xref: /netbsd-src/sys/netinet6/ip6_output.c (revision b7b7574d3bf8eeb51a1fa3977b59142ec6434a55)
1 /*	$NetBSD: ip6_output.c,v 1.157 2014/05/30 01:39:03 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 1.157 2014/05/30 01:39:03 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 #endif
185 
186 	memset(&ip6route, 0, sizeof(ip6route));
187 
188 #ifdef  DIAGNOSTIC
189 	if ((m->m_flags & M_PKTHDR) == 0)
190 		panic("ip6_output: no HDR");
191 
192 	if ((m->m_pkthdr.csum_flags &
193 	    (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) {
194 		panic("ip6_output: IPv4 checksum offload flags: %d",
195 		    m->m_pkthdr.csum_flags);
196 	}
197 
198 	if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) ==
199 	    (M_CSUM_TCPv6|M_CSUM_UDPv6)) {
200 		panic("ip6_output: conflicting checksum offload flags: %d",
201 		    m->m_pkthdr.csum_flags);
202 	}
203 #endif
204 
205 	M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr));
206 
207 #define MAKE_EXTHDR(hp, mp)						\
208     do {								\
209 	if (hp) {							\
210 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
211 		error = ip6_copyexthdr((mp), (void *)(hp), 		\
212 		    ((eh)->ip6e_len + 1) << 3);				\
213 		if (error)						\
214 			goto freehdrs;					\
215 	}								\
216     } while (/*CONSTCOND*/ 0)
217 
218 	memset(&exthdrs, 0, sizeof(exthdrs));
219 	if (opt) {
220 		/* Hop-by-Hop options header */
221 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
222 		/* Destination options header(1st part) */
223 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
224 		/* Routing header */
225 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
226 		/* Destination options header(2nd part) */
227 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
228 	}
229 
230 	/*
231 	 * Calculate the total length of the extension header chain.
232 	 * Keep the length of the unfragmentable part for fragmentation.
233 	 */
234 	optlen = 0;
235 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
236 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
237 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
238 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
239 	/* NOTE: we don't add AH/ESP length here. do that later. */
240 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
241 
242 #ifdef IPSEC
243 	if (ipsec_used) {
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 	}
259 #endif /* IPSEC */
260 
261 
262 	if (needipsec &&
263 	    (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
264 		in6_delayed_cksum(m);
265 		m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
266 	}
267 
268 
269 	/*
270 	 * If we need IPsec, or there is at least one extension header,
271 	 * separate IP6 header from the payload.
272 	 */
273 	if ((needipsec || optlen) && !hdrsplit) {
274 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
275 			m = NULL;
276 			goto freehdrs;
277 		}
278 		m = exthdrs.ip6e_ip6;
279 		hdrsplit++;
280 	}
281 
282 	/* adjust pointer */
283 	ip6 = mtod(m, struct ip6_hdr *);
284 
285 	/* adjust mbuf packet header length */
286 	m->m_pkthdr.len += optlen;
287 	plen = m->m_pkthdr.len - sizeof(*ip6);
288 
289 	/* If this is a jumbo payload, insert a jumbo payload option. */
290 	if (plen > IPV6_MAXPACKET) {
291 		if (!hdrsplit) {
292 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
293 				m = NULL;
294 				goto freehdrs;
295 			}
296 			m = exthdrs.ip6e_ip6;
297 			hdrsplit++;
298 		}
299 		/* adjust pointer */
300 		ip6 = mtod(m, struct ip6_hdr *);
301 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
302 			goto freehdrs;
303 		optlen += 8; /* XXX JUMBOOPTLEN */
304 		ip6->ip6_plen = 0;
305 	} else
306 		ip6->ip6_plen = htons(plen);
307 
308 	/*
309 	 * Concatenate headers and fill in next header fields.
310 	 * Here we have, on "m"
311 	 *	IPv6 payload
312 	 * and we insert headers accordingly.  Finally, we should be getting:
313 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
314 	 *
315 	 * during the header composing process, "m" points to IPv6 header.
316 	 * "mprev" points to an extension header prior to esp.
317 	 */
318 	{
319 		u_char *nexthdrp = &ip6->ip6_nxt;
320 		struct mbuf *mprev = m;
321 
322 		/*
323 		 * we treat dest2 specially.  this makes IPsec processing
324 		 * much easier.  the goal here is to make mprev point the
325 		 * mbuf prior to dest2.
326 		 *
327 		 * result: IPv6 dest2 payload
328 		 * m and mprev will point to IPv6 header.
329 		 */
330 		if (exthdrs.ip6e_dest2) {
331 			if (!hdrsplit)
332 				panic("assumption failed: hdr not split");
333 			exthdrs.ip6e_dest2->m_next = m->m_next;
334 			m->m_next = exthdrs.ip6e_dest2;
335 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
336 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
337 		}
338 
339 #define MAKE_CHAIN(m, mp, p, i)\
340     do {\
341 	if (m) {\
342 		if (!hdrsplit) \
343 			panic("assumption failed: hdr not split"); \
344 		*mtod((m), u_char *) = *(p);\
345 		*(p) = (i);\
346 		p = mtod((m), u_char *);\
347 		(m)->m_next = (mp)->m_next;\
348 		(mp)->m_next = (m);\
349 		(mp) = (m);\
350 	}\
351     } while (/*CONSTCOND*/ 0)
352 		/*
353 		 * result: IPv6 hbh dest1 rthdr dest2 payload
354 		 * m will point to IPv6 header.  mprev will point to the
355 		 * extension header prior to dest2 (rthdr in the above case).
356 		 */
357 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
358 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
359 		    IPPROTO_DSTOPTS);
360 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
361 		    IPPROTO_ROUTING);
362 
363 		M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data,
364 		    sizeof(struct ip6_hdr) + optlen);
365 	}
366 
367 	/*
368 	 * If there is a routing header, replace destination address field
369 	 * with the first hop of the routing header.
370 	 */
371 	if (exthdrs.ip6e_rthdr) {
372 		struct ip6_rthdr *rh;
373 		struct ip6_rthdr0 *rh0;
374 		struct in6_addr *addr;
375 		struct sockaddr_in6 sa;
376 
377 		rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
378 		    struct ip6_rthdr *));
379 		finaldst = ip6->ip6_dst;
380 		switch (rh->ip6r_type) {
381 		case IPV6_RTHDR_TYPE_0:
382 			 rh0 = (struct ip6_rthdr0 *)rh;
383 			 addr = (struct in6_addr *)(rh0 + 1);
384 
385 			 /*
386 			  * construct a sockaddr_in6 form of
387 			  * the first hop.
388 			  *
389 			  * XXX: we may not have enough
390 			  * information about its scope zone;
391 			  * there is no standard API to pass
392 			  * the information from the
393 			  * application.
394 			  */
395 			 sockaddr_in6_init(&sa, addr, 0, 0, 0);
396 			 if ((error = sa6_embedscope(&sa,
397 			     ip6_use_defzone)) != 0) {
398 				 goto bad;
399 			 }
400 			 ip6->ip6_dst = sa.sin6_addr;
401 			 (void)memmove(&addr[0], &addr[1],
402 			     sizeof(struct in6_addr) *
403 			     (rh0->ip6r0_segleft - 1));
404 			 addr[rh0->ip6r0_segleft - 1] = finaldst;
405 			 /* XXX */
406 			 in6_clearscope(addr + rh0->ip6r0_segleft - 1);
407 			 break;
408 		default:	/* is it possible? */
409 			 error = EINVAL;
410 			 goto bad;
411 		}
412 	}
413 
414 	/* Source address validation */
415 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
416 	    (flags & IPV6_UNSPECSRC) == 0) {
417 		error = EOPNOTSUPP;
418 		IP6_STATINC(IP6_STAT_BADSCOPE);
419 		goto bad;
420 	}
421 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
422 		error = EOPNOTSUPP;
423 		IP6_STATINC(IP6_STAT_BADSCOPE);
424 		goto bad;
425 	}
426 
427 	IP6_STATINC(IP6_STAT_LOCALOUT);
428 
429 	/*
430 	 * Route packet.
431 	 */
432 	/* initialize cached route */
433 	if (ro == NULL) {
434 		ro = &ip6route;
435 	}
436 	ro_pmtu = ro;
437 	if (opt && opt->ip6po_rthdr)
438 		ro = &opt->ip6po_route;
439 
440  	/*
441 	 * if specified, try to fill in the traffic class field.
442 	 * do not override if a non-zero value is already set.
443 	 * we check the diffserv field and the ecn field separately.
444 	 */
445 	if (opt && opt->ip6po_tclass >= 0) {
446 		int mask = 0;
447 
448 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
449 			mask |= 0xfc;
450 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
451 			mask |= 0x03;
452 		if (mask != 0)
453 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
454 	}
455 
456 	/* fill in or override the hop limit field, if necessary. */
457 	if (opt && opt->ip6po_hlim != -1)
458 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
459 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
460 		if (im6o != NULL)
461 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
462 		else
463 			ip6->ip6_hlim = ip6_defmcasthlim;
464 	}
465 
466 #ifdef IPSEC
467 	if (needipsec) {
468 		int s = splsoftnet();
469 		error = ipsec6_process_packet(m, sp->req);
470 
471 		/*
472 		 * Preserve KAME behaviour: ENOENT can be returned
473 		 * when an SA acquire is in progress.  Don't propagate
474 		 * this to user-level; it confuses applications.
475 		 * XXX this will go away when the SADB is redone.
476 		 */
477 		if (error == ENOENT)
478 			error = 0;
479 		splx(s);
480 		goto done;
481 	}
482 #endif /* IPSEC */
483 
484 	/* adjust pointer */
485 	ip6 = mtod(m, struct ip6_hdr *);
486 
487 	sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
488 	if ((error = in6_selectroute(&dst_sa, opt, im6o, ro,
489 	    &ifp, &rt, 0)) != 0) {
490 		if (ifp != NULL)
491 			in6_ifstat_inc(ifp, ifs6_out_discard);
492 		goto bad;
493 	}
494 	if (rt == NULL) {
495 		/*
496 		 * If in6_selectroute() does not return a route entry,
497 		 * dst may not have been updated.
498 		 */
499 		error = rtcache_setdst(ro, sin6tosa(&dst_sa));
500 		if (error) {
501 			goto bad;
502 		}
503 	}
504 
505 	/*
506 	 * then rt (for unicast) and ifp must be non-NULL valid values.
507 	 */
508 	if ((flags & IPV6_FORWARDING) == 0) {
509 		/* XXX: the FORWARDING flag can be set for mrouting. */
510 		in6_ifstat_inc(ifp, ifs6_out_request);
511 	}
512 	if (rt != NULL) {
513 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
514 		rt->rt_use++;
515 	}
516 
517 	/*
518 	 * The outgoing interface must be in the zone of source and
519 	 * destination addresses.  We should use ia_ifp to support the
520 	 * case of sending packets to an address of our own.
521 	 */
522 	if (ia != NULL && ia->ia_ifp)
523 		origifp = ia->ia_ifp;
524 	else
525 		origifp = ifp;
526 
527 	src0 = ip6->ip6_src;
528 	if (in6_setscope(&src0, origifp, &zone))
529 		goto badscope;
530 	sockaddr_in6_init(&src_sa, &ip6->ip6_src, 0, 0, 0);
531 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
532 		goto badscope;
533 
534 	dst0 = ip6->ip6_dst;
535 	if (in6_setscope(&dst0, origifp, &zone))
536 		goto badscope;
537 	/* re-initialize to be sure */
538 	sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
539 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id)
540 		goto badscope;
541 
542 	/* scope check is done. */
543 
544 	if (rt == NULL || IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
545 		if (dst == NULL)
546 			dst = satocsin6(rtcache_getdst(ro));
547 		KASSERT(dst != NULL);
548 	} else if (opt && rtcache_validate(&opt->ip6po_nextroute) != NULL) {
549 		/*
550 		 * The nexthop is explicitly specified by the
551 		 * application.  We assume the next hop is an IPv6
552 		 * address.
553 		 */
554 		dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
555 	} else if ((rt->rt_flags & RTF_GATEWAY))
556 		dst = (struct sockaddr_in6 *)rt->rt_gateway;
557 	else if (dst == NULL)
558 		dst = satocsin6(rtcache_getdst(ro));
559 
560 	/*
561 	 * XXXXXX: original code follows:
562 	 */
563 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
564 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
565 	else {
566 		struct	in6_multi *in6m;
567 
568 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
569 
570 		in6_ifstat_inc(ifp, ifs6_out_mcast);
571 
572 		/*
573 		 * Confirm that the outgoing interface supports multicast.
574 		 */
575 		if (!(ifp->if_flags & IFF_MULTICAST)) {
576 			IP6_STATINC(IP6_STAT_NOROUTE);
577 			in6_ifstat_inc(ifp, ifs6_out_discard);
578 			error = ENETUNREACH;
579 			goto bad;
580 		}
581 
582 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
583 		if (in6m != NULL &&
584 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
585 			/*
586 			 * If we belong to the destination multicast group
587 			 * on the outgoing interface, and the caller did not
588 			 * forbid loopback, loop back a copy.
589 			 */
590 			KASSERT(dst != NULL);
591 			ip6_mloopback(ifp, m, dst);
592 		} else {
593 			/*
594 			 * If we are acting as a multicast router, perform
595 			 * multicast forwarding as if the packet had just
596 			 * arrived on the interface to which we are about
597 			 * to send.  The multicast forwarding function
598 			 * recursively calls this function, using the
599 			 * IPV6_FORWARDING flag to prevent infinite recursion.
600 			 *
601 			 * Multicasts that are looped back by ip6_mloopback(),
602 			 * above, will be forwarded by the ip6_input() routine,
603 			 * if necessary.
604 			 */
605 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
606 				if (ip6_mforward(ip6, ifp, m) != 0) {
607 					m_freem(m);
608 					goto done;
609 				}
610 			}
611 		}
612 		/*
613 		 * Multicasts with a hoplimit of zero may be looped back,
614 		 * above, but must not be transmitted on a network.
615 		 * Also, multicasts addressed to the loopback interface
616 		 * are not sent -- the above call to ip6_mloopback() will
617 		 * loop back a copy if this host actually belongs to the
618 		 * destination group on the loopback interface.
619 		 */
620 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
621 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
622 			m_freem(m);
623 			goto done;
624 		}
625 	}
626 
627 	/*
628 	 * Fill the outgoing inteface to tell the upper layer
629 	 * to increment per-interface statistics.
630 	 */
631 	if (ifpp)
632 		*ifpp = ifp;
633 
634 	/* Determine path MTU. */
635 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
636 	    &alwaysfrag)) != 0)
637 		goto bad;
638 
639 	/*
640 	 * The caller of this function may specify to use the minimum MTU
641 	 * in some cases.
642 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
643 	 * setting.  The logic is a bit complicated; by default, unicast
644 	 * packets will follow path MTU while multicast packets will be sent at
645 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
646 	 * including unicast ones will be sent at the minimum MTU.  Multicast
647 	 * packets will always be sent at the minimum MTU unless
648 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
649 	 * See RFC 3542 for more details.
650 	 */
651 	if (mtu > IPV6_MMTU) {
652 		if ((flags & IPV6_MINMTU))
653 			mtu = IPV6_MMTU;
654 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
655 			mtu = IPV6_MMTU;
656 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
657 			 (opt == NULL ||
658 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
659 			mtu = IPV6_MMTU;
660 		}
661 	}
662 
663 	/*
664 	 * clear embedded scope identifiers if necessary.
665 	 * in6_clearscope will touch the addresses only when necessary.
666 	 */
667 	in6_clearscope(&ip6->ip6_src);
668 	in6_clearscope(&ip6->ip6_dst);
669 
670 	/*
671 	 * If the outgoing packet contains a hop-by-hop options header,
672 	 * it must be examined and processed even by the source node.
673 	 * (RFC 2460, section 4.)
674 	 */
675 	if (ip6->ip6_nxt == IPV6_HOPOPTS) {
676 		u_int32_t dummy1; /* XXX unused */
677 		u_int32_t dummy2; /* XXX unused */
678 		int hoff = sizeof(struct ip6_hdr);
679 
680 		if (ip6_hopopts_input(&dummy1, &dummy2, &m, &hoff)) {
681 			/* m was already freed at this point */
682 			error = EINVAL;/* better error? */
683 			goto done;
684 		}
685 
686 		ip6 = mtod(m, struct ip6_hdr *);
687 	}
688 
689 	/*
690 	 * Run through list of hooks for output packets.
691 	 */
692 	if ((error = pfil_run_hooks(inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0)
693 		goto done;
694 	if (m == NULL)
695 		goto done;
696 	ip6 = mtod(m, struct ip6_hdr *);
697 
698 	/*
699 	 * Send the packet to the outgoing interface.
700 	 * If necessary, do IPv6 fragmentation before sending.
701 	 *
702 	 * the logic here is rather complex:
703 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
704 	 * 1-a:	send as is if tlen <= path mtu
705 	 * 1-b:	fragment if tlen > path mtu
706 	 *
707 	 * 2: if user asks us not to fragment (dontfrag == 1)
708 	 * 2-a:	send as is if tlen <= interface mtu
709 	 * 2-b:	error if tlen > interface mtu
710 	 *
711 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
712 	 *	always fragment
713 	 *
714 	 * 4: if dontfrag == 1 && alwaysfrag == 1
715 	 *	error, as we cannot handle this conflicting request
716 	 */
717 	tlen = m->m_pkthdr.len;
718 	tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0;
719 	if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
720 		dontfrag = 1;
721 	else
722 		dontfrag = 0;
723 
724 	if (dontfrag && alwaysfrag) {	/* case 4 */
725 		/* conflicting request - can't transmit */
726 		error = EMSGSIZE;
727 		goto bad;
728 	}
729 	if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) {	/* case 2-b */
730 		/*
731 		 * Even if the DONTFRAG option is specified, we cannot send the
732 		 * packet when the data length is larger than the MTU of the
733 		 * outgoing interface.
734 		 * Notify the error by sending IPV6_PATHMTU ancillary data as
735 		 * well as returning an error code (the latter is not described
736 		 * in the API spec.)
737 		 */
738 		u_int32_t mtu32;
739 		struct ip6ctlparam ip6cp;
740 
741 		mtu32 = (u_int32_t)mtu;
742 		memset(&ip6cp, 0, sizeof(ip6cp));
743 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
744 		pfctlinput2(PRC_MSGSIZE,
745 		    rtcache_getdst(ro_pmtu), &ip6cp);
746 
747 		error = EMSGSIZE;
748 		goto bad;
749 	}
750 
751 	/*
752 	 * transmit packet without fragmentation
753 	 */
754 	if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) {
755 		/* case 1-a and 2-a */
756 		struct in6_ifaddr *ia6;
757 		int sw_csum;
758 
759 		ip6 = mtod(m, struct ip6_hdr *);
760 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
761 		if (ia6) {
762 			/* Record statistics for this interface address. */
763 			ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len;
764 		}
765 
766 		sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
767 		if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
768 			if (IN6_NEED_CHECKSUM(ifp,
769 			    sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) {
770 				in6_delayed_cksum(m);
771 			}
772 			m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
773 		}
774 
775 		KASSERT(dst != NULL);
776 		if (__predict_true(!tso ||
777 		    (ifp->if_capenable & IFCAP_TSOv6) != 0)) {
778 			error = nd6_output(ifp, origifp, m, dst, rt);
779 		} else {
780 			error = ip6_tso_output(ifp, origifp, m, dst, rt);
781 		}
782 		goto done;
783 	}
784 
785 	if (tso) {
786 		error = EINVAL; /* XXX */
787 		goto bad;
788 	}
789 
790 	/*
791 	 * try to fragment the packet.  case 1-b and 3
792 	 */
793 	if (mtu < IPV6_MMTU) {
794 		/* path MTU cannot be less than IPV6_MMTU */
795 		error = EMSGSIZE;
796 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
797 		goto bad;
798 	} else if (ip6->ip6_plen == 0) {
799 		/* jumbo payload cannot be fragmented */
800 		error = EMSGSIZE;
801 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
802 		goto bad;
803 	} else {
804 		struct mbuf **mnext, *m_frgpart;
805 		struct ip6_frag *ip6f;
806 		u_int32_t id = htonl(ip6_randomid());
807 		u_char nextproto;
808 #if 0				/* see below */
809 		struct ip6ctlparam ip6cp;
810 		u_int32_t mtu32;
811 #endif
812 
813 		/*
814 		 * Too large for the destination or interface;
815 		 * fragment if possible.
816 		 * Must be able to put at least 8 bytes per fragment.
817 		 */
818 		hlen = unfragpartlen;
819 		if (mtu > IPV6_MAXPACKET)
820 			mtu = IPV6_MAXPACKET;
821 
822 #if 0
823 		/*
824 		 * It is believed this code is a leftover from the
825 		 * development of the IPV6_RECVPATHMTU sockopt and
826 		 * associated work to implement RFC3542.
827 		 * It's not entirely clear what the intent of the API
828 		 * is at this point, so disable this code for now.
829 		 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG
830 		 * will send notifications if the application requests.
831 		 */
832 
833 		/* Notify a proper path MTU to applications. */
834 		mtu32 = (u_int32_t)mtu;
835 		memset(&ip6cp, 0, sizeof(ip6cp));
836 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
837 		pfctlinput2(PRC_MSGSIZE,
838 		    rtcache_getdst(ro_pmtu), &ip6cp);
839 #endif
840 
841 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
842 		if (len < 8) {
843 			error = EMSGSIZE;
844 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
845 			goto bad;
846 		}
847 
848 		mnext = &m->m_nextpkt;
849 
850 		/*
851 		 * Change the next header field of the last header in the
852 		 * unfragmentable part.
853 		 */
854 		if (exthdrs.ip6e_rthdr) {
855 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
856 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
857 		} else if (exthdrs.ip6e_dest1) {
858 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
859 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
860 		} else if (exthdrs.ip6e_hbh) {
861 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
862 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
863 		} else {
864 			nextproto = ip6->ip6_nxt;
865 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
866 		}
867 
868 		if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6))
869 		    != 0) {
870 			if (IN6_NEED_CHECKSUM(ifp,
871 			    m->m_pkthdr.csum_flags &
872 			    (M_CSUM_UDPv6|M_CSUM_TCPv6))) {
873 				in6_delayed_cksum(m);
874 			}
875 			m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
876 		}
877 
878 		/*
879 		 * Loop through length of segment after first fragment,
880 		 * make new header and copy data of each part and link onto
881 		 * chain.
882 		 */
883 		m0 = m;
884 		for (off = hlen; off < tlen; off += len) {
885 			struct mbuf *mlast;
886 
887 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
888 			if (!m) {
889 				error = ENOBUFS;
890 				IP6_STATINC(IP6_STAT_ODROPPED);
891 				goto sendorfree;
892 			}
893 			m->m_pkthdr.rcvif = NULL;
894 			m->m_flags = m0->m_flags & M_COPYFLAGS;
895 			*mnext = m;
896 			mnext = &m->m_nextpkt;
897 			m->m_data += max_linkhdr;
898 			mhip6 = mtod(m, struct ip6_hdr *);
899 			*mhip6 = *ip6;
900 			m->m_len = sizeof(*mhip6);
901 			/*
902 			 * ip6f must be valid if error is 0.  But how
903 			 * can a compiler be expected to infer this?
904 			 */
905 			ip6f = NULL;
906 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
907 			if (error) {
908 				IP6_STATINC(IP6_STAT_ODROPPED);
909 				goto sendorfree;
910 			}
911 			ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7));
912 			if (off + len >= tlen)
913 				len = tlen - off;
914 			else
915 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
916 			mhip6->ip6_plen = htons((u_int16_t)(len + hlen +
917 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
918 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
919 				error = ENOBUFS;
920 				IP6_STATINC(IP6_STAT_ODROPPED);
921 				goto sendorfree;
922 			}
923 			for (mlast = m; mlast->m_next; mlast = mlast->m_next)
924 				;
925 			mlast->m_next = m_frgpart;
926 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
927 			m->m_pkthdr.rcvif = NULL;
928 			ip6f->ip6f_reserved = 0;
929 			ip6f->ip6f_ident = id;
930 			ip6f->ip6f_nxt = nextproto;
931 			IP6_STATINC(IP6_STAT_OFRAGMENTS);
932 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
933 		}
934 
935 		in6_ifstat_inc(ifp, ifs6_out_fragok);
936 	}
937 
938 	/*
939 	 * Remove leading garbages.
940 	 */
941 sendorfree:
942 	m = m0->m_nextpkt;
943 	m0->m_nextpkt = 0;
944 	m_freem(m0);
945 	for (m0 = m; m; m = m0) {
946 		m0 = m->m_nextpkt;
947 		m->m_nextpkt = 0;
948 		if (error == 0) {
949 			struct in6_ifaddr *ia6;
950 			ip6 = mtod(m, struct ip6_hdr *);
951 			ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
952 			if (ia6) {
953 				/*
954 				 * Record statistics for this interface
955 				 * address.
956 				 */
957 				ia6->ia_ifa.ifa_data.ifad_outbytes +=
958 				    m->m_pkthdr.len;
959 			}
960 			KASSERT(dst != NULL);
961 			error = nd6_output(ifp, origifp, m, dst, rt);
962 		} else
963 			m_freem(m);
964 	}
965 
966 	if (error == 0)
967 		IP6_STATINC(IP6_STAT_FRAGMENTED);
968 
969 done:
970 	rtcache_free(&ip6route);
971 
972 #ifdef IPSEC
973 	if (sp != NULL)
974 		KEY_FREESP(&sp);
975 #endif /* IPSEC */
976 
977 
978 	return (error);
979 
980 freehdrs:
981 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
982 	m_freem(exthdrs.ip6e_dest1);
983 	m_freem(exthdrs.ip6e_rthdr);
984 	m_freem(exthdrs.ip6e_dest2);
985 	/* FALLTHROUGH */
986 bad:
987 	m_freem(m);
988 	goto done;
989 badscope:
990 	IP6_STATINC(IP6_STAT_BADSCOPE);
991 	in6_ifstat_inc(origifp, ifs6_out_discard);
992 	if (error == 0)
993 		error = EHOSTUNREACH; /* XXX */
994 	goto bad;
995 }
996 
997 static int
998 ip6_copyexthdr(struct mbuf **mp, void *hdr, int hlen)
999 {
1000 	struct mbuf *m;
1001 
1002 	if (hlen > MCLBYTES)
1003 		return (ENOBUFS); /* XXX */
1004 
1005 	MGET(m, M_DONTWAIT, MT_DATA);
1006 	if (!m)
1007 		return (ENOBUFS);
1008 
1009 	if (hlen > MLEN) {
1010 		MCLGET(m, M_DONTWAIT);
1011 		if ((m->m_flags & M_EXT) == 0) {
1012 			m_free(m);
1013 			return (ENOBUFS);
1014 		}
1015 	}
1016 	m->m_len = hlen;
1017 	if (hdr)
1018 		bcopy(hdr, mtod(m, void *), hlen);
1019 
1020 	*mp = m;
1021 	return (0);
1022 }
1023 
1024 /*
1025  * Process a delayed payload checksum calculation.
1026  */
1027 void
1028 in6_delayed_cksum(struct mbuf *m)
1029 {
1030 	uint16_t csum, offset;
1031 
1032 	KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0);
1033 	KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0);
1034 	KASSERT((m->m_pkthdr.csum_flags
1035 	    & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0);
1036 
1037 	offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data);
1038 	csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset);
1039 	if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) {
1040 		csum = 0xffff;
1041 	}
1042 
1043 	offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data);
1044 	if ((offset + sizeof(csum)) > m->m_len) {
1045 		m_copyback(m, offset, sizeof(csum), &csum);
1046 	} else {
1047 		*(uint16_t *)(mtod(m, char *) + offset) = csum;
1048 	}
1049 }
1050 
1051 /*
1052  * Insert jumbo payload option.
1053  */
1054 static int
1055 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1056 {
1057 	struct mbuf *mopt;
1058 	u_int8_t *optbuf;
1059 	u_int32_t v;
1060 
1061 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1062 
1063 	/*
1064 	 * If there is no hop-by-hop options header, allocate new one.
1065 	 * If there is one but it doesn't have enough space to store the
1066 	 * jumbo payload option, allocate a cluster to store the whole options.
1067 	 * Otherwise, use it to store the options.
1068 	 */
1069 	if (exthdrs->ip6e_hbh == 0) {
1070 		MGET(mopt, M_DONTWAIT, MT_DATA);
1071 		if (mopt == 0)
1072 			return (ENOBUFS);
1073 		mopt->m_len = JUMBOOPTLEN;
1074 		optbuf = mtod(mopt, u_int8_t *);
1075 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1076 		exthdrs->ip6e_hbh = mopt;
1077 	} else {
1078 		struct ip6_hbh *hbh;
1079 
1080 		mopt = exthdrs->ip6e_hbh;
1081 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1082 			/*
1083 			 * XXX assumption:
1084 			 * - exthdrs->ip6e_hbh is not referenced from places
1085 			 *   other than exthdrs.
1086 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1087 			 */
1088 			int oldoptlen = mopt->m_len;
1089 			struct mbuf *n;
1090 
1091 			/*
1092 			 * XXX: give up if the whole (new) hbh header does
1093 			 * not fit even in an mbuf cluster.
1094 			 */
1095 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1096 				return (ENOBUFS);
1097 
1098 			/*
1099 			 * As a consequence, we must always prepare a cluster
1100 			 * at this point.
1101 			 */
1102 			MGET(n, M_DONTWAIT, MT_DATA);
1103 			if (n) {
1104 				MCLGET(n, M_DONTWAIT);
1105 				if ((n->m_flags & M_EXT) == 0) {
1106 					m_freem(n);
1107 					n = NULL;
1108 				}
1109 			}
1110 			if (!n)
1111 				return (ENOBUFS);
1112 			n->m_len = oldoptlen + JUMBOOPTLEN;
1113 			bcopy(mtod(mopt, void *), mtod(n, void *),
1114 			    oldoptlen);
1115 			optbuf = mtod(n, u_int8_t *) + oldoptlen;
1116 			m_freem(mopt);
1117 			mopt = exthdrs->ip6e_hbh = n;
1118 		} else {
1119 			optbuf = mtod(mopt, u_int8_t *) + mopt->m_len;
1120 			mopt->m_len += JUMBOOPTLEN;
1121 		}
1122 		optbuf[0] = IP6OPT_PADN;
1123 		optbuf[1] = 0;
1124 
1125 		/*
1126 		 * Adjust the header length according to the pad and
1127 		 * the jumbo payload option.
1128 		 */
1129 		hbh = mtod(mopt, struct ip6_hbh *);
1130 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1131 	}
1132 
1133 	/* fill in the option. */
1134 	optbuf[2] = IP6OPT_JUMBO;
1135 	optbuf[3] = 4;
1136 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1137 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1138 
1139 	/* finally, adjust the packet header length */
1140 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1141 
1142 	return (0);
1143 #undef JUMBOOPTLEN
1144 }
1145 
1146 /*
1147  * Insert fragment header and copy unfragmentable header portions.
1148  *
1149  * *frghdrp will not be read, and it is guaranteed that either an
1150  * error is returned or that *frghdrp will point to space allocated
1151  * for the fragment header.
1152  */
1153 static int
1154 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1155 	struct ip6_frag **frghdrp)
1156 {
1157 	struct mbuf *n, *mlast;
1158 
1159 	if (hlen > sizeof(struct ip6_hdr)) {
1160 		n = m_copym(m0, sizeof(struct ip6_hdr),
1161 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1162 		if (n == 0)
1163 			return (ENOBUFS);
1164 		m->m_next = n;
1165 	} else
1166 		n = m;
1167 
1168 	/* Search for the last mbuf of unfragmentable part. */
1169 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1170 		;
1171 
1172 	if ((mlast->m_flags & M_EXT) == 0 &&
1173 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1174 		/* use the trailing space of the last mbuf for the fragment hdr */
1175 		*frghdrp = (struct ip6_frag *)(mtod(mlast, char *) +
1176 		    mlast->m_len);
1177 		mlast->m_len += sizeof(struct ip6_frag);
1178 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1179 	} else {
1180 		/* allocate a new mbuf for the fragment header */
1181 		struct mbuf *mfrg;
1182 
1183 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1184 		if (mfrg == 0)
1185 			return (ENOBUFS);
1186 		mfrg->m_len = sizeof(struct ip6_frag);
1187 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1188 		mlast->m_next = mfrg;
1189 	}
1190 
1191 	return (0);
1192 }
1193 
1194 static int
1195 ip6_getpmtu(struct route *ro_pmtu, struct route *ro, struct ifnet *ifp,
1196     const struct in6_addr *dst, u_long *mtup, int *alwaysfragp)
1197 {
1198 	struct rtentry *rt;
1199 	u_int32_t mtu = 0;
1200 	int alwaysfrag = 0;
1201 	int error = 0;
1202 
1203 	if (ro_pmtu != ro) {
1204 		union {
1205 			struct sockaddr		dst;
1206 			struct sockaddr_in6	dst6;
1207 		} u;
1208 
1209 		/* The first hop and the final destination may differ. */
1210 		sockaddr_in6_init(&u.dst6, dst, 0, 0, 0);
1211 		rt = rtcache_lookup(ro_pmtu, &u.dst);
1212 	} else
1213 		rt = rtcache_validate(ro_pmtu);
1214 	if (rt != NULL) {
1215 		u_int32_t ifmtu;
1216 
1217 		if (ifp == NULL)
1218 			ifp = rt->rt_ifp;
1219 		ifmtu = IN6_LINKMTU(ifp);
1220 		mtu = rt->rt_rmx.rmx_mtu;
1221 		if (mtu == 0)
1222 			mtu = ifmtu;
1223 		else if (mtu < IPV6_MMTU) {
1224 			/*
1225 			 * RFC2460 section 5, last paragraph:
1226 			 * if we record ICMPv6 too big message with
1227 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1228 			 * or smaller, with fragment header attached.
1229 			 * (fragment header is needed regardless from the
1230 			 * packet size, for translators to identify packets)
1231 			 */
1232 			alwaysfrag = 1;
1233 			mtu = IPV6_MMTU;
1234 		} else if (mtu > ifmtu) {
1235 			/*
1236 			 * The MTU on the route is larger than the MTU on
1237 			 * the interface!  This shouldn't happen, unless the
1238 			 * MTU of the interface has been changed after the
1239 			 * interface was brought up.  Change the MTU in the
1240 			 * route to match the interface MTU (as long as the
1241 			 * field isn't locked).
1242 			 */
1243 			mtu = ifmtu;
1244 			if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
1245 				rt->rt_rmx.rmx_mtu = mtu;
1246 		}
1247 	} else if (ifp) {
1248 		mtu = IN6_LINKMTU(ifp);
1249 	} else
1250 		error = EHOSTUNREACH; /* XXX */
1251 
1252 	*mtup = mtu;
1253 	if (alwaysfragp)
1254 		*alwaysfragp = alwaysfrag;
1255 	return (error);
1256 }
1257 
1258 /*
1259  * IP6 socket option processing.
1260  */
1261 int
1262 ip6_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1263 {
1264 	int optdatalen, uproto;
1265 	void *optdata;
1266 	struct in6pcb *in6p = sotoin6pcb(so);
1267 	int error, optval;
1268 	int level, optname;
1269 
1270 	KASSERT(sopt != NULL);
1271 
1272 	level = sopt->sopt_level;
1273 	optname = sopt->sopt_name;
1274 
1275 	error = optval = 0;
1276 	uproto = (int)so->so_proto->pr_protocol;
1277 
1278 	if (level != IPPROTO_IPV6) {
1279 		return ENOPROTOOPT;
1280 	}
1281 	switch (op) {
1282 	case PRCO_SETOPT:
1283 		switch (optname) {
1284 #ifdef RFC2292
1285 		case IPV6_2292PKTOPTIONS:
1286 			error = ip6_pcbopts(&in6p->in6p_outputopts, so, sopt);
1287 			break;
1288 #endif
1289 
1290 		/*
1291 		 * Use of some Hop-by-Hop options or some
1292 		 * Destination options, might require special
1293 		 * privilege.  That is, normal applications
1294 		 * (without special privilege) might be forbidden
1295 		 * from setting certain options in outgoing packets,
1296 		 * and might never see certain options in received
1297 		 * packets. [RFC 2292 Section 6]
1298 		 * KAME specific note:
1299 		 *  KAME prevents non-privileged users from sending or
1300 		 *  receiving ANY hbh/dst options in order to avoid
1301 		 *  overhead of parsing options in the kernel.
1302 		 */
1303 		case IPV6_RECVHOPOPTS:
1304 		case IPV6_RECVDSTOPTS:
1305 		case IPV6_RECVRTHDRDSTOPTS:
1306 			error = kauth_authorize_network(kauth_cred_get(),
1307 			    KAUTH_NETWORK_IPV6, KAUTH_REQ_NETWORK_IPV6_HOPBYHOP,
1308 			    NULL, NULL, NULL);
1309 			if (error)
1310 				break;
1311 			/* FALLTHROUGH */
1312 		case IPV6_UNICAST_HOPS:
1313 		case IPV6_HOPLIMIT:
1314 		case IPV6_FAITH:
1315 
1316 		case IPV6_RECVPKTINFO:
1317 		case IPV6_RECVHOPLIMIT:
1318 		case IPV6_RECVRTHDR:
1319 		case IPV6_RECVPATHMTU:
1320 		case IPV6_RECVTCLASS:
1321 		case IPV6_V6ONLY:
1322 			error = sockopt_getint(sopt, &optval);
1323 			if (error)
1324 				break;
1325 			switch (optname) {
1326 			case IPV6_UNICAST_HOPS:
1327 				if (optval < -1 || optval >= 256)
1328 					error = EINVAL;
1329 				else {
1330 					/* -1 = kernel default */
1331 					in6p->in6p_hops = optval;
1332 				}
1333 				break;
1334 #define OPTSET(bit) \
1335 do { \
1336 if (optval) \
1337 	in6p->in6p_flags |= (bit); \
1338 else \
1339 	in6p->in6p_flags &= ~(bit); \
1340 } while (/*CONSTCOND*/ 0)
1341 
1342 #ifdef RFC2292
1343 #define OPTSET2292(bit) 			\
1344 do { 						\
1345 in6p->in6p_flags |= IN6P_RFC2292; 	\
1346 if (optval) 				\
1347 	in6p->in6p_flags |= (bit); 	\
1348 else 					\
1349 	in6p->in6p_flags &= ~(bit); 	\
1350 } while (/*CONSTCOND*/ 0)
1351 #endif
1352 
1353 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1354 
1355 			case IPV6_RECVPKTINFO:
1356 #ifdef RFC2292
1357 				/* cannot mix with RFC2292 */
1358 				if (OPTBIT(IN6P_RFC2292)) {
1359 					error = EINVAL;
1360 					break;
1361 				}
1362 #endif
1363 				OPTSET(IN6P_PKTINFO);
1364 				break;
1365 
1366 			case IPV6_HOPLIMIT:
1367 			{
1368 				struct ip6_pktopts **optp;
1369 
1370 #ifdef RFC2292
1371 				/* cannot mix with RFC2292 */
1372 				if (OPTBIT(IN6P_RFC2292)) {
1373 					error = EINVAL;
1374 					break;
1375 				}
1376 #endif
1377 				optp = &in6p->in6p_outputopts;
1378 				error = ip6_pcbopt(IPV6_HOPLIMIT,
1379 						   (u_char *)&optval,
1380 						   sizeof(optval),
1381 						   optp,
1382 						   kauth_cred_get(), uproto);
1383 				break;
1384 			}
1385 
1386 			case IPV6_RECVHOPLIMIT:
1387 #ifdef RFC2292
1388 				/* cannot mix with RFC2292 */
1389 				if (OPTBIT(IN6P_RFC2292)) {
1390 					error = EINVAL;
1391 					break;
1392 				}
1393 #endif
1394 				OPTSET(IN6P_HOPLIMIT);
1395 				break;
1396 
1397 			case IPV6_RECVHOPOPTS:
1398 #ifdef RFC2292
1399 				/* cannot mix with RFC2292 */
1400 				if (OPTBIT(IN6P_RFC2292)) {
1401 					error = EINVAL;
1402 					break;
1403 				}
1404 #endif
1405 				OPTSET(IN6P_HOPOPTS);
1406 				break;
1407 
1408 			case IPV6_RECVDSTOPTS:
1409 #ifdef RFC2292
1410 				/* cannot mix with RFC2292 */
1411 				if (OPTBIT(IN6P_RFC2292)) {
1412 					error = EINVAL;
1413 					break;
1414 				}
1415 #endif
1416 				OPTSET(IN6P_DSTOPTS);
1417 				break;
1418 
1419 			case IPV6_RECVRTHDRDSTOPTS:
1420 #ifdef RFC2292
1421 				/* cannot mix with RFC2292 */
1422 				if (OPTBIT(IN6P_RFC2292)) {
1423 					error = EINVAL;
1424 					break;
1425 				}
1426 #endif
1427 				OPTSET(IN6P_RTHDRDSTOPTS);
1428 				break;
1429 
1430 			case IPV6_RECVRTHDR:
1431 #ifdef RFC2292
1432 				/* cannot mix with RFC2292 */
1433 				if (OPTBIT(IN6P_RFC2292)) {
1434 					error = EINVAL;
1435 					break;
1436 				}
1437 #endif
1438 				OPTSET(IN6P_RTHDR);
1439 				break;
1440 
1441 			case IPV6_FAITH:
1442 				OPTSET(IN6P_FAITH);
1443 				break;
1444 
1445 			case IPV6_RECVPATHMTU:
1446 				/*
1447 				 * We ignore this option for TCP
1448 				 * sockets.
1449 				 * (RFC3542 leaves this case
1450 				 * unspecified.)
1451 				 */
1452 				if (uproto != IPPROTO_TCP)
1453 					OPTSET(IN6P_MTU);
1454 				break;
1455 
1456 			case IPV6_V6ONLY:
1457 				/*
1458 				 * make setsockopt(IPV6_V6ONLY)
1459 				 * available only prior to bind(2).
1460 				 * see ipng mailing list, Jun 22 2001.
1461 				 */
1462 				if (in6p->in6p_lport ||
1463 				    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1464 					error = EINVAL;
1465 					break;
1466 				}
1467 #ifdef INET6_BINDV6ONLY
1468 				if (!optval)
1469 					error = EINVAL;
1470 #else
1471 				OPTSET(IN6P_IPV6_V6ONLY);
1472 #endif
1473 				break;
1474 			case IPV6_RECVTCLASS:
1475 #ifdef RFC2292
1476 				/* cannot mix with RFC2292 XXX */
1477 				if (OPTBIT(IN6P_RFC2292)) {
1478 					error = EINVAL;
1479 					break;
1480 				}
1481 #endif
1482 				OPTSET(IN6P_TCLASS);
1483 				break;
1484 
1485 			}
1486 			break;
1487 
1488 		case IPV6_OTCLASS:
1489 		{
1490 			struct ip6_pktopts **optp;
1491 			u_int8_t tclass;
1492 
1493 			error = sockopt_get(sopt, &tclass, sizeof(tclass));
1494 			if (error)
1495 				break;
1496 			optp = &in6p->in6p_outputopts;
1497 			error = ip6_pcbopt(optname,
1498 					   (u_char *)&tclass,
1499 					   sizeof(tclass),
1500 					   optp,
1501 					   kauth_cred_get(), uproto);
1502 			break;
1503 		}
1504 
1505 		case IPV6_TCLASS:
1506 		case IPV6_DONTFRAG:
1507 		case IPV6_USE_MIN_MTU:
1508 			error = sockopt_getint(sopt, &optval);
1509 			if (error)
1510 				break;
1511 			{
1512 				struct ip6_pktopts **optp;
1513 				optp = &in6p->in6p_outputopts;
1514 				error = ip6_pcbopt(optname,
1515 						   (u_char *)&optval,
1516 						   sizeof(optval),
1517 						   optp,
1518 						   kauth_cred_get(), uproto);
1519 				break;
1520 			}
1521 
1522 #ifdef RFC2292
1523 		case IPV6_2292PKTINFO:
1524 		case IPV6_2292HOPLIMIT:
1525 		case IPV6_2292HOPOPTS:
1526 		case IPV6_2292DSTOPTS:
1527 		case IPV6_2292RTHDR:
1528 			/* RFC 2292 */
1529 			error = sockopt_getint(sopt, &optval);
1530 			if (error)
1531 				break;
1532 
1533 			switch (optname) {
1534 			case IPV6_2292PKTINFO:
1535 				OPTSET2292(IN6P_PKTINFO);
1536 				break;
1537 			case IPV6_2292HOPLIMIT:
1538 				OPTSET2292(IN6P_HOPLIMIT);
1539 				break;
1540 			case IPV6_2292HOPOPTS:
1541 				/*
1542 				 * Check super-user privilege.
1543 				 * See comments for IPV6_RECVHOPOPTS.
1544 				 */
1545 				error =
1546 				    kauth_authorize_network(kauth_cred_get(),
1547 				    KAUTH_NETWORK_IPV6,
1548 				    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL,
1549 				    NULL, NULL);
1550 				if (error)
1551 					return (error);
1552 				OPTSET2292(IN6P_HOPOPTS);
1553 				break;
1554 			case IPV6_2292DSTOPTS:
1555 				error =
1556 				    kauth_authorize_network(kauth_cred_get(),
1557 				    KAUTH_NETWORK_IPV6,
1558 				    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL,
1559 				    NULL, NULL);
1560 				if (error)
1561 					return (error);
1562 				OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1563 				break;
1564 			case IPV6_2292RTHDR:
1565 				OPTSET2292(IN6P_RTHDR);
1566 				break;
1567 			}
1568 			break;
1569 #endif
1570 		case IPV6_PKTINFO:
1571 		case IPV6_HOPOPTS:
1572 		case IPV6_RTHDR:
1573 		case IPV6_DSTOPTS:
1574 		case IPV6_RTHDRDSTOPTS:
1575 		case IPV6_NEXTHOP: {
1576 			/* new advanced API (RFC3542) */
1577 			void *optbuf;
1578 			int optbuflen;
1579 			struct ip6_pktopts **optp;
1580 
1581 #ifdef RFC2292
1582 			/* cannot mix with RFC2292 */
1583 			if (OPTBIT(IN6P_RFC2292)) {
1584 				error = EINVAL;
1585 				break;
1586 			}
1587 #endif
1588 
1589 			optbuflen = sopt->sopt_size;
1590 			optbuf = malloc(optbuflen, M_IP6OPT, M_NOWAIT);
1591 			if (optbuf == NULL) {
1592 				error = ENOBUFS;
1593 				break;
1594 			}
1595 
1596 			error = sockopt_get(sopt, optbuf, optbuflen);
1597 			if (error) {
1598 				free(optbuf, M_IP6OPT);
1599 				break;
1600 			}
1601 			optp = &in6p->in6p_outputopts;
1602 			error = ip6_pcbopt(optname, optbuf, optbuflen,
1603 			    optp, kauth_cred_get(), uproto);
1604 			break;
1605 			}
1606 #undef OPTSET
1607 
1608 		case IPV6_MULTICAST_IF:
1609 		case IPV6_MULTICAST_HOPS:
1610 		case IPV6_MULTICAST_LOOP:
1611 		case IPV6_JOIN_GROUP:
1612 		case IPV6_LEAVE_GROUP:
1613 			error = ip6_setmoptions(sopt, &in6p->in6p_moptions);
1614 			break;
1615 
1616 		case IPV6_PORTRANGE:
1617 			error = sockopt_getint(sopt, &optval);
1618 			if (error)
1619 				break;
1620 
1621 			switch (optval) {
1622 			case IPV6_PORTRANGE_DEFAULT:
1623 				in6p->in6p_flags &= ~(IN6P_LOWPORT);
1624 				in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1625 				break;
1626 
1627 			case IPV6_PORTRANGE_HIGH:
1628 				in6p->in6p_flags &= ~(IN6P_LOWPORT);
1629 				in6p->in6p_flags |= IN6P_HIGHPORT;
1630 				break;
1631 
1632 			case IPV6_PORTRANGE_LOW:
1633 				in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1634 				in6p->in6p_flags |= IN6P_LOWPORT;
1635 				break;
1636 
1637 			default:
1638 				error = EINVAL;
1639 				break;
1640 			}
1641 			break;
1642 
1643 		case IPV6_PORTALGO:
1644 			error = sockopt_getint(sopt, &optval);
1645 			if (error)
1646 				break;
1647 
1648 			error = portalgo_algo_index_select(
1649 			    (struct inpcb_hdr *)in6p, optval);
1650 			break;
1651 
1652 #if defined(IPSEC)
1653 		case IPV6_IPSEC_POLICY:
1654 			if (ipsec_enabled) {
1655 				error = ipsec6_set_policy(in6p, optname,
1656 				    sopt->sopt_data, sopt->sopt_size,
1657 				    kauth_cred_get());
1658 				break;
1659 			}
1660 			/*FALLTHROUGH*/
1661 #endif /* IPSEC */
1662 
1663 		default:
1664 			error = ENOPROTOOPT;
1665 			break;
1666 		}
1667 		break;
1668 
1669 	case PRCO_GETOPT:
1670 		switch (optname) {
1671 #ifdef RFC2292
1672 		case IPV6_2292PKTOPTIONS:
1673 			/*
1674 			 * RFC3542 (effectively) deprecated the
1675 			 * semantics of the 2292-style pktoptions.
1676 			 * Since it was not reliable in nature (i.e.,
1677 			 * applications had to expect the lack of some
1678 			 * information after all), it would make sense
1679 			 * to simplify this part by always returning
1680 			 * empty data.
1681 			 */
1682 			break;
1683 #endif
1684 
1685 		case IPV6_RECVHOPOPTS:
1686 		case IPV6_RECVDSTOPTS:
1687 		case IPV6_RECVRTHDRDSTOPTS:
1688 		case IPV6_UNICAST_HOPS:
1689 		case IPV6_RECVPKTINFO:
1690 		case IPV6_RECVHOPLIMIT:
1691 		case IPV6_RECVRTHDR:
1692 		case IPV6_RECVPATHMTU:
1693 
1694 		case IPV6_FAITH:
1695 		case IPV6_V6ONLY:
1696 		case IPV6_PORTRANGE:
1697 		case IPV6_RECVTCLASS:
1698 			switch (optname) {
1699 
1700 			case IPV6_RECVHOPOPTS:
1701 				optval = OPTBIT(IN6P_HOPOPTS);
1702 				break;
1703 
1704 			case IPV6_RECVDSTOPTS:
1705 				optval = OPTBIT(IN6P_DSTOPTS);
1706 				break;
1707 
1708 			case IPV6_RECVRTHDRDSTOPTS:
1709 				optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1710 				break;
1711 
1712 			case IPV6_UNICAST_HOPS:
1713 				optval = in6p->in6p_hops;
1714 				break;
1715 
1716 			case IPV6_RECVPKTINFO:
1717 				optval = OPTBIT(IN6P_PKTINFO);
1718 				break;
1719 
1720 			case IPV6_RECVHOPLIMIT:
1721 				optval = OPTBIT(IN6P_HOPLIMIT);
1722 				break;
1723 
1724 			case IPV6_RECVRTHDR:
1725 				optval = OPTBIT(IN6P_RTHDR);
1726 				break;
1727 
1728 			case IPV6_RECVPATHMTU:
1729 				optval = OPTBIT(IN6P_MTU);
1730 				break;
1731 
1732 			case IPV6_FAITH:
1733 				optval = OPTBIT(IN6P_FAITH);
1734 				break;
1735 
1736 			case IPV6_V6ONLY:
1737 				optval = OPTBIT(IN6P_IPV6_V6ONLY);
1738 				break;
1739 
1740 			case IPV6_PORTRANGE:
1741 			    {
1742 				int flags;
1743 				flags = in6p->in6p_flags;
1744 				if (flags & IN6P_HIGHPORT)
1745 					optval = IPV6_PORTRANGE_HIGH;
1746 				else if (flags & IN6P_LOWPORT)
1747 					optval = IPV6_PORTRANGE_LOW;
1748 				else
1749 					optval = 0;
1750 				break;
1751 			    }
1752 			case IPV6_RECVTCLASS:
1753 				optval = OPTBIT(IN6P_TCLASS);
1754 				break;
1755 
1756 			}
1757 			if (error)
1758 				break;
1759 			error = sockopt_setint(sopt, optval);
1760 			break;
1761 
1762 		case IPV6_PATHMTU:
1763 		    {
1764 			u_long pmtu = 0;
1765 			struct ip6_mtuinfo mtuinfo;
1766 			struct route *ro = &in6p->in6p_route;
1767 
1768 			if (!(so->so_state & SS_ISCONNECTED))
1769 				return (ENOTCONN);
1770 			/*
1771 			 * XXX: we dot not consider the case of source
1772 			 * routing, or optional information to specify
1773 			 * the outgoing interface.
1774 			 */
1775 			error = ip6_getpmtu(ro, NULL, NULL,
1776 			    &in6p->in6p_faddr, &pmtu, NULL);
1777 			if (error)
1778 				break;
1779 			if (pmtu > IPV6_MAXPACKET)
1780 				pmtu = IPV6_MAXPACKET;
1781 
1782 			memset(&mtuinfo, 0, sizeof(mtuinfo));
1783 			mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1784 			optdata = (void *)&mtuinfo;
1785 			optdatalen = sizeof(mtuinfo);
1786 			if (optdatalen > MCLBYTES)
1787 				return (EMSGSIZE); /* XXX */
1788 			error = sockopt_set(sopt, optdata, optdatalen);
1789 			break;
1790 		    }
1791 
1792 #ifdef RFC2292
1793 		case IPV6_2292PKTINFO:
1794 		case IPV6_2292HOPLIMIT:
1795 		case IPV6_2292HOPOPTS:
1796 		case IPV6_2292RTHDR:
1797 		case IPV6_2292DSTOPTS:
1798 			switch (optname) {
1799 			case IPV6_2292PKTINFO:
1800 				optval = OPTBIT(IN6P_PKTINFO);
1801 				break;
1802 			case IPV6_2292HOPLIMIT:
1803 				optval = OPTBIT(IN6P_HOPLIMIT);
1804 				break;
1805 			case IPV6_2292HOPOPTS:
1806 				optval = OPTBIT(IN6P_HOPOPTS);
1807 				break;
1808 			case IPV6_2292RTHDR:
1809 				optval = OPTBIT(IN6P_RTHDR);
1810 				break;
1811 			case IPV6_2292DSTOPTS:
1812 				optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1813 				break;
1814 			}
1815 			error = sockopt_setint(sopt, optval);
1816 			break;
1817 #endif
1818 		case IPV6_PKTINFO:
1819 		case IPV6_HOPOPTS:
1820 		case IPV6_RTHDR:
1821 		case IPV6_DSTOPTS:
1822 		case IPV6_RTHDRDSTOPTS:
1823 		case IPV6_NEXTHOP:
1824 		case IPV6_OTCLASS:
1825 		case IPV6_TCLASS:
1826 		case IPV6_DONTFRAG:
1827 		case IPV6_USE_MIN_MTU:
1828 			error = ip6_getpcbopt(in6p->in6p_outputopts,
1829 			    optname, sopt);
1830 			break;
1831 
1832 		case IPV6_MULTICAST_IF:
1833 		case IPV6_MULTICAST_HOPS:
1834 		case IPV6_MULTICAST_LOOP:
1835 		case IPV6_JOIN_GROUP:
1836 		case IPV6_LEAVE_GROUP:
1837 			error = ip6_getmoptions(sopt, in6p->in6p_moptions);
1838 			break;
1839 
1840 		case IPV6_PORTALGO:
1841 			optval = ((struct inpcb_hdr *)in6p)->inph_portalgo;
1842 			error = sockopt_setint(sopt, optval);
1843 			break;
1844 
1845 #if defined(IPSEC)
1846 		case IPV6_IPSEC_POLICY:
1847 			if (ipsec_used) {
1848 				struct mbuf *m = NULL;
1849 
1850 				/*
1851 				 * XXX: this will return EINVAL as sopt is
1852 				 * empty
1853 				 */
1854 				error = ipsec6_get_policy(in6p, sopt->sopt_data,
1855 				    sopt->sopt_size, &m);
1856 				if (!error)
1857 					error = sockopt_setmbuf(sopt, m);
1858 				break;
1859 			}
1860 			/*FALLTHROUGH*/
1861 #endif /* IPSEC */
1862 
1863 		default:
1864 			error = ENOPROTOOPT;
1865 			break;
1866 		}
1867 		break;
1868 	}
1869 	return (error);
1870 }
1871 
1872 int
1873 ip6_raw_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1874 {
1875 	int error = 0, optval;
1876 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
1877 	struct in6pcb *in6p = sotoin6pcb(so);
1878 	int level, optname;
1879 
1880 	KASSERT(sopt != NULL);
1881 
1882 	level = sopt->sopt_level;
1883 	optname = sopt->sopt_name;
1884 
1885 	if (level != IPPROTO_IPV6) {
1886 		return ENOPROTOOPT;
1887 	}
1888 
1889 	switch (optname) {
1890 	case IPV6_CHECKSUM:
1891 		/*
1892 		 * For ICMPv6 sockets, no modification allowed for checksum
1893 		 * offset, permit "no change" values to help existing apps.
1894 		 *
1895 		 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM
1896 		 * for an ICMPv6 socket will fail."  The current
1897 		 * behavior does not meet RFC3542.
1898 		 */
1899 		switch (op) {
1900 		case PRCO_SETOPT:
1901 			error = sockopt_getint(sopt, &optval);
1902 			if (error)
1903 				break;
1904 			if ((optval % 2) != 0) {
1905 				/* the API assumes even offset values */
1906 				error = EINVAL;
1907 			} else if (so->so_proto->pr_protocol ==
1908 			    IPPROTO_ICMPV6) {
1909 				if (optval != icmp6off)
1910 					error = EINVAL;
1911 			} else
1912 				in6p->in6p_cksum = optval;
1913 			break;
1914 
1915 		case PRCO_GETOPT:
1916 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
1917 				optval = icmp6off;
1918 			else
1919 				optval = in6p->in6p_cksum;
1920 
1921 			error = sockopt_setint(sopt, optval);
1922 			break;
1923 
1924 		default:
1925 			error = EINVAL;
1926 			break;
1927 		}
1928 		break;
1929 
1930 	default:
1931 		error = ENOPROTOOPT;
1932 		break;
1933 	}
1934 
1935 	return (error);
1936 }
1937 
1938 #ifdef RFC2292
1939 /*
1940  * Set up IP6 options in pcb for insertion in output packets or
1941  * specifying behavior of outgoing packets.
1942  */
1943 static int
1944 ip6_pcbopts(struct ip6_pktopts **pktopt, struct socket *so,
1945     struct sockopt *sopt)
1946 {
1947 	struct ip6_pktopts *opt = *pktopt;
1948 	struct mbuf *m;
1949 	int error = 0;
1950 
1951 	/* turn off any old options. */
1952 	if (opt) {
1953 #ifdef DIAGNOSTIC
1954 	    if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
1955 		opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
1956 		opt->ip6po_rhinfo.ip6po_rhi_rthdr)
1957 		    printf("ip6_pcbopts: all specified options are cleared.\n");
1958 #endif
1959 		ip6_clearpktopts(opt, -1);
1960 	} else {
1961 		opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
1962 		if (opt == NULL)
1963 			return (ENOBUFS);
1964 	}
1965 	*pktopt = NULL;
1966 
1967 	if (sopt == NULL || sopt->sopt_size == 0) {
1968 		/*
1969 		 * Only turning off any previous options, regardless of
1970 		 * whether the opt is just created or given.
1971 		 */
1972 		free(opt, M_IP6OPT);
1973 		return (0);
1974 	}
1975 
1976 	/*  set options specified by user. */
1977 	m = sockopt_getmbuf(sopt);
1978 	if (m == NULL) {
1979 		free(opt, M_IP6OPT);
1980 		return (ENOBUFS);
1981 	}
1982 
1983 	error = ip6_setpktopts(m, opt, NULL, kauth_cred_get(),
1984 	    so->so_proto->pr_protocol);
1985 	m_freem(m);
1986 	if (error != 0) {
1987 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
1988 		free(opt, M_IP6OPT);
1989 		return (error);
1990 	}
1991 	*pktopt = opt;
1992 	return (0);
1993 }
1994 #endif
1995 
1996 /*
1997  * initialize ip6_pktopts.  beware that there are non-zero default values in
1998  * the struct.
1999  */
2000 void
2001 ip6_initpktopts(struct ip6_pktopts *opt)
2002 {
2003 
2004 	memset(opt, 0, sizeof(*opt));
2005 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2006 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2007 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2008 }
2009 
2010 #define sin6tosa(sin6)	((struct sockaddr *)(sin6)) /* XXX */
2011 static int
2012 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2013     kauth_cred_t cred, int uproto)
2014 {
2015 	struct ip6_pktopts *opt;
2016 
2017 	if (*pktopt == NULL) {
2018 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2019 		    M_NOWAIT);
2020 		if (*pktopt == NULL)
2021 			return (ENOBUFS);
2022 
2023 		ip6_initpktopts(*pktopt);
2024 	}
2025 	opt = *pktopt;
2026 
2027 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2028 }
2029 
2030 static int
2031 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2032 {
2033 	void *optdata = NULL;
2034 	int optdatalen = 0;
2035 	struct ip6_ext *ip6e;
2036 	int error = 0;
2037 	struct in6_pktinfo null_pktinfo;
2038 	int deftclass = 0, on;
2039 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2040 
2041 	switch (optname) {
2042 	case IPV6_PKTINFO:
2043 		if (pktopt && pktopt->ip6po_pktinfo)
2044 			optdata = (void *)pktopt->ip6po_pktinfo;
2045 		else {
2046 			/* XXX: we don't have to do this every time... */
2047 			memset(&null_pktinfo, 0, sizeof(null_pktinfo));
2048 			optdata = (void *)&null_pktinfo;
2049 		}
2050 		optdatalen = sizeof(struct in6_pktinfo);
2051 		break;
2052 	case IPV6_OTCLASS:
2053 		/* XXX */
2054 		return (EINVAL);
2055 	case IPV6_TCLASS:
2056 		if (pktopt && pktopt->ip6po_tclass >= 0)
2057 			optdata = (void *)&pktopt->ip6po_tclass;
2058 		else
2059 			optdata = (void *)&deftclass;
2060 		optdatalen = sizeof(int);
2061 		break;
2062 	case IPV6_HOPOPTS:
2063 		if (pktopt && pktopt->ip6po_hbh) {
2064 			optdata = (void *)pktopt->ip6po_hbh;
2065 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2066 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2067 		}
2068 		break;
2069 	case IPV6_RTHDR:
2070 		if (pktopt && pktopt->ip6po_rthdr) {
2071 			optdata = (void *)pktopt->ip6po_rthdr;
2072 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2073 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2074 		}
2075 		break;
2076 	case IPV6_RTHDRDSTOPTS:
2077 		if (pktopt && pktopt->ip6po_dest1) {
2078 			optdata = (void *)pktopt->ip6po_dest1;
2079 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2080 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2081 		}
2082 		break;
2083 	case IPV6_DSTOPTS:
2084 		if (pktopt && pktopt->ip6po_dest2) {
2085 			optdata = (void *)pktopt->ip6po_dest2;
2086 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2087 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2088 		}
2089 		break;
2090 	case IPV6_NEXTHOP:
2091 		if (pktopt && pktopt->ip6po_nexthop) {
2092 			optdata = (void *)pktopt->ip6po_nexthop;
2093 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2094 		}
2095 		break;
2096 	case IPV6_USE_MIN_MTU:
2097 		if (pktopt)
2098 			optdata = (void *)&pktopt->ip6po_minmtu;
2099 		else
2100 			optdata = (void *)&defminmtu;
2101 		optdatalen = sizeof(int);
2102 		break;
2103 	case IPV6_DONTFRAG:
2104 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2105 			on = 1;
2106 		else
2107 			on = 0;
2108 		optdata = (void *)&on;
2109 		optdatalen = sizeof(on);
2110 		break;
2111 	default:		/* should not happen */
2112 #ifdef DIAGNOSTIC
2113 		panic("ip6_getpcbopt: unexpected option\n");
2114 #endif
2115 		return (ENOPROTOOPT);
2116 	}
2117 
2118 	error = sockopt_set(sopt, optdata, optdatalen);
2119 
2120 	return (error);
2121 }
2122 
2123 void
2124 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2125 {
2126 	if (optname == -1 || optname == IPV6_PKTINFO) {
2127 		if (pktopt->ip6po_pktinfo)
2128 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2129 		pktopt->ip6po_pktinfo = NULL;
2130 	}
2131 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2132 		pktopt->ip6po_hlim = -1;
2133 	if (optname == -1 || optname == IPV6_TCLASS)
2134 		pktopt->ip6po_tclass = -1;
2135 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2136 		rtcache_free(&pktopt->ip6po_nextroute);
2137 		if (pktopt->ip6po_nexthop)
2138 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2139 		pktopt->ip6po_nexthop = NULL;
2140 	}
2141 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2142 		if (pktopt->ip6po_hbh)
2143 			free(pktopt->ip6po_hbh, M_IP6OPT);
2144 		pktopt->ip6po_hbh = NULL;
2145 	}
2146 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2147 		if (pktopt->ip6po_dest1)
2148 			free(pktopt->ip6po_dest1, M_IP6OPT);
2149 		pktopt->ip6po_dest1 = NULL;
2150 	}
2151 	if (optname == -1 || optname == IPV6_RTHDR) {
2152 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2153 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2154 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2155 		rtcache_free(&pktopt->ip6po_route);
2156 	}
2157 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2158 		if (pktopt->ip6po_dest2)
2159 			free(pktopt->ip6po_dest2, M_IP6OPT);
2160 		pktopt->ip6po_dest2 = NULL;
2161 	}
2162 }
2163 
2164 #define PKTOPT_EXTHDRCPY(type) 					\
2165 do {								\
2166 	if (src->type) {					\
2167 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2168 		dst->type = malloc(hlen, M_IP6OPT, canwait);	\
2169 		if (dst->type == NULL)				\
2170 			goto bad;				\
2171 		memcpy(dst->type, src->type, hlen);		\
2172 	}							\
2173 } while (/*CONSTCOND*/ 0)
2174 
2175 static int
2176 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2177 {
2178 	dst->ip6po_hlim = src->ip6po_hlim;
2179 	dst->ip6po_tclass = src->ip6po_tclass;
2180 	dst->ip6po_flags = src->ip6po_flags;
2181 	if (src->ip6po_pktinfo) {
2182 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2183 		    M_IP6OPT, canwait);
2184 		if (dst->ip6po_pktinfo == NULL)
2185 			goto bad;
2186 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2187 	}
2188 	if (src->ip6po_nexthop) {
2189 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2190 		    M_IP6OPT, canwait);
2191 		if (dst->ip6po_nexthop == NULL)
2192 			goto bad;
2193 		memcpy(dst->ip6po_nexthop, src->ip6po_nexthop,
2194 		    src->ip6po_nexthop->sa_len);
2195 	}
2196 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2197 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2198 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2199 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2200 	return (0);
2201 
2202   bad:
2203 	if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
2204 	if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
2205 	if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
2206 	if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
2207 	if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
2208 	if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
2209 
2210 	return (ENOBUFS);
2211 }
2212 #undef PKTOPT_EXTHDRCPY
2213 
2214 struct ip6_pktopts *
2215 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2216 {
2217 	int error;
2218 	struct ip6_pktopts *dst;
2219 
2220 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2221 	if (dst == NULL)
2222 		return (NULL);
2223 	ip6_initpktopts(dst);
2224 
2225 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2226 		free(dst, M_IP6OPT);
2227 		return (NULL);
2228 	}
2229 
2230 	return (dst);
2231 }
2232 
2233 void
2234 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2235 {
2236 	if (pktopt == NULL)
2237 		return;
2238 
2239 	ip6_clearpktopts(pktopt, -1);
2240 
2241 	free(pktopt, M_IP6OPT);
2242 }
2243 
2244 /*
2245  * Set the IP6 multicast options in response to user setsockopt().
2246  */
2247 static int
2248 ip6_setmoptions(const struct sockopt *sopt, struct ip6_moptions **im6op)
2249 {
2250 	int error = 0;
2251 	u_int loop, ifindex;
2252 	struct ipv6_mreq mreq;
2253 	struct ifnet *ifp;
2254 	struct ip6_moptions *im6o = *im6op;
2255 	struct route ro;
2256 	struct in6_multi_mship *imm;
2257 	struct lwp *l = curlwp;	/* XXX */
2258 
2259 	if (im6o == NULL) {
2260 		/*
2261 		 * No multicast option buffer attached to the pcb;
2262 		 * allocate one and initialize to default values.
2263 		 */
2264 		im6o = malloc(sizeof(*im6o), M_IPMOPTS, M_NOWAIT);
2265 		if (im6o == NULL)
2266 			return (ENOBUFS);
2267 
2268 		*im6op = im6o;
2269 		im6o->im6o_multicast_ifp = NULL;
2270 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2271 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2272 		LIST_INIT(&im6o->im6o_memberships);
2273 	}
2274 
2275 	switch (sopt->sopt_name) {
2276 
2277 	case IPV6_MULTICAST_IF:
2278 		/*
2279 		 * Select the interface for outgoing multicast packets.
2280 		 */
2281 		error = sockopt_get(sopt, &ifindex, sizeof(ifindex));
2282 		if (error != 0)
2283 			break;
2284 
2285 		if (ifindex != 0) {
2286 			if ((ifp = if_byindex(ifindex)) == NULL) {
2287 				error = ENXIO;	/* XXX EINVAL? */
2288 				break;
2289 			}
2290 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
2291 				error = EADDRNOTAVAIL;
2292 				break;
2293 			}
2294 		} else
2295 			ifp = NULL;
2296 		im6o->im6o_multicast_ifp = ifp;
2297 		break;
2298 
2299 	case IPV6_MULTICAST_HOPS:
2300 	    {
2301 		/*
2302 		 * Set the IP6 hoplimit for outgoing multicast packets.
2303 		 */
2304 		int optval;
2305 
2306 		error = sockopt_getint(sopt, &optval);
2307 		if (error != 0)
2308 			break;
2309 
2310 		if (optval < -1 || optval >= 256)
2311 			error = EINVAL;
2312 		else if (optval == -1)
2313 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2314 		else
2315 			im6o->im6o_multicast_hlim = optval;
2316 		break;
2317 	    }
2318 
2319 	case IPV6_MULTICAST_LOOP:
2320 		/*
2321 		 * Set the loopback flag for outgoing multicast packets.
2322 		 * Must be zero or one.
2323 		 */
2324 		error = sockopt_get(sopt, &loop, sizeof(loop));
2325 		if (error != 0)
2326 			break;
2327 		if (loop > 1) {
2328 			error = EINVAL;
2329 			break;
2330 		}
2331 		im6o->im6o_multicast_loop = loop;
2332 		break;
2333 
2334 	case IPV6_JOIN_GROUP:
2335 		/*
2336 		 * Add a multicast group membership.
2337 		 * Group must be a valid IP6 multicast address.
2338 		 */
2339 		error = sockopt_get(sopt, &mreq, sizeof(mreq));
2340 		if (error != 0)
2341 			break;
2342 
2343 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq.ipv6mr_multiaddr)) {
2344 			/*
2345 			 * We use the unspecified address to specify to accept
2346 			 * all multicast addresses. Only super user is allowed
2347 			 * to do this.
2348 			 */
2349 			if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_IPV6,
2350 			    KAUTH_REQ_NETWORK_IPV6_JOIN_MULTICAST, NULL, NULL, NULL))
2351 			{
2352 				error = EACCES;
2353 				break;
2354 			}
2355 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq.ipv6mr_multiaddr)) {
2356 			error = EINVAL;
2357 			break;
2358 		}
2359 
2360 		/*
2361 		 * If no interface was explicitly specified, choose an
2362 		 * appropriate one according to the given multicast address.
2363 		 */
2364 		if (mreq.ipv6mr_interface == 0) {
2365 			struct rtentry *rt;
2366 			union {
2367 				struct sockaddr		dst;
2368 				struct sockaddr_in6	dst6;
2369 			} u;
2370 
2371 			/*
2372 			 * Look up the routing table for the
2373 			 * address, and choose the outgoing interface.
2374 			 *   XXX: is it a good approach?
2375 			 */
2376 			memset(&ro, 0, sizeof(ro));
2377 			sockaddr_in6_init(&u.dst6, &mreq.ipv6mr_multiaddr, 0,
2378 			    0, 0);
2379 			rtcache_setdst(&ro, &u.dst);
2380 			ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp
2381 			                                        : NULL;
2382 			rtcache_free(&ro);
2383 		} else {
2384 			/*
2385 			 * If the interface is specified, validate it.
2386 			 */
2387 			if ((ifp = if_byindex(mreq.ipv6mr_interface)) == NULL) {
2388 				error = ENXIO;	/* XXX EINVAL? */
2389 				break;
2390 			}
2391 		}
2392 
2393 		/*
2394 		 * See if we found an interface, and confirm that it
2395 		 * supports multicast
2396 		 */
2397 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2398 			error = EADDRNOTAVAIL;
2399 			break;
2400 		}
2401 
2402 		if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) {
2403 			error = EADDRNOTAVAIL; /* XXX: should not happen */
2404 			break;
2405 		}
2406 
2407 		/*
2408 		 * See if the membership already exists.
2409 		 */
2410 		for (imm = im6o->im6o_memberships.lh_first;
2411 		     imm != NULL; imm = imm->i6mm_chain.le_next)
2412 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
2413 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2414 			    &mreq.ipv6mr_multiaddr))
2415 				break;
2416 		if (imm != NULL) {
2417 			error = EADDRINUSE;
2418 			break;
2419 		}
2420 		/*
2421 		 * Everything looks good; add a new record to the multicast
2422 		 * address list for the given interface.
2423 		 */
2424 		imm = in6_joingroup(ifp, &mreq.ipv6mr_multiaddr, &error, 0);
2425 		if (imm == NULL)
2426 			break;
2427 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2428 		break;
2429 
2430 	case IPV6_LEAVE_GROUP:
2431 		/*
2432 		 * Drop a multicast group membership.
2433 		 * Group must be a valid IP6 multicast address.
2434 		 */
2435 		error = sockopt_get(sopt, &mreq, sizeof(mreq));
2436 		if (error != 0)
2437 			break;
2438 
2439 		/*
2440 		 * If an interface address was specified, get a pointer
2441 		 * to its ifnet structure.
2442 		 */
2443 		if (mreq.ipv6mr_interface != 0) {
2444 			if ((ifp = if_byindex(mreq.ipv6mr_interface)) == NULL) {
2445 				error = ENXIO;	/* XXX EINVAL? */
2446 				break;
2447 			}
2448 		} else
2449 			ifp = NULL;
2450 
2451 		/* Fill in the scope zone ID */
2452 		if (ifp) {
2453 			if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) {
2454 				/* XXX: should not happen */
2455 				error = EADDRNOTAVAIL;
2456 				break;
2457 			}
2458 		} else if (mreq.ipv6mr_interface != 0) {
2459 			/*
2460 			 * XXX: This case would happens when the (positive)
2461 			 * index is in the valid range, but the corresponding
2462 			 * interface has been detached dynamically.  The above
2463 			 * check probably avoids such case to happen here, but
2464 			 * we check it explicitly for safety.
2465 			 */
2466 			error = EADDRNOTAVAIL;
2467 			break;
2468 		} else {	/* ipv6mr_interface == 0 */
2469 			struct sockaddr_in6 sa6_mc;
2470 
2471 			/*
2472 			 * The API spec says as follows:
2473 			 *  If the interface index is specified as 0, the
2474 			 *  system may choose a multicast group membership to
2475 			 *  drop by matching the multicast address only.
2476 			 * On the other hand, we cannot disambiguate the scope
2477 			 * zone unless an interface is provided.  Thus, we
2478 			 * check if there's ambiguity with the default scope
2479 			 * zone as the last resort.
2480 			 */
2481 			sockaddr_in6_init(&sa6_mc, &mreq.ipv6mr_multiaddr,
2482 			    0, 0, 0);
2483 			error = sa6_embedscope(&sa6_mc, ip6_use_defzone);
2484 			if (error != 0)
2485 				break;
2486 			mreq.ipv6mr_multiaddr = sa6_mc.sin6_addr;
2487 		}
2488 
2489 		/*
2490 		 * Find the membership in the membership list.
2491 		 */
2492 		for (imm = im6o->im6o_memberships.lh_first;
2493 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
2494 			if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2495 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2496 			    &mreq.ipv6mr_multiaddr))
2497 				break;
2498 		}
2499 		if (imm == NULL) {
2500 			/* Unable to resolve interface */
2501 			error = EADDRNOTAVAIL;
2502 			break;
2503 		}
2504 		/*
2505 		 * Give up the multicast address record to which the
2506 		 * membership points.
2507 		 */
2508 		LIST_REMOVE(imm, i6mm_chain);
2509 		in6_leavegroup(imm);
2510 		break;
2511 
2512 	default:
2513 		error = EOPNOTSUPP;
2514 		break;
2515 	}
2516 
2517 	/*
2518 	 * If all options have default values, no need to keep the mbuf.
2519 	 */
2520 	if (im6o->im6o_multicast_ifp == NULL &&
2521 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2522 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2523 	    im6o->im6o_memberships.lh_first == NULL) {
2524 		free(*im6op, M_IPMOPTS);
2525 		*im6op = NULL;
2526 	}
2527 
2528 	return (error);
2529 }
2530 
2531 /*
2532  * Return the IP6 multicast options in response to user getsockopt().
2533  */
2534 static int
2535 ip6_getmoptions(struct sockopt *sopt, struct ip6_moptions *im6o)
2536 {
2537 	u_int optval;
2538 	int error;
2539 
2540 	switch (sopt->sopt_name) {
2541 	case IPV6_MULTICAST_IF:
2542 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2543 			optval = 0;
2544 		else
2545 			optval = im6o->im6o_multicast_ifp->if_index;
2546 
2547 		error = sockopt_set(sopt, &optval, sizeof(optval));
2548 		break;
2549 
2550 	case IPV6_MULTICAST_HOPS:
2551 		if (im6o == NULL)
2552 			optval = ip6_defmcasthlim;
2553 		else
2554 			optval = im6o->im6o_multicast_hlim;
2555 
2556 		error = sockopt_set(sopt, &optval, sizeof(optval));
2557 		break;
2558 
2559 	case IPV6_MULTICAST_LOOP:
2560 		if (im6o == NULL)
2561 			optval = IPV6_DEFAULT_MULTICAST_LOOP;
2562 		else
2563 			optval = im6o->im6o_multicast_loop;
2564 
2565 		error = sockopt_set(sopt, &optval, sizeof(optval));
2566 		break;
2567 
2568 	default:
2569 		error = EOPNOTSUPP;
2570 	}
2571 
2572 	return (error);
2573 }
2574 
2575 /*
2576  * Discard the IP6 multicast options.
2577  */
2578 void
2579 ip6_freemoptions(struct ip6_moptions *im6o)
2580 {
2581 	struct in6_multi_mship *imm;
2582 
2583 	if (im6o == NULL)
2584 		return;
2585 
2586 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2587 		LIST_REMOVE(imm, i6mm_chain);
2588 		in6_leavegroup(imm);
2589 	}
2590 	free(im6o, M_IPMOPTS);
2591 }
2592 
2593 /*
2594  * Set IPv6 outgoing packet options based on advanced API.
2595  */
2596 int
2597 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2598 	struct ip6_pktopts *stickyopt, kauth_cred_t cred, int uproto)
2599 {
2600 	struct cmsghdr *cm = 0;
2601 
2602 	if (control == NULL || opt == NULL)
2603 		return (EINVAL);
2604 
2605 	ip6_initpktopts(opt);
2606 	if (stickyopt) {
2607 		int error;
2608 
2609 		/*
2610 		 * If stickyopt is provided, make a local copy of the options
2611 		 * for this particular packet, then override them by ancillary
2612 		 * objects.
2613 		 * XXX: copypktopts() does not copy the cached route to a next
2614 		 * hop (if any).  This is not very good in terms of efficiency,
2615 		 * but we can allow this since this option should be rarely
2616 		 * used.
2617 		 */
2618 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2619 			return (error);
2620 	}
2621 
2622 	/*
2623 	 * XXX: Currently, we assume all the optional information is stored
2624 	 * in a single mbuf.
2625 	 */
2626 	if (control->m_next)
2627 		return (EINVAL);
2628 
2629 	/* XXX if cm->cmsg_len is not aligned, control->m_len can become <0 */
2630 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2631 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2632 		int error;
2633 
2634 		if (control->m_len < CMSG_LEN(0))
2635 			return (EINVAL);
2636 
2637 		cm = mtod(control, struct cmsghdr *);
2638 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2639 			return (EINVAL);
2640 		if (cm->cmsg_level != IPPROTO_IPV6)
2641 			continue;
2642 
2643 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2644 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2645 		if (error)
2646 			return (error);
2647 	}
2648 
2649 	return (0);
2650 }
2651 
2652 /*
2653  * Set a particular packet option, as a sticky option or an ancillary data
2654  * item.  "len" can be 0 only when it's a sticky option.
2655  * We have 4 cases of combination of "sticky" and "cmsg":
2656  * "sticky=0, cmsg=0": impossible
2657  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2658  * "sticky=1, cmsg=0": RFC3542 socket option
2659  * "sticky=1, cmsg=1": RFC2292 socket option
2660  */
2661 static int
2662 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2663     kauth_cred_t cred, int sticky, int cmsg, int uproto)
2664 {
2665 	int minmtupolicy;
2666 	int error;
2667 
2668 	if (!sticky && !cmsg) {
2669 #ifdef DIAGNOSTIC
2670 		printf("ip6_setpktopt: impossible case\n");
2671 #endif
2672 		return (EINVAL);
2673 	}
2674 
2675 	/*
2676 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2677 	 * not be specified in the context of RFC3542.  Conversely,
2678 	 * RFC3542 types should not be specified in the context of RFC2292.
2679 	 */
2680 	if (!cmsg) {
2681 		switch (optname) {
2682 		case IPV6_2292PKTINFO:
2683 		case IPV6_2292HOPLIMIT:
2684 		case IPV6_2292NEXTHOP:
2685 		case IPV6_2292HOPOPTS:
2686 		case IPV6_2292DSTOPTS:
2687 		case IPV6_2292RTHDR:
2688 		case IPV6_2292PKTOPTIONS:
2689 			return (ENOPROTOOPT);
2690 		}
2691 	}
2692 	if (sticky && cmsg) {
2693 		switch (optname) {
2694 		case IPV6_PKTINFO:
2695 		case IPV6_HOPLIMIT:
2696 		case IPV6_NEXTHOP:
2697 		case IPV6_HOPOPTS:
2698 		case IPV6_DSTOPTS:
2699 		case IPV6_RTHDRDSTOPTS:
2700 		case IPV6_RTHDR:
2701 		case IPV6_USE_MIN_MTU:
2702 		case IPV6_DONTFRAG:
2703 		case IPV6_OTCLASS:
2704 		case IPV6_TCLASS:
2705 			return (ENOPROTOOPT);
2706 		}
2707 	}
2708 
2709 	switch (optname) {
2710 #ifdef RFC2292
2711 	case IPV6_2292PKTINFO:
2712 #endif
2713 	case IPV6_PKTINFO:
2714 	{
2715 		struct ifnet *ifp = NULL;
2716 		struct in6_pktinfo *pktinfo;
2717 
2718 		if (len != sizeof(struct in6_pktinfo))
2719 			return (EINVAL);
2720 
2721 		pktinfo = (struct in6_pktinfo *)buf;
2722 
2723 		/*
2724 		 * An application can clear any sticky IPV6_PKTINFO option by
2725 		 * doing a "regular" setsockopt with ipi6_addr being
2726 		 * in6addr_any and ipi6_ifindex being zero.
2727 		 * [RFC 3542, Section 6]
2728 		 */
2729 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2730 		    pktinfo->ipi6_ifindex == 0 &&
2731 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2732 			ip6_clearpktopts(opt, optname);
2733 			break;
2734 		}
2735 
2736 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2737 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2738 			return (EINVAL);
2739 		}
2740 
2741 		/* Validate the interface index if specified. */
2742 		if (pktinfo->ipi6_ifindex) {
2743 			ifp = if_byindex(pktinfo->ipi6_ifindex);
2744 			if (ifp == NULL)
2745 				return (ENXIO);
2746 		}
2747 
2748 		/*
2749 		 * We store the address anyway, and let in6_selectsrc()
2750 		 * validate the specified address.  This is because ipi6_addr
2751 		 * may not have enough information about its scope zone, and
2752 		 * we may need additional information (such as outgoing
2753 		 * interface or the scope zone of a destination address) to
2754 		 * disambiguate the scope.
2755 		 * XXX: the delay of the validation may confuse the
2756 		 * application when it is used as a sticky option.
2757 		 */
2758 		if (opt->ip6po_pktinfo == NULL) {
2759 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2760 			    M_IP6OPT, M_NOWAIT);
2761 			if (opt->ip6po_pktinfo == NULL)
2762 				return (ENOBUFS);
2763 		}
2764 		memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo));
2765 		break;
2766 	}
2767 
2768 #ifdef RFC2292
2769 	case IPV6_2292HOPLIMIT:
2770 #endif
2771 	case IPV6_HOPLIMIT:
2772 	{
2773 		int *hlimp;
2774 
2775 		/*
2776 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2777 		 * to simplify the ordering among hoplimit options.
2778 		 */
2779 		if (optname == IPV6_HOPLIMIT && sticky)
2780 			return (ENOPROTOOPT);
2781 
2782 		if (len != sizeof(int))
2783 			return (EINVAL);
2784 		hlimp = (int *)buf;
2785 		if (*hlimp < -1 || *hlimp > 255)
2786 			return (EINVAL);
2787 
2788 		opt->ip6po_hlim = *hlimp;
2789 		break;
2790 	}
2791 
2792 	case IPV6_OTCLASS:
2793 		if (len != sizeof(u_int8_t))
2794 			return (EINVAL);
2795 
2796 		opt->ip6po_tclass = *(u_int8_t *)buf;
2797 		break;
2798 
2799 	case IPV6_TCLASS:
2800 	{
2801 		int tclass;
2802 
2803 		if (len != sizeof(int))
2804 			return (EINVAL);
2805 		tclass = *(int *)buf;
2806 		if (tclass < -1 || tclass > 255)
2807 			return (EINVAL);
2808 
2809 		opt->ip6po_tclass = tclass;
2810 		break;
2811 	}
2812 
2813 #ifdef RFC2292
2814 	case IPV6_2292NEXTHOP:
2815 #endif
2816 	case IPV6_NEXTHOP:
2817 		error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
2818 		    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2819 		if (error)
2820 			return (error);
2821 
2822 		if (len == 0) {	/* just remove the option */
2823 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2824 			break;
2825 		}
2826 
2827 		/* check if cmsg_len is large enough for sa_len */
2828 		if (len < sizeof(struct sockaddr) || len < *buf)
2829 			return (EINVAL);
2830 
2831 		switch (((struct sockaddr *)buf)->sa_family) {
2832 		case AF_INET6:
2833 		{
2834 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2835 
2836 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2837 				return (EINVAL);
2838 
2839 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2840 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2841 				return (EINVAL);
2842 			}
2843 			if ((error = sa6_embedscope(sa6, ip6_use_defzone))
2844 			    != 0) {
2845 				return (error);
2846 			}
2847 			break;
2848 		}
2849 		case AF_LINK:	/* eventually be supported? */
2850 		default:
2851 			return (EAFNOSUPPORT);
2852 		}
2853 
2854 		/* turn off the previous option, then set the new option. */
2855 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2856 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2857 		if (opt->ip6po_nexthop == NULL)
2858 			return (ENOBUFS);
2859 		memcpy(opt->ip6po_nexthop, buf, *buf);
2860 		break;
2861 
2862 #ifdef RFC2292
2863 	case IPV6_2292HOPOPTS:
2864 #endif
2865 	case IPV6_HOPOPTS:
2866 	{
2867 		struct ip6_hbh *hbh;
2868 		int hbhlen;
2869 
2870 		/*
2871 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2872 		 * options, since per-option restriction has too much
2873 		 * overhead.
2874 		 */
2875 		error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
2876 		    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2877 		if (error)
2878 			return (error);
2879 
2880 		if (len == 0) {
2881 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2882 			break;	/* just remove the option */
2883 		}
2884 
2885 		/* message length validation */
2886 		if (len < sizeof(struct ip6_hbh))
2887 			return (EINVAL);
2888 		hbh = (struct ip6_hbh *)buf;
2889 		hbhlen = (hbh->ip6h_len + 1) << 3;
2890 		if (len != hbhlen)
2891 			return (EINVAL);
2892 
2893 		/* turn off the previous option, then set the new option. */
2894 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2895 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2896 		if (opt->ip6po_hbh == NULL)
2897 			return (ENOBUFS);
2898 		memcpy(opt->ip6po_hbh, hbh, hbhlen);
2899 
2900 		break;
2901 	}
2902 
2903 #ifdef RFC2292
2904 	case IPV6_2292DSTOPTS:
2905 #endif
2906 	case IPV6_DSTOPTS:
2907 	case IPV6_RTHDRDSTOPTS:
2908 	{
2909 		struct ip6_dest *dest, **newdest = NULL;
2910 		int destlen;
2911 
2912 		/* XXX: see the comment for IPV6_HOPOPTS */
2913 		error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
2914 		    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2915 		if (error)
2916 			return (error);
2917 
2918 		if (len == 0) {
2919 			ip6_clearpktopts(opt, optname);
2920 			break;	/* just remove the option */
2921 		}
2922 
2923 		/* message length validation */
2924 		if (len < sizeof(struct ip6_dest))
2925 			return (EINVAL);
2926 		dest = (struct ip6_dest *)buf;
2927 		destlen = (dest->ip6d_len + 1) << 3;
2928 		if (len != destlen)
2929 			return (EINVAL);
2930 		/*
2931 		 * Determine the position that the destination options header
2932 		 * should be inserted; before or after the routing header.
2933 		 */
2934 		switch (optname) {
2935 		case IPV6_2292DSTOPTS:
2936 			/*
2937 			 * The old advanced API is ambiguous on this point.
2938 			 * Our approach is to determine the position based
2939 			 * according to the existence of a routing header.
2940 			 * Note, however, that this depends on the order of the
2941 			 * extension headers in the ancillary data; the 1st
2942 			 * part of the destination options header must appear
2943 			 * before the routing header in the ancillary data,
2944 			 * too.
2945 			 * RFC3542 solved the ambiguity by introducing
2946 			 * separate ancillary data or option types.
2947 			 */
2948 			if (opt->ip6po_rthdr == NULL)
2949 				newdest = &opt->ip6po_dest1;
2950 			else
2951 				newdest = &opt->ip6po_dest2;
2952 			break;
2953 		case IPV6_RTHDRDSTOPTS:
2954 			newdest = &opt->ip6po_dest1;
2955 			break;
2956 		case IPV6_DSTOPTS:
2957 			newdest = &opt->ip6po_dest2;
2958 			break;
2959 		}
2960 
2961 		/* turn off the previous option, then set the new option. */
2962 		ip6_clearpktopts(opt, optname);
2963 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2964 		if (*newdest == NULL)
2965 			return (ENOBUFS);
2966 		memcpy(*newdest, dest, destlen);
2967 
2968 		break;
2969 	}
2970 
2971 #ifdef RFC2292
2972 	case IPV6_2292RTHDR:
2973 #endif
2974 	case IPV6_RTHDR:
2975 	{
2976 		struct ip6_rthdr *rth;
2977 		int rthlen;
2978 
2979 		if (len == 0) {
2980 			ip6_clearpktopts(opt, IPV6_RTHDR);
2981 			break;	/* just remove the option */
2982 		}
2983 
2984 		/* message length validation */
2985 		if (len < sizeof(struct ip6_rthdr))
2986 			return (EINVAL);
2987 		rth = (struct ip6_rthdr *)buf;
2988 		rthlen = (rth->ip6r_len + 1) << 3;
2989 		if (len != rthlen)
2990 			return (EINVAL);
2991 		switch (rth->ip6r_type) {
2992 		case IPV6_RTHDR_TYPE_0:
2993 			if (rth->ip6r_len == 0)	/* must contain one addr */
2994 				return (EINVAL);
2995 			if (rth->ip6r_len % 2) /* length must be even */
2996 				return (EINVAL);
2997 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2998 				return (EINVAL);
2999 			break;
3000 		default:
3001 			return (EINVAL);	/* not supported */
3002 		}
3003 		/* turn off the previous option */
3004 		ip6_clearpktopts(opt, IPV6_RTHDR);
3005 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3006 		if (opt->ip6po_rthdr == NULL)
3007 			return (ENOBUFS);
3008 		memcpy(opt->ip6po_rthdr, rth, rthlen);
3009 		break;
3010 	}
3011 
3012 	case IPV6_USE_MIN_MTU:
3013 		if (len != sizeof(int))
3014 			return (EINVAL);
3015 		minmtupolicy = *(int *)buf;
3016 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3017 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3018 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3019 			return (EINVAL);
3020 		}
3021 		opt->ip6po_minmtu = minmtupolicy;
3022 		break;
3023 
3024 	case IPV6_DONTFRAG:
3025 		if (len != sizeof(int))
3026 			return (EINVAL);
3027 
3028 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3029 			/*
3030 			 * we ignore this option for TCP sockets.
3031 			 * (RFC3542 leaves this case unspecified.)
3032 			 */
3033 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3034 		} else
3035 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3036 		break;
3037 
3038 	default:
3039 		return (ENOPROTOOPT);
3040 	} /* end of switch */
3041 
3042 	return (0);
3043 }
3044 
3045 /*
3046  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3047  * packet to the input queue of a specified interface.  Note that this
3048  * calls the output routine of the loopback "driver", but with an interface
3049  * pointer that might NOT be lo0ifp -- easier than replicating that code here.
3050  */
3051 void
3052 ip6_mloopback(struct ifnet *ifp, struct mbuf *m,
3053 	const struct sockaddr_in6 *dst)
3054 {
3055 	struct mbuf *copym;
3056 	struct ip6_hdr *ip6;
3057 
3058 	copym = m_copy(m, 0, M_COPYALL);
3059 	if (copym == NULL)
3060 		return;
3061 
3062 	/*
3063 	 * Make sure to deep-copy IPv6 header portion in case the data
3064 	 * is in an mbuf cluster, so that we can safely override the IPv6
3065 	 * header portion later.
3066 	 */
3067 	if ((copym->m_flags & M_EXT) != 0 ||
3068 	    copym->m_len < sizeof(struct ip6_hdr)) {
3069 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3070 		if (copym == NULL)
3071 			return;
3072 	}
3073 
3074 #ifdef DIAGNOSTIC
3075 	if (copym->m_len < sizeof(*ip6)) {
3076 		m_freem(copym);
3077 		return;
3078 	}
3079 #endif
3080 
3081 	ip6 = mtod(copym, struct ip6_hdr *);
3082 	/*
3083 	 * clear embedded scope identifiers if necessary.
3084 	 * in6_clearscope will touch the addresses only when necessary.
3085 	 */
3086 	in6_clearscope(&ip6->ip6_src);
3087 	in6_clearscope(&ip6->ip6_dst);
3088 
3089 	(void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL);
3090 }
3091 
3092 /*
3093  * Chop IPv6 header off from the payload.
3094  */
3095 static int
3096 ip6_splithdr(struct mbuf *m,  struct ip6_exthdrs *exthdrs)
3097 {
3098 	struct mbuf *mh;
3099 	struct ip6_hdr *ip6;
3100 
3101 	ip6 = mtod(m, struct ip6_hdr *);
3102 	if (m->m_len > sizeof(*ip6)) {
3103 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3104 		if (mh == 0) {
3105 			m_freem(m);
3106 			return ENOBUFS;
3107 		}
3108 		M_MOVE_PKTHDR(mh, m);
3109 		MH_ALIGN(mh, sizeof(*ip6));
3110 		m->m_len -= sizeof(*ip6);
3111 		m->m_data += sizeof(*ip6);
3112 		mh->m_next = m;
3113 		m = mh;
3114 		m->m_len = sizeof(*ip6);
3115 		bcopy((void *)ip6, mtod(m, void *), sizeof(*ip6));
3116 	}
3117 	exthdrs->ip6e_ip6 = m;
3118 	return 0;
3119 }
3120 
3121 /*
3122  * Compute IPv6 extension header length.
3123  */
3124 int
3125 ip6_optlen(struct in6pcb *in6p)
3126 {
3127 	int len;
3128 
3129 	if (!in6p->in6p_outputopts)
3130 		return 0;
3131 
3132 	len = 0;
3133 #define elen(x) \
3134     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3135 
3136 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3137 	len += elen(in6p->in6p_outputopts->ip6po_dest1);
3138 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3139 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3140 	return len;
3141 #undef elen
3142 }
3143