Home | History | Annotate | Line # | Download | only in netinet6
ip6_output.c revision 1.52
      1 /*	$NetBSD: ip6_output.c,v 1.52 2002/06/07 17:13:56 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 1.52 2002/06/07 17:13:56 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  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
    142  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
    143  * which is rt_rmx.rmx_mtu.
    144  */
    145 int
    146 ip6_output(m0, opt, ro, flags, im6o, ifpp)
    147 	struct mbuf *m0;
    148 	struct ip6_pktopts *opt;
    149 	struct route_in6 *ro;
    150 	int flags;
    151 	struct ip6_moptions *im6o;
    152 	struct ifnet **ifpp;		/* XXX: just for statistics */
    153 {
    154 	struct ip6_hdr *ip6, *mhip6;
    155 	struct ifnet *ifp, *origifp;
    156 	struct mbuf *m = m0;
    157 	int hlen, tlen, len, off;
    158 	struct route_in6 ip6route;
    159 	struct sockaddr_in6 *dst;
    160 	int error = 0;
    161 	struct in6_ifaddr *ia;
    162 	u_long mtu;
    163 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
    164 	struct ip6_exthdrs exthdrs;
    165 	struct in6_addr finaldst;
    166 	struct route_in6 *ro_pmtu = NULL;
    167 	int hdrsplit = 0;
    168 	int needipsec = 0;
    169 #ifdef IPSEC
    170 	int needipsectun = 0;
    171 	struct socket *so;
    172 	struct secpolicy *sp = NULL;
    173 
    174 	/* for AH processing. stupid to have "socket" variable in IP layer... */
    175 	so = ipsec_getsocket(m);
    176 	(void)ipsec_setsocket(m, NULL);
    177 	ip6 = mtod(m, struct ip6_hdr *);
    178 #endif /* IPSEC */
    179 
    180 #define MAKE_EXTHDR(hp, mp)						\
    181     do {								\
    182 	if (hp) {							\
    183 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
    184 		error = ip6_copyexthdr((mp), (caddr_t)(hp), 		\
    185 		    ((eh)->ip6e_len + 1) << 3);				\
    186 		if (error)						\
    187 			goto freehdrs;					\
    188 	}								\
    189     } while (0)
    190 
    191 	bzero(&exthdrs, sizeof(exthdrs));
    192 	if (opt) {
    193 		/* Hop-by-Hop options header */
    194 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
    195 		/* Destination options header(1st part) */
    196 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
    197 		/* Routing header */
    198 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
    199 		/* Destination options header(2nd part) */
    200 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
    201 	}
    202 
    203 #ifdef IPSEC
    204 	/* get a security policy for this packet */
    205 	if (so == NULL)
    206 		sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
    207 	else
    208 		sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
    209 
    210 	if (sp == NULL) {
    211 		ipsec6stat.out_inval++;
    212 		goto freehdrs;
    213 	}
    214 
    215 	error = 0;
    216 
    217 	/* check policy */
    218 	switch (sp->policy) {
    219 	case IPSEC_POLICY_DISCARD:
    220 		/*
    221 		 * This packet is just discarded.
    222 		 */
    223 		ipsec6stat.out_polvio++;
    224 		goto freehdrs;
    225 
    226 	case IPSEC_POLICY_BYPASS:
    227 	case IPSEC_POLICY_NONE:
    228 		/* no need to do IPsec. */
    229 		needipsec = 0;
    230 		break;
    231 
    232 	case IPSEC_POLICY_IPSEC:
    233 		if (sp->req == NULL) {
    234 			/* XXX should be panic ? */
    235 			printf("ip6_output: No IPsec request specified.\n");
    236 			error = EINVAL;
    237 			goto freehdrs;
    238 		}
    239 		needipsec = 1;
    240 		break;
    241 
    242 	case IPSEC_POLICY_ENTRUST:
    243 	default:
    244 		printf("ip6_output: Invalid policy found. %d\n", sp->policy);
    245 	}
    246 #endif /* IPSEC */
    247 
    248 	/*
    249 	 * Calculate the total length of the extension header chain.
    250 	 * Keep the length of the unfragmentable part for fragmentation.
    251 	 */
    252 	optlen = 0;
    253 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
    254 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
    255 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
    256 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
    257 	/* NOTE: we don't add AH/ESP length here. do that later. */
    258 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
    259 
    260 	/*
    261 	 * If we need IPsec, or there is at least one extension header,
    262 	 * separate IP6 header from the payload.
    263 	 */
    264 	if ((needipsec || optlen) && !hdrsplit) {
    265 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
    266 			m = NULL;
    267 			goto freehdrs;
    268 		}
    269 		m = exthdrs.ip6e_ip6;
    270 		hdrsplit++;
    271 	}
    272 
    273 	/* adjust pointer */
    274 	ip6 = mtod(m, struct ip6_hdr *);
    275 
    276 	/* adjust mbuf packet header length */
    277 	m->m_pkthdr.len += optlen;
    278 	plen = m->m_pkthdr.len - sizeof(*ip6);
    279 
    280 	/* If this is a jumbo payload, insert a jumbo payload option. */
    281 	if (plen > IPV6_MAXPACKET) {
    282 		if (!hdrsplit) {
    283 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
    284 				m = NULL;
    285 				goto freehdrs;
    286 			}
    287 			m = exthdrs.ip6e_ip6;
    288 			hdrsplit++;
    289 		}
    290 		/* adjust pointer */
    291 		ip6 = mtod(m, struct ip6_hdr *);
    292 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
    293 			goto freehdrs;
    294 		ip6->ip6_plen = 0;
    295 	} else
    296 		ip6->ip6_plen = htons(plen);
    297 
    298 	/*
    299 	 * Concatenate headers and fill in next header fields.
    300 	 * Here we have, on "m"
    301 	 *	IPv6 payload
    302 	 * and we insert headers accordingly.  Finally, we should be getting:
    303 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
    304 	 *
    305 	 * during the header composing process, "m" points to IPv6 header.
    306 	 * "mprev" points to an extension header prior to esp.
    307 	 */
    308 	{
    309 		u_char *nexthdrp = &ip6->ip6_nxt;
    310 		struct mbuf *mprev = m;
    311 
    312 		/*
    313 		 * we treat dest2 specially.  this makes IPsec processing
    314 		 * much easier.
    315 		 *
    316 		 * result: IPv6 dest2 payload
    317 		 * m and mprev will point to IPv6 header.
    318 		 */
    319 		if (exthdrs.ip6e_dest2) {
    320 			if (!hdrsplit)
    321 				panic("assumption failed: hdr not split");
    322 			exthdrs.ip6e_dest2->m_next = m->m_next;
    323 			m->m_next = exthdrs.ip6e_dest2;
    324 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
    325 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
    326 		}
    327 
    328 #define MAKE_CHAIN(m, mp, p, i)\
    329     do {\
    330 	if (m) {\
    331 		if (!hdrsplit) \
    332 			panic("assumption failed: hdr not split"); \
    333 		*mtod((m), u_char *) = *(p);\
    334 		*(p) = (i);\
    335 		p = mtod((m), u_char *);\
    336 		(m)->m_next = (mp)->m_next;\
    337 		(mp)->m_next = (m);\
    338 		(mp) = (m);\
    339 	}\
    340     } while (0)
    341 		/*
    342 		 * result: IPv6 hbh dest1 rthdr dest2 payload
    343 		 * m will point to IPv6 header.  mprev will point to the
    344 		 * extension header prior to dest2 (rthdr in the above case).
    345 		 */
    346 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
    347 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
    348 		    IPPROTO_DSTOPTS);
    349 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
    350 		    IPPROTO_ROUTING);
    351 
    352 #ifdef IPSEC
    353 		if (!needipsec)
    354 			goto skip_ipsec2;
    355 
    356 		/*
    357 		 * pointers after IPsec headers are not valid any more.
    358 		 * other pointers need a great care too.
    359 		 * (IPsec routines should not mangle mbufs prior to AH/ESP)
    360 		 */
    361 		exthdrs.ip6e_dest2 = NULL;
    362 
    363 	    {
    364 		struct ip6_rthdr *rh = NULL;
    365 		int segleft_org = 0;
    366 		struct ipsec_output_state state;
    367 
    368 		if (exthdrs.ip6e_rthdr) {
    369 			rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
    370 			segleft_org = rh->ip6r_segleft;
    371 			rh->ip6r_segleft = 0;
    372 		}
    373 
    374 		bzero(&state, sizeof(state));
    375 		state.m = m;
    376 		error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
    377 			&needipsectun);
    378 		m = state.m;
    379 		if (error) {
    380 			/* mbuf is already reclaimed in ipsec6_output_trans. */
    381 			m = NULL;
    382 			switch (error) {
    383 			case EHOSTUNREACH:
    384 			case ENETUNREACH:
    385 			case EMSGSIZE:
    386 			case ENOBUFS:
    387 			case ENOMEM:
    388 				break;
    389 			default:
    390 				printf("ip6_output (ipsec): error code %d\n", error);
    391 				/* FALLTHROUGH */
    392 			case ENOENT:
    393 				/* don't show these error codes to the user */
    394 				error = 0;
    395 				break;
    396 			}
    397 			goto bad;
    398 		}
    399 		if (exthdrs.ip6e_rthdr) {
    400 			/* ah6_output doesn't modify mbuf chain */
    401 			rh->ip6r_segleft = segleft_org;
    402 		}
    403 	    }
    404 skip_ipsec2:;
    405 #endif
    406 	}
    407 
    408 	/*
    409 	 * If there is a routing header, replace destination address field
    410 	 * with the first hop of the routing header.
    411 	 */
    412 	if (exthdrs.ip6e_rthdr) {
    413 		struct ip6_rthdr *rh;
    414 		struct ip6_rthdr0 *rh0;
    415 
    416 		rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
    417 		    struct ip6_rthdr *));
    418 		finaldst = ip6->ip6_dst;
    419 		switch (rh->ip6r_type) {
    420 		case IPV6_RTHDR_TYPE_0:
    421 			 rh0 = (struct ip6_rthdr0 *)rh;
    422 			 ip6->ip6_dst = rh0->ip6r0_addr[0];
    423 			 bcopy((caddr_t)&rh0->ip6r0_addr[1],
    424 			     (caddr_t)&rh0->ip6r0_addr[0],
    425 			     sizeof(struct in6_addr) * (rh0->ip6r0_segleft - 1));
    426 			 rh0->ip6r0_addr[rh0->ip6r0_segleft - 1] = finaldst;
    427 			 break;
    428 		default:	/* is it possible? */
    429 			 error = EINVAL;
    430 			 goto bad;
    431 		}
    432 	}
    433 
    434 	/* Source address validation */
    435 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
    436 	    (flags & IPV6_DADOUTPUT) == 0) {
    437 		error = EOPNOTSUPP;
    438 		ip6stat.ip6s_badscope++;
    439 		goto bad;
    440 	}
    441 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
    442 		error = EOPNOTSUPP;
    443 		ip6stat.ip6s_badscope++;
    444 		goto bad;
    445 	}
    446 
    447 	ip6stat.ip6s_localout++;
    448 
    449 	/*
    450 	 * Route packet.
    451 	 */
    452 	if (ro == 0) {
    453 		ro = &ip6route;
    454 		bzero((caddr_t)ro, sizeof(*ro));
    455 	}
    456 	ro_pmtu = ro;
    457 	if (opt && opt->ip6po_rthdr)
    458 		ro = &opt->ip6po_route;
    459 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
    460 	/*
    461 	 * If there is a cached route,
    462 	 * check that it is to the same destination
    463 	 * and is still up. If not, free it and try again.
    464 	 */
    465 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
    466 	    dst->sin6_family != AF_INET6 ||
    467 	    !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
    468 		RTFREE(ro->ro_rt);
    469 		ro->ro_rt = (struct rtentry *)0;
    470 	}
    471 	if (ro->ro_rt == 0) {
    472 		bzero(dst, sizeof(*dst));
    473 		dst->sin6_family = AF_INET6;
    474 		dst->sin6_len = sizeof(struct sockaddr_in6);
    475 		dst->sin6_addr = ip6->ip6_dst;
    476 	}
    477 #ifdef IPSEC
    478 	if (needipsec && needipsectun) {
    479 		struct ipsec_output_state state;
    480 
    481 		/*
    482 		 * All the extension headers will become inaccessible
    483 		 * (since they can be encrypted).
    484 		 * Don't panic, we need no more updates to extension headers
    485 		 * on inner IPv6 packet (since they are now encapsulated).
    486 		 *
    487 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
    488 		 */
    489 		bzero(&exthdrs, sizeof(exthdrs));
    490 		exthdrs.ip6e_ip6 = m;
    491 
    492 		bzero(&state, sizeof(state));
    493 		state.m = m;
    494 		state.ro = (struct route *)ro;
    495 		state.dst = (struct sockaddr *)dst;
    496 
    497 		error = ipsec6_output_tunnel(&state, sp, flags);
    498 
    499 		m = state.m;
    500 		ro = (struct route_in6 *)state.ro;
    501 		dst = (struct sockaddr_in6 *)state.dst;
    502 		if (error) {
    503 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
    504 			m0 = m = NULL;
    505 			m = NULL;
    506 			switch (error) {
    507 			case EHOSTUNREACH:
    508 			case ENETUNREACH:
    509 			case EMSGSIZE:
    510 			case ENOBUFS:
    511 			case ENOMEM:
    512 				break;
    513 			default:
    514 				printf("ip6_output (ipsec): error code %d\n", error);
    515 				/* FALLTHROUGH */
    516 			case ENOENT:
    517 				/* don't show these error codes to the user */
    518 				error = 0;
    519 				break;
    520 			}
    521 			goto bad;
    522 		}
    523 
    524 		exthdrs.ip6e_ip6 = m;
    525 	}
    526 #endif /* IPSEC */
    527 
    528 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
    529 		/* Unicast */
    530 
    531 #define ifatoia6(ifa)	((struct in6_ifaddr *)(ifa))
    532 #define sin6tosa(sin6)	((struct sockaddr *)(sin6))
    533 		/* xxx
    534 		 * interface selection comes here
    535 		 * if an interface is specified from an upper layer,
    536 		 * ifp must point it.
    537 		 */
    538 		if (ro->ro_rt == 0) {
    539 			/*
    540 			 * non-bsdi always clone routes, if parent is
    541 			 * PRF_CLONING.
    542 			 */
    543 			rtalloc((struct route *)ro);
    544 		}
    545 		if (ro->ro_rt == 0) {
    546 			ip6stat.ip6s_noroute++;
    547 			error = EHOSTUNREACH;
    548 			/* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
    549 			goto bad;
    550 		}
    551 		ia = ifatoia6(ro->ro_rt->rt_ifa);
    552 		ifp = ro->ro_rt->rt_ifp;
    553 		ro->ro_rt->rt_use++;
    554 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
    555 			dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
    556 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
    557 
    558 		in6_ifstat_inc(ifp, ifs6_out_request);
    559 
    560 		/*
    561 		 * Check if the outgoing interface conflicts with
    562 		 * the interface specified by ifi6_ifindex (if specified).
    563 		 * Note that loopback interface is always okay.
    564 		 * (this may happen when we are sending a packet to one of
    565 		 *  our own addresses.)
    566 		 */
    567 		if (opt && opt->ip6po_pktinfo
    568 		 && opt->ip6po_pktinfo->ipi6_ifindex) {
    569 			if (!(ifp->if_flags & IFF_LOOPBACK)
    570 			 && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
    571 				ip6stat.ip6s_noroute++;
    572 				in6_ifstat_inc(ifp, ifs6_out_discard);
    573 				error = EHOSTUNREACH;
    574 				goto bad;
    575 			}
    576 		}
    577 
    578 		if (opt && opt->ip6po_hlim != -1)
    579 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
    580 	} else {
    581 		/* Multicast */
    582 		struct	in6_multi *in6m;
    583 
    584 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
    585 
    586 		/*
    587 		 * See if the caller provided any multicast options
    588 		 */
    589 		ifp = NULL;
    590 		if (im6o != NULL) {
    591 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
    592 			if (im6o->im6o_multicast_ifp != NULL)
    593 				ifp = im6o->im6o_multicast_ifp;
    594 		} else
    595 			ip6->ip6_hlim = ip6_defmcasthlim;
    596 
    597 		/*
    598 		 * See if the caller provided the outgoing interface
    599 		 * as an ancillary data.
    600 		 * Boundary check for ifindex is assumed to be already done.
    601 		 */
    602 		if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
    603 			ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex];
    604 
    605 		/*
    606 		 * If the destination is a node-local scope multicast,
    607 		 * the packet should be loop-backed only.
    608 		 */
    609 		if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
    610 			/*
    611 			 * If the outgoing interface is already specified,
    612 			 * it should be a loopback interface.
    613 			 */
    614 			if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) {
    615 				ip6stat.ip6s_badscope++;
    616 				error = ENETUNREACH; /* XXX: better error? */
    617 				/* XXX correct ifp? */
    618 				in6_ifstat_inc(ifp, ifs6_out_discard);
    619 				goto bad;
    620 			} else {
    621 				ifp = &loif[0];
    622 			}
    623 		}
    624 
    625 		if (opt && opt->ip6po_hlim != -1)
    626 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
    627 
    628 		/*
    629 		 * If caller did not provide an interface lookup a
    630 		 * default in the routing table.  This is either a
    631 		 * default for the speicfied group (i.e. a host
    632 		 * route), or a multicast default (a route for the
    633 		 * ``net'' ff00::/8).
    634 		 */
    635 		if (ifp == NULL) {
    636 			if (ro->ro_rt == 0) {
    637 				ro->ro_rt = rtalloc1((struct sockaddr *)
    638 				    &ro->ro_dst, 0);
    639 			}
    640 			if (ro->ro_rt == 0) {
    641 				ip6stat.ip6s_noroute++;
    642 				error = EHOSTUNREACH;
    643 				/* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
    644 				goto bad;
    645 			}
    646 			ia = ifatoia6(ro->ro_rt->rt_ifa);
    647 			ifp = ro->ro_rt->rt_ifp;
    648 			ro->ro_rt->rt_use++;
    649 		}
    650 
    651 		if ((flags & IPV6_FORWARDING) == 0)
    652 			in6_ifstat_inc(ifp, ifs6_out_request);
    653 		in6_ifstat_inc(ifp, ifs6_out_mcast);
    654 
    655 		/*
    656 		 * Confirm that the outgoing interface supports multicast.
    657 		 */
    658 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
    659 			ip6stat.ip6s_noroute++;
    660 			in6_ifstat_inc(ifp, ifs6_out_discard);
    661 			error = ENETUNREACH;
    662 			goto bad;
    663 		}
    664 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
    665 		if (in6m != NULL &&
    666 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
    667 			/*
    668 			 * If we belong to the destination multicast group
    669 			 * on the outgoing interface, and the caller did not
    670 			 * forbid loopback, loop back a copy.
    671 			 */
    672 			ip6_mloopback(ifp, m, dst);
    673 		} else {
    674 			/*
    675 			 * If we are acting as a multicast router, perform
    676 			 * multicast forwarding as if the packet had just
    677 			 * arrived on the interface to which we are about
    678 			 * to send.  The multicast forwarding function
    679 			 * recursively calls this function, using the
    680 			 * IPV6_FORWARDING flag to prevent infinite recursion.
    681 			 *
    682 			 * Multicasts that are looped back by ip6_mloopback(),
    683 			 * above, will be forwarded by the ip6_input() routine,
    684 			 * if necessary.
    685 			 */
    686 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
    687 				if (ip6_mforward(ip6, ifp, m) != 0) {
    688 					m_freem(m);
    689 					goto done;
    690 				}
    691 			}
    692 		}
    693 		/*
    694 		 * Multicasts with a hoplimit of zero may be looped back,
    695 		 * above, but must not be transmitted on a network.
    696 		 * Also, multicasts addressed to the loopback interface
    697 		 * are not sent -- the above call to ip6_mloopback() will
    698 		 * loop back a copy if this host actually belongs to the
    699 		 * destination group on the loopback interface.
    700 		 */
    701 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
    702 			m_freem(m);
    703 			goto done;
    704 		}
    705 	}
    706 
    707 	/*
    708 	 * Fill the outgoing inteface to tell the upper layer
    709 	 * to increment per-interface statistics.
    710 	 */
    711 	if (ifpp)
    712 		*ifpp = ifp;
    713 
    714 	/* Determine path MTU. */
    715 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu)) != 0)
    716 		goto bad;
    717 
    718 	/*
    719 	 * The caller of this function may specify to use the minimum MTU
    720 	 * in some cases.
    721 	 */
    722 	if (mtu > IPV6_MMTU) {
    723 		if ((flags & IPV6_MINMTU))
    724 			mtu = IPV6_MMTU;
    725 	}
    726 
    727 	/* Fake scoped addresses */
    728 	if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
    729 		/*
    730 		 * If source or destination address is a scoped address, and
    731 		 * the packet is going to be sent to a loopback interface,
    732 		 * we should keep the original interface.
    733 		 */
    734 
    735 		/*
    736 		 * XXX: this is a very experimental and temporary solution.
    737 		 * We eventually have sockaddr_in6 and use the sin6_scope_id
    738 		 * field of the structure here.
    739 		 * We rely on the consistency between two scope zone ids
    740 		 * of source add destination, which should already be assured
    741 		 * Larger scopes than link will be supported in the near
    742 		 * future.
    743 		 */
    744 		origifp = NULL;
    745 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
    746 			origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
    747 		else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
    748 			origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
    749 		/*
    750 		 * XXX: origifp can be NULL even in those two cases above.
    751 		 * For example, if we remove the (only) link-local address
    752 		 * from the loopback interface, and try to send a link-local
    753 		 * address without link-id information.  Then the source
    754 		 * address is ::1, and the destination address is the
    755 		 * link-local address with its s6_addr16[1] being zero.
    756 		 * What is worse, if the packet goes to the loopback interface
    757 		 * by a default rejected route, the null pointer would be
    758 		 * passed to looutput, and the kernel would hang.
    759 		 * The following last resort would prevent such disaster.
    760 		 */
    761 		if (origifp == NULL)
    762 			origifp = ifp;
    763 	} else
    764 		origifp = ifp;
    765 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
    766 		ip6->ip6_src.s6_addr16[1] = 0;
    767 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
    768 		ip6->ip6_dst.s6_addr16[1] = 0;
    769 
    770 	/*
    771 	 * If the outgoing packet contains a hop-by-hop options header,
    772 	 * it must be examined and processed even by the source node.
    773 	 * (RFC 2460, section 4.)
    774 	 */
    775 	if (exthdrs.ip6e_hbh) {
    776 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
    777 		u_int32_t dummy1; /* XXX unused */
    778 		u_int32_t dummy2; /* XXX unused */
    779 
    780 		/*
    781 		 *  XXX: if we have to send an ICMPv6 error to the sender,
    782 		 *       we need the M_LOOP flag since icmp6_error() expects
    783 		 *       the IPv6 and the hop-by-hop options header are
    784 		 *       continuous unless the flag is set.
    785 		 */
    786 		m->m_flags |= M_LOOP;
    787 		m->m_pkthdr.rcvif = ifp;
    788 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
    789 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
    790 		    &dummy1, &dummy2) < 0) {
    791 			/* m was already freed at this point */
    792 			error = EINVAL;/* better error? */
    793 			goto done;
    794 		}
    795 		m->m_flags &= ~M_LOOP; /* XXX */
    796 		m->m_pkthdr.rcvif = NULL;
    797 	}
    798 
    799 #ifdef PFIL_HOOKS
    800 	/*
    801 	 * Run through list of hooks for output packets.
    802 	 */
    803 	if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0)
    804 		goto done;
    805 	if (m == NULL)
    806 		goto done;
    807 	ip6 = mtod(m, struct ip6_hdr *);
    808 #endif /* PFIL_HOOKS */
    809 	/*
    810 	 * Send the packet to the outgoing interface.
    811 	 * If necessary, do IPv6 fragmentation before sending.
    812 	 */
    813 	tlen = m->m_pkthdr.len;
    814 	if (tlen <= mtu) {
    815 #ifdef IFA_STATS
    816 		struct in6_ifaddr *ia6;
    817 		ip6 = mtod(m, struct ip6_hdr *);
    818 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
    819 		if (ia6) {
    820 			/* Record statistics for this interface address. */
    821 			ia6->ia_ifa.ifa_data.ifad_outbytes +=
    822 				m->m_pkthdr.len;
    823 		}
    824 #endif
    825 #ifdef IPSEC
    826 		/* clean ipsec history once it goes out of the node */
    827 		ipsec_delaux(m);
    828 #endif
    829 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
    830 		goto done;
    831 	} else if (mtu < IPV6_MMTU) {
    832 		/*
    833 		 * note that path MTU is never less than IPV6_MMTU
    834 		 * (see icmp6_input).
    835 		 */
    836 		error = EMSGSIZE;
    837 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
    838 		goto bad;
    839 	} else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
    840 		error = EMSGSIZE;
    841 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
    842 		goto bad;
    843 	} else {
    844 		struct mbuf **mnext, *m_frgpart;
    845 		struct ip6_frag *ip6f;
    846 		u_int32_t id = htonl(ip6_id++);
    847 		u_char nextproto;
    848 
    849 		/*
    850 		 * Too large for the destination or interface;
    851 		 * fragment if possible.
    852 		 * Must be able to put at least 8 bytes per fragment.
    853 		 */
    854 		hlen = unfragpartlen;
    855 		if (mtu > IPV6_MAXPACKET)
    856 			mtu = IPV6_MAXPACKET;
    857 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
    858 		if (len < 8) {
    859 			error = EMSGSIZE;
    860 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
    861 			goto bad;
    862 		}
    863 
    864 		mnext = &m->m_nextpkt;
    865 
    866 		/*
    867 		 * Change the next header field of the last header in the
    868 		 * unfragmentable part.
    869 		 */
    870 		if (exthdrs.ip6e_rthdr) {
    871 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
    872 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
    873 		} else if (exthdrs.ip6e_dest1) {
    874 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
    875 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
    876 		} else if (exthdrs.ip6e_hbh) {
    877 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
    878 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
    879 		} else {
    880 			nextproto = ip6->ip6_nxt;
    881 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
    882 		}
    883 
    884 		/*
    885 		 * Loop through length of segment after first fragment,
    886 		 * make new header and copy data of each part and link onto
    887 		 * chain.
    888 		 */
    889 		m0 = m;
    890 		for (off = hlen; off < tlen; off += len) {
    891 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
    892 			if (!m) {
    893 				error = ENOBUFS;
    894 				ip6stat.ip6s_odropped++;
    895 				goto sendorfree;
    896 			}
    897 			m->m_flags = m0->m_flags & M_COPYFLAGS;
    898 			*mnext = m;
    899 			mnext = &m->m_nextpkt;
    900 			m->m_data += max_linkhdr;
    901 			mhip6 = mtod(m, struct ip6_hdr *);
    902 			*mhip6 = *ip6;
    903 			m->m_len = sizeof(*mhip6);
    904 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
    905 			if (error) {
    906 				ip6stat.ip6s_odropped++;
    907 				goto sendorfree;
    908 			}
    909 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
    910 			if (off + len >= tlen)
    911 				len = tlen - off;
    912 			else
    913 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
    914 			mhip6->ip6_plen = htons((u_short)(len + hlen +
    915 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
    916 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
    917 				error = ENOBUFS;
    918 				ip6stat.ip6s_odropped++;
    919 				goto sendorfree;
    920 			}
    921 			m_cat(m, m_frgpart);
    922 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
    923 			m->m_pkthdr.rcvif = (struct ifnet *)0;
    924 			ip6f->ip6f_reserved = 0;
    925 			ip6f->ip6f_ident = id;
    926 			ip6f->ip6f_nxt = nextproto;
    927 			ip6stat.ip6s_ofragments++;
    928 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
    929 		}
    930 
    931 		in6_ifstat_inc(ifp, ifs6_out_fragok);
    932 	}
    933 
    934 	/*
    935 	 * Remove leading garbages.
    936 	 */
    937 sendorfree:
    938 	m = m0->m_nextpkt;
    939 	m0->m_nextpkt = 0;
    940 	m_freem(m0);
    941 	for (m0 = m; m; m = m0) {
    942 		m0 = m->m_nextpkt;
    943 		m->m_nextpkt = 0;
    944 		if (error == 0) {
    945 #ifdef IFA_STATS
    946 			struct in6_ifaddr *ia6;
    947 			ip6 = mtod(m, struct ip6_hdr *);
    948 			ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
    949 			if (ia6) {
    950 				/*
    951 				 * Record statistics for this interface
    952 				 * address.
    953 				 */
    954 				ia6->ia_ifa.ifa_data.ifad_outbytes +=
    955 					m->m_pkthdr.len;
    956 			}
    957 #endif
    958 #ifdef IPSEC
    959 			/* clean ipsec history once it goes out of the node */
    960 			ipsec_delaux(m);
    961 #endif
    962 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
    963 		} else
    964 			m_freem(m);
    965 	}
    966 
    967 	if (error == 0)
    968 		ip6stat.ip6s_fragmented++;
    969 
    970 done:
    971 	if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
    972 		RTFREE(ro->ro_rt);
    973 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
    974 		RTFREE(ro_pmtu->ro_rt);
    975 	}
    976 
    977 #ifdef IPSEC
    978 	if (sp != NULL)
    979 		key_freesp(sp);
    980 #endif /* IPSEC */
    981 
    982 	return(error);
    983 
    984 freehdrs:
    985 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
    986 	m_freem(exthdrs.ip6e_dest1);
    987 	m_freem(exthdrs.ip6e_rthdr);
    988 	m_freem(exthdrs.ip6e_dest2);
    989 	/* FALLTHROUGH */
    990 bad:
    991 	m_freem(m);
    992 	goto done;
    993 }
    994 
    995 static int
    996 ip6_copyexthdr(mp, hdr, hlen)
    997 	struct mbuf **mp;
    998 	caddr_t hdr;
    999 	int hlen;
   1000 {
   1001 	struct mbuf *m;
   1002 
   1003 	if (hlen > MCLBYTES)
   1004 		return(ENOBUFS); /* XXX */
   1005 
   1006 	MGET(m, M_DONTWAIT, MT_DATA);
   1007 	if (!m)
   1008 		return(ENOBUFS);
   1009 
   1010 	if (hlen > MLEN) {
   1011 		MCLGET(m, M_DONTWAIT);
   1012 		if ((m->m_flags & M_EXT) == 0) {
   1013 			m_free(m);
   1014 			return(ENOBUFS);
   1015 		}
   1016 	}
   1017 	m->m_len = hlen;
   1018 	if (hdr)
   1019 		bcopy(hdr, mtod(m, caddr_t), hlen);
   1020 
   1021 	*mp = m;
   1022 	return(0);
   1023 }
   1024 
   1025 /*
   1026  * Insert jumbo payload option.
   1027  */
   1028 static int
   1029 ip6_insert_jumboopt(exthdrs, plen)
   1030 	struct ip6_exthdrs *exthdrs;
   1031 	u_int32_t plen;
   1032 {
   1033 	struct mbuf *mopt;
   1034 	u_char *optbuf;
   1035 	u_int32_t v;
   1036 
   1037 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
   1038 
   1039 	/*
   1040 	 * If there is no hop-by-hop options header, allocate new one.
   1041 	 * If there is one but it doesn't have enough space to store the
   1042 	 * jumbo payload option, allocate a cluster to store the whole options.
   1043 	 * Otherwise, use it to store the options.
   1044 	 */
   1045 	if (exthdrs->ip6e_hbh == 0) {
   1046 		MGET(mopt, M_DONTWAIT, MT_DATA);
   1047 		if (mopt == 0)
   1048 			return(ENOBUFS);
   1049 		mopt->m_len = JUMBOOPTLEN;
   1050 		optbuf = mtod(mopt, u_char *);
   1051 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
   1052 		exthdrs->ip6e_hbh = mopt;
   1053 	} else {
   1054 		struct ip6_hbh *hbh;
   1055 
   1056 		mopt = exthdrs->ip6e_hbh;
   1057 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
   1058 			/*
   1059 			 * XXX assumption:
   1060 			 * - exthdrs->ip6e_hbh is not referenced from places
   1061 			 *   other than exthdrs.
   1062 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
   1063 			 */
   1064 			int oldoptlen = mopt->m_len;
   1065 			struct mbuf *n;
   1066 
   1067 			/*
   1068 			 * XXX: give up if the whole (new) hbh header does
   1069 			 * not fit even in an mbuf cluster.
   1070 			 */
   1071 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
   1072 				return(ENOBUFS);
   1073 
   1074 			/*
   1075 			 * As a consequence, we must always prepare a cluster
   1076 			 * at this point.
   1077 			 */
   1078 			MGET(n, M_DONTWAIT, MT_DATA);
   1079 			if (n) {
   1080 				MCLGET(n, M_DONTWAIT);
   1081 				if ((n->m_flags & M_EXT) == 0) {
   1082 					m_freem(n);
   1083 					n = NULL;
   1084 				}
   1085 			}
   1086 			if (!n)
   1087 				return(ENOBUFS);
   1088 			n->m_len = oldoptlen + JUMBOOPTLEN;
   1089 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
   1090 			      oldoptlen);
   1091 			optbuf = mtod(n, caddr_t) + oldoptlen;
   1092 			m_freem(mopt);
   1093 			mopt = exthdrs->ip6e_hbh = n;
   1094 		} else {
   1095 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
   1096 			mopt->m_len += JUMBOOPTLEN;
   1097 		}
   1098 		optbuf[0] = IP6OPT_PADN;
   1099 		optbuf[1] = 1;
   1100 
   1101 		/*
   1102 		 * Adjust the header length according to the pad and
   1103 		 * the jumbo payload option.
   1104 		 */
   1105 		hbh = mtod(mopt, struct ip6_hbh *);
   1106 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
   1107 	}
   1108 
   1109 	/* fill in the option. */
   1110 	optbuf[2] = IP6OPT_JUMBO;
   1111 	optbuf[3] = 4;
   1112 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
   1113 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
   1114 
   1115 	/* finally, adjust the packet header length */
   1116 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
   1117 
   1118 	return(0);
   1119 #undef JUMBOOPTLEN
   1120 }
   1121 
   1122 /*
   1123  * Insert fragment header and copy unfragmentable header portions.
   1124  */
   1125 static int
   1126 ip6_insertfraghdr(m0, m, hlen, frghdrp)
   1127 	struct mbuf *m0, *m;
   1128 	int hlen;
   1129 	struct ip6_frag **frghdrp;
   1130 {
   1131 	struct mbuf *n, *mlast;
   1132 
   1133 	if (hlen > sizeof(struct ip6_hdr)) {
   1134 		n = m_copym(m0, sizeof(struct ip6_hdr),
   1135 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
   1136 		if (n == 0)
   1137 			return(ENOBUFS);
   1138 		m->m_next = n;
   1139 	} else
   1140 		n = m;
   1141 
   1142 	/* Search for the last mbuf of unfragmentable part. */
   1143 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
   1144 		;
   1145 
   1146 	if ((mlast->m_flags & M_EXT) == 0 &&
   1147 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
   1148 		/* use the trailing space of the last mbuf for the fragment hdr */
   1149 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
   1150 		    mlast->m_len);
   1151 		mlast->m_len += sizeof(struct ip6_frag);
   1152 		m->m_pkthdr.len += sizeof(struct ip6_frag);
   1153 	} else {
   1154 		/* allocate a new mbuf for the fragment header */
   1155 		struct mbuf *mfrg;
   1156 
   1157 		MGET(mfrg, M_DONTWAIT, MT_DATA);
   1158 		if (mfrg == 0)
   1159 			return(ENOBUFS);
   1160 		mfrg->m_len = sizeof(struct ip6_frag);
   1161 		*frghdrp = mtod(mfrg, struct ip6_frag *);
   1162 		mlast->m_next = mfrg;
   1163 	}
   1164 
   1165 	return(0);
   1166 }
   1167 
   1168 static int
   1169 ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup)
   1170 	struct route_in6 *ro_pmtu, *ro;
   1171 	struct ifnet *ifp;
   1172 	struct in6_addr *dst;
   1173 	u_long *mtup;
   1174 {
   1175 	u_int32_t mtu = 0;
   1176 	int error = 0;
   1177 
   1178 	if (ro_pmtu != ro) {
   1179 		/* The first hop and the final destination may differ. */
   1180 		struct sockaddr_in6 *sa6_dst =
   1181 		    (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
   1182 		if (ro_pmtu->ro_rt &&
   1183 		    ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
   1184 		     !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
   1185 			RTFREE(ro_pmtu->ro_rt);
   1186 			ro_pmtu->ro_rt = (struct rtentry *)0;
   1187 		}
   1188 		if (ro_pmtu->ro_rt == 0) {
   1189 			bzero(sa6_dst, sizeof(*sa6_dst));
   1190 			sa6_dst->sin6_family = AF_INET6;
   1191 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
   1192 			sa6_dst->sin6_addr = *dst;
   1193 
   1194 			rtalloc((struct route *)ro_pmtu);
   1195 		}
   1196 	}
   1197 	if (ro_pmtu->ro_rt) {
   1198 		u_int32_t ifmtu;
   1199 
   1200 		if (ifp == NULL)
   1201 			ifp = ro_pmtu->ro_rt->rt_ifp;
   1202 		ifmtu = IN6_LINKMTU(ifp);
   1203 		mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
   1204 		if (mtu == 0)
   1205 			mtu = ifmtu;
   1206 		else if (mtu > ifmtu) {
   1207 			/*
   1208 			 * The MTU on the route is larger than the MTU on
   1209 			 * the interface!  This shouldn't happen, unless the
   1210 			 * MTU of the interface has been changed after the
   1211 			 * interface was brought up.  Change the MTU in the
   1212 			 * route to match the interface MTU (as long as the
   1213 			 * field isn't locked).
   1214 			 *
   1215 			 * if MTU on the route is 0, we need to fix the MTU.
   1216 			 * this case happens with path MTU discovery timeouts.
   1217 			 */
   1218 			mtu = ifmtu;
   1219 			if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU))
   1220 				ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
   1221 		}
   1222 	} else if (ifp) {
   1223 		mtu = IN6_LINKMTU(ifp);
   1224 	} else
   1225 		error = EHOSTUNREACH; /* XXX */
   1226 
   1227 	*mtup = mtu;
   1228 	return(error);
   1229 }
   1230 
   1231 /*
   1232  * IP6 socket option processing.
   1233  */
   1234 int
   1235 ip6_ctloutput(op, so, level, optname, mp)
   1236 	int op;
   1237 	struct socket *so;
   1238 	int level, optname;
   1239 	struct mbuf **mp;
   1240 {
   1241 	struct in6pcb *in6p = sotoin6pcb(so);
   1242 	struct mbuf *m = *mp;
   1243 	int optval = 0;
   1244 	int error = 0;
   1245 	struct proc *p = curproc;	/* XXX */
   1246 
   1247 	if (level == IPPROTO_IPV6) {
   1248 		switch (op) {
   1249 		case PRCO_SETOPT:
   1250 			switch (optname) {
   1251 			case IPV6_PKTOPTIONS:
   1252 				/* m is freed in ip6_pcbopts */
   1253 				return(ip6_pcbopts(&in6p->in6p_outputopts,
   1254 				    m, so));
   1255 			case IPV6_HOPOPTS:
   1256 			case IPV6_DSTOPTS:
   1257 				if (p == 0 || suser(p->p_ucred, &p->p_acflag)) {
   1258 					error = EPERM;
   1259 					break;
   1260 				}
   1261 				/* FALLTHROUGH */
   1262 			case IPV6_UNICAST_HOPS:
   1263 			case IPV6_RECVOPTS:
   1264 			case IPV6_RECVRETOPTS:
   1265 			case IPV6_RECVDSTADDR:
   1266 			case IPV6_PKTINFO:
   1267 			case IPV6_HOPLIMIT:
   1268 			case IPV6_RTHDR:
   1269 			case IPV6_FAITH:
   1270 			case IPV6_V6ONLY:
   1271 				if (!m || m->m_len != sizeof(int)) {
   1272 					error = EINVAL;
   1273 					break;
   1274 				}
   1275 				optval = *mtod(m, int *);
   1276 				switch (optname) {
   1277 
   1278 				case IPV6_UNICAST_HOPS:
   1279 					if (optval < -1 || optval >= 256)
   1280 						error = EINVAL;
   1281 					else {
   1282 						/* -1 = kernel default */
   1283 						in6p->in6p_hops = optval;
   1284 					}
   1285 					break;
   1286 #define OPTSET(bit) \
   1287 do { \
   1288 	if (optval) \
   1289 		in6p->in6p_flags |= (bit); \
   1290 	else \
   1291 		in6p->in6p_flags &= ~(bit); \
   1292 } while (0)
   1293 
   1294 				case IPV6_RECVOPTS:
   1295 					OPTSET(IN6P_RECVOPTS);
   1296 					break;
   1297 
   1298 				case IPV6_RECVRETOPTS:
   1299 					OPTSET(IN6P_RECVRETOPTS);
   1300 					break;
   1301 
   1302 				case IPV6_RECVDSTADDR:
   1303 					OPTSET(IN6P_RECVDSTADDR);
   1304 					break;
   1305 
   1306 				case IPV6_PKTINFO:
   1307 					OPTSET(IN6P_PKTINFO);
   1308 					break;
   1309 
   1310 				case IPV6_HOPLIMIT:
   1311 					OPTSET(IN6P_HOPLIMIT);
   1312 					break;
   1313 
   1314 				case IPV6_HOPOPTS:
   1315 					OPTSET(IN6P_HOPOPTS);
   1316 					break;
   1317 
   1318 				case IPV6_DSTOPTS:
   1319 					OPTSET(IN6P_DSTOPTS);
   1320 					break;
   1321 
   1322 				case IPV6_RTHDR:
   1323 					OPTSET(IN6P_RTHDR);
   1324 					break;
   1325 
   1326 				case IPV6_FAITH:
   1327 					OPTSET(IN6P_FAITH);
   1328 					break;
   1329 
   1330 				case IPV6_V6ONLY:
   1331 					/*
   1332 					 * make setsockopt(IPV6_V6ONLY)
   1333 					 * available only prior to bind(2).
   1334 					 * see ipng mailing list, Jun 22 2001.
   1335 					 */
   1336 					if (in6p->in6p_lport ||
   1337 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
   1338 						error = EINVAL;
   1339 						break;
   1340 					}
   1341 #ifdef INET6_BINDV6ONLY
   1342 					if (!optval)
   1343 						error = EINVAL;
   1344 #else
   1345 					OPTSET(IN6P_IPV6_V6ONLY);
   1346 #endif
   1347 					break;
   1348 				}
   1349 				break;
   1350 #undef OPTSET
   1351 
   1352 			case IPV6_MULTICAST_IF:
   1353 			case IPV6_MULTICAST_HOPS:
   1354 			case IPV6_MULTICAST_LOOP:
   1355 			case IPV6_JOIN_GROUP:
   1356 			case IPV6_LEAVE_GROUP:
   1357 				error =	ip6_setmoptions(optname,
   1358 				    &in6p->in6p_moptions, m);
   1359 				break;
   1360 
   1361 			case IPV6_PORTRANGE:
   1362 				optval = *mtod(m, int *);
   1363 
   1364 				switch (optval) {
   1365 				case IPV6_PORTRANGE_DEFAULT:
   1366 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
   1367 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
   1368 					break;
   1369 
   1370 				case IPV6_PORTRANGE_HIGH:
   1371 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
   1372 					in6p->in6p_flags |= IN6P_HIGHPORT;
   1373 					break;
   1374 
   1375 				case IPV6_PORTRANGE_LOW:
   1376 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
   1377 					in6p->in6p_flags |= IN6P_LOWPORT;
   1378 					break;
   1379 
   1380 				default:
   1381 					error = EINVAL;
   1382 					break;
   1383 				}
   1384 				break;
   1385 
   1386 #ifdef IPSEC
   1387 			case IPV6_IPSEC_POLICY:
   1388 			    {
   1389 				caddr_t req = NULL;
   1390 				size_t len = 0;
   1391 
   1392 				int priv = 0;
   1393 				if (p == 0 || suser(p->p_ucred, &p->p_acflag))
   1394 					priv = 0;
   1395 				else
   1396 					priv = 1;
   1397 				if (m) {
   1398 					req = mtod(m, caddr_t);
   1399 					len = m->m_len;
   1400 				}
   1401 				error = ipsec6_set_policy(in6p,
   1402 				                   optname, req, len, priv);
   1403 			    }
   1404 				break;
   1405 #endif /* IPSEC */
   1406 
   1407 			default:
   1408 				error = ENOPROTOOPT;
   1409 				break;
   1410 			}
   1411 			if (m)
   1412 				(void)m_free(m);
   1413 			break;
   1414 
   1415 		case PRCO_GETOPT:
   1416 			switch (optname) {
   1417 
   1418 			case IPV6_OPTIONS:
   1419 			case IPV6_RETOPTS:
   1420 				error = ENOPROTOOPT;
   1421 				break;
   1422 
   1423 			case IPV6_PKTOPTIONS:
   1424 				if (in6p->in6p_options) {
   1425 					*mp = m_copym(in6p->in6p_options, 0,
   1426 					    M_COPYALL, M_WAIT);
   1427 				} else {
   1428 					*mp = m_get(M_WAIT, MT_SOOPTS);
   1429 					(*mp)->m_len = 0;
   1430 				}
   1431 				break;
   1432 
   1433 			case IPV6_HOPOPTS:
   1434 			case IPV6_DSTOPTS:
   1435 				if (p == 0 || suser(p->p_ucred, &p->p_acflag)) {
   1436 					error = EPERM;
   1437 					break;
   1438 				}
   1439 				/* FALLTHROUGH */
   1440 			case IPV6_UNICAST_HOPS:
   1441 			case IPV6_RECVOPTS:
   1442 			case IPV6_RECVRETOPTS:
   1443 			case IPV6_RECVDSTADDR:
   1444 			case IPV6_PORTRANGE:
   1445 			case IPV6_PKTINFO:
   1446 			case IPV6_HOPLIMIT:
   1447 			case IPV6_RTHDR:
   1448 			case IPV6_FAITH:
   1449 			case IPV6_V6ONLY:
   1450 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
   1451 				m->m_len = sizeof(int);
   1452 				switch (optname) {
   1453 
   1454 				case IPV6_UNICAST_HOPS:
   1455 					optval = in6p->in6p_hops;
   1456 					break;
   1457 
   1458 #define OPTBIT(bit) (in6p->in6p_flags & bit ? 1 : 0)
   1459 
   1460 				case IPV6_RECVOPTS:
   1461 					optval = OPTBIT(IN6P_RECVOPTS);
   1462 					break;
   1463 
   1464 				case IPV6_RECVRETOPTS:
   1465 					optval = OPTBIT(IN6P_RECVRETOPTS);
   1466 					break;
   1467 
   1468 				case IPV6_RECVDSTADDR:
   1469 					optval = OPTBIT(IN6P_RECVDSTADDR);
   1470 					break;
   1471 
   1472 				case IPV6_PORTRANGE:
   1473 				    {
   1474 					int flags;
   1475 					flags = in6p->in6p_flags;
   1476 					if (flags & IN6P_HIGHPORT)
   1477 						optval = IPV6_PORTRANGE_HIGH;
   1478 					else if (flags & IN6P_LOWPORT)
   1479 						optval = IPV6_PORTRANGE_LOW;
   1480 					else
   1481 						optval = 0;
   1482 					break;
   1483 				    }
   1484 
   1485 				case IPV6_PKTINFO:
   1486 					optval = OPTBIT(IN6P_PKTINFO);
   1487 					break;
   1488 
   1489 				case IPV6_HOPLIMIT:
   1490 					optval = OPTBIT(IN6P_HOPLIMIT);
   1491 					break;
   1492 
   1493 				case IPV6_HOPOPTS:
   1494 					optval = OPTBIT(IN6P_HOPOPTS);
   1495 					break;
   1496 
   1497 				case IPV6_DSTOPTS:
   1498 					optval = OPTBIT(IN6P_DSTOPTS);
   1499 					break;
   1500 
   1501 				case IPV6_RTHDR:
   1502 					optval = OPTBIT(IN6P_RTHDR);
   1503 					break;
   1504 
   1505 				case IPV6_FAITH:
   1506 					optval = OPTBIT(IN6P_FAITH);
   1507 					break;
   1508 
   1509 				case IPV6_V6ONLY:
   1510 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
   1511 					break;
   1512 				}
   1513 				*mtod(m, int *) = optval;
   1514 				break;
   1515 
   1516 			case IPV6_MULTICAST_IF:
   1517 			case IPV6_MULTICAST_HOPS:
   1518 			case IPV6_MULTICAST_LOOP:
   1519 			case IPV6_JOIN_GROUP:
   1520 			case IPV6_LEAVE_GROUP:
   1521 				error = ip6_getmoptions(optname, in6p->in6p_moptions, mp);
   1522 				break;
   1523 
   1524 #ifdef IPSEC
   1525 			case IPV6_IPSEC_POLICY:
   1526 			{
   1527 				caddr_t req = NULL;
   1528 				size_t len = 0;
   1529 
   1530 				if (m) {
   1531 					req = mtod(m, caddr_t);
   1532 					len = m->m_len;
   1533 				}
   1534 				error = ipsec6_get_policy(in6p, req, len, mp);
   1535 				break;
   1536 			}
   1537 #endif /* IPSEC */
   1538 
   1539 			default:
   1540 				error = ENOPROTOOPT;
   1541 				break;
   1542 			}
   1543 			break;
   1544 		}
   1545 	} else {
   1546 		error = EINVAL;
   1547 		if (op == PRCO_SETOPT && *mp)
   1548 			(void)m_free(*mp);
   1549 	}
   1550 	return(error);
   1551 }
   1552 
   1553 /*
   1554  * Set up IP6 options in pcb for insertion in output packets.
   1555  * Store in mbuf with pointer in pcbopt, adding pseudo-option
   1556  * with destination address if source routed.
   1557  */
   1558 static int
   1559 ip6_pcbopts(pktopt, m, so)
   1560 	struct ip6_pktopts **pktopt;
   1561 	struct mbuf *m;
   1562 	struct socket *so;
   1563 {
   1564 	struct ip6_pktopts *opt = *pktopt;
   1565 	int error = 0;
   1566 	struct proc *p = curproc;	/* XXX */
   1567 	int priv = 0;
   1568 
   1569 	/* turn off any old options. */
   1570 	if (opt) {
   1571 		if (opt->ip6po_m)
   1572 			(void)m_free(opt->ip6po_m);
   1573 	} else
   1574 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
   1575 	*pktopt = 0;
   1576 
   1577 	if (!m || m->m_len == 0) {
   1578 		/*
   1579 		 * Only turning off any previous options.
   1580 		 */
   1581 		if (opt)
   1582 			free(opt, M_IP6OPT);
   1583 		if (m)
   1584 			(void)m_free(m);
   1585 		return(0);
   1586 	}
   1587 
   1588 	/*  set options specified by user. */
   1589 	if (p && !suser(p->p_ucred, &p->p_acflag))
   1590 		priv = 1;
   1591 	if ((error = ip6_setpktoptions(m, opt, priv)) != 0) {
   1592 		(void)m_free(m);
   1593 		return(error);
   1594 	}
   1595 	*pktopt = opt;
   1596 	return(0);
   1597 }
   1598 
   1599 /*
   1600  * Set the IP6 multicast options in response to user setsockopt().
   1601  */
   1602 static int
   1603 ip6_setmoptions(optname, im6op, m)
   1604 	int optname;
   1605 	struct ip6_moptions **im6op;
   1606 	struct mbuf *m;
   1607 {
   1608 	int error = 0;
   1609 	u_int loop, ifindex;
   1610 	struct ipv6_mreq *mreq;
   1611 	struct ifnet *ifp;
   1612 	struct ip6_moptions *im6o = *im6op;
   1613 	struct route_in6 ro;
   1614 	struct sockaddr_in6 *dst;
   1615 	struct in6_multi_mship *imm;
   1616 	struct proc *p = curproc;	/* XXX */
   1617 
   1618 	if (im6o == NULL) {
   1619 		/*
   1620 		 * No multicast option buffer attached to the pcb;
   1621 		 * allocate one and initialize to default values.
   1622 		 */
   1623 		im6o = (struct ip6_moptions *)
   1624 			malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
   1625 
   1626 		if (im6o == NULL)
   1627 			return(ENOBUFS);
   1628 		*im6op = im6o;
   1629 		im6o->im6o_multicast_ifp = NULL;
   1630 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
   1631 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
   1632 		LIST_INIT(&im6o->im6o_memberships);
   1633 	}
   1634 
   1635 	switch (optname) {
   1636 
   1637 	case IPV6_MULTICAST_IF:
   1638 		/*
   1639 		 * Select the interface for outgoing multicast packets.
   1640 		 */
   1641 		if (m == NULL || m->m_len != sizeof(u_int)) {
   1642 			error = EINVAL;
   1643 			break;
   1644 		}
   1645 		bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
   1646 		if (ifindex < 0 || if_index < ifindex) {
   1647 			error = ENXIO;	/* XXX EINVAL? */
   1648 			break;
   1649 		}
   1650 		ifp = ifindex2ifnet[ifindex];
   1651 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
   1652 			error = EADDRNOTAVAIL;
   1653 			break;
   1654 		}
   1655 		im6o->im6o_multicast_ifp = ifp;
   1656 		break;
   1657 
   1658 	case IPV6_MULTICAST_HOPS:
   1659 	    {
   1660 		/*
   1661 		 * Set the IP6 hoplimit for outgoing multicast packets.
   1662 		 */
   1663 		int optval;
   1664 		if (m == NULL || m->m_len != sizeof(int)) {
   1665 			error = EINVAL;
   1666 			break;
   1667 		}
   1668 		bcopy(mtod(m, u_int *), &optval, sizeof(optval));
   1669 		if (optval < -1 || optval >= 256)
   1670 			error = EINVAL;
   1671 		else if (optval == -1)
   1672 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
   1673 		else
   1674 			im6o->im6o_multicast_hlim = optval;
   1675 		break;
   1676 	    }
   1677 
   1678 	case IPV6_MULTICAST_LOOP:
   1679 		/*
   1680 		 * Set the loopback flag for outgoing multicast packets.
   1681 		 * Must be zero or one.
   1682 		 */
   1683 		if (m == NULL || m->m_len != sizeof(u_int)) {
   1684 			error = EINVAL;
   1685 			break;
   1686 		}
   1687 		bcopy(mtod(m, u_int *), &loop, sizeof(loop));
   1688 		if (loop > 1) {
   1689 			error = EINVAL;
   1690 			break;
   1691 		}
   1692 		im6o->im6o_multicast_loop = loop;
   1693 		break;
   1694 
   1695 	case IPV6_JOIN_GROUP:
   1696 		/*
   1697 		 * Add a multicast group membership.
   1698 		 * Group must be a valid IP6 multicast address.
   1699 		 */
   1700 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
   1701 			error = EINVAL;
   1702 			break;
   1703 		}
   1704 		mreq = mtod(m, struct ipv6_mreq *);
   1705 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
   1706 			/*
   1707 			 * We use the unspecified address to specify to accept
   1708 			 * all multicast addresses. Only super user is allowed
   1709 			 * to do this.
   1710 			 */
   1711 			if (suser(p->p_ucred, &p->p_acflag))
   1712 			{
   1713 				error = EACCES;
   1714 				break;
   1715 			}
   1716 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
   1717 			error = EINVAL;
   1718 			break;
   1719 		}
   1720 
   1721 		/*
   1722 		 * If the interface is specified, validate it.
   1723 		 */
   1724 		if (mreq->ipv6mr_interface < 0
   1725 		 || if_index < mreq->ipv6mr_interface) {
   1726 			error = ENXIO;	/* XXX EINVAL? */
   1727 			break;
   1728 		}
   1729 		/*
   1730 		 * If no interface was explicitly specified, choose an
   1731 		 * appropriate one according to the given multicast address.
   1732 		 */
   1733 		if (mreq->ipv6mr_interface == 0) {
   1734 			/*
   1735 			 * If the multicast address is in node-local scope,
   1736 			 * the interface should be a loopback interface.
   1737 			 * Otherwise, look up the routing table for the
   1738 			 * address, and choose the outgoing interface.
   1739 			 *   XXX: is it a good approach?
   1740 			 */
   1741 			if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
   1742 				ifp = &loif[0];
   1743 			} else {
   1744 				ro.ro_rt = NULL;
   1745 				dst = (struct sockaddr_in6 *)&ro.ro_dst;
   1746 				bzero(dst, sizeof(*dst));
   1747 				dst->sin6_len = sizeof(struct sockaddr_in6);
   1748 				dst->sin6_family = AF_INET6;
   1749 				dst->sin6_addr = mreq->ipv6mr_multiaddr;
   1750 				rtalloc((struct route *)&ro);
   1751 				if (ro.ro_rt == NULL) {
   1752 					error = EADDRNOTAVAIL;
   1753 					break;
   1754 				}
   1755 				ifp = ro.ro_rt->rt_ifp;
   1756 				rtfree(ro.ro_rt);
   1757 			}
   1758 		} else
   1759 			ifp = ifindex2ifnet[mreq->ipv6mr_interface];
   1760 
   1761 		/*
   1762 		 * See if we found an interface, and confirm that it
   1763 		 * supports multicast
   1764 		 */
   1765 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
   1766 			error = EADDRNOTAVAIL;
   1767 			break;
   1768 		}
   1769 		/*
   1770 		 * Put interface index into the multicast address,
   1771 		 * if the address has link-local scope.
   1772 		 */
   1773 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
   1774 			mreq->ipv6mr_multiaddr.s6_addr16[1] =
   1775 			    htons(mreq->ipv6mr_interface);
   1776 		}
   1777 		/*
   1778 		 * See if the membership already exists.
   1779 		 */
   1780 		for (imm = im6o->im6o_memberships.lh_first;
   1781 		     imm != NULL; imm = imm->i6mm_chain.le_next)
   1782 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
   1783 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
   1784 			    &mreq->ipv6mr_multiaddr))
   1785 				break;
   1786 		if (imm != NULL) {
   1787 			error = EADDRINUSE;
   1788 			break;
   1789 		}
   1790 		/*
   1791 		 * Everything looks good; add a new record to the multicast
   1792 		 * address list for the given interface.
   1793 		 */
   1794 		imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error);
   1795 		if (!imm)
   1796 			break;
   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 			{
   1813 				error = EACCES;
   1814 				break;
   1815 			}
   1816 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
   1817 			error = EINVAL;
   1818 			break;
   1819 		}
   1820 		/*
   1821 		 * If an interface address was specified, get a pointer
   1822 		 * to its ifnet structure.
   1823 		 */
   1824 		if (mreq->ipv6mr_interface < 0
   1825 		 || if_index < mreq->ipv6mr_interface) {
   1826 			error = ENXIO;	/* XXX EINVAL? */
   1827 			break;
   1828 		}
   1829 		ifp = ifindex2ifnet[mreq->ipv6mr_interface];
   1830 		/*
   1831 		 * Put interface index into the multicast address,
   1832 		 * if the address has link-local scope.
   1833 		 */
   1834 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
   1835 			mreq->ipv6mr_multiaddr.s6_addr16[1] =
   1836 			    htons(mreq->ipv6mr_interface);
   1837 		}
   1838 		/*
   1839 		 * Find the membership in the membership list.
   1840 		 */
   1841 		for (imm = im6o->im6o_memberships.lh_first;
   1842 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
   1843 			if ((ifp == NULL || 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_leavegroup(imm);
   1859 		break;
   1860 
   1861 	default:
   1862 		error = EOPNOTSUPP;
   1863 		break;
   1864 	}
   1865 
   1866 	/*
   1867 	 * If all options have default values, no need to keep the mbuf.
   1868 	 */
   1869 	if (im6o->im6o_multicast_ifp == NULL &&
   1870 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
   1871 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
   1872 	    im6o->im6o_memberships.lh_first == NULL) {
   1873 		free(*im6op, M_IPMOPTS);
   1874 		*im6op = NULL;
   1875 	}
   1876 
   1877 	return(error);
   1878 }
   1879 
   1880 /*
   1881  * Return the IP6 multicast options in response to user getsockopt().
   1882  */
   1883 static int
   1884 ip6_getmoptions(optname, im6o, mp)
   1885 	int optname;
   1886 	struct ip6_moptions *im6o;
   1887 	struct mbuf **mp;
   1888 {
   1889 	u_int *hlim, *loop, *ifindex;
   1890 
   1891 	*mp = m_get(M_WAIT, MT_SOOPTS);
   1892 
   1893 	switch (optname) {
   1894 
   1895 	case IPV6_MULTICAST_IF:
   1896 		ifindex = mtod(*mp, u_int *);
   1897 		(*mp)->m_len = sizeof(u_int);
   1898 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
   1899 			*ifindex = 0;
   1900 		else
   1901 			*ifindex = im6o->im6o_multicast_ifp->if_index;
   1902 		return(0);
   1903 
   1904 	case IPV6_MULTICAST_HOPS:
   1905 		hlim = mtod(*mp, u_int *);
   1906 		(*mp)->m_len = sizeof(u_int);
   1907 		if (im6o == NULL)
   1908 			*hlim = ip6_defmcasthlim;
   1909 		else
   1910 			*hlim = im6o->im6o_multicast_hlim;
   1911 		return(0);
   1912 
   1913 	case IPV6_MULTICAST_LOOP:
   1914 		loop = mtod(*mp, u_int *);
   1915 		(*mp)->m_len = sizeof(u_int);
   1916 		if (im6o == NULL)
   1917 			*loop = ip6_defmcasthlim;
   1918 		else
   1919 			*loop = im6o->im6o_multicast_loop;
   1920 		return(0);
   1921 
   1922 	default:
   1923 		return(EOPNOTSUPP);
   1924 	}
   1925 }
   1926 
   1927 /*
   1928  * Discard the IP6 multicast options.
   1929  */
   1930 void
   1931 ip6_freemoptions(im6o)
   1932 	struct ip6_moptions *im6o;
   1933 {
   1934 	struct in6_multi_mship *imm;
   1935 
   1936 	if (im6o == NULL)
   1937 		return;
   1938 
   1939 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
   1940 		LIST_REMOVE(imm, i6mm_chain);
   1941 		in6_leavegroup(imm);
   1942 	}
   1943 	free(im6o, M_IPMOPTS);
   1944 }
   1945 
   1946 /*
   1947  * Set IPv6 outgoing packet options based on advanced API.
   1948  */
   1949 int
   1950 ip6_setpktoptions(control, opt, priv)
   1951 	struct mbuf *control;
   1952 	struct ip6_pktopts *opt;
   1953 	int priv;
   1954 {
   1955 	struct cmsghdr *cm = 0;
   1956 
   1957 	if (control == 0 || opt == 0)
   1958 		return(EINVAL);
   1959 
   1960 	bzero(opt, sizeof(*opt));
   1961 	opt->ip6po_hlim = -1; /* -1 means to use default hop limit */
   1962 
   1963 	/*
   1964 	 * XXX: Currently, we assume all the optional information is stored
   1965 	 * in a single mbuf.
   1966 	 */
   1967 	if (control->m_next)
   1968 		return(EINVAL);
   1969 
   1970 	opt->ip6po_m = control;
   1971 
   1972 	for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
   1973 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
   1974 		cm = mtod(control, struct cmsghdr *);
   1975 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
   1976 			return(EINVAL);
   1977 		if (cm->cmsg_level != IPPROTO_IPV6)
   1978 			continue;
   1979 
   1980 		switch (cm->cmsg_type) {
   1981 		case IPV6_PKTINFO:
   1982 			if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo)))
   1983 				return(EINVAL);
   1984 			opt->ip6po_pktinfo = (struct in6_pktinfo *)CMSG_DATA(cm);
   1985 			if (opt->ip6po_pktinfo->ipi6_ifindex &&
   1986 			    IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr))
   1987 				opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] =
   1988 					htons(opt->ip6po_pktinfo->ipi6_ifindex);
   1989 
   1990 			if (opt->ip6po_pktinfo->ipi6_ifindex > if_index ||
   1991 			    opt->ip6po_pktinfo->ipi6_ifindex < 0) {
   1992 				return(ENXIO);
   1993 			}
   1994 
   1995 			/*
   1996 			 * Check if the requested source address is indeed a
   1997 			 * unicast address assigned to the node, and can be
   1998 			 * used as the packet's source address.
   1999 			 */
   2000 			if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
   2001 				struct ifaddr *ia;
   2002 				struct in6_ifaddr *ia6;
   2003 				struct sockaddr_in6 sin6;
   2004 
   2005 				bzero(&sin6, sizeof(sin6));
   2006 				sin6.sin6_len = sizeof(sin6);
   2007 				sin6.sin6_family = AF_INET6;
   2008 				sin6.sin6_addr =
   2009 					opt->ip6po_pktinfo->ipi6_addr;
   2010 				ia = ifa_ifwithaddr(sin6tosa(&sin6));
   2011 				if (ia == NULL ||
   2012 				    (opt->ip6po_pktinfo->ipi6_ifindex &&
   2013 				     (ia->ifa_ifp->if_index !=
   2014 				      opt->ip6po_pktinfo->ipi6_ifindex))) {
   2015 					return(EADDRNOTAVAIL);
   2016 				}
   2017 				ia6 = (struct in6_ifaddr *)ia;
   2018 				if ((ia6->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) != 0) {
   2019 					return(EADDRNOTAVAIL);
   2020 				}
   2021 
   2022 				/*
   2023 				 * Check if the requested source address is
   2024 				 * indeed a unicast address assigned to the
   2025 				 * node.
   2026 				 */
   2027 				if (IN6_IS_ADDR_MULTICAST(&opt->ip6po_pktinfo->ipi6_addr))
   2028 					return(EADDRNOTAVAIL);
   2029 			}
   2030 			break;
   2031 
   2032 		case IPV6_HOPLIMIT:
   2033 			if (cm->cmsg_len != CMSG_LEN(sizeof(int)))
   2034 				return(EINVAL);
   2035 
   2036 			bcopy(CMSG_DATA(cm), &opt->ip6po_hlim,
   2037 			    sizeof(opt->ip6po_hlim));
   2038 			if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255)
   2039 				return(EINVAL);
   2040 			break;
   2041 
   2042 		case IPV6_NEXTHOP:
   2043 			if (!priv)
   2044 				return(EPERM);
   2045 
   2046 			/* check if cmsg_len is large enough for sa_len */
   2047 			if (cm->cmsg_len < sizeof(u_char) ||
   2048 			    cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm)))
   2049 				return(EINVAL);
   2050 
   2051 			opt->ip6po_nexthop = (struct sockaddr *)CMSG_DATA(cm);
   2052 
   2053 			break;
   2054 
   2055 		case IPV6_HOPOPTS:
   2056 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh)))
   2057 				return(EINVAL);
   2058 			opt->ip6po_hbh = (struct ip6_hbh *)CMSG_DATA(cm);
   2059 			if (cm->cmsg_len !=
   2060 			    CMSG_LEN((opt->ip6po_hbh->ip6h_len + 1) << 3))
   2061 				return(EINVAL);
   2062 			break;
   2063 
   2064 		case IPV6_DSTOPTS:
   2065 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest)))
   2066 				return(EINVAL);
   2067 
   2068 			/*
   2069 			 * If there is no routing header yet, the destination
   2070 			 * options header should be put on the 1st part.
   2071 			 * Otherwise, the header should be on the 2nd part.
   2072 			 * (See RFC 2460, section 4.1)
   2073 			 */
   2074 			if (opt->ip6po_rthdr == NULL) {
   2075 				opt->ip6po_dest1 =
   2076 				    (struct ip6_dest *)CMSG_DATA(cm);
   2077 				if (cm->cmsg_len !=
   2078 				    CMSG_LEN((opt->ip6po_dest1->ip6d_len + 1) << 3));
   2079 					return(EINVAL);
   2080 			}
   2081 			else {
   2082 				opt->ip6po_dest2 =
   2083 				    (struct ip6_dest *)CMSG_DATA(cm);
   2084 				if (cm->cmsg_len !=
   2085 				    CMSG_LEN((opt->ip6po_dest2->ip6d_len + 1) << 3))
   2086 					return(EINVAL);
   2087 			}
   2088 			break;
   2089 
   2090 		case IPV6_RTHDR:
   2091 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr)))
   2092 				return(EINVAL);
   2093 			opt->ip6po_rthdr = (struct ip6_rthdr *)CMSG_DATA(cm);
   2094 			if (cm->cmsg_len !=
   2095 			    CMSG_LEN((opt->ip6po_rthdr->ip6r_len + 1) << 3))
   2096 				return(EINVAL);
   2097 			switch (opt->ip6po_rthdr->ip6r_type) {
   2098 			case IPV6_RTHDR_TYPE_0:
   2099 				if (opt->ip6po_rthdr->ip6r_segleft == 0)
   2100 					return(EINVAL);
   2101 				break;
   2102 			default:
   2103 				return(EINVAL);
   2104 			}
   2105 			break;
   2106 
   2107 		default:
   2108 			return(ENOPROTOOPT);
   2109 		}
   2110 	}
   2111 
   2112 	return(0);
   2113 }
   2114 
   2115 /*
   2116  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
   2117  * packet to the input queue of a specified interface.  Note that this
   2118  * calls the output routine of the loopback "driver", but with an interface
   2119  * pointer that might NOT be &loif -- easier than replicating that code here.
   2120  */
   2121 void
   2122 ip6_mloopback(ifp, m, dst)
   2123 	struct ifnet *ifp;
   2124 	struct mbuf *m;
   2125 	struct sockaddr_in6 *dst;
   2126 {
   2127 	struct mbuf *copym;
   2128 	struct ip6_hdr *ip6;
   2129 
   2130 	copym = m_copy(m, 0, M_COPYALL);
   2131 	if (copym == NULL)
   2132 		return;
   2133 
   2134 	/*
   2135 	 * Make sure to deep-copy IPv6 header portion in case the data
   2136 	 * is in an mbuf cluster, so that we can safely override the IPv6
   2137 	 * header portion later.
   2138 	 */
   2139 	if ((copym->m_flags & M_EXT) != 0 ||
   2140 	    copym->m_len < sizeof(struct ip6_hdr)) {
   2141 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
   2142 		if (copym == NULL)
   2143 			return;
   2144 	}
   2145 
   2146 #ifdef DIAGNOSTIC
   2147 	if (copym->m_len < sizeof(*ip6)) {
   2148 		m_freem(copym);
   2149 		return;
   2150 	}
   2151 #endif
   2152 
   2153 	ip6 = mtod(copym, struct ip6_hdr *);
   2154 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
   2155 		ip6->ip6_src.s6_addr16[1] = 0;
   2156 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
   2157 		ip6->ip6_dst.s6_addr16[1] = 0;
   2158 
   2159 	(void)looutput(ifp, copym, (struct sockaddr *)dst, NULL);
   2160 }
   2161 
   2162 /*
   2163  * Chop IPv6 header off from the payload.
   2164  */
   2165 static int
   2166 ip6_splithdr(m, exthdrs)
   2167 	struct mbuf *m;
   2168 	struct ip6_exthdrs *exthdrs;
   2169 {
   2170 	struct mbuf *mh;
   2171 	struct ip6_hdr *ip6;
   2172 
   2173 	ip6 = mtod(m, struct ip6_hdr *);
   2174 	if (m->m_len > sizeof(*ip6)) {
   2175 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
   2176 		if (mh == 0) {
   2177 			m_freem(m);
   2178 			return ENOBUFS;
   2179 		}
   2180 		M_COPY_PKTHDR(mh, m);
   2181 		MH_ALIGN(mh, sizeof(*ip6));
   2182 		m->m_flags &= ~M_PKTHDR;
   2183 		m->m_len -= sizeof(*ip6);
   2184 		m->m_data += sizeof(*ip6);
   2185 		mh->m_next = m;
   2186 		m = mh;
   2187 		m->m_len = sizeof(*ip6);
   2188 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
   2189 	}
   2190 	exthdrs->ip6e_ip6 = m;
   2191 	return 0;
   2192 }
   2193 
   2194 /*
   2195  * Compute IPv6 extension header length.
   2196  */
   2197 int
   2198 ip6_optlen(in6p)
   2199 	struct in6pcb *in6p;
   2200 {
   2201 	int len;
   2202 
   2203 	if (!in6p->in6p_outputopts)
   2204 		return 0;
   2205 
   2206 	len = 0;
   2207 #define elen(x) \
   2208     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
   2209 
   2210 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
   2211 	len += elen(in6p->in6p_outputopts->ip6po_dest1);
   2212 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
   2213 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
   2214 	return len;
   2215 #undef elen
   2216 }
   2217