Home | History | Annotate | Line # | Download | only in netinet6
ip6_output.c revision 1.51
      1 /*	$NetBSD: ip6_output.c,v 1.51 2002/06/07 14:48:56 itojun Exp $	*/
      2 /*	$KAME: ip6_output.c,v 1.172 2001/03/25 09:55:56 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, 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 <sys/cdefs.h>
     69 __KERNEL_RCSID(0, "$NetBSD: ip6_output.c,v 1.51 2002/06/07 14:48:56 itojun Exp $");
     70 
     71 #include "opt_inet.h"
     72 #include "opt_ipsec.h"
     73 #include "opt_pfil_hooks.h"
     74 
     75 #include <sys/param.h>
     76 #include <sys/malloc.h>
     77 #include <sys/mbuf.h>
     78 #include <sys/errno.h>
     79 #include <sys/protosw.h>
     80 #include <sys/socket.h>
     81 #include <sys/socketvar.h>
     82 #include <sys/systm.h>
     83 #include <sys/proc.h>
     84 
     85 #include <net/if.h>
     86 #include <net/route.h>
     87 #ifdef PFIL_HOOKS
     88 #include <net/pfil.h>
     89 #endif
     90 
     91 #include <netinet/in.h>
     92 #include <netinet/in_var.h>
     93 #include <netinet/ip6.h>
     94 #include <netinet/icmp6.h>
     95 #include <netinet6/ip6_var.h>
     96 #include <netinet6/in6_pcb.h>
     97 #include <netinet6/nd6.h>
     98 
     99 #ifdef IPSEC
    100 #include <netinet6/ipsec.h>
    101 #include <netkey/key.h>
    102 #endif /* IPSEC */
    103 
    104 #include "loop.h"
    105 
    106 #include <net/net_osdep.h>
    107 
    108 #ifdef PFIL_HOOKS
    109 extern struct pfil_head inet6_pfil_hook;	/* XXX */
    110 #endif
    111 
    112 struct ip6_exthdrs {
    113 	struct mbuf *ip6e_ip6;
    114 	struct mbuf *ip6e_hbh;
    115 	struct mbuf *ip6e_dest1;
    116 	struct mbuf *ip6e_rthdr;
    117 	struct mbuf *ip6e_dest2;
    118 };
    119 
    120 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
    121 	struct socket *));
    122 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *));
    123 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **));
    124 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int));
    125 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
    126 	struct ip6_frag **));
    127 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t));
    128 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *));
    129 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *,
    130 	struct ifnet *, struct in6_addr *, u_long *));
    131 
    132 extern struct ifnet loif[NLOOP];
    133 
    134 /*
    135  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
    136  * header (with pri, len, nxt, hlim, src, dst).
    137  * This function may modify ver and hlim only.
    138  * The mbuf chain containing the packet will be freed.
    139  * The mbuf opt, if present, will not be freed.
    140  */
    141 int
    142 ip6_output(m0, opt, ro, flags, im6o, ifpp)
    143 	struct mbuf *m0;
    144 	struct ip6_pktopts *opt;
    145 	struct route_in6 *ro;
    146 	int flags;
    147 	struct ip6_moptions *im6o;
    148 	struct ifnet **ifpp;		/* XXX: just for statistics */
    149 {
    150 	struct ip6_hdr *ip6, *mhip6;
    151 	struct ifnet *ifp, *origifp;
    152 	struct mbuf *m = m0;
    153 	int hlen, tlen, len, off;
    154 	struct route_in6 ip6route;
    155 	struct sockaddr_in6 *dst;
    156 	int error = 0;
    157 	struct in6_ifaddr *ia;
    158 	u_long mtu;
    159 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
    160 	struct ip6_exthdrs exthdrs;
    161 	struct in6_addr finaldst;
    162 	struct route_in6 *ro_pmtu = NULL;
    163 	int hdrsplit = 0;
    164 	int needipsec = 0;
    165 #ifdef IPSEC
    166 	int needipsectun = 0;
    167 	struct socket *so;
    168 	struct secpolicy *sp = NULL;
    169 
    170 	/* for AH processing. stupid to have "socket" variable in IP layer... */
    171 	so = ipsec_getsocket(m);
    172 	(void)ipsec_setsocket(m, NULL);
    173 	ip6 = mtod(m, struct ip6_hdr *);
    174 #endif /* IPSEC */
    175 
    176 #define MAKE_EXTHDR(hp, mp)						\
    177     do {								\
    178 	if (hp) {							\
    179 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
    180 		error = ip6_copyexthdr((mp), (caddr_t)(hp), 		\
    181 		    ((eh)->ip6e_len + 1) << 3);				\
    182 		if (error)						\
    183 			goto freehdrs;					\
    184 	}								\
    185     } while (0)
    186 
    187 	bzero(&exthdrs, sizeof(exthdrs));
    188 	if (opt) {
    189 		/* Hop-by-Hop options header */
    190 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
    191 		/* Destination options header(1st part) */
    192 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
    193 		/* Routing header */
    194 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
    195 		/* Destination options header(2nd part) */
    196 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
    197 	}
    198 
    199 #ifdef IPSEC
    200 	/* get a security policy for this packet */
    201 	if (so == NULL)
    202 		sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
    203 	else
    204 		sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
    205 
    206 	if (sp == NULL) {
    207 		ipsec6stat.out_inval++;
    208 		goto freehdrs;
    209 	}
    210 
    211 	error = 0;
    212 
    213 	/* check policy */
    214 	switch (sp->policy) {
    215 	case IPSEC_POLICY_DISCARD:
    216 		/*
    217 		 * This packet is just discarded.
    218 		 */
    219 		ipsec6stat.out_polvio++;
    220 		goto freehdrs;
    221 
    222 	case IPSEC_POLICY_BYPASS:
    223 	case IPSEC_POLICY_NONE:
    224 		/* no need to do IPsec. */
    225 		needipsec = 0;
    226 		break;
    227 
    228 	case IPSEC_POLICY_IPSEC:
    229 		if (sp->req == NULL) {
    230 			/* XXX should be panic ? */
    231 			printf("ip6_output: No IPsec request specified.\n");
    232 			error = EINVAL;
    233 			goto freehdrs;
    234 		}
    235 		needipsec = 1;
    236 		break;
    237 
    238 	case IPSEC_POLICY_ENTRUST:
    239 	default:
    240 		printf("ip6_output: Invalid policy found. %d\n", sp->policy);
    241 	}
    242 #endif /* IPSEC */
    243 
    244 	/*
    245 	 * Calculate the total length of the extension header chain.
    246 	 * Keep the length of the unfragmentable part for fragmentation.
    247 	 */
    248 	optlen = 0;
    249 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
    250 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
    251 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
    252 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
    253 	/* NOTE: we don't add AH/ESP length here. do that later. */
    254 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
    255 
    256 	/*
    257 	 * If we need IPsec, or there is at least one extension header,
    258 	 * separate IP6 header from the payload.
    259 	 */
    260 	if ((needipsec || optlen) && !hdrsplit) {
    261 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
    262 			m = NULL;
    263 			goto freehdrs;
    264 		}
    265 		m = exthdrs.ip6e_ip6;
    266 		hdrsplit++;
    267 	}
    268 
    269 	/* adjust pointer */
    270 	ip6 = mtod(m, struct ip6_hdr *);
    271 
    272 	/* adjust mbuf packet header length */
    273 	m->m_pkthdr.len += optlen;
    274 	plen = m->m_pkthdr.len - sizeof(*ip6);
    275 
    276 	/* If this is a jumbo payload, insert a jumbo payload option. */
    277 	if (plen > IPV6_MAXPACKET) {
    278 		if (!hdrsplit) {
    279 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
    280 				m = NULL;
    281 				goto freehdrs;
    282 			}
    283 			m = exthdrs.ip6e_ip6;
    284 			hdrsplit++;
    285 		}
    286 		/* adjust pointer */
    287 		ip6 = mtod(m, struct ip6_hdr *);
    288 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
    289 			goto freehdrs;
    290 		ip6->ip6_plen = 0;
    291 	} else
    292 		ip6->ip6_plen = htons(plen);
    293 
    294 	/*
    295 	 * Concatenate headers and fill in next header fields.
    296 	 * Here we have, on "m"
    297 	 *	IPv6 payload
    298 	 * and we insert headers accordingly.  Finally, we should be getting:
    299 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
    300 	 *
    301 	 * during the header composing process, "m" points to IPv6 header.
    302 	 * "mprev" points to an extension header prior to esp.
    303 	 */
    304 	{
    305 		u_char *nexthdrp = &ip6->ip6_nxt;
    306 		struct mbuf *mprev = m;
    307 
    308 		/*
    309 		 * we treat dest2 specially.  this makes IPsec processing
    310 		 * much easier.
    311 		 *
    312 		 * result: IPv6 dest2 payload
    313 		 * m and mprev will point to IPv6 header.
    314 		 */
    315 		if (exthdrs.ip6e_dest2) {
    316 			if (!hdrsplit)
    317 				panic("assumption failed: hdr not split");
    318 			exthdrs.ip6e_dest2->m_next = m->m_next;
    319 			m->m_next = exthdrs.ip6e_dest2;
    320 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
    321 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
    322 		}
    323 
    324 #define MAKE_CHAIN(m, mp, p, i)\
    325     do {\
    326 	if (m) {\
    327 		if (!hdrsplit) \
    328 			panic("assumption failed: hdr not split"); \
    329 		*mtod((m), u_char *) = *(p);\
    330 		*(p) = (i);\
    331 		p = mtod((m), u_char *);\
    332 		(m)->m_next = (mp)->m_next;\
    333 		(mp)->m_next = (m);\
    334 		(mp) = (m);\
    335 	}\
    336     } while (0)
    337 		/*
    338 		 * result: IPv6 hbh dest1 rthdr dest2 payload
    339 		 * m will point to IPv6 header.  mprev will point to the
    340 		 * extension header prior to dest2 (rthdr in the above case).
    341 		 */
    342 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
    343 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
    344 		    IPPROTO_DSTOPTS);
    345 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
    346 		    IPPROTO_ROUTING);
    347 
    348 #ifdef IPSEC
    349 		if (!needipsec)
    350 			goto skip_ipsec2;
    351 
    352 		/*
    353 		 * pointers after IPsec headers are not valid any more.
    354 		 * other pointers need a great care too.
    355 		 * (IPsec routines should not mangle mbufs prior to AH/ESP)
    356 		 */
    357 		exthdrs.ip6e_dest2 = NULL;
    358 
    359 	    {
    360 		struct ip6_rthdr *rh = NULL;
    361 		int segleft_org = 0;
    362 		struct ipsec_output_state state;
    363 
    364 		if (exthdrs.ip6e_rthdr) {
    365 			rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
    366 			segleft_org = rh->ip6r_segleft;
    367 			rh->ip6r_segleft = 0;
    368 		}
    369 
    370 		bzero(&state, sizeof(state));
    371 		state.m = m;
    372 		error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
    373 			&needipsectun);
    374 		m = state.m;
    375 		if (error) {
    376 			/* mbuf is already reclaimed in ipsec6_output_trans. */
    377 			m = NULL;
    378 			switch (error) {
    379 			case EHOSTUNREACH:
    380 			case ENETUNREACH:
    381 			case EMSGSIZE:
    382 			case ENOBUFS:
    383 			case ENOMEM:
    384 				break;
    385 			default:
    386 				printf("ip6_output (ipsec): error code %d\n", error);
    387 				/* FALLTHROUGH */
    388 			case ENOENT:
    389 				/* don't show these error codes to the user */
    390 				error = 0;
    391 				break;
    392 			}
    393 			goto bad;
    394 		}
    395 		if (exthdrs.ip6e_rthdr) {
    396 			/* ah6_output doesn't modify mbuf chain */
    397 			rh->ip6r_segleft = segleft_org;
    398 		}
    399 	    }
    400 skip_ipsec2:;
    401 #endif
    402 	}
    403 
    404 	/*
    405 	 * If there is a routing header, replace destination address field
    406 	 * with the first hop of the routing header.
    407 	 */
    408 	if (exthdrs.ip6e_rthdr) {
    409 		struct ip6_rthdr *rh;
    410 		struct ip6_rthdr0 *rh0;
    411 
    412 		rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
    413 		    struct ip6_rthdr *));
    414 		finaldst = ip6->ip6_dst;
    415 		switch (rh->ip6r_type) {
    416 		case IPV6_RTHDR_TYPE_0:
    417 			 rh0 = (struct ip6_rthdr0 *)rh;
    418 			 ip6->ip6_dst = rh0->ip6r0_addr[0];
    419 			 bcopy((caddr_t)&rh0->ip6r0_addr[1],
    420 			     (caddr_t)&rh0->ip6r0_addr[0],
    421 			     sizeof(struct in6_addr) * (rh0->ip6r0_segleft - 1));
    422 			 rh0->ip6r0_addr[rh0->ip6r0_segleft - 1] = finaldst;
    423 			 break;
    424 		default:	/* is it possible? */
    425 			 error = EINVAL;
    426 			 goto bad;
    427 		}
    428 	}
    429 
    430 	/* Source address validation */
    431 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
    432 	    (flags & IPV6_DADOUTPUT) == 0) {
    433 		error = EOPNOTSUPP;
    434 		ip6stat.ip6s_badscope++;
    435 		goto bad;
    436 	}
    437 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
    438 		error = EOPNOTSUPP;
    439 		ip6stat.ip6s_badscope++;
    440 		goto bad;
    441 	}
    442 
    443 	ip6stat.ip6s_localout++;
    444 
    445 	/*
    446 	 * Route packet.
    447 	 */
    448 	if (ro == 0) {
    449 		ro = &ip6route;
    450 		bzero((caddr_t)ro, sizeof(*ro));
    451 	}
    452 	ro_pmtu = ro;
    453 	if (opt && opt->ip6po_rthdr)
    454 		ro = &opt->ip6po_route;
    455 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
    456 	/*
    457 	 * If there is a cached route,
    458 	 * check that it is to the same destination
    459 	 * and is still up. If not, free it and try again.
    460 	 */
    461 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
    462 	    dst->sin6_family != AF_INET6 ||
    463 	    !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
    464 		RTFREE(ro->ro_rt);
    465 		ro->ro_rt = (struct rtentry *)0;
    466 	}
    467 	if (ro->ro_rt == 0) {
    468 		bzero(dst, sizeof(*dst));
    469 		dst->sin6_family = AF_INET6;
    470 		dst->sin6_len = sizeof(struct sockaddr_in6);
    471 		dst->sin6_addr = ip6->ip6_dst;
    472 	}
    473 #ifdef IPSEC
    474 	if (needipsec && needipsectun) {
    475 		struct ipsec_output_state state;
    476 
    477 		/*
    478 		 * All the extension headers will become inaccessible
    479 		 * (since they can be encrypted).
    480 		 * Don't panic, we need no more updates to extension headers
    481 		 * on inner IPv6 packet (since they are now encapsulated).
    482 		 *
    483 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
    484 		 */
    485 		bzero(&exthdrs, sizeof(exthdrs));
    486 		exthdrs.ip6e_ip6 = m;
    487 
    488 		bzero(&state, sizeof(state));
    489 		state.m = m;
    490 		state.ro = (struct route *)ro;
    491 		state.dst = (struct sockaddr *)dst;
    492 
    493 		error = ipsec6_output_tunnel(&state, sp, flags);
    494 
    495 		m = state.m;
    496 		ro = (struct route_in6 *)state.ro;
    497 		dst = (struct sockaddr_in6 *)state.dst;
    498 		if (error) {
    499 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
    500 			m0 = m = NULL;
    501 			m = NULL;
    502 			switch (error) {
    503 			case EHOSTUNREACH:
    504 			case ENETUNREACH:
    505 			case EMSGSIZE:
    506 			case ENOBUFS:
    507 			case ENOMEM:
    508 				break;
    509 			default:
    510 				printf("ip6_output (ipsec): error code %d\n", error);
    511 				/* FALLTHROUGH */
    512 			case ENOENT:
    513 				/* don't show these error codes to the user */
    514 				error = 0;
    515 				break;
    516 			}
    517 			goto bad;
    518 		}
    519 
    520 		exthdrs.ip6e_ip6 = m;
    521 	}
    522 #endif /* IPSEC */
    523 
    524 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
    525 		/* Unicast */
    526 
    527 #define ifatoia6(ifa)	((struct in6_ifaddr *)(ifa))
    528 #define sin6tosa(sin6)	((struct sockaddr *)(sin6))
    529 		/* xxx
    530 		 * interface selection comes here
    531 		 * if an interface is specified from an upper layer,
    532 		 * ifp must point it.
    533 		 */
    534 		if (ro->ro_rt == 0) {
    535 			/*
    536 			 * non-bsdi always clone routes, if parent is
    537 			 * PRF_CLONING.
    538 			 */
    539 			rtalloc((struct route *)ro);
    540 		}
    541 		if (ro->ro_rt == 0) {
    542 			ip6stat.ip6s_noroute++;
    543 			error = EHOSTUNREACH;
    544 			/* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
    545 			goto bad;
    546 		}
    547 		ia = ifatoia6(ro->ro_rt->rt_ifa);
    548 		ifp = ro->ro_rt->rt_ifp;
    549 		ro->ro_rt->rt_use++;
    550 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
    551 			dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
    552 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
    553 
    554 		in6_ifstat_inc(ifp, ifs6_out_request);
    555 
    556 		/*
    557 		 * Check if the outgoing interface conflicts with
    558 		 * the interface specified by ifi6_ifindex (if specified).
    559 		 * Note that loopback interface is always okay.
    560 		 * (this may happen when we are sending a packet to one of
    561 		 *  our own addresses.)
    562 		 */
    563 		if (opt && opt->ip6po_pktinfo
    564 		 && opt->ip6po_pktinfo->ipi6_ifindex) {
    565 			if (!(ifp->if_flags & IFF_LOOPBACK)
    566 			 && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
    567 				ip6stat.ip6s_noroute++;
    568 				in6_ifstat_inc(ifp, ifs6_out_discard);
    569 				error = EHOSTUNREACH;
    570 				goto bad;
    571 			}
    572 		}
    573 
    574 		if (opt && opt->ip6po_hlim != -1)
    575 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
    576 	} else {
    577 		/* Multicast */
    578 		struct	in6_multi *in6m;
    579 
    580 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
    581 
    582 		/*
    583 		 * See if the caller provided any multicast options
    584 		 */
    585 		ifp = NULL;
    586 		if (im6o != NULL) {
    587 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
    588 			if (im6o->im6o_multicast_ifp != NULL)
    589 				ifp = im6o->im6o_multicast_ifp;
    590 		} else
    591 			ip6->ip6_hlim = ip6_defmcasthlim;
    592 
    593 		/*
    594 		 * See if the caller provided the outgoing interface
    595 		 * as an ancillary data.
    596 		 * Boundary check for ifindex is assumed to be already done.
    597 		 */
    598 		if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
    599 			ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex];
    600 
    601 		/*
    602 		 * If the destination is a node-local scope multicast,
    603 		 * the packet should be loop-backed only.
    604 		 */
    605 		if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
    606 			/*
    607 			 * If the outgoing interface is already specified,
    608 			 * it should be a loopback interface.
    609 			 */
    610 			if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) {
    611 				ip6stat.ip6s_badscope++;
    612 				error = ENETUNREACH; /* XXX: better error? */
    613 				/* XXX correct ifp? */
    614 				in6_ifstat_inc(ifp, ifs6_out_discard);
    615 				goto bad;
    616 			} else {
    617 				ifp = &loif[0];
    618 			}
    619 		}
    620 
    621 		if (opt && opt->ip6po_hlim != -1)
    622 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
    623 
    624 		/*
    625 		 * If caller did not provide an interface lookup a
    626 		 * default in the routing table.  This is either a
    627 		 * default for the speicfied group (i.e. a host
    628 		 * route), or a multicast default (a route for the
    629 		 * ``net'' ff00::/8).
    630 		 */
    631 		if (ifp == NULL) {
    632 			if (ro->ro_rt == 0) {
    633 				ro->ro_rt = rtalloc1((struct sockaddr *)
    634 				    &ro->ro_dst, 0);
    635 			}
    636 			if (ro->ro_rt == 0) {
    637 				ip6stat.ip6s_noroute++;
    638 				error = EHOSTUNREACH;
    639 				/* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
    640 				goto bad;
    641 			}
    642 			ia = ifatoia6(ro->ro_rt->rt_ifa);
    643 			ifp = ro->ro_rt->rt_ifp;
    644 			ro->ro_rt->rt_use++;
    645 		}
    646 
    647 		if ((flags & IPV6_FORWARDING) == 0)
    648 			in6_ifstat_inc(ifp, ifs6_out_request);
    649 		in6_ifstat_inc(ifp, ifs6_out_mcast);
    650 
    651 		/*
    652 		 * Confirm that the outgoing interface supports multicast.
    653 		 */
    654 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
    655 			ip6stat.ip6s_noroute++;
    656 			in6_ifstat_inc(ifp, ifs6_out_discard);
    657 			error = ENETUNREACH;
    658 			goto bad;
    659 		}
    660 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
    661 		if (in6m != NULL &&
    662 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
    663 			/*
    664 			 * If we belong to the destination multicast group
    665 			 * on the outgoing interface, and the caller did not
    666 			 * forbid loopback, loop back a copy.
    667 			 */
    668 			ip6_mloopback(ifp, m, dst);
    669 		} else {
    670 			/*
    671 			 * If we are acting as a multicast router, perform
    672 			 * multicast forwarding as if the packet had just
    673 			 * arrived on the interface to which we are about
    674 			 * to send.  The multicast forwarding function
    675 			 * recursively calls this function, using the
    676 			 * IPV6_FORWARDING flag to prevent infinite recursion.
    677 			 *
    678 			 * Multicasts that are looped back by ip6_mloopback(),
    679 			 * above, will be forwarded by the ip6_input() routine,
    680 			 * if necessary.
    681 			 */
    682 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
    683 				if (ip6_mforward(ip6, ifp, m) != 0) {
    684 					m_freem(m);
    685 					goto done;
    686 				}
    687 			}
    688 		}
    689 		/*
    690 		 * Multicasts with a hoplimit of zero may be looped back,
    691 		 * above, but must not be transmitted on a network.
    692 		 * Also, multicasts addressed to the loopback interface
    693 		 * are not sent -- the above call to ip6_mloopback() will
    694 		 * loop back a copy if this host actually belongs to the
    695 		 * destination group on the loopback interface.
    696 		 */
    697 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
    698 			m_freem(m);
    699 			goto done;
    700 		}
    701 	}
    702 
    703 	/*
    704 	 * Fill the outgoing inteface to tell the upper layer
    705 	 * to increment per-interface statistics.
    706 	 */
    707 	if (ifpp)
    708 		*ifpp = ifp;
    709 
    710 	/* Determine path MTU. */
    711 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu)) != 0)
    712 		goto bad;
    713 
    714 	/*
    715 	 * The caller of this function may specify to use the minimum MTU
    716 	 * in some cases.
    717 	 */
    718 	if (mtu > IPV6_MMTU) {
    719 		if ((flags & IPV6_MINMTU))
    720 			mtu = IPV6_MMTU;
    721 	}
    722 
    723 	/* Fake scoped addresses */
    724 	if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
    725 		/*
    726 		 * If source or destination address is a scoped address, and
    727 		 * the packet is going to be sent to a loopback interface,
    728 		 * we should keep the original interface.
    729 		 */
    730 
    731 		/*
    732 		 * XXX: this is a very experimental and temporary solution.
    733 		 * We eventually have sockaddr_in6 and use the sin6_scope_id
    734 		 * field of the structure here.
    735 		 * We rely on the consistency between two scope zone ids
    736 		 * of source add destination, which should already be assured
    737 		 * Larger scopes than link will be supported in the near
    738 		 * future.
    739 		 */
    740 		origifp = NULL;
    741 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
    742 			origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
    743 		else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
    744 			origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
    745 		/*
    746 		 * XXX: origifp can be NULL even in those two cases above.
    747 		 * For example, if we remove the (only) link-local address
    748 		 * from the loopback interface, and try to send a link-local
    749 		 * address without link-id information.  Then the source
    750 		 * address is ::1, and the destination address is the
    751 		 * link-local address with its s6_addr16[1] being zero.
    752 		 * What is worse, if the packet goes to the loopback interface
    753 		 * by a default rejected route, the null pointer would be
    754 		 * passed to looutput, and the kernel would hang.
    755 		 * The following last resort would prevent such disaster.
    756 		 */
    757 		if (origifp == NULL)
    758 			origifp = ifp;
    759 	} else
    760 		origifp = ifp;
    761 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
    762 		ip6->ip6_src.s6_addr16[1] = 0;
    763 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
    764 		ip6->ip6_dst.s6_addr16[1] = 0;
    765 
    766 	/*
    767 	 * If the outgoing packet contains a hop-by-hop options header,
    768 	 * it must be examined and processed even by the source node.
    769 	 * (RFC 2460, section 4.)
    770 	 */
    771 	if (exthdrs.ip6e_hbh) {
    772 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
    773 		u_int32_t dummy1; /* XXX unused */
    774 		u_int32_t dummy2; /* XXX unused */
    775 
    776 		/*
    777 		 *  XXX: if we have to send an ICMPv6 error to the sender,
    778 		 *       we need the M_LOOP flag since icmp6_error() expects
    779 		 *       the IPv6 and the hop-by-hop options header are
    780 		 *       continuous unless the flag is set.
    781 		 */
    782 		m->m_flags |= M_LOOP;
    783 		m->m_pkthdr.rcvif = ifp;
    784 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
    785 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
    786 		    &dummy1, &dummy2) < 0) {
    787 			/* m was already freed at this point */
    788 			error = EINVAL;/* better error? */
    789 			goto done;
    790 		}
    791 		m->m_flags &= ~M_LOOP; /* XXX */
    792 		m->m_pkthdr.rcvif = NULL;
    793 	}
    794 
    795 #ifdef PFIL_HOOKS
    796 	/*
    797 	 * Run through list of hooks for output packets.
    798 	 */
    799 	if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0)
    800 		goto done;
    801 	if (m == NULL)
    802 		goto done;
    803 	ip6 = mtod(m, struct ip6_hdr *);
    804 #endif /* PFIL_HOOKS */
    805 	/*
    806 	 * Send the packet to the outgoing interface.
    807 	 * If necessary, do IPv6 fragmentation before sending.
    808 	 */
    809 	tlen = m->m_pkthdr.len;
    810 	if (tlen <= mtu) {
    811 #ifdef IFA_STATS
    812 		struct in6_ifaddr *ia6;
    813 		ip6 = mtod(m, struct ip6_hdr *);
    814 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
    815 		if (ia6) {
    816 			/* Record statistics for this interface address. */
    817 			ia6->ia_ifa.ifa_data.ifad_outbytes +=
    818 				m->m_pkthdr.len;
    819 		}
    820 #endif
    821 #ifdef IPSEC
    822 		/* clean ipsec history once it goes out of the node */
    823 		ipsec_delaux(m);
    824 #endif
    825 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
    826 		goto done;
    827 	} else if (mtu < IPV6_MMTU) {
    828 		/*
    829 		 * note that path MTU is never less than IPV6_MMTU
    830 		 * (see icmp6_input).
    831 		 */
    832 		error = EMSGSIZE;
    833 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
    834 		goto bad;
    835 	} else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
    836 		error = EMSGSIZE;
    837 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
    838 		goto bad;
    839 	} else {
    840 		struct mbuf **mnext, *m_frgpart;
    841 		struct ip6_frag *ip6f;
    842 		u_int32_t id = htonl(ip6_id++);
    843 		u_char nextproto;
    844 
    845 		/*
    846 		 * Too large for the destination or interface;
    847 		 * fragment if possible.
    848 		 * Must be able to put at least 8 bytes per fragment.
    849 		 */
    850 		hlen = unfragpartlen;
    851 		if (mtu > IPV6_MAXPACKET)
    852 			mtu = IPV6_MAXPACKET;
    853 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
    854 		if (len < 8) {
    855 			error = EMSGSIZE;
    856 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
    857 			goto bad;
    858 		}
    859 
    860 		mnext = &m->m_nextpkt;
    861 
    862 		/*
    863 		 * Change the next header field of the last header in the
    864 		 * unfragmentable part.
    865 		 */
    866 		if (exthdrs.ip6e_rthdr) {
    867 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
    868 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
    869 		} else if (exthdrs.ip6e_dest1) {
    870 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
    871 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
    872 		} else if (exthdrs.ip6e_hbh) {
    873 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
    874 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
    875 		} else {
    876 			nextproto = ip6->ip6_nxt;
    877 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
    878 		}
    879 
    880 		/*
    881 		 * Loop through length of segment after first fragment,
    882 		 * make new header and copy data of each part and link onto
    883 		 * chain.
    884 		 */
    885 		m0 = m;
    886 		for (off = hlen; off < tlen; off += len) {
    887 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
    888 			if (!m) {
    889 				error = ENOBUFS;
    890 				ip6stat.ip6s_odropped++;
    891 				goto sendorfree;
    892 			}
    893 			m->m_flags = m0->m_flags & M_COPYFLAGS;
    894 			*mnext = m;
    895 			mnext = &m->m_nextpkt;
    896 			m->m_data += max_linkhdr;
    897 			mhip6 = mtod(m, struct ip6_hdr *);
    898 			*mhip6 = *ip6;
    899 			m->m_len = sizeof(*mhip6);
    900 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
    901 			if (error) {
    902 				ip6stat.ip6s_odropped++;
    903 				goto sendorfree;
    904 			}
    905 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
    906 			if (off + len >= tlen)
    907 				len = tlen - off;
    908 			else
    909 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
    910 			mhip6->ip6_plen = htons((u_short)(len + hlen +
    911 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
    912 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
    913 				error = ENOBUFS;
    914 				ip6stat.ip6s_odropped++;
    915 				goto sendorfree;
    916 			}
    917 			m_cat(m, m_frgpart);
    918 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
    919 			m->m_pkthdr.rcvif = (struct ifnet *)0;
    920 			ip6f->ip6f_reserved = 0;
    921 			ip6f->ip6f_ident = id;
    922 			ip6f->ip6f_nxt = nextproto;
    923 			ip6stat.ip6s_ofragments++;
    924 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
    925 		}
    926 
    927 		in6_ifstat_inc(ifp, ifs6_out_fragok);
    928 	}
    929 
    930 	/*
    931 	 * Remove leading garbages.
    932 	 */
    933 sendorfree:
    934 	m = m0->m_nextpkt;
    935 	m0->m_nextpkt = 0;
    936 	m_freem(m0);
    937 	for (m0 = m; m; m = m0) {
    938 		m0 = m->m_nextpkt;
    939 		m->m_nextpkt = 0;
    940 		if (error == 0) {
    941 #ifdef IFA_STATS
    942 			struct in6_ifaddr *ia6;
    943 			ip6 = mtod(m, struct ip6_hdr *);
    944 			ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
    945 			if (ia6) {
    946 				/*
    947 				 * Record statistics for this interface
    948 				 * address.
    949 				 */
    950 				ia6->ia_ifa.ifa_data.ifad_outbytes +=
    951 					m->m_pkthdr.len;
    952 			}
    953 #endif
    954 #ifdef IPSEC
    955 			/* clean ipsec history once it goes out of the node */
    956 			ipsec_delaux(m);
    957 #endif
    958 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
    959 		} else
    960 			m_freem(m);
    961 	}
    962 
    963 	if (error == 0)
    964 		ip6stat.ip6s_fragmented++;
    965 
    966 done:
    967 	if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
    968 		RTFREE(ro->ro_rt);
    969 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
    970 		RTFREE(ro_pmtu->ro_rt);
    971 	}
    972 
    973 #ifdef IPSEC
    974 	if (sp != NULL)
    975 		key_freesp(sp);
    976 #endif /* IPSEC */
    977 
    978 	return(error);
    979 
    980 freehdrs:
    981 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
    982 	m_freem(exthdrs.ip6e_dest1);
    983 	m_freem(exthdrs.ip6e_rthdr);
    984 	m_freem(exthdrs.ip6e_dest2);
    985 	/* FALLTHROUGH */
    986 bad:
    987 	m_freem(m);
    988 	goto done;
    989 }
    990 
    991 static int
    992 ip6_copyexthdr(mp, hdr, hlen)
    993 	struct mbuf **mp;
    994 	caddr_t hdr;
    995 	int hlen;
    996 {
    997 	struct mbuf *m;
    998 
    999 	if (hlen > MCLBYTES)
   1000 		return(ENOBUFS); /* XXX */
   1001 
   1002 	MGET(m, M_DONTWAIT, MT_DATA);
   1003 	if (!m)
   1004 		return(ENOBUFS);
   1005 
   1006 	if (hlen > MLEN) {
   1007 		MCLGET(m, M_DONTWAIT);
   1008 		if ((m->m_flags & M_EXT) == 0) {
   1009 			m_free(m);
   1010 			return(ENOBUFS);
   1011 		}
   1012 	}
   1013 	m->m_len = hlen;
   1014 	if (hdr)
   1015 		bcopy(hdr, mtod(m, caddr_t), hlen);
   1016 
   1017 	*mp = m;
   1018 	return(0);
   1019 }
   1020 
   1021 /*
   1022  * Insert jumbo payload option.
   1023  */
   1024 static int
   1025 ip6_insert_jumboopt(exthdrs, plen)
   1026 	struct ip6_exthdrs *exthdrs;
   1027 	u_int32_t plen;
   1028 {
   1029 	struct mbuf *mopt;
   1030 	u_char *optbuf;
   1031 	u_int32_t v;
   1032 
   1033 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
   1034 
   1035 	/*
   1036 	 * If there is no hop-by-hop options header, allocate new one.
   1037 	 * If there is one but it doesn't have enough space to store the
   1038 	 * jumbo payload option, allocate a cluster to store the whole options.
   1039 	 * Otherwise, use it to store the options.
   1040 	 */
   1041 	if (exthdrs->ip6e_hbh == 0) {
   1042 		MGET(mopt, M_DONTWAIT, MT_DATA);
   1043 		if (mopt == 0)
   1044 			return(ENOBUFS);
   1045 		mopt->m_len = JUMBOOPTLEN;
   1046 		optbuf = mtod(mopt, u_char *);
   1047 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
   1048 		exthdrs->ip6e_hbh = mopt;
   1049 	} else {
   1050 		struct ip6_hbh *hbh;
   1051 
   1052 		mopt = exthdrs->ip6e_hbh;
   1053 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
   1054 			/*
   1055 			 * XXX assumption:
   1056 			 * - exthdrs->ip6e_hbh is not referenced from places
   1057 			 *   other than exthdrs.
   1058 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
   1059 			 */
   1060 			int oldoptlen = mopt->m_len;
   1061 			struct mbuf *n;
   1062 
   1063 			/*
   1064 			 * XXX: give up if the whole (new) hbh header does
   1065 			 * not fit even in an mbuf cluster.
   1066 			 */
   1067 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
   1068 				return(ENOBUFS);
   1069 
   1070 			/*
   1071 			 * As a consequence, we must always prepare a cluster
   1072 			 * at this point.
   1073 			 */
   1074 			MGET(n, M_DONTWAIT, MT_DATA);
   1075 			if (n) {
   1076 				MCLGET(n, M_DONTWAIT);
   1077 				if ((n->m_flags & M_EXT) == 0) {
   1078 					m_freem(n);
   1079 					n = NULL;
   1080 				}
   1081 			}
   1082 			if (!n)
   1083 				return(ENOBUFS);
   1084 			n->m_len = oldoptlen + JUMBOOPTLEN;
   1085 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
   1086 			      oldoptlen);
   1087 			optbuf = mtod(n, caddr_t) + oldoptlen;
   1088 			m_freem(mopt);
   1089 			mopt = exthdrs->ip6e_hbh = n;
   1090 		} else {
   1091 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
   1092 			mopt->m_len += JUMBOOPTLEN;
   1093 		}
   1094 		optbuf[0] = IP6OPT_PADN;
   1095 		optbuf[1] = 1;
   1096 
   1097 		/*
   1098 		 * Adjust the header length according to the pad and
   1099 		 * the jumbo payload option.
   1100 		 */
   1101 		hbh = mtod(mopt, struct ip6_hbh *);
   1102 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
   1103 	}
   1104 
   1105 	/* fill in the option. */
   1106 	optbuf[2] = IP6OPT_JUMBO;
   1107 	optbuf[3] = 4;
   1108 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
   1109 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
   1110 
   1111 	/* finally, adjust the packet header length */
   1112 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
   1113 
   1114 	return(0);
   1115 #undef JUMBOOPTLEN
   1116 }
   1117 
   1118 /*
   1119  * Insert fragment header and copy unfragmentable header portions.
   1120  */
   1121 static int
   1122 ip6_insertfraghdr(m0, m, hlen, frghdrp)
   1123 	struct mbuf *m0, *m;
   1124 	int hlen;
   1125 	struct ip6_frag **frghdrp;
   1126 {
   1127 	struct mbuf *n, *mlast;
   1128 
   1129 	if (hlen > sizeof(struct ip6_hdr)) {
   1130 		n = m_copym(m0, sizeof(struct ip6_hdr),
   1131 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
   1132 		if (n == 0)
   1133 			return(ENOBUFS);
   1134 		m->m_next = n;
   1135 	} else
   1136 		n = m;
   1137 
   1138 	/* Search for the last mbuf of unfragmentable part. */
   1139 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
   1140 		;
   1141 
   1142 	if ((mlast->m_flags & M_EXT) == 0 &&
   1143 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
   1144 		/* use the trailing space of the last mbuf for the fragment hdr */
   1145 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
   1146 		    mlast->m_len);
   1147 		mlast->m_len += sizeof(struct ip6_frag);
   1148 		m->m_pkthdr.len += sizeof(struct ip6_frag);
   1149 	} else {
   1150 		/* allocate a new mbuf for the fragment header */
   1151 		struct mbuf *mfrg;
   1152 
   1153 		MGET(mfrg, M_DONTWAIT, MT_DATA);
   1154 		if (mfrg == 0)
   1155 			return(ENOBUFS);
   1156 		mfrg->m_len = sizeof(struct ip6_frag);
   1157 		*frghdrp = mtod(mfrg, struct ip6_frag *);
   1158 		mlast->m_next = mfrg;
   1159 	}
   1160 
   1161 	return(0);
   1162 }
   1163 
   1164 static int
   1165 ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup)
   1166 	struct route_in6 *ro_pmtu, *ro;
   1167 	struct ifnet *ifp;
   1168 	struct in6_addr *dst;
   1169 	u_long *mtup;
   1170 {
   1171 	u_int32_t mtu = 0;
   1172 	int error = 0;
   1173 
   1174 	if (ro_pmtu != ro) {
   1175 		/* The first hop and the final destination may differ. */
   1176 		struct sockaddr_in6 *sa6_dst =
   1177 		    (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
   1178 		if (ro_pmtu->ro_rt &&
   1179 		    ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
   1180 		     !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
   1181 			RTFREE(ro_pmtu->ro_rt);
   1182 			ro_pmtu->ro_rt = (struct rtentry *)0;
   1183 		}
   1184 		if (ro_pmtu->ro_rt == 0) {
   1185 			bzero(sa6_dst, sizeof(*sa6_dst));
   1186 			sa6_dst->sin6_family = AF_INET6;
   1187 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
   1188 			sa6_dst->sin6_addr = *dst;
   1189 
   1190 			rtalloc((struct route *)ro_pmtu);
   1191 		}
   1192 	}
   1193 	if (ro_pmtu->ro_rt) {
   1194 		u_int32_t ifmtu;
   1195 
   1196 		if (ifp == NULL)
   1197 			ifp = ro_pmtu->ro_rt->rt_ifp;
   1198 		ifmtu = IN6_LINKMTU(ifp);
   1199 		mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
   1200 		if (mtu == 0)
   1201 			mtu = ifmtu;
   1202 		else if (mtu > ifmtu) {
   1203 			/*
   1204 			 * The MTU on the route is larger than the MTU on
   1205 			 * the interface!  This shouldn't happen, unless the
   1206 			 * MTU of the interface has been changed after the
   1207 			 * interface was brought up.  Change the MTU in the
   1208 			 * route to match the interface MTU (as long as the
   1209 			 * field isn't locked).
   1210 			 *
   1211 			 * if MTU on the route is 0, we need to fix the MTU.
   1212 			 * this case happens with path MTU discovery timeouts.
   1213 			 */
   1214 			mtu = ifmtu;
   1215 			if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU))
   1216 				ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
   1217 		}
   1218 	} else if (ifp) {
   1219 		mtu = IN6_LINKMTU(ifp);
   1220 	} else
   1221 		error = EHOSTUNREACH; /* XXX */
   1222 
   1223 	*mtup = mtu;
   1224 	return(error);
   1225 }
   1226 
   1227 /*
   1228  * IP6 socket option processing.
   1229  */
   1230 int
   1231 ip6_ctloutput(op, so, level, optname, mp)
   1232 	int op;
   1233 	struct socket *so;
   1234 	int level, optname;
   1235 	struct mbuf **mp;
   1236 {
   1237 	struct in6pcb *in6p = sotoin6pcb(so);
   1238 	struct mbuf *m = *mp;
   1239 	int optval = 0;
   1240 	int error = 0;
   1241 	struct proc *p = curproc;	/* XXX */
   1242 
   1243 	if (level == IPPROTO_IPV6) {
   1244 		switch (op) {
   1245 		case PRCO_SETOPT:
   1246 			switch (optname) {
   1247 			case IPV6_PKTOPTIONS:
   1248 				/* m is freed in ip6_pcbopts */
   1249 				return(ip6_pcbopts(&in6p->in6p_outputopts,
   1250 				    m, so));
   1251 			case IPV6_HOPOPTS:
   1252 			case IPV6_DSTOPTS:
   1253 				if (p == 0 || suser(p->p_ucred, &p->p_acflag)) {
   1254 					error = EPERM;
   1255 					break;
   1256 				}
   1257 				/* FALLTHROUGH */
   1258 			case IPV6_UNICAST_HOPS:
   1259 			case IPV6_RECVOPTS:
   1260 			case IPV6_RECVRETOPTS:
   1261 			case IPV6_RECVDSTADDR:
   1262 			case IPV6_PKTINFO:
   1263 			case IPV6_HOPLIMIT:
   1264 			case IPV6_RTHDR:
   1265 			case IPV6_FAITH:
   1266 			case IPV6_V6ONLY:
   1267 				if (!m || m->m_len != sizeof(int)) {
   1268 					error = EINVAL;
   1269 					break;
   1270 				}
   1271 				optval = *mtod(m, int *);
   1272 				switch (optname) {
   1273 
   1274 				case IPV6_UNICAST_HOPS:
   1275 					if (optval < -1 || optval >= 256)
   1276 						error = EINVAL;
   1277 					else {
   1278 						/* -1 = kernel default */
   1279 						in6p->in6p_hops = optval;
   1280 					}
   1281 					break;
   1282 #define OPTSET(bit) \
   1283 do { \
   1284 	if (optval) \
   1285 		in6p->in6p_flags |= (bit); \
   1286 	else \
   1287 		in6p->in6p_flags &= ~(bit); \
   1288 } while (0)
   1289 
   1290 				case IPV6_RECVOPTS:
   1291 					OPTSET(IN6P_RECVOPTS);
   1292 					break;
   1293 
   1294 				case IPV6_RECVRETOPTS:
   1295 					OPTSET(IN6P_RECVRETOPTS);
   1296 					break;
   1297 
   1298 				case IPV6_RECVDSTADDR:
   1299 					OPTSET(IN6P_RECVDSTADDR);
   1300 					break;
   1301 
   1302 				case IPV6_PKTINFO:
   1303 					OPTSET(IN6P_PKTINFO);
   1304 					break;
   1305 
   1306 				case IPV6_HOPLIMIT:
   1307 					OPTSET(IN6P_HOPLIMIT);
   1308 					break;
   1309 
   1310 				case IPV6_HOPOPTS:
   1311 					OPTSET(IN6P_HOPOPTS);
   1312 					break;
   1313 
   1314 				case IPV6_DSTOPTS:
   1315 					OPTSET(IN6P_DSTOPTS);
   1316 					break;
   1317 
   1318 				case IPV6_RTHDR:
   1319 					OPTSET(IN6P_RTHDR);
   1320 					break;
   1321 
   1322 				case IPV6_FAITH:
   1323 					OPTSET(IN6P_FAITH);
   1324 					break;
   1325 
   1326 				case IPV6_V6ONLY:
   1327 					/*
   1328 					 * make setsockopt(IPV6_V6ONLY)
   1329 					 * available only prior to bind(2).
   1330 					 * see ipng mailing list, Jun 22 2001.
   1331 					 */
   1332 					if (in6p->in6p_lport ||
   1333 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
   1334 						error = EINVAL;
   1335 						break;
   1336 					}
   1337 #ifdef INET6_BINDV6ONLY
   1338 					if (!optval)
   1339 						error = EINVAL;
   1340 #else
   1341 					OPTSET(IN6P_IPV6_V6ONLY);
   1342 #endif
   1343 					break;
   1344 				}
   1345 				break;
   1346 #undef OPTSET
   1347 
   1348 			case IPV6_MULTICAST_IF:
   1349 			case IPV6_MULTICAST_HOPS:
   1350 			case IPV6_MULTICAST_LOOP:
   1351 			case IPV6_JOIN_GROUP:
   1352 			case IPV6_LEAVE_GROUP:
   1353 				error =	ip6_setmoptions(optname,
   1354 				    &in6p->in6p_moptions, m);
   1355 				break;
   1356 
   1357 			case IPV6_PORTRANGE:
   1358 				optval = *mtod(m, int *);
   1359 
   1360 				switch (optval) {
   1361 				case IPV6_PORTRANGE_DEFAULT:
   1362 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
   1363 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
   1364 					break;
   1365 
   1366 				case IPV6_PORTRANGE_HIGH:
   1367 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
   1368 					in6p->in6p_flags |= IN6P_HIGHPORT;
   1369 					break;
   1370 
   1371 				case IPV6_PORTRANGE_LOW:
   1372 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
   1373 					in6p->in6p_flags |= IN6P_LOWPORT;
   1374 					break;
   1375 
   1376 				default:
   1377 					error = EINVAL;
   1378 					break;
   1379 				}
   1380 				break;
   1381 
   1382 #ifdef IPSEC
   1383 			case IPV6_IPSEC_POLICY:
   1384 			    {
   1385 				caddr_t req = NULL;
   1386 				size_t len = 0;
   1387 
   1388 				int priv = 0;
   1389 				if (p == 0 || suser(p->p_ucred, &p->p_acflag))
   1390 					priv = 0;
   1391 				else
   1392 					priv = 1;
   1393 				if (m) {
   1394 					req = mtod(m, caddr_t);
   1395 					len = m->m_len;
   1396 				}
   1397 				error = ipsec6_set_policy(in6p,
   1398 				                   optname, req, len, priv);
   1399 			    }
   1400 				break;
   1401 #endif /* IPSEC */
   1402 
   1403 			default:
   1404 				error = ENOPROTOOPT;
   1405 				break;
   1406 			}
   1407 			if (m)
   1408 				(void)m_free(m);
   1409 			break;
   1410 
   1411 		case PRCO_GETOPT:
   1412 			switch (optname) {
   1413 
   1414 			case IPV6_OPTIONS:
   1415 			case IPV6_RETOPTS:
   1416 				error = ENOPROTOOPT;
   1417 				break;
   1418 
   1419 			case IPV6_PKTOPTIONS:
   1420 				if (in6p->in6p_options) {
   1421 					*mp = m_copym(in6p->in6p_options, 0,
   1422 					    M_COPYALL, M_WAIT);
   1423 				} else {
   1424 					*mp = m_get(M_WAIT, MT_SOOPTS);
   1425 					(*mp)->m_len = 0;
   1426 				}
   1427 				break;
   1428 
   1429 			case IPV6_HOPOPTS:
   1430 			case IPV6_DSTOPTS:
   1431 				if (p == 0 || suser(p->p_ucred, &p->p_acflag)) {
   1432 					error = EPERM;
   1433 					break;
   1434 				}
   1435 				/* FALLTHROUGH */
   1436 			case IPV6_UNICAST_HOPS:
   1437 			case IPV6_RECVOPTS:
   1438 			case IPV6_RECVRETOPTS:
   1439 			case IPV6_RECVDSTADDR:
   1440 			case IPV6_PORTRANGE:
   1441 			case IPV6_PKTINFO:
   1442 			case IPV6_HOPLIMIT:
   1443 			case IPV6_RTHDR:
   1444 			case IPV6_FAITH:
   1445 			case IPV6_V6ONLY:
   1446 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
   1447 				m->m_len = sizeof(int);
   1448 				switch (optname) {
   1449 
   1450 				case IPV6_UNICAST_HOPS:
   1451 					optval = in6p->in6p_hops;
   1452 					break;
   1453 
   1454 #define OPTBIT(bit) (in6p->in6p_flags & bit ? 1 : 0)
   1455 
   1456 				case IPV6_RECVOPTS:
   1457 					optval = OPTBIT(IN6P_RECVOPTS);
   1458 					break;
   1459 
   1460 				case IPV6_RECVRETOPTS:
   1461 					optval = OPTBIT(IN6P_RECVRETOPTS);
   1462 					break;
   1463 
   1464 				case IPV6_RECVDSTADDR:
   1465 					optval = OPTBIT(IN6P_RECVDSTADDR);
   1466 					break;
   1467 
   1468 				case IPV6_PORTRANGE:
   1469 				    {
   1470 					int flags;
   1471 					flags = in6p->in6p_flags;
   1472 					if (flags & IN6P_HIGHPORT)
   1473 						optval = IPV6_PORTRANGE_HIGH;
   1474 					else if (flags & IN6P_LOWPORT)
   1475 						optval = IPV6_PORTRANGE_LOW;
   1476 					else
   1477 						optval = 0;
   1478 					break;
   1479 				    }
   1480 
   1481 				case IPV6_PKTINFO:
   1482 					optval = OPTBIT(IN6P_PKTINFO);
   1483 					break;
   1484 
   1485 				case IPV6_HOPLIMIT:
   1486 					optval = OPTBIT(IN6P_HOPLIMIT);
   1487 					break;
   1488 
   1489 				case IPV6_HOPOPTS:
   1490 					optval = OPTBIT(IN6P_HOPOPTS);
   1491 					break;
   1492 
   1493 				case IPV6_DSTOPTS:
   1494 					optval = OPTBIT(IN6P_DSTOPTS);
   1495 					break;
   1496 
   1497 				case IPV6_RTHDR:
   1498 					optval = OPTBIT(IN6P_RTHDR);
   1499 					break;
   1500 
   1501 				case IPV6_FAITH:
   1502 					optval = OPTBIT(IN6P_FAITH);
   1503 					break;
   1504 
   1505 				case IPV6_V6ONLY:
   1506 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
   1507 					break;
   1508 				}
   1509 				*mtod(m, int *) = optval;
   1510 				break;
   1511 
   1512 			case IPV6_MULTICAST_IF:
   1513 			case IPV6_MULTICAST_HOPS:
   1514 			case IPV6_MULTICAST_LOOP:
   1515 			case IPV6_JOIN_GROUP:
   1516 			case IPV6_LEAVE_GROUP:
   1517 				error = ip6_getmoptions(optname, in6p->in6p_moptions, mp);
   1518 				break;
   1519 
   1520 #ifdef IPSEC
   1521 			case IPV6_IPSEC_POLICY:
   1522 			{
   1523 				caddr_t req = NULL;
   1524 				size_t len = 0;
   1525 
   1526 				if (m) {
   1527 					req = mtod(m, caddr_t);
   1528 					len = m->m_len;
   1529 				}
   1530 				error = ipsec6_get_policy(in6p, req, len, mp);
   1531 				break;
   1532 			}
   1533 #endif /* IPSEC */
   1534 
   1535 			default:
   1536 				error = ENOPROTOOPT;
   1537 				break;
   1538 			}
   1539 			break;
   1540 		}
   1541 	} else {
   1542 		error = EINVAL;
   1543 		if (op == PRCO_SETOPT && *mp)
   1544 			(void)m_free(*mp);
   1545 	}
   1546 	return(error);
   1547 }
   1548 
   1549 /*
   1550  * Set up IP6 options in pcb for insertion in output packets.
   1551  * Store in mbuf with pointer in pcbopt, adding pseudo-option
   1552  * with destination address if source routed.
   1553  */
   1554 static int
   1555 ip6_pcbopts(pktopt, m, so)
   1556 	struct ip6_pktopts **pktopt;
   1557 	struct mbuf *m;
   1558 	struct socket *so;
   1559 {
   1560 	struct ip6_pktopts *opt = *pktopt;
   1561 	int error = 0;
   1562 	struct proc *p = curproc;	/* XXX */
   1563 	int priv = 0;
   1564 
   1565 	/* turn off any old options. */
   1566 	if (opt) {
   1567 		if (opt->ip6po_m)
   1568 			(void)m_free(opt->ip6po_m);
   1569 	} else
   1570 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
   1571 	*pktopt = 0;
   1572 
   1573 	if (!m || m->m_len == 0) {
   1574 		/*
   1575 		 * Only turning off any previous options.
   1576 		 */
   1577 		if (opt)
   1578 			free(opt, M_IP6OPT);
   1579 		if (m)
   1580 			(void)m_free(m);
   1581 		return(0);
   1582 	}
   1583 
   1584 	/*  set options specified by user. */
   1585 	if (p && !suser(p->p_ucred, &p->p_acflag))
   1586 		priv = 1;
   1587 	if ((error = ip6_setpktoptions(m, opt, priv)) != 0) {
   1588 		(void)m_free(m);
   1589 		return(error);
   1590 	}
   1591 	*pktopt = opt;
   1592 	return(0);
   1593 }
   1594 
   1595 /*
   1596  * Set the IP6 multicast options in response to user setsockopt().
   1597  */
   1598 static int
   1599 ip6_setmoptions(optname, im6op, m)
   1600 	int optname;
   1601 	struct ip6_moptions **im6op;
   1602 	struct mbuf *m;
   1603 {
   1604 	int error = 0;
   1605 	u_int loop, ifindex;
   1606 	struct ipv6_mreq *mreq;
   1607 	struct ifnet *ifp;
   1608 	struct ip6_moptions *im6o = *im6op;
   1609 	struct route_in6 ro;
   1610 	struct sockaddr_in6 *dst;
   1611 	struct in6_multi_mship *imm;
   1612 	struct proc *p = curproc;	/* XXX */
   1613 
   1614 	if (im6o == NULL) {
   1615 		/*
   1616 		 * No multicast option buffer attached to the pcb;
   1617 		 * allocate one and initialize to default values.
   1618 		 */
   1619 		im6o = (struct ip6_moptions *)
   1620 			malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
   1621 
   1622 		if (im6o == NULL)
   1623 			return(ENOBUFS);
   1624 		*im6op = im6o;
   1625 		im6o->im6o_multicast_ifp = NULL;
   1626 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
   1627 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
   1628 		LIST_INIT(&im6o->im6o_memberships);
   1629 	}
   1630 
   1631 	switch (optname) {
   1632 
   1633 	case IPV6_MULTICAST_IF:
   1634 		/*
   1635 		 * Select the interface for outgoing multicast packets.
   1636 		 */
   1637 		if (m == NULL || m->m_len != sizeof(u_int)) {
   1638 			error = EINVAL;
   1639 			break;
   1640 		}
   1641 		bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
   1642 		if (ifindex < 0 || if_index < ifindex) {
   1643 			error = ENXIO;	/* XXX EINVAL? */
   1644 			break;
   1645 		}
   1646 		ifp = ifindex2ifnet[ifindex];
   1647 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
   1648 			error = EADDRNOTAVAIL;
   1649 			break;
   1650 		}
   1651 		im6o->im6o_multicast_ifp = ifp;
   1652 		break;
   1653 
   1654 	case IPV6_MULTICAST_HOPS:
   1655 	    {
   1656 		/*
   1657 		 * Set the IP6 hoplimit for outgoing multicast packets.
   1658 		 */
   1659 		int optval;
   1660 		if (m == NULL || m->m_len != sizeof(int)) {
   1661 			error = EINVAL;
   1662 			break;
   1663 		}
   1664 		bcopy(mtod(m, u_int *), &optval, sizeof(optval));
   1665 		if (optval < -1 || optval >= 256)
   1666 			error = EINVAL;
   1667 		else if (optval == -1)
   1668 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
   1669 		else
   1670 			im6o->im6o_multicast_hlim = optval;
   1671 		break;
   1672 	    }
   1673 
   1674 	case IPV6_MULTICAST_LOOP:
   1675 		/*
   1676 		 * Set the loopback flag for outgoing multicast packets.
   1677 		 * Must be zero or one.
   1678 		 */
   1679 		if (m == NULL || m->m_len != sizeof(u_int)) {
   1680 			error = EINVAL;
   1681 			break;
   1682 		}
   1683 		bcopy(mtod(m, u_int *), &loop, sizeof(loop));
   1684 		if (loop > 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 			{
   1709 				error = EACCES;
   1710 				break;
   1711 			}
   1712 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
   1713 			error = EINVAL;
   1714 			break;
   1715 		}
   1716 
   1717 		/*
   1718 		 * If the interface is specified, validate it.
   1719 		 */
   1720 		if (mreq->ipv6mr_interface < 0
   1721 		 || if_index < mreq->ipv6mr_interface) {
   1722 			error = ENXIO;	/* XXX EINVAL? */
   1723 			break;
   1724 		}
   1725 		/*
   1726 		 * If no interface was explicitly specified, choose an
   1727 		 * appropriate one according to the given multicast address.
   1728 		 */
   1729 		if (mreq->ipv6mr_interface == 0) {
   1730 			/*
   1731 			 * If the multicast address is in node-local scope,
   1732 			 * the interface should be a loopback interface.
   1733 			 * Otherwise, look up the routing table for the
   1734 			 * address, and choose the outgoing interface.
   1735 			 *   XXX: is it a good approach?
   1736 			 */
   1737 			if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
   1738 				ifp = &loif[0];
   1739 			} else {
   1740 				ro.ro_rt = NULL;
   1741 				dst = (struct sockaddr_in6 *)&ro.ro_dst;
   1742 				bzero(dst, sizeof(*dst));
   1743 				dst->sin6_len = sizeof(struct sockaddr_in6);
   1744 				dst->sin6_family = AF_INET6;
   1745 				dst->sin6_addr = mreq->ipv6mr_multiaddr;
   1746 				rtalloc((struct route *)&ro);
   1747 				if (ro.ro_rt == NULL) {
   1748 					error = EADDRNOTAVAIL;
   1749 					break;
   1750 				}
   1751 				ifp = ro.ro_rt->rt_ifp;
   1752 				rtfree(ro.ro_rt);
   1753 			}
   1754 		} else
   1755 			ifp = ifindex2ifnet[mreq->ipv6mr_interface];
   1756 
   1757 		/*
   1758 		 * See if we found an interface, and confirm that it
   1759 		 * supports multicast
   1760 		 */
   1761 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
   1762 			error = EADDRNOTAVAIL;
   1763 			break;
   1764 		}
   1765 		/*
   1766 		 * Put interface index into the multicast address,
   1767 		 * if the address has link-local scope.
   1768 		 */
   1769 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
   1770 			mreq->ipv6mr_multiaddr.s6_addr16[1] =
   1771 			    htons(mreq->ipv6mr_interface);
   1772 		}
   1773 		/*
   1774 		 * See if the membership already exists.
   1775 		 */
   1776 		for (imm = im6o->im6o_memberships.lh_first;
   1777 		     imm != NULL; imm = imm->i6mm_chain.le_next)
   1778 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
   1779 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
   1780 			    &mreq->ipv6mr_multiaddr))
   1781 				break;
   1782 		if (imm != NULL) {
   1783 			error = EADDRINUSE;
   1784 			break;
   1785 		}
   1786 		/*
   1787 		 * Everything looks good; add a new record to the multicast
   1788 		 * address list for the given interface.
   1789 		 */
   1790 		imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error);
   1791 		if (!imm)
   1792 			break;
   1793 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
   1794 		break;
   1795 
   1796 	case IPV6_LEAVE_GROUP:
   1797 		/*
   1798 		 * Drop a multicast group membership.
   1799 		 * Group must be a valid IP6 multicast address.
   1800 		 */
   1801 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
   1802 			error = EINVAL;
   1803 			break;
   1804 		}
   1805 		mreq = mtod(m, struct ipv6_mreq *);
   1806 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
   1807 			if (suser(p->p_ucred, &p->p_acflag))
   1808 			{
   1809 				error = EACCES;
   1810 				break;
   1811 			}
   1812 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
   1813 			error = EINVAL;
   1814 			break;
   1815 		}
   1816 		/*
   1817 		 * If an interface address was specified, get a pointer
   1818 		 * to its ifnet structure.
   1819 		 */
   1820 		if (mreq->ipv6mr_interface < 0
   1821 		 || if_index < mreq->ipv6mr_interface) {
   1822 			error = ENXIO;	/* XXX EINVAL? */
   1823 			break;
   1824 		}
   1825 		ifp = ifindex2ifnet[mreq->ipv6mr_interface];
   1826 		/*
   1827 		 * Put interface index into the multicast address,
   1828 		 * if the address has link-local scope.
   1829 		 */
   1830 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
   1831 			mreq->ipv6mr_multiaddr.s6_addr16[1] =
   1832 			    htons(mreq->ipv6mr_interface);
   1833 		}
   1834 		/*
   1835 		 * Find the membership in the membership list.
   1836 		 */
   1837 		for (imm = im6o->im6o_memberships.lh_first;
   1838 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
   1839 			if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
   1840 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
   1841 			    &mreq->ipv6mr_multiaddr))
   1842 				break;
   1843 		}
   1844 		if (imm == NULL) {
   1845 			/* Unable to resolve interface */
   1846 			error = EADDRNOTAVAIL;
   1847 			break;
   1848 		}
   1849 		/*
   1850 		 * Give up the multicast address record to which the
   1851 		 * membership points.
   1852 		 */
   1853 		LIST_REMOVE(imm, i6mm_chain);
   1854 		in6_leavegroup(imm);
   1855 		break;
   1856 
   1857 	default:
   1858 		error = EOPNOTSUPP;
   1859 		break;
   1860 	}
   1861 
   1862 	/*
   1863 	 * If all options have default values, no need to keep the mbuf.
   1864 	 */
   1865 	if (im6o->im6o_multicast_ifp == NULL &&
   1866 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
   1867 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
   1868 	    im6o->im6o_memberships.lh_first == NULL) {
   1869 		free(*im6op, M_IPMOPTS);
   1870 		*im6op = NULL;
   1871 	}
   1872 
   1873 	return(error);
   1874 }
   1875 
   1876 /*
   1877  * Return the IP6 multicast options in response to user getsockopt().
   1878  */
   1879 static int
   1880 ip6_getmoptions(optname, im6o, mp)
   1881 	int optname;
   1882 	struct ip6_moptions *im6o;
   1883 	struct mbuf **mp;
   1884 {
   1885 	u_int *hlim, *loop, *ifindex;
   1886 
   1887 	*mp = m_get(M_WAIT, MT_SOOPTS);
   1888 
   1889 	switch (optname) {
   1890 
   1891 	case IPV6_MULTICAST_IF:
   1892 		ifindex = mtod(*mp, u_int *);
   1893 		(*mp)->m_len = sizeof(u_int);
   1894 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
   1895 			*ifindex = 0;
   1896 		else
   1897 			*ifindex = im6o->im6o_multicast_ifp->if_index;
   1898 		return(0);
   1899 
   1900 	case IPV6_MULTICAST_HOPS:
   1901 		hlim = mtod(*mp, u_int *);
   1902 		(*mp)->m_len = sizeof(u_int);
   1903 		if (im6o == NULL)
   1904 			*hlim = ip6_defmcasthlim;
   1905 		else
   1906 			*hlim = im6o->im6o_multicast_hlim;
   1907 		return(0);
   1908 
   1909 	case IPV6_MULTICAST_LOOP:
   1910 		loop = mtod(*mp, u_int *);
   1911 		(*mp)->m_len = sizeof(u_int);
   1912 		if (im6o == NULL)
   1913 			*loop = ip6_defmcasthlim;
   1914 		else
   1915 			*loop = im6o->im6o_multicast_loop;
   1916 		return(0);
   1917 
   1918 	default:
   1919 		return(EOPNOTSUPP);
   1920 	}
   1921 }
   1922 
   1923 /*
   1924  * Discard the IP6 multicast options.
   1925  */
   1926 void
   1927 ip6_freemoptions(im6o)
   1928 	struct ip6_moptions *im6o;
   1929 {
   1930 	struct in6_multi_mship *imm;
   1931 
   1932 	if (im6o == NULL)
   1933 		return;
   1934 
   1935 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
   1936 		LIST_REMOVE(imm, i6mm_chain);
   1937 		in6_leavegroup(imm);
   1938 	}
   1939 	free(im6o, M_IPMOPTS);
   1940 }
   1941 
   1942 /*
   1943  * Set IPv6 outgoing packet options based on advanced API.
   1944  */
   1945 int
   1946 ip6_setpktoptions(control, opt, priv)
   1947 	struct mbuf *control;
   1948 	struct ip6_pktopts *opt;
   1949 	int priv;
   1950 {
   1951 	struct cmsghdr *cm = 0;
   1952 
   1953 	if (control == 0 || opt == 0)
   1954 		return(EINVAL);
   1955 
   1956 	bzero(opt, sizeof(*opt));
   1957 	opt->ip6po_hlim = -1; /* -1 means to use default hop limit */
   1958 
   1959 	/*
   1960 	 * XXX: Currently, we assume all the optional information is stored
   1961 	 * in a single mbuf.
   1962 	 */
   1963 	if (control->m_next)
   1964 		return(EINVAL);
   1965 
   1966 	opt->ip6po_m = control;
   1967 
   1968 	for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
   1969 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
   1970 		cm = mtod(control, struct cmsghdr *);
   1971 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
   1972 			return(EINVAL);
   1973 		if (cm->cmsg_level != IPPROTO_IPV6)
   1974 			continue;
   1975 
   1976 		switch (cm->cmsg_type) {
   1977 		case IPV6_PKTINFO:
   1978 			if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo)))
   1979 				return(EINVAL);
   1980 			opt->ip6po_pktinfo = (struct in6_pktinfo *)CMSG_DATA(cm);
   1981 			if (opt->ip6po_pktinfo->ipi6_ifindex &&
   1982 			    IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr))
   1983 				opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] =
   1984 					htons(opt->ip6po_pktinfo->ipi6_ifindex);
   1985 
   1986 			if (opt->ip6po_pktinfo->ipi6_ifindex > if_index ||
   1987 			    opt->ip6po_pktinfo->ipi6_ifindex < 0) {
   1988 				return(ENXIO);
   1989 			}
   1990 
   1991 			/*
   1992 			 * Check if the requested source address is indeed a
   1993 			 * unicast address assigned to the node, and can be
   1994 			 * used as the packet's source address.
   1995 			 */
   1996 			if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
   1997 				struct ifaddr *ia;
   1998 				struct in6_ifaddr *ia6;
   1999 				struct sockaddr_in6 sin6;
   2000 
   2001 				bzero(&sin6, sizeof(sin6));
   2002 				sin6.sin6_len = sizeof(sin6);
   2003 				sin6.sin6_family = AF_INET6;
   2004 				sin6.sin6_addr =
   2005 					opt->ip6po_pktinfo->ipi6_addr;
   2006 				ia = ifa_ifwithaddr(sin6tosa(&sin6));
   2007 				if (ia == NULL ||
   2008 				    (opt->ip6po_pktinfo->ipi6_ifindex &&
   2009 				     (ia->ifa_ifp->if_index !=
   2010 				      opt->ip6po_pktinfo->ipi6_ifindex))) {
   2011 					return(EADDRNOTAVAIL);
   2012 				}
   2013 				ia6 = (struct in6_ifaddr *)ia;
   2014 				if ((ia6->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) != 0) {
   2015 					return(EADDRNOTAVAIL);
   2016 				}
   2017 
   2018 				/*
   2019 				 * Check if the requested source address is
   2020 				 * indeed a unicast address assigned to the
   2021 				 * node.
   2022 				 */
   2023 				if (IN6_IS_ADDR_MULTICAST(&opt->ip6po_pktinfo->ipi6_addr))
   2024 					return(EADDRNOTAVAIL);
   2025 			}
   2026 			break;
   2027 
   2028 		case IPV6_HOPLIMIT:
   2029 			if (cm->cmsg_len != CMSG_LEN(sizeof(int)))
   2030 				return(EINVAL);
   2031 
   2032 			bcopy(CMSG_DATA(cm), &opt->ip6po_hlim,
   2033 			    sizeof(opt->ip6po_hlim));
   2034 			if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255)
   2035 				return(EINVAL);
   2036 			break;
   2037 
   2038 		case IPV6_NEXTHOP:
   2039 			if (!priv)
   2040 				return(EPERM);
   2041 
   2042 			/* check if cmsg_len is large enough for sa_len */
   2043 			if (cm->cmsg_len < sizeof(u_char) ||
   2044 			    cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm)))
   2045 				return(EINVAL);
   2046 
   2047 			opt->ip6po_nexthop = (struct sockaddr *)CMSG_DATA(cm);
   2048 
   2049 			break;
   2050 
   2051 		case IPV6_HOPOPTS:
   2052 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh)))
   2053 				return(EINVAL);
   2054 			opt->ip6po_hbh = (struct ip6_hbh *)CMSG_DATA(cm);
   2055 			if (cm->cmsg_len !=
   2056 			    CMSG_LEN((opt->ip6po_hbh->ip6h_len + 1) << 3))
   2057 				return(EINVAL);
   2058 			break;
   2059 
   2060 		case IPV6_DSTOPTS:
   2061 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest)))
   2062 				return(EINVAL);
   2063 
   2064 			/*
   2065 			 * If there is no routing header yet, the destination
   2066 			 * options header should be put on the 1st part.
   2067 			 * Otherwise, the header should be on the 2nd part.
   2068 			 * (See RFC 2460, section 4.1)
   2069 			 */
   2070 			if (opt->ip6po_rthdr == NULL) {
   2071 				opt->ip6po_dest1 =
   2072 				    (struct ip6_dest *)CMSG_DATA(cm);
   2073 				if (cm->cmsg_len !=
   2074 				    CMSG_LEN((opt->ip6po_dest1->ip6d_len + 1) << 3));
   2075 					return(EINVAL);
   2076 			}
   2077 			else {
   2078 				opt->ip6po_dest2 =
   2079 				    (struct ip6_dest *)CMSG_DATA(cm);
   2080 				if (cm->cmsg_len !=
   2081 				    CMSG_LEN((opt->ip6po_dest2->ip6d_len + 1) << 3))
   2082 					return(EINVAL);
   2083 			}
   2084 			break;
   2085 
   2086 		case IPV6_RTHDR:
   2087 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr)))
   2088 				return(EINVAL);
   2089 			opt->ip6po_rthdr = (struct ip6_rthdr *)CMSG_DATA(cm);
   2090 			if (cm->cmsg_len !=
   2091 			    CMSG_LEN((opt->ip6po_rthdr->ip6r_len + 1) << 3))
   2092 				return(EINVAL);
   2093 			switch (opt->ip6po_rthdr->ip6r_type) {
   2094 			case IPV6_RTHDR_TYPE_0:
   2095 				if (opt->ip6po_rthdr->ip6r_segleft == 0)
   2096 					return(EINVAL);
   2097 				break;
   2098 			default:
   2099 				return(EINVAL);
   2100 			}
   2101 			break;
   2102 
   2103 		default:
   2104 			return(ENOPROTOOPT);
   2105 		}
   2106 	}
   2107 
   2108 	return(0);
   2109 }
   2110 
   2111 /*
   2112  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
   2113  * packet to the input queue of a specified interface.  Note that this
   2114  * calls the output routine of the loopback "driver", but with an interface
   2115  * pointer that might NOT be &loif -- easier than replicating that code here.
   2116  */
   2117 void
   2118 ip6_mloopback(ifp, m, dst)
   2119 	struct ifnet *ifp;
   2120 	struct mbuf *m;
   2121 	struct sockaddr_in6 *dst;
   2122 {
   2123 	struct mbuf *copym;
   2124 	struct ip6_hdr *ip6;
   2125 
   2126 	copym = m_copy(m, 0, M_COPYALL);
   2127 	if (copym == NULL)
   2128 		return;
   2129 
   2130 	/*
   2131 	 * Make sure to deep-copy IPv6 header portion in case the data
   2132 	 * is in an mbuf cluster, so that we can safely override the IPv6
   2133 	 * header portion later.
   2134 	 */
   2135 	if ((copym->m_flags & M_EXT) != 0 ||
   2136 	    copym->m_len < sizeof(struct ip6_hdr)) {
   2137 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
   2138 		if (copym == NULL)
   2139 			return;
   2140 	}
   2141 
   2142 #ifdef DIAGNOSTIC
   2143 	if (copym->m_len < sizeof(*ip6)) {
   2144 		m_freem(copym);
   2145 		return;
   2146 	}
   2147 #endif
   2148 
   2149 	ip6 = mtod(copym, struct ip6_hdr *);
   2150 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
   2151 		ip6->ip6_src.s6_addr16[1] = 0;
   2152 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
   2153 		ip6->ip6_dst.s6_addr16[1] = 0;
   2154 
   2155 	(void)looutput(ifp, copym, (struct sockaddr *)dst, NULL);
   2156 }
   2157 
   2158 /*
   2159  * Chop IPv6 header off from the payload.
   2160  */
   2161 static int
   2162 ip6_splithdr(m, exthdrs)
   2163 	struct mbuf *m;
   2164 	struct ip6_exthdrs *exthdrs;
   2165 {
   2166 	struct mbuf *mh;
   2167 	struct ip6_hdr *ip6;
   2168 
   2169 	ip6 = mtod(m, struct ip6_hdr *);
   2170 	if (m->m_len > sizeof(*ip6)) {
   2171 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
   2172 		if (mh == 0) {
   2173 			m_freem(m);
   2174 			return ENOBUFS;
   2175 		}
   2176 		M_COPY_PKTHDR(mh, m);
   2177 		MH_ALIGN(mh, sizeof(*ip6));
   2178 		m->m_flags &= ~M_PKTHDR;
   2179 		m->m_len -= sizeof(*ip6);
   2180 		m->m_data += sizeof(*ip6);
   2181 		mh->m_next = m;
   2182 		m = mh;
   2183 		m->m_len = sizeof(*ip6);
   2184 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
   2185 	}
   2186 	exthdrs->ip6e_ip6 = m;
   2187 	return 0;
   2188 }
   2189 
   2190 /*
   2191  * Compute IPv6 extension header length.
   2192  */
   2193 int
   2194 ip6_optlen(in6p)
   2195 	struct in6pcb *in6p;
   2196 {
   2197 	int len;
   2198 
   2199 	if (!in6p->in6p_outputopts)
   2200 		return 0;
   2201 
   2202 	len = 0;
   2203 #define elen(x) \
   2204     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
   2205 
   2206 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
   2207 	len += elen(in6p->in6p_outputopts->ip6po_dest1);
   2208 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
   2209 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
   2210 	return len;
   2211 #undef elen
   2212 }
   2213