xref: /openbsd-src/sys/netinet6/ip6_output.c (revision d13be5d47e4149db2549a9828e244d59dbc43f15)
1 /*	$OpenBSD: ip6_output.c,v 1.122 2011/07/04 06:54:49 claudio 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/if_enc.h>
78 #include <net/route.h>
79 
80 #include <netinet/in.h>
81 #include <netinet/in_var.h>
82 #include <netinet/in_systm.h>
83 #include <netinet/ip.h>
84 #include <netinet/in_pcb.h>
85 
86 #include <netinet/ip6.h>
87 #include <netinet/icmp6.h>
88 #include <netinet6/ip6_var.h>
89 #include <netinet6/nd6.h>
90 #include <netinet6/ip6protosw.h>
91 
92 #include <crypto/idgen.h>
93 
94 #if NPF > 0
95 #include <net/pfvar.h>
96 #endif
97 
98 #ifdef IPSEC
99 #include <netinet/ip_ipsp.h>
100 #include <netinet/ip_ah.h>
101 #include <netinet/ip_esp.h>
102 #include <netinet/udp.h>
103 #include <netinet/tcp.h>
104 #include <net/pfkeyv2.h>
105 
106 extern u_int8_t get_sa_require(struct inpcb *);
107 
108 extern int ipsec_auth_default_level;
109 extern int ipsec_esp_trans_default_level;
110 extern int ipsec_esp_network_default_level;
111 extern int ipsec_ipcomp_default_level;
112 #endif /* IPSEC */
113 
114 struct ip6_exthdrs {
115 	struct mbuf *ip6e_ip6;
116 	struct mbuf *ip6e_hbh;
117 	struct mbuf *ip6e_dest1;
118 	struct mbuf *ip6e_rthdr;
119 	struct mbuf *ip6e_dest2;
120 };
121 
122 int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **, int, int);
123 int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *, struct socket *);
124 int ip6_getpcbopt(struct ip6_pktopts *, int, struct mbuf **);
125 int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, int, int,
126 	int, int);
127 int ip6_setmoptions(int, struct ip6_moptions **, struct mbuf *);
128 int ip6_getmoptions(int, struct ip6_moptions *, struct mbuf **);
129 int ip6_copyexthdr(struct mbuf **, caddr_t, int);
130 int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
131 	struct ip6_frag **);
132 int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
133 int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
134 int ip6_getpmtu(struct route_in6 *, struct route_in6 *,
135 	struct ifnet *, struct in6_addr *, u_long *, int *);
136 int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
137 
138 /* Context for non-repeating IDs */
139 struct idgen32_ctx ip6_id_ctx;
140 
141 /*
142  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
143  * header (with pri, len, nxt, hlim, src, dst).
144  * This function may modify ver and hlim only.
145  * The mbuf chain containing the packet will be freed.
146  * The mbuf opt, if present, will not be freed.
147  *
148  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
149  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
150  * which is rt_rmx.rmx_mtu.
151  *
152  * ifpp - XXX: just for statistics
153  */
154 int
155 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt, struct route_in6 *ro,
156     int flags, struct ip6_moptions *im6o, struct ifnet **ifpp,
157     struct inpcb *inp)
158 {
159 	struct ip6_hdr *ip6;
160 	struct ifnet *ifp, *origifp = NULL;
161 	struct mbuf *m = m0;
162 	int hlen, tlen;
163 	struct route_in6 ip6route;
164 	struct rtentry *rt = NULL;
165 	struct sockaddr_in6 *dst, dstsock;
166 	int error = 0;
167 	struct in6_ifaddr *ia = NULL;
168 	u_long mtu;
169 	int alwaysfrag, dontfrag;
170 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
171 	struct ip6_exthdrs exthdrs;
172 	struct in6_addr finaldst;
173 	struct route_in6 *ro_pmtu = NULL;
174 	int hdrsplit = 0;
175 	u_int8_t sproto = 0;
176 #ifdef IPSEC
177 	struct m_tag *mtag;
178 	union sockaddr_union sdst;
179 	struct tdb_ident *tdbi;
180 	u_int32_t sspi;
181 	struct tdb *tdb;
182 	int s;
183 #if NPF > 0
184 	struct ifnet *encif;
185 #endif
186 #endif /* IPSEC */
187 
188 #ifdef IPSEC
189 	if (inp && (inp->inp_flags & INP_IPV6) == 0)
190 		panic("ip6_output: IPv4 pcb is passed");
191 #endif /* IPSEC */
192 
193 	ip6 = mtod(m, struct ip6_hdr *);
194 	finaldst = ip6->ip6_dst;
195 
196 #define MAKE_EXTHDR(hp, mp)						\
197     do {								\
198 	if (hp) {							\
199 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
200 		error = ip6_copyexthdr((mp), (caddr_t)(hp), 		\
201 		    ((eh)->ip6e_len + 1) << 3);				\
202 		if (error)						\
203 			goto freehdrs;					\
204 	}								\
205     } while (0)
206 
207 	bzero(&exthdrs, sizeof(exthdrs));
208 
209 	if (opt) {
210 		/* Hop-by-Hop options header */
211 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
212 		/* Destination options header(1st part) */
213 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
214 		/* Routing header */
215 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
216 		/* Destination options header(2nd part) */
217 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
218 	}
219 
220 #ifdef IPSEC
221 	if (!ipsec_in_use && !inp)
222 		goto done_spd;
223 
224 	/*
225 	 * splnet is chosen over spltdb because we are not allowed to
226 	 * lower the level, and udp6_output calls us in splnet(). XXX check
227 	 */
228 	s = splnet();
229 
230 	/*
231 	 * Check if there was an outgoing SA bound to the flow
232 	 * from a transport protocol.
233 	 */
234 	ip6 = mtod(m, struct ip6_hdr *);
235 
236 	/* Do we have any pending SAs to apply ? */
237 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
238 	if (mtag != NULL) {
239 #ifdef DIAGNOSTIC
240 		if (mtag->m_tag_len != sizeof (struct tdb_ident))
241 			panic("ip6_output: tag of length %d (should be %d",
242 			    mtag->m_tag_len, sizeof (struct tdb_ident));
243 #endif
244 		tdbi = (struct tdb_ident *)(mtag + 1);
245 		tdb = gettdb(tdbi->rdomain, tdbi->spi, &tdbi->dst, tdbi->proto);
246 		if (tdb == NULL)
247 			error = -EINVAL;
248 		m_tag_delete(m, mtag);
249 	} else
250 		tdb = ipsp_spd_lookup(m, AF_INET6, sizeof(struct ip6_hdr),
251 		    &error, IPSP_DIRECTION_OUT, NULL, inp);
252 
253 	if (tdb == NULL) {
254 	        splx(s);
255 
256 		if (error == 0) {
257 		        /*
258 			 * No IPsec processing required, we'll just send the
259 			 * packet out.
260 			 */
261 		        sproto = 0;
262 
263 			/* Fall through to routing/multicast handling */
264 		} else {
265 		        /*
266 			 * -EINVAL is used to indicate that the packet should
267 			 * be silently dropped, typically because we've asked
268 			 * key management for an SA.
269 			 */
270 		        if (error == -EINVAL) /* Should silently drop packet */
271 				error = 0;
272 
273 			goto freehdrs;
274 		}
275 	} else {
276 		/* Loop detection */
277 		for (mtag = m_tag_first(m); mtag != NULL;
278 		    mtag = m_tag_next(m, mtag)) {
279 			if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
280 			    mtag->m_tag_id !=
281 			    PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
282 				continue;
283 			tdbi = (struct tdb_ident *)(mtag + 1);
284 			if (tdbi->spi == tdb->tdb_spi &&
285 			    tdbi->proto == tdb->tdb_sproto &&
286 			    tdbi->rdomain == tdb->tdb_rdomain &&
287 			    !bcmp(&tdbi->dst, &tdb->tdb_dst,
288 			    sizeof(union sockaddr_union))) {
289 				splx(s);
290 				sproto = 0; /* mark as no-IPsec-needed */
291 				goto done_spd;
292 			}
293 		}
294 
295 	        /* We need to do IPsec */
296 	        bcopy(&tdb->tdb_dst, &sdst, sizeof(sdst));
297 		sspi = tdb->tdb_spi;
298 		sproto = tdb->tdb_sproto;
299 	        splx(s);
300 	}
301 
302 	/* Fall through to the routing/multicast handling code */
303  done_spd:
304 #endif /* IPSEC */
305 
306 	/*
307 	 * Calculate the total length of the extension header chain.
308 	 * Keep the length of the unfragmentable part for fragmentation.
309 	 */
310 	optlen = 0;
311 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
312 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
313 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
314 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
315 	/* NOTE: we don't add AH/ESP length here. do that later. */
316 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
317 
318 	/*
319 	 * If we need IPsec, or there is at least one extension header,
320 	 * separate IP6 header from the payload.
321 	 */
322 	if ((sproto || optlen) && !hdrsplit) {
323 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
324 			m = NULL;
325 			goto freehdrs;
326 		}
327 		m = exthdrs.ip6e_ip6;
328 		hdrsplit++;
329 	}
330 
331 	/* adjust pointer */
332 	ip6 = mtod(m, struct ip6_hdr *);
333 
334 	/* adjust mbuf packet header length */
335 	m->m_pkthdr.len += optlen;
336 	plen = m->m_pkthdr.len - sizeof(*ip6);
337 
338 	/* If this is a jumbo payload, insert a jumbo payload option. */
339 	if (plen > IPV6_MAXPACKET) {
340 		if (!hdrsplit) {
341 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
342 				m = NULL;
343 				goto freehdrs;
344 			}
345 			m = exthdrs.ip6e_ip6;
346 			hdrsplit++;
347 		}
348 		/* adjust pointer */
349 		ip6 = mtod(m, struct ip6_hdr *);
350 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
351 			goto freehdrs;
352 		ip6->ip6_plen = 0;
353 	} else
354 		ip6->ip6_plen = htons(plen);
355 
356 	/*
357 	 * Concatenate headers and fill in next header fields.
358 	 * Here we have, on "m"
359 	 *	IPv6 payload
360 	 * and we insert headers accordingly.  Finally, we should be getting:
361 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
362 	 *
363 	 * during the header composing process, "m" points to IPv6 header.
364 	 * "mprev" points to an extension header prior to esp.
365 	 */
366 	{
367 		u_char *nexthdrp = &ip6->ip6_nxt;
368 		struct mbuf *mprev = m;
369 
370 		/*
371 		 * we treat dest2 specially.  this makes IPsec processing
372 		 * much easier.  the goal here is to make mprev point the
373 		 * mbuf prior to dest2.
374 		 *
375 		 * result: IPv6 dest2 payload
376 		 * m and mprev will point to IPv6 header.
377 		 */
378 		if (exthdrs.ip6e_dest2) {
379 			if (!hdrsplit)
380 				panic("assumption failed: hdr not split");
381 			exthdrs.ip6e_dest2->m_next = m->m_next;
382 			m->m_next = exthdrs.ip6e_dest2;
383 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
384 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
385 		}
386 
387 #define MAKE_CHAIN(m, mp, p, i)\
388     do {\
389 	if (m) {\
390 		if (!hdrsplit) \
391 			panic("assumption failed: hdr not split"); \
392 		*mtod((m), u_char *) = *(p);\
393 		*(p) = (i);\
394 		p = mtod((m), u_char *);\
395 		(m)->m_next = (mp)->m_next;\
396 		(mp)->m_next = (m);\
397 		(mp) = (m);\
398 	}\
399     } while (0)
400 		/*
401 		 * result: IPv6 hbh dest1 rthdr dest2 payload
402 		 * m will point to IPv6 header.  mprev will point to the
403 		 * extension header prior to dest2 (rthdr in the above case).
404 		 */
405 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
406 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
407 		    IPPROTO_DSTOPTS);
408 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
409 		    IPPROTO_ROUTING);
410 	}
411 
412 	/*
413 	 * If there is a routing header, replace the destination address field
414 	 * with the first hop of the routing header.
415 	 */
416 	if (exthdrs.ip6e_rthdr) {
417 		struct ip6_rthdr *rh;
418 		struct ip6_rthdr0 *rh0;
419 		struct in6_addr *addr;
420 
421 		rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
422 		    struct ip6_rthdr *));
423 		switch (rh->ip6r_type) {
424 		case IPV6_RTHDR_TYPE_0:
425 			 rh0 = (struct ip6_rthdr0 *)rh;
426 			 addr = (struct in6_addr *)(rh0 + 1);
427 			 ip6->ip6_dst = addr[0];
428 			 bcopy(&addr[1], &addr[0],
429 			     sizeof(struct in6_addr) * (rh0->ip6r0_segleft - 1));
430 			 addr[rh0->ip6r0_segleft - 1] = finaldst;
431 			 break;
432 		default:	/* is it possible? */
433 			 error = EINVAL;
434 			 goto bad;
435 		}
436 	}
437 
438 	/* Source address validation */
439 	if (!(flags & IPV6_UNSPECSRC) &&
440 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
441 		/*
442 		 * XXX: we can probably assume validation in the caller, but
443 		 * we explicitly check the address here for safety.
444 		 */
445 		error = EOPNOTSUPP;
446 		ip6stat.ip6s_badscope++;
447 		goto bad;
448 	}
449 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
450 		error = EOPNOTSUPP;
451 		ip6stat.ip6s_badscope++;
452 		goto bad;
453 	}
454 
455 	ip6stat.ip6s_localout++;
456 
457 	/*
458 	 * Route packet.
459 	 */
460 #if NPF > 0
461 reroute:
462 #endif
463 
464 	/* initialize cached route */
465 	if (ro == 0) {
466 		ro = &ip6route;
467 		bzero((caddr_t)ro, sizeof(*ro));
468 	}
469 	ro_pmtu = ro;
470 	if (opt && opt->ip6po_rthdr)
471 		ro = &opt->ip6po_route;
472 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
473 
474 	/*
475 	 * if specified, try to fill in the traffic class field.
476 	 * do not override if a non-zero value is already set.
477 	 * we check the diffserv field and the ecn field separately.
478 	 */
479 	if (opt && opt->ip6po_tclass >= 0) {
480 		int mask = 0;
481 
482 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
483 			mask |= 0xfc;
484 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
485 			mask |= 0x03;
486 		if (mask != 0)
487 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
488 	}
489 
490 	/* fill in or override the hop limit field, if necessary. */
491 	if (opt && opt->ip6po_hlim != -1)
492 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
493 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
494 		if (im6o != NULL)
495 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
496 		else
497 			ip6->ip6_hlim = ip6_defmcasthlim;
498 	}
499 
500 #ifdef IPSEC
501 	/*
502 	 * Check if the packet needs encapsulation.
503 	 * ipsp_process_packet will never come back to here.
504 	 */
505 	if (sproto != 0) {
506 	        s = splnet();
507 
508 		/*
509 		 * XXX what should we do if ip6_hlim == 0 and the
510 		 * packet gets tunneled?
511 		 */
512 
513 		tdb = gettdb(rtable_l2(m->m_pkthdr.rdomain),
514 		    sspi, &sdst, sproto);
515 		if (tdb == NULL) {
516 			splx(s);
517 			error = EHOSTUNREACH;
518 			m_freem(m);
519 			goto done;
520 		}
521 
522 #if NPF > 0
523 		if ((encif = enc_getif(tdb->tdb_rdomain,
524 		    tdb->tdb_tap)) == NULL ||
525 		    pf_test(AF_INET6, PF_OUT, encif, &m, NULL) != PF_PASS) {
526 			splx(s);
527 			error = EHOSTUNREACH;
528 			m_freem(m);
529 			goto done;
530 		}
531 		if (m == NULL) {
532 			splx(s);
533 			goto done;
534 		}
535 		ip6 = mtod(m, struct ip6_hdr *);
536 		/*
537 		 * PF_TAG_REROUTE handling or not...
538 		 * Packet is entering IPsec so the routing is
539 		 * already overruled by the IPsec policy.
540 		 * Until now the change was not reconsidered.
541 		 * What's the behaviour?
542 		 */
543 #endif
544 
545 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
546 
547 		/* Callee frees mbuf */
548 		/*
549 		 * if we are source-routing, do not attempt to tunnel the
550 		 * packet just because ip6_dst is different from what tdb has.
551 		 * XXX
552 		 */
553 		error = ipsp_process_packet(m, tdb, AF_INET6,
554 		    exthdrs.ip6e_rthdr ? 1 : 0);
555 		splx(s);
556 
557 		return error;  /* Nothing more to be done */
558 	}
559 #endif /* IPSEC */
560 
561 	bzero(&dstsock, sizeof(dstsock));
562 	dstsock.sin6_family = AF_INET6;
563 	dstsock.sin6_addr = ip6->ip6_dst;
564 	dstsock.sin6_len = sizeof(dstsock);
565 	if ((error = in6_selectroute(&dstsock, opt, im6o, ro, &ifp,
566 	    &rt)) != 0) {
567 		switch (error) {
568 		case EHOSTUNREACH:
569 			ip6stat.ip6s_noroute++;
570 			break;
571 		case EADDRNOTAVAIL:
572 		default:
573 			break;	/* XXX statistics? */
574 		}
575 		if (ifp != NULL)
576 			in6_ifstat_inc(ifp, ifs6_out_discard);
577 		goto bad;
578 	}
579 	if (rt == NULL) {
580 		/*
581 		 * If in6_selectroute() does not return a route entry,
582 		 * dst may not have been updated.
583 		 */
584 		*dst = dstsock;	/* XXX */
585 	}
586 
587 	/*
588 	 * then rt (for unicast) and ifp must be non-NULL valid values.
589 	 */
590 	if (rt) {
591 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
592 		rt->rt_use++;
593 	}
594 
595 	if ((flags & IPV6_FORWARDING) == 0) {
596 		/* XXX: the FORWARDING flag can be set for mrouting. */
597 		in6_ifstat_inc(ifp, ifs6_out_request);
598 	}
599 
600 	/*
601 	 * The outgoing interface must be in the zone of source and
602 	 * destination addresses.  We should use ia_ifp to support the
603 	 * case of sending packets to an address of our own.
604 	 */
605 	if (ia != NULL && ia->ia_ifp)
606 		origifp = ia->ia_ifp;
607 	else
608 		origifp = ifp;
609 
610 	if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
611 		if (opt && opt->ip6po_nextroute.ro_rt) {
612 			/*
613 			 * The nexthop is explicitly specified by the
614 			 * application.  We assume the next hop is an IPv6
615 			 * address.
616 			 */
617 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
618 		} else if ((rt->rt_flags & RTF_GATEWAY))
619 			dst = (struct sockaddr_in6 *)rt->rt_gateway;
620 	}
621 
622 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
623 		/* Unicast */
624 
625 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
626 	} else {
627 		/* Multicast */
628 		struct	in6_multi *in6m;
629 
630 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
631 
632 		in6_ifstat_inc(ifp, ifs6_out_mcast);
633 
634 		/*
635 		 * Confirm that the outgoing interface supports multicast.
636 		 */
637 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
638 			ip6stat.ip6s_noroute++;
639 			in6_ifstat_inc(ifp, ifs6_out_discard);
640 			error = ENETUNREACH;
641 			goto bad;
642 		}
643 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
644 		if (in6m != NULL &&
645 		    (im6o == NULL || im6o->im6o_multicast_loop)) {
646 			/*
647 			 * If we belong to the destination multicast group
648 			 * on the outgoing interface, and the caller did not
649 			 * forbid loopback, loop back a copy.
650 			 */
651 			ip6_mloopback(ifp, m, dst);
652 		} else {
653 			/*
654 			 * If we are acting as a multicast router, perform
655 			 * multicast forwarding as if the packet had just
656 			 * arrived on the interface to which we are about
657 			 * to send.  The multicast forwarding function
658 			 * recursively calls this function, using the
659 			 * IPV6_FORWARDING flag to prevent infinite recursion.
660 			 *
661 			 * Multicasts that are looped back by ip6_mloopback(),
662 			 * above, will be forwarded by the ip6_input() routine,
663 			 * if necessary.
664 			 */
665 #ifdef MROUTING
666 			if (ip6_mforwarding && ip6_mrouter &&
667 			    (flags & IPV6_FORWARDING) == 0) {
668 				if (ip6_mforward(ip6, ifp, m) != 0) {
669 					m_freem(m);
670 					goto done;
671 				}
672 			}
673 #endif
674 		}
675 		/*
676 		 * Multicasts with a hoplimit of zero may be looped back,
677 		 * above, but must not be transmitted on a network.
678 		 * Also, multicasts addressed to the loopback interface
679 		 * are not sent -- the above call to ip6_mloopback() will
680 		 * loop back a copy if this host actually belongs to the
681 		 * destination group on the loopback interface.
682 		 */
683 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
684 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
685 			m_freem(m);
686 			goto done;
687 		}
688 	}
689 
690 	/*
691 	 * Fill the outgoing interface to tell the upper layer
692 	 * to increment per-interface statistics.
693 	 */
694 	if (ifpp)
695 		*ifpp = ifp;
696 
697 	/* Determine path MTU. */
698 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
699 	    &alwaysfrag)) != 0)
700 		goto bad;
701 
702 	/*
703 	 * The caller of this function may specify to use the minimum MTU
704 	 * in some cases.
705 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
706 	 * setting.  The logic is a bit complicated; by default, unicast
707 	 * packets will follow path MTU while multicast packets will be sent at
708 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
709 	 * including unicast ones will be sent at the minimum MTU.  Multicast
710 	 * packets will always be sent at the minimum MTU unless
711 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
712 	 * See RFC 3542 for more details.
713 	 */
714 	if (mtu > IPV6_MMTU) {
715 		if ((flags & IPV6_MINMTU))
716 			mtu = IPV6_MMTU;
717 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
718 			mtu = IPV6_MMTU;
719 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
720 			 (opt == NULL ||
721 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
722 			mtu = IPV6_MMTU;
723 		}
724 	}
725 
726 	/* Fake scoped addresses */
727 	if ((ifp->if_flags & IFF_LOOPBACK) != 0) {
728 		/*
729 		 * If source or destination address is a scoped address, and
730 		 * the packet is going to be sent to a loopback interface,
731 		 * we should keep the original interface.
732 		 */
733 
734 		/*
735 		 * XXX: this is a very experimental and temporary solution.
736 		 * We eventually have sockaddr_in6 and use the sin6_scope_id
737 		 * field of the structure here.
738 		 * We rely on the consistency between two scope zone ids
739 		 * of source add destination, which should already be assured
740 		 * Larger scopes than link will be supported in the near
741 		 * future.
742 		 */
743 		origifp = NULL;
744 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
745 			origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
746 		else if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
747 			origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
748 		/*
749 		 * XXX: origifp can be NULL even in those two cases above.
750 		 * For example, if we remove the (only) link-local address
751 		 * from the loopback interface, and try to send a link-local
752 		 * address without link-id information.  Then the source
753 		 * address is ::1, and the destination address is the
754 		 * link-local address with its s6_addr16[1] being zero.
755 		 * What is worse, if the packet goes to the loopback interface
756 		 * by a default rejected route, the null pointer would be
757 		 * passed to looutput, and the kernel would hang.
758 		 * The following last resort would prevent such disaster.
759 		 */
760 		if (origifp == NULL)
761 			origifp = ifp;
762 	} else
763 		origifp = ifp;
764 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
765 		ip6->ip6_src.s6_addr16[1] = 0;
766 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
767 		ip6->ip6_dst.s6_addr16[1] = 0;
768 
769 	/*
770 	 * If the outgoing packet contains a hop-by-hop options header,
771 	 * it must be examined and processed even by the source node.
772 	 * (RFC 2460, section 4.)
773 	 */
774 	if (exthdrs.ip6e_hbh) {
775 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
776 		u_int32_t dummy1; /* XXX unused */
777 		u_int32_t dummy2; /* XXX unused */
778 
779 		/*
780 		 *  XXX: if we have to send an ICMPv6 error to the sender,
781 		 *       we need the M_LOOP flag since icmp6_error() expects
782 		 *       the IPv6 and the hop-by-hop options header are
783 		 *       continuous unless the flag is set.
784 		 */
785 		m->m_flags |= M_LOOP;
786 		m->m_pkthdr.rcvif = ifp;
787 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
788 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
789 		    &dummy1, &dummy2) < 0) {
790 			/* m was already freed at this point */
791 			error = EINVAL;/* better error? */
792 			goto done;
793 		}
794 		m->m_flags &= ~M_LOOP; /* XXX */
795 		m->m_pkthdr.rcvif = NULL;
796 	}
797 
798 #if NPF > 0
799 	if (pf_test(AF_INET6, PF_OUT, ifp, &m, NULL) != PF_PASS) {
800 		error = EHOSTUNREACH;
801 		m_freem(m);
802 		goto done;
803 	}
804 	if (m == NULL)
805 		goto done;
806 	ip6 = mtod(m, struct ip6_hdr *);
807 	if ((m->m_pkthdr.pf.flags & (PF_TAG_REROUTE | PF_TAG_GENERATED)) ==
808 	    (PF_TAG_REROUTE | PF_TAG_GENERATED)) {
809 		/* already rerun the route lookup, go on */
810 		m->m_pkthdr.pf.flags &= ~(PF_TAG_GENERATED | PF_TAG_REROUTE);
811 	} else if (m->m_pkthdr.pf.flags & PF_TAG_REROUTE) {
812 		/* tag as generated to skip over pf_test on rerun */
813 		m->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
814 		finaldst = ip6->ip6_dst;
815 		ro = NULL;
816 		goto reroute;
817 	}
818 #endif
819 
820 	/*
821 	 * Send the packet to the outgoing interface.
822 	 * If necessary, do IPv6 fragmentation before sending.
823 	 *
824 	 * the logic here is rather complex:
825 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
826 	 * 1-a: send as is if tlen <= path mtu
827 	 * 1-b: fragment if tlen > path mtu
828 	 *
829 	 * 2: if user asks us not to fragment (dontfrag == 1)
830 	 * 2-a: send as is if tlen <= interface mtu
831 	 * 2-b: error if tlen > interface mtu
832 	 *
833 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
834 	 *      always fragment
835 	 *
836 	 * 4: if dontfrag == 1 && alwaysfrag == 1
837 	 *      error, as we cannot handle this conflicting request
838 	 */
839 	tlen = m->m_pkthdr.len;
840 
841 	if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
842 		dontfrag = 1;
843 	else
844 		dontfrag = 0;
845 	if (dontfrag && alwaysfrag) {	/* case 4 */
846 		/* conflicting request - can't transmit */
847 		error = EMSGSIZE;
848 		goto bad;
849 	}
850 	if (dontfrag && tlen > IN6_LINKMTU(ifp)) {	/* case 2-b */
851 		/*
852 		 * Even if the DONTFRAG option is specified, we cannot send the
853 		 * packet when the data length is larger than the MTU of the
854 		 * outgoing interface.
855 		 * Notify the error by sending IPV6_PATHMTU ancillary data as
856 		 * well as returning an error code (the latter is not described
857 		 * in the API spec.)
858 		 */
859 #if 0
860 		u_int32_t mtu32;
861 		struct ip6ctlparam ip6cp;
862 
863 		mtu32 = (u_int32_t)mtu;
864 		bzero(&ip6cp, sizeof(ip6cp));
865 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
866 		pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
867 		    (void *)&ip6cp);
868 #endif
869 
870 		error = EMSGSIZE;
871 		goto bad;
872 	}
873 
874 	/*
875 	 * transmit packet without fragmentation
876 	 */
877 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
878 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
879 		goto done;
880 	}
881 
882 	/*
883 	 * try to fragment the packet.  case 1-b and 3
884 	 */
885 	if (mtu < IPV6_MMTU) {
886 		/* path MTU cannot be less than IPV6_MMTU */
887 		error = EMSGSIZE;
888 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
889 		goto bad;
890 	} else if (ip6->ip6_plen == 0) {
891 		/* jumbo payload cannot be fragmented */
892 		error = EMSGSIZE;
893 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
894 		goto bad;
895 	} else {
896 		u_char nextproto;
897 #if 0
898 		struct ip6ctlparam ip6cp;
899 		u_int32_t mtu32;
900 #endif
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 
911 #if 0
912 		/* Notify a proper path MTU to applications. */
913 		mtu32 = (u_int32_t)mtu;
914 		bzero(&ip6cp, sizeof(ip6cp));
915 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
916 		pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
917 		    (void *)&ip6cp);
918 #endif
919 
920 		/*
921 		 * Change the next header field of the last header in the
922 		 * unfragmentable part.
923 		 */
924 		if (exthdrs.ip6e_rthdr) {
925 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
926 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
927 		} else if (exthdrs.ip6e_dest1) {
928 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
929 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
930 		} else if (exthdrs.ip6e_hbh) {
931 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
932 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
933 		} else {
934 			nextproto = ip6->ip6_nxt;
935 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
936 		}
937 
938 		m0 = m;
939 		error = ip6_fragment(m0, hlen, nextproto, mtu);
940 
941 		switch (error) {
942 		case 0:
943 			in6_ifstat_inc(ifp, ifs6_out_fragok);
944 			break;
945 		case EMSGSIZE:
946 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
947 			break;
948 		default:
949 			ip6stat.ip6s_odropped++;
950 			break;
951 		}
952 	}
953 
954 	/*
955 	 * Remove leading garbages.
956 	 */
957 	m = m0->m_nextpkt;
958 	m0->m_nextpkt = 0;
959 	m_freem(m0);
960 	for (m0 = m; m; m = m0) {
961 		m0 = m->m_nextpkt;
962 		m->m_nextpkt = 0;
963 		if (error == 0) {
964 			ip6stat.ip6s_ofragments++;
965 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
966 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
967 		} else
968 			m_freem(m);
969 	}
970 
971 	if (error == 0)
972 		ip6stat.ip6s_fragmented++;
973 
974 done:
975 	if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
976 		RTFREE(ro->ro_rt);
977 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
978 		RTFREE(ro_pmtu->ro_rt);
979 	}
980 
981 	return (error);
982 
983 freehdrs:
984 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
985 	m_freem(exthdrs.ip6e_dest1);
986 	m_freem(exthdrs.ip6e_rthdr);
987 	m_freem(exthdrs.ip6e_dest2);
988 	/* FALLTHROUGH */
989 bad:
990 	m_freem(m);
991 	goto done;
992 }
993 
994 int
995 ip6_fragment(struct mbuf *m0, int hlen, u_char nextproto, u_long mtu)
996 {
997 	struct mbuf	*m, **mnext, *m_frgpart;
998 	struct ip6_hdr	*mhip6;
999 	struct ip6_frag	*ip6f;
1000 	u_int32_t	 id;
1001 	int		 tlen, len, off;
1002 	int		 error;
1003 
1004 	id = htonl(ip6_randomid());
1005 
1006 	mnext = &m0->m_nextpkt;
1007 	*mnext = NULL;
1008 
1009 	tlen = m0->m_pkthdr.len;
1010 	len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1011 	if (len < 8)
1012 		return (EMSGSIZE);
1013 
1014 	/*
1015 	 * Loop through length of segment after first fragment,
1016 	 * make new header and copy data of each part and link onto
1017 	 * chain.
1018 	 */
1019 	for (off = hlen; off < tlen; off += len) {
1020 		struct mbuf *mlast;
1021 
1022 		if ((m = m_gethdr(M_DONTWAIT, MT_HEADER)) == NULL)
1023 			return (ENOBUFS);
1024 		*mnext = m;
1025 		mnext = &m->m_nextpkt;
1026 		if ((error = m_dup_pkthdr(m, m0, M_DONTWAIT)) != 0)
1027 			return (error);
1028 		m->m_data += max_linkhdr;
1029 		mhip6 = mtod(m, struct ip6_hdr *);
1030 		*mhip6 = *mtod(m0, struct ip6_hdr *);
1031 		m->m_len = sizeof(*mhip6);
1032 		if ((error = ip6_insertfraghdr(m0, m, hlen, &ip6f)) != 0)
1033 			return (error);
1034 		ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7));
1035 		if (off + len >= tlen)
1036 			len = tlen - off;
1037 		else
1038 			ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1039 		mhip6->ip6_plen = htons((u_int16_t)(len + hlen +
1040 		    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1041 		if ((m_frgpart = m_copym(m0, off, len, M_DONTWAIT)) == NULL)
1042 			return (ENOBUFS);
1043 		for (mlast = m; mlast->m_next; mlast = mlast->m_next)
1044 			;
1045 		mlast->m_next = m_frgpart;
1046 		m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1047 		ip6f->ip6f_reserved = 0;
1048 		ip6f->ip6f_ident = id;
1049 		ip6f->ip6f_nxt = nextproto;
1050 	}
1051 
1052 	return (0);
1053 }
1054 
1055 int
1056 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1057 {
1058 	struct mbuf *m;
1059 
1060 	if (hlen > MCLBYTES)
1061 		return (ENOBUFS); /* XXX */
1062 
1063 	MGET(m, M_DONTWAIT, MT_DATA);
1064 	if (!m)
1065 		return (ENOBUFS);
1066 
1067 	if (hlen > MLEN) {
1068 		MCLGET(m, M_DONTWAIT);
1069 		if ((m->m_flags & M_EXT) == 0) {
1070 			m_free(m);
1071 			return (ENOBUFS);
1072 		}
1073 	}
1074 	m->m_len = hlen;
1075 	if (hdr)
1076 		bcopy(hdr, mtod(m, caddr_t), hlen);
1077 
1078 	*mp = m;
1079 	return (0);
1080 }
1081 
1082 /*
1083  * Insert jumbo payload option.
1084  */
1085 int
1086 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1087 {
1088 	struct mbuf *mopt;
1089 	u_int8_t *optbuf;
1090 	u_int32_t v;
1091 
1092 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1093 
1094 	/*
1095 	 * If there is no hop-by-hop options header, allocate new one.
1096 	 * If there is one but it doesn't have enough space to store the
1097 	 * jumbo payload option, allocate a cluster to store the whole options.
1098 	 * Otherwise, use it to store the options.
1099 	 */
1100 	if (exthdrs->ip6e_hbh == 0) {
1101 		MGET(mopt, M_DONTWAIT, MT_DATA);
1102 		if (mopt == 0)
1103 			return (ENOBUFS);
1104 		mopt->m_len = JUMBOOPTLEN;
1105 		optbuf = mtod(mopt, u_int8_t *);
1106 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1107 		exthdrs->ip6e_hbh = mopt;
1108 	} else {
1109 		struct ip6_hbh *hbh;
1110 
1111 		mopt = exthdrs->ip6e_hbh;
1112 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1113 			/*
1114 			 * XXX assumption:
1115 			 * - exthdrs->ip6e_hbh is not referenced from places
1116 			 *   other than exthdrs.
1117 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1118 			 */
1119 			int oldoptlen = mopt->m_len;
1120 			struct mbuf *n;
1121 
1122 			/*
1123 			 * XXX: give up if the whole (new) hbh header does
1124 			 * not fit even in an mbuf cluster.
1125 			 */
1126 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1127 				return (ENOBUFS);
1128 
1129 			/*
1130 			 * As a consequence, we must always prepare a cluster
1131 			 * at this point.
1132 			 */
1133 			MGET(n, M_DONTWAIT, MT_DATA);
1134 			if (n) {
1135 				MCLGET(n, M_DONTWAIT);
1136 				if ((n->m_flags & M_EXT) == 0) {
1137 					m_freem(n);
1138 					n = NULL;
1139 				}
1140 			}
1141 			if (!n)
1142 				return (ENOBUFS);
1143 			n->m_len = oldoptlen + JUMBOOPTLEN;
1144 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1145 			      oldoptlen);
1146 			optbuf = mtod(n, u_int8_t *) + oldoptlen;
1147 			m_freem(mopt);
1148 			mopt = exthdrs->ip6e_hbh = n;
1149 		} else {
1150 			optbuf = mtod(mopt, u_int8_t *) + mopt->m_len;
1151 			mopt->m_len += JUMBOOPTLEN;
1152 		}
1153 		optbuf[0] = IP6OPT_PADN;
1154 		optbuf[1] = 0;
1155 
1156 		/*
1157 		 * Adjust the header length according to the pad and
1158 		 * the jumbo payload option.
1159 		 */
1160 		hbh = mtod(mopt, struct ip6_hbh *);
1161 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1162 	}
1163 
1164 	/* fill in the option. */
1165 	optbuf[2] = IP6OPT_JUMBO;
1166 	optbuf[3] = 4;
1167 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1168 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1169 
1170 	/* finally, adjust the packet header length */
1171 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1172 
1173 	return (0);
1174 #undef JUMBOOPTLEN
1175 }
1176 
1177 /*
1178  * Insert fragment header and copy unfragmentable header portions.
1179  */
1180 int
1181 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1182     struct ip6_frag **frghdrp)
1183 {
1184 	struct mbuf *n, *mlast;
1185 
1186 	if (hlen > sizeof(struct ip6_hdr)) {
1187 		n = m_copym(m0, sizeof(struct ip6_hdr),
1188 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1189 		if (n == 0)
1190 			return (ENOBUFS);
1191 		m->m_next = n;
1192 	} else
1193 		n = m;
1194 
1195 	/* Search for the last mbuf of unfragmentable part. */
1196 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1197 		;
1198 
1199 	if ((mlast->m_flags & M_EXT) == 0 &&
1200 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1201 		/* use the trailing space of the last mbuf for the fragment hdr */
1202 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1203 		    mlast->m_len);
1204 		mlast->m_len += sizeof(struct ip6_frag);
1205 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1206 	} else {
1207 		/* allocate a new mbuf for the fragment header */
1208 		struct mbuf *mfrg;
1209 
1210 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1211 		if (mfrg == 0)
1212 			return (ENOBUFS);
1213 		mfrg->m_len = sizeof(struct ip6_frag);
1214 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1215 		mlast->m_next = mfrg;
1216 	}
1217 
1218 	return (0);
1219 }
1220 
1221 int
1222 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1223     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup, int *alwaysfragp)
1224 {
1225 	u_int32_t mtu = 0;
1226 	int alwaysfrag = 0;
1227 	int error = 0;
1228 
1229 	if (ro_pmtu != ro) {
1230 		/* The first hop and the final destination may differ. */
1231 		struct sockaddr_in6 *sa6_dst =
1232 		    (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1233 		if (ro_pmtu->ro_rt &&
1234 		    ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1235 		     !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1236 			RTFREE(ro_pmtu->ro_rt);
1237 			ro_pmtu->ro_rt = (struct rtentry *)NULL;
1238 		}
1239 		if (ro_pmtu->ro_rt == 0) {
1240 			bzero(ro_pmtu, sizeof(*ro_pmtu));
1241 			sa6_dst->sin6_family = AF_INET6;
1242 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1243 			sa6_dst->sin6_addr = *dst;
1244 
1245 			rtalloc((struct route *)ro_pmtu);
1246 		}
1247 	}
1248 	if (ro_pmtu->ro_rt) {
1249 		u_int32_t ifmtu;
1250 
1251 		if (ifp == NULL)
1252 			ifp = ro_pmtu->ro_rt->rt_ifp;
1253 		ifmtu = IN6_LINKMTU(ifp);
1254 		mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1255 		if (mtu == 0)
1256 			mtu = ifmtu;
1257 		else if (mtu < IPV6_MMTU) {
1258 			/*
1259 			 * RFC2460 section 5, last paragraph:
1260 			 * if we record ICMPv6 too big message with
1261 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1262 			 * or smaller, with fragment header attached.
1263 			 * (fragment header is needed regardless from the
1264 			 * packet size, for translators to identify packets)
1265 			 */
1266 			alwaysfrag = 1;
1267 			mtu = IPV6_MMTU;
1268 		} else if (mtu > ifmtu) {
1269 			/*
1270 			 * The MTU on the route is larger than the MTU on
1271 			 * the interface!  This shouldn't happen, unless the
1272 			 * MTU of the interface has been changed after the
1273 			 * interface was brought up.  Change the MTU in the
1274 			 * route to match the interface MTU (as long as the
1275 			 * field isn't locked).
1276 			 */
1277 			mtu = ifmtu;
1278 			if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU))
1279 				ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1280 		}
1281 	} else if (ifp) {
1282 		mtu = IN6_LINKMTU(ifp);
1283 	} else
1284 		error = EHOSTUNREACH; /* XXX */
1285 
1286 	*mtup = mtu;
1287 	if (alwaysfragp)
1288 		*alwaysfragp = alwaysfrag;
1289 	return (error);
1290 }
1291 
1292 /*
1293  * IP6 socket option processing.
1294  */
1295 int
1296 ip6_ctloutput(int op, struct socket *so, int level, int optname,
1297     struct mbuf **mp)
1298 {
1299 	int privileged, optdatalen, uproto;
1300 	void *optdata;
1301 	struct inpcb *inp = sotoinpcb(so);
1302 	struct mbuf *m = *mp;
1303 	int error, optval;
1304 #ifdef IPSEC
1305 	struct proc *p = curproc; /* XXX */
1306 	struct tdb *tdb;
1307 	struct tdb_ident *tdbip, tdbi;
1308 	int s;
1309 #endif
1310 
1311 	error = optval = 0;
1312 
1313 	privileged = (inp->inp_socket->so_state & SS_PRIV);
1314 	uproto = (int)so->so_proto->pr_protocol;
1315 
1316 	if (level == IPPROTO_IPV6) {
1317 		switch (op) {
1318 		case PRCO_SETOPT:
1319 			switch (optname) {
1320 			case IPV6_2292PKTOPTIONS:
1321 			{
1322 				error = ip6_pcbopts(&inp->inp_outputopts6,
1323 						    m, so);
1324 				break;
1325 			}
1326 
1327 			/*
1328 			 * Use of some Hop-by-Hop options or some
1329 			 * Destination options, might require special
1330 			 * privilege.  That is, normal applications
1331 			 * (without special privilege) might be forbidden
1332 			 * from setting certain options in outgoing packets,
1333 			 * and might never see certain options in received
1334 			 * packets. [RFC 2292 Section 6]
1335 			 * KAME specific note:
1336 			 *  KAME prevents non-privileged users from sending or
1337 			 *  receiving ANY hbh/dst options in order to avoid
1338 			 *  overhead of parsing options in the kernel.
1339 			 */
1340 			case IPV6_RECVHOPOPTS:
1341 			case IPV6_RECVDSTOPTS:
1342 			case IPV6_RECVRTHDRDSTOPTS:
1343 				if (!privileged) {
1344 					error = EPERM;
1345 					break;
1346 				}
1347 				/* FALLTHROUGH */
1348 			case IPV6_UNICAST_HOPS:
1349 			case IPV6_HOPLIMIT:
1350 			case IPV6_FAITH:
1351 
1352 			case IPV6_RECVPKTINFO:
1353 			case IPV6_RECVHOPLIMIT:
1354 			case IPV6_RECVRTHDR:
1355 			case IPV6_RECVPATHMTU:
1356 			case IPV6_RECVTCLASS:
1357 			case IPV6_V6ONLY:
1358 			case IPV6_AUTOFLOWLABEL:
1359 				if (m == NULL || m->m_len != sizeof(int)) {
1360 					error = EINVAL;
1361 					break;
1362 				}
1363 				optval = *mtod(m, int *);
1364 				switch (optname) {
1365 
1366 				case IPV6_UNICAST_HOPS:
1367 					if (optval < -1 || optval >= 256)
1368 						error = EINVAL;
1369 					else {
1370 						/* -1 = kernel default */
1371 						inp->inp_hops = optval;
1372 					}
1373 					break;
1374 #define OPTSET(bit) \
1375 do { \
1376 	if (optval) \
1377 		inp->inp_flags |= (bit); \
1378 	else \
1379 		inp->inp_flags &= ~(bit); \
1380 } while (/*CONSTCOND*/ 0)
1381 #define OPTSET2292(bit) \
1382 do { \
1383 	inp->inp_flags |= IN6P_RFC2292; \
1384 	if (optval) \
1385 		inp->inp_flags |= (bit); \
1386 	else \
1387 		inp->inp_flags &= ~(bit); \
1388 } while (/*CONSTCOND*/ 0)
1389 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1390 
1391 				case IPV6_RECVPKTINFO:
1392 					/* cannot mix with RFC2292 */
1393 					if (OPTBIT(IN6P_RFC2292)) {
1394 						error = EINVAL;
1395 						break;
1396 					}
1397 					OPTSET(IN6P_PKTINFO);
1398 					break;
1399 
1400 				case IPV6_HOPLIMIT:
1401 				{
1402 					struct ip6_pktopts **optp;
1403 
1404 					/* cannot mix with RFC2292 */
1405 					if (OPTBIT(IN6P_RFC2292)) {
1406 						error = EINVAL;
1407 						break;
1408 					}
1409 					optp = &inp->inp_outputopts6;
1410 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1411 							   (u_char *)&optval,
1412 							   sizeof(optval),
1413 							   optp,
1414 							   privileged, uproto);
1415 					break;
1416 				}
1417 
1418 				case IPV6_RECVHOPLIMIT:
1419 					/* cannot mix with RFC2292 */
1420 					if (OPTBIT(IN6P_RFC2292)) {
1421 						error = EINVAL;
1422 						break;
1423 					}
1424 					OPTSET(IN6P_HOPLIMIT);
1425 					break;
1426 
1427 				case IPV6_RECVHOPOPTS:
1428 					/* cannot mix with RFC2292 */
1429 					if (OPTBIT(IN6P_RFC2292)) {
1430 						error = EINVAL;
1431 						break;
1432 					}
1433 					OPTSET(IN6P_HOPOPTS);
1434 					break;
1435 
1436 				case IPV6_RECVDSTOPTS:
1437 					/* cannot mix with RFC2292 */
1438 					if (OPTBIT(IN6P_RFC2292)) {
1439 						error = EINVAL;
1440 						break;
1441 					}
1442 					OPTSET(IN6P_DSTOPTS);
1443 					break;
1444 
1445 				case IPV6_RECVRTHDRDSTOPTS:
1446 					/* cannot mix with RFC2292 */
1447 					if (OPTBIT(IN6P_RFC2292)) {
1448 						error = EINVAL;
1449 						break;
1450 					}
1451 					OPTSET(IN6P_RTHDRDSTOPTS);
1452 					break;
1453 
1454 				case IPV6_RECVRTHDR:
1455 					/* cannot mix with RFC2292 */
1456 					if (OPTBIT(IN6P_RFC2292)) {
1457 						error = EINVAL;
1458 						break;
1459 					}
1460 					OPTSET(IN6P_RTHDR);
1461 					break;
1462 
1463 				case IPV6_FAITH:
1464 					OPTSET(IN6P_FAITH);
1465 					break;
1466 
1467 				case IPV6_RECVPATHMTU:
1468 					/*
1469 					 * We ignore this option for TCP
1470 					 * sockets.
1471 					 * (RFC3542 leaves this case
1472 					 * unspecified.)
1473 					 */
1474 					if (uproto != IPPROTO_TCP)
1475 						OPTSET(IN6P_MTU);
1476 					break;
1477 
1478 				case IPV6_V6ONLY:
1479 					/*
1480 					 * make setsockopt(IPV6_V6ONLY)
1481 					 * available only prior to bind(2).
1482 					 * see ipng mailing list, Jun 22 2001.
1483 					 */
1484 					if (inp->inp_lport ||
1485 					    !IN6_IS_ADDR_UNSPECIFIED(&inp->inp_laddr6)) {
1486 						error = EINVAL;
1487 						break;
1488 					}
1489 					if ((ip6_v6only && optval) ||
1490 					    (!ip6_v6only && !optval))
1491 						error = 0;
1492 					else
1493 						error = EINVAL;
1494 					break;
1495 				case IPV6_RECVTCLASS:
1496 					/* cannot mix with RFC2292 XXX */
1497 					if (OPTBIT(IN6P_RFC2292)) {
1498 						error = EINVAL;
1499 						break;
1500 					}
1501 					OPTSET(IN6P_TCLASS);
1502 					break;
1503 				case IPV6_AUTOFLOWLABEL:
1504 					OPTSET(IN6P_AUTOFLOWLABEL);
1505 					break;
1506 
1507 				}
1508 				break;
1509 
1510 			case IPV6_TCLASS:
1511 			case IPV6_DONTFRAG:
1512 			case IPV6_USE_MIN_MTU:
1513 				if (m == NULL || m->m_len != sizeof(optval)) {
1514 					error = EINVAL;
1515 					break;
1516 				}
1517 				optval = *mtod(m, int *);
1518 				{
1519 					struct ip6_pktopts **optp;
1520 					optp = &inp->inp_outputopts6;
1521 					error = ip6_pcbopt(optname,
1522 							   (u_char *)&optval,
1523 							   sizeof(optval),
1524 							   optp,
1525 							   privileged, uproto);
1526 					break;
1527 				}
1528 
1529 			case IPV6_2292PKTINFO:
1530 			case IPV6_2292HOPLIMIT:
1531 			case IPV6_2292HOPOPTS:
1532 			case IPV6_2292DSTOPTS:
1533 			case IPV6_2292RTHDR:
1534 				/* RFC 2292 */
1535 				if (m == NULL || m->m_len != sizeof(int)) {
1536 					error = EINVAL;
1537 					break;
1538 				}
1539 				optval = *mtod(m, int *);
1540 				switch (optname) {
1541 				case IPV6_2292PKTINFO:
1542 					OPTSET2292(IN6P_PKTINFO);
1543 					break;
1544 				case IPV6_2292HOPLIMIT:
1545 					OPTSET2292(IN6P_HOPLIMIT);
1546 					break;
1547 				case IPV6_2292HOPOPTS:
1548 					/*
1549 					 * Check super-user privilege.
1550 					 * See comments for IPV6_RECVHOPOPTS.
1551 					 */
1552 					if (!privileged)
1553 						return (EPERM);
1554 					OPTSET2292(IN6P_HOPOPTS);
1555 					break;
1556 				case IPV6_2292DSTOPTS:
1557 					if (!privileged)
1558 						return (EPERM);
1559 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1560 					break;
1561 				case IPV6_2292RTHDR:
1562 					OPTSET2292(IN6P_RTHDR);
1563 					break;
1564 				}
1565 				break;
1566 			case IPV6_PKTINFO:
1567 			case IPV6_HOPOPTS:
1568 			case IPV6_RTHDR:
1569 			case IPV6_DSTOPTS:
1570 			case IPV6_RTHDRDSTOPTS:
1571 			case IPV6_NEXTHOP:
1572 			{
1573 				/* new advanced API (RFC3542) */
1574 				u_char *optbuf;
1575 				int optbuflen;
1576 				struct ip6_pktopts **optp;
1577 
1578 				/* cannot mix with RFC2292 */
1579 				if (OPTBIT(IN6P_RFC2292)) {
1580 					error = EINVAL;
1581 					break;
1582 				}
1583 
1584 				if (m && m->m_next) {
1585 					error = EINVAL;	/* XXX */
1586 					break;
1587 				}
1588 				if (m) {
1589 					optbuf = mtod(m, u_char *);
1590 					optbuflen = m->m_len;
1591 				} else {
1592 					optbuf = NULL;
1593 					optbuflen = 0;
1594 				}
1595 				optp = &inp->inp_outputopts6;
1596 				error = ip6_pcbopt(optname,
1597 						   optbuf, optbuflen,
1598 						   optp, privileged, uproto);
1599 				break;
1600 			}
1601 #undef OPTSET
1602 
1603 			case IPV6_MULTICAST_IF:
1604 			case IPV6_MULTICAST_HOPS:
1605 			case IPV6_MULTICAST_LOOP:
1606 			case IPV6_JOIN_GROUP:
1607 			case IPV6_LEAVE_GROUP:
1608 				error =	ip6_setmoptions(optname,
1609 							&inp->inp_moptions6,
1610 							m);
1611 				break;
1612 
1613 			case IPV6_PORTRANGE:
1614 				if (m == NULL || m->m_len != sizeof(int)) {
1615 					error = EINVAL;
1616 					break;
1617 				}
1618 				optval = *mtod(m, int *);
1619 
1620 				switch (optval) {
1621 				case IPV6_PORTRANGE_DEFAULT:
1622 					inp->inp_flags &= ~(IN6P_LOWPORT);
1623 					inp->inp_flags &= ~(IN6P_HIGHPORT);
1624 					break;
1625 
1626 				case IPV6_PORTRANGE_HIGH:
1627 					inp->inp_flags &= ~(IN6P_LOWPORT);
1628 					inp->inp_flags |= IN6P_HIGHPORT;
1629 					break;
1630 
1631 				case IPV6_PORTRANGE_LOW:
1632 					inp->inp_flags &= ~(IN6P_HIGHPORT);
1633 					inp->inp_flags |= IN6P_LOWPORT;
1634 					break;
1635 
1636 				default:
1637 					error = EINVAL;
1638 					break;
1639 				}
1640 				break;
1641 
1642 			case IPSEC6_OUTSA:
1643 #ifndef IPSEC
1644 				error = EINVAL;
1645 #else
1646 				if (m == NULL ||
1647 				    m->m_len != sizeof(struct tdb_ident)) {
1648 					error = EINVAL;
1649 					break;
1650 				}
1651 				tdbip = mtod(m, struct tdb_ident *);
1652 				s = spltdb();
1653 				tdb = gettdb(tdbip->rdomain, tdbip->spi,
1654 				    &tdbip->dst, tdbip->proto);
1655 				if (tdb == NULL)
1656 					error = ESRCH;
1657 				else
1658 					tdb_add_inp(tdb, inp, 0);
1659 				splx(s);
1660 #endif
1661 				break;
1662 
1663 			case IPV6_AUTH_LEVEL:
1664 			case IPV6_ESP_TRANS_LEVEL:
1665 			case IPV6_ESP_NETWORK_LEVEL:
1666 			case IPV6_IPCOMP_LEVEL:
1667 #ifndef IPSEC
1668 				error = EINVAL;
1669 #else
1670 				if (m == 0 || m->m_len != sizeof(int)) {
1671 					error = EINVAL;
1672 					break;
1673 				}
1674 				optval = *mtod(m, int *);
1675 
1676 				if (optval < IPSEC_LEVEL_BYPASS ||
1677 				    optval > IPSEC_LEVEL_UNIQUE) {
1678 					error = EINVAL;
1679 					break;
1680 				}
1681 
1682 				switch (optname) {
1683 				case IPV6_AUTH_LEVEL:
1684 				        if (optval < ipsec_auth_default_level &&
1685 					    suser(p, 0)) {
1686 						error = EACCES;
1687 						break;
1688 					}
1689 					inp->inp_seclevel[SL_AUTH] = optval;
1690 					break;
1691 
1692 				case IPV6_ESP_TRANS_LEVEL:
1693 				        if (optval < ipsec_esp_trans_default_level &&
1694 					    suser(p, 0)) {
1695 						error = EACCES;
1696 						break;
1697 					}
1698 					inp->inp_seclevel[SL_ESP_TRANS] = optval;
1699 					break;
1700 
1701 				case IPV6_ESP_NETWORK_LEVEL:
1702 				        if (optval < ipsec_esp_network_default_level &&
1703 					    suser(p, 0)) {
1704 						error = EACCES;
1705 						break;
1706 					}
1707 					inp->inp_seclevel[SL_ESP_NETWORK] = optval;
1708 					break;
1709 
1710 				case IPV6_IPCOMP_LEVEL:
1711 				        if (optval < ipsec_ipcomp_default_level &&
1712 					    suser(p, 0)) {
1713 						error = EACCES;
1714 						break;
1715 					}
1716 					inp->inp_seclevel[SL_IPCOMP] = optval;
1717 					break;
1718 				}
1719 				if (!error)
1720 					inp->inp_secrequire = get_sa_require(inp);
1721 #endif
1722 				break;
1723 			case IPV6_PIPEX:
1724 				if (m != NULL && m->m_len == sizeof(int))
1725 					inp->inp_pipex = *mtod(m, int *);
1726 				else
1727 					error = EINVAL;
1728 				break;
1729 
1730 			default:
1731 				error = ENOPROTOOPT;
1732 				break;
1733 			}
1734 			if (m)
1735 				(void)m_free(m);
1736 			break;
1737 
1738 		case PRCO_GETOPT:
1739 			switch (optname) {
1740 
1741 			case IPV6_2292PKTOPTIONS:
1742 				/*
1743 				 * RFC3542 (effectively) deprecated the
1744 				 * semantics of the 2292-style pktoptions.
1745 				 * Since it was not reliable in nature (i.e.,
1746 				 * applications had to expect the lack of some
1747 				 * information after all), it would make sense
1748 				 * to simplify this part by always returning
1749 				 * empty data.
1750 				 */
1751 				*mp = m_get(M_WAIT, MT_SOOPTS);
1752 				(*mp)->m_len = 0;
1753 				break;
1754 
1755 			case IPV6_RECVHOPOPTS:
1756 			case IPV6_RECVDSTOPTS:
1757 			case IPV6_RECVRTHDRDSTOPTS:
1758 			case IPV6_UNICAST_HOPS:
1759 			case IPV6_RECVPKTINFO:
1760 			case IPV6_RECVHOPLIMIT:
1761 			case IPV6_RECVRTHDR:
1762 			case IPV6_RECVPATHMTU:
1763 
1764 			case IPV6_FAITH:
1765 			case IPV6_V6ONLY:
1766 			case IPV6_PORTRANGE:
1767 			case IPV6_RECVTCLASS:
1768 			case IPV6_AUTOFLOWLABEL:
1769 				switch (optname) {
1770 
1771 				case IPV6_RECVHOPOPTS:
1772 					optval = OPTBIT(IN6P_HOPOPTS);
1773 					break;
1774 
1775 				case IPV6_RECVDSTOPTS:
1776 					optval = OPTBIT(IN6P_DSTOPTS);
1777 					break;
1778 
1779 				case IPV6_RECVRTHDRDSTOPTS:
1780 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1781 					break;
1782 
1783 				case IPV6_UNICAST_HOPS:
1784 					optval = inp->inp_hops;
1785 					break;
1786 
1787 				case IPV6_RECVPKTINFO:
1788 					optval = OPTBIT(IN6P_PKTINFO);
1789 					break;
1790 
1791 				case IPV6_RECVHOPLIMIT:
1792 					optval = OPTBIT(IN6P_HOPLIMIT);
1793 					break;
1794 
1795 				case IPV6_RECVRTHDR:
1796 					optval = OPTBIT(IN6P_RTHDR);
1797 					break;
1798 
1799 				case IPV6_RECVPATHMTU:
1800 					optval = OPTBIT(IN6P_MTU);
1801 					break;
1802 
1803 				case IPV6_FAITH:
1804 					optval = OPTBIT(IN6P_FAITH);
1805 					break;
1806 
1807 				case IPV6_V6ONLY:
1808 					optval = (ip6_v6only != 0); /* XXX */
1809 					break;
1810 
1811 				case IPV6_PORTRANGE:
1812 				    {
1813 					int flags;
1814 					flags = inp->inp_flags;
1815 					if (flags & IN6P_HIGHPORT)
1816 						optval = IPV6_PORTRANGE_HIGH;
1817 					else if (flags & IN6P_LOWPORT)
1818 						optval = IPV6_PORTRANGE_LOW;
1819 					else
1820 						optval = 0;
1821 					break;
1822 				    }
1823 				case IPV6_RECVTCLASS:
1824 					optval = OPTBIT(IN6P_TCLASS);
1825 					break;
1826 
1827 				case IPV6_AUTOFLOWLABEL:
1828 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1829 					break;
1830 				}
1831 				if (error)
1832 					break;
1833 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1834 				m->m_len = sizeof(int);
1835 				*mtod(m, int *) = optval;
1836 				break;
1837 
1838 			case IPV6_PATHMTU:
1839 			{
1840 				u_long pmtu = 0;
1841 				struct ip6_mtuinfo mtuinfo;
1842 				struct route_in6 *ro = (struct route_in6 *)&inp->inp_route6;
1843 
1844 				if (!(so->so_state & SS_ISCONNECTED))
1845 					return (ENOTCONN);
1846 				/*
1847 				 * XXX: we dot not consider the case of source
1848 				 * routing, or optional information to specify
1849 				 * the outgoing interface.
1850 				 */
1851 				error = ip6_getpmtu(ro, NULL, NULL,
1852 				    &inp->inp_faddr6, &pmtu, NULL);
1853 				if (error)
1854 					break;
1855 				if (pmtu > IPV6_MAXPACKET)
1856 					pmtu = IPV6_MAXPACKET;
1857 
1858 				bzero(&mtuinfo, sizeof(mtuinfo));
1859 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1860 				optdata = (void *)&mtuinfo;
1861 				optdatalen = sizeof(mtuinfo);
1862 				if (optdatalen > MCLBYTES)
1863 					return (EMSGSIZE); /* XXX */
1864 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1865 				if (optdatalen > MLEN)
1866 					MCLGET(m, M_WAIT);
1867 				m->m_len = optdatalen;
1868 				bcopy(optdata, mtod(m, void *), optdatalen);
1869 				break;
1870 			}
1871 
1872 			case IPV6_2292PKTINFO:
1873 			case IPV6_2292HOPLIMIT:
1874 			case IPV6_2292HOPOPTS:
1875 			case IPV6_2292RTHDR:
1876 			case IPV6_2292DSTOPTS:
1877 				switch (optname) {
1878 				case IPV6_2292PKTINFO:
1879 					optval = OPTBIT(IN6P_PKTINFO);
1880 					break;
1881 				case IPV6_2292HOPLIMIT:
1882 					optval = OPTBIT(IN6P_HOPLIMIT);
1883 					break;
1884 				case IPV6_2292HOPOPTS:
1885 					optval = OPTBIT(IN6P_HOPOPTS);
1886 					break;
1887 				case IPV6_2292RTHDR:
1888 					optval = OPTBIT(IN6P_RTHDR);
1889 					break;
1890 				case IPV6_2292DSTOPTS:
1891 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1892 					break;
1893 				}
1894 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1895 				m->m_len = sizeof(int);
1896 				*mtod(m, int *) = optval;
1897 				break;
1898 			case IPV6_PKTINFO:
1899 			case IPV6_HOPOPTS:
1900 			case IPV6_RTHDR:
1901 			case IPV6_DSTOPTS:
1902 			case IPV6_RTHDRDSTOPTS:
1903 			case IPV6_NEXTHOP:
1904 			case IPV6_TCLASS:
1905 			case IPV6_DONTFRAG:
1906 			case IPV6_USE_MIN_MTU:
1907 				error = ip6_getpcbopt(inp->inp_outputopts6,
1908 				    optname, mp);
1909 				break;
1910 
1911 			case IPV6_MULTICAST_IF:
1912 			case IPV6_MULTICAST_HOPS:
1913 			case IPV6_MULTICAST_LOOP:
1914 			case IPV6_JOIN_GROUP:
1915 			case IPV6_LEAVE_GROUP:
1916 				error = ip6_getmoptions(optname,
1917 				    inp->inp_moptions6, mp);
1918 				break;
1919 
1920 			case IPSEC6_OUTSA:
1921 #ifndef IPSEC
1922 				error = EINVAL;
1923 #else
1924 				s = spltdb();
1925 				if (inp->inp_tdb_out == NULL) {
1926 					error = ENOENT;
1927 				} else {
1928 					tdbi.spi = inp->inp_tdb_out->tdb_spi;
1929 					tdbi.dst = inp->inp_tdb_out->tdb_dst;
1930 					tdbi.proto = inp->inp_tdb_out->tdb_sproto;
1931 					tdbi.rdomain =
1932 					    inp->inp_tdb_out->tdb_rdomain;
1933 					*mp = m = m_get(M_WAIT, MT_SOOPTS);
1934 					m->m_len = sizeof(tdbi);
1935 					bcopy((caddr_t)&tdbi, mtod(m, caddr_t),
1936 					    (unsigned)m->m_len);
1937 				}
1938 				splx(s);
1939 #endif
1940 				break;
1941 
1942 			case IPV6_AUTH_LEVEL:
1943 			case IPV6_ESP_TRANS_LEVEL:
1944 			case IPV6_ESP_NETWORK_LEVEL:
1945 			case IPV6_IPCOMP_LEVEL:
1946 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1947 #ifndef IPSEC
1948 				m->m_len = sizeof(int);
1949 				*mtod(m, int *) = IPSEC_LEVEL_NONE;
1950 #else
1951 				m->m_len = sizeof(int);
1952 				switch (optname) {
1953 				case IPV6_AUTH_LEVEL:
1954 					optval = inp->inp_seclevel[SL_AUTH];
1955 					break;
1956 
1957 				case IPV6_ESP_TRANS_LEVEL:
1958 					optval =
1959 					    inp->inp_seclevel[SL_ESP_TRANS];
1960 					break;
1961 
1962 				case IPV6_ESP_NETWORK_LEVEL:
1963 					optval =
1964 					    inp->inp_seclevel[SL_ESP_NETWORK];
1965 					break;
1966 
1967 				case IPV6_IPCOMP_LEVEL:
1968 					optval = inp->inp_seclevel[SL_IPCOMP];
1969 					break;
1970 				}
1971 				*mtod(m, int *) = optval;
1972 #endif
1973 				break;
1974 			case IPV6_PIPEX:
1975 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1976 				m->m_len = sizeof(int);
1977 				*mtod(m, int *) = optval;
1978 				break;
1979 
1980 			default:
1981 				error = ENOPROTOOPT;
1982 				break;
1983 			}
1984 			break;
1985 		}
1986 	} else {
1987 		error = EINVAL;
1988 		if (op == PRCO_SETOPT && *mp)
1989 			(void)m_free(*mp);
1990 	}
1991 	return (error);
1992 }
1993 
1994 int
1995 ip6_raw_ctloutput(int op, struct socket *so, int level, int optname,
1996     struct mbuf **mp)
1997 {
1998 	int error = 0, optval;
1999 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2000 	struct inpcb *inp = sotoinpcb(so);
2001 	struct mbuf *m = *mp;
2002 
2003 	if (level != IPPROTO_IPV6) {
2004 		if (op == PRCO_SETOPT && *mp)
2005 			(void)m_free(*mp);
2006 		return (EINVAL);
2007 	}
2008 
2009 	switch (optname) {
2010 	case IPV6_CHECKSUM:
2011 		/*
2012 		 * For ICMPv6 sockets, no modification allowed for checksum
2013 		 * offset, permit "no change" values to help existing apps.
2014 		 *
2015 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2016 		 * for an ICMPv6 socket will fail."
2017 		 * The current behavior does not meet RFC3542.
2018 		 */
2019 		switch (op) {
2020 		case PRCO_SETOPT:
2021 			if (m == NULL || m->m_len != sizeof(int)) {
2022 				error = EINVAL;
2023 				break;
2024 			}
2025 			optval = *mtod(m, int *);
2026 			if ((optval % 2) != 0) {
2027 				/* the API assumes even offset values */
2028 				error = EINVAL;
2029 			} else if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) {
2030 				if (optval != icmp6off)
2031 					error = EINVAL;
2032 			} else
2033 				inp->in6p_cksum = optval;
2034 			break;
2035 
2036 		case PRCO_GETOPT:
2037 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2038 				optval = icmp6off;
2039 			else
2040 				optval = inp->in6p_cksum;
2041 
2042 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
2043 			m->m_len = sizeof(int);
2044 			*mtod(m, int *) = optval;
2045 			break;
2046 
2047 		default:
2048 			error = EINVAL;
2049 			break;
2050 		}
2051 		break;
2052 
2053 	default:
2054 		error = ENOPROTOOPT;
2055 		break;
2056 	}
2057 
2058 	if (op == PRCO_SETOPT && m)
2059 		(void)m_free(m);
2060 
2061 	return (error);
2062 }
2063 
2064 /*
2065  * Set up IP6 options in pcb for insertion in output packets.
2066  * Store in mbuf with pointer in pcbopt, adding pseudo-option
2067  * with destination address if source routed.
2068  */
2069 int
2070 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, struct socket *so)
2071 {
2072 	struct ip6_pktopts *opt = *pktopt;
2073 	int error = 0;
2074 	struct proc *p = curproc;	/* XXX */
2075 	int priv = 0;
2076 
2077 	/* turn off any old options. */
2078 	if (opt)
2079 		ip6_clearpktopts(opt, -1);
2080 	else
2081 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2082 	*pktopt = 0;
2083 
2084 	if (!m || m->m_len == 0) {
2085 		/*
2086 		 * Only turning off any previous options, regardless of
2087 		 * whether the opt is just created or given.
2088 		 */
2089 		free(opt, M_IP6OPT);
2090 		return (0);
2091 	}
2092 
2093 	/*  set options specified by user. */
2094 	if (p && !suser(p, 0))
2095 		priv = 1;
2096 	if ((error = ip6_setpktopts(m, opt, NULL, priv,
2097 	    so->so_proto->pr_protocol)) != 0) {
2098 		ip6_clearpktopts(opt, -1);	/* XXX discard all options */
2099 		free(opt, M_IP6OPT);
2100 		return (error);
2101 	}
2102 	*pktopt = opt;
2103 	return (0);
2104 }
2105 
2106 /*
2107  * initialize ip6_pktopts.  beware that there are non-zero default values in
2108  * the struct.
2109  */
2110 void
2111 ip6_initpktopts(struct ip6_pktopts *opt)
2112 {
2113 
2114 	bzero(opt, sizeof(*opt));
2115 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2116 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2117 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2118 }
2119 
2120 #define sin6tosa(sin6)	((struct sockaddr *)(sin6)) /* XXX */
2121 int
2122 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2123     int priv, int uproto)
2124 {
2125 	struct ip6_pktopts *opt;
2126 
2127 	if (*pktopt == NULL) {
2128 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2129 		    M_WAITOK);
2130 		ip6_initpktopts(*pktopt);
2131 	}
2132 	opt = *pktopt;
2133 
2134 	return (ip6_setpktopt(optname, buf, len, opt, priv, 1, 0, uproto));
2135 }
2136 
2137 int
2138 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct mbuf **mp)
2139 {
2140 	void *optdata = NULL;
2141 	int optdatalen = 0;
2142 	struct ip6_ext *ip6e;
2143 	int error = 0;
2144 	struct in6_pktinfo null_pktinfo;
2145 	int deftclass = 0, on;
2146 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2147 	struct mbuf *m;
2148 
2149 	switch (optname) {
2150 	case IPV6_PKTINFO:
2151 		if (pktopt && pktopt->ip6po_pktinfo)
2152 			optdata = (void *)pktopt->ip6po_pktinfo;
2153 		else {
2154 			/* XXX: we don't have to do this every time... */
2155 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2156 			optdata = (void *)&null_pktinfo;
2157 		}
2158 		optdatalen = sizeof(struct in6_pktinfo);
2159 		break;
2160 	case IPV6_TCLASS:
2161 		if (pktopt && pktopt->ip6po_tclass >= 0)
2162 			optdata = (void *)&pktopt->ip6po_tclass;
2163 		else
2164 			optdata = (void *)&deftclass;
2165 		optdatalen = sizeof(int);
2166 		break;
2167 	case IPV6_HOPOPTS:
2168 		if (pktopt && pktopt->ip6po_hbh) {
2169 			optdata = (void *)pktopt->ip6po_hbh;
2170 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2171 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2172 		}
2173 		break;
2174 	case IPV6_RTHDR:
2175 		if (pktopt && pktopt->ip6po_rthdr) {
2176 			optdata = (void *)pktopt->ip6po_rthdr;
2177 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2178 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2179 		}
2180 		break;
2181 	case IPV6_RTHDRDSTOPTS:
2182 		if (pktopt && pktopt->ip6po_dest1) {
2183 			optdata = (void *)pktopt->ip6po_dest1;
2184 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2185 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2186 		}
2187 		break;
2188 	case IPV6_DSTOPTS:
2189 		if (pktopt && pktopt->ip6po_dest2) {
2190 			optdata = (void *)pktopt->ip6po_dest2;
2191 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2192 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2193 		}
2194 		break;
2195 	case IPV6_NEXTHOP:
2196 		if (pktopt && pktopt->ip6po_nexthop) {
2197 			optdata = (void *)pktopt->ip6po_nexthop;
2198 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2199 		}
2200 		break;
2201 	case IPV6_USE_MIN_MTU:
2202 		if (pktopt)
2203 			optdata = (void *)&pktopt->ip6po_minmtu;
2204 		else
2205 			optdata = (void *)&defminmtu;
2206 		optdatalen = sizeof(int);
2207 		break;
2208 	case IPV6_DONTFRAG:
2209 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2210 			on = 1;
2211 		else
2212 			on = 0;
2213 		optdata = (void *)&on;
2214 		optdatalen = sizeof(on);
2215 		break;
2216 	default:		/* should not happen */
2217 #ifdef DIAGNOSTIC
2218 		panic("ip6_getpcbopt: unexpected option");
2219 #endif
2220 		return (ENOPROTOOPT);
2221 	}
2222 
2223 	if (optdatalen > MCLBYTES)
2224 		return (EMSGSIZE); /* XXX */
2225 	*mp = m = m_get(M_WAIT, MT_SOOPTS);
2226 	if (optdatalen > MLEN)
2227 		MCLGET(m, M_WAIT);
2228 	m->m_len = optdatalen;
2229 	if (optdatalen)
2230 		bcopy(optdata, mtod(m, void *), optdatalen);
2231 
2232 	return (error);
2233 }
2234 
2235 void
2236 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2237 {
2238 	if (optname == -1 || optname == IPV6_PKTINFO) {
2239 		if (pktopt->ip6po_pktinfo)
2240 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2241 		pktopt->ip6po_pktinfo = NULL;
2242 	}
2243 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2244 		pktopt->ip6po_hlim = -1;
2245 	if (optname == -1 || optname == IPV6_TCLASS)
2246 		pktopt->ip6po_tclass = -1;
2247 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2248 		if (pktopt->ip6po_nextroute.ro_rt) {
2249 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2250 			pktopt->ip6po_nextroute.ro_rt = NULL;
2251 		}
2252 		if (pktopt->ip6po_nexthop)
2253 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2254 		pktopt->ip6po_nexthop = NULL;
2255 	}
2256 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2257 		if (pktopt->ip6po_hbh)
2258 			free(pktopt->ip6po_hbh, M_IP6OPT);
2259 		pktopt->ip6po_hbh = NULL;
2260 	}
2261 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2262 		if (pktopt->ip6po_dest1)
2263 			free(pktopt->ip6po_dest1, M_IP6OPT);
2264 		pktopt->ip6po_dest1 = NULL;
2265 	}
2266 	if (optname == -1 || optname == IPV6_RTHDR) {
2267 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2268 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2269 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2270 		if (pktopt->ip6po_route.ro_rt) {
2271 			RTFREE(pktopt->ip6po_route.ro_rt);
2272 			pktopt->ip6po_route.ro_rt = NULL;
2273 		}
2274 	}
2275 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2276 		if (pktopt->ip6po_dest2)
2277 			free(pktopt->ip6po_dest2, M_IP6OPT);
2278 		pktopt->ip6po_dest2 = NULL;
2279 	}
2280 }
2281 
2282 #define PKTOPT_EXTHDRCPY(type) \
2283 do {\
2284 	if (src->type) {\
2285 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2286 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2287 		if (dst->type == NULL && canwait == M_NOWAIT)\
2288 			goto bad;\
2289 		bcopy(src->type, dst->type, hlen);\
2290 	}\
2291 } while (/*CONSTCOND*/ 0)
2292 
2293 int
2294 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2295 {
2296 	dst->ip6po_hlim = src->ip6po_hlim;
2297 	dst->ip6po_tclass = src->ip6po_tclass;
2298 	dst->ip6po_flags = src->ip6po_flags;
2299 	if (src->ip6po_pktinfo) {
2300 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2301 		    M_IP6OPT, canwait);
2302 		if (dst->ip6po_pktinfo == NULL)
2303 			goto bad;
2304 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2305 	}
2306 	if (src->ip6po_nexthop) {
2307 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2308 		    M_IP6OPT, canwait);
2309 		if (dst->ip6po_nexthop == NULL)
2310 			goto bad;
2311 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2312 		    src->ip6po_nexthop->sa_len);
2313 	}
2314 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2315 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2316 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2317 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2318 	return (0);
2319 
2320   bad:
2321 	ip6_clearpktopts(dst, -1);
2322 	return (ENOBUFS);
2323 }
2324 #undef PKTOPT_EXTHDRCPY
2325 
2326 void
2327 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2328 {
2329 	if (pktopt == NULL)
2330 		return;
2331 
2332 	ip6_clearpktopts(pktopt, -1);
2333 
2334 	free(pktopt, M_IP6OPT);
2335 }
2336 
2337 /*
2338  * Set the IP6 multicast options in response to user setsockopt().
2339  */
2340 int
2341 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
2342 {
2343 	int error = 0;
2344 	u_int loop, ifindex;
2345 	struct ipv6_mreq *mreq;
2346 	struct ifnet *ifp;
2347 	struct ip6_moptions *im6o = *im6op;
2348 	struct route_in6 ro;
2349 	struct sockaddr_in6 *dst;
2350 	struct in6_multi_mship *imm;
2351 	struct proc *p = curproc;	/* XXX */
2352 
2353 	if (im6o == NULL) {
2354 		/*
2355 		 * No multicast option buffer attached to the pcb;
2356 		 * allocate one and initialize to default values.
2357 		 */
2358 		im6o = (struct ip6_moptions *)
2359 			malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
2360 
2361 		if (im6o == NULL)
2362 			return (ENOBUFS);
2363 		*im6op = im6o;
2364 		im6o->im6o_multicast_ifp = NULL;
2365 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2366 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2367 		LIST_INIT(&im6o->im6o_memberships);
2368 	}
2369 
2370 	switch (optname) {
2371 
2372 	case IPV6_MULTICAST_IF:
2373 		/*
2374 		 * Select the interface for outgoing multicast packets.
2375 		 */
2376 		if (m == NULL || m->m_len != sizeof(u_int)) {
2377 			error = EINVAL;
2378 			break;
2379 		}
2380 		bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
2381 		if (ifindex == 0)
2382 			ifp = NULL;
2383 		else {
2384 			if (ifindex < 0 || if_indexlim <= ifindex ||
2385 			    !ifindex2ifnet[ifindex]) {
2386 				error = ENXIO;	/* XXX EINVAL? */
2387 				break;
2388 			}
2389 			ifp = ifindex2ifnet[ifindex];
2390 			if (ifp == NULL ||
2391 			    (ifp->if_flags & IFF_MULTICAST) == 0) {
2392 				error = EADDRNOTAVAIL;
2393 				break;
2394 			}
2395 		}
2396 		im6o->im6o_multicast_ifp = ifp;
2397 		break;
2398 
2399 	case IPV6_MULTICAST_HOPS:
2400 	    {
2401 		/*
2402 		 * Set the IP6 hoplimit for outgoing multicast packets.
2403 		 */
2404 		int optval;
2405 		if (m == NULL || m->m_len != sizeof(int)) {
2406 			error = EINVAL;
2407 			break;
2408 		}
2409 		bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2410 		if (optval < -1 || optval >= 256)
2411 			error = EINVAL;
2412 		else if (optval == -1)
2413 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
2414 		else
2415 			im6o->im6o_multicast_hlim = optval;
2416 		break;
2417 	    }
2418 
2419 	case IPV6_MULTICAST_LOOP:
2420 		/*
2421 		 * Set the loopback flag for outgoing multicast packets.
2422 		 * Must be zero or one.
2423 		 */
2424 		if (m == NULL || m->m_len != sizeof(u_int)) {
2425 			error = EINVAL;
2426 			break;
2427 		}
2428 		bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2429 		if (loop > 1) {
2430 			error = EINVAL;
2431 			break;
2432 		}
2433 		im6o->im6o_multicast_loop = loop;
2434 		break;
2435 
2436 	case IPV6_JOIN_GROUP:
2437 		/*
2438 		 * Add a multicast group membership.
2439 		 * Group must be a valid IP6 multicast address.
2440 		 */
2441 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2442 			error = EINVAL;
2443 			break;
2444 		}
2445 		mreq = mtod(m, struct ipv6_mreq *);
2446 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2447 			/*
2448 			 * We use the unspecified address to specify to accept
2449 			 * all multicast addresses. Only super user is allowed
2450 			 * to do this.
2451 			 */
2452 			if (suser(p, 0))
2453 			{
2454 				error = EACCES;
2455 				break;
2456 			}
2457 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2458 			error = EINVAL;
2459 			break;
2460 		}
2461 
2462 		/*
2463 		 * If no interface was explicitly specified, choose an
2464 		 * appropriate one according to the given multicast address.
2465 		 */
2466 		if (mreq->ipv6mr_interface == 0) {
2467 			/*
2468 			 * Look up the routing table for the
2469 			 * address, and choose the outgoing interface.
2470 			 *   XXX: is it a good approach?
2471 			 */
2472 			bzero(&ro, sizeof(ro));
2473 			dst = (struct sockaddr_in6 *)&ro.ro_dst;
2474 			dst->sin6_len = sizeof(struct sockaddr_in6);
2475 			dst->sin6_family = AF_INET6;
2476 			dst->sin6_addr = mreq->ipv6mr_multiaddr;
2477 			rtalloc((struct route *)&ro);
2478 			if (ro.ro_rt == NULL) {
2479 				error = EADDRNOTAVAIL;
2480 				break;
2481 			}
2482 			ifp = ro.ro_rt->rt_ifp;
2483 			rtfree(ro.ro_rt);
2484 		} else {
2485 			/*
2486 			 * If the interface is specified, validate it.
2487 			 */
2488 			if (mreq->ipv6mr_interface < 0 ||
2489 			    if_indexlim <= mreq->ipv6mr_interface ||
2490 			    !ifindex2ifnet[mreq->ipv6mr_interface]) {
2491 				error = ENXIO;	/* XXX EINVAL? */
2492 				break;
2493 			}
2494 			ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2495 		}
2496 
2497 		/*
2498 		 * See if we found an interface, and confirm that it
2499 		 * supports multicast
2500 		 */
2501 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2502 			error = EADDRNOTAVAIL;
2503 			break;
2504 		}
2505 		/*
2506 		 * Put interface index into the multicast address,
2507 		 * if the address has link/interface-local scope.
2508 		 */
2509 		if (IN6_IS_SCOPE_EMBED(&mreq->ipv6mr_multiaddr)) {
2510 			mreq->ipv6mr_multiaddr.s6_addr16[1] =
2511 			    htons(ifp->if_index);
2512 		}
2513 		/*
2514 		 * See if the membership already exists.
2515 		 */
2516 		LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain)
2517 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
2518 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2519 			    &mreq->ipv6mr_multiaddr))
2520 				break;
2521 		if (imm != NULL) {
2522 			error = EADDRINUSE;
2523 			break;
2524 		}
2525 		/*
2526 		 * Everything looks good; add a new record to the multicast
2527 		 * address list for the given interface.
2528 		 */
2529 		imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error);
2530 		if (!imm)
2531 			break;
2532 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2533 		break;
2534 
2535 	case IPV6_LEAVE_GROUP:
2536 		/*
2537 		 * Drop a multicast group membership.
2538 		 * Group must be a valid IP6 multicast address.
2539 		 */
2540 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2541 			error = EINVAL;
2542 			break;
2543 		}
2544 		mreq = mtod(m, struct ipv6_mreq *);
2545 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2546 			if (suser(p, 0))
2547 			{
2548 				error = EACCES;
2549 				break;
2550 			}
2551 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2552 			error = EINVAL;
2553 			break;
2554 		}
2555 		/*
2556 		 * If an interface address was specified, get a pointer
2557 		 * to its ifnet structure.
2558 		 */
2559 		if (mreq->ipv6mr_interface == 0)
2560 			ifp = NULL;
2561 		else {
2562 			if (mreq->ipv6mr_interface < 0 ||
2563 			    if_indexlim <= mreq->ipv6mr_interface ||
2564 			    !ifindex2ifnet[mreq->ipv6mr_interface]) {
2565 				error = ENXIO;	/* XXX EINVAL? */
2566 				break;
2567 			}
2568 			ifp = ifindex2ifnet[mreq->ipv6mr_interface];
2569 		}
2570 
2571 		/*
2572 		 * Put interface index into the multicast address,
2573 		 * if the address has link-local scope.
2574 		 */
2575 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2576 			mreq->ipv6mr_multiaddr.s6_addr16[1] =
2577 			    htons(mreq->ipv6mr_interface);
2578 		}
2579 		/*
2580 		 * Find the membership in the membership list.
2581 		 */
2582 		LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain) {
2583 			if ((ifp == NULL || imm->i6mm_maddr->in6m_ifp == ifp) &&
2584 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2585 			    &mreq->ipv6mr_multiaddr))
2586 				break;
2587 		}
2588 		if (imm == NULL) {
2589 			/* Unable to resolve interface */
2590 			error = EADDRNOTAVAIL;
2591 			break;
2592 		}
2593 		/*
2594 		 * Give up the multicast address record to which the
2595 		 * membership points.
2596 		 */
2597 		LIST_REMOVE(imm, i6mm_chain);
2598 		in6_leavegroup(imm);
2599 		break;
2600 
2601 	default:
2602 		error = EOPNOTSUPP;
2603 		break;
2604 	}
2605 
2606 	/*
2607 	 * If all options have default values, no need to keep the option
2608 	 * structure.
2609 	 */
2610 	if (im6o->im6o_multicast_ifp == NULL &&
2611 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
2612 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2613 	    LIST_EMPTY(&im6o->im6o_memberships)) {
2614 		free(*im6op, M_IPMOPTS);
2615 		*im6op = NULL;
2616 	}
2617 
2618 	return (error);
2619 }
2620 
2621 /*
2622  * Return the IP6 multicast options in response to user getsockopt().
2623  */
2624 int
2625 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp)
2626 {
2627 	u_int *hlim, *loop, *ifindex;
2628 
2629 	*mp = m_get(M_WAIT, MT_SOOPTS);
2630 
2631 	switch (optname) {
2632 
2633 	case IPV6_MULTICAST_IF:
2634 		ifindex = mtod(*mp, u_int *);
2635 		(*mp)->m_len = sizeof(u_int);
2636 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
2637 			*ifindex = 0;
2638 		else
2639 			*ifindex = im6o->im6o_multicast_ifp->if_index;
2640 		return (0);
2641 
2642 	case IPV6_MULTICAST_HOPS:
2643 		hlim = mtod(*mp, u_int *);
2644 		(*mp)->m_len = sizeof(u_int);
2645 		if (im6o == NULL)
2646 			*hlim = ip6_defmcasthlim;
2647 		else
2648 			*hlim = im6o->im6o_multicast_hlim;
2649 		return (0);
2650 
2651 	case IPV6_MULTICAST_LOOP:
2652 		loop = mtod(*mp, u_int *);
2653 		(*mp)->m_len = sizeof(u_int);
2654 		if (im6o == NULL)
2655 			*loop = ip6_defmcasthlim;
2656 		else
2657 			*loop = im6o->im6o_multicast_loop;
2658 		return (0);
2659 
2660 	default:
2661 		return (EOPNOTSUPP);
2662 	}
2663 }
2664 
2665 /*
2666  * Discard the IP6 multicast options.
2667  */
2668 void
2669 ip6_freemoptions(struct ip6_moptions *im6o)
2670 {
2671 	struct in6_multi_mship *imm;
2672 
2673 	if (im6o == NULL)
2674 		return;
2675 
2676 	while (!LIST_EMPTY(&im6o->im6o_memberships)) {
2677 		imm = LIST_FIRST(&im6o->im6o_memberships);
2678 		LIST_REMOVE(imm, i6mm_chain);
2679 		in6_leavegroup(imm);
2680 	}
2681 	free(im6o, M_IPMOPTS);
2682 }
2683 
2684 /*
2685  * Set IPv6 outgoing packet options based on advanced API.
2686  */
2687 int
2688 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2689     struct ip6_pktopts *stickyopt, int priv, int uproto)
2690 {
2691 	u_int clen;
2692 	struct cmsghdr *cm = 0;
2693 	caddr_t cmsgs;
2694 	int error;
2695 
2696 	if (control == NULL || opt == NULL)
2697 		return (EINVAL);
2698 
2699 	ip6_initpktopts(opt);
2700 	if (stickyopt) {
2701 		int error;
2702 
2703 		/*
2704 		 * If stickyopt is provided, make a local copy of the options
2705 		 * for this particular packet, then override them by ancillary
2706 		 * objects.
2707 		 * XXX: copypktopts() does not copy the cached route to a next
2708 		 * hop (if any).  This is not very good in terms of efficiency,
2709 		 * but we can allow this since this option should be rarely
2710 		 * used.
2711 		 */
2712 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2713 			return (error);
2714 	}
2715 
2716 	/*
2717 	 * XXX: Currently, we assume all the optional information is stored
2718 	 * in a single mbuf.
2719 	 */
2720 	if (control->m_next)
2721 		return (EINVAL);
2722 
2723 	clen = control->m_len;
2724 	cmsgs = mtod(control, caddr_t);
2725 	do {
2726 		if (clen < CMSG_LEN(0))
2727 			return (EINVAL);
2728 		cm = (struct cmsghdr *)cmsgs;
2729 		if (cm->cmsg_len < CMSG_LEN(0) ||
2730 		    CMSG_ALIGN(cm->cmsg_len) > clen)
2731 			return (EINVAL);
2732 		if (cm->cmsg_level == IPPROTO_IPV6) {
2733 			error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2734 			    cm->cmsg_len - CMSG_LEN(0), opt, priv, 0, 1, uproto);
2735 			if (error)
2736 				return (error);
2737 		}
2738 
2739 		clen -= CMSG_ALIGN(cm->cmsg_len);
2740 		cmsgs += CMSG_ALIGN(cm->cmsg_len);
2741 	} while (clen);
2742 
2743 	return (0);
2744 }
2745 
2746 /*
2747  * Set a particular packet option, as a sticky option or an ancillary data
2748  * item.  "len" can be 0 only when it's a sticky option.
2749  * We have 4 cases of combination of "sticky" and "cmsg":
2750  * "sticky=0, cmsg=0": impossible
2751  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2752  * "sticky=1, cmsg=0": RFC3542 socket option
2753  * "sticky=1, cmsg=1": RFC2292 socket option
2754  */
2755 int
2756 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2757     int priv, int sticky, int cmsg, int uproto)
2758 {
2759 	int minmtupolicy;
2760 
2761 	if (!sticky && !cmsg) {
2762 #ifdef DIAGNOSTIC
2763 		printf("ip6_setpktopt: impossible case\n");
2764 #endif
2765 		return (EINVAL);
2766 	}
2767 
2768 	/*
2769 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2770 	 * not be specified in the context of RFC3542.  Conversely,
2771 	 * RFC3542 types should not be specified in the context of RFC2292.
2772 	 */
2773 	if (!cmsg) {
2774 		switch (optname) {
2775 		case IPV6_2292PKTINFO:
2776 		case IPV6_2292HOPLIMIT:
2777 		case IPV6_2292NEXTHOP:
2778 		case IPV6_2292HOPOPTS:
2779 		case IPV6_2292DSTOPTS:
2780 		case IPV6_2292RTHDR:
2781 		case IPV6_2292PKTOPTIONS:
2782 			return (ENOPROTOOPT);
2783 		}
2784 	}
2785 	if (sticky && cmsg) {
2786 		switch (optname) {
2787 		case IPV6_PKTINFO:
2788 		case IPV6_HOPLIMIT:
2789 		case IPV6_NEXTHOP:
2790 		case IPV6_HOPOPTS:
2791 		case IPV6_DSTOPTS:
2792 		case IPV6_RTHDRDSTOPTS:
2793 		case IPV6_RTHDR:
2794 		case IPV6_USE_MIN_MTU:
2795 		case IPV6_DONTFRAG:
2796 		case IPV6_TCLASS:
2797 			return (ENOPROTOOPT);
2798 		}
2799 	}
2800 
2801 	switch (optname) {
2802 	case IPV6_2292PKTINFO:
2803 	case IPV6_PKTINFO:
2804 	{
2805 		struct ifnet *ifp = NULL;
2806 		struct in6_pktinfo *pktinfo;
2807 
2808 		if (len != sizeof(struct in6_pktinfo))
2809 			return (EINVAL);
2810 
2811 		pktinfo = (struct in6_pktinfo *)buf;
2812 
2813 		/*
2814 		 * An application can clear any sticky IPV6_PKTINFO option by
2815 		 * doing a "regular" setsockopt with ipi6_addr being
2816 		 * in6addr_any and ipi6_ifindex being zero.
2817 		 * [RFC 3542, Section 6]
2818 		 */
2819 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2820 		    pktinfo->ipi6_ifindex == 0 &&
2821 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2822 			ip6_clearpktopts(opt, optname);
2823 			break;
2824 		}
2825 
2826 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2827 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2828 			return (EINVAL);
2829 		}
2830 
2831 		/* validate the interface index if specified. */
2832 		if (pktinfo->ipi6_ifindex >= if_indexlim ||
2833 		    pktinfo->ipi6_ifindex < 0) {
2834 			 return (ENXIO);
2835 		}
2836 		if (pktinfo->ipi6_ifindex) {
2837 			ifp = ifindex2ifnet[pktinfo->ipi6_ifindex];
2838 			if (ifp == NULL)
2839 				return (ENXIO);
2840 		}
2841 
2842 		/*
2843 		 * We store the address anyway, and let in6_selectsrc()
2844 		 * validate the specified address.  This is because ipi6_addr
2845 		 * may not have enough information about its scope zone, and
2846 		 * we may need additional information (such as outgoing
2847 		 * interface or the scope zone of a destination address) to
2848 		 * disambiguate the scope.
2849 		 * XXX: the delay of the validation may confuse the
2850 		 * application when it is used as a sticky option.
2851 		 */
2852 		if (opt->ip6po_pktinfo == NULL) {
2853 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2854 			    M_IP6OPT, M_NOWAIT);
2855 			if (opt->ip6po_pktinfo == NULL)
2856 				return (ENOBUFS);
2857 		}
2858 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2859 		break;
2860 	}
2861 
2862 	case IPV6_2292HOPLIMIT:
2863 	case IPV6_HOPLIMIT:
2864 	{
2865 		int *hlimp;
2866 
2867 		/*
2868 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2869 		 * to simplify the ordering among hoplimit options.
2870 		 */
2871 		if (optname == IPV6_HOPLIMIT && sticky)
2872 			return (ENOPROTOOPT);
2873 
2874 		if (len != sizeof(int))
2875 			return (EINVAL);
2876 		hlimp = (int *)buf;
2877 		if (*hlimp < -1 || *hlimp > 255)
2878 			return (EINVAL);
2879 
2880 		opt->ip6po_hlim = *hlimp;
2881 		break;
2882 	}
2883 
2884 	case IPV6_TCLASS:
2885 	{
2886 		int tclass;
2887 
2888 		if (len != sizeof(int))
2889 			return (EINVAL);
2890 		tclass = *(int *)buf;
2891 		if (tclass < -1 || tclass > 255)
2892 			return (EINVAL);
2893 
2894 		opt->ip6po_tclass = tclass;
2895 		break;
2896 	}
2897 
2898 	case IPV6_2292NEXTHOP:
2899 	case IPV6_NEXTHOP:
2900 		if (!priv)
2901 			return (EPERM);
2902 
2903 		if (len == 0) {	/* just remove the option */
2904 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2905 			break;
2906 		}
2907 
2908 		/* check if cmsg_len is large enough for sa_len */
2909 		if (len < sizeof(struct sockaddr) || len < *buf)
2910 			return (EINVAL);
2911 
2912 		switch (((struct sockaddr *)buf)->sa_family) {
2913 		case AF_INET6:
2914 		{
2915 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2916 
2917 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2918 				return (EINVAL);
2919 
2920 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2921 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2922 				return (EINVAL);
2923 			}
2924 			if (IN6_IS_SCOPE_EMBED(&sa6->sin6_addr)) {
2925 				if (sa6->sin6_scope_id < 0 ||
2926 				    if_indexlim <= sa6->sin6_scope_id ||
2927 				    !ifindex2ifnet[sa6->sin6_scope_id])
2928 					return (EINVAL);
2929 				sa6->sin6_addr.s6_addr16[1] =
2930 				    htonl(sa6->sin6_scope_id);
2931 			} else if (sa6->sin6_scope_id)
2932 				return (EINVAL);
2933 			break;
2934 		}
2935 		case AF_LINK:	/* eventually be supported? */
2936 		default:
2937 			return (EAFNOSUPPORT);
2938 		}
2939 
2940 		/* turn off the previous option, then set the new option. */
2941 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2942 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2943 		if (opt->ip6po_nexthop == NULL)
2944 			return (ENOBUFS);
2945 		bcopy(buf, opt->ip6po_nexthop, *buf);
2946 		break;
2947 
2948 	case IPV6_2292HOPOPTS:
2949 	case IPV6_HOPOPTS:
2950 	{
2951 		struct ip6_hbh *hbh;
2952 		int hbhlen;
2953 
2954 		/*
2955 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2956 		 * options, since per-option restriction has too much
2957 		 * overhead.
2958 		 */
2959 		if (!priv)
2960 			return (EPERM);
2961 
2962 		if (len == 0) {
2963 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2964 			break;	/* just remove the option */
2965 		}
2966 
2967 		/* message length validation */
2968 		if (len < sizeof(struct ip6_hbh))
2969 			return (EINVAL);
2970 		hbh = (struct ip6_hbh *)buf;
2971 		hbhlen = (hbh->ip6h_len + 1) << 3;
2972 		if (len != hbhlen)
2973 			return (EINVAL);
2974 
2975 		/* turn off the previous option, then set the new option. */
2976 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2977 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2978 		if (opt->ip6po_hbh == NULL)
2979 			return (ENOBUFS);
2980 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
2981 
2982 		break;
2983 	}
2984 
2985 	case IPV6_2292DSTOPTS:
2986 	case IPV6_DSTOPTS:
2987 	case IPV6_RTHDRDSTOPTS:
2988 	{
2989 		struct ip6_dest *dest, **newdest = NULL;
2990 		int destlen;
2991 
2992 		if (!priv)	/* XXX: see the comment for IPV6_HOPOPTS */
2993 			return (EPERM);
2994 
2995 		if (len == 0) {
2996 			ip6_clearpktopts(opt, optname);
2997 			break;	/* just remove the option */
2998 		}
2999 
3000 		/* message length validation */
3001 		if (len < sizeof(struct ip6_dest))
3002 			return (EINVAL);
3003 		dest = (struct ip6_dest *)buf;
3004 		destlen = (dest->ip6d_len + 1) << 3;
3005 		if (len != destlen)
3006 			return (EINVAL);
3007 		/*
3008 		 * Determine the position that the destination options header
3009 		 * should be inserted; before or after the routing header.
3010 		 */
3011 		switch (optname) {
3012 		case IPV6_2292DSTOPTS:
3013 			/*
3014 			 * The old advanced API is ambiguous on this point.
3015 			 * Our approach is to determine the position based
3016 			 * according to the existence of a routing header.
3017 			 * Note, however, that this depends on the order of the
3018 			 * extension headers in the ancillary data; the 1st
3019 			 * part of the destination options header must appear
3020 			 * before the routing header in the ancillary data,
3021 			 * too.
3022 			 * RFC3542 solved the ambiguity by introducing
3023 			 * separate ancillary data or option types.
3024 			 */
3025 			if (opt->ip6po_rthdr == NULL)
3026 				newdest = &opt->ip6po_dest1;
3027 			else
3028 				newdest = &opt->ip6po_dest2;
3029 			break;
3030 		case IPV6_RTHDRDSTOPTS:
3031 			newdest = &opt->ip6po_dest1;
3032 			break;
3033 		case IPV6_DSTOPTS:
3034 			newdest = &opt->ip6po_dest2;
3035 			break;
3036 		}
3037 
3038 		/* turn off the previous option, then set the new option. */
3039 		ip6_clearpktopts(opt, optname);
3040 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
3041 		if (*newdest == NULL)
3042 			return (ENOBUFS);
3043 		bcopy(dest, *newdest, destlen);
3044 
3045 		break;
3046 	}
3047 
3048 	case IPV6_2292RTHDR:
3049 	case IPV6_RTHDR:
3050 	{
3051 		struct ip6_rthdr *rth;
3052 		int rthlen;
3053 
3054 		if (len == 0) {
3055 			ip6_clearpktopts(opt, IPV6_RTHDR);
3056 			break;	/* just remove the option */
3057 		}
3058 
3059 		/* message length validation */
3060 		if (len < sizeof(struct ip6_rthdr))
3061 			return (EINVAL);
3062 		rth = (struct ip6_rthdr *)buf;
3063 		rthlen = (rth->ip6r_len + 1) << 3;
3064 		if (len != rthlen)
3065 			return (EINVAL);
3066 
3067 		switch (rth->ip6r_type) {
3068 		case IPV6_RTHDR_TYPE_0:
3069 			if (rth->ip6r_len == 0)	/* must contain one addr */
3070 				return (EINVAL);
3071 			if (rth->ip6r_len % 2) /* length must be even */
3072 				return (EINVAL);
3073 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
3074 				return (EINVAL);
3075 			break;
3076 		default:
3077 			return (EINVAL);	/* not supported */
3078 		}
3079 		/* turn off the previous option */
3080 		ip6_clearpktopts(opt, IPV6_RTHDR);
3081 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
3082 		if (opt->ip6po_rthdr == NULL)
3083 			return (ENOBUFS);
3084 		bcopy(rth, opt->ip6po_rthdr, rthlen);
3085 		break;
3086 	}
3087 
3088 	case IPV6_USE_MIN_MTU:
3089 		if (len != sizeof(int))
3090 			return (EINVAL);
3091 		minmtupolicy = *(int *)buf;
3092 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
3093 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
3094 		    minmtupolicy != IP6PO_MINMTU_ALL) {
3095 			return (EINVAL);
3096 		}
3097 		opt->ip6po_minmtu = minmtupolicy;
3098 		break;
3099 
3100 	case IPV6_DONTFRAG:
3101 		if (len != sizeof(int))
3102 			return (EINVAL);
3103 
3104 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
3105 			/*
3106 			 * we ignore this option for TCP sockets.
3107 			 * (RFC3542 leaves this case unspecified.)
3108 			 */
3109 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
3110 		} else
3111 			opt->ip6po_flags |= IP6PO_DONTFRAG;
3112 		break;
3113 
3114 	default:
3115 		return (ENOPROTOOPT);
3116 	} /* end of switch */
3117 
3118 	return (0);
3119 }
3120 
3121 /*
3122  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3123  * packet to the input queue of a specified interface.  Note that this
3124  * calls the output routine of the loopback "driver", but with an interface
3125  * pointer that might NOT be lo0ifp -- easier than replicating that code here.
3126  */
3127 void
3128 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
3129 {
3130 	struct mbuf *copym;
3131 	struct ip6_hdr *ip6;
3132 
3133 	/*
3134 	 * Duplicate the packet.
3135 	 */
3136 	copym = m_copy(m, 0, M_COPYALL);
3137 	if (copym == NULL)
3138 		return;
3139 
3140 	/*
3141 	 * Make sure to deep-copy IPv6 header portion in case the data
3142 	 * is in an mbuf cluster, so that we can safely override the IPv6
3143 	 * header portion later.
3144 	 */
3145 	if ((copym->m_flags & M_EXT) != 0 ||
3146 	    copym->m_len < sizeof(struct ip6_hdr)) {
3147 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3148 		if (copym == NULL)
3149 			return;
3150 	}
3151 
3152 #ifdef DIAGNOSTIC
3153 	if (copym->m_len < sizeof(*ip6)) {
3154 		m_freem(copym);
3155 		return;
3156 	}
3157 #endif
3158 
3159 	ip6 = mtod(copym, struct ip6_hdr *);
3160 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
3161 		ip6->ip6_src.s6_addr16[1] = 0;
3162 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
3163 		ip6->ip6_dst.s6_addr16[1] = 0;
3164 
3165 	(void)looutput(ifp, copym, (struct sockaddr *)dst, NULL);
3166 }
3167 
3168 /*
3169  * Chop IPv6 header off from the payload.
3170  */
3171 int
3172 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3173 {
3174 	struct mbuf *mh;
3175 	struct ip6_hdr *ip6;
3176 
3177 	ip6 = mtod(m, struct ip6_hdr *);
3178 	if (m->m_len > sizeof(*ip6)) {
3179 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3180 		if (mh == 0) {
3181 			m_freem(m);
3182 			return ENOBUFS;
3183 		}
3184 		M_MOVE_PKTHDR(mh, m);
3185 		MH_ALIGN(mh, sizeof(*ip6));
3186 		m->m_len -= sizeof(*ip6);
3187 		m->m_data += sizeof(*ip6);
3188 		mh->m_next = m;
3189 		m = mh;
3190 		m->m_len = sizeof(*ip6);
3191 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3192 	}
3193 	exthdrs->ip6e_ip6 = m;
3194 	return 0;
3195 }
3196 
3197 u_int32_t
3198 ip6_randomid(void)
3199 {
3200 	return idgen32(&ip6_id_ctx);
3201 }
3202 
3203 void
3204 ip6_randomid_init(void)
3205 {
3206 	idgen32_init(&ip6_id_ctx);
3207 }
3208