xref: /netbsd-src/sys/netinet6/ip6_output.c (revision 975a152cfcdb39ae6e496af647af0c7275ca0b61)
1 /*	$NetBSD: ip6_output.c,v 1.154 2013/06/29 21:06:58 rmind 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 1.154 2013/06/29 21:06:58 rmind Exp $");
66 
67 #include "opt_inet.h"
68 #include "opt_inet6.h"
69 #include "opt_ipsec.h"
70 
71 #include <sys/param.h>
72 #include <sys/malloc.h>
73 #include <sys/mbuf.h>
74 #include <sys/errno.h>
75 #include <sys/protosw.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/systm.h>
79 #include <sys/proc.h>
80 #include <sys/kauth.h>
81 
82 #include <net/if.h>
83 #include <net/route.h>
84 #include <net/pfil.h>
85 
86 #include <netinet/in.h>
87 #include <netinet/in_var.h>
88 #include <netinet/ip6.h>
89 #include <netinet/icmp6.h>
90 #include <netinet/in_offload.h>
91 #include <netinet/portalgo.h>
92 #include <netinet6/in6_offload.h>
93 #include <netinet6/ip6_var.h>
94 #include <netinet6/ip6_private.h>
95 #include <netinet6/in6_pcb.h>
96 #include <netinet6/nd6.h>
97 #include <netinet6/ip6protosw.h>
98 #include <netinet6/scope6_var.h>
99 
100 #ifdef IPSEC
101 #include <netipsec/ipsec.h>
102 #include <netipsec/ipsec6.h>
103 #include <netipsec/key.h>
104 #include <netipsec/xform.h>
105 #endif
106 
107 
108 #include <net/net_osdep.h>
109 
110 extern pfil_head_t *inet6_pfil_hook;	/* XXX */
111 
112 struct ip6_exthdrs {
113 	struct mbuf *ip6e_ip6;
114 	struct mbuf *ip6e_hbh;
115 	struct mbuf *ip6e_dest1;
116 	struct mbuf *ip6e_rthdr;
117 	struct mbuf *ip6e_dest2;
118 };
119 
120 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
121 	kauth_cred_t, int);
122 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
123 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, kauth_cred_t,
124 	int, int, int);
125 static int ip6_setmoptions(const struct sockopt *, struct ip6_moptions **);
126 static int ip6_getmoptions(struct sockopt *, struct ip6_moptions *);
127 static int ip6_copyexthdr(struct mbuf **, void *, int);
128 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
129 	struct ip6_frag **);
130 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
131 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
132 static int ip6_getpmtu(struct route *, struct route *, struct ifnet *,
133     const struct in6_addr *, u_long *, int *);
134 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
135 
136 #ifdef RFC2292
137 static int ip6_pcbopts(struct ip6_pktopts **, struct socket *, struct sockopt *);
138 #endif
139 
140 /*
141  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
142  * header (with pri, len, nxt, hlim, src, dst).
143  * This function may modify ver and hlim only.
144  * The mbuf chain containing the packet will be freed.
145  * The mbuf opt, if present, will not be freed.
146  *
147  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
148  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
149  * which is rt_rmx.rmx_mtu.
150  */
151 int
152 ip6_output(
153     struct mbuf *m0,
154     struct ip6_pktopts *opt,
155     struct route *ro,
156     int flags,
157     struct ip6_moptions *im6o,
158     struct socket *so,
159     struct ifnet **ifpp		/* XXX: just for statistics */
160 )
161 {
162 	struct ip6_hdr *ip6, *mhip6;
163 	struct ifnet *ifp, *origifp;
164 	struct mbuf *m = m0;
165 	int hlen, tlen, len, off;
166 	bool tso;
167 	struct route ip6route;
168 	struct rtentry *rt = NULL;
169 	const struct sockaddr_in6 *dst = NULL;
170 	struct sockaddr_in6 src_sa, dst_sa;
171 	int error = 0;
172 	struct in6_ifaddr *ia = NULL;
173 	u_long mtu;
174 	int alwaysfrag, dontfrag;
175 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
176 	struct ip6_exthdrs exthdrs;
177 	struct in6_addr finaldst, src0, dst0;
178 	u_int32_t zone;
179 	struct route *ro_pmtu = NULL;
180 	int hdrsplit = 0;
181 	int needipsec = 0;
182 #ifdef IPSEC
183 	struct secpolicy *sp = NULL;
184 	int s;
185 #endif
186 
187 	memset(&ip6route, 0, sizeof(ip6route));
188 
189 #ifdef  DIAGNOSTIC
190 	if ((m->m_flags & M_PKTHDR) == 0)
191 		panic("ip6_output: no HDR");
192 
193 	if ((m->m_pkthdr.csum_flags &
194 	    (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) {
195 		panic("ip6_output: IPv4 checksum offload flags: %d",
196 		    m->m_pkthdr.csum_flags);
197 	}
198 
199 	if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) ==
200 	    (M_CSUM_TCPv6|M_CSUM_UDPv6)) {
201 		panic("ip6_output: conflicting checksum offload flags: %d",
202 		    m->m_pkthdr.csum_flags);
203 	}
204 #endif
205 
206 	M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr));
207 
208 #define MAKE_EXTHDR(hp, mp)						\
209     do {								\
210 	if (hp) {							\
211 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
212 		error = ip6_copyexthdr((mp), (void *)(hp), 		\
213 		    ((eh)->ip6e_len + 1) << 3);				\
214 		if (error)						\
215 			goto freehdrs;					\
216 	}								\
217     } while (/*CONSTCOND*/ 0)
218 
219 	memset(&exthdrs, 0, sizeof(exthdrs));
220 	if (opt) {
221 		/* Hop-by-Hop options header */
222 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
223 		/* Destination options header(1st part) */
224 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
225 		/* Routing header */
226 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
227 		/* Destination options header(2nd part) */
228 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
229 	}
230 
231 	/*
232 	 * Calculate the total length of the extension header chain.
233 	 * Keep the length of the unfragmentable part for fragmentation.
234 	 */
235 	optlen = 0;
236 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
237 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
238 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
239 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
240 	/* NOTE: we don't add AH/ESP length here. do that later. */
241 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
242 
243 #ifdef IPSEC
244 	/* Check the security policy (SP) for the packet */
245 
246 	sp = ipsec6_check_policy(m,so,flags,&needipsec,&error);
247 	if (error != 0) {
248 		/*
249 		 * Hack: -EINVAL is used to signal that a packet
250 		 * should be silently discarded.  This is typically
251 		 * because we asked key management for an SA and
252 		 * it was delayed (e.g. kicked up to IKE).
253 		 */
254 	if (error == -EINVAL)
255 		error = 0;
256 	goto freehdrs;
257     }
258 #endif /* IPSEC */
259 
260 
261 	if (needipsec &&
262 	    (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
263 		in6_delayed_cksum(m);
264 		m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
265 	}
266 
267 
268 	/*
269 	 * If we need IPsec, or there is at least one extension header,
270 	 * separate IP6 header from the payload.
271 	 */
272 	if ((needipsec || optlen) && !hdrsplit) {
273 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
274 			m = NULL;
275 			goto freehdrs;
276 		}
277 		m = exthdrs.ip6e_ip6;
278 		hdrsplit++;
279 	}
280 
281 	/* adjust pointer */
282 	ip6 = mtod(m, struct ip6_hdr *);
283 
284 	/* adjust mbuf packet header length */
285 	m->m_pkthdr.len += optlen;
286 	plen = m->m_pkthdr.len - sizeof(*ip6);
287 
288 	/* If this is a jumbo payload, insert a jumbo payload option. */
289 	if (plen > IPV6_MAXPACKET) {
290 		if (!hdrsplit) {
291 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
292 				m = NULL;
293 				goto freehdrs;
294 			}
295 			m = exthdrs.ip6e_ip6;
296 			hdrsplit++;
297 		}
298 		/* adjust pointer */
299 		ip6 = mtod(m, struct ip6_hdr *);
300 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
301 			goto freehdrs;
302 		optlen += 8; /* XXX JUMBOOPTLEN */
303 		ip6->ip6_plen = 0;
304 	} else
305 		ip6->ip6_plen = htons(plen);
306 
307 	/*
308 	 * Concatenate headers and fill in next header fields.
309 	 * Here we have, on "m"
310 	 *	IPv6 payload
311 	 * and we insert headers accordingly.  Finally, we should be getting:
312 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
313 	 *
314 	 * during the header composing process, "m" points to IPv6 header.
315 	 * "mprev" points to an extension header prior to esp.
316 	 */
317 	{
318 		u_char *nexthdrp = &ip6->ip6_nxt;
319 		struct mbuf *mprev = m;
320 
321 		/*
322 		 * we treat dest2 specially.  this makes IPsec processing
323 		 * much easier.  the goal here is to make mprev point the
324 		 * mbuf prior to dest2.
325 		 *
326 		 * result: IPv6 dest2 payload
327 		 * m and mprev will point to IPv6 header.
328 		 */
329 		if (exthdrs.ip6e_dest2) {
330 			if (!hdrsplit)
331 				panic("assumption failed: hdr not split");
332 			exthdrs.ip6e_dest2->m_next = m->m_next;
333 			m->m_next = exthdrs.ip6e_dest2;
334 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
335 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
336 		}
337 
338 #define MAKE_CHAIN(m, mp, p, i)\
339     do {\
340 	if (m) {\
341 		if (!hdrsplit) \
342 			panic("assumption failed: hdr not split"); \
343 		*mtod((m), u_char *) = *(p);\
344 		*(p) = (i);\
345 		p = mtod((m), u_char *);\
346 		(m)->m_next = (mp)->m_next;\
347 		(mp)->m_next = (m);\
348 		(mp) = (m);\
349 	}\
350     } while (/*CONSTCOND*/ 0)
351 		/*
352 		 * result: IPv6 hbh dest1 rthdr dest2 payload
353 		 * m will point to IPv6 header.  mprev will point to the
354 		 * extension header prior to dest2 (rthdr in the above case).
355 		 */
356 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
357 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
358 		    IPPROTO_DSTOPTS);
359 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
360 		    IPPROTO_ROUTING);
361 
362 		M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data,
363 		    sizeof(struct ip6_hdr) + optlen);
364 	}
365 
366 	/*
367 	 * If there is a routing header, replace destination address field
368 	 * with the first hop of the routing header.
369 	 */
370 	if (exthdrs.ip6e_rthdr) {
371 		struct ip6_rthdr *rh;
372 		struct ip6_rthdr0 *rh0;
373 		struct in6_addr *addr;
374 		struct sockaddr_in6 sa;
375 
376 		rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
377 		    struct ip6_rthdr *));
378 		finaldst = ip6->ip6_dst;
379 		switch (rh->ip6r_type) {
380 		case IPV6_RTHDR_TYPE_0:
381 			 rh0 = (struct ip6_rthdr0 *)rh;
382 			 addr = (struct in6_addr *)(rh0 + 1);
383 
384 			 /*
385 			  * construct a sockaddr_in6 form of
386 			  * the first hop.
387 			  *
388 			  * XXX: we may not have enough
389 			  * information about its scope zone;
390 			  * there is no standard API to pass
391 			  * the information from the
392 			  * application.
393 			  */
394 			 sockaddr_in6_init(&sa, addr, 0, 0, 0);
395 			 if ((error = sa6_embedscope(&sa,
396 			     ip6_use_defzone)) != 0) {
397 				 goto bad;
398 			 }
399 			 ip6->ip6_dst = sa.sin6_addr;
400 			 (void)memmove(&addr[0], &addr[1],
401 			     sizeof(struct in6_addr) *
402 			     (rh0->ip6r0_segleft - 1));
403 			 addr[rh0->ip6r0_segleft - 1] = finaldst;
404 			 /* XXX */
405 			 in6_clearscope(addr + rh0->ip6r0_segleft - 1);
406 			 break;
407 		default:	/* is it possible? */
408 			 error = EINVAL;
409 			 goto bad;
410 		}
411 	}
412 
413 	/* Source address validation */
414 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
415 	    (flags & IPV6_UNSPECSRC) == 0) {
416 		error = EOPNOTSUPP;
417 		IP6_STATINC(IP6_STAT_BADSCOPE);
418 		goto bad;
419 	}
420 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
421 		error = EOPNOTSUPP;
422 		IP6_STATINC(IP6_STAT_BADSCOPE);
423 		goto bad;
424 	}
425 
426 	IP6_STATINC(IP6_STAT_LOCALOUT);
427 
428 	/*
429 	 * Route packet.
430 	 */
431 	/* initialize cached route */
432 	if (ro == NULL) {
433 		ro = &ip6route;
434 	}
435 	ro_pmtu = ro;
436 	if (opt && opt->ip6po_rthdr)
437 		ro = &opt->ip6po_route;
438 
439  	/*
440 	 * if specified, try to fill in the traffic class field.
441 	 * do not override if a non-zero value is already set.
442 	 * we check the diffserv field and the ecn field separately.
443 	 */
444 	if (opt && opt->ip6po_tclass >= 0) {
445 		int mask = 0;
446 
447 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
448 			mask |= 0xfc;
449 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
450 			mask |= 0x03;
451 		if (mask != 0)
452 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
453 	}
454 
455 	/* fill in or override the hop limit field, if necessary. */
456 	if (opt && opt->ip6po_hlim != -1)
457 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
458 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
459 		if (im6o != NULL)
460 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
461 		else
462 			ip6->ip6_hlim = ip6_defmcasthlim;
463 	}
464 
465 #ifdef IPSEC
466 	if (needipsec) {
467 		s = splsoftnet();
468 		error = ipsec6_process_packet(m,sp->req);
469 
470 		/*
471 		 * Preserve KAME behaviour: ENOENT can be returned
472 		 * when an SA acquire is in progress.  Don't propagate
473 		 * this to user-level; it confuses applications.
474 		 * XXX this will go away when the SADB is redone.
475 		 */
476 		if (error == ENOENT)
477 			error = 0;
478 		splx(s);
479 		goto done;
480 	}
481 #endif /* IPSEC */
482 
483 
484 
485 	/* adjust pointer */
486 	ip6 = mtod(m, struct ip6_hdr *);
487 
488 	sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
489 	if ((error = in6_selectroute(&dst_sa, opt, im6o, ro,
490 	    &ifp, &rt, 0)) != 0) {
491 		if (ifp != NULL)
492 			in6_ifstat_inc(ifp, ifs6_out_discard);
493 		goto bad;
494 	}
495 	if (rt == NULL) {
496 		/*
497 		 * If in6_selectroute() does not return a route entry,
498 		 * dst may not have been updated.
499 		 */
500 		error = rtcache_setdst(ro, sin6tosa(&dst_sa));
501 		if (error) {
502 			goto bad;
503 		}
504 	}
505 
506 	/*
507 	 * then rt (for unicast) and ifp must be non-NULL valid values.
508 	 */
509 	if ((flags & IPV6_FORWARDING) == 0) {
510 		/* XXX: the FORWARDING flag can be set for mrouting. */
511 		in6_ifstat_inc(ifp, ifs6_out_request);
512 	}
513 	if (rt != NULL) {
514 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
515 		rt->rt_use++;
516 	}
517 
518 	/*
519 	 * The outgoing interface must be in the zone of source and
520 	 * destination addresses.  We should use ia_ifp to support the
521 	 * case of sending packets to an address of our own.
522 	 */
523 	if (ia != NULL && ia->ia_ifp)
524 		origifp = ia->ia_ifp;
525 	else
526 		origifp = ifp;
527 
528 	src0 = ip6->ip6_src;
529 	if (in6_setscope(&src0, origifp, &zone))
530 		goto badscope;
531 	sockaddr_in6_init(&src_sa, &ip6->ip6_src, 0, 0, 0);
532 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
533 		goto badscope;
534 
535 	dst0 = ip6->ip6_dst;
536 	if (in6_setscope(&dst0, origifp, &zone))
537 		goto badscope;
538 	/* re-initialize to be sure */
539 	sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
540 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id)
541 		goto badscope;
542 
543 	/* scope check is done. */
544 
545 	if (rt == NULL || IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
546 		if (dst == NULL)
547 			dst = satocsin6(rtcache_getdst(ro));
548 		KASSERT(dst != NULL);
549 	} else if (opt && rtcache_validate(&opt->ip6po_nextroute) != NULL) {
550 		/*
551 		 * The nexthop is explicitly specified by the
552 		 * application.  We assume the next hop is an IPv6
553 		 * address.
554 		 */
555 		dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
556 	} else if ((rt->rt_flags & RTF_GATEWAY))
557 		dst = (struct sockaddr_in6 *)rt->rt_gateway;
558 	else if (dst == NULL)
559 		dst = satocsin6(rtcache_getdst(ro));
560 
561 	/*
562 	 * XXXXXX: original code follows:
563 	 */
564 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
565 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
566 	else {
567 		struct	in6_multi *in6m;
568 
569 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
570 
571 		in6_ifstat_inc(ifp, ifs6_out_mcast);
572 
573 		/*
574 		 * Confirm that the outgoing interface supports multicast.
575 		 */
576 		if (!(ifp->if_flags & IFF_MULTICAST)) {
577 			IP6_STATINC(IP6_STAT_NOROUTE);
578 			in6_ifstat_inc(ifp, ifs6_out_discard);
579 			error = ENETUNREACH;
580 			goto bad;
581 		}
582 
583 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
584 		if (in6m != NULL &&
585 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
586 			/*
587 			 * If we belong to the destination multicast group
588 			 * on the outgoing interface, and the caller did not
589 			 * forbid loopback, loop back a copy.
590 			 */
591 			KASSERT(dst != NULL);
592 			ip6_mloopback(ifp, m, dst);
593 		} else {
594 			/*
595 			 * If we are acting as a multicast router, perform
596 			 * multicast forwarding as if the packet had just
597 			 * arrived on the interface to which we are about
598 			 * to send.  The multicast forwarding function
599 			 * recursively calls this function, using the
600 			 * IPV6_FORWARDING flag to prevent infinite recursion.
601 			 *
602 			 * Multicasts that are looped back by ip6_mloopback(),
603 			 * above, will be forwarded by the ip6_input() routine,
604 			 * if necessary.
605 			 */
606 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
607 				if (ip6_mforward(ip6, ifp, m) != 0) {
608 					m_freem(m);
609 					goto done;
610 				}
611 			}
612 		}
613 		/*
614 		 * Multicasts with a hoplimit of zero may be looped back,
615 		 * above, but must not be transmitted on a network.
616 		 * Also, multicasts addressed to the loopback interface
617 		 * are not sent -- the above call to ip6_mloopback() will
618 		 * loop back a copy if this host actually belongs to the
619 		 * destination group on the loopback interface.
620 		 */
621 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
622 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
623 			m_freem(m);
624 			goto done;
625 		}
626 	}
627 
628 	/*
629 	 * Fill the outgoing inteface to tell the upper layer
630 	 * to increment per-interface statistics.
631 	 */
632 	if (ifpp)
633 		*ifpp = ifp;
634 
635 	/* Determine path MTU. */
636 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
637 	    &alwaysfrag)) != 0)
638 		goto bad;
639 
640 	/*
641 	 * The caller of this function may specify to use the minimum MTU
642 	 * in some cases.
643 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
644 	 * setting.  The logic is a bit complicated; by default, unicast
645 	 * packets will follow path MTU while multicast packets will be sent at
646 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
647 	 * including unicast ones will be sent at the minimum MTU.  Multicast
648 	 * packets will always be sent at the minimum MTU unless
649 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
650 	 * See RFC 3542 for more details.
651 	 */
652 	if (mtu > IPV6_MMTU) {
653 		if ((flags & IPV6_MINMTU))
654 			mtu = IPV6_MMTU;
655 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
656 			mtu = IPV6_MMTU;
657 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
658 			 (opt == NULL ||
659 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
660 			mtu = IPV6_MMTU;
661 		}
662 	}
663 
664 	/*
665 	 * clear embedded scope identifiers if necessary.
666 	 * in6_clearscope will touch the addresses only when necessary.
667 	 */
668 	in6_clearscope(&ip6->ip6_src);
669 	in6_clearscope(&ip6->ip6_dst);
670 
671 	/*
672 	 * If the outgoing packet contains a hop-by-hop options header,
673 	 * it must be examined and processed even by the source node.
674 	 * (RFC 2460, section 4.)
675 	 */
676 	if (ip6->ip6_nxt == IPV6_HOPOPTS) {
677 		u_int32_t dummy1; /* XXX unused */
678 		u_int32_t dummy2; /* XXX unused */
679 		int hoff = sizeof(struct ip6_hdr);
680 
681 		if (ip6_hopopts_input(&dummy1, &dummy2, &m, &hoff)) {
682 			/* m was already freed at this point */
683 			error = EINVAL;/* better error? */
684 			goto done;
685 		}
686 
687 		ip6 = mtod(m, struct ip6_hdr *);
688 	}
689 
690 	/*
691 	 * Run through list of hooks for output packets.
692 	 */
693 	if ((error = pfil_run_hooks(inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0)
694 		goto done;
695 	if (m == NULL)
696 		goto done;
697 	ip6 = mtod(m, struct ip6_hdr *);
698 
699 	/*
700 	 * Send the packet to the outgoing interface.
701 	 * If necessary, do IPv6 fragmentation before sending.
702 	 *
703 	 * the logic here is rather complex:
704 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
705 	 * 1-a:	send as is if tlen <= path mtu
706 	 * 1-b:	fragment if tlen > path mtu
707 	 *
708 	 * 2: if user asks us not to fragment (dontfrag == 1)
709 	 * 2-a:	send as is if tlen <= interface mtu
710 	 * 2-b:	error if tlen > interface mtu
711 	 *
712 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
713 	 *	always fragment
714 	 *
715 	 * 4: if dontfrag == 1 && alwaysfrag == 1
716 	 *	error, as we cannot handle this conflicting request
717 	 */
718 	tlen = m->m_pkthdr.len;
719 	tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0;
720 	if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
721 		dontfrag = 1;
722 	else
723 		dontfrag = 0;
724 
725 	if (dontfrag && alwaysfrag) {	/* case 4 */
726 		/* conflicting request - can't transmit */
727 		error = EMSGSIZE;
728 		goto bad;
729 	}
730 	if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) {	/* case 2-b */
731 		/*
732 		 * Even if the DONTFRAG option is specified, we cannot send the
733 		 * packet when the data length is larger than the MTU of the
734 		 * outgoing interface.
735 		 * Notify the error by sending IPV6_PATHMTU ancillary data as
736 		 * well as returning an error code (the latter is not described
737 		 * in the API spec.)
738 		 */
739 		u_int32_t mtu32;
740 		struct ip6ctlparam ip6cp;
741 
742 		mtu32 = (u_int32_t)mtu;
743 		memset(&ip6cp, 0, sizeof(ip6cp));
744 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
745 		pfctlinput2(PRC_MSGSIZE,
746 		    rtcache_getdst(ro_pmtu), &ip6cp);
747 
748 		error = EMSGSIZE;
749 		goto bad;
750 	}
751 
752 	/*
753 	 * transmit packet without fragmentation
754 	 */
755 	if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) {
756 		/* case 1-a and 2-a */
757 		struct in6_ifaddr *ia6;
758 		int sw_csum;
759 
760 		ip6 = mtod(m, struct ip6_hdr *);
761 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
762 		if (ia6) {
763 			/* Record statistics for this interface address. */
764 			ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len;
765 		}
766 
767 		sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
768 		if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
769 			if (IN6_NEED_CHECKSUM(ifp,
770 			    sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) {
771 				in6_delayed_cksum(m);
772 			}
773 			m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
774 		}
775 
776 		KASSERT(dst != NULL);
777 		if (__predict_true(!tso ||
778 		    (ifp->if_capenable & IFCAP_TSOv6) != 0)) {
779 			error = nd6_output(ifp, origifp, m, dst, rt);
780 		} else {
781 			error = ip6_tso_output(ifp, origifp, m, dst, rt);
782 		}
783 		goto done;
784 	}
785 
786 	if (tso) {
787 		error = EINVAL; /* XXX */
788 		goto bad;
789 	}
790 
791 	/*
792 	 * try to fragment the packet.  case 1-b and 3
793 	 */
794 	if (mtu < IPV6_MMTU) {
795 		/* path MTU cannot be less than IPV6_MMTU */
796 		error = EMSGSIZE;
797 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
798 		goto bad;
799 	} else if (ip6->ip6_plen == 0) {
800 		/* jumbo payload cannot be fragmented */
801 		error = EMSGSIZE;
802 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
803 		goto bad;
804 	} else {
805 		struct mbuf **mnext, *m_frgpart;
806 		struct ip6_frag *ip6f;
807 		u_int32_t id = htonl(ip6_randomid());
808 		u_char nextproto;
809 #if 0				/* see below */
810 		struct ip6ctlparam ip6cp;
811 		u_int32_t mtu32;
812 #endif
813 
814 		/*
815 		 * Too large for the destination or interface;
816 		 * fragment if possible.
817 		 * Must be able to put at least 8 bytes per fragment.
818 		 */
819 		hlen = unfragpartlen;
820 		if (mtu > IPV6_MAXPACKET)
821 			mtu = IPV6_MAXPACKET;
822 
823 #if 0
824 		/*
825 		 * It is believed this code is a leftover from the
826 		 * development of the IPV6_RECVPATHMTU sockopt and
827 		 * associated work to implement RFC3542.
828 		 * It's not entirely clear what the intent of the API
829 		 * is at this point, so disable this code for now.
830 		 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG
831 		 * will send notifications if the application requests.
832 		 */
833 
834 		/* Notify a proper path MTU to applications. */
835 		mtu32 = (u_int32_t)mtu;
836 		memset(&ip6cp, 0, sizeof(ip6cp));
837 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
838 		pfctlinput2(PRC_MSGSIZE,
839 		    rtcache_getdst(ro_pmtu), &ip6cp);
840 #endif
841 
842 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
843 		if (len < 8) {
844 			error = EMSGSIZE;
845 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
846 			goto bad;
847 		}
848 
849 		mnext = &m->m_nextpkt;
850 
851 		/*
852 		 * Change the next header field of the last header in the
853 		 * unfragmentable part.
854 		 */
855 		if (exthdrs.ip6e_rthdr) {
856 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
857 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
858 		} else if (exthdrs.ip6e_dest1) {
859 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
860 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
861 		} else if (exthdrs.ip6e_hbh) {
862 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
863 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
864 		} else {
865 			nextproto = ip6->ip6_nxt;
866 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
867 		}
868 
869 		if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6))
870 		    != 0) {
871 			if (IN6_NEED_CHECKSUM(ifp,
872 			    m->m_pkthdr.csum_flags &
873 			    (M_CSUM_UDPv6|M_CSUM_TCPv6))) {
874 				in6_delayed_cksum(m);
875 			}
876 			m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
877 		}
878 
879 		/*
880 		 * Loop through length of segment after first fragment,
881 		 * make new header and copy data of each part and link onto
882 		 * chain.
883 		 */
884 		m0 = m;
885 		for (off = hlen; off < tlen; off += len) {
886 			struct mbuf *mlast;
887 
888 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
889 			if (!m) {
890 				error = ENOBUFS;
891 				IP6_STATINC(IP6_STAT_ODROPPED);
892 				goto sendorfree;
893 			}
894 			m->m_pkthdr.rcvif = NULL;
895 			m->m_flags = m0->m_flags & M_COPYFLAGS;
896 			*mnext = m;
897 			mnext = &m->m_nextpkt;
898 			m->m_data += max_linkhdr;
899 			mhip6 = mtod(m, struct ip6_hdr *);
900 			*mhip6 = *ip6;
901 			m->m_len = sizeof(*mhip6);
902 			/*
903 			 * ip6f must be valid if error is 0.  But how
904 			 * can a compiler be expected to infer this?
905 			 */
906 			ip6f = NULL;
907 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
908 			if (error) {
909 				IP6_STATINC(IP6_STAT_ODROPPED);
910 				goto sendorfree;
911 			}
912 			ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7));
913 			if (off + len >= tlen)
914 				len = tlen - off;
915 			else
916 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
917 			mhip6->ip6_plen = htons((u_int16_t)(len + hlen +
918 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
919 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
920 				error = ENOBUFS;
921 				IP6_STATINC(IP6_STAT_ODROPPED);
922 				goto sendorfree;
923 			}
924 			for (mlast = m; mlast->m_next; mlast = mlast->m_next)
925 				;
926 			mlast->m_next = m_frgpart;
927 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
928 			m->m_pkthdr.rcvif = NULL;
929 			ip6f->ip6f_reserved = 0;
930 			ip6f->ip6f_ident = id;
931 			ip6f->ip6f_nxt = nextproto;
932 			IP6_STATINC(IP6_STAT_OFRAGMENTS);
933 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
934 		}
935 
936 		in6_ifstat_inc(ifp, ifs6_out_fragok);
937 	}
938 
939 	/*
940 	 * Remove leading garbages.
941 	 */
942 sendorfree:
943 	m = m0->m_nextpkt;
944 	m0->m_nextpkt = 0;
945 	m_freem(m0);
946 	for (m0 = m; m; m = m0) {
947 		m0 = m->m_nextpkt;
948 		m->m_nextpkt = 0;
949 		if (error == 0) {
950 			struct in6_ifaddr *ia6;
951 			ip6 = mtod(m, struct ip6_hdr *);
952 			ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
953 			if (ia6) {
954 				/*
955 				 * Record statistics for this interface
956 				 * address.
957 				 */
958 				ia6->ia_ifa.ifa_data.ifad_outbytes +=
959 				    m->m_pkthdr.len;
960 			}
961 			KASSERT(dst != NULL);
962 			error = nd6_output(ifp, origifp, m, dst, rt);
963 		} else
964 			m_freem(m);
965 	}
966 
967 	if (error == 0)
968 		IP6_STATINC(IP6_STAT_FRAGMENTED);
969 
970 done:
971 	rtcache_free(&ip6route);
972 
973 #ifdef IPSEC
974 	if (sp != NULL)
975 		KEY_FREESP(&sp);
976 #endif /* IPSEC */
977 
978 
979 	return (error);
980 
981 freehdrs:
982 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
983 	m_freem(exthdrs.ip6e_dest1);
984 	m_freem(exthdrs.ip6e_rthdr);
985 	m_freem(exthdrs.ip6e_dest2);
986 	/* FALLTHROUGH */
987 bad:
988 	m_freem(m);
989 	goto done;
990 badscope:
991 	IP6_STATINC(IP6_STAT_BADSCOPE);
992 	in6_ifstat_inc(origifp, ifs6_out_discard);
993 	if (error == 0)
994 		error = EHOSTUNREACH; /* XXX */
995 	goto bad;
996 }
997 
998 static int
999 ip6_copyexthdr(struct mbuf **mp, void *hdr, int hlen)
1000 {
1001 	struct mbuf *m;
1002 
1003 	if (hlen > MCLBYTES)
1004 		return (ENOBUFS); /* XXX */
1005 
1006 	MGET(m, M_DONTWAIT, MT_DATA);
1007 	if (!m)
1008 		return (ENOBUFS);
1009 
1010 	if (hlen > MLEN) {
1011 		MCLGET(m, M_DONTWAIT);
1012 		if ((m->m_flags & M_EXT) == 0) {
1013 			m_free(m);
1014 			return (ENOBUFS);
1015 		}
1016 	}
1017 	m->m_len = hlen;
1018 	if (hdr)
1019 		bcopy(hdr, mtod(m, void *), hlen);
1020 
1021 	*mp = m;
1022 	return (0);
1023 }
1024 
1025 /*
1026  * Process a delayed payload checksum calculation.
1027  */
1028 void
1029 in6_delayed_cksum(struct mbuf *m)
1030 {
1031 	uint16_t csum, offset;
1032 
1033 	KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0);
1034 	KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0);
1035 	KASSERT((m->m_pkthdr.csum_flags
1036 	    & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0);
1037 
1038 	offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data);
1039 	csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset);
1040 	if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) {
1041 		csum = 0xffff;
1042 	}
1043 
1044 	offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data);
1045 	if ((offset + sizeof(csum)) > m->m_len) {
1046 		m_copyback(m, offset, sizeof(csum), &csum);
1047 	} else {
1048 		*(uint16_t *)(mtod(m, char *) + offset) = csum;
1049 	}
1050 }
1051 
1052 /*
1053  * Insert jumbo payload option.
1054  */
1055 static int
1056 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1057 {
1058 	struct mbuf *mopt;
1059 	u_int8_t *optbuf;
1060 	u_int32_t v;
1061 
1062 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1063 
1064 	/*
1065 	 * If there is no hop-by-hop options header, allocate new one.
1066 	 * If there is one but it doesn't have enough space to store the
1067 	 * jumbo payload option, allocate a cluster to store the whole options.
1068 	 * Otherwise, use it to store the options.
1069 	 */
1070 	if (exthdrs->ip6e_hbh == 0) {
1071 		MGET(mopt, M_DONTWAIT, MT_DATA);
1072 		if (mopt == 0)
1073 			return (ENOBUFS);
1074 		mopt->m_len = JUMBOOPTLEN;
1075 		optbuf = mtod(mopt, u_int8_t *);
1076 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1077 		exthdrs->ip6e_hbh = mopt;
1078 	} else {
1079 		struct ip6_hbh *hbh;
1080 
1081 		mopt = exthdrs->ip6e_hbh;
1082 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1083 			/*
1084 			 * XXX assumption:
1085 			 * - exthdrs->ip6e_hbh is not referenced from places
1086 			 *   other than exthdrs.
1087 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1088 			 */
1089 			int oldoptlen = mopt->m_len;
1090 			struct mbuf *n;
1091 
1092 			/*
1093 			 * XXX: give up if the whole (new) hbh header does
1094 			 * not fit even in an mbuf cluster.
1095 			 */
1096 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1097 				return (ENOBUFS);
1098 
1099 			/*
1100 			 * As a consequence, we must always prepare a cluster
1101 			 * at this point.
1102 			 */
1103 			MGET(n, M_DONTWAIT, MT_DATA);
1104 			if (n) {
1105 				MCLGET(n, M_DONTWAIT);
1106 				if ((n->m_flags & M_EXT) == 0) {
1107 					m_freem(n);
1108 					n = NULL;
1109 				}
1110 			}
1111 			if (!n)
1112 				return (ENOBUFS);
1113 			n->m_len = oldoptlen + JUMBOOPTLEN;
1114 			bcopy(mtod(mopt, void *), mtod(n, void *),
1115 			    oldoptlen);
1116 			optbuf = mtod(n, u_int8_t *) + oldoptlen;
1117 			m_freem(mopt);
1118 			mopt = exthdrs->ip6e_hbh = n;
1119 		} else {
1120 			optbuf = mtod(mopt, u_int8_t *) + mopt->m_len;
1121 			mopt->m_len += JUMBOOPTLEN;
1122 		}
1123 		optbuf[0] = IP6OPT_PADN;
1124 		optbuf[1] = 0;
1125 
1126 		/*
1127 		 * Adjust the header length according to the pad and
1128 		 * the jumbo payload option.
1129 		 */
1130 		hbh = mtod(mopt, struct ip6_hbh *);
1131 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1132 	}
1133 
1134 	/* fill in the option. */
1135 	optbuf[2] = IP6OPT_JUMBO;
1136 	optbuf[3] = 4;
1137 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1138 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1139 
1140 	/* finally, adjust the packet header length */
1141 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1142 
1143 	return (0);
1144 #undef JUMBOOPTLEN
1145 }
1146 
1147 /*
1148  * Insert fragment header and copy unfragmentable header portions.
1149  *
1150  * *frghdrp will not be read, and it is guaranteed that either an
1151  * error is returned or that *frghdrp will point to space allocated
1152  * for the fragment header.
1153  */
1154 static int
1155 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1156 	struct ip6_frag **frghdrp)
1157 {
1158 	struct mbuf *n, *mlast;
1159 
1160 	if (hlen > sizeof(struct ip6_hdr)) {
1161 		n = m_copym(m0, sizeof(struct ip6_hdr),
1162 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1163 		if (n == 0)
1164 			return (ENOBUFS);
1165 		m->m_next = n;
1166 	} else
1167 		n = m;
1168 
1169 	/* Search for the last mbuf of unfragmentable part. */
1170 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1171 		;
1172 
1173 	if ((mlast->m_flags & M_EXT) == 0 &&
1174 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1175 		/* use the trailing space of the last mbuf for the fragment hdr */
1176 		*frghdrp = (struct ip6_frag *)(mtod(mlast, char *) +
1177 		    mlast->m_len);
1178 		mlast->m_len += sizeof(struct ip6_frag);
1179 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1180 	} else {
1181 		/* allocate a new mbuf for the fragment header */
1182 		struct mbuf *mfrg;
1183 
1184 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1185 		if (mfrg == 0)
1186 			return (ENOBUFS);
1187 		mfrg->m_len = sizeof(struct ip6_frag);
1188 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1189 		mlast->m_next = mfrg;
1190 	}
1191 
1192 	return (0);
1193 }
1194 
1195 static int
1196 ip6_getpmtu(struct route *ro_pmtu, struct route *ro, struct ifnet *ifp,
1197     const struct in6_addr *dst, u_long *mtup, int *alwaysfragp)
1198 {
1199 	struct rtentry *rt;
1200 	u_int32_t mtu = 0;
1201 	int alwaysfrag = 0;
1202 	int error = 0;
1203 
1204 	if (ro_pmtu != ro) {
1205 		union {
1206 			struct sockaddr		dst;
1207 			struct sockaddr_in6	dst6;
1208 		} u;
1209 
1210 		/* The first hop and the final destination may differ. */
1211 		sockaddr_in6_init(&u.dst6, dst, 0, 0, 0);
1212 		rt = rtcache_lookup(ro_pmtu, &u.dst);
1213 	} else
1214 		rt = rtcache_validate(ro_pmtu);
1215 	if (rt != NULL) {
1216 		u_int32_t ifmtu;
1217 
1218 		if (ifp == NULL)
1219 			ifp = rt->rt_ifp;
1220 		ifmtu = IN6_LINKMTU(ifp);
1221 		mtu = rt->rt_rmx.rmx_mtu;
1222 		if (mtu == 0)
1223 			mtu = ifmtu;
1224 		else if (mtu < IPV6_MMTU) {
1225 			/*
1226 			 * RFC2460 section 5, last paragraph:
1227 			 * if we record ICMPv6 too big message with
1228 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1229 			 * or smaller, with fragment header attached.
1230 			 * (fragment header is needed regardless from the
1231 			 * packet size, for translators to identify packets)
1232 			 */
1233 			alwaysfrag = 1;
1234 			mtu = IPV6_MMTU;
1235 		} else if (mtu > ifmtu) {
1236 			/*
1237 			 * The MTU on the route is larger than the MTU on
1238 			 * the interface!  This shouldn't happen, unless the
1239 			 * MTU of the interface has been changed after the
1240 			 * interface was brought up.  Change the MTU in the
1241 			 * route to match the interface MTU (as long as the
1242 			 * field isn't locked).
1243 			 */
1244 			mtu = ifmtu;
1245 			if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
1246 				rt->rt_rmx.rmx_mtu = mtu;
1247 		}
1248 	} else if (ifp) {
1249 		mtu = IN6_LINKMTU(ifp);
1250 	} else
1251 		error = EHOSTUNREACH; /* XXX */
1252 
1253 	*mtup = mtu;
1254 	if (alwaysfragp)
1255 		*alwaysfragp = alwaysfrag;
1256 	return (error);
1257 }
1258 
1259 /*
1260  * IP6 socket option processing.
1261  */
1262 int
1263 ip6_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1264 {
1265 	int optdatalen, uproto;
1266 	void *optdata;
1267 	struct in6pcb *in6p = sotoin6pcb(so);
1268 	int error, optval;
1269 	int level, optname;
1270 
1271 	KASSERT(sopt != NULL);
1272 
1273 	level = sopt->sopt_level;
1274 	optname = sopt->sopt_name;
1275 
1276 	error = optval = 0;
1277 	uproto = (int)so->so_proto->pr_protocol;
1278 
1279 	if (level != IPPROTO_IPV6) {
1280 		return ENOPROTOOPT;
1281 	}
1282 	switch (op) {
1283 	case PRCO_SETOPT:
1284 		switch (optname) {
1285 #ifdef RFC2292
1286 		case IPV6_2292PKTOPTIONS:
1287 			error = ip6_pcbopts(&in6p->in6p_outputopts, so, sopt);
1288 			break;
1289 #endif
1290 
1291 		/*
1292 		 * Use of some Hop-by-Hop options or some
1293 		 * Destination options, might require special
1294 		 * privilege.  That is, normal applications
1295 		 * (without special privilege) might be forbidden
1296 		 * from setting certain options in outgoing packets,
1297 		 * and might never see certain options in received
1298 		 * packets. [RFC 2292 Section 6]
1299 		 * KAME specific note:
1300 		 *  KAME prevents non-privileged users from sending or
1301 		 *  receiving ANY hbh/dst options in order to avoid
1302 		 *  overhead of parsing options in the kernel.
1303 		 */
1304 		case IPV6_RECVHOPOPTS:
1305 		case IPV6_RECVDSTOPTS:
1306 		case IPV6_RECVRTHDRDSTOPTS:
1307 			error = kauth_authorize_network(kauth_cred_get(),
1308 			    KAUTH_NETWORK_IPV6, KAUTH_REQ_NETWORK_IPV6_HOPBYHOP,
1309 			    NULL, NULL, NULL);
1310 			if (error)
1311 				break;
1312 			/* FALLTHROUGH */
1313 		case IPV6_UNICAST_HOPS:
1314 		case IPV6_HOPLIMIT:
1315 		case IPV6_FAITH:
1316 
1317 		case IPV6_RECVPKTINFO:
1318 		case IPV6_RECVHOPLIMIT:
1319 		case IPV6_RECVRTHDR:
1320 		case IPV6_RECVPATHMTU:
1321 		case IPV6_RECVTCLASS:
1322 		case IPV6_V6ONLY:
1323 			error = sockopt_getint(sopt, &optval);
1324 			if (error)
1325 				break;
1326 			switch (optname) {
1327 			case IPV6_UNICAST_HOPS:
1328 				if (optval < -1 || optval >= 256)
1329 					error = EINVAL;
1330 				else {
1331 					/* -1 = kernel default */
1332 					in6p->in6p_hops = optval;
1333 				}
1334 				break;
1335 #define OPTSET(bit) \
1336 do { \
1337 if (optval) \
1338 	in6p->in6p_flags |= (bit); \
1339 else \
1340 	in6p->in6p_flags &= ~(bit); \
1341 } while (/*CONSTCOND*/ 0)
1342 
1343 #ifdef RFC2292
1344 #define OPTSET2292(bit) 			\
1345 do { 						\
1346 in6p->in6p_flags |= IN6P_RFC2292; 	\
1347 if (optval) 				\
1348 	in6p->in6p_flags |= (bit); 	\
1349 else 					\
1350 	in6p->in6p_flags &= ~(bit); 	\
1351 } while (/*CONSTCOND*/ 0)
1352 #endif
1353 
1354 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1355 
1356 			case IPV6_RECVPKTINFO:
1357 #ifdef RFC2292
1358 				/* cannot mix with RFC2292 */
1359 				if (OPTBIT(IN6P_RFC2292)) {
1360 					error = EINVAL;
1361 					break;
1362 				}
1363 #endif
1364 				OPTSET(IN6P_PKTINFO);
1365 				break;
1366 
1367 			case IPV6_HOPLIMIT:
1368 			{
1369 				struct ip6_pktopts **optp;
1370 
1371 #ifdef RFC2292
1372 				/* cannot mix with RFC2292 */
1373 				if (OPTBIT(IN6P_RFC2292)) {
1374 					error = EINVAL;
1375 					break;
1376 				}
1377 #endif
1378 				optp = &in6p->in6p_outputopts;
1379 				error = ip6_pcbopt(IPV6_HOPLIMIT,
1380 						   (u_char *)&optval,
1381 						   sizeof(optval),
1382 						   optp,
1383 						   kauth_cred_get(), uproto);
1384 				break;
1385 			}
1386 
1387 			case IPV6_RECVHOPLIMIT:
1388 #ifdef RFC2292
1389 				/* cannot mix with RFC2292 */
1390 				if (OPTBIT(IN6P_RFC2292)) {
1391 					error = EINVAL;
1392 					break;
1393 				}
1394 #endif
1395 				OPTSET(IN6P_HOPLIMIT);
1396 				break;
1397 
1398 			case IPV6_RECVHOPOPTS:
1399 #ifdef RFC2292
1400 				/* cannot mix with RFC2292 */
1401 				if (OPTBIT(IN6P_RFC2292)) {
1402 					error = EINVAL;
1403 					break;
1404 				}
1405 #endif
1406 				OPTSET(IN6P_HOPOPTS);
1407 				break;
1408 
1409 			case IPV6_RECVDSTOPTS:
1410 #ifdef RFC2292
1411 				/* cannot mix with RFC2292 */
1412 				if (OPTBIT(IN6P_RFC2292)) {
1413 					error = EINVAL;
1414 					break;
1415 				}
1416 #endif
1417 				OPTSET(IN6P_DSTOPTS);
1418 				break;
1419 
1420 			case IPV6_RECVRTHDRDSTOPTS:
1421 #ifdef RFC2292
1422 				/* cannot mix with RFC2292 */
1423 				if (OPTBIT(IN6P_RFC2292)) {
1424 					error = EINVAL;
1425 					break;
1426 				}
1427 #endif
1428 				OPTSET(IN6P_RTHDRDSTOPTS);
1429 				break;
1430 
1431 			case IPV6_RECVRTHDR:
1432 #ifdef RFC2292
1433 				/* cannot mix with RFC2292 */
1434 				if (OPTBIT(IN6P_RFC2292)) {
1435 					error = EINVAL;
1436 					break;
1437 				}
1438 #endif
1439 				OPTSET(IN6P_RTHDR);
1440 				break;
1441 
1442 			case IPV6_FAITH:
1443 				OPTSET(IN6P_FAITH);
1444 				break;
1445 
1446 			case IPV6_RECVPATHMTU:
1447 				/*
1448 				 * We ignore this option for TCP
1449 				 * sockets.
1450 				 * (RFC3542 leaves this case
1451 				 * unspecified.)
1452 				 */
1453 				if (uproto != IPPROTO_TCP)
1454 					OPTSET(IN6P_MTU);
1455 				break;
1456 
1457 			case IPV6_V6ONLY:
1458 				/*
1459 				 * make setsockopt(IPV6_V6ONLY)
1460 				 * available only prior to bind(2).
1461 				 * see ipng mailing list, Jun 22 2001.
1462 				 */
1463 				if (in6p->in6p_lport ||
1464 				    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1465 					error = EINVAL;
1466 					break;
1467 				}
1468 #ifdef INET6_BINDV6ONLY
1469 				if (!optval)
1470 					error = EINVAL;
1471 #else
1472 				OPTSET(IN6P_IPV6_V6ONLY);
1473 #endif
1474 				break;
1475 			case IPV6_RECVTCLASS:
1476 #ifdef RFC2292
1477 				/* cannot mix with RFC2292 XXX */
1478 				if (OPTBIT(IN6P_RFC2292)) {
1479 					error = EINVAL;
1480 					break;
1481 				}
1482 #endif
1483 				OPTSET(IN6P_TCLASS);
1484 				break;
1485 
1486 			}
1487 			break;
1488 
1489 		case IPV6_OTCLASS:
1490 		{
1491 			struct ip6_pktopts **optp;
1492 			u_int8_t tclass;
1493 
1494 			error = sockopt_get(sopt, &tclass, sizeof(tclass));
1495 			if (error)
1496 				break;
1497 			optp = &in6p->in6p_outputopts;
1498 			error = ip6_pcbopt(optname,
1499 					   (u_char *)&tclass,
1500 					   sizeof(tclass),
1501 					   optp,
1502 					   kauth_cred_get(), uproto);
1503 			break;
1504 		}
1505 
1506 		case IPV6_TCLASS:
1507 		case IPV6_DONTFRAG:
1508 		case IPV6_USE_MIN_MTU:
1509 			error = sockopt_getint(sopt, &optval);
1510 			if (error)
1511 				break;
1512 			{
1513 				struct ip6_pktopts **optp;
1514 				optp = &in6p->in6p_outputopts;
1515 				error = ip6_pcbopt(optname,
1516 						   (u_char *)&optval,
1517 						   sizeof(optval),
1518 						   optp,
1519 						   kauth_cred_get(), uproto);
1520 				break;
1521 			}
1522 
1523 #ifdef RFC2292
1524 		case IPV6_2292PKTINFO:
1525 		case IPV6_2292HOPLIMIT:
1526 		case IPV6_2292HOPOPTS:
1527 		case IPV6_2292DSTOPTS:
1528 		case IPV6_2292RTHDR:
1529 			/* RFC 2292 */
1530 			error = sockopt_getint(sopt, &optval);
1531 			if (error)
1532 				break;
1533 
1534 			switch (optname) {
1535 			case IPV6_2292PKTINFO:
1536 				OPTSET2292(IN6P_PKTINFO);
1537 				break;
1538 			case IPV6_2292HOPLIMIT:
1539 				OPTSET2292(IN6P_HOPLIMIT);
1540 				break;
1541 			case IPV6_2292HOPOPTS:
1542 				/*
1543 				 * Check super-user privilege.
1544 				 * See comments for IPV6_RECVHOPOPTS.
1545 				 */
1546 				error =
1547 				    kauth_authorize_network(kauth_cred_get(),
1548 				    KAUTH_NETWORK_IPV6,
1549 				    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL,
1550 				    NULL, NULL);
1551 				if (error)
1552 					return (error);
1553 				OPTSET2292(IN6P_HOPOPTS);
1554 				break;
1555 			case IPV6_2292DSTOPTS:
1556 				error =
1557 				    kauth_authorize_network(kauth_cred_get(),
1558 				    KAUTH_NETWORK_IPV6,
1559 				    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL,
1560 				    NULL, NULL);
1561 				if (error)
1562 					return (error);
1563 				OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1564 				break;
1565 			case IPV6_2292RTHDR:
1566 				OPTSET2292(IN6P_RTHDR);
1567 				break;
1568 			}
1569 			break;
1570 #endif
1571 		case IPV6_PKTINFO:
1572 		case IPV6_HOPOPTS:
1573 		case IPV6_RTHDR:
1574 		case IPV6_DSTOPTS:
1575 		case IPV6_RTHDRDSTOPTS:
1576 		case IPV6_NEXTHOP: {
1577 			/* new advanced API (RFC3542) */
1578 			void *optbuf;
1579 			int optbuflen;
1580 			struct ip6_pktopts **optp;
1581 
1582 #ifdef RFC2292
1583 			/* cannot mix with RFC2292 */
1584 			if (OPTBIT(IN6P_RFC2292)) {
1585 				error = EINVAL;
1586 				break;
1587 			}
1588 #endif
1589 
1590 			optbuflen = sopt->sopt_size;
1591 			optbuf = malloc(optbuflen, M_IP6OPT, M_NOWAIT);
1592 			if (optbuf == NULL) {
1593 				error = ENOBUFS;
1594 				break;
1595 			}
1596 
1597 			sockopt_get(sopt, optbuf, optbuflen);
1598 			optp = &in6p->in6p_outputopts;
1599 			error = ip6_pcbopt(optname, optbuf, optbuflen,
1600 			    optp, kauth_cred_get(), uproto);
1601 			break;
1602 			}
1603 #undef OPTSET
1604 
1605 		case IPV6_MULTICAST_IF:
1606 		case IPV6_MULTICAST_HOPS:
1607 		case IPV6_MULTICAST_LOOP:
1608 		case IPV6_JOIN_GROUP:
1609 		case IPV6_LEAVE_GROUP:
1610 			error = ip6_setmoptions(sopt, &in6p->in6p_moptions);
1611 			break;
1612 
1613 		case IPV6_PORTRANGE:
1614 			error = sockopt_getint(sopt, &optval);
1615 			if (error)
1616 				break;
1617 
1618 			switch (optval) {
1619 			case IPV6_PORTRANGE_DEFAULT:
1620 				in6p->in6p_flags &= ~(IN6P_LOWPORT);
1621 				in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1622 				break;
1623 
1624 			case IPV6_PORTRANGE_HIGH:
1625 				in6p->in6p_flags &= ~(IN6P_LOWPORT);
1626 				in6p->in6p_flags |= IN6P_HIGHPORT;
1627 				break;
1628 
1629 			case IPV6_PORTRANGE_LOW:
1630 				in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1631 				in6p->in6p_flags |= IN6P_LOWPORT;
1632 				break;
1633 
1634 			default:
1635 				error = EINVAL;
1636 				break;
1637 			}
1638 			break;
1639 
1640 		case IPV6_PORTALGO:
1641 			error = sockopt_getint(sopt, &optval);
1642 			if (error)
1643 				break;
1644 
1645 			error = portalgo_algo_index_select(
1646 			    (struct inpcb_hdr *)in6p, optval);
1647 			break;
1648 
1649 #if defined(IPSEC)
1650 		case IPV6_IPSEC_POLICY:
1651 			error = ipsec6_set_policy(in6p, optname,
1652 			    sopt->sopt_data, sopt->sopt_size, kauth_cred_get());
1653 			break;
1654 #endif /* IPSEC */
1655 
1656 		default:
1657 			error = ENOPROTOOPT;
1658 			break;
1659 		}
1660 		break;
1661 
1662 	case PRCO_GETOPT:
1663 		switch (optname) {
1664 #ifdef RFC2292
1665 		case IPV6_2292PKTOPTIONS:
1666 			/*
1667 			 * RFC3542 (effectively) deprecated the
1668 			 * semantics of the 2292-style pktoptions.
1669 			 * Since it was not reliable in nature (i.e.,
1670 			 * applications had to expect the lack of some
1671 			 * information after all), it would make sense
1672 			 * to simplify this part by always returning
1673 			 * empty data.
1674 			 */
1675 			break;
1676 #endif
1677 
1678 		case IPV6_RECVHOPOPTS:
1679 		case IPV6_RECVDSTOPTS:
1680 		case IPV6_RECVRTHDRDSTOPTS:
1681 		case IPV6_UNICAST_HOPS:
1682 		case IPV6_RECVPKTINFO:
1683 		case IPV6_RECVHOPLIMIT:
1684 		case IPV6_RECVRTHDR:
1685 		case IPV6_RECVPATHMTU:
1686 
1687 		case IPV6_FAITH:
1688 		case IPV6_V6ONLY:
1689 		case IPV6_PORTRANGE:
1690 		case IPV6_RECVTCLASS:
1691 			switch (optname) {
1692 
1693 			case IPV6_RECVHOPOPTS:
1694 				optval = OPTBIT(IN6P_HOPOPTS);
1695 				break;
1696 
1697 			case IPV6_RECVDSTOPTS:
1698 				optval = OPTBIT(IN6P_DSTOPTS);
1699 				break;
1700 
1701 			case IPV6_RECVRTHDRDSTOPTS:
1702 				optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1703 				break;
1704 
1705 			case IPV6_UNICAST_HOPS:
1706 				optval = in6p->in6p_hops;
1707 				break;
1708 
1709 			case IPV6_RECVPKTINFO:
1710 				optval = OPTBIT(IN6P_PKTINFO);
1711 				break;
1712 
1713 			case IPV6_RECVHOPLIMIT:
1714 				optval = OPTBIT(IN6P_HOPLIMIT);
1715 				break;
1716 
1717 			case IPV6_RECVRTHDR:
1718 				optval = OPTBIT(IN6P_RTHDR);
1719 				break;
1720 
1721 			case IPV6_RECVPATHMTU:
1722 				optval = OPTBIT(IN6P_MTU);
1723 				break;
1724 
1725 			case IPV6_FAITH:
1726 				optval = OPTBIT(IN6P_FAITH);
1727 				break;
1728 
1729 			case IPV6_V6ONLY:
1730 				optval = OPTBIT(IN6P_IPV6_V6ONLY);
1731 				break;
1732 
1733 			case IPV6_PORTRANGE:
1734 			    {
1735 				int flags;
1736 				flags = in6p->in6p_flags;
1737 				if (flags & IN6P_HIGHPORT)
1738 					optval = IPV6_PORTRANGE_HIGH;
1739 				else if (flags & IN6P_LOWPORT)
1740 					optval = IPV6_PORTRANGE_LOW;
1741 				else
1742 					optval = 0;
1743 				break;
1744 			    }
1745 			case IPV6_RECVTCLASS:
1746 				optval = OPTBIT(IN6P_TCLASS);
1747 				break;
1748 
1749 			}
1750 			if (error)
1751 				break;
1752 			error = sockopt_setint(sopt, optval);
1753 			break;
1754 
1755 		case IPV6_PATHMTU:
1756 		    {
1757 			u_long pmtu = 0;
1758 			struct ip6_mtuinfo mtuinfo;
1759 			struct route *ro = &in6p->in6p_route;
1760 
1761 			if (!(so->so_state & SS_ISCONNECTED))
1762 				return (ENOTCONN);
1763 			/*
1764 			 * XXX: we dot not consider the case of source
1765 			 * routing, or optional information to specify
1766 			 * the outgoing interface.
1767 			 */
1768 			error = ip6_getpmtu(ro, NULL, NULL,
1769 			    &in6p->in6p_faddr, &pmtu, NULL);
1770 			if (error)
1771 				break;
1772 			if (pmtu > IPV6_MAXPACKET)
1773 				pmtu = IPV6_MAXPACKET;
1774 
1775 			memset(&mtuinfo, 0, sizeof(mtuinfo));
1776 			mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1777 			optdata = (void *)&mtuinfo;
1778 			optdatalen = sizeof(mtuinfo);
1779 			if (optdatalen > MCLBYTES)
1780 				return (EMSGSIZE); /* XXX */
1781 			error = sockopt_set(sopt, optdata, optdatalen);
1782 			break;
1783 		    }
1784 
1785 #ifdef RFC2292
1786 		case IPV6_2292PKTINFO:
1787 		case IPV6_2292HOPLIMIT:
1788 		case IPV6_2292HOPOPTS:
1789 		case IPV6_2292RTHDR:
1790 		case IPV6_2292DSTOPTS:
1791 			switch (optname) {
1792 			case IPV6_2292PKTINFO:
1793 				optval = OPTBIT(IN6P_PKTINFO);
1794 				break;
1795 			case IPV6_2292HOPLIMIT:
1796 				optval = OPTBIT(IN6P_HOPLIMIT);
1797 				break;
1798 			case IPV6_2292HOPOPTS:
1799 				optval = OPTBIT(IN6P_HOPOPTS);
1800 				break;
1801 			case IPV6_2292RTHDR:
1802 				optval = OPTBIT(IN6P_RTHDR);
1803 				break;
1804 			case IPV6_2292DSTOPTS:
1805 				optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1806 				break;
1807 			}
1808 			error = sockopt_setint(sopt, optval);
1809 			break;
1810 #endif
1811 		case IPV6_PKTINFO:
1812 		case IPV6_HOPOPTS:
1813 		case IPV6_RTHDR:
1814 		case IPV6_DSTOPTS:
1815 		case IPV6_RTHDRDSTOPTS:
1816 		case IPV6_NEXTHOP:
1817 		case IPV6_OTCLASS:
1818 		case IPV6_TCLASS:
1819 		case IPV6_DONTFRAG:
1820 		case IPV6_USE_MIN_MTU:
1821 			error = ip6_getpcbopt(in6p->in6p_outputopts,
1822 			    optname, sopt);
1823 			break;
1824 
1825 		case IPV6_MULTICAST_IF:
1826 		case IPV6_MULTICAST_HOPS:
1827 		case IPV6_MULTICAST_LOOP:
1828 		case IPV6_JOIN_GROUP:
1829 		case IPV6_LEAVE_GROUP:
1830 			error = ip6_getmoptions(sopt, in6p->in6p_moptions);
1831 			break;
1832 
1833 		case IPV6_PORTALGO:
1834 			optval = ((struct inpcb_hdr *)in6p)->inph_portalgo;
1835 			error = sockopt_setint(sopt, optval);
1836 			break;
1837 
1838 #if defined(IPSEC)
1839 		case IPV6_IPSEC_POLICY:
1840 		    {
1841 			struct mbuf *m = NULL;
1842 
1843 			/* XXX this will return EINVAL as sopt is empty */
1844 			error = ipsec6_get_policy(in6p, sopt->sopt_data,
1845 			    sopt->sopt_size, &m);
1846 			if (!error)
1847 				error = sockopt_setmbuf(sopt, m);
1848 
1849 			break;
1850 		    }
1851 #endif /* IPSEC */
1852 
1853 		default:
1854 			error = ENOPROTOOPT;
1855 			break;
1856 		}
1857 		break;
1858 	}
1859 	return (error);
1860 }
1861 
1862 int
1863 ip6_raw_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1864 {
1865 	int error = 0, optval;
1866 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
1867 	struct in6pcb *in6p = sotoin6pcb(so);
1868 	int level, optname;
1869 
1870 	KASSERT(sopt != NULL);
1871 
1872 	level = sopt->sopt_level;
1873 	optname = sopt->sopt_name;
1874 
1875 	if (level != IPPROTO_IPV6) {
1876 		return ENOPROTOOPT;
1877 	}
1878 
1879 	switch (optname) {
1880 	case IPV6_CHECKSUM:
1881 		/*
1882 		 * For ICMPv6 sockets, no modification allowed for checksum
1883 		 * offset, permit "no change" values to help existing apps.
1884 		 *
1885 		 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM
1886 		 * for an ICMPv6 socket will fail."  The current
1887 		 * behavior does not meet RFC3542.
1888 		 */
1889 		switch (op) {
1890 		case PRCO_SETOPT:
1891 			error = sockopt_getint(sopt, &optval);
1892 			if (error)
1893 				break;
1894 			if ((optval % 2) != 0) {
1895 				/* the API assumes even offset values */
1896 				error = EINVAL;
1897 			} else if (so->so_proto->pr_protocol ==
1898 			    IPPROTO_ICMPV6) {
1899 				if (optval != icmp6off)
1900 					error = EINVAL;
1901 			} else
1902 				in6p->in6p_cksum = optval;
1903 			break;
1904 
1905 		case PRCO_GETOPT:
1906 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
1907 				optval = icmp6off;
1908 			else
1909 				optval = in6p->in6p_cksum;
1910 
1911 			error = sockopt_setint(sopt, optval);
1912 			break;
1913 
1914 		default:
1915 			error = EINVAL;
1916 			break;
1917 		}
1918 		break;
1919 
1920 	default:
1921 		error = ENOPROTOOPT;
1922 		break;
1923 	}
1924 
1925 	return (error);
1926 }
1927 
1928 #ifdef RFC2292
1929 /*
1930  * Set up IP6 options in pcb for insertion in output packets or
1931  * specifying behavior of outgoing packets.
1932  */
1933 static int
1934 ip6_pcbopts(struct ip6_pktopts **pktopt, struct socket *so,
1935     struct sockopt *sopt)
1936 {
1937 	struct ip6_pktopts *opt = *pktopt;
1938 	struct mbuf *m;
1939 	int error = 0;
1940 
1941 	/* turn off any old options. */
1942 	if (opt) {
1943 #ifdef DIAGNOSTIC
1944 	    if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
1945 		opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
1946 		opt->ip6po_rhinfo.ip6po_rhi_rthdr)
1947 		    printf("ip6_pcbopts: all specified options are cleared.\n");
1948 #endif
1949 		ip6_clearpktopts(opt, -1);
1950 	} else {
1951 		opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
1952 		if (opt == NULL)
1953 			return (ENOBUFS);
1954 	}
1955 	*pktopt = NULL;
1956 
1957 	if (sopt == NULL || sopt->sopt_size == 0) {
1958 		/*
1959 		 * Only turning off any previous options, regardless of
1960 		 * whether the opt is just created or given.
1961 		 */
1962 		free(opt, M_IP6OPT);
1963 		return (0);
1964 	}
1965 
1966 	/*  set options specified by user. */
1967 	m = sockopt_getmbuf(sopt);
1968 	if (m == NULL) {
1969 		free(opt, M_IP6OPT);
1970 		return (ENOBUFS);
1971 	}
1972 
1973 	error = ip6_setpktopts(m, opt, NULL, kauth_cred_get(),
1974 	    so->so_proto->pr_protocol);
1975 	m_freem(m);
1976 	if (error != 0) {
1977 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
1978 		free(opt, M_IP6OPT);
1979 		return (error);
1980 	}
1981 	*pktopt = opt;
1982 	return (0);
1983 }
1984 #endif
1985 
1986 /*
1987  * initialize ip6_pktopts.  beware that there are non-zero default values in
1988  * the struct.
1989  */
1990 void
1991 ip6_initpktopts(struct ip6_pktopts *opt)
1992 {
1993 
1994 	memset(opt, 0, sizeof(*opt));
1995 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
1996 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
1997 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
1998 }
1999 
2000 #define sin6tosa(sin6)	((struct sockaddr *)(sin6)) /* XXX */
2001 static int
2002 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2003     kauth_cred_t cred, int uproto)
2004 {
2005 	struct ip6_pktopts *opt;
2006 
2007 	if (*pktopt == NULL) {
2008 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2009 		    M_NOWAIT);
2010 		if (*pktopt == NULL)
2011 			return (ENOBUFS);
2012 
2013 		ip6_initpktopts(*pktopt);
2014 	}
2015 	opt = *pktopt;
2016 
2017 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2018 }
2019 
2020 static int
2021 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2022 {
2023 	void *optdata = NULL;
2024 	int optdatalen = 0;
2025 	struct ip6_ext *ip6e;
2026 	int error = 0;
2027 	struct in6_pktinfo null_pktinfo;
2028 	int deftclass = 0, on;
2029 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2030 
2031 	switch (optname) {
2032 	case IPV6_PKTINFO:
2033 		if (pktopt && pktopt->ip6po_pktinfo)
2034 			optdata = (void *)pktopt->ip6po_pktinfo;
2035 		else {
2036 			/* XXX: we don't have to do this every time... */
2037 			memset(&null_pktinfo, 0, sizeof(null_pktinfo));
2038 			optdata = (void *)&null_pktinfo;
2039 		}
2040 		optdatalen = sizeof(struct in6_pktinfo);
2041 		break;
2042 	case IPV6_OTCLASS:
2043 		/* XXX */
2044 		return (EINVAL);
2045 	case IPV6_TCLASS:
2046 		if (pktopt && pktopt->ip6po_tclass >= 0)
2047 			optdata = (void *)&pktopt->ip6po_tclass;
2048 		else
2049 			optdata = (void *)&deftclass;
2050 		optdatalen = sizeof(int);
2051 		break;
2052 	case IPV6_HOPOPTS:
2053 		if (pktopt && pktopt->ip6po_hbh) {
2054 			optdata = (void *)pktopt->ip6po_hbh;
2055 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2056 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2057 		}
2058 		break;
2059 	case IPV6_RTHDR:
2060 		if (pktopt && pktopt->ip6po_rthdr) {
2061 			optdata = (void *)pktopt->ip6po_rthdr;
2062 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2063 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2064 		}
2065 		break;
2066 	case IPV6_RTHDRDSTOPTS:
2067 		if (pktopt && pktopt->ip6po_dest1) {
2068 			optdata = (void *)pktopt->ip6po_dest1;
2069 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2070 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2071 		}
2072 		break;
2073 	case IPV6_DSTOPTS:
2074 		if (pktopt && pktopt->ip6po_dest2) {
2075 			optdata = (void *)pktopt->ip6po_dest2;
2076 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2077 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2078 		}
2079 		break;
2080 	case IPV6_NEXTHOP:
2081 		if (pktopt && pktopt->ip6po_nexthop) {
2082 			optdata = (void *)pktopt->ip6po_nexthop;
2083 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2084 		}
2085 		break;
2086 	case IPV6_USE_MIN_MTU:
2087 		if (pktopt)
2088 			optdata = (void *)&pktopt->ip6po_minmtu;
2089 		else
2090 			optdata = (void *)&defminmtu;
2091 		optdatalen = sizeof(int);
2092 		break;
2093 	case IPV6_DONTFRAG:
2094 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2095 			on = 1;
2096 		else
2097 			on = 0;
2098 		optdata = (void *)&on;
2099 		optdatalen = sizeof(on);
2100 		break;
2101 	default:		/* should not happen */
2102 #ifdef DIAGNOSTIC
2103 		panic("ip6_getpcbopt: unexpected option\n");
2104 #endif
2105 		return (ENOPROTOOPT);
2106 	}
2107 
2108 	error = sockopt_set(sopt, optdata, optdatalen);
2109 
2110 	return (error);
2111 }
2112 
2113 void
2114 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2115 {
2116 	if (optname == -1 || optname == IPV6_PKTINFO) {
2117 		if (pktopt->ip6po_pktinfo)
2118 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2119 		pktopt->ip6po_pktinfo = NULL;
2120 	}
2121 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2122 		pktopt->ip6po_hlim = -1;
2123 	if (optname == -1 || optname == IPV6_TCLASS)
2124 		pktopt->ip6po_tclass = -1;
2125 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2126 		rtcache_free(&pktopt->ip6po_nextroute);
2127 		if (pktopt->ip6po_nexthop)
2128 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2129 		pktopt->ip6po_nexthop = NULL;
2130 	}
2131 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2132 		if (pktopt->ip6po_hbh)
2133 			free(pktopt->ip6po_hbh, M_IP6OPT);
2134 		pktopt->ip6po_hbh = NULL;
2135 	}
2136 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2137 		if (pktopt->ip6po_dest1)
2138 			free(pktopt->ip6po_dest1, M_IP6OPT);
2139 		pktopt->ip6po_dest1 = NULL;
2140 	}
2141 	if (optname == -1 || optname == IPV6_RTHDR) {
2142 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2143 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2144 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2145 		rtcache_free(&pktopt->ip6po_route);
2146 	}
2147 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2148 		if (pktopt->ip6po_dest2)
2149 			free(pktopt->ip6po_dest2, M_IP6OPT);
2150 		pktopt->ip6po_dest2 = NULL;
2151 	}
2152 }
2153 
2154 #define PKTOPT_EXTHDRCPY(type) 					\
2155 do {								\
2156 	if (src->type) {					\
2157 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2158 		dst->type = malloc(hlen, M_IP6OPT, canwait);	\
2159 		if (dst->type == NULL)				\
2160 			goto bad;				\
2161 		memcpy(dst->type, src->type, hlen);		\
2162 	}							\
2163 } while (/*CONSTCOND*/ 0)
2164 
2165 static int
2166 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2167 {
2168 	dst->ip6po_hlim = src->ip6po_hlim;
2169 	dst->ip6po_tclass = src->ip6po_tclass;
2170 	dst->ip6po_flags = src->ip6po_flags;
2171 	if (src->ip6po_pktinfo) {
2172 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2173 		    M_IP6OPT, canwait);
2174 		if (dst->ip6po_pktinfo == NULL)
2175 			goto bad;
2176 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2177 	}
2178 	if (src->ip6po_nexthop) {
2179 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2180 		    M_IP6OPT, canwait);
2181 		if (dst->ip6po_nexthop == NULL)
2182 			goto bad;
2183 		memcpy(dst->ip6po_nexthop, src->ip6po_nexthop,
2184 		    src->ip6po_nexthop->sa_len);
2185 	}
2186 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2187 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2188 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2189 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2190 	return (0);
2191 
2192   bad:
2193 	if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
2194 	if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
2195 	if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
2196 	if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
2197 	if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
2198 	if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
2199 
2200 	return (ENOBUFS);
2201 }
2202 #undef PKTOPT_EXTHDRCPY
2203 
2204 struct ip6_pktopts *
2205 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2206 {
2207 	int error;
2208 	struct ip6_pktopts *dst;
2209 
2210 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2211 	if (dst == NULL)
2212 		return (NULL);
2213 	ip6_initpktopts(dst);
2214 
2215 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2216 		free(dst, M_IP6OPT);
2217 		return (NULL);
2218 	}
2219 
2220 	return (dst);
2221 }
2222 
2223 void
2224 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2225 {
2226 	if (pktopt == NULL)
2227 		return;
2228 
2229 	ip6_clearpktopts(pktopt, -1);
2230 
2231 	free(pktopt, M_IP6OPT);
2232 }
2233 
2234 /*
2235  * Set the IP6 multicast options in response to user setsockopt().
2236  */
2237 static int
2238 ip6_setmoptions(const struct sockopt *sopt, struct ip6_moptions **im6op)
2239 {
2240 	int error = 0;
2241 	u_int loop, ifindex;
2242 	struct ipv6_mreq mreq;
2243 	struct ifnet *ifp;
2244 	struct ip6_moptions *im6o = *im6op;
2245 	struct route ro;
2246 	struct in6_multi_mship *imm;
2247 	struct lwp *l = curlwp;	/* XXX */
2248 
2249 	if (im6o == NULL) {
2250 		/*
2251 		 * No multicast option buffer attached to the pcb;
2252 		 * allocate one and initialize to default values.
2253 		 */
2254 		im6o = malloc(sizeof(*im6o), M_IPMOPTS, M_NOWAIT);
2255 		if (im6o == NULL)
2256 			return (ENOBUFS);
2257 
2258 		*im6op = im6o;
2259 		im6o->im6o_multicast_ifp = NULL;
2260 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2261 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2262 		LIST_INIT(&im6o->im6o_memberships);
2263 	}
2264 
2265 	switch (sopt->sopt_name) {
2266 
2267 	case IPV6_MULTICAST_IF:
2268 		/*
2269 		 * Select the interface for outgoing multicast packets.
2270 		 */
2271 		error = sockopt_get(sopt, &ifindex, sizeof(ifindex));
2272 		if (error != 0)
2273 			break;
2274 
2275 		if (ifindex != 0) {
2276 			if (if_indexlim <= ifindex || !ifindex2ifnet[ifindex]) {
2277 				error = ENXIO;	/* XXX EINVAL? */
2278 				break;
2279 			}
2280 			ifp = ifindex2ifnet[ifindex];
2281 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
2282 				error = EADDRNOTAVAIL;
2283 				break;
2284 			}
2285 		} else
2286 			ifp = NULL;
2287 		im6o->im6o_multicast_ifp = ifp;
2288 		break;
2289 
2290 	case IPV6_MULTICAST_HOPS:
2291 	    {
2292 		/*
2293 		 * Set the IP6 hoplimit for outgoing multicast packets.
2294 		 */
2295 		int optval;
2296 
2297 		error = sockopt_getint(sopt, &optval);
2298 		if (error != 0)
2299 			break;
2300 
2301 		if (optval < -1 || optval >= 256)
2302 			error = EINVAL;
2303 		else if (optval == -1)
2304 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2305 		else
2306 			im6o->im6o_multicast_hlim = optval;
2307 		break;
2308 	    }
2309 
2310 	case IPV6_MULTICAST_LOOP:
2311 		/*
2312 		 * Set the loopback flag for outgoing multicast packets.
2313 		 * Must be zero or one.
2314 		 */
2315 		error = sockopt_get(sopt, &loop, sizeof(loop));
2316 		if (error != 0)
2317 			break;
2318 		if (loop > 1) {
2319 			error = EINVAL;
2320 			break;
2321 		}
2322 		im6o->im6o_multicast_loop = loop;
2323 		break;
2324 
2325 	case IPV6_JOIN_GROUP:
2326 		/*
2327 		 * Add a multicast group membership.
2328 		 * Group must be a valid IP6 multicast address.
2329 		 */
2330 		error = sockopt_get(sopt, &mreq, sizeof(mreq));
2331 		if (error != 0)
2332 			break;
2333 
2334 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq.ipv6mr_multiaddr)) {
2335 			/*
2336 			 * We use the unspecified address to specify to accept
2337 			 * all multicast addresses. Only super user is allowed
2338 			 * to do this.
2339 			 */
2340 			if (kauth_authorize_network(l->l_cred, KAUTH_NETWORK_IPV6,
2341 			    KAUTH_REQ_NETWORK_IPV6_JOIN_MULTICAST, NULL, NULL, NULL))
2342 			{
2343 				error = EACCES;
2344 				break;
2345 			}
2346 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq.ipv6mr_multiaddr)) {
2347 			error = EINVAL;
2348 			break;
2349 		}
2350 
2351 		/*
2352 		 * If no interface was explicitly specified, choose an
2353 		 * appropriate one according to the given multicast address.
2354 		 */
2355 		if (mreq.ipv6mr_interface == 0) {
2356 			struct rtentry *rt;
2357 			union {
2358 				struct sockaddr		dst;
2359 				struct sockaddr_in6	dst6;
2360 			} u;
2361 
2362 			/*
2363 			 * Look up the routing table for the
2364 			 * address, and choose the outgoing interface.
2365 			 *   XXX: is it a good approach?
2366 			 */
2367 			memset(&ro, 0, sizeof(ro));
2368 			sockaddr_in6_init(&u.dst6, &mreq.ipv6mr_multiaddr, 0,
2369 			    0, 0);
2370 			rtcache_setdst(&ro, &u.dst);
2371 			ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp
2372 			                                        : NULL;
2373 			rtcache_free(&ro);
2374 		} else {
2375 			/*
2376 			 * If the interface is specified, validate it.
2377 			 */
2378 			if (if_indexlim <= mreq.ipv6mr_interface ||
2379 			    !ifindex2ifnet[mreq.ipv6mr_interface]) {
2380 				error = ENXIO;	/* XXX EINVAL? */
2381 				break;
2382 			}
2383 			ifp = ifindex2ifnet[mreq.ipv6mr_interface];
2384 		}
2385 
2386 		/*
2387 		 * See if we found an interface, and confirm that it
2388 		 * supports multicast
2389 		 */
2390 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2391 			error = EADDRNOTAVAIL;
2392 			break;
2393 		}
2394 
2395 		if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) {
2396 			error = EADDRNOTAVAIL; /* XXX: should not happen */
2397 			break;
2398 		}
2399 
2400 		/*
2401 		 * See if the membership already exists.
2402 		 */
2403 		for (imm = im6o->im6o_memberships.lh_first;
2404 		     imm != NULL; imm = imm->i6mm_chain.le_next)
2405 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
2406 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2407 			    &mreq.ipv6mr_multiaddr))
2408 				break;
2409 		if (imm != NULL) {
2410 			error = EADDRINUSE;
2411 			break;
2412 		}
2413 		/*
2414 		 * Everything looks good; add a new record to the multicast
2415 		 * address list for the given interface.
2416 		 */
2417 		imm = in6_joingroup(ifp, &mreq.ipv6mr_multiaddr, &error, 0);
2418 		if (imm == NULL)
2419 			break;
2420 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2421 		break;
2422 
2423 	case IPV6_LEAVE_GROUP:
2424 		/*
2425 		 * Drop a multicast group membership.
2426 		 * Group must be a valid IP6 multicast address.
2427 		 */
2428 		error = sockopt_get(sopt, &mreq, sizeof(mreq));
2429 		if (error != 0)
2430 			break;
2431 
2432 		/*
2433 		 * If an interface address was specified, get a pointer
2434 		 * to its ifnet structure.
2435 		 */
2436 		if (mreq.ipv6mr_interface != 0) {
2437 			if (if_indexlim <= mreq.ipv6mr_interface ||
2438 			    !ifindex2ifnet[mreq.ipv6mr_interface]) {
2439 				error = ENXIO;	/* XXX EINVAL? */
2440 				break;
2441 			}
2442 			ifp = ifindex2ifnet[mreq.ipv6mr_interface];
2443 		} else
2444 			ifp = NULL;
2445 
2446 		/* Fill in the scope zone ID */
2447 		if (ifp) {
2448 			if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) {
2449 				/* XXX: should not happen */
2450 				error = EADDRNOTAVAIL;
2451 				break;
2452 			}
2453 		} else if (mreq.ipv6mr_interface != 0) {
2454 			/*
2455 			 * XXX: This case would happens when the (positive)
2456 			 * index is in the valid range, but the corresponding
2457 			 * interface has been detached dynamically.  The above
2458 			 * check probably avoids such case to happen here, but
2459 			 * we check it explicitly for safety.
2460 			 */
2461 			error = EADDRNOTAVAIL;
2462 			break;
2463 		} else {	/* ipv6mr_interface == 0 */
2464 			struct sockaddr_in6 sa6_mc;
2465 
2466 			/*
2467 			 * The API spec says as follows:
2468 			 *  If the interface index is specified as 0, the
2469 			 *  system may choose a multicast group membership to
2470 			 *  drop by matching the multicast address only.
2471 			 * On the other hand, we cannot disambiguate the scope
2472 			 * zone unless an interface is provided.  Thus, we
2473 			 * check if there's ambiguity with the default scope
2474 			 * zone as the last resort.
2475 			 */
2476 			sockaddr_in6_init(&sa6_mc, &mreq.ipv6mr_multiaddr,
2477 			    0, 0, 0);
2478 			error = sa6_embedscope(&sa6_mc, ip6_use_defzone);
2479 			if (error != 0)
2480 				break;
2481 			mreq.ipv6mr_multiaddr = sa6_mc.sin6_addr;
2482 		}
2483 
2484 		/*
2485 		 * Find the membership in the membership list.
2486 		 */
2487 		for (imm = im6o->im6o_memberships.lh_first;
2488 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
2489 			if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2490 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2491 			    &mreq.ipv6mr_multiaddr))
2492 				break;
2493 		}
2494 		if (imm == NULL) {
2495 			/* Unable to resolve interface */
2496 			error = EADDRNOTAVAIL;
2497 			break;
2498 		}
2499 		/*
2500 		 * Give up the multicast address record to which the
2501 		 * membership points.
2502 		 */
2503 		LIST_REMOVE(imm, i6mm_chain);
2504 		in6_leavegroup(imm);
2505 		break;
2506 
2507 	default:
2508 		error = EOPNOTSUPP;
2509 		break;
2510 	}
2511 
2512 	/*
2513 	 * If all options have default values, no need to keep the mbuf.
2514 	 */
2515 	if (im6o->im6o_multicast_ifp == NULL &&
2516 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2517 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2518 	    im6o->im6o_memberships.lh_first == NULL) {
2519 		free(*im6op, M_IPMOPTS);
2520 		*im6op = NULL;
2521 	}
2522 
2523 	return (error);
2524 }
2525 
2526 /*
2527  * Return the IP6 multicast options in response to user getsockopt().
2528  */
2529 static int
2530 ip6_getmoptions(struct sockopt *sopt, struct ip6_moptions *im6o)
2531 {
2532 	u_int optval;
2533 	int error;
2534 
2535 	switch (sopt->sopt_name) {
2536 	case IPV6_MULTICAST_IF:
2537 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2538 			optval = 0;
2539 		else
2540 			optval = im6o->im6o_multicast_ifp->if_index;
2541 
2542 		error = sockopt_set(sopt, &optval, sizeof(optval));
2543 		break;
2544 
2545 	case IPV6_MULTICAST_HOPS:
2546 		if (im6o == NULL)
2547 			optval = ip6_defmcasthlim;
2548 		else
2549 			optval = im6o->im6o_multicast_hlim;
2550 
2551 		error = sockopt_set(sopt, &optval, sizeof(optval));
2552 		break;
2553 
2554 	case IPV6_MULTICAST_LOOP:
2555 		if (im6o == NULL)
2556 			optval = IPV6_DEFAULT_MULTICAST_LOOP;
2557 		else
2558 			optval = im6o->im6o_multicast_loop;
2559 
2560 		error = sockopt_set(sopt, &optval, sizeof(optval));
2561 		break;
2562 
2563 	default:
2564 		error = EOPNOTSUPP;
2565 	}
2566 
2567 	return (error);
2568 }
2569 
2570 /*
2571  * Discard the IP6 multicast options.
2572  */
2573 void
2574 ip6_freemoptions(struct ip6_moptions *im6o)
2575 {
2576 	struct in6_multi_mship *imm;
2577 
2578 	if (im6o == NULL)
2579 		return;
2580 
2581 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2582 		LIST_REMOVE(imm, i6mm_chain);
2583 		in6_leavegroup(imm);
2584 	}
2585 	free(im6o, M_IPMOPTS);
2586 }
2587 
2588 /*
2589  * Set IPv6 outgoing packet options based on advanced API.
2590  */
2591 int
2592 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2593 	struct ip6_pktopts *stickyopt, kauth_cred_t cred, int uproto)
2594 {
2595 	struct cmsghdr *cm = 0;
2596 
2597 	if (control == NULL || opt == NULL)
2598 		return (EINVAL);
2599 
2600 	ip6_initpktopts(opt);
2601 	if (stickyopt) {
2602 		int error;
2603 
2604 		/*
2605 		 * If stickyopt is provided, make a local copy of the options
2606 		 * for this particular packet, then override them by ancillary
2607 		 * objects.
2608 		 * XXX: copypktopts() does not copy the cached route to a next
2609 		 * hop (if any).  This is not very good in terms of efficiency,
2610 		 * but we can allow this since this option should be rarely
2611 		 * used.
2612 		 */
2613 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2614 			return (error);
2615 	}
2616 
2617 	/*
2618 	 * XXX: Currently, we assume all the optional information is stored
2619 	 * in a single mbuf.
2620 	 */
2621 	if (control->m_next)
2622 		return (EINVAL);
2623 
2624 	/* XXX if cm->cmsg_len is not aligned, control->m_len can become <0 */
2625 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2626 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2627 		int error;
2628 
2629 		if (control->m_len < CMSG_LEN(0))
2630 			return (EINVAL);
2631 
2632 		cm = mtod(control, struct cmsghdr *);
2633 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2634 			return (EINVAL);
2635 		if (cm->cmsg_level != IPPROTO_IPV6)
2636 			continue;
2637 
2638 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2639 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2640 		if (error)
2641 			return (error);
2642 	}
2643 
2644 	return (0);
2645 }
2646 
2647 /*
2648  * Set a particular packet option, as a sticky option or an ancillary data
2649  * item.  "len" can be 0 only when it's a sticky option.
2650  * We have 4 cases of combination of "sticky" and "cmsg":
2651  * "sticky=0, cmsg=0": impossible
2652  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2653  * "sticky=1, cmsg=0": RFC3542 socket option
2654  * "sticky=1, cmsg=1": RFC2292 socket option
2655  */
2656 static int
2657 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2658     kauth_cred_t cred, int sticky, int cmsg, int uproto)
2659 {
2660 	int minmtupolicy;
2661 	int error;
2662 
2663 	if (!sticky && !cmsg) {
2664 #ifdef DIAGNOSTIC
2665 		printf("ip6_setpktopt: impossible case\n");
2666 #endif
2667 		return (EINVAL);
2668 	}
2669 
2670 	/*
2671 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2672 	 * not be specified in the context of RFC3542.  Conversely,
2673 	 * RFC3542 types should not be specified in the context of RFC2292.
2674 	 */
2675 	if (!cmsg) {
2676 		switch (optname) {
2677 		case IPV6_2292PKTINFO:
2678 		case IPV6_2292HOPLIMIT:
2679 		case IPV6_2292NEXTHOP:
2680 		case IPV6_2292HOPOPTS:
2681 		case IPV6_2292DSTOPTS:
2682 		case IPV6_2292RTHDR:
2683 		case IPV6_2292PKTOPTIONS:
2684 			return (ENOPROTOOPT);
2685 		}
2686 	}
2687 	if (sticky && cmsg) {
2688 		switch (optname) {
2689 		case IPV6_PKTINFO:
2690 		case IPV6_HOPLIMIT:
2691 		case IPV6_NEXTHOP:
2692 		case IPV6_HOPOPTS:
2693 		case IPV6_DSTOPTS:
2694 		case IPV6_RTHDRDSTOPTS:
2695 		case IPV6_RTHDR:
2696 		case IPV6_USE_MIN_MTU:
2697 		case IPV6_DONTFRAG:
2698 		case IPV6_OTCLASS:
2699 		case IPV6_TCLASS:
2700 			return (ENOPROTOOPT);
2701 		}
2702 	}
2703 
2704 	switch (optname) {
2705 #ifdef RFC2292
2706 	case IPV6_2292PKTINFO:
2707 #endif
2708 	case IPV6_PKTINFO:
2709 	{
2710 		struct ifnet *ifp = NULL;
2711 		struct in6_pktinfo *pktinfo;
2712 
2713 		if (len != sizeof(struct in6_pktinfo))
2714 			return (EINVAL);
2715 
2716 		pktinfo = (struct in6_pktinfo *)buf;
2717 
2718 		/*
2719 		 * An application can clear any sticky IPV6_PKTINFO option by
2720 		 * doing a "regular" setsockopt with ipi6_addr being
2721 		 * in6addr_any and ipi6_ifindex being zero.
2722 		 * [RFC 3542, Section 6]
2723 		 */
2724 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2725 		    pktinfo->ipi6_ifindex == 0 &&
2726 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2727 			ip6_clearpktopts(opt, optname);
2728 			break;
2729 		}
2730 
2731 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2732 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2733 			return (EINVAL);
2734 		}
2735 
2736 		/* validate the interface index if specified. */
2737 		if (pktinfo->ipi6_ifindex >= if_indexlim) {
2738 			 return (ENXIO);
2739 		}
2740 		if (pktinfo->ipi6_ifindex) {
2741 			ifp = ifindex2ifnet[pktinfo->ipi6_ifindex];
2742 			if (ifp == NULL)
2743 				return (ENXIO);
2744 		}
2745 
2746 		/*
2747 		 * We store the address anyway, and let in6_selectsrc()
2748 		 * validate the specified address.  This is because ipi6_addr
2749 		 * may not have enough information about its scope zone, and
2750 		 * we may need additional information (such as outgoing
2751 		 * interface or the scope zone of a destination address) to
2752 		 * disambiguate the scope.
2753 		 * XXX: the delay of the validation may confuse the
2754 		 * application when it is used as a sticky option.
2755 		 */
2756 		if (opt->ip6po_pktinfo == NULL) {
2757 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2758 			    M_IP6OPT, M_NOWAIT);
2759 			if (opt->ip6po_pktinfo == NULL)
2760 				return (ENOBUFS);
2761 		}
2762 		memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo));
2763 		break;
2764 	}
2765 
2766 #ifdef RFC2292
2767 	case IPV6_2292HOPLIMIT:
2768 #endif
2769 	case IPV6_HOPLIMIT:
2770 	{
2771 		int *hlimp;
2772 
2773 		/*
2774 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2775 		 * to simplify the ordering among hoplimit options.
2776 		 */
2777 		if (optname == IPV6_HOPLIMIT && sticky)
2778 			return (ENOPROTOOPT);
2779 
2780 		if (len != sizeof(int))
2781 			return (EINVAL);
2782 		hlimp = (int *)buf;
2783 		if (*hlimp < -1 || *hlimp > 255)
2784 			return (EINVAL);
2785 
2786 		opt->ip6po_hlim = *hlimp;
2787 		break;
2788 	}
2789 
2790 	case IPV6_OTCLASS:
2791 		if (len != sizeof(u_int8_t))
2792 			return (EINVAL);
2793 
2794 		opt->ip6po_tclass = *(u_int8_t *)buf;
2795 		break;
2796 
2797 	case IPV6_TCLASS:
2798 	{
2799 		int tclass;
2800 
2801 		if (len != sizeof(int))
2802 			return (EINVAL);
2803 		tclass = *(int *)buf;
2804 		if (tclass < -1 || tclass > 255)
2805 			return (EINVAL);
2806 
2807 		opt->ip6po_tclass = tclass;
2808 		break;
2809 	}
2810 
2811 #ifdef RFC2292
2812 	case IPV6_2292NEXTHOP:
2813 #endif
2814 	case IPV6_NEXTHOP:
2815 		error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
2816 		    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2817 		if (error)
2818 			return (error);
2819 
2820 		if (len == 0) {	/* just remove the option */
2821 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2822 			break;
2823 		}
2824 
2825 		/* check if cmsg_len is large enough for sa_len */
2826 		if (len < sizeof(struct sockaddr) || len < *buf)
2827 			return (EINVAL);
2828 
2829 		switch (((struct sockaddr *)buf)->sa_family) {
2830 		case AF_INET6:
2831 		{
2832 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2833 
2834 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2835 				return (EINVAL);
2836 
2837 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2838 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2839 				return (EINVAL);
2840 			}
2841 			if ((error = sa6_embedscope(sa6, ip6_use_defzone))
2842 			    != 0) {
2843 				return (error);
2844 			}
2845 			break;
2846 		}
2847 		case AF_LINK:	/* eventually be supported? */
2848 		default:
2849 			return (EAFNOSUPPORT);
2850 		}
2851 
2852 		/* turn off the previous option, then set the new option. */
2853 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2854 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2855 		if (opt->ip6po_nexthop == NULL)
2856 			return (ENOBUFS);
2857 		memcpy(opt->ip6po_nexthop, buf, *buf);
2858 		break;
2859 
2860 #ifdef RFC2292
2861 	case IPV6_2292HOPOPTS:
2862 #endif
2863 	case IPV6_HOPOPTS:
2864 	{
2865 		struct ip6_hbh *hbh;
2866 		int hbhlen;
2867 
2868 		/*
2869 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2870 		 * options, since per-option restriction has too much
2871 		 * overhead.
2872 		 */
2873 		error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
2874 		    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2875 		if (error)
2876 			return (error);
2877 
2878 		if (len == 0) {
2879 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2880 			break;	/* just remove the option */
2881 		}
2882 
2883 		/* message length validation */
2884 		if (len < sizeof(struct ip6_hbh))
2885 			return (EINVAL);
2886 		hbh = (struct ip6_hbh *)buf;
2887 		hbhlen = (hbh->ip6h_len + 1) << 3;
2888 		if (len != hbhlen)
2889 			return (EINVAL);
2890 
2891 		/* turn off the previous option, then set the new option. */
2892 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2893 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2894 		if (opt->ip6po_hbh == NULL)
2895 			return (ENOBUFS);
2896 		memcpy(opt->ip6po_hbh, hbh, hbhlen);
2897 
2898 		break;
2899 	}
2900 
2901 #ifdef RFC2292
2902 	case IPV6_2292DSTOPTS:
2903 #endif
2904 	case IPV6_DSTOPTS:
2905 	case IPV6_RTHDRDSTOPTS:
2906 	{
2907 		struct ip6_dest *dest, **newdest = NULL;
2908 		int destlen;
2909 
2910 		/* XXX: see the comment for IPV6_HOPOPTS */
2911 		error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
2912 		    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2913 		if (error)
2914 			return (error);
2915 
2916 		if (len == 0) {
2917 			ip6_clearpktopts(opt, optname);
2918 			break;	/* just remove the option */
2919 		}
2920 
2921 		/* message length validation */
2922 		if (len < sizeof(struct ip6_dest))
2923 			return (EINVAL);
2924 		dest = (struct ip6_dest *)buf;
2925 		destlen = (dest->ip6d_len + 1) << 3;
2926 		if (len != destlen)
2927 			return (EINVAL);
2928 		/*
2929 		 * Determine the position that the destination options header
2930 		 * should be inserted; before or after the routing header.
2931 		 */
2932 		switch (optname) {
2933 		case IPV6_2292DSTOPTS:
2934 			/*
2935 			 * The old advanced API is ambiguous on this point.
2936 			 * Our approach is to determine the position based
2937 			 * according to the existence of a routing header.
2938 			 * Note, however, that this depends on the order of the
2939 			 * extension headers in the ancillary data; the 1st
2940 			 * part of the destination options header must appear
2941 			 * before the routing header in the ancillary data,
2942 			 * too.
2943 			 * RFC3542 solved the ambiguity by introducing
2944 			 * separate ancillary data or option types.
2945 			 */
2946 			if (opt->ip6po_rthdr == NULL)
2947 				newdest = &opt->ip6po_dest1;
2948 			else
2949 				newdest = &opt->ip6po_dest2;
2950 			break;
2951 		case IPV6_RTHDRDSTOPTS:
2952 			newdest = &opt->ip6po_dest1;
2953 			break;
2954 		case IPV6_DSTOPTS:
2955 			newdest = &opt->ip6po_dest2;
2956 			break;
2957 		}
2958 
2959 		/* turn off the previous option, then set the new option. */
2960 		ip6_clearpktopts(opt, optname);
2961 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2962 		if (*newdest == NULL)
2963 			return (ENOBUFS);
2964 		memcpy(*newdest, dest, destlen);
2965 
2966 		break;
2967 	}
2968 
2969 #ifdef RFC2292
2970 	case IPV6_2292RTHDR:
2971 #endif
2972 	case IPV6_RTHDR:
2973 	{
2974 		struct ip6_rthdr *rth;
2975 		int rthlen;
2976 
2977 		if (len == 0) {
2978 			ip6_clearpktopts(opt, IPV6_RTHDR);
2979 			break;	/* just remove the option */
2980 		}
2981 
2982 		/* message length validation */
2983 		if (len < sizeof(struct ip6_rthdr))
2984 			return (EINVAL);
2985 		rth = (struct ip6_rthdr *)buf;
2986 		rthlen = (rth->ip6r_len + 1) << 3;
2987 		if (len != rthlen)
2988 			return (EINVAL);
2989 		switch (rth->ip6r_type) {
2990 		case IPV6_RTHDR_TYPE_0:
2991 			if (rth->ip6r_len == 0)	/* must contain one addr */
2992 				return (EINVAL);
2993 			if (rth->ip6r_len % 2) /* length must be even */
2994 				return (EINVAL);
2995 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2996 				return (EINVAL);
2997 			break;
2998 		default:
2999 			return (EINVAL);	/* not supported */
3000 		}
3001 		/* turn off the previous option */
3002 		ip6_clearpktopts(opt, IPV6_RTHDR);
3003 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3004 		if (opt->ip6po_rthdr == NULL)
3005 			return (ENOBUFS);
3006 		memcpy(opt->ip6po_rthdr, rth, rthlen);
3007 		break;
3008 	}
3009 
3010 	case IPV6_USE_MIN_MTU:
3011 		if (len != sizeof(int))
3012 			return (EINVAL);
3013 		minmtupolicy = *(int *)buf;
3014 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3015 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3016 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3017 			return (EINVAL);
3018 		}
3019 		opt->ip6po_minmtu = minmtupolicy;
3020 		break;
3021 
3022 	case IPV6_DONTFRAG:
3023 		if (len != sizeof(int))
3024 			return (EINVAL);
3025 
3026 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3027 			/*
3028 			 * we ignore this option for TCP sockets.
3029 			 * (RFC3542 leaves this case unspecified.)
3030 			 */
3031 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3032 		} else
3033 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3034 		break;
3035 
3036 	default:
3037 		return (ENOPROTOOPT);
3038 	} /* end of switch */
3039 
3040 	return (0);
3041 }
3042 
3043 /*
3044  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3045  * packet to the input queue of a specified interface.  Note that this
3046  * calls the output routine of the loopback "driver", but with an interface
3047  * pointer that might NOT be lo0ifp -- easier than replicating that code here.
3048  */
3049 void
3050 ip6_mloopback(struct ifnet *ifp, struct mbuf *m,
3051 	const struct sockaddr_in6 *dst)
3052 {
3053 	struct mbuf *copym;
3054 	struct ip6_hdr *ip6;
3055 
3056 	copym = m_copy(m, 0, M_COPYALL);
3057 	if (copym == NULL)
3058 		return;
3059 
3060 	/*
3061 	 * Make sure to deep-copy IPv6 header portion in case the data
3062 	 * is in an mbuf cluster, so that we can safely override the IPv6
3063 	 * header portion later.
3064 	 */
3065 	if ((copym->m_flags & M_EXT) != 0 ||
3066 	    copym->m_len < sizeof(struct ip6_hdr)) {
3067 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3068 		if (copym == NULL)
3069 			return;
3070 	}
3071 
3072 #ifdef DIAGNOSTIC
3073 	if (copym->m_len < sizeof(*ip6)) {
3074 		m_freem(copym);
3075 		return;
3076 	}
3077 #endif
3078 
3079 	ip6 = mtod(copym, struct ip6_hdr *);
3080 	/*
3081 	 * clear embedded scope identifiers if necessary.
3082 	 * in6_clearscope will touch the addresses only when necessary.
3083 	 */
3084 	in6_clearscope(&ip6->ip6_src);
3085 	in6_clearscope(&ip6->ip6_dst);
3086 
3087 	(void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL);
3088 }
3089 
3090 /*
3091  * Chop IPv6 header off from the payload.
3092  */
3093 static int
3094 ip6_splithdr(struct mbuf *m,  struct ip6_exthdrs *exthdrs)
3095 {
3096 	struct mbuf *mh;
3097 	struct ip6_hdr *ip6;
3098 
3099 	ip6 = mtod(m, struct ip6_hdr *);
3100 	if (m->m_len > sizeof(*ip6)) {
3101 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3102 		if (mh == 0) {
3103 			m_freem(m);
3104 			return ENOBUFS;
3105 		}
3106 		M_MOVE_PKTHDR(mh, m);
3107 		MH_ALIGN(mh, sizeof(*ip6));
3108 		m->m_len -= sizeof(*ip6);
3109 		m->m_data += sizeof(*ip6);
3110 		mh->m_next = m;
3111 		m = mh;
3112 		m->m_len = sizeof(*ip6);
3113 		bcopy((void *)ip6, mtod(m, void *), sizeof(*ip6));
3114 	}
3115 	exthdrs->ip6e_ip6 = m;
3116 	return 0;
3117 }
3118 
3119 /*
3120  * Compute IPv6 extension header length.
3121  */
3122 int
3123 ip6_optlen(struct in6pcb *in6p)
3124 {
3125 	int len;
3126 
3127 	if (!in6p->in6p_outputopts)
3128 		return 0;
3129 
3130 	len = 0;
3131 #define elen(x) \
3132     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3133 
3134 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3135 	len += elen(in6p->in6p_outputopts->ip6po_dest1);
3136 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3137 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3138 	return len;
3139 #undef elen
3140 }
3141