xref: /netbsd-src/sys/netinet6/ip6_output.c (revision 5aefcfdc06931dd97e76246d2fe0302f7b3fe094)
1 /*	$NetBSD: ip6_output.c,v 1.27 2000/11/11 00:52:39 thorpej Exp $	*/
2 /*	$KAME: ip6_output.c,v 1.122 2000/08/19 02:12:02 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 	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 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
772 			origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
773 		else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
774 			origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
775 		else
776 			origifp = ifp;
777 	}
778 	else
779 		origifp = ifp;
780 #ifndef FAKE_LOOPBACK_IF
781 	if ((ifp->if_flags & IFF_LOOPBACK) == 0)
782 #else
783 	if (1)
784 #endif
785 	{
786 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
787 			ip6->ip6_src.s6_addr16[1] = 0;
788 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
789 			ip6->ip6_dst.s6_addr16[1] = 0;
790 	}
791 
792 	/*
793 	 * If the outgoing packet contains a hop-by-hop options header,
794 	 * it must be examined and processed even by the source node.
795 	 * (RFC 2460, section 4.)
796 	 */
797 	if (exthdrs.ip6e_hbh) {
798 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh,
799 					   struct ip6_hbh *);
800 		u_int32_t dummy1; /* XXX unused */
801 		u_int32_t dummy2; /* XXX unused */
802 
803 		/*
804 		 *  XXX: if we have to send an ICMPv6 error to the sender,
805 		 *       we need the M_LOOP flag since icmp6_error() expects
806 		 *       the IPv6 and the hop-by-hop options header are
807 		 *       continuous unless the flag is set.
808 		 */
809 		m->m_flags |= M_LOOP;
810 		m->m_pkthdr.rcvif = ifp;
811 		if (ip6_process_hopopts(m,
812 					(u_int8_t *)(hbh + 1),
813 					((hbh->ip6h_len + 1) << 3) -
814 					sizeof(struct ip6_hbh),
815 					&dummy1, &dummy2) < 0) {
816 			/* m was already freed at this point */
817 			error = EINVAL;/* better error? */
818 			goto done;
819 		}
820 		m->m_flags &= ~M_LOOP; /* XXX */
821 		m->m_pkthdr.rcvif = NULL;
822 	}
823 
824 #ifdef PFIL_HOOKS
825 	/*
826 	 * Run through list of hooks for output packets.
827 	 */
828 	if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp,
829 				    PFIL_OUT)) != 0)
830 		goto done;
831 	if (m == NULL)
832 		goto done;
833 	ip6 = mtod(m, struct ip6_hdr *);
834 #endif /* PFIL_HOOKS */
835 	/*
836 	 * Send the packet to the outgoing interface.
837 	 * If necessary, do IPv6 fragmentation before sending.
838 	 */
839 	tlen = m->m_pkthdr.len;
840 	if (tlen <= mtu
841 #ifdef notyet
842 	    /*
843 	     * On any link that cannot convey a 1280-octet packet in one piece,
844 	     * link-specific fragmentation and reassembly must be provided at
845 	     * a layer below IPv6. [RFC 2460, sec.5]
846 	     * Thus if the interface has ability of link-level fragmentation,
847 	     * we can just send the packet even if the packet size is
848 	     * larger than the link's MTU.
849 	     * XXX: IFF_FRAGMENTABLE (or such) flag has not been defined yet...
850 	     */
851 
852 	    || ifp->if_flags & IFF_FRAGMENTABLE
853 #endif
854 	    )
855 	{
856 #ifdef IFA_STATS
857 		struct in6_ifaddr *ia6;
858 		ip6 = mtod(m, struct ip6_hdr *);
859 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
860 		if (ia6) {
861 			ia6->ia_ifa.ifa_data.ifad_outbytes +=
862 				m->m_pkthdr.len;
863 		}
864 #endif
865 #ifdef OLDIP6OUTPUT
866 		error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst,
867 					  ro->ro_rt);
868 #else
869 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
870 #endif
871 		goto done;
872 	} else if (mtu < IPV6_MMTU) {
873 		/*
874 		 * note that path MTU is never less than IPV6_MMTU
875 		 * (see icmp6_input).
876 		 */
877 		error = EMSGSIZE;
878 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
879 		goto bad;
880 	} else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
881 		error = EMSGSIZE;
882 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
883 		goto bad;
884 	} else {
885 		struct mbuf **mnext, *m_frgpart;
886 		struct ip6_frag *ip6f;
887 		u_int32_t id = htonl(ip6_id++);
888 		u_char nextproto;
889 
890 		/*
891 		 * Too large for the destination or interface;
892 		 * fragment if possible.
893 		 * Must be able to put at least 8 bytes per fragment.
894 		 */
895 		hlen = unfragpartlen;
896 		if (mtu > IPV6_MAXPACKET)
897 			mtu = IPV6_MAXPACKET;
898 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
899 		if (len < 8) {
900 			error = EMSGSIZE;
901 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
902 			goto bad;
903 		}
904 
905 		mnext = &m->m_nextpkt;
906 
907 		/*
908 		 * Change the next header field of the last header in the
909 		 * unfragmentable part.
910 		 */
911 		if (exthdrs.ip6e_rthdr) {
912 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
913 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
914 		} else if (exthdrs.ip6e_dest1) {
915 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
916 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
917 		} else if (exthdrs.ip6e_hbh) {
918 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
919 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
920 		} else {
921 			nextproto = ip6->ip6_nxt;
922 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
923 		}
924 
925 		/*
926 		 * Loop through length of segment after first fragment,
927 		 * make new header and copy data of each part and link onto chain.
928 		 */
929 		m0 = m;
930 		for (off = hlen; off < tlen; off += len) {
931 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
932 			if (!m) {
933 				error = ENOBUFS;
934 				ip6stat.ip6s_odropped++;
935 				goto sendorfree;
936 			}
937 			m->m_flags = m0->m_flags & M_COPYFLAGS;
938 			*mnext = m;
939 			mnext = &m->m_nextpkt;
940 			m->m_data += max_linkhdr;
941 			mhip6 = mtod(m, struct ip6_hdr *);
942 			*mhip6 = *ip6;
943 			m->m_len = sizeof(*mhip6);
944  			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
945  			if (error) {
946 				ip6stat.ip6s_odropped++;
947 				goto sendorfree;
948 			}
949 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
950 			if (off + len >= tlen)
951 				len = tlen - off;
952 			else
953 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
954 			mhip6->ip6_plen = htons((u_short)(len + hlen +
955 							  sizeof(*ip6f) -
956 							  sizeof(struct ip6_hdr)));
957 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
958 				error = ENOBUFS;
959 				ip6stat.ip6s_odropped++;
960 				goto sendorfree;
961 			}
962 			m_cat(m, m_frgpart);
963 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
964 			m->m_pkthdr.rcvif = (struct ifnet *)0;
965 			ip6f->ip6f_reserved = 0;
966 			ip6f->ip6f_ident = id;
967 			ip6f->ip6f_nxt = nextproto;
968 			ip6stat.ip6s_ofragments++;
969 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
970 		}
971 
972 		in6_ifstat_inc(ifp, ifs6_out_fragok);
973 	}
974 
975 	/*
976 	 * Remove leading garbages.
977 	 */
978 sendorfree:
979 	m = m0->m_nextpkt;
980 	m0->m_nextpkt = 0;
981 	m_freem(m0);
982 	for (m0 = m; m; m = m0) {
983 		m0 = m->m_nextpkt;
984 		m->m_nextpkt = 0;
985 		if (error == 0) {
986 #ifdef IFA_STATS
987 			struct in6_ifaddr *ia6;
988 			ip6 = mtod(m, struct ip6_hdr *);
989 			ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
990 			if (ia6) {
991 				ia6->ia_ifa.ifa_data.ifad_outbytes +=
992 					m->m_pkthdr.len;
993 			}
994 #endif
995 #ifdef OLDIP6OUTPUT
996 			error = (*ifp->if_output)(ifp, m,
997 						  (struct sockaddr *)dst,
998 						  ro->ro_rt);
999 #else
1000 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1001 #endif
1002 		} else
1003 			m_freem(m);
1004 	}
1005 
1006 	if (error == 0)
1007 		ip6stat.ip6s_fragmented++;
1008 
1009 done:
1010 	if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1011 		RTFREE(ro->ro_rt);
1012 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1013 		RTFREE(ro_pmtu->ro_rt);
1014 	}
1015 
1016 #ifdef IPSEC
1017 	if (sp != NULL)
1018 		key_freesp(sp);
1019 #endif /* IPSEC */
1020 
1021 	return(error);
1022 
1023 freehdrs:
1024 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1025 	m_freem(exthdrs.ip6e_dest1);
1026 	m_freem(exthdrs.ip6e_rthdr);
1027 	m_freem(exthdrs.ip6e_dest2);
1028 	/* fall through */
1029 bad:
1030 	m_freem(m);
1031 	goto done;
1032 }
1033 
1034 static int
1035 ip6_copyexthdr(mp, hdr, hlen)
1036 	struct mbuf **mp;
1037 	caddr_t hdr;
1038 	int hlen;
1039 {
1040 	struct mbuf *m;
1041 
1042 	if (hlen > MCLBYTES)
1043 		return(ENOBUFS); /* XXX */
1044 
1045 	MGET(m, M_DONTWAIT, MT_DATA);
1046 	if (!m)
1047 		return(ENOBUFS);
1048 
1049 	if (hlen > MLEN) {
1050 		MCLGET(m, M_DONTWAIT);
1051 		if ((m->m_flags & M_EXT) == 0) {
1052 			m_free(m);
1053 			return(ENOBUFS);
1054 		}
1055 	}
1056 	m->m_len = hlen;
1057 	if (hdr)
1058 		bcopy(hdr, mtod(m, caddr_t), hlen);
1059 
1060 	*mp = m;
1061 	return(0);
1062 }
1063 
1064 /*
1065  * Insert jumbo payload option.
1066  */
1067 static int
1068 ip6_insert_jumboopt(exthdrs, plen)
1069 	struct ip6_exthdrs *exthdrs;
1070 	u_int32_t plen;
1071 {
1072 	struct mbuf *mopt;
1073 	u_char *optbuf;
1074 	u_int32_t v;
1075 
1076 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1077 
1078 	/*
1079 	 * If there is no hop-by-hop options header, allocate new one.
1080 	 * If there is one but it doesn't have enough space to store the
1081 	 * jumbo payload option, allocate a cluster to store the whole options.
1082 	 * Otherwise, use it to store the options.
1083 	 */
1084 	if (exthdrs->ip6e_hbh == 0) {
1085 		MGET(mopt, M_DONTWAIT, MT_DATA);
1086 		if (mopt == 0)
1087 			return(ENOBUFS);
1088 		mopt->m_len = JUMBOOPTLEN;
1089 		optbuf = mtod(mopt, u_char *);
1090 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1091 		exthdrs->ip6e_hbh = mopt;
1092 	} else {
1093 		struct ip6_hbh *hbh;
1094 
1095 		mopt = exthdrs->ip6e_hbh;
1096 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1097 			/*
1098 			 * XXX assumption:
1099 			 * - exthdrs->ip6e_hbh is not referenced from places
1100 			 *   other than exthdrs.
1101 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1102 			 */
1103 			int oldoptlen = mopt->m_len;
1104 			struct mbuf *n;
1105 
1106 			/*
1107 			 * XXX: give up if the whole (new) hbh header does
1108 			 * not fit even in an mbuf cluster.
1109 			 */
1110 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1111 				return(ENOBUFS);
1112 
1113 			/*
1114 			 * As a consequence, we must always prepare a cluster
1115 			 * at this point.
1116 			 */
1117 			MGET(n, M_DONTWAIT, MT_DATA);
1118 			if (n) {
1119 				MCLGET(n, M_DONTWAIT);
1120 				if ((n->m_flags & M_EXT) == 0) {
1121 					m_freem(n);
1122 					n = NULL;
1123 				}
1124 			}
1125 			if (!n)
1126 				return(ENOBUFS);
1127 			n->m_len = oldoptlen + JUMBOOPTLEN;
1128 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1129 			      oldoptlen);
1130 			optbuf = mtod(n, caddr_t) + oldoptlen;
1131 			m_freem(mopt);
1132 			exthdrs->ip6e_hbh = n;
1133 		} else {
1134 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1135 			mopt->m_len += JUMBOOPTLEN;
1136 		}
1137 		optbuf[0] = IP6OPT_PADN;
1138 		optbuf[1] = 1;
1139 
1140 		/*
1141 		 * Adjust the header length according to the pad and
1142 		 * the jumbo payload option.
1143 		 */
1144 		hbh = mtod(mopt, struct ip6_hbh *);
1145 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1146 	}
1147 
1148 	/* fill in the option. */
1149 	optbuf[2] = IP6OPT_JUMBO;
1150 	optbuf[3] = 4;
1151 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1152 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1153 
1154 	/* finally, adjust the packet header length */
1155 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1156 
1157 	return(0);
1158 #undef JUMBOOPTLEN
1159 }
1160 
1161 /*
1162  * Insert fragment header and copy unfragmentable header portions.
1163  */
1164 static int
1165 ip6_insertfraghdr(m0, m, hlen, frghdrp)
1166 	struct mbuf *m0, *m;
1167 	int hlen;
1168 	struct ip6_frag **frghdrp;
1169 {
1170 	struct mbuf *n, *mlast;
1171 
1172 	if (hlen > sizeof(struct ip6_hdr)) {
1173 		n = m_copym(m0, sizeof(struct ip6_hdr),
1174 			    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1175 		if (n == 0)
1176 			return(ENOBUFS);
1177 		m->m_next = n;
1178 	} else
1179 		n = m;
1180 
1181 	/* Search for the last mbuf of unfragmentable part. */
1182 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1183 		;
1184 
1185 	if ((mlast->m_flags & M_EXT) == 0 &&
1186 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1187 		/* use the trailing space of the last mbuf for the fragment hdr */
1188 		*frghdrp =
1189 			(struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len);
1190 		mlast->m_len += sizeof(struct ip6_frag);
1191 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1192 	} else {
1193 		/* allocate a new mbuf for the fragment header */
1194 		struct mbuf *mfrg;
1195 
1196 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1197 		if (mfrg == 0)
1198 			return(ENOBUFS);
1199 		mfrg->m_len = sizeof(struct ip6_frag);
1200 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1201 		mlast->m_next = mfrg;
1202 	}
1203 
1204 	return(0);
1205 }
1206 
1207 /*
1208  * IP6 socket option processing.
1209  */
1210 int
1211 ip6_ctloutput(op, so, level, optname, mp)
1212 	int op;
1213 	struct socket *so;
1214 	int level, optname;
1215 	struct mbuf **mp;
1216 {
1217 	register struct in6pcb *in6p = sotoin6pcb(so);
1218 	register struct mbuf *m = *mp;
1219 	register int optval = 0;
1220 	int error = 0;
1221 	struct proc *p = curproc;	/* XXX */
1222 
1223 	if (level == IPPROTO_IPV6) {
1224 		switch (op) {
1225 
1226 		case PRCO_SETOPT:
1227 			switch (optname) {
1228 			case IPV6_PKTOPTIONS:
1229 				/* m is freed in ip6_pcbopts */
1230 				return(ip6_pcbopts(&in6p->in6p_outputopts,
1231 						   m, so));
1232 			case IPV6_HOPOPTS:
1233 			case IPV6_DSTOPTS:
1234 				if (p == 0 || suser(p->p_ucred, &p->p_acflag)) {
1235 					error = EPERM;
1236 					break;
1237 				}
1238 				/* fall through */
1239 			case IPV6_UNICAST_HOPS:
1240 			case IPV6_RECVOPTS:
1241 			case IPV6_RECVRETOPTS:
1242 			case IPV6_RECVDSTADDR:
1243 			case IPV6_PKTINFO:
1244 			case IPV6_HOPLIMIT:
1245 			case IPV6_RTHDR:
1246 			case IPV6_CHECKSUM:
1247 			case IPV6_FAITH:
1248 #ifndef INET6_BINDV6ONLY
1249 			case IPV6_BINDV6ONLY:
1250 #endif
1251 				if (!m || m->m_len != sizeof(int))
1252 					error = EINVAL;
1253 				else {
1254 					optval = *mtod(m, int *);
1255 					switch (optname) {
1256 
1257 					case IPV6_UNICAST_HOPS:
1258 						if (optval < -1 || optval >= 256)
1259 							error = EINVAL;
1260 						else {
1261 							/* -1 = kernel default */
1262 							in6p->in6p_hops = optval;
1263 						}
1264 						break;
1265 #define OPTSET(bit) \
1266 	if (optval) \
1267 		in6p->in6p_flags |= bit; \
1268 	else \
1269 		in6p->in6p_flags &= ~bit;
1270 
1271 					case IPV6_RECVOPTS:
1272 						OPTSET(IN6P_RECVOPTS);
1273 						break;
1274 
1275 					case IPV6_RECVRETOPTS:
1276 						OPTSET(IN6P_RECVRETOPTS);
1277 						break;
1278 
1279 					case IPV6_RECVDSTADDR:
1280 						OPTSET(IN6P_RECVDSTADDR);
1281 						break;
1282 
1283 					case IPV6_PKTINFO:
1284 						OPTSET(IN6P_PKTINFO);
1285 						break;
1286 
1287 					case IPV6_HOPLIMIT:
1288 						OPTSET(IN6P_HOPLIMIT);
1289 						break;
1290 
1291 					case IPV6_HOPOPTS:
1292 						OPTSET(IN6P_HOPOPTS);
1293 						break;
1294 
1295 					case IPV6_DSTOPTS:
1296 						OPTSET(IN6P_DSTOPTS);
1297 						break;
1298 
1299 					case IPV6_RTHDR:
1300 						OPTSET(IN6P_RTHDR);
1301 						break;
1302 
1303 					case IPV6_CHECKSUM:
1304 						in6p->in6p_cksum = optval;
1305 						break;
1306 
1307 					case IPV6_FAITH:
1308 						OPTSET(IN6P_FAITH);
1309 						break;
1310 
1311 #ifndef INET6_BINDV6ONLY
1312 					case IPV6_BINDV6ONLY:
1313 						OPTSET(IN6P_BINDV6ONLY);
1314 						break;
1315 #endif
1316 					}
1317 				}
1318 				break;
1319 #undef OPTSET
1320 
1321 			case IPV6_MULTICAST_IF:
1322 			case IPV6_MULTICAST_HOPS:
1323 			case IPV6_MULTICAST_LOOP:
1324 			case IPV6_JOIN_GROUP:
1325 			case IPV6_LEAVE_GROUP:
1326 				error =	ip6_setmoptions(optname, &in6p->in6p_moptions, m);
1327 				break;
1328 
1329 			case IPV6_PORTRANGE:
1330 				optval = *mtod(m, int *);
1331 
1332 				switch (optval) {
1333 				case IPV6_PORTRANGE_DEFAULT:
1334 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1335 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1336 					break;
1337 
1338 				case IPV6_PORTRANGE_HIGH:
1339 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1340 					in6p->in6p_flags |= IN6P_HIGHPORT;
1341 					break;
1342 
1343 				case IPV6_PORTRANGE_LOW:
1344 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1345 					in6p->in6p_flags |= IN6P_LOWPORT;
1346 					break;
1347 
1348 				default:
1349 					error = EINVAL;
1350 					break;
1351 				}
1352 				break;
1353 
1354 #ifdef IPSEC
1355 			case IPV6_IPSEC_POLICY:
1356 			    {
1357 				caddr_t req = NULL;
1358 				size_t len = 0;
1359 
1360 				int priv = 0;
1361 				if (p == 0 || suser(p->p_ucred, &p->p_acflag))
1362 					priv = 0;
1363 				else
1364 					priv = 1;
1365 				if (m) {
1366 					req = mtod(m, caddr_t);
1367 					len = m->m_len;
1368 				}
1369 				error = ipsec6_set_policy(in6p,
1370 				                   optname, req, len, priv);
1371 			    }
1372 				break;
1373 #endif /* IPSEC */
1374 
1375 			default:
1376 				error = ENOPROTOOPT;
1377 				break;
1378 			}
1379 			if (m)
1380 				(void)m_free(m);
1381 			break;
1382 
1383 		case PRCO_GETOPT:
1384 			switch (optname) {
1385 
1386 			case IPV6_OPTIONS:
1387 			case IPV6_RETOPTS:
1388 #if 0
1389 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1390 				if (in6p->in6p_options) {
1391 					m->m_len = in6p->in6p_options->m_len;
1392 					bcopy(mtod(in6p->in6p_options, caddr_t),
1393 					      mtod(m, caddr_t),
1394 					      (unsigned)m->m_len);
1395 				} else
1396 					m->m_len = 0;
1397 				break;
1398 #else
1399 				error = ENOPROTOOPT;
1400 				break;
1401 #endif
1402 
1403 			case IPV6_PKTOPTIONS:
1404 				if (in6p->in6p_options) {
1405 					*mp = m_copym(in6p->in6p_options, 0,
1406 						      M_COPYALL, M_WAIT);
1407 				} else {
1408 					*mp = m_get(M_WAIT, MT_SOOPTS);
1409 					(*mp)->m_len = 0;
1410 				}
1411 				break;
1412 
1413 			case IPV6_HOPOPTS:
1414 			case IPV6_DSTOPTS:
1415 				if (p == 0 || suser(p->p_ucred, &p->p_acflag)) {
1416 					error = EPERM;
1417 					break;
1418 				}
1419 				/* fall through */
1420 			case IPV6_UNICAST_HOPS:
1421 			case IPV6_RECVOPTS:
1422 			case IPV6_RECVRETOPTS:
1423 			case IPV6_RECVDSTADDR:
1424 			case IPV6_PORTRANGE:
1425 			case IPV6_PKTINFO:
1426 			case IPV6_HOPLIMIT:
1427 			case IPV6_RTHDR:
1428 			case IPV6_CHECKSUM:
1429 			case IPV6_FAITH:
1430 #ifndef INET6_BINDV6ONLY
1431 			case IPV6_BINDV6ONLY:
1432 #endif
1433 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1434 				m->m_len = sizeof(int);
1435 				switch (optname) {
1436 
1437 				case IPV6_UNICAST_HOPS:
1438 					optval = in6p->in6p_hops;
1439 					break;
1440 
1441 #define OPTBIT(bit) (in6p->in6p_flags & bit ? 1 : 0)
1442 
1443 				case IPV6_RECVOPTS:
1444 					optval = OPTBIT(IN6P_RECVOPTS);
1445 					break;
1446 
1447 				case IPV6_RECVRETOPTS:
1448 					optval = OPTBIT(IN6P_RECVRETOPTS);
1449 					break;
1450 
1451 				case IPV6_RECVDSTADDR:
1452 					optval = OPTBIT(IN6P_RECVDSTADDR);
1453 					break;
1454 
1455 				case IPV6_PORTRANGE:
1456 				    {
1457 					int flags;
1458 					flags = in6p->in6p_flags;
1459 					if (flags & IN6P_HIGHPORT)
1460 						optval = IPV6_PORTRANGE_HIGH;
1461 					else if (flags & IN6P_LOWPORT)
1462 						optval = IPV6_PORTRANGE_LOW;
1463 					else
1464 						optval = 0;
1465 					break;
1466 				    }
1467 
1468 				case IPV6_PKTINFO:
1469 					optval = OPTBIT(IN6P_PKTINFO);
1470 					break;
1471 
1472 				case IPV6_HOPLIMIT:
1473 					optval = OPTBIT(IN6P_HOPLIMIT);
1474 					break;
1475 
1476 				case IPV6_HOPOPTS:
1477 					optval = OPTBIT(IN6P_HOPOPTS);
1478 					break;
1479 
1480 				case IPV6_DSTOPTS:
1481 					optval = OPTBIT(IN6P_DSTOPTS);
1482 					break;
1483 
1484 				case IPV6_RTHDR:
1485 					optval = OPTBIT(IN6P_RTHDR);
1486 					break;
1487 
1488 				case IPV6_CHECKSUM:
1489 					optval = in6p->in6p_cksum;
1490 					break;
1491 
1492 				case IPV6_FAITH:
1493 					optval = OPTBIT(IN6P_FAITH);
1494 					break;
1495 
1496 #ifndef INET6_BINDV6ONLY
1497 				case IPV6_BINDV6ONLY:
1498 					optval = OPTBIT(IN6P_BINDV6ONLY);
1499 					break;
1500 #endif
1501 				}
1502 				*mtod(m, int *) = optval;
1503 				break;
1504 
1505 			case IPV6_MULTICAST_IF:
1506 			case IPV6_MULTICAST_HOPS:
1507 			case IPV6_MULTICAST_LOOP:
1508 			case IPV6_JOIN_GROUP:
1509 			case IPV6_LEAVE_GROUP:
1510 				error = ip6_getmoptions(optname, in6p->in6p_moptions, mp);
1511 				break;
1512 
1513 #ifdef IPSEC
1514 			case IPV6_IPSEC_POLICY:
1515 			{
1516 				caddr_t req = NULL;
1517 				size_t len = 0;
1518 
1519 				if (m) {
1520 					req = mtod(m, caddr_t);
1521 					len = m->m_len;
1522 				}
1523 				error = ipsec6_get_policy(in6p, req, len, mp);
1524 				break;
1525 			}
1526 #endif /* IPSEC */
1527 
1528 			default:
1529 				error = ENOPROTOOPT;
1530 				break;
1531 			}
1532 			break;
1533 		}
1534 	} else {
1535 		error = EINVAL;
1536 		if (op == PRCO_SETOPT && *mp)
1537 			(void)m_free(*mp);
1538 	}
1539 	return(error);
1540 }
1541 
1542 /*
1543  * Set up IP6 options in pcb for insertion in output packets.
1544  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1545  * with destination address if source routed.
1546  */
1547 static int
1548 ip6_pcbopts(pktopt, m, so)
1549 	struct ip6_pktopts **pktopt;
1550 	register struct mbuf *m;
1551 	struct socket *so;
1552 {
1553 	register struct ip6_pktopts *opt = *pktopt;
1554 	int error = 0;
1555 	struct proc *p = curproc;	/* XXX */
1556 	int priv = 0;
1557 
1558 	/* turn off any old options. */
1559 	if (opt) {
1560 		if (opt->ip6po_m)
1561 			(void)m_free(opt->ip6po_m);
1562 	} else
1563 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
1564 	*pktopt = 0;
1565 
1566 	if (!m || m->m_len == 0) {
1567 		/*
1568 		 * Only turning off any previous options.
1569 		 */
1570 		if (opt)
1571 			free(opt, M_IP6OPT);
1572 		if (m)
1573 			(void)m_free(m);
1574 		return(0);
1575 	}
1576 
1577 	/*  set options specified by user. */
1578 	if (p && !suser(p->p_ucred, &p->p_acflag))
1579 		priv = 1;
1580 	if ((error = ip6_setpktoptions(m, opt, priv)) != 0) {
1581 		(void)m_free(m);
1582 		return(error);
1583 	}
1584 	*pktopt = opt;
1585 	return(0);
1586 }
1587 
1588 /*
1589  * Set the IP6 multicast options in response to user setsockopt().
1590  */
1591 static int
1592 ip6_setmoptions(optname, im6op, m)
1593 	int optname;
1594 	struct ip6_moptions **im6op;
1595 	struct mbuf *m;
1596 {
1597 	int error = 0;
1598 	u_int loop, ifindex;
1599 	struct ipv6_mreq *mreq;
1600 	struct ifnet *ifp;
1601 	struct ip6_moptions *im6o = *im6op;
1602 	struct route_in6 ro;
1603 	struct sockaddr_in6 *dst;
1604 	struct in6_multi_mship *imm;
1605 	struct proc *p = curproc;	/* XXX */
1606 
1607 	if (im6o == NULL) {
1608 		/*
1609 		 * No multicast option buffer attached to the pcb;
1610 		 * allocate one and initialize to default values.
1611 		 */
1612 		im6o = (struct ip6_moptions *)
1613 			malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
1614 
1615 		if (im6o == NULL)
1616 			return(ENOBUFS);
1617 		*im6op = im6o;
1618 		im6o->im6o_multicast_ifp = NULL;
1619 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1620 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
1621 		LIST_INIT(&im6o->im6o_memberships);
1622 	}
1623 
1624 	switch (optname) {
1625 
1626 	case IPV6_MULTICAST_IF:
1627 		/*
1628 		 * Select the interface for outgoing multicast packets.
1629 		 */
1630 		if (m == NULL || m->m_len != sizeof(u_int)) {
1631 			error = EINVAL;
1632 			break;
1633 		}
1634 		bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
1635 		if (ifindex < 0 || if_index < ifindex) {
1636 			error = ENXIO;	/* XXX EINVAL? */
1637 			break;
1638 		}
1639 		ifp = ifindex2ifnet[ifindex];
1640 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1641 			error = EADDRNOTAVAIL;
1642 			break;
1643 		}
1644 		im6o->im6o_multicast_ifp = ifp;
1645 		break;
1646 
1647 	case IPV6_MULTICAST_HOPS:
1648 	    {
1649 		/*
1650 		 * Set the IP6 hoplimit for outgoing multicast packets.
1651 		 */
1652 		int optval;
1653 		if (m == NULL || m->m_len != sizeof(int)) {
1654 			error = EINVAL;
1655 			break;
1656 		}
1657 		bcopy(mtod(m, u_int *), &optval, sizeof(optval));
1658 		if (optval < -1 || optval >= 256)
1659 			error = EINVAL;
1660 		else if (optval == -1)
1661 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1662 		else
1663 			im6o->im6o_multicast_hlim = optval;
1664 		break;
1665 	    }
1666 
1667 	case IPV6_MULTICAST_LOOP:
1668 		/*
1669 		 * Set the loopback flag for outgoing multicast packets.
1670 		 * Must be zero or one.
1671 		 */
1672 		if (m == NULL || m->m_len != sizeof(u_int)) {
1673 			error = EINVAL;
1674 			break;
1675 		}
1676 		bcopy(mtod(m, u_int *), &loop, sizeof(loop));
1677 		if (loop > 1) {
1678 			error = EINVAL;
1679 			break;
1680 		}
1681 		im6o->im6o_multicast_loop = loop;
1682 		break;
1683 
1684 	case IPV6_JOIN_GROUP:
1685 		/*
1686 		 * Add a multicast group membership.
1687 		 * Group must be a valid IP6 multicast address.
1688 		 */
1689 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
1690 			error = EINVAL;
1691 			break;
1692 		}
1693 		mreq = mtod(m, struct ipv6_mreq *);
1694 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
1695 			/*
1696 			 * We use the unspecified address to specify to accept
1697 			 * all multicast addresses. Only super user is allowed
1698 			 * to do this.
1699 			 */
1700 			if (suser(p->p_ucred, &p->p_acflag)) {
1701 				error = EACCES;
1702 				break;
1703 			}
1704 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
1705 			error = EINVAL;
1706 			break;
1707 		}
1708 
1709 		/*
1710 		 * If the interface is specified, validate it.
1711 		 */
1712 		if (mreq->ipv6mr_interface < 0
1713 		 || if_index < mreq->ipv6mr_interface) {
1714 			error = ENXIO;	/* XXX EINVAL? */
1715 			break;
1716 		}
1717 		/*
1718 		 * If no interface was explicitly specified, choose an
1719 		 * appropriate one according to the given multicast address.
1720 		 */
1721 		if (mreq->ipv6mr_interface == 0) {
1722 			/*
1723 			 * If the multicast address is in node-local scope,
1724 			 * the interface should be a loopback interface.
1725 			 * Otherwise, look up the routing table for the
1726 			 * address, and choose the outgoing interface.
1727 			 *   XXX: is it a good approach?
1728 			 */
1729 			if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
1730 				ifp = &loif[0];
1731 			} else {
1732 				ro.ro_rt = NULL;
1733 				dst = (struct sockaddr_in6 *)&ro.ro_dst;
1734 				bzero(dst, sizeof(*dst));
1735 				dst->sin6_len = sizeof(struct sockaddr_in6);
1736 				dst->sin6_family = AF_INET6;
1737 				dst->sin6_addr = mreq->ipv6mr_multiaddr;
1738 				rtalloc((struct route *)&ro);
1739 				if (ro.ro_rt == NULL) {
1740 					error = EADDRNOTAVAIL;
1741 					break;
1742 				}
1743 				ifp = ro.ro_rt->rt_ifp;
1744 				rtfree(ro.ro_rt);
1745 			}
1746 		} else
1747 			ifp = ifindex2ifnet[mreq->ipv6mr_interface];
1748 
1749 		/*
1750 		 * See if we found an interface, and confirm that it
1751 		 * supports multicast
1752 		 */
1753 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1754 			error = EADDRNOTAVAIL;
1755 			break;
1756 		}
1757 		/*
1758 		 * Put interface index into the multicast address,
1759 		 * if the address has link-local scope.
1760 		 */
1761 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
1762 			mreq->ipv6mr_multiaddr.s6_addr16[1]
1763 				= htons(mreq->ipv6mr_interface);
1764 		}
1765 		/*
1766 		 * See if the membership already exists.
1767 		 */
1768 		for (imm = im6o->im6o_memberships.lh_first;
1769 		     imm != NULL; imm = imm->i6mm_chain.le_next)
1770 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
1771 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
1772 					       &mreq->ipv6mr_multiaddr))
1773 				break;
1774 		if (imm != NULL) {
1775 			error = EADDRINUSE;
1776 			break;
1777 		}
1778 		/*
1779 		 * Everything looks good; add a new record to the multicast
1780 		 * address list for the given interface.
1781 		 */
1782 		imm = malloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
1783 		if (imm == NULL) {
1784 			error = ENOBUFS;
1785 			break;
1786 		}
1787 		if ((imm->i6mm_maddr =
1788 		     in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
1789 			free(imm, M_IPMADDR);
1790 			break;
1791 		}
1792 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
1793 		break;
1794 
1795 	case IPV6_LEAVE_GROUP:
1796 		/*
1797 		 * Drop a multicast group membership.
1798 		 * Group must be a valid IP6 multicast address.
1799 		 */
1800 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
1801 			error = EINVAL;
1802 			break;
1803 		}
1804 		mreq = mtod(m, struct ipv6_mreq *);
1805 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
1806 			if (suser(p->p_ucred, &p->p_acflag)) {
1807 				error = EACCES;
1808 				break;
1809 			}
1810 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
1811 			error = EINVAL;
1812 			break;
1813 		}
1814 		/*
1815 		 * If an interface address was specified, get a pointer
1816 		 * to its ifnet structure.
1817 		 */
1818 		if (mreq->ipv6mr_interface < 0
1819 		 || if_index < mreq->ipv6mr_interface) {
1820 			error = ENXIO;	/* XXX EINVAL? */
1821 			break;
1822 		}
1823 		ifp = ifindex2ifnet[mreq->ipv6mr_interface];
1824 		/*
1825 		 * Put interface index into the multicast address,
1826 		 * if the address has link-local scope.
1827 		 */
1828 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
1829 			mreq->ipv6mr_multiaddr.s6_addr16[1]
1830 				= htons(mreq->ipv6mr_interface);
1831 		}
1832 		/*
1833 		 * Find the membership in the membership list.
1834 		 */
1835 		for (imm = im6o->im6o_memberships.lh_first;
1836 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
1837 			if ((ifp == NULL ||
1838 			     imm->i6mm_maddr->in6m_ifp == ifp) &&
1839 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
1840 					       &mreq->ipv6mr_multiaddr))
1841 				break;
1842 		}
1843 		if (imm == NULL) {
1844 			/* Unable to resolve interface */
1845 			error = EADDRNOTAVAIL;
1846 			break;
1847 		}
1848 		/*
1849 		 * Give up the multicast address record to which the
1850 		 * membership points.
1851 		 */
1852 		LIST_REMOVE(imm, i6mm_chain);
1853 		in6_delmulti(imm->i6mm_maddr);
1854 		free(imm, M_IPMADDR);
1855 		break;
1856 
1857 	default:
1858 		error = EOPNOTSUPP;
1859 		break;
1860 	}
1861 
1862 	/*
1863 	 * If all options have default values, no need to keep the mbuf.
1864 	 */
1865 	if (im6o->im6o_multicast_ifp == NULL &&
1866 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
1867 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
1868 	    im6o->im6o_memberships.lh_first == NULL) {
1869 		free(*im6op, M_IPMOPTS);
1870 		*im6op = NULL;
1871 	}
1872 
1873 	return(error);
1874 }
1875 
1876 /*
1877  * Return the IP6 multicast options in response to user getsockopt().
1878  */
1879 static int
1880 ip6_getmoptions(optname, im6o, mp)
1881 	int optname;
1882 	register struct ip6_moptions *im6o;
1883 	register struct mbuf **mp;
1884 {
1885 	u_int *hlim, *loop, *ifindex;
1886 
1887 	*mp = m_get(M_WAIT, MT_SOOPTS);
1888 
1889 	switch (optname) {
1890 
1891 	case IPV6_MULTICAST_IF:
1892 		ifindex = mtod(*mp, u_int *);
1893 		(*mp)->m_len = sizeof(u_int);
1894 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
1895 			*ifindex = 0;
1896 		else
1897 			*ifindex = im6o->im6o_multicast_ifp->if_index;
1898 		return(0);
1899 
1900 	case IPV6_MULTICAST_HOPS:
1901 		hlim = mtod(*mp, u_int *);
1902 		(*mp)->m_len = sizeof(u_int);
1903 		if (im6o == NULL)
1904 			*hlim = ip6_defmcasthlim;
1905 		else
1906 			*hlim = im6o->im6o_multicast_hlim;
1907 		return(0);
1908 
1909 	case IPV6_MULTICAST_LOOP:
1910 		loop = mtod(*mp, u_int *);
1911 		(*mp)->m_len = sizeof(u_int);
1912 		if (im6o == NULL)
1913 			*loop = ip6_defmcasthlim;
1914 		else
1915 			*loop = im6o->im6o_multicast_loop;
1916 		return(0);
1917 
1918 	default:
1919 		return(EOPNOTSUPP);
1920 	}
1921 }
1922 
1923 /*
1924  * Discard the IP6 multicast options.
1925  */
1926 void
1927 ip6_freemoptions(im6o)
1928 	register struct ip6_moptions *im6o;
1929 {
1930 	struct in6_multi_mship *imm;
1931 
1932 	if (im6o == NULL)
1933 		return;
1934 
1935 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
1936 		LIST_REMOVE(imm, i6mm_chain);
1937 		if (imm->i6mm_maddr)
1938 			in6_delmulti(imm->i6mm_maddr);
1939 		free(imm, M_IPMADDR);
1940 	}
1941 	free(im6o, M_IPMOPTS);
1942 }
1943 
1944 /*
1945  * Set IPv6 outgoing packet options based on advanced API.
1946  */
1947 int
1948 ip6_setpktoptions(control, opt, priv)
1949 	struct mbuf *control;
1950 	struct ip6_pktopts *opt;
1951 	int priv;
1952 {
1953 	register struct cmsghdr *cm = 0;
1954 
1955 	if (control == 0 || opt == 0)
1956 		return(EINVAL);
1957 
1958 	bzero(opt, sizeof(*opt));
1959 	opt->ip6po_hlim = -1; /* -1 means to use default hop limit */
1960 
1961 	/*
1962 	 * XXX: Currently, we assume all the optional information is stored
1963 	 * in a single mbuf.
1964 	 */
1965 	if (control->m_next)
1966 		return(EINVAL);
1967 
1968 	opt->ip6po_m = control;
1969 
1970 	for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
1971 		     control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
1972 		cm = mtod(control, struct cmsghdr *);
1973 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
1974 			return(EINVAL);
1975 		if (cm->cmsg_level != IPPROTO_IPV6)
1976 			continue;
1977 
1978 		switch(cm->cmsg_type) {
1979 		case IPV6_PKTINFO:
1980 			if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo)))
1981 				return(EINVAL);
1982 			opt->ip6po_pktinfo = (struct in6_pktinfo *)CMSG_DATA(cm);
1983 			if (opt->ip6po_pktinfo->ipi6_ifindex &&
1984 			    IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr))
1985 				opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] =
1986 					htons(opt->ip6po_pktinfo->ipi6_ifindex);
1987 
1988 			if (opt->ip6po_pktinfo->ipi6_ifindex > if_index
1989 			 || opt->ip6po_pktinfo->ipi6_ifindex < 0) {
1990 				return(ENXIO);
1991 			}
1992 
1993 			/*
1994 			 * Check if the requested source address is indeed a
1995 			 * unicast address assigned to the node.
1996 			 */
1997 			if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
1998 				struct ifaddr *ia;
1999 				struct sockaddr_in6 sin6;
2000 
2001 				bzero(&sin6, sizeof(sin6));
2002 				sin6.sin6_len = sizeof(sin6);
2003 				sin6.sin6_family = AF_INET6;
2004 				sin6.sin6_addr =
2005 					opt->ip6po_pktinfo->ipi6_addr;
2006 				ia = ifa_ifwithaddr(sin6tosa(&sin6));
2007 				if (ia == NULL ||
2008 				    (opt->ip6po_pktinfo->ipi6_ifindex &&
2009 				     (ia->ifa_ifp->if_index !=
2010 				      opt->ip6po_pktinfo->ipi6_ifindex))) {
2011 					return(EADDRNOTAVAIL);
2012 				}
2013 				/*
2014 				 * Check if the requested source address is
2015 				 * indeed a unicast address assigned to the
2016 				 * node.
2017 				 */
2018 				if (IN6_IS_ADDR_MULTICAST(&opt->ip6po_pktinfo->ipi6_addr))
2019 					return(EADDRNOTAVAIL);
2020 			}
2021 			break;
2022 
2023 		case IPV6_HOPLIMIT:
2024 			if (cm->cmsg_len != CMSG_LEN(sizeof(int)))
2025 				return(EINVAL);
2026 
2027 			bcopy(CMSG_DATA(cm), &opt->ip6po_hlim,
2028 			    sizeof(opt->ip6po_hlim));
2029 			if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255)
2030 				return(EINVAL);
2031 			break;
2032 
2033 		case IPV6_NEXTHOP:
2034 			if (!priv)
2035 				return(EPERM);
2036 
2037 			if (cm->cmsg_len < sizeof(u_char) ||
2038 			    /* check if cmsg_len is large enough for sa_len */
2039 			    cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm)))
2040 				return(EINVAL);
2041 
2042 			opt->ip6po_nexthop = (struct sockaddr *)CMSG_DATA(cm);
2043 
2044 			break;
2045 
2046 		case IPV6_HOPOPTS:
2047 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh)))
2048 				return(EINVAL);
2049 			opt->ip6po_hbh = (struct ip6_hbh *)CMSG_DATA(cm);
2050 			if (cm->cmsg_len !=
2051 			    CMSG_LEN((opt->ip6po_hbh->ip6h_len + 1) << 3))
2052 				return(EINVAL);
2053 			break;
2054 
2055 		case IPV6_DSTOPTS:
2056 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest)))
2057 				return(EINVAL);
2058 
2059 			/*
2060 			 * If there is no routing header yet, the destination
2061 			 * options header should be put on the 1st part.
2062 			 * Otherwise, the header should be on the 2nd part.
2063 			 * (See RFC 2460, section 4.1)
2064 			 */
2065 			if (opt->ip6po_rthdr == NULL) {
2066 				opt->ip6po_dest1 =
2067 					(struct ip6_dest *)CMSG_DATA(cm);
2068 				if (cm->cmsg_len !=
2069 				    CMSG_LEN((opt->ip6po_dest1->ip6d_len + 1)
2070 					     << 3))
2071 					return(EINVAL);
2072 			}
2073 			else {
2074 				opt->ip6po_dest2 =
2075 					(struct ip6_dest *)CMSG_DATA(cm);
2076 				if (cm->cmsg_len !=
2077 				    CMSG_LEN((opt->ip6po_dest2->ip6d_len + 1)
2078 					     << 3))
2079 					return(EINVAL);
2080 			}
2081 			break;
2082 
2083 		case IPV6_RTHDR:
2084 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr)))
2085 				return(EINVAL);
2086 			opt->ip6po_rthdr = (struct ip6_rthdr *)CMSG_DATA(cm);
2087 			if (cm->cmsg_len !=
2088 			    CMSG_LEN((opt->ip6po_rthdr->ip6r_len + 1) << 3))
2089 				return(EINVAL);
2090 			switch(opt->ip6po_rthdr->ip6r_type) {
2091 			case IPV6_RTHDR_TYPE_0:
2092 				if (opt->ip6po_rthdr->ip6r_segleft == 0)
2093 					return(EINVAL);
2094 				break;
2095 			default:
2096 				return(EINVAL);
2097 			}
2098 			break;
2099 
2100 		default:
2101 			return(ENOPROTOOPT);
2102 		}
2103 	}
2104 
2105 	return(0);
2106 }
2107 
2108 /*
2109  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2110  * packet to the input queue of a specified interface.  Note that this
2111  * calls the output routine of the loopback "driver", but with an interface
2112  * pointer that might NOT be &loif -- easier than replicating that code here.
2113  */
2114 void
2115 ip6_mloopback(ifp, m, dst)
2116 	struct ifnet *ifp;
2117 	register struct mbuf *m;
2118 	register struct sockaddr_in6 *dst;
2119 {
2120 	struct mbuf *copym;
2121 	struct ip6_hdr *ip6;
2122 
2123 	copym = m_copy(m, 0, M_COPYALL);
2124 	if (copym == NULL)
2125 		return;
2126 
2127 	/*
2128 	 * Make sure to deep-copy IPv6 header portion in case the data
2129 	 * is in an mbuf cluster, so that we can safely override the IPv6
2130 	 * header portion later.
2131 	 */
2132 	if ((copym->m_flags & M_EXT) != 0 ||
2133 	    copym->m_len < sizeof(struct ip6_hdr)) {
2134 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
2135 		if (copym == NULL)
2136 			return;
2137 	}
2138 
2139 #ifdef DIAGNOSTIC
2140 	if (copym->m_len < sizeof(*ip6)) {
2141 		m_freem(copym);
2142 		return;
2143 	}
2144 #endif
2145 
2146 #ifndef FAKE_LOOPBACK_IF
2147 	if ((ifp->if_flags & IFF_LOOPBACK) == 0)
2148 #else
2149 	if (1)
2150 #endif
2151 	{
2152 		ip6 = mtod(copym, struct ip6_hdr *);
2153 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
2154 			ip6->ip6_src.s6_addr16[1] = 0;
2155 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
2156 			ip6->ip6_dst.s6_addr16[1] = 0;
2157 	}
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