xref: /netbsd-src/sys/netinet6/ip6_output.c (revision b1c86f5f087524e68db12794ee9c3e3da1ab17a0)
1 /*	$NetBSD: ip6_output.c,v 1.139 2009/05/07 21:51:47 elad Exp $	*/
2 /*	$KAME: ip6_output.c,v 1.172 2001/03/25 09:55:56 itojun Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1988, 1990, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
62  */
63 
64 #include <sys/cdefs.h>
65 __KERNEL_RCSID(0, "$NetBSD: ip6_output.c,v 1.139 2009/05/07 21:51:47 elad Exp $");
66 
67 #include "opt_inet.h"
68 #include "opt_inet6.h"
69 #include "opt_ipsec.h"
70 #include "opt_pfil_hooks.h"
71 
72 #include <sys/param.h>
73 #include <sys/malloc.h>
74 #include <sys/mbuf.h>
75 #include <sys/errno.h>
76 #include <sys/protosw.h>
77 #include <sys/socket.h>
78 #include <sys/socketvar.h>
79 #include <sys/systm.h>
80 #include <sys/proc.h>
81 #include <sys/kauth.h>
82 
83 #include <net/if.h>
84 #include <net/route.h>
85 #ifdef PFIL_HOOKS
86 #include <net/pfil.h>
87 #endif
88 
89 #include <netinet/in.h>
90 #include <netinet/in_var.h>
91 #include <netinet/ip6.h>
92 #include <netinet/icmp6.h>
93 #include <netinet/in_offload.h>
94 #include <netinet6/in6_offload.h>
95 #include <netinet6/ip6_var.h>
96 #include <netinet6/ip6_private.h>
97 #include <netinet6/in6_pcb.h>
98 #include <netinet6/nd6.h>
99 #include <netinet6/ip6protosw.h>
100 #include <netinet6/scope6_var.h>
101 
102 #ifdef IPSEC
103 #include <netinet6/ipsec.h>
104 #include <netinet6/ipsec_private.h>
105 #include <netkey/key.h>
106 #endif /* IPSEC */
107 
108 #ifdef FAST_IPSEC
109 #include <netipsec/ipsec.h>
110 #include <netipsec/ipsec6.h>
111 #include <netipsec/key.h>
112 #include <netipsec/xform.h>
113 #endif
114 
115 
116 #include <net/net_osdep.h>
117 
118 #ifdef PFIL_HOOKS
119 extern struct pfil_head inet6_pfil_hook;	/* XXX */
120 #endif
121 
122 struct ip6_exthdrs {
123 	struct mbuf *ip6e_ip6;
124 	struct mbuf *ip6e_hbh;
125 	struct mbuf *ip6e_dest1;
126 	struct mbuf *ip6e_rthdr;
127 	struct mbuf *ip6e_dest2;
128 };
129 
130 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
131 	kauth_cred_t, int);
132 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
133 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, kauth_cred_t,
134 	int, int, int);
135 static int ip6_setmoptions(const struct sockopt *, struct ip6_moptions **);
136 static int ip6_getmoptions(struct sockopt *, struct ip6_moptions *);
137 static int ip6_copyexthdr(struct mbuf **, void *, int);
138 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
139 	struct ip6_frag **);
140 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
141 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
142 static int ip6_getpmtu(struct route *, struct route *, struct ifnet *,
143     const struct in6_addr *, u_long *, int *);
144 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
145 
146 #ifdef RFC2292
147 static int ip6_pcbopts(struct ip6_pktopts **, struct socket *, struct sockopt *);
148 #endif
149 
150 #define	IN6_NEED_CHECKSUM(ifp, csum_flags) \
151 	(__predict_true(((ifp)->if_flags & IFF_LOOPBACK) == 0 || \
152 	(((csum_flags) & M_CSUM_UDPv6) != 0 && udp_do_loopback_cksum) || \
153 	(((csum_flags) & M_CSUM_TCPv6) != 0 && tcp_do_loopback_cksum)))
154 
155 /*
156  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
157  * header (with pri, len, nxt, hlim, src, dst).
158  * This function may modify ver and hlim only.
159  * The mbuf chain containing the packet will be freed.
160  * The mbuf opt, if present, will not be freed.
161  *
162  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
163  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
164  * which is rt_rmx.rmx_mtu.
165  */
166 int
167 ip6_output(
168     struct mbuf *m0,
169     struct ip6_pktopts *opt,
170     struct route *ro,
171     int flags,
172     struct ip6_moptions *im6o,
173     struct socket *so,
174     struct ifnet **ifpp		/* XXX: just for statistics */
175 )
176 {
177 	struct ip6_hdr *ip6, *mhip6;
178 	struct ifnet *ifp, *origifp;
179 	struct mbuf *m = m0;
180 	int hlen, tlen, len, off;
181 	bool tso;
182 	struct route ip6route;
183 	struct rtentry *rt = NULL;
184 	const struct sockaddr_in6 *dst = NULL;
185 	struct sockaddr_in6 src_sa, dst_sa;
186 	int error = 0;
187 	struct in6_ifaddr *ia = NULL;
188 	u_long mtu;
189 	int alwaysfrag, dontfrag;
190 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
191 	struct ip6_exthdrs exthdrs;
192 	struct in6_addr finaldst, src0, dst0;
193 	u_int32_t zone;
194 	struct route *ro_pmtu = NULL;
195 	int hdrsplit = 0;
196 	int needipsec = 0;
197 #ifdef IPSEC
198 	int needipsectun = 0;
199 	struct secpolicy *sp = NULL;
200 
201 	ip6 = mtod(m, struct ip6_hdr *);
202 #endif /* IPSEC */
203 #ifdef FAST_IPSEC
204 	struct secpolicy *sp = NULL;
205 	int s;
206 #endif
207 
208 	memset(&ip6route, 0, sizeof(ip6route));
209 
210 #ifdef  DIAGNOSTIC
211 	if ((m->m_flags & M_PKTHDR) == 0)
212 		panic("ip6_output: no HDR");
213 
214 	if ((m->m_pkthdr.csum_flags &
215 	    (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) {
216 		panic("ip6_output: IPv4 checksum offload flags: %d",
217 		    m->m_pkthdr.csum_flags);
218 	}
219 
220 	if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) ==
221 	    (M_CSUM_TCPv6|M_CSUM_UDPv6)) {
222 		panic("ip6_output: conflicting checksum offload flags: %d",
223 		    m->m_pkthdr.csum_flags);
224 	}
225 #endif
226 
227 	M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr));
228 
229 #define MAKE_EXTHDR(hp, mp)						\
230     do {								\
231 	if (hp) {							\
232 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
233 		error = ip6_copyexthdr((mp), (void *)(hp), 		\
234 		    ((eh)->ip6e_len + 1) << 3);				\
235 		if (error)						\
236 			goto freehdrs;					\
237 	}								\
238     } while (/*CONSTCOND*/ 0)
239 
240 	memset(&exthdrs, 0, sizeof(exthdrs));
241 	if (opt) {
242 		/* Hop-by-Hop options header */
243 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
244 		/* Destination options header(1st part) */
245 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
246 		/* Routing header */
247 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
248 		/* Destination options header(2nd part) */
249 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
250 	}
251 
252 #ifdef IPSEC
253 	if ((flags & IPV6_FORWARDING) != 0) {
254 		needipsec = 0;
255 		goto skippolicycheck;
256 	}
257 
258 	/* get a security policy for this packet */
259 	if (so == NULL)
260 		sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
261 	else {
262 		if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp,
263 					 IPSEC_DIR_OUTBOUND)) {
264 			needipsec = 0;
265 			goto skippolicycheck;
266 		}
267 		sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
268 	}
269 
270 	if (sp == NULL) {
271 		IPSEC6_STATINC(IPSEC_STAT_OUT_INVAL);
272 		goto freehdrs;
273 	}
274 
275 	error = 0;
276 
277 	/* check policy */
278 	switch (sp->policy) {
279 	case IPSEC_POLICY_DISCARD:
280 		/*
281 		 * This packet is just discarded.
282 		 */
283 		IPSEC6_STATINC(IPSEC_STAT_OUT_POLVIO);
284 		goto freehdrs;
285 
286 	case IPSEC_POLICY_BYPASS:
287 	case IPSEC_POLICY_NONE:
288 		/* no need to do IPsec. */
289 		needipsec = 0;
290 		break;
291 
292 	case IPSEC_POLICY_IPSEC:
293 		if (sp->req == NULL) {
294 			/* XXX should be panic ? */
295 			printf("ip6_output: No IPsec request specified.\n");
296 			error = EINVAL;
297 			goto freehdrs;
298 		}
299 		needipsec = 1;
300 		break;
301 
302 	case IPSEC_POLICY_ENTRUST:
303 	default:
304 		printf("ip6_output: Invalid policy found. %d\n", sp->policy);
305 	}
306 
307   skippolicycheck:;
308 #endif /* IPSEC */
309 
310 	/*
311 	 * Calculate the total length of the extension header chain.
312 	 * Keep the length of the unfragmentable part for fragmentation.
313 	 */
314 	optlen = 0;
315 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
316 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
317 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
318 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
319 	/* NOTE: we don't add AH/ESP length here. do that later. */
320 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
321 
322 #ifdef FAST_IPSEC
323 	/* Check the security policy (SP) for the packet */
324 
325 	/* XXX For moment, we doesn't support packet with extented action */
326 	if (optlen !=0)
327 		goto freehdrs;
328 
329 	sp = ipsec6_check_policy(m,so,flags,&needipsec,&error);
330 	if (error != 0) {
331 		/*
332 		 * Hack: -EINVAL is used to signal that a packet
333 		 * should be silently discarded.  This is typically
334 		 * because we asked key management for an SA and
335 		 * it was delayed (e.g. kicked up to IKE).
336 		 */
337 	if (error == -EINVAL)
338 		error = 0;
339 	goto freehdrs;
340     }
341 #endif /* FAST_IPSEC */
342 
343 
344 	if (needipsec &&
345 	    (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
346 		in6_delayed_cksum(m);
347 		m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
348 	}
349 
350 
351 	/*
352 	 * If we need IPsec, or there is at least one extension header,
353 	 * separate IP6 header from the payload.
354 	 */
355 	if ((needipsec || optlen) && !hdrsplit) {
356 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
357 			m = NULL;
358 			goto freehdrs;
359 		}
360 		m = exthdrs.ip6e_ip6;
361 		hdrsplit++;
362 	}
363 
364 	/* adjust pointer */
365 	ip6 = mtod(m, struct ip6_hdr *);
366 
367 	/* adjust mbuf packet header length */
368 	m->m_pkthdr.len += optlen;
369 	plen = m->m_pkthdr.len - sizeof(*ip6);
370 
371 	/* If this is a jumbo payload, insert a jumbo payload option. */
372 	if (plen > IPV6_MAXPACKET) {
373 		if (!hdrsplit) {
374 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
375 				m = NULL;
376 				goto freehdrs;
377 			}
378 			m = exthdrs.ip6e_ip6;
379 			hdrsplit++;
380 		}
381 		/* adjust pointer */
382 		ip6 = mtod(m, struct ip6_hdr *);
383 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
384 			goto freehdrs;
385 		optlen += 8; /* XXX JUMBOOPTLEN */
386 		ip6->ip6_plen = 0;
387 	} else
388 		ip6->ip6_plen = htons(plen);
389 
390 	/*
391 	 * Concatenate headers and fill in next header fields.
392 	 * Here we have, on "m"
393 	 *	IPv6 payload
394 	 * and we insert headers accordingly.  Finally, we should be getting:
395 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
396 	 *
397 	 * during the header composing process, "m" points to IPv6 header.
398 	 * "mprev" points to an extension header prior to esp.
399 	 */
400 	{
401 		u_char *nexthdrp = &ip6->ip6_nxt;
402 		struct mbuf *mprev = m;
403 
404 		/*
405 		 * we treat dest2 specially.  this makes IPsec processing
406 		 * much easier.  the goal here is to make mprev point the
407 		 * mbuf prior to dest2.
408 		 *
409 		 * result: IPv6 dest2 payload
410 		 * m and mprev will point to IPv6 header.
411 		 */
412 		if (exthdrs.ip6e_dest2) {
413 			if (!hdrsplit)
414 				panic("assumption failed: hdr not split");
415 			exthdrs.ip6e_dest2->m_next = m->m_next;
416 			m->m_next = exthdrs.ip6e_dest2;
417 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
418 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
419 		}
420 
421 #define MAKE_CHAIN(m, mp, p, i)\
422     do {\
423 	if (m) {\
424 		if (!hdrsplit) \
425 			panic("assumption failed: hdr not split"); \
426 		*mtod((m), u_char *) = *(p);\
427 		*(p) = (i);\
428 		p = mtod((m), u_char *);\
429 		(m)->m_next = (mp)->m_next;\
430 		(mp)->m_next = (m);\
431 		(mp) = (m);\
432 	}\
433     } while (/*CONSTCOND*/ 0)
434 		/*
435 		 * result: IPv6 hbh dest1 rthdr dest2 payload
436 		 * m will point to IPv6 header.  mprev will point to the
437 		 * extension header prior to dest2 (rthdr in the above case).
438 		 */
439 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
440 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
441 		    IPPROTO_DSTOPTS);
442 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
443 		    IPPROTO_ROUTING);
444 
445 		M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data,
446 		    sizeof(struct ip6_hdr) + optlen);
447 
448 #ifdef IPSEC
449 		if (!needipsec)
450 			goto skip_ipsec2;
451 
452 		/*
453 		 * pointers after IPsec headers are not valid any more.
454 		 * other pointers need a great care too.
455 		 * (IPsec routines should not mangle mbufs prior to AH/ESP)
456 		 */
457 		exthdrs.ip6e_dest2 = NULL;
458 
459 	    {
460 		struct ip6_rthdr *rh = NULL;
461 		int segleft_org = 0;
462 		struct ipsec_output_state state;
463 
464 		if (exthdrs.ip6e_rthdr) {
465 			rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
466 			segleft_org = rh->ip6r_segleft;
467 			rh->ip6r_segleft = 0;
468 		}
469 
470 		memset(&state, 0, sizeof(state));
471 		state.m = m;
472 		error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
473 		    &needipsectun);
474 		m = state.m;
475 		if (error) {
476 			rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
477 			/* mbuf is already reclaimed in ipsec6_output_trans. */
478 			m = NULL;
479 			switch (error) {
480 			case EHOSTUNREACH:
481 			case ENETUNREACH:
482 			case EMSGSIZE:
483 			case ENOBUFS:
484 			case ENOMEM:
485 				break;
486 			default:
487 				printf("ip6_output (ipsec): error code %d\n", error);
488 				/* FALLTHROUGH */
489 			case ENOENT:
490 				/* don't show these error codes to the user */
491 				error = 0;
492 				break;
493 			}
494 			goto bad;
495 		}
496 		if (exthdrs.ip6e_rthdr) {
497 			/* ah6_output doesn't modify mbuf chain */
498 			rh->ip6r_segleft = segleft_org;
499 		}
500 	    }
501 skip_ipsec2:;
502 #endif
503 	}
504 
505 	/*
506 	 * If there is a routing header, replace destination address field
507 	 * with the first hop of the routing header.
508 	 */
509 	if (exthdrs.ip6e_rthdr) {
510 		struct ip6_rthdr *rh;
511 		struct ip6_rthdr0 *rh0;
512 		struct in6_addr *addr;
513 		struct sockaddr_in6 sa;
514 
515 		rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
516 		    struct ip6_rthdr *));
517 		finaldst = ip6->ip6_dst;
518 		switch (rh->ip6r_type) {
519 		case IPV6_RTHDR_TYPE_0:
520 			 rh0 = (struct ip6_rthdr0 *)rh;
521 			 addr = (struct in6_addr *)(rh0 + 1);
522 
523 			 /*
524 			  * construct a sockaddr_in6 form of
525 			  * the first hop.
526 			  *
527 			  * XXX: we may not have enough
528 			  * information about its scope zone;
529 			  * there is no standard API to pass
530 			  * the information from the
531 			  * application.
532 			  */
533 			 sockaddr_in6_init(&sa, addr, 0, 0, 0);
534 			 if ((error = sa6_embedscope(&sa,
535 			     ip6_use_defzone)) != 0) {
536 				 goto bad;
537 			 }
538 			 ip6->ip6_dst = sa.sin6_addr;
539 			 (void)memmove(&addr[0], &addr[1],
540 			     sizeof(struct in6_addr) *
541 			     (rh0->ip6r0_segleft - 1));
542 			 addr[rh0->ip6r0_segleft - 1] = finaldst;
543 			 /* XXX */
544 			 in6_clearscope(addr + rh0->ip6r0_segleft - 1);
545 			 break;
546 		default:	/* is it possible? */
547 			 error = EINVAL;
548 			 goto bad;
549 		}
550 	}
551 
552 	/* Source address validation */
553 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
554 	    (flags & IPV6_UNSPECSRC) == 0) {
555 		error = EOPNOTSUPP;
556 		IP6_STATINC(IP6_STAT_BADSCOPE);
557 		goto bad;
558 	}
559 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
560 		error = EOPNOTSUPP;
561 		IP6_STATINC(IP6_STAT_BADSCOPE);
562 		goto bad;
563 	}
564 
565 	IP6_STATINC(IP6_STAT_LOCALOUT);
566 
567 	/*
568 	 * Route packet.
569 	 */
570 	/* initialize cached route */
571 	if (ro == NULL) {
572 		ro = &ip6route;
573 	}
574 	ro_pmtu = ro;
575 	if (opt && opt->ip6po_rthdr)
576 		ro = &opt->ip6po_route;
577 
578  	/*
579 	 * if specified, try to fill in the traffic class field.
580 	 * do not override if a non-zero value is already set.
581 	 * we check the diffserv field and the ecn field separately.
582 	 */
583 	if (opt && opt->ip6po_tclass >= 0) {
584 		int mask = 0;
585 
586 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
587 			mask |= 0xfc;
588 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
589 			mask |= 0x03;
590 		if (mask != 0)
591 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
592 	}
593 
594 	/* fill in or override the hop limit field, if necessary. */
595 	if (opt && opt->ip6po_hlim != -1)
596 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
597 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
598 		if (im6o != NULL)
599 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
600 		else
601 			ip6->ip6_hlim = ip6_defmcasthlim;
602 	}
603 
604 #ifdef IPSEC
605 	if (needipsec && needipsectun) {
606 		struct ipsec_output_state state;
607 
608 		/*
609 		 * All the extension headers will become inaccessible
610 		 * (since they can be encrypted).
611 		 * Don't panic, we need no more updates to extension headers
612 		 * on inner IPv6 packet (since they are now encapsulated).
613 		 *
614 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
615 		 */
616 		memset(&exthdrs, 0, sizeof(exthdrs));
617 		exthdrs.ip6e_ip6 = m;
618 
619 		memset(&state, 0, sizeof(state));
620 		state.m = m;
621 		state.ro = ro;
622 		state.dst = rtcache_getdst(ro);
623 
624 		error = ipsec6_output_tunnel(&state, sp, flags);
625 
626 		m = state.m;
627 		ro_pmtu = ro = state.ro;
628 		dst = satocsin6(state.dst);
629 		if (error) {
630 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
631 			m0 = m = NULL;
632 			m = NULL;
633 			switch (error) {
634 			case EHOSTUNREACH:
635 			case ENETUNREACH:
636 			case EMSGSIZE:
637 			case ENOBUFS:
638 			case ENOMEM:
639 				break;
640 			default:
641 				printf("ip6_output (ipsec): error code %d\n", error);
642 				/* FALLTHROUGH */
643 			case ENOENT:
644 				/* don't show these error codes to the user */
645 				error = 0;
646 				break;
647 			}
648 			goto bad;
649 		}
650 
651 		exthdrs.ip6e_ip6 = m;
652 	}
653 #endif /* IPSEC */
654 #ifdef FAST_IPSEC
655 	if (needipsec) {
656 		s = splsoftnet();
657 		error = ipsec6_process_packet(m,sp->req);
658 
659 		/*
660 		 * Preserve KAME behaviour: ENOENT can be returned
661 		 * when an SA acquire is in progress.  Don't propagate
662 		 * this to user-level; it confuses applications.
663 		 * XXX this will go away when the SADB is redone.
664 		 */
665 		if (error == ENOENT)
666 			error = 0;
667 		splx(s);
668 		goto done;
669 	}
670 #endif /* FAST_IPSEC */
671 
672 
673 
674 	/* adjust pointer */
675 	ip6 = mtod(m, struct ip6_hdr *);
676 
677 	sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
678 	if ((error = in6_selectroute(&dst_sa, opt, im6o, ro,
679 	    &ifp, &rt, 0)) != 0) {
680 		if (ifp != NULL)
681 			in6_ifstat_inc(ifp, ifs6_out_discard);
682 		goto bad;
683 	}
684 	if (rt == NULL) {
685 		/*
686 		 * If in6_selectroute() does not return a route entry,
687 		 * dst may not have been updated.
688 		 */
689 		rtcache_setdst(ro, sin6tosa(&dst_sa));
690 	}
691 
692 	/*
693 	 * then rt (for unicast) and ifp must be non-NULL valid values.
694 	 */
695 	if ((flags & IPV6_FORWARDING) == 0) {
696 		/* XXX: the FORWARDING flag can be set for mrouting. */
697 		in6_ifstat_inc(ifp, ifs6_out_request);
698 	}
699 	if (rt != NULL) {
700 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
701 		rt->rt_use++;
702 	}
703 
704 	/*
705 	 * The outgoing interface must be in the zone of source and
706 	 * destination addresses.  We should use ia_ifp to support the
707 	 * case of sending packets to an address of our own.
708 	 */
709 	if (ia != NULL && ia->ia_ifp)
710 		origifp = ia->ia_ifp;
711 	else
712 		origifp = ifp;
713 
714 	src0 = ip6->ip6_src;
715 	if (in6_setscope(&src0, origifp, &zone))
716 		goto badscope;
717 	sockaddr_in6_init(&src_sa, &ip6->ip6_src, 0, 0, 0);
718 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
719 		goto badscope;
720 
721 	dst0 = ip6->ip6_dst;
722 	if (in6_setscope(&dst0, origifp, &zone))
723 		goto badscope;
724 	/* re-initialize to be sure */
725 	sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
726 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id)
727 		goto badscope;
728 
729 	/* scope check is done. */
730 
731 	if (rt == NULL || IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
732 		if (dst == NULL)
733 			dst = satocsin6(rtcache_getdst(ro));
734 		KASSERT(dst != NULL);
735 	} else if (opt && rtcache_validate(&opt->ip6po_nextroute) != NULL) {
736 		/*
737 		 * The nexthop is explicitly specified by the
738 		 * application.  We assume the next hop is an IPv6
739 		 * address.
740 		 */
741 		dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
742 	} else if ((rt->rt_flags & RTF_GATEWAY))
743 		dst = (struct sockaddr_in6 *)rt->rt_gateway;
744 	else if (dst == NULL)
745 		dst = satocsin6(rtcache_getdst(ro));
746 
747 	/*
748 	 * XXXXXX: original code follows:
749 	 */
750 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
751 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
752 	else {
753 		struct	in6_multi *in6m;
754 
755 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
756 
757 		in6_ifstat_inc(ifp, ifs6_out_mcast);
758 
759 		/*
760 		 * Confirm that the outgoing interface supports multicast.
761 		 */
762 		if (!(ifp->if_flags & IFF_MULTICAST)) {
763 			IP6_STATINC(IP6_STAT_NOROUTE);
764 			in6_ifstat_inc(ifp, ifs6_out_discard);
765 			error = ENETUNREACH;
766 			goto bad;
767 		}
768 
769 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
770 		if (in6m != NULL &&
771 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
772 			/*
773 			 * If we belong to the destination multicast group
774 			 * on the outgoing interface, and the caller did not
775 			 * forbid loopback, loop back a copy.
776 			 */
777 			KASSERT(dst != NULL);
778 			ip6_mloopback(ifp, m, dst);
779 		} else {
780 			/*
781 			 * If we are acting as a multicast router, perform
782 			 * multicast forwarding as if the packet had just
783 			 * arrived on the interface to which we are about
784 			 * to send.  The multicast forwarding function
785 			 * recursively calls this function, using the
786 			 * IPV6_FORWARDING flag to prevent infinite recursion.
787 			 *
788 			 * Multicasts that are looped back by ip6_mloopback(),
789 			 * above, will be forwarded by the ip6_input() routine,
790 			 * if necessary.
791 			 */
792 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
793 				if (ip6_mforward(ip6, ifp, m) != 0) {
794 					m_freem(m);
795 					goto done;
796 				}
797 			}
798 		}
799 		/*
800 		 * Multicasts with a hoplimit of zero may be looped back,
801 		 * above, but must not be transmitted on a network.
802 		 * Also, multicasts addressed to the loopback interface
803 		 * are not sent -- the above call to ip6_mloopback() will
804 		 * loop back a copy if this host actually belongs to the
805 		 * destination group on the loopback interface.
806 		 */
807 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
808 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
809 			m_freem(m);
810 			goto done;
811 		}
812 	}
813 
814 	/*
815 	 * Fill the outgoing inteface to tell the upper layer
816 	 * to increment per-interface statistics.
817 	 */
818 	if (ifpp)
819 		*ifpp = ifp;
820 
821 	/* Determine path MTU. */
822 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
823 	    &alwaysfrag)) != 0)
824 		goto bad;
825 #ifdef IPSEC
826 	if (needipsectun)
827 		mtu = IPV6_MMTU;
828 #endif
829 
830 	/*
831 	 * The caller of this function may specify to use the minimum MTU
832 	 * in some cases.
833 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
834 	 * setting.  The logic is a bit complicated; by default, unicast
835 	 * packets will follow path MTU while multicast packets will be sent at
836 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
837 	 * including unicast ones will be sent at the minimum MTU.  Multicast
838 	 * packets will always be sent at the minimum MTU unless
839 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
840 	 * See RFC 3542 for more details.
841 	 */
842 	if (mtu > IPV6_MMTU) {
843 		if ((flags & IPV6_MINMTU))
844 			mtu = IPV6_MMTU;
845 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
846 			mtu = IPV6_MMTU;
847 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
848 			 (opt == NULL ||
849 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
850 			mtu = IPV6_MMTU;
851 		}
852 	}
853 
854 	/*
855 	 * clear embedded scope identifiers if necessary.
856 	 * in6_clearscope will touch the addresses only when necessary.
857 	 */
858 	in6_clearscope(&ip6->ip6_src);
859 	in6_clearscope(&ip6->ip6_dst);
860 
861 	/*
862 	 * If the outgoing packet contains a hop-by-hop options header,
863 	 * it must be examined and processed even by the source node.
864 	 * (RFC 2460, section 4.)
865 	 */
866 	if (exthdrs.ip6e_hbh) {
867 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
868 		u_int32_t dummy1; /* XXX unused */
869 		u_int32_t dummy2; /* XXX unused */
870 
871 		/*
872 		 *  XXX: if we have to send an ICMPv6 error to the sender,
873 		 *       we need the M_LOOP flag since icmp6_error() expects
874 		 *       the IPv6 and the hop-by-hop options header are
875 		 *       continuous unless the flag is set.
876 		 */
877 		m->m_flags |= M_LOOP;
878 		m->m_pkthdr.rcvif = ifp;
879 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
880 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
881 		    &dummy1, &dummy2) < 0) {
882 			/* m was already freed at this point */
883 			error = EINVAL;/* better error? */
884 			goto done;
885 		}
886 		m->m_flags &= ~M_LOOP; /* XXX */
887 		m->m_pkthdr.rcvif = NULL;
888 	}
889 
890 #ifdef PFIL_HOOKS
891 	/*
892 	 * Run through list of hooks for output packets.
893 	 */
894 	if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0)
895 		goto done;
896 	if (m == NULL)
897 		goto done;
898 	ip6 = mtod(m, struct ip6_hdr *);
899 #endif /* PFIL_HOOKS */
900 	/*
901 	 * Send the packet to the outgoing interface.
902 	 * If necessary, do IPv6 fragmentation before sending.
903 	 *
904 	 * the logic here is rather complex:
905 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
906 	 * 1-a:	send as is if tlen <= path mtu
907 	 * 1-b:	fragment if tlen > path mtu
908 	 *
909 	 * 2: if user asks us not to fragment (dontfrag == 1)
910 	 * 2-a:	send as is if tlen <= interface mtu
911 	 * 2-b:	error if tlen > interface mtu
912 	 *
913 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
914 	 *	always fragment
915 	 *
916 	 * 4: if dontfrag == 1 && alwaysfrag == 1
917 	 *	error, as we cannot handle this conflicting request
918 	 */
919 	tlen = m->m_pkthdr.len;
920 	tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0;
921 	if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
922 		dontfrag = 1;
923 	else
924 		dontfrag = 0;
925 
926 	if (dontfrag && alwaysfrag) {	/* case 4 */
927 		/* conflicting request - can't transmit */
928 		error = EMSGSIZE;
929 		goto bad;
930 	}
931 	if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) {	/* case 2-b */
932 		/*
933 		 * Even if the DONTFRAG option is specified, we cannot send the
934 		 * packet when the data length is larger than the MTU of the
935 		 * outgoing interface.
936 		 * Notify the error by sending IPV6_PATHMTU ancillary data as
937 		 * well as returning an error code (the latter is not described
938 		 * in the API spec.)
939 		 */
940 		u_int32_t mtu32;
941 		struct ip6ctlparam ip6cp;
942 
943 		mtu32 = (u_int32_t)mtu;
944 		memset(&ip6cp, 0, sizeof(ip6cp));
945 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
946 		pfctlinput2(PRC_MSGSIZE,
947 		    rtcache_getdst(ro_pmtu), &ip6cp);
948 
949 		error = EMSGSIZE;
950 		goto bad;
951 	}
952 
953 	/*
954 	 * transmit packet without fragmentation
955 	 */
956 	if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) {
957 		/* case 1-a and 2-a */
958 		struct in6_ifaddr *ia6;
959 		int sw_csum;
960 
961 		ip6 = mtod(m, struct ip6_hdr *);
962 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
963 		if (ia6) {
964 			/* Record statistics for this interface address. */
965 			ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len;
966 		}
967 #ifdef IPSEC
968 		/* clean ipsec history once it goes out of the node */
969 		ipsec_delaux(m);
970 #endif
971 
972 		sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
973 		if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
974 			if (IN6_NEED_CHECKSUM(ifp,
975 			    sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) {
976 				in6_delayed_cksum(m);
977 			}
978 			m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
979 		}
980 
981 		KASSERT(dst != NULL);
982 		if (__predict_true(!tso ||
983 		    (ifp->if_capenable & IFCAP_TSOv6) != 0)) {
984 			error = nd6_output(ifp, origifp, m, dst, rt);
985 		} else {
986 			error = ip6_tso_output(ifp, origifp, m, dst, rt);
987 		}
988 		goto done;
989 	}
990 
991 	if (tso) {
992 		error = EINVAL; /* XXX */
993 		goto bad;
994 	}
995 
996 	/*
997 	 * try to fragment the packet.  case 1-b and 3
998 	 */
999 	if (mtu < IPV6_MMTU) {
1000 		/* path MTU cannot be less than IPV6_MMTU */
1001 		error = EMSGSIZE;
1002 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1003 		goto bad;
1004 	} else if (ip6->ip6_plen == 0) {
1005 		/* jumbo payload cannot be fragmented */
1006 		error = EMSGSIZE;
1007 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1008 		goto bad;
1009 	} else {
1010 		struct mbuf **mnext, *m_frgpart;
1011 		struct ip6_frag *ip6f;
1012 		u_int32_t id = htonl(ip6_randomid());
1013 		u_char nextproto;
1014 #if 0				/* see below */
1015 		struct ip6ctlparam ip6cp;
1016 		u_int32_t mtu32;
1017 #endif
1018 
1019 		/*
1020 		 * Too large for the destination or interface;
1021 		 * fragment if possible.
1022 		 * Must be able to put at least 8 bytes per fragment.
1023 		 */
1024 		hlen = unfragpartlen;
1025 		if (mtu > IPV6_MAXPACKET)
1026 			mtu = IPV6_MAXPACKET;
1027 
1028 #if 0
1029 		/*
1030 		 * It is believed this code is a leftover from the
1031 		 * development of the IPV6_RECVPATHMTU sockopt and
1032 		 * associated work to implement RFC3542.
1033 		 * It's not entirely clear what the intent of the API
1034 		 * is at this point, so disable this code for now.
1035 		 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG
1036 		 * will send notifications if the application requests.
1037 		 */
1038 
1039 		/* Notify a proper path MTU to applications. */
1040 		mtu32 = (u_int32_t)mtu;
1041 		memset(&ip6cp, 0, sizeof(ip6cp));
1042 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
1043 		pfctlinput2(PRC_MSGSIZE,
1044 		    rtcache_getdst(ro_pmtu), &ip6cp);
1045 #endif
1046 
1047 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1048 		if (len < 8) {
1049 			error = EMSGSIZE;
1050 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
1051 			goto bad;
1052 		}
1053 
1054 		mnext = &m->m_nextpkt;
1055 
1056 		/*
1057 		 * Change the next header field of the last header in the
1058 		 * unfragmentable part.
1059 		 */
1060 		if (exthdrs.ip6e_rthdr) {
1061 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1062 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1063 		} else if (exthdrs.ip6e_dest1) {
1064 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1065 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1066 		} else if (exthdrs.ip6e_hbh) {
1067 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1068 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1069 		} else {
1070 			nextproto = ip6->ip6_nxt;
1071 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1072 		}
1073 
1074 		if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6))
1075 		    != 0) {
1076 			if (IN6_NEED_CHECKSUM(ifp,
1077 			    m->m_pkthdr.csum_flags &
1078 			    (M_CSUM_UDPv6|M_CSUM_TCPv6))) {
1079 				in6_delayed_cksum(m);
1080 			}
1081 			m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
1082 		}
1083 
1084 		/*
1085 		 * Loop through length of segment after first fragment,
1086 		 * make new header and copy data of each part and link onto
1087 		 * chain.
1088 		 */
1089 		m0 = m;
1090 		for (off = hlen; off < tlen; off += len) {
1091 			struct mbuf *mlast;
1092 
1093 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
1094 			if (!m) {
1095 				error = ENOBUFS;
1096 				IP6_STATINC(IP6_STAT_ODROPPED);
1097 				goto sendorfree;
1098 			}
1099 			m->m_pkthdr.rcvif = NULL;
1100 			m->m_flags = m0->m_flags & M_COPYFLAGS;
1101 			*mnext = m;
1102 			mnext = &m->m_nextpkt;
1103 			m->m_data += max_linkhdr;
1104 			mhip6 = mtod(m, struct ip6_hdr *);
1105 			*mhip6 = *ip6;
1106 			m->m_len = sizeof(*mhip6);
1107 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1108 			if (error) {
1109 				IP6_STATINC(IP6_STAT_ODROPPED);
1110 				goto sendorfree;
1111 			}
1112 			ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7));
1113 			if (off + len >= tlen)
1114 				len = tlen - off;
1115 			else
1116 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1117 			mhip6->ip6_plen = htons((u_int16_t)(len + hlen +
1118 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1119 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1120 				error = ENOBUFS;
1121 				IP6_STATINC(IP6_STAT_ODROPPED);
1122 				goto sendorfree;
1123 			}
1124 			for (mlast = m; mlast->m_next; mlast = mlast->m_next)
1125 				;
1126 			mlast->m_next = m_frgpart;
1127 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1128 			m->m_pkthdr.rcvif = (struct ifnet *)0;
1129 			ip6f->ip6f_reserved = 0;
1130 			ip6f->ip6f_ident = id;
1131 			ip6f->ip6f_nxt = nextproto;
1132 			IP6_STATINC(IP6_STAT_OFRAGMENTS);
1133 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1134 		}
1135 
1136 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1137 	}
1138 
1139 	/*
1140 	 * Remove leading garbages.
1141 	 */
1142 sendorfree:
1143 	m = m0->m_nextpkt;
1144 	m0->m_nextpkt = 0;
1145 	m_freem(m0);
1146 	for (m0 = m; m; m = m0) {
1147 		m0 = m->m_nextpkt;
1148 		m->m_nextpkt = 0;
1149 		if (error == 0) {
1150 			struct in6_ifaddr *ia6;
1151 			ip6 = mtod(m, struct ip6_hdr *);
1152 			ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1153 			if (ia6) {
1154 				/*
1155 				 * Record statistics for this interface
1156 				 * address.
1157 				 */
1158 				ia6->ia_ifa.ifa_data.ifad_outbytes +=
1159 				    m->m_pkthdr.len;
1160 			}
1161 #ifdef IPSEC
1162 			/* clean ipsec history once it goes out of the node */
1163 			ipsec_delaux(m);
1164 #endif
1165 			KASSERT(dst != NULL);
1166 			error = nd6_output(ifp, origifp, m, dst, rt);
1167 		} else
1168 			m_freem(m);
1169 	}
1170 
1171 	if (error == 0)
1172 		IP6_STATINC(IP6_STAT_FRAGMENTED);
1173 
1174 done:
1175 	rtcache_free(&ip6route);
1176 
1177 #ifdef IPSEC
1178 	if (sp != NULL)
1179 		key_freesp(sp);
1180 #endif /* IPSEC */
1181 #ifdef FAST_IPSEC
1182 	if (sp != NULL)
1183 		KEY_FREESP(&sp);
1184 #endif /* FAST_IPSEC */
1185 
1186 
1187 	return (error);
1188 
1189 freehdrs:
1190 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1191 	m_freem(exthdrs.ip6e_dest1);
1192 	m_freem(exthdrs.ip6e_rthdr);
1193 	m_freem(exthdrs.ip6e_dest2);
1194 	/* FALLTHROUGH */
1195 bad:
1196 	m_freem(m);
1197 	goto done;
1198 badscope:
1199 	IP6_STATINC(IP6_STAT_BADSCOPE);
1200 	in6_ifstat_inc(origifp, ifs6_out_discard);
1201 	if (error == 0)
1202 		error = EHOSTUNREACH; /* XXX */
1203 	goto bad;
1204 }
1205 
1206 static int
1207 ip6_copyexthdr(struct mbuf **mp, void *hdr, int hlen)
1208 {
1209 	struct mbuf *m;
1210 
1211 	if (hlen > MCLBYTES)
1212 		return (ENOBUFS); /* XXX */
1213 
1214 	MGET(m, M_DONTWAIT, MT_DATA);
1215 	if (!m)
1216 		return (ENOBUFS);
1217 
1218 	if (hlen > MLEN) {
1219 		MCLGET(m, M_DONTWAIT);
1220 		if ((m->m_flags & M_EXT) == 0) {
1221 			m_free(m);
1222 			return (ENOBUFS);
1223 		}
1224 	}
1225 	m->m_len = hlen;
1226 	if (hdr)
1227 		bcopy(hdr, mtod(m, void *), hlen);
1228 
1229 	*mp = m;
1230 	return (0);
1231 }
1232 
1233 /*
1234  * Process a delayed payload checksum calculation.
1235  */
1236 void
1237 in6_delayed_cksum(struct mbuf *m)
1238 {
1239 	uint16_t csum, offset;
1240 
1241 	KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0);
1242 	KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0);
1243 	KASSERT((m->m_pkthdr.csum_flags
1244 	    & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0);
1245 
1246 	offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data);
1247 	csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset);
1248 	if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) {
1249 		csum = 0xffff;
1250 	}
1251 
1252 	offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data);
1253 	if ((offset + sizeof(csum)) > m->m_len) {
1254 		m_copyback(m, offset, sizeof(csum), &csum);
1255 	} else {
1256 		*(uint16_t *)(mtod(m, char *) + offset) = csum;
1257 	}
1258 }
1259 
1260 /*
1261  * Insert jumbo payload option.
1262  */
1263 static int
1264 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1265 {
1266 	struct mbuf *mopt;
1267 	u_int8_t *optbuf;
1268 	u_int32_t v;
1269 
1270 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1271 
1272 	/*
1273 	 * If there is no hop-by-hop options header, allocate new one.
1274 	 * If there is one but it doesn't have enough space to store the
1275 	 * jumbo payload option, allocate a cluster to store the whole options.
1276 	 * Otherwise, use it to store the options.
1277 	 */
1278 	if (exthdrs->ip6e_hbh == 0) {
1279 		MGET(mopt, M_DONTWAIT, MT_DATA);
1280 		if (mopt == 0)
1281 			return (ENOBUFS);
1282 		mopt->m_len = JUMBOOPTLEN;
1283 		optbuf = mtod(mopt, u_int8_t *);
1284 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1285 		exthdrs->ip6e_hbh = mopt;
1286 	} else {
1287 		struct ip6_hbh *hbh;
1288 
1289 		mopt = exthdrs->ip6e_hbh;
1290 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1291 			/*
1292 			 * XXX assumption:
1293 			 * - exthdrs->ip6e_hbh is not referenced from places
1294 			 *   other than exthdrs.
1295 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1296 			 */
1297 			int oldoptlen = mopt->m_len;
1298 			struct mbuf *n;
1299 
1300 			/*
1301 			 * XXX: give up if the whole (new) hbh header does
1302 			 * not fit even in an mbuf cluster.
1303 			 */
1304 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1305 				return (ENOBUFS);
1306 
1307 			/*
1308 			 * As a consequence, we must always prepare a cluster
1309 			 * at this point.
1310 			 */
1311 			MGET(n, M_DONTWAIT, MT_DATA);
1312 			if (n) {
1313 				MCLGET(n, M_DONTWAIT);
1314 				if ((n->m_flags & M_EXT) == 0) {
1315 					m_freem(n);
1316 					n = NULL;
1317 				}
1318 			}
1319 			if (!n)
1320 				return (ENOBUFS);
1321 			n->m_len = oldoptlen + JUMBOOPTLEN;
1322 			bcopy(mtod(mopt, void *), mtod(n, void *),
1323 			    oldoptlen);
1324 			optbuf = mtod(n, u_int8_t *) + oldoptlen;
1325 			m_freem(mopt);
1326 			mopt = exthdrs->ip6e_hbh = n;
1327 		} else {
1328 			optbuf = mtod(mopt, u_int8_t *) + mopt->m_len;
1329 			mopt->m_len += JUMBOOPTLEN;
1330 		}
1331 		optbuf[0] = IP6OPT_PADN;
1332 		optbuf[1] = 0;
1333 
1334 		/*
1335 		 * Adjust the header length according to the pad and
1336 		 * the jumbo payload option.
1337 		 */
1338 		hbh = mtod(mopt, struct ip6_hbh *);
1339 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1340 	}
1341 
1342 	/* fill in the option. */
1343 	optbuf[2] = IP6OPT_JUMBO;
1344 	optbuf[3] = 4;
1345 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1346 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1347 
1348 	/* finally, adjust the packet header length */
1349 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1350 
1351 	return (0);
1352 #undef JUMBOOPTLEN
1353 }
1354 
1355 /*
1356  * Insert fragment header and copy unfragmentable header portions.
1357  */
1358 static int
1359 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1360 	struct ip6_frag **frghdrp)
1361 {
1362 	struct mbuf *n, *mlast;
1363 
1364 	if (hlen > sizeof(struct ip6_hdr)) {
1365 		n = m_copym(m0, sizeof(struct ip6_hdr),
1366 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1367 		if (n == 0)
1368 			return (ENOBUFS);
1369 		m->m_next = n;
1370 	} else
1371 		n = m;
1372 
1373 	/* Search for the last mbuf of unfragmentable part. */
1374 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1375 		;
1376 
1377 	if ((mlast->m_flags & M_EXT) == 0 &&
1378 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1379 		/* use the trailing space of the last mbuf for the fragment hdr */
1380 		*frghdrp = (struct ip6_frag *)(mtod(mlast, char *) +
1381 		    mlast->m_len);
1382 		mlast->m_len += sizeof(struct ip6_frag);
1383 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1384 	} else {
1385 		/* allocate a new mbuf for the fragment header */
1386 		struct mbuf *mfrg;
1387 
1388 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1389 		if (mfrg == 0)
1390 			return (ENOBUFS);
1391 		mfrg->m_len = sizeof(struct ip6_frag);
1392 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1393 		mlast->m_next = mfrg;
1394 	}
1395 
1396 	return (0);
1397 }
1398 
1399 static int
1400 ip6_getpmtu(struct route *ro_pmtu, struct route *ro, struct ifnet *ifp,
1401     const struct in6_addr *dst, u_long *mtup, int *alwaysfragp)
1402 {
1403 	struct rtentry *rt;
1404 	u_int32_t mtu = 0;
1405 	int alwaysfrag = 0;
1406 	int error = 0;
1407 
1408 	if (ro_pmtu != ro) {
1409 		union {
1410 			struct sockaddr		dst;
1411 			struct sockaddr_in6	dst6;
1412 		} u;
1413 
1414 		/* The first hop and the final destination may differ. */
1415 		sockaddr_in6_init(&u.dst6, dst, 0, 0, 0);
1416 		rt = rtcache_lookup(ro_pmtu, &u.dst);
1417 	} else
1418 		rt = rtcache_validate(ro_pmtu);
1419 	if (rt != NULL) {
1420 		u_int32_t ifmtu;
1421 
1422 		if (ifp == NULL)
1423 			ifp = rt->rt_ifp;
1424 		ifmtu = IN6_LINKMTU(ifp);
1425 		mtu = rt->rt_rmx.rmx_mtu;
1426 		if (mtu == 0)
1427 			mtu = ifmtu;
1428 		else if (mtu < IPV6_MMTU) {
1429 			/*
1430 			 * RFC2460 section 5, last paragraph:
1431 			 * if we record ICMPv6 too big message with
1432 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1433 			 * or smaller, with fragment header attached.
1434 			 * (fragment header is needed regardless from the
1435 			 * packet size, for translators to identify packets)
1436 			 */
1437 			alwaysfrag = 1;
1438 			mtu = IPV6_MMTU;
1439 		} else if (mtu > ifmtu) {
1440 			/*
1441 			 * The MTU on the route is larger than the MTU on
1442 			 * the interface!  This shouldn't happen, unless the
1443 			 * MTU of the interface has been changed after the
1444 			 * interface was brought up.  Change the MTU in the
1445 			 * route to match the interface MTU (as long as the
1446 			 * field isn't locked).
1447 			 */
1448 			mtu = ifmtu;
1449 			if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
1450 				rt->rt_rmx.rmx_mtu = mtu;
1451 		}
1452 	} else if (ifp) {
1453 		mtu = IN6_LINKMTU(ifp);
1454 	} else
1455 		error = EHOSTUNREACH; /* XXX */
1456 
1457 	*mtup = mtu;
1458 	if (alwaysfragp)
1459 		*alwaysfragp = alwaysfrag;
1460 	return (error);
1461 }
1462 
1463 /*
1464  * IP6 socket option processing.
1465  */
1466 int
1467 ip6_ctloutput(int op, struct socket *so, struct sockopt *sopt)
1468 {
1469 	int optdatalen, uproto;
1470 	void *optdata;
1471 	struct in6pcb *in6p = sotoin6pcb(so);
1472 	int error, optval;
1473 	int level, optname;
1474 
1475 	KASSERT(sopt != NULL);
1476 
1477 	level = sopt->sopt_level;
1478 	optname = sopt->sopt_name;
1479 
1480 	error = optval = 0;
1481 	uproto = (int)so->so_proto->pr_protocol;
1482 
1483 	if (level != IPPROTO_IPV6) {
1484 		return ENOPROTOOPT;
1485 	}
1486 	switch (op) {
1487 	case PRCO_SETOPT:
1488 		switch (optname) {
1489 #ifdef RFC2292
1490 		case IPV6_2292PKTOPTIONS:
1491 			error = ip6_pcbopts(&in6p->in6p_outputopts, so, sopt);
1492 			break;
1493 #endif
1494 
1495 		/*
1496 		 * Use of some Hop-by-Hop options or some
1497 		 * Destination options, might require special
1498 		 * privilege.  That is, normal applications
1499 		 * (without special privilege) might be forbidden
1500 		 * from setting certain options in outgoing packets,
1501 		 * and might never see certain options in received
1502 		 * packets. [RFC 2292 Section 6]
1503 		 * KAME specific note:
1504 		 *  KAME prevents non-privileged users from sending or
1505 		 *  receiving ANY hbh/dst options in order to avoid
1506 		 *  overhead of parsing options in the kernel.
1507 		 */
1508 		case IPV6_RECVHOPOPTS:
1509 		case IPV6_RECVDSTOPTS:
1510 		case IPV6_RECVRTHDRDSTOPTS:
1511 			error = kauth_authorize_generic(kauth_cred_get(),
1512 			    KAUTH_GENERIC_ISSUSER, NULL);
1513 			if (error)
1514 				break;
1515 			/* FALLTHROUGH */
1516 		case IPV6_UNICAST_HOPS:
1517 		case IPV6_HOPLIMIT:
1518 		case IPV6_FAITH:
1519 
1520 		case IPV6_RECVPKTINFO:
1521 		case IPV6_RECVHOPLIMIT:
1522 		case IPV6_RECVRTHDR:
1523 		case IPV6_RECVPATHMTU:
1524 		case IPV6_RECVTCLASS:
1525 		case IPV6_V6ONLY:
1526 			error = sockopt_getint(sopt, &optval);
1527 			if (error)
1528 				break;
1529 			switch (optname) {
1530 			case IPV6_UNICAST_HOPS:
1531 				if (optval < -1 || optval >= 256)
1532 					error = EINVAL;
1533 				else {
1534 					/* -1 = kernel default */
1535 					in6p->in6p_hops = optval;
1536 				}
1537 				break;
1538 #define OPTSET(bit) \
1539 do { \
1540 if (optval) \
1541 	in6p->in6p_flags |= (bit); \
1542 else \
1543 	in6p->in6p_flags &= ~(bit); \
1544 } while (/*CONSTCOND*/ 0)
1545 
1546 #ifdef RFC2292
1547 #define OPTSET2292(bit) 			\
1548 do { 						\
1549 in6p->in6p_flags |= IN6P_RFC2292; 	\
1550 if (optval) 				\
1551 	in6p->in6p_flags |= (bit); 	\
1552 else 					\
1553 	in6p->in6p_flags &= ~(bit); 	\
1554 } while (/*CONSTCOND*/ 0)
1555 #endif
1556 
1557 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1558 
1559 			case IPV6_RECVPKTINFO:
1560 #ifdef RFC2292
1561 				/* cannot mix with RFC2292 */
1562 				if (OPTBIT(IN6P_RFC2292)) {
1563 					error = EINVAL;
1564 					break;
1565 				}
1566 #endif
1567 				OPTSET(IN6P_PKTINFO);
1568 				break;
1569 
1570 			case IPV6_HOPLIMIT:
1571 			{
1572 				struct ip6_pktopts **optp;
1573 
1574 #ifdef RFC2292
1575 				/* cannot mix with RFC2292 */
1576 				if (OPTBIT(IN6P_RFC2292)) {
1577 					error = EINVAL;
1578 					break;
1579 				}
1580 #endif
1581 				optp = &in6p->in6p_outputopts;
1582 				error = ip6_pcbopt(IPV6_HOPLIMIT,
1583 						   (u_char *)&optval,
1584 						   sizeof(optval),
1585 						   optp,
1586 						   kauth_cred_get(), uproto);
1587 				break;
1588 			}
1589 
1590 			case IPV6_RECVHOPLIMIT:
1591 #ifdef RFC2292
1592 				/* cannot mix with RFC2292 */
1593 				if (OPTBIT(IN6P_RFC2292)) {
1594 					error = EINVAL;
1595 					break;
1596 				}
1597 #endif
1598 				OPTSET(IN6P_HOPLIMIT);
1599 				break;
1600 
1601 			case IPV6_RECVHOPOPTS:
1602 #ifdef RFC2292
1603 				/* cannot mix with RFC2292 */
1604 				if (OPTBIT(IN6P_RFC2292)) {
1605 					error = EINVAL;
1606 					break;
1607 				}
1608 #endif
1609 				OPTSET(IN6P_HOPOPTS);
1610 				break;
1611 
1612 			case IPV6_RECVDSTOPTS:
1613 #ifdef RFC2292
1614 				/* cannot mix with RFC2292 */
1615 				if (OPTBIT(IN6P_RFC2292)) {
1616 					error = EINVAL;
1617 					break;
1618 				}
1619 #endif
1620 				OPTSET(IN6P_DSTOPTS);
1621 				break;
1622 
1623 			case IPV6_RECVRTHDRDSTOPTS:
1624 #ifdef RFC2292
1625 				/* cannot mix with RFC2292 */
1626 				if (OPTBIT(IN6P_RFC2292)) {
1627 					error = EINVAL;
1628 					break;
1629 				}
1630 #endif
1631 				OPTSET(IN6P_RTHDRDSTOPTS);
1632 				break;
1633 
1634 			case IPV6_RECVRTHDR:
1635 #ifdef RFC2292
1636 				/* cannot mix with RFC2292 */
1637 				if (OPTBIT(IN6P_RFC2292)) {
1638 					error = EINVAL;
1639 					break;
1640 				}
1641 #endif
1642 				OPTSET(IN6P_RTHDR);
1643 				break;
1644 
1645 			case IPV6_FAITH:
1646 				OPTSET(IN6P_FAITH);
1647 				break;
1648 
1649 			case IPV6_RECVPATHMTU:
1650 				/*
1651 				 * We ignore this option for TCP
1652 				 * sockets.
1653 				 * (RFC3542 leaves this case
1654 				 * unspecified.)
1655 				 */
1656 				if (uproto != IPPROTO_TCP)
1657 					OPTSET(IN6P_MTU);
1658 				break;
1659 
1660 			case IPV6_V6ONLY:
1661 				/*
1662 				 * make setsockopt(IPV6_V6ONLY)
1663 				 * available only prior to bind(2).
1664 				 * see ipng mailing list, Jun 22 2001.
1665 				 */
1666 				if (in6p->in6p_lport ||
1667 				    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1668 					error = EINVAL;
1669 					break;
1670 				}
1671 #ifdef INET6_BINDV6ONLY
1672 				if (!optval)
1673 					error = EINVAL;
1674 #else
1675 				OPTSET(IN6P_IPV6_V6ONLY);
1676 #endif
1677 				break;
1678 			case IPV6_RECVTCLASS:
1679 #ifdef RFC2292
1680 				/* cannot mix with RFC2292 XXX */
1681 				if (OPTBIT(IN6P_RFC2292)) {
1682 					error = EINVAL;
1683 					break;
1684 				}
1685 #endif
1686 				OPTSET(IN6P_TCLASS);
1687 				break;
1688 
1689 			}
1690 			break;
1691 
1692 		case IPV6_OTCLASS:
1693 		{
1694 			struct ip6_pktopts **optp;
1695 			u_int8_t tclass;
1696 
1697 			error = sockopt_get(sopt, &tclass, sizeof(tclass));
1698 			if (error)
1699 				break;
1700 			optp = &in6p->in6p_outputopts;
1701 			error = ip6_pcbopt(optname,
1702 					   (u_char *)&tclass,
1703 					   sizeof(tclass),
1704 					   optp,
1705 					   kauth_cred_get(), uproto);
1706 			break;
1707 		}
1708 
1709 		case IPV6_TCLASS:
1710 		case IPV6_DONTFRAG:
1711 		case IPV6_USE_MIN_MTU:
1712 			error = sockopt_getint(sopt, &optval);
1713 			if (error)
1714 				break;
1715 			{
1716 				struct ip6_pktopts **optp;
1717 				optp = &in6p->in6p_outputopts;
1718 				error = ip6_pcbopt(optname,
1719 						   (u_char *)&optval,
1720 						   sizeof(optval),
1721 						   optp,
1722 						   kauth_cred_get(), uproto);
1723 				break;
1724 			}
1725 
1726 #ifdef RFC2292
1727 		case IPV6_2292PKTINFO:
1728 		case IPV6_2292HOPLIMIT:
1729 		case IPV6_2292HOPOPTS:
1730 		case IPV6_2292DSTOPTS:
1731 		case IPV6_2292RTHDR:
1732 			/* RFC 2292 */
1733 			error = sockopt_getint(sopt, &optval);
1734 			if (error)
1735 				break;
1736 
1737 			switch (optname) {
1738 			case IPV6_2292PKTINFO:
1739 				OPTSET2292(IN6P_PKTINFO);
1740 				break;
1741 			case IPV6_2292HOPLIMIT:
1742 				OPTSET2292(IN6P_HOPLIMIT);
1743 				break;
1744 			case IPV6_2292HOPOPTS:
1745 				/*
1746 				 * Check super-user privilege.
1747 				 * See comments for IPV6_RECVHOPOPTS.
1748 				 */
1749 				error =
1750 				    kauth_authorize_generic(kauth_cred_get(),
1751 				    KAUTH_GENERIC_ISSUSER, NULL);
1752 				if (error)
1753 					return (error);
1754 				OPTSET2292(IN6P_HOPOPTS);
1755 				break;
1756 			case IPV6_2292DSTOPTS:
1757 				error =
1758 				    kauth_authorize_generic(kauth_cred_get(),
1759 				    KAUTH_GENERIC_ISSUSER, NULL);
1760 				if (error)
1761 					return (error);
1762 				OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1763 				break;
1764 			case IPV6_2292RTHDR:
1765 				OPTSET2292(IN6P_RTHDR);
1766 				break;
1767 			}
1768 			break;
1769 #endif
1770 		case IPV6_PKTINFO:
1771 		case IPV6_HOPOPTS:
1772 		case IPV6_RTHDR:
1773 		case IPV6_DSTOPTS:
1774 		case IPV6_RTHDRDSTOPTS:
1775 		case IPV6_NEXTHOP: {
1776 			/* new advanced API (RFC3542) */
1777 			void *optbuf;
1778 			int optbuflen;
1779 			struct ip6_pktopts **optp;
1780 
1781 #ifdef RFC2292
1782 			/* cannot mix with RFC2292 */
1783 			if (OPTBIT(IN6P_RFC2292)) {
1784 				error = EINVAL;
1785 				break;
1786 			}
1787 #endif
1788 
1789 			optbuflen = sopt->sopt_size;
1790 			optbuf = malloc(optbuflen, M_IP6OPT, M_NOWAIT);
1791 			if (optbuf == NULL) {
1792 				error = ENOBUFS;
1793 				break;
1794 			}
1795 
1796 			sockopt_get(sopt, optbuf, optbuflen);
1797 			optp = &in6p->in6p_outputopts;
1798 			error = ip6_pcbopt(optname, optbuf, optbuflen,
1799 			    optp, kauth_cred_get(), uproto);
1800 			break;
1801 			}
1802 #undef OPTSET
1803 
1804 		case IPV6_MULTICAST_IF:
1805 		case IPV6_MULTICAST_HOPS:
1806 		case IPV6_MULTICAST_LOOP:
1807 		case IPV6_JOIN_GROUP:
1808 		case IPV6_LEAVE_GROUP:
1809 			error = ip6_setmoptions(sopt, &in6p->in6p_moptions);
1810 			break;
1811 
1812 		case IPV6_PORTRANGE:
1813 			error = sockopt_getint(sopt, &optval);
1814 			if (error)
1815 				break;
1816 
1817 			switch (optval) {
1818 			case IPV6_PORTRANGE_DEFAULT:
1819 				in6p->in6p_flags &= ~(IN6P_LOWPORT);
1820 				in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1821 				break;
1822 
1823 			case IPV6_PORTRANGE_HIGH:
1824 				in6p->in6p_flags &= ~(IN6P_LOWPORT);
1825 				in6p->in6p_flags |= IN6P_HIGHPORT;
1826 				break;
1827 
1828 			case IPV6_PORTRANGE_LOW:
1829 				in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1830 				in6p->in6p_flags |= IN6P_LOWPORT;
1831 				break;
1832 
1833 			default:
1834 				error = EINVAL;
1835 				break;
1836 			}
1837 			break;
1838 
1839 
1840 #if defined(IPSEC) || defined(FAST_IPSEC)
1841 		case IPV6_IPSEC_POLICY:
1842 			error = ipsec6_set_policy(in6p, optname,
1843 			    sopt->sopt_data, sopt->sopt_size, kauth_cred_get());
1844 			break;
1845 #endif /* IPSEC */
1846 
1847 		default:
1848 			error = ENOPROTOOPT;
1849 			break;
1850 		}
1851 		break;
1852 
1853 	case PRCO_GETOPT:
1854 		switch (optname) {
1855 #ifdef RFC2292
1856 		case IPV6_2292PKTOPTIONS:
1857 			/*
1858 			 * RFC3542 (effectively) deprecated the
1859 			 * semantics of the 2292-style pktoptions.
1860 			 * Since it was not reliable in nature (i.e.,
1861 			 * applications had to expect the lack of some
1862 			 * information after all), it would make sense
1863 			 * to simplify this part by always returning
1864 			 * empty data.
1865 			 */
1866 			break;
1867 #endif
1868 
1869 		case IPV6_RECVHOPOPTS:
1870 		case IPV6_RECVDSTOPTS:
1871 		case IPV6_RECVRTHDRDSTOPTS:
1872 		case IPV6_UNICAST_HOPS:
1873 		case IPV6_RECVPKTINFO:
1874 		case IPV6_RECVHOPLIMIT:
1875 		case IPV6_RECVRTHDR:
1876 		case IPV6_RECVPATHMTU:
1877 
1878 		case IPV6_FAITH:
1879 		case IPV6_V6ONLY:
1880 		case IPV6_PORTRANGE:
1881 		case IPV6_RECVTCLASS:
1882 			switch (optname) {
1883 
1884 			case IPV6_RECVHOPOPTS:
1885 				optval = OPTBIT(IN6P_HOPOPTS);
1886 				break;
1887 
1888 			case IPV6_RECVDSTOPTS:
1889 				optval = OPTBIT(IN6P_DSTOPTS);
1890 				break;
1891 
1892 			case IPV6_RECVRTHDRDSTOPTS:
1893 				optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1894 				break;
1895 
1896 			case IPV6_UNICAST_HOPS:
1897 				optval = in6p->in6p_hops;
1898 				break;
1899 
1900 			case IPV6_RECVPKTINFO:
1901 				optval = OPTBIT(IN6P_PKTINFO);
1902 				break;
1903 
1904 			case IPV6_RECVHOPLIMIT:
1905 				optval = OPTBIT(IN6P_HOPLIMIT);
1906 				break;
1907 
1908 			case IPV6_RECVRTHDR:
1909 				optval = OPTBIT(IN6P_RTHDR);
1910 				break;
1911 
1912 			case IPV6_RECVPATHMTU:
1913 				optval = OPTBIT(IN6P_MTU);
1914 				break;
1915 
1916 			case IPV6_FAITH:
1917 				optval = OPTBIT(IN6P_FAITH);
1918 				break;
1919 
1920 			case IPV6_V6ONLY:
1921 				optval = OPTBIT(IN6P_IPV6_V6ONLY);
1922 				break;
1923 
1924 			case IPV6_PORTRANGE:
1925 			    {
1926 				int flags;
1927 				flags = in6p->in6p_flags;
1928 				if (flags & IN6P_HIGHPORT)
1929 					optval = IPV6_PORTRANGE_HIGH;
1930 				else if (flags & IN6P_LOWPORT)
1931 					optval = IPV6_PORTRANGE_LOW;
1932 				else
1933 					optval = 0;
1934 				break;
1935 			    }
1936 			case IPV6_RECVTCLASS:
1937 				optval = OPTBIT(IN6P_TCLASS);
1938 				break;
1939 
1940 			}
1941 			if (error)
1942 				break;
1943 			error = sockopt_setint(sopt, optval);
1944 			break;
1945 
1946 		case IPV6_PATHMTU:
1947 		    {
1948 			u_long pmtu = 0;
1949 			struct ip6_mtuinfo mtuinfo;
1950 			struct route *ro = &in6p->in6p_route;
1951 
1952 			if (!(so->so_state & SS_ISCONNECTED))
1953 				return (ENOTCONN);
1954 			/*
1955 			 * XXX: we dot not consider the case of source
1956 			 * routing, or optional information to specify
1957 			 * the outgoing interface.
1958 			 */
1959 			error = ip6_getpmtu(ro, NULL, NULL,
1960 			    &in6p->in6p_faddr, &pmtu, NULL);
1961 			if (error)
1962 				break;
1963 			if (pmtu > IPV6_MAXPACKET)
1964 				pmtu = IPV6_MAXPACKET;
1965 
1966 			memset(&mtuinfo, 0, sizeof(mtuinfo));
1967 			mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1968 			optdata = (void *)&mtuinfo;
1969 			optdatalen = sizeof(mtuinfo);
1970 			if (optdatalen > MCLBYTES)
1971 				return (EMSGSIZE); /* XXX */
1972 			error = sockopt_set(sopt, optdata, optdatalen);
1973 			break;
1974 		    }
1975 
1976 #ifdef RFC2292
1977 		case IPV6_2292PKTINFO:
1978 		case IPV6_2292HOPLIMIT:
1979 		case IPV6_2292HOPOPTS:
1980 		case IPV6_2292RTHDR:
1981 		case IPV6_2292DSTOPTS:
1982 			switch (optname) {
1983 			case IPV6_2292PKTINFO:
1984 				optval = OPTBIT(IN6P_PKTINFO);
1985 				break;
1986 			case IPV6_2292HOPLIMIT:
1987 				optval = OPTBIT(IN6P_HOPLIMIT);
1988 				break;
1989 			case IPV6_2292HOPOPTS:
1990 				optval = OPTBIT(IN6P_HOPOPTS);
1991 				break;
1992 			case IPV6_2292RTHDR:
1993 				optval = OPTBIT(IN6P_RTHDR);
1994 				break;
1995 			case IPV6_2292DSTOPTS:
1996 				optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1997 				break;
1998 			}
1999 			error = sockopt_setint(sopt, optval);
2000 			break;
2001 #endif
2002 		case IPV6_PKTINFO:
2003 		case IPV6_HOPOPTS:
2004 		case IPV6_RTHDR:
2005 		case IPV6_DSTOPTS:
2006 		case IPV6_RTHDRDSTOPTS:
2007 		case IPV6_NEXTHOP:
2008 		case IPV6_OTCLASS:
2009 		case IPV6_TCLASS:
2010 		case IPV6_DONTFRAG:
2011 		case IPV6_USE_MIN_MTU:
2012 			error = ip6_getpcbopt(in6p->in6p_outputopts,
2013 			    optname, sopt);
2014 			break;
2015 
2016 		case IPV6_MULTICAST_IF:
2017 		case IPV6_MULTICAST_HOPS:
2018 		case IPV6_MULTICAST_LOOP:
2019 		case IPV6_JOIN_GROUP:
2020 		case IPV6_LEAVE_GROUP:
2021 			error = ip6_getmoptions(sopt, in6p->in6p_moptions);
2022 			break;
2023 
2024 #if defined(IPSEC) || defined(FAST_IPSEC)
2025 		case IPV6_IPSEC_POLICY:
2026 		    {
2027 			struct mbuf *m = NULL;
2028 
2029 			/* XXX this will return EINVAL as sopt is empty */
2030 			error = ipsec6_get_policy(in6p, sopt->sopt_data,
2031 			    sopt->sopt_size, &m);
2032 			if (!error)
2033 				error = sockopt_setmbuf(sopt, m);
2034 
2035 			break;
2036 		    }
2037 #endif /* IPSEC */
2038 
2039 		default:
2040 			error = ENOPROTOOPT;
2041 			break;
2042 		}
2043 		break;
2044 	}
2045 	return (error);
2046 }
2047 
2048 int
2049 ip6_raw_ctloutput(int op, struct socket *so, struct sockopt *sopt)
2050 {
2051 	int error = 0, optval;
2052 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2053 	struct in6pcb *in6p = sotoin6pcb(so);
2054 	int level, optname;
2055 
2056 	KASSERT(sopt != NULL);
2057 
2058 	level = sopt->sopt_level;
2059 	optname = sopt->sopt_name;
2060 
2061 	if (level != IPPROTO_IPV6) {
2062 		return ENOPROTOOPT;
2063 	}
2064 
2065 	switch (optname) {
2066 	case IPV6_CHECKSUM:
2067 		/*
2068 		 * For ICMPv6 sockets, no modification allowed for checksum
2069 		 * offset, permit "no change" values to help existing apps.
2070 		 *
2071 		 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM
2072 		 * for an ICMPv6 socket will fail."  The current
2073 		 * behavior does not meet RFC3542.
2074 		 */
2075 		switch (op) {
2076 		case PRCO_SETOPT:
2077 			error = sockopt_getint(sopt, &optval);
2078 			if (error)
2079 				break;
2080 			if ((optval % 2) != 0) {
2081 				/* the API assumes even offset values */
2082 				error = EINVAL;
2083 			} else if (so->so_proto->pr_protocol ==
2084 			    IPPROTO_ICMPV6) {
2085 				if (optval != icmp6off)
2086 					error = EINVAL;
2087 			} else
2088 				in6p->in6p_cksum = optval;
2089 			break;
2090 
2091 		case PRCO_GETOPT:
2092 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2093 				optval = icmp6off;
2094 			else
2095 				optval = in6p->in6p_cksum;
2096 
2097 			error = sockopt_setint(sopt, optval);
2098 			break;
2099 
2100 		default:
2101 			error = EINVAL;
2102 			break;
2103 		}
2104 		break;
2105 
2106 	default:
2107 		error = ENOPROTOOPT;
2108 		break;
2109 	}
2110 
2111 	return (error);
2112 }
2113 
2114 #ifdef RFC2292
2115 /*
2116  * Set up IP6 options in pcb for insertion in output packets or
2117  * specifying behavior of outgoing packets.
2118  */
2119 static int
2120 ip6_pcbopts(struct ip6_pktopts **pktopt, struct socket *so,
2121     struct sockopt *sopt)
2122 {
2123 	struct ip6_pktopts *opt = *pktopt;
2124 	struct mbuf *m;
2125 	int error = 0;
2126 
2127 	/* turn off any old options. */
2128 	if (opt) {
2129 #ifdef DIAGNOSTIC
2130 	    if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2131 		opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2132 		opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2133 		    printf("ip6_pcbopts: all specified options are cleared.\n");
2134 #endif
2135 		ip6_clearpktopts(opt, -1);
2136 	} else {
2137 		opt = malloc(sizeof(*opt), M_IP6OPT, M_NOWAIT);
2138 		if (opt == NULL)
2139 			return (ENOBUFS);
2140 	}
2141 	*pktopt = NULL;
2142 
2143 	if (sopt == NULL || sopt->sopt_size == 0) {
2144 		/*
2145 		 * Only turning off any previous options, regardless of
2146 		 * whether the opt is just created or given.
2147 		 */
2148 		free(opt, M_IP6OPT);
2149 		return (0);
2150 	}
2151 
2152 	/*  set options specified by user. */
2153 	m = sockopt_getmbuf(sopt);
2154 	if (m == NULL) {
2155 		free(opt, M_IP6OPT);
2156 		return (ENOBUFS);
2157 	}
2158 
2159 	error = ip6_setpktopts(m, opt, NULL, kauth_cred_get(),
2160 	    so->so_proto->pr_protocol);
2161 	m_freem(m);
2162 	if (error != 0) {
2163 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2164 		free(opt, M_IP6OPT);
2165 		return (error);
2166 	}
2167 	*pktopt = opt;
2168 	return (0);
2169 }
2170 #endif
2171 
2172 /*
2173  * initialize ip6_pktopts.  beware that there are non-zero default values in
2174  * the struct.
2175  */
2176 void
2177 ip6_initpktopts(struct ip6_pktopts *opt)
2178 {
2179 
2180 	memset(opt, 0, sizeof(*opt));
2181 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2182 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2183 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2184 }
2185 
2186 #define sin6tosa(sin6)	((struct sockaddr *)(sin6)) /* XXX */
2187 static int
2188 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2189     kauth_cred_t cred, int uproto)
2190 {
2191 	struct ip6_pktopts *opt;
2192 
2193 	if (*pktopt == NULL) {
2194 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2195 		    M_NOWAIT);
2196 		if (*pktopt == NULL)
2197 			return (ENOBUFS);
2198 
2199 		ip6_initpktopts(*pktopt);
2200 	}
2201 	opt = *pktopt;
2202 
2203 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2204 }
2205 
2206 static int
2207 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2208 {
2209 	void *optdata = NULL;
2210 	int optdatalen = 0;
2211 	struct ip6_ext *ip6e;
2212 	int error = 0;
2213 	struct in6_pktinfo null_pktinfo;
2214 	int deftclass = 0, on;
2215 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2216 
2217 	switch (optname) {
2218 	case IPV6_PKTINFO:
2219 		if (pktopt && pktopt->ip6po_pktinfo)
2220 			optdata = (void *)pktopt->ip6po_pktinfo;
2221 		else {
2222 			/* XXX: we don't have to do this every time... */
2223 			memset(&null_pktinfo, 0, sizeof(null_pktinfo));
2224 			optdata = (void *)&null_pktinfo;
2225 		}
2226 		optdatalen = sizeof(struct in6_pktinfo);
2227 		break;
2228 	case IPV6_OTCLASS:
2229 		/* XXX */
2230 		return (EINVAL);
2231 	case IPV6_TCLASS:
2232 		if (pktopt && pktopt->ip6po_tclass >= 0)
2233 			optdata = (void *)&pktopt->ip6po_tclass;
2234 		else
2235 			optdata = (void *)&deftclass;
2236 		optdatalen = sizeof(int);
2237 		break;
2238 	case IPV6_HOPOPTS:
2239 		if (pktopt && pktopt->ip6po_hbh) {
2240 			optdata = (void *)pktopt->ip6po_hbh;
2241 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2242 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2243 		}
2244 		break;
2245 	case IPV6_RTHDR:
2246 		if (pktopt && pktopt->ip6po_rthdr) {
2247 			optdata = (void *)pktopt->ip6po_rthdr;
2248 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2249 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2250 		}
2251 		break;
2252 	case IPV6_RTHDRDSTOPTS:
2253 		if (pktopt && pktopt->ip6po_dest1) {
2254 			optdata = (void *)pktopt->ip6po_dest1;
2255 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2256 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2257 		}
2258 		break;
2259 	case IPV6_DSTOPTS:
2260 		if (pktopt && pktopt->ip6po_dest2) {
2261 			optdata = (void *)pktopt->ip6po_dest2;
2262 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2263 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2264 		}
2265 		break;
2266 	case IPV6_NEXTHOP:
2267 		if (pktopt && pktopt->ip6po_nexthop) {
2268 			optdata = (void *)pktopt->ip6po_nexthop;
2269 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2270 		}
2271 		break;
2272 	case IPV6_USE_MIN_MTU:
2273 		if (pktopt)
2274 			optdata = (void *)&pktopt->ip6po_minmtu;
2275 		else
2276 			optdata = (void *)&defminmtu;
2277 		optdatalen = sizeof(int);
2278 		break;
2279 	case IPV6_DONTFRAG:
2280 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2281 			on = 1;
2282 		else
2283 			on = 0;
2284 		optdata = (void *)&on;
2285 		optdatalen = sizeof(on);
2286 		break;
2287 	default:		/* should not happen */
2288 #ifdef DIAGNOSTIC
2289 		panic("ip6_getpcbopt: unexpected option\n");
2290 #endif
2291 		return (ENOPROTOOPT);
2292 	}
2293 
2294 	error = sockopt_set(sopt, optdata, optdatalen);
2295 
2296 	return (error);
2297 }
2298 
2299 void
2300 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2301 {
2302 	if (optname == -1 || optname == IPV6_PKTINFO) {
2303 		if (pktopt->ip6po_pktinfo)
2304 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2305 		pktopt->ip6po_pktinfo = NULL;
2306 	}
2307 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2308 		pktopt->ip6po_hlim = -1;
2309 	if (optname == -1 || optname == IPV6_TCLASS)
2310 		pktopt->ip6po_tclass = -1;
2311 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2312 		rtcache_free(&pktopt->ip6po_nextroute);
2313 		if (pktopt->ip6po_nexthop)
2314 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2315 		pktopt->ip6po_nexthop = NULL;
2316 	}
2317 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2318 		if (pktopt->ip6po_hbh)
2319 			free(pktopt->ip6po_hbh, M_IP6OPT);
2320 		pktopt->ip6po_hbh = NULL;
2321 	}
2322 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2323 		if (pktopt->ip6po_dest1)
2324 			free(pktopt->ip6po_dest1, M_IP6OPT);
2325 		pktopt->ip6po_dest1 = NULL;
2326 	}
2327 	if (optname == -1 || optname == IPV6_RTHDR) {
2328 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2329 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2330 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2331 		rtcache_free(&pktopt->ip6po_route);
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 		memcpy(dst->type, src->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 	dst->ip6po_hlim = src->ip6po_hlim;
2355 	dst->ip6po_tclass = src->ip6po_tclass;
2356 	dst->ip6po_flags = src->ip6po_flags;
2357 	if (src->ip6po_pktinfo) {
2358 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2359 		    M_IP6OPT, canwait);
2360 		if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT)
2361 			goto bad;
2362 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2363 	}
2364 	if (src->ip6po_nexthop) {
2365 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2366 		    M_IP6OPT, canwait);
2367 		if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT)
2368 			goto bad;
2369 		memcpy(dst->ip6po_nexthop, src->ip6po_nexthop,
2370 		    src->ip6po_nexthop->sa_len);
2371 	}
2372 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2373 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2374 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2375 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2376 	return (0);
2377 
2378   bad:
2379 	if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
2380 	if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
2381 	if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
2382 	if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
2383 	if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
2384 	if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
2385 
2386 	return (ENOBUFS);
2387 }
2388 #undef PKTOPT_EXTHDRCPY
2389 
2390 struct ip6_pktopts *
2391 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2392 {
2393 	int error;
2394 	struct ip6_pktopts *dst;
2395 
2396 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2397 	if (dst == NULL && canwait == M_NOWAIT)
2398 		return (NULL);
2399 	ip6_initpktopts(dst);
2400 
2401 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2402 		free(dst, M_IP6OPT);
2403 		return (NULL);
2404 	}
2405 
2406 	return (dst);
2407 }
2408 
2409 void
2410 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2411 {
2412 	if (pktopt == NULL)
2413 		return;
2414 
2415 	ip6_clearpktopts(pktopt, -1);
2416 
2417 	free(pktopt, M_IP6OPT);
2418 }
2419 
2420 /*
2421  * Set the IP6 multicast options in response to user setsockopt().
2422  */
2423 static int
2424 ip6_setmoptions(const struct sockopt *sopt, struct ip6_moptions **im6op)
2425 {
2426 	int error = 0;
2427 	u_int loop, ifindex;
2428 	struct ipv6_mreq mreq;
2429 	struct ifnet *ifp;
2430 	struct ip6_moptions *im6o = *im6op;
2431 	struct route ro;
2432 	struct in6_multi_mship *imm;
2433 	struct lwp *l = curlwp;	/* XXX */
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 = malloc(sizeof(*im6o), M_IPMOPTS, M_NOWAIT);
2441 		if (im6o == NULL)
2442 			return (ENOBUFS);
2443 
2444 		*im6op = im6o;
2445 		im6o->im6o_multicast_ifp = NULL;
2446 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2447 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2448 		LIST_INIT(&im6o->im6o_memberships);
2449 	}
2450 
2451 	switch (sopt->sopt_name) {
2452 
2453 	case IPV6_MULTICAST_IF:
2454 		/*
2455 		 * Select the interface for outgoing multicast packets.
2456 		 */
2457 		error = sockopt_get(sopt, &ifindex, sizeof(ifindex));
2458 		if (error != 0)
2459 			break;
2460 
2461 		if (ifindex != 0) {
2462 			if (if_indexlim <= ifindex || !ifindex2ifnet[ifindex]) {
2463 				error = ENXIO;	/* XXX EINVAL? */
2464 				break;
2465 			}
2466 			ifp = ifindex2ifnet[ifindex];
2467 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
2468 				error = EADDRNOTAVAIL;
2469 				break;
2470 			}
2471 		} else
2472 			ifp = NULL;
2473 		im6o->im6o_multicast_ifp = ifp;
2474 		break;
2475 
2476 	case IPV6_MULTICAST_HOPS:
2477 	    {
2478 		/*
2479 		 * Set the IP6 hoplimit for outgoing multicast packets.
2480 		 */
2481 		int optval;
2482 
2483 		error = sockopt_getint(sopt, &optval);
2484 		if (error != 0)
2485 			break;
2486 
2487 		if (optval < -1 || optval >= 256)
2488 			error = EINVAL;
2489 		else if (optval == -1)
2490 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2491 		else
2492 			im6o->im6o_multicast_hlim = optval;
2493 		break;
2494 	    }
2495 
2496 	case IPV6_MULTICAST_LOOP:
2497 		/*
2498 		 * Set the loopback flag for outgoing multicast packets.
2499 		 * Must be zero or one.
2500 		 */
2501 		error = sockopt_get(sopt, &loop, sizeof(loop));
2502 		if (error != 0)
2503 			break;
2504 		if (loop > 1) {
2505 			error = EINVAL;
2506 			break;
2507 		}
2508 		im6o->im6o_multicast_loop = loop;
2509 		break;
2510 
2511 	case IPV6_JOIN_GROUP:
2512 		/*
2513 		 * Add a multicast group membership.
2514 		 * Group must be a valid IP6 multicast address.
2515 		 */
2516 		error = sockopt_get(sopt, &mreq, sizeof(mreq));
2517 		if (error != 0)
2518 			break;
2519 
2520 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq.ipv6mr_multiaddr)) {
2521 			/*
2522 			 * We use the unspecified address to specify to accept
2523 			 * all multicast addresses. Only super user is allowed
2524 			 * to do this.
2525 			 */
2526 			if (kauth_authorize_generic(l->l_cred,
2527 			    KAUTH_GENERIC_ISSUSER, NULL))
2528 			{
2529 				error = EACCES;
2530 				break;
2531 			}
2532 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq.ipv6mr_multiaddr)) {
2533 			error = EINVAL;
2534 			break;
2535 		}
2536 
2537 		/*
2538 		 * If no interface was explicitly specified, choose an
2539 		 * appropriate one according to the given multicast address.
2540 		 */
2541 		if (mreq.ipv6mr_interface == 0) {
2542 			struct rtentry *rt;
2543 			union {
2544 				struct sockaddr		dst;
2545 				struct sockaddr_in6	dst6;
2546 			} u;
2547 
2548 			/*
2549 			 * Look up the routing table for the
2550 			 * address, and choose the outgoing interface.
2551 			 *   XXX: is it a good approach?
2552 			 */
2553 			memset(&ro, 0, sizeof(ro));
2554 			sockaddr_in6_init(&u.dst6, &mreq.ipv6mr_multiaddr, 0,
2555 			    0, 0);
2556 			rtcache_setdst(&ro, &u.dst);
2557 			ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp
2558 			                                        : NULL;
2559 			rtcache_free(&ro);
2560 		} else {
2561 			/*
2562 			 * If the interface is specified, validate it.
2563 			 */
2564 			if (if_indexlim <= mreq.ipv6mr_interface ||
2565 			    !ifindex2ifnet[mreq.ipv6mr_interface]) {
2566 				error = ENXIO;	/* XXX EINVAL? */
2567 				break;
2568 			}
2569 			ifp = ifindex2ifnet[mreq.ipv6mr_interface];
2570 		}
2571 
2572 		/*
2573 		 * See if we found an interface, and confirm that it
2574 		 * supports multicast
2575 		 */
2576 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2577 			error = EADDRNOTAVAIL;
2578 			break;
2579 		}
2580 
2581 		if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) {
2582 			error = EADDRNOTAVAIL; /* XXX: should not happen */
2583 			break;
2584 		}
2585 
2586 		/*
2587 		 * See if the membership already exists.
2588 		 */
2589 		for (imm = im6o->im6o_memberships.lh_first;
2590 		     imm != NULL; imm = imm->i6mm_chain.le_next)
2591 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
2592 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2593 			    &mreq.ipv6mr_multiaddr))
2594 				break;
2595 		if (imm != NULL) {
2596 			error = EADDRINUSE;
2597 			break;
2598 		}
2599 		/*
2600 		 * Everything looks good; add a new record to the multicast
2601 		 * address list for the given interface.
2602 		 */
2603 		imm = in6_joingroup(ifp, &mreq.ipv6mr_multiaddr, &error, 0);
2604 		if (imm == NULL)
2605 			break;
2606 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2607 		break;
2608 
2609 	case IPV6_LEAVE_GROUP:
2610 		/*
2611 		 * Drop a multicast group membership.
2612 		 * Group must be a valid IP6 multicast address.
2613 		 */
2614 		error = sockopt_get(sopt, &mreq, sizeof(mreq));
2615 		if (error != 0)
2616 			break;
2617 
2618 		/*
2619 		 * If an interface address was specified, get a pointer
2620 		 * to its ifnet structure.
2621 		 */
2622 		if (mreq.ipv6mr_interface != 0) {
2623 			if (if_indexlim <= mreq.ipv6mr_interface ||
2624 			    !ifindex2ifnet[mreq.ipv6mr_interface]) {
2625 				error = ENXIO;	/* XXX EINVAL? */
2626 				break;
2627 			}
2628 			ifp = ifindex2ifnet[mreq.ipv6mr_interface];
2629 		} else
2630 			ifp = NULL;
2631 
2632 		/* Fill in the scope zone ID */
2633 		if (ifp) {
2634 			if (in6_setscope(&mreq.ipv6mr_multiaddr, ifp, NULL)) {
2635 				/* XXX: should not happen */
2636 				error = EADDRNOTAVAIL;
2637 				break;
2638 			}
2639 		} else if (mreq.ipv6mr_interface != 0) {
2640 			/*
2641 			 * XXX: This case would happens when the (positive)
2642 			 * index is in the valid range, but the corresponding
2643 			 * interface has been detached dynamically.  The above
2644 			 * check probably avoids such case to happen here, but
2645 			 * we check it explicitly for safety.
2646 			 */
2647 			error = EADDRNOTAVAIL;
2648 			break;
2649 		} else {	/* ipv6mr_interface == 0 */
2650 			struct sockaddr_in6 sa6_mc;
2651 
2652 			/*
2653 			 * The API spec says as follows:
2654 			 *  If the interface index is specified as 0, the
2655 			 *  system may choose a multicast group membership to
2656 			 *  drop by matching the multicast address only.
2657 			 * On the other hand, we cannot disambiguate the scope
2658 			 * zone unless an interface is provided.  Thus, we
2659 			 * check if there's ambiguity with the default scope
2660 			 * zone as the last resort.
2661 			 */
2662 			sockaddr_in6_init(&sa6_mc, &mreq.ipv6mr_multiaddr,
2663 			    0, 0, 0);
2664 			error = sa6_embedscope(&sa6_mc, ip6_use_defzone);
2665 			if (error != 0)
2666 				break;
2667 			mreq.ipv6mr_multiaddr = sa6_mc.sin6_addr;
2668 		}
2669 
2670 		/*
2671 		 * Find the membership in the membership list.
2672 		 */
2673 		for (imm = im6o->im6o_memberships.lh_first;
2674 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
2675 			if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2676 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2677 			    &mreq.ipv6mr_multiaddr))
2678 				break;
2679 		}
2680 		if (imm == NULL) {
2681 			/* Unable to resolve interface */
2682 			error = EADDRNOTAVAIL;
2683 			break;
2684 		}
2685 		/*
2686 		 * Give up the multicast address record to which the
2687 		 * membership points.
2688 		 */
2689 		LIST_REMOVE(imm, i6mm_chain);
2690 		in6_leavegroup(imm);
2691 		break;
2692 
2693 	default:
2694 		error = EOPNOTSUPP;
2695 		break;
2696 	}
2697 
2698 	/*
2699 	 * If all options have default values, no need to keep the mbuf.
2700 	 */
2701 	if (im6o->im6o_multicast_ifp == NULL &&
2702 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2703 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2704 	    im6o->im6o_memberships.lh_first == NULL) {
2705 		free(*im6op, M_IPMOPTS);
2706 		*im6op = NULL;
2707 	}
2708 
2709 	return (error);
2710 }
2711 
2712 /*
2713  * Return the IP6 multicast options in response to user getsockopt().
2714  */
2715 static int
2716 ip6_getmoptions(struct sockopt *sopt, struct ip6_moptions *im6o)
2717 {
2718 	u_int optval;
2719 	int error;
2720 
2721 	switch (sopt->sopt_name) {
2722 	case IPV6_MULTICAST_IF:
2723 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2724 			optval = 0;
2725 		else
2726 			optval = im6o->im6o_multicast_ifp->if_index;
2727 
2728 		error = sockopt_set(sopt, &optval, sizeof(optval));
2729 		break;
2730 
2731 	case IPV6_MULTICAST_HOPS:
2732 		if (im6o == NULL)
2733 			optval = ip6_defmcasthlim;
2734 		else
2735 			optval = im6o->im6o_multicast_hlim;
2736 
2737 		error = sockopt_set(sopt, &optval, sizeof(optval));
2738 		break;
2739 
2740 	case IPV6_MULTICAST_LOOP:
2741 		if (im6o == NULL)
2742 			optval = ip6_defmcasthlim;
2743 		else
2744 			optval = im6o->im6o_multicast_loop;
2745 
2746 		error = sockopt_set(sopt, &optval, sizeof(optval));
2747 		break;
2748 
2749 	default:
2750 		error = EOPNOTSUPP;
2751 	}
2752 
2753 	return (error);
2754 }
2755 
2756 /*
2757  * Discard the IP6 multicast options.
2758  */
2759 void
2760 ip6_freemoptions(struct ip6_moptions *im6o)
2761 {
2762 	struct in6_multi_mship *imm;
2763 
2764 	if (im6o == NULL)
2765 		return;
2766 
2767 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2768 		LIST_REMOVE(imm, i6mm_chain);
2769 		in6_leavegroup(imm);
2770 	}
2771 	free(im6o, M_IPMOPTS);
2772 }
2773 
2774 /*
2775  * Set IPv6 outgoing packet options based on advanced API.
2776  */
2777 int
2778 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2779 	struct ip6_pktopts *stickyopt, kauth_cred_t cred, int uproto)
2780 {
2781 	struct cmsghdr *cm = 0;
2782 
2783 	if (control == NULL || opt == NULL)
2784 		return (EINVAL);
2785 
2786 	ip6_initpktopts(opt);
2787 	if (stickyopt) {
2788 		int error;
2789 
2790 		/*
2791 		 * If stickyopt is provided, make a local copy of the options
2792 		 * for this particular packet, then override them by ancillary
2793 		 * objects.
2794 		 * XXX: copypktopts() does not copy the cached route to a next
2795 		 * hop (if any).  This is not very good in terms of efficiency,
2796 		 * but we can allow this since this option should be rarely
2797 		 * used.
2798 		 */
2799 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2800 			return (error);
2801 	}
2802 
2803 	/*
2804 	 * XXX: Currently, we assume all the optional information is stored
2805 	 * in a single mbuf.
2806 	 */
2807 	if (control->m_next)
2808 		return (EINVAL);
2809 
2810 	/* XXX if cm->cmsg_len is not aligned, control->m_len can become <0 */
2811 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2812 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2813 		int error;
2814 
2815 		if (control->m_len < CMSG_LEN(0))
2816 			return (EINVAL);
2817 
2818 		cm = mtod(control, struct cmsghdr *);
2819 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2820 			return (EINVAL);
2821 		if (cm->cmsg_level != IPPROTO_IPV6)
2822 			continue;
2823 
2824 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2825 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2826 		if (error)
2827 			return (error);
2828 	}
2829 
2830 	return (0);
2831 }
2832 
2833 /*
2834  * Set a particular packet option, as a sticky option or an ancillary data
2835  * item.  "len" can be 0 only when it's a sticky option.
2836  * We have 4 cases of combination of "sticky" and "cmsg":
2837  * "sticky=0, cmsg=0": impossible
2838  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2839  * "sticky=1, cmsg=0": RFC3542 socket option
2840  * "sticky=1, cmsg=1": RFC2292 socket option
2841  */
2842 static int
2843 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2844     kauth_cred_t cred, int sticky, int cmsg, int uproto)
2845 {
2846 	int minmtupolicy;
2847 	int error;
2848 
2849 	if (!sticky && !cmsg) {
2850 #ifdef DIAGNOSTIC
2851 		printf("ip6_setpktopt: impossible case\n");
2852 #endif
2853 		return (EINVAL);
2854 	}
2855 
2856 	/*
2857 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2858 	 * not be specified in the context of RFC3542.  Conversely,
2859 	 * RFC3542 types should not be specified in the context of RFC2292.
2860 	 */
2861 	if (!cmsg) {
2862 		switch (optname) {
2863 		case IPV6_2292PKTINFO:
2864 		case IPV6_2292HOPLIMIT:
2865 		case IPV6_2292NEXTHOP:
2866 		case IPV6_2292HOPOPTS:
2867 		case IPV6_2292DSTOPTS:
2868 		case IPV6_2292RTHDR:
2869 		case IPV6_2292PKTOPTIONS:
2870 			return (ENOPROTOOPT);
2871 		}
2872 	}
2873 	if (sticky && cmsg) {
2874 		switch (optname) {
2875 		case IPV6_PKTINFO:
2876 		case IPV6_HOPLIMIT:
2877 		case IPV6_NEXTHOP:
2878 		case IPV6_HOPOPTS:
2879 		case IPV6_DSTOPTS:
2880 		case IPV6_RTHDRDSTOPTS:
2881 		case IPV6_RTHDR:
2882 		case IPV6_USE_MIN_MTU:
2883 		case IPV6_DONTFRAG:
2884 		case IPV6_OTCLASS:
2885 		case IPV6_TCLASS:
2886 			return (ENOPROTOOPT);
2887 		}
2888 	}
2889 
2890 	switch (optname) {
2891 #ifdef RFC2292
2892 	case IPV6_2292PKTINFO:
2893 #endif
2894 	case IPV6_PKTINFO:
2895 	{
2896 		struct ifnet *ifp = NULL;
2897 		struct in6_pktinfo *pktinfo;
2898 
2899 		if (len != sizeof(struct in6_pktinfo))
2900 			return (EINVAL);
2901 
2902 		pktinfo = (struct in6_pktinfo *)buf;
2903 
2904 		/*
2905 		 * An application can clear any sticky IPV6_PKTINFO option by
2906 		 * doing a "regular" setsockopt with ipi6_addr being
2907 		 * in6addr_any and ipi6_ifindex being zero.
2908 		 * [RFC 3542, Section 6]
2909 		 */
2910 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2911 		    pktinfo->ipi6_ifindex == 0 &&
2912 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2913 			ip6_clearpktopts(opt, optname);
2914 			break;
2915 		}
2916 
2917 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2918 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2919 			return (EINVAL);
2920 		}
2921 
2922 		/* validate the interface index if specified. */
2923 		if (pktinfo->ipi6_ifindex >= if_indexlim) {
2924 			 return (ENXIO);
2925 		}
2926 		if (pktinfo->ipi6_ifindex) {
2927 			ifp = ifindex2ifnet[pktinfo->ipi6_ifindex];
2928 			if (ifp == NULL)
2929 				return (ENXIO);
2930 		}
2931 
2932 		/*
2933 		 * We store the address anyway, and let in6_selectsrc()
2934 		 * validate the specified address.  This is because ipi6_addr
2935 		 * may not have enough information about its scope zone, and
2936 		 * we may need additional information (such as outgoing
2937 		 * interface or the scope zone of a destination address) to
2938 		 * disambiguate the scope.
2939 		 * XXX: the delay of the validation may confuse the
2940 		 * application when it is used as a sticky option.
2941 		 */
2942 		if (opt->ip6po_pktinfo == NULL) {
2943 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2944 			    M_IP6OPT, M_NOWAIT);
2945 			if (opt->ip6po_pktinfo == NULL)
2946 				return (ENOBUFS);
2947 		}
2948 		memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo));
2949 		break;
2950 	}
2951 
2952 #ifdef RFC2292
2953 	case IPV6_2292HOPLIMIT:
2954 #endif
2955 	case IPV6_HOPLIMIT:
2956 	{
2957 		int *hlimp;
2958 
2959 		/*
2960 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2961 		 * to simplify the ordering among hoplimit options.
2962 		 */
2963 		if (optname == IPV6_HOPLIMIT && sticky)
2964 			return (ENOPROTOOPT);
2965 
2966 		if (len != sizeof(int))
2967 			return (EINVAL);
2968 		hlimp = (int *)buf;
2969 		if (*hlimp < -1 || *hlimp > 255)
2970 			return (EINVAL);
2971 
2972 		opt->ip6po_hlim = *hlimp;
2973 		break;
2974 	}
2975 
2976 	case IPV6_OTCLASS:
2977 		if (len != sizeof(u_int8_t))
2978 			return (EINVAL);
2979 
2980 		opt->ip6po_tclass = *(u_int8_t *)buf;
2981 		break;
2982 
2983 	case IPV6_TCLASS:
2984 	{
2985 		int tclass;
2986 
2987 		if (len != sizeof(int))
2988 			return (EINVAL);
2989 		tclass = *(int *)buf;
2990 		if (tclass < -1 || tclass > 255)
2991 			return (EINVAL);
2992 
2993 		opt->ip6po_tclass = tclass;
2994 		break;
2995 	}
2996 
2997 #ifdef RFC2292
2998 	case IPV6_2292NEXTHOP:
2999 #endif
3000 	case IPV6_NEXTHOP:
3001 		error = kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER,
3002 		    NULL);
3003 		if (error)
3004 			return (error);
3005 
3006 		if (len == 0) {	/* just remove the option */
3007 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
3008 			break;
3009 		}
3010 
3011 		/* check if cmsg_len is large enough for sa_len */
3012 		if (len < sizeof(struct sockaddr) || len < *buf)
3013 			return (EINVAL);
3014 
3015 		switch (((struct sockaddr *)buf)->sa_family) {
3016 		case AF_INET6:
3017 		{
3018 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3019 
3020 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3021 				return (EINVAL);
3022 
3023 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3024 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3025 				return (EINVAL);
3026 			}
3027 			if ((error = sa6_embedscope(sa6, ip6_use_defzone))
3028 			    != 0) {
3029 				return (error);
3030 			}
3031 			break;
3032 		}
3033 		case AF_LINK:	/* eventually be supported? */
3034 		default:
3035 			return (EAFNOSUPPORT);
3036 		}
3037 
3038 		/* turn off the previous option, then set the new option. */
3039 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
3040 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3041 		if (opt->ip6po_nexthop == NULL)
3042 			return (ENOBUFS);
3043 		memcpy(opt->ip6po_nexthop, buf, *buf);
3044 		break;
3045 
3046 #ifdef RFC2292
3047 	case IPV6_2292HOPOPTS:
3048 #endif
3049 	case IPV6_HOPOPTS:
3050 	{
3051 		struct ip6_hbh *hbh;
3052 		int hbhlen;
3053 
3054 		/*
3055 		 * XXX: We don't allow a non-privileged user to set ANY HbH
3056 		 * options, since per-option restriction has too much
3057 		 * overhead.
3058 		 */
3059 		error = kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER,
3060 		    NULL);
3061 		if (error)
3062 			return (error);
3063 
3064 		if (len == 0) {
3065 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
3066 			break;	/* just remove the option */
3067 		}
3068 
3069 		/* message length validation */
3070 		if (len < sizeof(struct ip6_hbh))
3071 			return (EINVAL);
3072 		hbh = (struct ip6_hbh *)buf;
3073 		hbhlen = (hbh->ip6h_len + 1) << 3;
3074 		if (len != hbhlen)
3075 			return (EINVAL);
3076 
3077 		/* turn off the previous option, then set the new option. */
3078 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
3079 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3080 		if (opt->ip6po_hbh == NULL)
3081 			return (ENOBUFS);
3082 		memcpy(opt->ip6po_hbh, hbh, hbhlen);
3083 
3084 		break;
3085 	}
3086 
3087 #ifdef RFC2292
3088 	case IPV6_2292DSTOPTS:
3089 #endif
3090 	case IPV6_DSTOPTS:
3091 	case IPV6_RTHDRDSTOPTS:
3092 	{
3093 		struct ip6_dest *dest, **newdest = NULL;
3094 		int destlen;
3095 
3096 		/* XXX: see the comment for IPV6_HOPOPTS */
3097 		error = kauth_authorize_generic(cred, KAUTH_GENERIC_ISSUSER,
3098 		    NULL);
3099 		if (error)
3100 			return (error);
3101 
3102 		if (len == 0) {
3103 			ip6_clearpktopts(opt, optname);
3104 			break;	/* just remove the option */
3105 		}
3106 
3107 		/* message length validation */
3108 		if (len < sizeof(struct ip6_dest))
3109 			return (EINVAL);
3110 		dest = (struct ip6_dest *)buf;
3111 		destlen = (dest->ip6d_len + 1) << 3;
3112 		if (len != destlen)
3113 			return (EINVAL);
3114 		/*
3115 		 * Determine the position that the destination options header
3116 		 * should be inserted; before or after the routing header.
3117 		 */
3118 		switch (optname) {
3119 		case IPV6_2292DSTOPTS:
3120 			/*
3121 			 * The old advanced API is ambiguous on this point.
3122 			 * Our approach is to determine the position based
3123 			 * according to the existence of a routing header.
3124 			 * Note, however, that this depends on the order of the
3125 			 * extension headers in the ancillary data; the 1st
3126 			 * part of the destination options header must appear
3127 			 * before the routing header in the ancillary data,
3128 			 * too.
3129 			 * RFC3542 solved the ambiguity by introducing
3130 			 * separate ancillary data or option types.
3131 			 */
3132 			if (opt->ip6po_rthdr == NULL)
3133 				newdest = &opt->ip6po_dest1;
3134 			else
3135 				newdest = &opt->ip6po_dest2;
3136 			break;
3137 		case IPV6_RTHDRDSTOPTS:
3138 			newdest = &opt->ip6po_dest1;
3139 			break;
3140 		case IPV6_DSTOPTS:
3141 			newdest = &opt->ip6po_dest2;
3142 			break;
3143 		}
3144 
3145 		/* turn off the previous option, then set the new option. */
3146 		ip6_clearpktopts(opt, optname);
3147 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3148 		if (*newdest == NULL)
3149 			return (ENOBUFS);
3150 		memcpy(*newdest, dest, destlen);
3151 
3152 		break;
3153 	}
3154 
3155 #ifdef RFC2292
3156 	case IPV6_2292RTHDR:
3157 #endif
3158 	case IPV6_RTHDR:
3159 	{
3160 		struct ip6_rthdr *rth;
3161 		int rthlen;
3162 
3163 		if (len == 0) {
3164 			ip6_clearpktopts(opt, IPV6_RTHDR);
3165 			break;	/* just remove the option */
3166 		}
3167 
3168 		/* message length validation */
3169 		if (len < sizeof(struct ip6_rthdr))
3170 			return (EINVAL);
3171 		rth = (struct ip6_rthdr *)buf;
3172 		rthlen = (rth->ip6r_len + 1) << 3;
3173 		if (len != rthlen)
3174 			return (EINVAL);
3175 		switch (rth->ip6r_type) {
3176 		case IPV6_RTHDR_TYPE_0:
3177 			if (rth->ip6r_len == 0)	/* must contain one addr */
3178 				return (EINVAL);
3179 			if (rth->ip6r_len % 2) /* length must be even */
3180 				return (EINVAL);
3181 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3182 				return (EINVAL);
3183 			break;
3184 		default:
3185 			return (EINVAL);	/* not supported */
3186 		}
3187 		/* turn off the previous option */
3188 		ip6_clearpktopts(opt, IPV6_RTHDR);
3189 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3190 		if (opt->ip6po_rthdr == NULL)
3191 			return (ENOBUFS);
3192 		memcpy(opt->ip6po_rthdr, rth, rthlen);
3193 		break;
3194 	}
3195 
3196 	case IPV6_USE_MIN_MTU:
3197 		if (len != sizeof(int))
3198 			return (EINVAL);
3199 		minmtupolicy = *(int *)buf;
3200 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3201 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3202 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3203 			return (EINVAL);
3204 		}
3205 		opt->ip6po_minmtu = minmtupolicy;
3206 		break;
3207 
3208 	case IPV6_DONTFRAG:
3209 		if (len != sizeof(int))
3210 			return (EINVAL);
3211 
3212 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3213 			/*
3214 			 * we ignore this option for TCP sockets.
3215 			 * (RFC3542 leaves this case unspecified.)
3216 			 */
3217 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3218 		} else
3219 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3220 		break;
3221 
3222 	default:
3223 		return (ENOPROTOOPT);
3224 	} /* end of switch */
3225 
3226 	return (0);
3227 }
3228 
3229 /*
3230  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3231  * packet to the input queue of a specified interface.  Note that this
3232  * calls the output routine of the loopback "driver", but with an interface
3233  * pointer that might NOT be lo0ifp -- easier than replicating that code here.
3234  */
3235 void
3236 ip6_mloopback(struct ifnet *ifp, struct mbuf *m,
3237 	const struct sockaddr_in6 *dst)
3238 {
3239 	struct mbuf *copym;
3240 	struct ip6_hdr *ip6;
3241 
3242 	copym = m_copy(m, 0, M_COPYALL);
3243 	if (copym == NULL)
3244 		return;
3245 
3246 	/*
3247 	 * Make sure to deep-copy IPv6 header portion in case the data
3248 	 * is in an mbuf cluster, so that we can safely override the IPv6
3249 	 * header portion later.
3250 	 */
3251 	if ((copym->m_flags & M_EXT) != 0 ||
3252 	    copym->m_len < sizeof(struct ip6_hdr)) {
3253 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3254 		if (copym == NULL)
3255 			return;
3256 	}
3257 
3258 #ifdef DIAGNOSTIC
3259 	if (copym->m_len < sizeof(*ip6)) {
3260 		m_freem(copym);
3261 		return;
3262 	}
3263 #endif
3264 
3265 	ip6 = mtod(copym, struct ip6_hdr *);
3266 	/*
3267 	 * clear embedded scope identifiers if necessary.
3268 	 * in6_clearscope will touch the addresses only when necessary.
3269 	 */
3270 	in6_clearscope(&ip6->ip6_src);
3271 	in6_clearscope(&ip6->ip6_dst);
3272 
3273 	(void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL);
3274 }
3275 
3276 /*
3277  * Chop IPv6 header off from the payload.
3278  */
3279 static int
3280 ip6_splithdr(struct mbuf *m,  struct ip6_exthdrs *exthdrs)
3281 {
3282 	struct mbuf *mh;
3283 	struct ip6_hdr *ip6;
3284 
3285 	ip6 = mtod(m, struct ip6_hdr *);
3286 	if (m->m_len > sizeof(*ip6)) {
3287 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3288 		if (mh == 0) {
3289 			m_freem(m);
3290 			return ENOBUFS;
3291 		}
3292 		M_MOVE_PKTHDR(mh, m);
3293 		MH_ALIGN(mh, sizeof(*ip6));
3294 		m->m_len -= sizeof(*ip6);
3295 		m->m_data += sizeof(*ip6);
3296 		mh->m_next = m;
3297 		m = mh;
3298 		m->m_len = sizeof(*ip6);
3299 		bcopy((void *)ip6, mtod(m, void *), sizeof(*ip6));
3300 	}
3301 	exthdrs->ip6e_ip6 = m;
3302 	return 0;
3303 }
3304 
3305 /*
3306  * Compute IPv6 extension header length.
3307  */
3308 int
3309 ip6_optlen(struct in6pcb *in6p)
3310 {
3311 	int len;
3312 
3313 	if (!in6p->in6p_outputopts)
3314 		return 0;
3315 
3316 	len = 0;
3317 #define elen(x) \
3318     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3319 
3320 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3321 	len += elen(in6p->in6p_outputopts->ip6po_dest1);
3322 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3323 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3324 	return len;
3325 #undef elen
3326 }
3327