xref: /openbsd-src/sys/netinet6/ip6_output.c (revision fb8aa7497fded39583f40e800732f9c046411717)
1 /*	$OpenBSD: ip6_output.c,v 1.211 2016/07/01 18:18:57 jca 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_pcbopts(struct ip6_pktopts **, struct mbuf *, struct socket *);
121 int ip6_getpcbopt(struct ip6_pktopts *, int, struct mbuf **);
122 int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, int, int,
123 	int, int);
124 int ip6_setmoptions(int, struct ip6_moptions **, struct mbuf *);
125 int ip6_getmoptions(int, struct ip6_moptions *, struct mbuf **);
126 int ip6_copyexthdr(struct mbuf **, caddr_t, int);
127 int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
128 	struct ip6_frag **);
129 int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
130 int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
131 int ip6_getpmtu(struct route_in6 *, struct route_in6 *, struct ifnet *,
132     unsigned int, struct in6_addr *, u_long *, int *);
133 int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
134 static __inline u_int16_t __attribute__((__unused__))
135     in6_cksum_phdr(const struct in6_addr *, const struct in6_addr *,
136     u_int32_t, u_int32_t);
137 void in6_delayed_cksum(struct mbuf *, u_int8_t);
138 
139 /* Context for non-repeating IDs */
140 struct idgen32_ctx ip6_id_ctx;
141 
142 /*
143  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
144  * header (with pri, len, nxt, hlim, src, dst).
145  * This function may modify ver and hlim only.
146  * The mbuf chain containing the packet will be freed.
147  * The mbuf opt, if present, will not be freed.
148  *
149  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
150  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
151  * which is rt_rmx.rmx_mtu.
152  */
153 int
154 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt, struct route_in6 *ro,
155     int flags, struct ip6_moptions *im6o, struct inpcb *inp)
156 {
157 	struct ip6_hdr *ip6;
158 	struct ifnet *ifp = NULL;
159 	struct mbuf *m = m0;
160 	int hlen, tlen;
161 	struct route_in6 ip6route;
162 	struct rtentry *rt = NULL;
163 	struct sockaddr_in6 *dst, dstsock;
164 	int error = 0;
165 	u_long mtu;
166 	int alwaysfrag, dontfrag;
167 	u_int16_t src_scope, dst_scope;
168 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
169 	struct ip6_exthdrs exthdrs;
170 	struct in6_addr finaldst;
171 	struct route_in6 *ro_pmtu = NULL;
172 	int hdrsplit = 0;
173 	u_int8_t sproto = 0;
174 #ifdef IPSEC
175 	struct tdb *tdb = NULL;
176 #endif /* IPSEC */
177 
178 #ifdef IPSEC
179 	if (inp && (inp->inp_flags & INP_IPV6) == 0)
180 		panic("ip6_output: IPv4 pcb is passed");
181 #endif /* IPSEC */
182 
183 	ip6 = mtod(m, struct ip6_hdr *);
184 	finaldst = ip6->ip6_dst;
185 
186 #define MAKE_EXTHDR(hp, mp)						\
187     do {								\
188 	if (hp) {							\
189 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
190 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
191 		    ((eh)->ip6e_len + 1) << 3);				\
192 		if (error)						\
193 			goto freehdrs;					\
194 	}								\
195     } while (0)
196 
197 	bzero(&exthdrs, sizeof(exthdrs));
198 
199 	if (opt) {
200 		/* Hop-by-Hop options header */
201 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
202 		/* Destination options header(1st part) */
203 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
204 		/* Routing header */
205 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
206 		/* Destination options header(2nd part) */
207 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
208 	}
209 
210 #ifdef IPSEC
211 	if (ipsec_in_use || inp) {
212 		tdb = ip6_output_ipsec_lookup(m, &error, inp);
213 		if (error != 0) {
214 		        /*
215 			 * -EINVAL is used to indicate that the packet should
216 			 * be silently dropped, typically because we've asked
217 			 * key management for an SA.
218 			 */
219 		        if (error == -EINVAL) /* Should silently drop packet */
220 				error = 0;
221 
222 			goto freehdrs;
223 		}
224 	}
225 #endif /* IPSEC */
226 
227 	/*
228 	 * Calculate the total length of the extension header chain.
229 	 * Keep the length of the unfragmentable part for fragmentation.
230 	 */
231 	optlen = 0;
232 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
233 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
234 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
235 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
236 	/* NOTE: we don't add AH/ESP length here. do that later. */
237 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
238 
239 	/*
240 	 * If we need IPsec, or there is at least one extension header,
241 	 * separate IP6 header from the payload.
242 	 */
243 	if ((sproto || optlen) && !hdrsplit) {
244 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
245 			m = NULL;
246 			goto freehdrs;
247 		}
248 		m = exthdrs.ip6e_ip6;
249 		hdrsplit++;
250 	}
251 
252 	/* adjust pointer */
253 	ip6 = mtod(m, struct ip6_hdr *);
254 
255 	/* adjust mbuf packet header length */
256 	m->m_pkthdr.len += optlen;
257 	plen = m->m_pkthdr.len - sizeof(*ip6);
258 
259 	/* If this is a jumbo payload, insert a jumbo payload option. */
260 	if (plen > IPV6_MAXPACKET) {
261 		if (!hdrsplit) {
262 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
263 				m = NULL;
264 				goto freehdrs;
265 			}
266 			m = exthdrs.ip6e_ip6;
267 			hdrsplit++;
268 		}
269 		/* adjust pointer */
270 		ip6 = mtod(m, struct ip6_hdr *);
271 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
272 			goto freehdrs;
273 		ip6->ip6_plen = 0;
274 	} else
275 		ip6->ip6_plen = htons(plen);
276 
277 	/*
278 	 * Concatenate headers and fill in next header fields.
279 	 * Here we have, on "m"
280 	 *	IPv6 payload
281 	 * and we insert headers accordingly.  Finally, we should be getting:
282 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
283 	 *
284 	 * during the header composing process, "m" points to IPv6 header.
285 	 * "mprev" points to an extension header prior to esp.
286 	 */
287 	{
288 		u_char *nexthdrp = &ip6->ip6_nxt;
289 		struct mbuf *mprev = m;
290 
291 		/*
292 		 * we treat dest2 specially.  this makes IPsec processing
293 		 * much easier.  the goal here is to make mprev point the
294 		 * mbuf prior to dest2.
295 		 *
296 		 * result: IPv6 dest2 payload
297 		 * m and mprev will point to IPv6 header.
298 		 */
299 		if (exthdrs.ip6e_dest2) {
300 			if (!hdrsplit)
301 				panic("assumption failed: hdr not split");
302 			exthdrs.ip6e_dest2->m_next = m->m_next;
303 			m->m_next = exthdrs.ip6e_dest2;
304 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
305 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
306 		}
307 
308 #define MAKE_CHAIN(m, mp, p, i)\
309     do {\
310 	if (m) {\
311 		if (!hdrsplit) \
312 			panic("assumption failed: hdr not split"); \
313 		*mtod((m), u_char *) = *(p);\
314 		*(p) = (i);\
315 		p = mtod((m), u_char *);\
316 		(m)->m_next = (mp)->m_next;\
317 		(mp)->m_next = (m);\
318 		(mp) = (m);\
319 	}\
320     } while (0)
321 		/*
322 		 * result: IPv6 hbh dest1 rthdr dest2 payload
323 		 * m will point to IPv6 header.  mprev will point to the
324 		 * extension header prior to dest2 (rthdr in the above case).
325 		 */
326 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
327 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
328 		    IPPROTO_DSTOPTS);
329 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
330 		    IPPROTO_ROUTING);
331 	}
332 
333 	/*
334 	 * If there is a routing header, replace the destination address field
335 	 * with the first hop of the routing header.
336 	 */
337 	if (exthdrs.ip6e_rthdr) {
338 		struct ip6_rthdr *rh;
339 		struct ip6_rthdr0 *rh0;
340 		struct in6_addr *addr;
341 
342 		rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
343 		    struct ip6_rthdr *));
344 		switch (rh->ip6r_type) {
345 		case IPV6_RTHDR_TYPE_0:
346 			 rh0 = (struct ip6_rthdr0 *)rh;
347 			 addr = (struct in6_addr *)(rh0 + 1);
348 			 ip6->ip6_dst = addr[0];
349 			 bcopy(&addr[1], &addr[0],
350 			     sizeof(struct in6_addr) * (rh0->ip6r0_segleft - 1));
351 			 addr[rh0->ip6r0_segleft - 1] = finaldst;
352 			 break;
353 		default:	/* is it possible? */
354 			 error = EINVAL;
355 			 goto bad;
356 		}
357 	}
358 
359 	/* Source address validation */
360 	if (!(flags & IPV6_UNSPECSRC) &&
361 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
362 		/*
363 		 * XXX: we can probably assume validation in the caller, but
364 		 * we explicitly check the address here for safety.
365 		 */
366 		error = EOPNOTSUPP;
367 		ip6stat.ip6s_badscope++;
368 		goto bad;
369 	}
370 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
371 		error = EOPNOTSUPP;
372 		ip6stat.ip6s_badscope++;
373 		goto bad;
374 	}
375 
376 	ip6stat.ip6s_localout++;
377 
378 	/*
379 	 * Route packet.
380 	 */
381 #if NPF > 0
382 reroute:
383 #endif
384 
385 	/* initialize cached route */
386 	if (ro == NULL) {
387 		ro = &ip6route;
388 		bzero((caddr_t)ro, sizeof(*ro));
389 	}
390 	ro_pmtu = ro;
391 	if (opt && opt->ip6po_rthdr)
392 		ro = &opt->ip6po_route;
393 	dst = &ro->ro_dst;
394 
395 	/*
396 	 * if specified, try to fill in the traffic class field.
397 	 * do not override if a non-zero value is already set.
398 	 * we check the diffserv field and the ecn field separately.
399 	 */
400 	if (opt && opt->ip6po_tclass >= 0) {
401 		int mask = 0;
402 
403 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
404 			mask |= 0xfc;
405 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
406 			mask |= 0x03;
407 		if (mask != 0)
408 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
409 	}
410 
411 	/* fill in or override the hop limit field, if necessary. */
412 	if (opt && opt->ip6po_hlim != -1)
413 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
414 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
415 		if (im6o != NULL)
416 			ip6->ip6_hlim = im6o->im6o_hlim;
417 		else
418 			ip6->ip6_hlim = ip6_defmcasthlim;
419 	}
420 
421 #ifdef IPSEC
422 	if (tdb) {
423 		/*
424 		 * XXX what should we do if ip6_hlim == 0 and the
425 		 * packet gets tunneled?
426 		 */
427 		/*
428 		 * if we are source-routing, do not attempt to tunnel the
429 		 * packet just because ip6_dst is different from what tdb has.
430 		 * XXX
431 		 */
432 		error = ip6_output_ipsec_send(tdb, m,
433 		    exthdrs.ip6e_rthdr ? 1 : 0, 0);
434 		goto done;
435 	}
436 #endif /* IPSEC */
437 
438 	bzero(&dstsock, sizeof(dstsock));
439 	dstsock.sin6_family = AF_INET6;
440 	dstsock.sin6_addr = ip6->ip6_dst;
441 	dstsock.sin6_len = sizeof(dstsock);
442 	ro->ro_tableid = m->m_pkthdr.ph_rtableid;
443 
444 	if (IN6_IS_ADDR_MULTICAST(&dstsock.sin6_addr)) {
445 		struct in6_pktinfo *pi = NULL;
446 
447 		/*
448 		 * If the caller specify the outgoing interface
449 		 * explicitly, use it.
450 		 */
451 		if (opt != NULL && (pi = opt->ip6po_pktinfo) != NULL)
452 			ifp = if_get(pi->ipi6_ifindex);
453 
454 		if (ifp == NULL && im6o != NULL)
455 			ifp = if_get(im6o->im6o_ifidx);
456 	}
457 
458 	if (ifp == NULL) {
459 		rt = in6_selectroute(&dstsock, opt, ro, ro->ro_tableid);
460 		if (rt == NULL) {
461 			ip6stat.ip6s_noroute++;
462 			error = EHOSTUNREACH;
463 			goto bad;
464 		}
465 		if (ISSET(rt->rt_flags, RTF_LOCAL))
466 			ifp = if_get(lo0ifidx);
467 		else
468 			ifp = if_get(rt->rt_ifidx);
469 	} else {
470 		*dst = dstsock;
471 	}
472 
473 	if (rt && (rt->rt_flags & RTF_GATEWAY) &&
474 	    !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
475 		dst = satosin6(rt->rt_gateway);
476 
477 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
478 		/* Unicast */
479 
480 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
481 	} else {
482 		/* Multicast */
483 
484 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
485 
486 		/*
487 		 * Confirm that the outgoing interface supports multicast.
488 		 */
489 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
490 			ip6stat.ip6s_noroute++;
491 			error = ENETUNREACH;
492 			goto bad;
493 		}
494 
495 		if ((im6o == NULL || im6o->im6o_loop) &&
496 		    in6_hasmulti(&ip6->ip6_dst, ifp)) {
497 			/*
498 			 * If we belong to the destination multicast group
499 			 * on the outgoing interface, and the caller did not
500 			 * forbid loopback, loop back a copy.
501 			 * Can't defer TCP/UDP checksumming, do the
502 			 * computation now.
503 			 */
504 			in6_proto_cksum_out(m, NULL);
505 			ip6_mloopback(ifp, m, dst);
506 		}
507 #ifdef MROUTING
508 		else {
509 			/*
510 			 * If we are acting as a multicast router, perform
511 			 * multicast forwarding as if the packet had just
512 			 * arrived on the interface to which we are about
513 			 * to send.  The multicast forwarding function
514 			 * recursively calls this function, using the
515 			 * IPV6_FORWARDING flag to prevent infinite recursion.
516 			 *
517 			 * Multicasts that are looped back by ip6_mloopback(),
518 			 * above, will be forwarded by the ip6_input() routine,
519 			 * if necessary.
520 			 */
521 			if (ip6_mforwarding && ip6_mrouter &&
522 			    (flags & IPV6_FORWARDING) == 0) {
523 				if (ip6_mforward(ip6, ifp, m) != 0) {
524 					m_freem(m);
525 					goto done;
526 				}
527 			}
528 		}
529 #endif
530 		/*
531 		 * Multicasts with a hoplimit of zero may be looped back,
532 		 * above, but must not be transmitted on a network.
533 		 * Also, multicasts addressed to the loopback interface
534 		 * are not sent -- the above call to ip6_mloopback() will
535 		 * loop back a copy if this host actually belongs to the
536 		 * destination group on the loopback interface.
537 		 */
538 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
539 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
540 			m_freem(m);
541 			goto done;
542 		}
543 	}
544 
545 	/*
546 	 * If this packet is going trough a loopback interface we wont
547 	 * be able to restore its scope ID using the interface index.
548 	 */
549 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src)) {
550 		if (ifp->if_flags & IFF_LOOPBACK)
551 			src_scope = ip6->ip6_src.s6_addr16[1];
552 		ip6->ip6_src.s6_addr16[1] = 0;
553 	}
554 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) {
555 		if (ifp->if_flags & IFF_LOOPBACK)
556 			dst_scope = ip6->ip6_dst.s6_addr16[1];
557 		ip6->ip6_dst.s6_addr16[1] = 0;
558 	}
559 
560 	/* Determine path MTU. */
561 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, ro->ro_tableid, &finaldst,
562 	    &mtu, &alwaysfrag)) != 0)
563 		goto bad;
564 
565 	/*
566 	 * The caller of this function may specify to use the minimum MTU
567 	 * in some cases.
568 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
569 	 * setting.  The logic is a bit complicated; by default, unicast
570 	 * packets will follow path MTU while multicast packets will be sent at
571 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
572 	 * including unicast ones will be sent at the minimum MTU.  Multicast
573 	 * packets will always be sent at the minimum MTU unless
574 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
575 	 * See RFC 3542 for more details.
576 	 */
577 	if (mtu > IPV6_MMTU) {
578 		if ((flags & IPV6_MINMTU))
579 			mtu = IPV6_MMTU;
580 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
581 			mtu = IPV6_MMTU;
582 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
583 			 (opt == NULL ||
584 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
585 			mtu = IPV6_MMTU;
586 		}
587 	}
588 
589 	/*
590 	 * If the outgoing packet contains a hop-by-hop options header,
591 	 * it must be examined and processed even by the source node.
592 	 * (RFC 2460, section 4.)
593 	 */
594 	if (exthdrs.ip6e_hbh) {
595 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
596 		u_int32_t dummy1; /* XXX unused */
597 		u_int32_t dummy2; /* XXX unused */
598 
599 		/*
600 		 *  XXX: if we have to send an ICMPv6 error to the sender,
601 		 *       we need the M_LOOP flag since icmp6_error() expects
602 		 *       the IPv6 and the hop-by-hop options header are
603 		 *       continuous unless the flag is set.
604 		 */
605 		m->m_flags |= M_LOOP;
606 		m->m_pkthdr.ph_ifidx = ifp->if_index;
607 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
608 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
609 		    &dummy1, &dummy2) < 0) {
610 			/* m was already freed at this point */
611 			error = EINVAL;/* better error? */
612 			goto done;
613 		}
614 		m->m_flags &= ~M_LOOP; /* XXX */
615 		m->m_pkthdr.ph_ifidx = 0;
616 	}
617 
618 #if NPF > 0
619 	if (pf_test(AF_INET6, PF_OUT, ifp, &m) != PF_PASS) {
620 		error = EHOSTUNREACH;
621 		m_freem(m);
622 		goto done;
623 	}
624 	if (m == NULL)
625 		goto done;
626 	ip6 = mtod(m, struct ip6_hdr *);
627 	if ((m->m_pkthdr.pf.flags & (PF_TAG_REROUTE | PF_TAG_GENERATED)) ==
628 	    (PF_TAG_REROUTE | PF_TAG_GENERATED)) {
629 		/* already rerun the route lookup, go on */
630 		m->m_pkthdr.pf.flags &= ~(PF_TAG_GENERATED | PF_TAG_REROUTE);
631 	} else if (m->m_pkthdr.pf.flags & PF_TAG_REROUTE) {
632 		/* tag as generated to skip over pf_test on rerun */
633 		m->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
634 		finaldst = ip6->ip6_dst;
635 		ro = NULL;
636 		if_put(ifp); /* drop reference since destination changed */
637 		ifp = NULL;
638 		goto reroute;
639 	}
640 #endif
641 
642 	/*
643 	 * If the packet is not going on the wire it can be destinated
644 	 * to any local address.  In this case do not clear its scopes
645 	 * to let ip6_input() find a matching local route.
646 	 */
647 	if (ifp->if_flags & IFF_LOOPBACK) {
648 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
649 			ip6->ip6_src.s6_addr16[1] = src_scope;
650 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
651 			ip6->ip6_dst.s6_addr16[1] = dst_scope;
652 	}
653 
654 	in6_proto_cksum_out(m, ifp);
655 
656 	/*
657 	 * Send the packet to the outgoing interface.
658 	 * If necessary, do IPv6 fragmentation before sending.
659 	 *
660 	 * the logic here is rather complex:
661 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
662 	 * 1-a: send as is if tlen <= path mtu
663 	 * 1-b: fragment if tlen > path mtu
664 	 *
665 	 * 2: if user asks us not to fragment (dontfrag == 1)
666 	 * 2-a: send as is if tlen <= interface mtu
667 	 * 2-b: error if tlen > interface mtu
668 	 *
669 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
670 	 *      always fragment
671 	 *
672 	 * 4: if dontfrag == 1 && alwaysfrag == 1
673 	 *      error, as we cannot handle this conflicting request
674 	 */
675 	tlen = m->m_pkthdr.len;
676 
677 	if (opt && (opt->ip6po_flags & IP6PO_DONTFRAG))
678 		dontfrag = 1;
679 	else
680 		dontfrag = 0;
681 	if (dontfrag && alwaysfrag) {	/* case 4 */
682 		/* conflicting request - can't transmit */
683 		error = EMSGSIZE;
684 		goto bad;
685 	}
686 	if (dontfrag && tlen > ifp->if_mtu) {	/* case 2-b */
687 		error = EMSGSIZE;
688 		goto bad;
689 	}
690 
691 	/*
692 	 * transmit packet without fragmentation
693 	 */
694 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
695 		error = ifp->if_output(ifp, m, sin6tosa(dst), ro->ro_rt);
696 		goto done;
697 	}
698 
699 	/*
700 	 * try to fragment the packet.  case 1-b and 3
701 	 */
702 	if (mtu < IPV6_MMTU) {
703 		/* path MTU cannot be less than IPV6_MMTU */
704 		error = EMSGSIZE;
705 		goto bad;
706 	} else if (ip6->ip6_plen == 0) {
707 		/* jumbo payload cannot be fragmented */
708 		error = EMSGSIZE;
709 		goto bad;
710 	} else {
711 		u_char nextproto;
712 #if 0
713 		struct ip6ctlparam ip6cp;
714 		u_int32_t mtu32;
715 #endif
716 
717 		/*
718 		 * Too large for the destination or interface;
719 		 * fragment if possible.
720 		 * Must be able to put at least 8 bytes per fragment.
721 		 */
722 		hlen = unfragpartlen;
723 		if (mtu > IPV6_MAXPACKET)
724 			mtu = IPV6_MAXPACKET;
725 
726 #if 0
727 		/* Notify a proper path MTU to applications. */
728 		mtu32 = (u_int32_t)mtu;
729 		bzero(&ip6cp, sizeof(ip6cp));
730 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
731 		pfctlinput2(PRC_MSGSIZE, sin6tosa(&ro_pmtu->ro_dst),
732 		    (void *)&ip6cp);
733 #endif
734 
735 		/*
736 		 * Change the next header field of the last header in the
737 		 * unfragmentable part.
738 		 */
739 		if (exthdrs.ip6e_rthdr) {
740 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
741 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
742 		} else if (exthdrs.ip6e_dest1) {
743 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
744 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
745 		} else if (exthdrs.ip6e_hbh) {
746 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
747 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
748 		} else {
749 			nextproto = ip6->ip6_nxt;
750 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
751 		}
752 
753 		m0 = m;
754 		error = ip6_fragment(m0, hlen, nextproto, mtu);
755 		if (error)
756 			ip6stat.ip6s_odropped++;
757 	}
758 
759 	/*
760 	 * Remove leading garbages.
761 	 */
762 	m = m0->m_nextpkt;
763 	m0->m_nextpkt = 0;
764 	m_freem(m0);
765 	for (m0 = m; m; m = m0) {
766 		m0 = m->m_nextpkt;
767 		m->m_nextpkt = 0;
768 		if (error == 0) {
769 			ip6stat.ip6s_ofragments++;
770 			error = ifp->if_output(ifp, m, sin6tosa(dst),
771 			    ro->ro_rt);
772 		} else
773 			m_freem(m);
774 	}
775 
776 	if (error == 0)
777 		ip6stat.ip6s_fragmented++;
778 
779 done:
780 	if_put(ifp);
781 	if (ro == &ip6route && ro->ro_rt) {
782 		rtfree(ro->ro_rt);
783 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
784 		rtfree(ro_pmtu->ro_rt);
785 	}
786 
787 	return (error);
788 
789 freehdrs:
790 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
791 	m_freem(exthdrs.ip6e_dest1);
792 	m_freem(exthdrs.ip6e_rthdr);
793 	m_freem(exthdrs.ip6e_dest2);
794 	/* FALLTHROUGH */
795 bad:
796 	m_freem(m);
797 	goto done;
798 }
799 
800 int
801 ip6_fragment(struct mbuf *m0, int hlen, u_char nextproto, u_long mtu)
802 {
803 	struct mbuf	*m, **mnext, *m_frgpart;
804 	struct ip6_hdr	*mhip6;
805 	struct ip6_frag	*ip6f;
806 	u_int32_t	 id;
807 	int		 tlen, len, off;
808 	int		 error;
809 
810 	id = htonl(ip6_randomid());
811 
812 	mnext = &m0->m_nextpkt;
813 	*mnext = NULL;
814 
815 	tlen = m0->m_pkthdr.len;
816 	len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
817 	if (len < 8)
818 		return (EMSGSIZE);
819 
820 	/*
821 	 * Loop through length of segment after first fragment,
822 	 * make new header and copy data of each part and link onto
823 	 * chain.
824 	 */
825 	for (off = hlen; off < tlen; off += len) {
826 		struct mbuf *mlast;
827 
828 		if ((m = m_gethdr(M_DONTWAIT, MT_HEADER)) == NULL)
829 			return (ENOBUFS);
830 		*mnext = m;
831 		mnext = &m->m_nextpkt;
832 		if ((error = m_dup_pkthdr(m, m0, M_DONTWAIT)) != 0)
833 			return (error);
834 		m->m_data += max_linkhdr;
835 		mhip6 = mtod(m, struct ip6_hdr *);
836 		*mhip6 = *mtod(m0, struct ip6_hdr *);
837 		m->m_len = sizeof(*mhip6);
838 		if ((error = ip6_insertfraghdr(m0, m, hlen, &ip6f)) != 0)
839 			return (error);
840 		ip6f->ip6f_offlg = htons((u_int16_t)((off - hlen) & ~7));
841 		if (off + len >= tlen)
842 			len = tlen - off;
843 		else
844 			ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
845 		mhip6->ip6_plen = htons((u_int16_t)(len + hlen +
846 		    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
847 		if ((m_frgpart = m_copym(m0, off, len, M_DONTWAIT)) == NULL)
848 			return (ENOBUFS);
849 		for (mlast = m; mlast->m_next; mlast = mlast->m_next)
850 			;
851 		mlast->m_next = m_frgpart;
852 		m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
853 		ip6f->ip6f_reserved = 0;
854 		ip6f->ip6f_ident = id;
855 		ip6f->ip6f_nxt = nextproto;
856 	}
857 
858 	return (0);
859 }
860 
861 int
862 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
863 {
864 	struct mbuf *m;
865 
866 	if (hlen > MCLBYTES)
867 		return (ENOBUFS); /* XXX */
868 
869 	MGET(m, M_DONTWAIT, MT_DATA);
870 	if (!m)
871 		return (ENOBUFS);
872 
873 	if (hlen > MLEN) {
874 		MCLGET(m, M_DONTWAIT);
875 		if ((m->m_flags & M_EXT) == 0) {
876 			m_free(m);
877 			return (ENOBUFS);
878 		}
879 	}
880 	m->m_len = hlen;
881 	if (hdr)
882 		bcopy(hdr, mtod(m, caddr_t), hlen);
883 
884 	*mp = m;
885 	return (0);
886 }
887 
888 /*
889  * Insert jumbo payload option.
890  */
891 int
892 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
893 {
894 	struct mbuf *mopt;
895 	u_int8_t *optbuf;
896 	u_int32_t v;
897 
898 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
899 
900 	/*
901 	 * If there is no hop-by-hop options header, allocate new one.
902 	 * If there is one but it doesn't have enough space to store the
903 	 * jumbo payload option, allocate a cluster to store the whole options.
904 	 * Otherwise, use it to store the options.
905 	 */
906 	if (exthdrs->ip6e_hbh == 0) {
907 		MGET(mopt, M_DONTWAIT, MT_DATA);
908 		if (mopt == NULL)
909 			return (ENOBUFS);
910 		mopt->m_len = JUMBOOPTLEN;
911 		optbuf = mtod(mopt, u_int8_t *);
912 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
913 		exthdrs->ip6e_hbh = mopt;
914 	} else {
915 		struct ip6_hbh *hbh;
916 
917 		mopt = exthdrs->ip6e_hbh;
918 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
919 			/*
920 			 * XXX assumption:
921 			 * - exthdrs->ip6e_hbh is not referenced from places
922 			 *   other than exthdrs.
923 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
924 			 */
925 			int oldoptlen = mopt->m_len;
926 			struct mbuf *n;
927 
928 			/*
929 			 * XXX: give up if the whole (new) hbh header does
930 			 * not fit even in an mbuf cluster.
931 			 */
932 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
933 				return (ENOBUFS);
934 
935 			/*
936 			 * As a consequence, we must always prepare a cluster
937 			 * at this point.
938 			 */
939 			MGET(n, M_DONTWAIT, MT_DATA);
940 			if (n) {
941 				MCLGET(n, M_DONTWAIT);
942 				if ((n->m_flags & M_EXT) == 0) {
943 					m_freem(n);
944 					n = NULL;
945 				}
946 			}
947 			if (!n)
948 				return (ENOBUFS);
949 			n->m_len = oldoptlen + JUMBOOPTLEN;
950 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
951 			      oldoptlen);
952 			optbuf = mtod(n, u_int8_t *) + oldoptlen;
953 			m_freem(mopt);
954 			mopt = exthdrs->ip6e_hbh = n;
955 		} else {
956 			optbuf = mtod(mopt, u_int8_t *) + mopt->m_len;
957 			mopt->m_len += JUMBOOPTLEN;
958 		}
959 		optbuf[0] = IP6OPT_PADN;
960 		optbuf[1] = 0;
961 
962 		/*
963 		 * Adjust the header length according to the pad and
964 		 * the jumbo payload option.
965 		 */
966 		hbh = mtod(mopt, struct ip6_hbh *);
967 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
968 	}
969 
970 	/* fill in the option. */
971 	optbuf[2] = IP6OPT_JUMBO;
972 	optbuf[3] = 4;
973 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
974 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
975 
976 	/* finally, adjust the packet header length */
977 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
978 
979 	return (0);
980 #undef JUMBOOPTLEN
981 }
982 
983 /*
984  * Insert fragment header and copy unfragmentable header portions.
985  */
986 int
987 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
988     struct ip6_frag **frghdrp)
989 {
990 	struct mbuf *n, *mlast;
991 
992 	if (hlen > sizeof(struct ip6_hdr)) {
993 		n = m_copym(m0, sizeof(struct ip6_hdr),
994 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
995 		if (n == NULL)
996 			return (ENOBUFS);
997 		m->m_next = n;
998 	} else
999 		n = m;
1000 
1001 	/* Search for the last mbuf of unfragmentable part. */
1002 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1003 		;
1004 
1005 	if ((mlast->m_flags & M_EXT) == 0 &&
1006 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1007 		/* use the trailing space of the last mbuf for the fragment hdr */
1008 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1009 		    mlast->m_len);
1010 		mlast->m_len += sizeof(struct ip6_frag);
1011 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1012 	} else {
1013 		/* allocate a new mbuf for the fragment header */
1014 		struct mbuf *mfrg;
1015 
1016 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1017 		if (mfrg == NULL)
1018 			return (ENOBUFS);
1019 		mfrg->m_len = sizeof(struct ip6_frag);
1020 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1021 		mlast->m_next = mfrg;
1022 	}
1023 
1024 	return (0);
1025 }
1026 
1027 int
1028 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro, struct ifnet *ifp0,
1029     unsigned int rtableid, struct in6_addr *dst, u_long *mtup, int *alwaysfragp)
1030 {
1031 	u_int32_t mtu = 0;
1032 	int alwaysfrag = 0;
1033 	int error = 0;
1034 
1035 	if (ro_pmtu != ro) {
1036 		/* The first hop and the final destination may differ. */
1037 		struct sockaddr_in6 *sa6_dst = &ro_pmtu->ro_dst;
1038 
1039 		if (!rtisvalid(ro_pmtu->ro_rt) ||
1040 		    (ro_pmtu->ro_tableid != rtableid) ||
1041 		     !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst)) {
1042 			rtfree(ro_pmtu->ro_rt);
1043 			ro_pmtu->ro_rt = NULL;
1044 		}
1045 		if (ro_pmtu->ro_rt == NULL) {
1046 			bzero(ro_pmtu, sizeof(*ro_pmtu));
1047 			ro_pmtu->ro_tableid = rtableid;
1048 			sa6_dst->sin6_family = AF_INET6;
1049 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1050 			sa6_dst->sin6_addr = *dst;
1051 
1052 			ro_pmtu->ro_rt = rtalloc(sin6tosa(&ro_pmtu->ro_dst),
1053 			    RT_RESOLVE, ro_pmtu->ro_tableid);
1054 		}
1055 	}
1056 	if (ro_pmtu->ro_rt) {
1057 		struct ifnet *ifp;
1058 
1059 		if (ifp0 == NULL) {
1060 			ifp = if_get(ro_pmtu->ro_rt->rt_ifidx);
1061 			if (ifp == NULL)
1062 				return (EHOSTUNREACH);
1063 		} else
1064 			ifp = ifp0;
1065 
1066 		mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1067 		if (mtu == 0)
1068 			mtu = ifp->if_mtu;
1069 		else if (mtu < IPV6_MMTU) {
1070 			/*
1071 			 * RFC2460 section 5, last paragraph:
1072 			 * if we record ICMPv6 too big message with
1073 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1074 			 * or smaller, with fragment header attached.
1075 			 * (fragment header is needed regardless from the
1076 			 * packet size, for translators to identify packets)
1077 			 */
1078 			alwaysfrag = 1;
1079 			mtu = IPV6_MMTU;
1080 		} else if (mtu > ifp->if_mtu) {
1081 			/*
1082 			 * The MTU on the route is larger than the MTU on
1083 			 * the interface!  This shouldn't happen, unless the
1084 			 * MTU of the interface has been changed after the
1085 			 * interface was brought up.  Change the MTU in the
1086 			 * route to match the interface MTU (as long as the
1087 			 * field isn't locked).
1088 			 */
1089 			mtu = ifp->if_mtu;
1090 			if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU))
1091 				ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1092 		}
1093 
1094 		if (ifp0 == NULL)
1095 			if_put(ifp);
1096 	} else if (ifp0) {
1097 		mtu = ifp0->if_mtu;
1098 	} else
1099 		error = EHOSTUNREACH; /* XXX */
1100 
1101 	*mtup = mtu;
1102 	if (alwaysfragp)
1103 		*alwaysfragp = alwaysfrag;
1104 	return (error);
1105 }
1106 
1107 /*
1108  * IP6 socket option processing.
1109  */
1110 int
1111 ip6_ctloutput(int op, struct socket *so, int level, int optname,
1112     struct mbuf **mp)
1113 {
1114 	int privileged, optdatalen, uproto;
1115 	void *optdata;
1116 	struct inpcb *inp = sotoinpcb(so);
1117 	struct mbuf *m = *mp;
1118 	int error, optval;
1119 	struct proc *p = curproc; /* For IPSec and rdomain */
1120 	u_int rtid = 0;
1121 
1122 	error = optval = 0;
1123 
1124 	privileged = (inp->inp_socket->so_state & SS_PRIV);
1125 	uproto = (int)so->so_proto->pr_protocol;
1126 
1127 	if (level == IPPROTO_IPV6) {
1128 		switch (op) {
1129 		case PRCO_SETOPT:
1130 			switch (optname) {
1131 			/*
1132 			 * Use of some Hop-by-Hop options or some
1133 			 * Destination options, might require special
1134 			 * privilege.  That is, normal applications
1135 			 * (without special privilege) might be forbidden
1136 			 * from setting certain options in outgoing packets,
1137 			 * and might never see certain options in received
1138 			 * packets. [RFC 2292 Section 6]
1139 			 * KAME specific note:
1140 			 *  KAME prevents non-privileged users from sending or
1141 			 *  receiving ANY hbh/dst options in order to avoid
1142 			 *  overhead of parsing options in the kernel.
1143 			 */
1144 			case IPV6_RECVHOPOPTS:
1145 			case IPV6_RECVDSTOPTS:
1146 				if (!privileged) {
1147 					error = EPERM;
1148 					break;
1149 				}
1150 				/* FALLTHROUGH */
1151 			case IPV6_UNICAST_HOPS:
1152 			case IPV6_MINHOPCOUNT:
1153 			case IPV6_HOPLIMIT:
1154 
1155 			case IPV6_RECVPKTINFO:
1156 			case IPV6_RECVHOPLIMIT:
1157 			case IPV6_RECVRTHDR:
1158 			case IPV6_RECVPATHMTU:
1159 			case IPV6_RECVTCLASS:
1160 			case IPV6_V6ONLY:
1161 			case IPV6_AUTOFLOWLABEL:
1162 			case IPV6_RECVDSTPORT:
1163 				if (m == NULL || m->m_len != sizeof(int)) {
1164 					error = EINVAL;
1165 					break;
1166 				}
1167 				optval = *mtod(m, int *);
1168 				switch (optname) {
1169 
1170 				case IPV6_UNICAST_HOPS:
1171 					if (optval < -1 || optval >= 256)
1172 						error = EINVAL;
1173 					else {
1174 						/* -1 = kernel default */
1175 						inp->inp_hops = optval;
1176 					}
1177 					break;
1178 
1179 				case IPV6_MINHOPCOUNT:
1180 					if (optval < 0 || optval > 255)
1181 						error = EINVAL;
1182 					else
1183 						inp->inp_ip6_minhlim = optval;
1184 					break;
1185 
1186 #define OPTSET(bit) \
1187 do { \
1188 	if (optval) \
1189 		inp->inp_flags |= (bit); \
1190 	else \
1191 		inp->inp_flags &= ~(bit); \
1192 } while (/*CONSTCOND*/ 0)
1193 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1194 
1195 				case IPV6_RECVPKTINFO:
1196 					OPTSET(IN6P_PKTINFO);
1197 					break;
1198 
1199 				case IPV6_HOPLIMIT:
1200 				{
1201 					struct ip6_pktopts **optp;
1202 
1203 					optp = &inp->inp_outputopts6;
1204 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1205 							   (u_char *)&optval,
1206 							   sizeof(optval),
1207 							   optp,
1208 							   privileged, uproto);
1209 					break;
1210 				}
1211 
1212 				case IPV6_RECVHOPLIMIT:
1213 					OPTSET(IN6P_HOPLIMIT);
1214 					break;
1215 
1216 				case IPV6_RECVHOPOPTS:
1217 					OPTSET(IN6P_HOPOPTS);
1218 					break;
1219 
1220 				case IPV6_RECVDSTOPTS:
1221 					OPTSET(IN6P_DSTOPTS);
1222 					break;
1223 
1224 				case IPV6_RECVRTHDR:
1225 					OPTSET(IN6P_RTHDR);
1226 					break;
1227 
1228 				case IPV6_RECVPATHMTU:
1229 					/*
1230 					 * We ignore this option for TCP
1231 					 * sockets.
1232 					 * (RFC3542 leaves this case
1233 					 * unspecified.)
1234 					 */
1235 					if (uproto != IPPROTO_TCP)
1236 						OPTSET(IN6P_MTU);
1237 					break;
1238 
1239 				case IPV6_V6ONLY:
1240 					/*
1241 					 * make setsockopt(IPV6_V6ONLY)
1242 					 * available only prior to bind(2).
1243 					 * see ipng mailing list, Jun 22 2001.
1244 					 */
1245 					if (inp->inp_lport ||
1246 					    !IN6_IS_ADDR_UNSPECIFIED(&inp->inp_laddr6)) {
1247 						error = EINVAL;
1248 						break;
1249 					}
1250 					/* No support for IPv4-mapped addresses. */
1251 					if (!optval)
1252 						error = EINVAL;
1253 					else
1254 						error = 0;
1255 					break;
1256 				case IPV6_RECVTCLASS:
1257 					OPTSET(IN6P_TCLASS);
1258 					break;
1259 				case IPV6_AUTOFLOWLABEL:
1260 					OPTSET(IN6P_AUTOFLOWLABEL);
1261 					break;
1262 
1263 				case IPV6_RECVDSTPORT:
1264 					OPTSET(IN6P_RECVDSTPORT);
1265 					break;
1266 				}
1267 				break;
1268 
1269 			case IPV6_TCLASS:
1270 			case IPV6_DONTFRAG:
1271 			case IPV6_USE_MIN_MTU:
1272 				if (m == NULL || m->m_len != sizeof(optval)) {
1273 					error = EINVAL;
1274 					break;
1275 				}
1276 				optval = *mtod(m, int *);
1277 				{
1278 					struct ip6_pktopts **optp;
1279 					optp = &inp->inp_outputopts6;
1280 					error = ip6_pcbopt(optname,
1281 							   (u_char *)&optval,
1282 							   sizeof(optval),
1283 							   optp,
1284 							   privileged, uproto);
1285 					break;
1286 				}
1287 
1288 			case IPV6_PKTINFO:
1289 			case IPV6_HOPOPTS:
1290 			case IPV6_RTHDR:
1291 			case IPV6_DSTOPTS:
1292 			case IPV6_RTHDRDSTOPTS:
1293 			{
1294 				/* new advanced API (RFC3542) */
1295 				u_char *optbuf;
1296 				int optbuflen;
1297 				struct ip6_pktopts **optp;
1298 
1299 				if (m && m->m_next) {
1300 					error = EINVAL;	/* XXX */
1301 					break;
1302 				}
1303 				if (m) {
1304 					optbuf = mtod(m, u_char *);
1305 					optbuflen = m->m_len;
1306 				} else {
1307 					optbuf = NULL;
1308 					optbuflen = 0;
1309 				}
1310 				optp = &inp->inp_outputopts6;
1311 				error = ip6_pcbopt(optname,
1312 						   optbuf, optbuflen,
1313 						   optp, privileged, uproto);
1314 				break;
1315 			}
1316 #undef OPTSET
1317 
1318 			case IPV6_MULTICAST_IF:
1319 			case IPV6_MULTICAST_HOPS:
1320 			case IPV6_MULTICAST_LOOP:
1321 			case IPV6_JOIN_GROUP:
1322 			case IPV6_LEAVE_GROUP:
1323 				error =	ip6_setmoptions(optname,
1324 							&inp->inp_moptions6,
1325 							m);
1326 				break;
1327 
1328 			case IPV6_PORTRANGE:
1329 				if (m == NULL || m->m_len != sizeof(int)) {
1330 					error = EINVAL;
1331 					break;
1332 				}
1333 				optval = *mtod(m, int *);
1334 
1335 				switch (optval) {
1336 				case IPV6_PORTRANGE_DEFAULT:
1337 					inp->inp_flags &= ~(IN6P_LOWPORT);
1338 					inp->inp_flags &= ~(IN6P_HIGHPORT);
1339 					break;
1340 
1341 				case IPV6_PORTRANGE_HIGH:
1342 					inp->inp_flags &= ~(IN6P_LOWPORT);
1343 					inp->inp_flags |= IN6P_HIGHPORT;
1344 					break;
1345 
1346 				case IPV6_PORTRANGE_LOW:
1347 					inp->inp_flags &= ~(IN6P_HIGHPORT);
1348 					inp->inp_flags |= IN6P_LOWPORT;
1349 					break;
1350 
1351 				default:
1352 					error = EINVAL;
1353 					break;
1354 				}
1355 				break;
1356 
1357 			case IPSEC6_OUTSA:
1358 				error = EINVAL;
1359 				break;
1360 
1361 			case IPV6_AUTH_LEVEL:
1362 			case IPV6_ESP_TRANS_LEVEL:
1363 			case IPV6_ESP_NETWORK_LEVEL:
1364 			case IPV6_IPCOMP_LEVEL:
1365 #ifndef IPSEC
1366 				error = EINVAL;
1367 #else
1368 				if (m == NULL || m->m_len != sizeof(int)) {
1369 					error = EINVAL;
1370 					break;
1371 				}
1372 				optval = *mtod(m, int *);
1373 
1374 				if (optval < IPSEC_LEVEL_BYPASS ||
1375 				    optval > IPSEC_LEVEL_UNIQUE) {
1376 					error = EINVAL;
1377 					break;
1378 				}
1379 
1380 				switch (optname) {
1381 				case IPV6_AUTH_LEVEL:
1382 				        if (optval < IPSEC_AUTH_LEVEL_DEFAULT &&
1383 					    suser(p, 0)) {
1384 						error = EACCES;
1385 						break;
1386 					}
1387 					inp->inp_seclevel[SL_AUTH] = optval;
1388 					break;
1389 
1390 				case IPV6_ESP_TRANS_LEVEL:
1391 				        if (optval < IPSEC_ESP_TRANS_LEVEL_DEFAULT &&
1392 					    suser(p, 0)) {
1393 						error = EACCES;
1394 						break;
1395 					}
1396 					inp->inp_seclevel[SL_ESP_TRANS] = optval;
1397 					break;
1398 
1399 				case IPV6_ESP_NETWORK_LEVEL:
1400 				        if (optval < IPSEC_ESP_NETWORK_LEVEL_DEFAULT &&
1401 					    suser(p, 0)) {
1402 						error = EACCES;
1403 						break;
1404 					}
1405 					inp->inp_seclevel[SL_ESP_NETWORK] = optval;
1406 					break;
1407 
1408 				case IPV6_IPCOMP_LEVEL:
1409 				        if (optval < IPSEC_IPCOMP_LEVEL_DEFAULT &&
1410 					    suser(p, 0)) {
1411 						error = EACCES;
1412 						break;
1413 					}
1414 					inp->inp_seclevel[SL_IPCOMP] = optval;
1415 					break;
1416 				}
1417 #endif
1418 				break;
1419 			case SO_RTABLE:
1420 				if (m == NULL || m->m_len < sizeof(u_int)) {
1421 					error = EINVAL;
1422 					break;
1423 				}
1424 				rtid = *mtod(m, u_int *);
1425 				if (inp->inp_rtableid == rtid)
1426 					break;
1427 				/* needs privileges to switch when already set */
1428 				if (p->p_p->ps_rtableid != rtid &&
1429 				    p->p_p->ps_rtableid != 0 &&
1430 				    (error = suser(p, 0)) != 0)
1431 					break;
1432 				/* table must exist */
1433 				if (!rtable_exists(rtid)) {
1434 					error = EINVAL;
1435 					break;
1436 				}
1437 				if (inp->inp_lport) {
1438 					error = EBUSY;
1439 					break;
1440 				}
1441 				inp->inp_rtableid = rtid;
1442 				in_pcbrehash(inp);
1443 				break;
1444 			case IPV6_PIPEX:
1445 				if (m != NULL && m->m_len == sizeof(int))
1446 					inp->inp_pipex = *mtod(m, int *);
1447 				else
1448 					error = EINVAL;
1449 				break;
1450 
1451 			default:
1452 				error = ENOPROTOOPT;
1453 				break;
1454 			}
1455 			if (m)
1456 				(void)m_free(m);
1457 			break;
1458 
1459 		case PRCO_GETOPT:
1460 			switch (optname) {
1461 
1462 			case IPV6_RECVHOPOPTS:
1463 			case IPV6_RECVDSTOPTS:
1464 			case IPV6_UNICAST_HOPS:
1465 			case IPV6_MINHOPCOUNT:
1466 			case IPV6_RECVPKTINFO:
1467 			case IPV6_RECVHOPLIMIT:
1468 			case IPV6_RECVRTHDR:
1469 			case IPV6_RECVPATHMTU:
1470 
1471 			case IPV6_V6ONLY:
1472 			case IPV6_PORTRANGE:
1473 			case IPV6_RECVTCLASS:
1474 			case IPV6_AUTOFLOWLABEL:
1475 			case IPV6_RECVDSTPORT:
1476 				switch (optname) {
1477 
1478 				case IPV6_RECVHOPOPTS:
1479 					optval = OPTBIT(IN6P_HOPOPTS);
1480 					break;
1481 
1482 				case IPV6_RECVDSTOPTS:
1483 					optval = OPTBIT(IN6P_DSTOPTS);
1484 					break;
1485 
1486 				case IPV6_UNICAST_HOPS:
1487 					optval = inp->inp_hops;
1488 					break;
1489 
1490 				case IPV6_MINHOPCOUNT:
1491 					optval = inp->inp_ip6_minhlim;
1492 					break;
1493 
1494 				case IPV6_RECVPKTINFO:
1495 					optval = OPTBIT(IN6P_PKTINFO);
1496 					break;
1497 
1498 				case IPV6_RECVHOPLIMIT:
1499 					optval = OPTBIT(IN6P_HOPLIMIT);
1500 					break;
1501 
1502 				case IPV6_RECVRTHDR:
1503 					optval = OPTBIT(IN6P_RTHDR);
1504 					break;
1505 
1506 				case IPV6_RECVPATHMTU:
1507 					optval = OPTBIT(IN6P_MTU);
1508 					break;
1509 
1510 				case IPV6_V6ONLY:
1511 					optval = 1;
1512 					break;
1513 
1514 				case IPV6_PORTRANGE:
1515 				    {
1516 					int flags;
1517 					flags = inp->inp_flags;
1518 					if (flags & IN6P_HIGHPORT)
1519 						optval = IPV6_PORTRANGE_HIGH;
1520 					else if (flags & IN6P_LOWPORT)
1521 						optval = IPV6_PORTRANGE_LOW;
1522 					else
1523 						optval = 0;
1524 					break;
1525 				    }
1526 				case IPV6_RECVTCLASS:
1527 					optval = OPTBIT(IN6P_TCLASS);
1528 					break;
1529 
1530 				case IPV6_AUTOFLOWLABEL:
1531 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1532 					break;
1533 
1534 				case IPV6_RECVDSTPORT:
1535 					optval = OPTBIT(IN6P_RECVDSTPORT);
1536 					break;
1537 				}
1538 				if (error)
1539 					break;
1540 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1541 				m->m_len = sizeof(int);
1542 				*mtod(m, int *) = optval;
1543 				break;
1544 
1545 			case IPV6_PATHMTU:
1546 			{
1547 				u_long pmtu = 0;
1548 				struct ip6_mtuinfo mtuinfo;
1549 				struct route_in6 *ro = (struct route_in6 *)&inp->inp_route6;
1550 
1551 				if (!(so->so_state & SS_ISCONNECTED))
1552 					return (ENOTCONN);
1553 				/*
1554 				 * XXX: we dot not consider the case of source
1555 				 * routing, or optional information to specify
1556 				 * the outgoing interface.
1557 				 */
1558 				error = ip6_getpmtu(ro, NULL, NULL,
1559 				    inp->inp_rtableid, &inp->inp_faddr6, &pmtu,
1560 				    NULL);
1561 				if (error)
1562 					break;
1563 				if (pmtu > IPV6_MAXPACKET)
1564 					pmtu = IPV6_MAXPACKET;
1565 
1566 				bzero(&mtuinfo, sizeof(mtuinfo));
1567 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1568 				optdata = (void *)&mtuinfo;
1569 				optdatalen = sizeof(mtuinfo);
1570 				if (optdatalen > MCLBYTES)
1571 					return (EMSGSIZE); /* XXX */
1572 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1573 				if (optdatalen > MLEN)
1574 					MCLGET(m, M_WAIT);
1575 				m->m_len = optdatalen;
1576 				bcopy(optdata, mtod(m, void *), optdatalen);
1577 				break;
1578 			}
1579 
1580 			case IPV6_PKTINFO:
1581 			case IPV6_HOPOPTS:
1582 			case IPV6_RTHDR:
1583 			case IPV6_DSTOPTS:
1584 			case IPV6_RTHDRDSTOPTS:
1585 			case IPV6_TCLASS:
1586 			case IPV6_DONTFRAG:
1587 			case IPV6_USE_MIN_MTU:
1588 				error = ip6_getpcbopt(inp->inp_outputopts6,
1589 				    optname, mp);
1590 				break;
1591 
1592 			case IPV6_MULTICAST_IF:
1593 			case IPV6_MULTICAST_HOPS:
1594 			case IPV6_MULTICAST_LOOP:
1595 			case IPV6_JOIN_GROUP:
1596 			case IPV6_LEAVE_GROUP:
1597 				error = ip6_getmoptions(optname,
1598 				    inp->inp_moptions6, mp);
1599 				break;
1600 
1601 			case IPSEC6_OUTSA:
1602 				error = EINVAL;
1603 				break;
1604 
1605 			case IPV6_AUTH_LEVEL:
1606 			case IPV6_ESP_TRANS_LEVEL:
1607 			case IPV6_ESP_NETWORK_LEVEL:
1608 			case IPV6_IPCOMP_LEVEL:
1609 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1610 #ifndef IPSEC
1611 				m->m_len = sizeof(int);
1612 				*mtod(m, int *) = IPSEC_LEVEL_NONE;
1613 #else
1614 				m->m_len = sizeof(int);
1615 				switch (optname) {
1616 				case IPV6_AUTH_LEVEL:
1617 					optval = inp->inp_seclevel[SL_AUTH];
1618 					break;
1619 
1620 				case IPV6_ESP_TRANS_LEVEL:
1621 					optval =
1622 					    inp->inp_seclevel[SL_ESP_TRANS];
1623 					break;
1624 
1625 				case IPV6_ESP_NETWORK_LEVEL:
1626 					optval =
1627 					    inp->inp_seclevel[SL_ESP_NETWORK];
1628 					break;
1629 
1630 				case IPV6_IPCOMP_LEVEL:
1631 					optval = inp->inp_seclevel[SL_IPCOMP];
1632 					break;
1633 				}
1634 				*mtod(m, int *) = optval;
1635 #endif
1636 				break;
1637 			case SO_RTABLE:
1638 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1639 				m->m_len = sizeof(u_int);
1640 				*mtod(m, u_int *) = optval;
1641 				break;
1642 			case IPV6_PIPEX:
1643 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1644 				m->m_len = sizeof(int);
1645 				*mtod(m, int *) = optval;
1646 				break;
1647 
1648 			default:
1649 				error = ENOPROTOOPT;
1650 				break;
1651 			}
1652 			break;
1653 		}
1654 	} else {
1655 		error = EINVAL;
1656 		if (op == PRCO_SETOPT)
1657 			(void)m_free(*mp);
1658 	}
1659 	return (error);
1660 }
1661 
1662 int
1663 ip6_raw_ctloutput(int op, struct socket *so, int level, int optname,
1664     struct mbuf **mp)
1665 {
1666 	int error = 0, optval;
1667 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
1668 	struct inpcb *inp = sotoinpcb(so);
1669 	struct mbuf *m = *mp;
1670 
1671 	if (level != IPPROTO_IPV6) {
1672 		if (op == PRCO_SETOPT)
1673 			(void)m_free(*mp);
1674 		return (EINVAL);
1675 	}
1676 
1677 	switch (optname) {
1678 	case IPV6_CHECKSUM:
1679 		/*
1680 		 * For ICMPv6 sockets, no modification allowed for checksum
1681 		 * offset, permit "no change" values to help existing apps.
1682 		 *
1683 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
1684 		 * for an ICMPv6 socket will fail."
1685 		 * The current behavior does not meet RFC3542.
1686 		 */
1687 		switch (op) {
1688 		case PRCO_SETOPT:
1689 			if (m == NULL || m->m_len != sizeof(int)) {
1690 				error = EINVAL;
1691 				break;
1692 			}
1693 			optval = *mtod(m, int *);
1694 			if ((optval % 2) != 0) {
1695 				/* the API assumes even offset values */
1696 				error = EINVAL;
1697 			} else if (so->so_proto->pr_protocol == IPPROTO_ICMPV6) {
1698 				if (optval != icmp6off)
1699 					error = EINVAL;
1700 			} else
1701 				inp->inp_cksum6 = optval;
1702 			break;
1703 
1704 		case PRCO_GETOPT:
1705 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
1706 				optval = icmp6off;
1707 			else
1708 				optval = inp->inp_cksum6;
1709 
1710 			*mp = m = m_get(M_WAIT, MT_SOOPTS);
1711 			m->m_len = sizeof(int);
1712 			*mtod(m, int *) = optval;
1713 			break;
1714 
1715 		default:
1716 			error = EINVAL;
1717 			break;
1718 		}
1719 		break;
1720 
1721 	default:
1722 		error = ENOPROTOOPT;
1723 		break;
1724 	}
1725 
1726 	if (op == PRCO_SETOPT)
1727 		(void)m_free(m);
1728 
1729 	return (error);
1730 }
1731 
1732 /*
1733  * Set up IP6 options in pcb for insertion in output packets.
1734  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1735  * with destination address if source routed.
1736  */
1737 int
1738 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m, struct socket *so)
1739 {
1740 	struct ip6_pktopts *opt = *pktopt;
1741 	int error = 0;
1742 	struct proc *p = curproc;	/* XXX */
1743 	int priv = 0;
1744 
1745 	/* turn off any old options. */
1746 	if (opt)
1747 		ip6_clearpktopts(opt, -1);
1748 	else
1749 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
1750 	*pktopt = 0;
1751 
1752 	if (!m || m->m_len == 0) {
1753 		/*
1754 		 * Only turning off any previous options, regardless of
1755 		 * whether the opt is just created or given.
1756 		 */
1757 		free(opt, M_IP6OPT, 0);
1758 		return (0);
1759 	}
1760 
1761 	/*  set options specified by user. */
1762 	if (p && !suser(p, 0))
1763 		priv = 1;
1764 	if ((error = ip6_setpktopts(m, opt, NULL, priv,
1765 	    so->so_proto->pr_protocol)) != 0) {
1766 		ip6_clearpktopts(opt, -1);	/* XXX discard all options */
1767 		free(opt, M_IP6OPT, 0);
1768 		return (error);
1769 	}
1770 	*pktopt = opt;
1771 	return (0);
1772 }
1773 
1774 /*
1775  * initialize ip6_pktopts.  beware that there are non-zero default values in
1776  * the struct.
1777  */
1778 void
1779 ip6_initpktopts(struct ip6_pktopts *opt)
1780 {
1781 
1782 	bzero(opt, sizeof(*opt));
1783 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
1784 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
1785 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
1786 }
1787 
1788 int
1789 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
1790     int priv, int uproto)
1791 {
1792 	struct ip6_pktopts *opt;
1793 
1794 	if (*pktopt == NULL) {
1795 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
1796 		    M_WAITOK);
1797 		ip6_initpktopts(*pktopt);
1798 	}
1799 	opt = *pktopt;
1800 
1801 	return (ip6_setpktopt(optname, buf, len, opt, priv, 1, 0, uproto));
1802 }
1803 
1804 int
1805 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct mbuf **mp)
1806 {
1807 	void *optdata = NULL;
1808 	int optdatalen = 0;
1809 	struct ip6_ext *ip6e;
1810 	int error = 0;
1811 	struct in6_pktinfo null_pktinfo;
1812 	int deftclass = 0, on;
1813 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
1814 	struct mbuf *m;
1815 
1816 	switch (optname) {
1817 	case IPV6_PKTINFO:
1818 		if (pktopt && pktopt->ip6po_pktinfo)
1819 			optdata = (void *)pktopt->ip6po_pktinfo;
1820 		else {
1821 			/* XXX: we don't have to do this every time... */
1822 			bzero(&null_pktinfo, sizeof(null_pktinfo));
1823 			optdata = (void *)&null_pktinfo;
1824 		}
1825 		optdatalen = sizeof(struct in6_pktinfo);
1826 		break;
1827 	case IPV6_TCLASS:
1828 		if (pktopt && pktopt->ip6po_tclass >= 0)
1829 			optdata = (void *)&pktopt->ip6po_tclass;
1830 		else
1831 			optdata = (void *)&deftclass;
1832 		optdatalen = sizeof(int);
1833 		break;
1834 	case IPV6_HOPOPTS:
1835 		if (pktopt && pktopt->ip6po_hbh) {
1836 			optdata = (void *)pktopt->ip6po_hbh;
1837 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
1838 			optdatalen = (ip6e->ip6e_len + 1) << 3;
1839 		}
1840 		break;
1841 	case IPV6_RTHDR:
1842 		if (pktopt && pktopt->ip6po_rthdr) {
1843 			optdata = (void *)pktopt->ip6po_rthdr;
1844 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
1845 			optdatalen = (ip6e->ip6e_len + 1) << 3;
1846 		}
1847 		break;
1848 	case IPV6_RTHDRDSTOPTS:
1849 		if (pktopt && pktopt->ip6po_dest1) {
1850 			optdata = (void *)pktopt->ip6po_dest1;
1851 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
1852 			optdatalen = (ip6e->ip6e_len + 1) << 3;
1853 		}
1854 		break;
1855 	case IPV6_DSTOPTS:
1856 		if (pktopt && pktopt->ip6po_dest2) {
1857 			optdata = (void *)pktopt->ip6po_dest2;
1858 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
1859 			optdatalen = (ip6e->ip6e_len + 1) << 3;
1860 		}
1861 		break;
1862 	case IPV6_USE_MIN_MTU:
1863 		if (pktopt)
1864 			optdata = (void *)&pktopt->ip6po_minmtu;
1865 		else
1866 			optdata = (void *)&defminmtu;
1867 		optdatalen = sizeof(int);
1868 		break;
1869 	case IPV6_DONTFRAG:
1870 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
1871 			on = 1;
1872 		else
1873 			on = 0;
1874 		optdata = (void *)&on;
1875 		optdatalen = sizeof(on);
1876 		break;
1877 	default:		/* should not happen */
1878 #ifdef DIAGNOSTIC
1879 		panic("ip6_getpcbopt: unexpected option");
1880 #endif
1881 		return (ENOPROTOOPT);
1882 	}
1883 
1884 	if (optdatalen > MCLBYTES)
1885 		return (EMSGSIZE); /* XXX */
1886 	*mp = m = m_get(M_WAIT, MT_SOOPTS);
1887 	if (optdatalen > MLEN)
1888 		MCLGET(m, M_WAIT);
1889 	m->m_len = optdatalen;
1890 	if (optdatalen)
1891 		bcopy(optdata, mtod(m, void *), optdatalen);
1892 
1893 	return (error);
1894 }
1895 
1896 void
1897 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
1898 {
1899 	if (optname == -1 || optname == IPV6_PKTINFO) {
1900 		if (pktopt->ip6po_pktinfo)
1901 			free(pktopt->ip6po_pktinfo, M_IP6OPT, 0);
1902 		pktopt->ip6po_pktinfo = NULL;
1903 	}
1904 	if (optname == -1 || optname == IPV6_HOPLIMIT)
1905 		pktopt->ip6po_hlim = -1;
1906 	if (optname == -1 || optname == IPV6_TCLASS)
1907 		pktopt->ip6po_tclass = -1;
1908 	if (optname == -1 || optname == IPV6_HOPOPTS) {
1909 		if (pktopt->ip6po_hbh)
1910 			free(pktopt->ip6po_hbh, M_IP6OPT, 0);
1911 		pktopt->ip6po_hbh = NULL;
1912 	}
1913 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
1914 		if (pktopt->ip6po_dest1)
1915 			free(pktopt->ip6po_dest1, M_IP6OPT, 0);
1916 		pktopt->ip6po_dest1 = NULL;
1917 	}
1918 	if (optname == -1 || optname == IPV6_RTHDR) {
1919 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
1920 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT, 0);
1921 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
1922 		if (pktopt->ip6po_route.ro_rt) {
1923 			rtfree(pktopt->ip6po_route.ro_rt);
1924 			pktopt->ip6po_route.ro_rt = NULL;
1925 		}
1926 	}
1927 	if (optname == -1 || optname == IPV6_DSTOPTS) {
1928 		if (pktopt->ip6po_dest2)
1929 			free(pktopt->ip6po_dest2, M_IP6OPT, 0);
1930 		pktopt->ip6po_dest2 = NULL;
1931 	}
1932 }
1933 
1934 #define PKTOPT_EXTHDRCPY(type) \
1935 do {\
1936 	if (src->type) {\
1937 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
1938 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
1939 		if (dst->type == NULL && canwait == M_NOWAIT)\
1940 			goto bad;\
1941 		bcopy(src->type, dst->type, hlen);\
1942 	}\
1943 } while (/*CONSTCOND*/ 0)
1944 
1945 int
1946 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
1947 {
1948 	dst->ip6po_hlim = src->ip6po_hlim;
1949 	dst->ip6po_tclass = src->ip6po_tclass;
1950 	dst->ip6po_flags = src->ip6po_flags;
1951 	if (src->ip6po_pktinfo) {
1952 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
1953 		    M_IP6OPT, canwait);
1954 		if (dst->ip6po_pktinfo == NULL)
1955 			goto bad;
1956 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
1957 	}
1958 	PKTOPT_EXTHDRCPY(ip6po_hbh);
1959 	PKTOPT_EXTHDRCPY(ip6po_dest1);
1960 	PKTOPT_EXTHDRCPY(ip6po_dest2);
1961 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
1962 	return (0);
1963 
1964   bad:
1965 	ip6_clearpktopts(dst, -1);
1966 	return (ENOBUFS);
1967 }
1968 #undef PKTOPT_EXTHDRCPY
1969 
1970 void
1971 ip6_freepcbopts(struct ip6_pktopts *pktopt)
1972 {
1973 	if (pktopt == NULL)
1974 		return;
1975 
1976 	ip6_clearpktopts(pktopt, -1);
1977 
1978 	free(pktopt, M_IP6OPT, 0);
1979 }
1980 
1981 /*
1982  * Set the IP6 multicast options in response to user setsockopt().
1983  */
1984 int
1985 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m)
1986 {
1987 	int error = 0;
1988 	u_int loop, ifindex;
1989 	struct ipv6_mreq *mreq;
1990 	struct ifnet *ifp;
1991 	struct ip6_moptions *im6o = *im6op;
1992 	struct route_in6 ro;
1993 	struct sockaddr_in6 *dst;
1994 	struct in6_multi_mship *imm;
1995 	struct proc *p = curproc;	/* XXX */
1996 
1997 	if (im6o == NULL) {
1998 		/*
1999 		 * No multicast option buffer attached to the pcb;
2000 		 * allocate one and initialize to default values.
2001 		 */
2002 		im6o = (struct ip6_moptions *)
2003 			malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
2004 
2005 		if (im6o == NULL)
2006 			return (ENOBUFS);
2007 		*im6op = im6o;
2008 		im6o->im6o_ifidx = 0;
2009 		im6o->im6o_hlim = ip6_defmcasthlim;
2010 		im6o->im6o_loop = IPV6_DEFAULT_MULTICAST_LOOP;
2011 		LIST_INIT(&im6o->im6o_memberships);
2012 	}
2013 
2014 	switch (optname) {
2015 
2016 	case IPV6_MULTICAST_IF:
2017 		/*
2018 		 * Select the interface for outgoing multicast packets.
2019 		 */
2020 		if (m == NULL || m->m_len != sizeof(u_int)) {
2021 			error = EINVAL;
2022 			break;
2023 		}
2024 		bcopy(mtod(m, u_int *), &ifindex, sizeof(ifindex));
2025 		if (ifindex != 0) {
2026 			ifp = if_get(ifindex);
2027 			if (ifp == NULL) {
2028 				error = ENXIO;	/* XXX EINVAL? */
2029 				break;
2030 			}
2031 			if ((ifp->if_flags & IFF_MULTICAST) == 0) {
2032 				error = EADDRNOTAVAIL;
2033 				if_put(ifp);
2034 				break;
2035 			}
2036 			if_put(ifp);
2037 		}
2038 		im6o->im6o_ifidx = ifindex;
2039 		break;
2040 
2041 	case IPV6_MULTICAST_HOPS:
2042 	    {
2043 		/*
2044 		 * Set the IP6 hoplimit for outgoing multicast packets.
2045 		 */
2046 		int optval;
2047 		if (m == NULL || m->m_len != sizeof(int)) {
2048 			error = EINVAL;
2049 			break;
2050 		}
2051 		bcopy(mtod(m, u_int *), &optval, sizeof(optval));
2052 		if (optval < -1 || optval >= 256)
2053 			error = EINVAL;
2054 		else if (optval == -1)
2055 			im6o->im6o_hlim = ip6_defmcasthlim;
2056 		else
2057 			im6o->im6o_hlim = optval;
2058 		break;
2059 	    }
2060 
2061 	case IPV6_MULTICAST_LOOP:
2062 		/*
2063 		 * Set the loopback flag for outgoing multicast packets.
2064 		 * Must be zero or one.
2065 		 */
2066 		if (m == NULL || m->m_len != sizeof(u_int)) {
2067 			error = EINVAL;
2068 			break;
2069 		}
2070 		bcopy(mtod(m, u_int *), &loop, sizeof(loop));
2071 		if (loop > 1) {
2072 			error = EINVAL;
2073 			break;
2074 		}
2075 		im6o->im6o_loop = loop;
2076 		break;
2077 
2078 	case IPV6_JOIN_GROUP:
2079 		/*
2080 		 * Add a multicast group membership.
2081 		 * Group must be a valid IP6 multicast address.
2082 		 */
2083 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2084 			error = EINVAL;
2085 			break;
2086 		}
2087 		mreq = mtod(m, struct ipv6_mreq *);
2088 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2089 			/*
2090 			 * We use the unspecified address to specify to accept
2091 			 * all multicast addresses. Only super user is allowed
2092 			 * to do this.
2093 			 */
2094 			if (suser(p, 0))
2095 			{
2096 				error = EACCES;
2097 				break;
2098 			}
2099 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2100 			error = EINVAL;
2101 			break;
2102 		}
2103 
2104 		/*
2105 		 * If no interface was explicitly specified, choose an
2106 		 * appropriate one according to the given multicast address.
2107 		 */
2108 		if (mreq->ipv6mr_interface == 0) {
2109 			/*
2110 			 * Look up the routing table for the
2111 			 * address, and choose the outgoing interface.
2112 			 *   XXX: is it a good approach?
2113 			 */
2114 			bzero(&ro, sizeof(ro));
2115 			ro.ro_tableid = m->m_pkthdr.ph_rtableid;
2116 			dst = &ro.ro_dst;
2117 			dst->sin6_len = sizeof(struct sockaddr_in6);
2118 			dst->sin6_family = AF_INET6;
2119 			dst->sin6_addr = mreq->ipv6mr_multiaddr;
2120 			ro.ro_rt = rtalloc(sin6tosa(&ro.ro_dst),
2121 			    RT_RESOLVE, ro.ro_tableid);
2122 			if (ro.ro_rt == NULL) {
2123 				error = EADDRNOTAVAIL;
2124 				break;
2125 			}
2126 			ifp = if_get(ro.ro_rt->rt_ifidx);
2127 			rtfree(ro.ro_rt);
2128 		} else {
2129 			/*
2130 			 * If the interface is specified, validate it.
2131 			 */
2132 			ifp = if_get(mreq->ipv6mr_interface);
2133 			if (ifp == NULL) {
2134 				error = ENXIO;	/* XXX EINVAL? */
2135 				break;
2136 			}
2137 		}
2138 
2139 		/*
2140 		 * See if we found an interface, and confirm that it
2141 		 * supports multicast
2142 		 */
2143 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
2144 			if_put(ifp);
2145 			error = EADDRNOTAVAIL;
2146 			break;
2147 		}
2148 		/*
2149 		 * Put interface index into the multicast address,
2150 		 * if the address has link/interface-local scope.
2151 		 */
2152 		if (IN6_IS_SCOPE_EMBED(&mreq->ipv6mr_multiaddr)) {
2153 			mreq->ipv6mr_multiaddr.s6_addr16[1] =
2154 			    htons(ifp->if_index);
2155 		}
2156 		/*
2157 		 * See if the membership already exists.
2158 		 */
2159 		LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain)
2160 			if (imm->i6mm_maddr->in6m_ifidx == ifp->if_index &&
2161 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2162 			    &mreq->ipv6mr_multiaddr))
2163 				break;
2164 		if (imm != NULL) {
2165 			if_put(ifp);
2166 			error = EADDRINUSE;
2167 			break;
2168 		}
2169 		/*
2170 		 * Everything looks good; add a new record to the multicast
2171 		 * address list for the given interface.
2172 		 */
2173 		imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error);
2174 		if_put(ifp);
2175 		if (!imm)
2176 			break;
2177 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2178 		break;
2179 
2180 	case IPV6_LEAVE_GROUP:
2181 		/*
2182 		 * Drop a multicast group membership.
2183 		 * Group must be a valid IP6 multicast address.
2184 		 */
2185 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2186 			error = EINVAL;
2187 			break;
2188 		}
2189 		mreq = mtod(m, struct ipv6_mreq *);
2190 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2191 			if (suser(p, 0))
2192 			{
2193 				error = EACCES;
2194 				break;
2195 			}
2196 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2197 			error = EINVAL;
2198 			break;
2199 		}
2200 
2201 		/*
2202 		 * Put interface index into the multicast address,
2203 		 * if the address has link-local scope.
2204 		 */
2205 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2206 			mreq->ipv6mr_multiaddr.s6_addr16[1] =
2207 			    htons(mreq->ipv6mr_interface);
2208 		}
2209 
2210 		/*
2211 		 * If an interface address was specified, get a pointer
2212 		 * to its ifnet structure.
2213 		 */
2214 		if (mreq->ipv6mr_interface == 0)
2215 			ifp = NULL;
2216 		else {
2217 			ifp = if_get(mreq->ipv6mr_interface);
2218 			if (ifp == NULL) {
2219 				error = ENXIO;	/* XXX EINVAL? */
2220 				break;
2221 			}
2222 		}
2223 
2224 		/*
2225 		 * Find the membership in the membership list.
2226 		 */
2227 		LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain) {
2228 			if ((ifp == NULL ||
2229 			    imm->i6mm_maddr->in6m_ifidx == ifp->if_index) &&
2230 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2231 			    &mreq->ipv6mr_multiaddr))
2232 				break;
2233 		}
2234 
2235 		if_put(ifp);
2236 
2237 		if (imm == NULL) {
2238 			/* Unable to resolve interface */
2239 			error = EADDRNOTAVAIL;
2240 			break;
2241 		}
2242 		/*
2243 		 * Give up the multicast address record to which the
2244 		 * membership points.
2245 		 */
2246 		LIST_REMOVE(imm, i6mm_chain);
2247 		in6_leavegroup(imm);
2248 		break;
2249 
2250 	default:
2251 		error = EOPNOTSUPP;
2252 		break;
2253 	}
2254 
2255 	/*
2256 	 * If all options have default values, no need to keep the option
2257 	 * structure.
2258 	 */
2259 	if (im6o->im6o_ifidx == 0 &&
2260 	    im6o->im6o_hlim == ip6_defmcasthlim &&
2261 	    im6o->im6o_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2262 	    LIST_EMPTY(&im6o->im6o_memberships)) {
2263 		free(*im6op, M_IPMOPTS, 0);
2264 		*im6op = NULL;
2265 	}
2266 
2267 	return (error);
2268 }
2269 
2270 /*
2271  * Return the IP6 multicast options in response to user getsockopt().
2272  */
2273 int
2274 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf **mp)
2275 {
2276 	u_int *hlim, *loop, *ifindex;
2277 
2278 	*mp = m_get(M_WAIT, MT_SOOPTS);
2279 
2280 	switch (optname) {
2281 
2282 	case IPV6_MULTICAST_IF:
2283 		ifindex = mtod(*mp, u_int *);
2284 		(*mp)->m_len = sizeof(u_int);
2285 		if (im6o == NULL || im6o->im6o_ifidx == 0)
2286 			*ifindex = 0;
2287 		else
2288 			*ifindex = im6o->im6o_ifidx;
2289 		return (0);
2290 
2291 	case IPV6_MULTICAST_HOPS:
2292 		hlim = mtod(*mp, u_int *);
2293 		(*mp)->m_len = sizeof(u_int);
2294 		if (im6o == NULL)
2295 			*hlim = ip6_defmcasthlim;
2296 		else
2297 			*hlim = im6o->im6o_hlim;
2298 		return (0);
2299 
2300 	case IPV6_MULTICAST_LOOP:
2301 		loop = mtod(*mp, u_int *);
2302 		(*mp)->m_len = sizeof(u_int);
2303 		if (im6o == NULL)
2304 			*loop = ip6_defmcasthlim;
2305 		else
2306 			*loop = im6o->im6o_loop;
2307 		return (0);
2308 
2309 	default:
2310 		return (EOPNOTSUPP);
2311 	}
2312 }
2313 
2314 /*
2315  * Discard the IP6 multicast options.
2316  */
2317 void
2318 ip6_freemoptions(struct ip6_moptions *im6o)
2319 {
2320 	struct in6_multi_mship *imm;
2321 
2322 	if (im6o == NULL)
2323 		return;
2324 
2325 	while (!LIST_EMPTY(&im6o->im6o_memberships)) {
2326 		imm = LIST_FIRST(&im6o->im6o_memberships);
2327 		LIST_REMOVE(imm, i6mm_chain);
2328 		in6_leavegroup(imm);
2329 	}
2330 	free(im6o, M_IPMOPTS, 0);
2331 }
2332 
2333 /*
2334  * Set IPv6 outgoing packet options based on advanced API.
2335  */
2336 int
2337 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2338     struct ip6_pktopts *stickyopt, int priv, int uproto)
2339 {
2340 	u_int clen;
2341 	struct cmsghdr *cm = 0;
2342 	caddr_t cmsgs;
2343 	int error;
2344 
2345 	if (control == NULL || opt == NULL)
2346 		return (EINVAL);
2347 
2348 	ip6_initpktopts(opt);
2349 	if (stickyopt) {
2350 		int error;
2351 
2352 		/*
2353 		 * If stickyopt is provided, make a local copy of the options
2354 		 * for this particular packet, then override them by ancillary
2355 		 * objects.
2356 		 * XXX: copypktopts() does not copy the cached route to a next
2357 		 * hop (if any).  This is not very good in terms of efficiency,
2358 		 * but we can allow this since this option should be rarely
2359 		 * used.
2360 		 */
2361 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2362 			return (error);
2363 	}
2364 
2365 	/*
2366 	 * XXX: Currently, we assume all the optional information is stored
2367 	 * in a single mbuf.
2368 	 */
2369 	if (control->m_next)
2370 		return (EINVAL);
2371 
2372 	clen = control->m_len;
2373 	cmsgs = mtod(control, caddr_t);
2374 	do {
2375 		if (clen < CMSG_LEN(0))
2376 			return (EINVAL);
2377 		cm = (struct cmsghdr *)cmsgs;
2378 		if (cm->cmsg_len < CMSG_LEN(0) || cm->cmsg_len > clen ||
2379 		    CMSG_ALIGN(cm->cmsg_len) > clen)
2380 			return (EINVAL);
2381 		if (cm->cmsg_level == IPPROTO_IPV6) {
2382 			error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2383 			    cm->cmsg_len - CMSG_LEN(0), opt, priv, 0, 1, uproto);
2384 			if (error)
2385 				return (error);
2386 		}
2387 
2388 		clen -= CMSG_ALIGN(cm->cmsg_len);
2389 		cmsgs += CMSG_ALIGN(cm->cmsg_len);
2390 	} while (clen);
2391 
2392 	return (0);
2393 }
2394 
2395 /*
2396  * Set a particular packet option, as a sticky option or an ancillary data
2397  * item.  "len" can be 0 only when it's a sticky option.
2398  * We have 4 cases of combination of "sticky" and "cmsg":
2399  * "sticky=0, cmsg=0": impossible
2400  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2401  * "sticky=1, cmsg=0": RFC3542 socket option
2402  * "sticky=1, cmsg=1": RFC2292 socket option
2403  */
2404 int
2405 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2406     int priv, int sticky, int cmsg, int uproto)
2407 {
2408 	int minmtupolicy;
2409 
2410 	if (!sticky && !cmsg) {
2411 #ifdef DIAGNOSTIC
2412 		printf("ip6_setpktopt: impossible case\n");
2413 #endif
2414 		return (EINVAL);
2415 	}
2416 
2417 	if (sticky && cmsg) {
2418 		switch (optname) {
2419 		case IPV6_PKTINFO:
2420 		case IPV6_HOPLIMIT:
2421 		case IPV6_HOPOPTS:
2422 		case IPV6_DSTOPTS:
2423 		case IPV6_RTHDRDSTOPTS:
2424 		case IPV6_RTHDR:
2425 		case IPV6_USE_MIN_MTU:
2426 		case IPV6_DONTFRAG:
2427 		case IPV6_TCLASS:
2428 			return (ENOPROTOOPT);
2429 		}
2430 	}
2431 
2432 	switch (optname) {
2433 	case IPV6_PKTINFO:
2434 	{
2435 		struct ifnet *ifp = NULL;
2436 		struct in6_pktinfo *pktinfo;
2437 
2438 		if (len != sizeof(struct in6_pktinfo))
2439 			return (EINVAL);
2440 
2441 		pktinfo = (struct in6_pktinfo *)buf;
2442 
2443 		/*
2444 		 * An application can clear any sticky IPV6_PKTINFO option by
2445 		 * doing a "regular" setsockopt with ipi6_addr being
2446 		 * in6addr_any and ipi6_ifindex being zero.
2447 		 * [RFC 3542, Section 6]
2448 		 */
2449 		if (opt->ip6po_pktinfo &&
2450 		    pktinfo->ipi6_ifindex == 0 &&
2451 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2452 			ip6_clearpktopts(opt, optname);
2453 			break;
2454 		}
2455 
2456 		if (uproto == IPPROTO_TCP &&
2457 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2458 			return (EINVAL);
2459 		}
2460 
2461 		if (pktinfo->ipi6_ifindex) {
2462 			ifp = if_get(pktinfo->ipi6_ifindex);
2463 			if (ifp == NULL)
2464 				return (ENXIO);
2465 			if_put(ifp);
2466 		}
2467 
2468 		/*
2469 		 * We store the address anyway, and let in6_selectsrc()
2470 		 * validate the specified address.  This is because ipi6_addr
2471 		 * may not have enough information about its scope zone, and
2472 		 * we may need additional information (such as outgoing
2473 		 * interface or the scope zone of a destination address) to
2474 		 * disambiguate the scope.
2475 		 * XXX: the delay of the validation may confuse the
2476 		 * application when it is used as a sticky option.
2477 		 */
2478 		if (opt->ip6po_pktinfo == NULL) {
2479 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2480 			    M_IP6OPT, M_NOWAIT);
2481 			if (opt->ip6po_pktinfo == NULL)
2482 				return (ENOBUFS);
2483 		}
2484 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2485 		break;
2486 	}
2487 
2488 	case IPV6_HOPLIMIT:
2489 	{
2490 		int *hlimp;
2491 
2492 		/*
2493 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2494 		 * to simplify the ordering among hoplimit options.
2495 		 */
2496 		if (sticky)
2497 			return (ENOPROTOOPT);
2498 
2499 		if (len != sizeof(int))
2500 			return (EINVAL);
2501 		hlimp = (int *)buf;
2502 		if (*hlimp < -1 || *hlimp > 255)
2503 			return (EINVAL);
2504 
2505 		opt->ip6po_hlim = *hlimp;
2506 		break;
2507 	}
2508 
2509 	case IPV6_TCLASS:
2510 	{
2511 		int tclass;
2512 
2513 		if (len != sizeof(int))
2514 			return (EINVAL);
2515 		tclass = *(int *)buf;
2516 		if (tclass < -1 || tclass > 255)
2517 			return (EINVAL);
2518 
2519 		opt->ip6po_tclass = tclass;
2520 		break;
2521 	}
2522 	case IPV6_HOPOPTS:
2523 	{
2524 		struct ip6_hbh *hbh;
2525 		int hbhlen;
2526 
2527 		/*
2528 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2529 		 * options, since per-option restriction has too much
2530 		 * overhead.
2531 		 */
2532 		if (!priv)
2533 			return (EPERM);
2534 
2535 		if (len == 0) {
2536 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2537 			break;	/* just remove the option */
2538 		}
2539 
2540 		/* message length validation */
2541 		if (len < sizeof(struct ip6_hbh))
2542 			return (EINVAL);
2543 		hbh = (struct ip6_hbh *)buf;
2544 		hbhlen = (hbh->ip6h_len + 1) << 3;
2545 		if (len != hbhlen)
2546 			return (EINVAL);
2547 
2548 		/* turn off the previous option, then set the new option. */
2549 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2550 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2551 		if (opt->ip6po_hbh == NULL)
2552 			return (ENOBUFS);
2553 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
2554 
2555 		break;
2556 	}
2557 
2558 	case IPV6_DSTOPTS:
2559 	case IPV6_RTHDRDSTOPTS:
2560 	{
2561 		struct ip6_dest *dest, **newdest = NULL;
2562 		int destlen;
2563 
2564 		if (!priv)	/* XXX: see the comment for IPV6_HOPOPTS */
2565 			return (EPERM);
2566 
2567 		if (len == 0) {
2568 			ip6_clearpktopts(opt, optname);
2569 			break;	/* just remove the option */
2570 		}
2571 
2572 		/* message length validation */
2573 		if (len < sizeof(struct ip6_dest))
2574 			return (EINVAL);
2575 		dest = (struct ip6_dest *)buf;
2576 		destlen = (dest->ip6d_len + 1) << 3;
2577 		if (len != destlen)
2578 			return (EINVAL);
2579 		/*
2580 		 * Determine the position that the destination options header
2581 		 * should be inserted; before or after the routing header.
2582 		 */
2583 		switch (optname) {
2584 		case IPV6_RTHDRDSTOPTS:
2585 			newdest = &opt->ip6po_dest1;
2586 			break;
2587 		case IPV6_DSTOPTS:
2588 			newdest = &opt->ip6po_dest2;
2589 			break;
2590 		}
2591 
2592 		/* turn off the previous option, then set the new option. */
2593 		ip6_clearpktopts(opt, optname);
2594 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2595 		if (*newdest == NULL)
2596 			return (ENOBUFS);
2597 		bcopy(dest, *newdest, destlen);
2598 
2599 		break;
2600 	}
2601 
2602 	case IPV6_RTHDR:
2603 	{
2604 		struct ip6_rthdr *rth;
2605 		int rthlen;
2606 
2607 		if (len == 0) {
2608 			ip6_clearpktopts(opt, IPV6_RTHDR);
2609 			break;	/* just remove the option */
2610 		}
2611 
2612 		/* message length validation */
2613 		if (len < sizeof(struct ip6_rthdr))
2614 			return (EINVAL);
2615 		rth = (struct ip6_rthdr *)buf;
2616 		rthlen = (rth->ip6r_len + 1) << 3;
2617 		if (len != rthlen)
2618 			return (EINVAL);
2619 
2620 		switch (rth->ip6r_type) {
2621 		case IPV6_RTHDR_TYPE_0:
2622 			if (rth->ip6r_len == 0)	/* must contain one addr */
2623 				return (EINVAL);
2624 			if (rth->ip6r_len % 2) /* length must be even */
2625 				return (EINVAL);
2626 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2627 				return (EINVAL);
2628 			break;
2629 		default:
2630 			return (EINVAL);	/* not supported */
2631 		}
2632 		/* turn off the previous option */
2633 		ip6_clearpktopts(opt, IPV6_RTHDR);
2634 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2635 		if (opt->ip6po_rthdr == NULL)
2636 			return (ENOBUFS);
2637 		bcopy(rth, opt->ip6po_rthdr, rthlen);
2638 		break;
2639 	}
2640 
2641 	case IPV6_USE_MIN_MTU:
2642 		if (len != sizeof(int))
2643 			return (EINVAL);
2644 		minmtupolicy = *(int *)buf;
2645 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2646 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
2647 		    minmtupolicy != IP6PO_MINMTU_ALL) {
2648 			return (EINVAL);
2649 		}
2650 		opt->ip6po_minmtu = minmtupolicy;
2651 		break;
2652 
2653 	case IPV6_DONTFRAG:
2654 		if (len != sizeof(int))
2655 			return (EINVAL);
2656 
2657 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2658 			/*
2659 			 * we ignore this option for TCP sockets.
2660 			 * (RFC3542 leaves this case unspecified.)
2661 			 */
2662 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2663 		} else
2664 			opt->ip6po_flags |= IP6PO_DONTFRAG;
2665 		break;
2666 
2667 	default:
2668 		return (ENOPROTOOPT);
2669 	} /* end of switch */
2670 
2671 	return (0);
2672 }
2673 
2674 /*
2675  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2676  * packet to the input queue of a specified interface.
2677  */
2678 void
2679 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2680 {
2681 	struct mbuf *copym;
2682 	struct ip6_hdr *ip6;
2683 
2684 	/*
2685 	 * Duplicate the packet.
2686 	 */
2687 	copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
2688 	if (copym == NULL)
2689 		return;
2690 
2691 	/*
2692 	 * Make sure to deep-copy IPv6 header portion in case the data
2693 	 * is in an mbuf cluster, so that we can safely override the IPv6
2694 	 * header portion later.
2695 	 */
2696 	if ((copym->m_flags & M_EXT) != 0 ||
2697 	    copym->m_len < sizeof(struct ip6_hdr)) {
2698 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
2699 		if (copym == NULL)
2700 			return;
2701 	}
2702 
2703 #ifdef DIAGNOSTIC
2704 	if (copym->m_len < sizeof(*ip6)) {
2705 		m_freem(copym);
2706 		return;
2707 	}
2708 #endif
2709 
2710 	ip6 = mtod(copym, struct ip6_hdr *);
2711 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
2712 		ip6->ip6_src.s6_addr16[1] = 0;
2713 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
2714 		ip6->ip6_dst.s6_addr16[1] = 0;
2715 
2716 	if_input_local(ifp, copym, dst->sin6_family);
2717 }
2718 
2719 /*
2720  * Chop IPv6 header off from the payload.
2721  */
2722 int
2723 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
2724 {
2725 	struct mbuf *mh;
2726 	struct ip6_hdr *ip6;
2727 
2728 	ip6 = mtod(m, struct ip6_hdr *);
2729 	if (m->m_len > sizeof(*ip6)) {
2730 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
2731 		if (mh == NULL) {
2732 			m_freem(m);
2733 			return ENOBUFS;
2734 		}
2735 		M_MOVE_PKTHDR(mh, m);
2736 		MH_ALIGN(mh, sizeof(*ip6));
2737 		m->m_len -= sizeof(*ip6);
2738 		m->m_data += sizeof(*ip6);
2739 		mh->m_next = m;
2740 		m = mh;
2741 		m->m_len = sizeof(*ip6);
2742 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2743 	}
2744 	exthdrs->ip6e_ip6 = m;
2745 	return 0;
2746 }
2747 
2748 u_int32_t
2749 ip6_randomid(void)
2750 {
2751 	return idgen32(&ip6_id_ctx);
2752 }
2753 
2754 void
2755 ip6_randomid_init(void)
2756 {
2757 	idgen32_init(&ip6_id_ctx);
2758 }
2759 
2760 /*
2761  *	Compute significant parts of the IPv6 checksum pseudo-header
2762  *	for use in a delayed TCP/UDP checksum calculation.
2763  */
2764 static __inline u_int16_t __attribute__((__unused__))
2765 in6_cksum_phdr(const struct in6_addr *src, const struct in6_addr *dst,
2766     u_int32_t len, u_int32_t nxt)
2767 {
2768 	u_int32_t sum = 0;
2769 	const u_int16_t *w;
2770 
2771 	w = (const u_int16_t *) src;
2772 	sum += w[0];
2773 	if (!IN6_IS_SCOPE_EMBED(src))
2774 		sum += w[1];
2775 	sum += w[2]; sum += w[3]; sum += w[4]; sum += w[5];
2776 	sum += w[6]; sum += w[7];
2777 
2778 	w = (const u_int16_t *) dst;
2779 	sum += w[0];
2780 	if (!IN6_IS_SCOPE_EMBED(dst))
2781 		sum += w[1];
2782 	sum += w[2]; sum += w[3]; sum += w[4]; sum += w[5];
2783 	sum += w[6]; sum += w[7];
2784 
2785 	sum += (u_int16_t)(len >> 16) + (u_int16_t)(len /*& 0xffff*/);
2786 
2787 	sum += (u_int16_t)(nxt >> 16) + (u_int16_t)(nxt /*& 0xffff*/);
2788 
2789 	sum = (u_int16_t)(sum >> 16) + (u_int16_t)(sum /*& 0xffff*/);
2790 
2791 	if (sum > 0xffff)
2792 		sum -= 0xffff;
2793 
2794 	return (sum);
2795 }
2796 
2797 /*
2798  * Process a delayed payload checksum calculation.
2799  */
2800 void
2801 in6_delayed_cksum(struct mbuf *m, u_int8_t nxt)
2802 {
2803 	int nxtp, offset;
2804 	u_int16_t csum;
2805 
2806 	offset = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxtp);
2807 	if (offset <= 0 || nxtp != nxt)
2808 		/* If the desired next protocol isn't found, punt. */
2809 		return;
2810 	csum = (u_int16_t)(in6_cksum(m, 0, offset, m->m_pkthdr.len - offset));
2811 
2812 	switch (nxt) {
2813 	case IPPROTO_TCP:
2814 		offset += offsetof(struct tcphdr, th_sum);
2815 		break;
2816 
2817 	case IPPROTO_UDP:
2818 		offset += offsetof(struct udphdr, uh_sum);
2819 		if (csum == 0)
2820 			csum = 0xffff;
2821 		break;
2822 
2823 	case IPPROTO_ICMPV6:
2824 		offset += offsetof(struct icmp6_hdr, icmp6_cksum);
2825 		break;
2826 	}
2827 
2828 	if ((offset + sizeof(u_int16_t)) > m->m_len)
2829 		m_copyback(m, offset, sizeof(csum), &csum, M_NOWAIT);
2830 	else
2831 		*(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
2832 }
2833 
2834 void
2835 in6_proto_cksum_out(struct mbuf *m, struct ifnet *ifp)
2836 {
2837 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2838 
2839 	/* some hw and in6_delayed_cksum need the pseudo header cksum */
2840 	if (m->m_pkthdr.csum_flags &
2841 	    (M_TCP_CSUM_OUT|M_UDP_CSUM_OUT|M_ICMP_CSUM_OUT)) {
2842 		int nxt, offset;
2843 		u_int16_t csum;
2844 
2845 		offset = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
2846 		csum = in6_cksum_phdr(&ip6->ip6_src, &ip6->ip6_dst,
2847 		    htonl(m->m_pkthdr.len - offset), htonl(nxt));
2848 		if (nxt == IPPROTO_TCP)
2849 			offset += offsetof(struct tcphdr, th_sum);
2850 		else if (nxt == IPPROTO_UDP)
2851 			offset += offsetof(struct udphdr, uh_sum);
2852 		else if (nxt == IPPROTO_ICMPV6)
2853 			offset += offsetof(struct icmp6_hdr, icmp6_cksum);
2854 		if ((offset + sizeof(u_int16_t)) > m->m_len)
2855 			m_copyback(m, offset, sizeof(csum), &csum, M_NOWAIT);
2856 		else
2857 			*(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
2858 	}
2859 
2860 	if (m->m_pkthdr.csum_flags & M_TCP_CSUM_OUT) {
2861 		if (!ifp || !(ifp->if_capabilities & IFCAP_CSUM_TCPv6) ||
2862 		    ip6->ip6_nxt != IPPROTO_TCP ||
2863 		    ifp->if_bridgeport != NULL) {
2864 			tcpstat.tcps_outswcsum++;
2865 			in6_delayed_cksum(m, IPPROTO_TCP);
2866 			m->m_pkthdr.csum_flags &= ~M_TCP_CSUM_OUT; /* Clear */
2867 		}
2868 	} else if (m->m_pkthdr.csum_flags & M_UDP_CSUM_OUT) {
2869 		if (!ifp || !(ifp->if_capabilities & IFCAP_CSUM_UDPv6) ||
2870 		    ip6->ip6_nxt != IPPROTO_UDP ||
2871 		    ifp->if_bridgeport != NULL) {
2872 			udpstat.udps_outswcsum++;
2873 			in6_delayed_cksum(m, IPPROTO_UDP);
2874 			m->m_pkthdr.csum_flags &= ~M_UDP_CSUM_OUT; /* Clear */
2875 		}
2876 	} else if (m->m_pkthdr.csum_flags & M_ICMP_CSUM_OUT) {
2877 		in6_delayed_cksum(m, IPPROTO_ICMPV6);
2878 		m->m_pkthdr.csum_flags &= ~M_ICMP_CSUM_OUT; /* Clear */
2879 	}
2880 }
2881 
2882 #ifdef IPSEC
2883 struct tdb *
2884 ip6_output_ipsec_lookup(struct mbuf *m, int *error, struct inpcb *inp)
2885 {
2886 	struct tdb *tdb;
2887 	struct m_tag *mtag;
2888 	struct tdb_ident *tdbi;
2889 
2890 	/*
2891 	 * Check if there was an outgoing SA bound to the flow
2892 	 * from a transport protocol.
2893 	 */
2894 
2895 	/* Do we have any pending SAs to apply ? */
2896 	tdb = ipsp_spd_lookup(m, AF_INET6, sizeof(struct ip6_hdr),
2897 	    error, IPSP_DIRECTION_OUT, NULL, inp, 0);
2898 
2899 	if (tdb == NULL)
2900 		return NULL;
2901 	/* Loop detection */
2902 	for (mtag = m_tag_first(m); mtag != NULL; mtag = m_tag_next(m, mtag)) {
2903 		if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE)
2904 			continue;
2905 		tdbi = (struct tdb_ident *)(mtag + 1);
2906 		if (tdbi->spi == tdb->tdb_spi &&
2907 		    tdbi->proto == tdb->tdb_sproto &&
2908 		    tdbi->rdomain == tdb->tdb_rdomain &&
2909 		    !memcmp(&tdbi->dst, &tdb->tdb_dst,
2910 		    sizeof(union sockaddr_union))) {
2911 			/* no IPsec needed */
2912 			return NULL;
2913 		}
2914 	}
2915 	return tdb;
2916 }
2917 
2918 int
2919 ip6_output_ipsec_send(struct tdb *tdb, struct mbuf *m, int tunalready, int fwd)
2920 {
2921 #if NPF > 0
2922 	struct ifnet *encif;
2923 #endif
2924 
2925 #if NPF > 0
2926 	if ((encif = enc_getif(tdb->tdb_rdomain, tdb->tdb_tap)) == NULL ||
2927 	    pf_test(AF_INET6, fwd ? PF_FWD : PF_OUT, encif, &m) != PF_PASS) {
2928 		m_freem(m);
2929 		return EHOSTUNREACH;
2930 	}
2931 	if (m == NULL)
2932 		return 0;
2933 	/*
2934 	 * PF_TAG_REROUTE handling or not...
2935 	 * Packet is entering IPsec so the routing is
2936 	 * already overruled by the IPsec policy.
2937 	 * Until now the change was not reconsidered.
2938 	 * What's the behaviour?
2939 	 */
2940 	in6_proto_cksum_out(m, encif);
2941 #endif
2942 	m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
2943 
2944 	/* Callee frees mbuf */
2945 	return ipsp_process_packet(m, tdb, AF_INET6, tunalready);
2946 }
2947 #endif /* IPSEC */
2948