xref: /netbsd-src/sys/netinet6/ip6_output.c (revision 3cec974c61d7fac0a37c0377723a33214a458c8b)
1 /*	$NetBSD: ip6_output.c,v 1.31 2001/02/10 04:14:28 itojun Exp $	*/
2 /*	$KAME: ip6_output.c,v 1.152 2001/02/02 15:36:33 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 #endif /* IPSEC */
100 
101 #include "loop.h"
102 
103 #include <net/net_osdep.h>
104 
105 #ifdef IPV6FIREWALL
106 #include <netinet6/ip6_fw.h>
107 #endif
108 
109 #ifdef PFIL_HOOKS
110 extern struct pfil_head inet6_pfil_hook;	/* XXX */
111 #endif
112 
113 struct ip6_exthdrs {
114 	struct mbuf *ip6e_ip6;
115 	struct mbuf *ip6e_hbh;
116 	struct mbuf *ip6e_dest1;
117 	struct mbuf *ip6e_rthdr;
118 	struct mbuf *ip6e_dest2;
119 };
120 
121 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
122 			    struct socket *));
123 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *));
124 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **));
125 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int));
126 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
127 				  struct ip6_frag **));
128 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t));
129 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *));
130 
131 extern struct ifnet loif[NLOOP];
132 
133 /*
134  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
135  * header (with pri, len, nxt, hlim, src, dst).
136  * This function may modify ver and hlim only.
137  * The mbuf chain containing the packet will be freed.
138  * The mbuf opt, if present, will not be freed.
139  */
140 int
141 ip6_output(m0, opt, ro, flags, im6o, ifpp)
142 	struct mbuf *m0;
143 	struct ip6_pktopts *opt;
144 	struct route_in6 *ro;
145 	int flags;
146 	struct ip6_moptions *im6o;
147 	struct ifnet **ifpp;		/* XXX: just for statistics */
148 {
149 	struct ip6_hdr *ip6, *mhip6;
150 	struct ifnet *ifp, *origifp;
151 	struct mbuf *m = m0;
152 	int hlen, tlen, len, off;
153 	struct route_in6 ip6route;
154 	struct sockaddr_in6 *dst;
155 	int error = 0;
156 	struct in6_ifaddr *ia;
157 	u_long mtu;
158 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
159 	struct ip6_exthdrs exthdrs;
160 	struct in6_addr finaldst;
161 	struct route_in6 *ro_pmtu = NULL;
162 	int hdrsplit = 0;
163 	int needipsec = 0;
164 #ifdef IPSEC
165 	int needipsectun = 0;
166 	struct socket *so;
167 	struct secpolicy *sp = NULL;
168 
169 	/* for AH processing. stupid to have "socket" variable in IP layer... */
170 	so = ipsec_getsocket(m);
171 	(void)ipsec_setsocket(m, NULL);
172 	ip6 = mtod(m, struct ip6_hdr *);
173 #endif /* IPSEC */
174 
175 #define MAKE_EXTHDR(hp, mp)						\
176     do {								\
177 	if (hp) {							\
178 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
179 		error = ip6_copyexthdr((mp), (caddr_t)(hp), 		\
180 				       ((eh)->ip6e_len + 1) << 3);	\
181 		if (error)						\
182 			goto freehdrs;					\
183 	}								\
184     } while (0)
185 
186 	bzero(&exthdrs, sizeof(exthdrs));
187 	if (opt) {
188 		/* Hop-by-Hop options header */
189 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
190 		/* Destination options header(1st part) */
191 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
192 		/* Routing header */
193 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
194 		/* Destination options header(2nd part) */
195 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
196 	}
197 
198 #ifdef IPSEC
199 	/* get a security policy for this packet */
200 	if (so == NULL)
201 		sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
202 	else
203 		sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
204 
205 	if (sp == NULL) {
206 		ipsec6stat.out_inval++;
207 		goto freehdrs;
208 	}
209 
210 	error = 0;
211 
212 	/* check policy */
213 	switch (sp->policy) {
214 	case IPSEC_POLICY_DISCARD:
215 		/*
216 		 * This packet is just discarded.
217 		 */
218 		ipsec6stat.out_polvio++;
219 		goto freehdrs;
220 
221 	case IPSEC_POLICY_BYPASS:
222 	case IPSEC_POLICY_NONE:
223 		/* no need to do IPsec. */
224 		needipsec = 0;
225 		break;
226 
227 	case IPSEC_POLICY_IPSEC:
228 		if (sp->req == NULL) {
229 			/* XXX should be panic ? */
230 			printf("ip6_output: No IPsec request specified.\n");
231 			error = EINVAL;
232 			goto freehdrs;
233 		}
234 		needipsec = 1;
235 		break;
236 
237 	case IPSEC_POLICY_ENTRUST:
238 	default:
239 		printf("ip6_output: Invalid policy found. %d\n", sp->policy);
240 	}
241 #endif /* IPSEC */
242 
243 	/*
244 	 * Calculate the total length of the extension header chain.
245 	 * Keep the length of the unfragmentable part for fragmentation.
246 	 */
247 	optlen = 0;
248 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
249 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
250 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
251 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
252 	/* NOTE: we don't add AH/ESP length here. do that later. */
253 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
254 
255 	/*
256 	 * If we need IPsec, or there is at least one extension header,
257 	 * separate IP6 header from the payload.
258 	 */
259 	if ((needipsec || optlen) && !hdrsplit) {
260 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
261 			m = NULL;
262 			goto freehdrs;
263 		}
264 		m = exthdrs.ip6e_ip6;
265 		hdrsplit++;
266 	}
267 
268 	/* adjust pointer */
269 	ip6 = mtod(m, struct ip6_hdr *);
270 
271 	/* adjust mbuf packet header length */
272 	m->m_pkthdr.len += optlen;
273 	plen = m->m_pkthdr.len - sizeof(*ip6);
274 
275 	/* If this is a jumbo payload, insert a jumbo payload option. */
276 	if (plen > IPV6_MAXPACKET) {
277 		if (!hdrsplit) {
278 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
279 				m = NULL;
280 				goto freehdrs;
281 			}
282 			m = exthdrs.ip6e_ip6;
283 			hdrsplit++;
284 		}
285 		/* adjust pointer */
286 		ip6 = mtod(m, struct ip6_hdr *);
287 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
288 			goto freehdrs;
289 		ip6->ip6_plen = 0;
290 	} else
291 		ip6->ip6_plen = htons(plen);
292 
293 	/*
294 	 * Concatenate headers and fill in next header fields.
295 	 * Here we have, on "m"
296 	 *	IPv6 payload
297 	 * and we insert headers accordingly.  Finally, we should be getting:
298 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
299 	 *
300 	 * during the header composing process, "m" points to IPv6 header.
301 	 * "mprev" points to an extension header prior to esp.
302 	 */
303 	{
304 		u_char *nexthdrp = &ip6->ip6_nxt;
305 		struct mbuf *mprev = m;
306 
307 		/*
308 		 * we treat dest2 specially.  this makes IPsec processing
309 		 * much easier.
310 		 *
311 		 * result: IPv6 dest2 payload
312 		 * m and mprev will point to IPv6 header.
313 		 */
314 		if (exthdrs.ip6e_dest2) {
315 			if (!hdrsplit)
316 				panic("assumption failed: hdr not split");
317 			exthdrs.ip6e_dest2->m_next = m->m_next;
318 			m->m_next = exthdrs.ip6e_dest2;
319 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
320 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
321 		}
322 
323 #define MAKE_CHAIN(m, mp, p, i)\
324     do {\
325 	if (m) {\
326 		if (!hdrsplit) \
327 			panic("assumption failed: hdr not split"); \
328 		*mtod((m), u_char *) = *(p);\
329 		*(p) = (i);\
330 		p = mtod((m), u_char *);\
331 		(m)->m_next = (mp)->m_next;\
332 		(mp)->m_next = (m);\
333 		(mp) = (m);\
334 	}\
335     } while (0)
336 		/*
337 		 * result: IPv6 hbh dest1 rthdr dest2 payload
338 		 * m will point to IPv6 header.  mprev will point to the
339 		 * extension header prior to dest2 (rthdr in the above case).
340 		 */
341 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev,
342 			   nexthdrp, IPPROTO_HOPOPTS);
343 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev,
344 			   nexthdrp, IPPROTO_DSTOPTS);
345 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev,
346 			   nexthdrp, IPPROTO_ROUTING);
347 
348 #ifdef IPSEC
349 		if (!needipsec)
350 			goto skip_ipsec2;
351 
352 		/*
353 		 * pointers after IPsec headers are not valid any more.
354 		 * other pointers need a great care too.
355 		 * (IPsec routines should not mangle mbufs prior to AH/ESP)
356 		 */
357 		exthdrs.ip6e_dest2 = NULL;
358 
359 	    {
360 		struct ip6_rthdr *rh = NULL;
361 		int segleft_org = 0;
362 		struct ipsec_output_state state;
363 
364 		if (exthdrs.ip6e_rthdr) {
365 			rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
366 			segleft_org = rh->ip6r_segleft;
367 			rh->ip6r_segleft = 0;
368 		}
369 
370 		bzero(&state, sizeof(state));
371 		state.m = m;
372 		error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
373 			&needipsectun);
374 		m = state.m;
375 		if (error) {
376 			/* mbuf is already reclaimed in ipsec6_output_trans. */
377 			m = NULL;
378 			switch (error) {
379 			case EHOSTUNREACH:
380 			case ENETUNREACH:
381 			case EMSGSIZE:
382 			case ENOBUFS:
383 			case ENOMEM:
384 				break;
385 			default:
386 				printf("ip6_output (ipsec): error code %d\n", error);
387 				/*fall through*/
388 			case ENOENT:
389 				/* don't show these error codes to the user */
390 				error = 0;
391 				break;
392 			}
393 			goto bad;
394 		}
395 		if (exthdrs.ip6e_rthdr) {
396 			/* ah6_output doesn't modify mbuf chain */
397 			rh->ip6r_segleft = segleft_org;
398 		}
399 	    }
400 skip_ipsec2:;
401 #endif
402 	}
403 
404 	/*
405 	 * If there is a routing header, replace destination address field
406 	 * with the first hop of the routing header.
407 	 */
408 	if (exthdrs.ip6e_rthdr) {
409 		struct ip6_rthdr *rh =
410 			(struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
411 						  struct ip6_rthdr *));
412 		struct ip6_rthdr0 *rh0;
413 
414 		finaldst = ip6->ip6_dst;
415 		switch (rh->ip6r_type) {
416 		case IPV6_RTHDR_TYPE_0:
417 			 rh0 = (struct ip6_rthdr0 *)rh;
418 			 ip6->ip6_dst = rh0->ip6r0_addr[0];
419 			 bcopy((caddr_t)&rh0->ip6r0_addr[1],
420 				 (caddr_t)&rh0->ip6r0_addr[0],
421 				 sizeof(struct in6_addr) * (rh0->ip6r0_segleft - 1)
422 				 );
423 			 rh0->ip6r0_addr[rh0->ip6r0_segleft - 1] = finaldst;
424 			 break;
425 		default:	/* is it possible? */
426 			 error = EINVAL;
427 			 goto bad;
428 		}
429 	}
430 
431 	/* Source address validation */
432 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
433 	    (flags & IPV6_DADOUTPUT) == 0) {
434 		error = EOPNOTSUPP;
435 		ip6stat.ip6s_badscope++;
436 		goto bad;
437 	}
438 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
439 		error = EOPNOTSUPP;
440 		ip6stat.ip6s_badscope++;
441 		goto bad;
442 	}
443 
444 	ip6stat.ip6s_localout++;
445 
446 	/*
447 	 * Route packet.
448 	 */
449 	if (ro == 0) {
450 		ro = &ip6route;
451 		bzero((caddr_t)ro, sizeof(*ro));
452 	}
453 	ro_pmtu = ro;
454 	if (opt && opt->ip6po_rthdr)
455 		ro = &opt->ip6po_route;
456 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
457 	/*
458 	 * If there is a cached route,
459 	 * check that it is to the same destination
460 	 * and is still up. If not, free it and try again.
461 	 */
462 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
463 			 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
464 		RTFREE(ro->ro_rt);
465 		ro->ro_rt = (struct rtentry *)0;
466 	}
467 	if (ro->ro_rt == 0) {
468 		bzero(dst, sizeof(*dst));
469 		dst->sin6_family = AF_INET6;
470 		dst->sin6_len = sizeof(struct sockaddr_in6);
471 		dst->sin6_addr = ip6->ip6_dst;
472 	}
473 #ifdef IPSEC
474 	if (needipsec && needipsectun) {
475 		struct ipsec_output_state state;
476 
477 		/*
478 		 * All the extension headers will become inaccessible
479 		 * (since they can be encrypted).
480 		 * Don't panic, we need no more updates to extension headers
481 		 * on inner IPv6 packet (since they are now encapsulated).
482 		 *
483 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
484 		 */
485 		bzero(&exthdrs, sizeof(exthdrs));
486 		exthdrs.ip6e_ip6 = m;
487 
488 		bzero(&state, sizeof(state));
489 		state.m = m;
490 		state.ro = (struct route *)ro;
491 		state.dst = (struct sockaddr *)dst;
492 
493 		error = ipsec6_output_tunnel(&state, sp, flags);
494 
495 		m = state.m;
496 		ro = (struct route_in6 *)state.ro;
497 		dst = (struct sockaddr_in6 *)state.dst;
498 		if (error) {
499 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
500 			m0 = m = NULL;
501 			m = NULL;
502 			switch (error) {
503 			case EHOSTUNREACH:
504 			case ENETUNREACH:
505 			case EMSGSIZE:
506 			case ENOBUFS:
507 			case ENOMEM:
508 				break;
509 			default:
510 				printf("ip6_output (ipsec): error code %d\n", error);
511 				/*fall through*/
512 			case ENOENT:
513 				/* don't show these error codes to the user */
514 				error = 0;
515 				break;
516 			}
517 			goto bad;
518 		}
519 
520 		exthdrs.ip6e_ip6 = m;
521 	}
522 #endif /*IPSEC*/
523 
524 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
525 		/* Unicast */
526 
527 #define ifatoia6(ifa)	((struct in6_ifaddr *)(ifa))
528 #define sin6tosa(sin6)	((struct sockaddr *)(sin6))
529 		/* xxx
530 		 * interface selection comes here
531 		 * if an interface is specified from an upper layer,
532 		 * ifp must point it.
533 		 */
534 		if (ro->ro_rt == 0) {
535 			/*
536 			 * non-bsdi always clone routes, if parent is
537 			 * PRF_CLONING.
538 			 */
539 			rtalloc((struct route *)ro);
540 		}
541 		if (ro->ro_rt == 0) {
542 			ip6stat.ip6s_noroute++;
543 			error = EHOSTUNREACH;
544 			/* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
545 			goto bad;
546 		}
547 		ia = ifatoia6(ro->ro_rt->rt_ifa);
548 		ifp = ro->ro_rt->rt_ifp;
549 		ro->ro_rt->rt_use++;
550 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
551 			dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
552 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
553 
554 		in6_ifstat_inc(ifp, ifs6_out_request);
555 
556 		/*
557 		 * Check if the outgoing interface conflicts with
558 		 * the interface specified by ifi6_ifindex (if specified).
559 		 * Note that loopback interface is always okay.
560 		 * (this may happen when we are sending a packet to one of
561 		 *  our own addresses.)
562 		 */
563 		if (opt && opt->ip6po_pktinfo
564 		 && opt->ip6po_pktinfo->ipi6_ifindex) {
565 			if (!(ifp->if_flags & IFF_LOOPBACK)
566 			 && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
567 				ip6stat.ip6s_noroute++;
568 				in6_ifstat_inc(ifp, ifs6_out_discard);
569 				error = EHOSTUNREACH;
570 				goto bad;
571 			}
572 		}
573 
574 		if (opt && opt->ip6po_hlim != -1)
575 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
576 	} else {
577 		/* Multicast */
578 		struct	in6_multi *in6m;
579 
580 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
581 
582 		/*
583 		 * See if the caller provided any multicast options
584 		 */
585 		ifp = NULL;
586 		if (im6o != NULL) {
587 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
588 			if (im6o->im6o_multicast_ifp != NULL)
589 				ifp = im6o->im6o_multicast_ifp;
590 		} else
591 			ip6->ip6_hlim = ip6_defmcasthlim;
592 
593 		/*
594 		 * See if the caller provided the outgoing interface
595 		 * as an ancillary data.
596 		 * Boundary check for ifindex is assumed to be already done.
597 		 */
598 		if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
599 			ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex];
600 
601 		/*
602 		 * If the destination is a node-local scope multicast,
603 		 * the packet should be loop-backed only.
604 		 */
605 		if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
606 			/*
607 			 * If the outgoing interface is already specified,
608 			 * it should be a loopback interface.
609 			 */
610 			if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) {
611 				ip6stat.ip6s_badscope++;
612 				error = ENETUNREACH; /* XXX: better error? */
613 				/* XXX correct ifp? */
614 				in6_ifstat_inc(ifp, ifs6_out_discard);
615 				goto bad;
616 			} else {
617 				ifp = &loif[0];
618 			}
619 		}
620 
621 		if (opt && opt->ip6po_hlim != -1)
622 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
623 
624 		/*
625 		 * If caller did not provide an interface lookup a
626 		 * default in the routing table.  This is either a
627 		 * default for the speicfied group (i.e. a host
628 		 * route), or a multicast default (a route for the
629 		 * ``net'' ff00::/8).
630 		 */
631 		if (ifp == NULL) {
632 			if (ro->ro_rt == 0) {
633 				ro->ro_rt = rtalloc1((struct sockaddr *)
634 						&ro->ro_dst, 0
635 						);
636 			}
637 			if (ro->ro_rt == 0) {
638 				ip6stat.ip6s_noroute++;
639 				error = EHOSTUNREACH;
640 				/* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
641 				goto bad;
642 			}
643 			ia = ifatoia6(ro->ro_rt->rt_ifa);
644 			ifp = ro->ro_rt->rt_ifp;
645 			ro->ro_rt->rt_use++;
646 		}
647 
648 		if ((flags & IPV6_FORWARDING) == 0)
649 			in6_ifstat_inc(ifp, ifs6_out_request);
650 		in6_ifstat_inc(ifp, ifs6_out_mcast);
651 
652 		/*
653 		 * Confirm that the outgoing interface supports multicast.
654 		 */
655 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
656 			ip6stat.ip6s_noroute++;
657 			in6_ifstat_inc(ifp, ifs6_out_discard);
658 			error = ENETUNREACH;
659 			goto bad;
660 		}
661 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
662 		if (in6m != NULL &&
663 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
664 			/*
665 			 * If we belong to the destination multicast group
666 			 * on the outgoing interface, and the caller did not
667 			 * forbid loopback, loop back a copy.
668 			 */
669 			ip6_mloopback(ifp, m, dst);
670 		} else {
671 			/*
672 			 * If we are acting as a multicast router, perform
673 			 * multicast forwarding as if the packet had just
674 			 * arrived on the interface to which we are about
675 			 * to send.  The multicast forwarding function
676 			 * recursively calls this function, using the
677 			 * IPV6_FORWARDING flag to prevent infinite recursion.
678 			 *
679 			 * Multicasts that are looped back by ip6_mloopback(),
680 			 * above, will be forwarded by the ip6_input() routine,
681 			 * if necessary.
682 			 */
683 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
684 				if (ip6_mforward(ip6, ifp, m) != 0) {
685 					m_freem(m);
686 					goto done;
687 				}
688 			}
689 		}
690 		/*
691 		 * Multicasts with a hoplimit of zero may be looped back,
692 		 * above, but must not be transmitted on a network.
693 		 * Also, multicasts addressed to the loopback interface
694 		 * are not sent -- the above call to ip6_mloopback() will
695 		 * loop back a copy if this host actually belongs to the
696 		 * destination group on the loopback interface.
697 		 */
698 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
699 			m_freem(m);
700 			goto done;
701 		}
702 	}
703 
704 	/*
705 	 * Fill the outgoing inteface to tell the upper layer
706 	 * to increment per-interface statistics.
707 	 */
708 	if (ifpp)
709 		*ifpp = ifp;
710 
711 	/*
712 	 * Determine path MTU.
713 	 */
714 	if (ro_pmtu != ro) {
715 		/* The first hop and the final destination may differ. */
716 		struct sockaddr_in6 *sin6_fin =
717 			(struct sockaddr_in6 *)&ro_pmtu->ro_dst;
718 		if (ro_pmtu->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
719 				       !IN6_ARE_ADDR_EQUAL(&sin6_fin->sin6_addr,
720 							   &finaldst))) {
721 			RTFREE(ro_pmtu->ro_rt);
722 			ro_pmtu->ro_rt = (struct rtentry *)0;
723 		}
724 		if (ro_pmtu->ro_rt == 0) {
725 			bzero(sin6_fin, sizeof(*sin6_fin));
726 			sin6_fin->sin6_family = AF_INET6;
727 			sin6_fin->sin6_len = sizeof(struct sockaddr_in6);
728 			sin6_fin->sin6_addr = finaldst;
729 
730 			rtalloc((struct route *)ro_pmtu);
731 		}
732 	}
733 	if (ro_pmtu->ro_rt != NULL) {
734 		u_int32_t ifmtu = nd_ifinfo[ifp->if_index].linkmtu;
735 
736 		mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
737 		if (mtu > ifmtu) {
738 			/*
739 			 * The MTU on the route is larger than the MTU on
740 			 * the interface!  This shouldn't happen, unless the
741 			 * MTU of the interface has been changed after the
742 			 * interface was brought up.  Change the MTU in the
743 			 * route to match the interface MTU (as long as the
744 			 * field isn't locked).
745 			 */
746 			 mtu = ifmtu;
747 			 if ((ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
748 				 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; /* XXX */
749 		}
750 	} else {
751 		mtu = nd_ifinfo[ifp->if_index].linkmtu;
752 	}
753 
754 	/* Fake scoped addresses */
755 	if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
756 		/*
757 		 * If source or destination address is a scoped address, and
758 		 * the packet is going to be sent to a loopback interface,
759 		 * we should keep the original interface.
760 		 */
761 
762 		/*
763 		 * XXX: this is a very experimental and temporary solution.
764 		 * We eventually have sockaddr_in6 and use the sin6_scope_id
765 		 * field of the structure here.
766 		 * We rely on the consistency between two scope zone ids
767 		 * of source add destination, which should already be assured
768 		 * Larger scopes than link will be supported in the near
769 		 * future.
770 		 */
771 		origifp = NULL;
772 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
773 			origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
774 		else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
775 			origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
776 		/*
777 		 * XXX: origifp can be NULL even in those two cases above.
778 		 * For example, if we remove the (only) link-local address
779 		 * from the loopback interface, and try to send a link-local
780 		 * address without link-id information.  Then the source
781 		 * address is ::1, and the destination address is the
782 		 * link-local address with its s6_addr16[1] being zero.
783 		 * What is worse, if the packet goes to the loopback interface
784 		 * by a default rejected route, the null pointer would be
785 		 * passed to looutput, and the kernel would hang.
786 		 * The following last resort would prevent such disaster.
787 		 */
788 		if (origifp == NULL)
789 			origifp = ifp;
790 	}
791 	else
792 		origifp = ifp;
793 #ifndef FAKE_LOOPBACK_IF
794 	if ((ifp->if_flags & IFF_LOOPBACK) == 0)
795 #else
796 	if (1)
797 #endif
798 	{
799 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
800 			ip6->ip6_src.s6_addr16[1] = 0;
801 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
802 			ip6->ip6_dst.s6_addr16[1] = 0;
803 	}
804 
805 	/*
806 	 * If the outgoing packet contains a hop-by-hop options header,
807 	 * it must be examined and processed even by the source node.
808 	 * (RFC 2460, section 4.)
809 	 */
810 	if (exthdrs.ip6e_hbh) {
811 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
812 		u_int32_t dummy1; /* XXX unused */
813 		u_int32_t dummy2; /* XXX unused */
814 
815 		/*
816 		 *  XXX: if we have to send an ICMPv6 error to the sender,
817 		 *       we need the M_LOOP flag since icmp6_error() expects
818 		 *       the IPv6 and the hop-by-hop options header are
819 		 *       continuous unless the flag is set.
820 		 */
821 		m->m_flags |= M_LOOP;
822 		m->m_pkthdr.rcvif = ifp;
823 		if (ip6_process_hopopts(m,
824 					(u_int8_t *)(hbh + 1),
825 					((hbh->ip6h_len + 1) << 3) -
826 					sizeof(struct ip6_hbh),
827 					&dummy1, &dummy2) < 0) {
828 			/* m was already freed at this point */
829 			error = EINVAL;/* better error? */
830 			goto done;
831 		}
832 		m->m_flags &= ~M_LOOP; /* XXX */
833 		m->m_pkthdr.rcvif = NULL;
834 	}
835 
836 #ifdef PFIL_HOOKS
837 	/*
838 	 * Run through list of hooks for output packets.
839 	 */
840 	if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp,
841 				    PFIL_OUT)) != 0)
842 		goto done;
843 	if (m == NULL)
844 		goto done;
845 	ip6 = mtod(m, struct ip6_hdr *);
846 #endif /* PFIL_HOOKS */
847 	/*
848 	 * Send the packet to the outgoing interface.
849 	 * If necessary, do IPv6 fragmentation before sending.
850 	 */
851 	tlen = m->m_pkthdr.len;
852 	if (tlen <= mtu
853 #ifdef notyet
854 	    /*
855 	     * On any link that cannot convey a 1280-octet packet in one piece,
856 	     * link-specific fragmentation and reassembly must be provided at
857 	     * a layer below IPv6. [RFC 2460, sec.5]
858 	     * Thus if the interface has ability of link-level fragmentation,
859 	     * we can just send the packet even if the packet size is
860 	     * larger than the link's MTU.
861 	     * XXX: IFF_FRAGMENTABLE (or such) flag has not been defined yet...
862 	     */
863 
864 	    || ifp->if_flags & IFF_FRAGMENTABLE
865 #endif
866 	    )
867 	{
868 #ifdef IFA_STATS
869 		struct in6_ifaddr *ia6;
870 		ip6 = mtod(m, struct ip6_hdr *);
871 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
872 		if (ia6) {
873 			ia6->ia_ifa.ifa_data.ifad_outbytes +=
874 				m->m_pkthdr.len;
875 		}
876 #endif
877 #ifdef IPSEC
878 		/* clean ipsec history once it goes out of the node */
879 		ipsec_delaux(m);
880 #endif
881 #ifdef OLDIP6OUTPUT
882 		error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst,
883 					  ro->ro_rt);
884 #else
885 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
886 #endif
887 		goto done;
888 	} else if (mtu < IPV6_MMTU) {
889 		/*
890 		 * note that path MTU is never less than IPV6_MMTU
891 		 * (see icmp6_input).
892 		 */
893 		error = EMSGSIZE;
894 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
895 		goto bad;
896 	} else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
897 		error = EMSGSIZE;
898 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
899 		goto bad;
900 	} else {
901 		struct mbuf **mnext, *m_frgpart;
902 		struct ip6_frag *ip6f;
903 		u_int32_t id = htonl(ip6_id++);
904 		u_char nextproto;
905 
906 		/*
907 		 * Too large for the destination or interface;
908 		 * fragment if possible.
909 		 * Must be able to put at least 8 bytes per fragment.
910 		 */
911 		hlen = unfragpartlen;
912 		if (mtu > IPV6_MAXPACKET)
913 			mtu = IPV6_MAXPACKET;
914 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
915 		if (len < 8) {
916 			error = EMSGSIZE;
917 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
918 			goto bad;
919 		}
920 
921 		mnext = &m->m_nextpkt;
922 
923 		/*
924 		 * Change the next header field of the last header in the
925 		 * unfragmentable part.
926 		 */
927 		if (exthdrs.ip6e_rthdr) {
928 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
929 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
930 		} else if (exthdrs.ip6e_dest1) {
931 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
932 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
933 		} else if (exthdrs.ip6e_hbh) {
934 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
935 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
936 		} else {
937 			nextproto = ip6->ip6_nxt;
938 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
939 		}
940 
941 		/*
942 		 * Loop through length of segment after first fragment,
943 		 * make new header and copy data of each part and link onto chain.
944 		 */
945 		m0 = m;
946 		for (off = hlen; off < tlen; off += len) {
947 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
948 			if (!m) {
949 				error = ENOBUFS;
950 				ip6stat.ip6s_odropped++;
951 				goto sendorfree;
952 			}
953 			m->m_flags = m0->m_flags & M_COPYFLAGS;
954 			*mnext = m;
955 			mnext = &m->m_nextpkt;
956 			m->m_data += max_linkhdr;
957 			mhip6 = mtod(m, struct ip6_hdr *);
958 			*mhip6 = *ip6;
959 			m->m_len = sizeof(*mhip6);
960  			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
961  			if (error) {
962 				ip6stat.ip6s_odropped++;
963 				goto sendorfree;
964 			}
965 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
966 			if (off + len >= tlen)
967 				len = tlen - off;
968 			else
969 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
970 			mhip6->ip6_plen = htons((u_short)(len + hlen +
971 							  sizeof(*ip6f) -
972 							  sizeof(struct ip6_hdr)));
973 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
974 				error = ENOBUFS;
975 				ip6stat.ip6s_odropped++;
976 				goto sendorfree;
977 			}
978 			m_cat(m, m_frgpart);
979 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
980 			m->m_pkthdr.rcvif = (struct ifnet *)0;
981 			ip6f->ip6f_reserved = 0;
982 			ip6f->ip6f_ident = id;
983 			ip6f->ip6f_nxt = nextproto;
984 			ip6stat.ip6s_ofragments++;
985 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
986 		}
987 
988 		in6_ifstat_inc(ifp, ifs6_out_fragok);
989 	}
990 
991 	/*
992 	 * Remove leading garbages.
993 	 */
994 sendorfree:
995 	m = m0->m_nextpkt;
996 	m0->m_nextpkt = 0;
997 	m_freem(m0);
998 	for (m0 = m; m; m = m0) {
999 		m0 = m->m_nextpkt;
1000 		m->m_nextpkt = 0;
1001 		if (error == 0) {
1002 #ifdef IFA_STATS
1003 			struct in6_ifaddr *ia6;
1004 			ip6 = mtod(m, struct ip6_hdr *);
1005 			ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1006 			if (ia6) {
1007 				ia6->ia_ifa.ifa_data.ifad_outbytes +=
1008 					m->m_pkthdr.len;
1009 			}
1010 #endif
1011 #ifdef IPSEC
1012 			/* clean ipsec history once it goes out of the node */
1013 			ipsec_delaux(m);
1014 #endif
1015 #ifdef OLDIP6OUTPUT
1016 			error = (*ifp->if_output)(ifp, m,
1017 						  (struct sockaddr *)dst,
1018 						  ro->ro_rt);
1019 #else
1020 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1021 #endif
1022 		} else
1023 			m_freem(m);
1024 	}
1025 
1026 	if (error == 0)
1027 		ip6stat.ip6s_fragmented++;
1028 
1029 done:
1030 	if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1031 		RTFREE(ro->ro_rt);
1032 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1033 		RTFREE(ro_pmtu->ro_rt);
1034 	}
1035 
1036 #ifdef IPSEC
1037 	if (sp != NULL)
1038 		key_freesp(sp);
1039 #endif /* IPSEC */
1040 
1041 	return(error);
1042 
1043 freehdrs:
1044 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1045 	m_freem(exthdrs.ip6e_dest1);
1046 	m_freem(exthdrs.ip6e_rthdr);
1047 	m_freem(exthdrs.ip6e_dest2);
1048 	/* fall through */
1049 bad:
1050 	m_freem(m);
1051 	goto done;
1052 }
1053 
1054 static int
1055 ip6_copyexthdr(mp, hdr, hlen)
1056 	struct mbuf **mp;
1057 	caddr_t hdr;
1058 	int hlen;
1059 {
1060 	struct mbuf *m;
1061 
1062 	if (hlen > MCLBYTES)
1063 		return(ENOBUFS); /* XXX */
1064 
1065 	MGET(m, M_DONTWAIT, MT_DATA);
1066 	if (!m)
1067 		return(ENOBUFS);
1068 
1069 	if (hlen > MLEN) {
1070 		MCLGET(m, M_DONTWAIT);
1071 		if ((m->m_flags & M_EXT) == 0) {
1072 			m_free(m);
1073 			return(ENOBUFS);
1074 		}
1075 	}
1076 	m->m_len = hlen;
1077 	if (hdr)
1078 		bcopy(hdr, mtod(m, caddr_t), hlen);
1079 
1080 	*mp = m;
1081 	return(0);
1082 }
1083 
1084 /*
1085  * Insert jumbo payload option.
1086  */
1087 static int
1088 ip6_insert_jumboopt(exthdrs, plen)
1089 	struct ip6_exthdrs *exthdrs;
1090 	u_int32_t plen;
1091 {
1092 	struct mbuf *mopt;
1093 	u_char *optbuf;
1094 	u_int32_t v;
1095 
1096 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1097 
1098 	/*
1099 	 * If there is no hop-by-hop options header, allocate new one.
1100 	 * If there is one but it doesn't have enough space to store the
1101 	 * jumbo payload option, allocate a cluster to store the whole options.
1102 	 * Otherwise, use it to store the options.
1103 	 */
1104 	if (exthdrs->ip6e_hbh == 0) {
1105 		MGET(mopt, M_DONTWAIT, MT_DATA);
1106 		if (mopt == 0)
1107 			return(ENOBUFS);
1108 		mopt->m_len = JUMBOOPTLEN;
1109 		optbuf = mtod(mopt, u_char *);
1110 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1111 		exthdrs->ip6e_hbh = mopt;
1112 	} else {
1113 		struct ip6_hbh *hbh;
1114 
1115 		mopt = exthdrs->ip6e_hbh;
1116 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1117 			/*
1118 			 * XXX assumption:
1119 			 * - exthdrs->ip6e_hbh is not referenced from places
1120 			 *   other than exthdrs.
1121 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1122 			 */
1123 			int oldoptlen = mopt->m_len;
1124 			struct mbuf *n;
1125 
1126 			/*
1127 			 * XXX: give up if the whole (new) hbh header does
1128 			 * not fit even in an mbuf cluster.
1129 			 */
1130 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1131 				return(ENOBUFS);
1132 
1133 			/*
1134 			 * As a consequence, we must always prepare a cluster
1135 			 * at this point.
1136 			 */
1137 			MGET(n, M_DONTWAIT, MT_DATA);
1138 			if (n) {
1139 				MCLGET(n, M_DONTWAIT);
1140 				if ((n->m_flags & M_EXT) == 0) {
1141 					m_freem(n);
1142 					n = NULL;
1143 				}
1144 			}
1145 			if (!n)
1146 				return(ENOBUFS);
1147 			n->m_len = oldoptlen + JUMBOOPTLEN;
1148 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1149 			      oldoptlen);
1150 			optbuf = mtod(n, caddr_t) + oldoptlen;
1151 			m_freem(mopt);
1152 			exthdrs->ip6e_hbh = n;
1153 		} else {
1154 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1155 			mopt->m_len += JUMBOOPTLEN;
1156 		}
1157 		optbuf[0] = IP6OPT_PADN;
1158 		optbuf[1] = 1;
1159 
1160 		/*
1161 		 * Adjust the header length according to the pad and
1162 		 * the jumbo payload option.
1163 		 */
1164 		hbh = mtod(mopt, struct ip6_hbh *);
1165 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1166 	}
1167 
1168 	/* fill in the option. */
1169 	optbuf[2] = IP6OPT_JUMBO;
1170 	optbuf[3] = 4;
1171 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1172 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1173 
1174 	/* finally, adjust the packet header length */
1175 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1176 
1177 	return(0);
1178 #undef JUMBOOPTLEN
1179 }
1180 
1181 /*
1182  * Insert fragment header and copy unfragmentable header portions.
1183  */
1184 static int
1185 ip6_insertfraghdr(m0, m, hlen, frghdrp)
1186 	struct mbuf *m0, *m;
1187 	int hlen;
1188 	struct ip6_frag **frghdrp;
1189 {
1190 	struct mbuf *n, *mlast;
1191 
1192 	if (hlen > sizeof(struct ip6_hdr)) {
1193 		n = m_copym(m0, sizeof(struct ip6_hdr),
1194 			    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1195 		if (n == 0)
1196 			return(ENOBUFS);
1197 		m->m_next = n;
1198 	} else
1199 		n = m;
1200 
1201 	/* Search for the last mbuf of unfragmentable part. */
1202 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1203 		;
1204 
1205 	if ((mlast->m_flags & M_EXT) == 0 &&
1206 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1207 		/* use the trailing space of the last mbuf for the fragment hdr */
1208 		*frghdrp =
1209 			(struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len);
1210 		mlast->m_len += sizeof(struct ip6_frag);
1211 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1212 	} else {
1213 		/* allocate a new mbuf for the fragment header */
1214 		struct mbuf *mfrg;
1215 
1216 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1217 		if (mfrg == 0)
1218 			return(ENOBUFS);
1219 		mfrg->m_len = sizeof(struct ip6_frag);
1220 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1221 		mlast->m_next = mfrg;
1222 	}
1223 
1224 	return(0);
1225 }
1226 
1227 /*
1228  * IP6 socket option processing.
1229  */
1230 int
1231 ip6_ctloutput(op, so, level, optname, mp)
1232 	int op;
1233 	struct socket *so;
1234 	int level, optname;
1235 	struct mbuf **mp;
1236 {
1237 	struct in6pcb *in6p = sotoin6pcb(so);
1238 	struct mbuf *m = *mp;
1239 	int optval = 0;
1240 	int error = 0;
1241 	struct proc *p = curproc;	/* XXX */
1242 
1243 	if (level == IPPROTO_IPV6) {
1244 		switch (op) {
1245 
1246 		case PRCO_SETOPT:
1247 			switch (optname) {
1248 			case IPV6_PKTOPTIONS:
1249 				/* m is freed in ip6_pcbopts */
1250 				return(ip6_pcbopts(&in6p->in6p_outputopts,
1251 						   m, so));
1252 			case IPV6_HOPOPTS:
1253 			case IPV6_DSTOPTS:
1254 				if (p == 0 || suser(p->p_ucred, &p->p_acflag)) {
1255 					error = EPERM;
1256 					break;
1257 				}
1258 				/* fall through */
1259 			case IPV6_UNICAST_HOPS:
1260 			case IPV6_RECVOPTS:
1261 			case IPV6_RECVRETOPTS:
1262 			case IPV6_RECVDSTADDR:
1263 			case IPV6_PKTINFO:
1264 			case IPV6_HOPLIMIT:
1265 			case IPV6_RTHDR:
1266 			case IPV6_CHECKSUM:
1267 			case IPV6_FAITH:
1268 #ifndef INET6_BINDV6ONLY
1269 			case IPV6_BINDV6ONLY:
1270 #endif
1271 				if (!m || m->m_len != sizeof(int))
1272 					error = EINVAL;
1273 				else {
1274 					optval = *mtod(m, int *);
1275 					switch (optname) {
1276 
1277 					case IPV6_UNICAST_HOPS:
1278 						if (optval < -1 || optval >= 256)
1279 							error = EINVAL;
1280 						else {
1281 							/* -1 = kernel default */
1282 							in6p->in6p_hops = optval;
1283 						}
1284 						break;
1285 #define OPTSET(bit) \
1286 	if (optval) \
1287 		in6p->in6p_flags |= bit; \
1288 	else \
1289 		in6p->in6p_flags &= ~bit;
1290 
1291 					case IPV6_RECVOPTS:
1292 						OPTSET(IN6P_RECVOPTS);
1293 						break;
1294 
1295 					case IPV6_RECVRETOPTS:
1296 						OPTSET(IN6P_RECVRETOPTS);
1297 						break;
1298 
1299 					case IPV6_RECVDSTADDR:
1300 						OPTSET(IN6P_RECVDSTADDR);
1301 						break;
1302 
1303 					case IPV6_PKTINFO:
1304 						OPTSET(IN6P_PKTINFO);
1305 						break;
1306 
1307 					case IPV6_HOPLIMIT:
1308 						OPTSET(IN6P_HOPLIMIT);
1309 						break;
1310 
1311 					case IPV6_HOPOPTS:
1312 						OPTSET(IN6P_HOPOPTS);
1313 						break;
1314 
1315 					case IPV6_DSTOPTS:
1316 						OPTSET(IN6P_DSTOPTS);
1317 						break;
1318 
1319 					case IPV6_RTHDR:
1320 						OPTSET(IN6P_RTHDR);
1321 						break;
1322 
1323 					case IPV6_CHECKSUM:
1324 						in6p->in6p_cksum = optval;
1325 						break;
1326 
1327 					case IPV6_FAITH:
1328 						OPTSET(IN6P_FAITH);
1329 						break;
1330 
1331 #ifndef INET6_BINDV6ONLY
1332 					case IPV6_BINDV6ONLY:
1333 						OPTSET(IN6P_BINDV6ONLY);
1334 						break;
1335 #endif
1336 					}
1337 				}
1338 				break;
1339 #undef OPTSET
1340 
1341 			case IPV6_MULTICAST_IF:
1342 			case IPV6_MULTICAST_HOPS:
1343 			case IPV6_MULTICAST_LOOP:
1344 			case IPV6_JOIN_GROUP:
1345 			case IPV6_LEAVE_GROUP:
1346 				error =	ip6_setmoptions(optname, &in6p->in6p_moptions, m);
1347 				break;
1348 
1349 			case IPV6_PORTRANGE:
1350 				optval = *mtod(m, int *);
1351 
1352 				switch (optval) {
1353 				case IPV6_PORTRANGE_DEFAULT:
1354 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1355 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1356 					break;
1357 
1358 				case IPV6_PORTRANGE_HIGH:
1359 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1360 					in6p->in6p_flags |= IN6P_HIGHPORT;
1361 					break;
1362 
1363 				case IPV6_PORTRANGE_LOW:
1364 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1365 					in6p->in6p_flags |= IN6P_LOWPORT;
1366 					break;
1367 
1368 				default:
1369 					error = EINVAL;
1370 					break;
1371 				}
1372 				break;
1373 
1374 #ifdef IPSEC
1375 			case IPV6_IPSEC_POLICY:
1376 			    {
1377 				caddr_t req = NULL;
1378 				size_t len = 0;
1379 
1380 				int priv = 0;
1381 				if (p == 0 || suser(p->p_ucred, &p->p_acflag))
1382 					priv = 0;
1383 				else
1384 					priv = 1;
1385 				if (m) {
1386 					req = mtod(m, caddr_t);
1387 					len = m->m_len;
1388 				}
1389 				error = ipsec6_set_policy(in6p,
1390 				                   optname, req, len, priv);
1391 			    }
1392 				break;
1393 #endif /* IPSEC */
1394 
1395 			default:
1396 				error = ENOPROTOOPT;
1397 				break;
1398 			}
1399 			if (m)
1400 				(void)m_free(m);
1401 			break;
1402 
1403 		case PRCO_GETOPT:
1404 			switch (optname) {
1405 
1406 			case IPV6_OPTIONS:
1407 			case IPV6_RETOPTS:
1408 #if 0
1409 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1410 				if (in6p->in6p_options) {
1411 					m->m_len = in6p->in6p_options->m_len;
1412 					bcopy(mtod(in6p->in6p_options, caddr_t),
1413 					      mtod(m, caddr_t),
1414 					      (unsigned)m->m_len);
1415 				} else
1416 					m->m_len = 0;
1417 				break;
1418 #else
1419 				error = ENOPROTOOPT;
1420 				break;
1421 #endif
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 				/* fall through */
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_CHECKSUM:
1449 			case IPV6_FAITH:
1450 #ifndef INET6_BINDV6ONLY
1451 			case IPV6_BINDV6ONLY:
1452 #endif
1453 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1454 				m->m_len = sizeof(int);
1455 				switch (optname) {
1456 
1457 				case IPV6_UNICAST_HOPS:
1458 					optval = in6p->in6p_hops;
1459 					break;
1460 
1461 #define OPTBIT(bit) (in6p->in6p_flags & bit ? 1 : 0)
1462 
1463 				case IPV6_RECVOPTS:
1464 					optval = OPTBIT(IN6P_RECVOPTS);
1465 					break;
1466 
1467 				case IPV6_RECVRETOPTS:
1468 					optval = OPTBIT(IN6P_RECVRETOPTS);
1469 					break;
1470 
1471 				case IPV6_RECVDSTADDR:
1472 					optval = OPTBIT(IN6P_RECVDSTADDR);
1473 					break;
1474 
1475 				case IPV6_PORTRANGE:
1476 				    {
1477 					int flags;
1478 					flags = in6p->in6p_flags;
1479 					if (flags & IN6P_HIGHPORT)
1480 						optval = IPV6_PORTRANGE_HIGH;
1481 					else if (flags & IN6P_LOWPORT)
1482 						optval = IPV6_PORTRANGE_LOW;
1483 					else
1484 						optval = 0;
1485 					break;
1486 				    }
1487 
1488 				case IPV6_PKTINFO:
1489 					optval = OPTBIT(IN6P_PKTINFO);
1490 					break;
1491 
1492 				case IPV6_HOPLIMIT:
1493 					optval = OPTBIT(IN6P_HOPLIMIT);
1494 					break;
1495 
1496 				case IPV6_HOPOPTS:
1497 					optval = OPTBIT(IN6P_HOPOPTS);
1498 					break;
1499 
1500 				case IPV6_DSTOPTS:
1501 					optval = OPTBIT(IN6P_DSTOPTS);
1502 					break;
1503 
1504 				case IPV6_RTHDR:
1505 					optval = OPTBIT(IN6P_RTHDR);
1506 					break;
1507 
1508 				case IPV6_CHECKSUM:
1509 					optval = in6p->in6p_cksum;
1510 					break;
1511 
1512 				case IPV6_FAITH:
1513 					optval = OPTBIT(IN6P_FAITH);
1514 					break;
1515 
1516 #ifndef INET6_BINDV6ONLY
1517 				case IPV6_BINDV6ONLY:
1518 					optval = OPTBIT(IN6P_BINDV6ONLY);
1519 					break;
1520 #endif
1521 				}
1522 				*mtod(m, int *) = optval;
1523 				break;
1524 
1525 			case IPV6_MULTICAST_IF:
1526 			case IPV6_MULTICAST_HOPS:
1527 			case IPV6_MULTICAST_LOOP:
1528 			case IPV6_JOIN_GROUP:
1529 			case IPV6_LEAVE_GROUP:
1530 				error = ip6_getmoptions(optname, in6p->in6p_moptions, mp);
1531 				break;
1532 
1533 #ifdef IPSEC
1534 			case IPV6_IPSEC_POLICY:
1535 			{
1536 				caddr_t req = NULL;
1537 				size_t len = 0;
1538 
1539 				if (m) {
1540 					req = mtod(m, caddr_t);
1541 					len = m->m_len;
1542 				}
1543 				error = ipsec6_get_policy(in6p, req, len, mp);
1544 				break;
1545 			}
1546 #endif /* IPSEC */
1547 
1548 			default:
1549 				error = ENOPROTOOPT;
1550 				break;
1551 			}
1552 			break;
1553 		}
1554 	} else {
1555 		error = EINVAL;
1556 		if (op == PRCO_SETOPT && *mp)
1557 			(void)m_free(*mp);
1558 	}
1559 	return(error);
1560 }
1561 
1562 /*
1563  * Set up IP6 options in pcb for insertion in output packets.
1564  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1565  * with destination address if source routed.
1566  */
1567 static int
1568 ip6_pcbopts(pktopt, m, so)
1569 	struct ip6_pktopts **pktopt;
1570 	struct mbuf *m;
1571 	struct socket *so;
1572 {
1573 	struct ip6_pktopts *opt = *pktopt;
1574 	int error = 0;
1575 	struct proc *p = curproc;	/* XXX */
1576 	int priv = 0;
1577 
1578 	/* turn off any old options. */
1579 	if (opt) {
1580 		if (opt->ip6po_m)
1581 			(void)m_free(opt->ip6po_m);
1582 	} else
1583 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
1584 	*pktopt = 0;
1585 
1586 	if (!m || m->m_len == 0) {
1587 		/*
1588 		 * Only turning off any previous options.
1589 		 */
1590 		if (opt)
1591 			free(opt, M_IP6OPT);
1592 		if (m)
1593 			(void)m_free(m);
1594 		return(0);
1595 	}
1596 
1597 	/*  set options specified by user. */
1598 	if (p && !suser(p->p_ucred, &p->p_acflag))
1599 		priv = 1;
1600 	if ((error = ip6_setpktoptions(m, opt, priv)) != 0) {
1601 		(void)m_free(m);
1602 		return(error);
1603 	}
1604 	*pktopt = opt;
1605 	return(0);
1606 }
1607 
1608 /*
1609  * Set the IP6 multicast options in response to user setsockopt().
1610  */
1611 static int
1612 ip6_setmoptions(optname, im6op, m)
1613 	int optname;
1614 	struct ip6_moptions **im6op;
1615 	struct mbuf *m;
1616 {
1617 	int error = 0;
1618 	u_int loop, ifindex;
1619 	struct ipv6_mreq *mreq;
1620 	struct ifnet *ifp;
1621 	struct ip6_moptions *im6o = *im6op;
1622 	struct route_in6 ro;
1623 	struct sockaddr_in6 *dst;
1624 	struct in6_multi_mship *imm;
1625 	struct proc *p = curproc;	/* XXX */
1626 
1627 	if (im6o == NULL) {
1628 		/*
1629 		 * No multicast option buffer attached to the pcb;
1630 		 * allocate one and initialize to default values.
1631 		 */
1632 		im6o = (struct ip6_moptions *)
1633 			malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
1634 
1635 		if (im6o == NULL)
1636 			return(ENOBUFS);
1637 		*im6op = im6o;
1638 		im6o->im6o_multicast_ifp = NULL;
1639 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1640 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
1641 		LIST_INIT(&im6o->im6o_memberships);
1642 	}
1643 
1644 	switch (optname) {
1645 
1646 	case IPV6_MULTICAST_IF:
1647 		/*
1648 		 * Select the interface for outgoing multicast packets.
1649 		 */
1650 		if (m == NULL || m->m_len != sizeof(u_int)) {
1651 			error = EINVAL;
1652 			break;
1653 		}
1654 		bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
1655 		if (ifindex < 0 || if_index < ifindex) {
1656 			error = ENXIO;	/* XXX EINVAL? */
1657 			break;
1658 		}
1659 		ifp = ifindex2ifnet[ifindex];
1660 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1661 			error = EADDRNOTAVAIL;
1662 			break;
1663 		}
1664 		im6o->im6o_multicast_ifp = ifp;
1665 		break;
1666 
1667 	case IPV6_MULTICAST_HOPS:
1668 	    {
1669 		/*
1670 		 * Set the IP6 hoplimit for outgoing multicast packets.
1671 		 */
1672 		int optval;
1673 		if (m == NULL || m->m_len != sizeof(int)) {
1674 			error = EINVAL;
1675 			break;
1676 		}
1677 		bcopy(mtod(m, u_int *), &optval, sizeof(optval));
1678 		if (optval < -1 || optval >= 256)
1679 			error = EINVAL;
1680 		else if (optval == -1)
1681 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1682 		else
1683 			im6o->im6o_multicast_hlim = optval;
1684 		break;
1685 	    }
1686 
1687 	case IPV6_MULTICAST_LOOP:
1688 		/*
1689 		 * Set the loopback flag for outgoing multicast packets.
1690 		 * Must be zero or one.
1691 		 */
1692 		if (m == NULL || m->m_len != sizeof(u_int)) {
1693 			error = EINVAL;
1694 			break;
1695 		}
1696 		bcopy(mtod(m, u_int *), &loop, sizeof(loop));
1697 		if (loop > 1) {
1698 			error = EINVAL;
1699 			break;
1700 		}
1701 		im6o->im6o_multicast_loop = loop;
1702 		break;
1703 
1704 	case IPV6_JOIN_GROUP:
1705 		/*
1706 		 * Add a multicast group membership.
1707 		 * Group must be a valid IP6 multicast address.
1708 		 */
1709 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
1710 			error = EINVAL;
1711 			break;
1712 		}
1713 		mreq = mtod(m, struct ipv6_mreq *);
1714 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
1715 			/*
1716 			 * We use the unspecified address to specify to accept
1717 			 * all multicast addresses. Only super user is allowed
1718 			 * to do this.
1719 			 */
1720 			if (suser(p->p_ucred, &p->p_acflag))
1721 			{
1722 				error = EACCES;
1723 				break;
1724 			}
1725 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
1726 			error = EINVAL;
1727 			break;
1728 		}
1729 
1730 		/*
1731 		 * If the interface is specified, validate it.
1732 		 */
1733 		if (mreq->ipv6mr_interface < 0
1734 		 || if_index < mreq->ipv6mr_interface) {
1735 			error = ENXIO;	/* XXX EINVAL? */
1736 			break;
1737 		}
1738 		/*
1739 		 * If no interface was explicitly specified, choose an
1740 		 * appropriate one according to the given multicast address.
1741 		 */
1742 		if (mreq->ipv6mr_interface == 0) {
1743 			/*
1744 			 * If the multicast address is in node-local scope,
1745 			 * the interface should be a loopback interface.
1746 			 * Otherwise, look up the routing table for the
1747 			 * address, and choose the outgoing interface.
1748 			 *   XXX: is it a good approach?
1749 			 */
1750 			if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
1751 				ifp = &loif[0];
1752 			} else {
1753 				ro.ro_rt = NULL;
1754 				dst = (struct sockaddr_in6 *)&ro.ro_dst;
1755 				bzero(dst, sizeof(*dst));
1756 				dst->sin6_len = sizeof(struct sockaddr_in6);
1757 				dst->sin6_family = AF_INET6;
1758 				dst->sin6_addr = mreq->ipv6mr_multiaddr;
1759 				rtalloc((struct route *)&ro);
1760 				if (ro.ro_rt == NULL) {
1761 					error = EADDRNOTAVAIL;
1762 					break;
1763 				}
1764 				ifp = ro.ro_rt->rt_ifp;
1765 				rtfree(ro.ro_rt);
1766 			}
1767 		} else
1768 			ifp = ifindex2ifnet[mreq->ipv6mr_interface];
1769 
1770 		/*
1771 		 * See if we found an interface, and confirm that it
1772 		 * supports multicast
1773 		 */
1774 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1775 			error = EADDRNOTAVAIL;
1776 			break;
1777 		}
1778 		/*
1779 		 * Put interface index into the multicast address,
1780 		 * if the address has link-local scope.
1781 		 */
1782 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
1783 			mreq->ipv6mr_multiaddr.s6_addr16[1]
1784 				= htons(mreq->ipv6mr_interface);
1785 		}
1786 		/*
1787 		 * See if the membership already exists.
1788 		 */
1789 		for (imm = im6o->im6o_memberships.lh_first;
1790 		     imm != NULL; imm = imm->i6mm_chain.le_next)
1791 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
1792 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
1793 					       &mreq->ipv6mr_multiaddr))
1794 				break;
1795 		if (imm != NULL) {
1796 			error = EADDRINUSE;
1797 			break;
1798 		}
1799 		/*
1800 		 * Everything looks good; add a new record to the multicast
1801 		 * address list for the given interface.
1802 		 */
1803 		imm = malloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
1804 		if (imm == NULL) {
1805 			error = ENOBUFS;
1806 			break;
1807 		}
1808 		if ((imm->i6mm_maddr =
1809 		     in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
1810 			free(imm, M_IPMADDR);
1811 			break;
1812 		}
1813 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
1814 		break;
1815 
1816 	case IPV6_LEAVE_GROUP:
1817 		/*
1818 		 * Drop a multicast group membership.
1819 		 * Group must be a valid IP6 multicast address.
1820 		 */
1821 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
1822 			error = EINVAL;
1823 			break;
1824 		}
1825 		mreq = mtod(m, struct ipv6_mreq *);
1826 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
1827 			if (suser(p->p_ucred, &p->p_acflag))
1828 			{
1829 				error = EACCES;
1830 				break;
1831 			}
1832 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
1833 			error = EINVAL;
1834 			break;
1835 		}
1836 		/*
1837 		 * If an interface address was specified, get a pointer
1838 		 * to its ifnet structure.
1839 		 */
1840 		if (mreq->ipv6mr_interface < 0
1841 		 || if_index < mreq->ipv6mr_interface) {
1842 			error = ENXIO;	/* XXX EINVAL? */
1843 			break;
1844 		}
1845 		ifp = ifindex2ifnet[mreq->ipv6mr_interface];
1846 		/*
1847 		 * Put interface index into the multicast address,
1848 		 * if the address has link-local scope.
1849 		 */
1850 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
1851 			mreq->ipv6mr_multiaddr.s6_addr16[1]
1852 				= htons(mreq->ipv6mr_interface);
1853 		}
1854 		/*
1855 		 * Find the membership in the membership list.
1856 		 */
1857 		for (imm = im6o->im6o_memberships.lh_first;
1858 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
1859 			if ((ifp == NULL ||
1860 			     imm->i6mm_maddr->in6m_ifp == ifp) &&
1861 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
1862 					       &mreq->ipv6mr_multiaddr))
1863 				break;
1864 		}
1865 		if (imm == NULL) {
1866 			/* Unable to resolve interface */
1867 			error = EADDRNOTAVAIL;
1868 			break;
1869 		}
1870 		/*
1871 		 * Give up the multicast address record to which the
1872 		 * membership points.
1873 		 */
1874 		LIST_REMOVE(imm, i6mm_chain);
1875 		in6_delmulti(imm->i6mm_maddr);
1876 		free(imm, M_IPMADDR);
1877 		break;
1878 
1879 	default:
1880 		error = EOPNOTSUPP;
1881 		break;
1882 	}
1883 
1884 	/*
1885 	 * If all options have default values, no need to keep the mbuf.
1886 	 */
1887 	if (im6o->im6o_multicast_ifp == NULL &&
1888 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
1889 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
1890 	    im6o->im6o_memberships.lh_first == NULL) {
1891 		free(*im6op, M_IPMOPTS);
1892 		*im6op = NULL;
1893 	}
1894 
1895 	return(error);
1896 }
1897 
1898 /*
1899  * Return the IP6 multicast options in response to user getsockopt().
1900  */
1901 static int
1902 ip6_getmoptions(optname, im6o, mp)
1903 	int optname;
1904 	struct ip6_moptions *im6o;
1905 	struct mbuf **mp;
1906 {
1907 	u_int *hlim, *loop, *ifindex;
1908 
1909 	*mp = m_get(M_WAIT, MT_SOOPTS);
1910 
1911 	switch (optname) {
1912 
1913 	case IPV6_MULTICAST_IF:
1914 		ifindex = mtod(*mp, u_int *);
1915 		(*mp)->m_len = sizeof(u_int);
1916 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
1917 			*ifindex = 0;
1918 		else
1919 			*ifindex = im6o->im6o_multicast_ifp->if_index;
1920 		return(0);
1921 
1922 	case IPV6_MULTICAST_HOPS:
1923 		hlim = mtod(*mp, u_int *);
1924 		(*mp)->m_len = sizeof(u_int);
1925 		if (im6o == NULL)
1926 			*hlim = ip6_defmcasthlim;
1927 		else
1928 			*hlim = im6o->im6o_multicast_hlim;
1929 		return(0);
1930 
1931 	case IPV6_MULTICAST_LOOP:
1932 		loop = mtod(*mp, u_int *);
1933 		(*mp)->m_len = sizeof(u_int);
1934 		if (im6o == NULL)
1935 			*loop = ip6_defmcasthlim;
1936 		else
1937 			*loop = im6o->im6o_multicast_loop;
1938 		return(0);
1939 
1940 	default:
1941 		return(EOPNOTSUPP);
1942 	}
1943 }
1944 
1945 /*
1946  * Discard the IP6 multicast options.
1947  */
1948 void
1949 ip6_freemoptions(im6o)
1950 	struct ip6_moptions *im6o;
1951 {
1952 	struct in6_multi_mship *imm;
1953 
1954 	if (im6o == NULL)
1955 		return;
1956 
1957 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
1958 		LIST_REMOVE(imm, i6mm_chain);
1959 		if (imm->i6mm_maddr)
1960 			in6_delmulti(imm->i6mm_maddr);
1961 		free(imm, M_IPMADDR);
1962 	}
1963 	free(im6o, M_IPMOPTS);
1964 }
1965 
1966 /*
1967  * Set IPv6 outgoing packet options based on advanced API.
1968  */
1969 int
1970 ip6_setpktoptions(control, opt, priv)
1971 	struct mbuf *control;
1972 	struct ip6_pktopts *opt;
1973 	int priv;
1974 {
1975 	struct cmsghdr *cm = 0;
1976 
1977 	if (control == 0 || opt == 0)
1978 		return(EINVAL);
1979 
1980 	bzero(opt, sizeof(*opt));
1981 	opt->ip6po_hlim = -1; /* -1 means to use default hop limit */
1982 
1983 	/*
1984 	 * XXX: Currently, we assume all the optional information is stored
1985 	 * in a single mbuf.
1986 	 */
1987 	if (control->m_next)
1988 		return(EINVAL);
1989 
1990 	opt->ip6po_m = control;
1991 
1992 	for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
1993 		     control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
1994 		cm = mtod(control, struct cmsghdr *);
1995 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
1996 			return(EINVAL);
1997 		if (cm->cmsg_level != IPPROTO_IPV6)
1998 			continue;
1999 
2000 		switch (cm->cmsg_type) {
2001 		case IPV6_PKTINFO:
2002 			if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo)))
2003 				return(EINVAL);
2004 			opt->ip6po_pktinfo = (struct in6_pktinfo *)CMSG_DATA(cm);
2005 			if (opt->ip6po_pktinfo->ipi6_ifindex &&
2006 			    IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr))
2007 				opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] =
2008 					htons(opt->ip6po_pktinfo->ipi6_ifindex);
2009 
2010 			if (opt->ip6po_pktinfo->ipi6_ifindex > if_index
2011 			 || opt->ip6po_pktinfo->ipi6_ifindex < 0) {
2012 				return(ENXIO);
2013 			}
2014 
2015 			/*
2016 			 * Check if the requested source address is indeed a
2017 			 * unicast address assigned to the node.
2018 			 */
2019 			if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
2020 				struct ifaddr *ia;
2021 				struct sockaddr_in6 sin6;
2022 
2023 				bzero(&sin6, sizeof(sin6));
2024 				sin6.sin6_len = sizeof(sin6);
2025 				sin6.sin6_family = AF_INET6;
2026 				sin6.sin6_addr =
2027 					opt->ip6po_pktinfo->ipi6_addr;
2028 				ia = ifa_ifwithaddr(sin6tosa(&sin6));
2029 				if (ia == NULL ||
2030 				    (opt->ip6po_pktinfo->ipi6_ifindex &&
2031 				     (ia->ifa_ifp->if_index !=
2032 				      opt->ip6po_pktinfo->ipi6_ifindex))) {
2033 					return(EADDRNOTAVAIL);
2034 				}
2035 				/*
2036 				 * Check if the requested source address is
2037 				 * indeed a unicast address assigned to the
2038 				 * node.
2039 				 */
2040 				if (IN6_IS_ADDR_MULTICAST(&opt->ip6po_pktinfo->ipi6_addr))
2041 					return(EADDRNOTAVAIL);
2042 			}
2043 			break;
2044 
2045 		case IPV6_HOPLIMIT:
2046 			if (cm->cmsg_len != CMSG_LEN(sizeof(int)))
2047 				return(EINVAL);
2048 
2049 			bcopy(CMSG_DATA(cm), &opt->ip6po_hlim,
2050 			    sizeof(opt->ip6po_hlim));
2051 			if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255)
2052 				return(EINVAL);
2053 			break;
2054 
2055 		case IPV6_NEXTHOP:
2056 			if (!priv)
2057 				return(EPERM);
2058 
2059 			if (cm->cmsg_len < sizeof(u_char) ||
2060 			    /* check if cmsg_len is large enough for sa_len */
2061 			    cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm)))
2062 				return(EINVAL);
2063 
2064 			opt->ip6po_nexthop = (struct sockaddr *)CMSG_DATA(cm);
2065 
2066 			break;
2067 
2068 		case IPV6_HOPOPTS:
2069 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh)))
2070 				return(EINVAL);
2071 			opt->ip6po_hbh = (struct ip6_hbh *)CMSG_DATA(cm);
2072 			if (cm->cmsg_len !=
2073 			    CMSG_LEN((opt->ip6po_hbh->ip6h_len + 1) << 3))
2074 				return(EINVAL);
2075 			break;
2076 
2077 		case IPV6_DSTOPTS:
2078 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest)))
2079 				return(EINVAL);
2080 
2081 			/*
2082 			 * If there is no routing header yet, the destination
2083 			 * options header should be put on the 1st part.
2084 			 * Otherwise, the header should be on the 2nd part.
2085 			 * (See RFC 2460, section 4.1)
2086 			 */
2087 			if (opt->ip6po_rthdr == NULL) {
2088 				opt->ip6po_dest1 =
2089 					(struct ip6_dest *)CMSG_DATA(cm);
2090 				if (cm->cmsg_len !=
2091 				    CMSG_LEN((opt->ip6po_dest1->ip6d_len + 1)
2092 					     << 3))
2093 					return(EINVAL);
2094 			}
2095 			else {
2096 				opt->ip6po_dest2 =
2097 					(struct ip6_dest *)CMSG_DATA(cm);
2098 				if (cm->cmsg_len !=
2099 				    CMSG_LEN((opt->ip6po_dest2->ip6d_len + 1)
2100 					     << 3))
2101 					return(EINVAL);
2102 			}
2103 			break;
2104 
2105 		case IPV6_RTHDR:
2106 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr)))
2107 				return(EINVAL);
2108 			opt->ip6po_rthdr = (struct ip6_rthdr *)CMSG_DATA(cm);
2109 			if (cm->cmsg_len !=
2110 			    CMSG_LEN((opt->ip6po_rthdr->ip6r_len + 1) << 3))
2111 				return(EINVAL);
2112 			switch (opt->ip6po_rthdr->ip6r_type) {
2113 			case IPV6_RTHDR_TYPE_0:
2114 				if (opt->ip6po_rthdr->ip6r_segleft == 0)
2115 					return(EINVAL);
2116 				break;
2117 			default:
2118 				return(EINVAL);
2119 			}
2120 			break;
2121 
2122 		default:
2123 			return(ENOPROTOOPT);
2124 		}
2125 	}
2126 
2127 	return(0);
2128 }
2129 
2130 /*
2131  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2132  * packet to the input queue of a specified interface.  Note that this
2133  * calls the output routine of the loopback "driver", but with an interface
2134  * pointer that might NOT be &loif -- easier than replicating that code here.
2135  */
2136 void
2137 ip6_mloopback(ifp, m, dst)
2138 	struct ifnet *ifp;
2139 	struct mbuf *m;
2140 	struct sockaddr_in6 *dst;
2141 {
2142 	struct mbuf *copym;
2143 	struct ip6_hdr *ip6;
2144 
2145 	copym = m_copy(m, 0, M_COPYALL);
2146 	if (copym == NULL)
2147 		return;
2148 
2149 	/*
2150 	 * Make sure to deep-copy IPv6 header portion in case the data
2151 	 * is in an mbuf cluster, so that we can safely override the IPv6
2152 	 * header portion later.
2153 	 */
2154 	if ((copym->m_flags & M_EXT) != 0 ||
2155 	    copym->m_len < sizeof(struct ip6_hdr)) {
2156 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
2157 		if (copym == NULL)
2158 			return;
2159 	}
2160 
2161 #ifdef DIAGNOSTIC
2162 	if (copym->m_len < sizeof(*ip6)) {
2163 		m_freem(copym);
2164 		return;
2165 	}
2166 #endif
2167 
2168 #ifndef FAKE_LOOPBACK_IF
2169 	if ((ifp->if_flags & IFF_LOOPBACK) == 0)
2170 #else
2171 	if (1)
2172 #endif
2173 	{
2174 		ip6 = mtod(copym, struct ip6_hdr *);
2175 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
2176 			ip6->ip6_src.s6_addr16[1] = 0;
2177 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
2178 			ip6->ip6_dst.s6_addr16[1] = 0;
2179 	}
2180 
2181 	(void)looutput(ifp, copym, (struct sockaddr *)dst, NULL);
2182 }
2183 
2184 /*
2185  * Chop IPv6 header off from the payload.
2186  */
2187 static int
2188 ip6_splithdr(m, exthdrs)
2189 	struct mbuf *m;
2190 	struct ip6_exthdrs *exthdrs;
2191 {
2192 	struct mbuf *mh;
2193 	struct ip6_hdr *ip6;
2194 
2195 	ip6 = mtod(m, struct ip6_hdr *);
2196 	if (m->m_len > sizeof(*ip6)) {
2197 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
2198 		if (mh == 0) {
2199 			m_freem(m);
2200 			return ENOBUFS;
2201 		}
2202 		M_COPY_PKTHDR(mh, m);
2203 		MH_ALIGN(mh, sizeof(*ip6));
2204 		m->m_flags &= ~M_PKTHDR;
2205 		m->m_len -= sizeof(*ip6);
2206 		m->m_data += sizeof(*ip6);
2207 		mh->m_next = m;
2208 		m = mh;
2209 		m->m_len = sizeof(*ip6);
2210 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2211 	}
2212 	exthdrs->ip6e_ip6 = m;
2213 	return 0;
2214 }
2215 
2216 /*
2217  * Compute IPv6 extension header length.
2218  */
2219 int
2220 ip6_optlen(in6p)
2221 	struct in6pcb *in6p;
2222 {
2223 	int len;
2224 
2225 	if (!in6p->in6p_outputopts)
2226 		return 0;
2227 
2228 	len = 0;
2229 #define elen(x) \
2230     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2231 
2232 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
2233 	len += elen(in6p->in6p_outputopts->ip6po_dest1);
2234 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
2235 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
2236 	return len;
2237 #undef elen
2238 }
2239