Home | History | Annotate | Line # | Download | only in netinet6
ip6_output.c revision 1.23.2.5
      1 /*	$NetBSD: ip6_output.c,v 1.23.2.5 2001/04/22 18:11:17 he 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 
     96 #ifdef IPSEC
     97 #include <netinet6/ipsec.h>
     98 #include <netkey/key.h>
     99 #include <netkey/key_debug.h>
    100 #endif /* IPSEC */
    101 
    102 #include "loop.h"
    103 
    104 #include <net/net_osdep.h>
    105 
    106 #ifdef IPV6FIREWALL
    107 #include <netinet6/ip6_fw.h>
    108 #endif
    109 
    110 struct ip6_exthdrs {
    111 	struct mbuf *ip6e_ip6;
    112 	struct mbuf *ip6e_hbh;
    113 	struct mbuf *ip6e_dest1;
    114 	struct mbuf *ip6e_rthdr;
    115 	struct mbuf *ip6e_dest2;
    116 };
    117 
    118 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
    119 			    struct socket *));
    120 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *));
    121 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **));
    122 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int));
    123 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
    124 				  struct ip6_frag **));
    125 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t));
    126 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *));
    127 
    128 extern struct ifnet loif[NLOOP];
    129 
    130 /*
    131  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
    132  * header (with pri, len, nxt, hlim, src, dst).
    133  * This function may modify ver and hlim only.
    134  * The mbuf chain containing the packet will be freed.
    135  * The mbuf opt, if present, will not be freed.
    136  */
    137 int
    138 ip6_output(m0, opt, ro, flags, im6o, ifpp)
    139 	struct mbuf *m0;
    140 	struct ip6_pktopts *opt;
    141 	struct route_in6 *ro;
    142 	int flags;
    143 	struct ip6_moptions *im6o;
    144 	struct ifnet **ifpp;		/* XXX: just for statistics */
    145 {
    146 	struct ip6_hdr *ip6, *mhip6;
    147 	struct ifnet *ifp, *origifp;
    148 	struct mbuf *m = m0;
    149 	int hlen, tlen, len, off;
    150 	struct route_in6 ip6route;
    151 	struct sockaddr_in6 *dst;
    152 	int error = 0;
    153 	struct in6_ifaddr *ia;
    154 	u_long mtu;
    155 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
    156 	struct ip6_exthdrs exthdrs;
    157 	struct in6_addr finaldst;
    158 	struct route_in6 *ro_pmtu = NULL;
    159 	int hdrsplit = 0;
    160 	int needipsec = 0;
    161 #ifdef PFIL_HOOKS
    162 	struct packet_filter_hook *pfh;
    163 	struct mbuf *m1;
    164 	int rv;
    165 #endif /* PFIL_HOOKS */
    166 #ifdef IPSEC
    167 	int needipsectun = 0;
    168 	struct socket *so;
    169 	struct secpolicy *sp = NULL;
    170 
    171 	/* for AH processing. stupid to have "socket" variable in IP layer... */
    172 	so = ipsec_getsocket(m);
    173 	(void)ipsec_setsocket(m, NULL);
    174 	ip6 = mtod(m, struct ip6_hdr *);
    175 #endif /* IPSEC */
    176 
    177 #define MAKE_EXTHDR(hp, mp)						\
    178     do {								\
    179 	if (hp) {							\
    180 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
    181 		error = ip6_copyexthdr((mp), (caddr_t)(hp), 		\
    182 				       ((eh)->ip6e_len + 1) << 3);	\
    183 		if (error)						\
    184 			goto freehdrs;					\
    185 	}								\
    186     } while (0)
    187 
    188 	bzero(&exthdrs, sizeof(exthdrs));
    189 	if (opt) {
    190 		/* Hop-by-Hop options header */
    191 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
    192 		/* Destination options header(1st part) */
    193 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
    194 		/* Routing header */
    195 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
    196 		/* Destination options header(2nd part) */
    197 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
    198 	}
    199 
    200 #ifdef IPSEC
    201 	/* get a security policy for this packet */
    202 	if (so == NULL)
    203 		sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
    204 	else
    205 		sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
    206 
    207 	if (sp == NULL) {
    208 		ipsec6stat.out_inval++;
    209 		goto freehdrs;
    210 	}
    211 
    212 	error = 0;
    213 
    214 	/* check policy */
    215 	switch (sp->policy) {
    216 	case IPSEC_POLICY_DISCARD:
    217 		/*
    218 		 * This packet is just discarded.
    219 		 */
    220 		ipsec6stat.out_polvio++;
    221 		goto freehdrs;
    222 
    223 	case IPSEC_POLICY_BYPASS:
    224 	case IPSEC_POLICY_NONE:
    225 		/* no need to do IPsec. */
    226 		needipsec = 0;
    227 		break;
    228 
    229 	case IPSEC_POLICY_IPSEC:
    230 		if (sp->req == NULL) {
    231 			/* XXX should be panic ? */
    232 			printf("ip6_output: No IPsec request specified.\n");
    233 			error = EINVAL;
    234 			goto freehdrs;
    235 		}
    236 		needipsec = 1;
    237 		break;
    238 
    239 	case IPSEC_POLICY_ENTRUST:
    240 	default:
    241 		printf("ip6_output: Invalid policy found. %d\n", sp->policy);
    242 	}
    243 #endif /* IPSEC */
    244 
    245 	/*
    246 	 * Calculate the total length of the extension header chain.
    247 	 * Keep the length of the unfragmentable part for fragmentation.
    248 	 */
    249 	optlen = 0;
    250 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
    251 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
    252 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
    253 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
    254 	/* NOTE: we don't add AH/ESP length here. do that later. */
    255 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
    256 
    257 	/*
    258 	 * If we need IPsec, or there is at least one extension header,
    259 	 * separate IP6 header from the payload.
    260 	 */
    261 	if ((needipsec || optlen) && !hdrsplit) {
    262 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
    263 			m = NULL;
    264 			goto freehdrs;
    265 		}
    266 		m = exthdrs.ip6e_ip6;
    267 		hdrsplit++;
    268 	}
    269 
    270 	/* adjust pointer */
    271 	ip6 = mtod(m, struct ip6_hdr *);
    272 
    273 	/* adjust mbuf packet header length */
    274 	m->m_pkthdr.len += optlen;
    275 	plen = m->m_pkthdr.len - sizeof(*ip6);
    276 
    277 	/* If this is a jumbo payload, insert a jumbo payload option. */
    278 	if (plen > IPV6_MAXPACKET) {
    279 		if (!hdrsplit) {
    280 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
    281 				m = NULL;
    282 				goto freehdrs;
    283 			}
    284 			m = exthdrs.ip6e_ip6;
    285 			hdrsplit++;
    286 		}
    287 		/* adjust pointer */
    288 		ip6 = mtod(m, struct ip6_hdr *);
    289 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
    290 			goto freehdrs;
    291 		ip6->ip6_plen = 0;
    292 	} else
    293 		ip6->ip6_plen = htons(plen);
    294 
    295 	/*
    296 	 * Concatenate headers and fill in next header fields.
    297 	 * Here we have, on "m"
    298 	 *	IPv6 payload
    299 	 * and we insert headers accordingly.  Finally, we should be getting:
    300 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
    301 	 *
    302 	 * during the header composing process, "m" points to IPv6 header.
    303 	 * "mprev" points to an extension header prior to esp.
    304 	 */
    305 	{
    306 		u_char *nexthdrp = &ip6->ip6_nxt;
    307 		struct mbuf *mprev = m;
    308 
    309 		/*
    310 		 * we treat dest2 specially.  this makes IPsec processing
    311 		 * much easier.
    312 		 *
    313 		 * result: IPv6 dest2 payload
    314 		 * m and mprev will point to IPv6 header.
    315 		 */
    316 		if (exthdrs.ip6e_dest2) {
    317 			if (!hdrsplit)
    318 				panic("assumption failed: hdr not split");
    319 			exthdrs.ip6e_dest2->m_next = m->m_next;
    320 			m->m_next = exthdrs.ip6e_dest2;
    321 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
    322 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
    323 		}
    324 
    325 #define MAKE_CHAIN(m, mp, p, i)\
    326     do {\
    327 	if (m) {\
    328 		if (!hdrsplit) \
    329 			panic("assumption failed: hdr not split"); \
    330 		*mtod((m), u_char *) = *(p);\
    331 		*(p) = (i);\
    332 		p = mtod((m), u_char *);\
    333 		(m)->m_next = (mp)->m_next;\
    334 		(mp)->m_next = (m);\
    335 		(mp) = (m);\
    336 	}\
    337     } while (0)
    338 		/*
    339 		 * result: IPv6 hbh dest1 rthdr dest2 payload
    340 		 * m will point to IPv6 header.  mprev will point to the
    341 		 * extension header prior to dest2 (rthdr in the above case).
    342 		 */
    343 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev,
    344 			   nexthdrp, IPPROTO_HOPOPTS);
    345 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev,
    346 			   nexthdrp, IPPROTO_DSTOPTS);
    347 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev,
    348 			   nexthdrp, IPPROTO_ROUTING);
    349 
    350 #ifdef IPSEC
    351 		if (!needipsec)
    352 			goto skip_ipsec2;
    353 
    354 		/*
    355 		 * pointers after IPsec headers are not valid any more.
    356 		 * other pointers need a great care too.
    357 		 * (IPsec routines should not mangle mbufs prior to AH/ESP)
    358 		 */
    359 		exthdrs.ip6e_dest2 = NULL;
    360 
    361 	    {
    362 		struct ip6_rthdr *rh = NULL;
    363 		int segleft_org = 0;
    364 		struct ipsec_output_state state;
    365 
    366 		if (exthdrs.ip6e_rthdr) {
    367 			rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
    368 			segleft_org = rh->ip6r_segleft;
    369 			rh->ip6r_segleft = 0;
    370 		}
    371 
    372 		bzero(&state, sizeof(state));
    373 		state.m = m;
    374 		error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
    375 			&needipsectun);
    376 		m = state.m;
    377 		if (error) {
    378 			/* mbuf is already reclaimed in ipsec6_output_trans. */
    379 			m = NULL;
    380 			switch (error) {
    381 			case EHOSTUNREACH:
    382 			case ENETUNREACH:
    383 			case EMSGSIZE:
    384 			case ENOBUFS:
    385 			case ENOMEM:
    386 				break;
    387 			default:
    388 				printf("ip6_output (ipsec): error code %d\n", error);
    389 				/*fall through*/
    390 			case ENOENT:
    391 				/* don't show these error codes to the user */
    392 				error = 0;
    393 				break;
    394 			}
    395 			goto bad;
    396 		}
    397 		if (exthdrs.ip6e_rthdr) {
    398 			/* ah6_output doesn't modify mbuf chain */
    399 			rh->ip6r_segleft = segleft_org;
    400 		}
    401 	    }
    402 skip_ipsec2:;
    403 #endif
    404 	}
    405 
    406 	/*
    407 	 * If there is a routing header, replace destination address field
    408 	 * with the first hop of the routing header.
    409 	 */
    410 	if (exthdrs.ip6e_rthdr) {
    411 		struct ip6_rthdr *rh =
    412 			(struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
    413 						  struct ip6_rthdr *));
    414 		struct ip6_rthdr0 *rh0;
    415 
    416 		finaldst = ip6->ip6_dst;
    417 		switch(rh->ip6r_type) {
    418 		case IPV6_RTHDR_TYPE_0:
    419 			 rh0 = (struct ip6_rthdr0 *)rh;
    420 			 ip6->ip6_dst = rh0->ip6r0_addr[0];
    421 			 bcopy((caddr_t)&rh0->ip6r0_addr[1],
    422 				 (caddr_t)&rh0->ip6r0_addr[0],
    423 				 sizeof(struct in6_addr)*(rh0->ip6r0_segleft - 1)
    424 				 );
    425 			 rh0->ip6r0_addr[rh0->ip6r0_segleft - 1] = finaldst;
    426 			 break;
    427 		default:	/* is it possible? */
    428 			 error = EINVAL;
    429 			 goto bad;
    430 		}
    431 	}
    432 
    433 	/* Source address validation */
    434 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
    435 	    (flags & IPV6_DADOUTPUT) == 0) {
    436 		error = EOPNOTSUPP;
    437 		ip6stat.ip6s_badscope++;
    438 		goto bad;
    439 	}
    440 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
    441 		error = EOPNOTSUPP;
    442 		ip6stat.ip6s_badscope++;
    443 		goto bad;
    444 	}
    445 
    446 	ip6stat.ip6s_localout++;
    447 
    448 	/*
    449 	 * Route packet.
    450 	 */
    451 	if (ro == 0) {
    452 		ro = &ip6route;
    453 		bzero((caddr_t)ro, sizeof(*ro));
    454 	}
    455 	ro_pmtu = ro;
    456 	if (opt && opt->ip6po_rthdr)
    457 		ro = &opt->ip6po_route;
    458 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
    459 	/*
    460 	 * If there is a cached route,
    461 	 * check that it is to the same destination
    462 	 * and is still up. If not, free it and try again.
    463 	 */
    464 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
    465 			 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
    466 		RTFREE(ro->ro_rt);
    467 		ro->ro_rt = (struct rtentry *)0;
    468 	}
    469 	if (ro->ro_rt == 0) {
    470 		bzero(dst, sizeof(*dst));
    471 		dst->sin6_family = AF_INET6;
    472 		dst->sin6_len = sizeof(struct sockaddr_in6);
    473 		dst->sin6_addr = ip6->ip6_dst;
    474 	}
    475 #ifdef IPSEC
    476 	if (needipsec && needipsectun) {
    477 		struct ipsec_output_state state;
    478 
    479 		/*
    480 		 * All the extension headers will become inaccessible
    481 		 * (since they can be encrypted).
    482 		 * Don't panic, we need no more updates to extension headers
    483 		 * on inner IPv6 packet (since they are now encapsulated).
    484 		 *
    485 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
    486 		 */
    487 		bzero(&exthdrs, sizeof(exthdrs));
    488 		exthdrs.ip6e_ip6 = m;
    489 
    490 		bzero(&state, sizeof(state));
    491 		state.m = m;
    492 		state.ro = (struct route *)ro;
    493 		state.dst = (struct sockaddr *)dst;
    494 
    495 		error = ipsec6_output_tunnel(&state, sp, flags);
    496 
    497 		m = state.m;
    498 		ro = (struct route_in6 *)state.ro;
    499 		dst = (struct sockaddr_in6 *)state.dst;
    500 		if (error) {
    501 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
    502 			m0 = m = NULL;
    503 			m = NULL;
    504 			switch (error) {
    505 			case EHOSTUNREACH:
    506 			case ENETUNREACH:
    507 			case EMSGSIZE:
    508 			case ENOBUFS:
    509 			case ENOMEM:
    510 				break;
    511 			default:
    512 				printf("ip6_output (ipsec): error code %d\n", error);
    513 				/*fall through*/
    514 			case ENOENT:
    515 				/* don't show these error codes to the user */
    516 				error = 0;
    517 				break;
    518 			}
    519 			goto bad;
    520 		}
    521 
    522 		exthdrs.ip6e_ip6 = m;
    523 	}
    524 #endif /*IPSEC*/
    525 
    526 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
    527 		/* Unicast */
    528 
    529 #define ifatoia6(ifa)	((struct in6_ifaddr *)(ifa))
    530 #define sin6tosa(sin6)	((struct sockaddr *)(sin6))
    531 		/* xxx
    532 		 * interface selection comes here
    533 		 * if an interface is specified from an upper layer,
    534 		 * ifp must point it.
    535 		 */
    536 		if (ro->ro_rt == 0) {
    537 			/*
    538 			 * non-bsdi always clone routes, if parent is
    539 			 * PRF_CLONING.
    540 			 */
    541 			rtalloc((struct route *)ro);
    542 		}
    543 		if (ro->ro_rt == 0) {
    544 			ip6stat.ip6s_noroute++;
    545 			error = EHOSTUNREACH;
    546 			/* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
    547 			goto bad;
    548 		}
    549 		ia = ifatoia6(ro->ro_rt->rt_ifa);
    550 		ifp = ro->ro_rt->rt_ifp;
    551 		ro->ro_rt->rt_use++;
    552 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
    553 			dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
    554 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
    555 
    556 		in6_ifstat_inc(ifp, ifs6_out_request);
    557 
    558 		/*
    559 		 * Check if the outgoing interface conflicts with
    560 		 * the interface specified by ifi6_ifindex (if specified).
    561 		 * Note that loopback interface is always okay.
    562 		 * (this may happen when we are sending a packet to one of
    563 		 *  our own addresses.)
    564 		 */
    565 		if (opt && opt->ip6po_pktinfo
    566 		 && opt->ip6po_pktinfo->ipi6_ifindex) {
    567 			if (!(ifp->if_flags & IFF_LOOPBACK)
    568 			 && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
    569 				ip6stat.ip6s_noroute++;
    570 				in6_ifstat_inc(ifp, ifs6_out_discard);
    571 				error = EHOSTUNREACH;
    572 				goto bad;
    573 			}
    574 		}
    575 
    576 		if (opt && opt->ip6po_hlim != -1)
    577 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
    578 	} else {
    579 		/* Multicast */
    580 		struct	in6_multi *in6m;
    581 
    582 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
    583 
    584 		/*
    585 		 * See if the caller provided any multicast options
    586 		 */
    587 		ifp = NULL;
    588 		if (im6o != NULL) {
    589 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
    590 			if (im6o->im6o_multicast_ifp != NULL)
    591 				ifp = im6o->im6o_multicast_ifp;
    592 		} else
    593 			ip6->ip6_hlim = ip6_defmcasthlim;
    594 
    595 		/*
    596 		 * See if the caller provided the outgoing interface
    597 		 * as an ancillary data.
    598 		 * Boundary check for ifindex is assumed to be already done.
    599 		 */
    600 		if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
    601 			ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex];
    602 
    603 		/*
    604 		 * If the destination is a node-local scope multicast,
    605 		 * the packet should be loop-backed only.
    606 		 */
    607 		if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
    608 			/*
    609 			 * If the outgoing interface is already specified,
    610 			 * it should be a loopback interface.
    611 			 */
    612 			if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) {
    613 				ip6stat.ip6s_badscope++;
    614 				error = ENETUNREACH; /* XXX: better error? */
    615 				/* XXX correct ifp? */
    616 				in6_ifstat_inc(ifp, ifs6_out_discard);
    617 				goto bad;
    618 			} else {
    619 				ifp = &loif[0];
    620 			}
    621 		}
    622 
    623 		if (opt && opt->ip6po_hlim != -1)
    624 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
    625 
    626 		/*
    627 		 * If caller did not provide an interface lookup a
    628 		 * default in the routing table.  This is either a
    629 		 * default for the speicfied group (i.e. a host
    630 		 * route), or a multicast default (a route for the
    631 		 * ``net'' ff00::/8).
    632 		 */
    633 		if (ifp == NULL) {
    634 			if (ro->ro_rt == 0) {
    635 				ro->ro_rt = rtalloc1((struct sockaddr *)
    636 						&ro->ro_dst, 0
    637 						);
    638 			}
    639 			if (ro->ro_rt == 0) {
    640 				ip6stat.ip6s_noroute++;
    641 				error = EHOSTUNREACH;
    642 				/* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
    643 				goto bad;
    644 			}
    645 			ia = ifatoia6(ro->ro_rt->rt_ifa);
    646 			ifp = ro->ro_rt->rt_ifp;
    647 			ro->ro_rt->rt_use++;
    648 		}
    649 
    650 		if ((flags & IPV6_FORWARDING) == 0)
    651 			in6_ifstat_inc(ifp, ifs6_out_request);
    652 		in6_ifstat_inc(ifp, ifs6_out_mcast);
    653 
    654 		/*
    655 		 * Confirm that the outgoing interface supports multicast.
    656 		 */
    657 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
    658 			ip6stat.ip6s_noroute++;
    659 			in6_ifstat_inc(ifp, ifs6_out_discard);
    660 			error = ENETUNREACH;
    661 			goto bad;
    662 		}
    663 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
    664 		if (in6m != NULL &&
    665 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
    666 			/*
    667 			 * If we belong to the destination multicast group
    668 			 * on the outgoing interface, and the caller did not
    669 			 * forbid loopback, loop back a copy.
    670 			 */
    671 			ip6_mloopback(ifp, m, dst);
    672 		} else {
    673 			/*
    674 			 * If we are acting as a multicast router, perform
    675 			 * multicast forwarding as if the packet had just
    676 			 * arrived on the interface to which we are about
    677 			 * to send.  The multicast forwarding function
    678 			 * recursively calls this function, using the
    679 			 * IPV6_FORWARDING flag to prevent infinite recursion.
    680 			 *
    681 			 * Multicasts that are looped back by ip6_mloopback(),
    682 			 * above, will be forwarded by the ip6_input() routine,
    683 			 * if necessary.
    684 			 */
    685 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
    686 				if (ip6_mforward(ip6, ifp, m) != 0) {
    687 					m_freem(m);
    688 					goto done;
    689 				}
    690 			}
    691 		}
    692 		/*
    693 		 * Multicasts with a hoplimit of zero may be looped back,
    694 		 * above, but must not be transmitted on a network.
    695 		 * Also, multicasts addressed to the loopback interface
    696 		 * are not sent -- the above call to ip6_mloopback() will
    697 		 * loop back a copy if this host actually belongs to the
    698 		 * destination group on the loopback interface.
    699 		 */
    700 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
    701 			m_freem(m);
    702 			goto done;
    703 		}
    704 	}
    705 
    706 	/*
    707 	 * Fill the outgoing inteface to tell the upper layer
    708 	 * to increment per-interface statistics.
    709 	 */
    710 	if (ifpp)
    711 		*ifpp = ifp;
    712 
    713 	/*
    714 	 * Determine path MTU.
    715 	 */
    716 	if (ro_pmtu != ro) {
    717 		/* The first hop and the final destination may differ. */
    718 		struct sockaddr_in6 *sin6_fin =
    719 			(struct sockaddr_in6 *)&ro_pmtu->ro_dst;
    720 		if (ro_pmtu->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
    721 				       !IN6_ARE_ADDR_EQUAL(&sin6_fin->sin6_addr,
    722 							   &finaldst))) {
    723 			RTFREE(ro_pmtu->ro_rt);
    724 			ro_pmtu->ro_rt = (struct rtentry *)0;
    725 		}
    726 		if (ro_pmtu->ro_rt == 0) {
    727 			bzero(sin6_fin, sizeof(*sin6_fin));
    728 			sin6_fin->sin6_family = AF_INET6;
    729 			sin6_fin->sin6_len = sizeof(struct sockaddr_in6);
    730 			sin6_fin->sin6_addr = finaldst;
    731 
    732 			rtalloc((struct route *)ro_pmtu);
    733 		}
    734 	}
    735 	if (ro_pmtu->ro_rt != NULL) {
    736 		u_int32_t ifmtu = nd_ifinfo[ifp->if_index].linkmtu;
    737 
    738 		mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
    739 		if (mtu > ifmtu) {
    740 			/*
    741 			 * The MTU on the route is larger than the MTU on
    742 			 * the interface!  This shouldn't happen, unless the
    743 			 * MTU of the interface has been changed after the
    744 			 * interface was brought up.  Change the MTU in the
    745 			 * route to match the interface MTU (as long as the
    746 			 * field isn't locked).
    747 			 */
    748 			 mtu = ifmtu;
    749 			 if ((ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
    750 				 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; /* XXX */
    751 		}
    752 	} else {
    753 		mtu = nd_ifinfo[ifp->if_index].linkmtu;
    754 	}
    755 
    756 	/* Fake scoped addresses */
    757 	if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
    758 		/*
    759 		 * If source or destination address is a scoped address, and
    760 		 * the packet is going to be sent to a loopback interface,
    761 		 * we should keep the original interface.
    762 		 */
    763 
    764 		/*
    765 		 * XXX: this is a very experimental and temporary solution.
    766 		 * We eventually have sockaddr_in6 and use the sin6_scope_id
    767 		 * field of the structure here.
    768 		 * We rely on the consistency between two scope zone ids
    769 		 * of source add destination, which should already be assured
    770 		 * Larger scopes than link will be supported in the near
    771 		 * future.
    772 		 */
    773 		origifp = NULL;
    774 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
    775 			origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
    776 		else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
    777 			origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
    778 		/*
    779 		 * XXX: origifp can be NULL even in those two cases above.
    780 		 * For example, if we remove the (only) link-local address
    781 		 * from the loopback interface, and try to send a link-local
    782 		 * address without link-id information.  Then the source
    783 		 * address is ::1, and the destination address is the
    784 		 * link-local address with its s6_addr16[1] being zero.
    785 		 * What is worse, if the packet goes to the loopback interface
    786 		 * by a default rejected route, the null pointer would be
    787 		 * passed to looutput, and the kernel would hang.
    788 		 * The following last resort would prevent such disaster.
    789 		 */
    790 		if (origifp == NULL)
    791 			origifp = ifp;
    792 	}
    793 	else
    794 		origifp = ifp;
    795 #ifndef FAKE_LOOPBACK_IF
    796 	if ((ifp->if_flags & IFF_LOOPBACK) == 0)
    797 #else
    798 	if (1)
    799 #endif
    800 	{
    801 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
    802 			ip6->ip6_src.s6_addr16[1] = 0;
    803 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
    804 			ip6->ip6_dst.s6_addr16[1] = 0;
    805 	}
    806 
    807 	/*
    808 	 * If the outgoing packet contains a hop-by-hop options header,
    809 	 * it must be examined and processed even by the source node.
    810 	 * (RFC 2460, section 4.)
    811 	 */
    812 	if (exthdrs.ip6e_hbh) {
    813 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh,
    814 					   struct ip6_hbh *);
    815 		u_int32_t dummy1; /* XXX unused */
    816 		u_int32_t dummy2; /* XXX unused */
    817 
    818 		/*
    819 		 *  XXX: if we have to send an ICMPv6 error to the sender,
    820 		 *       we need the M_LOOP flag since icmp6_error() expects
    821 		 *       the IPv6 and the hop-by-hop options header are
    822 		 *       continuous unless the flag is set.
    823 		 */
    824 		m->m_flags |= M_LOOP;
    825 		m->m_pkthdr.rcvif = ifp;
    826 		if (ip6_process_hopopts(m,
    827 					(u_int8_t *)(hbh + 1),
    828 					((hbh->ip6h_len + 1) << 3) -
    829 					sizeof(struct ip6_hbh),
    830 					&dummy1, &dummy2) < 0) {
    831 			/* m was already freed at this point */
    832 			error = EINVAL;/* better error? */
    833 			goto done;
    834 		}
    835 		m->m_flags &= ~M_LOOP; /* XXX */
    836 		m->m_pkthdr.rcvif = NULL;
    837 	}
    838 
    839 #ifdef PFIL_HOOKS
    840 	/*
    841 	 * Run through list of hooks for output packets.
    842 	 */
    843 	m1 = m;
    844 	pfh = pfil_hook_get(PFIL_OUT, &inetsw[ip_protox[IPPROTO_IPV6]].pr_pfh);
    845 	for (; pfh; pfh = pfh->pfil_link.tqe_next)
    846 		if (pfh->pfil_func) {
    847 		    	rv = pfh->pfil_func(ip6, sizeof(*ip6), ifp, 1, &m1);
    848 			if (rv) {
    849 				error = EHOSTUNREACH;
    850 				goto done;
    851 			}
    852 			m = m1;
    853 			if (m == NULL)
    854 				goto done;
    855 			ip6 = mtod(m, struct ip6_hdr *);
    856 		}
    857 #endif /* PFIL_HOOKS */
    858 	/*
    859 	 * Send the packet to the outgoing interface.
    860 	 * If necessary, do IPv6 fragmentation before sending.
    861 	 */
    862 	tlen = m->m_pkthdr.len;
    863 	if (tlen <= mtu
    864 #ifdef notyet
    865 	    /*
    866 	     * On any link that cannot convey a 1280-octet packet in one piece,
    867 	     * link-specific fragmentation and reassembly must be provided at
    868 	     * a layer below IPv6. [RFC 2460, sec.5]
    869 	     * Thus if the interface has ability of link-level fragmentation,
    870 	     * we can just send the packet even if the packet size is
    871 	     * larger than the link's MTU.
    872 	     * XXX: IFF_FRAGMENTABLE (or such) flag has not been defined yet...
    873 	     */
    874 
    875 	    || ifp->if_flags & IFF_FRAGMENTABLE
    876 #endif
    877 	    )
    878 	{
    879 #ifdef IFA_STATS
    880 		if (IFA_STATS) {
    881 			struct in6_ifaddr *ia6;
    882 			ip6 = mtod(m, struct ip6_hdr *);
    883 			ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
    884 			if (ia6) {
    885 				ia->ia_ifa.ifa_data.ifad_outbytes +=
    886 					m->m_pkthdr.len;
    887 			}
    888 		}
    889 #endif
    890 #ifdef IPSEC
    891 		/* clean ipsec history once it goes out of the node */
    892 		ipsec_delaux(m);
    893 #endif
    894 #ifdef OLDIP6OUTPUT
    895 		error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst,
    896 					  ro->ro_rt);
    897 #else
    898 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
    899 #endif
    900 		goto done;
    901 	} else if (mtu < IPV6_MMTU) {
    902 		/*
    903 		 * note that path MTU is never less than IPV6_MMTU
    904 		 * (see icmp6_input).
    905 		 */
    906 		error = EMSGSIZE;
    907 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
    908 		goto bad;
    909 	} else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
    910 		error = EMSGSIZE;
    911 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
    912 		goto bad;
    913 	} else {
    914 		struct mbuf **mnext, *m_frgpart;
    915 		struct ip6_frag *ip6f;
    916 		u_int32_t id = htonl(ip6_id++);
    917 		u_char nextproto;
    918 
    919 		/*
    920 		 * Too large for the destination or interface;
    921 		 * fragment if possible.
    922 		 * Must be able to put at least 8 bytes per fragment.
    923 		 */
    924 		hlen = unfragpartlen;
    925 		if (mtu > IPV6_MAXPACKET)
    926 			mtu = IPV6_MAXPACKET;
    927 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
    928 		if (len < 8) {
    929 			error = EMSGSIZE;
    930 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
    931 			goto bad;
    932 		}
    933 
    934 		mnext = &m->m_nextpkt;
    935 
    936 		/*
    937 		 * Change the next header field of the last header in the
    938 		 * unfragmentable part.
    939 		 */
    940 		if (exthdrs.ip6e_rthdr) {
    941 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
    942 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
    943 		} else if (exthdrs.ip6e_dest1) {
    944 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
    945 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
    946 		} else if (exthdrs.ip6e_hbh) {
    947 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
    948 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
    949 		} else {
    950 			nextproto = ip6->ip6_nxt;
    951 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
    952 		}
    953 
    954 		/*
    955 		 * Loop through length of segment after first fragment,
    956 		 * make new header and copy data of each part and link onto chain.
    957 		 */
    958 		m0 = m;
    959 		for (off = hlen; off < tlen; off += len) {
    960 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
    961 			if (!m) {
    962 				error = ENOBUFS;
    963 				ip6stat.ip6s_odropped++;
    964 				goto sendorfree;
    965 			}
    966 			m->m_flags = m0->m_flags & M_COPYFLAGS;
    967 			*mnext = m;
    968 			mnext = &m->m_nextpkt;
    969 			m->m_data += max_linkhdr;
    970 			mhip6 = mtod(m, struct ip6_hdr *);
    971 			*mhip6 = *ip6;
    972 			m->m_len = sizeof(*mhip6);
    973  			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
    974  			if (error) {
    975 				ip6stat.ip6s_odropped++;
    976 				goto sendorfree;
    977 			}
    978 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
    979 			if (off + len >= tlen)
    980 				len = tlen - off;
    981 			else
    982 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
    983 			mhip6->ip6_plen = htons((u_short)(len + hlen +
    984 							  sizeof(*ip6f) -
    985 							  sizeof(struct ip6_hdr)));
    986 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
    987 				error = ENOBUFS;
    988 				ip6stat.ip6s_odropped++;
    989 				goto sendorfree;
    990 			}
    991 			m_cat(m, m_frgpart);
    992 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
    993 			m->m_pkthdr.rcvif = (struct ifnet *)0;
    994 			ip6f->ip6f_reserved = 0;
    995 			ip6f->ip6f_ident = id;
    996 			ip6f->ip6f_nxt = nextproto;
    997 			ip6stat.ip6s_ofragments++;
    998 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
    999 		}
   1000 
   1001 		in6_ifstat_inc(ifp, ifs6_out_fragok);
   1002 	}
   1003 
   1004 	/*
   1005 	 * Remove leading garbages.
   1006 	 */
   1007 sendorfree:
   1008 	m = m0->m_nextpkt;
   1009 	m0->m_nextpkt = 0;
   1010 	m_freem(m0);
   1011 	for (m0 = m; m; m = m0) {
   1012 		m0 = m->m_nextpkt;
   1013 		m->m_nextpkt = 0;
   1014 		if (error == 0) {
   1015 #ifdef IFA_STATS
   1016 			if (IFA_STATS) {
   1017 				struct in6_ifaddr *ia6;
   1018 				ip6 = mtod(m, struct ip6_hdr *);
   1019 				ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
   1020 				if (ia6) {
   1021 					ia->ia_ifa.ifa_data.ifad_outbytes +=
   1022 						m->m_pkthdr.len;
   1023 				}
   1024 			}
   1025 #endif
   1026 #ifdef IPSEC
   1027 			/* clean ipsec history once it goes out of the node */
   1028 			ipsec_delaux(m);
   1029 #endif
   1030 #ifdef OLDIP6OUTPUT
   1031 			error = (*ifp->if_output)(ifp, m,
   1032 						  (struct sockaddr *)dst,
   1033 						  ro->ro_rt);
   1034 #else
   1035 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
   1036 #endif
   1037 		} else
   1038 			m_freem(m);
   1039 	}
   1040 
   1041 	if (error == 0)
   1042 		ip6stat.ip6s_fragmented++;
   1043 
   1044 done:
   1045 	if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
   1046 		RTFREE(ro->ro_rt);
   1047 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
   1048 		RTFREE(ro_pmtu->ro_rt);
   1049 	}
   1050 
   1051 #ifdef IPSEC
   1052 	if (sp != NULL)
   1053 		key_freesp(sp);
   1054 #endif /* IPSEC */
   1055 
   1056 	return(error);
   1057 
   1058 freehdrs:
   1059 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
   1060 	m_freem(exthdrs.ip6e_dest1);
   1061 	m_freem(exthdrs.ip6e_rthdr);
   1062 	m_freem(exthdrs.ip6e_dest2);
   1063 	/* fall through */
   1064 bad:
   1065 	m_freem(m);
   1066 	goto done;
   1067 }
   1068 
   1069 static int
   1070 ip6_copyexthdr(mp, hdr, hlen)
   1071 	struct mbuf **mp;
   1072 	caddr_t hdr;
   1073 	int hlen;
   1074 {
   1075 	struct mbuf *m;
   1076 
   1077 	if (hlen > MCLBYTES)
   1078 		return(ENOBUFS); /* XXX */
   1079 
   1080 	MGET(m, M_DONTWAIT, MT_DATA);
   1081 	if (!m)
   1082 		return(ENOBUFS);
   1083 
   1084 	if (hlen > MLEN) {
   1085 		MCLGET(m, M_DONTWAIT);
   1086 		if ((m->m_flags & M_EXT) == 0) {
   1087 			m_free(m);
   1088 			return(ENOBUFS);
   1089 		}
   1090 	}
   1091 	m->m_len = hlen;
   1092 	if (hdr)
   1093 		bcopy(hdr, mtod(m, caddr_t), hlen);
   1094 
   1095 	*mp = m;
   1096 	return(0);
   1097 }
   1098 
   1099 /*
   1100  * Insert jumbo payload option.
   1101  */
   1102 static int
   1103 ip6_insert_jumboopt(exthdrs, plen)
   1104 	struct ip6_exthdrs *exthdrs;
   1105 	u_int32_t plen;
   1106 {
   1107 	struct mbuf *mopt;
   1108 	u_char *optbuf;
   1109 
   1110 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
   1111 
   1112 	/*
   1113 	 * If there is no hop-by-hop options header, allocate new one.
   1114 	 * If there is one but it doesn't have enough space to store the
   1115 	 * jumbo payload option, allocate a cluster to store the whole options.
   1116 	 * Otherwise, use it to store the options.
   1117 	 */
   1118 	if (exthdrs->ip6e_hbh == 0) {
   1119 		MGET(mopt, M_DONTWAIT, MT_DATA);
   1120 		if (mopt == 0)
   1121 			return(ENOBUFS);
   1122 		mopt->m_len = JUMBOOPTLEN;
   1123 		optbuf = mtod(mopt, u_char *);
   1124 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
   1125 		exthdrs->ip6e_hbh = mopt;
   1126 	} else {
   1127 		struct ip6_hbh *hbh;
   1128 
   1129 		mopt = exthdrs->ip6e_hbh;
   1130 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
   1131 			caddr_t oldoptp = mtod(mopt, caddr_t);
   1132 			int oldoptlen = mopt->m_len;
   1133 
   1134 			if (mopt->m_flags & M_EXT)
   1135 				return(ENOBUFS); /* XXX */
   1136 			MCLGET(mopt, M_DONTWAIT);
   1137 			if ((mopt->m_flags & M_EXT) == 0)
   1138 				return(ENOBUFS);
   1139 
   1140 			bcopy(oldoptp, mtod(mopt, caddr_t), oldoptlen);
   1141 			optbuf = mtod(mopt, caddr_t) + oldoptlen;
   1142 			mopt->m_len = oldoptlen + JUMBOOPTLEN;
   1143 		} else {
   1144 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
   1145 			mopt->m_len += JUMBOOPTLEN;
   1146 		}
   1147 		optbuf[0] = IP6OPT_PADN;
   1148 		optbuf[1] = 1;
   1149 
   1150 		/*
   1151 		 * Adjust the header length according to the pad and
   1152 		 * the jumbo payload option.
   1153 		 */
   1154 		hbh = mtod(mopt, struct ip6_hbh *);
   1155 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
   1156 	}
   1157 
   1158 	/* fill in the option. */
   1159 	optbuf[2] = IP6OPT_JUMBO;
   1160 	optbuf[3] = 4;
   1161 	*(u_int32_t *)&optbuf[4] = htonl(plen + JUMBOOPTLEN);
   1162 
   1163 	/* finally, adjust the packet header length */
   1164 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
   1165 
   1166 	return(0);
   1167 #undef JUMBOOPTLEN
   1168 }
   1169 
   1170 /*
   1171  * Insert fragment header and copy unfragmentable header portions.
   1172  */
   1173 static int
   1174 ip6_insertfraghdr(m0, m, hlen, frghdrp)
   1175 	struct mbuf *m0, *m;
   1176 	int hlen;
   1177 	struct ip6_frag **frghdrp;
   1178 {
   1179 	struct mbuf *n, *mlast;
   1180 
   1181 	if (hlen > sizeof(struct ip6_hdr)) {
   1182 		n = m_copym(m0, sizeof(struct ip6_hdr),
   1183 			    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
   1184 		if (n == 0)
   1185 			return(ENOBUFS);
   1186 		m->m_next = n;
   1187 	} else
   1188 		n = m;
   1189 
   1190 	/* Search for the last mbuf of unfragmentable part. */
   1191 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
   1192 		;
   1193 
   1194 	if ((mlast->m_flags & M_EXT) == 0 &&
   1195 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
   1196 		/* use the trailing space of the last mbuf for the fragment hdr */
   1197 		*frghdrp =
   1198 			(struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len);
   1199 		mlast->m_len += sizeof(struct ip6_frag);
   1200 		m->m_pkthdr.len += sizeof(struct ip6_frag);
   1201 	} else {
   1202 		/* allocate a new mbuf for the fragment header */
   1203 		struct mbuf *mfrg;
   1204 
   1205 		MGET(mfrg, M_DONTWAIT, MT_DATA);
   1206 		if (mfrg == 0)
   1207 			return(ENOBUFS);
   1208 		mfrg->m_len = sizeof(struct ip6_frag);
   1209 		*frghdrp = mtod(mfrg, struct ip6_frag *);
   1210 		mlast->m_next = mfrg;
   1211 	}
   1212 
   1213 	return(0);
   1214 }
   1215 
   1216 /*
   1217  * IP6 socket option processing.
   1218  */
   1219 int
   1220 ip6_ctloutput(op, so, level, optname, mp)
   1221 	int op;
   1222 	struct socket *so;
   1223 	int level, optname;
   1224 	struct mbuf **mp;
   1225 {
   1226 	register struct in6pcb *in6p = sotoin6pcb(so);
   1227 	register struct mbuf *m = *mp;
   1228 	register int optval = 0;
   1229 	int error = 0;
   1230 	struct proc *p = curproc;	/* XXX */
   1231 
   1232 	if (level == IPPROTO_IPV6) {
   1233 		switch (op) {
   1234 
   1235 		case PRCO_SETOPT:
   1236 			switch (optname) {
   1237 			case IPV6_PKTOPTIONS:
   1238 				/* m is freed in ip6_pcbopts */
   1239 				return(ip6_pcbopts(&in6p->in6p_outputopts,
   1240 						   m, so));
   1241 			case IPV6_HOPOPTS:
   1242 			case IPV6_DSTOPTS:
   1243 				if (p == 0 || suser(p->p_ucred, &p->p_acflag)) {
   1244 					error = EPERM;
   1245 					break;
   1246 				}
   1247 				/* fall through */
   1248 			case IPV6_UNICAST_HOPS:
   1249 			case IPV6_RECVOPTS:
   1250 			case IPV6_RECVRETOPTS:
   1251 			case IPV6_RECVDSTADDR:
   1252 			case IPV6_PKTINFO:
   1253 			case IPV6_HOPLIMIT:
   1254 			case IPV6_RTHDR:
   1255 			case IPV6_CHECKSUM:
   1256 			case IPV6_FAITH:
   1257 #ifndef INET6_BINDV6ONLY
   1258 			case IPV6_BINDV6ONLY:
   1259 #endif
   1260 				if (!m || m->m_len != sizeof(int))
   1261 					error = EINVAL;
   1262 				else {
   1263 					optval = *mtod(m, int *);
   1264 					switch (optname) {
   1265 
   1266 					case IPV6_UNICAST_HOPS:
   1267 						if (optval < -1 || optval >= 256)
   1268 							error = EINVAL;
   1269 						else {
   1270 							/* -1 = kernel default */
   1271 							in6p->in6p_hops = optval;
   1272 						}
   1273 						break;
   1274 #define OPTSET(bit) \
   1275 	if (optval) \
   1276 		in6p->in6p_flags |= bit; \
   1277 	else \
   1278 		in6p->in6p_flags &= ~bit;
   1279 
   1280 					case IPV6_RECVOPTS:
   1281 						OPTSET(IN6P_RECVOPTS);
   1282 						break;
   1283 
   1284 					case IPV6_RECVRETOPTS:
   1285 						OPTSET(IN6P_RECVRETOPTS);
   1286 						break;
   1287 
   1288 					case IPV6_RECVDSTADDR:
   1289 						OPTSET(IN6P_RECVDSTADDR);
   1290 						break;
   1291 
   1292 					case IPV6_PKTINFO:
   1293 						OPTSET(IN6P_PKTINFO);
   1294 						break;
   1295 
   1296 					case IPV6_HOPLIMIT:
   1297 						OPTSET(IN6P_HOPLIMIT);
   1298 						break;
   1299 
   1300 					case IPV6_HOPOPTS:
   1301 						OPTSET(IN6P_HOPOPTS);
   1302 						break;
   1303 
   1304 					case IPV6_DSTOPTS:
   1305 						OPTSET(IN6P_DSTOPTS);
   1306 						break;
   1307 
   1308 					case IPV6_RTHDR:
   1309 						OPTSET(IN6P_RTHDR);
   1310 						break;
   1311 
   1312 					case IPV6_CHECKSUM:
   1313 						in6p->in6p_cksum = optval;
   1314 						break;
   1315 
   1316 					case IPV6_FAITH:
   1317 						OPTSET(IN6P_FAITH);
   1318 						break;
   1319 
   1320 #ifndef INET6_BINDV6ONLY
   1321 					case IPV6_BINDV6ONLY:
   1322 						OPTSET(IN6P_BINDV6ONLY);
   1323 						break;
   1324 #endif
   1325 					}
   1326 				}
   1327 				break;
   1328 #undef OPTSET
   1329 
   1330 			case IPV6_MULTICAST_IF:
   1331 			case IPV6_MULTICAST_HOPS:
   1332 			case IPV6_MULTICAST_LOOP:
   1333 			case IPV6_JOIN_GROUP:
   1334 			case IPV6_LEAVE_GROUP:
   1335 				error =	ip6_setmoptions(optname, &in6p->in6p_moptions, m);
   1336 				break;
   1337 
   1338 			case IPV6_PORTRANGE:
   1339 				optval = *mtod(m, int *);
   1340 
   1341 				switch (optval) {
   1342 				case IPV6_PORTRANGE_DEFAULT:
   1343 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
   1344 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
   1345 					break;
   1346 
   1347 				case IPV6_PORTRANGE_HIGH:
   1348 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
   1349 					in6p->in6p_flags |= IN6P_HIGHPORT;
   1350 					break;
   1351 
   1352 				case IPV6_PORTRANGE_LOW:
   1353 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
   1354 					in6p->in6p_flags |= IN6P_LOWPORT;
   1355 					break;
   1356 
   1357 				default:
   1358 					error = EINVAL;
   1359 					break;
   1360 				}
   1361 				break;
   1362 
   1363 #ifdef IPSEC
   1364 			case IPV6_IPSEC_POLICY:
   1365 			    {
   1366 				caddr_t req = NULL;
   1367 				size_t len = 0;
   1368 
   1369 				int priv = 0;
   1370 				if (p == 0 || suser(p->p_ucred, &p->p_acflag))
   1371 					priv = 0;
   1372 				else
   1373 					priv = 1;
   1374 				if (m) {
   1375 					req = mtod(m, caddr_t);
   1376 					len = m->m_len;
   1377 				}
   1378 				error = ipsec6_set_policy(in6p,
   1379 				                   optname, req, len, priv);
   1380 			    }
   1381 				break;
   1382 #endif /* IPSEC */
   1383 
   1384 			default:
   1385 				error = ENOPROTOOPT;
   1386 				break;
   1387 			}
   1388 			if (m)
   1389 				(void)m_free(m);
   1390 			break;
   1391 
   1392 		case PRCO_GETOPT:
   1393 			switch (optname) {
   1394 
   1395 			case IPV6_OPTIONS:
   1396 			case IPV6_RETOPTS:
   1397 #if 0
   1398 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
   1399 				if (in6p->in6p_options) {
   1400 					m->m_len = in6p->in6p_options->m_len;
   1401 					bcopy(mtod(in6p->in6p_options, caddr_t),
   1402 					      mtod(m, caddr_t),
   1403 					      (unsigned)m->m_len);
   1404 				} else
   1405 					m->m_len = 0;
   1406 				break;
   1407 #else
   1408 				error = ENOPROTOOPT;
   1409 				break;
   1410 #endif
   1411 
   1412 			case IPV6_PKTOPTIONS:
   1413 				if (in6p->in6p_options) {
   1414 					*mp = m_copym(in6p->in6p_options, 0,
   1415 						      M_COPYALL, M_WAIT);
   1416 				} else {
   1417 					*mp = m_get(M_WAIT, MT_SOOPTS);
   1418 					(*mp)->m_len = 0;
   1419 				}
   1420 				break;
   1421 
   1422 			case IPV6_HOPOPTS:
   1423 			case IPV6_DSTOPTS:
   1424 				if (p == 0 || suser(p->p_ucred, &p->p_acflag)) {
   1425 					error = EPERM;
   1426 					break;
   1427 				}
   1428 				/* fall through */
   1429 			case IPV6_UNICAST_HOPS:
   1430 			case IPV6_RECVOPTS:
   1431 			case IPV6_RECVRETOPTS:
   1432 			case IPV6_RECVDSTADDR:
   1433 			case IPV6_PORTRANGE:
   1434 			case IPV6_PKTINFO:
   1435 			case IPV6_HOPLIMIT:
   1436 			case IPV6_RTHDR:
   1437 			case IPV6_CHECKSUM:
   1438 			case IPV6_FAITH:
   1439 #ifndef INET6_BINDV6ONLY
   1440 			case IPV6_BINDV6ONLY:
   1441 #endif
   1442 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
   1443 				m->m_len = sizeof(int);
   1444 				switch (optname) {
   1445 
   1446 				case IPV6_UNICAST_HOPS:
   1447 					optval = in6p->in6p_hops;
   1448 					break;
   1449 
   1450 #define OPTBIT(bit) (in6p->in6p_flags & bit ? 1 : 0)
   1451 
   1452 				case IPV6_RECVOPTS:
   1453 					optval = OPTBIT(IN6P_RECVOPTS);
   1454 					break;
   1455 
   1456 				case IPV6_RECVRETOPTS:
   1457 					optval = OPTBIT(IN6P_RECVRETOPTS);
   1458 					break;
   1459 
   1460 				case IPV6_RECVDSTADDR:
   1461 					optval = OPTBIT(IN6P_RECVDSTADDR);
   1462 					break;
   1463 
   1464 				case IPV6_PORTRANGE:
   1465 				    {
   1466 					int flags;
   1467 					flags = in6p->in6p_flags;
   1468 					if (flags & IN6P_HIGHPORT)
   1469 						optval = IPV6_PORTRANGE_HIGH;
   1470 					else if (flags & IN6P_LOWPORT)
   1471 						optval = IPV6_PORTRANGE_LOW;
   1472 					else
   1473 						optval = 0;
   1474 					break;
   1475 				    }
   1476 
   1477 				case IPV6_PKTINFO:
   1478 					optval = OPTBIT(IN6P_PKTINFO);
   1479 					break;
   1480 
   1481 				case IPV6_HOPLIMIT:
   1482 					optval = OPTBIT(IN6P_HOPLIMIT);
   1483 					break;
   1484 
   1485 				case IPV6_HOPOPTS:
   1486 					optval = OPTBIT(IN6P_HOPOPTS);
   1487 					break;
   1488 
   1489 				case IPV6_DSTOPTS:
   1490 					optval = OPTBIT(IN6P_DSTOPTS);
   1491 					break;
   1492 
   1493 				case IPV6_RTHDR:
   1494 					optval = OPTBIT(IN6P_RTHDR);
   1495 					break;
   1496 
   1497 				case IPV6_CHECKSUM:
   1498 					optval = in6p->in6p_cksum;
   1499 					break;
   1500 
   1501 				case IPV6_FAITH:
   1502 					optval = OPTBIT(IN6P_FAITH);
   1503 					break;
   1504 
   1505 #ifndef INET6_BINDV6ONLY
   1506 				case IPV6_BINDV6ONLY:
   1507 					optval = OPTBIT(IN6P_BINDV6ONLY);
   1508 					break;
   1509 #endif
   1510 				}
   1511 				*mtod(m, int *) = optval;
   1512 				break;
   1513 
   1514 			case IPV6_MULTICAST_IF:
   1515 			case IPV6_MULTICAST_HOPS:
   1516 			case IPV6_MULTICAST_LOOP:
   1517 			case IPV6_JOIN_GROUP:
   1518 			case IPV6_LEAVE_GROUP:
   1519 				error = ip6_getmoptions(optname, in6p->in6p_moptions, mp);
   1520 				break;
   1521 
   1522 #ifdef IPSEC
   1523 			case IPV6_IPSEC_POLICY:
   1524 			{
   1525 				caddr_t req = NULL;
   1526 				size_t len = 0;
   1527 
   1528 				if (m) {
   1529 					req = mtod(m, caddr_t);
   1530 					len = m->m_len;
   1531 				}
   1532 				error = ipsec6_get_policy(in6p, req, len, mp);
   1533 				break;
   1534 			}
   1535 #endif /* IPSEC */
   1536 
   1537 			default:
   1538 				error = ENOPROTOOPT;
   1539 				break;
   1540 			}
   1541 			break;
   1542 		}
   1543 	} else {
   1544 		error = EINVAL;
   1545 		if (op == PRCO_SETOPT && *mp)
   1546 			(void)m_free(*mp);
   1547 	}
   1548 	return(error);
   1549 }
   1550 
   1551 /*
   1552  * Set up IP6 options in pcb for insertion in output packets.
   1553  * Store in mbuf with pointer in pcbopt, adding pseudo-option
   1554  * with destination address if source routed.
   1555  */
   1556 static int
   1557 ip6_pcbopts(pktopt, m, so)
   1558 	struct ip6_pktopts **pktopt;
   1559 	register struct mbuf *m;
   1560 	struct socket *so;
   1561 {
   1562 	register struct ip6_pktopts *opt = *pktopt;
   1563 	int error = 0;
   1564 	struct proc *p = curproc;	/* XXX */
   1565 	int priv = 0;
   1566 
   1567 	/* turn off any old options. */
   1568 	if (opt) {
   1569 		if (opt->ip6po_m)
   1570 			(void)m_free(opt->ip6po_m);
   1571 	} else
   1572 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
   1573 	*pktopt = 0;
   1574 
   1575 	if (!m || m->m_len == 0) {
   1576 		/*
   1577 		 * Only turning off any previous options.
   1578 		 */
   1579 		if (opt)
   1580 			free(opt, M_IP6OPT);
   1581 		if (m)
   1582 			(void)m_free(m);
   1583 		return(0);
   1584 	}
   1585 
   1586 	/*  set options specified by user. */
   1587 	if (p && !suser(p->p_ucred, &p->p_acflag))
   1588 		priv = 1;
   1589 	if ((error = ip6_setpktoptions(m, opt, priv)) != 0) {
   1590 		(void)m_free(m);
   1591 		return(error);
   1592 	}
   1593 	*pktopt = opt;
   1594 	return(0);
   1595 }
   1596 
   1597 /*
   1598  * Set the IP6 multicast options in response to user setsockopt().
   1599  */
   1600 static int
   1601 ip6_setmoptions(optname, im6op, m)
   1602 	int optname;
   1603 	struct ip6_moptions **im6op;
   1604 	struct mbuf *m;
   1605 {
   1606 	int error = 0;
   1607 	u_int loop, ifindex;
   1608 	struct ipv6_mreq *mreq;
   1609 	struct ifnet *ifp;
   1610 	struct ip6_moptions *im6o = *im6op;
   1611 	struct route_in6 ro;
   1612 	struct sockaddr_in6 *dst;
   1613 	struct in6_multi_mship *imm;
   1614 	struct proc *p = curproc;	/* XXX */
   1615 
   1616 	if (im6o == NULL) {
   1617 		/*
   1618 		 * No multicast option buffer attached to the pcb;
   1619 		 * allocate one and initialize to default values.
   1620 		 */
   1621 		im6o = (struct ip6_moptions *)
   1622 			malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
   1623 
   1624 		if (im6o == NULL)
   1625 			return(ENOBUFS);
   1626 		*im6op = im6o;
   1627 		im6o->im6o_multicast_ifp = NULL;
   1628 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
   1629 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
   1630 		LIST_INIT(&im6o->im6o_memberships);
   1631 	}
   1632 
   1633 	switch (optname) {
   1634 
   1635 	case IPV6_MULTICAST_IF:
   1636 		/*
   1637 		 * Select the interface for outgoing multicast packets.
   1638 		 */
   1639 		if (m == NULL || m->m_len != sizeof(u_int)) {
   1640 			error = EINVAL;
   1641 			break;
   1642 		}
   1643 		ifindex = *(mtod(m, u_int *));
   1644 		if (ifindex < 0 || if_index < ifindex) {
   1645 			error = ENXIO;	/* XXX EINVAL? */
   1646 			break;
   1647 		}
   1648 		ifp = ifindex2ifnet[ifindex];
   1649 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
   1650 			error = EADDRNOTAVAIL;
   1651 			break;
   1652 		}
   1653 		im6o->im6o_multicast_ifp = ifp;
   1654 		break;
   1655 
   1656 	case IPV6_MULTICAST_HOPS:
   1657 	    {
   1658 		/*
   1659 		 * Set the IP6 hoplimit for outgoing multicast packets.
   1660 		 */
   1661 		int optval;
   1662 		if (m == NULL || m->m_len != sizeof(int)) {
   1663 			error = EINVAL;
   1664 			break;
   1665 		}
   1666 		optval = *(mtod(m, u_int *));
   1667 		if (optval < -1 || optval >= 256)
   1668 			error = EINVAL;
   1669 		else if (optval == -1)
   1670 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
   1671 		else
   1672 			im6o->im6o_multicast_hlim = optval;
   1673 		break;
   1674 	    }
   1675 
   1676 	case IPV6_MULTICAST_LOOP:
   1677 		/*
   1678 		 * Set the loopback flag for outgoing multicast packets.
   1679 		 * Must be zero or one.
   1680 		 */
   1681 		if (m == NULL || m->m_len != sizeof(u_int) ||
   1682 		   (loop = *(mtod(m, u_int *))) > 1) {
   1683 			error = EINVAL;
   1684 			break;
   1685 		}
   1686 		im6o->im6o_multicast_loop = loop;
   1687 		break;
   1688 
   1689 	case IPV6_JOIN_GROUP:
   1690 		/*
   1691 		 * Add a multicast group membership.
   1692 		 * Group must be a valid IP6 multicast address.
   1693 		 */
   1694 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
   1695 			error = EINVAL;
   1696 			break;
   1697 		}
   1698 		mreq = mtod(m, struct ipv6_mreq *);
   1699 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
   1700 			/*
   1701 			 * We use the unspecified address to specify to accept
   1702 			 * all multicast addresses. Only super user is allowed
   1703 			 * to do this.
   1704 			 */
   1705 			if (suser(p->p_ucred, &p->p_acflag)) {
   1706 				error = EACCES;
   1707 				break;
   1708 			}
   1709 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
   1710 			error = EINVAL;
   1711 			break;
   1712 		}
   1713 
   1714 		/*
   1715 		 * If the interface is specified, validate it.
   1716 		 */
   1717 		if (mreq->ipv6mr_interface < 0
   1718 		 || if_index < mreq->ipv6mr_interface) {
   1719 			error = ENXIO;	/* XXX EINVAL? */
   1720 			break;
   1721 		}
   1722 		/*
   1723 		 * If no interface was explicitly specified, choose an
   1724 		 * appropriate one according to the given multicast address.
   1725 		 */
   1726 		if (mreq->ipv6mr_interface == 0) {
   1727 			/*
   1728 			 * If the multicast address is in node-local scope,
   1729 			 * the interface should be a loopback interface.
   1730 			 * Otherwise, look up the routing table for the
   1731 			 * address, and choose the outgoing interface.
   1732 			 *   XXX: is it a good approach?
   1733 			 */
   1734 			if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
   1735 				ifp = &loif[0];
   1736 			} else {
   1737 				ro.ro_rt = NULL;
   1738 				dst = (struct sockaddr_in6 *)&ro.ro_dst;
   1739 				bzero(dst, sizeof(*dst));
   1740 				dst->sin6_len = sizeof(struct sockaddr_in6);
   1741 				dst->sin6_family = AF_INET6;
   1742 				dst->sin6_addr = mreq->ipv6mr_multiaddr;
   1743 				rtalloc((struct route *)&ro);
   1744 				if (ro.ro_rt == NULL) {
   1745 					error = EADDRNOTAVAIL;
   1746 					break;
   1747 				}
   1748 				ifp = ro.ro_rt->rt_ifp;
   1749 				rtfree(ro.ro_rt);
   1750 			}
   1751 		} else
   1752 			ifp = ifindex2ifnet[mreq->ipv6mr_interface];
   1753 
   1754 		/*
   1755 		 * See if we found an interface, and confirm that it
   1756 		 * supports multicast
   1757 		 */
   1758 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
   1759 			error = EADDRNOTAVAIL;
   1760 			break;
   1761 		}
   1762 		/*
   1763 		 * Put interface index into the multicast address,
   1764 		 * if the address has link-local scope.
   1765 		 */
   1766 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
   1767 			mreq->ipv6mr_multiaddr.s6_addr16[1]
   1768 				= htons(mreq->ipv6mr_interface);
   1769 		}
   1770 		/*
   1771 		 * See if the membership already exists.
   1772 		 */
   1773 		for (imm = im6o->im6o_memberships.lh_first;
   1774 		     imm != NULL; imm = imm->i6mm_chain.le_next)
   1775 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
   1776 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
   1777 					       &mreq->ipv6mr_multiaddr))
   1778 				break;
   1779 		if (imm != NULL) {
   1780 			error = EADDRINUSE;
   1781 			break;
   1782 		}
   1783 		/*
   1784 		 * Everything looks good; add a new record to the multicast
   1785 		 * address list for the given interface.
   1786 		 */
   1787 		imm = malloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
   1788 		if (imm == NULL) {
   1789 			error = ENOBUFS;
   1790 			break;
   1791 		}
   1792 		if ((imm->i6mm_maddr =
   1793 		     in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
   1794 			free(imm, M_IPMADDR);
   1795 			break;
   1796 		}
   1797 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
   1798 		break;
   1799 
   1800 	case IPV6_LEAVE_GROUP:
   1801 		/*
   1802 		 * Drop a multicast group membership.
   1803 		 * Group must be a valid IP6 multicast address.
   1804 		 */
   1805 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
   1806 			error = EINVAL;
   1807 			break;
   1808 		}
   1809 		mreq = mtod(m, struct ipv6_mreq *);
   1810 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
   1811 			if (suser(p->p_ucred, &p->p_acflag)) {
   1812 				error = EACCES;
   1813 				break;
   1814 			}
   1815 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
   1816 			error = EINVAL;
   1817 			break;
   1818 		}
   1819 		/*
   1820 		 * If an interface address was specified, get a pointer
   1821 		 * to its ifnet structure.
   1822 		 */
   1823 		if (mreq->ipv6mr_interface < 0
   1824 		 || if_index < mreq->ipv6mr_interface) {
   1825 			error = ENXIO;	/* XXX EINVAL? */
   1826 			break;
   1827 		}
   1828 		ifp = ifindex2ifnet[mreq->ipv6mr_interface];
   1829 		/*
   1830 		 * Put interface index into the multicast address,
   1831 		 * if the address has link-local scope.
   1832 		 */
   1833 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
   1834 			mreq->ipv6mr_multiaddr.s6_addr16[1]
   1835 				= htons(mreq->ipv6mr_interface);
   1836 		}
   1837 		/*
   1838 		 * Find the membership in the membership list.
   1839 		 */
   1840 		for (imm = im6o->im6o_memberships.lh_first;
   1841 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
   1842 			if ((ifp == NULL ||
   1843 			     imm->i6mm_maddr->in6m_ifp == ifp) &&
   1844 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
   1845 					       &mreq->ipv6mr_multiaddr))
   1846 				break;
   1847 		}
   1848 		if (imm == NULL) {
   1849 			/* Unable to resolve interface */
   1850 			error = EADDRNOTAVAIL;
   1851 			break;
   1852 		}
   1853 		/*
   1854 		 * Give up the multicast address record to which the
   1855 		 * membership points.
   1856 		 */
   1857 		LIST_REMOVE(imm, i6mm_chain);
   1858 		in6_delmulti(imm->i6mm_maddr);
   1859 		free(imm, M_IPMADDR);
   1860 		break;
   1861 
   1862 	default:
   1863 		error = EOPNOTSUPP;
   1864 		break;
   1865 	}
   1866 
   1867 	/*
   1868 	 * If all options have default values, no need to keep the mbuf.
   1869 	 */
   1870 	if (im6o->im6o_multicast_ifp == NULL &&
   1871 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
   1872 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
   1873 	    im6o->im6o_memberships.lh_first == NULL) {
   1874 		free(*im6op, M_IPMOPTS);
   1875 		*im6op = NULL;
   1876 	}
   1877 
   1878 	return(error);
   1879 }
   1880 
   1881 /*
   1882  * Return the IP6 multicast options in response to user getsockopt().
   1883  */
   1884 static int
   1885 ip6_getmoptions(optname, im6o, mp)
   1886 	int optname;
   1887 	register struct ip6_moptions *im6o;
   1888 	register struct mbuf **mp;
   1889 {
   1890 	u_int *hlim, *loop, *ifindex;
   1891 
   1892 	*mp = m_get(M_WAIT, MT_SOOPTS);
   1893 
   1894 	switch (optname) {
   1895 
   1896 	case IPV6_MULTICAST_IF:
   1897 		ifindex = mtod(*mp, u_int *);
   1898 		(*mp)->m_len = sizeof(u_int);
   1899 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
   1900 			*ifindex = 0;
   1901 		else
   1902 			*ifindex = im6o->im6o_multicast_ifp->if_index;
   1903 		return(0);
   1904 
   1905 	case IPV6_MULTICAST_HOPS:
   1906 		hlim = mtod(*mp, u_int *);
   1907 		(*mp)->m_len = sizeof(u_int);
   1908 		if (im6o == NULL)
   1909 			*hlim = ip6_defmcasthlim;
   1910 		else
   1911 			*hlim = im6o->im6o_multicast_hlim;
   1912 		return(0);
   1913 
   1914 	case IPV6_MULTICAST_LOOP:
   1915 		loop = mtod(*mp, u_int *);
   1916 		(*mp)->m_len = sizeof(u_int);
   1917 		if (im6o == NULL)
   1918 			*loop = ip6_defmcasthlim;
   1919 		else
   1920 			*loop = im6o->im6o_multicast_loop;
   1921 		return(0);
   1922 
   1923 	default:
   1924 		return(EOPNOTSUPP);
   1925 	}
   1926 }
   1927 
   1928 /*
   1929  * Discard the IP6 multicast options.
   1930  */
   1931 void
   1932 ip6_freemoptions(im6o)
   1933 	register struct ip6_moptions *im6o;
   1934 {
   1935 	struct in6_multi_mship *imm;
   1936 
   1937 	if (im6o == NULL)
   1938 		return;
   1939 
   1940 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
   1941 		LIST_REMOVE(imm, i6mm_chain);
   1942 		if (imm->i6mm_maddr)
   1943 			in6_delmulti(imm->i6mm_maddr);
   1944 		free(imm, M_IPMADDR);
   1945 	}
   1946 	free(im6o, M_IPMOPTS);
   1947 }
   1948 
   1949 /*
   1950  * Set IPv6 outgoing packet options based on advanced API.
   1951  */
   1952 int
   1953 ip6_setpktoptions(control, opt, priv)
   1954 	struct mbuf *control;
   1955 	struct ip6_pktopts *opt;
   1956 	int priv;
   1957 {
   1958 	register struct cmsghdr *cm = 0;
   1959 
   1960 	if (control == 0 || opt == 0)
   1961 		return(EINVAL);
   1962 
   1963 	bzero(opt, sizeof(*opt));
   1964 	opt->ip6po_hlim = -1; /* -1 means to use default hop limit */
   1965 
   1966 	/*
   1967 	 * XXX: Currently, we assume all the optional information is stored
   1968 	 * in a single mbuf.
   1969 	 */
   1970 	if (control->m_next)
   1971 		return(EINVAL);
   1972 
   1973 	opt->ip6po_m = control;
   1974 
   1975 	for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
   1976 		     control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
   1977 		cm = mtod(control, struct cmsghdr *);
   1978 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
   1979 			return(EINVAL);
   1980 		if (cm->cmsg_level != IPPROTO_IPV6)
   1981 			continue;
   1982 
   1983 		switch(cm->cmsg_type) {
   1984 		case IPV6_PKTINFO:
   1985 			if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo)))
   1986 				return(EINVAL);
   1987 			opt->ip6po_pktinfo = (struct in6_pktinfo *)CMSG_DATA(cm);
   1988 			if (opt->ip6po_pktinfo->ipi6_ifindex &&
   1989 			    IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr))
   1990 				opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] =
   1991 					htons(opt->ip6po_pktinfo->ipi6_ifindex);
   1992 
   1993 			if (opt->ip6po_pktinfo->ipi6_ifindex > if_index
   1994 			 || opt->ip6po_pktinfo->ipi6_ifindex < 0) {
   1995 				return(ENXIO);
   1996 			}
   1997 
   1998 			/*
   1999 			 * Check if the requested source address is indeed a
   2000 			 * unicast address assigned to the node, and can be
   2001 			 * used as the packet's source address.
   2002 			 */
   2003 			if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
   2004 				struct ifaddr *ia;
   2005 				struct in6_ifaddr *ia6;
   2006 				struct sockaddr_in6 sin6;
   2007 
   2008 				bzero(&sin6, sizeof(sin6));
   2009 				sin6.sin6_len = sizeof(sin6);
   2010 				sin6.sin6_family = AF_INET6;
   2011 				sin6.sin6_addr =
   2012 					opt->ip6po_pktinfo->ipi6_addr;
   2013 				ia = ifa_ifwithaddr(sin6tosa(&sin6));
   2014 				if (ia == NULL ||
   2015 				    (opt->ip6po_pktinfo->ipi6_ifindex &&
   2016 				     (ia->ifa_ifp->if_index !=
   2017 				      opt->ip6po_pktinfo->ipi6_ifindex))) {
   2018 					return(EADDRNOTAVAIL);
   2019 				}
   2020 				ia6 = (struct in6_ifaddr *)ia;
   2021 				if ((ia6->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) != 0) {
   2022 					return(EADDRNOTAVAIL);
   2023 				}
   2024 
   2025 				/*
   2026 				 * Check if the requested source address is
   2027 				 * indeed a unicast address assigned to the
   2028 				 * node.
   2029 				 */
   2030 				if (IN6_IS_ADDR_MULTICAST(&opt->ip6po_pktinfo->ipi6_addr))
   2031 					return(EADDRNOTAVAIL);
   2032 			}
   2033 			break;
   2034 
   2035 		case IPV6_HOPLIMIT:
   2036 			if (cm->cmsg_len != CMSG_LEN(sizeof(int)))
   2037 				return(EINVAL);
   2038 
   2039 			opt->ip6po_hlim = *(int *)CMSG_DATA(cm);
   2040 			if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255)
   2041 				return(EINVAL);
   2042 			break;
   2043 
   2044 		case IPV6_NEXTHOP:
   2045 			if (!priv)
   2046 				return(EPERM);
   2047 
   2048 			if (cm->cmsg_len < sizeof(u_char) ||
   2049 			    /* check if cmsg_len is large enough for sa_len */
   2050 			    cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm)))
   2051 				return(EINVAL);
   2052 
   2053 			opt->ip6po_nexthop = (struct sockaddr *)CMSG_DATA(cm);
   2054 
   2055 			break;
   2056 
   2057 		case IPV6_HOPOPTS:
   2058 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh)))
   2059 				return(EINVAL);
   2060 			opt->ip6po_hbh = (struct ip6_hbh *)CMSG_DATA(cm);
   2061 			if (cm->cmsg_len !=
   2062 			    CMSG_LEN((opt->ip6po_hbh->ip6h_len + 1) << 3))
   2063 				return(EINVAL);
   2064 			break;
   2065 
   2066 		case IPV6_DSTOPTS:
   2067 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest)))
   2068 				return(EINVAL);
   2069 
   2070 			/*
   2071 			 * If there is no routing header yet, the destination
   2072 			 * options header should be put on the 1st part.
   2073 			 * Otherwise, the header should be on the 2nd part.
   2074 			 * (See RFC 2460, section 4.1)
   2075 			 */
   2076 			if (opt->ip6po_rthdr == NULL) {
   2077 				opt->ip6po_dest1 =
   2078 					(struct ip6_dest *)CMSG_DATA(cm);
   2079 				if (cm->cmsg_len !=
   2080 				    CMSG_LEN((opt->ip6po_dest1->ip6d_len + 1)
   2081 					     << 3))
   2082 					return(EINVAL);
   2083 			}
   2084 			else {
   2085 				opt->ip6po_dest2 =
   2086 					(struct ip6_dest *)CMSG_DATA(cm);
   2087 				if (cm->cmsg_len !=
   2088 				    CMSG_LEN((opt->ip6po_dest2->ip6d_len + 1)
   2089 					     << 3))
   2090 					return(EINVAL);
   2091 			}
   2092 			break;
   2093 
   2094 		case IPV6_RTHDR:
   2095 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr)))
   2096 				return(EINVAL);
   2097 			opt->ip6po_rthdr = (struct ip6_rthdr *)CMSG_DATA(cm);
   2098 			if (cm->cmsg_len !=
   2099 			    CMSG_LEN((opt->ip6po_rthdr->ip6r_len + 1) << 3))
   2100 				return(EINVAL);
   2101 			switch(opt->ip6po_rthdr->ip6r_type) {
   2102 			case IPV6_RTHDR_TYPE_0:
   2103 				if (opt->ip6po_rthdr->ip6r_segleft == 0)
   2104 					return(EINVAL);
   2105 				break;
   2106 			default:
   2107 				return(EINVAL);
   2108 			}
   2109 			break;
   2110 
   2111 		default:
   2112 			return(ENOPROTOOPT);
   2113 		}
   2114 	}
   2115 
   2116 	return(0);
   2117 }
   2118 
   2119 /*
   2120  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
   2121  * packet to the input queue of a specified interface.  Note that this
   2122  * calls the output routine of the loopback "driver", but with an interface
   2123  * pointer that might NOT be &loif -- easier than replicating that code here.
   2124  */
   2125 void
   2126 ip6_mloopback(ifp, m, dst)
   2127 	struct ifnet *ifp;
   2128 	register struct mbuf *m;
   2129 	register struct sockaddr_in6 *dst;
   2130 {
   2131 	struct mbuf *copym;
   2132 	struct ip6_hdr *ip6;
   2133 
   2134 	copym = m_copy(m, 0, M_COPYALL);
   2135 	if (copym == NULL)
   2136 		return;
   2137 
   2138 	/*
   2139 	 * Make sure to deep-copy IPv6 header portion in case the data
   2140 	 * is in an mbuf cluster, so that we can safely override the IPv6
   2141 	 * header portion later.
   2142 	 */
   2143 	if ((copym->m_flags & M_EXT) != 0 ||
   2144 	    copym->m_len < sizeof(struct ip6_hdr)) {
   2145 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
   2146 		if (copym == NULL)
   2147 			return;
   2148 	}
   2149 
   2150 #ifdef DIAGNOSTIC
   2151 	if (copym->m_len < sizeof(*ip6)) {
   2152 		m_freem(copym);
   2153 		return;
   2154 	}
   2155 #endif
   2156 
   2157 #ifndef FAKE_LOOPBACK_IF
   2158 	if ((ifp->if_flags & IFF_LOOPBACK) == 0)
   2159 #else
   2160 	if (1)
   2161 #endif
   2162 	{
   2163 		ip6 = mtod(copym, struct ip6_hdr *);
   2164 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
   2165 			ip6->ip6_src.s6_addr16[1] = 0;
   2166 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
   2167 			ip6->ip6_dst.s6_addr16[1] = 0;
   2168 	}
   2169 
   2170 	(void)looutput(ifp, copym, (struct sockaddr *)dst, NULL);
   2171 }
   2172 
   2173 /*
   2174  * Chop IPv6 header off from the payload.
   2175  */
   2176 static int
   2177 ip6_splithdr(m, exthdrs)
   2178 	struct mbuf *m;
   2179 	struct ip6_exthdrs *exthdrs;
   2180 {
   2181 	struct mbuf *mh;
   2182 	struct ip6_hdr *ip6;
   2183 
   2184 	ip6 = mtod(m, struct ip6_hdr *);
   2185 	if (m->m_len > sizeof(*ip6)) {
   2186 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
   2187 		if (mh == 0) {
   2188 			m_freem(m);
   2189 			return ENOBUFS;
   2190 		}
   2191 		M_COPY_PKTHDR(mh, m);
   2192 		MH_ALIGN(mh, sizeof(*ip6));
   2193 		m->m_flags &= ~M_PKTHDR;
   2194 		m->m_len -= sizeof(*ip6);
   2195 		m->m_data += sizeof(*ip6);
   2196 		mh->m_next = m;
   2197 		m = mh;
   2198 		m->m_len = sizeof(*ip6);
   2199 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
   2200 	}
   2201 	exthdrs->ip6e_ip6 = m;
   2202 	return 0;
   2203 }
   2204 
   2205 /*
   2206  * Compute IPv6 extension header length.
   2207  */
   2208 int
   2209 ip6_optlen(in6p)
   2210 	struct in6pcb *in6p;
   2211 {
   2212 	int len;
   2213 
   2214 	if (!in6p->in6p_outputopts)
   2215 		return 0;
   2216 
   2217 	len = 0;
   2218 #define elen(x) \
   2219     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
   2220 
   2221 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
   2222 	len += elen(in6p->in6p_outputopts->ip6po_dest1);
   2223 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
   2224 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
   2225 	return len;
   2226 #undef elen
   2227 }
   2228