xref: /freebsd/sys/netinet6/ip6_output.c (revision 8e648814b040d481f0429a2675e62af92b23bdd0)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
30  */
31 
32 /*-
33  * Copyright (c) 1982, 1986, 1988, 1990, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
61  */
62 
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
65 
66 #include "opt_inet.h"
67 #include "opt_inet6.h"
68 #include "opt_ipfw.h"
69 #include "opt_ipsec.h"
70 #include "opt_sctp.h"
71 #include "opt_route.h"
72 
73 #include <sys/param.h>
74 #include <sys/kernel.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/errno.h>
78 #include <sys/priv.h>
79 #include <sys/proc.h>
80 #include <sys/protosw.h>
81 #include <sys/socket.h>
82 #include <sys/socketvar.h>
83 #include <sys/syslog.h>
84 #include <sys/ucred.h>
85 
86 #include <machine/in_cksum.h>
87 
88 #include <net/if.h>
89 #include <net/if_var.h>
90 #include <net/netisr.h>
91 #include <net/route.h>
92 #include <net/pfil.h>
93 #include <net/vnet.h>
94 
95 #include <netinet/in.h>
96 #include <netinet/in_var.h>
97 #include <netinet/ip_var.h>
98 #include <netinet6/in6_var.h>
99 #include <netinet/ip6.h>
100 #include <netinet/icmp6.h>
101 #include <netinet6/ip6_var.h>
102 #include <netinet/in_pcb.h>
103 #include <netinet/tcp_var.h>
104 #include <netinet6/nd6.h>
105 
106 #ifdef IPSEC
107 #include <netipsec/ipsec.h>
108 #include <netipsec/ipsec6.h>
109 #include <netipsec/key.h>
110 #include <netinet6/ip6_ipsec.h>
111 #endif /* IPSEC */
112 #ifdef SCTP
113 #include <netinet/sctp.h>
114 #include <netinet/sctp_crc32.h>
115 #endif
116 
117 #include <netinet6/ip6protosw.h>
118 #include <netinet6/scope6_var.h>
119 
120 #ifdef FLOWTABLE
121 #include <net/flowtable.h>
122 #endif
123 
124 extern int in6_mcast_loop;
125 
126 struct ip6_exthdrs {
127 	struct mbuf *ip6e_ip6;
128 	struct mbuf *ip6e_hbh;
129 	struct mbuf *ip6e_dest1;
130 	struct mbuf *ip6e_rthdr;
131 	struct mbuf *ip6e_dest2;
132 };
133 
134 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
135 			   struct ucred *, int);
136 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
137 	struct socket *, struct sockopt *);
138 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
139 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
140 	struct ucred *, int, int, int);
141 
142 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
143 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
144 	struct ip6_frag **);
145 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
146 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
147 static int ip6_getpmtu(struct route_in6 *, struct route_in6 *,
148 	struct ifnet *, struct in6_addr *, u_long *, int *, u_int);
149 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
150 
151 
152 /*
153  * Make an extension header from option data.  hp is the source, and
154  * mp is the destination.
155  */
156 #define MAKE_EXTHDR(hp, mp)						\
157     do {								\
158 	if (hp) {							\
159 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
160 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
161 		    ((eh)->ip6e_len + 1) << 3);				\
162 		if (error)						\
163 			goto freehdrs;					\
164 	}								\
165     } while (/*CONSTCOND*/ 0)
166 
167 /*
168  * Form a chain of extension headers.
169  * m is the extension header mbuf
170  * mp is the previous mbuf in the chain
171  * p is the next header
172  * i is the type of option.
173  */
174 #define MAKE_CHAIN(m, mp, p, i)\
175     do {\
176 	if (m) {\
177 		if (!hdrsplit) \
178 			panic("assumption failed: hdr not split"); \
179 		*mtod((m), u_char *) = *(p);\
180 		*(p) = (i);\
181 		p = mtod((m), u_char *);\
182 		(m)->m_next = (mp)->m_next;\
183 		(mp)->m_next = (m);\
184 		(mp) = (m);\
185 	}\
186     } while (/*CONSTCOND*/ 0)
187 
188 void
189 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
190 {
191 	u_short csum;
192 
193 	csum = in_cksum_skip(m, offset + plen, offset);
194 	if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
195 		csum = 0xffff;
196 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
197 
198 	if (offset + sizeof(u_short) > m->m_len) {
199 		printf("%s: delayed m_pullup, m->len: %d plen %u off %u "
200 		    "csum_flags=%b\n", __func__, m->m_len, plen, offset,
201 		    (int)m->m_pkthdr.csum_flags, CSUM_BITS);
202 		/*
203 		 * XXX this should not happen, but if it does, the correct
204 		 * behavior may be to insert the checksum in the appropriate
205 		 * next mbuf in the chain.
206 		 */
207 		return;
208 	}
209 	*(u_short *)(m->m_data + offset) = csum;
210 }
211 
212 /*
213  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
214  * header (with pri, len, nxt, hlim, src, dst).
215  * This function may modify ver and hlim only.
216  * The mbuf chain containing the packet will be freed.
217  * The mbuf opt, if present, will not be freed.
218  * If route_in6 ro is present and has ro_rt initialized, route lookup would be
219  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
220  * then result of route lookup is stored in ro->ro_rt.
221  *
222  * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and
223  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
224  * which is rt_mtu.
225  *
226  * ifpp - XXX: just for statistics
227  */
228 int
229 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
230     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
231     struct ifnet **ifpp, struct inpcb *inp)
232 {
233 	struct ip6_hdr *ip6, *mhip6;
234 	struct ifnet *ifp, *origifp;
235 	struct mbuf *m = m0;
236 	struct mbuf *mprev = NULL;
237 	int hlen, tlen, len, off;
238 	struct route_in6 ip6route;
239 	struct rtentry *rt = NULL;
240 	struct sockaddr_in6 *dst, src_sa, dst_sa;
241 	struct in6_addr odst;
242 	int error = 0;
243 	struct in6_ifaddr *ia = NULL;
244 	u_long mtu;
245 	int alwaysfrag, dontfrag;
246 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
247 	struct ip6_exthdrs exthdrs;
248 	struct in6_addr finaldst, src0, dst0;
249 	u_int32_t zone;
250 	struct route_in6 *ro_pmtu = NULL;
251 	int hdrsplit = 0;
252 	int sw_csum, tso;
253 	struct m_tag *fwd_tag = NULL;
254 
255 	ip6 = mtod(m, struct ip6_hdr *);
256 	if (ip6 == NULL) {
257 		printf ("ip6 is NULL");
258 		goto bad;
259 	}
260 
261 	if (inp != NULL)
262 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
263 
264 	finaldst = ip6->ip6_dst;
265 	bzero(&exthdrs, sizeof(exthdrs));
266 	if (opt) {
267 		/* Hop-by-Hop options header */
268 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
269 		/* Destination options header(1st part) */
270 		if (opt->ip6po_rthdr) {
271 			/*
272 			 * Destination options header(1st part)
273 			 * This only makes sense with a routing header.
274 			 * See Section 9.2 of RFC 3542.
275 			 * Disabling this part just for MIP6 convenience is
276 			 * a bad idea.  We need to think carefully about a
277 			 * way to make the advanced API coexist with MIP6
278 			 * options, which might automatically be inserted in
279 			 * the kernel.
280 			 */
281 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
282 		}
283 		/* Routing header */
284 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
285 		/* Destination options header(2nd part) */
286 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
287 	}
288 
289 #ifdef IPSEC
290 	/*
291 	 * IPSec checking which handles several cases.
292 	 * FAST IPSEC: We re-injected the packet.
293 	 */
294 	switch(ip6_ipsec_output(&m, inp, &flags, &error, &ifp))
295 	{
296 	case 1:                 /* Bad packet */
297 		goto freehdrs;
298 	case -1:                /* IPSec done */
299 		goto done;
300 	case 0:                 /* No IPSec */
301 	default:
302 		break;
303 	}
304 #endif /* IPSEC */
305 
306 	/*
307 	 * Calculate the total length of the extension header chain.
308 	 * Keep the length of the unfragmentable part for fragmentation.
309 	 */
310 	optlen = 0;
311 	if (exthdrs.ip6e_hbh)
312 		optlen += exthdrs.ip6e_hbh->m_len;
313 	if (exthdrs.ip6e_dest1)
314 		optlen += exthdrs.ip6e_dest1->m_len;
315 	if (exthdrs.ip6e_rthdr)
316 		optlen += exthdrs.ip6e_rthdr->m_len;
317 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
318 
319 	/* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */
320 	if (exthdrs.ip6e_dest2)
321 		optlen += exthdrs.ip6e_dest2->m_len;
322 
323 	/*
324 	 * If there is at least one extension header,
325 	 * separate IP6 header from the payload.
326 	 */
327 	if (optlen && !hdrsplit) {
328 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
329 			m = NULL;
330 			goto freehdrs;
331 		}
332 		m = exthdrs.ip6e_ip6;
333 		hdrsplit++;
334 	}
335 
336 	/* adjust pointer */
337 	ip6 = mtod(m, struct ip6_hdr *);
338 
339 	/* adjust mbuf packet header length */
340 	m->m_pkthdr.len += optlen;
341 	plen = m->m_pkthdr.len - sizeof(*ip6);
342 
343 	/* If this is a jumbo payload, insert a jumbo payload option. */
344 	if (plen > IPV6_MAXPACKET) {
345 		if (!hdrsplit) {
346 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
347 				m = NULL;
348 				goto freehdrs;
349 			}
350 			m = exthdrs.ip6e_ip6;
351 			hdrsplit++;
352 		}
353 		/* adjust pointer */
354 		ip6 = mtod(m, struct ip6_hdr *);
355 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
356 			goto freehdrs;
357 		ip6->ip6_plen = 0;
358 	} else
359 		ip6->ip6_plen = htons(plen);
360 
361 	/*
362 	 * Concatenate headers and fill in next header fields.
363 	 * Here we have, on "m"
364 	 *	IPv6 payload
365 	 * and we insert headers accordingly.  Finally, we should be getting:
366 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
367 	 *
368 	 * during the header composing process, "m" points to IPv6 header.
369 	 * "mprev" points to an extension header prior to esp.
370 	 */
371 	u_char *nexthdrp = &ip6->ip6_nxt;
372 	mprev = m;
373 
374 	/*
375 	 * we treat dest2 specially.  this makes IPsec processing
376 	 * much easier.  the goal here is to make mprev point the
377 	 * mbuf prior to dest2.
378 	 *
379 	 * result: IPv6 dest2 payload
380 	 * m and mprev will point to IPv6 header.
381 	 */
382 	if (exthdrs.ip6e_dest2) {
383 		if (!hdrsplit)
384 			panic("assumption failed: hdr not split");
385 		exthdrs.ip6e_dest2->m_next = m->m_next;
386 		m->m_next = exthdrs.ip6e_dest2;
387 		*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
388 		ip6->ip6_nxt = IPPROTO_DSTOPTS;
389 	}
390 
391 	/*
392 	 * result: IPv6 hbh dest1 rthdr dest2 payload
393 	 * m will point to IPv6 header.  mprev will point to the
394 	 * extension header prior to dest2 (rthdr in the above case).
395 	 */
396 	MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
397 	MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
398 		   IPPROTO_DSTOPTS);
399 	MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
400 		   IPPROTO_ROUTING);
401 
402 	/*
403 	 * If there is a routing header, discard the packet.
404 	 */
405 	if (exthdrs.ip6e_rthdr) {
406 		 error = EINVAL;
407 		 goto bad;
408 	}
409 
410 	/* Source address validation */
411 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
412 	    (flags & IPV6_UNSPECSRC) == 0) {
413 		error = EOPNOTSUPP;
414 		IP6STAT_INC(ip6s_badscope);
415 		goto bad;
416 	}
417 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
418 		error = EOPNOTSUPP;
419 		IP6STAT_INC(ip6s_badscope);
420 		goto bad;
421 	}
422 
423 	IP6STAT_INC(ip6s_localout);
424 
425 	/*
426 	 * Route packet.
427 	 */
428 	if (ro == 0) {
429 		ro = &ip6route;
430 		bzero((caddr_t)ro, sizeof(*ro));
431 	}
432 	ro_pmtu = ro;
433 	if (opt && opt->ip6po_rthdr)
434 		ro = &opt->ip6po_route;
435 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
436 #ifdef FLOWTABLE
437 	if (ro->ro_rt == NULL)
438 		(void )flowtable_lookup(AF_INET6, m, (struct route *)ro);
439 #endif
440 again:
441 	/*
442 	 * if specified, try to fill in the traffic class field.
443 	 * do not override if a non-zero value is already set.
444 	 * we check the diffserv field and the ecn field separately.
445 	 */
446 	if (opt && opt->ip6po_tclass >= 0) {
447 		int mask = 0;
448 
449 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
450 			mask |= 0xfc;
451 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
452 			mask |= 0x03;
453 		if (mask != 0)
454 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
455 	}
456 
457 	/* fill in or override the hop limit field, if necessary. */
458 	if (opt && opt->ip6po_hlim != -1)
459 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
460 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
461 		if (im6o != NULL)
462 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
463 		else
464 			ip6->ip6_hlim = V_ip6_defmcasthlim;
465 	}
466 
467 	/* adjust pointer */
468 	ip6 = mtod(m, struct ip6_hdr *);
469 
470 	if (ro->ro_rt && fwd_tag == NULL) {
471 		rt = ro->ro_rt;
472 		ifp = ro->ro_rt->rt_ifp;
473 	} else {
474 		if (fwd_tag == NULL) {
475 			bzero(&dst_sa, sizeof(dst_sa));
476 			dst_sa.sin6_family = AF_INET6;
477 			dst_sa.sin6_len = sizeof(dst_sa);
478 			dst_sa.sin6_addr = ip6->ip6_dst;
479 		}
480 		error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
481 		    &rt, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
482 		if (error != 0) {
483 			if (ifp != NULL)
484 				in6_ifstat_inc(ifp, ifs6_out_discard);
485 			goto bad;
486 		}
487 	}
488 	if (rt == NULL) {
489 		/*
490 		 * If in6_selectroute() does not return a route entry,
491 		 * dst may not have been updated.
492 		 */
493 		*dst = dst_sa;	/* XXX */
494 	}
495 
496 	/*
497 	 * then rt (for unicast) and ifp must be non-NULL valid values.
498 	 */
499 	if ((flags & IPV6_FORWARDING) == 0) {
500 		/* XXX: the FORWARDING flag can be set for mrouting. */
501 		in6_ifstat_inc(ifp, ifs6_out_request);
502 	}
503 	if (rt != NULL) {
504 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
505 		counter_u64_add(rt->rt_pksent, 1);
506 	}
507 
508 
509 	/*
510 	 * The outgoing interface must be in the zone of source and
511 	 * destination addresses.
512 	 */
513 	origifp = ifp;
514 
515 	src0 = ip6->ip6_src;
516 	if (in6_setscope(&src0, origifp, &zone))
517 		goto badscope;
518 	bzero(&src_sa, sizeof(src_sa));
519 	src_sa.sin6_family = AF_INET6;
520 	src_sa.sin6_len = sizeof(src_sa);
521 	src_sa.sin6_addr = ip6->ip6_src;
522 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
523 		goto badscope;
524 
525 	dst0 = ip6->ip6_dst;
526 	if (in6_setscope(&dst0, origifp, &zone))
527 		goto badscope;
528 	/* re-initialize to be sure */
529 	bzero(&dst_sa, sizeof(dst_sa));
530 	dst_sa.sin6_family = AF_INET6;
531 	dst_sa.sin6_len = sizeof(dst_sa);
532 	dst_sa.sin6_addr = ip6->ip6_dst;
533 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
534 		goto badscope;
535 	}
536 
537 	/* We should use ia_ifp to support the case of
538 	 * sending packets to an address of our own.
539 	 */
540 	if (ia != NULL && ia->ia_ifp)
541 		ifp = ia->ia_ifp;
542 
543 	/* scope check is done. */
544 	goto routefound;
545 
546   badscope:
547 	IP6STAT_INC(ip6s_badscope);
548 	in6_ifstat_inc(origifp, ifs6_out_discard);
549 	if (error == 0)
550 		error = EHOSTUNREACH; /* XXX */
551 	goto bad;
552 
553   routefound:
554 	if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
555 		if (opt && opt->ip6po_nextroute.ro_rt) {
556 			/*
557 			 * The nexthop is explicitly specified by the
558 			 * application.  We assume the next hop is an IPv6
559 			 * address.
560 			 */
561 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
562 		}
563 		else if ((rt->rt_flags & RTF_GATEWAY))
564 			dst = (struct sockaddr_in6 *)rt->rt_gateway;
565 	}
566 
567 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
568 		m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
569 	} else {
570 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
571 		in6_ifstat_inc(ifp, ifs6_out_mcast);
572 		/*
573 		 * Confirm that the outgoing interface supports multicast.
574 		 */
575 		if (!(ifp->if_flags & IFF_MULTICAST)) {
576 			IP6STAT_INC(ip6s_noroute);
577 			in6_ifstat_inc(ifp, ifs6_out_discard);
578 			error = ENETUNREACH;
579 			goto bad;
580 		}
581 		if ((im6o == NULL && in6_mcast_loop) ||
582 		    (im6o && im6o->im6o_multicast_loop)) {
583 			/*
584 			 * Loop back multicast datagram if not expressly
585 			 * forbidden to do so, even if we have not joined
586 			 * the address; protocols will filter it later,
587 			 * thus deferring a hash lookup and lock acquisition
588 			 * at the expense of an m_copym().
589 			 */
590 			ip6_mloopback(ifp, m, dst);
591 		} else {
592 			/*
593 			 * If we are acting as a multicast router, perform
594 			 * multicast forwarding as if the packet had just
595 			 * arrived on the interface to which we are about
596 			 * to send.  The multicast forwarding function
597 			 * recursively calls this function, using the
598 			 * IPV6_FORWARDING flag to prevent infinite recursion.
599 			 *
600 			 * Multicasts that are looped back by ip6_mloopback(),
601 			 * above, will be forwarded by the ip6_input() routine,
602 			 * if necessary.
603 			 */
604 			if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
605 				/*
606 				 * XXX: ip6_mforward expects that rcvif is NULL
607 				 * when it is called from the originating path.
608 				 * However, it may not always be the case.
609 				 */
610 				m->m_pkthdr.rcvif = NULL;
611 				if (ip6_mforward(ip6, ifp, m) != 0) {
612 					m_freem(m);
613 					goto done;
614 				}
615 			}
616 		}
617 		/*
618 		 * Multicasts with a hoplimit of zero may be looped back,
619 		 * above, but must not be transmitted on a network.
620 		 * Also, multicasts addressed to the loopback interface
621 		 * are not sent -- the above call to ip6_mloopback() will
622 		 * loop back a copy if this host actually belongs to the
623 		 * destination group on the loopback interface.
624 		 */
625 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
626 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
627 			m_freem(m);
628 			goto done;
629 		}
630 	}
631 
632 	/*
633 	 * Fill the outgoing inteface to tell the upper layer
634 	 * to increment per-interface statistics.
635 	 */
636 	if (ifpp)
637 		*ifpp = ifp;
638 
639 	/* Determine path MTU. */
640 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
641 	    &alwaysfrag, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m))) != 0)
642 		goto bad;
643 
644 	/*
645 	 * The caller of this function may specify to use the minimum MTU
646 	 * in some cases.
647 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
648 	 * setting.  The logic is a bit complicated; by default, unicast
649 	 * packets will follow path MTU while multicast packets will be sent at
650 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
651 	 * including unicast ones will be sent at the minimum MTU.  Multicast
652 	 * packets will always be sent at the minimum MTU unless
653 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
654 	 * See RFC 3542 for more details.
655 	 */
656 	if (mtu > IPV6_MMTU) {
657 		if ((flags & IPV6_MINMTU))
658 			mtu = IPV6_MMTU;
659 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
660 			mtu = IPV6_MMTU;
661 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
662 			 (opt == NULL ||
663 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
664 			mtu = IPV6_MMTU;
665 		}
666 	}
667 
668 	/*
669 	 * clear embedded scope identifiers if necessary.
670 	 * in6_clearscope will touch the addresses only when necessary.
671 	 */
672 	in6_clearscope(&ip6->ip6_src);
673 	in6_clearscope(&ip6->ip6_dst);
674 
675 	/*
676 	 * If the outgoing packet contains a hop-by-hop options header,
677 	 * it must be examined and processed even by the source node.
678 	 * (RFC 2460, section 4.)
679 	 */
680 	if (exthdrs.ip6e_hbh) {
681 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
682 		u_int32_t dummy; /* XXX unused */
683 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
684 
685 #ifdef DIAGNOSTIC
686 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
687 			panic("ip6e_hbh is not contiguous");
688 #endif
689 		/*
690 		 *  XXX: if we have to send an ICMPv6 error to the sender,
691 		 *       we need the M_LOOP flag since icmp6_error() expects
692 		 *       the IPv6 and the hop-by-hop options header are
693 		 *       contiguous unless the flag is set.
694 		 */
695 		m->m_flags |= M_LOOP;
696 		m->m_pkthdr.rcvif = ifp;
697 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
698 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
699 		    &dummy, &plen) < 0) {
700 			/* m was already freed at this point */
701 			error = EINVAL;/* better error? */
702 			goto done;
703 		}
704 		m->m_flags &= ~M_LOOP; /* XXX */
705 		m->m_pkthdr.rcvif = NULL;
706 	}
707 
708 	/* Jump over all PFIL processing if hooks are not active. */
709 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
710 		goto passout;
711 
712 	odst = ip6->ip6_dst;
713 	/* Run through list of hooks for output packets. */
714 	error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
715 	if (error != 0 || m == NULL)
716 		goto done;
717 	ip6 = mtod(m, struct ip6_hdr *);
718 
719 	/* See if destination IP address was changed by packet filter. */
720 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
721 		m->m_flags |= M_SKIP_FIREWALL;
722 		/* If destination is now ourself drop to ip6_input(). */
723 		if (in6_localip(&ip6->ip6_dst)) {
724 			m->m_flags |= M_FASTFWD_OURS;
725 			if (m->m_pkthdr.rcvif == NULL)
726 				m->m_pkthdr.rcvif = V_loif;
727 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
728 				m->m_pkthdr.csum_flags |=
729 				    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
730 				m->m_pkthdr.csum_data = 0xffff;
731 			}
732 #ifdef SCTP
733 			if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
734 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
735 #endif
736 			error = netisr_queue(NETISR_IPV6, m);
737 			goto done;
738 		} else
739 			goto again;	/* Redo the routing table lookup. */
740 	}
741 
742 	/* See if local, if yes, send it to netisr. */
743 	if (m->m_flags & M_FASTFWD_OURS) {
744 		if (m->m_pkthdr.rcvif == NULL)
745 			m->m_pkthdr.rcvif = V_loif;
746 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
747 			m->m_pkthdr.csum_flags |=
748 			    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
749 			m->m_pkthdr.csum_data = 0xffff;
750 		}
751 #ifdef SCTP
752 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
753 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
754 #endif
755 		error = netisr_queue(NETISR_IPV6, m);
756 		goto done;
757 	}
758 	/* Or forward to some other address? */
759 	if ((m->m_flags & M_IP6_NEXTHOP) &&
760 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
761 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
762 		bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
763 		m->m_flags |= M_SKIP_FIREWALL;
764 		m->m_flags &= ~M_IP6_NEXTHOP;
765 		m_tag_delete(m, fwd_tag);
766 		goto again;
767 	}
768 
769 passout:
770 	/*
771 	 * Send the packet to the outgoing interface.
772 	 * If necessary, do IPv6 fragmentation before sending.
773 	 *
774 	 * the logic here is rather complex:
775 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
776 	 * 1-a:	send as is if tlen <= path mtu
777 	 * 1-b:	fragment if tlen > path mtu
778 	 *
779 	 * 2: if user asks us not to fragment (dontfrag == 1)
780 	 * 2-a:	send as is if tlen <= interface mtu
781 	 * 2-b:	error if tlen > interface mtu
782 	 *
783 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
784 	 *	always fragment
785 	 *
786 	 * 4: if dontfrag == 1 && alwaysfrag == 1
787 	 *	error, as we cannot handle this conflicting request
788 	 */
789 	sw_csum = m->m_pkthdr.csum_flags;
790 	if (!hdrsplit) {
791 		tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
792 		sw_csum &= ~ifp->if_hwassist;
793 	} else
794 		tso = 0;
795 	/*
796 	 * If we added extension headers, we will not do TSO and calculate the
797 	 * checksums ourselves for now.
798 	 * XXX-BZ  Need a framework to know when the NIC can handle it, even
799 	 * with ext. hdrs.
800 	 */
801 	if (sw_csum & CSUM_DELAY_DATA_IPV6) {
802 		sw_csum &= ~CSUM_DELAY_DATA_IPV6;
803 		in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
804 	}
805 #ifdef SCTP
806 	if (sw_csum & CSUM_SCTP_IPV6) {
807 		sw_csum &= ~CSUM_SCTP_IPV6;
808 		sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
809 	}
810 #endif
811 	m->m_pkthdr.csum_flags &= ifp->if_hwassist;
812 	tlen = m->m_pkthdr.len;
813 
814 	if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
815 		dontfrag = 1;
816 	else
817 		dontfrag = 0;
818 	if (dontfrag && alwaysfrag) {	/* case 4 */
819 		/* conflicting request - can't transmit */
820 		error = EMSGSIZE;
821 		goto bad;
822 	}
823 	if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {	/* case 2-b */
824 		/*
825 		 * Even if the DONTFRAG option is specified, we cannot send the
826 		 * packet when the data length is larger than the MTU of the
827 		 * outgoing interface.
828 		 * Notify the error by sending IPV6_PATHMTU ancillary data as
829 		 * well as returning an error code (the latter is not described
830 		 * in the API spec.)
831 		 */
832 		u_int32_t mtu32;
833 		struct ip6ctlparam ip6cp;
834 
835 		mtu32 = (u_int32_t)mtu;
836 		bzero(&ip6cp, sizeof(ip6cp));
837 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
838 		pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
839 		    (void *)&ip6cp);
840 
841 		error = EMSGSIZE;
842 		goto bad;
843 	}
844 
845 	/*
846 	 * transmit packet without fragmentation
847 	 */
848 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
849 		struct in6_ifaddr *ia6;
850 
851 		ip6 = mtod(m, struct ip6_hdr *);
852 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
853 		if (ia6) {
854 			/* Record statistics for this interface address. */
855 			counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
856 			counter_u64_add(ia6->ia_ifa.ifa_obytes,
857 			    m->m_pkthdr.len);
858 			ifa_free(&ia6->ia_ifa);
859 		}
860 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
861 		goto done;
862 	}
863 
864 	/*
865 	 * try to fragment the packet.  case 1-b and 3
866 	 */
867 	if (mtu < IPV6_MMTU) {
868 		/* path MTU cannot be less than IPV6_MMTU */
869 		error = EMSGSIZE;
870 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
871 		goto bad;
872 	} else if (ip6->ip6_plen == 0) {
873 		/* jumbo payload cannot be fragmented */
874 		error = EMSGSIZE;
875 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
876 		goto bad;
877 	} else {
878 		struct mbuf **mnext, *m_frgpart;
879 		struct ip6_frag *ip6f;
880 		u_int32_t id = htonl(ip6_randomid());
881 		u_char nextproto;
882 
883 		int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
884 
885 		/*
886 		 * Too large for the destination or interface;
887 		 * fragment if possible.
888 		 * Must be able to put at least 8 bytes per fragment.
889 		 */
890 		hlen = unfragpartlen;
891 		if (mtu > IPV6_MAXPACKET)
892 			mtu = IPV6_MAXPACKET;
893 
894 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
895 		if (len < 8) {
896 			error = EMSGSIZE;
897 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
898 			goto bad;
899 		}
900 
901 		/*
902 		 * Verify that we have any chance at all of being able to queue
903 		 *      the packet or packet fragments
904 		 */
905 		if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
906 		    < tlen  /* - hlen */)) {
907 			error = ENOBUFS;
908 			IP6STAT_INC(ip6s_odropped);
909 			goto bad;
910 		}
911 
912 
913 		/*
914 		 * If the interface will not calculate checksums on
915 		 * fragmented packets, then do it here.
916 		 * XXX-BZ handle the hw offloading case.  Need flags.
917 		 */
918 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
919 			in6_delayed_cksum(m, plen, hlen);
920 			m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
921 		}
922 #ifdef SCTP
923 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
924 			sctp_delayed_cksum(m, hlen);
925 			m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
926 		}
927 #endif
928 		mnext = &m->m_nextpkt;
929 
930 		/*
931 		 * Change the next header field of the last header in the
932 		 * unfragmentable part.
933 		 */
934 		if (exthdrs.ip6e_rthdr) {
935 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
936 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
937 		} else if (exthdrs.ip6e_dest1) {
938 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
939 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
940 		} else if (exthdrs.ip6e_hbh) {
941 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
942 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
943 		} else {
944 			nextproto = ip6->ip6_nxt;
945 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
946 		}
947 
948 		/*
949 		 * Loop through length of segment after first fragment,
950 		 * make new header and copy data of each part and link onto
951 		 * chain.
952 		 */
953 		m0 = m;
954 		for (off = hlen; off < tlen; off += len) {
955 			m = m_gethdr(M_NOWAIT, MT_DATA);
956 			if (!m) {
957 				error = ENOBUFS;
958 				IP6STAT_INC(ip6s_odropped);
959 				goto sendorfree;
960 			}
961 			m->m_flags = m0->m_flags & M_COPYFLAGS;
962 			*mnext = m;
963 			mnext = &m->m_nextpkt;
964 			m->m_data += max_linkhdr;
965 			mhip6 = mtod(m, struct ip6_hdr *);
966 			*mhip6 = *ip6;
967 			m->m_len = sizeof(*mhip6);
968 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
969 			if (error) {
970 				IP6STAT_INC(ip6s_odropped);
971 				goto sendorfree;
972 			}
973 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
974 			if (off + len >= tlen)
975 				len = tlen - off;
976 			else
977 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
978 			mhip6->ip6_plen = htons((u_short)(len + hlen +
979 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
980 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
981 				error = ENOBUFS;
982 				IP6STAT_INC(ip6s_odropped);
983 				goto sendorfree;
984 			}
985 			m_cat(m, m_frgpart);
986 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
987 			m->m_pkthdr.fibnum = m0->m_pkthdr.fibnum;
988 			m->m_pkthdr.rcvif = NULL;
989 			ip6f->ip6f_reserved = 0;
990 			ip6f->ip6f_ident = id;
991 			ip6f->ip6f_nxt = nextproto;
992 			IP6STAT_INC(ip6s_ofragments);
993 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
994 		}
995 
996 		in6_ifstat_inc(ifp, ifs6_out_fragok);
997 	}
998 
999 	/*
1000 	 * Remove leading garbages.
1001 	 */
1002 sendorfree:
1003 	m = m0->m_nextpkt;
1004 	m0->m_nextpkt = 0;
1005 	m_freem(m0);
1006 	for (m0 = m; m; m = m0) {
1007 		m0 = m->m_nextpkt;
1008 		m->m_nextpkt = 0;
1009 		if (error == 0) {
1010 			/* Record statistics for this interface address. */
1011 			if (ia) {
1012 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1013 				counter_u64_add(ia->ia_ifa.ifa_obytes,
1014 				    m->m_pkthdr.len);
1015 			}
1016 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1017 		} else
1018 			m_freem(m);
1019 	}
1020 
1021 	if (error == 0)
1022 		IP6STAT_INC(ip6s_fragmented);
1023 
1024 done:
1025 	if (ro == &ip6route)
1026 		RO_RTFREE(ro);
1027 	if (ro_pmtu == &ip6route)
1028 		RO_RTFREE(ro_pmtu);
1029 	return (error);
1030 
1031 freehdrs:
1032 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1033 	m_freem(exthdrs.ip6e_dest1);
1034 	m_freem(exthdrs.ip6e_rthdr);
1035 	m_freem(exthdrs.ip6e_dest2);
1036 	/* FALLTHROUGH */
1037 bad:
1038 	if (m)
1039 		m_freem(m);
1040 	goto done;
1041 }
1042 
1043 static int
1044 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1045 {
1046 	struct mbuf *m;
1047 
1048 	if (hlen > MCLBYTES)
1049 		return (ENOBUFS); /* XXX */
1050 
1051 	if (hlen > MLEN)
1052 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1053 	else
1054 		m = m_get(M_NOWAIT, MT_DATA);
1055 	if (m == NULL)
1056 		return (ENOBUFS);
1057 	m->m_len = hlen;
1058 	if (hdr)
1059 		bcopy(hdr, mtod(m, caddr_t), hlen);
1060 
1061 	*mp = m;
1062 	return (0);
1063 }
1064 
1065 /*
1066  * Insert jumbo payload option.
1067  */
1068 static int
1069 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1070 {
1071 	struct mbuf *mopt;
1072 	u_char *optbuf;
1073 	u_int32_t v;
1074 
1075 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1076 
1077 	/*
1078 	 * If there is no hop-by-hop options header, allocate new one.
1079 	 * If there is one but it doesn't have enough space to store the
1080 	 * jumbo payload option, allocate a cluster to store the whole options.
1081 	 * Otherwise, use it to store the options.
1082 	 */
1083 	if (exthdrs->ip6e_hbh == 0) {
1084 		mopt = m_get(M_NOWAIT, MT_DATA);
1085 		if (mopt == NULL)
1086 			return (ENOBUFS);
1087 		mopt->m_len = JUMBOOPTLEN;
1088 		optbuf = mtod(mopt, u_char *);
1089 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1090 		exthdrs->ip6e_hbh = mopt;
1091 	} else {
1092 		struct ip6_hbh *hbh;
1093 
1094 		mopt = exthdrs->ip6e_hbh;
1095 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1096 			/*
1097 			 * XXX assumption:
1098 			 * - exthdrs->ip6e_hbh is not referenced from places
1099 			 *   other than exthdrs.
1100 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1101 			 */
1102 			int oldoptlen = mopt->m_len;
1103 			struct mbuf *n;
1104 
1105 			/*
1106 			 * XXX: give up if the whole (new) hbh header does
1107 			 * not fit even in an mbuf cluster.
1108 			 */
1109 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1110 				return (ENOBUFS);
1111 
1112 			/*
1113 			 * As a consequence, we must always prepare a cluster
1114 			 * at this point.
1115 			 */
1116 			n = m_getcl(M_NOWAIT, MT_DATA, 0);
1117 			if (n == NULL)
1118 				return (ENOBUFS);
1119 			n->m_len = oldoptlen + JUMBOOPTLEN;
1120 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1121 			    oldoptlen);
1122 			optbuf = mtod(n, caddr_t) + oldoptlen;
1123 			m_freem(mopt);
1124 			mopt = exthdrs->ip6e_hbh = n;
1125 		} else {
1126 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1127 			mopt->m_len += JUMBOOPTLEN;
1128 		}
1129 		optbuf[0] = IP6OPT_PADN;
1130 		optbuf[1] = 1;
1131 
1132 		/*
1133 		 * Adjust the header length according to the pad and
1134 		 * the jumbo payload option.
1135 		 */
1136 		hbh = mtod(mopt, struct ip6_hbh *);
1137 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1138 	}
1139 
1140 	/* fill in the option. */
1141 	optbuf[2] = IP6OPT_JUMBO;
1142 	optbuf[3] = 4;
1143 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1144 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1145 
1146 	/* finally, adjust the packet header length */
1147 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1148 
1149 	return (0);
1150 #undef JUMBOOPTLEN
1151 }
1152 
1153 /*
1154  * Insert fragment header and copy unfragmentable header portions.
1155  */
1156 static int
1157 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1158     struct ip6_frag **frghdrp)
1159 {
1160 	struct mbuf *n, *mlast;
1161 
1162 	if (hlen > sizeof(struct ip6_hdr)) {
1163 		n = m_copym(m0, sizeof(struct ip6_hdr),
1164 		    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1165 		if (n == 0)
1166 			return (ENOBUFS);
1167 		m->m_next = n;
1168 	} else
1169 		n = m;
1170 
1171 	/* Search for the last mbuf of unfragmentable part. */
1172 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1173 		;
1174 
1175 	if ((mlast->m_flags & M_EXT) == 0 &&
1176 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1177 		/* use the trailing space of the last mbuf for the fragment hdr */
1178 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1179 		    mlast->m_len);
1180 		mlast->m_len += sizeof(struct ip6_frag);
1181 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1182 	} else {
1183 		/* allocate a new mbuf for the fragment header */
1184 		struct mbuf *mfrg;
1185 
1186 		mfrg = m_get(M_NOWAIT, MT_DATA);
1187 		if (mfrg == NULL)
1188 			return (ENOBUFS);
1189 		mfrg->m_len = sizeof(struct ip6_frag);
1190 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1191 		mlast->m_next = mfrg;
1192 	}
1193 
1194 	return (0);
1195 }
1196 
1197 static int
1198 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1199     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1200     int *alwaysfragp, u_int fibnum)
1201 {
1202 	u_int32_t mtu = 0;
1203 	int alwaysfrag = 0;
1204 	int error = 0;
1205 
1206 	if (ro_pmtu != ro) {
1207 		/* The first hop and the final destination may differ. */
1208 		struct sockaddr_in6 *sa6_dst =
1209 		    (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1210 		if (ro_pmtu->ro_rt &&
1211 		    ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1212 		     !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1213 			RTFREE(ro_pmtu->ro_rt);
1214 			ro_pmtu->ro_rt = (struct rtentry *)NULL;
1215 		}
1216 		if (ro_pmtu->ro_rt == NULL) {
1217 			bzero(sa6_dst, sizeof(*sa6_dst));
1218 			sa6_dst->sin6_family = AF_INET6;
1219 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1220 			sa6_dst->sin6_addr = *dst;
1221 
1222 			in6_rtalloc(ro_pmtu, fibnum);
1223 		}
1224 	}
1225 	if (ro_pmtu->ro_rt) {
1226 		u_int32_t ifmtu;
1227 		struct in_conninfo inc;
1228 
1229 		bzero(&inc, sizeof(inc));
1230 		inc.inc_flags |= INC_ISIPV6;
1231 		inc.inc6_faddr = *dst;
1232 
1233 		if (ifp == NULL)
1234 			ifp = ro_pmtu->ro_rt->rt_ifp;
1235 		ifmtu = IN6_LINKMTU(ifp);
1236 		mtu = tcp_hc_getmtu(&inc);
1237 		if (mtu)
1238 			mtu = min(mtu, ro_pmtu->ro_rt->rt_mtu);
1239 		else
1240 			mtu = ro_pmtu->ro_rt->rt_mtu;
1241 		if (mtu == 0)
1242 			mtu = ifmtu;
1243 		else if (mtu < IPV6_MMTU) {
1244 			/*
1245 			 * RFC2460 section 5, last paragraph:
1246 			 * if we record ICMPv6 too big message with
1247 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1248 			 * or smaller, with framgent header attached.
1249 			 * (fragment header is needed regardless from the
1250 			 * packet size, for translators to identify packets)
1251 			 */
1252 			alwaysfrag = 1;
1253 			mtu = IPV6_MMTU;
1254 		} else if (mtu > ifmtu) {
1255 			/*
1256 			 * The MTU on the route is larger than the MTU on
1257 			 * the interface!  This shouldn't happen, unless the
1258 			 * MTU of the interface has been changed after the
1259 			 * interface was brought up.  Change the MTU in the
1260 			 * route to match the interface MTU (as long as the
1261 			 * field isn't locked).
1262 			 */
1263 			mtu = ifmtu;
1264 			ro_pmtu->ro_rt->rt_mtu = mtu;
1265 		}
1266 	} else if (ifp) {
1267 		mtu = IN6_LINKMTU(ifp);
1268 	} else
1269 		error = EHOSTUNREACH; /* XXX */
1270 
1271 	*mtup = mtu;
1272 	if (alwaysfragp)
1273 		*alwaysfragp = alwaysfrag;
1274 	return (error);
1275 }
1276 
1277 /*
1278  * IP6 socket option processing.
1279  */
1280 int
1281 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1282 {
1283 	int optdatalen, uproto;
1284 	void *optdata;
1285 	struct inpcb *in6p = sotoinpcb(so);
1286 	int error, optval;
1287 	int level, op, optname;
1288 	int optlen;
1289 	struct thread *td;
1290 
1291 	level = sopt->sopt_level;
1292 	op = sopt->sopt_dir;
1293 	optname = sopt->sopt_name;
1294 	optlen = sopt->sopt_valsize;
1295 	td = sopt->sopt_td;
1296 	error = 0;
1297 	optval = 0;
1298 	uproto = (int)so->so_proto->pr_protocol;
1299 
1300 	if (level != IPPROTO_IPV6) {
1301 		error = EINVAL;
1302 
1303 		if (sopt->sopt_level == SOL_SOCKET &&
1304 		    sopt->sopt_dir == SOPT_SET) {
1305 			switch (sopt->sopt_name) {
1306 			case SO_REUSEADDR:
1307 				INP_WLOCK(in6p);
1308 				if ((so->so_options & SO_REUSEADDR) != 0)
1309 					in6p->inp_flags2 |= INP_REUSEADDR;
1310 				else
1311 					in6p->inp_flags2 &= ~INP_REUSEADDR;
1312 				INP_WUNLOCK(in6p);
1313 				error = 0;
1314 				break;
1315 			case SO_REUSEPORT:
1316 				INP_WLOCK(in6p);
1317 				if ((so->so_options & SO_REUSEPORT) != 0)
1318 					in6p->inp_flags2 |= INP_REUSEPORT;
1319 				else
1320 					in6p->inp_flags2 &= ~INP_REUSEPORT;
1321 				INP_WUNLOCK(in6p);
1322 				error = 0;
1323 				break;
1324 			case SO_SETFIB:
1325 				INP_WLOCK(in6p);
1326 				in6p->inp_inc.inc_fibnum = so->so_fibnum;
1327 				INP_WUNLOCK(in6p);
1328 				error = 0;
1329 				break;
1330 			default:
1331 				break;
1332 			}
1333 		}
1334 	} else {		/* level == IPPROTO_IPV6 */
1335 		switch (op) {
1336 
1337 		case SOPT_SET:
1338 			switch (optname) {
1339 			case IPV6_2292PKTOPTIONS:
1340 #ifdef IPV6_PKTOPTIONS
1341 			case IPV6_PKTOPTIONS:
1342 #endif
1343 			{
1344 				struct mbuf *m;
1345 
1346 				error = soopt_getm(sopt, &m); /* XXX */
1347 				if (error != 0)
1348 					break;
1349 				error = soopt_mcopyin(sopt, m); /* XXX */
1350 				if (error != 0)
1351 					break;
1352 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1353 						    m, so, sopt);
1354 				m_freem(m); /* XXX */
1355 				break;
1356 			}
1357 
1358 			/*
1359 			 * Use of some Hop-by-Hop options or some
1360 			 * Destination options, might require special
1361 			 * privilege.  That is, normal applications
1362 			 * (without special privilege) might be forbidden
1363 			 * from setting certain options in outgoing packets,
1364 			 * and might never see certain options in received
1365 			 * packets. [RFC 2292 Section 6]
1366 			 * KAME specific note:
1367 			 *  KAME prevents non-privileged users from sending or
1368 			 *  receiving ANY hbh/dst options in order to avoid
1369 			 *  overhead of parsing options in the kernel.
1370 			 */
1371 			case IPV6_RECVHOPOPTS:
1372 			case IPV6_RECVDSTOPTS:
1373 			case IPV6_RECVRTHDRDSTOPTS:
1374 				if (td != NULL) {
1375 					error = priv_check(td,
1376 					    PRIV_NETINET_SETHDROPTS);
1377 					if (error)
1378 						break;
1379 				}
1380 				/* FALLTHROUGH */
1381 			case IPV6_UNICAST_HOPS:
1382 			case IPV6_HOPLIMIT:
1383 			case IPV6_FAITH:
1384 
1385 			case IPV6_RECVPKTINFO:
1386 			case IPV6_RECVHOPLIMIT:
1387 			case IPV6_RECVRTHDR:
1388 			case IPV6_RECVPATHMTU:
1389 			case IPV6_RECVTCLASS:
1390 			case IPV6_V6ONLY:
1391 			case IPV6_AUTOFLOWLABEL:
1392 			case IPV6_BINDANY:
1393 				if (optname == IPV6_BINDANY && td != NULL) {
1394 					error = priv_check(td,
1395 					    PRIV_NETINET_BINDANY);
1396 					if (error)
1397 						break;
1398 				}
1399 
1400 				if (optlen != sizeof(int)) {
1401 					error = EINVAL;
1402 					break;
1403 				}
1404 				error = sooptcopyin(sopt, &optval,
1405 					sizeof optval, sizeof optval);
1406 				if (error)
1407 					break;
1408 				switch (optname) {
1409 
1410 				case IPV6_UNICAST_HOPS:
1411 					if (optval < -1 || optval >= 256)
1412 						error = EINVAL;
1413 					else {
1414 						/* -1 = kernel default */
1415 						in6p->in6p_hops = optval;
1416 						if ((in6p->inp_vflag &
1417 						     INP_IPV4) != 0)
1418 							in6p->inp_ip_ttl = optval;
1419 					}
1420 					break;
1421 #define OPTSET(bit) \
1422 do { \
1423 	INP_WLOCK(in6p); \
1424 	if (optval) \
1425 		in6p->inp_flags |= (bit); \
1426 	else \
1427 		in6p->inp_flags &= ~(bit); \
1428 	INP_WUNLOCK(in6p); \
1429 } while (/*CONSTCOND*/ 0)
1430 #define OPTSET2292(bit) \
1431 do { \
1432 	INP_WLOCK(in6p); \
1433 	in6p->inp_flags |= IN6P_RFC2292; \
1434 	if (optval) \
1435 		in6p->inp_flags |= (bit); \
1436 	else \
1437 		in6p->inp_flags &= ~(bit); \
1438 	INP_WUNLOCK(in6p); \
1439 } while (/*CONSTCOND*/ 0)
1440 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1441 
1442 				case IPV6_RECVPKTINFO:
1443 					/* cannot mix with RFC2292 */
1444 					if (OPTBIT(IN6P_RFC2292)) {
1445 						error = EINVAL;
1446 						break;
1447 					}
1448 					OPTSET(IN6P_PKTINFO);
1449 					break;
1450 
1451 				case IPV6_HOPLIMIT:
1452 				{
1453 					struct ip6_pktopts **optp;
1454 
1455 					/* cannot mix with RFC2292 */
1456 					if (OPTBIT(IN6P_RFC2292)) {
1457 						error = EINVAL;
1458 						break;
1459 					}
1460 					optp = &in6p->in6p_outputopts;
1461 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1462 					    (u_char *)&optval, sizeof(optval),
1463 					    optp, (td != NULL) ? td->td_ucred :
1464 					    NULL, uproto);
1465 					break;
1466 				}
1467 
1468 				case IPV6_RECVHOPLIMIT:
1469 					/* cannot mix with RFC2292 */
1470 					if (OPTBIT(IN6P_RFC2292)) {
1471 						error = EINVAL;
1472 						break;
1473 					}
1474 					OPTSET(IN6P_HOPLIMIT);
1475 					break;
1476 
1477 				case IPV6_RECVHOPOPTS:
1478 					/* cannot mix with RFC2292 */
1479 					if (OPTBIT(IN6P_RFC2292)) {
1480 						error = EINVAL;
1481 						break;
1482 					}
1483 					OPTSET(IN6P_HOPOPTS);
1484 					break;
1485 
1486 				case IPV6_RECVDSTOPTS:
1487 					/* cannot mix with RFC2292 */
1488 					if (OPTBIT(IN6P_RFC2292)) {
1489 						error = EINVAL;
1490 						break;
1491 					}
1492 					OPTSET(IN6P_DSTOPTS);
1493 					break;
1494 
1495 				case IPV6_RECVRTHDRDSTOPTS:
1496 					/* cannot mix with RFC2292 */
1497 					if (OPTBIT(IN6P_RFC2292)) {
1498 						error = EINVAL;
1499 						break;
1500 					}
1501 					OPTSET(IN6P_RTHDRDSTOPTS);
1502 					break;
1503 
1504 				case IPV6_RECVRTHDR:
1505 					/* cannot mix with RFC2292 */
1506 					if (OPTBIT(IN6P_RFC2292)) {
1507 						error = EINVAL;
1508 						break;
1509 					}
1510 					OPTSET(IN6P_RTHDR);
1511 					break;
1512 
1513 				case IPV6_FAITH:
1514 					OPTSET(INP_FAITH);
1515 					break;
1516 
1517 				case IPV6_RECVPATHMTU:
1518 					/*
1519 					 * We ignore this option for TCP
1520 					 * sockets.
1521 					 * (RFC3542 leaves this case
1522 					 * unspecified.)
1523 					 */
1524 					if (uproto != IPPROTO_TCP)
1525 						OPTSET(IN6P_MTU);
1526 					break;
1527 
1528 				case IPV6_V6ONLY:
1529 					/*
1530 					 * make setsockopt(IPV6_V6ONLY)
1531 					 * available only prior to bind(2).
1532 					 * see ipng mailing list, Jun 22 2001.
1533 					 */
1534 					if (in6p->inp_lport ||
1535 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1536 						error = EINVAL;
1537 						break;
1538 					}
1539 					OPTSET(IN6P_IPV6_V6ONLY);
1540 					if (optval)
1541 						in6p->inp_vflag &= ~INP_IPV4;
1542 					else
1543 						in6p->inp_vflag |= INP_IPV4;
1544 					break;
1545 				case IPV6_RECVTCLASS:
1546 					/* cannot mix with RFC2292 XXX */
1547 					if (OPTBIT(IN6P_RFC2292)) {
1548 						error = EINVAL;
1549 						break;
1550 					}
1551 					OPTSET(IN6P_TCLASS);
1552 					break;
1553 				case IPV6_AUTOFLOWLABEL:
1554 					OPTSET(IN6P_AUTOFLOWLABEL);
1555 					break;
1556 
1557 				case IPV6_BINDANY:
1558 					OPTSET(INP_BINDANY);
1559 					break;
1560 				}
1561 				break;
1562 
1563 			case IPV6_TCLASS:
1564 			case IPV6_DONTFRAG:
1565 			case IPV6_USE_MIN_MTU:
1566 			case IPV6_PREFER_TEMPADDR:
1567 				if (optlen != sizeof(optval)) {
1568 					error = EINVAL;
1569 					break;
1570 				}
1571 				error = sooptcopyin(sopt, &optval,
1572 					sizeof optval, sizeof optval);
1573 				if (error)
1574 					break;
1575 				{
1576 					struct ip6_pktopts **optp;
1577 					optp = &in6p->in6p_outputopts;
1578 					error = ip6_pcbopt(optname,
1579 					    (u_char *)&optval, sizeof(optval),
1580 					    optp, (td != NULL) ? td->td_ucred :
1581 					    NULL, uproto);
1582 					break;
1583 				}
1584 
1585 			case IPV6_2292PKTINFO:
1586 			case IPV6_2292HOPLIMIT:
1587 			case IPV6_2292HOPOPTS:
1588 			case IPV6_2292DSTOPTS:
1589 			case IPV6_2292RTHDR:
1590 				/* RFC 2292 */
1591 				if (optlen != sizeof(int)) {
1592 					error = EINVAL;
1593 					break;
1594 				}
1595 				error = sooptcopyin(sopt, &optval,
1596 					sizeof optval, sizeof optval);
1597 				if (error)
1598 					break;
1599 				switch (optname) {
1600 				case IPV6_2292PKTINFO:
1601 					OPTSET2292(IN6P_PKTINFO);
1602 					break;
1603 				case IPV6_2292HOPLIMIT:
1604 					OPTSET2292(IN6P_HOPLIMIT);
1605 					break;
1606 				case IPV6_2292HOPOPTS:
1607 					/*
1608 					 * Check super-user privilege.
1609 					 * See comments for IPV6_RECVHOPOPTS.
1610 					 */
1611 					if (td != NULL) {
1612 						error = priv_check(td,
1613 						    PRIV_NETINET_SETHDROPTS);
1614 						if (error)
1615 							return (error);
1616 					}
1617 					OPTSET2292(IN6P_HOPOPTS);
1618 					break;
1619 				case IPV6_2292DSTOPTS:
1620 					if (td != NULL) {
1621 						error = priv_check(td,
1622 						    PRIV_NETINET_SETHDROPTS);
1623 						if (error)
1624 							return (error);
1625 					}
1626 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1627 					break;
1628 				case IPV6_2292RTHDR:
1629 					OPTSET2292(IN6P_RTHDR);
1630 					break;
1631 				}
1632 				break;
1633 			case IPV6_PKTINFO:
1634 			case IPV6_HOPOPTS:
1635 			case IPV6_RTHDR:
1636 			case IPV6_DSTOPTS:
1637 			case IPV6_RTHDRDSTOPTS:
1638 			case IPV6_NEXTHOP:
1639 			{
1640 				/* new advanced API (RFC3542) */
1641 				u_char *optbuf;
1642 				u_char optbuf_storage[MCLBYTES];
1643 				int optlen;
1644 				struct ip6_pktopts **optp;
1645 
1646 				/* cannot mix with RFC2292 */
1647 				if (OPTBIT(IN6P_RFC2292)) {
1648 					error = EINVAL;
1649 					break;
1650 				}
1651 
1652 				/*
1653 				 * We only ensure valsize is not too large
1654 				 * here.  Further validation will be done
1655 				 * later.
1656 				 */
1657 				error = sooptcopyin(sopt, optbuf_storage,
1658 				    sizeof(optbuf_storage), 0);
1659 				if (error)
1660 					break;
1661 				optlen = sopt->sopt_valsize;
1662 				optbuf = optbuf_storage;
1663 				optp = &in6p->in6p_outputopts;
1664 				error = ip6_pcbopt(optname, optbuf, optlen,
1665 				    optp, (td != NULL) ? td->td_ucred : NULL,
1666 				    uproto);
1667 				break;
1668 			}
1669 #undef OPTSET
1670 
1671 			case IPV6_MULTICAST_IF:
1672 			case IPV6_MULTICAST_HOPS:
1673 			case IPV6_MULTICAST_LOOP:
1674 			case IPV6_JOIN_GROUP:
1675 			case IPV6_LEAVE_GROUP:
1676 			case IPV6_MSFILTER:
1677 			case MCAST_BLOCK_SOURCE:
1678 			case MCAST_UNBLOCK_SOURCE:
1679 			case MCAST_JOIN_GROUP:
1680 			case MCAST_LEAVE_GROUP:
1681 			case MCAST_JOIN_SOURCE_GROUP:
1682 			case MCAST_LEAVE_SOURCE_GROUP:
1683 				error = ip6_setmoptions(in6p, sopt);
1684 				break;
1685 
1686 			case IPV6_PORTRANGE:
1687 				error = sooptcopyin(sopt, &optval,
1688 				    sizeof optval, sizeof optval);
1689 				if (error)
1690 					break;
1691 
1692 				INP_WLOCK(in6p);
1693 				switch (optval) {
1694 				case IPV6_PORTRANGE_DEFAULT:
1695 					in6p->inp_flags &= ~(INP_LOWPORT);
1696 					in6p->inp_flags &= ~(INP_HIGHPORT);
1697 					break;
1698 
1699 				case IPV6_PORTRANGE_HIGH:
1700 					in6p->inp_flags &= ~(INP_LOWPORT);
1701 					in6p->inp_flags |= INP_HIGHPORT;
1702 					break;
1703 
1704 				case IPV6_PORTRANGE_LOW:
1705 					in6p->inp_flags &= ~(INP_HIGHPORT);
1706 					in6p->inp_flags |= INP_LOWPORT;
1707 					break;
1708 
1709 				default:
1710 					error = EINVAL;
1711 					break;
1712 				}
1713 				INP_WUNLOCK(in6p);
1714 				break;
1715 
1716 #ifdef IPSEC
1717 			case IPV6_IPSEC_POLICY:
1718 			{
1719 				caddr_t req;
1720 				struct mbuf *m;
1721 
1722 				if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1723 					break;
1724 				if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1725 					break;
1726 				req = mtod(m, caddr_t);
1727 				error = ipsec_set_policy(in6p, optname, req,
1728 				    m->m_len, (sopt->sopt_td != NULL) ?
1729 				    sopt->sopt_td->td_ucred : NULL);
1730 				m_freem(m);
1731 				break;
1732 			}
1733 #endif /* IPSEC */
1734 
1735 			default:
1736 				error = ENOPROTOOPT;
1737 				break;
1738 			}
1739 			break;
1740 
1741 		case SOPT_GET:
1742 			switch (optname) {
1743 
1744 			case IPV6_2292PKTOPTIONS:
1745 #ifdef IPV6_PKTOPTIONS
1746 			case IPV6_PKTOPTIONS:
1747 #endif
1748 				/*
1749 				 * RFC3542 (effectively) deprecated the
1750 				 * semantics of the 2292-style pktoptions.
1751 				 * Since it was not reliable in nature (i.e.,
1752 				 * applications had to expect the lack of some
1753 				 * information after all), it would make sense
1754 				 * to simplify this part by always returning
1755 				 * empty data.
1756 				 */
1757 				sopt->sopt_valsize = 0;
1758 				break;
1759 
1760 			case IPV6_RECVHOPOPTS:
1761 			case IPV6_RECVDSTOPTS:
1762 			case IPV6_RECVRTHDRDSTOPTS:
1763 			case IPV6_UNICAST_HOPS:
1764 			case IPV6_RECVPKTINFO:
1765 			case IPV6_RECVHOPLIMIT:
1766 			case IPV6_RECVRTHDR:
1767 			case IPV6_RECVPATHMTU:
1768 
1769 			case IPV6_FAITH:
1770 			case IPV6_V6ONLY:
1771 			case IPV6_PORTRANGE:
1772 			case IPV6_RECVTCLASS:
1773 			case IPV6_AUTOFLOWLABEL:
1774 			case IPV6_BINDANY:
1775 				switch (optname) {
1776 
1777 				case IPV6_RECVHOPOPTS:
1778 					optval = OPTBIT(IN6P_HOPOPTS);
1779 					break;
1780 
1781 				case IPV6_RECVDSTOPTS:
1782 					optval = OPTBIT(IN6P_DSTOPTS);
1783 					break;
1784 
1785 				case IPV6_RECVRTHDRDSTOPTS:
1786 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1787 					break;
1788 
1789 				case IPV6_UNICAST_HOPS:
1790 					optval = in6p->in6p_hops;
1791 					break;
1792 
1793 				case IPV6_RECVPKTINFO:
1794 					optval = OPTBIT(IN6P_PKTINFO);
1795 					break;
1796 
1797 				case IPV6_RECVHOPLIMIT:
1798 					optval = OPTBIT(IN6P_HOPLIMIT);
1799 					break;
1800 
1801 				case IPV6_RECVRTHDR:
1802 					optval = OPTBIT(IN6P_RTHDR);
1803 					break;
1804 
1805 				case IPV6_RECVPATHMTU:
1806 					optval = OPTBIT(IN6P_MTU);
1807 					break;
1808 
1809 				case IPV6_FAITH:
1810 					optval = OPTBIT(INP_FAITH);
1811 					break;
1812 
1813 				case IPV6_V6ONLY:
1814 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
1815 					break;
1816 
1817 				case IPV6_PORTRANGE:
1818 				    {
1819 					int flags;
1820 					flags = in6p->inp_flags;
1821 					if (flags & INP_HIGHPORT)
1822 						optval = IPV6_PORTRANGE_HIGH;
1823 					else if (flags & INP_LOWPORT)
1824 						optval = IPV6_PORTRANGE_LOW;
1825 					else
1826 						optval = 0;
1827 					break;
1828 				    }
1829 				case IPV6_RECVTCLASS:
1830 					optval = OPTBIT(IN6P_TCLASS);
1831 					break;
1832 
1833 				case IPV6_AUTOFLOWLABEL:
1834 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1835 					break;
1836 
1837 				case IPV6_BINDANY:
1838 					optval = OPTBIT(INP_BINDANY);
1839 					break;
1840 				}
1841 				if (error)
1842 					break;
1843 				error = sooptcopyout(sopt, &optval,
1844 					sizeof optval);
1845 				break;
1846 
1847 			case IPV6_PATHMTU:
1848 			{
1849 				u_long pmtu = 0;
1850 				struct ip6_mtuinfo mtuinfo;
1851 				struct route_in6 sro;
1852 
1853 				bzero(&sro, sizeof(sro));
1854 
1855 				if (!(so->so_state & SS_ISCONNECTED))
1856 					return (ENOTCONN);
1857 				/*
1858 				 * XXX: we dot not consider the case of source
1859 				 * routing, or optional information to specify
1860 				 * the outgoing interface.
1861 				 */
1862 				error = ip6_getpmtu(&sro, NULL, NULL,
1863 				    &in6p->in6p_faddr, &pmtu, NULL,
1864 				    so->so_fibnum);
1865 				if (sro.ro_rt)
1866 					RTFREE(sro.ro_rt);
1867 				if (error)
1868 					break;
1869 				if (pmtu > IPV6_MAXPACKET)
1870 					pmtu = IPV6_MAXPACKET;
1871 
1872 				bzero(&mtuinfo, sizeof(mtuinfo));
1873 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
1874 				optdata = (void *)&mtuinfo;
1875 				optdatalen = sizeof(mtuinfo);
1876 				error = sooptcopyout(sopt, optdata,
1877 				    optdatalen);
1878 				break;
1879 			}
1880 
1881 			case IPV6_2292PKTINFO:
1882 			case IPV6_2292HOPLIMIT:
1883 			case IPV6_2292HOPOPTS:
1884 			case IPV6_2292RTHDR:
1885 			case IPV6_2292DSTOPTS:
1886 				switch (optname) {
1887 				case IPV6_2292PKTINFO:
1888 					optval = OPTBIT(IN6P_PKTINFO);
1889 					break;
1890 				case IPV6_2292HOPLIMIT:
1891 					optval = OPTBIT(IN6P_HOPLIMIT);
1892 					break;
1893 				case IPV6_2292HOPOPTS:
1894 					optval = OPTBIT(IN6P_HOPOPTS);
1895 					break;
1896 				case IPV6_2292RTHDR:
1897 					optval = OPTBIT(IN6P_RTHDR);
1898 					break;
1899 				case IPV6_2292DSTOPTS:
1900 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
1901 					break;
1902 				}
1903 				error = sooptcopyout(sopt, &optval,
1904 				    sizeof optval);
1905 				break;
1906 			case IPV6_PKTINFO:
1907 			case IPV6_HOPOPTS:
1908 			case IPV6_RTHDR:
1909 			case IPV6_DSTOPTS:
1910 			case IPV6_RTHDRDSTOPTS:
1911 			case IPV6_NEXTHOP:
1912 			case IPV6_TCLASS:
1913 			case IPV6_DONTFRAG:
1914 			case IPV6_USE_MIN_MTU:
1915 			case IPV6_PREFER_TEMPADDR:
1916 				error = ip6_getpcbopt(in6p->in6p_outputopts,
1917 				    optname, sopt);
1918 				break;
1919 
1920 			case IPV6_MULTICAST_IF:
1921 			case IPV6_MULTICAST_HOPS:
1922 			case IPV6_MULTICAST_LOOP:
1923 			case IPV6_MSFILTER:
1924 				error = ip6_getmoptions(in6p, sopt);
1925 				break;
1926 
1927 #ifdef IPSEC
1928 			case IPV6_IPSEC_POLICY:
1929 			  {
1930 				caddr_t req = NULL;
1931 				size_t len = 0;
1932 				struct mbuf *m = NULL;
1933 				struct mbuf **mp = &m;
1934 				size_t ovalsize = sopt->sopt_valsize;
1935 				caddr_t oval = (caddr_t)sopt->sopt_val;
1936 
1937 				error = soopt_getm(sopt, &m); /* XXX */
1938 				if (error != 0)
1939 					break;
1940 				error = soopt_mcopyin(sopt, m); /* XXX */
1941 				if (error != 0)
1942 					break;
1943 				sopt->sopt_valsize = ovalsize;
1944 				sopt->sopt_val = oval;
1945 				if (m) {
1946 					req = mtod(m, caddr_t);
1947 					len = m->m_len;
1948 				}
1949 				error = ipsec_get_policy(in6p, req, len, mp);
1950 				if (error == 0)
1951 					error = soopt_mcopyout(sopt, m); /* XXX */
1952 				if (error == 0 && m)
1953 					m_freem(m);
1954 				break;
1955 			  }
1956 #endif /* IPSEC */
1957 
1958 			default:
1959 				error = ENOPROTOOPT;
1960 				break;
1961 			}
1962 			break;
1963 		}
1964 	}
1965 	return (error);
1966 }
1967 
1968 int
1969 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
1970 {
1971 	int error = 0, optval, optlen;
1972 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
1973 	struct inpcb *in6p = sotoinpcb(so);
1974 	int level, op, optname;
1975 
1976 	level = sopt->sopt_level;
1977 	op = sopt->sopt_dir;
1978 	optname = sopt->sopt_name;
1979 	optlen = sopt->sopt_valsize;
1980 
1981 	if (level != IPPROTO_IPV6) {
1982 		return (EINVAL);
1983 	}
1984 
1985 	switch (optname) {
1986 	case IPV6_CHECKSUM:
1987 		/*
1988 		 * For ICMPv6 sockets, no modification allowed for checksum
1989 		 * offset, permit "no change" values to help existing apps.
1990 		 *
1991 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
1992 		 * for an ICMPv6 socket will fail."
1993 		 * The current behavior does not meet RFC3542.
1994 		 */
1995 		switch (op) {
1996 		case SOPT_SET:
1997 			if (optlen != sizeof(int)) {
1998 				error = EINVAL;
1999 				break;
2000 			}
2001 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2002 					    sizeof(optval));
2003 			if (error)
2004 				break;
2005 			if ((optval % 2) != 0) {
2006 				/* the API assumes even offset values */
2007 				error = EINVAL;
2008 			} else if (so->so_proto->pr_protocol ==
2009 			    IPPROTO_ICMPV6) {
2010 				if (optval != icmp6off)
2011 					error = EINVAL;
2012 			} else
2013 				in6p->in6p_cksum = optval;
2014 			break;
2015 
2016 		case SOPT_GET:
2017 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2018 				optval = icmp6off;
2019 			else
2020 				optval = in6p->in6p_cksum;
2021 
2022 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2023 			break;
2024 
2025 		default:
2026 			error = EINVAL;
2027 			break;
2028 		}
2029 		break;
2030 
2031 	default:
2032 		error = ENOPROTOOPT;
2033 		break;
2034 	}
2035 
2036 	return (error);
2037 }
2038 
2039 /*
2040  * Set up IP6 options in pcb for insertion in output packets or
2041  * specifying behavior of outgoing packets.
2042  */
2043 static int
2044 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2045     struct socket *so, struct sockopt *sopt)
2046 {
2047 	struct ip6_pktopts *opt = *pktopt;
2048 	int error = 0;
2049 	struct thread *td = sopt->sopt_td;
2050 
2051 	/* turn off any old options. */
2052 	if (opt) {
2053 #ifdef DIAGNOSTIC
2054 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2055 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2056 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2057 			printf("ip6_pcbopts: all specified options are cleared.\n");
2058 #endif
2059 		ip6_clearpktopts(opt, -1);
2060 	} else
2061 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2062 	*pktopt = NULL;
2063 
2064 	if (!m || m->m_len == 0) {
2065 		/*
2066 		 * Only turning off any previous options, regardless of
2067 		 * whether the opt is just created or given.
2068 		 */
2069 		free(opt, M_IP6OPT);
2070 		return (0);
2071 	}
2072 
2073 	/*  set options specified by user. */
2074 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2075 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2076 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2077 		free(opt, M_IP6OPT);
2078 		return (error);
2079 	}
2080 	*pktopt = opt;
2081 	return (0);
2082 }
2083 
2084 /*
2085  * initialize ip6_pktopts.  beware that there are non-zero default values in
2086  * the struct.
2087  */
2088 void
2089 ip6_initpktopts(struct ip6_pktopts *opt)
2090 {
2091 
2092 	bzero(opt, sizeof(*opt));
2093 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2094 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2095 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2096 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2097 }
2098 
2099 static int
2100 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2101     struct ucred *cred, int uproto)
2102 {
2103 	struct ip6_pktopts *opt;
2104 
2105 	if (*pktopt == NULL) {
2106 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2107 		    M_WAITOK);
2108 		ip6_initpktopts(*pktopt);
2109 	}
2110 	opt = *pktopt;
2111 
2112 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2113 }
2114 
2115 static int
2116 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2117 {
2118 	void *optdata = NULL;
2119 	int optdatalen = 0;
2120 	struct ip6_ext *ip6e;
2121 	int error = 0;
2122 	struct in6_pktinfo null_pktinfo;
2123 	int deftclass = 0, on;
2124 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2125 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2126 
2127 	switch (optname) {
2128 	case IPV6_PKTINFO:
2129 		if (pktopt && pktopt->ip6po_pktinfo)
2130 			optdata = (void *)pktopt->ip6po_pktinfo;
2131 		else {
2132 			/* XXX: we don't have to do this every time... */
2133 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2134 			optdata = (void *)&null_pktinfo;
2135 		}
2136 		optdatalen = sizeof(struct in6_pktinfo);
2137 		break;
2138 	case IPV6_TCLASS:
2139 		if (pktopt && pktopt->ip6po_tclass >= 0)
2140 			optdata = (void *)&pktopt->ip6po_tclass;
2141 		else
2142 			optdata = (void *)&deftclass;
2143 		optdatalen = sizeof(int);
2144 		break;
2145 	case IPV6_HOPOPTS:
2146 		if (pktopt && pktopt->ip6po_hbh) {
2147 			optdata = (void *)pktopt->ip6po_hbh;
2148 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2149 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2150 		}
2151 		break;
2152 	case IPV6_RTHDR:
2153 		if (pktopt && pktopt->ip6po_rthdr) {
2154 			optdata = (void *)pktopt->ip6po_rthdr;
2155 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2156 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2157 		}
2158 		break;
2159 	case IPV6_RTHDRDSTOPTS:
2160 		if (pktopt && pktopt->ip6po_dest1) {
2161 			optdata = (void *)pktopt->ip6po_dest1;
2162 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2163 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2164 		}
2165 		break;
2166 	case IPV6_DSTOPTS:
2167 		if (pktopt && pktopt->ip6po_dest2) {
2168 			optdata = (void *)pktopt->ip6po_dest2;
2169 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2170 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2171 		}
2172 		break;
2173 	case IPV6_NEXTHOP:
2174 		if (pktopt && pktopt->ip6po_nexthop) {
2175 			optdata = (void *)pktopt->ip6po_nexthop;
2176 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2177 		}
2178 		break;
2179 	case IPV6_USE_MIN_MTU:
2180 		if (pktopt)
2181 			optdata = (void *)&pktopt->ip6po_minmtu;
2182 		else
2183 			optdata = (void *)&defminmtu;
2184 		optdatalen = sizeof(int);
2185 		break;
2186 	case IPV6_DONTFRAG:
2187 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2188 			on = 1;
2189 		else
2190 			on = 0;
2191 		optdata = (void *)&on;
2192 		optdatalen = sizeof(on);
2193 		break;
2194 	case IPV6_PREFER_TEMPADDR:
2195 		if (pktopt)
2196 			optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2197 		else
2198 			optdata = (void *)&defpreftemp;
2199 		optdatalen = sizeof(int);
2200 		break;
2201 	default:		/* should not happen */
2202 #ifdef DIAGNOSTIC
2203 		panic("ip6_getpcbopt: unexpected option\n");
2204 #endif
2205 		return (ENOPROTOOPT);
2206 	}
2207 
2208 	error = sooptcopyout(sopt, optdata, optdatalen);
2209 
2210 	return (error);
2211 }
2212 
2213 void
2214 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2215 {
2216 	if (pktopt == NULL)
2217 		return;
2218 
2219 	if (optname == -1 || optname == IPV6_PKTINFO) {
2220 		if (pktopt->ip6po_pktinfo)
2221 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2222 		pktopt->ip6po_pktinfo = NULL;
2223 	}
2224 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2225 		pktopt->ip6po_hlim = -1;
2226 	if (optname == -1 || optname == IPV6_TCLASS)
2227 		pktopt->ip6po_tclass = -1;
2228 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2229 		if (pktopt->ip6po_nextroute.ro_rt) {
2230 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2231 			pktopt->ip6po_nextroute.ro_rt = NULL;
2232 		}
2233 		if (pktopt->ip6po_nexthop)
2234 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2235 		pktopt->ip6po_nexthop = NULL;
2236 	}
2237 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2238 		if (pktopt->ip6po_hbh)
2239 			free(pktopt->ip6po_hbh, M_IP6OPT);
2240 		pktopt->ip6po_hbh = NULL;
2241 	}
2242 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2243 		if (pktopt->ip6po_dest1)
2244 			free(pktopt->ip6po_dest1, M_IP6OPT);
2245 		pktopt->ip6po_dest1 = NULL;
2246 	}
2247 	if (optname == -1 || optname == IPV6_RTHDR) {
2248 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2249 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2250 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2251 		if (pktopt->ip6po_route.ro_rt) {
2252 			RTFREE(pktopt->ip6po_route.ro_rt);
2253 			pktopt->ip6po_route.ro_rt = NULL;
2254 		}
2255 	}
2256 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2257 		if (pktopt->ip6po_dest2)
2258 			free(pktopt->ip6po_dest2, M_IP6OPT);
2259 		pktopt->ip6po_dest2 = NULL;
2260 	}
2261 }
2262 
2263 #define PKTOPT_EXTHDRCPY(type) \
2264 do {\
2265 	if (src->type) {\
2266 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2267 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2268 		if (dst->type == NULL && canwait == M_NOWAIT)\
2269 			goto bad;\
2270 		bcopy(src->type, dst->type, hlen);\
2271 	}\
2272 } while (/*CONSTCOND*/ 0)
2273 
2274 static int
2275 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2276 {
2277 	if (dst == NULL || src == NULL)  {
2278 		printf("ip6_clearpktopts: invalid argument\n");
2279 		return (EINVAL);
2280 	}
2281 
2282 	dst->ip6po_hlim = src->ip6po_hlim;
2283 	dst->ip6po_tclass = src->ip6po_tclass;
2284 	dst->ip6po_flags = src->ip6po_flags;
2285 	dst->ip6po_minmtu = src->ip6po_minmtu;
2286 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2287 	if (src->ip6po_pktinfo) {
2288 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2289 		    M_IP6OPT, canwait);
2290 		if (dst->ip6po_pktinfo == NULL)
2291 			goto bad;
2292 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2293 	}
2294 	if (src->ip6po_nexthop) {
2295 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2296 		    M_IP6OPT, canwait);
2297 		if (dst->ip6po_nexthop == NULL)
2298 			goto bad;
2299 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2300 		    src->ip6po_nexthop->sa_len);
2301 	}
2302 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2303 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2304 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2305 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2306 	return (0);
2307 
2308   bad:
2309 	ip6_clearpktopts(dst, -1);
2310 	return (ENOBUFS);
2311 }
2312 #undef PKTOPT_EXTHDRCPY
2313 
2314 struct ip6_pktopts *
2315 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2316 {
2317 	int error;
2318 	struct ip6_pktopts *dst;
2319 
2320 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2321 	if (dst == NULL)
2322 		return (NULL);
2323 	ip6_initpktopts(dst);
2324 
2325 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2326 		free(dst, M_IP6OPT);
2327 		return (NULL);
2328 	}
2329 
2330 	return (dst);
2331 }
2332 
2333 void
2334 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2335 {
2336 	if (pktopt == NULL)
2337 		return;
2338 
2339 	ip6_clearpktopts(pktopt, -1);
2340 
2341 	free(pktopt, M_IP6OPT);
2342 }
2343 
2344 /*
2345  * Set IPv6 outgoing packet options based on advanced API.
2346  */
2347 int
2348 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2349     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2350 {
2351 	struct cmsghdr *cm = 0;
2352 
2353 	if (control == NULL || opt == NULL)
2354 		return (EINVAL);
2355 
2356 	ip6_initpktopts(opt);
2357 	if (stickyopt) {
2358 		int error;
2359 
2360 		/*
2361 		 * If stickyopt is provided, make a local copy of the options
2362 		 * for this particular packet, then override them by ancillary
2363 		 * objects.
2364 		 * XXX: copypktopts() does not copy the cached route to a next
2365 		 * hop (if any).  This is not very good in terms of efficiency,
2366 		 * but we can allow this since this option should be rarely
2367 		 * used.
2368 		 */
2369 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2370 			return (error);
2371 	}
2372 
2373 	/*
2374 	 * XXX: Currently, we assume all the optional information is stored
2375 	 * in a single mbuf.
2376 	 */
2377 	if (control->m_next)
2378 		return (EINVAL);
2379 
2380 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2381 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2382 		int error;
2383 
2384 		if (control->m_len < CMSG_LEN(0))
2385 			return (EINVAL);
2386 
2387 		cm = mtod(control, struct cmsghdr *);
2388 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2389 			return (EINVAL);
2390 		if (cm->cmsg_level != IPPROTO_IPV6)
2391 			continue;
2392 
2393 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2394 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2395 		if (error)
2396 			return (error);
2397 	}
2398 
2399 	return (0);
2400 }
2401 
2402 /*
2403  * Set a particular packet option, as a sticky option or an ancillary data
2404  * item.  "len" can be 0 only when it's a sticky option.
2405  * We have 4 cases of combination of "sticky" and "cmsg":
2406  * "sticky=0, cmsg=0": impossible
2407  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2408  * "sticky=1, cmsg=0": RFC3542 socket option
2409  * "sticky=1, cmsg=1": RFC2292 socket option
2410  */
2411 static int
2412 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2413     struct ucred *cred, int sticky, int cmsg, int uproto)
2414 {
2415 	int minmtupolicy, preftemp;
2416 	int error;
2417 
2418 	if (!sticky && !cmsg) {
2419 #ifdef DIAGNOSTIC
2420 		printf("ip6_setpktopt: impossible case\n");
2421 #endif
2422 		return (EINVAL);
2423 	}
2424 
2425 	/*
2426 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2427 	 * not be specified in the context of RFC3542.  Conversely,
2428 	 * RFC3542 types should not be specified in the context of RFC2292.
2429 	 */
2430 	if (!cmsg) {
2431 		switch (optname) {
2432 		case IPV6_2292PKTINFO:
2433 		case IPV6_2292HOPLIMIT:
2434 		case IPV6_2292NEXTHOP:
2435 		case IPV6_2292HOPOPTS:
2436 		case IPV6_2292DSTOPTS:
2437 		case IPV6_2292RTHDR:
2438 		case IPV6_2292PKTOPTIONS:
2439 			return (ENOPROTOOPT);
2440 		}
2441 	}
2442 	if (sticky && cmsg) {
2443 		switch (optname) {
2444 		case IPV6_PKTINFO:
2445 		case IPV6_HOPLIMIT:
2446 		case IPV6_NEXTHOP:
2447 		case IPV6_HOPOPTS:
2448 		case IPV6_DSTOPTS:
2449 		case IPV6_RTHDRDSTOPTS:
2450 		case IPV6_RTHDR:
2451 		case IPV6_USE_MIN_MTU:
2452 		case IPV6_DONTFRAG:
2453 		case IPV6_TCLASS:
2454 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2455 			return (ENOPROTOOPT);
2456 		}
2457 	}
2458 
2459 	switch (optname) {
2460 	case IPV6_2292PKTINFO:
2461 	case IPV6_PKTINFO:
2462 	{
2463 		struct ifnet *ifp = NULL;
2464 		struct in6_pktinfo *pktinfo;
2465 
2466 		if (len != sizeof(struct in6_pktinfo))
2467 			return (EINVAL);
2468 
2469 		pktinfo = (struct in6_pktinfo *)buf;
2470 
2471 		/*
2472 		 * An application can clear any sticky IPV6_PKTINFO option by
2473 		 * doing a "regular" setsockopt with ipi6_addr being
2474 		 * in6addr_any and ipi6_ifindex being zero.
2475 		 * [RFC 3542, Section 6]
2476 		 */
2477 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2478 		    pktinfo->ipi6_ifindex == 0 &&
2479 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2480 			ip6_clearpktopts(opt, optname);
2481 			break;
2482 		}
2483 
2484 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2485 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2486 			return (EINVAL);
2487 		}
2488 
2489 		/* validate the interface index if specified. */
2490 		if (pktinfo->ipi6_ifindex > V_if_index)
2491 			 return (ENXIO);
2492 		if (pktinfo->ipi6_ifindex) {
2493 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2494 			if (ifp == NULL)
2495 				return (ENXIO);
2496 		}
2497 
2498 		/*
2499 		 * We store the address anyway, and let in6_selectsrc()
2500 		 * validate the specified address.  This is because ipi6_addr
2501 		 * may not have enough information about its scope zone, and
2502 		 * we may need additional information (such as outgoing
2503 		 * interface or the scope zone of a destination address) to
2504 		 * disambiguate the scope.
2505 		 * XXX: the delay of the validation may confuse the
2506 		 * application when it is used as a sticky option.
2507 		 */
2508 		if (opt->ip6po_pktinfo == NULL) {
2509 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2510 			    M_IP6OPT, M_NOWAIT);
2511 			if (opt->ip6po_pktinfo == NULL)
2512 				return (ENOBUFS);
2513 		}
2514 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2515 		break;
2516 	}
2517 
2518 	case IPV6_2292HOPLIMIT:
2519 	case IPV6_HOPLIMIT:
2520 	{
2521 		int *hlimp;
2522 
2523 		/*
2524 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2525 		 * to simplify the ordering among hoplimit options.
2526 		 */
2527 		if (optname == IPV6_HOPLIMIT && sticky)
2528 			return (ENOPROTOOPT);
2529 
2530 		if (len != sizeof(int))
2531 			return (EINVAL);
2532 		hlimp = (int *)buf;
2533 		if (*hlimp < -1 || *hlimp > 255)
2534 			return (EINVAL);
2535 
2536 		opt->ip6po_hlim = *hlimp;
2537 		break;
2538 	}
2539 
2540 	case IPV6_TCLASS:
2541 	{
2542 		int tclass;
2543 
2544 		if (len != sizeof(int))
2545 			return (EINVAL);
2546 		tclass = *(int *)buf;
2547 		if (tclass < -1 || tclass > 255)
2548 			return (EINVAL);
2549 
2550 		opt->ip6po_tclass = tclass;
2551 		break;
2552 	}
2553 
2554 	case IPV6_2292NEXTHOP:
2555 	case IPV6_NEXTHOP:
2556 		if (cred != NULL) {
2557 			error = priv_check_cred(cred,
2558 			    PRIV_NETINET_SETHDROPTS, 0);
2559 			if (error)
2560 				return (error);
2561 		}
2562 
2563 		if (len == 0) {	/* just remove the option */
2564 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2565 			break;
2566 		}
2567 
2568 		/* check if cmsg_len is large enough for sa_len */
2569 		if (len < sizeof(struct sockaddr) || len < *buf)
2570 			return (EINVAL);
2571 
2572 		switch (((struct sockaddr *)buf)->sa_family) {
2573 		case AF_INET6:
2574 		{
2575 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2576 			int error;
2577 
2578 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2579 				return (EINVAL);
2580 
2581 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2582 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2583 				return (EINVAL);
2584 			}
2585 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2586 			    != 0) {
2587 				return (error);
2588 			}
2589 			break;
2590 		}
2591 		case AF_LINK:	/* should eventually be supported */
2592 		default:
2593 			return (EAFNOSUPPORT);
2594 		}
2595 
2596 		/* turn off the previous option, then set the new option. */
2597 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2598 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2599 		if (opt->ip6po_nexthop == NULL)
2600 			return (ENOBUFS);
2601 		bcopy(buf, opt->ip6po_nexthop, *buf);
2602 		break;
2603 
2604 	case IPV6_2292HOPOPTS:
2605 	case IPV6_HOPOPTS:
2606 	{
2607 		struct ip6_hbh *hbh;
2608 		int hbhlen;
2609 
2610 		/*
2611 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2612 		 * options, since per-option restriction has too much
2613 		 * overhead.
2614 		 */
2615 		if (cred != NULL) {
2616 			error = priv_check_cred(cred,
2617 			    PRIV_NETINET_SETHDROPTS, 0);
2618 			if (error)
2619 				return (error);
2620 		}
2621 
2622 		if (len == 0) {
2623 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2624 			break;	/* just remove the option */
2625 		}
2626 
2627 		/* message length validation */
2628 		if (len < sizeof(struct ip6_hbh))
2629 			return (EINVAL);
2630 		hbh = (struct ip6_hbh *)buf;
2631 		hbhlen = (hbh->ip6h_len + 1) << 3;
2632 		if (len != hbhlen)
2633 			return (EINVAL);
2634 
2635 		/* turn off the previous option, then set the new option. */
2636 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2637 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2638 		if (opt->ip6po_hbh == NULL)
2639 			return (ENOBUFS);
2640 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
2641 
2642 		break;
2643 	}
2644 
2645 	case IPV6_2292DSTOPTS:
2646 	case IPV6_DSTOPTS:
2647 	case IPV6_RTHDRDSTOPTS:
2648 	{
2649 		struct ip6_dest *dest, **newdest = NULL;
2650 		int destlen;
2651 
2652 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2653 			error = priv_check_cred(cred,
2654 			    PRIV_NETINET_SETHDROPTS, 0);
2655 			if (error)
2656 				return (error);
2657 		}
2658 
2659 		if (len == 0) {
2660 			ip6_clearpktopts(opt, optname);
2661 			break;	/* just remove the option */
2662 		}
2663 
2664 		/* message length validation */
2665 		if (len < sizeof(struct ip6_dest))
2666 			return (EINVAL);
2667 		dest = (struct ip6_dest *)buf;
2668 		destlen = (dest->ip6d_len + 1) << 3;
2669 		if (len != destlen)
2670 			return (EINVAL);
2671 
2672 		/*
2673 		 * Determine the position that the destination options header
2674 		 * should be inserted; before or after the routing header.
2675 		 */
2676 		switch (optname) {
2677 		case IPV6_2292DSTOPTS:
2678 			/*
2679 			 * The old advacned API is ambiguous on this point.
2680 			 * Our approach is to determine the position based
2681 			 * according to the existence of a routing header.
2682 			 * Note, however, that this depends on the order of the
2683 			 * extension headers in the ancillary data; the 1st
2684 			 * part of the destination options header must appear
2685 			 * before the routing header in the ancillary data,
2686 			 * too.
2687 			 * RFC3542 solved the ambiguity by introducing
2688 			 * separate ancillary data or option types.
2689 			 */
2690 			if (opt->ip6po_rthdr == NULL)
2691 				newdest = &opt->ip6po_dest1;
2692 			else
2693 				newdest = &opt->ip6po_dest2;
2694 			break;
2695 		case IPV6_RTHDRDSTOPTS:
2696 			newdest = &opt->ip6po_dest1;
2697 			break;
2698 		case IPV6_DSTOPTS:
2699 			newdest = &opt->ip6po_dest2;
2700 			break;
2701 		}
2702 
2703 		/* turn off the previous option, then set the new option. */
2704 		ip6_clearpktopts(opt, optname);
2705 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2706 		if (*newdest == NULL)
2707 			return (ENOBUFS);
2708 		bcopy(dest, *newdest, destlen);
2709 
2710 		break;
2711 	}
2712 
2713 	case IPV6_2292RTHDR:
2714 	case IPV6_RTHDR:
2715 	{
2716 		struct ip6_rthdr *rth;
2717 		int rthlen;
2718 
2719 		if (len == 0) {
2720 			ip6_clearpktopts(opt, IPV6_RTHDR);
2721 			break;	/* just remove the option */
2722 		}
2723 
2724 		/* message length validation */
2725 		if (len < sizeof(struct ip6_rthdr))
2726 			return (EINVAL);
2727 		rth = (struct ip6_rthdr *)buf;
2728 		rthlen = (rth->ip6r_len + 1) << 3;
2729 		if (len != rthlen)
2730 			return (EINVAL);
2731 
2732 		switch (rth->ip6r_type) {
2733 		case IPV6_RTHDR_TYPE_0:
2734 			if (rth->ip6r_len == 0)	/* must contain one addr */
2735 				return (EINVAL);
2736 			if (rth->ip6r_len % 2) /* length must be even */
2737 				return (EINVAL);
2738 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2739 				return (EINVAL);
2740 			break;
2741 		default:
2742 			return (EINVAL);	/* not supported */
2743 		}
2744 
2745 		/* turn off the previous option */
2746 		ip6_clearpktopts(opt, IPV6_RTHDR);
2747 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2748 		if (opt->ip6po_rthdr == NULL)
2749 			return (ENOBUFS);
2750 		bcopy(rth, opt->ip6po_rthdr, rthlen);
2751 
2752 		break;
2753 	}
2754 
2755 	case IPV6_USE_MIN_MTU:
2756 		if (len != sizeof(int))
2757 			return (EINVAL);
2758 		minmtupolicy = *(int *)buf;
2759 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2760 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
2761 		    minmtupolicy != IP6PO_MINMTU_ALL) {
2762 			return (EINVAL);
2763 		}
2764 		opt->ip6po_minmtu = minmtupolicy;
2765 		break;
2766 
2767 	case IPV6_DONTFRAG:
2768 		if (len != sizeof(int))
2769 			return (EINVAL);
2770 
2771 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2772 			/*
2773 			 * we ignore this option for TCP sockets.
2774 			 * (RFC3542 leaves this case unspecified.)
2775 			 */
2776 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2777 		} else
2778 			opt->ip6po_flags |= IP6PO_DONTFRAG;
2779 		break;
2780 
2781 	case IPV6_PREFER_TEMPADDR:
2782 		if (len != sizeof(int))
2783 			return (EINVAL);
2784 		preftemp = *(int *)buf;
2785 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2786 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2787 		    preftemp != IP6PO_TEMPADDR_PREFER) {
2788 			return (EINVAL);
2789 		}
2790 		opt->ip6po_prefer_tempaddr = preftemp;
2791 		break;
2792 
2793 	default:
2794 		return (ENOPROTOOPT);
2795 	} /* end of switch */
2796 
2797 	return (0);
2798 }
2799 
2800 /*
2801  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2802  * packet to the input queue of a specified interface.  Note that this
2803  * calls the output routine of the loopback "driver", but with an interface
2804  * pointer that might NOT be &loif -- easier than replicating that code here.
2805  */
2806 void
2807 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2808 {
2809 	struct mbuf *copym;
2810 	struct ip6_hdr *ip6;
2811 
2812 	copym = m_copy(m, 0, M_COPYALL);
2813 	if (copym == NULL)
2814 		return;
2815 
2816 	/*
2817 	 * Make sure to deep-copy IPv6 header portion in case the data
2818 	 * is in an mbuf cluster, so that we can safely override the IPv6
2819 	 * header portion later.
2820 	 */
2821 	if ((copym->m_flags & M_EXT) != 0 ||
2822 	    copym->m_len < sizeof(struct ip6_hdr)) {
2823 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
2824 		if (copym == NULL)
2825 			return;
2826 	}
2827 
2828 #ifdef DIAGNOSTIC
2829 	if (copym->m_len < sizeof(*ip6)) {
2830 		m_freem(copym);
2831 		return;
2832 	}
2833 #endif
2834 
2835 	ip6 = mtod(copym, struct ip6_hdr *);
2836 	/*
2837 	 * clear embedded scope identifiers if necessary.
2838 	 * in6_clearscope will touch the addresses only when necessary.
2839 	 */
2840 	in6_clearscope(&ip6->ip6_src);
2841 	in6_clearscope(&ip6->ip6_dst);
2842 
2843 	(void)if_simloop(ifp, copym, dst->sin6_family, 0);
2844 }
2845 
2846 /*
2847  * Chop IPv6 header off from the payload.
2848  */
2849 static int
2850 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
2851 {
2852 	struct mbuf *mh;
2853 	struct ip6_hdr *ip6;
2854 
2855 	ip6 = mtod(m, struct ip6_hdr *);
2856 	if (m->m_len > sizeof(*ip6)) {
2857 		mh = m_gethdr(M_NOWAIT, MT_DATA);
2858 		if (mh == NULL) {
2859 			m_freem(m);
2860 			return ENOBUFS;
2861 		}
2862 		m_move_pkthdr(mh, m);
2863 		MH_ALIGN(mh, sizeof(*ip6));
2864 		m->m_len -= sizeof(*ip6);
2865 		m->m_data += sizeof(*ip6);
2866 		mh->m_next = m;
2867 		m = mh;
2868 		m->m_len = sizeof(*ip6);
2869 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2870 	}
2871 	exthdrs->ip6e_ip6 = m;
2872 	return 0;
2873 }
2874 
2875 /*
2876  * Compute IPv6 extension header length.
2877  */
2878 int
2879 ip6_optlen(struct inpcb *in6p)
2880 {
2881 	int len;
2882 
2883 	if (!in6p->in6p_outputopts)
2884 		return 0;
2885 
2886 	len = 0;
2887 #define elen(x) \
2888     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2889 
2890 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
2891 	if (in6p->in6p_outputopts->ip6po_rthdr)
2892 		/* dest1 is valid with rthdr only */
2893 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
2894 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
2895 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
2896 	return len;
2897 #undef elen
2898 }
2899