xref: /netbsd-src/sys/netinet6/ip6_output.c (revision 5bbd2a12505d72a8177929a37b5cee489d0a1cfd)
1 /*	$NetBSD: ip6_output.c,v 1.150 2012/07/21 14:52:40 gdt 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 1.150 2012/07/21 14:52:40 gdt Exp $");
66 
67 #include "opt_inet.h"
68 #include "opt_inet6.h"
69 #include "opt_ipsec.h"
70 #include "opt_pfil_hooks.h"
71 
72 #include <sys/param.h>
73 #include <sys/malloc.h>
74 #include <sys/mbuf.h>
75 #include <sys/errno.h>
76 #include <sys/protosw.h>
77 #include <sys/socket.h>
78 #include <sys/socketvar.h>
79 #include <sys/systm.h>
80 #include <sys/proc.h>
81 #include <sys/kauth.h>
82 
83 #include <net/if.h>
84 #include <net/route.h>
85 #ifdef PFIL_HOOKS
86 #include <net/pfil.h>
87 #endif
88 
89 #include <netinet/in.h>
90 #include <netinet/in_var.h>
91 #include <netinet/ip6.h>
92 #include <netinet/icmp6.h>
93 #include <netinet/in_offload.h>
94 #include <netinet/portalgo.h>
95 #include <netinet6/in6_offload.h>
96 #include <netinet6/ip6_var.h>
97 #include <netinet6/ip6_private.h>
98 #include <netinet6/in6_pcb.h>
99 #include <netinet6/nd6.h>
100 #include <netinet6/ip6protosw.h>
101 #include <netinet6/scope6_var.h>
102 
103 #ifdef FAST_IPSEC
104 #include <netipsec/ipsec.h>
105 #include <netipsec/ipsec6.h>
106 #include <netipsec/key.h>
107 #include <netipsec/xform.h>
108 #endif
109 
110 
111 #include <net/net_osdep.h>
112 
113 #ifdef PFIL_HOOKS
114 extern struct pfil_head inet6_pfil_hook;	/* XXX */
115 #endif
116 
117 struct ip6_exthdrs {
118 	struct mbuf *ip6e_ip6;
119 	struct mbuf *ip6e_hbh;
120 	struct mbuf *ip6e_dest1;
121 	struct mbuf *ip6e_rthdr;
122 	struct mbuf *ip6e_dest2;
123 };
124 
125 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
126 	kauth_cred_t, int);
127 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
128 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, kauth_cred_t,
129 	int, int, int);
130 static int ip6_setmoptions(const struct sockopt *, struct ip6_moptions **);
131 static int ip6_getmoptions(struct sockopt *, struct ip6_moptions *);
132 static int ip6_copyexthdr(struct mbuf **, void *, int);
133 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
134 	struct ip6_frag **);
135 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
136 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
137 static int ip6_getpmtu(struct route *, struct route *, struct ifnet *,
138     const struct in6_addr *, u_long *, int *);
139 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
140 
141 #ifdef RFC2292
142 static int ip6_pcbopts(struct ip6_pktopts **, struct socket *, struct sockopt *);
143 #endif
144 
145 /*
146  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
147  * header (with pri, len, nxt, hlim, src, dst).
148  * This function may modify ver and hlim only.
149  * The mbuf chain containing the packet will be freed.
150  * The mbuf opt, if present, will not be freed.
151  *
152  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
153  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
154  * which is rt_rmx.rmx_mtu.
155  */
156 int
157 ip6_output(
158     struct mbuf *m0,
159     struct ip6_pktopts *opt,
160     struct route *ro,
161     int flags,
162     struct ip6_moptions *im6o,
163     struct socket *so,
164     struct ifnet **ifpp		/* XXX: just for statistics */
165 )
166 {
167 	struct ip6_hdr *ip6, *mhip6;
168 	struct ifnet *ifp, *origifp;
169 	struct mbuf *m = m0;
170 	int hlen, tlen, len, off;
171 	bool tso;
172 	struct route ip6route;
173 	struct rtentry *rt = NULL;
174 	const struct sockaddr_in6 *dst = NULL;
175 	struct sockaddr_in6 src_sa, dst_sa;
176 	int error = 0;
177 	struct in6_ifaddr *ia = NULL;
178 	u_long mtu;
179 	int alwaysfrag, dontfrag;
180 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
181 	struct ip6_exthdrs exthdrs;
182 	struct in6_addr finaldst, src0, dst0;
183 	u_int32_t zone;
184 	struct route *ro_pmtu = NULL;
185 	int hdrsplit = 0;
186 	int needipsec = 0;
187 #ifdef FAST_IPSEC
188 	struct secpolicy *sp = NULL;
189 	int s;
190 #endif
191 
192 	memset(&ip6route, 0, sizeof(ip6route));
193 
194 #ifdef  DIAGNOSTIC
195 	if ((m->m_flags & M_PKTHDR) == 0)
196 		panic("ip6_output: no HDR");
197 
198 	if ((m->m_pkthdr.csum_flags &
199 	    (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) {
200 		panic("ip6_output: IPv4 checksum offload flags: %d",
201 		    m->m_pkthdr.csum_flags);
202 	}
203 
204 	if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) ==
205 	    (M_CSUM_TCPv6|M_CSUM_UDPv6)) {
206 		panic("ip6_output: conflicting checksum offload flags: %d",
207 		    m->m_pkthdr.csum_flags);
208 	}
209 #endif
210 
211 	M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr));
212 
213 #define MAKE_EXTHDR(hp, mp)						\
214     do {								\
215 	if (hp) {							\
216 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
217 		error = ip6_copyexthdr((mp), (void *)(hp), 		\
218 		    ((eh)->ip6e_len + 1) << 3);				\
219 		if (error)						\
220 			goto freehdrs;					\
221 	}								\
222     } while (/*CONSTCOND*/ 0)
223 
224 	memset(&exthdrs, 0, sizeof(exthdrs));
225 	if (opt) {
226 		/* Hop-by-Hop options header */
227 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
228 		/* Destination options header(1st part) */
229 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
230 		/* Routing header */
231 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
232 		/* Destination options header(2nd part) */
233 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
234 	}
235 
236 	/*
237 	 * Calculate the total length of the extension header chain.
238 	 * Keep the length of the unfragmentable part for fragmentation.
239 	 */
240 	optlen = 0;
241 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
242 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
243 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
244 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
245 	/* NOTE: we don't add AH/ESP length here. do that later. */
246 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
247 
248 #ifdef FAST_IPSEC
249 	/* Check the security policy (SP) for the packet */
250 
251 	sp = ipsec6_check_policy(m,so,flags,&needipsec,&error);
252 	if (error != 0) {
253 		/*
254 		 * Hack: -EINVAL is used to signal that a packet
255 		 * should be silently discarded.  This is typically
256 		 * because we asked key management for an SA and
257 		 * it was delayed (e.g. kicked up to IKE).
258 		 */
259 	if (error == -EINVAL)
260 		error = 0;
261 	goto freehdrs;
262     }
263 #endif /* FAST_IPSEC */
264 
265 
266 	if (needipsec &&
267 	    (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
268 		in6_delayed_cksum(m);
269 		m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
270 	}
271 
272 
273 	/*
274 	 * If we need IPsec, or there is at least one extension header,
275 	 * separate IP6 header from the payload.
276 	 */
277 	if ((needipsec || optlen) && !hdrsplit) {
278 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
279 			m = NULL;
280 			goto freehdrs;
281 		}
282 		m = exthdrs.ip6e_ip6;
283 		hdrsplit++;
284 	}
285 
286 	/* adjust pointer */
287 	ip6 = mtod(m, struct ip6_hdr *);
288 
289 	/* adjust mbuf packet header length */
290 	m->m_pkthdr.len += optlen;
291 	plen = m->m_pkthdr.len - sizeof(*ip6);
292 
293 	/* If this is a jumbo payload, insert a jumbo payload option. */
294 	if (plen > IPV6_MAXPACKET) {
295 		if (!hdrsplit) {
296 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
297 				m = NULL;
298 				goto freehdrs;
299 			}
300 			m = exthdrs.ip6e_ip6;
301 			hdrsplit++;
302 		}
303 		/* adjust pointer */
304 		ip6 = mtod(m, struct ip6_hdr *);
305 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
306 			goto freehdrs;
307 		optlen += 8; /* XXX JUMBOOPTLEN */
308 		ip6->ip6_plen = 0;
309 	} else
310 		ip6->ip6_plen = htons(plen);
311 
312 	/*
313 	 * Concatenate headers and fill in next header fields.
314 	 * Here we have, on "m"
315 	 *	IPv6 payload
316 	 * and we insert headers accordingly.  Finally, we should be getting:
317 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
318 	 *
319 	 * during the header composing process, "m" points to IPv6 header.
320 	 * "mprev" points to an extension header prior to esp.
321 	 */
322 	{
323 		u_char *nexthdrp = &ip6->ip6_nxt;
324 		struct mbuf *mprev = m;
325 
326 		/*
327 		 * we treat dest2 specially.  this makes IPsec processing
328 		 * much easier.  the goal here is to make mprev point the
329 		 * mbuf prior to dest2.
330 		 *
331 		 * result: IPv6 dest2 payload
332 		 * m and mprev will point to IPv6 header.
333 		 */
334 		if (exthdrs.ip6e_dest2) {
335 			if (!hdrsplit)
336 				panic("assumption failed: hdr not split");
337 			exthdrs.ip6e_dest2->m_next = m->m_next;
338 			m->m_next = exthdrs.ip6e_dest2;
339 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
340 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
341 		}
342 
343 #define MAKE_CHAIN(m, mp, p, i)\
344     do {\
345 	if (m) {\
346 		if (!hdrsplit) \
347 			panic("assumption failed: hdr not split"); \
348 		*mtod((m), u_char *) = *(p);\
349 		*(p) = (i);\
350 		p = mtod((m), u_char *);\
351 		(m)->m_next = (mp)->m_next;\
352 		(mp)->m_next = (m);\
353 		(mp) = (m);\
354 	}\
355     } while (/*CONSTCOND*/ 0)
356 		/*
357 		 * result: IPv6 hbh dest1 rthdr dest2 payload
358 		 * m will point to IPv6 header.  mprev will point to the
359 		 * extension header prior to dest2 (rthdr in the above case).
360 		 */
361 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
362 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
363 		    IPPROTO_DSTOPTS);
364 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
365 		    IPPROTO_ROUTING);
366 
367 		M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data,
368 		    sizeof(struct ip6_hdr) + optlen);
369 	}
370 
371 	/*
372 	 * If there is a routing header, replace destination address field
373 	 * with the first hop of the routing header.
374 	 */
375 	if (exthdrs.ip6e_rthdr) {
376 		struct ip6_rthdr *rh;
377 		struct ip6_rthdr0 *rh0;
378 		struct in6_addr *addr;
379 		struct sockaddr_in6 sa;
380 
381 		rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
382 		    struct ip6_rthdr *));
383 		finaldst = ip6->ip6_dst;
384 		switch (rh->ip6r_type) {
385 		case IPV6_RTHDR_TYPE_0:
386 			 rh0 = (struct ip6_rthdr0 *)rh;
387 			 addr = (struct in6_addr *)(rh0 + 1);
388 
389 			 /*
390 			  * construct a sockaddr_in6 form of
391 			  * the first hop.
392 			  *
393 			  * XXX: we may not have enough
394 			  * information about its scope zone;
395 			  * there is no standard API to pass
396 			  * the information from the
397 			  * application.
398 			  */
399 			 sockaddr_in6_init(&sa, addr, 0, 0, 0);
400 			 if ((error = sa6_embedscope(&sa,
401 			     ip6_use_defzone)) != 0) {
402 				 goto bad;
403 			 }
404 			 ip6->ip6_dst = sa.sin6_addr;
405 			 (void)memmove(&addr[0], &addr[1],
406 			     sizeof(struct in6_addr) *
407 			     (rh0->ip6r0_segleft - 1));
408 			 addr[rh0->ip6r0_segleft - 1] = finaldst;
409 			 /* XXX */
410 			 in6_clearscope(addr + rh0->ip6r0_segleft - 1);
411 			 break;
412 		default:	/* is it possible? */
413 			 error = EINVAL;
414 			 goto bad;
415 		}
416 	}
417 
418 	/* Source address validation */
419 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
420 	    (flags & IPV6_UNSPECSRC) == 0) {
421 		error = EOPNOTSUPP;
422 		IP6_STATINC(IP6_STAT_BADSCOPE);
423 		goto bad;
424 	}
425 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
426 		error = EOPNOTSUPP;
427 		IP6_STATINC(IP6_STAT_BADSCOPE);
428 		goto bad;
429 	}
430 
431 	IP6_STATINC(IP6_STAT_LOCALOUT);
432 
433 	/*
434 	 * Route packet.
435 	 */
436 	/* initialize cached route */
437 	if (ro == NULL) {
438 		ro = &ip6route;
439 	}
440 	ro_pmtu = ro;
441 	if (opt && opt->ip6po_rthdr)
442 		ro = &opt->ip6po_route;
443 
444  	/*
445 	 * if specified, try to fill in the traffic class field.
446 	 * do not override if a non-zero value is already set.
447 	 * we check the diffserv field and the ecn field separately.
448 	 */
449 	if (opt && opt->ip6po_tclass >= 0) {
450 		int mask = 0;
451 
452 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
453 			mask |= 0xfc;
454 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
455 			mask |= 0x03;
456 		if (mask != 0)
457 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
458 	}
459 
460 	/* fill in or override the hop limit field, if necessary. */
461 	if (opt && opt->ip6po_hlim != -1)
462 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
463 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
464 		if (im6o != NULL)
465 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
466 		else
467 			ip6->ip6_hlim = ip6_defmcasthlim;
468 	}
469 
470 #ifdef FAST_IPSEC
471 	if (needipsec) {
472 		s = splsoftnet();
473 		error = ipsec6_process_packet(m,sp->req);
474 
475 		/*
476 		 * Preserve KAME behaviour: ENOENT can be returned
477 		 * when an SA acquire is in progress.  Don't propagate
478 		 * this to user-level; it confuses applications.
479 		 * XXX this will go away when the SADB is redone.
480 		 */
481 		if (error == ENOENT)
482 			error = 0;
483 		splx(s);
484 		goto done;
485 	}
486 #endif /* FAST_IPSEC */
487 
488 
489 
490 	/* adjust pointer */
491 	ip6 = mtod(m, struct ip6_hdr *);
492 
493 	sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
494 	if ((error = in6_selectroute(&dst_sa, opt, im6o, ro,
495 	    &ifp, &rt, 0)) != 0) {
496 		if (ifp != NULL)
497 			in6_ifstat_inc(ifp, ifs6_out_discard);
498 		goto bad;
499 	}
500 	if (rt == NULL) {
501 		/*
502 		 * If in6_selectroute() does not return a route entry,
503 		 * dst may not have been updated.
504 		 */
505 		error = rtcache_setdst(ro, sin6tosa(&dst_sa));
506 		if (error) {
507 			goto bad;
508 		}
509 	}
510 
511 	/*
512 	 * then rt (for unicast) and ifp must be non-NULL valid values.
513 	 */
514 	if ((flags & IPV6_FORWARDING) == 0) {
515 		/* XXX: the FORWARDING flag can be set for mrouting. */
516 		in6_ifstat_inc(ifp, ifs6_out_request);
517 	}
518 	if (rt != NULL) {
519 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
520 		rt->rt_use++;
521 	}
522 
523 	/*
524 	 * The outgoing interface must be in the zone of source and
525 	 * destination addresses.  We should use ia_ifp to support the
526 	 * case of sending packets to an address of our own.
527 	 */
528 	if (ia != NULL && ia->ia_ifp)
529 		origifp = ia->ia_ifp;
530 	else
531 		origifp = ifp;
532 
533 	src0 = ip6->ip6_src;
534 	if (in6_setscope(&src0, origifp, &zone))
535 		goto badscope;
536 	sockaddr_in6_init(&src_sa, &ip6->ip6_src, 0, 0, 0);
537 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
538 		goto badscope;
539 
540 	dst0 = ip6->ip6_dst;
541 	if (in6_setscope(&dst0, origifp, &zone))
542 		goto badscope;
543 	/* re-initialize to be sure */
544 	sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
545 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id)
546 		goto badscope;
547 
548 	/* scope check is done. */
549 
550 	if (rt == NULL || IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
551 		if (dst == NULL)
552 			dst = satocsin6(rtcache_getdst(ro));
553 		KASSERT(dst != NULL);
554 	} else if (opt && rtcache_validate(&opt->ip6po_nextroute) != NULL) {
555 		/*
556 		 * The nexthop is explicitly specified by the
557 		 * application.  We assume the next hop is an IPv6
558 		 * address.
559 		 */
560 		dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
561 	} else if ((rt->rt_flags & RTF_GATEWAY))
562 		dst = (struct sockaddr_in6 *)rt->rt_gateway;
563 	else if (dst == NULL)
564 		dst = satocsin6(rtcache_getdst(ro));
565 
566 	/*
567 	 * XXXXXX: original code follows:
568 	 */
569 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
570 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
571 	else {
572 		struct	in6_multi *in6m;
573 
574 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
575 
576 		in6_ifstat_inc(ifp, ifs6_out_mcast);
577 
578 		/*
579 		 * Confirm that the outgoing interface supports multicast.
580 		 */
581 		if (!(ifp->if_flags & IFF_MULTICAST)) {
582 			IP6_STATINC(IP6_STAT_NOROUTE);
583 			in6_ifstat_inc(ifp, ifs6_out_discard);
584 			error = ENETUNREACH;
585 			goto bad;
586 		}
587 
588 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
589 		if (in6m != NULL &&
590 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
591 			/*
592 			 * If we belong to the destination multicast group
593 			 * on the outgoing interface, and the caller did not
594 			 * forbid loopback, loop back a copy.
595 			 */
596 			KASSERT(dst != NULL);
597 			ip6_mloopback(ifp, m, dst);
598 		} else {
599 			/*
600 			 * If we are acting as a multicast router, perform
601 			 * multicast forwarding as if the packet had just
602 			 * arrived on the interface to which we are about
603 			 * to send.  The multicast forwarding function
604 			 * recursively calls this function, using the
605 			 * IPV6_FORWARDING flag to prevent infinite recursion.
606 			 *
607 			 * Multicasts that are looped back by ip6_mloopback(),
608 			 * above, will be forwarded by the ip6_input() routine,
609 			 * if necessary.
610 			 */
611 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
612 				if (ip6_mforward(ip6, ifp, m) != 0) {
613 					m_freem(m);
614 					goto done;
615 				}
616 			}
617 		}
618 		/*
619 		 * Multicasts with a hoplimit of zero may be looped back,
620 		 * above, but must not be transmitted on a network.
621 		 * Also, multicasts addressed to the loopback interface
622 		 * are not sent -- the above call to ip6_mloopback() will
623 		 * loop back a copy if this host actually belongs to the
624 		 * destination group on the loopback interface.
625 		 */
626 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
627 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
628 			m_freem(m);
629 			goto done;
630 		}
631 	}
632 
633 	/*
634 	 * Fill the outgoing inteface to tell the upper layer
635 	 * to increment per-interface statistics.
636 	 */
637 	if (ifpp)
638 		*ifpp = ifp;
639 
640 	/* Determine path MTU. */
641 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
642 	    &alwaysfrag)) != 0)
643 		goto bad;
644 
645 	/*
646 	 * The caller of this function may specify to use the minimum MTU
647 	 * in some cases.
648 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
649 	 * setting.  The logic is a bit complicated; by default, unicast
650 	 * packets will follow path MTU while multicast packets will be sent at
651 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
652 	 * including unicast ones will be sent at the minimum MTU.  Multicast
653 	 * packets will always be sent at the minimum MTU unless
654 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
655 	 * See RFC 3542 for more details.
656 	 */
657 	if (mtu > IPV6_MMTU) {
658 		if ((flags & IPV6_MINMTU))
659 			mtu = IPV6_MMTU;
660 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
661 			mtu = IPV6_MMTU;
662 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
663 			 (opt == NULL ||
664 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
665 			mtu = IPV6_MMTU;
666 		}
667 	}
668 
669 	/*
670 	 * clear embedded scope identifiers if necessary.
671 	 * in6_clearscope will touch the addresses only when necessary.
672 	 */
673 	in6_clearscope(&ip6->ip6_src);
674 	in6_clearscope(&ip6->ip6_dst);
675 
676 	/*
677 	 * If the outgoing packet contains a hop-by-hop options header,
678 	 * it must be examined and processed even by the source node.
679 	 * (RFC 2460, section 4.)
680 	 */
681 	if (ip6->ip6_nxt == IPV6_HOPOPTS) {
682 		u_int32_t dummy1; /* XXX unused */
683 		u_int32_t dummy2; /* XXX unused */
684 		int hoff = sizeof(struct ip6_hdr);
685 
686 		if (ip6_hopopts_input(&dummy1, &dummy2, &m, &hoff)) {
687 			/* m was already freed at this point */
688 			error = EINVAL;/* better error? */
689 			goto done;
690 		}
691 
692 		ip6 = mtod(m, struct ip6_hdr *);
693 	}
694 
695 #ifdef PFIL_HOOKS
696 	/*
697 	 * Run through list of hooks for output packets.
698 	 */
699 	if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0)
700 		goto done;
701 	if (m == NULL)
702 		goto done;
703 	ip6 = mtod(m, struct ip6_hdr *);
704 #endif /* PFIL_HOOKS */
705 	/*
706 	 * Send the packet to the outgoing interface.
707 	 * If necessary, do IPv6 fragmentation before sending.
708 	 *
709 	 * the logic here is rather complex:
710 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
711 	 * 1-a:	send as is if tlen <= path mtu
712 	 * 1-b:	fragment if tlen > path mtu
713 	 *
714 	 * 2: if user asks us not to fragment (dontfrag == 1)
715 	 * 2-a:	send as is if tlen <= interface mtu
716 	 * 2-b:	error if tlen > interface mtu
717 	 *
718 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
719 	 *	always fragment
720 	 *
721 	 * 4: if dontfrag == 1 && alwaysfrag == 1
722 	 *	error, as we cannot handle this conflicting request
723 	 */
724 	tlen = m->m_pkthdr.len;
725 	tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0;
726 	if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
727 		dontfrag = 1;
728 	else
729 		dontfrag = 0;
730 
731 	if (dontfrag && alwaysfrag) {	/* case 4 */
732 		/* conflicting request - can't transmit */
733 		error = EMSGSIZE;
734 		goto bad;
735 	}
736 	if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) {	/* case 2-b */
737 		/*
738 		 * Even if the DONTFRAG option is specified, we cannot send the
739 		 * packet when the data length is larger than the MTU of the
740 		 * outgoing interface.
741 		 * Notify the error by sending IPV6_PATHMTU ancillary data as
742 		 * well as returning an error code (the latter is not described
743 		 * in the API spec.)
744 		 */
745 		u_int32_t mtu32;
746 		struct ip6ctlparam ip6cp;
747 
748 		mtu32 = (u_int32_t)mtu;
749 		memset(&ip6cp, 0, sizeof(ip6cp));
750 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
751 		pfctlinput2(PRC_MSGSIZE,
752 		    rtcache_getdst(ro_pmtu), &ip6cp);
753 
754 		error = EMSGSIZE;
755 		goto bad;
756 	}
757 
758 	/*
759 	 * transmit packet without fragmentation
760 	 */
761 	if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) {
762 		/* case 1-a and 2-a */
763 		struct in6_ifaddr *ia6;
764 		int sw_csum;
765 
766 		ip6 = mtod(m, struct ip6_hdr *);
767 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
768 		if (ia6) {
769 			/* Record statistics for this interface address. */
770 			ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len;
771 		}
772 
773 		sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
774 		if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
775 			if (IN6_NEED_CHECKSUM(ifp,
776 			    sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) {
777 				in6_delayed_cksum(m);
778 			}
779 			m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
780 		}
781 
782 		KASSERT(dst != NULL);
783 		if (__predict_true(!tso ||
784 		    (ifp->if_capenable & IFCAP_TSOv6) != 0)) {
785 			error = nd6_output(ifp, origifp, m, dst, rt);
786 		} else {
787 			error = ip6_tso_output(ifp, origifp, m, dst, rt);
788 		}
789 		goto done;
790 	}
791 
792 	if (tso) {
793 		error = EINVAL; /* XXX */
794 		goto bad;
795 	}
796 
797 	/*
798 	 * try to fragment the packet.  case 1-b and 3
799 	 */
800 	if (mtu < IPV6_MMTU) {
801 		/* path MTU cannot be less than IPV6_MMTU */
802 		error = EMSGSIZE;
803 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
804 		goto bad;
805 	} else if (ip6->ip6_plen == 0) {
806 		/* jumbo payload cannot be fragmented */
807 		error = EMSGSIZE;
808 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
809 		goto bad;
810 	} else {
811 		struct mbuf **mnext, *m_frgpart;
812 		struct ip6_frag *ip6f;
813 		u_int32_t id = htonl(ip6_randomid());
814 		u_char nextproto;
815 #if 0				/* see below */
816 		struct ip6ctlparam ip6cp;
817 		u_int32_t mtu32;
818 #endif
819 
820 		/*
821 		 * Too large for the destination or interface;
822 		 * fragment if possible.
823 		 * Must be able to put at least 8 bytes per fragment.
824 		 */
825 		hlen = unfragpartlen;
826 		if (mtu > IPV6_MAXPACKET)
827 			mtu = IPV6_MAXPACKET;
828 
829 #if 0
830 		/*
831 		 * It is believed this code is a leftover from the
832 		 * development of the IPV6_RECVPATHMTU sockopt and
833 		 * associated work to implement RFC3542.
834 		 * It's not entirely clear what the intent of the API
835 		 * is at this point, so disable this code for now.
836 		 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG
837 		 * will send notifications if the application requests.
838 		 */
839 
840 		/* Notify a proper path MTU to applications. */
841 		mtu32 = (u_int32_t)mtu;
842 		memset(&ip6cp, 0, sizeof(ip6cp));
843 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
844 		pfctlinput2(PRC_MSGSIZE,
845 		    rtcache_getdst(ro_pmtu), &ip6cp);
846 #endif
847 
848 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
849 		if (len < 8) {
850 			error = EMSGSIZE;
851 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
852 			goto bad;
853 		}
854 
855 		mnext = &m->m_nextpkt;
856 
857 		/*
858 		 * Change the next header field of the last header in the
859 		 * unfragmentable part.
860 		 */
861 		if (exthdrs.ip6e_rthdr) {
862 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
863 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
864 		} else if (exthdrs.ip6e_dest1) {
865 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
866 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
867 		} else if (exthdrs.ip6e_hbh) {
868 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
869 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
870 		} else {
871 			nextproto = ip6->ip6_nxt;
872 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
873 		}
874 
875 		if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6))
876 		    != 0) {
877 			if (IN6_NEED_CHECKSUM(ifp,
878 			    m->m_pkthdr.csum_flags &
879 			    (M_CSUM_UDPv6|M_CSUM_TCPv6))) {
880 				in6_delayed_cksum(m);
881 			}
882 			m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
883 		}
884 
885 		/*
886 		 * Loop through length of segment after first fragment,
887 		 * make new header and copy data of each part and link onto
888 		 * chain.
889 		 */
890 		m0 = m;
891 		for (off = hlen; off < tlen; off += len) {
892 			struct mbuf *mlast;
893 
894 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
895 			if (!m) {
896 				error = ENOBUFS;
897 				IP6_STATINC(IP6_STAT_ODROPPED);
898 				goto sendorfree;
899 			}
900 			m->m_pkthdr.rcvif = NULL;
901 			m->m_flags = m0->m_flags & M_COPYFLAGS;
902 			*mnext = m;
903 			mnext = &m->m_nextpkt;
904 			m->m_data += max_linkhdr;
905 			mhip6 = mtod(m, struct ip6_hdr *);
906 			*mhip6 = *ip6;
907 			m->m_len = sizeof(*mhip6);
908 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
909 			if (error) {
910 				IP6_STATINC(IP6_STAT_ODROPPED);
911 				goto sendorfree;
912 			}
913 			ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7));
914 			if (off + len >= tlen)
915 				len = tlen - off;
916 			else
917 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
918 			mhip6->ip6_plen = htons((u_int16_t)(len + hlen +
919 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
920 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
921 				error = ENOBUFS;
922 				IP6_STATINC(IP6_STAT_ODROPPED);
923 				goto sendorfree;
924 			}
925 			for (mlast = m; mlast->m_next; mlast = mlast->m_next)
926 				;
927 			mlast->m_next = m_frgpart;
928 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
929 			m->m_pkthdr.rcvif = NULL;
930 			ip6f->ip6f_reserved = 0;
931 			ip6f->ip6f_ident = id;
932 			ip6f->ip6f_nxt = nextproto;
933 			IP6_STATINC(IP6_STAT_OFRAGMENTS);
934 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
935 		}
936 
937 		in6_ifstat_inc(ifp, ifs6_out_fragok);
938 	}
939 
940 	/*
941 	 * Remove leading garbages.
942 	 */
943 sendorfree:
944 	m = m0->m_nextpkt;
945 	m0->m_nextpkt = 0;
946 	m_freem(m0);
947 	for (m0 = m; m; m = m0) {
948 		m0 = m->m_nextpkt;
949 		m->m_nextpkt = 0;
950 		if (error == 0) {
951 			struct in6_ifaddr *ia6;
952 			ip6 = mtod(m, struct ip6_hdr *);
953 			ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
954 			if (ia6) {
955 				/*
956 				 * Record statistics for this interface
957 				 * address.
958 				 */
959 				ia6->ia_ifa.ifa_data.ifad_outbytes +=
960 				    m->m_pkthdr.len;
961 			}
962 			KASSERT(dst != NULL);
963 			error = nd6_output(ifp, origifp, m, dst, rt);
964 		} else
965 			m_freem(m);
966 	}
967 
968 	if (error == 0)
969 		IP6_STATINC(IP6_STAT_FRAGMENTED);
970 
971 done:
972 	rtcache_free(&ip6route);
973 
974 #ifdef FAST_IPSEC
975 	if (sp != NULL)
976 		KEY_FREESP(&sp);
977 #endif /* FAST_IPSEC */
978 
979 
980 	return (error);
981 
982 freehdrs:
983 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
984 	m_freem(exthdrs.ip6e_dest1);
985 	m_freem(exthdrs.ip6e_rthdr);
986 	m_freem(exthdrs.ip6e_dest2);
987 	/* FALLTHROUGH */
988 bad:
989 	m_freem(m);
990 	goto done;
991 badscope:
992 	IP6_STATINC(IP6_STAT_BADSCOPE);
993 	in6_ifstat_inc(origifp, ifs6_out_discard);
994 	if (error == 0)
995 		error = EHOSTUNREACH; /* XXX */
996 	goto bad;
997 }
998 
999 static int
1000 ip6_copyexthdr(struct mbuf **mp, void *hdr, int hlen)
1001 {
1002 	struct mbuf *m;
1003 
1004 	if (hlen > MCLBYTES)
1005 		return (ENOBUFS); /* XXX */
1006 
1007 	MGET(m, M_DONTWAIT, MT_DATA);
1008 	if (!m)
1009 		return (ENOBUFS);
1010 
1011 	if (hlen > MLEN) {
1012 		MCLGET(m, M_DONTWAIT);
1013 		if ((m->m_flags & M_EXT) == 0) {
1014 			m_free(m);
1015 			return (ENOBUFS);
1016 		}
1017 	}
1018 	m->m_len = hlen;
1019 	if (hdr)
1020 		bcopy(hdr, mtod(m, void *), hlen);
1021 
1022 	*mp = m;
1023 	return (0);
1024 }
1025 
1026 /*
1027  * Process a delayed payload checksum calculation.
1028  */
1029 void
1030 in6_delayed_cksum(struct mbuf *m)
1031 {
1032 	uint16_t csum, offset;
1033 
1034 	KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0);
1035 	KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0);
1036 	KASSERT((m->m_pkthdr.csum_flags
1037 	    & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0);
1038 
1039 	offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data);
1040 	csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset);
1041 	if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) {
1042 		csum = 0xffff;
1043 	}
1044 
1045 	offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data);
1046 	if ((offset + sizeof(csum)) > m->m_len) {
1047 		m_copyback(m, offset, sizeof(csum), &csum);
1048 	} else {
1049 		*(uint16_t *)(mtod(m, char *) + offset) = csum;
1050 	}
1051 }
1052 
1053 /*
1054  * Insert jumbo payload option.
1055  */
1056 static int
1057 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1058 {
1059 	struct mbuf *mopt;
1060 	u_int8_t *optbuf;
1061 	u_int32_t v;
1062 
1063 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1064 
1065 	/*
1066 	 * If there is no hop-by-hop options header, allocate new one.
1067 	 * If there is one but it doesn't have enough space to store the
1068 	 * jumbo payload option, allocate a cluster to store the whole options.
1069 	 * Otherwise, use it to store the options.
1070 	 */
1071 	if (exthdrs->ip6e_hbh == 0) {
1072 		MGET(mopt, M_DONTWAIT, MT_DATA);
1073 		if (mopt == 0)
1074 			return (ENOBUFS);
1075 		mopt->m_len = JUMBOOPTLEN;
1076 		optbuf = mtod(mopt, u_int8_t *);
1077 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1078 		exthdrs->ip6e_hbh = mopt;
1079 	} else {
1080 		struct ip6_hbh *hbh;
1081 
1082 		mopt = exthdrs->ip6e_hbh;
1083 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1084 			/*
1085 			 * XXX assumption:
1086 			 * - exthdrs->ip6e_hbh is not referenced from places
1087 			 *   other than exthdrs.
1088 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1089 			 */
1090 			int oldoptlen = mopt->m_len;
1091 			struct mbuf *n;
1092 
1093 			/*
1094 			 * XXX: give up if the whole (new) hbh header does
1095 			 * not fit even in an mbuf cluster.
1096 			 */
1097 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1098 				return (ENOBUFS);
1099 
1100 			/*
1101 			 * As a consequence, we must always prepare a cluster
1102 			 * at this point.
1103 			 */
1104 			MGET(n, M_DONTWAIT, MT_DATA);
1105 			if (n) {
1106 				MCLGET(n, M_DONTWAIT);
1107 				if ((n->m_flags & M_EXT) == 0) {
1108 					m_freem(n);
1109 					n = NULL;
1110 				}
1111 			}
1112 			if (!n)
1113 				return (ENOBUFS);
1114 			n->m_len = oldoptlen + JUMBOOPTLEN;
1115 			bcopy(mtod(mopt, void *), mtod(n, void *),
1116 			    oldoptlen);
1117 			optbuf = mtod(n, u_int8_t *) + oldoptlen;
1118 			m_freem(mopt);
1119 			mopt = exthdrs->ip6e_hbh = n;
1120 		} else {
1121 			optbuf = mtod(mopt, u_int8_t *) + mopt->m_len;
1122 			mopt->m_len += JUMBOOPTLEN;
1123 		}
1124 		optbuf[0] = IP6OPT_PADN;
1125 		optbuf[1] = 0;
1126 
1127 		/*
1128 		 * Adjust the header length according to the pad and
1129 		 * the jumbo payload option.
1130 		 */
1131 		hbh = mtod(mopt, struct ip6_hbh *);
1132 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1133 	}
1134 
1135 	/* fill in the option. */
1136 	optbuf[2] = IP6OPT_JUMBO;
1137 	optbuf[3] = 4;
1138 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1139 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1140 
1141 	/* finally, adjust the packet header length */
1142 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1143 
1144 	return (0);
1145 #undef JUMBOOPTLEN
1146 }
1147 
1148 /*
1149  * Insert fragment header and copy unfragmentable header portions.
1150  *
1151  * *frghdrp will not be read, and it is guaranteed that either an
1152  * error is returned or that *frghdrp will point to space allocated
1153  * for the fragment header.
1154  */
1155 static int
1156 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1157 	struct ip6_frag **frghdrp)
1158 {
1159 	struct mbuf *n, *mlast;
1160 
1161 	if (hlen > sizeof(struct ip6_hdr)) {
1162 		n = m_copym(m0, sizeof(struct ip6_hdr),
1163 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1164 		if (n == 0)
1165 			return (ENOBUFS);
1166 		m->m_next = n;
1167 	} else
1168 		n = m;
1169 
1170 	/* Search for the last mbuf of unfragmentable part. */
1171 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1172 		;
1173 
1174 	if ((mlast->m_flags & M_EXT) == 0 &&
1175 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1176 		/* use the trailing space of the last mbuf for the fragment hdr */
1177 		*frghdrp = (struct ip6_frag *)(mtod(mlast, char *) +
1178 		    mlast->m_len);
1179 		mlast->m_len += sizeof(struct ip6_frag);
1180 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1181 	} else {
1182 		/* allocate a new mbuf for the fragment header */
1183 		struct mbuf *mfrg;
1184 
1185 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1186 		if (mfrg == 0)
1187 			return (ENOBUFS);
1188 		mfrg->m_len = sizeof(struct ip6_frag);
1189 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1190 		mlast->m_next = mfrg;
1191 	}
1192 
1193 	return (0);
1194 }
1195 
1196 static int
1197 ip6_getpmtu(struct route *ro_pmtu, struct route *ro, struct ifnet *ifp,
1198     const struct in6_addr *dst, u_long *mtup, int *alwaysfragp)
1199 {
1200 	struct rtentry *rt;
1201 	u_int32_t mtu = 0;
1202 	int alwaysfrag = 0;
1203 	int error = 0;
1204 
1205 	if (ro_pmtu != ro) {
1206 		union {
1207 			struct sockaddr		dst;
1208 			struct sockaddr_in6	dst6;
1209 		} u;
1210 
1211 		/* The first hop and the final destination may differ. */
1212 		sockaddr_in6_init(&u.dst6, dst, 0, 0, 0);
1213 		rt = rtcache_lookup(ro_pmtu, &u.dst);
1214 	} else
1215 		rt = rtcache_validate(ro_pmtu);
1216 	if (rt != NULL) {
1217 		u_int32_t ifmtu;
1218 
1219 		if (ifp == NULL)
1220 			ifp = rt->rt_ifp;
1221 		ifmtu = IN6_LINKMTU(ifp);
1222 		mtu = rt->rt_rmx.rmx_mtu;
1223 		if (mtu == 0)
1224 			mtu = ifmtu;
1225 		else if (mtu < IPV6_MMTU) {
1226 			/*
1227 			 * RFC2460 section 5, last paragraph:
1228 			 * if we record ICMPv6 too big message with
1229 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1230 			 * or smaller, with fragment header attached.
1231 			 * (fragment header is needed regardless from the
1232 			 * packet size, for translators to identify packets)
1233 			 */
1234 			alwaysfrag = 1;
1235 			mtu = IPV6_MMTU;
1236 		} else if (mtu > ifmtu) {
1237 			/*
1238 			 * The MTU on the route is larger than the MTU on
1239 			 * the interface!  This shouldn't happen, unless the
1240 			 * MTU of the interface has been changed after the
1241 			 * interface was brought up.  Change the MTU in the
1242 			 * route to match the interface MTU (as long as the
1243 			 * field isn't locked).
1244 			 */
1245 			mtu = ifmtu;
1246 			if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
1247 				rt->rt_rmx.rmx_mtu = mtu;
1248 		}
1249 	} else if (ifp) {
1250 		mtu = IN6_LINKMTU(ifp);
1251 	} else
1252 		error = EHOSTUNREACH; /* XXX */
1253 
1254 	*mtup = mtu;
1255 	if (alwaysfragp)
1256 		*alwaysfragp = alwaysfrag;
1257 	return (error);
1258 }
1259 
1260 /*
1261  * IP6 socket option processing.
1262  */
1263 int
1264 ip6_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1265 {
1266 	int optdatalen, uproto;
1267 	void *optdata;
1268 	struct in6pcb *in6p = sotoin6pcb(so);
1269 	int error, optval;
1270 	int level, optname;
1271 
1272 	KASSERT(sopt != NULL);
1273 
1274 	level = sopt->sopt_level;
1275 	optname = sopt->sopt_name;
1276 
1277 	error = optval = 0;
1278 	uproto = (int)so->so_proto->pr_protocol;
1279 
1280 	if (level != IPPROTO_IPV6) {
1281 		return ENOPROTOOPT;
1282 	}
1283 	switch (op) {
1284 	case PRCO_SETOPT:
1285 		switch (optname) {
1286 #ifdef RFC2292
1287 		case IPV6_2292PKTOPTIONS:
1288 			error = ip6_pcbopts(&in6p->in6p_outputopts, so, sopt);
1289 			break;
1290 #endif
1291 
1292 		/*
1293 		 * Use of some Hop-by-Hop options or some
1294 		 * Destination options, might require special
1295 		 * privilege.  That is, normal applications
1296 		 * (without special privilege) might be forbidden
1297 		 * from setting certain options in outgoing packets,
1298 		 * and might never see certain options in received
1299 		 * packets. [RFC 2292 Section 6]
1300 		 * KAME specific note:
1301 		 *  KAME prevents non-privileged users from sending or
1302 		 *  receiving ANY hbh/dst options in order to avoid
1303 		 *  overhead of parsing options in the kernel.
1304 		 */
1305 		case IPV6_RECVHOPOPTS:
1306 		case IPV6_RECVDSTOPTS:
1307 		case IPV6_RECVRTHDRDSTOPTS:
1308 			error = kauth_authorize_network(kauth_cred_get(),
1309 			    KAUTH_NETWORK_IPV6, KAUTH_REQ_NETWORK_IPV6_HOPBYHOP,
1310 			    NULL, NULL, NULL);
1311 			if (error)
1312 				break;
1313 			/* FALLTHROUGH */
1314 		case IPV6_UNICAST_HOPS:
1315 		case IPV6_HOPLIMIT:
1316 		case IPV6_FAITH:
1317 
1318 		case IPV6_RECVPKTINFO:
1319 		case IPV6_RECVHOPLIMIT:
1320 		case IPV6_RECVRTHDR:
1321 		case IPV6_RECVPATHMTU:
1322 		case IPV6_RECVTCLASS:
1323 		case IPV6_V6ONLY:
1324 			error = sockopt_getint(sopt, &optval);
1325 			if (error)
1326 				break;
1327 			switch (optname) {
1328 			case IPV6_UNICAST_HOPS:
1329 				if (optval < -1 || optval >= 256)
1330 					error = EINVAL;
1331 				else {
1332 					/* -1 = kernel default */
1333 					in6p->in6p_hops = optval;
1334 				}
1335 				break;
1336 #define OPTSET(bit) \
1337 do { \
1338 if (optval) \
1339 	in6p->in6p_flags |= (bit); \
1340 else \
1341 	in6p->in6p_flags &= ~(bit); \
1342 } while (/*CONSTCOND*/ 0)
1343 
1344 #ifdef RFC2292
1345 #define OPTSET2292(bit) 			\
1346 do { 						\
1347 in6p->in6p_flags |= IN6P_RFC2292; 	\
1348 if (optval) 				\
1349 	in6p->in6p_flags |= (bit); 	\
1350 else 					\
1351 	in6p->in6p_flags &= ~(bit); 	\
1352 } while (/*CONSTCOND*/ 0)
1353 #endif
1354 
1355 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1356 
1357 			case IPV6_RECVPKTINFO:
1358 #ifdef RFC2292
1359 				/* cannot mix with RFC2292 */
1360 				if (OPTBIT(IN6P_RFC2292)) {
1361 					error = EINVAL;
1362 					break;
1363 				}
1364 #endif
1365 				OPTSET(IN6P_PKTINFO);
1366 				break;
1367 
1368 			case IPV6_HOPLIMIT:
1369 			{
1370 				struct ip6_pktopts **optp;
1371 
1372 #ifdef RFC2292
1373 				/* cannot mix with RFC2292 */
1374 				if (OPTBIT(IN6P_RFC2292)) {
1375 					error = EINVAL;
1376 					break;
1377 				}
1378 #endif
1379 				optp = &in6p->in6p_outputopts;
1380 				error = ip6_pcbopt(IPV6_HOPLIMIT,
1381 						   (u_char *)&optval,
1382 						   sizeof(optval),
1383 						   optp,
1384 						   kauth_cred_get(), uproto);
1385 				break;
1386 			}
1387 
1388 			case IPV6_RECVHOPLIMIT:
1389 #ifdef RFC2292
1390 				/* cannot mix with RFC2292 */
1391 				if (OPTBIT(IN6P_RFC2292)) {
1392 					error = EINVAL;
1393 					break;
1394 				}
1395 #endif
1396 				OPTSET(IN6P_HOPLIMIT);
1397 				break;
1398 
1399 			case IPV6_RECVHOPOPTS:
1400 #ifdef RFC2292
1401 				/* cannot mix with RFC2292 */
1402 				if (OPTBIT(IN6P_RFC2292)) {
1403 					error = EINVAL;
1404 					break;
1405 				}
1406 #endif
1407 				OPTSET(IN6P_HOPOPTS);
1408 				break;
1409 
1410 			case IPV6_RECVDSTOPTS:
1411 #ifdef RFC2292
1412 				/* cannot mix with RFC2292 */
1413 				if (OPTBIT(IN6P_RFC2292)) {
1414 					error = EINVAL;
1415 					break;
1416 				}
1417 #endif
1418 				OPTSET(IN6P_DSTOPTS);
1419 				break;
1420 
1421 			case IPV6_RECVRTHDRDSTOPTS:
1422 #ifdef RFC2292
1423 				/* cannot mix with RFC2292 */
1424 				if (OPTBIT(IN6P_RFC2292)) {
1425 					error = EINVAL;
1426 					break;
1427 				}
1428 #endif
1429 				OPTSET(IN6P_RTHDRDSTOPTS);
1430 				break;
1431 
1432 			case IPV6_RECVRTHDR:
1433 #ifdef RFC2292
1434 				/* cannot mix with RFC2292 */
1435 				if (OPTBIT(IN6P_RFC2292)) {
1436 					error = EINVAL;
1437 					break;
1438 				}
1439 #endif
1440 				OPTSET(IN6P_RTHDR);
1441 				break;
1442 
1443 			case IPV6_FAITH:
1444 				OPTSET(IN6P_FAITH);
1445 				break;
1446 
1447 			case IPV6_RECVPATHMTU:
1448 				/*
1449 				 * We ignore this option for TCP
1450 				 * sockets.
1451 				 * (RFC3542 leaves this case
1452 				 * unspecified.)
1453 				 */
1454 				if (uproto != IPPROTO_TCP)
1455 					OPTSET(IN6P_MTU);
1456 				break;
1457 
1458 			case IPV6_V6ONLY:
1459 				/*
1460 				 * make setsockopt(IPV6_V6ONLY)
1461 				 * available only prior to bind(2).
1462 				 * see ipng mailing list, Jun 22 2001.
1463 				 */
1464 				if (in6p->in6p_lport ||
1465 				    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1466 					error = EINVAL;
1467 					break;
1468 				}
1469 #ifdef INET6_BINDV6ONLY
1470 				if (!optval)
1471 					error = EINVAL;
1472 #else
1473 				OPTSET(IN6P_IPV6_V6ONLY);
1474 #endif
1475 				break;
1476 			case IPV6_RECVTCLASS:
1477 #ifdef RFC2292
1478 				/* cannot mix with RFC2292 XXX */
1479 				if (OPTBIT(IN6P_RFC2292)) {
1480 					error = EINVAL;
1481 					break;
1482 				}
1483 #endif
1484 				OPTSET(IN6P_TCLASS);
1485 				break;
1486 
1487 			}
1488 			break;
1489 
1490 		case IPV6_OTCLASS:
1491 		{
1492 			struct ip6_pktopts **optp;
1493 			u_int8_t tclass;
1494 
1495 			error = sockopt_get(sopt, &tclass, sizeof(tclass));
1496 			if (error)
1497 				break;
1498 			optp = &in6p->in6p_outputopts;
1499 			error = ip6_pcbopt(optname,
1500 					   (u_char *)&tclass,
1501 					   sizeof(tclass),
1502 					   optp,
1503 					   kauth_cred_get(), uproto);
1504 			break;
1505 		}
1506 
1507 		case IPV6_TCLASS:
1508 		case IPV6_DONTFRAG:
1509 		case IPV6_USE_MIN_MTU:
1510 			error = sockopt_getint(sopt, &optval);
1511 			if (error)
1512 				break;
1513 			{
1514 				struct ip6_pktopts **optp;
1515 				optp = &in6p->in6p_outputopts;
1516 				error = ip6_pcbopt(optname,
1517 						   (u_char *)&optval,
1518 						   sizeof(optval),
1519 						   optp,
1520 						   kauth_cred_get(), uproto);
1521 				break;
1522 			}
1523 
1524 #ifdef RFC2292
1525 		case IPV6_2292PKTINFO:
1526 		case IPV6_2292HOPLIMIT:
1527 		case IPV6_2292HOPOPTS:
1528 		case IPV6_2292DSTOPTS:
1529 		case IPV6_2292RTHDR:
1530 			/* RFC 2292 */
1531 			error = sockopt_getint(sopt, &optval);
1532 			if (error)
1533 				break;
1534 
1535 			switch (optname) {
1536 			case IPV6_2292PKTINFO:
1537 				OPTSET2292(IN6P_PKTINFO);
1538 				break;
1539 			case IPV6_2292HOPLIMIT:
1540 				OPTSET2292(IN6P_HOPLIMIT);
1541 				break;
1542 			case IPV6_2292HOPOPTS:
1543 				/*
1544 				 * Check super-user privilege.
1545 				 * See comments for IPV6_RECVHOPOPTS.
1546 				 */
1547 				error =
1548 				    kauth_authorize_network(kauth_cred_get(),
1549 				    KAUTH_NETWORK_IPV6,
1550 				    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL,
1551 				    NULL, NULL);
1552 				if (error)
1553 					return (error);
1554 				OPTSET2292(IN6P_HOPOPTS);
1555 				break;
1556 			case IPV6_2292DSTOPTS:
1557 				error =
1558 				    kauth_authorize_network(kauth_cred_get(),
1559 				    KAUTH_NETWORK_IPV6,
1560 				    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL,
1561 				    NULL, NULL);
1562 				if (error)
1563 					return (error);
1564 				OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1565 				break;
1566 			case IPV6_2292RTHDR:
1567 				OPTSET2292(IN6P_RTHDR);
1568 				break;
1569 			}
1570 			break;
1571 #endif
1572 		case IPV6_PKTINFO:
1573 		case IPV6_HOPOPTS:
1574 		case IPV6_RTHDR:
1575 		case IPV6_DSTOPTS:
1576 		case IPV6_RTHDRDSTOPTS:
1577 		case IPV6_NEXTHOP: {
1578 			/* new advanced API (RFC3542) */
1579 			void *optbuf;
1580 			int optbuflen;
1581 			struct ip6_pktopts **optp;
1582 
1583 #ifdef RFC2292
1584 			/* cannot mix with RFC2292 */
1585 			if (OPTBIT(IN6P_RFC2292)) {
1586 				error = EINVAL;
1587 				break;
1588 			}
1589 #endif
1590 
1591 			optbuflen = sopt->sopt_size;
1592 			optbuf = malloc(optbuflen, M_IP6OPT, M_NOWAIT);
1593 			if (optbuf == NULL) {
1594 				error = ENOBUFS;
1595 				break;
1596 			}
1597 
1598 			sockopt_get(sopt, optbuf, optbuflen);
1599 			optp = &in6p->in6p_outputopts;
1600 			error = ip6_pcbopt(optname, optbuf, optbuflen,
1601 			    optp, kauth_cred_get(), uproto);
1602 			break;
1603 			}
1604 #undef OPTSET
1605 
1606 		case IPV6_MULTICAST_IF:
1607 		case IPV6_MULTICAST_HOPS:
1608 		case IPV6_MULTICAST_LOOP:
1609 		case IPV6_JOIN_GROUP:
1610 		case IPV6_LEAVE_GROUP:
1611 			error = ip6_setmoptions(sopt, &in6p->in6p_moptions);
1612 			break;
1613 
1614 		case IPV6_PORTRANGE:
1615 			error = sockopt_getint(sopt, &optval);
1616 			if (error)
1617 				break;
1618 
1619 			switch (optval) {
1620 			case IPV6_PORTRANGE_DEFAULT:
1621 				in6p->in6p_flags &= ~(IN6P_LOWPORT);
1622 				in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1623 				break;
1624 
1625 			case IPV6_PORTRANGE_HIGH:
1626 				in6p->in6p_flags &= ~(IN6P_LOWPORT);
1627 				in6p->in6p_flags |= IN6P_HIGHPORT;
1628 				break;
1629 
1630 			case IPV6_PORTRANGE_LOW:
1631 				in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1632 				in6p->in6p_flags |= IN6P_LOWPORT;
1633 				break;
1634 
1635 			default:
1636 				error = EINVAL;
1637 				break;
1638 			}
1639 			break;
1640 
1641 		case IPV6_PORTALGO:
1642 			error = sockopt_getint(sopt, &optval);
1643 			if (error)
1644 				break;
1645 
1646 			error = portalgo_algo_index_select(
1647 			    (struct inpcb_hdr *)in6p, optval);
1648 			break;
1649 
1650 #if defined(FAST_IPSEC)
1651 		case IPV6_IPSEC_POLICY:
1652 			error = ipsec6_set_policy(in6p, optname,
1653 			    sopt->sopt_data, sopt->sopt_size, kauth_cred_get());
1654 			break;
1655 #endif /* IPSEC */
1656 
1657 		default:
1658 			error = ENOPROTOOPT;
1659 			break;
1660 		}
1661 		break;
1662 
1663 	case PRCO_GETOPT:
1664 		switch (optname) {
1665 #ifdef RFC2292
1666 		case IPV6_2292PKTOPTIONS:
1667 			/*
1668 			 * RFC3542 (effectively) deprecated the
1669 			 * semantics of the 2292-style pktoptions.
1670 			 * Since it was not reliable in nature (i.e.,
1671 			 * applications had to expect the lack of some
1672 			 * information after all), it would make sense
1673 			 * to simplify this part by always returning
1674 			 * empty data.
1675 			 */
1676 			break;
1677 #endif
1678 
1679 		case IPV6_RECVHOPOPTS:
1680 		case IPV6_RECVDSTOPTS:
1681 		case IPV6_RECVRTHDRDSTOPTS:
1682 		case IPV6_UNICAST_HOPS:
1683 		case IPV6_RECVPKTINFO:
1684 		case IPV6_RECVHOPLIMIT:
1685 		case IPV6_RECVRTHDR:
1686 		case IPV6_RECVPATHMTU:
1687 
1688 		case IPV6_FAITH:
1689 		case IPV6_V6ONLY:
1690 		case IPV6_PORTRANGE:
1691 		case IPV6_RECVTCLASS:
1692 			switch (optname) {
1693 
1694 			case IPV6_RECVHOPOPTS:
1695 				optval = OPTBIT(IN6P_HOPOPTS);
1696 				break;
1697 
1698 			case IPV6_RECVDSTOPTS:
1699 				optval = OPTBIT(IN6P_DSTOPTS);
1700 				break;
1701 
1702 			case IPV6_RECVRTHDRDSTOPTS:
1703 				optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1704 				break;
1705 
1706 			case IPV6_UNICAST_HOPS:
1707 				optval = in6p->in6p_hops;
1708 				break;
1709 
1710 			case IPV6_RECVPKTINFO:
1711 				optval = OPTBIT(IN6P_PKTINFO);
1712 				break;
1713 
1714 			case IPV6_RECVHOPLIMIT:
1715 				optval = OPTBIT(IN6P_HOPLIMIT);
1716 				break;
1717 
1718 			case IPV6_RECVRTHDR:
1719 				optval = OPTBIT(IN6P_RTHDR);
1720 				break;
1721 
1722 			case IPV6_RECVPATHMTU:
1723 				optval = OPTBIT(IN6P_MTU);
1724 				break;
1725 
1726 			case IPV6_FAITH:
1727 				optval = OPTBIT(IN6P_FAITH);
1728 				break;
1729 
1730 			case IPV6_V6ONLY:
1731 				optval = OPTBIT(IN6P_IPV6_V6ONLY);
1732 				break;
1733 
1734 			case IPV6_PORTRANGE:
1735 			    {
1736 				int flags;
1737 				flags = in6p->in6p_flags;
1738 				if (flags & IN6P_HIGHPORT)
1739 					optval = IPV6_PORTRANGE_HIGH;
1740 				else if (flags & IN6P_LOWPORT)
1741 					optval = IPV6_PORTRANGE_LOW;
1742 				else
1743 					optval = 0;
1744 				break;
1745 			    }
1746 			case IPV6_RECVTCLASS:
1747 				optval = OPTBIT(IN6P_TCLASS);
1748 				break;
1749 
1750 			}
1751 			if (error)
1752 				break;
1753 			error = sockopt_setint(sopt, optval);
1754 			break;
1755 
1756 		case IPV6_PATHMTU:
1757 		    {
1758 			u_long pmtu = 0;
1759 			struct ip6_mtuinfo mtuinfo;
1760 			struct route *ro = &in6p->in6p_route;
1761 
1762 			if (!(so->so_state & SS_ISCONNECTED))
1763 				return (ENOTCONN);
1764 			/*
1765 			 * XXX: we dot not consider the case of source
1766 			 * routing, or optional information to specify
1767 			 * the outgoing interface.
1768 			 */
1769 			error = ip6_getpmtu(ro, NULL, NULL,
1770 			    &in6p->in6p_faddr, &pmtu, NULL);
1771 			if (error)
1772 				break;
1773 			if (pmtu > IPV6_MAXPACKET)
1774 				pmtu = IPV6_MAXPACKET;
1775 
1776 			memset(&mtuinfo, 0, sizeof(mtuinfo));
1777 			mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1778 			optdata = (void *)&mtuinfo;
1779 			optdatalen = sizeof(mtuinfo);
1780 			if (optdatalen > MCLBYTES)
1781 				return (EMSGSIZE); /* XXX */
1782 			error = sockopt_set(sopt, optdata, optdatalen);
1783 			break;
1784 		    }
1785 
1786 #ifdef RFC2292
1787 		case IPV6_2292PKTINFO:
1788 		case IPV6_2292HOPLIMIT:
1789 		case IPV6_2292HOPOPTS:
1790 		case IPV6_2292RTHDR:
1791 		case IPV6_2292DSTOPTS:
1792 			switch (optname) {
1793 			case IPV6_2292PKTINFO:
1794 				optval = OPTBIT(IN6P_PKTINFO);
1795 				break;
1796 			case IPV6_2292HOPLIMIT:
1797 				optval = OPTBIT(IN6P_HOPLIMIT);
1798 				break;
1799 			case IPV6_2292HOPOPTS:
1800 				optval = OPTBIT(IN6P_HOPOPTS);
1801 				break;
1802 			case IPV6_2292RTHDR:
1803 				optval = OPTBIT(IN6P_RTHDR);
1804 				break;
1805 			case IPV6_2292DSTOPTS:
1806 				optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1807 				break;
1808 			}
1809 			error = sockopt_setint(sopt, optval);
1810 			break;
1811 #endif
1812 		case IPV6_PKTINFO:
1813 		case IPV6_HOPOPTS:
1814 		case IPV6_RTHDR:
1815 		case IPV6_DSTOPTS:
1816 		case IPV6_RTHDRDSTOPTS:
1817 		case IPV6_NEXTHOP:
1818 		case IPV6_OTCLASS:
1819 		case IPV6_TCLASS:
1820 		case IPV6_DONTFRAG:
1821 		case IPV6_USE_MIN_MTU:
1822 			error = ip6_getpcbopt(in6p->in6p_outputopts,
1823 			    optname, sopt);
1824 			break;
1825 
1826 		case IPV6_MULTICAST_IF:
1827 		case IPV6_MULTICAST_HOPS:
1828 		case IPV6_MULTICAST_LOOP:
1829 		case IPV6_JOIN_GROUP:
1830 		case IPV6_LEAVE_GROUP:
1831 			error = ip6_getmoptions(sopt, in6p->in6p_moptions);
1832 			break;
1833 
1834 		case IPV6_PORTALGO:
1835 			optval = ((struct inpcb_hdr *)in6p)->inph_portalgo;
1836 			error = sockopt_setint(sopt, optval);
1837 			break;
1838 
1839 #if defined(FAST_IPSEC)
1840 		case IPV6_IPSEC_POLICY:
1841 		    {
1842 			struct mbuf *m = NULL;
1843 
1844 			/* XXX this will return EINVAL as sopt is empty */
1845 			error = ipsec6_get_policy(in6p, sopt->sopt_data,
1846 			    sopt->sopt_size, &m);
1847 			if (!error)
1848 				error = sockopt_setmbuf(sopt, m);
1849 
1850 			break;
1851 		    }
1852 #endif /* IPSEC */
1853 
1854 		default:
1855 			error = ENOPROTOOPT;
1856 			break;
1857 		}
1858 		break;
1859 	}
1860 	return (error);
1861 }
1862 
1863 int
1864 ip6_raw_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1865 {
1866 	int error = 0, optval;
1867 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
1868 	struct in6pcb *in6p = sotoin6pcb(so);
1869 	int level, optname;
1870 
1871 	KASSERT(sopt != NULL);
1872 
1873 	level = sopt->sopt_level;
1874 	optname = sopt->sopt_name;
1875 
1876 	if (level != IPPROTO_IPV6) {
1877 		return ENOPROTOOPT;
1878 	}
1879 
1880 	switch (optname) {
1881 	case IPV6_CHECKSUM:
1882 		/*
1883 		 * For ICMPv6 sockets, no modification allowed for checksum
1884 		 * offset, permit "no change" values to help existing apps.
1885 		 *
1886 		 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM
1887 		 * for an ICMPv6 socket will fail."  The current
1888 		 * behavior does not meet RFC3542.
1889 		 */
1890 		switch (op) {
1891 		case PRCO_SETOPT:
1892 			error = sockopt_getint(sopt, &optval);
1893 			if (error)
1894 				break;
1895 			if ((optval % 2) != 0) {
1896 				/* the API assumes even offset values */
1897 				error = EINVAL;
1898 			} else if (so->so_proto->pr_protocol ==
1899 			    IPPROTO_ICMPV6) {
1900 				if (optval != icmp6off)
1901 					error = EINVAL;
1902 			} else
1903 				in6p->in6p_cksum = optval;
1904 			break;
1905 
1906 		case PRCO_GETOPT:
1907 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
1908 				optval = icmp6off;
1909 			else
1910 				optval = in6p->in6p_cksum;
1911 
1912 			error = sockopt_setint(sopt, optval);
1913 			break;
1914 
1915 		default:
1916 			error = EINVAL;
1917 			break;
1918 		}
1919 		break;
1920 
1921 	default:
1922 		error = ENOPROTOOPT;
1923 		break;
1924 	}
1925 
1926 	return (error);
1927 }
1928 
1929 #ifdef RFC2292
1930 /*
1931  * Set up IP6 options in pcb for insertion in output packets or
1932  * specifying behavior of outgoing packets.
1933  */
1934 static int
1935 ip6_pcbopts(struct ip6_pktopts **pktopt, struct socket *so,
1936     struct sockopt *sopt)
1937 {
1938 	struct ip6_pktopts *opt = *pktopt;
1939 	struct mbuf *m;
1940 	int error = 0;
1941 
1942 	/* turn off any old options. */
1943 	if (opt) {
1944 #ifdef DIAGNOSTIC
1945 	    if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
1946 		opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
1947 		opt->ip6po_rhinfo.ip6po_rhi_rthdr)
1948 		    printf("ip6_pcbopts: all specified options are cleared.\n");
1949 #endif
1950 		ip6_clearpktopts(opt, -1);
1951 	} else {
1952 		opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
1953 		if (opt == NULL)
1954 			return (ENOBUFS);
1955 	}
1956 	*pktopt = NULL;
1957 
1958 	if (sopt == NULL || sopt->sopt_size == 0) {
1959 		/*
1960 		 * Only turning off any previous options, regardless of
1961 		 * whether the opt is just created or given.
1962 		 */
1963 		free(opt, M_IP6OPT);
1964 		return (0);
1965 	}
1966 
1967 	/*  set options specified by user. */
1968 	m = sockopt_getmbuf(sopt);
1969 	if (m == NULL) {
1970 		free(opt, M_IP6OPT);
1971 		return (ENOBUFS);
1972 	}
1973 
1974 	error = ip6_setpktopts(m, opt, NULL, kauth_cred_get(),
1975 	    so->so_proto->pr_protocol);
1976 	m_freem(m);
1977 	if (error != 0) {
1978 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
1979 		free(opt, M_IP6OPT);
1980 		return (error);
1981 	}
1982 	*pktopt = opt;
1983 	return (0);
1984 }
1985 #endif
1986 
1987 /*
1988  * initialize ip6_pktopts.  beware that there are non-zero default values in
1989  * the struct.
1990  */
1991 void
1992 ip6_initpktopts(struct ip6_pktopts *opt)
1993 {
1994 
1995 	memset(opt, 0, sizeof(*opt));
1996 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
1997 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
1998 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
1999 }
2000 
2001 #define sin6tosa(sin6)	((struct sockaddr *)(sin6)) /* XXX */
2002 static int
2003 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2004     kauth_cred_t cred, int uproto)
2005 {
2006 	struct ip6_pktopts *opt;
2007 
2008 	if (*pktopt == NULL) {
2009 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2010 		    M_NOWAIT);
2011 		if (*pktopt == NULL)
2012 			return (ENOBUFS);
2013 
2014 		ip6_initpktopts(*pktopt);
2015 	}
2016 	opt = *pktopt;
2017 
2018 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2019 }
2020 
2021 static int
2022 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2023 {
2024 	void *optdata = NULL;
2025 	int optdatalen = 0;
2026 	struct ip6_ext *ip6e;
2027 	int error = 0;
2028 	struct in6_pktinfo null_pktinfo;
2029 	int deftclass = 0, on;
2030 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2031 
2032 	switch (optname) {
2033 	case IPV6_PKTINFO:
2034 		if (pktopt && pktopt->ip6po_pktinfo)
2035 			optdata = (void *)pktopt->ip6po_pktinfo;
2036 		else {
2037 			/* XXX: we don't have to do this every time... */
2038 			memset(&null_pktinfo, 0, sizeof(null_pktinfo));
2039 			optdata = (void *)&null_pktinfo;
2040 		}
2041 		optdatalen = sizeof(struct in6_pktinfo);
2042 		break;
2043 	case IPV6_OTCLASS:
2044 		/* XXX */
2045 		return (EINVAL);
2046 	case IPV6_TCLASS:
2047 		if (pktopt && pktopt->ip6po_tclass >= 0)
2048 			optdata = (void *)&pktopt->ip6po_tclass;
2049 		else
2050 			optdata = (void *)&deftclass;
2051 		optdatalen = sizeof(int);
2052 		break;
2053 	case IPV6_HOPOPTS:
2054 		if (pktopt && pktopt->ip6po_hbh) {
2055 			optdata = (void *)pktopt->ip6po_hbh;
2056 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2057 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2058 		}
2059 		break;
2060 	case IPV6_RTHDR:
2061 		if (pktopt && pktopt->ip6po_rthdr) {
2062 			optdata = (void *)pktopt->ip6po_rthdr;
2063 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2064 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2065 		}
2066 		break;
2067 	case IPV6_RTHDRDSTOPTS:
2068 		if (pktopt && pktopt->ip6po_dest1) {
2069 			optdata = (void *)pktopt->ip6po_dest1;
2070 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2071 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2072 		}
2073 		break;
2074 	case IPV6_DSTOPTS:
2075 		if (pktopt && pktopt->ip6po_dest2) {
2076 			optdata = (void *)pktopt->ip6po_dest2;
2077 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2078 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2079 		}
2080 		break;
2081 	case IPV6_NEXTHOP:
2082 		if (pktopt && pktopt->ip6po_nexthop) {
2083 			optdata = (void *)pktopt->ip6po_nexthop;
2084 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2085 		}
2086 		break;
2087 	case IPV6_USE_MIN_MTU:
2088 		if (pktopt)
2089 			optdata = (void *)&pktopt->ip6po_minmtu;
2090 		else
2091 			optdata = (void *)&defminmtu;
2092 		optdatalen = sizeof(int);
2093 		break;
2094 	case IPV6_DONTFRAG:
2095 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2096 			on = 1;
2097 		else
2098 			on = 0;
2099 		optdata = (void *)&on;
2100 		optdatalen = sizeof(on);
2101 		break;
2102 	default:		/* should not happen */
2103 #ifdef DIAGNOSTIC
2104 		panic("ip6_getpcbopt: unexpected option\n");
2105 #endif
2106 		return (ENOPROTOOPT);
2107 	}
2108 
2109 	error = sockopt_set(sopt, optdata, optdatalen);
2110 
2111 	return (error);
2112 }
2113 
2114 void
2115 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2116 {
2117 	if (optname == -1 || optname == IPV6_PKTINFO) {
2118 		if (pktopt->ip6po_pktinfo)
2119 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2120 		pktopt->ip6po_pktinfo = NULL;
2121 	}
2122 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2123 		pktopt->ip6po_hlim = -1;
2124 	if (optname == -1 || optname == IPV6_TCLASS)
2125 		pktopt->ip6po_tclass = -1;
2126 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2127 		rtcache_free(&pktopt->ip6po_nextroute);
2128 		if (pktopt->ip6po_nexthop)
2129 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2130 		pktopt->ip6po_nexthop = NULL;
2131 	}
2132 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2133 		if (pktopt->ip6po_hbh)
2134 			free(pktopt->ip6po_hbh, M_IP6OPT);
2135 		pktopt->ip6po_hbh = NULL;
2136 	}
2137 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2138 		if (pktopt->ip6po_dest1)
2139 			free(pktopt->ip6po_dest1, M_IP6OPT);
2140 		pktopt->ip6po_dest1 = NULL;
2141 	}
2142 	if (optname == -1 || optname == IPV6_RTHDR) {
2143 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2144 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2145 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2146 		rtcache_free(&pktopt->ip6po_route);
2147 	}
2148 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2149 		if (pktopt->ip6po_dest2)
2150 			free(pktopt->ip6po_dest2, M_IP6OPT);
2151 		pktopt->ip6po_dest2 = NULL;
2152 	}
2153 }
2154 
2155 #define PKTOPT_EXTHDRCPY(type) 					\
2156 do {								\
2157 	if (src->type) {					\
2158 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2159 		dst->type = malloc(hlen, M_IP6OPT, canwait);	\
2160 		if (dst->type == NULL)				\
2161 			goto bad;				\
2162 		memcpy(dst->type, src->type, hlen);		\
2163 	}							\
2164 } while (/*CONSTCOND*/ 0)
2165 
2166 static int
2167 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2168 {
2169 	dst->ip6po_hlim = src->ip6po_hlim;
2170 	dst->ip6po_tclass = src->ip6po_tclass;
2171 	dst->ip6po_flags = src->ip6po_flags;
2172 	if (src->ip6po_pktinfo) {
2173 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2174 		    M_IP6OPT, canwait);
2175 		if (dst->ip6po_pktinfo == NULL)
2176 			goto bad;
2177 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2178 	}
2179 	if (src->ip6po_nexthop) {
2180 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2181 		    M_IP6OPT, canwait);
2182 		if (dst->ip6po_nexthop == NULL)
2183 			goto bad;
2184 		memcpy(dst->ip6po_nexthop, src->ip6po_nexthop,
2185 		    src->ip6po_nexthop->sa_len);
2186 	}
2187 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2188 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2189 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2190 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2191 	return (0);
2192 
2193   bad:
2194 	if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
2195 	if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
2196 	if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
2197 	if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
2198 	if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
2199 	if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
2200 
2201 	return (ENOBUFS);
2202 }
2203 #undef PKTOPT_EXTHDRCPY
2204 
2205 struct ip6_pktopts *
2206 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2207 {
2208 	int error;
2209 	struct ip6_pktopts *dst;
2210 
2211 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2212 	if (dst == NULL)
2213 		return (NULL);
2214 	ip6_initpktopts(dst);
2215 
2216 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2217 		free(dst, M_IP6OPT);
2218 		return (NULL);
2219 	}
2220 
2221 	return (dst);
2222 }
2223 
2224 void
2225 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2226 {
2227 	if (pktopt == NULL)
2228 		return;
2229 
2230 	ip6_clearpktopts(pktopt, -1);
2231 
2232 	free(pktopt, M_IP6OPT);
2233 }
2234 
2235 /*
2236  * Set the IP6 multicast options in response to user setsockopt().
2237  */
2238 static int
2239 ip6_setmoptions(const struct sockopt *sopt, struct ip6_moptions **im6op)
2240 {
2241 	int error = 0;
2242 	u_int loop, ifindex;
2243 	struct ipv6_mreq mreq;
2244 	struct ifnet *ifp;
2245 	struct ip6_moptions *im6o = *im6op;
2246 	struct route ro;
2247 	struct in6_multi_mship *imm;
2248 	struct lwp *l = curlwp;	/* XXX */
2249 
2250 	if (im6o == NULL) {
2251 		/*
2252 		 * No multicast option buffer attached to the pcb;
2253 		 * allocate one and initialize to default values.
2254 		 */
2255 		im6o = malloc(sizeof(*im6o), M_IPMOPTS, M_NOWAIT);
2256 		if (im6o == NULL)
2257 			return (ENOBUFS);
2258 
2259 		*im6op = im6o;
2260 		im6o->im6o_multicast_ifp = NULL;
2261 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2262 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2263 		LIST_INIT(&im6o->im6o_memberships);
2264 	}
2265 
2266 	switch (sopt->sopt_name) {
2267 
2268 	case IPV6_MULTICAST_IF:
2269 		/*
2270 		 * Select the interface for outgoing multicast packets.
2271 		 */
2272 		error = sockopt_get(sopt, &ifindex, sizeof(ifindex));
2273 		if (error != 0)
2274 			break;
2275 
2276 		if (ifindex != 0) {
2277 			if (if_indexlim <= ifindex || !ifindex2ifnet[ifindex]) {
2278 				error = ENXIO;	/* XXX EINVAL? */
2279 				break;
2280 			}
2281 			ifp = ifindex2ifnet[ifindex];
2282 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
2283 				error = EADDRNOTAVAIL;
2284 				break;
2285 			}
2286 		} else
2287 			ifp = NULL;
2288 		im6o->im6o_multicast_ifp = ifp;
2289 		break;
2290 
2291 	case IPV6_MULTICAST_HOPS:
2292 	    {
2293 		/*
2294 		 * Set the IP6 hoplimit for outgoing multicast packets.
2295 		 */
2296 		int optval;
2297 
2298 		error = sockopt_getint(sopt, &optval);
2299 		if (error != 0)
2300 			break;
2301 
2302 		if (optval < -1 || optval >= 256)
2303 			error = EINVAL;
2304 		else if (optval == -1)
2305 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2306 		else
2307 			im6o->im6o_multicast_hlim = optval;
2308 		break;
2309 	    }
2310 
2311 	case IPV6_MULTICAST_LOOP:
2312 		/*
2313 		 * Set the loopback flag for outgoing multicast packets.
2314 		 * Must be zero or one.
2315 		 */
2316 		error = sockopt_get(sopt, &loop, sizeof(loop));
2317 		if (error != 0)
2318 			break;
2319 		if (loop > 1) {
2320 			error = EINVAL;
2321 			break;
2322 		}
2323 		im6o->im6o_multicast_loop = loop;
2324 		break;
2325 
2326 	case IPV6_JOIN_GROUP:
2327 		/*
2328 		 * Add a multicast group membership.
2329 		 * Group must be a valid IP6 multicast address.
2330 		 */
2331 		error = sockopt_get(sopt, &mreq, sizeof(mreq));
2332 		if (error != 0)
2333 			break;
2334 
2335 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq.ipv6mr_multiaddr)) {
2336 			/*
2337 			 * We use the unspecified address to specify to accept
2338 			 * all multicast addresses. Only super user is allowed
2339 			 * to do this.
2340 			 */
2341 			if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_IPV6,
2342 			    KAUTH_REQ_NETWORK_IPV6_JOIN_MULTICAST, NULL, NULL, NULL))
2343 			{
2344 				error = EACCES;
2345 				break;
2346 			}
2347 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq.ipv6mr_multiaddr)) {
2348 			error = EINVAL;
2349 			break;
2350 		}
2351 
2352 		/*
2353 		 * If no interface was explicitly specified, choose an
2354 		 * appropriate one according to the given multicast address.
2355 		 */
2356 		if (mreq.ipv6mr_interface == 0) {
2357 			struct rtentry *rt;
2358 			union {
2359 				struct sockaddr		dst;
2360 				struct sockaddr_in6	dst6;
2361 			} u;
2362 
2363 			/*
2364 			 * Look up the routing table for the
2365 			 * address, and choose the outgoing interface.
2366 			 *   XXX: is it a good approach?
2367 			 */
2368 			memset(&ro, 0, sizeof(ro));
2369 			sockaddr_in6_init(&u.dst6, &mreq.ipv6mr_multiaddr, 0,
2370 			    0, 0);
2371 			rtcache_setdst(&ro, &u.dst);
2372 			ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp
2373 			                                        : NULL;
2374 			rtcache_free(&ro);
2375 		} else {
2376 			/*
2377 			 * If the interface is specified, validate it.
2378 			 */
2379 			if (if_indexlim <= mreq.ipv6mr_interface ||
2380 			    !ifindex2ifnet[mreq.ipv6mr_interface]) {
2381 				error = ENXIO;	/* XXX EINVAL? */
2382 				break;
2383 			}
2384 			ifp = ifindex2ifnet[mreq.ipv6mr_interface];
2385 		}
2386 
2387 		/*
2388 		 * See if we found an interface, and confirm that it
2389 		 * supports multicast
2390 		 */
2391 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2392 			error = EADDRNOTAVAIL;
2393 			break;
2394 		}
2395 
2396 		if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) {
2397 			error = EADDRNOTAVAIL; /* XXX: should not happen */
2398 			break;
2399 		}
2400 
2401 		/*
2402 		 * See if the membership already exists.
2403 		 */
2404 		for (imm = im6o->im6o_memberships.lh_first;
2405 		     imm != NULL; imm = imm->i6mm_chain.le_next)
2406 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
2407 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2408 			    &mreq.ipv6mr_multiaddr))
2409 				break;
2410 		if (imm != NULL) {
2411 			error = EADDRINUSE;
2412 			break;
2413 		}
2414 		/*
2415 		 * Everything looks good; add a new record to the multicast
2416 		 * address list for the given interface.
2417 		 */
2418 		imm = in6_joingroup(ifp, &mreq.ipv6mr_multiaddr, &error, 0);
2419 		if (imm == NULL)
2420 			break;
2421 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2422 		break;
2423 
2424 	case IPV6_LEAVE_GROUP:
2425 		/*
2426 		 * Drop a multicast group membership.
2427 		 * Group must be a valid IP6 multicast address.
2428 		 */
2429 		error = sockopt_get(sopt, &mreq, sizeof(mreq));
2430 		if (error != 0)
2431 			break;
2432 
2433 		/*
2434 		 * If an interface address was specified, get a pointer
2435 		 * to its ifnet structure.
2436 		 */
2437 		if (mreq.ipv6mr_interface != 0) {
2438 			if (if_indexlim <= mreq.ipv6mr_interface ||
2439 			    !ifindex2ifnet[mreq.ipv6mr_interface]) {
2440 				error = ENXIO;	/* XXX EINVAL? */
2441 				break;
2442 			}
2443 			ifp = ifindex2ifnet[mreq.ipv6mr_interface];
2444 		} else
2445 			ifp = NULL;
2446 
2447 		/* Fill in the scope zone ID */
2448 		if (ifp) {
2449 			if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) {
2450 				/* XXX: should not happen */
2451 				error = EADDRNOTAVAIL;
2452 				break;
2453 			}
2454 		} else if (mreq.ipv6mr_interface != 0) {
2455 			/*
2456 			 * XXX: This case would happens when the (positive)
2457 			 * index is in the valid range, but the corresponding
2458 			 * interface has been detached dynamically.  The above
2459 			 * check probably avoids such case to happen here, but
2460 			 * we check it explicitly for safety.
2461 			 */
2462 			error = EADDRNOTAVAIL;
2463 			break;
2464 		} else {	/* ipv6mr_interface == 0 */
2465 			struct sockaddr_in6 sa6_mc;
2466 
2467 			/*
2468 			 * The API spec says as follows:
2469 			 *  If the interface index is specified as 0, the
2470 			 *  system may choose a multicast group membership to
2471 			 *  drop by matching the multicast address only.
2472 			 * On the other hand, we cannot disambiguate the scope
2473 			 * zone unless an interface is provided.  Thus, we
2474 			 * check if there's ambiguity with the default scope
2475 			 * zone as the last resort.
2476 			 */
2477 			sockaddr_in6_init(&sa6_mc, &mreq.ipv6mr_multiaddr,
2478 			    0, 0, 0);
2479 			error = sa6_embedscope(&sa6_mc, ip6_use_defzone);
2480 			if (error != 0)
2481 				break;
2482 			mreq.ipv6mr_multiaddr = sa6_mc.sin6_addr;
2483 		}
2484 
2485 		/*
2486 		 * Find the membership in the membership list.
2487 		 */
2488 		for (imm = im6o->im6o_memberships.lh_first;
2489 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
2490 			if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2491 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2492 			    &mreq.ipv6mr_multiaddr))
2493 				break;
2494 		}
2495 		if (imm == NULL) {
2496 			/* Unable to resolve interface */
2497 			error = EADDRNOTAVAIL;
2498 			break;
2499 		}
2500 		/*
2501 		 * Give up the multicast address record to which the
2502 		 * membership points.
2503 		 */
2504 		LIST_REMOVE(imm, i6mm_chain);
2505 		in6_leavegroup(imm);
2506 		break;
2507 
2508 	default:
2509 		error = EOPNOTSUPP;
2510 		break;
2511 	}
2512 
2513 	/*
2514 	 * If all options have default values, no need to keep the mbuf.
2515 	 */
2516 	if (im6o->im6o_multicast_ifp == NULL &&
2517 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2518 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2519 	    im6o->im6o_memberships.lh_first == NULL) {
2520 		free(*im6op, M_IPMOPTS);
2521 		*im6op = NULL;
2522 	}
2523 
2524 	return (error);
2525 }
2526 
2527 /*
2528  * Return the IP6 multicast options in response to user getsockopt().
2529  */
2530 static int
2531 ip6_getmoptions(struct sockopt *sopt, struct ip6_moptions *im6o)
2532 {
2533 	u_int optval;
2534 	int error;
2535 
2536 	switch (sopt->sopt_name) {
2537 	case IPV6_MULTICAST_IF:
2538 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2539 			optval = 0;
2540 		else
2541 			optval = im6o->im6o_multicast_ifp->if_index;
2542 
2543 		error = sockopt_set(sopt, &optval, sizeof(optval));
2544 		break;
2545 
2546 	case IPV6_MULTICAST_HOPS:
2547 		if (im6o == NULL)
2548 			optval = ip6_defmcasthlim;
2549 		else
2550 			optval = im6o->im6o_multicast_hlim;
2551 
2552 		error = sockopt_set(sopt, &optval, sizeof(optval));
2553 		break;
2554 
2555 	case IPV6_MULTICAST_LOOP:
2556 		if (im6o == NULL)
2557 			optval = ip6_defmcasthlim;
2558 		else
2559 			optval = im6o->im6o_multicast_loop;
2560 
2561 		error = sockopt_set(sopt, &optval, sizeof(optval));
2562 		break;
2563 
2564 	default:
2565 		error = EOPNOTSUPP;
2566 	}
2567 
2568 	return (error);
2569 }
2570 
2571 /*
2572  * Discard the IP6 multicast options.
2573  */
2574 void
2575 ip6_freemoptions(struct ip6_moptions *im6o)
2576 {
2577 	struct in6_multi_mship *imm;
2578 
2579 	if (im6o == NULL)
2580 		return;
2581 
2582 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2583 		LIST_REMOVE(imm, i6mm_chain);
2584 		in6_leavegroup(imm);
2585 	}
2586 	free(im6o, M_IPMOPTS);
2587 }
2588 
2589 /*
2590  * Set IPv6 outgoing packet options based on advanced API.
2591  */
2592 int
2593 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2594 	struct ip6_pktopts *stickyopt, kauth_cred_t cred, int uproto)
2595 {
2596 	struct cmsghdr *cm = 0;
2597 
2598 	if (control == NULL || opt == NULL)
2599 		return (EINVAL);
2600 
2601 	ip6_initpktopts(opt);
2602 	if (stickyopt) {
2603 		int error;
2604 
2605 		/*
2606 		 * If stickyopt is provided, make a local copy of the options
2607 		 * for this particular packet, then override them by ancillary
2608 		 * objects.
2609 		 * XXX: copypktopts() does not copy the cached route to a next
2610 		 * hop (if any).  This is not very good in terms of efficiency,
2611 		 * but we can allow this since this option should be rarely
2612 		 * used.
2613 		 */
2614 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2615 			return (error);
2616 	}
2617 
2618 	/*
2619 	 * XXX: Currently, we assume all the optional information is stored
2620 	 * in a single mbuf.
2621 	 */
2622 	if (control->m_next)
2623 		return (EINVAL);
2624 
2625 	/* XXX if cm->cmsg_len is not aligned, control->m_len can become <0 */
2626 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2627 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2628 		int error;
2629 
2630 		if (control->m_len < CMSG_LEN(0))
2631 			return (EINVAL);
2632 
2633 		cm = mtod(control, struct cmsghdr *);
2634 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2635 			return (EINVAL);
2636 		if (cm->cmsg_level != IPPROTO_IPV6)
2637 			continue;
2638 
2639 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2640 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2641 		if (error)
2642 			return (error);
2643 	}
2644 
2645 	return (0);
2646 }
2647 
2648 /*
2649  * Set a particular packet option, as a sticky option or an ancillary data
2650  * item.  "len" can be 0 only when it's a sticky option.
2651  * We have 4 cases of combination of "sticky" and "cmsg":
2652  * "sticky=0, cmsg=0": impossible
2653  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2654  * "sticky=1, cmsg=0": RFC3542 socket option
2655  * "sticky=1, cmsg=1": RFC2292 socket option
2656  */
2657 static int
2658 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2659     kauth_cred_t cred, int sticky, int cmsg, int uproto)
2660 {
2661 	int minmtupolicy;
2662 	int error;
2663 
2664 	if (!sticky && !cmsg) {
2665 #ifdef DIAGNOSTIC
2666 		printf("ip6_setpktopt: impossible case\n");
2667 #endif
2668 		return (EINVAL);
2669 	}
2670 
2671 	/*
2672 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2673 	 * not be specified in the context of RFC3542.  Conversely,
2674 	 * RFC3542 types should not be specified in the context of RFC2292.
2675 	 */
2676 	if (!cmsg) {
2677 		switch (optname) {
2678 		case IPV6_2292PKTINFO:
2679 		case IPV6_2292HOPLIMIT:
2680 		case IPV6_2292NEXTHOP:
2681 		case IPV6_2292HOPOPTS:
2682 		case IPV6_2292DSTOPTS:
2683 		case IPV6_2292RTHDR:
2684 		case IPV6_2292PKTOPTIONS:
2685 			return (ENOPROTOOPT);
2686 		}
2687 	}
2688 	if (sticky && cmsg) {
2689 		switch (optname) {
2690 		case IPV6_PKTINFO:
2691 		case IPV6_HOPLIMIT:
2692 		case IPV6_NEXTHOP:
2693 		case IPV6_HOPOPTS:
2694 		case IPV6_DSTOPTS:
2695 		case IPV6_RTHDRDSTOPTS:
2696 		case IPV6_RTHDR:
2697 		case IPV6_USE_MIN_MTU:
2698 		case IPV6_DONTFRAG:
2699 		case IPV6_OTCLASS:
2700 		case IPV6_TCLASS:
2701 			return (ENOPROTOOPT);
2702 		}
2703 	}
2704 
2705 	switch (optname) {
2706 #ifdef RFC2292
2707 	case IPV6_2292PKTINFO:
2708 #endif
2709 	case IPV6_PKTINFO:
2710 	{
2711 		struct ifnet *ifp = NULL;
2712 		struct in6_pktinfo *pktinfo;
2713 
2714 		if (len != sizeof(struct in6_pktinfo))
2715 			return (EINVAL);
2716 
2717 		pktinfo = (struct in6_pktinfo *)buf;
2718 
2719 		/*
2720 		 * An application can clear any sticky IPV6_PKTINFO option by
2721 		 * doing a "regular" setsockopt with ipi6_addr being
2722 		 * in6addr_any and ipi6_ifindex being zero.
2723 		 * [RFC 3542, Section 6]
2724 		 */
2725 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2726 		    pktinfo->ipi6_ifindex == 0 &&
2727 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2728 			ip6_clearpktopts(opt, optname);
2729 			break;
2730 		}
2731 
2732 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2733 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2734 			return (EINVAL);
2735 		}
2736 
2737 		/* validate the interface index if specified. */
2738 		if (pktinfo->ipi6_ifindex >= if_indexlim) {
2739 			 return (ENXIO);
2740 		}
2741 		if (pktinfo->ipi6_ifindex) {
2742 			ifp = ifindex2ifnet[pktinfo->ipi6_ifindex];
2743 			if (ifp == NULL)
2744 				return (ENXIO);
2745 		}
2746 
2747 		/*
2748 		 * We store the address anyway, and let in6_selectsrc()
2749 		 * validate the specified address.  This is because ipi6_addr
2750 		 * may not have enough information about its scope zone, and
2751 		 * we may need additional information (such as outgoing
2752 		 * interface or the scope zone of a destination address) to
2753 		 * disambiguate the scope.
2754 		 * XXX: the delay of the validation may confuse the
2755 		 * application when it is used as a sticky option.
2756 		 */
2757 		if (opt->ip6po_pktinfo == NULL) {
2758 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2759 			    M_IP6OPT, M_NOWAIT);
2760 			if (opt->ip6po_pktinfo == NULL)
2761 				return (ENOBUFS);
2762 		}
2763 		memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo));
2764 		break;
2765 	}
2766 
2767 #ifdef RFC2292
2768 	case IPV6_2292HOPLIMIT:
2769 #endif
2770 	case IPV6_HOPLIMIT:
2771 	{
2772 		int *hlimp;
2773 
2774 		/*
2775 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2776 		 * to simplify the ordering among hoplimit options.
2777 		 */
2778 		if (optname == IPV6_HOPLIMIT && sticky)
2779 			return (ENOPROTOOPT);
2780 
2781 		if (len != sizeof(int))
2782 			return (EINVAL);
2783 		hlimp = (int *)buf;
2784 		if (*hlimp < -1 || *hlimp > 255)
2785 			return (EINVAL);
2786 
2787 		opt->ip6po_hlim = *hlimp;
2788 		break;
2789 	}
2790 
2791 	case IPV6_OTCLASS:
2792 		if (len != sizeof(u_int8_t))
2793 			return (EINVAL);
2794 
2795 		opt->ip6po_tclass = *(u_int8_t *)buf;
2796 		break;
2797 
2798 	case IPV6_TCLASS:
2799 	{
2800 		int tclass;
2801 
2802 		if (len != sizeof(int))
2803 			return (EINVAL);
2804 		tclass = *(int *)buf;
2805 		if (tclass < -1 || tclass > 255)
2806 			return (EINVAL);
2807 
2808 		opt->ip6po_tclass = tclass;
2809 		break;
2810 	}
2811 
2812 #ifdef RFC2292
2813 	case IPV6_2292NEXTHOP:
2814 #endif
2815 	case IPV6_NEXTHOP:
2816 		error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
2817 		    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2818 		if (error)
2819 			return (error);
2820 
2821 		if (len == 0) {	/* just remove the option */
2822 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2823 			break;
2824 		}
2825 
2826 		/* check if cmsg_len is large enough for sa_len */
2827 		if (len < sizeof(struct sockaddr) || len < *buf)
2828 			return (EINVAL);
2829 
2830 		switch (((struct sockaddr *)buf)->sa_family) {
2831 		case AF_INET6:
2832 		{
2833 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2834 
2835 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2836 				return (EINVAL);
2837 
2838 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2839 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2840 				return (EINVAL);
2841 			}
2842 			if ((error = sa6_embedscope(sa6, ip6_use_defzone))
2843 			    != 0) {
2844 				return (error);
2845 			}
2846 			break;
2847 		}
2848 		case AF_LINK:	/* eventually be supported? */
2849 		default:
2850 			return (EAFNOSUPPORT);
2851 		}
2852 
2853 		/* turn off the previous option, then set the new option. */
2854 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2855 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2856 		if (opt->ip6po_nexthop == NULL)
2857 			return (ENOBUFS);
2858 		memcpy(opt->ip6po_nexthop, buf, *buf);
2859 		break;
2860 
2861 #ifdef RFC2292
2862 	case IPV6_2292HOPOPTS:
2863 #endif
2864 	case IPV6_HOPOPTS:
2865 	{
2866 		struct ip6_hbh *hbh;
2867 		int hbhlen;
2868 
2869 		/*
2870 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2871 		 * options, since per-option restriction has too much
2872 		 * overhead.
2873 		 */
2874 		error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
2875 		    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2876 		if (error)
2877 			return (error);
2878 
2879 		if (len == 0) {
2880 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2881 			break;	/* just remove the option */
2882 		}
2883 
2884 		/* message length validation */
2885 		if (len < sizeof(struct ip6_hbh))
2886 			return (EINVAL);
2887 		hbh = (struct ip6_hbh *)buf;
2888 		hbhlen = (hbh->ip6h_len + 1) << 3;
2889 		if (len != hbhlen)
2890 			return (EINVAL);
2891 
2892 		/* turn off the previous option, then set the new option. */
2893 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2894 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2895 		if (opt->ip6po_hbh == NULL)
2896 			return (ENOBUFS);
2897 		memcpy(opt->ip6po_hbh, hbh, hbhlen);
2898 
2899 		break;
2900 	}
2901 
2902 #ifdef RFC2292
2903 	case IPV6_2292DSTOPTS:
2904 #endif
2905 	case IPV6_DSTOPTS:
2906 	case IPV6_RTHDRDSTOPTS:
2907 	{
2908 		struct ip6_dest *dest, **newdest = NULL;
2909 		int destlen;
2910 
2911 		/* XXX: see the comment for IPV6_HOPOPTS */
2912 		error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
2913 		    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2914 		if (error)
2915 			return (error);
2916 
2917 		if (len == 0) {
2918 			ip6_clearpktopts(opt, optname);
2919 			break;	/* just remove the option */
2920 		}
2921 
2922 		/* message length validation */
2923 		if (len < sizeof(struct ip6_dest))
2924 			return (EINVAL);
2925 		dest = (struct ip6_dest *)buf;
2926 		destlen = (dest->ip6d_len + 1) << 3;
2927 		if (len != destlen)
2928 			return (EINVAL);
2929 		/*
2930 		 * Determine the position that the destination options header
2931 		 * should be inserted; before or after the routing header.
2932 		 */
2933 		switch (optname) {
2934 		case IPV6_2292DSTOPTS:
2935 			/*
2936 			 * The old advanced API is ambiguous on this point.
2937 			 * Our approach is to determine the position based
2938 			 * according to the existence of a routing header.
2939 			 * Note, however, that this depends on the order of the
2940 			 * extension headers in the ancillary data; the 1st
2941 			 * part of the destination options header must appear
2942 			 * before the routing header in the ancillary data,
2943 			 * too.
2944 			 * RFC3542 solved the ambiguity by introducing
2945 			 * separate ancillary data or option types.
2946 			 */
2947 			if (opt->ip6po_rthdr == NULL)
2948 				newdest = &opt->ip6po_dest1;
2949 			else
2950 				newdest = &opt->ip6po_dest2;
2951 			break;
2952 		case IPV6_RTHDRDSTOPTS:
2953 			newdest = &opt->ip6po_dest1;
2954 			break;
2955 		case IPV6_DSTOPTS:
2956 			newdest = &opt->ip6po_dest2;
2957 			break;
2958 		}
2959 
2960 		/* turn off the previous option, then set the new option. */
2961 		ip6_clearpktopts(opt, optname);
2962 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2963 		if (*newdest == NULL)
2964 			return (ENOBUFS);
2965 		memcpy(*newdest, dest, destlen);
2966 
2967 		break;
2968 	}
2969 
2970 #ifdef RFC2292
2971 	case IPV6_2292RTHDR:
2972 #endif
2973 	case IPV6_RTHDR:
2974 	{
2975 		struct ip6_rthdr *rth;
2976 		int rthlen;
2977 
2978 		if (len == 0) {
2979 			ip6_clearpktopts(opt, IPV6_RTHDR);
2980 			break;	/* just remove the option */
2981 		}
2982 
2983 		/* message length validation */
2984 		if (len < sizeof(struct ip6_rthdr))
2985 			return (EINVAL);
2986 		rth = (struct ip6_rthdr *)buf;
2987 		rthlen = (rth->ip6r_len + 1) << 3;
2988 		if (len != rthlen)
2989 			return (EINVAL);
2990 		switch (rth->ip6r_type) {
2991 		case IPV6_RTHDR_TYPE_0:
2992 			if (rth->ip6r_len == 0)	/* must contain one addr */
2993 				return (EINVAL);
2994 			if (rth->ip6r_len % 2) /* length must be even */
2995 				return (EINVAL);
2996 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2997 				return (EINVAL);
2998 			break;
2999 		default:
3000 			return (EINVAL);	/* not supported */
3001 		}
3002 		/* turn off the previous option */
3003 		ip6_clearpktopts(opt, IPV6_RTHDR);
3004 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3005 		if (opt->ip6po_rthdr == NULL)
3006 			return (ENOBUFS);
3007 		memcpy(opt->ip6po_rthdr, rth, rthlen);
3008 		break;
3009 	}
3010 
3011 	case IPV6_USE_MIN_MTU:
3012 		if (len != sizeof(int))
3013 			return (EINVAL);
3014 		minmtupolicy = *(int *)buf;
3015 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3016 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3017 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3018 			return (EINVAL);
3019 		}
3020 		opt->ip6po_minmtu = minmtupolicy;
3021 		break;
3022 
3023 	case IPV6_DONTFRAG:
3024 		if (len != sizeof(int))
3025 			return (EINVAL);
3026 
3027 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3028 			/*
3029 			 * we ignore this option for TCP sockets.
3030 			 * (RFC3542 leaves this case unspecified.)
3031 			 */
3032 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3033 		} else
3034 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3035 		break;
3036 
3037 	default:
3038 		return (ENOPROTOOPT);
3039 	} /* end of switch */
3040 
3041 	return (0);
3042 }
3043 
3044 /*
3045  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3046  * packet to the input queue of a specified interface.  Note that this
3047  * calls the output routine of the loopback "driver", but with an interface
3048  * pointer that might NOT be lo0ifp -- easier than replicating that code here.
3049  */
3050 void
3051 ip6_mloopback(struct ifnet *ifp, struct mbuf *m,
3052 	const struct sockaddr_in6 *dst)
3053 {
3054 	struct mbuf *copym;
3055 	struct ip6_hdr *ip6;
3056 
3057 	copym = m_copy(m, 0, M_COPYALL);
3058 	if (copym == NULL)
3059 		return;
3060 
3061 	/*
3062 	 * Make sure to deep-copy IPv6 header portion in case the data
3063 	 * is in an mbuf cluster, so that we can safely override the IPv6
3064 	 * header portion later.
3065 	 */
3066 	if ((copym->m_flags & M_EXT) != 0 ||
3067 	    copym->m_len < sizeof(struct ip6_hdr)) {
3068 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3069 		if (copym == NULL)
3070 			return;
3071 	}
3072 
3073 #ifdef DIAGNOSTIC
3074 	if (copym->m_len < sizeof(*ip6)) {
3075 		m_freem(copym);
3076 		return;
3077 	}
3078 #endif
3079 
3080 	ip6 = mtod(copym, struct ip6_hdr *);
3081 	/*
3082 	 * clear embedded scope identifiers if necessary.
3083 	 * in6_clearscope will touch the addresses only when necessary.
3084 	 */
3085 	in6_clearscope(&ip6->ip6_src);
3086 	in6_clearscope(&ip6->ip6_dst);
3087 
3088 	(void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL);
3089 }
3090 
3091 /*
3092  * Chop IPv6 header off from the payload.
3093  */
3094 static int
3095 ip6_splithdr(struct mbuf *m,  struct ip6_exthdrs *exthdrs)
3096 {
3097 	struct mbuf *mh;
3098 	struct ip6_hdr *ip6;
3099 
3100 	ip6 = mtod(m, struct ip6_hdr *);
3101 	if (m->m_len > sizeof(*ip6)) {
3102 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3103 		if (mh == 0) {
3104 			m_freem(m);
3105 			return ENOBUFS;
3106 		}
3107 		M_MOVE_PKTHDR(mh, m);
3108 		MH_ALIGN(mh, sizeof(*ip6));
3109 		m->m_len -= sizeof(*ip6);
3110 		m->m_data += sizeof(*ip6);
3111 		mh->m_next = m;
3112 		m = mh;
3113 		m->m_len = sizeof(*ip6);
3114 		bcopy((void *)ip6, mtod(m, void *), sizeof(*ip6));
3115 	}
3116 	exthdrs->ip6e_ip6 = m;
3117 	return 0;
3118 }
3119 
3120 /*
3121  * Compute IPv6 extension header length.
3122  */
3123 int
3124 ip6_optlen(struct in6pcb *in6p)
3125 {
3126 	int len;
3127 
3128 	if (!in6p->in6p_outputopts)
3129 		return 0;
3130 
3131 	len = 0;
3132 #define elen(x) \
3133     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3134 
3135 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3136 	len += elen(in6p->in6p_outputopts->ip6po_dest1);
3137 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3138 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3139 	return len;
3140 #undef elen
3141 }
3142