xref: /netbsd-src/sys/netinet6/ip6_output.c (revision a4ddc2c8fb9af816efe3b1c375a5530aef0e89e9)
1 /*	$NetBSD: ip6_output.c,v 1.152 2013/03/18 19:31:39 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 1.152 2013/03/18 19:31:39 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 			/*
909 			 * ip6f must be valid if error is 0.  But how
910 			 * can a compiler be expected to infer this?
911 			 */
912 			ip6f = NULL;
913 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
914 			if (error) {
915 				IP6_STATINC(IP6_STAT_ODROPPED);
916 				goto sendorfree;
917 			}
918 			ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7));
919 			if (off + len >= tlen)
920 				len = tlen - off;
921 			else
922 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
923 			mhip6->ip6_plen = htons((u_int16_t)(len + hlen +
924 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
925 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
926 				error = ENOBUFS;
927 				IP6_STATINC(IP6_STAT_ODROPPED);
928 				goto sendorfree;
929 			}
930 			for (mlast = m; mlast->m_next; mlast = mlast->m_next)
931 				;
932 			mlast->m_next = m_frgpart;
933 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
934 			m->m_pkthdr.rcvif = NULL;
935 			ip6f->ip6f_reserved = 0;
936 			ip6f->ip6f_ident = id;
937 			ip6f->ip6f_nxt = nextproto;
938 			IP6_STATINC(IP6_STAT_OFRAGMENTS);
939 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
940 		}
941 
942 		in6_ifstat_inc(ifp, ifs6_out_fragok);
943 	}
944 
945 	/*
946 	 * Remove leading garbages.
947 	 */
948 sendorfree:
949 	m = m0->m_nextpkt;
950 	m0->m_nextpkt = 0;
951 	m_freem(m0);
952 	for (m0 = m; m; m = m0) {
953 		m0 = m->m_nextpkt;
954 		m->m_nextpkt = 0;
955 		if (error == 0) {
956 			struct in6_ifaddr *ia6;
957 			ip6 = mtod(m, struct ip6_hdr *);
958 			ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
959 			if (ia6) {
960 				/*
961 				 * Record statistics for this interface
962 				 * address.
963 				 */
964 				ia6->ia_ifa.ifa_data.ifad_outbytes +=
965 				    m->m_pkthdr.len;
966 			}
967 			KASSERT(dst != NULL);
968 			error = nd6_output(ifp, origifp, m, dst, rt);
969 		} else
970 			m_freem(m);
971 	}
972 
973 	if (error == 0)
974 		IP6_STATINC(IP6_STAT_FRAGMENTED);
975 
976 done:
977 	rtcache_free(&ip6route);
978 
979 #ifdef FAST_IPSEC
980 	if (sp != NULL)
981 		KEY_FREESP(&sp);
982 #endif /* FAST_IPSEC */
983 
984 
985 	return (error);
986 
987 freehdrs:
988 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
989 	m_freem(exthdrs.ip6e_dest1);
990 	m_freem(exthdrs.ip6e_rthdr);
991 	m_freem(exthdrs.ip6e_dest2);
992 	/* FALLTHROUGH */
993 bad:
994 	m_freem(m);
995 	goto done;
996 badscope:
997 	IP6_STATINC(IP6_STAT_BADSCOPE);
998 	in6_ifstat_inc(origifp, ifs6_out_discard);
999 	if (error == 0)
1000 		error = EHOSTUNREACH; /* XXX */
1001 	goto bad;
1002 }
1003 
1004 static int
1005 ip6_copyexthdr(struct mbuf **mp, void *hdr, int hlen)
1006 {
1007 	struct mbuf *m;
1008 
1009 	if (hlen > MCLBYTES)
1010 		return (ENOBUFS); /* XXX */
1011 
1012 	MGET(m, M_DONTWAIT, MT_DATA);
1013 	if (!m)
1014 		return (ENOBUFS);
1015 
1016 	if (hlen > MLEN) {
1017 		MCLGET(m, M_DONTWAIT);
1018 		if ((m->m_flags & M_EXT) == 0) {
1019 			m_free(m);
1020 			return (ENOBUFS);
1021 		}
1022 	}
1023 	m->m_len = hlen;
1024 	if (hdr)
1025 		bcopy(hdr, mtod(m, void *), hlen);
1026 
1027 	*mp = m;
1028 	return (0);
1029 }
1030 
1031 /*
1032  * Process a delayed payload checksum calculation.
1033  */
1034 void
1035 in6_delayed_cksum(struct mbuf *m)
1036 {
1037 	uint16_t csum, offset;
1038 
1039 	KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0);
1040 	KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0);
1041 	KASSERT((m->m_pkthdr.csum_flags
1042 	    & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0);
1043 
1044 	offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data);
1045 	csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset);
1046 	if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) {
1047 		csum = 0xffff;
1048 	}
1049 
1050 	offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data);
1051 	if ((offset + sizeof(csum)) > m->m_len) {
1052 		m_copyback(m, offset, sizeof(csum), &csum);
1053 	} else {
1054 		*(uint16_t *)(mtod(m, char *) + offset) = csum;
1055 	}
1056 }
1057 
1058 /*
1059  * Insert jumbo payload option.
1060  */
1061 static int
1062 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1063 {
1064 	struct mbuf *mopt;
1065 	u_int8_t *optbuf;
1066 	u_int32_t v;
1067 
1068 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1069 
1070 	/*
1071 	 * If there is no hop-by-hop options header, allocate new one.
1072 	 * If there is one but it doesn't have enough space to store the
1073 	 * jumbo payload option, allocate a cluster to store the whole options.
1074 	 * Otherwise, use it to store the options.
1075 	 */
1076 	if (exthdrs->ip6e_hbh == 0) {
1077 		MGET(mopt, M_DONTWAIT, MT_DATA);
1078 		if (mopt == 0)
1079 			return (ENOBUFS);
1080 		mopt->m_len = JUMBOOPTLEN;
1081 		optbuf = mtod(mopt, u_int8_t *);
1082 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1083 		exthdrs->ip6e_hbh = mopt;
1084 	} else {
1085 		struct ip6_hbh *hbh;
1086 
1087 		mopt = exthdrs->ip6e_hbh;
1088 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1089 			/*
1090 			 * XXX assumption:
1091 			 * - exthdrs->ip6e_hbh is not referenced from places
1092 			 *   other than exthdrs.
1093 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1094 			 */
1095 			int oldoptlen = mopt->m_len;
1096 			struct mbuf *n;
1097 
1098 			/*
1099 			 * XXX: give up if the whole (new) hbh header does
1100 			 * not fit even in an mbuf cluster.
1101 			 */
1102 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1103 				return (ENOBUFS);
1104 
1105 			/*
1106 			 * As a consequence, we must always prepare a cluster
1107 			 * at this point.
1108 			 */
1109 			MGET(n, M_DONTWAIT, MT_DATA);
1110 			if (n) {
1111 				MCLGET(n, M_DONTWAIT);
1112 				if ((n->m_flags & M_EXT) == 0) {
1113 					m_freem(n);
1114 					n = NULL;
1115 				}
1116 			}
1117 			if (!n)
1118 				return (ENOBUFS);
1119 			n->m_len = oldoptlen + JUMBOOPTLEN;
1120 			bcopy(mtod(mopt, void *), mtod(n, void *),
1121 			    oldoptlen);
1122 			optbuf = mtod(n, u_int8_t *) + oldoptlen;
1123 			m_freem(mopt);
1124 			mopt = exthdrs->ip6e_hbh = n;
1125 		} else {
1126 			optbuf = mtod(mopt, u_int8_t *) + mopt->m_len;
1127 			mopt->m_len += JUMBOOPTLEN;
1128 		}
1129 		optbuf[0] = IP6OPT_PADN;
1130 		optbuf[1] = 0;
1131 
1132 		/*
1133 		 * Adjust the header length according to the pad and
1134 		 * the jumbo payload option.
1135 		 */
1136 		hbh = mtod(mopt, struct ip6_hbh *);
1137 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1138 	}
1139 
1140 	/* fill in the option. */
1141 	optbuf[2] = IP6OPT_JUMBO;
1142 	optbuf[3] = 4;
1143 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1144 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1145 
1146 	/* finally, adjust the packet header length */
1147 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1148 
1149 	return (0);
1150 #undef JUMBOOPTLEN
1151 }
1152 
1153 /*
1154  * Insert fragment header and copy unfragmentable header portions.
1155  *
1156  * *frghdrp will not be read, and it is guaranteed that either an
1157  * error is returned or that *frghdrp will point to space allocated
1158  * for the fragment header.
1159  */
1160 static int
1161 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1162 	struct ip6_frag **frghdrp)
1163 {
1164 	struct mbuf *n, *mlast;
1165 
1166 	if (hlen > sizeof(struct ip6_hdr)) {
1167 		n = m_copym(m0, sizeof(struct ip6_hdr),
1168 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1169 		if (n == 0)
1170 			return (ENOBUFS);
1171 		m->m_next = n;
1172 	} else
1173 		n = m;
1174 
1175 	/* Search for the last mbuf of unfragmentable part. */
1176 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1177 		;
1178 
1179 	if ((mlast->m_flags & M_EXT) == 0 &&
1180 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1181 		/* use the trailing space of the last mbuf for the fragment hdr */
1182 		*frghdrp = (struct ip6_frag *)(mtod(mlast, char *) +
1183 		    mlast->m_len);
1184 		mlast->m_len += sizeof(struct ip6_frag);
1185 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1186 	} else {
1187 		/* allocate a new mbuf for the fragment header */
1188 		struct mbuf *mfrg;
1189 
1190 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1191 		if (mfrg == 0)
1192 			return (ENOBUFS);
1193 		mfrg->m_len = sizeof(struct ip6_frag);
1194 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1195 		mlast->m_next = mfrg;
1196 	}
1197 
1198 	return (0);
1199 }
1200 
1201 static int
1202 ip6_getpmtu(struct route *ro_pmtu, struct route *ro, struct ifnet *ifp,
1203     const struct in6_addr *dst, u_long *mtup, int *alwaysfragp)
1204 {
1205 	struct rtentry *rt;
1206 	u_int32_t mtu = 0;
1207 	int alwaysfrag = 0;
1208 	int error = 0;
1209 
1210 	if (ro_pmtu != ro) {
1211 		union {
1212 			struct sockaddr		dst;
1213 			struct sockaddr_in6	dst6;
1214 		} u;
1215 
1216 		/* The first hop and the final destination may differ. */
1217 		sockaddr_in6_init(&u.dst6, dst, 0, 0, 0);
1218 		rt = rtcache_lookup(ro_pmtu, &u.dst);
1219 	} else
1220 		rt = rtcache_validate(ro_pmtu);
1221 	if (rt != NULL) {
1222 		u_int32_t ifmtu;
1223 
1224 		if (ifp == NULL)
1225 			ifp = rt->rt_ifp;
1226 		ifmtu = IN6_LINKMTU(ifp);
1227 		mtu = rt->rt_rmx.rmx_mtu;
1228 		if (mtu == 0)
1229 			mtu = ifmtu;
1230 		else if (mtu < IPV6_MMTU) {
1231 			/*
1232 			 * RFC2460 section 5, last paragraph:
1233 			 * if we record ICMPv6 too big message with
1234 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1235 			 * or smaller, with fragment header attached.
1236 			 * (fragment header is needed regardless from the
1237 			 * packet size, for translators to identify packets)
1238 			 */
1239 			alwaysfrag = 1;
1240 			mtu = IPV6_MMTU;
1241 		} else if (mtu > ifmtu) {
1242 			/*
1243 			 * The MTU on the route is larger than the MTU on
1244 			 * the interface!  This shouldn't happen, unless the
1245 			 * MTU of the interface has been changed after the
1246 			 * interface was brought up.  Change the MTU in the
1247 			 * route to match the interface MTU (as long as the
1248 			 * field isn't locked).
1249 			 */
1250 			mtu = ifmtu;
1251 			if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
1252 				rt->rt_rmx.rmx_mtu = mtu;
1253 		}
1254 	} else if (ifp) {
1255 		mtu = IN6_LINKMTU(ifp);
1256 	} else
1257 		error = EHOSTUNREACH; /* XXX */
1258 
1259 	*mtup = mtu;
1260 	if (alwaysfragp)
1261 		*alwaysfragp = alwaysfrag;
1262 	return (error);
1263 }
1264 
1265 /*
1266  * IP6 socket option processing.
1267  */
1268 int
1269 ip6_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1270 {
1271 	int optdatalen, uproto;
1272 	void *optdata;
1273 	struct in6pcb *in6p = sotoin6pcb(so);
1274 	int error, optval;
1275 	int level, optname;
1276 
1277 	KASSERT(sopt != NULL);
1278 
1279 	level = sopt->sopt_level;
1280 	optname = sopt->sopt_name;
1281 
1282 	error = optval = 0;
1283 	uproto = (int)so->so_proto->pr_protocol;
1284 
1285 	if (level != IPPROTO_IPV6) {
1286 		return ENOPROTOOPT;
1287 	}
1288 	switch (op) {
1289 	case PRCO_SETOPT:
1290 		switch (optname) {
1291 #ifdef RFC2292
1292 		case IPV6_2292PKTOPTIONS:
1293 			error = ip6_pcbopts(&in6p->in6p_outputopts, so, sopt);
1294 			break;
1295 #endif
1296 
1297 		/*
1298 		 * Use of some Hop-by-Hop options or some
1299 		 * Destination options, might require special
1300 		 * privilege.  That is, normal applications
1301 		 * (without special privilege) might be forbidden
1302 		 * from setting certain options in outgoing packets,
1303 		 * and might never see certain options in received
1304 		 * packets. [RFC 2292 Section 6]
1305 		 * KAME specific note:
1306 		 *  KAME prevents non-privileged users from sending or
1307 		 *  receiving ANY hbh/dst options in order to avoid
1308 		 *  overhead of parsing options in the kernel.
1309 		 */
1310 		case IPV6_RECVHOPOPTS:
1311 		case IPV6_RECVDSTOPTS:
1312 		case IPV6_RECVRTHDRDSTOPTS:
1313 			error = kauth_authorize_network(kauth_cred_get(),
1314 			    KAUTH_NETWORK_IPV6, KAUTH_REQ_NETWORK_IPV6_HOPBYHOP,
1315 			    NULL, NULL, NULL);
1316 			if (error)
1317 				break;
1318 			/* FALLTHROUGH */
1319 		case IPV6_UNICAST_HOPS:
1320 		case IPV6_HOPLIMIT:
1321 		case IPV6_FAITH:
1322 
1323 		case IPV6_RECVPKTINFO:
1324 		case IPV6_RECVHOPLIMIT:
1325 		case IPV6_RECVRTHDR:
1326 		case IPV6_RECVPATHMTU:
1327 		case IPV6_RECVTCLASS:
1328 		case IPV6_V6ONLY:
1329 			error = sockopt_getint(sopt, &optval);
1330 			if (error)
1331 				break;
1332 			switch (optname) {
1333 			case IPV6_UNICAST_HOPS:
1334 				if (optval < -1 || optval >= 256)
1335 					error = EINVAL;
1336 				else {
1337 					/* -1 = kernel default */
1338 					in6p->in6p_hops = optval;
1339 				}
1340 				break;
1341 #define OPTSET(bit) \
1342 do { \
1343 if (optval) \
1344 	in6p->in6p_flags |= (bit); \
1345 else \
1346 	in6p->in6p_flags &= ~(bit); \
1347 } while (/*CONSTCOND*/ 0)
1348 
1349 #ifdef RFC2292
1350 #define OPTSET2292(bit) 			\
1351 do { 						\
1352 in6p->in6p_flags |= IN6P_RFC2292; 	\
1353 if (optval) 				\
1354 	in6p->in6p_flags |= (bit); 	\
1355 else 					\
1356 	in6p->in6p_flags &= ~(bit); 	\
1357 } while (/*CONSTCOND*/ 0)
1358 #endif
1359 
1360 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1361 
1362 			case IPV6_RECVPKTINFO:
1363 #ifdef RFC2292
1364 				/* cannot mix with RFC2292 */
1365 				if (OPTBIT(IN6P_RFC2292)) {
1366 					error = EINVAL;
1367 					break;
1368 				}
1369 #endif
1370 				OPTSET(IN6P_PKTINFO);
1371 				break;
1372 
1373 			case IPV6_HOPLIMIT:
1374 			{
1375 				struct ip6_pktopts **optp;
1376 
1377 #ifdef RFC2292
1378 				/* cannot mix with RFC2292 */
1379 				if (OPTBIT(IN6P_RFC2292)) {
1380 					error = EINVAL;
1381 					break;
1382 				}
1383 #endif
1384 				optp = &in6p->in6p_outputopts;
1385 				error = ip6_pcbopt(IPV6_HOPLIMIT,
1386 						   (u_char *)&optval,
1387 						   sizeof(optval),
1388 						   optp,
1389 						   kauth_cred_get(), uproto);
1390 				break;
1391 			}
1392 
1393 			case IPV6_RECVHOPLIMIT:
1394 #ifdef RFC2292
1395 				/* cannot mix with RFC2292 */
1396 				if (OPTBIT(IN6P_RFC2292)) {
1397 					error = EINVAL;
1398 					break;
1399 				}
1400 #endif
1401 				OPTSET(IN6P_HOPLIMIT);
1402 				break;
1403 
1404 			case IPV6_RECVHOPOPTS:
1405 #ifdef RFC2292
1406 				/* cannot mix with RFC2292 */
1407 				if (OPTBIT(IN6P_RFC2292)) {
1408 					error = EINVAL;
1409 					break;
1410 				}
1411 #endif
1412 				OPTSET(IN6P_HOPOPTS);
1413 				break;
1414 
1415 			case IPV6_RECVDSTOPTS:
1416 #ifdef RFC2292
1417 				/* cannot mix with RFC2292 */
1418 				if (OPTBIT(IN6P_RFC2292)) {
1419 					error = EINVAL;
1420 					break;
1421 				}
1422 #endif
1423 				OPTSET(IN6P_DSTOPTS);
1424 				break;
1425 
1426 			case IPV6_RECVRTHDRDSTOPTS:
1427 #ifdef RFC2292
1428 				/* cannot mix with RFC2292 */
1429 				if (OPTBIT(IN6P_RFC2292)) {
1430 					error = EINVAL;
1431 					break;
1432 				}
1433 #endif
1434 				OPTSET(IN6P_RTHDRDSTOPTS);
1435 				break;
1436 
1437 			case IPV6_RECVRTHDR:
1438 #ifdef RFC2292
1439 				/* cannot mix with RFC2292 */
1440 				if (OPTBIT(IN6P_RFC2292)) {
1441 					error = EINVAL;
1442 					break;
1443 				}
1444 #endif
1445 				OPTSET(IN6P_RTHDR);
1446 				break;
1447 
1448 			case IPV6_FAITH:
1449 				OPTSET(IN6P_FAITH);
1450 				break;
1451 
1452 			case IPV6_RECVPATHMTU:
1453 				/*
1454 				 * We ignore this option for TCP
1455 				 * sockets.
1456 				 * (RFC3542 leaves this case
1457 				 * unspecified.)
1458 				 */
1459 				if (uproto != IPPROTO_TCP)
1460 					OPTSET(IN6P_MTU);
1461 				break;
1462 
1463 			case IPV6_V6ONLY:
1464 				/*
1465 				 * make setsockopt(IPV6_V6ONLY)
1466 				 * available only prior to bind(2).
1467 				 * see ipng mailing list, Jun 22 2001.
1468 				 */
1469 				if (in6p->in6p_lport ||
1470 				    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1471 					error = EINVAL;
1472 					break;
1473 				}
1474 #ifdef INET6_BINDV6ONLY
1475 				if (!optval)
1476 					error = EINVAL;
1477 #else
1478 				OPTSET(IN6P_IPV6_V6ONLY);
1479 #endif
1480 				break;
1481 			case IPV6_RECVTCLASS:
1482 #ifdef RFC2292
1483 				/* cannot mix with RFC2292 XXX */
1484 				if (OPTBIT(IN6P_RFC2292)) {
1485 					error = EINVAL;
1486 					break;
1487 				}
1488 #endif
1489 				OPTSET(IN6P_TCLASS);
1490 				break;
1491 
1492 			}
1493 			break;
1494 
1495 		case IPV6_OTCLASS:
1496 		{
1497 			struct ip6_pktopts **optp;
1498 			u_int8_t tclass;
1499 
1500 			error = sockopt_get(sopt, &tclass, sizeof(tclass));
1501 			if (error)
1502 				break;
1503 			optp = &in6p->in6p_outputopts;
1504 			error = ip6_pcbopt(optname,
1505 					   (u_char *)&tclass,
1506 					   sizeof(tclass),
1507 					   optp,
1508 					   kauth_cred_get(), uproto);
1509 			break;
1510 		}
1511 
1512 		case IPV6_TCLASS:
1513 		case IPV6_DONTFRAG:
1514 		case IPV6_USE_MIN_MTU:
1515 			error = sockopt_getint(sopt, &optval);
1516 			if (error)
1517 				break;
1518 			{
1519 				struct ip6_pktopts **optp;
1520 				optp = &in6p->in6p_outputopts;
1521 				error = ip6_pcbopt(optname,
1522 						   (u_char *)&optval,
1523 						   sizeof(optval),
1524 						   optp,
1525 						   kauth_cred_get(), uproto);
1526 				break;
1527 			}
1528 
1529 #ifdef RFC2292
1530 		case IPV6_2292PKTINFO:
1531 		case IPV6_2292HOPLIMIT:
1532 		case IPV6_2292HOPOPTS:
1533 		case IPV6_2292DSTOPTS:
1534 		case IPV6_2292RTHDR:
1535 			/* RFC 2292 */
1536 			error = sockopt_getint(sopt, &optval);
1537 			if (error)
1538 				break;
1539 
1540 			switch (optname) {
1541 			case IPV6_2292PKTINFO:
1542 				OPTSET2292(IN6P_PKTINFO);
1543 				break;
1544 			case IPV6_2292HOPLIMIT:
1545 				OPTSET2292(IN6P_HOPLIMIT);
1546 				break;
1547 			case IPV6_2292HOPOPTS:
1548 				/*
1549 				 * Check super-user privilege.
1550 				 * See comments for IPV6_RECVHOPOPTS.
1551 				 */
1552 				error =
1553 				    kauth_authorize_network(kauth_cred_get(),
1554 				    KAUTH_NETWORK_IPV6,
1555 				    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL,
1556 				    NULL, NULL);
1557 				if (error)
1558 					return (error);
1559 				OPTSET2292(IN6P_HOPOPTS);
1560 				break;
1561 			case IPV6_2292DSTOPTS:
1562 				error =
1563 				    kauth_authorize_network(kauth_cred_get(),
1564 				    KAUTH_NETWORK_IPV6,
1565 				    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL,
1566 				    NULL, NULL);
1567 				if (error)
1568 					return (error);
1569 				OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1570 				break;
1571 			case IPV6_2292RTHDR:
1572 				OPTSET2292(IN6P_RTHDR);
1573 				break;
1574 			}
1575 			break;
1576 #endif
1577 		case IPV6_PKTINFO:
1578 		case IPV6_HOPOPTS:
1579 		case IPV6_RTHDR:
1580 		case IPV6_DSTOPTS:
1581 		case IPV6_RTHDRDSTOPTS:
1582 		case IPV6_NEXTHOP: {
1583 			/* new advanced API (RFC3542) */
1584 			void *optbuf;
1585 			int optbuflen;
1586 			struct ip6_pktopts **optp;
1587 
1588 #ifdef RFC2292
1589 			/* cannot mix with RFC2292 */
1590 			if (OPTBIT(IN6P_RFC2292)) {
1591 				error = EINVAL;
1592 				break;
1593 			}
1594 #endif
1595 
1596 			optbuflen = sopt->sopt_size;
1597 			optbuf = malloc(optbuflen, M_IP6OPT, M_NOWAIT);
1598 			if (optbuf == NULL) {
1599 				error = ENOBUFS;
1600 				break;
1601 			}
1602 
1603 			sockopt_get(sopt, optbuf, optbuflen);
1604 			optp = &in6p->in6p_outputopts;
1605 			error = ip6_pcbopt(optname, optbuf, optbuflen,
1606 			    optp, kauth_cred_get(), uproto);
1607 			break;
1608 			}
1609 #undef OPTSET
1610 
1611 		case IPV6_MULTICAST_IF:
1612 		case IPV6_MULTICAST_HOPS:
1613 		case IPV6_MULTICAST_LOOP:
1614 		case IPV6_JOIN_GROUP:
1615 		case IPV6_LEAVE_GROUP:
1616 			error = ip6_setmoptions(sopt, &in6p->in6p_moptions);
1617 			break;
1618 
1619 		case IPV6_PORTRANGE:
1620 			error = sockopt_getint(sopt, &optval);
1621 			if (error)
1622 				break;
1623 
1624 			switch (optval) {
1625 			case IPV6_PORTRANGE_DEFAULT:
1626 				in6p->in6p_flags &= ~(IN6P_LOWPORT);
1627 				in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1628 				break;
1629 
1630 			case IPV6_PORTRANGE_HIGH:
1631 				in6p->in6p_flags &= ~(IN6P_LOWPORT);
1632 				in6p->in6p_flags |= IN6P_HIGHPORT;
1633 				break;
1634 
1635 			case IPV6_PORTRANGE_LOW:
1636 				in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1637 				in6p->in6p_flags |= IN6P_LOWPORT;
1638 				break;
1639 
1640 			default:
1641 				error = EINVAL;
1642 				break;
1643 			}
1644 			break;
1645 
1646 		case IPV6_PORTALGO:
1647 			error = sockopt_getint(sopt, &optval);
1648 			if (error)
1649 				break;
1650 
1651 			error = portalgo_algo_index_select(
1652 			    (struct inpcb_hdr *)in6p, optval);
1653 			break;
1654 
1655 #if defined(FAST_IPSEC)
1656 		case IPV6_IPSEC_POLICY:
1657 			error = ipsec6_set_policy(in6p, optname,
1658 			    sopt->sopt_data, sopt->sopt_size, kauth_cred_get());
1659 			break;
1660 #endif /* IPSEC */
1661 
1662 		default:
1663 			error = ENOPROTOOPT;
1664 			break;
1665 		}
1666 		break;
1667 
1668 	case PRCO_GETOPT:
1669 		switch (optname) {
1670 #ifdef RFC2292
1671 		case IPV6_2292PKTOPTIONS:
1672 			/*
1673 			 * RFC3542 (effectively) deprecated the
1674 			 * semantics of the 2292-style pktoptions.
1675 			 * Since it was not reliable in nature (i.e.,
1676 			 * applications had to expect the lack of some
1677 			 * information after all), it would make sense
1678 			 * to simplify this part by always returning
1679 			 * empty data.
1680 			 */
1681 			break;
1682 #endif
1683 
1684 		case IPV6_RECVHOPOPTS:
1685 		case IPV6_RECVDSTOPTS:
1686 		case IPV6_RECVRTHDRDSTOPTS:
1687 		case IPV6_UNICAST_HOPS:
1688 		case IPV6_RECVPKTINFO:
1689 		case IPV6_RECVHOPLIMIT:
1690 		case IPV6_RECVRTHDR:
1691 		case IPV6_RECVPATHMTU:
1692 
1693 		case IPV6_FAITH:
1694 		case IPV6_V6ONLY:
1695 		case IPV6_PORTRANGE:
1696 		case IPV6_RECVTCLASS:
1697 			switch (optname) {
1698 
1699 			case IPV6_RECVHOPOPTS:
1700 				optval = OPTBIT(IN6P_HOPOPTS);
1701 				break;
1702 
1703 			case IPV6_RECVDSTOPTS:
1704 				optval = OPTBIT(IN6P_DSTOPTS);
1705 				break;
1706 
1707 			case IPV6_RECVRTHDRDSTOPTS:
1708 				optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1709 				break;
1710 
1711 			case IPV6_UNICAST_HOPS:
1712 				optval = in6p->in6p_hops;
1713 				break;
1714 
1715 			case IPV6_RECVPKTINFO:
1716 				optval = OPTBIT(IN6P_PKTINFO);
1717 				break;
1718 
1719 			case IPV6_RECVHOPLIMIT:
1720 				optval = OPTBIT(IN6P_HOPLIMIT);
1721 				break;
1722 
1723 			case IPV6_RECVRTHDR:
1724 				optval = OPTBIT(IN6P_RTHDR);
1725 				break;
1726 
1727 			case IPV6_RECVPATHMTU:
1728 				optval = OPTBIT(IN6P_MTU);
1729 				break;
1730 
1731 			case IPV6_FAITH:
1732 				optval = OPTBIT(IN6P_FAITH);
1733 				break;
1734 
1735 			case IPV6_V6ONLY:
1736 				optval = OPTBIT(IN6P_IPV6_V6ONLY);
1737 				break;
1738 
1739 			case IPV6_PORTRANGE:
1740 			    {
1741 				int flags;
1742 				flags = in6p->in6p_flags;
1743 				if (flags & IN6P_HIGHPORT)
1744 					optval = IPV6_PORTRANGE_HIGH;
1745 				else if (flags & IN6P_LOWPORT)
1746 					optval = IPV6_PORTRANGE_LOW;
1747 				else
1748 					optval = 0;
1749 				break;
1750 			    }
1751 			case IPV6_RECVTCLASS:
1752 				optval = OPTBIT(IN6P_TCLASS);
1753 				break;
1754 
1755 			}
1756 			if (error)
1757 				break;
1758 			error = sockopt_setint(sopt, optval);
1759 			break;
1760 
1761 		case IPV6_PATHMTU:
1762 		    {
1763 			u_long pmtu = 0;
1764 			struct ip6_mtuinfo mtuinfo;
1765 			struct route *ro = &in6p->in6p_route;
1766 
1767 			if (!(so->so_state & SS_ISCONNECTED))
1768 				return (ENOTCONN);
1769 			/*
1770 			 * XXX: we dot not consider the case of source
1771 			 * routing, or optional information to specify
1772 			 * the outgoing interface.
1773 			 */
1774 			error = ip6_getpmtu(ro, NULL, NULL,
1775 			    &in6p->in6p_faddr, &pmtu, NULL);
1776 			if (error)
1777 				break;
1778 			if (pmtu > IPV6_MAXPACKET)
1779 				pmtu = IPV6_MAXPACKET;
1780 
1781 			memset(&mtuinfo, 0, sizeof(mtuinfo));
1782 			mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1783 			optdata = (void *)&mtuinfo;
1784 			optdatalen = sizeof(mtuinfo);
1785 			if (optdatalen > MCLBYTES)
1786 				return (EMSGSIZE); /* XXX */
1787 			error = sockopt_set(sopt, optdata, optdatalen);
1788 			break;
1789 		    }
1790 
1791 #ifdef RFC2292
1792 		case IPV6_2292PKTINFO:
1793 		case IPV6_2292HOPLIMIT:
1794 		case IPV6_2292HOPOPTS:
1795 		case IPV6_2292RTHDR:
1796 		case IPV6_2292DSTOPTS:
1797 			switch (optname) {
1798 			case IPV6_2292PKTINFO:
1799 				optval = OPTBIT(IN6P_PKTINFO);
1800 				break;
1801 			case IPV6_2292HOPLIMIT:
1802 				optval = OPTBIT(IN6P_HOPLIMIT);
1803 				break;
1804 			case IPV6_2292HOPOPTS:
1805 				optval = OPTBIT(IN6P_HOPOPTS);
1806 				break;
1807 			case IPV6_2292RTHDR:
1808 				optval = OPTBIT(IN6P_RTHDR);
1809 				break;
1810 			case IPV6_2292DSTOPTS:
1811 				optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1812 				break;
1813 			}
1814 			error = sockopt_setint(sopt, optval);
1815 			break;
1816 #endif
1817 		case IPV6_PKTINFO:
1818 		case IPV6_HOPOPTS:
1819 		case IPV6_RTHDR:
1820 		case IPV6_DSTOPTS:
1821 		case IPV6_RTHDRDSTOPTS:
1822 		case IPV6_NEXTHOP:
1823 		case IPV6_OTCLASS:
1824 		case IPV6_TCLASS:
1825 		case IPV6_DONTFRAG:
1826 		case IPV6_USE_MIN_MTU:
1827 			error = ip6_getpcbopt(in6p->in6p_outputopts,
1828 			    optname, sopt);
1829 			break;
1830 
1831 		case IPV6_MULTICAST_IF:
1832 		case IPV6_MULTICAST_HOPS:
1833 		case IPV6_MULTICAST_LOOP:
1834 		case IPV6_JOIN_GROUP:
1835 		case IPV6_LEAVE_GROUP:
1836 			error = ip6_getmoptions(sopt, in6p->in6p_moptions);
1837 			break;
1838 
1839 		case IPV6_PORTALGO:
1840 			optval = ((struct inpcb_hdr *)in6p)->inph_portalgo;
1841 			error = sockopt_setint(sopt, optval);
1842 			break;
1843 
1844 #if defined(FAST_IPSEC)
1845 		case IPV6_IPSEC_POLICY:
1846 		    {
1847 			struct mbuf *m = NULL;
1848 
1849 			/* XXX this will return EINVAL as sopt is empty */
1850 			error = ipsec6_get_policy(in6p, sopt->sopt_data,
1851 			    sopt->sopt_size, &m);
1852 			if (!error)
1853 				error = sockopt_setmbuf(sopt, m);
1854 
1855 			break;
1856 		    }
1857 #endif /* IPSEC */
1858 
1859 		default:
1860 			error = ENOPROTOOPT;
1861 			break;
1862 		}
1863 		break;
1864 	}
1865 	return (error);
1866 }
1867 
1868 int
1869 ip6_raw_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1870 {
1871 	int error = 0, optval;
1872 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
1873 	struct in6pcb *in6p = sotoin6pcb(so);
1874 	int level, optname;
1875 
1876 	KASSERT(sopt != NULL);
1877 
1878 	level = sopt->sopt_level;
1879 	optname = sopt->sopt_name;
1880 
1881 	if (level != IPPROTO_IPV6) {
1882 		return ENOPROTOOPT;
1883 	}
1884 
1885 	switch (optname) {
1886 	case IPV6_CHECKSUM:
1887 		/*
1888 		 * For ICMPv6 sockets, no modification allowed for checksum
1889 		 * offset, permit "no change" values to help existing apps.
1890 		 *
1891 		 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM
1892 		 * for an ICMPv6 socket will fail."  The current
1893 		 * behavior does not meet RFC3542.
1894 		 */
1895 		switch (op) {
1896 		case PRCO_SETOPT:
1897 			error = sockopt_getint(sopt, &optval);
1898 			if (error)
1899 				break;
1900 			if ((optval % 2) != 0) {
1901 				/* the API assumes even offset values */
1902 				error = EINVAL;
1903 			} else if (so->so_proto->pr_protocol ==
1904 			    IPPROTO_ICMPV6) {
1905 				if (optval != icmp6off)
1906 					error = EINVAL;
1907 			} else
1908 				in6p->in6p_cksum = optval;
1909 			break;
1910 
1911 		case PRCO_GETOPT:
1912 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
1913 				optval = icmp6off;
1914 			else
1915 				optval = in6p->in6p_cksum;
1916 
1917 			error = sockopt_setint(sopt, optval);
1918 			break;
1919 
1920 		default:
1921 			error = EINVAL;
1922 			break;
1923 		}
1924 		break;
1925 
1926 	default:
1927 		error = ENOPROTOOPT;
1928 		break;
1929 	}
1930 
1931 	return (error);
1932 }
1933 
1934 #ifdef RFC2292
1935 /*
1936  * Set up IP6 options in pcb for insertion in output packets or
1937  * specifying behavior of outgoing packets.
1938  */
1939 static int
1940 ip6_pcbopts(struct ip6_pktopts **pktopt, struct socket *so,
1941     struct sockopt *sopt)
1942 {
1943 	struct ip6_pktopts *opt = *pktopt;
1944 	struct mbuf *m;
1945 	int error = 0;
1946 
1947 	/* turn off any old options. */
1948 	if (opt) {
1949 #ifdef DIAGNOSTIC
1950 	    if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
1951 		opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
1952 		opt->ip6po_rhinfo.ip6po_rhi_rthdr)
1953 		    printf("ip6_pcbopts: all specified options are cleared.\n");
1954 #endif
1955 		ip6_clearpktopts(opt, -1);
1956 	} else {
1957 		opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
1958 		if (opt == NULL)
1959 			return (ENOBUFS);
1960 	}
1961 	*pktopt = NULL;
1962 
1963 	if (sopt == NULL || sopt->sopt_size == 0) {
1964 		/*
1965 		 * Only turning off any previous options, regardless of
1966 		 * whether the opt is just created or given.
1967 		 */
1968 		free(opt, M_IP6OPT);
1969 		return (0);
1970 	}
1971 
1972 	/*  set options specified by user. */
1973 	m = sockopt_getmbuf(sopt);
1974 	if (m == NULL) {
1975 		free(opt, M_IP6OPT);
1976 		return (ENOBUFS);
1977 	}
1978 
1979 	error = ip6_setpktopts(m, opt, NULL, kauth_cred_get(),
1980 	    so->so_proto->pr_protocol);
1981 	m_freem(m);
1982 	if (error != 0) {
1983 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
1984 		free(opt, M_IP6OPT);
1985 		return (error);
1986 	}
1987 	*pktopt = opt;
1988 	return (0);
1989 }
1990 #endif
1991 
1992 /*
1993  * initialize ip6_pktopts.  beware that there are non-zero default values in
1994  * the struct.
1995  */
1996 void
1997 ip6_initpktopts(struct ip6_pktopts *opt)
1998 {
1999 
2000 	memset(opt, 0, sizeof(*opt));
2001 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2002 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2003 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2004 }
2005 
2006 #define sin6tosa(sin6)	((struct sockaddr *)(sin6)) /* XXX */
2007 static int
2008 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2009     kauth_cred_t cred, int uproto)
2010 {
2011 	struct ip6_pktopts *opt;
2012 
2013 	if (*pktopt == NULL) {
2014 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2015 		    M_NOWAIT);
2016 		if (*pktopt == NULL)
2017 			return (ENOBUFS);
2018 
2019 		ip6_initpktopts(*pktopt);
2020 	}
2021 	opt = *pktopt;
2022 
2023 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2024 }
2025 
2026 static int
2027 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2028 {
2029 	void *optdata = NULL;
2030 	int optdatalen = 0;
2031 	struct ip6_ext *ip6e;
2032 	int error = 0;
2033 	struct in6_pktinfo null_pktinfo;
2034 	int deftclass = 0, on;
2035 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2036 
2037 	switch (optname) {
2038 	case IPV6_PKTINFO:
2039 		if (pktopt && pktopt->ip6po_pktinfo)
2040 			optdata = (void *)pktopt->ip6po_pktinfo;
2041 		else {
2042 			/* XXX: we don't have to do this every time... */
2043 			memset(&null_pktinfo, 0, sizeof(null_pktinfo));
2044 			optdata = (void *)&null_pktinfo;
2045 		}
2046 		optdatalen = sizeof(struct in6_pktinfo);
2047 		break;
2048 	case IPV6_OTCLASS:
2049 		/* XXX */
2050 		return (EINVAL);
2051 	case IPV6_TCLASS:
2052 		if (pktopt && pktopt->ip6po_tclass >= 0)
2053 			optdata = (void *)&pktopt->ip6po_tclass;
2054 		else
2055 			optdata = (void *)&deftclass;
2056 		optdatalen = sizeof(int);
2057 		break;
2058 	case IPV6_HOPOPTS:
2059 		if (pktopt && pktopt->ip6po_hbh) {
2060 			optdata = (void *)pktopt->ip6po_hbh;
2061 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2062 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2063 		}
2064 		break;
2065 	case IPV6_RTHDR:
2066 		if (pktopt && pktopt->ip6po_rthdr) {
2067 			optdata = (void *)pktopt->ip6po_rthdr;
2068 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2069 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2070 		}
2071 		break;
2072 	case IPV6_RTHDRDSTOPTS:
2073 		if (pktopt && pktopt->ip6po_dest1) {
2074 			optdata = (void *)pktopt->ip6po_dest1;
2075 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2076 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2077 		}
2078 		break;
2079 	case IPV6_DSTOPTS:
2080 		if (pktopt && pktopt->ip6po_dest2) {
2081 			optdata = (void *)pktopt->ip6po_dest2;
2082 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2083 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2084 		}
2085 		break;
2086 	case IPV6_NEXTHOP:
2087 		if (pktopt && pktopt->ip6po_nexthop) {
2088 			optdata = (void *)pktopt->ip6po_nexthop;
2089 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2090 		}
2091 		break;
2092 	case IPV6_USE_MIN_MTU:
2093 		if (pktopt)
2094 			optdata = (void *)&pktopt->ip6po_minmtu;
2095 		else
2096 			optdata = (void *)&defminmtu;
2097 		optdatalen = sizeof(int);
2098 		break;
2099 	case IPV6_DONTFRAG:
2100 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2101 			on = 1;
2102 		else
2103 			on = 0;
2104 		optdata = (void *)&on;
2105 		optdatalen = sizeof(on);
2106 		break;
2107 	default:		/* should not happen */
2108 #ifdef DIAGNOSTIC
2109 		panic("ip6_getpcbopt: unexpected option\n");
2110 #endif
2111 		return (ENOPROTOOPT);
2112 	}
2113 
2114 	error = sockopt_set(sopt, optdata, optdatalen);
2115 
2116 	return (error);
2117 }
2118 
2119 void
2120 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2121 {
2122 	if (optname == -1 || optname == IPV6_PKTINFO) {
2123 		if (pktopt->ip6po_pktinfo)
2124 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2125 		pktopt->ip6po_pktinfo = NULL;
2126 	}
2127 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2128 		pktopt->ip6po_hlim = -1;
2129 	if (optname == -1 || optname == IPV6_TCLASS)
2130 		pktopt->ip6po_tclass = -1;
2131 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2132 		rtcache_free(&pktopt->ip6po_nextroute);
2133 		if (pktopt->ip6po_nexthop)
2134 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2135 		pktopt->ip6po_nexthop = NULL;
2136 	}
2137 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2138 		if (pktopt->ip6po_hbh)
2139 			free(pktopt->ip6po_hbh, M_IP6OPT);
2140 		pktopt->ip6po_hbh = NULL;
2141 	}
2142 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2143 		if (pktopt->ip6po_dest1)
2144 			free(pktopt->ip6po_dest1, M_IP6OPT);
2145 		pktopt->ip6po_dest1 = NULL;
2146 	}
2147 	if (optname == -1 || optname == IPV6_RTHDR) {
2148 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2149 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2150 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2151 		rtcache_free(&pktopt->ip6po_route);
2152 	}
2153 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2154 		if (pktopt->ip6po_dest2)
2155 			free(pktopt->ip6po_dest2, M_IP6OPT);
2156 		pktopt->ip6po_dest2 = NULL;
2157 	}
2158 }
2159 
2160 #define PKTOPT_EXTHDRCPY(type) 					\
2161 do {								\
2162 	if (src->type) {					\
2163 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2164 		dst->type = malloc(hlen, M_IP6OPT, canwait);	\
2165 		if (dst->type == NULL)				\
2166 			goto bad;				\
2167 		memcpy(dst->type, src->type, hlen);		\
2168 	}							\
2169 } while (/*CONSTCOND*/ 0)
2170 
2171 static int
2172 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2173 {
2174 	dst->ip6po_hlim = src->ip6po_hlim;
2175 	dst->ip6po_tclass = src->ip6po_tclass;
2176 	dst->ip6po_flags = src->ip6po_flags;
2177 	if (src->ip6po_pktinfo) {
2178 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2179 		    M_IP6OPT, canwait);
2180 		if (dst->ip6po_pktinfo == NULL)
2181 			goto bad;
2182 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2183 	}
2184 	if (src->ip6po_nexthop) {
2185 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2186 		    M_IP6OPT, canwait);
2187 		if (dst->ip6po_nexthop == NULL)
2188 			goto bad;
2189 		memcpy(dst->ip6po_nexthop, src->ip6po_nexthop,
2190 		    src->ip6po_nexthop->sa_len);
2191 	}
2192 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2193 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2194 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2195 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2196 	return (0);
2197 
2198   bad:
2199 	if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
2200 	if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
2201 	if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
2202 	if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
2203 	if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
2204 	if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
2205 
2206 	return (ENOBUFS);
2207 }
2208 #undef PKTOPT_EXTHDRCPY
2209 
2210 struct ip6_pktopts *
2211 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2212 {
2213 	int error;
2214 	struct ip6_pktopts *dst;
2215 
2216 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2217 	if (dst == NULL)
2218 		return (NULL);
2219 	ip6_initpktopts(dst);
2220 
2221 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2222 		free(dst, M_IP6OPT);
2223 		return (NULL);
2224 	}
2225 
2226 	return (dst);
2227 }
2228 
2229 void
2230 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2231 {
2232 	if (pktopt == NULL)
2233 		return;
2234 
2235 	ip6_clearpktopts(pktopt, -1);
2236 
2237 	free(pktopt, M_IP6OPT);
2238 }
2239 
2240 /*
2241  * Set the IP6 multicast options in response to user setsockopt().
2242  */
2243 static int
2244 ip6_setmoptions(const struct sockopt *sopt, struct ip6_moptions **im6op)
2245 {
2246 	int error = 0;
2247 	u_int loop, ifindex;
2248 	struct ipv6_mreq mreq;
2249 	struct ifnet *ifp;
2250 	struct ip6_moptions *im6o = *im6op;
2251 	struct route ro;
2252 	struct in6_multi_mship *imm;
2253 	struct lwp *l = curlwp;	/* XXX */
2254 
2255 	if (im6o == NULL) {
2256 		/*
2257 		 * No multicast option buffer attached to the pcb;
2258 		 * allocate one and initialize to default values.
2259 		 */
2260 		im6o = malloc(sizeof(*im6o), M_IPMOPTS, M_NOWAIT);
2261 		if (im6o == NULL)
2262 			return (ENOBUFS);
2263 
2264 		*im6op = im6o;
2265 		im6o->im6o_multicast_ifp = NULL;
2266 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2267 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2268 		LIST_INIT(&im6o->im6o_memberships);
2269 	}
2270 
2271 	switch (sopt->sopt_name) {
2272 
2273 	case IPV6_MULTICAST_IF:
2274 		/*
2275 		 * Select the interface for outgoing multicast packets.
2276 		 */
2277 		error = sockopt_get(sopt, &ifindex, sizeof(ifindex));
2278 		if (error != 0)
2279 			break;
2280 
2281 		if (ifindex != 0) {
2282 			if (if_indexlim <= ifindex || !ifindex2ifnet[ifindex]) {
2283 				error = ENXIO;	/* XXX EINVAL? */
2284 				break;
2285 			}
2286 			ifp = ifindex2ifnet[ifindex];
2287 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
2288 				error = EADDRNOTAVAIL;
2289 				break;
2290 			}
2291 		} else
2292 			ifp = NULL;
2293 		im6o->im6o_multicast_ifp = ifp;
2294 		break;
2295 
2296 	case IPV6_MULTICAST_HOPS:
2297 	    {
2298 		/*
2299 		 * Set the IP6 hoplimit for outgoing multicast packets.
2300 		 */
2301 		int optval;
2302 
2303 		error = sockopt_getint(sopt, &optval);
2304 		if (error != 0)
2305 			break;
2306 
2307 		if (optval < -1 || optval >= 256)
2308 			error = EINVAL;
2309 		else if (optval == -1)
2310 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2311 		else
2312 			im6o->im6o_multicast_hlim = optval;
2313 		break;
2314 	    }
2315 
2316 	case IPV6_MULTICAST_LOOP:
2317 		/*
2318 		 * Set the loopback flag for outgoing multicast packets.
2319 		 * Must be zero or one.
2320 		 */
2321 		error = sockopt_get(sopt, &loop, sizeof(loop));
2322 		if (error != 0)
2323 			break;
2324 		if (loop > 1) {
2325 			error = EINVAL;
2326 			break;
2327 		}
2328 		im6o->im6o_multicast_loop = loop;
2329 		break;
2330 
2331 	case IPV6_JOIN_GROUP:
2332 		/*
2333 		 * Add a multicast group membership.
2334 		 * Group must be a valid IP6 multicast address.
2335 		 */
2336 		error = sockopt_get(sopt, &mreq, sizeof(mreq));
2337 		if (error != 0)
2338 			break;
2339 
2340 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq.ipv6mr_multiaddr)) {
2341 			/*
2342 			 * We use the unspecified address to specify to accept
2343 			 * all multicast addresses. Only super user is allowed
2344 			 * to do this.
2345 			 */
2346 			if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_IPV6,
2347 			    KAUTH_REQ_NETWORK_IPV6_JOIN_MULTICAST, NULL, NULL, NULL))
2348 			{
2349 				error = EACCES;
2350 				break;
2351 			}
2352 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq.ipv6mr_multiaddr)) {
2353 			error = EINVAL;
2354 			break;
2355 		}
2356 
2357 		/*
2358 		 * If no interface was explicitly specified, choose an
2359 		 * appropriate one according to the given multicast address.
2360 		 */
2361 		if (mreq.ipv6mr_interface == 0) {
2362 			struct rtentry *rt;
2363 			union {
2364 				struct sockaddr		dst;
2365 				struct sockaddr_in6	dst6;
2366 			} u;
2367 
2368 			/*
2369 			 * Look up the routing table for the
2370 			 * address, and choose the outgoing interface.
2371 			 *   XXX: is it a good approach?
2372 			 */
2373 			memset(&ro, 0, sizeof(ro));
2374 			sockaddr_in6_init(&u.dst6, &mreq.ipv6mr_multiaddr, 0,
2375 			    0, 0);
2376 			rtcache_setdst(&ro, &u.dst);
2377 			ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp
2378 			                                        : NULL;
2379 			rtcache_free(&ro);
2380 		} else {
2381 			/*
2382 			 * If the interface is specified, validate it.
2383 			 */
2384 			if (if_indexlim <= mreq.ipv6mr_interface ||
2385 			    !ifindex2ifnet[mreq.ipv6mr_interface]) {
2386 				error = ENXIO;	/* XXX EINVAL? */
2387 				break;
2388 			}
2389 			ifp = ifindex2ifnet[mreq.ipv6mr_interface];
2390 		}
2391 
2392 		/*
2393 		 * See if we found an interface, and confirm that it
2394 		 * supports multicast
2395 		 */
2396 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2397 			error = EADDRNOTAVAIL;
2398 			break;
2399 		}
2400 
2401 		if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) {
2402 			error = EADDRNOTAVAIL; /* XXX: should not happen */
2403 			break;
2404 		}
2405 
2406 		/*
2407 		 * See if the membership already exists.
2408 		 */
2409 		for (imm = im6o->im6o_memberships.lh_first;
2410 		     imm != NULL; imm = imm->i6mm_chain.le_next)
2411 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
2412 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2413 			    &mreq.ipv6mr_multiaddr))
2414 				break;
2415 		if (imm != NULL) {
2416 			error = EADDRINUSE;
2417 			break;
2418 		}
2419 		/*
2420 		 * Everything looks good; add a new record to the multicast
2421 		 * address list for the given interface.
2422 		 */
2423 		imm = in6_joingroup(ifp, &mreq.ipv6mr_multiaddr, &error, 0);
2424 		if (imm == NULL)
2425 			break;
2426 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2427 		break;
2428 
2429 	case IPV6_LEAVE_GROUP:
2430 		/*
2431 		 * Drop a multicast group membership.
2432 		 * Group must be a valid IP6 multicast address.
2433 		 */
2434 		error = sockopt_get(sopt, &mreq, sizeof(mreq));
2435 		if (error != 0)
2436 			break;
2437 
2438 		/*
2439 		 * If an interface address was specified, get a pointer
2440 		 * to its ifnet structure.
2441 		 */
2442 		if (mreq.ipv6mr_interface != 0) {
2443 			if (if_indexlim <= mreq.ipv6mr_interface ||
2444 			    !ifindex2ifnet[mreq.ipv6mr_interface]) {
2445 				error = ENXIO;	/* XXX EINVAL? */
2446 				break;
2447 			}
2448 			ifp = ifindex2ifnet[mreq.ipv6mr_interface];
2449 		} else
2450 			ifp = NULL;
2451 
2452 		/* Fill in the scope zone ID */
2453 		if (ifp) {
2454 			if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) {
2455 				/* XXX: should not happen */
2456 				error = EADDRNOTAVAIL;
2457 				break;
2458 			}
2459 		} else if (mreq.ipv6mr_interface != 0) {
2460 			/*
2461 			 * XXX: This case would happens when the (positive)
2462 			 * index is in the valid range, but the corresponding
2463 			 * interface has been detached dynamically.  The above
2464 			 * check probably avoids such case to happen here, but
2465 			 * we check it explicitly for safety.
2466 			 */
2467 			error = EADDRNOTAVAIL;
2468 			break;
2469 		} else {	/* ipv6mr_interface == 0 */
2470 			struct sockaddr_in6 sa6_mc;
2471 
2472 			/*
2473 			 * The API spec says as follows:
2474 			 *  If the interface index is specified as 0, the
2475 			 *  system may choose a multicast group membership to
2476 			 *  drop by matching the multicast address only.
2477 			 * On the other hand, we cannot disambiguate the scope
2478 			 * zone unless an interface is provided.  Thus, we
2479 			 * check if there's ambiguity with the default scope
2480 			 * zone as the last resort.
2481 			 */
2482 			sockaddr_in6_init(&sa6_mc, &mreq.ipv6mr_multiaddr,
2483 			    0, 0, 0);
2484 			error = sa6_embedscope(&sa6_mc, ip6_use_defzone);
2485 			if (error != 0)
2486 				break;
2487 			mreq.ipv6mr_multiaddr = sa6_mc.sin6_addr;
2488 		}
2489 
2490 		/*
2491 		 * Find the membership in the membership list.
2492 		 */
2493 		for (imm = im6o->im6o_memberships.lh_first;
2494 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
2495 			if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2496 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2497 			    &mreq.ipv6mr_multiaddr))
2498 				break;
2499 		}
2500 		if (imm == NULL) {
2501 			/* Unable to resolve interface */
2502 			error = EADDRNOTAVAIL;
2503 			break;
2504 		}
2505 		/*
2506 		 * Give up the multicast address record to which the
2507 		 * membership points.
2508 		 */
2509 		LIST_REMOVE(imm, i6mm_chain);
2510 		in6_leavegroup(imm);
2511 		break;
2512 
2513 	default:
2514 		error = EOPNOTSUPP;
2515 		break;
2516 	}
2517 
2518 	/*
2519 	 * If all options have default values, no need to keep the mbuf.
2520 	 */
2521 	if (im6o->im6o_multicast_ifp == NULL &&
2522 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2523 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2524 	    im6o->im6o_memberships.lh_first == NULL) {
2525 		free(*im6op, M_IPMOPTS);
2526 		*im6op = NULL;
2527 	}
2528 
2529 	return (error);
2530 }
2531 
2532 /*
2533  * Return the IP6 multicast options in response to user getsockopt().
2534  */
2535 static int
2536 ip6_getmoptions(struct sockopt *sopt, struct ip6_moptions *im6o)
2537 {
2538 	u_int optval;
2539 	int error;
2540 
2541 	switch (sopt->sopt_name) {
2542 	case IPV6_MULTICAST_IF:
2543 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2544 			optval = 0;
2545 		else
2546 			optval = im6o->im6o_multicast_ifp->if_index;
2547 
2548 		error = sockopt_set(sopt, &optval, sizeof(optval));
2549 		break;
2550 
2551 	case IPV6_MULTICAST_HOPS:
2552 		if (im6o == NULL)
2553 			optval = ip6_defmcasthlim;
2554 		else
2555 			optval = im6o->im6o_multicast_hlim;
2556 
2557 		error = sockopt_set(sopt, &optval, sizeof(optval));
2558 		break;
2559 
2560 	case IPV6_MULTICAST_LOOP:
2561 		if (im6o == NULL)
2562 			optval = IPV6_DEFAULT_MULTICAST_LOOP;
2563 		else
2564 			optval = im6o->im6o_multicast_loop;
2565 
2566 		error = sockopt_set(sopt, &optval, sizeof(optval));
2567 		break;
2568 
2569 	default:
2570 		error = EOPNOTSUPP;
2571 	}
2572 
2573 	return (error);
2574 }
2575 
2576 /*
2577  * Discard the IP6 multicast options.
2578  */
2579 void
2580 ip6_freemoptions(struct ip6_moptions *im6o)
2581 {
2582 	struct in6_multi_mship *imm;
2583 
2584 	if (im6o == NULL)
2585 		return;
2586 
2587 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2588 		LIST_REMOVE(imm, i6mm_chain);
2589 		in6_leavegroup(imm);
2590 	}
2591 	free(im6o, M_IPMOPTS);
2592 }
2593 
2594 /*
2595  * Set IPv6 outgoing packet options based on advanced API.
2596  */
2597 int
2598 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2599 	struct ip6_pktopts *stickyopt, kauth_cred_t cred, int uproto)
2600 {
2601 	struct cmsghdr *cm = 0;
2602 
2603 	if (control == NULL || opt == NULL)
2604 		return (EINVAL);
2605 
2606 	ip6_initpktopts(opt);
2607 	if (stickyopt) {
2608 		int error;
2609 
2610 		/*
2611 		 * If stickyopt is provided, make a local copy of the options
2612 		 * for this particular packet, then override them by ancillary
2613 		 * objects.
2614 		 * XXX: copypktopts() does not copy the cached route to a next
2615 		 * hop (if any).  This is not very good in terms of efficiency,
2616 		 * but we can allow this since this option should be rarely
2617 		 * used.
2618 		 */
2619 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2620 			return (error);
2621 	}
2622 
2623 	/*
2624 	 * XXX: Currently, we assume all the optional information is stored
2625 	 * in a single mbuf.
2626 	 */
2627 	if (control->m_next)
2628 		return (EINVAL);
2629 
2630 	/* XXX if cm->cmsg_len is not aligned, control->m_len can become <0 */
2631 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2632 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2633 		int error;
2634 
2635 		if (control->m_len < CMSG_LEN(0))
2636 			return (EINVAL);
2637 
2638 		cm = mtod(control, struct cmsghdr *);
2639 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2640 			return (EINVAL);
2641 		if (cm->cmsg_level != IPPROTO_IPV6)
2642 			continue;
2643 
2644 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2645 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2646 		if (error)
2647 			return (error);
2648 	}
2649 
2650 	return (0);
2651 }
2652 
2653 /*
2654  * Set a particular packet option, as a sticky option or an ancillary data
2655  * item.  "len" can be 0 only when it's a sticky option.
2656  * We have 4 cases of combination of "sticky" and "cmsg":
2657  * "sticky=0, cmsg=0": impossible
2658  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2659  * "sticky=1, cmsg=0": RFC3542 socket option
2660  * "sticky=1, cmsg=1": RFC2292 socket option
2661  */
2662 static int
2663 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2664     kauth_cred_t cred, int sticky, int cmsg, int uproto)
2665 {
2666 	int minmtupolicy;
2667 	int error;
2668 
2669 	if (!sticky && !cmsg) {
2670 #ifdef DIAGNOSTIC
2671 		printf("ip6_setpktopt: impossible case\n");
2672 #endif
2673 		return (EINVAL);
2674 	}
2675 
2676 	/*
2677 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2678 	 * not be specified in the context of RFC3542.  Conversely,
2679 	 * RFC3542 types should not be specified in the context of RFC2292.
2680 	 */
2681 	if (!cmsg) {
2682 		switch (optname) {
2683 		case IPV6_2292PKTINFO:
2684 		case IPV6_2292HOPLIMIT:
2685 		case IPV6_2292NEXTHOP:
2686 		case IPV6_2292HOPOPTS:
2687 		case IPV6_2292DSTOPTS:
2688 		case IPV6_2292RTHDR:
2689 		case IPV6_2292PKTOPTIONS:
2690 			return (ENOPROTOOPT);
2691 		}
2692 	}
2693 	if (sticky && cmsg) {
2694 		switch (optname) {
2695 		case IPV6_PKTINFO:
2696 		case IPV6_HOPLIMIT:
2697 		case IPV6_NEXTHOP:
2698 		case IPV6_HOPOPTS:
2699 		case IPV6_DSTOPTS:
2700 		case IPV6_RTHDRDSTOPTS:
2701 		case IPV6_RTHDR:
2702 		case IPV6_USE_MIN_MTU:
2703 		case IPV6_DONTFRAG:
2704 		case IPV6_OTCLASS:
2705 		case IPV6_TCLASS:
2706 			return (ENOPROTOOPT);
2707 		}
2708 	}
2709 
2710 	switch (optname) {
2711 #ifdef RFC2292
2712 	case IPV6_2292PKTINFO:
2713 #endif
2714 	case IPV6_PKTINFO:
2715 	{
2716 		struct ifnet *ifp = NULL;
2717 		struct in6_pktinfo *pktinfo;
2718 
2719 		if (len != sizeof(struct in6_pktinfo))
2720 			return (EINVAL);
2721 
2722 		pktinfo = (struct in6_pktinfo *)buf;
2723 
2724 		/*
2725 		 * An application can clear any sticky IPV6_PKTINFO option by
2726 		 * doing a "regular" setsockopt with ipi6_addr being
2727 		 * in6addr_any and ipi6_ifindex being zero.
2728 		 * [RFC 3542, Section 6]
2729 		 */
2730 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2731 		    pktinfo->ipi6_ifindex == 0 &&
2732 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2733 			ip6_clearpktopts(opt, optname);
2734 			break;
2735 		}
2736 
2737 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2738 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2739 			return (EINVAL);
2740 		}
2741 
2742 		/* validate the interface index if specified. */
2743 		if (pktinfo->ipi6_ifindex >= if_indexlim) {
2744 			 return (ENXIO);
2745 		}
2746 		if (pktinfo->ipi6_ifindex) {
2747 			ifp = ifindex2ifnet[pktinfo->ipi6_ifindex];
2748 			if (ifp == NULL)
2749 				return (ENXIO);
2750 		}
2751 
2752 		/*
2753 		 * We store the address anyway, and let in6_selectsrc()
2754 		 * validate the specified address.  This is because ipi6_addr
2755 		 * may not have enough information about its scope zone, and
2756 		 * we may need additional information (such as outgoing
2757 		 * interface or the scope zone of a destination address) to
2758 		 * disambiguate the scope.
2759 		 * XXX: the delay of the validation may confuse the
2760 		 * application when it is used as a sticky option.
2761 		 */
2762 		if (opt->ip6po_pktinfo == NULL) {
2763 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2764 			    M_IP6OPT, M_NOWAIT);
2765 			if (opt->ip6po_pktinfo == NULL)
2766 				return (ENOBUFS);
2767 		}
2768 		memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo));
2769 		break;
2770 	}
2771 
2772 #ifdef RFC2292
2773 	case IPV6_2292HOPLIMIT:
2774 #endif
2775 	case IPV6_HOPLIMIT:
2776 	{
2777 		int *hlimp;
2778 
2779 		/*
2780 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2781 		 * to simplify the ordering among hoplimit options.
2782 		 */
2783 		if (optname == IPV6_HOPLIMIT && sticky)
2784 			return (ENOPROTOOPT);
2785 
2786 		if (len != sizeof(int))
2787 			return (EINVAL);
2788 		hlimp = (int *)buf;
2789 		if (*hlimp < -1 || *hlimp > 255)
2790 			return (EINVAL);
2791 
2792 		opt->ip6po_hlim = *hlimp;
2793 		break;
2794 	}
2795 
2796 	case IPV6_OTCLASS:
2797 		if (len != sizeof(u_int8_t))
2798 			return (EINVAL);
2799 
2800 		opt->ip6po_tclass = *(u_int8_t *)buf;
2801 		break;
2802 
2803 	case IPV6_TCLASS:
2804 	{
2805 		int tclass;
2806 
2807 		if (len != sizeof(int))
2808 			return (EINVAL);
2809 		tclass = *(int *)buf;
2810 		if (tclass < -1 || tclass > 255)
2811 			return (EINVAL);
2812 
2813 		opt->ip6po_tclass = tclass;
2814 		break;
2815 	}
2816 
2817 #ifdef RFC2292
2818 	case IPV6_2292NEXTHOP:
2819 #endif
2820 	case IPV6_NEXTHOP:
2821 		error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
2822 		    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2823 		if (error)
2824 			return (error);
2825 
2826 		if (len == 0) {	/* just remove the option */
2827 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2828 			break;
2829 		}
2830 
2831 		/* check if cmsg_len is large enough for sa_len */
2832 		if (len < sizeof(struct sockaddr) || len < *buf)
2833 			return (EINVAL);
2834 
2835 		switch (((struct sockaddr *)buf)->sa_family) {
2836 		case AF_INET6:
2837 		{
2838 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2839 
2840 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2841 				return (EINVAL);
2842 
2843 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2844 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2845 				return (EINVAL);
2846 			}
2847 			if ((error = sa6_embedscope(sa6, ip6_use_defzone))
2848 			    != 0) {
2849 				return (error);
2850 			}
2851 			break;
2852 		}
2853 		case AF_LINK:	/* eventually be supported? */
2854 		default:
2855 			return (EAFNOSUPPORT);
2856 		}
2857 
2858 		/* turn off the previous option, then set the new option. */
2859 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2860 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2861 		if (opt->ip6po_nexthop == NULL)
2862 			return (ENOBUFS);
2863 		memcpy(opt->ip6po_nexthop, buf, *buf);
2864 		break;
2865 
2866 #ifdef RFC2292
2867 	case IPV6_2292HOPOPTS:
2868 #endif
2869 	case IPV6_HOPOPTS:
2870 	{
2871 		struct ip6_hbh *hbh;
2872 		int hbhlen;
2873 
2874 		/*
2875 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2876 		 * options, since per-option restriction has too much
2877 		 * overhead.
2878 		 */
2879 		error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
2880 		    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2881 		if (error)
2882 			return (error);
2883 
2884 		if (len == 0) {
2885 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2886 			break;	/* just remove the option */
2887 		}
2888 
2889 		/* message length validation */
2890 		if (len < sizeof(struct ip6_hbh))
2891 			return (EINVAL);
2892 		hbh = (struct ip6_hbh *)buf;
2893 		hbhlen = (hbh->ip6h_len + 1) << 3;
2894 		if (len != hbhlen)
2895 			return (EINVAL);
2896 
2897 		/* turn off the previous option, then set the new option. */
2898 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2899 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2900 		if (opt->ip6po_hbh == NULL)
2901 			return (ENOBUFS);
2902 		memcpy(opt->ip6po_hbh, hbh, hbhlen);
2903 
2904 		break;
2905 	}
2906 
2907 #ifdef RFC2292
2908 	case IPV6_2292DSTOPTS:
2909 #endif
2910 	case IPV6_DSTOPTS:
2911 	case IPV6_RTHDRDSTOPTS:
2912 	{
2913 		struct ip6_dest *dest, **newdest = NULL;
2914 		int destlen;
2915 
2916 		/* XXX: see the comment for IPV6_HOPOPTS */
2917 		error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
2918 		    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2919 		if (error)
2920 			return (error);
2921 
2922 		if (len == 0) {
2923 			ip6_clearpktopts(opt, optname);
2924 			break;	/* just remove the option */
2925 		}
2926 
2927 		/* message length validation */
2928 		if (len < sizeof(struct ip6_dest))
2929 			return (EINVAL);
2930 		dest = (struct ip6_dest *)buf;
2931 		destlen = (dest->ip6d_len + 1) << 3;
2932 		if (len != destlen)
2933 			return (EINVAL);
2934 		/*
2935 		 * Determine the position that the destination options header
2936 		 * should be inserted; before or after the routing header.
2937 		 */
2938 		switch (optname) {
2939 		case IPV6_2292DSTOPTS:
2940 			/*
2941 			 * The old advanced API is ambiguous on this point.
2942 			 * Our approach is to determine the position based
2943 			 * according to the existence of a routing header.
2944 			 * Note, however, that this depends on the order of the
2945 			 * extension headers in the ancillary data; the 1st
2946 			 * part of the destination options header must appear
2947 			 * before the routing header in the ancillary data,
2948 			 * too.
2949 			 * RFC3542 solved the ambiguity by introducing
2950 			 * separate ancillary data or option types.
2951 			 */
2952 			if (opt->ip6po_rthdr == NULL)
2953 				newdest = &opt->ip6po_dest1;
2954 			else
2955 				newdest = &opt->ip6po_dest2;
2956 			break;
2957 		case IPV6_RTHDRDSTOPTS:
2958 			newdest = &opt->ip6po_dest1;
2959 			break;
2960 		case IPV6_DSTOPTS:
2961 			newdest = &opt->ip6po_dest2;
2962 			break;
2963 		}
2964 
2965 		/* turn off the previous option, then set the new option. */
2966 		ip6_clearpktopts(opt, optname);
2967 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2968 		if (*newdest == NULL)
2969 			return (ENOBUFS);
2970 		memcpy(*newdest, dest, destlen);
2971 
2972 		break;
2973 	}
2974 
2975 #ifdef RFC2292
2976 	case IPV6_2292RTHDR:
2977 #endif
2978 	case IPV6_RTHDR:
2979 	{
2980 		struct ip6_rthdr *rth;
2981 		int rthlen;
2982 
2983 		if (len == 0) {
2984 			ip6_clearpktopts(opt, IPV6_RTHDR);
2985 			break;	/* just remove the option */
2986 		}
2987 
2988 		/* message length validation */
2989 		if (len < sizeof(struct ip6_rthdr))
2990 			return (EINVAL);
2991 		rth = (struct ip6_rthdr *)buf;
2992 		rthlen = (rth->ip6r_len + 1) << 3;
2993 		if (len != rthlen)
2994 			return (EINVAL);
2995 		switch (rth->ip6r_type) {
2996 		case IPV6_RTHDR_TYPE_0:
2997 			if (rth->ip6r_len == 0)	/* must contain one addr */
2998 				return (EINVAL);
2999 			if (rth->ip6r_len % 2) /* length must be even */
3000 				return (EINVAL);
3001 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3002 				return (EINVAL);
3003 			break;
3004 		default:
3005 			return (EINVAL);	/* not supported */
3006 		}
3007 		/* turn off the previous option */
3008 		ip6_clearpktopts(opt, IPV6_RTHDR);
3009 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3010 		if (opt->ip6po_rthdr == NULL)
3011 			return (ENOBUFS);
3012 		memcpy(opt->ip6po_rthdr, rth, rthlen);
3013 		break;
3014 	}
3015 
3016 	case IPV6_USE_MIN_MTU:
3017 		if (len != sizeof(int))
3018 			return (EINVAL);
3019 		minmtupolicy = *(int *)buf;
3020 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3021 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3022 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3023 			return (EINVAL);
3024 		}
3025 		opt->ip6po_minmtu = minmtupolicy;
3026 		break;
3027 
3028 	case IPV6_DONTFRAG:
3029 		if (len != sizeof(int))
3030 			return (EINVAL);
3031 
3032 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3033 			/*
3034 			 * we ignore this option for TCP sockets.
3035 			 * (RFC3542 leaves this case unspecified.)
3036 			 */
3037 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3038 		} else
3039 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3040 		break;
3041 
3042 	default:
3043 		return (ENOPROTOOPT);
3044 	} /* end of switch */
3045 
3046 	return (0);
3047 }
3048 
3049 /*
3050  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3051  * packet to the input queue of a specified interface.  Note that this
3052  * calls the output routine of the loopback "driver", but with an interface
3053  * pointer that might NOT be lo0ifp -- easier than replicating that code here.
3054  */
3055 void
3056 ip6_mloopback(struct ifnet *ifp, struct mbuf *m,
3057 	const struct sockaddr_in6 *dst)
3058 {
3059 	struct mbuf *copym;
3060 	struct ip6_hdr *ip6;
3061 
3062 	copym = m_copy(m, 0, M_COPYALL);
3063 	if (copym == NULL)
3064 		return;
3065 
3066 	/*
3067 	 * Make sure to deep-copy IPv6 header portion in case the data
3068 	 * is in an mbuf cluster, so that we can safely override the IPv6
3069 	 * header portion later.
3070 	 */
3071 	if ((copym->m_flags & M_EXT) != 0 ||
3072 	    copym->m_len < sizeof(struct ip6_hdr)) {
3073 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3074 		if (copym == NULL)
3075 			return;
3076 	}
3077 
3078 #ifdef DIAGNOSTIC
3079 	if (copym->m_len < sizeof(*ip6)) {
3080 		m_freem(copym);
3081 		return;
3082 	}
3083 #endif
3084 
3085 	ip6 = mtod(copym, struct ip6_hdr *);
3086 	/*
3087 	 * clear embedded scope identifiers if necessary.
3088 	 * in6_clearscope will touch the addresses only when necessary.
3089 	 */
3090 	in6_clearscope(&ip6->ip6_src);
3091 	in6_clearscope(&ip6->ip6_dst);
3092 
3093 	(void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL);
3094 }
3095 
3096 /*
3097  * Chop IPv6 header off from the payload.
3098  */
3099 static int
3100 ip6_splithdr(struct mbuf *m,  struct ip6_exthdrs *exthdrs)
3101 {
3102 	struct mbuf *mh;
3103 	struct ip6_hdr *ip6;
3104 
3105 	ip6 = mtod(m, struct ip6_hdr *);
3106 	if (m->m_len > sizeof(*ip6)) {
3107 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3108 		if (mh == 0) {
3109 			m_freem(m);
3110 			return ENOBUFS;
3111 		}
3112 		M_MOVE_PKTHDR(mh, m);
3113 		MH_ALIGN(mh, sizeof(*ip6));
3114 		m->m_len -= sizeof(*ip6);
3115 		m->m_data += sizeof(*ip6);
3116 		mh->m_next = m;
3117 		m = mh;
3118 		m->m_len = sizeof(*ip6);
3119 		bcopy((void *)ip6, mtod(m, void *), sizeof(*ip6));
3120 	}
3121 	exthdrs->ip6e_ip6 = m;
3122 	return 0;
3123 }
3124 
3125 /*
3126  * Compute IPv6 extension header length.
3127  */
3128 int
3129 ip6_optlen(struct in6pcb *in6p)
3130 {
3131 	int len;
3132 
3133 	if (!in6p->in6p_outputopts)
3134 		return 0;
3135 
3136 	len = 0;
3137 #define elen(x) \
3138     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3139 
3140 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3141 	len += elen(in6p->in6p_outputopts->ip6po_dest1);
3142 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3143 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3144 	return len;
3145 #undef elen
3146 }
3147