xref: /openbsd/sys/netinet6/ip6_output.c (revision 4bdff4be)
1 /*	$OpenBSD: ip6_output.c,v 1.283 2024/01/18 11:03:16 claudio Exp $	*/
2 /*	$KAME: ip6_output.c,v 1.172 2001/03/25 09:55:56 itojun Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1988, 1990, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
62  */
63 
64 #include "pf.h"
65 
66 #include <sys/param.h>
67 #include <sys/malloc.h>
68 #include <sys/mbuf.h>
69 #include <sys/errno.h>
70 #include <sys/protosw.h>
71 #include <sys/socket.h>
72 #include <sys/socketvar.h>
73 #include <sys/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 
98 #include <crypto/idgen.h>
99 
100 #if NPF > 0
101 #include <net/pfvar.h>
102 #endif
103 
104 #ifdef IPSEC
105 #include <netinet/ip_ipsp.h>
106 #include <netinet/ip_ah.h>
107 #include <netinet/ip_esp.h>
108 
109 #ifdef ENCDEBUG
110 #define DPRINTF(fmt, args...)						\
111 	do {								\
112 		if (encdebug)						\
113 			printf("%s: " fmt "\n", __func__, ## args);	\
114 	} while (0)
115 #else
116 #define DPRINTF(fmt, args...)						\
117 	do { } while (0)
118 #endif
119 #endif /* IPSEC */
120 
121 struct ip6_exthdrs {
122 	struct mbuf *ip6e_ip6;
123 	struct mbuf *ip6e_hbh;
124 	struct mbuf *ip6e_dest1;
125 	struct mbuf *ip6e_rthdr;
126 	struct mbuf *ip6e_dest2;
127 };
128 
129 int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **, int, int);
130 int ip6_getpcbopt(struct ip6_pktopts *, int, struct mbuf *);
131 int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *, int, int, int);
132 int ip6_setmoptions(int, struct ip6_moptions **, struct mbuf *, unsigned int);
133 int ip6_getmoptions(int, struct ip6_moptions *, struct mbuf *);
134 int ip6_copyexthdr(struct mbuf **, caddr_t, int);
135 int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
136 	struct ip6_frag **);
137 int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
138 int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
139 int ip6_getpmtu(struct rtentry *, struct ifnet *, u_long *);
140 int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *);
141 static __inline u_int16_t __attribute__((__unused__))
142     in6_cksum_phdr(const struct in6_addr *, const struct in6_addr *,
143     u_int32_t, u_int32_t);
144 void in6_delayed_cksum(struct mbuf *, u_int8_t);
145 
146 int ip6_output_ipsec_pmtu_update(struct tdb *, struct route_in6 *,
147     struct in6_addr *, int, int, int);
148 
149 /* Context for non-repeating IDs */
150 struct idgen32_ctx ip6_id_ctx;
151 
152 /*
153  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
154  * header (with pri, len, nxt, hlim, src, dst).
155  * This function may modify ver and hlim only.
156  * The mbuf chain containing the packet will be freed.
157  * The mbuf opt, if present, will not be freed.
158  *
159  * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int.
160  * We use u_long to hold largest one, * which is rt_mtu.
161  */
162 int
163 ip6_output(struct mbuf *m, struct ip6_pktopts *opt, struct route_in6 *ro,
164     int flags, struct ip6_moptions *im6o, const u_char seclevel[])
165 {
166 	struct ip6_hdr *ip6;
167 	struct ifnet *ifp = NULL;
168 	struct mbuf_list ml;
169 	int hlen, tlen;
170 	struct route_in6 ip6route;
171 	struct rtentry *rt = NULL;
172 	struct sockaddr_in6 *dst, dstsock;
173 	int error = 0;
174 	u_long mtu;
175 	int dontfrag;
176 	u_int16_t src_scope, dst_scope;
177 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
178 	struct ip6_exthdrs exthdrs;
179 	struct in6_addr finaldst;
180 	struct route_in6 *ro_pmtu = NULL;
181 	int hdrsplit = 0;
182 	u_int8_t sproto = 0;
183 	u_char nextproto;
184 #ifdef IPSEC
185 	struct tdb *tdb = NULL;
186 #endif /* IPSEC */
187 
188 	ip6 = mtod(m, struct ip6_hdr *);
189 	finaldst = ip6->ip6_dst;
190 
191 #define MAKE_EXTHDR(hp, mp)						\
192     do {								\
193 	if (hp) {							\
194 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
195 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
196 		    ((eh)->ip6e_len + 1) << 3);				\
197 		if (error)						\
198 			goto freehdrs;					\
199 	}								\
200     } while (0)
201 
202 	bzero(&exthdrs, sizeof(exthdrs));
203 
204 	if (opt) {
205 		/* Hop-by-Hop options header */
206 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
207 		/* Destination options header(1st part) */
208 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
209 		/* Routing header */
210 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
211 		/* Destination options header(2nd part) */
212 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
213 	}
214 
215 #ifdef IPSEC
216 	if (ipsec_in_use || seclevel != NULL) {
217 		error = ip6_output_ipsec_lookup(m, seclevel, &tdb);
218 		if (error) {
219 			/*
220 			 * -EINVAL is used to indicate that the packet should
221 			 * be silently dropped, typically because we've asked
222 			 * key management for an SA.
223 			 */
224 			if (error == -EINVAL) /* Should silently drop packet */
225 				error = 0;
226 
227 			goto freehdrs;
228 		}
229 	}
230 #endif /* IPSEC */
231 
232 	/*
233 	 * Calculate the total length of the extension header chain.
234 	 * Keep the length of the unfragmentable part for fragmentation.
235 	 */
236 	optlen = 0;
237 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
238 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
239 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
240 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
241 	/* NOTE: we don't add AH/ESP length here. do that later. */
242 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
243 
244 	/*
245 	 * If we need IPsec, or there is at least one extension header,
246 	 * separate IP6 header from the payload.
247 	 */
248 	if ((sproto || optlen) && !hdrsplit) {
249 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
250 			m = NULL;
251 			goto freehdrs;
252 		}
253 		m = exthdrs.ip6e_ip6;
254 		hdrsplit++;
255 	}
256 
257 	/* adjust pointer */
258 	ip6 = mtod(m, struct ip6_hdr *);
259 
260 	/* adjust mbuf packet header length */
261 	m->m_pkthdr.len += optlen;
262 	plen = m->m_pkthdr.len - sizeof(*ip6);
263 
264 	/* If this is a jumbo payload, insert a jumbo payload option. */
265 	if (plen > IPV6_MAXPACKET) {
266 		if (!hdrsplit) {
267 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
268 				m = NULL;
269 				goto freehdrs;
270 			}
271 			m = exthdrs.ip6e_ip6;
272 			hdrsplit++;
273 		}
274 		/* adjust pointer */
275 		ip6 = mtod(m, struct ip6_hdr *);
276 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
277 			goto freehdrs;
278 		ip6->ip6_plen = 0;
279 	} else
280 		ip6->ip6_plen = htons(plen);
281 
282 	/*
283 	 * Concatenate headers and fill in next header fields.
284 	 * Here we have, on "m"
285 	 *	IPv6 payload
286 	 * and we insert headers accordingly.  Finally, we should be getting:
287 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
288 	 *
289 	 * during the header composing process, "m" points to IPv6 header.
290 	 * "mprev" points to an extension header prior to esp.
291 	 */
292 	{
293 		u_char *nexthdrp = &ip6->ip6_nxt;
294 		struct mbuf *mprev = m;
295 
296 		/*
297 		 * we treat dest2 specially.  this makes IPsec processing
298 		 * much easier.  the goal here is to make mprev point the
299 		 * mbuf prior to dest2.
300 		 *
301 		 * result: IPv6 dest2 payload
302 		 * m and mprev will point to IPv6 header.
303 		 */
304 		if (exthdrs.ip6e_dest2) {
305 			if (!hdrsplit)
306 				panic("%s: assumption failed: hdr not split",
307 				    __func__);
308 			exthdrs.ip6e_dest2->m_next = m->m_next;
309 			m->m_next = exthdrs.ip6e_dest2;
310 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
311 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
312 		}
313 
314 #define MAKE_CHAIN(m, mp, p, i)\
315     do {\
316 	if (m) {\
317 		if (!hdrsplit) \
318 			panic("assumption failed: hdr not split"); \
319 		*mtod((m), u_char *) = *(p);\
320 		*(p) = (i);\
321 		p = mtod((m), u_char *);\
322 		(m)->m_next = (mp)->m_next;\
323 		(mp)->m_next = (m);\
324 		(mp) = (m);\
325 	}\
326     } while (0)
327 		/*
328 		 * result: IPv6 hbh dest1 rthdr dest2 payload
329 		 * m will point to IPv6 header.  mprev will point to the
330 		 * extension header prior to dest2 (rthdr in the above case).
331 		 */
332 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
333 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
334 		    IPPROTO_DSTOPTS);
335 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
336 		    IPPROTO_ROUTING);
337 	}
338 
339 	/*
340 	 * If there is a routing header, replace the destination address field
341 	 * with the first hop of the routing header.
342 	 */
343 	if (exthdrs.ip6e_rthdr) {
344 		struct ip6_rthdr *rh;
345 		struct ip6_rthdr0 *rh0;
346 		struct in6_addr *addr;
347 
348 		rh = (struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
349 		    struct ip6_rthdr *));
350 		switch (rh->ip6r_type) {
351 		case IPV6_RTHDR_TYPE_0:
352 			rh0 = (struct ip6_rthdr0 *)rh;
353 			addr = (struct in6_addr *)(rh0 + 1);
354 			ip6->ip6_dst = addr[0];
355 			bcopy(&addr[1], &addr[0],
356 			    sizeof(struct in6_addr) * (rh0->ip6r0_segleft - 1));
357 			addr[rh0->ip6r0_segleft - 1] = finaldst;
358 			break;
359 		default:	/* is it possible? */
360 			error = EINVAL;
361 			goto bad;
362 		}
363 	}
364 
365 	/* Source address validation */
366 	if (!(flags & IPV6_UNSPECSRC) &&
367 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
368 		/*
369 		 * XXX: we can probably assume validation in the caller, but
370 		 * we explicitly check the address here for safety.
371 		 */
372 		error = EOPNOTSUPP;
373 		ip6stat_inc(ip6s_badscope);
374 		goto bad;
375 	}
376 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
377 		error = EOPNOTSUPP;
378 		ip6stat_inc(ip6s_badscope);
379 		goto bad;
380 	}
381 
382 	ip6stat_inc(ip6s_localout);
383 
384 	/*
385 	 * Route packet.
386 	 */
387 #if NPF > 0
388 reroute:
389 #endif
390 
391 	/* initialize cached route */
392 	if (ro == NULL) {
393 		ro = &ip6route;
394 		bzero((caddr_t)ro, sizeof(*ro));
395 	}
396 	ro_pmtu = ro;
397 	if (opt && opt->ip6po_rthdr)
398 		ro = &opt->ip6po_route;
399 	dst = &ro->ro_dst;
400 
401 	/*
402 	 * if specified, try to fill in the traffic class field.
403 	 * do not override if a non-zero value is already set.
404 	 * we check the diffserv field and the ecn field separately.
405 	 */
406 	if (opt && opt->ip6po_tclass >= 0) {
407 		int mask = 0;
408 
409 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
410 			mask |= 0xfc;
411 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
412 			mask |= 0x03;
413 		if (mask != 0)
414 			ip6->ip6_flow |=
415 			    htonl((opt->ip6po_tclass & mask) << 20);
416 	}
417 
418 	/* fill in or override the hop limit field, if necessary. */
419 	if (opt && opt->ip6po_hlim != -1)
420 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
421 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
422 		if (im6o != NULL)
423 			ip6->ip6_hlim = im6o->im6o_hlim;
424 		else
425 			ip6->ip6_hlim = ip6_defmcasthlim;
426 	}
427 
428 #ifdef IPSEC
429 	if (tdb != NULL) {
430 		/*
431 		 * XXX what should we do if ip6_hlim == 0 and the
432 		 * packet gets tunneled?
433 		 */
434 		/*
435 		 * if we are source-routing, do not attempt to tunnel the
436 		 * packet just because ip6_dst is different from what tdb has.
437 		 * XXX
438 		 */
439 		error = ip6_output_ipsec_send(tdb, m, ro,
440 		    exthdrs.ip6e_rthdr ? 1 : 0, 0);
441 		goto done;
442 	}
443 #endif /* IPSEC */
444 
445 	bzero(&dstsock, sizeof(dstsock));
446 	dstsock.sin6_family = AF_INET6;
447 	dstsock.sin6_addr = ip6->ip6_dst;
448 	dstsock.sin6_len = sizeof(dstsock);
449 	ro->ro_tableid = m->m_pkthdr.ph_rtableid;
450 
451 	if (IN6_IS_ADDR_MULTICAST(&dstsock.sin6_addr)) {
452 		struct in6_pktinfo *pi = NULL;
453 
454 		/*
455 		 * If the caller specify the outgoing interface
456 		 * explicitly, use it.
457 		 */
458 		if (opt != NULL && (pi = opt->ip6po_pktinfo) != NULL)
459 			ifp = if_get(pi->ipi6_ifindex);
460 
461 		if (ifp == NULL && im6o != NULL)
462 			ifp = if_get(im6o->im6o_ifidx);
463 	}
464 
465 	if (ifp == NULL) {
466 		rt = in6_selectroute(&dstsock, opt, ro, ro->ro_tableid);
467 		if (rt == NULL) {
468 			ip6stat_inc(ip6s_noroute);
469 			error = EHOSTUNREACH;
470 			goto bad;
471 		}
472 		if (ISSET(rt->rt_flags, RTF_LOCAL))
473 			ifp = if_get(rtable_loindex(m->m_pkthdr.ph_rtableid));
474 		else
475 			ifp = if_get(rt->rt_ifidx);
476 		/*
477 		 * We aren't using rtisvalid() here because the UP/DOWN state
478 		 * machine is broken with some Ethernet drivers like em(4).
479 		 * As a result we might try to use an invalid cached route
480 		 * entry while an interface is being detached.
481 		 */
482 		if (ifp == NULL) {
483 			ip6stat_inc(ip6s_noroute);
484 			error = EHOSTUNREACH;
485 			goto bad;
486 		}
487 	} else {
488 		*dst = dstsock;
489 	}
490 
491 	if (rt && (rt->rt_flags & RTF_GATEWAY) &&
492 	    !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
493 		dst = satosin6(rt->rt_gateway);
494 
495 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
496 		/* Unicast */
497 
498 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
499 	} else {
500 		/* Multicast */
501 
502 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
503 
504 		/*
505 		 * Confirm that the outgoing interface supports multicast.
506 		 */
507 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
508 			ip6stat_inc(ip6s_noroute);
509 			error = ENETUNREACH;
510 			goto bad;
511 		}
512 
513 		if ((im6o == NULL || im6o->im6o_loop) &&
514 		    in6_hasmulti(&ip6->ip6_dst, ifp)) {
515 			/*
516 			 * If we belong to the destination multicast group
517 			 * on the outgoing interface, and the caller did not
518 			 * forbid loopback, loop back a copy.
519 			 * Can't defer TCP/UDP checksumming, do the
520 			 * computation now.
521 			 */
522 			in6_proto_cksum_out(m, NULL);
523 			ip6_mloopback(ifp, m, dst);
524 		}
525 #ifdef MROUTING
526 		else {
527 			/*
528 			 * If we are acting as a multicast router, perform
529 			 * multicast forwarding as if the packet had just
530 			 * arrived on the interface to which we are about
531 			 * to send.  The multicast forwarding function
532 			 * recursively calls this function, using the
533 			 * IPV6_FORWARDING flag to prevent infinite recursion.
534 			 *
535 			 * Multicasts that are looped back by ip6_mloopback(),
536 			 * above, will be forwarded by the ip6_input() routine,
537 			 * if necessary.
538 			 */
539 			if (ip6_mforwarding && ip6_mrouter[ifp->if_rdomain] &&
540 			    (flags & IPV6_FORWARDING) == 0) {
541 				if (ip6_mforward(ip6, ifp, m) != 0) {
542 					m_freem(m);
543 					goto done;
544 				}
545 			}
546 		}
547 #endif
548 		/*
549 		 * Multicasts with a hoplimit of zero may be looped back,
550 		 * above, but must not be transmitted on a network.
551 		 * Also, multicasts addressed to the loopback interface
552 		 * are not sent -- the above call to ip6_mloopback() will
553 		 * loop back a copy if this host actually belongs to the
554 		 * destination group on the loopback interface.
555 		 */
556 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
557 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
558 			m_freem(m);
559 			goto done;
560 		}
561 	}
562 
563 	/*
564 	 * If this packet is going through a loopback interface we won't
565 	 * be able to restore its scope ID using the interface index.
566 	 */
567 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src)) {
568 		if (ifp->if_flags & IFF_LOOPBACK)
569 			src_scope = ip6->ip6_src.s6_addr16[1];
570 		ip6->ip6_src.s6_addr16[1] = 0;
571 	}
572 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) {
573 		if (ifp->if_flags & IFF_LOOPBACK)
574 			dst_scope = ip6->ip6_dst.s6_addr16[1];
575 		ip6->ip6_dst.s6_addr16[1] = 0;
576 	}
577 
578 	/* Determine path MTU. */
579 	if ((error = ip6_getpmtu(ro_pmtu->ro_rt, ifp, &mtu)) != 0)
580 		goto bad;
581 
582 	/*
583 	 * The caller of this function may specify to use the minimum MTU
584 	 * in some cases.
585 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
586 	 * setting.  The logic is a bit complicated; by default, unicast
587 	 * packets will follow path MTU while multicast packets will be sent at
588 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
589 	 * including unicast ones will be sent at the minimum MTU.  Multicast
590 	 * packets will always be sent at the minimum MTU unless
591 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
592 	 * See RFC 3542 for more details.
593 	 */
594 	if (mtu > IPV6_MMTU) {
595 		if ((flags & IPV6_MINMTU))
596 			mtu = IPV6_MMTU;
597 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
598 			mtu = IPV6_MMTU;
599 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && (opt == NULL ||
600 		    opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
601 			mtu = IPV6_MMTU;
602 		}
603 	}
604 
605 	/*
606 	 * If the outgoing packet contains a hop-by-hop options header,
607 	 * it must be examined and processed even by the source node.
608 	 * (RFC 2460, section 4.)
609 	 */
610 	if (exthdrs.ip6e_hbh) {
611 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
612 		u_int32_t rtalert; /* returned value is ignored */
613 		u_int32_t plen = 0; /* no more than 1 jumbo payload option! */
614 
615 		m->m_pkthdr.ph_ifidx = ifp->if_index;
616 		if (ip6_process_hopopts(&m, (u_int8_t *)(hbh + 1),
617 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
618 		    &rtalert, &plen) < 0) {
619 			/* m was already freed at this point */
620 			error = EINVAL;/* better error? */
621 			goto done;
622 		}
623 		m->m_pkthdr.ph_ifidx = 0;
624 	}
625 
626 #if NPF > 0
627 	if (pf_test(AF_INET6, PF_OUT, ifp, &m) != PF_PASS) {
628 		error = EACCES;
629 		m_freem(m);
630 		goto done;
631 	}
632 	if (m == NULL)
633 		goto done;
634 	ip6 = mtod(m, struct ip6_hdr *);
635 	if ((m->m_pkthdr.pf.flags & (PF_TAG_REROUTE | PF_TAG_GENERATED)) ==
636 	    (PF_TAG_REROUTE | PF_TAG_GENERATED)) {
637 		/* already rerun the route lookup, go on */
638 		m->m_pkthdr.pf.flags &= ~(PF_TAG_GENERATED | PF_TAG_REROUTE);
639 	} else if (m->m_pkthdr.pf.flags & PF_TAG_REROUTE) {
640 		/* tag as generated to skip over pf_test on rerun */
641 		m->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
642 		finaldst = ip6->ip6_dst;
643 		ro = NULL;
644 		if_put(ifp); /* drop reference since destination changed */
645 		ifp = NULL;
646 		goto reroute;
647 	}
648 #endif
649 
650 	/*
651 	 * If the packet is not going on the wire it can be destined
652 	 * to any local address.  In this case do not clear its scopes
653 	 * to let ip6_input() find a matching local route.
654 	 */
655 	if (ifp->if_flags & IFF_LOOPBACK) {
656 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
657 			ip6->ip6_src.s6_addr16[1] = src_scope;
658 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
659 			ip6->ip6_dst.s6_addr16[1] = dst_scope;
660 	}
661 
662 	/*
663 	 * Send the packet to the outgoing interface.
664 	 * If necessary, do IPv6 fragmentation before sending.
665 	 *
666 	 * the logic here is rather complex:
667 	 * 1: normal case (dontfrag == 0)
668 	 * 1-a: send as is if tlen <= path mtu
669 	 * 1-b: fragment if tlen > path mtu
670 	 *
671 	 * 2: if user asks us not to fragment (dontfrag == 1)
672 	 * 2-a: send as is if tlen <= interface mtu
673 	 * 2-b: error if tlen > interface mtu
674 	 */
675 	tlen = ISSET(m->m_pkthdr.csum_flags, M_TCP_TSO) ?
676 	    m->m_pkthdr.ph_mss : m->m_pkthdr.len;
677 
678 	if (ISSET(m->m_pkthdr.csum_flags, M_IPV6_DF_OUT)) {
679 		CLR(m->m_pkthdr.csum_flags, M_IPV6_DF_OUT);
680 		dontfrag = 1;
681 	} else if (opt && ISSET(opt->ip6po_flags, IP6PO_DONTFRAG))
682 		dontfrag = 1;
683 	else
684 		dontfrag = 0;
685 
686 	if (dontfrag && tlen > ifp->if_mtu) {		/* case 2-b */
687 #ifdef IPSEC
688 		if (ip_mtudisc)
689 			ipsec_adjust_mtu(m, mtu);
690 #endif
691 		error = EMSGSIZE;
692 		goto bad;
693 	}
694 
695 	/*
696 	 * transmit packet without fragmentation
697 	 */
698 	if (dontfrag || tlen <= mtu) {			/* case 1-a and 2-a */
699 		error = if_output_tso(ifp, &m, sin6tosa(dst), ro->ro_rt,
700 		    ifp->if_mtu);
701 		if (error || m == NULL)
702 			goto done;
703 		goto bad;				/* should not happen */
704 	}
705 
706 	/*
707 	 * try to fragment the packet.  case 1-b
708 	 */
709 	if (mtu < IPV6_MMTU) {
710 		/* path MTU cannot be less than IPV6_MMTU */
711 		error = EMSGSIZE;
712 		goto bad;
713 	} else if (ip6->ip6_plen == 0) {
714 		/* jumbo payload cannot be fragmented */
715 		error = EMSGSIZE;
716 		goto bad;
717 	}
718 
719 	/*
720 	 * Too large for the destination or interface;
721 	 * fragment if possible.
722 	 * Must be able to put at least 8 bytes per fragment.
723 	 */
724 	hlen = unfragpartlen;
725 	if (mtu > IPV6_MAXPACKET)
726 		mtu = IPV6_MAXPACKET;
727 
728 	/*
729 	 * If we are doing fragmentation, we can't defer TCP/UDP
730 	 * checksumming; compute the checksum and clear the flag.
731 	 */
732         in6_proto_cksum_out(m, NULL);
733 
734 	/*
735 	 * Change the next header field of the last header in the
736 	 * unfragmentable part.
737 	 */
738 	if (exthdrs.ip6e_rthdr) {
739 		nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
740 		*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
741 	} else if (exthdrs.ip6e_dest1) {
742 		nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
743 		*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
744 	} else if (exthdrs.ip6e_hbh) {
745 		nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
746 		*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
747 	} else {
748 		nextproto = ip6->ip6_nxt;
749 		ip6->ip6_nxt = IPPROTO_FRAGMENT;
750 	}
751 
752 	if ((error = ip6_fragment(m, &ml, hlen, nextproto, mtu)) ||
753 	    (error = if_output_ml(ifp, &ml, sin6tosa(dst), ro->ro_rt)))
754 		goto done;
755 	ip6stat_inc(ip6s_fragmented);
756 
757 done:
758 	if (ro == &ip6route && ro->ro_rt) {
759 		rtfree(ro->ro_rt);
760 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
761 		rtfree(ro_pmtu->ro_rt);
762 	}
763 	if_put(ifp);
764 #ifdef IPSEC
765 	tdb_unref(tdb);
766 #endif /* IPSEC */
767 	return (error);
768 
769 freehdrs:
770 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
771 	m_freem(exthdrs.ip6e_dest1);
772 	m_freem(exthdrs.ip6e_rthdr);
773 	m_freem(exthdrs.ip6e_dest2);
774 	/* FALLTHROUGH */
775 bad:
776 	m_freem(m);
777 	goto done;
778 }
779 
780 int
781 ip6_fragment(struct mbuf *m0, struct mbuf_list *ml, int hlen, u_char nextproto,
782     u_long mtu)
783 {
784 	struct ip6_hdr *ip6;
785 	u_int32_t id;
786 	int tlen, len, off;
787 	int error;
788 
789 	ml_init(ml);
790 
791 	ip6 = mtod(m0, struct ip6_hdr *);
792 	tlen = m0->m_pkthdr.len;
793 	len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
794 	if (len < 8) {
795 		error = EMSGSIZE;
796 		goto bad;
797 	}
798 	id = htonl(ip6_randomid());
799 
800 	/*
801 	 * Loop through length of payload,
802 	 * make new header and copy data of each part and link onto chain.
803 	 */
804 	for (off = hlen; off < tlen; off += len) {
805 		struct mbuf *m;
806 		struct mbuf *mlast;
807 		struct ip6_hdr *mhip6;
808 		struct ip6_frag *ip6f;
809 
810 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
811 		if (m == NULL) {
812 			error = ENOBUFS;
813 			goto bad;
814 		}
815 		ml_enqueue(ml, m);
816 		if ((error = m_dup_pkthdr(m, m0, M_DONTWAIT)) != 0)
817 			goto bad;
818 		m->m_data += max_linkhdr;
819 		mhip6 = mtod(m, struct ip6_hdr *);
820 		*mhip6 = *ip6;
821 		m->m_len = sizeof(struct ip6_hdr);
822 
823 		if ((error = ip6_insertfraghdr(m0, m, hlen, &ip6f)) != 0)
824 			goto bad;
825 		ip6f->ip6f_offlg = htons((off - hlen) & ~7);
826 		if (off + len >= tlen)
827 			len = tlen - off;
828 		else
829 			ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
830 
831 		m->m_pkthdr.len = hlen + sizeof(struct ip6_frag) + len;
832 		mhip6->ip6_plen = htons(m->m_pkthdr.len -
833 		    sizeof(struct ip6_hdr));
834 		for (mlast = m; mlast->m_next; mlast = mlast->m_next)
835 			;
836 		mlast->m_next = m_copym(m0, off, len, M_DONTWAIT);
837 		if (mlast->m_next == NULL) {
838 			error = ENOBUFS;
839 			goto bad;
840 		}
841 
842 		ip6f->ip6f_reserved = 0;
843 		ip6f->ip6f_ident = id;
844 		ip6f->ip6f_nxt = nextproto;
845 	}
846 
847 	ip6stat_add(ip6s_ofragments, ml_len(ml));
848 	m_freem(m0);
849 	return (0);
850 
851 bad:
852 	ip6stat_inc(ip6s_odropped);
853 	ml_purge(ml);
854 	m_freem(m0);
855 	return (error);
856 }
857 
858 int
859 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
860 {
861 	struct mbuf *m;
862 
863 	if (hlen > MCLBYTES)
864 		return (ENOBUFS); /* XXX */
865 
866 	MGET(m, M_DONTWAIT, MT_DATA);
867 	if (!m)
868 		return (ENOBUFS);
869 
870 	if (hlen > MLEN) {
871 		MCLGET(m, M_DONTWAIT);
872 		if ((m->m_flags & M_EXT) == 0) {
873 			m_free(m);
874 			return (ENOBUFS);
875 		}
876 	}
877 	m->m_len = hlen;
878 	if (hdr)
879 		memcpy(mtod(m, caddr_t), hdr, hlen);
880 
881 	*mp = m;
882 	return (0);
883 }
884 
885 /*
886  * Insert jumbo payload option.
887  */
888 int
889 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
890 {
891 	struct mbuf *mopt;
892 	u_int8_t *optbuf;
893 	u_int32_t v;
894 
895 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
896 
897 	/*
898 	 * If there is no hop-by-hop options header, allocate new one.
899 	 * If there is one but it doesn't have enough space to store the
900 	 * jumbo payload option, allocate a cluster to store the whole options.
901 	 * Otherwise, use it to store the options.
902 	 */
903 	if (exthdrs->ip6e_hbh == 0) {
904 		MGET(mopt, M_DONTWAIT, MT_DATA);
905 		if (mopt == NULL)
906 			return (ENOBUFS);
907 		mopt->m_len = JUMBOOPTLEN;
908 		optbuf = mtod(mopt, u_int8_t *);
909 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
910 		exthdrs->ip6e_hbh = mopt;
911 	} else {
912 		struct ip6_hbh *hbh;
913 
914 		mopt = exthdrs->ip6e_hbh;
915 		if (m_trailingspace(mopt) < JUMBOOPTLEN) {
916 			/*
917 			 * XXX assumption:
918 			 * - exthdrs->ip6e_hbh is not referenced from places
919 			 *   other than exthdrs.
920 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
921 			 */
922 			int oldoptlen = mopt->m_len;
923 			struct mbuf *n;
924 
925 			/*
926 			 * XXX: give up if the whole (new) hbh header does
927 			 * not fit even in an mbuf cluster.
928 			 */
929 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
930 				return (ENOBUFS);
931 
932 			/*
933 			 * As a consequence, we must always prepare a cluster
934 			 * at this point.
935 			 */
936 			MGET(n, M_DONTWAIT, MT_DATA);
937 			if (n) {
938 				MCLGET(n, M_DONTWAIT);
939 				if ((n->m_flags & M_EXT) == 0) {
940 					m_freem(n);
941 					n = NULL;
942 				}
943 			}
944 			if (!n)
945 				return (ENOBUFS);
946 			n->m_len = oldoptlen + JUMBOOPTLEN;
947 			memcpy(mtod(n, caddr_t), mtod(mopt, caddr_t),
948 			      oldoptlen);
949 			optbuf = mtod(n, u_int8_t *) + oldoptlen;
950 			m_freem(mopt);
951 			mopt = exthdrs->ip6e_hbh = n;
952 		} else {
953 			optbuf = mtod(mopt, u_int8_t *) + mopt->m_len;
954 			mopt->m_len += JUMBOOPTLEN;
955 		}
956 		optbuf[0] = IP6OPT_PADN;
957 		optbuf[1] = 0;
958 
959 		/*
960 		 * Adjust the header length according to the pad and
961 		 * the jumbo payload option.
962 		 */
963 		hbh = mtod(mopt, struct ip6_hbh *);
964 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
965 	}
966 
967 	/* fill in the option. */
968 	optbuf[2] = IP6OPT_JUMBO;
969 	optbuf[3] = 4;
970 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
971 	memcpy(&optbuf[4], &v, sizeof(u_int32_t));
972 
973 	/* finally, adjust the packet header length */
974 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
975 
976 	return (0);
977 #undef JUMBOOPTLEN
978 }
979 
980 /*
981  * Insert fragment header and copy unfragmentable header portions.
982  */
983 int
984 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
985     struct ip6_frag **frghdrp)
986 {
987 	struct mbuf *n, *mlast;
988 
989 	if (hlen > sizeof(struct ip6_hdr)) {
990 		n = m_copym(m0, sizeof(struct ip6_hdr),
991 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
992 		if (n == NULL)
993 			return (ENOBUFS);
994 		m->m_next = n;
995 	} else
996 		n = m;
997 
998 	/* Search for the last mbuf of unfragmentable part. */
999 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1000 		;
1001 
1002 	if ((mlast->m_flags & M_EXT) == 0 &&
1003 	    m_trailingspace(mlast) >= sizeof(struct ip6_frag)) {
1004 		/* use the trailing space of the last mbuf for fragment hdr */
1005 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1006 		    mlast->m_len);
1007 		mlast->m_len += sizeof(struct ip6_frag);
1008 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1009 	} else {
1010 		/* allocate a new mbuf for the fragment header */
1011 		struct mbuf *mfrg;
1012 
1013 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1014 		if (mfrg == NULL)
1015 			return (ENOBUFS);
1016 		mfrg->m_len = sizeof(struct ip6_frag);
1017 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1018 		mlast->m_next = mfrg;
1019 	}
1020 
1021 	return (0);
1022 }
1023 
1024 int
1025 ip6_getpmtu(struct rtentry *rt, struct ifnet *ifp, u_long *mtup)
1026 {
1027 	u_int32_t mtu = 0;
1028 	int error = 0;
1029 
1030 	if (rt != NULL) {
1031 		mtu = rt->rt_mtu;
1032 		if (mtu == 0)
1033 			mtu = ifp->if_mtu;
1034 		else if (mtu < IPV6_MMTU) {
1035 			/* RFC8021 IPv6 Atomic Fragments Considered Harmful */
1036 			mtu = IPV6_MMTU;
1037 		} else if (mtu > ifp->if_mtu) {
1038 			/*
1039 			 * The MTU on the route is larger than the MTU on
1040 			 * the interface!  This shouldn't happen, unless the
1041 			 * MTU of the interface has been changed after the
1042 			 * interface was brought up.  Change the MTU in the
1043 			 * route to match the interface MTU (as long as the
1044 			 * field isn't locked).
1045 			 */
1046 			mtu = ifp->if_mtu;
1047 			if (!(rt->rt_locks & RTV_MTU))
1048 				rt->rt_mtu = mtu;
1049 		}
1050 	} else {
1051 		mtu = ifp->if_mtu;
1052 	}
1053 
1054 	*mtup = mtu;
1055 	return (error);
1056 }
1057 
1058 /*
1059  * IP6 socket option processing.
1060  */
1061 int
1062 ip6_ctloutput(int op, struct socket *so, int level, int optname,
1063     struct mbuf *m)
1064 {
1065 	int privileged, optdatalen, uproto;
1066 	void *optdata;
1067 	struct inpcb *inp = sotoinpcb(so);
1068 	int error, optval;
1069 	struct proc *p = curproc; /* For IPsec and rdomain */
1070 	u_int rtableid, rtid = 0;
1071 
1072 	error = optval = 0;
1073 
1074 	privileged = (inp->inp_socket->so_state & SS_PRIV);
1075 	uproto = (int)so->so_proto->pr_protocol;
1076 
1077 	if (level != IPPROTO_IPV6)
1078 		return (EINVAL);
1079 
1080 	rtableid = p->p_p->ps_rtableid;
1081 
1082 	switch (op) {
1083 	case PRCO_SETOPT:
1084 		switch (optname) {
1085 		/*
1086 		 * Use of some Hop-by-Hop options or some
1087 		 * Destination options, might require special
1088 		 * privilege.  That is, normal applications
1089 		 * (without special privilege) might be forbidden
1090 		 * from setting certain options in outgoing packets,
1091 		 * and might never see certain options in received
1092 		 * packets. [RFC 2292 Section 6]
1093 		 * KAME specific note:
1094 		 *  KAME prevents non-privileged users from sending or
1095 		 *  receiving ANY hbh/dst options in order to avoid
1096 		 *  overhead of parsing options in the kernel.
1097 		 */
1098 		case IPV6_RECVHOPOPTS:
1099 		case IPV6_RECVDSTOPTS:
1100 			if (!privileged) {
1101 				error = EPERM;
1102 				break;
1103 			}
1104 			/* FALLTHROUGH */
1105 		case IPV6_UNICAST_HOPS:
1106 		case IPV6_MINHOPCOUNT:
1107 		case IPV6_HOPLIMIT:
1108 
1109 		case IPV6_RECVPKTINFO:
1110 		case IPV6_RECVHOPLIMIT:
1111 		case IPV6_RECVRTHDR:
1112 		case IPV6_RECVPATHMTU:
1113 		case IPV6_RECVTCLASS:
1114 		case IPV6_V6ONLY:
1115 		case IPV6_AUTOFLOWLABEL:
1116 		case IPV6_RECVDSTPORT:
1117 			if (m == NULL || m->m_len != sizeof(int)) {
1118 				error = EINVAL;
1119 				break;
1120 			}
1121 			optval = *mtod(m, int *);
1122 			switch (optname) {
1123 
1124 			case IPV6_UNICAST_HOPS:
1125 				if (optval < -1 || optval >= 256)
1126 					error = EINVAL;
1127 				else {
1128 					/* -1 = kernel default */
1129 					inp->inp_hops = optval;
1130 				}
1131 				break;
1132 
1133 			case IPV6_MINHOPCOUNT:
1134 				if (optval < 0 || optval > 255)
1135 					error = EINVAL;
1136 				else
1137 					inp->inp_ip6_minhlim = optval;
1138 				break;
1139 
1140 #define OPTSET(bit) \
1141 do { \
1142 	if (optval) \
1143 		inp->inp_flags |= (bit); \
1144 	else \
1145 		inp->inp_flags &= ~(bit); \
1146 } while (/*CONSTCOND*/ 0)
1147 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1148 
1149 			case IPV6_RECVPKTINFO:
1150 				OPTSET(IN6P_PKTINFO);
1151 				break;
1152 
1153 			case IPV6_HOPLIMIT:
1154 			{
1155 				struct ip6_pktopts **optp;
1156 
1157 				optp = &inp->inp_outputopts6;
1158 				error = ip6_pcbopt(IPV6_HOPLIMIT,
1159 				    (u_char *)&optval, sizeof(optval), optp,
1160 				    privileged, uproto);
1161 				break;
1162 			}
1163 
1164 			case IPV6_RECVHOPLIMIT:
1165 				OPTSET(IN6P_HOPLIMIT);
1166 				break;
1167 
1168 			case IPV6_RECVHOPOPTS:
1169 				OPTSET(IN6P_HOPOPTS);
1170 				break;
1171 
1172 			case IPV6_RECVDSTOPTS:
1173 				OPTSET(IN6P_DSTOPTS);
1174 				break;
1175 
1176 			case IPV6_RECVRTHDR:
1177 				OPTSET(IN6P_RTHDR);
1178 				break;
1179 
1180 			case IPV6_RECVPATHMTU:
1181 				/*
1182 				 * We ignore this option for TCP
1183 				 * sockets.
1184 				 * (RFC3542 leaves this case
1185 				 * unspecified.)
1186 				 */
1187 				if (uproto != IPPROTO_TCP)
1188 					OPTSET(IN6P_MTU);
1189 				break;
1190 
1191 			case IPV6_V6ONLY:
1192 				/*
1193 				 * make setsockopt(IPV6_V6ONLY)
1194 				 * available only prior to bind(2).
1195 				 * see ipng mailing list, Jun 22 2001.
1196 				 */
1197 				if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(
1198 				    &inp->inp_laddr6)) {
1199 					error = EINVAL;
1200 					break;
1201 				}
1202 				/* No support for IPv4-mapped addresses. */
1203 				if (!optval)
1204 					error = EINVAL;
1205 				else
1206 					error = 0;
1207 				break;
1208 			case IPV6_RECVTCLASS:
1209 				OPTSET(IN6P_TCLASS);
1210 				break;
1211 			case IPV6_AUTOFLOWLABEL:
1212 				OPTSET(IN6P_AUTOFLOWLABEL);
1213 				break;
1214 
1215 			case IPV6_RECVDSTPORT:
1216 				OPTSET(IN6P_RECVDSTPORT);
1217 				break;
1218 			}
1219 			break;
1220 
1221 		case IPV6_TCLASS:
1222 		case IPV6_DONTFRAG:
1223 		case IPV6_USE_MIN_MTU:
1224 			if (m == NULL || m->m_len != sizeof(optval)) {
1225 				error = EINVAL;
1226 				break;
1227 			}
1228 			optval = *mtod(m, int *);
1229 			{
1230 				struct ip6_pktopts **optp;
1231 				optp = &inp->inp_outputopts6;
1232 				error = ip6_pcbopt(optname, (u_char *)&optval,
1233 				    sizeof(optval), optp, privileged, uproto);
1234 				break;
1235 			}
1236 
1237 		case IPV6_PKTINFO:
1238 		case IPV6_HOPOPTS:
1239 		case IPV6_RTHDR:
1240 		case IPV6_DSTOPTS:
1241 		case IPV6_RTHDRDSTOPTS:
1242 		{
1243 			/* new advanced API (RFC3542) */
1244 			u_char *optbuf;
1245 			int optbuflen;
1246 			struct ip6_pktopts **optp;
1247 
1248 			if (m && m->m_next) {
1249 				error = EINVAL;	/* XXX */
1250 				break;
1251 			}
1252 			if (m) {
1253 				optbuf = mtod(m, u_char *);
1254 				optbuflen = m->m_len;
1255 			} else {
1256 				optbuf = NULL;
1257 				optbuflen = 0;
1258 			}
1259 			optp = &inp->inp_outputopts6;
1260 			error = ip6_pcbopt(optname, optbuf, optbuflen, optp,
1261 			    privileged, uproto);
1262 			break;
1263 		}
1264 #undef OPTSET
1265 
1266 		case IPV6_MULTICAST_IF:
1267 		case IPV6_MULTICAST_HOPS:
1268 		case IPV6_MULTICAST_LOOP:
1269 		case IPV6_JOIN_GROUP:
1270 		case IPV6_LEAVE_GROUP:
1271 			error =	ip6_setmoptions(optname,
1272 						&inp->inp_moptions6,
1273 						m, inp->inp_rtableid);
1274 			break;
1275 
1276 		case IPV6_PORTRANGE:
1277 			if (m == NULL || m->m_len != sizeof(int)) {
1278 				error = EINVAL;
1279 				break;
1280 			}
1281 			optval = *mtod(m, int *);
1282 
1283 			switch (optval) {
1284 			case IPV6_PORTRANGE_DEFAULT:
1285 				inp->inp_flags &= ~(IN6P_LOWPORT);
1286 				inp->inp_flags &= ~(IN6P_HIGHPORT);
1287 				break;
1288 
1289 			case IPV6_PORTRANGE_HIGH:
1290 				inp->inp_flags &= ~(IN6P_LOWPORT);
1291 				inp->inp_flags |= IN6P_HIGHPORT;
1292 				break;
1293 
1294 			case IPV6_PORTRANGE_LOW:
1295 				inp->inp_flags &= ~(IN6P_HIGHPORT);
1296 				inp->inp_flags |= IN6P_LOWPORT;
1297 				break;
1298 
1299 			default:
1300 				error = EINVAL;
1301 				break;
1302 			}
1303 			break;
1304 
1305 		case IPSEC6_OUTSA:
1306 			error = EINVAL;
1307 			break;
1308 
1309 		case IPV6_AUTH_LEVEL:
1310 		case IPV6_ESP_TRANS_LEVEL:
1311 		case IPV6_ESP_NETWORK_LEVEL:
1312 		case IPV6_IPCOMP_LEVEL:
1313 #ifndef IPSEC
1314 			error = EINVAL;
1315 #else
1316 			if (m == NULL || m->m_len != sizeof(int)) {
1317 				error = EINVAL;
1318 				break;
1319 			}
1320 			optval = *mtod(m, int *);
1321 
1322 			if (optval < IPSEC_LEVEL_BYPASS ||
1323 			    optval > IPSEC_LEVEL_UNIQUE) {
1324 				error = EINVAL;
1325 				break;
1326 			}
1327 
1328 			switch (optname) {
1329 			case IPV6_AUTH_LEVEL:
1330 				if (optval < IPSEC_AUTH_LEVEL_DEFAULT &&
1331 				    suser(p)) {
1332 					error = EACCES;
1333 					break;
1334 				}
1335 				inp->inp_seclevel[SL_AUTH] = optval;
1336 				break;
1337 
1338 			case IPV6_ESP_TRANS_LEVEL:
1339 				if (optval < IPSEC_ESP_TRANS_LEVEL_DEFAULT &&
1340 				    suser(p)) {
1341 					error = EACCES;
1342 					break;
1343 				}
1344 				inp->inp_seclevel[SL_ESP_TRANS] = optval;
1345 				break;
1346 
1347 			case IPV6_ESP_NETWORK_LEVEL:
1348 				if (optval < IPSEC_ESP_NETWORK_LEVEL_DEFAULT &&
1349 				    suser(p)) {
1350 					error = EACCES;
1351 					break;
1352 				}
1353 				inp->inp_seclevel[SL_ESP_NETWORK] = optval;
1354 				break;
1355 
1356 			case IPV6_IPCOMP_LEVEL:
1357 				if (optval < IPSEC_IPCOMP_LEVEL_DEFAULT &&
1358 				    suser(p)) {
1359 					error = EACCES;
1360 					break;
1361 				}
1362 				inp->inp_seclevel[SL_IPCOMP] = optval;
1363 				break;
1364 			}
1365 #endif
1366 			break;
1367 		case SO_RTABLE:
1368 			if (m == NULL || m->m_len < sizeof(u_int)) {
1369 				error = EINVAL;
1370 				break;
1371 			}
1372 			rtid = *mtod(m, u_int *);
1373 			if (inp->inp_rtableid == rtid)
1374 				break;
1375 			/* needs privileges to switch when already set */
1376 			if (rtableid != rtid && rtableid != 0 &&
1377 			    (error = suser(p)) != 0)
1378 				break;
1379 			error = in_pcbset_rtableid(inp, rtid);
1380 			break;
1381 		case IPV6_PIPEX:
1382 			if (m != NULL && m->m_len == sizeof(int))
1383 				inp->inp_pipex = *mtod(m, int *);
1384 			else
1385 				error = EINVAL;
1386 			break;
1387 
1388 		default:
1389 			error = ENOPROTOOPT;
1390 			break;
1391 		}
1392 		break;
1393 
1394 	case PRCO_GETOPT:
1395 		switch (optname) {
1396 
1397 		case IPV6_RECVHOPOPTS:
1398 		case IPV6_RECVDSTOPTS:
1399 		case IPV6_UNICAST_HOPS:
1400 		case IPV6_MINHOPCOUNT:
1401 		case IPV6_RECVPKTINFO:
1402 		case IPV6_RECVHOPLIMIT:
1403 		case IPV6_RECVRTHDR:
1404 		case IPV6_RECVPATHMTU:
1405 
1406 		case IPV6_V6ONLY:
1407 		case IPV6_PORTRANGE:
1408 		case IPV6_RECVTCLASS:
1409 		case IPV6_AUTOFLOWLABEL:
1410 		case IPV6_RECVDSTPORT:
1411 			switch (optname) {
1412 
1413 			case IPV6_RECVHOPOPTS:
1414 				optval = OPTBIT(IN6P_HOPOPTS);
1415 				break;
1416 
1417 			case IPV6_RECVDSTOPTS:
1418 				optval = OPTBIT(IN6P_DSTOPTS);
1419 				break;
1420 
1421 			case IPV6_UNICAST_HOPS:
1422 				optval = inp->inp_hops;
1423 				break;
1424 
1425 			case IPV6_MINHOPCOUNT:
1426 				optval = inp->inp_ip6_minhlim;
1427 				break;
1428 
1429 			case IPV6_RECVPKTINFO:
1430 				optval = OPTBIT(IN6P_PKTINFO);
1431 				break;
1432 
1433 			case IPV6_RECVHOPLIMIT:
1434 				optval = OPTBIT(IN6P_HOPLIMIT);
1435 				break;
1436 
1437 			case IPV6_RECVRTHDR:
1438 				optval = OPTBIT(IN6P_RTHDR);
1439 				break;
1440 
1441 			case IPV6_RECVPATHMTU:
1442 				optval = OPTBIT(IN6P_MTU);
1443 				break;
1444 
1445 			case IPV6_V6ONLY:
1446 				optval = 1;
1447 				break;
1448 
1449 			case IPV6_PORTRANGE:
1450 			    {
1451 				int flags;
1452 				flags = inp->inp_flags;
1453 				if (flags & IN6P_HIGHPORT)
1454 					optval = IPV6_PORTRANGE_HIGH;
1455 				else if (flags & IN6P_LOWPORT)
1456 					optval = IPV6_PORTRANGE_LOW;
1457 				else
1458 					optval = 0;
1459 				break;
1460 			    }
1461 			case IPV6_RECVTCLASS:
1462 				optval = OPTBIT(IN6P_TCLASS);
1463 				break;
1464 
1465 			case IPV6_AUTOFLOWLABEL:
1466 				optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1467 				break;
1468 
1469 			case IPV6_RECVDSTPORT:
1470 				optval = OPTBIT(IN6P_RECVDSTPORT);
1471 				break;
1472 			}
1473 			if (error)
1474 				break;
1475 			m->m_len = sizeof(int);
1476 			*mtod(m, int *) = optval;
1477 			break;
1478 
1479 		case IPV6_PATHMTU:
1480 		{
1481 			u_long pmtu = 0;
1482 			struct ip6_mtuinfo mtuinfo;
1483 			struct ifnet *ifp;
1484 			struct rtentry *rt;
1485 
1486 			if (!(so->so_state & SS_ISCONNECTED))
1487 				return (ENOTCONN);
1488 
1489 			rt = in_pcbrtentry(inp);
1490 			if (!rtisvalid(rt))
1491 				return (EHOSTUNREACH);
1492 
1493 			ifp = if_get(rt->rt_ifidx);
1494 			if (ifp == NULL)
1495 				return (EHOSTUNREACH);
1496 			/*
1497 			 * XXX: we dot not consider the case of source
1498 			 * routing, or optional information to specify
1499 			 * the outgoing interface.
1500 			 */
1501 			error = ip6_getpmtu(rt, ifp, &pmtu);
1502 			if_put(ifp);
1503 			if (error)
1504 				break;
1505 			if (pmtu > IPV6_MAXPACKET)
1506 				pmtu = IPV6_MAXPACKET;
1507 
1508 			bzero(&mtuinfo, sizeof(mtuinfo));
1509 			mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1510 			optdata = (void *)&mtuinfo;
1511 			optdatalen = sizeof(mtuinfo);
1512 			if (optdatalen > MCLBYTES)
1513 				return (EMSGSIZE); /* XXX */
1514 			if (optdatalen > MLEN)
1515 				MCLGET(m, M_WAIT);
1516 			m->m_len = optdatalen;
1517 			bcopy(optdata, mtod(m, void *), optdatalen);
1518 			break;
1519 		}
1520 
1521 		case IPV6_PKTINFO:
1522 		case IPV6_HOPOPTS:
1523 		case IPV6_RTHDR:
1524 		case IPV6_DSTOPTS:
1525 		case IPV6_RTHDRDSTOPTS:
1526 		case IPV6_TCLASS:
1527 		case IPV6_DONTFRAG:
1528 		case IPV6_USE_MIN_MTU:
1529 			error = ip6_getpcbopt(inp->inp_outputopts6,
1530 			    optname, m);
1531 			break;
1532 
1533 		case IPV6_MULTICAST_IF:
1534 		case IPV6_MULTICAST_HOPS:
1535 		case IPV6_MULTICAST_LOOP:
1536 		case IPV6_JOIN_GROUP:
1537 		case IPV6_LEAVE_GROUP:
1538 			error = ip6_getmoptions(optname,
1539 			    inp->inp_moptions6, m);
1540 			break;
1541 
1542 		case IPSEC6_OUTSA:
1543 			error = EINVAL;
1544 			break;
1545 
1546 		case IPV6_AUTH_LEVEL:
1547 		case IPV6_ESP_TRANS_LEVEL:
1548 		case IPV6_ESP_NETWORK_LEVEL:
1549 		case IPV6_IPCOMP_LEVEL:
1550 #ifndef IPSEC
1551 			m->m_len = sizeof(int);
1552 			*mtod(m, int *) = IPSEC_LEVEL_NONE;
1553 #else
1554 			m->m_len = sizeof(int);
1555 			switch (optname) {
1556 			case IPV6_AUTH_LEVEL:
1557 				optval = inp->inp_seclevel[SL_AUTH];
1558 				break;
1559 
1560 			case IPV6_ESP_TRANS_LEVEL:
1561 				optval =
1562 				    inp->inp_seclevel[SL_ESP_TRANS];
1563 				break;
1564 
1565 			case IPV6_ESP_NETWORK_LEVEL:
1566 				optval =
1567 				    inp->inp_seclevel[SL_ESP_NETWORK];
1568 				break;
1569 
1570 			case IPV6_IPCOMP_LEVEL:
1571 				optval = inp->inp_seclevel[SL_IPCOMP];
1572 				break;
1573 			}
1574 			*mtod(m, int *) = optval;
1575 #endif
1576 			break;
1577 		case SO_RTABLE:
1578 			m->m_len = sizeof(u_int);
1579 			*mtod(m, u_int *) = inp->inp_rtableid;
1580 			break;
1581 		case IPV6_PIPEX:
1582 			m->m_len = sizeof(int);
1583 			*mtod(m, int *) = inp->inp_pipex;
1584 			break;
1585 
1586 		default:
1587 			error = ENOPROTOOPT;
1588 			break;
1589 		}
1590 		break;
1591 	}
1592 	return (error);
1593 }
1594 
1595 int
1596 ip6_raw_ctloutput(int op, struct socket *so, int level, int optname,
1597     struct mbuf *m)
1598 {
1599 	int error = 0, optval;
1600 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
1601 	struct inpcb *inp = sotoinpcb(so);
1602 
1603 	if (level != IPPROTO_IPV6)
1604 		return (EINVAL);
1605 
1606 	switch (optname) {
1607 	case IPV6_CHECKSUM:
1608 		/*
1609 		 * For ICMPv6 sockets, no modification allowed for checksum
1610 		 * offset, permit "no change" values to help existing apps.
1611 		 *
1612 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
1613 		 * for an ICMPv6 socket will fail."
1614 		 * The current behavior does not meet RFC3542.
1615 		 */
1616 		switch (op) {
1617 		case PRCO_SETOPT:
1618 			if (m == NULL || m->m_len != sizeof(int)) {
1619 				error = EINVAL;
1620 				break;
1621 			}
1622 			optval = *mtod(m, int *);
1623 			if (optval < -1 ||
1624 			    (optval > 0 && (optval % 2) != 0)) {
1625 				/*
1626 				 * The API assumes non-negative even offset
1627 				 * values or -1 as a special value.
1628 				 */
1629 				error = EINVAL;
1630 			} else if (so->so_proto->pr_protocol ==
1631 			    IPPROTO_ICMPV6) {
1632 				if (optval != icmp6off)
1633 					error = EINVAL;
1634 			} else
1635 				inp->inp_cksum6 = optval;
1636 			break;
1637 
1638 		case PRCO_GETOPT:
1639 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
1640 				optval = icmp6off;
1641 			else
1642 				optval = inp->inp_cksum6;
1643 
1644 			m->m_len = sizeof(int);
1645 			*mtod(m, int *) = optval;
1646 			break;
1647 
1648 		default:
1649 			error = EINVAL;
1650 			break;
1651 		}
1652 		break;
1653 
1654 	default:
1655 		error = ENOPROTOOPT;
1656 		break;
1657 	}
1658 
1659 	return (error);
1660 }
1661 
1662 /*
1663  * initialize ip6_pktopts.  beware that there are non-zero default values in
1664  * the struct.
1665  */
1666 void
1667 ip6_initpktopts(struct ip6_pktopts *opt)
1668 {
1669 	bzero(opt, sizeof(*opt));
1670 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
1671 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
1672 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
1673 }
1674 
1675 int
1676 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
1677     int priv, int uproto)
1678 {
1679 	struct ip6_pktopts *opt;
1680 
1681 	if (*pktopt == NULL) {
1682 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
1683 		    M_WAITOK);
1684 		ip6_initpktopts(*pktopt);
1685 	}
1686 	opt = *pktopt;
1687 
1688 	return (ip6_setpktopt(optname, buf, len, opt, priv, 1, uproto));
1689 }
1690 
1691 int
1692 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct mbuf *m)
1693 {
1694 	void *optdata = NULL;
1695 	int optdatalen = 0;
1696 	struct ip6_ext *ip6e;
1697 	int error = 0;
1698 	struct in6_pktinfo null_pktinfo;
1699 	int deftclass = 0, on;
1700 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
1701 
1702 	switch (optname) {
1703 	case IPV6_PKTINFO:
1704 		if (pktopt && pktopt->ip6po_pktinfo)
1705 			optdata = (void *)pktopt->ip6po_pktinfo;
1706 		else {
1707 			/* XXX: we don't have to do this every time... */
1708 			bzero(&null_pktinfo, sizeof(null_pktinfo));
1709 			optdata = (void *)&null_pktinfo;
1710 		}
1711 		optdatalen = sizeof(struct in6_pktinfo);
1712 		break;
1713 	case IPV6_TCLASS:
1714 		if (pktopt && pktopt->ip6po_tclass >= 0)
1715 			optdata = (void *)&pktopt->ip6po_tclass;
1716 		else
1717 			optdata = (void *)&deftclass;
1718 		optdatalen = sizeof(int);
1719 		break;
1720 	case IPV6_HOPOPTS:
1721 		if (pktopt && pktopt->ip6po_hbh) {
1722 			optdata = (void *)pktopt->ip6po_hbh;
1723 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
1724 			optdatalen = (ip6e->ip6e_len + 1) << 3;
1725 		}
1726 		break;
1727 	case IPV6_RTHDR:
1728 		if (pktopt && pktopt->ip6po_rthdr) {
1729 			optdata = (void *)pktopt->ip6po_rthdr;
1730 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
1731 			optdatalen = (ip6e->ip6e_len + 1) << 3;
1732 		}
1733 		break;
1734 	case IPV6_RTHDRDSTOPTS:
1735 		if (pktopt && pktopt->ip6po_dest1) {
1736 			optdata = (void *)pktopt->ip6po_dest1;
1737 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
1738 			optdatalen = (ip6e->ip6e_len + 1) << 3;
1739 		}
1740 		break;
1741 	case IPV6_DSTOPTS:
1742 		if (pktopt && pktopt->ip6po_dest2) {
1743 			optdata = (void *)pktopt->ip6po_dest2;
1744 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
1745 			optdatalen = (ip6e->ip6e_len + 1) << 3;
1746 		}
1747 		break;
1748 	case IPV6_USE_MIN_MTU:
1749 		if (pktopt)
1750 			optdata = (void *)&pktopt->ip6po_minmtu;
1751 		else
1752 			optdata = (void *)&defminmtu;
1753 		optdatalen = sizeof(int);
1754 		break;
1755 	case IPV6_DONTFRAG:
1756 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
1757 			on = 1;
1758 		else
1759 			on = 0;
1760 		optdata = (void *)&on;
1761 		optdatalen = sizeof(on);
1762 		break;
1763 	default:		/* should not happen */
1764 #ifdef DIAGNOSTIC
1765 		panic("%s: unexpected option", __func__);
1766 #endif
1767 		return (ENOPROTOOPT);
1768 	}
1769 
1770 	if (optdatalen > MCLBYTES)
1771 		return (EMSGSIZE); /* XXX */
1772 	if (optdatalen > MLEN)
1773 		MCLGET(m, M_WAIT);
1774 	m->m_len = optdatalen;
1775 	if (optdatalen)
1776 		bcopy(optdata, mtod(m, void *), optdatalen);
1777 
1778 	return (error);
1779 }
1780 
1781 void
1782 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
1783 {
1784 	if (optname == -1 || optname == IPV6_PKTINFO) {
1785 		if (pktopt->ip6po_pktinfo)
1786 			free(pktopt->ip6po_pktinfo, M_IP6OPT, 0);
1787 		pktopt->ip6po_pktinfo = NULL;
1788 	}
1789 	if (optname == -1 || optname == IPV6_HOPLIMIT)
1790 		pktopt->ip6po_hlim = -1;
1791 	if (optname == -1 || optname == IPV6_TCLASS)
1792 		pktopt->ip6po_tclass = -1;
1793 	if (optname == -1 || optname == IPV6_HOPOPTS) {
1794 		if (pktopt->ip6po_hbh)
1795 			free(pktopt->ip6po_hbh, M_IP6OPT, 0);
1796 		pktopt->ip6po_hbh = NULL;
1797 	}
1798 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
1799 		if (pktopt->ip6po_dest1)
1800 			free(pktopt->ip6po_dest1, M_IP6OPT, 0);
1801 		pktopt->ip6po_dest1 = NULL;
1802 	}
1803 	if (optname == -1 || optname == IPV6_RTHDR) {
1804 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
1805 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT, 0);
1806 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
1807 		if (pktopt->ip6po_route.ro_rt) {
1808 			rtfree(pktopt->ip6po_route.ro_rt);
1809 			pktopt->ip6po_route.ro_rt = NULL;
1810 		}
1811 	}
1812 	if (optname == -1 || optname == IPV6_DSTOPTS) {
1813 		if (pktopt->ip6po_dest2)
1814 			free(pktopt->ip6po_dest2, M_IP6OPT, 0);
1815 		pktopt->ip6po_dest2 = NULL;
1816 	}
1817 }
1818 
1819 #define PKTOPT_EXTHDRCPY(type) \
1820 do {\
1821 	if (src->type) {\
1822 		size_t hlen;\
1823 		hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
1824 		dst->type = malloc(hlen, M_IP6OPT, M_NOWAIT);\
1825 		if (dst->type == NULL)\
1826 			goto bad;\
1827 		memcpy(dst->type, src->type, hlen);\
1828 	}\
1829 } while (/*CONSTCOND*/ 0)
1830 
1831 int
1832 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src)
1833 {
1834 	dst->ip6po_hlim = src->ip6po_hlim;
1835 	dst->ip6po_tclass = src->ip6po_tclass;
1836 	dst->ip6po_flags = src->ip6po_flags;
1837 	if (src->ip6po_pktinfo) {
1838 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
1839 		    M_IP6OPT, M_NOWAIT);
1840 		if (dst->ip6po_pktinfo == NULL)
1841 			goto bad;
1842 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
1843 	}
1844 	PKTOPT_EXTHDRCPY(ip6po_hbh);
1845 	PKTOPT_EXTHDRCPY(ip6po_dest1);
1846 	PKTOPT_EXTHDRCPY(ip6po_dest2);
1847 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
1848 	return (0);
1849 
1850   bad:
1851 	ip6_clearpktopts(dst, -1);
1852 	return (ENOBUFS);
1853 }
1854 #undef PKTOPT_EXTHDRCPY
1855 
1856 void
1857 ip6_freepcbopts(struct ip6_pktopts *pktopt)
1858 {
1859 	if (pktopt == NULL)
1860 		return;
1861 
1862 	ip6_clearpktopts(pktopt, -1);
1863 
1864 	free(pktopt, M_IP6OPT, 0);
1865 }
1866 
1867 /*
1868  * Set the IP6 multicast options in response to user setsockopt().
1869  */
1870 int
1871 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m,
1872     unsigned int rtableid)
1873 {
1874 	int error = 0;
1875 	u_int loop, ifindex;
1876 	struct ipv6_mreq *mreq;
1877 	struct ifnet *ifp;
1878 	struct ip6_moptions *im6o = *im6op;
1879 	struct in6_multi_mship *imm;
1880 	struct proc *p = curproc;	/* XXX */
1881 
1882 	if (im6o == NULL) {
1883 		/*
1884 		 * No multicast option buffer attached to the pcb;
1885 		 * allocate one and initialize to default values.
1886 		 */
1887 		im6o = malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
1888 		if (im6o == NULL)
1889 			return (ENOBUFS);
1890 		*im6op = im6o;
1891 		im6o->im6o_ifidx = 0;
1892 		im6o->im6o_hlim = ip6_defmcasthlim;
1893 		im6o->im6o_loop = IPV6_DEFAULT_MULTICAST_LOOP;
1894 		LIST_INIT(&im6o->im6o_memberships);
1895 	}
1896 
1897 	switch (optname) {
1898 
1899 	case IPV6_MULTICAST_IF:
1900 		/*
1901 		 * Select the interface for outgoing multicast packets.
1902 		 */
1903 		if (m == NULL || m->m_len != sizeof(u_int)) {
1904 			error = EINVAL;
1905 			break;
1906 		}
1907 		memcpy(&ifindex, mtod(m, u_int *), sizeof(ifindex));
1908 		if (ifindex != 0) {
1909 			ifp = if_get(ifindex);
1910 			if (ifp == NULL) {
1911 				error = ENXIO;	/* XXX EINVAL? */
1912 				break;
1913 			}
1914 			if (ifp->if_rdomain != rtable_l2(rtableid) ||
1915 			    (ifp->if_flags & IFF_MULTICAST) == 0) {
1916 				error = EADDRNOTAVAIL;
1917 				if_put(ifp);
1918 				break;
1919 			}
1920 			if_put(ifp);
1921 		}
1922 		im6o->im6o_ifidx = ifindex;
1923 		break;
1924 
1925 	case IPV6_MULTICAST_HOPS:
1926 	    {
1927 		/*
1928 		 * Set the IP6 hoplimit for outgoing multicast packets.
1929 		 */
1930 		int optval;
1931 		if (m == NULL || m->m_len != sizeof(int)) {
1932 			error = EINVAL;
1933 			break;
1934 		}
1935 		memcpy(&optval, mtod(m, u_int *), sizeof(optval));
1936 		if (optval < -1 || optval >= 256)
1937 			error = EINVAL;
1938 		else if (optval == -1)
1939 			im6o->im6o_hlim = ip6_defmcasthlim;
1940 		else
1941 			im6o->im6o_hlim = optval;
1942 		break;
1943 	    }
1944 
1945 	case IPV6_MULTICAST_LOOP:
1946 		/*
1947 		 * Set the loopback flag for outgoing multicast packets.
1948 		 * Must be zero or one.
1949 		 */
1950 		if (m == NULL || m->m_len != sizeof(u_int)) {
1951 			error = EINVAL;
1952 			break;
1953 		}
1954 		memcpy(&loop, mtod(m, u_int *), sizeof(loop));
1955 		if (loop > 1) {
1956 			error = EINVAL;
1957 			break;
1958 		}
1959 		im6o->im6o_loop = loop;
1960 		break;
1961 
1962 	case IPV6_JOIN_GROUP:
1963 		/*
1964 		 * Add a multicast group membership.
1965 		 * Group must be a valid IP6 multicast address.
1966 		 */
1967 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
1968 			error = EINVAL;
1969 			break;
1970 		}
1971 		mreq = mtod(m, struct ipv6_mreq *);
1972 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
1973 			/*
1974 			 * We use the unspecified address to specify to accept
1975 			 * all multicast addresses. Only super user is allowed
1976 			 * to do this.
1977 			 */
1978 			if (suser(p))
1979 			{
1980 				error = EACCES;
1981 				break;
1982 			}
1983 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
1984 			error = EINVAL;
1985 			break;
1986 		}
1987 
1988 		/*
1989 		 * If no interface was explicitly specified, choose an
1990 		 * appropriate one according to the given multicast address.
1991 		 */
1992 		if (mreq->ipv6mr_interface == 0) {
1993 			struct rtentry *rt;
1994 			struct sockaddr_in6 dst;
1995 
1996 			memset(&dst, 0, sizeof(dst));
1997 			dst.sin6_len = sizeof(dst);
1998 			dst.sin6_family = AF_INET6;
1999 			dst.sin6_addr = mreq->ipv6mr_multiaddr;
2000 			rt = rtalloc(sin6tosa(&dst), RT_RESOLVE, rtableid);
2001 			if (rt == NULL) {
2002 				error = EADDRNOTAVAIL;
2003 				break;
2004 			}
2005 			ifp = if_get(rt->rt_ifidx);
2006 			rtfree(rt);
2007 		} else {
2008 			/*
2009 			 * If the interface is specified, validate it.
2010 			 */
2011 			ifp = if_get(mreq->ipv6mr_interface);
2012 			if (ifp == NULL) {
2013 				error = ENXIO;	/* XXX EINVAL? */
2014 				break;
2015 			}
2016 		}
2017 
2018 		/*
2019 		 * See if we found an interface, and confirm that it
2020 		 * supports multicast
2021 		 */
2022 		if (ifp == NULL || ifp->if_rdomain != rtable_l2(rtableid) ||
2023 		    (ifp->if_flags & IFF_MULTICAST) == 0) {
2024 			if_put(ifp);
2025 			error = EADDRNOTAVAIL;
2026 			break;
2027 		}
2028 		/*
2029 		 * Put interface index into the multicast address,
2030 		 * if the address has link/interface-local scope.
2031 		 */
2032 		if (IN6_IS_SCOPE_EMBED(&mreq->ipv6mr_multiaddr)) {
2033 			mreq->ipv6mr_multiaddr.s6_addr16[1] =
2034 			    htons(ifp->if_index);
2035 		}
2036 		/*
2037 		 * See if the membership already exists.
2038 		 */
2039 		LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain)
2040 			if (imm->i6mm_maddr->in6m_ifidx == ifp->if_index &&
2041 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2042 			    &mreq->ipv6mr_multiaddr))
2043 				break;
2044 		if (imm != NULL) {
2045 			if_put(ifp);
2046 			error = EADDRINUSE;
2047 			break;
2048 		}
2049 		/*
2050 		 * Everything looks good; add a new record to the multicast
2051 		 * address list for the given interface.
2052 		 */
2053 		imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error);
2054 		if_put(ifp);
2055 		if (!imm)
2056 			break;
2057 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2058 		break;
2059 
2060 	case IPV6_LEAVE_GROUP:
2061 		/*
2062 		 * Drop a multicast group membership.
2063 		 * Group must be a valid IP6 multicast address.
2064 		 */
2065 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2066 			error = EINVAL;
2067 			break;
2068 		}
2069 		mreq = mtod(m, struct ipv6_mreq *);
2070 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2071 			if (suser(p)) {
2072 				error = EACCES;
2073 				break;
2074 			}
2075 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2076 			error = EINVAL;
2077 			break;
2078 		}
2079 
2080 		/*
2081 		 * Put interface index into the multicast address,
2082 		 * if the address has link-local scope.
2083 		 */
2084 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2085 			mreq->ipv6mr_multiaddr.s6_addr16[1] =
2086 			    htons(mreq->ipv6mr_interface);
2087 		}
2088 
2089 		/*
2090 		 * If an interface address was specified, get a pointer
2091 		 * to its ifnet structure.
2092 		 */
2093 		if (mreq->ipv6mr_interface == 0)
2094 			ifp = NULL;
2095 		else {
2096 			ifp = if_get(mreq->ipv6mr_interface);
2097 			if (ifp == NULL) {
2098 				error = ENXIO;	/* XXX EINVAL? */
2099 				break;
2100 			}
2101 		}
2102 
2103 		/*
2104 		 * Find the membership in the membership list.
2105 		 */
2106 		LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain) {
2107 			if ((ifp == NULL ||
2108 			    imm->i6mm_maddr->in6m_ifidx == ifp->if_index) &&
2109 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2110 			    &mreq->ipv6mr_multiaddr))
2111 				break;
2112 		}
2113 
2114 		if_put(ifp);
2115 
2116 		if (imm == NULL) {
2117 			/* Unable to resolve interface */
2118 			error = EADDRNOTAVAIL;
2119 			break;
2120 		}
2121 		/*
2122 		 * Give up the multicast address record to which the
2123 		 * membership points.
2124 		 */
2125 		LIST_REMOVE(imm, i6mm_chain);
2126 		in6_leavegroup(imm);
2127 		break;
2128 
2129 	default:
2130 		error = EOPNOTSUPP;
2131 		break;
2132 	}
2133 
2134 	/*
2135 	 * If all options have default values, no need to keep the option
2136 	 * structure.
2137 	 */
2138 	if (im6o->im6o_ifidx == 0 &&
2139 	    im6o->im6o_hlim == ip6_defmcasthlim &&
2140 	    im6o->im6o_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2141 	    LIST_EMPTY(&im6o->im6o_memberships)) {
2142 		free(*im6op, M_IPMOPTS, sizeof(**im6op));
2143 		*im6op = NULL;
2144 	}
2145 
2146 	return (error);
2147 }
2148 
2149 /*
2150  * Return the IP6 multicast options in response to user getsockopt().
2151  */
2152 int
2153 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf *m)
2154 {
2155 	u_int *hlim, *loop, *ifindex;
2156 
2157 	switch (optname) {
2158 	case IPV6_MULTICAST_IF:
2159 		ifindex = mtod(m, u_int *);
2160 		m->m_len = sizeof(u_int);
2161 		if (im6o == NULL || im6o->im6o_ifidx == 0)
2162 			*ifindex = 0;
2163 		else
2164 			*ifindex = im6o->im6o_ifidx;
2165 		return (0);
2166 
2167 	case IPV6_MULTICAST_HOPS:
2168 		hlim = mtod(m, u_int *);
2169 		m->m_len = sizeof(u_int);
2170 		if (im6o == NULL)
2171 			*hlim = ip6_defmcasthlim;
2172 		else
2173 			*hlim = im6o->im6o_hlim;
2174 		return (0);
2175 
2176 	case IPV6_MULTICAST_LOOP:
2177 		loop = mtod(m, u_int *);
2178 		m->m_len = sizeof(u_int);
2179 		if (im6o == NULL)
2180 			*loop = ip6_defmcasthlim;
2181 		else
2182 			*loop = im6o->im6o_loop;
2183 		return (0);
2184 
2185 	default:
2186 		return (EOPNOTSUPP);
2187 	}
2188 }
2189 
2190 /*
2191  * Discard the IP6 multicast options.
2192  */
2193 void
2194 ip6_freemoptions(struct ip6_moptions *im6o)
2195 {
2196 	struct in6_multi_mship *imm;
2197 
2198 	if (im6o == NULL)
2199 		return;
2200 
2201 	while (!LIST_EMPTY(&im6o->im6o_memberships)) {
2202 		imm = LIST_FIRST(&im6o->im6o_memberships);
2203 		LIST_REMOVE(imm, i6mm_chain);
2204 		in6_leavegroup(imm);
2205 	}
2206 	free(im6o, M_IPMOPTS, sizeof(*im6o));
2207 }
2208 
2209 /*
2210  * Set IPv6 outgoing packet options based on advanced API.
2211  */
2212 int
2213 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2214     struct ip6_pktopts *stickyopt, int priv, int uproto)
2215 {
2216 	u_int clen;
2217 	struct cmsghdr *cm = 0;
2218 	caddr_t cmsgs;
2219 	int error;
2220 
2221 	if (control == NULL || opt == NULL)
2222 		return (EINVAL);
2223 
2224 	ip6_initpktopts(opt);
2225 	if (stickyopt) {
2226 		int error;
2227 
2228 		/*
2229 		 * If stickyopt is provided, make a local copy of the options
2230 		 * for this particular packet, then override them by ancillary
2231 		 * objects.
2232 		 * XXX: copypktopts() does not copy the cached route to a next
2233 		 * hop (if any).  This is not very good in terms of efficiency,
2234 		 * but we can allow this since this option should be rarely
2235 		 * used.
2236 		 */
2237 		if ((error = copypktopts(opt, stickyopt)) != 0)
2238 			return (error);
2239 	}
2240 
2241 	/*
2242 	 * XXX: Currently, we assume all the optional information is stored
2243 	 * in a single mbuf.
2244 	 */
2245 	if (control->m_next)
2246 		return (EINVAL);
2247 
2248 	clen = control->m_len;
2249 	cmsgs = mtod(control, caddr_t);
2250 	do {
2251 		if (clen < CMSG_LEN(0))
2252 			return (EINVAL);
2253 		cm = (struct cmsghdr *)cmsgs;
2254 		if (cm->cmsg_len < CMSG_LEN(0) || cm->cmsg_len > clen ||
2255 		    CMSG_ALIGN(cm->cmsg_len) > clen)
2256 			return (EINVAL);
2257 		if (cm->cmsg_level == IPPROTO_IPV6) {
2258 			error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2259 			    cm->cmsg_len - CMSG_LEN(0), opt, priv, 0, uproto);
2260 			if (error)
2261 				return (error);
2262 		}
2263 
2264 		clen -= CMSG_ALIGN(cm->cmsg_len);
2265 		cmsgs += CMSG_ALIGN(cm->cmsg_len);
2266 	} while (clen);
2267 
2268 	return (0);
2269 }
2270 
2271 /*
2272  * Set a particular packet option, as a sticky option or an ancillary data
2273  * item.  "len" can be 0 only when it's a sticky option.
2274  */
2275 int
2276 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2277     int priv, int sticky, int uproto)
2278 {
2279 	int minmtupolicy;
2280 
2281 	switch (optname) {
2282 	case IPV6_PKTINFO:
2283 	{
2284 		struct ifnet *ifp = NULL;
2285 		struct in6_pktinfo *pktinfo;
2286 
2287 		if (len != sizeof(struct in6_pktinfo))
2288 			return (EINVAL);
2289 
2290 		pktinfo = (struct in6_pktinfo *)buf;
2291 
2292 		/*
2293 		 * An application can clear any sticky IPV6_PKTINFO option by
2294 		 * doing a "regular" setsockopt with ipi6_addr being
2295 		 * in6addr_any and ipi6_ifindex being zero.
2296 		 * [RFC 3542, Section 6]
2297 		 */
2298 		if (opt->ip6po_pktinfo &&
2299 		    pktinfo->ipi6_ifindex == 0 &&
2300 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2301 			ip6_clearpktopts(opt, optname);
2302 			break;
2303 		}
2304 
2305 		if (uproto == IPPROTO_TCP &&
2306 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2307 			return (EINVAL);
2308 		}
2309 
2310 		if (pktinfo->ipi6_ifindex) {
2311 			ifp = if_get(pktinfo->ipi6_ifindex);
2312 			if (ifp == NULL)
2313 				return (ENXIO);
2314 			if_put(ifp);
2315 		}
2316 
2317 		/*
2318 		 * We store the address anyway, and let in6_selectsrc()
2319 		 * validate the specified address.  This is because ipi6_addr
2320 		 * may not have enough information about its scope zone, and
2321 		 * we may need additional information (such as outgoing
2322 		 * interface or the scope zone of a destination address) to
2323 		 * disambiguate the scope.
2324 		 * XXX: the delay of the validation may confuse the
2325 		 * application when it is used as a sticky option.
2326 		 */
2327 		if (opt->ip6po_pktinfo == NULL) {
2328 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2329 			    M_IP6OPT, M_NOWAIT);
2330 			if (opt->ip6po_pktinfo == NULL)
2331 				return (ENOBUFS);
2332 		}
2333 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2334 		break;
2335 	}
2336 
2337 	case IPV6_HOPLIMIT:
2338 	{
2339 		int *hlimp;
2340 
2341 		/*
2342 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2343 		 * to simplify the ordering among hoplimit options.
2344 		 */
2345 		if (sticky)
2346 			return (ENOPROTOOPT);
2347 
2348 		if (len != sizeof(int))
2349 			return (EINVAL);
2350 		hlimp = (int *)buf;
2351 		if (*hlimp < -1 || *hlimp > 255)
2352 			return (EINVAL);
2353 
2354 		opt->ip6po_hlim = *hlimp;
2355 		break;
2356 	}
2357 
2358 	case IPV6_TCLASS:
2359 	{
2360 		int tclass;
2361 
2362 		if (len != sizeof(int))
2363 			return (EINVAL);
2364 		tclass = *(int *)buf;
2365 		if (tclass < -1 || tclass > 255)
2366 			return (EINVAL);
2367 
2368 		opt->ip6po_tclass = tclass;
2369 		break;
2370 	}
2371 	case IPV6_HOPOPTS:
2372 	{
2373 		struct ip6_hbh *hbh;
2374 		int hbhlen;
2375 
2376 		/*
2377 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2378 		 * options, since per-option restriction has too much
2379 		 * overhead.
2380 		 */
2381 		if (!priv)
2382 			return (EPERM);
2383 
2384 		if (len == 0) {
2385 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2386 			break;	/* just remove the option */
2387 		}
2388 
2389 		/* message length validation */
2390 		if (len < sizeof(struct ip6_hbh))
2391 			return (EINVAL);
2392 		hbh = (struct ip6_hbh *)buf;
2393 		hbhlen = (hbh->ip6h_len + 1) << 3;
2394 		if (len != hbhlen)
2395 			return (EINVAL);
2396 
2397 		/* turn off the previous option, then set the new option. */
2398 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2399 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2400 		if (opt->ip6po_hbh == NULL)
2401 			return (ENOBUFS);
2402 		memcpy(opt->ip6po_hbh, hbh, hbhlen);
2403 
2404 		break;
2405 	}
2406 
2407 	case IPV6_DSTOPTS:
2408 	case IPV6_RTHDRDSTOPTS:
2409 	{
2410 		struct ip6_dest *dest, **newdest = NULL;
2411 		int destlen;
2412 
2413 		if (!priv)	/* XXX: see the comment for IPV6_HOPOPTS */
2414 			return (EPERM);
2415 
2416 		if (len == 0) {
2417 			ip6_clearpktopts(opt, optname);
2418 			break;	/* just remove the option */
2419 		}
2420 
2421 		/* message length validation */
2422 		if (len < sizeof(struct ip6_dest))
2423 			return (EINVAL);
2424 		dest = (struct ip6_dest *)buf;
2425 		destlen = (dest->ip6d_len + 1) << 3;
2426 		if (len != destlen)
2427 			return (EINVAL);
2428 		/*
2429 		 * Determine the position that the destination options header
2430 		 * should be inserted; before or after the routing header.
2431 		 */
2432 		switch (optname) {
2433 		case IPV6_RTHDRDSTOPTS:
2434 			newdest = &opt->ip6po_dest1;
2435 			break;
2436 		case IPV6_DSTOPTS:
2437 			newdest = &opt->ip6po_dest2;
2438 			break;
2439 		}
2440 
2441 		/* turn off the previous option, then set the new option. */
2442 		ip6_clearpktopts(opt, optname);
2443 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2444 		if (*newdest == NULL)
2445 			return (ENOBUFS);
2446 		memcpy(*newdest, dest, destlen);
2447 
2448 		break;
2449 	}
2450 
2451 	case IPV6_RTHDR:
2452 	{
2453 		struct ip6_rthdr *rth;
2454 		int rthlen;
2455 
2456 		if (len == 0) {
2457 			ip6_clearpktopts(opt, IPV6_RTHDR);
2458 			break;	/* just remove the option */
2459 		}
2460 
2461 		/* message length validation */
2462 		if (len < sizeof(struct ip6_rthdr))
2463 			return (EINVAL);
2464 		rth = (struct ip6_rthdr *)buf;
2465 		rthlen = (rth->ip6r_len + 1) << 3;
2466 		if (len != rthlen)
2467 			return (EINVAL);
2468 
2469 		switch (rth->ip6r_type) {
2470 		case IPV6_RTHDR_TYPE_0:
2471 			if (rth->ip6r_len == 0)	/* must contain one addr */
2472 				return (EINVAL);
2473 			if (rth->ip6r_len % 2) /* length must be even */
2474 				return (EINVAL);
2475 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2476 				return (EINVAL);
2477 			break;
2478 		default:
2479 			return (EINVAL);	/* not supported */
2480 		}
2481 		/* turn off the previous option */
2482 		ip6_clearpktopts(opt, IPV6_RTHDR);
2483 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2484 		if (opt->ip6po_rthdr == NULL)
2485 			return (ENOBUFS);
2486 		memcpy(opt->ip6po_rthdr, rth, rthlen);
2487 		break;
2488 	}
2489 
2490 	case IPV6_USE_MIN_MTU:
2491 		if (len != sizeof(int))
2492 			return (EINVAL);
2493 		minmtupolicy = *(int *)buf;
2494 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2495 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
2496 		    minmtupolicy != IP6PO_MINMTU_ALL) {
2497 			return (EINVAL);
2498 		}
2499 		opt->ip6po_minmtu = minmtupolicy;
2500 		break;
2501 
2502 	case IPV6_DONTFRAG:
2503 		if (len != sizeof(int))
2504 			return (EINVAL);
2505 
2506 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2507 			/*
2508 			 * we ignore this option for TCP sockets.
2509 			 * (RFC3542 leaves this case unspecified.)
2510 			 */
2511 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2512 		} else
2513 			opt->ip6po_flags |= IP6PO_DONTFRAG;
2514 		break;
2515 
2516 	default:
2517 		return (ENOPROTOOPT);
2518 	} /* end of switch */
2519 
2520 	return (0);
2521 }
2522 
2523 /*
2524  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2525  * packet to the input queue of a specified interface.
2526  */
2527 void
2528 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2529 {
2530 	struct mbuf *copym;
2531 	struct ip6_hdr *ip6;
2532 
2533 	/*
2534 	 * Duplicate the packet.
2535 	 */
2536 	copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
2537 	if (copym == NULL)
2538 		return;
2539 
2540 	/*
2541 	 * Make sure to deep-copy IPv6 header portion in case the data
2542 	 * is in an mbuf cluster, so that we can safely override the IPv6
2543 	 * header portion later.
2544 	 */
2545 	if ((copym->m_flags & M_EXT) != 0 ||
2546 	    copym->m_len < sizeof(struct ip6_hdr)) {
2547 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
2548 		if (copym == NULL)
2549 			return;
2550 	}
2551 
2552 #ifdef DIAGNOSTIC
2553 	if (copym->m_len < sizeof(*ip6)) {
2554 		m_freem(copym);
2555 		return;
2556 	}
2557 #endif
2558 
2559 	ip6 = mtod(copym, struct ip6_hdr *);
2560 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
2561 		ip6->ip6_src.s6_addr16[1] = 0;
2562 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
2563 		ip6->ip6_dst.s6_addr16[1] = 0;
2564 
2565 	if_input_local(ifp, copym, dst->sin6_family);
2566 }
2567 
2568 /*
2569  * Chop IPv6 header off from the payload.
2570  */
2571 int
2572 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
2573 {
2574 	struct mbuf *mh;
2575 	struct ip6_hdr *ip6;
2576 
2577 	ip6 = mtod(m, struct ip6_hdr *);
2578 	if (m->m_len > sizeof(*ip6)) {
2579 		MGET(mh, M_DONTWAIT, MT_HEADER);
2580 		if (mh == NULL) {
2581 			m_freem(m);
2582 			return ENOBUFS;
2583 		}
2584 		M_MOVE_PKTHDR(mh, m);
2585 		m_align(mh, sizeof(*ip6));
2586 		m->m_len -= sizeof(*ip6);
2587 		m->m_data += sizeof(*ip6);
2588 		mh->m_next = m;
2589 		m = mh;
2590 		m->m_len = sizeof(*ip6);
2591 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2592 	}
2593 	exthdrs->ip6e_ip6 = m;
2594 	return 0;
2595 }
2596 
2597 u_int32_t
2598 ip6_randomid(void)
2599 {
2600 	return idgen32(&ip6_id_ctx);
2601 }
2602 
2603 void
2604 ip6_randomid_init(void)
2605 {
2606 	idgen32_init(&ip6_id_ctx);
2607 }
2608 
2609 /*
2610  *	Compute significant parts of the IPv6 checksum pseudo-header
2611  *	for use in a delayed TCP/UDP checksum calculation.
2612  */
2613 static __inline u_int16_t __attribute__((__unused__))
2614 in6_cksum_phdr(const struct in6_addr *src, const struct in6_addr *dst,
2615     u_int32_t len, u_int32_t nxt)
2616 {
2617 	u_int32_t sum = 0;
2618 	const u_int16_t *w;
2619 
2620 	w = (const u_int16_t *) src;
2621 	sum += w[0];
2622 	if (!IN6_IS_SCOPE_EMBED(src))
2623 		sum += w[1];
2624 	sum += w[2]; sum += w[3]; sum += w[4]; sum += w[5];
2625 	sum += w[6]; sum += w[7];
2626 
2627 	w = (const u_int16_t *) dst;
2628 	sum += w[0];
2629 	if (!IN6_IS_SCOPE_EMBED(dst))
2630 		sum += w[1];
2631 	sum += w[2]; sum += w[3]; sum += w[4]; sum += w[5];
2632 	sum += w[6]; sum += w[7];
2633 
2634 	sum += (u_int16_t)(len >> 16) + (u_int16_t)(len /*& 0xffff*/);
2635 
2636 	sum += (u_int16_t)(nxt >> 16) + (u_int16_t)(nxt /*& 0xffff*/);
2637 
2638 	sum = (u_int16_t)(sum >> 16) + (u_int16_t)(sum /*& 0xffff*/);
2639 
2640 	if (sum > 0xffff)
2641 		sum -= 0xffff;
2642 
2643 	return (sum);
2644 }
2645 
2646 /*
2647  * Process a delayed payload checksum calculation.
2648  */
2649 void
2650 in6_delayed_cksum(struct mbuf *m, u_int8_t nxt)
2651 {
2652 	int nxtp, offset;
2653 	u_int16_t csum;
2654 
2655 	offset = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxtp);
2656 	if (offset <= 0 || nxtp != nxt)
2657 		/* If the desired next protocol isn't found, punt. */
2658 		return;
2659 	csum = (u_int16_t)(in6_cksum(m, 0, offset, m->m_pkthdr.len - offset));
2660 
2661 	switch (nxt) {
2662 	case IPPROTO_TCP:
2663 		offset += offsetof(struct tcphdr, th_sum);
2664 		break;
2665 
2666 	case IPPROTO_UDP:
2667 		offset += offsetof(struct udphdr, uh_sum);
2668 		if (csum == 0)
2669 			csum = 0xffff;
2670 		break;
2671 
2672 	case IPPROTO_ICMPV6:
2673 		offset += offsetof(struct icmp6_hdr, icmp6_cksum);
2674 		break;
2675 	}
2676 
2677 	if ((offset + sizeof(u_int16_t)) > m->m_len)
2678 		m_copyback(m, offset, sizeof(csum), &csum, M_NOWAIT);
2679 	else
2680 		*(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
2681 }
2682 
2683 void
2684 in6_proto_cksum_out(struct mbuf *m, struct ifnet *ifp)
2685 {
2686 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2687 
2688 	/* some hw and in6_delayed_cksum need the pseudo header cksum */
2689 	if (m->m_pkthdr.csum_flags &
2690 	    (M_TCP_CSUM_OUT|M_UDP_CSUM_OUT|M_ICMP_CSUM_OUT)) {
2691 		int nxt, offset;
2692 		u_int16_t csum;
2693 
2694 		offset = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
2695 		if (ISSET(m->m_pkthdr.csum_flags, M_TCP_TSO) &&
2696 		    in_ifcap_cksum(m, ifp, IFCAP_TSOv6)) {
2697 			csum = in6_cksum_phdr(&ip6->ip6_src, &ip6->ip6_dst,
2698 			    htonl(0), htonl(nxt));
2699 		} else {
2700 			csum = in6_cksum_phdr(&ip6->ip6_src, &ip6->ip6_dst,
2701 			    htonl(m->m_pkthdr.len - offset), htonl(nxt));
2702 		}
2703 		if (nxt == IPPROTO_TCP)
2704 			offset += offsetof(struct tcphdr, th_sum);
2705 		else if (nxt == IPPROTO_UDP)
2706 			offset += offsetof(struct udphdr, uh_sum);
2707 		else if (nxt == IPPROTO_ICMPV6)
2708 			offset += offsetof(struct icmp6_hdr, icmp6_cksum);
2709 		if ((offset + sizeof(u_int16_t)) > m->m_len)
2710 			m_copyback(m, offset, sizeof(csum), &csum, M_NOWAIT);
2711 		else
2712 			*(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
2713 	}
2714 
2715 	if (m->m_pkthdr.csum_flags & M_TCP_CSUM_OUT) {
2716 		if (!ifp || !(ifp->if_capabilities & IFCAP_CSUM_TCPv6) ||
2717 		    ip6->ip6_nxt != IPPROTO_TCP ||
2718 		    ifp->if_bridgeidx != 0) {
2719 			tcpstat_inc(tcps_outswcsum);
2720 			in6_delayed_cksum(m, IPPROTO_TCP);
2721 			m->m_pkthdr.csum_flags &= ~M_TCP_CSUM_OUT; /* Clear */
2722 		}
2723 	} else if (m->m_pkthdr.csum_flags & M_UDP_CSUM_OUT) {
2724 		if (!ifp || !(ifp->if_capabilities & IFCAP_CSUM_UDPv6) ||
2725 		    ip6->ip6_nxt != IPPROTO_UDP ||
2726 		    ifp->if_bridgeidx != 0) {
2727 			udpstat_inc(udps_outswcsum);
2728 			in6_delayed_cksum(m, IPPROTO_UDP);
2729 			m->m_pkthdr.csum_flags &= ~M_UDP_CSUM_OUT; /* Clear */
2730 		}
2731 	} else if (m->m_pkthdr.csum_flags & M_ICMP_CSUM_OUT) {
2732 		in6_delayed_cksum(m, IPPROTO_ICMPV6);
2733 		m->m_pkthdr.csum_flags &= ~M_ICMP_CSUM_OUT; /* Clear */
2734 	}
2735 }
2736 
2737 #ifdef IPSEC
2738 int
2739 ip6_output_ipsec_lookup(struct mbuf *m, const u_char seclevel[],
2740     struct tdb **tdbout)
2741 {
2742 	struct tdb *tdb;
2743 	struct m_tag *mtag;
2744 	struct tdb_ident *tdbi;
2745 	int error;
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 	error = ipsp_spd_lookup(m, AF_INET6, sizeof(struct ip6_hdr),
2754 	    IPSP_DIRECTION_OUT, NULL, seclevel, &tdb, NULL);
2755 	if (error || tdb == NULL) {
2756 		*tdbout = NULL;
2757 		return error;
2758 	}
2759 	/* Loop detection */
2760 	for (mtag = m_tag_first(m); mtag != NULL; mtag = m_tag_next(m, mtag)) {
2761 		if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE)
2762 			continue;
2763 		tdbi = (struct tdb_ident *)(mtag + 1);
2764 		if (tdbi->spi == tdb->tdb_spi &&
2765 		    tdbi->proto == tdb->tdb_sproto &&
2766 		    tdbi->rdomain == tdb->tdb_rdomain &&
2767 		    !memcmp(&tdbi->dst, &tdb->tdb_dst,
2768 		    sizeof(union sockaddr_union))) {
2769 			/* no IPsec needed */
2770 			tdb_unref(tdb);
2771 			*tdbout = NULL;
2772 			return 0;
2773 		}
2774 	}
2775 	*tdbout = tdb;
2776 	return 0;
2777 }
2778 
2779 int
2780 ip6_output_ipsec_pmtu_update(struct tdb *tdb, struct route_in6 *ro,
2781     struct in6_addr *dst, int ifidx, int rtableid, int transportmode)
2782 {
2783 	struct rtentry *rt = NULL;
2784 	int rt_mtucloned = 0;
2785 
2786 	/* Find a host route to store the mtu in */
2787 	if (ro != NULL)
2788 		rt = ro->ro_rt;
2789 	/* but don't add a PMTU route for transport mode SAs */
2790 	if (transportmode)
2791 		rt = NULL;
2792 	else if (rt == NULL || (rt->rt_flags & RTF_HOST) == 0) {
2793 		struct sockaddr_in6 sin6;
2794 		int error;
2795 
2796 		memset(&sin6, 0, sizeof(sin6));
2797 		sin6.sin6_family = AF_INET6;
2798 		sin6.sin6_len = sizeof(sin6);
2799 		sin6.sin6_addr = *dst;
2800 		sin6.sin6_scope_id = in6_addr2scopeid(ifidx, dst);
2801 		error = in6_embedscope(dst, &sin6, NULL, NULL);
2802 		if (error) {
2803 			/* should be impossible */
2804 			return error;
2805 		}
2806 		rt = icmp6_mtudisc_clone(&sin6, rtableid, 1);
2807 		rt_mtucloned = 1;
2808 	}
2809 	DPRINTF("spi %08x mtu %d rt %p cloned %d",
2810 	    ntohl(tdb->tdb_spi), tdb->tdb_mtu, rt, rt_mtucloned);
2811 	if (rt != NULL) {
2812 		rt->rt_mtu = tdb->tdb_mtu;
2813 		if (ro != NULL && ro->ro_rt != NULL) {
2814 			rtfree(ro->ro_rt);
2815 			ro->ro_rt = rtalloc(sin6tosa(&ro->ro_dst), RT_RESOLVE,
2816 			    rtableid);
2817 		}
2818 		if (rt_mtucloned)
2819 			rtfree(rt);
2820 	}
2821 	return 0;
2822 }
2823 
2824 int
2825 ip6_output_ipsec_send(struct tdb *tdb, struct mbuf *m, struct route_in6 *ro,
2826     int tunalready, int fwd)
2827 {
2828 	struct mbuf_list ml;
2829 	struct ifnet *encif = NULL;
2830 	struct ip6_hdr *ip6;
2831 	struct in6_addr dst;
2832 	u_int len;
2833 	int error, ifidx, rtableid, tso = 0;
2834 
2835 #if NPF > 0
2836 	/*
2837 	 * Packet filter
2838 	 */
2839 	if ((encif = enc_getif(tdb->tdb_rdomain, tdb->tdb_tap)) == NULL ||
2840 	    pf_test(AF_INET6, fwd ? PF_FWD : PF_OUT, encif, &m) != PF_PASS) {
2841 		m_freem(m);
2842 		return EACCES;
2843 	}
2844 	if (m == NULL)
2845 		return 0;
2846 	/*
2847 	 * PF_TAG_REROUTE handling or not...
2848 	 * Packet is entering IPsec so the routing is
2849 	 * already overruled by the IPsec policy.
2850 	 * Until now the change was not reconsidered.
2851 	 * What's the behaviour?
2852 	 */
2853 #endif
2854 
2855 	/* Check if we can chop the TCP packet */
2856 	ip6 = mtod(m, struct ip6_hdr *);
2857 	if (ISSET(m->m_pkthdr.csum_flags, M_TCP_TSO) &&
2858 	    m->m_pkthdr.ph_mss <= tdb->tdb_mtu) {
2859 		tso = 1;
2860 		len = m->m_pkthdr.ph_mss;
2861 	} else
2862 		len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
2863 
2864 	/* Check if we are allowed to fragment */
2865 	dst = ip6->ip6_dst;
2866 	ifidx = m->m_pkthdr.ph_ifidx;
2867 	rtableid = m->m_pkthdr.ph_rtableid;
2868 	if (ip_mtudisc && tdb->tdb_mtu &&
2869 	    len > tdb->tdb_mtu && tdb->tdb_mtutimeout > gettime()) {
2870 		int transportmode;
2871 
2872 		transportmode = (tdb->tdb_dst.sa.sa_family == AF_INET6) &&
2873 		    (IN6_ARE_ADDR_EQUAL(&tdb->tdb_dst.sin6.sin6_addr, &dst));
2874 		error = ip6_output_ipsec_pmtu_update(tdb, ro, &dst, ifidx,
2875 		    rtableid, transportmode);
2876 		if (error) {
2877 			ipsecstat_inc(ipsec_odrops);
2878 			tdbstat_inc(tdb, tdb_odrops);
2879 			m_freem(m);
2880 			return error;
2881 		}
2882 		ipsec_adjust_mtu(m, tdb->tdb_mtu);
2883 		m_freem(m);
2884 		return EMSGSIZE;
2885 	}
2886 	/* propagate don't fragment for v6-over-v6 */
2887 	if (ip_mtudisc)
2888 		SET(m->m_pkthdr.csum_flags, M_IPV6_DF_OUT);
2889 
2890 	/*
2891 	 * Clear these -- they'll be set in the recursive invocation
2892 	 * as needed.
2893 	 */
2894 	m->m_flags &= ~(M_BCAST | M_MCAST);
2895 
2896 	if (tso) {
2897 		error = tcp_chopper(m, &ml, encif, len);
2898 		if (error)
2899 			goto done;
2900 	} else {
2901 		CLR(m->m_pkthdr.csum_flags, M_TCP_TSO);
2902 		in6_proto_cksum_out(m, encif);
2903 		ml_init(&ml);
2904 		ml_enqueue(&ml, m);
2905 	}
2906 
2907 	KERNEL_LOCK();
2908 	while ((m = ml_dequeue(&ml)) != NULL) {
2909 		/* Callee frees mbuf */
2910 		error = ipsp_process_packet(m, tdb, AF_INET6, tunalready);
2911 		if (error)
2912 			break;
2913 	}
2914 	KERNEL_UNLOCK();
2915  done:
2916 	if (error) {
2917 		ml_purge(&ml);
2918 		ipsecstat_inc(ipsec_odrops);
2919 		tdbstat_inc(tdb, tdb_odrops);
2920 	}
2921 	if (!error && tso)
2922 		tcpstat_inc(tcps_outswtso);
2923 	if (ip_mtudisc && error == EMSGSIZE)
2924 		ip6_output_ipsec_pmtu_update(tdb, ro, &dst, ifidx, rtableid, 0);
2925 	return error;
2926 }
2927 #endif /* IPSEC */
2928