xref: /freebsd/sys/netinet6/ip6_output.c (revision 6b3455a7665208c366849f0b2b3bc916fb97516e)
1 /*	$FreeBSD$	*/
2 /*	$KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei 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  * 4. 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 "opt_ip6fw.h"
65 #include "opt_inet.h"
66 #include "opt_inet6.h"
67 #include "opt_ipsec.h"
68 #include "opt_pfil_hooks.h"
69 #include "opt_random_ip_id.h"
70 
71 #include <sys/param.h>
72 #include <sys/malloc.h>
73 #include <sys/mbuf.h>
74 #include <sys/proc.h>
75 #include <sys/errno.h>
76 #include <sys/protosw.h>
77 #include <sys/socket.h>
78 #include <sys/socketvar.h>
79 #include <sys/systm.h>
80 #include <sys/kernel.h>
81 
82 #include <net/if.h>
83 #include <net/route.h>
84 #ifdef PFIL_HOOKS
85 #include <net/pfil.h>
86 #endif
87 
88 #include <netinet/in.h>
89 #include <netinet/in_var.h>
90 #include <netinet6/in6_var.h>
91 #include <netinet/ip6.h>
92 #include <netinet/icmp6.h>
93 #include <netinet6/ip6_var.h>
94 #include <netinet/in_pcb.h>
95 #include <netinet/tcp_var.h>
96 #include <netinet6/nd6.h>
97 
98 #ifdef IPSEC
99 #include <netinet6/ipsec.h>
100 #ifdef INET6
101 #include <netinet6/ipsec6.h>
102 #endif
103 #include <netkey/key.h>
104 #endif /* IPSEC */
105 
106 #ifdef FAST_IPSEC
107 #include <netipsec/ipsec.h>
108 #include <netipsec/ipsec6.h>
109 #include <netipsec/key.h>
110 #endif /* FAST_IPSEC */
111 
112 #include <netinet6/ip6_fw.h>
113 
114 #include <net/net_osdep.h>
115 
116 #include <netinet6/ip6protosw.h>
117 
118 static MALLOC_DEFINE(M_IPMOPTS, "ip6_moptions", "internet multicast options");
119 
120 struct ip6_exthdrs {
121 	struct mbuf *ip6e_ip6;
122 	struct mbuf *ip6e_hbh;
123 	struct mbuf *ip6e_dest1;
124 	struct mbuf *ip6e_rthdr;
125 	struct mbuf *ip6e_dest2;
126 };
127 
128 static int ip6_pcbopt __P((int, u_char *, int, struct ip6_pktopts **,
129 			   int, int));
130 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
131 	struct socket *, struct sockopt *));
132 static int ip6_getpcbopt __P((struct ip6_pktopts *, int, struct sockopt *));
133 static int ip6_setpktoption __P((int, u_char *, int, struct ip6_pktopts *, int,
134 	int, int, int));
135 
136 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *));
137 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **));
138 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int));
139 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
140 	struct ip6_frag **));
141 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t));
142 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *));
143 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *,
144 	struct ifnet *, struct in6_addr *, u_long *, int *));
145 
146 
147 /*
148  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
149  * header (with pri, len, nxt, hlim, src, dst).
150  * This function may modify ver and hlim only.
151  * The mbuf chain containing the packet will be freed.
152  * The mbuf opt, if present, will not be freed.
153  *
154  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
155  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
156  * which is rt_rmx.rmx_mtu.
157  */
158 int
159 ip6_output(m0, opt, ro, flags, im6o, ifpp, inp)
160 	struct mbuf *m0;
161 	struct ip6_pktopts *opt;
162 	struct route_in6 *ro;
163 	int flags;
164 	struct ip6_moptions *im6o;
165 	struct ifnet **ifpp;		/* XXX: just for statistics */
166 	struct inpcb *inp;
167 {
168 	struct ip6_hdr *ip6, *mhip6;
169 	struct ifnet *ifp, *origifp;
170 	struct mbuf *m = m0;
171 	int hlen, tlen, len, off;
172 	struct route_in6 ip6route;
173 	struct sockaddr_in6 *dst;
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;
181 	struct route_in6 *ro_pmtu = NULL;
182 	int hdrsplit = 0;
183 	int needipsec = 0;
184 #if defined(IPSEC) || defined(FAST_IPSEC)
185 	int needipsectun = 0;
186 	struct secpolicy *sp = NULL;
187 #endif /*IPSEC || FAST_IPSEC*/
188 
189 	ip6 = mtod(m, struct ip6_hdr *);
190 	finaldst = ip6->ip6_dst;
191 
192 #define MAKE_EXTHDR(hp, mp)						\
193     do {								\
194 	if (hp) {							\
195 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
196 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
197 		    ((eh)->ip6e_len + 1) << 3);				\
198 		if (error)						\
199 			goto freehdrs;					\
200 	}								\
201     } while (/*CONSTCOND*/ 0)
202 
203 	bzero(&exthdrs, sizeof(exthdrs));
204 
205 	if (opt) {
206 		/* Hop-by-Hop options header */
207 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
208 		/* Destination options header(1st part) */
209 		if (opt->ip6po_rthdr) {
210 			/*
211 			 * Destination options header(1st part)
212 			 * This only makes sence with a routing header.
213 			 * See Section 9.2 of RFC 3542.
214 			 * Disabling this part just for MIP6 convenience is
215 			 * a bad idea.  We need to think carefully about a
216 			 * way to make the advanced API coexist with MIP6
217 			 * options, which might automatically be inserted in
218 			 * the kernel.
219 			 */
220 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
221 		}
222 		/* Routing header */
223 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
224 		/* Destination options header(2nd part) */
225 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
226 	}
227 
228 #ifdef IPSEC
229 	/* get a security policy for this packet */
230 	if (inp == NULL)
231 		sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
232 	else
233 		sp = ipsec6_getpolicybypcb(m, IPSEC_DIR_OUTBOUND, inp, &error);
234 
235 	if (sp == NULL) {
236 		ipsec6stat.out_inval++;
237 		goto freehdrs;
238 	}
239 
240 	error = 0;
241 
242 	/* check policy */
243 	switch (sp->policy) {
244 	case IPSEC_POLICY_DISCARD:
245 		/*
246 		 * This packet is just discarded.
247 		 */
248 		ipsec6stat.out_polvio++;
249 		goto freehdrs;
250 
251 	case IPSEC_POLICY_BYPASS:
252 	case IPSEC_POLICY_NONE:
253 		/* no need to do IPsec. */
254 		needipsec = 0;
255 		break;
256 
257 	case IPSEC_POLICY_IPSEC:
258 		if (sp->req == NULL) {
259 			/* acquire a policy */
260 			error = key_spdacquire(sp);
261 			goto freehdrs;
262 		}
263 		needipsec = 1;
264 		break;
265 
266 	case IPSEC_POLICY_ENTRUST:
267 	default:
268 		printf("ip6_output: Invalid policy found. %d\n", sp->policy);
269 	}
270 #endif /* IPSEC */
271 #ifdef FAST_IPSEC
272 	/* get a security policy for this packet */
273 	if (inp == NULL)
274 		sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
275 	else
276 		sp = ipsec_getpolicybysock(m, IPSEC_DIR_OUTBOUND, inp, &error);
277 
278 	if (sp == NULL) {
279 		newipsecstat.ips_out_inval++;
280 		goto freehdrs;
281 	}
282 
283 	error = 0;
284 
285 	/* check policy */
286 	switch (sp->policy) {
287 	case IPSEC_POLICY_DISCARD:
288 		/*
289 		 * This packet is just discarded.
290 		 */
291 		newipsecstat.ips_out_polvio++;
292 		goto freehdrs;
293 
294 	case IPSEC_POLICY_BYPASS:
295 	case IPSEC_POLICY_NONE:
296 		/* no need to do IPsec. */
297 		needipsec = 0;
298 		break;
299 
300 	case IPSEC_POLICY_IPSEC:
301 		if (sp->req == NULL) {
302 			/* acquire a policy */
303 			error = key_spdacquire(sp);
304 			goto freehdrs;
305 		}
306 		needipsec = 1;
307 		break;
308 
309 	case IPSEC_POLICY_ENTRUST:
310 	default:
311 		printf("ip6_output: Invalid policy found. %d\n", sp->policy);
312 	}
313 #endif /* FAST_IPSEC */
314 
315 	/*
316 	 * Calculate the total length of the extension header chain.
317 	 * Keep the length of the unfragmentable part for fragmentation.
318 	 */
319 	optlen = 0;
320 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
321 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
322 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
323 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
324 	/* NOTE: we don't add AH/ESP length here. do that later. */
325 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
326 
327 	/*
328 	 * If we need IPsec, or there is at least one extension header,
329 	 * separate IP6 header from the payload.
330 	 */
331 	if ((needipsec || optlen) && !hdrsplit) {
332 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
333 			m = NULL;
334 			goto freehdrs;
335 		}
336 		m = exthdrs.ip6e_ip6;
337 		hdrsplit++;
338 	}
339 
340 	/* adjust pointer */
341 	ip6 = mtod(m, struct ip6_hdr *);
342 
343 	/* adjust mbuf packet header length */
344 	m->m_pkthdr.len += optlen;
345 	plen = m->m_pkthdr.len - sizeof(*ip6);
346 
347 	/* If this is a jumbo payload, insert a jumbo payload option. */
348 	if (plen > IPV6_MAXPACKET) {
349 		if (!hdrsplit) {
350 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
351 				m = NULL;
352 				goto freehdrs;
353 			}
354 			m = exthdrs.ip6e_ip6;
355 			hdrsplit++;
356 		}
357 		/* adjust pointer */
358 		ip6 = mtod(m, struct ip6_hdr *);
359 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
360 			goto freehdrs;
361 		ip6->ip6_plen = 0;
362 	} else
363 		ip6->ip6_plen = htons(plen);
364 
365 	/*
366 	 * Concatenate headers and fill in next header fields.
367 	 * Here we have, on "m"
368 	 *	IPv6 payload
369 	 * and we insert headers accordingly.  Finally, we should be getting:
370 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
371 	 *
372 	 * during the header composing process, "m" points to IPv6 header.
373 	 * "mprev" points to an extension header prior to esp.
374 	 */
375 	{
376 		u_char *nexthdrp = &ip6->ip6_nxt;
377 		struct mbuf *mprev = m;
378 
379 		/*
380 		 * we treat dest2 specially.  this makes IPsec processing
381 		 * much easier.  the goal here is to make mprev point the
382 		 * mbuf prior to dest2.
383 		 *
384 		 * result: IPv6 dest2 payload
385 		 * m and mprev will point to IPv6 header.
386 		 */
387 		if (exthdrs.ip6e_dest2) {
388 			if (!hdrsplit)
389 				panic("assumption failed: hdr not split");
390 			exthdrs.ip6e_dest2->m_next = m->m_next;
391 			m->m_next = exthdrs.ip6e_dest2;
392 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
393 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
394 		}
395 
396 #define MAKE_CHAIN(m, mp, p, i)\
397     do {\
398 	if (m) {\
399 		if (!hdrsplit) \
400 			panic("assumption failed: hdr not split"); \
401 		*mtod((m), u_char *) = *(p);\
402 		*(p) = (i);\
403 		p = mtod((m), u_char *);\
404 		(m)->m_next = (mp)->m_next;\
405 		(mp)->m_next = (m);\
406 		(mp) = (m);\
407 	}\
408     } while (/*CONSTCOND*/ 0)
409 		/*
410 		 * result: IPv6 hbh dest1 rthdr dest2 payload
411 		 * m will point to IPv6 header.  mprev will point to the
412 		 * extension header prior to dest2 (rthdr in the above case).
413 		 */
414 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
415 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
416 		    IPPROTO_DSTOPTS);
417 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
418 		    IPPROTO_ROUTING);
419 
420 #if defined(IPSEC) || defined(FAST_IPSEC)
421 		if (!needipsec)
422 			goto skip_ipsec2;
423 
424 		/*
425 		 * pointers after IPsec headers are not valid any more.
426 		 * other pointers need a great care too.
427 		 * (IPsec routines should not mangle mbufs prior to AH/ESP)
428 		 */
429 		exthdrs.ip6e_dest2 = NULL;
430 
431 	    {
432 		struct ip6_rthdr *rh = NULL;
433 		int segleft_org = 0;
434 		struct ipsec_output_state state;
435 
436 		if (exthdrs.ip6e_rthdr) {
437 			rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
438 			segleft_org = rh->ip6r_segleft;
439 			rh->ip6r_segleft = 0;
440 		}
441 
442 		bzero(&state, sizeof(state));
443 		state.m = m;
444 		error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
445 		    &needipsectun);
446 		m = state.m;
447 		if (error) {
448 			/* mbuf is already reclaimed in ipsec6_output_trans. */
449 			m = NULL;
450 			switch (error) {
451 			case EHOSTUNREACH:
452 			case ENETUNREACH:
453 			case EMSGSIZE:
454 			case ENOBUFS:
455 			case ENOMEM:
456 				break;
457 			default:
458 				printf("ip6_output (ipsec): error code %d\n", error);
459 				/* FALLTHROUGH */
460 			case ENOENT:
461 				/* don't show these error codes to the user */
462 				error = 0;
463 				break;
464 			}
465 			goto bad;
466 		}
467 		if (exthdrs.ip6e_rthdr) {
468 			/* ah6_output doesn't modify mbuf chain */
469 			rh->ip6r_segleft = segleft_org;
470 		}
471 	    }
472 skip_ipsec2:;
473 #endif
474 	}
475 
476 	/*
477 	 * If there is a routing header, replace the destination address field
478 	 * with the first hop of the routing header.
479 	 */
480 	if (exthdrs.ip6e_rthdr) {
481 		struct ip6_rthdr *rh =
482 			(struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
483 						  struct ip6_rthdr *));
484 		struct ip6_rthdr0 *rh0;
485 		struct in6_addr *addrs;
486 
487 		switch (rh->ip6r_type) {
488 		case IPV6_RTHDR_TYPE_0:
489 			 rh0 = (struct ip6_rthdr0 *)rh;
490 			 addrs = (struct in6_addr *)(rh0 + 1);
491 
492 			 ip6->ip6_dst = *addrs;
493 			 bcopy((caddr_t)(addrs + 1), (caddr_t)addrs,
494 			       sizeof(struct in6_addr)*(rh0->ip6r0_segleft - 1)
495 				 );
496 			 *(addrs + rh0->ip6r0_segleft - 1) = finaldst;
497 			 break;
498 		default:	/* is it possible? */
499 			 error = EINVAL;
500 			 goto bad;
501 		}
502 	}
503 
504 	/* Source address validation */
505 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
506 	    (flags & IPV6_DADOUTPUT) == 0) {
507 		error = EOPNOTSUPP;
508 		ip6stat.ip6s_badscope++;
509 		goto bad;
510 	}
511 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
512 		error = EOPNOTSUPP;
513 		ip6stat.ip6s_badscope++;
514 		goto bad;
515 	}
516 
517 	ip6stat.ip6s_localout++;
518 
519 	/*
520 	 * Route packet.
521 	 */
522 	if (ro == 0) {
523 		ro = &ip6route;
524 		bzero((caddr_t)ro, sizeof(*ro));
525 	}
526 	ro_pmtu = ro;
527 	if (opt && opt->ip6po_rthdr)
528 		ro = &opt->ip6po_route;
529 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
530 
531 	/*
532 	 * If there is a cached route,
533 	 * check that it is to the same destination
534 	 * and is still up. If not, free it and try again.
535 	 */
536 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
537 			 dst->sin6_family != AF_INET6 ||
538 			 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
539 		RTFREE(ro->ro_rt);
540 		ro->ro_rt = (struct rtentry *)0;
541 	}
542 	if (ro->ro_rt == 0) {
543 		bzero(dst, sizeof(*dst));
544 		dst->sin6_family = AF_INET6;
545 		dst->sin6_len = sizeof(struct sockaddr_in6);
546 		dst->sin6_addr = ip6->ip6_dst;
547 	}
548 
549  	/*
550 	 * if specified, try to fill in the traffic class field.
551 	 * do not override if a non-zero value is already set.
552 	 * we check the diffserv field and the ecn field separately.
553 	 */
554 	if (opt && opt->ip6po_tclass >= 0) {
555 		int mask = 0;
556 
557 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
558 			mask |= 0xfc;
559 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
560 			mask |= 0x03;
561 		if (mask != 0)
562 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
563 	}
564 
565 	/* fill in or override the hop limit field, if necessary. */
566 	if (opt && opt->ip6po_hlim != -1)
567 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
568 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
569 		if (im6o != NULL)
570 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
571 		else
572 			ip6->ip6_hlim = ip6_defmcasthlim;
573 	}
574 
575 #if defined(IPSEC) || defined(FAST_IPSEC)
576 	if (needipsec && needipsectun) {
577 		struct ipsec_output_state state;
578 
579 		/*
580 		 * All the extension headers will become inaccessible
581 		 * (since they can be encrypted).
582 		 * Don't panic, we need no more updates to extension headers
583 		 * on inner IPv6 packet (since they are now encapsulated).
584 		 *
585 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
586 		 */
587 		bzero(&exthdrs, sizeof(exthdrs));
588 		exthdrs.ip6e_ip6 = m;
589 
590 		bzero(&state, sizeof(state));
591 		state.m = m;
592 		state.ro = (struct route *)ro;
593 		state.dst = (struct sockaddr *)dst;
594 
595 		error = ipsec6_output_tunnel(&state, sp, flags);
596 
597 		m = state.m;
598 		ro = (struct route_in6 *)state.ro;
599 		dst = (struct sockaddr_in6 *)state.dst;
600 		if (error) {
601 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
602 			m0 = m = NULL;
603 			m = NULL;
604 			switch (error) {
605 			case EHOSTUNREACH:
606 			case ENETUNREACH:
607 			case EMSGSIZE:
608 			case ENOBUFS:
609 			case ENOMEM:
610 				break;
611 			default:
612 				printf("ip6_output (ipsec): error code %d\n", error);
613 				/* FALLTHROUGH */
614 			case ENOENT:
615 				/* don't show these error codes to the user */
616 				error = 0;
617 				break;
618 			}
619 			goto bad;
620 		}
621 
622 		exthdrs.ip6e_ip6 = m;
623 	}
624 #endif /* IPSEC */
625 
626 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
627 		/* Unicast */
628 
629 #define ifatoia6(ifa)	((struct in6_ifaddr *)(ifa))
630 #define sin6tosa(sin6)	((struct sockaddr *)(sin6))
631 		/* xxx
632 		 * interface selection comes here
633 		 * if an interface is specified from an upper layer,
634 		 * ifp must point it.
635 		 */
636 		if (ro->ro_rt == 0) {
637 			/*
638 			 * non-bsdi always clone routes, if parent is
639 			 * PRF_CLONING.
640 			 */
641 			rtalloc((struct route *)ro);
642 		}
643 		if (ro->ro_rt == 0) {
644 			ip6stat.ip6s_noroute++;
645 			error = EHOSTUNREACH;
646 			/* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
647 			goto bad;
648 		}
649 		/* XXX rt not locked */
650 		ia = ifatoia6(ro->ro_rt->rt_ifa);
651 		ifp = ro->ro_rt->rt_ifp;
652 		ro->ro_rt->rt_rmx.rmx_pksent++;
653 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
654 			dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
655 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
656 
657 		in6_ifstat_inc(ifp, ifs6_out_request);
658 
659 		/*
660 		 * Check if the outgoing interface conflicts with
661 		 * the interface specified by ifi6_ifindex (if specified).
662 		 * Note that loopback interface is always okay.
663 		 * (this may happen when we are sending a packet to one of
664 		 *  our own addresses.)
665 		 */
666 		if (opt && opt->ip6po_pktinfo
667 		 && opt->ip6po_pktinfo->ipi6_ifindex) {
668 			if (!(ifp->if_flags & IFF_LOOPBACK)
669 			 && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
670 				ip6stat.ip6s_noroute++;
671 				in6_ifstat_inc(ifp, ifs6_out_discard);
672 				error = EHOSTUNREACH;
673 				goto bad;
674 			}
675 		}
676 
677 		if (opt && opt->ip6po_hlim != -1)
678 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
679 	} else {
680 		/* Multicast */
681 		struct	in6_multi *in6m;
682 
683 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
684 
685 		/*
686 		 * See if the caller provided any multicast options
687 		 */
688 		ifp = NULL;
689 		if (im6o != NULL) {
690 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
691 			if (im6o->im6o_multicast_ifp != NULL)
692 				ifp = im6o->im6o_multicast_ifp;
693 		} else
694 			ip6->ip6_hlim = ip6_defmcasthlim;
695 
696 		/*
697 		 * See if the caller provided the outgoing interface
698 		 * as an ancillary data.
699 		 * Boundary check for ifindex is assumed to be already done.
700 		 */
701 		if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
702 			ifp = ifnet_byindex(opt->ip6po_pktinfo->ipi6_ifindex);
703 
704 		/*
705 		 * If the destination is a node-local scope multicast,
706 		 * the packet should be loop-backed only.
707 		 */
708 		if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
709 			/*
710 			 * If the outgoing interface is already specified,
711 			 * it should be a loopback interface.
712 			 */
713 			if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) {
714 				ip6stat.ip6s_badscope++;
715 				error = ENETUNREACH; /* XXX: better error? */
716 				/* XXX correct ifp? */
717 				in6_ifstat_inc(ifp, ifs6_out_discard);
718 				goto bad;
719 			} else {
720 				ifp = &loif[0];
721 			}
722 		}
723 
724 		if (opt && opt->ip6po_hlim != -1)
725 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
726 
727 		/*
728 		 * If caller did not provide an interface lookup a
729 		 * default in the routing table.  This is either a
730 		 * default for the speicfied group (i.e. a host
731 		 * route), or a multicast default (a route for the
732 		 * ``net'' ff00::/8).
733 		 */
734 		if (ifp == NULL) {
735 			if (ro->ro_rt == 0)
736 				ro->ro_rt = rtalloc1((struct sockaddr *)
737 						&ro->ro_dst, 0, 0UL);
738 			else
739 				RT_LOCK(ro->ro_rt);
740 			if (ro->ro_rt == 0) {
741 				ip6stat.ip6s_noroute++;
742 				error = EHOSTUNREACH;
743 				/* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
744 				goto bad;
745 			}
746 			ia = ifatoia6(ro->ro_rt->rt_ifa);
747 			ifp = ro->ro_rt->rt_ifp;
748 			ro->ro_rt->rt_rmx.rmx_pksent++;
749 			RT_UNLOCK(ro->ro_rt);
750 		}
751 
752 		if ((flags & IPV6_FORWARDING) == 0)
753 			in6_ifstat_inc(ifp, ifs6_out_request);
754 		in6_ifstat_inc(ifp, ifs6_out_mcast);
755 
756 		/*
757 		 * Confirm that the outgoing interface supports multicast.
758 		 */
759 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
760 			ip6stat.ip6s_noroute++;
761 			in6_ifstat_inc(ifp, ifs6_out_discard);
762 			error = ENETUNREACH;
763 			goto bad;
764 		}
765 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
766 		if (in6m != NULL &&
767 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
768 			/*
769 			 * If we belong to the destination multicast group
770 			 * on the outgoing interface, and the caller did not
771 			 * forbid loopback, loop back a copy.
772 			 */
773 			ip6_mloopback(ifp, m, dst);
774 		} else {
775 			/*
776 			 * If we are acting as a multicast router, perform
777 			 * multicast forwarding as if the packet had just
778 			 * arrived on the interface to which we are about
779 			 * to send.  The multicast forwarding function
780 			 * recursively calls this function, using the
781 			 * IPV6_FORWARDING flag to prevent infinite recursion.
782 			 *
783 			 * Multicasts that are looped back by ip6_mloopback(),
784 			 * above, will be forwarded by the ip6_input() routine,
785 			 * if necessary.
786 			 */
787 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
788 				if (ip6_mforward(ip6, ifp, m) != 0) {
789 					m_freem(m);
790 					goto done;
791 				}
792 			}
793 		}
794 		/*
795 		 * Multicasts with a hoplimit of zero may be looped back,
796 		 * above, but must not be transmitted on a network.
797 		 * Also, multicasts addressed to the loopback interface
798 		 * are not sent -- the above call to ip6_mloopback() will
799 		 * loop back a copy if this host actually belongs to the
800 		 * destination group on the loopback interface.
801 		 */
802 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
803 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
804 			m_freem(m);
805 			goto done;
806 		}
807 	}
808 
809 	/*
810 	 * Fill the outgoing inteface to tell the upper layer
811 	 * to increment per-interface statistics.
812 	 */
813 	if (ifpp)
814 		*ifpp = ifp;
815 
816 	/* Determine path MTU. */
817 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
818 	    &alwaysfrag)) != 0)
819 		goto bad;
820 
821 	/*
822 	 * The caller of this function may specify to use the minimum MTU
823 	 * in some cases.
824 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
825 	 * setting.  The logic is a bit complicated; by default, unicast
826 	 * packets will follow path MTU while multicast packets will be sent at
827 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
828 	 * including unicast ones will be sent at the minimum MTU.  Multicast
829 	 * packets will always be sent at the minimum MTU unless
830 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
831 	 * See RFC 3542 for more details.
832 	 */
833 	if (mtu > IPV6_MMTU) {
834 		if ((flags & IPV6_MINMTU))
835 			mtu = IPV6_MMTU;
836 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
837 			mtu = IPV6_MMTU;
838 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
839 			 (opt == NULL ||
840 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
841 			mtu = IPV6_MMTU;
842 		}
843 	}
844 
845 	/* Fake scoped addresses */
846 	if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
847 		/*
848 		 * If source or destination address is a scoped address, and
849 		 * the packet is going to be sent to a loopback interface,
850 		 * we should keep the original interface.
851 		 */
852 
853 		/*
854 		 * XXX: this is a very experimental and temporary solution.
855 		 * We eventually have sockaddr_in6 and use the sin6_scope_id
856 		 * field of the structure here.
857 		 * We rely on the consistency between two scope zone ids
858 		 * of source and destination, which should already be assured.
859 		 * Larger scopes than link will be supported in the future.
860 		 */
861 		origifp = NULL;
862 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
863 			origifp = ifnet_byindex(ntohs(ip6->ip6_src.s6_addr16[1]));
864 		else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
865 			origifp = ifnet_byindex(ntohs(ip6->ip6_dst.s6_addr16[1]));
866 		/*
867 		 * XXX: origifp can be NULL even in those two cases above.
868 		 * For example, if we remove the (only) link-local address
869 		 * from the loopback interface, and try to send a link-local
870 		 * address without link-id information.  Then the source
871 		 * address is ::1, and the destination address is the
872 		 * link-local address with its s6_addr16[1] being zero.
873 		 * What is worse, if the packet goes to the loopback interface
874 		 * by a default rejected route, the null pointer would be
875 		 * passed to looutput, and the kernel would hang.
876 		 * The following last resort would prevent such disaster.
877 		 */
878 		if (origifp == NULL)
879 			origifp = ifp;
880 	}
881 	else
882 		origifp = ifp;
883 	/*
884 	 * clear embedded scope identifiers if necessary.
885 	 * in6_clearscope will touch the addresses only when necessary.
886 	 */
887 	in6_clearscope(&ip6->ip6_src);
888 	in6_clearscope(&ip6->ip6_dst);
889 
890 	/*
891 	 * Check with the firewall...
892 	 */
893 	if (ip6_fw_enable && ip6_fw_chk_ptr) {
894 		u_short port = 0;
895 		m->m_pkthdr.rcvif = NULL;	/* XXX */
896 		/* If ipfw says divert, we have to just drop packet */
897 		if ((*ip6_fw_chk_ptr)(&ip6, ifp, &port, &m)) {
898 			m_freem(m);
899 			goto done;
900 		}
901 		if (!m) {
902 			error = EACCES;
903 			goto done;
904 		}
905 	}
906 
907 	/*
908 	 * If the outgoing packet contains a hop-by-hop options header,
909 	 * it must be examined and processed even by the source node.
910 	 * (RFC 2460, section 4.)
911 	 */
912 	if (exthdrs.ip6e_hbh) {
913 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
914 		u_int32_t dummy1; /* XXX unused */
915 		u_int32_t dummy2; /* XXX unused */
916 
917 #ifdef DIAGNOSTIC
918 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
919 			panic("ip6e_hbh is not continuous");
920 #endif
921 		/*
922 		 *  XXX: if we have to send an ICMPv6 error to the sender,
923 		 *       we need the M_LOOP flag since icmp6_error() expects
924 		 *       the IPv6 and the hop-by-hop options header are
925 		 *       continuous unless the flag is set.
926 		 */
927 		m->m_flags |= M_LOOP;
928 		m->m_pkthdr.rcvif = ifp;
929 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
930 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
931 		    &dummy1, &dummy2) < 0) {
932 			/* m was already freed at this point */
933 			error = EINVAL;/* better error? */
934 			goto done;
935 		}
936 		m->m_flags &= ~M_LOOP; /* XXX */
937 		m->m_pkthdr.rcvif = NULL;
938 	}
939 
940 #ifdef PFIL_HOOKS
941 	/*
942 	 * Run through list of hooks for output packets.
943 	 */
944 	error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT);
945 	if (error != 0 || m == NULL)
946 		goto done;
947 	ip6 = mtod(m, struct ip6_hdr *);
948 #endif /* PFIL_HOOKS */
949 
950 	/*
951 	 * Send the packet to the outgoing interface.
952 	 * If necessary, do IPv6 fragmentation before sending.
953 	 *
954 	 * the logic here is rather complex:
955 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
956 	 * 1-a:	send as is if tlen <= path mtu
957 	 * 1-b:	fragment if tlen > path mtu
958 	 *
959 	 * 2: if user asks us not to fragment (dontfrag == 1)
960 	 * 2-a:	send as is if tlen <= interface mtu
961 	 * 2-b:	error if tlen > interface mtu
962 	 *
963 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
964 	 *	always fragment
965 	 *
966 	 * 4: if dontfrag == 1 && alwaysfrag == 1
967 	 *	error, as we cannot handle this conflicting request
968 	 */
969 	tlen = m->m_pkthdr.len;
970 
971 	if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
972 		dontfrag = 1;
973 	else
974 		dontfrag = 0;
975 	if (dontfrag && alwaysfrag) {	/* case 4 */
976 		/* conflicting request - can't transmit */
977 		error = EMSGSIZE;
978 		goto bad;
979 	}
980 	if (dontfrag && tlen > IN6_LINKMTU(ifp)) {	/* case 2-b */
981 		/*
982 		 * Even if the DONTFRAG option is specified, we cannot send the
983 		 * packet when the data length is larger than the MTU of the
984 		 * outgoing interface.
985 		 * Notify the error by sending IPV6_PATHMTU ancillary data as
986 		 * well as returning an error code (the latter is not described
987 		 * in the API spec.)
988 		 */
989 		u_int32_t mtu32;
990 		struct ip6ctlparam ip6cp;
991 
992 		mtu32 = (u_int32_t)mtu;
993 		bzero(&ip6cp, sizeof(ip6cp));
994 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
995 		pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
996 		    (void *)&ip6cp);
997 
998 		error = EMSGSIZE;
999 		goto bad;
1000 	}
1001 
1002 	/*
1003 	 * transmit packet without fragmentation
1004 	 */
1005 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
1006 		struct in6_ifaddr *ia6;
1007 
1008 		ip6 = mtod(m, struct ip6_hdr *);
1009 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1010 		if (ia6) {
1011 			/* Record statistics for this interface address. */
1012 			ia6->ia_ifa.if_opackets++;
1013 			ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
1014 		}
1015 #ifdef IPSEC
1016 		/* clean ipsec history once it goes out of the node */
1017 		ipsec_delaux(m);
1018 #endif
1019 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1020 		goto done;
1021 	}
1022 
1023 	/*
1024 	 * try to fragment the packet.  case 1-b and 3
1025 	 */
1026 	if (mtu < IPV6_MMTU) {
1027 		/* path MTU cannot be less than IPV6_MMTU */
1028 		error = EMSGSIZE;
1029 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1030 		goto bad;
1031 	} else if (ip6->ip6_plen == 0) {
1032 		/* jumbo payload cannot be fragmented */
1033 		error = EMSGSIZE;
1034 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1035 		goto bad;
1036 	} else {
1037 		struct mbuf **mnext, *m_frgpart;
1038 		struct ip6_frag *ip6f;
1039 #ifdef RANDOM_IP_ID
1040 		u_int32_t id = htonl(ip6_randomid());
1041 #else
1042 		u_int32_t id = htonl(ip6_id++);
1043 #endif
1044 		u_char nextproto;
1045 		struct ip6ctlparam ip6cp;
1046 		u_int32_t mtu32;
1047 		int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
1048 
1049 		/*
1050 		 * Too large for the destination or interface;
1051 		 * fragment if possible.
1052 		 * Must be able to put at least 8 bytes per fragment.
1053 		 */
1054 		hlen = unfragpartlen;
1055 		if (mtu > IPV6_MAXPACKET)
1056 			mtu = IPV6_MAXPACKET;
1057 
1058 		/* Notify a proper path MTU to applications. */
1059 		mtu32 = (u_int32_t)mtu;
1060 		bzero(&ip6cp, sizeof(ip6cp));
1061 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
1062 		pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
1063 		    (void *)&ip6cp);
1064 
1065 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1066 		if (len < 8) {
1067 			error = EMSGSIZE;
1068 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
1069 			goto bad;
1070 		}
1071 
1072 		/*
1073 		 * Verify that we have any chance at all of being able to queue
1074 		 *      the packet or packet fragments
1075 		 */
1076 		if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
1077 		    < tlen  /* - hlen */)) {
1078 			error = ENOBUFS;
1079 			ip6stat.ip6s_odropped++;
1080 			goto bad;
1081 		}
1082 
1083 		mnext = &m->m_nextpkt;
1084 
1085 		/*
1086 		 * Change the next header field of the last header in the
1087 		 * unfragmentable part.
1088 		 */
1089 		if (exthdrs.ip6e_rthdr) {
1090 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1091 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1092 		} else if (exthdrs.ip6e_dest1) {
1093 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1094 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1095 		} else if (exthdrs.ip6e_hbh) {
1096 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1097 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1098 		} else {
1099 			nextproto = ip6->ip6_nxt;
1100 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1101 		}
1102 
1103 		/*
1104 		 * Loop through length of segment after first fragment,
1105 		 * make new header and copy data of each part and link onto
1106 		 * chain.
1107 		 */
1108 		m0 = m;
1109 		for (off = hlen; off < tlen; off += len) {
1110 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
1111 			if (!m) {
1112 				error = ENOBUFS;
1113 				ip6stat.ip6s_odropped++;
1114 				goto sendorfree;
1115 			}
1116 			m->m_pkthdr.rcvif = NULL;
1117 			m->m_flags = m0->m_flags & M_COPYFLAGS;
1118 			*mnext = m;
1119 			mnext = &m->m_nextpkt;
1120 			m->m_data += max_linkhdr;
1121 			mhip6 = mtod(m, struct ip6_hdr *);
1122 			*mhip6 = *ip6;
1123 			m->m_len = sizeof(*mhip6);
1124 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1125 			if (error) {
1126 				ip6stat.ip6s_odropped++;
1127 				goto sendorfree;
1128 			}
1129 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1130 			if (off + len >= tlen)
1131 				len = tlen - off;
1132 			else
1133 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1134 			mhip6->ip6_plen = htons((u_short)(len + hlen +
1135 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1136 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1137 				error = ENOBUFS;
1138 				ip6stat.ip6s_odropped++;
1139 				goto sendorfree;
1140 			}
1141 			m_cat(m, m_frgpart);
1142 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1143 			m->m_pkthdr.rcvif = (struct ifnet *)0;
1144 			ip6f->ip6f_reserved = 0;
1145 			ip6f->ip6f_ident = id;
1146 			ip6f->ip6f_nxt = nextproto;
1147 			ip6stat.ip6s_ofragments++;
1148 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1149 		}
1150 
1151 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1152 	}
1153 
1154 	/*
1155 	 * Remove leading garbages.
1156 	 */
1157 sendorfree:
1158 	m = m0->m_nextpkt;
1159 	m0->m_nextpkt = 0;
1160 	m_freem(m0);
1161 	for (m0 = m; m; m = m0) {
1162 		m0 = m->m_nextpkt;
1163 		m->m_nextpkt = 0;
1164 		if (error == 0) {
1165  			/* Record statistics for this interface address. */
1166  			if (ia) {
1167  				ia->ia_ifa.if_opackets++;
1168  				ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1169  			}
1170 #ifdef IPSEC
1171 			/* clean ipsec history once it goes out of the node */
1172 			ipsec_delaux(m);
1173 #endif
1174 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1175 		} else
1176 			m_freem(m);
1177 	}
1178 
1179 	if (error == 0)
1180 		ip6stat.ip6s_fragmented++;
1181 
1182 done:
1183 	if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1184 		RTFREE(ro->ro_rt);
1185 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1186 		RTFREE(ro_pmtu->ro_rt);
1187 	}
1188 
1189 #ifdef IPSEC
1190 	if (sp != NULL)
1191 		key_freesp(sp);
1192 #endif /* IPSEC */
1193 #ifdef FAST_IPSEC
1194 	if (sp != NULL)
1195 		KEY_FREESP(&sp);
1196 #endif /* FAST_IPSEC */
1197 
1198 	return (error);
1199 
1200 freehdrs:
1201 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1202 	m_freem(exthdrs.ip6e_dest1);
1203 	m_freem(exthdrs.ip6e_rthdr);
1204 	m_freem(exthdrs.ip6e_dest2);
1205 	/* FALLTHROUGH */
1206 bad:
1207 	m_freem(m);
1208 	goto done;
1209 }
1210 
1211 static int
1212 ip6_copyexthdr(mp, hdr, hlen)
1213 	struct mbuf **mp;
1214 	caddr_t hdr;
1215 	int hlen;
1216 {
1217 	struct mbuf *m;
1218 
1219 	if (hlen > MCLBYTES)
1220 		return (ENOBUFS); /* XXX */
1221 
1222 	MGET(m, M_DONTWAIT, MT_DATA);
1223 	if (!m)
1224 		return (ENOBUFS);
1225 
1226 	if (hlen > MLEN) {
1227 		MCLGET(m, M_DONTWAIT);
1228 		if ((m->m_flags & M_EXT) == 0) {
1229 			m_free(m);
1230 			return (ENOBUFS);
1231 		}
1232 	}
1233 	m->m_len = hlen;
1234 	if (hdr)
1235 		bcopy(hdr, mtod(m, caddr_t), hlen);
1236 
1237 	*mp = m;
1238 	return (0);
1239 }
1240 
1241 /*
1242  * Insert jumbo payload option.
1243  */
1244 static int
1245 ip6_insert_jumboopt(exthdrs, plen)
1246 	struct ip6_exthdrs *exthdrs;
1247 	u_int32_t plen;
1248 {
1249 	struct mbuf *mopt;
1250 	u_char *optbuf;
1251 	u_int32_t v;
1252 
1253 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1254 
1255 	/*
1256 	 * If there is no hop-by-hop options header, allocate new one.
1257 	 * If there is one but it doesn't have enough space to store the
1258 	 * jumbo payload option, allocate a cluster to store the whole options.
1259 	 * Otherwise, use it to store the options.
1260 	 */
1261 	if (exthdrs->ip6e_hbh == 0) {
1262 		MGET(mopt, M_DONTWAIT, MT_DATA);
1263 		if (mopt == 0)
1264 			return (ENOBUFS);
1265 		mopt->m_len = JUMBOOPTLEN;
1266 		optbuf = mtod(mopt, u_char *);
1267 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1268 		exthdrs->ip6e_hbh = mopt;
1269 	} else {
1270 		struct ip6_hbh *hbh;
1271 
1272 		mopt = exthdrs->ip6e_hbh;
1273 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1274 			/*
1275 			 * XXX assumption:
1276 			 * - exthdrs->ip6e_hbh is not referenced from places
1277 			 *   other than exthdrs.
1278 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1279 			 */
1280 			int oldoptlen = mopt->m_len;
1281 			struct mbuf *n;
1282 
1283 			/*
1284 			 * XXX: give up if the whole (new) hbh header does
1285 			 * not fit even in an mbuf cluster.
1286 			 */
1287 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1288 				return (ENOBUFS);
1289 
1290 			/*
1291 			 * As a consequence, we must always prepare a cluster
1292 			 * at this point.
1293 			 */
1294 			MGET(n, M_DONTWAIT, MT_DATA);
1295 			if (n) {
1296 				MCLGET(n, M_DONTWAIT);
1297 				if ((n->m_flags & M_EXT) == 0) {
1298 					m_freem(n);
1299 					n = NULL;
1300 				}
1301 			}
1302 			if (!n)
1303 				return (ENOBUFS);
1304 			n->m_len = oldoptlen + JUMBOOPTLEN;
1305 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1306 			    oldoptlen);
1307 			optbuf = mtod(n, caddr_t) + oldoptlen;
1308 			m_freem(mopt);
1309 			mopt = exthdrs->ip6e_hbh = n;
1310 		} else {
1311 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1312 			mopt->m_len += JUMBOOPTLEN;
1313 		}
1314 		optbuf[0] = IP6OPT_PADN;
1315 		optbuf[1] = 1;
1316 
1317 		/*
1318 		 * Adjust the header length according to the pad and
1319 		 * the jumbo payload option.
1320 		 */
1321 		hbh = mtod(mopt, struct ip6_hbh *);
1322 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1323 	}
1324 
1325 	/* fill in the option. */
1326 	optbuf[2] = IP6OPT_JUMBO;
1327 	optbuf[3] = 4;
1328 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1329 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1330 
1331 	/* finally, adjust the packet header length */
1332 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1333 
1334 	return (0);
1335 #undef JUMBOOPTLEN
1336 }
1337 
1338 /*
1339  * Insert fragment header and copy unfragmentable header portions.
1340  */
1341 static int
1342 ip6_insertfraghdr(m0, m, hlen, frghdrp)
1343 	struct mbuf *m0, *m;
1344 	int hlen;
1345 	struct ip6_frag **frghdrp;
1346 {
1347 	struct mbuf *n, *mlast;
1348 
1349 	if (hlen > sizeof(struct ip6_hdr)) {
1350 		n = m_copym(m0, sizeof(struct ip6_hdr),
1351 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1352 		if (n == 0)
1353 			return (ENOBUFS);
1354 		m->m_next = n;
1355 	} else
1356 		n = m;
1357 
1358 	/* Search for the last mbuf of unfragmentable part. */
1359 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1360 		;
1361 
1362 	if ((mlast->m_flags & M_EXT) == 0 &&
1363 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1364 		/* use the trailing space of the last mbuf for the fragment hdr */
1365 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1366 		    mlast->m_len);
1367 		mlast->m_len += sizeof(struct ip6_frag);
1368 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1369 	} else {
1370 		/* allocate a new mbuf for the fragment header */
1371 		struct mbuf *mfrg;
1372 
1373 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1374 		if (mfrg == 0)
1375 			return (ENOBUFS);
1376 		mfrg->m_len = sizeof(struct ip6_frag);
1377 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1378 		mlast->m_next = mfrg;
1379 	}
1380 
1381 	return (0);
1382 }
1383 
1384 static int
1385 ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup, alwaysfragp)
1386 	struct route_in6 *ro_pmtu, *ro;
1387 	struct ifnet *ifp;
1388 	struct in6_addr *dst;
1389 	u_long *mtup;
1390 	int *alwaysfragp;
1391 {
1392 	u_int32_t mtu = 0;
1393 	int alwaysfrag = 0;
1394 	int error = 0;
1395 
1396 	if (ro_pmtu != ro) {
1397 		/* The first hop and the final destination may differ. */
1398 		struct sockaddr_in6 *sa6_dst =
1399 		    (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1400 		if (ro_pmtu->ro_rt &&
1401 		    ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1402 		     !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1403 			RTFREE(ro_pmtu->ro_rt);
1404 			ro_pmtu->ro_rt = (struct rtentry *)NULL;
1405 		}
1406 		if (ro_pmtu->ro_rt == NULL) {
1407 			bzero(sa6_dst, sizeof(*sa6_dst));
1408 			sa6_dst->sin6_family = AF_INET6;
1409 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1410 			sa6_dst->sin6_addr = *dst;
1411 
1412 			rtalloc((struct route *)ro_pmtu);
1413 		}
1414 	}
1415 	if (ro_pmtu->ro_rt) {
1416 		u_int32_t ifmtu;
1417 		struct in_conninfo inc;
1418 
1419 		bzero(&inc, sizeof(inc));
1420 		inc.inc_flags = 1; /* IPv6 */
1421 		inc.inc6_faddr = *dst;
1422 
1423 		if (ifp == NULL)
1424 			ifp = ro_pmtu->ro_rt->rt_ifp;
1425 		ifmtu = IN6_LINKMTU(ifp);
1426 		mtu = tcp_hc_getmtu(&inc);
1427 		if (mtu)
1428 			mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu);
1429 		else
1430 			mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1431 		if (mtu == 0)
1432 			mtu = ifmtu;
1433 		else if (mtu < IPV6_MMTU) {
1434 			/*
1435 			 * RFC2460 section 5, last paragraph:
1436 			 * if we record ICMPv6 too big message with
1437 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1438 			 * or smaller, with framgent header attached.
1439 			 * (fragment header is needed regardless from the
1440 			 * packet size, for translators to identify packets)
1441 			 */
1442 			alwaysfrag = 1;
1443 			mtu = IPV6_MMTU;
1444 		} else if (mtu > ifmtu) {
1445 			/*
1446 			 * The MTU on the route is larger than the MTU on
1447 			 * the interface!  This shouldn't happen, unless the
1448 			 * MTU of the interface has been changed after the
1449 			 * interface was brought up.  Change the MTU in the
1450 			 * route to match the interface MTU (as long as the
1451 			 * field isn't locked).
1452 			 */
1453 			mtu = ifmtu;
1454 			ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1455 		}
1456 	} else if (ifp) {
1457 		mtu = IN6_LINKMTU(ifp);
1458 	} else
1459 		error = EHOSTUNREACH; /* XXX */
1460 
1461 	*mtup = mtu;
1462 	if (alwaysfragp)
1463 		*alwaysfragp = alwaysfrag;
1464 	return (error);
1465 }
1466 
1467 /*
1468  * IP6 socket option processing.
1469  */
1470 int
1471 ip6_ctloutput(so, sopt)
1472 	struct socket *so;
1473 	struct sockopt *sopt;
1474 {
1475 	int privileged, optdatalen, uproto;
1476 	void *optdata;
1477 	struct inpcb *in6p = sotoinpcb(so);
1478 	int error, optval;
1479 	int level, op, optname;
1480 	int optlen;
1481 	struct thread *td;
1482 
1483 	if (sopt) {
1484 		level = sopt->sopt_level;
1485 		op = sopt->sopt_dir;
1486 		optname = sopt->sopt_name;
1487 		optlen = sopt->sopt_valsize;
1488 		td = sopt->sopt_td;
1489 	} else {
1490 		panic("ip6_ctloutput: arg soopt is NULL");
1491 	}
1492 	error = optval = 0;
1493 
1494 	privileged = (td == 0 || suser(td)) ? 0 : 1;
1495 	uproto = (int)so->so_proto->pr_protocol;
1496 
1497 	if (level == IPPROTO_IPV6) {
1498 		switch (op) {
1499 
1500 		case SOPT_SET:
1501 			switch (optname) {
1502 			case IPV6_2292PKTOPTIONS:
1503 #ifdef IPV6_PKTOPTIONS
1504 			case IPV6_PKTOPTIONS:
1505 #endif
1506 			{
1507 				struct mbuf *m;
1508 
1509 				error = soopt_getm(sopt, &m); /* XXX */
1510 				if (error != 0)
1511 					break;
1512 				error = soopt_mcopyin(sopt, m); /* XXX */
1513 				if (error != 0)
1514 					break;
1515 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1516 						    m, so, sopt);
1517 				m_freem(m); /* XXX */
1518 				break;
1519 			}
1520 
1521 			/*
1522 			 * Use of some Hop-by-Hop options or some
1523 			 * Destination options, might require special
1524 			 * privilege.  That is, normal applications
1525 			 * (without special privilege) might be forbidden
1526 			 * from setting certain options in outgoing packets,
1527 			 * and might never see certain options in received
1528 			 * packets. [RFC 2292 Section 6]
1529 			 * KAME specific note:
1530 			 *  KAME prevents non-privileged users from sending or
1531 			 *  receiving ANY hbh/dst options in order to avoid
1532 			 *  overhead of parsing options in the kernel.
1533 			 */
1534 			case IPV6_RECVHOPOPTS:
1535 			case IPV6_RECVDSTOPTS:
1536 			case IPV6_RECVRTHDRDSTOPTS:
1537 				if (!privileged) {
1538 					error = EPERM;
1539 					break;
1540 				}
1541 				/* FALLTHROUGH */
1542 			case IPV6_UNICAST_HOPS:
1543 			case IPV6_HOPLIMIT:
1544 			case IPV6_FAITH:
1545 
1546 			case IPV6_RECVPKTINFO:
1547 			case IPV6_RECVHOPLIMIT:
1548 			case IPV6_RECVRTHDR:
1549 			case IPV6_RECVPATHMTU:
1550 			case IPV6_RECVTCLASS:
1551 			case IPV6_V6ONLY:
1552 			case IPV6_AUTOFLOWLABEL:
1553 				if (optlen != sizeof(int)) {
1554 					error = EINVAL;
1555 					break;
1556 				}
1557 				error = sooptcopyin(sopt, &optval,
1558 					sizeof optval, sizeof optval);
1559 				if (error)
1560 					break;
1561 				switch (optname) {
1562 
1563 				case IPV6_UNICAST_HOPS:
1564 					if (optval < -1 || optval >= 256)
1565 						error = EINVAL;
1566 					else {
1567 						/* -1 = kernel default */
1568 						in6p->in6p_hops = optval;
1569 						if ((in6p->in6p_vflag &
1570 						     INP_IPV4) != 0)
1571 							in6p->inp_ip_ttl = optval;
1572 					}
1573 					break;
1574 #define OPTSET(bit) \
1575 do { \
1576 	if (optval) \
1577 		in6p->in6p_flags |= (bit); \
1578 	else \
1579 		in6p->in6p_flags &= ~(bit); \
1580 } while (/*CONSTCOND*/ 0)
1581 #define OPTSET2292(bit) \
1582 do { \
1583 	in6p->in6p_flags |= IN6P_RFC2292; \
1584 	if (optval) \
1585 		in6p->in6p_flags |= (bit); \
1586 	else \
1587 		in6p->in6p_flags &= ~(bit); \
1588 } while (/*CONSTCOND*/ 0)
1589 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1590 
1591 				case IPV6_RECVPKTINFO:
1592 					/* cannot mix with RFC2292 */
1593 					if (OPTBIT(IN6P_RFC2292)) {
1594 						error = EINVAL;
1595 						break;
1596 					}
1597 					OPTSET(IN6P_PKTINFO);
1598 					break;
1599 
1600 				case IPV6_HOPLIMIT:
1601 				{
1602 					struct ip6_pktopts **optp;
1603 
1604 					/* cannot mix with RFC2292 */
1605 					if (OPTBIT(IN6P_RFC2292)) {
1606 						error = EINVAL;
1607 						break;
1608 					}
1609 					optp = &in6p->in6p_outputopts;
1610 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1611 							   (u_char *)&optval,
1612 							   sizeof(optval),
1613 							   optp,
1614 							   privileged, uproto);
1615 					break;
1616 				}
1617 
1618 				case IPV6_RECVHOPLIMIT:
1619 					/* cannot mix with RFC2292 */
1620 					if (OPTBIT(IN6P_RFC2292)) {
1621 						error = EINVAL;
1622 						break;
1623 					}
1624 					OPTSET(IN6P_HOPLIMIT);
1625 					break;
1626 
1627 				case IPV6_RECVHOPOPTS:
1628 					/* cannot mix with RFC2292 */
1629 					if (OPTBIT(IN6P_RFC2292)) {
1630 						error = EINVAL;
1631 						break;
1632 					}
1633 					OPTSET(IN6P_HOPOPTS);
1634 					break;
1635 
1636 				case IPV6_RECVDSTOPTS:
1637 					/* cannot mix with RFC2292 */
1638 					if (OPTBIT(IN6P_RFC2292)) {
1639 						error = EINVAL;
1640 						break;
1641 					}
1642 					OPTSET(IN6P_DSTOPTS);
1643 					break;
1644 
1645 				case IPV6_RECVRTHDRDSTOPTS:
1646 					/* cannot mix with RFC2292 */
1647 					if (OPTBIT(IN6P_RFC2292)) {
1648 						error = EINVAL;
1649 						break;
1650 					}
1651 					OPTSET(IN6P_RTHDRDSTOPTS);
1652 					break;
1653 
1654 				case IPV6_RECVRTHDR:
1655 					/* cannot mix with RFC2292 */
1656 					if (OPTBIT(IN6P_RFC2292)) {
1657 						error = EINVAL;
1658 						break;
1659 					}
1660 					OPTSET(IN6P_RTHDR);
1661 					break;
1662 
1663 				case IPV6_FAITH:
1664 					OPTSET(IN6P_FAITH);
1665 					break;
1666 
1667 				case IPV6_RECVPATHMTU:
1668 					/*
1669 					 * We ignore this option for TCP
1670 					 * sockets.
1671 					 * (rfc2292bis leaves this case
1672 					 * unspecified.)
1673 					 */
1674 					if (uproto != IPPROTO_TCP)
1675 						OPTSET(IN6P_MTU);
1676 					break;
1677 
1678 				case IPV6_V6ONLY:
1679 					/*
1680 					 * make setsockopt(IPV6_V6ONLY)
1681 					 * available only prior to bind(2).
1682 					 * see ipng mailing list, Jun 22 2001.
1683 					 */
1684 					if (in6p->in6p_lport ||
1685 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1686 						error = EINVAL;
1687 						break;
1688 					}
1689 					OPTSET(IN6P_IPV6_V6ONLY);
1690 					if (optval)
1691 						in6p->in6p_vflag &= ~INP_IPV4;
1692 					else
1693 						in6p->in6p_vflag |= INP_IPV4;
1694 					break;
1695 				case IPV6_RECVTCLASS:
1696 					/* cannot mix with RFC2292 XXX */
1697 					if (OPTBIT(IN6P_RFC2292)) {
1698 						error = EINVAL;
1699 						break;
1700 					}
1701 					OPTSET(IN6P_TCLASS);
1702 					break;
1703 				case IPV6_AUTOFLOWLABEL:
1704 					OPTSET(IN6P_AUTOFLOWLABEL);
1705 					break;
1706 
1707 				}
1708 				break;
1709 
1710 			case IPV6_TCLASS:
1711 			case IPV6_DONTFRAG:
1712 			case IPV6_USE_MIN_MTU:
1713 			case IPV6_PREFER_TEMPADDR:
1714 				if (optlen != sizeof(optval)) {
1715 					error = EINVAL;
1716 					break;
1717 				}
1718 				error = sooptcopyin(sopt, &optval,
1719 					sizeof optval, sizeof optval);
1720 				if (error)
1721 					break;
1722 				{
1723 					struct ip6_pktopts **optp;
1724 					optp = &in6p->in6p_outputopts;
1725 					error = ip6_pcbopt(optname,
1726 							   (u_char *)&optval,
1727 							   sizeof(optval),
1728 							   optp,
1729 							   privileged, uproto);
1730 					break;
1731 				}
1732 
1733 			case IPV6_2292PKTINFO:
1734 			case IPV6_2292HOPLIMIT:
1735 			case IPV6_2292HOPOPTS:
1736 			case IPV6_2292DSTOPTS:
1737 			case IPV6_2292RTHDR:
1738 				/* RFC 2292 */
1739 				if (optlen != sizeof(int)) {
1740 					error = EINVAL;
1741 					break;
1742 				}
1743 				error = sooptcopyin(sopt, &optval,
1744 					sizeof optval, sizeof optval);
1745 				if (error)
1746 					break;
1747 				switch (optname) {
1748 				case IPV6_2292PKTINFO:
1749 					OPTSET2292(IN6P_PKTINFO);
1750 					break;
1751 				case IPV6_2292HOPLIMIT:
1752 					OPTSET2292(IN6P_HOPLIMIT);
1753 					break;
1754 				case IPV6_2292HOPOPTS:
1755 					/*
1756 					 * Check super-user privilege.
1757 					 * See comments for IPV6_RECVHOPOPTS.
1758 					 */
1759 					if (!privileged)
1760 						return (EPERM);
1761 					OPTSET2292(IN6P_HOPOPTS);
1762 					break;
1763 				case IPV6_2292DSTOPTS:
1764 					if (!privileged)
1765 						return (EPERM);
1766 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1767 					break;
1768 				case IPV6_2292RTHDR:
1769 					OPTSET2292(IN6P_RTHDR);
1770 					break;
1771 				}
1772 				break;
1773 			case IPV6_PKTINFO:
1774 			case IPV6_HOPOPTS:
1775 			case IPV6_RTHDR:
1776 			case IPV6_DSTOPTS:
1777 			case IPV6_RTHDRDSTOPTS:
1778 			case IPV6_NEXTHOP:
1779 			{
1780 				/* new advanced API (2292bis) */
1781 				u_char *optbuf;
1782 				int optlen;
1783 				struct ip6_pktopts **optp;
1784 
1785 				/* cannot mix with RFC2292 */
1786 				if (OPTBIT(IN6P_RFC2292)) {
1787 					error = EINVAL;
1788 					break;
1789 				}
1790 
1791 				switch (optname) {
1792 				case IPV6_HOPOPTS:
1793 				case IPV6_DSTOPTS:
1794 				case IPV6_RTHDRDSTOPTS:
1795 				case IPV6_NEXTHOP:
1796 					if (!privileged)
1797 						error = EPERM;
1798 					break;
1799 				}
1800 				if (error)
1801 					break;
1802 
1803 				switch (optname) {
1804 				case IPV6_PKTINFO:
1805 					optlen = sizeof(struct in6_pktinfo);
1806 					break;
1807 				case IPV6_NEXTHOP:
1808 					optlen = SOCK_MAXADDRLEN;
1809 					break;
1810 				default:
1811 					optlen = IPV6_MAXOPTHDR;
1812 					break;
1813 				}
1814 				if (sopt->sopt_valsize > optlen) {
1815 					error = EINVAL;
1816 					break;
1817 				}
1818 
1819 				optlen = sopt->sopt_valsize;
1820 				optbuf = malloc(optlen, M_TEMP, M_WAITOK);
1821 				error = sooptcopyin(sopt, optbuf, optlen,
1822 				    optlen);
1823 				if (error) {
1824 					free(optbuf, M_TEMP);
1825 					break;
1826 				}
1827 
1828 				optp = &in6p->in6p_outputopts;
1829 				error = ip6_pcbopt(optname,
1830 						   optbuf, optlen,
1831 						   optp, privileged, uproto);
1832 				free(optbuf, M_TEMP);
1833 				break;
1834 			}
1835 #undef OPTSET
1836 
1837 			case IPV6_MULTICAST_IF:
1838 			case IPV6_MULTICAST_HOPS:
1839 			case IPV6_MULTICAST_LOOP:
1840 			case IPV6_JOIN_GROUP:
1841 			case IPV6_LEAVE_GROUP:
1842 			    {
1843 				if (sopt->sopt_valsize > MLEN) {
1844 					error = EMSGSIZE;
1845 					break;
1846 				}
1847 				/* XXX */
1848 			    }
1849 			    /* FALLTHROUGH */
1850 			    {
1851 				struct mbuf *m;
1852 
1853 				if (sopt->sopt_valsize > MCLBYTES) {
1854 					error = EMSGSIZE;
1855 					break;
1856 				}
1857 				/* XXX */
1858 				MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_HEADER);
1859 				if (m == 0) {
1860 					error = ENOBUFS;
1861 					break;
1862 				}
1863 				if (sopt->sopt_valsize > MLEN) {
1864 					MCLGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT);
1865 					if ((m->m_flags & M_EXT) == 0) {
1866 						m_free(m);
1867 						error = ENOBUFS;
1868 						break;
1869 					}
1870 				}
1871 				m->m_len = sopt->sopt_valsize;
1872 				error = sooptcopyin(sopt, mtod(m, char *),
1873 						    m->m_len, m->m_len);
1874 				if (error) {
1875 					(void)m_free(m);
1876 					break;
1877 				}
1878 				error =	ip6_setmoptions(sopt->sopt_name,
1879 							&in6p->in6p_moptions,
1880 							m);
1881 				(void)m_free(m);
1882 			    }
1883 				break;
1884 
1885 			case IPV6_PORTRANGE:
1886 				error = sooptcopyin(sopt, &optval,
1887 				    sizeof optval, sizeof optval);
1888 				if (error)
1889 					break;
1890 
1891 				switch (optval) {
1892 				case IPV6_PORTRANGE_DEFAULT:
1893 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1894 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1895 					break;
1896 
1897 				case IPV6_PORTRANGE_HIGH:
1898 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1899 					in6p->in6p_flags |= IN6P_HIGHPORT;
1900 					break;
1901 
1902 				case IPV6_PORTRANGE_LOW:
1903 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1904 					in6p->in6p_flags |= IN6P_LOWPORT;
1905 					break;
1906 
1907 				default:
1908 					error = EINVAL;
1909 					break;
1910 				}
1911 				break;
1912 
1913 #if defined(IPSEC) || defined(FAST_IPSEC)
1914 			case IPV6_IPSEC_POLICY:
1915 			    {
1916 				caddr_t req = NULL;
1917 				size_t len = 0;
1918 				struct mbuf *m;
1919 
1920 				if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1921 					break;
1922 				if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1923 					break;
1924 				if (m) {
1925 					req = mtod(m, caddr_t);
1926 					len = m->m_len;
1927 				}
1928 				error = ipsec6_set_policy(in6p, optname, req,
1929 							  len, privileged);
1930 				m_freem(m);
1931 			    }
1932 				break;
1933 #endif /* KAME IPSEC */
1934 
1935 			case IPV6_FW_ADD:
1936 			case IPV6_FW_DEL:
1937 			case IPV6_FW_FLUSH:
1938 			case IPV6_FW_ZERO:
1939 			    {
1940 				struct mbuf *m;
1941 				struct mbuf **mp = &m;
1942 
1943 				if (ip6_fw_ctl_ptr == NULL)
1944 					return EINVAL;
1945 				/* XXX */
1946 				if ((error = soopt_getm(sopt, &m)) != 0)
1947 					break;
1948 				/* XXX */
1949 				if ((error = soopt_mcopyin(sopt, m)) != 0)
1950 					break;
1951 				error = (*ip6_fw_ctl_ptr)(optname, mp);
1952 				m = *mp;
1953 			    }
1954 				break;
1955 
1956 			default:
1957 				error = ENOPROTOOPT;
1958 				break;
1959 			}
1960 			break;
1961 
1962 		case SOPT_GET:
1963 			switch (optname) {
1964 
1965 			case IPV6_2292PKTOPTIONS:
1966 #ifdef IPV6_PKTOPTIONS
1967 			case IPV6_PKTOPTIONS:
1968 #endif
1969 				/*
1970 				 * RFC3542 (effectively) deprecated the
1971 				 * semantics of the 2292-style pktoptions.
1972 				 * Since it was not reliable in nature (i.e.,
1973 				 * applications had to expect the lack of some
1974 				 * information after all), it would make sense
1975 				 * to simplify this part by always returning
1976 				 * empty data.
1977 				 */
1978 				sopt->sopt_valsize = 0;
1979 				break;
1980 
1981 			case IPV6_RECVHOPOPTS:
1982 			case IPV6_RECVDSTOPTS:
1983 			case IPV6_RECVRTHDRDSTOPTS:
1984 			case IPV6_UNICAST_HOPS:
1985 			case IPV6_RECVPKTINFO:
1986 			case IPV6_RECVHOPLIMIT:
1987 			case IPV6_RECVRTHDR:
1988 			case IPV6_RECVPATHMTU:
1989 
1990 			case IPV6_FAITH:
1991 			case IPV6_V6ONLY:
1992 			case IPV6_PORTRANGE:
1993 			case IPV6_RECVTCLASS:
1994 			case IPV6_AUTOFLOWLABEL:
1995 				switch (optname) {
1996 
1997 				case IPV6_RECVHOPOPTS:
1998 					optval = OPTBIT(IN6P_HOPOPTS);
1999 					break;
2000 
2001 				case IPV6_RECVDSTOPTS:
2002 					optval = OPTBIT(IN6P_DSTOPTS);
2003 					break;
2004 
2005 				case IPV6_RECVRTHDRDSTOPTS:
2006 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
2007 					break;
2008 
2009 				case IPV6_UNICAST_HOPS:
2010 					optval = in6p->in6p_hops;
2011 					break;
2012 
2013 				case IPV6_RECVPKTINFO:
2014 					optval = OPTBIT(IN6P_PKTINFO);
2015 					break;
2016 
2017 				case IPV6_RECVHOPLIMIT:
2018 					optval = OPTBIT(IN6P_HOPLIMIT);
2019 					break;
2020 
2021 				case IPV6_RECVRTHDR:
2022 					optval = OPTBIT(IN6P_RTHDR);
2023 					break;
2024 
2025 				case IPV6_RECVPATHMTU:
2026 					optval = OPTBIT(IN6P_MTU);
2027 					break;
2028 
2029 				case IPV6_FAITH:
2030 					optval = OPTBIT(IN6P_FAITH);
2031 					break;
2032 
2033 				case IPV6_V6ONLY:
2034 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
2035 					break;
2036 
2037 				case IPV6_PORTRANGE:
2038 				    {
2039 					int flags;
2040 					flags = in6p->in6p_flags;
2041 					if (flags & IN6P_HIGHPORT)
2042 						optval = IPV6_PORTRANGE_HIGH;
2043 					else if (flags & IN6P_LOWPORT)
2044 						optval = IPV6_PORTRANGE_LOW;
2045 					else
2046 						optval = 0;
2047 					break;
2048 				    }
2049 				case IPV6_RECVTCLASS:
2050 					optval = OPTBIT(IN6P_TCLASS);
2051 					break;
2052 
2053 				case IPV6_AUTOFLOWLABEL:
2054 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2055 					break;
2056 				}
2057 				if (error)
2058 					break;
2059 				error = sooptcopyout(sopt, &optval,
2060 					sizeof optval);
2061 				break;
2062 
2063 			case IPV6_PATHMTU:
2064 			{
2065 				u_long pmtu = 0;
2066 				struct ip6_mtuinfo mtuinfo;
2067 				struct route_in6 sro;
2068 
2069 				bzero(&sro, sizeof(sro));
2070 
2071 				if (!(so->so_state & SS_ISCONNECTED))
2072 					return (ENOTCONN);
2073 				/*
2074 				 * XXX: we dot not consider the case of source
2075 				 * routing, or optional information to specify
2076 				 * the outgoing interface.
2077 				 */
2078 				error = ip6_getpmtu(&sro, NULL, NULL,
2079 				    &in6p->in6p_faddr, &pmtu, NULL);
2080 				if (sro.ro_rt)
2081 					RTFREE(sro.ro_rt);
2082 				if (error)
2083 					break;
2084 				if (pmtu > IPV6_MAXPACKET)
2085 					pmtu = IPV6_MAXPACKET;
2086 
2087 				bzero(&mtuinfo, sizeof(mtuinfo));
2088 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2089 				optdata = (void *)&mtuinfo;
2090 				optdatalen = sizeof(mtuinfo);
2091 				error = sooptcopyout(sopt, optdata,
2092 				    optdatalen);
2093 				break;
2094 			}
2095 
2096 			case IPV6_2292PKTINFO:
2097 			case IPV6_2292HOPLIMIT:
2098 			case IPV6_2292HOPOPTS:
2099 			case IPV6_2292RTHDR:
2100 			case IPV6_2292DSTOPTS:
2101 				switch (optname) {
2102 				case IPV6_2292PKTINFO:
2103 					optval = OPTBIT(IN6P_PKTINFO);
2104 					break;
2105 				case IPV6_2292HOPLIMIT:
2106 					optval = OPTBIT(IN6P_HOPLIMIT);
2107 					break;
2108 				case IPV6_2292HOPOPTS:
2109 					optval = OPTBIT(IN6P_HOPOPTS);
2110 					break;
2111 				case IPV6_2292RTHDR:
2112 					optval = OPTBIT(IN6P_RTHDR);
2113 					break;
2114 				case IPV6_2292DSTOPTS:
2115 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2116 					break;
2117 				}
2118 				error = sooptcopyout(sopt, &optval,
2119 				    sizeof optval);
2120 				break;
2121 			case IPV6_PKTINFO:
2122 			case IPV6_HOPOPTS:
2123 			case IPV6_RTHDR:
2124 			case IPV6_DSTOPTS:
2125 			case IPV6_RTHDRDSTOPTS:
2126 			case IPV6_NEXTHOP:
2127 			case IPV6_TCLASS:
2128 			case IPV6_DONTFRAG:
2129 			case IPV6_USE_MIN_MTU:
2130 			case IPV6_PREFER_TEMPADDR:
2131 				error = ip6_getpcbopt(in6p->in6p_outputopts,
2132 				    optname, sopt);
2133 				break;
2134 
2135 			case IPV6_MULTICAST_IF:
2136 			case IPV6_MULTICAST_HOPS:
2137 			case IPV6_MULTICAST_LOOP:
2138 			case IPV6_JOIN_GROUP:
2139 			case IPV6_LEAVE_GROUP:
2140 			    {
2141 				struct mbuf *m;
2142 				error = ip6_getmoptions(sopt->sopt_name,
2143 				    in6p->in6p_moptions, &m);
2144 				if (error == 0)
2145 					error = sooptcopyout(sopt,
2146 					    mtod(m, char *), m->m_len);
2147 				m_freem(m);
2148 			    }
2149 				break;
2150 
2151 #if defined(IPSEC) || defined(FAST_IPSEC)
2152 			case IPV6_IPSEC_POLICY:
2153 			  {
2154 				caddr_t req = NULL;
2155 				size_t len = 0;
2156 				struct mbuf *m = NULL;
2157 				struct mbuf **mp = &m;
2158 				size_t ovalsize = sopt->sopt_valsize;
2159 				caddr_t oval = (caddr_t)sopt->sopt_val;
2160 
2161 				error = soopt_getm(sopt, &m); /* XXX */
2162 				if (error != 0)
2163 					break;
2164 				error = soopt_mcopyin(sopt, m); /* XXX */
2165 				if (error != 0)
2166 					break;
2167 				sopt->sopt_valsize = ovalsize;
2168 				sopt->sopt_val = oval;
2169 				if (m) {
2170 					req = mtod(m, caddr_t);
2171 					len = m->m_len;
2172 				}
2173 				error = ipsec6_get_policy(in6p, req, len, mp);
2174 				if (error == 0)
2175 					error = soopt_mcopyout(sopt, m); /* XXX */
2176 				if (error == 0 && m)
2177 					m_freem(m);
2178 				break;
2179 			  }
2180 #endif /* KAME IPSEC */
2181 
2182 			case IPV6_FW_GET:
2183 			  {
2184 				struct mbuf *m;
2185 				struct mbuf **mp = &m;
2186 
2187 				if (ip6_fw_ctl_ptr == NULL)
2188 			        {
2189 					return EINVAL;
2190 				}
2191 				error = (*ip6_fw_ctl_ptr)(optname, mp);
2192 				if (error == 0)
2193 					error = soopt_mcopyout(sopt, m); /* XXX */
2194 				if (error == 0 && m)
2195 					m_freem(m);
2196 			  }
2197 				break;
2198 
2199 			default:
2200 				error = ENOPROTOOPT;
2201 				break;
2202 			}
2203 			break;
2204 		}
2205 	} else {		/* level != IPPROTO_IPV6 */
2206 		error = EINVAL;
2207 	}
2208 	return (error);
2209 }
2210 
2211 int
2212 ip6_raw_ctloutput(so, sopt)
2213 	struct socket *so;
2214 	struct sockopt *sopt;
2215 {
2216 	int error = 0, optval, optlen;
2217 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2218 	struct in6pcb *in6p = sotoin6pcb(so);
2219 	int level, op, optname;
2220 
2221 	if (sopt) {
2222 		level = sopt->sopt_level;
2223 		op = sopt->sopt_dir;
2224 		optname = sopt->sopt_name;
2225 		optlen = sopt->sopt_valsize;
2226 	} else
2227 		panic("ip6_raw_ctloutput: arg soopt is NULL");
2228 
2229 	if (level != IPPROTO_IPV6) {
2230 		return (EINVAL);
2231 	}
2232 
2233 	switch (optname) {
2234 	case IPV6_CHECKSUM:
2235 		/*
2236 		 * For ICMPv6 sockets, no modification allowed for checksum
2237 		 * offset, permit "no change" values to help existing apps.
2238 		 *
2239 		 * XXX 2292bis says: "An attempt to set IPV6_CHECKSUM
2240 		 * for an ICMPv6 socket will fail."
2241 		 * The current behavior does not meet 2292bis.
2242 		 */
2243 		switch (op) {
2244 		case SOPT_SET:
2245 			if (optlen != sizeof(int)) {
2246 				error = EINVAL;
2247 				break;
2248 			}
2249 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2250 					    sizeof(optval));
2251 			if (error)
2252 				break;
2253 			if ((optval % 2) != 0) {
2254 				/* the API assumes even offset values */
2255 				error = EINVAL;
2256 			} else if (so->so_proto->pr_protocol ==
2257 			    IPPROTO_ICMPV6) {
2258 				if (optval != icmp6off)
2259 					error = EINVAL;
2260 			} else
2261 				in6p->in6p_cksum = optval;
2262 			break;
2263 
2264 		case SOPT_GET:
2265 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2266 				optval = icmp6off;
2267 			else
2268 				optval = in6p->in6p_cksum;
2269 
2270 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2271 			break;
2272 
2273 		default:
2274 			error = EINVAL;
2275 			break;
2276 		}
2277 		break;
2278 
2279 	default:
2280 		error = ENOPROTOOPT;
2281 		break;
2282 	}
2283 
2284 	return (error);
2285 }
2286 
2287 /*
2288  * Set up IP6 options in pcb for insertion in output packets or
2289  * specifying behavior of outgoing packets.
2290  */
2291 static int
2292 ip6_pcbopts(pktopt, m, so, sopt)
2293 	struct ip6_pktopts **pktopt;
2294 	struct mbuf *m;
2295 	struct socket *so;
2296 	struct sockopt *sopt;
2297 {
2298 	struct ip6_pktopts *opt = *pktopt;
2299 	int error = 0;
2300 	struct thread *td = sopt->sopt_td;
2301 	int priv = 0;
2302 
2303 	/* turn off any old options. */
2304 	if (opt) {
2305 #ifdef DIAGNOSTIC
2306 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2307 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2308 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2309 			printf("ip6_pcbopts: all specified options are cleared.\n");
2310 #endif
2311 		ip6_clearpktopts(opt, -1);
2312 	} else
2313 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2314 	*pktopt = NULL;
2315 
2316 	if (!m || m->m_len == 0) {
2317 		/*
2318 		 * Only turning off any previous options, regardless of
2319 		 * whether the opt is just created or given.
2320 		 */
2321 		free(opt, M_IP6OPT);
2322 		return (0);
2323 	}
2324 
2325 	/*  set options specified by user. */
2326 	if (td && !suser(td))
2327 		priv = 1;
2328 	if ((error = ip6_setpktoptions(m, opt, NULL, priv, 1,
2329 	    so->so_proto->pr_protocol)) != 0) {
2330 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2331 		free(opt, M_IP6OPT);
2332 		return (error);
2333 	}
2334 	*pktopt = opt;
2335 	return (0);
2336 }
2337 
2338 /*
2339  * initialize ip6_pktopts.  beware that there are non-zero default values in
2340  * the struct.
2341  */
2342 void
2343 init_ip6pktopts(opt)
2344 	struct ip6_pktopts *opt;
2345 {
2346 
2347 	bzero(opt, sizeof(*opt));
2348 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2349 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2350 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2351 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2352 }
2353 
2354 static int
2355 ip6_pcbopt(optname, buf, len, pktopt, priv, uproto)
2356 	int optname, len, priv;
2357 	u_char *buf;
2358 	struct ip6_pktopts **pktopt;
2359 	int uproto;
2360 {
2361 	struct ip6_pktopts *opt;
2362 
2363 	if (*pktopt == NULL) {
2364 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2365 		    M_WAITOK);
2366 		init_ip6pktopts(*pktopt);
2367 		(*pktopt)->needfree = 1;
2368 	}
2369 	opt = *pktopt;
2370 
2371 	return (ip6_setpktoption(optname, buf, len, opt, priv, 1, 0, uproto));
2372 }
2373 
2374 static int
2375 ip6_getpcbopt(pktopt, optname, sopt)
2376 	struct ip6_pktopts *pktopt;
2377 	struct sockopt *sopt;
2378 	int optname;
2379 {
2380 	void *optdata = NULL;
2381 	int optdatalen = 0;
2382 	struct ip6_ext *ip6e;
2383 	int error = 0;
2384 	struct in6_pktinfo null_pktinfo;
2385 	int deftclass = 0, on;
2386 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2387 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2388 
2389 	switch (optname) {
2390 	case IPV6_PKTINFO:
2391 		if (pktopt && pktopt->ip6po_pktinfo)
2392 			optdata = (void *)pktopt->ip6po_pktinfo;
2393 		else {
2394 			/* XXX: we don't have to do this every time... */
2395 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2396 			optdata = (void *)&null_pktinfo;
2397 		}
2398 		optdatalen = sizeof(struct in6_pktinfo);
2399 		break;
2400 	case IPV6_TCLASS:
2401 		if (pktopt && pktopt->ip6po_tclass >= 0)
2402 			optdata = (void *)&pktopt->ip6po_tclass;
2403 		else
2404 			optdata = (void *)&deftclass;
2405 		optdatalen = sizeof(int);
2406 		break;
2407 	case IPV6_HOPOPTS:
2408 		if (pktopt && pktopt->ip6po_hbh) {
2409 			optdata = (void *)pktopt->ip6po_hbh;
2410 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2411 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2412 		}
2413 		break;
2414 	case IPV6_RTHDR:
2415 		if (pktopt && pktopt->ip6po_rthdr) {
2416 			optdata = (void *)pktopt->ip6po_rthdr;
2417 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2418 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2419 		}
2420 		break;
2421 	case IPV6_RTHDRDSTOPTS:
2422 		if (pktopt && pktopt->ip6po_dest1) {
2423 			optdata = (void *)pktopt->ip6po_dest1;
2424 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2425 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2426 		}
2427 		break;
2428 	case IPV6_DSTOPTS:
2429 		if (pktopt && pktopt->ip6po_dest2) {
2430 			optdata = (void *)pktopt->ip6po_dest2;
2431 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2432 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2433 		}
2434 		break;
2435 	case IPV6_NEXTHOP:
2436 		if (pktopt && pktopt->ip6po_nexthop) {
2437 			optdata = (void *)pktopt->ip6po_nexthop;
2438 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2439 		}
2440 		break;
2441 	case IPV6_USE_MIN_MTU:
2442 		if (pktopt)
2443 			optdata = (void *)&pktopt->ip6po_minmtu;
2444 		else
2445 			optdata = (void *)&defminmtu;
2446 		optdatalen = sizeof(int);
2447 		break;
2448 	case IPV6_DONTFRAG:
2449 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2450 			on = 1;
2451 		else
2452 			on = 0;
2453 		optdata = (void *)&on;
2454 		optdatalen = sizeof(on);
2455 		break;
2456 	case IPV6_PREFER_TEMPADDR:
2457 		if (pktopt)
2458 			optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2459 		else
2460 			optdata = (void *)&defpreftemp;
2461 		optdatalen = sizeof(int);
2462 		break;
2463 	default:		/* should not happen */
2464 #ifdef DIAGNOSTIC
2465 		panic("ip6_getpcbopt: unexpected option\n");
2466 #endif
2467 		return (ENOPROTOOPT);
2468 	}
2469 
2470 	error = sooptcopyout(sopt, optdata, optdatalen);
2471 
2472 	return (error);
2473 }
2474 
2475 void
2476 ip6_clearpktopts(pktopt, optname)
2477 	struct ip6_pktopts *pktopt;
2478 	int optname;
2479 {
2480 	int needfree;
2481 
2482 	if (pktopt == NULL)
2483 		return;
2484 
2485 	needfree = pktopt->needfree;
2486 
2487 	if (optname == -1 || optname == IPV6_PKTINFO) {
2488 		if (needfree && pktopt->ip6po_pktinfo)
2489 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2490 		pktopt->ip6po_pktinfo = NULL;
2491 	}
2492 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2493 		pktopt->ip6po_hlim = -1;
2494 	if (optname == -1 || optname == IPV6_TCLASS)
2495 		pktopt->ip6po_tclass = -1;
2496 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2497 		if (pktopt->ip6po_nextroute.ro_rt) {
2498 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2499 			pktopt->ip6po_nextroute.ro_rt = NULL;
2500 		}
2501 		if (needfree && pktopt->ip6po_nexthop)
2502 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2503 		pktopt->ip6po_nexthop = NULL;
2504 	}
2505 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2506 		if (needfree && pktopt->ip6po_hbh)
2507 			free(pktopt->ip6po_hbh, M_IP6OPT);
2508 		pktopt->ip6po_hbh = NULL;
2509 	}
2510 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2511 		if (needfree && pktopt->ip6po_dest1)
2512 			free(pktopt->ip6po_dest1, M_IP6OPT);
2513 		pktopt->ip6po_dest1 = NULL;
2514 	}
2515 	if (optname == -1 || optname == IPV6_RTHDR) {
2516 		if (needfree && pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2517 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2518 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2519 		if (pktopt->ip6po_route.ro_rt) {
2520 			RTFREE(pktopt->ip6po_route.ro_rt);
2521 			pktopt->ip6po_route.ro_rt = NULL;
2522 		}
2523 	}
2524 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2525 		if (needfree && pktopt->ip6po_dest2)
2526 			free(pktopt->ip6po_dest2, M_IP6OPT);
2527 		pktopt->ip6po_dest2 = NULL;
2528 	}
2529 }
2530 
2531 #define PKTOPT_EXTHDRCPY(type) \
2532 do {\
2533 	if (src->type) {\
2534 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2535 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2536 		if (dst->type == NULL && canwait == M_NOWAIT)\
2537 			goto bad;\
2538 		bcopy(src->type, dst->type, hlen);\
2539 	}\
2540 } while (/*CONSTCOND*/ 0)
2541 
2542 struct ip6_pktopts *
2543 ip6_copypktopts(src, canwait)
2544 	struct ip6_pktopts *src;
2545 	int canwait;
2546 {
2547 	struct ip6_pktopts *dst;
2548 
2549 	if (src == NULL) {
2550 		printf("ip6_clearpktopts: invalid argument\n");
2551 		return (NULL);
2552 	}
2553 
2554 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2555 	if (dst == NULL && canwait == M_NOWAIT)
2556 		return (NULL);
2557 	bzero(dst, sizeof(*dst));
2558 	dst->needfree = 1;
2559 
2560 	dst->ip6po_hlim = src->ip6po_hlim;
2561 	dst->ip6po_tclass = src->ip6po_tclass;
2562 	dst->ip6po_flags = src->ip6po_flags;
2563 	if (src->ip6po_pktinfo) {
2564 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2565 		    M_IP6OPT, canwait);
2566 		if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT)
2567 			goto bad;
2568 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2569 	}
2570 	if (src->ip6po_nexthop) {
2571 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2572 		    M_IP6OPT, canwait);
2573 		if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT)
2574 			goto bad;
2575 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2576 		    src->ip6po_nexthop->sa_len);
2577 	}
2578 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2579 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2580 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2581 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2582 	return (dst);
2583 
2584   bad:
2585 	if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
2586 	if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
2587 	if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
2588 	if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
2589 	if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
2590 	if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
2591 	free(dst, M_IP6OPT);
2592 	return (NULL);
2593 }
2594 #undef PKTOPT_EXTHDRCPY
2595 
2596 void
2597 ip6_freepcbopts(pktopt)
2598 	struct ip6_pktopts *pktopt;
2599 {
2600 	if (pktopt == NULL)
2601 		return;
2602 
2603 	ip6_clearpktopts(pktopt, -1);
2604 
2605 	free(pktopt, M_IP6OPT);
2606 }
2607 
2608 /*
2609  * Set the IP6 multicast options in response to user setsockopt().
2610  */
2611 static int
2612 ip6_setmoptions(optname, im6op, m)
2613 	int optname;
2614 	struct ip6_moptions **im6op;
2615 	struct mbuf *m;
2616 {
2617 	int error = 0;
2618 	u_int loop, ifindex;
2619 	struct ipv6_mreq *mreq;
2620 	struct ifnet *ifp;
2621 	struct ip6_moptions *im6o = *im6op;
2622 	struct route_in6 ro;
2623 	struct sockaddr_in6 *dst;
2624 	struct in6_multi_mship *imm;
2625 	struct thread *td = curthread;
2626 
2627 	if (im6o == NULL) {
2628 		/*
2629 		 * No multicast option buffer attached to the pcb;
2630 		 * allocate one and initialize to default values.
2631 		 */
2632 		im6o = (struct ip6_moptions *)
2633 			malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
2634 
2635 		if (im6o == NULL)
2636 			return (ENOBUFS);
2637 		*im6op = im6o;
2638 		im6o->im6o_multicast_ifp = NULL;
2639 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2640 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2641 		LIST_INIT(&im6o->im6o_memberships);
2642 	}
2643 
2644 	switch (optname) {
2645 
2646 	case IPV6_MULTICAST_IF:
2647 		/*
2648 		 * Select the interface for outgoing multicast packets.
2649 		 */
2650 		if (m == NULL || m->m_len != sizeof(u_int)) {
2651 			error = EINVAL;
2652 			break;
2653 		}
2654 		bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
2655 		if (ifindex < 0 || if_index < ifindex) {
2656 			error = ENXIO;	/* XXX EINVAL? */
2657 			break;
2658 		}
2659 		ifp = ifnet_byindex(ifindex);
2660 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2661 			error = EADDRNOTAVAIL;
2662 			break;
2663 		}
2664 		im6o->im6o_multicast_ifp = ifp;
2665 		break;
2666 
2667 	case IPV6_MULTICAST_HOPS:
2668 	    {
2669 		/*
2670 		 * Set the IP6 hoplimit for outgoing multicast packets.
2671 		 */
2672 		int optval;
2673 		if (m == NULL || m->m_len != sizeof(int)) {
2674 			error = EINVAL;
2675 			break;
2676 		}
2677 		bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2678 		if (optval < -1 || optval >= 256)
2679 			error = EINVAL;
2680 		else if (optval == -1)
2681 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2682 		else
2683 			im6o->im6o_multicast_hlim = optval;
2684 		break;
2685 	    }
2686 
2687 	case IPV6_MULTICAST_LOOP:
2688 		/*
2689 		 * Set the loopback flag for outgoing multicast packets.
2690 		 * Must be zero or one.
2691 		 */
2692 		if (m == NULL || m->m_len != sizeof(u_int)) {
2693 			error = EINVAL;
2694 			break;
2695 		}
2696 		bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2697 		if (loop > 1) {
2698 			error = EINVAL;
2699 			break;
2700 		}
2701 		im6o->im6o_multicast_loop = loop;
2702 		break;
2703 
2704 	case IPV6_JOIN_GROUP:
2705 		/*
2706 		 * Add a multicast group membership.
2707 		 * Group must be a valid IP6 multicast address.
2708 		 */
2709 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2710 			error = EINVAL;
2711 			break;
2712 		}
2713 		mreq = mtod(m, struct ipv6_mreq *);
2714 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2715 			/*
2716 			 * We use the unspecified address to specify to accept
2717 			 * all multicast addresses. Only super user is allowed
2718 			 * to do this.
2719 			 */
2720 			if (suser(td)) {
2721 				error = EACCES;
2722 				break;
2723 			}
2724 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2725 			error = EINVAL;
2726 			break;
2727 		}
2728 
2729 		/*
2730 		 * If the interface is specified, validate it.
2731 		 */
2732 		if (mreq->ipv6mr_interface < 0 ||
2733 		    if_index < mreq->ipv6mr_interface) {
2734 			error = ENXIO;	/* XXX EINVAL? */
2735 			break;
2736 		}
2737 		/*
2738 		 * If no interface was explicitly specified, choose an
2739 		 * appropriate one according to the given multicast address.
2740 		 */
2741 		if (mreq->ipv6mr_interface == 0) {
2742 			/*
2743 			 * If the multicast address is in node-local scope,
2744 			 * the interface should be a loopback interface.
2745 			 * Otherwise, look up the routing table for the
2746 			 * address, and choose the outgoing interface.
2747 			 *   XXX: is it a good approach?
2748 			 */
2749 			if (IN6_IS_ADDR_MC_INTFACELOCAL(&mreq->ipv6mr_multiaddr)) {
2750 				ifp = &loif[0];
2751 			} else {
2752 				ro.ro_rt = NULL;
2753 				dst = (struct sockaddr_in6 *)&ro.ro_dst;
2754 				bzero(dst, sizeof(*dst));
2755 				dst->sin6_len = sizeof(struct sockaddr_in6);
2756 				dst->sin6_family = AF_INET6;
2757 				dst->sin6_addr = mreq->ipv6mr_multiaddr;
2758 				rtalloc((struct route *)&ro);
2759 				if (ro.ro_rt == NULL) {
2760 					error = EADDRNOTAVAIL;
2761 					break;
2762 				}
2763 				ifp = ro.ro_rt->rt_ifp;
2764 				RTFREE(ro.ro_rt);
2765 			}
2766 		} else
2767 			ifp = ifnet_byindex(mreq->ipv6mr_interface);
2768 
2769 		/*
2770 		 * See if we found an interface, and confirm that it
2771 		 * supports multicast
2772 		 */
2773 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2774 			error = EADDRNOTAVAIL;
2775 			break;
2776 		}
2777 		/*
2778 		 * Put interface index into the multicast address,
2779 		 * if the address has link-local scope.
2780 		 */
2781 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2782 			mreq->ipv6mr_multiaddr.s6_addr16[1] =
2783 			    htons(ifp->if_index);
2784 		}
2785 		/*
2786 		 * See if the membership already exists.
2787 		 */
2788 		for (imm = im6o->im6o_memberships.lh_first;
2789 		     imm != NULL; imm = imm->i6mm_chain.le_next)
2790 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
2791 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2792 					       &mreq->ipv6mr_multiaddr))
2793 				break;
2794 		if (imm != NULL) {
2795 			error = EADDRINUSE;
2796 			break;
2797 		}
2798 		/*
2799 		 * Everything looks good; add a new record to the multicast
2800 		 * address list for the given interface.
2801 		 */
2802 		imm = malloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
2803 		if (imm == NULL) {
2804 			error = ENOBUFS;
2805 			break;
2806 		}
2807 		if ((imm->i6mm_maddr =
2808 		     in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
2809 			free(imm, M_IPMADDR);
2810 			break;
2811 		}
2812 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2813 		break;
2814 
2815 	case IPV6_LEAVE_GROUP:
2816 		/*
2817 		 * Drop a multicast group membership.
2818 		 * Group must be a valid IP6 multicast address.
2819 		 */
2820 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2821 			error = EINVAL;
2822 			break;
2823 		}
2824 		mreq = mtod(m, struct ipv6_mreq *);
2825 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2826 			if (suser(td)) {
2827 				error = EACCES;
2828 				break;
2829 			}
2830 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2831 			error = EINVAL;
2832 			break;
2833 		}
2834 		/*
2835 		 * If an interface address was specified, get a pointer
2836 		 * to its ifnet structure.
2837 		 */
2838 		if (mreq->ipv6mr_interface < 0
2839 		 || if_index < mreq->ipv6mr_interface) {
2840 			error = ENXIO;	/* XXX EINVAL? */
2841 			break;
2842 		}
2843 		ifp = ifnet_byindex(mreq->ipv6mr_interface);
2844 		/*
2845 		 * Put interface index into the multicast address,
2846 		 * if the address has link-local scope.
2847 		 */
2848 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2849 			mreq->ipv6mr_multiaddr.s6_addr16[1]
2850 				= htons(mreq->ipv6mr_interface);
2851 		}
2852 
2853 		/*
2854 		 * Find the membership in the membership list.
2855 		 */
2856 		for (imm = im6o->im6o_memberships.lh_first;
2857 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
2858 			if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2859 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2860 			    &mreq->ipv6mr_multiaddr))
2861 				break;
2862 		}
2863 		if (imm == NULL) {
2864 			/* Unable to resolve interface */
2865 			error = EADDRNOTAVAIL;
2866 			break;
2867 		}
2868 		/*
2869 		 * Give up the multicast address record to which the
2870 		 * membership points.
2871 		 */
2872 		LIST_REMOVE(imm, i6mm_chain);
2873 		in6_delmulti(imm->i6mm_maddr);
2874 		free(imm, M_IPMADDR);
2875 		break;
2876 
2877 	default:
2878 		error = EOPNOTSUPP;
2879 		break;
2880 	}
2881 
2882 	/*
2883 	 * If all options have default values, no need to keep the mbuf.
2884 	 */
2885 	if (im6o->im6o_multicast_ifp == NULL &&
2886 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2887 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2888 	    im6o->im6o_memberships.lh_first == NULL) {
2889 		free(*im6op, M_IPMOPTS);
2890 		*im6op = NULL;
2891 	}
2892 
2893 	return (error);
2894 }
2895 
2896 /*
2897  * Return the IP6 multicast options in response to user getsockopt().
2898  */
2899 static int
2900 ip6_getmoptions(optname, im6o, mp)
2901 	int optname;
2902 	struct ip6_moptions *im6o;
2903 	struct mbuf **mp;
2904 {
2905 	u_int *hlim, *loop, *ifindex;
2906 
2907 	*mp = m_get(M_TRYWAIT, MT_HEADER);		/* XXX */
2908 
2909 	switch (optname) {
2910 
2911 	case IPV6_MULTICAST_IF:
2912 		ifindex = mtod(*mp, u_int *);
2913 		(*mp)->m_len = sizeof(u_int);
2914 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2915 			*ifindex = 0;
2916 		else
2917 			*ifindex = im6o->im6o_multicast_ifp->if_index;
2918 		return (0);
2919 
2920 	case IPV6_MULTICAST_HOPS:
2921 		hlim = mtod(*mp, u_int *);
2922 		(*mp)->m_len = sizeof(u_int);
2923 		if (im6o == NULL)
2924 			*hlim = ip6_defmcasthlim;
2925 		else
2926 			*hlim = im6o->im6o_multicast_hlim;
2927 		return (0);
2928 
2929 	case IPV6_MULTICAST_LOOP:
2930 		loop = mtod(*mp, u_int *);
2931 		(*mp)->m_len = sizeof(u_int);
2932 		if (im6o == NULL)
2933 			*loop = ip6_defmcasthlim;
2934 		else
2935 			*loop = im6o->im6o_multicast_loop;
2936 		return (0);
2937 
2938 	default:
2939 		return (EOPNOTSUPP);
2940 	}
2941 }
2942 
2943 /*
2944  * Discard the IP6 multicast options.
2945  */
2946 void
2947 ip6_freemoptions(im6o)
2948 	struct ip6_moptions *im6o;
2949 {
2950 	struct in6_multi_mship *imm;
2951 
2952 	if (im6o == NULL)
2953 		return;
2954 
2955 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2956 		LIST_REMOVE(imm, i6mm_chain);
2957 		if (imm->i6mm_maddr)
2958 			in6_delmulti(imm->i6mm_maddr);
2959 		free(imm, M_IPMADDR);
2960 	}
2961 	free(im6o, M_IPMOPTS);
2962 }
2963 
2964 /*
2965  * Set IPv6 outgoing packet options based on advanced API.
2966  */
2967 int
2968 ip6_setpktoptions(control, opt, stickyopt, priv, needcopy, uproto)
2969 	struct mbuf *control;
2970 	struct ip6_pktopts *opt, *stickyopt;
2971 	int priv, needcopy, uproto;
2972 {
2973 	struct cmsghdr *cm = 0;
2974 
2975 	if (control == 0 || opt == 0)
2976 		return (EINVAL);
2977 
2978 	if (stickyopt) {
2979 		/*
2980 		 * If stickyopt is provided, make a local copy of the options
2981 		 * for this particular packet, then override them by ancillary
2982 		 * objects.
2983 		 * XXX: need to gain a reference for the cached route of the
2984 		 * next hop in case of the overriding.
2985 		 */
2986 		*opt = *stickyopt;
2987 		if (opt->ip6po_nextroute.ro_rt) {
2988 			RT_LOCK(opt->ip6po_nextroute.ro_rt);
2989 			RT_ADDREF(opt->ip6po_nextroute.ro_rt);
2990 			RT_UNLOCK(opt->ip6po_nextroute.ro_rt);
2991 		}
2992 	} else
2993 		init_ip6pktopts(opt);
2994 	opt->needfree = needcopy;
2995 
2996 	/*
2997 	 * XXX: Currently, we assume all the optional information is stored
2998 	 * in a single mbuf.
2999 	 */
3000 	if (control->m_next)
3001 		return (EINVAL);
3002 
3003 	for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
3004 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
3005 		int error;
3006 
3007 		if (control->m_len < CMSG_LEN(0))
3008 			return (EINVAL);
3009 
3010 		cm = mtod(control, struct cmsghdr *);
3011 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
3012 			return (EINVAL);
3013 		if (cm->cmsg_level != IPPROTO_IPV6)
3014 			continue;
3015 
3016 		error = ip6_setpktoption(cm->cmsg_type, CMSG_DATA(cm),
3017 		    cm->cmsg_len - CMSG_LEN(0), opt, priv, needcopy, 1, uproto);
3018 		if (error)
3019 			return (error);
3020 	}
3021 
3022 	return (0);
3023 }
3024 
3025 /*
3026  * Set a particular packet option, as a sticky option or an ancillary data
3027  * item.  "len" can be 0 only when it's a sticky option.
3028  * We have 4 cases of combination of "sticky" and "cmsg":
3029  * "sticky=0, cmsg=0": impossible
3030  * "sticky=0, cmsg=1": RFC2292 or rfc2292bis ancillary data
3031  * "sticky=1, cmsg=0": rfc2292bis socket option
3032  * "sticky=1, cmsg=1": RFC2292 socket option
3033  */
3034 static int
3035 ip6_setpktoption(optname, buf, len, opt, priv, sticky, cmsg, uproto)
3036 	int optname, len, priv, sticky, cmsg, uproto;
3037 	u_char *buf;
3038 	struct ip6_pktopts *opt;
3039 {
3040 	int minmtupolicy, preftemp;
3041 
3042 	if (!sticky && !cmsg) {
3043 #ifdef DIAGNOSTIC
3044 		printf("ip6_setpktoption: impossible case\n");
3045 #endif
3046 		return (EINVAL);
3047 	}
3048 
3049 	/*
3050 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
3051 	 * not be specified in the context of rfc2292bis.  Conversely,
3052 	 * rfc2292bis types should not be specified in the context of RFC2292.
3053 	 */
3054 	if (!cmsg) {
3055 		switch (optname) {
3056 		case IPV6_2292PKTINFO:
3057 		case IPV6_2292HOPLIMIT:
3058 		case IPV6_2292NEXTHOP:
3059 		case IPV6_2292HOPOPTS:
3060 		case IPV6_2292DSTOPTS:
3061 		case IPV6_2292RTHDR:
3062 		case IPV6_2292PKTOPTIONS:
3063 			return (ENOPROTOOPT);
3064 		}
3065 	}
3066 	if (sticky && cmsg) {
3067 		switch (optname) {
3068 		case IPV6_PKTINFO:
3069 		case IPV6_HOPLIMIT:
3070 		case IPV6_NEXTHOP:
3071 		case IPV6_HOPOPTS:
3072 		case IPV6_DSTOPTS:
3073 		case IPV6_RTHDRDSTOPTS:
3074 		case IPV6_RTHDR:
3075 		case IPV6_USE_MIN_MTU:
3076 		case IPV6_DONTFRAG:
3077 		case IPV6_TCLASS:
3078 		case IPV6_PREFER_TEMPADDR: /* XXX: not an rfc2292bis option */
3079 			return (ENOPROTOOPT);
3080 		}
3081 	}
3082 
3083 	switch (optname) {
3084 	case IPV6_2292PKTINFO:
3085 	case IPV6_PKTINFO:
3086 	{
3087 		struct ifnet *ifp = NULL;
3088 		struct in6_pktinfo *pktinfo;
3089 
3090 		if (len != sizeof(struct in6_pktinfo))
3091 			return (EINVAL);
3092 
3093 		pktinfo = (struct in6_pktinfo *)buf;
3094 
3095 		/*
3096 		 * An application can clear any sticky IPV6_PKTINFO option by
3097 		 * doing a "regular" setsockopt with ipi6_addr being
3098 		 * in6addr_any and ipi6_ifindex being zero.
3099 		 * [RFC 3542, Section 6]
3100 		 */
3101 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
3102 		    pktinfo->ipi6_ifindex == 0 &&
3103 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
3104 			ip6_clearpktopts(opt, optname);
3105 			break;
3106 		}
3107 
3108 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
3109 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
3110 			return (EINVAL);
3111 		}
3112 
3113 		/* validate the interface index if specified. */
3114 		if (pktinfo->ipi6_ifindex > if_index ||
3115 		    pktinfo->ipi6_ifindex < 0) {
3116 			 return (ENXIO);
3117 		}
3118 		if (pktinfo->ipi6_ifindex) {
3119 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
3120 			if (ifp == NULL)
3121 				return (ENXIO);
3122 		}
3123 
3124 		/*
3125 		 * We store the address anyway, and let in6_selectsrc()
3126 		 * validate the specified address.  This is because ipi6_addr
3127 		 * may not have enough information about its scope zone, and
3128 		 * we may need additional information (such as outgoing
3129 		 * interface or the scope zone of a destination address) to
3130 		 * disambiguate the scope.
3131 		 * XXX: the delay of the validation may confuse the
3132 		 * application when it is used as a sticky option.
3133 		 */
3134 		if (sticky) {
3135 			if (opt->ip6po_pktinfo == NULL) {
3136 				opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
3137 				    M_IP6OPT, M_WAITOK);
3138 			}
3139 			bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
3140 		} else
3141 			opt->ip6po_pktinfo = pktinfo;
3142 		break;
3143 	}
3144 
3145 	case IPV6_2292HOPLIMIT:
3146 	case IPV6_HOPLIMIT:
3147 	{
3148 		int *hlimp;
3149 
3150 		/*
3151 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
3152 		 * to simplify the ordering among hoplimit options.
3153 		 */
3154 		if (optname == IPV6_HOPLIMIT && sticky)
3155 			return (ENOPROTOOPT);
3156 
3157 		if (len != sizeof(int))
3158 			return (EINVAL);
3159 		hlimp = (int *)buf;
3160 		if (*hlimp < -1 || *hlimp > 255)
3161 			return (EINVAL);
3162 
3163 		opt->ip6po_hlim = *hlimp;
3164 		break;
3165 	}
3166 
3167 	case IPV6_TCLASS:
3168 	{
3169 		int tclass;
3170 
3171 		if (len != sizeof(int))
3172 			return (EINVAL);
3173 		tclass = *(int *)buf;
3174 		if (tclass < -1 || tclass > 255)
3175 			return (EINVAL);
3176 
3177 		opt->ip6po_tclass = tclass;
3178 		break;
3179 	}
3180 
3181 	case IPV6_2292NEXTHOP:
3182 	case IPV6_NEXTHOP:
3183 		if (!priv)
3184 			return (EPERM);
3185 
3186 		if (len == 0) {	/* just remove the option */
3187 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
3188 			break;
3189 		}
3190 
3191 		/* check if cmsg_len is large enough for sa_len */
3192 		if (len < sizeof(struct sockaddr) || len < *buf)
3193 			return (EINVAL);
3194 
3195 		switch (((struct sockaddr *)buf)->sa_family) {
3196 		case AF_INET6:
3197 		{
3198 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3199 #if 0
3200 			int error;
3201 #endif
3202 
3203 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3204 				return (EINVAL);
3205 
3206 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3207 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3208 				return (EINVAL);
3209 			}
3210 #if 0
3211 			if ((error = scope6_check_id(sa6, ip6_use_defzone))
3212 			    != 0) {
3213 				return (error);
3214 			}
3215 #endif
3216 			sa6->sin6_scope_id = 0; /* XXX */
3217 			break;
3218 		}
3219 		case AF_LINK:	/* should eventually be supported */
3220 		default:
3221 			return (EAFNOSUPPORT);
3222 		}
3223 
3224 		/* turn off the previous option, then set the new option. */
3225 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
3226 		if (sticky) {
3227 			opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_WAITOK);
3228 			bcopy(buf, opt->ip6po_nexthop, *buf);
3229 		} else
3230 			opt->ip6po_nexthop = (struct sockaddr *)buf;
3231 		break;
3232 
3233 	case IPV6_2292HOPOPTS:
3234 	case IPV6_HOPOPTS:
3235 	{
3236 		struct ip6_hbh *hbh;
3237 		int hbhlen;
3238 
3239 		/*
3240 		 * XXX: We don't allow a non-privileged user to set ANY HbH
3241 		 * options, since per-option restriction has too much
3242 		 * overhead.
3243 		 */
3244 		if (!priv)
3245 			return (EPERM);
3246 
3247 		if (len == 0) {
3248 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
3249 			break;	/* just remove the option */
3250 		}
3251 
3252 		/* message length validation */
3253 		if (len < sizeof(struct ip6_hbh))
3254 			return (EINVAL);
3255 		hbh = (struct ip6_hbh *)buf;
3256 		hbhlen = (hbh->ip6h_len + 1) << 3;
3257 		if (len != hbhlen)
3258 			return (EINVAL);
3259 
3260 		/* turn off the previous option, then set the new option. */
3261 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
3262 		if (sticky) {
3263 			opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_WAITOK);
3264 			bcopy(hbh, opt->ip6po_hbh, hbhlen);
3265 		} else
3266 			opt->ip6po_hbh = hbh;
3267 
3268 		break;
3269 	}
3270 
3271 	case IPV6_2292DSTOPTS:
3272 	case IPV6_DSTOPTS:
3273 	case IPV6_RTHDRDSTOPTS:
3274 	{
3275 		struct ip6_dest *dest, **newdest = NULL;
3276 		int destlen;
3277 
3278 		if (!priv)	/* XXX: see the comment for IPV6_HOPOPTS */
3279 			return (EPERM);
3280 
3281 		if (len == 0) {
3282 			ip6_clearpktopts(opt, optname);
3283 			break;	/* just remove the option */
3284 		}
3285 
3286 		/* message length validation */
3287 		if (len < sizeof(struct ip6_dest))
3288 			return (EINVAL);
3289 		dest = (struct ip6_dest *)buf;
3290 		destlen = (dest->ip6d_len + 1) << 3;
3291 		if (len != destlen)
3292 			return (EINVAL);
3293 
3294 		/*
3295 		 * Determine the position that the destination options header
3296 		 * should be inserted; before or after the routing header.
3297 		 */
3298 		switch (optname) {
3299 		case IPV6_2292DSTOPTS:
3300 			/*
3301 			 * The old advacned API is ambiguous on this point.
3302 			 * Our approach is to determine the position based
3303 			 * according to the existence of a routing header.
3304 			 * Note, however, that this depends on the order of the
3305 			 * extension headers in the ancillary data; the 1st
3306 			 * part of the destination options header must appear
3307 			 * before the routing header in the ancillary data,
3308 			 * too.
3309 			 * RFC2292bis solved the ambiguity by introducing
3310 			 * separate ancillary data or option types.
3311 			 */
3312 			if (opt->ip6po_rthdr == NULL)
3313 				newdest = &opt->ip6po_dest1;
3314 			else
3315 				newdest = &opt->ip6po_dest2;
3316 			break;
3317 		case IPV6_RTHDRDSTOPTS:
3318 			newdest = &opt->ip6po_dest1;
3319 			break;
3320 		case IPV6_DSTOPTS:
3321 			newdest = &opt->ip6po_dest2;
3322 			break;
3323 		}
3324 
3325 		/* turn off the previous option, then set the new option. */
3326 		ip6_clearpktopts(opt, optname);
3327 		if (sticky) {
3328 			*newdest = malloc(destlen, M_IP6OPT, M_WAITOK);
3329 			bcopy(dest, *newdest, destlen);
3330 		} else
3331 			*newdest = dest;
3332 
3333 		break;
3334 	}
3335 
3336 	case IPV6_2292RTHDR:
3337 	case IPV6_RTHDR:
3338 	{
3339 		struct ip6_rthdr *rth;
3340 		int rthlen;
3341 
3342 		if (len == 0) {
3343 			ip6_clearpktopts(opt, IPV6_RTHDR);
3344 			break;	/* just remove the option */
3345 		}
3346 
3347 		/* message length validation */
3348 		if (len < sizeof(struct ip6_rthdr))
3349 			return (EINVAL);
3350 		rth = (struct ip6_rthdr *)buf;
3351 		rthlen = (rth->ip6r_len + 1) << 3;
3352 		if (len != rthlen)
3353 			return (EINVAL);
3354 
3355 		switch (rth->ip6r_type) {
3356 		case IPV6_RTHDR_TYPE_0:
3357 			if (rth->ip6r_len == 0)	/* must contain one addr */
3358 				return (EINVAL);
3359 			if (rth->ip6r_len % 2) /* length must be even */
3360 				return (EINVAL);
3361 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3362 				return (EINVAL);
3363 			break;
3364 		default:
3365 			return (EINVAL);	/* not supported */
3366 		}
3367 
3368 		/* turn off the previous option */
3369 		ip6_clearpktopts(opt, IPV6_RTHDR);
3370 		if (sticky) {
3371 			opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_WAITOK);
3372 			bcopy(rth, opt->ip6po_rthdr, rthlen);
3373 		} else
3374 			opt->ip6po_rthdr = rth;
3375 
3376 		break;
3377 	}
3378 
3379 	case IPV6_USE_MIN_MTU:
3380 		if (len != sizeof(int))
3381 			return (EINVAL);
3382 		minmtupolicy = *(int *)buf;
3383 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3384 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3385 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3386 			return (EINVAL);
3387 		}
3388 		opt->ip6po_minmtu = minmtupolicy;
3389 		break;
3390 
3391 	case IPV6_DONTFRAG:
3392 		if (len != sizeof(int))
3393 			return (EINVAL);
3394 
3395 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3396 			/*
3397 			 * we ignore this option for TCP sockets.
3398 			 * (rfc2292bis leaves this case unspecified.)
3399 			 */
3400 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3401 		} else
3402 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3403 		break;
3404 
3405 	case IPV6_PREFER_TEMPADDR:
3406 		if (len != sizeof(int))
3407 			return (EINVAL);
3408 		preftemp = *(int *)buf;
3409 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3410 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3411 		    preftemp != IP6PO_TEMPADDR_PREFER) {
3412 			return (EINVAL);
3413 		}
3414 		opt->ip6po_prefer_tempaddr = preftemp;
3415 		break;
3416 
3417 	default:
3418 		return (ENOPROTOOPT);
3419 	} /* end of switch */
3420 
3421 	return (0);
3422 }
3423 
3424 /*
3425  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3426  * packet to the input queue of a specified interface.  Note that this
3427  * calls the output routine of the loopback "driver", but with an interface
3428  * pointer that might NOT be &loif -- easier than replicating that code here.
3429  */
3430 void
3431 ip6_mloopback(ifp, m, dst)
3432 	struct ifnet *ifp;
3433 	struct mbuf *m;
3434 	struct sockaddr_in6 *dst;
3435 {
3436 	struct mbuf *copym;
3437 	struct ip6_hdr *ip6;
3438 
3439 	copym = m_copy(m, 0, M_COPYALL);
3440 	if (copym == NULL)
3441 		return;
3442 
3443 	/*
3444 	 * Make sure to deep-copy IPv6 header portion in case the data
3445 	 * is in an mbuf cluster, so that we can safely override the IPv6
3446 	 * header portion later.
3447 	 */
3448 	if ((copym->m_flags & M_EXT) != 0 ||
3449 	    copym->m_len < sizeof(struct ip6_hdr)) {
3450 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3451 		if (copym == NULL)
3452 			return;
3453 	}
3454 
3455 #ifdef DIAGNOSTIC
3456 	if (copym->m_len < sizeof(*ip6)) {
3457 		m_freem(copym);
3458 		return;
3459 	}
3460 #endif
3461 
3462 	ip6 = mtod(copym, struct ip6_hdr *);
3463 	/*
3464 	 * clear embedded scope identifiers if necessary.
3465 	 * in6_clearscope will touch the addresses only when necessary.
3466 	 */
3467 	in6_clearscope(&ip6->ip6_src);
3468 	in6_clearscope(&ip6->ip6_dst);
3469 
3470 	(void)if_simloop(ifp, copym, dst->sin6_family, 0);
3471 }
3472 
3473 /*
3474  * Chop IPv6 header off from the payload.
3475  */
3476 static int
3477 ip6_splithdr(m, exthdrs)
3478 	struct mbuf *m;
3479 	struct ip6_exthdrs *exthdrs;
3480 {
3481 	struct mbuf *mh;
3482 	struct ip6_hdr *ip6;
3483 
3484 	ip6 = mtod(m, struct ip6_hdr *);
3485 	if (m->m_len > sizeof(*ip6)) {
3486 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3487 		if (mh == 0) {
3488 			m_freem(m);
3489 			return ENOBUFS;
3490 		}
3491 		M_MOVE_PKTHDR(mh, m);
3492 		MH_ALIGN(mh, sizeof(*ip6));
3493 		m->m_len -= sizeof(*ip6);
3494 		m->m_data += sizeof(*ip6);
3495 		mh->m_next = m;
3496 		m = mh;
3497 		m->m_len = sizeof(*ip6);
3498 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3499 	}
3500 	exthdrs->ip6e_ip6 = m;
3501 	return 0;
3502 }
3503 
3504 /*
3505  * Compute IPv6 extension header length.
3506  */
3507 int
3508 ip6_optlen(in6p)
3509 	struct in6pcb *in6p;
3510 {
3511 	int len;
3512 
3513 	if (!in6p->in6p_outputopts)
3514 		return 0;
3515 
3516 	len = 0;
3517 #define elen(x) \
3518     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3519 
3520 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3521 	if (in6p->in6p_outputopts->ip6po_rthdr)
3522 		/* dest1 is valid with rthdr only */
3523 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
3524 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3525 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3526 	return len;
3527 #undef elen
3528 }
3529