xref: /freebsd/sys/netinet6/ip6_output.c (revision 4a5216a6dc0c3ce4cf5f2d3ee8af0c3ff3402c4f)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
30  */
31 
32 /*-
33  * Copyright (c) 1982, 1986, 1988, 1990, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
61  */
62 
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
65 
66 #include "opt_inet.h"
67 #include "opt_inet6.h"
68 #include "opt_ipsec.h"
69 
70 #include <sys/param.h>
71 #include <sys/kernel.h>
72 #include <sys/malloc.h>
73 #include <sys/mbuf.h>
74 #include <sys/errno.h>
75 #include <sys/priv.h>
76 #include <sys/proc.h>
77 #include <sys/protosw.h>
78 #include <sys/socket.h>
79 #include <sys/socketvar.h>
80 #include <sys/ucred.h>
81 #include <sys/vimage.h>
82 
83 #include <net/if.h>
84 #include <net/netisr.h>
85 #include <net/route.h>
86 #include <net/pfil.h>
87 
88 #include <netinet/in.h>
89 #include <netinet/in_var.h>
90 #include <netinet6/in6_var.h>
91 #include <netinet/ip6.h>
92 #include <netinet/icmp6.h>
93 #include <netinet6/ip6_var.h>
94 #include <netinet/in_pcb.h>
95 #include <netinet/tcp_var.h>
96 #include <netinet6/nd6.h>
97 
98 #ifdef IPSEC
99 #include <netipsec/ipsec.h>
100 #include <netipsec/ipsec6.h>
101 #include <netipsec/key.h>
102 #include <netinet6/ip6_ipsec.h>
103 #endif /* IPSEC */
104 
105 #include <netinet6/ip6protosw.h>
106 #include <netinet6/scope6_var.h>
107 
108 static MALLOC_DEFINE(M_IP6MOPTS, "ip6_moptions", "internet multicast options");
109 
110 struct ip6_exthdrs {
111 	struct mbuf *ip6e_ip6;
112 	struct mbuf *ip6e_hbh;
113 	struct mbuf *ip6e_dest1;
114 	struct mbuf *ip6e_rthdr;
115 	struct mbuf *ip6e_dest2;
116 };
117 
118 static int ip6_pcbopt __P((int, u_char *, int, struct ip6_pktopts **,
119 			   struct ucred *, int));
120 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
121 	struct socket *, struct sockopt *));
122 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
123 static int ip6_setpktopt __P((int, u_char *, int, struct ip6_pktopts *,
124 	struct ucred *, int, int, int));
125 
126 static int ip6_setmoptions(int, struct ip6_moptions **, struct mbuf *);
127 static int ip6_getmoptions(int, struct ip6_moptions *, struct mbuf **);
128 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
129 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
130 	struct ip6_frag **));
131 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
132 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
133 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *,
134 	struct ifnet *, struct in6_addr *, u_long *, int *));
135 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
136 
137 
138 /*
139  * Make an extension header from option data.  hp is the source, and
140  * mp is the destination.
141  */
142 #define MAKE_EXTHDR(hp, mp)						\
143     do {								\
144 	if (hp) {							\
145 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
146 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
147 		    ((eh)->ip6e_len + 1) << 3);				\
148 		if (error)						\
149 			goto freehdrs;					\
150 	}								\
151     } while (/*CONSTCOND*/ 0)
152 
153 /*
154  * Form a chain of extension headers.
155  * m is the extension header mbuf
156  * mp is the previous mbuf in the chain
157  * p is the next header
158  * i is the type of option.
159  */
160 #define MAKE_CHAIN(m, mp, p, i)\
161     do {\
162 	if (m) {\
163 		if (!hdrsplit) \
164 			panic("assumption failed: hdr not split"); \
165 		*mtod((m), u_char *) = *(p);\
166 		*(p) = (i);\
167 		p = mtod((m), u_char *);\
168 		(m)->m_next = (mp)->m_next;\
169 		(mp)->m_next = (m);\
170 		(mp) = (m);\
171 	}\
172     } while (/*CONSTCOND*/ 0)
173 
174 /*
175  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
176  * header (with pri, len, nxt, hlim, src, dst).
177  * This function may modify ver and hlim only.
178  * The mbuf chain containing the packet will be freed.
179  * The mbuf opt, if present, will not be freed.
180  *
181  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
182  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
183  * which is rt_rmx.rmx_mtu.
184  *
185  * ifpp - XXX: just for statistics
186  */
187 int
188 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
189     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
190     struct ifnet **ifpp, struct inpcb *inp)
191 {
192 	INIT_VNET_NET(curvnet);
193 	INIT_VNET_INET6(curvnet);
194 	struct ip6_hdr *ip6, *mhip6;
195 	struct ifnet *ifp, *origifp;
196 	struct mbuf *m = m0;
197 	struct mbuf *mprev = NULL;
198 	int hlen, tlen, len, off;
199 	struct route_in6 ip6route;
200 	struct rtentry *rt = NULL;
201 	struct sockaddr_in6 *dst, src_sa, dst_sa;
202 	struct in6_addr odst;
203 	int error = 0;
204 	struct in6_ifaddr *ia = NULL;
205 	u_long mtu;
206 	int alwaysfrag, dontfrag;
207 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
208 	struct ip6_exthdrs exthdrs;
209 	struct in6_addr finaldst, src0, dst0;
210 	u_int32_t zone;
211 	struct route_in6 *ro_pmtu = NULL;
212 	int hdrsplit = 0;
213 	int needipsec = 0;
214 #ifdef IPSEC
215 	struct ipsec_output_state state;
216 	struct ip6_rthdr *rh = NULL;
217 	int needipsectun = 0;
218 	int segleft_org = 0;
219 	struct secpolicy *sp = NULL;
220 #endif /* IPSEC */
221 
222 	ip6 = mtod(m, struct ip6_hdr *);
223 	if (ip6 == NULL) {
224 		printf ("ip6 is NULL");
225 		goto bad;
226 	}
227 
228 	finaldst = ip6->ip6_dst;
229 
230 	bzero(&exthdrs, sizeof(exthdrs));
231 
232 	if (opt) {
233 		/* Hop-by-Hop options header */
234 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
235 		/* Destination options header(1st part) */
236 		if (opt->ip6po_rthdr) {
237 			/*
238 			 * Destination options header(1st part)
239 			 * This only makes sense with a routing header.
240 			 * See Section 9.2 of RFC 3542.
241 			 * Disabling this part just for MIP6 convenience is
242 			 * a bad idea.  We need to think carefully about a
243 			 * way to make the advanced API coexist with MIP6
244 			 * options, which might automatically be inserted in
245 			 * the kernel.
246 			 */
247 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
248 		}
249 		/* Routing header */
250 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
251 		/* Destination options header(2nd part) */
252 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
253 	}
254 
255 	/*
256 	 * IPSec checking which handles several cases.
257 	 * FAST IPSEC: We re-injected the packet.
258 	 */
259 #ifdef IPSEC
260 	switch(ip6_ipsec_output(&m, inp, &flags, &error, &ifp, &sp))
261 	{
262 	case 1:                 /* Bad packet */
263 		goto freehdrs;
264 	case -1:                /* Do IPSec */
265 		needipsec = 1;
266 	case 0:                 /* No IPSec */
267 	default:
268 		break;
269 	}
270 #endif /* IPSEC */
271 
272 	/*
273 	 * Calculate the total length of the extension header chain.
274 	 * Keep the length of the unfragmentable part for fragmentation.
275 	 */
276 	optlen = 0;
277 	if (exthdrs.ip6e_hbh)
278 		optlen += exthdrs.ip6e_hbh->m_len;
279 	if (exthdrs.ip6e_dest1)
280 		optlen += exthdrs.ip6e_dest1->m_len;
281 	if (exthdrs.ip6e_rthdr)
282 		optlen += exthdrs.ip6e_rthdr->m_len;
283 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
284 
285 	/* NOTE: we don't add AH/ESP length here. do that later. */
286 	if (exthdrs.ip6e_dest2)
287 		optlen += exthdrs.ip6e_dest2->m_len;
288 
289 	/*
290 	 * If we need IPsec, or there is at least one extension header,
291 	 * separate IP6 header from the payload.
292 	 */
293 	if ((needipsec || optlen) && !hdrsplit) {
294 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
295 			m = NULL;
296 			goto freehdrs;
297 		}
298 		m = exthdrs.ip6e_ip6;
299 		hdrsplit++;
300 	}
301 
302 	/* adjust pointer */
303 	ip6 = mtod(m, struct ip6_hdr *);
304 
305 	/* adjust mbuf packet header length */
306 	m->m_pkthdr.len += optlen;
307 	plen = m->m_pkthdr.len - sizeof(*ip6);
308 
309 	/* If this is a jumbo payload, insert a jumbo payload option. */
310 	if (plen > IPV6_MAXPACKET) {
311 		if (!hdrsplit) {
312 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
313 				m = NULL;
314 				goto freehdrs;
315 			}
316 			m = exthdrs.ip6e_ip6;
317 			hdrsplit++;
318 		}
319 		/* adjust pointer */
320 		ip6 = mtod(m, struct ip6_hdr *);
321 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
322 			goto freehdrs;
323 		ip6->ip6_plen = 0;
324 	} else
325 		ip6->ip6_plen = htons(plen);
326 
327 	/*
328 	 * Concatenate headers and fill in next header fields.
329 	 * Here we have, on "m"
330 	 *	IPv6 payload
331 	 * and we insert headers accordingly.  Finally, we should be getting:
332 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
333 	 *
334 	 * during the header composing process, "m" points to IPv6 header.
335 	 * "mprev" points to an extension header prior to esp.
336 	 */
337 	u_char *nexthdrp = &ip6->ip6_nxt;
338 	mprev = m;
339 
340 	/*
341 	 * we treat dest2 specially.  this makes IPsec processing
342 	 * much easier.  the goal here is to make mprev point the
343 	 * mbuf prior to dest2.
344 	 *
345 	 * result: IPv6 dest2 payload
346 	 * m and mprev will point to IPv6 header.
347 	 */
348 	if (exthdrs.ip6e_dest2) {
349 		if (!hdrsplit)
350 			panic("assumption failed: hdr not split");
351 		exthdrs.ip6e_dest2->m_next = m->m_next;
352 		m->m_next = exthdrs.ip6e_dest2;
353 		*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
354 		ip6->ip6_nxt = IPPROTO_DSTOPTS;
355 	}
356 
357 	/*
358 	 * result: IPv6 hbh dest1 rthdr dest2 payload
359 	 * m will point to IPv6 header.  mprev will point to the
360 	 * extension header prior to dest2 (rthdr in the above case).
361 	 */
362 	MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
363 	MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
364 		   IPPROTO_DSTOPTS);
365 	MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
366 		   IPPROTO_ROUTING);
367 
368 #ifdef IPSEC
369 	if (!needipsec)
370 		goto skip_ipsec2;
371 
372 	/*
373 	 * pointers after IPsec headers are not valid any more.
374 	 * other pointers need a great care too.
375 	 * (IPsec routines should not mangle mbufs prior to AH/ESP)
376 	 */
377 	exthdrs.ip6e_dest2 = NULL;
378 
379 	if (exthdrs.ip6e_rthdr) {
380 		rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
381 		segleft_org = rh->ip6r_segleft;
382 		rh->ip6r_segleft = 0;
383 	}
384 
385 	bzero(&state, sizeof(state));
386 	state.m = m;
387 	error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
388 				    &needipsectun);
389 	m = state.m;
390 	if (error == EJUSTRETURN) {
391 		/*
392 		 * We had a SP with a level of 'use' and no SA. We
393 		 * will just continue to process the packet without
394 		 * IPsec processing.
395 		 */
396 		;
397 	} else if (error) {
398 		/* mbuf is already reclaimed in ipsec6_output_trans. */
399 		m = NULL;
400 		switch (error) {
401 		case EHOSTUNREACH:
402 		case ENETUNREACH:
403 		case EMSGSIZE:
404 		case ENOBUFS:
405 		case ENOMEM:
406 			break;
407 		default:
408 			printf("[%s:%d] (ipsec): error code %d\n",
409 			    __func__, __LINE__, error);
410 			/* FALLTHROUGH */
411 		case ENOENT:
412 			/* don't show these error codes to the user */
413 			error = 0;
414 			break;
415 		}
416 		goto bad;
417 	} else if (!needipsectun) {
418 		/*
419 		 * In the FAST IPSec case we have already
420 		 * re-injected the packet and it has been freed
421 		 * by the ipsec_done() function.  So, just clean
422 		 * up after ourselves.
423 		 */
424 		m = NULL;
425 		goto done;
426 	}
427 	if (exthdrs.ip6e_rthdr) {
428 		/* ah6_output doesn't modify mbuf chain */
429 		rh->ip6r_segleft = segleft_org;
430 	}
431 skip_ipsec2:;
432 #endif /* IPSEC */
433 
434 	/*
435 	 * If there is a routing header, replace the destination address field
436 	 * with the first hop of the routing header.
437 	 */
438 	if (exthdrs.ip6e_rthdr) {
439 		struct ip6_rthdr *rh =
440 			(struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
441 						  struct ip6_rthdr *));
442 		struct ip6_rthdr0 *rh0;
443 		struct in6_addr *addr;
444 		struct sockaddr_in6 sa;
445 
446 		switch (rh->ip6r_type) {
447 		case IPV6_RTHDR_TYPE_0:
448 			 rh0 = (struct ip6_rthdr0 *)rh;
449 			 addr = (struct in6_addr *)(rh0 + 1);
450 
451 			 /*
452 			  * construct a sockaddr_in6 form of
453 			  * the first hop.
454 			  *
455 			  * XXX: we may not have enough
456 			  * information about its scope zone;
457 			  * there is no standard API to pass
458 			  * the information from the
459 			  * application.
460 			  */
461 			 bzero(&sa, sizeof(sa));
462 			 sa.sin6_family = AF_INET6;
463 			 sa.sin6_len = sizeof(sa);
464 			 sa.sin6_addr = addr[0];
465 			 if ((error = sa6_embedscope(&sa,
466 			     V_ip6_use_defzone)) != 0) {
467 				 goto bad;
468 			 }
469 			 ip6->ip6_dst = sa.sin6_addr;
470 			 bcopy(&addr[1], &addr[0], sizeof(struct in6_addr)
471 			     * (rh0->ip6r0_segleft - 1));
472 			 addr[rh0->ip6r0_segleft - 1] = finaldst;
473 			 /* XXX */
474 			 in6_clearscope(addr + rh0->ip6r0_segleft - 1);
475 			 break;
476 		default:	/* is it possible? */
477 			 error = EINVAL;
478 			 goto bad;
479 		}
480 	}
481 
482 	/* Source address validation */
483 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
484 	    (flags & IPV6_UNSPECSRC) == 0) {
485 		error = EOPNOTSUPP;
486 		V_ip6stat.ip6s_badscope++;
487 		goto bad;
488 	}
489 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
490 		error = EOPNOTSUPP;
491 		V_ip6stat.ip6s_badscope++;
492 		goto bad;
493 	}
494 
495 	V_ip6stat.ip6s_localout++;
496 
497 	/*
498 	 * Route packet.
499 	 */
500 	if (ro == 0) {
501 		ro = &ip6route;
502 		bzero((caddr_t)ro, sizeof(*ro));
503 	}
504 	ro_pmtu = ro;
505 	if (opt && opt->ip6po_rthdr)
506 		ro = &opt->ip6po_route;
507 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
508 
509 again:
510 	/*
511 	 * if specified, try to fill in the traffic class field.
512 	 * do not override if a non-zero value is already set.
513 	 * we check the diffserv field and the ecn field separately.
514 	 */
515 	if (opt && opt->ip6po_tclass >= 0) {
516 		int mask = 0;
517 
518 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
519 			mask |= 0xfc;
520 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
521 			mask |= 0x03;
522 		if (mask != 0)
523 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
524 	}
525 
526 	/* fill in or override the hop limit field, if necessary. */
527 	if (opt && opt->ip6po_hlim != -1)
528 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
529 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
530 		if (im6o != NULL)
531 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
532 		else
533 			ip6->ip6_hlim = V_ip6_defmcasthlim;
534 	}
535 
536 #ifdef IPSEC
537 	/*
538 	 * We may re-inject packets into the stack here.
539 	 */
540 	if (needipsec && needipsectun) {
541 		struct ipsec_output_state state;
542 
543 		/*
544 		 * All the extension headers will become inaccessible
545 		 * (since they can be encrypted).
546 		 * Don't panic, we need no more updates to extension headers
547 		 * on inner IPv6 packet (since they are now encapsulated).
548 		 *
549 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
550 		 */
551 		bzero(&exthdrs, sizeof(exthdrs));
552 		exthdrs.ip6e_ip6 = m;
553 
554 		bzero(&state, sizeof(state));
555 		state.m = m;
556 		state.ro = (struct route *)ro;
557 		state.dst = (struct sockaddr *)dst;
558 
559 		error = ipsec6_output_tunnel(&state, sp, flags);
560 
561 		m = state.m;
562 		ro = (struct route_in6 *)state.ro;
563 		dst = (struct sockaddr_in6 *)state.dst;
564 		if (error == EJUSTRETURN) {
565 			/*
566 			 * We had a SP with a level of 'use' and no SA. We
567 			 * will just continue to process the packet without
568 			 * IPsec processing.
569 			 */
570 			;
571 		} else if (error) {
572 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
573 			m0 = m = NULL;
574 			m = NULL;
575 			switch (error) {
576 			case EHOSTUNREACH:
577 			case ENETUNREACH:
578 			case EMSGSIZE:
579 			case ENOBUFS:
580 			case ENOMEM:
581 				break;
582 			default:
583 				printf("[%s:%d] (ipsec): error code %d\n",
584 				    __func__, __LINE__, error);
585 				/* FALLTHROUGH */
586 			case ENOENT:
587 				/* don't show these error codes to the user */
588 				error = 0;
589 				break;
590 			}
591 			goto bad;
592 		} else {
593 			/*
594 			 * In the FAST IPSec case we have already
595 			 * re-injected the packet and it has been freed
596 			 * by the ipsec_done() function.  So, just clean
597 			 * up after ourselves.
598 			 */
599 			m = NULL;
600 			goto done;
601 		}
602 
603 		exthdrs.ip6e_ip6 = m;
604 	}
605 #endif /* IPSEC */
606 
607 	/* adjust pointer */
608 	ip6 = mtod(m, struct ip6_hdr *);
609 
610 	bzero(&dst_sa, sizeof(dst_sa));
611 	dst_sa.sin6_family = AF_INET6;
612 	dst_sa.sin6_len = sizeof(dst_sa);
613 	dst_sa.sin6_addr = ip6->ip6_dst;
614 	if ((error = in6_selectroute(&dst_sa, opt, im6o, ro,
615 	    &ifp, &rt, 0)) != 0) {
616 		switch (error) {
617 		case EHOSTUNREACH:
618 			V_ip6stat.ip6s_noroute++;
619 			break;
620 		case EADDRNOTAVAIL:
621 		default:
622 			break; /* XXX statistics? */
623 		}
624 		if (ifp != NULL)
625 			in6_ifstat_inc(ifp, ifs6_out_discard);
626 		goto bad;
627 	}
628 	if (rt == NULL) {
629 		/*
630 		 * If in6_selectroute() does not return a route entry,
631 		 * dst may not have been updated.
632 		 */
633 		*dst = dst_sa;	/* XXX */
634 	}
635 
636 	/*
637 	 * then rt (for unicast) and ifp must be non-NULL valid values.
638 	 */
639 	if ((flags & IPV6_FORWARDING) == 0) {
640 		/* XXX: the FORWARDING flag can be set for mrouting. */
641 		in6_ifstat_inc(ifp, ifs6_out_request);
642 	}
643 	if (rt != NULL) {
644 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
645 		rt->rt_use++;
646 	}
647 
648 	/*
649 	 * The outgoing interface must be in the zone of source and
650 	 * destination addresses.  We should use ia_ifp to support the
651 	 * case of sending packets to an address of our own.
652 	 */
653 	if (ia != NULL && ia->ia_ifp)
654 		origifp = ia->ia_ifp;
655 	else
656 		origifp = ifp;
657 
658 	src0 = ip6->ip6_src;
659 	if (in6_setscope(&src0, origifp, &zone))
660 		goto badscope;
661 	bzero(&src_sa, sizeof(src_sa));
662 	src_sa.sin6_family = AF_INET6;
663 	src_sa.sin6_len = sizeof(src_sa);
664 	src_sa.sin6_addr = ip6->ip6_src;
665 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
666 		goto badscope;
667 
668 	dst0 = ip6->ip6_dst;
669 	if (in6_setscope(&dst0, origifp, &zone))
670 		goto badscope;
671 	/* re-initialize to be sure */
672 	bzero(&dst_sa, sizeof(dst_sa));
673 	dst_sa.sin6_family = AF_INET6;
674 	dst_sa.sin6_len = sizeof(dst_sa);
675 	dst_sa.sin6_addr = ip6->ip6_dst;
676 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
677 		goto badscope;
678 	}
679 
680 	/* scope check is done. */
681 	goto routefound;
682 
683   badscope:
684 	V_ip6stat.ip6s_badscope++;
685 	in6_ifstat_inc(origifp, ifs6_out_discard);
686 	if (error == 0)
687 		error = EHOSTUNREACH; /* XXX */
688 	goto bad;
689 
690   routefound:
691 	if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
692 		if (opt && opt->ip6po_nextroute.ro_rt) {
693 			/*
694 			 * The nexthop is explicitly specified by the
695 			 * application.  We assume the next hop is an IPv6
696 			 * address.
697 			 */
698 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
699 		}
700 		else if ((rt->rt_flags & RTF_GATEWAY))
701 			dst = (struct sockaddr_in6 *)rt->rt_gateway;
702 	}
703 
704 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
705 		m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
706 	} else {
707 		struct	in6_multi *in6m;
708 
709 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
710 
711 		in6_ifstat_inc(ifp, ifs6_out_mcast);
712 
713 		/*
714 		 * Confirm that the outgoing interface supports multicast.
715 		 */
716 		if (!(ifp->if_flags & IFF_MULTICAST)) {
717 			V_ip6stat.ip6s_noroute++;
718 			in6_ifstat_inc(ifp, ifs6_out_discard);
719 			error = ENETUNREACH;
720 			goto bad;
721 		}
722 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
723 		if (in6m != NULL &&
724 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
725 			/*
726 			 * If we belong to the destination multicast group
727 			 * on the outgoing interface, and the caller did not
728 			 * forbid loopback, loop back a copy.
729 			 */
730 			ip6_mloopback(ifp, m, dst);
731 		} else {
732 			/*
733 			 * If we are acting as a multicast router, perform
734 			 * multicast forwarding as if the packet had just
735 			 * arrived on the interface to which we are about
736 			 * to send.  The multicast forwarding function
737 			 * recursively calls this function, using the
738 			 * IPV6_FORWARDING flag to prevent infinite recursion.
739 			 *
740 			 * Multicasts that are looped back by ip6_mloopback(),
741 			 * above, will be forwarded by the ip6_input() routine,
742 			 * if necessary.
743 			 */
744 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
745 				/*
746 				 * XXX: ip6_mforward expects that rcvif is NULL
747 				 * when it is called from the originating path.
748 				 * However, it is not always the case, since
749 				 * some versions of MGETHDR() does not
750 				 * initialize the field.
751 				 */
752 				m->m_pkthdr.rcvif = NULL;
753 				if (ip6_mforward(ip6, ifp, m) != 0) {
754 					m_freem(m);
755 					goto done;
756 				}
757 			}
758 		}
759 		/*
760 		 * Multicasts with a hoplimit of zero may be looped back,
761 		 * above, but must not be transmitted on a network.
762 		 * Also, multicasts addressed to the loopback interface
763 		 * are not sent -- the above call to ip6_mloopback() will
764 		 * loop back a copy if this host actually belongs to the
765 		 * destination group on the loopback interface.
766 		 */
767 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
768 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
769 			m_freem(m);
770 			goto done;
771 		}
772 	}
773 
774 	/*
775 	 * Fill the outgoing inteface to tell the upper layer
776 	 * to increment per-interface statistics.
777 	 */
778 	if (ifpp)
779 		*ifpp = ifp;
780 
781 	/* Determine path MTU. */
782 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
783 	    &alwaysfrag)) != 0)
784 		goto bad;
785 
786 	/*
787 	 * The caller of this function may specify to use the minimum MTU
788 	 * in some cases.
789 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
790 	 * setting.  The logic is a bit complicated; by default, unicast
791 	 * packets will follow path MTU while multicast packets will be sent at
792 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
793 	 * including unicast ones will be sent at the minimum MTU.  Multicast
794 	 * packets will always be sent at the minimum MTU unless
795 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
796 	 * See RFC 3542 for more details.
797 	 */
798 	if (mtu > IPV6_MMTU) {
799 		if ((flags & IPV6_MINMTU))
800 			mtu = IPV6_MMTU;
801 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
802 			mtu = IPV6_MMTU;
803 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
804 			 (opt == NULL ||
805 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
806 			mtu = IPV6_MMTU;
807 		}
808 	}
809 
810 	/*
811 	 * clear embedded scope identifiers if necessary.
812 	 * in6_clearscope will touch the addresses only when necessary.
813 	 */
814 	in6_clearscope(&ip6->ip6_src);
815 	in6_clearscope(&ip6->ip6_dst);
816 
817 	/*
818 	 * If the outgoing packet contains a hop-by-hop options header,
819 	 * it must be examined and processed even by the source node.
820 	 * (RFC 2460, section 4.)
821 	 */
822 	if (exthdrs.ip6e_hbh) {
823 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
824 		u_int32_t dummy; /* XXX unused */
825 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
826 
827 #ifdef DIAGNOSTIC
828 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
829 			panic("ip6e_hbh is not continuous");
830 #endif
831 		/*
832 		 *  XXX: if we have to send an ICMPv6 error to the sender,
833 		 *       we need the M_LOOP flag since icmp6_error() expects
834 		 *       the IPv6 and the hop-by-hop options header are
835 		 *       continuous unless the flag is set.
836 		 */
837 		m->m_flags |= M_LOOP;
838 		m->m_pkthdr.rcvif = ifp;
839 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
840 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
841 		    &dummy, &plen) < 0) {
842 			/* m was already freed at this point */
843 			error = EINVAL;/* better error? */
844 			goto done;
845 		}
846 		m->m_flags &= ~M_LOOP; /* XXX */
847 		m->m_pkthdr.rcvif = NULL;
848 	}
849 
850 	/* Jump over all PFIL processing if hooks are not active. */
851 	if (!PFIL_HOOKED(&inet6_pfil_hook))
852 		goto passout;
853 
854 	odst = ip6->ip6_dst;
855 	/* Run through list of hooks for output packets. */
856 	error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
857 	if (error != 0 || m == NULL)
858 		goto done;
859 	ip6 = mtod(m, struct ip6_hdr *);
860 
861 	/* See if destination IP address was changed by packet filter. */
862 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
863 		m->m_flags |= M_SKIP_FIREWALL;
864 		/* If destination is now ourself drop to ip6_input(). */
865 		if (in6_localaddr(&ip6->ip6_dst)) {
866 			if (m->m_pkthdr.rcvif == NULL)
867 				m->m_pkthdr.rcvif = V_loif;
868 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
869 				m->m_pkthdr.csum_flags |=
870 				    CSUM_DATA_VALID | CSUM_PSEUDO_HDR;
871 				m->m_pkthdr.csum_data = 0xffff;
872 			}
873 			m->m_pkthdr.csum_flags |=
874 			    CSUM_IP_CHECKED | CSUM_IP_VALID;
875 			error = netisr_queue(NETISR_IPV6, m);
876 			goto done;
877 		} else
878 			goto again;	/* Redo the routing table lookup. */
879 	}
880 
881 	/* XXX: IPFIREWALL_FORWARD */
882 
883 passout:
884 	/*
885 	 * Send the packet to the outgoing interface.
886 	 * If necessary, do IPv6 fragmentation before sending.
887 	 *
888 	 * the logic here is rather complex:
889 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
890 	 * 1-a:	send as is if tlen <= path mtu
891 	 * 1-b:	fragment if tlen > path mtu
892 	 *
893 	 * 2: if user asks us not to fragment (dontfrag == 1)
894 	 * 2-a:	send as is if tlen <= interface mtu
895 	 * 2-b:	error if tlen > interface mtu
896 	 *
897 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
898 	 *	always fragment
899 	 *
900 	 * 4: if dontfrag == 1 && alwaysfrag == 1
901 	 *	error, as we cannot handle this conflicting request
902 	 */
903 	tlen = m->m_pkthdr.len;
904 
905 	if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
906 		dontfrag = 1;
907 	else
908 		dontfrag = 0;
909 	if (dontfrag && alwaysfrag) {	/* case 4 */
910 		/* conflicting request - can't transmit */
911 		error = EMSGSIZE;
912 		goto bad;
913 	}
914 	if (dontfrag && tlen > IN6_LINKMTU(ifp)) {	/* case 2-b */
915 		/*
916 		 * Even if the DONTFRAG option is specified, we cannot send the
917 		 * packet when the data length is larger than the MTU of the
918 		 * outgoing interface.
919 		 * Notify the error by sending IPV6_PATHMTU ancillary data as
920 		 * well as returning an error code (the latter is not described
921 		 * in the API spec.)
922 		 */
923 		u_int32_t mtu32;
924 		struct ip6ctlparam ip6cp;
925 
926 		mtu32 = (u_int32_t)mtu;
927 		bzero(&ip6cp, sizeof(ip6cp));
928 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
929 		pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
930 		    (void *)&ip6cp);
931 
932 		error = EMSGSIZE;
933 		goto bad;
934 	}
935 
936 	/*
937 	 * transmit packet without fragmentation
938 	 */
939 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
940 		struct in6_ifaddr *ia6;
941 
942 		ip6 = mtod(m, struct ip6_hdr *);
943 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
944 		if (ia6) {
945 			/* Record statistics for this interface address. */
946 			ia6->ia_ifa.if_opackets++;
947 			ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
948 		}
949 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
950 		goto done;
951 	}
952 
953 	/*
954 	 * try to fragment the packet.  case 1-b and 3
955 	 */
956 	if (mtu < IPV6_MMTU) {
957 		/* path MTU cannot be less than IPV6_MMTU */
958 		error = EMSGSIZE;
959 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
960 		goto bad;
961 	} else if (ip6->ip6_plen == 0) {
962 		/* jumbo payload cannot be fragmented */
963 		error = EMSGSIZE;
964 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
965 		goto bad;
966 	} else {
967 		struct mbuf **mnext, *m_frgpart;
968 		struct ip6_frag *ip6f;
969 		u_int32_t id = htonl(ip6_randomid());
970 		u_char nextproto;
971 
972 		int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
973 
974 		/*
975 		 * Too large for the destination or interface;
976 		 * fragment if possible.
977 		 * Must be able to put at least 8 bytes per fragment.
978 		 */
979 		hlen = unfragpartlen;
980 		if (mtu > IPV6_MAXPACKET)
981 			mtu = IPV6_MAXPACKET;
982 
983 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
984 		if (len < 8) {
985 			error = EMSGSIZE;
986 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
987 			goto bad;
988 		}
989 
990 		/*
991 		 * Verify that we have any chance at all of being able to queue
992 		 *      the packet or packet fragments
993 		 */
994 		if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
995 		    < tlen  /* - hlen */)) {
996 			error = ENOBUFS;
997 			V_ip6stat.ip6s_odropped++;
998 			goto bad;
999 		}
1000 
1001 		mnext = &m->m_nextpkt;
1002 
1003 		/*
1004 		 * Change the next header field of the last header in the
1005 		 * unfragmentable part.
1006 		 */
1007 		if (exthdrs.ip6e_rthdr) {
1008 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1009 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1010 		} else if (exthdrs.ip6e_dest1) {
1011 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1012 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1013 		} else if (exthdrs.ip6e_hbh) {
1014 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1015 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1016 		} else {
1017 			nextproto = ip6->ip6_nxt;
1018 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1019 		}
1020 
1021 		/*
1022 		 * Loop through length of segment after first fragment,
1023 		 * make new header and copy data of each part and link onto
1024 		 * chain.
1025 		 */
1026 		m0 = m;
1027 		for (off = hlen; off < tlen; off += len) {
1028 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
1029 			if (!m) {
1030 				error = ENOBUFS;
1031 				V_ip6stat.ip6s_odropped++;
1032 				goto sendorfree;
1033 			}
1034 			m->m_pkthdr.rcvif = NULL;
1035 			m->m_flags = m0->m_flags & M_COPYFLAGS;
1036 			*mnext = m;
1037 			mnext = &m->m_nextpkt;
1038 			m->m_data += max_linkhdr;
1039 			mhip6 = mtod(m, struct ip6_hdr *);
1040 			*mhip6 = *ip6;
1041 			m->m_len = sizeof(*mhip6);
1042 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1043 			if (error) {
1044 				V_ip6stat.ip6s_odropped++;
1045 				goto sendorfree;
1046 			}
1047 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1048 			if (off + len >= tlen)
1049 				len = tlen - off;
1050 			else
1051 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1052 			mhip6->ip6_plen = htons((u_short)(len + hlen +
1053 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1054 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1055 				error = ENOBUFS;
1056 				V_ip6stat.ip6s_odropped++;
1057 				goto sendorfree;
1058 			}
1059 			m_cat(m, m_frgpart);
1060 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1061 			m->m_pkthdr.rcvif = NULL;
1062 			ip6f->ip6f_reserved = 0;
1063 			ip6f->ip6f_ident = id;
1064 			ip6f->ip6f_nxt = nextproto;
1065 			V_ip6stat.ip6s_ofragments++;
1066 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1067 		}
1068 
1069 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1070 	}
1071 
1072 	/*
1073 	 * Remove leading garbages.
1074 	 */
1075 sendorfree:
1076 	m = m0->m_nextpkt;
1077 	m0->m_nextpkt = 0;
1078 	m_freem(m0);
1079 	for (m0 = m; m; m = m0) {
1080 		m0 = m->m_nextpkt;
1081 		m->m_nextpkt = 0;
1082 		if (error == 0) {
1083 			/* Record statistics for this interface address. */
1084 			if (ia) {
1085 				ia->ia_ifa.if_opackets++;
1086 				ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1087 			}
1088 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1089 		} else
1090 			m_freem(m);
1091 	}
1092 
1093 	if (error == 0)
1094 		V_ip6stat.ip6s_fragmented++;
1095 
1096 done:
1097 	if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1098 		RTFREE(ro->ro_rt);
1099 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1100 		RTFREE(ro_pmtu->ro_rt);
1101 	}
1102 #ifdef IPSEC
1103 	if (sp != NULL)
1104 		KEY_FREESP(&sp);
1105 #endif
1106 
1107 	return (error);
1108 
1109 freehdrs:
1110 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1111 	m_freem(exthdrs.ip6e_dest1);
1112 	m_freem(exthdrs.ip6e_rthdr);
1113 	m_freem(exthdrs.ip6e_dest2);
1114 	/* FALLTHROUGH */
1115 bad:
1116 	if (m)
1117 		m_freem(m);
1118 	goto done;
1119 }
1120 
1121 static int
1122 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1123 {
1124 	struct mbuf *m;
1125 
1126 	if (hlen > MCLBYTES)
1127 		return (ENOBUFS); /* XXX */
1128 
1129 	MGET(m, M_DONTWAIT, MT_DATA);
1130 	if (!m)
1131 		return (ENOBUFS);
1132 
1133 	if (hlen > MLEN) {
1134 		MCLGET(m, M_DONTWAIT);
1135 		if ((m->m_flags & M_EXT) == 0) {
1136 			m_free(m);
1137 			return (ENOBUFS);
1138 		}
1139 	}
1140 	m->m_len = hlen;
1141 	if (hdr)
1142 		bcopy(hdr, mtod(m, caddr_t), hlen);
1143 
1144 	*mp = m;
1145 	return (0);
1146 }
1147 
1148 /*
1149  * Insert jumbo payload option.
1150  */
1151 static int
1152 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1153 {
1154 	struct mbuf *mopt;
1155 	u_char *optbuf;
1156 	u_int32_t v;
1157 
1158 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1159 
1160 	/*
1161 	 * If there is no hop-by-hop options header, allocate new one.
1162 	 * If there is one but it doesn't have enough space to store the
1163 	 * jumbo payload option, allocate a cluster to store the whole options.
1164 	 * Otherwise, use it to store the options.
1165 	 */
1166 	if (exthdrs->ip6e_hbh == 0) {
1167 		MGET(mopt, M_DONTWAIT, MT_DATA);
1168 		if (mopt == 0)
1169 			return (ENOBUFS);
1170 		mopt->m_len = JUMBOOPTLEN;
1171 		optbuf = mtod(mopt, u_char *);
1172 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1173 		exthdrs->ip6e_hbh = mopt;
1174 	} else {
1175 		struct ip6_hbh *hbh;
1176 
1177 		mopt = exthdrs->ip6e_hbh;
1178 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1179 			/*
1180 			 * XXX assumption:
1181 			 * - exthdrs->ip6e_hbh is not referenced from places
1182 			 *   other than exthdrs.
1183 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1184 			 */
1185 			int oldoptlen = mopt->m_len;
1186 			struct mbuf *n;
1187 
1188 			/*
1189 			 * XXX: give up if the whole (new) hbh header does
1190 			 * not fit even in an mbuf cluster.
1191 			 */
1192 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1193 				return (ENOBUFS);
1194 
1195 			/*
1196 			 * As a consequence, we must always prepare a cluster
1197 			 * at this point.
1198 			 */
1199 			MGET(n, M_DONTWAIT, MT_DATA);
1200 			if (n) {
1201 				MCLGET(n, M_DONTWAIT);
1202 				if ((n->m_flags & M_EXT) == 0) {
1203 					m_freem(n);
1204 					n = NULL;
1205 				}
1206 			}
1207 			if (!n)
1208 				return (ENOBUFS);
1209 			n->m_len = oldoptlen + JUMBOOPTLEN;
1210 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1211 			    oldoptlen);
1212 			optbuf = mtod(n, caddr_t) + oldoptlen;
1213 			m_freem(mopt);
1214 			mopt = exthdrs->ip6e_hbh = n;
1215 		} else {
1216 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1217 			mopt->m_len += JUMBOOPTLEN;
1218 		}
1219 		optbuf[0] = IP6OPT_PADN;
1220 		optbuf[1] = 1;
1221 
1222 		/*
1223 		 * Adjust the header length according to the pad and
1224 		 * the jumbo payload option.
1225 		 */
1226 		hbh = mtod(mopt, struct ip6_hbh *);
1227 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1228 	}
1229 
1230 	/* fill in the option. */
1231 	optbuf[2] = IP6OPT_JUMBO;
1232 	optbuf[3] = 4;
1233 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1234 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1235 
1236 	/* finally, adjust the packet header length */
1237 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1238 
1239 	return (0);
1240 #undef JUMBOOPTLEN
1241 }
1242 
1243 /*
1244  * Insert fragment header and copy unfragmentable header portions.
1245  */
1246 static int
1247 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1248     struct ip6_frag **frghdrp)
1249 {
1250 	struct mbuf *n, *mlast;
1251 
1252 	if (hlen > sizeof(struct ip6_hdr)) {
1253 		n = m_copym(m0, sizeof(struct ip6_hdr),
1254 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1255 		if (n == 0)
1256 			return (ENOBUFS);
1257 		m->m_next = n;
1258 	} else
1259 		n = m;
1260 
1261 	/* Search for the last mbuf of unfragmentable part. */
1262 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1263 		;
1264 
1265 	if ((mlast->m_flags & M_EXT) == 0 &&
1266 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1267 		/* use the trailing space of the last mbuf for the fragment hdr */
1268 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1269 		    mlast->m_len);
1270 		mlast->m_len += sizeof(struct ip6_frag);
1271 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1272 	} else {
1273 		/* allocate a new mbuf for the fragment header */
1274 		struct mbuf *mfrg;
1275 
1276 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1277 		if (mfrg == 0)
1278 			return (ENOBUFS);
1279 		mfrg->m_len = sizeof(struct ip6_frag);
1280 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1281 		mlast->m_next = mfrg;
1282 	}
1283 
1284 	return (0);
1285 }
1286 
1287 static int
1288 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1289     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1290     int *alwaysfragp)
1291 {
1292 	u_int32_t mtu = 0;
1293 	int alwaysfrag = 0;
1294 	int error = 0;
1295 
1296 	if (ro_pmtu != ro) {
1297 		/* The first hop and the final destination may differ. */
1298 		struct sockaddr_in6 *sa6_dst =
1299 		    (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1300 		if (ro_pmtu->ro_rt &&
1301 		    ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1302 		     !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1303 			RTFREE(ro_pmtu->ro_rt);
1304 			ro_pmtu->ro_rt = (struct rtentry *)NULL;
1305 		}
1306 		if (ro_pmtu->ro_rt == NULL) {
1307 			bzero(sa6_dst, sizeof(*sa6_dst));
1308 			sa6_dst->sin6_family = AF_INET6;
1309 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1310 			sa6_dst->sin6_addr = *dst;
1311 
1312 			rtalloc((struct route *)ro_pmtu);
1313 		}
1314 	}
1315 	if (ro_pmtu->ro_rt) {
1316 		u_int32_t ifmtu;
1317 		struct in_conninfo inc;
1318 
1319 		bzero(&inc, sizeof(inc));
1320 		inc.inc_flags = 1; /* IPv6 */
1321 		inc.inc6_faddr = *dst;
1322 
1323 		if (ifp == NULL)
1324 			ifp = ro_pmtu->ro_rt->rt_ifp;
1325 		ifmtu = IN6_LINKMTU(ifp);
1326 		mtu = tcp_hc_getmtu(&inc);
1327 		if (mtu)
1328 			mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu);
1329 		else
1330 			mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1331 		if (mtu == 0)
1332 			mtu = ifmtu;
1333 		else if (mtu < IPV6_MMTU) {
1334 			/*
1335 			 * RFC2460 section 5, last paragraph:
1336 			 * if we record ICMPv6 too big message with
1337 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1338 			 * or smaller, with framgent header attached.
1339 			 * (fragment header is needed regardless from the
1340 			 * packet size, for translators to identify packets)
1341 			 */
1342 			alwaysfrag = 1;
1343 			mtu = IPV6_MMTU;
1344 		} else if (mtu > ifmtu) {
1345 			/*
1346 			 * The MTU on the route is larger than the MTU on
1347 			 * the interface!  This shouldn't happen, unless the
1348 			 * MTU of the interface has been changed after the
1349 			 * interface was brought up.  Change the MTU in the
1350 			 * route to match the interface MTU (as long as the
1351 			 * field isn't locked).
1352 			 */
1353 			mtu = ifmtu;
1354 			ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1355 		}
1356 	} else if (ifp) {
1357 		mtu = IN6_LINKMTU(ifp);
1358 	} else
1359 		error = EHOSTUNREACH; /* XXX */
1360 
1361 	*mtup = mtu;
1362 	if (alwaysfragp)
1363 		*alwaysfragp = alwaysfrag;
1364 	return (error);
1365 }
1366 
1367 /*
1368  * IP6 socket option processing.
1369  */
1370 int
1371 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1372 {
1373 	int optdatalen, uproto;
1374 	void *optdata;
1375 	struct inpcb *in6p = sotoinpcb(so);
1376 	int error, optval;
1377 	int level, op, optname;
1378 	int optlen;
1379 	struct thread *td;
1380 
1381 	level = sopt->sopt_level;
1382 	op = sopt->sopt_dir;
1383 	optname = sopt->sopt_name;
1384 	optlen = sopt->sopt_valsize;
1385 	td = sopt->sopt_td;
1386 	error = 0;
1387 	optval = 0;
1388 	uproto = (int)so->so_proto->pr_protocol;
1389 
1390 	if (level == IPPROTO_IPV6) {
1391 		switch (op) {
1392 
1393 		case SOPT_SET:
1394 			switch (optname) {
1395 			case IPV6_2292PKTOPTIONS:
1396 #ifdef IPV6_PKTOPTIONS
1397 			case IPV6_PKTOPTIONS:
1398 #endif
1399 			{
1400 				struct mbuf *m;
1401 
1402 				error = soopt_getm(sopt, &m); /* XXX */
1403 				if (error != 0)
1404 					break;
1405 				error = soopt_mcopyin(sopt, m); /* XXX */
1406 				if (error != 0)
1407 					break;
1408 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1409 						    m, so, sopt);
1410 				m_freem(m); /* XXX */
1411 				break;
1412 			}
1413 
1414 			/*
1415 			 * Use of some Hop-by-Hop options or some
1416 			 * Destination options, might require special
1417 			 * privilege.  That is, normal applications
1418 			 * (without special privilege) might be forbidden
1419 			 * from setting certain options in outgoing packets,
1420 			 * and might never see certain options in received
1421 			 * packets. [RFC 2292 Section 6]
1422 			 * KAME specific note:
1423 			 *  KAME prevents non-privileged users from sending or
1424 			 *  receiving ANY hbh/dst options in order to avoid
1425 			 *  overhead of parsing options in the kernel.
1426 			 */
1427 			case IPV6_RECVHOPOPTS:
1428 			case IPV6_RECVDSTOPTS:
1429 			case IPV6_RECVRTHDRDSTOPTS:
1430 				if (td != NULL) {
1431 					error = priv_check(td,
1432 					    PRIV_NETINET_SETHDROPTS);
1433 					if (error)
1434 						break;
1435 				}
1436 				/* FALLTHROUGH */
1437 			case IPV6_UNICAST_HOPS:
1438 			case IPV6_HOPLIMIT:
1439 			case IPV6_FAITH:
1440 
1441 			case IPV6_RECVPKTINFO:
1442 			case IPV6_RECVHOPLIMIT:
1443 			case IPV6_RECVRTHDR:
1444 			case IPV6_RECVPATHMTU:
1445 			case IPV6_RECVTCLASS:
1446 			case IPV6_V6ONLY:
1447 			case IPV6_AUTOFLOWLABEL:
1448 				if (optlen != sizeof(int)) {
1449 					error = EINVAL;
1450 					break;
1451 				}
1452 				error = sooptcopyin(sopt, &optval,
1453 					sizeof optval, sizeof optval);
1454 				if (error)
1455 					break;
1456 				switch (optname) {
1457 
1458 				case IPV6_UNICAST_HOPS:
1459 					if (optval < -1 || optval >= 256)
1460 						error = EINVAL;
1461 					else {
1462 						/* -1 = kernel default */
1463 						in6p->in6p_hops = optval;
1464 						if ((in6p->in6p_vflag &
1465 						     INP_IPV4) != 0)
1466 							in6p->inp_ip_ttl = optval;
1467 					}
1468 					break;
1469 #define OPTSET(bit) \
1470 do { \
1471 	if (optval) \
1472 		in6p->in6p_flags |= (bit); \
1473 	else \
1474 		in6p->in6p_flags &= ~(bit); \
1475 } while (/*CONSTCOND*/ 0)
1476 #define OPTSET2292(bit) \
1477 do { \
1478 	in6p->in6p_flags |= IN6P_RFC2292; \
1479 	if (optval) \
1480 		in6p->in6p_flags |= (bit); \
1481 	else \
1482 		in6p->in6p_flags &= ~(bit); \
1483 } while (/*CONSTCOND*/ 0)
1484 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1485 
1486 				case IPV6_RECVPKTINFO:
1487 					/* cannot mix with RFC2292 */
1488 					if (OPTBIT(IN6P_RFC2292)) {
1489 						error = EINVAL;
1490 						break;
1491 					}
1492 					OPTSET(IN6P_PKTINFO);
1493 					break;
1494 
1495 				case IPV6_HOPLIMIT:
1496 				{
1497 					struct ip6_pktopts **optp;
1498 
1499 					/* cannot mix with RFC2292 */
1500 					if (OPTBIT(IN6P_RFC2292)) {
1501 						error = EINVAL;
1502 						break;
1503 					}
1504 					optp = &in6p->in6p_outputopts;
1505 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1506 					    (u_char *)&optval, sizeof(optval),
1507 					    optp, (td != NULL) ? td->td_ucred :
1508 					    NULL, uproto);
1509 					break;
1510 				}
1511 
1512 				case IPV6_RECVHOPLIMIT:
1513 					/* cannot mix with RFC2292 */
1514 					if (OPTBIT(IN6P_RFC2292)) {
1515 						error = EINVAL;
1516 						break;
1517 					}
1518 					OPTSET(IN6P_HOPLIMIT);
1519 					break;
1520 
1521 				case IPV6_RECVHOPOPTS:
1522 					/* cannot mix with RFC2292 */
1523 					if (OPTBIT(IN6P_RFC2292)) {
1524 						error = EINVAL;
1525 						break;
1526 					}
1527 					OPTSET(IN6P_HOPOPTS);
1528 					break;
1529 
1530 				case IPV6_RECVDSTOPTS:
1531 					/* cannot mix with RFC2292 */
1532 					if (OPTBIT(IN6P_RFC2292)) {
1533 						error = EINVAL;
1534 						break;
1535 					}
1536 					OPTSET(IN6P_DSTOPTS);
1537 					break;
1538 
1539 				case IPV6_RECVRTHDRDSTOPTS:
1540 					/* cannot mix with RFC2292 */
1541 					if (OPTBIT(IN6P_RFC2292)) {
1542 						error = EINVAL;
1543 						break;
1544 					}
1545 					OPTSET(IN6P_RTHDRDSTOPTS);
1546 					break;
1547 
1548 				case IPV6_RECVRTHDR:
1549 					/* cannot mix with RFC2292 */
1550 					if (OPTBIT(IN6P_RFC2292)) {
1551 						error = EINVAL;
1552 						break;
1553 					}
1554 					OPTSET(IN6P_RTHDR);
1555 					break;
1556 
1557 				case IPV6_FAITH:
1558 					OPTSET(IN6P_FAITH);
1559 					break;
1560 
1561 				case IPV6_RECVPATHMTU:
1562 					/*
1563 					 * We ignore this option for TCP
1564 					 * sockets.
1565 					 * (RFC3542 leaves this case
1566 					 * unspecified.)
1567 					 */
1568 					if (uproto != IPPROTO_TCP)
1569 						OPTSET(IN6P_MTU);
1570 					break;
1571 
1572 				case IPV6_V6ONLY:
1573 					/*
1574 					 * make setsockopt(IPV6_V6ONLY)
1575 					 * available only prior to bind(2).
1576 					 * see ipng mailing list, Jun 22 2001.
1577 					 */
1578 					if (in6p->in6p_lport ||
1579 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1580 						error = EINVAL;
1581 						break;
1582 					}
1583 					OPTSET(IN6P_IPV6_V6ONLY);
1584 					if (optval)
1585 						in6p->in6p_vflag &= ~INP_IPV4;
1586 					else
1587 						in6p->in6p_vflag |= INP_IPV4;
1588 					break;
1589 				case IPV6_RECVTCLASS:
1590 					/* cannot mix with RFC2292 XXX */
1591 					if (OPTBIT(IN6P_RFC2292)) {
1592 						error = EINVAL;
1593 						break;
1594 					}
1595 					OPTSET(IN6P_TCLASS);
1596 					break;
1597 				case IPV6_AUTOFLOWLABEL:
1598 					OPTSET(IN6P_AUTOFLOWLABEL);
1599 					break;
1600 
1601 				}
1602 				break;
1603 
1604 			case IPV6_TCLASS:
1605 			case IPV6_DONTFRAG:
1606 			case IPV6_USE_MIN_MTU:
1607 			case IPV6_PREFER_TEMPADDR:
1608 				if (optlen != sizeof(optval)) {
1609 					error = EINVAL;
1610 					break;
1611 				}
1612 				error = sooptcopyin(sopt, &optval,
1613 					sizeof optval, sizeof optval);
1614 				if (error)
1615 					break;
1616 				{
1617 					struct ip6_pktopts **optp;
1618 					optp = &in6p->in6p_outputopts;
1619 					error = ip6_pcbopt(optname,
1620 					    (u_char *)&optval, sizeof(optval),
1621 					    optp, (td != NULL) ? td->td_ucred :
1622 					    NULL, uproto);
1623 					break;
1624 				}
1625 
1626 			case IPV6_2292PKTINFO:
1627 			case IPV6_2292HOPLIMIT:
1628 			case IPV6_2292HOPOPTS:
1629 			case IPV6_2292DSTOPTS:
1630 			case IPV6_2292RTHDR:
1631 				/* RFC 2292 */
1632 				if (optlen != sizeof(int)) {
1633 					error = EINVAL;
1634 					break;
1635 				}
1636 				error = sooptcopyin(sopt, &optval,
1637 					sizeof optval, sizeof optval);
1638 				if (error)
1639 					break;
1640 				switch (optname) {
1641 				case IPV6_2292PKTINFO:
1642 					OPTSET2292(IN6P_PKTINFO);
1643 					break;
1644 				case IPV6_2292HOPLIMIT:
1645 					OPTSET2292(IN6P_HOPLIMIT);
1646 					break;
1647 				case IPV6_2292HOPOPTS:
1648 					/*
1649 					 * Check super-user privilege.
1650 					 * See comments for IPV6_RECVHOPOPTS.
1651 					 */
1652 					if (td != NULL) {
1653 						error = priv_check(td,
1654 						    PRIV_NETINET_SETHDROPTS);
1655 						if (error)
1656 							return (error);
1657 					}
1658 					OPTSET2292(IN6P_HOPOPTS);
1659 					break;
1660 				case IPV6_2292DSTOPTS:
1661 					if (td != NULL) {
1662 						error = priv_check(td,
1663 						    PRIV_NETINET_SETHDROPTS);
1664 						if (error)
1665 							return (error);
1666 					}
1667 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1668 					break;
1669 				case IPV6_2292RTHDR:
1670 					OPTSET2292(IN6P_RTHDR);
1671 					break;
1672 				}
1673 				break;
1674 			case IPV6_PKTINFO:
1675 			case IPV6_HOPOPTS:
1676 			case IPV6_RTHDR:
1677 			case IPV6_DSTOPTS:
1678 			case IPV6_RTHDRDSTOPTS:
1679 			case IPV6_NEXTHOP:
1680 			{
1681 				/* new advanced API (RFC3542) */
1682 				u_char *optbuf;
1683 				u_char optbuf_storage[MCLBYTES];
1684 				int optlen;
1685 				struct ip6_pktopts **optp;
1686 
1687 				/* cannot mix with RFC2292 */
1688 				if (OPTBIT(IN6P_RFC2292)) {
1689 					error = EINVAL;
1690 					break;
1691 				}
1692 
1693 				/*
1694 				 * We only ensure valsize is not too large
1695 				 * here.  Further validation will be done
1696 				 * later.
1697 				 */
1698 				error = sooptcopyin(sopt, optbuf_storage,
1699 				    sizeof(optbuf_storage), 0);
1700 				if (error)
1701 					break;
1702 				optlen = sopt->sopt_valsize;
1703 				optbuf = optbuf_storage;
1704 				optp = &in6p->in6p_outputopts;
1705 				error = ip6_pcbopt(optname, optbuf, optlen,
1706 				    optp, (td != NULL) ? td->td_ucred : NULL,
1707 				    uproto);
1708 				break;
1709 			}
1710 #undef OPTSET
1711 
1712 			case IPV6_MULTICAST_IF:
1713 			case IPV6_MULTICAST_HOPS:
1714 			case IPV6_MULTICAST_LOOP:
1715 			case IPV6_JOIN_GROUP:
1716 			case IPV6_LEAVE_GROUP:
1717 			    {
1718 				if (sopt->sopt_valsize > MLEN) {
1719 					error = EMSGSIZE;
1720 					break;
1721 				}
1722 				/* XXX */
1723 			    }
1724 			    /* FALLTHROUGH */
1725 			    {
1726 				struct mbuf *m;
1727 
1728 				if (sopt->sopt_valsize > MCLBYTES) {
1729 					error = EMSGSIZE;
1730 					break;
1731 				}
1732 				/* XXX */
1733 				MGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT, MT_DATA);
1734 				if (m == 0) {
1735 					error = ENOBUFS;
1736 					break;
1737 				}
1738 				if (sopt->sopt_valsize > MLEN) {
1739 					MCLGET(m, sopt->sopt_td ? M_WAIT : M_DONTWAIT);
1740 					if ((m->m_flags & M_EXT) == 0) {
1741 						m_free(m);
1742 						error = ENOBUFS;
1743 						break;
1744 					}
1745 				}
1746 				m->m_len = sopt->sopt_valsize;
1747 				error = sooptcopyin(sopt, mtod(m, char *),
1748 						    m->m_len, m->m_len);
1749 				if (error) {
1750 					(void)m_free(m);
1751 					break;
1752 				}
1753 				error =	ip6_setmoptions(sopt->sopt_name,
1754 							&in6p->in6p_moptions,
1755 							m);
1756 				(void)m_free(m);
1757 			    }
1758 				break;
1759 
1760 			case IPV6_PORTRANGE:
1761 				error = sooptcopyin(sopt, &optval,
1762 				    sizeof optval, sizeof optval);
1763 				if (error)
1764 					break;
1765 
1766 				switch (optval) {
1767 				case IPV6_PORTRANGE_DEFAULT:
1768 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1769 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1770 					break;
1771 
1772 				case IPV6_PORTRANGE_HIGH:
1773 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1774 					in6p->in6p_flags |= IN6P_HIGHPORT;
1775 					break;
1776 
1777 				case IPV6_PORTRANGE_LOW:
1778 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1779 					in6p->in6p_flags |= IN6P_LOWPORT;
1780 					break;
1781 
1782 				default:
1783 					error = EINVAL;
1784 					break;
1785 				}
1786 				break;
1787 
1788 #ifdef IPSEC
1789 			case IPV6_IPSEC_POLICY:
1790 			{
1791 				caddr_t req;
1792 				struct mbuf *m;
1793 
1794 				if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1795 					break;
1796 				if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1797 					break;
1798 				req = mtod(m, caddr_t);
1799 				error = ipsec6_set_policy(in6p, optname, req,
1800 				    m->m_len, (sopt->sopt_td != NULL) ?
1801 				    sopt->sopt_td->td_ucred : NULL);
1802 				m_freem(m);
1803 				break;
1804 			}
1805 #endif /* IPSEC */
1806 
1807 			default:
1808 				error = ENOPROTOOPT;
1809 				break;
1810 			}
1811 			break;
1812 
1813 		case SOPT_GET:
1814 			switch (optname) {
1815 
1816 			case IPV6_2292PKTOPTIONS:
1817 #ifdef IPV6_PKTOPTIONS
1818 			case IPV6_PKTOPTIONS:
1819 #endif
1820 				/*
1821 				 * RFC3542 (effectively) deprecated the
1822 				 * semantics of the 2292-style pktoptions.
1823 				 * Since it was not reliable in nature (i.e.,
1824 				 * applications had to expect the lack of some
1825 				 * information after all), it would make sense
1826 				 * to simplify this part by always returning
1827 				 * empty data.
1828 				 */
1829 				sopt->sopt_valsize = 0;
1830 				break;
1831 
1832 			case IPV6_RECVHOPOPTS:
1833 			case IPV6_RECVDSTOPTS:
1834 			case IPV6_RECVRTHDRDSTOPTS:
1835 			case IPV6_UNICAST_HOPS:
1836 			case IPV6_RECVPKTINFO:
1837 			case IPV6_RECVHOPLIMIT:
1838 			case IPV6_RECVRTHDR:
1839 			case IPV6_RECVPATHMTU:
1840 
1841 			case IPV6_FAITH:
1842 			case IPV6_V6ONLY:
1843 			case IPV6_PORTRANGE:
1844 			case IPV6_RECVTCLASS:
1845 			case IPV6_AUTOFLOWLABEL:
1846 				switch (optname) {
1847 
1848 				case IPV6_RECVHOPOPTS:
1849 					optval = OPTBIT(IN6P_HOPOPTS);
1850 					break;
1851 
1852 				case IPV6_RECVDSTOPTS:
1853 					optval = OPTBIT(IN6P_DSTOPTS);
1854 					break;
1855 
1856 				case IPV6_RECVRTHDRDSTOPTS:
1857 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1858 					break;
1859 
1860 				case IPV6_UNICAST_HOPS:
1861 					optval = in6p->in6p_hops;
1862 					break;
1863 
1864 				case IPV6_RECVPKTINFO:
1865 					optval = OPTBIT(IN6P_PKTINFO);
1866 					break;
1867 
1868 				case IPV6_RECVHOPLIMIT:
1869 					optval = OPTBIT(IN6P_HOPLIMIT);
1870 					break;
1871 
1872 				case IPV6_RECVRTHDR:
1873 					optval = OPTBIT(IN6P_RTHDR);
1874 					break;
1875 
1876 				case IPV6_RECVPATHMTU:
1877 					optval = OPTBIT(IN6P_MTU);
1878 					break;
1879 
1880 				case IPV6_FAITH:
1881 					optval = OPTBIT(IN6P_FAITH);
1882 					break;
1883 
1884 				case IPV6_V6ONLY:
1885 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
1886 					break;
1887 
1888 				case IPV6_PORTRANGE:
1889 				    {
1890 					int flags;
1891 					flags = in6p->in6p_flags;
1892 					if (flags & IN6P_HIGHPORT)
1893 						optval = IPV6_PORTRANGE_HIGH;
1894 					else if (flags & IN6P_LOWPORT)
1895 						optval = IPV6_PORTRANGE_LOW;
1896 					else
1897 						optval = 0;
1898 					break;
1899 				    }
1900 				case IPV6_RECVTCLASS:
1901 					optval = OPTBIT(IN6P_TCLASS);
1902 					break;
1903 
1904 				case IPV6_AUTOFLOWLABEL:
1905 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1906 					break;
1907 				}
1908 				if (error)
1909 					break;
1910 				error = sooptcopyout(sopt, &optval,
1911 					sizeof optval);
1912 				break;
1913 
1914 			case IPV6_PATHMTU:
1915 			{
1916 				u_long pmtu = 0;
1917 				struct ip6_mtuinfo mtuinfo;
1918 				struct route_in6 sro;
1919 
1920 				bzero(&sro, sizeof(sro));
1921 
1922 				if (!(so->so_state & SS_ISCONNECTED))
1923 					return (ENOTCONN);
1924 				/*
1925 				 * XXX: we dot not consider the case of source
1926 				 * routing, or optional information to specify
1927 				 * the outgoing interface.
1928 				 */
1929 				error = ip6_getpmtu(&sro, NULL, NULL,
1930 				    &in6p->in6p_faddr, &pmtu, NULL);
1931 				if (sro.ro_rt)
1932 					RTFREE(sro.ro_rt);
1933 				if (error)
1934 					break;
1935 				if (pmtu > IPV6_MAXPACKET)
1936 					pmtu = IPV6_MAXPACKET;
1937 
1938 				bzero(&mtuinfo, sizeof(mtuinfo));
1939 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1940 				optdata = (void *)&mtuinfo;
1941 				optdatalen = sizeof(mtuinfo);
1942 				error = sooptcopyout(sopt, optdata,
1943 				    optdatalen);
1944 				break;
1945 			}
1946 
1947 			case IPV6_2292PKTINFO:
1948 			case IPV6_2292HOPLIMIT:
1949 			case IPV6_2292HOPOPTS:
1950 			case IPV6_2292RTHDR:
1951 			case IPV6_2292DSTOPTS:
1952 				switch (optname) {
1953 				case IPV6_2292PKTINFO:
1954 					optval = OPTBIT(IN6P_PKTINFO);
1955 					break;
1956 				case IPV6_2292HOPLIMIT:
1957 					optval = OPTBIT(IN6P_HOPLIMIT);
1958 					break;
1959 				case IPV6_2292HOPOPTS:
1960 					optval = OPTBIT(IN6P_HOPOPTS);
1961 					break;
1962 				case IPV6_2292RTHDR:
1963 					optval = OPTBIT(IN6P_RTHDR);
1964 					break;
1965 				case IPV6_2292DSTOPTS:
1966 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1967 					break;
1968 				}
1969 				error = sooptcopyout(sopt, &optval,
1970 				    sizeof optval);
1971 				break;
1972 			case IPV6_PKTINFO:
1973 			case IPV6_HOPOPTS:
1974 			case IPV6_RTHDR:
1975 			case IPV6_DSTOPTS:
1976 			case IPV6_RTHDRDSTOPTS:
1977 			case IPV6_NEXTHOP:
1978 			case IPV6_TCLASS:
1979 			case IPV6_DONTFRAG:
1980 			case IPV6_USE_MIN_MTU:
1981 			case IPV6_PREFER_TEMPADDR:
1982 				error = ip6_getpcbopt(in6p->in6p_outputopts,
1983 				    optname, sopt);
1984 				break;
1985 
1986 			case IPV6_MULTICAST_IF:
1987 			case IPV6_MULTICAST_HOPS:
1988 			case IPV6_MULTICAST_LOOP:
1989 			case IPV6_JOIN_GROUP:
1990 			case IPV6_LEAVE_GROUP:
1991 			    {
1992 				struct mbuf *m;
1993 				error = ip6_getmoptions(sopt->sopt_name,
1994 				    in6p->in6p_moptions, &m);
1995 				if (error == 0)
1996 					error = sooptcopyout(sopt,
1997 					    mtod(m, char *), m->m_len);
1998 				m_freem(m);
1999 			    }
2000 				break;
2001 
2002 #ifdef IPSEC
2003 			case IPV6_IPSEC_POLICY:
2004 			  {
2005 				caddr_t req = NULL;
2006 				size_t len = 0;
2007 				struct mbuf *m = NULL;
2008 				struct mbuf **mp = &m;
2009 				size_t ovalsize = sopt->sopt_valsize;
2010 				caddr_t oval = (caddr_t)sopt->sopt_val;
2011 
2012 				error = soopt_getm(sopt, &m); /* XXX */
2013 				if (error != 0)
2014 					break;
2015 				error = soopt_mcopyin(sopt, m); /* XXX */
2016 				if (error != 0)
2017 					break;
2018 				sopt->sopt_valsize = ovalsize;
2019 				sopt->sopt_val = oval;
2020 				if (m) {
2021 					req = mtod(m, caddr_t);
2022 					len = m->m_len;
2023 				}
2024 				error = ipsec6_get_policy(in6p, req, len, mp);
2025 				if (error == 0)
2026 					error = soopt_mcopyout(sopt, m); /* XXX */
2027 				if (error == 0 && m)
2028 					m_freem(m);
2029 				break;
2030 			  }
2031 #endif /* IPSEC */
2032 
2033 			default:
2034 				error = ENOPROTOOPT;
2035 				break;
2036 			}
2037 			break;
2038 		}
2039 	} else {		/* level != IPPROTO_IPV6 */
2040 		error = EINVAL;
2041 	}
2042 	return (error);
2043 }
2044 
2045 int
2046 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2047 {
2048 	int error = 0, optval, optlen;
2049 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2050 	struct in6pcb *in6p = sotoin6pcb(so);
2051 	int level, op, optname;
2052 
2053 	level = sopt->sopt_level;
2054 	op = sopt->sopt_dir;
2055 	optname = sopt->sopt_name;
2056 	optlen = sopt->sopt_valsize;
2057 
2058 	if (level != IPPROTO_IPV6) {
2059 		return (EINVAL);
2060 	}
2061 
2062 	switch (optname) {
2063 	case IPV6_CHECKSUM:
2064 		/*
2065 		 * For ICMPv6 sockets, no modification allowed for checksum
2066 		 * offset, permit "no change" values to help existing apps.
2067 		 *
2068 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2069 		 * for an ICMPv6 socket will fail."
2070 		 * The current behavior does not meet RFC3542.
2071 		 */
2072 		switch (op) {
2073 		case SOPT_SET:
2074 			if (optlen != sizeof(int)) {
2075 				error = EINVAL;
2076 				break;
2077 			}
2078 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2079 					    sizeof(optval));
2080 			if (error)
2081 				break;
2082 			if ((optval % 2) != 0) {
2083 				/* the API assumes even offset values */
2084 				error = EINVAL;
2085 			} else if (so->so_proto->pr_protocol ==
2086 			    IPPROTO_ICMPV6) {
2087 				if (optval != icmp6off)
2088 					error = EINVAL;
2089 			} else
2090 				in6p->in6p_cksum = optval;
2091 			break;
2092 
2093 		case SOPT_GET:
2094 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2095 				optval = icmp6off;
2096 			else
2097 				optval = in6p->in6p_cksum;
2098 
2099 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2100 			break;
2101 
2102 		default:
2103 			error = EINVAL;
2104 			break;
2105 		}
2106 		break;
2107 
2108 	default:
2109 		error = ENOPROTOOPT;
2110 		break;
2111 	}
2112 
2113 	return (error);
2114 }
2115 
2116 /*
2117  * Set up IP6 options in pcb for insertion in output packets or
2118  * specifying behavior of outgoing packets.
2119  */
2120 static int
2121 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2122     struct socket *so, struct sockopt *sopt)
2123 {
2124 	struct ip6_pktopts *opt = *pktopt;
2125 	int error = 0;
2126 	struct thread *td = sopt->sopt_td;
2127 
2128 	/* turn off any old options. */
2129 	if (opt) {
2130 #ifdef DIAGNOSTIC
2131 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2132 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2133 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2134 			printf("ip6_pcbopts: all specified options are cleared.\n");
2135 #endif
2136 		ip6_clearpktopts(opt, -1);
2137 	} else
2138 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2139 	*pktopt = NULL;
2140 
2141 	if (!m || m->m_len == 0) {
2142 		/*
2143 		 * Only turning off any previous options, regardless of
2144 		 * whether the opt is just created or given.
2145 		 */
2146 		free(opt, M_IP6OPT);
2147 		return (0);
2148 	}
2149 
2150 	/*  set options specified by user. */
2151 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2152 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2153 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2154 		free(opt, M_IP6OPT);
2155 		return (error);
2156 	}
2157 	*pktopt = opt;
2158 	return (0);
2159 }
2160 
2161 /*
2162  * initialize ip6_pktopts.  beware that there are non-zero default values in
2163  * the struct.
2164  */
2165 void
2166 ip6_initpktopts(struct ip6_pktopts *opt)
2167 {
2168 
2169 	bzero(opt, sizeof(*opt));
2170 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2171 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2172 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2173 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2174 }
2175 
2176 static int
2177 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2178     struct ucred *cred, int uproto)
2179 {
2180 	struct ip6_pktopts *opt;
2181 
2182 	if (*pktopt == NULL) {
2183 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2184 		    M_WAITOK);
2185 		ip6_initpktopts(*pktopt);
2186 	}
2187 	opt = *pktopt;
2188 
2189 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2190 }
2191 
2192 static int
2193 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2194 {
2195 	void *optdata = NULL;
2196 	int optdatalen = 0;
2197 	struct ip6_ext *ip6e;
2198 	int error = 0;
2199 	struct in6_pktinfo null_pktinfo;
2200 	int deftclass = 0, on;
2201 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2202 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2203 
2204 	switch (optname) {
2205 	case IPV6_PKTINFO:
2206 		if (pktopt && pktopt->ip6po_pktinfo)
2207 			optdata = (void *)pktopt->ip6po_pktinfo;
2208 		else {
2209 			/* XXX: we don't have to do this every time... */
2210 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2211 			optdata = (void *)&null_pktinfo;
2212 		}
2213 		optdatalen = sizeof(struct in6_pktinfo);
2214 		break;
2215 	case IPV6_TCLASS:
2216 		if (pktopt && pktopt->ip6po_tclass >= 0)
2217 			optdata = (void *)&pktopt->ip6po_tclass;
2218 		else
2219 			optdata = (void *)&deftclass;
2220 		optdatalen = sizeof(int);
2221 		break;
2222 	case IPV6_HOPOPTS:
2223 		if (pktopt && pktopt->ip6po_hbh) {
2224 			optdata = (void *)pktopt->ip6po_hbh;
2225 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2226 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2227 		}
2228 		break;
2229 	case IPV6_RTHDR:
2230 		if (pktopt && pktopt->ip6po_rthdr) {
2231 			optdata = (void *)pktopt->ip6po_rthdr;
2232 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2233 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2234 		}
2235 		break;
2236 	case IPV6_RTHDRDSTOPTS:
2237 		if (pktopt && pktopt->ip6po_dest1) {
2238 			optdata = (void *)pktopt->ip6po_dest1;
2239 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2240 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2241 		}
2242 		break;
2243 	case IPV6_DSTOPTS:
2244 		if (pktopt && pktopt->ip6po_dest2) {
2245 			optdata = (void *)pktopt->ip6po_dest2;
2246 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2247 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2248 		}
2249 		break;
2250 	case IPV6_NEXTHOP:
2251 		if (pktopt && pktopt->ip6po_nexthop) {
2252 			optdata = (void *)pktopt->ip6po_nexthop;
2253 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2254 		}
2255 		break;
2256 	case IPV6_USE_MIN_MTU:
2257 		if (pktopt)
2258 			optdata = (void *)&pktopt->ip6po_minmtu;
2259 		else
2260 			optdata = (void *)&defminmtu;
2261 		optdatalen = sizeof(int);
2262 		break;
2263 	case IPV6_DONTFRAG:
2264 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2265 			on = 1;
2266 		else
2267 			on = 0;
2268 		optdata = (void *)&on;
2269 		optdatalen = sizeof(on);
2270 		break;
2271 	case IPV6_PREFER_TEMPADDR:
2272 		if (pktopt)
2273 			optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2274 		else
2275 			optdata = (void *)&defpreftemp;
2276 		optdatalen = sizeof(int);
2277 		break;
2278 	default:		/* should not happen */
2279 #ifdef DIAGNOSTIC
2280 		panic("ip6_getpcbopt: unexpected option\n");
2281 #endif
2282 		return (ENOPROTOOPT);
2283 	}
2284 
2285 	error = sooptcopyout(sopt, optdata, optdatalen);
2286 
2287 	return (error);
2288 }
2289 
2290 void
2291 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2292 {
2293 	if (pktopt == NULL)
2294 		return;
2295 
2296 	if (optname == -1 || optname == IPV6_PKTINFO) {
2297 		if (pktopt->ip6po_pktinfo)
2298 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2299 		pktopt->ip6po_pktinfo = NULL;
2300 	}
2301 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2302 		pktopt->ip6po_hlim = -1;
2303 	if (optname == -1 || optname == IPV6_TCLASS)
2304 		pktopt->ip6po_tclass = -1;
2305 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2306 		if (pktopt->ip6po_nextroute.ro_rt) {
2307 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2308 			pktopt->ip6po_nextroute.ro_rt = NULL;
2309 		}
2310 		if (pktopt->ip6po_nexthop)
2311 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2312 		pktopt->ip6po_nexthop = NULL;
2313 	}
2314 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2315 		if (pktopt->ip6po_hbh)
2316 			free(pktopt->ip6po_hbh, M_IP6OPT);
2317 		pktopt->ip6po_hbh = NULL;
2318 	}
2319 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2320 		if (pktopt->ip6po_dest1)
2321 			free(pktopt->ip6po_dest1, M_IP6OPT);
2322 		pktopt->ip6po_dest1 = NULL;
2323 	}
2324 	if (optname == -1 || optname == IPV6_RTHDR) {
2325 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2326 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2327 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2328 		if (pktopt->ip6po_route.ro_rt) {
2329 			RTFREE(pktopt->ip6po_route.ro_rt);
2330 			pktopt->ip6po_route.ro_rt = NULL;
2331 		}
2332 	}
2333 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2334 		if (pktopt->ip6po_dest2)
2335 			free(pktopt->ip6po_dest2, M_IP6OPT);
2336 		pktopt->ip6po_dest2 = NULL;
2337 	}
2338 }
2339 
2340 #define PKTOPT_EXTHDRCPY(type) \
2341 do {\
2342 	if (src->type) {\
2343 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2344 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2345 		if (dst->type == NULL && canwait == M_NOWAIT)\
2346 			goto bad;\
2347 		bcopy(src->type, dst->type, hlen);\
2348 	}\
2349 } while (/*CONSTCOND*/ 0)
2350 
2351 static int
2352 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2353 {
2354 	if (dst == NULL || src == NULL)  {
2355 		printf("ip6_clearpktopts: invalid argument\n");
2356 		return (EINVAL);
2357 	}
2358 
2359 	dst->ip6po_hlim = src->ip6po_hlim;
2360 	dst->ip6po_tclass = src->ip6po_tclass;
2361 	dst->ip6po_flags = src->ip6po_flags;
2362 	if (src->ip6po_pktinfo) {
2363 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2364 		    M_IP6OPT, canwait);
2365 		if (dst->ip6po_pktinfo == NULL)
2366 			goto bad;
2367 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2368 	}
2369 	if (src->ip6po_nexthop) {
2370 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2371 		    M_IP6OPT, canwait);
2372 		if (dst->ip6po_nexthop == NULL)
2373 			goto bad;
2374 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2375 		    src->ip6po_nexthop->sa_len);
2376 	}
2377 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2378 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2379 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2380 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2381 	return (0);
2382 
2383   bad:
2384 	ip6_clearpktopts(dst, -1);
2385 	return (ENOBUFS);
2386 }
2387 #undef PKTOPT_EXTHDRCPY
2388 
2389 struct ip6_pktopts *
2390 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2391 {
2392 	int error;
2393 	struct ip6_pktopts *dst;
2394 
2395 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2396 	if (dst == NULL)
2397 		return (NULL);
2398 	ip6_initpktopts(dst);
2399 
2400 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2401 		free(dst, M_IP6OPT);
2402 		return (NULL);
2403 	}
2404 
2405 	return (dst);
2406 }
2407 
2408 void
2409 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2410 {
2411 	if (pktopt == NULL)
2412 		return;
2413 
2414 	ip6_clearpktopts(pktopt, -1);
2415 
2416 	free(pktopt, M_IP6OPT);
2417 }
2418 
2419 /*
2420  * Set the IP6 multicast options in response to user setsockopt().
2421  */
2422 static int
2423 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
2424 {
2425 	INIT_VNET_NET(curvnet);
2426 	INIT_VNET_INET6(curvnet);
2427 	int error = 0;
2428 	u_int loop, ifindex;
2429 	struct ipv6_mreq *mreq;
2430 	struct ifnet *ifp;
2431 	struct ip6_moptions *im6o = *im6op;
2432 	struct route_in6 ro;
2433 	struct in6_multi_mship *imm;
2434 
2435 	if (im6o == NULL) {
2436 		/*
2437 		 * No multicast option buffer attached to the pcb;
2438 		 * allocate one and initialize to default values.
2439 		 */
2440 		im6o = (struct ip6_moptions *)
2441 			malloc(sizeof(*im6o), M_IP6MOPTS, M_WAITOK);
2442 
2443 		if (im6o == NULL)
2444 			return (ENOBUFS);
2445 		*im6op = im6o;
2446 		im6o->im6o_multicast_ifp = NULL;
2447 		im6o->im6o_multicast_hlim = V_ip6_defmcasthlim;
2448 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2449 		LIST_INIT(&im6o->im6o_memberships);
2450 	}
2451 
2452 	switch (optname) {
2453 
2454 	case IPV6_MULTICAST_IF:
2455 		/*
2456 		 * Select the interface for outgoing multicast packets.
2457 		 */
2458 		if (m == NULL || m->m_len != sizeof(u_int)) {
2459 			error = EINVAL;
2460 			break;
2461 		}
2462 		bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
2463 		if (ifindex < 0 || V_if_index < ifindex) {
2464 			error = ENXIO;	/* XXX EINVAL? */
2465 			break;
2466 		}
2467 		ifp = ifnet_byindex(ifindex);
2468 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2469 			error = EADDRNOTAVAIL;
2470 			break;
2471 		}
2472 		im6o->im6o_multicast_ifp = ifp;
2473 		break;
2474 
2475 	case IPV6_MULTICAST_HOPS:
2476 	    {
2477 		/*
2478 		 * Set the IP6 hoplimit for outgoing multicast packets.
2479 		 */
2480 		int optval;
2481 		if (m == NULL || m->m_len != sizeof(int)) {
2482 			error = EINVAL;
2483 			break;
2484 		}
2485 		bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2486 		if (optval < -1 || optval >= 256)
2487 			error = EINVAL;
2488 		else if (optval == -1)
2489 			im6o->im6o_multicast_hlim = V_ip6_defmcasthlim;
2490 		else
2491 			im6o->im6o_multicast_hlim = optval;
2492 		break;
2493 	    }
2494 
2495 	case IPV6_MULTICAST_LOOP:
2496 		/*
2497 		 * Set the loopback flag for outgoing multicast packets.
2498 		 * Must be zero or one.
2499 		 */
2500 		if (m == NULL || m->m_len != sizeof(u_int)) {
2501 			error = EINVAL;
2502 			break;
2503 		}
2504 		bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2505 		if (loop > 1) {
2506 			error = EINVAL;
2507 			break;
2508 		}
2509 		im6o->im6o_multicast_loop = loop;
2510 		break;
2511 
2512 	case IPV6_JOIN_GROUP:
2513 		/*
2514 		 * Add a multicast group membership.
2515 		 * Group must be a valid IP6 multicast address.
2516 		 */
2517 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2518 			error = EINVAL;
2519 			break;
2520 		}
2521 		mreq = mtod(m, struct ipv6_mreq *);
2522 
2523 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2524 			/*
2525 			 * We use the unspecified address to specify to accept
2526 			 * all multicast addresses. Only super user is allowed
2527 			 * to do this.
2528 			 */
2529 			/* XXX-BZ might need a better PRIV_NETINET_x for this */
2530 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
2531 			if (error)
2532 				break;
2533 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2534 			error = EINVAL;
2535 			break;
2536 		}
2537 
2538 		/*
2539 		 * If no interface was explicitly specified, choose an
2540 		 * appropriate one according to the given multicast address.
2541 		 */
2542 		if (mreq->ipv6mr_interface == 0) {
2543 			struct sockaddr_in6 *dst;
2544 
2545 			/*
2546 			 * Look up the routing table for the
2547 			 * address, and choose the outgoing interface.
2548 			 *   XXX: is it a good approach?
2549 			 */
2550 			ro.ro_rt = NULL;
2551 			dst = (struct sockaddr_in6 *)&ro.ro_dst;
2552 			bzero(dst, sizeof(*dst));
2553 			dst->sin6_family = AF_INET6;
2554 			dst->sin6_len = sizeof(*dst);
2555 			dst->sin6_addr = mreq->ipv6mr_multiaddr;
2556 			rtalloc((struct route *)&ro);
2557 			if (ro.ro_rt == NULL) {
2558 				error = EADDRNOTAVAIL;
2559 				break;
2560 			}
2561 			ifp = ro.ro_rt->rt_ifp;
2562 			RTFREE(ro.ro_rt);
2563 		} else {
2564 			/*
2565 			 * If the interface is specified, validate it.
2566 			 */
2567 			if (mreq->ipv6mr_interface < 0 ||
2568 			    V_if_index < mreq->ipv6mr_interface) {
2569 				error = ENXIO;	/* XXX EINVAL? */
2570 				break;
2571 			}
2572 			ifp = ifnet_byindex(mreq->ipv6mr_interface);
2573 			if (!ifp) {
2574 				error = ENXIO;	/* XXX EINVAL? */
2575 				break;
2576 			}
2577 		}
2578 
2579 		/*
2580 		 * See if we found an interface, and confirm that it
2581 		 * supports multicast
2582 		 */
2583 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2584 			error = EADDRNOTAVAIL;
2585 			break;
2586 		}
2587 
2588 		if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) {
2589 			error = EADDRNOTAVAIL; /* XXX: should not happen */
2590 			break;
2591 		}
2592 
2593 		/*
2594 		 * See if the membership already exists.
2595 		 */
2596 		for (imm = im6o->im6o_memberships.lh_first;
2597 		     imm != NULL; imm = imm->i6mm_chain.le_next)
2598 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
2599 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2600 					       &mreq->ipv6mr_multiaddr))
2601 				break;
2602 		if (imm != NULL) {
2603 			error = EADDRINUSE;
2604 			break;
2605 		}
2606 		/*
2607 		 * Everything looks good; add a new record to the multicast
2608 		 * address list for the given interface.
2609 		 */
2610 		imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr,  &error, 0);
2611 		if (imm == NULL)
2612 			break;
2613 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2614 		break;
2615 
2616 	case IPV6_LEAVE_GROUP:
2617 		/*
2618 		 * Drop a multicast group membership.
2619 		 * Group must be a valid IP6 multicast address.
2620 		 */
2621 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2622 			error = EINVAL;
2623 			break;
2624 		}
2625 		mreq = mtod(m, struct ipv6_mreq *);
2626 
2627 		/*
2628 		 * If an interface address was specified, get a pointer
2629 		 * to its ifnet structure.
2630 		 */
2631 		if (mreq->ipv6mr_interface < 0 ||
2632 		    V_if_index < mreq->ipv6mr_interface) {
2633 			error = ENXIO;	/* XXX EINVAL? */
2634 			break;
2635 		}
2636 		if (mreq->ipv6mr_interface == 0)
2637 			ifp = NULL;
2638 		else
2639 			ifp = ifnet_byindex(mreq->ipv6mr_interface);
2640 
2641 		/* Fill in the scope zone ID */
2642 		if (ifp) {
2643 			if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) {
2644 				/* XXX: should not happen */
2645 				error = EADDRNOTAVAIL;
2646 				break;
2647 			}
2648 		} else if (mreq->ipv6mr_interface != 0) {
2649 			/*
2650 			 * This case happens when the (positive) index is in
2651 			 * the valid range, but the corresponding interface has
2652 			 * been detached dynamically (XXX).
2653 			 */
2654 			error = EADDRNOTAVAIL;
2655 			break;
2656 		} else {	/* ipv6mr_interface == 0 */
2657 			struct sockaddr_in6 sa6_mc;
2658 
2659 			/*
2660 			 * The API spec says as follows:
2661 			 *  If the interface index is specified as 0, the
2662 			 *  system may choose a multicast group membership to
2663 			 *  drop by matching the multicast address only.
2664 			 * On the other hand, we cannot disambiguate the scope
2665 			 * zone unless an interface is provided.  Thus, we
2666 			 * check if there's ambiguity with the default scope
2667 			 * zone as the last resort.
2668 			 */
2669 			bzero(&sa6_mc, sizeof(sa6_mc));
2670 			sa6_mc.sin6_family = AF_INET6;
2671 			sa6_mc.sin6_len = sizeof(sa6_mc);
2672 			sa6_mc.sin6_addr = mreq->ipv6mr_multiaddr;
2673 			error = sa6_embedscope(&sa6_mc, V_ip6_use_defzone);
2674 			if (error != 0)
2675 				break;
2676 			mreq->ipv6mr_multiaddr = sa6_mc.sin6_addr;
2677 		}
2678 
2679 		/*
2680 		 * Find the membership in the membership list.
2681 		 */
2682 		for (imm = im6o->im6o_memberships.lh_first;
2683 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
2684 			if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2685 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2686 			    &mreq->ipv6mr_multiaddr))
2687 				break;
2688 		}
2689 		if (imm == NULL) {
2690 			/* Unable to resolve interface */
2691 			error = EADDRNOTAVAIL;
2692 			break;
2693 		}
2694 		/*
2695 		 * Give up the multicast address record to which the
2696 		 * membership points.
2697 		 */
2698 		LIST_REMOVE(imm, i6mm_chain);
2699 		in6_delmulti(imm->i6mm_maddr);
2700 		free(imm, M_IP6MADDR);
2701 		break;
2702 
2703 	default:
2704 		error = EOPNOTSUPP;
2705 		break;
2706 	}
2707 
2708 	/*
2709 	 * If all options have default values, no need to keep the mbuf.
2710 	 */
2711 	if (im6o->im6o_multicast_ifp == NULL &&
2712 	    im6o->im6o_multicast_hlim == V_ip6_defmcasthlim &&
2713 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2714 	    im6o->im6o_memberships.lh_first == NULL) {
2715 		free(*im6op, M_IP6MOPTS);
2716 		*im6op = NULL;
2717 	}
2718 
2719 	return (error);
2720 }
2721 
2722 /*
2723  * Return the IP6 multicast options in response to user getsockopt().
2724  */
2725 static int
2726 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp)
2727 {
2728 	INIT_VNET_INET6(curvnet);
2729 	u_int *hlim, *loop, *ifindex;
2730 
2731 	*mp = m_get(M_WAIT, MT_HEADER);		/* XXX */
2732 
2733 	switch (optname) {
2734 
2735 	case IPV6_MULTICAST_IF:
2736 		ifindex = mtod(*mp, u_int *);
2737 		(*mp)->m_len = sizeof(u_int);
2738 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2739 			*ifindex = 0;
2740 		else
2741 			*ifindex = im6o->im6o_multicast_ifp->if_index;
2742 		return (0);
2743 
2744 	case IPV6_MULTICAST_HOPS:
2745 		hlim = mtod(*mp, u_int *);
2746 		(*mp)->m_len = sizeof(u_int);
2747 		if (im6o == NULL)
2748 			*hlim = V_ip6_defmcasthlim;
2749 		else
2750 			*hlim = im6o->im6o_multicast_hlim;
2751 		return (0);
2752 
2753 	case IPV6_MULTICAST_LOOP:
2754 		loop = mtod(*mp, u_int *);
2755 		(*mp)->m_len = sizeof(u_int);
2756 		if (im6o == NULL)
2757 			*loop = V_ip6_defmcasthlim;
2758 		else
2759 			*loop = im6o->im6o_multicast_loop;
2760 		return (0);
2761 
2762 	default:
2763 		return (EOPNOTSUPP);
2764 	}
2765 }
2766 
2767 /*
2768  * Discard the IP6 multicast options.
2769  */
2770 void
2771 ip6_freemoptions(struct ip6_moptions *im6o)
2772 {
2773 	struct in6_multi_mship *imm;
2774 
2775 	if (im6o == NULL)
2776 		return;
2777 
2778 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2779 		LIST_REMOVE(imm, i6mm_chain);
2780 		if (imm->i6mm_maddr)
2781 			in6_delmulti(imm->i6mm_maddr);
2782 		free(imm, M_IP6MADDR);
2783 	}
2784 	free(im6o, M_IP6MOPTS);
2785 }
2786 
2787 /*
2788  * Set IPv6 outgoing packet options based on advanced API.
2789  */
2790 int
2791 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2792     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2793 {
2794 	struct cmsghdr *cm = 0;
2795 
2796 	if (control == NULL || opt == NULL)
2797 		return (EINVAL);
2798 
2799 	ip6_initpktopts(opt);
2800 	if (stickyopt) {
2801 		int error;
2802 
2803 		/*
2804 		 * If stickyopt is provided, make a local copy of the options
2805 		 * for this particular packet, then override them by ancillary
2806 		 * objects.
2807 		 * XXX: copypktopts() does not copy the cached route to a next
2808 		 * hop (if any).  This is not very good in terms of efficiency,
2809 		 * but we can allow this since this option should be rarely
2810 		 * used.
2811 		 */
2812 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2813 			return (error);
2814 	}
2815 
2816 	/*
2817 	 * XXX: Currently, we assume all the optional information is stored
2818 	 * in a single mbuf.
2819 	 */
2820 	if (control->m_next)
2821 		return (EINVAL);
2822 
2823 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2824 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2825 		int error;
2826 
2827 		if (control->m_len < CMSG_LEN(0))
2828 			return (EINVAL);
2829 
2830 		cm = mtod(control, struct cmsghdr *);
2831 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2832 			return (EINVAL);
2833 		if (cm->cmsg_level != IPPROTO_IPV6)
2834 			continue;
2835 
2836 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2837 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2838 		if (error)
2839 			return (error);
2840 	}
2841 
2842 	return (0);
2843 }
2844 
2845 /*
2846  * Set a particular packet option, as a sticky option or an ancillary data
2847  * item.  "len" can be 0 only when it's a sticky option.
2848  * We have 4 cases of combination of "sticky" and "cmsg":
2849  * "sticky=0, cmsg=0": impossible
2850  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2851  * "sticky=1, cmsg=0": RFC3542 socket option
2852  * "sticky=1, cmsg=1": RFC2292 socket option
2853  */
2854 static int
2855 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2856     struct ucred *cred, int sticky, int cmsg, int uproto)
2857 {
2858 	INIT_VNET_NET(curvnet);
2859 	INIT_VNET_INET6(curvnet);
2860 	int minmtupolicy, preftemp;
2861 	int error;
2862 
2863 	if (!sticky && !cmsg) {
2864 #ifdef DIAGNOSTIC
2865 		printf("ip6_setpktopt: impossible case\n");
2866 #endif
2867 		return (EINVAL);
2868 	}
2869 
2870 	/*
2871 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2872 	 * not be specified in the context of RFC3542.  Conversely,
2873 	 * RFC3542 types should not be specified in the context of RFC2292.
2874 	 */
2875 	if (!cmsg) {
2876 		switch (optname) {
2877 		case IPV6_2292PKTINFO:
2878 		case IPV6_2292HOPLIMIT:
2879 		case IPV6_2292NEXTHOP:
2880 		case IPV6_2292HOPOPTS:
2881 		case IPV6_2292DSTOPTS:
2882 		case IPV6_2292RTHDR:
2883 		case IPV6_2292PKTOPTIONS:
2884 			return (ENOPROTOOPT);
2885 		}
2886 	}
2887 	if (sticky && cmsg) {
2888 		switch (optname) {
2889 		case IPV6_PKTINFO:
2890 		case IPV6_HOPLIMIT:
2891 		case IPV6_NEXTHOP:
2892 		case IPV6_HOPOPTS:
2893 		case IPV6_DSTOPTS:
2894 		case IPV6_RTHDRDSTOPTS:
2895 		case IPV6_RTHDR:
2896 		case IPV6_USE_MIN_MTU:
2897 		case IPV6_DONTFRAG:
2898 		case IPV6_TCLASS:
2899 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2900 			return (ENOPROTOOPT);
2901 		}
2902 	}
2903 
2904 	switch (optname) {
2905 	case IPV6_2292PKTINFO:
2906 	case IPV6_PKTINFO:
2907 	{
2908 		struct ifnet *ifp = NULL;
2909 		struct in6_pktinfo *pktinfo;
2910 
2911 		if (len != sizeof(struct in6_pktinfo))
2912 			return (EINVAL);
2913 
2914 		pktinfo = (struct in6_pktinfo *)buf;
2915 
2916 		/*
2917 		 * An application can clear any sticky IPV6_PKTINFO option by
2918 		 * doing a "regular" setsockopt with ipi6_addr being
2919 		 * in6addr_any and ipi6_ifindex being zero.
2920 		 * [RFC 3542, Section 6]
2921 		 */
2922 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2923 		    pktinfo->ipi6_ifindex == 0 &&
2924 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2925 			ip6_clearpktopts(opt, optname);
2926 			break;
2927 		}
2928 
2929 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2930 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2931 			return (EINVAL);
2932 		}
2933 
2934 		/* validate the interface index if specified. */
2935 		if (pktinfo->ipi6_ifindex > V_if_index ||
2936 		    pktinfo->ipi6_ifindex < 0) {
2937 			 return (ENXIO);
2938 		}
2939 		if (pktinfo->ipi6_ifindex) {
2940 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2941 			if (ifp == NULL)
2942 				return (ENXIO);
2943 		}
2944 
2945 		/*
2946 		 * We store the address anyway, and let in6_selectsrc()
2947 		 * validate the specified address.  This is because ipi6_addr
2948 		 * may not have enough information about its scope zone, and
2949 		 * we may need additional information (such as outgoing
2950 		 * interface or the scope zone of a destination address) to
2951 		 * disambiguate the scope.
2952 		 * XXX: the delay of the validation may confuse the
2953 		 * application when it is used as a sticky option.
2954 		 */
2955 		if (opt->ip6po_pktinfo == NULL) {
2956 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2957 			    M_IP6OPT, M_NOWAIT);
2958 			if (opt->ip6po_pktinfo == NULL)
2959 				return (ENOBUFS);
2960 		}
2961 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2962 		break;
2963 	}
2964 
2965 	case IPV6_2292HOPLIMIT:
2966 	case IPV6_HOPLIMIT:
2967 	{
2968 		int *hlimp;
2969 
2970 		/*
2971 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2972 		 * to simplify the ordering among hoplimit options.
2973 		 */
2974 		if (optname == IPV6_HOPLIMIT && sticky)
2975 			return (ENOPROTOOPT);
2976 
2977 		if (len != sizeof(int))
2978 			return (EINVAL);
2979 		hlimp = (int *)buf;
2980 		if (*hlimp < -1 || *hlimp > 255)
2981 			return (EINVAL);
2982 
2983 		opt->ip6po_hlim = *hlimp;
2984 		break;
2985 	}
2986 
2987 	case IPV6_TCLASS:
2988 	{
2989 		int tclass;
2990 
2991 		if (len != sizeof(int))
2992 			return (EINVAL);
2993 		tclass = *(int *)buf;
2994 		if (tclass < -1 || tclass > 255)
2995 			return (EINVAL);
2996 
2997 		opt->ip6po_tclass = tclass;
2998 		break;
2999 	}
3000 
3001 	case IPV6_2292NEXTHOP:
3002 	case IPV6_NEXTHOP:
3003 		if (cred != NULL) {
3004 			error = priv_check_cred(cred,
3005 			    PRIV_NETINET_SETHDROPTS, 0);
3006 			if (error)
3007 				return (error);
3008 		}
3009 
3010 		if (len == 0) {	/* just remove the option */
3011 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
3012 			break;
3013 		}
3014 
3015 		/* check if cmsg_len is large enough for sa_len */
3016 		if (len < sizeof(struct sockaddr) || len < *buf)
3017 			return (EINVAL);
3018 
3019 		switch (((struct sockaddr *)buf)->sa_family) {
3020 		case AF_INET6:
3021 		{
3022 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3023 			int error;
3024 
3025 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3026 				return (EINVAL);
3027 
3028 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3029 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3030 				return (EINVAL);
3031 			}
3032 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
3033 			    != 0) {
3034 				return (error);
3035 			}
3036 			break;
3037 		}
3038 		case AF_LINK:	/* should eventually be supported */
3039 		default:
3040 			return (EAFNOSUPPORT);
3041 		}
3042 
3043 		/* turn off the previous option, then set the new option. */
3044 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
3045 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3046 		if (opt->ip6po_nexthop == NULL)
3047 			return (ENOBUFS);
3048 		bcopy(buf, opt->ip6po_nexthop, *buf);
3049 		break;
3050 
3051 	case IPV6_2292HOPOPTS:
3052 	case IPV6_HOPOPTS:
3053 	{
3054 		struct ip6_hbh *hbh;
3055 		int hbhlen;
3056 
3057 		/*
3058 		 * XXX: We don't allow a non-privileged user to set ANY HbH
3059 		 * options, since per-option restriction has too much
3060 		 * overhead.
3061 		 */
3062 		if (cred != NULL) {
3063 			error = priv_check_cred(cred,
3064 			    PRIV_NETINET_SETHDROPTS, 0);
3065 			if (error)
3066 				return (error);
3067 		}
3068 
3069 		if (len == 0) {
3070 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
3071 			break;	/* just remove the option */
3072 		}
3073 
3074 		/* message length validation */
3075 		if (len < sizeof(struct ip6_hbh))
3076 			return (EINVAL);
3077 		hbh = (struct ip6_hbh *)buf;
3078 		hbhlen = (hbh->ip6h_len + 1) << 3;
3079 		if (len != hbhlen)
3080 			return (EINVAL);
3081 
3082 		/* turn off the previous option, then set the new option. */
3083 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
3084 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3085 		if (opt->ip6po_hbh == NULL)
3086 			return (ENOBUFS);
3087 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
3088 
3089 		break;
3090 	}
3091 
3092 	case IPV6_2292DSTOPTS:
3093 	case IPV6_DSTOPTS:
3094 	case IPV6_RTHDRDSTOPTS:
3095 	{
3096 		struct ip6_dest *dest, **newdest = NULL;
3097 		int destlen;
3098 
3099 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
3100 			error = priv_check_cred(cred,
3101 			    PRIV_NETINET_SETHDROPTS, 0);
3102 			if (error)
3103 				return (error);
3104 		}
3105 
3106 		if (len == 0) {
3107 			ip6_clearpktopts(opt, optname);
3108 			break;	/* just remove the option */
3109 		}
3110 
3111 		/* message length validation */
3112 		if (len < sizeof(struct ip6_dest))
3113 			return (EINVAL);
3114 		dest = (struct ip6_dest *)buf;
3115 		destlen = (dest->ip6d_len + 1) << 3;
3116 		if (len != destlen)
3117 			return (EINVAL);
3118 
3119 		/*
3120 		 * Determine the position that the destination options header
3121 		 * should be inserted; before or after the routing header.
3122 		 */
3123 		switch (optname) {
3124 		case IPV6_2292DSTOPTS:
3125 			/*
3126 			 * The old advacned API is ambiguous on this point.
3127 			 * Our approach is to determine the position based
3128 			 * according to the existence of a routing header.
3129 			 * Note, however, that this depends on the order of the
3130 			 * extension headers in the ancillary data; the 1st
3131 			 * part of the destination options header must appear
3132 			 * before the routing header in the ancillary data,
3133 			 * too.
3134 			 * RFC3542 solved the ambiguity by introducing
3135 			 * separate ancillary data or option types.
3136 			 */
3137 			if (opt->ip6po_rthdr == NULL)
3138 				newdest = &opt->ip6po_dest1;
3139 			else
3140 				newdest = &opt->ip6po_dest2;
3141 			break;
3142 		case IPV6_RTHDRDSTOPTS:
3143 			newdest = &opt->ip6po_dest1;
3144 			break;
3145 		case IPV6_DSTOPTS:
3146 			newdest = &opt->ip6po_dest2;
3147 			break;
3148 		}
3149 
3150 		/* turn off the previous option, then set the new option. */
3151 		ip6_clearpktopts(opt, optname);
3152 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3153 		if (*newdest == NULL)
3154 			return (ENOBUFS);
3155 		bcopy(dest, *newdest, destlen);
3156 
3157 		break;
3158 	}
3159 
3160 	case IPV6_2292RTHDR:
3161 	case IPV6_RTHDR:
3162 	{
3163 		struct ip6_rthdr *rth;
3164 		int rthlen;
3165 
3166 		if (len == 0) {
3167 			ip6_clearpktopts(opt, IPV6_RTHDR);
3168 			break;	/* just remove the option */
3169 		}
3170 
3171 		/* message length validation */
3172 		if (len < sizeof(struct ip6_rthdr))
3173 			return (EINVAL);
3174 		rth = (struct ip6_rthdr *)buf;
3175 		rthlen = (rth->ip6r_len + 1) << 3;
3176 		if (len != rthlen)
3177 			return (EINVAL);
3178 
3179 		switch (rth->ip6r_type) {
3180 		case IPV6_RTHDR_TYPE_0:
3181 			if (rth->ip6r_len == 0)	/* must contain one addr */
3182 				return (EINVAL);
3183 			if (rth->ip6r_len % 2) /* length must be even */
3184 				return (EINVAL);
3185 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3186 				return (EINVAL);
3187 			break;
3188 		default:
3189 			return (EINVAL);	/* not supported */
3190 		}
3191 
3192 		/* turn off the previous option */
3193 		ip6_clearpktopts(opt, IPV6_RTHDR);
3194 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3195 		if (opt->ip6po_rthdr == NULL)
3196 			return (ENOBUFS);
3197 		bcopy(rth, opt->ip6po_rthdr, rthlen);
3198 
3199 		break;
3200 	}
3201 
3202 	case IPV6_USE_MIN_MTU:
3203 		if (len != sizeof(int))
3204 			return (EINVAL);
3205 		minmtupolicy = *(int *)buf;
3206 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3207 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3208 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3209 			return (EINVAL);
3210 		}
3211 		opt->ip6po_minmtu = minmtupolicy;
3212 		break;
3213 
3214 	case IPV6_DONTFRAG:
3215 		if (len != sizeof(int))
3216 			return (EINVAL);
3217 
3218 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3219 			/*
3220 			 * we ignore this option for TCP sockets.
3221 			 * (RFC3542 leaves this case unspecified.)
3222 			 */
3223 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3224 		} else
3225 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3226 		break;
3227 
3228 	case IPV6_PREFER_TEMPADDR:
3229 		if (len != sizeof(int))
3230 			return (EINVAL);
3231 		preftemp = *(int *)buf;
3232 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3233 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3234 		    preftemp != IP6PO_TEMPADDR_PREFER) {
3235 			return (EINVAL);
3236 		}
3237 		opt->ip6po_prefer_tempaddr = preftemp;
3238 		break;
3239 
3240 	default:
3241 		return (ENOPROTOOPT);
3242 	} /* end of switch */
3243 
3244 	return (0);
3245 }
3246 
3247 /*
3248  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3249  * packet to the input queue of a specified interface.  Note that this
3250  * calls the output routine of the loopback "driver", but with an interface
3251  * pointer that might NOT be &loif -- easier than replicating that code here.
3252  */
3253 void
3254 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
3255 {
3256 	struct mbuf *copym;
3257 	struct ip6_hdr *ip6;
3258 
3259 	copym = m_copy(m, 0, M_COPYALL);
3260 	if (copym == NULL)
3261 		return;
3262 
3263 	/*
3264 	 * Make sure to deep-copy IPv6 header portion in case the data
3265 	 * is in an mbuf cluster, so that we can safely override the IPv6
3266 	 * header portion later.
3267 	 */
3268 	if ((copym->m_flags & M_EXT) != 0 ||
3269 	    copym->m_len < sizeof(struct ip6_hdr)) {
3270 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3271 		if (copym == NULL)
3272 			return;
3273 	}
3274 
3275 #ifdef DIAGNOSTIC
3276 	if (copym->m_len < sizeof(*ip6)) {
3277 		m_freem(copym);
3278 		return;
3279 	}
3280 #endif
3281 
3282 	ip6 = mtod(copym, struct ip6_hdr *);
3283 	/*
3284 	 * clear embedded scope identifiers if necessary.
3285 	 * in6_clearscope will touch the addresses only when necessary.
3286 	 */
3287 	in6_clearscope(&ip6->ip6_src);
3288 	in6_clearscope(&ip6->ip6_dst);
3289 
3290 	(void)if_simloop(ifp, copym, dst->sin6_family, 0);
3291 }
3292 
3293 /*
3294  * Chop IPv6 header off from the payload.
3295  */
3296 static int
3297 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3298 {
3299 	struct mbuf *mh;
3300 	struct ip6_hdr *ip6;
3301 
3302 	ip6 = mtod(m, struct ip6_hdr *);
3303 	if (m->m_len > sizeof(*ip6)) {
3304 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3305 		if (mh == 0) {
3306 			m_freem(m);
3307 			return ENOBUFS;
3308 		}
3309 		M_MOVE_PKTHDR(mh, m);
3310 		MH_ALIGN(mh, sizeof(*ip6));
3311 		m->m_len -= sizeof(*ip6);
3312 		m->m_data += sizeof(*ip6);
3313 		mh->m_next = m;
3314 		m = mh;
3315 		m->m_len = sizeof(*ip6);
3316 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3317 	}
3318 	exthdrs->ip6e_ip6 = m;
3319 	return 0;
3320 }
3321 
3322 /*
3323  * Compute IPv6 extension header length.
3324  */
3325 int
3326 ip6_optlen(struct in6pcb *in6p)
3327 {
3328 	int len;
3329 
3330 	if (!in6p->in6p_outputopts)
3331 		return 0;
3332 
3333 	len = 0;
3334 #define elen(x) \
3335     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3336 
3337 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3338 	if (in6p->in6p_outputopts->ip6po_rthdr)
3339 		/* dest1 is valid with rthdr only */
3340 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
3341 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3342 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3343 	return len;
3344 #undef elen
3345 }
3346