xref: /netbsd-src/sys/netinet6/ip6_output.c (revision 20e85ad185ab16980f1219a557c42e057edb42ea)
1 /*	$NetBSD: ip6_output.c,v 1.88 2005/02/28 09:27:07 itojun 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 1.88 2005/02/28 09:27:07 itojun Exp $");
66 
67 #include "opt_inet.h"
68 #include "opt_ipsec.h"
69 #include "opt_pfil_hooks.h"
70 
71 #include <sys/param.h>
72 #include <sys/malloc.h>
73 #include <sys/mbuf.h>
74 #include <sys/errno.h>
75 #include <sys/protosw.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/systm.h>
79 #include <sys/proc.h>
80 
81 #include <net/if.h>
82 #include <net/route.h>
83 #ifdef PFIL_HOOKS
84 #include <net/pfil.h>
85 #endif
86 
87 #include <netinet/in.h>
88 #include <netinet/in_var.h>
89 #include <netinet/ip6.h>
90 #include <netinet/icmp6.h>
91 #include <netinet6/ip6_var.h>
92 #include <netinet6/in6_pcb.h>
93 #include <netinet6/nd6.h>
94 #include <netinet6/ip6protosw.h>
95 
96 #ifdef IPSEC
97 #include <netinet6/ipsec.h>
98 #include <netkey/key.h>
99 #endif /* IPSEC */
100 
101 #include <net/net_osdep.h>
102 
103 #ifdef PFIL_HOOKS
104 extern struct pfil_head inet6_pfil_hook;	/* XXX */
105 #endif
106 
107 struct ip6_exthdrs {
108 	struct mbuf *ip6e_ip6;
109 	struct mbuf *ip6e_hbh;
110 	struct mbuf *ip6e_dest1;
111 	struct mbuf *ip6e_rthdr;
112 	struct mbuf *ip6e_dest2;
113 };
114 
115 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
116 	struct socket *));
117 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *));
118 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **));
119 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int));
120 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
121 	struct ip6_frag **));
122 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t));
123 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *));
124 static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *,
125 	struct ifnet *, struct in6_addr *, u_long *, int *));
126 
127 /*
128  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
129  * header (with pri, len, nxt, hlim, src, dst).
130  * This function may modify ver and hlim only.
131  * The mbuf chain containing the packet will be freed.
132  * The mbuf opt, if present, will not be freed.
133  *
134  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
135  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
136  * which is rt_rmx.rmx_mtu.
137  */
138 int
139 ip6_output(m0, opt, ro, flags, im6o, so, ifpp)
140 	struct mbuf *m0;
141 	struct ip6_pktopts *opt;
142 	struct route_in6 *ro;
143 	int flags;
144 	struct ip6_moptions *im6o;
145 	struct socket *so;
146 	struct ifnet **ifpp;		/* XXX: just for statistics */
147 {
148 	struct ip6_hdr *ip6, *mhip6;
149 	struct ifnet *ifp, *origifp;
150 	struct mbuf *m = m0;
151 	int hlen, tlen, len, off;
152 	struct route_in6 ip6route;
153 	struct sockaddr_in6 *dst;
154 	int error = 0;
155 	u_long mtu;
156 	int alwaysfrag, dontfrag;
157 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
158 	struct ip6_exthdrs exthdrs;
159 	struct in6_addr finaldst;
160 	struct route_in6 *ro_pmtu = NULL;
161 	int hdrsplit = 0;
162 	int needipsec = 0;
163 #ifdef IPSEC
164 	int needipsectun = 0;
165 	struct secpolicy *sp = NULL;
166 
167 	ip6 = mtod(m, struct ip6_hdr *);
168 #endif /* IPSEC */
169 
170 #define MAKE_EXTHDR(hp, mp)						\
171     do {								\
172 	if (hp) {							\
173 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
174 		error = ip6_copyexthdr((mp), (caddr_t)(hp), 		\
175 		    ((eh)->ip6e_len + 1) << 3);				\
176 		if (error)						\
177 			goto freehdrs;					\
178 	}								\
179     } while (/*CONSTCOND*/ 0)
180 
181 	bzero(&exthdrs, sizeof(exthdrs));
182 	if (opt) {
183 		/* Hop-by-Hop options header */
184 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
185 		/* Destination options header(1st part) */
186 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
187 		/* Routing header */
188 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
189 		/* Destination options header(2nd part) */
190 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
191 	}
192 
193 #ifdef IPSEC
194 	if ((flags & IPV6_FORWARDING) != 0) {
195 		needipsec = 0;
196 		goto skippolicycheck;
197 	}
198 
199 	/* get a security policy for this packet */
200 	if (so == NULL)
201 		sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
202 	else {
203 		if (IPSEC_PCB_SKIP_IPSEC(sotoinpcb_hdr(so)->inph_sp,
204 					 IPSEC_DIR_OUTBOUND)) {
205 			needipsec = 0;
206 			goto skippolicycheck;
207 		}
208 		sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
209 	}
210 
211 	if (sp == NULL) {
212 		ipsec6stat.out_inval++;
213 		goto freehdrs;
214 	}
215 
216 	error = 0;
217 
218 	/* check policy */
219 	switch (sp->policy) {
220 	case IPSEC_POLICY_DISCARD:
221 		/*
222 		 * This packet is just discarded.
223 		 */
224 		ipsec6stat.out_polvio++;
225 		goto freehdrs;
226 
227 	case IPSEC_POLICY_BYPASS:
228 	case IPSEC_POLICY_NONE:
229 		/* no need to do IPsec. */
230 		needipsec = 0;
231 		break;
232 
233 	case IPSEC_POLICY_IPSEC:
234 		if (sp->req == NULL) {
235 			/* XXX should be panic ? */
236 			printf("ip6_output: No IPsec request specified.\n");
237 			error = EINVAL;
238 			goto freehdrs;
239 		}
240 		needipsec = 1;
241 		break;
242 
243 	case IPSEC_POLICY_ENTRUST:
244 	default:
245 		printf("ip6_output: Invalid policy found. %d\n", sp->policy);
246 	}
247 
248   skippolicycheck:;
249 #endif /* IPSEC */
250 
251 	/*
252 	 * Calculate the total length of the extension header chain.
253 	 * Keep the length of the unfragmentable part for fragmentation.
254 	 */
255 	optlen = 0;
256 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
257 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
258 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
259 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
260 	/* NOTE: we don't add AH/ESP length here. do that later. */
261 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
262 
263 	/*
264 	 * If we need IPsec, or there is at least one extension header,
265 	 * separate IP6 header from the payload.
266 	 */
267 	if ((needipsec || optlen) && !hdrsplit) {
268 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
269 			m = NULL;
270 			goto freehdrs;
271 		}
272 		m = exthdrs.ip6e_ip6;
273 		hdrsplit++;
274 	}
275 
276 	/* adjust pointer */
277 	ip6 = mtod(m, struct ip6_hdr *);
278 
279 	/* adjust mbuf packet header length */
280 	m->m_pkthdr.len += optlen;
281 	plen = m->m_pkthdr.len - sizeof(*ip6);
282 
283 	/* If this is a jumbo payload, insert a jumbo payload option. */
284 	if (plen > IPV6_MAXPACKET) {
285 		if (!hdrsplit) {
286 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
287 				m = NULL;
288 				goto freehdrs;
289 			}
290 			m = exthdrs.ip6e_ip6;
291 			hdrsplit++;
292 		}
293 		/* adjust pointer */
294 		ip6 = mtod(m, struct ip6_hdr *);
295 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
296 			goto freehdrs;
297 		ip6->ip6_plen = 0;
298 	} else
299 		ip6->ip6_plen = htons(plen);
300 
301 	/*
302 	 * Concatenate headers and fill in next header fields.
303 	 * Here we have, on "m"
304 	 *	IPv6 payload
305 	 * and we insert headers accordingly.  Finally, we should be getting:
306 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
307 	 *
308 	 * during the header composing process, "m" points to IPv6 header.
309 	 * "mprev" points to an extension header prior to esp.
310 	 */
311 	{
312 		u_char *nexthdrp = &ip6->ip6_nxt;
313 		struct mbuf *mprev = m;
314 
315 		/*
316 		 * we treat dest2 specially.  this makes IPsec processing
317 		 * much easier.  the goal here is to make mprev point the
318 		 * mbuf prior to dest2.
319 		 *
320 		 * result: IPv6 dest2 payload
321 		 * m and mprev will point to IPv6 header.
322 		 */
323 		if (exthdrs.ip6e_dest2) {
324 			if (!hdrsplit)
325 				panic("assumption failed: hdr not split");
326 			exthdrs.ip6e_dest2->m_next = m->m_next;
327 			m->m_next = exthdrs.ip6e_dest2;
328 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
329 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
330 		}
331 
332 #define MAKE_CHAIN(m, mp, p, i)\
333     do {\
334 	if (m) {\
335 		if (!hdrsplit) \
336 			panic("assumption failed: hdr not split"); \
337 		*mtod((m), u_char *) = *(p);\
338 		*(p) = (i);\
339 		p = mtod((m), u_char *);\
340 		(m)->m_next = (mp)->m_next;\
341 		(mp)->m_next = (m);\
342 		(mp) = (m);\
343 	}\
344     } while (/*CONSTCOND*/ 0)
345 		/*
346 		 * result: IPv6 hbh dest1 rthdr dest2 payload
347 		 * m will point to IPv6 header.  mprev will point to the
348 		 * extension header prior to dest2 (rthdr in the above case).
349 		 */
350 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
351 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
352 		    IPPROTO_DSTOPTS);
353 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
354 		    IPPROTO_ROUTING);
355 
356 #ifdef IPSEC
357 		if (!needipsec)
358 			goto skip_ipsec2;
359 
360 		/*
361 		 * pointers after IPsec headers are not valid any more.
362 		 * other pointers need a great care too.
363 		 * (IPsec routines should not mangle mbufs prior to AH/ESP)
364 		 */
365 		exthdrs.ip6e_dest2 = NULL;
366 
367 	    {
368 		struct ip6_rthdr *rh = NULL;
369 		int segleft_org = 0;
370 		struct ipsec_output_state state;
371 
372 		if (exthdrs.ip6e_rthdr) {
373 			rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
374 			segleft_org = rh->ip6r_segleft;
375 			rh->ip6r_segleft = 0;
376 		}
377 
378 		bzero(&state, sizeof(state));
379 		state.m = m;
380 		error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
381 		    &needipsectun);
382 		m = state.m;
383 		if (error) {
384 			/* mbuf is already reclaimed in ipsec6_output_trans. */
385 			m = NULL;
386 			switch (error) {
387 			case EHOSTUNREACH:
388 			case ENETUNREACH:
389 			case EMSGSIZE:
390 			case ENOBUFS:
391 			case ENOMEM:
392 				break;
393 			default:
394 				printf("ip6_output (ipsec): error code %d\n", error);
395 				/* FALLTHROUGH */
396 			case ENOENT:
397 				/* don't show these error codes to the user */
398 				error = 0;
399 				break;
400 			}
401 			goto bad;
402 		}
403 		if (exthdrs.ip6e_rthdr) {
404 			/* ah6_output doesn't modify mbuf chain */
405 			rh->ip6r_segleft = segleft_org;
406 		}
407 	    }
408 skip_ipsec2:;
409 #endif
410 	}
411 
412 	/*
413 	 * If there is a routing header, replace destination address field
414 	 * with the first hop of the routing header.
415 	 */
416 	if (exthdrs.ip6e_rthdr) {
417 		struct ip6_rthdr *rh;
418 		struct ip6_rthdr0 *rh0;
419 		struct in6_addr *addr;
420 
421 		rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
422 		    struct ip6_rthdr *));
423 		finaldst = ip6->ip6_dst;
424 		switch (rh->ip6r_type) {
425 		case IPV6_RTHDR_TYPE_0:
426 			 rh0 = (struct ip6_rthdr0 *)rh;
427 			 addr = (struct in6_addr *)(rh0 + 1);
428 			 ip6->ip6_dst = addr[0];
429 			 bcopy(&addr[1], &addr[0],
430 			     sizeof(struct in6_addr) * (rh0->ip6r0_segleft - 1));
431 			 addr[rh0->ip6r0_segleft - 1] = finaldst;
432 			 break;
433 		default:	/* is it possible? */
434 			 error = EINVAL;
435 			 goto bad;
436 		}
437 	}
438 
439 	/* Source address validation */
440 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
441 	    (flags & IPV6_UNSPECSRC) == 0) {
442 		error = EOPNOTSUPP;
443 		ip6stat.ip6s_badscope++;
444 		goto bad;
445 	}
446 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
447 		error = EOPNOTSUPP;
448 		ip6stat.ip6s_badscope++;
449 		goto bad;
450 	}
451 
452 	ip6stat.ip6s_localout++;
453 
454 	/*
455 	 * Route packet.
456 	 */
457 	/* initialize cached route */
458 	if (ro == 0) {
459 		ro = &ip6route;
460 		bzero((caddr_t)ro, sizeof(*ro));
461 	}
462 	ro_pmtu = ro;
463 	if (opt && opt->ip6po_rthdr)
464 		ro = &opt->ip6po_route;
465 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
466 	/*
467 	 * If there is a cached route,
468 	 * check that it is to the same destination
469 	 * and is still up. If not, free it and try again.
470 	 */
471 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
472 	    dst->sin6_family != AF_INET6 ||
473 	    !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
474 		RTFREE(ro->ro_rt);
475 		ro->ro_rt = (struct rtentry *)0;
476 	}
477 	if (ro->ro_rt == 0) {
478 		bzero(dst, sizeof(*dst));
479 		dst->sin6_family = AF_INET6;
480 		dst->sin6_len = sizeof(struct sockaddr_in6);
481 		dst->sin6_addr = ip6->ip6_dst;
482 	}
483 #ifdef IPSEC
484 	if (needipsec && needipsectun) {
485 		struct ipsec_output_state state;
486 
487 		/*
488 		 * All the extension headers will become inaccessible
489 		 * (since they can be encrypted).
490 		 * Don't panic, we need no more updates to extension headers
491 		 * on inner IPv6 packet (since they are now encapsulated).
492 		 *
493 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
494 		 */
495 		bzero(&exthdrs, sizeof(exthdrs));
496 		exthdrs.ip6e_ip6 = m;
497 
498 		bzero(&state, sizeof(state));
499 		state.m = m;
500 		state.ro = (struct route *)ro;
501 		state.dst = (struct sockaddr *)dst;
502 
503 		error = ipsec6_output_tunnel(&state, sp, flags);
504 
505 		m = state.m;
506 		ro_pmtu = ro = (struct route_in6 *)state.ro;
507 		dst = (struct sockaddr_in6 *)state.dst;
508 		if (error) {
509 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
510 			m0 = m = NULL;
511 			m = NULL;
512 			switch (error) {
513 			case EHOSTUNREACH:
514 			case ENETUNREACH:
515 			case EMSGSIZE:
516 			case ENOBUFS:
517 			case ENOMEM:
518 				break;
519 			default:
520 				printf("ip6_output (ipsec): error code %d\n", error);
521 				/* FALLTHROUGH */
522 			case ENOENT:
523 				/* don't show these error codes to the user */
524 				error = 0;
525 				break;
526 			}
527 			goto bad;
528 		}
529 
530 		exthdrs.ip6e_ip6 = m;
531 	}
532 #endif /* IPSEC */
533 
534 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
535 		/* Unicast */
536 
537 #define ifatoia6(ifa)	((struct in6_ifaddr *)(ifa))
538 #define sin6tosa(sin6)	((struct sockaddr *)(sin6))
539 		/* xxx
540 		 * interface selection comes here
541 		 * if an interface is specified from an upper layer,
542 		 * ifp must point it.
543 		 */
544 		if (ro->ro_rt == 0) {
545 			/*
546 			 * non-bsdi always clone routes, if parent is
547 			 * PRF_CLONING.
548 			 */
549 			rtalloc((struct route *)ro);
550 		}
551 		if (ro->ro_rt == 0) {
552 			ip6stat.ip6s_noroute++;
553 			error = EHOSTUNREACH;
554 			/* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
555 			goto bad;
556 		}
557 		ifp = ro->ro_rt->rt_ifp;
558 		ro->ro_rt->rt_use++;
559 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
560 			dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
561 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
562 
563 		in6_ifstat_inc(ifp, ifs6_out_request);
564 
565 		/*
566 		 * Check if the outgoing interface conflicts with
567 		 * the interface specified by ifi6_ifindex (if specified).
568 		 * Note that loopback interface is always okay.
569 		 * (this may happen when we are sending a packet to one of
570 		 *  our own addresses.)
571 		 */
572 		if (opt && opt->ip6po_pktinfo &&
573 		    opt->ip6po_pktinfo->ipi6_ifindex) {
574 			if (!(ifp->if_flags & IFF_LOOPBACK) &&
575 			    ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
576 				ip6stat.ip6s_noroute++;
577 				in6_ifstat_inc(ifp, ifs6_out_discard);
578 				error = EHOSTUNREACH;
579 				goto bad;
580 			}
581 		}
582 
583 		if (opt && opt->ip6po_hlim != -1)
584 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
585 	} else {
586 		/* Multicast */
587 		struct	in6_multi *in6m;
588 
589 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
590 
591 		/*
592 		 * See if the caller provided any multicast options
593 		 */
594 		ifp = NULL;
595 		if (im6o != NULL) {
596 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
597 			if (im6o->im6o_multicast_ifp != NULL)
598 				ifp = im6o->im6o_multicast_ifp;
599 		} else
600 			ip6->ip6_hlim = ip6_defmcasthlim;
601 
602 		/*
603 		 * See if the caller provided the outgoing interface
604 		 * as an ancillary data.
605 		 * Boundary check for ifindex is assumed to be already done.
606 		 */
607 		if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
608 			ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex];
609 
610 		/*
611 		 * If the destination is a node-local scope multicast,
612 		 * the packet should be loop-backed only.
613 		 */
614 		if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
615 			/*
616 			 * If the outgoing interface is already specified,
617 			 * it should be a loopback interface.
618 			 */
619 			if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) {
620 				ip6stat.ip6s_badscope++;
621 				error = ENETUNREACH; /* XXX: better error? */
622 				/* XXX correct ifp? */
623 				in6_ifstat_inc(ifp, ifs6_out_discard);
624 				goto bad;
625 			} else
626 				ifp = lo0ifp;
627 		}
628 
629 		if (opt && opt->ip6po_hlim != -1)
630 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
631 
632 		/*
633 		 * If caller did not provide an interface lookup a
634 		 * default in the routing table.  This is either a
635 		 * default for the speicfied group (i.e. a host
636 		 * route), or a multicast default (a route for the
637 		 * ``net'' ff00::/8).
638 		 */
639 		if (ifp == NULL) {
640 			if (ro->ro_rt == 0) {
641 				ro->ro_rt = rtalloc1((struct sockaddr *)
642 				    &ro->ro_dst, 0);
643 			}
644 			if (ro->ro_rt == 0) {
645 				ip6stat.ip6s_noroute++;
646 				error = EHOSTUNREACH;
647 				/* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
648 				goto bad;
649 			}
650 			ifp = ro->ro_rt->rt_ifp;
651 			ro->ro_rt->rt_use++;
652 		}
653 
654 		if ((flags & IPV6_FORWARDING) == 0)
655 			in6_ifstat_inc(ifp, ifs6_out_request);
656 		in6_ifstat_inc(ifp, ifs6_out_mcast);
657 
658 		/*
659 		 * Confirm that the outgoing interface supports multicast.
660 		 */
661 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
662 			ip6stat.ip6s_noroute++;
663 			in6_ifstat_inc(ifp, ifs6_out_discard);
664 			error = ENETUNREACH;
665 			goto bad;
666 		}
667 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
668 		if (in6m != NULL &&
669 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
670 			/*
671 			 * If we belong to the destination multicast group
672 			 * on the outgoing interface, and the caller did not
673 			 * forbid loopback, loop back a copy.
674 			 */
675 			ip6_mloopback(ifp, m, dst);
676 		} else {
677 			/*
678 			 * If we are acting as a multicast router, perform
679 			 * multicast forwarding as if the packet had just
680 			 * arrived on the interface to which we are about
681 			 * to send.  The multicast forwarding function
682 			 * recursively calls this function, using the
683 			 * IPV6_FORWARDING flag to prevent infinite recursion.
684 			 *
685 			 * Multicasts that are looped back by ip6_mloopback(),
686 			 * above, will be forwarded by the ip6_input() routine,
687 			 * if necessary.
688 			 */
689 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
690 				if (ip6_mforward(ip6, ifp, m) != 0) {
691 					m_freem(m);
692 					goto done;
693 				}
694 			}
695 		}
696 		/*
697 		 * Multicasts with a hoplimit of zero may be looped back,
698 		 * above, but must not be transmitted on a network.
699 		 * Also, multicasts addressed to the loopback interface
700 		 * are not sent -- the above call to ip6_mloopback() will
701 		 * loop back a copy if this host actually belongs to the
702 		 * destination group on the loopback interface.
703 		 */
704 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
705 			m_freem(m);
706 			goto done;
707 		}
708 	}
709 
710 	/*
711 	 * Fill the outgoing inteface to tell the upper layer
712 	 * to increment per-interface statistics.
713 	 */
714 	if (ifpp)
715 		*ifpp = ifp;
716 
717 	/* Determine path MTU. */
718 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
719 	    &alwaysfrag)) != 0)
720 		goto bad;
721 #ifdef IPSEC
722 	if (needipsectun)
723 		mtu = IPV6_MMTU;
724 #endif
725 
726 	/*
727 	 * The caller of this function may specify to use the minimum MTU
728 	 * in some cases.
729 	 */
730 	if (mtu > IPV6_MMTU) {
731 		if ((flags & IPV6_MINMTU))
732 			mtu = IPV6_MMTU;
733 	}
734 
735 	/* Fake scoped addresses */
736 	if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
737 		/*
738 		 * If source or destination address is a scoped address, and
739 		 * the packet is going to be sent to a loopback interface,
740 		 * we should keep the original interface.
741 		 */
742 
743 		/*
744 		 * XXX: this is a very experimental and temporary solution.
745 		 * We eventually have sockaddr_in6 and use the sin6_scope_id
746 		 * field of the structure here.
747 		 * We rely on the consistency between two scope zone ids
748 		 * of source add destination, which should already be assured
749 		 * Larger scopes than link will be supported in the near
750 		 * future.
751 		 */
752 		origifp = NULL;
753 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
754 			origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
755 		else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
756 			origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
757 		/*
758 		 * XXX: origifp can be NULL even in those two cases above.
759 		 * For example, if we remove the (only) link-local address
760 		 * from the loopback interface, and try to send a link-local
761 		 * address without link-id information.  Then the source
762 		 * address is ::1, and the destination address is the
763 		 * link-local address with its s6_addr16[1] being zero.
764 		 * What is worse, if the packet goes to the loopback interface
765 		 * by a default rejected route, the null pointer would be
766 		 * passed to looutput, and the kernel would hang.
767 		 * The following last resort would prevent such disaster.
768 		 */
769 		if (origifp == NULL)
770 			origifp = ifp;
771 	} else
772 		origifp = ifp;
773 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
774 		ip6->ip6_src.s6_addr16[1] = 0;
775 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
776 		ip6->ip6_dst.s6_addr16[1] = 0;
777 
778 	/*
779 	 * If the outgoing packet contains a hop-by-hop options header,
780 	 * it must be examined and processed even by the source node.
781 	 * (RFC 2460, section 4.)
782 	 */
783 	if (exthdrs.ip6e_hbh) {
784 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
785 		u_int32_t dummy1; /* XXX unused */
786 		u_int32_t dummy2; /* XXX unused */
787 
788 		/*
789 		 *  XXX: if we have to send an ICMPv6 error to the sender,
790 		 *       we need the M_LOOP flag since icmp6_error() expects
791 		 *       the IPv6 and the hop-by-hop options header are
792 		 *       continuous unless the flag is set.
793 		 */
794 		m->m_flags |= M_LOOP;
795 		m->m_pkthdr.rcvif = ifp;
796 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
797 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
798 		    &dummy1, &dummy2) < 0) {
799 			/* m was already freed at this point */
800 			error = EINVAL;/* better error? */
801 			goto done;
802 		}
803 		m->m_flags &= ~M_LOOP; /* XXX */
804 		m->m_pkthdr.rcvif = NULL;
805 	}
806 
807 #ifdef PFIL_HOOKS
808 	/*
809 	 * Run through list of hooks for output packets.
810 	 */
811 	if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, ifp, PFIL_OUT)) != 0)
812 		goto done;
813 	if (m == NULL)
814 		goto done;
815 	ip6 = mtod(m, struct ip6_hdr *);
816 #endif /* PFIL_HOOKS */
817 	/*
818 	 * Send the packet to the outgoing interface.
819 	 * If necessary, do IPv6 fragmentation before sending.
820 	 *
821 	 * the logic here is rather complex:
822 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
823 	 * 1-a:	send as is if tlen <= path mtu
824 	 * 1-b:	fragment if tlen > path mtu
825 	 *
826 	 * 2: if user asks us not to fragment (dontfrag == 1)
827 	 * 2-a:	send as is if tlen <= interface mtu
828 	 * 2-b:	error if tlen > interface mtu
829 	 *
830 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
831 	 *	always fragment
832 	 *
833 	 * 4: if dontfrag == 1 && alwaysfrag == 1
834 	 *	error, as we cannot handle this conflicting request
835 	 */
836 	tlen = m->m_pkthdr.len;
837 
838 	dontfrag = 0;
839 	if (dontfrag && alwaysfrag) {	/* case 4 */
840 		/* conflicting request - can't transmit */
841 		error = EMSGSIZE;
842 		goto bad;
843 	}
844 	if (dontfrag && tlen > IN6_LINKMTU(ifp)) {	/* case 2-b */
845 		/*
846 		 * Even if the DONTFRAG option is specified, we cannot send the
847 		 * packet when the data length is larger than the MTU of the
848 		 * outgoing interface.
849 		 * Notify the error by sending IPV6_PATHMTU ancillary data as
850 		 * well as returning an error code (the latter is not described
851 		 * in the API spec.)
852 		 */
853 		u_int32_t mtu32;
854 		struct ip6ctlparam ip6cp;
855 
856 		mtu32 = (u_int32_t)mtu;
857 		bzero(&ip6cp, sizeof(ip6cp));
858 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
859 		pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
860 		    (void *)&ip6cp);
861 
862 		error = EMSGSIZE;
863 		goto bad;
864 	}
865 
866 	/*
867 	 * transmit packet without fragmentation
868 	 */
869 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
870 		struct in6_ifaddr *ia6;
871 
872 		ip6 = mtod(m, struct ip6_hdr *);
873 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
874 		if (ia6) {
875 			/* Record statistics for this interface address. */
876 			ia6->ia_ifa.ifa_data.ifad_outbytes += m->m_pkthdr.len;
877 		}
878 #ifdef IPSEC
879 		/* clean ipsec history once it goes out of the node */
880 		ipsec_delaux(m);
881 #endif
882 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
883 		goto done;
884 	}
885 
886 	/*
887 	 * try to fragment the packet.  case 1-b and 3
888 	 */
889 	if (mtu < IPV6_MMTU) {
890 		/* path MTU cannot be less than IPV6_MMTU */
891 		error = EMSGSIZE;
892 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
893 		goto bad;
894 	} else if (ip6->ip6_plen == 0) {
895 		/* jumbo payload cannot be fragmented */
896 		error = EMSGSIZE;
897 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
898 		goto bad;
899 	} else {
900 		struct mbuf **mnext, *m_frgpart;
901 		struct ip6_frag *ip6f;
902 		u_int32_t id = htonl(ip6_randomid());
903 		u_char nextproto;
904 		struct ip6ctlparam ip6cp;
905 		u_int32_t mtu32;
906 
907 		/*
908 		 * Too large for the destination or interface;
909 		 * fragment if possible.
910 		 * Must be able to put at least 8 bytes per fragment.
911 		 */
912 		hlen = unfragpartlen;
913 		if (mtu > IPV6_MAXPACKET)
914 			mtu = IPV6_MAXPACKET;
915 
916 		/* Notify a proper path MTU to applications. */
917 		mtu32 = (u_int32_t)mtu;
918 		bzero(&ip6cp, sizeof(ip6cp));
919 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
920 		pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
921 		    (void *)&ip6cp);
922 
923 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
924 		if (len < 8) {
925 			error = EMSGSIZE;
926 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
927 			goto bad;
928 		}
929 
930 		mnext = &m->m_nextpkt;
931 
932 		/*
933 		 * Change the next header field of the last header in the
934 		 * unfragmentable part.
935 		 */
936 		if (exthdrs.ip6e_rthdr) {
937 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
938 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
939 		} else if (exthdrs.ip6e_dest1) {
940 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
941 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
942 		} else if (exthdrs.ip6e_hbh) {
943 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
944 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
945 		} else {
946 			nextproto = ip6->ip6_nxt;
947 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
948 		}
949 
950 		/*
951 		 * Loop through length of segment after first fragment,
952 		 * make new header and copy data of each part and link onto
953 		 * chain.
954 		 */
955 		m0 = m;
956 		for (off = hlen; off < tlen; off += len) {
957 			struct mbuf *mlast;
958 
959 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
960 			if (!m) {
961 				error = ENOBUFS;
962 				ip6stat.ip6s_odropped++;
963 				goto sendorfree;
964 			}
965 			m->m_pkthdr.rcvif = NULL;
966 			m->m_flags = m0->m_flags & M_COPYFLAGS;
967 			*mnext = m;
968 			mnext = &m->m_nextpkt;
969 			m->m_data += max_linkhdr;
970 			mhip6 = mtod(m, struct ip6_hdr *);
971 			*mhip6 = *ip6;
972 			m->m_len = sizeof(*mhip6);
973 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
974 			if (error) {
975 				ip6stat.ip6s_odropped++;
976 				goto sendorfree;
977 			}
978 			ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7));
979 			if (off + len >= tlen)
980 				len = tlen - off;
981 			else
982 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
983 			mhip6->ip6_plen = htons((u_int16_t)(len + hlen +
984 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
985 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
986 				error = ENOBUFS;
987 				ip6stat.ip6s_odropped++;
988 				goto sendorfree;
989 			}
990 			for (mlast = m; mlast->m_next; mlast = mlast->m_next)
991 				;
992 			mlast->m_next = m_frgpart;
993 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
994 			m->m_pkthdr.rcvif = (struct ifnet *)0;
995 			ip6f->ip6f_reserved = 0;
996 			ip6f->ip6f_ident = id;
997 			ip6f->ip6f_nxt = nextproto;
998 			ip6stat.ip6s_ofragments++;
999 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1000 		}
1001 
1002 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1003 	}
1004 
1005 	/*
1006 	 * Remove leading garbages.
1007 	 */
1008 sendorfree:
1009 	m = m0->m_nextpkt;
1010 	m0->m_nextpkt = 0;
1011 	m_freem(m0);
1012 	for (m0 = m; m; m = m0) {
1013 		m0 = m->m_nextpkt;
1014 		m->m_nextpkt = 0;
1015 		if (error == 0) {
1016 			struct in6_ifaddr *ia6;
1017 			ip6 = mtod(m, struct ip6_hdr *);
1018 			ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1019 			if (ia6) {
1020 				/*
1021 				 * Record statistics for this interface
1022 				 * address.
1023 				 */
1024 				ia6->ia_ifa.ifa_data.ifad_outbytes +=
1025 				    m->m_pkthdr.len;
1026 			}
1027 #ifdef IPSEC
1028 			/* clean ipsec history once it goes out of the node */
1029 			ipsec_delaux(m);
1030 #endif
1031 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1032 		} else
1033 			m_freem(m);
1034 	}
1035 
1036 	if (error == 0)
1037 		ip6stat.ip6s_fragmented++;
1038 
1039 done:
1040 	if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1041 		RTFREE(ro->ro_rt);
1042 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1043 		RTFREE(ro_pmtu->ro_rt);
1044 	}
1045 
1046 #ifdef IPSEC
1047 	if (sp != NULL)
1048 		key_freesp(sp);
1049 #endif /* IPSEC */
1050 
1051 	return (error);
1052 
1053 freehdrs:
1054 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1055 	m_freem(exthdrs.ip6e_dest1);
1056 	m_freem(exthdrs.ip6e_rthdr);
1057 	m_freem(exthdrs.ip6e_dest2);
1058 	/* FALLTHROUGH */
1059 bad:
1060 	m_freem(m);
1061 	goto done;
1062 }
1063 
1064 static int
1065 ip6_copyexthdr(mp, hdr, hlen)
1066 	struct mbuf **mp;
1067 	caddr_t hdr;
1068 	int hlen;
1069 {
1070 	struct mbuf *m;
1071 
1072 	if (hlen > MCLBYTES)
1073 		return (ENOBUFS); /* XXX */
1074 
1075 	MGET(m, M_DONTWAIT, MT_DATA);
1076 	if (!m)
1077 		return (ENOBUFS);
1078 
1079 	if (hlen > MLEN) {
1080 		MCLGET(m, M_DONTWAIT);
1081 		if ((m->m_flags & M_EXT) == 0) {
1082 			m_free(m);
1083 			return (ENOBUFS);
1084 		}
1085 	}
1086 	m->m_len = hlen;
1087 	if (hdr)
1088 		bcopy(hdr, mtod(m, caddr_t), hlen);
1089 
1090 	*mp = m;
1091 	return (0);
1092 }
1093 
1094 /*
1095  * Insert jumbo payload option.
1096  */
1097 static int
1098 ip6_insert_jumboopt(exthdrs, plen)
1099 	struct ip6_exthdrs *exthdrs;
1100 	u_int32_t plen;
1101 {
1102 	struct mbuf *mopt;
1103 	u_int8_t *optbuf;
1104 	u_int32_t v;
1105 
1106 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1107 
1108 	/*
1109 	 * If there is no hop-by-hop options header, allocate new one.
1110 	 * If there is one but it doesn't have enough space to store the
1111 	 * jumbo payload option, allocate a cluster to store the whole options.
1112 	 * Otherwise, use it to store the options.
1113 	 */
1114 	if (exthdrs->ip6e_hbh == 0) {
1115 		MGET(mopt, M_DONTWAIT, MT_DATA);
1116 		if (mopt == 0)
1117 			return (ENOBUFS);
1118 		mopt->m_len = JUMBOOPTLEN;
1119 		optbuf = mtod(mopt, u_int8_t *);
1120 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1121 		exthdrs->ip6e_hbh = mopt;
1122 	} else {
1123 		struct ip6_hbh *hbh;
1124 
1125 		mopt = exthdrs->ip6e_hbh;
1126 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1127 			/*
1128 			 * XXX assumption:
1129 			 * - exthdrs->ip6e_hbh is not referenced from places
1130 			 *   other than exthdrs.
1131 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1132 			 */
1133 			int oldoptlen = mopt->m_len;
1134 			struct mbuf *n;
1135 
1136 			/*
1137 			 * XXX: give up if the whole (new) hbh header does
1138 			 * not fit even in an mbuf cluster.
1139 			 */
1140 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1141 				return (ENOBUFS);
1142 
1143 			/*
1144 			 * As a consequence, we must always prepare a cluster
1145 			 * at this point.
1146 			 */
1147 			MGET(n, M_DONTWAIT, MT_DATA);
1148 			if (n) {
1149 				MCLGET(n, M_DONTWAIT);
1150 				if ((n->m_flags & M_EXT) == 0) {
1151 					m_freem(n);
1152 					n = NULL;
1153 				}
1154 			}
1155 			if (!n)
1156 				return (ENOBUFS);
1157 			n->m_len = oldoptlen + JUMBOOPTLEN;
1158 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1159 			    oldoptlen);
1160 			optbuf = mtod(n, u_int8_t *) + oldoptlen;
1161 			m_freem(mopt);
1162 			mopt = exthdrs->ip6e_hbh = n;
1163 		} else {
1164 			optbuf = mtod(mopt, u_int8_t *) + mopt->m_len;
1165 			mopt->m_len += JUMBOOPTLEN;
1166 		}
1167 		optbuf[0] = IP6OPT_PADN;
1168 		optbuf[1] = 0;
1169 
1170 		/*
1171 		 * Adjust the header length according to the pad and
1172 		 * the jumbo payload option.
1173 		 */
1174 		hbh = mtod(mopt, struct ip6_hbh *);
1175 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1176 	}
1177 
1178 	/* fill in the option. */
1179 	optbuf[2] = IP6OPT_JUMBO;
1180 	optbuf[3] = 4;
1181 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1182 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1183 
1184 	/* finally, adjust the packet header length */
1185 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1186 
1187 	return (0);
1188 #undef JUMBOOPTLEN
1189 }
1190 
1191 /*
1192  * Insert fragment header and copy unfragmentable header portions.
1193  */
1194 static int
1195 ip6_insertfraghdr(m0, m, hlen, frghdrp)
1196 	struct mbuf *m0, *m;
1197 	int hlen;
1198 	struct ip6_frag **frghdrp;
1199 {
1200 	struct mbuf *n, *mlast;
1201 
1202 	if (hlen > sizeof(struct ip6_hdr)) {
1203 		n = m_copym(m0, sizeof(struct ip6_hdr),
1204 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1205 		if (n == 0)
1206 			return (ENOBUFS);
1207 		m->m_next = n;
1208 	} else
1209 		n = m;
1210 
1211 	/* Search for the last mbuf of unfragmentable part. */
1212 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1213 		;
1214 
1215 	if ((mlast->m_flags & M_EXT) == 0 &&
1216 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1217 		/* use the trailing space of the last mbuf for the fragment hdr */
1218 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1219 		    mlast->m_len);
1220 		mlast->m_len += sizeof(struct ip6_frag);
1221 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1222 	} else {
1223 		/* allocate a new mbuf for the fragment header */
1224 		struct mbuf *mfrg;
1225 
1226 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1227 		if (mfrg == 0)
1228 			return (ENOBUFS);
1229 		mfrg->m_len = sizeof(struct ip6_frag);
1230 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1231 		mlast->m_next = mfrg;
1232 	}
1233 
1234 	return (0);
1235 }
1236 
1237 static int
1238 ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup, alwaysfragp)
1239 	struct route_in6 *ro_pmtu, *ro;
1240 	struct ifnet *ifp;
1241 	struct in6_addr *dst;
1242 	u_long *mtup;
1243 	int *alwaysfragp;
1244 {
1245 	u_int32_t mtu = 0;
1246 	int alwaysfrag = 0;
1247 	int error = 0;
1248 
1249 	if (ro_pmtu != ro) {
1250 		/* The first hop and the final destination may differ. */
1251 		struct sockaddr_in6 *sa6_dst =
1252 		    (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1253 		if (ro_pmtu->ro_rt &&
1254 		    ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1255 		      !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1256 			RTFREE(ro_pmtu->ro_rt);
1257 			ro_pmtu->ro_rt = (struct rtentry *)NULL;
1258 		}
1259 		if (ro_pmtu->ro_rt == NULL) {
1260 			bzero(sa6_dst, sizeof(*sa6_dst)); /* for safety */
1261 			sa6_dst->sin6_family = AF_INET6;
1262 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1263 			sa6_dst->sin6_addr = *dst;
1264 
1265 			rtalloc((struct route *)ro_pmtu);
1266 		}
1267 	}
1268 	if (ro_pmtu->ro_rt) {
1269 		u_int32_t ifmtu;
1270 
1271 		if (ifp == NULL)
1272 			ifp = ro_pmtu->ro_rt->rt_ifp;
1273 		ifmtu = IN6_LINKMTU(ifp);
1274 		mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1275 		if (mtu == 0)
1276 			mtu = ifmtu;
1277 		else if (mtu < IPV6_MMTU) {
1278 			/*
1279 			 * RFC2460 section 5, last paragraph:
1280 			 * if we record ICMPv6 too big message with
1281 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1282 			 * or smaller, with fragment header attached.
1283 			 * (fragment header is needed regardless from the
1284 			 * packet size, for translators to identify packets)
1285 			 */
1286 			alwaysfrag = 1;
1287 			mtu = IPV6_MMTU;
1288 		} else if (mtu > ifmtu) {
1289 			/*
1290 			 * The MTU on the route is larger than the MTU on
1291 			 * the interface!  This shouldn't happen, unless the
1292 			 * MTU of the interface has been changed after the
1293 			 * interface was brought up.  Change the MTU in the
1294 			 * route to match the interface MTU (as long as the
1295 			 * field isn't locked).
1296 			 */
1297 			mtu = ifmtu;
1298 			if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU))
1299 				ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1300 		}
1301 	} else if (ifp) {
1302 		mtu = IN6_LINKMTU(ifp);
1303 	} else
1304 		error = EHOSTUNREACH; /* XXX */
1305 
1306 	*mtup = mtu;
1307 	if (alwaysfragp)
1308 		*alwaysfragp = alwaysfrag;
1309 	return (error);
1310 }
1311 
1312 /*
1313  * IP6 socket option processing.
1314  */
1315 int
1316 ip6_ctloutput(op, so, level, optname, mp)
1317 	int op;
1318 	struct socket *so;
1319 	int level, optname;
1320 	struct mbuf **mp;
1321 {
1322 	struct in6pcb *in6p = sotoin6pcb(so);
1323 	struct mbuf *m = *mp;
1324 	int optval = 0;
1325 	int error = 0;
1326 	struct proc *p = curproc;	/* XXX */
1327 
1328 	if (level == IPPROTO_IPV6) {
1329 		switch (op) {
1330 		case PRCO_SETOPT:
1331 			switch (optname) {
1332 			case IPV6_PKTOPTIONS:
1333 				/* m is freed in ip6_pcbopts */
1334 				return (ip6_pcbopts(&in6p->in6p_outputopts,
1335 				    m, so));
1336 			case IPV6_HOPOPTS:
1337 			case IPV6_DSTOPTS:
1338 				if (p == 0 || suser(p->p_ucred, &p->p_acflag)) {
1339 					error = EPERM;
1340 					break;
1341 				}
1342 				/* FALLTHROUGH */
1343 			case IPV6_UNICAST_HOPS:
1344 			case IPV6_RECVOPTS:
1345 			case IPV6_RECVRETOPTS:
1346 			case IPV6_RECVDSTADDR:
1347 			case IPV6_PKTINFO:
1348 			case IPV6_HOPLIMIT:
1349 			case IPV6_RTHDR:
1350 			case IPV6_FAITH:
1351 			case IPV6_V6ONLY:
1352 			case IPV6_USE_MIN_MTU:
1353 				if (!m || m->m_len != sizeof(int)) {
1354 					error = EINVAL;
1355 					break;
1356 				}
1357 				optval = *mtod(m, int *);
1358 				switch (optname) {
1359 
1360 				case IPV6_UNICAST_HOPS:
1361 					if (optval < -1 || optval >= 256)
1362 						error = EINVAL;
1363 					else {
1364 						/* -1 = kernel default */
1365 						in6p->in6p_hops = optval;
1366 					}
1367 					break;
1368 #define OPTSET(bit) \
1369 do { \
1370 	if (optval) \
1371 		in6p->in6p_flags |= (bit); \
1372 	else \
1373 		in6p->in6p_flags &= ~(bit); \
1374 } while (/*CONSTCOND*/ 0)
1375 
1376 				case IPV6_RECVOPTS:
1377 					OPTSET(IN6P_RECVOPTS);
1378 					break;
1379 
1380 				case IPV6_RECVRETOPTS:
1381 					OPTSET(IN6P_RECVRETOPTS);
1382 					break;
1383 
1384 				case IPV6_RECVDSTADDR:
1385 					OPTSET(IN6P_RECVDSTADDR);
1386 					break;
1387 
1388 				case IPV6_PKTINFO:
1389 					OPTSET(IN6P_PKTINFO);
1390 					break;
1391 
1392 				case IPV6_HOPLIMIT:
1393 					OPTSET(IN6P_HOPLIMIT);
1394 					break;
1395 
1396 				case IPV6_HOPOPTS:
1397 					OPTSET(IN6P_HOPOPTS);
1398 					break;
1399 
1400 				case IPV6_DSTOPTS:
1401 					OPTSET(IN6P_DSTOPTS);
1402 					break;
1403 
1404 				case IPV6_RTHDR:
1405 					OPTSET(IN6P_RTHDR);
1406 					break;
1407 
1408 				case IPV6_FAITH:
1409 					OPTSET(IN6P_FAITH);
1410 					break;
1411 
1412 				case IPV6_USE_MIN_MTU:
1413 					OPTSET(IN6P_MINMTU);
1414 					break;
1415 
1416 				case IPV6_V6ONLY:
1417 					/*
1418 					 * make setsockopt(IPV6_V6ONLY)
1419 					 * available only prior to bind(2).
1420 					 * see ipng mailing list, Jun 22 2001.
1421 					 */
1422 					if (in6p->in6p_lport ||
1423 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1424 						error = EINVAL;
1425 						break;
1426 					}
1427 #ifdef INET6_BINDV6ONLY
1428 					if (!optval)
1429 						error = EINVAL;
1430 #else
1431 					OPTSET(IN6P_IPV6_V6ONLY);
1432 #endif
1433 					break;
1434 				}
1435 				break;
1436 #undef OPTSET
1437 
1438 			case IPV6_MULTICAST_IF:
1439 			case IPV6_MULTICAST_HOPS:
1440 			case IPV6_MULTICAST_LOOP:
1441 			case IPV6_JOIN_GROUP:
1442 			case IPV6_LEAVE_GROUP:
1443 				error =	ip6_setmoptions(optname,
1444 				    &in6p->in6p_moptions, m);
1445 				break;
1446 
1447 			case IPV6_PORTRANGE:
1448 				optval = *mtod(m, int *);
1449 
1450 				switch (optval) {
1451 				case IPV6_PORTRANGE_DEFAULT:
1452 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1453 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1454 					break;
1455 
1456 				case IPV6_PORTRANGE_HIGH:
1457 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1458 					in6p->in6p_flags |= IN6P_HIGHPORT;
1459 					break;
1460 
1461 				case IPV6_PORTRANGE_LOW:
1462 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1463 					in6p->in6p_flags |= IN6P_LOWPORT;
1464 					break;
1465 
1466 				default:
1467 					error = EINVAL;
1468 					break;
1469 				}
1470 				break;
1471 
1472 #ifdef IPSEC
1473 			case IPV6_IPSEC_POLICY:
1474 			    {
1475 				caddr_t req = NULL;
1476 				size_t len = 0;
1477 
1478 				int priv = 0;
1479 				if (p == 0 || suser(p->p_ucred, &p->p_acflag))
1480 					priv = 0;
1481 				else
1482 					priv = 1;
1483 				if (m) {
1484 					req = mtod(m, caddr_t);
1485 					len = m->m_len;
1486 				}
1487 				error = ipsec6_set_policy(in6p,
1488 				                   optname, req, len, priv);
1489 			    }
1490 				break;
1491 #endif /* IPSEC */
1492 
1493 			default:
1494 				error = ENOPROTOOPT;
1495 				break;
1496 			}
1497 			if (m)
1498 				(void)m_free(m);
1499 			break;
1500 
1501 		case PRCO_GETOPT:
1502 			switch (optname) {
1503 
1504 			case IPV6_OPTIONS:
1505 			case IPV6_RETOPTS:
1506 				error = ENOPROTOOPT;
1507 				break;
1508 
1509 			case IPV6_PKTOPTIONS:
1510 				if (in6p->in6p_options) {
1511 					*mp = m_copym(in6p->in6p_options, 0,
1512 					    M_COPYALL, M_WAIT);
1513 				} else {
1514 					*mp = m_get(M_WAIT, MT_SOOPTS);
1515 					(*mp)->m_len = 0;
1516 				}
1517 				break;
1518 
1519 			case IPV6_HOPOPTS:
1520 			case IPV6_DSTOPTS:
1521 				if (p == 0 || suser(p->p_ucred, &p->p_acflag)) {
1522 					error = EPERM;
1523 					break;
1524 				}
1525 				/* FALLTHROUGH */
1526 			case IPV6_UNICAST_HOPS:
1527 			case IPV6_RECVOPTS:
1528 			case IPV6_RECVRETOPTS:
1529 			case IPV6_RECVDSTADDR:
1530 			case IPV6_PORTRANGE:
1531 			case IPV6_PKTINFO:
1532 			case IPV6_HOPLIMIT:
1533 			case IPV6_RTHDR:
1534 			case IPV6_FAITH:
1535 			case IPV6_V6ONLY:
1536 			case IPV6_USE_MIN_MTU:
1537 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1538 				m->m_len = sizeof(int);
1539 				switch (optname) {
1540 
1541 				case IPV6_UNICAST_HOPS:
1542 					optval = in6p->in6p_hops;
1543 					break;
1544 
1545 #define OPTBIT(bit) (in6p->in6p_flags & bit ? 1 : 0)
1546 
1547 				case IPV6_RECVOPTS:
1548 					optval = OPTBIT(IN6P_RECVOPTS);
1549 					break;
1550 
1551 				case IPV6_RECVRETOPTS:
1552 					optval = OPTBIT(IN6P_RECVRETOPTS);
1553 					break;
1554 
1555 				case IPV6_RECVDSTADDR:
1556 					optval = OPTBIT(IN6P_RECVDSTADDR);
1557 					break;
1558 
1559 				case IPV6_PORTRANGE:
1560 				    {
1561 					int flags;
1562 					flags = in6p->in6p_flags;
1563 					if (flags & IN6P_HIGHPORT)
1564 						optval = IPV6_PORTRANGE_HIGH;
1565 					else if (flags & IN6P_LOWPORT)
1566 						optval = IPV6_PORTRANGE_LOW;
1567 					else
1568 						optval = 0;
1569 					break;
1570 				    }
1571 
1572 				case IPV6_PKTINFO:
1573 					optval = OPTBIT(IN6P_PKTINFO);
1574 					break;
1575 
1576 				case IPV6_HOPLIMIT:
1577 					optval = OPTBIT(IN6P_HOPLIMIT);
1578 					break;
1579 
1580 				case IPV6_HOPOPTS:
1581 					optval = OPTBIT(IN6P_HOPOPTS);
1582 					break;
1583 
1584 				case IPV6_DSTOPTS:
1585 					optval = OPTBIT(IN6P_DSTOPTS);
1586 					break;
1587 
1588 				case IPV6_RTHDR:
1589 					optval = OPTBIT(IN6P_RTHDR);
1590 					break;
1591 
1592 				case IPV6_FAITH:
1593 					optval = OPTBIT(IN6P_FAITH);
1594 					break;
1595 
1596 				case IPV6_V6ONLY:
1597 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
1598 					break;
1599 
1600 				case IPV6_USE_MIN_MTU:
1601 					optval = OPTBIT(IN6P_MINMTU);
1602 					break;
1603 				}
1604 				*mtod(m, int *) = optval;
1605 				break;
1606 
1607 			case IPV6_MULTICAST_IF:
1608 			case IPV6_MULTICAST_HOPS:
1609 			case IPV6_MULTICAST_LOOP:
1610 			case IPV6_JOIN_GROUP:
1611 			case IPV6_LEAVE_GROUP:
1612 				error = ip6_getmoptions(optname, in6p->in6p_moptions, mp);
1613 				break;
1614 
1615 #if 0	/* defined(IPSEC) */
1616 			/* XXX: code broken */
1617 			case IPV6_IPSEC_POLICY:
1618 			{
1619 				caddr_t req = NULL;
1620 				size_t len = 0;
1621 
1622 				if (m) {
1623 					req = mtod(m, caddr_t);
1624 					len = m->m_len;
1625 				}
1626 				error = ipsec6_get_policy(in6p, req, len, mp);
1627 				break;
1628 			}
1629 #endif /* IPSEC */
1630 
1631 			default:
1632 				error = ENOPROTOOPT;
1633 				break;
1634 			}
1635 			break;
1636 		}
1637 	} else {
1638 		error = EINVAL;
1639 		if (op == PRCO_SETOPT && *mp)
1640 			(void)m_free(*mp);
1641 	}
1642 	return (error);
1643 }
1644 
1645 int
1646 ip6_raw_ctloutput(op, so, level, optname, mp)
1647 	int op;
1648 	struct socket *so;
1649 	int level, optname;
1650 	struct mbuf **mp;
1651 {
1652 	int error = 0, optval, optlen;
1653 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
1654 	struct in6pcb *in6p = sotoin6pcb(so);
1655 	struct mbuf *m = *mp;
1656 
1657 	optlen = m ? m->m_len : 0;
1658 
1659 	if (level != IPPROTO_IPV6) {
1660 		if (op == PRCO_SETOPT && *mp)
1661 			(void)m_free(*mp);
1662 		return (EINVAL);
1663 	}
1664 
1665 	switch (optname) {
1666 	case IPV6_CHECKSUM:
1667 		/*
1668 		 * For ICMPv6 sockets, no modification allowed for checksum
1669 		 * offset, permit "no change" values to help existing apps.
1670 		 *
1671 		 * XXX 2292bis says: "An attempt to set IPV6_CHECKSUM
1672 		 * for an ICMPv6 socket will fail."
1673 		 * The current behavior does not meet 2292bis.
1674 		 */
1675 		switch (op) {
1676 		case PRCO_SETOPT:
1677 			if (optlen != sizeof(int)) {
1678 				error = EINVAL;
1679 				break;
1680 			}
1681 			optval = *mtod(m, int *);
1682 			if ((optval % 2) != 0) {
1683 				/* the API assumes even offset values */
1684 				error = EINVAL;
1685 			} else if (so->so_proto->pr_protocol ==
1686 			    IPPROTO_ICMPV6) {
1687 				if (optval != icmp6off)
1688 					error = EINVAL;
1689 			} else
1690 				in6p->in6p_cksum = optval;
1691 			break;
1692 
1693 		case PRCO_GETOPT:
1694 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
1695 				optval = icmp6off;
1696 			else
1697 				optval = in6p->in6p_cksum;
1698 
1699 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1700 			m->m_len = sizeof(int);
1701 			*mtod(m, int *) = optval;
1702 			break;
1703 
1704 		default:
1705 			error = EINVAL;
1706 			break;
1707 		}
1708 		break;
1709 
1710 	default:
1711 		error = ENOPROTOOPT;
1712 		break;
1713 	}
1714 
1715 	if (op == PRCO_SETOPT && m)
1716 		(void)m_free(m);
1717 
1718 	return (error);
1719 }
1720 
1721 /*
1722  * Set up IP6 options in pcb for insertion in output packets.
1723  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1724  * with destination address if source routed.
1725  */
1726 static int
1727 ip6_pcbopts(pktopt, m, so)
1728 	struct ip6_pktopts **pktopt;
1729 	struct mbuf *m;
1730 	struct socket *so;
1731 {
1732 	struct ip6_pktopts *opt = *pktopt;
1733 	int error = 0;
1734 	struct proc *p = curproc;	/* XXX */
1735 	int priv = 0;
1736 
1737 	/* turn off any old options. */
1738 	if (opt) {
1739 		if (opt->ip6po_m)
1740 			(void)m_free(opt->ip6po_m);
1741 	} else
1742 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
1743 	*pktopt = 0;
1744 
1745 	if (!m || m->m_len == 0) {
1746 		/*
1747 		 * Only turning off any previous options.
1748 		 */
1749 		free(opt, M_IP6OPT);
1750 		if (m)
1751 			(void)m_free(m);
1752 		return (0);
1753 	}
1754 
1755 	/*  set options specified by user. */
1756 	if (p && !suser(p->p_ucred, &p->p_acflag))
1757 		priv = 1;
1758 	if ((error = ip6_setpktoptions(m, opt, priv)) != 0) {
1759 		(void)m_free(m);
1760 		free(opt, M_IP6OPT);
1761 		return (error);
1762 	}
1763 	*pktopt = opt;
1764 	return (0);
1765 }
1766 
1767 /*
1768  * Set the IP6 multicast options in response to user setsockopt().
1769  */
1770 static int
1771 ip6_setmoptions(optname, im6op, m)
1772 	int optname;
1773 	struct ip6_moptions **im6op;
1774 	struct mbuf *m;
1775 {
1776 	int error = 0;
1777 	u_int loop, ifindex;
1778 	struct ipv6_mreq *mreq;
1779 	struct ifnet *ifp;
1780 	struct ip6_moptions *im6o = *im6op;
1781 	struct route_in6 ro;
1782 	struct sockaddr_in6 *dst;
1783 	struct in6_multi_mship *imm;
1784 	struct proc *p = curproc;	/* XXX */
1785 
1786 	if (im6o == NULL) {
1787 		/*
1788 		 * No multicast option buffer attached to the pcb;
1789 		 * allocate one and initialize to default values.
1790 		 */
1791 		im6o = (struct ip6_moptions *)
1792 			malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
1793 
1794 		if (im6o == NULL)
1795 			return (ENOBUFS);
1796 		*im6op = im6o;
1797 		im6o->im6o_multicast_ifp = NULL;
1798 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1799 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
1800 		LIST_INIT(&im6o->im6o_memberships);
1801 	}
1802 
1803 	switch (optname) {
1804 
1805 	case IPV6_MULTICAST_IF:
1806 		/*
1807 		 * Select the interface for outgoing multicast packets.
1808 		 */
1809 		if (m == NULL || m->m_len != sizeof(u_int)) {
1810 			error = EINVAL;
1811 			break;
1812 		}
1813 		bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
1814 		if (ifindex != 0) {
1815 			if (ifindex < 0 || if_indexlim <= ifindex ||
1816 			    !ifindex2ifnet[ifindex]) {
1817 				error = ENXIO;	/* XXX EINVAL? */
1818 				break;
1819 			}
1820 			ifp = ifindex2ifnet[ifindex];
1821 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
1822 				error = EADDRNOTAVAIL;
1823 				break;
1824 			}
1825 		} else
1826 			ifp = NULL;
1827 		im6o->im6o_multicast_ifp = ifp;
1828 		break;
1829 
1830 	case IPV6_MULTICAST_HOPS:
1831 	    {
1832 		/*
1833 		 * Set the IP6 hoplimit for outgoing multicast packets.
1834 		 */
1835 		int optval;
1836 		if (m == NULL || m->m_len != sizeof(int)) {
1837 			error = EINVAL;
1838 			break;
1839 		}
1840 		bcopy(mtod(m, u_int *), &optval, sizeof(optval));
1841 		if (optval < -1 || optval >= 256)
1842 			error = EINVAL;
1843 		else if (optval == -1)
1844 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1845 		else
1846 			im6o->im6o_multicast_hlim = optval;
1847 		break;
1848 	    }
1849 
1850 	case IPV6_MULTICAST_LOOP:
1851 		/*
1852 		 * Set the loopback flag for outgoing multicast packets.
1853 		 * Must be zero or one.
1854 		 */
1855 		if (m == NULL || m->m_len != sizeof(u_int)) {
1856 			error = EINVAL;
1857 			break;
1858 		}
1859 		bcopy(mtod(m, u_int *), &loop, sizeof(loop));
1860 		if (loop > 1) {
1861 			error = EINVAL;
1862 			break;
1863 		}
1864 		im6o->im6o_multicast_loop = loop;
1865 		break;
1866 
1867 	case IPV6_JOIN_GROUP:
1868 		/*
1869 		 * Add a multicast group membership.
1870 		 * Group must be a valid IP6 multicast address.
1871 		 */
1872 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
1873 			error = EINVAL;
1874 			break;
1875 		}
1876 		mreq = mtod(m, struct ipv6_mreq *);
1877 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
1878 			/*
1879 			 * We use the unspecified address to specify to accept
1880 			 * all multicast addresses. Only super user is allowed
1881 			 * to do this.
1882 			 */
1883 			if (suser(p->p_ucred, &p->p_acflag))
1884 			{
1885 				error = EACCES;
1886 				break;
1887 			}
1888 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
1889 			error = EINVAL;
1890 			break;
1891 		}
1892 
1893 		/*
1894 		 * If the interface is specified, validate it.
1895 		 * If no interface was explicitly specified, choose an
1896 		 * appropriate one according to the given multicast address.
1897 		 */
1898 		if (mreq->ipv6mr_interface != 0) {
1899 			if (mreq->ipv6mr_interface < 0 ||
1900 			    if_indexlim <= mreq->ipv6mr_interface ||
1901 			    !ifindex2ifnet[mreq->ipv6mr_interface]) {
1902 				error = ENXIO;	/* XXX EINVAL? */
1903 				break;
1904 			}
1905 			ifp = ifindex2ifnet[mreq->ipv6mr_interface];
1906 		} else {
1907 			/*
1908 			 * If the multicast address is in node-local scope,
1909 			 * the interface should be a loopback interface.
1910 			 * Otherwise, look up the routing table for the
1911 			 * address, and choose the outgoing interface.
1912 			 *   XXX: is it a good approach?
1913 			 */
1914 			if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
1915 				ifp = lo0ifp;
1916 			} else {
1917 				ro.ro_rt = NULL;
1918 				dst = (struct sockaddr_in6 *)&ro.ro_dst;
1919 				bzero(dst, sizeof(*dst));
1920 				dst->sin6_len = sizeof(struct sockaddr_in6);
1921 				dst->sin6_family = AF_INET6;
1922 				dst->sin6_addr = mreq->ipv6mr_multiaddr;
1923 				rtalloc((struct route *)&ro);
1924 				if (ro.ro_rt == NULL) {
1925 					error = EADDRNOTAVAIL;
1926 					break;
1927 				}
1928 				ifp = ro.ro_rt->rt_ifp;
1929 				rtfree(ro.ro_rt);
1930 			}
1931 		}
1932 
1933 		/*
1934 		 * See if we found an interface, and confirm that it
1935 		 * supports multicast
1936 		 */
1937 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1938 			error = EADDRNOTAVAIL;
1939 			break;
1940 		}
1941 		/*
1942 		 * Put interface index into the multicast address,
1943 		 * if the address has link-local scope.
1944 		 */
1945 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
1946 			mreq->ipv6mr_multiaddr.s6_addr16[1] =
1947 			    htons(ifp->if_index);
1948 		}
1949 		/*
1950 		 * See if the membership already exists.
1951 		 */
1952 		for (imm = im6o->im6o_memberships.lh_first;
1953 		     imm != NULL; imm = imm->i6mm_chain.le_next)
1954 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
1955 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
1956 			    &mreq->ipv6mr_multiaddr))
1957 				break;
1958 		if (imm != NULL) {
1959 			error = EADDRINUSE;
1960 			break;
1961 		}
1962 		/*
1963 		 * Everything looks good; add a new record to the multicast
1964 		 * address list for the given interface.
1965 		 */
1966 		imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error);
1967 		if (!imm)
1968 			break;
1969 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
1970 		break;
1971 
1972 	case IPV6_LEAVE_GROUP:
1973 		/*
1974 		 * Drop a multicast group membership.
1975 		 * Group must be a valid IP6 multicast address.
1976 		 */
1977 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
1978 			error = EINVAL;
1979 			break;
1980 		}
1981 		mreq = mtod(m, struct ipv6_mreq *);
1982 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
1983 			if (suser(p->p_ucred, &p->p_acflag))
1984 			{
1985 				error = EACCES;
1986 				break;
1987 			}
1988 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
1989 			error = EINVAL;
1990 			break;
1991 		}
1992 		/*
1993 		 * If an interface address was specified, get a pointer
1994 		 * to its ifnet structure.
1995 		 */
1996 		if (mreq->ipv6mr_interface != 0) {
1997 			if (mreq->ipv6mr_interface < 0 ||
1998 			    if_indexlim <= mreq->ipv6mr_interface ||
1999 			    !ifindex2ifnet[mreq->ipv6mr_interface]) {
2000 				error = ENXIO;	/* XXX EINVAL? */
2001 				break;
2002 			}
2003 			ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2004 		} else
2005 			ifp = NULL;
2006 		/*
2007 		 * Put interface index into the multicast address,
2008 		 * if the address has link-local scope.
2009 		 */
2010 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2011 			mreq->ipv6mr_multiaddr.s6_addr16[1] =
2012 			    htons(mreq->ipv6mr_interface);
2013 		}
2014 		/*
2015 		 * Find the membership in the membership list.
2016 		 */
2017 		for (imm = im6o->im6o_memberships.lh_first;
2018 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
2019 			if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2020 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2021 			    &mreq->ipv6mr_multiaddr))
2022 				break;
2023 		}
2024 		if (imm == NULL) {
2025 			/* Unable to resolve interface */
2026 			error = EADDRNOTAVAIL;
2027 			break;
2028 		}
2029 		/*
2030 		 * Give up the multicast address record to which the
2031 		 * membership points.
2032 		 */
2033 		LIST_REMOVE(imm, i6mm_chain);
2034 		in6_leavegroup(imm);
2035 		break;
2036 
2037 	default:
2038 		error = EOPNOTSUPP;
2039 		break;
2040 	}
2041 
2042 	/*
2043 	 * If all options have default values, no need to keep the mbuf.
2044 	 */
2045 	if (im6o->im6o_multicast_ifp == NULL &&
2046 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2047 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2048 	    im6o->im6o_memberships.lh_first == NULL) {
2049 		free(*im6op, M_IPMOPTS);
2050 		*im6op = NULL;
2051 	}
2052 
2053 	return (error);
2054 }
2055 
2056 /*
2057  * Return the IP6 multicast options in response to user getsockopt().
2058  */
2059 static int
2060 ip6_getmoptions(optname, im6o, mp)
2061 	int optname;
2062 	struct ip6_moptions *im6o;
2063 	struct mbuf **mp;
2064 {
2065 	u_int *hlim, *loop, *ifindex;
2066 
2067 	*mp = m_get(M_WAIT, MT_SOOPTS);
2068 
2069 	switch (optname) {
2070 
2071 	case IPV6_MULTICAST_IF:
2072 		ifindex = mtod(*mp, u_int *);
2073 		(*mp)->m_len = sizeof(u_int);
2074 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2075 			*ifindex = 0;
2076 		else
2077 			*ifindex = im6o->im6o_multicast_ifp->if_index;
2078 		return (0);
2079 
2080 	case IPV6_MULTICAST_HOPS:
2081 		hlim = mtod(*mp, u_int *);
2082 		(*mp)->m_len = sizeof(u_int);
2083 		if (im6o == NULL)
2084 			*hlim = ip6_defmcasthlim;
2085 		else
2086 			*hlim = im6o->im6o_multicast_hlim;
2087 		return (0);
2088 
2089 	case IPV6_MULTICAST_LOOP:
2090 		loop = mtod(*mp, u_int *);
2091 		(*mp)->m_len = sizeof(u_int);
2092 		if (im6o == NULL)
2093 			*loop = ip6_defmcasthlim;
2094 		else
2095 			*loop = im6o->im6o_multicast_loop;
2096 		return (0);
2097 
2098 	default:
2099 		return (EOPNOTSUPP);
2100 	}
2101 }
2102 
2103 /*
2104  * Discard the IP6 multicast options.
2105  */
2106 void
2107 ip6_freemoptions(im6o)
2108 	struct ip6_moptions *im6o;
2109 {
2110 	struct in6_multi_mship *imm;
2111 
2112 	if (im6o == NULL)
2113 		return;
2114 
2115 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2116 		LIST_REMOVE(imm, i6mm_chain);
2117 		in6_leavegroup(imm);
2118 	}
2119 	free(im6o, M_IPMOPTS);
2120 }
2121 
2122 /*
2123  * Set IPv6 outgoing packet options based on advanced API.
2124  */
2125 int
2126 ip6_setpktoptions(control, opt, priv)
2127 	struct mbuf *control;
2128 	struct ip6_pktopts *opt;
2129 	int priv;
2130 {
2131 	struct cmsghdr *cm = 0;
2132 
2133 	if (control == 0 || opt == 0)
2134 		return (EINVAL);
2135 
2136 	bzero(opt, sizeof(*opt));
2137 	opt->ip6po_hlim = -1; /* -1 means to use default hop limit */
2138 
2139 	/*
2140 	 * XXX: Currently, we assume all the optional information is stored
2141 	 * in a single mbuf.
2142 	 */
2143 	if (control->m_next)
2144 		return (EINVAL);
2145 
2146 	opt->ip6po_m = control;
2147 
2148 	for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2149 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2150 		cm = mtod(control, struct cmsghdr *);
2151 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2152 			return (EINVAL);
2153 		if (cm->cmsg_level != IPPROTO_IPV6)
2154 			continue;
2155 
2156 		switch (cm->cmsg_type) {
2157 		case IPV6_PKTINFO:
2158 			if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo)))
2159 				return (EINVAL);
2160 			opt->ip6po_pktinfo = (struct in6_pktinfo *)CMSG_DATA(cm);
2161 			if (opt->ip6po_pktinfo->ipi6_ifindex &&
2162 			    IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr))
2163 				opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] =
2164 					htons(opt->ip6po_pktinfo->ipi6_ifindex);
2165 
2166 			if (opt->ip6po_pktinfo->ipi6_ifindex >= if_indexlim ||
2167 			    opt->ip6po_pktinfo->ipi6_ifindex < 0)
2168 				return (ENXIO);
2169 			if (opt->ip6po_pktinfo->ipi6_ifindex > 0 &&
2170 			    !ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex])
2171 				return (ENXIO);
2172 
2173 			/*
2174 			 * Check if the requested source address is indeed a
2175 			 * unicast address assigned to the node, and can be
2176 			 * used as the packet's source address.
2177 			 */
2178 			if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
2179 				struct ifaddr *ia;
2180 				struct in6_ifaddr *ia6;
2181 				struct sockaddr_in6 sin6;
2182 
2183 				bzero(&sin6, sizeof(sin6));
2184 				sin6.sin6_len = sizeof(sin6);
2185 				sin6.sin6_family = AF_INET6;
2186 				sin6.sin6_addr =
2187 					opt->ip6po_pktinfo->ipi6_addr;
2188 				ia = ifa_ifwithaddr(sin6tosa(&sin6));
2189 				if (ia == NULL ||
2190 				    (opt->ip6po_pktinfo->ipi6_ifindex &&
2191 				     (ia->ifa_ifp->if_index !=
2192 				      opt->ip6po_pktinfo->ipi6_ifindex))) {
2193 					return (EADDRNOTAVAIL);
2194 				}
2195 				ia6 = (struct in6_ifaddr *)ia;
2196 				if ((ia6->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) != 0) {
2197 					return (EADDRNOTAVAIL);
2198 				}
2199 
2200 				/*
2201 				 * Check if the requested source address is
2202 				 * indeed a unicast address assigned to the
2203 				 * node.
2204 				 */
2205 				if (IN6_IS_ADDR_MULTICAST(&opt->ip6po_pktinfo->ipi6_addr))
2206 					return (EADDRNOTAVAIL);
2207 			}
2208 			break;
2209 
2210 		case IPV6_HOPLIMIT:
2211 			if (cm->cmsg_len != CMSG_LEN(sizeof(int)))
2212 				return (EINVAL);
2213 			else {
2214 				int t;
2215 
2216 				bcopy(CMSG_DATA(cm), &t, sizeof(t));
2217 				if (t < -1 || t > 255)
2218 					return (EINVAL);
2219 				opt->ip6po_hlim = t;
2220 			}
2221 			break;
2222 
2223 		case IPV6_NEXTHOP:
2224 			if (!priv)
2225 				return (EPERM);
2226 
2227 			/* check if cmsg_len is large enough for sa_len */
2228 			if (cm->cmsg_len < sizeof(u_char) ||
2229 			    cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm)))
2230 				return (EINVAL);
2231 
2232 			opt->ip6po_nexthop = (struct sockaddr *)CMSG_DATA(cm);
2233 
2234 			break;
2235 
2236 		case IPV6_HOPOPTS:
2237 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh)))
2238 				return (EINVAL);
2239 			else {
2240 				struct  ip6_hbh *t;
2241 
2242 				t = (struct ip6_hbh *)CMSG_DATA(cm);
2243 				if (cm->cmsg_len !=
2244 				    CMSG_LEN((t->ip6h_len + 1) << 3))
2245 					return (EINVAL);
2246 				opt->ip6po_hbh = t;
2247 			}
2248 			break;
2249 
2250 		case IPV6_DSTOPTS:
2251 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest)))
2252 				return (EINVAL);
2253 
2254 			/*
2255 			 * If there is no routing header yet, the destination
2256 			 * options header should be put on the 1st part.
2257 			 * Otherwise, the header should be on the 2nd part.
2258 			 * (See RFC 2460, section 4.1)
2259 			 */
2260 			if (opt->ip6po_rthdr == NULL) {
2261 				struct ip6_dest *t;
2262 
2263 				t = (struct ip6_dest *)CMSG_DATA(cm);
2264 				if (cm->cmsg_len !=
2265 				    CMSG_LEN((t->ip6d_len + 1) << 3));
2266 					return (EINVAL);
2267 				opt->ip6po_dest1 = t;
2268 			}
2269 			else {
2270 				struct ip6_dest *t;
2271 
2272 				t = (struct ip6_dest *)CMSG_DATA(cm);
2273 				if (cm->cmsg_len !=
2274 				    CMSG_LEN((opt->ip6po_dest2->ip6d_len + 1) << 3))
2275 					return (EINVAL);
2276 				opt->ip6po_dest2 = t;
2277 			}
2278 			break;
2279 
2280 		case IPV6_RTHDR:
2281 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr)))
2282 				return (EINVAL);
2283 			else {
2284 				struct ip6_rthdr *t;
2285 
2286 				t = (struct ip6_rthdr *)CMSG_DATA(cm);
2287 				if (cm->cmsg_len !=
2288 				    CMSG_LEN((t->ip6r_len + 1) << 3))
2289 					return (EINVAL);
2290 				switch (t->ip6r_type) {
2291 				case IPV6_RTHDR_TYPE_0:
2292 					if (t->ip6r_segleft == 0)
2293 						return (EINVAL);
2294 					break;
2295 				default:
2296 					return (EINVAL);
2297 				}
2298 				opt->ip6po_rthdr = t;
2299 			}
2300 			break;
2301 
2302 		default:
2303 			return (ENOPROTOOPT);
2304 		}
2305 	}
2306 
2307 	return (0);
2308 }
2309 
2310 /*
2311  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2312  * packet to the input queue of a specified interface.  Note that this
2313  * calls the output routine of the loopback "driver", but with an interface
2314  * pointer that might NOT be lo0ifp -- easier than replicating that code here.
2315  */
2316 void
2317 ip6_mloopback(ifp, m, dst)
2318 	struct ifnet *ifp;
2319 	struct mbuf *m;
2320 	struct sockaddr_in6 *dst;
2321 {
2322 	struct mbuf *copym;
2323 	struct ip6_hdr *ip6;
2324 
2325 	copym = m_copy(m, 0, M_COPYALL);
2326 	if (copym == NULL)
2327 		return;
2328 
2329 	/*
2330 	 * Make sure to deep-copy IPv6 header portion in case the data
2331 	 * is in an mbuf cluster, so that we can safely override the IPv6
2332 	 * header portion later.
2333 	 */
2334 	if ((copym->m_flags & M_EXT) != 0 ||
2335 	    copym->m_len < sizeof(struct ip6_hdr)) {
2336 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
2337 		if (copym == NULL)
2338 			return;
2339 	}
2340 
2341 #ifdef DIAGNOSTIC
2342 	if (copym->m_len < sizeof(*ip6)) {
2343 		m_freem(copym);
2344 		return;
2345 	}
2346 #endif
2347 
2348 	ip6 = mtod(copym, struct ip6_hdr *);
2349 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
2350 		ip6->ip6_src.s6_addr16[1] = 0;
2351 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
2352 		ip6->ip6_dst.s6_addr16[1] = 0;
2353 
2354 	(void)looutput(ifp, copym, (struct sockaddr *)dst, NULL);
2355 }
2356 
2357 /*
2358  * Chop IPv6 header off from the payload.
2359  */
2360 static int
2361 ip6_splithdr(m, exthdrs)
2362 	struct mbuf *m;
2363 	struct ip6_exthdrs *exthdrs;
2364 {
2365 	struct mbuf *mh;
2366 	struct ip6_hdr *ip6;
2367 
2368 	ip6 = mtod(m, struct ip6_hdr *);
2369 	if (m->m_len > sizeof(*ip6)) {
2370 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
2371 		if (mh == 0) {
2372 			m_freem(m);
2373 			return ENOBUFS;
2374 		}
2375 		M_COPY_PKTHDR(mh, m);
2376 		MH_ALIGN(mh, sizeof(*ip6));
2377 		m_tag_delete_chain(m, NULL);
2378 		m->m_flags &= ~M_PKTHDR;
2379 		m->m_len -= sizeof(*ip6);
2380 		m->m_data += sizeof(*ip6);
2381 		mh->m_next = m;
2382 		m = mh;
2383 		m->m_len = sizeof(*ip6);
2384 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2385 	}
2386 	exthdrs->ip6e_ip6 = m;
2387 	return 0;
2388 }
2389 
2390 /*
2391  * Compute IPv6 extension header length.
2392  */
2393 int
2394 ip6_optlen(in6p)
2395 	struct in6pcb *in6p;
2396 {
2397 	int len;
2398 
2399 	if (!in6p->in6p_outputopts)
2400 		return 0;
2401 
2402 	len = 0;
2403 #define elen(x) \
2404     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2405 
2406 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
2407 	len += elen(in6p->in6p_outputopts->ip6po_dest1);
2408 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
2409 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
2410 	return len;
2411 #undef elen
2412 }
2413