xref: /netbsd-src/sys/netinet6/ip6_output.c (revision b757af438b42b93f8c6571f026d8b8ef3eaf5fc9)
1 /*	$NetBSD: ip6_output.c,v 1.147 2012/03/22 20:34:41 drochner 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. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
62  */
63 
64 #include <sys/cdefs.h>
65 __KERNEL_RCSID(0, "$NetBSD: ip6_output.c,v 1.147 2012/03/22 20:34:41 drochner Exp $");
66 
67 #include "opt_inet.h"
68 #include "opt_inet6.h"
69 #include "opt_ipsec.h"
70 #include "opt_pfil_hooks.h"
71 
72 #include <sys/param.h>
73 #include <sys/malloc.h>
74 #include <sys/mbuf.h>
75 #include <sys/errno.h>
76 #include <sys/protosw.h>
77 #include <sys/socket.h>
78 #include <sys/socketvar.h>
79 #include <sys/systm.h>
80 #include <sys/proc.h>
81 #include <sys/kauth.h>
82 
83 #include <net/if.h>
84 #include <net/route.h>
85 #ifdef PFIL_HOOKS
86 #include <net/pfil.h>
87 #endif
88 
89 #include <netinet/in.h>
90 #include <netinet/in_var.h>
91 #include <netinet/ip6.h>
92 #include <netinet/icmp6.h>
93 #include <netinet/in_offload.h>
94 #include <netinet6/in6_offload.h>
95 #include <netinet6/ip6_var.h>
96 #include <netinet6/ip6_private.h>
97 #include <netinet6/in6_pcb.h>
98 #include <netinet6/nd6.h>
99 #include <netinet6/ip6protosw.h>
100 #include <netinet6/scope6_var.h>
101 
102 #ifdef FAST_IPSEC
103 #include <netipsec/ipsec.h>
104 #include <netipsec/ipsec6.h>
105 #include <netipsec/key.h>
106 #include <netipsec/xform.h>
107 #endif
108 
109 
110 #include <net/net_osdep.h>
111 
112 #ifdef PFIL_HOOKS
113 extern struct pfil_head inet6_pfil_hook;	/* XXX */
114 #endif
115 
116 struct ip6_exthdrs {
117 	struct mbuf *ip6e_ip6;
118 	struct mbuf *ip6e_hbh;
119 	struct mbuf *ip6e_dest1;
120 	struct mbuf *ip6e_rthdr;
121 	struct mbuf *ip6e_dest2;
122 };
123 
124 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
125 	kauth_cred_t, int);
126 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
127 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, kauth_cred_t,
128 	int, int, int);
129 static int ip6_setmoptions(const struct sockopt *, struct ip6_moptions **);
130 static int ip6_getmoptions(struct sockopt *, struct ip6_moptions *);
131 static int ip6_copyexthdr(struct mbuf **, void *, int);
132 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
133 	struct ip6_frag **);
134 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
135 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
136 static int ip6_getpmtu(struct route *, struct route *, struct ifnet *,
137     const struct in6_addr *, u_long *, int *);
138 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
139 
140 #ifdef RFC2292
141 static int ip6_pcbopts(struct ip6_pktopts **, struct socket *, struct sockopt *);
142 #endif
143 
144 /*
145  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
146  * header (with pri, len, nxt, hlim, src, dst).
147  * This function may modify ver and hlim only.
148  * The mbuf chain containing the packet will be freed.
149  * The mbuf opt, if present, will not be freed.
150  *
151  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
152  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
153  * which is rt_rmx.rmx_mtu.
154  */
155 int
156 ip6_output(
157     struct mbuf *m0,
158     struct ip6_pktopts *opt,
159     struct route *ro,
160     int flags,
161     struct ip6_moptions *im6o,
162     struct socket *so,
163     struct ifnet **ifpp		/* XXX: just for statistics */
164 )
165 {
166 	struct ip6_hdr *ip6, *mhip6;
167 	struct ifnet *ifp, *origifp;
168 	struct mbuf *m = m0;
169 	int hlen, tlen, len, off;
170 	bool tso;
171 	struct route ip6route;
172 	struct rtentry *rt = NULL;
173 	const struct sockaddr_in6 *dst = NULL;
174 	struct sockaddr_in6 src_sa, dst_sa;
175 	int error = 0;
176 	struct in6_ifaddr *ia = NULL;
177 	u_long mtu;
178 	int alwaysfrag, dontfrag;
179 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
180 	struct ip6_exthdrs exthdrs;
181 	struct in6_addr finaldst, src0, dst0;
182 	u_int32_t zone;
183 	struct route *ro_pmtu = NULL;
184 	int hdrsplit = 0;
185 	int needipsec = 0;
186 #ifdef FAST_IPSEC
187 	struct secpolicy *sp = NULL;
188 	int s;
189 #endif
190 
191 	memset(&ip6route, 0, sizeof(ip6route));
192 
193 #ifdef  DIAGNOSTIC
194 	if ((m->m_flags & M_PKTHDR) == 0)
195 		panic("ip6_output: no HDR");
196 
197 	if ((m->m_pkthdr.csum_flags &
198 	    (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) {
199 		panic("ip6_output: IPv4 checksum offload flags: %d",
200 		    m->m_pkthdr.csum_flags);
201 	}
202 
203 	if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) ==
204 	    (M_CSUM_TCPv6|M_CSUM_UDPv6)) {
205 		panic("ip6_output: conflicting checksum offload flags: %d",
206 		    m->m_pkthdr.csum_flags);
207 	}
208 #endif
209 
210 	M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr));
211 
212 #define MAKE_EXTHDR(hp, mp)						\
213     do {								\
214 	if (hp) {							\
215 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
216 		error = ip6_copyexthdr((mp), (void *)(hp), 		\
217 		    ((eh)->ip6e_len + 1) << 3);				\
218 		if (error)						\
219 			goto freehdrs;					\
220 	}								\
221     } while (/*CONSTCOND*/ 0)
222 
223 	memset(&exthdrs, 0, sizeof(exthdrs));
224 	if (opt) {
225 		/* Hop-by-Hop options header */
226 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
227 		/* Destination options header(1st part) */
228 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
229 		/* Routing header */
230 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
231 		/* Destination options header(2nd part) */
232 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
233 	}
234 
235 	/*
236 	 * Calculate the total length of the extension header chain.
237 	 * Keep the length of the unfragmentable part for fragmentation.
238 	 */
239 	optlen = 0;
240 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
241 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
242 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
243 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
244 	/* NOTE: we don't add AH/ESP length here. do that later. */
245 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
246 
247 #ifdef FAST_IPSEC
248 	/* Check the security policy (SP) for the packet */
249 
250 	sp = ipsec6_check_policy(m,so,flags,&needipsec,&error);
251 	if (error != 0) {
252 		/*
253 		 * Hack: -EINVAL is used to signal that a packet
254 		 * should be silently discarded.  This is typically
255 		 * because we asked key management for an SA and
256 		 * it was delayed (e.g. kicked up to IKE).
257 		 */
258 	if (error == -EINVAL)
259 		error = 0;
260 	goto freehdrs;
261     }
262 #endif /* FAST_IPSEC */
263 
264 
265 	if (needipsec &&
266 	    (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
267 		in6_delayed_cksum(m);
268 		m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
269 	}
270 
271 
272 	/*
273 	 * If we need IPsec, or there is at least one extension header,
274 	 * separate IP6 header from the payload.
275 	 */
276 	if ((needipsec || optlen) && !hdrsplit) {
277 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
278 			m = NULL;
279 			goto freehdrs;
280 		}
281 		m = exthdrs.ip6e_ip6;
282 		hdrsplit++;
283 	}
284 
285 	/* adjust pointer */
286 	ip6 = mtod(m, struct ip6_hdr *);
287 
288 	/* adjust mbuf packet header length */
289 	m->m_pkthdr.len += optlen;
290 	plen = m->m_pkthdr.len - sizeof(*ip6);
291 
292 	/* If this is a jumbo payload, insert a jumbo payload option. */
293 	if (plen > IPV6_MAXPACKET) {
294 		if (!hdrsplit) {
295 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
296 				m = NULL;
297 				goto freehdrs;
298 			}
299 			m = exthdrs.ip6e_ip6;
300 			hdrsplit++;
301 		}
302 		/* adjust pointer */
303 		ip6 = mtod(m, struct ip6_hdr *);
304 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
305 			goto freehdrs;
306 		optlen += 8; /* XXX JUMBOOPTLEN */
307 		ip6->ip6_plen = 0;
308 	} else
309 		ip6->ip6_plen = htons(plen);
310 
311 	/*
312 	 * Concatenate headers and fill in next header fields.
313 	 * Here we have, on "m"
314 	 *	IPv6 payload
315 	 * and we insert headers accordingly.  Finally, we should be getting:
316 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
317 	 *
318 	 * during the header composing process, "m" points to IPv6 header.
319 	 * "mprev" points to an extension header prior to esp.
320 	 */
321 	{
322 		u_char *nexthdrp = &ip6->ip6_nxt;
323 		struct mbuf *mprev = m;
324 
325 		/*
326 		 * we treat dest2 specially.  this makes IPsec processing
327 		 * much easier.  the goal here is to make mprev point the
328 		 * mbuf prior to dest2.
329 		 *
330 		 * result: IPv6 dest2 payload
331 		 * m and mprev will point to IPv6 header.
332 		 */
333 		if (exthdrs.ip6e_dest2) {
334 			if (!hdrsplit)
335 				panic("assumption failed: hdr not split");
336 			exthdrs.ip6e_dest2->m_next = m->m_next;
337 			m->m_next = exthdrs.ip6e_dest2;
338 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
339 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
340 		}
341 
342 #define MAKE_CHAIN(m, mp, p, i)\
343     do {\
344 	if (m) {\
345 		if (!hdrsplit) \
346 			panic("assumption failed: hdr not split"); \
347 		*mtod((m), u_char *) = *(p);\
348 		*(p) = (i);\
349 		p = mtod((m), u_char *);\
350 		(m)->m_next = (mp)->m_next;\
351 		(mp)->m_next = (m);\
352 		(mp) = (m);\
353 	}\
354     } while (/*CONSTCOND*/ 0)
355 		/*
356 		 * result: IPv6 hbh dest1 rthdr dest2 payload
357 		 * m will point to IPv6 header.  mprev will point to the
358 		 * extension header prior to dest2 (rthdr in the above case).
359 		 */
360 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
361 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
362 		    IPPROTO_DSTOPTS);
363 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
364 		    IPPROTO_ROUTING);
365 
366 		M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data,
367 		    sizeof(struct ip6_hdr) + optlen);
368 	}
369 
370 	/*
371 	 * If there is a routing header, replace destination address field
372 	 * with the first hop of the routing header.
373 	 */
374 	if (exthdrs.ip6e_rthdr) {
375 		struct ip6_rthdr *rh;
376 		struct ip6_rthdr0 *rh0;
377 		struct in6_addr *addr;
378 		struct sockaddr_in6 sa;
379 
380 		rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
381 		    struct ip6_rthdr *));
382 		finaldst = ip6->ip6_dst;
383 		switch (rh->ip6r_type) {
384 		case IPV6_RTHDR_TYPE_0:
385 			 rh0 = (struct ip6_rthdr0 *)rh;
386 			 addr = (struct in6_addr *)(rh0 + 1);
387 
388 			 /*
389 			  * construct a sockaddr_in6 form of
390 			  * the first hop.
391 			  *
392 			  * XXX: we may not have enough
393 			  * information about its scope zone;
394 			  * there is no standard API to pass
395 			  * the information from the
396 			  * application.
397 			  */
398 			 sockaddr_in6_init(&sa, addr, 0, 0, 0);
399 			 if ((error = sa6_embedscope(&sa,
400 			     ip6_use_defzone)) != 0) {
401 				 goto bad;
402 			 }
403 			 ip6->ip6_dst = sa.sin6_addr;
404 			 (void)memmove(&addr[0], &addr[1],
405 			     sizeof(struct in6_addr) *
406 			     (rh0->ip6r0_segleft - 1));
407 			 addr[rh0->ip6r0_segleft - 1] = finaldst;
408 			 /* XXX */
409 			 in6_clearscope(addr + rh0->ip6r0_segleft - 1);
410 			 break;
411 		default:	/* is it possible? */
412 			 error = EINVAL;
413 			 goto bad;
414 		}
415 	}
416 
417 	/* Source address validation */
418 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
419 	    (flags & IPV6_UNSPECSRC) == 0) {
420 		error = EOPNOTSUPP;
421 		IP6_STATINC(IP6_STAT_BADSCOPE);
422 		goto bad;
423 	}
424 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
425 		error = EOPNOTSUPP;
426 		IP6_STATINC(IP6_STAT_BADSCOPE);
427 		goto bad;
428 	}
429 
430 	IP6_STATINC(IP6_STAT_LOCALOUT);
431 
432 	/*
433 	 * Route packet.
434 	 */
435 	/* initialize cached route */
436 	if (ro == NULL) {
437 		ro = &ip6route;
438 	}
439 	ro_pmtu = ro;
440 	if (opt && opt->ip6po_rthdr)
441 		ro = &opt->ip6po_route;
442 
443  	/*
444 	 * if specified, try to fill in the traffic class field.
445 	 * do not override if a non-zero value is already set.
446 	 * we check the diffserv field and the ecn field separately.
447 	 */
448 	if (opt && opt->ip6po_tclass >= 0) {
449 		int mask = 0;
450 
451 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
452 			mask |= 0xfc;
453 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
454 			mask |= 0x03;
455 		if (mask != 0)
456 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
457 	}
458 
459 	/* fill in or override the hop limit field, if necessary. */
460 	if (opt && opt->ip6po_hlim != -1)
461 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
462 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
463 		if (im6o != NULL)
464 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
465 		else
466 			ip6->ip6_hlim = ip6_defmcasthlim;
467 	}
468 
469 #ifdef FAST_IPSEC
470 	if (needipsec) {
471 		s = splsoftnet();
472 		error = ipsec6_process_packet(m,sp->req);
473 
474 		/*
475 		 * Preserve KAME behaviour: ENOENT can be returned
476 		 * when an SA acquire is in progress.  Don't propagate
477 		 * this to user-level; it confuses applications.
478 		 * XXX this will go away when the SADB is redone.
479 		 */
480 		if (error == ENOENT)
481 			error = 0;
482 		splx(s);
483 		goto done;
484 	}
485 #endif /* FAST_IPSEC */
486 
487 
488 
489 	/* adjust pointer */
490 	ip6 = mtod(m, struct ip6_hdr *);
491 
492 	sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
493 	if ((error = in6_selectroute(&dst_sa, opt, im6o, ro,
494 	    &ifp, &rt, 0)) != 0) {
495 		if (ifp != NULL)
496 			in6_ifstat_inc(ifp, ifs6_out_discard);
497 		goto bad;
498 	}
499 	if (rt == NULL) {
500 		/*
501 		 * If in6_selectroute() does not return a route entry,
502 		 * dst may not have been updated.
503 		 */
504 		error = rtcache_setdst(ro, sin6tosa(&dst_sa));
505 		if (error) {
506 			goto bad;
507 		}
508 	}
509 
510 	/*
511 	 * then rt (for unicast) and ifp must be non-NULL valid values.
512 	 */
513 	if ((flags & IPV6_FORWARDING) == 0) {
514 		/* XXX: the FORWARDING flag can be set for mrouting. */
515 		in6_ifstat_inc(ifp, ifs6_out_request);
516 	}
517 	if (rt != NULL) {
518 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
519 		rt->rt_use++;
520 	}
521 
522 	/*
523 	 * The outgoing interface must be in the zone of source and
524 	 * destination addresses.  We should use ia_ifp to support the
525 	 * case of sending packets to an address of our own.
526 	 */
527 	if (ia != NULL && ia->ia_ifp)
528 		origifp = ia->ia_ifp;
529 	else
530 		origifp = ifp;
531 
532 	src0 = ip6->ip6_src;
533 	if (in6_setscope(&src0, origifp, &zone))
534 		goto badscope;
535 	sockaddr_in6_init(&src_sa, &ip6->ip6_src, 0, 0, 0);
536 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
537 		goto badscope;
538 
539 	dst0 = ip6->ip6_dst;
540 	if (in6_setscope(&dst0, origifp, &zone))
541 		goto badscope;
542 	/* re-initialize to be sure */
543 	sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
544 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id)
545 		goto badscope;
546 
547 	/* scope check is done. */
548 
549 	if (rt == NULL || IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
550 		if (dst == NULL)
551 			dst = satocsin6(rtcache_getdst(ro));
552 		KASSERT(dst != NULL);
553 	} else if (opt && rtcache_validate(&opt->ip6po_nextroute) != NULL) {
554 		/*
555 		 * The nexthop is explicitly specified by the
556 		 * application.  We assume the next hop is an IPv6
557 		 * address.
558 		 */
559 		dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
560 	} else if ((rt->rt_flags & RTF_GATEWAY))
561 		dst = (struct sockaddr_in6 *)rt->rt_gateway;
562 	else if (dst == NULL)
563 		dst = satocsin6(rtcache_getdst(ro));
564 
565 	/*
566 	 * XXXXXX: original code follows:
567 	 */
568 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
569 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
570 	else {
571 		struct	in6_multi *in6m;
572 
573 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
574 
575 		in6_ifstat_inc(ifp, ifs6_out_mcast);
576 
577 		/*
578 		 * Confirm that the outgoing interface supports multicast.
579 		 */
580 		if (!(ifp->if_flags & IFF_MULTICAST)) {
581 			IP6_STATINC(IP6_STAT_NOROUTE);
582 			in6_ifstat_inc(ifp, ifs6_out_discard);
583 			error = ENETUNREACH;
584 			goto bad;
585 		}
586 
587 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
588 		if (in6m != NULL &&
589 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
590 			/*
591 			 * If we belong to the destination multicast group
592 			 * on the outgoing interface, and the caller did not
593 			 * forbid loopback, loop back a copy.
594 			 */
595 			KASSERT(dst != NULL);
596 			ip6_mloopback(ifp, m, dst);
597 		} else {
598 			/*
599 			 * If we are acting as a multicast router, perform
600 			 * multicast forwarding as if the packet had just
601 			 * arrived on the interface to which we are about
602 			 * to send.  The multicast forwarding function
603 			 * recursively calls this function, using the
604 			 * IPV6_FORWARDING flag to prevent infinite recursion.
605 			 *
606 			 * Multicasts that are looped back by ip6_mloopback(),
607 			 * above, will be forwarded by the ip6_input() routine,
608 			 * if necessary.
609 			 */
610 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
611 				if (ip6_mforward(ip6, ifp, m) != 0) {
612 					m_freem(m);
613 					goto done;
614 				}
615 			}
616 		}
617 		/*
618 		 * Multicasts with a hoplimit of zero may be looped back,
619 		 * above, but must not be transmitted on a network.
620 		 * Also, multicasts addressed to the loopback interface
621 		 * are not sent -- the above call to ip6_mloopback() will
622 		 * loop back a copy if this host actually belongs to the
623 		 * destination group on the loopback interface.
624 		 */
625 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
626 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
627 			m_freem(m);
628 			goto done;
629 		}
630 	}
631 
632 	/*
633 	 * Fill the outgoing inteface to tell the upper layer
634 	 * to increment per-interface statistics.
635 	 */
636 	if (ifpp)
637 		*ifpp = ifp;
638 
639 	/* Determine path MTU. */
640 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
641 	    &alwaysfrag)) != 0)
642 		goto bad;
643 
644 	/*
645 	 * The caller of this function may specify to use the minimum MTU
646 	 * in some cases.
647 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
648 	 * setting.  The logic is a bit complicated; by default, unicast
649 	 * packets will follow path MTU while multicast packets will be sent at
650 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
651 	 * including unicast ones will be sent at the minimum MTU.  Multicast
652 	 * packets will always be sent at the minimum MTU unless
653 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
654 	 * See RFC 3542 for more details.
655 	 */
656 	if (mtu > IPV6_MMTU) {
657 		if ((flags & IPV6_MINMTU))
658 			mtu = IPV6_MMTU;
659 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
660 			mtu = IPV6_MMTU;
661 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
662 			 (opt == NULL ||
663 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
664 			mtu = IPV6_MMTU;
665 		}
666 	}
667 
668 	/*
669 	 * clear embedded scope identifiers if necessary.
670 	 * in6_clearscope will touch the addresses only when necessary.
671 	 */
672 	in6_clearscope(&ip6->ip6_src);
673 	in6_clearscope(&ip6->ip6_dst);
674 
675 	/*
676 	 * If the outgoing packet contains a hop-by-hop options header,
677 	 * it must be examined and processed even by the source node.
678 	 * (RFC 2460, section 4.)
679 	 */
680 	if (ip6->ip6_nxt == IPV6_HOPOPTS) {
681 		u_int32_t dummy1; /* XXX unused */
682 		u_int32_t dummy2; /* XXX unused */
683 		int hoff = sizeof(struct ip6_hdr);
684 
685 		if (ip6_hopopts_input(&dummy1, &dummy2, &m, &hoff)) {
686 			/* m was already freed at this point */
687 			error = EINVAL;/* better error? */
688 			goto done;
689 		}
690 
691 		ip6 = mtod(m, struct ip6_hdr *);
692 	}
693 
694 #ifdef PFIL_HOOKS
695 	/*
696 	 * Run through list of hooks for output packets.
697 	 */
698 	if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0)
699 		goto done;
700 	if (m == NULL)
701 		goto done;
702 	ip6 = mtod(m, struct ip6_hdr *);
703 #endif /* PFIL_HOOKS */
704 	/*
705 	 * Send the packet to the outgoing interface.
706 	 * If necessary, do IPv6 fragmentation before sending.
707 	 *
708 	 * the logic here is rather complex:
709 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
710 	 * 1-a:	send as is if tlen <= path mtu
711 	 * 1-b:	fragment if tlen > path mtu
712 	 *
713 	 * 2: if user asks us not to fragment (dontfrag == 1)
714 	 * 2-a:	send as is if tlen <= interface mtu
715 	 * 2-b:	error if tlen > interface mtu
716 	 *
717 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
718 	 *	always fragment
719 	 *
720 	 * 4: if dontfrag == 1 && alwaysfrag == 1
721 	 *	error, as we cannot handle this conflicting request
722 	 */
723 	tlen = m->m_pkthdr.len;
724 	tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0;
725 	if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
726 		dontfrag = 1;
727 	else
728 		dontfrag = 0;
729 
730 	if (dontfrag && alwaysfrag) {	/* case 4 */
731 		/* conflicting request - can't transmit */
732 		error = EMSGSIZE;
733 		goto bad;
734 	}
735 	if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) {	/* case 2-b */
736 		/*
737 		 * Even if the DONTFRAG option is specified, we cannot send the
738 		 * packet when the data length is larger than the MTU of the
739 		 * outgoing interface.
740 		 * Notify the error by sending IPV6_PATHMTU ancillary data as
741 		 * well as returning an error code (the latter is not described
742 		 * in the API spec.)
743 		 */
744 		u_int32_t mtu32;
745 		struct ip6ctlparam ip6cp;
746 
747 		mtu32 = (u_int32_t)mtu;
748 		memset(&ip6cp, 0, sizeof(ip6cp));
749 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
750 		pfctlinput2(PRC_MSGSIZE,
751 		    rtcache_getdst(ro_pmtu), &ip6cp);
752 
753 		error = EMSGSIZE;
754 		goto bad;
755 	}
756 
757 	/*
758 	 * transmit packet without fragmentation
759 	 */
760 	if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) {
761 		/* case 1-a and 2-a */
762 		struct in6_ifaddr *ia6;
763 		int sw_csum;
764 
765 		ip6 = mtod(m, struct ip6_hdr *);
766 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
767 		if (ia6) {
768 			/* Record statistics for this interface address. */
769 			ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len;
770 		}
771 
772 		sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
773 		if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
774 			if (IN6_NEED_CHECKSUM(ifp,
775 			    sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) {
776 				in6_delayed_cksum(m);
777 			}
778 			m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
779 		}
780 
781 		KASSERT(dst != NULL);
782 		if (__predict_true(!tso ||
783 		    (ifp->if_capenable & IFCAP_TSOv6) != 0)) {
784 			error = nd6_output(ifp, origifp, m, dst, rt);
785 		} else {
786 			error = ip6_tso_output(ifp, origifp, m, dst, rt);
787 		}
788 		goto done;
789 	}
790 
791 	if (tso) {
792 		error = EINVAL; /* XXX */
793 		goto bad;
794 	}
795 
796 	/*
797 	 * try to fragment the packet.  case 1-b and 3
798 	 */
799 	if (mtu < IPV6_MMTU) {
800 		/* path MTU cannot be less than IPV6_MMTU */
801 		error = EMSGSIZE;
802 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
803 		goto bad;
804 	} else if (ip6->ip6_plen == 0) {
805 		/* jumbo payload cannot be fragmented */
806 		error = EMSGSIZE;
807 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
808 		goto bad;
809 	} else {
810 		struct mbuf **mnext, *m_frgpart;
811 		struct ip6_frag *ip6f;
812 		u_int32_t id = htonl(ip6_randomid());
813 		u_char nextproto;
814 #if 0				/* see below */
815 		struct ip6ctlparam ip6cp;
816 		u_int32_t mtu32;
817 #endif
818 
819 		/*
820 		 * Too large for the destination or interface;
821 		 * fragment if possible.
822 		 * Must be able to put at least 8 bytes per fragment.
823 		 */
824 		hlen = unfragpartlen;
825 		if (mtu > IPV6_MAXPACKET)
826 			mtu = IPV6_MAXPACKET;
827 
828 #if 0
829 		/*
830 		 * It is believed this code is a leftover from the
831 		 * development of the IPV6_RECVPATHMTU sockopt and
832 		 * associated work to implement RFC3542.
833 		 * It's not entirely clear what the intent of the API
834 		 * is at this point, so disable this code for now.
835 		 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG
836 		 * will send notifications if the application requests.
837 		 */
838 
839 		/* Notify a proper path MTU to applications. */
840 		mtu32 = (u_int32_t)mtu;
841 		memset(&ip6cp, 0, sizeof(ip6cp));
842 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
843 		pfctlinput2(PRC_MSGSIZE,
844 		    rtcache_getdst(ro_pmtu), &ip6cp);
845 #endif
846 
847 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
848 		if (len < 8) {
849 			error = EMSGSIZE;
850 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
851 			goto bad;
852 		}
853 
854 		mnext = &m->m_nextpkt;
855 
856 		/*
857 		 * Change the next header field of the last header in the
858 		 * unfragmentable part.
859 		 */
860 		if (exthdrs.ip6e_rthdr) {
861 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
862 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
863 		} else if (exthdrs.ip6e_dest1) {
864 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
865 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
866 		} else if (exthdrs.ip6e_hbh) {
867 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
868 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
869 		} else {
870 			nextproto = ip6->ip6_nxt;
871 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
872 		}
873 
874 		if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6))
875 		    != 0) {
876 			if (IN6_NEED_CHECKSUM(ifp,
877 			    m->m_pkthdr.csum_flags &
878 			    (M_CSUM_UDPv6|M_CSUM_TCPv6))) {
879 				in6_delayed_cksum(m);
880 			}
881 			m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
882 		}
883 
884 		/*
885 		 * Loop through length of segment after first fragment,
886 		 * make new header and copy data of each part and link onto
887 		 * chain.
888 		 */
889 		m0 = m;
890 		for (off = hlen; off < tlen; off += len) {
891 			struct mbuf *mlast;
892 
893 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
894 			if (!m) {
895 				error = ENOBUFS;
896 				IP6_STATINC(IP6_STAT_ODROPPED);
897 				goto sendorfree;
898 			}
899 			m->m_pkthdr.rcvif = NULL;
900 			m->m_flags = m0->m_flags & M_COPYFLAGS;
901 			*mnext = m;
902 			mnext = &m->m_nextpkt;
903 			m->m_data += max_linkhdr;
904 			mhip6 = mtod(m, struct ip6_hdr *);
905 			*mhip6 = *ip6;
906 			m->m_len = sizeof(*mhip6);
907 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
908 			if (error) {
909 				IP6_STATINC(IP6_STAT_ODROPPED);
910 				goto sendorfree;
911 			}
912 			ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7));
913 			if (off + len >= tlen)
914 				len = tlen - off;
915 			else
916 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
917 			mhip6->ip6_plen = htons((u_int16_t)(len + hlen +
918 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
919 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
920 				error = ENOBUFS;
921 				IP6_STATINC(IP6_STAT_ODROPPED);
922 				goto sendorfree;
923 			}
924 			for (mlast = m; mlast->m_next; mlast = mlast->m_next)
925 				;
926 			mlast->m_next = m_frgpart;
927 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
928 			m->m_pkthdr.rcvif = NULL;
929 			ip6f->ip6f_reserved = 0;
930 			ip6f->ip6f_ident = id;
931 			ip6f->ip6f_nxt = nextproto;
932 			IP6_STATINC(IP6_STAT_OFRAGMENTS);
933 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
934 		}
935 
936 		in6_ifstat_inc(ifp, ifs6_out_fragok);
937 	}
938 
939 	/*
940 	 * Remove leading garbages.
941 	 */
942 sendorfree:
943 	m = m0->m_nextpkt;
944 	m0->m_nextpkt = 0;
945 	m_freem(m0);
946 	for (m0 = m; m; m = m0) {
947 		m0 = m->m_nextpkt;
948 		m->m_nextpkt = 0;
949 		if (error == 0) {
950 			struct in6_ifaddr *ia6;
951 			ip6 = mtod(m, struct ip6_hdr *);
952 			ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
953 			if (ia6) {
954 				/*
955 				 * Record statistics for this interface
956 				 * address.
957 				 */
958 				ia6->ia_ifa.ifa_data.ifad_outbytes +=
959 				    m->m_pkthdr.len;
960 			}
961 			KASSERT(dst != NULL);
962 			error = nd6_output(ifp, origifp, m, dst, rt);
963 		} else
964 			m_freem(m);
965 	}
966 
967 	if (error == 0)
968 		IP6_STATINC(IP6_STAT_FRAGMENTED);
969 
970 done:
971 	rtcache_free(&ip6route);
972 
973 #ifdef FAST_IPSEC
974 	if (sp != NULL)
975 		KEY_FREESP(&sp);
976 #endif /* FAST_IPSEC */
977 
978 
979 	return (error);
980 
981 freehdrs:
982 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
983 	m_freem(exthdrs.ip6e_dest1);
984 	m_freem(exthdrs.ip6e_rthdr);
985 	m_freem(exthdrs.ip6e_dest2);
986 	/* FALLTHROUGH */
987 bad:
988 	m_freem(m);
989 	goto done;
990 badscope:
991 	IP6_STATINC(IP6_STAT_BADSCOPE);
992 	in6_ifstat_inc(origifp, ifs6_out_discard);
993 	if (error == 0)
994 		error = EHOSTUNREACH; /* XXX */
995 	goto bad;
996 }
997 
998 static int
999 ip6_copyexthdr(struct mbuf **mp, void *hdr, int hlen)
1000 {
1001 	struct mbuf *m;
1002 
1003 	if (hlen > MCLBYTES)
1004 		return (ENOBUFS); /* XXX */
1005 
1006 	MGET(m, M_DONTWAIT, MT_DATA);
1007 	if (!m)
1008 		return (ENOBUFS);
1009 
1010 	if (hlen > MLEN) {
1011 		MCLGET(m, M_DONTWAIT);
1012 		if ((m->m_flags & M_EXT) == 0) {
1013 			m_free(m);
1014 			return (ENOBUFS);
1015 		}
1016 	}
1017 	m->m_len = hlen;
1018 	if (hdr)
1019 		bcopy(hdr, mtod(m, void *), hlen);
1020 
1021 	*mp = m;
1022 	return (0);
1023 }
1024 
1025 /*
1026  * Process a delayed payload checksum calculation.
1027  */
1028 void
1029 in6_delayed_cksum(struct mbuf *m)
1030 {
1031 	uint16_t csum, offset;
1032 
1033 	KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0);
1034 	KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0);
1035 	KASSERT((m->m_pkthdr.csum_flags
1036 	    & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0);
1037 
1038 	offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data);
1039 	csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset);
1040 	if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) {
1041 		csum = 0xffff;
1042 	}
1043 
1044 	offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data);
1045 	if ((offset + sizeof(csum)) > m->m_len) {
1046 		m_copyback(m, offset, sizeof(csum), &csum);
1047 	} else {
1048 		*(uint16_t *)(mtod(m, char *) + offset) = csum;
1049 	}
1050 }
1051 
1052 /*
1053  * Insert jumbo payload option.
1054  */
1055 static int
1056 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1057 {
1058 	struct mbuf *mopt;
1059 	u_int8_t *optbuf;
1060 	u_int32_t v;
1061 
1062 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1063 
1064 	/*
1065 	 * If there is no hop-by-hop options header, allocate new one.
1066 	 * If there is one but it doesn't have enough space to store the
1067 	 * jumbo payload option, allocate a cluster to store the whole options.
1068 	 * Otherwise, use it to store the options.
1069 	 */
1070 	if (exthdrs->ip6e_hbh == 0) {
1071 		MGET(mopt, M_DONTWAIT, MT_DATA);
1072 		if (mopt == 0)
1073 			return (ENOBUFS);
1074 		mopt->m_len = JUMBOOPTLEN;
1075 		optbuf = mtod(mopt, u_int8_t *);
1076 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1077 		exthdrs->ip6e_hbh = mopt;
1078 	} else {
1079 		struct ip6_hbh *hbh;
1080 
1081 		mopt = exthdrs->ip6e_hbh;
1082 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1083 			/*
1084 			 * XXX assumption:
1085 			 * - exthdrs->ip6e_hbh is not referenced from places
1086 			 *   other than exthdrs.
1087 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1088 			 */
1089 			int oldoptlen = mopt->m_len;
1090 			struct mbuf *n;
1091 
1092 			/*
1093 			 * XXX: give up if the whole (new) hbh header does
1094 			 * not fit even in an mbuf cluster.
1095 			 */
1096 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1097 				return (ENOBUFS);
1098 
1099 			/*
1100 			 * As a consequence, we must always prepare a cluster
1101 			 * at this point.
1102 			 */
1103 			MGET(n, M_DONTWAIT, MT_DATA);
1104 			if (n) {
1105 				MCLGET(n, M_DONTWAIT);
1106 				if ((n->m_flags & M_EXT) == 0) {
1107 					m_freem(n);
1108 					n = NULL;
1109 				}
1110 			}
1111 			if (!n)
1112 				return (ENOBUFS);
1113 			n->m_len = oldoptlen + JUMBOOPTLEN;
1114 			bcopy(mtod(mopt, void *), mtod(n, void *),
1115 			    oldoptlen);
1116 			optbuf = mtod(n, u_int8_t *) + oldoptlen;
1117 			m_freem(mopt);
1118 			mopt = exthdrs->ip6e_hbh = n;
1119 		} else {
1120 			optbuf = mtod(mopt, u_int8_t *) + mopt->m_len;
1121 			mopt->m_len += JUMBOOPTLEN;
1122 		}
1123 		optbuf[0] = IP6OPT_PADN;
1124 		optbuf[1] = 0;
1125 
1126 		/*
1127 		 * Adjust the header length according to the pad and
1128 		 * the jumbo payload option.
1129 		 */
1130 		hbh = mtod(mopt, struct ip6_hbh *);
1131 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1132 	}
1133 
1134 	/* fill in the option. */
1135 	optbuf[2] = IP6OPT_JUMBO;
1136 	optbuf[3] = 4;
1137 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1138 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1139 
1140 	/* finally, adjust the packet header length */
1141 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1142 
1143 	return (0);
1144 #undef JUMBOOPTLEN
1145 }
1146 
1147 /*
1148  * Insert fragment header and copy unfragmentable header portions.
1149  */
1150 static int
1151 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1152 	struct ip6_frag **frghdrp)
1153 {
1154 	struct mbuf *n, *mlast;
1155 
1156 	if (hlen > sizeof(struct ip6_hdr)) {
1157 		n = m_copym(m0, sizeof(struct ip6_hdr),
1158 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1159 		if (n == 0)
1160 			return (ENOBUFS);
1161 		m->m_next = n;
1162 	} else
1163 		n = m;
1164 
1165 	/* Search for the last mbuf of unfragmentable part. */
1166 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1167 		;
1168 
1169 	if ((mlast->m_flags & M_EXT) == 0 &&
1170 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1171 		/* use the trailing space of the last mbuf for the fragment hdr */
1172 		*frghdrp = (struct ip6_frag *)(mtod(mlast, char *) +
1173 		    mlast->m_len);
1174 		mlast->m_len += sizeof(struct ip6_frag);
1175 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1176 	} else {
1177 		/* allocate a new mbuf for the fragment header */
1178 		struct mbuf *mfrg;
1179 
1180 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1181 		if (mfrg == 0)
1182 			return (ENOBUFS);
1183 		mfrg->m_len = sizeof(struct ip6_frag);
1184 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1185 		mlast->m_next = mfrg;
1186 	}
1187 
1188 	return (0);
1189 }
1190 
1191 static int
1192 ip6_getpmtu(struct route *ro_pmtu, struct route *ro, struct ifnet *ifp,
1193     const struct in6_addr *dst, u_long *mtup, int *alwaysfragp)
1194 {
1195 	struct rtentry *rt;
1196 	u_int32_t mtu = 0;
1197 	int alwaysfrag = 0;
1198 	int error = 0;
1199 
1200 	if (ro_pmtu != ro) {
1201 		union {
1202 			struct sockaddr		dst;
1203 			struct sockaddr_in6	dst6;
1204 		} u;
1205 
1206 		/* The first hop and the final destination may differ. */
1207 		sockaddr_in6_init(&u.dst6, dst, 0, 0, 0);
1208 		rt = rtcache_lookup(ro_pmtu, &u.dst);
1209 	} else
1210 		rt = rtcache_validate(ro_pmtu);
1211 	if (rt != NULL) {
1212 		u_int32_t ifmtu;
1213 
1214 		if (ifp == NULL)
1215 			ifp = rt->rt_ifp;
1216 		ifmtu = IN6_LINKMTU(ifp);
1217 		mtu = rt->rt_rmx.rmx_mtu;
1218 		if (mtu == 0)
1219 			mtu = ifmtu;
1220 		else if (mtu < IPV6_MMTU) {
1221 			/*
1222 			 * RFC2460 section 5, last paragraph:
1223 			 * if we record ICMPv6 too big message with
1224 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1225 			 * or smaller, with fragment header attached.
1226 			 * (fragment header is needed regardless from the
1227 			 * packet size, for translators to identify packets)
1228 			 */
1229 			alwaysfrag = 1;
1230 			mtu = IPV6_MMTU;
1231 		} else if (mtu > ifmtu) {
1232 			/*
1233 			 * The MTU on the route is larger than the MTU on
1234 			 * the interface!  This shouldn't happen, unless the
1235 			 * MTU of the interface has been changed after the
1236 			 * interface was brought up.  Change the MTU in the
1237 			 * route to match the interface MTU (as long as the
1238 			 * field isn't locked).
1239 			 */
1240 			mtu = ifmtu;
1241 			if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
1242 				rt->rt_rmx.rmx_mtu = mtu;
1243 		}
1244 	} else if (ifp) {
1245 		mtu = IN6_LINKMTU(ifp);
1246 	} else
1247 		error = EHOSTUNREACH; /* XXX */
1248 
1249 	*mtup = mtu;
1250 	if (alwaysfragp)
1251 		*alwaysfragp = alwaysfrag;
1252 	return (error);
1253 }
1254 
1255 /*
1256  * IP6 socket option processing.
1257  */
1258 int
1259 ip6_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1260 {
1261 	int optdatalen, uproto;
1262 	void *optdata;
1263 	struct in6pcb *in6p = sotoin6pcb(so);
1264 	int error, optval;
1265 	int level, optname;
1266 
1267 	KASSERT(sopt != NULL);
1268 
1269 	level = sopt->sopt_level;
1270 	optname = sopt->sopt_name;
1271 
1272 	error = optval = 0;
1273 	uproto = (int)so->so_proto->pr_protocol;
1274 
1275 	if (level != IPPROTO_IPV6) {
1276 		return ENOPROTOOPT;
1277 	}
1278 	switch (op) {
1279 	case PRCO_SETOPT:
1280 		switch (optname) {
1281 #ifdef RFC2292
1282 		case IPV6_2292PKTOPTIONS:
1283 			error = ip6_pcbopts(&in6p->in6p_outputopts, so, sopt);
1284 			break;
1285 #endif
1286 
1287 		/*
1288 		 * Use of some Hop-by-Hop options or some
1289 		 * Destination options, might require special
1290 		 * privilege.  That is, normal applications
1291 		 * (without special privilege) might be forbidden
1292 		 * from setting certain options in outgoing packets,
1293 		 * and might never see certain options in received
1294 		 * packets. [RFC 2292 Section 6]
1295 		 * KAME specific note:
1296 		 *  KAME prevents non-privileged users from sending or
1297 		 *  receiving ANY hbh/dst options in order to avoid
1298 		 *  overhead of parsing options in the kernel.
1299 		 */
1300 		case IPV6_RECVHOPOPTS:
1301 		case IPV6_RECVDSTOPTS:
1302 		case IPV6_RECVRTHDRDSTOPTS:
1303 			error = kauth_authorize_network(kauth_cred_get(),
1304 			    KAUTH_NETWORK_IPV6, KAUTH_REQ_NETWORK_IPV6_HOPBYHOP,
1305 			    NULL, NULL, NULL);
1306 			if (error)
1307 				break;
1308 			/* FALLTHROUGH */
1309 		case IPV6_UNICAST_HOPS:
1310 		case IPV6_HOPLIMIT:
1311 		case IPV6_FAITH:
1312 
1313 		case IPV6_RECVPKTINFO:
1314 		case IPV6_RECVHOPLIMIT:
1315 		case IPV6_RECVRTHDR:
1316 		case IPV6_RECVPATHMTU:
1317 		case IPV6_RECVTCLASS:
1318 		case IPV6_V6ONLY:
1319 			error = sockopt_getint(sopt, &optval);
1320 			if (error)
1321 				break;
1322 			switch (optname) {
1323 			case IPV6_UNICAST_HOPS:
1324 				if (optval < -1 || optval >= 256)
1325 					error = EINVAL;
1326 				else {
1327 					/* -1 = kernel default */
1328 					in6p->in6p_hops = optval;
1329 				}
1330 				break;
1331 #define OPTSET(bit) \
1332 do { \
1333 if (optval) \
1334 	in6p->in6p_flags |= (bit); \
1335 else \
1336 	in6p->in6p_flags &= ~(bit); \
1337 } while (/*CONSTCOND*/ 0)
1338 
1339 #ifdef RFC2292
1340 #define OPTSET2292(bit) 			\
1341 do { 						\
1342 in6p->in6p_flags |= IN6P_RFC2292; 	\
1343 if (optval) 				\
1344 	in6p->in6p_flags |= (bit); 	\
1345 else 					\
1346 	in6p->in6p_flags &= ~(bit); 	\
1347 } while (/*CONSTCOND*/ 0)
1348 #endif
1349 
1350 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1351 
1352 			case IPV6_RECVPKTINFO:
1353 #ifdef RFC2292
1354 				/* cannot mix with RFC2292 */
1355 				if (OPTBIT(IN6P_RFC2292)) {
1356 					error = EINVAL;
1357 					break;
1358 				}
1359 #endif
1360 				OPTSET(IN6P_PKTINFO);
1361 				break;
1362 
1363 			case IPV6_HOPLIMIT:
1364 			{
1365 				struct ip6_pktopts **optp;
1366 
1367 #ifdef RFC2292
1368 				/* cannot mix with RFC2292 */
1369 				if (OPTBIT(IN6P_RFC2292)) {
1370 					error = EINVAL;
1371 					break;
1372 				}
1373 #endif
1374 				optp = &in6p->in6p_outputopts;
1375 				error = ip6_pcbopt(IPV6_HOPLIMIT,
1376 						   (u_char *)&optval,
1377 						   sizeof(optval),
1378 						   optp,
1379 						   kauth_cred_get(), uproto);
1380 				break;
1381 			}
1382 
1383 			case IPV6_RECVHOPLIMIT:
1384 #ifdef RFC2292
1385 				/* cannot mix with RFC2292 */
1386 				if (OPTBIT(IN6P_RFC2292)) {
1387 					error = EINVAL;
1388 					break;
1389 				}
1390 #endif
1391 				OPTSET(IN6P_HOPLIMIT);
1392 				break;
1393 
1394 			case IPV6_RECVHOPOPTS:
1395 #ifdef RFC2292
1396 				/* cannot mix with RFC2292 */
1397 				if (OPTBIT(IN6P_RFC2292)) {
1398 					error = EINVAL;
1399 					break;
1400 				}
1401 #endif
1402 				OPTSET(IN6P_HOPOPTS);
1403 				break;
1404 
1405 			case IPV6_RECVDSTOPTS:
1406 #ifdef RFC2292
1407 				/* cannot mix with RFC2292 */
1408 				if (OPTBIT(IN6P_RFC2292)) {
1409 					error = EINVAL;
1410 					break;
1411 				}
1412 #endif
1413 				OPTSET(IN6P_DSTOPTS);
1414 				break;
1415 
1416 			case IPV6_RECVRTHDRDSTOPTS:
1417 #ifdef RFC2292
1418 				/* cannot mix with RFC2292 */
1419 				if (OPTBIT(IN6P_RFC2292)) {
1420 					error = EINVAL;
1421 					break;
1422 				}
1423 #endif
1424 				OPTSET(IN6P_RTHDRDSTOPTS);
1425 				break;
1426 
1427 			case IPV6_RECVRTHDR:
1428 #ifdef RFC2292
1429 				/* cannot mix with RFC2292 */
1430 				if (OPTBIT(IN6P_RFC2292)) {
1431 					error = EINVAL;
1432 					break;
1433 				}
1434 #endif
1435 				OPTSET(IN6P_RTHDR);
1436 				break;
1437 
1438 			case IPV6_FAITH:
1439 				OPTSET(IN6P_FAITH);
1440 				break;
1441 
1442 			case IPV6_RECVPATHMTU:
1443 				/*
1444 				 * We ignore this option for TCP
1445 				 * sockets.
1446 				 * (RFC3542 leaves this case
1447 				 * unspecified.)
1448 				 */
1449 				if (uproto != IPPROTO_TCP)
1450 					OPTSET(IN6P_MTU);
1451 				break;
1452 
1453 			case IPV6_V6ONLY:
1454 				/*
1455 				 * make setsockopt(IPV6_V6ONLY)
1456 				 * available only prior to bind(2).
1457 				 * see ipng mailing list, Jun 22 2001.
1458 				 */
1459 				if (in6p->in6p_lport ||
1460 				    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1461 					error = EINVAL;
1462 					break;
1463 				}
1464 #ifdef INET6_BINDV6ONLY
1465 				if (!optval)
1466 					error = EINVAL;
1467 #else
1468 				OPTSET(IN6P_IPV6_V6ONLY);
1469 #endif
1470 				break;
1471 			case IPV6_RECVTCLASS:
1472 #ifdef RFC2292
1473 				/* cannot mix with RFC2292 XXX */
1474 				if (OPTBIT(IN6P_RFC2292)) {
1475 					error = EINVAL;
1476 					break;
1477 				}
1478 #endif
1479 				OPTSET(IN6P_TCLASS);
1480 				break;
1481 
1482 			}
1483 			break;
1484 
1485 		case IPV6_OTCLASS:
1486 		{
1487 			struct ip6_pktopts **optp;
1488 			u_int8_t tclass;
1489 
1490 			error = sockopt_get(sopt, &tclass, sizeof(tclass));
1491 			if (error)
1492 				break;
1493 			optp = &in6p->in6p_outputopts;
1494 			error = ip6_pcbopt(optname,
1495 					   (u_char *)&tclass,
1496 					   sizeof(tclass),
1497 					   optp,
1498 					   kauth_cred_get(), uproto);
1499 			break;
1500 		}
1501 
1502 		case IPV6_TCLASS:
1503 		case IPV6_DONTFRAG:
1504 		case IPV6_USE_MIN_MTU:
1505 			error = sockopt_getint(sopt, &optval);
1506 			if (error)
1507 				break;
1508 			{
1509 				struct ip6_pktopts **optp;
1510 				optp = &in6p->in6p_outputopts;
1511 				error = ip6_pcbopt(optname,
1512 						   (u_char *)&optval,
1513 						   sizeof(optval),
1514 						   optp,
1515 						   kauth_cred_get(), uproto);
1516 				break;
1517 			}
1518 
1519 #ifdef RFC2292
1520 		case IPV6_2292PKTINFO:
1521 		case IPV6_2292HOPLIMIT:
1522 		case IPV6_2292HOPOPTS:
1523 		case IPV6_2292DSTOPTS:
1524 		case IPV6_2292RTHDR:
1525 			/* RFC 2292 */
1526 			error = sockopt_getint(sopt, &optval);
1527 			if (error)
1528 				break;
1529 
1530 			switch (optname) {
1531 			case IPV6_2292PKTINFO:
1532 				OPTSET2292(IN6P_PKTINFO);
1533 				break;
1534 			case IPV6_2292HOPLIMIT:
1535 				OPTSET2292(IN6P_HOPLIMIT);
1536 				break;
1537 			case IPV6_2292HOPOPTS:
1538 				/*
1539 				 * Check super-user privilege.
1540 				 * See comments for IPV6_RECVHOPOPTS.
1541 				 */
1542 				error =
1543 				    kauth_authorize_network(kauth_cred_get(),
1544 				    KAUTH_NETWORK_IPV6,
1545 				    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL,
1546 				    NULL, NULL);
1547 				if (error)
1548 					return (error);
1549 				OPTSET2292(IN6P_HOPOPTS);
1550 				break;
1551 			case IPV6_2292DSTOPTS:
1552 				error =
1553 				    kauth_authorize_network(kauth_cred_get(),
1554 				    KAUTH_NETWORK_IPV6,
1555 				    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL,
1556 				    NULL, NULL);
1557 				if (error)
1558 					return (error);
1559 				OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1560 				break;
1561 			case IPV6_2292RTHDR:
1562 				OPTSET2292(IN6P_RTHDR);
1563 				break;
1564 			}
1565 			break;
1566 #endif
1567 		case IPV6_PKTINFO:
1568 		case IPV6_HOPOPTS:
1569 		case IPV6_RTHDR:
1570 		case IPV6_DSTOPTS:
1571 		case IPV6_RTHDRDSTOPTS:
1572 		case IPV6_NEXTHOP: {
1573 			/* new advanced API (RFC3542) */
1574 			void *optbuf;
1575 			int optbuflen;
1576 			struct ip6_pktopts **optp;
1577 
1578 #ifdef RFC2292
1579 			/* cannot mix with RFC2292 */
1580 			if (OPTBIT(IN6P_RFC2292)) {
1581 				error = EINVAL;
1582 				break;
1583 			}
1584 #endif
1585 
1586 			optbuflen = sopt->sopt_size;
1587 			optbuf = malloc(optbuflen, M_IP6OPT, M_NOWAIT);
1588 			if (optbuf == NULL) {
1589 				error = ENOBUFS;
1590 				break;
1591 			}
1592 
1593 			sockopt_get(sopt, optbuf, optbuflen);
1594 			optp = &in6p->in6p_outputopts;
1595 			error = ip6_pcbopt(optname, optbuf, optbuflen,
1596 			    optp, kauth_cred_get(), uproto);
1597 			break;
1598 			}
1599 #undef OPTSET
1600 
1601 		case IPV6_MULTICAST_IF:
1602 		case IPV6_MULTICAST_HOPS:
1603 		case IPV6_MULTICAST_LOOP:
1604 		case IPV6_JOIN_GROUP:
1605 		case IPV6_LEAVE_GROUP:
1606 			error = ip6_setmoptions(sopt, &in6p->in6p_moptions);
1607 			break;
1608 
1609 		case IPV6_PORTRANGE:
1610 			error = sockopt_getint(sopt, &optval);
1611 			if (error)
1612 				break;
1613 
1614 			switch (optval) {
1615 			case IPV6_PORTRANGE_DEFAULT:
1616 				in6p->in6p_flags &= ~(IN6P_LOWPORT);
1617 				in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1618 				break;
1619 
1620 			case IPV6_PORTRANGE_HIGH:
1621 				in6p->in6p_flags &= ~(IN6P_LOWPORT);
1622 				in6p->in6p_flags |= IN6P_HIGHPORT;
1623 				break;
1624 
1625 			case IPV6_PORTRANGE_LOW:
1626 				in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1627 				in6p->in6p_flags |= IN6P_LOWPORT;
1628 				break;
1629 
1630 			default:
1631 				error = EINVAL;
1632 				break;
1633 			}
1634 			break;
1635 
1636 
1637 #if defined(FAST_IPSEC)
1638 		case IPV6_IPSEC_POLICY:
1639 			error = ipsec6_set_policy(in6p, optname,
1640 			    sopt->sopt_data, sopt->sopt_size, kauth_cred_get());
1641 			break;
1642 #endif /* IPSEC */
1643 
1644 		default:
1645 			error = ENOPROTOOPT;
1646 			break;
1647 		}
1648 		break;
1649 
1650 	case PRCO_GETOPT:
1651 		switch (optname) {
1652 #ifdef RFC2292
1653 		case IPV6_2292PKTOPTIONS:
1654 			/*
1655 			 * RFC3542 (effectively) deprecated the
1656 			 * semantics of the 2292-style pktoptions.
1657 			 * Since it was not reliable in nature (i.e.,
1658 			 * applications had to expect the lack of some
1659 			 * information after all), it would make sense
1660 			 * to simplify this part by always returning
1661 			 * empty data.
1662 			 */
1663 			break;
1664 #endif
1665 
1666 		case IPV6_RECVHOPOPTS:
1667 		case IPV6_RECVDSTOPTS:
1668 		case IPV6_RECVRTHDRDSTOPTS:
1669 		case IPV6_UNICAST_HOPS:
1670 		case IPV6_RECVPKTINFO:
1671 		case IPV6_RECVHOPLIMIT:
1672 		case IPV6_RECVRTHDR:
1673 		case IPV6_RECVPATHMTU:
1674 
1675 		case IPV6_FAITH:
1676 		case IPV6_V6ONLY:
1677 		case IPV6_PORTRANGE:
1678 		case IPV6_RECVTCLASS:
1679 			switch (optname) {
1680 
1681 			case IPV6_RECVHOPOPTS:
1682 				optval = OPTBIT(IN6P_HOPOPTS);
1683 				break;
1684 
1685 			case IPV6_RECVDSTOPTS:
1686 				optval = OPTBIT(IN6P_DSTOPTS);
1687 				break;
1688 
1689 			case IPV6_RECVRTHDRDSTOPTS:
1690 				optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1691 				break;
1692 
1693 			case IPV6_UNICAST_HOPS:
1694 				optval = in6p->in6p_hops;
1695 				break;
1696 
1697 			case IPV6_RECVPKTINFO:
1698 				optval = OPTBIT(IN6P_PKTINFO);
1699 				break;
1700 
1701 			case IPV6_RECVHOPLIMIT:
1702 				optval = OPTBIT(IN6P_HOPLIMIT);
1703 				break;
1704 
1705 			case IPV6_RECVRTHDR:
1706 				optval = OPTBIT(IN6P_RTHDR);
1707 				break;
1708 
1709 			case IPV6_RECVPATHMTU:
1710 				optval = OPTBIT(IN6P_MTU);
1711 				break;
1712 
1713 			case IPV6_FAITH:
1714 				optval = OPTBIT(IN6P_FAITH);
1715 				break;
1716 
1717 			case IPV6_V6ONLY:
1718 				optval = OPTBIT(IN6P_IPV6_V6ONLY);
1719 				break;
1720 
1721 			case IPV6_PORTRANGE:
1722 			    {
1723 				int flags;
1724 				flags = in6p->in6p_flags;
1725 				if (flags & IN6P_HIGHPORT)
1726 					optval = IPV6_PORTRANGE_HIGH;
1727 				else if (flags & IN6P_LOWPORT)
1728 					optval = IPV6_PORTRANGE_LOW;
1729 				else
1730 					optval = 0;
1731 				break;
1732 			    }
1733 			case IPV6_RECVTCLASS:
1734 				optval = OPTBIT(IN6P_TCLASS);
1735 				break;
1736 
1737 			}
1738 			if (error)
1739 				break;
1740 			error = sockopt_setint(sopt, optval);
1741 			break;
1742 
1743 		case IPV6_PATHMTU:
1744 		    {
1745 			u_long pmtu = 0;
1746 			struct ip6_mtuinfo mtuinfo;
1747 			struct route *ro = &in6p->in6p_route;
1748 
1749 			if (!(so->so_state & SS_ISCONNECTED))
1750 				return (ENOTCONN);
1751 			/*
1752 			 * XXX: we dot not consider the case of source
1753 			 * routing, or optional information to specify
1754 			 * the outgoing interface.
1755 			 */
1756 			error = ip6_getpmtu(ro, NULL, NULL,
1757 			    &in6p->in6p_faddr, &pmtu, NULL);
1758 			if (error)
1759 				break;
1760 			if (pmtu > IPV6_MAXPACKET)
1761 				pmtu = IPV6_MAXPACKET;
1762 
1763 			memset(&mtuinfo, 0, sizeof(mtuinfo));
1764 			mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1765 			optdata = (void *)&mtuinfo;
1766 			optdatalen = sizeof(mtuinfo);
1767 			if (optdatalen > MCLBYTES)
1768 				return (EMSGSIZE); /* XXX */
1769 			error = sockopt_set(sopt, optdata, optdatalen);
1770 			break;
1771 		    }
1772 
1773 #ifdef RFC2292
1774 		case IPV6_2292PKTINFO:
1775 		case IPV6_2292HOPLIMIT:
1776 		case IPV6_2292HOPOPTS:
1777 		case IPV6_2292RTHDR:
1778 		case IPV6_2292DSTOPTS:
1779 			switch (optname) {
1780 			case IPV6_2292PKTINFO:
1781 				optval = OPTBIT(IN6P_PKTINFO);
1782 				break;
1783 			case IPV6_2292HOPLIMIT:
1784 				optval = OPTBIT(IN6P_HOPLIMIT);
1785 				break;
1786 			case IPV6_2292HOPOPTS:
1787 				optval = OPTBIT(IN6P_HOPOPTS);
1788 				break;
1789 			case IPV6_2292RTHDR:
1790 				optval = OPTBIT(IN6P_RTHDR);
1791 				break;
1792 			case IPV6_2292DSTOPTS:
1793 				optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1794 				break;
1795 			}
1796 			error = sockopt_setint(sopt, optval);
1797 			break;
1798 #endif
1799 		case IPV6_PKTINFO:
1800 		case IPV6_HOPOPTS:
1801 		case IPV6_RTHDR:
1802 		case IPV6_DSTOPTS:
1803 		case IPV6_RTHDRDSTOPTS:
1804 		case IPV6_NEXTHOP:
1805 		case IPV6_OTCLASS:
1806 		case IPV6_TCLASS:
1807 		case IPV6_DONTFRAG:
1808 		case IPV6_USE_MIN_MTU:
1809 			error = ip6_getpcbopt(in6p->in6p_outputopts,
1810 			    optname, sopt);
1811 			break;
1812 
1813 		case IPV6_MULTICAST_IF:
1814 		case IPV6_MULTICAST_HOPS:
1815 		case IPV6_MULTICAST_LOOP:
1816 		case IPV6_JOIN_GROUP:
1817 		case IPV6_LEAVE_GROUP:
1818 			error = ip6_getmoptions(sopt, in6p->in6p_moptions);
1819 			break;
1820 
1821 #if defined(FAST_IPSEC)
1822 		case IPV6_IPSEC_POLICY:
1823 		    {
1824 			struct mbuf *m = NULL;
1825 
1826 			/* XXX this will return EINVAL as sopt is empty */
1827 			error = ipsec6_get_policy(in6p, sopt->sopt_data,
1828 			    sopt->sopt_size, &m);
1829 			if (!error)
1830 				error = sockopt_setmbuf(sopt, m);
1831 
1832 			break;
1833 		    }
1834 #endif /* IPSEC */
1835 
1836 		default:
1837 			error = ENOPROTOOPT;
1838 			break;
1839 		}
1840 		break;
1841 	}
1842 	return (error);
1843 }
1844 
1845 int
1846 ip6_raw_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1847 {
1848 	int error = 0, optval;
1849 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
1850 	struct in6pcb *in6p = sotoin6pcb(so);
1851 	int level, optname;
1852 
1853 	KASSERT(sopt != NULL);
1854 
1855 	level = sopt->sopt_level;
1856 	optname = sopt->sopt_name;
1857 
1858 	if (level != IPPROTO_IPV6) {
1859 		return ENOPROTOOPT;
1860 	}
1861 
1862 	switch (optname) {
1863 	case IPV6_CHECKSUM:
1864 		/*
1865 		 * For ICMPv6 sockets, no modification allowed for checksum
1866 		 * offset, permit "no change" values to help existing apps.
1867 		 *
1868 		 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM
1869 		 * for an ICMPv6 socket will fail."  The current
1870 		 * behavior does not meet RFC3542.
1871 		 */
1872 		switch (op) {
1873 		case PRCO_SETOPT:
1874 			error = sockopt_getint(sopt, &optval);
1875 			if (error)
1876 				break;
1877 			if ((optval % 2) != 0) {
1878 				/* the API assumes even offset values */
1879 				error = EINVAL;
1880 			} else if (so->so_proto->pr_protocol ==
1881 			    IPPROTO_ICMPV6) {
1882 				if (optval != icmp6off)
1883 					error = EINVAL;
1884 			} else
1885 				in6p->in6p_cksum = optval;
1886 			break;
1887 
1888 		case PRCO_GETOPT:
1889 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
1890 				optval = icmp6off;
1891 			else
1892 				optval = in6p->in6p_cksum;
1893 
1894 			error = sockopt_setint(sopt, optval);
1895 			break;
1896 
1897 		default:
1898 			error = EINVAL;
1899 			break;
1900 		}
1901 		break;
1902 
1903 	default:
1904 		error = ENOPROTOOPT;
1905 		break;
1906 	}
1907 
1908 	return (error);
1909 }
1910 
1911 #ifdef RFC2292
1912 /*
1913  * Set up IP6 options in pcb for insertion in output packets or
1914  * specifying behavior of outgoing packets.
1915  */
1916 static int
1917 ip6_pcbopts(struct ip6_pktopts **pktopt, struct socket *so,
1918     struct sockopt *sopt)
1919 {
1920 	struct ip6_pktopts *opt = *pktopt;
1921 	struct mbuf *m;
1922 	int error = 0;
1923 
1924 	/* turn off any old options. */
1925 	if (opt) {
1926 #ifdef DIAGNOSTIC
1927 	    if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
1928 		opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
1929 		opt->ip6po_rhinfo.ip6po_rhi_rthdr)
1930 		    printf("ip6_pcbopts: all specified options are cleared.\n");
1931 #endif
1932 		ip6_clearpktopts(opt, -1);
1933 	} else {
1934 		opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
1935 		if (opt == NULL)
1936 			return (ENOBUFS);
1937 	}
1938 	*pktopt = NULL;
1939 
1940 	if (sopt == NULL || sopt->sopt_size == 0) {
1941 		/*
1942 		 * Only turning off any previous options, regardless of
1943 		 * whether the opt is just created or given.
1944 		 */
1945 		free(opt, M_IP6OPT);
1946 		return (0);
1947 	}
1948 
1949 	/*  set options specified by user. */
1950 	m = sockopt_getmbuf(sopt);
1951 	if (m == NULL) {
1952 		free(opt, M_IP6OPT);
1953 		return (ENOBUFS);
1954 	}
1955 
1956 	error = ip6_setpktopts(m, opt, NULL, kauth_cred_get(),
1957 	    so->so_proto->pr_protocol);
1958 	m_freem(m);
1959 	if (error != 0) {
1960 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
1961 		free(opt, M_IP6OPT);
1962 		return (error);
1963 	}
1964 	*pktopt = opt;
1965 	return (0);
1966 }
1967 #endif
1968 
1969 /*
1970  * initialize ip6_pktopts.  beware that there are non-zero default values in
1971  * the struct.
1972  */
1973 void
1974 ip6_initpktopts(struct ip6_pktopts *opt)
1975 {
1976 
1977 	memset(opt, 0, sizeof(*opt));
1978 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
1979 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
1980 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
1981 }
1982 
1983 #define sin6tosa(sin6)	((struct sockaddr *)(sin6)) /* XXX */
1984 static int
1985 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
1986     kauth_cred_t cred, int uproto)
1987 {
1988 	struct ip6_pktopts *opt;
1989 
1990 	if (*pktopt == NULL) {
1991 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
1992 		    M_NOWAIT);
1993 		if (*pktopt == NULL)
1994 			return (ENOBUFS);
1995 
1996 		ip6_initpktopts(*pktopt);
1997 	}
1998 	opt = *pktopt;
1999 
2000 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2001 }
2002 
2003 static int
2004 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2005 {
2006 	void *optdata = NULL;
2007 	int optdatalen = 0;
2008 	struct ip6_ext *ip6e;
2009 	int error = 0;
2010 	struct in6_pktinfo null_pktinfo;
2011 	int deftclass = 0, on;
2012 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2013 
2014 	switch (optname) {
2015 	case IPV6_PKTINFO:
2016 		if (pktopt && pktopt->ip6po_pktinfo)
2017 			optdata = (void *)pktopt->ip6po_pktinfo;
2018 		else {
2019 			/* XXX: we don't have to do this every time... */
2020 			memset(&null_pktinfo, 0, sizeof(null_pktinfo));
2021 			optdata = (void *)&null_pktinfo;
2022 		}
2023 		optdatalen = sizeof(struct in6_pktinfo);
2024 		break;
2025 	case IPV6_OTCLASS:
2026 		/* XXX */
2027 		return (EINVAL);
2028 	case IPV6_TCLASS:
2029 		if (pktopt && pktopt->ip6po_tclass >= 0)
2030 			optdata = (void *)&pktopt->ip6po_tclass;
2031 		else
2032 			optdata = (void *)&deftclass;
2033 		optdatalen = sizeof(int);
2034 		break;
2035 	case IPV6_HOPOPTS:
2036 		if (pktopt && pktopt->ip6po_hbh) {
2037 			optdata = (void *)pktopt->ip6po_hbh;
2038 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2039 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2040 		}
2041 		break;
2042 	case IPV6_RTHDR:
2043 		if (pktopt && pktopt->ip6po_rthdr) {
2044 			optdata = (void *)pktopt->ip6po_rthdr;
2045 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2046 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2047 		}
2048 		break;
2049 	case IPV6_RTHDRDSTOPTS:
2050 		if (pktopt && pktopt->ip6po_dest1) {
2051 			optdata = (void *)pktopt->ip6po_dest1;
2052 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2053 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2054 		}
2055 		break;
2056 	case IPV6_DSTOPTS:
2057 		if (pktopt && pktopt->ip6po_dest2) {
2058 			optdata = (void *)pktopt->ip6po_dest2;
2059 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2060 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2061 		}
2062 		break;
2063 	case IPV6_NEXTHOP:
2064 		if (pktopt && pktopt->ip6po_nexthop) {
2065 			optdata = (void *)pktopt->ip6po_nexthop;
2066 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2067 		}
2068 		break;
2069 	case IPV6_USE_MIN_MTU:
2070 		if (pktopt)
2071 			optdata = (void *)&pktopt->ip6po_minmtu;
2072 		else
2073 			optdata = (void *)&defminmtu;
2074 		optdatalen = sizeof(int);
2075 		break;
2076 	case IPV6_DONTFRAG:
2077 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2078 			on = 1;
2079 		else
2080 			on = 0;
2081 		optdata = (void *)&on;
2082 		optdatalen = sizeof(on);
2083 		break;
2084 	default:		/* should not happen */
2085 #ifdef DIAGNOSTIC
2086 		panic("ip6_getpcbopt: unexpected option\n");
2087 #endif
2088 		return (ENOPROTOOPT);
2089 	}
2090 
2091 	error = sockopt_set(sopt, optdata, optdatalen);
2092 
2093 	return (error);
2094 }
2095 
2096 void
2097 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2098 {
2099 	if (optname == -1 || optname == IPV6_PKTINFO) {
2100 		if (pktopt->ip6po_pktinfo)
2101 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2102 		pktopt->ip6po_pktinfo = NULL;
2103 	}
2104 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2105 		pktopt->ip6po_hlim = -1;
2106 	if (optname == -1 || optname == IPV6_TCLASS)
2107 		pktopt->ip6po_tclass = -1;
2108 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2109 		rtcache_free(&pktopt->ip6po_nextroute);
2110 		if (pktopt->ip6po_nexthop)
2111 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2112 		pktopt->ip6po_nexthop = NULL;
2113 	}
2114 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2115 		if (pktopt->ip6po_hbh)
2116 			free(pktopt->ip6po_hbh, M_IP6OPT);
2117 		pktopt->ip6po_hbh = NULL;
2118 	}
2119 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2120 		if (pktopt->ip6po_dest1)
2121 			free(pktopt->ip6po_dest1, M_IP6OPT);
2122 		pktopt->ip6po_dest1 = NULL;
2123 	}
2124 	if (optname == -1 || optname == IPV6_RTHDR) {
2125 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2126 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2127 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2128 		rtcache_free(&pktopt->ip6po_route);
2129 	}
2130 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2131 		if (pktopt->ip6po_dest2)
2132 			free(pktopt->ip6po_dest2, M_IP6OPT);
2133 		pktopt->ip6po_dest2 = NULL;
2134 	}
2135 }
2136 
2137 #define PKTOPT_EXTHDRCPY(type) 					\
2138 do {								\
2139 	if (src->type) {					\
2140 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2141 		dst->type = malloc(hlen, M_IP6OPT, canwait);	\
2142 		if (dst->type == NULL)				\
2143 			goto bad;				\
2144 		memcpy(dst->type, src->type, hlen);		\
2145 	}							\
2146 } while (/*CONSTCOND*/ 0)
2147 
2148 static int
2149 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2150 {
2151 	dst->ip6po_hlim = src->ip6po_hlim;
2152 	dst->ip6po_tclass = src->ip6po_tclass;
2153 	dst->ip6po_flags = src->ip6po_flags;
2154 	if (src->ip6po_pktinfo) {
2155 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2156 		    M_IP6OPT, canwait);
2157 		if (dst->ip6po_pktinfo == NULL)
2158 			goto bad;
2159 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2160 	}
2161 	if (src->ip6po_nexthop) {
2162 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2163 		    M_IP6OPT, canwait);
2164 		if (dst->ip6po_nexthop == NULL)
2165 			goto bad;
2166 		memcpy(dst->ip6po_nexthop, src->ip6po_nexthop,
2167 		    src->ip6po_nexthop->sa_len);
2168 	}
2169 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2170 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2171 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2172 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2173 	return (0);
2174 
2175   bad:
2176 	if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
2177 	if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
2178 	if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
2179 	if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
2180 	if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
2181 	if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
2182 
2183 	return (ENOBUFS);
2184 }
2185 #undef PKTOPT_EXTHDRCPY
2186 
2187 struct ip6_pktopts *
2188 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2189 {
2190 	int error;
2191 	struct ip6_pktopts *dst;
2192 
2193 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2194 	if (dst == NULL)
2195 		return (NULL);
2196 	ip6_initpktopts(dst);
2197 
2198 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2199 		free(dst, M_IP6OPT);
2200 		return (NULL);
2201 	}
2202 
2203 	return (dst);
2204 }
2205 
2206 void
2207 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2208 {
2209 	if (pktopt == NULL)
2210 		return;
2211 
2212 	ip6_clearpktopts(pktopt, -1);
2213 
2214 	free(pktopt, M_IP6OPT);
2215 }
2216 
2217 /*
2218  * Set the IP6 multicast options in response to user setsockopt().
2219  */
2220 static int
2221 ip6_setmoptions(const struct sockopt *sopt, struct ip6_moptions **im6op)
2222 {
2223 	int error = 0;
2224 	u_int loop, ifindex;
2225 	struct ipv6_mreq mreq;
2226 	struct ifnet *ifp;
2227 	struct ip6_moptions *im6o = *im6op;
2228 	struct route ro;
2229 	struct in6_multi_mship *imm;
2230 	struct lwp *l = curlwp;	/* XXX */
2231 
2232 	if (im6o == NULL) {
2233 		/*
2234 		 * No multicast option buffer attached to the pcb;
2235 		 * allocate one and initialize to default values.
2236 		 */
2237 		im6o = malloc(sizeof(*im6o), M_IPMOPTS, M_NOWAIT);
2238 		if (im6o == NULL)
2239 			return (ENOBUFS);
2240 
2241 		*im6op = im6o;
2242 		im6o->im6o_multicast_ifp = NULL;
2243 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2244 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2245 		LIST_INIT(&im6o->im6o_memberships);
2246 	}
2247 
2248 	switch (sopt->sopt_name) {
2249 
2250 	case IPV6_MULTICAST_IF:
2251 		/*
2252 		 * Select the interface for outgoing multicast packets.
2253 		 */
2254 		error = sockopt_get(sopt, &ifindex, sizeof(ifindex));
2255 		if (error != 0)
2256 			break;
2257 
2258 		if (ifindex != 0) {
2259 			if (if_indexlim <= ifindex || !ifindex2ifnet[ifindex]) {
2260 				error = ENXIO;	/* XXX EINVAL? */
2261 				break;
2262 			}
2263 			ifp = ifindex2ifnet[ifindex];
2264 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
2265 				error = EADDRNOTAVAIL;
2266 				break;
2267 			}
2268 		} else
2269 			ifp = NULL;
2270 		im6o->im6o_multicast_ifp = ifp;
2271 		break;
2272 
2273 	case IPV6_MULTICAST_HOPS:
2274 	    {
2275 		/*
2276 		 * Set the IP6 hoplimit for outgoing multicast packets.
2277 		 */
2278 		int optval;
2279 
2280 		error = sockopt_getint(sopt, &optval);
2281 		if (error != 0)
2282 			break;
2283 
2284 		if (optval < -1 || optval >= 256)
2285 			error = EINVAL;
2286 		else if (optval == -1)
2287 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2288 		else
2289 			im6o->im6o_multicast_hlim = optval;
2290 		break;
2291 	    }
2292 
2293 	case IPV6_MULTICAST_LOOP:
2294 		/*
2295 		 * Set the loopback flag for outgoing multicast packets.
2296 		 * Must be zero or one.
2297 		 */
2298 		error = sockopt_get(sopt, &loop, sizeof(loop));
2299 		if (error != 0)
2300 			break;
2301 		if (loop > 1) {
2302 			error = EINVAL;
2303 			break;
2304 		}
2305 		im6o->im6o_multicast_loop = loop;
2306 		break;
2307 
2308 	case IPV6_JOIN_GROUP:
2309 		/*
2310 		 * Add a multicast group membership.
2311 		 * Group must be a valid IP6 multicast address.
2312 		 */
2313 		error = sockopt_get(sopt, &mreq, sizeof(mreq));
2314 		if (error != 0)
2315 			break;
2316 
2317 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq.ipv6mr_multiaddr)) {
2318 			/*
2319 			 * We use the unspecified address to specify to accept
2320 			 * all multicast addresses. Only super user is allowed
2321 			 * to do this.
2322 			 */
2323 			if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_IPV6,
2324 			    KAUTH_REQ_NETWORK_IPV6_JOIN_MULTICAST, NULL, NULL, NULL))
2325 			{
2326 				error = EACCES;
2327 				break;
2328 			}
2329 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq.ipv6mr_multiaddr)) {
2330 			error = EINVAL;
2331 			break;
2332 		}
2333 
2334 		/*
2335 		 * If no interface was explicitly specified, choose an
2336 		 * appropriate one according to the given multicast address.
2337 		 */
2338 		if (mreq.ipv6mr_interface == 0) {
2339 			struct rtentry *rt;
2340 			union {
2341 				struct sockaddr		dst;
2342 				struct sockaddr_in6	dst6;
2343 			} u;
2344 
2345 			/*
2346 			 * Look up the routing table for the
2347 			 * address, and choose the outgoing interface.
2348 			 *   XXX: is it a good approach?
2349 			 */
2350 			memset(&ro, 0, sizeof(ro));
2351 			sockaddr_in6_init(&u.dst6, &mreq.ipv6mr_multiaddr, 0,
2352 			    0, 0);
2353 			rtcache_setdst(&ro, &u.dst);
2354 			ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp
2355 			                                        : NULL;
2356 			rtcache_free(&ro);
2357 		} else {
2358 			/*
2359 			 * If the interface is specified, validate it.
2360 			 */
2361 			if (if_indexlim <= mreq.ipv6mr_interface ||
2362 			    !ifindex2ifnet[mreq.ipv6mr_interface]) {
2363 				error = ENXIO;	/* XXX EINVAL? */
2364 				break;
2365 			}
2366 			ifp = ifindex2ifnet[mreq.ipv6mr_interface];
2367 		}
2368 
2369 		/*
2370 		 * See if we found an interface, and confirm that it
2371 		 * supports multicast
2372 		 */
2373 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2374 			error = EADDRNOTAVAIL;
2375 			break;
2376 		}
2377 
2378 		if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) {
2379 			error = EADDRNOTAVAIL; /* XXX: should not happen */
2380 			break;
2381 		}
2382 
2383 		/*
2384 		 * See if the membership already exists.
2385 		 */
2386 		for (imm = im6o->im6o_memberships.lh_first;
2387 		     imm != NULL; imm = imm->i6mm_chain.le_next)
2388 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
2389 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2390 			    &mreq.ipv6mr_multiaddr))
2391 				break;
2392 		if (imm != NULL) {
2393 			error = EADDRINUSE;
2394 			break;
2395 		}
2396 		/*
2397 		 * Everything looks good; add a new record to the multicast
2398 		 * address list for the given interface.
2399 		 */
2400 		imm = in6_joingroup(ifp, &mreq.ipv6mr_multiaddr, &error, 0);
2401 		if (imm == NULL)
2402 			break;
2403 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2404 		break;
2405 
2406 	case IPV6_LEAVE_GROUP:
2407 		/*
2408 		 * Drop a multicast group membership.
2409 		 * Group must be a valid IP6 multicast address.
2410 		 */
2411 		error = sockopt_get(sopt, &mreq, sizeof(mreq));
2412 		if (error != 0)
2413 			break;
2414 
2415 		/*
2416 		 * If an interface address was specified, get a pointer
2417 		 * to its ifnet structure.
2418 		 */
2419 		if (mreq.ipv6mr_interface != 0) {
2420 			if (if_indexlim <= mreq.ipv6mr_interface ||
2421 			    !ifindex2ifnet[mreq.ipv6mr_interface]) {
2422 				error = ENXIO;	/* XXX EINVAL? */
2423 				break;
2424 			}
2425 			ifp = ifindex2ifnet[mreq.ipv6mr_interface];
2426 		} else
2427 			ifp = NULL;
2428 
2429 		/* Fill in the scope zone ID */
2430 		if (ifp) {
2431 			if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) {
2432 				/* XXX: should not happen */
2433 				error = EADDRNOTAVAIL;
2434 				break;
2435 			}
2436 		} else if (mreq.ipv6mr_interface != 0) {
2437 			/*
2438 			 * XXX: This case would happens when the (positive)
2439 			 * index is in the valid range, but the corresponding
2440 			 * interface has been detached dynamically.  The above
2441 			 * check probably avoids such case to happen here, but
2442 			 * we check it explicitly for safety.
2443 			 */
2444 			error = EADDRNOTAVAIL;
2445 			break;
2446 		} else {	/* ipv6mr_interface == 0 */
2447 			struct sockaddr_in6 sa6_mc;
2448 
2449 			/*
2450 			 * The API spec says as follows:
2451 			 *  If the interface index is specified as 0, the
2452 			 *  system may choose a multicast group membership to
2453 			 *  drop by matching the multicast address only.
2454 			 * On the other hand, we cannot disambiguate the scope
2455 			 * zone unless an interface is provided.  Thus, we
2456 			 * check if there's ambiguity with the default scope
2457 			 * zone as the last resort.
2458 			 */
2459 			sockaddr_in6_init(&sa6_mc, &mreq.ipv6mr_multiaddr,
2460 			    0, 0, 0);
2461 			error = sa6_embedscope(&sa6_mc, ip6_use_defzone);
2462 			if (error != 0)
2463 				break;
2464 			mreq.ipv6mr_multiaddr = sa6_mc.sin6_addr;
2465 		}
2466 
2467 		/*
2468 		 * Find the membership in the membership list.
2469 		 */
2470 		for (imm = im6o->im6o_memberships.lh_first;
2471 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
2472 			if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2473 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2474 			    &mreq.ipv6mr_multiaddr))
2475 				break;
2476 		}
2477 		if (imm == NULL) {
2478 			/* Unable to resolve interface */
2479 			error = EADDRNOTAVAIL;
2480 			break;
2481 		}
2482 		/*
2483 		 * Give up the multicast address record to which the
2484 		 * membership points.
2485 		 */
2486 		LIST_REMOVE(imm, i6mm_chain);
2487 		in6_leavegroup(imm);
2488 		break;
2489 
2490 	default:
2491 		error = EOPNOTSUPP;
2492 		break;
2493 	}
2494 
2495 	/*
2496 	 * If all options have default values, no need to keep the mbuf.
2497 	 */
2498 	if (im6o->im6o_multicast_ifp == NULL &&
2499 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2500 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2501 	    im6o->im6o_memberships.lh_first == NULL) {
2502 		free(*im6op, M_IPMOPTS);
2503 		*im6op = NULL;
2504 	}
2505 
2506 	return (error);
2507 }
2508 
2509 /*
2510  * Return the IP6 multicast options in response to user getsockopt().
2511  */
2512 static int
2513 ip6_getmoptions(struct sockopt *sopt, struct ip6_moptions *im6o)
2514 {
2515 	u_int optval;
2516 	int error;
2517 
2518 	switch (sopt->sopt_name) {
2519 	case IPV6_MULTICAST_IF:
2520 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2521 			optval = 0;
2522 		else
2523 			optval = im6o->im6o_multicast_ifp->if_index;
2524 
2525 		error = sockopt_set(sopt, &optval, sizeof(optval));
2526 		break;
2527 
2528 	case IPV6_MULTICAST_HOPS:
2529 		if (im6o == NULL)
2530 			optval = ip6_defmcasthlim;
2531 		else
2532 			optval = im6o->im6o_multicast_hlim;
2533 
2534 		error = sockopt_set(sopt, &optval, sizeof(optval));
2535 		break;
2536 
2537 	case IPV6_MULTICAST_LOOP:
2538 		if (im6o == NULL)
2539 			optval = ip6_defmcasthlim;
2540 		else
2541 			optval = im6o->im6o_multicast_loop;
2542 
2543 		error = sockopt_set(sopt, &optval, sizeof(optval));
2544 		break;
2545 
2546 	default:
2547 		error = EOPNOTSUPP;
2548 	}
2549 
2550 	return (error);
2551 }
2552 
2553 /*
2554  * Discard the IP6 multicast options.
2555  */
2556 void
2557 ip6_freemoptions(struct ip6_moptions *im6o)
2558 {
2559 	struct in6_multi_mship *imm;
2560 
2561 	if (im6o == NULL)
2562 		return;
2563 
2564 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2565 		LIST_REMOVE(imm, i6mm_chain);
2566 		in6_leavegroup(imm);
2567 	}
2568 	free(im6o, M_IPMOPTS);
2569 }
2570 
2571 /*
2572  * Set IPv6 outgoing packet options based on advanced API.
2573  */
2574 int
2575 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2576 	struct ip6_pktopts *stickyopt, kauth_cred_t cred, int uproto)
2577 {
2578 	struct cmsghdr *cm = 0;
2579 
2580 	if (control == NULL || opt == NULL)
2581 		return (EINVAL);
2582 
2583 	ip6_initpktopts(opt);
2584 	if (stickyopt) {
2585 		int error;
2586 
2587 		/*
2588 		 * If stickyopt is provided, make a local copy of the options
2589 		 * for this particular packet, then override them by ancillary
2590 		 * objects.
2591 		 * XXX: copypktopts() does not copy the cached route to a next
2592 		 * hop (if any).  This is not very good in terms of efficiency,
2593 		 * but we can allow this since this option should be rarely
2594 		 * used.
2595 		 */
2596 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2597 			return (error);
2598 	}
2599 
2600 	/*
2601 	 * XXX: Currently, we assume all the optional information is stored
2602 	 * in a single mbuf.
2603 	 */
2604 	if (control->m_next)
2605 		return (EINVAL);
2606 
2607 	/* XXX if cm->cmsg_len is not aligned, control->m_len can become <0 */
2608 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2609 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2610 		int error;
2611 
2612 		if (control->m_len < CMSG_LEN(0))
2613 			return (EINVAL);
2614 
2615 		cm = mtod(control, struct cmsghdr *);
2616 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2617 			return (EINVAL);
2618 		if (cm->cmsg_level != IPPROTO_IPV6)
2619 			continue;
2620 
2621 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2622 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2623 		if (error)
2624 			return (error);
2625 	}
2626 
2627 	return (0);
2628 }
2629 
2630 /*
2631  * Set a particular packet option, as a sticky option or an ancillary data
2632  * item.  "len" can be 0 only when it's a sticky option.
2633  * We have 4 cases of combination of "sticky" and "cmsg":
2634  * "sticky=0, cmsg=0": impossible
2635  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2636  * "sticky=1, cmsg=0": RFC3542 socket option
2637  * "sticky=1, cmsg=1": RFC2292 socket option
2638  */
2639 static int
2640 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2641     kauth_cred_t cred, int sticky, int cmsg, int uproto)
2642 {
2643 	int minmtupolicy;
2644 	int error;
2645 
2646 	if (!sticky && !cmsg) {
2647 #ifdef DIAGNOSTIC
2648 		printf("ip6_setpktopt: impossible case\n");
2649 #endif
2650 		return (EINVAL);
2651 	}
2652 
2653 	/*
2654 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2655 	 * not be specified in the context of RFC3542.  Conversely,
2656 	 * RFC3542 types should not be specified in the context of RFC2292.
2657 	 */
2658 	if (!cmsg) {
2659 		switch (optname) {
2660 		case IPV6_2292PKTINFO:
2661 		case IPV6_2292HOPLIMIT:
2662 		case IPV6_2292NEXTHOP:
2663 		case IPV6_2292HOPOPTS:
2664 		case IPV6_2292DSTOPTS:
2665 		case IPV6_2292RTHDR:
2666 		case IPV6_2292PKTOPTIONS:
2667 			return (ENOPROTOOPT);
2668 		}
2669 	}
2670 	if (sticky && cmsg) {
2671 		switch (optname) {
2672 		case IPV6_PKTINFO:
2673 		case IPV6_HOPLIMIT:
2674 		case IPV6_NEXTHOP:
2675 		case IPV6_HOPOPTS:
2676 		case IPV6_DSTOPTS:
2677 		case IPV6_RTHDRDSTOPTS:
2678 		case IPV6_RTHDR:
2679 		case IPV6_USE_MIN_MTU:
2680 		case IPV6_DONTFRAG:
2681 		case IPV6_OTCLASS:
2682 		case IPV6_TCLASS:
2683 			return (ENOPROTOOPT);
2684 		}
2685 	}
2686 
2687 	switch (optname) {
2688 #ifdef RFC2292
2689 	case IPV6_2292PKTINFO:
2690 #endif
2691 	case IPV6_PKTINFO:
2692 	{
2693 		struct ifnet *ifp = NULL;
2694 		struct in6_pktinfo *pktinfo;
2695 
2696 		if (len != sizeof(struct in6_pktinfo))
2697 			return (EINVAL);
2698 
2699 		pktinfo = (struct in6_pktinfo *)buf;
2700 
2701 		/*
2702 		 * An application can clear any sticky IPV6_PKTINFO option by
2703 		 * doing a "regular" setsockopt with ipi6_addr being
2704 		 * in6addr_any and ipi6_ifindex being zero.
2705 		 * [RFC 3542, Section 6]
2706 		 */
2707 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2708 		    pktinfo->ipi6_ifindex == 0 &&
2709 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2710 			ip6_clearpktopts(opt, optname);
2711 			break;
2712 		}
2713 
2714 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2715 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2716 			return (EINVAL);
2717 		}
2718 
2719 		/* validate the interface index if specified. */
2720 		if (pktinfo->ipi6_ifindex >= if_indexlim) {
2721 			 return (ENXIO);
2722 		}
2723 		if (pktinfo->ipi6_ifindex) {
2724 			ifp = ifindex2ifnet[pktinfo->ipi6_ifindex];
2725 			if (ifp == NULL)
2726 				return (ENXIO);
2727 		}
2728 
2729 		/*
2730 		 * We store the address anyway, and let in6_selectsrc()
2731 		 * validate the specified address.  This is because ipi6_addr
2732 		 * may not have enough information about its scope zone, and
2733 		 * we may need additional information (such as outgoing
2734 		 * interface or the scope zone of a destination address) to
2735 		 * disambiguate the scope.
2736 		 * XXX: the delay of the validation may confuse the
2737 		 * application when it is used as a sticky option.
2738 		 */
2739 		if (opt->ip6po_pktinfo == NULL) {
2740 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2741 			    M_IP6OPT, M_NOWAIT);
2742 			if (opt->ip6po_pktinfo == NULL)
2743 				return (ENOBUFS);
2744 		}
2745 		memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo));
2746 		break;
2747 	}
2748 
2749 #ifdef RFC2292
2750 	case IPV6_2292HOPLIMIT:
2751 #endif
2752 	case IPV6_HOPLIMIT:
2753 	{
2754 		int *hlimp;
2755 
2756 		/*
2757 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2758 		 * to simplify the ordering among hoplimit options.
2759 		 */
2760 		if (optname == IPV6_HOPLIMIT && sticky)
2761 			return (ENOPROTOOPT);
2762 
2763 		if (len != sizeof(int))
2764 			return (EINVAL);
2765 		hlimp = (int *)buf;
2766 		if (*hlimp < -1 || *hlimp > 255)
2767 			return (EINVAL);
2768 
2769 		opt->ip6po_hlim = *hlimp;
2770 		break;
2771 	}
2772 
2773 	case IPV6_OTCLASS:
2774 		if (len != sizeof(u_int8_t))
2775 			return (EINVAL);
2776 
2777 		opt->ip6po_tclass = *(u_int8_t *)buf;
2778 		break;
2779 
2780 	case IPV6_TCLASS:
2781 	{
2782 		int tclass;
2783 
2784 		if (len != sizeof(int))
2785 			return (EINVAL);
2786 		tclass = *(int *)buf;
2787 		if (tclass < -1 || tclass > 255)
2788 			return (EINVAL);
2789 
2790 		opt->ip6po_tclass = tclass;
2791 		break;
2792 	}
2793 
2794 #ifdef RFC2292
2795 	case IPV6_2292NEXTHOP:
2796 #endif
2797 	case IPV6_NEXTHOP:
2798 		error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
2799 		    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2800 		if (error)
2801 			return (error);
2802 
2803 		if (len == 0) {	/* just remove the option */
2804 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2805 			break;
2806 		}
2807 
2808 		/* check if cmsg_len is large enough for sa_len */
2809 		if (len < sizeof(struct sockaddr) || len < *buf)
2810 			return (EINVAL);
2811 
2812 		switch (((struct sockaddr *)buf)->sa_family) {
2813 		case AF_INET6:
2814 		{
2815 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2816 
2817 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2818 				return (EINVAL);
2819 
2820 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2821 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2822 				return (EINVAL);
2823 			}
2824 			if ((error = sa6_embedscope(sa6, ip6_use_defzone))
2825 			    != 0) {
2826 				return (error);
2827 			}
2828 			break;
2829 		}
2830 		case AF_LINK:	/* eventually be supported? */
2831 		default:
2832 			return (EAFNOSUPPORT);
2833 		}
2834 
2835 		/* turn off the previous option, then set the new option. */
2836 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2837 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2838 		if (opt->ip6po_nexthop == NULL)
2839 			return (ENOBUFS);
2840 		memcpy(opt->ip6po_nexthop, buf, *buf);
2841 		break;
2842 
2843 #ifdef RFC2292
2844 	case IPV6_2292HOPOPTS:
2845 #endif
2846 	case IPV6_HOPOPTS:
2847 	{
2848 		struct ip6_hbh *hbh;
2849 		int hbhlen;
2850 
2851 		/*
2852 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2853 		 * options, since per-option restriction has too much
2854 		 * overhead.
2855 		 */
2856 		error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
2857 		    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2858 		if (error)
2859 			return (error);
2860 
2861 		if (len == 0) {
2862 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2863 			break;	/* just remove the option */
2864 		}
2865 
2866 		/* message length validation */
2867 		if (len < sizeof(struct ip6_hbh))
2868 			return (EINVAL);
2869 		hbh = (struct ip6_hbh *)buf;
2870 		hbhlen = (hbh->ip6h_len + 1) << 3;
2871 		if (len != hbhlen)
2872 			return (EINVAL);
2873 
2874 		/* turn off the previous option, then set the new option. */
2875 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2876 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2877 		if (opt->ip6po_hbh == NULL)
2878 			return (ENOBUFS);
2879 		memcpy(opt->ip6po_hbh, hbh, hbhlen);
2880 
2881 		break;
2882 	}
2883 
2884 #ifdef RFC2292
2885 	case IPV6_2292DSTOPTS:
2886 #endif
2887 	case IPV6_DSTOPTS:
2888 	case IPV6_RTHDRDSTOPTS:
2889 	{
2890 		struct ip6_dest *dest, **newdest = NULL;
2891 		int destlen;
2892 
2893 		/* XXX: see the comment for IPV6_HOPOPTS */
2894 		error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
2895 		    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2896 		if (error)
2897 			return (error);
2898 
2899 		if (len == 0) {
2900 			ip6_clearpktopts(opt, optname);
2901 			break;	/* just remove the option */
2902 		}
2903 
2904 		/* message length validation */
2905 		if (len < sizeof(struct ip6_dest))
2906 			return (EINVAL);
2907 		dest = (struct ip6_dest *)buf;
2908 		destlen = (dest->ip6d_len + 1) << 3;
2909 		if (len != destlen)
2910 			return (EINVAL);
2911 		/*
2912 		 * Determine the position that the destination options header
2913 		 * should be inserted; before or after the routing header.
2914 		 */
2915 		switch (optname) {
2916 		case IPV6_2292DSTOPTS:
2917 			/*
2918 			 * The old advanced API is ambiguous on this point.
2919 			 * Our approach is to determine the position based
2920 			 * according to the existence of a routing header.
2921 			 * Note, however, that this depends on the order of the
2922 			 * extension headers in the ancillary data; the 1st
2923 			 * part of the destination options header must appear
2924 			 * before the routing header in the ancillary data,
2925 			 * too.
2926 			 * RFC3542 solved the ambiguity by introducing
2927 			 * separate ancillary data or option types.
2928 			 */
2929 			if (opt->ip6po_rthdr == NULL)
2930 				newdest = &opt->ip6po_dest1;
2931 			else
2932 				newdest = &opt->ip6po_dest2;
2933 			break;
2934 		case IPV6_RTHDRDSTOPTS:
2935 			newdest = &opt->ip6po_dest1;
2936 			break;
2937 		case IPV6_DSTOPTS:
2938 			newdest = &opt->ip6po_dest2;
2939 			break;
2940 		}
2941 
2942 		/* turn off the previous option, then set the new option. */
2943 		ip6_clearpktopts(opt, optname);
2944 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2945 		if (*newdest == NULL)
2946 			return (ENOBUFS);
2947 		memcpy(*newdest, dest, destlen);
2948 
2949 		break;
2950 	}
2951 
2952 #ifdef RFC2292
2953 	case IPV6_2292RTHDR:
2954 #endif
2955 	case IPV6_RTHDR:
2956 	{
2957 		struct ip6_rthdr *rth;
2958 		int rthlen;
2959 
2960 		if (len == 0) {
2961 			ip6_clearpktopts(opt, IPV6_RTHDR);
2962 			break;	/* just remove the option */
2963 		}
2964 
2965 		/* message length validation */
2966 		if (len < sizeof(struct ip6_rthdr))
2967 			return (EINVAL);
2968 		rth = (struct ip6_rthdr *)buf;
2969 		rthlen = (rth->ip6r_len + 1) << 3;
2970 		if (len != rthlen)
2971 			return (EINVAL);
2972 		switch (rth->ip6r_type) {
2973 		case IPV6_RTHDR_TYPE_0:
2974 			if (rth->ip6r_len == 0)	/* must contain one addr */
2975 				return (EINVAL);
2976 			if (rth->ip6r_len % 2) /* length must be even */
2977 				return (EINVAL);
2978 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2979 				return (EINVAL);
2980 			break;
2981 		default:
2982 			return (EINVAL);	/* not supported */
2983 		}
2984 		/* turn off the previous option */
2985 		ip6_clearpktopts(opt, IPV6_RTHDR);
2986 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2987 		if (opt->ip6po_rthdr == NULL)
2988 			return (ENOBUFS);
2989 		memcpy(opt->ip6po_rthdr, rth, rthlen);
2990 		break;
2991 	}
2992 
2993 	case IPV6_USE_MIN_MTU:
2994 		if (len != sizeof(int))
2995 			return (EINVAL);
2996 		minmtupolicy = *(int *)buf;
2997 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2998 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
2999 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3000 			return (EINVAL);
3001 		}
3002 		opt->ip6po_minmtu = minmtupolicy;
3003 		break;
3004 
3005 	case IPV6_DONTFRAG:
3006 		if (len != sizeof(int))
3007 			return (EINVAL);
3008 
3009 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3010 			/*
3011 			 * we ignore this option for TCP sockets.
3012 			 * (RFC3542 leaves this case unspecified.)
3013 			 */
3014 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3015 		} else
3016 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3017 		break;
3018 
3019 	default:
3020 		return (ENOPROTOOPT);
3021 	} /* end of switch */
3022 
3023 	return (0);
3024 }
3025 
3026 /*
3027  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3028  * packet to the input queue of a specified interface.  Note that this
3029  * calls the output routine of the loopback "driver", but with an interface
3030  * pointer that might NOT be lo0ifp -- easier than replicating that code here.
3031  */
3032 void
3033 ip6_mloopback(struct ifnet *ifp, struct mbuf *m,
3034 	const struct sockaddr_in6 *dst)
3035 {
3036 	struct mbuf *copym;
3037 	struct ip6_hdr *ip6;
3038 
3039 	copym = m_copy(m, 0, M_COPYALL);
3040 	if (copym == NULL)
3041 		return;
3042 
3043 	/*
3044 	 * Make sure to deep-copy IPv6 header portion in case the data
3045 	 * is in an mbuf cluster, so that we can safely override the IPv6
3046 	 * header portion later.
3047 	 */
3048 	if ((copym->m_flags & M_EXT) != 0 ||
3049 	    copym->m_len < sizeof(struct ip6_hdr)) {
3050 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3051 		if (copym == NULL)
3052 			return;
3053 	}
3054 
3055 #ifdef DIAGNOSTIC
3056 	if (copym->m_len < sizeof(*ip6)) {
3057 		m_freem(copym);
3058 		return;
3059 	}
3060 #endif
3061 
3062 	ip6 = mtod(copym, struct ip6_hdr *);
3063 	/*
3064 	 * clear embedded scope identifiers if necessary.
3065 	 * in6_clearscope will touch the addresses only when necessary.
3066 	 */
3067 	in6_clearscope(&ip6->ip6_src);
3068 	in6_clearscope(&ip6->ip6_dst);
3069 
3070 	(void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL);
3071 }
3072 
3073 /*
3074  * Chop IPv6 header off from the payload.
3075  */
3076 static int
3077 ip6_splithdr(struct mbuf *m,  struct ip6_exthdrs *exthdrs)
3078 {
3079 	struct mbuf *mh;
3080 	struct ip6_hdr *ip6;
3081 
3082 	ip6 = mtod(m, struct ip6_hdr *);
3083 	if (m->m_len > sizeof(*ip6)) {
3084 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3085 		if (mh == 0) {
3086 			m_freem(m);
3087 			return ENOBUFS;
3088 		}
3089 		M_MOVE_PKTHDR(mh, m);
3090 		MH_ALIGN(mh, sizeof(*ip6));
3091 		m->m_len -= sizeof(*ip6);
3092 		m->m_data += sizeof(*ip6);
3093 		mh->m_next = m;
3094 		m = mh;
3095 		m->m_len = sizeof(*ip6);
3096 		bcopy((void *)ip6, mtod(m, void *), sizeof(*ip6));
3097 	}
3098 	exthdrs->ip6e_ip6 = m;
3099 	return 0;
3100 }
3101 
3102 /*
3103  * Compute IPv6 extension header length.
3104  */
3105 int
3106 ip6_optlen(struct in6pcb *in6p)
3107 {
3108 	int len;
3109 
3110 	if (!in6p->in6p_outputopts)
3111 		return 0;
3112 
3113 	len = 0;
3114 #define elen(x) \
3115     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3116 
3117 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3118 	len += elen(in6p->in6p_outputopts->ip6po_dest1);
3119 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3120 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3121 	return len;
3122 #undef elen
3123 }
3124