xref: /netbsd-src/sys/netinet6/ip6_output.c (revision 267197ec1eebfcb9810ea27a89625b6ddf68e3e7)
1 /*	$NetBSD: ip6_output.c,v 1.126 2008/01/14 04:16:45 dyoung 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 1.126 2008/01/14 04:16:45 dyoung 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/in6_pcb.h>
97 #include <netinet6/nd6.h>
98 #include <netinet6/ip6protosw.h>
99 #include <netinet6/scope6_var.h>
100 
101 #ifdef IPSEC
102 #include <netinet6/ipsec.h>
103 #include <netkey/key.h>
104 #endif /* IPSEC */
105 
106 #ifdef FAST_IPSEC
107 #include <netipsec/ipsec.h>
108 #include <netipsec/ipsec6.h>
109 #include <netipsec/key.h>
110 #include <netipsec/xform.h>
111 #endif
112 
113 
114 #include <net/net_osdep.h>
115 
116 #ifdef PFIL_HOOKS
117 extern struct pfil_head inet6_pfil_hook;	/* XXX */
118 #endif
119 
120 struct ip6_exthdrs {
121 	struct mbuf *ip6e_ip6;
122 	struct mbuf *ip6e_hbh;
123 	struct mbuf *ip6e_dest1;
124 	struct mbuf *ip6e_rthdr;
125 	struct mbuf *ip6e_dest2;
126 };
127 
128 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
129 	int, int);
130 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct mbuf **);
131 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, int,
132 	int, int, int);
133 static int ip6_setmoptions(int, struct ip6_moptions **, struct mbuf *);
134 static int ip6_getmoptions(int, struct ip6_moptions *, struct mbuf **);
135 static int ip6_copyexthdr(struct mbuf **, void *, int);
136 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
137 	struct ip6_frag **);
138 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
139 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
140 static int ip6_getpmtu(struct route *, struct route *, struct ifnet *,
141     const struct in6_addr *, u_long *, int *);
142 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
143 
144 #ifdef RFC2292
145 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
146 	struct socket *);
147 #endif
148 
149 #define	IN6_NEED_CHECKSUM(ifp, csum_flags) \
150 	(__predict_true(((ifp)->if_flags & IFF_LOOPBACK) == 0 || \
151 	(((csum_flags) & M_CSUM_UDPv6) != 0 && udp_do_loopback_cksum) || \
152 	(((csum_flags) & M_CSUM_TCPv6) != 0 && tcp_do_loopback_cksum)))
153 
154 /*
155  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
156  * header (with pri, len, nxt, hlim, src, dst).
157  * This function may modify ver and hlim only.
158  * The mbuf chain containing the packet will be freed.
159  * The mbuf opt, if present, will not be freed.
160  *
161  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
162  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
163  * which is rt_rmx.rmx_mtu.
164  */
165 int
166 ip6_output(
167     struct mbuf *m0,
168     struct ip6_pktopts *opt,
169     struct route *ro,
170     int flags,
171     struct ip6_moptions *im6o,
172     struct socket *so,
173     struct ifnet **ifpp		/* XXX: just for statistics */
174 )
175 {
176 	struct ip6_hdr *ip6, *mhip6;
177 	struct ifnet *ifp, *origifp;
178 	struct mbuf *m = m0;
179 	int hlen, tlen, len, off;
180 	bool tso;
181 	struct route ip6route;
182 	struct rtentry *rt = NULL;
183 	const struct sockaddr_in6 *dst = NULL;
184 	struct sockaddr_in6 src_sa, dst_sa;
185 	int error = 0;
186 	struct in6_ifaddr *ia = NULL;
187 	u_long mtu;
188 	int alwaysfrag, dontfrag;
189 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
190 	struct ip6_exthdrs exthdrs;
191 	struct in6_addr finaldst, src0, dst0;
192 	u_int32_t zone;
193 	struct route *ro_pmtu = NULL;
194 	int hdrsplit = 0;
195 	int needipsec = 0;
196 #ifdef IPSEC
197 	int needipsectun = 0;
198 	struct secpolicy *sp = NULL;
199 
200 	ip6 = mtod(m, struct ip6_hdr *);
201 #endif /* IPSEC */
202 #ifdef FAST_IPSEC
203 	struct secpolicy *sp = NULL;
204 	int s;
205 #endif
206 
207 	memset(&ip6route, 0, sizeof(ip6route));
208 
209 #ifdef  DIAGNOSTIC
210 	if ((m->m_flags & M_PKTHDR) == 0)
211 		panic("ip6_output: no HDR");
212 
213 	if ((m->m_pkthdr.csum_flags &
214 	    (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) {
215 		panic("ip6_output: IPv4 checksum offload flags: %d",
216 		    m->m_pkthdr.csum_flags);
217 	}
218 
219 	if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) ==
220 	    (M_CSUM_TCPv6|M_CSUM_UDPv6)) {
221 		panic("ip6_output: conflicting checksum offload flags: %d",
222 		    m->m_pkthdr.csum_flags);
223 	}
224 #endif
225 
226 	M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr));
227 
228 #define MAKE_EXTHDR(hp, mp)						\
229     do {								\
230 	if (hp) {							\
231 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
232 		error = ip6_copyexthdr((mp), (void *)(hp), 		\
233 		    ((eh)->ip6e_len + 1) << 3);				\
234 		if (error)						\
235 			goto freehdrs;					\
236 	}								\
237     } while (/*CONSTCOND*/ 0)
238 
239 	bzero(&exthdrs, sizeof(exthdrs));
240 	if (opt) {
241 		/* Hop-by-Hop options header */
242 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
243 		/* Destination options header(1st part) */
244 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
245 		/* Routing header */
246 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
247 		/* Destination options header(2nd part) */
248 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
249 	}
250 
251 #ifdef IPSEC
252 	if ((flags & IPV6_FORWARDING) != 0) {
253 		needipsec = 0;
254 		goto skippolicycheck;
255 	}
256 
257 	/* get a security policy for this packet */
258 	if (so == NULL)
259 		sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
260 	else {
261 		if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp,
262 					 IPSEC_DIR_OUTBOUND)) {
263 			needipsec = 0;
264 			goto skippolicycheck;
265 		}
266 		sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
267 	}
268 
269 	if (sp == NULL) {
270 		ipsec6stat.out_inval++;
271 		goto freehdrs;
272 	}
273 
274 	error = 0;
275 
276 	/* check policy */
277 	switch (sp->policy) {
278 	case IPSEC_POLICY_DISCARD:
279 		/*
280 		 * This packet is just discarded.
281 		 */
282 		ipsec6stat.out_polvio++;
283 		goto freehdrs;
284 
285 	case IPSEC_POLICY_BYPASS:
286 	case IPSEC_POLICY_NONE:
287 		/* no need to do IPsec. */
288 		needipsec = 0;
289 		break;
290 
291 	case IPSEC_POLICY_IPSEC:
292 		if (sp->req == NULL) {
293 			/* XXX should be panic ? */
294 			printf("ip6_output: No IPsec request specified.\n");
295 			error = EINVAL;
296 			goto freehdrs;
297 		}
298 		needipsec = 1;
299 		break;
300 
301 	case IPSEC_POLICY_ENTRUST:
302 	default:
303 		printf("ip6_output: Invalid policy found. %d\n", sp->policy);
304 	}
305 
306   skippolicycheck:;
307 #endif /* IPSEC */
308 
309 	/*
310 	 * Calculate the total length of the extension header chain.
311 	 * Keep the length of the unfragmentable part for fragmentation.
312 	 */
313 	optlen = 0;
314 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
315 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
316 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
317 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
318 	/* NOTE: we don't add AH/ESP length here. do that later. */
319 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
320 
321 #ifdef FAST_IPSEC
322 	/* Check the security policy (SP) for the packet */
323 
324 	/* XXX For moment, we doesn't support packet with extented action */
325 	if (optlen !=0)
326 		goto freehdrs;
327 
328 	sp = ipsec6_check_policy(m,so,flags,&needipsec,&error);
329 	if (error != 0) {
330 		/*
331 		 * Hack: -EINVAL is used to signal that a packet
332 		 * should be silently discarded.  This is typically
333 		 * because we asked key management for an SA and
334 		 * it was delayed (e.g. kicked up to IKE).
335 		 */
336 	if (error == -EINVAL)
337 		error = 0;
338 	goto freehdrs;
339     }
340 #endif /* FAST_IPSEC */
341 
342 
343 	if (needipsec &&
344 	    (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
345 		in6_delayed_cksum(m);
346 		m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
347 	}
348 
349 
350 	/*
351 	 * If we need IPsec, or there is at least one extension header,
352 	 * separate IP6 header from the payload.
353 	 */
354 	if ((needipsec || optlen) && !hdrsplit) {
355 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
356 			m = NULL;
357 			goto freehdrs;
358 		}
359 		m = exthdrs.ip6e_ip6;
360 		hdrsplit++;
361 	}
362 
363 	/* adjust pointer */
364 	ip6 = mtod(m, struct ip6_hdr *);
365 
366 	/* adjust mbuf packet header length */
367 	m->m_pkthdr.len += optlen;
368 	plen = m->m_pkthdr.len - sizeof(*ip6);
369 
370 	/* If this is a jumbo payload, insert a jumbo payload option. */
371 	if (plen > IPV6_MAXPACKET) {
372 		if (!hdrsplit) {
373 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
374 				m = NULL;
375 				goto freehdrs;
376 			}
377 			m = exthdrs.ip6e_ip6;
378 			hdrsplit++;
379 		}
380 		/* adjust pointer */
381 		ip6 = mtod(m, struct ip6_hdr *);
382 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
383 			goto freehdrs;
384 		optlen += 8; /* XXX JUMBOOPTLEN */
385 		ip6->ip6_plen = 0;
386 	} else
387 		ip6->ip6_plen = htons(plen);
388 
389 	/*
390 	 * Concatenate headers and fill in next header fields.
391 	 * Here we have, on "m"
392 	 *	IPv6 payload
393 	 * and we insert headers accordingly.  Finally, we should be getting:
394 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
395 	 *
396 	 * during the header composing process, "m" points to IPv6 header.
397 	 * "mprev" points to an extension header prior to esp.
398 	 */
399 	{
400 		u_char *nexthdrp = &ip6->ip6_nxt;
401 		struct mbuf *mprev = m;
402 
403 		/*
404 		 * we treat dest2 specially.  this makes IPsec processing
405 		 * much easier.  the goal here is to make mprev point the
406 		 * mbuf prior to dest2.
407 		 *
408 		 * result: IPv6 dest2 payload
409 		 * m and mprev will point to IPv6 header.
410 		 */
411 		if (exthdrs.ip6e_dest2) {
412 			if (!hdrsplit)
413 				panic("assumption failed: hdr not split");
414 			exthdrs.ip6e_dest2->m_next = m->m_next;
415 			m->m_next = exthdrs.ip6e_dest2;
416 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
417 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
418 		}
419 
420 #define MAKE_CHAIN(m, mp, p, i)\
421     do {\
422 	if (m) {\
423 		if (!hdrsplit) \
424 			panic("assumption failed: hdr not split"); \
425 		*mtod((m), u_char *) = *(p);\
426 		*(p) = (i);\
427 		p = mtod((m), u_char *);\
428 		(m)->m_next = (mp)->m_next;\
429 		(mp)->m_next = (m);\
430 		(mp) = (m);\
431 	}\
432     } while (/*CONSTCOND*/ 0)
433 		/*
434 		 * result: IPv6 hbh dest1 rthdr dest2 payload
435 		 * m will point to IPv6 header.  mprev will point to the
436 		 * extension header prior to dest2 (rthdr in the above case).
437 		 */
438 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
439 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
440 		    IPPROTO_DSTOPTS);
441 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
442 		    IPPROTO_ROUTING);
443 
444 		M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data,
445 		    sizeof(struct ip6_hdr) + optlen);
446 
447 #ifdef IPSEC
448 		if (!needipsec)
449 			goto skip_ipsec2;
450 
451 		/*
452 		 * pointers after IPsec headers are not valid any more.
453 		 * other pointers need a great care too.
454 		 * (IPsec routines should not mangle mbufs prior to AH/ESP)
455 		 */
456 		exthdrs.ip6e_dest2 = NULL;
457 
458 	    {
459 		struct ip6_rthdr *rh = NULL;
460 		int segleft_org = 0;
461 		struct ipsec_output_state state;
462 
463 		if (exthdrs.ip6e_rthdr) {
464 			rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
465 			segleft_org = rh->ip6r_segleft;
466 			rh->ip6r_segleft = 0;
467 		}
468 
469 		bzero(&state, sizeof(state));
470 		state.m = m;
471 		error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
472 		    &needipsectun);
473 		m = state.m;
474 		if (error) {
475 			rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
476 			/* mbuf is already reclaimed in ipsec6_output_trans. */
477 			m = NULL;
478 			switch (error) {
479 			case EHOSTUNREACH:
480 			case ENETUNREACH:
481 			case EMSGSIZE:
482 			case ENOBUFS:
483 			case ENOMEM:
484 				break;
485 			default:
486 				printf("ip6_output (ipsec): error code %d\n", error);
487 				/* FALLTHROUGH */
488 			case ENOENT:
489 				/* don't show these error codes to the user */
490 				error = 0;
491 				break;
492 			}
493 			goto bad;
494 		}
495 		if (exthdrs.ip6e_rthdr) {
496 			/* ah6_output doesn't modify mbuf chain */
497 			rh->ip6r_segleft = segleft_org;
498 		}
499 	    }
500 skip_ipsec2:;
501 #endif
502 	}
503 
504 	/*
505 	 * If there is a routing header, replace destination address field
506 	 * with the first hop of the routing header.
507 	 */
508 	if (exthdrs.ip6e_rthdr) {
509 		struct ip6_rthdr *rh;
510 		struct ip6_rthdr0 *rh0;
511 		struct in6_addr *addr;
512 		struct sockaddr_in6 sa;
513 
514 		rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
515 		    struct ip6_rthdr *));
516 		finaldst = ip6->ip6_dst;
517 		switch (rh->ip6r_type) {
518 		case IPV6_RTHDR_TYPE_0:
519 			 rh0 = (struct ip6_rthdr0 *)rh;
520 			 addr = (struct in6_addr *)(rh0 + 1);
521 
522 			 /*
523 			  * construct a sockaddr_in6 form of
524 			  * the first hop.
525 			  *
526 			  * XXX: we may not have enough
527 			  * information about its scope zone;
528 			  * there is no standard API to pass
529 			  * the information from the
530 			  * application.
531 			  */
532 			 sockaddr_in6_init(&sa, addr, 0, 0, 0);
533 			 if ((error = sa6_embedscope(&sa,
534 			     ip6_use_defzone)) != 0) {
535 				 goto bad;
536 			 }
537 			 ip6->ip6_dst = sa.sin6_addr;
538 			 (void)memmove(&addr[0], &addr[1],
539 			     sizeof(struct in6_addr) *
540 			     (rh0->ip6r0_segleft - 1));
541 			 addr[rh0->ip6r0_segleft - 1] = finaldst;
542 			 /* XXX */
543 			 in6_clearscope(addr + rh0->ip6r0_segleft - 1);
544 			 break;
545 		default:	/* is it possible? */
546 			 error = EINVAL;
547 			 goto bad;
548 		}
549 	}
550 
551 	/* Source address validation */
552 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
553 	    (flags & IPV6_UNSPECSRC) == 0) {
554 		error = EOPNOTSUPP;
555 		ip6stat.ip6s_badscope++;
556 		goto bad;
557 	}
558 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
559 		error = EOPNOTSUPP;
560 		ip6stat.ip6s_badscope++;
561 		goto bad;
562 	}
563 
564 	ip6stat.ip6s_localout++;
565 
566 	/*
567 	 * Route packet.
568 	 */
569 	/* initialize cached route */
570 	if (ro == NULL) {
571 		ro = &ip6route;
572 	}
573 	ro_pmtu = ro;
574 	if (opt && opt->ip6po_rthdr)
575 		ro = &opt->ip6po_route;
576 
577  	/*
578 	 * if specified, try to fill in the traffic class field.
579 	 * do not override if a non-zero value is already set.
580 	 * we check the diffserv field and the ecn field separately.
581 	 */
582 	if (opt && opt->ip6po_tclass >= 0) {
583 		int mask = 0;
584 
585 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
586 			mask |= 0xfc;
587 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
588 			mask |= 0x03;
589 		if (mask != 0)
590 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
591 	}
592 
593 	/* fill in or override the hop limit field, if necessary. */
594 	if (opt && opt->ip6po_hlim != -1)
595 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
596 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
597 		if (im6o != NULL)
598 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
599 		else
600 			ip6->ip6_hlim = ip6_defmcasthlim;
601 	}
602 
603 #ifdef IPSEC
604 	if (needipsec && needipsectun) {
605 		struct ipsec_output_state state;
606 
607 		/*
608 		 * All the extension headers will become inaccessible
609 		 * (since they can be encrypted).
610 		 * Don't panic, we need no more updates to extension headers
611 		 * on inner IPv6 packet (since they are now encapsulated).
612 		 *
613 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
614 		 */
615 		bzero(&exthdrs, sizeof(exthdrs));
616 		exthdrs.ip6e_ip6 = m;
617 
618 		bzero(&state, sizeof(state));
619 		state.m = m;
620 		state.ro = ro;
621 		state.dst = rtcache_getdst(ro);
622 
623 		error = ipsec6_output_tunnel(&state, sp, flags);
624 
625 		m = state.m;
626 		ro_pmtu = ro = state.ro;
627 		dst = satocsin6(state.dst);
628 		if (error) {
629 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
630 			m0 = m = NULL;
631 			m = NULL;
632 			switch (error) {
633 			case EHOSTUNREACH:
634 			case ENETUNREACH:
635 			case EMSGSIZE:
636 			case ENOBUFS:
637 			case ENOMEM:
638 				break;
639 			default:
640 				printf("ip6_output (ipsec): error code %d\n", error);
641 				/* FALLTHROUGH */
642 			case ENOENT:
643 				/* don't show these error codes to the user */
644 				error = 0;
645 				break;
646 			}
647 			goto bad;
648 		}
649 
650 		exthdrs.ip6e_ip6 = m;
651 	}
652 #endif /* IPSEC */
653 #ifdef FAST_IPSEC
654 	if (needipsec) {
655 		s = splsoftnet();
656 		error = ipsec6_process_packet(m,sp->req);
657 
658 		/*
659 		 * Preserve KAME behaviour: ENOENT can be returned
660 		 * when an SA acquire is in progress.  Don't propagate
661 		 * this to user-level; it confuses applications.
662 		 * XXX this will go away when the SADB is redone.
663 		 */
664 		if (error == ENOENT)
665 			error = 0;
666 		splx(s);
667 		goto done;
668 	}
669 #endif /* FAST_IPSEC */
670 
671 
672 
673 	/* adjust pointer */
674 	ip6 = mtod(m, struct ip6_hdr *);
675 
676 	sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
677 	if ((error = in6_selectroute(&dst_sa, opt, im6o, ro,
678 	    &ifp, &rt, 0)) != 0) {
679 		if (ifp != NULL)
680 			in6_ifstat_inc(ifp, ifs6_out_discard);
681 		goto bad;
682 	}
683 	if (rt == NULL) {
684 		/*
685 		 * If in6_selectroute() does not return a route entry,
686 		 * dst may not have been updated.
687 		 */
688 		rtcache_setdst(ro, sin6tosa(&dst_sa));
689 	}
690 
691 	/*
692 	 * then rt (for unicast) and ifp must be non-NULL valid values.
693 	 */
694 	if ((flags & IPV6_FORWARDING) == 0) {
695 		/* XXX: the FORWARDING flag can be set for mrouting. */
696 		in6_ifstat_inc(ifp, ifs6_out_request);
697 	}
698 	if (rt != NULL) {
699 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
700 		rt->rt_use++;
701 	}
702 
703 	/*
704 	 * The outgoing interface must be in the zone of source and
705 	 * destination addresses.  We should use ia_ifp to support the
706 	 * case of sending packets to an address of our own.
707 	 */
708 	if (ia != NULL && ia->ia_ifp)
709 		origifp = ia->ia_ifp;
710 	else
711 		origifp = ifp;
712 
713 	src0 = ip6->ip6_src;
714 	if (in6_setscope(&src0, origifp, &zone))
715 		goto badscope;
716 	sockaddr_in6_init(&src_sa, &ip6->ip6_src, 0, 0, 0);
717 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
718 		goto badscope;
719 
720 	dst0 = ip6->ip6_dst;
721 	if (in6_setscope(&dst0, origifp, &zone))
722 		goto badscope;
723 	/* re-initialize to be sure */
724 	sockaddr_in6_init(&dst_sa, &ip6->ip6_dst, 0, 0, 0);
725 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id)
726 		goto badscope;
727 
728 	/* scope check is done. */
729 
730 	if (rt == NULL || IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
731 		if (dst == NULL)
732 			dst = satocsin6(rtcache_getdst(ro));
733 		KASSERT(dst != NULL);
734 	} else if (opt && rtcache_validate(&opt->ip6po_nextroute) != NULL) {
735 		/*
736 		 * The nexthop is explicitly specified by the
737 		 * application.  We assume the next hop is an IPv6
738 		 * address.
739 		 */
740 		dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
741 	} else if ((rt->rt_flags & RTF_GATEWAY))
742 		dst = (struct sockaddr_in6 *)rt->rt_gateway;
743 	else if (dst == NULL)
744 		dst = satocsin6(rtcache_getdst(ro));
745 
746 	/*
747 	 * XXXXXX: original code follows:
748 	 */
749 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
750 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
751 	else {
752 		struct	in6_multi *in6m;
753 
754 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
755 
756 		in6_ifstat_inc(ifp, ifs6_out_mcast);
757 
758 		/*
759 		 * Confirm that the outgoing interface supports multicast.
760 		 */
761 		if (!(ifp->if_flags & IFF_MULTICAST)) {
762 			ip6stat.ip6s_noroute++;
763 			in6_ifstat_inc(ifp, ifs6_out_discard);
764 			error = ENETUNREACH;
765 			goto bad;
766 		}
767 
768 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
769 		if (in6m != NULL &&
770 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
771 			/*
772 			 * If we belong to the destination multicast group
773 			 * on the outgoing interface, and the caller did not
774 			 * forbid loopback, loop back a copy.
775 			 */
776 			KASSERT(dst != NULL);
777 			ip6_mloopback(ifp, m, dst);
778 		} else {
779 			/*
780 			 * If we are acting as a multicast router, perform
781 			 * multicast forwarding as if the packet had just
782 			 * arrived on the interface to which we are about
783 			 * to send.  The multicast forwarding function
784 			 * recursively calls this function, using the
785 			 * IPV6_FORWARDING flag to prevent infinite recursion.
786 			 *
787 			 * Multicasts that are looped back by ip6_mloopback(),
788 			 * above, will be forwarded by the ip6_input() routine,
789 			 * if necessary.
790 			 */
791 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
792 				if (ip6_mforward(ip6, ifp, m) != 0) {
793 					m_freem(m);
794 					goto done;
795 				}
796 			}
797 		}
798 		/*
799 		 * Multicasts with a hoplimit of zero may be looped back,
800 		 * above, but must not be transmitted on a network.
801 		 * Also, multicasts addressed to the loopback interface
802 		 * are not sent -- the above call to ip6_mloopback() will
803 		 * loop back a copy if this host actually belongs to the
804 		 * destination group on the loopback interface.
805 		 */
806 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
807 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
808 			m_freem(m);
809 			goto done;
810 		}
811 	}
812 
813 	/*
814 	 * Fill the outgoing inteface to tell the upper layer
815 	 * to increment per-interface statistics.
816 	 */
817 	if (ifpp)
818 		*ifpp = ifp;
819 
820 	/* Determine path MTU. */
821 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
822 	    &alwaysfrag)) != 0)
823 		goto bad;
824 #ifdef IPSEC
825 	if (needipsectun)
826 		mtu = IPV6_MMTU;
827 #endif
828 
829 	/*
830 	 * The caller of this function may specify to use the minimum MTU
831 	 * in some cases.
832 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
833 	 * setting.  The logic is a bit complicated; by default, unicast
834 	 * packets will follow path MTU while multicast packets will be sent at
835 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
836 	 * including unicast ones will be sent at the minimum MTU.  Multicast
837 	 * packets will always be sent at the minimum MTU unless
838 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
839 	 * See RFC 3542 for more details.
840 	 */
841 	if (mtu > IPV6_MMTU) {
842 		if ((flags & IPV6_MINMTU))
843 			mtu = IPV6_MMTU;
844 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
845 			mtu = IPV6_MMTU;
846 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
847 			 (opt == NULL ||
848 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
849 			mtu = IPV6_MMTU;
850 		}
851 	}
852 
853 	/*
854 	 * clear embedded scope identifiers if necessary.
855 	 * in6_clearscope will touch the addresses only when necessary.
856 	 */
857 	in6_clearscope(&ip6->ip6_src);
858 	in6_clearscope(&ip6->ip6_dst);
859 
860 	/*
861 	 * If the outgoing packet contains a hop-by-hop options header,
862 	 * it must be examined and processed even by the source node.
863 	 * (RFC 2460, section 4.)
864 	 */
865 	if (exthdrs.ip6e_hbh) {
866 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
867 		u_int32_t dummy1; /* XXX unused */
868 		u_int32_t dummy2; /* XXX unused */
869 
870 		/*
871 		 *  XXX: if we have to send an ICMPv6 error to the sender,
872 		 *       we need the M_LOOP flag since icmp6_error() expects
873 		 *       the IPv6 and the hop-by-hop options header are
874 		 *       continuous unless the flag is set.
875 		 */
876 		m->m_flags |= M_LOOP;
877 		m->m_pkthdr.rcvif = ifp;
878 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
879 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
880 		    &dummy1, &dummy2) < 0) {
881 			/* m was already freed at this point */
882 			error = EINVAL;/* better error? */
883 			goto done;
884 		}
885 		m->m_flags &= ~M_LOOP; /* XXX */
886 		m->m_pkthdr.rcvif = NULL;
887 	}
888 
889 #ifdef PFIL_HOOKS
890 	/*
891 	 * Run through list of hooks for output packets.
892 	 */
893 	if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0)
894 		goto done;
895 	if (m == NULL)
896 		goto done;
897 	ip6 = mtod(m, struct ip6_hdr *);
898 #endif /* PFIL_HOOKS */
899 	/*
900 	 * Send the packet to the outgoing interface.
901 	 * If necessary, do IPv6 fragmentation before sending.
902 	 *
903 	 * the logic here is rather complex:
904 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
905 	 * 1-a:	send as is if tlen <= path mtu
906 	 * 1-b:	fragment if tlen > path mtu
907 	 *
908 	 * 2: if user asks us not to fragment (dontfrag == 1)
909 	 * 2-a:	send as is if tlen <= interface mtu
910 	 * 2-b:	error if tlen > interface mtu
911 	 *
912 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
913 	 *	always fragment
914 	 *
915 	 * 4: if dontfrag == 1 && alwaysfrag == 1
916 	 *	error, as we cannot handle this conflicting request
917 	 */
918 	tlen = m->m_pkthdr.len;
919 	tso = (m->m_pkthdr.csum_flags & M_CSUM_TSOv6) != 0;
920 	if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
921 		dontfrag = 1;
922 	else
923 		dontfrag = 0;
924 
925 	if (dontfrag && alwaysfrag) {	/* case 4 */
926 		/* conflicting request - can't transmit */
927 		error = EMSGSIZE;
928 		goto bad;
929 	}
930 	if (dontfrag && (!tso && tlen > IN6_LINKMTU(ifp))) {	/* case 2-b */
931 		/*
932 		 * Even if the DONTFRAG option is specified, we cannot send the
933 		 * packet when the data length is larger than the MTU of the
934 		 * outgoing interface.
935 		 * Notify the error by sending IPV6_PATHMTU ancillary data as
936 		 * well as returning an error code (the latter is not described
937 		 * in the API spec.)
938 		 */
939 		u_int32_t mtu32;
940 		struct ip6ctlparam ip6cp;
941 
942 		mtu32 = (u_int32_t)mtu;
943 		bzero(&ip6cp, sizeof(ip6cp));
944 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
945 		pfctlinput2(PRC_MSGSIZE,
946 		    rtcache_getdst(ro_pmtu), &ip6cp);
947 
948 		error = EMSGSIZE;
949 		goto bad;
950 	}
951 
952 	/*
953 	 * transmit packet without fragmentation
954 	 */
955 	if (dontfrag || (!alwaysfrag && (tlen <= mtu || tso))) {
956 		/* case 1-a and 2-a */
957 		struct in6_ifaddr *ia6;
958 		int sw_csum;
959 
960 		ip6 = mtod(m, struct ip6_hdr *);
961 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
962 		if (ia6) {
963 			/* Record statistics for this interface address. */
964 			ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len;
965 		}
966 #ifdef IPSEC
967 		/* clean ipsec history once it goes out of the node */
968 		ipsec_delaux(m);
969 #endif
970 
971 		sw_csum = m->m_pkthdr.csum_flags & ~ifp->if_csum_flags_tx;
972 		if ((sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
973 			if (IN6_NEED_CHECKSUM(ifp,
974 			    sw_csum & (M_CSUM_UDPv6|M_CSUM_TCPv6))) {
975 				in6_delayed_cksum(m);
976 			}
977 			m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
978 		}
979 
980 		KASSERT(dst != NULL);
981 		if (__predict_true(!tso ||
982 		    (ifp->if_capenable & IFCAP_TSOv6) != 0)) {
983 			error = nd6_output(ifp, origifp, m, dst, rt);
984 		} else {
985 			error = ip6_tso_output(ifp, origifp, m, dst, rt);
986 		}
987 		goto done;
988 	}
989 
990 	if (tso) {
991 		error = EINVAL; /* XXX */
992 		goto bad;
993 	}
994 
995 	/*
996 	 * try to fragment the packet.  case 1-b and 3
997 	 */
998 	if (mtu < IPV6_MMTU) {
999 		/* path MTU cannot be less than IPV6_MMTU */
1000 		error = EMSGSIZE;
1001 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1002 		goto bad;
1003 	} else if (ip6->ip6_plen == 0) {
1004 		/* jumbo payload cannot be fragmented */
1005 		error = EMSGSIZE;
1006 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1007 		goto bad;
1008 	} else {
1009 		struct mbuf **mnext, *m_frgpart;
1010 		struct ip6_frag *ip6f;
1011 		u_int32_t id = htonl(ip6_randomid());
1012 		u_char nextproto;
1013 #if 0				/* see below */
1014 		struct ip6ctlparam ip6cp;
1015 		u_int32_t mtu32;
1016 #endif
1017 
1018 		/*
1019 		 * Too large for the destination or interface;
1020 		 * fragment if possible.
1021 		 * Must be able to put at least 8 bytes per fragment.
1022 		 */
1023 		hlen = unfragpartlen;
1024 		if (mtu > IPV6_MAXPACKET)
1025 			mtu = IPV6_MAXPACKET;
1026 
1027 #if 0
1028 		/*
1029 		 * It is believed this code is a leftover from the
1030 		 * development of the IPV6_RECVPATHMTU sockopt and
1031 		 * associated work to implement RFC3542.
1032 		 * It's not entirely clear what the intent of the API
1033 		 * is at this point, so disable this code for now.
1034 		 * The IPV6_RECVPATHMTU sockopt and/or IPV6_DONTFRAG
1035 		 * will send notifications if the application requests.
1036 		 */
1037 
1038 		/* Notify a proper path MTU to applications. */
1039 		mtu32 = (u_int32_t)mtu;
1040 		bzero(&ip6cp, sizeof(ip6cp));
1041 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
1042 		pfctlinput2(PRC_MSGSIZE,
1043 		    rtcache_getdst(ro_pmtu), &ip6cp);
1044 #endif
1045 
1046 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1047 		if (len < 8) {
1048 			error = EMSGSIZE;
1049 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
1050 			goto bad;
1051 		}
1052 
1053 		mnext = &m->m_nextpkt;
1054 
1055 		/*
1056 		 * Change the next header field of the last header in the
1057 		 * unfragmentable part.
1058 		 */
1059 		if (exthdrs.ip6e_rthdr) {
1060 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1061 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1062 		} else if (exthdrs.ip6e_dest1) {
1063 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1064 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1065 		} else if (exthdrs.ip6e_hbh) {
1066 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1067 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1068 		} else {
1069 			nextproto = ip6->ip6_nxt;
1070 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1071 		}
1072 
1073 		if ((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6))
1074 		    != 0) {
1075 			if (IN6_NEED_CHECKSUM(ifp,
1076 			    m->m_pkthdr.csum_flags &
1077 			    (M_CSUM_UDPv6|M_CSUM_TCPv6))) {
1078 				in6_delayed_cksum(m);
1079 			}
1080 			m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
1081 		}
1082 
1083 		/*
1084 		 * Loop through length of segment after first fragment,
1085 		 * make new header and copy data of each part and link onto
1086 		 * chain.
1087 		 */
1088 		m0 = m;
1089 		for (off = hlen; off < tlen; off += len) {
1090 			struct mbuf *mlast;
1091 
1092 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
1093 			if (!m) {
1094 				error = ENOBUFS;
1095 				ip6stat.ip6s_odropped++;
1096 				goto sendorfree;
1097 			}
1098 			m->m_pkthdr.rcvif = NULL;
1099 			m->m_flags = m0->m_flags & M_COPYFLAGS;
1100 			*mnext = m;
1101 			mnext = &m->m_nextpkt;
1102 			m->m_data += max_linkhdr;
1103 			mhip6 = mtod(m, struct ip6_hdr *);
1104 			*mhip6 = *ip6;
1105 			m->m_len = sizeof(*mhip6);
1106 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1107 			if (error) {
1108 				ip6stat.ip6s_odropped++;
1109 				goto sendorfree;
1110 			}
1111 			ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7));
1112 			if (off + len >= tlen)
1113 				len = tlen - off;
1114 			else
1115 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1116 			mhip6->ip6_plen = htons((u_int16_t)(len + hlen +
1117 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1118 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1119 				error = ENOBUFS;
1120 				ip6stat.ip6s_odropped++;
1121 				goto sendorfree;
1122 			}
1123 			for (mlast = m; mlast->m_next; mlast = mlast->m_next)
1124 				;
1125 			mlast->m_next = m_frgpart;
1126 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1127 			m->m_pkthdr.rcvif = (struct ifnet *)0;
1128 			ip6f->ip6f_reserved = 0;
1129 			ip6f->ip6f_ident = id;
1130 			ip6f->ip6f_nxt = nextproto;
1131 			ip6stat.ip6s_ofragments++;
1132 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1133 		}
1134 
1135 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1136 	}
1137 
1138 	/*
1139 	 * Remove leading garbages.
1140 	 */
1141 sendorfree:
1142 	m = m0->m_nextpkt;
1143 	m0->m_nextpkt = 0;
1144 	m_freem(m0);
1145 	for (m0 = m; m; m = m0) {
1146 		m0 = m->m_nextpkt;
1147 		m->m_nextpkt = 0;
1148 		if (error == 0) {
1149 			struct in6_ifaddr *ia6;
1150 			ip6 = mtod(m, struct ip6_hdr *);
1151 			ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1152 			if (ia6) {
1153 				/*
1154 				 * Record statistics for this interface
1155 				 * address.
1156 				 */
1157 				ia6->ia_ifa.ifa_data.ifad_outbytes +=
1158 				    m->m_pkthdr.len;
1159 			}
1160 #ifdef IPSEC
1161 			/* clean ipsec history once it goes out of the node */
1162 			ipsec_delaux(m);
1163 #endif
1164 			KASSERT(dst != NULL);
1165 			error = nd6_output(ifp, origifp, m, dst, rt);
1166 		} else
1167 			m_freem(m);
1168 	}
1169 
1170 	if (error == 0)
1171 		ip6stat.ip6s_fragmented++;
1172 
1173 done:
1174 	rtcache_free(&ip6route);
1175 
1176 #ifdef IPSEC
1177 	if (sp != NULL)
1178 		key_freesp(sp);
1179 #endif /* IPSEC */
1180 #ifdef FAST_IPSEC
1181 	if (sp != NULL)
1182 		KEY_FREESP(&sp);
1183 #endif /* FAST_IPSEC */
1184 
1185 
1186 	return (error);
1187 
1188 freehdrs:
1189 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1190 	m_freem(exthdrs.ip6e_dest1);
1191 	m_freem(exthdrs.ip6e_rthdr);
1192 	m_freem(exthdrs.ip6e_dest2);
1193 	/* FALLTHROUGH */
1194 bad:
1195 	m_freem(m);
1196 	goto done;
1197 badscope:
1198 	ip6stat.ip6s_badscope++;
1199 	in6_ifstat_inc(origifp, ifs6_out_discard);
1200 	if (error == 0)
1201 		error = EHOSTUNREACH; /* XXX */
1202 	goto bad;
1203 }
1204 
1205 static int
1206 ip6_copyexthdr(struct mbuf **mp, void *hdr, int hlen)
1207 {
1208 	struct mbuf *m;
1209 
1210 	if (hlen > MCLBYTES)
1211 		return (ENOBUFS); /* XXX */
1212 
1213 	MGET(m, M_DONTWAIT, MT_DATA);
1214 	if (!m)
1215 		return (ENOBUFS);
1216 
1217 	if (hlen > MLEN) {
1218 		MCLGET(m, M_DONTWAIT);
1219 		if ((m->m_flags & M_EXT) == 0) {
1220 			m_free(m);
1221 			return (ENOBUFS);
1222 		}
1223 	}
1224 	m->m_len = hlen;
1225 	if (hdr)
1226 		bcopy(hdr, mtod(m, void *), hlen);
1227 
1228 	*mp = m;
1229 	return (0);
1230 }
1231 
1232 /*
1233  * Process a delayed payload checksum calculation.
1234  */
1235 void
1236 in6_delayed_cksum(struct mbuf *m)
1237 {
1238 	uint16_t csum, offset;
1239 
1240 	KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0);
1241 	KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0);
1242 	KASSERT((m->m_pkthdr.csum_flags
1243 	    & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0);
1244 
1245 	offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data);
1246 	csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset);
1247 	if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) {
1248 		csum = 0xffff;
1249 	}
1250 
1251 	offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data);
1252 	if ((offset + sizeof(csum)) > m->m_len) {
1253 		m_copyback(m, offset, sizeof(csum), &csum);
1254 	} else {
1255 		*(uint16_t *)(mtod(m, char *) + offset) = csum;
1256 	}
1257 }
1258 
1259 /*
1260  * Insert jumbo payload option.
1261  */
1262 static int
1263 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1264 {
1265 	struct mbuf *mopt;
1266 	u_int8_t *optbuf;
1267 	u_int32_t v;
1268 
1269 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1270 
1271 	/*
1272 	 * If there is no hop-by-hop options header, allocate new one.
1273 	 * If there is one but it doesn't have enough space to store the
1274 	 * jumbo payload option, allocate a cluster to store the whole options.
1275 	 * Otherwise, use it to store the options.
1276 	 */
1277 	if (exthdrs->ip6e_hbh == 0) {
1278 		MGET(mopt, M_DONTWAIT, MT_DATA);
1279 		if (mopt == 0)
1280 			return (ENOBUFS);
1281 		mopt->m_len = JUMBOOPTLEN;
1282 		optbuf = mtod(mopt, u_int8_t *);
1283 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1284 		exthdrs->ip6e_hbh = mopt;
1285 	} else {
1286 		struct ip6_hbh *hbh;
1287 
1288 		mopt = exthdrs->ip6e_hbh;
1289 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1290 			/*
1291 			 * XXX assumption:
1292 			 * - exthdrs->ip6e_hbh is not referenced from places
1293 			 *   other than exthdrs.
1294 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1295 			 */
1296 			int oldoptlen = mopt->m_len;
1297 			struct mbuf *n;
1298 
1299 			/*
1300 			 * XXX: give up if the whole (new) hbh header does
1301 			 * not fit even in an mbuf cluster.
1302 			 */
1303 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1304 				return (ENOBUFS);
1305 
1306 			/*
1307 			 * As a consequence, we must always prepare a cluster
1308 			 * at this point.
1309 			 */
1310 			MGET(n, M_DONTWAIT, MT_DATA);
1311 			if (n) {
1312 				MCLGET(n, M_DONTWAIT);
1313 				if ((n->m_flags & M_EXT) == 0) {
1314 					m_freem(n);
1315 					n = NULL;
1316 				}
1317 			}
1318 			if (!n)
1319 				return (ENOBUFS);
1320 			n->m_len = oldoptlen + JUMBOOPTLEN;
1321 			bcopy(mtod(mopt, void *), mtod(n, void *),
1322 			    oldoptlen);
1323 			optbuf = mtod(n, u_int8_t *) + oldoptlen;
1324 			m_freem(mopt);
1325 			mopt = exthdrs->ip6e_hbh = n;
1326 		} else {
1327 			optbuf = mtod(mopt, u_int8_t *) + mopt->m_len;
1328 			mopt->m_len += JUMBOOPTLEN;
1329 		}
1330 		optbuf[0] = IP6OPT_PADN;
1331 		optbuf[1] = 0;
1332 
1333 		/*
1334 		 * Adjust the header length according to the pad and
1335 		 * the jumbo payload option.
1336 		 */
1337 		hbh = mtod(mopt, struct ip6_hbh *);
1338 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1339 	}
1340 
1341 	/* fill in the option. */
1342 	optbuf[2] = IP6OPT_JUMBO;
1343 	optbuf[3] = 4;
1344 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1345 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1346 
1347 	/* finally, adjust the packet header length */
1348 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1349 
1350 	return (0);
1351 #undef JUMBOOPTLEN
1352 }
1353 
1354 /*
1355  * Insert fragment header and copy unfragmentable header portions.
1356  */
1357 static int
1358 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1359 	struct ip6_frag **frghdrp)
1360 {
1361 	struct mbuf *n, *mlast;
1362 
1363 	if (hlen > sizeof(struct ip6_hdr)) {
1364 		n = m_copym(m0, sizeof(struct ip6_hdr),
1365 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1366 		if (n == 0)
1367 			return (ENOBUFS);
1368 		m->m_next = n;
1369 	} else
1370 		n = m;
1371 
1372 	/* Search for the last mbuf of unfragmentable part. */
1373 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1374 		;
1375 
1376 	if ((mlast->m_flags & M_EXT) == 0 &&
1377 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1378 		/* use the trailing space of the last mbuf for the fragment hdr */
1379 		*frghdrp = (struct ip6_frag *)(mtod(mlast, char *) +
1380 		    mlast->m_len);
1381 		mlast->m_len += sizeof(struct ip6_frag);
1382 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1383 	} else {
1384 		/* allocate a new mbuf for the fragment header */
1385 		struct mbuf *mfrg;
1386 
1387 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1388 		if (mfrg == 0)
1389 			return (ENOBUFS);
1390 		mfrg->m_len = sizeof(struct ip6_frag);
1391 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1392 		mlast->m_next = mfrg;
1393 	}
1394 
1395 	return (0);
1396 }
1397 
1398 static int
1399 ip6_getpmtu(struct route *ro_pmtu, struct route *ro, struct ifnet *ifp,
1400     const struct in6_addr *dst, u_long *mtup, int *alwaysfragp)
1401 {
1402 	struct rtentry *rt;
1403 	u_int32_t mtu = 0;
1404 	int alwaysfrag = 0;
1405 	int error = 0;
1406 
1407 	if (ro_pmtu != ro) {
1408 		union {
1409 			struct sockaddr		dst;
1410 			struct sockaddr_in6	dst6;
1411 		} u;
1412 
1413 		/* The first hop and the final destination may differ. */
1414 		sockaddr_in6_init(&u.dst6, dst, 0, 0, 0);
1415 		rt = rtcache_lookup(ro_pmtu, &u.dst);
1416 	} else
1417 		rt = rtcache_validate(ro_pmtu);
1418 	if (rt != NULL) {
1419 		u_int32_t ifmtu;
1420 
1421 		if (ifp == NULL)
1422 			ifp = rt->rt_ifp;
1423 		ifmtu = IN6_LINKMTU(ifp);
1424 		mtu = rt->rt_rmx.rmx_mtu;
1425 		if (mtu == 0)
1426 			mtu = ifmtu;
1427 		else if (mtu < IPV6_MMTU) {
1428 			/*
1429 			 * RFC2460 section 5, last paragraph:
1430 			 * if we record ICMPv6 too big message with
1431 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1432 			 * or smaller, with fragment header attached.
1433 			 * (fragment header is needed regardless from the
1434 			 * packet size, for translators to identify packets)
1435 			 */
1436 			alwaysfrag = 1;
1437 			mtu = IPV6_MMTU;
1438 		} else if (mtu > ifmtu) {
1439 			/*
1440 			 * The MTU on the route is larger than the MTU on
1441 			 * the interface!  This shouldn't happen, unless the
1442 			 * MTU of the interface has been changed after the
1443 			 * interface was brought up.  Change the MTU in the
1444 			 * route to match the interface MTU (as long as the
1445 			 * field isn't locked).
1446 			 */
1447 			mtu = ifmtu;
1448 			if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
1449 				rt->rt_rmx.rmx_mtu = mtu;
1450 		}
1451 	} else if (ifp) {
1452 		mtu = IN6_LINKMTU(ifp);
1453 	} else
1454 		error = EHOSTUNREACH; /* XXX */
1455 
1456 	*mtup = mtu;
1457 	if (alwaysfragp)
1458 		*alwaysfragp = alwaysfrag;
1459 	return (error);
1460 }
1461 
1462 /*
1463  * IP6 socket option processing.
1464  */
1465 int
1466 ip6_ctloutput(int op, struct socket *so, int level, int optname,
1467     struct mbuf **mp)
1468 {
1469 	int privileged, optdatalen, uproto;
1470 	void *optdata;
1471 	struct in6pcb *in6p = sotoin6pcb(so);
1472 	struct mbuf *m = *mp;
1473 	int error, optval;
1474 	int optlen;
1475 	struct lwp *l = curlwp;	/* XXX */
1476 
1477 	optlen = m ? m->m_len : 0;
1478 	error = optval = 0;
1479 	privileged = (l == 0 || kauth_authorize_generic(l->l_cred,
1480 	    KAUTH_GENERIC_ISSUSER, NULL)) ? 0 : 1;
1481 	uproto = (int)so->so_proto->pr_protocol;
1482 
1483 	if (level != IPPROTO_IPV6) {
1484 		if (op == PRCO_SETOPT && *mp)
1485 			(void)m_free(*mp);
1486 		return ENOPROTOOPT;
1487 	}
1488 	switch (op) {
1489 	case PRCO_SETOPT:
1490 		switch (optname) {
1491 #ifdef RFC2292
1492 		case IPV6_2292PKTOPTIONS:
1493 			/* m is freed in ip6_pcbopts */
1494 			error = ip6_pcbopts(&in6p->in6p_outputopts,
1495 			    m, so);
1496 			break;
1497 #endif
1498 
1499 		/*
1500 		 * Use of some Hop-by-Hop options or some
1501 		 * Destination options, might require special
1502 		 * privilege.  That is, normal applications
1503 		 * (without special privilege) might be forbidden
1504 		 * from setting certain options in outgoing packets,
1505 		 * and might never see certain options in received
1506 		 * packets. [RFC 2292 Section 6]
1507 		 * KAME specific note:
1508 		 *  KAME prevents non-privileged users from sending or
1509 		 *  receiving ANY hbh/dst options in order to avoid
1510 		 *  overhead of parsing options in the kernel.
1511 		 */
1512 		case IPV6_RECVHOPOPTS:
1513 		case IPV6_RECVDSTOPTS:
1514 		case IPV6_RECVRTHDRDSTOPTS:
1515 			if (!privileged) {
1516 				error = EPERM;
1517 				break;
1518 			}
1519 			/* FALLTHROUGH */
1520 		case IPV6_UNICAST_HOPS:
1521 		case IPV6_HOPLIMIT:
1522 		case IPV6_FAITH:
1523 
1524 		case IPV6_RECVPKTINFO:
1525 		case IPV6_RECVHOPLIMIT:
1526 		case IPV6_RECVRTHDR:
1527 		case IPV6_RECVPATHMTU:
1528 		case IPV6_RECVTCLASS:
1529 		case IPV6_V6ONLY:
1530 			if (optlen != sizeof(int)) {
1531 				error = EINVAL;
1532 				break;
1533 			}
1534 			optval = *mtod(m, int *);
1535 			switch (optname) {
1536 
1537 			case IPV6_UNICAST_HOPS:
1538 				if (optval < -1 || optval >= 256)
1539 					error = EINVAL;
1540 				else {
1541 					/* -1 = kernel default */
1542 					in6p->in6p_hops = optval;
1543 				}
1544 				break;
1545 #define OPTSET(bit) \
1546 do { \
1547 if (optval) \
1548 	in6p->in6p_flags |= (bit); \
1549 else \
1550 	in6p->in6p_flags &= ~(bit); \
1551 } while (/*CONSTCOND*/ 0)
1552 
1553 #ifdef RFC2292
1554 #define OPTSET2292(bit) 			\
1555 do { 						\
1556 in6p->in6p_flags |= IN6P_RFC2292; 	\
1557 if (optval) 				\
1558 	in6p->in6p_flags |= (bit); 	\
1559 else 					\
1560 	in6p->in6p_flags &= ~(bit); 	\
1561 } while (/*CONSTCOND*/ 0)
1562 #endif
1563 
1564 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1565 
1566 			case IPV6_RECVPKTINFO:
1567 #ifdef RFC2292
1568 				/* cannot mix with RFC2292 */
1569 				if (OPTBIT(IN6P_RFC2292)) {
1570 					error = EINVAL;
1571 					break;
1572 				}
1573 #endif
1574 				OPTSET(IN6P_PKTINFO);
1575 				break;
1576 
1577 			case IPV6_HOPLIMIT:
1578 			{
1579 				struct ip6_pktopts **optp;
1580 
1581 #ifdef RFC2292
1582 				/* cannot mix with RFC2292 */
1583 				if (OPTBIT(IN6P_RFC2292)) {
1584 					error = EINVAL;
1585 					break;
1586 				}
1587 #endif
1588 				optp = &in6p->in6p_outputopts;
1589 				error = ip6_pcbopt(IPV6_HOPLIMIT,
1590 						   (u_char *)&optval,
1591 						   sizeof(optval),
1592 						   optp,
1593 						   privileged, uproto);
1594 				break;
1595 			}
1596 
1597 			case IPV6_RECVHOPLIMIT:
1598 #ifdef RFC2292
1599 				/* cannot mix with RFC2292 */
1600 				if (OPTBIT(IN6P_RFC2292)) {
1601 					error = EINVAL;
1602 					break;
1603 				}
1604 #endif
1605 				OPTSET(IN6P_HOPLIMIT);
1606 				break;
1607 
1608 			case IPV6_RECVHOPOPTS:
1609 #ifdef RFC2292
1610 				/* cannot mix with RFC2292 */
1611 				if (OPTBIT(IN6P_RFC2292)) {
1612 					error = EINVAL;
1613 					break;
1614 				}
1615 #endif
1616 				OPTSET(IN6P_HOPOPTS);
1617 				break;
1618 
1619 			case IPV6_RECVDSTOPTS:
1620 #ifdef RFC2292
1621 				/* cannot mix with RFC2292 */
1622 				if (OPTBIT(IN6P_RFC2292)) {
1623 					error = EINVAL;
1624 					break;
1625 				}
1626 #endif
1627 				OPTSET(IN6P_DSTOPTS);
1628 				break;
1629 
1630 			case IPV6_RECVRTHDRDSTOPTS:
1631 #ifdef RFC2292
1632 				/* cannot mix with RFC2292 */
1633 				if (OPTBIT(IN6P_RFC2292)) {
1634 					error = EINVAL;
1635 					break;
1636 				}
1637 #endif
1638 				OPTSET(IN6P_RTHDRDSTOPTS);
1639 				break;
1640 
1641 			case IPV6_RECVRTHDR:
1642 #ifdef RFC2292
1643 				/* cannot mix with RFC2292 */
1644 				if (OPTBIT(IN6P_RFC2292)) {
1645 					error = EINVAL;
1646 					break;
1647 				}
1648 #endif
1649 				OPTSET(IN6P_RTHDR);
1650 				break;
1651 
1652 			case IPV6_FAITH:
1653 				OPTSET(IN6P_FAITH);
1654 				break;
1655 
1656 			case IPV6_RECVPATHMTU:
1657 				/*
1658 				 * We ignore this option for TCP
1659 				 * sockets.
1660 				 * (RFC3542 leaves this case
1661 				 * unspecified.)
1662 				 */
1663 				if (uproto != IPPROTO_TCP)
1664 					OPTSET(IN6P_MTU);
1665 				break;
1666 
1667 			case IPV6_V6ONLY:
1668 				/*
1669 				 * make setsockopt(IPV6_V6ONLY)
1670 				 * available only prior to bind(2).
1671 				 * see ipng mailing list, Jun 22 2001.
1672 				 */
1673 				if (in6p->in6p_lport ||
1674 				    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1675 					error = EINVAL;
1676 					break;
1677 				}
1678 #ifdef INET6_BINDV6ONLY
1679 				if (!optval)
1680 					error = EINVAL;
1681 #else
1682 				OPTSET(IN6P_IPV6_V6ONLY);
1683 #endif
1684 				break;
1685 			case IPV6_RECVTCLASS:
1686 #ifdef RFC2292
1687 				/* cannot mix with RFC2292 XXX */
1688 				if (OPTBIT(IN6P_RFC2292)) {
1689 					error = EINVAL;
1690 					break;
1691 				}
1692 #endif
1693 				OPTSET(IN6P_TCLASS);
1694 				break;
1695 
1696 			}
1697 			break;
1698 
1699 		case IPV6_OTCLASS:
1700 		{
1701 			struct ip6_pktopts **optp;
1702 			u_int8_t tclass;
1703 
1704 			if (optlen != sizeof(tclass)) {
1705 				error = EINVAL;
1706 				break;
1707 			}
1708 			tclass = *mtod(m, u_int8_t *);
1709 			optp = &in6p->in6p_outputopts;
1710 			error = ip6_pcbopt(optname,
1711 					   (u_char *)&tclass,
1712 					   sizeof(tclass),
1713 					   optp,
1714 					   privileged, uproto);
1715 			break;
1716 		}
1717 
1718 		case IPV6_TCLASS:
1719 		case IPV6_DONTFRAG:
1720 		case IPV6_USE_MIN_MTU:
1721 			if (optlen != sizeof(optval)) {
1722 				error = EINVAL;
1723 				break;
1724 			}
1725 			optval = *mtod(m, int *);
1726 			{
1727 				struct ip6_pktopts **optp;
1728 				optp = &in6p->in6p_outputopts;
1729 				error = ip6_pcbopt(optname,
1730 						   (u_char *)&optval,
1731 						   sizeof(optval),
1732 						   optp,
1733 						   privileged, uproto);
1734 				break;
1735 			}
1736 
1737 #ifdef RFC2292
1738 		case IPV6_2292PKTINFO:
1739 		case IPV6_2292HOPLIMIT:
1740 		case IPV6_2292HOPOPTS:
1741 		case IPV6_2292DSTOPTS:
1742 		case IPV6_2292RTHDR:
1743 			/* RFC 2292 */
1744 			if (optlen != sizeof(int)) {
1745 				error = EINVAL;
1746 				break;
1747 			}
1748 			optval = *mtod(m, int *);
1749 			switch (optname) {
1750 			case IPV6_2292PKTINFO:
1751 				OPTSET2292(IN6P_PKTINFO);
1752 				break;
1753 			case IPV6_2292HOPLIMIT:
1754 				OPTSET2292(IN6P_HOPLIMIT);
1755 				break;
1756 			case IPV6_2292HOPOPTS:
1757 				/*
1758 				 * Check super-user privilege.
1759 				 * See comments for IPV6_RECVHOPOPTS.
1760 				 */
1761 				if (!privileged)
1762 					return (EPERM);
1763 				OPTSET2292(IN6P_HOPOPTS);
1764 				break;
1765 			case IPV6_2292DSTOPTS:
1766 				if (!privileged)
1767 					return (EPERM);
1768 				OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1769 				break;
1770 			case IPV6_2292RTHDR:
1771 				OPTSET2292(IN6P_RTHDR);
1772 				break;
1773 			}
1774 			break;
1775 #endif
1776 		case IPV6_PKTINFO:
1777 		case IPV6_HOPOPTS:
1778 		case IPV6_RTHDR:
1779 		case IPV6_DSTOPTS:
1780 		case IPV6_RTHDRDSTOPTS:
1781 		case IPV6_NEXTHOP:
1782 		{
1783 			/* new advanced API (RFC3542) */
1784 			u_char *optbuf;
1785 			int optbuflen;
1786 			struct ip6_pktopts **optp;
1787 			if (!m) {
1788 				error = EINVAL;
1789 				break;
1790 			}
1791 
1792 #ifdef RFC2292
1793 			/* cannot mix with RFC2292 */
1794 			if (OPTBIT(IN6P_RFC2292)) {
1795 				error = EINVAL;
1796 				break;
1797 			}
1798 #endif
1799 
1800 			if (m && m->m_next) {
1801 				error = EINVAL;	/* XXX */
1802 				break;
1803 			}
1804 
1805 			optbuf = mtod(m, u_char *);
1806 			optbuflen = m->m_len;
1807 			optp = &in6p->in6p_outputopts;
1808 			error = ip6_pcbopt(optname, optbuf, optbuflen,
1809 			    optp, privileged, uproto);
1810 			break;
1811 		}
1812 #undef OPTSET
1813 
1814 		case IPV6_MULTICAST_IF:
1815 		case IPV6_MULTICAST_HOPS:
1816 		case IPV6_MULTICAST_LOOP:
1817 		case IPV6_JOIN_GROUP:
1818 		case IPV6_LEAVE_GROUP:
1819 			error = ip6_setmoptions(optname,
1820 			    &in6p->in6p_moptions, m);
1821 			break;
1822 
1823 		case IPV6_PORTRANGE:
1824 			if (!m) {
1825 				error = EINVAL;
1826 				break;
1827 			}
1828 			optval = *mtod(m, int *);
1829 
1830 			switch (optval) {
1831 			case IPV6_PORTRANGE_DEFAULT:
1832 				in6p->in6p_flags &= ~(IN6P_LOWPORT);
1833 				in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1834 				break;
1835 
1836 			case IPV6_PORTRANGE_HIGH:
1837 				in6p->in6p_flags &= ~(IN6P_LOWPORT);
1838 				in6p->in6p_flags |= IN6P_HIGHPORT;
1839 				break;
1840 
1841 			case IPV6_PORTRANGE_LOW:
1842 				in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1843 				in6p->in6p_flags |= IN6P_LOWPORT;
1844 				break;
1845 
1846 			default:
1847 				error = EINVAL;
1848 				break;
1849 			}
1850 			break;
1851 
1852 
1853 #if defined(IPSEC) || defined(FAST_IPSEC)
1854 		case IPV6_IPSEC_POLICY:
1855 		{
1856 			void *req = NULL;
1857 			size_t len = 0;
1858 			if (m) {
1859 				req = mtod(m, void *);
1860 				len = m->m_len;
1861 			}
1862 			error = ipsec6_set_policy(in6p, optname, req,
1863 						  len, privileged);
1864 		}
1865 			break;
1866 #endif /* IPSEC */
1867 
1868 		default:
1869 			error = ENOPROTOOPT;
1870 			break;
1871 		}
1872 		if (m)
1873 			(void)m_free(m);
1874 		break;
1875 
1876 	case PRCO_GETOPT:
1877 		switch (optname) {
1878 #ifdef RFC2292
1879 		case IPV6_2292PKTOPTIONS:
1880 			/*
1881 			 * RFC3542 (effectively) deprecated the
1882 			 * semantics of the 2292-style pktoptions.
1883 			 * Since it was not reliable in nature (i.e.,
1884 			 * applications had to expect the lack of some
1885 			 * information after all), it would make sense
1886 			 * to simplify this part by always returning
1887 			 * empty data.
1888 			 */
1889 			*mp = m_get(M_WAIT, MT_SOOPTS);
1890 			(*mp)->m_len = 0;
1891 			break;
1892 #endif
1893 
1894 		case IPV6_RECVHOPOPTS:
1895 		case IPV6_RECVDSTOPTS:
1896 		case IPV6_RECVRTHDRDSTOPTS:
1897 		case IPV6_UNICAST_HOPS:
1898 		case IPV6_RECVPKTINFO:
1899 		case IPV6_RECVHOPLIMIT:
1900 		case IPV6_RECVRTHDR:
1901 		case IPV6_RECVPATHMTU:
1902 
1903 		case IPV6_FAITH:
1904 		case IPV6_V6ONLY:
1905 		case IPV6_PORTRANGE:
1906 		case IPV6_RECVTCLASS:
1907 			switch (optname) {
1908 
1909 			case IPV6_RECVHOPOPTS:
1910 				optval = OPTBIT(IN6P_HOPOPTS);
1911 				break;
1912 
1913 			case IPV6_RECVDSTOPTS:
1914 				optval = OPTBIT(IN6P_DSTOPTS);
1915 				break;
1916 
1917 			case IPV6_RECVRTHDRDSTOPTS:
1918 				optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1919 				break;
1920 
1921 			case IPV6_UNICAST_HOPS:
1922 				optval = in6p->in6p_hops;
1923 				break;
1924 
1925 			case IPV6_RECVPKTINFO:
1926 				optval = OPTBIT(IN6P_PKTINFO);
1927 				break;
1928 
1929 			case IPV6_RECVHOPLIMIT:
1930 				optval = OPTBIT(IN6P_HOPLIMIT);
1931 				break;
1932 
1933 			case IPV6_RECVRTHDR:
1934 				optval = OPTBIT(IN6P_RTHDR);
1935 				break;
1936 
1937 			case IPV6_RECVPATHMTU:
1938 				optval = OPTBIT(IN6P_MTU);
1939 				break;
1940 
1941 			case IPV6_FAITH:
1942 				optval = OPTBIT(IN6P_FAITH);
1943 				break;
1944 
1945 			case IPV6_V6ONLY:
1946 				optval = OPTBIT(IN6P_IPV6_V6ONLY);
1947 				break;
1948 
1949 			case IPV6_PORTRANGE:
1950 			    {
1951 				int flags;
1952 				flags = in6p->in6p_flags;
1953 				if (flags & IN6P_HIGHPORT)
1954 					optval = IPV6_PORTRANGE_HIGH;
1955 				else if (flags & IN6P_LOWPORT)
1956 					optval = IPV6_PORTRANGE_LOW;
1957 				else
1958 					optval = 0;
1959 				break;
1960 			    }
1961 			case IPV6_RECVTCLASS:
1962 				optval = OPTBIT(IN6P_TCLASS);
1963 				break;
1964 
1965 			}
1966 			if (error)
1967 				break;
1968 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1969 			m->m_len = sizeof(int);
1970 			*mtod(m, int *) = optval;
1971 			break;
1972 
1973 		case IPV6_PATHMTU:
1974 		    {
1975 			u_long pmtu = 0;
1976 			struct ip6_mtuinfo mtuinfo;
1977 			struct route *ro = &in6p->in6p_route;
1978 
1979 			if (!(so->so_state & SS_ISCONNECTED))
1980 				return (ENOTCONN);
1981 			/*
1982 			 * XXX: we dot not consider the case of source
1983 			 * routing, or optional information to specify
1984 			 * the outgoing interface.
1985 			 */
1986 			error = ip6_getpmtu(ro, NULL, NULL,
1987 			    &in6p->in6p_faddr, &pmtu, NULL);
1988 			if (error)
1989 				break;
1990 			if (pmtu > IPV6_MAXPACKET)
1991 				pmtu = IPV6_MAXPACKET;
1992 
1993 			memset(&mtuinfo, 0, sizeof(mtuinfo));
1994 			mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1995 			optdata = (void *)&mtuinfo;
1996 			optdatalen = sizeof(mtuinfo);
1997 			if (optdatalen > MCLBYTES)
1998 				return (EMSGSIZE); /* XXX */
1999 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
2000 			if (optdatalen > MLEN)
2001 				MCLGET(m, M_WAIT);
2002 			m->m_len = optdatalen;
2003 			memcpy(mtod(m, void *), optdata, optdatalen);
2004 			break;
2005 		    }
2006 
2007 #ifdef RFC2292
2008 		case IPV6_2292PKTINFO:
2009 		case IPV6_2292HOPLIMIT:
2010 		case IPV6_2292HOPOPTS:
2011 		case IPV6_2292RTHDR:
2012 		case IPV6_2292DSTOPTS:
2013 			switch (optname) {
2014 			case IPV6_2292PKTINFO:
2015 				optval = OPTBIT(IN6P_PKTINFO);
2016 				break;
2017 			case IPV6_2292HOPLIMIT:
2018 				optval = OPTBIT(IN6P_HOPLIMIT);
2019 				break;
2020 			case IPV6_2292HOPOPTS:
2021 				optval = OPTBIT(IN6P_HOPOPTS);
2022 				break;
2023 			case IPV6_2292RTHDR:
2024 				optval = OPTBIT(IN6P_RTHDR);
2025 				break;
2026 			case IPV6_2292DSTOPTS:
2027 				optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2028 				break;
2029 			}
2030 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
2031 			m->m_len = sizeof(int);
2032 			*mtod(m, int *) = optval;
2033 			break;
2034 #endif
2035 		case IPV6_PKTINFO:
2036 		case IPV6_HOPOPTS:
2037 		case IPV6_RTHDR:
2038 		case IPV6_DSTOPTS:
2039 		case IPV6_RTHDRDSTOPTS:
2040 		case IPV6_NEXTHOP:
2041 		case IPV6_OTCLASS:
2042 		case IPV6_TCLASS:
2043 		case IPV6_DONTFRAG:
2044 		case IPV6_USE_MIN_MTU:
2045 			error = ip6_getpcbopt(in6p->in6p_outputopts,
2046 			    optname, mp);
2047 			break;
2048 
2049 		case IPV6_MULTICAST_IF:
2050 		case IPV6_MULTICAST_HOPS:
2051 		case IPV6_MULTICAST_LOOP:
2052 		case IPV6_JOIN_GROUP:
2053 		case IPV6_LEAVE_GROUP:
2054 			error = ip6_getmoptions(optname,
2055 			    in6p->in6p_moptions, mp);
2056 			break;
2057 
2058 #if defined(IPSEC) || defined(FAST_IPSEC)
2059 		case IPV6_IPSEC_POLICY:
2060 		    {
2061 			void *req = NULL;
2062 			size_t len = 0;
2063 			if (m) {
2064 				req = mtod(m, void *);
2065 				len = m->m_len;
2066 			}
2067 			error = ipsec6_get_policy(in6p, req, len, mp);
2068 			break;
2069 		    }
2070 #endif /* IPSEC */
2071 
2072 
2073 
2074 
2075 		default:
2076 			error = ENOPROTOOPT;
2077 			break;
2078 		}
2079 		break;
2080 	}
2081 	return (error);
2082 }
2083 
2084 int
2085 ip6_raw_ctloutput(int op, struct socket *so, int level, int optname,
2086 	struct mbuf **mp)
2087 {
2088 	int error = 0, optval, optlen;
2089 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2090 	struct in6pcb *in6p = sotoin6pcb(so);
2091 	struct mbuf *m = *mp;
2092 
2093 	optlen = m ? m->m_len : 0;
2094 
2095 	if (level != IPPROTO_IPV6) {
2096 		if (op == PRCO_SETOPT && *mp)
2097 			(void)m_free(*mp);
2098 		return ENOPROTOOPT;
2099 	}
2100 
2101 	switch (optname) {
2102 	case IPV6_CHECKSUM:
2103 		/*
2104 		 * For ICMPv6 sockets, no modification allowed for checksum
2105 		 * offset, permit "no change" values to help existing apps.
2106 		 *
2107 		 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM
2108 		 * for an ICMPv6 socket will fail."  The current
2109 		 * behavior does not meet RFC3542.
2110 		 */
2111 		switch (op) {
2112 		case PRCO_SETOPT:
2113 			if (optlen != sizeof(int)) {
2114 				error = EINVAL;
2115 				break;
2116 			}
2117 			optval = *mtod(m, int *);
2118 			if ((optval % 2) != 0) {
2119 				/* the API assumes even offset values */
2120 				error = EINVAL;
2121 			} else if (so->so_proto->pr_protocol ==
2122 			    IPPROTO_ICMPV6) {
2123 				if (optval != icmp6off)
2124 					error = EINVAL;
2125 			} else
2126 				in6p->in6p_cksum = optval;
2127 			break;
2128 
2129 		case PRCO_GETOPT:
2130 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2131 				optval = icmp6off;
2132 			else
2133 				optval = in6p->in6p_cksum;
2134 
2135 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
2136 			m->m_len = sizeof(int);
2137 			*mtod(m, int *) = optval;
2138 			break;
2139 
2140 		default:
2141 			error = EINVAL;
2142 			break;
2143 		}
2144 		break;
2145 
2146 	default:
2147 		error = ENOPROTOOPT;
2148 		break;
2149 	}
2150 
2151 	if (op == PRCO_SETOPT && m)
2152 		(void)m_free(m);
2153 
2154 	return (error);
2155 }
2156 
2157 #ifdef RFC2292
2158 /*
2159  * Set up IP6 options in pcb for insertion in output packets or
2160  * specifying behavior of outgoing packets.
2161  */
2162 static int
2163 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, struct socket *so)
2164 {
2165 	struct ip6_pktopts *opt = *pktopt;
2166 	int error = 0;
2167 	struct lwp *l = curlwp;	/* XXX */
2168 	int priv = 0;
2169 
2170 	/* turn off any old options. */
2171 	if (opt) {
2172 #ifdef DIAGNOSTIC
2173 	    if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2174 		opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2175 		opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2176 		    printf("ip6_pcbopts: all specified options are cleared.\n");
2177 #endif
2178 		ip6_clearpktopts(opt, -1);
2179 	} else
2180 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2181 	*pktopt = NULL;
2182 
2183 	if (!m || m->m_len == 0) {
2184 		/*
2185 		 * Only turning off any previous options, regardless of
2186 		 * whether the opt is just created or given.
2187 		 */
2188 		free(opt, M_IP6OPT);
2189 		return (0);
2190 	}
2191 
2192 	/*  set options specified by user. */
2193 	if (l && !kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER,
2194 	    NULL))
2195 		priv = 1;
2196 	if ((error = ip6_setpktopts(m, opt, NULL, priv,
2197 	    so->so_proto->pr_protocol)) != 0) {
2198 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2199 		free(opt, M_IP6OPT);
2200 		return (error);
2201 	}
2202 	*pktopt = opt;
2203 	return (0);
2204 }
2205 #endif
2206 
2207 /*
2208  * initialize ip6_pktopts.  beware that there are non-zero default values in
2209  * the struct.
2210  */
2211 void
2212 ip6_initpktopts(struct ip6_pktopts *opt)
2213 {
2214 
2215 	memset(opt, 0, sizeof(*opt));
2216 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2217 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2218 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2219 }
2220 
2221 #define sin6tosa(sin6)	((struct sockaddr *)(sin6)) /* XXX */
2222 static int
2223 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2224     int priv, int uproto)
2225 {
2226 	struct ip6_pktopts *opt;
2227 
2228 	if (*pktopt == NULL) {
2229 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2230 		    M_WAITOK);
2231 		ip6_initpktopts(*pktopt);
2232 	}
2233 	opt = *pktopt;
2234 
2235 	return (ip6_setpktopt(optname, buf, len, opt, priv, 1, 0, uproto));
2236 }
2237 
2238 static int
2239 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct mbuf **mp)
2240 {
2241 	void *optdata = NULL;
2242 	int optdatalen = 0;
2243 	struct ip6_ext *ip6e;
2244 	int error = 0;
2245 	struct in6_pktinfo null_pktinfo;
2246 	int deftclass = 0, on;
2247 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2248 	struct mbuf *m;
2249 
2250 	switch (optname) {
2251 	case IPV6_PKTINFO:
2252 		if (pktopt && pktopt->ip6po_pktinfo)
2253 			optdata = (void *)pktopt->ip6po_pktinfo;
2254 		else {
2255 			/* XXX: we don't have to do this every time... */
2256 			memset(&null_pktinfo, 0, sizeof(null_pktinfo));
2257 			optdata = (void *)&null_pktinfo;
2258 		}
2259 		optdatalen = sizeof(struct in6_pktinfo);
2260 		break;
2261 	case IPV6_OTCLASS:
2262 		/* XXX */
2263 		return (EINVAL);
2264 	case IPV6_TCLASS:
2265 		if (pktopt && pktopt->ip6po_tclass >= 0)
2266 			optdata = (void *)&pktopt->ip6po_tclass;
2267 		else
2268 			optdata = (void *)&deftclass;
2269 		optdatalen = sizeof(int);
2270 		break;
2271 	case IPV6_HOPOPTS:
2272 		if (pktopt && pktopt->ip6po_hbh) {
2273 			optdata = (void *)pktopt->ip6po_hbh;
2274 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2275 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2276 		}
2277 		break;
2278 	case IPV6_RTHDR:
2279 		if (pktopt && pktopt->ip6po_rthdr) {
2280 			optdata = (void *)pktopt->ip6po_rthdr;
2281 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2282 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2283 		}
2284 		break;
2285 	case IPV6_RTHDRDSTOPTS:
2286 		if (pktopt && pktopt->ip6po_dest1) {
2287 			optdata = (void *)pktopt->ip6po_dest1;
2288 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2289 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2290 		}
2291 		break;
2292 	case IPV6_DSTOPTS:
2293 		if (pktopt && pktopt->ip6po_dest2) {
2294 			optdata = (void *)pktopt->ip6po_dest2;
2295 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2296 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2297 		}
2298 		break;
2299 	case IPV6_NEXTHOP:
2300 		if (pktopt && pktopt->ip6po_nexthop) {
2301 			optdata = (void *)pktopt->ip6po_nexthop;
2302 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2303 		}
2304 		break;
2305 	case IPV6_USE_MIN_MTU:
2306 		if (pktopt)
2307 			optdata = (void *)&pktopt->ip6po_minmtu;
2308 		else
2309 			optdata = (void *)&defminmtu;
2310 		optdatalen = sizeof(int);
2311 		break;
2312 	case IPV6_DONTFRAG:
2313 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2314 			on = 1;
2315 		else
2316 			on = 0;
2317 		optdata = (void *)&on;
2318 		optdatalen = sizeof(on);
2319 		break;
2320 	default:		/* should not happen */
2321 #ifdef DIAGNOSTIC
2322 		panic("ip6_getpcbopt: unexpected option\n");
2323 #endif
2324 		return (ENOPROTOOPT);
2325 	}
2326 
2327 	if (optdatalen > MCLBYTES)
2328 		return (EMSGSIZE); /* XXX */
2329 	*mp = m = m_get(M_WAIT, MT_SOOPTS);
2330 	if (optdatalen > MLEN)
2331 		MCLGET(m, M_WAIT);
2332 	m->m_len = optdatalen;
2333 	if (optdatalen)
2334 		memcpy(mtod(m, void *), optdata, optdatalen);
2335 
2336 	return (error);
2337 }
2338 
2339 void
2340 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2341 {
2342 	if (optname == -1 || optname == IPV6_PKTINFO) {
2343 		if (pktopt->ip6po_pktinfo)
2344 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2345 		pktopt->ip6po_pktinfo = NULL;
2346 	}
2347 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2348 		pktopt->ip6po_hlim = -1;
2349 	if (optname == -1 || optname == IPV6_TCLASS)
2350 		pktopt->ip6po_tclass = -1;
2351 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2352 		rtcache_free(&pktopt->ip6po_nextroute);
2353 		if (pktopt->ip6po_nexthop)
2354 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2355 		pktopt->ip6po_nexthop = NULL;
2356 	}
2357 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2358 		if (pktopt->ip6po_hbh)
2359 			free(pktopt->ip6po_hbh, M_IP6OPT);
2360 		pktopt->ip6po_hbh = NULL;
2361 	}
2362 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2363 		if (pktopt->ip6po_dest1)
2364 			free(pktopt->ip6po_dest1, M_IP6OPT);
2365 		pktopt->ip6po_dest1 = NULL;
2366 	}
2367 	if (optname == -1 || optname == IPV6_RTHDR) {
2368 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2369 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2370 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2371 		rtcache_free(&pktopt->ip6po_route);
2372 	}
2373 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2374 		if (pktopt->ip6po_dest2)
2375 			free(pktopt->ip6po_dest2, M_IP6OPT);
2376 		pktopt->ip6po_dest2 = NULL;
2377 	}
2378 }
2379 
2380 #define PKTOPT_EXTHDRCPY(type) 					\
2381 do {								\
2382 	if (src->type) {					\
2383 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2384 		dst->type = malloc(hlen, M_IP6OPT, canwait);	\
2385 		if (dst->type == NULL && canwait == M_NOWAIT)	\
2386 			goto bad;				\
2387 		memcpy(dst->type, src->type, hlen);		\
2388 	}							\
2389 } while (/*CONSTCOND*/ 0)
2390 
2391 static int
2392 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2393 {
2394 	dst->ip6po_hlim = src->ip6po_hlim;
2395 	dst->ip6po_tclass = src->ip6po_tclass;
2396 	dst->ip6po_flags = src->ip6po_flags;
2397 	if (src->ip6po_pktinfo) {
2398 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2399 		    M_IP6OPT, canwait);
2400 		if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT)
2401 			goto bad;
2402 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2403 	}
2404 	if (src->ip6po_nexthop) {
2405 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2406 		    M_IP6OPT, canwait);
2407 		if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT)
2408 			goto bad;
2409 		memcpy(dst->ip6po_nexthop, src->ip6po_nexthop,
2410 		    src->ip6po_nexthop->sa_len);
2411 	}
2412 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2413 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2414 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2415 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2416 	return (0);
2417 
2418   bad:
2419 	if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
2420 	if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
2421 	if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
2422 	if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
2423 	if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
2424 	if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
2425 
2426 	return (ENOBUFS);
2427 }
2428 #undef PKTOPT_EXTHDRCPY
2429 
2430 struct ip6_pktopts *
2431 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2432 {
2433 	int error;
2434 	struct ip6_pktopts *dst;
2435 
2436 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2437 	if (dst == NULL && canwait == M_NOWAIT)
2438 		return (NULL);
2439 	ip6_initpktopts(dst);
2440 
2441 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2442 		free(dst, M_IP6OPT);
2443 		return (NULL);
2444 	}
2445 
2446 	return (dst);
2447 }
2448 
2449 void
2450 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2451 {
2452 	if (pktopt == NULL)
2453 		return;
2454 
2455 	ip6_clearpktopts(pktopt, -1);
2456 
2457 	free(pktopt, M_IP6OPT);
2458 }
2459 
2460 /*
2461  * Set the IP6 multicast options in response to user setsockopt().
2462  */
2463 static int
2464 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
2465 {
2466 	int error = 0;
2467 	u_int loop, ifindex;
2468 	struct ipv6_mreq *mreq;
2469 	struct ifnet *ifp;
2470 	struct ip6_moptions *im6o = *im6op;
2471 	struct route ro;
2472 	struct in6_multi_mship *imm;
2473 	struct lwp *l = curlwp;	/* XXX */
2474 
2475 	if (im6o == NULL) {
2476 		/*
2477 		 * No multicast option buffer attached to the pcb;
2478 		 * allocate one and initialize to default values.
2479 		 */
2480 		im6o = (struct ip6_moptions *)
2481 			malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
2482 
2483 		if (im6o == NULL)
2484 			return (ENOBUFS);
2485 		*im6op = im6o;
2486 		im6o->im6o_multicast_ifp = NULL;
2487 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2488 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2489 		LIST_INIT(&im6o->im6o_memberships);
2490 	}
2491 
2492 	switch (optname) {
2493 
2494 	case IPV6_MULTICAST_IF:
2495 		/*
2496 		 * Select the interface for outgoing multicast packets.
2497 		 */
2498 		if (m == NULL || m->m_len != sizeof(u_int)) {
2499 			error = EINVAL;
2500 			break;
2501 		}
2502 		bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
2503 		if (ifindex != 0) {
2504 			if (if_indexlim <= ifindex || !ifindex2ifnet[ifindex]) {
2505 				error = ENXIO;	/* XXX EINVAL? */
2506 				break;
2507 			}
2508 			ifp = ifindex2ifnet[ifindex];
2509 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
2510 				error = EADDRNOTAVAIL;
2511 				break;
2512 			}
2513 		} else
2514 			ifp = NULL;
2515 		im6o->im6o_multicast_ifp = ifp;
2516 		break;
2517 
2518 	case IPV6_MULTICAST_HOPS:
2519 	    {
2520 		/*
2521 		 * Set the IP6 hoplimit for outgoing multicast packets.
2522 		 */
2523 		int optval;
2524 		if (m == NULL || m->m_len != sizeof(int)) {
2525 			error = EINVAL;
2526 			break;
2527 		}
2528 		bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2529 		if (optval < -1 || optval >= 256)
2530 			error = EINVAL;
2531 		else if (optval == -1)
2532 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2533 		else
2534 			im6o->im6o_multicast_hlim = optval;
2535 		break;
2536 	    }
2537 
2538 	case IPV6_MULTICAST_LOOP:
2539 		/*
2540 		 * Set the loopback flag for outgoing multicast packets.
2541 		 * Must be zero or one.
2542 		 */
2543 		if (m == NULL || m->m_len != sizeof(u_int)) {
2544 			error = EINVAL;
2545 			break;
2546 		}
2547 		bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2548 		if (loop > 1) {
2549 			error = EINVAL;
2550 			break;
2551 		}
2552 		im6o->im6o_multicast_loop = loop;
2553 		break;
2554 
2555 	case IPV6_JOIN_GROUP:
2556 		/*
2557 		 * Add a multicast group membership.
2558 		 * Group must be a valid IP6 multicast address.
2559 		 */
2560 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2561 			error = EINVAL;
2562 			break;
2563 		}
2564 		mreq = mtod(m, struct ipv6_mreq *);
2565 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2566 			/*
2567 			 * We use the unspecified address to specify to accept
2568 			 * all multicast addresses. Only super user is allowed
2569 			 * to do this.
2570 			 */
2571 			if (kauth_authorize_generic(l->l_cred,
2572 			    KAUTH_GENERIC_ISSUSER, NULL))
2573 			{
2574 				error = EACCES;
2575 				break;
2576 			}
2577 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2578 			error = EINVAL;
2579 			break;
2580 		}
2581 
2582 		/*
2583 		 * If no interface was explicitly specified, choose an
2584 		 * appropriate one according to the given multicast address.
2585 		 */
2586 		if (mreq->ipv6mr_interface == 0) {
2587 			struct rtentry *rt;
2588 			union {
2589 				struct sockaddr		dst;
2590 				struct sockaddr_in6	dst6;
2591 			} u;
2592 
2593 			/*
2594 			 * Look up the routing table for the
2595 			 * address, and choose the outgoing interface.
2596 			 *   XXX: is it a good approach?
2597 			 */
2598 			memset(&ro, 0, sizeof(ro));
2599 			sockaddr_in6_init(&u.dst6, &mreq->ipv6mr_multiaddr, 0,
2600 			    0, 0);
2601 			rtcache_setdst(&ro, &u.dst);
2602 			ifp = (rt = rtcache_init(&ro)) != NULL ? rt->rt_ifp
2603 			                                        : NULL;
2604 			rtcache_free(&ro);
2605 		} else {
2606 			/*
2607 			 * If the interface is specified, validate it.
2608 			 */
2609 			if (if_indexlim <= mreq->ipv6mr_interface ||
2610 			    !ifindex2ifnet[mreq->ipv6mr_interface]) {
2611 				error = ENXIO;	/* XXX EINVAL? */
2612 				break;
2613 			}
2614 			ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2615 		}
2616 
2617 		/*
2618 		 * See if we found an interface, and confirm that it
2619 		 * supports multicast
2620 		 */
2621 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2622 			error = EADDRNOTAVAIL;
2623 			break;
2624 		}
2625 
2626 		if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) {
2627 			error = EADDRNOTAVAIL; /* XXX: should not happen */
2628 			break;
2629 		}
2630 
2631 		/*
2632 		 * See if the membership already exists.
2633 		 */
2634 		for (imm = im6o->im6o_memberships.lh_first;
2635 		     imm != NULL; imm = imm->i6mm_chain.le_next)
2636 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
2637 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2638 			    &mreq->ipv6mr_multiaddr))
2639 				break;
2640 		if (imm != NULL) {
2641 			error = EADDRINUSE;
2642 			break;
2643 		}
2644 		/*
2645 		 * Everything looks good; add a new record to the multicast
2646 		 * address list for the given interface.
2647 		 */
2648 		imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error, 0);
2649 		if (imm == NULL)
2650 			break;
2651 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2652 		break;
2653 
2654 	case IPV6_LEAVE_GROUP:
2655 		/*
2656 		 * Drop a multicast group membership.
2657 		 * Group must be a valid IP6 multicast address.
2658 		 */
2659 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2660 			error = EINVAL;
2661 			break;
2662 		}
2663 		mreq = mtod(m, struct ipv6_mreq *);
2664 
2665 		/*
2666 		 * If an interface address was specified, get a pointer
2667 		 * to its ifnet structure.
2668 		 */
2669 		if (mreq->ipv6mr_interface != 0) {
2670 			if (if_indexlim <= mreq->ipv6mr_interface ||
2671 			    !ifindex2ifnet[mreq->ipv6mr_interface]) {
2672 				error = ENXIO;	/* XXX EINVAL? */
2673 				break;
2674 			}
2675 			ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2676 		} else
2677 			ifp = NULL;
2678 
2679 		/* Fill in the scope zone ID */
2680 		if (ifp) {
2681 			if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) {
2682 				/* XXX: should not happen */
2683 				error = EADDRNOTAVAIL;
2684 				break;
2685 			}
2686 		} else if (mreq->ipv6mr_interface != 0) {
2687 			/*
2688 			 * XXX: This case would happens when the (positive)
2689 			 * index is in the valid range, but the corresponding
2690 			 * interface has been detached dynamically.  The above
2691 			 * check probably avoids such case to happen here, but
2692 			 * we check it explicitly for safety.
2693 			 */
2694 			error = EADDRNOTAVAIL;
2695 			break;
2696 		} else {	/* ipv6mr_interface == 0 */
2697 			struct sockaddr_in6 sa6_mc;
2698 
2699 			/*
2700 			 * The API spec says as follows:
2701 			 *  If the interface index is specified as 0, the
2702 			 *  system may choose a multicast group membership to
2703 			 *  drop by matching the multicast address only.
2704 			 * On the other hand, we cannot disambiguate the scope
2705 			 * zone unless an interface is provided.  Thus, we
2706 			 * check if there's ambiguity with the default scope
2707 			 * zone as the last resort.
2708 			 */
2709 			sockaddr_in6_init(&sa6_mc, &mreq->ipv6mr_multiaddr,
2710 			    0, 0, 0);
2711 			error = sa6_embedscope(&sa6_mc, ip6_use_defzone);
2712 			if (error != 0)
2713 				break;
2714 			mreq->ipv6mr_multiaddr = sa6_mc.sin6_addr;
2715 		}
2716 
2717 		/*
2718 		 * Find the membership in the membership list.
2719 		 */
2720 		for (imm = im6o->im6o_memberships.lh_first;
2721 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
2722 			if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2723 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2724 			    &mreq->ipv6mr_multiaddr))
2725 				break;
2726 		}
2727 		if (imm == NULL) {
2728 			/* Unable to resolve interface */
2729 			error = EADDRNOTAVAIL;
2730 			break;
2731 		}
2732 		/*
2733 		 * Give up the multicast address record to which the
2734 		 * membership points.
2735 		 */
2736 		LIST_REMOVE(imm, i6mm_chain);
2737 		in6_leavegroup(imm);
2738 		break;
2739 
2740 	default:
2741 		error = EOPNOTSUPP;
2742 		break;
2743 	}
2744 
2745 	/*
2746 	 * If all options have default values, no need to keep the mbuf.
2747 	 */
2748 	if (im6o->im6o_multicast_ifp == NULL &&
2749 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2750 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2751 	    im6o->im6o_memberships.lh_first == NULL) {
2752 		free(*im6op, M_IPMOPTS);
2753 		*im6op = NULL;
2754 	}
2755 
2756 	return (error);
2757 }
2758 
2759 /*
2760  * Return the IP6 multicast options in response to user getsockopt().
2761  */
2762 static int
2763 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp)
2764 {
2765 	u_int *hlim, *loop, *ifindex;
2766 
2767 	*mp = m_get(M_WAIT, MT_SOOPTS);
2768 
2769 	switch (optname) {
2770 
2771 	case IPV6_MULTICAST_IF:
2772 		ifindex = mtod(*mp, u_int *);
2773 		(*mp)->m_len = sizeof(u_int);
2774 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2775 			*ifindex = 0;
2776 		else
2777 			*ifindex = im6o->im6o_multicast_ifp->if_index;
2778 		return (0);
2779 
2780 	case IPV6_MULTICAST_HOPS:
2781 		hlim = mtod(*mp, u_int *);
2782 		(*mp)->m_len = sizeof(u_int);
2783 		if (im6o == NULL)
2784 			*hlim = ip6_defmcasthlim;
2785 		else
2786 			*hlim = im6o->im6o_multicast_hlim;
2787 		return (0);
2788 
2789 	case IPV6_MULTICAST_LOOP:
2790 		loop = mtod(*mp, u_int *);
2791 		(*mp)->m_len = sizeof(u_int);
2792 		if (im6o == NULL)
2793 			*loop = ip6_defmcasthlim;
2794 		else
2795 			*loop = im6o->im6o_multicast_loop;
2796 		return (0);
2797 
2798 	default:
2799 		return (EOPNOTSUPP);
2800 	}
2801 }
2802 
2803 /*
2804  * Discard the IP6 multicast options.
2805  */
2806 void
2807 ip6_freemoptions(struct ip6_moptions *im6o)
2808 {
2809 	struct in6_multi_mship *imm;
2810 
2811 	if (im6o == NULL)
2812 		return;
2813 
2814 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2815 		LIST_REMOVE(imm, i6mm_chain);
2816 		in6_leavegroup(imm);
2817 	}
2818 	free(im6o, M_IPMOPTS);
2819 }
2820 
2821 /*
2822  * Set IPv6 outgoing packet options based on advanced API.
2823  */
2824 int
2825 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2826 	struct ip6_pktopts *stickyopt, int priv, int uproto)
2827 {
2828 	struct cmsghdr *cm = 0;
2829 
2830 	if (control == NULL || opt == NULL)
2831 		return (EINVAL);
2832 
2833 	ip6_initpktopts(opt);
2834 	if (stickyopt) {
2835 		int error;
2836 
2837 		/*
2838 		 * If stickyopt is provided, make a local copy of the options
2839 		 * for this particular packet, then override them by ancillary
2840 		 * objects.
2841 		 * XXX: copypktopts() does not copy the cached route to a next
2842 		 * hop (if any).  This is not very good in terms of efficiency,
2843 		 * but we can allow this since this option should be rarely
2844 		 * used.
2845 		 */
2846 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2847 			return (error);
2848 	}
2849 
2850 	/*
2851 	 * XXX: Currently, we assume all the optional information is stored
2852 	 * in a single mbuf.
2853 	 */
2854 	if (control->m_next)
2855 		return (EINVAL);
2856 
2857 	for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2858 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2859 		int error;
2860 
2861 		if (control->m_len < CMSG_LEN(0))
2862 			return (EINVAL);
2863 
2864 		cm = mtod(control, struct cmsghdr *);
2865 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2866 			return (EINVAL);
2867 		if (cm->cmsg_level != IPPROTO_IPV6)
2868 			continue;
2869 
2870 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2871 		    cm->cmsg_len - CMSG_LEN(0), opt, priv, 0, 1, uproto);
2872 		if (error)
2873 			return (error);
2874 	}
2875 
2876 	return (0);
2877 }
2878 
2879 /*
2880  * Set a particular packet option, as a sticky option or an ancillary data
2881  * item.  "len" can be 0 only when it's a sticky option.
2882  * We have 4 cases of combination of "sticky" and "cmsg":
2883  * "sticky=0, cmsg=0": impossible
2884  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2885  * "sticky=1, cmsg=0": RFC3542 socket option
2886  * "sticky=1, cmsg=1": RFC2292 socket option
2887  */
2888 static int
2889 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2890     int priv, int sticky, int cmsg, int uproto)
2891 {
2892 	int minmtupolicy;
2893 
2894 	if (!sticky && !cmsg) {
2895 #ifdef DIAGNOSTIC
2896 		printf("ip6_setpktopt: impossible case\n");
2897 #endif
2898 		return (EINVAL);
2899 	}
2900 
2901 	/*
2902 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2903 	 * not be specified in the context of RFC3542.  Conversely,
2904 	 * RFC3542 types should not be specified in the context of RFC2292.
2905 	 */
2906 	if (!cmsg) {
2907 		switch (optname) {
2908 		case IPV6_2292PKTINFO:
2909 		case IPV6_2292HOPLIMIT:
2910 		case IPV6_2292NEXTHOP:
2911 		case IPV6_2292HOPOPTS:
2912 		case IPV6_2292DSTOPTS:
2913 		case IPV6_2292RTHDR:
2914 		case IPV6_2292PKTOPTIONS:
2915 			return (ENOPROTOOPT);
2916 		}
2917 	}
2918 	if (sticky && cmsg) {
2919 		switch (optname) {
2920 		case IPV6_PKTINFO:
2921 		case IPV6_HOPLIMIT:
2922 		case IPV6_NEXTHOP:
2923 		case IPV6_HOPOPTS:
2924 		case IPV6_DSTOPTS:
2925 		case IPV6_RTHDRDSTOPTS:
2926 		case IPV6_RTHDR:
2927 		case IPV6_USE_MIN_MTU:
2928 		case IPV6_DONTFRAG:
2929 		case IPV6_OTCLASS:
2930 		case IPV6_TCLASS:
2931 			return (ENOPROTOOPT);
2932 		}
2933 	}
2934 
2935 	switch (optname) {
2936 #ifdef RFC2292
2937 	case IPV6_2292PKTINFO:
2938 #endif
2939 	case IPV6_PKTINFO:
2940 	{
2941 		struct ifnet *ifp = NULL;
2942 		struct in6_pktinfo *pktinfo;
2943 
2944 		if (len != sizeof(struct in6_pktinfo))
2945 			return (EINVAL);
2946 
2947 		pktinfo = (struct in6_pktinfo *)buf;
2948 
2949 		/*
2950 		 * An application can clear any sticky IPV6_PKTINFO option by
2951 		 * doing a "regular" setsockopt with ipi6_addr being
2952 		 * in6addr_any and ipi6_ifindex being zero.
2953 		 * [RFC 3542, Section 6]
2954 		 */
2955 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2956 		    pktinfo->ipi6_ifindex == 0 &&
2957 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2958 			ip6_clearpktopts(opt, optname);
2959 			break;
2960 		}
2961 
2962 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2963 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2964 			return (EINVAL);
2965 		}
2966 
2967 		/* validate the interface index if specified. */
2968 		if (pktinfo->ipi6_ifindex >= if_indexlim) {
2969 			 return (ENXIO);
2970 		}
2971 		if (pktinfo->ipi6_ifindex) {
2972 			ifp = ifindex2ifnet[pktinfo->ipi6_ifindex];
2973 			if (ifp == NULL)
2974 				return (ENXIO);
2975 		}
2976 
2977 		/*
2978 		 * We store the address anyway, and let in6_selectsrc()
2979 		 * validate the specified address.  This is because ipi6_addr
2980 		 * may not have enough information about its scope zone, and
2981 		 * we may need additional information (such as outgoing
2982 		 * interface or the scope zone of a destination address) to
2983 		 * disambiguate the scope.
2984 		 * XXX: the delay of the validation may confuse the
2985 		 * application when it is used as a sticky option.
2986 		 */
2987 		if (opt->ip6po_pktinfo == NULL) {
2988 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2989 			    M_IP6OPT, M_NOWAIT);
2990 			if (opt->ip6po_pktinfo == NULL)
2991 				return (ENOBUFS);
2992 		}
2993 		memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo));
2994 		break;
2995 	}
2996 
2997 #ifdef RFC2292
2998 	case IPV6_2292HOPLIMIT:
2999 #endif
3000 	case IPV6_HOPLIMIT:
3001 	{
3002 		int *hlimp;
3003 
3004 		/*
3005 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
3006 		 * to simplify the ordering among hoplimit options.
3007 		 */
3008 		if (optname == IPV6_HOPLIMIT && sticky)
3009 			return (ENOPROTOOPT);
3010 
3011 		if (len != sizeof(int))
3012 			return (EINVAL);
3013 		hlimp = (int *)buf;
3014 		if (*hlimp < -1 || *hlimp > 255)
3015 			return (EINVAL);
3016 
3017 		opt->ip6po_hlim = *hlimp;
3018 		break;
3019 	}
3020 
3021 	case IPV6_OTCLASS:
3022 		if (len != sizeof(u_int8_t))
3023 			return (EINVAL);
3024 
3025 		opt->ip6po_tclass = *(u_int8_t *)buf;
3026 		break;
3027 
3028 	case IPV6_TCLASS:
3029 	{
3030 		int tclass;
3031 
3032 		if (len != sizeof(int))
3033 			return (EINVAL);
3034 		tclass = *(int *)buf;
3035 		if (tclass < -1 || tclass > 255)
3036 			return (EINVAL);
3037 
3038 		opt->ip6po_tclass = tclass;
3039 		break;
3040 	}
3041 
3042 #ifdef RFC2292
3043 	case IPV6_2292NEXTHOP:
3044 #endif
3045 	case IPV6_NEXTHOP:
3046 		if (!priv)
3047 			return (EPERM);
3048 
3049 		if (len == 0) {	/* just remove the option */
3050 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
3051 			break;
3052 		}
3053 
3054 		/* check if cmsg_len is large enough for sa_len */
3055 		if (len < sizeof(struct sockaddr) || len < *buf)
3056 			return (EINVAL);
3057 
3058 		switch (((struct sockaddr *)buf)->sa_family) {
3059 		case AF_INET6:
3060 		{
3061 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3062 			int error;
3063 
3064 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3065 				return (EINVAL);
3066 
3067 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3068 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3069 				return (EINVAL);
3070 			}
3071 			if ((error = sa6_embedscope(sa6, ip6_use_defzone))
3072 			    != 0) {
3073 				return (error);
3074 			}
3075 			break;
3076 		}
3077 		case AF_LINK:	/* eventually be supported? */
3078 		default:
3079 			return (EAFNOSUPPORT);
3080 		}
3081 
3082 		/* turn off the previous option, then set the new option. */
3083 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
3084 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3085 		if (opt->ip6po_nexthop == NULL)
3086 			return (ENOBUFS);
3087 		memcpy(opt->ip6po_nexthop, buf, *buf);
3088 		break;
3089 
3090 #ifdef RFC2292
3091 	case IPV6_2292HOPOPTS:
3092 #endif
3093 	case IPV6_HOPOPTS:
3094 	{
3095 		struct ip6_hbh *hbh;
3096 		int hbhlen;
3097 
3098 		/*
3099 		 * XXX: We don't allow a non-privileged user to set ANY HbH
3100 		 * options, since per-option restriction has too much
3101 		 * overhead.
3102 		 */
3103 		if (!priv)
3104 			return (EPERM);
3105 
3106 		if (len == 0) {
3107 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
3108 			break;	/* just remove the option */
3109 		}
3110 
3111 		/* message length validation */
3112 		if (len < sizeof(struct ip6_hbh))
3113 			return (EINVAL);
3114 		hbh = (struct ip6_hbh *)buf;
3115 		hbhlen = (hbh->ip6h_len + 1) << 3;
3116 		if (len != hbhlen)
3117 			return (EINVAL);
3118 
3119 		/* turn off the previous option, then set the new option. */
3120 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
3121 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3122 		if (opt->ip6po_hbh == NULL)
3123 			return (ENOBUFS);
3124 		memcpy(opt->ip6po_hbh, hbh, hbhlen);
3125 
3126 		break;
3127 	}
3128 
3129 #ifdef RFC2292
3130 	case IPV6_2292DSTOPTS:
3131 #endif
3132 	case IPV6_DSTOPTS:
3133 	case IPV6_RTHDRDSTOPTS:
3134 	{
3135 		struct ip6_dest *dest, **newdest = NULL;
3136 		int destlen;
3137 
3138 		if (!priv)	/* XXX: see the comment for IPV6_HOPOPTS */
3139 			return (EPERM);
3140 
3141 		if (len == 0) {
3142 			ip6_clearpktopts(opt, optname);
3143 			break;	/* just remove the option */
3144 		}
3145 
3146 		/* message length validation */
3147 		if (len < sizeof(struct ip6_dest))
3148 			return (EINVAL);
3149 		dest = (struct ip6_dest *)buf;
3150 		destlen = (dest->ip6d_len + 1) << 3;
3151 		if (len != destlen)
3152 			return (EINVAL);
3153 		/*
3154 		 * Determine the position that the destination options header
3155 		 * should be inserted; before or after the routing header.
3156 		 */
3157 		switch (optname) {
3158 		case IPV6_2292DSTOPTS:
3159 			/*
3160 			 * The old advanced API is ambiguous on this point.
3161 			 * Our approach is to determine the position based
3162 			 * according to the existence of a routing header.
3163 			 * Note, however, that this depends on the order of the
3164 			 * extension headers in the ancillary data; the 1st
3165 			 * part of the destination options header must appear
3166 			 * before the routing header in the ancillary data,
3167 			 * too.
3168 			 * RFC3542 solved the ambiguity by introducing
3169 			 * separate ancillary data or option types.
3170 			 */
3171 			if (opt->ip6po_rthdr == NULL)
3172 				newdest = &opt->ip6po_dest1;
3173 			else
3174 				newdest = &opt->ip6po_dest2;
3175 			break;
3176 		case IPV6_RTHDRDSTOPTS:
3177 			newdest = &opt->ip6po_dest1;
3178 			break;
3179 		case IPV6_DSTOPTS:
3180 			newdest = &opt->ip6po_dest2;
3181 			break;
3182 		}
3183 
3184 		/* turn off the previous option, then set the new option. */
3185 		ip6_clearpktopts(opt, optname);
3186 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3187 		if (*newdest == NULL)
3188 			return (ENOBUFS);
3189 		memcpy(*newdest, dest, destlen);
3190 
3191 		break;
3192 	}
3193 
3194 #ifdef RFC2292
3195 	case IPV6_2292RTHDR:
3196 #endif
3197 	case IPV6_RTHDR:
3198 	{
3199 		struct ip6_rthdr *rth;
3200 		int rthlen;
3201 
3202 		if (len == 0) {
3203 			ip6_clearpktopts(opt, IPV6_RTHDR);
3204 			break;	/* just remove the option */
3205 		}
3206 
3207 		/* message length validation */
3208 		if (len < sizeof(struct ip6_rthdr))
3209 			return (EINVAL);
3210 		rth = (struct ip6_rthdr *)buf;
3211 		rthlen = (rth->ip6r_len + 1) << 3;
3212 		if (len != rthlen)
3213 			return (EINVAL);
3214 		switch (rth->ip6r_type) {
3215 		case IPV6_RTHDR_TYPE_0:
3216 			if (rth->ip6r_len == 0)	/* must contain one addr */
3217 				return (EINVAL);
3218 			if (rth->ip6r_len % 2) /* length must be even */
3219 				return (EINVAL);
3220 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3221 				return (EINVAL);
3222 			break;
3223 		default:
3224 			return (EINVAL);	/* not supported */
3225 		}
3226 		/* turn off the previous option */
3227 		ip6_clearpktopts(opt, IPV6_RTHDR);
3228 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3229 		if (opt->ip6po_rthdr == NULL)
3230 			return (ENOBUFS);
3231 		memcpy(opt->ip6po_rthdr, rth, rthlen);
3232 		break;
3233 	}
3234 
3235 	case IPV6_USE_MIN_MTU:
3236 		if (len != sizeof(int))
3237 			return (EINVAL);
3238 		minmtupolicy = *(int *)buf;
3239 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3240 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3241 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3242 			return (EINVAL);
3243 		}
3244 		opt->ip6po_minmtu = minmtupolicy;
3245 		break;
3246 
3247 	case IPV6_DONTFRAG:
3248 		if (len != sizeof(int))
3249 			return (EINVAL);
3250 
3251 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3252 			/*
3253 			 * we ignore this option for TCP sockets.
3254 			 * (RFC3542 leaves this case unspecified.)
3255 			 */
3256 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3257 		} else
3258 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3259 		break;
3260 
3261 	default:
3262 		return (ENOPROTOOPT);
3263 	} /* end of switch */
3264 
3265 	return (0);
3266 }
3267 
3268 /*
3269  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3270  * packet to the input queue of a specified interface.  Note that this
3271  * calls the output routine of the loopback "driver", but with an interface
3272  * pointer that might NOT be lo0ifp -- easier than replicating that code here.
3273  */
3274 void
3275 ip6_mloopback(struct ifnet *ifp, struct mbuf *m,
3276 	const struct sockaddr_in6 *dst)
3277 {
3278 	struct mbuf *copym;
3279 	struct ip6_hdr *ip6;
3280 
3281 	copym = m_copy(m, 0, M_COPYALL);
3282 	if (copym == NULL)
3283 		return;
3284 
3285 	/*
3286 	 * Make sure to deep-copy IPv6 header portion in case the data
3287 	 * is in an mbuf cluster, so that we can safely override the IPv6
3288 	 * header portion later.
3289 	 */
3290 	if ((copym->m_flags & M_EXT) != 0 ||
3291 	    copym->m_len < sizeof(struct ip6_hdr)) {
3292 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3293 		if (copym == NULL)
3294 			return;
3295 	}
3296 
3297 #ifdef DIAGNOSTIC
3298 	if (copym->m_len < sizeof(*ip6)) {
3299 		m_freem(copym);
3300 		return;
3301 	}
3302 #endif
3303 
3304 	ip6 = mtod(copym, struct ip6_hdr *);
3305 	/*
3306 	 * clear embedded scope identifiers if necessary.
3307 	 * in6_clearscope will touch the addresses only when necessary.
3308 	 */
3309 	in6_clearscope(&ip6->ip6_src);
3310 	in6_clearscope(&ip6->ip6_dst);
3311 
3312 	(void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL);
3313 }
3314 
3315 /*
3316  * Chop IPv6 header off from the payload.
3317  */
3318 static int
3319 ip6_splithdr(struct mbuf *m,  struct ip6_exthdrs *exthdrs)
3320 {
3321 	struct mbuf *mh;
3322 	struct ip6_hdr *ip6;
3323 
3324 	ip6 = mtod(m, struct ip6_hdr *);
3325 	if (m->m_len > sizeof(*ip6)) {
3326 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3327 		if (mh == 0) {
3328 			m_freem(m);
3329 			return ENOBUFS;
3330 		}
3331 		M_MOVE_PKTHDR(mh, m);
3332 		MH_ALIGN(mh, sizeof(*ip6));
3333 		m->m_len -= sizeof(*ip6);
3334 		m->m_data += sizeof(*ip6);
3335 		mh->m_next = m;
3336 		m = mh;
3337 		m->m_len = sizeof(*ip6);
3338 		bcopy((void *)ip6, mtod(m, void *), sizeof(*ip6));
3339 	}
3340 	exthdrs->ip6e_ip6 = m;
3341 	return 0;
3342 }
3343 
3344 /*
3345  * Compute IPv6 extension header length.
3346  */
3347 int
3348 ip6_optlen(struct in6pcb *in6p)
3349 {
3350 	int len;
3351 
3352 	if (!in6p->in6p_outputopts)
3353 		return 0;
3354 
3355 	len = 0;
3356 #define elen(x) \
3357     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3358 
3359 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3360 	len += elen(in6p->in6p_outputopts->ip6po_dest1);
3361 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3362 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3363 	return len;
3364 #undef elen
3365 }
3366