xref: /netbsd-src/sys/netinet6/ip6_output.c (revision 6a493d6bc668897c91594964a732d38505b70cbb)
1 /*	$NetBSD: ip6_output.c,v 1.155 2013/10/03 20:27:55 christos 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 1.155 2013/10/03 20:27:55 christos Exp $");
66 
67 #include "opt_inet.h"
68 #include "opt_inet6.h"
69 #include "opt_ipsec.h"
70 
71 #include <sys/param.h>
72 #include <sys/malloc.h>
73 #include <sys/mbuf.h>
74 #include <sys/errno.h>
75 #include <sys/protosw.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/systm.h>
79 #include <sys/proc.h>
80 #include <sys/kauth.h>
81 
82 #include <net/if.h>
83 #include <net/route.h>
84 #include <net/pfil.h>
85 
86 #include <netinet/in.h>
87 #include <netinet/in_var.h>
88 #include <netinet/ip6.h>
89 #include <netinet/icmp6.h>
90 #include <netinet/in_offload.h>
91 #include <netinet/portalgo.h>
92 #include <netinet6/in6_offload.h>
93 #include <netinet6/ip6_var.h>
94 #include <netinet6/ip6_private.h>
95 #include <netinet6/in6_pcb.h>
96 #include <netinet6/nd6.h>
97 #include <netinet6/ip6protosw.h>
98 #include <netinet6/scope6_var.h>
99 
100 #ifdef IPSEC
101 #include <netipsec/ipsec.h>
102 #include <netipsec/ipsec6.h>
103 #include <netipsec/key.h>
104 #include <netipsec/xform.h>
105 #endif
106 
107 
108 #include <net/net_osdep.h>
109 
110 extern pfil_head_t *inet6_pfil_hook;	/* XXX */
111 
112 struct ip6_exthdrs {
113 	struct mbuf *ip6e_ip6;
114 	struct mbuf *ip6e_hbh;
115 	struct mbuf *ip6e_dest1;
116 	struct mbuf *ip6e_rthdr;
117 	struct mbuf *ip6e_dest2;
118 };
119 
120 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
121 	kauth_cred_t, int);
122 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
123 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, kauth_cred_t,
124 	int, int, int);
125 static int ip6_setmoptions(const struct sockopt *, struct ip6_moptions **);
126 static int ip6_getmoptions(struct sockopt *, struct ip6_moptions *);
127 static int ip6_copyexthdr(struct mbuf **, void *, int);
128 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
129 	struct ip6_frag **);
130 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
131 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
132 static int ip6_getpmtu(struct route *, struct route *, struct ifnet *,
133     const struct in6_addr *, u_long *, int *);
134 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
135 
136 #ifdef RFC2292
137 static int ip6_pcbopts(struct ip6_pktopts **, struct socket *, struct sockopt *);
138 #endif
139 
140 /*
141  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
142  * header (with pri, len, nxt, hlim, src, dst).
143  * This function may modify ver and hlim only.
144  * The mbuf chain containing the packet will be freed.
145  * The mbuf opt, if present, will not be freed.
146  *
147  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
148  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
149  * which is rt_rmx.rmx_mtu.
150  */
151 int
152 ip6_output(
153     struct mbuf *m0,
154     struct ip6_pktopts *opt,
155     struct route *ro,
156     int flags,
157     struct ip6_moptions *im6o,
158     struct socket *so,
159     struct ifnet **ifpp		/* XXX: just for statistics */
160 )
161 {
162 	struct ip6_hdr *ip6, *mhip6;
163 	struct ifnet *ifp, *origifp;
164 	struct mbuf *m = m0;
165 	int hlen, tlen, len, off;
166 	bool tso;
167 	struct route ip6route;
168 	struct rtentry *rt = NULL;
169 	const struct sockaddr_in6 *dst = NULL;
170 	struct sockaddr_in6 src_sa, dst_sa;
171 	int error = 0;
172 	struct in6_ifaddr *ia = NULL;
173 	u_long mtu;
174 	int alwaysfrag, dontfrag;
175 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
176 	struct ip6_exthdrs exthdrs;
177 	struct in6_addr finaldst, src0, dst0;
178 	u_int32_t zone;
179 	struct route *ro_pmtu = NULL;
180 	int hdrsplit = 0;
181 	int needipsec = 0;
182 #ifdef IPSEC
183 	struct secpolicy *sp = NULL;
184 	int s;
185 #endif
186 
187 	memset(&ip6route, 0, sizeof(ip6route));
188 
189 #ifdef  DIAGNOSTIC
190 	if ((m->m_flags & M_PKTHDR) == 0)
191 		panic("ip6_output: no HDR");
192 
193 	if ((m->m_pkthdr.csum_flags &
194 	    (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) {
195 		panic("ip6_output: IPv4 checksum offload flags: %d",
196 		    m->m_pkthdr.csum_flags);
197 	}
198 
199 	if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) ==
200 	    (M_CSUM_TCPv6|M_CSUM_UDPv6)) {
201 		panic("ip6_output: conflicting checksum offload flags: %d",
202 		    m->m_pkthdr.csum_flags);
203 	}
204 #endif
205 
206 	M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr));
207 
208 #define MAKE_EXTHDR(hp, mp)						\
209     do {								\
210 	if (hp) {							\
211 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
212 		error = ip6_copyexthdr((mp), (void *)(hp), 		\
213 		    ((eh)->ip6e_len + 1) << 3);				\
214 		if (error)						\
215 			goto freehdrs;					\
216 	}								\
217     } while (/*CONSTCOND*/ 0)
218 
219 	memset(&exthdrs, 0, sizeof(exthdrs));
220 	if (opt) {
221 		/* Hop-by-Hop options header */
222 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
223 		/* Destination options header(1st part) */
224 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
225 		/* Routing header */
226 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
227 		/* Destination options header(2nd part) */
228 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
229 	}
230 
231 	/*
232 	 * Calculate the total length of the extension header chain.
233 	 * Keep the length of the unfragmentable part for fragmentation.
234 	 */
235 	optlen = 0;
236 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
237 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
238 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
239 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
240 	/* NOTE: we don't add AH/ESP length here. do that later. */
241 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
242 
243 #ifdef IPSEC
244 	/* Check the security policy (SP) for the packet */
245 
246 	sp = ipsec6_check_policy(m,so,flags,&needipsec,&error);
247 	if (error != 0) {
248 		/*
249 		 * Hack: -EINVAL is used to signal that a packet
250 		 * should be silently discarded.  This is typically
251 		 * because we asked key management for an SA and
252 		 * it was delayed (e.g. kicked up to IKE).
253 		 */
254 	if (error == -EINVAL)
255 		error = 0;
256 	goto freehdrs;
257     }
258 #endif /* IPSEC */
259 
260 
261 	if (needipsec &&
262 	    (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
263 		in6_delayed_cksum(m);
264 		m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
265 	}
266 
267 
268 	/*
269 	 * If we need IPsec, or there is at least one extension header,
270 	 * separate IP6 header from the payload.
271 	 */
272 	if ((needipsec || optlen) && !hdrsplit) {
273 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
274 			m = NULL;
275 			goto freehdrs;
276 		}
277 		m = exthdrs.ip6e_ip6;
278 		hdrsplit++;
279 	}
280 
281 	/* adjust pointer */
282 	ip6 = mtod(m, struct ip6_hdr *);
283 
284 	/* adjust mbuf packet header length */
285 	m->m_pkthdr.len += optlen;
286 	plen = m->m_pkthdr.len - sizeof(*ip6);
287 
288 	/* If this is a jumbo payload, insert a jumbo payload option. */
289 	if (plen > IPV6_MAXPACKET) {
290 		if (!hdrsplit) {
291 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
292 				m = NULL;
293 				goto freehdrs;
294 			}
295 			m = exthdrs.ip6e_ip6;
296 			hdrsplit++;
297 		}
298 		/* adjust pointer */
299 		ip6 = mtod(m, struct ip6_hdr *);
300 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
301 			goto freehdrs;
302 		optlen += 8; /* XXX JUMBOOPTLEN */
303 		ip6->ip6_plen = 0;
304 	} else
305 		ip6->ip6_plen = htons(plen);
306 
307 	/*
308 	 * Concatenate headers and fill in next header fields.
309 	 * Here we have, on "m"
310 	 *	IPv6 payload
311 	 * and we insert headers accordingly.  Finally, we should be getting:
312 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
313 	 *
314 	 * during the header composing process, "m" points to IPv6 header.
315 	 * "mprev" points to an extension header prior to esp.
316 	 */
317 	{
318 		u_char *nexthdrp = &ip6->ip6_nxt;
319 		struct mbuf *mprev = m;
320 
321 		/*
322 		 * we treat dest2 specially.  this makes IPsec processing
323 		 * much easier.  the goal here is to make mprev point the
324 		 * mbuf prior to dest2.
325 		 *
326 		 * result: IPv6 dest2 payload
327 		 * m and mprev will point to IPv6 header.
328 		 */
329 		if (exthdrs.ip6e_dest2) {
330 			if (!hdrsplit)
331 				panic("assumption failed: hdr not split");
332 			exthdrs.ip6e_dest2->m_next = m->m_next;
333 			m->m_next = exthdrs.ip6e_dest2;
334 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
335 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
336 		}
337 
338 #define MAKE_CHAIN(m, mp, p, i)\
339     do {\
340 	if (m) {\
341 		if (!hdrsplit) \
342 			panic("assumption failed: hdr not split"); \
343 		*mtod((m), u_char *) = *(p);\
344 		*(p) = (i);\
345 		p = mtod((m), u_char *);\
346 		(m)->m_next = (mp)->m_next;\
347 		(mp)->m_next = (m);\
348 		(mp) = (m);\
349 	}\
350     } while (/*CONSTCOND*/ 0)
351 		/*
352 		 * result: IPv6 hbh dest1 rthdr dest2 payload
353 		 * m will point to IPv6 header.  mprev will point to the
354 		 * extension header prior to dest2 (rthdr in the above case).
355 		 */
356 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
357 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
358 		    IPPROTO_DSTOPTS);
359 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
360 		    IPPROTO_ROUTING);
361 
362 		M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data,
363 		    sizeof(struct ip6_hdr) + optlen);
364 	}
365 
366 	/*
367 	 * If there is a routing header, replace destination address field
368 	 * with the first hop of the routing header.
369 	 */
370 	if (exthdrs.ip6e_rthdr) {
371 		struct ip6_rthdr *rh;
372 		struct ip6_rthdr0 *rh0;
373 		struct in6_addr *addr;
374 		struct sockaddr_in6 sa;
375 
376 		rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
377 		    struct ip6_rthdr *));
378 		finaldst = ip6->ip6_dst;
379 		switch (rh->ip6r_type) {
380 		case IPV6_RTHDR_TYPE_0:
381 			 rh0 = (struct ip6_rthdr0 *)rh;
382 			 addr = (struct in6_addr *)(rh0 + 1);
383 
384 			 /*
385 			  * construct a sockaddr_in6 form of
386 			  * the first hop.
387 			  *
388 			  * XXX: we may not have enough
389 			  * information about its scope zone;
390 			  * there is no standard API to pass
391 			  * the information from the
392 			  * application.
393 			  */
394 			 sockaddr_in6_init(&sa, addr, 0, 0, 0);
395 			 if ((error = sa6_embedscope(&sa,
396 			     ip6_use_defzone)) != 0) {
397 				 goto bad;
398 			 }
399 			 ip6->ip6_dst = sa.sin6_addr;
400 			 (void)memmove(&addr[0], &addr[1],
401 			     sizeof(struct in6_addr) *
402 			     (rh0->ip6r0_segleft - 1));
403 			 addr[rh0->ip6r0_segleft - 1] = finaldst;
404 			 /* XXX */
405 			 in6_clearscope(addr + rh0->ip6r0_segleft - 1);
406 			 break;
407 		default:	/* is it possible? */
408 			 error = EINVAL;
409 			 goto bad;
410 		}
411 	}
412 
413 	/* Source address validation */
414 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
415 	    (flags & IPV6_UNSPECSRC) == 0) {
416 		error = EOPNOTSUPP;
417 		IP6_STATINC(IP6_STAT_BADSCOPE);
418 		goto bad;
419 	}
420 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
421 		error = EOPNOTSUPP;
422 		IP6_STATINC(IP6_STAT_BADSCOPE);
423 		goto bad;
424 	}
425 
426 	IP6_STATINC(IP6_STAT_LOCALOUT);
427 
428 	/*
429 	 * Route packet.
430 	 */
431 	/* initialize cached route */
432 	if (ro == NULL) {
433 		ro = &ip6route;
434 	}
435 	ro_pmtu = ro;
436 	if (opt && opt->ip6po_rthdr)
437 		ro = &opt->ip6po_route;
438 
439  	/*
440 	 * if specified, try to fill in the traffic class field.
441 	 * do not override if a non-zero value is already set.
442 	 * we check the diffserv field and the ecn field separately.
443 	 */
444 	if (opt && opt->ip6po_tclass >= 0) {
445 		int mask = 0;
446 
447 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
448 			mask |= 0xfc;
449 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
450 			mask |= 0x03;
451 		if (mask != 0)
452 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
453 	}
454 
455 	/* fill in or override the hop limit field, if necessary. */
456 	if (opt && opt->ip6po_hlim != -1)
457 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
458 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
459 		if (im6o != NULL)
460 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
461 		else
462 			ip6->ip6_hlim = ip6_defmcasthlim;
463 	}
464 
465 #ifdef IPSEC
466 	if (needipsec) {
467 		s = splsoftnet();
468 		error = ipsec6_process_packet(m,sp->req);
469 
470 		/*
471 		 * Preserve KAME behaviour: ENOENT can be returned
472 		 * when an SA acquire is in progress.  Don't propagate
473 		 * this to user-level; it confuses applications.
474 		 * XXX this will go away when the SADB is redone.
475 		 */
476 		if (error == ENOENT)
477 			error = 0;
478 		splx(s);
479 		goto done;
480 	}
481 #endif /* IPSEC */
482 
483 
484 
485 	/* adjust pointer */
486 	ip6 = mtod(m, struct ip6_hdr *);
487 
488 	sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
489 	if ((error = in6_selectroute(&dst_sa, opt, im6o, ro,
490 	    &ifp, &rt, 0)) != 0) {
491 		if (ifp != NULL)
492 			in6_ifstat_inc(ifp, ifs6_out_discard);
493 		goto bad;
494 	}
495 	if (rt == NULL) {
496 		/*
497 		 * If in6_selectroute() does not return a route entry,
498 		 * dst may not have been updated.
499 		 */
500 		error = rtcache_setdst(ro, sin6tosa(&dst_sa));
501 		if (error) {
502 			goto bad;
503 		}
504 	}
505 
506 	/*
507 	 * then rt (for unicast) and ifp must be non-NULL valid values.
508 	 */
509 	if ((flags & IPV6_FORWARDING) == 0) {
510 		/* XXX: the FORWARDING flag can be set for mrouting. */
511 		in6_ifstat_inc(ifp, ifs6_out_request);
512 	}
513 	if (rt != NULL) {
514 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
515 		rt->rt_use++;
516 	}
517 
518 	/*
519 	 * The outgoing interface must be in the zone of source and
520 	 * destination addresses.  We should use ia_ifp to support the
521 	 * case of sending packets to an address of our own.
522 	 */
523 	if (ia != NULL && ia->ia_ifp)
524 		origifp = ia->ia_ifp;
525 	else
526 		origifp = ifp;
527 
528 	src0 = ip6->ip6_src;
529 	if (in6_setscope(&src0, origifp, &zone))
530 		goto badscope;
531 	sockaddr_in6_init(&src_sa, &ip6->ip6_src, 0, 0, 0);
532 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
533 		goto badscope;
534 
535 	dst0 = ip6->ip6_dst;
536 	if (in6_setscope(&dst0, origifp, &zone))
537 		goto badscope;
538 	/* re-initialize to be sure */
539 	sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
540 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id)
541 		goto badscope;
542 
543 	/* scope check is done. */
544 
545 	if (rt == NULL || IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
546 		if (dst == NULL)
547 			dst = satocsin6(rtcache_getdst(ro));
548 		KASSERT(dst != NULL);
549 	} else if (opt && rtcache_validate(&opt->ip6po_nextroute) != NULL) {
550 		/*
551 		 * The nexthop is explicitly specified by the
552 		 * application.  We assume the next hop is an IPv6
553 		 * address.
554 		 */
555 		dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
556 	} else if ((rt->rt_flags & RTF_GATEWAY))
557 		dst = (struct sockaddr_in6 *)rt->rt_gateway;
558 	else if (dst == NULL)
559 		dst = satocsin6(rtcache_getdst(ro));
560 
561 	/*
562 	 * XXXXXX: original code follows:
563 	 */
564 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
565 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
566 	else {
567 		struct	in6_multi *in6m;
568 
569 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
570 
571 		in6_ifstat_inc(ifp, ifs6_out_mcast);
572 
573 		/*
574 		 * Confirm that the outgoing interface supports multicast.
575 		 */
576 		if (!(ifp->if_flags & IFF_MULTICAST)) {
577 			IP6_STATINC(IP6_STAT_NOROUTE);
578 			in6_ifstat_inc(ifp, ifs6_out_discard);
579 			error = ENETUNREACH;
580 			goto bad;
581 		}
582 
583 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
584 		if (in6m != NULL &&
585 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
586 			/*
587 			 * If we belong to the destination multicast group
588 			 * on the outgoing interface, and the caller did not
589 			 * forbid loopback, loop back a copy.
590 			 */
591 			KASSERT(dst != NULL);
592 			ip6_mloopback(ifp, m, dst);
593 		} else {
594 			/*
595 			 * If we are acting as a multicast router, perform
596 			 * multicast forwarding as if the packet had just
597 			 * arrived on the interface to which we are about
598 			 * to send.  The multicast forwarding function
599 			 * recursively calls this function, using the
600 			 * IPV6_FORWARDING flag to prevent infinite recursion.
601 			 *
602 			 * Multicasts that are looped back by ip6_mloopback(),
603 			 * above, will be forwarded by the ip6_input() routine,
604 			 * if necessary.
605 			 */
606 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
607 				if (ip6_mforward(ip6, ifp, m) != 0) {
608 					m_freem(m);
609 					goto done;
610 				}
611 			}
612 		}
613 		/*
614 		 * Multicasts with a hoplimit of zero may be looped back,
615 		 * above, but must not be transmitted on a network.
616 		 * Also, multicasts addressed to the loopback interface
617 		 * are not sent -- the above call to ip6_mloopback() will
618 		 * loop back a copy if this host actually belongs to the
619 		 * destination group on the loopback interface.
620 		 */
621 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
622 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
623 			m_freem(m);
624 			goto done;
625 		}
626 	}
627 
628 	/*
629 	 * Fill the outgoing inteface to tell the upper layer
630 	 * to increment per-interface statistics.
631 	 */
632 	if (ifpp)
633 		*ifpp = ifp;
634 
635 	/* Determine path MTU. */
636 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
637 	    &alwaysfrag)) != 0)
638 		goto bad;
639 
640 	/*
641 	 * The caller of this function may specify to use the minimum MTU
642 	 * in some cases.
643 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
644 	 * setting.  The logic is a bit complicated; by default, unicast
645 	 * packets will follow path MTU while multicast packets will be sent at
646 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
647 	 * including unicast ones will be sent at the minimum MTU.  Multicast
648 	 * packets will always be sent at the minimum MTU unless
649 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
650 	 * See RFC 3542 for more details.
651 	 */
652 	if (mtu > IPV6_MMTU) {
653 		if ((flags & IPV6_MINMTU))
654 			mtu = IPV6_MMTU;
655 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
656 			mtu = IPV6_MMTU;
657 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
658 			 (opt == NULL ||
659 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
660 			mtu = IPV6_MMTU;
661 		}
662 	}
663 
664 	/*
665 	 * clear embedded scope identifiers if necessary.
666 	 * in6_clearscope will touch the addresses only when necessary.
667 	 */
668 	in6_clearscope(&ip6->ip6_src);
669 	in6_clearscope(&ip6->ip6_dst);
670 
671 	/*
672 	 * If the outgoing packet contains a hop-by-hop options header,
673 	 * it must be examined and processed even by the source node.
674 	 * (RFC 2460, section 4.)
675 	 */
676 	if (ip6->ip6_nxt == IPV6_HOPOPTS) {
677 		u_int32_t dummy1; /* XXX unused */
678 		u_int32_t dummy2; /* XXX unused */
679 		int hoff = sizeof(struct ip6_hdr);
680 
681 		if (ip6_hopopts_input(&dummy1, &dummy2, &m, &hoff)) {
682 			/* m was already freed at this point */
683 			error = EINVAL;/* better error? */
684 			goto done;
685 		}
686 
687 		ip6 = mtod(m, struct ip6_hdr *);
688 	}
689 
690 	/*
691 	 * Run through list of hooks for output packets.
692 	 */
693 	if ((error = pfil_run_hooks(inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0)
694 		goto done;
695 	if (m == NULL)
696 		goto done;
697 	ip6 = mtod(m, struct ip6_hdr *);
698 
699 	/*
700 	 * Send the packet to the outgoing interface.
701 	 * If necessary, do IPv6 fragmentation before sending.
702 	 *
703 	 * the logic here is rather complex:
704 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
705 	 * 1-a:	send as is if tlen <= path mtu
706 	 * 1-b:	fragment if tlen > path mtu
707 	 *
708 	 * 2: if user asks us not to fragment (dontfrag == 1)
709 	 * 2-a:	send as is if tlen <= interface mtu
710 	 * 2-b:	error if tlen > interface mtu
711 	 *
712 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
713 	 *	always fragment
714 	 *
715 	 * 4: if dontfrag == 1 && alwaysfrag == 1
716 	 *	error, as we cannot handle this conflicting request
717 	 */
718 	tlen = m->m_pkthdr.len;
719 	tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0;
720 	if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
721 		dontfrag = 1;
722 	else
723 		dontfrag = 0;
724 
725 	if (dontfrag && alwaysfrag) {	/* case 4 */
726 		/* conflicting request - can't transmit */
727 		error = EMSGSIZE;
728 		goto bad;
729 	}
730 	if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) {	/* case 2-b */
731 		/*
732 		 * Even if the DONTFRAG option is specified, we cannot send the
733 		 * packet when the data length is larger than the MTU of the
734 		 * outgoing interface.
735 		 * Notify the error by sending IPV6_PATHMTU ancillary data as
736 		 * well as returning an error code (the latter is not described
737 		 * in the API spec.)
738 		 */
739 		u_int32_t mtu32;
740 		struct ip6ctlparam ip6cp;
741 
742 		mtu32 = (u_int32_t)mtu;
743 		memset(&ip6cp, 0, sizeof(ip6cp));
744 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
745 		pfctlinput2(PRC_MSGSIZE,
746 		    rtcache_getdst(ro_pmtu), &ip6cp);
747 
748 		error = EMSGSIZE;
749 		goto bad;
750 	}
751 
752 	/*
753 	 * transmit packet without fragmentation
754 	 */
755 	if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) {
756 		/* case 1-a and 2-a */
757 		struct in6_ifaddr *ia6;
758 		int sw_csum;
759 
760 		ip6 = mtod(m, struct ip6_hdr *);
761 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
762 		if (ia6) {
763 			/* Record statistics for this interface address. */
764 			ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len;
765 		}
766 
767 		sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
768 		if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
769 			if (IN6_NEED_CHECKSUM(ifp,
770 			    sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) {
771 				in6_delayed_cksum(m);
772 			}
773 			m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
774 		}
775 
776 		KASSERT(dst != NULL);
777 		if (__predict_true(!tso ||
778 		    (ifp->if_capenable & IFCAP_TSOv6) != 0)) {
779 			error = nd6_output(ifp, origifp, m, dst, rt);
780 		} else {
781 			error = ip6_tso_output(ifp, origifp, m, dst, rt);
782 		}
783 		goto done;
784 	}
785 
786 	if (tso) {
787 		error = EINVAL; /* XXX */
788 		goto bad;
789 	}
790 
791 	/*
792 	 * try to fragment the packet.  case 1-b and 3
793 	 */
794 	if (mtu < IPV6_MMTU) {
795 		/* path MTU cannot be less than IPV6_MMTU */
796 		error = EMSGSIZE;
797 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
798 		goto bad;
799 	} else if (ip6->ip6_plen == 0) {
800 		/* jumbo payload cannot be fragmented */
801 		error = EMSGSIZE;
802 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
803 		goto bad;
804 	} else {
805 		struct mbuf **mnext, *m_frgpart;
806 		struct ip6_frag *ip6f;
807 		u_int32_t id = htonl(ip6_randomid());
808 		u_char nextproto;
809 #if 0				/* see below */
810 		struct ip6ctlparam ip6cp;
811 		u_int32_t mtu32;
812 #endif
813 
814 		/*
815 		 * Too large for the destination or interface;
816 		 * fragment if possible.
817 		 * Must be able to put at least 8 bytes per fragment.
818 		 */
819 		hlen = unfragpartlen;
820 		if (mtu > IPV6_MAXPACKET)
821 			mtu = IPV6_MAXPACKET;
822 
823 #if 0
824 		/*
825 		 * It is believed this code is a leftover from the
826 		 * development of the IPV6_RECVPATHMTU sockopt and
827 		 * associated work to implement RFC3542.
828 		 * It's not entirely clear what the intent of the API
829 		 * is at this point, so disable this code for now.
830 		 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG
831 		 * will send notifications if the application requests.
832 		 */
833 
834 		/* Notify a proper path MTU to applications. */
835 		mtu32 = (u_int32_t)mtu;
836 		memset(&ip6cp, 0, sizeof(ip6cp));
837 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
838 		pfctlinput2(PRC_MSGSIZE,
839 		    rtcache_getdst(ro_pmtu), &ip6cp);
840 #endif
841 
842 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
843 		if (len < 8) {
844 			error = EMSGSIZE;
845 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
846 			goto bad;
847 		}
848 
849 		mnext = &m->m_nextpkt;
850 
851 		/*
852 		 * Change the next header field of the last header in the
853 		 * unfragmentable part.
854 		 */
855 		if (exthdrs.ip6e_rthdr) {
856 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
857 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
858 		} else if (exthdrs.ip6e_dest1) {
859 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
860 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
861 		} else if (exthdrs.ip6e_hbh) {
862 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
863 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
864 		} else {
865 			nextproto = ip6->ip6_nxt;
866 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
867 		}
868 
869 		if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6))
870 		    != 0) {
871 			if (IN6_NEED_CHECKSUM(ifp,
872 			    m->m_pkthdr.csum_flags &
873 			    (M_CSUM_UDPv6|M_CSUM_TCPv6))) {
874 				in6_delayed_cksum(m);
875 			}
876 			m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
877 		}
878 
879 		/*
880 		 * Loop through length of segment after first fragment,
881 		 * make new header and copy data of each part and link onto
882 		 * chain.
883 		 */
884 		m0 = m;
885 		for (off = hlen; off < tlen; off += len) {
886 			struct mbuf *mlast;
887 
888 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
889 			if (!m) {
890 				error = ENOBUFS;
891 				IP6_STATINC(IP6_STAT_ODROPPED);
892 				goto sendorfree;
893 			}
894 			m->m_pkthdr.rcvif = NULL;
895 			m->m_flags = m0->m_flags & M_COPYFLAGS;
896 			*mnext = m;
897 			mnext = &m->m_nextpkt;
898 			m->m_data += max_linkhdr;
899 			mhip6 = mtod(m, struct ip6_hdr *);
900 			*mhip6 = *ip6;
901 			m->m_len = sizeof(*mhip6);
902 			/*
903 			 * ip6f must be valid if error is 0.  But how
904 			 * can a compiler be expected to infer this?
905 			 */
906 			ip6f = NULL;
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 IPSEC
974 	if (sp != NULL)
975 		KEY_FREESP(&sp);
976 #endif /* 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  * *frghdrp will not be read, and it is guaranteed that either an
1151  * error is returned or that *frghdrp will point to space allocated
1152  * for the fragment header.
1153  */
1154 static int
1155 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1156 	struct ip6_frag **frghdrp)
1157 {
1158 	struct mbuf *n, *mlast;
1159 
1160 	if (hlen > sizeof(struct ip6_hdr)) {
1161 		n = m_copym(m0, sizeof(struct ip6_hdr),
1162 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1163 		if (n == 0)
1164 			return (ENOBUFS);
1165 		m->m_next = n;
1166 	} else
1167 		n = m;
1168 
1169 	/* Search for the last mbuf of unfragmentable part. */
1170 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1171 		;
1172 
1173 	if ((mlast->m_flags & M_EXT) == 0 &&
1174 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1175 		/* use the trailing space of the last mbuf for the fragment hdr */
1176 		*frghdrp = (struct ip6_frag *)(mtod(mlast, char *) +
1177 		    mlast->m_len);
1178 		mlast->m_len += sizeof(struct ip6_frag);
1179 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1180 	} else {
1181 		/* allocate a new mbuf for the fragment header */
1182 		struct mbuf *mfrg;
1183 
1184 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1185 		if (mfrg == 0)
1186 			return (ENOBUFS);
1187 		mfrg->m_len = sizeof(struct ip6_frag);
1188 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1189 		mlast->m_next = mfrg;
1190 	}
1191 
1192 	return (0);
1193 }
1194 
1195 static int
1196 ip6_getpmtu(struct route *ro_pmtu, struct route *ro, struct ifnet *ifp,
1197     const struct in6_addr *dst, u_long *mtup, int *alwaysfragp)
1198 {
1199 	struct rtentry *rt;
1200 	u_int32_t mtu = 0;
1201 	int alwaysfrag = 0;
1202 	int error = 0;
1203 
1204 	if (ro_pmtu != ro) {
1205 		union {
1206 			struct sockaddr		dst;
1207 			struct sockaddr_in6	dst6;
1208 		} u;
1209 
1210 		/* The first hop and the final destination may differ. */
1211 		sockaddr_in6_init(&u.dst6, dst, 0, 0, 0);
1212 		rt = rtcache_lookup(ro_pmtu, &u.dst);
1213 	} else
1214 		rt = rtcache_validate(ro_pmtu);
1215 	if (rt != NULL) {
1216 		u_int32_t ifmtu;
1217 
1218 		if (ifp == NULL)
1219 			ifp = rt->rt_ifp;
1220 		ifmtu = IN6_LINKMTU(ifp);
1221 		mtu = rt->rt_rmx.rmx_mtu;
1222 		if (mtu == 0)
1223 			mtu = ifmtu;
1224 		else if (mtu < IPV6_MMTU) {
1225 			/*
1226 			 * RFC2460 section 5, last paragraph:
1227 			 * if we record ICMPv6 too big message with
1228 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1229 			 * or smaller, with fragment header attached.
1230 			 * (fragment header is needed regardless from the
1231 			 * packet size, for translators to identify packets)
1232 			 */
1233 			alwaysfrag = 1;
1234 			mtu = IPV6_MMTU;
1235 		} else if (mtu > ifmtu) {
1236 			/*
1237 			 * The MTU on the route is larger than the MTU on
1238 			 * the interface!  This shouldn't happen, unless the
1239 			 * MTU of the interface has been changed after the
1240 			 * interface was brought up.  Change the MTU in the
1241 			 * route to match the interface MTU (as long as the
1242 			 * field isn't locked).
1243 			 */
1244 			mtu = ifmtu;
1245 			if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
1246 				rt->rt_rmx.rmx_mtu = mtu;
1247 		}
1248 	} else if (ifp) {
1249 		mtu = IN6_LINKMTU(ifp);
1250 	} else
1251 		error = EHOSTUNREACH; /* XXX */
1252 
1253 	*mtup = mtu;
1254 	if (alwaysfragp)
1255 		*alwaysfragp = alwaysfrag;
1256 	return (error);
1257 }
1258 
1259 /*
1260  * IP6 socket option processing.
1261  */
1262 int
1263 ip6_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1264 {
1265 	int optdatalen, uproto;
1266 	void *optdata;
1267 	struct in6pcb *in6p = sotoin6pcb(so);
1268 	int error, optval;
1269 	int level, optname;
1270 
1271 	KASSERT(sopt != NULL);
1272 
1273 	level = sopt->sopt_level;
1274 	optname = sopt->sopt_name;
1275 
1276 	error = optval = 0;
1277 	uproto = (int)so->so_proto->pr_protocol;
1278 
1279 	if (level != IPPROTO_IPV6) {
1280 		return ENOPROTOOPT;
1281 	}
1282 	switch (op) {
1283 	case PRCO_SETOPT:
1284 		switch (optname) {
1285 #ifdef RFC2292
1286 		case IPV6_2292PKTOPTIONS:
1287 			error = ip6_pcbopts(&in6p->in6p_outputopts, so, sopt);
1288 			break;
1289 #endif
1290 
1291 		/*
1292 		 * Use of some Hop-by-Hop options or some
1293 		 * Destination options, might require special
1294 		 * privilege.  That is, normal applications
1295 		 * (without special privilege) might be forbidden
1296 		 * from setting certain options in outgoing packets,
1297 		 * and might never see certain options in received
1298 		 * packets. [RFC 2292 Section 6]
1299 		 * KAME specific note:
1300 		 *  KAME prevents non-privileged users from sending or
1301 		 *  receiving ANY hbh/dst options in order to avoid
1302 		 *  overhead of parsing options in the kernel.
1303 		 */
1304 		case IPV6_RECVHOPOPTS:
1305 		case IPV6_RECVDSTOPTS:
1306 		case IPV6_RECVRTHDRDSTOPTS:
1307 			error = kauth_authorize_network(kauth_cred_get(),
1308 			    KAUTH_NETWORK_IPV6, KAUTH_REQ_NETWORK_IPV6_HOPBYHOP,
1309 			    NULL, NULL, NULL);
1310 			if (error)
1311 				break;
1312 			/* FALLTHROUGH */
1313 		case IPV6_UNICAST_HOPS:
1314 		case IPV6_HOPLIMIT:
1315 		case IPV6_FAITH:
1316 
1317 		case IPV6_RECVPKTINFO:
1318 		case IPV6_RECVHOPLIMIT:
1319 		case IPV6_RECVRTHDR:
1320 		case IPV6_RECVPATHMTU:
1321 		case IPV6_RECVTCLASS:
1322 		case IPV6_V6ONLY:
1323 			error = sockopt_getint(sopt, &optval);
1324 			if (error)
1325 				break;
1326 			switch (optname) {
1327 			case IPV6_UNICAST_HOPS:
1328 				if (optval < -1 || optval >= 256)
1329 					error = EINVAL;
1330 				else {
1331 					/* -1 = kernel default */
1332 					in6p->in6p_hops = optval;
1333 				}
1334 				break;
1335 #define OPTSET(bit) \
1336 do { \
1337 if (optval) \
1338 	in6p->in6p_flags |= (bit); \
1339 else \
1340 	in6p->in6p_flags &= ~(bit); \
1341 } while (/*CONSTCOND*/ 0)
1342 
1343 #ifdef RFC2292
1344 #define OPTSET2292(bit) 			\
1345 do { 						\
1346 in6p->in6p_flags |= IN6P_RFC2292; 	\
1347 if (optval) 				\
1348 	in6p->in6p_flags |= (bit); 	\
1349 else 					\
1350 	in6p->in6p_flags &= ~(bit); 	\
1351 } while (/*CONSTCOND*/ 0)
1352 #endif
1353 
1354 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1355 
1356 			case IPV6_RECVPKTINFO:
1357 #ifdef RFC2292
1358 				/* cannot mix with RFC2292 */
1359 				if (OPTBIT(IN6P_RFC2292)) {
1360 					error = EINVAL;
1361 					break;
1362 				}
1363 #endif
1364 				OPTSET(IN6P_PKTINFO);
1365 				break;
1366 
1367 			case IPV6_HOPLIMIT:
1368 			{
1369 				struct ip6_pktopts **optp;
1370 
1371 #ifdef RFC2292
1372 				/* cannot mix with RFC2292 */
1373 				if (OPTBIT(IN6P_RFC2292)) {
1374 					error = EINVAL;
1375 					break;
1376 				}
1377 #endif
1378 				optp = &in6p->in6p_outputopts;
1379 				error = ip6_pcbopt(IPV6_HOPLIMIT,
1380 						   (u_char *)&optval,
1381 						   sizeof(optval),
1382 						   optp,
1383 						   kauth_cred_get(), uproto);
1384 				break;
1385 			}
1386 
1387 			case IPV6_RECVHOPLIMIT:
1388 #ifdef RFC2292
1389 				/* cannot mix with RFC2292 */
1390 				if (OPTBIT(IN6P_RFC2292)) {
1391 					error = EINVAL;
1392 					break;
1393 				}
1394 #endif
1395 				OPTSET(IN6P_HOPLIMIT);
1396 				break;
1397 
1398 			case IPV6_RECVHOPOPTS:
1399 #ifdef RFC2292
1400 				/* cannot mix with RFC2292 */
1401 				if (OPTBIT(IN6P_RFC2292)) {
1402 					error = EINVAL;
1403 					break;
1404 				}
1405 #endif
1406 				OPTSET(IN6P_HOPOPTS);
1407 				break;
1408 
1409 			case IPV6_RECVDSTOPTS:
1410 #ifdef RFC2292
1411 				/* cannot mix with RFC2292 */
1412 				if (OPTBIT(IN6P_RFC2292)) {
1413 					error = EINVAL;
1414 					break;
1415 				}
1416 #endif
1417 				OPTSET(IN6P_DSTOPTS);
1418 				break;
1419 
1420 			case IPV6_RECVRTHDRDSTOPTS:
1421 #ifdef RFC2292
1422 				/* cannot mix with RFC2292 */
1423 				if (OPTBIT(IN6P_RFC2292)) {
1424 					error = EINVAL;
1425 					break;
1426 				}
1427 #endif
1428 				OPTSET(IN6P_RTHDRDSTOPTS);
1429 				break;
1430 
1431 			case IPV6_RECVRTHDR:
1432 #ifdef RFC2292
1433 				/* cannot mix with RFC2292 */
1434 				if (OPTBIT(IN6P_RFC2292)) {
1435 					error = EINVAL;
1436 					break;
1437 				}
1438 #endif
1439 				OPTSET(IN6P_RTHDR);
1440 				break;
1441 
1442 			case IPV6_FAITH:
1443 				OPTSET(IN6P_FAITH);
1444 				break;
1445 
1446 			case IPV6_RECVPATHMTU:
1447 				/*
1448 				 * We ignore this option for TCP
1449 				 * sockets.
1450 				 * (RFC3542 leaves this case
1451 				 * unspecified.)
1452 				 */
1453 				if (uproto != IPPROTO_TCP)
1454 					OPTSET(IN6P_MTU);
1455 				break;
1456 
1457 			case IPV6_V6ONLY:
1458 				/*
1459 				 * make setsockopt(IPV6_V6ONLY)
1460 				 * available only prior to bind(2).
1461 				 * see ipng mailing list, Jun 22 2001.
1462 				 */
1463 				if (in6p->in6p_lport ||
1464 				    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1465 					error = EINVAL;
1466 					break;
1467 				}
1468 #ifdef INET6_BINDV6ONLY
1469 				if (!optval)
1470 					error = EINVAL;
1471 #else
1472 				OPTSET(IN6P_IPV6_V6ONLY);
1473 #endif
1474 				break;
1475 			case IPV6_RECVTCLASS:
1476 #ifdef RFC2292
1477 				/* cannot mix with RFC2292 XXX */
1478 				if (OPTBIT(IN6P_RFC2292)) {
1479 					error = EINVAL;
1480 					break;
1481 				}
1482 #endif
1483 				OPTSET(IN6P_TCLASS);
1484 				break;
1485 
1486 			}
1487 			break;
1488 
1489 		case IPV6_OTCLASS:
1490 		{
1491 			struct ip6_pktopts **optp;
1492 			u_int8_t tclass;
1493 
1494 			error = sockopt_get(sopt, &tclass, sizeof(tclass));
1495 			if (error)
1496 				break;
1497 			optp = &in6p->in6p_outputopts;
1498 			error = ip6_pcbopt(optname,
1499 					   (u_char *)&tclass,
1500 					   sizeof(tclass),
1501 					   optp,
1502 					   kauth_cred_get(), uproto);
1503 			break;
1504 		}
1505 
1506 		case IPV6_TCLASS:
1507 		case IPV6_DONTFRAG:
1508 		case IPV6_USE_MIN_MTU:
1509 			error = sockopt_getint(sopt, &optval);
1510 			if (error)
1511 				break;
1512 			{
1513 				struct ip6_pktopts **optp;
1514 				optp = &in6p->in6p_outputopts;
1515 				error = ip6_pcbopt(optname,
1516 						   (u_char *)&optval,
1517 						   sizeof(optval),
1518 						   optp,
1519 						   kauth_cred_get(), uproto);
1520 				break;
1521 			}
1522 
1523 #ifdef RFC2292
1524 		case IPV6_2292PKTINFO:
1525 		case IPV6_2292HOPLIMIT:
1526 		case IPV6_2292HOPOPTS:
1527 		case IPV6_2292DSTOPTS:
1528 		case IPV6_2292RTHDR:
1529 			/* RFC 2292 */
1530 			error = sockopt_getint(sopt, &optval);
1531 			if (error)
1532 				break;
1533 
1534 			switch (optname) {
1535 			case IPV6_2292PKTINFO:
1536 				OPTSET2292(IN6P_PKTINFO);
1537 				break;
1538 			case IPV6_2292HOPLIMIT:
1539 				OPTSET2292(IN6P_HOPLIMIT);
1540 				break;
1541 			case IPV6_2292HOPOPTS:
1542 				/*
1543 				 * Check super-user privilege.
1544 				 * See comments for IPV6_RECVHOPOPTS.
1545 				 */
1546 				error =
1547 				    kauth_authorize_network(kauth_cred_get(),
1548 				    KAUTH_NETWORK_IPV6,
1549 				    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL,
1550 				    NULL, NULL);
1551 				if (error)
1552 					return (error);
1553 				OPTSET2292(IN6P_HOPOPTS);
1554 				break;
1555 			case IPV6_2292DSTOPTS:
1556 				error =
1557 				    kauth_authorize_network(kauth_cred_get(),
1558 				    KAUTH_NETWORK_IPV6,
1559 				    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL,
1560 				    NULL, NULL);
1561 				if (error)
1562 					return (error);
1563 				OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1564 				break;
1565 			case IPV6_2292RTHDR:
1566 				OPTSET2292(IN6P_RTHDR);
1567 				break;
1568 			}
1569 			break;
1570 #endif
1571 		case IPV6_PKTINFO:
1572 		case IPV6_HOPOPTS:
1573 		case IPV6_RTHDR:
1574 		case IPV6_DSTOPTS:
1575 		case IPV6_RTHDRDSTOPTS:
1576 		case IPV6_NEXTHOP: {
1577 			/* new advanced API (RFC3542) */
1578 			void *optbuf;
1579 			int optbuflen;
1580 			struct ip6_pktopts **optp;
1581 
1582 #ifdef RFC2292
1583 			/* cannot mix with RFC2292 */
1584 			if (OPTBIT(IN6P_RFC2292)) {
1585 				error = EINVAL;
1586 				break;
1587 			}
1588 #endif
1589 
1590 			optbuflen = sopt->sopt_size;
1591 			optbuf = malloc(optbuflen, M_IP6OPT, M_NOWAIT);
1592 			if (optbuf == NULL) {
1593 				error = ENOBUFS;
1594 				break;
1595 			}
1596 
1597 			error = sockopt_get(sopt, optbuf, optbuflen);
1598 			if (error) {
1599 				free(optbuf, M_IP6OPT);
1600 				break;
1601 			}
1602 			optp = &in6p->in6p_outputopts;
1603 			error = ip6_pcbopt(optname, optbuf, optbuflen,
1604 			    optp, kauth_cred_get(), uproto);
1605 			break;
1606 			}
1607 #undef OPTSET
1608 
1609 		case IPV6_MULTICAST_IF:
1610 		case IPV6_MULTICAST_HOPS:
1611 		case IPV6_MULTICAST_LOOP:
1612 		case IPV6_JOIN_GROUP:
1613 		case IPV6_LEAVE_GROUP:
1614 			error = ip6_setmoptions(sopt, &in6p->in6p_moptions);
1615 			break;
1616 
1617 		case IPV6_PORTRANGE:
1618 			error = sockopt_getint(sopt, &optval);
1619 			if (error)
1620 				break;
1621 
1622 			switch (optval) {
1623 			case IPV6_PORTRANGE_DEFAULT:
1624 				in6p->in6p_flags &= ~(IN6P_LOWPORT);
1625 				in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1626 				break;
1627 
1628 			case IPV6_PORTRANGE_HIGH:
1629 				in6p->in6p_flags &= ~(IN6P_LOWPORT);
1630 				in6p->in6p_flags |= IN6P_HIGHPORT;
1631 				break;
1632 
1633 			case IPV6_PORTRANGE_LOW:
1634 				in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1635 				in6p->in6p_flags |= IN6P_LOWPORT;
1636 				break;
1637 
1638 			default:
1639 				error = EINVAL;
1640 				break;
1641 			}
1642 			break;
1643 
1644 		case IPV6_PORTALGO:
1645 			error = sockopt_getint(sopt, &optval);
1646 			if (error)
1647 				break;
1648 
1649 			error = portalgo_algo_index_select(
1650 			    (struct inpcb_hdr *)in6p, optval);
1651 			break;
1652 
1653 #if defined(IPSEC)
1654 		case IPV6_IPSEC_POLICY:
1655 			error = ipsec6_set_policy(in6p, optname,
1656 			    sopt->sopt_data, sopt->sopt_size, kauth_cred_get());
1657 			break;
1658 #endif /* IPSEC */
1659 
1660 		default:
1661 			error = ENOPROTOOPT;
1662 			break;
1663 		}
1664 		break;
1665 
1666 	case PRCO_GETOPT:
1667 		switch (optname) {
1668 #ifdef RFC2292
1669 		case IPV6_2292PKTOPTIONS:
1670 			/*
1671 			 * RFC3542 (effectively) deprecated the
1672 			 * semantics of the 2292-style pktoptions.
1673 			 * Since it was not reliable in nature (i.e.,
1674 			 * applications had to expect the lack of some
1675 			 * information after all), it would make sense
1676 			 * to simplify this part by always returning
1677 			 * empty data.
1678 			 */
1679 			break;
1680 #endif
1681 
1682 		case IPV6_RECVHOPOPTS:
1683 		case IPV6_RECVDSTOPTS:
1684 		case IPV6_RECVRTHDRDSTOPTS:
1685 		case IPV6_UNICAST_HOPS:
1686 		case IPV6_RECVPKTINFO:
1687 		case IPV6_RECVHOPLIMIT:
1688 		case IPV6_RECVRTHDR:
1689 		case IPV6_RECVPATHMTU:
1690 
1691 		case IPV6_FAITH:
1692 		case IPV6_V6ONLY:
1693 		case IPV6_PORTRANGE:
1694 		case IPV6_RECVTCLASS:
1695 			switch (optname) {
1696 
1697 			case IPV6_RECVHOPOPTS:
1698 				optval = OPTBIT(IN6P_HOPOPTS);
1699 				break;
1700 
1701 			case IPV6_RECVDSTOPTS:
1702 				optval = OPTBIT(IN6P_DSTOPTS);
1703 				break;
1704 
1705 			case IPV6_RECVRTHDRDSTOPTS:
1706 				optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1707 				break;
1708 
1709 			case IPV6_UNICAST_HOPS:
1710 				optval = in6p->in6p_hops;
1711 				break;
1712 
1713 			case IPV6_RECVPKTINFO:
1714 				optval = OPTBIT(IN6P_PKTINFO);
1715 				break;
1716 
1717 			case IPV6_RECVHOPLIMIT:
1718 				optval = OPTBIT(IN6P_HOPLIMIT);
1719 				break;
1720 
1721 			case IPV6_RECVRTHDR:
1722 				optval = OPTBIT(IN6P_RTHDR);
1723 				break;
1724 
1725 			case IPV6_RECVPATHMTU:
1726 				optval = OPTBIT(IN6P_MTU);
1727 				break;
1728 
1729 			case IPV6_FAITH:
1730 				optval = OPTBIT(IN6P_FAITH);
1731 				break;
1732 
1733 			case IPV6_V6ONLY:
1734 				optval = OPTBIT(IN6P_IPV6_V6ONLY);
1735 				break;
1736 
1737 			case IPV6_PORTRANGE:
1738 			    {
1739 				int flags;
1740 				flags = in6p->in6p_flags;
1741 				if (flags & IN6P_HIGHPORT)
1742 					optval = IPV6_PORTRANGE_HIGH;
1743 				else if (flags & IN6P_LOWPORT)
1744 					optval = IPV6_PORTRANGE_LOW;
1745 				else
1746 					optval = 0;
1747 				break;
1748 			    }
1749 			case IPV6_RECVTCLASS:
1750 				optval = OPTBIT(IN6P_TCLASS);
1751 				break;
1752 
1753 			}
1754 			if (error)
1755 				break;
1756 			error = sockopt_setint(sopt, optval);
1757 			break;
1758 
1759 		case IPV6_PATHMTU:
1760 		    {
1761 			u_long pmtu = 0;
1762 			struct ip6_mtuinfo mtuinfo;
1763 			struct route *ro = &in6p->in6p_route;
1764 
1765 			if (!(so->so_state & SS_ISCONNECTED))
1766 				return (ENOTCONN);
1767 			/*
1768 			 * XXX: we dot not consider the case of source
1769 			 * routing, or optional information to specify
1770 			 * the outgoing interface.
1771 			 */
1772 			error = ip6_getpmtu(ro, NULL, NULL,
1773 			    &in6p->in6p_faddr, &pmtu, NULL);
1774 			if (error)
1775 				break;
1776 			if (pmtu > IPV6_MAXPACKET)
1777 				pmtu = IPV6_MAXPACKET;
1778 
1779 			memset(&mtuinfo, 0, sizeof(mtuinfo));
1780 			mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1781 			optdata = (void *)&mtuinfo;
1782 			optdatalen = sizeof(mtuinfo);
1783 			if (optdatalen > MCLBYTES)
1784 				return (EMSGSIZE); /* XXX */
1785 			error = sockopt_set(sopt, optdata, optdatalen);
1786 			break;
1787 		    }
1788 
1789 #ifdef RFC2292
1790 		case IPV6_2292PKTINFO:
1791 		case IPV6_2292HOPLIMIT:
1792 		case IPV6_2292HOPOPTS:
1793 		case IPV6_2292RTHDR:
1794 		case IPV6_2292DSTOPTS:
1795 			switch (optname) {
1796 			case IPV6_2292PKTINFO:
1797 				optval = OPTBIT(IN6P_PKTINFO);
1798 				break;
1799 			case IPV6_2292HOPLIMIT:
1800 				optval = OPTBIT(IN6P_HOPLIMIT);
1801 				break;
1802 			case IPV6_2292HOPOPTS:
1803 				optval = OPTBIT(IN6P_HOPOPTS);
1804 				break;
1805 			case IPV6_2292RTHDR:
1806 				optval = OPTBIT(IN6P_RTHDR);
1807 				break;
1808 			case IPV6_2292DSTOPTS:
1809 				optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1810 				break;
1811 			}
1812 			error = sockopt_setint(sopt, optval);
1813 			break;
1814 #endif
1815 		case IPV6_PKTINFO:
1816 		case IPV6_HOPOPTS:
1817 		case IPV6_RTHDR:
1818 		case IPV6_DSTOPTS:
1819 		case IPV6_RTHDRDSTOPTS:
1820 		case IPV6_NEXTHOP:
1821 		case IPV6_OTCLASS:
1822 		case IPV6_TCLASS:
1823 		case IPV6_DONTFRAG:
1824 		case IPV6_USE_MIN_MTU:
1825 			error = ip6_getpcbopt(in6p->in6p_outputopts,
1826 			    optname, sopt);
1827 			break;
1828 
1829 		case IPV6_MULTICAST_IF:
1830 		case IPV6_MULTICAST_HOPS:
1831 		case IPV6_MULTICAST_LOOP:
1832 		case IPV6_JOIN_GROUP:
1833 		case IPV6_LEAVE_GROUP:
1834 			error = ip6_getmoptions(sopt, in6p->in6p_moptions);
1835 			break;
1836 
1837 		case IPV6_PORTALGO:
1838 			optval = ((struct inpcb_hdr *)in6p)->inph_portalgo;
1839 			error = sockopt_setint(sopt, optval);
1840 			break;
1841 
1842 #if defined(IPSEC)
1843 		case IPV6_IPSEC_POLICY:
1844 		    {
1845 			struct mbuf *m = NULL;
1846 
1847 			/* XXX this will return EINVAL as sopt is empty */
1848 			error = ipsec6_get_policy(in6p, sopt->sopt_data,
1849 			    sopt->sopt_size, &m);
1850 			if (!error)
1851 				error = sockopt_setmbuf(sopt, m);
1852 
1853 			break;
1854 		    }
1855 #endif /* IPSEC */
1856 
1857 		default:
1858 			error = ENOPROTOOPT;
1859 			break;
1860 		}
1861 		break;
1862 	}
1863 	return (error);
1864 }
1865 
1866 int
1867 ip6_raw_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1868 {
1869 	int error = 0, optval;
1870 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
1871 	struct in6pcb *in6p = sotoin6pcb(so);
1872 	int level, optname;
1873 
1874 	KASSERT(sopt != NULL);
1875 
1876 	level = sopt->sopt_level;
1877 	optname = sopt->sopt_name;
1878 
1879 	if (level != IPPROTO_IPV6) {
1880 		return ENOPROTOOPT;
1881 	}
1882 
1883 	switch (optname) {
1884 	case IPV6_CHECKSUM:
1885 		/*
1886 		 * For ICMPv6 sockets, no modification allowed for checksum
1887 		 * offset, permit "no change" values to help existing apps.
1888 		 *
1889 		 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM
1890 		 * for an ICMPv6 socket will fail."  The current
1891 		 * behavior does not meet RFC3542.
1892 		 */
1893 		switch (op) {
1894 		case PRCO_SETOPT:
1895 			error = sockopt_getint(sopt, &optval);
1896 			if (error)
1897 				break;
1898 			if ((optval % 2) != 0) {
1899 				/* the API assumes even offset values */
1900 				error = EINVAL;
1901 			} else if (so->so_proto->pr_protocol ==
1902 			    IPPROTO_ICMPV6) {
1903 				if (optval != icmp6off)
1904 					error = EINVAL;
1905 			} else
1906 				in6p->in6p_cksum = optval;
1907 			break;
1908 
1909 		case PRCO_GETOPT:
1910 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
1911 				optval = icmp6off;
1912 			else
1913 				optval = in6p->in6p_cksum;
1914 
1915 			error = sockopt_setint(sopt, optval);
1916 			break;
1917 
1918 		default:
1919 			error = EINVAL;
1920 			break;
1921 		}
1922 		break;
1923 
1924 	default:
1925 		error = ENOPROTOOPT;
1926 		break;
1927 	}
1928 
1929 	return (error);
1930 }
1931 
1932 #ifdef RFC2292
1933 /*
1934  * Set up IP6 options in pcb for insertion in output packets or
1935  * specifying behavior of outgoing packets.
1936  */
1937 static int
1938 ip6_pcbopts(struct ip6_pktopts **pktopt, struct socket *so,
1939     struct sockopt *sopt)
1940 {
1941 	struct ip6_pktopts *opt = *pktopt;
1942 	struct mbuf *m;
1943 	int error = 0;
1944 
1945 	/* turn off any old options. */
1946 	if (opt) {
1947 #ifdef DIAGNOSTIC
1948 	    if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
1949 		opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
1950 		opt->ip6po_rhinfo.ip6po_rhi_rthdr)
1951 		    printf("ip6_pcbopts: all specified options are cleared.\n");
1952 #endif
1953 		ip6_clearpktopts(opt, -1);
1954 	} else {
1955 		opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
1956 		if (opt == NULL)
1957 			return (ENOBUFS);
1958 	}
1959 	*pktopt = NULL;
1960 
1961 	if (sopt == NULL || sopt->sopt_size == 0) {
1962 		/*
1963 		 * Only turning off any previous options, regardless of
1964 		 * whether the opt is just created or given.
1965 		 */
1966 		free(opt, M_IP6OPT);
1967 		return (0);
1968 	}
1969 
1970 	/*  set options specified by user. */
1971 	m = sockopt_getmbuf(sopt);
1972 	if (m == NULL) {
1973 		free(opt, M_IP6OPT);
1974 		return (ENOBUFS);
1975 	}
1976 
1977 	error = ip6_setpktopts(m, opt, NULL, kauth_cred_get(),
1978 	    so->so_proto->pr_protocol);
1979 	m_freem(m);
1980 	if (error != 0) {
1981 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
1982 		free(opt, M_IP6OPT);
1983 		return (error);
1984 	}
1985 	*pktopt = opt;
1986 	return (0);
1987 }
1988 #endif
1989 
1990 /*
1991  * initialize ip6_pktopts.  beware that there are non-zero default values in
1992  * the struct.
1993  */
1994 void
1995 ip6_initpktopts(struct ip6_pktopts *opt)
1996 {
1997 
1998 	memset(opt, 0, sizeof(*opt));
1999 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2000 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2001 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2002 }
2003 
2004 #define sin6tosa(sin6)	((struct sockaddr *)(sin6)) /* XXX */
2005 static int
2006 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2007     kauth_cred_t cred, int uproto)
2008 {
2009 	struct ip6_pktopts *opt;
2010 
2011 	if (*pktopt == NULL) {
2012 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2013 		    M_NOWAIT);
2014 		if (*pktopt == NULL)
2015 			return (ENOBUFS);
2016 
2017 		ip6_initpktopts(*pktopt);
2018 	}
2019 	opt = *pktopt;
2020 
2021 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2022 }
2023 
2024 static int
2025 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2026 {
2027 	void *optdata = NULL;
2028 	int optdatalen = 0;
2029 	struct ip6_ext *ip6e;
2030 	int error = 0;
2031 	struct in6_pktinfo null_pktinfo;
2032 	int deftclass = 0, on;
2033 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2034 
2035 	switch (optname) {
2036 	case IPV6_PKTINFO:
2037 		if (pktopt && pktopt->ip6po_pktinfo)
2038 			optdata = (void *)pktopt->ip6po_pktinfo;
2039 		else {
2040 			/* XXX: we don't have to do this every time... */
2041 			memset(&null_pktinfo, 0, sizeof(null_pktinfo));
2042 			optdata = (void *)&null_pktinfo;
2043 		}
2044 		optdatalen = sizeof(struct in6_pktinfo);
2045 		break;
2046 	case IPV6_OTCLASS:
2047 		/* XXX */
2048 		return (EINVAL);
2049 	case IPV6_TCLASS:
2050 		if (pktopt && pktopt->ip6po_tclass >= 0)
2051 			optdata = (void *)&pktopt->ip6po_tclass;
2052 		else
2053 			optdata = (void *)&deftclass;
2054 		optdatalen = sizeof(int);
2055 		break;
2056 	case IPV6_HOPOPTS:
2057 		if (pktopt && pktopt->ip6po_hbh) {
2058 			optdata = (void *)pktopt->ip6po_hbh;
2059 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2060 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2061 		}
2062 		break;
2063 	case IPV6_RTHDR:
2064 		if (pktopt && pktopt->ip6po_rthdr) {
2065 			optdata = (void *)pktopt->ip6po_rthdr;
2066 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2067 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2068 		}
2069 		break;
2070 	case IPV6_RTHDRDSTOPTS:
2071 		if (pktopt && pktopt->ip6po_dest1) {
2072 			optdata = (void *)pktopt->ip6po_dest1;
2073 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2074 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2075 		}
2076 		break;
2077 	case IPV6_DSTOPTS:
2078 		if (pktopt && pktopt->ip6po_dest2) {
2079 			optdata = (void *)pktopt->ip6po_dest2;
2080 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2081 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2082 		}
2083 		break;
2084 	case IPV6_NEXTHOP:
2085 		if (pktopt && pktopt->ip6po_nexthop) {
2086 			optdata = (void *)pktopt->ip6po_nexthop;
2087 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2088 		}
2089 		break;
2090 	case IPV6_USE_MIN_MTU:
2091 		if (pktopt)
2092 			optdata = (void *)&pktopt->ip6po_minmtu;
2093 		else
2094 			optdata = (void *)&defminmtu;
2095 		optdatalen = sizeof(int);
2096 		break;
2097 	case IPV6_DONTFRAG:
2098 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2099 			on = 1;
2100 		else
2101 			on = 0;
2102 		optdata = (void *)&on;
2103 		optdatalen = sizeof(on);
2104 		break;
2105 	default:		/* should not happen */
2106 #ifdef DIAGNOSTIC
2107 		panic("ip6_getpcbopt: unexpected option\n");
2108 #endif
2109 		return (ENOPROTOOPT);
2110 	}
2111 
2112 	error = sockopt_set(sopt, optdata, optdatalen);
2113 
2114 	return (error);
2115 }
2116 
2117 void
2118 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2119 {
2120 	if (optname == -1 || optname == IPV6_PKTINFO) {
2121 		if (pktopt->ip6po_pktinfo)
2122 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2123 		pktopt->ip6po_pktinfo = NULL;
2124 	}
2125 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2126 		pktopt->ip6po_hlim = -1;
2127 	if (optname == -1 || optname == IPV6_TCLASS)
2128 		pktopt->ip6po_tclass = -1;
2129 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2130 		rtcache_free(&pktopt->ip6po_nextroute);
2131 		if (pktopt->ip6po_nexthop)
2132 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2133 		pktopt->ip6po_nexthop = NULL;
2134 	}
2135 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2136 		if (pktopt->ip6po_hbh)
2137 			free(pktopt->ip6po_hbh, M_IP6OPT);
2138 		pktopt->ip6po_hbh = NULL;
2139 	}
2140 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2141 		if (pktopt->ip6po_dest1)
2142 			free(pktopt->ip6po_dest1, M_IP6OPT);
2143 		pktopt->ip6po_dest1 = NULL;
2144 	}
2145 	if (optname == -1 || optname == IPV6_RTHDR) {
2146 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2147 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2148 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2149 		rtcache_free(&pktopt->ip6po_route);
2150 	}
2151 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2152 		if (pktopt->ip6po_dest2)
2153 			free(pktopt->ip6po_dest2, M_IP6OPT);
2154 		pktopt->ip6po_dest2 = NULL;
2155 	}
2156 }
2157 
2158 #define PKTOPT_EXTHDRCPY(type) 					\
2159 do {								\
2160 	if (src->type) {					\
2161 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2162 		dst->type = malloc(hlen, M_IP6OPT, canwait);	\
2163 		if (dst->type == NULL)				\
2164 			goto bad;				\
2165 		memcpy(dst->type, src->type, hlen);		\
2166 	}							\
2167 } while (/*CONSTCOND*/ 0)
2168 
2169 static int
2170 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2171 {
2172 	dst->ip6po_hlim = src->ip6po_hlim;
2173 	dst->ip6po_tclass = src->ip6po_tclass;
2174 	dst->ip6po_flags = src->ip6po_flags;
2175 	if (src->ip6po_pktinfo) {
2176 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2177 		    M_IP6OPT, canwait);
2178 		if (dst->ip6po_pktinfo == NULL)
2179 			goto bad;
2180 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2181 	}
2182 	if (src->ip6po_nexthop) {
2183 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2184 		    M_IP6OPT, canwait);
2185 		if (dst->ip6po_nexthop == NULL)
2186 			goto bad;
2187 		memcpy(dst->ip6po_nexthop, src->ip6po_nexthop,
2188 		    src->ip6po_nexthop->sa_len);
2189 	}
2190 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2191 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2192 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2193 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2194 	return (0);
2195 
2196   bad:
2197 	if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
2198 	if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
2199 	if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
2200 	if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
2201 	if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
2202 	if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
2203 
2204 	return (ENOBUFS);
2205 }
2206 #undef PKTOPT_EXTHDRCPY
2207 
2208 struct ip6_pktopts *
2209 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2210 {
2211 	int error;
2212 	struct ip6_pktopts *dst;
2213 
2214 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2215 	if (dst == NULL)
2216 		return (NULL);
2217 	ip6_initpktopts(dst);
2218 
2219 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2220 		free(dst, M_IP6OPT);
2221 		return (NULL);
2222 	}
2223 
2224 	return (dst);
2225 }
2226 
2227 void
2228 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2229 {
2230 	if (pktopt == NULL)
2231 		return;
2232 
2233 	ip6_clearpktopts(pktopt, -1);
2234 
2235 	free(pktopt, M_IP6OPT);
2236 }
2237 
2238 /*
2239  * Set the IP6 multicast options in response to user setsockopt().
2240  */
2241 static int
2242 ip6_setmoptions(const struct sockopt *sopt, struct ip6_moptions **im6op)
2243 {
2244 	int error = 0;
2245 	u_int loop, ifindex;
2246 	struct ipv6_mreq mreq;
2247 	struct ifnet *ifp;
2248 	struct ip6_moptions *im6o = *im6op;
2249 	struct route ro;
2250 	struct in6_multi_mship *imm;
2251 	struct lwp *l = curlwp;	/* XXX */
2252 
2253 	if (im6o == NULL) {
2254 		/*
2255 		 * No multicast option buffer attached to the pcb;
2256 		 * allocate one and initialize to default values.
2257 		 */
2258 		im6o = malloc(sizeof(*im6o), M_IPMOPTS, M_NOWAIT);
2259 		if (im6o == NULL)
2260 			return (ENOBUFS);
2261 
2262 		*im6op = im6o;
2263 		im6o->im6o_multicast_ifp = NULL;
2264 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2265 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2266 		LIST_INIT(&im6o->im6o_memberships);
2267 	}
2268 
2269 	switch (sopt->sopt_name) {
2270 
2271 	case IPV6_MULTICAST_IF:
2272 		/*
2273 		 * Select the interface for outgoing multicast packets.
2274 		 */
2275 		error = sockopt_get(sopt, &ifindex, sizeof(ifindex));
2276 		if (error != 0)
2277 			break;
2278 
2279 		if (ifindex != 0) {
2280 			if (if_indexlim <= ifindex || !ifindex2ifnet[ifindex]) {
2281 				error = ENXIO;	/* XXX EINVAL? */
2282 				break;
2283 			}
2284 			ifp = ifindex2ifnet[ifindex];
2285 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
2286 				error = EADDRNOTAVAIL;
2287 				break;
2288 			}
2289 		} else
2290 			ifp = NULL;
2291 		im6o->im6o_multicast_ifp = ifp;
2292 		break;
2293 
2294 	case IPV6_MULTICAST_HOPS:
2295 	    {
2296 		/*
2297 		 * Set the IP6 hoplimit for outgoing multicast packets.
2298 		 */
2299 		int optval;
2300 
2301 		error = sockopt_getint(sopt, &optval);
2302 		if (error != 0)
2303 			break;
2304 
2305 		if (optval < -1 || optval >= 256)
2306 			error = EINVAL;
2307 		else if (optval == -1)
2308 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2309 		else
2310 			im6o->im6o_multicast_hlim = optval;
2311 		break;
2312 	    }
2313 
2314 	case IPV6_MULTICAST_LOOP:
2315 		/*
2316 		 * Set the loopback flag for outgoing multicast packets.
2317 		 * Must be zero or one.
2318 		 */
2319 		error = sockopt_get(sopt, &loop, sizeof(loop));
2320 		if (error != 0)
2321 			break;
2322 		if (loop > 1) {
2323 			error = EINVAL;
2324 			break;
2325 		}
2326 		im6o->im6o_multicast_loop = loop;
2327 		break;
2328 
2329 	case IPV6_JOIN_GROUP:
2330 		/*
2331 		 * Add a multicast group membership.
2332 		 * Group must be a valid IP6 multicast address.
2333 		 */
2334 		error = sockopt_get(sopt, &mreq, sizeof(mreq));
2335 		if (error != 0)
2336 			break;
2337 
2338 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq.ipv6mr_multiaddr)) {
2339 			/*
2340 			 * We use the unspecified address to specify to accept
2341 			 * all multicast addresses. Only super user is allowed
2342 			 * to do this.
2343 			 */
2344 			if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_IPV6,
2345 			    KAUTH_REQ_NETWORK_IPV6_JOIN_MULTICAST, NULL, NULL, NULL))
2346 			{
2347 				error = EACCES;
2348 				break;
2349 			}
2350 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq.ipv6mr_multiaddr)) {
2351 			error = EINVAL;
2352 			break;
2353 		}
2354 
2355 		/*
2356 		 * If no interface was explicitly specified, choose an
2357 		 * appropriate one according to the given multicast address.
2358 		 */
2359 		if (mreq.ipv6mr_interface == 0) {
2360 			struct rtentry *rt;
2361 			union {
2362 				struct sockaddr		dst;
2363 				struct sockaddr_in6	dst6;
2364 			} u;
2365 
2366 			/*
2367 			 * Look up the routing table for the
2368 			 * address, and choose the outgoing interface.
2369 			 *   XXX: is it a good approach?
2370 			 */
2371 			memset(&ro, 0, sizeof(ro));
2372 			sockaddr_in6_init(&u.dst6, &mreq.ipv6mr_multiaddr, 0,
2373 			    0, 0);
2374 			rtcache_setdst(&ro, &u.dst);
2375 			ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp
2376 			                                        : NULL;
2377 			rtcache_free(&ro);
2378 		} else {
2379 			/*
2380 			 * If the interface is specified, validate it.
2381 			 */
2382 			if (if_indexlim <= mreq.ipv6mr_interface ||
2383 			    !ifindex2ifnet[mreq.ipv6mr_interface]) {
2384 				error = ENXIO;	/* XXX EINVAL? */
2385 				break;
2386 			}
2387 			ifp = ifindex2ifnet[mreq.ipv6mr_interface];
2388 		}
2389 
2390 		/*
2391 		 * See if we found an interface, and confirm that it
2392 		 * supports multicast
2393 		 */
2394 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2395 			error = EADDRNOTAVAIL;
2396 			break;
2397 		}
2398 
2399 		if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) {
2400 			error = EADDRNOTAVAIL; /* XXX: should not happen */
2401 			break;
2402 		}
2403 
2404 		/*
2405 		 * See if the membership already exists.
2406 		 */
2407 		for (imm = im6o->im6o_memberships.lh_first;
2408 		     imm != NULL; imm = imm->i6mm_chain.le_next)
2409 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
2410 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2411 			    &mreq.ipv6mr_multiaddr))
2412 				break;
2413 		if (imm != NULL) {
2414 			error = EADDRINUSE;
2415 			break;
2416 		}
2417 		/*
2418 		 * Everything looks good; add a new record to the multicast
2419 		 * address list for the given interface.
2420 		 */
2421 		imm = in6_joingroup(ifp, &mreq.ipv6mr_multiaddr, &error, 0);
2422 		if (imm == NULL)
2423 			break;
2424 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2425 		break;
2426 
2427 	case IPV6_LEAVE_GROUP:
2428 		/*
2429 		 * Drop a multicast group membership.
2430 		 * Group must be a valid IP6 multicast address.
2431 		 */
2432 		error = sockopt_get(sopt, &mreq, sizeof(mreq));
2433 		if (error != 0)
2434 			break;
2435 
2436 		/*
2437 		 * If an interface address was specified, get a pointer
2438 		 * to its ifnet structure.
2439 		 */
2440 		if (mreq.ipv6mr_interface != 0) {
2441 			if (if_indexlim <= mreq.ipv6mr_interface ||
2442 			    !ifindex2ifnet[mreq.ipv6mr_interface]) {
2443 				error = ENXIO;	/* XXX EINVAL? */
2444 				break;
2445 			}
2446 			ifp = ifindex2ifnet[mreq.ipv6mr_interface];
2447 		} else
2448 			ifp = NULL;
2449 
2450 		/* Fill in the scope zone ID */
2451 		if (ifp) {
2452 			if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) {
2453 				/* XXX: should not happen */
2454 				error = EADDRNOTAVAIL;
2455 				break;
2456 			}
2457 		} else if (mreq.ipv6mr_interface != 0) {
2458 			/*
2459 			 * XXX: This case would happens when the (positive)
2460 			 * index is in the valid range, but the corresponding
2461 			 * interface has been detached dynamically.  The above
2462 			 * check probably avoids such case to happen here, but
2463 			 * we check it explicitly for safety.
2464 			 */
2465 			error = EADDRNOTAVAIL;
2466 			break;
2467 		} else {	/* ipv6mr_interface == 0 */
2468 			struct sockaddr_in6 sa6_mc;
2469 
2470 			/*
2471 			 * The API spec says as follows:
2472 			 *  If the interface index is specified as 0, the
2473 			 *  system may choose a multicast group membership to
2474 			 *  drop by matching the multicast address only.
2475 			 * On the other hand, we cannot disambiguate the scope
2476 			 * zone unless an interface is provided.  Thus, we
2477 			 * check if there's ambiguity with the default scope
2478 			 * zone as the last resort.
2479 			 */
2480 			sockaddr_in6_init(&sa6_mc, &mreq.ipv6mr_multiaddr,
2481 			    0, 0, 0);
2482 			error = sa6_embedscope(&sa6_mc, ip6_use_defzone);
2483 			if (error != 0)
2484 				break;
2485 			mreq.ipv6mr_multiaddr = sa6_mc.sin6_addr;
2486 		}
2487 
2488 		/*
2489 		 * Find the membership in the membership list.
2490 		 */
2491 		for (imm = im6o->im6o_memberships.lh_first;
2492 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
2493 			if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2494 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2495 			    &mreq.ipv6mr_multiaddr))
2496 				break;
2497 		}
2498 		if (imm == NULL) {
2499 			/* Unable to resolve interface */
2500 			error = EADDRNOTAVAIL;
2501 			break;
2502 		}
2503 		/*
2504 		 * Give up the multicast address record to which the
2505 		 * membership points.
2506 		 */
2507 		LIST_REMOVE(imm, i6mm_chain);
2508 		in6_leavegroup(imm);
2509 		break;
2510 
2511 	default:
2512 		error = EOPNOTSUPP;
2513 		break;
2514 	}
2515 
2516 	/*
2517 	 * If all options have default values, no need to keep the mbuf.
2518 	 */
2519 	if (im6o->im6o_multicast_ifp == NULL &&
2520 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2521 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2522 	    im6o->im6o_memberships.lh_first == NULL) {
2523 		free(*im6op, M_IPMOPTS);
2524 		*im6op = NULL;
2525 	}
2526 
2527 	return (error);
2528 }
2529 
2530 /*
2531  * Return the IP6 multicast options in response to user getsockopt().
2532  */
2533 static int
2534 ip6_getmoptions(struct sockopt *sopt, struct ip6_moptions *im6o)
2535 {
2536 	u_int optval;
2537 	int error;
2538 
2539 	switch (sopt->sopt_name) {
2540 	case IPV6_MULTICAST_IF:
2541 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2542 			optval = 0;
2543 		else
2544 			optval = im6o->im6o_multicast_ifp->if_index;
2545 
2546 		error = sockopt_set(sopt, &optval, sizeof(optval));
2547 		break;
2548 
2549 	case IPV6_MULTICAST_HOPS:
2550 		if (im6o == NULL)
2551 			optval = ip6_defmcasthlim;
2552 		else
2553 			optval = im6o->im6o_multicast_hlim;
2554 
2555 		error = sockopt_set(sopt, &optval, sizeof(optval));
2556 		break;
2557 
2558 	case IPV6_MULTICAST_LOOP:
2559 		if (im6o == NULL)
2560 			optval = IPV6_DEFAULT_MULTICAST_LOOP;
2561 		else
2562 			optval = im6o->im6o_multicast_loop;
2563 
2564 		error = sockopt_set(sopt, &optval, sizeof(optval));
2565 		break;
2566 
2567 	default:
2568 		error = EOPNOTSUPP;
2569 	}
2570 
2571 	return (error);
2572 }
2573 
2574 /*
2575  * Discard the IP6 multicast options.
2576  */
2577 void
2578 ip6_freemoptions(struct ip6_moptions *im6o)
2579 {
2580 	struct in6_multi_mship *imm;
2581 
2582 	if (im6o == NULL)
2583 		return;
2584 
2585 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2586 		LIST_REMOVE(imm, i6mm_chain);
2587 		in6_leavegroup(imm);
2588 	}
2589 	free(im6o, M_IPMOPTS);
2590 }
2591 
2592 /*
2593  * Set IPv6 outgoing packet options based on advanced API.
2594  */
2595 int
2596 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2597 	struct ip6_pktopts *stickyopt, kauth_cred_t cred, int uproto)
2598 {
2599 	struct cmsghdr *cm = 0;
2600 
2601 	if (control == NULL || opt == NULL)
2602 		return (EINVAL);
2603 
2604 	ip6_initpktopts(opt);
2605 	if (stickyopt) {
2606 		int error;
2607 
2608 		/*
2609 		 * If stickyopt is provided, make a local copy of the options
2610 		 * for this particular packet, then override them by ancillary
2611 		 * objects.
2612 		 * XXX: copypktopts() does not copy the cached route to a next
2613 		 * hop (if any).  This is not very good in terms of efficiency,
2614 		 * but we can allow this since this option should be rarely
2615 		 * used.
2616 		 */
2617 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2618 			return (error);
2619 	}
2620 
2621 	/*
2622 	 * XXX: Currently, we assume all the optional information is stored
2623 	 * in a single mbuf.
2624 	 */
2625 	if (control->m_next)
2626 		return (EINVAL);
2627 
2628 	/* XXX if cm->cmsg_len is not aligned, control->m_len can become <0 */
2629 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2630 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2631 		int error;
2632 
2633 		if (control->m_len < CMSG_LEN(0))
2634 			return (EINVAL);
2635 
2636 		cm = mtod(control, struct cmsghdr *);
2637 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2638 			return (EINVAL);
2639 		if (cm->cmsg_level != IPPROTO_IPV6)
2640 			continue;
2641 
2642 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2643 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2644 		if (error)
2645 			return (error);
2646 	}
2647 
2648 	return (0);
2649 }
2650 
2651 /*
2652  * Set a particular packet option, as a sticky option or an ancillary data
2653  * item.  "len" can be 0 only when it's a sticky option.
2654  * We have 4 cases of combination of "sticky" and "cmsg":
2655  * "sticky=0, cmsg=0": impossible
2656  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2657  * "sticky=1, cmsg=0": RFC3542 socket option
2658  * "sticky=1, cmsg=1": RFC2292 socket option
2659  */
2660 static int
2661 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2662     kauth_cred_t cred, int sticky, int cmsg, int uproto)
2663 {
2664 	int minmtupolicy;
2665 	int error;
2666 
2667 	if (!sticky && !cmsg) {
2668 #ifdef DIAGNOSTIC
2669 		printf("ip6_setpktopt: impossible case\n");
2670 #endif
2671 		return (EINVAL);
2672 	}
2673 
2674 	/*
2675 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2676 	 * not be specified in the context of RFC3542.  Conversely,
2677 	 * RFC3542 types should not be specified in the context of RFC2292.
2678 	 */
2679 	if (!cmsg) {
2680 		switch (optname) {
2681 		case IPV6_2292PKTINFO:
2682 		case IPV6_2292HOPLIMIT:
2683 		case IPV6_2292NEXTHOP:
2684 		case IPV6_2292HOPOPTS:
2685 		case IPV6_2292DSTOPTS:
2686 		case IPV6_2292RTHDR:
2687 		case IPV6_2292PKTOPTIONS:
2688 			return (ENOPROTOOPT);
2689 		}
2690 	}
2691 	if (sticky && cmsg) {
2692 		switch (optname) {
2693 		case IPV6_PKTINFO:
2694 		case IPV6_HOPLIMIT:
2695 		case IPV6_NEXTHOP:
2696 		case IPV6_HOPOPTS:
2697 		case IPV6_DSTOPTS:
2698 		case IPV6_RTHDRDSTOPTS:
2699 		case IPV6_RTHDR:
2700 		case IPV6_USE_MIN_MTU:
2701 		case IPV6_DONTFRAG:
2702 		case IPV6_OTCLASS:
2703 		case IPV6_TCLASS:
2704 			return (ENOPROTOOPT);
2705 		}
2706 	}
2707 
2708 	switch (optname) {
2709 #ifdef RFC2292
2710 	case IPV6_2292PKTINFO:
2711 #endif
2712 	case IPV6_PKTINFO:
2713 	{
2714 		struct ifnet *ifp = NULL;
2715 		struct in6_pktinfo *pktinfo;
2716 
2717 		if (len != sizeof(struct in6_pktinfo))
2718 			return (EINVAL);
2719 
2720 		pktinfo = (struct in6_pktinfo *)buf;
2721 
2722 		/*
2723 		 * An application can clear any sticky IPV6_PKTINFO option by
2724 		 * doing a "regular" setsockopt with ipi6_addr being
2725 		 * in6addr_any and ipi6_ifindex being zero.
2726 		 * [RFC 3542, Section 6]
2727 		 */
2728 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2729 		    pktinfo->ipi6_ifindex == 0 &&
2730 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2731 			ip6_clearpktopts(opt, optname);
2732 			break;
2733 		}
2734 
2735 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2736 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2737 			return (EINVAL);
2738 		}
2739 
2740 		/* validate the interface index if specified. */
2741 		if (pktinfo->ipi6_ifindex >= if_indexlim) {
2742 			 return (ENXIO);
2743 		}
2744 		if (pktinfo->ipi6_ifindex) {
2745 			ifp = ifindex2ifnet[pktinfo->ipi6_ifindex];
2746 			if (ifp == NULL)
2747 				return (ENXIO);
2748 		}
2749 
2750 		/*
2751 		 * We store the address anyway, and let in6_selectsrc()
2752 		 * validate the specified address.  This is because ipi6_addr
2753 		 * may not have enough information about its scope zone, and
2754 		 * we may need additional information (such as outgoing
2755 		 * interface or the scope zone of a destination address) to
2756 		 * disambiguate the scope.
2757 		 * XXX: the delay of the validation may confuse the
2758 		 * application when it is used as a sticky option.
2759 		 */
2760 		if (opt->ip6po_pktinfo == NULL) {
2761 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2762 			    M_IP6OPT, M_NOWAIT);
2763 			if (opt->ip6po_pktinfo == NULL)
2764 				return (ENOBUFS);
2765 		}
2766 		memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo));
2767 		break;
2768 	}
2769 
2770 #ifdef RFC2292
2771 	case IPV6_2292HOPLIMIT:
2772 #endif
2773 	case IPV6_HOPLIMIT:
2774 	{
2775 		int *hlimp;
2776 
2777 		/*
2778 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2779 		 * to simplify the ordering among hoplimit options.
2780 		 */
2781 		if (optname == IPV6_HOPLIMIT && sticky)
2782 			return (ENOPROTOOPT);
2783 
2784 		if (len != sizeof(int))
2785 			return (EINVAL);
2786 		hlimp = (int *)buf;
2787 		if (*hlimp < -1 || *hlimp > 255)
2788 			return (EINVAL);
2789 
2790 		opt->ip6po_hlim = *hlimp;
2791 		break;
2792 	}
2793 
2794 	case IPV6_OTCLASS:
2795 		if (len != sizeof(u_int8_t))
2796 			return (EINVAL);
2797 
2798 		opt->ip6po_tclass = *(u_int8_t *)buf;
2799 		break;
2800 
2801 	case IPV6_TCLASS:
2802 	{
2803 		int tclass;
2804 
2805 		if (len != sizeof(int))
2806 			return (EINVAL);
2807 		tclass = *(int *)buf;
2808 		if (tclass < -1 || tclass > 255)
2809 			return (EINVAL);
2810 
2811 		opt->ip6po_tclass = tclass;
2812 		break;
2813 	}
2814 
2815 #ifdef RFC2292
2816 	case IPV6_2292NEXTHOP:
2817 #endif
2818 	case IPV6_NEXTHOP:
2819 		error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
2820 		    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2821 		if (error)
2822 			return (error);
2823 
2824 		if (len == 0) {	/* just remove the option */
2825 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2826 			break;
2827 		}
2828 
2829 		/* check if cmsg_len is large enough for sa_len */
2830 		if (len < sizeof(struct sockaddr) || len < *buf)
2831 			return (EINVAL);
2832 
2833 		switch (((struct sockaddr *)buf)->sa_family) {
2834 		case AF_INET6:
2835 		{
2836 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2837 
2838 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2839 				return (EINVAL);
2840 
2841 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2842 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2843 				return (EINVAL);
2844 			}
2845 			if ((error = sa6_embedscope(sa6, ip6_use_defzone))
2846 			    != 0) {
2847 				return (error);
2848 			}
2849 			break;
2850 		}
2851 		case AF_LINK:	/* eventually be supported? */
2852 		default:
2853 			return (EAFNOSUPPORT);
2854 		}
2855 
2856 		/* turn off the previous option, then set the new option. */
2857 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2858 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2859 		if (opt->ip6po_nexthop == NULL)
2860 			return (ENOBUFS);
2861 		memcpy(opt->ip6po_nexthop, buf, *buf);
2862 		break;
2863 
2864 #ifdef RFC2292
2865 	case IPV6_2292HOPOPTS:
2866 #endif
2867 	case IPV6_HOPOPTS:
2868 	{
2869 		struct ip6_hbh *hbh;
2870 		int hbhlen;
2871 
2872 		/*
2873 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2874 		 * options, since per-option restriction has too much
2875 		 * overhead.
2876 		 */
2877 		error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
2878 		    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2879 		if (error)
2880 			return (error);
2881 
2882 		if (len == 0) {
2883 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2884 			break;	/* just remove the option */
2885 		}
2886 
2887 		/* message length validation */
2888 		if (len < sizeof(struct ip6_hbh))
2889 			return (EINVAL);
2890 		hbh = (struct ip6_hbh *)buf;
2891 		hbhlen = (hbh->ip6h_len + 1) << 3;
2892 		if (len != hbhlen)
2893 			return (EINVAL);
2894 
2895 		/* turn off the previous option, then set the new option. */
2896 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2897 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2898 		if (opt->ip6po_hbh == NULL)
2899 			return (ENOBUFS);
2900 		memcpy(opt->ip6po_hbh, hbh, hbhlen);
2901 
2902 		break;
2903 	}
2904 
2905 #ifdef RFC2292
2906 	case IPV6_2292DSTOPTS:
2907 #endif
2908 	case IPV6_DSTOPTS:
2909 	case IPV6_RTHDRDSTOPTS:
2910 	{
2911 		struct ip6_dest *dest, **newdest = NULL;
2912 		int destlen;
2913 
2914 		/* XXX: see the comment for IPV6_HOPOPTS */
2915 		error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
2916 		    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2917 		if (error)
2918 			return (error);
2919 
2920 		if (len == 0) {
2921 			ip6_clearpktopts(opt, optname);
2922 			break;	/* just remove the option */
2923 		}
2924 
2925 		/* message length validation */
2926 		if (len < sizeof(struct ip6_dest))
2927 			return (EINVAL);
2928 		dest = (struct ip6_dest *)buf;
2929 		destlen = (dest->ip6d_len + 1) << 3;
2930 		if (len != destlen)
2931 			return (EINVAL);
2932 		/*
2933 		 * Determine the position that the destination options header
2934 		 * should be inserted; before or after the routing header.
2935 		 */
2936 		switch (optname) {
2937 		case IPV6_2292DSTOPTS:
2938 			/*
2939 			 * The old advanced API is ambiguous on this point.
2940 			 * Our approach is to determine the position based
2941 			 * according to the existence of a routing header.
2942 			 * Note, however, that this depends on the order of the
2943 			 * extension headers in the ancillary data; the 1st
2944 			 * part of the destination options header must appear
2945 			 * before the routing header in the ancillary data,
2946 			 * too.
2947 			 * RFC3542 solved the ambiguity by introducing
2948 			 * separate ancillary data or option types.
2949 			 */
2950 			if (opt->ip6po_rthdr == NULL)
2951 				newdest = &opt->ip6po_dest1;
2952 			else
2953 				newdest = &opt->ip6po_dest2;
2954 			break;
2955 		case IPV6_RTHDRDSTOPTS:
2956 			newdest = &opt->ip6po_dest1;
2957 			break;
2958 		case IPV6_DSTOPTS:
2959 			newdest = &opt->ip6po_dest2;
2960 			break;
2961 		}
2962 
2963 		/* turn off the previous option, then set the new option. */
2964 		ip6_clearpktopts(opt, optname);
2965 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2966 		if (*newdest == NULL)
2967 			return (ENOBUFS);
2968 		memcpy(*newdest, dest, destlen);
2969 
2970 		break;
2971 	}
2972 
2973 #ifdef RFC2292
2974 	case IPV6_2292RTHDR:
2975 #endif
2976 	case IPV6_RTHDR:
2977 	{
2978 		struct ip6_rthdr *rth;
2979 		int rthlen;
2980 
2981 		if (len == 0) {
2982 			ip6_clearpktopts(opt, IPV6_RTHDR);
2983 			break;	/* just remove the option */
2984 		}
2985 
2986 		/* message length validation */
2987 		if (len < sizeof(struct ip6_rthdr))
2988 			return (EINVAL);
2989 		rth = (struct ip6_rthdr *)buf;
2990 		rthlen = (rth->ip6r_len + 1) << 3;
2991 		if (len != rthlen)
2992 			return (EINVAL);
2993 		switch (rth->ip6r_type) {
2994 		case IPV6_RTHDR_TYPE_0:
2995 			if (rth->ip6r_len == 0)	/* must contain one addr */
2996 				return (EINVAL);
2997 			if (rth->ip6r_len % 2) /* length must be even */
2998 				return (EINVAL);
2999 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3000 				return (EINVAL);
3001 			break;
3002 		default:
3003 			return (EINVAL);	/* not supported */
3004 		}
3005 		/* turn off the previous option */
3006 		ip6_clearpktopts(opt, IPV6_RTHDR);
3007 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3008 		if (opt->ip6po_rthdr == NULL)
3009 			return (ENOBUFS);
3010 		memcpy(opt->ip6po_rthdr, rth, rthlen);
3011 		break;
3012 	}
3013 
3014 	case IPV6_USE_MIN_MTU:
3015 		if (len != sizeof(int))
3016 			return (EINVAL);
3017 		minmtupolicy = *(int *)buf;
3018 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3019 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3020 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3021 			return (EINVAL);
3022 		}
3023 		opt->ip6po_minmtu = minmtupolicy;
3024 		break;
3025 
3026 	case IPV6_DONTFRAG:
3027 		if (len != sizeof(int))
3028 			return (EINVAL);
3029 
3030 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3031 			/*
3032 			 * we ignore this option for TCP sockets.
3033 			 * (RFC3542 leaves this case unspecified.)
3034 			 */
3035 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3036 		} else
3037 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3038 		break;
3039 
3040 	default:
3041 		return (ENOPROTOOPT);
3042 	} /* end of switch */
3043 
3044 	return (0);
3045 }
3046 
3047 /*
3048  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3049  * packet to the input queue of a specified interface.  Note that this
3050  * calls the output routine of the loopback "driver", but with an interface
3051  * pointer that might NOT be lo0ifp -- easier than replicating that code here.
3052  */
3053 void
3054 ip6_mloopback(struct ifnet *ifp, struct mbuf *m,
3055 	const struct sockaddr_in6 *dst)
3056 {
3057 	struct mbuf *copym;
3058 	struct ip6_hdr *ip6;
3059 
3060 	copym = m_copy(m, 0, M_COPYALL);
3061 	if (copym == NULL)
3062 		return;
3063 
3064 	/*
3065 	 * Make sure to deep-copy IPv6 header portion in case the data
3066 	 * is in an mbuf cluster, so that we can safely override the IPv6
3067 	 * header portion later.
3068 	 */
3069 	if ((copym->m_flags & M_EXT) != 0 ||
3070 	    copym->m_len < sizeof(struct ip6_hdr)) {
3071 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3072 		if (copym == NULL)
3073 			return;
3074 	}
3075 
3076 #ifdef DIAGNOSTIC
3077 	if (copym->m_len < sizeof(*ip6)) {
3078 		m_freem(copym);
3079 		return;
3080 	}
3081 #endif
3082 
3083 	ip6 = mtod(copym, struct ip6_hdr *);
3084 	/*
3085 	 * clear embedded scope identifiers if necessary.
3086 	 * in6_clearscope will touch the addresses only when necessary.
3087 	 */
3088 	in6_clearscope(&ip6->ip6_src);
3089 	in6_clearscope(&ip6->ip6_dst);
3090 
3091 	(void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL);
3092 }
3093 
3094 /*
3095  * Chop IPv6 header off from the payload.
3096  */
3097 static int
3098 ip6_splithdr(struct mbuf *m,  struct ip6_exthdrs *exthdrs)
3099 {
3100 	struct mbuf *mh;
3101 	struct ip6_hdr *ip6;
3102 
3103 	ip6 = mtod(m, struct ip6_hdr *);
3104 	if (m->m_len > sizeof(*ip6)) {
3105 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3106 		if (mh == 0) {
3107 			m_freem(m);
3108 			return ENOBUFS;
3109 		}
3110 		M_MOVE_PKTHDR(mh, m);
3111 		MH_ALIGN(mh, sizeof(*ip6));
3112 		m->m_len -= sizeof(*ip6);
3113 		m->m_data += sizeof(*ip6);
3114 		mh->m_next = m;
3115 		m = mh;
3116 		m->m_len = sizeof(*ip6);
3117 		bcopy((void *)ip6, mtod(m, void *), sizeof(*ip6));
3118 	}
3119 	exthdrs->ip6e_ip6 = m;
3120 	return 0;
3121 }
3122 
3123 /*
3124  * Compute IPv6 extension header length.
3125  */
3126 int
3127 ip6_optlen(struct in6pcb *in6p)
3128 {
3129 	int len;
3130 
3131 	if (!in6p->in6p_outputopts)
3132 		return 0;
3133 
3134 	len = 0;
3135 #define elen(x) \
3136     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3137 
3138 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3139 	len += elen(in6p->in6p_outputopts->ip6po_dest1);
3140 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3141 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3142 	return len;
3143 #undef elen
3144 }
3145