xref: /freebsd/sys/netinet6/ip6_output.c (revision 35a04710d7286aa9538917fd7f8e417dbee95b82)
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_inet.h"
65 #include "opt_inet6.h"
66 #include "opt_ipsec.h"
67 
68 #include <sys/param.h>
69 #include <sys/malloc.h>
70 #include <sys/mbuf.h>
71 #include <sys/proc.h>
72 #include <sys/errno.h>
73 #include <sys/priv.h>
74 #include <sys/protosw.h>
75 #include <sys/socket.h>
76 #include <sys/socketvar.h>
77 #include <sys/kernel.h>
78 
79 #include <net/if.h>
80 #include <net/netisr.h>
81 #include <net/route.h>
82 #include <net/pfil.h>
83 
84 #include <netinet/in.h>
85 #include <netinet/in_var.h>
86 #include <netinet6/in6_var.h>
87 #include <netinet/ip6.h>
88 #include <netinet/icmp6.h>
89 #include <netinet6/ip6_var.h>
90 #include <netinet/in_pcb.h>
91 #include <netinet/tcp_var.h>
92 #include <netinet6/nd6.h>
93 
94 #ifdef IPSEC
95 #include <netipsec/ipsec.h>
96 #include <netipsec/ipsec6.h>
97 #include <netipsec/key.h>
98 #include <netinet6/ip6_ipsec.h>
99 #endif /* IPSEC */
100 
101 #include <netinet6/ip6protosw.h>
102 #include <netinet6/scope6_var.h>
103 
104 static MALLOC_DEFINE(M_IP6MOPTS, "ip6_moptions", "internet multicast options");
105 
106 struct ip6_exthdrs {
107 	struct mbuf *ip6e_ip6;
108 	struct mbuf *ip6e_hbh;
109 	struct mbuf *ip6e_dest1;
110 	struct mbuf *ip6e_rthdr;
111 	struct mbuf *ip6e_dest2;
112 };
113 
114 static int ip6_pcbopt __P((int, u_char *, int, struct ip6_pktopts **,
115 			   int, int));
116 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
117 	struct socket *, struct sockopt *));
118 static int ip6_getpcbopt __P((struct ip6_pktopts *, int, struct sockopt *));
119 static int ip6_setpktopt __P((int, u_char *, int, struct ip6_pktopts *, int,
120 	int, int, int));
121 
122 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *));
123 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **));
124 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int));
125 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
126 	struct ip6_frag **));
127 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t));
128 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *));
129 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *,
130 	struct ifnet *, struct in6_addr *, u_long *, int *));
131 static int copypktopts __P((struct ip6_pktopts *, struct ip6_pktopts *, int));
132 
133 
134 /*
135  * Make an extension header from option data.  hp is the source, and
136  * mp is the destination.
137  */
138 #define MAKE_EXTHDR(hp, mp)						\
139     do {								\
140 	if (hp) {							\
141 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
142 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
143 		    ((eh)->ip6e_len + 1) << 3);				\
144 		if (error)						\
145 			goto freehdrs;					\
146 	}								\
147     } while (/*CONSTCOND*/ 0)
148 
149 /*
150  * Form a chain of extension headers.
151  * m is the extension header mbuf
152  * mp is the previous mbuf in the chain
153  * p is the next header
154  * i is the type of option.
155  */
156 #define MAKE_CHAIN(m, mp, p, i)\
157     do {\
158 	if (m) {\
159 		if (!hdrsplit) \
160 			panic("assumption failed: hdr not split"); \
161 		*mtod((m), u_char *) = *(p);\
162 		*(p) = (i);\
163 		p = mtod((m), u_char *);\
164 		(m)->m_next = (mp)->m_next;\
165 		(mp)->m_next = (m);\
166 		(mp) = (m);\
167 	}\
168     } while (/*CONSTCOND*/ 0)
169 
170 /*
171  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
172  * header (with pri, len, nxt, hlim, src, dst).
173  * This function may modify ver and hlim only.
174  * The mbuf chain containing the packet will be freed.
175  * The mbuf opt, if present, will not be freed.
176  *
177  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
178  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
179  * which is rt_rmx.rmx_mtu.
180  *
181  * ifpp - XXX: just for statistics
182  */
183 int
184 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
185     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
186     struct ifnet **ifpp, struct inpcb *inp)
187 {
188 	struct ip6_hdr *ip6, *mhip6;
189 	struct ifnet *ifp, *origifp;
190 	struct mbuf *m = m0;
191 	struct mbuf *mprev = NULL;
192 	int hlen, tlen, len, off;
193 	struct route_in6 ip6route;
194 	struct rtentry *rt = NULL;
195 	struct sockaddr_in6 *dst, src_sa, dst_sa;
196 	struct in6_addr odst;
197 	int error = 0;
198 	struct in6_ifaddr *ia = NULL;
199 	u_long mtu;
200 	int alwaysfrag, dontfrag;
201 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
202 	struct ip6_exthdrs exthdrs;
203 	struct in6_addr finaldst, src0, dst0;
204 	u_int32_t zone;
205 	struct route_in6 *ro_pmtu = NULL;
206 	int hdrsplit = 0;
207 	int needipsec = 0;
208 #ifdef IPSEC
209 	struct ipsec_output_state state;
210 	struct ip6_rthdr *rh = NULL;
211 	int needipsectun = 0;
212 	int segleft_org = 0;
213 	struct secpolicy *sp = NULL;
214 #endif /* IPSEC */
215 
216 	ip6 = mtod(m, struct ip6_hdr *);
217 	if (ip6 == NULL) {
218 		printf ("ip6 is NULL");
219 		goto bad;
220 	}
221 
222 	finaldst = ip6->ip6_dst;
223 
224 	bzero(&exthdrs, sizeof(exthdrs));
225 
226 	if (opt) {
227 		/* Hop-by-Hop options header */
228 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
229 		/* Destination options header(1st part) */
230 		if (opt->ip6po_rthdr) {
231 			/*
232 			 * Destination options header(1st part)
233 			 * This only makes sense with a routing header.
234 			 * See Section 9.2 of RFC 3542.
235 			 * Disabling this part just for MIP6 convenience is
236 			 * a bad idea.  We need to think carefully about a
237 			 * way to make the advanced API coexist with MIP6
238 			 * options, which might automatically be inserted in
239 			 * the kernel.
240 			 */
241 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
242 		}
243 		/* Routing header */
244 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
245 		/* Destination options header(2nd part) */
246 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
247 	}
248 
249 	/*
250 	 * IPSec checking which handles several cases.
251 	 * FAST IPSEC: We re-injected the packet.
252 	 */
253 #ifdef IPSEC
254 	switch(ip6_ipsec_output(&m, inp, &flags, &error, &ifp, &sp))
255 	{
256 	case 1:                 /* Bad packet */
257 		goto freehdrs;
258 	case -1:                /* Do IPSec */
259 		needipsec = 1;
260 	case 0:                 /* No IPSec */
261 	default:
262 		break;
263 	}
264 #endif /* IPSEC */
265 
266 	/*
267 	 * Calculate the total length of the extension header chain.
268 	 * Keep the length of the unfragmentable part for fragmentation.
269 	 */
270 	optlen = 0;
271 	if (exthdrs.ip6e_hbh)
272 		optlen += exthdrs.ip6e_hbh->m_len;
273 	if (exthdrs.ip6e_dest1)
274 		optlen += exthdrs.ip6e_dest1->m_len;
275 	if (exthdrs.ip6e_rthdr)
276 		optlen += exthdrs.ip6e_rthdr->m_len;
277 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
278 
279 	/* NOTE: we don't add AH/ESP length here. do that later. */
280 	if (exthdrs.ip6e_dest2)
281 		optlen += exthdrs.ip6e_dest2->m_len;
282 
283 	/*
284 	 * If we need IPsec, or there is at least one extension header,
285 	 * separate IP6 header from the payload.
286 	 */
287 	if ((needipsec || optlen) && !hdrsplit) {
288 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
289 			m = NULL;
290 			goto freehdrs;
291 		}
292 		m = exthdrs.ip6e_ip6;
293 		hdrsplit++;
294 	}
295 
296 	/* adjust pointer */
297 	ip6 = mtod(m, struct ip6_hdr *);
298 
299 	/* adjust mbuf packet header length */
300 	m->m_pkthdr.len += optlen;
301 	plen = m->m_pkthdr.len - sizeof(*ip6);
302 
303 	/* If this is a jumbo payload, insert a jumbo payload option. */
304 	if (plen > IPV6_MAXPACKET) {
305 		if (!hdrsplit) {
306 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
307 				m = NULL;
308 				goto freehdrs;
309 			}
310 			m = exthdrs.ip6e_ip6;
311 			hdrsplit++;
312 		}
313 		/* adjust pointer */
314 		ip6 = mtod(m, struct ip6_hdr *);
315 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
316 			goto freehdrs;
317 		ip6->ip6_plen = 0;
318 	} else
319 		ip6->ip6_plen = htons(plen);
320 
321 	/*
322 	 * Concatenate headers and fill in next header fields.
323 	 * Here we have, on "m"
324 	 *	IPv6 payload
325 	 * and we insert headers accordingly.  Finally, we should be getting:
326 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
327 	 *
328 	 * during the header composing process, "m" points to IPv6 header.
329 	 * "mprev" points to an extension header prior to esp.
330 	 */
331 	u_char *nexthdrp = &ip6->ip6_nxt;
332 	mprev = m;
333 
334 	/*
335 	 * we treat dest2 specially.  this makes IPsec processing
336 	 * much easier.  the goal here is to make mprev point the
337 	 * mbuf prior to dest2.
338 	 *
339 	 * result: IPv6 dest2 payload
340 	 * m and mprev will point to IPv6 header.
341 	 */
342 	if (exthdrs.ip6e_dest2) {
343 		if (!hdrsplit)
344 			panic("assumption failed: hdr not split");
345 		exthdrs.ip6e_dest2->m_next = m->m_next;
346 		m->m_next = exthdrs.ip6e_dest2;
347 		*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
348 		ip6->ip6_nxt = IPPROTO_DSTOPTS;
349 	}
350 
351 	/*
352 	 * result: IPv6 hbh dest1 rthdr dest2 payload
353 	 * m will point to IPv6 header.  mprev will point to the
354 	 * extension header prior to dest2 (rthdr in the above case).
355 	 */
356 	MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
357 	MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
358 		   IPPROTO_DSTOPTS);
359 	MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
360 		   IPPROTO_ROUTING);
361 
362 #ifdef IPSEC
363 	if (!needipsec)
364 		goto skip_ipsec2;
365 
366 	/*
367 	 * pointers after IPsec headers are not valid any more.
368 	 * other pointers need a great care too.
369 	 * (IPsec routines should not mangle mbufs prior to AH/ESP)
370 	 */
371 	exthdrs.ip6e_dest2 = NULL;
372 
373 	if (exthdrs.ip6e_rthdr) {
374 		rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
375 		segleft_org = rh->ip6r_segleft;
376 		rh->ip6r_segleft = 0;
377 	}
378 
379 	bzero(&state, sizeof(state));
380 	state.m = m;
381 	error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
382 				    &needipsectun);
383 	m = state.m;
384 	if (error) {
385 		/* mbuf is already reclaimed in ipsec6_output_trans. */
386 		m = NULL;
387 		switch (error) {
388 		case EHOSTUNREACH:
389 		case ENETUNREACH:
390 		case EMSGSIZE:
391 		case ENOBUFS:
392 		case ENOMEM:
393 			break;
394 		default:
395 			printf("ip6_output (ipsec): error code %d\n", error);
396 			/* FALLTHROUGH */
397 		case ENOENT:
398 			/* don't show these error codes to the user */
399 			error = 0;
400 			break;
401 		}
402 		goto bad;
403 	} else if (!needipsectun) {
404 		/*
405 		 * In the FAST IPSec case we have already
406 		 * re-injected the packet and it has been freed
407 		 * by the ipsec_done() function.  So, just clean
408 		 * up after ourselves.
409 		 */
410 		m = NULL;
411 		goto done;
412 	}
413 	if (exthdrs.ip6e_rthdr) {
414 		/* ah6_output doesn't modify mbuf chain */
415 		rh->ip6r_segleft = segleft_org;
416 	}
417 skip_ipsec2:;
418 #endif /* IPSEC */
419 
420 	/*
421 	 * If there is a routing header, replace the destination address field
422 	 * with the first hop of the routing header.
423 	 */
424 	if (exthdrs.ip6e_rthdr) {
425 		struct ip6_rthdr *rh =
426 			(struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
427 						  struct ip6_rthdr *));
428 		struct ip6_rthdr0 *rh0;
429 		struct in6_addr *addr;
430 		struct sockaddr_in6 sa;
431 
432 		switch (rh->ip6r_type) {
433 		case IPV6_RTHDR_TYPE_0:
434 			 rh0 = (struct ip6_rthdr0 *)rh;
435 			 addr = (struct in6_addr *)(rh0 + 1);
436 
437 			 /*
438 			  * construct a sockaddr_in6 form of
439 			  * the first hop.
440 			  *
441 			  * XXX: we may not have enough
442 			  * information about its scope zone;
443 			  * there is no standard API to pass
444 			  * the information from the
445 			  * application.
446 			  */
447 			 bzero(&sa, sizeof(sa));
448 			 sa.sin6_family = AF_INET6;
449 			 sa.sin6_len = sizeof(sa);
450 			 sa.sin6_addr = addr[0];
451 			 if ((error = sa6_embedscope(&sa,
452 			     ip6_use_defzone)) != 0) {
453 				 goto bad;
454 			 }
455 			 ip6->ip6_dst = sa.sin6_addr;
456 			 bcopy(&addr[1], &addr[0], sizeof(struct in6_addr)
457 			     * (rh0->ip6r0_segleft - 1));
458 			 addr[rh0->ip6r0_segleft - 1] = finaldst;
459 			 /* XXX */
460 			 in6_clearscope(addr + rh0->ip6r0_segleft - 1);
461 			 break;
462 		default:	/* is it possible? */
463 			 error = EINVAL;
464 			 goto bad;
465 		}
466 	}
467 
468 	/* Source address validation */
469 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
470 	    (flags & IPV6_UNSPECSRC) == 0) {
471 		error = EOPNOTSUPP;
472 		ip6stat.ip6s_badscope++;
473 		goto bad;
474 	}
475 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
476 		error = EOPNOTSUPP;
477 		ip6stat.ip6s_badscope++;
478 		goto bad;
479 	}
480 
481 	ip6stat.ip6s_localout++;
482 
483 	/*
484 	 * Route packet.
485 	 */
486 	if (ro == 0) {
487 		ro = &ip6route;
488 		bzero((caddr_t)ro, sizeof(*ro));
489 	}
490 	ro_pmtu = ro;
491 	if (opt && opt->ip6po_rthdr)
492 		ro = &opt->ip6po_route;
493 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
494 
495 again:
496 	/*
497 	 * if specified, try to fill in the traffic class field.
498 	 * do not override if a non-zero value is already set.
499 	 * we check the diffserv field and the ecn field separately.
500 	 */
501 	if (opt && opt->ip6po_tclass >= 0) {
502 		int mask = 0;
503 
504 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
505 			mask |= 0xfc;
506 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
507 			mask |= 0x03;
508 		if (mask != 0)
509 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
510 	}
511 
512 	/* fill in or override the hop limit field, if necessary. */
513 	if (opt && opt->ip6po_hlim != -1)
514 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
515 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
516 		if (im6o != NULL)
517 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
518 		else
519 			ip6->ip6_hlim = ip6_defmcasthlim;
520 	}
521 
522 #ifdef IPSEC
523 	/*
524 	 * We may re-inject packets into the stack here.
525 	 */
526 	if (needipsec && needipsectun) {
527 		struct ipsec_output_state state;
528 
529 		/*
530 		 * All the extension headers will become inaccessible
531 		 * (since they can be encrypted).
532 		 * Don't panic, we need no more updates to extension headers
533 		 * on inner IPv6 packet (since they are now encapsulated).
534 		 *
535 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
536 		 */
537 		bzero(&exthdrs, sizeof(exthdrs));
538 		exthdrs.ip6e_ip6 = m;
539 
540 		bzero(&state, sizeof(state));
541 		state.m = m;
542 		state.ro = (struct route *)ro;
543 		state.dst = (struct sockaddr *)dst;
544 
545 		error = ipsec6_output_tunnel(&state, sp, flags);
546 
547 		m = state.m;
548 		ro = (struct route_in6 *)state.ro;
549 		dst = (struct sockaddr_in6 *)state.dst;
550 		if (error) {
551 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
552 			m0 = m = NULL;
553 			m = NULL;
554 			switch (error) {
555 			case EHOSTUNREACH:
556 			case ENETUNREACH:
557 			case EMSGSIZE:
558 			case ENOBUFS:
559 			case ENOMEM:
560 				break;
561 			default:
562 				printf("ip6_output (ipsec): error code %d\n", error);
563 				/* FALLTHROUGH */
564 			case ENOENT:
565 				/* don't show these error codes to the user */
566 				error = 0;
567 				break;
568 			}
569 			goto bad;
570 		} else {
571 			/*
572 			 * In the FAST IPSec case we have already
573 			 * re-injected the packet and it has been freed
574 			 * by the ipsec_done() function.  So, just clean
575 			 * up after ourselves.
576 			 */
577 			m = NULL;
578 			goto done;
579 		}
580 
581 		exthdrs.ip6e_ip6 = m;
582 	}
583 #endif /* IPSEC */
584 
585 	/* adjust pointer */
586 	ip6 = mtod(m, struct ip6_hdr *);
587 
588 	bzero(&dst_sa, sizeof(dst_sa));
589 	dst_sa.sin6_family = AF_INET6;
590 	dst_sa.sin6_len = sizeof(dst_sa);
591 	dst_sa.sin6_addr = ip6->ip6_dst;
592 	if ((error = in6_selectroute(&dst_sa, opt, im6o, ro,
593 	    &ifp, &rt, 0)) != 0) {
594 		switch (error) {
595 		case EHOSTUNREACH:
596 			ip6stat.ip6s_noroute++;
597 			break;
598 		case EADDRNOTAVAIL:
599 		default:
600 			break; /* XXX statistics? */
601 		}
602 		if (ifp != NULL)
603 			in6_ifstat_inc(ifp, ifs6_out_discard);
604 		goto bad;
605 	}
606 	if (rt == NULL) {
607 		/*
608 		 * If in6_selectroute() does not return a route entry,
609 		 * dst may not have been updated.
610 		 */
611 		*dst = dst_sa;	/* XXX */
612 	}
613 
614 	/*
615 	 * then rt (for unicast) and ifp must be non-NULL valid values.
616 	 */
617 	if ((flags & IPV6_FORWARDING) == 0) {
618 		/* XXX: the FORWARDING flag can be set for mrouting. */
619 		in6_ifstat_inc(ifp, ifs6_out_request);
620 	}
621 	if (rt != NULL) {
622 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
623 		rt->rt_use++;
624 	}
625 
626 	/*
627 	 * The outgoing interface must be in the zone of source and
628 	 * destination addresses.  We should use ia_ifp to support the
629 	 * case of sending packets to an address of our own.
630 	 */
631 	if (ia != NULL && ia->ia_ifp)
632 		origifp = ia->ia_ifp;
633 	else
634 		origifp = ifp;
635 
636 	src0 = ip6->ip6_src;
637 	if (in6_setscope(&src0, origifp, &zone))
638 		goto badscope;
639 	bzero(&src_sa, sizeof(src_sa));
640 	src_sa.sin6_family = AF_INET6;
641 	src_sa.sin6_len = sizeof(src_sa);
642 	src_sa.sin6_addr = ip6->ip6_src;
643 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
644 		goto badscope;
645 
646 	dst0 = ip6->ip6_dst;
647 	if (in6_setscope(&dst0, origifp, &zone))
648 		goto badscope;
649 	/* re-initialize to be sure */
650 	bzero(&dst_sa, sizeof(dst_sa));
651 	dst_sa.sin6_family = AF_INET6;
652 	dst_sa.sin6_len = sizeof(dst_sa);
653 	dst_sa.sin6_addr = ip6->ip6_dst;
654 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
655 		goto badscope;
656 	}
657 
658 	/* scope check is done. */
659 	goto routefound;
660 
661   badscope:
662 	ip6stat.ip6s_badscope++;
663 	in6_ifstat_inc(origifp, ifs6_out_discard);
664 	if (error == 0)
665 		error = EHOSTUNREACH; /* XXX */
666 	goto bad;
667 
668   routefound:
669 	if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
670 		if (opt && opt->ip6po_nextroute.ro_rt) {
671 			/*
672 			 * The nexthop is explicitly specified by the
673 			 * application.  We assume the next hop is an IPv6
674 			 * address.
675 			 */
676 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
677 		}
678 		else if ((rt->rt_flags & RTF_GATEWAY))
679 			dst = (struct sockaddr_in6 *)rt->rt_gateway;
680 	}
681 
682 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
683 		m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
684 	} else {
685 		struct	in6_multi *in6m;
686 
687 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
688 
689 		in6_ifstat_inc(ifp, ifs6_out_mcast);
690 
691 		/*
692 		 * Confirm that the outgoing interface supports multicast.
693 		 */
694 		if (!(ifp->if_flags & IFF_MULTICAST)) {
695 			ip6stat.ip6s_noroute++;
696 			in6_ifstat_inc(ifp, ifs6_out_discard);
697 			error = ENETUNREACH;
698 			goto bad;
699 		}
700 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
701 		if (in6m != NULL &&
702 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
703 			/*
704 			 * If we belong to the destination multicast group
705 			 * on the outgoing interface, and the caller did not
706 			 * forbid loopback, loop back a copy.
707 			 */
708 			ip6_mloopback(ifp, m, dst);
709 		} else {
710 			/*
711 			 * If we are acting as a multicast router, perform
712 			 * multicast forwarding as if the packet had just
713 			 * arrived on the interface to which we are about
714 			 * to send.  The multicast forwarding function
715 			 * recursively calls this function, using the
716 			 * IPV6_FORWARDING flag to prevent infinite recursion.
717 			 *
718 			 * Multicasts that are looped back by ip6_mloopback(),
719 			 * above, will be forwarded by the ip6_input() routine,
720 			 * if necessary.
721 			 */
722 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
723 				/*
724 				 * XXX: ip6_mforward expects that rcvif is NULL
725 				 * when it is called from the originating path.
726 				 * However, it is not always the case, since
727 				 * some versions of MGETHDR() does not
728 				 * initialize the field.
729 				 */
730 				m->m_pkthdr.rcvif = NULL;
731 				if (ip6_mforward(ip6, ifp, m) != 0) {
732 					m_freem(m);
733 					goto done;
734 				}
735 			}
736 		}
737 		/*
738 		 * Multicasts with a hoplimit of zero may be looped back,
739 		 * above, but must not be transmitted on a network.
740 		 * Also, multicasts addressed to the loopback interface
741 		 * are not sent -- the above call to ip6_mloopback() will
742 		 * loop back a copy if this host actually belongs to the
743 		 * destination group on the loopback interface.
744 		 */
745 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
746 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
747 			m_freem(m);
748 			goto done;
749 		}
750 	}
751 
752 	/*
753 	 * Fill the outgoing inteface to tell the upper layer
754 	 * to increment per-interface statistics.
755 	 */
756 	if (ifpp)
757 		*ifpp = ifp;
758 
759 	/* Determine path MTU. */
760 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
761 	    &alwaysfrag)) != 0)
762 		goto bad;
763 
764 	/*
765 	 * The caller of this function may specify to use the minimum MTU
766 	 * in some cases.
767 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
768 	 * setting.  The logic is a bit complicated; by default, unicast
769 	 * packets will follow path MTU while multicast packets will be sent at
770 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
771 	 * including unicast ones will be sent at the minimum MTU.  Multicast
772 	 * packets will always be sent at the minimum MTU unless
773 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
774 	 * See RFC 3542 for more details.
775 	 */
776 	if (mtu > IPV6_MMTU) {
777 		if ((flags & IPV6_MINMTU))
778 			mtu = IPV6_MMTU;
779 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
780 			mtu = IPV6_MMTU;
781 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
782 			 (opt == NULL ||
783 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
784 			mtu = IPV6_MMTU;
785 		}
786 	}
787 
788 	/*
789 	 * clear embedded scope identifiers if necessary.
790 	 * in6_clearscope will touch the addresses only when necessary.
791 	 */
792 	in6_clearscope(&ip6->ip6_src);
793 	in6_clearscope(&ip6->ip6_dst);
794 
795 	/*
796 	 * If the outgoing packet contains a hop-by-hop options header,
797 	 * it must be examined and processed even by the source node.
798 	 * (RFC 2460, section 4.)
799 	 */
800 	if (exthdrs.ip6e_hbh) {
801 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
802 		u_int32_t dummy; /* XXX unused */
803 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
804 
805 #ifdef DIAGNOSTIC
806 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
807 			panic("ip6e_hbh is not continuous");
808 #endif
809 		/*
810 		 *  XXX: if we have to send an ICMPv6 error to the sender,
811 		 *       we need the M_LOOP flag since icmp6_error() expects
812 		 *       the IPv6 and the hop-by-hop options header are
813 		 *       continuous unless the flag is set.
814 		 */
815 		m->m_flags |= M_LOOP;
816 		m->m_pkthdr.rcvif = ifp;
817 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
818 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
819 		    &dummy, &plen) < 0) {
820 			/* m was already freed at this point */
821 			error = EINVAL;/* better error? */
822 			goto done;
823 		}
824 		m->m_flags &= ~M_LOOP; /* XXX */
825 		m->m_pkthdr.rcvif = NULL;
826 	}
827 
828 	/* Jump over all PFIL processing if hooks are not active. */
829 	if (!PFIL_HOOKED(&inet6_pfil_hook))
830 		goto passout;
831 
832 	odst = ip6->ip6_dst;
833 	/* Run through list of hooks for output packets. */
834 	error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
835 	if (error != 0 || m == NULL)
836 		goto done;
837 	ip6 = mtod(m, struct ip6_hdr *);
838 
839 	/* See if destination IP address was changed by packet filter. */
840 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
841 		m->m_flags |= M_SKIP_FIREWALL;
842 		/* If destination is now ourself drop to ip6_input(). */
843 		if (in6_localaddr(&ip6->ip6_dst)) {
844 			if (m->m_pkthdr.rcvif == NULL)
845 				m->m_pkthdr.rcvif = loif;
846 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
847 				m->m_pkthdr.csum_flags |=
848 				    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
849 				m->m_pkthdr.csum_data = 0xffff;
850 			}
851 			m->m_pkthdr.csum_flags |=
852 			    CSUM_IP_CHECKED | CSUM_IP_VALID;
853 			error = netisr_queue(NETISR_IPV6, m);
854 			goto done;
855 		} else
856 			goto again;	/* Redo the routing table lookup. */
857 	}
858 
859 	/* XXX: IPFIREWALL_FORWARD */
860 
861 passout:
862 	/*
863 	 * Send the packet to the outgoing interface.
864 	 * If necessary, do IPv6 fragmentation before sending.
865 	 *
866 	 * the logic here is rather complex:
867 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
868 	 * 1-a:	send as is if tlen <= path mtu
869 	 * 1-b:	fragment if tlen > path mtu
870 	 *
871 	 * 2: if user asks us not to fragment (dontfrag == 1)
872 	 * 2-a:	send as is if tlen <= interface mtu
873 	 * 2-b:	error if tlen > interface mtu
874 	 *
875 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
876 	 *	always fragment
877 	 *
878 	 * 4: if dontfrag == 1 && alwaysfrag == 1
879 	 *	error, as we cannot handle this conflicting request
880 	 */
881 	tlen = m->m_pkthdr.len;
882 
883 	if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
884 		dontfrag = 1;
885 	else
886 		dontfrag = 0;
887 	if (dontfrag && alwaysfrag) {	/* case 4 */
888 		/* conflicting request - can't transmit */
889 		error = EMSGSIZE;
890 		goto bad;
891 	}
892 	if (dontfrag && tlen > IN6_LINKMTU(ifp)) {	/* case 2-b */
893 		/*
894 		 * Even if the DONTFRAG option is specified, we cannot send the
895 		 * packet when the data length is larger than the MTU of the
896 		 * outgoing interface.
897 		 * Notify the error by sending IPV6_PATHMTU ancillary data as
898 		 * well as returning an error code (the latter is not described
899 		 * in the API spec.)
900 		 */
901 		u_int32_t mtu32;
902 		struct ip6ctlparam ip6cp;
903 
904 		mtu32 = (u_int32_t)mtu;
905 		bzero(&ip6cp, sizeof(ip6cp));
906 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
907 		pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
908 		    (void *)&ip6cp);
909 
910 		error = EMSGSIZE;
911 		goto bad;
912 	}
913 
914 	/*
915 	 * transmit packet without fragmentation
916 	 */
917 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
918 		struct in6_ifaddr *ia6;
919 
920 		ip6 = mtod(m, struct ip6_hdr *);
921 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
922 		if (ia6) {
923 			/* Record statistics for this interface address. */
924 			ia6->ia_ifa.if_opackets++;
925 			ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
926 		}
927 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
928 		goto done;
929 	}
930 
931 	/*
932 	 * try to fragment the packet.  case 1-b and 3
933 	 */
934 	if (mtu < IPV6_MMTU) {
935 		/* path MTU cannot be less than IPV6_MMTU */
936 		error = EMSGSIZE;
937 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
938 		goto bad;
939 	} else if (ip6->ip6_plen == 0) {
940 		/* jumbo payload cannot be fragmented */
941 		error = EMSGSIZE;
942 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
943 		goto bad;
944 	} else {
945 		struct mbuf **mnext, *m_frgpart;
946 		struct ip6_frag *ip6f;
947 		u_int32_t id = htonl(ip6_randomid());
948 		u_char nextproto;
949 
950 		int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
951 
952 		/*
953 		 * Too large for the destination or interface;
954 		 * fragment if possible.
955 		 * Must be able to put at least 8 bytes per fragment.
956 		 */
957 		hlen = unfragpartlen;
958 		if (mtu > IPV6_MAXPACKET)
959 			mtu = IPV6_MAXPACKET;
960 
961 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
962 		if (len < 8) {
963 			error = EMSGSIZE;
964 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
965 			goto bad;
966 		}
967 
968 		/*
969 		 * Verify that we have any chance at all of being able to queue
970 		 *      the packet or packet fragments
971 		 */
972 		if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
973 		    < tlen  /* - hlen */)) {
974 			error = ENOBUFS;
975 			ip6stat.ip6s_odropped++;
976 			goto bad;
977 		}
978 
979 		mnext = &m->m_nextpkt;
980 
981 		/*
982 		 * Change the next header field of the last header in the
983 		 * unfragmentable part.
984 		 */
985 		if (exthdrs.ip6e_rthdr) {
986 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
987 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
988 		} else if (exthdrs.ip6e_dest1) {
989 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
990 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
991 		} else if (exthdrs.ip6e_hbh) {
992 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
993 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
994 		} else {
995 			nextproto = ip6->ip6_nxt;
996 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
997 		}
998 
999 		/*
1000 		 * Loop through length of segment after first fragment,
1001 		 * make new header and copy data of each part and link onto
1002 		 * chain.
1003 		 */
1004 		m0 = m;
1005 		for (off = hlen; off < tlen; off += len) {
1006 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
1007 			if (!m) {
1008 				error = ENOBUFS;
1009 				ip6stat.ip6s_odropped++;
1010 				goto sendorfree;
1011 			}
1012 			m->m_pkthdr.rcvif = NULL;
1013 			m->m_flags = m0->m_flags & M_COPYFLAGS;
1014 			*mnext = m;
1015 			mnext = &m->m_nextpkt;
1016 			m->m_data += max_linkhdr;
1017 			mhip6 = mtod(m, struct ip6_hdr *);
1018 			*mhip6 = *ip6;
1019 			m->m_len = sizeof(*mhip6);
1020 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1021 			if (error) {
1022 				ip6stat.ip6s_odropped++;
1023 				goto sendorfree;
1024 			}
1025 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1026 			if (off + len >= tlen)
1027 				len = tlen - off;
1028 			else
1029 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1030 			mhip6->ip6_plen = htons((u_short)(len + hlen +
1031 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1032 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1033 				error = ENOBUFS;
1034 				ip6stat.ip6s_odropped++;
1035 				goto sendorfree;
1036 			}
1037 			m_cat(m, m_frgpart);
1038 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1039 			m->m_pkthdr.rcvif = NULL;
1040 			ip6f->ip6f_reserved = 0;
1041 			ip6f->ip6f_ident = id;
1042 			ip6f->ip6f_nxt = nextproto;
1043 			ip6stat.ip6s_ofragments++;
1044 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1045 		}
1046 
1047 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1048 	}
1049 
1050 	/*
1051 	 * Remove leading garbages.
1052 	 */
1053 sendorfree:
1054 	m = m0->m_nextpkt;
1055 	m0->m_nextpkt = 0;
1056 	m_freem(m0);
1057 	for (m0 = m; m; m = m0) {
1058 		m0 = m->m_nextpkt;
1059 		m->m_nextpkt = 0;
1060 		if (error == 0) {
1061 			/* Record statistics for this interface address. */
1062 			if (ia) {
1063 				ia->ia_ifa.if_opackets++;
1064 				ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1065 			}
1066 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1067 		} else
1068 			m_freem(m);
1069 	}
1070 
1071 	if (error == 0)
1072 		ip6stat.ip6s_fragmented++;
1073 
1074 done:
1075 	if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1076 		RTFREE(ro->ro_rt);
1077 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1078 		RTFREE(ro_pmtu->ro_rt);
1079 	}
1080 
1081 	return (error);
1082 
1083 freehdrs:
1084 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1085 	m_freem(exthdrs.ip6e_dest1);
1086 	m_freem(exthdrs.ip6e_rthdr);
1087 	m_freem(exthdrs.ip6e_dest2);
1088 	/* FALLTHROUGH */
1089 bad:
1090 	if (m)
1091 		m_freem(m);
1092 	goto done;
1093 }
1094 
1095 static int
1096 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1097 {
1098 	struct mbuf *m;
1099 
1100 	if (hlen > MCLBYTES)
1101 		return (ENOBUFS); /* XXX */
1102 
1103 	MGET(m, M_DONTWAIT, MT_DATA);
1104 	if (!m)
1105 		return (ENOBUFS);
1106 
1107 	if (hlen > MLEN) {
1108 		MCLGET(m, M_DONTWAIT);
1109 		if ((m->m_flags & M_EXT) == 0) {
1110 			m_free(m);
1111 			return (ENOBUFS);
1112 		}
1113 	}
1114 	m->m_len = hlen;
1115 	if (hdr)
1116 		bcopy(hdr, mtod(m, caddr_t), hlen);
1117 
1118 	*mp = m;
1119 	return (0);
1120 }
1121 
1122 /*
1123  * Insert jumbo payload option.
1124  */
1125 static int
1126 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1127 {
1128 	struct mbuf *mopt;
1129 	u_char *optbuf;
1130 	u_int32_t v;
1131 
1132 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1133 
1134 	/*
1135 	 * If there is no hop-by-hop options header, allocate new one.
1136 	 * If there is one but it doesn't have enough space to store the
1137 	 * jumbo payload option, allocate a cluster to store the whole options.
1138 	 * Otherwise, use it to store the options.
1139 	 */
1140 	if (exthdrs->ip6e_hbh == 0) {
1141 		MGET(mopt, M_DONTWAIT, MT_DATA);
1142 		if (mopt == 0)
1143 			return (ENOBUFS);
1144 		mopt->m_len = JUMBOOPTLEN;
1145 		optbuf = mtod(mopt, u_char *);
1146 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1147 		exthdrs->ip6e_hbh = mopt;
1148 	} else {
1149 		struct ip6_hbh *hbh;
1150 
1151 		mopt = exthdrs->ip6e_hbh;
1152 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1153 			/*
1154 			 * XXX assumption:
1155 			 * - exthdrs->ip6e_hbh is not referenced from places
1156 			 *   other than exthdrs.
1157 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1158 			 */
1159 			int oldoptlen = mopt->m_len;
1160 			struct mbuf *n;
1161 
1162 			/*
1163 			 * XXX: give up if the whole (new) hbh header does
1164 			 * not fit even in an mbuf cluster.
1165 			 */
1166 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1167 				return (ENOBUFS);
1168 
1169 			/*
1170 			 * As a consequence, we must always prepare a cluster
1171 			 * at this point.
1172 			 */
1173 			MGET(n, M_DONTWAIT, MT_DATA);
1174 			if (n) {
1175 				MCLGET(n, M_DONTWAIT);
1176 				if ((n->m_flags & M_EXT) == 0) {
1177 					m_freem(n);
1178 					n = NULL;
1179 				}
1180 			}
1181 			if (!n)
1182 				return (ENOBUFS);
1183 			n->m_len = oldoptlen + JUMBOOPTLEN;
1184 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1185 			    oldoptlen);
1186 			optbuf = mtod(n, caddr_t) + oldoptlen;
1187 			m_freem(mopt);
1188 			mopt = exthdrs->ip6e_hbh = n;
1189 		} else {
1190 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1191 			mopt->m_len += JUMBOOPTLEN;
1192 		}
1193 		optbuf[0] = IP6OPT_PADN;
1194 		optbuf[1] = 1;
1195 
1196 		/*
1197 		 * Adjust the header length according to the pad and
1198 		 * the jumbo payload option.
1199 		 */
1200 		hbh = mtod(mopt, struct ip6_hbh *);
1201 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1202 	}
1203 
1204 	/* fill in the option. */
1205 	optbuf[2] = IP6OPT_JUMBO;
1206 	optbuf[3] = 4;
1207 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1208 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1209 
1210 	/* finally, adjust the packet header length */
1211 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1212 
1213 	return (0);
1214 #undef JUMBOOPTLEN
1215 }
1216 
1217 /*
1218  * Insert fragment header and copy unfragmentable header portions.
1219  */
1220 static int
1221 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1222     struct ip6_frag **frghdrp)
1223 {
1224 	struct mbuf *n, *mlast;
1225 
1226 	if (hlen > sizeof(struct ip6_hdr)) {
1227 		n = m_copym(m0, sizeof(struct ip6_hdr),
1228 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1229 		if (n == 0)
1230 			return (ENOBUFS);
1231 		m->m_next = n;
1232 	} else
1233 		n = m;
1234 
1235 	/* Search for the last mbuf of unfragmentable part. */
1236 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1237 		;
1238 
1239 	if ((mlast->m_flags & M_EXT) == 0 &&
1240 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1241 		/* use the trailing space of the last mbuf for the fragment hdr */
1242 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1243 		    mlast->m_len);
1244 		mlast->m_len += sizeof(struct ip6_frag);
1245 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1246 	} else {
1247 		/* allocate a new mbuf for the fragment header */
1248 		struct mbuf *mfrg;
1249 
1250 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1251 		if (mfrg == 0)
1252 			return (ENOBUFS);
1253 		mfrg->m_len = sizeof(struct ip6_frag);
1254 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1255 		mlast->m_next = mfrg;
1256 	}
1257 
1258 	return (0);
1259 }
1260 
1261 static int
1262 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1263     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1264     int *alwaysfragp)
1265 {
1266 	u_int32_t mtu = 0;
1267 	int alwaysfrag = 0;
1268 	int error = 0;
1269 
1270 	if (ro_pmtu != ro) {
1271 		/* The first hop and the final destination may differ. */
1272 		struct sockaddr_in6 *sa6_dst =
1273 		    (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1274 		if (ro_pmtu->ro_rt &&
1275 		    ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1276 		     !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1277 			RTFREE(ro_pmtu->ro_rt);
1278 			ro_pmtu->ro_rt = (struct rtentry *)NULL;
1279 		}
1280 		if (ro_pmtu->ro_rt == NULL) {
1281 			bzero(sa6_dst, sizeof(*sa6_dst));
1282 			sa6_dst->sin6_family = AF_INET6;
1283 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1284 			sa6_dst->sin6_addr = *dst;
1285 
1286 			rtalloc((struct route *)ro_pmtu);
1287 		}
1288 	}
1289 	if (ro_pmtu->ro_rt) {
1290 		u_int32_t ifmtu;
1291 		struct in_conninfo inc;
1292 
1293 		bzero(&inc, sizeof(inc));
1294 		inc.inc_flags = 1; /* IPv6 */
1295 		inc.inc6_faddr = *dst;
1296 
1297 		if (ifp == NULL)
1298 			ifp = ro_pmtu->ro_rt->rt_ifp;
1299 		ifmtu = IN6_LINKMTU(ifp);
1300 		mtu = tcp_hc_getmtu(&inc);
1301 		if (mtu)
1302 			mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu);
1303 		else
1304 			mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1305 		if (mtu == 0)
1306 			mtu = ifmtu;
1307 		else if (mtu < IPV6_MMTU) {
1308 			/*
1309 			 * RFC2460 section 5, last paragraph:
1310 			 * if we record ICMPv6 too big message with
1311 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1312 			 * or smaller, with framgent header attached.
1313 			 * (fragment header is needed regardless from the
1314 			 * packet size, for translators to identify packets)
1315 			 */
1316 			alwaysfrag = 1;
1317 			mtu = IPV6_MMTU;
1318 		} else if (mtu > ifmtu) {
1319 			/*
1320 			 * The MTU on the route is larger than the MTU on
1321 			 * the interface!  This shouldn't happen, unless the
1322 			 * MTU of the interface has been changed after the
1323 			 * interface was brought up.  Change the MTU in the
1324 			 * route to match the interface MTU (as long as the
1325 			 * field isn't locked).
1326 			 */
1327 			mtu = ifmtu;
1328 			ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1329 		}
1330 	} else if (ifp) {
1331 		mtu = IN6_LINKMTU(ifp);
1332 	} else
1333 		error = EHOSTUNREACH; /* XXX */
1334 
1335 	*mtup = mtu;
1336 	if (alwaysfragp)
1337 		*alwaysfragp = alwaysfrag;
1338 	return (error);
1339 }
1340 
1341 /*
1342  * IP6 socket option processing.
1343  */
1344 int
1345 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1346 {
1347 	int privileged, optdatalen, uproto;
1348 	void *optdata;
1349 	struct inpcb *in6p = sotoinpcb(so);
1350 	int error, optval;
1351 	int level, op, optname;
1352 	int optlen;
1353 	struct thread *td;
1354 
1355 	if (sopt) {
1356 		level = sopt->sopt_level;
1357 		op = sopt->sopt_dir;
1358 		optname = sopt->sopt_name;
1359 		optlen = sopt->sopt_valsize;
1360 		td = sopt->sopt_td;
1361 	} else {
1362 		panic("ip6_ctloutput: arg soopt is NULL");
1363 	}
1364 	error = optval = 0;
1365 
1366 	privileged = (td == 0 || suser(td)) ? 0 : 1;
1367 	uproto = (int)so->so_proto->pr_protocol;
1368 
1369 	if (level == IPPROTO_IPV6) {
1370 		switch (op) {
1371 
1372 		case SOPT_SET:
1373 			switch (optname) {
1374 			case IPV6_2292PKTOPTIONS:
1375 #ifdef IPV6_PKTOPTIONS
1376 			case IPV6_PKTOPTIONS:
1377 #endif
1378 			{
1379 				struct mbuf *m;
1380 
1381 				error = soopt_getm(sopt, &m); /* XXX */
1382 				if (error != 0)
1383 					break;
1384 				error = soopt_mcopyin(sopt, m); /* XXX */
1385 				if (error != 0)
1386 					break;
1387 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1388 						    m, so, sopt);
1389 				m_freem(m); /* XXX */
1390 				break;
1391 			}
1392 
1393 			/*
1394 			 * Use of some Hop-by-Hop options or some
1395 			 * Destination options, might require special
1396 			 * privilege.  That is, normal applications
1397 			 * (without special privilege) might be forbidden
1398 			 * from setting certain options in outgoing packets,
1399 			 * and might never see certain options in received
1400 			 * packets. [RFC 2292 Section 6]
1401 			 * KAME specific note:
1402 			 *  KAME prevents non-privileged users from sending or
1403 			 *  receiving ANY hbh/dst options in order to avoid
1404 			 *  overhead of parsing options in the kernel.
1405 			 */
1406 			case IPV6_RECVHOPOPTS:
1407 			case IPV6_RECVDSTOPTS:
1408 			case IPV6_RECVRTHDRDSTOPTS:
1409 				if (!privileged) {
1410 					error = EPERM;
1411 					break;
1412 				}
1413 				/* FALLTHROUGH */
1414 			case IPV6_UNICAST_HOPS:
1415 			case IPV6_HOPLIMIT:
1416 			case IPV6_FAITH:
1417 
1418 			case IPV6_RECVPKTINFO:
1419 			case IPV6_RECVHOPLIMIT:
1420 			case IPV6_RECVRTHDR:
1421 			case IPV6_RECVPATHMTU:
1422 			case IPV6_RECVTCLASS:
1423 			case IPV6_V6ONLY:
1424 			case IPV6_AUTOFLOWLABEL:
1425 				if (optlen != sizeof(int)) {
1426 					error = EINVAL;
1427 					break;
1428 				}
1429 				error = sooptcopyin(sopt, &optval,
1430 					sizeof optval, sizeof optval);
1431 				if (error)
1432 					break;
1433 				switch (optname) {
1434 
1435 				case IPV6_UNICAST_HOPS:
1436 					if (optval < -1 || optval >= 256)
1437 						error = EINVAL;
1438 					else {
1439 						/* -1 = kernel default */
1440 						in6p->in6p_hops = optval;
1441 						if ((in6p->in6p_vflag &
1442 						     INP_IPV4) != 0)
1443 							in6p->inp_ip_ttl = optval;
1444 					}
1445 					break;
1446 #define OPTSET(bit) \
1447 do { \
1448 	if (optval) \
1449 		in6p->in6p_flags |= (bit); \
1450 	else \
1451 		in6p->in6p_flags &= ~(bit); \
1452 } while (/*CONSTCOND*/ 0)
1453 #define OPTSET2292(bit) \
1454 do { \
1455 	in6p->in6p_flags |= IN6P_RFC2292; \
1456 	if (optval) \
1457 		in6p->in6p_flags |= (bit); \
1458 	else \
1459 		in6p->in6p_flags &= ~(bit); \
1460 } while (/*CONSTCOND*/ 0)
1461 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1462 
1463 				case IPV6_RECVPKTINFO:
1464 					/* cannot mix with RFC2292 */
1465 					if (OPTBIT(IN6P_RFC2292)) {
1466 						error = EINVAL;
1467 						break;
1468 					}
1469 					OPTSET(IN6P_PKTINFO);
1470 					break;
1471 
1472 				case IPV6_HOPLIMIT:
1473 				{
1474 					struct ip6_pktopts **optp;
1475 
1476 					/* cannot mix with RFC2292 */
1477 					if (OPTBIT(IN6P_RFC2292)) {
1478 						error = EINVAL;
1479 						break;
1480 					}
1481 					optp = &in6p->in6p_outputopts;
1482 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1483 							   (u_char *)&optval,
1484 							   sizeof(optval),
1485 							   optp,
1486 							   privileged, uproto);
1487 					break;
1488 				}
1489 
1490 				case IPV6_RECVHOPLIMIT:
1491 					/* cannot mix with RFC2292 */
1492 					if (OPTBIT(IN6P_RFC2292)) {
1493 						error = EINVAL;
1494 						break;
1495 					}
1496 					OPTSET(IN6P_HOPLIMIT);
1497 					break;
1498 
1499 				case IPV6_RECVHOPOPTS:
1500 					/* cannot mix with RFC2292 */
1501 					if (OPTBIT(IN6P_RFC2292)) {
1502 						error = EINVAL;
1503 						break;
1504 					}
1505 					OPTSET(IN6P_HOPOPTS);
1506 					break;
1507 
1508 				case IPV6_RECVDSTOPTS:
1509 					/* cannot mix with RFC2292 */
1510 					if (OPTBIT(IN6P_RFC2292)) {
1511 						error = EINVAL;
1512 						break;
1513 					}
1514 					OPTSET(IN6P_DSTOPTS);
1515 					break;
1516 
1517 				case IPV6_RECVRTHDRDSTOPTS:
1518 					/* cannot mix with RFC2292 */
1519 					if (OPTBIT(IN6P_RFC2292)) {
1520 						error = EINVAL;
1521 						break;
1522 					}
1523 					OPTSET(IN6P_RTHDRDSTOPTS);
1524 					break;
1525 
1526 				case IPV6_RECVRTHDR:
1527 					/* cannot mix with RFC2292 */
1528 					if (OPTBIT(IN6P_RFC2292)) {
1529 						error = EINVAL;
1530 						break;
1531 					}
1532 					OPTSET(IN6P_RTHDR);
1533 					break;
1534 
1535 				case IPV6_FAITH:
1536 					OPTSET(IN6P_FAITH);
1537 					break;
1538 
1539 				case IPV6_RECVPATHMTU:
1540 					/*
1541 					 * We ignore this option for TCP
1542 					 * sockets.
1543 					 * (RFC3542 leaves this case
1544 					 * unspecified.)
1545 					 */
1546 					if (uproto != IPPROTO_TCP)
1547 						OPTSET(IN6P_MTU);
1548 					break;
1549 
1550 				case IPV6_V6ONLY:
1551 					/*
1552 					 * make setsockopt(IPV6_V6ONLY)
1553 					 * available only prior to bind(2).
1554 					 * see ipng mailing list, Jun 22 2001.
1555 					 */
1556 					if (in6p->in6p_lport ||
1557 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1558 						error = EINVAL;
1559 						break;
1560 					}
1561 					OPTSET(IN6P_IPV6_V6ONLY);
1562 					if (optval)
1563 						in6p->in6p_vflag &= ~INP_IPV4;
1564 					else
1565 						in6p->in6p_vflag |= INP_IPV4;
1566 					break;
1567 				case IPV6_RECVTCLASS:
1568 					/* cannot mix with RFC2292 XXX */
1569 					if (OPTBIT(IN6P_RFC2292)) {
1570 						error = EINVAL;
1571 						break;
1572 					}
1573 					OPTSET(IN6P_TCLASS);
1574 					break;
1575 				case IPV6_AUTOFLOWLABEL:
1576 					OPTSET(IN6P_AUTOFLOWLABEL);
1577 					break;
1578 
1579 				}
1580 				break;
1581 
1582 			case IPV6_TCLASS:
1583 			case IPV6_DONTFRAG:
1584 			case IPV6_USE_MIN_MTU:
1585 			case IPV6_PREFER_TEMPADDR:
1586 				if (optlen != sizeof(optval)) {
1587 					error = EINVAL;
1588 					break;
1589 				}
1590 				error = sooptcopyin(sopt, &optval,
1591 					sizeof optval, sizeof optval);
1592 				if (error)
1593 					break;
1594 				{
1595 					struct ip6_pktopts **optp;
1596 					optp = &in6p->in6p_outputopts;
1597 					error = ip6_pcbopt(optname,
1598 							   (u_char *)&optval,
1599 							   sizeof(optval),
1600 							   optp,
1601 							   privileged, uproto);
1602 					break;
1603 				}
1604 
1605 			case IPV6_2292PKTINFO:
1606 			case IPV6_2292HOPLIMIT:
1607 			case IPV6_2292HOPOPTS:
1608 			case IPV6_2292DSTOPTS:
1609 			case IPV6_2292RTHDR:
1610 				/* RFC 2292 */
1611 				if (optlen != sizeof(int)) {
1612 					error = EINVAL;
1613 					break;
1614 				}
1615 				error = sooptcopyin(sopt, &optval,
1616 					sizeof optval, sizeof optval);
1617 				if (error)
1618 					break;
1619 				switch (optname) {
1620 				case IPV6_2292PKTINFO:
1621 					OPTSET2292(IN6P_PKTINFO);
1622 					break;
1623 				case IPV6_2292HOPLIMIT:
1624 					OPTSET2292(IN6P_HOPLIMIT);
1625 					break;
1626 				case IPV6_2292HOPOPTS:
1627 					/*
1628 					 * Check super-user privilege.
1629 					 * See comments for IPV6_RECVHOPOPTS.
1630 					 */
1631 					if (!privileged)
1632 						return (EPERM);
1633 					OPTSET2292(IN6P_HOPOPTS);
1634 					break;
1635 				case IPV6_2292DSTOPTS:
1636 					if (!privileged)
1637 						return (EPERM);
1638 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1639 					break;
1640 				case IPV6_2292RTHDR:
1641 					OPTSET2292(IN6P_RTHDR);
1642 					break;
1643 				}
1644 				break;
1645 			case IPV6_PKTINFO:
1646 			case IPV6_HOPOPTS:
1647 			case IPV6_RTHDR:
1648 			case IPV6_DSTOPTS:
1649 			case IPV6_RTHDRDSTOPTS:
1650 			case IPV6_NEXTHOP:
1651 			{
1652 				/* new advanced API (RFC3542) */
1653 				u_char *optbuf;
1654 				u_char optbuf_storage[MCLBYTES];
1655 				int optlen;
1656 				struct ip6_pktopts **optp;
1657 
1658 				/* cannot mix with RFC2292 */
1659 				if (OPTBIT(IN6P_RFC2292)) {
1660 					error = EINVAL;
1661 					break;
1662 				}
1663 
1664 				/*
1665 				 * We only ensure valsize is not too large
1666 				 * here.  Further validation will be done
1667 				 * later.
1668 				 */
1669 				error = sooptcopyin(sopt, optbuf_storage,
1670 				    sizeof(optbuf_storage), 0);
1671 				if (error)
1672 					break;
1673 				optlen = sopt->sopt_valsize;
1674 				optbuf = optbuf_storage;
1675 				optp = &in6p->in6p_outputopts;
1676 				error = ip6_pcbopt(optname,
1677 						   optbuf, optlen,
1678 						   optp, privileged, uproto);
1679 				break;
1680 			}
1681 #undef OPTSET
1682 
1683 			case IPV6_MULTICAST_IF:
1684 			case IPV6_MULTICAST_HOPS:
1685 			case IPV6_MULTICAST_LOOP:
1686 			case IPV6_JOIN_GROUP:
1687 			case IPV6_LEAVE_GROUP:
1688 			    {
1689 				if (sopt->sopt_valsize > MLEN) {
1690 					error = EMSGSIZE;
1691 					break;
1692 				}
1693 				/* XXX */
1694 			    }
1695 			    /* FALLTHROUGH */
1696 			    {
1697 				struct mbuf *m;
1698 
1699 				if (sopt->sopt_valsize > MCLBYTES) {
1700 					error = EMSGSIZE;
1701 					break;
1702 				}
1703 				/* XXX */
1704 				MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA);
1705 				if (m == 0) {
1706 					error = ENOBUFS;
1707 					break;
1708 				}
1709 				if (sopt->sopt_valsize > MLEN) {
1710 					MCLGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT);
1711 					if ((m->m_flags & M_EXT) == 0) {
1712 						m_free(m);
1713 						error = ENOBUFS;
1714 						break;
1715 					}
1716 				}
1717 				m->m_len = sopt->sopt_valsize;
1718 				error = sooptcopyin(sopt, mtod(m, char *),
1719 						    m->m_len, m->m_len);
1720 				if (error) {
1721 					(void)m_free(m);
1722 					break;
1723 				}
1724 				error =	ip6_setmoptions(sopt->sopt_name,
1725 							&in6p->in6p_moptions,
1726 							m);
1727 				(void)m_free(m);
1728 			    }
1729 				break;
1730 
1731 			case IPV6_PORTRANGE:
1732 				error = sooptcopyin(sopt, &optval,
1733 				    sizeof optval, sizeof optval);
1734 				if (error)
1735 					break;
1736 
1737 				switch (optval) {
1738 				case IPV6_PORTRANGE_DEFAULT:
1739 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1740 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1741 					break;
1742 
1743 				case IPV6_PORTRANGE_HIGH:
1744 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1745 					in6p->in6p_flags |= IN6P_HIGHPORT;
1746 					break;
1747 
1748 				case IPV6_PORTRANGE_LOW:
1749 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1750 					in6p->in6p_flags |= IN6P_LOWPORT;
1751 					break;
1752 
1753 				default:
1754 					error = EINVAL;
1755 					break;
1756 				}
1757 				break;
1758 
1759 #ifdef IPSEC
1760 			case IPV6_IPSEC_POLICY:
1761 			    {
1762 				caddr_t req = NULL;
1763 				size_t len = 0;
1764 				struct mbuf *m;
1765 
1766 				if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1767 					break;
1768 				if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1769 					break;
1770 				if (m) {
1771 					req = mtod(m, caddr_t);
1772 					len = m->m_len;
1773 				}
1774 				error = ipsec6_set_policy(in6p, optname, req,
1775 							  len, privileged);
1776 				m_freem(m);
1777 			    }
1778 				break;
1779 #endif /* IPSEC */
1780 
1781 			default:
1782 				error = ENOPROTOOPT;
1783 				break;
1784 			}
1785 			break;
1786 
1787 		case SOPT_GET:
1788 			switch (optname) {
1789 
1790 			case IPV6_2292PKTOPTIONS:
1791 #ifdef IPV6_PKTOPTIONS
1792 			case IPV6_PKTOPTIONS:
1793 #endif
1794 				/*
1795 				 * RFC3542 (effectively) deprecated the
1796 				 * semantics of the 2292-style pktoptions.
1797 				 * Since it was not reliable in nature (i.e.,
1798 				 * applications had to expect the lack of some
1799 				 * information after all), it would make sense
1800 				 * to simplify this part by always returning
1801 				 * empty data.
1802 				 */
1803 				sopt->sopt_valsize = 0;
1804 				break;
1805 
1806 			case IPV6_RECVHOPOPTS:
1807 			case IPV6_RECVDSTOPTS:
1808 			case IPV6_RECVRTHDRDSTOPTS:
1809 			case IPV6_UNICAST_HOPS:
1810 			case IPV6_RECVPKTINFO:
1811 			case IPV6_RECVHOPLIMIT:
1812 			case IPV6_RECVRTHDR:
1813 			case IPV6_RECVPATHMTU:
1814 
1815 			case IPV6_FAITH:
1816 			case IPV6_V6ONLY:
1817 			case IPV6_PORTRANGE:
1818 			case IPV6_RECVTCLASS:
1819 			case IPV6_AUTOFLOWLABEL:
1820 				switch (optname) {
1821 
1822 				case IPV6_RECVHOPOPTS:
1823 					optval = OPTBIT(IN6P_HOPOPTS);
1824 					break;
1825 
1826 				case IPV6_RECVDSTOPTS:
1827 					optval = OPTBIT(IN6P_DSTOPTS);
1828 					break;
1829 
1830 				case IPV6_RECVRTHDRDSTOPTS:
1831 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1832 					break;
1833 
1834 				case IPV6_UNICAST_HOPS:
1835 					optval = in6p->in6p_hops;
1836 					break;
1837 
1838 				case IPV6_RECVPKTINFO:
1839 					optval = OPTBIT(IN6P_PKTINFO);
1840 					break;
1841 
1842 				case IPV6_RECVHOPLIMIT:
1843 					optval = OPTBIT(IN6P_HOPLIMIT);
1844 					break;
1845 
1846 				case IPV6_RECVRTHDR:
1847 					optval = OPTBIT(IN6P_RTHDR);
1848 					break;
1849 
1850 				case IPV6_RECVPATHMTU:
1851 					optval = OPTBIT(IN6P_MTU);
1852 					break;
1853 
1854 				case IPV6_FAITH:
1855 					optval = OPTBIT(IN6P_FAITH);
1856 					break;
1857 
1858 				case IPV6_V6ONLY:
1859 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
1860 					break;
1861 
1862 				case IPV6_PORTRANGE:
1863 				    {
1864 					int flags;
1865 					flags = in6p->in6p_flags;
1866 					if (flags & IN6P_HIGHPORT)
1867 						optval = IPV6_PORTRANGE_HIGH;
1868 					else if (flags & IN6P_LOWPORT)
1869 						optval = IPV6_PORTRANGE_LOW;
1870 					else
1871 						optval = 0;
1872 					break;
1873 				    }
1874 				case IPV6_RECVTCLASS:
1875 					optval = OPTBIT(IN6P_TCLASS);
1876 					break;
1877 
1878 				case IPV6_AUTOFLOWLABEL:
1879 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1880 					break;
1881 				}
1882 				if (error)
1883 					break;
1884 				error = sooptcopyout(sopt, &optval,
1885 					sizeof optval);
1886 				break;
1887 
1888 			case IPV6_PATHMTU:
1889 			{
1890 				u_long pmtu = 0;
1891 				struct ip6_mtuinfo mtuinfo;
1892 				struct route_in6 sro;
1893 
1894 				bzero(&sro, sizeof(sro));
1895 
1896 				if (!(so->so_state & SS_ISCONNECTED))
1897 					return (ENOTCONN);
1898 				/*
1899 				 * XXX: we dot not consider the case of source
1900 				 * routing, or optional information to specify
1901 				 * the outgoing interface.
1902 				 */
1903 				error = ip6_getpmtu(&sro, NULL, NULL,
1904 				    &in6p->in6p_faddr, &pmtu, NULL);
1905 				if (sro.ro_rt)
1906 					RTFREE(sro.ro_rt);
1907 				if (error)
1908 					break;
1909 				if (pmtu > IPV6_MAXPACKET)
1910 					pmtu = IPV6_MAXPACKET;
1911 
1912 				bzero(&mtuinfo, sizeof(mtuinfo));
1913 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1914 				optdata = (void *)&mtuinfo;
1915 				optdatalen = sizeof(mtuinfo);
1916 				error = sooptcopyout(sopt, optdata,
1917 				    optdatalen);
1918 				break;
1919 			}
1920 
1921 			case IPV6_2292PKTINFO:
1922 			case IPV6_2292HOPLIMIT:
1923 			case IPV6_2292HOPOPTS:
1924 			case IPV6_2292RTHDR:
1925 			case IPV6_2292DSTOPTS:
1926 				switch (optname) {
1927 				case IPV6_2292PKTINFO:
1928 					optval = OPTBIT(IN6P_PKTINFO);
1929 					break;
1930 				case IPV6_2292HOPLIMIT:
1931 					optval = OPTBIT(IN6P_HOPLIMIT);
1932 					break;
1933 				case IPV6_2292HOPOPTS:
1934 					optval = OPTBIT(IN6P_HOPOPTS);
1935 					break;
1936 				case IPV6_2292RTHDR:
1937 					optval = OPTBIT(IN6P_RTHDR);
1938 					break;
1939 				case IPV6_2292DSTOPTS:
1940 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1941 					break;
1942 				}
1943 				error = sooptcopyout(sopt, &optval,
1944 				    sizeof optval);
1945 				break;
1946 			case IPV6_PKTINFO:
1947 			case IPV6_HOPOPTS:
1948 			case IPV6_RTHDR:
1949 			case IPV6_DSTOPTS:
1950 			case IPV6_RTHDRDSTOPTS:
1951 			case IPV6_NEXTHOP:
1952 			case IPV6_TCLASS:
1953 			case IPV6_DONTFRAG:
1954 			case IPV6_USE_MIN_MTU:
1955 			case IPV6_PREFER_TEMPADDR:
1956 				error = ip6_getpcbopt(in6p->in6p_outputopts,
1957 				    optname, sopt);
1958 				break;
1959 
1960 			case IPV6_MULTICAST_IF:
1961 			case IPV6_MULTICAST_HOPS:
1962 			case IPV6_MULTICAST_LOOP:
1963 			case IPV6_JOIN_GROUP:
1964 			case IPV6_LEAVE_GROUP:
1965 			    {
1966 				struct mbuf *m;
1967 				error = ip6_getmoptions(sopt->sopt_name,
1968 				    in6p->in6p_moptions, &m);
1969 				if (error == 0)
1970 					error = sooptcopyout(sopt,
1971 					    mtod(m, char *), m->m_len);
1972 				m_freem(m);
1973 			    }
1974 				break;
1975 
1976 #ifdef IPSEC
1977 			case IPV6_IPSEC_POLICY:
1978 			  {
1979 				caddr_t req = NULL;
1980 				size_t len = 0;
1981 				struct mbuf *m = NULL;
1982 				struct mbuf **mp = &m;
1983 				size_t ovalsize = sopt->sopt_valsize;
1984 				caddr_t oval = (caddr_t)sopt->sopt_val;
1985 
1986 				error = soopt_getm(sopt, &m); /* XXX */
1987 				if (error != 0)
1988 					break;
1989 				error = soopt_mcopyin(sopt, m); /* XXX */
1990 				if (error != 0)
1991 					break;
1992 				sopt->sopt_valsize = ovalsize;
1993 				sopt->sopt_val = oval;
1994 				if (m) {
1995 					req = mtod(m, caddr_t);
1996 					len = m->m_len;
1997 				}
1998 				error = ipsec6_get_policy(in6p, req, len, mp);
1999 				if (error == 0)
2000 					error = soopt_mcopyout(sopt, m); /* XXX */
2001 				if (error == 0 && m)
2002 					m_freem(m);
2003 				break;
2004 			  }
2005 #endif /* IPSEC */
2006 
2007 			default:
2008 				error = ENOPROTOOPT;
2009 				break;
2010 			}
2011 			break;
2012 		}
2013 	} else {		/* level != IPPROTO_IPV6 */
2014 		error = EINVAL;
2015 	}
2016 	return (error);
2017 }
2018 
2019 int
2020 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2021 {
2022 	int error = 0, optval, optlen;
2023 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2024 	struct in6pcb *in6p = sotoin6pcb(so);
2025 	int level, op, optname;
2026 
2027 	if (sopt) {
2028 		level = sopt->sopt_level;
2029 		op = sopt->sopt_dir;
2030 		optname = sopt->sopt_name;
2031 		optlen = sopt->sopt_valsize;
2032 	} else
2033 		panic("ip6_raw_ctloutput: arg soopt is NULL");
2034 
2035 	if (level != IPPROTO_IPV6) {
2036 		return (EINVAL);
2037 	}
2038 
2039 	switch (optname) {
2040 	case IPV6_CHECKSUM:
2041 		/*
2042 		 * For ICMPv6 sockets, no modification allowed for checksum
2043 		 * offset, permit "no change" values to help existing apps.
2044 		 *
2045 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2046 		 * for an ICMPv6 socket will fail."
2047 		 * The current behavior does not meet RFC3542.
2048 		 */
2049 		switch (op) {
2050 		case SOPT_SET:
2051 			if (optlen != sizeof(int)) {
2052 				error = EINVAL;
2053 				break;
2054 			}
2055 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2056 					    sizeof(optval));
2057 			if (error)
2058 				break;
2059 			if ((optval % 2) != 0) {
2060 				/* the API assumes even offset values */
2061 				error = EINVAL;
2062 			} else if (so->so_proto->pr_protocol ==
2063 			    IPPROTO_ICMPV6) {
2064 				if (optval != icmp6off)
2065 					error = EINVAL;
2066 			} else
2067 				in6p->in6p_cksum = optval;
2068 			break;
2069 
2070 		case SOPT_GET:
2071 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2072 				optval = icmp6off;
2073 			else
2074 				optval = in6p->in6p_cksum;
2075 
2076 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2077 			break;
2078 
2079 		default:
2080 			error = EINVAL;
2081 			break;
2082 		}
2083 		break;
2084 
2085 	default:
2086 		error = ENOPROTOOPT;
2087 		break;
2088 	}
2089 
2090 	return (error);
2091 }
2092 
2093 /*
2094  * Set up IP6 options in pcb for insertion in output packets or
2095  * specifying behavior of outgoing packets.
2096  */
2097 static int
2098 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2099     struct socket *so, struct sockopt *sopt)
2100 {
2101 	struct ip6_pktopts *opt = *pktopt;
2102 	int error = 0;
2103 	struct thread *td = sopt->sopt_td;
2104 	int priv = 0;
2105 
2106 	/* turn off any old options. */
2107 	if (opt) {
2108 #ifdef DIAGNOSTIC
2109 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2110 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2111 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2112 			printf("ip6_pcbopts: all specified options are cleared.\n");
2113 #endif
2114 		ip6_clearpktopts(opt, -1);
2115 	} else
2116 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2117 	*pktopt = NULL;
2118 
2119 	if (!m || m->m_len == 0) {
2120 		/*
2121 		 * Only turning off any previous options, regardless of
2122 		 * whether the opt is just created or given.
2123 		 */
2124 		free(opt, M_IP6OPT);
2125 		return (0);
2126 	}
2127 
2128 	/*  set options specified by user. */
2129 	if (td && !suser(td))
2130 		priv = 1;
2131 	if ((error = ip6_setpktopts(m, opt, NULL, priv,
2132 	    so->so_proto->pr_protocol)) != 0) {
2133 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2134 		free(opt, M_IP6OPT);
2135 		return (error);
2136 	}
2137 	*pktopt = opt;
2138 	return (0);
2139 }
2140 
2141 /*
2142  * initialize ip6_pktopts.  beware that there are non-zero default values in
2143  * the struct.
2144  */
2145 void
2146 ip6_initpktopts(struct ip6_pktopts *opt)
2147 {
2148 
2149 	bzero(opt, sizeof(*opt));
2150 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2151 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2152 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2153 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2154 }
2155 
2156 static int
2157 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2158     int priv, int uproto)
2159 {
2160 	struct ip6_pktopts *opt;
2161 
2162 	if (*pktopt == NULL) {
2163 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2164 		    M_WAITOK);
2165 		ip6_initpktopts(*pktopt);
2166 	}
2167 	opt = *pktopt;
2168 
2169 	return (ip6_setpktopt(optname, buf, len, opt, priv, 1, 0, uproto));
2170 }
2171 
2172 static int
2173 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2174 {
2175 	void *optdata = NULL;
2176 	int optdatalen = 0;
2177 	struct ip6_ext *ip6e;
2178 	int error = 0;
2179 	struct in6_pktinfo null_pktinfo;
2180 	int deftclass = 0, on;
2181 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2182 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2183 
2184 	switch (optname) {
2185 	case IPV6_PKTINFO:
2186 		if (pktopt && pktopt->ip6po_pktinfo)
2187 			optdata = (void *)pktopt->ip6po_pktinfo;
2188 		else {
2189 			/* XXX: we don't have to do this every time... */
2190 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2191 			optdata = (void *)&null_pktinfo;
2192 		}
2193 		optdatalen = sizeof(struct in6_pktinfo);
2194 		break;
2195 	case IPV6_TCLASS:
2196 		if (pktopt && pktopt->ip6po_tclass >= 0)
2197 			optdata = (void *)&pktopt->ip6po_tclass;
2198 		else
2199 			optdata = (void *)&deftclass;
2200 		optdatalen = sizeof(int);
2201 		break;
2202 	case IPV6_HOPOPTS:
2203 		if (pktopt && pktopt->ip6po_hbh) {
2204 			optdata = (void *)pktopt->ip6po_hbh;
2205 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2206 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2207 		}
2208 		break;
2209 	case IPV6_RTHDR:
2210 		if (pktopt && pktopt->ip6po_rthdr) {
2211 			optdata = (void *)pktopt->ip6po_rthdr;
2212 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2213 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2214 		}
2215 		break;
2216 	case IPV6_RTHDRDSTOPTS:
2217 		if (pktopt && pktopt->ip6po_dest1) {
2218 			optdata = (void *)pktopt->ip6po_dest1;
2219 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2220 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2221 		}
2222 		break;
2223 	case IPV6_DSTOPTS:
2224 		if (pktopt && pktopt->ip6po_dest2) {
2225 			optdata = (void *)pktopt->ip6po_dest2;
2226 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2227 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2228 		}
2229 		break;
2230 	case IPV6_NEXTHOP:
2231 		if (pktopt && pktopt->ip6po_nexthop) {
2232 			optdata = (void *)pktopt->ip6po_nexthop;
2233 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2234 		}
2235 		break;
2236 	case IPV6_USE_MIN_MTU:
2237 		if (pktopt)
2238 			optdata = (void *)&pktopt->ip6po_minmtu;
2239 		else
2240 			optdata = (void *)&defminmtu;
2241 		optdatalen = sizeof(int);
2242 		break;
2243 	case IPV6_DONTFRAG:
2244 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2245 			on = 1;
2246 		else
2247 			on = 0;
2248 		optdata = (void *)&on;
2249 		optdatalen = sizeof(on);
2250 		break;
2251 	case IPV6_PREFER_TEMPADDR:
2252 		if (pktopt)
2253 			optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2254 		else
2255 			optdata = (void *)&defpreftemp;
2256 		optdatalen = sizeof(int);
2257 		break;
2258 	default:		/* should not happen */
2259 #ifdef DIAGNOSTIC
2260 		panic("ip6_getpcbopt: unexpected option\n");
2261 #endif
2262 		return (ENOPROTOOPT);
2263 	}
2264 
2265 	error = sooptcopyout(sopt, optdata, optdatalen);
2266 
2267 	return (error);
2268 }
2269 
2270 void
2271 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2272 {
2273 	if (pktopt == NULL)
2274 		return;
2275 
2276 	if (optname == -1 || optname == IPV6_PKTINFO) {
2277 		if (pktopt->ip6po_pktinfo)
2278 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2279 		pktopt->ip6po_pktinfo = NULL;
2280 	}
2281 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2282 		pktopt->ip6po_hlim = -1;
2283 	if (optname == -1 || optname == IPV6_TCLASS)
2284 		pktopt->ip6po_tclass = -1;
2285 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2286 		if (pktopt->ip6po_nextroute.ro_rt) {
2287 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2288 			pktopt->ip6po_nextroute.ro_rt = NULL;
2289 		}
2290 		if (pktopt->ip6po_nexthop)
2291 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2292 		pktopt->ip6po_nexthop = NULL;
2293 	}
2294 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2295 		if (pktopt->ip6po_hbh)
2296 			free(pktopt->ip6po_hbh, M_IP6OPT);
2297 		pktopt->ip6po_hbh = NULL;
2298 	}
2299 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2300 		if (pktopt->ip6po_dest1)
2301 			free(pktopt->ip6po_dest1, M_IP6OPT);
2302 		pktopt->ip6po_dest1 = NULL;
2303 	}
2304 	if (optname == -1 || optname == IPV6_RTHDR) {
2305 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2306 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2307 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2308 		if (pktopt->ip6po_route.ro_rt) {
2309 			RTFREE(pktopt->ip6po_route.ro_rt);
2310 			pktopt->ip6po_route.ro_rt = NULL;
2311 		}
2312 	}
2313 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2314 		if (pktopt->ip6po_dest2)
2315 			free(pktopt->ip6po_dest2, M_IP6OPT);
2316 		pktopt->ip6po_dest2 = NULL;
2317 	}
2318 }
2319 
2320 #define PKTOPT_EXTHDRCPY(type) \
2321 do {\
2322 	if (src->type) {\
2323 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2324 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2325 		if (dst->type == NULL && canwait == M_NOWAIT)\
2326 			goto bad;\
2327 		bcopy(src->type, dst->type, hlen);\
2328 	}\
2329 } while (/*CONSTCOND*/ 0)
2330 
2331 static int
2332 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2333 {
2334 	if (dst == NULL || src == NULL)  {
2335 		printf("ip6_clearpktopts: invalid argument\n");
2336 		return (EINVAL);
2337 	}
2338 
2339 	dst->ip6po_hlim = src->ip6po_hlim;
2340 	dst->ip6po_tclass = src->ip6po_tclass;
2341 	dst->ip6po_flags = src->ip6po_flags;
2342 	if (src->ip6po_pktinfo) {
2343 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2344 		    M_IP6OPT, canwait);
2345 		if (dst->ip6po_pktinfo == NULL)
2346 			goto bad;
2347 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2348 	}
2349 	if (src->ip6po_nexthop) {
2350 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2351 		    M_IP6OPT, canwait);
2352 		if (dst->ip6po_nexthop == NULL)
2353 			goto bad;
2354 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2355 		    src->ip6po_nexthop->sa_len);
2356 	}
2357 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2358 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2359 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2360 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2361 	return (0);
2362 
2363   bad:
2364 	ip6_clearpktopts(dst, -1);
2365 	return (ENOBUFS);
2366 }
2367 #undef PKTOPT_EXTHDRCPY
2368 
2369 struct ip6_pktopts *
2370 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2371 {
2372 	int error;
2373 	struct ip6_pktopts *dst;
2374 
2375 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2376 	if (dst == NULL)
2377 		return (NULL);
2378 	ip6_initpktopts(dst);
2379 
2380 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2381 		free(dst, M_IP6OPT);
2382 		return (NULL);
2383 	}
2384 
2385 	return (dst);
2386 }
2387 
2388 void
2389 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2390 {
2391 	if (pktopt == NULL)
2392 		return;
2393 
2394 	ip6_clearpktopts(pktopt, -1);
2395 
2396 	free(pktopt, M_IP6OPT);
2397 }
2398 
2399 /*
2400  * Set the IP6 multicast options in response to user setsockopt().
2401  */
2402 static int
2403 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
2404 {
2405 	int error = 0;
2406 	u_int loop, ifindex;
2407 	struct ipv6_mreq *mreq;
2408 	struct ifnet *ifp;
2409 	struct ip6_moptions *im6o = *im6op;
2410 	struct route_in6 ro;
2411 	struct in6_multi_mship *imm;
2412 	struct thread *td = curthread;
2413 
2414 	if (im6o == NULL) {
2415 		/*
2416 		 * No multicast option buffer attached to the pcb;
2417 		 * allocate one and initialize to default values.
2418 		 */
2419 		im6o = (struct ip6_moptions *)
2420 			malloc(sizeof(*im6o), M_IP6MOPTS, M_WAITOK);
2421 
2422 		if (im6o == NULL)
2423 			return (ENOBUFS);
2424 		*im6op = im6o;
2425 		im6o->im6o_multicast_ifp = NULL;
2426 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2427 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2428 		LIST_INIT(&im6o->im6o_memberships);
2429 	}
2430 
2431 	switch (optname) {
2432 
2433 	case IPV6_MULTICAST_IF:
2434 		/*
2435 		 * Select the interface for outgoing multicast packets.
2436 		 */
2437 		if (m == NULL || m->m_len != sizeof(u_int)) {
2438 			error = EINVAL;
2439 			break;
2440 		}
2441 		bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
2442 		if (ifindex < 0 || if_index < ifindex) {
2443 			error = ENXIO;	/* XXX EINVAL? */
2444 			break;
2445 		}
2446 		ifp = ifnet_byindex(ifindex);
2447 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2448 			error = EADDRNOTAVAIL;
2449 			break;
2450 		}
2451 		im6o->im6o_multicast_ifp = ifp;
2452 		break;
2453 
2454 	case IPV6_MULTICAST_HOPS:
2455 	    {
2456 		/*
2457 		 * Set the IP6 hoplimit for outgoing multicast packets.
2458 		 */
2459 		int optval;
2460 		if (m == NULL || m->m_len != sizeof(int)) {
2461 			error = EINVAL;
2462 			break;
2463 		}
2464 		bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2465 		if (optval < -1 || optval >= 256)
2466 			error = EINVAL;
2467 		else if (optval == -1)
2468 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2469 		else
2470 			im6o->im6o_multicast_hlim = optval;
2471 		break;
2472 	    }
2473 
2474 	case IPV6_MULTICAST_LOOP:
2475 		/*
2476 		 * Set the loopback flag for outgoing multicast packets.
2477 		 * Must be zero or one.
2478 		 */
2479 		if (m == NULL || m->m_len != sizeof(u_int)) {
2480 			error = EINVAL;
2481 			break;
2482 		}
2483 		bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2484 		if (loop > 1) {
2485 			error = EINVAL;
2486 			break;
2487 		}
2488 		im6o->im6o_multicast_loop = loop;
2489 		break;
2490 
2491 	case IPV6_JOIN_GROUP:
2492 		/*
2493 		 * Add a multicast group membership.
2494 		 * Group must be a valid IP6 multicast address.
2495 		 */
2496 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2497 			error = EINVAL;
2498 			break;
2499 		}
2500 		mreq = mtod(m, struct ipv6_mreq *);
2501 
2502 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2503 			/*
2504 			 * We use the unspecified address to specify to accept
2505 			 * all multicast addresses. Only super user is allowed
2506 			 * to do this.
2507 			 */
2508 			if (suser(td)) {
2509 				error = EACCES;
2510 				break;
2511 			}
2512 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2513 			error = EINVAL;
2514 			break;
2515 		}
2516 
2517 		/*
2518 		 * If no interface was explicitly specified, choose an
2519 		 * appropriate one according to the given multicast address.
2520 		 */
2521 		if (mreq->ipv6mr_interface == 0) {
2522 			struct sockaddr_in6 *dst;
2523 
2524 			/*
2525 			 * Look up the routing table for the
2526 			 * address, and choose the outgoing interface.
2527 			 *   XXX: is it a good approach?
2528 			 */
2529 			ro.ro_rt = NULL;
2530 			dst = (struct sockaddr_in6 *)&ro.ro_dst;
2531 			bzero(dst, sizeof(*dst));
2532 			dst->sin6_family = AF_INET6;
2533 			dst->sin6_len = sizeof(*dst);
2534 			dst->sin6_addr = mreq->ipv6mr_multiaddr;
2535 			rtalloc((struct route *)&ro);
2536 			if (ro.ro_rt == NULL) {
2537 				error = EADDRNOTAVAIL;
2538 				break;
2539 			}
2540 			ifp = ro.ro_rt->rt_ifp;
2541 			RTFREE(ro.ro_rt);
2542 		} else {
2543 			/*
2544 			 * If the interface is specified, validate it.
2545 			 */
2546 			if (mreq->ipv6mr_interface < 0 ||
2547 			    if_index < mreq->ipv6mr_interface) {
2548 				error = ENXIO;	/* XXX EINVAL? */
2549 				break;
2550 			}
2551 			ifp = ifnet_byindex(mreq->ipv6mr_interface);
2552 			if (!ifp) {
2553 				error = ENXIO;	/* XXX EINVAL? */
2554 				break;
2555 			}
2556 		}
2557 
2558 		/*
2559 		 * See if we found an interface, and confirm that it
2560 		 * supports multicast
2561 		 */
2562 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2563 			error = EADDRNOTAVAIL;
2564 			break;
2565 		}
2566 
2567 		if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) {
2568 			error = EADDRNOTAVAIL; /* XXX: should not happen */
2569 			break;
2570 		}
2571 
2572 		/*
2573 		 * See if the membership already exists.
2574 		 */
2575 		for (imm = im6o->im6o_memberships.lh_first;
2576 		     imm != NULL; imm = imm->i6mm_chain.le_next)
2577 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
2578 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2579 					       &mreq->ipv6mr_multiaddr))
2580 				break;
2581 		if (imm != NULL) {
2582 			error = EADDRINUSE;
2583 			break;
2584 		}
2585 		/*
2586 		 * Everything looks good; add a new record to the multicast
2587 		 * address list for the given interface.
2588 		 */
2589 		imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr,  &error, 0);
2590 		if (imm == NULL)
2591 			break;
2592 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2593 		break;
2594 
2595 	case IPV6_LEAVE_GROUP:
2596 		/*
2597 		 * Drop a multicast group membership.
2598 		 * Group must be a valid IP6 multicast address.
2599 		 */
2600 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2601 			error = EINVAL;
2602 			break;
2603 		}
2604 		mreq = mtod(m, struct ipv6_mreq *);
2605 
2606 		/*
2607 		 * If an interface address was specified, get a pointer
2608 		 * to its ifnet structure.
2609 		 */
2610 		if (mreq->ipv6mr_interface < 0 ||
2611 		    if_index < mreq->ipv6mr_interface) {
2612 			error = ENXIO;	/* XXX EINVAL? */
2613 			break;
2614 		}
2615 		if (mreq->ipv6mr_interface == 0)
2616 			ifp = NULL;
2617 		else
2618 			ifp = ifnet_byindex(mreq->ipv6mr_interface);
2619 
2620 		/* Fill in the scope zone ID */
2621 		if (ifp) {
2622 			if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) {
2623 				/* XXX: should not happen */
2624 				error = EADDRNOTAVAIL;
2625 				break;
2626 			}
2627 		} else if (mreq->ipv6mr_interface != 0) {
2628 			/*
2629 			 * This case happens when the (positive) index is in
2630 			 * the valid range, but the corresponding interface has
2631 			 * been detached dynamically (XXX).
2632 			 */
2633 			error = EADDRNOTAVAIL;
2634 			break;
2635 		} else {	/* ipv6mr_interface == 0 */
2636 			struct sockaddr_in6 sa6_mc;
2637 
2638 			/*
2639 			 * The API spec says as follows:
2640 			 *  If the interface index is specified as 0, the
2641 			 *  system may choose a multicast group membership to
2642 			 *  drop by matching the multicast address only.
2643 			 * On the other hand, we cannot disambiguate the scope
2644 			 * zone unless an interface is provided.  Thus, we
2645 			 * check if there's ambiguity with the default scope
2646 			 * zone as the last resort.
2647 			 */
2648 			bzero(&sa6_mc, sizeof(sa6_mc));
2649 			sa6_mc.sin6_family = AF_INET6;
2650 			sa6_mc.sin6_len = sizeof(sa6_mc);
2651 			sa6_mc.sin6_addr = mreq->ipv6mr_multiaddr;
2652 			error = sa6_embedscope(&sa6_mc, ip6_use_defzone);
2653 			if (error != 0)
2654 				break;
2655 			mreq->ipv6mr_multiaddr = sa6_mc.sin6_addr;
2656 		}
2657 
2658 		/*
2659 		 * Find the membership in the membership list.
2660 		 */
2661 		for (imm = im6o->im6o_memberships.lh_first;
2662 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
2663 			if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2664 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2665 			    &mreq->ipv6mr_multiaddr))
2666 				break;
2667 		}
2668 		if (imm == NULL) {
2669 			/* Unable to resolve interface */
2670 			error = EADDRNOTAVAIL;
2671 			break;
2672 		}
2673 		/*
2674 		 * Give up the multicast address record to which the
2675 		 * membership points.
2676 		 */
2677 		LIST_REMOVE(imm, i6mm_chain);
2678 		in6_delmulti(imm->i6mm_maddr);
2679 		free(imm, M_IP6MADDR);
2680 		break;
2681 
2682 	default:
2683 		error = EOPNOTSUPP;
2684 		break;
2685 	}
2686 
2687 	/*
2688 	 * If all options have default values, no need to keep the mbuf.
2689 	 */
2690 	if (im6o->im6o_multicast_ifp == NULL &&
2691 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2692 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2693 	    im6o->im6o_memberships.lh_first == NULL) {
2694 		free(*im6op, M_IP6MOPTS);
2695 		*im6op = NULL;
2696 	}
2697 
2698 	return (error);
2699 }
2700 
2701 /*
2702  * Return the IP6 multicast options in response to user getsockopt().
2703  */
2704 static int
2705 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp)
2706 {
2707 	u_int *hlim, *loop, *ifindex;
2708 
2709 	*mp = m_get(M_TRYWAIT, MT_HEADER);		/* XXX */
2710 
2711 	switch (optname) {
2712 
2713 	case IPV6_MULTICAST_IF:
2714 		ifindex = mtod(*mp, u_int *);
2715 		(*mp)->m_len = sizeof(u_int);
2716 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2717 			*ifindex = 0;
2718 		else
2719 			*ifindex = im6o->im6o_multicast_ifp->if_index;
2720 		return (0);
2721 
2722 	case IPV6_MULTICAST_HOPS:
2723 		hlim = mtod(*mp, u_int *);
2724 		(*mp)->m_len = sizeof(u_int);
2725 		if (im6o == NULL)
2726 			*hlim = ip6_defmcasthlim;
2727 		else
2728 			*hlim = im6o->im6o_multicast_hlim;
2729 		return (0);
2730 
2731 	case IPV6_MULTICAST_LOOP:
2732 		loop = mtod(*mp, u_int *);
2733 		(*mp)->m_len = sizeof(u_int);
2734 		if (im6o == NULL)
2735 			*loop = ip6_defmcasthlim;
2736 		else
2737 			*loop = im6o->im6o_multicast_loop;
2738 		return (0);
2739 
2740 	default:
2741 		return (EOPNOTSUPP);
2742 	}
2743 }
2744 
2745 /*
2746  * Discard the IP6 multicast options.
2747  */
2748 void
2749 ip6_freemoptions(struct ip6_moptions *im6o)
2750 {
2751 	struct in6_multi_mship *imm;
2752 
2753 	if (im6o == NULL)
2754 		return;
2755 
2756 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2757 		LIST_REMOVE(imm, i6mm_chain);
2758 		if (imm->i6mm_maddr)
2759 			in6_delmulti(imm->i6mm_maddr);
2760 		free(imm, M_IP6MADDR);
2761 	}
2762 	free(im6o, M_IP6MOPTS);
2763 }
2764 
2765 /*
2766  * Set IPv6 outgoing packet options based on advanced API.
2767  */
2768 int
2769 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2770     struct ip6_pktopts *stickyopt, int priv, int uproto)
2771 {
2772 	struct cmsghdr *cm = 0;
2773 
2774 	if (control == NULL || opt == NULL)
2775 		return (EINVAL);
2776 
2777 	ip6_initpktopts(opt);
2778 	if (stickyopt) {
2779 		int error;
2780 
2781 		/*
2782 		 * If stickyopt is provided, make a local copy of the options
2783 		 * for this particular packet, then override them by ancillary
2784 		 * objects.
2785 		 * XXX: copypktopts() does not copy the cached route to a next
2786 		 * hop (if any).  This is not very good in terms of efficiency,
2787 		 * but we can allow this since this option should be rarely
2788 		 * used.
2789 		 */
2790 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2791 			return (error);
2792 	}
2793 
2794 	/*
2795 	 * XXX: Currently, we assume all the optional information is stored
2796 	 * in a single mbuf.
2797 	 */
2798 	if (control->m_next)
2799 		return (EINVAL);
2800 
2801 	for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2802 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2803 		int error;
2804 
2805 		if (control->m_len < CMSG_LEN(0))
2806 			return (EINVAL);
2807 
2808 		cm = mtod(control, struct cmsghdr *);
2809 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2810 			return (EINVAL);
2811 		if (cm->cmsg_level != IPPROTO_IPV6)
2812 			continue;
2813 
2814 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2815 		    cm->cmsg_len - CMSG_LEN(0), opt, priv, 0, 1, uproto);
2816 		if (error)
2817 			return (error);
2818 	}
2819 
2820 	return (0);
2821 }
2822 
2823 /*
2824  * Set a particular packet option, as a sticky option or an ancillary data
2825  * item.  "len" can be 0 only when it's a sticky option.
2826  * We have 4 cases of combination of "sticky" and "cmsg":
2827  * "sticky=0, cmsg=0": impossible
2828  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2829  * "sticky=1, cmsg=0": RFC3542 socket option
2830  * "sticky=1, cmsg=1": RFC2292 socket option
2831  */
2832 static int
2833 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2834     int priv, int sticky, int cmsg, int uproto)
2835 {
2836 	int minmtupolicy, preftemp;
2837 
2838 	if (!sticky && !cmsg) {
2839 #ifdef DIAGNOSTIC
2840 		printf("ip6_setpktopt: impossible case\n");
2841 #endif
2842 		return (EINVAL);
2843 	}
2844 
2845 	/*
2846 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2847 	 * not be specified in the context of RFC3542.  Conversely,
2848 	 * RFC3542 types should not be specified in the context of RFC2292.
2849 	 */
2850 	if (!cmsg) {
2851 		switch (optname) {
2852 		case IPV6_2292PKTINFO:
2853 		case IPV6_2292HOPLIMIT:
2854 		case IPV6_2292NEXTHOP:
2855 		case IPV6_2292HOPOPTS:
2856 		case IPV6_2292DSTOPTS:
2857 		case IPV6_2292RTHDR:
2858 		case IPV6_2292PKTOPTIONS:
2859 			return (ENOPROTOOPT);
2860 		}
2861 	}
2862 	if (sticky && cmsg) {
2863 		switch (optname) {
2864 		case IPV6_PKTINFO:
2865 		case IPV6_HOPLIMIT:
2866 		case IPV6_NEXTHOP:
2867 		case IPV6_HOPOPTS:
2868 		case IPV6_DSTOPTS:
2869 		case IPV6_RTHDRDSTOPTS:
2870 		case IPV6_RTHDR:
2871 		case IPV6_USE_MIN_MTU:
2872 		case IPV6_DONTFRAG:
2873 		case IPV6_TCLASS:
2874 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2875 			return (ENOPROTOOPT);
2876 		}
2877 	}
2878 
2879 	switch (optname) {
2880 	case IPV6_2292PKTINFO:
2881 	case IPV6_PKTINFO:
2882 	{
2883 		struct ifnet *ifp = NULL;
2884 		struct in6_pktinfo *pktinfo;
2885 
2886 		if (len != sizeof(struct in6_pktinfo))
2887 			return (EINVAL);
2888 
2889 		pktinfo = (struct in6_pktinfo *)buf;
2890 
2891 		/*
2892 		 * An application can clear any sticky IPV6_PKTINFO option by
2893 		 * doing a "regular" setsockopt with ipi6_addr being
2894 		 * in6addr_any and ipi6_ifindex being zero.
2895 		 * [RFC 3542, Section 6]
2896 		 */
2897 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2898 		    pktinfo->ipi6_ifindex == 0 &&
2899 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2900 			ip6_clearpktopts(opt, optname);
2901 			break;
2902 		}
2903 
2904 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2905 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2906 			return (EINVAL);
2907 		}
2908 
2909 		/* validate the interface index if specified. */
2910 		if (pktinfo->ipi6_ifindex > if_index ||
2911 		    pktinfo->ipi6_ifindex < 0) {
2912 			 return (ENXIO);
2913 		}
2914 		if (pktinfo->ipi6_ifindex) {
2915 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2916 			if (ifp == NULL)
2917 				return (ENXIO);
2918 		}
2919 
2920 		/*
2921 		 * We store the address anyway, and let in6_selectsrc()
2922 		 * validate the specified address.  This is because ipi6_addr
2923 		 * may not have enough information about its scope zone, and
2924 		 * we may need additional information (such as outgoing
2925 		 * interface or the scope zone of a destination address) to
2926 		 * disambiguate the scope.
2927 		 * XXX: the delay of the validation may confuse the
2928 		 * application when it is used as a sticky option.
2929 		 */
2930 		if (opt->ip6po_pktinfo == NULL) {
2931 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2932 			    M_IP6OPT, M_NOWAIT);
2933 			if (opt->ip6po_pktinfo == NULL)
2934 				return (ENOBUFS);
2935 		}
2936 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2937 		break;
2938 	}
2939 
2940 	case IPV6_2292HOPLIMIT:
2941 	case IPV6_HOPLIMIT:
2942 	{
2943 		int *hlimp;
2944 
2945 		/*
2946 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2947 		 * to simplify the ordering among hoplimit options.
2948 		 */
2949 		if (optname == IPV6_HOPLIMIT && sticky)
2950 			return (ENOPROTOOPT);
2951 
2952 		if (len != sizeof(int))
2953 			return (EINVAL);
2954 		hlimp = (int *)buf;
2955 		if (*hlimp < -1 || *hlimp > 255)
2956 			return (EINVAL);
2957 
2958 		opt->ip6po_hlim = *hlimp;
2959 		break;
2960 	}
2961 
2962 	case IPV6_TCLASS:
2963 	{
2964 		int tclass;
2965 
2966 		if (len != sizeof(int))
2967 			return (EINVAL);
2968 		tclass = *(int *)buf;
2969 		if (tclass < -1 || tclass > 255)
2970 			return (EINVAL);
2971 
2972 		opt->ip6po_tclass = tclass;
2973 		break;
2974 	}
2975 
2976 	case IPV6_2292NEXTHOP:
2977 	case IPV6_NEXTHOP:
2978 		if (!priv)
2979 			return (EPERM);
2980 
2981 		if (len == 0) {	/* just remove the option */
2982 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2983 			break;
2984 		}
2985 
2986 		/* check if cmsg_len is large enough for sa_len */
2987 		if (len < sizeof(struct sockaddr) || len < *buf)
2988 			return (EINVAL);
2989 
2990 		switch (((struct sockaddr *)buf)->sa_family) {
2991 		case AF_INET6:
2992 		{
2993 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2994 			int error;
2995 
2996 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2997 				return (EINVAL);
2998 
2999 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3000 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3001 				return (EINVAL);
3002 			}
3003 			if ((error = sa6_embedscope(sa6, ip6_use_defzone))
3004 			    != 0) {
3005 				return (error);
3006 			}
3007 			break;
3008 		}
3009 		case AF_LINK:	/* should eventually be supported */
3010 		default:
3011 			return (EAFNOSUPPORT);
3012 		}
3013 
3014 		/* turn off the previous option, then set the new option. */
3015 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
3016 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3017 		if (opt->ip6po_nexthop == NULL)
3018 			return (ENOBUFS);
3019 		bcopy(buf, opt->ip6po_nexthop, *buf);
3020 		break;
3021 
3022 	case IPV6_2292HOPOPTS:
3023 	case IPV6_HOPOPTS:
3024 	{
3025 		struct ip6_hbh *hbh;
3026 		int hbhlen;
3027 
3028 		/*
3029 		 * XXX: We don't allow a non-privileged user to set ANY HbH
3030 		 * options, since per-option restriction has too much
3031 		 * overhead.
3032 		 */
3033 		if (!priv)
3034 			return (EPERM);
3035 
3036 		if (len == 0) {
3037 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
3038 			break;	/* just remove the option */
3039 		}
3040 
3041 		/* message length validation */
3042 		if (len < sizeof(struct ip6_hbh))
3043 			return (EINVAL);
3044 		hbh = (struct ip6_hbh *)buf;
3045 		hbhlen = (hbh->ip6h_len + 1) << 3;
3046 		if (len != hbhlen)
3047 			return (EINVAL);
3048 
3049 		/* turn off the previous option, then set the new option. */
3050 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
3051 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3052 		if (opt->ip6po_hbh == NULL)
3053 			return (ENOBUFS);
3054 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
3055 
3056 		break;
3057 	}
3058 
3059 	case IPV6_2292DSTOPTS:
3060 	case IPV6_DSTOPTS:
3061 	case IPV6_RTHDRDSTOPTS:
3062 	{
3063 		struct ip6_dest *dest, **newdest = NULL;
3064 		int destlen;
3065 
3066 		if (!priv)	/* XXX: see the comment for IPV6_HOPOPTS */
3067 			return (EPERM);
3068 
3069 		if (len == 0) {
3070 			ip6_clearpktopts(opt, optname);
3071 			break;	/* just remove the option */
3072 		}
3073 
3074 		/* message length validation */
3075 		if (len < sizeof(struct ip6_dest))
3076 			return (EINVAL);
3077 		dest = (struct ip6_dest *)buf;
3078 		destlen = (dest->ip6d_len + 1) << 3;
3079 		if (len != destlen)
3080 			return (EINVAL);
3081 
3082 		/*
3083 		 * Determine the position that the destination options header
3084 		 * should be inserted; before or after the routing header.
3085 		 */
3086 		switch (optname) {
3087 		case IPV6_2292DSTOPTS:
3088 			/*
3089 			 * The old advacned API is ambiguous on this point.
3090 			 * Our approach is to determine the position based
3091 			 * according to the existence of a routing header.
3092 			 * Note, however, that this depends on the order of the
3093 			 * extension headers in the ancillary data; the 1st
3094 			 * part of the destination options header must appear
3095 			 * before the routing header in the ancillary data,
3096 			 * too.
3097 			 * RFC3542 solved the ambiguity by introducing
3098 			 * separate ancillary data or option types.
3099 			 */
3100 			if (opt->ip6po_rthdr == NULL)
3101 				newdest = &opt->ip6po_dest1;
3102 			else
3103 				newdest = &opt->ip6po_dest2;
3104 			break;
3105 		case IPV6_RTHDRDSTOPTS:
3106 			newdest = &opt->ip6po_dest1;
3107 			break;
3108 		case IPV6_DSTOPTS:
3109 			newdest = &opt->ip6po_dest2;
3110 			break;
3111 		}
3112 
3113 		/* turn off the previous option, then set the new option. */
3114 		ip6_clearpktopts(opt, optname);
3115 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3116 		if (*newdest == NULL)
3117 			return (ENOBUFS);
3118 		bcopy(dest, *newdest, destlen);
3119 
3120 		break;
3121 	}
3122 
3123 	case IPV6_2292RTHDR:
3124 	case IPV6_RTHDR:
3125 	{
3126 		struct ip6_rthdr *rth;
3127 		int rthlen;
3128 
3129 		if (len == 0) {
3130 			ip6_clearpktopts(opt, IPV6_RTHDR);
3131 			break;	/* just remove the option */
3132 		}
3133 
3134 		/* message length validation */
3135 		if (len < sizeof(struct ip6_rthdr))
3136 			return (EINVAL);
3137 		rth = (struct ip6_rthdr *)buf;
3138 		rthlen = (rth->ip6r_len + 1) << 3;
3139 		if (len != rthlen)
3140 			return (EINVAL);
3141 
3142 		switch (rth->ip6r_type) {
3143 		case IPV6_RTHDR_TYPE_0:
3144 			if (rth->ip6r_len == 0)	/* must contain one addr */
3145 				return (EINVAL);
3146 			if (rth->ip6r_len % 2) /* length must be even */
3147 				return (EINVAL);
3148 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3149 				return (EINVAL);
3150 			break;
3151 		default:
3152 			return (EINVAL);	/* not supported */
3153 		}
3154 
3155 		/* turn off the previous option */
3156 		ip6_clearpktopts(opt, IPV6_RTHDR);
3157 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3158 		if (opt->ip6po_rthdr == NULL)
3159 			return (ENOBUFS);
3160 		bcopy(rth, opt->ip6po_rthdr, rthlen);
3161 
3162 		break;
3163 	}
3164 
3165 	case IPV6_USE_MIN_MTU:
3166 		if (len != sizeof(int))
3167 			return (EINVAL);
3168 		minmtupolicy = *(int *)buf;
3169 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3170 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3171 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3172 			return (EINVAL);
3173 		}
3174 		opt->ip6po_minmtu = minmtupolicy;
3175 		break;
3176 
3177 	case IPV6_DONTFRAG:
3178 		if (len != sizeof(int))
3179 			return (EINVAL);
3180 
3181 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3182 			/*
3183 			 * we ignore this option for TCP sockets.
3184 			 * (RFC3542 leaves this case unspecified.)
3185 			 */
3186 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3187 		} else
3188 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3189 		break;
3190 
3191 	case IPV6_PREFER_TEMPADDR:
3192 		if (len != sizeof(int))
3193 			return (EINVAL);
3194 		preftemp = *(int *)buf;
3195 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3196 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3197 		    preftemp != IP6PO_TEMPADDR_PREFER) {
3198 			return (EINVAL);
3199 		}
3200 		opt->ip6po_prefer_tempaddr = preftemp;
3201 		break;
3202 
3203 	default:
3204 		return (ENOPROTOOPT);
3205 	} /* end of switch */
3206 
3207 	return (0);
3208 }
3209 
3210 /*
3211  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3212  * packet to the input queue of a specified interface.  Note that this
3213  * calls the output routine of the loopback "driver", but with an interface
3214  * pointer that might NOT be &loif -- easier than replicating that code here.
3215  */
3216 void
3217 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
3218 {
3219 	struct mbuf *copym;
3220 	struct ip6_hdr *ip6;
3221 
3222 	copym = m_copy(m, 0, M_COPYALL);
3223 	if (copym == NULL)
3224 		return;
3225 
3226 	/*
3227 	 * Make sure to deep-copy IPv6 header portion in case the data
3228 	 * is in an mbuf cluster, so that we can safely override the IPv6
3229 	 * header portion later.
3230 	 */
3231 	if ((copym->m_flags & M_EXT) != 0 ||
3232 	    copym->m_len < sizeof(struct ip6_hdr)) {
3233 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3234 		if (copym == NULL)
3235 			return;
3236 	}
3237 
3238 #ifdef DIAGNOSTIC
3239 	if (copym->m_len < sizeof(*ip6)) {
3240 		m_freem(copym);
3241 		return;
3242 	}
3243 #endif
3244 
3245 	ip6 = mtod(copym, struct ip6_hdr *);
3246 	/*
3247 	 * clear embedded scope identifiers if necessary.
3248 	 * in6_clearscope will touch the addresses only when necessary.
3249 	 */
3250 	in6_clearscope(&ip6->ip6_src);
3251 	in6_clearscope(&ip6->ip6_dst);
3252 
3253 	(void)if_simloop(ifp, copym, dst->sin6_family, 0);
3254 }
3255 
3256 /*
3257  * Chop IPv6 header off from the payload.
3258  */
3259 static int
3260 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3261 {
3262 	struct mbuf *mh;
3263 	struct ip6_hdr *ip6;
3264 
3265 	ip6 = mtod(m, struct ip6_hdr *);
3266 	if (m->m_len > sizeof(*ip6)) {
3267 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3268 		if (mh == 0) {
3269 			m_freem(m);
3270 			return ENOBUFS;
3271 		}
3272 		M_MOVE_PKTHDR(mh, m);
3273 		MH_ALIGN(mh, sizeof(*ip6));
3274 		m->m_len -= sizeof(*ip6);
3275 		m->m_data += sizeof(*ip6);
3276 		mh->m_next = m;
3277 		m = mh;
3278 		m->m_len = sizeof(*ip6);
3279 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3280 	}
3281 	exthdrs->ip6e_ip6 = m;
3282 	return 0;
3283 }
3284 
3285 /*
3286  * Compute IPv6 extension header length.
3287  */
3288 int
3289 ip6_optlen(struct in6pcb *in6p)
3290 {
3291 	int len;
3292 
3293 	if (!in6p->in6p_outputopts)
3294 		return 0;
3295 
3296 	len = 0;
3297 #define elen(x) \
3298     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3299 
3300 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3301 	if (in6p->in6p_outputopts->ip6po_rthdr)
3302 		/* dest1 is valid with rthdr only */
3303 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
3304 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3305 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3306 	return len;
3307 #undef elen
3308 }
3309