xref: /openbsd-src/sys/netinet6/ip6_output.c (revision b8851fcc53cbe24fd20b090f26dd149e353f6174)
1 /*	$OpenBSD: ip6_output.c,v 1.223 2017/02/01 20:59:47 dhill 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 *);
130 int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *);
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 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(rtable_loindex(m->m_pkthdr.ph_rtableid));
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)) != 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)
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 	tlen = m->m_pkthdr.len;
666 
667 	if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
668 		dontfrag = 1;
669 	else
670 		dontfrag = 0;
671 	if (dontfrag && tlen > ifp->if_mtu) {	/* case 2-b */
672 		error = EMSGSIZE;
673 		goto bad;
674 	}
675 
676 	/*
677 	 * transmit packet without fragmentation
678 	 */
679 	if (dontfrag || (tlen <= mtu)) {	/* case 1-a and 2-a */
680 		error = ifp->if_output(ifp, m, sin6tosa(dst), ro->ro_rt);
681 		goto done;
682 	}
683 
684 	/*
685 	 * try to fragment the packet.  case 1-b
686 	 */
687 	if (mtu < IPV6_MMTU) {
688 		/* path MTU cannot be less than IPV6_MMTU */
689 		error = EMSGSIZE;
690 		goto bad;
691 	} else if (ip6->ip6_plen == 0) {
692 		/* jumbo payload cannot be fragmented */
693 		error = EMSGSIZE;
694 		goto bad;
695 	} else {
696 		u_char nextproto;
697 #if 0
698 		struct ip6ctlparam ip6cp;
699 		u_int32_t mtu32;
700 #endif
701 
702 		/*
703 		 * Too large for the destination or interface;
704 		 * fragment if possible.
705 		 * Must be able to put at least 8 bytes per fragment.
706 		 */
707 		hlen = unfragpartlen;
708 		if (mtu > IPV6_MAXPACKET)
709 			mtu = IPV6_MAXPACKET;
710 
711 #if 0
712 		/* Notify a proper path MTU to applications. */
713 		mtu32 = (u_int32_t)mtu;
714 		bzero(&ip6cp, sizeof(ip6cp));
715 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
716 		pfctlinput2(PRC_MSGSIZE, sin6tosa(&ro_pmtu->ro_dst),
717 		    (void *)&ip6cp);
718 #endif
719 
720 		/*
721 		 * Change the next header field of the last header in the
722 		 * unfragmentable part.
723 		 */
724 		if (exthdrs.ip6e_rthdr) {
725 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
726 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
727 		} else if (exthdrs.ip6e_dest1) {
728 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
729 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
730 		} else if (exthdrs.ip6e_hbh) {
731 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
732 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
733 		} else {
734 			nextproto = ip6->ip6_nxt;
735 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
736 		}
737 
738 		m0 = m;
739 		error = ip6_fragment(m0, hlen, nextproto, mtu);
740 		if (error)
741 			ip6stat.ip6s_odropped++;
742 	}
743 
744 	/*
745 	 * Remove leading garbages.
746 	 */
747 	m = m0->m_nextpkt;
748 	m0->m_nextpkt = 0;
749 	m_freem(m0);
750 	for (m0 = m; m; m = m0) {
751 		m0 = m->m_nextpkt;
752 		m->m_nextpkt = 0;
753 		if (error == 0) {
754 			ip6stat.ip6s_ofragments++;
755 			error = ifp->if_output(ifp, m, sin6tosa(dst),
756 			    ro->ro_rt);
757 		} else
758 			m_freem(m);
759 	}
760 
761 	if (error == 0)
762 		ip6stat.ip6s_fragmented++;
763 
764 done:
765 	if_put(ifp);
766 	if (ro == &ip6route && ro->ro_rt) {
767 		rtfree(ro->ro_rt);
768 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
769 		rtfree(ro_pmtu->ro_rt);
770 	}
771 
772 	return (error);
773 
774 freehdrs:
775 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
776 	m_freem(exthdrs.ip6e_dest1);
777 	m_freem(exthdrs.ip6e_rthdr);
778 	m_freem(exthdrs.ip6e_dest2);
779 	/* FALLTHROUGH */
780 bad:
781 	m_freem(m);
782 	goto done;
783 }
784 
785 int
786 ip6_fragment(struct mbuf *m0, int hlen, u_char nextproto, u_long mtu)
787 {
788 	struct mbuf	*m, **mnext, *m_frgpart;
789 	struct ip6_hdr	*mhip6;
790 	struct ip6_frag	*ip6f;
791 	u_int32_t	 id;
792 	int		 tlen, len, off;
793 	int		 error;
794 
795 	id = htonl(ip6_randomid());
796 
797 	mnext = &m0->m_nextpkt;
798 	*mnext = NULL;
799 
800 	tlen = m0->m_pkthdr.len;
801 	len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
802 	if (len < 8)
803 		return (EMSGSIZE);
804 
805 	/*
806 	 * Loop through length of segment after first fragment,
807 	 * make new header and copy data of each part and link onto
808 	 * chain.
809 	 */
810 	for (off = hlen; off < tlen; off += len) {
811 		struct mbuf *mlast;
812 
813 		if ((m = m_gethdr(M_DONTWAIT, MT_HEADER)) == NULL)
814 			return (ENOBUFS);
815 		*mnext = m;
816 		mnext = &m->m_nextpkt;
817 		if ((error = m_dup_pkthdr(m, m0, M_DONTWAIT)) != 0)
818 			return (error);
819 		m->m_data += max_linkhdr;
820 		mhip6 = mtod(m, struct ip6_hdr *);
821 		*mhip6 = *mtod(m0, struct ip6_hdr *);
822 		m->m_len = sizeof(*mhip6);
823 		if ((error = ip6_insertfraghdr(m0, m, hlen, &ip6f)) != 0)
824 			return (error);
825 		ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7));
826 		if (off + len >= tlen)
827 			len = tlen - off;
828 		else
829 			ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
830 		mhip6->ip6_plen = htons((u_int16_t)(len + hlen +
831 		    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
832 		if ((m_frgpart = m_copym(m0, off, len, M_DONTWAIT)) == NULL)
833 			return (ENOBUFS);
834 		for (mlast = m; mlast->m_next; mlast = mlast->m_next)
835 			;
836 		mlast->m_next = m_frgpart;
837 		m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
838 		ip6f->ip6f_reserved = 0;
839 		ip6f->ip6f_ident = id;
840 		ip6f->ip6f_nxt = nextproto;
841 	}
842 
843 	return (0);
844 }
845 
846 int
847 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
848 {
849 	struct mbuf *m;
850 
851 	if (hlen > MCLBYTES)
852 		return (ENOBUFS); /* XXX */
853 
854 	MGET(m, M_DONTWAIT, MT_DATA);
855 	if (!m)
856 		return (ENOBUFS);
857 
858 	if (hlen > MLEN) {
859 		MCLGET(m, M_DONTWAIT);
860 		if ((m->m_flags & M_EXT) == 0) {
861 			m_free(m);
862 			return (ENOBUFS);
863 		}
864 	}
865 	m->m_len = hlen;
866 	if (hdr)
867 		bcopy(hdr, mtod(m, caddr_t), hlen);
868 
869 	*mp = m;
870 	return (0);
871 }
872 
873 /*
874  * Insert jumbo payload option.
875  */
876 int
877 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
878 {
879 	struct mbuf *mopt;
880 	u_int8_t *optbuf;
881 	u_int32_t v;
882 
883 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
884 
885 	/*
886 	 * If there is no hop-by-hop options header, allocate new one.
887 	 * If there is one but it doesn't have enough space to store the
888 	 * jumbo payload option, allocate a cluster to store the whole options.
889 	 * Otherwise, use it to store the options.
890 	 */
891 	if (exthdrs->ip6e_hbh == 0) {
892 		MGET(mopt, M_DONTWAIT, MT_DATA);
893 		if (mopt == NULL)
894 			return (ENOBUFS);
895 		mopt->m_len = JUMBOOPTLEN;
896 		optbuf = mtod(mopt, u_int8_t *);
897 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
898 		exthdrs->ip6e_hbh = mopt;
899 	} else {
900 		struct ip6_hbh *hbh;
901 
902 		mopt = exthdrs->ip6e_hbh;
903 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
904 			/*
905 			 * XXX assumption:
906 			 * - exthdrs->ip6e_hbh is not referenced from places
907 			 *   other than exthdrs.
908 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
909 			 */
910 			int oldoptlen = mopt->m_len;
911 			struct mbuf *n;
912 
913 			/*
914 			 * XXX: give up if the whole (new) hbh header does
915 			 * not fit even in an mbuf cluster.
916 			 */
917 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
918 				return (ENOBUFS);
919 
920 			/*
921 			 * As a consequence, we must always prepare a cluster
922 			 * at this point.
923 			 */
924 			MGET(n, M_DONTWAIT, MT_DATA);
925 			if (n) {
926 				MCLGET(n, M_DONTWAIT);
927 				if ((n->m_flags & M_EXT) == 0) {
928 					m_freem(n);
929 					n = NULL;
930 				}
931 			}
932 			if (!n)
933 				return (ENOBUFS);
934 			n->m_len = oldoptlen + JUMBOOPTLEN;
935 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
936 			      oldoptlen);
937 			optbuf = mtod(n, u_int8_t *) + oldoptlen;
938 			m_freem(mopt);
939 			mopt = exthdrs->ip6e_hbh = n;
940 		} else {
941 			optbuf = mtod(mopt, u_int8_t *) + mopt->m_len;
942 			mopt->m_len += JUMBOOPTLEN;
943 		}
944 		optbuf[0] = IP6OPT_PADN;
945 		optbuf[1] = 0;
946 
947 		/*
948 		 * Adjust the header length according to the pad and
949 		 * the jumbo payload option.
950 		 */
951 		hbh = mtod(mopt, struct ip6_hbh *);
952 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
953 	}
954 
955 	/* fill in the option. */
956 	optbuf[2] = IP6OPT_JUMBO;
957 	optbuf[3] = 4;
958 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
959 	memcpy(&optbuf[4], &v, sizeof(u_int32_t));
960 
961 	/* finally, adjust the packet header length */
962 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
963 
964 	return (0);
965 #undef JUMBOOPTLEN
966 }
967 
968 /*
969  * Insert fragment header and copy unfragmentable header portions.
970  */
971 int
972 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
973     struct ip6_frag **frghdrp)
974 {
975 	struct mbuf *n, *mlast;
976 
977 	if (hlen > sizeof(struct ip6_hdr)) {
978 		n = m_copym(m0, sizeof(struct ip6_hdr),
979 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
980 		if (n == NULL)
981 			return (ENOBUFS);
982 		m->m_next = n;
983 	} else
984 		n = m;
985 
986 	/* Search for the last mbuf of unfragmentable part. */
987 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
988 		;
989 
990 	if ((mlast->m_flags & M_EXT) == 0 &&
991 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
992 		/* use the trailing space of the last mbuf for the fragment hdr */
993 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
994 		    mlast->m_len);
995 		mlast->m_len += sizeof(struct ip6_frag);
996 		m->m_pkthdr.len += sizeof(struct ip6_frag);
997 	} else {
998 		/* allocate a new mbuf for the fragment header */
999 		struct mbuf *mfrg;
1000 
1001 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1002 		if (mfrg == NULL)
1003 			return (ENOBUFS);
1004 		mfrg->m_len = sizeof(struct ip6_frag);
1005 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1006 		mlast->m_next = mfrg;
1007 	}
1008 
1009 	return (0);
1010 }
1011 
1012 int
1013 ip6_getpmtu(struct rtentry *rt, struct ifnet *ifp, u_long *mtup)
1014 {
1015 	u_int32_t mtu = 0;
1016 	int error = 0;
1017 
1018 	if (rt != NULL) {
1019 		mtu = rt->rt_rmx.rmx_mtu;
1020 		if (mtu == 0)
1021 			mtu = ifp->if_mtu;
1022 		else if (mtu < IPV6_MMTU) {
1023 			/* RFC8021 IPv6 Atomic Fragments Considered Harmful */
1024 			mtu = IPV6_MMTU;
1025 		} else if (mtu > ifp->if_mtu) {
1026 			/*
1027 			 * The MTU on the route is larger than the MTU on
1028 			 * the interface!  This shouldn't happen, unless the
1029 			 * MTU of the interface has been changed after the
1030 			 * interface was brought up.  Change the MTU in the
1031 			 * route to match the interface MTU (as long as the
1032 			 * field isn't locked).
1033 			 */
1034 			mtu = ifp->if_mtu;
1035 			if (!(rt->rt_rmx.rmx_locks & RTV_MTU))
1036 				rt->rt_rmx.rmx_mtu = mtu;
1037 		}
1038 	} else {
1039 		mtu = ifp->if_mtu;
1040 	}
1041 
1042 	*mtup = mtu;
1043 	return (error);
1044 }
1045 
1046 /*
1047  * IP6 socket option processing.
1048  */
1049 int
1050 ip6_ctloutput(int op, struct socket *so, int level, int optname,
1051     struct mbuf *m)
1052 {
1053 	int privileged, optdatalen, uproto;
1054 	void *optdata;
1055 	struct inpcb *inp = sotoinpcb(so);
1056 	int error, optval;
1057 	struct proc *p = curproc; /* For IPSec and rdomain */
1058 	u_int rtid = 0;
1059 
1060 	error = optval = 0;
1061 
1062 	privileged = (inp->inp_socket->so_state & SS_PRIV);
1063 	uproto = (int)so->so_proto->pr_protocol;
1064 
1065 	if (level == IPPROTO_IPV6) {
1066 		switch (op) {
1067 		case PRCO_SETOPT:
1068 			switch (optname) {
1069 			/*
1070 			 * Use of some Hop-by-Hop options or some
1071 			 * Destination options, might require special
1072 			 * privilege.  That is, normal applications
1073 			 * (without special privilege) might be forbidden
1074 			 * from setting certain options in outgoing packets,
1075 			 * and might never see certain options in received
1076 			 * packets. [RFC 2292 Section 6]
1077 			 * KAME specific note:
1078 			 *  KAME prevents non-privileged users from sending or
1079 			 *  receiving ANY hbh/dst options in order to avoid
1080 			 *  overhead of parsing options in the kernel.
1081 			 */
1082 			case IPV6_RECVHOPOPTS:
1083 			case IPV6_RECVDSTOPTS:
1084 				if (!privileged) {
1085 					error = EPERM;
1086 					break;
1087 				}
1088 				/* FALLTHROUGH */
1089 			case IPV6_UNICAST_HOPS:
1090 			case IPV6_MINHOPCOUNT:
1091 			case IPV6_HOPLIMIT:
1092 
1093 			case IPV6_RECVPKTINFO:
1094 			case IPV6_RECVHOPLIMIT:
1095 			case IPV6_RECVRTHDR:
1096 			case IPV6_RECVPATHMTU:
1097 			case IPV6_RECVTCLASS:
1098 			case IPV6_V6ONLY:
1099 			case IPV6_AUTOFLOWLABEL:
1100 			case IPV6_RECVDSTPORT:
1101 				if (m == NULL || m->m_len != sizeof(int)) {
1102 					error = EINVAL;
1103 					break;
1104 				}
1105 				optval = *mtod(m, int *);
1106 				switch (optname) {
1107 
1108 				case IPV6_UNICAST_HOPS:
1109 					if (optval < -1 || optval >= 256)
1110 						error = EINVAL;
1111 					else {
1112 						/* -1 = kernel default */
1113 						inp->inp_hops = optval;
1114 					}
1115 					break;
1116 
1117 				case IPV6_MINHOPCOUNT:
1118 					if (optval < 0 || optval > 255)
1119 						error = EINVAL;
1120 					else
1121 						inp->inp_ip6_minhlim = optval;
1122 					break;
1123 
1124 #define OPTSET(bit) \
1125 do { \
1126 	if (optval) \
1127 		inp->inp_flags |= (bit); \
1128 	else \
1129 		inp->inp_flags &= ~(bit); \
1130 } while (/*CONSTCOND*/ 0)
1131 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1132 
1133 				case IPV6_RECVPKTINFO:
1134 					OPTSET(IN6P_PKTINFO);
1135 					break;
1136 
1137 				case IPV6_HOPLIMIT:
1138 				{
1139 					struct ip6_pktopts **optp;
1140 
1141 					optp = &inp->inp_outputopts6;
1142 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1143 							   (u_char *)&optval,
1144 							   sizeof(optval),
1145 							   optp,
1146 							   privileged, uproto);
1147 					break;
1148 				}
1149 
1150 				case IPV6_RECVHOPLIMIT:
1151 					OPTSET(IN6P_HOPLIMIT);
1152 					break;
1153 
1154 				case IPV6_RECVHOPOPTS:
1155 					OPTSET(IN6P_HOPOPTS);
1156 					break;
1157 
1158 				case IPV6_RECVDSTOPTS:
1159 					OPTSET(IN6P_DSTOPTS);
1160 					break;
1161 
1162 				case IPV6_RECVRTHDR:
1163 					OPTSET(IN6P_RTHDR);
1164 					break;
1165 
1166 				case IPV6_RECVPATHMTU:
1167 					/*
1168 					 * We ignore this option for TCP
1169 					 * sockets.
1170 					 * (RFC3542 leaves this case
1171 					 * unspecified.)
1172 					 */
1173 					if (uproto != IPPROTO_TCP)
1174 						OPTSET(IN6P_MTU);
1175 					break;
1176 
1177 				case IPV6_V6ONLY:
1178 					/*
1179 					 * make setsockopt(IPV6_V6ONLY)
1180 					 * available only prior to bind(2).
1181 					 * see ipng mailing list, Jun 22 2001.
1182 					 */
1183 					if (inp->inp_lport ||
1184 					    !IN6_IS_ADDR_UNSPECIFIED(&inp->inp_laddr6)) {
1185 						error = EINVAL;
1186 						break;
1187 					}
1188 					/* No support for IPv4-mapped addresses. */
1189 					if (!optval)
1190 						error = EINVAL;
1191 					else
1192 						error = 0;
1193 					break;
1194 				case IPV6_RECVTCLASS:
1195 					OPTSET(IN6P_TCLASS);
1196 					break;
1197 				case IPV6_AUTOFLOWLABEL:
1198 					OPTSET(IN6P_AUTOFLOWLABEL);
1199 					break;
1200 
1201 				case IPV6_RECVDSTPORT:
1202 					OPTSET(IN6P_RECVDSTPORT);
1203 					break;
1204 				}
1205 				break;
1206 
1207 			case IPV6_TCLASS:
1208 			case IPV6_DONTFRAG:
1209 			case IPV6_USE_MIN_MTU:
1210 				if (m == NULL || m->m_len != sizeof(optval)) {
1211 					error = EINVAL;
1212 					break;
1213 				}
1214 				optval = *mtod(m, int *);
1215 				{
1216 					struct ip6_pktopts **optp;
1217 					optp = &inp->inp_outputopts6;
1218 					error = ip6_pcbopt(optname,
1219 							   (u_char *)&optval,
1220 							   sizeof(optval),
1221 							   optp,
1222 							   privileged, uproto);
1223 					break;
1224 				}
1225 
1226 			case IPV6_PKTINFO:
1227 			case IPV6_HOPOPTS:
1228 			case IPV6_RTHDR:
1229 			case IPV6_DSTOPTS:
1230 			case IPV6_RTHDRDSTOPTS:
1231 			{
1232 				/* new advanced API (RFC3542) */
1233 				u_char *optbuf;
1234 				int optbuflen;
1235 				struct ip6_pktopts **optp;
1236 
1237 				if (m && m->m_next) {
1238 					error = EINVAL;	/* XXX */
1239 					break;
1240 				}
1241 				if (m) {
1242 					optbuf = mtod(m, u_char *);
1243 					optbuflen = m->m_len;
1244 				} else {
1245 					optbuf = NULL;
1246 					optbuflen = 0;
1247 				}
1248 				optp = &inp->inp_outputopts6;
1249 				error = ip6_pcbopt(optname,
1250 						   optbuf, optbuflen,
1251 						   optp, privileged, uproto);
1252 				break;
1253 			}
1254 #undef OPTSET
1255 
1256 			case IPV6_MULTICAST_IF:
1257 			case IPV6_MULTICAST_HOPS:
1258 			case IPV6_MULTICAST_LOOP:
1259 			case IPV6_JOIN_GROUP:
1260 			case IPV6_LEAVE_GROUP:
1261 				error =	ip6_setmoptions(optname,
1262 							&inp->inp_moptions6,
1263 							m);
1264 				break;
1265 
1266 			case IPV6_PORTRANGE:
1267 				if (m == NULL || m->m_len != sizeof(int)) {
1268 					error = EINVAL;
1269 					break;
1270 				}
1271 				optval = *mtod(m, int *);
1272 
1273 				switch (optval) {
1274 				case IPV6_PORTRANGE_DEFAULT:
1275 					inp->inp_flags &= ~(IN6P_LOWPORT);
1276 					inp->inp_flags &= ~(IN6P_HIGHPORT);
1277 					break;
1278 
1279 				case IPV6_PORTRANGE_HIGH:
1280 					inp->inp_flags &= ~(IN6P_LOWPORT);
1281 					inp->inp_flags |= IN6P_HIGHPORT;
1282 					break;
1283 
1284 				case IPV6_PORTRANGE_LOW:
1285 					inp->inp_flags &= ~(IN6P_HIGHPORT);
1286 					inp->inp_flags |= IN6P_LOWPORT;
1287 					break;
1288 
1289 				default:
1290 					error = EINVAL;
1291 					break;
1292 				}
1293 				break;
1294 
1295 			case IPSEC6_OUTSA:
1296 				error = EINVAL;
1297 				break;
1298 
1299 			case IPV6_AUTH_LEVEL:
1300 			case IPV6_ESP_TRANS_LEVEL:
1301 			case IPV6_ESP_NETWORK_LEVEL:
1302 			case IPV6_IPCOMP_LEVEL:
1303 #ifndef IPSEC
1304 				error = EINVAL;
1305 #else
1306 				if (m == NULL || m->m_len != sizeof(int)) {
1307 					error = EINVAL;
1308 					break;
1309 				}
1310 				optval = *mtod(m, int *);
1311 
1312 				if (optval < IPSEC_LEVEL_BYPASS ||
1313 				    optval > IPSEC_LEVEL_UNIQUE) {
1314 					error = EINVAL;
1315 					break;
1316 				}
1317 
1318 				switch (optname) {
1319 				case IPV6_AUTH_LEVEL:
1320 				        if (optval < IPSEC_AUTH_LEVEL_DEFAULT &&
1321 					    suser(p, 0)) {
1322 						error = EACCES;
1323 						break;
1324 					}
1325 					inp->inp_seclevel[SL_AUTH] = optval;
1326 					break;
1327 
1328 				case IPV6_ESP_TRANS_LEVEL:
1329 				        if (optval < IPSEC_ESP_TRANS_LEVEL_DEFAULT &&
1330 					    suser(p, 0)) {
1331 						error = EACCES;
1332 						break;
1333 					}
1334 					inp->inp_seclevel[SL_ESP_TRANS] = optval;
1335 					break;
1336 
1337 				case IPV6_ESP_NETWORK_LEVEL:
1338 				        if (optval < IPSEC_ESP_NETWORK_LEVEL_DEFAULT &&
1339 					    suser(p, 0)) {
1340 						error = EACCES;
1341 						break;
1342 					}
1343 					inp->inp_seclevel[SL_ESP_NETWORK] = optval;
1344 					break;
1345 
1346 				case IPV6_IPCOMP_LEVEL:
1347 				        if (optval < IPSEC_IPCOMP_LEVEL_DEFAULT &&
1348 					    suser(p, 0)) {
1349 						error = EACCES;
1350 						break;
1351 					}
1352 					inp->inp_seclevel[SL_IPCOMP] = optval;
1353 					break;
1354 				}
1355 #endif
1356 				break;
1357 			case SO_RTABLE:
1358 				if (m == NULL || m->m_len < sizeof(u_int)) {
1359 					error = EINVAL;
1360 					break;
1361 				}
1362 				rtid = *mtod(m, u_int *);
1363 				if (inp->inp_rtableid == rtid)
1364 					break;
1365 				/* needs privileges to switch when already set */
1366 				if (p->p_p->ps_rtableid != rtid &&
1367 				    p->p_p->ps_rtableid != 0 &&
1368 				    (error = suser(p, 0)) != 0)
1369 					break;
1370 				/* table must exist */
1371 				if (!rtable_exists(rtid)) {
1372 					error = EINVAL;
1373 					break;
1374 				}
1375 				if (inp->inp_lport) {
1376 					error = EBUSY;
1377 					break;
1378 				}
1379 				inp->inp_rtableid = rtid;
1380 				in_pcbrehash(inp);
1381 				break;
1382 			case IPV6_PIPEX:
1383 				if (m != NULL && m->m_len == sizeof(int))
1384 					inp->inp_pipex = *mtod(m, int *);
1385 				else
1386 					error = EINVAL;
1387 				break;
1388 
1389 			default:
1390 				error = ENOPROTOOPT;
1391 				break;
1392 			}
1393 			m_free(m);
1394 			break;
1395 
1396 		case PRCO_GETOPT:
1397 			switch (optname) {
1398 
1399 			case IPV6_RECVHOPOPTS:
1400 			case IPV6_RECVDSTOPTS:
1401 			case IPV6_UNICAST_HOPS:
1402 			case IPV6_MINHOPCOUNT:
1403 			case IPV6_RECVPKTINFO:
1404 			case IPV6_RECVHOPLIMIT:
1405 			case IPV6_RECVRTHDR:
1406 			case IPV6_RECVPATHMTU:
1407 
1408 			case IPV6_V6ONLY:
1409 			case IPV6_PORTRANGE:
1410 			case IPV6_RECVTCLASS:
1411 			case IPV6_AUTOFLOWLABEL:
1412 			case IPV6_RECVDSTPORT:
1413 				switch (optname) {
1414 
1415 				case IPV6_RECVHOPOPTS:
1416 					optval = OPTBIT(IN6P_HOPOPTS);
1417 					break;
1418 
1419 				case IPV6_RECVDSTOPTS:
1420 					optval = OPTBIT(IN6P_DSTOPTS);
1421 					break;
1422 
1423 				case IPV6_UNICAST_HOPS:
1424 					optval = inp->inp_hops;
1425 					break;
1426 
1427 				case IPV6_MINHOPCOUNT:
1428 					optval = inp->inp_ip6_minhlim;
1429 					break;
1430 
1431 				case IPV6_RECVPKTINFO:
1432 					optval = OPTBIT(IN6P_PKTINFO);
1433 					break;
1434 
1435 				case IPV6_RECVHOPLIMIT:
1436 					optval = OPTBIT(IN6P_HOPLIMIT);
1437 					break;
1438 
1439 				case IPV6_RECVRTHDR:
1440 					optval = OPTBIT(IN6P_RTHDR);
1441 					break;
1442 
1443 				case IPV6_RECVPATHMTU:
1444 					optval = OPTBIT(IN6P_MTU);
1445 					break;
1446 
1447 				case IPV6_V6ONLY:
1448 					optval = 1;
1449 					break;
1450 
1451 				case IPV6_PORTRANGE:
1452 				    {
1453 					int flags;
1454 					flags = inp->inp_flags;
1455 					if (flags & IN6P_HIGHPORT)
1456 						optval = IPV6_PORTRANGE_HIGH;
1457 					else if (flags & IN6P_LOWPORT)
1458 						optval = IPV6_PORTRANGE_LOW;
1459 					else
1460 						optval = 0;
1461 					break;
1462 				    }
1463 				case IPV6_RECVTCLASS:
1464 					optval = OPTBIT(IN6P_TCLASS);
1465 					break;
1466 
1467 				case IPV6_AUTOFLOWLABEL:
1468 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1469 					break;
1470 
1471 				case IPV6_RECVDSTPORT:
1472 					optval = OPTBIT(IN6P_RECVDSTPORT);
1473 					break;
1474 				}
1475 				if (error)
1476 					break;
1477 				m->m_len = sizeof(int);
1478 				*mtod(m, int *) = optval;
1479 				break;
1480 
1481 			case IPV6_PATHMTU:
1482 			{
1483 				u_long pmtu = 0;
1484 				struct ip6_mtuinfo mtuinfo;
1485 				struct ifnet *ifp;
1486 				struct rtentry *rt;
1487 
1488 				if (!(so->so_state & SS_ISCONNECTED))
1489 					return (ENOTCONN);
1490 
1491 				rt = in_pcbrtentry(inp);
1492 				if (!rtisvalid(rt))
1493 					return (EHOSTUNREACH);
1494 
1495 				ifp = if_get(rt->rt_ifidx);
1496 				if (ifp == NULL)
1497 					return (EHOSTUNREACH);
1498 				/*
1499 				 * XXX: we dot not consider the case of source
1500 				 * routing, or optional information to specify
1501 				 * the outgoing interface.
1502 				 */
1503 				error = ip6_getpmtu(rt, ifp, &pmtu);
1504 				if_put(ifp);
1505 				if (error)
1506 					break;
1507 				if (pmtu > IPV6_MAXPACKET)
1508 					pmtu = IPV6_MAXPACKET;
1509 
1510 				bzero(&mtuinfo, sizeof(mtuinfo));
1511 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1512 				optdata = (void *)&mtuinfo;
1513 				optdatalen = sizeof(mtuinfo);
1514 				if (optdatalen > MCLBYTES)
1515 					return (EMSGSIZE); /* XXX */
1516 				if (optdatalen > MLEN)
1517 					MCLGET(m, M_WAIT);
1518 				m->m_len = optdatalen;
1519 				bcopy(optdata, mtod(m, void *), optdatalen);
1520 				break;
1521 			}
1522 
1523 			case IPV6_PKTINFO:
1524 			case IPV6_HOPOPTS:
1525 			case IPV6_RTHDR:
1526 			case IPV6_DSTOPTS:
1527 			case IPV6_RTHDRDSTOPTS:
1528 			case IPV6_TCLASS:
1529 			case IPV6_DONTFRAG:
1530 			case IPV6_USE_MIN_MTU:
1531 				error = ip6_getpcbopt(inp->inp_outputopts6,
1532 				    optname, m);
1533 				break;
1534 
1535 			case IPV6_MULTICAST_IF:
1536 			case IPV6_MULTICAST_HOPS:
1537 			case IPV6_MULTICAST_LOOP:
1538 			case IPV6_JOIN_GROUP:
1539 			case IPV6_LEAVE_GROUP:
1540 				error = ip6_getmoptions(optname,
1541 				    inp->inp_moptions6, m);
1542 				break;
1543 
1544 			case IPSEC6_OUTSA:
1545 				error = EINVAL;
1546 				break;
1547 
1548 			case IPV6_AUTH_LEVEL:
1549 			case IPV6_ESP_TRANS_LEVEL:
1550 			case IPV6_ESP_NETWORK_LEVEL:
1551 			case IPV6_IPCOMP_LEVEL:
1552 #ifndef IPSEC
1553 				m->m_len = sizeof(int);
1554 				*mtod(m, int *) = IPSEC_LEVEL_NONE;
1555 #else
1556 				m->m_len = sizeof(int);
1557 				switch (optname) {
1558 				case IPV6_AUTH_LEVEL:
1559 					optval = inp->inp_seclevel[SL_AUTH];
1560 					break;
1561 
1562 				case IPV6_ESP_TRANS_LEVEL:
1563 					optval =
1564 					    inp->inp_seclevel[SL_ESP_TRANS];
1565 					break;
1566 
1567 				case IPV6_ESP_NETWORK_LEVEL:
1568 					optval =
1569 					    inp->inp_seclevel[SL_ESP_NETWORK];
1570 					break;
1571 
1572 				case IPV6_IPCOMP_LEVEL:
1573 					optval = inp->inp_seclevel[SL_IPCOMP];
1574 					break;
1575 				}
1576 				*mtod(m, int *) = optval;
1577 #endif
1578 				break;
1579 			case SO_RTABLE:
1580 				m->m_len = sizeof(u_int);
1581 				*mtod(m, u_int *) = optval;
1582 				break;
1583 			case IPV6_PIPEX:
1584 				m->m_len = sizeof(int);
1585 				*mtod(m, int *) = optval;
1586 				break;
1587 
1588 			default:
1589 				error = ENOPROTOOPT;
1590 				break;
1591 			}
1592 			break;
1593 		}
1594 	} else {
1595 		error = EINVAL;
1596 		if (op == PRCO_SETOPT)
1597 			(void)m_free(m);
1598 	}
1599 	return (error);
1600 }
1601 
1602 int
1603 ip6_raw_ctloutput(int op, struct socket *so, int level, int optname,
1604     struct mbuf *m)
1605 {
1606 	int error = 0, optval;
1607 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
1608 	struct inpcb *inp = sotoinpcb(so);
1609 
1610 	if (level != IPPROTO_IPV6) {
1611 		if (op == PRCO_SETOPT)
1612 			(void)m_free(m);
1613 		return (EINVAL);
1614 	}
1615 
1616 	switch (optname) {
1617 	case IPV6_CHECKSUM:
1618 		/*
1619 		 * For ICMPv6 sockets, no modification allowed for checksum
1620 		 * offset, permit "no change" values to help existing apps.
1621 		 *
1622 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
1623 		 * for an ICMPv6 socket will fail."
1624 		 * The current behavior does not meet RFC3542.
1625 		 */
1626 		switch (op) {
1627 		case PRCO_SETOPT:
1628 			if (m == NULL || m->m_len != sizeof(int)) {
1629 				error = EINVAL;
1630 				break;
1631 			}
1632 			optval = *mtod(m, int *);
1633 			if ((optval % 2) != 0) {
1634 				/* the API assumes even offset values */
1635 				error = EINVAL;
1636 			} else if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) {
1637 				if (optval != icmp6off)
1638 					error = EINVAL;
1639 			} else
1640 				inp->inp_cksum6 = optval;
1641 			break;
1642 
1643 		case PRCO_GETOPT:
1644 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
1645 				optval = icmp6off;
1646 			else
1647 				optval = inp->inp_cksum6;
1648 
1649 			m->m_len = sizeof(int);
1650 			*mtod(m, int *) = optval;
1651 			break;
1652 
1653 		default:
1654 			error = EINVAL;
1655 			break;
1656 		}
1657 		break;
1658 
1659 	default:
1660 		error = ENOPROTOOPT;
1661 		break;
1662 	}
1663 
1664 	if (op == PRCO_SETOPT)
1665 		(void)m_free(m);
1666 
1667 	return (error);
1668 }
1669 
1670 /*
1671  * initialize ip6_pktopts.  beware that there are non-zero default values in
1672  * the struct.
1673  */
1674 void
1675 ip6_initpktopts(struct ip6_pktopts *opt)
1676 {
1677 	bzero(opt, sizeof(*opt));
1678 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
1679 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
1680 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
1681 }
1682 
1683 int
1684 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
1685     int priv, int uproto)
1686 {
1687 	struct ip6_pktopts *opt;
1688 
1689 	if (*pktopt == NULL) {
1690 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
1691 		    M_WAITOK);
1692 		ip6_initpktopts(*pktopt);
1693 	}
1694 	opt = *pktopt;
1695 
1696 	return (ip6_setpktopt(optname, buf, len, opt, priv, 1, uproto));
1697 }
1698 
1699 int
1700 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct mbuf *m)
1701 {
1702 	void *optdata = NULL;
1703 	int optdatalen = 0;
1704 	struct ip6_ext *ip6e;
1705 	int error = 0;
1706 	struct in6_pktinfo null_pktinfo;
1707 	int deftclass = 0, on;
1708 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
1709 
1710 	switch (optname) {
1711 	case IPV6_PKTINFO:
1712 		if (pktopt && pktopt->ip6po_pktinfo)
1713 			optdata = (void *)pktopt->ip6po_pktinfo;
1714 		else {
1715 			/* XXX: we don't have to do this every time... */
1716 			bzero(&null_pktinfo, sizeof(null_pktinfo));
1717 			optdata = (void *)&null_pktinfo;
1718 		}
1719 		optdatalen = sizeof(struct in6_pktinfo);
1720 		break;
1721 	case IPV6_TCLASS:
1722 		if (pktopt && pktopt->ip6po_tclass >= 0)
1723 			optdata = (void *)&pktopt->ip6po_tclass;
1724 		else
1725 			optdata = (void *)&deftclass;
1726 		optdatalen = sizeof(int);
1727 		break;
1728 	case IPV6_HOPOPTS:
1729 		if (pktopt && pktopt->ip6po_hbh) {
1730 			optdata = (void *)pktopt->ip6po_hbh;
1731 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
1732 			optdatalen = (ip6e->ip6e_len + 1) << 3;
1733 		}
1734 		break;
1735 	case IPV6_RTHDR:
1736 		if (pktopt && pktopt->ip6po_rthdr) {
1737 			optdata = (void *)pktopt->ip6po_rthdr;
1738 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
1739 			optdatalen = (ip6e->ip6e_len + 1) << 3;
1740 		}
1741 		break;
1742 	case IPV6_RTHDRDSTOPTS:
1743 		if (pktopt && pktopt->ip6po_dest1) {
1744 			optdata = (void *)pktopt->ip6po_dest1;
1745 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
1746 			optdatalen = (ip6e->ip6e_len + 1) << 3;
1747 		}
1748 		break;
1749 	case IPV6_DSTOPTS:
1750 		if (pktopt && pktopt->ip6po_dest2) {
1751 			optdata = (void *)pktopt->ip6po_dest2;
1752 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
1753 			optdatalen = (ip6e->ip6e_len + 1) << 3;
1754 		}
1755 		break;
1756 	case IPV6_USE_MIN_MTU:
1757 		if (pktopt)
1758 			optdata = (void *)&pktopt->ip6po_minmtu;
1759 		else
1760 			optdata = (void *)&defminmtu;
1761 		optdatalen = sizeof(int);
1762 		break;
1763 	case IPV6_DONTFRAG:
1764 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
1765 			on = 1;
1766 		else
1767 			on = 0;
1768 		optdata = (void *)&on;
1769 		optdatalen = sizeof(on);
1770 		break;
1771 	default:		/* should not happen */
1772 #ifdef DIAGNOSTIC
1773 		panic("ip6_getpcbopt: unexpected option");
1774 #endif
1775 		return (ENOPROTOOPT);
1776 	}
1777 
1778 	if (optdatalen > MCLBYTES)
1779 		return (EMSGSIZE); /* XXX */
1780 	if (optdatalen > MLEN)
1781 		MCLGET(m, M_WAIT);
1782 	m->m_len = optdatalen;
1783 	if (optdatalen)
1784 		bcopy(optdata, mtod(m, void *), optdatalen);
1785 
1786 	return (error);
1787 }
1788 
1789 void
1790 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
1791 {
1792 	if (optname == -1 || optname == IPV6_PKTINFO) {
1793 		if (pktopt->ip6po_pktinfo)
1794 			free(pktopt->ip6po_pktinfo, M_IP6OPT, 0);
1795 		pktopt->ip6po_pktinfo = NULL;
1796 	}
1797 	if (optname == -1 || optname == IPV6_HOPLIMIT)
1798 		pktopt->ip6po_hlim = -1;
1799 	if (optname == -1 || optname == IPV6_TCLASS)
1800 		pktopt->ip6po_tclass = -1;
1801 	if (optname == -1 || optname == IPV6_HOPOPTS) {
1802 		if (pktopt->ip6po_hbh)
1803 			free(pktopt->ip6po_hbh, M_IP6OPT, 0);
1804 		pktopt->ip6po_hbh = NULL;
1805 	}
1806 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
1807 		if (pktopt->ip6po_dest1)
1808 			free(pktopt->ip6po_dest1, M_IP6OPT, 0);
1809 		pktopt->ip6po_dest1 = NULL;
1810 	}
1811 	if (optname == -1 || optname == IPV6_RTHDR) {
1812 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
1813 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT, 0);
1814 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
1815 		if (pktopt->ip6po_route.ro_rt) {
1816 			rtfree(pktopt->ip6po_route.ro_rt);
1817 			pktopt->ip6po_route.ro_rt = NULL;
1818 		}
1819 	}
1820 	if (optname == -1 || optname == IPV6_DSTOPTS) {
1821 		if (pktopt->ip6po_dest2)
1822 			free(pktopt->ip6po_dest2, M_IP6OPT, 0);
1823 		pktopt->ip6po_dest2 = NULL;
1824 	}
1825 }
1826 
1827 #define PKTOPT_EXTHDRCPY(type) \
1828 do {\
1829 	if (src->type) {\
1830 		size_t hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
1831 		dst->type = malloc(hlen, M_IP6OPT, M_NOWAIT);\
1832 		if (dst->type == NULL)\
1833 			goto bad;\
1834 		memcpy(dst->type, src->type, hlen);\
1835 	}\
1836 } while (/*CONSTCOND*/ 0)
1837 
1838 int
1839 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src)
1840 {
1841 	dst->ip6po_hlim = src->ip6po_hlim;
1842 	dst->ip6po_tclass = src->ip6po_tclass;
1843 	dst->ip6po_flags = src->ip6po_flags;
1844 	if (src->ip6po_pktinfo) {
1845 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
1846 		    M_IP6OPT, M_NOWAIT);
1847 		if (dst->ip6po_pktinfo == NULL)
1848 			goto bad;
1849 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
1850 	}
1851 	PKTOPT_EXTHDRCPY(ip6po_hbh);
1852 	PKTOPT_EXTHDRCPY(ip6po_dest1);
1853 	PKTOPT_EXTHDRCPY(ip6po_dest2);
1854 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
1855 	return (0);
1856 
1857   bad:
1858 	ip6_clearpktopts(dst, -1);
1859 	return (ENOBUFS);
1860 }
1861 #undef PKTOPT_EXTHDRCPY
1862 
1863 void
1864 ip6_freepcbopts(struct ip6_pktopts *pktopt)
1865 {
1866 	if (pktopt == NULL)
1867 		return;
1868 
1869 	ip6_clearpktopts(pktopt, -1);
1870 
1871 	free(pktopt, M_IP6OPT, 0);
1872 }
1873 
1874 /*
1875  * Set the IP6 multicast options in response to user setsockopt().
1876  */
1877 int
1878 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
1879 {
1880 	int error = 0;
1881 	u_int loop, ifindex;
1882 	struct ipv6_mreq *mreq;
1883 	struct ifnet *ifp;
1884 	struct ip6_moptions *im6o = *im6op;
1885 	struct in6_multi_mship *imm;
1886 	struct proc *p = curproc;	/* XXX */
1887 
1888 	if (im6o == NULL) {
1889 		/*
1890 		 * No multicast option buffer attached to the pcb;
1891 		 * allocate one and initialize to default values.
1892 		 */
1893 		im6o = (struct ip6_moptions *)
1894 			malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
1895 
1896 		if (im6o == NULL)
1897 			return (ENOBUFS);
1898 		*im6op = im6o;
1899 		im6o->im6o_ifidx = 0;
1900 		im6o->im6o_hlim = ip6_defmcasthlim;
1901 		im6o->im6o_loop = IPV6_DEFAULT_MULTICAST_LOOP;
1902 		LIST_INIT(&im6o->im6o_memberships);
1903 	}
1904 
1905 	switch (optname) {
1906 
1907 	case IPV6_MULTICAST_IF:
1908 		/*
1909 		 * Select the interface for outgoing multicast packets.
1910 		 */
1911 		if (m == NULL || m->m_len != sizeof(u_int)) {
1912 			error = EINVAL;
1913 			break;
1914 		}
1915 		memcpy(&ifindex, mtod(m, u_int *), sizeof(ifindex));
1916 		if (ifindex != 0) {
1917 			ifp = if_get(ifindex);
1918 			if (ifp == NULL) {
1919 				error = ENXIO;	/* XXX EINVAL? */
1920 				break;
1921 			}
1922 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
1923 				error = EADDRNOTAVAIL;
1924 				if_put(ifp);
1925 				break;
1926 			}
1927 			if_put(ifp);
1928 		}
1929 		im6o->im6o_ifidx = ifindex;
1930 		break;
1931 
1932 	case IPV6_MULTICAST_HOPS:
1933 	    {
1934 		/*
1935 		 * Set the IP6 hoplimit for outgoing multicast packets.
1936 		 */
1937 		int optval;
1938 		if (m == NULL || m->m_len != sizeof(int)) {
1939 			error = EINVAL;
1940 			break;
1941 		}
1942 		memcpy(&optval, mtod(m, u_int *), sizeof(optval));
1943 		if (optval < -1 || optval >= 256)
1944 			error = EINVAL;
1945 		else if (optval == -1)
1946 			im6o->im6o_hlim = ip6_defmcasthlim;
1947 		else
1948 			im6o->im6o_hlim = optval;
1949 		break;
1950 	    }
1951 
1952 	case IPV6_MULTICAST_LOOP:
1953 		/*
1954 		 * Set the loopback flag for outgoing multicast packets.
1955 		 * Must be zero or one.
1956 		 */
1957 		if (m == NULL || m->m_len != sizeof(u_int)) {
1958 			error = EINVAL;
1959 			break;
1960 		}
1961 		memcpy(&loop, mtod(m, u_int *), sizeof(loop));
1962 		if (loop > 1) {
1963 			error = EINVAL;
1964 			break;
1965 		}
1966 		im6o->im6o_loop = loop;
1967 		break;
1968 
1969 	case IPV6_JOIN_GROUP:
1970 		/*
1971 		 * Add a multicast group membership.
1972 		 * Group must be a valid IP6 multicast address.
1973 		 */
1974 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
1975 			error = EINVAL;
1976 			break;
1977 		}
1978 		mreq = mtod(m, struct ipv6_mreq *);
1979 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
1980 			/*
1981 			 * We use the unspecified address to specify to accept
1982 			 * all multicast addresses. Only super user is allowed
1983 			 * to do this.
1984 			 */
1985 			if (suser(p, 0))
1986 			{
1987 				error = EACCES;
1988 				break;
1989 			}
1990 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
1991 			error = EINVAL;
1992 			break;
1993 		}
1994 
1995 		/*
1996 		 * If no interface was explicitly specified, choose an
1997 		 * appropriate one according to the given multicast address.
1998 		 */
1999 		if (mreq->ipv6mr_interface == 0) {
2000 			struct rtentry *rt;
2001 			struct sockaddr_in6 dst;
2002 
2003 			memset(&dst, 0, sizeof(dst));
2004 			dst.sin6_len = sizeof(dst);
2005 			dst.sin6_family = AF_INET6;
2006 			dst.sin6_addr = mreq->ipv6mr_multiaddr;
2007 			rt = rtalloc(sin6tosa(&dst), RT_RESOLVE,
2008 			    m->m_pkthdr.ph_rtableid);
2009 			if (rt == NULL) {
2010 				error = EADDRNOTAVAIL;
2011 				break;
2012 			}
2013 			ifp = if_get(rt->rt_ifidx);
2014 			rtfree(rt);
2015 		} else {
2016 			/*
2017 			 * If the interface is specified, validate it.
2018 			 */
2019 			ifp = if_get(mreq->ipv6mr_interface);
2020 			if (ifp == NULL) {
2021 				error = ENXIO;	/* XXX EINVAL? */
2022 				break;
2023 			}
2024 		}
2025 
2026 		/*
2027 		 * See if we found an interface, and confirm that it
2028 		 * supports multicast
2029 		 */
2030 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2031 			if_put(ifp);
2032 			error = EADDRNOTAVAIL;
2033 			break;
2034 		}
2035 		/*
2036 		 * Put interface index into the multicast address,
2037 		 * if the address has link/interface-local scope.
2038 		 */
2039 		if (IN6_IS_SCOPE_EMBED(&mreq->ipv6mr_multiaddr)) {
2040 			mreq->ipv6mr_multiaddr.s6_addr16[1] =
2041 			    htons(ifp->if_index);
2042 		}
2043 		/*
2044 		 * See if the membership already exists.
2045 		 */
2046 		LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain)
2047 			if (imm->i6mm_maddr->in6m_ifidx == ifp->if_index &&
2048 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2049 			    &mreq->ipv6mr_multiaddr))
2050 				break;
2051 		if (imm != NULL) {
2052 			if_put(ifp);
2053 			error = EADDRINUSE;
2054 			break;
2055 		}
2056 		/*
2057 		 * Everything looks good; add a new record to the multicast
2058 		 * address list for the given interface.
2059 		 */
2060 		imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error);
2061 		if_put(ifp);
2062 		if (!imm)
2063 			break;
2064 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2065 		break;
2066 
2067 	case IPV6_LEAVE_GROUP:
2068 		/*
2069 		 * Drop a multicast group membership.
2070 		 * Group must be a valid IP6 multicast address.
2071 		 */
2072 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2073 			error = EINVAL;
2074 			break;
2075 		}
2076 		mreq = mtod(m, struct ipv6_mreq *);
2077 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2078 			if (suser(p, 0))
2079 			{
2080 				error = EACCES;
2081 				break;
2082 			}
2083 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2084 			error = EINVAL;
2085 			break;
2086 		}
2087 
2088 		/*
2089 		 * Put interface index into the multicast address,
2090 		 * if the address has link-local scope.
2091 		 */
2092 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2093 			mreq->ipv6mr_multiaddr.s6_addr16[1] =
2094 			    htons(mreq->ipv6mr_interface);
2095 		}
2096 
2097 		/*
2098 		 * If an interface address was specified, get a pointer
2099 		 * to its ifnet structure.
2100 		 */
2101 		if (mreq->ipv6mr_interface == 0)
2102 			ifp = NULL;
2103 		else {
2104 			ifp = if_get(mreq->ipv6mr_interface);
2105 			if (ifp == NULL) {
2106 				error = ENXIO;	/* XXX EINVAL? */
2107 				break;
2108 			}
2109 		}
2110 
2111 		/*
2112 		 * Find the membership in the membership list.
2113 		 */
2114 		LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain) {
2115 			if ((ifp == NULL ||
2116 			    imm->i6mm_maddr->in6m_ifidx == ifp->if_index) &&
2117 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2118 			    &mreq->ipv6mr_multiaddr))
2119 				break;
2120 		}
2121 
2122 		if_put(ifp);
2123 
2124 		if (imm == NULL) {
2125 			/* Unable to resolve interface */
2126 			error = EADDRNOTAVAIL;
2127 			break;
2128 		}
2129 		/*
2130 		 * Give up the multicast address record to which the
2131 		 * membership points.
2132 		 */
2133 		LIST_REMOVE(imm, i6mm_chain);
2134 		in6_leavegroup(imm);
2135 		break;
2136 
2137 	default:
2138 		error = EOPNOTSUPP;
2139 		break;
2140 	}
2141 
2142 	/*
2143 	 * If all options have default values, no need to keep the option
2144 	 * structure.
2145 	 */
2146 	if (im6o->im6o_ifidx == 0 &&
2147 	    im6o->im6o_hlim == ip6_defmcasthlim &&
2148 	    im6o->im6o_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2149 	    LIST_EMPTY(&im6o->im6o_memberships)) {
2150 		free(*im6op, M_IPMOPTS, 0);
2151 		*im6op = NULL;
2152 	}
2153 
2154 	return (error);
2155 }
2156 
2157 /*
2158  * Return the IP6 multicast options in response to user getsockopt().
2159  */
2160 int
2161 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf *m)
2162 {
2163 	u_int *hlim, *loop, *ifindex;
2164 
2165 	switch (optname) {
2166 	case IPV6_MULTICAST_IF:
2167 		ifindex = mtod(m, u_int *);
2168 		m->m_len = sizeof(u_int);
2169 		if (im6o == NULL || im6o->im6o_ifidx == 0)
2170 			*ifindex = 0;
2171 		else
2172 			*ifindex = im6o->im6o_ifidx;
2173 		return (0);
2174 
2175 	case IPV6_MULTICAST_HOPS:
2176 		hlim = mtod(m, u_int *);
2177 		m->m_len = sizeof(u_int);
2178 		if (im6o == NULL)
2179 			*hlim = ip6_defmcasthlim;
2180 		else
2181 			*hlim = im6o->im6o_hlim;
2182 		return (0);
2183 
2184 	case IPV6_MULTICAST_LOOP:
2185 		loop = mtod(m, u_int *);
2186 		m->m_len = sizeof(u_int);
2187 		if (im6o == NULL)
2188 			*loop = ip6_defmcasthlim;
2189 		else
2190 			*loop = im6o->im6o_loop;
2191 		return (0);
2192 
2193 	default:
2194 		return (EOPNOTSUPP);
2195 	}
2196 }
2197 
2198 /*
2199  * Discard the IP6 multicast options.
2200  */
2201 void
2202 ip6_freemoptions(struct ip6_moptions *im6o)
2203 {
2204 	struct in6_multi_mship *imm;
2205 
2206 	if (im6o == NULL)
2207 		return;
2208 
2209 	while (!LIST_EMPTY(&im6o->im6o_memberships)) {
2210 		imm = LIST_FIRST(&im6o->im6o_memberships);
2211 		LIST_REMOVE(imm, i6mm_chain);
2212 		in6_leavegroup(imm);
2213 	}
2214 	free(im6o, M_IPMOPTS, 0);
2215 }
2216 
2217 /*
2218  * Set IPv6 outgoing packet options based on advanced API.
2219  */
2220 int
2221 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2222     struct ip6_pktopts *stickyopt, int priv, int uproto)
2223 {
2224 	u_int clen;
2225 	struct cmsghdr *cm = 0;
2226 	caddr_t cmsgs;
2227 	int error;
2228 
2229 	if (control == NULL || opt == NULL)
2230 		return (EINVAL);
2231 
2232 	ip6_initpktopts(opt);
2233 	if (stickyopt) {
2234 		int error;
2235 
2236 		/*
2237 		 * If stickyopt is provided, make a local copy of the options
2238 		 * for this particular packet, then override them by ancillary
2239 		 * objects.
2240 		 * XXX: copypktopts() does not copy the cached route to a next
2241 		 * hop (if any).  This is not very good in terms of efficiency,
2242 		 * but we can allow this since this option should be rarely
2243 		 * used.
2244 		 */
2245 		if ((error = copypktopts(opt, stickyopt)) != 0)
2246 			return (error);
2247 	}
2248 
2249 	/*
2250 	 * XXX: Currently, we assume all the optional information is stored
2251 	 * in a single mbuf.
2252 	 */
2253 	if (control->m_next)
2254 		return (EINVAL);
2255 
2256 	clen = control->m_len;
2257 	cmsgs = mtod(control, caddr_t);
2258 	do {
2259 		if (clen < CMSG_LEN(0))
2260 			return (EINVAL);
2261 		cm = (struct cmsghdr *)cmsgs;
2262 		if (cm->cmsg_len < CMSG_LEN(0) || cm->cmsg_len > clen ||
2263 		    CMSG_ALIGN(cm->cmsg_len) > clen)
2264 			return (EINVAL);
2265 		if (cm->cmsg_level == IPPROTO_IPV6) {
2266 			error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2267 			    cm->cmsg_len - CMSG_LEN(0), opt, priv, 0, uproto);
2268 			if (error)
2269 				return (error);
2270 		}
2271 
2272 		clen -= CMSG_ALIGN(cm->cmsg_len);
2273 		cmsgs += CMSG_ALIGN(cm->cmsg_len);
2274 	} while (clen);
2275 
2276 	return (0);
2277 }
2278 
2279 /*
2280  * Set a particular packet option, as a sticky option or an ancillary data
2281  * item.  "len" can be 0 only when it's a sticky option.
2282  */
2283 int
2284 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2285     int priv, int sticky, int uproto)
2286 {
2287 	int minmtupolicy;
2288 
2289 	switch (optname) {
2290 	case IPV6_PKTINFO:
2291 	{
2292 		struct ifnet *ifp = NULL;
2293 		struct in6_pktinfo *pktinfo;
2294 
2295 		if (len != sizeof(struct in6_pktinfo))
2296 			return (EINVAL);
2297 
2298 		pktinfo = (struct in6_pktinfo *)buf;
2299 
2300 		/*
2301 		 * An application can clear any sticky IPV6_PKTINFO option by
2302 		 * doing a "regular" setsockopt with ipi6_addr being
2303 		 * in6addr_any and ipi6_ifindex being zero.
2304 		 * [RFC 3542, Section 6]
2305 		 */
2306 		if (opt->ip6po_pktinfo &&
2307 		    pktinfo->ipi6_ifindex == 0 &&
2308 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2309 			ip6_clearpktopts(opt, optname);
2310 			break;
2311 		}
2312 
2313 		if (uproto == IPPROTO_TCP &&
2314 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2315 			return (EINVAL);
2316 		}
2317 
2318 		if (pktinfo->ipi6_ifindex) {
2319 			ifp = if_get(pktinfo->ipi6_ifindex);
2320 			if (ifp == NULL)
2321 				return (ENXIO);
2322 			if_put(ifp);
2323 		}
2324 
2325 		/*
2326 		 * We store the address anyway, and let in6_selectsrc()
2327 		 * validate the specified address.  This is because ipi6_addr
2328 		 * may not have enough information about its scope zone, and
2329 		 * we may need additional information (such as outgoing
2330 		 * interface or the scope zone of a destination address) to
2331 		 * disambiguate the scope.
2332 		 * XXX: the delay of the validation may confuse the
2333 		 * application when it is used as a sticky option.
2334 		 */
2335 		if (opt->ip6po_pktinfo == NULL) {
2336 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2337 			    M_IP6OPT, M_NOWAIT);
2338 			if (opt->ip6po_pktinfo == NULL)
2339 				return (ENOBUFS);
2340 		}
2341 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2342 		break;
2343 	}
2344 
2345 	case IPV6_HOPLIMIT:
2346 	{
2347 		int *hlimp;
2348 
2349 		/*
2350 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2351 		 * to simplify the ordering among hoplimit options.
2352 		 */
2353 		if (sticky)
2354 			return (ENOPROTOOPT);
2355 
2356 		if (len != sizeof(int))
2357 			return (EINVAL);
2358 		hlimp = (int *)buf;
2359 		if (*hlimp < -1 || *hlimp > 255)
2360 			return (EINVAL);
2361 
2362 		opt->ip6po_hlim = *hlimp;
2363 		break;
2364 	}
2365 
2366 	case IPV6_TCLASS:
2367 	{
2368 		int tclass;
2369 
2370 		if (len != sizeof(int))
2371 			return (EINVAL);
2372 		tclass = *(int *)buf;
2373 		if (tclass < -1 || tclass > 255)
2374 			return (EINVAL);
2375 
2376 		opt->ip6po_tclass = tclass;
2377 		break;
2378 	}
2379 	case IPV6_HOPOPTS:
2380 	{
2381 		struct ip6_hbh *hbh;
2382 		int hbhlen;
2383 
2384 		/*
2385 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2386 		 * options, since per-option restriction has too much
2387 		 * overhead.
2388 		 */
2389 		if (!priv)
2390 			return (EPERM);
2391 
2392 		if (len == 0) {
2393 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2394 			break;	/* just remove the option */
2395 		}
2396 
2397 		/* message length validation */
2398 		if (len < sizeof(struct ip6_hbh))
2399 			return (EINVAL);
2400 		hbh = (struct ip6_hbh *)buf;
2401 		hbhlen = (hbh->ip6h_len + 1) << 3;
2402 		if (len != hbhlen)
2403 			return (EINVAL);
2404 
2405 		/* turn off the previous option, then set the new option. */
2406 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2407 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2408 		if (opt->ip6po_hbh == NULL)
2409 			return (ENOBUFS);
2410 		memcpy(opt->ip6po_hbh, hbh, hbhlen);
2411 
2412 		break;
2413 	}
2414 
2415 	case IPV6_DSTOPTS:
2416 	case IPV6_RTHDRDSTOPTS:
2417 	{
2418 		struct ip6_dest *dest, **newdest = NULL;
2419 		int destlen;
2420 
2421 		if (!priv)	/* XXX: see the comment for IPV6_HOPOPTS */
2422 			return (EPERM);
2423 
2424 		if (len == 0) {
2425 			ip6_clearpktopts(opt, optname);
2426 			break;	/* just remove the option */
2427 		}
2428 
2429 		/* message length validation */
2430 		if (len < sizeof(struct ip6_dest))
2431 			return (EINVAL);
2432 		dest = (struct ip6_dest *)buf;
2433 		destlen = (dest->ip6d_len + 1) << 3;
2434 		if (len != destlen)
2435 			return (EINVAL);
2436 		/*
2437 		 * Determine the position that the destination options header
2438 		 * should be inserted; before or after the routing header.
2439 		 */
2440 		switch (optname) {
2441 		case IPV6_RTHDRDSTOPTS:
2442 			newdest = &opt->ip6po_dest1;
2443 			break;
2444 		case IPV6_DSTOPTS:
2445 			newdest = &opt->ip6po_dest2;
2446 			break;
2447 		}
2448 
2449 		/* turn off the previous option, then set the new option. */
2450 		ip6_clearpktopts(opt, optname);
2451 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2452 		if (*newdest == NULL)
2453 			return (ENOBUFS);
2454 		memcpy(*newdest, dest, destlen);
2455 
2456 		break;
2457 	}
2458 
2459 	case IPV6_RTHDR:
2460 	{
2461 		struct ip6_rthdr *rth;
2462 		int rthlen;
2463 
2464 		if (len == 0) {
2465 			ip6_clearpktopts(opt, IPV6_RTHDR);
2466 			break;	/* just remove the option */
2467 		}
2468 
2469 		/* message length validation */
2470 		if (len < sizeof(struct ip6_rthdr))
2471 			return (EINVAL);
2472 		rth = (struct ip6_rthdr *)buf;
2473 		rthlen = (rth->ip6r_len + 1) << 3;
2474 		if (len != rthlen)
2475 			return (EINVAL);
2476 
2477 		switch (rth->ip6r_type) {
2478 		case IPV6_RTHDR_TYPE_0:
2479 			if (rth->ip6r_len == 0)	/* must contain one addr */
2480 				return (EINVAL);
2481 			if (rth->ip6r_len % 2) /* length must be even */
2482 				return (EINVAL);
2483 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2484 				return (EINVAL);
2485 			break;
2486 		default:
2487 			return (EINVAL);	/* not supported */
2488 		}
2489 		/* turn off the previous option */
2490 		ip6_clearpktopts(opt, IPV6_RTHDR);
2491 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2492 		if (opt->ip6po_rthdr == NULL)
2493 			return (ENOBUFS);
2494 		memcpy(opt->ip6po_rthdr, rth, rthlen);
2495 		break;
2496 	}
2497 
2498 	case IPV6_USE_MIN_MTU:
2499 		if (len != sizeof(int))
2500 			return (EINVAL);
2501 		minmtupolicy = *(int *)buf;
2502 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2503 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
2504 		    minmtupolicy != IP6PO_MINMTU_ALL) {
2505 			return (EINVAL);
2506 		}
2507 		opt->ip6po_minmtu = minmtupolicy;
2508 		break;
2509 
2510 	case IPV6_DONTFRAG:
2511 		if (len != sizeof(int))
2512 			return (EINVAL);
2513 
2514 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2515 			/*
2516 			 * we ignore this option for TCP sockets.
2517 			 * (RFC3542 leaves this case unspecified.)
2518 			 */
2519 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2520 		} else
2521 			opt->ip6po_flags |= IP6PO_DONTFRAG;
2522 		break;
2523 
2524 	default:
2525 		return (ENOPROTOOPT);
2526 	} /* end of switch */
2527 
2528 	return (0);
2529 }
2530 
2531 /*
2532  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2533  * packet to the input queue of a specified interface.
2534  */
2535 void
2536 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2537 {
2538 	struct mbuf *copym;
2539 	struct ip6_hdr *ip6;
2540 
2541 	/*
2542 	 * Duplicate the packet.
2543 	 */
2544 	copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
2545 	if (copym == NULL)
2546 		return;
2547 
2548 	/*
2549 	 * Make sure to deep-copy IPv6 header portion in case the data
2550 	 * is in an mbuf cluster, so that we can safely override the IPv6
2551 	 * header portion later.
2552 	 */
2553 	if ((copym->m_flags & M_EXT) != 0 ||
2554 	    copym->m_len < sizeof(struct ip6_hdr)) {
2555 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
2556 		if (copym == NULL)
2557 			return;
2558 	}
2559 
2560 #ifdef DIAGNOSTIC
2561 	if (copym->m_len < sizeof(*ip6)) {
2562 		m_freem(copym);
2563 		return;
2564 	}
2565 #endif
2566 
2567 	ip6 = mtod(copym, struct ip6_hdr *);
2568 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
2569 		ip6->ip6_src.s6_addr16[1] = 0;
2570 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
2571 		ip6->ip6_dst.s6_addr16[1] = 0;
2572 
2573 	if_input_local(ifp, copym, dst->sin6_family);
2574 }
2575 
2576 /*
2577  * Chop IPv6 header off from the payload.
2578  */
2579 int
2580 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
2581 {
2582 	struct mbuf *mh;
2583 	struct ip6_hdr *ip6;
2584 
2585 	ip6 = mtod(m, struct ip6_hdr *);
2586 	if (m->m_len > sizeof(*ip6)) {
2587 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
2588 		if (mh == NULL) {
2589 			m_freem(m);
2590 			return ENOBUFS;
2591 		}
2592 		M_MOVE_PKTHDR(mh, m);
2593 		MH_ALIGN(mh, sizeof(*ip6));
2594 		m->m_len -= sizeof(*ip6);
2595 		m->m_data += sizeof(*ip6);
2596 		mh->m_next = m;
2597 		m = mh;
2598 		m->m_len = sizeof(*ip6);
2599 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2600 	}
2601 	exthdrs->ip6e_ip6 = m;
2602 	return 0;
2603 }
2604 
2605 u_int32_t
2606 ip6_randomid(void)
2607 {
2608 	return idgen32(&ip6_id_ctx);
2609 }
2610 
2611 void
2612 ip6_randomid_init(void)
2613 {
2614 	idgen32_init(&ip6_id_ctx);
2615 }
2616 
2617 /*
2618  *	Compute significant parts of the IPv6 checksum pseudo-header
2619  *	for use in a delayed TCP/UDP checksum calculation.
2620  */
2621 static __inline u_int16_t __attribute__((__unused__))
2622 in6_cksum_phdr(const struct in6_addr *src, const struct in6_addr *dst,
2623     u_int32_t len, u_int32_t nxt)
2624 {
2625 	u_int32_t sum = 0;
2626 	const u_int16_t *w;
2627 
2628 	w = (const u_int16_t *) src;
2629 	sum += w[0];
2630 	if (!IN6_IS_SCOPE_EMBED(src))
2631 		sum += w[1];
2632 	sum += w[2]; sum += w[3]; sum += w[4]; sum += w[5];
2633 	sum += w[6]; sum += w[7];
2634 
2635 	w = (const u_int16_t *) dst;
2636 	sum += w[0];
2637 	if (!IN6_IS_SCOPE_EMBED(dst))
2638 		sum += w[1];
2639 	sum += w[2]; sum += w[3]; sum += w[4]; sum += w[5];
2640 	sum += w[6]; sum += w[7];
2641 
2642 	sum += (u_int16_t)(len >> 16) + (u_int16_t)(len /*& 0xffff*/);
2643 
2644 	sum += (u_int16_t)(nxt >> 16) + (u_int16_t)(nxt /*& 0xffff*/);
2645 
2646 	sum = (u_int16_t)(sum >> 16) + (u_int16_t)(sum /*& 0xffff*/);
2647 
2648 	if (sum > 0xffff)
2649 		sum -= 0xffff;
2650 
2651 	return (sum);
2652 }
2653 
2654 /*
2655  * Process a delayed payload checksum calculation.
2656  */
2657 void
2658 in6_delayed_cksum(struct mbuf *m, u_int8_t nxt)
2659 {
2660 	int nxtp, offset;
2661 	u_int16_t csum;
2662 
2663 	offset = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxtp);
2664 	if (offset <= 0 || nxtp != nxt)
2665 		/* If the desired next protocol isn't found, punt. */
2666 		return;
2667 	csum = (u_int16_t)(in6_cksum(m, 0, offset, m->m_pkthdr.len - offset));
2668 
2669 	switch (nxt) {
2670 	case IPPROTO_TCP:
2671 		offset += offsetof(struct tcphdr, th_sum);
2672 		break;
2673 
2674 	case IPPROTO_UDP:
2675 		offset += offsetof(struct udphdr, uh_sum);
2676 		if (csum == 0)
2677 			csum = 0xffff;
2678 		break;
2679 
2680 	case IPPROTO_ICMPV6:
2681 		offset += offsetof(struct icmp6_hdr, icmp6_cksum);
2682 		break;
2683 	}
2684 
2685 	if ((offset + sizeof(u_int16_t)) > m->m_len)
2686 		m_copyback(m, offset, sizeof(csum), &csum, M_NOWAIT);
2687 	else
2688 		*(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
2689 }
2690 
2691 void
2692 in6_proto_cksum_out(struct mbuf *m, struct ifnet *ifp)
2693 {
2694 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2695 
2696 	/* some hw and in6_delayed_cksum need the pseudo header cksum */
2697 	if (m->m_pkthdr.csum_flags &
2698 	    (M_TCP_CSUM_OUT|M_UDP_CSUM_OUT|M_ICMP_CSUM_OUT)) {
2699 		int nxt, offset;
2700 		u_int16_t csum;
2701 
2702 		offset = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
2703 		csum = in6_cksum_phdr(&ip6->ip6_src, &ip6->ip6_dst,
2704 		    htonl(m->m_pkthdr.len - offset), htonl(nxt));
2705 		if (nxt == IPPROTO_TCP)
2706 			offset += offsetof(struct tcphdr, th_sum);
2707 		else if (nxt == IPPROTO_UDP)
2708 			offset += offsetof(struct udphdr, uh_sum);
2709 		else if (nxt == IPPROTO_ICMPV6)
2710 			offset += offsetof(struct icmp6_hdr, icmp6_cksum);
2711 		if ((offset + sizeof(u_int16_t)) > m->m_len)
2712 			m_copyback(m, offset, sizeof(csum), &csum, M_NOWAIT);
2713 		else
2714 			*(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
2715 	}
2716 
2717 	if (m->m_pkthdr.csum_flags & M_TCP_CSUM_OUT) {
2718 		if (!ifp || !(ifp->if_capabilities & IFCAP_CSUM_TCPv6) ||
2719 		    ip6->ip6_nxt != IPPROTO_TCP ||
2720 		    ifp->if_bridgeport != NULL) {
2721 			tcpstat.tcps_outswcsum++;
2722 			in6_delayed_cksum(m, IPPROTO_TCP);
2723 			m->m_pkthdr.csum_flags &= ~M_TCP_CSUM_OUT; /* Clear */
2724 		}
2725 	} else if (m->m_pkthdr.csum_flags & M_UDP_CSUM_OUT) {
2726 		if (!ifp || !(ifp->if_capabilities & IFCAP_CSUM_UDPv6) ||
2727 		    ip6->ip6_nxt != IPPROTO_UDP ||
2728 		    ifp->if_bridgeport != NULL) {
2729 			udpstat_inc(udps_outswcsum);
2730 			in6_delayed_cksum(m, IPPROTO_UDP);
2731 			m->m_pkthdr.csum_flags &= ~M_UDP_CSUM_OUT; /* Clear */
2732 		}
2733 	} else if (m->m_pkthdr.csum_flags & M_ICMP_CSUM_OUT) {
2734 		in6_delayed_cksum(m, IPPROTO_ICMPV6);
2735 		m->m_pkthdr.csum_flags &= ~M_ICMP_CSUM_OUT; /* Clear */
2736 	}
2737 }
2738 
2739 #ifdef IPSEC
2740 struct tdb *
2741 ip6_output_ipsec_lookup(struct mbuf *m, int *error, struct inpcb *inp)
2742 {
2743 	struct tdb *tdb;
2744 	struct m_tag *mtag;
2745 	struct tdb_ident *tdbi;
2746 
2747 	/*
2748 	 * Check if there was an outgoing SA bound to the flow
2749 	 * from a transport protocol.
2750 	 */
2751 
2752 	/* Do we have any pending SAs to apply ? */
2753 	tdb = ipsp_spd_lookup(m, AF_INET6, sizeof(struct ip6_hdr),
2754 	    error, IPSP_DIRECTION_OUT, NULL, inp, 0);
2755 
2756 	if (tdb == NULL)
2757 		return NULL;
2758 	/* Loop detection */
2759 	for (mtag = m_tag_first(m); mtag != NULL; mtag = m_tag_next(m, mtag)) {
2760 		if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE)
2761 			continue;
2762 		tdbi = (struct tdb_ident *)(mtag + 1);
2763 		if (tdbi->spi == tdb->tdb_spi &&
2764 		    tdbi->proto == tdb->tdb_sproto &&
2765 		    tdbi->rdomain == tdb->tdb_rdomain &&
2766 		    !memcmp(&tdbi->dst, &tdb->tdb_dst,
2767 		    sizeof(union sockaddr_union))) {
2768 			/* no IPsec needed */
2769 			return NULL;
2770 		}
2771 	}
2772 	return tdb;
2773 }
2774 
2775 int
2776 ip6_output_ipsec_send(struct tdb *tdb, struct mbuf *m, int tunalready, int fwd)
2777 {
2778 #if NPF > 0
2779 	struct ifnet *encif;
2780 #endif
2781 
2782 #if NPF > 0
2783 	if ((encif = enc_getif(tdb->tdb_rdomain, tdb->tdb_tap)) == NULL ||
2784 	    pf_test(AF_INET6, fwd ? PF_FWD : PF_OUT, encif, &m) != PF_PASS) {
2785 		m_freem(m);
2786 		return EHOSTUNREACH;
2787 	}
2788 	if (m == NULL)
2789 		return 0;
2790 	/*
2791 	 * PF_TAG_REROUTE handling or not...
2792 	 * Packet is entering IPsec so the routing is
2793 	 * already overruled by the IPsec policy.
2794 	 * Until now the change was not reconsidered.
2795 	 * What's the behaviour?
2796 	 */
2797 	in6_proto_cksum_out(m, encif);
2798 #endif
2799 	m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
2800 
2801 	/* Callee frees mbuf */
2802 	return ipsp_process_packet(m, tdb, AF_INET6, tunalready);
2803 }
2804 #endif /* IPSEC */
2805