xref: /freebsd/sys/netinet6/ip6_output.c (revision f9218d3d4fd34f082473b3a021c6d4d109fb47cf)
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 #ifdef FAST_IPSEC
109 #include <netipsec/ipsec.h>
110 #include <netipsec/ipsec6.h>
111 #include <netipsec/key.h>
112 #endif /* FAST_IPSEC */
113 
114 #include <netinet6/ip6_fw.h>
115 
116 #include <net/net_osdep.h>
117 
118 #include <netinet6/ip6protosw.h>
119 
120 static MALLOC_DEFINE(M_IPMOPTS, "ip6_moptions", "internet multicast options");
121 
122 struct ip6_exthdrs {
123 	struct mbuf *ip6e_ip6;
124 	struct mbuf *ip6e_hbh;
125 	struct mbuf *ip6e_dest1;
126 	struct mbuf *ip6e_rthdr;
127 	struct mbuf *ip6e_dest2;
128 };
129 
130 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
131 			    struct socket *, struct sockopt *sopt));
132 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *));
133 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **));
134 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int));
135 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
136 				  struct ip6_frag **));
137 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t));
138 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *));
139 
140 /*
141  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
142  * header (with pri, len, nxt, hlim, src, dst).
143  * This function may modify ver and hlim only.
144  * The mbuf chain containing the packet will be freed.
145  * The mbuf opt, if present, will not be freed.
146  *
147  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
148  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
149  * which is rt_rmx.rmx_mtu.
150  */
151 int
152 ip6_output(m0, opt, ro, flags, im6o, ifpp, inp)
153 	struct mbuf *m0;
154 	struct ip6_pktopts *opt;
155 	struct route_in6 *ro;
156 	int flags;
157 	struct ip6_moptions *im6o;
158 	struct ifnet **ifpp;		/* XXX: just for statistics */
159 	struct inpcb *inp;
160 {
161 	struct ip6_hdr *ip6, *mhip6;
162 	struct ifnet *ifp, *origifp;
163 	struct mbuf *m = m0;
164 	int hlen, tlen, len, off;
165 	struct route_in6 ip6route;
166 	struct sockaddr_in6 *dst;
167 	int error = 0;
168 	struct in6_ifaddr *ia = NULL;
169 	u_long mtu;
170 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
171 	struct ip6_exthdrs exthdrs;
172 	struct in6_addr finaldst;
173 	struct route_in6 *ro_pmtu = NULL;
174 	int hdrsplit = 0;
175 	int needipsec = 0;
176 #ifdef PFIL_HOOKS
177 	struct packet_filter_hook *pfh;
178 	struct mbuf *m1;
179 	int rv;
180 #endif /* PFIL_HOOKS */
181 #ifdef IPSEC
182 	int needipsectun = 0;
183 	struct secpolicy *sp = NULL;
184 
185 	ip6 = mtod(m, struct ip6_hdr *);
186 #endif /* IPSEC */
187 #ifdef FAST_IPSEC
188 	int needipsectun = 0;
189 	struct secpolicy *sp = NULL;
190 
191 	ip6 = mtod(m, struct ip6_hdr *);
192 #endif /* FAST_IPSEC */
193 
194 #define MAKE_EXTHDR(hp, mp)						\
195     do {								\
196 	if (hp) {							\
197 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
198 		error = ip6_copyexthdr((mp), (caddr_t)(hp), 		\
199 				       ((eh)->ip6e_len + 1) << 3);	\
200 		if (error)						\
201 			goto freehdrs;					\
202 	}								\
203     } while (0)
204 
205 	bzero(&exthdrs, sizeof(exthdrs));
206 
207 	if (opt) {
208 		/* Hop-by-Hop options header */
209 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
210 		/* Destination options header(1st part) */
211 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
212 		/* Routing header */
213 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
214 		/* Destination options header(2nd part) */
215 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
216 	}
217 
218 #ifdef IPSEC
219 	/* get a security policy for this packet */
220 	if (inp == NULL)
221 		sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
222 	else
223 		sp = ipsec6_getpolicybypcb(m, IPSEC_DIR_OUTBOUND, inp, &error);
224 
225 	if (sp == NULL) {
226 		ipsec6stat.out_inval++;
227 		goto freehdrs;
228 	}
229 
230 	error = 0;
231 
232 	/* check policy */
233 	switch (sp->policy) {
234 	case IPSEC_POLICY_DISCARD:
235 		/*
236 		 * This packet is just discarded.
237 		 */
238 		ipsec6stat.out_polvio++;
239 		goto freehdrs;
240 
241 	case IPSEC_POLICY_BYPASS:
242 	case IPSEC_POLICY_NONE:
243 		/* no need to do IPsec. */
244 		needipsec = 0;
245 		break;
246 
247 	case IPSEC_POLICY_IPSEC:
248 		if (sp->req == NULL) {
249 			/* acquire a policy */
250 			error = key_spdacquire(sp);
251 			goto freehdrs;
252 		}
253 		needipsec = 1;
254 		break;
255 
256 	case IPSEC_POLICY_ENTRUST:
257 	default:
258 		printf("ip6_output: Invalid policy found. %d\n", sp->policy);
259 	}
260 #endif /* IPSEC */
261 #ifdef FAST_IPSEC
262 	/* get a security policy for this packet */
263 	if (inp == NULL)
264 		sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
265 	else
266 		sp = ipsec_getpolicybysock(m, IPSEC_DIR_OUTBOUND, inp, &error);
267 
268 	if (sp == NULL) {
269 		newipsecstat.ips_out_inval++;
270 		goto freehdrs;
271 	}
272 
273 	error = 0;
274 
275 	/* check policy */
276 	switch (sp->policy) {
277 	case IPSEC_POLICY_DISCARD:
278 		/*
279 		 * This packet is just discarded.
280 		 */
281 		newipsecstat.ips_out_polvio++;
282 		goto freehdrs;
283 
284 	case IPSEC_POLICY_BYPASS:
285 	case IPSEC_POLICY_NONE:
286 		/* no need to do IPsec. */
287 		needipsec = 0;
288 		break;
289 
290 	case IPSEC_POLICY_IPSEC:
291 		if (sp->req == NULL) {
292 			/* acquire a policy */
293 			error = key_spdacquire(sp);
294 			goto freehdrs;
295 		}
296 		needipsec = 1;
297 		break;
298 
299 	case IPSEC_POLICY_ENTRUST:
300 	default:
301 		printf("ip6_output: Invalid policy found. %d\n", sp->policy);
302 	}
303 #endif /* FAST_IPSEC */
304 
305 	/*
306 	 * Calculate the total length of the extension header chain.
307 	 * Keep the length of the unfragmentable part for fragmentation.
308 	 */
309 	optlen = 0;
310 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
311 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
312 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
313 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
314 	/* NOTE: we don't add AH/ESP length here. do that later. */
315 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
316 
317 	/*
318 	 * If we need IPsec, or there is at least one extension header,
319 	 * separate IP6 header from the payload.
320 	 */
321 	if ((needipsec || optlen) && !hdrsplit) {
322 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
323 			m = NULL;
324 			goto freehdrs;
325 		}
326 		m = exthdrs.ip6e_ip6;
327 		hdrsplit++;
328 	}
329 
330 	/* adjust pointer */
331 	ip6 = mtod(m, struct ip6_hdr *);
332 
333 	/* adjust mbuf packet header length */
334 	m->m_pkthdr.len += optlen;
335 	plen = m->m_pkthdr.len - sizeof(*ip6);
336 
337 	/* If this is a jumbo payload, insert a jumbo payload option. */
338 	if (plen > IPV6_MAXPACKET) {
339 		if (!hdrsplit) {
340 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
341 				m = NULL;
342 				goto freehdrs;
343 			}
344 			m = exthdrs.ip6e_ip6;
345 			hdrsplit++;
346 		}
347 		/* adjust pointer */
348 		ip6 = mtod(m, struct ip6_hdr *);
349 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
350 			goto freehdrs;
351 		ip6->ip6_plen = 0;
352 	} else
353 		ip6->ip6_plen = htons(plen);
354 
355 	/*
356 	 * Concatenate headers and fill in next header fields.
357 	 * Here we have, on "m"
358 	 *	IPv6 payload
359 	 * and we insert headers accordingly.  Finally, we should be getting:
360 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
361 	 *
362 	 * during the header composing process, "m" points to IPv6 header.
363 	 * "mprev" points to an extension header prior to esp.
364 	 */
365 	{
366 		u_char *nexthdrp = &ip6->ip6_nxt;
367 		struct mbuf *mprev = m;
368 
369 		/*
370 		 * we treat dest2 specially.  this makes IPsec processing
371 		 * much easier.  the goal here is to make mprev point the
372 		 * mbuf prior to dest2.
373 		 *
374 		 * result: IPv6 dest2 payload
375 		 * m and mprev will point to IPv6 header.
376 		 */
377 		if (exthdrs.ip6e_dest2) {
378 			if (!hdrsplit)
379 				panic("assumption failed: hdr not split");
380 			exthdrs.ip6e_dest2->m_next = m->m_next;
381 			m->m_next = exthdrs.ip6e_dest2;
382 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
383 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
384 		}
385 
386 #define MAKE_CHAIN(m, mp, p, i)\
387     do {\
388 	if (m) {\
389 		if (!hdrsplit) \
390 			panic("assumption failed: hdr not split"); \
391 		*mtod((m), u_char *) = *(p);\
392 		*(p) = (i);\
393 		p = mtod((m), u_char *);\
394 		(m)->m_next = (mp)->m_next;\
395 		(mp)->m_next = (m);\
396 		(mp) = (m);\
397 	}\
398     } while (0)
399 		/*
400 		 * result: IPv6 hbh dest1 rthdr dest2 payload
401 		 * m will point to IPv6 header.  mprev will point to the
402 		 * extension header prior to dest2 (rthdr in the above case).
403 		 */
404 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev,
405 			   nexthdrp, IPPROTO_HOPOPTS);
406 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev,
407 			   nexthdrp, IPPROTO_DSTOPTS);
408 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev,
409 			   nexthdrp, IPPROTO_ROUTING);
410 
411 #if defined(IPSEC) || defined(FAST_IPSEC)
412 		if (!needipsec)
413 			goto skip_ipsec2;
414 
415 		/*
416 		 * pointers after IPsec headers are not valid any more.
417 		 * other pointers need a great care too.
418 		 * (IPsec routines should not mangle mbufs prior to AH/ESP)
419 		 */
420 		exthdrs.ip6e_dest2 = NULL;
421 
422 	    {
423 		struct ip6_rthdr *rh = NULL;
424 		int segleft_org = 0;
425 		struct ipsec_output_state state;
426 
427 		if (exthdrs.ip6e_rthdr) {
428 			rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
429 			segleft_org = rh->ip6r_segleft;
430 			rh->ip6r_segleft = 0;
431 		}
432 
433 		bzero(&state, sizeof(state));
434 		state.m = m;
435 		error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
436 			&needipsectun);
437 		m = state.m;
438 		if (error) {
439 			/* mbuf is already reclaimed in ipsec6_output_trans. */
440 			m = NULL;
441 			switch (error) {
442 			case EHOSTUNREACH:
443 			case ENETUNREACH:
444 			case EMSGSIZE:
445 			case ENOBUFS:
446 			case ENOMEM:
447 				break;
448 			default:
449 				printf("ip6_output (ipsec): error code %d\n", error);
450 				/* fall through */
451 			case ENOENT:
452 				/* don't show these error codes to the user */
453 				error = 0;
454 				break;
455 			}
456 			goto bad;
457 		}
458 		if (exthdrs.ip6e_rthdr) {
459 			/* ah6_output doesn't modify mbuf chain */
460 			rh->ip6r_segleft = segleft_org;
461 		}
462 	    }
463 skip_ipsec2:;
464 #endif
465 	}
466 
467 	/*
468 	 * If there is a routing header, replace destination address field
469 	 * with the first hop of the routing header.
470 	 */
471 	if (exthdrs.ip6e_rthdr) {
472 		struct ip6_rthdr *rh =
473 			(struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
474 						  struct ip6_rthdr *));
475 		struct ip6_rthdr0 *rh0;
476 
477 		finaldst = ip6->ip6_dst;
478 		switch (rh->ip6r_type) {
479 		case IPV6_RTHDR_TYPE_0:
480 			 rh0 = (struct ip6_rthdr0 *)rh;
481 			 ip6->ip6_dst = rh0->ip6r0_addr[0];
482 			 bcopy((caddr_t)&rh0->ip6r0_addr[1],
483 			       (caddr_t)&rh0->ip6r0_addr[0],
484 			       sizeof(struct in6_addr)*(rh0->ip6r0_segleft - 1)
485 				 );
486 			 rh0->ip6r0_addr[rh0->ip6r0_segleft - 1] = finaldst;
487 			 break;
488 		default:	/* is it possible? */
489 			 error = EINVAL;
490 			 goto bad;
491 		}
492 	}
493 
494 	/* Source address validation */
495 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
496 	    (flags & IPV6_DADOUTPUT) == 0) {
497 		error = EOPNOTSUPP;
498 		ip6stat.ip6s_badscope++;
499 		goto bad;
500 	}
501 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
502 		error = EOPNOTSUPP;
503 		ip6stat.ip6s_badscope++;
504 		goto bad;
505 	}
506 
507 	ip6stat.ip6s_localout++;
508 
509 	/*
510 	 * Route packet.
511 	 */
512 	if (ro == 0) {
513 		ro = &ip6route;
514 		bzero((caddr_t)ro, sizeof(*ro));
515 	}
516 	ro_pmtu = ro;
517 	if (opt && opt->ip6po_rthdr)
518 		ro = &opt->ip6po_route;
519 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
520 	/*
521 	 * If there is a cached route,
522 	 * check that it is to the same destination
523 	 * and is still up. If not, free it and try again.
524 	 */
525 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
526 			 dst->sin6_family != AF_INET6 ||
527 			 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
528 		RTFREE(ro->ro_rt);
529 		ro->ro_rt = (struct rtentry *)0;
530 	}
531 	if (ro->ro_rt == 0) {
532 		bzero(dst, sizeof(*dst));
533 		dst->sin6_family = AF_INET6;
534 		dst->sin6_len = sizeof(struct sockaddr_in6);
535 		dst->sin6_addr = ip6->ip6_dst;
536 #ifdef SCOPEDROUTING
537 		/* XXX: sin6_scope_id should already be fixed at this point */
538 		if (IN6_IS_SCOPE_LINKLOCAL(&dst->sin6_addr))
539 			dst->sin6_scope_id = ntohs(dst->sin6_addr.s6_addr16[1]);
540 #endif
541 	}
542 #if defined(IPSEC) || defined(FAST_IPSEC)
543 	if (needipsec && needipsectun) {
544 		struct ipsec_output_state state;
545 
546 		/*
547 		 * All the extension headers will become inaccessible
548 		 * (since they can be encrypted).
549 		 * Don't panic, we need no more updates to extension headers
550 		 * on inner IPv6 packet (since they are now encapsulated).
551 		 *
552 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
553 		 */
554 		bzero(&exthdrs, sizeof(exthdrs));
555 		exthdrs.ip6e_ip6 = m;
556 
557 		bzero(&state, sizeof(state));
558 		state.m = m;
559 		state.ro = (struct route *)ro;
560 		state.dst = (struct sockaddr *)dst;
561 
562 		error = ipsec6_output_tunnel(&state, sp, flags);
563 
564 		m = state.m;
565 		ro = (struct route_in6 *)state.ro;
566 		dst = (struct sockaddr_in6 *)state.dst;
567 		if (error) {
568 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
569 			m0 = m = NULL;
570 			m = NULL;
571 			switch (error) {
572 			case EHOSTUNREACH:
573 			case ENETUNREACH:
574 			case EMSGSIZE:
575 			case ENOBUFS:
576 			case ENOMEM:
577 				break;
578 			default:
579 				printf("ip6_output (ipsec): error code %d\n", error);
580 				/* fall through */
581 			case ENOENT:
582 				/* don't show these error codes to the user */
583 				error = 0;
584 				break;
585 			}
586 			goto bad;
587 		}
588 
589 		exthdrs.ip6e_ip6 = m;
590 	}
591 #endif /* IPSEC */
592 
593 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
594 		/* Unicast */
595 
596 #define ifatoia6(ifa)	((struct in6_ifaddr *)(ifa))
597 #define sin6tosa(sin6)	((struct sockaddr *)(sin6))
598 		/* xxx
599 		 * interface selection comes here
600 		 * if an interface is specified from an upper layer,
601 		 * ifp must point it.
602 		 */
603 		if (ro->ro_rt == 0) {
604 			/*
605 			 * non-bsdi always clone routes, if parent is
606 			 * PRF_CLONING.
607 			 */
608 			rtalloc((struct route *)ro);
609 		}
610 		if (ro->ro_rt == 0) {
611 			ip6stat.ip6s_noroute++;
612 			error = EHOSTUNREACH;
613 			/* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
614 			goto bad;
615 		}
616 		ia = ifatoia6(ro->ro_rt->rt_ifa);
617 		ifp = ro->ro_rt->rt_ifp;
618 		ro->ro_rt->rt_use++;
619 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
620 			dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
621 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
622 
623 		in6_ifstat_inc(ifp, ifs6_out_request);
624 
625 		/*
626 		 * Check if the outgoing interface conflicts with
627 		 * the interface specified by ifi6_ifindex (if specified).
628 		 * Note that loopback interface is always okay.
629 		 * (this may happen when we are sending a packet to one of
630 		 *  our own addresses.)
631 		 */
632 		if (opt && opt->ip6po_pktinfo
633 		 && opt->ip6po_pktinfo->ipi6_ifindex) {
634 			if (!(ifp->if_flags & IFF_LOOPBACK)
635 			 && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
636 				ip6stat.ip6s_noroute++;
637 				in6_ifstat_inc(ifp, ifs6_out_discard);
638 				error = EHOSTUNREACH;
639 				goto bad;
640 			}
641 		}
642 
643 		if (opt && opt->ip6po_hlim != -1)
644 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
645 	} else {
646 		/* Multicast */
647 		struct	in6_multi *in6m;
648 
649 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
650 
651 		/*
652 		 * See if the caller provided any multicast options
653 		 */
654 		ifp = NULL;
655 		if (im6o != NULL) {
656 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
657 			if (im6o->im6o_multicast_ifp != NULL)
658 				ifp = im6o->im6o_multicast_ifp;
659 		} else
660 			ip6->ip6_hlim = ip6_defmcasthlim;
661 
662 		/*
663 		 * See if the caller provided the outgoing interface
664 		 * as an ancillary data.
665 		 * Boundary check for ifindex is assumed to be already done.
666 		 */
667 		if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
668 			ifp = ifnet_byindex(opt->ip6po_pktinfo->ipi6_ifindex);
669 
670 		/*
671 		 * If the destination is a node-local scope multicast,
672 		 * the packet should be loop-backed only.
673 		 */
674 		if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
675 			/*
676 			 * If the outgoing interface is already specified,
677 			 * it should be a loopback interface.
678 			 */
679 			if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) {
680 				ip6stat.ip6s_badscope++;
681 				error = ENETUNREACH; /* XXX: better error? */
682 				/* XXX correct ifp? */
683 				in6_ifstat_inc(ifp, ifs6_out_discard);
684 				goto bad;
685 			} else {
686 				ifp = &loif[0];
687 			}
688 		}
689 
690 		if (opt && opt->ip6po_hlim != -1)
691 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
692 
693 		/*
694 		 * If caller did not provide an interface lookup a
695 		 * default in the routing table.  This is either a
696 		 * default for the speicfied group (i.e. a host
697 		 * route), or a multicast default (a route for the
698 		 * ``net'' ff00::/8).
699 		 */
700 		if (ifp == NULL) {
701 			if (ro->ro_rt == 0) {
702 				ro->ro_rt = rtalloc1((struct sockaddr *)
703 						&ro->ro_dst, 0, 0UL);
704 			}
705 			if (ro->ro_rt == 0) {
706 				ip6stat.ip6s_noroute++;
707 				error = EHOSTUNREACH;
708 				/* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
709 				goto bad;
710 			}
711 			ia = ifatoia6(ro->ro_rt->rt_ifa);
712 			ifp = ro->ro_rt->rt_ifp;
713 			ro->ro_rt->rt_use++;
714 		}
715 
716 		if ((flags & IPV6_FORWARDING) == 0)
717 			in6_ifstat_inc(ifp, ifs6_out_request);
718 		in6_ifstat_inc(ifp, ifs6_out_mcast);
719 
720 		/*
721 		 * Confirm that the outgoing interface supports multicast.
722 		 */
723 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
724 			ip6stat.ip6s_noroute++;
725 			in6_ifstat_inc(ifp, ifs6_out_discard);
726 			error = ENETUNREACH;
727 			goto bad;
728 		}
729 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
730 		if (in6m != NULL &&
731 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
732 			/*
733 			 * If we belong to the destination multicast group
734 			 * on the outgoing interface, and the caller did not
735 			 * forbid loopback, loop back a copy.
736 			 */
737 			ip6_mloopback(ifp, m, dst);
738 		} else {
739 			/*
740 			 * If we are acting as a multicast router, perform
741 			 * multicast forwarding as if the packet had just
742 			 * arrived on the interface to which we are about
743 			 * to send.  The multicast forwarding function
744 			 * recursively calls this function, using the
745 			 * IPV6_FORWARDING flag to prevent infinite recursion.
746 			 *
747 			 * Multicasts that are looped back by ip6_mloopback(),
748 			 * above, will be forwarded by the ip6_input() routine,
749 			 * if necessary.
750 			 */
751 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
752 				if (ip6_mforward(ip6, ifp, m) != 0) {
753 					m_freem(m);
754 					goto done;
755 				}
756 			}
757 		}
758 		/*
759 		 * Multicasts with a hoplimit of zero may be looped back,
760 		 * above, but must not be transmitted on a network.
761 		 * Also, multicasts addressed to the loopback interface
762 		 * are not sent -- the above call to ip6_mloopback() will
763 		 * loop back a copy if this host actually belongs to the
764 		 * destination group on the loopback interface.
765 		 */
766 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
767 			m_freem(m);
768 			goto done;
769 		}
770 	}
771 
772 	/*
773 	 * Fill the outgoing inteface to tell the upper layer
774 	 * to increment per-interface statistics.
775 	 */
776 	if (ifpp)
777 		*ifpp = ifp;
778 
779 	/*
780 	 * Determine path MTU.
781 	 */
782 	if (ro_pmtu != ro) {
783 		/* The first hop and the final destination may differ. */
784 		struct sockaddr_in6 *sin6_fin =
785 			(struct sockaddr_in6 *)&ro_pmtu->ro_dst;
786 		if (ro_pmtu->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
787 				       !IN6_ARE_ADDR_EQUAL(&sin6_fin->sin6_addr,
788 							   &finaldst))) {
789 			RTFREE(ro_pmtu->ro_rt);
790 			ro_pmtu->ro_rt = (struct rtentry *)0;
791 		}
792 		if (ro_pmtu->ro_rt == 0) {
793 			bzero(sin6_fin, sizeof(*sin6_fin));
794 			sin6_fin->sin6_family = AF_INET6;
795 			sin6_fin->sin6_len = sizeof(struct sockaddr_in6);
796 			sin6_fin->sin6_addr = finaldst;
797 
798 			rtalloc((struct route *)ro_pmtu);
799 		}
800 	}
801 	if (ro_pmtu->ro_rt != NULL) {
802 		u_int32_t ifmtu = nd_ifinfo[ifp->if_index].linkmtu;
803 
804 		mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
805 		if (mtu > ifmtu || mtu == 0) {
806 			/*
807 			 * The MTU on the route is larger than the MTU on
808 			 * the interface!  This shouldn't happen, unless the
809 			 * MTU of the interface has been changed after the
810 			 * interface was brought up.  Change the MTU in the
811 			 * route to match the interface MTU (as long as the
812 			 * field isn't locked).
813 			 *
814 			 * if MTU on the route is 0, we need to fix the MTU.
815 			 * this case happens with path MTU discovery timeouts.
816 			 */
817 			 mtu = ifmtu;
818 			 if ((ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
819 				 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; /* XXX */
820 		}
821 	} else {
822 		mtu = nd_ifinfo[ifp->if_index].linkmtu;
823 	}
824 
825 	/*
826 	 * advanced API (IPV6_USE_MIN_MTU) overrides mtu setting
827 	 */
828 	if ((flags & IPV6_MINMTU) != 0 && mtu > IPV6_MMTU)
829 		mtu = IPV6_MMTU;
830 
831 	/* Fake scoped addresses */
832 	if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
833 		/*
834 		 * If source or destination address is a scoped address, and
835 		 * the packet is going to be sent to a loopback interface,
836 		 * we should keep the original interface.
837 		 */
838 
839 		/*
840 		 * XXX: this is a very experimental and temporary solution.
841 		 * We eventually have sockaddr_in6 and use the sin6_scope_id
842 		 * field of the structure here.
843 		 * We rely on the consistency between two scope zone ids
844 		 * of source and destination, which should already be assured.
845 		 * Larger scopes than link will be supported in the future.
846 		 */
847 		origifp = NULL;
848 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
849 			origifp = ifnet_byindex(ntohs(ip6->ip6_src.s6_addr16[1]));
850 		else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
851 			origifp = ifnet_byindex(ntohs(ip6->ip6_dst.s6_addr16[1]));
852 		/*
853 		 * XXX: origifp can be NULL even in those two cases above.
854 		 * For example, if we remove the (only) link-local address
855 		 * from the loopback interface, and try to send a link-local
856 		 * address without link-id information.  Then the source
857 		 * address is ::1, and the destination address is the
858 		 * link-local address with its s6_addr16[1] being zero.
859 		 * What is worse, if the packet goes to the loopback interface
860 		 * by a default rejected route, the null pointer would be
861 		 * passed to looutput, and the kernel would hang.
862 		 * The following last resort would prevent such disaster.
863 		 */
864 		if (origifp == NULL)
865 			origifp = ifp;
866 	}
867 	else
868 		origifp = ifp;
869 #ifndef SCOPEDROUTING
870 	/*
871 	 * clear embedded scope identifiers if necessary.
872 	 * in6_clearscope will touch the addresses only when necessary.
873 	 */
874 	in6_clearscope(&ip6->ip6_src);
875 	in6_clearscope(&ip6->ip6_dst);
876 #endif
877 
878 	/*
879 	 * Check with the firewall...
880 	 */
881         if (ip6_fw_enable && ip6_fw_chk_ptr) {
882 		u_short port = 0;
883 		m->m_pkthdr.rcvif = NULL;	/* XXX */
884 		/* If ipfw says divert, we have to just drop packet */
885 		if ((*ip6_fw_chk_ptr)(&ip6, ifp, &port, &m)) {
886 			m_freem(m);
887 			goto done;
888 		}
889 		if (!m) {
890 			error = EACCES;
891 			goto done;
892 		}
893 	}
894 
895 	/*
896 	 * If the outgoing packet contains a hop-by-hop options header,
897 	 * it must be examined and processed even by the source node.
898 	 * (RFC 2460, section 4.)
899 	 */
900 	if (exthdrs.ip6e_hbh) {
901 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
902 		u_int32_t dummy1; /* XXX unused */
903 		u_int32_t dummy2; /* XXX unused */
904 
905 #ifdef DIAGNOSTIC
906 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
907 			panic("ip6e_hbh is not continuous");
908 #endif
909 		/*
910 		 *  XXX: if we have to send an ICMPv6 error to the sender,
911 		 *       we need the M_LOOP flag since icmp6_error() expects
912 		 *       the IPv6 and the hop-by-hop options header are
913 		 *       continuous unless the flag is set.
914 		 */
915 		m->m_flags |= M_LOOP;
916 		m->m_pkthdr.rcvif = ifp;
917 		if (ip6_process_hopopts(m,
918 					(u_int8_t *)(hbh + 1),
919 					((hbh->ip6h_len + 1) << 3) -
920 					sizeof(struct ip6_hbh),
921 					&dummy1, &dummy2) < 0) {
922 			/* m was already freed at this point */
923 			error = EINVAL;/* better error? */
924 			goto done;
925 		}
926 		m->m_flags &= ~M_LOOP; /* XXX */
927 		m->m_pkthdr.rcvif = NULL;
928 	}
929 
930 #ifdef PFIL_HOOKS
931 	/*
932 	 * Run through list of hooks for output packets.
933 	 */
934 	m1 = m;
935 	pfh = pfil_hook_get(PFIL_OUT, &inet6sw[ip6_protox[IPPROTO_IPV6]].pr_pfh);
936 	for (; pfh; pfh = pfh->pfil_link.tqe_next)
937 		if (pfh->pfil_func) {
938 			rv = pfh->pfil_func(ip6, sizeof(*ip6), ifp, 1, &m1);
939 			if (rv) {
940 				error = EHOSTUNREACH;
941 				goto done;
942 			}
943 			m = m1;
944 			if (m == NULL)
945 				goto done;
946 			ip6 = mtod(m, struct ip6_hdr *);
947 		}
948 #endif /* PFIL_HOOKS */
949 	/*
950 	 * Send the packet to the outgoing interface.
951 	 * If necessary, do IPv6 fragmentation before sending.
952 	 */
953 	tlen = m->m_pkthdr.len;
954 	if (tlen <= mtu
955 #ifdef notyet
956 	    /*
957 	     * On any link that cannot convey a 1280-octet packet in one piece,
958 	     * link-specific fragmentation and reassembly must be provided at
959 	     * a layer below IPv6. [RFC 2460, sec.5]
960 	     * Thus if the interface has ability of link-level fragmentation,
961 	     * we can just send the packet even if the packet size is
962 	     * larger than the link's MTU.
963 	     * XXX: IFF_FRAGMENTABLE (or such) flag has not been defined yet...
964 	     */
965 
966 	    || ifp->if_flags & IFF_FRAGMENTABLE
967 #endif
968 	    )
969 	{
970  		/* Record statistics for this interface address. */
971  		if (ia && !(flags & IPV6_FORWARDING)) {
972  			ia->ia_ifa.if_opackets++;
973  			ia->ia_ifa.if_obytes += m->m_pkthdr.len;
974  		}
975 #ifdef IPSEC
976 		/* clean ipsec history once it goes out of the node */
977 		ipsec_delaux(m);
978 #endif
979 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
980 		goto done;
981 	} else if (mtu < IPV6_MMTU) {
982 		/*
983 		 * note that path MTU is never less than IPV6_MMTU
984 		 * (see icmp6_input).
985 		 */
986 		error = EMSGSIZE;
987 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
988 		goto bad;
989 	} else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
990 		error = EMSGSIZE;
991 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
992 		goto bad;
993 	} else {
994 		struct mbuf **mnext, *m_frgpart;
995 		struct ip6_frag *ip6f;
996 		u_int32_t id = htonl(ip6_id++);
997 		u_char nextproto;
998 
999 		/*
1000 		 * Too large for the destination or interface;
1001 		 * fragment if possible.
1002 		 * Must be able to put at least 8 bytes per fragment.
1003 		 */
1004 		hlen = unfragpartlen;
1005 		if (mtu > IPV6_MAXPACKET)
1006 			mtu = IPV6_MAXPACKET;
1007 
1008 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1009 		if (len < 8) {
1010 			error = EMSGSIZE;
1011 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
1012 			goto bad;
1013 		}
1014 
1015 		mnext = &m->m_nextpkt;
1016 
1017 		/*
1018 		 * Change the next header field of the last header in the
1019 		 * unfragmentable part.
1020 		 */
1021 		if (exthdrs.ip6e_rthdr) {
1022 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1023 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1024 		} else if (exthdrs.ip6e_dest1) {
1025 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1026 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1027 		} else if (exthdrs.ip6e_hbh) {
1028 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1029 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1030 		} else {
1031 			nextproto = ip6->ip6_nxt;
1032 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1033 		}
1034 
1035 		/*
1036 		 * Loop through length of segment after first fragment,
1037 		 * make new header and copy data of each part and link onto
1038 		 * chain.
1039 		 */
1040 		m0 = m;
1041 		for (off = hlen; off < tlen; off += len) {
1042 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
1043 			if (!m) {
1044 				error = ENOBUFS;
1045 				ip6stat.ip6s_odropped++;
1046 				goto sendorfree;
1047 			}
1048 			m->m_pkthdr.rcvif = NULL;
1049 			m->m_flags = m0->m_flags & M_COPYFLAGS;
1050 			*mnext = m;
1051 			mnext = &m->m_nextpkt;
1052 			m->m_data += max_linkhdr;
1053 			mhip6 = mtod(m, struct ip6_hdr *);
1054 			*mhip6 = *ip6;
1055 			m->m_len = sizeof(*mhip6);
1056  			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1057  			if (error) {
1058 				ip6stat.ip6s_odropped++;
1059 				goto sendorfree;
1060 			}
1061 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1062 			if (off + len >= tlen)
1063 				len = tlen - off;
1064 			else
1065 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1066 			mhip6->ip6_plen = htons((u_short)(len + hlen +
1067 							  sizeof(*ip6f) -
1068 							  sizeof(struct ip6_hdr)));
1069 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1070 				error = ENOBUFS;
1071 				ip6stat.ip6s_odropped++;
1072 				goto sendorfree;
1073 			}
1074 			m_cat(m, m_frgpart);
1075 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1076 			m->m_pkthdr.rcvif = (struct ifnet *)0;
1077 			ip6f->ip6f_reserved = 0;
1078 			ip6f->ip6f_ident = id;
1079 			ip6f->ip6f_nxt = nextproto;
1080 			ip6stat.ip6s_ofragments++;
1081 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1082 		}
1083 
1084 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1085 	}
1086 
1087 	/*
1088 	 * Remove leading garbages.
1089 	 */
1090 sendorfree:
1091 	m = m0->m_nextpkt;
1092 	m0->m_nextpkt = 0;
1093 	m_freem(m0);
1094 	for (m0 = m; m; m = m0) {
1095 		m0 = m->m_nextpkt;
1096 		m->m_nextpkt = 0;
1097 		if (error == 0) {
1098  			/* Record statistics for this interface address. */
1099  			if (ia) {
1100  				ia->ia_ifa.if_opackets++;
1101  				ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1102  			}
1103 #ifdef IPSEC
1104 			/* clean ipsec history once it goes out of the node */
1105 			ipsec_delaux(m);
1106 #endif
1107 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1108 		} else
1109 			m_freem(m);
1110 	}
1111 
1112 	if (error == 0)
1113 		ip6stat.ip6s_fragmented++;
1114 
1115 done:
1116 	if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1117 		RTFREE(ro->ro_rt);
1118 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1119 		RTFREE(ro_pmtu->ro_rt);
1120 	}
1121 
1122 #ifdef IPSEC
1123 	if (sp != NULL)
1124 		key_freesp(sp);
1125 #endif /* IPSEC */
1126 #ifdef FAST_IPSEC
1127 	if (sp != NULL)
1128 		KEY_FREESP(&sp);
1129 #endif /* FAST_IPSEC */
1130 
1131 	return(error);
1132 
1133 freehdrs:
1134 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1135 	m_freem(exthdrs.ip6e_dest1);
1136 	m_freem(exthdrs.ip6e_rthdr);
1137 	m_freem(exthdrs.ip6e_dest2);
1138 	/* fall through */
1139 bad:
1140 	m_freem(m);
1141 	goto done;
1142 }
1143 
1144 static int
1145 ip6_copyexthdr(mp, hdr, hlen)
1146 	struct mbuf **mp;
1147 	caddr_t hdr;
1148 	int hlen;
1149 {
1150 	struct mbuf *m;
1151 
1152 	if (hlen > MCLBYTES)
1153 		return(ENOBUFS); /* XXX */
1154 
1155 	MGET(m, M_DONTWAIT, MT_DATA);
1156 	if (!m)
1157 		return(ENOBUFS);
1158 
1159 	if (hlen > MLEN) {
1160 		MCLGET(m, M_DONTWAIT);
1161 		if ((m->m_flags & M_EXT) == 0) {
1162 			m_free(m);
1163 			return(ENOBUFS);
1164 		}
1165 	}
1166 	m->m_len = hlen;
1167 	if (hdr)
1168 		bcopy(hdr, mtod(m, caddr_t), hlen);
1169 
1170 	*mp = m;
1171 	return(0);
1172 }
1173 
1174 /*
1175  * Insert jumbo payload option.
1176  */
1177 static int
1178 ip6_insert_jumboopt(exthdrs, plen)
1179 	struct ip6_exthdrs *exthdrs;
1180 	u_int32_t plen;
1181 {
1182 	struct mbuf *mopt;
1183 	u_char *optbuf;
1184 	u_int32_t v;
1185 
1186 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1187 
1188 	/*
1189 	 * If there is no hop-by-hop options header, allocate new one.
1190 	 * If there is one but it doesn't have enough space to store the
1191 	 * jumbo payload option, allocate a cluster to store the whole options.
1192 	 * Otherwise, use it to store the options.
1193 	 */
1194 	if (exthdrs->ip6e_hbh == 0) {
1195 		MGET(mopt, M_DONTWAIT, MT_DATA);
1196 		if (mopt == 0)
1197 			return(ENOBUFS);
1198 		mopt->m_len = JUMBOOPTLEN;
1199 		optbuf = mtod(mopt, u_char *);
1200 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1201 		exthdrs->ip6e_hbh = mopt;
1202 	} else {
1203 		struct ip6_hbh *hbh;
1204 
1205 		mopt = exthdrs->ip6e_hbh;
1206 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1207 			/*
1208 			 * XXX assumption:
1209 			 * - exthdrs->ip6e_hbh is not referenced from places
1210 			 *   other than exthdrs.
1211 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1212 			 */
1213 			int oldoptlen = mopt->m_len;
1214 			struct mbuf *n;
1215 
1216 			/*
1217 			 * XXX: give up if the whole (new) hbh header does
1218 			 * not fit even in an mbuf cluster.
1219 			 */
1220 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1221 				return(ENOBUFS);
1222 
1223 			/*
1224 			 * As a consequence, we must always prepare a cluster
1225 			 * at this point.
1226 			 */
1227 			MGET(n, M_DONTWAIT, MT_DATA);
1228 			if (n) {
1229 				MCLGET(n, M_DONTWAIT);
1230 				if ((n->m_flags & M_EXT) == 0) {
1231 					m_freem(n);
1232 					n = NULL;
1233 				}
1234 			}
1235 			if (!n)
1236 				return(ENOBUFS);
1237 			n->m_len = oldoptlen + JUMBOOPTLEN;
1238 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1239 			      oldoptlen);
1240 			optbuf = mtod(n, caddr_t) + oldoptlen;
1241 			m_freem(mopt);
1242 			mopt = exthdrs->ip6e_hbh = n;
1243 		} else {
1244 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1245 			mopt->m_len += JUMBOOPTLEN;
1246 		}
1247 		optbuf[0] = IP6OPT_PADN;
1248 		optbuf[1] = 1;
1249 
1250 		/*
1251 		 * Adjust the header length according to the pad and
1252 		 * the jumbo payload option.
1253 		 */
1254 		hbh = mtod(mopt, struct ip6_hbh *);
1255 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1256 	}
1257 
1258 	/* fill in the option. */
1259 	optbuf[2] = IP6OPT_JUMBO;
1260 	optbuf[3] = 4;
1261 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1262 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1263 
1264 	/* finally, adjust the packet header length */
1265 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1266 
1267 	return(0);
1268 #undef JUMBOOPTLEN
1269 }
1270 
1271 /*
1272  * Insert fragment header and copy unfragmentable header portions.
1273  */
1274 static int
1275 ip6_insertfraghdr(m0, m, hlen, frghdrp)
1276 	struct mbuf *m0, *m;
1277 	int hlen;
1278 	struct ip6_frag **frghdrp;
1279 {
1280 	struct mbuf *n, *mlast;
1281 
1282 	if (hlen > sizeof(struct ip6_hdr)) {
1283 		n = m_copym(m0, sizeof(struct ip6_hdr),
1284 			    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1285 		if (n == 0)
1286 			return(ENOBUFS);
1287 		m->m_next = n;
1288 	} else
1289 		n = m;
1290 
1291 	/* Search for the last mbuf of unfragmentable part. */
1292 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1293 		;
1294 
1295 	if ((mlast->m_flags & M_EXT) == 0 &&
1296 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1297 		/* use the trailing space of the last mbuf for the fragment hdr */
1298 		*frghdrp =
1299 			(struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len);
1300 		mlast->m_len += sizeof(struct ip6_frag);
1301 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1302 	} else {
1303 		/* allocate a new mbuf for the fragment header */
1304 		struct mbuf *mfrg;
1305 
1306 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1307 		if (mfrg == 0)
1308 			return(ENOBUFS);
1309 		mfrg->m_len = sizeof(struct ip6_frag);
1310 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1311 		mlast->m_next = mfrg;
1312 	}
1313 
1314 	return(0);
1315 }
1316 
1317 /*
1318  * IP6 socket option processing.
1319  */
1320 int
1321 ip6_ctloutput(so, sopt)
1322 	struct socket *so;
1323 	struct sockopt *sopt;
1324 {
1325 	int privileged;
1326 	struct inpcb *in6p = sotoinpcb(so);
1327 	int error, optval;
1328 	int level, op, optname;
1329 	int optlen;
1330 	struct thread *td;
1331 
1332 	if (sopt) {
1333 		level = sopt->sopt_level;
1334 		op = sopt->sopt_dir;
1335 		optname = sopt->sopt_name;
1336 		optlen = sopt->sopt_valsize;
1337 		td = sopt->sopt_td;
1338 	} else {
1339 		panic("ip6_ctloutput: arg soopt is NULL");
1340 	}
1341 	error = optval = 0;
1342 
1343 	privileged = (td == 0 || suser(td)) ? 0 : 1;
1344 
1345 	if (level == IPPROTO_IPV6) {
1346 		switch (op) {
1347 
1348 		case SOPT_SET:
1349 			switch (optname) {
1350 			case IPV6_PKTOPTIONS:
1351 			{
1352 				struct mbuf *m;
1353 
1354 				error = soopt_getm(sopt, &m); /* XXX */
1355 				if (error != 0)
1356 					break;
1357 				error = soopt_mcopyin(sopt, m); /* XXX */
1358 				if (error != 0)
1359 					break;
1360 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1361 						    m, so, sopt);
1362 				m_freem(m); /* XXX */
1363 				break;
1364 			}
1365 
1366 			/*
1367 			 * Use of some Hop-by-Hop options or some
1368 			 * Destination options, might require special
1369 			 * privilege.  That is, normal applications
1370 			 * (without special privilege) might be forbidden
1371 			 * from setting certain options in outgoing packets,
1372 			 * and might never see certain options in received
1373 			 * packets. [RFC 2292 Section 6]
1374 			 * KAME specific note:
1375 			 *  KAME prevents non-privileged users from sending or
1376 			 *  receiving ANY hbh/dst options in order to avoid
1377 			 *  overhead of parsing options in the kernel.
1378 			 */
1379 			case IPV6_UNICAST_HOPS:
1380 			case IPV6_CHECKSUM:
1381 			case IPV6_FAITH:
1382 
1383 			case IPV6_V6ONLY:
1384 				if (optlen != sizeof(int)) {
1385 					error = EINVAL;
1386 					break;
1387 				}
1388 				error = sooptcopyin(sopt, &optval,
1389 					sizeof optval, sizeof optval);
1390 				if (error)
1391 					break;
1392 				switch (optname) {
1393 
1394 				case IPV6_UNICAST_HOPS:
1395 					if (optval < -1 || optval >= 256)
1396 						error = EINVAL;
1397 					else {
1398 						/* -1 = kernel default */
1399 						in6p->in6p_hops = optval;
1400 
1401 						if ((in6p->in6p_vflag &
1402 						     INP_IPV4) != 0)
1403 							in6p->inp_ip_ttl = optval;
1404 					}
1405 					break;
1406 #define OPTSET(bit) \
1407 do { \
1408 	if (optval) \
1409 		in6p->in6p_flags |= (bit); \
1410 	else \
1411 		in6p->in6p_flags &= ~(bit); \
1412 } while (0)
1413 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1414 
1415 				case IPV6_CHECKSUM:
1416 					in6p->in6p_cksum = optval;
1417 					break;
1418 
1419 				case IPV6_FAITH:
1420 					OPTSET(IN6P_FAITH);
1421 					break;
1422 
1423 				case IPV6_V6ONLY:
1424 					/*
1425 					 * make setsockopt(IPV6_V6ONLY)
1426 					 * available only prior to bind(2).
1427 					 * see ipng mailing list, Jun 22 2001.
1428 					 */
1429 					if (in6p->in6p_lport ||
1430 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
1431 					{
1432 						error = EINVAL;
1433 						break;
1434 					}
1435 					OPTSET(IN6P_IPV6_V6ONLY);
1436 					if (optval)
1437 						in6p->in6p_vflag &= ~INP_IPV4;
1438 					else
1439 						in6p->in6p_vflag |= INP_IPV4;
1440 					break;
1441 				}
1442 				break;
1443 
1444 			case IPV6_PKTINFO:
1445 			case IPV6_HOPLIMIT:
1446 			case IPV6_HOPOPTS:
1447 			case IPV6_DSTOPTS:
1448 			case IPV6_RTHDR:
1449 				/* RFC 2292 */
1450 				if (optlen != sizeof(int)) {
1451 					error = EINVAL;
1452 					break;
1453 				}
1454 				error = sooptcopyin(sopt, &optval,
1455 					sizeof optval, sizeof optval);
1456 				if (error)
1457 					break;
1458 				switch (optname) {
1459 				case IPV6_PKTINFO:
1460 					OPTSET(IN6P_PKTINFO);
1461 					break;
1462 				case IPV6_HOPLIMIT:
1463 					OPTSET(IN6P_HOPLIMIT);
1464 					break;
1465 				case IPV6_HOPOPTS:
1466 					/*
1467 					 * Check super-user privilege.
1468 					 * See comments for IPV6_RECVHOPOPTS.
1469 					 */
1470 					if (!privileged)
1471 						return(EPERM);
1472 					OPTSET(IN6P_HOPOPTS);
1473 					break;
1474 				case IPV6_DSTOPTS:
1475 					if (!privileged)
1476 						return(EPERM);
1477 					OPTSET(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1478 					break;
1479 				case IPV6_RTHDR:
1480 					OPTSET(IN6P_RTHDR);
1481 					break;
1482 				}
1483 				break;
1484 #undef OPTSET
1485 
1486 			case IPV6_MULTICAST_IF:
1487 			case IPV6_MULTICAST_HOPS:
1488 			case IPV6_MULTICAST_LOOP:
1489 			case IPV6_JOIN_GROUP:
1490 			case IPV6_LEAVE_GROUP:
1491 			    {
1492 				struct mbuf *m;
1493 				if (sopt->sopt_valsize > MLEN) {
1494 					error = EMSGSIZE;
1495 					break;
1496 				}
1497 				/* XXX */
1498 				MGET(m, sopt->sopt_td ? M_TRYWAIT : M_DONTWAIT, MT_HEADER);
1499 				if (m == 0) {
1500 					error = ENOBUFS;
1501 					break;
1502 				}
1503 				m->m_len = sopt->sopt_valsize;
1504 				error = sooptcopyin(sopt, mtod(m, char *),
1505 						    m->m_len, m->m_len);
1506 				error =	ip6_setmoptions(sopt->sopt_name,
1507 							&in6p->in6p_moptions,
1508 							m);
1509 				(void)m_free(m);
1510 			    }
1511 				break;
1512 
1513 			case IPV6_PORTRANGE:
1514 				error = sooptcopyin(sopt, &optval,
1515 				    sizeof optval, sizeof optval);
1516 				if (error)
1517 					break;
1518 
1519 				switch (optval) {
1520 				case IPV6_PORTRANGE_DEFAULT:
1521 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1522 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1523 					break;
1524 
1525 				case IPV6_PORTRANGE_HIGH:
1526 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1527 					in6p->in6p_flags |= IN6P_HIGHPORT;
1528 					break;
1529 
1530 				case IPV6_PORTRANGE_LOW:
1531 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1532 					in6p->in6p_flags |= IN6P_LOWPORT;
1533 					break;
1534 
1535 				default:
1536 					error = EINVAL;
1537 					break;
1538 				}
1539 				break;
1540 
1541 #if defined(IPSEC) || defined(FAST_IPSEC)
1542 			case IPV6_IPSEC_POLICY:
1543 			    {
1544 				caddr_t req = NULL;
1545 				size_t len = 0;
1546 				struct mbuf *m;
1547 
1548 				if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1549 					break;
1550 				if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1551 					break;
1552 				if (m) {
1553 					req = mtod(m, caddr_t);
1554 					len = m->m_len;
1555 				}
1556 				error = ipsec6_set_policy(in6p, optname, req,
1557 				                          len, privileged);
1558 				m_freem(m);
1559 			    }
1560 				break;
1561 #endif /* KAME IPSEC */
1562 
1563 			case IPV6_FW_ADD:
1564 			case IPV6_FW_DEL:
1565 			case IPV6_FW_FLUSH:
1566 			case IPV6_FW_ZERO:
1567 			    {
1568 				struct mbuf *m;
1569 				struct mbuf **mp = &m;
1570 
1571 				if (ip6_fw_ctl_ptr == NULL)
1572 					return EINVAL;
1573 				/* XXX */
1574 				if ((error = soopt_getm(sopt, &m)) != 0)
1575 					break;
1576 				/* XXX */
1577 				if ((error = soopt_mcopyin(sopt, m)) != 0)
1578 					break;
1579 				error = (*ip6_fw_ctl_ptr)(optname, mp);
1580 				m = *mp;
1581 			    }
1582 				break;
1583 
1584 			default:
1585 				error = ENOPROTOOPT;
1586 				break;
1587 			}
1588 			break;
1589 
1590 		case SOPT_GET:
1591 			switch (optname) {
1592 
1593 			case IPV6_PKTOPTIONS:
1594 				if (in6p->in6p_options) {
1595 					struct mbuf *m;
1596 					m = m_copym(in6p->in6p_options,
1597 					    0, M_COPYALL, M_TRYWAIT);
1598 					error = soopt_mcopyout(sopt, m);
1599 					if (error == 0)
1600 						m_freem(m);
1601 				} else
1602 					sopt->sopt_valsize = 0;
1603 				break;
1604 
1605 			case IPV6_UNICAST_HOPS:
1606 			case IPV6_CHECKSUM:
1607 
1608 			case IPV6_FAITH:
1609 			case IPV6_V6ONLY:
1610 			case IPV6_PORTRANGE:
1611 				switch (optname) {
1612 
1613 				case IPV6_UNICAST_HOPS:
1614 					optval = in6p->in6p_hops;
1615 					break;
1616 
1617 				case IPV6_CHECKSUM:
1618 					optval = in6p->in6p_cksum;
1619 					break;
1620 
1621 				case IPV6_FAITH:
1622 					optval = OPTBIT(IN6P_FAITH);
1623 					break;
1624 
1625 				case IPV6_V6ONLY:
1626 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
1627 					break;
1628 
1629 				case IPV6_PORTRANGE:
1630 				    {
1631 					int flags;
1632 					flags = in6p->in6p_flags;
1633 					if (flags & IN6P_HIGHPORT)
1634 						optval = IPV6_PORTRANGE_HIGH;
1635 					else if (flags & IN6P_LOWPORT)
1636 						optval = IPV6_PORTRANGE_LOW;
1637 					else
1638 						optval = 0;
1639 					break;
1640 				    }
1641 				}
1642 				error = sooptcopyout(sopt, &optval,
1643 					sizeof optval);
1644 				break;
1645 
1646 			case IPV6_PKTINFO:
1647 			case IPV6_HOPLIMIT:
1648 			case IPV6_HOPOPTS:
1649 			case IPV6_RTHDR:
1650 			case IPV6_DSTOPTS:
1651 				if (optname == IPV6_HOPOPTS ||
1652 				    optname == IPV6_DSTOPTS ||
1653 				    !privileged)
1654 					return(EPERM);
1655 				switch (optname) {
1656 				case IPV6_PKTINFO:
1657 					optval = OPTBIT(IN6P_PKTINFO);
1658 					break;
1659 				case IPV6_HOPLIMIT:
1660 					optval = OPTBIT(IN6P_HOPLIMIT);
1661 					break;
1662 				case IPV6_HOPOPTS:
1663 					if (!privileged)
1664 						return(EPERM);
1665 					optval = OPTBIT(IN6P_HOPOPTS);
1666 					break;
1667 				case IPV6_RTHDR:
1668 					optval = OPTBIT(IN6P_RTHDR);
1669 					break;
1670 				case IPV6_DSTOPTS:
1671 					if (!privileged)
1672 						return(EPERM);
1673 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1674 					break;
1675 				}
1676 				error = sooptcopyout(sopt, &optval,
1677 					sizeof optval);
1678 				break;
1679 
1680 			case IPV6_MULTICAST_IF:
1681 			case IPV6_MULTICAST_HOPS:
1682 			case IPV6_MULTICAST_LOOP:
1683 			case IPV6_JOIN_GROUP:
1684 			case IPV6_LEAVE_GROUP:
1685 			    {
1686 				struct mbuf *m;
1687 				error = ip6_getmoptions(sopt->sopt_name,
1688 						in6p->in6p_moptions, &m);
1689 				if (error == 0)
1690 					error = sooptcopyout(sopt,
1691 						mtod(m, char *), m->m_len);
1692 				m_freem(m);
1693 			    }
1694 				break;
1695 
1696 #if defined(IPSEC) || defined(FAST_IPSEC)
1697 			case IPV6_IPSEC_POLICY:
1698 			  {
1699 				caddr_t req = NULL;
1700 				size_t len = 0;
1701 				struct mbuf *m = NULL;
1702 				struct mbuf **mp = &m;
1703 
1704 				error = soopt_getm(sopt, &m); /* XXX */
1705 				if (error != 0)
1706 					break;
1707 				error = soopt_mcopyin(sopt, m); /* XXX */
1708 				if (error != 0)
1709 					break;
1710 				if (m) {
1711 					req = mtod(m, caddr_t);
1712 					len = m->m_len;
1713 				}
1714 				error = ipsec6_get_policy(in6p, req, len, mp);
1715 				if (error == 0)
1716 					error = soopt_mcopyout(sopt, m); /*XXX*/
1717 				if (error == 0 && m)
1718 					m_freem(m);
1719 				break;
1720 			  }
1721 #endif /* KAME IPSEC */
1722 
1723 			case IPV6_FW_GET:
1724 			  {
1725 				struct mbuf *m;
1726 				struct mbuf **mp = &m;
1727 
1728 				if (ip6_fw_ctl_ptr == NULL)
1729 			        {
1730 					return EINVAL;
1731 				}
1732 				error = (*ip6_fw_ctl_ptr)(optname, mp);
1733 				if (error == 0)
1734 					error = soopt_mcopyout(sopt, m); /* XXX */
1735 				if (error == 0 && m)
1736 					m_freem(m);
1737 			  }
1738 				break;
1739 
1740 			default:
1741 				error = ENOPROTOOPT;
1742 				break;
1743 			}
1744 			break;
1745 		}
1746 	} else {
1747 		error = EINVAL;
1748 	}
1749 	return(error);
1750 }
1751 
1752 /*
1753  * Set up IP6 options in pcb for insertion in output packets or
1754  * specifying behavior of outgoing packets.
1755  */
1756 static int
1757 ip6_pcbopts(pktopt, m, so, sopt)
1758 	struct ip6_pktopts **pktopt;
1759 	struct mbuf *m;
1760 	struct socket *so;
1761 	struct sockopt *sopt;
1762 {
1763 	struct ip6_pktopts *opt = *pktopt;
1764 	int error = 0;
1765 	struct thread *td = sopt->sopt_td;
1766 	int priv = 0;
1767 
1768 	/* turn off any old options. */
1769 	if (opt) {
1770 #ifdef DIAGNOSTIC
1771 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
1772 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
1773 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
1774 			printf("ip6_pcbopts: all specified options are cleared.\n");
1775 #endif
1776 		ip6_clearpktopts(opt, 1, -1);
1777 	} else
1778 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
1779 	*pktopt = NULL;
1780 
1781 	if (!m || m->m_len == 0) {
1782 		/*
1783 		 * Only turning off any previous options, regardless of
1784 		 * whether the opt is just created or given.
1785 		 */
1786 		free(opt, M_IP6OPT);
1787 		return(0);
1788 	}
1789 
1790 	/*  set options specified by user. */
1791 	if (td && !suser(td))
1792 		priv = 1;
1793 	if ((error = ip6_setpktoptions(m, opt, priv, 1)) != 0) {
1794 		ip6_clearpktopts(opt, 1, -1); /* XXX: discard all options */
1795 		free(opt, M_IP6OPT);
1796 		return(error);
1797 	}
1798 	*pktopt = opt;
1799 	return(0);
1800 }
1801 
1802 /*
1803  * initialize ip6_pktopts.  beware that there are non-zero default values in
1804  * the struct.
1805  */
1806 void
1807 init_ip6pktopts(opt)
1808 	struct ip6_pktopts *opt;
1809 {
1810 
1811 	bzero(opt, sizeof(*opt));
1812 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
1813 }
1814 
1815 void
1816 ip6_clearpktopts(pktopt, needfree, optname)
1817 	struct ip6_pktopts *pktopt;
1818 	int needfree, optname;
1819 {
1820 	if (pktopt == NULL)
1821 		return;
1822 
1823 	if (optname == -1) {
1824 		if (needfree && pktopt->ip6po_pktinfo)
1825 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
1826 		pktopt->ip6po_pktinfo = NULL;
1827 	}
1828 	if (optname == -1)
1829 		pktopt->ip6po_hlim = -1;
1830 	if (optname == -1) {
1831 		if (needfree && pktopt->ip6po_nexthop)
1832 			free(pktopt->ip6po_nexthop, M_IP6OPT);
1833 		pktopt->ip6po_nexthop = NULL;
1834 	}
1835 	if (optname == -1) {
1836 		if (needfree && pktopt->ip6po_hbh)
1837 			free(pktopt->ip6po_hbh, M_IP6OPT);
1838 		pktopt->ip6po_hbh = NULL;
1839 	}
1840 	if (optname == -1) {
1841 		if (needfree && pktopt->ip6po_dest1)
1842 			free(pktopt->ip6po_dest1, M_IP6OPT);
1843 		pktopt->ip6po_dest1 = NULL;
1844 	}
1845 	if (optname == -1) {
1846 		if (needfree && pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
1847 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
1848 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
1849 		if (pktopt->ip6po_route.ro_rt) {
1850 			RTFREE(pktopt->ip6po_route.ro_rt);
1851 			pktopt->ip6po_route.ro_rt = NULL;
1852 		}
1853 	}
1854 	if (optname == -1) {
1855 		if (needfree && pktopt->ip6po_dest2)
1856 			free(pktopt->ip6po_dest2, M_IP6OPT);
1857 		pktopt->ip6po_dest2 = NULL;
1858 	}
1859 }
1860 
1861 #define PKTOPT_EXTHDRCPY(type) \
1862 do {\
1863 	if (src->type) {\
1864 		int hlen =\
1865 			(((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
1866 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
1867 		if (dst->type == NULL && canwait == M_NOWAIT)\
1868 			goto bad;\
1869 		bcopy(src->type, dst->type, hlen);\
1870 	}\
1871 } while (0)
1872 
1873 struct ip6_pktopts *
1874 ip6_copypktopts(src, canwait)
1875 	struct ip6_pktopts *src;
1876 	int canwait;
1877 {
1878 	struct ip6_pktopts *dst;
1879 
1880 	if (src == NULL) {
1881 		printf("ip6_clearpktopts: invalid argument\n");
1882 		return(NULL);
1883 	}
1884 
1885 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
1886 	if (dst == NULL && canwait == M_NOWAIT)
1887 		return (NULL);
1888 	bzero(dst, sizeof(*dst));
1889 
1890 	dst->ip6po_hlim = src->ip6po_hlim;
1891 	if (src->ip6po_pktinfo) {
1892 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
1893 					    M_IP6OPT, canwait);
1894 		if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT)
1895 			goto bad;
1896 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
1897 	}
1898 	if (src->ip6po_nexthop) {
1899 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
1900 					    M_IP6OPT, canwait);
1901 		if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT)
1902 			goto bad;
1903 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
1904 		      src->ip6po_nexthop->sa_len);
1905 	}
1906 	PKTOPT_EXTHDRCPY(ip6po_hbh);
1907 	PKTOPT_EXTHDRCPY(ip6po_dest1);
1908 	PKTOPT_EXTHDRCPY(ip6po_dest2);
1909 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
1910 	return(dst);
1911 
1912   bad:
1913 	if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
1914 	if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
1915 	if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
1916 	if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
1917 	if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
1918 	if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
1919 	free(dst, M_IP6OPT);
1920 	return(NULL);
1921 }
1922 #undef PKTOPT_EXTHDRCPY
1923 
1924 void
1925 ip6_freepcbopts(pktopt)
1926 	struct ip6_pktopts *pktopt;
1927 {
1928 	if (pktopt == NULL)
1929 		return;
1930 
1931 	ip6_clearpktopts(pktopt, 1, -1);
1932 
1933 	free(pktopt, M_IP6OPT);
1934 }
1935 
1936 /*
1937  * Set the IP6 multicast options in response to user setsockopt().
1938  */
1939 static int
1940 ip6_setmoptions(optname, im6op, m)
1941 	int optname;
1942 	struct ip6_moptions **im6op;
1943 	struct mbuf *m;
1944 {
1945 	int error = 0;
1946 	u_int loop, ifindex;
1947 	struct ipv6_mreq *mreq;
1948 	struct ifnet *ifp;
1949 	struct ip6_moptions *im6o = *im6op;
1950 	struct route_in6 ro;
1951 	struct sockaddr_in6 *dst;
1952 	struct in6_multi_mship *imm;
1953 	struct thread *td = curthread;	/* XXX */
1954 
1955 	if (im6o == NULL) {
1956 		/*
1957 		 * No multicast option buffer attached to the pcb;
1958 		 * allocate one and initialize to default values.
1959 		 */
1960 		im6o = (struct ip6_moptions *)
1961 			malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
1962 
1963 		if (im6o == NULL)
1964 			return(ENOBUFS);
1965 		*im6op = im6o;
1966 		im6o->im6o_multicast_ifp = NULL;
1967 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1968 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
1969 		LIST_INIT(&im6o->im6o_memberships);
1970 	}
1971 
1972 	switch (optname) {
1973 
1974 	case IPV6_MULTICAST_IF:
1975 		/*
1976 		 * Select the interface for outgoing multicast packets.
1977 		 */
1978 		if (m == NULL || m->m_len != sizeof(u_int)) {
1979 			error = EINVAL;
1980 			break;
1981 		}
1982 		bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
1983 		if (ifindex < 0 || if_index < ifindex) {
1984 			error = ENXIO;	/* XXX EINVAL? */
1985 			break;
1986 		}
1987 		ifp = ifnet_byindex(ifindex);
1988 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1989 			error = EADDRNOTAVAIL;
1990 			break;
1991 		}
1992 		im6o->im6o_multicast_ifp = ifp;
1993 		break;
1994 
1995 	case IPV6_MULTICAST_HOPS:
1996 	    {
1997 		/*
1998 		 * Set the IP6 hoplimit for outgoing multicast packets.
1999 		 */
2000 		int optval;
2001 		if (m == NULL || m->m_len != sizeof(int)) {
2002 			error = EINVAL;
2003 			break;
2004 		}
2005 		bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2006 		if (optval < -1 || optval >= 256)
2007 			error = EINVAL;
2008 		else if (optval == -1)
2009 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2010 		else
2011 			im6o->im6o_multicast_hlim = optval;
2012 		break;
2013 	    }
2014 
2015 	case IPV6_MULTICAST_LOOP:
2016 		/*
2017 		 * Set the loopback flag for outgoing multicast packets.
2018 		 * Must be zero or one.
2019 		 */
2020 		if (m == NULL || m->m_len != sizeof(u_int)) {
2021 			error = EINVAL;
2022 			break;
2023 		}
2024 		bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2025 		if (loop > 1) {
2026 			error = EINVAL;
2027 			break;
2028 		}
2029 		im6o->im6o_multicast_loop = loop;
2030 		break;
2031 
2032 	case IPV6_JOIN_GROUP:
2033 		/*
2034 		 * Add a multicast group membership.
2035 		 * Group must be a valid IP6 multicast address.
2036 		 */
2037 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2038 			error = EINVAL;
2039 			break;
2040 		}
2041 		mreq = mtod(m, struct ipv6_mreq *);
2042 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2043 			/*
2044 			 * We use the unspecified address to specify to accept
2045 			 * all multicast addresses. Only super user is allowed
2046 			 * to do this.
2047 			 */
2048 			if (suser(td))
2049 			{
2050 				error = EACCES;
2051 				break;
2052 			}
2053 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2054 			error = EINVAL;
2055 			break;
2056 		}
2057 
2058 		/*
2059 		 * If the interface is specified, validate it.
2060 		 */
2061 		if (mreq->ipv6mr_interface < 0
2062 		 || if_index < mreq->ipv6mr_interface) {
2063 			error = ENXIO;	/* XXX EINVAL? */
2064 			break;
2065 		}
2066 		/*
2067 		 * If no interface was explicitly specified, choose an
2068 		 * appropriate one according to the given multicast address.
2069 		 */
2070 		if (mreq->ipv6mr_interface == 0) {
2071 			/*
2072 			 * If the multicast address is in node-local scope,
2073 			 * the interface should be a loopback interface.
2074 			 * Otherwise, look up the routing table for the
2075 			 * address, and choose the outgoing interface.
2076 			 *   XXX: is it a good approach?
2077 			 */
2078 			if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
2079 				ifp = &loif[0];
2080 			} else {
2081 				ro.ro_rt = NULL;
2082 				dst = (struct sockaddr_in6 *)&ro.ro_dst;
2083 				bzero(dst, sizeof(*dst));
2084 				dst->sin6_len = sizeof(struct sockaddr_in6);
2085 				dst->sin6_family = AF_INET6;
2086 				dst->sin6_addr = mreq->ipv6mr_multiaddr;
2087 				rtalloc((struct route *)&ro);
2088 				if (ro.ro_rt == NULL) {
2089 					error = EADDRNOTAVAIL;
2090 					break;
2091 				}
2092 				ifp = ro.ro_rt->rt_ifp;
2093 				rtfree(ro.ro_rt);
2094 			}
2095 		} else
2096 			ifp = ifnet_byindex(mreq->ipv6mr_interface);
2097 
2098 		/*
2099 		 * See if we found an interface, and confirm that it
2100 		 * supports multicast
2101 		 */
2102 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2103 			error = EADDRNOTAVAIL;
2104 			break;
2105 		}
2106 		/*
2107 		 * Put interface index into the multicast address,
2108 		 * if the address has link-local scope.
2109 		 */
2110 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2111 			mreq->ipv6mr_multiaddr.s6_addr16[1]
2112 				= htons(mreq->ipv6mr_interface);
2113 		}
2114 		/*
2115 		 * See if the membership already exists.
2116 		 */
2117 		for (imm = im6o->im6o_memberships.lh_first;
2118 		     imm != NULL; imm = imm->i6mm_chain.le_next)
2119 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
2120 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2121 					       &mreq->ipv6mr_multiaddr))
2122 				break;
2123 		if (imm != NULL) {
2124 			error = EADDRINUSE;
2125 			break;
2126 		}
2127 		/*
2128 		 * Everything looks good; add a new record to the multicast
2129 		 * address list for the given interface.
2130 		 */
2131 		imm = malloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
2132 		if (imm == NULL) {
2133 			error = ENOBUFS;
2134 			break;
2135 		}
2136 		if ((imm->i6mm_maddr =
2137 		     in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
2138 			free(imm, M_IPMADDR);
2139 			break;
2140 		}
2141 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2142 		break;
2143 
2144 	case IPV6_LEAVE_GROUP:
2145 		/*
2146 		 * Drop a multicast group membership.
2147 		 * Group must be a valid IP6 multicast address.
2148 		 */
2149 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2150 			error = EINVAL;
2151 			break;
2152 		}
2153 		mreq = mtod(m, struct ipv6_mreq *);
2154 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2155 			if (suser(td)) {
2156 				error = EACCES;
2157 				break;
2158 			}
2159 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2160 			error = EINVAL;
2161 			break;
2162 		}
2163 		/*
2164 		 * If an interface address was specified, get a pointer
2165 		 * to its ifnet structure.
2166 		 */
2167 		if (mreq->ipv6mr_interface < 0
2168 		 || if_index < mreq->ipv6mr_interface) {
2169 			error = ENXIO;	/* XXX EINVAL? */
2170 			break;
2171 		}
2172 		ifp = ifnet_byindex(mreq->ipv6mr_interface);
2173 		/*
2174 		 * Put interface index into the multicast address,
2175 		 * if the address has link-local scope.
2176 		 */
2177 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2178 			mreq->ipv6mr_multiaddr.s6_addr16[1]
2179 				= htons(mreq->ipv6mr_interface);
2180 		}
2181 		/*
2182 		 * Find the membership in the membership list.
2183 		 */
2184 		for (imm = im6o->im6o_memberships.lh_first;
2185 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
2186 			if ((ifp == NULL ||
2187 			     imm->i6mm_maddr->in6m_ifp == ifp) &&
2188 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2189 					       &mreq->ipv6mr_multiaddr))
2190 				break;
2191 		}
2192 		if (imm == NULL) {
2193 			/* Unable to resolve interface */
2194 			error = EADDRNOTAVAIL;
2195 			break;
2196 		}
2197 		/*
2198 		 * Give up the multicast address record to which the
2199 		 * membership points.
2200 		 */
2201 		LIST_REMOVE(imm, i6mm_chain);
2202 		in6_delmulti(imm->i6mm_maddr);
2203 		free(imm, M_IPMADDR);
2204 		break;
2205 
2206 	default:
2207 		error = EOPNOTSUPP;
2208 		break;
2209 	}
2210 
2211 	/*
2212 	 * If all options have default values, no need to keep the mbuf.
2213 	 */
2214 	if (im6o->im6o_multicast_ifp == NULL &&
2215 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2216 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2217 	    im6o->im6o_memberships.lh_first == NULL) {
2218 		free(*im6op, M_IPMOPTS);
2219 		*im6op = NULL;
2220 	}
2221 
2222 	return(error);
2223 }
2224 
2225 /*
2226  * Return the IP6 multicast options in response to user getsockopt().
2227  */
2228 static int
2229 ip6_getmoptions(optname, im6o, mp)
2230 	int optname;
2231 	struct ip6_moptions *im6o;
2232 	struct mbuf **mp;
2233 {
2234 	u_int *hlim, *loop, *ifindex;
2235 
2236 	*mp = m_get(M_TRYWAIT, MT_HEADER);		/* XXX */
2237 
2238 	switch (optname) {
2239 
2240 	case IPV6_MULTICAST_IF:
2241 		ifindex = mtod(*mp, u_int *);
2242 		(*mp)->m_len = sizeof(u_int);
2243 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2244 			*ifindex = 0;
2245 		else
2246 			*ifindex = im6o->im6o_multicast_ifp->if_index;
2247 		return(0);
2248 
2249 	case IPV6_MULTICAST_HOPS:
2250 		hlim = mtod(*mp, u_int *);
2251 		(*mp)->m_len = sizeof(u_int);
2252 		if (im6o == NULL)
2253 			*hlim = ip6_defmcasthlim;
2254 		else
2255 			*hlim = im6o->im6o_multicast_hlim;
2256 		return(0);
2257 
2258 	case IPV6_MULTICAST_LOOP:
2259 		loop = mtod(*mp, u_int *);
2260 		(*mp)->m_len = sizeof(u_int);
2261 		if (im6o == NULL)
2262 			*loop = ip6_defmcasthlim;
2263 		else
2264 			*loop = im6o->im6o_multicast_loop;
2265 		return(0);
2266 
2267 	default:
2268 		return(EOPNOTSUPP);
2269 	}
2270 }
2271 
2272 /*
2273  * Discard the IP6 multicast options.
2274  */
2275 void
2276 ip6_freemoptions(im6o)
2277 	struct ip6_moptions *im6o;
2278 {
2279 	struct in6_multi_mship *imm;
2280 
2281 	if (im6o == NULL)
2282 		return;
2283 
2284 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2285 		LIST_REMOVE(imm, i6mm_chain);
2286 		if (imm->i6mm_maddr)
2287 			in6_delmulti(imm->i6mm_maddr);
2288 		free(imm, M_IPMADDR);
2289 	}
2290 	free(im6o, M_IPMOPTS);
2291 }
2292 
2293 /*
2294  * Set IPv6 outgoing packet options based on advanced API.
2295  */
2296 int
2297 ip6_setpktoptions(control, opt, priv, needcopy)
2298 	struct mbuf *control;
2299 	struct ip6_pktopts *opt;
2300 	int priv, needcopy;
2301 {
2302 	struct cmsghdr *cm = 0;
2303 
2304 	if (control == 0 || opt == 0)
2305 		return(EINVAL);
2306 
2307 	init_ip6pktopts(opt);
2308 
2309 	/*
2310 	 * XXX: Currently, we assume all the optional information is stored
2311 	 * in a single mbuf.
2312 	 */
2313 	if (control->m_next)
2314 		return(EINVAL);
2315 
2316 	for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2317 		     control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2318 		cm = mtod(control, struct cmsghdr *);
2319 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2320 			return(EINVAL);
2321 		if (cm->cmsg_level != IPPROTO_IPV6)
2322 			continue;
2323 
2324 		/*
2325 		 * XXX should check if RFC2292 API is mixed with 2292bis API
2326 		 */
2327 		switch (cm->cmsg_type) {
2328 		case IPV6_PKTINFO:
2329 			if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo)))
2330 				return(EINVAL);
2331 			if (needcopy) {
2332 				/* XXX: Is it really WAITOK? */
2333 				opt->ip6po_pktinfo =
2334 					malloc(sizeof(struct in6_pktinfo),
2335 					       M_IP6OPT, M_WAITOK);
2336 				bcopy(CMSG_DATA(cm), opt->ip6po_pktinfo,
2337 				    sizeof(struct in6_pktinfo));
2338 			} else
2339 				opt->ip6po_pktinfo =
2340 					(struct in6_pktinfo *)CMSG_DATA(cm);
2341 			if (opt->ip6po_pktinfo->ipi6_ifindex &&
2342 			    IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr))
2343 				opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] =
2344 					htons(opt->ip6po_pktinfo->ipi6_ifindex);
2345 
2346 			if (opt->ip6po_pktinfo->ipi6_ifindex > if_index
2347 			 || opt->ip6po_pktinfo->ipi6_ifindex < 0) {
2348 				return(ENXIO);
2349 			}
2350 
2351 			/*
2352 			 * Check if the requested source address is indeed a
2353 			 * unicast address assigned to the node, and can be
2354 			 * used as the packet's source address.
2355 			 */
2356 			if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
2357 				struct in6_ifaddr *ia6;
2358 				struct sockaddr_in6 sin6;
2359 
2360 				bzero(&sin6, sizeof(sin6));
2361 				sin6.sin6_len = sizeof(sin6);
2362 				sin6.sin6_family = AF_INET6;
2363 				sin6.sin6_addr =
2364 					opt->ip6po_pktinfo->ipi6_addr;
2365 				ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(sin6tosa(&sin6));
2366 				if (ia6 == NULL ||
2367 				    (ia6->ia6_flags & (IN6_IFF_ANYCAST |
2368 						       IN6_IFF_NOTREADY)) != 0)
2369 					return(EADDRNOTAVAIL);
2370 			}
2371 			break;
2372 
2373 		case IPV6_HOPLIMIT:
2374 			if (cm->cmsg_len != CMSG_LEN(sizeof(int)))
2375 				return(EINVAL);
2376 
2377 			opt->ip6po_hlim = *(int *)CMSG_DATA(cm);
2378 			if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255)
2379 				return(EINVAL);
2380 			break;
2381 
2382 		case IPV6_NEXTHOP:
2383 			if (!priv)
2384 				return(EPERM);
2385 
2386 			if (cm->cmsg_len < sizeof(u_char) ||
2387 			    /* check if cmsg_len is large enough for sa_len */
2388 			    cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm)))
2389 				return(EINVAL);
2390 
2391 			if (needcopy) {
2392 				opt->ip6po_nexthop =
2393 					malloc(*CMSG_DATA(cm),
2394 					       M_IP6OPT, M_WAITOK);
2395 				bcopy(CMSG_DATA(cm),
2396 				      opt->ip6po_nexthop,
2397 				      *CMSG_DATA(cm));
2398 			} else
2399 				opt->ip6po_nexthop =
2400 					(struct sockaddr *)CMSG_DATA(cm);
2401 			break;
2402 
2403 		case IPV6_HOPOPTS:
2404 		{
2405 			struct ip6_hbh *hbh;
2406 			int hbhlen;
2407 
2408 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh)))
2409 				return(EINVAL);
2410 			hbh = (struct ip6_hbh *)CMSG_DATA(cm);
2411 			hbhlen = (hbh->ip6h_len + 1) << 3;
2412 			if (cm->cmsg_len != CMSG_LEN(hbhlen))
2413 				return(EINVAL);
2414 
2415 			if (needcopy) {
2416 				opt->ip6po_hbh =
2417 					malloc(hbhlen, M_IP6OPT, M_WAITOK);
2418 				bcopy(hbh, opt->ip6po_hbh, hbhlen);
2419 			} else
2420 				opt->ip6po_hbh = hbh;
2421 			break;
2422 		}
2423 
2424 		case IPV6_DSTOPTS:
2425 		{
2426 			struct ip6_dest *dest, **newdest;
2427 			int destlen;
2428 
2429 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest)))
2430 				return(EINVAL);
2431 			dest = (struct ip6_dest *)CMSG_DATA(cm);
2432 			destlen = (dest->ip6d_len + 1) << 3;
2433 			if (cm->cmsg_len != CMSG_LEN(destlen))
2434 				return(EINVAL);
2435 
2436 			/*
2437 			 * The old advacned API is ambiguous on this
2438 			 * point. Our approach is to determine the
2439 			 * position based according to the existence
2440 			 * of a routing header. Note, however, that
2441 			 * this depends on the order of the extension
2442 			 * headers in the ancillary data; the 1st part
2443 			 * of the destination options header must
2444 			 * appear before the routing header in the
2445 			 * ancillary data, too.
2446 			 * RFC2292bis solved the ambiguity by
2447 			 * introducing separate cmsg types.
2448 			 */
2449 			if (opt->ip6po_rthdr == NULL)
2450 				newdest = &opt->ip6po_dest1;
2451 			else
2452 				newdest = &opt->ip6po_dest2;
2453 
2454 			if (needcopy) {
2455 				*newdest = malloc(destlen, M_IP6OPT, M_WAITOK);
2456 				bcopy(dest, *newdest, destlen);
2457 			} else
2458 				*newdest = dest;
2459 
2460 			break;
2461 		}
2462 
2463 		case IPV6_RTHDR:
2464 		{
2465 			struct ip6_rthdr *rth;
2466 			int rthlen;
2467 
2468 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr)))
2469 				return(EINVAL);
2470 			rth = (struct ip6_rthdr *)CMSG_DATA(cm);
2471 			rthlen = (rth->ip6r_len + 1) << 3;
2472 			if (cm->cmsg_len != CMSG_LEN(rthlen))
2473 				return(EINVAL);
2474 
2475 			switch (rth->ip6r_type) {
2476 			case IPV6_RTHDR_TYPE_0:
2477 				/* must contain one addr */
2478 				if (rth->ip6r_len == 0)
2479 					return(EINVAL);
2480 				/* length must be even */
2481 				if (rth->ip6r_len % 2)
2482 					return(EINVAL);
2483 				if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2484 					return(EINVAL);
2485 				break;
2486 			default:
2487 				return(EINVAL);	/* not supported */
2488 			}
2489 
2490 			if (needcopy) {
2491 				opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT,
2492 							  M_WAITOK);
2493 				bcopy(rth, opt->ip6po_rthdr, rthlen);
2494 			} else
2495 				opt->ip6po_rthdr = rth;
2496 
2497 			break;
2498 		}
2499 
2500 		default:
2501 			return(ENOPROTOOPT);
2502 		}
2503 	}
2504 
2505 	return(0);
2506 }
2507 
2508 /*
2509  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2510  * packet to the input queue of a specified interface.  Note that this
2511  * calls the output routine of the loopback "driver", but with an interface
2512  * pointer that might NOT be &loif -- easier than replicating that code here.
2513  */
2514 void
2515 ip6_mloopback(ifp, m, dst)
2516 	struct ifnet *ifp;
2517 	struct mbuf *m;
2518 	struct sockaddr_in6 *dst;
2519 {
2520 	struct mbuf *copym;
2521 	struct ip6_hdr *ip6;
2522 
2523 	copym = m_copy(m, 0, M_COPYALL);
2524 	if (copym == NULL)
2525 		return;
2526 
2527 	/*
2528 	 * Make sure to deep-copy IPv6 header portion in case the data
2529 	 * is in an mbuf cluster, so that we can safely override the IPv6
2530 	 * header portion later.
2531 	 */
2532 	if ((copym->m_flags & M_EXT) != 0 ||
2533 	    copym->m_len < sizeof(struct ip6_hdr)) {
2534 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
2535 		if (copym == NULL)
2536 			return;
2537 	}
2538 
2539 #ifdef DIAGNOSTIC
2540 	if (copym->m_len < sizeof(*ip6)) {
2541 		m_freem(copym);
2542 		return;
2543 	}
2544 #endif
2545 
2546 	ip6 = mtod(copym, struct ip6_hdr *);
2547 #ifndef SCOPEDROUTING
2548 	/*
2549 	 * clear embedded scope identifiers if necessary.
2550 	 * in6_clearscope will touch the addresses only when necessary.
2551 	 */
2552 	in6_clearscope(&ip6->ip6_src);
2553 	in6_clearscope(&ip6->ip6_dst);
2554 #endif
2555 
2556 	(void)if_simloop(ifp, copym, dst->sin6_family, 0);
2557 }
2558 
2559 /*
2560  * Chop IPv6 header off from the payload.
2561  */
2562 static int
2563 ip6_splithdr(m, exthdrs)
2564 	struct mbuf *m;
2565 	struct ip6_exthdrs *exthdrs;
2566 {
2567 	struct mbuf *mh;
2568 	struct ip6_hdr *ip6;
2569 
2570 	ip6 = mtod(m, struct ip6_hdr *);
2571 	if (m->m_len > sizeof(*ip6)) {
2572 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
2573 		if (mh == 0) {
2574 			m_freem(m);
2575 			return ENOBUFS;
2576 		}
2577 		M_MOVE_PKTHDR(mh, m);
2578 		MH_ALIGN(mh, sizeof(*ip6));
2579 		m->m_len -= sizeof(*ip6);
2580 		m->m_data += sizeof(*ip6);
2581 		mh->m_next = m;
2582 		m = mh;
2583 		m->m_len = sizeof(*ip6);
2584 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2585 	}
2586 	exthdrs->ip6e_ip6 = m;
2587 	return 0;
2588 }
2589 
2590 /*
2591  * Compute IPv6 extension header length.
2592  */
2593 int
2594 ip6_optlen(in6p)
2595 	struct in6pcb *in6p;
2596 {
2597 	int len;
2598 
2599 	if (!in6p->in6p_outputopts)
2600 		return 0;
2601 
2602 	len = 0;
2603 #define elen(x) \
2604     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2605 
2606 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
2607 	if (in6p->in6p_outputopts->ip6po_rthdr)
2608 		/* dest1 is valid with rthdr only */
2609 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
2610 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
2611 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
2612 	return len;
2613 #undef elen
2614 }
2615