xref: /dragonfly/sys/netinet6/ip6_output.c (revision 113f6df6)
1 /*	$FreeBSD: src/sys/netinet6/ip6_output.c,v 1.13.2.18 2003/01/24 05:11:35 sam Exp $	*/
2 /*	$DragonFly: src/sys/netinet6/ip6_output.c,v 1.14 2004/10/15 22:59:10 hsu Exp $	*/
3 /*	$KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $	*/
4 
5 /*
6  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
7  * All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. Neither the name of the project nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 /*
35  * Copyright (c) 1982, 1986, 1988, 1990, 1993
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. All advertising materials mentioning features or use of this software
47  *    must display the following acknowledgement:
48  *	This product includes software developed by the University of
49  *	California, Berkeley and its contributors.
50  * 4. Neither the name of the University nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64  * SUCH DAMAGE.
65  *
66  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
67  */
68 
69 #include "opt_ip6fw.h"
70 #include "opt_inet.h"
71 #include "opt_inet6.h"
72 #include "opt_ipsec.h"
73 
74 #include <sys/param.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/errno.h>
78 #include <sys/protosw.h>
79 #include <sys/socket.h>
80 #include <sys/socketvar.h>
81 #include <sys/systm.h>
82 #include <sys/kernel.h>
83 #include <sys/proc.h>
84 
85 #include <net/if.h>
86 #include <net/route.h>
87 #include <net/pfil.h>
88 
89 #include <netinet/in.h>
90 #include <netinet/in_var.h>
91 #include <netinet6/in6_var.h>
92 #include <netinet/ip6.h>
93 #include <netinet/icmp6.h>
94 #include <netinet6/ip6_var.h>
95 #include <netinet/in_pcb.h>
96 #include <netinet6/nd6.h>
97 
98 #ifdef IPSEC
99 #include <netinet6/ipsec.h>
100 #ifdef INET6
101 #include <netinet6/ipsec6.h>
102 #endif
103 #include <netproto/key/key.h>
104 #endif /* IPSEC */
105 
106 #ifdef FAST_IPSEC
107 #include <netproto/ipsec/ipsec.h>
108 #include <netproto/ipsec/ipsec6.h>
109 #include <netproto/ipsec/key.h>
110 #endif /* FAST_IPSEC */
111 
112 #include <net/ip6fw/ip6_fw.h>
113 
114 #include <net/net_osdep.h>
115 
116 static MALLOC_DEFINE(M_IPMOPTS, "ip6_moptions", "internet multicast options");
117 
118 struct ip6_exthdrs {
119 	struct mbuf *ip6e_ip6;
120 	struct mbuf *ip6e_hbh;
121 	struct mbuf *ip6e_dest1;
122 	struct mbuf *ip6e_rthdr;
123 	struct mbuf *ip6e_dest2;
124 };
125 
126 static int ip6_pcbopts (struct ip6_pktopts **, struct mbuf *,
127 			    struct socket *, struct sockopt *sopt);
128 static int ip6_setmoptions (int, struct ip6_moptions **, struct mbuf *);
129 static int ip6_getmoptions (int, struct ip6_moptions *, struct mbuf **);
130 static int ip6_copyexthdr (struct mbuf **, caddr_t, int);
131 static int ip6_insertfraghdr (struct mbuf *, struct mbuf *, int,
132 				  struct ip6_frag **);
133 static int ip6_insert_jumboopt (struct ip6_exthdrs *, u_int32_t);
134 static int ip6_splithdr (struct mbuf *, struct ip6_exthdrs *);
135 
136 /*
137  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
138  * header (with pri, len, nxt, hlim, src, dst).
139  * This function may modify ver and hlim only.
140  * The mbuf chain containing the packet will be freed.
141  * The mbuf opt, if present, will not be freed.
142  *
143  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
144  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
145  * which is rt_rmx.rmx_mtu.
146  */
147 int
148 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt, struct route_in6 *ro,
149 	   int flags, struct ip6_moptions *im6o,
150 	   struct ifnet **ifpp,		/* XXX: just for statistics */
151 	   struct inpcb *inp)
152 {
153 	struct ip6_hdr *ip6, *mhip6;
154 	struct ifnet *ifp, *origifp;
155 	struct mbuf *m = m0;
156 	int hlen, tlen, len, off;
157 	struct route_in6 ip6route;
158 	struct sockaddr_in6 *dst;
159 	int error = 0;
160 	struct in6_ifaddr *ia = NULL;
161 	u_long mtu;
162 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
163 	struct ip6_exthdrs exthdrs;
164 	struct in6_addr finaldst;
165 	struct route_in6 *ro_pmtu = NULL;
166 	int hdrsplit = 0;
167 	int needipsec = 0;
168 #ifdef IPSEC
169 	int needipsectun = 0;
170 	struct secpolicy *sp = NULL;
171 	struct socket *so = inp ? inp->inp_socket : NULL;
172 
173 	ip6 = mtod(m, struct ip6_hdr *);
174 #endif /* IPSEC */
175 #ifdef FAST_IPSEC
176 	int needipsectun = 0;
177 	struct secpolicy *sp = NULL;
178 
179 	ip6 = mtod(m, struct ip6_hdr *);
180 #endif /* FAST_IPSEC */
181 
182 #define MAKE_EXTHDR(hp, mp)						\
183     do {								\
184 	if (hp) {							\
185 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
186 		error = ip6_copyexthdr((mp), (caddr_t)(hp), 		\
187 				       ((eh)->ip6e_len + 1) << 3);	\
188 		if (error)						\
189 			goto freehdrs;					\
190 	}								\
191     } while (0)
192 
193 	bzero(&exthdrs, sizeof(exthdrs));
194 
195 	if (opt) {
196 		/* Hop-by-Hop options header */
197 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
198 		/* Destination options header(1st part) */
199 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
200 		/* Routing header */
201 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
202 		/* Destination options header(2nd part) */
203 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
204 	}
205 
206 #ifdef IPSEC
207 	/* get a security policy for this packet */
208 	if (so == NULL)
209 		sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
210 	else
211 		sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
212 
213 	if (sp == NULL) {
214 		ipsec6stat.out_inval++;
215 		goto freehdrs;
216 	}
217 
218 	error = 0;
219 
220 	/* check policy */
221 	switch (sp->policy) {
222 	case IPSEC_POLICY_DISCARD:
223 		/*
224 		 * This packet is just discarded.
225 		 */
226 		ipsec6stat.out_polvio++;
227 		goto freehdrs;
228 
229 	case IPSEC_POLICY_BYPASS:
230 	case IPSEC_POLICY_NONE:
231 		/* no need to do IPsec. */
232 		needipsec = 0;
233 		break;
234 
235 	case IPSEC_POLICY_IPSEC:
236 		if (sp->req == NULL) {
237 			/* acquire a policy */
238 			error = key_spdacquire(sp);
239 			goto freehdrs;
240 		}
241 		needipsec = 1;
242 		break;
243 
244 	case IPSEC_POLICY_ENTRUST:
245 	default:
246 		printf("ip6_output: Invalid policy found. %d\n", sp->policy);
247 	}
248 #endif /* IPSEC */
249 #ifdef FAST_IPSEC
250 	/* get a security policy for this packet */
251 	if (inp == NULL)
252 		sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
253 	else
254 		sp = ipsec_getpolicybysock(m, IPSEC_DIR_OUTBOUND, inp, &error);
255 
256 	if (sp == NULL) {
257 		newipsecstat.ips_out_inval++;
258 		goto freehdrs;
259 	}
260 
261 	error = 0;
262 
263 	/* check policy */
264 	switch (sp->policy) {
265 	case IPSEC_POLICY_DISCARD:
266 		/*
267 		 * This packet is just discarded.
268 		 */
269 		newipsecstat.ips_out_polvio++;
270 		goto freehdrs;
271 
272 	case IPSEC_POLICY_BYPASS:
273 	case IPSEC_POLICY_NONE:
274 		/* no need to do IPsec. */
275 		needipsec = 0;
276 		break;
277 
278 	case IPSEC_POLICY_IPSEC:
279 		if (sp->req == NULL) {
280 			/* acquire a policy */
281 			error = key_spdacquire(sp);
282 			goto freehdrs;
283 		}
284 		needipsec = 1;
285 		break;
286 
287 	case IPSEC_POLICY_ENTRUST:
288 	default:
289 		printf("ip6_output: Invalid policy found. %d\n", sp->policy);
290 	}
291 #endif /* FAST_IPSEC */
292 
293 	/*
294 	 * Calculate the total length of the extension header chain.
295 	 * Keep the length of the unfragmentable part for fragmentation.
296 	 */
297 	optlen = 0;
298 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
299 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
300 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
301 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
302 	/* NOTE: we don't add AH/ESP length here. do that later. */
303 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
304 
305 	/*
306 	 * If we need IPsec, or there is at least one extension header,
307 	 * separate IP6 header from the payload.
308 	 */
309 	if ((needipsec || optlen) && !hdrsplit) {
310 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
311 			m = NULL;
312 			goto freehdrs;
313 		}
314 		m = exthdrs.ip6e_ip6;
315 		hdrsplit++;
316 	}
317 
318 	/* adjust pointer */
319 	ip6 = mtod(m, struct ip6_hdr *);
320 
321 	/* adjust mbuf packet header length */
322 	m->m_pkthdr.len += optlen;
323 	plen = m->m_pkthdr.len - sizeof(*ip6);
324 
325 	/* If this is a jumbo payload, insert a jumbo payload option. */
326 	if (plen > IPV6_MAXPACKET) {
327 		if (!hdrsplit) {
328 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
329 				m = NULL;
330 				goto freehdrs;
331 			}
332 			m = exthdrs.ip6e_ip6;
333 			hdrsplit++;
334 		}
335 		/* adjust pointer */
336 		ip6 = mtod(m, struct ip6_hdr *);
337 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
338 			goto freehdrs;
339 		ip6->ip6_plen = 0;
340 	} else
341 		ip6->ip6_plen = htons(plen);
342 
343 	/*
344 	 * Concatenate headers and fill in next header fields.
345 	 * Here we have, on "m"
346 	 *	IPv6 payload
347 	 * and we insert headers accordingly.  Finally, we should be getting:
348 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
349 	 *
350 	 * during the header composing process, "m" points to IPv6 header.
351 	 * "mprev" points to an extension header prior to esp.
352 	 */
353 	{
354 		u_char *nexthdrp = &ip6->ip6_nxt;
355 		struct mbuf *mprev = m;
356 
357 		/*
358 		 * we treat dest2 specially.  this makes IPsec processing
359 		 * much easier.  the goal here is to make mprev point the
360 		 * mbuf prior to dest2.
361 		 *
362 		 * result: IPv6 dest2 payload
363 		 * m and mprev will point to IPv6 header.
364 		 */
365 		if (exthdrs.ip6e_dest2) {
366 			if (!hdrsplit)
367 				panic("assumption failed: hdr not split");
368 			exthdrs.ip6e_dest2->m_next = m->m_next;
369 			m->m_next = exthdrs.ip6e_dest2;
370 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
371 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
372 		}
373 
374 #define MAKE_CHAIN(m, mp, p, i)\
375     do {\
376 	if (m) {\
377 		if (!hdrsplit) \
378 			panic("assumption failed: hdr not split"); \
379 		*mtod((m), u_char *) = *(p);\
380 		*(p) = (i);\
381 		p = mtod((m), u_char *);\
382 		(m)->m_next = (mp)->m_next;\
383 		(mp)->m_next = (m);\
384 		(mp) = (m);\
385 	}\
386     } while (0)
387 		/*
388 		 * result: IPv6 hbh dest1 rthdr dest2 payload
389 		 * m will point to IPv6 header.  mprev will point to the
390 		 * extension header prior to dest2 (rthdr in the above case).
391 		 */
392 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev,
393 			   nexthdrp, IPPROTO_HOPOPTS);
394 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev,
395 			   nexthdrp, IPPROTO_DSTOPTS);
396 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev,
397 			   nexthdrp, IPPROTO_ROUTING);
398 
399 #if defined(IPSEC) || defined(FAST_IPSEC)
400 		if (!needipsec)
401 			goto skip_ipsec2;
402 
403 		/*
404 		 * pointers after IPsec headers are not valid any more.
405 		 * other pointers need a great care too.
406 		 * (IPsec routines should not mangle mbufs prior to AH/ESP)
407 		 */
408 		exthdrs.ip6e_dest2 = NULL;
409 
410 	    {
411 		struct ip6_rthdr *rh = NULL;
412 		int segleft_org = 0;
413 		struct ipsec_output_state state;
414 
415 		if (exthdrs.ip6e_rthdr) {
416 			rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
417 			segleft_org = rh->ip6r_segleft;
418 			rh->ip6r_segleft = 0;
419 		}
420 
421 		bzero(&state, sizeof(state));
422 		state.m = m;
423 		error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
424 			&needipsectun);
425 		m = state.m;
426 		if (error) {
427 			/* mbuf is already reclaimed in ipsec6_output_trans. */
428 			m = NULL;
429 			switch (error) {
430 			case EHOSTUNREACH:
431 			case ENETUNREACH:
432 			case EMSGSIZE:
433 			case ENOBUFS:
434 			case ENOMEM:
435 				break;
436 			default:
437 				printf("ip6_output (ipsec): error code %d\n", error);
438 				/* fall through */
439 			case ENOENT:
440 				/* don't show these error codes to the user */
441 				error = 0;
442 				break;
443 			}
444 			goto bad;
445 		}
446 		if (exthdrs.ip6e_rthdr) {
447 			/* ah6_output doesn't modify mbuf chain */
448 			rh->ip6r_segleft = segleft_org;
449 		}
450 	    }
451 skip_ipsec2:;
452 #endif
453 	}
454 
455 	/*
456 	 * If there is a routing header, replace destination address field
457 	 * with the first hop of the routing header.
458 	 */
459 	if (exthdrs.ip6e_rthdr) {
460 		struct ip6_rthdr *rh =
461 			(struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
462 						  struct ip6_rthdr *));
463 		struct ip6_rthdr0 *rh0;
464 
465 		finaldst = ip6->ip6_dst;
466 		switch (rh->ip6r_type) {
467 		case IPV6_RTHDR_TYPE_0:
468 			 rh0 = (struct ip6_rthdr0 *)rh;
469 			 ip6->ip6_dst = rh0->ip6r0_addr[0];
470 			 bcopy((caddr_t)&rh0->ip6r0_addr[1],
471 			       (caddr_t)&rh0->ip6r0_addr[0],
472 			       sizeof(struct in6_addr)*(rh0->ip6r0_segleft - 1)
473 				 );
474 			 rh0->ip6r0_addr[rh0->ip6r0_segleft - 1] = finaldst;
475 			 break;
476 		default:	/* is it possible? */
477 			 error = EINVAL;
478 			 goto bad;
479 		}
480 	}
481 
482 	/* Source address validation */
483 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
484 	    (flags & IPV6_DADOUTPUT) == 0) {
485 		error = EOPNOTSUPP;
486 		ip6stat.ip6s_badscope++;
487 		goto bad;
488 	}
489 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
490 		error = EOPNOTSUPP;
491 		ip6stat.ip6s_badscope++;
492 		goto bad;
493 	}
494 
495 	ip6stat.ip6s_localout++;
496 
497 	/*
498 	 * Route packet.
499 	 */
500 	if (ro == 0) {
501 		ro = &ip6route;
502 		bzero((caddr_t)ro, sizeof(*ro));
503 	}
504 	ro_pmtu = ro;
505 	if (opt && opt->ip6po_rthdr)
506 		ro = &opt->ip6po_route;
507 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
508 	/*
509 	 * If there is a cached route,
510 	 * check that it is to the same destination
511 	 * and is still up. If not, free it and try again.
512 	 */
513 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
514 			 dst->sin6_family != AF_INET6 ||
515 			 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
516 		RTFREE(ro->ro_rt);
517 		ro->ro_rt = (struct rtentry *)0;
518 	}
519 	if (ro->ro_rt == 0) {
520 		bzero(dst, sizeof(*dst));
521 		dst->sin6_family = AF_INET6;
522 		dst->sin6_len = sizeof(struct sockaddr_in6);
523 		dst->sin6_addr = ip6->ip6_dst;
524 #ifdef SCOPEDROUTING
525 		/* XXX: sin6_scope_id should already be fixed at this point */
526 		if (IN6_IS_SCOPE_LINKLOCAL(&dst->sin6_addr))
527 			dst->sin6_scope_id = ntohs(dst->sin6_addr.s6_addr16[1]);
528 #endif
529 	}
530 #if defined(IPSEC) || defined(FAST_IPSEC)
531 	if (needipsec && needipsectun) {
532 		struct ipsec_output_state state;
533 
534 		/*
535 		 * All the extension headers will become inaccessible
536 		 * (since they can be encrypted).
537 		 * Don't panic, we need no more updates to extension headers
538 		 * on inner IPv6 packet (since they are now encapsulated).
539 		 *
540 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
541 		 */
542 		bzero(&exthdrs, sizeof(exthdrs));
543 		exthdrs.ip6e_ip6 = m;
544 
545 		bzero(&state, sizeof(state));
546 		state.m = m;
547 		state.ro = (struct route *)ro;
548 		state.dst = (struct sockaddr *)dst;
549 
550 		error = ipsec6_output_tunnel(&state, sp, flags);
551 
552 		m = state.m;
553 		ro = (struct route_in6 *)state.ro;
554 		dst = (struct sockaddr_in6 *)state.dst;
555 		if (error) {
556 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
557 			m0 = m = NULL;
558 			m = NULL;
559 			switch (error) {
560 			case EHOSTUNREACH:
561 			case ENETUNREACH:
562 			case EMSGSIZE:
563 			case ENOBUFS:
564 			case ENOMEM:
565 				break;
566 			default:
567 				printf("ip6_output (ipsec): error code %d\n", error);
568 				/* fall through */
569 			case ENOENT:
570 				/* don't show these error codes to the user */
571 				error = 0;
572 				break;
573 			}
574 			goto bad;
575 		}
576 
577 		exthdrs.ip6e_ip6 = m;
578 	}
579 #endif /* IPSEC */
580 
581 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
582 		/* Unicast */
583 
584 #define ifatoia6(ifa)	((struct in6_ifaddr *)(ifa))
585 #define sin6tosa(sin6)	((struct sockaddr *)(sin6))
586 		/* xxx
587 		 * interface selection comes here
588 		 * if an interface is specified from an upper layer,
589 		 * ifp must point it.
590 		 */
591 		if (ro->ro_rt == 0) {
592 			/*
593 			 * non-bsdi always clone routes, if parent is
594 			 * PRF_CLONING.
595 			 */
596 			rtalloc((struct route *)ro);
597 		}
598 		if (ro->ro_rt == 0) {
599 			ip6stat.ip6s_noroute++;
600 			error = EHOSTUNREACH;
601 			/* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
602 			goto bad;
603 		}
604 		ia = ifatoia6(ro->ro_rt->rt_ifa);
605 		ifp = ro->ro_rt->rt_ifp;
606 		ro->ro_rt->rt_use++;
607 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
608 			dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
609 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
610 
611 		in6_ifstat_inc(ifp, ifs6_out_request);
612 
613 		/*
614 		 * Check if the outgoing interface conflicts with
615 		 * the interface specified by ifi6_ifindex (if specified).
616 		 * Note that loopback interface is always okay.
617 		 * (this may happen when we are sending a packet to one of
618 		 *  our own addresses.)
619 		 */
620 		if (opt && opt->ip6po_pktinfo
621 		 && opt->ip6po_pktinfo->ipi6_ifindex) {
622 			if (!(ifp->if_flags & IFF_LOOPBACK)
623 			 && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
624 				ip6stat.ip6s_noroute++;
625 				in6_ifstat_inc(ifp, ifs6_out_discard);
626 				error = EHOSTUNREACH;
627 				goto bad;
628 			}
629 		}
630 
631 		if (opt && opt->ip6po_hlim != -1)
632 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
633 	} else {
634 		/* Multicast */
635 		struct	in6_multi *in6m;
636 
637 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
638 
639 		/*
640 		 * See if the caller provided any multicast options
641 		 */
642 		ifp = NULL;
643 		if (im6o != NULL) {
644 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
645 			if (im6o->im6o_multicast_ifp != NULL)
646 				ifp = im6o->im6o_multicast_ifp;
647 		} else
648 			ip6->ip6_hlim = ip6_defmcasthlim;
649 
650 		/*
651 		 * See if the caller provided the outgoing interface
652 		 * as an ancillary data.
653 		 * Boundary check for ifindex is assumed to be already done.
654 		 */
655 		if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
656 			ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex];
657 
658 		/*
659 		 * If the destination is a node-local scope multicast,
660 		 * the packet should be loop-backed only.
661 		 */
662 		if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
663 			/*
664 			 * If the outgoing interface is already specified,
665 			 * it should be a loopback interface.
666 			 */
667 			if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) {
668 				ip6stat.ip6s_badscope++;
669 				error = ENETUNREACH; /* XXX: better error? */
670 				/* XXX correct ifp? */
671 				in6_ifstat_inc(ifp, ifs6_out_discard);
672 				goto bad;
673 			} else {
674 				ifp = &loif[0];
675 			}
676 		}
677 
678 		if (opt && opt->ip6po_hlim != -1)
679 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
680 
681 		/*
682 		 * If caller did not provide an interface lookup a
683 		 * default in the routing table.  This is either a
684 		 * default for the speicfied group (i.e. a host
685 		 * route), or a multicast default (a route for the
686 		 * ``net'' ff00::/8).
687 		 */
688 		if (ifp == NULL) {
689 			if (ro->ro_rt == 0) {
690 				ro->ro_rt = rtalloc1((struct sockaddr *)
691 						&ro->ro_dst, 0, 0UL);
692 			}
693 			if (ro->ro_rt == 0) {
694 				ip6stat.ip6s_noroute++;
695 				error = EHOSTUNREACH;
696 				/* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
697 				goto bad;
698 			}
699 			ia = ifatoia6(ro->ro_rt->rt_ifa);
700 			ifp = ro->ro_rt->rt_ifp;
701 			ro->ro_rt->rt_use++;
702 		}
703 
704 		if ((flags & IPV6_FORWARDING) == 0)
705 			in6_ifstat_inc(ifp, ifs6_out_request);
706 		in6_ifstat_inc(ifp, ifs6_out_mcast);
707 
708 		/*
709 		 * Confirm that the outgoing interface supports multicast.
710 		 */
711 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
712 			ip6stat.ip6s_noroute++;
713 			in6_ifstat_inc(ifp, ifs6_out_discard);
714 			error = ENETUNREACH;
715 			goto bad;
716 		}
717 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
718 		if (in6m != NULL &&
719 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
720 			/*
721 			 * If we belong to the destination multicast group
722 			 * on the outgoing interface, and the caller did not
723 			 * forbid loopback, loop back a copy.
724 			 */
725 			ip6_mloopback(ifp, m, dst);
726 		} else {
727 			/*
728 			 * If we are acting as a multicast router, perform
729 			 * multicast forwarding as if the packet had just
730 			 * arrived on the interface to which we are about
731 			 * to send.  The multicast forwarding function
732 			 * recursively calls this function, using the
733 			 * IPV6_FORWARDING flag to prevent infinite recursion.
734 			 *
735 			 * Multicasts that are looped back by ip6_mloopback(),
736 			 * above, will be forwarded by the ip6_input() routine,
737 			 * if necessary.
738 			 */
739 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
740 				if (ip6_mforward(ip6, ifp, m) != 0) {
741 					m_freem(m);
742 					goto done;
743 				}
744 			}
745 		}
746 		/*
747 		 * Multicasts with a hoplimit of zero may be looped back,
748 		 * above, but must not be transmitted on a network.
749 		 * Also, multicasts addressed to the loopback interface
750 		 * are not sent -- the above call to ip6_mloopback() will
751 		 * loop back a copy if this host actually belongs to the
752 		 * destination group on the loopback interface.
753 		 */
754 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
755 			m_freem(m);
756 			goto done;
757 		}
758 	}
759 
760 	/*
761 	 * Fill the outgoing inteface to tell the upper layer
762 	 * to increment per-interface statistics.
763 	 */
764 	if (ifpp)
765 		*ifpp = ifp;
766 
767 	/*
768 	 * Determine path MTU.
769 	 */
770 	if (ro_pmtu != ro) {
771 		/* The first hop and the final destination may differ. */
772 		struct sockaddr_in6 *sin6_fin =
773 			(struct sockaddr_in6 *)&ro_pmtu->ro_dst;
774 		if (ro_pmtu->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
775 				       !IN6_ARE_ADDR_EQUAL(&sin6_fin->sin6_addr,
776 							   &finaldst))) {
777 			RTFREE(ro_pmtu->ro_rt);
778 			ro_pmtu->ro_rt = (struct rtentry *)0;
779 		}
780 		if (ro_pmtu->ro_rt == 0) {
781 			bzero(sin6_fin, sizeof(*sin6_fin));
782 			sin6_fin->sin6_family = AF_INET6;
783 			sin6_fin->sin6_len = sizeof(struct sockaddr_in6);
784 			sin6_fin->sin6_addr = finaldst;
785 
786 			rtalloc((struct route *)ro_pmtu);
787 		}
788 	}
789 	if (ro_pmtu->ro_rt != NULL) {
790 		u_int32_t ifmtu = nd_ifinfo[ifp->if_index].linkmtu;
791 
792 		mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
793 		if (mtu > ifmtu || mtu == 0) {
794 			/*
795 			 * The MTU on the route is larger than the MTU on
796 			 * the interface!  This shouldn't happen, unless the
797 			 * MTU of the interface has been changed after the
798 			 * interface was brought up.  Change the MTU in the
799 			 * route to match the interface MTU (as long as the
800 			 * field isn't locked).
801 			 *
802 			 * if MTU on the route is 0, we need to fix the MTU.
803 			 * this case happens with path MTU discovery timeouts.
804 			 */
805 			 mtu = ifmtu;
806 			 if ((ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
807 				 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; /* XXX */
808 		}
809 	} else {
810 		mtu = nd_ifinfo[ifp->if_index].linkmtu;
811 	}
812 
813 	/*
814 	 * advanced API (IPV6_USE_MIN_MTU) overrides mtu setting
815 	 */
816 	if ((flags & IPV6_MINMTU) != 0 && mtu > IPV6_MMTU)
817 		mtu = IPV6_MMTU;
818 
819 	/* Fake scoped addresses */
820 	if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
821 		/*
822 		 * If source or destination address is a scoped address, and
823 		 * the packet is going to be sent to a loopback interface,
824 		 * we should keep the original interface.
825 		 */
826 
827 		/*
828 		 * XXX: this is a very experimental and temporary solution.
829 		 * We eventually have sockaddr_in6 and use the sin6_scope_id
830 		 * field of the structure here.
831 		 * We rely on the consistency between two scope zone ids
832 		 * of source and destination, which should already be assured.
833 		 * Larger scopes than link will be supported in the future.
834 		 */
835 		origifp = NULL;
836 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
837 			origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
838 		else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
839 			origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
840 		/*
841 		 * XXX: origifp can be NULL even in those two cases above.
842 		 * For example, if we remove the (only) link-local address
843 		 * from the loopback interface, and try to send a link-local
844 		 * address without link-id information.  Then the source
845 		 * address is ::1, and the destination address is the
846 		 * link-local address with its s6_addr16[1] being zero.
847 		 * What is worse, if the packet goes to the loopback interface
848 		 * by a default rejected route, the null pointer would be
849 		 * passed to looutput, and the kernel would hang.
850 		 * The following last resort would prevent such disaster.
851 		 */
852 		if (origifp == NULL)
853 			origifp = ifp;
854 	}
855 	else
856 		origifp = ifp;
857 #ifndef SCOPEDROUTING
858 	/*
859 	 * clear embedded scope identifiers if necessary.
860 	 * in6_clearscope will touch the addresses only when necessary.
861 	 */
862 	in6_clearscope(&ip6->ip6_src);
863 	in6_clearscope(&ip6->ip6_dst);
864 #endif
865 
866 	/*
867 	 * Check with the firewall...
868 	 */
869 	if (ip6_fw_enable && ip6_fw_chk_ptr) {
870 		u_short port = 0;
871 		m->m_pkthdr.rcvif = NULL;	/* XXX */
872 		/* If ipfw says divert, we have to just drop packet */
873 		if ((*ip6_fw_chk_ptr)(&ip6, ifp, &port, &m)) {
874 			m_freem(m);
875 			goto done;
876 		}
877 		if (!m) {
878 			error = EACCES;
879 			goto done;
880 		}
881 	}
882 
883 	/*
884 	 * If the outgoing packet contains a hop-by-hop options header,
885 	 * it must be examined and processed even by the source node.
886 	 * (RFC 2460, section 4.)
887 	 */
888 	if (exthdrs.ip6e_hbh) {
889 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
890 		u_int32_t dummy1; /* XXX unused */
891 		u_int32_t dummy2; /* XXX unused */
892 
893 #ifdef DIAGNOSTIC
894 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
895 			panic("ip6e_hbh is not continuous");
896 #endif
897 		/*
898 		 *  XXX: if we have to send an ICMPv6 error to the sender,
899 		 *       we need the M_LOOP flag since icmp6_error() expects
900 		 *       the IPv6 and the hop-by-hop options header are
901 		 *       continuous unless the flag is set.
902 		 */
903 		m->m_flags |= M_LOOP;
904 		m->m_pkthdr.rcvif = ifp;
905 		if (ip6_process_hopopts(m,
906 					(u_int8_t *)(hbh + 1),
907 					((hbh->ip6h_len + 1) << 3) -
908 					sizeof(struct ip6_hbh),
909 					&dummy1, &dummy2) < 0) {
910 			/* m was already freed at this point */
911 			error = EINVAL;/* better error? */
912 			goto done;
913 		}
914 		m->m_flags &= ~M_LOOP; /* XXX */
915 		m->m_pkthdr.rcvif = NULL;
916 	}
917 
918 	/*
919 	 * Run through list of hooks for output packets.
920 	 */
921 	if (pfil_has_hooks(&inet6_pfil_hook)) {
922 		error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT);
923 		if (error != 0 || m == NULL)
924 			goto done;
925 		ip6 = mtod(m, struct ip6_hdr *);
926 	}
927 
928 	/*
929 	 * Send the packet to the outgoing interface.
930 	 * If necessary, do IPv6 fragmentation before sending.
931 	 */
932 	tlen = m->m_pkthdr.len;
933 	if (tlen <= mtu
934 #ifdef notyet
935 	    /*
936 	     * On any link that cannot convey a 1280-octet packet in one piece,
937 	     * link-specific fragmentation and reassembly must be provided at
938 	     * a layer below IPv6. [RFC 2460, sec.5]
939 	     * Thus if the interface has ability of link-level fragmentation,
940 	     * we can just send the packet even if the packet size is
941 	     * larger than the link's MTU.
942 	     * XXX: IFF_FRAGMENTABLE (or such) flag has not been defined yet...
943 	     */
944 
945 	    || ifp->if_flags & IFF_FRAGMENTABLE
946 #endif
947 	    )
948 	{
949  		/* Record statistics for this interface address. */
950  		if (ia && !(flags & IPV6_FORWARDING)) {
951  			ia->ia_ifa.if_opackets++;
952  			ia->ia_ifa.if_obytes += m->m_pkthdr.len;
953  		}
954 #ifdef IPSEC
955 		/* clean ipsec history once it goes out of the node */
956 		ipsec_delaux(m);
957 #endif
958 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
959 		goto done;
960 	} else if (mtu < IPV6_MMTU) {
961 		/*
962 		 * note that path MTU is never less than IPV6_MMTU
963 		 * (see icmp6_input).
964 		 */
965 		error = EMSGSIZE;
966 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
967 		goto bad;
968 	} else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
969 		error = EMSGSIZE;
970 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
971 		goto bad;
972 	} else {
973 		struct mbuf **mnext, *m_frgpart;
974 		struct ip6_frag *ip6f;
975 		u_int32_t id = htonl(ip6_id++);
976 		u_char nextproto;
977 
978 		/*
979 		 * Too large for the destination or interface;
980 		 * fragment if possible.
981 		 * Must be able to put at least 8 bytes per fragment.
982 		 */
983 		hlen = unfragpartlen;
984 		if (mtu > IPV6_MAXPACKET)
985 			mtu = IPV6_MAXPACKET;
986 
987 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
988 		if (len < 8) {
989 			error = EMSGSIZE;
990 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
991 			goto bad;
992 		}
993 
994 		mnext = &m->m_nextpkt;
995 
996 		/*
997 		 * Change the next header field of the last header in the
998 		 * unfragmentable part.
999 		 */
1000 		if (exthdrs.ip6e_rthdr) {
1001 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1002 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1003 		} else if (exthdrs.ip6e_dest1) {
1004 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1005 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1006 		} else if (exthdrs.ip6e_hbh) {
1007 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1008 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1009 		} else {
1010 			nextproto = ip6->ip6_nxt;
1011 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1012 		}
1013 
1014 		/*
1015 		 * Loop through length of segment after first fragment,
1016 		 * make new header and copy data of each part and link onto
1017 		 * chain.
1018 		 */
1019 		m0 = m;
1020 		for (off = hlen; off < tlen; off += len) {
1021 			MGETHDR(m, MB_DONTWAIT, MT_HEADER);
1022 			if (!m) {
1023 				error = ENOBUFS;
1024 				ip6stat.ip6s_odropped++;
1025 				goto sendorfree;
1026 			}
1027 			m->m_pkthdr.rcvif = NULL;
1028 			m->m_flags = m0->m_flags & M_COPYFLAGS;
1029 			*mnext = m;
1030 			mnext = &m->m_nextpkt;
1031 			m->m_data += max_linkhdr;
1032 			mhip6 = mtod(m, struct ip6_hdr *);
1033 			*mhip6 = *ip6;
1034 			m->m_len = sizeof(*mhip6);
1035  			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1036  			if (error) {
1037 				ip6stat.ip6s_odropped++;
1038 				goto sendorfree;
1039 			}
1040 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1041 			if (off + len >= tlen)
1042 				len = tlen - off;
1043 			else
1044 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1045 			mhip6->ip6_plen = htons((u_short)(len + hlen +
1046 							  sizeof(*ip6f) -
1047 							  sizeof(struct ip6_hdr)));
1048 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1049 				error = ENOBUFS;
1050 				ip6stat.ip6s_odropped++;
1051 				goto sendorfree;
1052 			}
1053 			m_cat(m, m_frgpart);
1054 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1055 			m->m_pkthdr.rcvif = (struct ifnet *)0;
1056 			ip6f->ip6f_reserved = 0;
1057 			ip6f->ip6f_ident = id;
1058 			ip6f->ip6f_nxt = nextproto;
1059 			ip6stat.ip6s_ofragments++;
1060 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1061 		}
1062 
1063 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1064 	}
1065 
1066 	/*
1067 	 * Remove leading garbages.
1068 	 */
1069 sendorfree:
1070 	m = m0->m_nextpkt;
1071 	m0->m_nextpkt = 0;
1072 	m_freem(m0);
1073 	for (m0 = m; m; m = m0) {
1074 		m0 = m->m_nextpkt;
1075 		m->m_nextpkt = 0;
1076 		if (error == 0) {
1077  			/* Record statistics for this interface address. */
1078  			if (ia) {
1079  				ia->ia_ifa.if_opackets++;
1080  				ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1081  			}
1082 #ifdef IPSEC
1083 			/* clean ipsec history once it goes out of the node */
1084 			ipsec_delaux(m);
1085 #endif
1086 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1087 		} else
1088 			m_freem(m);
1089 	}
1090 
1091 	if (error == 0)
1092 		ip6stat.ip6s_fragmented++;
1093 
1094 done:
1095 	if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1096 		RTFREE(ro->ro_rt);
1097 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1098 		RTFREE(ro_pmtu->ro_rt);
1099 	}
1100 
1101 #ifdef IPSEC
1102 	if (sp != NULL)
1103 		key_freesp(sp);
1104 #endif /* IPSEC */
1105 #ifdef FAST_IPSEC
1106 	if (sp != NULL)
1107 		KEY_FREESP(&sp);
1108 #endif /* FAST_IPSEC */
1109 
1110 	return(error);
1111 
1112 freehdrs:
1113 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1114 	m_freem(exthdrs.ip6e_dest1);
1115 	m_freem(exthdrs.ip6e_rthdr);
1116 	m_freem(exthdrs.ip6e_dest2);
1117 	/* fall through */
1118 bad:
1119 	m_freem(m);
1120 	goto done;
1121 }
1122 
1123 static int
1124 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1125 {
1126 	struct mbuf *m;
1127 
1128 	if (hlen > MCLBYTES)
1129 		return(ENOBUFS); /* XXX */
1130 
1131 	MGET(m, MB_DONTWAIT, MT_DATA);
1132 	if (!m)
1133 		return(ENOBUFS);
1134 
1135 	if (hlen > MLEN) {
1136 		MCLGET(m, MB_DONTWAIT);
1137 		if ((m->m_flags & M_EXT) == 0) {
1138 			m_free(m);
1139 			return(ENOBUFS);
1140 		}
1141 	}
1142 	m->m_len = hlen;
1143 	if (hdr)
1144 		bcopy(hdr, mtod(m, caddr_t), hlen);
1145 
1146 	*mp = m;
1147 	return(0);
1148 }
1149 
1150 /*
1151  * Insert jumbo payload option.
1152  */
1153 static int
1154 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1155 {
1156 	struct mbuf *mopt;
1157 	u_char *optbuf;
1158 	u_int32_t v;
1159 
1160 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1161 
1162 	/*
1163 	 * If there is no hop-by-hop options header, allocate new one.
1164 	 * If there is one but it doesn't have enough space to store the
1165 	 * jumbo payload option, allocate a cluster to store the whole options.
1166 	 * Otherwise, use it to store the options.
1167 	 */
1168 	if (exthdrs->ip6e_hbh == 0) {
1169 		MGET(mopt, MB_DONTWAIT, MT_DATA);
1170 		if (mopt == 0)
1171 			return(ENOBUFS);
1172 		mopt->m_len = JUMBOOPTLEN;
1173 		optbuf = mtod(mopt, u_char *);
1174 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1175 		exthdrs->ip6e_hbh = mopt;
1176 	} else {
1177 		struct ip6_hbh *hbh;
1178 
1179 		mopt = exthdrs->ip6e_hbh;
1180 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1181 			/*
1182 			 * XXX assumption:
1183 			 * - exthdrs->ip6e_hbh is not referenced from places
1184 			 *   other than exthdrs.
1185 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1186 			 */
1187 			int oldoptlen = mopt->m_len;
1188 			struct mbuf *n;
1189 
1190 			/*
1191 			 * XXX: give up if the whole (new) hbh header does
1192 			 * not fit even in an mbuf cluster.
1193 			 */
1194 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1195 				return(ENOBUFS);
1196 
1197 			/*
1198 			 * As a consequence, we must always prepare a cluster
1199 			 * at this point.
1200 			 */
1201 			MGET(n, MB_DONTWAIT, MT_DATA);
1202 			if (n) {
1203 				MCLGET(n, MB_DONTWAIT);
1204 				if ((n->m_flags & M_EXT) == 0) {
1205 					m_freem(n);
1206 					n = NULL;
1207 				}
1208 			}
1209 			if (!n)
1210 				return(ENOBUFS);
1211 			n->m_len = oldoptlen + JUMBOOPTLEN;
1212 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1213 			      oldoptlen);
1214 			optbuf = mtod(n, caddr_t) + oldoptlen;
1215 			m_freem(mopt);
1216 			mopt = exthdrs->ip6e_hbh = n;
1217 		} else {
1218 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1219 			mopt->m_len += JUMBOOPTLEN;
1220 		}
1221 		optbuf[0] = IP6OPT_PADN;
1222 		optbuf[1] = 1;
1223 
1224 		/*
1225 		 * Adjust the header length according to the pad and
1226 		 * the jumbo payload option.
1227 		 */
1228 		hbh = mtod(mopt, struct ip6_hbh *);
1229 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1230 	}
1231 
1232 	/* fill in the option. */
1233 	optbuf[2] = IP6OPT_JUMBO;
1234 	optbuf[3] = 4;
1235 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1236 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1237 
1238 	/* finally, adjust the packet header length */
1239 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1240 
1241 	return(0);
1242 #undef JUMBOOPTLEN
1243 }
1244 
1245 /*
1246  * Insert fragment header and copy unfragmentable header portions.
1247  */
1248 static int
1249 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1250 		  struct ip6_frag **frghdrp)
1251 {
1252 	struct mbuf *n, *mlast;
1253 
1254 	if (hlen > sizeof(struct ip6_hdr)) {
1255 		n = m_copym(m0, sizeof(struct ip6_hdr),
1256 			    hlen - sizeof(struct ip6_hdr), MB_DONTWAIT);
1257 		if (n == 0)
1258 			return(ENOBUFS);
1259 		m->m_next = n;
1260 	} else
1261 		n = m;
1262 
1263 	/* Search for the last mbuf of unfragmentable part. */
1264 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1265 		;
1266 
1267 	if ((mlast->m_flags & M_EXT) == 0 &&
1268 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1269 		/* use the trailing space of the last mbuf for the fragment hdr */
1270 		*frghdrp =
1271 			(struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len);
1272 		mlast->m_len += sizeof(struct ip6_frag);
1273 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1274 	} else {
1275 		/* allocate a new mbuf for the fragment header */
1276 		struct mbuf *mfrg;
1277 
1278 		MGET(mfrg, MB_DONTWAIT, MT_DATA);
1279 		if (mfrg == 0)
1280 			return(ENOBUFS);
1281 		mfrg->m_len = sizeof(struct ip6_frag);
1282 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1283 		mlast->m_next = mfrg;
1284 	}
1285 
1286 	return(0);
1287 }
1288 
1289 /*
1290  * IP6 socket option processing.
1291  */
1292 int
1293 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1294 {
1295 	int privileged;
1296 	struct inpcb *in6p = sotoinpcb(so);
1297 	int error, optval;
1298 	int level, op, optname;
1299 	int optlen;
1300 	struct thread *td;
1301 
1302 	if (sopt) {
1303 		level = sopt->sopt_level;
1304 		op = sopt->sopt_dir;
1305 		optname = sopt->sopt_name;
1306 		optlen = sopt->sopt_valsize;
1307 		td = sopt->sopt_td;
1308 	} else {
1309 		panic("ip6_ctloutput: arg soopt is NULL");
1310 		/* NOT REACHED */
1311 		td = NULL;
1312 	}
1313 	error = optval = 0;
1314 
1315 	privileged = (td == NULL || suser(td)) ? 0 : 1;
1316 
1317 	if (level == IPPROTO_IPV6) {
1318 		switch (op) {
1319 
1320 		case SOPT_SET:
1321 			switch (optname) {
1322 			case IPV6_PKTOPTIONS:
1323 			{
1324 				struct mbuf *m;
1325 
1326 				error = soopt_getm(sopt, &m); /* XXX */
1327 				if (error != NULL)
1328 					break;
1329 				error = soopt_mcopyin(sopt, m); /* XXX */
1330 				if (error != NULL)
1331 					break;
1332 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1333 						    m, so, sopt);
1334 				m_freem(m); /* XXX */
1335 				break;
1336 			}
1337 
1338 			/*
1339 			 * Use of some Hop-by-Hop options or some
1340 			 * Destination options, might require special
1341 			 * privilege.  That is, normal applications
1342 			 * (without special privilege) might be forbidden
1343 			 * from setting certain options in outgoing packets,
1344 			 * and might never see certain options in received
1345 			 * packets. [RFC 2292 Section 6]
1346 			 * KAME specific note:
1347 			 *  KAME prevents non-privileged users from sending or
1348 			 *  receiving ANY hbh/dst options in order to avoid
1349 			 *  overhead of parsing options in the kernel.
1350 			 */
1351 			case IPV6_UNICAST_HOPS:
1352 			case IPV6_CHECKSUM:
1353 			case IPV6_FAITH:
1354 
1355 			case IPV6_V6ONLY:
1356 				if (optlen != sizeof(int)) {
1357 					error = EINVAL;
1358 					break;
1359 				}
1360 				error = sooptcopyin(sopt, &optval,
1361 					sizeof optval, sizeof optval);
1362 				if (error)
1363 					break;
1364 				switch (optname) {
1365 
1366 				case IPV6_UNICAST_HOPS:
1367 					if (optval < -1 || optval >= 256)
1368 						error = EINVAL;
1369 					else {
1370 						/* -1 = kernel default */
1371 						in6p->in6p_hops = optval;
1372 
1373 						if ((in6p->in6p_vflag &
1374 						     INP_IPV4) != 0)
1375 							in6p->inp_ip_ttl = optval;
1376 					}
1377 					break;
1378 #define OPTSET(bit) \
1379 do { \
1380 	if (optval) \
1381 		in6p->in6p_flags |= (bit); \
1382 	else \
1383 		in6p->in6p_flags &= ~(bit); \
1384 } while (0)
1385 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1386 
1387 				case IPV6_CHECKSUM:
1388 					in6p->in6p_cksum = optval;
1389 					break;
1390 
1391 				case IPV6_FAITH:
1392 					OPTSET(IN6P_FAITH);
1393 					break;
1394 
1395 				case IPV6_V6ONLY:
1396 					/*
1397 					 * make setsockopt(IPV6_V6ONLY)
1398 					 * available only prior to bind(2).
1399 					 * see ipng mailing list, Jun 22 2001.
1400 					 */
1401 					if (in6p->in6p_lport ||
1402 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
1403 					{
1404 						error = EINVAL;
1405 						break;
1406 					}
1407 					OPTSET(IN6P_IPV6_V6ONLY);
1408 					if (optval)
1409 						in6p->in6p_vflag &= ~INP_IPV4;
1410 					else
1411 						in6p->in6p_vflag |= INP_IPV4;
1412 					break;
1413 				}
1414 				break;
1415 
1416 			case IPV6_PKTINFO:
1417 			case IPV6_HOPLIMIT:
1418 			case IPV6_HOPOPTS:
1419 			case IPV6_DSTOPTS:
1420 			case IPV6_RTHDR:
1421 				/* RFC 2292 */
1422 				if (optlen != sizeof(int)) {
1423 					error = EINVAL;
1424 					break;
1425 				}
1426 				error = sooptcopyin(sopt, &optval,
1427 					sizeof optval, sizeof optval);
1428 				if (error)
1429 					break;
1430 				switch (optname) {
1431 				case IPV6_PKTINFO:
1432 					OPTSET(IN6P_PKTINFO);
1433 					break;
1434 				case IPV6_HOPLIMIT:
1435 					OPTSET(IN6P_HOPLIMIT);
1436 					break;
1437 				case IPV6_HOPOPTS:
1438 					/*
1439 					 * Check super-user privilege.
1440 					 * See comments for IPV6_RECVHOPOPTS.
1441 					 */
1442 					if (!privileged)
1443 						return(EPERM);
1444 					OPTSET(IN6P_HOPOPTS);
1445 					break;
1446 				case IPV6_DSTOPTS:
1447 					if (!privileged)
1448 						return(EPERM);
1449 					OPTSET(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1450 					break;
1451 				case IPV6_RTHDR:
1452 					OPTSET(IN6P_RTHDR);
1453 					break;
1454 				}
1455 				break;
1456 #undef OPTSET
1457 
1458 			case IPV6_MULTICAST_IF:
1459 			case IPV6_MULTICAST_HOPS:
1460 			case IPV6_MULTICAST_LOOP:
1461 			case IPV6_JOIN_GROUP:
1462 			case IPV6_LEAVE_GROUP:
1463 			    {
1464 				struct mbuf *m;
1465 				if (sopt->sopt_valsize > MLEN) {
1466 					error = EMSGSIZE;
1467 					break;
1468 				}
1469 				/* XXX */
1470 				MGET(m, sopt->sopt_td ? MB_WAIT : MB_DONTWAIT, MT_HEADER);
1471 				if (m == 0) {
1472 					error = ENOBUFS;
1473 					break;
1474 				}
1475 				m->m_len = sopt->sopt_valsize;
1476 				error = sooptcopyin(sopt, mtod(m, char *),
1477 						    m->m_len, m->m_len);
1478 				error =	ip6_setmoptions(sopt->sopt_name,
1479 							&in6p->in6p_moptions,
1480 							m);
1481 				(void)m_free(m);
1482 			    }
1483 				break;
1484 
1485 			case IPV6_PORTRANGE:
1486 				error = sooptcopyin(sopt, &optval,
1487 				    sizeof optval, sizeof optval);
1488 				if (error)
1489 					break;
1490 
1491 				switch (optval) {
1492 				case IPV6_PORTRANGE_DEFAULT:
1493 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1494 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1495 					break;
1496 
1497 				case IPV6_PORTRANGE_HIGH:
1498 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1499 					in6p->in6p_flags |= IN6P_HIGHPORT;
1500 					break;
1501 
1502 				case IPV6_PORTRANGE_LOW:
1503 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1504 					in6p->in6p_flags |= IN6P_LOWPORT;
1505 					break;
1506 
1507 				default:
1508 					error = EINVAL;
1509 					break;
1510 				}
1511 				break;
1512 
1513 #if defined(IPSEC) || defined(FAST_IPSEC)
1514 			case IPV6_IPSEC_POLICY:
1515 			    {
1516 				caddr_t req = NULL;
1517 				size_t len = 0;
1518 				struct mbuf *m;
1519 
1520 				if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1521 					break;
1522 				if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1523 					break;
1524 				if (m) {
1525 					req = mtod(m, caddr_t);
1526 					len = m->m_len;
1527 				}
1528 				error = ipsec6_set_policy(in6p, optname, req,
1529 							  len, privileged);
1530 				m_freem(m);
1531 			    }
1532 				break;
1533 #endif /* KAME IPSEC */
1534 
1535 			case IPV6_FW_ADD:
1536 			case IPV6_FW_DEL:
1537 			case IPV6_FW_FLUSH:
1538 			case IPV6_FW_ZERO:
1539 			    {
1540 				struct mbuf *m;
1541 				struct mbuf **mp = &m;
1542 
1543 				if (ip6_fw_ctl_ptr == NULL)
1544 					return EINVAL;
1545 				/* XXX */
1546 				if ((error = soopt_getm(sopt, &m)) != 0)
1547 					break;
1548 				/* XXX */
1549 				if ((error = soopt_mcopyin(sopt, m)) != 0)
1550 					break;
1551 				error = (*ip6_fw_ctl_ptr)(optname, mp);
1552 				m = *mp;
1553 			    }
1554 				break;
1555 
1556 			default:
1557 				error = ENOPROTOOPT;
1558 				break;
1559 			}
1560 			break;
1561 
1562 		case SOPT_GET:
1563 			switch (optname) {
1564 
1565 			case IPV6_PKTOPTIONS:
1566 				if (in6p->in6p_options) {
1567 					struct mbuf *m;
1568 					m = m_copym(in6p->in6p_options,
1569 					    0, M_COPYALL, MB_WAIT);
1570 					error = soopt_mcopyout(sopt, m);
1571 					if (error == 0)
1572 						m_freem(m);
1573 				} else
1574 					sopt->sopt_valsize = 0;
1575 				break;
1576 
1577 			case IPV6_UNICAST_HOPS:
1578 			case IPV6_CHECKSUM:
1579 
1580 			case IPV6_FAITH:
1581 			case IPV6_V6ONLY:
1582 			case IPV6_PORTRANGE:
1583 				switch (optname) {
1584 
1585 				case IPV6_UNICAST_HOPS:
1586 					optval = in6p->in6p_hops;
1587 					break;
1588 
1589 				case IPV6_CHECKSUM:
1590 					optval = in6p->in6p_cksum;
1591 					break;
1592 
1593 				case IPV6_FAITH:
1594 					optval = OPTBIT(IN6P_FAITH);
1595 					break;
1596 
1597 				case IPV6_V6ONLY:
1598 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
1599 					break;
1600 
1601 				case IPV6_PORTRANGE:
1602 				    {
1603 					int flags;
1604 					flags = in6p->in6p_flags;
1605 					if (flags & IN6P_HIGHPORT)
1606 						optval = IPV6_PORTRANGE_HIGH;
1607 					else if (flags & IN6P_LOWPORT)
1608 						optval = IPV6_PORTRANGE_LOW;
1609 					else
1610 						optval = 0;
1611 					break;
1612 				    }
1613 				}
1614 				error = sooptcopyout(sopt, &optval,
1615 					sizeof optval);
1616 				break;
1617 
1618 			case IPV6_PKTINFO:
1619 			case IPV6_HOPLIMIT:
1620 			case IPV6_HOPOPTS:
1621 			case IPV6_RTHDR:
1622 			case IPV6_DSTOPTS:
1623 				if (optname == IPV6_HOPOPTS ||
1624 				    optname == IPV6_DSTOPTS ||
1625 				    !privileged)
1626 					return(EPERM);
1627 				switch (optname) {
1628 				case IPV6_PKTINFO:
1629 					optval = OPTBIT(IN6P_PKTINFO);
1630 					break;
1631 				case IPV6_HOPLIMIT:
1632 					optval = OPTBIT(IN6P_HOPLIMIT);
1633 					break;
1634 				case IPV6_HOPOPTS:
1635 					if (!privileged)
1636 						return(EPERM);
1637 					optval = OPTBIT(IN6P_HOPOPTS);
1638 					break;
1639 				case IPV6_RTHDR:
1640 					optval = OPTBIT(IN6P_RTHDR);
1641 					break;
1642 				case IPV6_DSTOPTS:
1643 					if (!privileged)
1644 						return(EPERM);
1645 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1646 					break;
1647 				}
1648 				error = sooptcopyout(sopt, &optval,
1649 					sizeof optval);
1650 				break;
1651 
1652 			case IPV6_MULTICAST_IF:
1653 			case IPV6_MULTICAST_HOPS:
1654 			case IPV6_MULTICAST_LOOP:
1655 			case IPV6_JOIN_GROUP:
1656 			case IPV6_LEAVE_GROUP:
1657 			    {
1658 				struct mbuf *m;
1659 				error = ip6_getmoptions(sopt->sopt_name,
1660 						in6p->in6p_moptions, &m);
1661 				if (error == 0)
1662 					error = sooptcopyout(sopt,
1663 						mtod(m, char *), m->m_len);
1664 				m_freem(m);
1665 			    }
1666 				break;
1667 
1668 #if defined(IPSEC) || defined(FAST_IPSEC)
1669 			case IPV6_IPSEC_POLICY:
1670 			  {
1671 				caddr_t req = NULL;
1672 				size_t len = 0;
1673 				struct mbuf *m = NULL;
1674 				struct mbuf **mp = &m;
1675 
1676 				error = soopt_getm(sopt, &m); /* XXX */
1677 				if (error != NULL)
1678 					break;
1679 				error = soopt_mcopyin(sopt, m); /* XXX */
1680 				if (error != NULL)
1681 					break;
1682 				if (m) {
1683 					req = mtod(m, caddr_t);
1684 					len = m->m_len;
1685 				}
1686 				error = ipsec6_get_policy(in6p, req, len, mp);
1687 				if (error == 0)
1688 					error = soopt_mcopyout(sopt, m); /*XXX*/
1689 				if (error == 0 && m)
1690 					m_freem(m);
1691 				break;
1692 			  }
1693 #endif /* KAME IPSEC */
1694 
1695 			case IPV6_FW_GET:
1696 			  {
1697 				struct mbuf *m;
1698 				struct mbuf **mp = &m;
1699 
1700 				if (ip6_fw_ctl_ptr == NULL)
1701 				{
1702 					return EINVAL;
1703 				}
1704 				error = (*ip6_fw_ctl_ptr)(optname, mp);
1705 				if (error == 0)
1706 					error = soopt_mcopyout(sopt, m); /* XXX */
1707 				if (error == 0 && m)
1708 					m_freem(m);
1709 			  }
1710 				break;
1711 
1712 			default:
1713 				error = ENOPROTOOPT;
1714 				break;
1715 			}
1716 			break;
1717 		}
1718 	} else {
1719 		error = EINVAL;
1720 	}
1721 	return(error);
1722 }
1723 
1724 /*
1725  * Set up IP6 options in pcb for insertion in output packets or
1726  * specifying behavior of outgoing packets.
1727  */
1728 static int
1729 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, struct socket *so,
1730 	    struct sockopt *sopt)
1731 {
1732 	struct ip6_pktopts *opt = *pktopt;
1733 	int error = 0;
1734 	struct thread *td = sopt->sopt_td;
1735 	int priv = 0;
1736 
1737 	/* turn off any old options. */
1738 	if (opt) {
1739 #ifdef DIAGNOSTIC
1740 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
1741 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
1742 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
1743 			printf("ip6_pcbopts: all specified options are cleared.\n");
1744 #endif
1745 		ip6_clearpktopts(opt, 1, -1);
1746 	} else
1747 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
1748 	*pktopt = NULL;
1749 
1750 	if (!m || m->m_len == 0) {
1751 		/*
1752 		 * Only turning off any previous options, regardless of
1753 		 * whether the opt is just created or given.
1754 		 */
1755 		free(opt, M_IP6OPT);
1756 		return(0);
1757 	}
1758 
1759 	/*  set options specified by user. */
1760 	if (suser(td) == 0)
1761 		priv = 1;
1762 	if ((error = ip6_setpktoptions(m, opt, priv, 1)) != 0) {
1763 		ip6_clearpktopts(opt, 1, -1); /* XXX: discard all options */
1764 		free(opt, M_IP6OPT);
1765 		return(error);
1766 	}
1767 	*pktopt = opt;
1768 	return(0);
1769 }
1770 
1771 /*
1772  * initialize ip6_pktopts.  beware that there are non-zero default values in
1773  * the struct.
1774  */
1775 void
1776 init_ip6pktopts(struct ip6_pktopts *opt)
1777 {
1778 
1779 	bzero(opt, sizeof(*opt));
1780 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
1781 }
1782 
1783 void
1784 ip6_clearpktopts(struct ip6_pktopts *pktopt, int needfree, int optname)
1785 {
1786 	if (pktopt == NULL)
1787 		return;
1788 
1789 	if (optname == -1) {
1790 		if (needfree && pktopt->ip6po_pktinfo)
1791 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
1792 		pktopt->ip6po_pktinfo = NULL;
1793 	}
1794 	if (optname == -1)
1795 		pktopt->ip6po_hlim = -1;
1796 	if (optname == -1) {
1797 		if (needfree && pktopt->ip6po_nexthop)
1798 			free(pktopt->ip6po_nexthop, M_IP6OPT);
1799 		pktopt->ip6po_nexthop = NULL;
1800 	}
1801 	if (optname == -1) {
1802 		if (needfree && pktopt->ip6po_hbh)
1803 			free(pktopt->ip6po_hbh, M_IP6OPT);
1804 		pktopt->ip6po_hbh = NULL;
1805 	}
1806 	if (optname == -1) {
1807 		if (needfree && pktopt->ip6po_dest1)
1808 			free(pktopt->ip6po_dest1, M_IP6OPT);
1809 		pktopt->ip6po_dest1 = NULL;
1810 	}
1811 	if (optname == -1) {
1812 		if (needfree && pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
1813 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
1814 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
1815 		if (pktopt->ip6po_route.ro_rt) {
1816 			RTFREE(pktopt->ip6po_route.ro_rt);
1817 			pktopt->ip6po_route.ro_rt = NULL;
1818 		}
1819 	}
1820 	if (optname == -1) {
1821 		if (needfree && pktopt->ip6po_dest2)
1822 			free(pktopt->ip6po_dest2, M_IP6OPT);
1823 		pktopt->ip6po_dest2 = NULL;
1824 	}
1825 }
1826 
1827 #define PKTOPT_EXTHDRCPY(type) \
1828 do {\
1829 	if (src->type) {\
1830 		int hlen =\
1831 			(((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
1832 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
1833 		if (dst->type == NULL && canwait == M_NOWAIT)\
1834 			goto bad;\
1835 		bcopy(src->type, dst->type, hlen);\
1836 	}\
1837 } while (0)
1838 
1839 struct ip6_pktopts *
1840 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
1841 {
1842 	struct ip6_pktopts *dst;
1843 
1844 	if (src == NULL) {
1845 		printf("ip6_clearpktopts: invalid argument\n");
1846 		return(NULL);
1847 	}
1848 
1849 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
1850 	if (dst == NULL && canwait == M_NOWAIT)
1851 		return (NULL);
1852 	bzero(dst, sizeof(*dst));
1853 
1854 	dst->ip6po_hlim = src->ip6po_hlim;
1855 	if (src->ip6po_pktinfo) {
1856 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
1857 					    M_IP6OPT, canwait);
1858 		if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT)
1859 			goto bad;
1860 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
1861 	}
1862 	if (src->ip6po_nexthop) {
1863 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
1864 					    M_IP6OPT, canwait);
1865 		if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT)
1866 			goto bad;
1867 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
1868 		      src->ip6po_nexthop->sa_len);
1869 	}
1870 	PKTOPT_EXTHDRCPY(ip6po_hbh);
1871 	PKTOPT_EXTHDRCPY(ip6po_dest1);
1872 	PKTOPT_EXTHDRCPY(ip6po_dest2);
1873 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
1874 	return(dst);
1875 
1876   bad:
1877 	if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
1878 	if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
1879 	if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
1880 	if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
1881 	if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
1882 	if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
1883 	free(dst, M_IP6OPT);
1884 	return(NULL);
1885 }
1886 #undef PKTOPT_EXTHDRCPY
1887 
1888 void
1889 ip6_freepcbopts(struct ip6_pktopts *pktopt)
1890 {
1891 	if (pktopt == NULL)
1892 		return;
1893 
1894 	ip6_clearpktopts(pktopt, 1, -1);
1895 
1896 	free(pktopt, M_IP6OPT);
1897 }
1898 
1899 /*
1900  * Set the IP6 multicast options in response to user setsockopt().
1901  */
1902 static int
1903 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
1904 {
1905 	int error = 0;
1906 	u_int loop, ifindex;
1907 	struct ipv6_mreq *mreq;
1908 	struct ifnet *ifp;
1909 	struct ip6_moptions *im6o = *im6op;
1910 	struct route_in6 ro;
1911 	struct sockaddr_in6 *dst;
1912 	struct in6_multi_mship *imm;
1913 	struct thread *td = curthread;	/* XXX */
1914 
1915 	if (im6o == NULL) {
1916 		/*
1917 		 * No multicast option buffer attached to the pcb;
1918 		 * allocate one and initialize to default values.
1919 		 */
1920 		im6o = (struct ip6_moptions *)
1921 			malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
1922 
1923 		if (im6o == NULL)
1924 			return(ENOBUFS);
1925 		*im6op = im6o;
1926 		im6o->im6o_multicast_ifp = NULL;
1927 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1928 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
1929 		LIST_INIT(&im6o->im6o_memberships);
1930 	}
1931 
1932 	switch (optname) {
1933 
1934 	case IPV6_MULTICAST_IF:
1935 		/*
1936 		 * Select the interface for outgoing multicast packets.
1937 		 */
1938 		if (m == NULL || m->m_len != sizeof(u_int)) {
1939 			error = EINVAL;
1940 			break;
1941 		}
1942 		bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
1943 		if (ifindex < 0 || if_index < ifindex) {
1944 			error = ENXIO;	/* XXX EINVAL? */
1945 			break;
1946 		}
1947 		ifp = ifindex2ifnet[ifindex];
1948 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1949 			error = EADDRNOTAVAIL;
1950 			break;
1951 		}
1952 		im6o->im6o_multicast_ifp = ifp;
1953 		break;
1954 
1955 	case IPV6_MULTICAST_HOPS:
1956 	    {
1957 		/*
1958 		 * Set the IP6 hoplimit for outgoing multicast packets.
1959 		 */
1960 		int optval;
1961 		if (m == NULL || m->m_len != sizeof(int)) {
1962 			error = EINVAL;
1963 			break;
1964 		}
1965 		bcopy(mtod(m, u_int *), &optval, sizeof(optval));
1966 		if (optval < -1 || optval >= 256)
1967 			error = EINVAL;
1968 		else if (optval == -1)
1969 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1970 		else
1971 			im6o->im6o_multicast_hlim = optval;
1972 		break;
1973 	    }
1974 
1975 	case IPV6_MULTICAST_LOOP:
1976 		/*
1977 		 * Set the loopback flag for outgoing multicast packets.
1978 		 * Must be zero or one.
1979 		 */
1980 		if (m == NULL || m->m_len != sizeof(u_int)) {
1981 			error = EINVAL;
1982 			break;
1983 		}
1984 		bcopy(mtod(m, u_int *), &loop, sizeof(loop));
1985 		if (loop > 1) {
1986 			error = EINVAL;
1987 			break;
1988 		}
1989 		im6o->im6o_multicast_loop = loop;
1990 		break;
1991 
1992 	case IPV6_JOIN_GROUP:
1993 		/*
1994 		 * Add a multicast group membership.
1995 		 * Group must be a valid IP6 multicast address.
1996 		 */
1997 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
1998 			error = EINVAL;
1999 			break;
2000 		}
2001 		mreq = mtod(m, struct ipv6_mreq *);
2002 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2003 			/*
2004 			 * We use the unspecified address to specify to accept
2005 			 * all multicast addresses. Only super user is allowed
2006 			 * to do this.
2007 			 */
2008 			if (suser(td))
2009 			{
2010 				error = EACCES;
2011 				break;
2012 			}
2013 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2014 			error = EINVAL;
2015 			break;
2016 		}
2017 
2018 		/*
2019 		 * If the interface is specified, validate it.
2020 		 */
2021 		if (mreq->ipv6mr_interface < 0
2022 		 || if_index < mreq->ipv6mr_interface) {
2023 			error = ENXIO;	/* XXX EINVAL? */
2024 			break;
2025 		}
2026 		/*
2027 		 * If no interface was explicitly specified, choose an
2028 		 * appropriate one according to the given multicast address.
2029 		 */
2030 		if (mreq->ipv6mr_interface == 0) {
2031 			/*
2032 			 * If the multicast address is in node-local scope,
2033 			 * the interface should be a loopback interface.
2034 			 * Otherwise, look up the routing table for the
2035 			 * address, and choose the outgoing interface.
2036 			 *   XXX: is it a good approach?
2037 			 */
2038 			if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
2039 				ifp = &loif[0];
2040 			} else {
2041 				ro.ro_rt = NULL;
2042 				dst = (struct sockaddr_in6 *)&ro.ro_dst;
2043 				bzero(dst, sizeof(*dst));
2044 				dst->sin6_len = sizeof(struct sockaddr_in6);
2045 				dst->sin6_family = AF_INET6;
2046 				dst->sin6_addr = mreq->ipv6mr_multiaddr;
2047 				rtalloc((struct route *)&ro);
2048 				if (ro.ro_rt == NULL) {
2049 					error = EADDRNOTAVAIL;
2050 					break;
2051 				}
2052 				ifp = ro.ro_rt->rt_ifp;
2053 				rtfree(ro.ro_rt);
2054 			}
2055 		} else
2056 			ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2057 
2058 		/*
2059 		 * See if we found an interface, and confirm that it
2060 		 * supports multicast
2061 		 */
2062 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2063 			error = EADDRNOTAVAIL;
2064 			break;
2065 		}
2066 		/*
2067 		 * Put interface index into the multicast address,
2068 		 * if the address has link-local scope.
2069 		 */
2070 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2071 			mreq->ipv6mr_multiaddr.s6_addr16[1]
2072 				= htons(mreq->ipv6mr_interface);
2073 		}
2074 		/*
2075 		 * See if the membership already exists.
2076 		 */
2077 		for (imm = im6o->im6o_memberships.lh_first;
2078 		     imm != NULL; imm = imm->i6mm_chain.le_next)
2079 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
2080 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2081 					       &mreq->ipv6mr_multiaddr))
2082 				break;
2083 		if (imm != NULL) {
2084 			error = EADDRINUSE;
2085 			break;
2086 		}
2087 		/*
2088 		 * Everything looks good; add a new record to the multicast
2089 		 * address list for the given interface.
2090 		 */
2091 		imm = malloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
2092 		if (imm == NULL) {
2093 			error = ENOBUFS;
2094 			break;
2095 		}
2096 		if ((imm->i6mm_maddr =
2097 		     in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
2098 			free(imm, M_IPMADDR);
2099 			break;
2100 		}
2101 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2102 		break;
2103 
2104 	case IPV6_LEAVE_GROUP:
2105 		/*
2106 		 * Drop a multicast group membership.
2107 		 * Group must be a valid IP6 multicast address.
2108 		 */
2109 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2110 			error = EINVAL;
2111 			break;
2112 		}
2113 		mreq = mtod(m, struct ipv6_mreq *);
2114 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2115 			if (suser(td)) {
2116 				error = EACCES;
2117 				break;
2118 			}
2119 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2120 			error = EINVAL;
2121 			break;
2122 		}
2123 		/*
2124 		 * If an interface address was specified, get a pointer
2125 		 * to its ifnet structure.
2126 		 */
2127 		if (mreq->ipv6mr_interface < 0
2128 		 || if_index < mreq->ipv6mr_interface) {
2129 			error = ENXIO;	/* XXX EINVAL? */
2130 			break;
2131 		}
2132 		ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2133 		/*
2134 		 * Put interface index into the multicast address,
2135 		 * if the address has link-local scope.
2136 		 */
2137 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2138 			mreq->ipv6mr_multiaddr.s6_addr16[1]
2139 				= htons(mreq->ipv6mr_interface);
2140 		}
2141 		/*
2142 		 * Find the membership in the membership list.
2143 		 */
2144 		for (imm = im6o->im6o_memberships.lh_first;
2145 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
2146 			if ((ifp == NULL ||
2147 			     imm->i6mm_maddr->in6m_ifp == ifp) &&
2148 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2149 					       &mreq->ipv6mr_multiaddr))
2150 				break;
2151 		}
2152 		if (imm == NULL) {
2153 			/* Unable to resolve interface */
2154 			error = EADDRNOTAVAIL;
2155 			break;
2156 		}
2157 		/*
2158 		 * Give up the multicast address record to which the
2159 		 * membership points.
2160 		 */
2161 		LIST_REMOVE(imm, i6mm_chain);
2162 		in6_delmulti(imm->i6mm_maddr);
2163 		free(imm, M_IPMADDR);
2164 		break;
2165 
2166 	default:
2167 		error = EOPNOTSUPP;
2168 		break;
2169 	}
2170 
2171 	/*
2172 	 * If all options have default values, no need to keep the mbuf.
2173 	 */
2174 	if (im6o->im6o_multicast_ifp == NULL &&
2175 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2176 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2177 	    im6o->im6o_memberships.lh_first == NULL) {
2178 		free(*im6op, M_IPMOPTS);
2179 		*im6op = NULL;
2180 	}
2181 
2182 	return(error);
2183 }
2184 
2185 /*
2186  * Return the IP6 multicast options in response to user getsockopt().
2187  */
2188 static int
2189 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp)
2190 {
2191 	u_int *hlim, *loop, *ifindex;
2192 
2193 	*mp = m_get(MB_WAIT, MT_HEADER);		/* XXX */
2194 
2195 	switch (optname) {
2196 
2197 	case IPV6_MULTICAST_IF:
2198 		ifindex = mtod(*mp, u_int *);
2199 		(*mp)->m_len = sizeof(u_int);
2200 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2201 			*ifindex = 0;
2202 		else
2203 			*ifindex = im6o->im6o_multicast_ifp->if_index;
2204 		return(0);
2205 
2206 	case IPV6_MULTICAST_HOPS:
2207 		hlim = mtod(*mp, u_int *);
2208 		(*mp)->m_len = sizeof(u_int);
2209 		if (im6o == NULL)
2210 			*hlim = ip6_defmcasthlim;
2211 		else
2212 			*hlim = im6o->im6o_multicast_hlim;
2213 		return(0);
2214 
2215 	case IPV6_MULTICAST_LOOP:
2216 		loop = mtod(*mp, u_int *);
2217 		(*mp)->m_len = sizeof(u_int);
2218 		if (im6o == NULL)
2219 			*loop = ip6_defmcasthlim;
2220 		else
2221 			*loop = im6o->im6o_multicast_loop;
2222 		return(0);
2223 
2224 	default:
2225 		return(EOPNOTSUPP);
2226 	}
2227 }
2228 
2229 /*
2230  * Discard the IP6 multicast options.
2231  */
2232 void
2233 ip6_freemoptions(struct ip6_moptions *im6o)
2234 {
2235 	struct in6_multi_mship *imm;
2236 
2237 	if (im6o == NULL)
2238 		return;
2239 
2240 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2241 		LIST_REMOVE(imm, i6mm_chain);
2242 		if (imm->i6mm_maddr)
2243 			in6_delmulti(imm->i6mm_maddr);
2244 		free(imm, M_IPMADDR);
2245 	}
2246 	free(im6o, M_IPMOPTS);
2247 }
2248 
2249 /*
2250  * Set IPv6 outgoing packet options based on advanced API.
2251  */
2252 int
2253 ip6_setpktoptions(struct mbuf *control, struct ip6_pktopts *opt, int priv,
2254 		  int needcopy)
2255 {
2256 	struct cmsghdr *cm = 0;
2257 
2258 	if (control == 0 || opt == 0)
2259 		return(EINVAL);
2260 
2261 	init_ip6pktopts(opt);
2262 
2263 	/*
2264 	 * XXX: Currently, we assume all the optional information is stored
2265 	 * in a single mbuf.
2266 	 */
2267 	if (control->m_next)
2268 		return(EINVAL);
2269 
2270 	for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2271 		     control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2272 		cm = mtod(control, struct cmsghdr *);
2273 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2274 			return(EINVAL);
2275 		if (cm->cmsg_level != IPPROTO_IPV6)
2276 			continue;
2277 
2278 		/*
2279 		 * XXX should check if RFC2292 API is mixed with 2292bis API
2280 		 */
2281 		switch (cm->cmsg_type) {
2282 		case IPV6_PKTINFO:
2283 			if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo)))
2284 				return(EINVAL);
2285 			if (needcopy) {
2286 				/* XXX: Is it really WAITOK? */
2287 				opt->ip6po_pktinfo =
2288 					malloc(sizeof(struct in6_pktinfo),
2289 					       M_IP6OPT, M_WAITOK);
2290 				bcopy(CMSG_DATA(cm), opt->ip6po_pktinfo,
2291 				    sizeof(struct in6_pktinfo));
2292 			} else
2293 				opt->ip6po_pktinfo =
2294 					(struct in6_pktinfo *)CMSG_DATA(cm);
2295 			if (opt->ip6po_pktinfo->ipi6_ifindex &&
2296 			    IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr))
2297 				opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] =
2298 					htons(opt->ip6po_pktinfo->ipi6_ifindex);
2299 
2300 			if (opt->ip6po_pktinfo->ipi6_ifindex > if_index
2301 			 || opt->ip6po_pktinfo->ipi6_ifindex < 0) {
2302 				return(ENXIO);
2303 			}
2304 
2305 			/*
2306 			 * Check if the requested source address is indeed a
2307 			 * unicast address assigned to the node, and can be
2308 			 * used as the packet's source address.
2309 			 */
2310 			if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
2311 				struct in6_ifaddr *ia6;
2312 				struct sockaddr_in6 sin6;
2313 
2314 				bzero(&sin6, sizeof(sin6));
2315 				sin6.sin6_len = sizeof(sin6);
2316 				sin6.sin6_family = AF_INET6;
2317 				sin6.sin6_addr =
2318 					opt->ip6po_pktinfo->ipi6_addr;
2319 				ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(sin6tosa(&sin6));
2320 				if (ia6 == NULL ||
2321 				    (ia6->ia6_flags & (IN6_IFF_ANYCAST |
2322 						       IN6_IFF_NOTREADY)) != 0)
2323 					return(EADDRNOTAVAIL);
2324 			}
2325 			break;
2326 
2327 		case IPV6_HOPLIMIT:
2328 			if (cm->cmsg_len != CMSG_LEN(sizeof(int)))
2329 				return(EINVAL);
2330 
2331 			opt->ip6po_hlim = *(int *)CMSG_DATA(cm);
2332 			if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255)
2333 				return(EINVAL);
2334 			break;
2335 
2336 		case IPV6_NEXTHOP:
2337 			if (!priv)
2338 				return(EPERM);
2339 
2340 			if (cm->cmsg_len < sizeof(u_char) ||
2341 			    /* check if cmsg_len is large enough for sa_len */
2342 			    cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm)))
2343 				return(EINVAL);
2344 
2345 			if (needcopy) {
2346 				opt->ip6po_nexthop =
2347 					malloc(*CMSG_DATA(cm),
2348 					       M_IP6OPT, M_WAITOK);
2349 				bcopy(CMSG_DATA(cm),
2350 				      opt->ip6po_nexthop,
2351 				      *CMSG_DATA(cm));
2352 			} else
2353 				opt->ip6po_nexthop =
2354 					(struct sockaddr *)CMSG_DATA(cm);
2355 			break;
2356 
2357 		case IPV6_HOPOPTS:
2358 		{
2359 			struct ip6_hbh *hbh;
2360 			int hbhlen;
2361 
2362 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh)))
2363 				return(EINVAL);
2364 			hbh = (struct ip6_hbh *)CMSG_DATA(cm);
2365 			hbhlen = (hbh->ip6h_len + 1) << 3;
2366 			if (cm->cmsg_len != CMSG_LEN(hbhlen))
2367 				return(EINVAL);
2368 
2369 			if (needcopy) {
2370 				opt->ip6po_hbh =
2371 					malloc(hbhlen, M_IP6OPT, M_WAITOK);
2372 				bcopy(hbh, opt->ip6po_hbh, hbhlen);
2373 			} else
2374 				opt->ip6po_hbh = hbh;
2375 			break;
2376 		}
2377 
2378 		case IPV6_DSTOPTS:
2379 		{
2380 			struct ip6_dest *dest, **newdest;
2381 			int destlen;
2382 
2383 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest)))
2384 				return(EINVAL);
2385 			dest = (struct ip6_dest *)CMSG_DATA(cm);
2386 			destlen = (dest->ip6d_len + 1) << 3;
2387 			if (cm->cmsg_len != CMSG_LEN(destlen))
2388 				return(EINVAL);
2389 
2390 			/*
2391 			 * The old advacned API is ambiguous on this
2392 			 * point. Our approach is to determine the
2393 			 * position based according to the existence
2394 			 * of a routing header. Note, however, that
2395 			 * this depends on the order of the extension
2396 			 * headers in the ancillary data; the 1st part
2397 			 * of the destination options header must
2398 			 * appear before the routing header in the
2399 			 * ancillary data, too.
2400 			 * RFC2292bis solved the ambiguity by
2401 			 * introducing separate cmsg types.
2402 			 */
2403 			if (opt->ip6po_rthdr == NULL)
2404 				newdest = &opt->ip6po_dest1;
2405 			else
2406 				newdest = &opt->ip6po_dest2;
2407 
2408 			if (needcopy) {
2409 				*newdest = malloc(destlen, M_IP6OPT, M_WAITOK);
2410 				bcopy(dest, *newdest, destlen);
2411 			} else
2412 				*newdest = dest;
2413 
2414 			break;
2415 		}
2416 
2417 		case IPV6_RTHDR:
2418 		{
2419 			struct ip6_rthdr *rth;
2420 			int rthlen;
2421 
2422 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr)))
2423 				return(EINVAL);
2424 			rth = (struct ip6_rthdr *)CMSG_DATA(cm);
2425 			rthlen = (rth->ip6r_len + 1) << 3;
2426 			if (cm->cmsg_len != CMSG_LEN(rthlen))
2427 				return(EINVAL);
2428 
2429 			switch (rth->ip6r_type) {
2430 			case IPV6_RTHDR_TYPE_0:
2431 				/* must contain one addr */
2432 				if (rth->ip6r_len == 0)
2433 					return(EINVAL);
2434 				/* length must be even */
2435 				if (rth->ip6r_len % 2)
2436 					return(EINVAL);
2437 				if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2438 					return(EINVAL);
2439 				break;
2440 			default:
2441 				return(EINVAL);	/* not supported */
2442 			}
2443 
2444 			if (needcopy) {
2445 				opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT,
2446 							  M_WAITOK);
2447 				bcopy(rth, opt->ip6po_rthdr, rthlen);
2448 			} else
2449 				opt->ip6po_rthdr = rth;
2450 
2451 			break;
2452 		}
2453 
2454 		default:
2455 			return(ENOPROTOOPT);
2456 		}
2457 	}
2458 
2459 	return(0);
2460 }
2461 
2462 /*
2463  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2464  * packet to the input queue of a specified interface.  Note that this
2465  * calls the output routine of the loopback "driver", but with an interface
2466  * pointer that might NOT be &loif -- easier than replicating that code here.
2467  */
2468 void
2469 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2470 {
2471 	struct mbuf *copym;
2472 	struct ip6_hdr *ip6;
2473 
2474 	copym = m_copy(m, 0, M_COPYALL);
2475 	if (copym == NULL)
2476 		return;
2477 
2478 	/*
2479 	 * Make sure to deep-copy IPv6 header portion in case the data
2480 	 * is in an mbuf cluster, so that we can safely override the IPv6
2481 	 * header portion later.
2482 	 */
2483 	if ((copym->m_flags & M_EXT) != 0 ||
2484 	    copym->m_len < sizeof(struct ip6_hdr)) {
2485 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
2486 		if (copym == NULL)
2487 			return;
2488 	}
2489 
2490 #ifdef DIAGNOSTIC
2491 	if (copym->m_len < sizeof(*ip6)) {
2492 		m_freem(copym);
2493 		return;
2494 	}
2495 #endif
2496 
2497 	ip6 = mtod(copym, struct ip6_hdr *);
2498 #ifndef SCOPEDROUTING
2499 	/*
2500 	 * clear embedded scope identifiers if necessary.
2501 	 * in6_clearscope will touch the addresses only when necessary.
2502 	 */
2503 	in6_clearscope(&ip6->ip6_src);
2504 	in6_clearscope(&ip6->ip6_dst);
2505 #endif
2506 
2507 	(void)if_simloop(ifp, copym, dst->sin6_family, NULL);
2508 }
2509 
2510 /*
2511  * Chop IPv6 header off from the payload.
2512  */
2513 static int
2514 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
2515 {
2516 	struct mbuf *mh;
2517 	struct ip6_hdr *ip6;
2518 
2519 	ip6 = mtod(m, struct ip6_hdr *);
2520 	if (m->m_len > sizeof(*ip6)) {
2521 		MGETHDR(mh, MB_DONTWAIT, MT_HEADER);
2522 		if (mh == 0) {
2523 			m_freem(m);
2524 			return ENOBUFS;
2525 		}
2526 		M_MOVE_PKTHDR(mh, m);
2527 		MH_ALIGN(mh, sizeof(*ip6));
2528 		m->m_len -= sizeof(*ip6);
2529 		m->m_data += sizeof(*ip6);
2530 		mh->m_next = m;
2531 		m = mh;
2532 		m->m_len = sizeof(*ip6);
2533 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2534 	}
2535 	exthdrs->ip6e_ip6 = m;
2536 	return 0;
2537 }
2538 
2539 /*
2540  * Compute IPv6 extension header length.
2541  */
2542 int
2543 ip6_optlen(struct in6pcb *in6p)
2544 {
2545 	int len;
2546 
2547 	if (!in6p->in6p_outputopts)
2548 		return 0;
2549 
2550 	len = 0;
2551 #define elen(x) \
2552     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2553 
2554 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
2555 	if (in6p->in6p_outputopts->ip6po_rthdr)
2556 		/* dest1 is valid with rthdr only */
2557 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
2558 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
2559 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
2560 	return len;
2561 #undef elen
2562 }
2563