xref: /openbsd-src/sys/netinet6/ip6_output.c (revision f2da64fbbbf1b03f09f390ab01267c93dfd77c4c)
1 /*	$OpenBSD: ip6_output.c,v 1.216 2016/09/19 18:09:09 tedu Exp $	*/
2 /*	$KAME: ip6_output.c,v 1.172 2001/03/25 09:55:56 itojun Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1988, 1990, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
62  */
63 
64 #include "pf.h"
65 
66 #include <sys/param.h>
67 #include <sys/malloc.h>
68 #include <sys/mbuf.h>
69 #include <sys/errno.h>
70 #include <sys/protosw.h>
71 #include <sys/socket.h>
72 #include <sys/socketvar.h>
73 #include <sys/proc.h>
74 #include <sys/systm.h>
75 
76 #include <net/if.h>
77 #include <net/if_var.h>
78 #include <net/if_enc.h>
79 #include <net/route.h>
80 
81 #include <netinet/in.h>
82 #include <netinet/ip.h>
83 #include <netinet/in_pcb.h>
84 #include <netinet/udp.h>
85 #include <netinet/tcp.h>
86 
87 #include <netinet/ip_var.h>
88 #include <netinet/tcp_timer.h>
89 #include <netinet/tcp_var.h>
90 #include <netinet/udp_var.h>
91 
92 #include <netinet6/in6_var.h>
93 #include <netinet/ip6.h>
94 #include <netinet/icmp6.h>
95 #include <netinet6/ip6_var.h>
96 #include <netinet6/nd6.h>
97 #include <netinet6/ip6protosw.h>
98 
99 #include <crypto/idgen.h>
100 
101 #if NPF > 0
102 #include <net/pfvar.h>
103 #endif
104 
105 #ifdef IPSEC
106 #include <netinet/ip_ipsp.h>
107 #include <netinet/ip_ah.h>
108 #include <netinet/ip_esp.h>
109 #endif /* IPSEC */
110 
111 struct ip6_exthdrs {
112 	struct mbuf *ip6e_ip6;
113 	struct mbuf *ip6e_hbh;
114 	struct mbuf *ip6e_dest1;
115 	struct mbuf *ip6e_rthdr;
116 	struct mbuf *ip6e_dest2;
117 };
118 
119 int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **, int, int);
120 int ip6_getpcbopt(struct ip6_pktopts *, int, struct mbuf **);
121 int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, int, int, int);
122 int ip6_setmoptions(int, struct ip6_moptions **, struct mbuf *);
123 int ip6_getmoptions(int, struct ip6_moptions *, struct mbuf **);
124 int ip6_copyexthdr(struct mbuf **, caddr_t, int);
125 int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
126 	struct ip6_frag **);
127 int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
128 int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
129 int ip6_getpmtu(struct rtentry *, struct ifnet *, u_long *, int *);
130 int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
131 static __inline u_int16_t __attribute__((__unused__))
132     in6_cksum_phdr(const struct in6_addr *, const struct in6_addr *,
133     u_int32_t, u_int32_t);
134 void in6_delayed_cksum(struct mbuf *, u_int8_t);
135 
136 /* Context for non-repeating IDs */
137 struct idgen32_ctx ip6_id_ctx;
138 
139 /*
140  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
141  * header (with pri, len, nxt, hlim, src, dst).
142  * This function may modify ver and hlim only.
143  * The mbuf chain containing the packet will be freed.
144  * The mbuf opt, if present, will not be freed.
145  *
146  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
147  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
148  * which is rt_rmx.rmx_mtu.
149  */
150 int
151 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt, struct route_in6 *ro,
152     int flags, struct ip6_moptions *im6o, struct inpcb *inp)
153 {
154 	struct ip6_hdr *ip6;
155 	struct ifnet *ifp = NULL;
156 	struct mbuf *m = m0;
157 	int hlen, tlen;
158 	struct route_in6 ip6route;
159 	struct rtentry *rt = NULL;
160 	struct sockaddr_in6 *dst, dstsock;
161 	int error = 0;
162 	u_long mtu;
163 	int alwaysfrag, dontfrag;
164 	u_int16_t src_scope, dst_scope;
165 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
166 	struct ip6_exthdrs exthdrs;
167 	struct in6_addr finaldst;
168 	struct route_in6 *ro_pmtu = NULL;
169 	int hdrsplit = 0;
170 	u_int8_t sproto = 0;
171 #ifdef IPSEC
172 	struct tdb *tdb = NULL;
173 #endif /* IPSEC */
174 
175 #ifdef IPSEC
176 	if (inp && (inp->inp_flags & INP_IPV6) == 0)
177 		panic("ip6_output: IPv4 pcb is passed");
178 #endif /* IPSEC */
179 
180 	ip6 = mtod(m, struct ip6_hdr *);
181 	finaldst = ip6->ip6_dst;
182 
183 #define MAKE_EXTHDR(hp, mp)						\
184     do {								\
185 	if (hp) {							\
186 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
187 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
188 		    ((eh)->ip6e_len + 1) << 3);				\
189 		if (error)						\
190 			goto freehdrs;					\
191 	}								\
192     } while (0)
193 
194 	bzero(&exthdrs, sizeof(exthdrs));
195 
196 	if (opt) {
197 		/* Hop-by-Hop options header */
198 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
199 		/* Destination options header(1st part) */
200 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
201 		/* Routing header */
202 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
203 		/* Destination options header(2nd part) */
204 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
205 	}
206 
207 #ifdef IPSEC
208 	if (ipsec_in_use || inp) {
209 		tdb = ip6_output_ipsec_lookup(m, &error, inp);
210 		if (error != 0) {
211 		        /*
212 			 * -EINVAL is used to indicate that the packet should
213 			 * be silently dropped, typically because we've asked
214 			 * key management for an SA.
215 			 */
216 		        if (error == -EINVAL) /* Should silently drop packet */
217 				error = 0;
218 
219 			goto freehdrs;
220 		}
221 	}
222 #endif /* IPSEC */
223 
224 	/*
225 	 * Calculate the total length of the extension header chain.
226 	 * Keep the length of the unfragmentable part for fragmentation.
227 	 */
228 	optlen = 0;
229 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
230 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
231 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
232 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
233 	/* NOTE: we don't add AH/ESP length here. do that later. */
234 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
235 
236 	/*
237 	 * If we need IPsec, or there is at least one extension header,
238 	 * separate IP6 header from the payload.
239 	 */
240 	if ((sproto || optlen) && !hdrsplit) {
241 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
242 			m = NULL;
243 			goto freehdrs;
244 		}
245 		m = exthdrs.ip6e_ip6;
246 		hdrsplit++;
247 	}
248 
249 	/* adjust pointer */
250 	ip6 = mtod(m, struct ip6_hdr *);
251 
252 	/* adjust mbuf packet header length */
253 	m->m_pkthdr.len += optlen;
254 	plen = m->m_pkthdr.len - sizeof(*ip6);
255 
256 	/* If this is a jumbo payload, insert a jumbo payload option. */
257 	if (plen > IPV6_MAXPACKET) {
258 		if (!hdrsplit) {
259 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
260 				m = NULL;
261 				goto freehdrs;
262 			}
263 			m = exthdrs.ip6e_ip6;
264 			hdrsplit++;
265 		}
266 		/* adjust pointer */
267 		ip6 = mtod(m, struct ip6_hdr *);
268 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
269 			goto freehdrs;
270 		ip6->ip6_plen = 0;
271 	} else
272 		ip6->ip6_plen = htons(plen);
273 
274 	/*
275 	 * Concatenate headers and fill in next header fields.
276 	 * Here we have, on "m"
277 	 *	IPv6 payload
278 	 * and we insert headers accordingly.  Finally, we should be getting:
279 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
280 	 *
281 	 * during the header composing process, "m" points to IPv6 header.
282 	 * "mprev" points to an extension header prior to esp.
283 	 */
284 	{
285 		u_char *nexthdrp = &ip6->ip6_nxt;
286 		struct mbuf *mprev = m;
287 
288 		/*
289 		 * we treat dest2 specially.  this makes IPsec processing
290 		 * much easier.  the goal here is to make mprev point the
291 		 * mbuf prior to dest2.
292 		 *
293 		 * result: IPv6 dest2 payload
294 		 * m and mprev will point to IPv6 header.
295 		 */
296 		if (exthdrs.ip6e_dest2) {
297 			if (!hdrsplit)
298 				panic("assumption failed: hdr not split");
299 			exthdrs.ip6e_dest2->m_next = m->m_next;
300 			m->m_next = exthdrs.ip6e_dest2;
301 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
302 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
303 		}
304 
305 #define MAKE_CHAIN(m, mp, p, i)\
306     do {\
307 	if (m) {\
308 		if (!hdrsplit) \
309 			panic("assumption failed: hdr not split"); \
310 		*mtod((m), u_char *) = *(p);\
311 		*(p) = (i);\
312 		p = mtod((m), u_char *);\
313 		(m)->m_next = (mp)->m_next;\
314 		(mp)->m_next = (m);\
315 		(mp) = (m);\
316 	}\
317     } while (0)
318 		/*
319 		 * result: IPv6 hbh dest1 rthdr dest2 payload
320 		 * m will point to IPv6 header.  mprev will point to the
321 		 * extension header prior to dest2 (rthdr in the above case).
322 		 */
323 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
324 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
325 		    IPPROTO_DSTOPTS);
326 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
327 		    IPPROTO_ROUTING);
328 	}
329 
330 	/*
331 	 * If there is a routing header, replace the destination address field
332 	 * with the first hop of the routing header.
333 	 */
334 	if (exthdrs.ip6e_rthdr) {
335 		struct ip6_rthdr *rh;
336 		struct ip6_rthdr0 *rh0;
337 		struct in6_addr *addr;
338 
339 		rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
340 		    struct ip6_rthdr *));
341 		switch (rh->ip6r_type) {
342 		case IPV6_RTHDR_TYPE_0:
343 			 rh0 = (struct ip6_rthdr0 *)rh;
344 			 addr = (struct in6_addr *)(rh0 + 1);
345 			 ip6->ip6_dst = addr[0];
346 			 bcopy(&addr[1], &addr[0],
347 			     sizeof(struct in6_addr) * (rh0->ip6r0_segleft - 1));
348 			 addr[rh0->ip6r0_segleft - 1] = finaldst;
349 			 break;
350 		default:	/* is it possible? */
351 			 error = EINVAL;
352 			 goto bad;
353 		}
354 	}
355 
356 	/* Source address validation */
357 	if (!(flags & IPV6_UNSPECSRC) &&
358 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
359 		/*
360 		 * XXX: we can probably assume validation in the caller, but
361 		 * we explicitly check the address here for safety.
362 		 */
363 		error = EOPNOTSUPP;
364 		ip6stat.ip6s_badscope++;
365 		goto bad;
366 	}
367 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
368 		error = EOPNOTSUPP;
369 		ip6stat.ip6s_badscope++;
370 		goto bad;
371 	}
372 
373 	ip6stat.ip6s_localout++;
374 
375 	/*
376 	 * Route packet.
377 	 */
378 #if NPF > 0
379 reroute:
380 #endif
381 
382 	/* initialize cached route */
383 	if (ro == NULL) {
384 		ro = &ip6route;
385 		bzero((caddr_t)ro, sizeof(*ro));
386 	}
387 	ro_pmtu = ro;
388 	if (opt && opt->ip6po_rthdr)
389 		ro = &opt->ip6po_route;
390 	dst = &ro->ro_dst;
391 
392 	/*
393 	 * if specified, try to fill in the traffic class field.
394 	 * do not override if a non-zero value is already set.
395 	 * we check the diffserv field and the ecn field separately.
396 	 */
397 	if (opt && opt->ip6po_tclass >= 0) {
398 		int mask = 0;
399 
400 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
401 			mask |= 0xfc;
402 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
403 			mask |= 0x03;
404 		if (mask != 0)
405 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
406 	}
407 
408 	/* fill in or override the hop limit field, if necessary. */
409 	if (opt && opt->ip6po_hlim != -1)
410 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
411 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
412 		if (im6o != NULL)
413 			ip6->ip6_hlim = im6o->im6o_hlim;
414 		else
415 			ip6->ip6_hlim = ip6_defmcasthlim;
416 	}
417 
418 #ifdef IPSEC
419 	if (tdb) {
420 		/*
421 		 * XXX what should we do if ip6_hlim == 0 and the
422 		 * packet gets tunneled?
423 		 */
424 		/*
425 		 * if we are source-routing, do not attempt to tunnel the
426 		 * packet just because ip6_dst is different from what tdb has.
427 		 * XXX
428 		 */
429 		error = ip6_output_ipsec_send(tdb, m,
430 		    exthdrs.ip6e_rthdr ? 1 : 0, 0);
431 		goto done;
432 	}
433 #endif /* IPSEC */
434 
435 	bzero(&dstsock, sizeof(dstsock));
436 	dstsock.sin6_family = AF_INET6;
437 	dstsock.sin6_addr = ip6->ip6_dst;
438 	dstsock.sin6_len = sizeof(dstsock);
439 	ro->ro_tableid = m->m_pkthdr.ph_rtableid;
440 
441 	if (IN6_IS_ADDR_MULTICAST(&dstsock.sin6_addr)) {
442 		struct in6_pktinfo *pi = NULL;
443 
444 		/*
445 		 * If the caller specify the outgoing interface
446 		 * explicitly, use it.
447 		 */
448 		if (opt != NULL && (pi = opt->ip6po_pktinfo) != NULL)
449 			ifp = if_get(pi->ipi6_ifindex);
450 
451 		if (ifp == NULL && im6o != NULL)
452 			ifp = if_get(im6o->im6o_ifidx);
453 	}
454 
455 	if (ifp == NULL) {
456 		rt = in6_selectroute(&dstsock, opt, ro, ro->ro_tableid);
457 		if (rt == NULL) {
458 			ip6stat.ip6s_noroute++;
459 			error = EHOSTUNREACH;
460 			goto bad;
461 		}
462 		if (ISSET(rt->rt_flags, RTF_LOCAL))
463 			ifp = if_get(lo0ifidx);
464 		else
465 			ifp = if_get(rt->rt_ifidx);
466 	} else {
467 		*dst = dstsock;
468 	}
469 
470 	if (rt && (rt->rt_flags & RTF_GATEWAY) &&
471 	    !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
472 		dst = satosin6(rt->rt_gateway);
473 
474 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
475 		/* Unicast */
476 
477 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
478 	} else {
479 		/* Multicast */
480 
481 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
482 
483 		/*
484 		 * Confirm that the outgoing interface supports multicast.
485 		 */
486 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
487 			ip6stat.ip6s_noroute++;
488 			error = ENETUNREACH;
489 			goto bad;
490 		}
491 
492 		if ((im6o == NULL || im6o->im6o_loop) &&
493 		    in6_hasmulti(&ip6->ip6_dst, ifp)) {
494 			/*
495 			 * If we belong to the destination multicast group
496 			 * on the outgoing interface, and the caller did not
497 			 * forbid loopback, loop back a copy.
498 			 * Can't defer TCP/UDP checksumming, do the
499 			 * computation now.
500 			 */
501 			in6_proto_cksum_out(m, NULL);
502 			ip6_mloopback(ifp, m, dst);
503 		}
504 #ifdef MROUTING
505 		else {
506 			/*
507 			 * If we are acting as a multicast router, perform
508 			 * multicast forwarding as if the packet had just
509 			 * arrived on the interface to which we are about
510 			 * to send.  The multicast forwarding function
511 			 * recursively calls this function, using the
512 			 * IPV6_FORWARDING flag to prevent infinite recursion.
513 			 *
514 			 * Multicasts that are looped back by ip6_mloopback(),
515 			 * above, will be forwarded by the ip6_input() routine,
516 			 * if necessary.
517 			 */
518 			if (ip6_mforwarding && ip6_mrouter &&
519 			    (flags & IPV6_FORWARDING) == 0) {
520 				if (ip6_mforward(ip6, ifp, m) != 0) {
521 					m_freem(m);
522 					goto done;
523 				}
524 			}
525 		}
526 #endif
527 		/*
528 		 * Multicasts with a hoplimit of zero may be looped back,
529 		 * above, but must not be transmitted on a network.
530 		 * Also, multicasts addressed to the loopback interface
531 		 * are not sent -- the above call to ip6_mloopback() will
532 		 * loop back a copy if this host actually belongs to the
533 		 * destination group on the loopback interface.
534 		 */
535 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
536 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
537 			m_freem(m);
538 			goto done;
539 		}
540 	}
541 
542 	/*
543 	 * If this packet is going trough a loopback interface we wont
544 	 * be able to restore its scope ID using the interface index.
545 	 */
546 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src)) {
547 		if (ifp->if_flags & IFF_LOOPBACK)
548 			src_scope = ip6->ip6_src.s6_addr16[1];
549 		ip6->ip6_src.s6_addr16[1] = 0;
550 	}
551 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) {
552 		if (ifp->if_flags & IFF_LOOPBACK)
553 			dst_scope = ip6->ip6_dst.s6_addr16[1];
554 		ip6->ip6_dst.s6_addr16[1] = 0;
555 	}
556 
557 	/* Determine path MTU. */
558 	if ((error = ip6_getpmtu(ro_pmtu->ro_rt, ifp, &mtu, &alwaysfrag)) != 0)
559 		goto bad;
560 
561 	/*
562 	 * The caller of this function may specify to use the minimum MTU
563 	 * in some cases.
564 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
565 	 * setting.  The logic is a bit complicated; by default, unicast
566 	 * packets will follow path MTU while multicast packets will be sent at
567 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
568 	 * including unicast ones will be sent at the minimum MTU.  Multicast
569 	 * packets will always be sent at the minimum MTU unless
570 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
571 	 * See RFC 3542 for more details.
572 	 */
573 	if (mtu > IPV6_MMTU) {
574 		if ((flags & IPV6_MINMTU))
575 			mtu = IPV6_MMTU;
576 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
577 			mtu = IPV6_MMTU;
578 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
579 			 (opt == NULL ||
580 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
581 			mtu = IPV6_MMTU;
582 		}
583 	}
584 
585 	/*
586 	 * If the outgoing packet contains a hop-by-hop options header,
587 	 * it must be examined and processed even by the source node.
588 	 * (RFC 2460, section 4.)
589 	 */
590 	if (exthdrs.ip6e_hbh) {
591 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
592 		u_int32_t dummy1; /* XXX unused */
593 		u_int32_t dummy2; /* XXX unused */
594 
595 		/*
596 		 *  XXX: if we have to send an ICMPv6 error to the sender,
597 		 *       we need the M_LOOP flag since icmp6_error() expects
598 		 *       the IPv6 and the hop-by-hop options header are
599 		 *       continuous unless the flag is set.
600 		 */
601 		m->m_flags |= M_LOOP;
602 		m->m_pkthdr.ph_ifidx = ifp->if_index;
603 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
604 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
605 		    &dummy1, &dummy2) < 0) {
606 			/* m was already freed at this point */
607 			error = EINVAL;/* better error? */
608 			goto done;
609 		}
610 		m->m_flags &= ~M_LOOP; /* XXX */
611 		m->m_pkthdr.ph_ifidx = 0;
612 	}
613 
614 #if NPF > 0
615 	if (pf_test(AF_INET6, PF_OUT, ifp, &m) != PF_PASS) {
616 		error = EHOSTUNREACH;
617 		m_freem(m);
618 		goto done;
619 	}
620 	if (m == NULL)
621 		goto done;
622 	ip6 = mtod(m, struct ip6_hdr *);
623 	if ((m->m_pkthdr.pf.flags & (PF_TAG_REROUTE | PF_TAG_GENERATED)) ==
624 	    (PF_TAG_REROUTE | PF_TAG_GENERATED)) {
625 		/* already rerun the route lookup, go on */
626 		m->m_pkthdr.pf.flags &= ~(PF_TAG_GENERATED | PF_TAG_REROUTE);
627 	} else if (m->m_pkthdr.pf.flags & PF_TAG_REROUTE) {
628 		/* tag as generated to skip over pf_test on rerun */
629 		m->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
630 		finaldst = ip6->ip6_dst;
631 		ro = NULL;
632 		if_put(ifp); /* drop reference since destination changed */
633 		ifp = NULL;
634 		goto reroute;
635 	}
636 #endif
637 
638 	/*
639 	 * If the packet is not going on the wire it can be destinated
640 	 * to any local address.  In this case do not clear its scopes
641 	 * to let ip6_input() find a matching local route.
642 	 */
643 	if (ifp->if_flags & IFF_LOOPBACK) {
644 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
645 			ip6->ip6_src.s6_addr16[1] = src_scope;
646 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
647 			ip6->ip6_dst.s6_addr16[1] = dst_scope;
648 	}
649 
650 	in6_proto_cksum_out(m, ifp);
651 
652 	/*
653 	 * Send the packet to the outgoing interface.
654 	 * If necessary, do IPv6 fragmentation before sending.
655 	 *
656 	 * the logic here is rather complex:
657 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
658 	 * 1-a: send as is if tlen <= path mtu
659 	 * 1-b: fragment if tlen > path mtu
660 	 *
661 	 * 2: if user asks us not to fragment (dontfrag == 1)
662 	 * 2-a: send as is if tlen <= interface mtu
663 	 * 2-b: error if tlen > interface mtu
664 	 *
665 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
666 	 *      always fragment
667 	 *
668 	 * 4: if dontfrag == 1 && alwaysfrag == 1
669 	 *      error, as we cannot handle this conflicting request
670 	 */
671 	tlen = m->m_pkthdr.len;
672 
673 	if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
674 		dontfrag = 1;
675 	else
676 		dontfrag = 0;
677 	if (dontfrag && alwaysfrag) {	/* case 4 */
678 		/* conflicting request - can't transmit */
679 		error = EMSGSIZE;
680 		goto bad;
681 	}
682 	if (dontfrag && tlen > ifp->if_mtu) {	/* case 2-b */
683 		error = EMSGSIZE;
684 		goto bad;
685 	}
686 
687 	/*
688 	 * transmit packet without fragmentation
689 	 */
690 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
691 		error = ifp->if_output(ifp, m, sin6tosa(dst), ro->ro_rt);
692 		goto done;
693 	}
694 
695 	/*
696 	 * try to fragment the packet.  case 1-b and 3
697 	 */
698 	if (mtu < IPV6_MMTU) {
699 		/* path MTU cannot be less than IPV6_MMTU */
700 		error = EMSGSIZE;
701 		goto bad;
702 	} else if (ip6->ip6_plen == 0) {
703 		/* jumbo payload cannot be fragmented */
704 		error = EMSGSIZE;
705 		goto bad;
706 	} else {
707 		u_char nextproto;
708 #if 0
709 		struct ip6ctlparam ip6cp;
710 		u_int32_t mtu32;
711 #endif
712 
713 		/*
714 		 * Too large for the destination or interface;
715 		 * fragment if possible.
716 		 * Must be able to put at least 8 bytes per fragment.
717 		 */
718 		hlen = unfragpartlen;
719 		if (mtu > IPV6_MAXPACKET)
720 			mtu = IPV6_MAXPACKET;
721 
722 #if 0
723 		/* Notify a proper path MTU to applications. */
724 		mtu32 = (u_int32_t)mtu;
725 		bzero(&ip6cp, sizeof(ip6cp));
726 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
727 		pfctlinput2(PRC_MSGSIZE, sin6tosa(&ro_pmtu->ro_dst),
728 		    (void *)&ip6cp);
729 #endif
730 
731 		/*
732 		 * Change the next header field of the last header in the
733 		 * unfragmentable part.
734 		 */
735 		if (exthdrs.ip6e_rthdr) {
736 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
737 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
738 		} else if (exthdrs.ip6e_dest1) {
739 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
740 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
741 		} else if (exthdrs.ip6e_hbh) {
742 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
743 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
744 		} else {
745 			nextproto = ip6->ip6_nxt;
746 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
747 		}
748 
749 		m0 = m;
750 		error = ip6_fragment(m0, hlen, nextproto, mtu);
751 		if (error)
752 			ip6stat.ip6s_odropped++;
753 	}
754 
755 	/*
756 	 * Remove leading garbages.
757 	 */
758 	m = m0->m_nextpkt;
759 	m0->m_nextpkt = 0;
760 	m_freem(m0);
761 	for (m0 = m; m; m = m0) {
762 		m0 = m->m_nextpkt;
763 		m->m_nextpkt = 0;
764 		if (error == 0) {
765 			ip6stat.ip6s_ofragments++;
766 			error = ifp->if_output(ifp, m, sin6tosa(dst),
767 			    ro->ro_rt);
768 		} else
769 			m_freem(m);
770 	}
771 
772 	if (error == 0)
773 		ip6stat.ip6s_fragmented++;
774 
775 done:
776 	if_put(ifp);
777 	if (ro == &ip6route && ro->ro_rt) {
778 		rtfree(ro->ro_rt);
779 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
780 		rtfree(ro_pmtu->ro_rt);
781 	}
782 
783 	return (error);
784 
785 freehdrs:
786 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
787 	m_freem(exthdrs.ip6e_dest1);
788 	m_freem(exthdrs.ip6e_rthdr);
789 	m_freem(exthdrs.ip6e_dest2);
790 	/* FALLTHROUGH */
791 bad:
792 	m_freem(m);
793 	goto done;
794 }
795 
796 int
797 ip6_fragment(struct mbuf *m0, int hlen, u_char nextproto, u_long mtu)
798 {
799 	struct mbuf	*m, **mnext, *m_frgpart;
800 	struct ip6_hdr	*mhip6;
801 	struct ip6_frag	*ip6f;
802 	u_int32_t	 id;
803 	int		 tlen, len, off;
804 	int		 error;
805 
806 	id = htonl(ip6_randomid());
807 
808 	mnext = &m0->m_nextpkt;
809 	*mnext = NULL;
810 
811 	tlen = m0->m_pkthdr.len;
812 	len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
813 	if (len < 8)
814 		return (EMSGSIZE);
815 
816 	/*
817 	 * Loop through length of segment after first fragment,
818 	 * make new header and copy data of each part and link onto
819 	 * chain.
820 	 */
821 	for (off = hlen; off < tlen; off += len) {
822 		struct mbuf *mlast;
823 
824 		if ((m = m_gethdr(M_DONTWAIT, MT_HEADER)) == NULL)
825 			return (ENOBUFS);
826 		*mnext = m;
827 		mnext = &m->m_nextpkt;
828 		if ((error = m_dup_pkthdr(m, m0, M_DONTWAIT)) != 0)
829 			return (error);
830 		m->m_data += max_linkhdr;
831 		mhip6 = mtod(m, struct ip6_hdr *);
832 		*mhip6 = *mtod(m0, struct ip6_hdr *);
833 		m->m_len = sizeof(*mhip6);
834 		if ((error = ip6_insertfraghdr(m0, m, hlen, &ip6f)) != 0)
835 			return (error);
836 		ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7));
837 		if (off + len >= tlen)
838 			len = tlen - off;
839 		else
840 			ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
841 		mhip6->ip6_plen = htons((u_int16_t)(len + hlen +
842 		    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
843 		if ((m_frgpart = m_copym(m0, off, len, M_DONTWAIT)) == NULL)
844 			return (ENOBUFS);
845 		for (mlast = m; mlast->m_next; mlast = mlast->m_next)
846 			;
847 		mlast->m_next = m_frgpart;
848 		m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
849 		ip6f->ip6f_reserved = 0;
850 		ip6f->ip6f_ident = id;
851 		ip6f->ip6f_nxt = nextproto;
852 	}
853 
854 	return (0);
855 }
856 
857 int
858 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
859 {
860 	struct mbuf *m;
861 
862 	if (hlen > MCLBYTES)
863 		return (ENOBUFS); /* XXX */
864 
865 	MGET(m, M_DONTWAIT, MT_DATA);
866 	if (!m)
867 		return (ENOBUFS);
868 
869 	if (hlen > MLEN) {
870 		MCLGET(m, M_DONTWAIT);
871 		if ((m->m_flags & M_EXT) == 0) {
872 			m_free(m);
873 			return (ENOBUFS);
874 		}
875 	}
876 	m->m_len = hlen;
877 	if (hdr)
878 		bcopy(hdr, mtod(m, caddr_t), hlen);
879 
880 	*mp = m;
881 	return (0);
882 }
883 
884 /*
885  * Insert jumbo payload option.
886  */
887 int
888 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
889 {
890 	struct mbuf *mopt;
891 	u_int8_t *optbuf;
892 	u_int32_t v;
893 
894 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
895 
896 	/*
897 	 * If there is no hop-by-hop options header, allocate new one.
898 	 * If there is one but it doesn't have enough space to store the
899 	 * jumbo payload option, allocate a cluster to store the whole options.
900 	 * Otherwise, use it to store the options.
901 	 */
902 	if (exthdrs->ip6e_hbh == 0) {
903 		MGET(mopt, M_DONTWAIT, MT_DATA);
904 		if (mopt == NULL)
905 			return (ENOBUFS);
906 		mopt->m_len = JUMBOOPTLEN;
907 		optbuf = mtod(mopt, u_int8_t *);
908 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
909 		exthdrs->ip6e_hbh = mopt;
910 	} else {
911 		struct ip6_hbh *hbh;
912 
913 		mopt = exthdrs->ip6e_hbh;
914 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
915 			/*
916 			 * XXX assumption:
917 			 * - exthdrs->ip6e_hbh is not referenced from places
918 			 *   other than exthdrs.
919 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
920 			 */
921 			int oldoptlen = mopt->m_len;
922 			struct mbuf *n;
923 
924 			/*
925 			 * XXX: give up if the whole (new) hbh header does
926 			 * not fit even in an mbuf cluster.
927 			 */
928 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
929 				return (ENOBUFS);
930 
931 			/*
932 			 * As a consequence, we must always prepare a cluster
933 			 * at this point.
934 			 */
935 			MGET(n, M_DONTWAIT, MT_DATA);
936 			if (n) {
937 				MCLGET(n, M_DONTWAIT);
938 				if ((n->m_flags & M_EXT) == 0) {
939 					m_freem(n);
940 					n = NULL;
941 				}
942 			}
943 			if (!n)
944 				return (ENOBUFS);
945 			n->m_len = oldoptlen + JUMBOOPTLEN;
946 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
947 			      oldoptlen);
948 			optbuf = mtod(n, u_int8_t *) + oldoptlen;
949 			m_freem(mopt);
950 			mopt = exthdrs->ip6e_hbh = n;
951 		} else {
952 			optbuf = mtod(mopt, u_int8_t *) + mopt->m_len;
953 			mopt->m_len += JUMBOOPTLEN;
954 		}
955 		optbuf[0] = IP6OPT_PADN;
956 		optbuf[1] = 0;
957 
958 		/*
959 		 * Adjust the header length according to the pad and
960 		 * the jumbo payload option.
961 		 */
962 		hbh = mtod(mopt, struct ip6_hbh *);
963 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
964 	}
965 
966 	/* fill in the option. */
967 	optbuf[2] = IP6OPT_JUMBO;
968 	optbuf[3] = 4;
969 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
970 	memcpy(&optbuf[4], &v, sizeof(u_int32_t));
971 
972 	/* finally, adjust the packet header length */
973 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
974 
975 	return (0);
976 #undef JUMBOOPTLEN
977 }
978 
979 /*
980  * Insert fragment header and copy unfragmentable header portions.
981  */
982 int
983 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
984     struct ip6_frag **frghdrp)
985 {
986 	struct mbuf *n, *mlast;
987 
988 	if (hlen > sizeof(struct ip6_hdr)) {
989 		n = m_copym(m0, sizeof(struct ip6_hdr),
990 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
991 		if (n == NULL)
992 			return (ENOBUFS);
993 		m->m_next = n;
994 	} else
995 		n = m;
996 
997 	/* Search for the last mbuf of unfragmentable part. */
998 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
999 		;
1000 
1001 	if ((mlast->m_flags & M_EXT) == 0 &&
1002 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1003 		/* use the trailing space of the last mbuf for the fragment hdr */
1004 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1005 		    mlast->m_len);
1006 		mlast->m_len += sizeof(struct ip6_frag);
1007 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1008 	} else {
1009 		/* allocate a new mbuf for the fragment header */
1010 		struct mbuf *mfrg;
1011 
1012 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1013 		if (mfrg == NULL)
1014 			return (ENOBUFS);
1015 		mfrg->m_len = sizeof(struct ip6_frag);
1016 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1017 		mlast->m_next = mfrg;
1018 	}
1019 
1020 	return (0);
1021 }
1022 
1023 int
1024 ip6_getpmtu(struct rtentry *rt, struct ifnet *ifp, u_long *mtup,
1025     int *alwaysfragp)
1026 {
1027 	u_int32_t mtu = 0;
1028 	int alwaysfrag = 0;
1029 	int error = 0;
1030 
1031 	if (rt != NULL) {
1032 		mtu = rt->rt_rmx.rmx_mtu;
1033 		if (mtu == 0)
1034 			mtu = ifp->if_mtu;
1035 		else if (mtu < IPV6_MMTU) {
1036 			/*
1037 			 * RFC2460 section 5, last paragraph:
1038 			 * if we record ICMPv6 too big message with
1039 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1040 			 * or smaller, with fragment header attached.
1041 			 * (fragment header is needed regardless from the
1042 			 * packet size, for translators to identify packets)
1043 			 */
1044 			alwaysfrag = 1;
1045 			mtu = IPV6_MMTU;
1046 		} else if (mtu > ifp->if_mtu) {
1047 			/*
1048 			 * The MTU on the route is larger than the MTU on
1049 			 * the interface!  This shouldn't happen, unless the
1050 			 * MTU of the interface has been changed after the
1051 			 * interface was brought up.  Change the MTU in the
1052 			 * route to match the interface MTU (as long as the
1053 			 * field isn't locked).
1054 			 */
1055 			mtu = ifp->if_mtu;
1056 			if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
1057 				rt->rt_rmx.rmx_mtu = mtu;
1058 		}
1059 	} else {
1060 		mtu = ifp->if_mtu;
1061 	}
1062 
1063 	*mtup = mtu;
1064 	if (alwaysfragp)
1065 		*alwaysfragp = alwaysfrag;
1066 	return (error);
1067 }
1068 
1069 /*
1070  * IP6 socket option processing.
1071  */
1072 int
1073 ip6_ctloutput(int op, struct socket *so, int level, int optname,
1074     struct mbuf **mp)
1075 {
1076 	int privileged, optdatalen, uproto;
1077 	void *optdata;
1078 	struct inpcb *inp = sotoinpcb(so);
1079 	struct mbuf *m = *mp;
1080 	int error, optval;
1081 	struct proc *p = curproc; /* For IPSec and rdomain */
1082 	u_int rtid = 0;
1083 
1084 	error = optval = 0;
1085 
1086 	privileged = (inp->inp_socket->so_state & SS_PRIV);
1087 	uproto = (int)so->so_proto->pr_protocol;
1088 
1089 	if (level == IPPROTO_IPV6) {
1090 		switch (op) {
1091 		case PRCO_SETOPT:
1092 			switch (optname) {
1093 			/*
1094 			 * Use of some Hop-by-Hop options or some
1095 			 * Destination options, might require special
1096 			 * privilege.  That is, normal applications
1097 			 * (without special privilege) might be forbidden
1098 			 * from setting certain options in outgoing packets,
1099 			 * and might never see certain options in received
1100 			 * packets. [RFC 2292 Section 6]
1101 			 * KAME specific note:
1102 			 *  KAME prevents non-privileged users from sending or
1103 			 *  receiving ANY hbh/dst options in order to avoid
1104 			 *  overhead of parsing options in the kernel.
1105 			 */
1106 			case IPV6_RECVHOPOPTS:
1107 			case IPV6_RECVDSTOPTS:
1108 				if (!privileged) {
1109 					error = EPERM;
1110 					break;
1111 				}
1112 				/* FALLTHROUGH */
1113 			case IPV6_UNICAST_HOPS:
1114 			case IPV6_MINHOPCOUNT:
1115 			case IPV6_HOPLIMIT:
1116 
1117 			case IPV6_RECVPKTINFO:
1118 			case IPV6_RECVHOPLIMIT:
1119 			case IPV6_RECVRTHDR:
1120 			case IPV6_RECVPATHMTU:
1121 			case IPV6_RECVTCLASS:
1122 			case IPV6_V6ONLY:
1123 			case IPV6_AUTOFLOWLABEL:
1124 			case IPV6_RECVDSTPORT:
1125 				if (m == NULL || m->m_len != sizeof(int)) {
1126 					error = EINVAL;
1127 					break;
1128 				}
1129 				optval = *mtod(m, int *);
1130 				switch (optname) {
1131 
1132 				case IPV6_UNICAST_HOPS:
1133 					if (optval < -1 || optval >= 256)
1134 						error = EINVAL;
1135 					else {
1136 						/* -1 = kernel default */
1137 						inp->inp_hops = optval;
1138 					}
1139 					break;
1140 
1141 				case IPV6_MINHOPCOUNT:
1142 					if (optval < 0 || optval > 255)
1143 						error = EINVAL;
1144 					else
1145 						inp->inp_ip6_minhlim = optval;
1146 					break;
1147 
1148 #define OPTSET(bit) \
1149 do { \
1150 	if (optval) \
1151 		inp->inp_flags |= (bit); \
1152 	else \
1153 		inp->inp_flags &= ~(bit); \
1154 } while (/*CONSTCOND*/ 0)
1155 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1156 
1157 				case IPV6_RECVPKTINFO:
1158 					OPTSET(IN6P_PKTINFO);
1159 					break;
1160 
1161 				case IPV6_HOPLIMIT:
1162 				{
1163 					struct ip6_pktopts **optp;
1164 
1165 					optp = &inp->inp_outputopts6;
1166 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1167 							   (u_char *)&optval,
1168 							   sizeof(optval),
1169 							   optp,
1170 							   privileged, uproto);
1171 					break;
1172 				}
1173 
1174 				case IPV6_RECVHOPLIMIT:
1175 					OPTSET(IN6P_HOPLIMIT);
1176 					break;
1177 
1178 				case IPV6_RECVHOPOPTS:
1179 					OPTSET(IN6P_HOPOPTS);
1180 					break;
1181 
1182 				case IPV6_RECVDSTOPTS:
1183 					OPTSET(IN6P_DSTOPTS);
1184 					break;
1185 
1186 				case IPV6_RECVRTHDR:
1187 					OPTSET(IN6P_RTHDR);
1188 					break;
1189 
1190 				case IPV6_RECVPATHMTU:
1191 					/*
1192 					 * We ignore this option for TCP
1193 					 * sockets.
1194 					 * (RFC3542 leaves this case
1195 					 * unspecified.)
1196 					 */
1197 					if (uproto != IPPROTO_TCP)
1198 						OPTSET(IN6P_MTU);
1199 					break;
1200 
1201 				case IPV6_V6ONLY:
1202 					/*
1203 					 * make setsockopt(IPV6_V6ONLY)
1204 					 * available only prior to bind(2).
1205 					 * see ipng mailing list, Jun 22 2001.
1206 					 */
1207 					if (inp->inp_lport ||
1208 					    !IN6_IS_ADDR_UNSPECIFIED(&inp->inp_laddr6)) {
1209 						error = EINVAL;
1210 						break;
1211 					}
1212 					/* No support for IPv4-mapped addresses. */
1213 					if (!optval)
1214 						error = EINVAL;
1215 					else
1216 						error = 0;
1217 					break;
1218 				case IPV6_RECVTCLASS:
1219 					OPTSET(IN6P_TCLASS);
1220 					break;
1221 				case IPV6_AUTOFLOWLABEL:
1222 					OPTSET(IN6P_AUTOFLOWLABEL);
1223 					break;
1224 
1225 				case IPV6_RECVDSTPORT:
1226 					OPTSET(IN6P_RECVDSTPORT);
1227 					break;
1228 				}
1229 				break;
1230 
1231 			case IPV6_TCLASS:
1232 			case IPV6_DONTFRAG:
1233 			case IPV6_USE_MIN_MTU:
1234 				if (m == NULL || m->m_len != sizeof(optval)) {
1235 					error = EINVAL;
1236 					break;
1237 				}
1238 				optval = *mtod(m, int *);
1239 				{
1240 					struct ip6_pktopts **optp;
1241 					optp = &inp->inp_outputopts6;
1242 					error = ip6_pcbopt(optname,
1243 							   (u_char *)&optval,
1244 							   sizeof(optval),
1245 							   optp,
1246 							   privileged, uproto);
1247 					break;
1248 				}
1249 
1250 			case IPV6_PKTINFO:
1251 			case IPV6_HOPOPTS:
1252 			case IPV6_RTHDR:
1253 			case IPV6_DSTOPTS:
1254 			case IPV6_RTHDRDSTOPTS:
1255 			{
1256 				/* new advanced API (RFC3542) */
1257 				u_char *optbuf;
1258 				int optbuflen;
1259 				struct ip6_pktopts **optp;
1260 
1261 				if (m && m->m_next) {
1262 					error = EINVAL;	/* XXX */
1263 					break;
1264 				}
1265 				if (m) {
1266 					optbuf = mtod(m, u_char *);
1267 					optbuflen = m->m_len;
1268 				} else {
1269 					optbuf = NULL;
1270 					optbuflen = 0;
1271 				}
1272 				optp = &inp->inp_outputopts6;
1273 				error = ip6_pcbopt(optname,
1274 						   optbuf, optbuflen,
1275 						   optp, privileged, uproto);
1276 				break;
1277 			}
1278 #undef OPTSET
1279 
1280 			case IPV6_MULTICAST_IF:
1281 			case IPV6_MULTICAST_HOPS:
1282 			case IPV6_MULTICAST_LOOP:
1283 			case IPV6_JOIN_GROUP:
1284 			case IPV6_LEAVE_GROUP:
1285 				error =	ip6_setmoptions(optname,
1286 							&inp->inp_moptions6,
1287 							m);
1288 				break;
1289 
1290 			case IPV6_PORTRANGE:
1291 				if (m == NULL || m->m_len != sizeof(int)) {
1292 					error = EINVAL;
1293 					break;
1294 				}
1295 				optval = *mtod(m, int *);
1296 
1297 				switch (optval) {
1298 				case IPV6_PORTRANGE_DEFAULT:
1299 					inp->inp_flags &= ~(IN6P_LOWPORT);
1300 					inp->inp_flags &= ~(IN6P_HIGHPORT);
1301 					break;
1302 
1303 				case IPV6_PORTRANGE_HIGH:
1304 					inp->inp_flags &= ~(IN6P_LOWPORT);
1305 					inp->inp_flags |= IN6P_HIGHPORT;
1306 					break;
1307 
1308 				case IPV6_PORTRANGE_LOW:
1309 					inp->inp_flags &= ~(IN6P_HIGHPORT);
1310 					inp->inp_flags |= IN6P_LOWPORT;
1311 					break;
1312 
1313 				default:
1314 					error = EINVAL;
1315 					break;
1316 				}
1317 				break;
1318 
1319 			case IPSEC6_OUTSA:
1320 				error = EINVAL;
1321 				break;
1322 
1323 			case IPV6_AUTH_LEVEL:
1324 			case IPV6_ESP_TRANS_LEVEL:
1325 			case IPV6_ESP_NETWORK_LEVEL:
1326 			case IPV6_IPCOMP_LEVEL:
1327 #ifndef IPSEC
1328 				error = EINVAL;
1329 #else
1330 				if (m == NULL || m->m_len != sizeof(int)) {
1331 					error = EINVAL;
1332 					break;
1333 				}
1334 				optval = *mtod(m, int *);
1335 
1336 				if (optval < IPSEC_LEVEL_BYPASS ||
1337 				    optval > IPSEC_LEVEL_UNIQUE) {
1338 					error = EINVAL;
1339 					break;
1340 				}
1341 
1342 				switch (optname) {
1343 				case IPV6_AUTH_LEVEL:
1344 				        if (optval < IPSEC_AUTH_LEVEL_DEFAULT &&
1345 					    suser(p, 0)) {
1346 						error = EACCES;
1347 						break;
1348 					}
1349 					inp->inp_seclevel[SL_AUTH] = optval;
1350 					break;
1351 
1352 				case IPV6_ESP_TRANS_LEVEL:
1353 				        if (optval < IPSEC_ESP_TRANS_LEVEL_DEFAULT &&
1354 					    suser(p, 0)) {
1355 						error = EACCES;
1356 						break;
1357 					}
1358 					inp->inp_seclevel[SL_ESP_TRANS] = optval;
1359 					break;
1360 
1361 				case IPV6_ESP_NETWORK_LEVEL:
1362 				        if (optval < IPSEC_ESP_NETWORK_LEVEL_DEFAULT &&
1363 					    suser(p, 0)) {
1364 						error = EACCES;
1365 						break;
1366 					}
1367 					inp->inp_seclevel[SL_ESP_NETWORK] = optval;
1368 					break;
1369 
1370 				case IPV6_IPCOMP_LEVEL:
1371 				        if (optval < IPSEC_IPCOMP_LEVEL_DEFAULT &&
1372 					    suser(p, 0)) {
1373 						error = EACCES;
1374 						break;
1375 					}
1376 					inp->inp_seclevel[SL_IPCOMP] = optval;
1377 					break;
1378 				}
1379 #endif
1380 				break;
1381 			case SO_RTABLE:
1382 				if (m == NULL || m->m_len < sizeof(u_int)) {
1383 					error = EINVAL;
1384 					break;
1385 				}
1386 				rtid = *mtod(m, u_int *);
1387 				if (inp->inp_rtableid == rtid)
1388 					break;
1389 				/* needs privileges to switch when already set */
1390 				if (p->p_p->ps_rtableid != rtid &&
1391 				    p->p_p->ps_rtableid != 0 &&
1392 				    (error = suser(p, 0)) != 0)
1393 					break;
1394 				/* table must exist */
1395 				if (!rtable_exists(rtid)) {
1396 					error = EINVAL;
1397 					break;
1398 				}
1399 				if (inp->inp_lport) {
1400 					error = EBUSY;
1401 					break;
1402 				}
1403 				inp->inp_rtableid = rtid;
1404 				in_pcbrehash(inp);
1405 				break;
1406 			case IPV6_PIPEX:
1407 				if (m != NULL && m->m_len == sizeof(int))
1408 					inp->inp_pipex = *mtod(m, int *);
1409 				else
1410 					error = EINVAL;
1411 				break;
1412 
1413 			default:
1414 				error = ENOPROTOOPT;
1415 				break;
1416 			}
1417 			if (m)
1418 				(void)m_free(m);
1419 			break;
1420 
1421 		case PRCO_GETOPT:
1422 			switch (optname) {
1423 
1424 			case IPV6_RECVHOPOPTS:
1425 			case IPV6_RECVDSTOPTS:
1426 			case IPV6_UNICAST_HOPS:
1427 			case IPV6_MINHOPCOUNT:
1428 			case IPV6_RECVPKTINFO:
1429 			case IPV6_RECVHOPLIMIT:
1430 			case IPV6_RECVRTHDR:
1431 			case IPV6_RECVPATHMTU:
1432 
1433 			case IPV6_V6ONLY:
1434 			case IPV6_PORTRANGE:
1435 			case IPV6_RECVTCLASS:
1436 			case IPV6_AUTOFLOWLABEL:
1437 			case IPV6_RECVDSTPORT:
1438 				switch (optname) {
1439 
1440 				case IPV6_RECVHOPOPTS:
1441 					optval = OPTBIT(IN6P_HOPOPTS);
1442 					break;
1443 
1444 				case IPV6_RECVDSTOPTS:
1445 					optval = OPTBIT(IN6P_DSTOPTS);
1446 					break;
1447 
1448 				case IPV6_UNICAST_HOPS:
1449 					optval = inp->inp_hops;
1450 					break;
1451 
1452 				case IPV6_MINHOPCOUNT:
1453 					optval = inp->inp_ip6_minhlim;
1454 					break;
1455 
1456 				case IPV6_RECVPKTINFO:
1457 					optval = OPTBIT(IN6P_PKTINFO);
1458 					break;
1459 
1460 				case IPV6_RECVHOPLIMIT:
1461 					optval = OPTBIT(IN6P_HOPLIMIT);
1462 					break;
1463 
1464 				case IPV6_RECVRTHDR:
1465 					optval = OPTBIT(IN6P_RTHDR);
1466 					break;
1467 
1468 				case IPV6_RECVPATHMTU:
1469 					optval = OPTBIT(IN6P_MTU);
1470 					break;
1471 
1472 				case IPV6_V6ONLY:
1473 					optval = 1;
1474 					break;
1475 
1476 				case IPV6_PORTRANGE:
1477 				    {
1478 					int flags;
1479 					flags = inp->inp_flags;
1480 					if (flags & IN6P_HIGHPORT)
1481 						optval = IPV6_PORTRANGE_HIGH;
1482 					else if (flags & IN6P_LOWPORT)
1483 						optval = IPV6_PORTRANGE_LOW;
1484 					else
1485 						optval = 0;
1486 					break;
1487 				    }
1488 				case IPV6_RECVTCLASS:
1489 					optval = OPTBIT(IN6P_TCLASS);
1490 					break;
1491 
1492 				case IPV6_AUTOFLOWLABEL:
1493 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1494 					break;
1495 
1496 				case IPV6_RECVDSTPORT:
1497 					optval = OPTBIT(IN6P_RECVDSTPORT);
1498 					break;
1499 				}
1500 				if (error)
1501 					break;
1502 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1503 				m->m_len = sizeof(int);
1504 				*mtod(m, int *) = optval;
1505 				break;
1506 
1507 			case IPV6_PATHMTU:
1508 			{
1509 				u_long pmtu = 0;
1510 				struct ip6_mtuinfo mtuinfo;
1511 				struct ifnet *ifp;
1512 				struct rtentry *rt;
1513 
1514 				if (!(so->so_state & SS_ISCONNECTED))
1515 					return (ENOTCONN);
1516 
1517 				rt = in_pcbrtentry(inp);
1518 				if (!rtisvalid(rt))
1519 					return (EHOSTUNREACH);
1520 
1521 				ifp = if_get(rt->rt_ifidx);
1522 				if (ifp == NULL)
1523 					return (EHOSTUNREACH);
1524 				/*
1525 				 * XXX: we dot not consider the case of source
1526 				 * routing, or optional information to specify
1527 				 * the outgoing interface.
1528 				 */
1529 				error = ip6_getpmtu(rt, ifp, &pmtu, NULL);
1530 				if_put(ifp);
1531 				if (error)
1532 					break;
1533 				if (pmtu > IPV6_MAXPACKET)
1534 					pmtu = IPV6_MAXPACKET;
1535 
1536 				bzero(&mtuinfo, sizeof(mtuinfo));
1537 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1538 				optdata = (void *)&mtuinfo;
1539 				optdatalen = sizeof(mtuinfo);
1540 				if (optdatalen > MCLBYTES)
1541 					return (EMSGSIZE); /* XXX */
1542 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1543 				if (optdatalen > MLEN)
1544 					MCLGET(m, M_WAIT);
1545 				m->m_len = optdatalen;
1546 				bcopy(optdata, mtod(m, void *), optdatalen);
1547 				break;
1548 			}
1549 
1550 			case IPV6_PKTINFO:
1551 			case IPV6_HOPOPTS:
1552 			case IPV6_RTHDR:
1553 			case IPV6_DSTOPTS:
1554 			case IPV6_RTHDRDSTOPTS:
1555 			case IPV6_TCLASS:
1556 			case IPV6_DONTFRAG:
1557 			case IPV6_USE_MIN_MTU:
1558 				error = ip6_getpcbopt(inp->inp_outputopts6,
1559 				    optname, mp);
1560 				break;
1561 
1562 			case IPV6_MULTICAST_IF:
1563 			case IPV6_MULTICAST_HOPS:
1564 			case IPV6_MULTICAST_LOOP:
1565 			case IPV6_JOIN_GROUP:
1566 			case IPV6_LEAVE_GROUP:
1567 				error = ip6_getmoptions(optname,
1568 				    inp->inp_moptions6, mp);
1569 				break;
1570 
1571 			case IPSEC6_OUTSA:
1572 				error = EINVAL;
1573 				break;
1574 
1575 			case IPV6_AUTH_LEVEL:
1576 			case IPV6_ESP_TRANS_LEVEL:
1577 			case IPV6_ESP_NETWORK_LEVEL:
1578 			case IPV6_IPCOMP_LEVEL:
1579 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1580 #ifndef IPSEC
1581 				m->m_len = sizeof(int);
1582 				*mtod(m, int *) = IPSEC_LEVEL_NONE;
1583 #else
1584 				m->m_len = sizeof(int);
1585 				switch (optname) {
1586 				case IPV6_AUTH_LEVEL:
1587 					optval = inp->inp_seclevel[SL_AUTH];
1588 					break;
1589 
1590 				case IPV6_ESP_TRANS_LEVEL:
1591 					optval =
1592 					    inp->inp_seclevel[SL_ESP_TRANS];
1593 					break;
1594 
1595 				case IPV6_ESP_NETWORK_LEVEL:
1596 					optval =
1597 					    inp->inp_seclevel[SL_ESP_NETWORK];
1598 					break;
1599 
1600 				case IPV6_IPCOMP_LEVEL:
1601 					optval = inp->inp_seclevel[SL_IPCOMP];
1602 					break;
1603 				}
1604 				*mtod(m, int *) = optval;
1605 #endif
1606 				break;
1607 			case SO_RTABLE:
1608 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1609 				m->m_len = sizeof(u_int);
1610 				*mtod(m, u_int *) = optval;
1611 				break;
1612 			case IPV6_PIPEX:
1613 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1614 				m->m_len = sizeof(int);
1615 				*mtod(m, int *) = optval;
1616 				break;
1617 
1618 			default:
1619 				error = ENOPROTOOPT;
1620 				break;
1621 			}
1622 			break;
1623 		}
1624 	} else {
1625 		error = EINVAL;
1626 		if (op == PRCO_SETOPT)
1627 			(void)m_free(*mp);
1628 	}
1629 	return (error);
1630 }
1631 
1632 int
1633 ip6_raw_ctloutput(int op, struct socket *so, int level, int optname,
1634     struct mbuf **mp)
1635 {
1636 	int error = 0, optval;
1637 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
1638 	struct inpcb *inp = sotoinpcb(so);
1639 	struct mbuf *m = *mp;
1640 
1641 	if (level != IPPROTO_IPV6) {
1642 		if (op == PRCO_SETOPT)
1643 			(void)m_free(*mp);
1644 		return (EINVAL);
1645 	}
1646 
1647 	switch (optname) {
1648 	case IPV6_CHECKSUM:
1649 		/*
1650 		 * For ICMPv6 sockets, no modification allowed for checksum
1651 		 * offset, permit "no change" values to help existing apps.
1652 		 *
1653 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
1654 		 * for an ICMPv6 socket will fail."
1655 		 * The current behavior does not meet RFC3542.
1656 		 */
1657 		switch (op) {
1658 		case PRCO_SETOPT:
1659 			if (m == NULL || m->m_len != sizeof(int)) {
1660 				error = EINVAL;
1661 				break;
1662 			}
1663 			optval = *mtod(m, int *);
1664 			if ((optval % 2) != 0) {
1665 				/* the API assumes even offset values */
1666 				error = EINVAL;
1667 			} else if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) {
1668 				if (optval != icmp6off)
1669 					error = EINVAL;
1670 			} else
1671 				inp->inp_cksum6 = optval;
1672 			break;
1673 
1674 		case PRCO_GETOPT:
1675 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
1676 				optval = icmp6off;
1677 			else
1678 				optval = inp->inp_cksum6;
1679 
1680 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1681 			m->m_len = sizeof(int);
1682 			*mtod(m, int *) = optval;
1683 			break;
1684 
1685 		default:
1686 			error = EINVAL;
1687 			break;
1688 		}
1689 		break;
1690 
1691 	default:
1692 		error = ENOPROTOOPT;
1693 		break;
1694 	}
1695 
1696 	if (op == PRCO_SETOPT)
1697 		(void)m_free(m);
1698 
1699 	return (error);
1700 }
1701 
1702 /*
1703  * initialize ip6_pktopts.  beware that there are non-zero default values in
1704  * the struct.
1705  */
1706 void
1707 ip6_initpktopts(struct ip6_pktopts *opt)
1708 {
1709 
1710 	bzero(opt, sizeof(*opt));
1711 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
1712 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
1713 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
1714 }
1715 
1716 int
1717 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
1718     int priv, int uproto)
1719 {
1720 	struct ip6_pktopts *opt;
1721 
1722 	if (*pktopt == NULL) {
1723 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
1724 		    M_WAITOK);
1725 		ip6_initpktopts(*pktopt);
1726 	}
1727 	opt = *pktopt;
1728 
1729 	return (ip6_setpktopt(optname, buf, len, opt, priv, 1, uproto));
1730 }
1731 
1732 int
1733 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct mbuf **mp)
1734 {
1735 	void *optdata = NULL;
1736 	int optdatalen = 0;
1737 	struct ip6_ext *ip6e;
1738 	int error = 0;
1739 	struct in6_pktinfo null_pktinfo;
1740 	int deftclass = 0, on;
1741 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
1742 	struct mbuf *m;
1743 
1744 	switch (optname) {
1745 	case IPV6_PKTINFO:
1746 		if (pktopt && pktopt->ip6po_pktinfo)
1747 			optdata = (void *)pktopt->ip6po_pktinfo;
1748 		else {
1749 			/* XXX: we don't have to do this every time... */
1750 			bzero(&null_pktinfo, sizeof(null_pktinfo));
1751 			optdata = (void *)&null_pktinfo;
1752 		}
1753 		optdatalen = sizeof(struct in6_pktinfo);
1754 		break;
1755 	case IPV6_TCLASS:
1756 		if (pktopt && pktopt->ip6po_tclass >= 0)
1757 			optdata = (void *)&pktopt->ip6po_tclass;
1758 		else
1759 			optdata = (void *)&deftclass;
1760 		optdatalen = sizeof(int);
1761 		break;
1762 	case IPV6_HOPOPTS:
1763 		if (pktopt && pktopt->ip6po_hbh) {
1764 			optdata = (void *)pktopt->ip6po_hbh;
1765 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
1766 			optdatalen = (ip6e->ip6e_len + 1) << 3;
1767 		}
1768 		break;
1769 	case IPV6_RTHDR:
1770 		if (pktopt && pktopt->ip6po_rthdr) {
1771 			optdata = (void *)pktopt->ip6po_rthdr;
1772 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
1773 			optdatalen = (ip6e->ip6e_len + 1) << 3;
1774 		}
1775 		break;
1776 	case IPV6_RTHDRDSTOPTS:
1777 		if (pktopt && pktopt->ip6po_dest1) {
1778 			optdata = (void *)pktopt->ip6po_dest1;
1779 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
1780 			optdatalen = (ip6e->ip6e_len + 1) << 3;
1781 		}
1782 		break;
1783 	case IPV6_DSTOPTS:
1784 		if (pktopt && pktopt->ip6po_dest2) {
1785 			optdata = (void *)pktopt->ip6po_dest2;
1786 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
1787 			optdatalen = (ip6e->ip6e_len + 1) << 3;
1788 		}
1789 		break;
1790 	case IPV6_USE_MIN_MTU:
1791 		if (pktopt)
1792 			optdata = (void *)&pktopt->ip6po_minmtu;
1793 		else
1794 			optdata = (void *)&defminmtu;
1795 		optdatalen = sizeof(int);
1796 		break;
1797 	case IPV6_DONTFRAG:
1798 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
1799 			on = 1;
1800 		else
1801 			on = 0;
1802 		optdata = (void *)&on;
1803 		optdatalen = sizeof(on);
1804 		break;
1805 	default:		/* should not happen */
1806 #ifdef DIAGNOSTIC
1807 		panic("ip6_getpcbopt: unexpected option");
1808 #endif
1809 		return (ENOPROTOOPT);
1810 	}
1811 
1812 	if (optdatalen > MCLBYTES)
1813 		return (EMSGSIZE); /* XXX */
1814 	*mp = m = m_get(M_WAIT, MT_SOOPTS);
1815 	if (optdatalen > MLEN)
1816 		MCLGET(m, M_WAIT);
1817 	m->m_len = optdatalen;
1818 	if (optdatalen)
1819 		bcopy(optdata, mtod(m, void *), optdatalen);
1820 
1821 	return (error);
1822 }
1823 
1824 void
1825 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
1826 {
1827 	if (optname == -1 || optname == IPV6_PKTINFO) {
1828 		if (pktopt->ip6po_pktinfo)
1829 			free(pktopt->ip6po_pktinfo, M_IP6OPT, 0);
1830 		pktopt->ip6po_pktinfo = NULL;
1831 	}
1832 	if (optname == -1 || optname == IPV6_HOPLIMIT)
1833 		pktopt->ip6po_hlim = -1;
1834 	if (optname == -1 || optname == IPV6_TCLASS)
1835 		pktopt->ip6po_tclass = -1;
1836 	if (optname == -1 || optname == IPV6_HOPOPTS) {
1837 		if (pktopt->ip6po_hbh)
1838 			free(pktopt->ip6po_hbh, M_IP6OPT, 0);
1839 		pktopt->ip6po_hbh = NULL;
1840 	}
1841 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
1842 		if (pktopt->ip6po_dest1)
1843 			free(pktopt->ip6po_dest1, M_IP6OPT, 0);
1844 		pktopt->ip6po_dest1 = NULL;
1845 	}
1846 	if (optname == -1 || optname == IPV6_RTHDR) {
1847 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
1848 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT, 0);
1849 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
1850 		if (pktopt->ip6po_route.ro_rt) {
1851 			rtfree(pktopt->ip6po_route.ro_rt);
1852 			pktopt->ip6po_route.ro_rt = NULL;
1853 		}
1854 	}
1855 	if (optname == -1 || optname == IPV6_DSTOPTS) {
1856 		if (pktopt->ip6po_dest2)
1857 			free(pktopt->ip6po_dest2, M_IP6OPT, 0);
1858 		pktopt->ip6po_dest2 = NULL;
1859 	}
1860 }
1861 
1862 #define PKTOPT_EXTHDRCPY(type) \
1863 do {\
1864 	if (src->type) {\
1865 		size_t hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
1866 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
1867 		if (dst->type == NULL && canwait == M_NOWAIT)\
1868 			goto bad;\
1869 		memcpy(dst->type, src->type, hlen);\
1870 	}\
1871 } while (/*CONSTCOND*/ 0)
1872 
1873 int
1874 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
1875 {
1876 	dst->ip6po_hlim = src->ip6po_hlim;
1877 	dst->ip6po_tclass = src->ip6po_tclass;
1878 	dst->ip6po_flags = src->ip6po_flags;
1879 	if (src->ip6po_pktinfo) {
1880 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
1881 		    M_IP6OPT, canwait);
1882 		if (dst->ip6po_pktinfo == NULL)
1883 			goto bad;
1884 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
1885 	}
1886 	PKTOPT_EXTHDRCPY(ip6po_hbh);
1887 	PKTOPT_EXTHDRCPY(ip6po_dest1);
1888 	PKTOPT_EXTHDRCPY(ip6po_dest2);
1889 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
1890 	return (0);
1891 
1892   bad:
1893 	ip6_clearpktopts(dst, -1);
1894 	return (ENOBUFS);
1895 }
1896 #undef PKTOPT_EXTHDRCPY
1897 
1898 void
1899 ip6_freepcbopts(struct ip6_pktopts *pktopt)
1900 {
1901 	if (pktopt == NULL)
1902 		return;
1903 
1904 	ip6_clearpktopts(pktopt, -1);
1905 
1906 	free(pktopt, M_IP6OPT, 0);
1907 }
1908 
1909 /*
1910  * Set the IP6 multicast options in response to user setsockopt().
1911  */
1912 int
1913 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
1914 {
1915 	int error = 0;
1916 	u_int loop, ifindex;
1917 	struct ipv6_mreq *mreq;
1918 	struct ifnet *ifp;
1919 	struct ip6_moptions *im6o = *im6op;
1920 	struct route_in6 ro;
1921 	struct sockaddr_in6 *dst;
1922 	struct in6_multi_mship *imm;
1923 	struct proc *p = curproc;	/* XXX */
1924 
1925 	if (im6o == NULL) {
1926 		/*
1927 		 * No multicast option buffer attached to the pcb;
1928 		 * allocate one and initialize to default values.
1929 		 */
1930 		im6o = (struct ip6_moptions *)
1931 			malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
1932 
1933 		if (im6o == NULL)
1934 			return (ENOBUFS);
1935 		*im6op = im6o;
1936 		im6o->im6o_ifidx = 0;
1937 		im6o->im6o_hlim = ip6_defmcasthlim;
1938 		im6o->im6o_loop = IPV6_DEFAULT_MULTICAST_LOOP;
1939 		LIST_INIT(&im6o->im6o_memberships);
1940 	}
1941 
1942 	switch (optname) {
1943 
1944 	case IPV6_MULTICAST_IF:
1945 		/*
1946 		 * Select the interface for outgoing multicast packets.
1947 		 */
1948 		if (m == NULL || m->m_len != sizeof(u_int)) {
1949 			error = EINVAL;
1950 			break;
1951 		}
1952 		memcpy(&ifindex, mtod(m, u_int *), sizeof(ifindex));
1953 		if (ifindex != 0) {
1954 			ifp = if_get(ifindex);
1955 			if (ifp == NULL) {
1956 				error = ENXIO;	/* XXX EINVAL? */
1957 				break;
1958 			}
1959 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
1960 				error = EADDRNOTAVAIL;
1961 				if_put(ifp);
1962 				break;
1963 			}
1964 			if_put(ifp);
1965 		}
1966 		im6o->im6o_ifidx = ifindex;
1967 		break;
1968 
1969 	case IPV6_MULTICAST_HOPS:
1970 	    {
1971 		/*
1972 		 * Set the IP6 hoplimit for outgoing multicast packets.
1973 		 */
1974 		int optval;
1975 		if (m == NULL || m->m_len != sizeof(int)) {
1976 			error = EINVAL;
1977 			break;
1978 		}
1979 		memcpy(&optval, mtod(m, u_int *), sizeof(optval));
1980 		if (optval < -1 || optval >= 256)
1981 			error = EINVAL;
1982 		else if (optval == -1)
1983 			im6o->im6o_hlim = ip6_defmcasthlim;
1984 		else
1985 			im6o->im6o_hlim = optval;
1986 		break;
1987 	    }
1988 
1989 	case IPV6_MULTICAST_LOOP:
1990 		/*
1991 		 * Set the loopback flag for outgoing multicast packets.
1992 		 * Must be zero or one.
1993 		 */
1994 		if (m == NULL || m->m_len != sizeof(u_int)) {
1995 			error = EINVAL;
1996 			break;
1997 		}
1998 		memcpy(&loop, mtod(m, u_int *), sizeof(loop));
1999 		if (loop > 1) {
2000 			error = EINVAL;
2001 			break;
2002 		}
2003 		im6o->im6o_loop = loop;
2004 		break;
2005 
2006 	case IPV6_JOIN_GROUP:
2007 		/*
2008 		 * Add a multicast group membership.
2009 		 * Group must be a valid IP6 multicast address.
2010 		 */
2011 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2012 			error = EINVAL;
2013 			break;
2014 		}
2015 		mreq = mtod(m, struct ipv6_mreq *);
2016 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2017 			/*
2018 			 * We use the unspecified address to specify to accept
2019 			 * all multicast addresses. Only super user is allowed
2020 			 * to do this.
2021 			 */
2022 			if (suser(p, 0))
2023 			{
2024 				error = EACCES;
2025 				break;
2026 			}
2027 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2028 			error = EINVAL;
2029 			break;
2030 		}
2031 
2032 		/*
2033 		 * If no interface was explicitly specified, choose an
2034 		 * appropriate one according to the given multicast address.
2035 		 */
2036 		if (mreq->ipv6mr_interface == 0) {
2037 			/*
2038 			 * Look up the routing table for the
2039 			 * address, and choose the outgoing interface.
2040 			 *   XXX: is it a good approach?
2041 			 */
2042 			bzero(&ro, sizeof(ro));
2043 			ro.ro_tableid = m->m_pkthdr.ph_rtableid;
2044 			dst = &ro.ro_dst;
2045 			dst->sin6_len = sizeof(struct sockaddr_in6);
2046 			dst->sin6_family = AF_INET6;
2047 			dst->sin6_addr = mreq->ipv6mr_multiaddr;
2048 			ro.ro_rt = rtalloc(sin6tosa(&ro.ro_dst),
2049 			    RT_RESOLVE, ro.ro_tableid);
2050 			if (ro.ro_rt == NULL) {
2051 				error = EADDRNOTAVAIL;
2052 				break;
2053 			}
2054 			ifp = if_get(ro.ro_rt->rt_ifidx);
2055 			rtfree(ro.ro_rt);
2056 		} else {
2057 			/*
2058 			 * If the interface is specified, validate it.
2059 			 */
2060 			ifp = if_get(mreq->ipv6mr_interface);
2061 			if (ifp == NULL) {
2062 				error = ENXIO;	/* XXX EINVAL? */
2063 				break;
2064 			}
2065 		}
2066 
2067 		/*
2068 		 * See if we found an interface, and confirm that it
2069 		 * supports multicast
2070 		 */
2071 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2072 			if_put(ifp);
2073 			error = EADDRNOTAVAIL;
2074 			break;
2075 		}
2076 		/*
2077 		 * Put interface index into the multicast address,
2078 		 * if the address has link/interface-local scope.
2079 		 */
2080 		if (IN6_IS_SCOPE_EMBED(&mreq->ipv6mr_multiaddr)) {
2081 			mreq->ipv6mr_multiaddr.s6_addr16[1] =
2082 			    htons(ifp->if_index);
2083 		}
2084 		/*
2085 		 * See if the membership already exists.
2086 		 */
2087 		LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain)
2088 			if (imm->i6mm_maddr->in6m_ifidx == ifp->if_index &&
2089 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2090 			    &mreq->ipv6mr_multiaddr))
2091 				break;
2092 		if (imm != NULL) {
2093 			if_put(ifp);
2094 			error = EADDRINUSE;
2095 			break;
2096 		}
2097 		/*
2098 		 * Everything looks good; add a new record to the multicast
2099 		 * address list for the given interface.
2100 		 */
2101 		imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error);
2102 		if_put(ifp);
2103 		if (!imm)
2104 			break;
2105 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2106 		break;
2107 
2108 	case IPV6_LEAVE_GROUP:
2109 		/*
2110 		 * Drop a multicast group membership.
2111 		 * Group must be a valid IP6 multicast address.
2112 		 */
2113 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2114 			error = EINVAL;
2115 			break;
2116 		}
2117 		mreq = mtod(m, struct ipv6_mreq *);
2118 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2119 			if (suser(p, 0))
2120 			{
2121 				error = EACCES;
2122 				break;
2123 			}
2124 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2125 			error = EINVAL;
2126 			break;
2127 		}
2128 
2129 		/*
2130 		 * Put interface index into the multicast address,
2131 		 * if the address has link-local scope.
2132 		 */
2133 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2134 			mreq->ipv6mr_multiaddr.s6_addr16[1] =
2135 			    htons(mreq->ipv6mr_interface);
2136 		}
2137 
2138 		/*
2139 		 * If an interface address was specified, get a pointer
2140 		 * to its ifnet structure.
2141 		 */
2142 		if (mreq->ipv6mr_interface == 0)
2143 			ifp = NULL;
2144 		else {
2145 			ifp = if_get(mreq->ipv6mr_interface);
2146 			if (ifp == NULL) {
2147 				error = ENXIO;	/* XXX EINVAL? */
2148 				break;
2149 			}
2150 		}
2151 
2152 		/*
2153 		 * Find the membership in the membership list.
2154 		 */
2155 		LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain) {
2156 			if ((ifp == NULL ||
2157 			    imm->i6mm_maddr->in6m_ifidx == ifp->if_index) &&
2158 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2159 			    &mreq->ipv6mr_multiaddr))
2160 				break;
2161 		}
2162 
2163 		if_put(ifp);
2164 
2165 		if (imm == NULL) {
2166 			/* Unable to resolve interface */
2167 			error = EADDRNOTAVAIL;
2168 			break;
2169 		}
2170 		/*
2171 		 * Give up the multicast address record to which the
2172 		 * membership points.
2173 		 */
2174 		LIST_REMOVE(imm, i6mm_chain);
2175 		in6_leavegroup(imm);
2176 		break;
2177 
2178 	default:
2179 		error = EOPNOTSUPP;
2180 		break;
2181 	}
2182 
2183 	/*
2184 	 * If all options have default values, no need to keep the option
2185 	 * structure.
2186 	 */
2187 	if (im6o->im6o_ifidx == 0 &&
2188 	    im6o->im6o_hlim == ip6_defmcasthlim &&
2189 	    im6o->im6o_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2190 	    LIST_EMPTY(&im6o->im6o_memberships)) {
2191 		free(*im6op, M_IPMOPTS, 0);
2192 		*im6op = NULL;
2193 	}
2194 
2195 	return (error);
2196 }
2197 
2198 /*
2199  * Return the IP6 multicast options in response to user getsockopt().
2200  */
2201 int
2202 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp)
2203 {
2204 	u_int *hlim, *loop, *ifindex;
2205 
2206 	*mp = m_get(M_WAIT, MT_SOOPTS);
2207 
2208 	switch (optname) {
2209 
2210 	case IPV6_MULTICAST_IF:
2211 		ifindex = mtod(*mp, u_int *);
2212 		(*mp)->m_len = sizeof(u_int);
2213 		if (im6o == NULL || im6o->im6o_ifidx == 0)
2214 			*ifindex = 0;
2215 		else
2216 			*ifindex = im6o->im6o_ifidx;
2217 		return (0);
2218 
2219 	case IPV6_MULTICAST_HOPS:
2220 		hlim = mtod(*mp, u_int *);
2221 		(*mp)->m_len = sizeof(u_int);
2222 		if (im6o == NULL)
2223 			*hlim = ip6_defmcasthlim;
2224 		else
2225 			*hlim = im6o->im6o_hlim;
2226 		return (0);
2227 
2228 	case IPV6_MULTICAST_LOOP:
2229 		loop = mtod(*mp, u_int *);
2230 		(*mp)->m_len = sizeof(u_int);
2231 		if (im6o == NULL)
2232 			*loop = ip6_defmcasthlim;
2233 		else
2234 			*loop = im6o->im6o_loop;
2235 		return (0);
2236 
2237 	default:
2238 		return (EOPNOTSUPP);
2239 	}
2240 }
2241 
2242 /*
2243  * Discard the IP6 multicast options.
2244  */
2245 void
2246 ip6_freemoptions(struct ip6_moptions *im6o)
2247 {
2248 	struct in6_multi_mship *imm;
2249 
2250 	if (im6o == NULL)
2251 		return;
2252 
2253 	while (!LIST_EMPTY(&im6o->im6o_memberships)) {
2254 		imm = LIST_FIRST(&im6o->im6o_memberships);
2255 		LIST_REMOVE(imm, i6mm_chain);
2256 		in6_leavegroup(imm);
2257 	}
2258 	free(im6o, M_IPMOPTS, 0);
2259 }
2260 
2261 /*
2262  * Set IPv6 outgoing packet options based on advanced API.
2263  */
2264 int
2265 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2266     struct ip6_pktopts *stickyopt, int priv, int uproto)
2267 {
2268 	u_int clen;
2269 	struct cmsghdr *cm = 0;
2270 	caddr_t cmsgs;
2271 	int error;
2272 
2273 	if (control == NULL || opt == NULL)
2274 		return (EINVAL);
2275 
2276 	ip6_initpktopts(opt);
2277 	if (stickyopt) {
2278 		int error;
2279 
2280 		/*
2281 		 * If stickyopt is provided, make a local copy of the options
2282 		 * for this particular packet, then override them by ancillary
2283 		 * objects.
2284 		 * XXX: copypktopts() does not copy the cached route to a next
2285 		 * hop (if any).  This is not very good in terms of efficiency,
2286 		 * but we can allow this since this option should be rarely
2287 		 * used.
2288 		 */
2289 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2290 			return (error);
2291 	}
2292 
2293 	/*
2294 	 * XXX: Currently, we assume all the optional information is stored
2295 	 * in a single mbuf.
2296 	 */
2297 	if (control->m_next)
2298 		return (EINVAL);
2299 
2300 	clen = control->m_len;
2301 	cmsgs = mtod(control, caddr_t);
2302 	do {
2303 		if (clen < CMSG_LEN(0))
2304 			return (EINVAL);
2305 		cm = (struct cmsghdr *)cmsgs;
2306 		if (cm->cmsg_len < CMSG_LEN(0) || cm->cmsg_len > clen ||
2307 		    CMSG_ALIGN(cm->cmsg_len) > clen)
2308 			return (EINVAL);
2309 		if (cm->cmsg_level == IPPROTO_IPV6) {
2310 			error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2311 			    cm->cmsg_len - CMSG_LEN(0), opt, priv, 0, uproto);
2312 			if (error)
2313 				return (error);
2314 		}
2315 
2316 		clen -= CMSG_ALIGN(cm->cmsg_len);
2317 		cmsgs += CMSG_ALIGN(cm->cmsg_len);
2318 	} while (clen);
2319 
2320 	return (0);
2321 }
2322 
2323 /*
2324  * Set a particular packet option, as a sticky option or an ancillary data
2325  * item.  "len" can be 0 only when it's a sticky option.
2326  */
2327 int
2328 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2329     int priv, int sticky, int uproto)
2330 {
2331 	int minmtupolicy;
2332 
2333 	switch (optname) {
2334 	case IPV6_PKTINFO:
2335 	{
2336 		struct ifnet *ifp = NULL;
2337 		struct in6_pktinfo *pktinfo;
2338 
2339 		if (len != sizeof(struct in6_pktinfo))
2340 			return (EINVAL);
2341 
2342 		pktinfo = (struct in6_pktinfo *)buf;
2343 
2344 		/*
2345 		 * An application can clear any sticky IPV6_PKTINFO option by
2346 		 * doing a "regular" setsockopt with ipi6_addr being
2347 		 * in6addr_any and ipi6_ifindex being zero.
2348 		 * [RFC 3542, Section 6]
2349 		 */
2350 		if (opt->ip6po_pktinfo &&
2351 		    pktinfo->ipi6_ifindex == 0 &&
2352 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2353 			ip6_clearpktopts(opt, optname);
2354 			break;
2355 		}
2356 
2357 		if (uproto == IPPROTO_TCP &&
2358 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2359 			return (EINVAL);
2360 		}
2361 
2362 		if (pktinfo->ipi6_ifindex) {
2363 			ifp = if_get(pktinfo->ipi6_ifindex);
2364 			if (ifp == NULL)
2365 				return (ENXIO);
2366 			if_put(ifp);
2367 		}
2368 
2369 		/*
2370 		 * We store the address anyway, and let in6_selectsrc()
2371 		 * validate the specified address.  This is because ipi6_addr
2372 		 * may not have enough information about its scope zone, and
2373 		 * we may need additional information (such as outgoing
2374 		 * interface or the scope zone of a destination address) to
2375 		 * disambiguate the scope.
2376 		 * XXX: the delay of the validation may confuse the
2377 		 * application when it is used as a sticky option.
2378 		 */
2379 		if (opt->ip6po_pktinfo == NULL) {
2380 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2381 			    M_IP6OPT, M_NOWAIT);
2382 			if (opt->ip6po_pktinfo == NULL)
2383 				return (ENOBUFS);
2384 		}
2385 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2386 		break;
2387 	}
2388 
2389 	case IPV6_HOPLIMIT:
2390 	{
2391 		int *hlimp;
2392 
2393 		/*
2394 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2395 		 * to simplify the ordering among hoplimit options.
2396 		 */
2397 		if (sticky)
2398 			return (ENOPROTOOPT);
2399 
2400 		if (len != sizeof(int))
2401 			return (EINVAL);
2402 		hlimp = (int *)buf;
2403 		if (*hlimp < -1 || *hlimp > 255)
2404 			return (EINVAL);
2405 
2406 		opt->ip6po_hlim = *hlimp;
2407 		break;
2408 	}
2409 
2410 	case IPV6_TCLASS:
2411 	{
2412 		int tclass;
2413 
2414 		if (len != sizeof(int))
2415 			return (EINVAL);
2416 		tclass = *(int *)buf;
2417 		if (tclass < -1 || tclass > 255)
2418 			return (EINVAL);
2419 
2420 		opt->ip6po_tclass = tclass;
2421 		break;
2422 	}
2423 	case IPV6_HOPOPTS:
2424 	{
2425 		struct ip6_hbh *hbh;
2426 		int hbhlen;
2427 
2428 		/*
2429 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2430 		 * options, since per-option restriction has too much
2431 		 * overhead.
2432 		 */
2433 		if (!priv)
2434 			return (EPERM);
2435 
2436 		if (len == 0) {
2437 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2438 			break;	/* just remove the option */
2439 		}
2440 
2441 		/* message length validation */
2442 		if (len < sizeof(struct ip6_hbh))
2443 			return (EINVAL);
2444 		hbh = (struct ip6_hbh *)buf;
2445 		hbhlen = (hbh->ip6h_len + 1) << 3;
2446 		if (len != hbhlen)
2447 			return (EINVAL);
2448 
2449 		/* turn off the previous option, then set the new option. */
2450 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2451 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2452 		if (opt->ip6po_hbh == NULL)
2453 			return (ENOBUFS);
2454 		memcpy(opt->ip6po_hbh, hbh, hbhlen);
2455 
2456 		break;
2457 	}
2458 
2459 	case IPV6_DSTOPTS:
2460 	case IPV6_RTHDRDSTOPTS:
2461 	{
2462 		struct ip6_dest *dest, **newdest = NULL;
2463 		int destlen;
2464 
2465 		if (!priv)	/* XXX: see the comment for IPV6_HOPOPTS */
2466 			return (EPERM);
2467 
2468 		if (len == 0) {
2469 			ip6_clearpktopts(opt, optname);
2470 			break;	/* just remove the option */
2471 		}
2472 
2473 		/* message length validation */
2474 		if (len < sizeof(struct ip6_dest))
2475 			return (EINVAL);
2476 		dest = (struct ip6_dest *)buf;
2477 		destlen = (dest->ip6d_len + 1) << 3;
2478 		if (len != destlen)
2479 			return (EINVAL);
2480 		/*
2481 		 * Determine the position that the destination options header
2482 		 * should be inserted; before or after the routing header.
2483 		 */
2484 		switch (optname) {
2485 		case IPV6_RTHDRDSTOPTS:
2486 			newdest = &opt->ip6po_dest1;
2487 			break;
2488 		case IPV6_DSTOPTS:
2489 			newdest = &opt->ip6po_dest2;
2490 			break;
2491 		}
2492 
2493 		/* turn off the previous option, then set the new option. */
2494 		ip6_clearpktopts(opt, optname);
2495 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2496 		if (*newdest == NULL)
2497 			return (ENOBUFS);
2498 		memcpy(*newdest, dest, destlen);
2499 
2500 		break;
2501 	}
2502 
2503 	case IPV6_RTHDR:
2504 	{
2505 		struct ip6_rthdr *rth;
2506 		int rthlen;
2507 
2508 		if (len == 0) {
2509 			ip6_clearpktopts(opt, IPV6_RTHDR);
2510 			break;	/* just remove the option */
2511 		}
2512 
2513 		/* message length validation */
2514 		if (len < sizeof(struct ip6_rthdr))
2515 			return (EINVAL);
2516 		rth = (struct ip6_rthdr *)buf;
2517 		rthlen = (rth->ip6r_len + 1) << 3;
2518 		if (len != rthlen)
2519 			return (EINVAL);
2520 
2521 		switch (rth->ip6r_type) {
2522 		case IPV6_RTHDR_TYPE_0:
2523 			if (rth->ip6r_len == 0)	/* must contain one addr */
2524 				return (EINVAL);
2525 			if (rth->ip6r_len % 2) /* length must be even */
2526 				return (EINVAL);
2527 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2528 				return (EINVAL);
2529 			break;
2530 		default:
2531 			return (EINVAL);	/* not supported */
2532 		}
2533 		/* turn off the previous option */
2534 		ip6_clearpktopts(opt, IPV6_RTHDR);
2535 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2536 		if (opt->ip6po_rthdr == NULL)
2537 			return (ENOBUFS);
2538 		memcpy(opt->ip6po_rthdr, rth, rthlen);
2539 		break;
2540 	}
2541 
2542 	case IPV6_USE_MIN_MTU:
2543 		if (len != sizeof(int))
2544 			return (EINVAL);
2545 		minmtupolicy = *(int *)buf;
2546 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2547 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
2548 		    minmtupolicy != IP6PO_MINMTU_ALL) {
2549 			return (EINVAL);
2550 		}
2551 		opt->ip6po_minmtu = minmtupolicy;
2552 		break;
2553 
2554 	case IPV6_DONTFRAG:
2555 		if (len != sizeof(int))
2556 			return (EINVAL);
2557 
2558 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2559 			/*
2560 			 * we ignore this option for TCP sockets.
2561 			 * (RFC3542 leaves this case unspecified.)
2562 			 */
2563 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2564 		} else
2565 			opt->ip6po_flags |= IP6PO_DONTFRAG;
2566 		break;
2567 
2568 	default:
2569 		return (ENOPROTOOPT);
2570 	} /* end of switch */
2571 
2572 	return (0);
2573 }
2574 
2575 /*
2576  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2577  * packet to the input queue of a specified interface.
2578  */
2579 void
2580 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2581 {
2582 	struct mbuf *copym;
2583 	struct ip6_hdr *ip6;
2584 
2585 	/*
2586 	 * Duplicate the packet.
2587 	 */
2588 	copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
2589 	if (copym == NULL)
2590 		return;
2591 
2592 	/*
2593 	 * Make sure to deep-copy IPv6 header portion in case the data
2594 	 * is in an mbuf cluster, so that we can safely override the IPv6
2595 	 * header portion later.
2596 	 */
2597 	if ((copym->m_flags & M_EXT) != 0 ||
2598 	    copym->m_len < sizeof(struct ip6_hdr)) {
2599 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
2600 		if (copym == NULL)
2601 			return;
2602 	}
2603 
2604 #ifdef DIAGNOSTIC
2605 	if (copym->m_len < sizeof(*ip6)) {
2606 		m_freem(copym);
2607 		return;
2608 	}
2609 #endif
2610 
2611 	ip6 = mtod(copym, struct ip6_hdr *);
2612 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
2613 		ip6->ip6_src.s6_addr16[1] = 0;
2614 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
2615 		ip6->ip6_dst.s6_addr16[1] = 0;
2616 
2617 	if_input_local(ifp, copym, dst->sin6_family);
2618 }
2619 
2620 /*
2621  * Chop IPv6 header off from the payload.
2622  */
2623 int
2624 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
2625 {
2626 	struct mbuf *mh;
2627 	struct ip6_hdr *ip6;
2628 
2629 	ip6 = mtod(m, struct ip6_hdr *);
2630 	if (m->m_len > sizeof(*ip6)) {
2631 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
2632 		if (mh == NULL) {
2633 			m_freem(m);
2634 			return ENOBUFS;
2635 		}
2636 		M_MOVE_PKTHDR(mh, m);
2637 		MH_ALIGN(mh, sizeof(*ip6));
2638 		m->m_len -= sizeof(*ip6);
2639 		m->m_data += sizeof(*ip6);
2640 		mh->m_next = m;
2641 		m = mh;
2642 		m->m_len = sizeof(*ip6);
2643 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2644 	}
2645 	exthdrs->ip6e_ip6 = m;
2646 	return 0;
2647 }
2648 
2649 u_int32_t
2650 ip6_randomid(void)
2651 {
2652 	return idgen32(&ip6_id_ctx);
2653 }
2654 
2655 void
2656 ip6_randomid_init(void)
2657 {
2658 	idgen32_init(&ip6_id_ctx);
2659 }
2660 
2661 /*
2662  *	Compute significant parts of the IPv6 checksum pseudo-header
2663  *	for use in a delayed TCP/UDP checksum calculation.
2664  */
2665 static __inline u_int16_t __attribute__((__unused__))
2666 in6_cksum_phdr(const struct in6_addr *src, const struct in6_addr *dst,
2667     u_int32_t len, u_int32_t nxt)
2668 {
2669 	u_int32_t sum = 0;
2670 	const u_int16_t *w;
2671 
2672 	w = (const u_int16_t *) src;
2673 	sum += w[0];
2674 	if (!IN6_IS_SCOPE_EMBED(src))
2675 		sum += w[1];
2676 	sum += w[2]; sum += w[3]; sum += w[4]; sum += w[5];
2677 	sum += w[6]; sum += w[7];
2678 
2679 	w = (const u_int16_t *) dst;
2680 	sum += w[0];
2681 	if (!IN6_IS_SCOPE_EMBED(dst))
2682 		sum += w[1];
2683 	sum += w[2]; sum += w[3]; sum += w[4]; sum += w[5];
2684 	sum += w[6]; sum += w[7];
2685 
2686 	sum += (u_int16_t)(len >> 16) + (u_int16_t)(len /*& 0xffff*/);
2687 
2688 	sum += (u_int16_t)(nxt >> 16) + (u_int16_t)(nxt /*& 0xffff*/);
2689 
2690 	sum = (u_int16_t)(sum >> 16) + (u_int16_t)(sum /*& 0xffff*/);
2691 
2692 	if (sum > 0xffff)
2693 		sum -= 0xffff;
2694 
2695 	return (sum);
2696 }
2697 
2698 /*
2699  * Process a delayed payload checksum calculation.
2700  */
2701 void
2702 in6_delayed_cksum(struct mbuf *m, u_int8_t nxt)
2703 {
2704 	int nxtp, offset;
2705 	u_int16_t csum;
2706 
2707 	offset = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxtp);
2708 	if (offset <= 0 || nxtp != nxt)
2709 		/* If the desired next protocol isn't found, punt. */
2710 		return;
2711 	csum = (u_int16_t)(in6_cksum(m, 0, offset, m->m_pkthdr.len - offset));
2712 
2713 	switch (nxt) {
2714 	case IPPROTO_TCP:
2715 		offset += offsetof(struct tcphdr, th_sum);
2716 		break;
2717 
2718 	case IPPROTO_UDP:
2719 		offset += offsetof(struct udphdr, uh_sum);
2720 		if (csum == 0)
2721 			csum = 0xffff;
2722 		break;
2723 
2724 	case IPPROTO_ICMPV6:
2725 		offset += offsetof(struct icmp6_hdr, icmp6_cksum);
2726 		break;
2727 	}
2728 
2729 	if ((offset + sizeof(u_int16_t)) > m->m_len)
2730 		m_copyback(m, offset, sizeof(csum), &csum, M_NOWAIT);
2731 	else
2732 		*(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
2733 }
2734 
2735 void
2736 in6_proto_cksum_out(struct mbuf *m, struct ifnet *ifp)
2737 {
2738 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2739 
2740 	/* some hw and in6_delayed_cksum need the pseudo header cksum */
2741 	if (m->m_pkthdr.csum_flags &
2742 	    (M_TCP_CSUM_OUT|M_UDP_CSUM_OUT|M_ICMP_CSUM_OUT)) {
2743 		int nxt, offset;
2744 		u_int16_t csum;
2745 
2746 		offset = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
2747 		csum = in6_cksum_phdr(&ip6->ip6_src, &ip6->ip6_dst,
2748 		    htonl(m->m_pkthdr.len - offset), htonl(nxt));
2749 		if (nxt == IPPROTO_TCP)
2750 			offset += offsetof(struct tcphdr, th_sum);
2751 		else if (nxt == IPPROTO_UDP)
2752 			offset += offsetof(struct udphdr, uh_sum);
2753 		else if (nxt == IPPROTO_ICMPV6)
2754 			offset += offsetof(struct icmp6_hdr, icmp6_cksum);
2755 		if ((offset + sizeof(u_int16_t)) > m->m_len)
2756 			m_copyback(m, offset, sizeof(csum), &csum, M_NOWAIT);
2757 		else
2758 			*(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
2759 	}
2760 
2761 	if (m->m_pkthdr.csum_flags & M_TCP_CSUM_OUT) {
2762 		if (!ifp || !(ifp->if_capabilities & IFCAP_CSUM_TCPv6) ||
2763 		    ip6->ip6_nxt != IPPROTO_TCP ||
2764 		    ifp->if_bridgeport != NULL) {
2765 			tcpstat.tcps_outswcsum++;
2766 			in6_delayed_cksum(m, IPPROTO_TCP);
2767 			m->m_pkthdr.csum_flags &= ~M_TCP_CSUM_OUT; /* Clear */
2768 		}
2769 	} else if (m->m_pkthdr.csum_flags & M_UDP_CSUM_OUT) {
2770 		if (!ifp || !(ifp->if_capabilities & IFCAP_CSUM_UDPv6) ||
2771 		    ip6->ip6_nxt != IPPROTO_UDP ||
2772 		    ifp->if_bridgeport != NULL) {
2773 			udpstat.udps_outswcsum++;
2774 			in6_delayed_cksum(m, IPPROTO_UDP);
2775 			m->m_pkthdr.csum_flags &= ~M_UDP_CSUM_OUT; /* Clear */
2776 		}
2777 	} else if (m->m_pkthdr.csum_flags & M_ICMP_CSUM_OUT) {
2778 		in6_delayed_cksum(m, IPPROTO_ICMPV6);
2779 		m->m_pkthdr.csum_flags &= ~M_ICMP_CSUM_OUT; /* Clear */
2780 	}
2781 }
2782 
2783 #ifdef IPSEC
2784 struct tdb *
2785 ip6_output_ipsec_lookup(struct mbuf *m, int *error, struct inpcb *inp)
2786 {
2787 	struct tdb *tdb;
2788 	struct m_tag *mtag;
2789 	struct tdb_ident *tdbi;
2790 
2791 	/*
2792 	 * Check if there was an outgoing SA bound to the flow
2793 	 * from a transport protocol.
2794 	 */
2795 
2796 	/* Do we have any pending SAs to apply ? */
2797 	tdb = ipsp_spd_lookup(m, AF_INET6, sizeof(struct ip6_hdr),
2798 	    error, IPSP_DIRECTION_OUT, NULL, inp, 0);
2799 
2800 	if (tdb == NULL)
2801 		return NULL;
2802 	/* Loop detection */
2803 	for (mtag = m_tag_first(m); mtag != NULL; mtag = m_tag_next(m, mtag)) {
2804 		if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE)
2805 			continue;
2806 		tdbi = (struct tdb_ident *)(mtag + 1);
2807 		if (tdbi->spi == tdb->tdb_spi &&
2808 		    tdbi->proto == tdb->tdb_sproto &&
2809 		    tdbi->rdomain == tdb->tdb_rdomain &&
2810 		    !memcmp(&tdbi->dst, &tdb->tdb_dst,
2811 		    sizeof(union sockaddr_union))) {
2812 			/* no IPsec needed */
2813 			return NULL;
2814 		}
2815 	}
2816 	return tdb;
2817 }
2818 
2819 int
2820 ip6_output_ipsec_send(struct tdb *tdb, struct mbuf *m, int tunalready, int fwd)
2821 {
2822 #if NPF > 0
2823 	struct ifnet *encif;
2824 #endif
2825 
2826 #if NPF > 0
2827 	if ((encif = enc_getif(tdb->tdb_rdomain, tdb->tdb_tap)) == NULL ||
2828 	    pf_test(AF_INET6, fwd ? PF_FWD : PF_OUT, encif, &m) != PF_PASS) {
2829 		m_freem(m);
2830 		return EHOSTUNREACH;
2831 	}
2832 	if (m == NULL)
2833 		return 0;
2834 	/*
2835 	 * PF_TAG_REROUTE handling or not...
2836 	 * Packet is entering IPsec so the routing is
2837 	 * already overruled by the IPsec policy.
2838 	 * Until now the change was not reconsidered.
2839 	 * What's the behaviour?
2840 	 */
2841 	in6_proto_cksum_out(m, encif);
2842 #endif
2843 	m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
2844 
2845 	/* Callee frees mbuf */
2846 	return ipsp_process_packet(m, tdb, AF_INET6, tunalready);
2847 }
2848 #endif /* IPSEC */
2849