xref: /netbsd-src/sys/netinet6/ip6_output.c (revision d909946ca08dceb44d7d0f22ec9488679695d976)
1 /*	$NetBSD: ip6_output.c,v 1.173 2016/08/01 03:15:31 ozaki-r 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 1.173 2016/08/01 03:15:31 ozaki-r Exp $");
66 
67 #ifdef _KERNEL_OPT
68 #include "opt_inet.h"
69 #include "opt_inet6.h"
70 #include "opt_ipsec.h"
71 #endif
72 
73 #include <sys/param.h>
74 #include <sys/malloc.h>
75 #include <sys/mbuf.h>
76 #include <sys/errno.h>
77 #include <sys/protosw.h>
78 #include <sys/socket.h>
79 #include <sys/socketvar.h>
80 #include <sys/systm.h>
81 #include <sys/proc.h>
82 #include <sys/kauth.h>
83 
84 #include <net/if.h>
85 #include <net/route.h>
86 #include <net/pfil.h>
87 
88 #include <netinet/in.h>
89 #include <netinet/in_var.h>
90 #include <netinet/ip6.h>
91 #include <netinet/ip_var.h>
92 #include <netinet/icmp6.h>
93 #include <netinet/in_offload.h>
94 #include <netinet/portalgo.h>
95 #include <netinet6/in6_offload.h>
96 #include <netinet6/ip6_var.h>
97 #include <netinet6/ip6_private.h>
98 #include <netinet6/in6_pcb.h>
99 #include <netinet6/nd6.h>
100 #include <netinet6/ip6protosw.h>
101 #include <netinet6/scope6_var.h>
102 
103 #ifdef IPSEC
104 #include <netipsec/ipsec.h>
105 #include <netipsec/ipsec6.h>
106 #include <netipsec/key.h>
107 #include <netipsec/xform.h>
108 #endif
109 
110 
111 #include <net/net_osdep.h>
112 
113 extern pfil_head_t *inet6_pfil_hook;	/* XXX */
114 
115 struct ip6_exthdrs {
116 	struct mbuf *ip6e_ip6;
117 	struct mbuf *ip6e_hbh;
118 	struct mbuf *ip6e_dest1;
119 	struct mbuf *ip6e_rthdr;
120 	struct mbuf *ip6e_dest2;
121 };
122 
123 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
124 	kauth_cred_t, int);
125 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
126 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, kauth_cred_t,
127 	int, int, int);
128 static int ip6_setmoptions(const struct sockopt *, struct in6pcb *);
129 static int ip6_getmoptions(struct sockopt *, struct in6pcb *);
130 static int ip6_copyexthdr(struct mbuf **, void *, int);
131 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
132 	struct ip6_frag **);
133 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
134 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
135 static int ip6_getpmtu(struct route *, struct route *, struct ifnet *,
136     const struct in6_addr *, u_long *, int *);
137 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
138 
139 #ifdef RFC2292
140 static int ip6_pcbopts(struct ip6_pktopts **, struct socket *, struct sockopt *);
141 #endif
142 
143 /*
144  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
145  * header (with pri, len, nxt, hlim, src, dst).
146  * This function may modify ver and hlim only.
147  * The mbuf chain containing the packet will be freed.
148  * The mbuf opt, if present, will not be freed.
149  *
150  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
151  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
152  * which is rt_rmx.rmx_mtu.
153  */
154 int
155 ip6_output(
156     struct mbuf *m0,
157     struct ip6_pktopts *opt,
158     struct route *ro,
159     int flags,
160     struct ip6_moptions *im6o,
161     struct socket *so,
162     struct ifnet **ifpp		/* XXX: just for statistics */
163 )
164 {
165 	struct ip6_hdr *ip6, *mhip6;
166 	struct ifnet *ifp = NULL, *origifp = NULL;
167 	struct mbuf *m = m0;
168 	int hlen, tlen, len, off;
169 	bool tso;
170 	struct route ip6route;
171 	struct rtentry *rt = NULL;
172 	const struct sockaddr_in6 *dst;
173 	struct sockaddr_in6 src_sa, dst_sa;
174 	int error = 0;
175 	struct in6_ifaddr *ia = NULL;
176 	u_long mtu;
177 	int alwaysfrag, dontfrag;
178 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
179 	struct ip6_exthdrs exthdrs;
180 	struct in6_addr finaldst, src0, dst0;
181 	u_int32_t zone;
182 	struct route *ro_pmtu = NULL;
183 	int hdrsplit = 0;
184 	int needipsec = 0;
185 #ifdef IPSEC
186 	struct secpolicy *sp = NULL;
187 #endif
188 	struct psref psref, psref_ia;
189 	int bound = curlwp_bind();
190 	bool release_psref_ia = false;
191 
192 #ifdef  DIAGNOSTIC
193 	if ((m->m_flags & M_PKTHDR) == 0)
194 		panic("ip6_output: no HDR");
195 
196 	if ((m->m_pkthdr.csum_flags &
197 	    (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) {
198 		panic("ip6_output: IPv4 checksum offload flags: %d",
199 		    m->m_pkthdr.csum_flags);
200 	}
201 
202 	if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) ==
203 	    (M_CSUM_TCPv6|M_CSUM_UDPv6)) {
204 		panic("ip6_output: conflicting checksum offload flags: %d",
205 		    m->m_pkthdr.csum_flags);
206 	}
207 #endif
208 
209 	M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr));
210 
211 #define MAKE_EXTHDR(hp, mp)						\
212     do {								\
213 	if (hp) {							\
214 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
215 		error = ip6_copyexthdr((mp), (void *)(hp), 		\
216 		    ((eh)->ip6e_len + 1) << 3);				\
217 		if (error)						\
218 			goto freehdrs;					\
219 	}								\
220     } while (/*CONSTCOND*/ 0)
221 
222 	memset(&exthdrs, 0, sizeof(exthdrs));
223 	if (opt) {
224 		/* Hop-by-Hop options header */
225 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
226 		/* Destination options header(1st part) */
227 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
228 		/* Routing header */
229 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
230 		/* Destination options header(2nd part) */
231 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
232 	}
233 
234 	/*
235 	 * Calculate the total length of the extension header chain.
236 	 * Keep the length of the unfragmentable part for fragmentation.
237 	 */
238 	optlen = 0;
239 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
240 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
241 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
242 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
243 	/* NOTE: we don't add AH/ESP length here. do that later. */
244 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
245 
246 #ifdef IPSEC
247 	if (ipsec_used) {
248 		/* Check the security policy (SP) for the packet */
249 
250 		sp = ipsec6_check_policy(m, so, flags, &needipsec, &error);
251 		if (error != 0) {
252 			/*
253 			 * Hack: -EINVAL is used to signal that a packet
254 			 * should be silently discarded.  This is typically
255 			 * because we asked key management for an SA and
256 			 * it was delayed (e.g. kicked up to IKE).
257 			 */
258 			if (error == -EINVAL)
259 				error = 0;
260 			goto freehdrs;
261 		}
262 	}
263 #endif /* IPSEC */
264 
265 
266 	if (needipsec &&
267 	    (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
268 		in6_delayed_cksum(m);
269 		m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
270 	}
271 
272 
273 	/*
274 	 * If we need IPsec, or there is at least one extension header,
275 	 * separate IP6 header from the payload.
276 	 */
277 	if ((needipsec || optlen) && !hdrsplit) {
278 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
279 			m = NULL;
280 			goto freehdrs;
281 		}
282 		m = exthdrs.ip6e_ip6;
283 		hdrsplit++;
284 	}
285 
286 	/* adjust pointer */
287 	ip6 = mtod(m, struct ip6_hdr *);
288 
289 	/* adjust mbuf packet header length */
290 	m->m_pkthdr.len += optlen;
291 	plen = m->m_pkthdr.len - sizeof(*ip6);
292 
293 	/* If this is a jumbo payload, insert a jumbo payload option. */
294 	if (plen > IPV6_MAXPACKET) {
295 		if (!hdrsplit) {
296 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
297 				m = NULL;
298 				goto freehdrs;
299 			}
300 			m = exthdrs.ip6e_ip6;
301 			hdrsplit++;
302 		}
303 		/* adjust pointer */
304 		ip6 = mtod(m, struct ip6_hdr *);
305 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
306 			goto freehdrs;
307 		optlen += 8; /* XXX JUMBOOPTLEN */
308 		ip6->ip6_plen = 0;
309 	} else
310 		ip6->ip6_plen = htons(plen);
311 
312 	/*
313 	 * Concatenate headers and fill in next header fields.
314 	 * Here we have, on "m"
315 	 *	IPv6 payload
316 	 * and we insert headers accordingly.  Finally, we should be getting:
317 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
318 	 *
319 	 * during the header composing process, "m" points to IPv6 header.
320 	 * "mprev" points to an extension header prior to esp.
321 	 */
322 	{
323 		u_char *nexthdrp = &ip6->ip6_nxt;
324 		struct mbuf *mprev = m;
325 
326 		/*
327 		 * we treat dest2 specially.  this makes IPsec processing
328 		 * much easier.  the goal here is to make mprev point the
329 		 * mbuf prior to dest2.
330 		 *
331 		 * result: IPv6 dest2 payload
332 		 * m and mprev will point to IPv6 header.
333 		 */
334 		if (exthdrs.ip6e_dest2) {
335 			if (!hdrsplit)
336 				panic("assumption failed: hdr not split");
337 			exthdrs.ip6e_dest2->m_next = m->m_next;
338 			m->m_next = exthdrs.ip6e_dest2;
339 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
340 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
341 		}
342 
343 #define MAKE_CHAIN(m, mp, p, i)\
344     do {\
345 	if (m) {\
346 		if (!hdrsplit) \
347 			panic("assumption failed: hdr not split"); \
348 		*mtod((m), u_char *) = *(p);\
349 		*(p) = (i);\
350 		p = mtod((m), u_char *);\
351 		(m)->m_next = (mp)->m_next;\
352 		(mp)->m_next = (m);\
353 		(mp) = (m);\
354 	}\
355     } while (/*CONSTCOND*/ 0)
356 		/*
357 		 * result: IPv6 hbh dest1 rthdr dest2 payload
358 		 * m will point to IPv6 header.  mprev will point to the
359 		 * extension header prior to dest2 (rthdr in the above case).
360 		 */
361 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
362 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
363 		    IPPROTO_DSTOPTS);
364 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
365 		    IPPROTO_ROUTING);
366 
367 		M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data,
368 		    sizeof(struct ip6_hdr) + optlen);
369 	}
370 
371 	/*
372 	 * If there is a routing header, replace destination address field
373 	 * with the first hop of the routing header.
374 	 */
375 	if (exthdrs.ip6e_rthdr) {
376 		struct ip6_rthdr *rh;
377 		struct ip6_rthdr0 *rh0;
378 		struct in6_addr *addr;
379 		struct sockaddr_in6 sa;
380 
381 		rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
382 		    struct ip6_rthdr *));
383 		finaldst = ip6->ip6_dst;
384 		switch (rh->ip6r_type) {
385 		case IPV6_RTHDR_TYPE_0:
386 			 rh0 = (struct ip6_rthdr0 *)rh;
387 			 addr = (struct in6_addr *)(rh0 + 1);
388 
389 			 /*
390 			  * construct a sockaddr_in6 form of
391 			  * the first hop.
392 			  *
393 			  * XXX: we may not have enough
394 			  * information about its scope zone;
395 			  * there is no standard API to pass
396 			  * the information from the
397 			  * application.
398 			  */
399 			 sockaddr_in6_init(&sa, addr, 0, 0, 0);
400 			 if ((error = sa6_embedscope(&sa,
401 			     ip6_use_defzone)) != 0) {
402 				 goto bad;
403 			 }
404 			 ip6->ip6_dst = sa.sin6_addr;
405 			 (void)memmove(&addr[0], &addr[1],
406 			     sizeof(struct in6_addr) *
407 			     (rh0->ip6r0_segleft - 1));
408 			 addr[rh0->ip6r0_segleft - 1] = finaldst;
409 			 /* XXX */
410 			 in6_clearscope(addr + rh0->ip6r0_segleft - 1);
411 			 break;
412 		default:	/* is it possible? */
413 			 error = EINVAL;
414 			 goto bad;
415 		}
416 	}
417 
418 	/* Source address validation */
419 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
420 	    (flags & IPV6_UNSPECSRC) == 0) {
421 		error = EOPNOTSUPP;
422 		IP6_STATINC(IP6_STAT_BADSCOPE);
423 		goto bad;
424 	}
425 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
426 		error = EOPNOTSUPP;
427 		IP6_STATINC(IP6_STAT_BADSCOPE);
428 		goto bad;
429 	}
430 
431 	IP6_STATINC(IP6_STAT_LOCALOUT);
432 
433 	/*
434 	 * Route packet.
435 	 */
436 	/* initialize cached route */
437 	if (ro == NULL) {
438 		memset(&ip6route, 0, sizeof(ip6route));
439 		ro = &ip6route;
440 	}
441 	ro_pmtu = ro;
442 	if (opt && opt->ip6po_rthdr)
443 		ro = &opt->ip6po_route;
444 
445  	/*
446 	 * if specified, try to fill in the traffic class field.
447 	 * do not override if a non-zero value is already set.
448 	 * we check the diffserv field and the ecn field separately.
449 	 */
450 	if (opt && opt->ip6po_tclass >= 0) {
451 		int mask = 0;
452 
453 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
454 			mask |= 0xfc;
455 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
456 			mask |= 0x03;
457 		if (mask != 0)
458 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
459 	}
460 
461 	/* fill in or override the hop limit field, if necessary. */
462 	if (opt && opt->ip6po_hlim != -1)
463 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
464 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
465 		if (im6o != NULL)
466 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
467 		else
468 			ip6->ip6_hlim = ip6_defmcasthlim;
469 	}
470 
471 #ifdef IPSEC
472 	if (needipsec) {
473 		int s = splsoftnet();
474 		error = ipsec6_process_packet(m, sp->req);
475 
476 		/*
477 		 * Preserve KAME behaviour: ENOENT can be returned
478 		 * when an SA acquire is in progress.  Don't propagate
479 		 * this to user-level; it confuses applications.
480 		 * XXX this will go away when the SADB is redone.
481 		 */
482 		if (error == ENOENT)
483 			error = 0;
484 		splx(s);
485 		goto done;
486 	}
487 #endif /* IPSEC */
488 
489 	/* adjust pointer */
490 	ip6 = mtod(m, struct ip6_hdr *);
491 
492 	sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
493 	if ((error = in6_selectroute(&dst_sa, opt, im6o, ro,
494 	    &ifp, &psref, &rt, 0)) != 0) {
495 		if (ifp != NULL)
496 			in6_ifstat_inc(ifp, ifs6_out_discard);
497 		goto bad;
498 	}
499 	if (rt == NULL) {
500 		/*
501 		 * If in6_selectroute() does not return a route entry,
502 		 * dst may not have been updated.
503 		 */
504 		error = rtcache_setdst(ro, sin6tosa(&dst_sa));
505 		if (error) {
506 			goto bad;
507 		}
508 	}
509 
510 	/*
511 	 * then rt (for unicast) and ifp must be non-NULL valid values.
512 	 */
513 	if ((flags & IPV6_FORWARDING) == 0) {
514 		/* XXX: the FORWARDING flag can be set for mrouting. */
515 		in6_ifstat_inc(ifp, ifs6_out_request);
516 	}
517 	if (rt != NULL) {
518 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
519 		rt->rt_use++;
520 	}
521 
522 	/*
523 	 * The outgoing interface must be in the zone of source and
524 	 * destination addresses.  We should use ia_ifp to support the
525 	 * case of sending packets to an address of our own.
526 	 */
527 	if (ia != NULL && ia->ia_ifp) {
528 		origifp = ia->ia_ifp;
529 		if (if_is_deactivated(origifp))
530 			goto bad;
531 		if_acquire_NOMPSAFE(origifp, &psref_ia);
532 		release_psref_ia = true;
533 	} else
534 		origifp = ifp;
535 
536 	src0 = ip6->ip6_src;
537 	if (in6_setscope(&src0, origifp, &zone))
538 		goto badscope;
539 	sockaddr_in6_init(&src_sa, &ip6->ip6_src, 0, 0, 0);
540 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
541 		goto badscope;
542 
543 	dst0 = ip6->ip6_dst;
544 	if (in6_setscope(&dst0, origifp, &zone))
545 		goto badscope;
546 	/* re-initialize to be sure */
547 	sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
548 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id)
549 		goto badscope;
550 
551 	/* scope check is done. */
552 
553 	if (rt == NULL || IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
554 		dst = satocsin6(rtcache_getdst(ro));
555 		KASSERT(dst != NULL);
556 	} else if (opt && rtcache_validate(&opt->ip6po_nextroute) != NULL) {
557 		/*
558 		 * The nexthop is explicitly specified by the
559 		 * application.  We assume the next hop is an IPv6
560 		 * address.
561 		 */
562 		dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
563 	} else if ((rt->rt_flags & RTF_GATEWAY))
564 		dst = (struct sockaddr_in6 *)rt->rt_gateway;
565 	else
566 		dst = satocsin6(rtcache_getdst(ro));
567 
568 	/*
569 	 * XXXXXX: original code follows:
570 	 */
571 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
572 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
573 	else {
574 		struct	in6_multi *in6m;
575 
576 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
577 
578 		in6_ifstat_inc(ifp, ifs6_out_mcast);
579 
580 		/*
581 		 * Confirm that the outgoing interface supports multicast.
582 		 */
583 		if (!(ifp->if_flags & IFF_MULTICAST)) {
584 			IP6_STATINC(IP6_STAT_NOROUTE);
585 			in6_ifstat_inc(ifp, ifs6_out_discard);
586 			error = ENETUNREACH;
587 			goto bad;
588 		}
589 
590 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
591 		if (in6m != NULL &&
592 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
593 			/*
594 			 * If we belong to the destination multicast group
595 			 * on the outgoing interface, and the caller did not
596 			 * forbid loopback, loop back a copy.
597 			 */
598 			KASSERT(dst != NULL);
599 			ip6_mloopback(ifp, m, dst);
600 		} else {
601 			/*
602 			 * If we are acting as a multicast router, perform
603 			 * multicast forwarding as if the packet had just
604 			 * arrived on the interface to which we are about
605 			 * to send.  The multicast forwarding function
606 			 * recursively calls this function, using the
607 			 * IPV6_FORWARDING flag to prevent infinite recursion.
608 			 *
609 			 * Multicasts that are looped back by ip6_mloopback(),
610 			 * above, will be forwarded by the ip6_input() routine,
611 			 * if necessary.
612 			 */
613 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
614 				if (ip6_mforward(ip6, ifp, m) != 0) {
615 					m_freem(m);
616 					goto done;
617 				}
618 			}
619 		}
620 		/*
621 		 * Multicasts with a hoplimit of zero may be looped back,
622 		 * above, but must not be transmitted on a network.
623 		 * Also, multicasts addressed to the loopback interface
624 		 * are not sent -- the above call to ip6_mloopback() will
625 		 * loop back a copy if this host actually belongs to the
626 		 * destination group on the loopback interface.
627 		 */
628 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
629 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
630 			m_freem(m);
631 			goto done;
632 		}
633 	}
634 
635 	/*
636 	 * Fill the outgoing inteface to tell the upper layer
637 	 * to increment per-interface statistics.
638 	 */
639 	if (ifpp)
640 		*ifpp = ifp;
641 
642 	/* Determine path MTU. */
643 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
644 	    &alwaysfrag)) != 0)
645 		goto bad;
646 
647 	/*
648 	 * The caller of this function may specify to use the minimum MTU
649 	 * in some cases.
650 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
651 	 * setting.  The logic is a bit complicated; by default, unicast
652 	 * packets will follow path MTU while multicast packets will be sent at
653 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
654 	 * including unicast ones will be sent at the minimum MTU.  Multicast
655 	 * packets will always be sent at the minimum MTU unless
656 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
657 	 * See RFC 3542 for more details.
658 	 */
659 	if (mtu > IPV6_MMTU) {
660 		if ((flags & IPV6_MINMTU))
661 			mtu = IPV6_MMTU;
662 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
663 			mtu = IPV6_MMTU;
664 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
665 			 (opt == NULL ||
666 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
667 			mtu = IPV6_MMTU;
668 		}
669 	}
670 
671 	/*
672 	 * clear embedded scope identifiers if necessary.
673 	 * in6_clearscope will touch the addresses only when necessary.
674 	 */
675 	in6_clearscope(&ip6->ip6_src);
676 	in6_clearscope(&ip6->ip6_dst);
677 
678 	/*
679 	 * If the outgoing packet contains a hop-by-hop options header,
680 	 * it must be examined and processed even by the source node.
681 	 * (RFC 2460, section 4.)
682 	 */
683 	if (ip6->ip6_nxt == IPV6_HOPOPTS) {
684 		u_int32_t dummy1; /* XXX unused */
685 		u_int32_t dummy2; /* XXX unused */
686 		int hoff = sizeof(struct ip6_hdr);
687 
688 		if (ip6_hopopts_input(&dummy1, &dummy2, &m, &hoff)) {
689 			/* m was already freed at this point */
690 			error = EINVAL;/* better error? */
691 			goto done;
692 		}
693 
694 		ip6 = mtod(m, struct ip6_hdr *);
695 	}
696 
697 	/*
698 	 * Run through list of hooks for output packets.
699 	 */
700 	if ((error = pfil_run_hooks(inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0)
701 		goto done;
702 	if (m == NULL)
703 		goto done;
704 	ip6 = mtod(m, struct ip6_hdr *);
705 
706 	/*
707 	 * Send the packet to the outgoing interface.
708 	 * If necessary, do IPv6 fragmentation before sending.
709 	 *
710 	 * the logic here is rather complex:
711 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
712 	 * 1-a:	send as is if tlen <= path mtu
713 	 * 1-b:	fragment if tlen > path mtu
714 	 *
715 	 * 2: if user asks us not to fragment (dontfrag == 1)
716 	 * 2-a:	send as is if tlen <= interface mtu
717 	 * 2-b:	error if tlen > interface mtu
718 	 *
719 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
720 	 *	always fragment
721 	 *
722 	 * 4: if dontfrag == 1 && alwaysfrag == 1
723 	 *	error, as we cannot handle this conflicting request
724 	 */
725 	tlen = m->m_pkthdr.len;
726 	tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0;
727 	if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
728 		dontfrag = 1;
729 	else
730 		dontfrag = 0;
731 
732 	if (dontfrag && alwaysfrag) {	/* case 4 */
733 		/* conflicting request - can't transmit */
734 		error = EMSGSIZE;
735 		goto bad;
736 	}
737 	if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) {	/* case 2-b */
738 		/*
739 		 * Even if the DONTFRAG option is specified, we cannot send the
740 		 * packet when the data length is larger than the MTU of the
741 		 * outgoing interface.
742 		 * Notify the error by sending IPV6_PATHMTU ancillary data as
743 		 * well as returning an error code (the latter is not described
744 		 * in the API spec.)
745 		 */
746 		u_int32_t mtu32;
747 		struct ip6ctlparam ip6cp;
748 
749 		mtu32 = (u_int32_t)mtu;
750 		memset(&ip6cp, 0, sizeof(ip6cp));
751 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
752 		pfctlinput2(PRC_MSGSIZE,
753 		    rtcache_getdst(ro_pmtu), &ip6cp);
754 
755 		error = EMSGSIZE;
756 		goto bad;
757 	}
758 
759 	/*
760 	 * transmit packet without fragmentation
761 	 */
762 	if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) {
763 		/* case 1-a and 2-a */
764 		struct in6_ifaddr *ia6;
765 		int sw_csum;
766 		int s;
767 
768 		ip6 = mtod(m, struct ip6_hdr *);
769 		s = pserialize_read_enter();
770 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
771 		if (ia6) {
772 			/* Record statistics for this interface address. */
773 			ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len;
774 		}
775 		pserialize_read_exit(s);
776 
777 		sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
778 		if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
779 			if (IN6_NEED_CHECKSUM(ifp,
780 			    sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) {
781 				in6_delayed_cksum(m);
782 			}
783 			m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
784 		}
785 
786 		KASSERT(dst != NULL);
787 		if (__predict_true(!tso ||
788 		    (ifp->if_capenable & IFCAP_TSOv6) != 0)) {
789 			error = nd6_output(ifp, origifp, m, dst, rt);
790 		} else {
791 			error = ip6_tso_output(ifp, origifp, m, dst, rt);
792 		}
793 		goto done;
794 	}
795 
796 	if (tso) {
797 		error = EINVAL; /* XXX */
798 		goto bad;
799 	}
800 
801 	/*
802 	 * try to fragment the packet.  case 1-b and 3
803 	 */
804 	if (mtu < IPV6_MMTU) {
805 		/* path MTU cannot be less than IPV6_MMTU */
806 		error = EMSGSIZE;
807 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
808 		goto bad;
809 	} else if (ip6->ip6_plen == 0) {
810 		/* jumbo payload cannot be fragmented */
811 		error = EMSGSIZE;
812 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
813 		goto bad;
814 	} else {
815 		struct mbuf **mnext, *m_frgpart;
816 		struct ip6_frag *ip6f;
817 		u_int32_t id = htonl(ip6_randomid());
818 		u_char nextproto;
819 #if 0				/* see below */
820 		struct ip6ctlparam ip6cp;
821 		u_int32_t mtu32;
822 #endif
823 
824 		/*
825 		 * Too large for the destination or interface;
826 		 * fragment if possible.
827 		 * Must be able to put at least 8 bytes per fragment.
828 		 */
829 		hlen = unfragpartlen;
830 		if (mtu > IPV6_MAXPACKET)
831 			mtu = IPV6_MAXPACKET;
832 
833 #if 0
834 		/*
835 		 * It is believed this code is a leftover from the
836 		 * development of the IPV6_RECVPATHMTU sockopt and
837 		 * associated work to implement RFC3542.
838 		 * It's not entirely clear what the intent of the API
839 		 * is at this point, so disable this code for now.
840 		 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG
841 		 * will send notifications if the application requests.
842 		 */
843 
844 		/* Notify a proper path MTU to applications. */
845 		mtu32 = (u_int32_t)mtu;
846 		memset(&ip6cp, 0, sizeof(ip6cp));
847 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
848 		pfctlinput2(PRC_MSGSIZE,
849 		    rtcache_getdst(ro_pmtu), &ip6cp);
850 #endif
851 
852 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
853 		if (len < 8) {
854 			error = EMSGSIZE;
855 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
856 			goto bad;
857 		}
858 
859 		mnext = &m->m_nextpkt;
860 
861 		/*
862 		 * Change the next header field of the last header in the
863 		 * unfragmentable part.
864 		 */
865 		if (exthdrs.ip6e_rthdr) {
866 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
867 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
868 		} else if (exthdrs.ip6e_dest1) {
869 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
870 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
871 		} else if (exthdrs.ip6e_hbh) {
872 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
873 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
874 		} else {
875 			nextproto = ip6->ip6_nxt;
876 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
877 		}
878 
879 		if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6))
880 		    != 0) {
881 			if (IN6_NEED_CHECKSUM(ifp,
882 			    m->m_pkthdr.csum_flags &
883 			    (M_CSUM_UDPv6|M_CSUM_TCPv6))) {
884 				in6_delayed_cksum(m);
885 			}
886 			m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
887 		}
888 
889 		/*
890 		 * Loop through length of segment after first fragment,
891 		 * make new header and copy data of each part and link onto
892 		 * chain.
893 		 */
894 		m0 = m;
895 		for (off = hlen; off < tlen; off += len) {
896 			struct mbuf *mlast;
897 
898 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
899 			if (!m) {
900 				error = ENOBUFS;
901 				IP6_STATINC(IP6_STAT_ODROPPED);
902 				goto sendorfree;
903 			}
904 			m_reset_rcvif(m);
905 			m->m_flags = m0->m_flags & M_COPYFLAGS;
906 			*mnext = m;
907 			mnext = &m->m_nextpkt;
908 			m->m_data += max_linkhdr;
909 			mhip6 = mtod(m, struct ip6_hdr *);
910 			*mhip6 = *ip6;
911 			m->m_len = sizeof(*mhip6);
912 			/*
913 			 * ip6f must be valid if error is 0.  But how
914 			 * can a compiler be expected to infer this?
915 			 */
916 			ip6f = NULL;
917 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
918 			if (error) {
919 				IP6_STATINC(IP6_STAT_ODROPPED);
920 				goto sendorfree;
921 			}
922 			ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7));
923 			if (off + len >= tlen)
924 				len = tlen - off;
925 			else
926 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
927 			mhip6->ip6_plen = htons((u_int16_t)(len + hlen +
928 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
929 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
930 				error = ENOBUFS;
931 				IP6_STATINC(IP6_STAT_ODROPPED);
932 				goto sendorfree;
933 			}
934 			for (mlast = m; mlast->m_next; mlast = mlast->m_next)
935 				;
936 			mlast->m_next = m_frgpart;
937 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
938 			m_reset_rcvif(m);
939 			ip6f->ip6f_reserved = 0;
940 			ip6f->ip6f_ident = id;
941 			ip6f->ip6f_nxt = nextproto;
942 			IP6_STATINC(IP6_STAT_OFRAGMENTS);
943 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
944 		}
945 
946 		in6_ifstat_inc(ifp, ifs6_out_fragok);
947 	}
948 
949 	/*
950 	 * Remove leading garbages.
951 	 */
952 sendorfree:
953 	m = m0->m_nextpkt;
954 	m0->m_nextpkt = 0;
955 	m_freem(m0);
956 	for (m0 = m; m; m = m0) {
957 		m0 = m->m_nextpkt;
958 		m->m_nextpkt = 0;
959 		if (error == 0) {
960 			struct in6_ifaddr *ia6;
961 			int s;
962 			ip6 = mtod(m, struct ip6_hdr *);
963 			s = pserialize_read_enter();
964 			ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
965 			if (ia6) {
966 				/*
967 				 * Record statistics for this interface
968 				 * address.
969 				 */
970 				ia6->ia_ifa.ifa_data.ifad_outbytes +=
971 				    m->m_pkthdr.len;
972 			}
973 			pserialize_read_exit(s);
974 			KASSERT(dst != NULL);
975 			error = nd6_output(ifp, origifp, m, dst, rt);
976 		} else
977 			m_freem(m);
978 	}
979 
980 	if (error == 0)
981 		IP6_STATINC(IP6_STAT_FRAGMENTED);
982 
983 done:
984 	if (ro == &ip6route)
985 		rtcache_free(&ip6route);
986 
987 #ifdef IPSEC
988 	if (sp != NULL)
989 		KEY_FREESP(&sp);
990 #endif /* IPSEC */
991 
992 	if_put(ifp, &psref);
993 	if (release_psref_ia)
994 		if_put(origifp, &psref_ia);
995 	curlwp_bindx(bound);
996 
997 	return (error);
998 
999 freehdrs:
1000 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1001 	m_freem(exthdrs.ip6e_dest1);
1002 	m_freem(exthdrs.ip6e_rthdr);
1003 	m_freem(exthdrs.ip6e_dest2);
1004 	/* FALLTHROUGH */
1005 bad:
1006 	m_freem(m);
1007 	goto done;
1008 badscope:
1009 	IP6_STATINC(IP6_STAT_BADSCOPE);
1010 	in6_ifstat_inc(origifp, ifs6_out_discard);
1011 	if (error == 0)
1012 		error = EHOSTUNREACH; /* XXX */
1013 	goto bad;
1014 }
1015 
1016 static int
1017 ip6_copyexthdr(struct mbuf **mp, void *hdr, int hlen)
1018 {
1019 	struct mbuf *m;
1020 
1021 	if (hlen > MCLBYTES)
1022 		return (ENOBUFS); /* XXX */
1023 
1024 	MGET(m, M_DONTWAIT, MT_DATA);
1025 	if (!m)
1026 		return (ENOBUFS);
1027 
1028 	if (hlen > MLEN) {
1029 		MCLGET(m, M_DONTWAIT);
1030 		if ((m->m_flags & M_EXT) == 0) {
1031 			m_free(m);
1032 			return (ENOBUFS);
1033 		}
1034 	}
1035 	m->m_len = hlen;
1036 	if (hdr)
1037 		bcopy(hdr, mtod(m, void *), hlen);
1038 
1039 	*mp = m;
1040 	return (0);
1041 }
1042 
1043 /*
1044  * Process a delayed payload checksum calculation.
1045  */
1046 void
1047 in6_delayed_cksum(struct mbuf *m)
1048 {
1049 	uint16_t csum, offset;
1050 
1051 	KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0);
1052 	KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0);
1053 	KASSERT((m->m_pkthdr.csum_flags
1054 	    & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0);
1055 
1056 	offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data);
1057 	csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset);
1058 	if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) {
1059 		csum = 0xffff;
1060 	}
1061 
1062 	offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data);
1063 	if ((offset + sizeof(csum)) > m->m_len) {
1064 		m_copyback(m, offset, sizeof(csum), &csum);
1065 	} else {
1066 		*(uint16_t *)(mtod(m, char *) + offset) = csum;
1067 	}
1068 }
1069 
1070 /*
1071  * Insert jumbo payload option.
1072  */
1073 static int
1074 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1075 {
1076 	struct mbuf *mopt;
1077 	u_int8_t *optbuf;
1078 	u_int32_t v;
1079 
1080 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1081 
1082 	/*
1083 	 * If there is no hop-by-hop options header, allocate new one.
1084 	 * If there is one but it doesn't have enough space to store the
1085 	 * jumbo payload option, allocate a cluster to store the whole options.
1086 	 * Otherwise, use it to store the options.
1087 	 */
1088 	if (exthdrs->ip6e_hbh == 0) {
1089 		MGET(mopt, M_DONTWAIT, MT_DATA);
1090 		if (mopt == 0)
1091 			return (ENOBUFS);
1092 		mopt->m_len = JUMBOOPTLEN;
1093 		optbuf = mtod(mopt, u_int8_t *);
1094 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1095 		exthdrs->ip6e_hbh = mopt;
1096 	} else {
1097 		struct ip6_hbh *hbh;
1098 
1099 		mopt = exthdrs->ip6e_hbh;
1100 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1101 			/*
1102 			 * XXX assumption:
1103 			 * - exthdrs->ip6e_hbh is not referenced from places
1104 			 *   other than exthdrs.
1105 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1106 			 */
1107 			int oldoptlen = mopt->m_len;
1108 			struct mbuf *n;
1109 
1110 			/*
1111 			 * XXX: give up if the whole (new) hbh header does
1112 			 * not fit even in an mbuf cluster.
1113 			 */
1114 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1115 				return (ENOBUFS);
1116 
1117 			/*
1118 			 * As a consequence, we must always prepare a cluster
1119 			 * at this point.
1120 			 */
1121 			MGET(n, M_DONTWAIT, MT_DATA);
1122 			if (n) {
1123 				MCLGET(n, M_DONTWAIT);
1124 				if ((n->m_flags & M_EXT) == 0) {
1125 					m_freem(n);
1126 					n = NULL;
1127 				}
1128 			}
1129 			if (!n)
1130 				return (ENOBUFS);
1131 			n->m_len = oldoptlen + JUMBOOPTLEN;
1132 			bcopy(mtod(mopt, void *), mtod(n, void *),
1133 			    oldoptlen);
1134 			optbuf = mtod(n, u_int8_t *) + oldoptlen;
1135 			m_freem(mopt);
1136 			mopt = exthdrs->ip6e_hbh = n;
1137 		} else {
1138 			optbuf = mtod(mopt, u_int8_t *) + mopt->m_len;
1139 			mopt->m_len += JUMBOOPTLEN;
1140 		}
1141 		optbuf[0] = IP6OPT_PADN;
1142 		optbuf[1] = 0;
1143 
1144 		/*
1145 		 * Adjust the header length according to the pad and
1146 		 * the jumbo payload option.
1147 		 */
1148 		hbh = mtod(mopt, struct ip6_hbh *);
1149 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1150 	}
1151 
1152 	/* fill in the option. */
1153 	optbuf[2] = IP6OPT_JUMBO;
1154 	optbuf[3] = 4;
1155 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1156 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1157 
1158 	/* finally, adjust the packet header length */
1159 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1160 
1161 	return (0);
1162 #undef JUMBOOPTLEN
1163 }
1164 
1165 /*
1166  * Insert fragment header and copy unfragmentable header portions.
1167  *
1168  * *frghdrp will not be read, and it is guaranteed that either an
1169  * error is returned or that *frghdrp will point to space allocated
1170  * for the fragment header.
1171  */
1172 static int
1173 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1174 	struct ip6_frag **frghdrp)
1175 {
1176 	struct mbuf *n, *mlast;
1177 
1178 	if (hlen > sizeof(struct ip6_hdr)) {
1179 		n = m_copym(m0, sizeof(struct ip6_hdr),
1180 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1181 		if (n == 0)
1182 			return (ENOBUFS);
1183 		m->m_next = n;
1184 	} else
1185 		n = m;
1186 
1187 	/* Search for the last mbuf of unfragmentable part. */
1188 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1189 		;
1190 
1191 	if ((mlast->m_flags & M_EXT) == 0 &&
1192 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1193 		/* use the trailing space of the last mbuf for the fragment hdr */
1194 		*frghdrp = (struct ip6_frag *)(mtod(mlast, char *) +
1195 		    mlast->m_len);
1196 		mlast->m_len += sizeof(struct ip6_frag);
1197 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1198 	} else {
1199 		/* allocate a new mbuf for the fragment header */
1200 		struct mbuf *mfrg;
1201 
1202 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1203 		if (mfrg == 0)
1204 			return (ENOBUFS);
1205 		mfrg->m_len = sizeof(struct ip6_frag);
1206 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1207 		mlast->m_next = mfrg;
1208 	}
1209 
1210 	return (0);
1211 }
1212 
1213 static int
1214 ip6_getpmtu(struct route *ro_pmtu, struct route *ro, struct ifnet *ifp,
1215     const struct in6_addr *dst, u_long *mtup, int *alwaysfragp)
1216 {
1217 	struct rtentry *rt;
1218 	u_int32_t mtu = 0;
1219 	int alwaysfrag = 0;
1220 	int error = 0;
1221 
1222 	if (ro_pmtu != ro) {
1223 		union {
1224 			struct sockaddr		dst;
1225 			struct sockaddr_in6	dst6;
1226 		} u;
1227 
1228 		/* The first hop and the final destination may differ. */
1229 		sockaddr_in6_init(&u.dst6, dst, 0, 0, 0);
1230 		rt = rtcache_lookup(ro_pmtu, &u.dst);
1231 	} else
1232 		rt = rtcache_validate(ro_pmtu);
1233 	if (rt != NULL) {
1234 		u_int32_t ifmtu;
1235 
1236 		if (ifp == NULL)
1237 			ifp = rt->rt_ifp;
1238 		ifmtu = IN6_LINKMTU(ifp);
1239 		mtu = rt->rt_rmx.rmx_mtu;
1240 		if (mtu == 0)
1241 			mtu = ifmtu;
1242 		else if (mtu < IPV6_MMTU) {
1243 			/*
1244 			 * RFC2460 section 5, last paragraph:
1245 			 * if we record ICMPv6 too big message with
1246 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1247 			 * or smaller, with fragment header attached.
1248 			 * (fragment header is needed regardless from the
1249 			 * packet size, for translators to identify packets)
1250 			 */
1251 			alwaysfrag = 1;
1252 			mtu = IPV6_MMTU;
1253 		} else if (mtu > ifmtu) {
1254 			/*
1255 			 * The MTU on the route is larger than the MTU on
1256 			 * the interface!  This shouldn't happen, unless the
1257 			 * MTU of the interface has been changed after the
1258 			 * interface was brought up.  Change the MTU in the
1259 			 * route to match the interface MTU (as long as the
1260 			 * field isn't locked).
1261 			 */
1262 			mtu = ifmtu;
1263 			if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
1264 				rt->rt_rmx.rmx_mtu = mtu;
1265 		}
1266 	} else if (ifp) {
1267 		mtu = IN6_LINKMTU(ifp);
1268 	} else
1269 		error = EHOSTUNREACH; /* XXX */
1270 
1271 	*mtup = mtu;
1272 	if (alwaysfragp)
1273 		*alwaysfragp = alwaysfrag;
1274 	return (error);
1275 }
1276 
1277 /*
1278  * IP6 socket option processing.
1279  */
1280 int
1281 ip6_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1282 {
1283 	int optdatalen, uproto;
1284 	void *optdata;
1285 	struct in6pcb *in6p = sotoin6pcb(so);
1286 	struct ip_moptions **mopts;
1287 	int error, optval;
1288 	int level, optname;
1289 
1290 	KASSERT(sopt != NULL);
1291 
1292 	level = sopt->sopt_level;
1293 	optname = sopt->sopt_name;
1294 
1295 	error = optval = 0;
1296 	uproto = (int)so->so_proto->pr_protocol;
1297 
1298 	switch (level) {
1299 	case IPPROTO_IP:
1300 		switch (optname) {
1301 		case IP_ADD_MEMBERSHIP:
1302 		case IP_DROP_MEMBERSHIP:
1303 		case IP_MULTICAST_IF:
1304 		case IP_MULTICAST_LOOP:
1305 		case IP_MULTICAST_TTL:
1306 			mopts = &in6p->in6p_v4moptions;
1307 			switch (op) {
1308 			case PRCO_GETOPT:
1309 				return ip_getmoptions(*mopts, sopt);
1310 			case PRCO_SETOPT:
1311 				return ip_setmoptions(mopts, sopt);
1312 			default:
1313 				return EINVAL;
1314 			}
1315 		default:
1316 			return ENOPROTOOPT;
1317 		}
1318 	case IPPROTO_IPV6:
1319 		break;
1320 	default:
1321 		return ENOPROTOOPT;
1322 	}
1323 	switch (op) {
1324 	case PRCO_SETOPT:
1325 		switch (optname) {
1326 #ifdef RFC2292
1327 		case IPV6_2292PKTOPTIONS:
1328 			error = ip6_pcbopts(&in6p->in6p_outputopts, so, sopt);
1329 			break;
1330 #endif
1331 
1332 		/*
1333 		 * Use of some Hop-by-Hop options or some
1334 		 * Destination options, might require special
1335 		 * privilege.  That is, normal applications
1336 		 * (without special privilege) might be forbidden
1337 		 * from setting certain options in outgoing packets,
1338 		 * and might never see certain options in received
1339 		 * packets. [RFC 2292 Section 6]
1340 		 * KAME specific note:
1341 		 *  KAME prevents non-privileged users from sending or
1342 		 *  receiving ANY hbh/dst options in order to avoid
1343 		 *  overhead of parsing options in the kernel.
1344 		 */
1345 		case IPV6_RECVHOPOPTS:
1346 		case IPV6_RECVDSTOPTS:
1347 		case IPV6_RECVRTHDRDSTOPTS:
1348 			error = kauth_authorize_network(kauth_cred_get(),
1349 			    KAUTH_NETWORK_IPV6, KAUTH_REQ_NETWORK_IPV6_HOPBYHOP,
1350 			    NULL, NULL, NULL);
1351 			if (error)
1352 				break;
1353 			/* FALLTHROUGH */
1354 		case IPV6_UNICAST_HOPS:
1355 		case IPV6_HOPLIMIT:
1356 		case IPV6_FAITH:
1357 
1358 		case IPV6_RECVPKTINFO:
1359 		case IPV6_RECVHOPLIMIT:
1360 		case IPV6_RECVRTHDR:
1361 		case IPV6_RECVPATHMTU:
1362 		case IPV6_RECVTCLASS:
1363 		case IPV6_V6ONLY:
1364 			error = sockopt_getint(sopt, &optval);
1365 			if (error)
1366 				break;
1367 			switch (optname) {
1368 			case IPV6_UNICAST_HOPS:
1369 				if (optval < -1 || optval >= 256)
1370 					error = EINVAL;
1371 				else {
1372 					/* -1 = kernel default */
1373 					in6p->in6p_hops = optval;
1374 				}
1375 				break;
1376 #define OPTSET(bit) \
1377 do { \
1378 if (optval) \
1379 	in6p->in6p_flags |= (bit); \
1380 else \
1381 	in6p->in6p_flags &= ~(bit); \
1382 } while (/*CONSTCOND*/ 0)
1383 
1384 #ifdef RFC2292
1385 #define OPTSET2292(bit) 			\
1386 do { 						\
1387 in6p->in6p_flags |= IN6P_RFC2292; 	\
1388 if (optval) 				\
1389 	in6p->in6p_flags |= (bit); 	\
1390 else 					\
1391 	in6p->in6p_flags &= ~(bit); 	\
1392 } while (/*CONSTCOND*/ 0)
1393 #endif
1394 
1395 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1396 
1397 			case IPV6_RECVPKTINFO:
1398 #ifdef RFC2292
1399 				/* cannot mix with RFC2292 */
1400 				if (OPTBIT(IN6P_RFC2292)) {
1401 					error = EINVAL;
1402 					break;
1403 				}
1404 #endif
1405 				OPTSET(IN6P_PKTINFO);
1406 				break;
1407 
1408 			case IPV6_HOPLIMIT:
1409 			{
1410 				struct ip6_pktopts **optp;
1411 
1412 #ifdef RFC2292
1413 				/* cannot mix with RFC2292 */
1414 				if (OPTBIT(IN6P_RFC2292)) {
1415 					error = EINVAL;
1416 					break;
1417 				}
1418 #endif
1419 				optp = &in6p->in6p_outputopts;
1420 				error = ip6_pcbopt(IPV6_HOPLIMIT,
1421 						   (u_char *)&optval,
1422 						   sizeof(optval),
1423 						   optp,
1424 						   kauth_cred_get(), uproto);
1425 				break;
1426 			}
1427 
1428 			case IPV6_RECVHOPLIMIT:
1429 #ifdef RFC2292
1430 				/* cannot mix with RFC2292 */
1431 				if (OPTBIT(IN6P_RFC2292)) {
1432 					error = EINVAL;
1433 					break;
1434 				}
1435 #endif
1436 				OPTSET(IN6P_HOPLIMIT);
1437 				break;
1438 
1439 			case IPV6_RECVHOPOPTS:
1440 #ifdef RFC2292
1441 				/* cannot mix with RFC2292 */
1442 				if (OPTBIT(IN6P_RFC2292)) {
1443 					error = EINVAL;
1444 					break;
1445 				}
1446 #endif
1447 				OPTSET(IN6P_HOPOPTS);
1448 				break;
1449 
1450 			case IPV6_RECVDSTOPTS:
1451 #ifdef RFC2292
1452 				/* cannot mix with RFC2292 */
1453 				if (OPTBIT(IN6P_RFC2292)) {
1454 					error = EINVAL;
1455 					break;
1456 				}
1457 #endif
1458 				OPTSET(IN6P_DSTOPTS);
1459 				break;
1460 
1461 			case IPV6_RECVRTHDRDSTOPTS:
1462 #ifdef RFC2292
1463 				/* cannot mix with RFC2292 */
1464 				if (OPTBIT(IN6P_RFC2292)) {
1465 					error = EINVAL;
1466 					break;
1467 				}
1468 #endif
1469 				OPTSET(IN6P_RTHDRDSTOPTS);
1470 				break;
1471 
1472 			case IPV6_RECVRTHDR:
1473 #ifdef RFC2292
1474 				/* cannot mix with RFC2292 */
1475 				if (OPTBIT(IN6P_RFC2292)) {
1476 					error = EINVAL;
1477 					break;
1478 				}
1479 #endif
1480 				OPTSET(IN6P_RTHDR);
1481 				break;
1482 
1483 			case IPV6_FAITH:
1484 				OPTSET(IN6P_FAITH);
1485 				break;
1486 
1487 			case IPV6_RECVPATHMTU:
1488 				/*
1489 				 * We ignore this option for TCP
1490 				 * sockets.
1491 				 * (RFC3542 leaves this case
1492 				 * unspecified.)
1493 				 */
1494 				if (uproto != IPPROTO_TCP)
1495 					OPTSET(IN6P_MTU);
1496 				break;
1497 
1498 			case IPV6_V6ONLY:
1499 				/*
1500 				 * make setsockopt(IPV6_V6ONLY)
1501 				 * available only prior to bind(2).
1502 				 * see ipng mailing list, Jun 22 2001.
1503 				 */
1504 				if (in6p->in6p_lport ||
1505 				    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1506 					error = EINVAL;
1507 					break;
1508 				}
1509 #ifdef INET6_BINDV6ONLY
1510 				if (!optval)
1511 					error = EINVAL;
1512 #else
1513 				OPTSET(IN6P_IPV6_V6ONLY);
1514 #endif
1515 				break;
1516 			case IPV6_RECVTCLASS:
1517 #ifdef RFC2292
1518 				/* cannot mix with RFC2292 XXX */
1519 				if (OPTBIT(IN6P_RFC2292)) {
1520 					error = EINVAL;
1521 					break;
1522 				}
1523 #endif
1524 				OPTSET(IN6P_TCLASS);
1525 				break;
1526 
1527 			}
1528 			break;
1529 
1530 		case IPV6_OTCLASS:
1531 		{
1532 			struct ip6_pktopts **optp;
1533 			u_int8_t tclass;
1534 
1535 			error = sockopt_get(sopt, &tclass, sizeof(tclass));
1536 			if (error)
1537 				break;
1538 			optp = &in6p->in6p_outputopts;
1539 			error = ip6_pcbopt(optname,
1540 					   (u_char *)&tclass,
1541 					   sizeof(tclass),
1542 					   optp,
1543 					   kauth_cred_get(), uproto);
1544 			break;
1545 		}
1546 
1547 		case IPV6_TCLASS:
1548 		case IPV6_DONTFRAG:
1549 		case IPV6_USE_MIN_MTU:
1550 		case IPV6_PREFER_TEMPADDR:
1551 			error = sockopt_getint(sopt, &optval);
1552 			if (error)
1553 				break;
1554 			{
1555 				struct ip6_pktopts **optp;
1556 				optp = &in6p->in6p_outputopts;
1557 				error = ip6_pcbopt(optname,
1558 						   (u_char *)&optval,
1559 						   sizeof(optval),
1560 						   optp,
1561 						   kauth_cred_get(), uproto);
1562 				break;
1563 			}
1564 
1565 #ifdef RFC2292
1566 		case IPV6_2292PKTINFO:
1567 		case IPV6_2292HOPLIMIT:
1568 		case IPV6_2292HOPOPTS:
1569 		case IPV6_2292DSTOPTS:
1570 		case IPV6_2292RTHDR:
1571 			/* RFC 2292 */
1572 			error = sockopt_getint(sopt, &optval);
1573 			if (error)
1574 				break;
1575 
1576 			switch (optname) {
1577 			case IPV6_2292PKTINFO:
1578 				OPTSET2292(IN6P_PKTINFO);
1579 				break;
1580 			case IPV6_2292HOPLIMIT:
1581 				OPTSET2292(IN6P_HOPLIMIT);
1582 				break;
1583 			case IPV6_2292HOPOPTS:
1584 				/*
1585 				 * Check super-user privilege.
1586 				 * See comments for IPV6_RECVHOPOPTS.
1587 				 */
1588 				error =
1589 				    kauth_authorize_network(kauth_cred_get(),
1590 				    KAUTH_NETWORK_IPV6,
1591 				    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL,
1592 				    NULL, NULL);
1593 				if (error)
1594 					return (error);
1595 				OPTSET2292(IN6P_HOPOPTS);
1596 				break;
1597 			case IPV6_2292DSTOPTS:
1598 				error =
1599 				    kauth_authorize_network(kauth_cred_get(),
1600 				    KAUTH_NETWORK_IPV6,
1601 				    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL,
1602 				    NULL, NULL);
1603 				if (error)
1604 					return (error);
1605 				OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1606 				break;
1607 			case IPV6_2292RTHDR:
1608 				OPTSET2292(IN6P_RTHDR);
1609 				break;
1610 			}
1611 			break;
1612 #endif
1613 		case IPV6_PKTINFO:
1614 		case IPV6_HOPOPTS:
1615 		case IPV6_RTHDR:
1616 		case IPV6_DSTOPTS:
1617 		case IPV6_RTHDRDSTOPTS:
1618 		case IPV6_NEXTHOP: {
1619 			/* new advanced API (RFC3542) */
1620 			void *optbuf;
1621 			int optbuflen;
1622 			struct ip6_pktopts **optp;
1623 
1624 #ifdef RFC2292
1625 			/* cannot mix with RFC2292 */
1626 			if (OPTBIT(IN6P_RFC2292)) {
1627 				error = EINVAL;
1628 				break;
1629 			}
1630 #endif
1631 
1632 			optbuflen = sopt->sopt_size;
1633 			optbuf = malloc(optbuflen, M_IP6OPT, M_NOWAIT);
1634 			if (optbuf == NULL) {
1635 				error = ENOBUFS;
1636 				break;
1637 			}
1638 
1639 			error = sockopt_get(sopt, optbuf, optbuflen);
1640 			if (error) {
1641 				free(optbuf, M_IP6OPT);
1642 				break;
1643 			}
1644 			optp = &in6p->in6p_outputopts;
1645 			error = ip6_pcbopt(optname, optbuf, optbuflen,
1646 			    optp, kauth_cred_get(), uproto);
1647 
1648 			free(optbuf, M_IP6OPT);
1649 			break;
1650 			}
1651 #undef OPTSET
1652 
1653 		case IPV6_MULTICAST_IF:
1654 		case IPV6_MULTICAST_HOPS:
1655 		case IPV6_MULTICAST_LOOP:
1656 		case IPV6_JOIN_GROUP:
1657 		case IPV6_LEAVE_GROUP:
1658 			error = ip6_setmoptions(sopt, in6p);
1659 			break;
1660 
1661 		case IPV6_PORTRANGE:
1662 			error = sockopt_getint(sopt, &optval);
1663 			if (error)
1664 				break;
1665 
1666 			switch (optval) {
1667 			case IPV6_PORTRANGE_DEFAULT:
1668 				in6p->in6p_flags &= ~(IN6P_LOWPORT);
1669 				in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1670 				break;
1671 
1672 			case IPV6_PORTRANGE_HIGH:
1673 				in6p->in6p_flags &= ~(IN6P_LOWPORT);
1674 				in6p->in6p_flags |= IN6P_HIGHPORT;
1675 				break;
1676 
1677 			case IPV6_PORTRANGE_LOW:
1678 				in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1679 				in6p->in6p_flags |= IN6P_LOWPORT;
1680 				break;
1681 
1682 			default:
1683 				error = EINVAL;
1684 				break;
1685 			}
1686 			break;
1687 
1688 		case IPV6_PORTALGO:
1689 			error = sockopt_getint(sopt, &optval);
1690 			if (error)
1691 				break;
1692 
1693 			error = portalgo_algo_index_select(
1694 			    (struct inpcb_hdr *)in6p, optval);
1695 			break;
1696 
1697 #if defined(IPSEC)
1698 		case IPV6_IPSEC_POLICY:
1699 			if (ipsec_enabled) {
1700 				error = ipsec6_set_policy(in6p, optname,
1701 				    sopt->sopt_data, sopt->sopt_size,
1702 				    kauth_cred_get());
1703 				break;
1704 			}
1705 			/*FALLTHROUGH*/
1706 #endif /* IPSEC */
1707 
1708 		default:
1709 			error = ENOPROTOOPT;
1710 			break;
1711 		}
1712 		break;
1713 
1714 	case PRCO_GETOPT:
1715 		switch (optname) {
1716 #ifdef RFC2292
1717 		case IPV6_2292PKTOPTIONS:
1718 			/*
1719 			 * RFC3542 (effectively) deprecated the
1720 			 * semantics of the 2292-style pktoptions.
1721 			 * Since it was not reliable in nature (i.e.,
1722 			 * applications had to expect the lack of some
1723 			 * information after all), it would make sense
1724 			 * to simplify this part by always returning
1725 			 * empty data.
1726 			 */
1727 			break;
1728 #endif
1729 
1730 		case IPV6_RECVHOPOPTS:
1731 		case IPV6_RECVDSTOPTS:
1732 		case IPV6_RECVRTHDRDSTOPTS:
1733 		case IPV6_UNICAST_HOPS:
1734 		case IPV6_RECVPKTINFO:
1735 		case IPV6_RECVHOPLIMIT:
1736 		case IPV6_RECVRTHDR:
1737 		case IPV6_RECVPATHMTU:
1738 
1739 		case IPV6_FAITH:
1740 		case IPV6_V6ONLY:
1741 		case IPV6_PORTRANGE:
1742 		case IPV6_RECVTCLASS:
1743 			switch (optname) {
1744 
1745 			case IPV6_RECVHOPOPTS:
1746 				optval = OPTBIT(IN6P_HOPOPTS);
1747 				break;
1748 
1749 			case IPV6_RECVDSTOPTS:
1750 				optval = OPTBIT(IN6P_DSTOPTS);
1751 				break;
1752 
1753 			case IPV6_RECVRTHDRDSTOPTS:
1754 				optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1755 				break;
1756 
1757 			case IPV6_UNICAST_HOPS:
1758 				optval = in6p->in6p_hops;
1759 				break;
1760 
1761 			case IPV6_RECVPKTINFO:
1762 				optval = OPTBIT(IN6P_PKTINFO);
1763 				break;
1764 
1765 			case IPV6_RECVHOPLIMIT:
1766 				optval = OPTBIT(IN6P_HOPLIMIT);
1767 				break;
1768 
1769 			case IPV6_RECVRTHDR:
1770 				optval = OPTBIT(IN6P_RTHDR);
1771 				break;
1772 
1773 			case IPV6_RECVPATHMTU:
1774 				optval = OPTBIT(IN6P_MTU);
1775 				break;
1776 
1777 			case IPV6_FAITH:
1778 				optval = OPTBIT(IN6P_FAITH);
1779 				break;
1780 
1781 			case IPV6_V6ONLY:
1782 				optval = OPTBIT(IN6P_IPV6_V6ONLY);
1783 				break;
1784 
1785 			case IPV6_PORTRANGE:
1786 			    {
1787 				int flags;
1788 				flags = in6p->in6p_flags;
1789 				if (flags & IN6P_HIGHPORT)
1790 					optval = IPV6_PORTRANGE_HIGH;
1791 				else if (flags & IN6P_LOWPORT)
1792 					optval = IPV6_PORTRANGE_LOW;
1793 				else
1794 					optval = 0;
1795 				break;
1796 			    }
1797 			case IPV6_RECVTCLASS:
1798 				optval = OPTBIT(IN6P_TCLASS);
1799 				break;
1800 
1801 			}
1802 			if (error)
1803 				break;
1804 			error = sockopt_setint(sopt, optval);
1805 			break;
1806 
1807 		case IPV6_PATHMTU:
1808 		    {
1809 			u_long pmtu = 0;
1810 			struct ip6_mtuinfo mtuinfo;
1811 			struct route *ro = &in6p->in6p_route;
1812 
1813 			if (!(so->so_state & SS_ISCONNECTED))
1814 				return (ENOTCONN);
1815 			/*
1816 			 * XXX: we dot not consider the case of source
1817 			 * routing, or optional information to specify
1818 			 * the outgoing interface.
1819 			 */
1820 			error = ip6_getpmtu(ro, NULL, NULL,
1821 			    &in6p->in6p_faddr, &pmtu, NULL);
1822 			if (error)
1823 				break;
1824 			if (pmtu > IPV6_MAXPACKET)
1825 				pmtu = IPV6_MAXPACKET;
1826 
1827 			memset(&mtuinfo, 0, sizeof(mtuinfo));
1828 			mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1829 			optdata = (void *)&mtuinfo;
1830 			optdatalen = sizeof(mtuinfo);
1831 			if (optdatalen > MCLBYTES)
1832 				return (EMSGSIZE); /* XXX */
1833 			error = sockopt_set(sopt, optdata, optdatalen);
1834 			break;
1835 		    }
1836 
1837 #ifdef RFC2292
1838 		case IPV6_2292PKTINFO:
1839 		case IPV6_2292HOPLIMIT:
1840 		case IPV6_2292HOPOPTS:
1841 		case IPV6_2292RTHDR:
1842 		case IPV6_2292DSTOPTS:
1843 			switch (optname) {
1844 			case IPV6_2292PKTINFO:
1845 				optval = OPTBIT(IN6P_PKTINFO);
1846 				break;
1847 			case IPV6_2292HOPLIMIT:
1848 				optval = OPTBIT(IN6P_HOPLIMIT);
1849 				break;
1850 			case IPV6_2292HOPOPTS:
1851 				optval = OPTBIT(IN6P_HOPOPTS);
1852 				break;
1853 			case IPV6_2292RTHDR:
1854 				optval = OPTBIT(IN6P_RTHDR);
1855 				break;
1856 			case IPV6_2292DSTOPTS:
1857 				optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1858 				break;
1859 			}
1860 			error = sockopt_setint(sopt, optval);
1861 			break;
1862 #endif
1863 		case IPV6_PKTINFO:
1864 		case IPV6_HOPOPTS:
1865 		case IPV6_RTHDR:
1866 		case IPV6_DSTOPTS:
1867 		case IPV6_RTHDRDSTOPTS:
1868 		case IPV6_NEXTHOP:
1869 		case IPV6_OTCLASS:
1870 		case IPV6_TCLASS:
1871 		case IPV6_DONTFRAG:
1872 		case IPV6_USE_MIN_MTU:
1873 		case IPV6_PREFER_TEMPADDR:
1874 			error = ip6_getpcbopt(in6p->in6p_outputopts,
1875 			    optname, sopt);
1876 			break;
1877 
1878 		case IPV6_MULTICAST_IF:
1879 		case IPV6_MULTICAST_HOPS:
1880 		case IPV6_MULTICAST_LOOP:
1881 		case IPV6_JOIN_GROUP:
1882 		case IPV6_LEAVE_GROUP:
1883 			error = ip6_getmoptions(sopt, in6p);
1884 			break;
1885 
1886 		case IPV6_PORTALGO:
1887 			optval = ((struct inpcb_hdr *)in6p)->inph_portalgo;
1888 			error = sockopt_setint(sopt, optval);
1889 			break;
1890 
1891 #if defined(IPSEC)
1892 		case IPV6_IPSEC_POLICY:
1893 			if (ipsec_used) {
1894 				struct mbuf *m = NULL;
1895 
1896 				/*
1897 				 * XXX: this will return EINVAL as sopt is
1898 				 * empty
1899 				 */
1900 				error = ipsec6_get_policy(in6p, sopt->sopt_data,
1901 				    sopt->sopt_size, &m);
1902 				if (!error)
1903 					error = sockopt_setmbuf(sopt, m);
1904 				break;
1905 			}
1906 			/*FALLTHROUGH*/
1907 #endif /* IPSEC */
1908 
1909 		default:
1910 			error = ENOPROTOOPT;
1911 			break;
1912 		}
1913 		break;
1914 	}
1915 	return (error);
1916 }
1917 
1918 int
1919 ip6_raw_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1920 {
1921 	int error = 0, optval;
1922 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
1923 	struct in6pcb *in6p = sotoin6pcb(so);
1924 	int level, optname;
1925 
1926 	KASSERT(sopt != NULL);
1927 
1928 	level = sopt->sopt_level;
1929 	optname = sopt->sopt_name;
1930 
1931 	if (level != IPPROTO_IPV6) {
1932 		return ENOPROTOOPT;
1933 	}
1934 
1935 	switch (optname) {
1936 	case IPV6_CHECKSUM:
1937 		/*
1938 		 * For ICMPv6 sockets, no modification allowed for checksum
1939 		 * offset, permit "no change" values to help existing apps.
1940 		 *
1941 		 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM
1942 		 * for an ICMPv6 socket will fail."  The current
1943 		 * behavior does not meet RFC3542.
1944 		 */
1945 		switch (op) {
1946 		case PRCO_SETOPT:
1947 			error = sockopt_getint(sopt, &optval);
1948 			if (error)
1949 				break;
1950 			if ((optval % 2) != 0) {
1951 				/* the API assumes even offset values */
1952 				error = EINVAL;
1953 			} else if (so->so_proto->pr_protocol ==
1954 			    IPPROTO_ICMPV6) {
1955 				if (optval != icmp6off)
1956 					error = EINVAL;
1957 			} else
1958 				in6p->in6p_cksum = optval;
1959 			break;
1960 
1961 		case PRCO_GETOPT:
1962 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
1963 				optval = icmp6off;
1964 			else
1965 				optval = in6p->in6p_cksum;
1966 
1967 			error = sockopt_setint(sopt, optval);
1968 			break;
1969 
1970 		default:
1971 			error = EINVAL;
1972 			break;
1973 		}
1974 		break;
1975 
1976 	default:
1977 		error = ENOPROTOOPT;
1978 		break;
1979 	}
1980 
1981 	return (error);
1982 }
1983 
1984 #ifdef RFC2292
1985 /*
1986  * Set up IP6 options in pcb for insertion in output packets or
1987  * specifying behavior of outgoing packets.
1988  */
1989 static int
1990 ip6_pcbopts(struct ip6_pktopts **pktopt, struct socket *so,
1991     struct sockopt *sopt)
1992 {
1993 	struct ip6_pktopts *opt = *pktopt;
1994 	struct mbuf *m;
1995 	int error = 0;
1996 
1997 	/* turn off any old options. */
1998 	if (opt) {
1999 #ifdef DIAGNOSTIC
2000 	    if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2001 		opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2002 		opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2003 		    printf("ip6_pcbopts: all specified options are cleared.\n");
2004 #endif
2005 		ip6_clearpktopts(opt, -1);
2006 	} else {
2007 		opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
2008 		if (opt == NULL)
2009 			return (ENOBUFS);
2010 	}
2011 	*pktopt = NULL;
2012 
2013 	if (sopt == NULL || sopt->sopt_size == 0) {
2014 		/*
2015 		 * Only turning off any previous options, regardless of
2016 		 * whether the opt is just created or given.
2017 		 */
2018 		free(opt, M_IP6OPT);
2019 		return (0);
2020 	}
2021 
2022 	/*  set options specified by user. */
2023 	m = sockopt_getmbuf(sopt);
2024 	if (m == NULL) {
2025 		free(opt, M_IP6OPT);
2026 		return (ENOBUFS);
2027 	}
2028 
2029 	error = ip6_setpktopts(m, opt, NULL, kauth_cred_get(),
2030 	    so->so_proto->pr_protocol);
2031 	m_freem(m);
2032 	if (error != 0) {
2033 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2034 		free(opt, M_IP6OPT);
2035 		return (error);
2036 	}
2037 	*pktopt = opt;
2038 	return (0);
2039 }
2040 #endif
2041 
2042 /*
2043  * initialize ip6_pktopts.  beware that there are non-zero default values in
2044  * the struct.
2045  */
2046 void
2047 ip6_initpktopts(struct ip6_pktopts *opt)
2048 {
2049 
2050 	memset(opt, 0, sizeof(*opt));
2051 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2052 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2053 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2054 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2055 }
2056 
2057 #define sin6tosa(sin6)	((struct sockaddr *)(sin6)) /* XXX */
2058 static int
2059 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2060     kauth_cred_t cred, int uproto)
2061 {
2062 	struct ip6_pktopts *opt;
2063 
2064 	if (*pktopt == NULL) {
2065 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2066 		    M_NOWAIT);
2067 		if (*pktopt == NULL)
2068 			return (ENOBUFS);
2069 
2070 		ip6_initpktopts(*pktopt);
2071 	}
2072 	opt = *pktopt;
2073 
2074 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2075 }
2076 
2077 static int
2078 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2079 {
2080 	void *optdata = NULL;
2081 	int optdatalen = 0;
2082 	struct ip6_ext *ip6e;
2083 	int error = 0;
2084 	struct in6_pktinfo null_pktinfo;
2085 	int deftclass = 0, on;
2086 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2087 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2088 
2089 	switch (optname) {
2090 	case IPV6_PKTINFO:
2091 		if (pktopt && pktopt->ip6po_pktinfo)
2092 			optdata = (void *)pktopt->ip6po_pktinfo;
2093 		else {
2094 			/* XXX: we don't have to do this every time... */
2095 			memset(&null_pktinfo, 0, sizeof(null_pktinfo));
2096 			optdata = (void *)&null_pktinfo;
2097 		}
2098 		optdatalen = sizeof(struct in6_pktinfo);
2099 		break;
2100 	case IPV6_OTCLASS:
2101 		/* XXX */
2102 		return (EINVAL);
2103 	case IPV6_TCLASS:
2104 		if (pktopt && pktopt->ip6po_tclass >= 0)
2105 			optdata = (void *)&pktopt->ip6po_tclass;
2106 		else
2107 			optdata = (void *)&deftclass;
2108 		optdatalen = sizeof(int);
2109 		break;
2110 	case IPV6_HOPOPTS:
2111 		if (pktopt && pktopt->ip6po_hbh) {
2112 			optdata = (void *)pktopt->ip6po_hbh;
2113 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2114 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2115 		}
2116 		break;
2117 	case IPV6_RTHDR:
2118 		if (pktopt && pktopt->ip6po_rthdr) {
2119 			optdata = (void *)pktopt->ip6po_rthdr;
2120 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2121 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2122 		}
2123 		break;
2124 	case IPV6_RTHDRDSTOPTS:
2125 		if (pktopt && pktopt->ip6po_dest1) {
2126 			optdata = (void *)pktopt->ip6po_dest1;
2127 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2128 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2129 		}
2130 		break;
2131 	case IPV6_DSTOPTS:
2132 		if (pktopt && pktopt->ip6po_dest2) {
2133 			optdata = (void *)pktopt->ip6po_dest2;
2134 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2135 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2136 		}
2137 		break;
2138 	case IPV6_NEXTHOP:
2139 		if (pktopt && pktopt->ip6po_nexthop) {
2140 			optdata = (void *)pktopt->ip6po_nexthop;
2141 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2142 		}
2143 		break;
2144 	case IPV6_USE_MIN_MTU:
2145 		if (pktopt)
2146 			optdata = (void *)&pktopt->ip6po_minmtu;
2147 		else
2148 			optdata = (void *)&defminmtu;
2149 		optdatalen = sizeof(int);
2150 		break;
2151 	case IPV6_DONTFRAG:
2152 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2153 			on = 1;
2154 		else
2155 			on = 0;
2156 		optdata = (void *)&on;
2157 		optdatalen = sizeof(on);
2158 		break;
2159 	case IPV6_PREFER_TEMPADDR:
2160 		if (pktopt)
2161 			optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2162 		else
2163 			optdata = (void *)&defpreftemp;
2164 		optdatalen = sizeof(int);
2165 		break;
2166 	default:		/* should not happen */
2167 #ifdef DIAGNOSTIC
2168 		panic("ip6_getpcbopt: unexpected option\n");
2169 #endif
2170 		return (ENOPROTOOPT);
2171 	}
2172 
2173 	error = sockopt_set(sopt, optdata, optdatalen);
2174 
2175 	return (error);
2176 }
2177 
2178 void
2179 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2180 {
2181 	if (optname == -1 || optname == IPV6_PKTINFO) {
2182 		if (pktopt->ip6po_pktinfo)
2183 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2184 		pktopt->ip6po_pktinfo = NULL;
2185 	}
2186 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2187 		pktopt->ip6po_hlim = -1;
2188 	if (optname == -1 || optname == IPV6_TCLASS)
2189 		pktopt->ip6po_tclass = -1;
2190 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2191 		rtcache_free(&pktopt->ip6po_nextroute);
2192 		if (pktopt->ip6po_nexthop)
2193 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2194 		pktopt->ip6po_nexthop = NULL;
2195 	}
2196 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2197 		if (pktopt->ip6po_hbh)
2198 			free(pktopt->ip6po_hbh, M_IP6OPT);
2199 		pktopt->ip6po_hbh = NULL;
2200 	}
2201 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2202 		if (pktopt->ip6po_dest1)
2203 			free(pktopt->ip6po_dest1, M_IP6OPT);
2204 		pktopt->ip6po_dest1 = NULL;
2205 	}
2206 	if (optname == -1 || optname == IPV6_RTHDR) {
2207 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2208 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2209 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2210 		rtcache_free(&pktopt->ip6po_route);
2211 	}
2212 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2213 		if (pktopt->ip6po_dest2)
2214 			free(pktopt->ip6po_dest2, M_IP6OPT);
2215 		pktopt->ip6po_dest2 = NULL;
2216 	}
2217 }
2218 
2219 #define PKTOPT_EXTHDRCPY(type) 					\
2220 do {								\
2221 	if (src->type) {					\
2222 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2223 		dst->type = malloc(hlen, M_IP6OPT, canwait);	\
2224 		if (dst->type == NULL)				\
2225 			goto bad;				\
2226 		memcpy(dst->type, src->type, hlen);		\
2227 	}							\
2228 } while (/*CONSTCOND*/ 0)
2229 
2230 static int
2231 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2232 {
2233 	dst->ip6po_hlim = src->ip6po_hlim;
2234 	dst->ip6po_tclass = src->ip6po_tclass;
2235 	dst->ip6po_flags = src->ip6po_flags;
2236 	dst->ip6po_minmtu = src->ip6po_minmtu;
2237 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2238 	if (src->ip6po_pktinfo) {
2239 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2240 		    M_IP6OPT, canwait);
2241 		if (dst->ip6po_pktinfo == NULL)
2242 			goto bad;
2243 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2244 	}
2245 	if (src->ip6po_nexthop) {
2246 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2247 		    M_IP6OPT, canwait);
2248 		if (dst->ip6po_nexthop == NULL)
2249 			goto bad;
2250 		memcpy(dst->ip6po_nexthop, src->ip6po_nexthop,
2251 		    src->ip6po_nexthop->sa_len);
2252 	}
2253 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2254 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2255 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2256 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2257 	return (0);
2258 
2259   bad:
2260 	if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
2261 	if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
2262 	if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
2263 	if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
2264 	if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
2265 	if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
2266 
2267 	return (ENOBUFS);
2268 }
2269 #undef PKTOPT_EXTHDRCPY
2270 
2271 struct ip6_pktopts *
2272 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2273 {
2274 	int error;
2275 	struct ip6_pktopts *dst;
2276 
2277 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2278 	if (dst == NULL)
2279 		return (NULL);
2280 	ip6_initpktopts(dst);
2281 
2282 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2283 		free(dst, M_IP6OPT);
2284 		return (NULL);
2285 	}
2286 
2287 	return (dst);
2288 }
2289 
2290 void
2291 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2292 {
2293 	if (pktopt == NULL)
2294 		return;
2295 
2296 	ip6_clearpktopts(pktopt, -1);
2297 
2298 	free(pktopt, M_IP6OPT);
2299 }
2300 
2301 int
2302 ip6_get_membership(const struct sockopt *sopt, struct ifnet **ifp, void *v,
2303     size_t l)
2304 {
2305 	struct ipv6_mreq mreq;
2306 	int error;
2307 	struct in6_addr *ia = &mreq.ipv6mr_multiaddr;
2308 	struct in_addr *ia4 = (void *)&ia->s6_addr32[3];
2309 	error = sockopt_get(sopt, &mreq, sizeof(mreq));
2310 	if (error != 0)
2311 		return error;
2312 
2313 	if (IN6_IS_ADDR_UNSPECIFIED(ia)) {
2314 		/*
2315 		 * We use the unspecified address to specify to accept
2316 		 * all multicast addresses. Only super user is allowed
2317 		 * to do this.
2318 		 */
2319 		if (kauth_authorize_network(curlwp->l_cred, KAUTH_NETWORK_IPV6,
2320 		    KAUTH_REQ_NETWORK_IPV6_JOIN_MULTICAST, NULL, NULL, NULL))
2321 			return EACCES;
2322 	} else if (IN6_IS_ADDR_V4MAPPED(ia)) {
2323 		// Don't bother if we are not going to use ifp.
2324 		if (l == sizeof(*ia)) {
2325 			memcpy(v, ia, l);
2326 			return 0;
2327 		}
2328 	} else if (!IN6_IS_ADDR_MULTICAST(ia)) {
2329 		return EINVAL;
2330 	}
2331 
2332 	/*
2333 	 * If no interface was explicitly specified, choose an
2334 	 * appropriate one according to the given multicast address.
2335 	 */
2336 	if (mreq.ipv6mr_interface == 0) {
2337 		struct rtentry *rt;
2338 		union {
2339 			struct sockaddr		dst;
2340 			struct sockaddr_in	dst4;
2341 			struct sockaddr_in6	dst6;
2342 		} u;
2343 		struct route ro;
2344 
2345 		/*
2346 		 * Look up the routing table for the
2347 		 * address, and choose the outgoing interface.
2348 		 *   XXX: is it a good approach?
2349 		 */
2350 		memset(&ro, 0, sizeof(ro));
2351 		if (IN6_IS_ADDR_V4MAPPED(ia))
2352 			sockaddr_in_init(&u.dst4, ia4, 0);
2353 		else
2354 			sockaddr_in6_init(&u.dst6, ia, 0, 0, 0);
2355 		error = rtcache_setdst(&ro, &u.dst);
2356 		if (error != 0)
2357 			return error;
2358 		*ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp : NULL;
2359 		rtcache_free(&ro);
2360 	} else {
2361 		/*
2362 		 * If the interface is specified, validate it.
2363 		 */
2364 		if ((*ifp = if_byindex(mreq.ipv6mr_interface)) == NULL)
2365 			return ENXIO;	/* XXX EINVAL? */
2366 	}
2367 	if (sizeof(*ia) == l)
2368 		memcpy(v, ia, l);
2369 	else
2370 		memcpy(v, ia4, l);
2371 	return 0;
2372 }
2373 
2374 /*
2375  * Set the IP6 multicast options in response to user setsockopt().
2376  */
2377 static int
2378 ip6_setmoptions(const struct sockopt *sopt, struct in6pcb *in6p)
2379 {
2380 	int error = 0;
2381 	u_int loop, ifindex;
2382 	struct ipv6_mreq mreq;
2383 	struct in6_addr ia;
2384 	struct ifnet *ifp;
2385 	struct ip6_moptions *im6o = in6p->in6p_moptions;
2386 	struct in6_multi_mship *imm;
2387 
2388 	if (im6o == NULL) {
2389 		/*
2390 		 * No multicast option buffer attached to the pcb;
2391 		 * allocate one and initialize to default values.
2392 		 */
2393 		im6o = malloc(sizeof(*im6o), M_IPMOPTS, M_NOWAIT);
2394 		if (im6o == NULL)
2395 			return (ENOBUFS);
2396 		in6p->in6p_moptions = im6o;
2397 		im6o->im6o_multicast_if_index = 0;
2398 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2399 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2400 		LIST_INIT(&im6o->im6o_memberships);
2401 	}
2402 
2403 	switch (sopt->sopt_name) {
2404 
2405 	case IPV6_MULTICAST_IF:
2406 		/*
2407 		 * Select the interface for outgoing multicast packets.
2408 		 */
2409 		error = sockopt_get(sopt, &ifindex, sizeof(ifindex));
2410 		if (error != 0)
2411 			break;
2412 
2413 		if (ifindex != 0) {
2414 			if ((ifp = if_byindex(ifindex)) == NULL) {
2415 				error = ENXIO;	/* XXX EINVAL? */
2416 				break;
2417 			}
2418 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
2419 				error = EADDRNOTAVAIL;
2420 				break;
2421 			}
2422 		} else
2423 			ifp = NULL;
2424 		im6o->im6o_multicast_if_index = if_get_index(ifp);
2425 		break;
2426 
2427 	case IPV6_MULTICAST_HOPS:
2428 	    {
2429 		/*
2430 		 * Set the IP6 hoplimit for outgoing multicast packets.
2431 		 */
2432 		int optval;
2433 
2434 		error = sockopt_getint(sopt, &optval);
2435 		if (error != 0)
2436 			break;
2437 
2438 		if (optval < -1 || optval >= 256)
2439 			error = EINVAL;
2440 		else if (optval == -1)
2441 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2442 		else
2443 			im6o->im6o_multicast_hlim = optval;
2444 		break;
2445 	    }
2446 
2447 	case IPV6_MULTICAST_LOOP:
2448 		/*
2449 		 * Set the loopback flag for outgoing multicast packets.
2450 		 * Must be zero or one.
2451 		 */
2452 		error = sockopt_get(sopt, &loop, sizeof(loop));
2453 		if (error != 0)
2454 			break;
2455 		if (loop > 1) {
2456 			error = EINVAL;
2457 			break;
2458 		}
2459 		im6o->im6o_multicast_loop = loop;
2460 		break;
2461 
2462 	case IPV6_JOIN_GROUP:
2463 		/*
2464 		 * Add a multicast group membership.
2465 		 * Group must be a valid IP6 multicast address.
2466 		 */
2467 		if ((error = ip6_get_membership(sopt, &ifp, &ia, sizeof(ia))))
2468 			return error;
2469 
2470 		if (IN6_IS_ADDR_V4MAPPED(&ia)) {
2471 			error = ip_setmoptions(&in6p->in6p_v4moptions, sopt);
2472 			break;
2473 		}
2474 		/*
2475 		 * See if we found an interface, and confirm that it
2476 		 * supports multicast
2477 		 */
2478 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2479 			error = EADDRNOTAVAIL;
2480 			break;
2481 		}
2482 
2483 		if (in6_setscope(&ia, ifp, NULL)) {
2484 			error = EADDRNOTAVAIL; /* XXX: should not happen */
2485 			break;
2486 		}
2487 
2488 		/*
2489 		 * See if the membership already exists.
2490 		 */
2491 		for (imm = im6o->im6o_memberships.lh_first;
2492 		     imm != NULL; imm = imm->i6mm_chain.le_next)
2493 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
2494 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2495 			    &ia))
2496 				break;
2497 		if (imm != NULL) {
2498 			error = EADDRINUSE;
2499 			break;
2500 		}
2501 		/*
2502 		 * Everything looks good; add a new record to the multicast
2503 		 * address list for the given interface.
2504 		 */
2505 		imm = in6_joingroup(ifp, &ia, &error, 0);
2506 		if (imm == NULL)
2507 			break;
2508 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2509 		break;
2510 
2511 	case IPV6_LEAVE_GROUP:
2512 		/*
2513 		 * Drop a multicast group membership.
2514 		 * Group must be a valid IP6 multicast address.
2515 		 */
2516 		error = sockopt_get(sopt, &mreq, sizeof(mreq));
2517 		if (error != 0)
2518 			break;
2519 
2520 		if (IN6_IS_ADDR_V4MAPPED(&mreq.ipv6mr_multiaddr)) {
2521 			error = ip_setmoptions(&in6p->in6p_v4moptions, sopt);
2522 			break;
2523 		}
2524 		/*
2525 		 * If an interface address was specified, get a pointer
2526 		 * to its ifnet structure.
2527 		 */
2528 		if (mreq.ipv6mr_interface != 0) {
2529 			if ((ifp = if_byindex(mreq.ipv6mr_interface)) == NULL) {
2530 				error = ENXIO;	/* XXX EINVAL? */
2531 				break;
2532 			}
2533 		} else
2534 			ifp = NULL;
2535 
2536 		/* Fill in the scope zone ID */
2537 		if (ifp) {
2538 			if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) {
2539 				/* XXX: should not happen */
2540 				error = EADDRNOTAVAIL;
2541 				break;
2542 			}
2543 		} else if (mreq.ipv6mr_interface != 0) {
2544 			/*
2545 			 * XXX: This case would happens when the (positive)
2546 			 * index is in the valid range, but the corresponding
2547 			 * interface has been detached dynamically.  The above
2548 			 * check probably avoids such case to happen here, but
2549 			 * we check it explicitly for safety.
2550 			 */
2551 			error = EADDRNOTAVAIL;
2552 			break;
2553 		} else {	/* ipv6mr_interface == 0 */
2554 			struct sockaddr_in6 sa6_mc;
2555 
2556 			/*
2557 			 * The API spec says as follows:
2558 			 *  If the interface index is specified as 0, the
2559 			 *  system may choose a multicast group membership to
2560 			 *  drop by matching the multicast address only.
2561 			 * On the other hand, we cannot disambiguate the scope
2562 			 * zone unless an interface is provided.  Thus, we
2563 			 * check if there's ambiguity with the default scope
2564 			 * zone as the last resort.
2565 			 */
2566 			sockaddr_in6_init(&sa6_mc, &mreq.ipv6mr_multiaddr,
2567 			    0, 0, 0);
2568 			error = sa6_embedscope(&sa6_mc, ip6_use_defzone);
2569 			if (error != 0)
2570 				break;
2571 			mreq.ipv6mr_multiaddr = sa6_mc.sin6_addr;
2572 		}
2573 
2574 		/*
2575 		 * Find the membership in the membership list.
2576 		 */
2577 		for (imm = im6o->im6o_memberships.lh_first;
2578 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
2579 			if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2580 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2581 			    &mreq.ipv6mr_multiaddr))
2582 				break;
2583 		}
2584 		if (imm == NULL) {
2585 			/* Unable to resolve interface */
2586 			error = EADDRNOTAVAIL;
2587 			break;
2588 		}
2589 		/*
2590 		 * Give up the multicast address record to which the
2591 		 * membership points.
2592 		 */
2593 		LIST_REMOVE(imm, i6mm_chain);
2594 		in6_leavegroup(imm);
2595 		break;
2596 
2597 	default:
2598 		error = EOPNOTSUPP;
2599 		break;
2600 	}
2601 
2602 	/*
2603 	 * If all options have default values, no need to keep the mbuf.
2604 	 */
2605 	if (im6o->im6o_multicast_if_index == 0 &&
2606 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2607 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2608 	    im6o->im6o_memberships.lh_first == NULL) {
2609 		free(in6p->in6p_moptions, M_IPMOPTS);
2610 		in6p->in6p_moptions = NULL;
2611 	}
2612 
2613 	return (error);
2614 }
2615 
2616 /*
2617  * Return the IP6 multicast options in response to user getsockopt().
2618  */
2619 static int
2620 ip6_getmoptions(struct sockopt *sopt, struct in6pcb *in6p)
2621 {
2622 	u_int optval;
2623 	int error;
2624 	struct ip6_moptions *im6o = in6p->in6p_moptions;
2625 
2626 	switch (sopt->sopt_name) {
2627 	case IPV6_MULTICAST_IF:
2628 		if (im6o == NULL || im6o->im6o_multicast_if_index == 0)
2629 			optval = 0;
2630 		else
2631 			optval = im6o->im6o_multicast_if_index;
2632 
2633 		error = sockopt_set(sopt, &optval, sizeof(optval));
2634 		break;
2635 
2636 	case IPV6_MULTICAST_HOPS:
2637 		if (im6o == NULL)
2638 			optval = ip6_defmcasthlim;
2639 		else
2640 			optval = im6o->im6o_multicast_hlim;
2641 
2642 		error = sockopt_set(sopt, &optval, sizeof(optval));
2643 		break;
2644 
2645 	case IPV6_MULTICAST_LOOP:
2646 		if (im6o == NULL)
2647 			optval = IPV6_DEFAULT_MULTICAST_LOOP;
2648 		else
2649 			optval = im6o->im6o_multicast_loop;
2650 
2651 		error = sockopt_set(sopt, &optval, sizeof(optval));
2652 		break;
2653 
2654 	default:
2655 		error = EOPNOTSUPP;
2656 	}
2657 
2658 	return (error);
2659 }
2660 
2661 /*
2662  * Discard the IP6 multicast options.
2663  */
2664 void
2665 ip6_freemoptions(struct ip6_moptions *im6o)
2666 {
2667 	struct in6_multi_mship *imm;
2668 
2669 	if (im6o == NULL)
2670 		return;
2671 
2672 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2673 		LIST_REMOVE(imm, i6mm_chain);
2674 		in6_leavegroup(imm);
2675 	}
2676 	free(im6o, M_IPMOPTS);
2677 }
2678 
2679 /*
2680  * Set IPv6 outgoing packet options based on advanced API.
2681  */
2682 int
2683 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2684 	struct ip6_pktopts *stickyopt, kauth_cred_t cred, int uproto)
2685 {
2686 	struct cmsghdr *cm = 0;
2687 
2688 	if (control == NULL || opt == NULL)
2689 		return (EINVAL);
2690 
2691 	ip6_initpktopts(opt);
2692 	if (stickyopt) {
2693 		int error;
2694 
2695 		/*
2696 		 * If stickyopt is provided, make a local copy of the options
2697 		 * for this particular packet, then override them by ancillary
2698 		 * objects.
2699 		 * XXX: copypktopts() does not copy the cached route to a next
2700 		 * hop (if any).  This is not very good in terms of efficiency,
2701 		 * but we can allow this since this option should be rarely
2702 		 * used.
2703 		 */
2704 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2705 			return (error);
2706 	}
2707 
2708 	/*
2709 	 * XXX: Currently, we assume all the optional information is stored
2710 	 * in a single mbuf.
2711 	 */
2712 	if (control->m_next)
2713 		return (EINVAL);
2714 
2715 	/* XXX if cm->cmsg_len is not aligned, control->m_len can become <0 */
2716 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2717 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2718 		int error;
2719 
2720 		if (control->m_len < CMSG_LEN(0))
2721 			return (EINVAL);
2722 
2723 		cm = mtod(control, struct cmsghdr *);
2724 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2725 			return (EINVAL);
2726 		if (cm->cmsg_level != IPPROTO_IPV6)
2727 			continue;
2728 
2729 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2730 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2731 		if (error)
2732 			return (error);
2733 	}
2734 
2735 	return (0);
2736 }
2737 
2738 /*
2739  * Set a particular packet option, as a sticky option or an ancillary data
2740  * item.  "len" can be 0 only when it's a sticky option.
2741  * We have 4 cases of combination of "sticky" and "cmsg":
2742  * "sticky=0, cmsg=0": impossible
2743  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2744  * "sticky=1, cmsg=0": RFC3542 socket option
2745  * "sticky=1, cmsg=1": RFC2292 socket option
2746  */
2747 static int
2748 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2749     kauth_cred_t cred, int sticky, int cmsg, int uproto)
2750 {
2751 	int minmtupolicy;
2752 	int error;
2753 
2754 	if (!sticky && !cmsg) {
2755 #ifdef DIAGNOSTIC
2756 		printf("ip6_setpktopt: impossible case\n");
2757 #endif
2758 		return (EINVAL);
2759 	}
2760 
2761 	/*
2762 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2763 	 * not be specified in the context of RFC3542.  Conversely,
2764 	 * RFC3542 types should not be specified in the context of RFC2292.
2765 	 */
2766 	if (!cmsg) {
2767 		switch (optname) {
2768 		case IPV6_2292PKTINFO:
2769 		case IPV6_2292HOPLIMIT:
2770 		case IPV6_2292NEXTHOP:
2771 		case IPV6_2292HOPOPTS:
2772 		case IPV6_2292DSTOPTS:
2773 		case IPV6_2292RTHDR:
2774 		case IPV6_2292PKTOPTIONS:
2775 			return (ENOPROTOOPT);
2776 		}
2777 	}
2778 	if (sticky && cmsg) {
2779 		switch (optname) {
2780 		case IPV6_PKTINFO:
2781 		case IPV6_HOPLIMIT:
2782 		case IPV6_NEXTHOP:
2783 		case IPV6_HOPOPTS:
2784 		case IPV6_DSTOPTS:
2785 		case IPV6_RTHDRDSTOPTS:
2786 		case IPV6_RTHDR:
2787 		case IPV6_USE_MIN_MTU:
2788 		case IPV6_DONTFRAG:
2789 		case IPV6_OTCLASS:
2790 		case IPV6_TCLASS:
2791 		case IPV6_PREFER_TEMPADDR: /* XXX not an RFC3542 option */
2792 			return (ENOPROTOOPT);
2793 		}
2794 	}
2795 
2796 	switch (optname) {
2797 #ifdef RFC2292
2798 	case IPV6_2292PKTINFO:
2799 #endif
2800 	case IPV6_PKTINFO:
2801 	{
2802 		struct in6_pktinfo *pktinfo;
2803 
2804 		if (len != sizeof(struct in6_pktinfo))
2805 			return (EINVAL);
2806 
2807 		pktinfo = (struct in6_pktinfo *)buf;
2808 
2809 		/*
2810 		 * An application can clear any sticky IPV6_PKTINFO option by
2811 		 * doing a "regular" setsockopt with ipi6_addr being
2812 		 * in6addr_any and ipi6_ifindex being zero.
2813 		 * [RFC 3542, Section 6]
2814 		 */
2815 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2816 		    pktinfo->ipi6_ifindex == 0 &&
2817 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2818 			ip6_clearpktopts(opt, optname);
2819 			break;
2820 		}
2821 
2822 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2823 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2824 			return (EINVAL);
2825 		}
2826 
2827 		/* Validate the interface index if specified. */
2828 		if (pktinfo->ipi6_ifindex) {
2829 			struct ifnet *ifp;
2830 			int s = pserialize_read_enter();
2831 			ifp = if_byindex(pktinfo->ipi6_ifindex);
2832 			if (ifp == NULL) {
2833 				pserialize_read_exit(s);
2834 				return ENXIO;
2835 			}
2836 			pserialize_read_exit(s);
2837 		}
2838 
2839 		/*
2840 		 * We store the address anyway, and let in6_selectsrc()
2841 		 * validate the specified address.  This is because ipi6_addr
2842 		 * may not have enough information about its scope zone, and
2843 		 * we may need additional information (such as outgoing
2844 		 * interface or the scope zone of a destination address) to
2845 		 * disambiguate the scope.
2846 		 * XXX: the delay of the validation may confuse the
2847 		 * application when it is used as a sticky option.
2848 		 */
2849 		if (opt->ip6po_pktinfo == NULL) {
2850 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2851 			    M_IP6OPT, M_NOWAIT);
2852 			if (opt->ip6po_pktinfo == NULL)
2853 				return (ENOBUFS);
2854 		}
2855 		memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo));
2856 		break;
2857 	}
2858 
2859 #ifdef RFC2292
2860 	case IPV6_2292HOPLIMIT:
2861 #endif
2862 	case IPV6_HOPLIMIT:
2863 	{
2864 		int *hlimp;
2865 
2866 		/*
2867 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2868 		 * to simplify the ordering among hoplimit options.
2869 		 */
2870 		if (optname == IPV6_HOPLIMIT && sticky)
2871 			return (ENOPROTOOPT);
2872 
2873 		if (len != sizeof(int))
2874 			return (EINVAL);
2875 		hlimp = (int *)buf;
2876 		if (*hlimp < -1 || *hlimp > 255)
2877 			return (EINVAL);
2878 
2879 		opt->ip6po_hlim = *hlimp;
2880 		break;
2881 	}
2882 
2883 	case IPV6_OTCLASS:
2884 		if (len != sizeof(u_int8_t))
2885 			return (EINVAL);
2886 
2887 		opt->ip6po_tclass = *(u_int8_t *)buf;
2888 		break;
2889 
2890 	case IPV6_TCLASS:
2891 	{
2892 		int tclass;
2893 
2894 		if (len != sizeof(int))
2895 			return (EINVAL);
2896 		tclass = *(int *)buf;
2897 		if (tclass < -1 || tclass > 255)
2898 			return (EINVAL);
2899 
2900 		opt->ip6po_tclass = tclass;
2901 		break;
2902 	}
2903 
2904 #ifdef RFC2292
2905 	case IPV6_2292NEXTHOP:
2906 #endif
2907 	case IPV6_NEXTHOP:
2908 		error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
2909 		    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2910 		if (error)
2911 			return (error);
2912 
2913 		if (len == 0) {	/* just remove the option */
2914 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2915 			break;
2916 		}
2917 
2918 		/* check if cmsg_len is large enough for sa_len */
2919 		if (len < sizeof(struct sockaddr) || len < *buf)
2920 			return (EINVAL);
2921 
2922 		switch (((struct sockaddr *)buf)->sa_family) {
2923 		case AF_INET6:
2924 		{
2925 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2926 
2927 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2928 				return (EINVAL);
2929 
2930 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2931 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2932 				return (EINVAL);
2933 			}
2934 			if ((error = sa6_embedscope(sa6, ip6_use_defzone))
2935 			    != 0) {
2936 				return (error);
2937 			}
2938 			break;
2939 		}
2940 		case AF_LINK:	/* eventually be supported? */
2941 		default:
2942 			return (EAFNOSUPPORT);
2943 		}
2944 
2945 		/* turn off the previous option, then set the new option. */
2946 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2947 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2948 		if (opt->ip6po_nexthop == NULL)
2949 			return (ENOBUFS);
2950 		memcpy(opt->ip6po_nexthop, buf, *buf);
2951 		break;
2952 
2953 #ifdef RFC2292
2954 	case IPV6_2292HOPOPTS:
2955 #endif
2956 	case IPV6_HOPOPTS:
2957 	{
2958 		struct ip6_hbh *hbh;
2959 		int hbhlen;
2960 
2961 		/*
2962 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2963 		 * options, since per-option restriction has too much
2964 		 * overhead.
2965 		 */
2966 		error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
2967 		    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
2968 		if (error)
2969 			return (error);
2970 
2971 		if (len == 0) {
2972 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2973 			break;	/* just remove the option */
2974 		}
2975 
2976 		/* message length validation */
2977 		if (len < sizeof(struct ip6_hbh))
2978 			return (EINVAL);
2979 		hbh = (struct ip6_hbh *)buf;
2980 		hbhlen = (hbh->ip6h_len + 1) << 3;
2981 		if (len != hbhlen)
2982 			return (EINVAL);
2983 
2984 		/* turn off the previous option, then set the new option. */
2985 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2986 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2987 		if (opt->ip6po_hbh == NULL)
2988 			return (ENOBUFS);
2989 		memcpy(opt->ip6po_hbh, hbh, hbhlen);
2990 
2991 		break;
2992 	}
2993 
2994 #ifdef RFC2292
2995 	case IPV6_2292DSTOPTS:
2996 #endif
2997 	case IPV6_DSTOPTS:
2998 	case IPV6_RTHDRDSTOPTS:
2999 	{
3000 		struct ip6_dest *dest, **newdest = NULL;
3001 		int destlen;
3002 
3003 		/* XXX: see the comment for IPV6_HOPOPTS */
3004 		error = kauth_authorize_network(cred, KAUTH_NETWORK_IPV6,
3005 		    KAUTH_REQ_NETWORK_IPV6_HOPBYHOP, NULL, NULL, NULL);
3006 		if (error)
3007 			return (error);
3008 
3009 		if (len == 0) {
3010 			ip6_clearpktopts(opt, optname);
3011 			break;	/* just remove the option */
3012 		}
3013 
3014 		/* message length validation */
3015 		if (len < sizeof(struct ip6_dest))
3016 			return (EINVAL);
3017 		dest = (struct ip6_dest *)buf;
3018 		destlen = (dest->ip6d_len + 1) << 3;
3019 		if (len != destlen)
3020 			return (EINVAL);
3021 		/*
3022 		 * Determine the position that the destination options header
3023 		 * should be inserted; before or after the routing header.
3024 		 */
3025 		switch (optname) {
3026 		case IPV6_2292DSTOPTS:
3027 			/*
3028 			 * The old advanced API is ambiguous on this point.
3029 			 * Our approach is to determine the position based
3030 			 * according to the existence of a routing header.
3031 			 * Note, however, that this depends on the order of the
3032 			 * extension headers in the ancillary data; the 1st
3033 			 * part of the destination options header must appear
3034 			 * before the routing header in the ancillary data,
3035 			 * too.
3036 			 * RFC3542 solved the ambiguity by introducing
3037 			 * separate ancillary data or option types.
3038 			 */
3039 			if (opt->ip6po_rthdr == NULL)
3040 				newdest = &opt->ip6po_dest1;
3041 			else
3042 				newdest = &opt->ip6po_dest2;
3043 			break;
3044 		case IPV6_RTHDRDSTOPTS:
3045 			newdest = &opt->ip6po_dest1;
3046 			break;
3047 		case IPV6_DSTOPTS:
3048 			newdest = &opt->ip6po_dest2;
3049 			break;
3050 		}
3051 
3052 		/* turn off the previous option, then set the new option. */
3053 		ip6_clearpktopts(opt, optname);
3054 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3055 		if (*newdest == NULL)
3056 			return (ENOBUFS);
3057 		memcpy(*newdest, dest, destlen);
3058 
3059 		break;
3060 	}
3061 
3062 #ifdef RFC2292
3063 	case IPV6_2292RTHDR:
3064 #endif
3065 	case IPV6_RTHDR:
3066 	{
3067 		struct ip6_rthdr *rth;
3068 		int rthlen;
3069 
3070 		if (len == 0) {
3071 			ip6_clearpktopts(opt, IPV6_RTHDR);
3072 			break;	/* just remove the option */
3073 		}
3074 
3075 		/* message length validation */
3076 		if (len < sizeof(struct ip6_rthdr))
3077 			return (EINVAL);
3078 		rth = (struct ip6_rthdr *)buf;
3079 		rthlen = (rth->ip6r_len + 1) << 3;
3080 		if (len != rthlen)
3081 			return (EINVAL);
3082 		switch (rth->ip6r_type) {
3083 		case IPV6_RTHDR_TYPE_0:
3084 			if (rth->ip6r_len == 0)	/* must contain one addr */
3085 				return (EINVAL);
3086 			if (rth->ip6r_len % 2) /* length must be even */
3087 				return (EINVAL);
3088 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3089 				return (EINVAL);
3090 			break;
3091 		default:
3092 			return (EINVAL);	/* not supported */
3093 		}
3094 		/* turn off the previous option */
3095 		ip6_clearpktopts(opt, IPV6_RTHDR);
3096 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3097 		if (opt->ip6po_rthdr == NULL)
3098 			return (ENOBUFS);
3099 		memcpy(opt->ip6po_rthdr, rth, rthlen);
3100 		break;
3101 	}
3102 
3103 	case IPV6_USE_MIN_MTU:
3104 		if (len != sizeof(int))
3105 			return (EINVAL);
3106 		minmtupolicy = *(int *)buf;
3107 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3108 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3109 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3110 			return (EINVAL);
3111 		}
3112 		opt->ip6po_minmtu = minmtupolicy;
3113 		break;
3114 
3115 	case IPV6_DONTFRAG:
3116 		if (len != sizeof(int))
3117 			return (EINVAL);
3118 
3119 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3120 			/*
3121 			 * we ignore this option for TCP sockets.
3122 			 * (RFC3542 leaves this case unspecified.)
3123 			 */
3124 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3125 		} else
3126 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3127 		break;
3128 
3129 	case IPV6_PREFER_TEMPADDR:
3130 	{
3131 		int preftemp;
3132 
3133 		if (len != sizeof(int))
3134 			return (EINVAL);
3135 		preftemp = *(int *)buf;
3136 		switch (preftemp) {
3137 		case IP6PO_TEMPADDR_SYSTEM:
3138 		case IP6PO_TEMPADDR_NOTPREFER:
3139 		case IP6PO_TEMPADDR_PREFER:
3140 			break;
3141 		default:
3142 			return (EINVAL);
3143 		}
3144 		opt->ip6po_prefer_tempaddr = preftemp;
3145 		break;
3146 	}
3147 
3148 	default:
3149 		return (ENOPROTOOPT);
3150 	} /* end of switch */
3151 
3152 	return (0);
3153 }
3154 
3155 /*
3156  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3157  * packet to the input queue of a specified interface.  Note that this
3158  * calls the output routine of the loopback "driver", but with an interface
3159  * pointer that might NOT be lo0ifp -- easier than replicating that code here.
3160  */
3161 void
3162 ip6_mloopback(struct ifnet *ifp, struct mbuf *m,
3163 	const struct sockaddr_in6 *dst)
3164 {
3165 	struct mbuf *copym;
3166 	struct ip6_hdr *ip6;
3167 
3168 	copym = m_copy(m, 0, M_COPYALL);
3169 	if (copym == NULL)
3170 		return;
3171 
3172 	/*
3173 	 * Make sure to deep-copy IPv6 header portion in case the data
3174 	 * is in an mbuf cluster, so that we can safely override the IPv6
3175 	 * header portion later.
3176 	 */
3177 	if ((copym->m_flags & M_EXT) != 0 ||
3178 	    copym->m_len < sizeof(struct ip6_hdr)) {
3179 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3180 		if (copym == NULL)
3181 			return;
3182 	}
3183 
3184 #ifdef DIAGNOSTIC
3185 	if (copym->m_len < sizeof(*ip6)) {
3186 		m_freem(copym);
3187 		return;
3188 	}
3189 #endif
3190 
3191 	ip6 = mtod(copym, struct ip6_hdr *);
3192 	/*
3193 	 * clear embedded scope identifiers if necessary.
3194 	 * in6_clearscope will touch the addresses only when necessary.
3195 	 */
3196 	in6_clearscope(&ip6->ip6_src);
3197 	in6_clearscope(&ip6->ip6_dst);
3198 
3199 	(void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL);
3200 }
3201 
3202 /*
3203  * Chop IPv6 header off from the payload.
3204  */
3205 static int
3206 ip6_splithdr(struct mbuf *m,  struct ip6_exthdrs *exthdrs)
3207 {
3208 	struct mbuf *mh;
3209 	struct ip6_hdr *ip6;
3210 
3211 	ip6 = mtod(m, struct ip6_hdr *);
3212 	if (m->m_len > sizeof(*ip6)) {
3213 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3214 		if (mh == 0) {
3215 			m_freem(m);
3216 			return ENOBUFS;
3217 		}
3218 		M_MOVE_PKTHDR(mh, m);
3219 		MH_ALIGN(mh, sizeof(*ip6));
3220 		m->m_len -= sizeof(*ip6);
3221 		m->m_data += sizeof(*ip6);
3222 		mh->m_next = m;
3223 		m = mh;
3224 		m->m_len = sizeof(*ip6);
3225 		bcopy((void *)ip6, mtod(m, void *), sizeof(*ip6));
3226 	}
3227 	exthdrs->ip6e_ip6 = m;
3228 	return 0;
3229 }
3230 
3231 /*
3232  * Compute IPv6 extension header length.
3233  */
3234 int
3235 ip6_optlen(struct in6pcb *in6p)
3236 {
3237 	int len;
3238 
3239 	if (!in6p->in6p_outputopts)
3240 		return 0;
3241 
3242 	len = 0;
3243 #define elen(x) \
3244     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3245 
3246 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3247 	len += elen(in6p->in6p_outputopts->ip6po_dest1);
3248 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3249 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3250 	return len;
3251 #undef elen
3252 }
3253