xref: /freebsd/sys/netinet6/ip6_output.c (revision ee2ea5ceafed78a5bd9810beb9e3ca927180c226)
1 /*	$FreeBSD$	*/
2 /*	$KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1988, 1990, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. All advertising materials mentioning features or use of this software
46  *    must display the following acknowledgement:
47  *	This product includes software developed by the University of
48  *	California, Berkeley and its contributors.
49  * 4. Neither the name of the University nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63  * SUCH DAMAGE.
64  *
65  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
66  */
67 
68 #include "opt_ip6fw.h"
69 #include "opt_inet.h"
70 #include "opt_inet6.h"
71 #include "opt_ipsec.h"
72 #include "opt_pfil_hooks.h"
73 
74 #include <sys/param.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/proc.h>
78 #include <sys/errno.h>
79 #include <sys/protosw.h>
80 #include <sys/socket.h>
81 #include <sys/socketvar.h>
82 #include <sys/systm.h>
83 #include <sys/kernel.h>
84 
85 #include <net/if.h>
86 #include <net/route.h>
87 #ifdef PFIL_HOOKS
88 #include <net/pfil.h>
89 #endif
90 
91 #include <netinet/in.h>
92 #include <netinet/in_var.h>
93 #include <netinet6/in6_var.h>
94 #include <netinet/ip6.h>
95 #include <netinet/icmp6.h>
96 #include <netinet6/ip6_var.h>
97 #include <netinet/in_pcb.h>
98 #include <netinet6/nd6.h>
99 
100 #ifdef IPSEC
101 #include <netinet6/ipsec.h>
102 #ifdef INET6
103 #include <netinet6/ipsec6.h>
104 #endif
105 #include <netkey/key.h>
106 #endif /* IPSEC */
107 
108 #include <netinet6/ip6_fw.h>
109 
110 #include <net/net_osdep.h>
111 
112 #include <netinet6/ip6protosw.h>
113 
114 static MALLOC_DEFINE(M_IPMOPTS, "ip6_moptions", "internet multicast options");
115 
116 struct ip6_exthdrs {
117 	struct mbuf *ip6e_ip6;
118 	struct mbuf *ip6e_hbh;
119 	struct mbuf *ip6e_dest1;
120 	struct mbuf *ip6e_rthdr;
121 	struct mbuf *ip6e_dest2;
122 };
123 
124 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
125 			    struct socket *, struct sockopt *sopt));
126 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *));
127 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **));
128 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int));
129 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
130 				  struct ip6_frag **));
131 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t));
132 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *));
133 
134 /*
135  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
136  * header (with pri, len, nxt, hlim, src, dst).
137  * This function may modify ver and hlim only.
138  * The mbuf chain containing the packet will be freed.
139  * The mbuf opt, if present, will not be freed.
140  *
141  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
142  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
143  * which is rt_rmx.rmx_mtu.
144  */
145 int
146 ip6_output(m0, opt, ro, flags, im6o, ifpp)
147 	struct mbuf *m0;
148 	struct ip6_pktopts *opt;
149 	struct route_in6 *ro;
150 	int flags;
151 	struct ip6_moptions *im6o;
152 	struct ifnet **ifpp;		/* XXX: just for statistics */
153 {
154 	struct ip6_hdr *ip6, *mhip6;
155 	struct ifnet *ifp, *origifp;
156 	struct mbuf *m = m0;
157 	int hlen, tlen, len, off;
158 	struct route_in6 ip6route;
159 	struct sockaddr_in6 *dst;
160 	int error = 0;
161 	struct in6_ifaddr *ia = NULL;
162 	u_long mtu;
163 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
164 	struct ip6_exthdrs exthdrs;
165 	struct in6_addr finaldst;
166 	struct route_in6 *ro_pmtu = NULL;
167 	int hdrsplit = 0;
168 	int needipsec = 0;
169 #ifdef PFIL_HOOKS
170 	struct packet_filter_hook *pfh;
171 	struct mbuf *m1;
172 	int rv;
173 #endif /* PFIL_HOOKS */
174 #ifdef IPSEC
175 	int needipsectun = 0;
176 	struct socket *so;
177 	struct secpolicy *sp = NULL;
178 
179 	/* for AH processing. stupid to have "socket" variable in IP layer... */
180 	so = ipsec_getsocket(m);
181 	(void)ipsec_setsocket(m, NULL);
182 	ip6 = mtod(m, struct ip6_hdr *);
183 #endif /* IPSEC */
184 
185 #define MAKE_EXTHDR(hp, mp)						\
186     do {								\
187 	if (hp) {							\
188 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
189 		error = ip6_copyexthdr((mp), (caddr_t)(hp), 		\
190 				       ((eh)->ip6e_len + 1) << 3);	\
191 		if (error)						\
192 			goto freehdrs;					\
193 	}								\
194     } while (0)
195 
196 	bzero(&exthdrs, sizeof(exthdrs));
197 
198 	if (opt) {
199 		/* Hop-by-Hop options header */
200 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
201 		/* Destination options header(1st part) */
202 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
203 		/* Routing header */
204 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
205 		/* Destination options header(2nd part) */
206 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
207 	}
208 
209 #ifdef IPSEC
210 	/* get a security policy for this packet */
211 	if (so == NULL)
212 		sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
213 	else
214 		sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
215 
216 	if (sp == NULL) {
217 		ipsec6stat.out_inval++;
218 		goto freehdrs;
219 	}
220 
221 	error = 0;
222 
223 	/* check policy */
224 	switch (sp->policy) {
225 	case IPSEC_POLICY_DISCARD:
226 		/*
227 		 * This packet is just discarded.
228 		 */
229 		ipsec6stat.out_polvio++;
230 		goto freehdrs;
231 
232 	case IPSEC_POLICY_BYPASS:
233 	case IPSEC_POLICY_NONE:
234 		/* no need to do IPsec. */
235 		needipsec = 0;
236 		break;
237 
238 	case IPSEC_POLICY_IPSEC:
239 		if (sp->req == NULL) {
240 			/* acquire a policy */
241 			error = key_spdacquire(sp);
242 			goto freehdrs;
243 		}
244 		needipsec = 1;
245 		break;
246 
247 	case IPSEC_POLICY_ENTRUST:
248 	default:
249 		printf("ip6_output: Invalid policy found. %d\n", sp->policy);
250 	}
251 #endif /* IPSEC */
252 
253 	/*
254 	 * Calculate the total length of the extension header chain.
255 	 * Keep the length of the unfragmentable part for fragmentation.
256 	 */
257 	optlen = 0;
258 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
259 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
260 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
261 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
262 	/* NOTE: we don't add AH/ESP length here. do that later. */
263 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
264 
265 	/*
266 	 * If we need IPsec, or there is at least one extension header,
267 	 * separate IP6 header from the payload.
268 	 */
269 	if ((needipsec || optlen) && !hdrsplit) {
270 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
271 			m = NULL;
272 			goto freehdrs;
273 		}
274 		m = exthdrs.ip6e_ip6;
275 		hdrsplit++;
276 	}
277 
278 	/* adjust pointer */
279 	ip6 = mtod(m, struct ip6_hdr *);
280 
281 	/* adjust mbuf packet header length */
282 	m->m_pkthdr.len += optlen;
283 	plen = m->m_pkthdr.len - sizeof(*ip6);
284 
285 	/* If this is a jumbo payload, insert a jumbo payload option. */
286 	if (plen > IPV6_MAXPACKET) {
287 		if (!hdrsplit) {
288 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
289 				m = NULL;
290 				goto freehdrs;
291 			}
292 			m = exthdrs.ip6e_ip6;
293 			hdrsplit++;
294 		}
295 		/* adjust pointer */
296 		ip6 = mtod(m, struct ip6_hdr *);
297 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
298 			goto freehdrs;
299 		ip6->ip6_plen = 0;
300 	} else
301 		ip6->ip6_plen = htons(plen);
302 
303 	/*
304 	 * Concatenate headers and fill in next header fields.
305 	 * Here we have, on "m"
306 	 *	IPv6 payload
307 	 * and we insert headers accordingly.  Finally, we should be getting:
308 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
309 	 *
310 	 * during the header composing process, "m" points to IPv6 header.
311 	 * "mprev" points to an extension header prior to esp.
312 	 */
313 	{
314 		u_char *nexthdrp = &ip6->ip6_nxt;
315 		struct mbuf *mprev = m;
316 
317 		/*
318 		 * we treat dest2 specially.  this makes IPsec processing
319 		 * much easier.  the goal here is to make mprev point the
320 		 * mbuf prior to dest2.
321 		 *
322 		 * result: IPv6 dest2 payload
323 		 * m and mprev will point to IPv6 header.
324 		 */
325 		if (exthdrs.ip6e_dest2) {
326 			if (!hdrsplit)
327 				panic("assumption failed: hdr not split");
328 			exthdrs.ip6e_dest2->m_next = m->m_next;
329 			m->m_next = exthdrs.ip6e_dest2;
330 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
331 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
332 		}
333 
334 #define MAKE_CHAIN(m, mp, p, i)\
335     do {\
336 	if (m) {\
337 		if (!hdrsplit) \
338 			panic("assumption failed: hdr not split"); \
339 		*mtod((m), u_char *) = *(p);\
340 		*(p) = (i);\
341 		p = mtod((m), u_char *);\
342 		(m)->m_next = (mp)->m_next;\
343 		(mp)->m_next = (m);\
344 		(mp) = (m);\
345 	}\
346     } while (0)
347 		/*
348 		 * result: IPv6 hbh dest1 rthdr dest2 payload
349 		 * m will point to IPv6 header.  mprev will point to the
350 		 * extension header prior to dest2 (rthdr in the above case).
351 		 */
352 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev,
353 			   nexthdrp, IPPROTO_HOPOPTS);
354 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev,
355 			   nexthdrp, IPPROTO_DSTOPTS);
356 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev,
357 			   nexthdrp, IPPROTO_ROUTING);
358 
359 #ifdef IPSEC
360 		if (!needipsec)
361 			goto skip_ipsec2;
362 
363 		/*
364 		 * pointers after IPsec headers are not valid any more.
365 		 * other pointers need a great care too.
366 		 * (IPsec routines should not mangle mbufs prior to AH/ESP)
367 		 */
368 		exthdrs.ip6e_dest2 = NULL;
369 
370 	    {
371 		struct ip6_rthdr *rh = NULL;
372 		int segleft_org = 0;
373 		struct ipsec_output_state state;
374 
375 		if (exthdrs.ip6e_rthdr) {
376 			rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
377 			segleft_org = rh->ip6r_segleft;
378 			rh->ip6r_segleft = 0;
379 		}
380 
381 		bzero(&state, sizeof(state));
382 		state.m = m;
383 		error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
384 			&needipsectun);
385 		m = state.m;
386 		if (error) {
387 			/* mbuf is already reclaimed in ipsec6_output_trans. */
388 			m = NULL;
389 			switch (error) {
390 			case EHOSTUNREACH:
391 			case ENETUNREACH:
392 			case EMSGSIZE:
393 			case ENOBUFS:
394 			case ENOMEM:
395 				break;
396 			default:
397 				printf("ip6_output (ipsec): error code %d\n", error);
398 				/* fall through */
399 			case ENOENT:
400 				/* don't show these error codes to the user */
401 				error = 0;
402 				break;
403 			}
404 			goto bad;
405 		}
406 		if (exthdrs.ip6e_rthdr) {
407 			/* ah6_output doesn't modify mbuf chain */
408 			rh->ip6r_segleft = segleft_org;
409 		}
410 	    }
411 skip_ipsec2:;
412 #endif
413 	}
414 
415 	/*
416 	 * If there is a routing header, replace destination address field
417 	 * with the first hop of the routing header.
418 	 */
419 	if (exthdrs.ip6e_rthdr) {
420 		struct ip6_rthdr *rh =
421 			(struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
422 						  struct ip6_rthdr *));
423 		struct ip6_rthdr0 *rh0;
424 
425 		finaldst = ip6->ip6_dst;
426 		switch (rh->ip6r_type) {
427 		case IPV6_RTHDR_TYPE_0:
428 			 rh0 = (struct ip6_rthdr0 *)rh;
429 			 ip6->ip6_dst = rh0->ip6r0_addr[0];
430 			 bcopy((caddr_t)&rh0->ip6r0_addr[1],
431 			       (caddr_t)&rh0->ip6r0_addr[0],
432 			       sizeof(struct in6_addr)*(rh0->ip6r0_segleft - 1)
433 				 );
434 			 rh0->ip6r0_addr[rh0->ip6r0_segleft - 1] = finaldst;
435 			 break;
436 		default:	/* is it possible? */
437 			 error = EINVAL;
438 			 goto bad;
439 		}
440 	}
441 
442 	/* Source address validation */
443 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
444 	    (flags & IPV6_DADOUTPUT) == 0) {
445 		error = EOPNOTSUPP;
446 		ip6stat.ip6s_badscope++;
447 		goto bad;
448 	}
449 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
450 		error = EOPNOTSUPP;
451 		ip6stat.ip6s_badscope++;
452 		goto bad;
453 	}
454 
455 	ip6stat.ip6s_localout++;
456 
457 	/*
458 	 * Route packet.
459 	 */
460 	if (ro == 0) {
461 		ro = &ip6route;
462 		bzero((caddr_t)ro, sizeof(*ro));
463 	}
464 	ro_pmtu = ro;
465 	if (opt && opt->ip6po_rthdr)
466 		ro = &opt->ip6po_route;
467 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
468 	/*
469 	 * If there is a cached route,
470 	 * check that it is to the same destination
471 	 * and is still up. If not, free it and try again.
472 	 */
473 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
474 			 dst->sin6_family != AF_INET6 ||
475 			 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
476 		RTFREE(ro->ro_rt);
477 		ro->ro_rt = (struct rtentry *)0;
478 	}
479 	if (ro->ro_rt == 0) {
480 		bzero(dst, sizeof(*dst));
481 		dst->sin6_family = AF_INET6;
482 		dst->sin6_len = sizeof(struct sockaddr_in6);
483 		dst->sin6_addr = ip6->ip6_dst;
484 #ifdef SCOPEDROUTING
485 		/* XXX: sin6_scope_id should already be fixed at this point */
486 		if (IN6_IS_SCOPE_LINKLOCAL(&dst->sin6_addr))
487 			dst->sin6_scope_id = ntohs(dst->sin6_addr.s6_addr16[1]);
488 #endif
489 	}
490 #ifdef IPSEC
491 	if (needipsec && needipsectun) {
492 		struct ipsec_output_state state;
493 
494 		/*
495 		 * All the extension headers will become inaccessible
496 		 * (since they can be encrypted).
497 		 * Don't panic, we need no more updates to extension headers
498 		 * on inner IPv6 packet (since they are now encapsulated).
499 		 *
500 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
501 		 */
502 		bzero(&exthdrs, sizeof(exthdrs));
503 		exthdrs.ip6e_ip6 = m;
504 
505 		bzero(&state, sizeof(state));
506 		state.m = m;
507 		state.ro = (struct route *)ro;
508 		state.dst = (struct sockaddr *)dst;
509 
510 		error = ipsec6_output_tunnel(&state, sp, flags);
511 
512 		m = state.m;
513 		ro = (struct route_in6 *)state.ro;
514 		dst = (struct sockaddr_in6 *)state.dst;
515 		if (error) {
516 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
517 			m0 = m = NULL;
518 			m = NULL;
519 			switch (error) {
520 			case EHOSTUNREACH:
521 			case ENETUNREACH:
522 			case EMSGSIZE:
523 			case ENOBUFS:
524 			case ENOMEM:
525 				break;
526 			default:
527 				printf("ip6_output (ipsec): error code %d\n", error);
528 				/* fall through */
529 			case ENOENT:
530 				/* don't show these error codes to the user */
531 				error = 0;
532 				break;
533 			}
534 			goto bad;
535 		}
536 
537 		exthdrs.ip6e_ip6 = m;
538 	}
539 #endif /* IPSEC */
540 
541 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
542 		/* Unicast */
543 
544 #define ifatoia6(ifa)	((struct in6_ifaddr *)(ifa))
545 #define sin6tosa(sin6)	((struct sockaddr *)(sin6))
546 		/* xxx
547 		 * interface selection comes here
548 		 * if an interface is specified from an upper layer,
549 		 * ifp must point it.
550 		 */
551 		if (ro->ro_rt == 0) {
552 			/*
553 			 * non-bsdi always clone routes, if parent is
554 			 * PRF_CLONING.
555 			 */
556 			rtalloc((struct route *)ro);
557 		}
558 		if (ro->ro_rt == 0) {
559 			ip6stat.ip6s_noroute++;
560 			error = EHOSTUNREACH;
561 			/* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
562 			goto bad;
563 		}
564 		ia = ifatoia6(ro->ro_rt->rt_ifa);
565 		ifp = ro->ro_rt->rt_ifp;
566 		ro->ro_rt->rt_use++;
567 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
568 			dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
569 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
570 
571 		in6_ifstat_inc(ifp, ifs6_out_request);
572 
573 		/*
574 		 * Check if the outgoing interface conflicts with
575 		 * the interface specified by ifi6_ifindex (if specified).
576 		 * Note that loopback interface is always okay.
577 		 * (this may happen when we are sending a packet to one of
578 		 *  our own addresses.)
579 		 */
580 		if (opt && opt->ip6po_pktinfo
581 		 && opt->ip6po_pktinfo->ipi6_ifindex) {
582 			if (!(ifp->if_flags & IFF_LOOPBACK)
583 			 && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
584 				ip6stat.ip6s_noroute++;
585 				in6_ifstat_inc(ifp, ifs6_out_discard);
586 				error = EHOSTUNREACH;
587 				goto bad;
588 			}
589 		}
590 
591 		if (opt && opt->ip6po_hlim != -1)
592 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
593 	} else {
594 		/* Multicast */
595 		struct	in6_multi *in6m;
596 
597 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
598 
599 		/*
600 		 * See if the caller provided any multicast options
601 		 */
602 		ifp = NULL;
603 		if (im6o != NULL) {
604 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
605 			if (im6o->im6o_multicast_ifp != NULL)
606 				ifp = im6o->im6o_multicast_ifp;
607 		} else
608 			ip6->ip6_hlim = ip6_defmcasthlim;
609 
610 		/*
611 		 * See if the caller provided the outgoing interface
612 		 * as an ancillary data.
613 		 * Boundary check for ifindex is assumed to be already done.
614 		 */
615 		if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
616 			ifp = ifnet_byindex(opt->ip6po_pktinfo->ipi6_ifindex);
617 
618 		/*
619 		 * If the destination is a node-local scope multicast,
620 		 * the packet should be loop-backed only.
621 		 */
622 		if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
623 			/*
624 			 * If the outgoing interface is already specified,
625 			 * it should be a loopback interface.
626 			 */
627 			if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) {
628 				ip6stat.ip6s_badscope++;
629 				error = ENETUNREACH; /* XXX: better error? */
630 				/* XXX correct ifp? */
631 				in6_ifstat_inc(ifp, ifs6_out_discard);
632 				goto bad;
633 			} else {
634 				ifp = &loif[0];
635 			}
636 		}
637 
638 		if (opt && opt->ip6po_hlim != -1)
639 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
640 
641 		/*
642 		 * If caller did not provide an interface lookup a
643 		 * default in the routing table.  This is either a
644 		 * default for the speicfied group (i.e. a host
645 		 * route), or a multicast default (a route for the
646 		 * ``net'' ff00::/8).
647 		 */
648 		if (ifp == NULL) {
649 			if (ro->ro_rt == 0) {
650 				ro->ro_rt = rtalloc1((struct sockaddr *)
651 						&ro->ro_dst, 0, 0UL);
652 			}
653 			if (ro->ro_rt == 0) {
654 				ip6stat.ip6s_noroute++;
655 				error = EHOSTUNREACH;
656 				/* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
657 				goto bad;
658 			}
659 			ia = ifatoia6(ro->ro_rt->rt_ifa);
660 			ifp = ro->ro_rt->rt_ifp;
661 			ro->ro_rt->rt_use++;
662 		}
663 
664 		if ((flags & IPV6_FORWARDING) == 0)
665 			in6_ifstat_inc(ifp, ifs6_out_request);
666 		in6_ifstat_inc(ifp, ifs6_out_mcast);
667 
668 		/*
669 		 * Confirm that the outgoing interface supports multicast.
670 		 */
671 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
672 			ip6stat.ip6s_noroute++;
673 			in6_ifstat_inc(ifp, ifs6_out_discard);
674 			error = ENETUNREACH;
675 			goto bad;
676 		}
677 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
678 		if (in6m != NULL &&
679 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
680 			/*
681 			 * If we belong to the destination multicast group
682 			 * on the outgoing interface, and the caller did not
683 			 * forbid loopback, loop back a copy.
684 			 */
685 			ip6_mloopback(ifp, m, dst);
686 		} else {
687 			/*
688 			 * If we are acting as a multicast router, perform
689 			 * multicast forwarding as if the packet had just
690 			 * arrived on the interface to which we are about
691 			 * to send.  The multicast forwarding function
692 			 * recursively calls this function, using the
693 			 * IPV6_FORWARDING flag to prevent infinite recursion.
694 			 *
695 			 * Multicasts that are looped back by ip6_mloopback(),
696 			 * above, will be forwarded by the ip6_input() routine,
697 			 * if necessary.
698 			 */
699 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
700 				if (ip6_mforward(ip6, ifp, m) != 0) {
701 					m_freem(m);
702 					goto done;
703 				}
704 			}
705 		}
706 		/*
707 		 * Multicasts with a hoplimit of zero may be looped back,
708 		 * above, but must not be transmitted on a network.
709 		 * Also, multicasts addressed to the loopback interface
710 		 * are not sent -- the above call to ip6_mloopback() will
711 		 * loop back a copy if this host actually belongs to the
712 		 * destination group on the loopback interface.
713 		 */
714 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
715 			m_freem(m);
716 			goto done;
717 		}
718 	}
719 
720 	/*
721 	 * Fill the outgoing inteface to tell the upper layer
722 	 * to increment per-interface statistics.
723 	 */
724 	if (ifpp)
725 		*ifpp = ifp;
726 
727 	/*
728 	 * Determine path MTU.
729 	 */
730 	if (ro_pmtu != ro) {
731 		/* The first hop and the final destination may differ. */
732 		struct sockaddr_in6 *sin6_fin =
733 			(struct sockaddr_in6 *)&ro_pmtu->ro_dst;
734 		if (ro_pmtu->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
735 				       !IN6_ARE_ADDR_EQUAL(&sin6_fin->sin6_addr,
736 							   &finaldst))) {
737 			RTFREE(ro_pmtu->ro_rt);
738 			ro_pmtu->ro_rt = (struct rtentry *)0;
739 		}
740 		if (ro_pmtu->ro_rt == 0) {
741 			bzero(sin6_fin, sizeof(*sin6_fin));
742 			sin6_fin->sin6_family = AF_INET6;
743 			sin6_fin->sin6_len = sizeof(struct sockaddr_in6);
744 			sin6_fin->sin6_addr = finaldst;
745 
746 			rtalloc((struct route *)ro_pmtu);
747 		}
748 	}
749 	if (ro_pmtu->ro_rt != NULL) {
750 		u_int32_t ifmtu = nd_ifinfo[ifp->if_index].linkmtu;
751 
752 		mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
753 		if (mtu > ifmtu || mtu == 0) {
754 			/*
755 			 * The MTU on the route is larger than the MTU on
756 			 * the interface!  This shouldn't happen, unless the
757 			 * MTU of the interface has been changed after the
758 			 * interface was brought up.  Change the MTU in the
759 			 * route to match the interface MTU (as long as the
760 			 * field isn't locked).
761 			 *
762 			 * if MTU on the route is 0, we need to fix the MTU.
763 			 * this case happens with path MTU discovery timeouts.
764 			 */
765 			 mtu = ifmtu;
766 			 if ((ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
767 				 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; /* XXX */
768 		}
769 	} else {
770 		mtu = nd_ifinfo[ifp->if_index].linkmtu;
771 	}
772 
773 	/*
774 	 * advanced API (IPV6_USE_MIN_MTU) overrides mtu setting
775 	 */
776 	if ((flags & IPV6_MINMTU) != 0 && mtu > IPV6_MMTU)
777 		mtu = IPV6_MMTU;
778 
779 	/* Fake scoped addresses */
780 	if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
781 		/*
782 		 * If source or destination address is a scoped address, and
783 		 * the packet is going to be sent to a loopback interface,
784 		 * we should keep the original interface.
785 		 */
786 
787 		/*
788 		 * XXX: this is a very experimental and temporary solution.
789 		 * We eventually have sockaddr_in6 and use the sin6_scope_id
790 		 * field of the structure here.
791 		 * We rely on the consistency between two scope zone ids
792 		 * of source and destination, which should already be assured.
793 		 * Larger scopes than link will be supported in the future.
794 		 */
795 		origifp = NULL;
796 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
797 			origifp = ifnet_byindex(ntohs(ip6->ip6_src.s6_addr16[1]));
798 		else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
799 			origifp = ifnet_byindex(ntohs(ip6->ip6_dst.s6_addr16[1]));
800 		/*
801 		 * XXX: origifp can be NULL even in those two cases above.
802 		 * For example, if we remove the (only) link-local address
803 		 * from the loopback interface, and try to send a link-local
804 		 * address without link-id information.  Then the source
805 		 * address is ::1, and the destination address is the
806 		 * link-local address with its s6_addr16[1] being zero.
807 		 * What is worse, if the packet goes to the loopback interface
808 		 * by a default rejected route, the null pointer would be
809 		 * passed to looutput, and the kernel would hang.
810 		 * The following last resort would prevent such disaster.
811 		 */
812 		if (origifp == NULL)
813 			origifp = ifp;
814 	}
815 	else
816 		origifp = ifp;
817 #ifndef SCOPEDROUTING
818 	/*
819 	 * clear embedded scope identifiers if necessary.
820 	 * in6_clearscope will touch the addresses only when necessary.
821 	 */
822 	in6_clearscope(&ip6->ip6_src);
823 	in6_clearscope(&ip6->ip6_dst);
824 #endif
825 
826 	/*
827 	 * Check with the firewall...
828 	 */
829         if (ip6_fw_enable && ip6_fw_chk_ptr) {
830 		u_short port = 0;
831 		m->m_pkthdr.rcvif = NULL;	/* XXX */
832 		/* If ipfw says divert, we have to just drop packet */
833 		if ((*ip6_fw_chk_ptr)(&ip6, ifp, &port, &m)) {
834 			m_freem(m);
835 			goto done;
836 		}
837 		if (!m) {
838 			error = EACCES;
839 			goto done;
840 		}
841 	}
842 
843 	/*
844 	 * If the outgoing packet contains a hop-by-hop options header,
845 	 * it must be examined and processed even by the source node.
846 	 * (RFC 2460, section 4.)
847 	 */
848 	if (exthdrs.ip6e_hbh) {
849 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
850 		u_int32_t dummy1; /* XXX unused */
851 		u_int32_t dummy2; /* XXX unused */
852 
853 #ifdef DIAGNOSTIC
854 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
855 			panic("ip6e_hbh is not continuous");
856 #endif
857 		/*
858 		 *  XXX: if we have to send an ICMPv6 error to the sender,
859 		 *       we need the M_LOOP flag since icmp6_error() expects
860 		 *       the IPv6 and the hop-by-hop options header are
861 		 *       continuous unless the flag is set.
862 		 */
863 		m->m_flags |= M_LOOP;
864 		m->m_pkthdr.rcvif = ifp;
865 		if (ip6_process_hopopts(m,
866 					(u_int8_t *)(hbh + 1),
867 					((hbh->ip6h_len + 1) << 3) -
868 					sizeof(struct ip6_hbh),
869 					&dummy1, &dummy2) < 0) {
870 			/* m was already freed at this point */
871 			error = EINVAL;/* better error? */
872 			goto done;
873 		}
874 		m->m_flags &= ~M_LOOP; /* XXX */
875 		m->m_pkthdr.rcvif = NULL;
876 	}
877 
878 #ifdef PFIL_HOOKS
879 	/*
880 	 * Run through list of hooks for output packets.
881 	 */
882 	m1 = m;
883 	pfh = pfil_hook_get(PFIL_OUT, &inet6sw[ip6_protox[IPPROTO_IPV6]].pr_pfh);
884 	for (; pfh; pfh = pfh->pfil_link.tqe_next)
885 		if (pfh->pfil_func) {
886 			rv = pfh->pfil_func(ip6, sizeof(*ip6), ifp, 1, &m1);
887 			if (rv) {
888 				error = EHOSTUNREACH;
889 				goto done;
890 			}
891 			m = m1;
892 			if (m == NULL)
893 				goto done;
894 			ip6 = mtod(m, struct ip6_hdr *);
895 		}
896 #endif /* PFIL_HOOKS */
897 	/*
898 	 * Send the packet to the outgoing interface.
899 	 * If necessary, do IPv6 fragmentation before sending.
900 	 */
901 	tlen = m->m_pkthdr.len;
902 	if (tlen <= mtu
903 #ifdef notyet
904 	    /*
905 	     * On any link that cannot convey a 1280-octet packet in one piece,
906 	     * link-specific fragmentation and reassembly must be provided at
907 	     * a layer below IPv6. [RFC 2460, sec.5]
908 	     * Thus if the interface has ability of link-level fragmentation,
909 	     * we can just send the packet even if the packet size is
910 	     * larger than the link's MTU.
911 	     * XXX: IFF_FRAGMENTABLE (or such) flag has not been defined yet...
912 	     */
913 
914 	    || ifp->if_flags & IFF_FRAGMENTABLE
915 #endif
916 	    )
917 	{
918  		/* Record statistics for this interface address. */
919  		if (ia && !(flags & IPV6_FORWARDING)) {
920  			ia->ia_ifa.if_opackets++;
921  			ia->ia_ifa.if_obytes += m->m_pkthdr.len;
922  		}
923 #ifdef IPSEC
924 		/* clean ipsec history once it goes out of the node */
925 		ipsec_delaux(m);
926 #endif
927 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
928 		goto done;
929 	} else if (mtu < IPV6_MMTU) {
930 		/*
931 		 * note that path MTU is never less than IPV6_MMTU
932 		 * (see icmp6_input).
933 		 */
934 		error = EMSGSIZE;
935 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
936 		goto bad;
937 	} else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
938 		error = EMSGSIZE;
939 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
940 		goto bad;
941 	} else {
942 		struct mbuf **mnext, *m_frgpart;
943 		struct ip6_frag *ip6f;
944 		u_int32_t id = htonl(ip6_id++);
945 		u_char nextproto;
946 
947 		/*
948 		 * Too large for the destination or interface;
949 		 * fragment if possible.
950 		 * Must be able to put at least 8 bytes per fragment.
951 		 */
952 		hlen = unfragpartlen;
953 		if (mtu > IPV6_MAXPACKET)
954 			mtu = IPV6_MAXPACKET;
955 
956 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
957 		if (len < 8) {
958 			error = EMSGSIZE;
959 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
960 			goto bad;
961 		}
962 
963 		mnext = &m->m_nextpkt;
964 
965 		/*
966 		 * Change the next header field of the last header in the
967 		 * unfragmentable part.
968 		 */
969 		if (exthdrs.ip6e_rthdr) {
970 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
971 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
972 		} else if (exthdrs.ip6e_dest1) {
973 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
974 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
975 		} else if (exthdrs.ip6e_hbh) {
976 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
977 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
978 		} else {
979 			nextproto = ip6->ip6_nxt;
980 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
981 		}
982 
983 		/*
984 		 * Loop through length of segment after first fragment,
985 		 * make new header and copy data of each part and link onto
986 		 * chain.
987 		 */
988 		m0 = m;
989 		for (off = hlen; off < tlen; off += len) {
990 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
991 			if (!m) {
992 				error = ENOBUFS;
993 				ip6stat.ip6s_odropped++;
994 				goto sendorfree;
995 			}
996 			m->m_pkthdr.rcvif = NULL;
997 			m->m_flags = m0->m_flags & M_COPYFLAGS;
998 			*mnext = m;
999 			mnext = &m->m_nextpkt;
1000 			m->m_data += max_linkhdr;
1001 			mhip6 = mtod(m, struct ip6_hdr *);
1002 			*mhip6 = *ip6;
1003 			m->m_len = sizeof(*mhip6);
1004  			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1005  			if (error) {
1006 				ip6stat.ip6s_odropped++;
1007 				goto sendorfree;
1008 			}
1009 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1010 			if (off + len >= tlen)
1011 				len = tlen - off;
1012 			else
1013 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1014 			mhip6->ip6_plen = htons((u_short)(len + hlen +
1015 							  sizeof(*ip6f) -
1016 							  sizeof(struct ip6_hdr)));
1017 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1018 				error = ENOBUFS;
1019 				ip6stat.ip6s_odropped++;
1020 				goto sendorfree;
1021 			}
1022 			m_cat(m, m_frgpart);
1023 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1024 			m->m_pkthdr.rcvif = (struct ifnet *)0;
1025 			ip6f->ip6f_reserved = 0;
1026 			ip6f->ip6f_ident = id;
1027 			ip6f->ip6f_nxt = nextproto;
1028 			ip6stat.ip6s_ofragments++;
1029 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1030 		}
1031 
1032 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1033 	}
1034 
1035 	/*
1036 	 * Remove leading garbages.
1037 	 */
1038 sendorfree:
1039 	m = m0->m_nextpkt;
1040 	m0->m_nextpkt = 0;
1041 	m_freem(m0);
1042 	for (m0 = m; m; m = m0) {
1043 		m0 = m->m_nextpkt;
1044 		m->m_nextpkt = 0;
1045 		if (error == 0) {
1046  			/* Record statistics for this interface address. */
1047  			if (ia) {
1048  				ia->ia_ifa.if_opackets++;
1049  				ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1050  			}
1051 #ifdef IPSEC
1052 			/* clean ipsec history once it goes out of the node */
1053 			ipsec_delaux(m);
1054 #endif
1055 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1056 		} else
1057 			m_freem(m);
1058 	}
1059 
1060 	if (error == 0)
1061 		ip6stat.ip6s_fragmented++;
1062 
1063 done:
1064 	if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1065 		RTFREE(ro->ro_rt);
1066 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1067 		RTFREE(ro_pmtu->ro_rt);
1068 	}
1069 
1070 #ifdef IPSEC
1071 	if (sp != NULL)
1072 		key_freesp(sp);
1073 #endif /* IPSEC */
1074 
1075 	return(error);
1076 
1077 freehdrs:
1078 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1079 	m_freem(exthdrs.ip6e_dest1);
1080 	m_freem(exthdrs.ip6e_rthdr);
1081 	m_freem(exthdrs.ip6e_dest2);
1082 	/* fall through */
1083 bad:
1084 	m_freem(m);
1085 	goto done;
1086 }
1087 
1088 static int
1089 ip6_copyexthdr(mp, hdr, hlen)
1090 	struct mbuf **mp;
1091 	caddr_t hdr;
1092 	int hlen;
1093 {
1094 	struct mbuf *m;
1095 
1096 	if (hlen > MCLBYTES)
1097 		return(ENOBUFS); /* XXX */
1098 
1099 	MGET(m, M_DONTWAIT, MT_DATA);
1100 	if (!m)
1101 		return(ENOBUFS);
1102 
1103 	if (hlen > MLEN) {
1104 		MCLGET(m, M_DONTWAIT);
1105 		if ((m->m_flags & M_EXT) == 0) {
1106 			m_free(m);
1107 			return(ENOBUFS);
1108 		}
1109 	}
1110 	m->m_len = hlen;
1111 	if (hdr)
1112 		bcopy(hdr, mtod(m, caddr_t), hlen);
1113 
1114 	*mp = m;
1115 	return(0);
1116 }
1117 
1118 /*
1119  * Insert jumbo payload option.
1120  */
1121 static int
1122 ip6_insert_jumboopt(exthdrs, plen)
1123 	struct ip6_exthdrs *exthdrs;
1124 	u_int32_t plen;
1125 {
1126 	struct mbuf *mopt;
1127 	u_char *optbuf;
1128 	u_int32_t v;
1129 
1130 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1131 
1132 	/*
1133 	 * If there is no hop-by-hop options header, allocate new one.
1134 	 * If there is one but it doesn't have enough space to store the
1135 	 * jumbo payload option, allocate a cluster to store the whole options.
1136 	 * Otherwise, use it to store the options.
1137 	 */
1138 	if (exthdrs->ip6e_hbh == 0) {
1139 		MGET(mopt, M_DONTWAIT, MT_DATA);
1140 		if (mopt == 0)
1141 			return(ENOBUFS);
1142 		mopt->m_len = JUMBOOPTLEN;
1143 		optbuf = mtod(mopt, u_char *);
1144 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1145 		exthdrs->ip6e_hbh = mopt;
1146 	} else {
1147 		struct ip6_hbh *hbh;
1148 
1149 		mopt = exthdrs->ip6e_hbh;
1150 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1151 			/*
1152 			 * XXX assumption:
1153 			 * - exthdrs->ip6e_hbh is not referenced from places
1154 			 *   other than exthdrs.
1155 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1156 			 */
1157 			int oldoptlen = mopt->m_len;
1158 			struct mbuf *n;
1159 
1160 			/*
1161 			 * XXX: give up if the whole (new) hbh header does
1162 			 * not fit even in an mbuf cluster.
1163 			 */
1164 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1165 				return(ENOBUFS);
1166 
1167 			/*
1168 			 * As a consequence, we must always prepare a cluster
1169 			 * at this point.
1170 			 */
1171 			MGET(n, M_DONTWAIT, MT_DATA);
1172 			if (n) {
1173 				MCLGET(n, M_DONTWAIT);
1174 				if ((n->m_flags & M_EXT) == 0) {
1175 					m_freem(n);
1176 					n = NULL;
1177 				}
1178 			}
1179 			if (!n)
1180 				return(ENOBUFS);
1181 			n->m_len = oldoptlen + JUMBOOPTLEN;
1182 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1183 			      oldoptlen);
1184 			optbuf = mtod(n, caddr_t) + oldoptlen;
1185 			m_freem(mopt);
1186 			mopt = exthdrs->ip6e_hbh = n;
1187 		} else {
1188 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1189 			mopt->m_len += JUMBOOPTLEN;
1190 		}
1191 		optbuf[0] = IP6OPT_PADN;
1192 		optbuf[1] = 1;
1193 
1194 		/*
1195 		 * Adjust the header length according to the pad and
1196 		 * the jumbo payload option.
1197 		 */
1198 		hbh = mtod(mopt, struct ip6_hbh *);
1199 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1200 	}
1201 
1202 	/* fill in the option. */
1203 	optbuf[2] = IP6OPT_JUMBO;
1204 	optbuf[3] = 4;
1205 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1206 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1207 
1208 	/* finally, adjust the packet header length */
1209 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1210 
1211 	return(0);
1212 #undef JUMBOOPTLEN
1213 }
1214 
1215 /*
1216  * Insert fragment header and copy unfragmentable header portions.
1217  */
1218 static int
1219 ip6_insertfraghdr(m0, m, hlen, frghdrp)
1220 	struct mbuf *m0, *m;
1221 	int hlen;
1222 	struct ip6_frag **frghdrp;
1223 {
1224 	struct mbuf *n, *mlast;
1225 
1226 	if (hlen > sizeof(struct ip6_hdr)) {
1227 		n = m_copym(m0, sizeof(struct ip6_hdr),
1228 			    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1229 		if (n == 0)
1230 			return(ENOBUFS);
1231 		m->m_next = n;
1232 	} else
1233 		n = m;
1234 
1235 	/* Search for the last mbuf of unfragmentable part. */
1236 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1237 		;
1238 
1239 	if ((mlast->m_flags & M_EXT) == 0 &&
1240 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1241 		/* use the trailing space of the last mbuf for the fragment hdr */
1242 		*frghdrp =
1243 			(struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len);
1244 		mlast->m_len += sizeof(struct ip6_frag);
1245 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1246 	} else {
1247 		/* allocate a new mbuf for the fragment header */
1248 		struct mbuf *mfrg;
1249 
1250 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1251 		if (mfrg == 0)
1252 			return(ENOBUFS);
1253 		mfrg->m_len = sizeof(struct ip6_frag);
1254 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1255 		mlast->m_next = mfrg;
1256 	}
1257 
1258 	return(0);
1259 }
1260 
1261 /*
1262  * IP6 socket option processing.
1263  */
1264 int
1265 ip6_ctloutput(so, sopt)
1266 	struct socket *so;
1267 	struct sockopt *sopt;
1268 {
1269 	int privileged;
1270 	struct inpcb *in6p = sotoinpcb(so);
1271 	int error, optval;
1272 	int level, op, optname;
1273 	int optlen;
1274 	struct thread *td;
1275 
1276 	if (sopt) {
1277 		level = sopt->sopt_level;
1278 		op = sopt->sopt_dir;
1279 		optname = sopt->sopt_name;
1280 		optlen = sopt->sopt_valsize;
1281 		td = sopt->sopt_td;
1282 	} else {
1283 		panic("ip6_ctloutput: arg soopt is NULL");
1284 	}
1285 	error = optval = 0;
1286 
1287 	privileged = (td == 0 || suser(td)) ? 0 : 1;
1288 
1289 	if (level == IPPROTO_IPV6) {
1290 		switch (op) {
1291 
1292 		case SOPT_SET:
1293 			switch (optname) {
1294 			case IPV6_PKTOPTIONS:
1295 			{
1296 				struct mbuf *m;
1297 
1298 				error = soopt_getm(sopt, &m); /* XXX */
1299 				if (error != NULL)
1300 					break;
1301 				error = soopt_mcopyin(sopt, m); /* XXX */
1302 				if (error != NULL)
1303 					break;
1304 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1305 						    m, so, sopt);
1306 				m_freem(m); /* XXX */
1307 				break;
1308 			}
1309 
1310 			/*
1311 			 * Use of some Hop-by-Hop options or some
1312 			 * Destination options, might require special
1313 			 * privilege.  That is, normal applications
1314 			 * (without special privilege) might be forbidden
1315 			 * from setting certain options in outgoing packets,
1316 			 * and might never see certain options in received
1317 			 * packets. [RFC 2292 Section 6]
1318 			 * KAME specific note:
1319 			 *  KAME prevents non-privileged users from sending or
1320 			 *  receiving ANY hbh/dst options in order to avoid
1321 			 *  overhead of parsing options in the kernel.
1322 			 */
1323 			case IPV6_UNICAST_HOPS:
1324 			case IPV6_CHECKSUM:
1325 			case IPV6_FAITH:
1326 
1327 			case IPV6_V6ONLY:
1328 				if (optlen != sizeof(int)) {
1329 					error = EINVAL;
1330 					break;
1331 				}
1332 				error = sooptcopyin(sopt, &optval,
1333 					sizeof optval, sizeof optval);
1334 				if (error)
1335 					break;
1336 				switch (optname) {
1337 
1338 				case IPV6_UNICAST_HOPS:
1339 					if (optval < -1 || optval >= 256)
1340 						error = EINVAL;
1341 					else {
1342 						/* -1 = kernel default */
1343 						in6p->in6p_hops = optval;
1344 
1345 						if ((in6p->in6p_vflag &
1346 						     INP_IPV4) != 0)
1347 							in6p->inp_ip_ttl = optval;
1348 					}
1349 					break;
1350 #define OPTSET(bit) \
1351 do { \
1352 	if (optval) \
1353 		in6p->in6p_flags |= (bit); \
1354 	else \
1355 		in6p->in6p_flags &= ~(bit); \
1356 } while (0)
1357 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1358 
1359 				case IPV6_CHECKSUM:
1360 					in6p->in6p_cksum = optval;
1361 					break;
1362 
1363 				case IPV6_FAITH:
1364 					OPTSET(IN6P_FAITH);
1365 					break;
1366 
1367 				case IPV6_V6ONLY:
1368 					/*
1369 					 * make setsockopt(IPV6_V6ONLY)
1370 					 * available only prior to bind(2).
1371 					 * see ipng mailing list, Jun 22 2001.
1372 					 */
1373 					if (in6p->in6p_lport ||
1374 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
1375 					{
1376 						error = EINVAL;
1377 						break;
1378 					}
1379 					/*
1380 					 * XXX: BINDV6ONLY should be integrated
1381 					 * into V6ONLY.
1382 					 */
1383 					OPTSET(IN6P_BINDV6ONLY);
1384 					OPTSET(IN6P_IPV6_V6ONLY);
1385 					break;
1386 				}
1387 				break;
1388 
1389 			case IPV6_PKTINFO:
1390 			case IPV6_HOPLIMIT:
1391 			case IPV6_HOPOPTS:
1392 			case IPV6_DSTOPTS:
1393 			case IPV6_RTHDR:
1394 				/* RFC 2292 */
1395 				if (optlen != sizeof(int)) {
1396 					error = EINVAL;
1397 					break;
1398 				}
1399 				error = sooptcopyin(sopt, &optval,
1400 					sizeof optval, sizeof optval);
1401 				if (error)
1402 					break;
1403 				switch (optname) {
1404 				case IPV6_PKTINFO:
1405 					OPTSET(IN6P_PKTINFO);
1406 					break;
1407 				case IPV6_HOPLIMIT:
1408 					OPTSET(IN6P_HOPLIMIT);
1409 					break;
1410 				case IPV6_HOPOPTS:
1411 					/*
1412 					 * Check super-user privilege.
1413 					 * See comments for IPV6_RECVHOPOPTS.
1414 					 */
1415 					if (!privileged)
1416 						return(EPERM);
1417 					OPTSET(IN6P_HOPOPTS);
1418 					break;
1419 				case IPV6_DSTOPTS:
1420 					if (!privileged)
1421 						return(EPERM);
1422 					OPTSET(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1423 					break;
1424 				case IPV6_RTHDR:
1425 					OPTSET(IN6P_RTHDR);
1426 					break;
1427 				}
1428 				break;
1429 #undef OPTSET
1430 
1431 			case IPV6_MULTICAST_IF:
1432 			case IPV6_MULTICAST_HOPS:
1433 			case IPV6_MULTICAST_LOOP:
1434 			case IPV6_JOIN_GROUP:
1435 			case IPV6_LEAVE_GROUP:
1436 			    {
1437 				struct mbuf *m;
1438 				if (sopt->sopt_valsize > MLEN) {
1439 					error = EMSGSIZE;
1440 					break;
1441 				}
1442 				/* XXX */
1443 				MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_HEADER);
1444 				if (m == 0) {
1445 					error = ENOBUFS;
1446 					break;
1447 				}
1448 				m->m_len = sopt->sopt_valsize;
1449 				error = sooptcopyin(sopt, mtod(m, char *),
1450 						    m->m_len, m->m_len);
1451 				error =	ip6_setmoptions(sopt->sopt_name,
1452 							&in6p->in6p_moptions,
1453 							m);
1454 				(void)m_free(m);
1455 			    }
1456 				break;
1457 
1458 			case IPV6_PORTRANGE:
1459 				error = sooptcopyin(sopt, &optval,
1460 				    sizeof optval, sizeof optval);
1461 				if (error)
1462 					break;
1463 
1464 				switch (optval) {
1465 				case IPV6_PORTRANGE_DEFAULT:
1466 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1467 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1468 					break;
1469 
1470 				case IPV6_PORTRANGE_HIGH:
1471 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1472 					in6p->in6p_flags |= IN6P_HIGHPORT;
1473 					break;
1474 
1475 				case IPV6_PORTRANGE_LOW:
1476 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1477 					in6p->in6p_flags |= IN6P_LOWPORT;
1478 					break;
1479 
1480 				default:
1481 					error = EINVAL;
1482 					break;
1483 				}
1484 				break;
1485 
1486 #ifdef IPSEC
1487 			case IPV6_IPSEC_POLICY:
1488 			    {
1489 				caddr_t req = NULL;
1490 				size_t len = 0;
1491 				struct mbuf *m;
1492 
1493 				if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1494 					break;
1495 				if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1496 					break;
1497 				if (m) {
1498 					req = mtod(m, caddr_t);
1499 					len = m->m_len;
1500 				}
1501 				error = ipsec6_set_policy(in6p, optname, req,
1502 				                          len, privileged);
1503 				m_freem(m);
1504 			    }
1505 				break;
1506 #endif /* KAME IPSEC */
1507 
1508 			case IPV6_FW_ADD:
1509 			case IPV6_FW_DEL:
1510 			case IPV6_FW_FLUSH:
1511 			case IPV6_FW_ZERO:
1512 			    {
1513 				struct mbuf *m;
1514 				struct mbuf **mp = &m;
1515 
1516 				if (ip6_fw_ctl_ptr == NULL)
1517 					return EINVAL;
1518 				/* XXX */
1519 				if ((error = soopt_getm(sopt, &m)) != 0)
1520 					break;
1521 				/* XXX */
1522 				if ((error = soopt_mcopyin(sopt, m)) != 0)
1523 					break;
1524 				error = (*ip6_fw_ctl_ptr)(optname, mp);
1525 				m = *mp;
1526 			    }
1527 				break;
1528 
1529 			default:
1530 				error = ENOPROTOOPT;
1531 				break;
1532 			}
1533 			break;
1534 
1535 		case SOPT_GET:
1536 			switch (optname) {
1537 
1538 			case IPV6_PKTOPTIONS:
1539 				if (in6p->in6p_options) {
1540 					struct mbuf *m;
1541 					m = m_copym(in6p->in6p_options,
1542 					    0, M_COPYALL, M_TRYWAIT);
1543 					error = soopt_mcopyout(sopt, m);
1544 					if (error == 0)
1545 						m_freem(m);
1546 				} else
1547 					sopt->sopt_valsize = 0;
1548 				break;
1549 
1550 			case IPV6_UNICAST_HOPS:
1551 			case IPV6_CHECKSUM:
1552 
1553 			case IPV6_FAITH:
1554 			case IPV6_V6ONLY:
1555 			case IPV6_PORTRANGE:
1556 				switch (optname) {
1557 
1558 				case IPV6_UNICAST_HOPS:
1559 					optval = in6p->in6p_hops;
1560 					break;
1561 
1562 				case IPV6_CHECKSUM:
1563 					optval = in6p->in6p_cksum;
1564 					break;
1565 
1566 				case IPV6_FAITH:
1567 					optval = OPTBIT(IN6P_FAITH);
1568 					break;
1569 
1570 				case IPV6_V6ONLY:
1571 					/* XXX: see the setopt case. */
1572 					optval = OPTBIT(IN6P_BINDV6ONLY);
1573 					break;
1574 
1575 				case IPV6_PORTRANGE:
1576 				    {
1577 					int flags;
1578 					flags = in6p->in6p_flags;
1579 					if (flags & IN6P_HIGHPORT)
1580 						optval = IPV6_PORTRANGE_HIGH;
1581 					else if (flags & IN6P_LOWPORT)
1582 						optval = IPV6_PORTRANGE_LOW;
1583 					else
1584 						optval = 0;
1585 					break;
1586 				    }
1587 				}
1588 				error = sooptcopyout(sopt, &optval,
1589 					sizeof optval);
1590 				break;
1591 
1592 			case IPV6_PKTINFO:
1593 			case IPV6_HOPLIMIT:
1594 			case IPV6_HOPOPTS:
1595 			case IPV6_RTHDR:
1596 			case IPV6_DSTOPTS:
1597 				if (optname == IPV6_HOPOPTS ||
1598 				    optname == IPV6_DSTOPTS ||
1599 				    !privileged)
1600 					return(EPERM);
1601 				switch (optname) {
1602 				case IPV6_PKTINFO:
1603 					optval = OPTBIT(IN6P_PKTINFO);
1604 					break;
1605 				case IPV6_HOPLIMIT:
1606 					optval = OPTBIT(IN6P_HOPLIMIT);
1607 					break;
1608 				case IPV6_HOPOPTS:
1609 					if (!privileged)
1610 						return(EPERM);
1611 					optval = OPTBIT(IN6P_HOPOPTS);
1612 					break;
1613 				case IPV6_RTHDR:
1614 					optval = OPTBIT(IN6P_RTHDR);
1615 					break;
1616 				case IPV6_DSTOPTS:
1617 					if (!privileged)
1618 						return(EPERM);
1619 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1620 					break;
1621 				}
1622 				error = sooptcopyout(sopt, &optval,
1623 					sizeof optval);
1624 				break;
1625 
1626 			case IPV6_MULTICAST_IF:
1627 			case IPV6_MULTICAST_HOPS:
1628 			case IPV6_MULTICAST_LOOP:
1629 			case IPV6_JOIN_GROUP:
1630 			case IPV6_LEAVE_GROUP:
1631 			    {
1632 				struct mbuf *m;
1633 				error = ip6_getmoptions(sopt->sopt_name,
1634 						in6p->in6p_moptions, &m);
1635 				if (error == 0)
1636 					error = sooptcopyout(sopt,
1637 						mtod(m, char *), m->m_len);
1638 				m_freem(m);
1639 			    }
1640 				break;
1641 
1642 #ifdef IPSEC
1643 			case IPV6_IPSEC_POLICY:
1644 			  {
1645 				caddr_t req = NULL;
1646 				size_t len = 0;
1647 				struct mbuf *m = NULL;
1648 				struct mbuf **mp = &m;
1649 
1650 				error = soopt_getm(sopt, &m); /* XXX */
1651 				if (error != NULL)
1652 					break;
1653 				error = soopt_mcopyin(sopt, m); /* XXX */
1654 				if (error != NULL)
1655 					break;
1656 				if (m) {
1657 					req = mtod(m, caddr_t);
1658 					len = m->m_len;
1659 				}
1660 				error = ipsec6_get_policy(in6p, req, len, mp);
1661 				if (error == 0)
1662 					error = soopt_mcopyout(sopt, m); /*XXX*/
1663 				if (error == 0 && m)
1664 					m_freem(m);
1665 				break;
1666 			  }
1667 #endif /* KAME IPSEC */
1668 
1669 			case IPV6_FW_GET:
1670 			  {
1671 				struct mbuf *m;
1672 				struct mbuf **mp = &m;
1673 
1674 				if (ip6_fw_ctl_ptr == NULL)
1675 			        {
1676 					return EINVAL;
1677 				}
1678 				error = (*ip6_fw_ctl_ptr)(optname, mp);
1679 				if (error == 0)
1680 					error = soopt_mcopyout(sopt, m); /* XXX */
1681 				if (error == 0 && m)
1682 					m_freem(m);
1683 			  }
1684 				break;
1685 
1686 			default:
1687 				error = ENOPROTOOPT;
1688 				break;
1689 			}
1690 			break;
1691 		}
1692 	} else {
1693 		error = EINVAL;
1694 	}
1695 	return(error);
1696 }
1697 
1698 /*
1699  * Set up IP6 options in pcb for insertion in output packets or
1700  * specifying behavior of outgoing packets.
1701  */
1702 static int
1703 ip6_pcbopts(pktopt, m, so, sopt)
1704 	struct ip6_pktopts **pktopt;
1705 	struct mbuf *m;
1706 	struct socket *so;
1707 	struct sockopt *sopt;
1708 {
1709 	struct ip6_pktopts *opt = *pktopt;
1710 	int error = 0;
1711 	struct thread *td = sopt->sopt_td;
1712 	int priv = 0;
1713 
1714 	/* turn off any old options. */
1715 	if (opt) {
1716 #ifdef DIAGNOSTIC
1717 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
1718 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
1719 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
1720 			printf("ip6_pcbopts: all specified options are cleared.\n");
1721 #endif
1722 		ip6_clearpktopts(opt, 1, -1);
1723 	} else
1724 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
1725 	*pktopt = NULL;
1726 
1727 	if (!m || m->m_len == 0) {
1728 		/*
1729 		 * Only turning off any previous options.
1730 		 */
1731 		if (opt)
1732 			free(opt, M_IP6OPT);
1733 		return(0);
1734 	}
1735 
1736 	/*  set options specified by user. */
1737 	if (td && !suser(td))
1738 		priv = 1;
1739 	if ((error = ip6_setpktoptions(m, opt, priv, 1)) != 0) {
1740 		ip6_clearpktopts(opt, 1, -1); /* XXX: discard all options */
1741 		return(error);
1742 	}
1743 	*pktopt = opt;
1744 	return(0);
1745 }
1746 
1747 /*
1748  * initialize ip6_pktopts.  beware that there are non-zero default values in
1749  * the struct.
1750  */
1751 void
1752 init_ip6pktopts(opt)
1753 	struct ip6_pktopts *opt;
1754 {
1755 
1756 	bzero(opt, sizeof(*opt));
1757 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
1758 }
1759 
1760 void
1761 ip6_clearpktopts(pktopt, needfree, optname)
1762 	struct ip6_pktopts *pktopt;
1763 	int needfree, optname;
1764 {
1765 	if (pktopt == NULL)
1766 		return;
1767 
1768 	if (optname == -1) {
1769 		if (needfree && pktopt->ip6po_pktinfo)
1770 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
1771 		pktopt->ip6po_pktinfo = NULL;
1772 	}
1773 	if (optname == -1)
1774 		pktopt->ip6po_hlim = -1;
1775 	if (optname == -1) {
1776 		if (needfree && pktopt->ip6po_nexthop)
1777 			free(pktopt->ip6po_nexthop, M_IP6OPT);
1778 		pktopt->ip6po_nexthop = NULL;
1779 	}
1780 	if (optname == -1) {
1781 		if (needfree && pktopt->ip6po_hbh)
1782 			free(pktopt->ip6po_hbh, M_IP6OPT);
1783 		pktopt->ip6po_hbh = NULL;
1784 	}
1785 	if (optname == -1) {
1786 		if (needfree && pktopt->ip6po_dest1)
1787 			free(pktopt->ip6po_dest1, M_IP6OPT);
1788 		pktopt->ip6po_dest1 = NULL;
1789 	}
1790 	if (optname == -1) {
1791 		if (needfree && pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
1792 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
1793 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
1794 		if (pktopt->ip6po_route.ro_rt) {
1795 			RTFREE(pktopt->ip6po_route.ro_rt);
1796 			pktopt->ip6po_route.ro_rt = NULL;
1797 		}
1798 	}
1799 	if (optname == -1) {
1800 		if (needfree && pktopt->ip6po_dest2)
1801 			free(pktopt->ip6po_dest2, M_IP6OPT);
1802 		pktopt->ip6po_dest2 = NULL;
1803 	}
1804 }
1805 
1806 #define PKTOPT_EXTHDRCPY(type) \
1807 do {\
1808 	if (src->type) {\
1809 		int hlen =\
1810 			(((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
1811 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
1812 		if (dst->type == NULL && canwait == M_NOWAIT)\
1813 			goto bad;\
1814 		bcopy(src->type, dst->type, hlen);\
1815 	}\
1816 } while (0)
1817 
1818 struct ip6_pktopts *
1819 ip6_copypktopts(src, canwait)
1820 	struct ip6_pktopts *src;
1821 	int canwait;
1822 {
1823 	struct ip6_pktopts *dst;
1824 
1825 	if (src == NULL) {
1826 		printf("ip6_clearpktopts: invalid argument\n");
1827 		return(NULL);
1828 	}
1829 
1830 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
1831 	if (dst == NULL && canwait == M_NOWAIT)
1832 		goto bad;
1833 	bzero(dst, sizeof(*dst));
1834 
1835 	dst->ip6po_hlim = src->ip6po_hlim;
1836 	if (src->ip6po_pktinfo) {
1837 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
1838 					    M_IP6OPT, canwait);
1839 		if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT)
1840 			goto bad;
1841 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
1842 	}
1843 	if (src->ip6po_nexthop) {
1844 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
1845 					    M_IP6OPT, canwait);
1846 		if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT)
1847 			goto bad;
1848 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
1849 		      src->ip6po_nexthop->sa_len);
1850 	}
1851 	PKTOPT_EXTHDRCPY(ip6po_hbh);
1852 	PKTOPT_EXTHDRCPY(ip6po_dest1);
1853 	PKTOPT_EXTHDRCPY(ip6po_dest2);
1854 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
1855 	return(dst);
1856 
1857   bad:
1858 	printf("ip6_copypktopts: copy failed");
1859 	if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
1860 	if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
1861 	if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
1862 	if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
1863 	if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
1864 	if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
1865 	return(NULL);
1866 }
1867 #undef PKTOPT_EXTHDRCPY
1868 
1869 void
1870 ip6_freepcbopts(pktopt)
1871 	struct ip6_pktopts *pktopt;
1872 {
1873 	if (pktopt == NULL)
1874 		return;
1875 
1876 	ip6_clearpktopts(pktopt, 1, -1);
1877 
1878 	free(pktopt, M_IP6OPT);
1879 }
1880 
1881 /*
1882  * Set the IP6 multicast options in response to user setsockopt().
1883  */
1884 static int
1885 ip6_setmoptions(optname, im6op, m)
1886 	int optname;
1887 	struct ip6_moptions **im6op;
1888 	struct mbuf *m;
1889 {
1890 	int error = 0;
1891 	u_int loop, ifindex;
1892 	struct ipv6_mreq *mreq;
1893 	struct ifnet *ifp;
1894 	struct ip6_moptions *im6o = *im6op;
1895 	struct route_in6 ro;
1896 	struct sockaddr_in6 *dst;
1897 	struct in6_multi_mship *imm;
1898 	struct thread *td = curthread;	/* XXX */
1899 
1900 	if (im6o == NULL) {
1901 		/*
1902 		 * No multicast option buffer attached to the pcb;
1903 		 * allocate one and initialize to default values.
1904 		 */
1905 		im6o = (struct ip6_moptions *)
1906 			malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
1907 
1908 		if (im6o == NULL)
1909 			return(ENOBUFS);
1910 		*im6op = im6o;
1911 		im6o->im6o_multicast_ifp = NULL;
1912 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1913 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
1914 		LIST_INIT(&im6o->im6o_memberships);
1915 	}
1916 
1917 	switch (optname) {
1918 
1919 	case IPV6_MULTICAST_IF:
1920 		/*
1921 		 * Select the interface for outgoing multicast packets.
1922 		 */
1923 		if (m == NULL || m->m_len != sizeof(u_int)) {
1924 			error = EINVAL;
1925 			break;
1926 		}
1927 		bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
1928 		if (ifindex < 0 || if_index < ifindex) {
1929 			error = ENXIO;	/* XXX EINVAL? */
1930 			break;
1931 		}
1932 		ifp = ifnet_byindex(ifindex);
1933 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1934 			error = EADDRNOTAVAIL;
1935 			break;
1936 		}
1937 		im6o->im6o_multicast_ifp = ifp;
1938 		break;
1939 
1940 	case IPV6_MULTICAST_HOPS:
1941 	    {
1942 		/*
1943 		 * Set the IP6 hoplimit for outgoing multicast packets.
1944 		 */
1945 		int optval;
1946 		if (m == NULL || m->m_len != sizeof(int)) {
1947 			error = EINVAL;
1948 			break;
1949 		}
1950 		bcopy(mtod(m, u_int *), &optval, sizeof(optval));
1951 		if (optval < -1 || optval >= 256)
1952 			error = EINVAL;
1953 		else if (optval == -1)
1954 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1955 		else
1956 			im6o->im6o_multicast_hlim = optval;
1957 		break;
1958 	    }
1959 
1960 	case IPV6_MULTICAST_LOOP:
1961 		/*
1962 		 * Set the loopback flag for outgoing multicast packets.
1963 		 * Must be zero or one.
1964 		 */
1965 		if (m == NULL || m->m_len != sizeof(u_int)) {
1966 			error = EINVAL;
1967 			break;
1968 		}
1969 		bcopy(mtod(m, u_int *), &loop, sizeof(loop));
1970 		if (loop > 1) {
1971 			error = EINVAL;
1972 			break;
1973 		}
1974 		im6o->im6o_multicast_loop = loop;
1975 		break;
1976 
1977 	case IPV6_JOIN_GROUP:
1978 		/*
1979 		 * Add a multicast group membership.
1980 		 * Group must be a valid IP6 multicast address.
1981 		 */
1982 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
1983 			error = EINVAL;
1984 			break;
1985 		}
1986 		mreq = mtod(m, struct ipv6_mreq *);
1987 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
1988 			/*
1989 			 * We use the unspecified address to specify to accept
1990 			 * all multicast addresses. Only super user is allowed
1991 			 * to do this.
1992 			 */
1993 			if (suser(td))
1994 			{
1995 				error = EACCES;
1996 				break;
1997 			}
1998 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
1999 			error = EINVAL;
2000 			break;
2001 		}
2002 
2003 		/*
2004 		 * If the interface is specified, validate it.
2005 		 */
2006 		if (mreq->ipv6mr_interface < 0
2007 		 || if_index < mreq->ipv6mr_interface) {
2008 			error = ENXIO;	/* XXX EINVAL? */
2009 			break;
2010 		}
2011 		/*
2012 		 * If no interface was explicitly specified, choose an
2013 		 * appropriate one according to the given multicast address.
2014 		 */
2015 		if (mreq->ipv6mr_interface == 0) {
2016 			/*
2017 			 * If the multicast address is in node-local scope,
2018 			 * the interface should be a loopback interface.
2019 			 * Otherwise, look up the routing table for the
2020 			 * address, and choose the outgoing interface.
2021 			 *   XXX: is it a good approach?
2022 			 */
2023 			if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
2024 				ifp = &loif[0];
2025 			} else {
2026 				ro.ro_rt = NULL;
2027 				dst = (struct sockaddr_in6 *)&ro.ro_dst;
2028 				bzero(dst, sizeof(*dst));
2029 				dst->sin6_len = sizeof(struct sockaddr_in6);
2030 				dst->sin6_family = AF_INET6;
2031 				dst->sin6_addr = mreq->ipv6mr_multiaddr;
2032 				rtalloc((struct route *)&ro);
2033 				if (ro.ro_rt == NULL) {
2034 					error = EADDRNOTAVAIL;
2035 					break;
2036 				}
2037 				ifp = ro.ro_rt->rt_ifp;
2038 				rtfree(ro.ro_rt);
2039 			}
2040 		} else
2041 			ifp = ifnet_byindex(mreq->ipv6mr_interface);
2042 
2043 		/*
2044 		 * See if we found an interface, and confirm that it
2045 		 * supports multicast
2046 		 */
2047 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2048 			error = EADDRNOTAVAIL;
2049 			break;
2050 		}
2051 		/*
2052 		 * Put interface index into the multicast address,
2053 		 * if the address has link-local scope.
2054 		 */
2055 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2056 			mreq->ipv6mr_multiaddr.s6_addr16[1]
2057 				= htons(mreq->ipv6mr_interface);
2058 		}
2059 		/*
2060 		 * See if the membership already exists.
2061 		 */
2062 		for (imm = im6o->im6o_memberships.lh_first;
2063 		     imm != NULL; imm = imm->i6mm_chain.le_next)
2064 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
2065 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2066 					       &mreq->ipv6mr_multiaddr))
2067 				break;
2068 		if (imm != NULL) {
2069 			error = EADDRINUSE;
2070 			break;
2071 		}
2072 		/*
2073 		 * Everything looks good; add a new record to the multicast
2074 		 * address list for the given interface.
2075 		 */
2076 		imm = malloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
2077 		if (imm == NULL) {
2078 			error = ENOBUFS;
2079 			break;
2080 		}
2081 		if ((imm->i6mm_maddr =
2082 		     in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
2083 			free(imm, M_IPMADDR);
2084 			break;
2085 		}
2086 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2087 		break;
2088 
2089 	case IPV6_LEAVE_GROUP:
2090 		/*
2091 		 * Drop a multicast group membership.
2092 		 * Group must be a valid IP6 multicast address.
2093 		 */
2094 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2095 			error = EINVAL;
2096 			break;
2097 		}
2098 		mreq = mtod(m, struct ipv6_mreq *);
2099 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2100 			if (suser(td)) {
2101 				error = EACCES;
2102 				break;
2103 			}
2104 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2105 			error = EINVAL;
2106 			break;
2107 		}
2108 		/*
2109 		 * If an interface address was specified, get a pointer
2110 		 * to its ifnet structure.
2111 		 */
2112 		if (mreq->ipv6mr_interface < 0
2113 		 || if_index < mreq->ipv6mr_interface) {
2114 			error = ENXIO;	/* XXX EINVAL? */
2115 			break;
2116 		}
2117 		ifp = ifnet_byindex(mreq->ipv6mr_interface);
2118 		/*
2119 		 * Put interface index into the multicast address,
2120 		 * if the address has link-local scope.
2121 		 */
2122 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2123 			mreq->ipv6mr_multiaddr.s6_addr16[1]
2124 				= htons(mreq->ipv6mr_interface);
2125 		}
2126 		/*
2127 		 * Find the membership in the membership list.
2128 		 */
2129 		for (imm = im6o->im6o_memberships.lh_first;
2130 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
2131 			if ((ifp == NULL ||
2132 			     imm->i6mm_maddr->in6m_ifp == ifp) &&
2133 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2134 					       &mreq->ipv6mr_multiaddr))
2135 				break;
2136 		}
2137 		if (imm == NULL) {
2138 			/* Unable to resolve interface */
2139 			error = EADDRNOTAVAIL;
2140 			break;
2141 		}
2142 		/*
2143 		 * Give up the multicast address record to which the
2144 		 * membership points.
2145 		 */
2146 		LIST_REMOVE(imm, i6mm_chain);
2147 		in6_delmulti(imm->i6mm_maddr);
2148 		free(imm, M_IPMADDR);
2149 		break;
2150 
2151 	default:
2152 		error = EOPNOTSUPP;
2153 		break;
2154 	}
2155 
2156 	/*
2157 	 * If all options have default values, no need to keep the mbuf.
2158 	 */
2159 	if (im6o->im6o_multicast_ifp == NULL &&
2160 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2161 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2162 	    im6o->im6o_memberships.lh_first == NULL) {
2163 		free(*im6op, M_IPMOPTS);
2164 		*im6op = NULL;
2165 	}
2166 
2167 	return(error);
2168 }
2169 
2170 /*
2171  * Return the IP6 multicast options in response to user getsockopt().
2172  */
2173 static int
2174 ip6_getmoptions(optname, im6o, mp)
2175 	int optname;
2176 	struct ip6_moptions *im6o;
2177 	struct mbuf **mp;
2178 {
2179 	u_int *hlim, *loop, *ifindex;
2180 
2181 	*mp = m_get(M_TRYWAIT, MT_HEADER);		/* XXX */
2182 
2183 	switch (optname) {
2184 
2185 	case IPV6_MULTICAST_IF:
2186 		ifindex = mtod(*mp, u_int *);
2187 		(*mp)->m_len = sizeof(u_int);
2188 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2189 			*ifindex = 0;
2190 		else
2191 			*ifindex = im6o->im6o_multicast_ifp->if_index;
2192 		return(0);
2193 
2194 	case IPV6_MULTICAST_HOPS:
2195 		hlim = mtod(*mp, u_int *);
2196 		(*mp)->m_len = sizeof(u_int);
2197 		if (im6o == NULL)
2198 			*hlim = ip6_defmcasthlim;
2199 		else
2200 			*hlim = im6o->im6o_multicast_hlim;
2201 		return(0);
2202 
2203 	case IPV6_MULTICAST_LOOP:
2204 		loop = mtod(*mp, u_int *);
2205 		(*mp)->m_len = sizeof(u_int);
2206 		if (im6o == NULL)
2207 			*loop = ip6_defmcasthlim;
2208 		else
2209 			*loop = im6o->im6o_multicast_loop;
2210 		return(0);
2211 
2212 	default:
2213 		return(EOPNOTSUPP);
2214 	}
2215 }
2216 
2217 /*
2218  * Discard the IP6 multicast options.
2219  */
2220 void
2221 ip6_freemoptions(im6o)
2222 	struct ip6_moptions *im6o;
2223 {
2224 	struct in6_multi_mship *imm;
2225 
2226 	if (im6o == NULL)
2227 		return;
2228 
2229 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2230 		LIST_REMOVE(imm, i6mm_chain);
2231 		if (imm->i6mm_maddr)
2232 			in6_delmulti(imm->i6mm_maddr);
2233 		free(imm, M_IPMADDR);
2234 	}
2235 	free(im6o, M_IPMOPTS);
2236 }
2237 
2238 /*
2239  * Set IPv6 outgoing packet options based on advanced API.
2240  */
2241 int
2242 ip6_setpktoptions(control, opt, priv, needcopy)
2243 	struct mbuf *control;
2244 	struct ip6_pktopts *opt;
2245 	int priv, needcopy;
2246 {
2247 	struct cmsghdr *cm = 0;
2248 
2249 	if (control == 0 || opt == 0)
2250 		return(EINVAL);
2251 
2252 	init_ip6pktopts(opt);
2253 
2254 	/*
2255 	 * XXX: Currently, we assume all the optional information is stored
2256 	 * in a single mbuf.
2257 	 */
2258 	if (control->m_next)
2259 		return(EINVAL);
2260 
2261 	for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2262 		     control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2263 		cm = mtod(control, struct cmsghdr *);
2264 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2265 			return(EINVAL);
2266 		if (cm->cmsg_level != IPPROTO_IPV6)
2267 			continue;
2268 
2269 		/*
2270 		 * XXX should check if RFC2292 API is mixed with 2292bis API
2271 		 */
2272 		switch (cm->cmsg_type) {
2273 		case IPV6_PKTINFO:
2274 			if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo)))
2275 				return(EINVAL);
2276 			if (needcopy) {
2277 				/* XXX: Is it really WAITOK? */
2278 				opt->ip6po_pktinfo =
2279 					malloc(sizeof(struct in6_pktinfo),
2280 					       M_IP6OPT, M_WAITOK);
2281 				bcopy(CMSG_DATA(cm), opt->ip6po_pktinfo,
2282 				    sizeof(struct in6_pktinfo));
2283 			} else
2284 				opt->ip6po_pktinfo =
2285 					(struct in6_pktinfo *)CMSG_DATA(cm);
2286 			if (opt->ip6po_pktinfo->ipi6_ifindex &&
2287 			    IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr))
2288 				opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] =
2289 					htons(opt->ip6po_pktinfo->ipi6_ifindex);
2290 
2291 			if (opt->ip6po_pktinfo->ipi6_ifindex > if_index
2292 			 || opt->ip6po_pktinfo->ipi6_ifindex < 0) {
2293 				return(ENXIO);
2294 			}
2295 
2296 			/*
2297 			 * Check if the requested source address is indeed a
2298 			 * unicast address assigned to the node, and can be
2299 			 * used as the packet's source address.
2300 			 */
2301 			if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
2302 				struct in6_ifaddr *ia6;
2303 				struct sockaddr_in6 sin6;
2304 
2305 				bzero(&sin6, sizeof(sin6));
2306 				sin6.sin6_len = sizeof(sin6);
2307 				sin6.sin6_family = AF_INET6;
2308 				sin6.sin6_addr =
2309 					opt->ip6po_pktinfo->ipi6_addr;
2310 				ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(sin6tosa(&sin6));
2311 				if (ia6 == NULL ||
2312 				    (ia6->ia6_flags & (IN6_IFF_ANYCAST |
2313 						       IN6_IFF_NOTREADY)) != 0)
2314 					return(EADDRNOTAVAIL);
2315 			}
2316 			break;
2317 
2318 		case IPV6_HOPLIMIT:
2319 			if (cm->cmsg_len != CMSG_LEN(sizeof(int)))
2320 				return(EINVAL);
2321 
2322 			opt->ip6po_hlim = *(int *)CMSG_DATA(cm);
2323 			if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255)
2324 				return(EINVAL);
2325 			break;
2326 
2327 		case IPV6_NEXTHOP:
2328 			if (!priv)
2329 				return(EPERM);
2330 
2331 			if (cm->cmsg_len < sizeof(u_char) ||
2332 			    /* check if cmsg_len is large enough for sa_len */
2333 			    cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm)))
2334 				return(EINVAL);
2335 
2336 			if (needcopy) {
2337 				opt->ip6po_nexthop =
2338 					malloc(*CMSG_DATA(cm),
2339 					       M_IP6OPT, M_WAITOK);
2340 				bcopy(CMSG_DATA(cm),
2341 				      opt->ip6po_nexthop,
2342 				      *CMSG_DATA(cm));
2343 			} else
2344 				opt->ip6po_nexthop =
2345 					(struct sockaddr *)CMSG_DATA(cm);
2346 			break;
2347 
2348 		case IPV6_HOPOPTS:
2349 		{
2350 			struct ip6_hbh *hbh;
2351 			int hbhlen;
2352 
2353 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh)))
2354 				return(EINVAL);
2355 			hbh = (struct ip6_hbh *)CMSG_DATA(cm);
2356 			hbhlen = (hbh->ip6h_len + 1) << 3;
2357 			if (cm->cmsg_len != CMSG_LEN(hbhlen))
2358 				return(EINVAL);
2359 
2360 			if (needcopy) {
2361 				opt->ip6po_hbh =
2362 					malloc(hbhlen, M_IP6OPT, M_WAITOK);
2363 				bcopy(hbh, opt->ip6po_hbh, hbhlen);
2364 			} else
2365 				opt->ip6po_hbh = hbh;
2366 			break;
2367 		}
2368 
2369 		case IPV6_DSTOPTS:
2370 		{
2371 			struct ip6_dest *dest, **newdest;
2372 			int destlen;
2373 
2374 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest)))
2375 				return(EINVAL);
2376 			dest = (struct ip6_dest *)CMSG_DATA(cm);
2377 			destlen = (dest->ip6d_len + 1) << 3;
2378 			if (cm->cmsg_len != CMSG_LEN(destlen))
2379 				return(EINVAL);
2380 
2381 			/*
2382 			 * The old advacned API is ambiguous on this
2383 			 * point. Our approach is to determine the
2384 			 * position based according to the existence
2385 			 * of a routing header. Note, however, that
2386 			 * this depends on the order of the extension
2387 			 * headers in the ancillary data; the 1st part
2388 			 * of the destination options header must
2389 			 * appear before the routing header in the
2390 			 * ancillary data, too.
2391 			 * RFC2292bis solved the ambiguity by
2392 			 * introducing separate cmsg types.
2393 			 */
2394 			if (opt->ip6po_rthdr == NULL)
2395 				newdest = &opt->ip6po_dest1;
2396 			else
2397 				newdest = &opt->ip6po_dest2;
2398 
2399 			if (needcopy) {
2400 				*newdest = malloc(destlen, M_IP6OPT, M_WAITOK);
2401 				bcopy(dest, *newdest, destlen);
2402 			} else
2403 				*newdest = dest;
2404 
2405 			break;
2406 		}
2407 
2408 		case IPV6_RTHDR:
2409 		{
2410 			struct ip6_rthdr *rth;
2411 			int rthlen;
2412 
2413 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr)))
2414 				return(EINVAL);
2415 			rth = (struct ip6_rthdr *)CMSG_DATA(cm);
2416 			rthlen = (rth->ip6r_len + 1) << 3;
2417 			if (cm->cmsg_len != CMSG_LEN(rthlen))
2418 				return(EINVAL);
2419 
2420 			switch (rth->ip6r_type) {
2421 			case IPV6_RTHDR_TYPE_0:
2422 				/* must contain one addr */
2423 				if (rth->ip6r_len == 0)
2424 					return(EINVAL);
2425 				/* length must be even */
2426 				if (rth->ip6r_len % 2)
2427 					return(EINVAL);
2428 				if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2429 					return(EINVAL);
2430 				break;
2431 			default:
2432 				return(EINVAL);	/* not supported */
2433 			}
2434 
2435 			if (needcopy) {
2436 				opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT,
2437 							  M_WAITOK);
2438 				bcopy(rth, opt->ip6po_rthdr, rthlen);
2439 			} else
2440 				opt->ip6po_rthdr = rth;
2441 
2442 			break;
2443 		}
2444 
2445 		default:
2446 			return(ENOPROTOOPT);
2447 		}
2448 	}
2449 
2450 	return(0);
2451 }
2452 
2453 /*
2454  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2455  * packet to the input queue of a specified interface.  Note that this
2456  * calls the output routine of the loopback "driver", but with an interface
2457  * pointer that might NOT be &loif -- easier than replicating that code here.
2458  */
2459 void
2460 ip6_mloopback(ifp, m, dst)
2461 	struct ifnet *ifp;
2462 	struct mbuf *m;
2463 	struct sockaddr_in6 *dst;
2464 {
2465 	struct mbuf *copym;
2466 	struct ip6_hdr *ip6;
2467 
2468 	copym = m_copy(m, 0, M_COPYALL);
2469 	if (copym == NULL)
2470 		return;
2471 
2472 	/*
2473 	 * Make sure to deep-copy IPv6 header portion in case the data
2474 	 * is in an mbuf cluster, so that we can safely override the IPv6
2475 	 * header portion later.
2476 	 */
2477 	if ((copym->m_flags & M_EXT) != 0 ||
2478 	    copym->m_len < sizeof(struct ip6_hdr)) {
2479 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
2480 		if (copym == NULL)
2481 			return;
2482 	}
2483 
2484 #ifdef DIAGNOSTIC
2485 	if (copym->m_len < sizeof(*ip6)) {
2486 		m_freem(copym);
2487 		return;
2488 	}
2489 #endif
2490 
2491 	ip6 = mtod(copym, struct ip6_hdr *);
2492 #ifndef SCOPEDROUTING
2493 	/*
2494 	 * clear embedded scope identifiers if necessary.
2495 	 * in6_clearscope will touch the addresses only when necessary.
2496 	 */
2497 	in6_clearscope(&ip6->ip6_src);
2498 	in6_clearscope(&ip6->ip6_dst);
2499 #endif
2500 
2501 	(void)if_simloop(ifp, copym, dst->sin6_family, NULL);
2502 }
2503 
2504 /*
2505  * Chop IPv6 header off from the payload.
2506  */
2507 static int
2508 ip6_splithdr(m, exthdrs)
2509 	struct mbuf *m;
2510 	struct ip6_exthdrs *exthdrs;
2511 {
2512 	struct mbuf *mh;
2513 	struct ip6_hdr *ip6;
2514 
2515 	ip6 = mtod(m, struct ip6_hdr *);
2516 	if (m->m_len > sizeof(*ip6)) {
2517 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
2518 		if (mh == 0) {
2519 			m_freem(m);
2520 			return ENOBUFS;
2521 		}
2522 		M_COPY_PKTHDR(mh, m);
2523 		MH_ALIGN(mh, sizeof(*ip6));
2524 		m->m_flags &= ~M_PKTHDR;
2525 		m->m_len -= sizeof(*ip6);
2526 		m->m_data += sizeof(*ip6);
2527 		mh->m_next = m;
2528 		m = mh;
2529 		m->m_len = sizeof(*ip6);
2530 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2531 	}
2532 	exthdrs->ip6e_ip6 = m;
2533 	return 0;
2534 }
2535 
2536 /*
2537  * Compute IPv6 extension header length.
2538  */
2539 int
2540 ip6_optlen(in6p)
2541 	struct in6pcb *in6p;
2542 {
2543 	int len;
2544 
2545 	if (!in6p->in6p_outputopts)
2546 		return 0;
2547 
2548 	len = 0;
2549 #define elen(x) \
2550     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2551 
2552 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
2553 	if (in6p->in6p_outputopts->ip6po_rthdr)
2554 		/* dest1 is valid with rthdr only */
2555 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
2556 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
2557 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
2558 	return len;
2559 #undef elen
2560 }
2561