xref: /netbsd-src/sys/netinet6/ip6_output.c (revision 4472dbe5e3bd91ef2540bada7a7ca7384627ff9b)
1 /*	$NetBSD: ip6_output.c,v 1.22 2000/06/03 14:36:37 itojun Exp $	*/
2 /*	$KAME: ip6_output.c,v 1.109 2000/05/31 05:03:09 jinmei 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. All advertising materials mentioning features or use of this software
46  *    must display the following acknowledgement:
47  *	This product includes software developed by the University of
48  *	California, Berkeley and its contributors.
49  * 4. Neither the name of the University nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63  * SUCH DAMAGE.
64  *
65  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
66  */
67 
68 #include "opt_inet.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 
82 #include <net/if.h>
83 #include <net/route.h>
84 #ifdef PFIL_HOOKS
85 #include <net/pfil.h>
86 #endif
87 
88 #include <netinet/in.h>
89 #include <netinet/in_var.h>
90 #include <netinet/ip6.h>
91 #include <netinet/icmp6.h>
92 #include <netinet6/ip6_var.h>
93 #include <netinet6/in6_pcb.h>
94 #include <netinet6/nd6.h>
95 
96 #ifdef IPSEC
97 #include <netinet6/ipsec.h>
98 #include <netkey/key.h>
99 #include <netkey/key_debug.h>
100 #endif /* IPSEC */
101 
102 #include "loop.h"
103 
104 #include <net/net_osdep.h>
105 
106 #ifdef IPV6FIREWALL
107 #include <netinet6/ip6_fw.h>
108 #endif
109 
110 struct ip6_exthdrs {
111 	struct mbuf *ip6e_ip6;
112 	struct mbuf *ip6e_hbh;
113 	struct mbuf *ip6e_dest1;
114 	struct mbuf *ip6e_rthdr;
115 	struct mbuf *ip6e_dest2;
116 };
117 
118 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
119 			    struct socket *));
120 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *));
121 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **));
122 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int));
123 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
124 				  struct ip6_frag **));
125 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t));
126 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *));
127 
128 extern struct ifnet loif[NLOOP];
129 
130 /*
131  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
132  * header (with pri, len, nxt, hlim, src, dst).
133  * This function may modify ver and hlim only.
134  * The mbuf chain containing the packet will be freed.
135  * The mbuf opt, if present, will not be freed.
136  */
137 int
138 ip6_output(m0, opt, ro, flags, im6o, ifpp)
139 	struct mbuf *m0;
140 	struct ip6_pktopts *opt;
141 	struct route_in6 *ro;
142 	int flags;
143 	struct ip6_moptions *im6o;
144 	struct ifnet **ifpp;		/* XXX: just for statistics */
145 {
146 	struct ip6_hdr *ip6, *mhip6;
147 	struct ifnet *ifp, *origifp;
148 	struct mbuf *m = m0;
149 	int hlen, tlen, len, off;
150 	struct route_in6 ip6route;
151 	struct sockaddr_in6 *dst;
152 	int error = 0;
153 	struct in6_ifaddr *ia;
154 	u_long mtu;
155 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
156 	struct ip6_exthdrs exthdrs;
157 	struct in6_addr finaldst;
158 	struct route_in6 *ro_pmtu = NULL;
159 	int hdrsplit = 0;
160 	int needipsec = 0;
161 #ifdef PFIL_HOOKS
162 	struct packet_filter_hook *pfh;
163 	struct mbuf *m1;
164 	int rv;
165 #endif /* PFIL_HOOKS */
166 #ifdef IPSEC
167 	int needipsectun = 0;
168 	struct socket *so;
169 	struct secpolicy *sp = NULL;
170 
171 	/* for AH processing. stupid to have "socket" variable in IP layer... */
172 	so = ipsec_getsocket(m);
173 	ipsec_setsocket(m, NULL);
174 	ip6 = mtod(m, struct ip6_hdr *);
175 #endif /* IPSEC */
176 
177 #define MAKE_EXTHDR(hp, mp)						\
178     do {								\
179 	if (hp) {							\
180 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
181 		error = ip6_copyexthdr((mp), (caddr_t)(hp), 		\
182 				       ((eh)->ip6e_len + 1) << 3);	\
183 		if (error)						\
184 			goto freehdrs;					\
185 	}								\
186     } while (0)
187 
188 	bzero(&exthdrs, sizeof(exthdrs));
189 	if (opt) {
190 		/* Hop-by-Hop options header */
191 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
192 		/* Destination options header(1st part) */
193 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
194 		/* Routing header */
195 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
196 		/* Destination options header(2nd part) */
197 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
198 	}
199 
200 #ifdef IPSEC
201 	/* get a security policy for this packet */
202 	if (so == NULL)
203 		sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
204 	else
205 		sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
206 
207 	if (sp == NULL) {
208 		ipsec6stat.out_inval++;
209 		goto bad;
210 	}
211 
212 	error = 0;
213 
214 	/* check policy */
215 	switch (sp->policy) {
216 	case IPSEC_POLICY_DISCARD:
217 		/*
218 		 * This packet is just discarded.
219 		 */
220 		ipsec6stat.out_polvio++;
221 		goto bad;
222 
223 	case IPSEC_POLICY_BYPASS:
224 	case IPSEC_POLICY_NONE:
225 		/* no need to do IPsec. */
226 		needipsec = 0;
227 		break;
228 
229 	case IPSEC_POLICY_IPSEC:
230 		if (sp->req == NULL) {
231 			/* XXX should be panic ? */
232 			printf("ip6_output: No IPsec request specified.\n");
233 			error = EINVAL;
234 			goto bad;
235 		}
236 		needipsec = 1;
237 		break;
238 
239 	case IPSEC_POLICY_ENTRUST:
240 	default:
241 		printf("ip6_output: Invalid policy found. %d\n", sp->policy);
242 	}
243 #endif /* IPSEC */
244 
245 	/*
246 	 * Calculate the total length of the extension header chain.
247 	 * Keep the length of the unfragmentable part for fragmentation.
248 	 */
249 	optlen = 0;
250 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
251 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
252 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
253 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
254 	/* NOTE: we don't add AH/ESP length here. do that later. */
255 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
256 
257 	/*
258 	 * If we need IPsec, or there is at least one extension header,
259 	 * separate IP6 header from the payload.
260 	 */
261 	if ((needipsec || optlen) && !hdrsplit) {
262 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
263 			m = NULL;
264 			goto freehdrs;
265 		}
266 		m = exthdrs.ip6e_ip6;
267 		hdrsplit++;
268 	}
269 
270 	/* adjust pointer */
271 	ip6 = mtod(m, struct ip6_hdr *);
272 
273 	/* adjust mbuf packet header length */
274 	m->m_pkthdr.len += optlen;
275 	plen = m->m_pkthdr.len - sizeof(*ip6);
276 
277 	/* If this is a jumbo payload, insert a jumbo payload option. */
278 	if (plen > IPV6_MAXPACKET) {
279 		if (!hdrsplit) {
280 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
281 				m = NULL;
282 				goto freehdrs;
283 			}
284 			m = exthdrs.ip6e_ip6;
285 			hdrsplit++;
286 		}
287 		/* adjust pointer */
288 		ip6 = mtod(m, struct ip6_hdr *);
289 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
290 			goto freehdrs;
291 		ip6->ip6_plen = 0;
292 	} else
293 		ip6->ip6_plen = htons(plen);
294 
295 	/*
296 	 * Concatenate headers and fill in next header fields.
297 	 * Here we have, on "m"
298 	 *	IPv6 payload
299 	 * and we insert headers accordingly.  Finally, we should be getting:
300 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
301 	 *
302 	 * during the header composing process, "m" points to IPv6 header.
303 	 * "mprev" points to an extension header prior to esp.
304 	 */
305 	{
306 		u_char *nexthdrp = &ip6->ip6_nxt;
307 		struct mbuf *mprev = m;
308 
309 		/*
310 		 * we treat dest2 specially.  this makes IPsec processing
311 		 * much easier.
312 		 *
313 		 * result: IPv6 dest2 payload
314 		 * m and mprev will point to IPv6 header.
315 		 */
316 		if (exthdrs.ip6e_dest2) {
317 			if (!hdrsplit)
318 				panic("assumption failed: hdr not split");
319 			exthdrs.ip6e_dest2->m_next = m->m_next;
320 			m->m_next = exthdrs.ip6e_dest2;
321 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
322 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
323 		}
324 
325 #define MAKE_CHAIN(m, mp, p, i)\
326     do {\
327 	if (m) {\
328 		if (!hdrsplit) \
329 			panic("assumption failed: hdr not split"); \
330 		*mtod((m), u_char *) = *(p);\
331 		*(p) = (i);\
332 		p = mtod((m), u_char *);\
333 		(m)->m_next = (mp)->m_next;\
334 		(mp)->m_next = (m);\
335 		(mp) = (m);\
336 	}\
337     } while (0)
338 		/*
339 		 * result: IPv6 hbh dest1 rthdr dest2 payload
340 		 * m will point to IPv6 header.  mprev will point to the
341 		 * extension header prior to dest2 (rthdr in the above case).
342 		 */
343 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev,
344 			   nexthdrp, IPPROTO_HOPOPTS);
345 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev,
346 			   nexthdrp, IPPROTO_DSTOPTS);
347 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev,
348 			   nexthdrp, IPPROTO_ROUTING);
349 
350 #ifdef IPSEC
351 		if (!needipsec)
352 			goto skip_ipsec2;
353 
354 		/*
355 		 * pointers after IPsec headers are not valid any more.
356 		 * other pointers need a great care too.
357 		 * (IPsec routines should not mangle mbufs prior to AH/ESP)
358 		 */
359 		exthdrs.ip6e_dest2 = NULL;
360 
361 	    {
362 		struct ip6_rthdr *rh = NULL;
363 		int segleft_org = 0;
364 		struct ipsec_output_state state;
365 
366 		if (exthdrs.ip6e_rthdr) {
367 			rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
368 			segleft_org = rh->ip6r_segleft;
369 			rh->ip6r_segleft = 0;
370 		}
371 
372 		bzero(&state, sizeof(state));
373 		state.m = m;
374 		error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
375 			&needipsectun);
376 		m = state.m;
377 		if (error) {
378 			/* mbuf is already reclaimed in ipsec6_output_trans. */
379 			m = NULL;
380 			switch (error) {
381 			case EHOSTUNREACH:
382 			case ENETUNREACH:
383 			case EMSGSIZE:
384 			case ENOBUFS:
385 			case ENOMEM:
386 				break;
387 			default:
388 				printf("ip6_output (ipsec): error code %d\n", error);
389 				/*fall through*/
390 			case ENOENT:
391 				/* don't show these error codes to the user */
392 				error = 0;
393 				break;
394 			}
395 			goto bad;
396 		}
397 		if (exthdrs.ip6e_rthdr) {
398 			/* ah6_output doesn't modify mbuf chain */
399 			rh->ip6r_segleft = segleft_org;
400 		}
401 	    }
402 skip_ipsec2:;
403 #endif
404 	}
405 
406 	/*
407 	 * If there is a routing header, replace destination address field
408 	 * with the first hop of the routing header.
409 	 */
410 	if (exthdrs.ip6e_rthdr) {
411 		struct ip6_rthdr *rh =
412 			(struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
413 						  struct ip6_rthdr *));
414 		struct ip6_rthdr0 *rh0;
415 
416 		finaldst = ip6->ip6_dst;
417 		switch(rh->ip6r_type) {
418 		case IPV6_RTHDR_TYPE_0:
419 			 rh0 = (struct ip6_rthdr0 *)rh;
420 			 ip6->ip6_dst = rh0->ip6r0_addr[0];
421 			 bcopy((caddr_t)&rh0->ip6r0_addr[1],
422 				 (caddr_t)&rh0->ip6r0_addr[0],
423 				 sizeof(struct in6_addr)*(rh0->ip6r0_segleft - 1)
424 				 );
425 			 rh0->ip6r0_addr[rh0->ip6r0_segleft - 1] = finaldst;
426 			 break;
427 		default:	/* is it possible? */
428 			 error = EINVAL;
429 			 goto bad;
430 		}
431 	}
432 
433 	/* Source address validation */
434 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
435 	    (flags & IPV6_DADOUTPUT) == 0) {
436 		error = EOPNOTSUPP;
437 		ip6stat.ip6s_badscope++;
438 		goto bad;
439 	}
440 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
441 		error = EOPNOTSUPP;
442 		ip6stat.ip6s_badscope++;
443 		goto bad;
444 	}
445 
446 	ip6stat.ip6s_localout++;
447 
448 	/*
449 	 * Route packet.
450 	 */
451 	if (ro == 0) {
452 		ro = &ip6route;
453 		bzero((caddr_t)ro, sizeof(*ro));
454 	}
455 	ro_pmtu = ro;
456 	if (opt && opt->ip6po_rthdr)
457 		ro = &opt->ip6po_route;
458 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
459 	/*
460 	 * If there is a cached route,
461 	 * check that it is to the same destination
462 	 * and is still up. If not, free it and try again.
463 	 */
464 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
465 			 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
466 		RTFREE(ro->ro_rt);
467 		ro->ro_rt = (struct rtentry *)0;
468 	}
469 	if (ro->ro_rt == 0) {
470 		bzero(dst, sizeof(*dst));
471 		dst->sin6_family = AF_INET6;
472 		dst->sin6_len = sizeof(struct sockaddr_in6);
473 		dst->sin6_addr = ip6->ip6_dst;
474 	}
475 #ifdef IPSEC
476 	if (needipsec && needipsectun) {
477 		struct ipsec_output_state state;
478 
479 		/*
480 		 * All the extension headers will become inaccessible
481 		 * (since they can be encrypted).
482 		 * Don't panic, we need no more updates to extension headers
483 		 * on inner IPv6 packet (since they are now encapsulated).
484 		 *
485 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
486 		 */
487 		bzero(&exthdrs, sizeof(exthdrs));
488 		exthdrs.ip6e_ip6 = m;
489 
490 		bzero(&state, sizeof(state));
491 		state.m = m;
492 		state.ro = (struct route *)ro;
493 		state.dst = (struct sockaddr *)dst;
494 
495 		error = ipsec6_output_tunnel(&state, sp, flags);
496 
497 		m = state.m;
498 		ro = (struct route_in6 *)state.ro;
499 		dst = (struct sockaddr_in6 *)state.dst;
500 		if (error) {
501 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
502 			m0 = m = NULL;
503 			m = NULL;
504 			switch (error) {
505 			case EHOSTUNREACH:
506 			case ENETUNREACH:
507 			case EMSGSIZE:
508 			case ENOBUFS:
509 			case ENOMEM:
510 				break;
511 			default:
512 				printf("ip6_output (ipsec): error code %d\n", error);
513 				/*fall through*/
514 			case ENOENT:
515 				/* don't show these error codes to the user */
516 				error = 0;
517 				break;
518 			}
519 			goto bad;
520 		}
521 
522 		exthdrs.ip6e_ip6 = m;
523 	}
524 #endif /*IPSEC*/
525 
526 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
527 		/* Unicast */
528 
529 #define ifatoia6(ifa)	((struct in6_ifaddr *)(ifa))
530 #define sin6tosa(sin6)	((struct sockaddr *)(sin6))
531 		/* xxx
532 		 * interface selection comes here
533 		 * if an interface is specified from an upper layer,
534 		 * ifp must point it.
535 		 */
536 		if (ro->ro_rt == 0) {
537 			/*
538 			 * non-bsdi always clone routes, if parent is
539 			 * PRF_CLONING.
540 			 */
541 			rtalloc((struct route *)ro);
542 		}
543 		if (ro->ro_rt == 0) {
544 			ip6stat.ip6s_noroute++;
545 			error = EHOSTUNREACH;
546 			/* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
547 			goto bad;
548 		}
549 		ia = ifatoia6(ro->ro_rt->rt_ifa);
550 		ifp = ro->ro_rt->rt_ifp;
551 		ro->ro_rt->rt_use++;
552 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
553 			dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
554 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
555 
556 		in6_ifstat_inc(ifp, ifs6_out_request);
557 
558 		/*
559 		 * Check if the outgoing interface conflicts with
560 		 * the interface specified by ifi6_ifindex (if specified).
561 		 * Note that loopback interface is always okay.
562 		 * (this may happen when we are sending a packet to one of
563 		 *  our own addresses.)
564 		 */
565 		if (opt && opt->ip6po_pktinfo
566 		 && opt->ip6po_pktinfo->ipi6_ifindex) {
567 			if (!(ifp->if_flags & IFF_LOOPBACK)
568 			 && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
569 				ip6stat.ip6s_noroute++;
570 				in6_ifstat_inc(ifp, ifs6_out_discard);
571 				error = EHOSTUNREACH;
572 				goto bad;
573 			}
574 		}
575 
576 		if (opt && opt->ip6po_hlim != -1)
577 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
578 	} else {
579 		/* Multicast */
580 		struct	in6_multi *in6m;
581 
582 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
583 
584 		/*
585 		 * See if the caller provided any multicast options
586 		 */
587 		ifp = NULL;
588 		if (im6o != NULL) {
589 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
590 			if (im6o->im6o_multicast_ifp != NULL)
591 				ifp = im6o->im6o_multicast_ifp;
592 		} else
593 			ip6->ip6_hlim = ip6_defmcasthlim;
594 
595 		/*
596 		 * See if the caller provided the outgoing interface
597 		 * as an ancillary data.
598 		 * Boundary check for ifindex is assumed to be already done.
599 		 */
600 		if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
601 			ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex];
602 
603 		/*
604 		 * If the destination is a node-local scope multicast,
605 		 * the packet should be loop-backed only.
606 		 */
607 		if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
608 			/*
609 			 * If the outgoing interface is already specified,
610 			 * it should be a loopback interface.
611 			 */
612 			if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) {
613 				ip6stat.ip6s_badscope++;
614 				error = ENETUNREACH; /* XXX: better error? */
615 				/* XXX correct ifp? */
616 				in6_ifstat_inc(ifp, ifs6_out_discard);
617 				goto bad;
618 			} else {
619 				ifp = &loif[0];
620 			}
621 		}
622 
623 		if (opt && opt->ip6po_hlim != -1)
624 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
625 
626 		/*
627 		 * If caller did not provide an interface lookup a
628 		 * default in the routing table.  This is either a
629 		 * default for the speicfied group (i.e. a host
630 		 * route), or a multicast default (a route for the
631 		 * ``net'' ff00::/8).
632 		 */
633 		if (ifp == NULL) {
634 			if (ro->ro_rt == 0) {
635 				ro->ro_rt = rtalloc1((struct sockaddr *)
636 						&ro->ro_dst, 0
637 						);
638 			}
639 			if (ro->ro_rt == 0) {
640 				ip6stat.ip6s_noroute++;
641 				error = EHOSTUNREACH;
642 				/* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
643 				goto bad;
644 			}
645 			ia = ifatoia6(ro->ro_rt->rt_ifa);
646 			ifp = ro->ro_rt->rt_ifp;
647 			ro->ro_rt->rt_use++;
648 		}
649 
650 		if ((flags & IPV6_FORWARDING) == 0)
651 			in6_ifstat_inc(ifp, ifs6_out_request);
652 		in6_ifstat_inc(ifp, ifs6_out_mcast);
653 
654 		/*
655 		 * Confirm that the outgoing interface supports multicast.
656 		 */
657 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
658 			ip6stat.ip6s_noroute++;
659 			in6_ifstat_inc(ifp, ifs6_out_discard);
660 			error = ENETUNREACH;
661 			goto bad;
662 		}
663 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
664 		if (in6m != NULL &&
665 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
666 			/*
667 			 * If we belong to the destination multicast group
668 			 * on the outgoing interface, and the caller did not
669 			 * forbid loopback, loop back a copy.
670 			 */
671 			ip6_mloopback(ifp, m, dst);
672 		} else {
673 			/*
674 			 * If we are acting as a multicast router, perform
675 			 * multicast forwarding as if the packet had just
676 			 * arrived on the interface to which we are about
677 			 * to send.  The multicast forwarding function
678 			 * recursively calls this function, using the
679 			 * IPV6_FORWARDING flag to prevent infinite recursion.
680 			 *
681 			 * Multicasts that are looped back by ip6_mloopback(),
682 			 * above, will be forwarded by the ip6_input() routine,
683 			 * if necessary.
684 			 */
685 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
686 				if (ip6_mforward(ip6, ifp, m) != 0) {
687 					m_freem(m);
688 					goto done;
689 				}
690 			}
691 		}
692 		/*
693 		 * Multicasts with a hoplimit of zero may be looped back,
694 		 * above, but must not be transmitted on a network.
695 		 * Also, multicasts addressed to the loopback interface
696 		 * are not sent -- the above call to ip6_mloopback() will
697 		 * loop back a copy if this host actually belongs to the
698 		 * destination group on the loopback interface.
699 		 */
700 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
701 			m_freem(m);
702 			goto done;
703 		}
704 	}
705 
706 	/*
707 	 * Fill the outgoing inteface to tell the upper layer
708 	 * to increment per-interface statistics.
709 	 */
710 	if (ifpp)
711 		*ifpp = ifp;
712 
713 	/*
714 	 * Determine path MTU.
715 	 */
716 	if (ro_pmtu != ro) {
717 		/* The first hop and the final destination may differ. */
718 		struct sockaddr_in6 *sin6_fin =
719 			(struct sockaddr_in6 *)&ro_pmtu->ro_dst;
720 		if (ro_pmtu->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
721 				       !IN6_ARE_ADDR_EQUAL(&sin6_fin->sin6_addr,
722 							   &finaldst))) {
723 			RTFREE(ro_pmtu->ro_rt);
724 			ro_pmtu->ro_rt = (struct rtentry *)0;
725 		}
726 		if (ro_pmtu->ro_rt == 0) {
727 			bzero(sin6_fin, sizeof(*sin6_fin));
728 			sin6_fin->sin6_family = AF_INET6;
729 			sin6_fin->sin6_len = sizeof(struct sockaddr_in6);
730 			sin6_fin->sin6_addr = finaldst;
731 
732 			rtalloc((struct route *)ro_pmtu);
733 		}
734 	}
735 	if (ro_pmtu->ro_rt != NULL) {
736 		u_int32_t ifmtu = nd_ifinfo[ifp->if_index].linkmtu;
737 
738 		mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
739 		if (mtu > ifmtu) {
740 			/*
741 			 * The MTU on the route is larger than the MTU on
742 			 * the interface!  This shouldn't happen, unless the
743 			 * MTU of the interface has been changed after the
744 			 * interface was brought up.  Change the MTU in the
745 			 * route to match the interface MTU (as long as the
746 			 * field isn't locked).
747 			 */
748 			 mtu = ifmtu;
749 			 if ((ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
750 				 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; /* XXX */
751 		}
752 	} else {
753 		mtu = nd_ifinfo[ifp->if_index].linkmtu;
754 	}
755 
756 	/* Fake scoped addresses */
757 	if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
758 		/*
759 		 * If source or destination address is a scoped address, and
760 		 * the packet is going to be sent to a loopback interface,
761 		 * we should keep the original interface.
762 		 */
763 
764 		/*
765 		 * XXX: this is a very experimental and temporary solution.
766 		 * We eventually have sockaddr_in6 and use the sin6_scope_id
767 		 * field of the structure here.
768 		 * We rely on the consistency between two scope zone ids
769 		 * of source add destination, which should already be assured
770 		 * Larger scopes than link will be supported in the near
771 		 * future.
772 		 */
773 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
774 			origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
775 		else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
776 			origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
777 		else
778 			origifp = ifp;
779 	}
780 	else
781 		origifp = ifp;
782 #ifndef FAKE_LOOPBACK_IF
783 	if ((ifp->if_flags & IFF_LOOPBACK) == 0)
784 #else
785 	if (1)
786 #endif
787 	{
788 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
789 			ip6->ip6_src.s6_addr16[1] = 0;
790 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
791 			ip6->ip6_dst.s6_addr16[1] = 0;
792 	}
793 
794 	/*
795 	 * If the outgoing packet contains a hop-by-hop options header,
796 	 * it must be examined and processed even by the source node.
797 	 * (RFC 2460, section 4.)
798 	 */
799 	if (exthdrs.ip6e_hbh) {
800 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh,
801 					   struct ip6_hbh *);
802 		u_int32_t dummy1; /* XXX unused */
803 		u_int32_t dummy2; /* XXX unused */
804 
805 		/*
806 		 *  XXX: if we have to send an ICMPv6 error to the sender,
807 		 *       we need the M_LOOP flag since icmp6_error() expects
808 		 *       the IPv6 and the hop-by-hop options header are
809 		 *       continuous unless the flag is set.
810 		 */
811 		m->m_flags |= M_LOOP;
812 		m->m_pkthdr.rcvif = ifp;
813 		if (ip6_process_hopopts(m,
814 					(u_int8_t *)(hbh + 1),
815 					((hbh->ip6h_len + 1) << 3) -
816 					sizeof(struct ip6_hbh),
817 					&dummy1, &dummy2) < 0) {
818 			/* m was already freed at this point */
819 			error = EINVAL;/* better error? */
820 			goto done;
821 		}
822 		m->m_flags &= ~M_LOOP; /* XXX */
823 		m->m_pkthdr.rcvif = NULL;
824 	}
825 
826 #ifdef PFIL_HOOKS
827 	/*
828 	 * Run through list of hooks for output packets.
829 	 */
830 	m1 = m;
831 	pfh = pfil_hook_get(PFIL_OUT, &inetsw[ip_protox[IPPROTO_IPV6]].pr_pfh);
832 	for (; pfh; pfh = pfh->pfil_link.tqe_next)
833 		if (pfh->pfil_func) {
834 		    	rv = pfh->pfil_func(ip6, sizeof(*ip6), ifp, 1, &m1);
835 			if (rv) {
836 				error = EHOSTUNREACH;
837 				goto done;
838 			}
839 			m = m1;
840 			if (m == NULL)
841 				goto done;
842 			ip6 = mtod(m, struct ip6_hdr *);
843 		}
844 #endif /* PFIL_HOOKS */
845 	/*
846 	 * Send the packet to the outgoing interface.
847 	 * If necessary, do IPv6 fragmentation before sending.
848 	 */
849 	tlen = m->m_pkthdr.len;
850 	if (tlen <= mtu
851 #ifdef notyet
852 	    /*
853 	     * On any link that cannot convey a 1280-octet packet in one piece,
854 	     * link-specific fragmentation and reassembly must be provided at
855 	     * a layer below IPv6. [RFC 2460, sec.5]
856 	     * Thus if the interface has ability of link-level fragmentation,
857 	     * we can just send the packet even if the packet size is
858 	     * larger than the link's MTU.
859 	     * XXX: IFF_FRAGMENTABLE (or such) flag has not been defined yet...
860 	     */
861 
862 	    || ifp->if_flags & IFF_FRAGMENTABLE
863 #endif
864 	    )
865 	{
866 #ifdef IFA_STATS
867 		if (IFA_STATS) {
868 			struct in6_ifaddr *ia6;
869 			ip6 = mtod(m, struct ip6_hdr *);
870 			ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
871 			if (ia6) {
872 				ia->ia_ifa.ifa_data.ifad_outbytes +=
873 					m->m_pkthdr.len;
874 			}
875 		}
876 #endif
877 #ifdef OLDIP6OUTPUT
878 		error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst,
879 					  ro->ro_rt);
880 #else
881 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
882 #endif
883 		goto done;
884 	} else if (mtu < IPV6_MMTU) {
885 		/*
886 		 * note that path MTU is never less than IPV6_MMTU
887 		 * (see icmp6_input).
888 		 */
889 		error = EMSGSIZE;
890 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
891 		goto bad;
892 	} else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
893 		error = EMSGSIZE;
894 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
895 		goto bad;
896 	} else {
897 		struct mbuf **mnext, *m_frgpart;
898 		struct ip6_frag *ip6f;
899 		u_int32_t id = htonl(ip6_id++);
900 		u_char nextproto;
901 
902 		/*
903 		 * Too large for the destination or interface;
904 		 * fragment if possible.
905 		 * Must be able to put at least 8 bytes per fragment.
906 		 */
907 		hlen = unfragpartlen;
908 		if (mtu > IPV6_MAXPACKET)
909 			mtu = IPV6_MAXPACKET;
910 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
911 		if (len < 8) {
912 			error = EMSGSIZE;
913 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
914 			goto bad;
915 		}
916 
917 		mnext = &m->m_nextpkt;
918 
919 		/*
920 		 * Change the next header field of the last header in the
921 		 * unfragmentable part.
922 		 */
923 		if (exthdrs.ip6e_rthdr) {
924 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
925 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
926 		} else if (exthdrs.ip6e_dest1) {
927 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
928 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
929 		} else if (exthdrs.ip6e_hbh) {
930 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
931 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
932 		} else {
933 			nextproto = ip6->ip6_nxt;
934 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
935 		}
936 
937 		/*
938 		 * Loop through length of segment after first fragment,
939 		 * make new header and copy data of each part and link onto chain.
940 		 */
941 		m0 = m;
942 		for (off = hlen; off < tlen; off += len) {
943 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
944 			if (!m) {
945 				error = ENOBUFS;
946 				ip6stat.ip6s_odropped++;
947 				goto sendorfree;
948 			}
949 			m->m_flags = m0->m_flags & M_COPYFLAGS;
950 			*mnext = m;
951 			mnext = &m->m_nextpkt;
952 			m->m_data += max_linkhdr;
953 			mhip6 = mtod(m, struct ip6_hdr *);
954 			*mhip6 = *ip6;
955 			m->m_len = sizeof(*mhip6);
956  			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
957  			if (error) {
958 				ip6stat.ip6s_odropped++;
959 				goto sendorfree;
960 			}
961 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
962 			if (off + len >= tlen)
963 				len = tlen - off;
964 			else
965 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
966 			mhip6->ip6_plen = htons((u_short)(len + hlen +
967 							  sizeof(*ip6f) -
968 							  sizeof(struct ip6_hdr)));
969 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
970 				error = ENOBUFS;
971 				ip6stat.ip6s_odropped++;
972 				goto sendorfree;
973 			}
974 			m_cat(m, m_frgpart);
975 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
976 			m->m_pkthdr.rcvif = (struct ifnet *)0;
977 			ip6f->ip6f_reserved = 0;
978 			ip6f->ip6f_ident = id;
979 			ip6f->ip6f_nxt = nextproto;
980 			ip6stat.ip6s_ofragments++;
981 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
982 		}
983 
984 		in6_ifstat_inc(ifp, ifs6_out_fragok);
985 	}
986 
987 	/*
988 	 * Remove leading garbages.
989 	 */
990 sendorfree:
991 	m = m0->m_nextpkt;
992 	m0->m_nextpkt = 0;
993 	m_freem(m0);
994 	for (m0 = m; m; m = m0) {
995 		m0 = m->m_nextpkt;
996 		m->m_nextpkt = 0;
997 		if (error == 0) {
998 #ifdef IFA_STATS
999 			if (IFA_STATS) {
1000 				struct in6_ifaddr *ia6;
1001 				ip6 = mtod(m, struct ip6_hdr *);
1002 				ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1003 				if (ia6) {
1004 					ia->ia_ifa.ifa_data.ifad_outbytes +=
1005 						m->m_pkthdr.len;
1006 				}
1007 			}
1008 #endif
1009 #ifdef OLDIP6OUTPUT
1010 			error = (*ifp->if_output)(ifp, m,
1011 						  (struct sockaddr *)dst,
1012 						  ro->ro_rt);
1013 #else
1014 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1015 #endif
1016 		} else
1017 			m_freem(m);
1018 	}
1019 
1020 	if (error == 0)
1021 		ip6stat.ip6s_fragmented++;
1022 
1023 done:
1024 	if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1025 		RTFREE(ro->ro_rt);
1026 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1027 		RTFREE(ro_pmtu->ro_rt);
1028 	}
1029 
1030 #ifdef IPSEC
1031 	if (sp != NULL)
1032 		key_freesp(sp);
1033 #endif /* IPSEC */
1034 
1035 	return(error);
1036 
1037 freehdrs:
1038 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1039 	m_freem(exthdrs.ip6e_dest1);
1040 	m_freem(exthdrs.ip6e_rthdr);
1041 	m_freem(exthdrs.ip6e_dest2);
1042 	/* fall through */
1043 bad:
1044 	m_freem(m);
1045 	goto done;
1046 }
1047 
1048 static int
1049 ip6_copyexthdr(mp, hdr, hlen)
1050 	struct mbuf **mp;
1051 	caddr_t hdr;
1052 	int hlen;
1053 {
1054 	struct mbuf *m;
1055 
1056 	if (hlen > MCLBYTES)
1057 		return(ENOBUFS); /* XXX */
1058 
1059 	MGET(m, M_DONTWAIT, MT_DATA);
1060 	if (!m)
1061 		return(ENOBUFS);
1062 
1063 	if (hlen > MLEN) {
1064 		MCLGET(m, M_DONTWAIT);
1065 		if ((m->m_flags & M_EXT) == 0) {
1066 			m_free(m);
1067 			return(ENOBUFS);
1068 		}
1069 	}
1070 	m->m_len = hlen;
1071 	if (hdr)
1072 		bcopy(hdr, mtod(m, caddr_t), hlen);
1073 
1074 	*mp = m;
1075 	return(0);
1076 }
1077 
1078 /*
1079  * Insert jumbo payload option.
1080  */
1081 static int
1082 ip6_insert_jumboopt(exthdrs, plen)
1083 	struct ip6_exthdrs *exthdrs;
1084 	u_int32_t plen;
1085 {
1086 	struct mbuf *mopt;
1087 	u_char *optbuf;
1088 
1089 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1090 
1091 	/*
1092 	 * If there is no hop-by-hop options header, allocate new one.
1093 	 * If there is one but it doesn't have enough space to store the
1094 	 * jumbo payload option, allocate a cluster to store the whole options.
1095 	 * Otherwise, use it to store the options.
1096 	 */
1097 	if (exthdrs->ip6e_hbh == 0) {
1098 		MGET(mopt, M_DONTWAIT, MT_DATA);
1099 		if (mopt == 0)
1100 			return(ENOBUFS);
1101 		mopt->m_len = JUMBOOPTLEN;
1102 		optbuf = mtod(mopt, u_char *);
1103 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1104 		exthdrs->ip6e_hbh = mopt;
1105 	} else {
1106 		struct ip6_hbh *hbh;
1107 
1108 		mopt = exthdrs->ip6e_hbh;
1109 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1110 			caddr_t oldoptp = mtod(mopt, caddr_t);
1111 			int oldoptlen = mopt->m_len;
1112 
1113 			if (mopt->m_flags & M_EXT)
1114 				return(ENOBUFS); /* XXX */
1115 			MCLGET(mopt, M_DONTWAIT);
1116 			if ((mopt->m_flags & M_EXT) == 0)
1117 				return(ENOBUFS);
1118 
1119 			bcopy(oldoptp, mtod(mopt, caddr_t), oldoptlen);
1120 			optbuf = mtod(mopt, caddr_t) + oldoptlen;
1121 			mopt->m_len = oldoptlen + JUMBOOPTLEN;
1122 		} else {
1123 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1124 			mopt->m_len += JUMBOOPTLEN;
1125 		}
1126 		optbuf[0] = IP6OPT_PADN;
1127 		optbuf[1] = 1;
1128 
1129 		/*
1130 		 * Adjust the header length according to the pad and
1131 		 * the jumbo payload option.
1132 		 */
1133 		hbh = mtod(mopt, struct ip6_hbh *);
1134 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1135 	}
1136 
1137 	/* fill in the option. */
1138 	optbuf[2] = IP6OPT_JUMBO;
1139 	optbuf[3] = 4;
1140 	*(u_int32_t *)&optbuf[4] = htonl(plen + JUMBOOPTLEN);
1141 
1142 	/* finally, adjust the packet header length */
1143 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1144 
1145 	return(0);
1146 #undef JUMBOOPTLEN
1147 }
1148 
1149 /*
1150  * Insert fragment header and copy unfragmentable header portions.
1151  */
1152 static int
1153 ip6_insertfraghdr(m0, m, hlen, frghdrp)
1154 	struct mbuf *m0, *m;
1155 	int hlen;
1156 	struct ip6_frag **frghdrp;
1157 {
1158 	struct mbuf *n, *mlast;
1159 
1160 	if (hlen > sizeof(struct ip6_hdr)) {
1161 		n = m_copym(m0, sizeof(struct ip6_hdr),
1162 			    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1163 		if (n == 0)
1164 			return(ENOBUFS);
1165 		m->m_next = n;
1166 	} else
1167 		n = m;
1168 
1169 	/* Search for the last mbuf of unfragmentable part. */
1170 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1171 		;
1172 
1173 	if ((mlast->m_flags & M_EXT) == 0 &&
1174 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1175 		/* use the trailing space of the last mbuf for the fragment hdr */
1176 		*frghdrp =
1177 			(struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len);
1178 		mlast->m_len += sizeof(struct ip6_frag);
1179 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1180 	} else {
1181 		/* allocate a new mbuf for the fragment header */
1182 		struct mbuf *mfrg;
1183 
1184 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1185 		if (mfrg == 0)
1186 			return(ENOBUFS);
1187 		mfrg->m_len = sizeof(struct ip6_frag);
1188 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1189 		mlast->m_next = mfrg;
1190 	}
1191 
1192 	return(0);
1193 }
1194 
1195 /*
1196  * IP6 socket option processing.
1197  */
1198 int
1199 ip6_ctloutput(op, so, level, optname, mp)
1200 	int op;
1201 	struct socket *so;
1202 	int level, optname;
1203 	struct mbuf **mp;
1204 {
1205 	register struct in6pcb *in6p = sotoin6pcb(so);
1206 	register struct mbuf *m = *mp;
1207 	register int optval = 0;
1208 	int error = 0;
1209 	struct proc *p = curproc;	/* XXX */
1210 
1211 	if (level == IPPROTO_IPV6) {
1212 		switch (op) {
1213 
1214 		case PRCO_SETOPT:
1215 			switch (optname) {
1216 			case IPV6_PKTOPTIONS:
1217 				/* m is freed in ip6_pcbopts */
1218 				return(ip6_pcbopts(&in6p->in6p_outputopts,
1219 						   m, so));
1220 			case IPV6_HOPOPTS:
1221 			case IPV6_DSTOPTS:
1222 				if (p == 0 || suser(p->p_ucred, &p->p_acflag)) {
1223 					error = EPERM;
1224 					break;
1225 				}
1226 				/* fall through */
1227 			case IPV6_UNICAST_HOPS:
1228 			case IPV6_RECVOPTS:
1229 			case IPV6_RECVRETOPTS:
1230 			case IPV6_RECVDSTADDR:
1231 			case IPV6_PKTINFO:
1232 			case IPV6_HOPLIMIT:
1233 			case IPV6_RTHDR:
1234 			case IPV6_CHECKSUM:
1235 			case IPV6_FAITH:
1236 #ifndef INET6_BINDV6ONLY
1237 			case IPV6_BINDV6ONLY:
1238 #endif
1239 				if (!m || m->m_len != sizeof(int))
1240 					error = EINVAL;
1241 				else {
1242 					optval = *mtod(m, int *);
1243 					switch (optname) {
1244 
1245 					case IPV6_UNICAST_HOPS:
1246 						if (optval < -1 || optval >= 256)
1247 							error = EINVAL;
1248 						else {
1249 							/* -1 = kernel default */
1250 							in6p->in6p_hops = optval;
1251 						}
1252 						break;
1253 #define OPTSET(bit) \
1254 	if (optval) \
1255 		in6p->in6p_flags |= bit; \
1256 	else \
1257 		in6p->in6p_flags &= ~bit;
1258 
1259 					case IPV6_RECVOPTS:
1260 						OPTSET(IN6P_RECVOPTS);
1261 						break;
1262 
1263 					case IPV6_RECVRETOPTS:
1264 						OPTSET(IN6P_RECVRETOPTS);
1265 						break;
1266 
1267 					case IPV6_RECVDSTADDR:
1268 						OPTSET(IN6P_RECVDSTADDR);
1269 						break;
1270 
1271 					case IPV6_PKTINFO:
1272 						OPTSET(IN6P_PKTINFO);
1273 						break;
1274 
1275 					case IPV6_HOPLIMIT:
1276 						OPTSET(IN6P_HOPLIMIT);
1277 						break;
1278 
1279 					case IPV6_HOPOPTS:
1280 						OPTSET(IN6P_HOPOPTS);
1281 						break;
1282 
1283 					case IPV6_DSTOPTS:
1284 						OPTSET(IN6P_DSTOPTS);
1285 						break;
1286 
1287 					case IPV6_RTHDR:
1288 						OPTSET(IN6P_RTHDR);
1289 						break;
1290 
1291 					case IPV6_CHECKSUM:
1292 						in6p->in6p_cksum = optval;
1293 						break;
1294 
1295 					case IPV6_FAITH:
1296 						OPTSET(IN6P_FAITH);
1297 						break;
1298 
1299 #ifndef INET6_BINDV6ONLY
1300 					case IPV6_BINDV6ONLY:
1301 						OPTSET(IN6P_BINDV6ONLY);
1302 						break;
1303 #endif
1304 					}
1305 				}
1306 				break;
1307 #undef OPTSET
1308 
1309 			case IPV6_MULTICAST_IF:
1310 			case IPV6_MULTICAST_HOPS:
1311 			case IPV6_MULTICAST_LOOP:
1312 			case IPV6_JOIN_GROUP:
1313 			case IPV6_LEAVE_GROUP:
1314 				error =	ip6_setmoptions(optname, &in6p->in6p_moptions, m);
1315 				break;
1316 
1317 			case IPV6_PORTRANGE:
1318 				optval = *mtod(m, int *);
1319 
1320 				switch (optval) {
1321 				case IPV6_PORTRANGE_DEFAULT:
1322 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1323 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1324 					break;
1325 
1326 				case IPV6_PORTRANGE_HIGH:
1327 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1328 					in6p->in6p_flags |= IN6P_HIGHPORT;
1329 					break;
1330 
1331 				case IPV6_PORTRANGE_LOW:
1332 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1333 					in6p->in6p_flags |= IN6P_LOWPORT;
1334 					break;
1335 
1336 				default:
1337 					error = EINVAL;
1338 					break;
1339 				}
1340 				break;
1341 
1342 #ifdef IPSEC
1343 			case IPV6_IPSEC_POLICY:
1344 			    {
1345 				caddr_t req = NULL;
1346 				size_t len = 0;
1347 
1348 				int priv = 0;
1349 				if (p == 0 || suser(p->p_ucred, &p->p_acflag))
1350 					priv = 0;
1351 				else
1352 					priv = 1;
1353 				if (m) {
1354 					req = mtod(m, caddr_t);
1355 					len = m->m_len;
1356 				}
1357 				error = ipsec6_set_policy(in6p,
1358 				                   optname, req, len, priv);
1359 			    }
1360 				break;
1361 #endif /* IPSEC */
1362 
1363 			default:
1364 				error = ENOPROTOOPT;
1365 				break;
1366 			}
1367 			if (m)
1368 				(void)m_free(m);
1369 			break;
1370 
1371 		case PRCO_GETOPT:
1372 			switch (optname) {
1373 
1374 			case IPV6_OPTIONS:
1375 			case IPV6_RETOPTS:
1376 #if 0
1377 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1378 				if (in6p->in6p_options) {
1379 					m->m_len = in6p->in6p_options->m_len;
1380 					bcopy(mtod(in6p->in6p_options, caddr_t),
1381 					      mtod(m, caddr_t),
1382 					      (unsigned)m->m_len);
1383 				} else
1384 					m->m_len = 0;
1385 				break;
1386 #else
1387 				error = ENOPROTOOPT;
1388 				break;
1389 #endif
1390 
1391 			case IPV6_PKTOPTIONS:
1392 				if (in6p->in6p_options) {
1393 					*mp = m_copym(in6p->in6p_options, 0,
1394 						      M_COPYALL, M_WAIT);
1395 				} else {
1396 					*mp = m_get(M_WAIT, MT_SOOPTS);
1397 					(*mp)->m_len = 0;
1398 				}
1399 				break;
1400 
1401 			case IPV6_HOPOPTS:
1402 			case IPV6_DSTOPTS:
1403 				if (p == 0 || suser(p->p_ucred, &p->p_acflag)) {
1404 					error = EPERM;
1405 					break;
1406 				}
1407 				/* fall through */
1408 			case IPV6_UNICAST_HOPS:
1409 			case IPV6_RECVOPTS:
1410 			case IPV6_RECVRETOPTS:
1411 			case IPV6_RECVDSTADDR:
1412 			case IPV6_PORTRANGE:
1413 			case IPV6_PKTINFO:
1414 			case IPV6_HOPLIMIT:
1415 			case IPV6_RTHDR:
1416 			case IPV6_CHECKSUM:
1417 			case IPV6_FAITH:
1418 #ifndef INET6_BINDV6ONLY
1419 			case IPV6_BINDV6ONLY:
1420 #endif
1421 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1422 				m->m_len = sizeof(int);
1423 				switch (optname) {
1424 
1425 				case IPV6_UNICAST_HOPS:
1426 					optval = in6p->in6p_hops;
1427 					break;
1428 
1429 #define OPTBIT(bit) (in6p->in6p_flags & bit ? 1 : 0)
1430 
1431 				case IPV6_RECVOPTS:
1432 					optval = OPTBIT(IN6P_RECVOPTS);
1433 					break;
1434 
1435 				case IPV6_RECVRETOPTS:
1436 					optval = OPTBIT(IN6P_RECVRETOPTS);
1437 					break;
1438 
1439 				case IPV6_RECVDSTADDR:
1440 					optval = OPTBIT(IN6P_RECVDSTADDR);
1441 					break;
1442 
1443 				case IPV6_PORTRANGE:
1444 				    {
1445 					int flags;
1446 					flags = in6p->in6p_flags;
1447 					if (flags & IN6P_HIGHPORT)
1448 						optval = IPV6_PORTRANGE_HIGH;
1449 					else if (flags & IN6P_LOWPORT)
1450 						optval = IPV6_PORTRANGE_LOW;
1451 					else
1452 						optval = 0;
1453 					break;
1454 				    }
1455 
1456 				case IPV6_PKTINFO:
1457 					optval = OPTBIT(IN6P_PKTINFO);
1458 					break;
1459 
1460 				case IPV6_HOPLIMIT:
1461 					optval = OPTBIT(IN6P_HOPLIMIT);
1462 					break;
1463 
1464 				case IPV6_HOPOPTS:
1465 					optval = OPTBIT(IN6P_HOPOPTS);
1466 					break;
1467 
1468 				case IPV6_DSTOPTS:
1469 					optval = OPTBIT(IN6P_DSTOPTS);
1470 					break;
1471 
1472 				case IPV6_RTHDR:
1473 					optval = OPTBIT(IN6P_RTHDR);
1474 					break;
1475 
1476 				case IPV6_CHECKSUM:
1477 					optval = in6p->in6p_cksum;
1478 					break;
1479 
1480 				case IPV6_FAITH:
1481 					optval = OPTBIT(IN6P_FAITH);
1482 					break;
1483 
1484 #ifndef INET6_BINDV6ONLY
1485 				case IPV6_BINDV6ONLY:
1486 					optval = OPTBIT(IN6P_BINDV6ONLY);
1487 					break;
1488 #endif
1489 				}
1490 				*mtod(m, int *) = optval;
1491 				break;
1492 
1493 			case IPV6_MULTICAST_IF:
1494 			case IPV6_MULTICAST_HOPS:
1495 			case IPV6_MULTICAST_LOOP:
1496 			case IPV6_JOIN_GROUP:
1497 			case IPV6_LEAVE_GROUP:
1498 				error = ip6_getmoptions(optname, in6p->in6p_moptions, mp);
1499 				break;
1500 
1501 #ifdef IPSEC
1502 			case IPV6_IPSEC_POLICY:
1503 			{
1504 				caddr_t req = NULL;
1505 				size_t len = 0;
1506 
1507 				if (m) {
1508 					req = mtod(m, caddr_t);
1509 					len = m->m_len;
1510 				}
1511 				error = ipsec6_get_policy(in6p, req, len, mp);
1512 				break;
1513 			}
1514 #endif /* IPSEC */
1515 
1516 			default:
1517 				error = ENOPROTOOPT;
1518 				break;
1519 			}
1520 			break;
1521 		}
1522 	} else {
1523 		error = EINVAL;
1524 		if (op == PRCO_SETOPT && *mp)
1525 			(void)m_free(*mp);
1526 	}
1527 	return(error);
1528 }
1529 
1530 /*
1531  * Set up IP6 options in pcb for insertion in output packets.
1532  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1533  * with destination address if source routed.
1534  */
1535 static int
1536 ip6_pcbopts(pktopt, m, so)
1537 	struct ip6_pktopts **pktopt;
1538 	register struct mbuf *m;
1539 	struct socket *so;
1540 {
1541 	register struct ip6_pktopts *opt = *pktopt;
1542 	int error = 0;
1543 	struct proc *p = curproc;	/* XXX */
1544 	int priv = 0;
1545 
1546 	/* turn off any old options. */
1547 	if (opt) {
1548 		if (opt->ip6po_m)
1549 			(void)m_free(opt->ip6po_m);
1550 	} else
1551 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
1552 	*pktopt = 0;
1553 
1554 	if (!m || m->m_len == 0) {
1555 		/*
1556 		 * Only turning off any previous options.
1557 		 */
1558 		if (opt)
1559 			free(opt, M_IP6OPT);
1560 		if (m)
1561 			(void)m_free(m);
1562 		return(0);
1563 	}
1564 
1565 	/*  set options specified by user. */
1566 	if (p && !suser(p->p_ucred, &p->p_acflag))
1567 		priv = 1;
1568 	if ((error = ip6_setpktoptions(m, opt, priv)) != 0) {
1569 		(void)m_free(m);
1570 		return(error);
1571 	}
1572 	*pktopt = opt;
1573 	return(0);
1574 }
1575 
1576 /*
1577  * Set the IP6 multicast options in response to user setsockopt().
1578  */
1579 static int
1580 ip6_setmoptions(optname, im6op, m)
1581 	int optname;
1582 	struct ip6_moptions **im6op;
1583 	struct mbuf *m;
1584 {
1585 	int error = 0;
1586 	u_int loop, ifindex;
1587 	struct ipv6_mreq *mreq;
1588 	struct ifnet *ifp;
1589 	struct ip6_moptions *im6o = *im6op;
1590 	struct route_in6 ro;
1591 	struct sockaddr_in6 *dst;
1592 	struct in6_multi_mship *imm;
1593 	struct proc *p = curproc;	/* XXX */
1594 
1595 	if (im6o == NULL) {
1596 		/*
1597 		 * No multicast option buffer attached to the pcb;
1598 		 * allocate one and initialize to default values.
1599 		 */
1600 		im6o = (struct ip6_moptions *)
1601 			malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
1602 
1603 		if (im6o == NULL)
1604 			return(ENOBUFS);
1605 		*im6op = im6o;
1606 		im6o->im6o_multicast_ifp = NULL;
1607 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1608 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
1609 		LIST_INIT(&im6o->im6o_memberships);
1610 	}
1611 
1612 	switch (optname) {
1613 
1614 	case IPV6_MULTICAST_IF:
1615 		/*
1616 		 * Select the interface for outgoing multicast packets.
1617 		 */
1618 		if (m == NULL || m->m_len != sizeof(u_int)) {
1619 			error = EINVAL;
1620 			break;
1621 		}
1622 		ifindex = *(mtod(m, u_int *));
1623 		if (ifindex < 0 || if_index < ifindex) {
1624 			error = ENXIO;	/* XXX EINVAL? */
1625 			break;
1626 		}
1627 		ifp = ifindex2ifnet[ifindex];
1628 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1629 			error = EADDRNOTAVAIL;
1630 			break;
1631 		}
1632 		im6o->im6o_multicast_ifp = ifp;
1633 		break;
1634 
1635 	case IPV6_MULTICAST_HOPS:
1636 	    {
1637 		/*
1638 		 * Set the IP6 hoplimit for outgoing multicast packets.
1639 		 */
1640 		int optval;
1641 		if (m == NULL || m->m_len != sizeof(int)) {
1642 			error = EINVAL;
1643 			break;
1644 		}
1645 		optval = *(mtod(m, u_int *));
1646 		if (optval < -1 || optval >= 256)
1647 			error = EINVAL;
1648 		else if (optval == -1)
1649 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1650 		else
1651 			im6o->im6o_multicast_hlim = optval;
1652 		break;
1653 	    }
1654 
1655 	case IPV6_MULTICAST_LOOP:
1656 		/*
1657 		 * Set the loopback flag for outgoing multicast packets.
1658 		 * Must be zero or one.
1659 		 */
1660 		if (m == NULL || m->m_len != sizeof(u_int) ||
1661 		   (loop = *(mtod(m, u_int *))) > 1) {
1662 			error = EINVAL;
1663 			break;
1664 		}
1665 		im6o->im6o_multicast_loop = loop;
1666 		break;
1667 
1668 	case IPV6_JOIN_GROUP:
1669 		/*
1670 		 * Add a multicast group membership.
1671 		 * Group must be a valid IP6 multicast address.
1672 		 */
1673 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
1674 			error = EINVAL;
1675 			break;
1676 		}
1677 		mreq = mtod(m, struct ipv6_mreq *);
1678 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
1679 			/*
1680 			 * We use the unspecified address to specify to accept
1681 			 * all multicast addresses. Only super user is allowed
1682 			 * to do this.
1683 			 */
1684 			if (suser(p->p_ucred, &p->p_acflag)) {
1685 				error = EACCES;
1686 				break;
1687 			}
1688 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
1689 			error = EINVAL;
1690 			break;
1691 		}
1692 
1693 		/*
1694 		 * If the interface is specified, validate it.
1695 		 */
1696 		if (mreq->ipv6mr_interface < 0
1697 		 || if_index < mreq->ipv6mr_interface) {
1698 			error = ENXIO;	/* XXX EINVAL? */
1699 			break;
1700 		}
1701 		/*
1702 		 * If no interface was explicitly specified, choose an
1703 		 * appropriate one according to the given multicast address.
1704 		 */
1705 		if (mreq->ipv6mr_interface == 0) {
1706 			/*
1707 			 * If the multicast address is in node-local scope,
1708 			 * the interface should be a loopback interface.
1709 			 * Otherwise, look up the routing table for the
1710 			 * address, and choose the outgoing interface.
1711 			 *   XXX: is it a good approach?
1712 			 */
1713 			if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
1714 				ifp = &loif[0];
1715 			} else {
1716 				ro.ro_rt = NULL;
1717 				dst = (struct sockaddr_in6 *)&ro.ro_dst;
1718 				bzero(dst, sizeof(*dst));
1719 				dst->sin6_len = sizeof(struct sockaddr_in6);
1720 				dst->sin6_family = AF_INET6;
1721 				dst->sin6_addr = mreq->ipv6mr_multiaddr;
1722 				rtalloc((struct route *)&ro);
1723 				if (ro.ro_rt == NULL) {
1724 					error = EADDRNOTAVAIL;
1725 					break;
1726 				}
1727 				ifp = ro.ro_rt->rt_ifp;
1728 				rtfree(ro.ro_rt);
1729 			}
1730 		} else
1731 			ifp = ifindex2ifnet[mreq->ipv6mr_interface];
1732 
1733 		/*
1734 		 * See if we found an interface, and confirm that it
1735 		 * supports multicast
1736 		 */
1737 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1738 			error = EADDRNOTAVAIL;
1739 			break;
1740 		}
1741 		/*
1742 		 * Put interface index into the multicast address,
1743 		 * if the address has link-local scope.
1744 		 */
1745 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
1746 			mreq->ipv6mr_multiaddr.s6_addr16[1]
1747 				= htons(mreq->ipv6mr_interface);
1748 		}
1749 		/*
1750 		 * See if the membership already exists.
1751 		 */
1752 		for (imm = im6o->im6o_memberships.lh_first;
1753 		     imm != NULL; imm = imm->i6mm_chain.le_next)
1754 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
1755 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
1756 					       &mreq->ipv6mr_multiaddr))
1757 				break;
1758 		if (imm != NULL) {
1759 			error = EADDRINUSE;
1760 			break;
1761 		}
1762 		/*
1763 		 * Everything looks good; add a new record to the multicast
1764 		 * address list for the given interface.
1765 		 */
1766 		imm = malloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
1767 		if (imm == NULL) {
1768 			error = ENOBUFS;
1769 			break;
1770 		}
1771 		if ((imm->i6mm_maddr =
1772 		     in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
1773 			free(imm, M_IPMADDR);
1774 			break;
1775 		}
1776 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
1777 		break;
1778 
1779 	case IPV6_LEAVE_GROUP:
1780 		/*
1781 		 * Drop a multicast group membership.
1782 		 * Group must be a valid IP6 multicast address.
1783 		 */
1784 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
1785 			error = EINVAL;
1786 			break;
1787 		}
1788 		mreq = mtod(m, struct ipv6_mreq *);
1789 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
1790 			if (suser(p->p_ucred, &p->p_acflag)) {
1791 				error = EACCES;
1792 				break;
1793 			}
1794 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
1795 			error = EINVAL;
1796 			break;
1797 		}
1798 		/*
1799 		 * If an interface address was specified, get a pointer
1800 		 * to its ifnet structure.
1801 		 */
1802 		if (mreq->ipv6mr_interface < 0
1803 		 || if_index < mreq->ipv6mr_interface) {
1804 			error = ENXIO;	/* XXX EINVAL? */
1805 			break;
1806 		}
1807 		ifp = ifindex2ifnet[mreq->ipv6mr_interface];
1808 		/*
1809 		 * Put interface index into the multicast address,
1810 		 * if the address has link-local scope.
1811 		 */
1812 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
1813 			mreq->ipv6mr_multiaddr.s6_addr16[1]
1814 				= htons(mreq->ipv6mr_interface);
1815 		}
1816 		/*
1817 		 * Find the membership in the membership list.
1818 		 */
1819 		for (imm = im6o->im6o_memberships.lh_first;
1820 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
1821 			if ((ifp == NULL ||
1822 			     imm->i6mm_maddr->in6m_ifp == ifp) &&
1823 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
1824 					       &mreq->ipv6mr_multiaddr))
1825 				break;
1826 		}
1827 		if (imm == NULL) {
1828 			/* Unable to resolve interface */
1829 			error = EADDRNOTAVAIL;
1830 			break;
1831 		}
1832 		/*
1833 		 * Give up the multicast address record to which the
1834 		 * membership points.
1835 		 */
1836 		LIST_REMOVE(imm, i6mm_chain);
1837 		in6_delmulti(imm->i6mm_maddr);
1838 		free(imm, M_IPMADDR);
1839 		break;
1840 
1841 	default:
1842 		error = EOPNOTSUPP;
1843 		break;
1844 	}
1845 
1846 	/*
1847 	 * If all options have default values, no need to keep the mbuf.
1848 	 */
1849 	if (im6o->im6o_multicast_ifp == NULL &&
1850 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
1851 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
1852 	    im6o->im6o_memberships.lh_first == NULL) {
1853 		free(*im6op, M_IPMOPTS);
1854 		*im6op = NULL;
1855 	}
1856 
1857 	return(error);
1858 }
1859 
1860 /*
1861  * Return the IP6 multicast options in response to user getsockopt().
1862  */
1863 static int
1864 ip6_getmoptions(optname, im6o, mp)
1865 	int optname;
1866 	register struct ip6_moptions *im6o;
1867 	register struct mbuf **mp;
1868 {
1869 	u_int *hlim, *loop, *ifindex;
1870 
1871 	*mp = m_get(M_WAIT, MT_SOOPTS);
1872 
1873 	switch (optname) {
1874 
1875 	case IPV6_MULTICAST_IF:
1876 		ifindex = mtod(*mp, u_int *);
1877 		(*mp)->m_len = sizeof(u_int);
1878 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
1879 			*ifindex = 0;
1880 		else
1881 			*ifindex = im6o->im6o_multicast_ifp->if_index;
1882 		return(0);
1883 
1884 	case IPV6_MULTICAST_HOPS:
1885 		hlim = mtod(*mp, u_int *);
1886 		(*mp)->m_len = sizeof(u_int);
1887 		if (im6o == NULL)
1888 			*hlim = ip6_defmcasthlim;
1889 		else
1890 			*hlim = im6o->im6o_multicast_hlim;
1891 		return(0);
1892 
1893 	case IPV6_MULTICAST_LOOP:
1894 		loop = mtod(*mp, u_int *);
1895 		(*mp)->m_len = sizeof(u_int);
1896 		if (im6o == NULL)
1897 			*loop = ip6_defmcasthlim;
1898 		else
1899 			*loop = im6o->im6o_multicast_loop;
1900 		return(0);
1901 
1902 	default:
1903 		return(EOPNOTSUPP);
1904 	}
1905 }
1906 
1907 /*
1908  * Discard the IP6 multicast options.
1909  */
1910 void
1911 ip6_freemoptions(im6o)
1912 	register struct ip6_moptions *im6o;
1913 {
1914 	struct in6_multi_mship *imm;
1915 
1916 	if (im6o == NULL)
1917 		return;
1918 
1919 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
1920 		LIST_REMOVE(imm, i6mm_chain);
1921 		if (imm->i6mm_maddr)
1922 			in6_delmulti(imm->i6mm_maddr);
1923 		free(imm, M_IPMADDR);
1924 	}
1925 	free(im6o, M_IPMOPTS);
1926 }
1927 
1928 /*
1929  * Set IPv6 outgoing packet options based on advanced API.
1930  */
1931 int
1932 ip6_setpktoptions(control, opt, priv)
1933 	struct mbuf *control;
1934 	struct ip6_pktopts *opt;
1935 	int priv;
1936 {
1937 	register struct cmsghdr *cm = 0;
1938 
1939 	if (control == 0 || opt == 0)
1940 		return(EINVAL);
1941 
1942 	bzero(opt, sizeof(*opt));
1943 	opt->ip6po_hlim = -1; /* -1 means to use default hop limit */
1944 
1945 	/*
1946 	 * XXX: Currently, we assume all the optional information is stored
1947 	 * in a single mbuf.
1948 	 */
1949 	if (control->m_next)
1950 		return(EINVAL);
1951 
1952 	opt->ip6po_m = control;
1953 
1954 	for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
1955 		     control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
1956 		cm = mtod(control, struct cmsghdr *);
1957 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
1958 			return(EINVAL);
1959 		if (cm->cmsg_level != IPPROTO_IPV6)
1960 			continue;
1961 
1962 		switch(cm->cmsg_type) {
1963 		case IPV6_PKTINFO:
1964 			if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo)))
1965 				return(EINVAL);
1966 			opt->ip6po_pktinfo = (struct in6_pktinfo *)CMSG_DATA(cm);
1967 			if (opt->ip6po_pktinfo->ipi6_ifindex &&
1968 			    IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr))
1969 				opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] =
1970 					htons(opt->ip6po_pktinfo->ipi6_ifindex);
1971 
1972 			if (opt->ip6po_pktinfo->ipi6_ifindex > if_index
1973 			 || opt->ip6po_pktinfo->ipi6_ifindex < 0) {
1974 				return(ENXIO);
1975 			}
1976 
1977 			/*
1978 			 * Check if the requested source address is indeed a
1979 			 * unicast address assigned to the node.
1980 			 */
1981 			if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
1982 				struct ifaddr *ia;
1983 				struct sockaddr_in6 sin6;
1984 
1985 				bzero(&sin6, sizeof(sin6));
1986 				sin6.sin6_len = sizeof(sin6);
1987 				sin6.sin6_family = AF_INET6;
1988 				sin6.sin6_addr =
1989 					opt->ip6po_pktinfo->ipi6_addr;
1990 				ia = ifa_ifwithaddr(sin6tosa(&sin6));
1991 				if (ia == NULL ||
1992 				    (opt->ip6po_pktinfo->ipi6_ifindex &&
1993 				     (ia->ifa_ifp->if_index !=
1994 				      opt->ip6po_pktinfo->ipi6_ifindex))) {
1995 					return(EADDRNOTAVAIL);
1996 				}
1997 				/*
1998 				 * Check if the requested source address is
1999 				 * indeed a unicast address assigned to the
2000 				 * node.
2001 				 */
2002 				if (IN6_IS_ADDR_MULTICAST(&opt->ip6po_pktinfo->ipi6_addr))
2003 					return(EADDRNOTAVAIL);
2004 			}
2005 			break;
2006 
2007 		case IPV6_HOPLIMIT:
2008 			if (cm->cmsg_len != CMSG_LEN(sizeof(int)))
2009 				return(EINVAL);
2010 
2011 			opt->ip6po_hlim = *(int *)CMSG_DATA(cm);
2012 			if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255)
2013 				return(EINVAL);
2014 			break;
2015 
2016 		case IPV6_NEXTHOP:
2017 			if (!priv)
2018 				return(EPERM);
2019 
2020 			if (cm->cmsg_len < sizeof(u_char) ||
2021 			    /* check if cmsg_len is large enough for sa_len */
2022 			    cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm)))
2023 				return(EINVAL);
2024 
2025 			opt->ip6po_nexthop = (struct sockaddr *)CMSG_DATA(cm);
2026 
2027 			break;
2028 
2029 		case IPV6_HOPOPTS:
2030 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh)))
2031 				return(EINVAL);
2032 			opt->ip6po_hbh = (struct ip6_hbh *)CMSG_DATA(cm);
2033 			if (cm->cmsg_len !=
2034 			    CMSG_LEN((opt->ip6po_hbh->ip6h_len + 1) << 3))
2035 				return(EINVAL);
2036 			break;
2037 
2038 		case IPV6_DSTOPTS:
2039 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest)))
2040 				return(EINVAL);
2041 
2042 			/*
2043 			 * If there is no routing header yet, the destination
2044 			 * options header should be put on the 1st part.
2045 			 * Otherwise, the header should be on the 2nd part.
2046 			 * (See RFC 2460, section 4.1)
2047 			 */
2048 			if (opt->ip6po_rthdr == NULL) {
2049 				opt->ip6po_dest1 =
2050 					(struct ip6_dest *)CMSG_DATA(cm);
2051 				if (cm->cmsg_len !=
2052 				    CMSG_LEN((opt->ip6po_dest1->ip6d_len + 1)
2053 					     << 3))
2054 					return(EINVAL);
2055 			}
2056 			else {
2057 				opt->ip6po_dest2 =
2058 					(struct ip6_dest *)CMSG_DATA(cm);
2059 				if (cm->cmsg_len !=
2060 				    CMSG_LEN((opt->ip6po_dest2->ip6d_len + 1)
2061 					     << 3))
2062 					return(EINVAL);
2063 			}
2064 			break;
2065 
2066 		case IPV6_RTHDR:
2067 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr)))
2068 				return(EINVAL);
2069 			opt->ip6po_rthdr = (struct ip6_rthdr *)CMSG_DATA(cm);
2070 			if (cm->cmsg_len !=
2071 			    CMSG_LEN((opt->ip6po_rthdr->ip6r_len + 1) << 3))
2072 				return(EINVAL);
2073 			switch(opt->ip6po_rthdr->ip6r_type) {
2074 			case IPV6_RTHDR_TYPE_0:
2075 				if (opt->ip6po_rthdr->ip6r_segleft == 0)
2076 					return(EINVAL);
2077 				break;
2078 			default:
2079 				return(EINVAL);
2080 			}
2081 			break;
2082 
2083 		default:
2084 			return(ENOPROTOOPT);
2085 		}
2086 	}
2087 
2088 	return(0);
2089 }
2090 
2091 /*
2092  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2093  * packet to the input queue of a specified interface.  Note that this
2094  * calls the output routine of the loopback "driver", but with an interface
2095  * pointer that might NOT be &loif -- easier than replicating that code here.
2096  */
2097 void
2098 ip6_mloopback(ifp, m, dst)
2099 	struct ifnet *ifp;
2100 	register struct mbuf *m;
2101 	register struct sockaddr_in6 *dst;
2102 {
2103 	struct mbuf *copym;
2104 	struct ip6_hdr *ip6;
2105 
2106 	copym = m_copy(m, 0, M_COPYALL);
2107 	if (copym == NULL)
2108 		return;
2109 
2110 	/*
2111 	 * Make sure to deep-copy IPv6 header portion in case the data
2112 	 * is in an mbuf cluster, so that we can safely override the IPv6
2113 	 * header portion later.
2114 	 */
2115 	if ((copym->m_flags & M_EXT) != 0 ||
2116 	    copym->m_len < sizeof(struct ip6_hdr)) {
2117 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
2118 		if (copym == NULL)
2119 			return;
2120 	}
2121 
2122 #ifdef DIAGNOSTIC
2123 	if (copym->m_len < sizeof(*ip6)) {
2124 		m_freem(copym);
2125 		return;
2126 	}
2127 #endif
2128 
2129 #ifndef FAKE_LOOPBACK_IF
2130 	if ((ifp->if_flags & IFF_LOOPBACK) == 0)
2131 #else
2132 	if (1)
2133 #endif
2134 	{
2135 		ip6 = mtod(copym, struct ip6_hdr *);
2136 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
2137 			ip6->ip6_src.s6_addr16[1] = 0;
2138 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
2139 			ip6->ip6_dst.s6_addr16[1] = 0;
2140 	}
2141 
2142 	(void)looutput(ifp, copym, (struct sockaddr *)dst, NULL);
2143 }
2144 
2145 /*
2146  * Chop IPv6 header off from the payload.
2147  */
2148 static int
2149 ip6_splithdr(m, exthdrs)
2150 	struct mbuf *m;
2151 	struct ip6_exthdrs *exthdrs;
2152 {
2153 	struct mbuf *mh;
2154 	struct ip6_hdr *ip6;
2155 
2156 	ip6 = mtod(m, struct ip6_hdr *);
2157 	if (m->m_len > sizeof(*ip6)) {
2158 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
2159 		if (mh == 0) {
2160 			m_freem(m);
2161 			return ENOBUFS;
2162 		}
2163 		M_COPY_PKTHDR(mh, m);
2164 		MH_ALIGN(mh, sizeof(*ip6));
2165 		m->m_flags &= ~M_PKTHDR;
2166 		m->m_len -= sizeof(*ip6);
2167 		m->m_data += sizeof(*ip6);
2168 		mh->m_next = m;
2169 		m = mh;
2170 		m->m_len = sizeof(*ip6);
2171 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2172 	}
2173 	exthdrs->ip6e_ip6 = m;
2174 	return 0;
2175 }
2176 
2177 /*
2178  * Compute IPv6 extension header length.
2179  */
2180 int
2181 ip6_optlen(in6p)
2182 	struct in6pcb *in6p;
2183 {
2184 	int len;
2185 
2186 	if (!in6p->in6p_outputopts)
2187 		return 0;
2188 
2189 	len = 0;
2190 #define elen(x) \
2191     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2192 
2193 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
2194 	len += elen(in6p->in6p_outputopts->ip6po_dest1);
2195 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
2196 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
2197 	return len;
2198 #undef elen
2199 }
2200