xref: /openbsd/sys/netinet6/ip6_output.c (revision 1821443c)
1 /*	$OpenBSD: ip6_output.c,v 1.76 2003/08/15 20:32:20 tedu 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 "pf.h"
65 
66 #include <sys/param.h>
67 #include <sys/malloc.h>
68 #include <sys/mbuf.h>
69 #include <sys/errno.h>
70 #include <sys/protosw.h>
71 #include <sys/socket.h>
72 #include <sys/socketvar.h>
73 #include <sys/systm.h>
74 #include <sys/proc.h>
75 
76 #include <net/if.h>
77 #include <net/route.h>
78 
79 #include <netinet/in.h>
80 #include <netinet/in_var.h>
81 #include <netinet/in_systm.h>
82 #include <netinet/ip.h>
83 #include <netinet/in_pcb.h>
84 
85 #include <netinet/ip6.h>
86 #include <netinet/icmp6.h>
87 #include <netinet6/ip6_var.h>
88 #include <netinet6/nd6.h>
89 
90 #if NPF > 0
91 #include <net/pfvar.h>
92 #endif
93 
94 #ifdef IPSEC
95 #include <netinet/ip_ah.h>
96 #include <netinet/ip_esp.h>
97 #include <netinet/udp.h>
98 #include <netinet/tcp.h>
99 #include <net/pfkeyv2.h>
100 
101 extern u_int8_t get_sa_require(struct inpcb *);
102 
103 extern int ipsec_auth_default_level;
104 extern int ipsec_esp_trans_default_level;
105 extern int ipsec_esp_network_default_level;
106 extern int ipsec_ipcomp_default_level;
107 #endif /* IPSEC */
108 
109 struct ip6_exthdrs {
110 	struct mbuf *ip6e_ip6;
111 	struct mbuf *ip6e_hbh;
112 	struct mbuf *ip6e_dest1;
113 	struct mbuf *ip6e_rthdr;
114 	struct mbuf *ip6e_dest2;
115 };
116 
117 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *, struct socket *);
118 static int ip6_setmoptions(int, struct ip6_moptions **, struct mbuf *);
119 static int ip6_getmoptions(int, struct ip6_moptions *, struct mbuf **);
120 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
121 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
122 	struct ip6_frag **);
123 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
124 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
125 static int ip6_getpmtu(struct route_in6 *, struct route_in6 *,
126 	struct ifnet *, struct in6_addr *, u_long *);
127 
128 /*
129  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
130  * header (with pri, len, nxt, hlim, src, dst).
131  * This function may modify ver and hlim only.
132  * The mbuf chain containing the packet will be freed.
133  * The mbuf opt, if present, will not be freed.
134  *
135  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
136  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
137  * which is rt_rmx.rmx_mtu.
138  */
139 int
140 ip6_output(m0, opt, ro, flags, im6o, ifpp)
141 	struct mbuf *m0;
142 	struct ip6_pktopts *opt;
143 	struct route_in6 *ro;
144 	int flags;
145 	struct ip6_moptions *im6o;
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 	struct in6_ifaddr *ia;
156 	u_long mtu;
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 	u_int8_t sproto = 0;
163 #ifdef IPSEC
164 	struct m_tag *mtag;
165 	union sockaddr_union sdst;
166 	struct tdb_ident *tdbi;
167 	u_int32_t sspi;
168 	struct inpcb *inp;
169 	struct tdb *tdb;
170 	int s;
171 #endif /* IPSEC */
172 
173 #ifdef IPSEC
174 	inp = NULL;	/*XXX*/
175 	if (inp && (inp->inp_flags & INP_IPV6) == 0)
176 		panic("ip6_output: IPv4 pcb is passed");
177 #endif /* IPSEC */
178 
179 #define MAKE_EXTHDR(hp, mp)						\
180     do {								\
181 	if (hp) {							\
182 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
183 		error = ip6_copyexthdr((mp), (caddr_t)(hp), 		\
184 		    ((eh)->ip6e_len + 1) << 3);				\
185 		if (error)						\
186 			goto freehdrs;					\
187 	}								\
188     } while (0)
189 
190 	bzero(&exthdrs, sizeof(exthdrs));
191 	if (opt) {
192 		/* Hop-by-Hop options header */
193 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
194 		/* Destination options header(1st part) */
195 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
196 		/* Routing header */
197 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
198 		/* Destination options header(2nd part) */
199 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
200 	}
201 
202 #ifdef IPSEC
203 	/*
204 	 * splnet is chosen over spltdb because we are not allowed to
205 	 * lower the level, and udp6_output calls us in splnet(). XXX check
206 	 */
207 	s = splnet();
208 
209 	/*
210 	 * Check if there was an outgoing SA bound to the flow
211 	 * from a transport protocol.
212 	 */
213 	ip6 = mtod(m, struct ip6_hdr *);
214 
215 	/* Do we have any pending SAs to apply ? */
216 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
217 	if (mtag != NULL) {
218 #ifdef DIAGNOSTIC
219 		if (mtag->m_tag_len != sizeof (struct tdb_ident))
220 			panic("ip6_output: tag of length %d (should be %d",
221 			    mtag->m_tag_len, sizeof (struct tdb_ident));
222 #endif
223 		tdbi = (struct tdb_ident *)(mtag + 1);
224 		tdb = gettdb(tdbi->spi, &tdbi->dst, tdbi->proto);
225 		if (tdb == NULL)
226 			error = -EINVAL;
227 		m_tag_delete(m, mtag);
228 	} else
229 		tdb = ipsp_spd_lookup(m, AF_INET6, sizeof(struct ip6_hdr),
230 		    &error, IPSP_DIRECTION_OUT, NULL, inp);
231 
232 	if (tdb == NULL) {
233 	        splx(s);
234 
235 		if (error == 0) {
236 		        /*
237 			 * No IPsec processing required, we'll just send the
238 			 * packet out.
239 			 */
240 		        sproto = 0;
241 
242 			/* Fall through to routing/multicast handling */
243 		} else {
244 		        /*
245 			 * -EINVAL is used to indicate that the packet should
246 			 * be silently dropped, typically because we've asked
247 			 * key management for an SA.
248 			 */
249 		        if (error == -EINVAL) /* Should silently drop packet */
250 				error = 0;
251 
252 			goto freehdrs;
253 		}
254 	} else {
255 		/* Loop detection */
256 		for (mtag = m_tag_first(m); mtag != NULL;
257 		    mtag = m_tag_next(m, mtag)) {
258 			if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
259 			    mtag->m_tag_id !=
260 			    PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
261 				continue;
262 			tdbi = (struct tdb_ident *)(mtag + 1);
263 			if (tdbi->spi == tdb->tdb_spi &&
264 			    tdbi->proto == tdb->tdb_sproto &&
265 			    !bcmp(&tdbi->dst, &tdb->tdb_dst,
266 			    sizeof(union sockaddr_union))) {
267 				splx(s);
268 				sproto = 0; /* mark as no-IPsec-needed */
269 				goto done_spd;
270 			}
271 		}
272 
273 	        /* We need to do IPsec */
274 	        bcopy(&tdb->tdb_dst, &sdst, sizeof(sdst));
275 		sspi = tdb->tdb_spi;
276 		sproto = tdb->tdb_sproto;
277 	        splx(s);
278 
279 #if 1 /* XXX */
280 		/* if we have any extension header, we cannot perform IPsec */
281 		if (exthdrs.ip6e_hbh || exthdrs.ip6e_dest1 ||
282 		    exthdrs.ip6e_rthdr || exthdrs.ip6e_dest2) {
283 			error = EHOSTUNREACH;
284 			goto freehdrs;
285 		}
286 #endif
287 	}
288 
289 	/* Fall through to the routing/multicast handling code */
290  done_spd:
291 #endif /* IPSEC */
292 
293 	/*
294 	 * Calculate the total length of the extension header chain.
295 	 * Keep the length of the unfragmentable part for fragmentation.
296 	 */
297 	optlen = 0;
298 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
299 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
300 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
301 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
302 	/* NOTE: we don't add AH/ESP length here. do that later. */
303 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
304 
305 	/*
306 	 * If we need IPsec, or there is at least one extension header,
307 	 * separate IP6 header from the payload.
308 	 */
309 	if ((sproto || optlen) && !hdrsplit) {
310 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
311 			m = NULL;
312 			goto freehdrs;
313 		}
314 		m = exthdrs.ip6e_ip6;
315 		hdrsplit++;
316 	}
317 
318 	/* adjust pointer */
319 	ip6 = mtod(m, struct ip6_hdr *);
320 
321 	/* adjust mbuf packet header length */
322 	m->m_pkthdr.len += optlen;
323 	plen = m->m_pkthdr.len - sizeof(*ip6);
324 
325 	/* If this is a jumbo payload, insert a jumbo payload option. */
326 	if (plen > IPV6_MAXPACKET) {
327 		if (!hdrsplit) {
328 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
329 				m = NULL;
330 				goto freehdrs;
331 			}
332 			m = exthdrs.ip6e_ip6;
333 			hdrsplit++;
334 		}
335 		/* adjust pointer */
336 		ip6 = mtod(m, struct ip6_hdr *);
337 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
338 			goto freehdrs;
339 		ip6->ip6_plen = 0;
340 	} else
341 		ip6->ip6_plen = htons(plen);
342 
343 	/*
344 	 * Concatenate headers and fill in next header fields.
345 	 * Here we have, on "m"
346 	 *	IPv6 payload
347 	 * and we insert headers accordingly.  Finally, we should be getting:
348 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
349 	 *
350 	 * during the header composing process, "m" points to IPv6 header.
351 	 * "mprev" points to an extension header prior to esp.
352 	 */
353 	{
354 		u_char *nexthdrp = &ip6->ip6_nxt;
355 		struct mbuf *mprev = m;
356 
357 		/*
358 		 * we treat dest2 specially.  this makes IPsec processing
359 		 * much easier.
360 		 *
361 		 * result: IPv6 dest2 payload
362 		 * m and mprev will point to IPv6 header.
363 		 */
364 		if (exthdrs.ip6e_dest2) {
365 			if (!hdrsplit)
366 				panic("assumption failed: hdr not split");
367 			exthdrs.ip6e_dest2->m_next = m->m_next;
368 			m->m_next = exthdrs.ip6e_dest2;
369 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
370 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
371 		}
372 
373 #define MAKE_CHAIN(m, mp, p, i)\
374     do {\
375 	if (m) {\
376 		if (!hdrsplit) \
377 			panic("assumption failed: hdr not split"); \
378 		*mtod((m), u_char *) = *(p);\
379 		*(p) = (i);\
380 		p = mtod((m), u_char *);\
381 		(m)->m_next = (mp)->m_next;\
382 		(mp)->m_next = (m);\
383 		(mp) = (m);\
384 	}\
385     } while (0)
386 		/*
387 		 * result: IPv6 hbh dest1 rthdr dest2 payload
388 		 * m will point to IPv6 header.  mprev will point to the
389 		 * extension header prior to dest2 (rthdr in the above case).
390 		 */
391 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
392 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
393 		    IPPROTO_DSTOPTS);
394 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
395 		    IPPROTO_ROUTING);
396 	}
397 
398 	/*
399 	 * If there is a routing header, replace destination address field
400 	 * with the first hop of the routing header.
401 	 */
402 	if (exthdrs.ip6e_rthdr) {
403 		struct ip6_rthdr *rh;
404 		struct ip6_rthdr0 *rh0;
405 		struct in6_addr *addr;
406 
407 		rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
408 		    struct ip6_rthdr *));
409 		finaldst = ip6->ip6_dst;
410 		switch (rh->ip6r_type) {
411 		case IPV6_RTHDR_TYPE_0:
412 			 rh0 = (struct ip6_rthdr0 *)rh;
413 			 addr = (struct in6_addr *)(rh0 + 1);
414 			 ip6->ip6_dst = addr[0];
415 			 bcopy(&addr[1], &addr[0],
416 			     sizeof(struct in6_addr) * (rh0->ip6r0_segleft - 1));
417 			 addr[rh0->ip6r0_segleft - 1] = finaldst;
418 			 break;
419 		default:	/* is it possible? */
420 			 error = EINVAL;
421 			 goto bad;
422 		}
423 	}
424 
425 	/* Source address validation */
426 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
427 	    (flags & IPV6_UNSPECSRC) == 0) {
428 		error = EOPNOTSUPP;
429 		ip6stat.ip6s_badscope++;
430 		goto bad;
431 	}
432 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
433 		error = EOPNOTSUPP;
434 		ip6stat.ip6s_badscope++;
435 		goto bad;
436 	}
437 
438 	ip6stat.ip6s_localout++;
439 
440 	/*
441 	 * Route packet.
442 	 */
443 	if (ro == 0) {
444 		ro = &ip6route;
445 		bzero((caddr_t)ro, sizeof(*ro));
446 	}
447 	ro_pmtu = ro;
448 	if (opt && opt->ip6po_rthdr)
449 		ro = &opt->ip6po_route;
450 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
451 	/*
452 	 * If there is a cached route,
453 	 * check that it is to the same destination
454 	 * and is still up. If not, free it and try again.
455 	 */
456 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
457 	    dst->sin6_family != AF_INET6 ||
458 	    !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
459 		RTFREE(ro->ro_rt);
460 		ro->ro_rt = (struct rtentry *)0;
461 	}
462 	if (ro->ro_rt == 0) {
463 		bzero(dst, sizeof(*dst));
464 		dst->sin6_family = AF_INET6;
465 		dst->sin6_len = sizeof(struct sockaddr_in6);
466 		dst->sin6_addr = ip6->ip6_dst;
467 	}
468 #ifdef IPSEC
469 	/*
470 	 * Check if the packet needs encapsulation.
471 	 * ipsp_process_packet will never come back to here.
472 	 */
473 	if (sproto != 0) {
474 	        s = splnet();
475 
476 		/* fill in IPv6 header which would be filled later */
477 		if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
478 			if (opt && opt->ip6po_hlim != -1)
479 				ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
480 		} else {
481 			if (im6o != NULL)
482 				ip6->ip6_hlim = im6o->im6o_multicast_hlim;
483 			else
484 				ip6->ip6_hlim = ip6_defmcasthlim;
485 			if (opt && opt->ip6po_hlim != -1)
486 				ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
487 
488 			/*
489 			 * XXX what should we do if ip6_hlim == 0 and the
490 			 * packet gets tunnelled?
491 			 */
492 		}
493 
494 		tdb = gettdb(sspi, &sdst, sproto);
495 		if (tdb == NULL) {
496 			splx(s);
497 			error = EHOSTUNREACH;
498 			m_freem(m);
499 			goto done;
500 		}
501 
502 		/* Latch to PCB */
503 		if (inp)
504 			tdb_add_inp(tdb, inp, 0);
505 
506 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
507 
508 		/* Callee frees mbuf */
509 		error = ipsp_process_packet(m, tdb, AF_INET6, 0);
510 		splx(s);
511 		return error;  /* Nothing more to be done */
512 	}
513 #endif /* IPSEC */
514 
515 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
516 		/* Unicast */
517 
518 #define ifatoia6(ifa)	((struct in6_ifaddr *)(ifa))
519 #define sin6tosa(sin6)	((struct sockaddr *)(sin6))
520 		/* xxx
521 		 * interface selection comes here
522 		 * if an interface is specified from an upper layer,
523 		 * ifp must point it.
524 		 */
525 		if (ro->ro_rt == 0) {
526 			/*
527 			 * non-bsdi always clone routes, if parent is
528 			 * PRF_CLONING.
529 			 */
530 			rtalloc((struct route *)ro);
531 		}
532 		if (ro->ro_rt == 0) {
533 			ip6stat.ip6s_noroute++;
534 			error = EHOSTUNREACH;
535 			/* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
536 			goto bad;
537 		}
538 		ia = ifatoia6(ro->ro_rt->rt_ifa);
539 		ifp = ro->ro_rt->rt_ifp;
540 		ro->ro_rt->rt_use++;
541 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
542 			dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
543 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
544 
545 		in6_ifstat_inc(ifp, ifs6_out_request);
546 
547 		/*
548 		 * Check if the outgoing interface conflicts with
549 		 * the interface specified by ifi6_ifindex (if specified).
550 		 * Note that loopback interface is always okay.
551 		 * (this may happen when we are sending a packet to one of
552 		 *  our own addresses.)
553 		 */
554 		if (opt && opt->ip6po_pktinfo
555 		 && opt->ip6po_pktinfo->ipi6_ifindex) {
556 			if (!(ifp->if_flags & IFF_LOOPBACK)
557 			 && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
558 				ip6stat.ip6s_noroute++;
559 				in6_ifstat_inc(ifp, ifs6_out_discard);
560 				error = EHOSTUNREACH;
561 				goto bad;
562 			}
563 		}
564 
565 		if (opt && opt->ip6po_hlim != -1)
566 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
567 	} else {
568 		/* Multicast */
569 		struct	in6_multi *in6m;
570 
571 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
572 
573 		/*
574 		 * See if the caller provided any multicast options
575 		 */
576 		ifp = NULL;
577 		if (im6o != NULL) {
578 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
579 			if (im6o->im6o_multicast_ifp != NULL)
580 				ifp = im6o->im6o_multicast_ifp;
581 		} else
582 			ip6->ip6_hlim = ip6_defmcasthlim;
583 
584 		/*
585 		 * See if the caller provided the outgoing interface
586 		 * as an ancillary data.
587 		 * Boundary check for ifindex is assumed to be already done.
588 		 */
589 		if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
590 			ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex];
591 
592 		/*
593 		 * If the destination is a node-local scope multicast,
594 		 * the packet should be loop-backed only.
595 		 */
596 		if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
597 			/*
598 			 * If the outgoing interface is already specified,
599 			 * it should be a loopback interface.
600 			 */
601 			if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) {
602 				ip6stat.ip6s_badscope++;
603 				error = ENETUNREACH; /* XXX: better error? */
604 				/* XXX correct ifp? */
605 				in6_ifstat_inc(ifp, ifs6_out_discard);
606 				goto bad;
607 			} else {
608 				ifp = lo0ifp;
609 			}
610 		}
611 
612 		if (opt && opt->ip6po_hlim != -1)
613 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
614 
615 		/*
616 		 * If caller did not provide an interface lookup a
617 		 * default in the routing table.  This is either a
618 		 * default for the speicfied group (i.e. a host
619 		 * route), or a multicast default (a route for the
620 		 * ``net'' ff00::/8).
621 		 */
622 		if (ifp == NULL) {
623 			if (ro->ro_rt == 0) {
624 				ro->ro_rt = rtalloc1((struct sockaddr *)
625 				    &ro->ro_dst, 0);
626 			}
627 			if (ro->ro_rt == 0) {
628 				ip6stat.ip6s_noroute++;
629 				error = EHOSTUNREACH;
630 				/* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
631 				goto bad;
632 			}
633 			ia = ifatoia6(ro->ro_rt->rt_ifa);
634 			ifp = ro->ro_rt->rt_ifp;
635 			ro->ro_rt->rt_use++;
636 		}
637 
638 		if ((flags & IPV6_FORWARDING) == 0)
639 			in6_ifstat_inc(ifp, ifs6_out_request);
640 		in6_ifstat_inc(ifp, ifs6_out_mcast);
641 
642 		/*
643 		 * Confirm that the outgoing interface supports multicast.
644 		 */
645 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
646 			ip6stat.ip6s_noroute++;
647 			in6_ifstat_inc(ifp, ifs6_out_discard);
648 			error = ENETUNREACH;
649 			goto bad;
650 		}
651 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
652 		if (in6m != NULL &&
653 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
654 			/*
655 			 * If we belong to the destination multicast group
656 			 * on the outgoing interface, and the caller did not
657 			 * forbid loopback, loop back a copy.
658 			 */
659 			ip6_mloopback(ifp, m, dst);
660 		} else {
661 			/*
662 			 * If we are acting as a multicast router, perform
663 			 * multicast forwarding as if the packet had just
664 			 * arrived on the interface to which we are about
665 			 * to send.  The multicast forwarding function
666 			 * recursively calls this function, using the
667 			 * IPV6_FORWARDING flag to prevent infinite recursion.
668 			 *
669 			 * Multicasts that are looped back by ip6_mloopback(),
670 			 * above, will be forwarded by the ip6_input() routine,
671 			 * if necessary.
672 			 */
673 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
674 				if (ip6_mforward(ip6, ifp, m) != 0) {
675 					m_freem(m);
676 					goto done;
677 				}
678 			}
679 		}
680 		/*
681 		 * Multicasts with a hoplimit of zero may be looped back,
682 		 * above, but must not be transmitted on a network.
683 		 * Also, multicasts addressed to the loopback interface
684 		 * are not sent -- the above call to ip6_mloopback() will
685 		 * loop back a copy if this host actually belongs to the
686 		 * destination group on the loopback interface.
687 		 */
688 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
689 			m_freem(m);
690 			goto done;
691 		}
692 	}
693 
694 	/*
695 	 * Fill the outgoing inteface to tell the upper layer
696 	 * to increment per-interface statistics.
697 	 */
698 	if (ifpp)
699 		*ifpp = ifp;
700 
701 	/* Determine path MTU. */
702 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu)) != 0)
703 		goto bad;
704 
705 	/*
706 	 * The caller of this function may specify to use the minimum MTU
707 	 * in some cases.
708 	 */
709 	if (mtu > IPV6_MMTU) {
710 		if ((flags & IPV6_MINMTU))
711 			mtu = IPV6_MMTU;
712 	}
713 
714 	/* Fake scoped addresses */
715 	if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
716 		/*
717 		 * If source or destination address is a scoped address, and
718 		 * the packet is going to be sent to a loopback interface,
719 		 * we should keep the original interface.
720 		 */
721 
722 		/*
723 		 * XXX: this is a very experimental and temporary solution.
724 		 * We eventually have sockaddr_in6 and use the sin6_scope_id
725 		 * field of the structure here.
726 		 * We rely on the consistency between two scope zone ids
727 		 * of source add destination, which should already be assured
728 		 * Larger scopes than link will be supported in the near
729 		 * future.
730 		 */
731 		origifp = NULL;
732 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
733 			origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
734 		else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
735 			origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
736 		/*
737 		 * XXX: origifp can be NULL even in those two cases above.
738 		 * For example, if we remove the (only) link-local address
739 		 * from the loopback interface, and try to send a link-local
740 		 * address without link-id information.  Then the source
741 		 * address is ::1, and the destination address is the
742 		 * link-local address with its s6_addr16[1] being zero.
743 		 * What is worse, if the packet goes to the loopback interface
744 		 * by a default rejected route, the null pointer would be
745 		 * passed to looutput, and the kernel would hang.
746 		 * The following last resort would prevent such disaster.
747 		 */
748 		if (origifp == NULL)
749 			origifp = ifp;
750 	} else
751 		origifp = ifp;
752 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
753 		ip6->ip6_src.s6_addr16[1] = 0;
754 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
755 		ip6->ip6_dst.s6_addr16[1] = 0;
756 
757 	/*
758 	 * If the outgoing packet contains a hop-by-hop options header,
759 	 * it must be examined and processed even by the source node.
760 	 * (RFC 2460, section 4.)
761 	 */
762 	if (exthdrs.ip6e_hbh) {
763 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
764 		u_int32_t dummy1; /* XXX unused */
765 		u_int32_t dummy2; /* XXX unused */
766 
767 		/*
768 		 *  XXX: if we have to send an ICMPv6 error to the sender,
769 		 *       we need the M_LOOP flag since icmp6_error() expects
770 		 *       the IPv6 and the hop-by-hop options header are
771 		 *       continuous unless the flag is set.
772 		 */
773 		m->m_flags |= M_LOOP;
774 		m->m_pkthdr.rcvif = ifp;
775 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
776 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
777 		    &dummy1, &dummy2) < 0) {
778 			/* m was already freed at this point */
779 			error = EINVAL;/* better error? */
780 			goto done;
781 		}
782 		m->m_flags &= ~M_LOOP; /* XXX */
783 		m->m_pkthdr.rcvif = NULL;
784 	}
785 
786 #if NPF > 0
787 	if (pf_test6(PF_OUT, ifp, &m) != PF_PASS) {
788 		error = EHOSTUNREACH;
789 		m_freem(m);
790 		goto done;
791 	}
792 	if (m == NULL)
793 		goto done;
794 	ip6 = mtod(m, struct ip6_hdr *);
795 #endif
796 
797 	/*
798 	 * Send the packet to the outgoing interface.
799 	 * If necessary, do IPv6 fragmentation before sending.
800 	 */
801 	tlen = m->m_pkthdr.len;
802 	if (tlen <= mtu) {
803 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
804 		goto done;
805 	} else if (mtu < IPV6_MMTU) {
806 		/*
807 		 * note that path MTU is never less than IPV6_MMTU
808 		 * (see icmp6_input).
809 		 */
810 		error = EMSGSIZE;
811 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
812 		goto bad;
813 	} else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
814 		error = EMSGSIZE;
815 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
816 		goto bad;
817 	} else {
818 		struct mbuf **mnext, *m_frgpart;
819 		struct ip6_frag *ip6f;
820 		u_int32_t id = htonl(ip6_id++);
821 		u_char nextproto;
822 
823 		/*
824 		 * Too large for the destination or interface;
825 		 * fragment if possible.
826 		 * Must be able to put at least 8 bytes per fragment.
827 		 */
828 		hlen = unfragpartlen;
829 		if (mtu > IPV6_MAXPACKET)
830 			mtu = IPV6_MAXPACKET;
831 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
832 		if (len < 8) {
833 			error = EMSGSIZE;
834 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
835 			goto bad;
836 		}
837 
838 		mnext = &m->m_nextpkt;
839 
840 		/*
841 		 * Change the next header field of the last header in the
842 		 * unfragmentable part.
843 		 */
844 		if (exthdrs.ip6e_rthdr) {
845 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
846 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
847 		} else if (exthdrs.ip6e_dest1) {
848 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
849 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
850 		} else if (exthdrs.ip6e_hbh) {
851 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
852 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
853 		} else {
854 			nextproto = ip6->ip6_nxt;
855 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
856 		}
857 
858 		/*
859 		 * Loop through length of segment after first fragment,
860 		 * make new header and copy data of each part and link onto
861 		 * chain.
862 		 */
863 		m0 = m;
864 		for (off = hlen; off < tlen; off += len) {
865 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
866 			if (!m) {
867 				error = ENOBUFS;
868 				ip6stat.ip6s_odropped++;
869 				goto sendorfree;
870 			}
871 			m->m_flags = m0->m_flags & M_COPYFLAGS;
872 			*mnext = m;
873 			mnext = &m->m_nextpkt;
874 			m->m_data += max_linkhdr;
875 			mhip6 = mtod(m, struct ip6_hdr *);
876 			*mhip6 = *ip6;
877 			m->m_len = sizeof(*mhip6);
878 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
879 			if (error) {
880 				ip6stat.ip6s_odropped++;
881 				goto sendorfree;
882 			}
883 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
884 			if (off + len >= tlen)
885 				len = tlen - off;
886 			else
887 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
888 			mhip6->ip6_plen = htons((u_short)(len + hlen +
889 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
890 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
891 				error = ENOBUFS;
892 				ip6stat.ip6s_odropped++;
893 				goto sendorfree;
894 			}
895 			m_cat(m, m_frgpart);
896 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
897 			m->m_pkthdr.rcvif = (struct ifnet *)0;
898 			ip6f->ip6f_reserved = 0;
899 			ip6f->ip6f_ident = id;
900 			ip6f->ip6f_nxt = nextproto;
901 			ip6stat.ip6s_ofragments++;
902 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
903 		}
904 
905 		in6_ifstat_inc(ifp, ifs6_out_fragok);
906 	}
907 
908 	/*
909 	 * Remove leading garbages.
910 	 */
911 sendorfree:
912 	m = m0->m_nextpkt;
913 	m0->m_nextpkt = 0;
914 	m_freem(m0);
915 	for (m0 = m; m; m = m0) {
916 		m0 = m->m_nextpkt;
917 		m->m_nextpkt = 0;
918 		if (error == 0) {
919 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
920 		} else
921 			m_freem(m);
922 	}
923 
924 	if (error == 0)
925 		ip6stat.ip6s_fragmented++;
926 
927 done:
928 	if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
929 		RTFREE(ro->ro_rt);
930 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
931 		RTFREE(ro_pmtu->ro_rt);
932 	}
933 
934 	return (error);
935 
936 freehdrs:
937 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
938 	m_freem(exthdrs.ip6e_dest1);
939 	m_freem(exthdrs.ip6e_rthdr);
940 	m_freem(exthdrs.ip6e_dest2);
941 	/* FALLTHROUGH */
942 bad:
943 	m_freem(m);
944 	goto done;
945 }
946 
947 static int
948 ip6_copyexthdr(mp, hdr, hlen)
949 	struct mbuf **mp;
950 	caddr_t hdr;
951 	int hlen;
952 {
953 	struct mbuf *m;
954 
955 	if (hlen > MCLBYTES)
956 		return (ENOBUFS); /* XXX */
957 
958 	MGET(m, M_DONTWAIT, MT_DATA);
959 	if (!m)
960 		return (ENOBUFS);
961 
962 	if (hlen > MLEN) {
963 		MCLGET(m, M_DONTWAIT);
964 		if ((m->m_flags & M_EXT) == 0) {
965 			m_free(m);
966 			return (ENOBUFS);
967 		}
968 	}
969 	m->m_len = hlen;
970 	if (hdr)
971 		bcopy(hdr, mtod(m, caddr_t), hlen);
972 
973 	*mp = m;
974 	return (0);
975 }
976 
977 /*
978  * Insert jumbo payload option.
979  */
980 static int
981 ip6_insert_jumboopt(exthdrs, plen)
982 	struct ip6_exthdrs *exthdrs;
983 	u_int32_t plen;
984 {
985 	struct mbuf *mopt;
986 	u_int8_t *optbuf;
987 	u_int32_t v;
988 
989 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
990 
991 	/*
992 	 * If there is no hop-by-hop options header, allocate new one.
993 	 * If there is one but it doesn't have enough space to store the
994 	 * jumbo payload option, allocate a cluster to store the whole options.
995 	 * Otherwise, use it to store the options.
996 	 */
997 	if (exthdrs->ip6e_hbh == 0) {
998 		MGET(mopt, M_DONTWAIT, MT_DATA);
999 		if (mopt == 0)
1000 			return (ENOBUFS);
1001 		mopt->m_len = JUMBOOPTLEN;
1002 		optbuf = mtod(mopt, u_int8_t *);
1003 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1004 		exthdrs->ip6e_hbh = mopt;
1005 	} else {
1006 		struct ip6_hbh *hbh;
1007 
1008 		mopt = exthdrs->ip6e_hbh;
1009 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1010 			/*
1011 			 * XXX assumption:
1012 			 * - exthdrs->ip6e_hbh is not referenced from places
1013 			 *   other than exthdrs.
1014 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1015 			 */
1016 			int oldoptlen = mopt->m_len;
1017 			struct mbuf *n;
1018 
1019 			/*
1020 			 * XXX: give up if the whole (new) hbh header does
1021 			 * not fit even in an mbuf cluster.
1022 			 */
1023 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1024 				return (ENOBUFS);
1025 
1026 			/*
1027 			 * As a consequence, we must always prepare a cluster
1028 			 * at this point.
1029 			 */
1030 			MGET(n, M_DONTWAIT, MT_DATA);
1031 			if (n) {
1032 				MCLGET(n, M_DONTWAIT);
1033 				if ((n->m_flags & M_EXT) == 0) {
1034 					m_freem(n);
1035 					n = NULL;
1036 				}
1037 			}
1038 			if (!n)
1039 				return (ENOBUFS);
1040 			n->m_len = oldoptlen + JUMBOOPTLEN;
1041 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1042 			      oldoptlen);
1043 			optbuf = mtod(n, u_int8_t *) + oldoptlen;
1044 			m_freem(mopt);
1045 			mopt = exthdrs->ip6e_hbh = n;
1046 		} else {
1047 			optbuf = mtod(mopt, u_int8_t *) + mopt->m_len;
1048 			mopt->m_len += JUMBOOPTLEN;
1049 		}
1050 		optbuf[0] = IP6OPT_PADN;
1051 		optbuf[1] = 0;
1052 
1053 		/*
1054 		 * Adjust the header length according to the pad and
1055 		 * the jumbo payload option.
1056 		 */
1057 		hbh = mtod(mopt, struct ip6_hbh *);
1058 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1059 	}
1060 
1061 	/* fill in the option. */
1062 	optbuf[2] = IP6OPT_JUMBO;
1063 	optbuf[3] = 4;
1064 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1065 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1066 
1067 	/* finally, adjust the packet header length */
1068 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1069 
1070 	return (0);
1071 #undef JUMBOOPTLEN
1072 }
1073 
1074 /*
1075  * Insert fragment header and copy unfragmentable header portions.
1076  */
1077 static int
1078 ip6_insertfraghdr(m0, m, hlen, frghdrp)
1079 	struct mbuf *m0, *m;
1080 	int hlen;
1081 	struct ip6_frag **frghdrp;
1082 {
1083 	struct mbuf *n, *mlast;
1084 
1085 	if (hlen > sizeof(struct ip6_hdr)) {
1086 		n = m_copym(m0, sizeof(struct ip6_hdr),
1087 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1088 		if (n == 0)
1089 			return (ENOBUFS);
1090 		m->m_next = n;
1091 	} else
1092 		n = m;
1093 
1094 	/* Search for the last mbuf of unfragmentable part. */
1095 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1096 		;
1097 
1098 	if ((mlast->m_flags & M_EXT) == 0 &&
1099 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1100 		/* use the trailing space of the last mbuf for the fragment hdr */
1101 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1102 		    mlast->m_len);
1103 		mlast->m_len += sizeof(struct ip6_frag);
1104 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1105 	} else {
1106 		/* allocate a new mbuf for the fragment header */
1107 		struct mbuf *mfrg;
1108 
1109 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1110 		if (mfrg == 0)
1111 			return (ENOBUFS);
1112 		mfrg->m_len = sizeof(struct ip6_frag);
1113 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1114 		mlast->m_next = mfrg;
1115 	}
1116 
1117 	return (0);
1118 }
1119 
1120 static int
1121 ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup)
1122 	struct route_in6 *ro_pmtu, *ro;
1123 	struct ifnet *ifp;
1124 	struct in6_addr *dst;
1125 	u_long *mtup;
1126 {
1127 	u_int32_t mtu = 0;
1128 	int error = 0;
1129 
1130 	if (ro_pmtu != ro) {
1131 		/* The first hop and the final destination may differ. */
1132 		struct sockaddr_in6 *sa6_dst =
1133 		    (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1134 		if (ro_pmtu->ro_rt &&
1135 		    ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1136 		     !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1137 			RTFREE(ro_pmtu->ro_rt);
1138 			ro_pmtu->ro_rt = (struct rtentry *)0;
1139 		}
1140 		if (ro_pmtu->ro_rt == 0) {
1141 			bzero(sa6_dst, sizeof(*sa6_dst));
1142 			sa6_dst->sin6_family = AF_INET6;
1143 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1144 			sa6_dst->sin6_addr = *dst;
1145 
1146 			rtalloc((struct route *)ro_pmtu);
1147 		}
1148 	}
1149 	if (ro_pmtu->ro_rt) {
1150 		u_int32_t ifmtu;
1151 
1152 		if (ifp == NULL)
1153 			ifp = ro_pmtu->ro_rt->rt_ifp;
1154 		ifmtu = IN6_LINKMTU(ifp);
1155 		mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1156 		if (mtu == 0)
1157 			mtu = ifmtu;
1158 		else if (mtu > ifmtu) {
1159 			/*
1160 			 * The MTU on the route is larger than the MTU on
1161 			 * the interface!  This shouldn't happen, unless the
1162 			 * MTU of the interface has been changed after the
1163 			 * interface was brought up.  Change the MTU in the
1164 			 * route to match the interface MTU (as long as the
1165 			 * field isn't locked).
1166 			 *
1167 			 * if MTU on the route is 0, we need to fix the MTU.
1168 			 * this case happens with path MTU discovery timeouts.
1169 			 */
1170 			mtu = ifmtu;
1171 			if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU))
1172 				ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1173 		}
1174 	} else if (ifp) {
1175 		mtu = IN6_LINKMTU(ifp);
1176 	} else
1177 		error = EHOSTUNREACH; /* XXX */
1178 
1179 	*mtup = mtu;
1180 	return (error);
1181 }
1182 
1183 /*
1184  * IP6 socket option processing.
1185  */
1186 int
1187 ip6_ctloutput(op, so, level, optname, mp)
1188 	int op;
1189 	struct socket *so;
1190 	int level, optname;
1191 	struct mbuf **mp;
1192 {
1193 	int privileged;
1194 	struct inpcb *inp = sotoinpcb(so);
1195 	struct mbuf *m = *mp;
1196 	int error, optval;
1197 	int optlen;
1198 #ifdef IPSEC
1199 	struct proc *p = curproc; /* XXX */
1200 	struct tdb *tdb;
1201 	struct tdb_ident *tdbip, tdbi;
1202 	int s;
1203 #endif
1204 
1205 	optlen = m ? m->m_len : 0;
1206 	error = optval = 0;
1207 
1208 	privileged = (inp->inp_socket->so_state & SS_PRIV);
1209 
1210 	if (level == IPPROTO_IPV6) {
1211 		switch (op) {
1212 		case PRCO_SETOPT:
1213 			switch (optname) {
1214 			case IPV6_PKTOPTIONS:
1215 				/* m is freed in ip6_pcbopts */
1216 				return (ip6_pcbopts(&inp->inp_outputopts6,
1217 				    m, so));
1218 			case IPV6_HOPOPTS:
1219 			case IPV6_DSTOPTS:
1220 				if (!privileged) {
1221 					error = EPERM;
1222 					break;
1223 				}
1224 				/* FALLTHROUGH */
1225 			case IPV6_UNICAST_HOPS:
1226 			case IPV6_RECVOPTS:
1227 			case IPV6_RECVRETOPTS:
1228 			case IPV6_RECVDSTADDR:
1229 			case IPV6_PKTINFO:
1230 			case IPV6_HOPLIMIT:
1231 			case IPV6_RTHDR:
1232 			case IPV6_FAITH:
1233 			case IPV6_V6ONLY:
1234 				if (optlen != sizeof(int)) {
1235 					error = EINVAL;
1236 					break;
1237 				}
1238 				optval = *mtod(m, int *);
1239 				switch (optname) {
1240 
1241 				case IPV6_UNICAST_HOPS:
1242 					if (optval < -1 || optval >= 256)
1243 						error = EINVAL;
1244 					else {
1245 						/* -1 = kernel default */
1246 						inp->inp_hops = optval;
1247 					}
1248 					break;
1249 #define OPTSET(bit) \
1250 do { \
1251 	if (optval) \
1252 		inp->inp_flags |= (bit); \
1253 	else \
1254 		inp->inp_flags &= ~(bit); \
1255 } while (0)
1256 				case IPV6_RECVOPTS:
1257 					OPTSET(IN6P_RECVOPTS);
1258 					break;
1259 
1260 				case IPV6_RECVRETOPTS:
1261 					OPTSET(IN6P_RECVRETOPTS);
1262 					break;
1263 
1264 				case IPV6_RECVDSTADDR:
1265 					OPTSET(IN6P_RECVDSTADDR);
1266 					break;
1267 
1268 				case IPV6_PKTINFO:
1269 					OPTSET(IN6P_PKTINFO);
1270 					break;
1271 
1272 				case IPV6_HOPLIMIT:
1273 					OPTSET(IN6P_HOPLIMIT);
1274 					break;
1275 
1276 				case IPV6_HOPOPTS:
1277 					OPTSET(IN6P_HOPOPTS);
1278 					break;
1279 
1280 				case IPV6_DSTOPTS:
1281 					OPTSET(IN6P_DSTOPTS);
1282 					break;
1283 
1284 				case IPV6_RTHDR:
1285 					OPTSET(IN6P_RTHDR);
1286 					break;
1287 
1288 				case IPV6_FAITH:
1289 					OPTSET(IN6P_FAITH);
1290 					break;
1291 
1292 				case IPV6_V6ONLY:
1293 					if (!optval)
1294 						error = EINVAL;
1295 					break;
1296 				}
1297 				break;
1298 #undef OPTSET
1299 
1300 			case IPV6_MULTICAST_IF:
1301 			case IPV6_MULTICAST_HOPS:
1302 			case IPV6_MULTICAST_LOOP:
1303 			case IPV6_JOIN_GROUP:
1304 			case IPV6_LEAVE_GROUP:
1305 				error =	ip6_setmoptions(optname,
1306 					&inp->inp_moptions6, m);
1307 				break;
1308 
1309 			case IPV6_PORTRANGE:
1310 				optval = *mtod(m, int *);
1311 
1312 				switch (optval) {
1313 				case IPV6_PORTRANGE_DEFAULT:
1314 					inp->inp_flags &= ~(IN6P_LOWPORT);
1315 					inp->inp_flags &= ~(IN6P_HIGHPORT);
1316 					break;
1317 
1318 				case IPV6_PORTRANGE_HIGH:
1319 					inp->inp_flags &= ~(IN6P_LOWPORT);
1320 					inp->inp_flags |= IN6P_HIGHPORT;
1321 					break;
1322 
1323 				case IPV6_PORTRANGE_LOW:
1324 					inp->inp_flags &= ~(IN6P_HIGHPORT);
1325 					inp->inp_flags |= IN6P_LOWPORT;
1326 					break;
1327 
1328 				default:
1329 					error = EINVAL;
1330 					break;
1331 				}
1332 				break;
1333 
1334 			case IPSEC6_OUTSA:
1335 #ifndef IPSEC
1336 				error = EINVAL;
1337 #else
1338 				s = spltdb();
1339 				if (m == 0 || m->m_len != sizeof(struct tdb_ident)) {
1340 					error = EINVAL;
1341 				} else {
1342 					tdbip = mtod(m, struct tdb_ident *);
1343 					tdb = gettdb(tdbip->spi, &tdbip->dst,
1344 						     tdbip->proto);
1345 					if (tdb == NULL)
1346 						error = ESRCH;
1347 					else
1348 						tdb_add_inp(tdb, inp, 0);
1349 				}
1350 				splx(s);
1351 #endif /* IPSEC */
1352 				break;
1353 
1354 			case IPV6_AUTH_LEVEL:
1355 			case IPV6_ESP_TRANS_LEVEL:
1356 			case IPV6_ESP_NETWORK_LEVEL:
1357 			case IPV6_IPCOMP_LEVEL:
1358 #ifndef IPSEC
1359 				error = EINVAL;
1360 #else
1361 				if (m == 0 || m->m_len != sizeof(int)) {
1362 					error = EINVAL;
1363 					break;
1364 				}
1365 				optval = *mtod(m, int *);
1366 
1367 				if (optval < IPSEC_LEVEL_BYPASS ||
1368 				    optval > IPSEC_LEVEL_UNIQUE) {
1369 					error = EINVAL;
1370 					break;
1371 				}
1372 
1373 				switch (optname) {
1374 				case IPV6_AUTH_LEVEL:
1375 				        if (optval < ipsec_auth_default_level &&
1376 					    suser(p, 0)) {
1377 						error = EACCES;
1378 						break;
1379 					}
1380 					inp->inp_seclevel[SL_AUTH] = optval;
1381 					break;
1382 
1383 				case IPV6_ESP_TRANS_LEVEL:
1384 				        if (optval < ipsec_esp_trans_default_level &&
1385 					    suser(p, 0)) {
1386 						error = EACCES;
1387 						break;
1388 					}
1389 					inp->inp_seclevel[SL_ESP_TRANS] = optval;
1390 					break;
1391 
1392 				case IPV6_ESP_NETWORK_LEVEL:
1393 				        if (optval < ipsec_esp_network_default_level &&
1394 					    suser(p, 0)) {
1395 						error = EACCES;
1396 						break;
1397 					}
1398 					inp->inp_seclevel[SL_ESP_NETWORK] = optval;
1399 					break;
1400 
1401 				case IPV6_IPCOMP_LEVEL:
1402 				        if (optval < ipsec_ipcomp_default_level &&
1403 					    suser(p, 0)) {
1404 						error = EACCES;
1405 						break;
1406 					}
1407 					inp->inp_seclevel[SL_IPCOMP] = optval;
1408 					break;
1409 				}
1410 				if (!error)
1411 					inp->inp_secrequire = get_sa_require(inp);
1412 #endif
1413 				break;
1414 
1415 
1416 			default:
1417 				error = ENOPROTOOPT;
1418 				break;
1419 			}
1420 			if (m)
1421 				(void)m_free(m);
1422 			break;
1423 
1424 		case PRCO_GETOPT:
1425 			switch (optname) {
1426 
1427 			case IPV6_OPTIONS:
1428 			case IPV6_RETOPTS:
1429 				error = ENOPROTOOPT;
1430 				break;
1431 
1432 			case IPV6_PKTOPTIONS:
1433 				if (inp->inp_options) {
1434 					*mp = m_copym(inp->inp_options, 0,
1435 					    M_COPYALL, M_WAIT);
1436 				} else {
1437 					*mp = m_get(M_WAIT, MT_SOOPTS);
1438 					(*mp)->m_len = 0;
1439 				}
1440 				break;
1441 
1442 			case IPV6_HOPOPTS:
1443 			case IPV6_DSTOPTS:
1444 				if (!privileged) {
1445 					error = EPERM;
1446 					break;
1447 				}
1448 				/* FALLTHROUGH */
1449 			case IPV6_UNICAST_HOPS:
1450 			case IPV6_RECVOPTS:
1451 			case IPV6_RECVRETOPTS:
1452 			case IPV6_RECVDSTADDR:
1453 			case IPV6_PKTINFO:
1454 			case IPV6_HOPLIMIT:
1455 			case IPV6_RTHDR:
1456 			case IPV6_FAITH:
1457 			case IPV6_V6ONLY:
1458 			case IPV6_PORTRANGE:
1459 				switch (optname) {
1460 
1461 				case IPV6_UNICAST_HOPS:
1462 					optval = inp->inp_hops;
1463 					break;
1464 
1465 #define OPTBIT(bit) (inp->inp_flags & bit ? 1 : 0)
1466 
1467 				case IPV6_RECVOPTS:
1468 					optval = OPTBIT(IN6P_RECVOPTS);
1469 					break;
1470 
1471 				case IPV6_RECVRETOPTS:
1472 					optval = OPTBIT(IN6P_RECVRETOPTS);
1473 					break;
1474 
1475 				case IPV6_RECVDSTADDR:
1476 					optval = OPTBIT(IN6P_RECVDSTADDR);
1477 					break;
1478 
1479 				case IPV6_PKTINFO:
1480 					optval = OPTBIT(IN6P_PKTINFO);
1481 					break;
1482 
1483 				case IPV6_HOPLIMIT:
1484 					optval = OPTBIT(IN6P_HOPLIMIT);
1485 					break;
1486 
1487 				case IPV6_HOPOPTS:
1488 					optval = OPTBIT(IN6P_HOPOPTS);
1489 					break;
1490 
1491 				case IPV6_DSTOPTS:
1492 					optval = OPTBIT(IN6P_DSTOPTS);
1493 					break;
1494 
1495 				case IPV6_RTHDR:
1496 					optval = OPTBIT(IN6P_RTHDR);
1497 					break;
1498 
1499 				case IPV6_FAITH:
1500 					optval = OPTBIT(IN6P_FAITH);
1501 					break;
1502 
1503 				case IPV6_V6ONLY:
1504 					optval = (ip6_v6only != 0); /* XXX */
1505 					break;
1506 
1507 				case IPV6_PORTRANGE:
1508 				    {
1509 					int flags;
1510 
1511 					flags = inp->inp_flags;
1512 					if (flags & IN6P_HIGHPORT)
1513 						optval = IPV6_PORTRANGE_HIGH;
1514 					else if (flags & IN6P_LOWPORT)
1515 						optval = IPV6_PORTRANGE_LOW;
1516 					else
1517 						optval = 0;
1518 					break;
1519 				    }
1520 				}
1521 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1522 				m->m_len = sizeof(int);
1523 				*mtod(m, int *) = optval;
1524 				break;
1525 
1526 			case IPV6_MULTICAST_IF:
1527 			case IPV6_MULTICAST_HOPS:
1528 			case IPV6_MULTICAST_LOOP:
1529 			case IPV6_JOIN_GROUP:
1530 			case IPV6_LEAVE_GROUP:
1531 				error = ip6_getmoptions(optname, inp->inp_moptions6, mp);
1532 				break;
1533 
1534 			case IPSEC6_OUTSA:
1535 #ifndef IPSEC
1536 				error = EINVAL;
1537 #else
1538 				s = spltdb();
1539 				if (inp->inp_tdb_out == NULL) {
1540 					error = ENOENT;
1541 				} else {
1542 					tdbi.spi = inp->inp_tdb_out->tdb_spi;
1543 					tdbi.dst = inp->inp_tdb_out->tdb_dst;
1544 					tdbi.proto = inp->inp_tdb_out->tdb_sproto;
1545 					*mp = m = m_get(M_WAIT, MT_SOOPTS);
1546 					m->m_len = sizeof(tdbi);
1547 					bcopy((caddr_t)&tdbi, mtod(m, caddr_t),
1548 					    (unsigned)m->m_len);
1549 				}
1550 				splx(s);
1551 #endif /* IPSEC */
1552 				break;
1553 
1554 			case IPV6_AUTH_LEVEL:
1555 			case IPV6_ESP_TRANS_LEVEL:
1556 			case IPV6_ESP_NETWORK_LEVEL:
1557 			case IPV6_IPCOMP_LEVEL:
1558 #ifndef IPSEC
1559 				m->m_len = sizeof(int);
1560 				*mtod(m, int *) = IPSEC_LEVEL_NONE;
1561 #else
1562 				m->m_len = sizeof(int);
1563 				switch (optname) {
1564 				case IPV6_AUTH_LEVEL:
1565 					optval = inp->inp_seclevel[SL_AUTH];
1566 					break;
1567 
1568 				case IPV6_ESP_TRANS_LEVEL:
1569 					optval =
1570 					    inp->inp_seclevel[SL_ESP_TRANS];
1571 					break;
1572 
1573 				case IPV6_ESP_NETWORK_LEVEL:
1574 					optval =
1575 					    inp->inp_seclevel[SL_ESP_NETWORK];
1576 					break;
1577 
1578 				case IPV6_IPCOMP_LEVEL:
1579 					optval = inp->inp_seclevel[SL_IPCOMP];
1580 					break;
1581 				}
1582 				*mtod(m, int *) = optval;
1583 #endif
1584 				break;
1585 
1586 			default:
1587 				error = ENOPROTOOPT;
1588 				break;
1589 			}
1590 			break;
1591 		}
1592 	} else {
1593 		error = EINVAL;
1594 		if (op == PRCO_SETOPT && *mp)
1595 			(void)m_free(*mp);
1596 	}
1597 	return (error);
1598 }
1599 
1600 int
1601 ip6_raw_ctloutput(op, so, level, optname, mp)
1602 	int op;
1603 	struct socket *so;
1604 	int level, optname;
1605 	struct mbuf **mp;
1606 {
1607 	int error = 0, optval, optlen;
1608 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
1609 	struct inpcb *inp = sotoinpcb(so);
1610 	struct mbuf *m = *mp;
1611 
1612 	optlen = m ? m->m_len : 0;
1613 
1614 	if (level != IPPROTO_IPV6) {
1615 		if (op == PRCO_SETOPT && *mp)
1616 			(void)m_free(*mp);
1617 		return (EINVAL);
1618 	}
1619 
1620 	switch (optname) {
1621 	case IPV6_CHECKSUM:
1622 		/*
1623 		 * For ICMPv6 sockets, no modification allowed for checksum
1624 		 * offset, permit "no change" values to help existing apps.
1625 		 *
1626 		 * XXX 2292bis says: "An attempt to set IPV6_CHECKSUM
1627 		 * for an ICMPv6 socket will fail."
1628 		 * The current behavior does not meet 2292bis.
1629 		 */
1630 		switch (op) {
1631 		case PRCO_SETOPT:
1632 			if (optlen != sizeof(int)) {
1633 				error = EINVAL;
1634 				break;
1635 			}
1636 			optval = *mtod(m, int *);
1637 			if ((optval % 2) != 0) {
1638 				/* the API assumes even offset values */
1639 				error = EINVAL;
1640 			} else if (so->so_proto->pr_protocol ==
1641 			    IPPROTO_ICMPV6) {
1642 				if (optval != icmp6off)
1643 					error = EINVAL;
1644 			} else
1645 				inp->in6p_cksum = optval;
1646 			break;
1647 
1648 		case PRCO_GETOPT:
1649 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
1650 				optval = icmp6off;
1651 			else
1652 				optval = inp->in6p_cksum;
1653 
1654 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1655 			m->m_len = sizeof(int);
1656 			*mtod(m, int *) = optval;
1657 			break;
1658 
1659 		default:
1660 			error = EINVAL;
1661 			break;
1662 		}
1663 		break;
1664 
1665 	default:
1666 		error = ENOPROTOOPT;
1667 		break;
1668 	}
1669 
1670 	if (op == PRCO_SETOPT && m)
1671 		(void)m_free(m);
1672 
1673 	return (error);
1674 }
1675 
1676 /*
1677  * Set up IP6 options in pcb for insertion in output packets.
1678  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1679  * with destination address if source routed.
1680  */
1681 static int
1682 ip6_pcbopts(pktopt, m, so)
1683 	struct ip6_pktopts **pktopt;
1684 	struct mbuf *m;
1685 	struct socket *so;
1686 {
1687 	struct ip6_pktopts *opt = *pktopt;
1688 	int error = 0;
1689 	struct proc *p = curproc;	/* XXX */
1690 	int priv = 0;
1691 
1692 	/* turn off any old options. */
1693 	if (opt) {
1694 		if (opt->ip6po_m)
1695 			(void)m_free(opt->ip6po_m);
1696 	} else
1697 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
1698 	*pktopt = 0;
1699 
1700 	if (!m || m->m_len == 0) {
1701 		/*
1702 		 * Only turning off any previous options.
1703 		 */
1704 		free(opt, M_IP6OPT);
1705 		if (m)
1706 			(void)m_free(m);
1707 		return (0);
1708 	}
1709 
1710 	/*  set options specified by user. */
1711 	if (p && !suser(p, 0))
1712 		priv = 1;
1713 	if ((error = ip6_setpktoptions(m, opt, priv)) != 0) {
1714 		(void)m_free(m);
1715 		free(opt, M_IP6OPT);
1716 		return (error);
1717 	}
1718 	*pktopt = opt;
1719 	return (0);
1720 }
1721 
1722 /*
1723  * Set the IP6 multicast options in response to user setsockopt().
1724  */
1725 static int
1726 ip6_setmoptions(optname, im6op, m)
1727 	int optname;
1728 	struct ip6_moptions **im6op;
1729 	struct mbuf *m;
1730 {
1731 	int error = 0;
1732 	u_int loop, ifindex;
1733 	struct ipv6_mreq *mreq;
1734 	struct ifnet *ifp;
1735 	struct ip6_moptions *im6o = *im6op;
1736 	struct route_in6 ro;
1737 	struct sockaddr_in6 *dst;
1738 	struct in6_multi_mship *imm;
1739 	struct proc *p = curproc;	/* XXX */
1740 
1741 	if (im6o == NULL) {
1742 		/*
1743 		 * No multicast option buffer attached to the pcb;
1744 		 * allocate one and initialize to default values.
1745 		 */
1746 		im6o = (struct ip6_moptions *)
1747 			malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
1748 
1749 		if (im6o == NULL)
1750 			return (ENOBUFS);
1751 		*im6op = im6o;
1752 		im6o->im6o_multicast_ifp = NULL;
1753 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1754 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
1755 		LIST_INIT(&im6o->im6o_memberships);
1756 	}
1757 
1758 	switch (optname) {
1759 
1760 	case IPV6_MULTICAST_IF:
1761 		/*
1762 		 * Select the interface for outgoing multicast packets.
1763 		 */
1764 		if (m == NULL || m->m_len != sizeof(u_int)) {
1765 			error = EINVAL;
1766 			break;
1767 		}
1768 		bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
1769 		if (ifindex < 0 || if_index < ifindex) {
1770 			error = ENXIO;	/* XXX EINVAL? */
1771 			break;
1772 		}
1773 		ifp = ifindex2ifnet[ifindex];
1774 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1775 			error = EADDRNOTAVAIL;
1776 			break;
1777 		}
1778 		im6o->im6o_multicast_ifp = ifp;
1779 		break;
1780 
1781 	case IPV6_MULTICAST_HOPS:
1782 	    {
1783 		/*
1784 		 * Set the IP6 hoplimit for outgoing multicast packets.
1785 		 */
1786 		int optval;
1787 		if (m == NULL || m->m_len != sizeof(int)) {
1788 			error = EINVAL;
1789 			break;
1790 		}
1791 		bcopy(mtod(m, u_int *), &optval, sizeof(optval));
1792 		if (optval < -1 || optval >= 256)
1793 			error = EINVAL;
1794 		else if (optval == -1)
1795 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1796 		else
1797 			im6o->im6o_multicast_hlim = optval;
1798 		break;
1799 	    }
1800 
1801 	case IPV6_MULTICAST_LOOP:
1802 		/*
1803 		 * Set the loopback flag for outgoing multicast packets.
1804 		 * Must be zero or one.
1805 		 */
1806 		if (m == NULL || m->m_len != sizeof(u_int)) {
1807 			error = EINVAL;
1808 			break;
1809 		}
1810 		bcopy(mtod(m, u_int *), &loop, sizeof(loop));
1811 		if (loop > 1) {
1812 			error = EINVAL;
1813 			break;
1814 		}
1815 		im6o->im6o_multicast_loop = loop;
1816 		break;
1817 
1818 	case IPV6_JOIN_GROUP:
1819 		/*
1820 		 * Add a multicast group membership.
1821 		 * Group must be a valid IP6 multicast address.
1822 		 */
1823 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
1824 			error = EINVAL;
1825 			break;
1826 		}
1827 		mreq = mtod(m, struct ipv6_mreq *);
1828 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
1829 			/*
1830 			 * We use the unspecified address to specify to accept
1831 			 * all multicast addresses. Only super user is allowed
1832 			 * to do this.
1833 			 */
1834 			if (suser(p, 0))
1835 			{
1836 				error = EACCES;
1837 				break;
1838 			}
1839 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
1840 			error = EINVAL;
1841 			break;
1842 		}
1843 
1844 		/*
1845 		 * If the interface is specified, validate it.
1846 		 */
1847 		if (mreq->ipv6mr_interface < 0
1848 		 || if_index < mreq->ipv6mr_interface) {
1849 			error = ENXIO;	/* XXX EINVAL? */
1850 			break;
1851 		}
1852 		/*
1853 		 * If no interface was explicitly specified, choose an
1854 		 * appropriate one according to the given multicast address.
1855 		 */
1856 		if (mreq->ipv6mr_interface == 0) {
1857 			/*
1858 			 * If the multicast address is in node-local scope,
1859 			 * the interface should be a loopback interface.
1860 			 * Otherwise, look up the routing table for the
1861 			 * address, and choose the outgoing interface.
1862 			 *   XXX: is it a good approach?
1863 			 */
1864 			if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
1865 				ifp = lo0ifp;
1866 			} else {
1867 				ro.ro_rt = NULL;
1868 				dst = (struct sockaddr_in6 *)&ro.ro_dst;
1869 				bzero(dst, sizeof(*dst));
1870 				dst->sin6_len = sizeof(struct sockaddr_in6);
1871 				dst->sin6_family = AF_INET6;
1872 				dst->sin6_addr = mreq->ipv6mr_multiaddr;
1873 				rtalloc((struct route *)&ro);
1874 				if (ro.ro_rt == NULL) {
1875 					error = EADDRNOTAVAIL;
1876 					break;
1877 				}
1878 				ifp = ro.ro_rt->rt_ifp;
1879 				rtfree(ro.ro_rt);
1880 			}
1881 		} else
1882 			ifp = ifindex2ifnet[mreq->ipv6mr_interface];
1883 
1884 		/*
1885 		 * See if we found an interface, and confirm that it
1886 		 * supports multicast
1887 		 */
1888 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1889 			error = EADDRNOTAVAIL;
1890 			break;
1891 		}
1892 		/*
1893 		 * Put interface index into the multicast address,
1894 		 * if the address has link-local scope.
1895 		 */
1896 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
1897 			mreq->ipv6mr_multiaddr.s6_addr16[1] =
1898 			    htons(mreq->ipv6mr_interface);
1899 		}
1900 		/*
1901 		 * See if the membership already exists.
1902 		 */
1903 		for (imm = im6o->im6o_memberships.lh_first;
1904 		     imm != NULL; imm = imm->i6mm_chain.le_next)
1905 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
1906 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
1907 			    &mreq->ipv6mr_multiaddr))
1908 				break;
1909 		if (imm != NULL) {
1910 			error = EADDRINUSE;
1911 			break;
1912 		}
1913 		/*
1914 		 * Everything looks good; add a new record to the multicast
1915 		 * address list for the given interface.
1916 		 */
1917 		imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error);
1918 		if (!imm)
1919 			break;
1920 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
1921 		break;
1922 
1923 	case IPV6_LEAVE_GROUP:
1924 		/*
1925 		 * Drop a multicast group membership.
1926 		 * Group must be a valid IP6 multicast address.
1927 		 */
1928 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
1929 			error = EINVAL;
1930 			break;
1931 		}
1932 		mreq = mtod(m, struct ipv6_mreq *);
1933 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
1934 			if (suser(p, 0))
1935 			{
1936 				error = EACCES;
1937 				break;
1938 			}
1939 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
1940 			error = EINVAL;
1941 			break;
1942 		}
1943 		/*
1944 		 * If an interface address was specified, get a pointer
1945 		 * to its ifnet structure.
1946 		 */
1947 		if (mreq->ipv6mr_interface < 0
1948 		 || if_index < mreq->ipv6mr_interface) {
1949 			error = ENXIO;	/* XXX EINVAL? */
1950 			break;
1951 		}
1952 		ifp = ifindex2ifnet[mreq->ipv6mr_interface];
1953 		/*
1954 		 * Put interface index into the multicast address,
1955 		 * if the address has link-local scope.
1956 		 */
1957 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
1958 			mreq->ipv6mr_multiaddr.s6_addr16[1] =
1959 			    htons(mreq->ipv6mr_interface);
1960 		}
1961 		/*
1962 		 * Find the membership in the membership list.
1963 		 */
1964 		for (imm = im6o->im6o_memberships.lh_first;
1965 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
1966 			if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
1967 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
1968 			    &mreq->ipv6mr_multiaddr))
1969 				break;
1970 		}
1971 		if (imm == NULL) {
1972 			/* Unable to resolve interface */
1973 			error = EADDRNOTAVAIL;
1974 			break;
1975 		}
1976 		/*
1977 		 * Give up the multicast address record to which the
1978 		 * membership points.
1979 		 */
1980 		LIST_REMOVE(imm, i6mm_chain);
1981 		in6_leavegroup(imm);
1982 		break;
1983 
1984 	default:
1985 		error = EOPNOTSUPP;
1986 		break;
1987 	}
1988 
1989 	/*
1990 	 * If all options have default values, no need to keep the mbuf.
1991 	 */
1992 	if (im6o->im6o_multicast_ifp == NULL &&
1993 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
1994 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
1995 	    im6o->im6o_memberships.lh_first == NULL) {
1996 		free(*im6op, M_IPMOPTS);
1997 		*im6op = NULL;
1998 	}
1999 
2000 	return (error);
2001 }
2002 
2003 /*
2004  * Return the IP6 multicast options in response to user getsockopt().
2005  */
2006 static int
2007 ip6_getmoptions(optname, im6o, mp)
2008 	int optname;
2009 	struct ip6_moptions *im6o;
2010 	struct mbuf **mp;
2011 {
2012 	u_int *hlim, *loop, *ifindex;
2013 
2014 	*mp = m_get(M_WAIT, MT_SOOPTS);
2015 
2016 	switch (optname) {
2017 
2018 	case IPV6_MULTICAST_IF:
2019 		ifindex = mtod(*mp, u_int *);
2020 		(*mp)->m_len = sizeof(u_int);
2021 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2022 			*ifindex = 0;
2023 		else
2024 			*ifindex = im6o->im6o_multicast_ifp->if_index;
2025 		return (0);
2026 
2027 	case IPV6_MULTICAST_HOPS:
2028 		hlim = mtod(*mp, u_int *);
2029 		(*mp)->m_len = sizeof(u_int);
2030 		if (im6o == NULL)
2031 			*hlim = ip6_defmcasthlim;
2032 		else
2033 			*hlim = im6o->im6o_multicast_hlim;
2034 		return (0);
2035 
2036 	case IPV6_MULTICAST_LOOP:
2037 		loop = mtod(*mp, u_int *);
2038 		(*mp)->m_len = sizeof(u_int);
2039 		if (im6o == NULL)
2040 			*loop = ip6_defmcasthlim;
2041 		else
2042 			*loop = im6o->im6o_multicast_loop;
2043 		return (0);
2044 
2045 	default:
2046 		return (EOPNOTSUPP);
2047 	}
2048 }
2049 
2050 /*
2051  * Discard the IP6 multicast options.
2052  */
2053 void
2054 ip6_freemoptions(im6o)
2055 	struct ip6_moptions *im6o;
2056 {
2057 	struct in6_multi_mship *imm;
2058 
2059 	if (im6o == NULL)
2060 		return;
2061 
2062 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
2063 		LIST_REMOVE(imm, i6mm_chain);
2064 		in6_leavegroup(imm);
2065 	}
2066 	free(im6o, M_IPMOPTS);
2067 }
2068 
2069 /*
2070  * Set IPv6 outgoing packet options based on advanced API.
2071  */
2072 int
2073 ip6_setpktoptions(control, opt, priv)
2074 	struct mbuf *control;
2075 	struct ip6_pktopts *opt;
2076 	int priv;
2077 {
2078 	struct cmsghdr *cm = 0;
2079 
2080 	if (control == 0 || opt == 0)
2081 		return (EINVAL);
2082 
2083 	bzero(opt, sizeof(*opt));
2084 	opt->ip6po_hlim = -1; /* -1 means to use default hop limit */
2085 
2086 	/*
2087 	 * XXX: Currently, we assume all the optional information is stored
2088 	 * in a single mbuf.
2089 	 */
2090 	if (control->m_next)
2091 		return (EINVAL);
2092 
2093 	opt->ip6po_m = control;
2094 
2095 	for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2096 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2097 		cm = mtod(control, struct cmsghdr *);
2098 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2099 			return (EINVAL);
2100 		if (cm->cmsg_level != IPPROTO_IPV6)
2101 			continue;
2102 
2103 		switch (cm->cmsg_type) {
2104 		case IPV6_PKTINFO:
2105 			if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo)))
2106 				return (EINVAL);
2107 			opt->ip6po_pktinfo = (struct in6_pktinfo *)CMSG_DATA(cm);
2108 			if (opt->ip6po_pktinfo->ipi6_ifindex &&
2109 			    IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr))
2110 				opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] =
2111 					htons(opt->ip6po_pktinfo->ipi6_ifindex);
2112 
2113 			if (opt->ip6po_pktinfo->ipi6_ifindex > if_index ||
2114 			    opt->ip6po_pktinfo->ipi6_ifindex < 0) {
2115 				return (ENXIO);
2116 			}
2117 
2118 			/*
2119 			 * Check if the requested source address is indeed a
2120 			 * unicast address assigned to the node, and can be
2121 			 * used as the packet's source address.
2122 			 */
2123 			if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
2124 				struct ifaddr *ia;
2125 				struct in6_ifaddr *ia6;
2126 				struct sockaddr_in6 sin6;
2127 
2128 				bzero(&sin6, sizeof(sin6));
2129 				sin6.sin6_len = sizeof(sin6);
2130 				sin6.sin6_family = AF_INET6;
2131 				sin6.sin6_addr =
2132 					opt->ip6po_pktinfo->ipi6_addr;
2133 				ia = ifa_ifwithaddr(sin6tosa(&sin6));
2134 				if (ia == NULL ||
2135 				    (opt->ip6po_pktinfo->ipi6_ifindex &&
2136 				     (ia->ifa_ifp->if_index !=
2137 				      opt->ip6po_pktinfo->ipi6_ifindex))) {
2138 					return (EADDRNOTAVAIL);
2139 				}
2140 				ia6 = (struct in6_ifaddr *)ia;
2141 				if ((ia6->ia6_flags & (IN6_IFF_ANYCAST|IN6_IFF_NOTREADY)) != 0) {
2142 					return (EADDRNOTAVAIL);
2143 				}
2144 
2145 				/*
2146 				 * Check if the requested source address is
2147 				 * indeed a unicast address assigned to the
2148 				 * node.
2149 				 */
2150 				if (IN6_IS_ADDR_MULTICAST(&opt->ip6po_pktinfo->ipi6_addr))
2151 					return (EADDRNOTAVAIL);
2152 			}
2153 			break;
2154 
2155 		case IPV6_HOPLIMIT:
2156 			if (cm->cmsg_len != CMSG_LEN(sizeof(int)))
2157 				return (EINVAL);
2158 
2159 			bcopy(CMSG_DATA(cm), &opt->ip6po_hlim,
2160 			    sizeof(opt->ip6po_hlim));
2161 			if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255)
2162 				return (EINVAL);
2163 			break;
2164 
2165 		case IPV6_NEXTHOP:
2166 			if (!priv)
2167 				return (EPERM);
2168 
2169 			/* check if cmsg_len is large enough for sa_len */
2170 			if (cm->cmsg_len < sizeof(u_char) ||
2171 			    cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm)))
2172 				return (EINVAL);
2173 
2174 			opt->ip6po_nexthop = (struct sockaddr *)CMSG_DATA(cm);
2175 
2176 			break;
2177 
2178 		case IPV6_HOPOPTS:
2179 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh)))
2180 				return (EINVAL);
2181 			opt->ip6po_hbh = (struct ip6_hbh *)CMSG_DATA(cm);
2182 			if (cm->cmsg_len !=
2183 			    CMSG_LEN((opt->ip6po_hbh->ip6h_len + 1) << 3))
2184 				return (EINVAL);
2185 			break;
2186 
2187 		case IPV6_DSTOPTS:
2188 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest)))
2189 				return (EINVAL);
2190 
2191 			/*
2192 			 * If there is no routing header yet, the destination
2193 			 * options header should be put on the 1st part.
2194 			 * Otherwise, the header should be on the 2nd part.
2195 			 * (See RFC 2460, section 4.1)
2196 			 */
2197 			if (opt->ip6po_rthdr == NULL) {
2198 				opt->ip6po_dest1 =
2199 				    (struct ip6_dest *)CMSG_DATA(cm);
2200 				if (cm->cmsg_len !=
2201 				    CMSG_LEN((opt->ip6po_dest1->ip6d_len + 1) << 3));
2202 					return (EINVAL);
2203 			}
2204 			else {
2205 				opt->ip6po_dest2 =
2206 				    (struct ip6_dest *)CMSG_DATA(cm);
2207 				if (cm->cmsg_len !=
2208 				    CMSG_LEN((opt->ip6po_dest2->ip6d_len + 1) << 3))
2209 					return (EINVAL);
2210 			}
2211 			break;
2212 
2213 		case IPV6_RTHDR:
2214 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr)))
2215 				return (EINVAL);
2216 			opt->ip6po_rthdr = (struct ip6_rthdr *)CMSG_DATA(cm);
2217 			if (cm->cmsg_len !=
2218 			    CMSG_LEN((opt->ip6po_rthdr->ip6r_len + 1) << 3))
2219 				return (EINVAL);
2220 			switch (opt->ip6po_rthdr->ip6r_type) {
2221 			case IPV6_RTHDR_TYPE_0:
2222 				if (opt->ip6po_rthdr->ip6r_segleft == 0)
2223 					return (EINVAL);
2224 				break;
2225 			default:
2226 				return (EINVAL);
2227 			}
2228 			break;
2229 
2230 		default:
2231 			return (ENOPROTOOPT);
2232 		}
2233 	}
2234 
2235 	return (0);
2236 }
2237 
2238 /*
2239  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2240  * packet to the input queue of a specified interface.  Note that this
2241  * calls the output routine of the loopback "driver", but with an interface
2242  * pointer that might NOT be lo0ifp -- easier than replicating that code here.
2243  */
2244 void
2245 ip6_mloopback(ifp, m, dst)
2246 	struct ifnet *ifp;
2247 	struct mbuf *m;
2248 	struct sockaddr_in6 *dst;
2249 {
2250 	struct mbuf *copym;
2251 	struct ip6_hdr *ip6;
2252 
2253 	copym = m_copy(m, 0, M_COPYALL);
2254 	if (copym == NULL)
2255 		return;
2256 
2257 	/*
2258 	 * Make sure to deep-copy IPv6 header portion in case the data
2259 	 * is in an mbuf cluster, so that we can safely override the IPv6
2260 	 * header portion later.
2261 	 */
2262 	if ((copym->m_flags & M_EXT) != 0 ||
2263 	    copym->m_len < sizeof(struct ip6_hdr)) {
2264 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
2265 		if (copym == NULL)
2266 			return;
2267 	}
2268 
2269 #ifdef DIAGNOSTIC
2270 	if (copym->m_len < sizeof(*ip6)) {
2271 		m_freem(copym);
2272 		return;
2273 	}
2274 #endif
2275 
2276 	ip6 = mtod(copym, struct ip6_hdr *);
2277 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
2278 		ip6->ip6_src.s6_addr16[1] = 0;
2279 	if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
2280 		ip6->ip6_dst.s6_addr16[1] = 0;
2281 
2282 	(void)looutput(ifp, copym, (struct sockaddr *)dst, NULL);
2283 }
2284 
2285 /*
2286  * Chop IPv6 header off from the payload.
2287  */
2288 static int
2289 ip6_splithdr(m, exthdrs)
2290 	struct mbuf *m;
2291 	struct ip6_exthdrs *exthdrs;
2292 {
2293 	struct mbuf *mh;
2294 	struct ip6_hdr *ip6;
2295 
2296 	ip6 = mtod(m, struct ip6_hdr *);
2297 	if (m->m_len > sizeof(*ip6)) {
2298 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
2299 		if (mh == 0) {
2300 			m_freem(m);
2301 			return ENOBUFS;
2302 		}
2303 		M_MOVE_PKTHDR(mh, m);
2304 		MH_ALIGN(mh, sizeof(*ip6));
2305 		m->m_len -= sizeof(*ip6);
2306 		m->m_data += sizeof(*ip6);
2307 		mh->m_next = m;
2308 		m = mh;
2309 		m->m_len = sizeof(*ip6);
2310 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2311 	}
2312 	exthdrs->ip6e_ip6 = m;
2313 	return 0;
2314 }
2315 
2316 /*
2317  * Compute IPv6 extension header length.
2318  */
2319 int
2320 ip6_optlen(inp)
2321 	struct inpcb *inp;
2322 {
2323 	int len;
2324 
2325 	if (!inp->inp_outputopts6)
2326 		return 0;
2327 
2328 	len = 0;
2329 #define elen(x) \
2330     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2331 
2332 	len += elen(inp->inp_outputopts6->ip6po_hbh);
2333 	len += elen(inp->inp_outputopts6->ip6po_dest1);
2334 	len += elen(inp->inp_outputopts6->ip6po_rthdr);
2335 	len += elen(inp->inp_outputopts6->ip6po_dest2);
2336 	return len;
2337 #undef elen
2338 }
2339