xref: /dragonfly/sys/netinet6/ip6_output.c (revision 1de703da)
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 1.2 2003/06/17 04:28:52 dillon 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 
88 #include <netinet/in.h>
89 #include <netinet/in_var.h>
90 #include <netinet6/in6_var.h>
91 #include <netinet/ip6.h>
92 #include <netinet/icmp6.h>
93 #include <netinet6/ip6_var.h>
94 #include <netinet/in_pcb.h>
95 #include <netinet6/nd6.h>
96 
97 #ifdef IPSEC
98 #include <netinet6/ipsec.h>
99 #ifdef INET6
100 #include <netinet6/ipsec6.h>
101 #endif
102 #include <netkey/key.h>
103 #endif /* IPSEC */
104 
105 #ifdef FAST_IPSEC
106 #include <netipsec/ipsec.h>
107 #include <netipsec/ipsec6.h>
108 #include <netipsec/key.h>
109 #endif /* FAST_IPSEC */
110 
111 #include <netinet6/ip6_fw.h>
112 
113 #include <net/net_osdep.h>
114 
115 extern int (*fr_checkp) __P((struct ip *, int, struct ifnet *, int, struct mbuf **));
116 
117 static MALLOC_DEFINE(M_IPMOPTS, "ip6_moptions", "internet multicast options");
118 
119 struct ip6_exthdrs {
120 	struct mbuf *ip6e_ip6;
121 	struct mbuf *ip6e_hbh;
122 	struct mbuf *ip6e_dest1;
123 	struct mbuf *ip6e_rthdr;
124 	struct mbuf *ip6e_dest2;
125 };
126 
127 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
128 			    struct socket *, struct sockopt *sopt));
129 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *));
130 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **));
131 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int));
132 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
133 				  struct ip6_frag **));
134 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t));
135 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *));
136 
137 /*
138  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
139  * header (with pri, len, nxt, hlim, src, dst).
140  * This function may modify ver and hlim only.
141  * The mbuf chain containing the packet will be freed.
142  * The mbuf opt, if present, will not be freed.
143  *
144  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
145  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
146  * which is rt_rmx.rmx_mtu.
147  */
148 int
149 ip6_output(m0, opt, ro, flags, im6o, ifpp, inp)
150 	struct mbuf *m0;
151 	struct ip6_pktopts *opt;
152 	struct route_in6 *ro;
153 	int flags;
154 	struct ip6_moptions *im6o;
155 	struct ifnet **ifpp;		/* XXX: just for statistics */
156 	struct inpcb *inp;
157 {
158 	struct ip6_hdr *ip6, *mhip6;
159 	struct ifnet *ifp, *origifp;
160 	struct mbuf *m = m0;
161 	int hlen, tlen, len, off;
162 	struct route_in6 ip6route;
163 	struct sockaddr_in6 *dst;
164 	int error = 0;
165 	struct in6_ifaddr *ia = NULL;
166 	u_long mtu;
167 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
168 	struct ip6_exthdrs exthdrs;
169 	struct in6_addr finaldst;
170 	struct route_in6 *ro_pmtu = NULL;
171 	int hdrsplit = 0;
172 	int needipsec = 0;
173 #ifdef IPSEC
174 	int needipsectun = 0;
175 	struct secpolicy *sp = NULL;
176 	struct socket *so = inp ? inp->inp_socket : NULL;
177 
178 	ip6 = mtod(m, struct ip6_hdr *);
179 #endif /* IPSEC */
180 #ifdef FAST_IPSEC
181 	int needipsectun = 0;
182 	struct secpolicy *sp = NULL;
183 
184 	ip6 = mtod(m, struct ip6_hdr *);
185 #endif /* FAST_IPSEC */
186 
187 #define MAKE_EXTHDR(hp, mp)						\
188     do {								\
189 	if (hp) {							\
190 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
191 		error = ip6_copyexthdr((mp), (caddr_t)(hp), 		\
192 				       ((eh)->ip6e_len + 1) << 3);	\
193 		if (error)						\
194 			goto freehdrs;					\
195 	}								\
196     } while (0)
197 
198 	bzero(&exthdrs, sizeof(exthdrs));
199 
200 	if (opt) {
201 		/* Hop-by-Hop options header */
202 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
203 		/* Destination options header(1st part) */
204 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
205 		/* Routing header */
206 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
207 		/* Destination options header(2nd part) */
208 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
209 	}
210 
211 #ifdef IPSEC
212 	/* get a security policy for this packet */
213 	if (so == NULL)
214 		sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
215 	else
216 		sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
217 
218 	if (sp == NULL) {
219 		ipsec6stat.out_inval++;
220 		goto freehdrs;
221 	}
222 
223 	error = 0;
224 
225 	/* check policy */
226 	switch (sp->policy) {
227 	case IPSEC_POLICY_DISCARD:
228 		/*
229 		 * This packet is just discarded.
230 		 */
231 		ipsec6stat.out_polvio++;
232 		goto freehdrs;
233 
234 	case IPSEC_POLICY_BYPASS:
235 	case IPSEC_POLICY_NONE:
236 		/* no need to do IPsec. */
237 		needipsec = 0;
238 		break;
239 
240 	case IPSEC_POLICY_IPSEC:
241 		if (sp->req == NULL) {
242 			/* acquire a policy */
243 			error = key_spdacquire(sp);
244 			goto freehdrs;
245 		}
246 		needipsec = 1;
247 		break;
248 
249 	case IPSEC_POLICY_ENTRUST:
250 	default:
251 		printf("ip6_output: Invalid policy found. %d\n", sp->policy);
252 	}
253 #endif /* IPSEC */
254 #ifdef FAST_IPSEC
255 	/* get a security policy for this packet */
256 	if (inp == NULL)
257 		sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
258 	else
259 		sp = ipsec_getpolicybysock(m, IPSEC_DIR_OUTBOUND, inp, &error);
260 
261 	if (sp == NULL) {
262 		newipsecstat.ips_out_inval++;
263 		goto freehdrs;
264 	}
265 
266 	error = 0;
267 
268 	/* check policy */
269 	switch (sp->policy) {
270 	case IPSEC_POLICY_DISCARD:
271 		/*
272 		 * This packet is just discarded.
273 		 */
274 		newipsecstat.ips_out_polvio++;
275 		goto freehdrs;
276 
277 	case IPSEC_POLICY_BYPASS:
278 	case IPSEC_POLICY_NONE:
279 		/* no need to do IPsec. */
280 		needipsec = 0;
281 		break;
282 
283 	case IPSEC_POLICY_IPSEC:
284 		if (sp->req == NULL) {
285 			/* acquire a policy */
286 			error = key_spdacquire(sp);
287 			goto freehdrs;
288 		}
289 		needipsec = 1;
290 		break;
291 
292 	case IPSEC_POLICY_ENTRUST:
293 	default:
294 		printf("ip6_output: Invalid policy found. %d\n", sp->policy);
295 	}
296 #endif /* FAST_IPSEC */
297 
298 	/*
299 	 * Calculate the total length of the extension header chain.
300 	 * Keep the length of the unfragmentable part for fragmentation.
301 	 */
302 	optlen = 0;
303 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
304 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
305 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
306 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
307 	/* NOTE: we don't add AH/ESP length here. do that later. */
308 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
309 
310 	/*
311 	 * If we need IPsec, or there is at least one extension header,
312 	 * separate IP6 header from the payload.
313 	 */
314 	if ((needipsec || optlen) && !hdrsplit) {
315 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
316 			m = NULL;
317 			goto freehdrs;
318 		}
319 		m = exthdrs.ip6e_ip6;
320 		hdrsplit++;
321 	}
322 
323 	/* adjust pointer */
324 	ip6 = mtod(m, struct ip6_hdr *);
325 
326 	/* adjust mbuf packet header length */
327 	m->m_pkthdr.len += optlen;
328 	plen = m->m_pkthdr.len - sizeof(*ip6);
329 
330 	/* If this is a jumbo payload, insert a jumbo payload option. */
331 	if (plen > IPV6_MAXPACKET) {
332 		if (!hdrsplit) {
333 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
334 				m = NULL;
335 				goto freehdrs;
336 			}
337 			m = exthdrs.ip6e_ip6;
338 			hdrsplit++;
339 		}
340 		/* adjust pointer */
341 		ip6 = mtod(m, struct ip6_hdr *);
342 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
343 			goto freehdrs;
344 		ip6->ip6_plen = 0;
345 	} else
346 		ip6->ip6_plen = htons(plen);
347 
348 	/*
349 	 * Concatenate headers and fill in next header fields.
350 	 * Here we have, on "m"
351 	 *	IPv6 payload
352 	 * and we insert headers accordingly.  Finally, we should be getting:
353 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
354 	 *
355 	 * during the header composing process, "m" points to IPv6 header.
356 	 * "mprev" points to an extension header prior to esp.
357 	 */
358 	{
359 		u_char *nexthdrp = &ip6->ip6_nxt;
360 		struct mbuf *mprev = m;
361 
362 		/*
363 		 * we treat dest2 specially.  this makes IPsec processing
364 		 * much easier.  the goal here is to make mprev point the
365 		 * mbuf prior to dest2.
366 		 *
367 		 * result: IPv6 dest2 payload
368 		 * m and mprev will point to IPv6 header.
369 		 */
370 		if (exthdrs.ip6e_dest2) {
371 			if (!hdrsplit)
372 				panic("assumption failed: hdr not split");
373 			exthdrs.ip6e_dest2->m_next = m->m_next;
374 			m->m_next = exthdrs.ip6e_dest2;
375 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
376 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
377 		}
378 
379 #define MAKE_CHAIN(m, mp, p, i)\
380     do {\
381 	if (m) {\
382 		if (!hdrsplit) \
383 			panic("assumption failed: hdr not split"); \
384 		*mtod((m), u_char *) = *(p);\
385 		*(p) = (i);\
386 		p = mtod((m), u_char *);\
387 		(m)->m_next = (mp)->m_next;\
388 		(mp)->m_next = (m);\
389 		(mp) = (m);\
390 	}\
391     } while (0)
392 		/*
393 		 * result: IPv6 hbh dest1 rthdr dest2 payload
394 		 * m will point to IPv6 header.  mprev will point to the
395 		 * extension header prior to dest2 (rthdr in the above case).
396 		 */
397 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev,
398 			   nexthdrp, IPPROTO_HOPOPTS);
399 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev,
400 			   nexthdrp, IPPROTO_DSTOPTS);
401 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev,
402 			   nexthdrp, IPPROTO_ROUTING);
403 
404 #if defined(IPSEC) || defined(FAST_IPSEC)
405 		if (!needipsec)
406 			goto skip_ipsec2;
407 
408 		/*
409 		 * pointers after IPsec headers are not valid any more.
410 		 * other pointers need a great care too.
411 		 * (IPsec routines should not mangle mbufs prior to AH/ESP)
412 		 */
413 		exthdrs.ip6e_dest2 = NULL;
414 
415 	    {
416 		struct ip6_rthdr *rh = NULL;
417 		int segleft_org = 0;
418 		struct ipsec_output_state state;
419 
420 		if (exthdrs.ip6e_rthdr) {
421 			rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
422 			segleft_org = rh->ip6r_segleft;
423 			rh->ip6r_segleft = 0;
424 		}
425 
426 		bzero(&state, sizeof(state));
427 		state.m = m;
428 		error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
429 			&needipsectun);
430 		m = state.m;
431 		if (error) {
432 			/* mbuf is already reclaimed in ipsec6_output_trans. */
433 			m = NULL;
434 			switch (error) {
435 			case EHOSTUNREACH:
436 			case ENETUNREACH:
437 			case EMSGSIZE:
438 			case ENOBUFS:
439 			case ENOMEM:
440 				break;
441 			default:
442 				printf("ip6_output (ipsec): error code %d\n", error);
443 				/* fall through */
444 			case ENOENT:
445 				/* don't show these error codes to the user */
446 				error = 0;
447 				break;
448 			}
449 			goto bad;
450 		}
451 		if (exthdrs.ip6e_rthdr) {
452 			/* ah6_output doesn't modify mbuf chain */
453 			rh->ip6r_segleft = segleft_org;
454 		}
455 	    }
456 skip_ipsec2:;
457 #endif
458 	}
459 
460 	/*
461 	 * If there is a routing header, replace destination address field
462 	 * with the first hop of the routing header.
463 	 */
464 	if (exthdrs.ip6e_rthdr) {
465 		struct ip6_rthdr *rh =
466 			(struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
467 						  struct ip6_rthdr *));
468 		struct ip6_rthdr0 *rh0;
469 
470 		finaldst = ip6->ip6_dst;
471 		switch (rh->ip6r_type) {
472 		case IPV6_RTHDR_TYPE_0:
473 			 rh0 = (struct ip6_rthdr0 *)rh;
474 			 ip6->ip6_dst = rh0->ip6r0_addr[0];
475 			 bcopy((caddr_t)&rh0->ip6r0_addr[1],
476 			       (caddr_t)&rh0->ip6r0_addr[0],
477 			       sizeof(struct in6_addr)*(rh0->ip6r0_segleft - 1)
478 				 );
479 			 rh0->ip6r0_addr[rh0->ip6r0_segleft - 1] = finaldst;
480 			 break;
481 		default:	/* is it possible? */
482 			 error = EINVAL;
483 			 goto bad;
484 		}
485 	}
486 
487 	/* Source address validation */
488 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
489 	    (flags & IPV6_DADOUTPUT) == 0) {
490 		error = EOPNOTSUPP;
491 		ip6stat.ip6s_badscope++;
492 		goto bad;
493 	}
494 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
495 		error = EOPNOTSUPP;
496 		ip6stat.ip6s_badscope++;
497 		goto bad;
498 	}
499 
500 	ip6stat.ip6s_localout++;
501 
502 	/*
503 	 * Route packet.
504 	 */
505 	if (ro == 0) {
506 		ro = &ip6route;
507 		bzero((caddr_t)ro, sizeof(*ro));
508 	}
509 	ro_pmtu = ro;
510 	if (opt && opt->ip6po_rthdr)
511 		ro = &opt->ip6po_route;
512 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
513 	/*
514 	 * If there is a cached route,
515 	 * check that it is to the same destination
516 	 * and is still up. If not, free it and try again.
517 	 */
518 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
519 			 dst->sin6_family != AF_INET6 ||
520 			 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
521 		RTFREE(ro->ro_rt);
522 		ro->ro_rt = (struct rtentry *)0;
523 	}
524 	if (ro->ro_rt == 0) {
525 		bzero(dst, sizeof(*dst));
526 		dst->sin6_family = AF_INET6;
527 		dst->sin6_len = sizeof(struct sockaddr_in6);
528 		dst->sin6_addr = ip6->ip6_dst;
529 #ifdef SCOPEDROUTING
530 		/* XXX: sin6_scope_id should already be fixed at this point */
531 		if (IN6_IS_SCOPE_LINKLOCAL(&dst->sin6_addr))
532 			dst->sin6_scope_id = ntohs(dst->sin6_addr.s6_addr16[1]);
533 #endif
534 	}
535 #if defined(IPSEC) || defined(FAST_IPSEC)
536 	if (needipsec && needipsectun) {
537 		struct ipsec_output_state state;
538 
539 		/*
540 		 * All the extension headers will become inaccessible
541 		 * (since they can be encrypted).
542 		 * Don't panic, we need no more updates to extension headers
543 		 * on inner IPv6 packet (since they are now encapsulated).
544 		 *
545 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
546 		 */
547 		bzero(&exthdrs, sizeof(exthdrs));
548 		exthdrs.ip6e_ip6 = m;
549 
550 		bzero(&state, sizeof(state));
551 		state.m = m;
552 		state.ro = (struct route *)ro;
553 		state.dst = (struct sockaddr *)dst;
554 
555 		error = ipsec6_output_tunnel(&state, sp, flags);
556 
557 		m = state.m;
558 		ro = (struct route_in6 *)state.ro;
559 		dst = (struct sockaddr_in6 *)state.dst;
560 		if (error) {
561 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
562 			m0 = m = NULL;
563 			m = NULL;
564 			switch (error) {
565 			case EHOSTUNREACH:
566 			case ENETUNREACH:
567 			case EMSGSIZE:
568 			case ENOBUFS:
569 			case ENOMEM:
570 				break;
571 			default:
572 				printf("ip6_output (ipsec): error code %d\n", error);
573 				/* fall through */
574 			case ENOENT:
575 				/* don't show these error codes to the user */
576 				error = 0;
577 				break;
578 			}
579 			goto bad;
580 		}
581 
582 		exthdrs.ip6e_ip6 = m;
583 	}
584 #endif /* IPSEC */
585 
586 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
587 		/* Unicast */
588 
589 #define ifatoia6(ifa)	((struct in6_ifaddr *)(ifa))
590 #define sin6tosa(sin6)	((struct sockaddr *)(sin6))
591 		/* xxx
592 		 * interface selection comes here
593 		 * if an interface is specified from an upper layer,
594 		 * ifp must point it.
595 		 */
596 		if (ro->ro_rt == 0) {
597 			/*
598 			 * non-bsdi always clone routes, if parent is
599 			 * PRF_CLONING.
600 			 */
601 			rtalloc((struct route *)ro);
602 		}
603 		if (ro->ro_rt == 0) {
604 			ip6stat.ip6s_noroute++;
605 			error = EHOSTUNREACH;
606 			/* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
607 			goto bad;
608 		}
609 		ia = ifatoia6(ro->ro_rt->rt_ifa);
610 		ifp = ro->ro_rt->rt_ifp;
611 		ro->ro_rt->rt_use++;
612 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
613 			dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
614 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
615 
616 		in6_ifstat_inc(ifp, ifs6_out_request);
617 
618 		/*
619 		 * Check if the outgoing interface conflicts with
620 		 * the interface specified by ifi6_ifindex (if specified).
621 		 * Note that loopback interface is always okay.
622 		 * (this may happen when we are sending a packet to one of
623 		 *  our own addresses.)
624 		 */
625 		if (opt && opt->ip6po_pktinfo
626 		 && opt->ip6po_pktinfo->ipi6_ifindex) {
627 			if (!(ifp->if_flags & IFF_LOOPBACK)
628 			 && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
629 				ip6stat.ip6s_noroute++;
630 				in6_ifstat_inc(ifp, ifs6_out_discard);
631 				error = EHOSTUNREACH;
632 				goto bad;
633 			}
634 		}
635 
636 		if (opt && opt->ip6po_hlim != -1)
637 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
638 	} else {
639 		/* Multicast */
640 		struct	in6_multi *in6m;
641 
642 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
643 
644 		/*
645 		 * See if the caller provided any multicast options
646 		 */
647 		ifp = NULL;
648 		if (im6o != NULL) {
649 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
650 			if (im6o->im6o_multicast_ifp != NULL)
651 				ifp = im6o->im6o_multicast_ifp;
652 		} else
653 			ip6->ip6_hlim = ip6_defmcasthlim;
654 
655 		/*
656 		 * See if the caller provided the outgoing interface
657 		 * as an ancillary data.
658 		 * Boundary check for ifindex is assumed to be already done.
659 		 */
660 		if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
661 			ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex];
662 
663 		/*
664 		 * If the destination is a node-local scope multicast,
665 		 * the packet should be loop-backed only.
666 		 */
667 		if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
668 			/*
669 			 * If the outgoing interface is already specified,
670 			 * it should be a loopback interface.
671 			 */
672 			if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) {
673 				ip6stat.ip6s_badscope++;
674 				error = ENETUNREACH; /* XXX: better error? */
675 				/* XXX correct ifp? */
676 				in6_ifstat_inc(ifp, ifs6_out_discard);
677 				goto bad;
678 			} else {
679 				ifp = &loif[0];
680 			}
681 		}
682 
683 		if (opt && opt->ip6po_hlim != -1)
684 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
685 
686 		/*
687 		 * If caller did not provide an interface lookup a
688 		 * default in the routing table.  This is either a
689 		 * default for the speicfied group (i.e. a host
690 		 * route), or a multicast default (a route for the
691 		 * ``net'' ff00::/8).
692 		 */
693 		if (ifp == NULL) {
694 			if (ro->ro_rt == 0) {
695 				ro->ro_rt = rtalloc1((struct sockaddr *)
696 						&ro->ro_dst, 0, 0UL);
697 			}
698 			if (ro->ro_rt == 0) {
699 				ip6stat.ip6s_noroute++;
700 				error = EHOSTUNREACH;
701 				/* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
702 				goto bad;
703 			}
704 			ia = ifatoia6(ro->ro_rt->rt_ifa);
705 			ifp = ro->ro_rt->rt_ifp;
706 			ro->ro_rt->rt_use++;
707 		}
708 
709 		if ((flags & IPV6_FORWARDING) == 0)
710 			in6_ifstat_inc(ifp, ifs6_out_request);
711 		in6_ifstat_inc(ifp, ifs6_out_mcast);
712 
713 		/*
714 		 * Confirm that the outgoing interface supports multicast.
715 		 */
716 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
717 			ip6stat.ip6s_noroute++;
718 			in6_ifstat_inc(ifp, ifs6_out_discard);
719 			error = ENETUNREACH;
720 			goto bad;
721 		}
722 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
723 		if (in6m != NULL &&
724 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
725 			/*
726 			 * If we belong to the destination multicast group
727 			 * on the outgoing interface, and the caller did not
728 			 * forbid loopback, loop back a copy.
729 			 */
730 			ip6_mloopback(ifp, m, dst);
731 		} else {
732 			/*
733 			 * If we are acting as a multicast router, perform
734 			 * multicast forwarding as if the packet had just
735 			 * arrived on the interface to which we are about
736 			 * to send.  The multicast forwarding function
737 			 * recursively calls this function, using the
738 			 * IPV6_FORWARDING flag to prevent infinite recursion.
739 			 *
740 			 * Multicasts that are looped back by ip6_mloopback(),
741 			 * above, will be forwarded by the ip6_input() routine,
742 			 * if necessary.
743 			 */
744 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
745 				if (ip6_mforward(ip6, ifp, m) != 0) {
746 					m_freem(m);
747 					goto done;
748 				}
749 			}
750 		}
751 		/*
752 		 * Multicasts with a hoplimit of zero may be looped back,
753 		 * above, but must not be transmitted on a network.
754 		 * Also, multicasts addressed to the loopback interface
755 		 * are not sent -- the above call to ip6_mloopback() will
756 		 * loop back a copy if this host actually belongs to the
757 		 * destination group on the loopback interface.
758 		 */
759 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
760 			m_freem(m);
761 			goto done;
762 		}
763 	}
764 
765 	/*
766 	 * Fill the outgoing inteface to tell the upper layer
767 	 * to increment per-interface statistics.
768 	 */
769 	if (ifpp)
770 		*ifpp = ifp;
771 
772 	/*
773 	 * Determine path MTU.
774 	 */
775 	if (ro_pmtu != ro) {
776 		/* The first hop and the final destination may differ. */
777 		struct sockaddr_in6 *sin6_fin =
778 			(struct sockaddr_in6 *)&ro_pmtu->ro_dst;
779 		if (ro_pmtu->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
780 				       !IN6_ARE_ADDR_EQUAL(&sin6_fin->sin6_addr,
781 							   &finaldst))) {
782 			RTFREE(ro_pmtu->ro_rt);
783 			ro_pmtu->ro_rt = (struct rtentry *)0;
784 		}
785 		if (ro_pmtu->ro_rt == 0) {
786 			bzero(sin6_fin, sizeof(*sin6_fin));
787 			sin6_fin->sin6_family = AF_INET6;
788 			sin6_fin->sin6_len = sizeof(struct sockaddr_in6);
789 			sin6_fin->sin6_addr = finaldst;
790 
791 			rtalloc((struct route *)ro_pmtu);
792 		}
793 	}
794 	if (ro_pmtu->ro_rt != NULL) {
795 		u_int32_t ifmtu = nd_ifinfo[ifp->if_index].linkmtu;
796 
797 		mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
798 		if (mtu > ifmtu || mtu == 0) {
799 			/*
800 			 * The MTU on the route is larger than the MTU on
801 			 * the interface!  This shouldn't happen, unless the
802 			 * MTU of the interface has been changed after the
803 			 * interface was brought up.  Change the MTU in the
804 			 * route to match the interface MTU (as long as the
805 			 * field isn't locked).
806 			 *
807 			 * if MTU on the route is 0, we need to fix the MTU.
808 			 * this case happens with path MTU discovery timeouts.
809 			 */
810 			 mtu = ifmtu;
811 			 if ((ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
812 				 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; /* XXX */
813 		}
814 	} else {
815 		mtu = nd_ifinfo[ifp->if_index].linkmtu;
816 	}
817 
818 	/*
819 	 * advanced API (IPV6_USE_MIN_MTU) overrides mtu setting
820 	 */
821 	if ((flags & IPV6_MINMTU) != 0 && mtu > IPV6_MMTU)
822 		mtu = IPV6_MMTU;
823 
824 	/* Fake scoped addresses */
825 	if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
826 		/*
827 		 * If source or destination address is a scoped address, and
828 		 * the packet is going to be sent to a loopback interface,
829 		 * we should keep the original interface.
830 		 */
831 
832 		/*
833 		 * XXX: this is a very experimental and temporary solution.
834 		 * We eventually have sockaddr_in6 and use the sin6_scope_id
835 		 * field of the structure here.
836 		 * We rely on the consistency between two scope zone ids
837 		 * of source and destination, which should already be assured.
838 		 * Larger scopes than link will be supported in the future.
839 		 */
840 		origifp = NULL;
841 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
842 			origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
843 		else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
844 			origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
845 		/*
846 		 * XXX: origifp can be NULL even in those two cases above.
847 		 * For example, if we remove the (only) link-local address
848 		 * from the loopback interface, and try to send a link-local
849 		 * address without link-id information.  Then the source
850 		 * address is ::1, and the destination address is the
851 		 * link-local address with its s6_addr16[1] being zero.
852 		 * What is worse, if the packet goes to the loopback interface
853 		 * by a default rejected route, the null pointer would be
854 		 * passed to looutput, and the kernel would hang.
855 		 * The following last resort would prevent such disaster.
856 		 */
857 		if (origifp == NULL)
858 			origifp = ifp;
859 	}
860 	else
861 		origifp = ifp;
862 #ifndef SCOPEDROUTING
863 	/*
864 	 * clear embedded scope identifiers if necessary.
865 	 * in6_clearscope will touch the addresses only when necessary.
866 	 */
867 	in6_clearscope(&ip6->ip6_src);
868 	in6_clearscope(&ip6->ip6_dst);
869 #endif
870 
871 	/*
872 	 * Check with the firewall...
873 	 */
874 	if (ip6_fw_enable && ip6_fw_chk_ptr) {
875 		u_short port = 0;
876 		m->m_pkthdr.rcvif = NULL;	/* XXX */
877 		/* If ipfw says divert, we have to just drop packet */
878 		if ((*ip6_fw_chk_ptr)(&ip6, ifp, &port, &m)) {
879 			m_freem(m);
880 			goto done;
881 		}
882 		if (!m) {
883 			error = EACCES;
884 			goto done;
885 		}
886 	}
887 
888 	/*
889 	 * If the outgoing packet contains a hop-by-hop options header,
890 	 * it must be examined and processed even by the source node.
891 	 * (RFC 2460, section 4.)
892 	 */
893 	if (exthdrs.ip6e_hbh) {
894 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
895 		u_int32_t dummy1; /* XXX unused */
896 		u_int32_t dummy2; /* XXX unused */
897 
898 #ifdef DIAGNOSTIC
899 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
900 			panic("ip6e_hbh is not continuous");
901 #endif
902 		/*
903 		 *  XXX: if we have to send an ICMPv6 error to the sender,
904 		 *       we need the M_LOOP flag since icmp6_error() expects
905 		 *       the IPv6 and the hop-by-hop options header are
906 		 *       continuous unless the flag is set.
907 		 */
908 		m->m_flags |= M_LOOP;
909 		m->m_pkthdr.rcvif = ifp;
910 		if (ip6_process_hopopts(m,
911 					(u_int8_t *)(hbh + 1),
912 					((hbh->ip6h_len + 1) << 3) -
913 					sizeof(struct ip6_hbh),
914 					&dummy1, &dummy2) < 0) {
915 			/* m was already freed at this point */
916 			error = EINVAL;/* better error? */
917 			goto done;
918 		}
919 		m->m_flags &= ~M_LOOP; /* XXX */
920 		m->m_pkthdr.rcvif = NULL;
921 	}
922 
923 	/*
924 	 * Check if we want to allow this packet to be processed.
925 	 * Consider it to be bad if not.
926 	 */
927 	if (fr_checkp) {
928 		struct	mbuf	*m1 = m;
929 
930 		if ((*fr_checkp)((struct ip *)ip6, sizeof(*ip6), ifp, 1, &m1))
931 			goto done;
932 		m = m1;
933 		if (m == NULL)
934 			goto done;
935 		ip6 = mtod(m, struct ip6_hdr *);
936 	}
937 
938 	/*
939 	 * Send the packet to the outgoing interface.
940 	 * If necessary, do IPv6 fragmentation before sending.
941 	 */
942 	tlen = m->m_pkthdr.len;
943 	if (tlen <= mtu
944 #ifdef notyet
945 	    /*
946 	     * On any link that cannot convey a 1280-octet packet in one piece,
947 	     * link-specific fragmentation and reassembly must be provided at
948 	     * a layer below IPv6. [RFC 2460, sec.5]
949 	     * Thus if the interface has ability of link-level fragmentation,
950 	     * we can just send the packet even if the packet size is
951 	     * larger than the link's MTU.
952 	     * XXX: IFF_FRAGMENTABLE (or such) flag has not been defined yet...
953 	     */
954 
955 	    || ifp->if_flags & IFF_FRAGMENTABLE
956 #endif
957 	    )
958 	{
959  		/* Record statistics for this interface address. */
960  		if (ia && !(flags & IPV6_FORWARDING)) {
961  			ia->ia_ifa.if_opackets++;
962  			ia->ia_ifa.if_obytes += m->m_pkthdr.len;
963  		}
964 #ifdef IPSEC
965 		/* clean ipsec history once it goes out of the node */
966 		ipsec_delaux(m);
967 #endif
968 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
969 		goto done;
970 	} else if (mtu < IPV6_MMTU) {
971 		/*
972 		 * note that path MTU is never less than IPV6_MMTU
973 		 * (see icmp6_input).
974 		 */
975 		error = EMSGSIZE;
976 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
977 		goto bad;
978 	} else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
979 		error = EMSGSIZE;
980 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
981 		goto bad;
982 	} else {
983 		struct mbuf **mnext, *m_frgpart;
984 		struct ip6_frag *ip6f;
985 		u_int32_t id = htonl(ip6_id++);
986 		u_char nextproto;
987 
988 		/*
989 		 * Too large for the destination or interface;
990 		 * fragment if possible.
991 		 * Must be able to put at least 8 bytes per fragment.
992 		 */
993 		hlen = unfragpartlen;
994 		if (mtu > IPV6_MAXPACKET)
995 			mtu = IPV6_MAXPACKET;
996 
997 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
998 		if (len < 8) {
999 			error = EMSGSIZE;
1000 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
1001 			goto bad;
1002 		}
1003 
1004 		mnext = &m->m_nextpkt;
1005 
1006 		/*
1007 		 * Change the next header field of the last header in the
1008 		 * unfragmentable part.
1009 		 */
1010 		if (exthdrs.ip6e_rthdr) {
1011 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1012 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1013 		} else if (exthdrs.ip6e_dest1) {
1014 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1015 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1016 		} else if (exthdrs.ip6e_hbh) {
1017 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1018 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1019 		} else {
1020 			nextproto = ip6->ip6_nxt;
1021 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1022 		}
1023 
1024 		/*
1025 		 * Loop through length of segment after first fragment,
1026 		 * make new header and copy data of each part and link onto
1027 		 * chain.
1028 		 */
1029 		m0 = m;
1030 		for (off = hlen; off < tlen; off += len) {
1031 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
1032 			if (!m) {
1033 				error = ENOBUFS;
1034 				ip6stat.ip6s_odropped++;
1035 				goto sendorfree;
1036 			}
1037 			m->m_pkthdr.rcvif = NULL;
1038 			m->m_flags = m0->m_flags & M_COPYFLAGS;
1039 			*mnext = m;
1040 			mnext = &m->m_nextpkt;
1041 			m->m_data += max_linkhdr;
1042 			mhip6 = mtod(m, struct ip6_hdr *);
1043 			*mhip6 = *ip6;
1044 			m->m_len = sizeof(*mhip6);
1045  			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1046  			if (error) {
1047 				ip6stat.ip6s_odropped++;
1048 				goto sendorfree;
1049 			}
1050 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1051 			if (off + len >= tlen)
1052 				len = tlen - off;
1053 			else
1054 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1055 			mhip6->ip6_plen = htons((u_short)(len + hlen +
1056 							  sizeof(*ip6f) -
1057 							  sizeof(struct ip6_hdr)));
1058 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1059 				error = ENOBUFS;
1060 				ip6stat.ip6s_odropped++;
1061 				goto sendorfree;
1062 			}
1063 			m_cat(m, m_frgpart);
1064 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1065 			m->m_pkthdr.rcvif = (struct ifnet *)0;
1066 			ip6f->ip6f_reserved = 0;
1067 			ip6f->ip6f_ident = id;
1068 			ip6f->ip6f_nxt = nextproto;
1069 			ip6stat.ip6s_ofragments++;
1070 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1071 		}
1072 
1073 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1074 	}
1075 
1076 	/*
1077 	 * Remove leading garbages.
1078 	 */
1079 sendorfree:
1080 	m = m0->m_nextpkt;
1081 	m0->m_nextpkt = 0;
1082 	m_freem(m0);
1083 	for (m0 = m; m; m = m0) {
1084 		m0 = m->m_nextpkt;
1085 		m->m_nextpkt = 0;
1086 		if (error == 0) {
1087  			/* Record statistics for this interface address. */
1088  			if (ia) {
1089  				ia->ia_ifa.if_opackets++;
1090  				ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1091  			}
1092 #ifdef IPSEC
1093 			/* clean ipsec history once it goes out of the node */
1094 			ipsec_delaux(m);
1095 #endif
1096 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1097 		} else
1098 			m_freem(m);
1099 	}
1100 
1101 	if (error == 0)
1102 		ip6stat.ip6s_fragmented++;
1103 
1104 done:
1105 	if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1106 		RTFREE(ro->ro_rt);
1107 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1108 		RTFREE(ro_pmtu->ro_rt);
1109 	}
1110 
1111 #ifdef IPSEC
1112 	if (sp != NULL)
1113 		key_freesp(sp);
1114 #endif /* IPSEC */
1115 #ifdef FAST_IPSEC
1116 	if (sp != NULL)
1117 		KEY_FREESP(&sp);
1118 #endif /* FAST_IPSEC */
1119 
1120 	return(error);
1121 
1122 freehdrs:
1123 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1124 	m_freem(exthdrs.ip6e_dest1);
1125 	m_freem(exthdrs.ip6e_rthdr);
1126 	m_freem(exthdrs.ip6e_dest2);
1127 	/* fall through */
1128 bad:
1129 	m_freem(m);
1130 	goto done;
1131 }
1132 
1133 static int
1134 ip6_copyexthdr(mp, hdr, hlen)
1135 	struct mbuf **mp;
1136 	caddr_t hdr;
1137 	int hlen;
1138 {
1139 	struct mbuf *m;
1140 
1141 	if (hlen > MCLBYTES)
1142 		return(ENOBUFS); /* XXX */
1143 
1144 	MGET(m, M_DONTWAIT, MT_DATA);
1145 	if (!m)
1146 		return(ENOBUFS);
1147 
1148 	if (hlen > MLEN) {
1149 		MCLGET(m, M_DONTWAIT);
1150 		if ((m->m_flags & M_EXT) == 0) {
1151 			m_free(m);
1152 			return(ENOBUFS);
1153 		}
1154 	}
1155 	m->m_len = hlen;
1156 	if (hdr)
1157 		bcopy(hdr, mtod(m, caddr_t), hlen);
1158 
1159 	*mp = m;
1160 	return(0);
1161 }
1162 
1163 /*
1164  * Insert jumbo payload option.
1165  */
1166 static int
1167 ip6_insert_jumboopt(exthdrs, plen)
1168 	struct ip6_exthdrs *exthdrs;
1169 	u_int32_t plen;
1170 {
1171 	struct mbuf *mopt;
1172 	u_char *optbuf;
1173 	u_int32_t v;
1174 
1175 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1176 
1177 	/*
1178 	 * If there is no hop-by-hop options header, allocate new one.
1179 	 * If there is one but it doesn't have enough space to store the
1180 	 * jumbo payload option, allocate a cluster to store the whole options.
1181 	 * Otherwise, use it to store the options.
1182 	 */
1183 	if (exthdrs->ip6e_hbh == 0) {
1184 		MGET(mopt, M_DONTWAIT, MT_DATA);
1185 		if (mopt == 0)
1186 			return(ENOBUFS);
1187 		mopt->m_len = JUMBOOPTLEN;
1188 		optbuf = mtod(mopt, u_char *);
1189 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1190 		exthdrs->ip6e_hbh = mopt;
1191 	} else {
1192 		struct ip6_hbh *hbh;
1193 
1194 		mopt = exthdrs->ip6e_hbh;
1195 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1196 			/*
1197 			 * XXX assumption:
1198 			 * - exthdrs->ip6e_hbh is not referenced from places
1199 			 *   other than exthdrs.
1200 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1201 			 */
1202 			int oldoptlen = mopt->m_len;
1203 			struct mbuf *n;
1204 
1205 			/*
1206 			 * XXX: give up if the whole (new) hbh header does
1207 			 * not fit even in an mbuf cluster.
1208 			 */
1209 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1210 				return(ENOBUFS);
1211 
1212 			/*
1213 			 * As a consequence, we must always prepare a cluster
1214 			 * at this point.
1215 			 */
1216 			MGET(n, M_DONTWAIT, MT_DATA);
1217 			if (n) {
1218 				MCLGET(n, M_DONTWAIT);
1219 				if ((n->m_flags & M_EXT) == 0) {
1220 					m_freem(n);
1221 					n = NULL;
1222 				}
1223 			}
1224 			if (!n)
1225 				return(ENOBUFS);
1226 			n->m_len = oldoptlen + JUMBOOPTLEN;
1227 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1228 			      oldoptlen);
1229 			optbuf = mtod(n, caddr_t) + oldoptlen;
1230 			m_freem(mopt);
1231 			mopt = exthdrs->ip6e_hbh = n;
1232 		} else {
1233 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1234 			mopt->m_len += JUMBOOPTLEN;
1235 		}
1236 		optbuf[0] = IP6OPT_PADN;
1237 		optbuf[1] = 1;
1238 
1239 		/*
1240 		 * Adjust the header length according to the pad and
1241 		 * the jumbo payload option.
1242 		 */
1243 		hbh = mtod(mopt, struct ip6_hbh *);
1244 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1245 	}
1246 
1247 	/* fill in the option. */
1248 	optbuf[2] = IP6OPT_JUMBO;
1249 	optbuf[3] = 4;
1250 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1251 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1252 
1253 	/* finally, adjust the packet header length */
1254 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1255 
1256 	return(0);
1257 #undef JUMBOOPTLEN
1258 }
1259 
1260 /*
1261  * Insert fragment header and copy unfragmentable header portions.
1262  */
1263 static int
1264 ip6_insertfraghdr(m0, m, hlen, frghdrp)
1265 	struct mbuf *m0, *m;
1266 	int hlen;
1267 	struct ip6_frag **frghdrp;
1268 {
1269 	struct mbuf *n, *mlast;
1270 
1271 	if (hlen > sizeof(struct ip6_hdr)) {
1272 		n = m_copym(m0, sizeof(struct ip6_hdr),
1273 			    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1274 		if (n == 0)
1275 			return(ENOBUFS);
1276 		m->m_next = n;
1277 	} else
1278 		n = m;
1279 
1280 	/* Search for the last mbuf of unfragmentable part. */
1281 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1282 		;
1283 
1284 	if ((mlast->m_flags & M_EXT) == 0 &&
1285 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1286 		/* use the trailing space of the last mbuf for the fragment hdr */
1287 		*frghdrp =
1288 			(struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len);
1289 		mlast->m_len += sizeof(struct ip6_frag);
1290 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1291 	} else {
1292 		/* allocate a new mbuf for the fragment header */
1293 		struct mbuf *mfrg;
1294 
1295 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1296 		if (mfrg == 0)
1297 			return(ENOBUFS);
1298 		mfrg->m_len = sizeof(struct ip6_frag);
1299 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1300 		mlast->m_next = mfrg;
1301 	}
1302 
1303 	return(0);
1304 }
1305 
1306 /*
1307  * IP6 socket option processing.
1308  */
1309 int
1310 ip6_ctloutput(so, sopt)
1311 	struct socket *so;
1312 	struct sockopt *sopt;
1313 {
1314 	int privileged;
1315 	struct inpcb *in6p = sotoinpcb(so);
1316 	int error, optval;
1317 	int level, op, optname;
1318 	int optlen;
1319 	struct proc *p;
1320 
1321 	if (sopt) {
1322 		level = sopt->sopt_level;
1323 		op = sopt->sopt_dir;
1324 		optname = sopt->sopt_name;
1325 		optlen = sopt->sopt_valsize;
1326 		p = sopt->sopt_p;
1327 	} else {
1328 		panic("ip6_ctloutput: arg soopt is NULL");
1329 	}
1330 	error = optval = 0;
1331 
1332 	privileged = (p == 0 || suser(p)) ? 0 : 1;
1333 
1334 	if (level == IPPROTO_IPV6) {
1335 		switch (op) {
1336 
1337 		case SOPT_SET:
1338 			switch (optname) {
1339 			case IPV6_PKTOPTIONS:
1340 			{
1341 				struct mbuf *m;
1342 
1343 				error = soopt_getm(sopt, &m); /* XXX */
1344 				if (error != NULL)
1345 					break;
1346 				error = soopt_mcopyin(sopt, m); /* XXX */
1347 				if (error != NULL)
1348 					break;
1349 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1350 						    m, so, sopt);
1351 				m_freem(m); /* XXX */
1352 				break;
1353 			}
1354 
1355 			/*
1356 			 * Use of some Hop-by-Hop options or some
1357 			 * Destination options, might require special
1358 			 * privilege.  That is, normal applications
1359 			 * (without special privilege) might be forbidden
1360 			 * from setting certain options in outgoing packets,
1361 			 * and might never see certain options in received
1362 			 * packets. [RFC 2292 Section 6]
1363 			 * KAME specific note:
1364 			 *  KAME prevents non-privileged users from sending or
1365 			 *  receiving ANY hbh/dst options in order to avoid
1366 			 *  overhead of parsing options in the kernel.
1367 			 */
1368 			case IPV6_UNICAST_HOPS:
1369 			case IPV6_CHECKSUM:
1370 			case IPV6_FAITH:
1371 
1372 			case IPV6_V6ONLY:
1373 				if (optlen != sizeof(int)) {
1374 					error = EINVAL;
1375 					break;
1376 				}
1377 				error = sooptcopyin(sopt, &optval,
1378 					sizeof optval, sizeof optval);
1379 				if (error)
1380 					break;
1381 				switch (optname) {
1382 
1383 				case IPV6_UNICAST_HOPS:
1384 					if (optval < -1 || optval >= 256)
1385 						error = EINVAL;
1386 					else {
1387 						/* -1 = kernel default */
1388 						in6p->in6p_hops = optval;
1389 
1390 						if ((in6p->in6p_vflag &
1391 						     INP_IPV4) != 0)
1392 							in6p->inp_ip_ttl = optval;
1393 					}
1394 					break;
1395 #define OPTSET(bit) \
1396 do { \
1397 	if (optval) \
1398 		in6p->in6p_flags |= (bit); \
1399 	else \
1400 		in6p->in6p_flags &= ~(bit); \
1401 } while (0)
1402 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1403 
1404 				case IPV6_CHECKSUM:
1405 					in6p->in6p_cksum = optval;
1406 					break;
1407 
1408 				case IPV6_FAITH:
1409 					OPTSET(IN6P_FAITH);
1410 					break;
1411 
1412 				case IPV6_V6ONLY:
1413 					/*
1414 					 * make setsockopt(IPV6_V6ONLY)
1415 					 * available only prior to bind(2).
1416 					 * see ipng mailing list, Jun 22 2001.
1417 					 */
1418 					if (in6p->in6p_lport ||
1419 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
1420 					{
1421 						error = EINVAL;
1422 						break;
1423 					}
1424 					OPTSET(IN6P_IPV6_V6ONLY);
1425 					if (optval)
1426 						in6p->in6p_vflag &= ~INP_IPV4;
1427 					else
1428 						in6p->in6p_vflag |= INP_IPV4;
1429 					break;
1430 				}
1431 				break;
1432 
1433 			case IPV6_PKTINFO:
1434 			case IPV6_HOPLIMIT:
1435 			case IPV6_HOPOPTS:
1436 			case IPV6_DSTOPTS:
1437 			case IPV6_RTHDR:
1438 				/* RFC 2292 */
1439 				if (optlen != sizeof(int)) {
1440 					error = EINVAL;
1441 					break;
1442 				}
1443 				error = sooptcopyin(sopt, &optval,
1444 					sizeof optval, sizeof optval);
1445 				if (error)
1446 					break;
1447 				switch (optname) {
1448 				case IPV6_PKTINFO:
1449 					OPTSET(IN6P_PKTINFO);
1450 					break;
1451 				case IPV6_HOPLIMIT:
1452 					OPTSET(IN6P_HOPLIMIT);
1453 					break;
1454 				case IPV6_HOPOPTS:
1455 					/*
1456 					 * Check super-user privilege.
1457 					 * See comments for IPV6_RECVHOPOPTS.
1458 					 */
1459 					if (!privileged)
1460 						return(EPERM);
1461 					OPTSET(IN6P_HOPOPTS);
1462 					break;
1463 				case IPV6_DSTOPTS:
1464 					if (!privileged)
1465 						return(EPERM);
1466 					OPTSET(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1467 					break;
1468 				case IPV6_RTHDR:
1469 					OPTSET(IN6P_RTHDR);
1470 					break;
1471 				}
1472 				break;
1473 #undef OPTSET
1474 
1475 			case IPV6_MULTICAST_IF:
1476 			case IPV6_MULTICAST_HOPS:
1477 			case IPV6_MULTICAST_LOOP:
1478 			case IPV6_JOIN_GROUP:
1479 			case IPV6_LEAVE_GROUP:
1480 			    {
1481 				struct mbuf *m;
1482 				if (sopt->sopt_valsize > MLEN) {
1483 					error = EMSGSIZE;
1484 					break;
1485 				}
1486 				/* XXX */
1487 				MGET(m, sopt->sopt_p ? M_WAIT : M_DONTWAIT, MT_HEADER);
1488 				if (m == 0) {
1489 					error = ENOBUFS;
1490 					break;
1491 				}
1492 				m->m_len = sopt->sopt_valsize;
1493 				error = sooptcopyin(sopt, mtod(m, char *),
1494 						    m->m_len, m->m_len);
1495 				error =	ip6_setmoptions(sopt->sopt_name,
1496 							&in6p->in6p_moptions,
1497 							m);
1498 				(void)m_free(m);
1499 			    }
1500 				break;
1501 
1502 			case IPV6_PORTRANGE:
1503 				error = sooptcopyin(sopt, &optval,
1504 				    sizeof optval, sizeof optval);
1505 				if (error)
1506 					break;
1507 
1508 				switch (optval) {
1509 				case IPV6_PORTRANGE_DEFAULT:
1510 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1511 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1512 					break;
1513 
1514 				case IPV6_PORTRANGE_HIGH:
1515 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1516 					in6p->in6p_flags |= IN6P_HIGHPORT;
1517 					break;
1518 
1519 				case IPV6_PORTRANGE_LOW:
1520 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1521 					in6p->in6p_flags |= IN6P_LOWPORT;
1522 					break;
1523 
1524 				default:
1525 					error = EINVAL;
1526 					break;
1527 				}
1528 				break;
1529 
1530 #if defined(IPSEC) || defined(FAST_IPSEC)
1531 			case IPV6_IPSEC_POLICY:
1532 			    {
1533 				caddr_t req = NULL;
1534 				size_t len = 0;
1535 				struct mbuf *m;
1536 
1537 				if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1538 					break;
1539 				if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1540 					break;
1541 				if (m) {
1542 					req = mtod(m, caddr_t);
1543 					len = m->m_len;
1544 				}
1545 				error = ipsec6_set_policy(in6p, optname, req,
1546 							  len, privileged);
1547 				m_freem(m);
1548 			    }
1549 				break;
1550 #endif /* KAME IPSEC */
1551 
1552 			case IPV6_FW_ADD:
1553 			case IPV6_FW_DEL:
1554 			case IPV6_FW_FLUSH:
1555 			case IPV6_FW_ZERO:
1556 			    {
1557 				struct mbuf *m;
1558 				struct mbuf **mp = &m;
1559 
1560 				if (ip6_fw_ctl_ptr == NULL)
1561 					return EINVAL;
1562 				/* XXX */
1563 				if ((error = soopt_getm(sopt, &m)) != 0)
1564 					break;
1565 				/* XXX */
1566 				if ((error = soopt_mcopyin(sopt, m)) != 0)
1567 					break;
1568 				error = (*ip6_fw_ctl_ptr)(optname, mp);
1569 				m = *mp;
1570 			    }
1571 				break;
1572 
1573 			default:
1574 				error = ENOPROTOOPT;
1575 				break;
1576 			}
1577 			break;
1578 
1579 		case SOPT_GET:
1580 			switch (optname) {
1581 
1582 			case IPV6_PKTOPTIONS:
1583 				if (in6p->in6p_options) {
1584 					struct mbuf *m;
1585 					m = m_copym(in6p->in6p_options,
1586 					    0, M_COPYALL, M_WAIT);
1587 					error = soopt_mcopyout(sopt, m);
1588 					if (error == 0)
1589 						m_freem(m);
1590 				} else
1591 					sopt->sopt_valsize = 0;
1592 				break;
1593 
1594 			case IPV6_UNICAST_HOPS:
1595 			case IPV6_CHECKSUM:
1596 
1597 			case IPV6_FAITH:
1598 			case IPV6_V6ONLY:
1599 			case IPV6_PORTRANGE:
1600 				switch (optname) {
1601 
1602 				case IPV6_UNICAST_HOPS:
1603 					optval = in6p->in6p_hops;
1604 					break;
1605 
1606 				case IPV6_CHECKSUM:
1607 					optval = in6p->in6p_cksum;
1608 					break;
1609 
1610 				case IPV6_FAITH:
1611 					optval = OPTBIT(IN6P_FAITH);
1612 					break;
1613 
1614 				case IPV6_V6ONLY:
1615 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
1616 					break;
1617 
1618 				case IPV6_PORTRANGE:
1619 				    {
1620 					int flags;
1621 					flags = in6p->in6p_flags;
1622 					if (flags & IN6P_HIGHPORT)
1623 						optval = IPV6_PORTRANGE_HIGH;
1624 					else if (flags & IN6P_LOWPORT)
1625 						optval = IPV6_PORTRANGE_LOW;
1626 					else
1627 						optval = 0;
1628 					break;
1629 				    }
1630 				}
1631 				error = sooptcopyout(sopt, &optval,
1632 					sizeof optval);
1633 				break;
1634 
1635 			case IPV6_PKTINFO:
1636 			case IPV6_HOPLIMIT:
1637 			case IPV6_HOPOPTS:
1638 			case IPV6_RTHDR:
1639 			case IPV6_DSTOPTS:
1640 				if (optname == IPV6_HOPOPTS ||
1641 				    optname == IPV6_DSTOPTS ||
1642 				    !privileged)
1643 					return(EPERM);
1644 				switch (optname) {
1645 				case IPV6_PKTINFO:
1646 					optval = OPTBIT(IN6P_PKTINFO);
1647 					break;
1648 				case IPV6_HOPLIMIT:
1649 					optval = OPTBIT(IN6P_HOPLIMIT);
1650 					break;
1651 				case IPV6_HOPOPTS:
1652 					if (!privileged)
1653 						return(EPERM);
1654 					optval = OPTBIT(IN6P_HOPOPTS);
1655 					break;
1656 				case IPV6_RTHDR:
1657 					optval = OPTBIT(IN6P_RTHDR);
1658 					break;
1659 				case IPV6_DSTOPTS:
1660 					if (!privileged)
1661 						return(EPERM);
1662 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1663 					break;
1664 				}
1665 				error = sooptcopyout(sopt, &optval,
1666 					sizeof optval);
1667 				break;
1668 
1669 			case IPV6_MULTICAST_IF:
1670 			case IPV6_MULTICAST_HOPS:
1671 			case IPV6_MULTICAST_LOOP:
1672 			case IPV6_JOIN_GROUP:
1673 			case IPV6_LEAVE_GROUP:
1674 			    {
1675 				struct mbuf *m;
1676 				error = ip6_getmoptions(sopt->sopt_name,
1677 						in6p->in6p_moptions, &m);
1678 				if (error == 0)
1679 					error = sooptcopyout(sopt,
1680 						mtod(m, char *), m->m_len);
1681 				m_freem(m);
1682 			    }
1683 				break;
1684 
1685 #if defined(IPSEC) || defined(FAST_IPSEC)
1686 			case IPV6_IPSEC_POLICY:
1687 			  {
1688 				caddr_t req = NULL;
1689 				size_t len = 0;
1690 				struct mbuf *m = NULL;
1691 				struct mbuf **mp = &m;
1692 
1693 				error = soopt_getm(sopt, &m); /* XXX */
1694 				if (error != NULL)
1695 					break;
1696 				error = soopt_mcopyin(sopt, m); /* XXX */
1697 				if (error != NULL)
1698 					break;
1699 				if (m) {
1700 					req = mtod(m, caddr_t);
1701 					len = m->m_len;
1702 				}
1703 				error = ipsec6_get_policy(in6p, req, len, mp);
1704 				if (error == 0)
1705 					error = soopt_mcopyout(sopt, m); /*XXX*/
1706 				if (error == 0 && m)
1707 					m_freem(m);
1708 				break;
1709 			  }
1710 #endif /* KAME IPSEC */
1711 
1712 			case IPV6_FW_GET:
1713 			  {
1714 				struct mbuf *m;
1715 				struct mbuf **mp = &m;
1716 
1717 				if (ip6_fw_ctl_ptr == NULL)
1718 				{
1719 					return EINVAL;
1720 				}
1721 				error = (*ip6_fw_ctl_ptr)(optname, mp);
1722 				if (error == 0)
1723 					error = soopt_mcopyout(sopt, m); /* XXX */
1724 				if (error == 0 && m)
1725 					m_freem(m);
1726 			  }
1727 				break;
1728 
1729 			default:
1730 				error = ENOPROTOOPT;
1731 				break;
1732 			}
1733 			break;
1734 		}
1735 	} else {
1736 		error = EINVAL;
1737 	}
1738 	return(error);
1739 }
1740 
1741 /*
1742  * Set up IP6 options in pcb for insertion in output packets or
1743  * specifying behavior of outgoing packets.
1744  */
1745 static int
1746 ip6_pcbopts(pktopt, m, so, sopt)
1747 	struct ip6_pktopts **pktopt;
1748 	struct mbuf *m;
1749 	struct socket *so;
1750 	struct sockopt *sopt;
1751 {
1752 	struct ip6_pktopts *opt = *pktopt;
1753 	int error = 0;
1754 	struct proc *p = sopt->sopt_p;
1755 	int priv = 0;
1756 
1757 	/* turn off any old options. */
1758 	if (opt) {
1759 #ifdef DIAGNOSTIC
1760 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
1761 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
1762 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
1763 			printf("ip6_pcbopts: all specified options are cleared.\n");
1764 #endif
1765 		ip6_clearpktopts(opt, 1, -1);
1766 	} else
1767 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
1768 	*pktopt = NULL;
1769 
1770 	if (!m || m->m_len == 0) {
1771 		/*
1772 		 * Only turning off any previous options, regardless of
1773 		 * whether the opt is just created or given.
1774 		 */
1775 		free(opt, M_IP6OPT);
1776 		return(0);
1777 	}
1778 
1779 	/*  set options specified by user. */
1780 	if (p && !suser(p))
1781 		priv = 1;
1782 	if ((error = ip6_setpktoptions(m, opt, priv, 1)) != 0) {
1783 		ip6_clearpktopts(opt, 1, -1); /* XXX: discard all options */
1784 		free(opt, M_IP6OPT);
1785 		return(error);
1786 	}
1787 	*pktopt = opt;
1788 	return(0);
1789 }
1790 
1791 /*
1792  * initialize ip6_pktopts.  beware that there are non-zero default values in
1793  * the struct.
1794  */
1795 void
1796 init_ip6pktopts(opt)
1797 	struct ip6_pktopts *opt;
1798 {
1799 
1800 	bzero(opt, sizeof(*opt));
1801 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
1802 }
1803 
1804 void
1805 ip6_clearpktopts(pktopt, needfree, optname)
1806 	struct ip6_pktopts *pktopt;
1807 	int needfree, optname;
1808 {
1809 	if (pktopt == NULL)
1810 		return;
1811 
1812 	if (optname == -1) {
1813 		if (needfree && pktopt->ip6po_pktinfo)
1814 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
1815 		pktopt->ip6po_pktinfo = NULL;
1816 	}
1817 	if (optname == -1)
1818 		pktopt->ip6po_hlim = -1;
1819 	if (optname == -1) {
1820 		if (needfree && pktopt->ip6po_nexthop)
1821 			free(pktopt->ip6po_nexthop, M_IP6OPT);
1822 		pktopt->ip6po_nexthop = NULL;
1823 	}
1824 	if (optname == -1) {
1825 		if (needfree && pktopt->ip6po_hbh)
1826 			free(pktopt->ip6po_hbh, M_IP6OPT);
1827 		pktopt->ip6po_hbh = NULL;
1828 	}
1829 	if (optname == -1) {
1830 		if (needfree && pktopt->ip6po_dest1)
1831 			free(pktopt->ip6po_dest1, M_IP6OPT);
1832 		pktopt->ip6po_dest1 = NULL;
1833 	}
1834 	if (optname == -1) {
1835 		if (needfree && pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
1836 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
1837 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
1838 		if (pktopt->ip6po_route.ro_rt) {
1839 			RTFREE(pktopt->ip6po_route.ro_rt);
1840 			pktopt->ip6po_route.ro_rt = NULL;
1841 		}
1842 	}
1843 	if (optname == -1) {
1844 		if (needfree && pktopt->ip6po_dest2)
1845 			free(pktopt->ip6po_dest2, M_IP6OPT);
1846 		pktopt->ip6po_dest2 = NULL;
1847 	}
1848 }
1849 
1850 #define PKTOPT_EXTHDRCPY(type) \
1851 do {\
1852 	if (src->type) {\
1853 		int hlen =\
1854 			(((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
1855 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
1856 		if (dst->type == NULL && canwait == M_NOWAIT)\
1857 			goto bad;\
1858 		bcopy(src->type, dst->type, hlen);\
1859 	}\
1860 } while (0)
1861 
1862 struct ip6_pktopts *
1863 ip6_copypktopts(src, canwait)
1864 	struct ip6_pktopts *src;
1865 	int canwait;
1866 {
1867 	struct ip6_pktopts *dst;
1868 
1869 	if (src == NULL) {
1870 		printf("ip6_clearpktopts: invalid argument\n");
1871 		return(NULL);
1872 	}
1873 
1874 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
1875 	if (dst == NULL && canwait == M_NOWAIT)
1876 		return (NULL);
1877 	bzero(dst, sizeof(*dst));
1878 
1879 	dst->ip6po_hlim = src->ip6po_hlim;
1880 	if (src->ip6po_pktinfo) {
1881 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
1882 					    M_IP6OPT, canwait);
1883 		if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT)
1884 			goto bad;
1885 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
1886 	}
1887 	if (src->ip6po_nexthop) {
1888 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
1889 					    M_IP6OPT, canwait);
1890 		if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT)
1891 			goto bad;
1892 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
1893 		      src->ip6po_nexthop->sa_len);
1894 	}
1895 	PKTOPT_EXTHDRCPY(ip6po_hbh);
1896 	PKTOPT_EXTHDRCPY(ip6po_dest1);
1897 	PKTOPT_EXTHDRCPY(ip6po_dest2);
1898 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
1899 	return(dst);
1900 
1901   bad:
1902 	if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
1903 	if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
1904 	if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
1905 	if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
1906 	if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
1907 	if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
1908 	free(dst, M_IP6OPT);
1909 	return(NULL);
1910 }
1911 #undef PKTOPT_EXTHDRCPY
1912 
1913 void
1914 ip6_freepcbopts(pktopt)
1915 	struct ip6_pktopts *pktopt;
1916 {
1917 	if (pktopt == NULL)
1918 		return;
1919 
1920 	ip6_clearpktopts(pktopt, 1, -1);
1921 
1922 	free(pktopt, M_IP6OPT);
1923 }
1924 
1925 /*
1926  * Set the IP6 multicast options in response to user setsockopt().
1927  */
1928 static int
1929 ip6_setmoptions(optname, im6op, m)
1930 	int optname;
1931 	struct ip6_moptions **im6op;
1932 	struct mbuf *m;
1933 {
1934 	int error = 0;
1935 	u_int loop, ifindex;
1936 	struct ipv6_mreq *mreq;
1937 	struct ifnet *ifp;
1938 	struct ip6_moptions *im6o = *im6op;
1939 	struct route_in6 ro;
1940 	struct sockaddr_in6 *dst;
1941 	struct in6_multi_mship *imm;
1942 	struct proc *p = curproc;	/* XXX */
1943 
1944 	if (im6o == NULL) {
1945 		/*
1946 		 * No multicast option buffer attached to the pcb;
1947 		 * allocate one and initialize to default values.
1948 		 */
1949 		im6o = (struct ip6_moptions *)
1950 			malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
1951 
1952 		if (im6o == NULL)
1953 			return(ENOBUFS);
1954 		*im6op = im6o;
1955 		im6o->im6o_multicast_ifp = NULL;
1956 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1957 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
1958 		LIST_INIT(&im6o->im6o_memberships);
1959 	}
1960 
1961 	switch (optname) {
1962 
1963 	case IPV6_MULTICAST_IF:
1964 		/*
1965 		 * Select the interface for outgoing multicast packets.
1966 		 */
1967 		if (m == NULL || m->m_len != sizeof(u_int)) {
1968 			error = EINVAL;
1969 			break;
1970 		}
1971 		bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
1972 		if (ifindex < 0 || if_index < ifindex) {
1973 			error = ENXIO;	/* XXX EINVAL? */
1974 			break;
1975 		}
1976 		ifp = ifindex2ifnet[ifindex];
1977 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1978 			error = EADDRNOTAVAIL;
1979 			break;
1980 		}
1981 		im6o->im6o_multicast_ifp = ifp;
1982 		break;
1983 
1984 	case IPV6_MULTICAST_HOPS:
1985 	    {
1986 		/*
1987 		 * Set the IP6 hoplimit for outgoing multicast packets.
1988 		 */
1989 		int optval;
1990 		if (m == NULL || m->m_len != sizeof(int)) {
1991 			error = EINVAL;
1992 			break;
1993 		}
1994 		bcopy(mtod(m, u_int *), &optval, sizeof(optval));
1995 		if (optval < -1 || optval >= 256)
1996 			error = EINVAL;
1997 		else if (optval == -1)
1998 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1999 		else
2000 			im6o->im6o_multicast_hlim = optval;
2001 		break;
2002 	    }
2003 
2004 	case IPV6_MULTICAST_LOOP:
2005 		/*
2006 		 * Set the loopback flag for outgoing multicast packets.
2007 		 * Must be zero or one.
2008 		 */
2009 		if (m == NULL || m->m_len != sizeof(u_int)) {
2010 			error = EINVAL;
2011 			break;
2012 		}
2013 		bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2014 		if (loop > 1) {
2015 			error = EINVAL;
2016 			break;
2017 		}
2018 		im6o->im6o_multicast_loop = loop;
2019 		break;
2020 
2021 	case IPV6_JOIN_GROUP:
2022 		/*
2023 		 * Add a multicast group membership.
2024 		 * Group must be a valid IP6 multicast address.
2025 		 */
2026 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2027 			error = EINVAL;
2028 			break;
2029 		}
2030 		mreq = mtod(m, struct ipv6_mreq *);
2031 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2032 			/*
2033 			 * We use the unspecified address to specify to accept
2034 			 * all multicast addresses. Only super user is allowed
2035 			 * to do this.
2036 			 */
2037 			if (suser(p))
2038 			{
2039 				error = EACCES;
2040 				break;
2041 			}
2042 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2043 			error = EINVAL;
2044 			break;
2045 		}
2046 
2047 		/*
2048 		 * If the interface is specified, validate it.
2049 		 */
2050 		if (mreq->ipv6mr_interface < 0
2051 		 || if_index < mreq->ipv6mr_interface) {
2052 			error = ENXIO;	/* XXX EINVAL? */
2053 			break;
2054 		}
2055 		/*
2056 		 * If no interface was explicitly specified, choose an
2057 		 * appropriate one according to the given multicast address.
2058 		 */
2059 		if (mreq->ipv6mr_interface == 0) {
2060 			/*
2061 			 * If the multicast address is in node-local scope,
2062 			 * the interface should be a loopback interface.
2063 			 * Otherwise, look up the routing table for the
2064 			 * address, and choose the outgoing interface.
2065 			 *   XXX: is it a good approach?
2066 			 */
2067 			if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
2068 				ifp = &loif[0];
2069 			} else {
2070 				ro.ro_rt = NULL;
2071 				dst = (struct sockaddr_in6 *)&ro.ro_dst;
2072 				bzero(dst, sizeof(*dst));
2073 				dst->sin6_len = sizeof(struct sockaddr_in6);
2074 				dst->sin6_family = AF_INET6;
2075 				dst->sin6_addr = mreq->ipv6mr_multiaddr;
2076 				rtalloc((struct route *)&ro);
2077 				if (ro.ro_rt == NULL) {
2078 					error = EADDRNOTAVAIL;
2079 					break;
2080 				}
2081 				ifp = ro.ro_rt->rt_ifp;
2082 				rtfree(ro.ro_rt);
2083 			}
2084 		} else
2085 			ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2086 
2087 		/*
2088 		 * See if we found an interface, and confirm that it
2089 		 * supports multicast
2090 		 */
2091 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2092 			error = EADDRNOTAVAIL;
2093 			break;
2094 		}
2095 		/*
2096 		 * Put interface index into the multicast address,
2097 		 * if the address has link-local scope.
2098 		 */
2099 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2100 			mreq->ipv6mr_multiaddr.s6_addr16[1]
2101 				= htons(mreq->ipv6mr_interface);
2102 		}
2103 		/*
2104 		 * See if the membership already exists.
2105 		 */
2106 		for (imm = im6o->im6o_memberships.lh_first;
2107 		     imm != NULL; imm = imm->i6mm_chain.le_next)
2108 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
2109 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2110 					       &mreq->ipv6mr_multiaddr))
2111 				break;
2112 		if (imm != NULL) {
2113 			error = EADDRINUSE;
2114 			break;
2115 		}
2116 		/*
2117 		 * Everything looks good; add a new record to the multicast
2118 		 * address list for the given interface.
2119 		 */
2120 		imm = malloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
2121 		if (imm == NULL) {
2122 			error = ENOBUFS;
2123 			break;
2124 		}
2125 		if ((imm->i6mm_maddr =
2126 		     in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
2127 			free(imm, M_IPMADDR);
2128 			break;
2129 		}
2130 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2131 		break;
2132 
2133 	case IPV6_LEAVE_GROUP:
2134 		/*
2135 		 * Drop a multicast group membership.
2136 		 * Group must be a valid IP6 multicast address.
2137 		 */
2138 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2139 			error = EINVAL;
2140 			break;
2141 		}
2142 		mreq = mtod(m, struct ipv6_mreq *);
2143 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2144 			if (suser(p)) {
2145 				error = EACCES;
2146 				break;
2147 			}
2148 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2149 			error = EINVAL;
2150 			break;
2151 		}
2152 		/*
2153 		 * If an interface address was specified, get a pointer
2154 		 * to its ifnet structure.
2155 		 */
2156 		if (mreq->ipv6mr_interface < 0
2157 		 || if_index < mreq->ipv6mr_interface) {
2158 			error = ENXIO;	/* XXX EINVAL? */
2159 			break;
2160 		}
2161 		ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2162 		/*
2163 		 * Put interface index into the multicast address,
2164 		 * if the address has link-local scope.
2165 		 */
2166 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2167 			mreq->ipv6mr_multiaddr.s6_addr16[1]
2168 				= htons(mreq->ipv6mr_interface);
2169 		}
2170 		/*
2171 		 * Find the membership in the membership list.
2172 		 */
2173 		for (imm = im6o->im6o_memberships.lh_first;
2174 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
2175 			if ((ifp == NULL ||
2176 			     imm->i6mm_maddr->in6m_ifp == ifp) &&
2177 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2178 					       &mreq->ipv6mr_multiaddr))
2179 				break;
2180 		}
2181 		if (imm == NULL) {
2182 			/* Unable to resolve interface */
2183 			error = EADDRNOTAVAIL;
2184 			break;
2185 		}
2186 		/*
2187 		 * Give up the multicast address record to which the
2188 		 * membership points.
2189 		 */
2190 		LIST_REMOVE(imm, i6mm_chain);
2191 		in6_delmulti(imm->i6mm_maddr);
2192 		free(imm, M_IPMADDR);
2193 		break;
2194 
2195 	default:
2196 		error = EOPNOTSUPP;
2197 		break;
2198 	}
2199 
2200 	/*
2201 	 * If all options have default values, no need to keep the mbuf.
2202 	 */
2203 	if (im6o->im6o_multicast_ifp == NULL &&
2204 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2205 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2206 	    im6o->im6o_memberships.lh_first == NULL) {
2207 		free(*im6op, M_IPMOPTS);
2208 		*im6op = NULL;
2209 	}
2210 
2211 	return(error);
2212 }
2213 
2214 /*
2215  * Return the IP6 multicast options in response to user getsockopt().
2216  */
2217 static int
2218 ip6_getmoptions(optname, im6o, mp)
2219 	int optname;
2220 	struct ip6_moptions *im6o;
2221 	struct mbuf **mp;
2222 {
2223 	u_int *hlim, *loop, *ifindex;
2224 
2225 	*mp = m_get(M_WAIT, MT_HEADER);		/* XXX */
2226 
2227 	switch (optname) {
2228 
2229 	case IPV6_MULTICAST_IF:
2230 		ifindex = mtod(*mp, u_int *);
2231 		(*mp)->m_len = sizeof(u_int);
2232 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2233 			*ifindex = 0;
2234 		else
2235 			*ifindex = im6o->im6o_multicast_ifp->if_index;
2236 		return(0);
2237 
2238 	case IPV6_MULTICAST_HOPS:
2239 		hlim = mtod(*mp, u_int *);
2240 		(*mp)->m_len = sizeof(u_int);
2241 		if (im6o == NULL)
2242 			*hlim = ip6_defmcasthlim;
2243 		else
2244 			*hlim = im6o->im6o_multicast_hlim;
2245 		return(0);
2246 
2247 	case IPV6_MULTICAST_LOOP:
2248 		loop = mtod(*mp, u_int *);
2249 		(*mp)->m_len = sizeof(u_int);
2250 		if (im6o == NULL)
2251 			*loop = ip6_defmcasthlim;
2252 		else
2253 			*loop = im6o->im6o_multicast_loop;
2254 		return(0);
2255 
2256 	default:
2257 		return(EOPNOTSUPP);
2258 	}
2259 }
2260 
2261 /*
2262  * Discard the IP6 multicast options.
2263  */
2264 void
2265 ip6_freemoptions(im6o)
2266 	struct ip6_moptions *im6o;
2267 {
2268 	struct in6_multi_mship *imm;
2269 
2270 	if (im6o == NULL)
2271 		return;
2272 
2273 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2274 		LIST_REMOVE(imm, i6mm_chain);
2275 		if (imm->i6mm_maddr)
2276 			in6_delmulti(imm->i6mm_maddr);
2277 		free(imm, M_IPMADDR);
2278 	}
2279 	free(im6o, M_IPMOPTS);
2280 }
2281 
2282 /*
2283  * Set IPv6 outgoing packet options based on advanced API.
2284  */
2285 int
2286 ip6_setpktoptions(control, opt, priv, needcopy)
2287 	struct mbuf *control;
2288 	struct ip6_pktopts *opt;
2289 	int priv, needcopy;
2290 {
2291 	struct cmsghdr *cm = 0;
2292 
2293 	if (control == 0 || opt == 0)
2294 		return(EINVAL);
2295 
2296 	init_ip6pktopts(opt);
2297 
2298 	/*
2299 	 * XXX: Currently, we assume all the optional information is stored
2300 	 * in a single mbuf.
2301 	 */
2302 	if (control->m_next)
2303 		return(EINVAL);
2304 
2305 	for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2306 		     control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2307 		cm = mtod(control, struct cmsghdr *);
2308 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2309 			return(EINVAL);
2310 		if (cm->cmsg_level != IPPROTO_IPV6)
2311 			continue;
2312 
2313 		/*
2314 		 * XXX should check if RFC2292 API is mixed with 2292bis API
2315 		 */
2316 		switch (cm->cmsg_type) {
2317 		case IPV6_PKTINFO:
2318 			if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo)))
2319 				return(EINVAL);
2320 			if (needcopy) {
2321 				/* XXX: Is it really WAITOK? */
2322 				opt->ip6po_pktinfo =
2323 					malloc(sizeof(struct in6_pktinfo),
2324 					       M_IP6OPT, M_WAITOK);
2325 				bcopy(CMSG_DATA(cm), opt->ip6po_pktinfo,
2326 				    sizeof(struct in6_pktinfo));
2327 			} else
2328 				opt->ip6po_pktinfo =
2329 					(struct in6_pktinfo *)CMSG_DATA(cm);
2330 			if (opt->ip6po_pktinfo->ipi6_ifindex &&
2331 			    IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr))
2332 				opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] =
2333 					htons(opt->ip6po_pktinfo->ipi6_ifindex);
2334 
2335 			if (opt->ip6po_pktinfo->ipi6_ifindex > if_index
2336 			 || opt->ip6po_pktinfo->ipi6_ifindex < 0) {
2337 				return(ENXIO);
2338 			}
2339 
2340 			/*
2341 			 * Check if the requested source address is indeed a
2342 			 * unicast address assigned to the node, and can be
2343 			 * used as the packet's source address.
2344 			 */
2345 			if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
2346 				struct in6_ifaddr *ia6;
2347 				struct sockaddr_in6 sin6;
2348 
2349 				bzero(&sin6, sizeof(sin6));
2350 				sin6.sin6_len = sizeof(sin6);
2351 				sin6.sin6_family = AF_INET6;
2352 				sin6.sin6_addr =
2353 					opt->ip6po_pktinfo->ipi6_addr;
2354 				ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(sin6tosa(&sin6));
2355 				if (ia6 == NULL ||
2356 				    (ia6->ia6_flags & (IN6_IFF_ANYCAST |
2357 						       IN6_IFF_NOTREADY)) != 0)
2358 					return(EADDRNOTAVAIL);
2359 			}
2360 			break;
2361 
2362 		case IPV6_HOPLIMIT:
2363 			if (cm->cmsg_len != CMSG_LEN(sizeof(int)))
2364 				return(EINVAL);
2365 
2366 			opt->ip6po_hlim = *(int *)CMSG_DATA(cm);
2367 			if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255)
2368 				return(EINVAL);
2369 			break;
2370 
2371 		case IPV6_NEXTHOP:
2372 			if (!priv)
2373 				return(EPERM);
2374 
2375 			if (cm->cmsg_len < sizeof(u_char) ||
2376 			    /* check if cmsg_len is large enough for sa_len */
2377 			    cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm)))
2378 				return(EINVAL);
2379 
2380 			if (needcopy) {
2381 				opt->ip6po_nexthop =
2382 					malloc(*CMSG_DATA(cm),
2383 					       M_IP6OPT, M_WAITOK);
2384 				bcopy(CMSG_DATA(cm),
2385 				      opt->ip6po_nexthop,
2386 				      *CMSG_DATA(cm));
2387 			} else
2388 				opt->ip6po_nexthop =
2389 					(struct sockaddr *)CMSG_DATA(cm);
2390 			break;
2391 
2392 		case IPV6_HOPOPTS:
2393 		{
2394 			struct ip6_hbh *hbh;
2395 			int hbhlen;
2396 
2397 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh)))
2398 				return(EINVAL);
2399 			hbh = (struct ip6_hbh *)CMSG_DATA(cm);
2400 			hbhlen = (hbh->ip6h_len + 1) << 3;
2401 			if (cm->cmsg_len != CMSG_LEN(hbhlen))
2402 				return(EINVAL);
2403 
2404 			if (needcopy) {
2405 				opt->ip6po_hbh =
2406 					malloc(hbhlen, M_IP6OPT, M_WAITOK);
2407 				bcopy(hbh, opt->ip6po_hbh, hbhlen);
2408 			} else
2409 				opt->ip6po_hbh = hbh;
2410 			break;
2411 		}
2412 
2413 		case IPV6_DSTOPTS:
2414 		{
2415 			struct ip6_dest *dest, **newdest;
2416 			int destlen;
2417 
2418 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest)))
2419 				return(EINVAL);
2420 			dest = (struct ip6_dest *)CMSG_DATA(cm);
2421 			destlen = (dest->ip6d_len + 1) << 3;
2422 			if (cm->cmsg_len != CMSG_LEN(destlen))
2423 				return(EINVAL);
2424 
2425 			/*
2426 			 * The old advacned API is ambiguous on this
2427 			 * point. Our approach is to determine the
2428 			 * position based according to the existence
2429 			 * of a routing header. Note, however, that
2430 			 * this depends on the order of the extension
2431 			 * headers in the ancillary data; the 1st part
2432 			 * of the destination options header must
2433 			 * appear before the routing header in the
2434 			 * ancillary data, too.
2435 			 * RFC2292bis solved the ambiguity by
2436 			 * introducing separate cmsg types.
2437 			 */
2438 			if (opt->ip6po_rthdr == NULL)
2439 				newdest = &opt->ip6po_dest1;
2440 			else
2441 				newdest = &opt->ip6po_dest2;
2442 
2443 			if (needcopy) {
2444 				*newdest = malloc(destlen, M_IP6OPT, M_WAITOK);
2445 				bcopy(dest, *newdest, destlen);
2446 			} else
2447 				*newdest = dest;
2448 
2449 			break;
2450 		}
2451 
2452 		case IPV6_RTHDR:
2453 		{
2454 			struct ip6_rthdr *rth;
2455 			int rthlen;
2456 
2457 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr)))
2458 				return(EINVAL);
2459 			rth = (struct ip6_rthdr *)CMSG_DATA(cm);
2460 			rthlen = (rth->ip6r_len + 1) << 3;
2461 			if (cm->cmsg_len != CMSG_LEN(rthlen))
2462 				return(EINVAL);
2463 
2464 			switch (rth->ip6r_type) {
2465 			case IPV6_RTHDR_TYPE_0:
2466 				/* must contain one addr */
2467 				if (rth->ip6r_len == 0)
2468 					return(EINVAL);
2469 				/* length must be even */
2470 				if (rth->ip6r_len % 2)
2471 					return(EINVAL);
2472 				if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2473 					return(EINVAL);
2474 				break;
2475 			default:
2476 				return(EINVAL);	/* not supported */
2477 			}
2478 
2479 			if (needcopy) {
2480 				opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT,
2481 							  M_WAITOK);
2482 				bcopy(rth, opt->ip6po_rthdr, rthlen);
2483 			} else
2484 				opt->ip6po_rthdr = rth;
2485 
2486 			break;
2487 		}
2488 
2489 		default:
2490 			return(ENOPROTOOPT);
2491 		}
2492 	}
2493 
2494 	return(0);
2495 }
2496 
2497 /*
2498  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2499  * packet to the input queue of a specified interface.  Note that this
2500  * calls the output routine of the loopback "driver", but with an interface
2501  * pointer that might NOT be &loif -- easier than replicating that code here.
2502  */
2503 void
2504 ip6_mloopback(ifp, m, dst)
2505 	struct ifnet *ifp;
2506 	struct mbuf *m;
2507 	struct sockaddr_in6 *dst;
2508 {
2509 	struct mbuf *copym;
2510 	struct ip6_hdr *ip6;
2511 
2512 	copym = m_copy(m, 0, M_COPYALL);
2513 	if (copym == NULL)
2514 		return;
2515 
2516 	/*
2517 	 * Make sure to deep-copy IPv6 header portion in case the data
2518 	 * is in an mbuf cluster, so that we can safely override the IPv6
2519 	 * header portion later.
2520 	 */
2521 	if ((copym->m_flags & M_EXT) != 0 ||
2522 	    copym->m_len < sizeof(struct ip6_hdr)) {
2523 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
2524 		if (copym == NULL)
2525 			return;
2526 	}
2527 
2528 #ifdef DIAGNOSTIC
2529 	if (copym->m_len < sizeof(*ip6)) {
2530 		m_freem(copym);
2531 		return;
2532 	}
2533 #endif
2534 
2535 	ip6 = mtod(copym, struct ip6_hdr *);
2536 #ifndef SCOPEDROUTING
2537 	/*
2538 	 * clear embedded scope identifiers if necessary.
2539 	 * in6_clearscope will touch the addresses only when necessary.
2540 	 */
2541 	in6_clearscope(&ip6->ip6_src);
2542 	in6_clearscope(&ip6->ip6_dst);
2543 #endif
2544 
2545 	(void)if_simloop(ifp, copym, dst->sin6_family, NULL);
2546 }
2547 
2548 /*
2549  * Chop IPv6 header off from the payload.
2550  */
2551 static int
2552 ip6_splithdr(m, exthdrs)
2553 	struct mbuf *m;
2554 	struct ip6_exthdrs *exthdrs;
2555 {
2556 	struct mbuf *mh;
2557 	struct ip6_hdr *ip6;
2558 
2559 	ip6 = mtod(m, struct ip6_hdr *);
2560 	if (m->m_len > sizeof(*ip6)) {
2561 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
2562 		if (mh == 0) {
2563 			m_freem(m);
2564 			return ENOBUFS;
2565 		}
2566 		M_MOVE_PKTHDR(mh, m);
2567 		MH_ALIGN(mh, sizeof(*ip6));
2568 		m->m_len -= sizeof(*ip6);
2569 		m->m_data += sizeof(*ip6);
2570 		mh->m_next = m;
2571 		m = mh;
2572 		m->m_len = sizeof(*ip6);
2573 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2574 	}
2575 	exthdrs->ip6e_ip6 = m;
2576 	return 0;
2577 }
2578 
2579 /*
2580  * Compute IPv6 extension header length.
2581  */
2582 int
2583 ip6_optlen(in6p)
2584 	struct in6pcb *in6p;
2585 {
2586 	int len;
2587 
2588 	if (!in6p->in6p_outputopts)
2589 		return 0;
2590 
2591 	len = 0;
2592 #define elen(x) \
2593     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2594 
2595 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
2596 	if (in6p->in6p_outputopts->ip6po_rthdr)
2597 		/* dest1 is valid with rthdr only */
2598 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
2599 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
2600 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
2601 	return len;
2602 #undef elen
2603 }
2604