xref: /netbsd-src/sys/netinet6/ip6_output.c (revision 001c68bd94f75ce9270b69227c4199fbf34ee396)
1 /*	$NetBSD: ip6_output.c,v 1.61 2003/06/06 08:13:43 itojun Exp $	*/
2 /*	$KAME: ip6_output.c,v 1.172 2001/03/25 09:55:56 itojun 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 <sys/cdefs.h>
69 __KERNEL_RCSID(0, "$NetBSD: ip6_output.c,v 1.61 2003/06/06 08:13:43 itojun Exp $");
70 
71 #include "opt_inet.h"
72 #include "opt_ipsec.h"
73 #include "opt_pfil_hooks.h"
74 
75 #include <sys/param.h>
76 #include <sys/malloc.h>
77 #include <sys/mbuf.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/proc.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 <netinet/ip6.h>
94 #include <netinet/icmp6.h>
95 #include <netinet6/ip6_var.h>
96 #include <netinet6/in6_pcb.h>
97 #include <netinet6/nd6.h>
98 
99 #ifdef IPSEC
100 #include <netinet6/ipsec.h>
101 #include <netkey/key.h>
102 #endif /* IPSEC */
103 
104 #include "loop.h"
105 
106 #include <net/net_osdep.h>
107 
108 #ifdef PFIL_HOOKS
109 extern struct pfil_head inet6_pfil_hook;	/* XXX */
110 #endif
111 
112 struct ip6_exthdrs {
113 	struct mbuf *ip6e_ip6;
114 	struct mbuf *ip6e_hbh;
115 	struct mbuf *ip6e_dest1;
116 	struct mbuf *ip6e_rthdr;
117 	struct mbuf *ip6e_dest2;
118 };
119 
120 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
121 	struct socket *));
122 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *));
123 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **));
124 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int));
125 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
126 	struct ip6_frag **));
127 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t));
128 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *));
129 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *,
130 	struct ifnet *, struct in6_addr *, u_long *));
131 
132 extern struct ifnet loif[NLOOP];
133 
134 /*
135  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
136  * header (with pri, len, nxt, hlim, src, dst).
137  * This function may modify ver and hlim only.
138  * The mbuf chain containing the packet will be freed.
139  * The mbuf opt, if present, will not be freed.
140  *
141  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
142  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
143  * which is rt_rmx.rmx_mtu.
144  */
145 int
146 ip6_output(m0, opt, ro, flags, im6o, ifpp)
147 	struct mbuf *m0;
148 	struct ip6_pktopts *opt;
149 	struct route_in6 *ro;
150 	int flags;
151 	struct ip6_moptions *im6o;
152 	struct ifnet **ifpp;		/* XXX: just for statistics */
153 {
154 	struct ip6_hdr *ip6, *mhip6;
155 	struct ifnet *ifp, *origifp;
156 	struct mbuf *m = m0;
157 	int hlen, tlen, len, off;
158 	struct route_in6 ip6route;
159 	struct sockaddr_in6 *dst;
160 	int error = 0;
161 	struct in6_ifaddr *ia;
162 	u_long mtu;
163 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
164 	struct ip6_exthdrs exthdrs;
165 	struct in6_addr finaldst;
166 	struct route_in6 *ro_pmtu = NULL;
167 	int hdrsplit = 0;
168 	int needipsec = 0;
169 #ifdef IPSEC
170 	int needipsectun = 0;
171 	struct socket *so;
172 	struct secpolicy *sp = NULL;
173 
174 	/* for AH processing. stupid to have "socket" variable in IP layer... */
175 	so = ipsec_getsocket(m);
176 	(void)ipsec_setsocket(m, NULL);
177 	ip6 = mtod(m, struct ip6_hdr *);
178 #endif /* IPSEC */
179 
180 #define MAKE_EXTHDR(hp, mp)						\
181     do {								\
182 	if (hp) {							\
183 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
184 		error = ip6_copyexthdr((mp), (caddr_t)(hp), 		\
185 		    ((eh)->ip6e_len + 1) << 3);				\
186 		if (error)						\
187 			goto freehdrs;					\
188 	}								\
189     } while (/*CONSTCOND*/ 0)
190 
191 	bzero(&exthdrs, sizeof(exthdrs));
192 	if (opt) {
193 		/* Hop-by-Hop options header */
194 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
195 		/* Destination options header(1st part) */
196 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
197 		/* Routing header */
198 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
199 		/* Destination options header(2nd part) */
200 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
201 	}
202 
203 #ifdef IPSEC
204 	/* get a security policy for this packet */
205 	if (so == NULL)
206 		sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
207 	else
208 		sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
209 
210 	if (sp == NULL) {
211 		ipsec6stat.out_inval++;
212 		goto freehdrs;
213 	}
214 
215 	error = 0;
216 
217 	/* check policy */
218 	switch (sp->policy) {
219 	case IPSEC_POLICY_DISCARD:
220 		/*
221 		 * This packet is just discarded.
222 		 */
223 		ipsec6stat.out_polvio++;
224 		goto freehdrs;
225 
226 	case IPSEC_POLICY_BYPASS:
227 	case IPSEC_POLICY_NONE:
228 		/* no need to do IPsec. */
229 		needipsec = 0;
230 		break;
231 
232 	case IPSEC_POLICY_IPSEC:
233 		if (sp->req == NULL) {
234 			/* XXX should be panic ? */
235 			printf("ip6_output: No IPsec request specified.\n");
236 			error = EINVAL;
237 			goto freehdrs;
238 		}
239 		needipsec = 1;
240 		break;
241 
242 	case IPSEC_POLICY_ENTRUST:
243 	default:
244 		printf("ip6_output: Invalid policy found. %d\n", sp->policy);
245 	}
246 #endif /* IPSEC */
247 
248 	/*
249 	 * Calculate the total length of the extension header chain.
250 	 * Keep the length of the unfragmentable part for fragmentation.
251 	 */
252 	optlen = 0;
253 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
254 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
255 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
256 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
257 	/* NOTE: we don't add AH/ESP length here. do that later. */
258 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
259 
260 	/*
261 	 * If we need IPsec, or there is at least one extension header,
262 	 * separate IP6 header from the payload.
263 	 */
264 	if ((needipsec || optlen) && !hdrsplit) {
265 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
266 			m = NULL;
267 			goto freehdrs;
268 		}
269 		m = exthdrs.ip6e_ip6;
270 		hdrsplit++;
271 	}
272 
273 	/* adjust pointer */
274 	ip6 = mtod(m, struct ip6_hdr *);
275 
276 	/* adjust mbuf packet header length */
277 	m->m_pkthdr.len += optlen;
278 	plen = m->m_pkthdr.len - sizeof(*ip6);
279 
280 	/* If this is a jumbo payload, insert a jumbo payload option. */
281 	if (plen > IPV6_MAXPACKET) {
282 		if (!hdrsplit) {
283 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
284 				m = NULL;
285 				goto freehdrs;
286 			}
287 			m = exthdrs.ip6e_ip6;
288 			hdrsplit++;
289 		}
290 		/* adjust pointer */
291 		ip6 = mtod(m, struct ip6_hdr *);
292 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
293 			goto freehdrs;
294 		ip6->ip6_plen = 0;
295 	} else
296 		ip6->ip6_plen = htons(plen);
297 
298 	/*
299 	 * Concatenate headers and fill in next header fields.
300 	 * Here we have, on "m"
301 	 *	IPv6 payload
302 	 * and we insert headers accordingly.  Finally, we should be getting:
303 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
304 	 *
305 	 * during the header composing process, "m" points to IPv6 header.
306 	 * "mprev" points to an extension header prior to esp.
307 	 */
308 	{
309 		u_char *nexthdrp = &ip6->ip6_nxt;
310 		struct mbuf *mprev = m;
311 
312 		/*
313 		 * we treat dest2 specially.  this makes IPsec processing
314 		 * much easier.
315 		 *
316 		 * result: IPv6 dest2 payload
317 		 * m and mprev will point to IPv6 header.
318 		 */
319 		if (exthdrs.ip6e_dest2) {
320 			if (!hdrsplit)
321 				panic("assumption failed: hdr not split");
322 			exthdrs.ip6e_dest2->m_next = m->m_next;
323 			m->m_next = exthdrs.ip6e_dest2;
324 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
325 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
326 		}
327 
328 #define MAKE_CHAIN(m, mp, p, i)\
329     do {\
330 	if (m) {\
331 		if (!hdrsplit) \
332 			panic("assumption failed: hdr not split"); \
333 		*mtod((m), u_char *) = *(p);\
334 		*(p) = (i);\
335 		p = mtod((m), u_char *);\
336 		(m)->m_next = (mp)->m_next;\
337 		(mp)->m_next = (m);\
338 		(mp) = (m);\
339 	}\
340     } while (/*CONSTCOND*/ 0)
341 		/*
342 		 * result: IPv6 hbh dest1 rthdr dest2 payload
343 		 * m will point to IPv6 header.  mprev will point to the
344 		 * extension header prior to dest2 (rthdr in the above case).
345 		 */
346 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
347 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
348 		    IPPROTO_DSTOPTS);
349 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
350 		    IPPROTO_ROUTING);
351 
352 #ifdef IPSEC
353 		if (!needipsec)
354 			goto skip_ipsec2;
355 
356 		/*
357 		 * pointers after IPsec headers are not valid any more.
358 		 * other pointers need a great care too.
359 		 * (IPsec routines should not mangle mbufs prior to AH/ESP)
360 		 */
361 		exthdrs.ip6e_dest2 = NULL;
362 
363 	    {
364 		struct ip6_rthdr *rh = NULL;
365 		int segleft_org = 0;
366 		struct ipsec_output_state state;
367 
368 		if (exthdrs.ip6e_rthdr) {
369 			rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
370 			segleft_org = rh->ip6r_segleft;
371 			rh->ip6r_segleft = 0;
372 		}
373 
374 		bzero(&state, sizeof(state));
375 		state.m = m;
376 		error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
377 			&needipsectun);
378 		m = state.m;
379 		if (error) {
380 			/* mbuf is already reclaimed in ipsec6_output_trans. */
381 			m = NULL;
382 			switch (error) {
383 			case EHOSTUNREACH:
384 			case ENETUNREACH:
385 			case EMSGSIZE:
386 			case ENOBUFS:
387 			case ENOMEM:
388 				break;
389 			default:
390 				printf("ip6_output (ipsec): error code %d\n", error);
391 				/* FALLTHROUGH */
392 			case ENOENT:
393 				/* don't show these error codes to the user */
394 				error = 0;
395 				break;
396 			}
397 			goto bad;
398 		}
399 		if (exthdrs.ip6e_rthdr) {
400 			/* ah6_output doesn't modify mbuf chain */
401 			rh->ip6r_segleft = segleft_org;
402 		}
403 	    }
404 skip_ipsec2:;
405 #endif
406 	}
407 
408 	/*
409 	 * If there is a routing header, replace destination address field
410 	 * with the first hop of the routing header.
411 	 */
412 	if (exthdrs.ip6e_rthdr) {
413 		struct ip6_rthdr *rh;
414 		struct ip6_rthdr0 *rh0;
415 		struct in6_addr *addr;
416 
417 		rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
418 		    struct ip6_rthdr *));
419 		finaldst = ip6->ip6_dst;
420 		switch (rh->ip6r_type) {
421 		case IPV6_RTHDR_TYPE_0:
422 			 rh0 = (struct ip6_rthdr0 *)rh;
423 			 addr = (struct in6_addr *)(rh0 + 1);
424 			 ip6->ip6_dst = addr[0];
425 			 bcopy(&addr[1], &addr[0],
426 			     sizeof(struct in6_addr) * (rh0->ip6r0_segleft - 1));
427 			 addr[rh0->ip6r0_segleft - 1] = finaldst;
428 			 break;
429 		default:	/* is it possible? */
430 			 error = EINVAL;
431 			 goto bad;
432 		}
433 	}
434 
435 	/* Source address validation */
436 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
437 	    (flags & IPV6_UNSPECSRC) == 0) {
438 		error = EOPNOTSUPP;
439 		ip6stat.ip6s_badscope++;
440 		goto bad;
441 	}
442 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
443 		error = EOPNOTSUPP;
444 		ip6stat.ip6s_badscope++;
445 		goto bad;
446 	}
447 
448 	ip6stat.ip6s_localout++;
449 
450 	/*
451 	 * Route packet.
452 	 */
453 	if (ro == 0) {
454 		ro = &ip6route;
455 		bzero((caddr_t)ro, sizeof(*ro));
456 	}
457 	ro_pmtu = ro;
458 	if (opt && opt->ip6po_rthdr)
459 		ro = &opt->ip6po_route;
460 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
461 	/*
462 	 * If there is a cached route,
463 	 * check that it is to the same destination
464 	 * and is still up. If not, free it and try again.
465 	 */
466 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
467 	    dst->sin6_family != AF_INET6 ||
468 	    !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
469 		RTFREE(ro->ro_rt);
470 		ro->ro_rt = (struct rtentry *)0;
471 	}
472 	if (ro->ro_rt == 0) {
473 		bzero(dst, sizeof(*dst));
474 		dst->sin6_family = AF_INET6;
475 		dst->sin6_len = sizeof(struct sockaddr_in6);
476 		dst->sin6_addr = ip6->ip6_dst;
477 	}
478 #ifdef IPSEC
479 	if (needipsec && needipsectun) {
480 		struct ipsec_output_state state;
481 
482 		/*
483 		 * All the extension headers will become inaccessible
484 		 * (since they can be encrypted).
485 		 * Don't panic, we need no more updates to extension headers
486 		 * on inner IPv6 packet (since they are now encapsulated).
487 		 *
488 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
489 		 */
490 		bzero(&exthdrs, sizeof(exthdrs));
491 		exthdrs.ip6e_ip6 = m;
492 
493 		bzero(&state, sizeof(state));
494 		state.m = m;
495 		state.ro = (struct route *)ro;
496 		state.dst = (struct sockaddr *)dst;
497 
498 		error = ipsec6_output_tunnel(&state, sp, flags);
499 
500 		m = state.m;
501 		ro = (struct route_in6 *)state.ro;
502 		dst = (struct sockaddr_in6 *)state.dst;
503 		if (error) {
504 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
505 			m0 = m = NULL;
506 			m = NULL;
507 			switch (error) {
508 			case EHOSTUNREACH:
509 			case ENETUNREACH:
510 			case EMSGSIZE:
511 			case ENOBUFS:
512 			case ENOMEM:
513 				break;
514 			default:
515 				printf("ip6_output (ipsec): error code %d\n", error);
516 				/* FALLTHROUGH */
517 			case ENOENT:
518 				/* don't show these error codes to the user */
519 				error = 0;
520 				break;
521 			}
522 			goto bad;
523 		}
524 
525 		exthdrs.ip6e_ip6 = m;
526 	}
527 #endif /* IPSEC */
528 
529 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
530 		/* Unicast */
531 
532 #define ifatoia6(ifa)	((struct in6_ifaddr *)(ifa))
533 #define sin6tosa(sin6)	((struct sockaddr *)(sin6))
534 		/* xxx
535 		 * interface selection comes here
536 		 * if an interface is specified from an upper layer,
537 		 * ifp must point it.
538 		 */
539 		if (ro->ro_rt == 0) {
540 			/*
541 			 * non-bsdi always clone routes, if parent is
542 			 * PRF_CLONING.
543 			 */
544 			rtalloc((struct route *)ro);
545 		}
546 		if (ro->ro_rt == 0) {
547 			ip6stat.ip6s_noroute++;
548 			error = EHOSTUNREACH;
549 			/* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
550 			goto bad;
551 		}
552 		ia = ifatoia6(ro->ro_rt->rt_ifa);
553 		ifp = ro->ro_rt->rt_ifp;
554 		ro->ro_rt->rt_use++;
555 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
556 			dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
557 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
558 
559 		in6_ifstat_inc(ifp, ifs6_out_request);
560 
561 		/*
562 		 * Check if the outgoing interface conflicts with
563 		 * the interface specified by ifi6_ifindex (if specified).
564 		 * Note that loopback interface is always okay.
565 		 * (this may happen when we are sending a packet to one of
566 		 *  our own addresses.)
567 		 */
568 		if (opt && opt->ip6po_pktinfo
569 		 && opt->ip6po_pktinfo->ipi6_ifindex) {
570 			if (!(ifp->if_flags & IFF_LOOPBACK)
571 			 && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
572 				ip6stat.ip6s_noroute++;
573 				in6_ifstat_inc(ifp, ifs6_out_discard);
574 				error = EHOSTUNREACH;
575 				goto bad;
576 			}
577 		}
578 
579 		if (opt && opt->ip6po_hlim != -1)
580 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
581 	} else {
582 		/* Multicast */
583 		struct	in6_multi *in6m;
584 
585 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
586 
587 		/*
588 		 * See if the caller provided any multicast options
589 		 */
590 		ifp = NULL;
591 		if (im6o != NULL) {
592 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
593 			if (im6o->im6o_multicast_ifp != NULL)
594 				ifp = im6o->im6o_multicast_ifp;
595 		} else
596 			ip6->ip6_hlim = ip6_defmcasthlim;
597 
598 		/*
599 		 * See if the caller provided the outgoing interface
600 		 * as an ancillary data.
601 		 * Boundary check for ifindex is assumed to be already done.
602 		 */
603 		if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
604 			ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex];
605 
606 		/*
607 		 * If the destination is a node-local scope multicast,
608 		 * the packet should be loop-backed only.
609 		 */
610 		if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
611 			/*
612 			 * If the outgoing interface is already specified,
613 			 * it should be a loopback interface.
614 			 */
615 			if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) {
616 				ip6stat.ip6s_badscope++;
617 				error = ENETUNREACH; /* XXX: better error? */
618 				/* XXX correct ifp? */
619 				in6_ifstat_inc(ifp, ifs6_out_discard);
620 				goto bad;
621 			} else {
622 				ifp = &loif[0];
623 			}
624 		}
625 
626 		if (opt && opt->ip6po_hlim != -1)
627 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
628 
629 		/*
630 		 * If caller did not provide an interface lookup a
631 		 * default in the routing table.  This is either a
632 		 * default for the speicfied group (i.e. a host
633 		 * route), or a multicast default (a route for the
634 		 * ``net'' ff00::/8).
635 		 */
636 		if (ifp == NULL) {
637 			if (ro->ro_rt == 0) {
638 				ro->ro_rt = rtalloc1((struct sockaddr *)
639 				    &ro->ro_dst, 0);
640 			}
641 			if (ro->ro_rt == 0) {
642 				ip6stat.ip6s_noroute++;
643 				error = EHOSTUNREACH;
644 				/* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
645 				goto bad;
646 			}
647 			ia = ifatoia6(ro->ro_rt->rt_ifa);
648 			ifp = ro->ro_rt->rt_ifp;
649 			ro->ro_rt->rt_use++;
650 		}
651 
652 		if ((flags & IPV6_FORWARDING) == 0)
653 			in6_ifstat_inc(ifp, ifs6_out_request);
654 		in6_ifstat_inc(ifp, ifs6_out_mcast);
655 
656 		/*
657 		 * Confirm that the outgoing interface supports multicast.
658 		 */
659 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
660 			ip6stat.ip6s_noroute++;
661 			in6_ifstat_inc(ifp, ifs6_out_discard);
662 			error = ENETUNREACH;
663 			goto bad;
664 		}
665 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
666 		if (in6m != NULL &&
667 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
668 			/*
669 			 * If we belong to the destination multicast group
670 			 * on the outgoing interface, and the caller did not
671 			 * forbid loopback, loop back a copy.
672 			 */
673 			ip6_mloopback(ifp, m, dst);
674 		} else {
675 			/*
676 			 * If we are acting as a multicast router, perform
677 			 * multicast forwarding as if the packet had just
678 			 * arrived on the interface to which we are about
679 			 * to send.  The multicast forwarding function
680 			 * recursively calls this function, using the
681 			 * IPV6_FORWARDING flag to prevent infinite recursion.
682 			 *
683 			 * Multicasts that are looped back by ip6_mloopback(),
684 			 * above, will be forwarded by the ip6_input() routine,
685 			 * if necessary.
686 			 */
687 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
688 				if (ip6_mforward(ip6, ifp, m) != 0) {
689 					m_freem(m);
690 					goto done;
691 				}
692 			}
693 		}
694 		/*
695 		 * Multicasts with a hoplimit of zero may be looped back,
696 		 * above, but must not be transmitted on a network.
697 		 * Also, multicasts addressed to the loopback interface
698 		 * are not sent -- the above call to ip6_mloopback() will
699 		 * loop back a copy if this host actually belongs to the
700 		 * destination group on the loopback interface.
701 		 */
702 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
703 			m_freem(m);
704 			goto done;
705 		}
706 	}
707 
708 	/*
709 	 * Fill the outgoing inteface to tell the upper layer
710 	 * to increment per-interface statistics.
711 	 */
712 	if (ifpp)
713 		*ifpp = ifp;
714 
715 	/* Determine path MTU. */
716 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu)) != 0)
717 		goto bad;
718 
719 	/*
720 	 * The caller of this function may specify to use the minimum MTU
721 	 * in some cases.
722 	 */
723 	if (mtu > IPV6_MMTU) {
724 		if ((flags & IPV6_MINMTU))
725 			mtu = IPV6_MMTU;
726 	}
727 
728 	/* Fake scoped addresses */
729 	if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
730 		/*
731 		 * If source or destination address is a scoped address, and
732 		 * the packet is going to be sent to a loopback interface,
733 		 * we should keep the original interface.
734 		 */
735 
736 		/*
737 		 * XXX: this is a very experimental and temporary solution.
738 		 * We eventually have sockaddr_in6 and use the sin6_scope_id
739 		 * field of the structure here.
740 		 * We rely on the consistency between two scope zone ids
741 		 * of source add destination, which should already be assured
742 		 * Larger scopes than link will be supported in the near
743 		 * future.
744 		 */
745 		origifp = NULL;
746 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
747 			origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
748 		else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
749 			origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
750 		/*
751 		 * XXX: origifp can be NULL even in those two cases above.
752 		 * For example, if we remove the (only) link-local address
753 		 * from the loopback interface, and try to send a link-local
754 		 * address without link-id information.  Then the source
755 		 * address is ::1, and the destination address is the
756 		 * link-local address with its s6_addr16[1] being zero.
757 		 * What is worse, if the packet goes to the loopback interface
758 		 * by a default rejected route, the null pointer would be
759 		 * passed to looutput, and the kernel would hang.
760 		 * The following last resort would prevent such disaster.
761 		 */
762 		if (origifp == NULL)
763 			origifp = ifp;
764 	} else
765 		origifp = ifp;
766 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
767 		ip6->ip6_src.s6_addr16[1] = 0;
768 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
769 		ip6->ip6_dst.s6_addr16[1] = 0;
770 
771 	/*
772 	 * If the outgoing packet contains a hop-by-hop options header,
773 	 * it must be examined and processed even by the source node.
774 	 * (RFC 2460, section 4.)
775 	 */
776 	if (exthdrs.ip6e_hbh) {
777 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
778 		u_int32_t dummy1; /* XXX unused */
779 		u_int32_t dummy2; /* XXX unused */
780 
781 		/*
782 		 *  XXX: if we have to send an ICMPv6 error to the sender,
783 		 *       we need the M_LOOP flag since icmp6_error() expects
784 		 *       the IPv6 and the hop-by-hop options header are
785 		 *       continuous unless the flag is set.
786 		 */
787 		m->m_flags |= M_LOOP;
788 		m->m_pkthdr.rcvif = ifp;
789 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
790 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
791 		    &dummy1, &dummy2) < 0) {
792 			/* m was already freed at this point */
793 			error = EINVAL;/* better error? */
794 			goto done;
795 		}
796 		m->m_flags &= ~M_LOOP; /* XXX */
797 		m->m_pkthdr.rcvif = NULL;
798 	}
799 
800 #ifdef PFIL_HOOKS
801 	/*
802 	 * Run through list of hooks for output packets.
803 	 */
804 	if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0)
805 		goto done;
806 	if (m == NULL)
807 		goto done;
808 	ip6 = mtod(m, struct ip6_hdr *);
809 #endif /* PFIL_HOOKS */
810 	/*
811 	 * Send the packet to the outgoing interface.
812 	 * If necessary, do IPv6 fragmentation before sending.
813 	 */
814 	tlen = m->m_pkthdr.len;
815 	if (tlen <= mtu) {
816 #ifdef IFA_STATS
817 		struct in6_ifaddr *ia6;
818 		ip6 = mtod(m, struct ip6_hdr *);
819 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
820 		if (ia6) {
821 			/* Record statistics for this interface address. */
822 			ia6->ia_ifa.ifa_data.ifad_outbytes +=
823 				m->m_pkthdr.len;
824 		}
825 #endif
826 #ifdef IPSEC
827 		/* clean ipsec history once it goes out of the node */
828 		ipsec_delaux(m);
829 #endif
830 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
831 		goto done;
832 	} else if (mtu < IPV6_MMTU) {
833 		/*
834 		 * note that path MTU is never less than IPV6_MMTU
835 		 * (see icmp6_input).
836 		 */
837 		error = EMSGSIZE;
838 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
839 		goto bad;
840 	} else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
841 		error = EMSGSIZE;
842 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
843 		goto bad;
844 	} else {
845 		struct mbuf **mnext, *m_frgpart;
846 		struct ip6_frag *ip6f;
847 		u_int32_t id = htonl(ip6_id++);
848 		u_char nextproto;
849 
850 		/*
851 		 * Too large for the destination or interface;
852 		 * fragment if possible.
853 		 * Must be able to put at least 8 bytes per fragment.
854 		 */
855 		hlen = unfragpartlen;
856 		if (mtu > IPV6_MAXPACKET)
857 			mtu = IPV6_MAXPACKET;
858 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
859 		if (len < 8) {
860 			error = EMSGSIZE;
861 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
862 			goto bad;
863 		}
864 
865 		mnext = &m->m_nextpkt;
866 
867 		/*
868 		 * Change the next header field of the last header in the
869 		 * unfragmentable part.
870 		 */
871 		if (exthdrs.ip6e_rthdr) {
872 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
873 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
874 		} else if (exthdrs.ip6e_dest1) {
875 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
876 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
877 		} else if (exthdrs.ip6e_hbh) {
878 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
879 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
880 		} else {
881 			nextproto = ip6->ip6_nxt;
882 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
883 		}
884 
885 		/*
886 		 * Loop through length of segment after first fragment,
887 		 * make new header and copy data of each part and link onto
888 		 * chain.
889 		 */
890 		m0 = m;
891 		for (off = hlen; off < tlen; off += len) {
892 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
893 			if (!m) {
894 				error = ENOBUFS;
895 				ip6stat.ip6s_odropped++;
896 				goto sendorfree;
897 			}
898 			m->m_flags = m0->m_flags & M_COPYFLAGS;
899 			*mnext = m;
900 			mnext = &m->m_nextpkt;
901 			m->m_data += max_linkhdr;
902 			mhip6 = mtod(m, struct ip6_hdr *);
903 			*mhip6 = *ip6;
904 			m->m_len = sizeof(*mhip6);
905 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
906 			if (error) {
907 				ip6stat.ip6s_odropped++;
908 				goto sendorfree;
909 			}
910 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
911 			if (off + len >= tlen)
912 				len = tlen - off;
913 			else
914 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
915 			mhip6->ip6_plen = htons((u_short)(len + hlen +
916 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
917 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
918 				error = ENOBUFS;
919 				ip6stat.ip6s_odropped++;
920 				goto sendorfree;
921 			}
922 			m_cat(m, m_frgpart);
923 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
924 			m->m_pkthdr.rcvif = (struct ifnet *)0;
925 			ip6f->ip6f_reserved = 0;
926 			ip6f->ip6f_ident = id;
927 			ip6f->ip6f_nxt = nextproto;
928 			ip6stat.ip6s_ofragments++;
929 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
930 		}
931 
932 		in6_ifstat_inc(ifp, ifs6_out_fragok);
933 	}
934 
935 	/*
936 	 * Remove leading garbages.
937 	 */
938 sendorfree:
939 	m = m0->m_nextpkt;
940 	m0->m_nextpkt = 0;
941 	m_freem(m0);
942 	for (m0 = m; m; m = m0) {
943 		m0 = m->m_nextpkt;
944 		m->m_nextpkt = 0;
945 		if (error == 0) {
946 #ifdef IFA_STATS
947 			struct in6_ifaddr *ia6;
948 			ip6 = mtod(m, struct ip6_hdr *);
949 			ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
950 			if (ia6) {
951 				/*
952 				 * Record statistics for this interface
953 				 * address.
954 				 */
955 				ia6->ia_ifa.ifa_data.ifad_outbytes +=
956 					m->m_pkthdr.len;
957 			}
958 #endif
959 #ifdef IPSEC
960 			/* clean ipsec history once it goes out of the node */
961 			ipsec_delaux(m);
962 #endif
963 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
964 		} else
965 			m_freem(m);
966 	}
967 
968 	if (error == 0)
969 		ip6stat.ip6s_fragmented++;
970 
971 done:
972 	if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
973 		RTFREE(ro->ro_rt);
974 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
975 		RTFREE(ro_pmtu->ro_rt);
976 	}
977 
978 #ifdef IPSEC
979 	if (sp != NULL)
980 		key_freesp(sp);
981 #endif /* IPSEC */
982 
983 	return (error);
984 
985 freehdrs:
986 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
987 	m_freem(exthdrs.ip6e_dest1);
988 	m_freem(exthdrs.ip6e_rthdr);
989 	m_freem(exthdrs.ip6e_dest2);
990 	/* FALLTHROUGH */
991 bad:
992 	m_freem(m);
993 	goto done;
994 }
995 
996 static int
997 ip6_copyexthdr(mp, hdr, hlen)
998 	struct mbuf **mp;
999 	caddr_t hdr;
1000 	int hlen;
1001 {
1002 	struct mbuf *m;
1003 
1004 	if (hlen > MCLBYTES)
1005 		return (ENOBUFS); /* XXX */
1006 
1007 	MGET(m, M_DONTWAIT, MT_DATA);
1008 	if (!m)
1009 		return (ENOBUFS);
1010 
1011 	if (hlen > MLEN) {
1012 		MCLGET(m, M_DONTWAIT);
1013 		if ((m->m_flags & M_EXT) == 0) {
1014 			m_free(m);
1015 			return (ENOBUFS);
1016 		}
1017 	}
1018 	m->m_len = hlen;
1019 	if (hdr)
1020 		bcopy(hdr, mtod(m, caddr_t), hlen);
1021 
1022 	*mp = m;
1023 	return (0);
1024 }
1025 
1026 /*
1027  * Insert jumbo payload option.
1028  */
1029 static int
1030 ip6_insert_jumboopt(exthdrs, plen)
1031 	struct ip6_exthdrs *exthdrs;
1032 	u_int32_t plen;
1033 {
1034 	struct mbuf *mopt;
1035 	u_int8_t *optbuf;
1036 	u_int32_t v;
1037 
1038 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1039 
1040 	/*
1041 	 * If there is no hop-by-hop options header, allocate new one.
1042 	 * If there is one but it doesn't have enough space to store the
1043 	 * jumbo payload option, allocate a cluster to store the whole options.
1044 	 * Otherwise, use it to store the options.
1045 	 */
1046 	if (exthdrs->ip6e_hbh == 0) {
1047 		MGET(mopt, M_DONTWAIT, MT_DATA);
1048 		if (mopt == 0)
1049 			return (ENOBUFS);
1050 		mopt->m_len = JUMBOOPTLEN;
1051 		optbuf = mtod(mopt, u_int8_t *);
1052 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1053 		exthdrs->ip6e_hbh = mopt;
1054 	} else {
1055 		struct ip6_hbh *hbh;
1056 
1057 		mopt = exthdrs->ip6e_hbh;
1058 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1059 			/*
1060 			 * XXX assumption:
1061 			 * - exthdrs->ip6e_hbh is not referenced from places
1062 			 *   other than exthdrs.
1063 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1064 			 */
1065 			int oldoptlen = mopt->m_len;
1066 			struct mbuf *n;
1067 
1068 			/*
1069 			 * XXX: give up if the whole (new) hbh header does
1070 			 * not fit even in an mbuf cluster.
1071 			 */
1072 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1073 				return (ENOBUFS);
1074 
1075 			/*
1076 			 * As a consequence, we must always prepare a cluster
1077 			 * at this point.
1078 			 */
1079 			MGET(n, M_DONTWAIT, MT_DATA);
1080 			if (n) {
1081 				MCLGET(n, M_DONTWAIT);
1082 				if ((n->m_flags & M_EXT) == 0) {
1083 					m_freem(n);
1084 					n = NULL;
1085 				}
1086 			}
1087 			if (!n)
1088 				return (ENOBUFS);
1089 			n->m_len = oldoptlen + JUMBOOPTLEN;
1090 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1091 			      oldoptlen);
1092 			optbuf = mtod(n, u_int8_t *) + oldoptlen;
1093 			m_freem(mopt);
1094 			mopt = exthdrs->ip6e_hbh = n;
1095 		} else {
1096 			optbuf = mtod(mopt, u_int8_t *) + mopt->m_len;
1097 			mopt->m_len += JUMBOOPTLEN;
1098 		}
1099 		optbuf[0] = IP6OPT_PADN;
1100 		optbuf[1] = 0;
1101 
1102 		/*
1103 		 * Adjust the header length according to the pad and
1104 		 * the jumbo payload option.
1105 		 */
1106 		hbh = mtod(mopt, struct ip6_hbh *);
1107 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1108 	}
1109 
1110 	/* fill in the option. */
1111 	optbuf[2] = IP6OPT_JUMBO;
1112 	optbuf[3] = 4;
1113 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1114 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1115 
1116 	/* finally, adjust the packet header length */
1117 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1118 
1119 	return (0);
1120 #undef JUMBOOPTLEN
1121 }
1122 
1123 /*
1124  * Insert fragment header and copy unfragmentable header portions.
1125  */
1126 static int
1127 ip6_insertfraghdr(m0, m, hlen, frghdrp)
1128 	struct mbuf *m0, *m;
1129 	int hlen;
1130 	struct ip6_frag **frghdrp;
1131 {
1132 	struct mbuf *n, *mlast;
1133 
1134 	if (hlen > sizeof(struct ip6_hdr)) {
1135 		n = m_copym(m0, sizeof(struct ip6_hdr),
1136 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1137 		if (n == 0)
1138 			return (ENOBUFS);
1139 		m->m_next = n;
1140 	} else
1141 		n = m;
1142 
1143 	/* Search for the last mbuf of unfragmentable part. */
1144 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1145 		;
1146 
1147 	if ((mlast->m_flags & M_EXT) == 0 &&
1148 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1149 		/* use the trailing space of the last mbuf for the fragment hdr */
1150 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1151 		    mlast->m_len);
1152 		mlast->m_len += sizeof(struct ip6_frag);
1153 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1154 	} else {
1155 		/* allocate a new mbuf for the fragment header */
1156 		struct mbuf *mfrg;
1157 
1158 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1159 		if (mfrg == 0)
1160 			return (ENOBUFS);
1161 		mfrg->m_len = sizeof(struct ip6_frag);
1162 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1163 		mlast->m_next = mfrg;
1164 	}
1165 
1166 	return (0);
1167 }
1168 
1169 static int
1170 ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup)
1171 	struct route_in6 *ro_pmtu, *ro;
1172 	struct ifnet *ifp;
1173 	struct in6_addr *dst;
1174 	u_long *mtup;
1175 {
1176 	u_int32_t mtu = 0;
1177 	int error = 0;
1178 
1179 	if (ro_pmtu != ro) {
1180 		/* The first hop and the final destination may differ. */
1181 		struct sockaddr_in6 *sa6_dst =
1182 		    (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1183 		if (ro_pmtu->ro_rt &&
1184 		    ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1185 		     !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1186 			RTFREE(ro_pmtu->ro_rt);
1187 			ro_pmtu->ro_rt = (struct rtentry *)0;
1188 		}
1189 		if (ro_pmtu->ro_rt == 0) {
1190 			bzero(sa6_dst, sizeof(*sa6_dst));
1191 			sa6_dst->sin6_family = AF_INET6;
1192 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1193 			sa6_dst->sin6_addr = *dst;
1194 
1195 			rtalloc((struct route *)ro_pmtu);
1196 		}
1197 	}
1198 	if (ro_pmtu->ro_rt) {
1199 		u_int32_t ifmtu;
1200 
1201 		if (ifp == NULL)
1202 			ifp = ro_pmtu->ro_rt->rt_ifp;
1203 		ifmtu = IN6_LINKMTU(ifp);
1204 		mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1205 		if (mtu == 0)
1206 			mtu = ifmtu;
1207 		else if (mtu > ifmtu) {
1208 			/*
1209 			 * The MTU on the route is larger than the MTU on
1210 			 * the interface!  This shouldn't happen, unless the
1211 			 * MTU of the interface has been changed after the
1212 			 * interface was brought up.  Change the MTU in the
1213 			 * route to match the interface MTU (as long as the
1214 			 * field isn't locked).
1215 			 *
1216 			 * if MTU on the route is 0, we need to fix the MTU.
1217 			 * this case happens with path MTU discovery timeouts.
1218 			 */
1219 			mtu = ifmtu;
1220 			if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU))
1221 				ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1222 		}
1223 	} else if (ifp) {
1224 		mtu = IN6_LINKMTU(ifp);
1225 	} else
1226 		error = EHOSTUNREACH; /* XXX */
1227 
1228 	*mtup = mtu;
1229 	return (error);
1230 }
1231 
1232 /*
1233  * IP6 socket option processing.
1234  */
1235 int
1236 ip6_ctloutput(op, so, level, optname, mp)
1237 	int op;
1238 	struct socket *so;
1239 	int level, optname;
1240 	struct mbuf **mp;
1241 {
1242 	struct in6pcb *in6p = sotoin6pcb(so);
1243 	struct mbuf *m = *mp;
1244 	int optval = 0;
1245 	int error = 0;
1246 	struct proc *p = curproc;	/* XXX */
1247 
1248 	if (level == IPPROTO_IPV6) {
1249 		switch (op) {
1250 		case PRCO_SETOPT:
1251 			switch (optname) {
1252 			case IPV6_PKTOPTIONS:
1253 				/* m is freed in ip6_pcbopts */
1254 				return (ip6_pcbopts(&in6p->in6p_outputopts,
1255 				    m, so));
1256 			case IPV6_HOPOPTS:
1257 			case IPV6_DSTOPTS:
1258 				if (p == 0 || suser(p->p_ucred, &p->p_acflag)) {
1259 					error = EPERM;
1260 					break;
1261 				}
1262 				/* FALLTHROUGH */
1263 			case IPV6_UNICAST_HOPS:
1264 			case IPV6_RECVOPTS:
1265 			case IPV6_RECVRETOPTS:
1266 			case IPV6_RECVDSTADDR:
1267 			case IPV6_PKTINFO:
1268 			case IPV6_HOPLIMIT:
1269 			case IPV6_RTHDR:
1270 			case IPV6_FAITH:
1271 			case IPV6_V6ONLY:
1272 				if (!m || m->m_len != sizeof(int)) {
1273 					error = EINVAL;
1274 					break;
1275 				}
1276 				optval = *mtod(m, int *);
1277 				switch (optname) {
1278 
1279 				case IPV6_UNICAST_HOPS:
1280 					if (optval < -1 || optval >= 256)
1281 						error = EINVAL;
1282 					else {
1283 						/* -1 = kernel default */
1284 						in6p->in6p_hops = optval;
1285 					}
1286 					break;
1287 #define OPTSET(bit) \
1288 do { \
1289 	if (optval) \
1290 		in6p->in6p_flags |= (bit); \
1291 	else \
1292 		in6p->in6p_flags &= ~(bit); \
1293 } while (/*CONSTCOND*/ 0)
1294 
1295 				case IPV6_RECVOPTS:
1296 					OPTSET(IN6P_RECVOPTS);
1297 					break;
1298 
1299 				case IPV6_RECVRETOPTS:
1300 					OPTSET(IN6P_RECVRETOPTS);
1301 					break;
1302 
1303 				case IPV6_RECVDSTADDR:
1304 					OPTSET(IN6P_RECVDSTADDR);
1305 					break;
1306 
1307 				case IPV6_PKTINFO:
1308 					OPTSET(IN6P_PKTINFO);
1309 					break;
1310 
1311 				case IPV6_HOPLIMIT:
1312 					OPTSET(IN6P_HOPLIMIT);
1313 					break;
1314 
1315 				case IPV6_HOPOPTS:
1316 					OPTSET(IN6P_HOPOPTS);
1317 					break;
1318 
1319 				case IPV6_DSTOPTS:
1320 					OPTSET(IN6P_DSTOPTS);
1321 					break;
1322 
1323 				case IPV6_RTHDR:
1324 					OPTSET(IN6P_RTHDR);
1325 					break;
1326 
1327 				case IPV6_FAITH:
1328 					OPTSET(IN6P_FAITH);
1329 					break;
1330 
1331 				case IPV6_V6ONLY:
1332 					/*
1333 					 * make setsockopt(IPV6_V6ONLY)
1334 					 * available only prior to bind(2).
1335 					 * see ipng mailing list, Jun 22 2001.
1336 					 */
1337 					if (in6p->in6p_lport ||
1338 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1339 						error = EINVAL;
1340 						break;
1341 					}
1342 #ifdef INET6_BINDV6ONLY
1343 					if (!optval)
1344 						error = EINVAL;
1345 #else
1346 					OPTSET(IN6P_IPV6_V6ONLY);
1347 #endif
1348 					break;
1349 				}
1350 				break;
1351 #undef OPTSET
1352 
1353 			case IPV6_MULTICAST_IF:
1354 			case IPV6_MULTICAST_HOPS:
1355 			case IPV6_MULTICAST_LOOP:
1356 			case IPV6_JOIN_GROUP:
1357 			case IPV6_LEAVE_GROUP:
1358 				error =	ip6_setmoptions(optname,
1359 				    &in6p->in6p_moptions, m);
1360 				break;
1361 
1362 			case IPV6_PORTRANGE:
1363 				optval = *mtod(m, int *);
1364 
1365 				switch (optval) {
1366 				case IPV6_PORTRANGE_DEFAULT:
1367 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1368 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1369 					break;
1370 
1371 				case IPV6_PORTRANGE_HIGH:
1372 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1373 					in6p->in6p_flags |= IN6P_HIGHPORT;
1374 					break;
1375 
1376 				case IPV6_PORTRANGE_LOW:
1377 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1378 					in6p->in6p_flags |= IN6P_LOWPORT;
1379 					break;
1380 
1381 				default:
1382 					error = EINVAL;
1383 					break;
1384 				}
1385 				break;
1386 
1387 #ifdef IPSEC
1388 			case IPV6_IPSEC_POLICY:
1389 			    {
1390 				caddr_t req = NULL;
1391 				size_t len = 0;
1392 
1393 				int priv = 0;
1394 				if (p == 0 || suser(p->p_ucred, &p->p_acflag))
1395 					priv = 0;
1396 				else
1397 					priv = 1;
1398 				if (m) {
1399 					req = mtod(m, caddr_t);
1400 					len = m->m_len;
1401 				}
1402 				error = ipsec6_set_policy(in6p,
1403 				                   optname, req, len, priv);
1404 			    }
1405 				break;
1406 #endif /* IPSEC */
1407 
1408 			default:
1409 				error = ENOPROTOOPT;
1410 				break;
1411 			}
1412 			if (m)
1413 				(void)m_free(m);
1414 			break;
1415 
1416 		case PRCO_GETOPT:
1417 			switch (optname) {
1418 
1419 			case IPV6_OPTIONS:
1420 			case IPV6_RETOPTS:
1421 				error = ENOPROTOOPT;
1422 				break;
1423 
1424 			case IPV6_PKTOPTIONS:
1425 				if (in6p->in6p_options) {
1426 					*mp = m_copym(in6p->in6p_options, 0,
1427 					    M_COPYALL, M_WAIT);
1428 				} else {
1429 					*mp = m_get(M_WAIT, MT_SOOPTS);
1430 					(*mp)->m_len = 0;
1431 				}
1432 				break;
1433 
1434 			case IPV6_HOPOPTS:
1435 			case IPV6_DSTOPTS:
1436 				if (p == 0 || suser(p->p_ucred, &p->p_acflag)) {
1437 					error = EPERM;
1438 					break;
1439 				}
1440 				/* FALLTHROUGH */
1441 			case IPV6_UNICAST_HOPS:
1442 			case IPV6_RECVOPTS:
1443 			case IPV6_RECVRETOPTS:
1444 			case IPV6_RECVDSTADDR:
1445 			case IPV6_PORTRANGE:
1446 			case IPV6_PKTINFO:
1447 			case IPV6_HOPLIMIT:
1448 			case IPV6_RTHDR:
1449 			case IPV6_FAITH:
1450 			case IPV6_V6ONLY:
1451 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1452 				m->m_len = sizeof(int);
1453 				switch (optname) {
1454 
1455 				case IPV6_UNICAST_HOPS:
1456 					optval = in6p->in6p_hops;
1457 					break;
1458 
1459 #define OPTBIT(bit) (in6p->in6p_flags & bit ? 1 : 0)
1460 
1461 				case IPV6_RECVOPTS:
1462 					optval = OPTBIT(IN6P_RECVOPTS);
1463 					break;
1464 
1465 				case IPV6_RECVRETOPTS:
1466 					optval = OPTBIT(IN6P_RECVRETOPTS);
1467 					break;
1468 
1469 				case IPV6_RECVDSTADDR:
1470 					optval = OPTBIT(IN6P_RECVDSTADDR);
1471 					break;
1472 
1473 				case IPV6_PORTRANGE:
1474 				    {
1475 					int flags;
1476 					flags = in6p->in6p_flags;
1477 					if (flags & IN6P_HIGHPORT)
1478 						optval = IPV6_PORTRANGE_HIGH;
1479 					else if (flags & IN6P_LOWPORT)
1480 						optval = IPV6_PORTRANGE_LOW;
1481 					else
1482 						optval = 0;
1483 					break;
1484 				    }
1485 
1486 				case IPV6_PKTINFO:
1487 					optval = OPTBIT(IN6P_PKTINFO);
1488 					break;
1489 
1490 				case IPV6_HOPLIMIT:
1491 					optval = OPTBIT(IN6P_HOPLIMIT);
1492 					break;
1493 
1494 				case IPV6_HOPOPTS:
1495 					optval = OPTBIT(IN6P_HOPOPTS);
1496 					break;
1497 
1498 				case IPV6_DSTOPTS:
1499 					optval = OPTBIT(IN6P_DSTOPTS);
1500 					break;
1501 
1502 				case IPV6_RTHDR:
1503 					optval = OPTBIT(IN6P_RTHDR);
1504 					break;
1505 
1506 				case IPV6_FAITH:
1507 					optval = OPTBIT(IN6P_FAITH);
1508 					break;
1509 
1510 				case IPV6_V6ONLY:
1511 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
1512 					break;
1513 				}
1514 				*mtod(m, int *) = optval;
1515 				break;
1516 
1517 			case IPV6_MULTICAST_IF:
1518 			case IPV6_MULTICAST_HOPS:
1519 			case IPV6_MULTICAST_LOOP:
1520 			case IPV6_JOIN_GROUP:
1521 			case IPV6_LEAVE_GROUP:
1522 				error = ip6_getmoptions(optname, in6p->in6p_moptions, mp);
1523 				break;
1524 
1525 #ifdef IPSEC
1526 			case IPV6_IPSEC_POLICY:
1527 			{
1528 				caddr_t req = NULL;
1529 				size_t len = 0;
1530 
1531 				if (m) {
1532 					req = mtod(m, caddr_t);
1533 					len = m->m_len;
1534 				}
1535 				error = ipsec6_get_policy(in6p, req, len, mp);
1536 				break;
1537 			}
1538 #endif /* IPSEC */
1539 
1540 			default:
1541 				error = ENOPROTOOPT;
1542 				break;
1543 			}
1544 			break;
1545 		}
1546 	} else {
1547 		error = EINVAL;
1548 		if (op == PRCO_SETOPT && *mp)
1549 			(void)m_free(*mp);
1550 	}
1551 	return (error);
1552 }
1553 
1554 int
1555 ip6_raw_ctloutput(op, so, level, optname, mp)
1556 	int op;
1557 	struct socket *so;
1558 	int level, optname;
1559 	struct mbuf **mp;
1560 {
1561 	int error = 0, optval, optlen;
1562 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
1563 	struct in6pcb *in6p = sotoin6pcb(so);
1564 	struct mbuf *m = *mp;
1565 
1566 	optlen = m ? m->m_len : 0;
1567 
1568 	if (level != IPPROTO_IPV6) {
1569 		if (op == PRCO_SETOPT && *mp)
1570 			(void)m_free(*mp);
1571 		return (EINVAL);
1572 	}
1573 
1574 	switch (optname) {
1575 	case IPV6_CHECKSUM:
1576 		/*
1577 		 * For ICMPv6 sockets, no modification allowed for checksum
1578 		 * offset, permit "no change" values to help existing apps.
1579 		 *
1580 		 * XXX 2292bis says: "An attempt to set IPV6_CHECKSUM
1581 		 * for an ICMPv6 socket will fail."
1582 		 * The current behavior does not meet 2292bis.
1583 		 */
1584 		switch (op) {
1585 		case PRCO_SETOPT:
1586 			if (optlen != sizeof(int)) {
1587 				error = EINVAL;
1588 				break;
1589 			}
1590 			optval = *mtod(m, int *);
1591 			if ((optval % 2) != 0) {
1592 				/* the API assumes even offset values */
1593 				error = EINVAL;
1594 			} else if (so->so_proto->pr_protocol ==
1595 			    IPPROTO_ICMPV6) {
1596 				if (optval != icmp6off)
1597 					error = EINVAL;
1598 			} else
1599 				in6p->in6p_cksum = optval;
1600 			break;
1601 
1602 		case PRCO_GETOPT:
1603 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
1604 				optval = icmp6off;
1605 			else
1606 				optval = in6p->in6p_cksum;
1607 
1608 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1609 			m->m_len = sizeof(int);
1610 			*mtod(m, int *) = optval;
1611 			break;
1612 
1613 		default:
1614 			error = EINVAL;
1615 			break;
1616 		}
1617 		break;
1618 
1619 	default:
1620 		error = ENOPROTOOPT;
1621 		break;
1622 	}
1623 
1624 	if (op == PRCO_SETOPT && m)
1625 		(void)m_free(m);
1626 
1627 	return (error);
1628 }
1629 
1630 /*
1631  * Set up IP6 options in pcb for insertion in output packets.
1632  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1633  * with destination address if source routed.
1634  */
1635 static int
1636 ip6_pcbopts(pktopt, m, so)
1637 	struct ip6_pktopts **pktopt;
1638 	struct mbuf *m;
1639 	struct socket *so;
1640 {
1641 	struct ip6_pktopts *opt = *pktopt;
1642 	int error = 0;
1643 	struct proc *p = curproc;	/* XXX */
1644 	int priv = 0;
1645 
1646 	/* turn off any old options. */
1647 	if (opt) {
1648 		if (opt->ip6po_m)
1649 			(void)m_free(opt->ip6po_m);
1650 	} else
1651 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
1652 	*pktopt = 0;
1653 
1654 	if (!m || m->m_len == 0) {
1655 		/*
1656 		 * Only turning off any previous options.
1657 		 */
1658 		free(opt, M_IP6OPT);
1659 		if (m)
1660 			(void)m_free(m);
1661 		return (0);
1662 	}
1663 
1664 	/*  set options specified by user. */
1665 	if (p && !suser(p->p_ucred, &p->p_acflag))
1666 		priv = 1;
1667 	if ((error = ip6_setpktoptions(m, opt, priv)) != 0) {
1668 		(void)m_free(m);
1669 		free(opt, M_IP6OPT);
1670 		return (error);
1671 	}
1672 	*pktopt = opt;
1673 	return (0);
1674 }
1675 
1676 /*
1677  * Set the IP6 multicast options in response to user setsockopt().
1678  */
1679 static int
1680 ip6_setmoptions(optname, im6op, m)
1681 	int optname;
1682 	struct ip6_moptions **im6op;
1683 	struct mbuf *m;
1684 {
1685 	int error = 0;
1686 	u_int loop, ifindex;
1687 	struct ipv6_mreq *mreq;
1688 	struct ifnet *ifp;
1689 	struct ip6_moptions *im6o = *im6op;
1690 	struct route_in6 ro;
1691 	struct sockaddr_in6 *dst;
1692 	struct in6_multi_mship *imm;
1693 	struct proc *p = curproc;	/* XXX */
1694 
1695 	if (im6o == NULL) {
1696 		/*
1697 		 * No multicast option buffer attached to the pcb;
1698 		 * allocate one and initialize to default values.
1699 		 */
1700 		im6o = (struct ip6_moptions *)
1701 			malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
1702 
1703 		if (im6o == NULL)
1704 			return (ENOBUFS);
1705 		*im6op = im6o;
1706 		im6o->im6o_multicast_ifp = NULL;
1707 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1708 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
1709 		LIST_INIT(&im6o->im6o_memberships);
1710 	}
1711 
1712 	switch (optname) {
1713 
1714 	case IPV6_MULTICAST_IF:
1715 		/*
1716 		 * Select the interface for outgoing multicast packets.
1717 		 */
1718 		if (m == NULL || m->m_len != sizeof(u_int)) {
1719 			error = EINVAL;
1720 			break;
1721 		}
1722 		bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
1723 		if (ifindex < 0 || if_index < ifindex) {
1724 			error = ENXIO;	/* XXX EINVAL? */
1725 			break;
1726 		}
1727 		ifp = ifindex2ifnet[ifindex];
1728 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1729 			error = EADDRNOTAVAIL;
1730 			break;
1731 		}
1732 		im6o->im6o_multicast_ifp = ifp;
1733 		break;
1734 
1735 	case IPV6_MULTICAST_HOPS:
1736 	    {
1737 		/*
1738 		 * Set the IP6 hoplimit for outgoing multicast packets.
1739 		 */
1740 		int optval;
1741 		if (m == NULL || m->m_len != sizeof(int)) {
1742 			error = EINVAL;
1743 			break;
1744 		}
1745 		bcopy(mtod(m, u_int *), &optval, sizeof(optval));
1746 		if (optval < -1 || optval >= 256)
1747 			error = EINVAL;
1748 		else if (optval == -1)
1749 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1750 		else
1751 			im6o->im6o_multicast_hlim = optval;
1752 		break;
1753 	    }
1754 
1755 	case IPV6_MULTICAST_LOOP:
1756 		/*
1757 		 * Set the loopback flag for outgoing multicast packets.
1758 		 * Must be zero or one.
1759 		 */
1760 		if (m == NULL || m->m_len != sizeof(u_int)) {
1761 			error = EINVAL;
1762 			break;
1763 		}
1764 		bcopy(mtod(m, u_int *), &loop, sizeof(loop));
1765 		if (loop > 1) {
1766 			error = EINVAL;
1767 			break;
1768 		}
1769 		im6o->im6o_multicast_loop = loop;
1770 		break;
1771 
1772 	case IPV6_JOIN_GROUP:
1773 		/*
1774 		 * Add a multicast group membership.
1775 		 * Group must be a valid IP6 multicast address.
1776 		 */
1777 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
1778 			error = EINVAL;
1779 			break;
1780 		}
1781 		mreq = mtod(m, struct ipv6_mreq *);
1782 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
1783 			/*
1784 			 * We use the unspecified address to specify to accept
1785 			 * all multicast addresses. Only super user is allowed
1786 			 * to do this.
1787 			 */
1788 			if (suser(p->p_ucred, &p->p_acflag))
1789 			{
1790 				error = EACCES;
1791 				break;
1792 			}
1793 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
1794 			error = EINVAL;
1795 			break;
1796 		}
1797 
1798 		/*
1799 		 * If the interface is specified, validate it.
1800 		 */
1801 		if (mreq->ipv6mr_interface < 0
1802 		 || if_index < mreq->ipv6mr_interface) {
1803 			error = ENXIO;	/* XXX EINVAL? */
1804 			break;
1805 		}
1806 		/*
1807 		 * If no interface was explicitly specified, choose an
1808 		 * appropriate one according to the given multicast address.
1809 		 */
1810 		if (mreq->ipv6mr_interface == 0) {
1811 			/*
1812 			 * If the multicast address is in node-local scope,
1813 			 * the interface should be a loopback interface.
1814 			 * Otherwise, look up the routing table for the
1815 			 * address, and choose the outgoing interface.
1816 			 *   XXX: is it a good approach?
1817 			 */
1818 			if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
1819 				ifp = &loif[0];
1820 			} else {
1821 				ro.ro_rt = NULL;
1822 				dst = (struct sockaddr_in6 *)&ro.ro_dst;
1823 				bzero(dst, sizeof(*dst));
1824 				dst->sin6_len = sizeof(struct sockaddr_in6);
1825 				dst->sin6_family = AF_INET6;
1826 				dst->sin6_addr = mreq->ipv6mr_multiaddr;
1827 				rtalloc((struct route *)&ro);
1828 				if (ro.ro_rt == NULL) {
1829 					error = EADDRNOTAVAIL;
1830 					break;
1831 				}
1832 				ifp = ro.ro_rt->rt_ifp;
1833 				rtfree(ro.ro_rt);
1834 			}
1835 		} else
1836 			ifp = ifindex2ifnet[mreq->ipv6mr_interface];
1837 
1838 		/*
1839 		 * See if we found an interface, and confirm that it
1840 		 * supports multicast
1841 		 */
1842 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1843 			error = EADDRNOTAVAIL;
1844 			break;
1845 		}
1846 		/*
1847 		 * Put interface index into the multicast address,
1848 		 * if the address has link-local scope.
1849 		 */
1850 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
1851 			mreq->ipv6mr_multiaddr.s6_addr16[1] =
1852 			    htons(mreq->ipv6mr_interface);
1853 		}
1854 		/*
1855 		 * See if the membership already exists.
1856 		 */
1857 		for (imm = im6o->im6o_memberships.lh_first;
1858 		     imm != NULL; imm = imm->i6mm_chain.le_next)
1859 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
1860 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
1861 			    &mreq->ipv6mr_multiaddr))
1862 				break;
1863 		if (imm != NULL) {
1864 			error = EADDRINUSE;
1865 			break;
1866 		}
1867 		/*
1868 		 * Everything looks good; add a new record to the multicast
1869 		 * address list for the given interface.
1870 		 */
1871 		imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error);
1872 		if (!imm)
1873 			break;
1874 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
1875 		break;
1876 
1877 	case IPV6_LEAVE_GROUP:
1878 		/*
1879 		 * Drop a multicast group membership.
1880 		 * Group must be a valid IP6 multicast address.
1881 		 */
1882 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
1883 			error = EINVAL;
1884 			break;
1885 		}
1886 		mreq = mtod(m, struct ipv6_mreq *);
1887 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
1888 			if (suser(p->p_ucred, &p->p_acflag))
1889 			{
1890 				error = EACCES;
1891 				break;
1892 			}
1893 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
1894 			error = EINVAL;
1895 			break;
1896 		}
1897 		/*
1898 		 * If an interface address was specified, get a pointer
1899 		 * to its ifnet structure.
1900 		 */
1901 		if (mreq->ipv6mr_interface < 0
1902 		 || if_index < mreq->ipv6mr_interface) {
1903 			error = ENXIO;	/* XXX EINVAL? */
1904 			break;
1905 		}
1906 		ifp = ifindex2ifnet[mreq->ipv6mr_interface];
1907 		/*
1908 		 * Put interface index into the multicast address,
1909 		 * if the address has link-local scope.
1910 		 */
1911 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
1912 			mreq->ipv6mr_multiaddr.s6_addr16[1] =
1913 			    htons(mreq->ipv6mr_interface);
1914 		}
1915 		/*
1916 		 * Find the membership in the membership list.
1917 		 */
1918 		for (imm = im6o->im6o_memberships.lh_first;
1919 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
1920 			if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
1921 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
1922 			    &mreq->ipv6mr_multiaddr))
1923 				break;
1924 		}
1925 		if (imm == NULL) {
1926 			/* Unable to resolve interface */
1927 			error = EADDRNOTAVAIL;
1928 			break;
1929 		}
1930 		/*
1931 		 * Give up the multicast address record to which the
1932 		 * membership points.
1933 		 */
1934 		LIST_REMOVE(imm, i6mm_chain);
1935 		in6_leavegroup(imm);
1936 		break;
1937 
1938 	default:
1939 		error = EOPNOTSUPP;
1940 		break;
1941 	}
1942 
1943 	/*
1944 	 * If all options have default values, no need to keep the mbuf.
1945 	 */
1946 	if (im6o->im6o_multicast_ifp == NULL &&
1947 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
1948 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
1949 	    im6o->im6o_memberships.lh_first == NULL) {
1950 		free(*im6op, M_IPMOPTS);
1951 		*im6op = NULL;
1952 	}
1953 
1954 	return (error);
1955 }
1956 
1957 /*
1958  * Return the IP6 multicast options in response to user getsockopt().
1959  */
1960 static int
1961 ip6_getmoptions(optname, im6o, mp)
1962 	int optname;
1963 	struct ip6_moptions *im6o;
1964 	struct mbuf **mp;
1965 {
1966 	u_int *hlim, *loop, *ifindex;
1967 
1968 	*mp = m_get(M_WAIT, MT_SOOPTS);
1969 
1970 	switch (optname) {
1971 
1972 	case IPV6_MULTICAST_IF:
1973 		ifindex = mtod(*mp, u_int *);
1974 		(*mp)->m_len = sizeof(u_int);
1975 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
1976 			*ifindex = 0;
1977 		else
1978 			*ifindex = im6o->im6o_multicast_ifp->if_index;
1979 		return (0);
1980 
1981 	case IPV6_MULTICAST_HOPS:
1982 		hlim = mtod(*mp, u_int *);
1983 		(*mp)->m_len = sizeof(u_int);
1984 		if (im6o == NULL)
1985 			*hlim = ip6_defmcasthlim;
1986 		else
1987 			*hlim = im6o->im6o_multicast_hlim;
1988 		return (0);
1989 
1990 	case IPV6_MULTICAST_LOOP:
1991 		loop = mtod(*mp, u_int *);
1992 		(*mp)->m_len = sizeof(u_int);
1993 		if (im6o == NULL)
1994 			*loop = ip6_defmcasthlim;
1995 		else
1996 			*loop = im6o->im6o_multicast_loop;
1997 		return (0);
1998 
1999 	default:
2000 		return (EOPNOTSUPP);
2001 	}
2002 }
2003 
2004 /*
2005  * Discard the IP6 multicast options.
2006  */
2007 void
2008 ip6_freemoptions(im6o)
2009 	struct ip6_moptions *im6o;
2010 {
2011 	struct in6_multi_mship *imm;
2012 
2013 	if (im6o == NULL)
2014 		return;
2015 
2016 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2017 		LIST_REMOVE(imm, i6mm_chain);
2018 		in6_leavegroup(imm);
2019 	}
2020 	free(im6o, M_IPMOPTS);
2021 }
2022 
2023 /*
2024  * Set IPv6 outgoing packet options based on advanced API.
2025  */
2026 int
2027 ip6_setpktoptions(control, opt, priv)
2028 	struct mbuf *control;
2029 	struct ip6_pktopts *opt;
2030 	int priv;
2031 {
2032 	struct cmsghdr *cm = 0;
2033 
2034 	if (control == 0 || opt == 0)
2035 		return (EINVAL);
2036 
2037 	bzero(opt, sizeof(*opt));
2038 	opt->ip6po_hlim = -1; /* -1 means to use default hop limit */
2039 
2040 	/*
2041 	 * XXX: Currently, we assume all the optional information is stored
2042 	 * in a single mbuf.
2043 	 */
2044 	if (control->m_next)
2045 		return (EINVAL);
2046 
2047 	opt->ip6po_m = control;
2048 
2049 	for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2050 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2051 		cm = mtod(control, struct cmsghdr *);
2052 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2053 			return (EINVAL);
2054 		if (cm->cmsg_level != IPPROTO_IPV6)
2055 			continue;
2056 
2057 		switch (cm->cmsg_type) {
2058 		case IPV6_PKTINFO:
2059 			if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo)))
2060 				return (EINVAL);
2061 			opt->ip6po_pktinfo = (struct in6_pktinfo *)CMSG_DATA(cm);
2062 			if (opt->ip6po_pktinfo->ipi6_ifindex &&
2063 			    IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr))
2064 				opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] =
2065 					htons(opt->ip6po_pktinfo->ipi6_ifindex);
2066 
2067 			if (opt->ip6po_pktinfo->ipi6_ifindex > if_index ||
2068 			    opt->ip6po_pktinfo->ipi6_ifindex < 0) {
2069 				return (ENXIO);
2070 			}
2071 
2072 			/*
2073 			 * Check if the requested source address is indeed a
2074 			 * unicast address assigned to the node, and can be
2075 			 * used as the packet's source address.
2076 			 */
2077 			if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
2078 				struct ifaddr *ia;
2079 				struct in6_ifaddr *ia6;
2080 				struct sockaddr_in6 sin6;
2081 
2082 				bzero(&sin6, sizeof(sin6));
2083 				sin6.sin6_len = sizeof(sin6);
2084 				sin6.sin6_family = AF_INET6;
2085 				sin6.sin6_addr =
2086 					opt->ip6po_pktinfo->ipi6_addr;
2087 				ia = ifa_ifwithaddr(sin6tosa(&sin6));
2088 				if (ia == NULL ||
2089 				    (opt->ip6po_pktinfo->ipi6_ifindex &&
2090 				     (ia->ifa_ifp->if_index !=
2091 				      opt->ip6po_pktinfo->ipi6_ifindex))) {
2092 					return (EADDRNOTAVAIL);
2093 				}
2094 				ia6 = (struct in6_ifaddr *)ia;
2095 				if ((ia6->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) != 0) {
2096 					return (EADDRNOTAVAIL);
2097 				}
2098 
2099 				/*
2100 				 * Check if the requested source address is
2101 				 * indeed a unicast address assigned to the
2102 				 * node.
2103 				 */
2104 				if (IN6_IS_ADDR_MULTICAST(&opt->ip6po_pktinfo->ipi6_addr))
2105 					return (EADDRNOTAVAIL);
2106 			}
2107 			break;
2108 
2109 		case IPV6_HOPLIMIT:
2110 			if (cm->cmsg_len != CMSG_LEN(sizeof(int)))
2111 				return (EINVAL);
2112 
2113 			bcopy(CMSG_DATA(cm), &opt->ip6po_hlim,
2114 			    sizeof(opt->ip6po_hlim));
2115 			if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255)
2116 				return (EINVAL);
2117 			break;
2118 
2119 		case IPV6_NEXTHOP:
2120 			if (!priv)
2121 				return (EPERM);
2122 
2123 			/* check if cmsg_len is large enough for sa_len */
2124 			if (cm->cmsg_len < sizeof(u_char) ||
2125 			    cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm)))
2126 				return (EINVAL);
2127 
2128 			opt->ip6po_nexthop = (struct sockaddr *)CMSG_DATA(cm);
2129 
2130 			break;
2131 
2132 		case IPV6_HOPOPTS:
2133 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh)))
2134 				return (EINVAL);
2135 			opt->ip6po_hbh = (struct ip6_hbh *)CMSG_DATA(cm);
2136 			if (cm->cmsg_len !=
2137 			    CMSG_LEN((opt->ip6po_hbh->ip6h_len + 1) << 3))
2138 				return (EINVAL);
2139 			break;
2140 
2141 		case IPV6_DSTOPTS:
2142 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest)))
2143 				return (EINVAL);
2144 
2145 			/*
2146 			 * If there is no routing header yet, the destination
2147 			 * options header should be put on the 1st part.
2148 			 * Otherwise, the header should be on the 2nd part.
2149 			 * (See RFC 2460, section 4.1)
2150 			 */
2151 			if (opt->ip6po_rthdr == NULL) {
2152 				opt->ip6po_dest1 =
2153 				    (struct ip6_dest *)CMSG_DATA(cm);
2154 				if (cm->cmsg_len !=
2155 				    CMSG_LEN((opt->ip6po_dest1->ip6d_len + 1) << 3));
2156 					return (EINVAL);
2157 			}
2158 			else {
2159 				opt->ip6po_dest2 =
2160 				    (struct ip6_dest *)CMSG_DATA(cm);
2161 				if (cm->cmsg_len !=
2162 				    CMSG_LEN((opt->ip6po_dest2->ip6d_len + 1) << 3))
2163 					return (EINVAL);
2164 			}
2165 			break;
2166 
2167 		case IPV6_RTHDR:
2168 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr)))
2169 				return (EINVAL);
2170 			opt->ip6po_rthdr = (struct ip6_rthdr *)CMSG_DATA(cm);
2171 			if (cm->cmsg_len !=
2172 			    CMSG_LEN((opt->ip6po_rthdr->ip6r_len + 1) << 3))
2173 				return (EINVAL);
2174 			switch (opt->ip6po_rthdr->ip6r_type) {
2175 			case IPV6_RTHDR_TYPE_0:
2176 				if (opt->ip6po_rthdr->ip6r_segleft == 0)
2177 					return (EINVAL);
2178 				break;
2179 			default:
2180 				return (EINVAL);
2181 			}
2182 			break;
2183 
2184 		default:
2185 			return (ENOPROTOOPT);
2186 		}
2187 	}
2188 
2189 	return (0);
2190 }
2191 
2192 /*
2193  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2194  * packet to the input queue of a specified interface.  Note that this
2195  * calls the output routine of the loopback "driver", but with an interface
2196  * pointer that might NOT be &loif -- easier than replicating that code here.
2197  */
2198 void
2199 ip6_mloopback(ifp, m, dst)
2200 	struct ifnet *ifp;
2201 	struct mbuf *m;
2202 	struct sockaddr_in6 *dst;
2203 {
2204 	struct mbuf *copym;
2205 	struct ip6_hdr *ip6;
2206 
2207 	copym = m_copy(m, 0, M_COPYALL);
2208 	if (copym == NULL)
2209 		return;
2210 
2211 	/*
2212 	 * Make sure to deep-copy IPv6 header portion in case the data
2213 	 * is in an mbuf cluster, so that we can safely override the IPv6
2214 	 * header portion later.
2215 	 */
2216 	if ((copym->m_flags & M_EXT) != 0 ||
2217 	    copym->m_len < sizeof(struct ip6_hdr)) {
2218 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
2219 		if (copym == NULL)
2220 			return;
2221 	}
2222 
2223 #ifdef DIAGNOSTIC
2224 	if (copym->m_len < sizeof(*ip6)) {
2225 		m_freem(copym);
2226 		return;
2227 	}
2228 #endif
2229 
2230 	ip6 = mtod(copym, struct ip6_hdr *);
2231 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
2232 		ip6->ip6_src.s6_addr16[1] = 0;
2233 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
2234 		ip6->ip6_dst.s6_addr16[1] = 0;
2235 
2236 	(void)looutput(ifp, copym, (struct sockaddr *)dst, NULL);
2237 }
2238 
2239 /*
2240  * Chop IPv6 header off from the payload.
2241  */
2242 static int
2243 ip6_splithdr(m, exthdrs)
2244 	struct mbuf *m;
2245 	struct ip6_exthdrs *exthdrs;
2246 {
2247 	struct mbuf *mh;
2248 	struct ip6_hdr *ip6;
2249 
2250 	ip6 = mtod(m, struct ip6_hdr *);
2251 	if (m->m_len > sizeof(*ip6)) {
2252 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
2253 		if (mh == 0) {
2254 			m_freem(m);
2255 			return ENOBUFS;
2256 		}
2257 		M_COPY_PKTHDR(mh, m);
2258 		MH_ALIGN(mh, sizeof(*ip6));
2259 		m->m_flags &= ~M_PKTHDR;
2260 		m->m_len -= sizeof(*ip6);
2261 		m->m_data += sizeof(*ip6);
2262 		mh->m_next = m;
2263 		m = mh;
2264 		m->m_len = sizeof(*ip6);
2265 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2266 	}
2267 	exthdrs->ip6e_ip6 = m;
2268 	return 0;
2269 }
2270 
2271 /*
2272  * Compute IPv6 extension header length.
2273  */
2274 int
2275 ip6_optlen(in6p)
2276 	struct in6pcb *in6p;
2277 {
2278 	int len;
2279 
2280 	if (!in6p->in6p_outputopts)
2281 		return 0;
2282 
2283 	len = 0;
2284 #define elen(x) \
2285     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2286 
2287 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
2288 	len += elen(in6p->in6p_outputopts->ip6po_dest1);
2289 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
2290 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
2291 	return len;
2292 #undef elen
2293 }
2294