xref: /netbsd-src/sys/netinet6/ip6_output.c (revision 8b0f9554ff8762542c4defc4f70e1eb76fb508fa)
1 /*	$NetBSD: ip6_output.c,v 1.123 2007/11/06 23:48:24 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 1.123 2007/11/06 23:48:24 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 
208 #ifdef  DIAGNOSTIC
209 	if ((m->m_flags & M_PKTHDR) == 0)
210 		panic("ip6_output: no HDR");
211 
212 	if ((m->m_pkthdr.csum_flags &
213 	    (M_CSUM_TCPv4|M_CSUM_UDPv4|M_CSUM_TSOv4)) != 0) {
214 		panic("ip6_output: IPv4 checksum offload flags: %d",
215 		    m->m_pkthdr.csum_flags);
216 	}
217 
218 	if ((m->m_pkthdr.csum_flags & (M_CSUM_TCPv6|M_CSUM_UDPv6)) ==
219 	    (M_CSUM_TCPv6|M_CSUM_UDPv6)) {
220 		panic("ip6_output: conflicting checksum offload flags: %d",
221 		    m->m_pkthdr.csum_flags);
222 	}
223 #endif
224 
225 	M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data, sizeof(struct ip6_hdr));
226 
227 #define MAKE_EXTHDR(hp, mp)						\
228     do {								\
229 	if (hp) {							\
230 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
231 		error = ip6_copyexthdr((mp), (void *)(hp), 		\
232 		    ((eh)->ip6e_len + 1) << 3);				\
233 		if (error)						\
234 			goto freehdrs;					\
235 	}								\
236     } while (/*CONSTCOND*/ 0)
237 
238 	bzero(&exthdrs, sizeof(exthdrs));
239 	if (opt) {
240 		/* Hop-by-Hop options header */
241 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
242 		/* Destination options header(1st part) */
243 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
244 		/* Routing header */
245 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
246 		/* Destination options header(2nd part) */
247 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
248 	}
249 
250 #ifdef IPSEC
251 	if ((flags & IPV6_FORWARDING) != 0) {
252 		needipsec = 0;
253 		goto skippolicycheck;
254 	}
255 
256 	/* get a security policy for this packet */
257 	if (so == NULL)
258 		sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
259 	else {
260 		if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp,
261 					 IPSEC_DIR_OUTBOUND)) {
262 			needipsec = 0;
263 			goto skippolicycheck;
264 		}
265 		sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
266 	}
267 
268 	if (sp == NULL) {
269 		ipsec6stat.out_inval++;
270 		goto freehdrs;
271 	}
272 
273 	error = 0;
274 
275 	/* check policy */
276 	switch (sp->policy) {
277 	case IPSEC_POLICY_DISCARD:
278 		/*
279 		 * This packet is just discarded.
280 		 */
281 		ipsec6stat.out_polvio++;
282 		goto freehdrs;
283 
284 	case IPSEC_POLICY_BYPASS:
285 	case IPSEC_POLICY_NONE:
286 		/* no need to do IPsec. */
287 		needipsec = 0;
288 		break;
289 
290 	case IPSEC_POLICY_IPSEC:
291 		if (sp->req == NULL) {
292 			/* XXX should be panic ? */
293 			printf("ip6_output: No IPsec request specified.\n");
294 			error = EINVAL;
295 			goto freehdrs;
296 		}
297 		needipsec = 1;
298 		break;
299 
300 	case IPSEC_POLICY_ENTRUST:
301 	default:
302 		printf("ip6_output: Invalid policy found. %d\n", sp->policy);
303 	}
304 
305   skippolicycheck:;
306 #endif /* IPSEC */
307 
308 	/*
309 	 * Calculate the total length of the extension header chain.
310 	 * Keep the length of the unfragmentable part for fragmentation.
311 	 */
312 	optlen = 0;
313 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
314 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
315 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
316 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
317 	/* NOTE: we don't add AH/ESP length here. do that later. */
318 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
319 
320 #ifdef FAST_IPSEC
321 	/* Check the security policy (SP) for the packet */
322 
323 	/* XXX For moment, we doesn't support packet with extented action */
324 	if (optlen !=0)
325 		goto freehdrs;
326 
327 	sp = ipsec6_check_policy(m,so,flags,&needipsec,&error);
328 	if (error != 0) {
329 		/*
330 		 * Hack: -EINVAL is used to signal that a packet
331 		 * should be silently discarded.  This is typically
332 		 * because we asked key management for an SA and
333 		 * it was delayed (e.g. kicked up to IKE).
334 		 */
335 	if (error == -EINVAL)
336 		error = 0;
337 	goto freehdrs;
338     }
339 #endif /* FAST_IPSEC */
340 
341 
342 	if (needipsec &&
343 	    (m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0) {
344 		in6_delayed_cksum(m);
345 		m->m_pkthdr.csum_flags &= ~(M_CSUM_UDPv6|M_CSUM_TCPv6);
346 	}
347 
348 
349 	/*
350 	 * If we need IPsec, or there is at least one extension header,
351 	 * separate IP6 header from the payload.
352 	 */
353 	if ((needipsec || optlen) && !hdrsplit) {
354 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
355 			m = NULL;
356 			goto freehdrs;
357 		}
358 		m = exthdrs.ip6e_ip6;
359 		hdrsplit++;
360 	}
361 
362 	/* adjust pointer */
363 	ip6 = mtod(m, struct ip6_hdr *);
364 
365 	/* adjust mbuf packet header length */
366 	m->m_pkthdr.len += optlen;
367 	plen = m->m_pkthdr.len - sizeof(*ip6);
368 
369 	/* If this is a jumbo payload, insert a jumbo payload option. */
370 	if (plen > IPV6_MAXPACKET) {
371 		if (!hdrsplit) {
372 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
373 				m = NULL;
374 				goto freehdrs;
375 			}
376 			m = exthdrs.ip6e_ip6;
377 			hdrsplit++;
378 		}
379 		/* adjust pointer */
380 		ip6 = mtod(m, struct ip6_hdr *);
381 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
382 			goto freehdrs;
383 		optlen += 8; /* XXX JUMBOOPTLEN */
384 		ip6->ip6_plen = 0;
385 	} else
386 		ip6->ip6_plen = htons(plen);
387 
388 	/*
389 	 * Concatenate headers and fill in next header fields.
390 	 * Here we have, on "m"
391 	 *	IPv6 payload
392 	 * and we insert headers accordingly.  Finally, we should be getting:
393 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
394 	 *
395 	 * during the header composing process, "m" points to IPv6 header.
396 	 * "mprev" points to an extension header prior to esp.
397 	 */
398 	{
399 		u_char *nexthdrp = &ip6->ip6_nxt;
400 		struct mbuf *mprev = m;
401 
402 		/*
403 		 * we treat dest2 specially.  this makes IPsec processing
404 		 * much easier.  the goal here is to make mprev point the
405 		 * mbuf prior to dest2.
406 		 *
407 		 * result: IPv6 dest2 payload
408 		 * m and mprev will point to IPv6 header.
409 		 */
410 		if (exthdrs.ip6e_dest2) {
411 			if (!hdrsplit)
412 				panic("assumption failed: hdr not split");
413 			exthdrs.ip6e_dest2->m_next = m->m_next;
414 			m->m_next = exthdrs.ip6e_dest2;
415 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
416 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
417 		}
418 
419 #define MAKE_CHAIN(m, mp, p, i)\
420     do {\
421 	if (m) {\
422 		if (!hdrsplit) \
423 			panic("assumption failed: hdr not split"); \
424 		*mtod((m), u_char *) = *(p);\
425 		*(p) = (i);\
426 		p = mtod((m), u_char *);\
427 		(m)->m_next = (mp)->m_next;\
428 		(mp)->m_next = (m);\
429 		(mp) = (m);\
430 	}\
431     } while (/*CONSTCOND*/ 0)
432 		/*
433 		 * result: IPv6 hbh dest1 rthdr dest2 payload
434 		 * m will point to IPv6 header.  mprev will point to the
435 		 * extension header prior to dest2 (rthdr in the above case).
436 		 */
437 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
438 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
439 		    IPPROTO_DSTOPTS);
440 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
441 		    IPPROTO_ROUTING);
442 
443 		M_CSUM_DATA_IPv6_HL_SET(m->m_pkthdr.csum_data,
444 		    sizeof(struct ip6_hdr) + optlen);
445 
446 #ifdef IPSEC
447 		if (!needipsec)
448 			goto skip_ipsec2;
449 
450 		/*
451 		 * pointers after IPsec headers are not valid any more.
452 		 * other pointers need a great care too.
453 		 * (IPsec routines should not mangle mbufs prior to AH/ESP)
454 		 */
455 		exthdrs.ip6e_dest2 = NULL;
456 
457 	    {
458 		struct ip6_rthdr *rh = NULL;
459 		int segleft_org = 0;
460 		struct ipsec_output_state state;
461 
462 		if (exthdrs.ip6e_rthdr) {
463 			rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
464 			segleft_org = rh->ip6r_segleft;
465 			rh->ip6r_segleft = 0;
466 		}
467 
468 		bzero(&state, sizeof(state));
469 		state.m = m;
470 		error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
471 		    &needipsectun);
472 		m = state.m;
473 		if (error) {
474 			rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
475 			/* mbuf is already reclaimed in ipsec6_output_trans. */
476 			m = NULL;
477 			switch (error) {
478 			case EHOSTUNREACH:
479 			case ENETUNREACH:
480 			case EMSGSIZE:
481 			case ENOBUFS:
482 			case ENOMEM:
483 				break;
484 			default:
485 				printf("ip6_output (ipsec): error code %d\n", error);
486 				/* FALLTHROUGH */
487 			case ENOENT:
488 				/* don't show these error codes to the user */
489 				error = 0;
490 				break;
491 			}
492 			goto bad;
493 		}
494 		if (exthdrs.ip6e_rthdr) {
495 			/* ah6_output doesn't modify mbuf chain */
496 			rh->ip6r_segleft = segleft_org;
497 		}
498 	    }
499 skip_ipsec2:;
500 #endif
501 	}
502 
503 	/*
504 	 * If there is a routing header, replace destination address field
505 	 * with the first hop of the routing header.
506 	 */
507 	if (exthdrs.ip6e_rthdr) {
508 		struct ip6_rthdr *rh;
509 		struct ip6_rthdr0 *rh0;
510 		struct in6_addr *addr;
511 		struct sockaddr_in6 sa;
512 
513 		rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
514 		    struct ip6_rthdr *));
515 		finaldst = ip6->ip6_dst;
516 		switch (rh->ip6r_type) {
517 		case IPV6_RTHDR_TYPE_0:
518 			 rh0 = (struct ip6_rthdr0 *)rh;
519 			 addr = (struct in6_addr *)(rh0 + 1);
520 
521 			 /*
522 			  * construct a sockaddr_in6 form of
523 			  * the first hop.
524 			  *
525 			  * XXX: we may not have enough
526 			  * information about its scope zone;
527 			  * there is no standard API to pass
528 			  * the information from the
529 			  * application.
530 			  */
531 			 sockaddr_in6_init(&sa, addr, 0, 0, 0);
532 			 if ((error = sa6_embedscope(&sa,
533 			     ip6_use_defzone)) != 0) {
534 				 goto bad;
535 			 }
536 			 ip6->ip6_dst = sa.sin6_addr;
537 			 (void)memmove(&addr[0], &addr[1],
538 			     sizeof(struct in6_addr) *
539 			     (rh0->ip6r0_segleft - 1));
540 			 addr[rh0->ip6r0_segleft - 1] = finaldst;
541 			 /* XXX */
542 			 in6_clearscope(addr + rh0->ip6r0_segleft - 1);
543 			 break;
544 		default:	/* is it possible? */
545 			 error = EINVAL;
546 			 goto bad;
547 		}
548 	}
549 
550 	/* Source address validation */
551 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
552 	    (flags & IPV6_UNSPECSRC) == 0) {
553 		error = EOPNOTSUPP;
554 		ip6stat.ip6s_badscope++;
555 		goto bad;
556 	}
557 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
558 		error = EOPNOTSUPP;
559 		ip6stat.ip6s_badscope++;
560 		goto bad;
561 	}
562 
563 	ip6stat.ip6s_localout++;
564 
565 	/*
566 	 * Route packet.
567 	 */
568 	/* initialize cached route */
569 	if (ro == NULL) {
570 		memset(&ip6route, 0, sizeof(ip6route));
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 && opt->ip6po_nextroute.ro_rt != 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 	/* XXX Second if is invariant? */
1175 	if (ro == &ip6route)
1176 		rtcache_free(ro);
1177 	else if (ro_pmtu == &ip6route)
1178 		rtcache_free(ro_pmtu);
1179 
1180 #ifdef IPSEC
1181 	if (sp != NULL)
1182 		key_freesp(sp);
1183 #endif /* IPSEC */
1184 #ifdef FAST_IPSEC
1185 	if (sp != NULL)
1186 		KEY_FREESP(&sp);
1187 #endif /* FAST_IPSEC */
1188 
1189 
1190 	return (error);
1191 
1192 freehdrs:
1193 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1194 	m_freem(exthdrs.ip6e_dest1);
1195 	m_freem(exthdrs.ip6e_rthdr);
1196 	m_freem(exthdrs.ip6e_dest2);
1197 	/* FALLTHROUGH */
1198 bad:
1199 	m_freem(m);
1200 	goto done;
1201 badscope:
1202 	ip6stat.ip6s_badscope++;
1203 	in6_ifstat_inc(origifp, ifs6_out_discard);
1204 	if (error == 0)
1205 		error = EHOSTUNREACH; /* XXX */
1206 	goto bad;
1207 }
1208 
1209 static int
1210 ip6_copyexthdr(struct mbuf **mp, void *hdr, int hlen)
1211 {
1212 	struct mbuf *m;
1213 
1214 	if (hlen > MCLBYTES)
1215 		return (ENOBUFS); /* XXX */
1216 
1217 	MGET(m, M_DONTWAIT, MT_DATA);
1218 	if (!m)
1219 		return (ENOBUFS);
1220 
1221 	if (hlen > MLEN) {
1222 		MCLGET(m, M_DONTWAIT);
1223 		if ((m->m_flags & M_EXT) == 0) {
1224 			m_free(m);
1225 			return (ENOBUFS);
1226 		}
1227 	}
1228 	m->m_len = hlen;
1229 	if (hdr)
1230 		bcopy(hdr, mtod(m, void *), hlen);
1231 
1232 	*mp = m;
1233 	return (0);
1234 }
1235 
1236 /*
1237  * Process a delayed payload checksum calculation.
1238  */
1239 void
1240 in6_delayed_cksum(struct mbuf *m)
1241 {
1242 	uint16_t csum, offset;
1243 
1244 	KASSERT((m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0);
1245 	KASSERT((~m->m_pkthdr.csum_flags & (M_CSUM_UDPv6|M_CSUM_TCPv6)) != 0);
1246 	KASSERT((m->m_pkthdr.csum_flags
1247 	    & (M_CSUM_UDPv4|M_CSUM_TCPv4|M_CSUM_TSOv4)) == 0);
1248 
1249 	offset = M_CSUM_DATA_IPv6_HL(m->m_pkthdr.csum_data);
1250 	csum = in6_cksum(m, 0, offset, m->m_pkthdr.len - offset);
1251 	if (csum == 0 && (m->m_pkthdr.csum_flags & M_CSUM_UDPv6) != 0) {
1252 		csum = 0xffff;
1253 	}
1254 
1255 	offset += M_CSUM_DATA_IPv6_OFFSET(m->m_pkthdr.csum_data);
1256 	if ((offset + sizeof(csum)) > m->m_len) {
1257 		m_copyback(m, offset, sizeof(csum), &csum);
1258 	} else {
1259 		*(uint16_t *)(mtod(m, char *) + offset) = csum;
1260 	}
1261 }
1262 
1263 /*
1264  * Insert jumbo payload option.
1265  */
1266 static int
1267 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1268 {
1269 	struct mbuf *mopt;
1270 	u_int8_t *optbuf;
1271 	u_int32_t v;
1272 
1273 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1274 
1275 	/*
1276 	 * If there is no hop-by-hop options header, allocate new one.
1277 	 * If there is one but it doesn't have enough space to store the
1278 	 * jumbo payload option, allocate a cluster to store the whole options.
1279 	 * Otherwise, use it to store the options.
1280 	 */
1281 	if (exthdrs->ip6e_hbh == 0) {
1282 		MGET(mopt, M_DONTWAIT, MT_DATA);
1283 		if (mopt == 0)
1284 			return (ENOBUFS);
1285 		mopt->m_len = JUMBOOPTLEN;
1286 		optbuf = mtod(mopt, u_int8_t *);
1287 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1288 		exthdrs->ip6e_hbh = mopt;
1289 	} else {
1290 		struct ip6_hbh *hbh;
1291 
1292 		mopt = exthdrs->ip6e_hbh;
1293 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1294 			/*
1295 			 * XXX assumption:
1296 			 * - exthdrs->ip6e_hbh is not referenced from places
1297 			 *   other than exthdrs.
1298 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1299 			 */
1300 			int oldoptlen = mopt->m_len;
1301 			struct mbuf *n;
1302 
1303 			/*
1304 			 * XXX: give up if the whole (new) hbh header does
1305 			 * not fit even in an mbuf cluster.
1306 			 */
1307 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1308 				return (ENOBUFS);
1309 
1310 			/*
1311 			 * As a consequence, we must always prepare a cluster
1312 			 * at this point.
1313 			 */
1314 			MGET(n, M_DONTWAIT, MT_DATA);
1315 			if (n) {
1316 				MCLGET(n, M_DONTWAIT);
1317 				if ((n->m_flags & M_EXT) == 0) {
1318 					m_freem(n);
1319 					n = NULL;
1320 				}
1321 			}
1322 			if (!n)
1323 				return (ENOBUFS);
1324 			n->m_len = oldoptlen + JUMBOOPTLEN;
1325 			bcopy(mtod(mopt, void *), mtod(n, void *),
1326 			    oldoptlen);
1327 			optbuf = mtod(n, u_int8_t *) + oldoptlen;
1328 			m_freem(mopt);
1329 			mopt = exthdrs->ip6e_hbh = n;
1330 		} else {
1331 			optbuf = mtod(mopt, u_int8_t *) + mopt->m_len;
1332 			mopt->m_len += JUMBOOPTLEN;
1333 		}
1334 		optbuf[0] = IP6OPT_PADN;
1335 		optbuf[1] = 0;
1336 
1337 		/*
1338 		 * Adjust the header length according to the pad and
1339 		 * the jumbo payload option.
1340 		 */
1341 		hbh = mtod(mopt, struct ip6_hbh *);
1342 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1343 	}
1344 
1345 	/* fill in the option. */
1346 	optbuf[2] = IP6OPT_JUMBO;
1347 	optbuf[3] = 4;
1348 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1349 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1350 
1351 	/* finally, adjust the packet header length */
1352 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1353 
1354 	return (0);
1355 #undef JUMBOOPTLEN
1356 }
1357 
1358 /*
1359  * Insert fragment header and copy unfragmentable header portions.
1360  */
1361 static int
1362 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1363 	struct ip6_frag **frghdrp)
1364 {
1365 	struct mbuf *n, *mlast;
1366 
1367 	if (hlen > sizeof(struct ip6_hdr)) {
1368 		n = m_copym(m0, sizeof(struct ip6_hdr),
1369 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1370 		if (n == 0)
1371 			return (ENOBUFS);
1372 		m->m_next = n;
1373 	} else
1374 		n = m;
1375 
1376 	/* Search for the last mbuf of unfragmentable part. */
1377 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1378 		;
1379 
1380 	if ((mlast->m_flags & M_EXT) == 0 &&
1381 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1382 		/* use the trailing space of the last mbuf for the fragment hdr */
1383 		*frghdrp = (struct ip6_frag *)(mtod(mlast, char *) +
1384 		    mlast->m_len);
1385 		mlast->m_len += sizeof(struct ip6_frag);
1386 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1387 	} else {
1388 		/* allocate a new mbuf for the fragment header */
1389 		struct mbuf *mfrg;
1390 
1391 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1392 		if (mfrg == 0)
1393 			return (ENOBUFS);
1394 		mfrg->m_len = sizeof(struct ip6_frag);
1395 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1396 		mlast->m_next = mfrg;
1397 	}
1398 
1399 	return (0);
1400 }
1401 
1402 static int
1403 ip6_getpmtu(struct route *ro_pmtu, struct route *ro, struct ifnet *ifp,
1404     const struct in6_addr *dst, u_long *mtup, int *alwaysfragp)
1405 {
1406 	u_int32_t mtu = 0;
1407 	int alwaysfrag = 0;
1408 	int error = 0;
1409 
1410 	if (ro_pmtu != ro) {
1411 		union {
1412 			struct sockaddr		dst;
1413 			struct sockaddr_in6	dst6;
1414 		} u;
1415 
1416 		/* The first hop and the final destination may differ. */
1417 		sockaddr_in6_init(&u.dst6, dst, 0, 0, 0);
1418 		rtcache_lookup(ro_pmtu, &u.dst);
1419 	}
1420 	if (ro_pmtu->ro_rt != NULL) {
1421 		u_int32_t ifmtu;
1422 
1423 		if (ifp == NULL)
1424 			ifp = ro_pmtu->ro_rt->rt_ifp;
1425 		ifmtu = IN6_LINKMTU(ifp);
1426 		mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1427 		if (mtu == 0)
1428 			mtu = ifmtu;
1429 		else if (mtu < IPV6_MMTU) {
1430 			/*
1431 			 * RFC2460 section 5, last paragraph:
1432 			 * if we record ICMPv6 too big message with
1433 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1434 			 * or smaller, with fragment header attached.
1435 			 * (fragment header is needed regardless from the
1436 			 * packet size, for translators to identify packets)
1437 			 */
1438 			alwaysfrag = 1;
1439 			mtu = IPV6_MMTU;
1440 		} else if (mtu > ifmtu) {
1441 			/*
1442 			 * The MTU on the route is larger than the MTU on
1443 			 * the interface!  This shouldn't happen, unless the
1444 			 * MTU of the interface has been changed after the
1445 			 * interface was brought up.  Change the MTU in the
1446 			 * route to match the interface MTU (as long as the
1447 			 * field isn't locked).
1448 			 */
1449 			mtu = ifmtu;
1450 			if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU))
1451 				ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1452 		}
1453 	} else if (ifp) {
1454 		mtu = IN6_LINKMTU(ifp);
1455 	} else
1456 		error = EHOSTUNREACH; /* XXX */
1457 
1458 	*mtup = mtu;
1459 	if (alwaysfragp)
1460 		*alwaysfragp = alwaysfrag;
1461 	return (error);
1462 }
1463 
1464 /*
1465  * IP6 socket option processing.
1466  */
1467 int
1468 ip6_ctloutput(int op, struct socket *so, int level, int optname,
1469     struct mbuf **mp)
1470 {
1471 	int privileged, optdatalen, uproto;
1472 	void *optdata;
1473 	struct in6pcb *in6p = sotoin6pcb(so);
1474 	struct mbuf *m = *mp;
1475 	int error, optval;
1476 	int optlen;
1477 	struct lwp *l = curlwp;	/* XXX */
1478 
1479 	optlen = m ? m->m_len : 0;
1480 	error = optval = 0;
1481 	privileged = (l == 0 || kauth_authorize_generic(l->l_cred,
1482 	    KAUTH_GENERIC_ISSUSER, NULL)) ? 0 : 1;
1483 	uproto = (int)so->so_proto->pr_protocol;
1484 
1485 	if (level != IPPROTO_IPV6) {
1486 		if (op == PRCO_SETOPT && *mp)
1487 			(void)m_free(*mp);
1488 		return ENOPROTOOPT;
1489 	}
1490 	switch (op) {
1491 	case PRCO_SETOPT:
1492 		switch (optname) {
1493 #ifdef RFC2292
1494 		case IPV6_2292PKTOPTIONS:
1495 			/* m is freed in ip6_pcbopts */
1496 			error = ip6_pcbopts(&in6p->in6p_outputopts,
1497 			    m, so);
1498 			break;
1499 #endif
1500 
1501 		/*
1502 		 * Use of some Hop-by-Hop options or some
1503 		 * Destination options, might require special
1504 		 * privilege.  That is, normal applications
1505 		 * (without special privilege) might be forbidden
1506 		 * from setting certain options in outgoing packets,
1507 		 * and might never see certain options in received
1508 		 * packets. [RFC 2292 Section 6]
1509 		 * KAME specific note:
1510 		 *  KAME prevents non-privileged users from sending or
1511 		 *  receiving ANY hbh/dst options in order to avoid
1512 		 *  overhead of parsing options in the kernel.
1513 		 */
1514 		case IPV6_RECVHOPOPTS:
1515 		case IPV6_RECVDSTOPTS:
1516 		case IPV6_RECVRTHDRDSTOPTS:
1517 			if (!privileged) {
1518 				error = EPERM;
1519 				break;
1520 			}
1521 			/* FALLTHROUGH */
1522 		case IPV6_UNICAST_HOPS:
1523 		case IPV6_HOPLIMIT:
1524 		case IPV6_FAITH:
1525 
1526 		case IPV6_RECVPKTINFO:
1527 		case IPV6_RECVHOPLIMIT:
1528 		case IPV6_RECVRTHDR:
1529 		case IPV6_RECVPATHMTU:
1530 		case IPV6_RECVTCLASS:
1531 		case IPV6_V6ONLY:
1532 			if (optlen != sizeof(int)) {
1533 				error = EINVAL;
1534 				break;
1535 			}
1536 			optval = *mtod(m, int *);
1537 			switch (optname) {
1538 
1539 			case IPV6_UNICAST_HOPS:
1540 				if (optval < -1 || optval >= 256)
1541 					error = EINVAL;
1542 				else {
1543 					/* -1 = kernel default */
1544 					in6p->in6p_hops = optval;
1545 				}
1546 				break;
1547 #define OPTSET(bit) \
1548 do { \
1549 if (optval) \
1550 	in6p->in6p_flags |= (bit); \
1551 else \
1552 	in6p->in6p_flags &= ~(bit); \
1553 } while (/*CONSTCOND*/ 0)
1554 
1555 #ifdef RFC2292
1556 #define OPTSET2292(bit) 			\
1557 do { 						\
1558 in6p->in6p_flags |= IN6P_RFC2292; 	\
1559 if (optval) 				\
1560 	in6p->in6p_flags |= (bit); 	\
1561 else 					\
1562 	in6p->in6p_flags &= ~(bit); 	\
1563 } while (/*CONSTCOND*/ 0)
1564 #endif
1565 
1566 #define OPTBIT(bit) (in6p->in6p_flags & (bit) ? 1 : 0)
1567 
1568 			case IPV6_RECVPKTINFO:
1569 #ifdef RFC2292
1570 				/* cannot mix with RFC2292 */
1571 				if (OPTBIT(IN6P_RFC2292)) {
1572 					error = EINVAL;
1573 					break;
1574 				}
1575 #endif
1576 				OPTSET(IN6P_PKTINFO);
1577 				break;
1578 
1579 			case IPV6_HOPLIMIT:
1580 			{
1581 				struct ip6_pktopts **optp;
1582 
1583 #ifdef RFC2292
1584 				/* cannot mix with RFC2292 */
1585 				if (OPTBIT(IN6P_RFC2292)) {
1586 					error = EINVAL;
1587 					break;
1588 				}
1589 #endif
1590 				optp = &in6p->in6p_outputopts;
1591 				error = ip6_pcbopt(IPV6_HOPLIMIT,
1592 						   (u_char *)&optval,
1593 						   sizeof(optval),
1594 						   optp,
1595 						   privileged, uproto);
1596 				break;
1597 			}
1598 
1599 			case IPV6_RECVHOPLIMIT:
1600 #ifdef RFC2292
1601 				/* cannot mix with RFC2292 */
1602 				if (OPTBIT(IN6P_RFC2292)) {
1603 					error = EINVAL;
1604 					break;
1605 				}
1606 #endif
1607 				OPTSET(IN6P_HOPLIMIT);
1608 				break;
1609 
1610 			case IPV6_RECVHOPOPTS:
1611 #ifdef RFC2292
1612 				/* cannot mix with RFC2292 */
1613 				if (OPTBIT(IN6P_RFC2292)) {
1614 					error = EINVAL;
1615 					break;
1616 				}
1617 #endif
1618 				OPTSET(IN6P_HOPOPTS);
1619 				break;
1620 
1621 			case IPV6_RECVDSTOPTS:
1622 #ifdef RFC2292
1623 				/* cannot mix with RFC2292 */
1624 				if (OPTBIT(IN6P_RFC2292)) {
1625 					error = EINVAL;
1626 					break;
1627 				}
1628 #endif
1629 				OPTSET(IN6P_DSTOPTS);
1630 				break;
1631 
1632 			case IPV6_RECVRTHDRDSTOPTS:
1633 #ifdef RFC2292
1634 				/* cannot mix with RFC2292 */
1635 				if (OPTBIT(IN6P_RFC2292)) {
1636 					error = EINVAL;
1637 					break;
1638 				}
1639 #endif
1640 				OPTSET(IN6P_RTHDRDSTOPTS);
1641 				break;
1642 
1643 			case IPV6_RECVRTHDR:
1644 #ifdef RFC2292
1645 				/* cannot mix with RFC2292 */
1646 				if (OPTBIT(IN6P_RFC2292)) {
1647 					error = EINVAL;
1648 					break;
1649 				}
1650 #endif
1651 				OPTSET(IN6P_RTHDR);
1652 				break;
1653 
1654 			case IPV6_FAITH:
1655 				OPTSET(IN6P_FAITH);
1656 				break;
1657 
1658 			case IPV6_RECVPATHMTU:
1659 				/*
1660 				 * We ignore this option for TCP
1661 				 * sockets.
1662 				 * (RFC3542 leaves this case
1663 				 * unspecified.)
1664 				 */
1665 				if (uproto != IPPROTO_TCP)
1666 					OPTSET(IN6P_MTU);
1667 				break;
1668 
1669 			case IPV6_V6ONLY:
1670 				/*
1671 				 * make setsockopt(IPV6_V6ONLY)
1672 				 * available only prior to bind(2).
1673 				 * see ipng mailing list, Jun 22 2001.
1674 				 */
1675 				if (in6p->in6p_lport ||
1676 				    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1677 					error = EINVAL;
1678 					break;
1679 				}
1680 #ifdef INET6_BINDV6ONLY
1681 				if (!optval)
1682 					error = EINVAL;
1683 #else
1684 				OPTSET(IN6P_IPV6_V6ONLY);
1685 #endif
1686 				break;
1687 			case IPV6_RECVTCLASS:
1688 #ifdef RFC2292
1689 				/* cannot mix with RFC2292 XXX */
1690 				if (OPTBIT(IN6P_RFC2292)) {
1691 					error = EINVAL;
1692 					break;
1693 				}
1694 #endif
1695 				OPTSET(IN6P_TCLASS);
1696 				break;
1697 
1698 			}
1699 			break;
1700 
1701 		case IPV6_OTCLASS:
1702 		{
1703 			struct ip6_pktopts **optp;
1704 			u_int8_t tclass;
1705 
1706 			if (optlen != sizeof(tclass)) {
1707 				error = EINVAL;
1708 				break;
1709 			}
1710 			tclass = *mtod(m, u_int8_t *);
1711 			optp = &in6p->in6p_outputopts;
1712 			error = ip6_pcbopt(optname,
1713 					   (u_char *)&tclass,
1714 					   sizeof(tclass),
1715 					   optp,
1716 					   privileged, uproto);
1717 			break;
1718 		}
1719 
1720 		case IPV6_TCLASS:
1721 		case IPV6_DONTFRAG:
1722 		case IPV6_USE_MIN_MTU:
1723 			if (optlen != sizeof(optval)) {
1724 				error = EINVAL;
1725 				break;
1726 			}
1727 			optval = *mtod(m, int *);
1728 			{
1729 				struct ip6_pktopts **optp;
1730 				optp = &in6p->in6p_outputopts;
1731 				error = ip6_pcbopt(optname,
1732 						   (u_char *)&optval,
1733 						   sizeof(optval),
1734 						   optp,
1735 						   privileged, uproto);
1736 				break;
1737 			}
1738 
1739 #ifdef RFC2292
1740 		case IPV6_2292PKTINFO:
1741 		case IPV6_2292HOPLIMIT:
1742 		case IPV6_2292HOPOPTS:
1743 		case IPV6_2292DSTOPTS:
1744 		case IPV6_2292RTHDR:
1745 			/* RFC 2292 */
1746 			if (optlen != sizeof(int)) {
1747 				error = EINVAL;
1748 				break;
1749 			}
1750 			optval = *mtod(m, int *);
1751 			switch (optname) {
1752 			case IPV6_2292PKTINFO:
1753 				OPTSET2292(IN6P_PKTINFO);
1754 				break;
1755 			case IPV6_2292HOPLIMIT:
1756 				OPTSET2292(IN6P_HOPLIMIT);
1757 				break;
1758 			case IPV6_2292HOPOPTS:
1759 				/*
1760 				 * Check super-user privilege.
1761 				 * See comments for IPV6_RECVHOPOPTS.
1762 				 */
1763 				if (!privileged)
1764 					return (EPERM);
1765 				OPTSET2292(IN6P_HOPOPTS);
1766 				break;
1767 			case IPV6_2292DSTOPTS:
1768 				if (!privileged)
1769 					return (EPERM);
1770 				OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1771 				break;
1772 			case IPV6_2292RTHDR:
1773 				OPTSET2292(IN6P_RTHDR);
1774 				break;
1775 			}
1776 			break;
1777 #endif
1778 		case IPV6_PKTINFO:
1779 		case IPV6_HOPOPTS:
1780 		case IPV6_RTHDR:
1781 		case IPV6_DSTOPTS:
1782 		case IPV6_RTHDRDSTOPTS:
1783 		case IPV6_NEXTHOP:
1784 		{
1785 			/* new advanced API (RFC3542) */
1786 			u_char *optbuf;
1787 			int optbuflen;
1788 			struct ip6_pktopts **optp;
1789 			if (!m) {
1790 				error = EINVAL;
1791 				break;
1792 			}
1793 
1794 #ifdef RFC2292
1795 			/* cannot mix with RFC2292 */
1796 			if (OPTBIT(IN6P_RFC2292)) {
1797 				error = EINVAL;
1798 				break;
1799 			}
1800 #endif
1801 
1802 			if (m && m->m_next) {
1803 				error = EINVAL;	/* XXX */
1804 				break;
1805 			}
1806 
1807 			optbuf = mtod(m, u_char *);
1808 			optbuflen = m->m_len;
1809 			optp = &in6p->in6p_outputopts;
1810 			error = ip6_pcbopt(optname, optbuf, optbuflen,
1811 			    optp, privileged, uproto);
1812 			break;
1813 		}
1814 #undef OPTSET
1815 
1816 		case IPV6_MULTICAST_IF:
1817 		case IPV6_MULTICAST_HOPS:
1818 		case IPV6_MULTICAST_LOOP:
1819 		case IPV6_JOIN_GROUP:
1820 		case IPV6_LEAVE_GROUP:
1821 			error = ip6_setmoptions(optname,
1822 			    &in6p->in6p_moptions, m);
1823 			break;
1824 
1825 		case IPV6_PORTRANGE:
1826 			if (!m) {
1827 				error = EINVAL;
1828 				break;
1829 			}
1830 			optval = *mtod(m, int *);
1831 
1832 			switch (optval) {
1833 			case IPV6_PORTRANGE_DEFAULT:
1834 				in6p->in6p_flags &= ~(IN6P_LOWPORT);
1835 				in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1836 				break;
1837 
1838 			case IPV6_PORTRANGE_HIGH:
1839 				in6p->in6p_flags &= ~(IN6P_LOWPORT);
1840 				in6p->in6p_flags |= IN6P_HIGHPORT;
1841 				break;
1842 
1843 			case IPV6_PORTRANGE_LOW:
1844 				in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1845 				in6p->in6p_flags |= IN6P_LOWPORT;
1846 				break;
1847 
1848 			default:
1849 				error = EINVAL;
1850 				break;
1851 			}
1852 			break;
1853 
1854 
1855 #if defined(IPSEC) || defined(FAST_IPSEC)
1856 		case IPV6_IPSEC_POLICY:
1857 		{
1858 			void *req = NULL;
1859 			size_t len = 0;
1860 			if (m) {
1861 				req = mtod(m, void *);
1862 				len = m->m_len;
1863 			}
1864 			error = ipsec6_set_policy(in6p, optname, req,
1865 						  len, privileged);
1866 		}
1867 			break;
1868 #endif /* IPSEC */
1869 
1870 		default:
1871 			error = ENOPROTOOPT;
1872 			break;
1873 		}
1874 		if (m)
1875 			(void)m_free(m);
1876 		break;
1877 
1878 	case PRCO_GETOPT:
1879 		switch (optname) {
1880 #ifdef RFC2292
1881 		case IPV6_2292PKTOPTIONS:
1882 			/*
1883 			 * RFC3542 (effectively) deprecated the
1884 			 * semantics of the 2292-style pktoptions.
1885 			 * Since it was not reliable in nature (i.e.,
1886 			 * applications had to expect the lack of some
1887 			 * information after all), it would make sense
1888 			 * to simplify this part by always returning
1889 			 * empty data.
1890 			 */
1891 			*mp = m_get(M_WAIT, MT_SOOPTS);
1892 			(*mp)->m_len = 0;
1893 			break;
1894 #endif
1895 
1896 		case IPV6_RECVHOPOPTS:
1897 		case IPV6_RECVDSTOPTS:
1898 		case IPV6_RECVRTHDRDSTOPTS:
1899 		case IPV6_UNICAST_HOPS:
1900 		case IPV6_RECVPKTINFO:
1901 		case IPV6_RECVHOPLIMIT:
1902 		case IPV6_RECVRTHDR:
1903 		case IPV6_RECVPATHMTU:
1904 
1905 		case IPV6_FAITH:
1906 		case IPV6_V6ONLY:
1907 		case IPV6_PORTRANGE:
1908 		case IPV6_RECVTCLASS:
1909 			switch (optname) {
1910 
1911 			case IPV6_RECVHOPOPTS:
1912 				optval = OPTBIT(IN6P_HOPOPTS);
1913 				break;
1914 
1915 			case IPV6_RECVDSTOPTS:
1916 				optval = OPTBIT(IN6P_DSTOPTS);
1917 				break;
1918 
1919 			case IPV6_RECVRTHDRDSTOPTS:
1920 				optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1921 				break;
1922 
1923 			case IPV6_UNICAST_HOPS:
1924 				optval = in6p->in6p_hops;
1925 				break;
1926 
1927 			case IPV6_RECVPKTINFO:
1928 				optval = OPTBIT(IN6P_PKTINFO);
1929 				break;
1930 
1931 			case IPV6_RECVHOPLIMIT:
1932 				optval = OPTBIT(IN6P_HOPLIMIT);
1933 				break;
1934 
1935 			case IPV6_RECVRTHDR:
1936 				optval = OPTBIT(IN6P_RTHDR);
1937 				break;
1938 
1939 			case IPV6_RECVPATHMTU:
1940 				optval = OPTBIT(IN6P_MTU);
1941 				break;
1942 
1943 			case IPV6_FAITH:
1944 				optval = OPTBIT(IN6P_FAITH);
1945 				break;
1946 
1947 			case IPV6_V6ONLY:
1948 				optval = OPTBIT(IN6P_IPV6_V6ONLY);
1949 				break;
1950 
1951 			case IPV6_PORTRANGE:
1952 			    {
1953 				int flags;
1954 				flags = in6p->in6p_flags;
1955 				if (flags & IN6P_HIGHPORT)
1956 					optval = IPV6_PORTRANGE_HIGH;
1957 				else if (flags & IN6P_LOWPORT)
1958 					optval = IPV6_PORTRANGE_LOW;
1959 				else
1960 					optval = 0;
1961 				break;
1962 			    }
1963 			case IPV6_RECVTCLASS:
1964 				optval = OPTBIT(IN6P_TCLASS);
1965 				break;
1966 
1967 			}
1968 			if (error)
1969 				break;
1970 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1971 			m->m_len = sizeof(int);
1972 			*mtod(m, int *) = optval;
1973 			break;
1974 
1975 		case IPV6_PATHMTU:
1976 		    {
1977 			u_long pmtu = 0;
1978 			struct ip6_mtuinfo mtuinfo;
1979 			struct route *ro = &in6p->in6p_route;
1980 
1981 			if (!(so->so_state & SS_ISCONNECTED))
1982 				return (ENOTCONN);
1983 			/*
1984 			 * XXX: we dot not consider the case of source
1985 			 * routing, or optional information to specify
1986 			 * the outgoing interface.
1987 			 */
1988 			error = ip6_getpmtu(ro, NULL, NULL,
1989 			    &in6p->in6p_faddr, &pmtu, NULL);
1990 			if (error)
1991 				break;
1992 			if (pmtu > IPV6_MAXPACKET)
1993 				pmtu = IPV6_MAXPACKET;
1994 
1995 			memset(&mtuinfo, 0, sizeof(mtuinfo));
1996 			mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1997 			optdata = (void *)&mtuinfo;
1998 			optdatalen = sizeof(mtuinfo);
1999 			if (optdatalen > MCLBYTES)
2000 				return (EMSGSIZE); /* XXX */
2001 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
2002 			if (optdatalen > MLEN)
2003 				MCLGET(m, M_WAIT);
2004 			m->m_len = optdatalen;
2005 			memcpy(mtod(m, void *), optdata, optdatalen);
2006 			break;
2007 		    }
2008 
2009 #ifdef RFC2292
2010 		case IPV6_2292PKTINFO:
2011 		case IPV6_2292HOPLIMIT:
2012 		case IPV6_2292HOPOPTS:
2013 		case IPV6_2292RTHDR:
2014 		case IPV6_2292DSTOPTS:
2015 			switch (optname) {
2016 			case IPV6_2292PKTINFO:
2017 				optval = OPTBIT(IN6P_PKTINFO);
2018 				break;
2019 			case IPV6_2292HOPLIMIT:
2020 				optval = OPTBIT(IN6P_HOPLIMIT);
2021 				break;
2022 			case IPV6_2292HOPOPTS:
2023 				optval = OPTBIT(IN6P_HOPOPTS);
2024 				break;
2025 			case IPV6_2292RTHDR:
2026 				optval = OPTBIT(IN6P_RTHDR);
2027 				break;
2028 			case IPV6_2292DSTOPTS:
2029 				optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2030 				break;
2031 			}
2032 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
2033 			m->m_len = sizeof(int);
2034 			*mtod(m, int *) = optval;
2035 			break;
2036 #endif
2037 		case IPV6_PKTINFO:
2038 		case IPV6_HOPOPTS:
2039 		case IPV6_RTHDR:
2040 		case IPV6_DSTOPTS:
2041 		case IPV6_RTHDRDSTOPTS:
2042 		case IPV6_NEXTHOP:
2043 		case IPV6_OTCLASS:
2044 		case IPV6_TCLASS:
2045 		case IPV6_DONTFRAG:
2046 		case IPV6_USE_MIN_MTU:
2047 			error = ip6_getpcbopt(in6p->in6p_outputopts,
2048 			    optname, mp);
2049 			break;
2050 
2051 		case IPV6_MULTICAST_IF:
2052 		case IPV6_MULTICAST_HOPS:
2053 		case IPV6_MULTICAST_LOOP:
2054 		case IPV6_JOIN_GROUP:
2055 		case IPV6_LEAVE_GROUP:
2056 			error = ip6_getmoptions(optname,
2057 			    in6p->in6p_moptions, mp);
2058 			break;
2059 
2060 #if defined(IPSEC) || defined(FAST_IPSEC)
2061 		case IPV6_IPSEC_POLICY:
2062 		    {
2063 			void *req = NULL;
2064 			size_t len = 0;
2065 			if (m) {
2066 				req = mtod(m, void *);
2067 				len = m->m_len;
2068 			}
2069 			error = ipsec6_get_policy(in6p, req, len, mp);
2070 			break;
2071 		    }
2072 #endif /* IPSEC */
2073 
2074 
2075 
2076 
2077 		default:
2078 			error = ENOPROTOOPT;
2079 			break;
2080 		}
2081 		break;
2082 	}
2083 	return (error);
2084 }
2085 
2086 int
2087 ip6_raw_ctloutput(int op, struct socket *so, int level, int optname,
2088 	struct mbuf **mp)
2089 {
2090 	int error = 0, optval, optlen;
2091 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2092 	struct in6pcb *in6p = sotoin6pcb(so);
2093 	struct mbuf *m = *mp;
2094 
2095 	optlen = m ? m->m_len : 0;
2096 
2097 	if (level != IPPROTO_IPV6) {
2098 		if (op == PRCO_SETOPT && *mp)
2099 			(void)m_free(*mp);
2100 		return ENOPROTOOPT;
2101 	}
2102 
2103 	switch (optname) {
2104 	case IPV6_CHECKSUM:
2105 		/*
2106 		 * For ICMPv6 sockets, no modification allowed for checksum
2107 		 * offset, permit "no change" values to help existing apps.
2108 		 *
2109 		 * XXX RFC3542 says: "An attempt to set IPV6_CHECKSUM
2110 		 * for an ICMPv6 socket will fail."  The current
2111 		 * behavior does not meet RFC3542.
2112 		 */
2113 		switch (op) {
2114 		case PRCO_SETOPT:
2115 			if (optlen != sizeof(int)) {
2116 				error = EINVAL;
2117 				break;
2118 			}
2119 			optval = *mtod(m, int *);
2120 			if ((optval % 2) != 0) {
2121 				/* the API assumes even offset values */
2122 				error = EINVAL;
2123 			} else if (so->so_proto->pr_protocol ==
2124 			    IPPROTO_ICMPV6) {
2125 				if (optval != icmp6off)
2126 					error = EINVAL;
2127 			} else
2128 				in6p->in6p_cksum = optval;
2129 			break;
2130 
2131 		case PRCO_GETOPT:
2132 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2133 				optval = icmp6off;
2134 			else
2135 				optval = in6p->in6p_cksum;
2136 
2137 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
2138 			m->m_len = sizeof(int);
2139 			*mtod(m, int *) = optval;
2140 			break;
2141 
2142 		default:
2143 			error = EINVAL;
2144 			break;
2145 		}
2146 		break;
2147 
2148 	default:
2149 		error = ENOPROTOOPT;
2150 		break;
2151 	}
2152 
2153 	if (op == PRCO_SETOPT && m)
2154 		(void)m_free(m);
2155 
2156 	return (error);
2157 }
2158 
2159 #ifdef RFC2292
2160 /*
2161  * Set up IP6 options in pcb for insertion in output packets or
2162  * specifying behavior of outgoing packets.
2163  */
2164 static int
2165 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, struct socket *so)
2166 {
2167 	struct ip6_pktopts *opt = *pktopt;
2168 	int error = 0;
2169 	struct lwp *l = curlwp;	/* XXX */
2170 	int priv = 0;
2171 
2172 	/* turn off any old options. */
2173 	if (opt) {
2174 #ifdef DIAGNOSTIC
2175 	    if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2176 		opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2177 		opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2178 		    printf("ip6_pcbopts: all specified options are cleared.\n");
2179 #endif
2180 		ip6_clearpktopts(opt, -1);
2181 	} else
2182 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2183 	*pktopt = NULL;
2184 
2185 	if (!m || m->m_len == 0) {
2186 		/*
2187 		 * Only turning off any previous options, regardless of
2188 		 * whether the opt is just created or given.
2189 		 */
2190 		free(opt, M_IP6OPT);
2191 		return (0);
2192 	}
2193 
2194 	/*  set options specified by user. */
2195 	if (l && !kauth_authorize_generic(l->l_cred, KAUTH_GENERIC_ISSUSER,
2196 	    NULL))
2197 		priv = 1;
2198 	if ((error = ip6_setpktopts(m, opt, NULL, priv,
2199 	    so->so_proto->pr_protocol)) != 0) {
2200 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2201 		free(opt, M_IP6OPT);
2202 		return (error);
2203 	}
2204 	*pktopt = opt;
2205 	return (0);
2206 }
2207 #endif
2208 
2209 /*
2210  * initialize ip6_pktopts.  beware that there are non-zero default values in
2211  * the struct.
2212  */
2213 void
2214 ip6_initpktopts(struct ip6_pktopts *opt)
2215 {
2216 
2217 	memset(opt, 0, sizeof(*opt));
2218 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2219 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2220 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2221 }
2222 
2223 #define sin6tosa(sin6)	((struct sockaddr *)(sin6)) /* XXX */
2224 static int
2225 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2226     int priv, int uproto)
2227 {
2228 	struct ip6_pktopts *opt;
2229 
2230 	if (*pktopt == NULL) {
2231 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2232 		    M_WAITOK);
2233 		ip6_initpktopts(*pktopt);
2234 	}
2235 	opt = *pktopt;
2236 
2237 	return (ip6_setpktopt(optname, buf, len, opt, priv, 1, 0, uproto));
2238 }
2239 
2240 static int
2241 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct mbuf **mp)
2242 {
2243 	void *optdata = NULL;
2244 	int optdatalen = 0;
2245 	struct ip6_ext *ip6e;
2246 	int error = 0;
2247 	struct in6_pktinfo null_pktinfo;
2248 	int deftclass = 0, on;
2249 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2250 	struct mbuf *m;
2251 
2252 	switch (optname) {
2253 	case IPV6_PKTINFO:
2254 		if (pktopt && pktopt->ip6po_pktinfo)
2255 			optdata = (void *)pktopt->ip6po_pktinfo;
2256 		else {
2257 			/* XXX: we don't have to do this every time... */
2258 			memset(&null_pktinfo, 0, sizeof(null_pktinfo));
2259 			optdata = (void *)&null_pktinfo;
2260 		}
2261 		optdatalen = sizeof(struct in6_pktinfo);
2262 		break;
2263 	case IPV6_OTCLASS:
2264 		/* XXX */
2265 		return (EINVAL);
2266 	case IPV6_TCLASS:
2267 		if (pktopt && pktopt->ip6po_tclass >= 0)
2268 			optdata = (void *)&pktopt->ip6po_tclass;
2269 		else
2270 			optdata = (void *)&deftclass;
2271 		optdatalen = sizeof(int);
2272 		break;
2273 	case IPV6_HOPOPTS:
2274 		if (pktopt && pktopt->ip6po_hbh) {
2275 			optdata = (void *)pktopt->ip6po_hbh;
2276 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2277 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2278 		}
2279 		break;
2280 	case IPV6_RTHDR:
2281 		if (pktopt && pktopt->ip6po_rthdr) {
2282 			optdata = (void *)pktopt->ip6po_rthdr;
2283 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2284 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2285 		}
2286 		break;
2287 	case IPV6_RTHDRDSTOPTS:
2288 		if (pktopt && pktopt->ip6po_dest1) {
2289 			optdata = (void *)pktopt->ip6po_dest1;
2290 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2291 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2292 		}
2293 		break;
2294 	case IPV6_DSTOPTS:
2295 		if (pktopt && pktopt->ip6po_dest2) {
2296 			optdata = (void *)pktopt->ip6po_dest2;
2297 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2298 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2299 		}
2300 		break;
2301 	case IPV6_NEXTHOP:
2302 		if (pktopt && pktopt->ip6po_nexthop) {
2303 			optdata = (void *)pktopt->ip6po_nexthop;
2304 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2305 		}
2306 		break;
2307 	case IPV6_USE_MIN_MTU:
2308 		if (pktopt)
2309 			optdata = (void *)&pktopt->ip6po_minmtu;
2310 		else
2311 			optdata = (void *)&defminmtu;
2312 		optdatalen = sizeof(int);
2313 		break;
2314 	case IPV6_DONTFRAG:
2315 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2316 			on = 1;
2317 		else
2318 			on = 0;
2319 		optdata = (void *)&on;
2320 		optdatalen = sizeof(on);
2321 		break;
2322 	default:		/* should not happen */
2323 #ifdef DIAGNOSTIC
2324 		panic("ip6_getpcbopt: unexpected option\n");
2325 #endif
2326 		return (ENOPROTOOPT);
2327 	}
2328 
2329 	if (optdatalen > MCLBYTES)
2330 		return (EMSGSIZE); /* XXX */
2331 	*mp = m = m_get(M_WAIT, MT_SOOPTS);
2332 	if (optdatalen > MLEN)
2333 		MCLGET(m, M_WAIT);
2334 	m->m_len = optdatalen;
2335 	if (optdatalen)
2336 		memcpy(mtod(m, void *), optdata, optdatalen);
2337 
2338 	return (error);
2339 }
2340 
2341 void
2342 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2343 {
2344 	if (optname == -1 || optname == IPV6_PKTINFO) {
2345 		if (pktopt->ip6po_pktinfo)
2346 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2347 		pktopt->ip6po_pktinfo = NULL;
2348 	}
2349 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2350 		pktopt->ip6po_hlim = -1;
2351 	if (optname == -1 || optname == IPV6_TCLASS)
2352 		pktopt->ip6po_tclass = -1;
2353 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2354 		rtcache_free(&pktopt->ip6po_nextroute);
2355 		if (pktopt->ip6po_nexthop)
2356 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2357 		pktopt->ip6po_nexthop = NULL;
2358 	}
2359 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2360 		if (pktopt->ip6po_hbh)
2361 			free(pktopt->ip6po_hbh, M_IP6OPT);
2362 		pktopt->ip6po_hbh = NULL;
2363 	}
2364 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2365 		if (pktopt->ip6po_dest1)
2366 			free(pktopt->ip6po_dest1, M_IP6OPT);
2367 		pktopt->ip6po_dest1 = NULL;
2368 	}
2369 	if (optname == -1 || optname == IPV6_RTHDR) {
2370 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2371 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2372 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2373 		rtcache_free(&pktopt->ip6po_route);
2374 	}
2375 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2376 		if (pktopt->ip6po_dest2)
2377 			free(pktopt->ip6po_dest2, M_IP6OPT);
2378 		pktopt->ip6po_dest2 = NULL;
2379 	}
2380 }
2381 
2382 #define PKTOPT_EXTHDRCPY(type) 					\
2383 do {								\
2384 	if (src->type) {					\
2385 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2386 		dst->type = malloc(hlen, M_IP6OPT, canwait);	\
2387 		if (dst->type == NULL && canwait == M_NOWAIT)	\
2388 			goto bad;				\
2389 		memcpy(dst->type, src->type, hlen);		\
2390 	}							\
2391 } while (/*CONSTCOND*/ 0)
2392 
2393 static int
2394 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2395 {
2396 	dst->ip6po_hlim = src->ip6po_hlim;
2397 	dst->ip6po_tclass = src->ip6po_tclass;
2398 	dst->ip6po_flags = src->ip6po_flags;
2399 	if (src->ip6po_pktinfo) {
2400 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2401 		    M_IP6OPT, canwait);
2402 		if (dst->ip6po_pktinfo == NULL && canwait == M_NOWAIT)
2403 			goto bad;
2404 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2405 	}
2406 	if (src->ip6po_nexthop) {
2407 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2408 		    M_IP6OPT, canwait);
2409 		if (dst->ip6po_nexthop == NULL && canwait == M_NOWAIT)
2410 			goto bad;
2411 		memcpy(dst->ip6po_nexthop, src->ip6po_nexthop,
2412 		    src->ip6po_nexthop->sa_len);
2413 	}
2414 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2415 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2416 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2417 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2418 	return (0);
2419 
2420   bad:
2421 	if (dst->ip6po_pktinfo) free(dst->ip6po_pktinfo, M_IP6OPT);
2422 	if (dst->ip6po_nexthop) free(dst->ip6po_nexthop, M_IP6OPT);
2423 	if (dst->ip6po_hbh) free(dst->ip6po_hbh, M_IP6OPT);
2424 	if (dst->ip6po_dest1) free(dst->ip6po_dest1, M_IP6OPT);
2425 	if (dst->ip6po_dest2) free(dst->ip6po_dest2, M_IP6OPT);
2426 	if (dst->ip6po_rthdr) free(dst->ip6po_rthdr, M_IP6OPT);
2427 
2428 	return (ENOBUFS);
2429 }
2430 #undef PKTOPT_EXTHDRCPY
2431 
2432 struct ip6_pktopts *
2433 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2434 {
2435 	int error;
2436 	struct ip6_pktopts *dst;
2437 
2438 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2439 	if (dst == NULL && canwait == M_NOWAIT)
2440 		return (NULL);
2441 	ip6_initpktopts(dst);
2442 
2443 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2444 		free(dst, M_IP6OPT);
2445 		return (NULL);
2446 	}
2447 
2448 	return (dst);
2449 }
2450 
2451 void
2452 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2453 {
2454 	if (pktopt == NULL)
2455 		return;
2456 
2457 	ip6_clearpktopts(pktopt, -1);
2458 
2459 	free(pktopt, M_IP6OPT);
2460 }
2461 
2462 /*
2463  * Set the IP6 multicast options in response to user setsockopt().
2464  */
2465 static int
2466 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
2467 {
2468 	int error = 0;
2469 	u_int loop, ifindex;
2470 	struct ipv6_mreq *mreq;
2471 	struct ifnet *ifp;
2472 	struct ip6_moptions *im6o = *im6op;
2473 	struct route ro;
2474 	struct in6_multi_mship *imm;
2475 	struct lwp *l = curlwp;	/* XXX */
2476 
2477 	if (im6o == NULL) {
2478 		/*
2479 		 * No multicast option buffer attached to the pcb;
2480 		 * allocate one and initialize to default values.
2481 		 */
2482 		im6o = (struct ip6_moptions *)
2483 			malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
2484 
2485 		if (im6o == NULL)
2486 			return (ENOBUFS);
2487 		*im6op = im6o;
2488 		im6o->im6o_multicast_ifp = NULL;
2489 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2490 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2491 		LIST_INIT(&im6o->im6o_memberships);
2492 	}
2493 
2494 	switch (optname) {
2495 
2496 	case IPV6_MULTICAST_IF:
2497 		/*
2498 		 * Select the interface for outgoing multicast packets.
2499 		 */
2500 		if (m == NULL || m->m_len != sizeof(u_int)) {
2501 			error = EINVAL;
2502 			break;
2503 		}
2504 		bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
2505 		if (ifindex != 0) {
2506 			if (if_indexlim <= ifindex || !ifindex2ifnet[ifindex]) {
2507 				error = ENXIO;	/* XXX EINVAL? */
2508 				break;
2509 			}
2510 			ifp = ifindex2ifnet[ifindex];
2511 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
2512 				error = EADDRNOTAVAIL;
2513 				break;
2514 			}
2515 		} else
2516 			ifp = NULL;
2517 		im6o->im6o_multicast_ifp = ifp;
2518 		break;
2519 
2520 	case IPV6_MULTICAST_HOPS:
2521 	    {
2522 		/*
2523 		 * Set the IP6 hoplimit for outgoing multicast packets.
2524 		 */
2525 		int optval;
2526 		if (m == NULL || m->m_len != sizeof(int)) {
2527 			error = EINVAL;
2528 			break;
2529 		}
2530 		bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2531 		if (optval < -1 || optval >= 256)
2532 			error = EINVAL;
2533 		else if (optval == -1)
2534 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2535 		else
2536 			im6o->im6o_multicast_hlim = optval;
2537 		break;
2538 	    }
2539 
2540 	case IPV6_MULTICAST_LOOP:
2541 		/*
2542 		 * Set the loopback flag for outgoing multicast packets.
2543 		 * Must be zero or one.
2544 		 */
2545 		if (m == NULL || m->m_len != sizeof(u_int)) {
2546 			error = EINVAL;
2547 			break;
2548 		}
2549 		bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2550 		if (loop > 1) {
2551 			error = EINVAL;
2552 			break;
2553 		}
2554 		im6o->im6o_multicast_loop = loop;
2555 		break;
2556 
2557 	case IPV6_JOIN_GROUP:
2558 		/*
2559 		 * Add a multicast group membership.
2560 		 * Group must be a valid IP6 multicast address.
2561 		 */
2562 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2563 			error = EINVAL;
2564 			break;
2565 		}
2566 		mreq = mtod(m, struct ipv6_mreq *);
2567 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2568 			/*
2569 			 * We use the unspecified address to specify to accept
2570 			 * all multicast addresses. Only super user is allowed
2571 			 * to do this.
2572 			 */
2573 			if (kauth_authorize_generic(l->l_cred,
2574 			    KAUTH_GENERIC_ISSUSER, NULL))
2575 			{
2576 				error = EACCES;
2577 				break;
2578 			}
2579 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2580 			error = EINVAL;
2581 			break;
2582 		}
2583 
2584 		/*
2585 		 * If no interface was explicitly specified, choose an
2586 		 * appropriate one according to the given multicast address.
2587 		 */
2588 		if (mreq->ipv6mr_interface == 0) {
2589 			union {
2590 				struct sockaddr		dst;
2591 				struct sockaddr_in6	dst6;
2592 			} u;
2593 
2594 			/*
2595 			 * Look up the routing table for the
2596 			 * address, and choose the outgoing interface.
2597 			 *   XXX: is it a good approach?
2598 			 */
2599 			memset(&ro, 0, sizeof(ro));
2600 			sockaddr_in6_init(&u.dst6, &mreq->ipv6mr_multiaddr, 0,
2601 			    0, 0);
2602 			rtcache_setdst(&ro, &u.dst);
2603 			rtcache_init(&ro);
2604 			ifp = (ro.ro_rt != NULL) ? ro.ro_rt->rt_ifp : NULL;
2605 			rtcache_free(&ro);
2606 		} else {
2607 			/*
2608 			 * If the interface is specified, validate it.
2609 			 */
2610 			if (if_indexlim <= mreq->ipv6mr_interface ||
2611 			    !ifindex2ifnet[mreq->ipv6mr_interface]) {
2612 				error = ENXIO;	/* XXX EINVAL? */
2613 				break;
2614 			}
2615 			ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2616 		}
2617 
2618 		/*
2619 		 * See if we found an interface, and confirm that it
2620 		 * supports multicast
2621 		 */
2622 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2623 			error = EADDRNOTAVAIL;
2624 			break;
2625 		}
2626 
2627 		if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) {
2628 			error = EADDRNOTAVAIL; /* XXX: should not happen */
2629 			break;
2630 		}
2631 
2632 		/*
2633 		 * See if the membership already exists.
2634 		 */
2635 		for (imm = im6o->im6o_memberships.lh_first;
2636 		     imm != NULL; imm = imm->i6mm_chain.le_next)
2637 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
2638 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2639 			    &mreq->ipv6mr_multiaddr))
2640 				break;
2641 		if (imm != NULL) {
2642 			error = EADDRINUSE;
2643 			break;
2644 		}
2645 		/*
2646 		 * Everything looks good; add a new record to the multicast
2647 		 * address list for the given interface.
2648 		 */
2649 		imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error, 0);
2650 		if (imm == NULL)
2651 			break;
2652 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2653 		break;
2654 
2655 	case IPV6_LEAVE_GROUP:
2656 		/*
2657 		 * Drop a multicast group membership.
2658 		 * Group must be a valid IP6 multicast address.
2659 		 */
2660 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2661 			error = EINVAL;
2662 			break;
2663 		}
2664 		mreq = mtod(m, struct ipv6_mreq *);
2665 
2666 		/*
2667 		 * If an interface address was specified, get a pointer
2668 		 * to its ifnet structure.
2669 		 */
2670 		if (mreq->ipv6mr_interface != 0) {
2671 			if (if_indexlim <= mreq->ipv6mr_interface ||
2672 			    !ifindex2ifnet[mreq->ipv6mr_interface]) {
2673 				error = ENXIO;	/* XXX EINVAL? */
2674 				break;
2675 			}
2676 			ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2677 		} else
2678 			ifp = NULL;
2679 
2680 		/* Fill in the scope zone ID */
2681 		if (ifp) {
2682 			if (in6_setscope(&mreq->ipv6mr_multiaddr, ifp, NULL)) {
2683 				/* XXX: should not happen */
2684 				error = EADDRNOTAVAIL;
2685 				break;
2686 			}
2687 		} else if (mreq->ipv6mr_interface != 0) {
2688 			/*
2689 			 * XXX: This case would happens when the (positive)
2690 			 * index is in the valid range, but the corresponding
2691 			 * interface has been detached dynamically.  The above
2692 			 * check probably avoids such case to happen here, but
2693 			 * we check it explicitly for safety.
2694 			 */
2695 			error = EADDRNOTAVAIL;
2696 			break;
2697 		} else {	/* ipv6mr_interface == 0 */
2698 			struct sockaddr_in6 sa6_mc;
2699 
2700 			/*
2701 			 * The API spec says as follows:
2702 			 *  If the interface index is specified as 0, the
2703 			 *  system may choose a multicast group membership to
2704 			 *  drop by matching the multicast address only.
2705 			 * On the other hand, we cannot disambiguate the scope
2706 			 * zone unless an interface is provided.  Thus, we
2707 			 * check if there's ambiguity with the default scope
2708 			 * zone as the last resort.
2709 			 */
2710 			sockaddr_in6_init(&sa6_mc, &mreq->ipv6mr_multiaddr,
2711 			    0, 0, 0);
2712 			error = sa6_embedscope(&sa6_mc, ip6_use_defzone);
2713 			if (error != 0)
2714 				break;
2715 			mreq->ipv6mr_multiaddr = sa6_mc.sin6_addr;
2716 		}
2717 
2718 		/*
2719 		 * Find the membership in the membership list.
2720 		 */
2721 		for (imm = im6o->im6o_memberships.lh_first;
2722 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
2723 			if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2724 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2725 			    &mreq->ipv6mr_multiaddr))
2726 				break;
2727 		}
2728 		if (imm == NULL) {
2729 			/* Unable to resolve interface */
2730 			error = EADDRNOTAVAIL;
2731 			break;
2732 		}
2733 		/*
2734 		 * Give up the multicast address record to which the
2735 		 * membership points.
2736 		 */
2737 		LIST_REMOVE(imm, i6mm_chain);
2738 		in6_leavegroup(imm);
2739 		break;
2740 
2741 	default:
2742 		error = EOPNOTSUPP;
2743 		break;
2744 	}
2745 
2746 	/*
2747 	 * If all options have default values, no need to keep the mbuf.
2748 	 */
2749 	if (im6o->im6o_multicast_ifp == NULL &&
2750 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2751 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2752 	    im6o->im6o_memberships.lh_first == NULL) {
2753 		free(*im6op, M_IPMOPTS);
2754 		*im6op = NULL;
2755 	}
2756 
2757 	return (error);
2758 }
2759 
2760 /*
2761  * Return the IP6 multicast options in response to user getsockopt().
2762  */
2763 static int
2764 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp)
2765 {
2766 	u_int *hlim, *loop, *ifindex;
2767 
2768 	*mp = m_get(M_WAIT, MT_SOOPTS);
2769 
2770 	switch (optname) {
2771 
2772 	case IPV6_MULTICAST_IF:
2773 		ifindex = mtod(*mp, u_int *);
2774 		(*mp)->m_len = sizeof(u_int);
2775 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2776 			*ifindex = 0;
2777 		else
2778 			*ifindex = im6o->im6o_multicast_ifp->if_index;
2779 		return (0);
2780 
2781 	case IPV6_MULTICAST_HOPS:
2782 		hlim = mtod(*mp, u_int *);
2783 		(*mp)->m_len = sizeof(u_int);
2784 		if (im6o == NULL)
2785 			*hlim = ip6_defmcasthlim;
2786 		else
2787 			*hlim = im6o->im6o_multicast_hlim;
2788 		return (0);
2789 
2790 	case IPV6_MULTICAST_LOOP:
2791 		loop = mtod(*mp, u_int *);
2792 		(*mp)->m_len = sizeof(u_int);
2793 		if (im6o == NULL)
2794 			*loop = ip6_defmcasthlim;
2795 		else
2796 			*loop = im6o->im6o_multicast_loop;
2797 		return (0);
2798 
2799 	default:
2800 		return (EOPNOTSUPP);
2801 	}
2802 }
2803 
2804 /*
2805  * Discard the IP6 multicast options.
2806  */
2807 void
2808 ip6_freemoptions(struct ip6_moptions *im6o)
2809 {
2810 	struct in6_multi_mship *imm;
2811 
2812 	if (im6o == NULL)
2813 		return;
2814 
2815 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2816 		LIST_REMOVE(imm, i6mm_chain);
2817 		in6_leavegroup(imm);
2818 	}
2819 	free(im6o, M_IPMOPTS);
2820 }
2821 
2822 /*
2823  * Set IPv6 outgoing packet options based on advanced API.
2824  */
2825 int
2826 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2827 	struct ip6_pktopts *stickyopt, int priv, int uproto)
2828 {
2829 	struct cmsghdr *cm = 0;
2830 
2831 	if (control == NULL || opt == NULL)
2832 		return (EINVAL);
2833 
2834 	ip6_initpktopts(opt);
2835 	if (stickyopt) {
2836 		int error;
2837 
2838 		/*
2839 		 * If stickyopt is provided, make a local copy of the options
2840 		 * for this particular packet, then override them by ancillary
2841 		 * objects.
2842 		 * XXX: copypktopts() does not copy the cached route to a next
2843 		 * hop (if any).  This is not very good in terms of efficiency,
2844 		 * but we can allow this since this option should be rarely
2845 		 * used.
2846 		 */
2847 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2848 			return (error);
2849 	}
2850 
2851 	/*
2852 	 * XXX: Currently, we assume all the optional information is stored
2853 	 * in a single mbuf.
2854 	 */
2855 	if (control->m_next)
2856 		return (EINVAL);
2857 
2858 	for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2859 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2860 		int error;
2861 
2862 		if (control->m_len < CMSG_LEN(0))
2863 			return (EINVAL);
2864 
2865 		cm = mtod(control, struct cmsghdr *);
2866 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2867 			return (EINVAL);
2868 		if (cm->cmsg_level != IPPROTO_IPV6)
2869 			continue;
2870 
2871 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2872 		    cm->cmsg_len - CMSG_LEN(0), opt, priv, 0, 1, uproto);
2873 		if (error)
2874 			return (error);
2875 	}
2876 
2877 	return (0);
2878 }
2879 
2880 /*
2881  * Set a particular packet option, as a sticky option or an ancillary data
2882  * item.  "len" can be 0 only when it's a sticky option.
2883  * We have 4 cases of combination of "sticky" and "cmsg":
2884  * "sticky=0, cmsg=0": impossible
2885  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2886  * "sticky=1, cmsg=0": RFC3542 socket option
2887  * "sticky=1, cmsg=1": RFC2292 socket option
2888  */
2889 static int
2890 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2891     int priv, int sticky, int cmsg, int uproto)
2892 {
2893 	int minmtupolicy;
2894 
2895 	if (!sticky && !cmsg) {
2896 #ifdef DIAGNOSTIC
2897 		printf("ip6_setpktopt: impossible case\n");
2898 #endif
2899 		return (EINVAL);
2900 	}
2901 
2902 	/*
2903 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2904 	 * not be specified in the context of RFC3542.  Conversely,
2905 	 * RFC3542 types should not be specified in the context of RFC2292.
2906 	 */
2907 	if (!cmsg) {
2908 		switch (optname) {
2909 		case IPV6_2292PKTINFO:
2910 		case IPV6_2292HOPLIMIT:
2911 		case IPV6_2292NEXTHOP:
2912 		case IPV6_2292HOPOPTS:
2913 		case IPV6_2292DSTOPTS:
2914 		case IPV6_2292RTHDR:
2915 		case IPV6_2292PKTOPTIONS:
2916 			return (ENOPROTOOPT);
2917 		}
2918 	}
2919 	if (sticky && cmsg) {
2920 		switch (optname) {
2921 		case IPV6_PKTINFO:
2922 		case IPV6_HOPLIMIT:
2923 		case IPV6_NEXTHOP:
2924 		case IPV6_HOPOPTS:
2925 		case IPV6_DSTOPTS:
2926 		case IPV6_RTHDRDSTOPTS:
2927 		case IPV6_RTHDR:
2928 		case IPV6_USE_MIN_MTU:
2929 		case IPV6_DONTFRAG:
2930 		case IPV6_OTCLASS:
2931 		case IPV6_TCLASS:
2932 			return (ENOPROTOOPT);
2933 		}
2934 	}
2935 
2936 	switch (optname) {
2937 #ifdef RFC2292
2938 	case IPV6_2292PKTINFO:
2939 #endif
2940 	case IPV6_PKTINFO:
2941 	{
2942 		struct ifnet *ifp = NULL;
2943 		struct in6_pktinfo *pktinfo;
2944 
2945 		if (len != sizeof(struct in6_pktinfo))
2946 			return (EINVAL);
2947 
2948 		pktinfo = (struct in6_pktinfo *)buf;
2949 
2950 		/*
2951 		 * An application can clear any sticky IPV6_PKTINFO option by
2952 		 * doing a "regular" setsockopt with ipi6_addr being
2953 		 * in6addr_any and ipi6_ifindex being zero.
2954 		 * [RFC 3542, Section 6]
2955 		 */
2956 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2957 		    pktinfo->ipi6_ifindex == 0 &&
2958 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2959 			ip6_clearpktopts(opt, optname);
2960 			break;
2961 		}
2962 
2963 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2964 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2965 			return (EINVAL);
2966 		}
2967 
2968 		/* validate the interface index if specified. */
2969 		if (pktinfo->ipi6_ifindex >= if_indexlim) {
2970 			 return (ENXIO);
2971 		}
2972 		if (pktinfo->ipi6_ifindex) {
2973 			ifp = ifindex2ifnet[pktinfo->ipi6_ifindex];
2974 			if (ifp == NULL)
2975 				return (ENXIO);
2976 		}
2977 
2978 		/*
2979 		 * We store the address anyway, and let in6_selectsrc()
2980 		 * validate the specified address.  This is because ipi6_addr
2981 		 * may not have enough information about its scope zone, and
2982 		 * we may need additional information (such as outgoing
2983 		 * interface or the scope zone of a destination address) to
2984 		 * disambiguate the scope.
2985 		 * XXX: the delay of the validation may confuse the
2986 		 * application when it is used as a sticky option.
2987 		 */
2988 		if (opt->ip6po_pktinfo == NULL) {
2989 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2990 			    M_IP6OPT, M_NOWAIT);
2991 			if (opt->ip6po_pktinfo == NULL)
2992 				return (ENOBUFS);
2993 		}
2994 		memcpy(opt->ip6po_pktinfo, pktinfo, sizeof(*pktinfo));
2995 		break;
2996 	}
2997 
2998 #ifdef RFC2292
2999 	case IPV6_2292HOPLIMIT:
3000 #endif
3001 	case IPV6_HOPLIMIT:
3002 	{
3003 		int *hlimp;
3004 
3005 		/*
3006 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
3007 		 * to simplify the ordering among hoplimit options.
3008 		 */
3009 		if (optname == IPV6_HOPLIMIT && sticky)
3010 			return (ENOPROTOOPT);
3011 
3012 		if (len != sizeof(int))
3013 			return (EINVAL);
3014 		hlimp = (int *)buf;
3015 		if (*hlimp < -1 || *hlimp > 255)
3016 			return (EINVAL);
3017 
3018 		opt->ip6po_hlim = *hlimp;
3019 		break;
3020 	}
3021 
3022 	case IPV6_OTCLASS:
3023 		if (len != sizeof(u_int8_t))
3024 			return (EINVAL);
3025 
3026 		opt->ip6po_tclass = *(u_int8_t *)buf;
3027 		break;
3028 
3029 	case IPV6_TCLASS:
3030 	{
3031 		int tclass;
3032 
3033 		if (len != sizeof(int))
3034 			return (EINVAL);
3035 		tclass = *(int *)buf;
3036 		if (tclass < -1 || tclass > 255)
3037 			return (EINVAL);
3038 
3039 		opt->ip6po_tclass = tclass;
3040 		break;
3041 	}
3042 
3043 #ifdef RFC2292
3044 	case IPV6_2292NEXTHOP:
3045 #endif
3046 	case IPV6_NEXTHOP:
3047 		if (!priv)
3048 			return (EPERM);
3049 
3050 		if (len == 0) {	/* just remove the option */
3051 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
3052 			break;
3053 		}
3054 
3055 		/* check if cmsg_len is large enough for sa_len */
3056 		if (len < sizeof(struct sockaddr) || len < *buf)
3057 			return (EINVAL);
3058 
3059 		switch (((struct sockaddr *)buf)->sa_family) {
3060 		case AF_INET6:
3061 		{
3062 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
3063 			int error;
3064 
3065 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
3066 				return (EINVAL);
3067 
3068 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
3069 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
3070 				return (EINVAL);
3071 			}
3072 			if ((error = sa6_embedscope(sa6, ip6_use_defzone))
3073 			    != 0) {
3074 				return (error);
3075 			}
3076 			break;
3077 		}
3078 		case AF_LINK:	/* eventually be supported? */
3079 		default:
3080 			return (EAFNOSUPPORT);
3081 		}
3082 
3083 		/* turn off the previous option, then set the new option. */
3084 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
3085 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
3086 		if (opt->ip6po_nexthop == NULL)
3087 			return (ENOBUFS);
3088 		memcpy(opt->ip6po_nexthop, buf, *buf);
3089 		break;
3090 
3091 #ifdef RFC2292
3092 	case IPV6_2292HOPOPTS:
3093 #endif
3094 	case IPV6_HOPOPTS:
3095 	{
3096 		struct ip6_hbh *hbh;
3097 		int hbhlen;
3098 
3099 		/*
3100 		 * XXX: We don't allow a non-privileged user to set ANY HbH
3101 		 * options, since per-option restriction has too much
3102 		 * overhead.
3103 		 */
3104 		if (!priv)
3105 			return (EPERM);
3106 
3107 		if (len == 0) {
3108 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
3109 			break;	/* just remove the option */
3110 		}
3111 
3112 		/* message length validation */
3113 		if (len < sizeof(struct ip6_hbh))
3114 			return (EINVAL);
3115 		hbh = (struct ip6_hbh *)buf;
3116 		hbhlen = (hbh->ip6h_len + 1) << 3;
3117 		if (len != hbhlen)
3118 			return (EINVAL);
3119 
3120 		/* turn off the previous option, then set the new option. */
3121 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
3122 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
3123 		if (opt->ip6po_hbh == NULL)
3124 			return (ENOBUFS);
3125 		memcpy(opt->ip6po_hbh, hbh, hbhlen);
3126 
3127 		break;
3128 	}
3129 
3130 #ifdef RFC2292
3131 	case IPV6_2292DSTOPTS:
3132 #endif
3133 	case IPV6_DSTOPTS:
3134 	case IPV6_RTHDRDSTOPTS:
3135 	{
3136 		struct ip6_dest *dest, **newdest = NULL;
3137 		int destlen;
3138 
3139 		if (!priv)	/* XXX: see the comment for IPV6_HOPOPTS */
3140 			return (EPERM);
3141 
3142 		if (len == 0) {
3143 			ip6_clearpktopts(opt, optname);
3144 			break;	/* just remove the option */
3145 		}
3146 
3147 		/* message length validation */
3148 		if (len < sizeof(struct ip6_dest))
3149 			return (EINVAL);
3150 		dest = (struct ip6_dest *)buf;
3151 		destlen = (dest->ip6d_len + 1) << 3;
3152 		if (len != destlen)
3153 			return (EINVAL);
3154 		/*
3155 		 * Determine the position that the destination options header
3156 		 * should be inserted; before or after the routing header.
3157 		 */
3158 		switch (optname) {
3159 		case IPV6_2292DSTOPTS:
3160 			/*
3161 			 * The old advanced API is ambiguous on this point.
3162 			 * Our approach is to determine the position based
3163 			 * according to the existence of a routing header.
3164 			 * Note, however, that this depends on the order of the
3165 			 * extension headers in the ancillary data; the 1st
3166 			 * part of the destination options header must appear
3167 			 * before the routing header in the ancillary data,
3168 			 * too.
3169 			 * RFC3542 solved the ambiguity by introducing
3170 			 * separate ancillary data or option types.
3171 			 */
3172 			if (opt->ip6po_rthdr == NULL)
3173 				newdest = &opt->ip6po_dest1;
3174 			else
3175 				newdest = &opt->ip6po_dest2;
3176 			break;
3177 		case IPV6_RTHDRDSTOPTS:
3178 			newdest = &opt->ip6po_dest1;
3179 			break;
3180 		case IPV6_DSTOPTS:
3181 			newdest = &opt->ip6po_dest2;
3182 			break;
3183 		}
3184 
3185 		/* turn off the previous option, then set the new option. */
3186 		ip6_clearpktopts(opt, optname);
3187 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3188 		if (*newdest == NULL)
3189 			return (ENOBUFS);
3190 		memcpy(*newdest, dest, destlen);
3191 
3192 		break;
3193 	}
3194 
3195 #ifdef RFC2292
3196 	case IPV6_2292RTHDR:
3197 #endif
3198 	case IPV6_RTHDR:
3199 	{
3200 		struct ip6_rthdr *rth;
3201 		int rthlen;
3202 
3203 		if (len == 0) {
3204 			ip6_clearpktopts(opt, IPV6_RTHDR);
3205 			break;	/* just remove the option */
3206 		}
3207 
3208 		/* message length validation */
3209 		if (len < sizeof(struct ip6_rthdr))
3210 			return (EINVAL);
3211 		rth = (struct ip6_rthdr *)buf;
3212 		rthlen = (rth->ip6r_len + 1) << 3;
3213 		if (len != rthlen)
3214 			return (EINVAL);
3215 		switch (rth->ip6r_type) {
3216 		case IPV6_RTHDR_TYPE_0:
3217 			if (rth->ip6r_len == 0)	/* must contain one addr */
3218 				return (EINVAL);
3219 			if (rth->ip6r_len % 2) /* length must be even */
3220 				return (EINVAL);
3221 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3222 				return (EINVAL);
3223 			break;
3224 		default:
3225 			return (EINVAL);	/* not supported */
3226 		}
3227 		/* turn off the previous option */
3228 		ip6_clearpktopts(opt, IPV6_RTHDR);
3229 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3230 		if (opt->ip6po_rthdr == NULL)
3231 			return (ENOBUFS);
3232 		memcpy(opt->ip6po_rthdr, rth, rthlen);
3233 		break;
3234 	}
3235 
3236 	case IPV6_USE_MIN_MTU:
3237 		if (len != sizeof(int))
3238 			return (EINVAL);
3239 		minmtupolicy = *(int *)buf;
3240 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3241 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3242 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3243 			return (EINVAL);
3244 		}
3245 		opt->ip6po_minmtu = minmtupolicy;
3246 		break;
3247 
3248 	case IPV6_DONTFRAG:
3249 		if (len != sizeof(int))
3250 			return (EINVAL);
3251 
3252 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3253 			/*
3254 			 * we ignore this option for TCP sockets.
3255 			 * (RFC3542 leaves this case unspecified.)
3256 			 */
3257 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3258 		} else
3259 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3260 		break;
3261 
3262 	default:
3263 		return (ENOPROTOOPT);
3264 	} /* end of switch */
3265 
3266 	return (0);
3267 }
3268 
3269 /*
3270  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3271  * packet to the input queue of a specified interface.  Note that this
3272  * calls the output routine of the loopback "driver", but with an interface
3273  * pointer that might NOT be lo0ifp -- easier than replicating that code here.
3274  */
3275 void
3276 ip6_mloopback(struct ifnet *ifp, struct mbuf *m,
3277 	const struct sockaddr_in6 *dst)
3278 {
3279 	struct mbuf *copym;
3280 	struct ip6_hdr *ip6;
3281 
3282 	copym = m_copy(m, 0, M_COPYALL);
3283 	if (copym == NULL)
3284 		return;
3285 
3286 	/*
3287 	 * Make sure to deep-copy IPv6 header portion in case the data
3288 	 * is in an mbuf cluster, so that we can safely override the IPv6
3289 	 * header portion later.
3290 	 */
3291 	if ((copym->m_flags & M_EXT) != 0 ||
3292 	    copym->m_len < sizeof(struct ip6_hdr)) {
3293 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3294 		if (copym == NULL)
3295 			return;
3296 	}
3297 
3298 #ifdef DIAGNOSTIC
3299 	if (copym->m_len < sizeof(*ip6)) {
3300 		m_freem(copym);
3301 		return;
3302 	}
3303 #endif
3304 
3305 	ip6 = mtod(copym, struct ip6_hdr *);
3306 	/*
3307 	 * clear embedded scope identifiers if necessary.
3308 	 * in6_clearscope will touch the addresses only when necessary.
3309 	 */
3310 	in6_clearscope(&ip6->ip6_src);
3311 	in6_clearscope(&ip6->ip6_dst);
3312 
3313 	(void)looutput(ifp, copym, (const struct sockaddr *)dst, NULL);
3314 }
3315 
3316 /*
3317  * Chop IPv6 header off from the payload.
3318  */
3319 static int
3320 ip6_splithdr(struct mbuf *m,  struct ip6_exthdrs *exthdrs)
3321 {
3322 	struct mbuf *mh;
3323 	struct ip6_hdr *ip6;
3324 
3325 	ip6 = mtod(m, struct ip6_hdr *);
3326 	if (m->m_len > sizeof(*ip6)) {
3327 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3328 		if (mh == 0) {
3329 			m_freem(m);
3330 			return ENOBUFS;
3331 		}
3332 		M_MOVE_PKTHDR(mh, m);
3333 		MH_ALIGN(mh, sizeof(*ip6));
3334 		m->m_len -= sizeof(*ip6);
3335 		m->m_data += sizeof(*ip6);
3336 		mh->m_next = m;
3337 		m = mh;
3338 		m->m_len = sizeof(*ip6);
3339 		bcopy((void *)ip6, mtod(m, void *), sizeof(*ip6));
3340 	}
3341 	exthdrs->ip6e_ip6 = m;
3342 	return 0;
3343 }
3344 
3345 /*
3346  * Compute IPv6 extension header length.
3347  */
3348 int
3349 ip6_optlen(struct in6pcb *in6p)
3350 {
3351 	int len;
3352 
3353 	if (!in6p->in6p_outputopts)
3354 		return 0;
3355 
3356 	len = 0;
3357 #define elen(x) \
3358     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3359 
3360 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3361 	len += elen(in6p->in6p_outputopts->ip6po_dest1);
3362 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3363 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3364 	return len;
3365 #undef elen
3366 }
3367