xref: /openbsd/sys/netinet6/ip6_output.c (revision 5dea098c)
1 /*	$OpenBSD: ip6_output.c,v 1.291 2024/04/17 20:48:51 bluhm 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 *,
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 *ro,
164     int flags, struct ip6_moptions *im6o, const struct ipsec_level *seclevel)
165 {
166 	struct ip6_hdr *ip6;
167 	struct ifnet *ifp = NULL;
168 	struct mbuf_list ml;
169 	int hlen, tlen;
170 	struct route iproute;
171 	struct rtentry *rt = NULL;
172 	struct sockaddr_in6 *dst;
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 *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 = &iproute;
394 		ro->ro_rt = NULL;
395 	}
396 	ro_pmtu = ro;
397 	if (opt && opt->ip6po_rthdr)
398 		ro = &opt->ip6po_route;
399 	dst = &ro->ro_dstsin6;
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 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
446 		struct in6_pktinfo *pi = NULL;
447 
448 		/*
449 		 * If the caller specify the outgoing interface
450 		 * explicitly, use it.
451 		 */
452 		if (opt != NULL && (pi = opt->ip6po_pktinfo) != NULL)
453 			ifp = if_get(pi->ipi6_ifindex);
454 
455 		if (ifp == NULL && im6o != NULL)
456 			ifp = if_get(im6o->im6o_ifidx);
457 	}
458 
459 	if (ifp == NULL) {
460 		rt = in6_selectroute(&ip6->ip6_dst, opt, ro,
461 		    m->m_pkthdr.ph_rtableid);
462 		if (rt == NULL) {
463 			ip6stat_inc(ip6s_noroute);
464 			error = EHOSTUNREACH;
465 			goto bad;
466 		}
467 		if (ISSET(rt->rt_flags, RTF_LOCAL))
468 			ifp = if_get(rtable_loindex(m->m_pkthdr.ph_rtableid));
469 		else
470 			ifp = if_get(rt->rt_ifidx);
471 		/*
472 		 * We aren't using rtisvalid() here because the UP/DOWN state
473 		 * machine is broken with some Ethernet drivers like em(4).
474 		 * As a result we might try to use an invalid cached route
475 		 * entry while an interface is being detached.
476 		 */
477 		if (ifp == NULL) {
478 			ip6stat_inc(ip6s_noroute);
479 			error = EHOSTUNREACH;
480 			goto bad;
481 		}
482 	} else {
483 		route6_cache(ro, &ip6->ip6_dst, NULL, m->m_pkthdr.ph_rtableid);
484 	}
485 
486 	if (rt && (rt->rt_flags & RTF_GATEWAY) &&
487 	    !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst))
488 		dst = satosin6(rt->rt_gateway);
489 
490 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
491 		/* Unicast */
492 
493 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
494 	} else {
495 		/* Multicast */
496 
497 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
498 
499 		/*
500 		 * Confirm that the outgoing interface supports multicast.
501 		 */
502 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
503 			ip6stat_inc(ip6s_noroute);
504 			error = ENETUNREACH;
505 			goto bad;
506 		}
507 
508 		if ((im6o == NULL || im6o->im6o_loop) &&
509 		    in6_hasmulti(&ip6->ip6_dst, ifp)) {
510 			/*
511 			 * If we belong to the destination multicast group
512 			 * on the outgoing interface, and the caller did not
513 			 * forbid loopback, loop back a copy.
514 			 * Can't defer TCP/UDP checksumming, do the
515 			 * computation now.
516 			 */
517 			in6_proto_cksum_out(m, NULL);
518 			ip6_mloopback(ifp, m, dst);
519 		}
520 #ifdef MROUTING
521 		else {
522 			/*
523 			 * If we are acting as a multicast router, perform
524 			 * multicast forwarding as if the packet had just
525 			 * arrived on the interface to which we are about
526 			 * to send.  The multicast forwarding function
527 			 * recursively calls this function, using the
528 			 * IPV6_FORWARDING flag to prevent infinite recursion.
529 			 *
530 			 * Multicasts that are looped back by ip6_mloopback(),
531 			 * above, will be forwarded by the ip6_input() routine,
532 			 * if necessary.
533 			 */
534 			if (ip6_mforwarding && ip6_mrouter[ifp->if_rdomain] &&
535 			    (flags & IPV6_FORWARDING) == 0) {
536 				if (ip6_mforward(ip6, ifp, m) != 0) {
537 					m_freem(m);
538 					goto done;
539 				}
540 			}
541 		}
542 #endif
543 		/*
544 		 * Multicasts with a hoplimit of zero may be looped back,
545 		 * above, but must not be transmitted on a network.
546 		 * Also, multicasts addressed to the loopback interface
547 		 * are not sent -- the above call to ip6_mloopback() will
548 		 * loop back a copy if this host actually belongs to the
549 		 * destination group on the loopback interface.
550 		 */
551 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
552 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
553 			m_freem(m);
554 			goto done;
555 		}
556 	}
557 
558 	/*
559 	 * If this packet is going through a loopback interface we won't
560 	 * be able to restore its scope ID using the interface index.
561 	 */
562 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src)) {
563 		if (ifp->if_flags & IFF_LOOPBACK)
564 			src_scope = ip6->ip6_src.s6_addr16[1];
565 		ip6->ip6_src.s6_addr16[1] = 0;
566 	}
567 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) {
568 		if (ifp->if_flags & IFF_LOOPBACK)
569 			dst_scope = ip6->ip6_dst.s6_addr16[1];
570 		ip6->ip6_dst.s6_addr16[1] = 0;
571 	}
572 
573 	/* Determine path MTU. */
574 	if ((error = ip6_getpmtu(ro_pmtu->ro_rt, ifp, &mtu)) != 0)
575 		goto bad;
576 
577 	/*
578 	 * The caller of this function may specify to use the minimum MTU
579 	 * in some cases.
580 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
581 	 * setting.  The logic is a bit complicated; by default, unicast
582 	 * packets will follow path MTU while multicast packets will be sent at
583 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
584 	 * including unicast ones will be sent at the minimum MTU.  Multicast
585 	 * packets will always be sent at the minimum MTU unless
586 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
587 	 * See RFC 3542 for more details.
588 	 */
589 	if (mtu > IPV6_MMTU) {
590 		if ((flags & IPV6_MINMTU))
591 			mtu = IPV6_MMTU;
592 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
593 			mtu = IPV6_MMTU;
594 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) && (opt == NULL ||
595 		    opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
596 			mtu = IPV6_MMTU;
597 		}
598 	}
599 
600 	/*
601 	 * If the outgoing packet contains a hop-by-hop options header,
602 	 * it must be examined and processed even by the source node.
603 	 * (RFC 2460, section 4.)
604 	 */
605 	if (exthdrs.ip6e_hbh) {
606 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
607 		u_int32_t rtalert; /* returned value is ignored */
608 		u_int32_t plen = 0; /* no more than 1 jumbo payload option! */
609 
610 		m->m_pkthdr.ph_ifidx = ifp->if_index;
611 		if (ip6_process_hopopts(&m, (u_int8_t *)(hbh + 1),
612 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
613 		    &rtalert, &plen) < 0) {
614 			/* m was already freed at this point */
615 			error = EINVAL;/* better error? */
616 			goto done;
617 		}
618 		m->m_pkthdr.ph_ifidx = 0;
619 	}
620 
621 #if NPF > 0
622 	if (pf_test(AF_INET6, PF_OUT, ifp, &m) != PF_PASS) {
623 		error = EACCES;
624 		m_freem(m);
625 		goto done;
626 	}
627 	if (m == NULL)
628 		goto done;
629 	ip6 = mtod(m, struct ip6_hdr *);
630 	if ((m->m_pkthdr.pf.flags & (PF_TAG_REROUTE | PF_TAG_GENERATED)) ==
631 	    (PF_TAG_REROUTE | PF_TAG_GENERATED)) {
632 		/* already rerun the route lookup, go on */
633 		m->m_pkthdr.pf.flags &= ~(PF_TAG_GENERATED | PF_TAG_REROUTE);
634 	} else if (m->m_pkthdr.pf.flags & PF_TAG_REROUTE) {
635 		/* tag as generated to skip over pf_test on rerun */
636 		m->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
637 		finaldst = ip6->ip6_dst;
638 		if (ro == &iproute)
639 			rtfree(ro->ro_rt);
640 		ro = NULL;
641 		if_put(ifp); /* drop reference since destination changed */
642 		ifp = NULL;
643 		goto reroute;
644 	}
645 #endif
646 
647 	/*
648 	 * If the packet is not going on the wire it can be destined
649 	 * to any local address.  In this case do not clear its scopes
650 	 * to let ip6_input() find a matching local route.
651 	 */
652 	if (ifp->if_flags & IFF_LOOPBACK) {
653 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
654 			ip6->ip6_src.s6_addr16[1] = src_scope;
655 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
656 			ip6->ip6_dst.s6_addr16[1] = dst_scope;
657 	}
658 
659 	/*
660 	 * Send the packet to the outgoing interface.
661 	 * If necessary, do IPv6 fragmentation before sending.
662 	 *
663 	 * the logic here is rather complex:
664 	 * 1: normal case (dontfrag == 0)
665 	 * 1-a: send as is if tlen <= path mtu
666 	 * 1-b: fragment if tlen > path mtu
667 	 *
668 	 * 2: if user asks us not to fragment (dontfrag == 1)
669 	 * 2-a: send as is if tlen <= interface mtu
670 	 * 2-b: error if tlen > interface mtu
671 	 */
672 	tlen = ISSET(m->m_pkthdr.csum_flags, M_TCP_TSO) ?
673 	    m->m_pkthdr.ph_mss : m->m_pkthdr.len;
674 
675 	if (ISSET(m->m_pkthdr.csum_flags, M_IPV6_DF_OUT)) {
676 		CLR(m->m_pkthdr.csum_flags, M_IPV6_DF_OUT);
677 		dontfrag = 1;
678 	} else if (opt && ISSET(opt->ip6po_flags, IP6PO_DONTFRAG))
679 		dontfrag = 1;
680 	else
681 		dontfrag = 0;
682 
683 	if (dontfrag && tlen > ifp->if_mtu) {		/* case 2-b */
684 #ifdef IPSEC
685 		if (ip_mtudisc)
686 			ipsec_adjust_mtu(m, mtu);
687 #endif
688 		error = EMSGSIZE;
689 		goto bad;
690 	}
691 
692 	/*
693 	 * transmit packet without fragmentation
694 	 */
695 	if (dontfrag || tlen <= mtu) {			/* case 1-a and 2-a */
696 		error = if_output_tso(ifp, &m, sin6tosa(dst), ro->ro_rt,
697 		    ifp->if_mtu);
698 		if (error || m == NULL)
699 			goto done;
700 		goto bad;				/* should not happen */
701 	}
702 
703 	/*
704 	 * try to fragment the packet.  case 1-b
705 	 */
706 	if (mtu < IPV6_MMTU) {
707 		/* path MTU cannot be less than IPV6_MMTU */
708 		error = EMSGSIZE;
709 		goto bad;
710 	} else if (ip6->ip6_plen == 0) {
711 		/* jumbo payload cannot be fragmented */
712 		error = EMSGSIZE;
713 		goto bad;
714 	}
715 
716 	/*
717 	 * Too large for the destination or interface;
718 	 * fragment if possible.
719 	 * Must be able to put at least 8 bytes per fragment.
720 	 */
721 	hlen = unfragpartlen;
722 	if (mtu > IPV6_MAXPACKET)
723 		mtu = IPV6_MAXPACKET;
724 
725 	/*
726 	 * If we are doing fragmentation, we can't defer TCP/UDP
727 	 * checksumming; compute the checksum and clear the flag.
728 	 */
729         in6_proto_cksum_out(m, NULL);
730 
731 	/*
732 	 * Change the next header field of the last header in the
733 	 * unfragmentable part.
734 	 */
735 	if (exthdrs.ip6e_rthdr) {
736 		nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
737 		*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
738 	} else if (exthdrs.ip6e_dest1) {
739 		nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
740 		*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
741 	} else if (exthdrs.ip6e_hbh) {
742 		nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
743 		*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
744 	} else {
745 		nextproto = ip6->ip6_nxt;
746 		ip6->ip6_nxt = IPPROTO_FRAGMENT;
747 	}
748 
749 	if ((error = ip6_fragment(m, &ml, hlen, nextproto, mtu)) ||
750 	    (error = if_output_ml(ifp, &ml, sin6tosa(dst), ro->ro_rt)))
751 		goto done;
752 	ip6stat_inc(ip6s_fragmented);
753 	goto done;
754 
755  freehdrs:
756 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
757 	m_freem(exthdrs.ip6e_dest1);
758 	m_freem(exthdrs.ip6e_rthdr);
759 	m_freem(exthdrs.ip6e_dest2);
760  bad:
761 	m_freem(m);
762  done:
763 	if (ro == &iproute)
764 		rtfree(ro->ro_rt);
765 	else if (ro_pmtu == &iproute)
766 		rtfree(ro_pmtu->ro_rt);
767 	if_put(ifp);
768 #ifdef IPSEC
769 	tdb_unref(tdb);
770 #endif /* IPSEC */
771 	return (error);
772 }
773 
774 int
775 ip6_fragment(struct mbuf *m0, struct mbuf_list *ml, int hlen, u_char nextproto,
776     u_long mtu)
777 {
778 	struct ip6_hdr *ip6;
779 	u_int32_t id;
780 	int tlen, len, off;
781 	int error;
782 
783 	ml_init(ml);
784 
785 	ip6 = mtod(m0, struct ip6_hdr *);
786 	tlen = m0->m_pkthdr.len;
787 	len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
788 	if (len < 8) {
789 		error = EMSGSIZE;
790 		goto bad;
791 	}
792 	id = htonl(ip6_randomid());
793 
794 	/*
795 	 * Loop through length of payload,
796 	 * make new header and copy data of each part and link onto chain.
797 	 */
798 	for (off = hlen; off < tlen; off += len) {
799 		struct mbuf *m;
800 		struct mbuf *mlast;
801 		struct ip6_hdr *mhip6;
802 		struct ip6_frag *ip6f;
803 
804 		MGETHDR(m, M_DONTWAIT, MT_HEADER);
805 		if (m == NULL) {
806 			error = ENOBUFS;
807 			goto bad;
808 		}
809 		ml_enqueue(ml, m);
810 		if ((error = m_dup_pkthdr(m, m0, M_DONTWAIT)) != 0)
811 			goto bad;
812 		m->m_data += max_linkhdr;
813 		mhip6 = mtod(m, struct ip6_hdr *);
814 		*mhip6 = *ip6;
815 		m->m_len = sizeof(struct ip6_hdr);
816 
817 		if ((error = ip6_insertfraghdr(m0, m, hlen, &ip6f)) != 0)
818 			goto bad;
819 		ip6f->ip6f_offlg = htons((off - hlen) & ~7);
820 		if (off + len >= tlen)
821 			len = tlen - off;
822 		else
823 			ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
824 
825 		m->m_pkthdr.len = hlen + sizeof(struct ip6_frag) + len;
826 		mhip6->ip6_plen = htons(m->m_pkthdr.len -
827 		    sizeof(struct ip6_hdr));
828 		for (mlast = m; mlast->m_next; mlast = mlast->m_next)
829 			;
830 		mlast->m_next = m_copym(m0, off, len, M_DONTWAIT);
831 		if (mlast->m_next == NULL) {
832 			error = ENOBUFS;
833 			goto bad;
834 		}
835 
836 		ip6f->ip6f_reserved = 0;
837 		ip6f->ip6f_ident = id;
838 		ip6f->ip6f_nxt = nextproto;
839 	}
840 
841 	ip6stat_add(ip6s_ofragments, ml_len(ml));
842 	m_freem(m0);
843 	return (0);
844 
845 bad:
846 	ip6stat_inc(ip6s_odropped);
847 	ml_purge(ml);
848 	m_freem(m0);
849 	return (error);
850 }
851 
852 int
853 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
854 {
855 	struct mbuf *m;
856 
857 	if (hlen > MCLBYTES)
858 		return (ENOBUFS); /* XXX */
859 
860 	MGET(m, M_DONTWAIT, MT_DATA);
861 	if (!m)
862 		return (ENOBUFS);
863 
864 	if (hlen > MLEN) {
865 		MCLGET(m, M_DONTWAIT);
866 		if ((m->m_flags & M_EXT) == 0) {
867 			m_free(m);
868 			return (ENOBUFS);
869 		}
870 	}
871 	m->m_len = hlen;
872 	if (hdr)
873 		memcpy(mtod(m, caddr_t), hdr, hlen);
874 
875 	*mp = m;
876 	return (0);
877 }
878 
879 /*
880  * Insert jumbo payload option.
881  */
882 int
883 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
884 {
885 	struct mbuf *mopt;
886 	u_int8_t *optbuf;
887 	u_int32_t v;
888 
889 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
890 
891 	/*
892 	 * If there is no hop-by-hop options header, allocate new one.
893 	 * If there is one but it doesn't have enough space to store the
894 	 * jumbo payload option, allocate a cluster to store the whole options.
895 	 * Otherwise, use it to store the options.
896 	 */
897 	if (exthdrs->ip6e_hbh == 0) {
898 		MGET(mopt, M_DONTWAIT, MT_DATA);
899 		if (mopt == NULL)
900 			return (ENOBUFS);
901 		mopt->m_len = JUMBOOPTLEN;
902 		optbuf = mtod(mopt, u_int8_t *);
903 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
904 		exthdrs->ip6e_hbh = mopt;
905 	} else {
906 		struct ip6_hbh *hbh;
907 
908 		mopt = exthdrs->ip6e_hbh;
909 		if (m_trailingspace(mopt) < JUMBOOPTLEN) {
910 			/*
911 			 * XXX assumption:
912 			 * - exthdrs->ip6e_hbh is not referenced from places
913 			 *   other than exthdrs.
914 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
915 			 */
916 			int oldoptlen = mopt->m_len;
917 			struct mbuf *n;
918 
919 			/*
920 			 * XXX: give up if the whole (new) hbh header does
921 			 * not fit even in an mbuf cluster.
922 			 */
923 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
924 				return (ENOBUFS);
925 
926 			/*
927 			 * As a consequence, we must always prepare a cluster
928 			 * at this point.
929 			 */
930 			MGET(n, M_DONTWAIT, MT_DATA);
931 			if (n) {
932 				MCLGET(n, M_DONTWAIT);
933 				if ((n->m_flags & M_EXT) == 0) {
934 					m_freem(n);
935 					n = NULL;
936 				}
937 			}
938 			if (!n)
939 				return (ENOBUFS);
940 			n->m_len = oldoptlen + JUMBOOPTLEN;
941 			memcpy(mtod(n, caddr_t), mtod(mopt, caddr_t),
942 			      oldoptlen);
943 			optbuf = mtod(n, u_int8_t *) + oldoptlen;
944 			m_freem(mopt);
945 			mopt = exthdrs->ip6e_hbh = n;
946 		} else {
947 			optbuf = mtod(mopt, u_int8_t *) + mopt->m_len;
948 			mopt->m_len += JUMBOOPTLEN;
949 		}
950 		optbuf[0] = IP6OPT_PADN;
951 		optbuf[1] = 0;
952 
953 		/*
954 		 * Adjust the header length according to the pad and
955 		 * the jumbo payload option.
956 		 */
957 		hbh = mtod(mopt, struct ip6_hbh *);
958 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
959 	}
960 
961 	/* fill in the option. */
962 	optbuf[2] = IP6OPT_JUMBO;
963 	optbuf[3] = 4;
964 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
965 	memcpy(&optbuf[4], &v, sizeof(u_int32_t));
966 
967 	/* finally, adjust the packet header length */
968 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
969 
970 	return (0);
971 #undef JUMBOOPTLEN
972 }
973 
974 /*
975  * Insert fragment header and copy unfragmentable header portions.
976  */
977 int
978 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
979     struct ip6_frag **frghdrp)
980 {
981 	struct mbuf *n, *mlast;
982 
983 	if (hlen > sizeof(struct ip6_hdr)) {
984 		n = m_copym(m0, sizeof(struct ip6_hdr),
985 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
986 		if (n == NULL)
987 			return (ENOBUFS);
988 		m->m_next = n;
989 	} else
990 		n = m;
991 
992 	/* Search for the last mbuf of unfragmentable part. */
993 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
994 		;
995 
996 	if ((mlast->m_flags & M_EXT) == 0 &&
997 	    m_trailingspace(mlast) >= sizeof(struct ip6_frag)) {
998 		/* use the trailing space of the last mbuf for fragment hdr */
999 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1000 		    mlast->m_len);
1001 		mlast->m_len += sizeof(struct ip6_frag);
1002 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1003 	} else {
1004 		/* allocate a new mbuf for the fragment header */
1005 		struct mbuf *mfrg;
1006 
1007 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1008 		if (mfrg == NULL)
1009 			return (ENOBUFS);
1010 		mfrg->m_len = sizeof(struct ip6_frag);
1011 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1012 		mlast->m_next = mfrg;
1013 	}
1014 
1015 	return (0);
1016 }
1017 
1018 int
1019 ip6_getpmtu(struct rtentry *rt, struct ifnet *ifp, u_long *mtup)
1020 {
1021 	u_int32_t mtu = 0;
1022 	int error = 0;
1023 
1024 	if (rt != NULL) {
1025 		mtu = rt->rt_mtu;
1026 		if (mtu == 0)
1027 			mtu = ifp->if_mtu;
1028 		else if (mtu < IPV6_MMTU) {
1029 			/* RFC8021 IPv6 Atomic Fragments Considered Harmful */
1030 			mtu = IPV6_MMTU;
1031 		} else if (mtu > ifp->if_mtu) {
1032 			/*
1033 			 * The MTU on the route is larger than the MTU on
1034 			 * the interface!  This shouldn't happen, unless the
1035 			 * MTU of the interface has been changed after the
1036 			 * interface was brought up.  Change the MTU in the
1037 			 * route to match the interface MTU (as long as the
1038 			 * field isn't locked).
1039 			 */
1040 			mtu = ifp->if_mtu;
1041 			if (!(rt->rt_locks & RTV_MTU))
1042 				rt->rt_mtu = mtu;
1043 		}
1044 	} else {
1045 		mtu = ifp->if_mtu;
1046 	}
1047 
1048 	*mtup = mtu;
1049 	return (error);
1050 }
1051 
1052 /*
1053  * IP6 socket option processing.
1054  */
1055 int
1056 ip6_ctloutput(int op, struct socket *so, int level, int optname,
1057     struct mbuf *m)
1058 {
1059 	int privileged, optdatalen, uproto;
1060 	void *optdata;
1061 	struct inpcb *inp = sotoinpcb(so);
1062 	int error, optval;
1063 	struct proc *p = curproc; /* For IPsec and rdomain */
1064 	u_int rtableid, rtid = 0;
1065 
1066 	error = optval = 0;
1067 
1068 	privileged = (inp->inp_socket->so_state & SS_PRIV);
1069 	uproto = (int)so->so_proto->pr_protocol;
1070 
1071 	if (level != IPPROTO_IPV6)
1072 		return (EINVAL);
1073 
1074 	rtableid = p->p_p->ps_rtableid;
1075 
1076 	switch (op) {
1077 	case PRCO_SETOPT:
1078 		switch (optname) {
1079 		/*
1080 		 * Use of some Hop-by-Hop options or some
1081 		 * Destination options, might require special
1082 		 * privilege.  That is, normal applications
1083 		 * (without special privilege) might be forbidden
1084 		 * from setting certain options in outgoing packets,
1085 		 * and might never see certain options in received
1086 		 * packets. [RFC 2292 Section 6]
1087 		 * KAME specific note:
1088 		 *  KAME prevents non-privileged users from sending or
1089 		 *  receiving ANY hbh/dst options in order to avoid
1090 		 *  overhead of parsing options in the kernel.
1091 		 */
1092 		case IPV6_RECVHOPOPTS:
1093 		case IPV6_RECVDSTOPTS:
1094 			if (!privileged) {
1095 				error = EPERM;
1096 				break;
1097 			}
1098 			/* FALLTHROUGH */
1099 		case IPV6_UNICAST_HOPS:
1100 		case IPV6_MINHOPCOUNT:
1101 		case IPV6_HOPLIMIT:
1102 
1103 		case IPV6_RECVPKTINFO:
1104 		case IPV6_RECVHOPLIMIT:
1105 		case IPV6_RECVRTHDR:
1106 		case IPV6_RECVPATHMTU:
1107 		case IPV6_RECVTCLASS:
1108 		case IPV6_V6ONLY:
1109 		case IPV6_AUTOFLOWLABEL:
1110 		case IPV6_RECVDSTPORT:
1111 			if (m == NULL || m->m_len != sizeof(int)) {
1112 				error = EINVAL;
1113 				break;
1114 			}
1115 			optval = *mtod(m, int *);
1116 			switch (optname) {
1117 
1118 			case IPV6_UNICAST_HOPS:
1119 				if (optval < -1 || optval >= 256)
1120 					error = EINVAL;
1121 				else {
1122 					/* -1 = kernel default */
1123 					inp->inp_hops = optval;
1124 				}
1125 				break;
1126 
1127 			case IPV6_MINHOPCOUNT:
1128 				if (optval < 0 || optval > 255)
1129 					error = EINVAL;
1130 				else
1131 					inp->inp_ip6_minhlim = optval;
1132 				break;
1133 
1134 #define OPTSET(bit) \
1135 do { \
1136 	if (optval) \
1137 		inp->inp_flags |= (bit); \
1138 	else \
1139 		inp->inp_flags &= ~(bit); \
1140 } while (/*CONSTCOND*/ 0)
1141 #define OPTBIT(bit) (inp->inp_flags & (bit) ? 1 : 0)
1142 
1143 			case IPV6_RECVPKTINFO:
1144 				OPTSET(IN6P_PKTINFO);
1145 				break;
1146 
1147 			case IPV6_HOPLIMIT:
1148 			{
1149 				struct ip6_pktopts **optp;
1150 
1151 				optp = &inp->inp_outputopts6;
1152 				error = ip6_pcbopt(IPV6_HOPLIMIT,
1153 				    (u_char *)&optval, sizeof(optval), optp,
1154 				    privileged, uproto);
1155 				break;
1156 			}
1157 
1158 			case IPV6_RECVHOPLIMIT:
1159 				OPTSET(IN6P_HOPLIMIT);
1160 				break;
1161 
1162 			case IPV6_RECVHOPOPTS:
1163 				OPTSET(IN6P_HOPOPTS);
1164 				break;
1165 
1166 			case IPV6_RECVDSTOPTS:
1167 				OPTSET(IN6P_DSTOPTS);
1168 				break;
1169 
1170 			case IPV6_RECVRTHDR:
1171 				OPTSET(IN6P_RTHDR);
1172 				break;
1173 
1174 			case IPV6_RECVPATHMTU:
1175 				/*
1176 				 * We ignore this option for TCP
1177 				 * sockets.
1178 				 * (RFC3542 leaves this case
1179 				 * unspecified.)
1180 				 */
1181 				if (uproto != IPPROTO_TCP)
1182 					OPTSET(IN6P_MTU);
1183 				break;
1184 
1185 			case IPV6_V6ONLY:
1186 				/*
1187 				 * make setsockopt(IPV6_V6ONLY)
1188 				 * available only prior to bind(2).
1189 				 * see ipng mailing list, Jun 22 2001.
1190 				 */
1191 				if (inp->inp_lport || !IN6_IS_ADDR_UNSPECIFIED(
1192 				    &inp->inp_laddr6)) {
1193 					error = EINVAL;
1194 					break;
1195 				}
1196 				/* No support for IPv4-mapped addresses. */
1197 				if (!optval)
1198 					error = EINVAL;
1199 				else
1200 					error = 0;
1201 				break;
1202 			case IPV6_RECVTCLASS:
1203 				OPTSET(IN6P_TCLASS);
1204 				break;
1205 			case IPV6_AUTOFLOWLABEL:
1206 				OPTSET(IN6P_AUTOFLOWLABEL);
1207 				break;
1208 
1209 			case IPV6_RECVDSTPORT:
1210 				OPTSET(IN6P_RECVDSTPORT);
1211 				break;
1212 			}
1213 			break;
1214 
1215 		case IPV6_TCLASS:
1216 		case IPV6_DONTFRAG:
1217 		case IPV6_USE_MIN_MTU:
1218 			if (m == NULL || m->m_len != sizeof(optval)) {
1219 				error = EINVAL;
1220 				break;
1221 			}
1222 			optval = *mtod(m, int *);
1223 			{
1224 				struct ip6_pktopts **optp;
1225 				optp = &inp->inp_outputopts6;
1226 				error = ip6_pcbopt(optname, (u_char *)&optval,
1227 				    sizeof(optval), optp, privileged, uproto);
1228 				break;
1229 			}
1230 
1231 		case IPV6_PKTINFO:
1232 		case IPV6_HOPOPTS:
1233 		case IPV6_RTHDR:
1234 		case IPV6_DSTOPTS:
1235 		case IPV6_RTHDRDSTOPTS:
1236 		{
1237 			/* new advanced API (RFC3542) */
1238 			u_char *optbuf;
1239 			int optbuflen;
1240 			struct ip6_pktopts **optp;
1241 
1242 			if (m && m->m_next) {
1243 				error = EINVAL;	/* XXX */
1244 				break;
1245 			}
1246 			if (m) {
1247 				optbuf = mtod(m, u_char *);
1248 				optbuflen = m->m_len;
1249 			} else {
1250 				optbuf = NULL;
1251 				optbuflen = 0;
1252 			}
1253 			optp = &inp->inp_outputopts6;
1254 			error = ip6_pcbopt(optname, optbuf, optbuflen, optp,
1255 			    privileged, uproto);
1256 			break;
1257 		}
1258 #undef OPTSET
1259 
1260 		case IPV6_MULTICAST_IF:
1261 		case IPV6_MULTICAST_HOPS:
1262 		case IPV6_MULTICAST_LOOP:
1263 		case IPV6_JOIN_GROUP:
1264 		case IPV6_LEAVE_GROUP:
1265 			error =	ip6_setmoptions(optname,
1266 						&inp->inp_moptions6,
1267 						m, inp->inp_rtableid);
1268 			break;
1269 
1270 		case IPV6_PORTRANGE:
1271 			if (m == NULL || m->m_len != sizeof(int)) {
1272 				error = EINVAL;
1273 				break;
1274 			}
1275 			optval = *mtod(m, int *);
1276 
1277 			switch (optval) {
1278 			case IPV6_PORTRANGE_DEFAULT:
1279 				inp->inp_flags &= ~(IN6P_LOWPORT);
1280 				inp->inp_flags &= ~(IN6P_HIGHPORT);
1281 				break;
1282 
1283 			case IPV6_PORTRANGE_HIGH:
1284 				inp->inp_flags &= ~(IN6P_LOWPORT);
1285 				inp->inp_flags |= IN6P_HIGHPORT;
1286 				break;
1287 
1288 			case IPV6_PORTRANGE_LOW:
1289 				inp->inp_flags &= ~(IN6P_HIGHPORT);
1290 				inp->inp_flags |= IN6P_LOWPORT;
1291 				break;
1292 
1293 			default:
1294 				error = EINVAL;
1295 				break;
1296 			}
1297 			break;
1298 
1299 		case IPSEC6_OUTSA:
1300 			error = EINVAL;
1301 			break;
1302 
1303 		case IPV6_AUTH_LEVEL:
1304 		case IPV6_ESP_TRANS_LEVEL:
1305 		case IPV6_ESP_NETWORK_LEVEL:
1306 		case IPV6_IPCOMP_LEVEL:
1307 #ifndef IPSEC
1308 			error = EINVAL;
1309 #else
1310 			if (m == NULL || m->m_len != sizeof(int)) {
1311 				error = EINVAL;
1312 				break;
1313 			}
1314 			optval = *mtod(m, int *);
1315 
1316 			if (optval < IPSEC_LEVEL_BYPASS ||
1317 			    optval > IPSEC_LEVEL_UNIQUE) {
1318 				error = EINVAL;
1319 				break;
1320 			}
1321 
1322 			switch (optname) {
1323 			case IPV6_AUTH_LEVEL:
1324 				if (optval < IPSEC_AUTH_LEVEL_DEFAULT &&
1325 				    suser(p)) {
1326 					error = EACCES;
1327 					break;
1328 				}
1329 				inp->inp_seclevel.sl_auth = optval;
1330 				break;
1331 
1332 			case IPV6_ESP_TRANS_LEVEL:
1333 				if (optval < IPSEC_ESP_TRANS_LEVEL_DEFAULT &&
1334 				    suser(p)) {
1335 					error = EACCES;
1336 					break;
1337 				}
1338 				inp->inp_seclevel.sl_esp_trans = optval;
1339 				break;
1340 
1341 			case IPV6_ESP_NETWORK_LEVEL:
1342 				if (optval < IPSEC_ESP_NETWORK_LEVEL_DEFAULT &&
1343 				    suser(p)) {
1344 					error = EACCES;
1345 					break;
1346 				}
1347 				inp->inp_seclevel.sl_esp_network = optval;
1348 				break;
1349 
1350 			case IPV6_IPCOMP_LEVEL:
1351 				if (optval < IPSEC_IPCOMP_LEVEL_DEFAULT &&
1352 				    suser(p)) {
1353 					error = EACCES;
1354 					break;
1355 				}
1356 				inp->inp_seclevel.sl_ipcomp = optval;
1357 				break;
1358 			}
1359 #endif
1360 			break;
1361 		case SO_RTABLE:
1362 			if (m == NULL || m->m_len < sizeof(u_int)) {
1363 				error = EINVAL;
1364 				break;
1365 			}
1366 			rtid = *mtod(m, u_int *);
1367 			if (inp->inp_rtableid == rtid)
1368 				break;
1369 			/* needs privileges to switch when already set */
1370 			if (rtableid != rtid && rtableid != 0 &&
1371 			    (error = suser(p)) != 0)
1372 				break;
1373 			error = in_pcbset_rtableid(inp, rtid);
1374 			break;
1375 		case IPV6_PIPEX:
1376 			if (m != NULL && m->m_len == sizeof(int))
1377 				inp->inp_pipex = *mtod(m, int *);
1378 			else
1379 				error = EINVAL;
1380 			break;
1381 
1382 		default:
1383 			error = ENOPROTOOPT;
1384 			break;
1385 		}
1386 		break;
1387 
1388 	case PRCO_GETOPT:
1389 		switch (optname) {
1390 
1391 		case IPV6_RECVHOPOPTS:
1392 		case IPV6_RECVDSTOPTS:
1393 		case IPV6_UNICAST_HOPS:
1394 		case IPV6_MINHOPCOUNT:
1395 		case IPV6_RECVPKTINFO:
1396 		case IPV6_RECVHOPLIMIT:
1397 		case IPV6_RECVRTHDR:
1398 		case IPV6_RECVPATHMTU:
1399 
1400 		case IPV6_V6ONLY:
1401 		case IPV6_PORTRANGE:
1402 		case IPV6_RECVTCLASS:
1403 		case IPV6_AUTOFLOWLABEL:
1404 		case IPV6_RECVDSTPORT:
1405 			switch (optname) {
1406 
1407 			case IPV6_RECVHOPOPTS:
1408 				optval = OPTBIT(IN6P_HOPOPTS);
1409 				break;
1410 
1411 			case IPV6_RECVDSTOPTS:
1412 				optval = OPTBIT(IN6P_DSTOPTS);
1413 				break;
1414 
1415 			case IPV6_UNICAST_HOPS:
1416 				optval = inp->inp_hops;
1417 				break;
1418 
1419 			case IPV6_MINHOPCOUNT:
1420 				optval = inp->inp_ip6_minhlim;
1421 				break;
1422 
1423 			case IPV6_RECVPKTINFO:
1424 				optval = OPTBIT(IN6P_PKTINFO);
1425 				break;
1426 
1427 			case IPV6_RECVHOPLIMIT:
1428 				optval = OPTBIT(IN6P_HOPLIMIT);
1429 				break;
1430 
1431 			case IPV6_RECVRTHDR:
1432 				optval = OPTBIT(IN6P_RTHDR);
1433 				break;
1434 
1435 			case IPV6_RECVPATHMTU:
1436 				optval = OPTBIT(IN6P_MTU);
1437 				break;
1438 
1439 			case IPV6_V6ONLY:
1440 				optval = 1;
1441 				break;
1442 
1443 			case IPV6_PORTRANGE:
1444 			    {
1445 				int flags;
1446 				flags = inp->inp_flags;
1447 				if (flags & IN6P_HIGHPORT)
1448 					optval = IPV6_PORTRANGE_HIGH;
1449 				else if (flags & IN6P_LOWPORT)
1450 					optval = IPV6_PORTRANGE_LOW;
1451 				else
1452 					optval = 0;
1453 				break;
1454 			    }
1455 			case IPV6_RECVTCLASS:
1456 				optval = OPTBIT(IN6P_TCLASS);
1457 				break;
1458 
1459 			case IPV6_AUTOFLOWLABEL:
1460 				optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1461 				break;
1462 
1463 			case IPV6_RECVDSTPORT:
1464 				optval = OPTBIT(IN6P_RECVDSTPORT);
1465 				break;
1466 			}
1467 			if (error)
1468 				break;
1469 			m->m_len = sizeof(int);
1470 			*mtod(m, int *) = optval;
1471 			break;
1472 
1473 		case IPV6_PATHMTU:
1474 		{
1475 			u_long pmtu = 0;
1476 			struct ip6_mtuinfo mtuinfo;
1477 			struct ifnet *ifp;
1478 			struct rtentry *rt;
1479 
1480 			if (!(so->so_state & SS_ISCONNECTED))
1481 				return (ENOTCONN);
1482 
1483 			rt = in6_pcbrtentry(inp);
1484 			if (!rtisvalid(rt))
1485 				return (EHOSTUNREACH);
1486 
1487 			ifp = if_get(rt->rt_ifidx);
1488 			if (ifp == NULL)
1489 				return (EHOSTUNREACH);
1490 			/*
1491 			 * XXX: we dot not consider the case of source
1492 			 * routing, or optional information to specify
1493 			 * the outgoing interface.
1494 			 */
1495 			error = ip6_getpmtu(rt, ifp, &pmtu);
1496 			if_put(ifp);
1497 			if (error)
1498 				break;
1499 			if (pmtu > IPV6_MAXPACKET)
1500 				pmtu = IPV6_MAXPACKET;
1501 
1502 			bzero(&mtuinfo, sizeof(mtuinfo));
1503 			mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1504 			optdata = (void *)&mtuinfo;
1505 			optdatalen = sizeof(mtuinfo);
1506 			if (optdatalen > MCLBYTES)
1507 				return (EMSGSIZE); /* XXX */
1508 			if (optdatalen > MLEN)
1509 				MCLGET(m, M_WAIT);
1510 			m->m_len = optdatalen;
1511 			bcopy(optdata, mtod(m, void *), optdatalen);
1512 			break;
1513 		}
1514 
1515 		case IPV6_PKTINFO:
1516 		case IPV6_HOPOPTS:
1517 		case IPV6_RTHDR:
1518 		case IPV6_DSTOPTS:
1519 		case IPV6_RTHDRDSTOPTS:
1520 		case IPV6_TCLASS:
1521 		case IPV6_DONTFRAG:
1522 		case IPV6_USE_MIN_MTU:
1523 			error = ip6_getpcbopt(inp->inp_outputopts6,
1524 			    optname, m);
1525 			break;
1526 
1527 		case IPV6_MULTICAST_IF:
1528 		case IPV6_MULTICAST_HOPS:
1529 		case IPV6_MULTICAST_LOOP:
1530 		case IPV6_JOIN_GROUP:
1531 		case IPV6_LEAVE_GROUP:
1532 			error = ip6_getmoptions(optname,
1533 			    inp->inp_moptions6, m);
1534 			break;
1535 
1536 		case IPSEC6_OUTSA:
1537 			error = EINVAL;
1538 			break;
1539 
1540 		case IPV6_AUTH_LEVEL:
1541 		case IPV6_ESP_TRANS_LEVEL:
1542 		case IPV6_ESP_NETWORK_LEVEL:
1543 		case IPV6_IPCOMP_LEVEL:
1544 #ifndef IPSEC
1545 			m->m_len = sizeof(int);
1546 			*mtod(m, int *) = IPSEC_LEVEL_NONE;
1547 #else
1548 			m->m_len = sizeof(int);
1549 			switch (optname) {
1550 			case IPV6_AUTH_LEVEL:
1551 				optval = inp->inp_seclevel.sl_auth;
1552 				break;
1553 
1554 			case IPV6_ESP_TRANS_LEVEL:
1555 				optval =
1556 				    inp->inp_seclevel.sl_esp_trans;
1557 				break;
1558 
1559 			case IPV6_ESP_NETWORK_LEVEL:
1560 				optval =
1561 				    inp->inp_seclevel.sl_esp_network;
1562 				break;
1563 
1564 			case IPV6_IPCOMP_LEVEL:
1565 				optval = inp->inp_seclevel.sl_ipcomp;
1566 				break;
1567 			}
1568 			*mtod(m, int *) = optval;
1569 #endif
1570 			break;
1571 		case SO_RTABLE:
1572 			m->m_len = sizeof(u_int);
1573 			*mtod(m, u_int *) = inp->inp_rtableid;
1574 			break;
1575 		case IPV6_PIPEX:
1576 			m->m_len = sizeof(int);
1577 			*mtod(m, int *) = inp->inp_pipex;
1578 			break;
1579 
1580 		default:
1581 			error = ENOPROTOOPT;
1582 			break;
1583 		}
1584 		break;
1585 	}
1586 	return (error);
1587 }
1588 
1589 int
1590 ip6_raw_ctloutput(int op, struct socket *so, int level, int optname,
1591     struct mbuf *m)
1592 {
1593 	int error = 0, optval;
1594 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
1595 	struct inpcb *inp = sotoinpcb(so);
1596 
1597 	if (level != IPPROTO_IPV6)
1598 		return (EINVAL);
1599 
1600 	switch (optname) {
1601 	case IPV6_CHECKSUM:
1602 		/*
1603 		 * For ICMPv6 sockets, no modification allowed for checksum
1604 		 * offset, permit "no change" values to help existing apps.
1605 		 *
1606 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
1607 		 * for an ICMPv6 socket will fail."
1608 		 * The current behavior does not meet RFC3542.
1609 		 */
1610 		switch (op) {
1611 		case PRCO_SETOPT:
1612 			if (m == NULL || m->m_len != sizeof(int)) {
1613 				error = EINVAL;
1614 				break;
1615 			}
1616 			optval = *mtod(m, int *);
1617 			if (optval < -1 ||
1618 			    (optval > 0 && (optval % 2) != 0)) {
1619 				/*
1620 				 * The API assumes non-negative even offset
1621 				 * values or -1 as a special value.
1622 				 */
1623 				error = EINVAL;
1624 			} else if (so->so_proto->pr_protocol ==
1625 			    IPPROTO_ICMPV6) {
1626 				if (optval != icmp6off)
1627 					error = EINVAL;
1628 			} else
1629 				inp->inp_cksum6 = optval;
1630 			break;
1631 
1632 		case PRCO_GETOPT:
1633 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
1634 				optval = icmp6off;
1635 			else
1636 				optval = inp->inp_cksum6;
1637 
1638 			m->m_len = sizeof(int);
1639 			*mtod(m, int *) = optval;
1640 			break;
1641 
1642 		default:
1643 			error = EINVAL;
1644 			break;
1645 		}
1646 		break;
1647 
1648 	default:
1649 		error = ENOPROTOOPT;
1650 		break;
1651 	}
1652 
1653 	return (error);
1654 }
1655 
1656 /*
1657  * initialize ip6_pktopts.  beware that there are non-zero default values in
1658  * the struct.
1659  */
1660 void
1661 ip6_initpktopts(struct ip6_pktopts *opt)
1662 {
1663 	bzero(opt, sizeof(*opt));
1664 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
1665 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
1666 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
1667 }
1668 
1669 int
1670 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
1671     int priv, int uproto)
1672 {
1673 	struct ip6_pktopts *opt;
1674 
1675 	if (*pktopt == NULL) {
1676 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
1677 		    M_WAITOK);
1678 		ip6_initpktopts(*pktopt);
1679 	}
1680 	opt = *pktopt;
1681 
1682 	return (ip6_setpktopt(optname, buf, len, opt, priv, 1, uproto));
1683 }
1684 
1685 int
1686 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct mbuf *m)
1687 {
1688 	void *optdata = NULL;
1689 	int optdatalen = 0;
1690 	struct ip6_ext *ip6e;
1691 	int error = 0;
1692 	struct in6_pktinfo null_pktinfo;
1693 	int deftclass = 0, on;
1694 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
1695 
1696 	switch (optname) {
1697 	case IPV6_PKTINFO:
1698 		if (pktopt && pktopt->ip6po_pktinfo)
1699 			optdata = (void *)pktopt->ip6po_pktinfo;
1700 		else {
1701 			/* XXX: we don't have to do this every time... */
1702 			bzero(&null_pktinfo, sizeof(null_pktinfo));
1703 			optdata = (void *)&null_pktinfo;
1704 		}
1705 		optdatalen = sizeof(struct in6_pktinfo);
1706 		break;
1707 	case IPV6_TCLASS:
1708 		if (pktopt && pktopt->ip6po_tclass >= 0)
1709 			optdata = (void *)&pktopt->ip6po_tclass;
1710 		else
1711 			optdata = (void *)&deftclass;
1712 		optdatalen = sizeof(int);
1713 		break;
1714 	case IPV6_HOPOPTS:
1715 		if (pktopt && pktopt->ip6po_hbh) {
1716 			optdata = (void *)pktopt->ip6po_hbh;
1717 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
1718 			optdatalen = (ip6e->ip6e_len + 1) << 3;
1719 		}
1720 		break;
1721 	case IPV6_RTHDR:
1722 		if (pktopt && pktopt->ip6po_rthdr) {
1723 			optdata = (void *)pktopt->ip6po_rthdr;
1724 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
1725 			optdatalen = (ip6e->ip6e_len + 1) << 3;
1726 		}
1727 		break;
1728 	case IPV6_RTHDRDSTOPTS:
1729 		if (pktopt && pktopt->ip6po_dest1) {
1730 			optdata = (void *)pktopt->ip6po_dest1;
1731 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
1732 			optdatalen = (ip6e->ip6e_len + 1) << 3;
1733 		}
1734 		break;
1735 	case IPV6_DSTOPTS:
1736 		if (pktopt && pktopt->ip6po_dest2) {
1737 			optdata = (void *)pktopt->ip6po_dest2;
1738 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
1739 			optdatalen = (ip6e->ip6e_len + 1) << 3;
1740 		}
1741 		break;
1742 	case IPV6_USE_MIN_MTU:
1743 		if (pktopt)
1744 			optdata = (void *)&pktopt->ip6po_minmtu;
1745 		else
1746 			optdata = (void *)&defminmtu;
1747 		optdatalen = sizeof(int);
1748 		break;
1749 	case IPV6_DONTFRAG:
1750 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
1751 			on = 1;
1752 		else
1753 			on = 0;
1754 		optdata = (void *)&on;
1755 		optdatalen = sizeof(on);
1756 		break;
1757 	default:		/* should not happen */
1758 #ifdef DIAGNOSTIC
1759 		panic("%s: unexpected option", __func__);
1760 #endif
1761 		return (ENOPROTOOPT);
1762 	}
1763 
1764 	if (optdatalen > MCLBYTES)
1765 		return (EMSGSIZE); /* XXX */
1766 	if (optdatalen > MLEN)
1767 		MCLGET(m, M_WAIT);
1768 	m->m_len = optdatalen;
1769 	if (optdatalen)
1770 		bcopy(optdata, mtod(m, void *), optdatalen);
1771 
1772 	return (error);
1773 }
1774 
1775 void
1776 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
1777 {
1778 	if (optname == -1 || optname == IPV6_PKTINFO) {
1779 		if (pktopt->ip6po_pktinfo)
1780 			free(pktopt->ip6po_pktinfo, M_IP6OPT, 0);
1781 		pktopt->ip6po_pktinfo = NULL;
1782 	}
1783 	if (optname == -1 || optname == IPV6_HOPLIMIT)
1784 		pktopt->ip6po_hlim = -1;
1785 	if (optname == -1 || optname == IPV6_TCLASS)
1786 		pktopt->ip6po_tclass = -1;
1787 	if (optname == -1 || optname == IPV6_HOPOPTS) {
1788 		if (pktopt->ip6po_hbh)
1789 			free(pktopt->ip6po_hbh, M_IP6OPT, 0);
1790 		pktopt->ip6po_hbh = NULL;
1791 	}
1792 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
1793 		if (pktopt->ip6po_dest1)
1794 			free(pktopt->ip6po_dest1, M_IP6OPT, 0);
1795 		pktopt->ip6po_dest1 = NULL;
1796 	}
1797 	if (optname == -1 || optname == IPV6_RTHDR) {
1798 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
1799 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT, 0);
1800 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
1801 		if (pktopt->ip6po_route.ro_rt) {
1802 			rtfree(pktopt->ip6po_route.ro_rt);
1803 			pktopt->ip6po_route.ro_rt = NULL;
1804 		}
1805 	}
1806 	if (optname == -1 || optname == IPV6_DSTOPTS) {
1807 		if (pktopt->ip6po_dest2)
1808 			free(pktopt->ip6po_dest2, M_IP6OPT, 0);
1809 		pktopt->ip6po_dest2 = NULL;
1810 	}
1811 }
1812 
1813 #define PKTOPT_EXTHDRCPY(type) \
1814 do {\
1815 	if (src->type) {\
1816 		size_t hlen;\
1817 		hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
1818 		dst->type = malloc(hlen, M_IP6OPT, M_NOWAIT);\
1819 		if (dst->type == NULL)\
1820 			goto bad;\
1821 		memcpy(dst->type, src->type, hlen);\
1822 	}\
1823 } while (/*CONSTCOND*/ 0)
1824 
1825 int
1826 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src)
1827 {
1828 	dst->ip6po_hlim = src->ip6po_hlim;
1829 	dst->ip6po_tclass = src->ip6po_tclass;
1830 	dst->ip6po_flags = src->ip6po_flags;
1831 	if (src->ip6po_pktinfo) {
1832 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
1833 		    M_IP6OPT, M_NOWAIT);
1834 		if (dst->ip6po_pktinfo == NULL)
1835 			goto bad;
1836 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
1837 	}
1838 	PKTOPT_EXTHDRCPY(ip6po_hbh);
1839 	PKTOPT_EXTHDRCPY(ip6po_dest1);
1840 	PKTOPT_EXTHDRCPY(ip6po_dest2);
1841 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
1842 	return (0);
1843 
1844   bad:
1845 	ip6_clearpktopts(dst, -1);
1846 	return (ENOBUFS);
1847 }
1848 #undef PKTOPT_EXTHDRCPY
1849 
1850 void
1851 ip6_freepcbopts(struct ip6_pktopts *pktopt)
1852 {
1853 	if (pktopt == NULL)
1854 		return;
1855 
1856 	ip6_clearpktopts(pktopt, -1);
1857 
1858 	free(pktopt, M_IP6OPT, 0);
1859 }
1860 
1861 /*
1862  * Set the IP6 multicast options in response to user setsockopt().
1863  */
1864 int
1865 ip6_setmoptions(int optname, struct ip6_moptions **im6op, struct mbuf *m,
1866     unsigned int rtableid)
1867 {
1868 	int error = 0;
1869 	u_int loop, ifindex;
1870 	struct ipv6_mreq *mreq;
1871 	struct ifnet *ifp;
1872 	struct ip6_moptions *im6o = *im6op;
1873 	struct in6_multi_mship *imm;
1874 	struct proc *p = curproc;	/* XXX */
1875 
1876 	if (im6o == NULL) {
1877 		/*
1878 		 * No multicast option buffer attached to the pcb;
1879 		 * allocate one and initialize to default values.
1880 		 */
1881 		im6o = malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
1882 		if (im6o == NULL)
1883 			return (ENOBUFS);
1884 		*im6op = im6o;
1885 		im6o->im6o_ifidx = 0;
1886 		im6o->im6o_hlim = ip6_defmcasthlim;
1887 		im6o->im6o_loop = IPV6_DEFAULT_MULTICAST_LOOP;
1888 		LIST_INIT(&im6o->im6o_memberships);
1889 	}
1890 
1891 	switch (optname) {
1892 
1893 	case IPV6_MULTICAST_IF:
1894 		/*
1895 		 * Select the interface for outgoing multicast packets.
1896 		 */
1897 		if (m == NULL || m->m_len != sizeof(u_int)) {
1898 			error = EINVAL;
1899 			break;
1900 		}
1901 		memcpy(&ifindex, mtod(m, u_int *), sizeof(ifindex));
1902 		if (ifindex != 0) {
1903 			ifp = if_get(ifindex);
1904 			if (ifp == NULL) {
1905 				error = ENXIO;	/* XXX EINVAL? */
1906 				break;
1907 			}
1908 			if (ifp->if_rdomain != rtable_l2(rtableid) ||
1909 			    (ifp->if_flags & IFF_MULTICAST) == 0) {
1910 				error = EADDRNOTAVAIL;
1911 				if_put(ifp);
1912 				break;
1913 			}
1914 			if_put(ifp);
1915 		}
1916 		im6o->im6o_ifidx = ifindex;
1917 		break;
1918 
1919 	case IPV6_MULTICAST_HOPS:
1920 	    {
1921 		/*
1922 		 * Set the IP6 hoplimit for outgoing multicast packets.
1923 		 */
1924 		int optval;
1925 		if (m == NULL || m->m_len != sizeof(int)) {
1926 			error = EINVAL;
1927 			break;
1928 		}
1929 		memcpy(&optval, mtod(m, u_int *), sizeof(optval));
1930 		if (optval < -1 || optval >= 256)
1931 			error = EINVAL;
1932 		else if (optval == -1)
1933 			im6o->im6o_hlim = ip6_defmcasthlim;
1934 		else
1935 			im6o->im6o_hlim = optval;
1936 		break;
1937 	    }
1938 
1939 	case IPV6_MULTICAST_LOOP:
1940 		/*
1941 		 * Set the loopback flag for outgoing multicast packets.
1942 		 * Must be zero or one.
1943 		 */
1944 		if (m == NULL || m->m_len != sizeof(u_int)) {
1945 			error = EINVAL;
1946 			break;
1947 		}
1948 		memcpy(&loop, mtod(m, u_int *), sizeof(loop));
1949 		if (loop > 1) {
1950 			error = EINVAL;
1951 			break;
1952 		}
1953 		im6o->im6o_loop = loop;
1954 		break;
1955 
1956 	case IPV6_JOIN_GROUP:
1957 		/*
1958 		 * Add a multicast group membership.
1959 		 * Group must be a valid IP6 multicast address.
1960 		 */
1961 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
1962 			error = EINVAL;
1963 			break;
1964 		}
1965 		mreq = mtod(m, struct ipv6_mreq *);
1966 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
1967 			/*
1968 			 * We use the unspecified address to specify to accept
1969 			 * all multicast addresses. Only super user is allowed
1970 			 * to do this.
1971 			 */
1972 			if (suser(p))
1973 			{
1974 				error = EACCES;
1975 				break;
1976 			}
1977 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
1978 			error = EINVAL;
1979 			break;
1980 		}
1981 
1982 		/*
1983 		 * If no interface was explicitly specified, choose an
1984 		 * appropriate one according to the given multicast address.
1985 		 */
1986 		if (mreq->ipv6mr_interface == 0) {
1987 			struct rtentry *rt;
1988 			struct sockaddr_in6 dst;
1989 
1990 			memset(&dst, 0, sizeof(dst));
1991 			dst.sin6_len = sizeof(dst);
1992 			dst.sin6_family = AF_INET6;
1993 			dst.sin6_addr = mreq->ipv6mr_multiaddr;
1994 			rt = rtalloc(sin6tosa(&dst), RT_RESOLVE, rtableid);
1995 			if (rt == NULL) {
1996 				error = EADDRNOTAVAIL;
1997 				break;
1998 			}
1999 			ifp = if_get(rt->rt_ifidx);
2000 			rtfree(rt);
2001 		} else {
2002 			/*
2003 			 * If the interface is specified, validate it.
2004 			 */
2005 			ifp = if_get(mreq->ipv6mr_interface);
2006 			if (ifp == NULL) {
2007 				error = ENXIO;	/* XXX EINVAL? */
2008 				break;
2009 			}
2010 		}
2011 
2012 		/*
2013 		 * See if we found an interface, and confirm that it
2014 		 * supports multicast
2015 		 */
2016 		if (ifp == NULL || ifp->if_rdomain != rtable_l2(rtableid) ||
2017 		    (ifp->if_flags & IFF_MULTICAST) == 0) {
2018 			if_put(ifp);
2019 			error = EADDRNOTAVAIL;
2020 			break;
2021 		}
2022 		/*
2023 		 * Put interface index into the multicast address,
2024 		 * if the address has link/interface-local scope.
2025 		 */
2026 		if (IN6_IS_SCOPE_EMBED(&mreq->ipv6mr_multiaddr)) {
2027 			mreq->ipv6mr_multiaddr.s6_addr16[1] =
2028 			    htons(ifp->if_index);
2029 		}
2030 		/*
2031 		 * See if the membership already exists.
2032 		 */
2033 		LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain)
2034 			if (imm->i6mm_maddr->in6m_ifidx == ifp->if_index &&
2035 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2036 			    &mreq->ipv6mr_multiaddr))
2037 				break;
2038 		if (imm != NULL) {
2039 			if_put(ifp);
2040 			error = EADDRINUSE;
2041 			break;
2042 		}
2043 		/*
2044 		 * Everything looks good; add a new record to the multicast
2045 		 * address list for the given interface.
2046 		 */
2047 		imm = in6_joingroup(ifp, &mreq->ipv6mr_multiaddr, &error);
2048 		if_put(ifp);
2049 		if (!imm)
2050 			break;
2051 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
2052 		break;
2053 
2054 	case IPV6_LEAVE_GROUP:
2055 		/*
2056 		 * Drop a multicast group membership.
2057 		 * Group must be a valid IP6 multicast address.
2058 		 */
2059 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
2060 			error = EINVAL;
2061 			break;
2062 		}
2063 		mreq = mtod(m, struct ipv6_mreq *);
2064 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
2065 			if (suser(p)) {
2066 				error = EACCES;
2067 				break;
2068 			}
2069 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
2070 			error = EINVAL;
2071 			break;
2072 		}
2073 
2074 		/*
2075 		 * Put interface index into the multicast address,
2076 		 * if the address has link-local scope.
2077 		 */
2078 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
2079 			mreq->ipv6mr_multiaddr.s6_addr16[1] =
2080 			    htons(mreq->ipv6mr_interface);
2081 		}
2082 
2083 		/*
2084 		 * If an interface address was specified, get a pointer
2085 		 * to its ifnet structure.
2086 		 */
2087 		if (mreq->ipv6mr_interface == 0)
2088 			ifp = NULL;
2089 		else {
2090 			ifp = if_get(mreq->ipv6mr_interface);
2091 			if (ifp == NULL) {
2092 				error = ENXIO;	/* XXX EINVAL? */
2093 				break;
2094 			}
2095 		}
2096 
2097 		/*
2098 		 * Find the membership in the membership list.
2099 		 */
2100 		LIST_FOREACH(imm, &im6o->im6o_memberships, i6mm_chain) {
2101 			if ((ifp == NULL ||
2102 			    imm->i6mm_maddr->in6m_ifidx == ifp->if_index) &&
2103 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
2104 			    &mreq->ipv6mr_multiaddr))
2105 				break;
2106 		}
2107 
2108 		if_put(ifp);
2109 
2110 		if (imm == NULL) {
2111 			/* Unable to resolve interface */
2112 			error = EADDRNOTAVAIL;
2113 			break;
2114 		}
2115 		/*
2116 		 * Give up the multicast address record to which the
2117 		 * membership points.
2118 		 */
2119 		LIST_REMOVE(imm, i6mm_chain);
2120 		in6_leavegroup(imm);
2121 		break;
2122 
2123 	default:
2124 		error = EOPNOTSUPP;
2125 		break;
2126 	}
2127 
2128 	/*
2129 	 * If all options have default values, no need to keep the option
2130 	 * structure.
2131 	 */
2132 	if (im6o->im6o_ifidx == 0 &&
2133 	    im6o->im6o_hlim == ip6_defmcasthlim &&
2134 	    im6o->im6o_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
2135 	    LIST_EMPTY(&im6o->im6o_memberships)) {
2136 		free(*im6op, M_IPMOPTS, sizeof(**im6op));
2137 		*im6op = NULL;
2138 	}
2139 
2140 	return (error);
2141 }
2142 
2143 /*
2144  * Return the IP6 multicast options in response to user getsockopt().
2145  */
2146 int
2147 ip6_getmoptions(int optname, struct ip6_moptions *im6o, struct mbuf *m)
2148 {
2149 	u_int *hlim, *loop, *ifindex;
2150 
2151 	switch (optname) {
2152 	case IPV6_MULTICAST_IF:
2153 		ifindex = mtod(m, u_int *);
2154 		m->m_len = sizeof(u_int);
2155 		if (im6o == NULL || im6o->im6o_ifidx == 0)
2156 			*ifindex = 0;
2157 		else
2158 			*ifindex = im6o->im6o_ifidx;
2159 		return (0);
2160 
2161 	case IPV6_MULTICAST_HOPS:
2162 		hlim = mtod(m, u_int *);
2163 		m->m_len = sizeof(u_int);
2164 		if (im6o == NULL)
2165 			*hlim = ip6_defmcasthlim;
2166 		else
2167 			*hlim = im6o->im6o_hlim;
2168 		return (0);
2169 
2170 	case IPV6_MULTICAST_LOOP:
2171 		loop = mtod(m, u_int *);
2172 		m->m_len = sizeof(u_int);
2173 		if (im6o == NULL)
2174 			*loop = ip6_defmcasthlim;
2175 		else
2176 			*loop = im6o->im6o_loop;
2177 		return (0);
2178 
2179 	default:
2180 		return (EOPNOTSUPP);
2181 	}
2182 }
2183 
2184 /*
2185  * Discard the IP6 multicast options.
2186  */
2187 void
2188 ip6_freemoptions(struct ip6_moptions *im6o)
2189 {
2190 	struct in6_multi_mship *imm;
2191 
2192 	if (im6o == NULL)
2193 		return;
2194 
2195 	while (!LIST_EMPTY(&im6o->im6o_memberships)) {
2196 		imm = LIST_FIRST(&im6o->im6o_memberships);
2197 		LIST_REMOVE(imm, i6mm_chain);
2198 		in6_leavegroup(imm);
2199 	}
2200 	free(im6o, M_IPMOPTS, sizeof(*im6o));
2201 }
2202 
2203 /*
2204  * Set IPv6 outgoing packet options based on advanced API.
2205  */
2206 int
2207 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2208     struct ip6_pktopts *stickyopt, int priv, int uproto)
2209 {
2210 	u_int clen;
2211 	struct cmsghdr *cm = 0;
2212 	caddr_t cmsgs;
2213 	int error;
2214 
2215 	if (control == NULL || opt == NULL)
2216 		return (EINVAL);
2217 
2218 	ip6_initpktopts(opt);
2219 	if (stickyopt) {
2220 		int error;
2221 
2222 		/*
2223 		 * If stickyopt is provided, make a local copy of the options
2224 		 * for this particular packet, then override them by ancillary
2225 		 * objects.
2226 		 * XXX: copypktopts() does not copy the cached route to a next
2227 		 * hop (if any).  This is not very good in terms of efficiency,
2228 		 * but we can allow this since this option should be rarely
2229 		 * used.
2230 		 */
2231 		if ((error = copypktopts(opt, stickyopt)) != 0)
2232 			return (error);
2233 	}
2234 
2235 	/*
2236 	 * XXX: Currently, we assume all the optional information is stored
2237 	 * in a single mbuf.
2238 	 */
2239 	if (control->m_next)
2240 		return (EINVAL);
2241 
2242 	clen = control->m_len;
2243 	cmsgs = mtod(control, caddr_t);
2244 	do {
2245 		if (clen < CMSG_LEN(0))
2246 			return (EINVAL);
2247 		cm = (struct cmsghdr *)cmsgs;
2248 		if (cm->cmsg_len < CMSG_LEN(0) || cm->cmsg_len > clen ||
2249 		    CMSG_ALIGN(cm->cmsg_len) > clen)
2250 			return (EINVAL);
2251 		if (cm->cmsg_level == IPPROTO_IPV6) {
2252 			error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2253 			    cm->cmsg_len - CMSG_LEN(0), opt, priv, 0, uproto);
2254 			if (error)
2255 				return (error);
2256 		}
2257 
2258 		clen -= CMSG_ALIGN(cm->cmsg_len);
2259 		cmsgs += CMSG_ALIGN(cm->cmsg_len);
2260 	} while (clen);
2261 
2262 	return (0);
2263 }
2264 
2265 /*
2266  * Set a particular packet option, as a sticky option or an ancillary data
2267  * item.  "len" can be 0 only when it's a sticky option.
2268  */
2269 int
2270 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2271     int priv, int sticky, int uproto)
2272 {
2273 	int minmtupolicy;
2274 
2275 	switch (optname) {
2276 	case IPV6_PKTINFO:
2277 	{
2278 		struct ifnet *ifp = NULL;
2279 		struct in6_pktinfo *pktinfo;
2280 
2281 		if (len != sizeof(struct in6_pktinfo))
2282 			return (EINVAL);
2283 
2284 		pktinfo = (struct in6_pktinfo *)buf;
2285 
2286 		/*
2287 		 * An application can clear any sticky IPV6_PKTINFO option by
2288 		 * doing a "regular" setsockopt with ipi6_addr being
2289 		 * in6addr_any and ipi6_ifindex being zero.
2290 		 * [RFC 3542, Section 6]
2291 		 */
2292 		if (opt->ip6po_pktinfo &&
2293 		    pktinfo->ipi6_ifindex == 0 &&
2294 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2295 			ip6_clearpktopts(opt, optname);
2296 			break;
2297 		}
2298 
2299 		if (uproto == IPPROTO_TCP &&
2300 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2301 			return (EINVAL);
2302 		}
2303 
2304 		if (pktinfo->ipi6_ifindex) {
2305 			ifp = if_get(pktinfo->ipi6_ifindex);
2306 			if (ifp == NULL)
2307 				return (ENXIO);
2308 			if_put(ifp);
2309 		}
2310 
2311 		/*
2312 		 * We store the address anyway, and let in6_selectsrc()
2313 		 * validate the specified address.  This is because ipi6_addr
2314 		 * may not have enough information about its scope zone, and
2315 		 * we may need additional information (such as outgoing
2316 		 * interface or the scope zone of a destination address) to
2317 		 * disambiguate the scope.
2318 		 * XXX: the delay of the validation may confuse the
2319 		 * application when it is used as a sticky option.
2320 		 */
2321 		if (opt->ip6po_pktinfo == NULL) {
2322 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2323 			    M_IP6OPT, M_NOWAIT);
2324 			if (opt->ip6po_pktinfo == NULL)
2325 				return (ENOBUFS);
2326 		}
2327 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2328 		break;
2329 	}
2330 
2331 	case IPV6_HOPLIMIT:
2332 	{
2333 		int *hlimp;
2334 
2335 		/*
2336 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2337 		 * to simplify the ordering among hoplimit options.
2338 		 */
2339 		if (sticky)
2340 			return (ENOPROTOOPT);
2341 
2342 		if (len != sizeof(int))
2343 			return (EINVAL);
2344 		hlimp = (int *)buf;
2345 		if (*hlimp < -1 || *hlimp > 255)
2346 			return (EINVAL);
2347 
2348 		opt->ip6po_hlim = *hlimp;
2349 		break;
2350 	}
2351 
2352 	case IPV6_TCLASS:
2353 	{
2354 		int tclass;
2355 
2356 		if (len != sizeof(int))
2357 			return (EINVAL);
2358 		tclass = *(int *)buf;
2359 		if (tclass < -1 || tclass > 255)
2360 			return (EINVAL);
2361 
2362 		opt->ip6po_tclass = tclass;
2363 		break;
2364 	}
2365 	case IPV6_HOPOPTS:
2366 	{
2367 		struct ip6_hbh *hbh;
2368 		int hbhlen;
2369 
2370 		/*
2371 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2372 		 * options, since per-option restriction has too much
2373 		 * overhead.
2374 		 */
2375 		if (!priv)
2376 			return (EPERM);
2377 
2378 		if (len == 0) {
2379 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2380 			break;	/* just remove the option */
2381 		}
2382 
2383 		/* message length validation */
2384 		if (len < sizeof(struct ip6_hbh))
2385 			return (EINVAL);
2386 		hbh = (struct ip6_hbh *)buf;
2387 		hbhlen = (hbh->ip6h_len + 1) << 3;
2388 		if (len != hbhlen)
2389 			return (EINVAL);
2390 
2391 		/* turn off the previous option, then set the new option. */
2392 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2393 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2394 		if (opt->ip6po_hbh == NULL)
2395 			return (ENOBUFS);
2396 		memcpy(opt->ip6po_hbh, hbh, hbhlen);
2397 
2398 		break;
2399 	}
2400 
2401 	case IPV6_DSTOPTS:
2402 	case IPV6_RTHDRDSTOPTS:
2403 	{
2404 		struct ip6_dest *dest, **newdest = NULL;
2405 		int destlen;
2406 
2407 		if (!priv)	/* XXX: see the comment for IPV6_HOPOPTS */
2408 			return (EPERM);
2409 
2410 		if (len == 0) {
2411 			ip6_clearpktopts(opt, optname);
2412 			break;	/* just remove the option */
2413 		}
2414 
2415 		/* message length validation */
2416 		if (len < sizeof(struct ip6_dest))
2417 			return (EINVAL);
2418 		dest = (struct ip6_dest *)buf;
2419 		destlen = (dest->ip6d_len + 1) << 3;
2420 		if (len != destlen)
2421 			return (EINVAL);
2422 		/*
2423 		 * Determine the position that the destination options header
2424 		 * should be inserted; before or after the routing header.
2425 		 */
2426 		switch (optname) {
2427 		case IPV6_RTHDRDSTOPTS:
2428 			newdest = &opt->ip6po_dest1;
2429 			break;
2430 		case IPV6_DSTOPTS:
2431 			newdest = &opt->ip6po_dest2;
2432 			break;
2433 		}
2434 
2435 		/* turn off the previous option, then set the new option. */
2436 		ip6_clearpktopts(opt, optname);
2437 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2438 		if (*newdest == NULL)
2439 			return (ENOBUFS);
2440 		memcpy(*newdest, dest, destlen);
2441 
2442 		break;
2443 	}
2444 
2445 	case IPV6_RTHDR:
2446 	{
2447 		struct ip6_rthdr *rth;
2448 		int rthlen;
2449 
2450 		if (len == 0) {
2451 			ip6_clearpktopts(opt, IPV6_RTHDR);
2452 			break;	/* just remove the option */
2453 		}
2454 
2455 		/* message length validation */
2456 		if (len < sizeof(struct ip6_rthdr))
2457 			return (EINVAL);
2458 		rth = (struct ip6_rthdr *)buf;
2459 		rthlen = (rth->ip6r_len + 1) << 3;
2460 		if (len != rthlen)
2461 			return (EINVAL);
2462 
2463 		switch (rth->ip6r_type) {
2464 		case IPV6_RTHDR_TYPE_0:
2465 			if (rth->ip6r_len == 0)	/* must contain one addr */
2466 				return (EINVAL);
2467 			if (rth->ip6r_len % 2) /* length must be even */
2468 				return (EINVAL);
2469 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2470 				return (EINVAL);
2471 			break;
2472 		default:
2473 			return (EINVAL);	/* not supported */
2474 		}
2475 		/* turn off the previous option */
2476 		ip6_clearpktopts(opt, IPV6_RTHDR);
2477 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2478 		if (opt->ip6po_rthdr == NULL)
2479 			return (ENOBUFS);
2480 		memcpy(opt->ip6po_rthdr, rth, rthlen);
2481 		break;
2482 	}
2483 
2484 	case IPV6_USE_MIN_MTU:
2485 		if (len != sizeof(int))
2486 			return (EINVAL);
2487 		minmtupolicy = *(int *)buf;
2488 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2489 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
2490 		    minmtupolicy != IP6PO_MINMTU_ALL) {
2491 			return (EINVAL);
2492 		}
2493 		opt->ip6po_minmtu = minmtupolicy;
2494 		break;
2495 
2496 	case IPV6_DONTFRAG:
2497 		if (len != sizeof(int))
2498 			return (EINVAL);
2499 
2500 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2501 			/*
2502 			 * we ignore this option for TCP sockets.
2503 			 * (RFC3542 leaves this case unspecified.)
2504 			 */
2505 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2506 		} else
2507 			opt->ip6po_flags |= IP6PO_DONTFRAG;
2508 		break;
2509 
2510 	default:
2511 		return (ENOPROTOOPT);
2512 	} /* end of switch */
2513 
2514 	return (0);
2515 }
2516 
2517 /*
2518  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2519  * packet to the input queue of a specified interface.
2520  */
2521 void
2522 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2523 {
2524 	struct mbuf *copym;
2525 	struct ip6_hdr *ip6;
2526 
2527 	/*
2528 	 * Duplicate the packet.
2529 	 */
2530 	copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
2531 	if (copym == NULL)
2532 		return;
2533 
2534 	/*
2535 	 * Make sure to deep-copy IPv6 header portion in case the data
2536 	 * is in an mbuf cluster, so that we can safely override the IPv6
2537 	 * header portion later.
2538 	 */
2539 	if ((copym->m_flags & M_EXT) != 0 ||
2540 	    copym->m_len < sizeof(struct ip6_hdr)) {
2541 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
2542 		if (copym == NULL)
2543 			return;
2544 	}
2545 
2546 #ifdef DIAGNOSTIC
2547 	if (copym->m_len < sizeof(*ip6)) {
2548 		m_freem(copym);
2549 		return;
2550 	}
2551 #endif
2552 
2553 	ip6 = mtod(copym, struct ip6_hdr *);
2554 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
2555 		ip6->ip6_src.s6_addr16[1] = 0;
2556 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
2557 		ip6->ip6_dst.s6_addr16[1] = 0;
2558 
2559 	if_input_local(ifp, copym, dst->sin6_family);
2560 }
2561 
2562 /*
2563  * Chop IPv6 header off from the payload.
2564  */
2565 int
2566 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
2567 {
2568 	struct mbuf *mh;
2569 	struct ip6_hdr *ip6;
2570 
2571 	ip6 = mtod(m, struct ip6_hdr *);
2572 	if (m->m_len > sizeof(*ip6)) {
2573 		MGET(mh, M_DONTWAIT, MT_HEADER);
2574 		if (mh == NULL) {
2575 			m_freem(m);
2576 			return ENOBUFS;
2577 		}
2578 		M_MOVE_PKTHDR(mh, m);
2579 		m_align(mh, sizeof(*ip6));
2580 		m->m_len -= sizeof(*ip6);
2581 		m->m_data += sizeof(*ip6);
2582 		mh->m_next = m;
2583 		m = mh;
2584 		m->m_len = sizeof(*ip6);
2585 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2586 	}
2587 	exthdrs->ip6e_ip6 = m;
2588 	return 0;
2589 }
2590 
2591 u_int32_t
2592 ip6_randomid(void)
2593 {
2594 	return idgen32(&ip6_id_ctx);
2595 }
2596 
2597 void
2598 ip6_randomid_init(void)
2599 {
2600 	idgen32_init(&ip6_id_ctx);
2601 }
2602 
2603 /*
2604  *	Compute significant parts of the IPv6 checksum pseudo-header
2605  *	for use in a delayed TCP/UDP checksum calculation.
2606  */
2607 static __inline u_int16_t __attribute__((__unused__))
2608 in6_cksum_phdr(const struct in6_addr *src, const struct in6_addr *dst,
2609     u_int32_t len, u_int32_t nxt)
2610 {
2611 	u_int32_t sum = 0;
2612 	const u_int16_t *w;
2613 
2614 	w = (const u_int16_t *) src;
2615 	sum += w[0];
2616 	if (!IN6_IS_SCOPE_EMBED(src))
2617 		sum += w[1];
2618 	sum += w[2]; sum += w[3]; sum += w[4]; sum += w[5];
2619 	sum += w[6]; sum += w[7];
2620 
2621 	w = (const u_int16_t *) dst;
2622 	sum += w[0];
2623 	if (!IN6_IS_SCOPE_EMBED(dst))
2624 		sum += w[1];
2625 	sum += w[2]; sum += w[3]; sum += w[4]; sum += w[5];
2626 	sum += w[6]; sum += w[7];
2627 
2628 	sum += (u_int16_t)(len >> 16) + (u_int16_t)(len /*& 0xffff*/);
2629 
2630 	sum += (u_int16_t)(nxt >> 16) + (u_int16_t)(nxt /*& 0xffff*/);
2631 
2632 	sum = (u_int16_t)(sum >> 16) + (u_int16_t)(sum /*& 0xffff*/);
2633 
2634 	if (sum > 0xffff)
2635 		sum -= 0xffff;
2636 
2637 	return (sum);
2638 }
2639 
2640 /*
2641  * Process a delayed payload checksum calculation.
2642  */
2643 void
2644 in6_delayed_cksum(struct mbuf *m, u_int8_t nxt)
2645 {
2646 	int nxtp, offset;
2647 	u_int16_t csum;
2648 
2649 	offset = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxtp);
2650 	if (offset <= 0 || nxtp != nxt)
2651 		/* If the desired next protocol isn't found, punt. */
2652 		return;
2653 	csum = (u_int16_t)(in6_cksum(m, 0, offset, m->m_pkthdr.len - offset));
2654 
2655 	switch (nxt) {
2656 	case IPPROTO_TCP:
2657 		offset += offsetof(struct tcphdr, th_sum);
2658 		break;
2659 
2660 	case IPPROTO_UDP:
2661 		offset += offsetof(struct udphdr, uh_sum);
2662 		if (csum == 0)
2663 			csum = 0xffff;
2664 		break;
2665 
2666 	case IPPROTO_ICMPV6:
2667 		offset += offsetof(struct icmp6_hdr, icmp6_cksum);
2668 		break;
2669 	}
2670 
2671 	if ((offset + sizeof(u_int16_t)) > m->m_len)
2672 		m_copyback(m, offset, sizeof(csum), &csum, M_NOWAIT);
2673 	else
2674 		*(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
2675 }
2676 
2677 void
2678 in6_proto_cksum_out(struct mbuf *m, struct ifnet *ifp)
2679 {
2680 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
2681 
2682 	/* some hw and in6_delayed_cksum need the pseudo header cksum */
2683 	if (m->m_pkthdr.csum_flags &
2684 	    (M_TCP_CSUM_OUT|M_UDP_CSUM_OUT|M_ICMP_CSUM_OUT)) {
2685 		int nxt, offset;
2686 		u_int16_t csum;
2687 
2688 		offset = ip6_lasthdr(m, 0, IPPROTO_IPV6, &nxt);
2689 		if (ISSET(m->m_pkthdr.csum_flags, M_TCP_TSO) &&
2690 		    in_ifcap_cksum(m, ifp, IFCAP_TSOv6)) {
2691 			csum = in6_cksum_phdr(&ip6->ip6_src, &ip6->ip6_dst,
2692 			    htonl(0), htonl(nxt));
2693 		} else {
2694 			csum = in6_cksum_phdr(&ip6->ip6_src, &ip6->ip6_dst,
2695 			    htonl(m->m_pkthdr.len - offset), htonl(nxt));
2696 		}
2697 		if (nxt == IPPROTO_TCP)
2698 			offset += offsetof(struct tcphdr, th_sum);
2699 		else if (nxt == IPPROTO_UDP)
2700 			offset += offsetof(struct udphdr, uh_sum);
2701 		else if (nxt == IPPROTO_ICMPV6)
2702 			offset += offsetof(struct icmp6_hdr, icmp6_cksum);
2703 		if ((offset + sizeof(u_int16_t)) > m->m_len)
2704 			m_copyback(m, offset, sizeof(csum), &csum, M_NOWAIT);
2705 		else
2706 			*(u_int16_t *)(mtod(m, caddr_t) + offset) = csum;
2707 	}
2708 
2709 	if (m->m_pkthdr.csum_flags & M_TCP_CSUM_OUT) {
2710 		if (!ifp || !(ifp->if_capabilities & IFCAP_CSUM_TCPv6) ||
2711 		    ip6->ip6_nxt != IPPROTO_TCP ||
2712 		    ifp->if_bridgeidx != 0) {
2713 			tcpstat_inc(tcps_outswcsum);
2714 			in6_delayed_cksum(m, IPPROTO_TCP);
2715 			m->m_pkthdr.csum_flags &= ~M_TCP_CSUM_OUT; /* Clear */
2716 		}
2717 	} else if (m->m_pkthdr.csum_flags & M_UDP_CSUM_OUT) {
2718 		if (!ifp || !(ifp->if_capabilities & IFCAP_CSUM_UDPv6) ||
2719 		    ip6->ip6_nxt != IPPROTO_UDP ||
2720 		    ifp->if_bridgeidx != 0) {
2721 			udpstat_inc(udps_outswcsum);
2722 			in6_delayed_cksum(m, IPPROTO_UDP);
2723 			m->m_pkthdr.csum_flags &= ~M_UDP_CSUM_OUT; /* Clear */
2724 		}
2725 	} else if (m->m_pkthdr.csum_flags & M_ICMP_CSUM_OUT) {
2726 		in6_delayed_cksum(m, IPPROTO_ICMPV6);
2727 		m->m_pkthdr.csum_flags &= ~M_ICMP_CSUM_OUT; /* Clear */
2728 	}
2729 }
2730 
2731 #ifdef IPSEC
2732 int
2733 ip6_output_ipsec_lookup(struct mbuf *m, const struct ipsec_level *seclevel,
2734     struct tdb **tdbout)
2735 {
2736 	struct tdb *tdb;
2737 	struct m_tag *mtag;
2738 	struct tdb_ident *tdbi;
2739 	int error;
2740 
2741 	/*
2742 	 * Check if there was an outgoing SA bound to the flow
2743 	 * from a transport protocol.
2744 	 */
2745 
2746 	/* Do we have any pending SAs to apply ? */
2747 	error = ipsp_spd_lookup(m, AF_INET6, sizeof(struct ip6_hdr),
2748 	    IPSP_DIRECTION_OUT, NULL, seclevel, &tdb, NULL);
2749 	if (error || tdb == NULL) {
2750 		*tdbout = NULL;
2751 		return error;
2752 	}
2753 	/* Loop detection */
2754 	for (mtag = m_tag_first(m); mtag != NULL; mtag = m_tag_next(m, mtag)) {
2755 		if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE)
2756 			continue;
2757 		tdbi = (struct tdb_ident *)(mtag + 1);
2758 		if (tdbi->spi == tdb->tdb_spi &&
2759 		    tdbi->proto == tdb->tdb_sproto &&
2760 		    tdbi->rdomain == tdb->tdb_rdomain &&
2761 		    !memcmp(&tdbi->dst, &tdb->tdb_dst,
2762 		    sizeof(union sockaddr_union))) {
2763 			/* no IPsec needed */
2764 			tdb_unref(tdb);
2765 			*tdbout = NULL;
2766 			return 0;
2767 		}
2768 	}
2769 	*tdbout = tdb;
2770 	return 0;
2771 }
2772 
2773 int
2774 ip6_output_ipsec_pmtu_update(struct tdb *tdb, struct route *ro,
2775     struct in6_addr *dst, int ifidx, int rtableid, int transportmode)
2776 {
2777 	struct rtentry *rt = NULL;
2778 	int rt_mtucloned = 0;
2779 
2780 	/* Find a host route to store the mtu in */
2781 	if (ro != NULL)
2782 		rt = ro->ro_rt;
2783 	/* but don't add a PMTU route for transport mode SAs */
2784 	if (transportmode)
2785 		rt = NULL;
2786 	else if (rt == NULL || (rt->rt_flags & RTF_HOST) == 0) {
2787 		struct sockaddr_in6 sin6;
2788 		int error;
2789 
2790 		memset(&sin6, 0, sizeof(sin6));
2791 		sin6.sin6_family = AF_INET6;
2792 		sin6.sin6_len = sizeof(sin6);
2793 		sin6.sin6_addr = *dst;
2794 		sin6.sin6_scope_id = in6_addr2scopeid(ifidx, dst);
2795 		error = in6_embedscope(dst, &sin6, NULL, NULL);
2796 		if (error) {
2797 			/* should be impossible */
2798 			return error;
2799 		}
2800 		rt = icmp6_mtudisc_clone(&sin6, rtableid, 1);
2801 		rt_mtucloned = 1;
2802 	}
2803 	DPRINTF("spi %08x mtu %d rt %p cloned %d",
2804 	    ntohl(tdb->tdb_spi), tdb->tdb_mtu, rt, rt_mtucloned);
2805 	if (rt != NULL) {
2806 		rt->rt_mtu = tdb->tdb_mtu;
2807 		if (ro != NULL && ro->ro_rt != NULL) {
2808 			rtfree(ro->ro_rt);
2809 			ro->ro_rt = rtalloc(&ro->ro_dstsa, RT_RESOLVE,
2810 			    rtableid);
2811 		}
2812 		if (rt_mtucloned)
2813 			rtfree(rt);
2814 	}
2815 	return 0;
2816 }
2817 
2818 int
2819 ip6_output_ipsec_send(struct tdb *tdb, struct mbuf *m, struct route *ro,
2820     int tunalready, int fwd)
2821 {
2822 	struct mbuf_list ml;
2823 	struct ifnet *encif = NULL;
2824 	struct ip6_hdr *ip6;
2825 	struct in6_addr dst;
2826 	u_int len;
2827 	int error, ifidx, rtableid, tso = 0;
2828 
2829 #if NPF > 0
2830 	/*
2831 	 * Packet filter
2832 	 */
2833 	if ((encif = enc_getif(tdb->tdb_rdomain, tdb->tdb_tap)) == NULL ||
2834 	    pf_test(AF_INET6, fwd ? PF_FWD : PF_OUT, encif, &m) != PF_PASS) {
2835 		m_freem(m);
2836 		return EACCES;
2837 	}
2838 	if (m == NULL)
2839 		return 0;
2840 	/*
2841 	 * PF_TAG_REROUTE handling or not...
2842 	 * Packet is entering IPsec so the routing is
2843 	 * already overruled by the IPsec policy.
2844 	 * Until now the change was not reconsidered.
2845 	 * What's the behaviour?
2846 	 */
2847 #endif
2848 
2849 	/* Check if we can chop the TCP packet */
2850 	ip6 = mtod(m, struct ip6_hdr *);
2851 	if (ISSET(m->m_pkthdr.csum_flags, M_TCP_TSO) &&
2852 	    m->m_pkthdr.ph_mss <= tdb->tdb_mtu) {
2853 		tso = 1;
2854 		len = m->m_pkthdr.ph_mss;
2855 	} else
2856 		len = sizeof(struct ip6_hdr) + ntohs(ip6->ip6_plen);
2857 
2858 	/* Check if we are allowed to fragment */
2859 	dst = ip6->ip6_dst;
2860 	ifidx = m->m_pkthdr.ph_ifidx;
2861 	rtableid = m->m_pkthdr.ph_rtableid;
2862 	if (ip_mtudisc && tdb->tdb_mtu &&
2863 	    len > tdb->tdb_mtu && tdb->tdb_mtutimeout > gettime()) {
2864 		int transportmode;
2865 
2866 		transportmode = (tdb->tdb_dst.sa.sa_family == AF_INET6) &&
2867 		    (IN6_ARE_ADDR_EQUAL(&tdb->tdb_dst.sin6.sin6_addr, &dst));
2868 		error = ip6_output_ipsec_pmtu_update(tdb, ro, &dst, ifidx,
2869 		    rtableid, transportmode);
2870 		if (error) {
2871 			ipsecstat_inc(ipsec_odrops);
2872 			tdbstat_inc(tdb, tdb_odrops);
2873 			m_freem(m);
2874 			return error;
2875 		}
2876 		ipsec_adjust_mtu(m, tdb->tdb_mtu);
2877 		m_freem(m);
2878 		return EMSGSIZE;
2879 	}
2880 	/* propagate don't fragment for v6-over-v6 */
2881 	if (ip_mtudisc)
2882 		SET(m->m_pkthdr.csum_flags, M_IPV6_DF_OUT);
2883 
2884 	/*
2885 	 * Clear these -- they'll be set in the recursive invocation
2886 	 * as needed.
2887 	 */
2888 	m->m_flags &= ~(M_BCAST | M_MCAST);
2889 
2890 	if (tso) {
2891 		error = tcp_chopper(m, &ml, encif, len);
2892 		if (error)
2893 			goto done;
2894 	} else {
2895 		CLR(m->m_pkthdr.csum_flags, M_TCP_TSO);
2896 		in6_proto_cksum_out(m, encif);
2897 		ml_init(&ml);
2898 		ml_enqueue(&ml, m);
2899 	}
2900 
2901 	KERNEL_LOCK();
2902 	while ((m = ml_dequeue(&ml)) != NULL) {
2903 		/* Callee frees mbuf */
2904 		error = ipsp_process_packet(m, tdb, AF_INET6, tunalready);
2905 		if (error)
2906 			break;
2907 	}
2908 	KERNEL_UNLOCK();
2909  done:
2910 	if (error) {
2911 		ml_purge(&ml);
2912 		ipsecstat_inc(ipsec_odrops);
2913 		tdbstat_inc(tdb, tdb_odrops);
2914 	}
2915 	if (!error && tso)
2916 		tcpstat_inc(tcps_outswtso);
2917 	if (ip_mtudisc && error == EMSGSIZE)
2918 		ip6_output_ipsec_pmtu_update(tdb, ro, &dst, ifidx, rtableid, 0);
2919 	return error;
2920 }
2921 #endif /* IPSEC */
2922