xref: /freebsd/sys/netinet6/ip6_output.c (revision 4b2eaea43fec8e8792be611dea204071a10b655a)
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 	struct socket *so = inp ? inp->inp_socket : NULL;
185 
186 	ip6 = mtod(m, struct ip6_hdr *);
187 #endif /* IPSEC */
188 #ifdef FAST_IPSEC
189 	int needipsectun = 0;
190 	struct secpolicy *sp = NULL;
191 
192 	ip6 = mtod(m, struct ip6_hdr *);
193 #endif /* FAST_IPSEC */
194 
195 #define MAKE_EXTHDR(hp, mp)						\
196     do {								\
197 	if (hp) {							\
198 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
199 		error = ip6_copyexthdr((mp), (caddr_t)(hp), 		\
200 				       ((eh)->ip6e_len + 1) << 3);	\
201 		if (error)						\
202 			goto freehdrs;					\
203 	}								\
204     } while (0)
205 
206 	bzero(&exthdrs, sizeof(exthdrs));
207 
208 	if (opt) {
209 		/* Hop-by-Hop options header */
210 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
211 		/* Destination options header(1st part) */
212 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
213 		/* Routing header */
214 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
215 		/* Destination options header(2nd part) */
216 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
217 	}
218 
219 #ifdef IPSEC
220 	/* get a security policy for this packet */
221 	if (so == NULL)
222 		sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
223 	else
224 		sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
225 
226 	if (sp == NULL) {
227 		ipsec6stat.out_inval++;
228 		goto freehdrs;
229 	}
230 
231 	error = 0;
232 
233 	/* check policy */
234 	switch (sp->policy) {
235 	case IPSEC_POLICY_DISCARD:
236 		/*
237 		 * This packet is just discarded.
238 		 */
239 		ipsec6stat.out_polvio++;
240 		goto freehdrs;
241 
242 	case IPSEC_POLICY_BYPASS:
243 	case IPSEC_POLICY_NONE:
244 		/* no need to do IPsec. */
245 		needipsec = 0;
246 		break;
247 
248 	case IPSEC_POLICY_IPSEC:
249 		if (sp->req == NULL) {
250 			/* acquire a policy */
251 			error = key_spdacquire(sp);
252 			goto freehdrs;
253 		}
254 		needipsec = 1;
255 		break;
256 
257 	case IPSEC_POLICY_ENTRUST:
258 	default:
259 		printf("ip6_output: Invalid policy found. %d\n", sp->policy);
260 	}
261 #endif /* IPSEC */
262 #ifdef FAST_IPSEC
263 	/* get a security policy for this packet */
264 	if (inp == NULL)
265 		sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
266 	else
267 		sp = ipsec_getpolicybysock(m, IPSEC_DIR_OUTBOUND, inp, &error);
268 
269 	if (sp == NULL) {
270 		newipsecstat.ips_out_inval++;
271 		goto freehdrs;
272 	}
273 
274 	error = 0;
275 
276 	/* check policy */
277 	switch (sp->policy) {
278 	case IPSEC_POLICY_DISCARD:
279 		/*
280 		 * This packet is just discarded.
281 		 */
282 		newipsecstat.ips_out_polvio++;
283 		goto freehdrs;
284 
285 	case IPSEC_POLICY_BYPASS:
286 	case IPSEC_POLICY_NONE:
287 		/* no need to do IPsec. */
288 		needipsec = 0;
289 		break;
290 
291 	case IPSEC_POLICY_IPSEC:
292 		if (sp->req == NULL) {
293 			/* acquire a policy */
294 			error = key_spdacquire(sp);
295 			goto freehdrs;
296 		}
297 		needipsec = 1;
298 		break;
299 
300 	case IPSEC_POLICY_ENTRUST:
301 	default:
302 		printf("ip6_output: Invalid policy found. %d\n", sp->policy);
303 	}
304 #endif /* FAST_IPSEC */
305 
306 	/*
307 	 * Calculate the total length of the extension header chain.
308 	 * Keep the length of the unfragmentable part for fragmentation.
309 	 */
310 	optlen = 0;
311 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
312 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
313 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
314 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
315 	/* NOTE: we don't add AH/ESP length here. do that later. */
316 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
317 
318 	/*
319 	 * If we need IPsec, or there is at least one extension header,
320 	 * separate IP6 header from the payload.
321 	 */
322 	if ((needipsec || optlen) && !hdrsplit) {
323 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
324 			m = NULL;
325 			goto freehdrs;
326 		}
327 		m = exthdrs.ip6e_ip6;
328 		hdrsplit++;
329 	}
330 
331 	/* adjust pointer */
332 	ip6 = mtod(m, struct ip6_hdr *);
333 
334 	/* adjust mbuf packet header length */
335 	m->m_pkthdr.len += optlen;
336 	plen = m->m_pkthdr.len - sizeof(*ip6);
337 
338 	/* If this is a jumbo payload, insert a jumbo payload option. */
339 	if (plen > IPV6_MAXPACKET) {
340 		if (!hdrsplit) {
341 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
342 				m = NULL;
343 				goto freehdrs;
344 			}
345 			m = exthdrs.ip6e_ip6;
346 			hdrsplit++;
347 		}
348 		/* adjust pointer */
349 		ip6 = mtod(m, struct ip6_hdr *);
350 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
351 			goto freehdrs;
352 		ip6->ip6_plen = 0;
353 	} else
354 		ip6->ip6_plen = htons(plen);
355 
356 	/*
357 	 * Concatenate headers and fill in next header fields.
358 	 * Here we have, on "m"
359 	 *	IPv6 payload
360 	 * and we insert headers accordingly.  Finally, we should be getting:
361 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
362 	 *
363 	 * during the header composing process, "m" points to IPv6 header.
364 	 * "mprev" points to an extension header prior to esp.
365 	 */
366 	{
367 		u_char *nexthdrp = &ip6->ip6_nxt;
368 		struct mbuf *mprev = m;
369 
370 		/*
371 		 * we treat dest2 specially.  this makes IPsec processing
372 		 * much easier.  the goal here is to make mprev point the
373 		 * mbuf prior to dest2.
374 		 *
375 		 * result: IPv6 dest2 payload
376 		 * m and mprev will point to IPv6 header.
377 		 */
378 		if (exthdrs.ip6e_dest2) {
379 			if (!hdrsplit)
380 				panic("assumption failed: hdr not split");
381 			exthdrs.ip6e_dest2->m_next = m->m_next;
382 			m->m_next = exthdrs.ip6e_dest2;
383 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
384 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
385 		}
386 
387 #define MAKE_CHAIN(m, mp, p, i)\
388     do {\
389 	if (m) {\
390 		if (!hdrsplit) \
391 			panic("assumption failed: hdr not split"); \
392 		*mtod((m), u_char *) = *(p);\
393 		*(p) = (i);\
394 		p = mtod((m), u_char *);\
395 		(m)->m_next = (mp)->m_next;\
396 		(mp)->m_next = (m);\
397 		(mp) = (m);\
398 	}\
399     } while (0)
400 		/*
401 		 * result: IPv6 hbh dest1 rthdr dest2 payload
402 		 * m will point to IPv6 header.  mprev will point to the
403 		 * extension header prior to dest2 (rthdr in the above case).
404 		 */
405 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev,
406 			   nexthdrp, IPPROTO_HOPOPTS);
407 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev,
408 			   nexthdrp, IPPROTO_DSTOPTS);
409 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev,
410 			   nexthdrp, IPPROTO_ROUTING);
411 
412 #if defined(IPSEC) || defined(FAST_IPSEC)
413 		if (!needipsec)
414 			goto skip_ipsec2;
415 
416 		/*
417 		 * pointers after IPsec headers are not valid any more.
418 		 * other pointers need a great care too.
419 		 * (IPsec routines should not mangle mbufs prior to AH/ESP)
420 		 */
421 		exthdrs.ip6e_dest2 = NULL;
422 
423 	    {
424 		struct ip6_rthdr *rh = NULL;
425 		int segleft_org = 0;
426 		struct ipsec_output_state state;
427 
428 		if (exthdrs.ip6e_rthdr) {
429 			rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
430 			segleft_org = rh->ip6r_segleft;
431 			rh->ip6r_segleft = 0;
432 		}
433 
434 		bzero(&state, sizeof(state));
435 		state.m = m;
436 		error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
437 			&needipsectun);
438 		m = state.m;
439 		if (error) {
440 			/* mbuf is already reclaimed in ipsec6_output_trans. */
441 			m = NULL;
442 			switch (error) {
443 			case EHOSTUNREACH:
444 			case ENETUNREACH:
445 			case EMSGSIZE:
446 			case ENOBUFS:
447 			case ENOMEM:
448 				break;
449 			default:
450 				printf("ip6_output (ipsec): error code %d\n", error);
451 				/* fall through */
452 			case ENOENT:
453 				/* don't show these error codes to the user */
454 				error = 0;
455 				break;
456 			}
457 			goto bad;
458 		}
459 		if (exthdrs.ip6e_rthdr) {
460 			/* ah6_output doesn't modify mbuf chain */
461 			rh->ip6r_segleft = segleft_org;
462 		}
463 	    }
464 skip_ipsec2:;
465 #endif
466 	}
467 
468 	/*
469 	 * If there is a routing header, replace destination address field
470 	 * with the first hop of the routing header.
471 	 */
472 	if (exthdrs.ip6e_rthdr) {
473 		struct ip6_rthdr *rh =
474 			(struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
475 						  struct ip6_rthdr *));
476 		struct ip6_rthdr0 *rh0;
477 
478 		finaldst = ip6->ip6_dst;
479 		switch (rh->ip6r_type) {
480 		case IPV6_RTHDR_TYPE_0:
481 			 rh0 = (struct ip6_rthdr0 *)rh;
482 			 ip6->ip6_dst = rh0->ip6r0_addr[0];
483 			 bcopy((caddr_t)&rh0->ip6r0_addr[1],
484 			       (caddr_t)&rh0->ip6r0_addr[0],
485 			       sizeof(struct in6_addr)*(rh0->ip6r0_segleft - 1)
486 				 );
487 			 rh0->ip6r0_addr[rh0->ip6r0_segleft - 1] = finaldst;
488 			 break;
489 		default:	/* is it possible? */
490 			 error = EINVAL;
491 			 goto bad;
492 		}
493 	}
494 
495 	/* Source address validation */
496 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
497 	    (flags & IPV6_DADOUTPUT) == 0) {
498 		error = EOPNOTSUPP;
499 		ip6stat.ip6s_badscope++;
500 		goto bad;
501 	}
502 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
503 		error = EOPNOTSUPP;
504 		ip6stat.ip6s_badscope++;
505 		goto bad;
506 	}
507 
508 	ip6stat.ip6s_localout++;
509 
510 	/*
511 	 * Route packet.
512 	 */
513 	if (ro == 0) {
514 		ro = &ip6route;
515 		bzero((caddr_t)ro, sizeof(*ro));
516 	}
517 	ro_pmtu = ro;
518 	if (opt && opt->ip6po_rthdr)
519 		ro = &opt->ip6po_route;
520 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
521 	/*
522 	 * If there is a cached route,
523 	 * check that it is to the same destination
524 	 * and is still up. If not, free it and try again.
525 	 */
526 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
527 			 dst->sin6_family != AF_INET6 ||
528 			 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
529 		RTFREE(ro->ro_rt);
530 		ro->ro_rt = (struct rtentry *)0;
531 	}
532 	if (ro->ro_rt == 0) {
533 		bzero(dst, sizeof(*dst));
534 		dst->sin6_family = AF_INET6;
535 		dst->sin6_len = sizeof(struct sockaddr_in6);
536 		dst->sin6_addr = ip6->ip6_dst;
537 #ifdef SCOPEDROUTING
538 		/* XXX: sin6_scope_id should already be fixed at this point */
539 		if (IN6_IS_SCOPE_LINKLOCAL(&dst->sin6_addr))
540 			dst->sin6_scope_id = ntohs(dst->sin6_addr.s6_addr16[1]);
541 #endif
542 	}
543 #if defined(IPSEC) || defined(FAST_IPSEC)
544 	if (needipsec && needipsectun) {
545 		struct ipsec_output_state state;
546 
547 		/*
548 		 * All the extension headers will become inaccessible
549 		 * (since they can be encrypted).
550 		 * Don't panic, we need no more updates to extension headers
551 		 * on inner IPv6 packet (since they are now encapsulated).
552 		 *
553 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
554 		 */
555 		bzero(&exthdrs, sizeof(exthdrs));
556 		exthdrs.ip6e_ip6 = m;
557 
558 		bzero(&state, sizeof(state));
559 		state.m = m;
560 		state.ro = (struct route *)ro;
561 		state.dst = (struct sockaddr *)dst;
562 
563 		error = ipsec6_output_tunnel(&state, sp, flags);
564 
565 		m = state.m;
566 		ro = (struct route_in6 *)state.ro;
567 		dst = (struct sockaddr_in6 *)state.dst;
568 		if (error) {
569 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
570 			m0 = m = NULL;
571 			m = NULL;
572 			switch (error) {
573 			case EHOSTUNREACH:
574 			case ENETUNREACH:
575 			case EMSGSIZE:
576 			case ENOBUFS:
577 			case ENOMEM:
578 				break;
579 			default:
580 				printf("ip6_output (ipsec): error code %d\n", error);
581 				/* fall through */
582 			case ENOENT:
583 				/* don't show these error codes to the user */
584 				error = 0;
585 				break;
586 			}
587 			goto bad;
588 		}
589 
590 		exthdrs.ip6e_ip6 = m;
591 	}
592 #endif /* IPSEC */
593 
594 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
595 		/* Unicast */
596 
597 #define ifatoia6(ifa)	((struct in6_ifaddr *)(ifa))
598 #define sin6tosa(sin6)	((struct sockaddr *)(sin6))
599 		/* xxx
600 		 * interface selection comes here
601 		 * if an interface is specified from an upper layer,
602 		 * ifp must point it.
603 		 */
604 		if (ro->ro_rt == 0) {
605 			/*
606 			 * non-bsdi always clone routes, if parent is
607 			 * PRF_CLONING.
608 			 */
609 			rtalloc((struct route *)ro);
610 		}
611 		if (ro->ro_rt == 0) {
612 			ip6stat.ip6s_noroute++;
613 			error = EHOSTUNREACH;
614 			/* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
615 			goto bad;
616 		}
617 		ia = ifatoia6(ro->ro_rt->rt_ifa);
618 		ifp = ro->ro_rt->rt_ifp;
619 		ro->ro_rt->rt_use++;
620 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
621 			dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
622 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
623 
624 		in6_ifstat_inc(ifp, ifs6_out_request);
625 
626 		/*
627 		 * Check if the outgoing interface conflicts with
628 		 * the interface specified by ifi6_ifindex (if specified).
629 		 * Note that loopback interface is always okay.
630 		 * (this may happen when we are sending a packet to one of
631 		 *  our own addresses.)
632 		 */
633 		if (opt && opt->ip6po_pktinfo
634 		 && opt->ip6po_pktinfo->ipi6_ifindex) {
635 			if (!(ifp->if_flags & IFF_LOOPBACK)
636 			 && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
637 				ip6stat.ip6s_noroute++;
638 				in6_ifstat_inc(ifp, ifs6_out_discard);
639 				error = EHOSTUNREACH;
640 				goto bad;
641 			}
642 		}
643 
644 		if (opt && opt->ip6po_hlim != -1)
645 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
646 	} else {
647 		/* Multicast */
648 		struct	in6_multi *in6m;
649 
650 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
651 
652 		/*
653 		 * See if the caller provided any multicast options
654 		 */
655 		ifp = NULL;
656 		if (im6o != NULL) {
657 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
658 			if (im6o->im6o_multicast_ifp != NULL)
659 				ifp = im6o->im6o_multicast_ifp;
660 		} else
661 			ip6->ip6_hlim = ip6_defmcasthlim;
662 
663 		/*
664 		 * See if the caller provided the outgoing interface
665 		 * as an ancillary data.
666 		 * Boundary check for ifindex is assumed to be already done.
667 		 */
668 		if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
669 			ifp = ifnet_byindex(opt->ip6po_pktinfo->ipi6_ifindex);
670 
671 		/*
672 		 * If the destination is a node-local scope multicast,
673 		 * the packet should be loop-backed only.
674 		 */
675 		if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
676 			/*
677 			 * If the outgoing interface is already specified,
678 			 * it should be a loopback interface.
679 			 */
680 			if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) {
681 				ip6stat.ip6s_badscope++;
682 				error = ENETUNREACH; /* XXX: better error? */
683 				/* XXX correct ifp? */
684 				in6_ifstat_inc(ifp, ifs6_out_discard);
685 				goto bad;
686 			} else {
687 				ifp = &loif[0];
688 			}
689 		}
690 
691 		if (opt && opt->ip6po_hlim != -1)
692 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
693 
694 		/*
695 		 * If caller did not provide an interface lookup a
696 		 * default in the routing table.  This is either a
697 		 * default for the speicfied group (i.e. a host
698 		 * route), or a multicast default (a route for the
699 		 * ``net'' ff00::/8).
700 		 */
701 		if (ifp == NULL) {
702 			if (ro->ro_rt == 0) {
703 				ro->ro_rt = rtalloc1((struct sockaddr *)
704 						&ro->ro_dst, 0, 0UL);
705 			}
706 			if (ro->ro_rt == 0) {
707 				ip6stat.ip6s_noroute++;
708 				error = EHOSTUNREACH;
709 				/* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
710 				goto bad;
711 			}
712 			ia = ifatoia6(ro->ro_rt->rt_ifa);
713 			ifp = ro->ro_rt->rt_ifp;
714 			ro->ro_rt->rt_use++;
715 		}
716 
717 		if ((flags & IPV6_FORWARDING) == 0)
718 			in6_ifstat_inc(ifp, ifs6_out_request);
719 		in6_ifstat_inc(ifp, ifs6_out_mcast);
720 
721 		/*
722 		 * Confirm that the outgoing interface supports multicast.
723 		 */
724 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
725 			ip6stat.ip6s_noroute++;
726 			in6_ifstat_inc(ifp, ifs6_out_discard);
727 			error = ENETUNREACH;
728 			goto bad;
729 		}
730 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
731 		if (in6m != NULL &&
732 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
733 			/*
734 			 * If we belong to the destination multicast group
735 			 * on the outgoing interface, and the caller did not
736 			 * forbid loopback, loop back a copy.
737 			 */
738 			ip6_mloopback(ifp, m, dst);
739 		} else {
740 			/*
741 			 * If we are acting as a multicast router, perform
742 			 * multicast forwarding as if the packet had just
743 			 * arrived on the interface to which we are about
744 			 * to send.  The multicast forwarding function
745 			 * recursively calls this function, using the
746 			 * IPV6_FORWARDING flag to prevent infinite recursion.
747 			 *
748 			 * Multicasts that are looped back by ip6_mloopback(),
749 			 * above, will be forwarded by the ip6_input() routine,
750 			 * if necessary.
751 			 */
752 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
753 				if (ip6_mforward(ip6, ifp, m) != 0) {
754 					m_freem(m);
755 					goto done;
756 				}
757 			}
758 		}
759 		/*
760 		 * Multicasts with a hoplimit of zero may be looped back,
761 		 * above, but must not be transmitted on a network.
762 		 * Also, multicasts addressed to the loopback interface
763 		 * are not sent -- the above call to ip6_mloopback() will
764 		 * loop back a copy if this host actually belongs to the
765 		 * destination group on the loopback interface.
766 		 */
767 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
768 			m_freem(m);
769 			goto done;
770 		}
771 	}
772 
773 	/*
774 	 * Fill the outgoing inteface to tell the upper layer
775 	 * to increment per-interface statistics.
776 	 */
777 	if (ifpp)
778 		*ifpp = ifp;
779 
780 	/*
781 	 * Determine path MTU.
782 	 */
783 	if (ro_pmtu != ro) {
784 		/* The first hop and the final destination may differ. */
785 		struct sockaddr_in6 *sin6_fin =
786 			(struct sockaddr_in6 *)&ro_pmtu->ro_dst;
787 		if (ro_pmtu->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
788 				       !IN6_ARE_ADDR_EQUAL(&sin6_fin->sin6_addr,
789 							   &finaldst))) {
790 			RTFREE(ro_pmtu->ro_rt);
791 			ro_pmtu->ro_rt = (struct rtentry *)0;
792 		}
793 		if (ro_pmtu->ro_rt == 0) {
794 			bzero(sin6_fin, sizeof(*sin6_fin));
795 			sin6_fin->sin6_family = AF_INET6;
796 			sin6_fin->sin6_len = sizeof(struct sockaddr_in6);
797 			sin6_fin->sin6_addr = finaldst;
798 
799 			rtalloc((struct route *)ro_pmtu);
800 		}
801 	}
802 	if (ro_pmtu->ro_rt != NULL) {
803 		u_int32_t ifmtu = nd_ifinfo[ifp->if_index].linkmtu;
804 
805 		mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
806 		if (mtu > ifmtu || mtu == 0) {
807 			/*
808 			 * The MTU on the route is larger than the MTU on
809 			 * the interface!  This shouldn't happen, unless the
810 			 * MTU of the interface has been changed after the
811 			 * interface was brought up.  Change the MTU in the
812 			 * route to match the interface MTU (as long as the
813 			 * field isn't locked).
814 			 *
815 			 * if MTU on the route is 0, we need to fix the MTU.
816 			 * this case happens with path MTU discovery timeouts.
817 			 */
818 			 mtu = ifmtu;
819 			 if ((ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
820 				 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; /* XXX */
821 		}
822 	} else {
823 		mtu = nd_ifinfo[ifp->if_index].linkmtu;
824 	}
825 
826 	/*
827 	 * advanced API (IPV6_USE_MIN_MTU) overrides mtu setting
828 	 */
829 	if ((flags & IPV6_MINMTU) != 0 && mtu > IPV6_MMTU)
830 		mtu = IPV6_MMTU;
831 
832 	/* Fake scoped addresses */
833 	if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
834 		/*
835 		 * If source or destination address is a scoped address, and
836 		 * the packet is going to be sent to a loopback interface,
837 		 * we should keep the original interface.
838 		 */
839 
840 		/*
841 		 * XXX: this is a very experimental and temporary solution.
842 		 * We eventually have sockaddr_in6 and use the sin6_scope_id
843 		 * field of the structure here.
844 		 * We rely on the consistency between two scope zone ids
845 		 * of source and destination, which should already be assured.
846 		 * Larger scopes than link will be supported in the future.
847 		 */
848 		origifp = NULL;
849 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
850 			origifp = ifnet_byindex(ntohs(ip6->ip6_src.s6_addr16[1]));
851 		else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
852 			origifp = ifnet_byindex(ntohs(ip6->ip6_dst.s6_addr16[1]));
853 		/*
854 		 * XXX: origifp can be NULL even in those two cases above.
855 		 * For example, if we remove the (only) link-local address
856 		 * from the loopback interface, and try to send a link-local
857 		 * address without link-id information.  Then the source
858 		 * address is ::1, and the destination address is the
859 		 * link-local address with its s6_addr16[1] being zero.
860 		 * What is worse, if the packet goes to the loopback interface
861 		 * by a default rejected route, the null pointer would be
862 		 * passed to looutput, and the kernel would hang.
863 		 * The following last resort would prevent such disaster.
864 		 */
865 		if (origifp == NULL)
866 			origifp = ifp;
867 	}
868 	else
869 		origifp = ifp;
870 #ifndef SCOPEDROUTING
871 	/*
872 	 * clear embedded scope identifiers if necessary.
873 	 * in6_clearscope will touch the addresses only when necessary.
874 	 */
875 	in6_clearscope(&ip6->ip6_src);
876 	in6_clearscope(&ip6->ip6_dst);
877 #endif
878 
879 	/*
880 	 * Check with the firewall...
881 	 */
882         if (ip6_fw_enable && ip6_fw_chk_ptr) {
883 		u_short port = 0;
884 		m->m_pkthdr.rcvif = NULL;	/* XXX */
885 		/* If ipfw says divert, we have to just drop packet */
886 		if ((*ip6_fw_chk_ptr)(&ip6, ifp, &port, &m)) {
887 			m_freem(m);
888 			goto done;
889 		}
890 		if (!m) {
891 			error = EACCES;
892 			goto done;
893 		}
894 	}
895 
896 	/*
897 	 * If the outgoing packet contains a hop-by-hop options header,
898 	 * it must be examined and processed even by the source node.
899 	 * (RFC 2460, section 4.)
900 	 */
901 	if (exthdrs.ip6e_hbh) {
902 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
903 		u_int32_t dummy1; /* XXX unused */
904 		u_int32_t dummy2; /* XXX unused */
905 
906 #ifdef DIAGNOSTIC
907 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
908 			panic("ip6e_hbh is not continuous");
909 #endif
910 		/*
911 		 *  XXX: if we have to send an ICMPv6 error to the sender,
912 		 *       we need the M_LOOP flag since icmp6_error() expects
913 		 *       the IPv6 and the hop-by-hop options header are
914 		 *       continuous unless the flag is set.
915 		 */
916 		m->m_flags |= M_LOOP;
917 		m->m_pkthdr.rcvif = ifp;
918 		if (ip6_process_hopopts(m,
919 					(u_int8_t *)(hbh + 1),
920 					((hbh->ip6h_len + 1) << 3) -
921 					sizeof(struct ip6_hbh),
922 					&dummy1, &dummy2) < 0) {
923 			/* m was already freed at this point */
924 			error = EINVAL;/* better error? */
925 			goto done;
926 		}
927 		m->m_flags &= ~M_LOOP; /* XXX */
928 		m->m_pkthdr.rcvif = NULL;
929 	}
930 
931 #ifdef PFIL_HOOKS
932 	/*
933 	 * Run through list of hooks for output packets.
934 	 */
935 	m1 = m;
936 	pfh = pfil_hook_get(PFIL_OUT, &inet6sw[ip6_protox[IPPROTO_IPV6]].pr_pfh);
937 	for (; pfh; pfh = pfh->pfil_link.tqe_next)
938 		if (pfh->pfil_func) {
939 			rv = pfh->pfil_func(ip6, sizeof(*ip6), ifp, 1, &m1);
940 			if (rv) {
941 				error = EHOSTUNREACH;
942 				goto done;
943 			}
944 			m = m1;
945 			if (m == NULL)
946 				goto done;
947 			ip6 = mtod(m, struct ip6_hdr *);
948 		}
949 #endif /* PFIL_HOOKS */
950 	/*
951 	 * Send the packet to the outgoing interface.
952 	 * If necessary, do IPv6 fragmentation before sending.
953 	 */
954 	tlen = m->m_pkthdr.len;
955 	if (tlen <= mtu
956 #ifdef notyet
957 	    /*
958 	     * On any link that cannot convey a 1280-octet packet in one piece,
959 	     * link-specific fragmentation and reassembly must be provided at
960 	     * a layer below IPv6. [RFC 2460, sec.5]
961 	     * Thus if the interface has ability of link-level fragmentation,
962 	     * we can just send the packet even if the packet size is
963 	     * larger than the link's MTU.
964 	     * XXX: IFF_FRAGMENTABLE (or such) flag has not been defined yet...
965 	     */
966 
967 	    || ifp->if_flags & IFF_FRAGMENTABLE
968 #endif
969 	    )
970 	{
971  		/* Record statistics for this interface address. */
972  		if (ia && !(flags & IPV6_FORWARDING)) {
973  			ia->ia_ifa.if_opackets++;
974  			ia->ia_ifa.if_obytes += m->m_pkthdr.len;
975  		}
976 #ifdef IPSEC
977 		/* clean ipsec history once it goes out of the node */
978 		ipsec_delaux(m);
979 #endif
980 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
981 		goto done;
982 	} else if (mtu < IPV6_MMTU) {
983 		/*
984 		 * note that path MTU is never less than IPV6_MMTU
985 		 * (see icmp6_input).
986 		 */
987 		error = EMSGSIZE;
988 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
989 		goto bad;
990 	} else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
991 		error = EMSGSIZE;
992 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
993 		goto bad;
994 	} else {
995 		struct mbuf **mnext, *m_frgpart;
996 		struct ip6_frag *ip6f;
997 		u_int32_t id = htonl(ip6_id++);
998 		u_char nextproto;
999 
1000 		/*
1001 		 * Too large for the destination or interface;
1002 		 * fragment if possible.
1003 		 * Must be able to put at least 8 bytes per fragment.
1004 		 */
1005 		hlen = unfragpartlen;
1006 		if (mtu > IPV6_MAXPACKET)
1007 			mtu = IPV6_MAXPACKET;
1008 
1009 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1010 		if (len < 8) {
1011 			error = EMSGSIZE;
1012 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
1013 			goto bad;
1014 		}
1015 
1016 		mnext = &m->m_nextpkt;
1017 
1018 		/*
1019 		 * Change the next header field of the last header in the
1020 		 * unfragmentable part.
1021 		 */
1022 		if (exthdrs.ip6e_rthdr) {
1023 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1024 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1025 		} else if (exthdrs.ip6e_dest1) {
1026 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1027 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1028 		} else if (exthdrs.ip6e_hbh) {
1029 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1030 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1031 		} else {
1032 			nextproto = ip6->ip6_nxt;
1033 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1034 		}
1035 
1036 		/*
1037 		 * Loop through length of segment after first fragment,
1038 		 * make new header and copy data of each part and link onto
1039 		 * chain.
1040 		 */
1041 		m0 = m;
1042 		for (off = hlen; off < tlen; off += len) {
1043 			MGETHDR(m, M_NOWAIT, MT_HEADER);
1044 			if (!m) {
1045 				error = ENOBUFS;
1046 				ip6stat.ip6s_odropped++;
1047 				goto sendorfree;
1048 			}
1049 			m->m_pkthdr.rcvif = NULL;
1050 			m->m_flags = m0->m_flags & M_COPYFLAGS;
1051 			*mnext = m;
1052 			mnext = &m->m_nextpkt;
1053 			m->m_data += max_linkhdr;
1054 			mhip6 = mtod(m, struct ip6_hdr *);
1055 			*mhip6 = *ip6;
1056 			m->m_len = sizeof(*mhip6);
1057  			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1058  			if (error) {
1059 				ip6stat.ip6s_odropped++;
1060 				goto sendorfree;
1061 			}
1062 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1063 			if (off + len >= tlen)
1064 				len = tlen - off;
1065 			else
1066 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1067 			mhip6->ip6_plen = htons((u_short)(len + hlen +
1068 							  sizeof(*ip6f) -
1069 							  sizeof(struct ip6_hdr)));
1070 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1071 				error = ENOBUFS;
1072 				ip6stat.ip6s_odropped++;
1073 				goto sendorfree;
1074 			}
1075 			m_cat(m, m_frgpart);
1076 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1077 			m->m_pkthdr.rcvif = (struct ifnet *)0;
1078 			ip6f->ip6f_reserved = 0;
1079 			ip6f->ip6f_ident = id;
1080 			ip6f->ip6f_nxt = nextproto;
1081 			ip6stat.ip6s_ofragments++;
1082 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1083 		}
1084 
1085 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1086 	}
1087 
1088 	/*
1089 	 * Remove leading garbages.
1090 	 */
1091 sendorfree:
1092 	m = m0->m_nextpkt;
1093 	m0->m_nextpkt = 0;
1094 	m_freem(m0);
1095 	for (m0 = m; m; m = m0) {
1096 		m0 = m->m_nextpkt;
1097 		m->m_nextpkt = 0;
1098 		if (error == 0) {
1099  			/* Record statistics for this interface address. */
1100  			if (ia) {
1101  				ia->ia_ifa.if_opackets++;
1102  				ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1103  			}
1104 #ifdef IPSEC
1105 			/* clean ipsec history once it goes out of the node */
1106 			ipsec_delaux(m);
1107 #endif
1108 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1109 		} else
1110 			m_freem(m);
1111 	}
1112 
1113 	if (error == 0)
1114 		ip6stat.ip6s_fragmented++;
1115 
1116 done:
1117 	if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1118 		RTFREE(ro->ro_rt);
1119 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1120 		RTFREE(ro_pmtu->ro_rt);
1121 	}
1122 
1123 #ifdef IPSEC
1124 	if (sp != NULL)
1125 		key_freesp(sp);
1126 #endif /* IPSEC */
1127 #ifdef FAST_IPSEC
1128 	if (sp != NULL)
1129 		KEY_FREESP(&sp);
1130 #endif /* FAST_IPSEC */
1131 
1132 	return(error);
1133 
1134 freehdrs:
1135 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1136 	m_freem(exthdrs.ip6e_dest1);
1137 	m_freem(exthdrs.ip6e_rthdr);
1138 	m_freem(exthdrs.ip6e_dest2);
1139 	/* fall through */
1140 bad:
1141 	m_freem(m);
1142 	goto done;
1143 }
1144 
1145 static int
1146 ip6_copyexthdr(mp, hdr, hlen)
1147 	struct mbuf **mp;
1148 	caddr_t hdr;
1149 	int hlen;
1150 {
1151 	struct mbuf *m;
1152 
1153 	if (hlen > MCLBYTES)
1154 		return(ENOBUFS); /* XXX */
1155 
1156 	MGET(m, M_NOWAIT, MT_DATA);
1157 	if (!m)
1158 		return(ENOBUFS);
1159 
1160 	if (hlen > MLEN) {
1161 		MCLGET(m, M_NOWAIT);
1162 		if ((m->m_flags & M_EXT) == 0) {
1163 			m_free(m);
1164 			return(ENOBUFS);
1165 		}
1166 	}
1167 	m->m_len = hlen;
1168 	if (hdr)
1169 		bcopy(hdr, mtod(m, caddr_t), hlen);
1170 
1171 	*mp = m;
1172 	return(0);
1173 }
1174 
1175 /*
1176  * Insert jumbo payload option.
1177  */
1178 static int
1179 ip6_insert_jumboopt(exthdrs, plen)
1180 	struct ip6_exthdrs *exthdrs;
1181 	u_int32_t plen;
1182 {
1183 	struct mbuf *mopt;
1184 	u_char *optbuf;
1185 	u_int32_t v;
1186 
1187 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1188 
1189 	/*
1190 	 * If there is no hop-by-hop options header, allocate new one.
1191 	 * If there is one but it doesn't have enough space to store the
1192 	 * jumbo payload option, allocate a cluster to store the whole options.
1193 	 * Otherwise, use it to store the options.
1194 	 */
1195 	if (exthdrs->ip6e_hbh == 0) {
1196 		MGET(mopt, M_NOWAIT, MT_DATA);
1197 		if (mopt == 0)
1198 			return(ENOBUFS);
1199 		mopt->m_len = JUMBOOPTLEN;
1200 		optbuf = mtod(mopt, u_char *);
1201 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1202 		exthdrs->ip6e_hbh = mopt;
1203 	} else {
1204 		struct ip6_hbh *hbh;
1205 
1206 		mopt = exthdrs->ip6e_hbh;
1207 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1208 			/*
1209 			 * XXX assumption:
1210 			 * - exthdrs->ip6e_hbh is not referenced from places
1211 			 *   other than exthdrs.
1212 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1213 			 */
1214 			int oldoptlen = mopt->m_len;
1215 			struct mbuf *n;
1216 
1217 			/*
1218 			 * XXX: give up if the whole (new) hbh header does
1219 			 * not fit even in an mbuf cluster.
1220 			 */
1221 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1222 				return(ENOBUFS);
1223 
1224 			/*
1225 			 * As a consequence, we must always prepare a cluster
1226 			 * at this point.
1227 			 */
1228 			MGET(n, M_NOWAIT, MT_DATA);
1229 			if (n) {
1230 				MCLGET(n, M_NOWAIT);
1231 				if ((n->m_flags & M_EXT) == 0) {
1232 					m_freem(n);
1233 					n = NULL;
1234 				}
1235 			}
1236 			if (!n)
1237 				return(ENOBUFS);
1238 			n->m_len = oldoptlen + JUMBOOPTLEN;
1239 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1240 			      oldoptlen);
1241 			optbuf = mtod(n, caddr_t) + oldoptlen;
1242 			m_freem(mopt);
1243 			mopt = exthdrs->ip6e_hbh = n;
1244 		} else {
1245 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1246 			mopt->m_len += JUMBOOPTLEN;
1247 		}
1248 		optbuf[0] = IP6OPT_PADN;
1249 		optbuf[1] = 1;
1250 
1251 		/*
1252 		 * Adjust the header length according to the pad and
1253 		 * the jumbo payload option.
1254 		 */
1255 		hbh = mtod(mopt, struct ip6_hbh *);
1256 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1257 	}
1258 
1259 	/* fill in the option. */
1260 	optbuf[2] = IP6OPT_JUMBO;
1261 	optbuf[3] = 4;
1262 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1263 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1264 
1265 	/* finally, adjust the packet header length */
1266 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1267 
1268 	return(0);
1269 #undef JUMBOOPTLEN
1270 }
1271 
1272 /*
1273  * Insert fragment header and copy unfragmentable header portions.
1274  */
1275 static int
1276 ip6_insertfraghdr(m0, m, hlen, frghdrp)
1277 	struct mbuf *m0, *m;
1278 	int hlen;
1279 	struct ip6_frag **frghdrp;
1280 {
1281 	struct mbuf *n, *mlast;
1282 
1283 	if (hlen > sizeof(struct ip6_hdr)) {
1284 		n = m_copym(m0, sizeof(struct ip6_hdr),
1285 			    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1286 		if (n == 0)
1287 			return(ENOBUFS);
1288 		m->m_next = n;
1289 	} else
1290 		n = m;
1291 
1292 	/* Search for the last mbuf of unfragmentable part. */
1293 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1294 		;
1295 
1296 	if ((mlast->m_flags & M_EXT) == 0 &&
1297 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1298 		/* use the trailing space of the last mbuf for the fragment hdr */
1299 		*frghdrp =
1300 			(struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len);
1301 		mlast->m_len += sizeof(struct ip6_frag);
1302 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1303 	} else {
1304 		/* allocate a new mbuf for the fragment header */
1305 		struct mbuf *mfrg;
1306 
1307 		MGET(mfrg, M_NOWAIT, MT_DATA);
1308 		if (mfrg == 0)
1309 			return(ENOBUFS);
1310 		mfrg->m_len = sizeof(struct ip6_frag);
1311 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1312 		mlast->m_next = mfrg;
1313 	}
1314 
1315 	return(0);
1316 }
1317 
1318 /*
1319  * IP6 socket option processing.
1320  */
1321 int
1322 ip6_ctloutput(so, sopt)
1323 	struct socket *so;
1324 	struct sockopt *sopt;
1325 {
1326 	int privileged;
1327 	struct inpcb *in6p = sotoinpcb(so);
1328 	int error, optval;
1329 	int level, op, optname;
1330 	int optlen;
1331 	struct thread *td;
1332 
1333 	if (sopt) {
1334 		level = sopt->sopt_level;
1335 		op = sopt->sopt_dir;
1336 		optname = sopt->sopt_name;
1337 		optlen = sopt->sopt_valsize;
1338 		td = sopt->sopt_td;
1339 	} else {
1340 		panic("ip6_ctloutput: arg soopt is NULL");
1341 	}
1342 	error = optval = 0;
1343 
1344 	privileged = (td == 0 || suser(td)) ? 0 : 1;
1345 
1346 	if (level == IPPROTO_IPV6) {
1347 		switch (op) {
1348 
1349 		case SOPT_SET:
1350 			switch (optname) {
1351 			case IPV6_PKTOPTIONS:
1352 			{
1353 				struct mbuf *m;
1354 
1355 				error = soopt_getm(sopt, &m); /* XXX */
1356 				if (error != 0)
1357 					break;
1358 				error = soopt_mcopyin(sopt, m); /* XXX */
1359 				if (error != 0)
1360 					break;
1361 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1362 						    m, so, sopt);
1363 				m_freem(m); /* XXX */
1364 				break;
1365 			}
1366 
1367 			/*
1368 			 * Use of some Hop-by-Hop options or some
1369 			 * Destination options, might require special
1370 			 * privilege.  That is, normal applications
1371 			 * (without special privilege) might be forbidden
1372 			 * from setting certain options in outgoing packets,
1373 			 * and might never see certain options in received
1374 			 * packets. [RFC 2292 Section 6]
1375 			 * KAME specific note:
1376 			 *  KAME prevents non-privileged users from sending or
1377 			 *  receiving ANY hbh/dst options in order to avoid
1378 			 *  overhead of parsing options in the kernel.
1379 			 */
1380 			case IPV6_UNICAST_HOPS:
1381 			case IPV6_CHECKSUM:
1382 			case IPV6_FAITH:
1383 
1384 			case IPV6_V6ONLY:
1385 				if (optlen != sizeof(int)) {
1386 					error = EINVAL;
1387 					break;
1388 				}
1389 				error = sooptcopyin(sopt, &optval,
1390 					sizeof optval, sizeof optval);
1391 				if (error)
1392 					break;
1393 				switch (optname) {
1394 
1395 				case IPV6_UNICAST_HOPS:
1396 					if (optval < -1 || optval >= 256)
1397 						error = EINVAL;
1398 					else {
1399 						/* -1 = kernel default */
1400 						in6p->in6p_hops = optval;
1401 
1402 						if ((in6p->in6p_vflag &
1403 						     INP_IPV4) != 0)
1404 							in6p->inp_ip_ttl = optval;
1405 					}
1406 					break;
1407 #define OPTSET(bit) \
1408 do { \
1409 	if (optval) \
1410 		in6p->in6p_flags |= (bit); \
1411 	else \
1412 		in6p->in6p_flags &= ~(bit); \
1413 } while (0)
1414 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1415 
1416 				case IPV6_CHECKSUM:
1417 					in6p->in6p_cksum = optval;
1418 					break;
1419 
1420 				case IPV6_FAITH:
1421 					OPTSET(IN6P_FAITH);
1422 					break;
1423 
1424 				case IPV6_V6ONLY:
1425 					/*
1426 					 * make setsockopt(IPV6_V6ONLY)
1427 					 * available only prior to bind(2).
1428 					 * see ipng mailing list, Jun 22 2001.
1429 					 */
1430 					if (in6p->in6p_lport ||
1431 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr))
1432 					{
1433 						error = EINVAL;
1434 						break;
1435 					}
1436 					OPTSET(IN6P_IPV6_V6ONLY);
1437 					if (optval)
1438 						in6p->in6p_vflag &= ~INP_IPV4;
1439 					else
1440 						in6p->in6p_vflag |= INP_IPV4;
1441 					break;
1442 				}
1443 				break;
1444 
1445 			case IPV6_PKTINFO:
1446 			case IPV6_HOPLIMIT:
1447 			case IPV6_HOPOPTS:
1448 			case IPV6_DSTOPTS:
1449 			case IPV6_RTHDR:
1450 				/* RFC 2292 */
1451 				if (optlen != sizeof(int)) {
1452 					error = EINVAL;
1453 					break;
1454 				}
1455 				error = sooptcopyin(sopt, &optval,
1456 					sizeof optval, sizeof optval);
1457 				if (error)
1458 					break;
1459 				switch (optname) {
1460 				case IPV6_PKTINFO:
1461 					OPTSET(IN6P_PKTINFO);
1462 					break;
1463 				case IPV6_HOPLIMIT:
1464 					OPTSET(IN6P_HOPLIMIT);
1465 					break;
1466 				case IPV6_HOPOPTS:
1467 					/*
1468 					 * Check super-user privilege.
1469 					 * See comments for IPV6_RECVHOPOPTS.
1470 					 */
1471 					if (!privileged)
1472 						return(EPERM);
1473 					OPTSET(IN6P_HOPOPTS);
1474 					break;
1475 				case IPV6_DSTOPTS:
1476 					if (!privileged)
1477 						return(EPERM);
1478 					OPTSET(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1479 					break;
1480 				case IPV6_RTHDR:
1481 					OPTSET(IN6P_RTHDR);
1482 					break;
1483 				}
1484 				break;
1485 #undef OPTSET
1486 
1487 			case IPV6_MULTICAST_IF:
1488 			case IPV6_MULTICAST_HOPS:
1489 			case IPV6_MULTICAST_LOOP:
1490 			case IPV6_JOIN_GROUP:
1491 			case IPV6_LEAVE_GROUP:
1492 			    {
1493 				struct mbuf *m;
1494 				if (sopt->sopt_valsize > MLEN) {
1495 					error = EMSGSIZE;
1496 					break;
1497 				}
1498 				/* XXX */
1499 				MGET(m, sopt->sopt_td ? 0 : M_NOWAIT, MT_HEADER);
1500 				if (m == 0) {
1501 					error = ENOBUFS;
1502 					break;
1503 				}
1504 				m->m_len = sopt->sopt_valsize;
1505 				error = sooptcopyin(sopt, mtod(m, char *),
1506 						    m->m_len, m->m_len);
1507 				error =	ip6_setmoptions(sopt->sopt_name,
1508 							&in6p->in6p_moptions,
1509 							m);
1510 				(void)m_free(m);
1511 			    }
1512 				break;
1513 
1514 			case IPV6_PORTRANGE:
1515 				error = sooptcopyin(sopt, &optval,
1516 				    sizeof optval, sizeof optval);
1517 				if (error)
1518 					break;
1519 
1520 				switch (optval) {
1521 				case IPV6_PORTRANGE_DEFAULT:
1522 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1523 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1524 					break;
1525 
1526 				case IPV6_PORTRANGE_HIGH:
1527 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1528 					in6p->in6p_flags |= IN6P_HIGHPORT;
1529 					break;
1530 
1531 				case IPV6_PORTRANGE_LOW:
1532 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1533 					in6p->in6p_flags |= IN6P_LOWPORT;
1534 					break;
1535 
1536 				default:
1537 					error = EINVAL;
1538 					break;
1539 				}
1540 				break;
1541 
1542 #if defined(IPSEC) || defined(FAST_IPSEC)
1543 			case IPV6_IPSEC_POLICY:
1544 			    {
1545 				caddr_t req = NULL;
1546 				size_t len = 0;
1547 				struct mbuf *m;
1548 
1549 				if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1550 					break;
1551 				if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1552 					break;
1553 				if (m) {
1554 					req = mtod(m, caddr_t);
1555 					len = m->m_len;
1556 				}
1557 				error = ipsec6_set_policy(in6p, optname, req,
1558 				                          len, privileged);
1559 				m_freem(m);
1560 			    }
1561 				break;
1562 #endif /* KAME IPSEC */
1563 
1564 			case IPV6_FW_ADD:
1565 			case IPV6_FW_DEL:
1566 			case IPV6_FW_FLUSH:
1567 			case IPV6_FW_ZERO:
1568 			    {
1569 				struct mbuf *m;
1570 				struct mbuf **mp = &m;
1571 
1572 				if (ip6_fw_ctl_ptr == NULL)
1573 					return EINVAL;
1574 				/* XXX */
1575 				if ((error = soopt_getm(sopt, &m)) != 0)
1576 					break;
1577 				/* XXX */
1578 				if ((error = soopt_mcopyin(sopt, m)) != 0)
1579 					break;
1580 				error = (*ip6_fw_ctl_ptr)(optname, mp);
1581 				m = *mp;
1582 			    }
1583 				break;
1584 
1585 			default:
1586 				error = ENOPROTOOPT;
1587 				break;
1588 			}
1589 			break;
1590 
1591 		case SOPT_GET:
1592 			switch (optname) {
1593 
1594 			case IPV6_PKTOPTIONS:
1595 				if (in6p->in6p_options) {
1596 					struct mbuf *m;
1597 					m = m_copym(in6p->in6p_options,
1598 					    0, M_COPYALL, 0);
1599 					error = soopt_mcopyout(sopt, m);
1600 					if (error == 0)
1601 						m_freem(m);
1602 				} else
1603 					sopt->sopt_valsize = 0;
1604 				break;
1605 
1606 			case IPV6_UNICAST_HOPS:
1607 			case IPV6_CHECKSUM:
1608 
1609 			case IPV6_FAITH:
1610 			case IPV6_V6ONLY:
1611 			case IPV6_PORTRANGE:
1612 				switch (optname) {
1613 
1614 				case IPV6_UNICAST_HOPS:
1615 					optval = in6p->in6p_hops;
1616 					break;
1617 
1618 				case IPV6_CHECKSUM:
1619 					optval = in6p->in6p_cksum;
1620 					break;
1621 
1622 				case IPV6_FAITH:
1623 					optval = OPTBIT(IN6P_FAITH);
1624 					break;
1625 
1626 				case IPV6_V6ONLY:
1627 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
1628 					break;
1629 
1630 				case IPV6_PORTRANGE:
1631 				    {
1632 					int flags;
1633 					flags = in6p->in6p_flags;
1634 					if (flags & IN6P_HIGHPORT)
1635 						optval = IPV6_PORTRANGE_HIGH;
1636 					else if (flags & IN6P_LOWPORT)
1637 						optval = IPV6_PORTRANGE_LOW;
1638 					else
1639 						optval = 0;
1640 					break;
1641 				    }
1642 				}
1643 				error = sooptcopyout(sopt, &optval,
1644 					sizeof optval);
1645 				break;
1646 
1647 			case IPV6_PKTINFO:
1648 			case IPV6_HOPLIMIT:
1649 			case IPV6_HOPOPTS:
1650 			case IPV6_RTHDR:
1651 			case IPV6_DSTOPTS:
1652 				if (optname == IPV6_HOPOPTS ||
1653 				    optname == IPV6_DSTOPTS ||
1654 				    !privileged)
1655 					return(EPERM);
1656 				switch (optname) {
1657 				case IPV6_PKTINFO:
1658 					optval = OPTBIT(IN6P_PKTINFO);
1659 					break;
1660 				case IPV6_HOPLIMIT:
1661 					optval = OPTBIT(IN6P_HOPLIMIT);
1662 					break;
1663 				case IPV6_HOPOPTS:
1664 					if (!privileged)
1665 						return(EPERM);
1666 					optval = OPTBIT(IN6P_HOPOPTS);
1667 					break;
1668 				case IPV6_RTHDR:
1669 					optval = OPTBIT(IN6P_RTHDR);
1670 					break;
1671 				case IPV6_DSTOPTS:
1672 					if (!privileged)
1673 						return(EPERM);
1674 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1675 					break;
1676 				}
1677 				error = sooptcopyout(sopt, &optval,
1678 					sizeof optval);
1679 				break;
1680 
1681 			case IPV6_MULTICAST_IF:
1682 			case IPV6_MULTICAST_HOPS:
1683 			case IPV6_MULTICAST_LOOP:
1684 			case IPV6_JOIN_GROUP:
1685 			case IPV6_LEAVE_GROUP:
1686 			    {
1687 				struct mbuf *m;
1688 				error = ip6_getmoptions(sopt->sopt_name,
1689 						in6p->in6p_moptions, &m);
1690 				if (error == 0)
1691 					error = sooptcopyout(sopt,
1692 						mtod(m, char *), m->m_len);
1693 				m_freem(m);
1694 			    }
1695 				break;
1696 
1697 #if defined(IPSEC) || defined(FAST_IPSEC)
1698 			case IPV6_IPSEC_POLICY:
1699 			  {
1700 				caddr_t req = NULL;
1701 				size_t len = 0;
1702 				struct mbuf *m = NULL;
1703 				struct mbuf **mp = &m;
1704 
1705 				error = soopt_getm(sopt, &m); /* XXX */
1706 				if (error != 0)
1707 					break;
1708 				error = soopt_mcopyin(sopt, m); /* XXX */
1709 				if (error != 0)
1710 					break;
1711 				if (m) {
1712 					req = mtod(m, caddr_t);
1713 					len = m->m_len;
1714 				}
1715 				error = ipsec6_get_policy(in6p, req, len, mp);
1716 				if (error == 0)
1717 					error = soopt_mcopyout(sopt, m); /*XXX*/
1718 				if (error == 0 && m)
1719 					m_freem(m);
1720 				break;
1721 			  }
1722 #endif /* KAME IPSEC */
1723 
1724 			case IPV6_FW_GET:
1725 			  {
1726 				struct mbuf *m;
1727 				struct mbuf **mp = &m;
1728 
1729 				if (ip6_fw_ctl_ptr == NULL)
1730 			        {
1731 					return EINVAL;
1732 				}
1733 				error = (*ip6_fw_ctl_ptr)(optname, mp);
1734 				if (error == 0)
1735 					error = soopt_mcopyout(sopt, m); /* XXX */
1736 				if (error == 0 && m)
1737 					m_freem(m);
1738 			  }
1739 				break;
1740 
1741 			default:
1742 				error = ENOPROTOOPT;
1743 				break;
1744 			}
1745 			break;
1746 		}
1747 	} else {
1748 		error = EINVAL;
1749 	}
1750 	return(error);
1751 }
1752 
1753 /*
1754  * Set up IP6 options in pcb for insertion in output packets or
1755  * specifying behavior of outgoing packets.
1756  */
1757 static int
1758 ip6_pcbopts(pktopt, m, so, sopt)
1759 	struct ip6_pktopts **pktopt;
1760 	struct mbuf *m;
1761 	struct socket *so;
1762 	struct sockopt *sopt;
1763 {
1764 	struct ip6_pktopts *opt = *pktopt;
1765 	int error = 0;
1766 	struct thread *td = sopt->sopt_td;
1767 	int priv = 0;
1768 
1769 	/* turn off any old options. */
1770 	if (opt) {
1771 #ifdef DIAGNOSTIC
1772 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
1773 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
1774 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
1775 			printf("ip6_pcbopts: all specified options are cleared.\n");
1776 #endif
1777 		ip6_clearpktopts(opt, 1, -1);
1778 	} else
1779 		opt = malloc(sizeof(*opt), M_IP6OPT, 0);
1780 	*pktopt = NULL;
1781 
1782 	if (!m || m->m_len == 0) {
1783 		/*
1784 		 * Only turning off any previous options, regardless of
1785 		 * whether the opt is just created or given.
1786 		 */
1787 		free(opt, M_IP6OPT);
1788 		return(0);
1789 	}
1790 
1791 	/*  set options specified by user. */
1792 	if (td && !suser(td))
1793 		priv = 1;
1794 	if ((error = ip6_setpktoptions(m, opt, priv, 1)) != 0) {
1795 		ip6_clearpktopts(opt, 1, -1); /* XXX: discard all options */
1796 		free(opt, M_IP6OPT);
1797 		return(error);
1798 	}
1799 	*pktopt = opt;
1800 	return(0);
1801 }
1802 
1803 /*
1804  * initialize ip6_pktopts.  beware that there are non-zero default values in
1805  * the struct.
1806  */
1807 void
1808 init_ip6pktopts(opt)
1809 	struct ip6_pktopts *opt;
1810 {
1811 
1812 	bzero(opt, sizeof(*opt));
1813 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
1814 }
1815 
1816 void
1817 ip6_clearpktopts(pktopt, needfree, optname)
1818 	struct ip6_pktopts *pktopt;
1819 	int needfree, optname;
1820 {
1821 	if (pktopt == NULL)
1822 		return;
1823 
1824 	if (optname == -1) {
1825 		if (needfree && pktopt->ip6po_pktinfo)
1826 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
1827 		pktopt->ip6po_pktinfo = NULL;
1828 	}
1829 	if (optname == -1)
1830 		pktopt->ip6po_hlim = -1;
1831 	if (optname == -1) {
1832 		if (needfree && pktopt->ip6po_nexthop)
1833 			free(pktopt->ip6po_nexthop, M_IP6OPT);
1834 		pktopt->ip6po_nexthop = NULL;
1835 	}
1836 	if (optname == -1) {
1837 		if (needfree && pktopt->ip6po_hbh)
1838 			free(pktopt->ip6po_hbh, M_IP6OPT);
1839 		pktopt->ip6po_hbh = NULL;
1840 	}
1841 	if (optname == -1) {
1842 		if (needfree && pktopt->ip6po_dest1)
1843 			free(pktopt->ip6po_dest1, M_IP6OPT);
1844 		pktopt->ip6po_dest1 = NULL;
1845 	}
1846 	if (optname == -1) {
1847 		if (needfree && pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
1848 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
1849 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
1850 		if (pktopt->ip6po_route.ro_rt) {
1851 			RTFREE(pktopt->ip6po_route.ro_rt);
1852 			pktopt->ip6po_route.ro_rt = NULL;
1853 		}
1854 	}
1855 	if (optname == -1) {
1856 		if (needfree && pktopt->ip6po_dest2)
1857 			free(pktopt->ip6po_dest2, M_IP6OPT);
1858 		pktopt->ip6po_dest2 = NULL;
1859 	}
1860 }
1861 
1862 #define PKTOPT_EXTHDRCPY(type) \
1863 do {\
1864 	if (src->type) {\
1865 		int hlen =\
1866 			(((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
1867 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
1868 		if (dst->type == NULL && canwait == M_NOWAIT)\
1869 			goto bad;\
1870 		bcopy(src->type, dst->type, hlen);\
1871 	}\
1872 } while (0)
1873 
1874 struct ip6_pktopts *
1875 ip6_copypktopts(src, canwait)
1876 	struct ip6_pktopts *src;
1877 	int canwait;
1878 {
1879 	struct ip6_pktopts *dst;
1880 
1881 	if (src == NULL) {
1882 		printf("ip6_clearpktopts: invalid argument\n");
1883 		return(NULL);
1884 	}
1885 
1886 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
1887 	if (dst == NULL && canwait == M_NOWAIT)
1888 		return (NULL);
1889 	bzero(dst, sizeof(*dst));
1890 
1891 	dst->ip6po_hlim = src->ip6po_hlim;
1892 	if (src->ip6po_pktinfo) {
1893 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
1894 					    M_IP6OPT, canwait);
1895 		if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT)
1896 			goto bad;
1897 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
1898 	}
1899 	if (src->ip6po_nexthop) {
1900 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
1901 					    M_IP6OPT, canwait);
1902 		if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT)
1903 			goto bad;
1904 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
1905 		      src->ip6po_nexthop->sa_len);
1906 	}
1907 	PKTOPT_EXTHDRCPY(ip6po_hbh);
1908 	PKTOPT_EXTHDRCPY(ip6po_dest1);
1909 	PKTOPT_EXTHDRCPY(ip6po_dest2);
1910 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
1911 	return(dst);
1912 
1913   bad:
1914 	if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
1915 	if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
1916 	if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
1917 	if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
1918 	if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
1919 	if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
1920 	free(dst, M_IP6OPT);
1921 	return(NULL);
1922 }
1923 #undef PKTOPT_EXTHDRCPY
1924 
1925 void
1926 ip6_freepcbopts(pktopt)
1927 	struct ip6_pktopts *pktopt;
1928 {
1929 	if (pktopt == NULL)
1930 		return;
1931 
1932 	ip6_clearpktopts(pktopt, 1, -1);
1933 
1934 	free(pktopt, M_IP6OPT);
1935 }
1936 
1937 /*
1938  * Set the IP6 multicast options in response to user setsockopt().
1939  */
1940 static int
1941 ip6_setmoptions(optname, im6op, m)
1942 	int optname;
1943 	struct ip6_moptions **im6op;
1944 	struct mbuf *m;
1945 {
1946 	int error = 0;
1947 	u_int loop, ifindex;
1948 	struct ipv6_mreq *mreq;
1949 	struct ifnet *ifp;
1950 	struct ip6_moptions *im6o = *im6op;
1951 	struct route_in6 ro;
1952 	struct sockaddr_in6 *dst;
1953 	struct in6_multi_mship *imm;
1954 	struct thread *td = curthread;	/* XXX */
1955 
1956 	if (im6o == NULL) {
1957 		/*
1958 		 * No multicast option buffer attached to the pcb;
1959 		 * allocate one and initialize to default values.
1960 		 */
1961 		im6o = (struct ip6_moptions *)
1962 			malloc(sizeof(*im6o), M_IPMOPTS, 0);
1963 
1964 		if (im6o == NULL)
1965 			return(ENOBUFS);
1966 		*im6op = im6o;
1967 		im6o->im6o_multicast_ifp = NULL;
1968 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1969 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
1970 		LIST_INIT(&im6o->im6o_memberships);
1971 	}
1972 
1973 	switch (optname) {
1974 
1975 	case IPV6_MULTICAST_IF:
1976 		/*
1977 		 * Select the interface for outgoing multicast packets.
1978 		 */
1979 		if (m == NULL || m->m_len != sizeof(u_int)) {
1980 			error = EINVAL;
1981 			break;
1982 		}
1983 		bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
1984 		if (ifindex < 0 || if_index < ifindex) {
1985 			error = ENXIO;	/* XXX EINVAL? */
1986 			break;
1987 		}
1988 		ifp = ifnet_byindex(ifindex);
1989 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1990 			error = EADDRNOTAVAIL;
1991 			break;
1992 		}
1993 		im6o->im6o_multicast_ifp = ifp;
1994 		break;
1995 
1996 	case IPV6_MULTICAST_HOPS:
1997 	    {
1998 		/*
1999 		 * Set the IP6 hoplimit for outgoing multicast packets.
2000 		 */
2001 		int optval;
2002 		if (m == NULL || m->m_len != sizeof(int)) {
2003 			error = EINVAL;
2004 			break;
2005 		}
2006 		bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2007 		if (optval < -1 || optval >= 256)
2008 			error = EINVAL;
2009 		else if (optval == -1)
2010 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2011 		else
2012 			im6o->im6o_multicast_hlim = optval;
2013 		break;
2014 	    }
2015 
2016 	case IPV6_MULTICAST_LOOP:
2017 		/*
2018 		 * Set the loopback flag for outgoing multicast packets.
2019 		 * Must be zero or one.
2020 		 */
2021 		if (m == NULL || m->m_len != sizeof(u_int)) {
2022 			error = EINVAL;
2023 			break;
2024 		}
2025 		bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2026 		if (loop > 1) {
2027 			error = EINVAL;
2028 			break;
2029 		}
2030 		im6o->im6o_multicast_loop = loop;
2031 		break;
2032 
2033 	case IPV6_JOIN_GROUP:
2034 		/*
2035 		 * Add a multicast group membership.
2036 		 * Group must be a valid IP6 multicast address.
2037 		 */
2038 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2039 			error = EINVAL;
2040 			break;
2041 		}
2042 		mreq = mtod(m, struct ipv6_mreq *);
2043 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2044 			/*
2045 			 * We use the unspecified address to specify to accept
2046 			 * all multicast addresses. Only super user is allowed
2047 			 * to do this.
2048 			 */
2049 			if (suser(td))
2050 			{
2051 				error = EACCES;
2052 				break;
2053 			}
2054 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2055 			error = EINVAL;
2056 			break;
2057 		}
2058 
2059 		/*
2060 		 * If the interface is specified, validate it.
2061 		 */
2062 		if (mreq->ipv6mr_interface < 0
2063 		 || if_index < mreq->ipv6mr_interface) {
2064 			error = ENXIO;	/* XXX EINVAL? */
2065 			break;
2066 		}
2067 		/*
2068 		 * If no interface was explicitly specified, choose an
2069 		 * appropriate one according to the given multicast address.
2070 		 */
2071 		if (mreq->ipv6mr_interface == 0) {
2072 			/*
2073 			 * If the multicast address is in node-local scope,
2074 			 * the interface should be a loopback interface.
2075 			 * Otherwise, look up the routing table for the
2076 			 * address, and choose the outgoing interface.
2077 			 *   XXX: is it a good approach?
2078 			 */
2079 			if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
2080 				ifp = &loif[0];
2081 			} else {
2082 				ro.ro_rt = NULL;
2083 				dst = (struct sockaddr_in6 *)&ro.ro_dst;
2084 				bzero(dst, sizeof(*dst));
2085 				dst->sin6_len = sizeof(struct sockaddr_in6);
2086 				dst->sin6_family = AF_INET6;
2087 				dst->sin6_addr = mreq->ipv6mr_multiaddr;
2088 				rtalloc((struct route *)&ro);
2089 				if (ro.ro_rt == NULL) {
2090 					error = EADDRNOTAVAIL;
2091 					break;
2092 				}
2093 				ifp = ro.ro_rt->rt_ifp;
2094 				rtfree(ro.ro_rt);
2095 			}
2096 		} else
2097 			ifp = ifnet_byindex(mreq->ipv6mr_interface);
2098 
2099 		/*
2100 		 * See if we found an interface, and confirm that it
2101 		 * supports multicast
2102 		 */
2103 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2104 			error = EADDRNOTAVAIL;
2105 			break;
2106 		}
2107 		/*
2108 		 * Put interface index into the multicast address,
2109 		 * if the address has link-local scope.
2110 		 */
2111 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2112 			mreq->ipv6mr_multiaddr.s6_addr16[1]
2113 				= htons(mreq->ipv6mr_interface);
2114 		}
2115 		/*
2116 		 * See if the membership already exists.
2117 		 */
2118 		for (imm = im6o->im6o_memberships.lh_first;
2119 		     imm != NULL; imm = imm->i6mm_chain.le_next)
2120 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
2121 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2122 					       &mreq->ipv6mr_multiaddr))
2123 				break;
2124 		if (imm != NULL) {
2125 			error = EADDRINUSE;
2126 			break;
2127 		}
2128 		/*
2129 		 * Everything looks good; add a new record to the multicast
2130 		 * address list for the given interface.
2131 		 */
2132 		imm = malloc(sizeof(*imm), M_IPMADDR, 0);
2133 		if (imm == NULL) {
2134 			error = ENOBUFS;
2135 			break;
2136 		}
2137 		if ((imm->i6mm_maddr =
2138 		     in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
2139 			free(imm, M_IPMADDR);
2140 			break;
2141 		}
2142 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2143 		break;
2144 
2145 	case IPV6_LEAVE_GROUP:
2146 		/*
2147 		 * Drop a multicast group membership.
2148 		 * Group must be a valid IP6 multicast address.
2149 		 */
2150 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2151 			error = EINVAL;
2152 			break;
2153 		}
2154 		mreq = mtod(m, struct ipv6_mreq *);
2155 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2156 			if (suser(td)) {
2157 				error = EACCES;
2158 				break;
2159 			}
2160 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2161 			error = EINVAL;
2162 			break;
2163 		}
2164 		/*
2165 		 * If an interface address was specified, get a pointer
2166 		 * to its ifnet structure.
2167 		 */
2168 		if (mreq->ipv6mr_interface < 0
2169 		 || if_index < mreq->ipv6mr_interface) {
2170 			error = ENXIO;	/* XXX EINVAL? */
2171 			break;
2172 		}
2173 		ifp = ifnet_byindex(mreq->ipv6mr_interface);
2174 		/*
2175 		 * Put interface index into the multicast address,
2176 		 * if the address has link-local scope.
2177 		 */
2178 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2179 			mreq->ipv6mr_multiaddr.s6_addr16[1]
2180 				= htons(mreq->ipv6mr_interface);
2181 		}
2182 		/*
2183 		 * Find the membership in the membership list.
2184 		 */
2185 		for (imm = im6o->im6o_memberships.lh_first;
2186 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
2187 			if ((ifp == NULL ||
2188 			     imm->i6mm_maddr->in6m_ifp == ifp) &&
2189 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2190 					       &mreq->ipv6mr_multiaddr))
2191 				break;
2192 		}
2193 		if (imm == NULL) {
2194 			/* Unable to resolve interface */
2195 			error = EADDRNOTAVAIL;
2196 			break;
2197 		}
2198 		/*
2199 		 * Give up the multicast address record to which the
2200 		 * membership points.
2201 		 */
2202 		LIST_REMOVE(imm, i6mm_chain);
2203 		in6_delmulti(imm->i6mm_maddr);
2204 		free(imm, M_IPMADDR);
2205 		break;
2206 
2207 	default:
2208 		error = EOPNOTSUPP;
2209 		break;
2210 	}
2211 
2212 	/*
2213 	 * If all options have default values, no need to keep the mbuf.
2214 	 */
2215 	if (im6o->im6o_multicast_ifp == NULL &&
2216 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2217 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2218 	    im6o->im6o_memberships.lh_first == NULL) {
2219 		free(*im6op, M_IPMOPTS);
2220 		*im6op = NULL;
2221 	}
2222 
2223 	return(error);
2224 }
2225 
2226 /*
2227  * Return the IP6 multicast options in response to user getsockopt().
2228  */
2229 static int
2230 ip6_getmoptions(optname, im6o, mp)
2231 	int optname;
2232 	struct ip6_moptions *im6o;
2233 	struct mbuf **mp;
2234 {
2235 	u_int *hlim, *loop, *ifindex;
2236 
2237 	*mp = m_get(0, MT_HEADER);		/* XXX */
2238 
2239 	switch (optname) {
2240 
2241 	case IPV6_MULTICAST_IF:
2242 		ifindex = mtod(*mp, u_int *);
2243 		(*mp)->m_len = sizeof(u_int);
2244 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2245 			*ifindex = 0;
2246 		else
2247 			*ifindex = im6o->im6o_multicast_ifp->if_index;
2248 		return(0);
2249 
2250 	case IPV6_MULTICAST_HOPS:
2251 		hlim = mtod(*mp, u_int *);
2252 		(*mp)->m_len = sizeof(u_int);
2253 		if (im6o == NULL)
2254 			*hlim = ip6_defmcasthlim;
2255 		else
2256 			*hlim = im6o->im6o_multicast_hlim;
2257 		return(0);
2258 
2259 	case IPV6_MULTICAST_LOOP:
2260 		loop = mtod(*mp, u_int *);
2261 		(*mp)->m_len = sizeof(u_int);
2262 		if (im6o == NULL)
2263 			*loop = ip6_defmcasthlim;
2264 		else
2265 			*loop = im6o->im6o_multicast_loop;
2266 		return(0);
2267 
2268 	default:
2269 		return(EOPNOTSUPP);
2270 	}
2271 }
2272 
2273 /*
2274  * Discard the IP6 multicast options.
2275  */
2276 void
2277 ip6_freemoptions(im6o)
2278 	struct ip6_moptions *im6o;
2279 {
2280 	struct in6_multi_mship *imm;
2281 
2282 	if (im6o == NULL)
2283 		return;
2284 
2285 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2286 		LIST_REMOVE(imm, i6mm_chain);
2287 		if (imm->i6mm_maddr)
2288 			in6_delmulti(imm->i6mm_maddr);
2289 		free(imm, M_IPMADDR);
2290 	}
2291 	free(im6o, M_IPMOPTS);
2292 }
2293 
2294 /*
2295  * Set IPv6 outgoing packet options based on advanced API.
2296  */
2297 int
2298 ip6_setpktoptions(control, opt, priv, needcopy)
2299 	struct mbuf *control;
2300 	struct ip6_pktopts *opt;
2301 	int priv, needcopy;
2302 {
2303 	struct cmsghdr *cm = 0;
2304 
2305 	if (control == 0 || opt == 0)
2306 		return(EINVAL);
2307 
2308 	init_ip6pktopts(opt);
2309 
2310 	/*
2311 	 * XXX: Currently, we assume all the optional information is stored
2312 	 * in a single mbuf.
2313 	 */
2314 	if (control->m_next)
2315 		return(EINVAL);
2316 
2317 	for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2318 		     control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2319 		cm = mtod(control, struct cmsghdr *);
2320 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2321 			return(EINVAL);
2322 		if (cm->cmsg_level != IPPROTO_IPV6)
2323 			continue;
2324 
2325 		/*
2326 		 * XXX should check if RFC2292 API is mixed with 2292bis API
2327 		 */
2328 		switch (cm->cmsg_type) {
2329 		case IPV6_PKTINFO:
2330 			if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo)))
2331 				return(EINVAL);
2332 			if (needcopy) {
2333 				/* XXX: Is it really WAITOK? */
2334 				opt->ip6po_pktinfo =
2335 					malloc(sizeof(struct in6_pktinfo),
2336 					       M_IP6OPT, 0);
2337 				bcopy(CMSG_DATA(cm), opt->ip6po_pktinfo,
2338 				    sizeof(struct in6_pktinfo));
2339 			} else
2340 				opt->ip6po_pktinfo =
2341 					(struct in6_pktinfo *)CMSG_DATA(cm);
2342 			if (opt->ip6po_pktinfo->ipi6_ifindex &&
2343 			    IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr))
2344 				opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] =
2345 					htons(opt->ip6po_pktinfo->ipi6_ifindex);
2346 
2347 			if (opt->ip6po_pktinfo->ipi6_ifindex > if_index
2348 			 || opt->ip6po_pktinfo->ipi6_ifindex < 0) {
2349 				return(ENXIO);
2350 			}
2351 
2352 			/*
2353 			 * Check if the requested source address is indeed a
2354 			 * unicast address assigned to the node, and can be
2355 			 * used as the packet's source address.
2356 			 */
2357 			if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
2358 				struct in6_ifaddr *ia6;
2359 				struct sockaddr_in6 sin6;
2360 
2361 				bzero(&sin6, sizeof(sin6));
2362 				sin6.sin6_len = sizeof(sin6);
2363 				sin6.sin6_family = AF_INET6;
2364 				sin6.sin6_addr =
2365 					opt->ip6po_pktinfo->ipi6_addr;
2366 				ia6 = (struct in6_ifaddr *)ifa_ifwithaddr(sin6tosa(&sin6));
2367 				if (ia6 == NULL ||
2368 				    (ia6->ia6_flags & (IN6_IFF_ANYCAST |
2369 						       IN6_IFF_NOTREADY)) != 0)
2370 					return(EADDRNOTAVAIL);
2371 			}
2372 			break;
2373 
2374 		case IPV6_HOPLIMIT:
2375 			if (cm->cmsg_len != CMSG_LEN(sizeof(int)))
2376 				return(EINVAL);
2377 
2378 			opt->ip6po_hlim = *(int *)CMSG_DATA(cm);
2379 			if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255)
2380 				return(EINVAL);
2381 			break;
2382 
2383 		case IPV6_NEXTHOP:
2384 			if (!priv)
2385 				return(EPERM);
2386 
2387 			if (cm->cmsg_len < sizeof(u_char) ||
2388 			    /* check if cmsg_len is large enough for sa_len */
2389 			    cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm)))
2390 				return(EINVAL);
2391 
2392 			if (needcopy) {
2393 				opt->ip6po_nexthop =
2394 					malloc(*CMSG_DATA(cm),
2395 					       M_IP6OPT, 0);
2396 				bcopy(CMSG_DATA(cm),
2397 				      opt->ip6po_nexthop,
2398 				      *CMSG_DATA(cm));
2399 			} else
2400 				opt->ip6po_nexthop =
2401 					(struct sockaddr *)CMSG_DATA(cm);
2402 			break;
2403 
2404 		case IPV6_HOPOPTS:
2405 		{
2406 			struct ip6_hbh *hbh;
2407 			int hbhlen;
2408 
2409 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh)))
2410 				return(EINVAL);
2411 			hbh = (struct ip6_hbh *)CMSG_DATA(cm);
2412 			hbhlen = (hbh->ip6h_len + 1) << 3;
2413 			if (cm->cmsg_len != CMSG_LEN(hbhlen))
2414 				return(EINVAL);
2415 
2416 			if (needcopy) {
2417 				opt->ip6po_hbh =
2418 					malloc(hbhlen, M_IP6OPT, 0);
2419 				bcopy(hbh, opt->ip6po_hbh, hbhlen);
2420 			} else
2421 				opt->ip6po_hbh = hbh;
2422 			break;
2423 		}
2424 
2425 		case IPV6_DSTOPTS:
2426 		{
2427 			struct ip6_dest *dest, **newdest;
2428 			int destlen;
2429 
2430 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest)))
2431 				return(EINVAL);
2432 			dest = (struct ip6_dest *)CMSG_DATA(cm);
2433 			destlen = (dest->ip6d_len + 1) << 3;
2434 			if (cm->cmsg_len != CMSG_LEN(destlen))
2435 				return(EINVAL);
2436 
2437 			/*
2438 			 * The old advacned API is ambiguous on this
2439 			 * point. Our approach is to determine the
2440 			 * position based according to the existence
2441 			 * of a routing header. Note, however, that
2442 			 * this depends on the order of the extension
2443 			 * headers in the ancillary data; the 1st part
2444 			 * of the destination options header must
2445 			 * appear before the routing header in the
2446 			 * ancillary data, too.
2447 			 * RFC2292bis solved the ambiguity by
2448 			 * introducing separate cmsg types.
2449 			 */
2450 			if (opt->ip6po_rthdr == NULL)
2451 				newdest = &opt->ip6po_dest1;
2452 			else
2453 				newdest = &opt->ip6po_dest2;
2454 
2455 			if (needcopy) {
2456 				*newdest = malloc(destlen, M_IP6OPT, 0);
2457 				bcopy(dest, *newdest, destlen);
2458 			} else
2459 				*newdest = dest;
2460 
2461 			break;
2462 		}
2463 
2464 		case IPV6_RTHDR:
2465 		{
2466 			struct ip6_rthdr *rth;
2467 			int rthlen;
2468 
2469 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr)))
2470 				return(EINVAL);
2471 			rth = (struct ip6_rthdr *)CMSG_DATA(cm);
2472 			rthlen = (rth->ip6r_len + 1) << 3;
2473 			if (cm->cmsg_len != CMSG_LEN(rthlen))
2474 				return(EINVAL);
2475 
2476 			switch (rth->ip6r_type) {
2477 			case IPV6_RTHDR_TYPE_0:
2478 				/* must contain one addr */
2479 				if (rth->ip6r_len == 0)
2480 					return(EINVAL);
2481 				/* length must be even */
2482 				if (rth->ip6r_len % 2)
2483 					return(EINVAL);
2484 				if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2485 					return(EINVAL);
2486 				break;
2487 			default:
2488 				return(EINVAL);	/* not supported */
2489 			}
2490 
2491 			if (needcopy) {
2492 				opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT,
2493 							  0);
2494 				bcopy(rth, opt->ip6po_rthdr, rthlen);
2495 			} else
2496 				opt->ip6po_rthdr = rth;
2497 
2498 			break;
2499 		}
2500 
2501 		default:
2502 			return(ENOPROTOOPT);
2503 		}
2504 	}
2505 
2506 	return(0);
2507 }
2508 
2509 /*
2510  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2511  * packet to the input queue of a specified interface.  Note that this
2512  * calls the output routine of the loopback "driver", but with an interface
2513  * pointer that might NOT be &loif -- easier than replicating that code here.
2514  */
2515 void
2516 ip6_mloopback(ifp, m, dst)
2517 	struct ifnet *ifp;
2518 	struct mbuf *m;
2519 	struct sockaddr_in6 *dst;
2520 {
2521 	struct mbuf *copym;
2522 	struct ip6_hdr *ip6;
2523 
2524 	copym = m_copy(m, 0, M_COPYALL);
2525 	if (copym == NULL)
2526 		return;
2527 
2528 	/*
2529 	 * Make sure to deep-copy IPv6 header portion in case the data
2530 	 * is in an mbuf cluster, so that we can safely override the IPv6
2531 	 * header portion later.
2532 	 */
2533 	if ((copym->m_flags & M_EXT) != 0 ||
2534 	    copym->m_len < sizeof(struct ip6_hdr)) {
2535 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
2536 		if (copym == NULL)
2537 			return;
2538 	}
2539 
2540 #ifdef DIAGNOSTIC
2541 	if (copym->m_len < sizeof(*ip6)) {
2542 		m_freem(copym);
2543 		return;
2544 	}
2545 #endif
2546 
2547 	ip6 = mtod(copym, struct ip6_hdr *);
2548 #ifndef SCOPEDROUTING
2549 	/*
2550 	 * clear embedded scope identifiers if necessary.
2551 	 * in6_clearscope will touch the addresses only when necessary.
2552 	 */
2553 	in6_clearscope(&ip6->ip6_src);
2554 	in6_clearscope(&ip6->ip6_dst);
2555 #endif
2556 
2557 	(void)if_simloop(ifp, copym, dst->sin6_family, 0);
2558 }
2559 
2560 /*
2561  * Chop IPv6 header off from the payload.
2562  */
2563 static int
2564 ip6_splithdr(m, exthdrs)
2565 	struct mbuf *m;
2566 	struct ip6_exthdrs *exthdrs;
2567 {
2568 	struct mbuf *mh;
2569 	struct ip6_hdr *ip6;
2570 
2571 	ip6 = mtod(m, struct ip6_hdr *);
2572 	if (m->m_len > sizeof(*ip6)) {
2573 		MGETHDR(mh, M_NOWAIT, MT_HEADER);
2574 		if (mh == 0) {
2575 			m_freem(m);
2576 			return ENOBUFS;
2577 		}
2578 		M_MOVE_PKTHDR(mh, m);
2579 		MH_ALIGN(mh, sizeof(*ip6));
2580 		m->m_len -= sizeof(*ip6);
2581 		m->m_data += sizeof(*ip6);
2582 		mh->m_next = m;
2583 		m = mh;
2584 		m->m_len = sizeof(*ip6);
2585 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2586 	}
2587 	exthdrs->ip6e_ip6 = m;
2588 	return 0;
2589 }
2590 
2591 /*
2592  * Compute IPv6 extension header length.
2593  */
2594 int
2595 ip6_optlen(in6p)
2596 	struct in6pcb *in6p;
2597 {
2598 	int len;
2599 
2600 	if (!in6p->in6p_outputopts)
2601 		return 0;
2602 
2603 	len = 0;
2604 #define elen(x) \
2605     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2606 
2607 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
2608 	if (in6p->in6p_outputopts->ip6po_rthdr)
2609 		/* dest1 is valid with rthdr only */
2610 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
2611 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
2612 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
2613 	return len;
2614 #undef elen
2615 }
2616