a multicast router, perform
    676 			 * multicast forwarding as if the packet had just
    677 			 * arrived on the interface to which we are about
    678 			 * to send.  The multicast forwarding function
    679 			 * recursively calls this function, using the
    680 			 * IPV6_FORWARDING flag to prevent infinite recursion.
    681 			 *
    682 			 * Multicasts that are looped back by ip6_mloopback(),
    683 			 * above, will be forwarded by the ip6_input() routine,
    684 			 * if necessary.
    685 			 */
    686 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
    687 				if (ip6_mforward(ip6, ifp, m) != 0) {
    688 					m_freem(m);
    689 					goto done;
    690 				}
    691 			}
    692 		}
    693 		/*
    694 		 * Multicasts with a hoplimit of zero may be looped back,
    695 		 * above, but must not be transmitted on a network.
    696 		 * Also, multicasts addressed to the loopback interface
    697 		 * are not sent -- the above call to ip6_mloopback() will
    698 		 * loop back a copy if this host actually belongs to the
    699 		 * destination group on the loopback interface.
    700 		 */
    701 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
    702 			m_freem(m);
    703 			goto done;
    704 		}
    705 	}
    706 
    707 	/*
    708 	 * Fill the outgoing inteface to tell the upper layer
    709 	 * to increment per-interface statistics.
    710 	 */
    711 	if (ifpp)
    712 		*ifpp = ifp;
    713 
    714 	/*
    715 	 * Determine path MTU.
    716 	 */
    717 	if (ro_pmtu != ro) {
    718 		/* The first hop and the final destination may differ. */
    719 		struct sockaddr_in6 *sin6_fin =
    720 			(struct sockaddr_in6 *)&ro_pmtu->ro_dst;
    721 		if (ro_pmtu->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
    722 				       !IN6_ARE_ADDR_EQUAL(&sin6_fin->sin6_addr,
    723 							   &finaldst))) {
    724 			RTFREE(ro_pmtu->ro_rt);
    725 			ro_pmtu->ro_rt = (struct rtentry *)0;
    726 		}
    727 		if (ro_pmtu->ro_rt == 0) {
    728 			bzero(sin6_fin, sizeof(*sin6_fin));
    729 			sin6_fin->sin6_family = AF_INET6;
    730 			sin6_fin->sin6_len = sizeof(struct sockaddr_in6);
    731 			sin6_fin->sin6_addr = finaldst;
    732 
    733 			rtalloc((struct route *)ro_pmtu);
    734 		}
    735 	}
    736 	if (ro_pmtu->ro_rt != NULL) {
    737 		u_int32_t ifmtu = nd_ifinfo[ifp->if_index].linkmtu;
    738 
    739 		mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
    740 		if (mtu > ifmtu) {
    741 			/*
    742 			 * The MTU on the route is larger than the MTU on
    743 			 * the interface!  This shouldn't happen, unless the
    744 			 * MTU of the interface has been changed after the
    745 			 * interface was brought up.  Change the MTU in the
    746 			 * route to match the interface MTU (as long as the
    747 			 * field isn't locked).
    748 			 */
    749 			 mtu = ifmtu;
    750 			 if ((ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
    751 				 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; /* XXX */
    752 		}
    753 	} else {
    754 		mtu = nd_ifinfo[ifp->if_index].linkmtu;
    755 	}
    756 
    757 	/* Fake scoped addresses */
    758 	if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
    759 		/*
    760 		 * If source or destination address is a scoped address, and
    761 		 * the packet is going to be sent to a loopback interface,
    762 		 * we should keep the original interface.
    763 		 */
    764 
    765 		/*
    766 		 * XXX: this is a very experimental and temporary solution.
    767 		 * We eventually have sockaddr_in6 and use the sin6_scope_id
    768 		 * field of the structure here.
    769 		 * We rely on the consistency between two scope zone ids
    770 		 * of source add destination, which should already be assured
    771 		 * Larger scopes than link will be supported in the near
    772 		 * future.
    773 		 */
    774 		origifp = NULL;
    775 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
    776 			origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
    777 		else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
    778 			origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
    779 		/*
    780 		 * XXX: origifp can be NULL even in those two cases above.
    781 		 * For example, if we remove the (only) link-local address
    782 		 * from the loopback interface, and try to send a link-local
    783 		 * address without link-id information.  Then the source
    784 		 * address is ::1, and the destination address is the
    785 		 * link-local address with its s6_addr16[1] being zero.
    786 		 * What is worse, if the packet goes to the loopback interface
    787 		 * by a default rejected route, the null pointer would be
    788 		 * passed to looutput, and the kernel would hang.
    789 		 * The following last resort would prevent such disaster.
    790 		 */
    791 		if (origifp == NULL)
    792 			origifp = ifp;
    793 	}
    794 	else
    795 		origifp = ifp;
    796 #ifndef FAKE_LOOPBACK_IF
    797 	if ((ifp->if_flags & IFF_LOOPBACK) == 0)
    798 #else
    799 	if (1)
    800 #endif
    801 	{
    802 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
    803 			ip6->ip6_src.s6_addr16[1] = 0;
    804 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
    805 			ip6->ip6_dst.s6_addr16[1] = 0;
    806 	}
    807 
    808 	/*
    809 	 * If the outgoing packet contains a hop-by-hop options header,
    810 	 * it must be examined and processed even by the source node.
    811 	 * (RFC 2460, section 4.)
    812 	 */
    813 	if (exthdrs.ip6e_hbh) {
    814 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh,
    815 					   struct ip6_hbh *);
    816 		u_int32_t dummy1; /* XXX unused */
    817 		u_int32_t dummy2; /* XXX unused */
    818 
    819 		/*
    820 		 *  XXX: if we have to send an ICMPv6 error to the sender,
    821 		 *       we need the M_LOOP flag since icmp6_error() expects
    822 		 *       the IPv6 and the hop-by-hop options header are
    823 		 *       continuous unless the flag is set.
    824 		 */
    825 		m->m_flags |= M_LOOP;
    826 		m->m_pkthdr.rcvif = ifp;
    827 		if (ip6_process_hopopts(m,
    828 					(u_int8_t *)(hbh + 1),
    829 					((hbh->ip6h_len + 1) << 3) -
    830 					sizeof(struct ip6_hbh),
    831 					&dummy1, &dummy2) < 0) {
    832 			/* m was already freed at this point */
    833 			error = EINVAL;/* better error? */
    834 			goto done;
    835 		}
    836 		m->m_flags &= ~M_LOOP; /* XXX */
    837 		m->m_pkthdr.rcvif = NULL;
    838 	}
    839 
    840 #ifdef PFIL_HOOKS
    841 	/*
    842 	 * Run through list of hooks for output packets.
    843 	 */
    844 	m1 = m;
    845 	pfh = pfil_hook_get(PFIL_OUT,
    846 		&inet6sw[ip6_protox[IPPROTO_IPV6]].pr_pfh);
    847 	for (; pfh; pfh = pfh->pfil_link.tqe_next)
    848 		if (pfh->pfil_func) {
    849 		    	rv = pfh->pfil_func(ip6, sizeof(*ip6), ifp, 1, &m1);
    850 			if (rv) {
    851 				error = EHOSTUNREACH;
    852 				goto done;
    853 			}
    854 			m = m1;
    855 			if (m == NULL)
    856 				goto done;
    857 			ip6 = mtod(m, struct ip6_hdr *);
    858 		}
    859 #endif /* PFIL_HOOKS */
    860 	/*
    861 	 * Send the packet to the outgoing interface.
    862 	 * If necessary, do IPv6 fragmentation before sending.
    863 	 */
    864 	tlen = m->m_pkthdr.len;
    865 	if (tlen <= mtu
    866 #ifdef notyet
    867 	    /*
    868 	     * On any link that cannot convey a 1280-octet packet in one piece,
    869 	     * link-specific fragmentation and reassembly must be provided at
    870 	     * a layer below IPv6. [RFC 2460, sec.5]
    871 	     * Thus if the interface has ability of link-level fragmentation,
    872 	     * we can just send the packet even if the packet size is
    873 	     * larger than the link's MTU.
    874 	     * XXX: IFF_FRAGMENTABLE (or such) flag has not been defined yet...
    875 	     */
    876 
    877 	    || ifp->if_flags & IFF_FRAGMENTABLE
    878 #endif
    879 	    )
    880 	{
    881 #ifdef IFA_STATS
    882 		if (IFA_STATS) {
    883 			struct in6_ifaddr *ia6;
    884 			ip6 = mtod(m, struct ip6_hdr *);
    885 			ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
    886 			if (ia6) {
    887 				ia->ia_ifa.ifa_data.ifad_outbytes +=
    888 					m->m_pkthdr.len;
    889 			}
    890 		}
    891 #endif
    892 #ifdef IPSEC
    893 		/* clean ipsec history once it goes out of the node */
    894 		ipsec_delaux(m);
    895 #endif
    896 #ifdef OLDIP6OUTPUT
    897 		error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst,
    898 					  ro->ro_rt);
    899 #else
    900 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
    901 #endif
    902 		goto done;
    903 	} else if (mtu < IPV6_MMTU) {
    904 		/*
    905 		 * note that path MTU is never less than IPV6_MMTU
    906 		 * (see icmp6_input).
    907 		 */
    908 		error = EMSGSIZE;
    909 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
    910 		goto bad;
    911 	} else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
    912 		error = EMSGSIZE;
    913 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
    914 		goto bad;
    915 	} else {
    916 		struct mbuf **mnext, *m_frgpart;
    917 		struct ip6_frag *ip6f;
    918 		u_int32_t id = htonl(ip6_id++);
    919 		u_char nextproto;
    920 
    921 		/*
    922 		 * Too large for the destination or interface;
    923 		 * fragment if possible.
    924 		 * Must be able to put at least 8 bytes per fragment.
    925 		 */
    926 		hlen = unfragpartlen;
    927 		if (mtu > IPV6_MAXPACKET)
    928 			mtu = IPV6_MAXPACKET;
    929 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
    930 		if (len < 8) {
    931 			error = EMSGSIZE;
    932 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
    933 			goto bad;
    934 		}
    935 
    936 		mnext = &m->m_nextpkt;
    937 
    938 		/*
    939 		 * Change the next header field of the last header in the
    940 		 * unfragmentable part.
    941 		 */
    942 		if (exthdrs.ip6e_rthdr) {
    943 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
    944 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
    945 		} else if (exthdrs.ip6e_dest1) {
    946 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
    947 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
    948 		} else if (exthdrs.ip6e_hbh) {
    949 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
    950 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
    951 		} else {
    952 			nextproto = ip6->ip6_nxt;
    953 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
    954 		}
    955 
    956 		/*
    957 		 * Loop through length of segment after first fragment,
    958 		 * make new header and copy data of each part and link onto chain.
    959 		 */
    960 		m0 = m;
    961 		for (off = hlen; off < tlen; off += len) {
    962 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
    963 			if (!m) {
    964 				error = ENOBUFS;
    965 				ip6stat.ip6s_odropped++;
    966 				goto sendorfree;
    967 			}
    968 			m->m_flags = m0->m_flags & M_COPYFLAGS;
    969 			*mnext = m;
    970 			mnext = &m->m_nextpkt;
    971 			m->m_data += max_linkhdr;
    972 			mhip6 = mtod(m, struct ip6_hdr *);
    973 			*mhip6 = *ip6;
    974 			m->m_len = sizeof(*mhip6);
    975  			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
    976  			if (error) {
    977 				ip6stat.ip6s_odropped++;
    978 				goto sendorfree;
    979 			}
    980 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
    981 			if (off + len >= tlen)
    982 				len = tlen - off;
    983 			else
    984 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
    985 			mhip6->ip6_plen = htons((u_short)(len + hlen +
    986 							  sizeof(*ip6f) -
    987 							  sizeof(struct ip6_hdr)));
    988 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
    989 				error = ENOBUFS;
    990 				ip6stat.ip6s_odropped++;
    991 				goto sendorfree;
    992 			}
    993 			m_cat(m, m_frgpart);
    994 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
    995 			m->m_pkthdr.rcvif = (struct ifnet *)0;
    996 			ip6f->ip6f_reserved = 0;
    997 			ip6f->ip6f_ident = id;
    998 			ip6f->ip6f_nxt = nextproto;
    999 			ip6stat.ip6s_ofragments++;
   1000 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
   1001 		}
   1002 
   1003 		in6_ifstat_inc(ifp, ifs6_out_fragok);
   1004 	}
   1005 
   1006 	/*
   1007 	 * Remove leading garbages.
   1008 	 */
   1009 sendorfree:
   1010 	m = m0->m_nextpkt;
   1011 	m0->m_nextpkt = 0;
   1012 	m_freem(m0);
   1013 	for (m0 = m; m; m = m0) {
   1014 		m0 = m->m_nextpkt;
   1015 		m->m_nextpkt = 0;
   1016 		if (error == 0) {
   1017 #ifdef IFA_STATS
   1018 			if (IFA_STATS) {
   1019 				struct in6_ifaddr *ia6;
   1020 				ip6 = mtod(m, struct ip6_hdr *);
   1021 				ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
   1022 				if (ia6) {
   1023 					ia->ia_ifa.ifa_data.ifad_outbytes +=
   1024 						m->m_pkthdr.len;
   1025 				}
   1026 			}
   1027 #endif
   1028 #ifdef IPSEC
   1029 			/* clean ipsec history once it goes out of the node */
   1030 			ipsec_delaux(m);
   1031 #endif
   1032 #ifdef OLDIP6OUTPUT
   1033 			error = (*ifp->if_output)(ifp, m,
   1034 						  (struct sockaddr *)dst,
   1035 						  ro->ro_rt);
   1036 #else
   1037 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
   1038 #endif
   1039 		} else
   1040 			m_freem(m);
   1041 	}
   1042 
   1043 	if (error == 0)
   1044 		ip6stat.ip6s_fragmented++;
   1045 
   1046 done:
   1047 	if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
   1048 		RTFREE(ro->ro_rt);
   1049 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
   1050 		RTFREE(ro_pmtu->ro_rt);
   1051 	}
   1052 
   1053 #ifdef IPSEC
   1054 	if (sp != NULL)
   1055 		key_freesp(sp);
   1056 #endif /* IPSEC */
   1057 
   1058 	return(error);
   1059 
   1060 freehdrs:
   1061 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
   1062 	m_freem(exthdrs.ip6e_dest1);
   1063 	m_freem(exthdrs.ip6e_rthdr);
   1064 	m_freem(exthdrs.ip6e_dest2);
   1065 	/* fall through */
   1066 bad:
   1067 	m_freem(m);
   1068 	goto done;
   1069 }
   1070 
   1071 static int
   1072 ip6_copyexthdr(mp, hdr, hlen)
   1073 	struct mbuf **mp;
   1074 	caddr_t hdr;
   1075 	int hlen;
   1076 {
   1077 	struct mbuf *m;
   1078 
   1079 	if (hlen > MCLBYTES)
   1080 		return(ENOBUFS); /* XXX */
   1081 
   1082 	MGET(m, M_DONTWAIT, MT_DATA);
   1083 	if (!m)
   1084 		return(ENOBUFS);
   1085 
   1086 	if (hlen > MLEN) {
   1087 		MCLGET(m, M_DONTWAIT);
   1088 		if ((m->m_flags & M_EXT) == 0) {
   1089 			m_free(m);
   1090 			return(ENOBUFS);
   1091 		}
   1092 	}
   1093 	m->m_len = hlen;
   1094 	if (hdr)
   1095 		bcopy(hdr, mtod(m, caddr_t), hlen);
   1096 
   1097 	*mp = m;
   1098 	return(0);
   1099 }
   1100 
   1101 /*
   1102  * Insert jumbo payload option.
   1103  */
   1104 static int
   1105 ip6_insert_jumboopt(exthdrs, plen)
   1106 	struct ip6_exthdrs *exthdrs;
   1107 	u_int32_t plen;
   1108 {
   1109 	struct mbuf *mopt;
   1110 	u_char *optbuf;
   1111 
   1112 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
   1113 
   1114 	/*
   1115 	 * If there is no hop-by-hop options header, allocate new one.
   1116 	 * If there is one but it doesn't have enough space to store the
   1117 	 * jumbo payload option, allocate a cluster to store the whole options.
   1118 	 * Otherwise, use it to store the options.
   1119 	 */
   1120 	if (exthdrs->ip6e_hbh == 0) {
   1121 		MGET(mopt, M_DONTWAIT, MT_DATA);
   1122 		if (mopt == 0)
   1123 			return(ENOBUFS);
   1124 		mopt->m_len = JUMBOOPTLEN;
   1125 		optbuf = mtod(mopt, u_char *);
   1126 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
   1127 		exthdrs->ip6e_hbh = mopt;
   1128 	} else {
   1129 		struct ip6_hbh *hbh;
   1130 
   1131 		mopt = exthdrs->ip6e_hbh;
   1132 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
   1133 			caddr_t oldoptp = mtod(mopt, caddr_t);
   1134 			int oldoptlen = mopt->m_len;
   1135 
   1136 			if (mopt->m_flags & M_EXT)
   1137 				return(ENOBUFS); /* XXX */
   1138 			MCLGET(mopt, M_DONTWAIT);
   1139 			if ((mopt->m_flags & M_EXT) == 0)
   1140 				return(ENOBUFS);
   1141 
   1142 			bcopy(oldoptp, mtod(mopt, caddr_t), oldoptlen);
   1143 			optbuf = mtod(mopt, caddr_t) + oldoptlen;
   1144 			mopt->m_len = oldoptlen + JUMBOOPTLEN;
   1145 		} else {
   1146 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
   1147 			mopt->m_len += JUMBOOPTLEN;
   1148 		}
   1149 		optbuf[0] = IP6OPT_PADN;
   1150 		optbuf[1] = 0;
   1151 
   1152 		/*
   1153 		 * Adjust the header length according to the pad and
   1154 		 * the jumbo payload option.
   1155 		 */
   1156 		hbh = mtod(mopt, struct ip6_hbh *);
   1157 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
   1158 	}
   1159 
   1160 	/* fill in the option. */
   1161 	optbuf[2] = IP6OPT_JUMBO;
   1162 	optbuf[3] = 4;
   1163 	*(u_int32_t *)&optbuf[4] = htonl(plen + JUMBOOPTLEN);
   1164 
   1165 	/* finally, adjust the packet header length */
   1166 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
   1167 
   1168 	return(0);
   1169 #undef JUMBOOPTLEN
   1170 }
   1171 
   1172 /*
   1173  * Insert fragment header and copy unfragmentable header portions.
   1174  */
   1175 static int
   1176 ip6_insertfraghdr(m0, m, hlen, frghdrp)
   1177 	struct mbuf *m0, *m;
   1178 	int hlen;
   1179 	struct ip6_frag **frghdrp;
   1180 {
   1181 	struct mbuf *n, *mlast;
   1182 
   1183 	if (hlen > sizeof(struct ip6_hdr)) {
   1184 		n = m_copym(m0, sizeof(struct ip6_hdr),
   1185 			    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
   1186 		if (n == 0)
   1187 			return(ENOBUFS);
   1188 		m->m_next = n;
   1189 	} else
   1190 		n = m;
   1191 
   1192 	/* Search for the last mbuf of unfragmentable part. */
   1193 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
   1194 		;
   1195 
   1196 	if ((mlast->m_flags & M_EXT) == 0 &&
   1197 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
   1198 		/* use the trailing space of the last mbuf for the fragment hdr */
   1199 		*frghdrp =
   1200 			(struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len);
   1201 		mlast->m_len += sizeof(struct ip6_frag);
   1202 		m->m_pkthdr.len += sizeof(struct ip6_frag);
   1203 	} else {
   1204 		/* allocate a new mbuf for the fragment header */
   1205 		struct mbuf *mfrg;
   1206 
   1207 		MGET(mfrg, M_DONTWAIT, MT_DATA);
   1208 		if (mfrg == 0)
   1209 			return(ENOBUFS);
   1210 		mfrg->m_len = sizeof(struct ip6_frag);
   1211 		*frghdrp = mtod(mfrg, struct ip6_frag *);
   1212 		mlast->m_next = mfrg;
   1213 	}
   1214 
   1215 	return(0);
   1216 }
   1217 
   1218 /*
   1219  * IP6 socket option processing.
   1220  */
   1221 int
   1222 ip6_ctloutput(op, so, level, optname, mp)
   1223 	int op;
   1224 	struct socket *so;
   1225 	int level, optname;
   1226 	struct mbuf **mp;
   1227 {
   1228 	register struct in6pcb *in6p = sotoin6pcb(so);
   1229 	register struct mbuf *m = *mp;
   1230 	register int optval = 0;
   1231 	int error = 0;
   1232 	struct proc *p = curproc;	/* XXX */
   1233 
   1234 	if (level == IPPROTO_IPV6) {
   1235 		switch (op) {
   1236 
   1237 		case PRCO_SETOPT:
   1238 			switch (optname) {
   1239 			case IPV6_PKTOPTIONS:
   1240 				/* m is freed in ip6_pcbopts */
   1241 				return(ip6_pcbopts(&in6p->in6p_outputopts,
   1242 						   m, so));
   1243 			case IPV6_HOPOPTS:
   1244 			case IPV6_DSTOPTS:
   1245 				if (p == 0 || suser(p->p_ucred, &p->p_acflag)) {
   1246 					error = EPERM;
   1247 					break;
   1248 				}
   1249 				/* fall through */
   1250 			case IPV6_UNICAST_HOPS:
   1251 			case IPV6_RECVOPTS:
   1252 			case IPV6_RECVRETOPTS:
   1253 			case IPV6_RECVDSTADDR:
   1254 			case IPV6_PKTINFO:
   1255 			case IPV6_HOPLIMIT:
   1256 			case IPV6_RTHDR:
   1257 			case IPV6_CHECKSUM:
   1258 			case IPV6_FAITH:
   1259 #ifndef INET6_BINDV6ONLY
   1260 			case IPV6_BINDV6ONLY:
   1261 #endif
   1262 				if (!m || m->m_len != sizeof(int))
   1263 					error = EINVAL;
   1264 				else {
   1265 					optval = *mtod(m, int *);
   1266 					switch (optname) {
   1267 
   1268 					case IPV6_UNICAST_HOPS:
   1269 						if (optval < -1 || optval >= 256)
   1270 							error = EINVAL;
   1271 						else {
   1272 							/* -1 = kernel default */
   1273 							in6p->in6p_hops = optval;
   1274 						}
   1275 						break;
   1276 #define OPTSET(bit) \
   1277 	if (optval) \
   1278 		in6p->in6p_flags |= bit; \
   1279 	else \
   1280 		in6p->in6p_flags &= ~bit;
   1281 
   1282 					case IPV6_RECVOPTS:
   1283 						OPTSET(IN6P_RECVOPTS);
   1284 						break;
   1285 
   1286 					case IPV6_RECVRETOPTS:
   1287 						OPTSET(IN6P_RECVRETOPTS);
   1288 						break;
   1289 
   1290 					case IPV6_RECVDSTADDR:
   1291 						OPTSET(IN6P_RECVDSTADDR);
   1292 						break;
   1293 
   1294 					case IPV6_PKTINFO:
   1295 						OPTSET(IN6P_PKTINFO);
   1296 						break;
   1297 
   1298 					case IPV6_HOPLIMIT:
   1299 						OPTSET(IN6P_HOPLIMIT);
   1300 						break;
   1301 
   1302 					case IPV6_HOPOPTS:
   1303 						OPTSET(IN6P_HOPOPTS);
   1304 						break;
   1305 
   1306 					case IPV6_DSTOPTS:
   1307 						OPTSET(IN6P_DSTOPTS);
   1308 						break;
   1309 
   1310 					case IPV6_RTHDR:
   1311 						OPTSET(IN6P_RTHDR);
   1312 						break;
   1313 
   1314 					case IPV6_CHECKSUM:
   1315 						in6p->in6p_cksum = optval;
   1316 						break;
   1317 
   1318 					case IPV6_FAITH:
   1319 						OPTSET(IN6P_FAITH);
   1320 						break;
   1321 
   1322 #ifndef INET6_BINDV6ONLY
   1323 					case IPV6_BINDV6ONLY:
   1324 						OPTSET(IN6P_BINDV6ONLY);
   1325 						break;
   1326 #endif
   1327 					}
   1328 				}
   1329 				break;
   1330 #undef OPTSET
   1331 
   1332 			case IPV6_MULTICAST_IF:
   1333 			case IPV6_MULTICAST_HOPS:
   1334 			case IPV6_MULTICAST_LOOP:
   1335 			case IPV6_JOIN_GROUP:
   1336 			case IPV6_LEAVE_GROUP:
   1337 				error =	ip6_setmoptions(optname, &in6p->in6p_moptions, m);
   1338 				break;
   1339 
   1340 			case IPV6_PORTRANGE:
   1341 				optval = *mtod(m, int *);
   1342 
   1343 				switch (optval) {
   1344 				case IPV6_PORTRANGE_DEFAULT:
   1345 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
   1346 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
   1347 					break;
   1348 
   1349 				case IPV6_PORTRANGE_HIGH:
   1350 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
   1351 					in6p->in6p_flags |= IN6P_HIGHPORT;
   1352 					break;
   1353 
   1354 				case IPV6_PORTRANGE_LOW:
   1355 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
   1356 					in6p->in6p_flags |= IN6P_LOWPORT;
   1357 					break;
   1358 
   1359 				default:
   1360 					error = EINVAL;
   1361 					break;
   1362 				}
   1363 				break;
   1364 
   1365 #ifdef IPSEC
   1366 			case IPV6_IPSEC_POLICY:
   1367 			    {
   1368 				caddr_t req = NULL;
   1369 				size_t len = 0;
   1370 
   1371 				int priv = 0;
   1372 				if (p == 0 || suser(p->p_ucred, &p->p_acflag))
   1373 					priv = 0;
   1374 				else
   1375 					priv = 1;
   1376 				if (m) {
   1377 					req = mtod(m, caddr_t);
   1378 					len = m->m_len;
   1379 				}
   1380 				error = ipsec6_set_policy(in6p,
   1381 				                   optname, req, len, priv);
   1382 			    }
   1383 				break;
   1384 #endif /* IPSEC */
   1385 
   1386 			default:
   1387 				error = ENOPROTOOPT;
   1388 				break;
   1389 			}
   1390 			if (m)
   1391 				(void)m_free(m);
   1392 			break;
   1393 
   1394 		case PRCO_GETOPT:
   1395 			switch (optname) {
   1396 
   1397 			case IPV6_OPTIONS:
   1398 			case IPV6_RETOPTS:
   1399 #if 0
   1400 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
   1401 				if (in6p->in6p_options) {
   1402 					m->m_len = in6p->in6p_options->m_len;
   1403 					bcopy(mtod(in6p->in6p_options, caddr_t),
   1404 					      mtod(m, caddr_t),
   1405 					      (unsigned)m->m_len);
   1406 				} else
   1407 					m->m_len = 0;
   1408 				break;
   1409 #else
   1410 				error = ENOPROTOOPT;
   1411 				break;
   1412 #endif
   1413 
   1414 			case IPV6_PKTOPTIONS:
   1415 				if (in6p->in6p_options) {
   1416 					*mp = m_copym(in6p->in6p_options, 0,
   1417 						      M_COPYALL, M_WAIT);
   1418 				} else {
   1419 					*mp = m_get(M_WAIT, MT_SOOPTS);
   1420 					(*mp)->m_len = 0;
   1421 				}
   1422 				break;
   1423 
   1424 			case IPV6_HOPOPTS:
   1425 			case IPV6_DSTOPTS:
   1426 				if (p == 0 || suser(p->p_ucred, &p->p_acflag)) {
   1427 					error = EPERM;
   1428 					break;
   1429 				}
   1430 				/* fall through */
   1431 			case IPV6_UNICAST_HOPS:
   1432 			case IPV6_RECVOPTS:
   1433 			case IPV6_RECVRETOPTS:
   1434 			case IPV6_RECVDSTADDR:
   1435 			case IPV6_PORTRANGE:
   1436 			case IPV6_PKTINFO:
   1437 			case IPV6_HOPLIMIT:
   1438 			case IPV6_RTHDR:
   1439 			case IPV6_CHECKSUM:
   1440 			case IPV6_FAITH:
   1441 #ifndef INET6_BINDV6ONLY
   1442 			case IPV6_BINDV6ONLY:
   1443 #endif
   1444 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
   1445 				m->m_len = sizeof(int);
   1446 				switch (optname) {
   1447 
   1448 				case IPV6_UNICAST_HOPS:
   1449 					optval = in6p->in6p_hops;
   1450 					break;
   1451 
   1452 #define OPTBIT(bit) (in6p->in6p_flags & bit ? 1 : 0)
   1453 
   1454 				case IPV6_RECVOPTS:
   1455 					optval = OPTBIT(IN6P_RECVOPTS);
   1456 					break;
   1457 
   1458 				case IPV6_RECVRETOPTS:
   1459 					optval = OPTBIT(IN6P_RECVRETOPTS);
   1460 					break;
   1461 
   1462 				case IPV6_RECVDSTADDR:
   1463 					optval = OPTBIT(IN6P_RECVDSTADDR);
   1464 					break;
   1465 
   1466 				case IPV6_PORTRANGE:
   1467 				    {
   1468 					int flags;
   1469 					flags = in6p->in6p_flags;
   1470 					if (flags & IN6P_HIGHPORT)
   1471 						optval = IPV6_PORTRANGE_HIGH;
   1472 					else if (flags & IN6P_LOWPORT)
   1473 						optval = IPV6_PORTRANGE_LOW;
   1474 					else
   1475 						optval = 0;
   1476 					break;
   1477 				    }
   1478 
   1479 				case IPV6_PKTINFO:
   1480 					optval = OPTBIT(IN6P_PKTINFO);
   1481 					break;
   1482 
   1483 				case IPV6_HOPLIMIT:
   1484 					optval = OPTBIT(IN6P_HOPLIMIT);
   1485 					break;
   1486 
   1487 				case IPV6_HOPOPTS:
   1488 					optval = OPTBIT(IN6P_HOPOPTS);
   1489 					break;
   1490 
   1491 				case IPV6_DSTOPTS:
   1492 					optval = OPTBIT(IN6P_DSTOPTS);
   1493 					break;
   1494 
   1495 				case IPV6_RTHDR:
   1496 					optval = OPTBIT(IN6P_RTHDR);
   1497 					break;
   1498 
   1499 				case IPV6_CHECKSUM:
   1500 					optval = in6p->in6p_cksum;
   1501 					break;
   1502 
   1503 				case IPV6_FAITH:
   1504 					optval = OPTBIT(IN6P_FAITH);
   1505 					break;
   1506 
   1507 #ifndef INET6_BINDV6ONLY
   1508 				case IPV6_BINDV6ONLY:
   1509 					optval = OPTBIT(IN6P_BINDV6ONLY);
   1510 					break;
   1511 #endif
   1512 				}
   1513 				*mtod(m, int *) = optval;
   1514 				break;
   1515 
   1516 			case IPV6_MULTICAST_IF:
   1517 			case IPV6_MULTICAST_HOPS:
   1518 			case IPV6_MULTICAST_LOOP:
   1519 			case IPV6_JOIN_GROUP:
   1520 			case IPV6_LEAVE_GROUP:
   1521 				error = ip6_getmoptions(optname, in6p->in6p_moptions, mp);
   1522 				break;
   1523 
   1524 #ifdef IPSEC
   1525 			case IPV6_IPSEC_POLICY:
   1526 			{
   1527 				caddr_t req = NULL;
   1528 				size_t len = 0;
   1529 
   1530 				if (m) {
   1531 					req = mtod(m, caddr_t);
   1532 					len = m->m_len;
   1533 				}
   1534 				error = ipsec6_get_policy(in6p, req, len, mp);
   1535 				break;
   1536 			}
   1537 #endif /* IPSEC */
   1538 
   1539 			default:
   1540 				error = ENOPROTOOPT;
   1541 				break;
   1542 			}
   1543 			break;
   1544 		}
   1545 	} else {
   1546 		error = EINVAL;
   1547 		if (op == PRCO_SETOPT && *mp)
   1548 			(void)m_free(*mp);
   1549 	}
   1550 	return(error);
   1551 }
   1552 
   1553 /*
   1554  * Set up IP6 options in pcb for insertion in output packets.
   1555  * Store in mbuf with pointer in pcbopt, adding pseudo-option
   1556  * with destination address if source routed.
   1557  */
   1558 static int
   1559 ip6_pcbopts(pktopt, m, so)
   1560 	struct ip6_pktopts **pktopt;
   1561 	register struct mbuf *m;
   1562 	struct socket *so;
   1563 {
   1564 	register struct ip6_pktopts *opt = *pktopt;
   1565 	int error = 0;
   1566 	struct proc *p = curproc;	/* XXX */
   1567 	int priv = 0;
   1568 
   1569 	/* turn off any old options. */
   1570 	if (opt) {
   1571 		if (opt->ip6po_m)
   1572 			(void)m_free(opt->ip6po_m);
   1573 	} else
   1574 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
   1575 	*pktopt = 0;
   1576 
   1577 	if (!m || m->m_len == 0) {
   1578 		/*
   1579 		 * Only turning off any previous options.
   1580 		 */
   1581 		if (opt)
   1582 			free(opt, M_IP6OPT);
   1583 		if (m)
   1584 			(void)m_free(m);
   1585 		return(0);
   1586 	}
   1587 
   1588 	/*  set options specified by user. */
   1589 	if (p && !suser(p->p_ucred, &p->p_acflag))
   1590 		priv = 1;
   1591 	if ((error = ip6_setpktoptions(m, opt, priv)) != 0) {
   1592 		(void)m_free(m);
   1593 		return(error);
   1594 	}
   1595 	*pktopt = opt;
   1596 	return(0);
   1597 }
   1598 
   1599 /*
   1600  * Set the IP6 multicast options in response to user setsockopt().
   1601  */
   1602 static int
   1603 ip6_setmoptions(optname, im6op, m)
   1604 	int optname;
   1605 	struct ip6_moptions **im6op;
   1606 	struct mbuf *m;
   1607 {
   1608 	int error = 0;
   1609 	u_int loop, ifindex;
   1610 	struct ipv6_mreq *mreq;
   1611 	struct ifnet *ifp;
   1612 	struct ip6_moptions *im6o = *im6op;
   1613 	struct route_in6 ro;
   1614 	struct sockaddr_in6 *dst;
   1615 	struct in6_multi_mship *imm;
   1616 	struct proc *p = curproc;	/* XXX */
   1617 
   1618 	if (im6o == NULL) {
   1619 		/*
   1620 		 * No multicast option buffer attached to the pcb;
   1621 		 * allocate one and initialize to default values.
   1622 		 */
   1623 		im6o = (struct ip6_moptions *)
   1624 			malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
   1625 
   1626 		if (im6o == NULL)
   1627 			return(ENOBUFS);
   1628 		*im6op = im6o;
   1629 		im6o->im6o_multicast_ifp = NULL;
   1630 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
   1631 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
   1632 		LIST_INIT(&im6o->im6o_memberships);
   1633 	}
   1634 
   1635 	switch (optname) {
   1636 
   1637 	case IPV6_MULTICAST_IF:
   1638 		/*
   1639 		 * Select the interface for outgoing multicast packets.
   1640 		 */
   1641 		if (m == NULL || m->m_len != sizeof(u_int)) {
   1642 			error = EINVAL;
   1643 			break;
   1644 		}
   1645 		ifindex = *(mtod(m, u_int *));
   1646 		if (ifindex < 0 || if_index < ifindex) {
   1647 			error = ENXIO;	/* XXX EINVAL? */
   1648 			break;
   1649 		}
   1650 		ifp = ifindex2ifnet[ifindex];
   1651 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
   1652 			error = EADDRNOTAVAIL;
   1653 			break;
   1654 		}
   1655 		im6o->im6o_multicast_ifp = ifp;
   1656 		break;
   1657 
   1658 	case IPV6_MULTICAST_HOPS:
   1659 	    {
   1660 		/*
   1661 		 * Set the IP6 hoplimit for outgoing multicast packets.
   1662 		 */
   1663 		int optval;
   1664 		if (m == NULL || m->m_len != sizeof(int)) {
   1665 			error = EINVAL;
   1666 			break;
   1667 		}
   1668 		optval = *(mtod(m, u_int *));
   1669 		if (optval < -1 || optval >= 256)
   1670 			error = EINVAL;
   1671 		else if (optval == -1)
   1672 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
   1673 		else
   1674 			im6o->im6o_multicast_hlim = optval;
   1675 		break;
   1676 	    }
   1677 
   1678 	case IPV6_MULTICAST_LOOP:
   1679 		/*
   1680 		 * Set the loopback flag for outgoing multicast packets.
   1681 		 * Must be zero or one.
   1682 		 */
   1683 		if (m == NULL || m->m_len != sizeof(u_int) ||
   1684 		   (loop = *(mtod(m, u_int *))) > 1) {
   1685 			error = EINVAL;
   1686 			break;
   1687 		}
   1688 		im6o->im6o_multicast_loop = loop;
   1689 		break;
   1690 
   1691 	case IPV6_JOIN_GROUP:
   1692 		/*
   1693 		 * Add a multicast group membership.
   1694 		 * Group must be a valid IP6 multicast address.
   1695 		 */
   1696 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
   1697 			error = EINVAL;
   1698 			break;
   1699 		}
   1700 		mreq = mtod(m, struct ipv6_mreq *);
   1701 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
   1702 			/*
   1703 			 * We use the unspecified address to specify to accept
   1704 			 * all multicast addresses. Only super user is allowed
   1705 			 * to do this.
   1706 			 */
   1707 			if (suser(p->p_ucred, &p->p_acflag)) {
   1708 				error = EACCES;
   1709 				break;
   1710 			}
   1711 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
   1712 			error = EINVAL;
   1713 			break;
   1714 		}
   1715 
   1716 		/*
   1717 		 * If the interface is specified, validate it.
   1718 		 */
   1719 		if (mreq->ipv6mr_interface < 0
   1720 		 || if_index < mreq->ipv6mr_interface) {
   1721 			error = ENXIO;	/* XXX EINVAL? */
   1722 			break;
   1723 		}
   1724 		/*
   1725 		 * If no interface was explicitly specified, choose an
   1726 		 * appropriate one according to the given multicast address.
   1727 		 */
   1728 		if (mreq->ipv6mr_interface == 0) {
   1729 			/*
   1730 			 * If the multicast address is in node-local scope,
   1731 			 * the interface should be a loopback interface.
   1732 			 * Otherwise, look up the routing table for the
   1733 			 * address, and choose the outgoing interface.
   1734 			 *   XXX: is it a good approach?
   1735 			 */
   1736 			if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
   1737 				ifp = &loif[0];
   1738 			} else {
   1739 				ro.ro_rt = NULL;
   1740 				dst = (struct sockaddr_in6 *)&ro.ro_dst;
   1741 				bzero(dst, sizeof(*dst));
   1742 				dst->sin6_len = sizeof(struct sockaddr_in6);
   1743 				dst->sin6_family = AF_INET6;
   1744 				dst->sin6_addr = mreq->ipv6mr_multiaddr;
   1745 				rtalloc((struct route *)&ro);
   1746 				if (ro.ro_rt == NULL) {
   1747 					error = EADDRNOTAVAIL;
   1748 					break;
   1749 				}
   1750 				ifp = ro.ro_rt->rt_ifp;
   1751 				rtfree(ro.ro_rt);
   1752 			}
   1753 		} else
   1754 			ifp = ifindex2ifnet[mreq->ipv6mr_interface];
   1755 
   1756 		/*
   1757 		 * See if we found an interface, and confirm that it
   1758 		 * supports multicast
   1759 		 */
   1760 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
   1761 			error = EADDRNOTAVAIL;
   1762 			break;
   1763 		}
   1764 		/*
   1765 		 * Put interface index into the multicast address,
   1766 		 * if the address has link-local scope.
   1767 		 */
   1768 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
   1769 			mreq->ipv6mr_multiaddr.s6_addr16[1]
   1770 				= htons(mreq->ipv6mr_interface);
   1771 		}
   1772 		/*
   1773 		 * See if the membership already exists.
   1774 		 */
   1775 		for (imm = im6o->im6o_memberships.lh_first;
   1776 		     imm != NULL; imm = imm->i6mm_chain.le_next)
   1777 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
   1778 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
   1779 					       &mreq->ipv6mr_multiaddr))
   1780 				break;
   1781 		if (imm != NULL) {
   1782 			error = EADDRINUSE;
   1783 			break;
   1784 		}
   1785 		/*
   1786 		 * Everything looks good; add a new record to the multicast
   1787 		 * address list for the given interface.
   1788 		 */
   1789 		imm = malloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
   1790 		if (imm == NULL) {
   1791 			error = ENOBUFS;
   1792 			break;
   1793 		}
   1794 		if ((imm->i6mm_maddr =
   1795 		     in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
   1796 			free(imm, M_IPMADDR);
   1797 			break;
   1798 		}
   1799 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
   1800 		break;
   1801 
   1802 	case IPV6_LEAVE_GROUP:
   1803 		/*
   1804 		 * Drop a multicast group membership.
   1805 		 * Group must be a valid IP6 multicast address.
   1806 		 */
   1807 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
   1808 			error = EINVAL;
   1809 			break;
   1810 		}
   1811 		mreq = mtod(m, struct ipv6_mreq *);
   1812 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
   1813 			if (suser(p->p_ucred, &p->p_acflag)) {
   1814 				error = EACCES;
   1815 				break;
   1816 			}
   1817 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
   1818 			error = EINVAL;
   1819 			break;
   1820 		}
   1821 		/*
   1822 		 * If an interface address was specified, get a pointer
   1823 		 * to its ifnet structure.
   1824 		 */
   1825 		if (mreq->ipv6mr_interface < 0
   1826 		 || if_index < mreq->ipv6mr_interface) {
   1827 			error = ENXIO;	/* XXX EINVAL? */
   1828 			break;
   1829 		}
   1830 		ifp = ifindex2ifnet[mreq->ipv6mr_interface];
   1831 		/*
   1832 		 * Put interface index into the multicast address,
   1833 		 * if the address has link-local scope.
   1834 		 */
   1835 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
   1836 			mreq->ipv6mr_multiaddr.s6_addr16[1]
   1837 				= htons(mreq->ipv6mr_interface);
   1838 		}
   1839 		/*
   1840 		 * Find the membership in the membership list.
   1841 		 */
   1842 		for (imm = im6o->im6o_memberships.lh_first;
   1843 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
   1844 			if ((ifp == NULL ||
   1845 			     imm->i6mm_maddr->in6m_ifp == ifp) &&
   1846 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
   1847 					       &mreq->ipv6mr_multiaddr))
   1848 				break;
   1849 		}
   1850 		if (imm == NULL) {
   1851 			/* Unable to resolve interface */
   1852 			error = EADDRNOTAVAIL;
   1853 			break;
   1854 		}
   1855 		/*
   1856 		 * Give up the multicast address record to which the
   1857 		 * membership points.
   1858 		 */
   1859 		LIST_REMOVE(imm, i6mm_chain);
   1860 		in6_delmulti(imm->i6mm_maddr);
   1861 		free(imm, M_IPMADDR);
   1862 		break;
   1863 
   1864 	default:
   1865 		error = EOPNOTSUPP;
   1866 		break;
   1867 	}
   1868 
   1869 	/*
   1870 	 * If all options have default values, no need to keep the mbuf.
   1871 	 */
   1872 	if (im6o->im6o_multicast_ifp == NULL &&
   1873 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
   1874 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
   1875 	    im6o->im6o_memberships.lh_first == NULL) {
   1876 		free(*im6op, M_IPMOPTS);
   1877 		*im6op = NULL;
   1878 	}
   1879 
   1880 	return(error);
   1881 }
   1882 
   1883 /*
   1884  * Return the IP6 multicast options in response to user getsockopt().
   1885  */
   1886 static int
   1887 ip6_getmoptions(optname, im6o, mp)
   1888 	int optname;
   1889 	register struct ip6_moptions *im6o;
   1890 	register struct mbuf **mp;
   1891 {
   1892 	u_int *hlim, *loop, *ifindex;
   1893 
   1894 	*mp = m_get(M_WAIT, MT_SOOPTS);
   1895 
   1896 	switch (optname) {
   1897 
   1898 	case IPV6_MULTICAST_IF:
   1899 		ifindex = mtod(*mp, u_int *);
   1900 		(*mp)->m_len = sizeof(u_int);
   1901 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
   1902 			*ifindex = 0;
   1903 		else
   1904 			*ifindex = im6o->im6o_multicast_ifp->if_index;
   1905 		return(0);
   1906 
   1907 	case IPV6_MULTICAST_HOPS:
   1908 		hlim = mtod(*mp, u_int *);
   1909 		(*mp)->m_len = sizeof(u_int);
   1910 		if (im6o == NULL)
   1911 			*hlim = ip6_defmcasthlim;
   1912 		else
   1913 			*hlim = im6o->im6o_multicast_hlim;
   1914 		return(0);
   1915 
   1916 	case IPV6_MULTICAST_LOOP:
   1917 		loop = mtod(*mp, u_int *);
   1918 		(*mp)->m_len = sizeof(u_int);
   1919 		if (im6o == NULL)
   1920 			*loop = ip6_defmcasthlim;
   1921 		else
   1922 			*loop = im6o->im6o_multicast_loop;
   1923 		return(0);
   1924 
   1925 	default:
   1926 		return(EOPNOTSUPP);
   1927 	}
   1928 }
   1929 
   1930 /*
   1931  * Discard the IP6 multicast options.
   1932  */
   1933 void
   1934 ip6_freemoptions(im6o)
   1935 	register struct ip6_moptions *im6o;
   1936 {
   1937 	struct in6_multi_mship *imm;
   1938 
   1939 	if (im6o == NULL)
   1940 		return;
   1941 
   1942 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
   1943 		LIST_REMOVE(imm, i6mm_chain);
   1944 		if (imm->i6mm_maddr)
   1945 			in6_delmulti(imm->i6mm_maddr);
   1946 		free(imm, M_IPMADDR);
   1947 	}
   1948 	free(im6o, M_IPMOPTS);
   1949 }
   1950 
   1951 /*
   1952  * Set IPv6 outgoing packet options based on advanced API.
   1953  */
   1954 int
   1955 ip6_setpktoptions(control, opt, priv)
   1956 	struct mbuf *control;
   1957 	struct ip6_pktopts *opt;
   1958 	int priv;
   1959 {
   1960 	register struct cmsghdr *cm = 0;
   1961 
   1962 	if (control == 0 || opt == 0)
   1963 		return(EINVAL);
   1964 
   1965 	bzero(opt, sizeof(*opt));
   1966 	opt->ip6po_hlim = -1; /* -1 means to use default hop limit */
   1967 
   1968 	/*
   1969 	 * XXX: Currently, we assume all the optional information is stored
   1970 	 * in a single mbuf.
   1971 	 */
   1972 	if (control->m_next)
   1973 		return(EINVAL);
   1974 
   1975 	opt->ip6po_m = control;
   1976 
   1977 	for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
   1978 		     control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
   1979 		cm = mtod(control, struct cmsghdr *);
   1980 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
   1981 			return(EINVAL);
   1982 		if (cm->cmsg_level != IPPROTO_IPV6)
   1983 			continue;
   1984 
   1985 		switch(cm->cmsg_type) {
   1986 		case IPV6_PKTINFO:
   1987 			if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo)))
   1988 				return(EINVAL);
   1989 			opt->ip6po_pktinfo = (struct in6_pktinfo *)CMSG_DATA(cm);
   1990 			if (opt->ip6po_pktinfo->ipi6_ifindex &&
   1991 			    IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr))
   1992 				opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] =
   1993 					htons(opt->ip6po_pktinfo->ipi6_ifindex);
   1994 
   1995 			if (opt->ip6po_pktinfo->ipi6_ifindex > if_index
   1996 			 || opt->ip6po_pktinfo->ipi6_ifindex < 0) {
   1997 				return(ENXIO);
   1998 			}
   1999 
   2000 			/*
   2001 			 * Check if the requested source address is indeed a
   2002 			 * unicast address assigned to the node, and can be
   2003 			 * used as the packet's source address.
   2004 			 */
   2005 			if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
   2006 				struct ifaddr *ia;
   2007 				struct in6_ifaddr *ia6;
   2008 				struct sockaddr_in6 sin6;
   2009 
   2010 				bzero(&sin6, sizeof(sin6));
   2011 				sin6.sin6_len = sizeof(sin6);
   2012 				sin6.sin6_family = AF_INET6;
   2013 				sin6.sin6_addr =
   2014 					opt->ip6po_pktinfo->ipi6_addr;
   2015 				ia = ifa_ifwithaddr(sin6tosa(&sin6));
   2016 				if (ia == NULL ||
   2017 				    (opt->ip6po_pktinfo->ipi6_ifindex &&
   2018 				     (ia->ifa_ifp->if_index !=
   2019 				      opt->ip6po_pktinfo->ipi6_ifindex))) {
   2020 					return(EADDRNOTAVAIL);
   2021 				}
   2022 				ia6 = (struct in6_ifaddr *)ia;
   2023 				if ((ia6->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) != 0) {
   2024 					return(EADDRNOTAVAIL);
   2025 				}
   2026 
   2027 				/*
   2028 				 * Check if the requested source address is
   2029 				 * indeed a unicast address assigned to the
   2030 				 * node.
   2031 				 */
   2032 				if (IN6_IS_ADDR_MULTICAST(&opt->ip6po_pktinfo->ipi6_addr))
   2033 					return(EADDRNOTAVAIL);
   2034 			}
   2035 			break;
   2036 
   2037 		case IPV6_HOPLIMIT:
   2038 			if (cm->cmsg_len != CMSG_LEN(sizeof(int)))
   2039 				return(EINVAL);
   2040 
   2041 			opt->ip6po_hlim = *(int *)CMSG_DATA(cm);
   2042 			if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255)
   2043 				return(EINVAL);
   2044 			break;
   2045 
   2046 		case IPV6_NEXTHOP:
   2047 			if (!priv)
   2048 				return(EPERM);
   2049 
   2050 			if (cm->cmsg_len < sizeof(u_char) ||
   2051 			    /* check if cmsg_len is large enough for sa_len */
   2052 			    cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm)))
   2053 				return(EINVAL);
   2054 
   2055 			opt->ip6po_nexthop = (struct sockaddr *)CMSG_DATA(cm);
   2056 
   2057 			break;
   2058 
   2059 		case IPV6_HOPOPTS:
   2060 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh)))
   2061 				return(EINVAL);
   2062 			opt->ip6po_hbh = (struct ip6_hbh *)CMSG_DATA(cm);
   2063 			if (cm->cmsg_len !=
   2064 			    CMSG_LEN((opt->ip6po_hbh->ip6h_len + 1) << 3))
   2065 				return(EINVAL);
   2066 			break;
   2067 
   2068 		case IPV6_DSTOPTS:
   2069 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest)))
   2070 				return(EINVAL);
   2071 
   2072 			/*
   2073 			 * If there is no routing header yet, the destination
   2074 			 * options header should be put on the 1st part.
   2075 			 * Otherwise, the header should be on the 2nd part.
   2076 			 * (See RFC 2460, section 4.1)
   2077 			 */
   2078 			if (opt->ip6po_rthdr == NULL) {
   2079 				opt->ip6po_dest1 =
   2080 					(struct ip6_dest *)CMSG_DATA(cm);
   2081 				if (cm->cmsg_len !=
   2082 				    CMSG_LEN((opt->ip6po_dest1->ip6d_len + 1)
   2083 					     << 3))
   2084 					return(EINVAL);
   2085 			}
   2086 			else {
   2087 				opt->ip6po_dest2 =
   2088 					(struct ip6_dest *)CMSG_DATA(cm);
   2089 				if (cm->cmsg_len !=
   2090 				    CMSG_LEN((opt->ip6po_dest2->ip6d_len + 1)
   2091 					     << 3))
   2092 					return(EINVAL);
   2093 			}
   2094 			break;
   2095 
   2096 		case IPV6_RTHDR:
   2097 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr)))
   2098 				return(EINVAL);
   2099 			opt->ip6po_rthdr = (struct ip6_rthdr *)CMSG_DATA(cm);
   2100 			if (cm->cmsg_len !=
   2101 			    CMSG_LEN((opt->ip6po_rthdr->ip6r_len + 1) << 3))
   2102 				return(EINVAL);
   2103 			switch(opt->ip6po_rthdr->ip6r_type) {
   2104 			case IPV6_RTHDR_TYPE_0:
   2105 				if (opt->ip6po_rthdr->ip6r_segleft == 0)
   2106 					return(EINVAL);
   2107 				break;
   2108 			default:
   2109 				return(EINVAL);
   2110 			}
   2111 			break;
   2112 
   2113 		default:
   2114 			return(ENOPROTOOPT);
   2115 		}
   2116 	}
   2117 
   2118 	return(0);
   2119 }
   2120 
   2121 /*
   2122  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
   2123  * packet to the input queue of a specified interface.  Note that this
   2124  * calls the output routine of the loopback "driver", but with an interface
   2125  * pointer that might NOT be &loif -- easier than replicating that code here.
   2126  */
   2127 void
   2128 ip6_mloopback(ifp, m, dst)
   2129 	struct ifnet *ifp;
   2130 	register struct mbuf *m;
   2131 	register struct sockaddr_in6 *dst;
   2132 {
   2133 	struct mbuf *copym;
   2134 	struct ip6_hdr *ip6;
   2135 
   2136 	copym = m_copy(m, 0, M_COPYALL);
   2137 	if (copym == NULL)
   2138 		return;
   2139 
   2140 	/*
   2141 	 * Make sure to deep-copy IPv6 header portion in case the data
   2142 	 * is in an mbuf cluster, so that we can safely override the IPv6
   2143 	 * header portion later.
   2144 	 */
   2145 	if ((copym->m_flags & M_EXT) != 0 ||
   2146 	    copym->m_len < sizeof(struct ip6_hdr)) {
   2147 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
   2148 		if (copym == NULL)
   2149 			return;
   2150 	}
   2151 
   2152 #ifdef DIAGNOSTIC
   2153 	if (copym->m_len < sizeof(*ip6)) {
   2154 		m_freem(copym);
   2155 		return;
   2156 	}
   2157 #endif
   2158 
   2159 #ifndef FAKE_LOOPBACK_IF
   2160 	if ((ifp->if_flags & IFF_LOOPBACK) == 0)
   2161 #else
   2162 	if (1)
   2163 #endif
   2164 	{
   2165 		ip6 = mtod(copym, struct ip6_hdr *);
   2166 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
   2167 			ip6->ip6_src.s6_addr16[1] = 0;
   2168 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
   2169 			ip6->ip6_dst.s6_addr16[1] = 0;
   2170 	}
   2171 
   2172 	(void)looutput(ifp, copym, (struct sockaddr *)dst, NULL);
   2173 }
   2174 
   2175 /*
   2176  * Chop IPv6 header off from the payload.
   2177  */
   2178 static int
   2179 ip6_splithdr(m, exthdrs)
   2180 	struct mbuf *m;
   2181 	struct ip6_exthdrs *exthdrs;
   2182 {
   2183 	struct mbuf *mh;
   2184 	struct ip6_hdr *ip6;
   2185 
   2186 	ip6 = mtod(m, struct ip6_hdr *);
   2187 	if (m->m_len > sizeof(*ip6)) {
   2188 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
   2189 		if (mh == 0) {
   2190 			m_freem(m);
   2191 			return ENOBUFS;
   2192 		}
   2193 		M_COPY_PKTHDR(mh, m);
   2194 		MH_ALIGN(mh, sizeof(*ip6));
   2195 		m->m_flags &= ~M_PKTHDR;
   2196 		m->m_len -= sizeof(*ip6);
   2197 		m->m_data += sizeof(*ip6);
   2198 		mh->m_next = m;
   2199 		m = mh;
   2200 		m->m_len = sizeof(*ip6);
   2201 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
   2202 	}
   2203 	exthdrs->ip6e_ip6 = m;
   2204 	return 0;
   2205 }
   2206 
   2207 /*
   2208  * Compute IPv6 extension header length.
   2209  */
   2210 int
   2211 ip6_optlen(in6p)
   2212 	struct in6pcb *in6p;
   2213 {
   2214 	int len;
   2215 
   2216 	if (!in6p->in6p_outputopts)
   2217 		return 0;
   2218 
   2219 	len = 0;
   2220 #define elen(x) \
   2221     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
   2222 
   2223 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
   2224 	len += elen(in6p->in6p_outputopts->ip6po_dest1);
   2225 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
   2226 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
   2227 	return len;
   2228 #undef elen
   2229 }
   2230