xref: /freebsd/sys/netinet6/ip6_output.c (revision c6ec7d31830ab1c80edae95ad5e4b9dba10c47ac)
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/netisr.h>
90 #include <net/route.h>
91 #include <net/pfil.h>
92 #include <net/vnet.h>
93 
94 #include <netinet/in.h>
95 #include <netinet/in_var.h>
96 #include <netinet/ip_var.h>
97 #include <netinet6/in6_var.h>
98 #include <netinet/ip6.h>
99 #include <netinet/icmp6.h>
100 #include <netinet6/ip6_var.h>
101 #include <netinet/in_pcb.h>
102 #include <netinet/tcp_var.h>
103 #include <netinet6/nd6.h>
104 
105 #ifdef IPSEC
106 #include <netipsec/ipsec.h>
107 #include <netipsec/ipsec6.h>
108 #include <netipsec/key.h>
109 #include <netinet6/ip6_ipsec.h>
110 #endif /* IPSEC */
111 #ifdef SCTP
112 #include <netinet/sctp.h>
113 #include <netinet/sctp_crc32.h>
114 #endif
115 
116 #include <netinet6/ip6protosw.h>
117 #include <netinet6/scope6_var.h>
118 
119 #ifdef FLOWTABLE
120 #include <net/flowtable.h>
121 #endif
122 
123 extern int in6_mcast_loop;
124 
125 struct ip6_exthdrs {
126 	struct mbuf *ip6e_ip6;
127 	struct mbuf *ip6e_hbh;
128 	struct mbuf *ip6e_dest1;
129 	struct mbuf *ip6e_rthdr;
130 	struct mbuf *ip6e_dest2;
131 };
132 
133 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
134 			   struct ucred *, int);
135 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
136 	struct socket *, struct sockopt *);
137 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
138 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
139 	struct ucred *, int, int, int);
140 
141 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
142 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
143 	struct ip6_frag **);
144 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
145 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
146 static int ip6_getpmtu(struct route_in6 *, struct route_in6 *,
147 	struct ifnet *, struct in6_addr *, u_long *, int *, u_int);
148 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
149 
150 
151 /*
152  * Make an extension header from option data.  hp is the source, and
153  * mp is the destination.
154  */
155 #define MAKE_EXTHDR(hp, mp)						\
156     do {								\
157 	if (hp) {							\
158 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
159 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
160 		    ((eh)->ip6e_len + 1) << 3);				\
161 		if (error)						\
162 			goto freehdrs;					\
163 	}								\
164     } while (/*CONSTCOND*/ 0)
165 
166 /*
167  * Form a chain of extension headers.
168  * m is the extension header mbuf
169  * mp is the previous mbuf in the chain
170  * p is the next header
171  * i is the type of option.
172  */
173 #define MAKE_CHAIN(m, mp, p, i)\
174     do {\
175 	if (m) {\
176 		if (!hdrsplit) \
177 			panic("assumption failed: hdr not split"); \
178 		*mtod((m), u_char *) = *(p);\
179 		*(p) = (i);\
180 		p = mtod((m), u_char *);\
181 		(m)->m_next = (mp)->m_next;\
182 		(mp)->m_next = (m);\
183 		(mp) = (m);\
184 	}\
185     } while (/*CONSTCOND*/ 0)
186 
187 static void
188 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
189 {
190 	u_short csum;
191 
192 	csum = in_cksum_skip(m, offset + plen, offset);
193 	if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
194 		csum = 0xffff;
195 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
196 
197 	if (offset + sizeof(u_short) > m->m_len) {
198 		printf("%s: delayed m_pullup, m->len: %d plen %u off %u "
199 		    "csum_flags=0x%04x\n", __func__, m->m_len, plen, offset,
200 		    m->m_pkthdr.csum_flags);
201 		/*
202 		 * XXX this should not happen, but if it does, the correct
203 		 * behavior may be to insert the checksum in the appropriate
204 		 * next mbuf in the chain.
205 		 */
206 		return;
207 	}
208 	*(u_short *)(m->m_data + offset) = csum;
209 }
210 
211 /*
212  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
213  * header (with pri, len, nxt, hlim, src, dst).
214  * This function may modify ver and hlim only.
215  * The mbuf chain containing the packet will be freed.
216  * The mbuf opt, if present, will not be freed.
217  * If route_in6 ro is present and has ro_rt initialized, route lookup would be
218  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
219  * then result of route lookup is stored in ro->ro_rt.
220  *
221  * type of "mtu": rt_rmx.rmx_mtu is u_long, ifnet.ifr_mtu is int, and
222  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
223  * which is rt_rmx.rmx_mtu.
224  *
225  * ifpp - XXX: just for statistics
226  */
227 int
228 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
229     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
230     struct ifnet **ifpp, struct inpcb *inp)
231 {
232 	struct ip6_hdr *ip6, *mhip6;
233 	struct ifnet *ifp, *origifp;
234 	struct mbuf *m = m0;
235 	struct mbuf *mprev = NULL;
236 	int hlen, tlen, len, off;
237 	struct route_in6 ip6route;
238 	struct rtentry *rt = NULL;
239 	struct sockaddr_in6 *dst, src_sa, dst_sa;
240 	struct in6_addr odst;
241 	int error = 0;
242 	struct in6_ifaddr *ia = NULL;
243 	u_long mtu;
244 	int alwaysfrag, dontfrag;
245 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
246 	struct ip6_exthdrs exthdrs;
247 	struct in6_addr finaldst, src0, dst0;
248 	u_int32_t zone;
249 	struct route_in6 *ro_pmtu = NULL;
250 	int hdrsplit = 0;
251 	int needipsec = 0;
252 	int sw_csum, tso;
253 #ifdef IPSEC
254 	struct ipsec_output_state state;
255 	struct ip6_rthdr *rh = NULL;
256 	int needipsectun = 0;
257 	int segleft_org = 0;
258 	struct secpolicy *sp = NULL;
259 #endif /* IPSEC */
260 	struct m_tag *fwd_tag = NULL;
261 
262 	ip6 = mtod(m, struct ip6_hdr *);
263 	if (ip6 == NULL) {
264 		printf ("ip6 is NULL");
265 		goto bad;
266 	}
267 
268 	if (inp != NULL)
269 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
270 
271 	finaldst = ip6->ip6_dst;
272 	bzero(&exthdrs, sizeof(exthdrs));
273 	if (opt) {
274 		/* Hop-by-Hop options header */
275 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
276 		/* Destination options header(1st part) */
277 		if (opt->ip6po_rthdr) {
278 			/*
279 			 * Destination options header(1st part)
280 			 * This only makes sense with a routing header.
281 			 * See Section 9.2 of RFC 3542.
282 			 * Disabling this part just for MIP6 convenience is
283 			 * a bad idea.  We need to think carefully about a
284 			 * way to make the advanced API coexist with MIP6
285 			 * options, which might automatically be inserted in
286 			 * the kernel.
287 			 */
288 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
289 		}
290 		/* Routing header */
291 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
292 		/* Destination options header(2nd part) */
293 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
294 	}
295 
296 #ifdef IPSEC
297 	/*
298 	 * IPSec checking which handles several cases.
299 	 * FAST IPSEC: We re-injected the packet.
300 	 */
301 	switch(ip6_ipsec_output(&m, inp, &flags, &error, &ifp, &sp))
302 	{
303 	case 1:                 /* Bad packet */
304 		goto freehdrs;
305 	case -1:                /* Do IPSec */
306 		needipsec = 1;
307 		/*
308 		 * Do delayed checksums now, as we may send before returning.
309 		 */
310 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
311 			plen = m->m_pkthdr.len - sizeof(*ip6);
312 			in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
313 			m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
314 		}
315 #ifdef SCTP
316 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
317 			sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
318 			m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
319 		}
320 #endif
321 	case 0:                 /* No IPSec */
322 	default:
323 		break;
324 	}
325 #endif /* IPSEC */
326 
327 	/*
328 	 * Calculate the total length of the extension header chain.
329 	 * Keep the length of the unfragmentable part for fragmentation.
330 	 */
331 	optlen = 0;
332 	if (exthdrs.ip6e_hbh)
333 		optlen += exthdrs.ip6e_hbh->m_len;
334 	if (exthdrs.ip6e_dest1)
335 		optlen += exthdrs.ip6e_dest1->m_len;
336 	if (exthdrs.ip6e_rthdr)
337 		optlen += exthdrs.ip6e_rthdr->m_len;
338 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
339 
340 	/* NOTE: we don't add AH/ESP length here. do that later. */
341 	if (exthdrs.ip6e_dest2)
342 		optlen += exthdrs.ip6e_dest2->m_len;
343 
344 	/*
345 	 * If we need IPsec, or there is at least one extension header,
346 	 * separate IP6 header from the payload.
347 	 */
348 	if ((needipsec || optlen) && !hdrsplit) {
349 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
350 			m = NULL;
351 			goto freehdrs;
352 		}
353 		m = exthdrs.ip6e_ip6;
354 		hdrsplit++;
355 	}
356 
357 	/* adjust pointer */
358 	ip6 = mtod(m, struct ip6_hdr *);
359 
360 	/* adjust mbuf packet header length */
361 	m->m_pkthdr.len += optlen;
362 	plen = m->m_pkthdr.len - sizeof(*ip6);
363 
364 	/* If this is a jumbo payload, insert a jumbo payload option. */
365 	if (plen > IPV6_MAXPACKET) {
366 		if (!hdrsplit) {
367 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
368 				m = NULL;
369 				goto freehdrs;
370 			}
371 			m = exthdrs.ip6e_ip6;
372 			hdrsplit++;
373 		}
374 		/* adjust pointer */
375 		ip6 = mtod(m, struct ip6_hdr *);
376 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
377 			goto freehdrs;
378 		ip6->ip6_plen = 0;
379 	} else
380 		ip6->ip6_plen = htons(plen);
381 
382 	/*
383 	 * Concatenate headers and fill in next header fields.
384 	 * Here we have, on "m"
385 	 *	IPv6 payload
386 	 * and we insert headers accordingly.  Finally, we should be getting:
387 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
388 	 *
389 	 * during the header composing process, "m" points to IPv6 header.
390 	 * "mprev" points to an extension header prior to esp.
391 	 */
392 	u_char *nexthdrp = &ip6->ip6_nxt;
393 	mprev = m;
394 
395 	/*
396 	 * we treat dest2 specially.  this makes IPsec processing
397 	 * much easier.  the goal here is to make mprev point the
398 	 * mbuf prior to dest2.
399 	 *
400 	 * result: IPv6 dest2 payload
401 	 * m and mprev will point to IPv6 header.
402 	 */
403 	if (exthdrs.ip6e_dest2) {
404 		if (!hdrsplit)
405 			panic("assumption failed: hdr not split");
406 		exthdrs.ip6e_dest2->m_next = m->m_next;
407 		m->m_next = exthdrs.ip6e_dest2;
408 		*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
409 		ip6->ip6_nxt = IPPROTO_DSTOPTS;
410 	}
411 
412 	/*
413 	 * result: IPv6 hbh dest1 rthdr dest2 payload
414 	 * m will point to IPv6 header.  mprev will point to the
415 	 * extension header prior to dest2 (rthdr in the above case).
416 	 */
417 	MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
418 	MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
419 		   IPPROTO_DSTOPTS);
420 	MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
421 		   IPPROTO_ROUTING);
422 
423 #ifdef IPSEC
424 	if (!needipsec)
425 		goto skip_ipsec2;
426 
427 	/*
428 	 * pointers after IPsec headers are not valid any more.
429 	 * other pointers need a great care too.
430 	 * (IPsec routines should not mangle mbufs prior to AH/ESP)
431 	 */
432 	exthdrs.ip6e_dest2 = NULL;
433 
434 	if (exthdrs.ip6e_rthdr) {
435 		rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
436 		segleft_org = rh->ip6r_segleft;
437 		rh->ip6r_segleft = 0;
438 	}
439 
440 	bzero(&state, sizeof(state));
441 	state.m = m;
442 	error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
443 				    &needipsectun);
444 	m = state.m;
445 	if (error == EJUSTRETURN) {
446 		/*
447 		 * We had a SP with a level of 'use' and no SA. We
448 		 * will just continue to process the packet without
449 		 * IPsec processing.
450 		 */
451 		;
452 	} else if (error) {
453 		/* mbuf is already reclaimed in ipsec6_output_trans. */
454 		m = NULL;
455 		switch (error) {
456 		case EHOSTUNREACH:
457 		case ENETUNREACH:
458 		case EMSGSIZE:
459 		case ENOBUFS:
460 		case ENOMEM:
461 			break;
462 		default:
463 			printf("[%s:%d] (ipsec): error code %d\n",
464 			    __func__, __LINE__, error);
465 			/* FALLTHROUGH */
466 		case ENOENT:
467 			/* don't show these error codes to the user */
468 			error = 0;
469 			break;
470 		}
471 		goto bad;
472 	} else if (!needipsectun) {
473 		/*
474 		 * In the FAST IPSec case we have already
475 		 * re-injected the packet and it has been freed
476 		 * by the ipsec_done() function.  So, just clean
477 		 * up after ourselves.
478 		 */
479 		m = NULL;
480 		goto done;
481 	}
482 	if (exthdrs.ip6e_rthdr) {
483 		/* ah6_output doesn't modify mbuf chain */
484 		rh->ip6r_segleft = segleft_org;
485 	}
486 skip_ipsec2:;
487 #endif /* IPSEC */
488 
489 	/*
490 	 * If there is a routing header, discard the packet.
491 	 */
492 	if (exthdrs.ip6e_rthdr) {
493 		 error = EINVAL;
494 		 goto bad;
495 	}
496 
497 	/* Source address validation */
498 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
499 	    (flags & IPV6_UNSPECSRC) == 0) {
500 		error = EOPNOTSUPP;
501 		V_ip6stat.ip6s_badscope++;
502 		goto bad;
503 	}
504 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
505 		error = EOPNOTSUPP;
506 		V_ip6stat.ip6s_badscope++;
507 		goto bad;
508 	}
509 
510 	V_ip6stat.ip6s_localout++;
511 
512 	/*
513 	 * Route packet.
514 	 */
515 	if (ro == 0) {
516 		ro = &ip6route;
517 		bzero((caddr_t)ro, sizeof(*ro));
518 	}
519 	ro_pmtu = ro;
520 	if (opt && opt->ip6po_rthdr)
521 		ro = &opt->ip6po_route;
522 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
523 #ifdef FLOWTABLE
524 	if (ro->ro_rt == NULL) {
525 		struct flentry *fle;
526 
527 		/*
528 		 * The flow table returns route entries valid for up to 30
529 		 * seconds; we rely on the remainder of ip_output() taking no
530 		 * longer than that long for the stability of ro_rt.  The
531 		 * flow ID assignment must have happened before this point.
532 		 */
533 		fle = flowtable_lookup_mbuf(V_ip6_ft, m, AF_INET6);
534 		if (fle != NULL)
535 			flow_to_route_in6(fle, ro);
536 	}
537 #endif
538 again:
539 	/*
540 	 * if specified, try to fill in the traffic class field.
541 	 * do not override if a non-zero value is already set.
542 	 * we check the diffserv field and the ecn field separately.
543 	 */
544 	if (opt && opt->ip6po_tclass >= 0) {
545 		int mask = 0;
546 
547 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
548 			mask |= 0xfc;
549 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
550 			mask |= 0x03;
551 		if (mask != 0)
552 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
553 	}
554 
555 	/* fill in or override the hop limit field, if necessary. */
556 	if (opt && opt->ip6po_hlim != -1)
557 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
558 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
559 		if (im6o != NULL)
560 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
561 		else
562 			ip6->ip6_hlim = V_ip6_defmcasthlim;
563 	}
564 
565 #ifdef IPSEC
566 	/*
567 	 * We may re-inject packets into the stack here.
568 	 */
569 	if (needipsec && needipsectun) {
570 		struct ipsec_output_state state;
571 
572 		/*
573 		 * All the extension headers will become inaccessible
574 		 * (since they can be encrypted).
575 		 * Don't panic, we need no more updates to extension headers
576 		 * on inner IPv6 packet (since they are now encapsulated).
577 		 *
578 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
579 		 */
580 		bzero(&exthdrs, sizeof(exthdrs));
581 		exthdrs.ip6e_ip6 = m;
582 
583 		bzero(&state, sizeof(state));
584 		state.m = m;
585 		state.ro = (struct route *)ro;
586 		state.dst = (struct sockaddr *)dst;
587 
588 		error = ipsec6_output_tunnel(&state, sp, flags);
589 
590 		m = state.m;
591 		ro = (struct route_in6 *)state.ro;
592 		dst = (struct sockaddr_in6 *)state.dst;
593 		if (error == EJUSTRETURN) {
594 			/*
595 			 * We had a SP with a level of 'use' and no SA. We
596 			 * will just continue to process the packet without
597 			 * IPsec processing.
598 			 */
599 			;
600 		} else if (error) {
601 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
602 			m0 = m = NULL;
603 			m = NULL;
604 			switch (error) {
605 			case EHOSTUNREACH:
606 			case ENETUNREACH:
607 			case EMSGSIZE:
608 			case ENOBUFS:
609 			case ENOMEM:
610 				break;
611 			default:
612 				printf("[%s:%d] (ipsec): error code %d\n",
613 				    __func__, __LINE__, error);
614 				/* FALLTHROUGH */
615 			case ENOENT:
616 				/* don't show these error codes to the user */
617 				error = 0;
618 				break;
619 			}
620 			goto bad;
621 		} else {
622 			/*
623 			 * In the FAST IPSec case we have already
624 			 * re-injected the packet and it has been freed
625 			 * by the ipsec_done() function.  So, just clean
626 			 * up after ourselves.
627 			 */
628 			m = NULL;
629 			goto done;
630 		}
631 
632 		exthdrs.ip6e_ip6 = m;
633 	}
634 #endif /* IPSEC */
635 
636 	/* adjust pointer */
637 	ip6 = mtod(m, struct ip6_hdr *);
638 
639 	if (ro->ro_rt && fwd_tag == NULL) {
640 		rt = ro->ro_rt;
641 		ifp = ro->ro_rt->rt_ifp;
642 	} else {
643 		if (fwd_tag == NULL) {
644 			bzero(&dst_sa, sizeof(dst_sa));
645 			dst_sa.sin6_family = AF_INET6;
646 			dst_sa.sin6_len = sizeof(dst_sa);
647 			dst_sa.sin6_addr = ip6->ip6_dst;
648 		}
649 		error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
650 		    &rt, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m));
651 		if (error != 0) {
652 			if (ifp != NULL)
653 				in6_ifstat_inc(ifp, ifs6_out_discard);
654 			goto bad;
655 		}
656 	}
657 	if (rt == NULL) {
658 		/*
659 		 * If in6_selectroute() does not return a route entry,
660 		 * dst may not have been updated.
661 		 */
662 		*dst = dst_sa;	/* XXX */
663 	}
664 
665 	/*
666 	 * then rt (for unicast) and ifp must be non-NULL valid values.
667 	 */
668 	if ((flags & IPV6_FORWARDING) == 0) {
669 		/* XXX: the FORWARDING flag can be set for mrouting. */
670 		in6_ifstat_inc(ifp, ifs6_out_request);
671 	}
672 	if (rt != NULL) {
673 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
674 		rt->rt_use++;
675 	}
676 
677 
678 	/*
679 	 * The outgoing interface must be in the zone of source and
680 	 * destination addresses.
681 	 */
682 	origifp = ifp;
683 
684 	src0 = ip6->ip6_src;
685 	if (in6_setscope(&src0, origifp, &zone))
686 		goto badscope;
687 	bzero(&src_sa, sizeof(src_sa));
688 	src_sa.sin6_family = AF_INET6;
689 	src_sa.sin6_len = sizeof(src_sa);
690 	src_sa.sin6_addr = ip6->ip6_src;
691 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
692 		goto badscope;
693 
694 	dst0 = ip6->ip6_dst;
695 	if (in6_setscope(&dst0, origifp, &zone))
696 		goto badscope;
697 	/* re-initialize to be sure */
698 	bzero(&dst_sa, sizeof(dst_sa));
699 	dst_sa.sin6_family = AF_INET6;
700 	dst_sa.sin6_len = sizeof(dst_sa);
701 	dst_sa.sin6_addr = ip6->ip6_dst;
702 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
703 		goto badscope;
704 	}
705 
706 	/* We should use ia_ifp to support the case of
707 	 * sending packets to an address of our own.
708 	 */
709 	if (ia != NULL && ia->ia_ifp)
710 		ifp = ia->ia_ifp;
711 
712 	/* scope check is done. */
713 	goto routefound;
714 
715   badscope:
716 	V_ip6stat.ip6s_badscope++;
717 	in6_ifstat_inc(origifp, ifs6_out_discard);
718 	if (error == 0)
719 		error = EHOSTUNREACH; /* XXX */
720 	goto bad;
721 
722   routefound:
723 	if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
724 		if (opt && opt->ip6po_nextroute.ro_rt) {
725 			/*
726 			 * The nexthop is explicitly specified by the
727 			 * application.  We assume the next hop is an IPv6
728 			 * address.
729 			 */
730 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
731 		}
732 		else if ((rt->rt_flags & RTF_GATEWAY))
733 			dst = (struct sockaddr_in6 *)rt->rt_gateway;
734 	}
735 
736 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
737 		m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
738 	} else {
739 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
740 		in6_ifstat_inc(ifp, ifs6_out_mcast);
741 		/*
742 		 * Confirm that the outgoing interface supports multicast.
743 		 */
744 		if (!(ifp->if_flags & IFF_MULTICAST)) {
745 			V_ip6stat.ip6s_noroute++;
746 			in6_ifstat_inc(ifp, ifs6_out_discard);
747 			error = ENETUNREACH;
748 			goto bad;
749 		}
750 		if ((im6o == NULL && in6_mcast_loop) ||
751 		    (im6o && im6o->im6o_multicast_loop)) {
752 			/*
753 			 * Loop back multicast datagram if not expressly
754 			 * forbidden to do so, even if we have not joined
755 			 * the address; protocols will filter it later,
756 			 * thus deferring a hash lookup and lock acquisition
757 			 * at the expense of an m_copym().
758 			 */
759 			ip6_mloopback(ifp, m, dst);
760 		} else {
761 			/*
762 			 * If we are acting as a multicast router, perform
763 			 * multicast forwarding as if the packet had just
764 			 * arrived on the interface to which we are about
765 			 * to send.  The multicast forwarding function
766 			 * recursively calls this function, using the
767 			 * IPV6_FORWARDING flag to prevent infinite recursion.
768 			 *
769 			 * Multicasts that are looped back by ip6_mloopback(),
770 			 * above, will be forwarded by the ip6_input() routine,
771 			 * if necessary.
772 			 */
773 			if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
774 				/*
775 				 * XXX: ip6_mforward expects that rcvif is NULL
776 				 * when it is called from the originating path.
777 				 * However, it is not always the case, since
778 				 * some versions of MGETHDR() does not
779 				 * initialize the field.
780 				 */
781 				m->m_pkthdr.rcvif = NULL;
782 				if (ip6_mforward(ip6, ifp, m) != 0) {
783 					m_freem(m);
784 					goto done;
785 				}
786 			}
787 		}
788 		/*
789 		 * Multicasts with a hoplimit of zero may be looped back,
790 		 * above, but must not be transmitted on a network.
791 		 * Also, multicasts addressed to the loopback interface
792 		 * are not sent -- the above call to ip6_mloopback() will
793 		 * loop back a copy if this host actually belongs to the
794 		 * destination group on the loopback interface.
795 		 */
796 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
797 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
798 			m_freem(m);
799 			goto done;
800 		}
801 	}
802 
803 	/*
804 	 * Fill the outgoing inteface to tell the upper layer
805 	 * to increment per-interface statistics.
806 	 */
807 	if (ifpp)
808 		*ifpp = ifp;
809 
810 	/* Determine path MTU. */
811 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
812 	    &alwaysfrag, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m))) != 0)
813 		goto bad;
814 
815 	/*
816 	 * The caller of this function may specify to use the minimum MTU
817 	 * in some cases.
818 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
819 	 * setting.  The logic is a bit complicated; by default, unicast
820 	 * packets will follow path MTU while multicast packets will be sent at
821 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
822 	 * including unicast ones will be sent at the minimum MTU.  Multicast
823 	 * packets will always be sent at the minimum MTU unless
824 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
825 	 * See RFC 3542 for more details.
826 	 */
827 	if (mtu > IPV6_MMTU) {
828 		if ((flags & IPV6_MINMTU))
829 			mtu = IPV6_MMTU;
830 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
831 			mtu = IPV6_MMTU;
832 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
833 			 (opt == NULL ||
834 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
835 			mtu = IPV6_MMTU;
836 		}
837 	}
838 
839 	/*
840 	 * clear embedded scope identifiers if necessary.
841 	 * in6_clearscope will touch the addresses only when necessary.
842 	 */
843 	in6_clearscope(&ip6->ip6_src);
844 	in6_clearscope(&ip6->ip6_dst);
845 
846 	/*
847 	 * If the outgoing packet contains a hop-by-hop options header,
848 	 * it must be examined and processed even by the source node.
849 	 * (RFC 2460, section 4.)
850 	 */
851 	if (exthdrs.ip6e_hbh) {
852 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
853 		u_int32_t dummy; /* XXX unused */
854 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
855 
856 #ifdef DIAGNOSTIC
857 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
858 			panic("ip6e_hbh is not contiguous");
859 #endif
860 		/*
861 		 *  XXX: if we have to send an ICMPv6 error to the sender,
862 		 *       we need the M_LOOP flag since icmp6_error() expects
863 		 *       the IPv6 and the hop-by-hop options header are
864 		 *       contiguous unless the flag is set.
865 		 */
866 		m->m_flags |= M_LOOP;
867 		m->m_pkthdr.rcvif = ifp;
868 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
869 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
870 		    &dummy, &plen) < 0) {
871 			/* m was already freed at this point */
872 			error = EINVAL;/* better error? */
873 			goto done;
874 		}
875 		m->m_flags &= ~M_LOOP; /* XXX */
876 		m->m_pkthdr.rcvif = NULL;
877 	}
878 
879 	/* Jump over all PFIL processing if hooks are not active. */
880 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
881 		goto passout;
882 
883 	odst = ip6->ip6_dst;
884 	/* Run through list of hooks for output packets. */
885 	error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
886 	if (error != 0 || m == NULL)
887 		goto done;
888 	ip6 = mtod(m, struct ip6_hdr *);
889 
890 	/* See if destination IP address was changed by packet filter. */
891 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
892 		m->m_flags |= M_SKIP_FIREWALL;
893 		/* If destination is now ourself drop to ip6_input(). */
894 		if (in6_localip(&ip6->ip6_dst)) {
895 			m->m_flags |= M_FASTFWD_OURS;
896 			if (m->m_pkthdr.rcvif == NULL)
897 				m->m_pkthdr.rcvif = V_loif;
898 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
899 				m->m_pkthdr.csum_flags |=
900 				    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
901 				m->m_pkthdr.csum_data = 0xffff;
902 			}
903 #ifdef SCTP
904 			if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
905 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
906 #endif
907 			error = netisr_queue(NETISR_IPV6, m);
908 			goto done;
909 		} else
910 			goto again;	/* Redo the routing table lookup. */
911 	}
912 
913 	/* See if local, if yes, send it to netisr. */
914 	if (m->m_flags & M_FASTFWD_OURS) {
915 		if (m->m_pkthdr.rcvif == NULL)
916 			m->m_pkthdr.rcvif = V_loif;
917 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
918 			m->m_pkthdr.csum_flags |=
919 			    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
920 			m->m_pkthdr.csum_data = 0xffff;
921 		}
922 #ifdef SCTP
923 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
924 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
925 #endif
926 		error = netisr_queue(NETISR_IPV6, m);
927 		goto done;
928 	}
929 	/* Or forward to some other address? */
930 	if ((m->m_flags & M_IP6_NEXTHOP) &&
931 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
932 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
933 		bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
934 		m->m_flags |= M_SKIP_FIREWALL;
935 		m->m_flags &= ~M_IP6_NEXTHOP;
936 		m_tag_delete(m, fwd_tag);
937 		goto again;
938 	}
939 
940 passout:
941 	/*
942 	 * Send the packet to the outgoing interface.
943 	 * If necessary, do IPv6 fragmentation before sending.
944 	 *
945 	 * the logic here is rather complex:
946 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
947 	 * 1-a:	send as is if tlen <= path mtu
948 	 * 1-b:	fragment if tlen > path mtu
949 	 *
950 	 * 2: if user asks us not to fragment (dontfrag == 1)
951 	 * 2-a:	send as is if tlen <= interface mtu
952 	 * 2-b:	error if tlen > interface mtu
953 	 *
954 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
955 	 *	always fragment
956 	 *
957 	 * 4: if dontfrag == 1 && alwaysfrag == 1
958 	 *	error, as we cannot handle this conflicting request
959 	 */
960 	sw_csum = m->m_pkthdr.csum_flags;
961 	if (!hdrsplit) {
962 		tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
963 		sw_csum &= ~ifp->if_hwassist;
964 	} else
965 		tso = 0;
966 	/*
967 	 * If we added extension headers, we will not do TSO and calculate the
968 	 * checksums ourselves for now.
969 	 * XXX-BZ  Need a framework to know when the NIC can handle it, even
970 	 * with ext. hdrs.
971 	 */
972 	if (sw_csum & CSUM_DELAY_DATA_IPV6) {
973 		sw_csum &= ~CSUM_DELAY_DATA_IPV6;
974 		in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
975 	}
976 #ifdef SCTP
977 	if (sw_csum & CSUM_SCTP_IPV6) {
978 		sw_csum &= ~CSUM_SCTP_IPV6;
979 		sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
980 	}
981 #endif
982 	m->m_pkthdr.csum_flags &= ifp->if_hwassist;
983 	tlen = m->m_pkthdr.len;
984 
985 	if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
986 		dontfrag = 1;
987 	else
988 		dontfrag = 0;
989 	if (dontfrag && alwaysfrag) {	/* case 4 */
990 		/* conflicting request - can't transmit */
991 		error = EMSGSIZE;
992 		goto bad;
993 	}
994 	if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {	/* case 2-b */
995 		/*
996 		 * Even if the DONTFRAG option is specified, we cannot send the
997 		 * packet when the data length is larger than the MTU of the
998 		 * outgoing interface.
999 		 * Notify the error by sending IPV6_PATHMTU ancillary data as
1000 		 * well as returning an error code (the latter is not described
1001 		 * in the API spec.)
1002 		 */
1003 		u_int32_t mtu32;
1004 		struct ip6ctlparam ip6cp;
1005 
1006 		mtu32 = (u_int32_t)mtu;
1007 		bzero(&ip6cp, sizeof(ip6cp));
1008 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
1009 		pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
1010 		    (void *)&ip6cp);
1011 
1012 		error = EMSGSIZE;
1013 		goto bad;
1014 	}
1015 
1016 	/*
1017 	 * transmit packet without fragmentation
1018 	 */
1019 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
1020 		struct in6_ifaddr *ia6;
1021 
1022 		ip6 = mtod(m, struct ip6_hdr *);
1023 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1024 		if (ia6) {
1025 			/* Record statistics for this interface address. */
1026 			ia6->ia_ifa.if_opackets++;
1027 			ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
1028 			ifa_free(&ia6->ia_ifa);
1029 		}
1030 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1031 		goto done;
1032 	}
1033 
1034 	/*
1035 	 * try to fragment the packet.  case 1-b and 3
1036 	 */
1037 	if (mtu < IPV6_MMTU) {
1038 		/* path MTU cannot be less than IPV6_MMTU */
1039 		error = EMSGSIZE;
1040 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1041 		goto bad;
1042 	} else if (ip6->ip6_plen == 0) {
1043 		/* jumbo payload cannot be fragmented */
1044 		error = EMSGSIZE;
1045 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1046 		goto bad;
1047 	} else {
1048 		struct mbuf **mnext, *m_frgpart;
1049 		struct ip6_frag *ip6f;
1050 		u_int32_t id = htonl(ip6_randomid());
1051 		u_char nextproto;
1052 
1053 		int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
1054 
1055 		/*
1056 		 * Too large for the destination or interface;
1057 		 * fragment if possible.
1058 		 * Must be able to put at least 8 bytes per fragment.
1059 		 */
1060 		hlen = unfragpartlen;
1061 		if (mtu > IPV6_MAXPACKET)
1062 			mtu = IPV6_MAXPACKET;
1063 
1064 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1065 		if (len < 8) {
1066 			error = EMSGSIZE;
1067 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
1068 			goto bad;
1069 		}
1070 
1071 		/*
1072 		 * Verify that we have any chance at all of being able to queue
1073 		 *      the packet or packet fragments
1074 		 */
1075 		if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
1076 		    < tlen  /* - hlen */)) {
1077 			error = ENOBUFS;
1078 			V_ip6stat.ip6s_odropped++;
1079 			goto bad;
1080 		}
1081 
1082 
1083 		/*
1084 		 * If the interface will not calculate checksums on
1085 		 * fragmented packets, then do it here.
1086 		 * XXX-BZ handle the hw offloading case.  Need flags.
1087 		 */
1088 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1089 			in6_delayed_cksum(m, plen, hlen);
1090 			m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
1091 		}
1092 #ifdef SCTP
1093 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1094 			sctp_delayed_cksum(m, hlen);
1095 			m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1096 		}
1097 #endif
1098 		mnext = &m->m_nextpkt;
1099 
1100 		/*
1101 		 * Change the next header field of the last header in the
1102 		 * unfragmentable part.
1103 		 */
1104 		if (exthdrs.ip6e_rthdr) {
1105 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1106 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1107 		} else if (exthdrs.ip6e_dest1) {
1108 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1109 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1110 		} else if (exthdrs.ip6e_hbh) {
1111 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1112 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1113 		} else {
1114 			nextproto = ip6->ip6_nxt;
1115 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1116 		}
1117 
1118 		/*
1119 		 * Loop through length of segment after first fragment,
1120 		 * make new header and copy data of each part and link onto
1121 		 * chain.
1122 		 */
1123 		m0 = m;
1124 		for (off = hlen; off < tlen; off += len) {
1125 			MGETHDR(m, M_NOWAIT, MT_HEADER);
1126 			if (!m) {
1127 				error = ENOBUFS;
1128 				V_ip6stat.ip6s_odropped++;
1129 				goto sendorfree;
1130 			}
1131 			m->m_pkthdr.rcvif = NULL;
1132 			m->m_flags = m0->m_flags & M_COPYFLAGS;	/* incl. FIB */
1133 			*mnext = m;
1134 			mnext = &m->m_nextpkt;
1135 			m->m_data += max_linkhdr;
1136 			mhip6 = mtod(m, struct ip6_hdr *);
1137 			*mhip6 = *ip6;
1138 			m->m_len = sizeof(*mhip6);
1139 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1140 			if (error) {
1141 				V_ip6stat.ip6s_odropped++;
1142 				goto sendorfree;
1143 			}
1144 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1145 			if (off + len >= tlen)
1146 				len = tlen - off;
1147 			else
1148 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1149 			mhip6->ip6_plen = htons((u_short)(len + hlen +
1150 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1151 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1152 				error = ENOBUFS;
1153 				V_ip6stat.ip6s_odropped++;
1154 				goto sendorfree;
1155 			}
1156 			m_cat(m, m_frgpart);
1157 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1158 			m->m_pkthdr.rcvif = NULL;
1159 			ip6f->ip6f_reserved = 0;
1160 			ip6f->ip6f_ident = id;
1161 			ip6f->ip6f_nxt = nextproto;
1162 			V_ip6stat.ip6s_ofragments++;
1163 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1164 		}
1165 
1166 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1167 	}
1168 
1169 	/*
1170 	 * Remove leading garbages.
1171 	 */
1172 sendorfree:
1173 	m = m0->m_nextpkt;
1174 	m0->m_nextpkt = 0;
1175 	m_freem(m0);
1176 	for (m0 = m; m; m = m0) {
1177 		m0 = m->m_nextpkt;
1178 		m->m_nextpkt = 0;
1179 		if (error == 0) {
1180 			/* Record statistics for this interface address. */
1181 			if (ia) {
1182 				ia->ia_ifa.if_opackets++;
1183 				ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1184 			}
1185 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1186 		} else
1187 			m_freem(m);
1188 	}
1189 
1190 	if (error == 0)
1191 		V_ip6stat.ip6s_fragmented++;
1192 
1193 done:
1194 	if (ro == &ip6route)
1195 		RO_RTFREE(ro);
1196 	if (ro_pmtu == &ip6route)
1197 		RO_RTFREE(ro_pmtu);
1198 #ifdef IPSEC
1199 	if (sp != NULL)
1200 		KEY_FREESP(&sp);
1201 #endif
1202 
1203 	return (error);
1204 
1205 freehdrs:
1206 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1207 	m_freem(exthdrs.ip6e_dest1);
1208 	m_freem(exthdrs.ip6e_rthdr);
1209 	m_freem(exthdrs.ip6e_dest2);
1210 	/* FALLTHROUGH */
1211 bad:
1212 	if (m)
1213 		m_freem(m);
1214 	goto done;
1215 }
1216 
1217 static int
1218 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1219 {
1220 	struct mbuf *m;
1221 
1222 	if (hlen > MCLBYTES)
1223 		return (ENOBUFS); /* XXX */
1224 
1225 	MGET(m, M_NOWAIT, MT_DATA);
1226 	if (!m)
1227 		return (ENOBUFS);
1228 
1229 	if (hlen > MLEN) {
1230 		MCLGET(m, M_NOWAIT);
1231 		if ((m->m_flags & M_EXT) == 0) {
1232 			m_free(m);
1233 			return (ENOBUFS);
1234 		}
1235 	}
1236 	m->m_len = hlen;
1237 	if (hdr)
1238 		bcopy(hdr, mtod(m, caddr_t), hlen);
1239 
1240 	*mp = m;
1241 	return (0);
1242 }
1243 
1244 /*
1245  * Insert jumbo payload option.
1246  */
1247 static int
1248 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1249 {
1250 	struct mbuf *mopt;
1251 	u_char *optbuf;
1252 	u_int32_t v;
1253 
1254 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1255 
1256 	/*
1257 	 * If there is no hop-by-hop options header, allocate new one.
1258 	 * If there is one but it doesn't have enough space to store the
1259 	 * jumbo payload option, allocate a cluster to store the whole options.
1260 	 * Otherwise, use it to store the options.
1261 	 */
1262 	if (exthdrs->ip6e_hbh == 0) {
1263 		MGET(mopt, M_NOWAIT, MT_DATA);
1264 		if (mopt == 0)
1265 			return (ENOBUFS);
1266 		mopt->m_len = JUMBOOPTLEN;
1267 		optbuf = mtod(mopt, u_char *);
1268 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1269 		exthdrs->ip6e_hbh = mopt;
1270 	} else {
1271 		struct ip6_hbh *hbh;
1272 
1273 		mopt = exthdrs->ip6e_hbh;
1274 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1275 			/*
1276 			 * XXX assumption:
1277 			 * - exthdrs->ip6e_hbh is not referenced from places
1278 			 *   other than exthdrs.
1279 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1280 			 */
1281 			int oldoptlen = mopt->m_len;
1282 			struct mbuf *n;
1283 
1284 			/*
1285 			 * XXX: give up if the whole (new) hbh header does
1286 			 * not fit even in an mbuf cluster.
1287 			 */
1288 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1289 				return (ENOBUFS);
1290 
1291 			/*
1292 			 * As a consequence, we must always prepare a cluster
1293 			 * at this point.
1294 			 */
1295 			MGET(n, M_NOWAIT, MT_DATA);
1296 			if (n) {
1297 				MCLGET(n, M_NOWAIT);
1298 				if ((n->m_flags & M_EXT) == 0) {
1299 					m_freem(n);
1300 					n = NULL;
1301 				}
1302 			}
1303 			if (!n)
1304 				return (ENOBUFS);
1305 			n->m_len = oldoptlen + JUMBOOPTLEN;
1306 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1307 			    oldoptlen);
1308 			optbuf = mtod(n, caddr_t) + oldoptlen;
1309 			m_freem(mopt);
1310 			mopt = exthdrs->ip6e_hbh = n;
1311 		} else {
1312 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1313 			mopt->m_len += JUMBOOPTLEN;
1314 		}
1315 		optbuf[0] = IP6OPT_PADN;
1316 		optbuf[1] = 1;
1317 
1318 		/*
1319 		 * Adjust the header length according to the pad and
1320 		 * the jumbo payload option.
1321 		 */
1322 		hbh = mtod(mopt, struct ip6_hbh *);
1323 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1324 	}
1325 
1326 	/* fill in the option. */
1327 	optbuf[2] = IP6OPT_JUMBO;
1328 	optbuf[3] = 4;
1329 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1330 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1331 
1332 	/* finally, adjust the packet header length */
1333 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1334 
1335 	return (0);
1336 #undef JUMBOOPTLEN
1337 }
1338 
1339 /*
1340  * Insert fragment header and copy unfragmentable header portions.
1341  */
1342 static int
1343 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1344     struct ip6_frag **frghdrp)
1345 {
1346 	struct mbuf *n, *mlast;
1347 
1348 	if (hlen > sizeof(struct ip6_hdr)) {
1349 		n = m_copym(m0, sizeof(struct ip6_hdr),
1350 		    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1351 		if (n == 0)
1352 			return (ENOBUFS);
1353 		m->m_next = n;
1354 	} else
1355 		n = m;
1356 
1357 	/* Search for the last mbuf of unfragmentable part. */
1358 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1359 		;
1360 
1361 	if ((mlast->m_flags & M_EXT) == 0 &&
1362 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1363 		/* use the trailing space of the last mbuf for the fragment hdr */
1364 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1365 		    mlast->m_len);
1366 		mlast->m_len += sizeof(struct ip6_frag);
1367 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1368 	} else {
1369 		/* allocate a new mbuf for the fragment header */
1370 		struct mbuf *mfrg;
1371 
1372 		MGET(mfrg, M_NOWAIT, MT_DATA);
1373 		if (mfrg == 0)
1374 			return (ENOBUFS);
1375 		mfrg->m_len = sizeof(struct ip6_frag);
1376 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1377 		mlast->m_next = mfrg;
1378 	}
1379 
1380 	return (0);
1381 }
1382 
1383 static int
1384 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1385     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1386     int *alwaysfragp, u_int fibnum)
1387 {
1388 	u_int32_t mtu = 0;
1389 	int alwaysfrag = 0;
1390 	int error = 0;
1391 
1392 	if (ro_pmtu != ro) {
1393 		/* The first hop and the final destination may differ. */
1394 		struct sockaddr_in6 *sa6_dst =
1395 		    (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1396 		if (ro_pmtu->ro_rt &&
1397 		    ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1398 		     !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1399 			RTFREE(ro_pmtu->ro_rt);
1400 			ro_pmtu->ro_rt = (struct rtentry *)NULL;
1401 		}
1402 		if (ro_pmtu->ro_rt == NULL) {
1403 			bzero(sa6_dst, sizeof(*sa6_dst));
1404 			sa6_dst->sin6_family = AF_INET6;
1405 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1406 			sa6_dst->sin6_addr = *dst;
1407 
1408 			in6_rtalloc(ro_pmtu, fibnum);
1409 		}
1410 	}
1411 	if (ro_pmtu->ro_rt) {
1412 		u_int32_t ifmtu;
1413 		struct in_conninfo inc;
1414 
1415 		bzero(&inc, sizeof(inc));
1416 		inc.inc_flags |= INC_ISIPV6;
1417 		inc.inc6_faddr = *dst;
1418 
1419 		if (ifp == NULL)
1420 			ifp = ro_pmtu->ro_rt->rt_ifp;
1421 		ifmtu = IN6_LINKMTU(ifp);
1422 		mtu = tcp_hc_getmtu(&inc);
1423 		if (mtu)
1424 			mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu);
1425 		else
1426 			mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1427 		if (mtu == 0)
1428 			mtu = ifmtu;
1429 		else if (mtu < IPV6_MMTU) {
1430 			/*
1431 			 * RFC2460 section 5, last paragraph:
1432 			 * if we record ICMPv6 too big message with
1433 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1434 			 * or smaller, with framgent header attached.
1435 			 * (fragment header is needed regardless from the
1436 			 * packet size, for translators to identify packets)
1437 			 */
1438 			alwaysfrag = 1;
1439 			mtu = IPV6_MMTU;
1440 		} else if (mtu > ifmtu) {
1441 			/*
1442 			 * The MTU on the route is larger than the MTU on
1443 			 * the interface!  This shouldn't happen, unless the
1444 			 * MTU of the interface has been changed after the
1445 			 * interface was brought up.  Change the MTU in the
1446 			 * route to match the interface MTU (as long as the
1447 			 * field isn't locked).
1448 			 */
1449 			mtu = ifmtu;
1450 			ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1451 		}
1452 	} else if (ifp) {
1453 		mtu = IN6_LINKMTU(ifp);
1454 	} else
1455 		error = EHOSTUNREACH; /* XXX */
1456 
1457 	*mtup = mtu;
1458 	if (alwaysfragp)
1459 		*alwaysfragp = alwaysfrag;
1460 	return (error);
1461 }
1462 
1463 /*
1464  * IP6 socket option processing.
1465  */
1466 int
1467 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1468 {
1469 	int optdatalen, uproto;
1470 	void *optdata;
1471 	struct inpcb *in6p = sotoinpcb(so);
1472 	int error, optval;
1473 	int level, op, optname;
1474 	int optlen;
1475 	struct thread *td;
1476 
1477 	level = sopt->sopt_level;
1478 	op = sopt->sopt_dir;
1479 	optname = sopt->sopt_name;
1480 	optlen = sopt->sopt_valsize;
1481 	td = sopt->sopt_td;
1482 	error = 0;
1483 	optval = 0;
1484 	uproto = (int)so->so_proto->pr_protocol;
1485 
1486 	if (level != IPPROTO_IPV6) {
1487 		error = EINVAL;
1488 
1489 		if (sopt->sopt_level == SOL_SOCKET &&
1490 		    sopt->sopt_dir == SOPT_SET) {
1491 			switch (sopt->sopt_name) {
1492 			case SO_REUSEADDR:
1493 				INP_WLOCK(in6p);
1494 				if (IN_MULTICAST(ntohl(in6p->inp_laddr.s_addr))) {
1495 					if ((so->so_options &
1496 					    (SO_REUSEADDR | SO_REUSEPORT)) != 0)
1497 						in6p->inp_flags2 |= INP_REUSEPORT;
1498 					else
1499 						in6p->inp_flags2 &= ~INP_REUSEPORT;
1500 				}
1501 				INP_WUNLOCK(in6p);
1502 				error = 0;
1503 				break;
1504 			case SO_REUSEPORT:
1505 				INP_WLOCK(in6p);
1506 				if ((so->so_options & SO_REUSEPORT) != 0)
1507 					in6p->inp_flags2 |= INP_REUSEPORT;
1508 				else
1509 					in6p->inp_flags2 &= ~INP_REUSEPORT;
1510 				INP_WUNLOCK(in6p);
1511 				error = 0;
1512 				break;
1513 			case SO_SETFIB:
1514 				INP_WLOCK(in6p);
1515 				in6p->inp_inc.inc_fibnum = so->so_fibnum;
1516 				INP_WUNLOCK(in6p);
1517 				error = 0;
1518 				break;
1519 			default:
1520 				break;
1521 			}
1522 		}
1523 	} else {		/* level == IPPROTO_IPV6 */
1524 		switch (op) {
1525 
1526 		case SOPT_SET:
1527 			switch (optname) {
1528 			case IPV6_2292PKTOPTIONS:
1529 #ifdef IPV6_PKTOPTIONS
1530 			case IPV6_PKTOPTIONS:
1531 #endif
1532 			{
1533 				struct mbuf *m;
1534 
1535 				error = soopt_getm(sopt, &m); /* XXX */
1536 				if (error != 0)
1537 					break;
1538 				error = soopt_mcopyin(sopt, m); /* XXX */
1539 				if (error != 0)
1540 					break;
1541 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1542 						    m, so, sopt);
1543 				m_freem(m); /* XXX */
1544 				break;
1545 			}
1546 
1547 			/*
1548 			 * Use of some Hop-by-Hop options or some
1549 			 * Destination options, might require special
1550 			 * privilege.  That is, normal applications
1551 			 * (without special privilege) might be forbidden
1552 			 * from setting certain options in outgoing packets,
1553 			 * and might never see certain options in received
1554 			 * packets. [RFC 2292 Section 6]
1555 			 * KAME specific note:
1556 			 *  KAME prevents non-privileged users from sending or
1557 			 *  receiving ANY hbh/dst options in order to avoid
1558 			 *  overhead of parsing options in the kernel.
1559 			 */
1560 			case IPV6_RECVHOPOPTS:
1561 			case IPV6_RECVDSTOPTS:
1562 			case IPV6_RECVRTHDRDSTOPTS:
1563 				if (td != NULL) {
1564 					error = priv_check(td,
1565 					    PRIV_NETINET_SETHDROPTS);
1566 					if (error)
1567 						break;
1568 				}
1569 				/* FALLTHROUGH */
1570 			case IPV6_UNICAST_HOPS:
1571 			case IPV6_HOPLIMIT:
1572 			case IPV6_FAITH:
1573 
1574 			case IPV6_RECVPKTINFO:
1575 			case IPV6_RECVHOPLIMIT:
1576 			case IPV6_RECVRTHDR:
1577 			case IPV6_RECVPATHMTU:
1578 			case IPV6_RECVTCLASS:
1579 			case IPV6_V6ONLY:
1580 			case IPV6_AUTOFLOWLABEL:
1581 			case IPV6_BINDANY:
1582 				if (optname == IPV6_BINDANY && td != NULL) {
1583 					error = priv_check(td,
1584 					    PRIV_NETINET_BINDANY);
1585 					if (error)
1586 						break;
1587 				}
1588 
1589 				if (optlen != sizeof(int)) {
1590 					error = EINVAL;
1591 					break;
1592 				}
1593 				error = sooptcopyin(sopt, &optval,
1594 					sizeof optval, sizeof optval);
1595 				if (error)
1596 					break;
1597 				switch (optname) {
1598 
1599 				case IPV6_UNICAST_HOPS:
1600 					if (optval < -1 || optval >= 256)
1601 						error = EINVAL;
1602 					else {
1603 						/* -1 = kernel default */
1604 						in6p->in6p_hops = optval;
1605 						if ((in6p->inp_vflag &
1606 						     INP_IPV4) != 0)
1607 							in6p->inp_ip_ttl = optval;
1608 					}
1609 					break;
1610 #define OPTSET(bit) \
1611 do { \
1612 	INP_WLOCK(in6p); \
1613 	if (optval) \
1614 		in6p->inp_flags |= (bit); \
1615 	else \
1616 		in6p->inp_flags &= ~(bit); \
1617 	INP_WUNLOCK(in6p); \
1618 } while (/*CONSTCOND*/ 0)
1619 #define OPTSET2292(bit) \
1620 do { \
1621 	INP_WLOCK(in6p); \
1622 	in6p->inp_flags |= IN6P_RFC2292; \
1623 	if (optval) \
1624 		in6p->inp_flags |= (bit); \
1625 	else \
1626 		in6p->inp_flags &= ~(bit); \
1627 	INP_WUNLOCK(in6p); \
1628 } while (/*CONSTCOND*/ 0)
1629 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1630 
1631 				case IPV6_RECVPKTINFO:
1632 					/* cannot mix with RFC2292 */
1633 					if (OPTBIT(IN6P_RFC2292)) {
1634 						error = EINVAL;
1635 						break;
1636 					}
1637 					OPTSET(IN6P_PKTINFO);
1638 					break;
1639 
1640 				case IPV6_HOPLIMIT:
1641 				{
1642 					struct ip6_pktopts **optp;
1643 
1644 					/* cannot mix with RFC2292 */
1645 					if (OPTBIT(IN6P_RFC2292)) {
1646 						error = EINVAL;
1647 						break;
1648 					}
1649 					optp = &in6p->in6p_outputopts;
1650 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1651 					    (u_char *)&optval, sizeof(optval),
1652 					    optp, (td != NULL) ? td->td_ucred :
1653 					    NULL, uproto);
1654 					break;
1655 				}
1656 
1657 				case IPV6_RECVHOPLIMIT:
1658 					/* cannot mix with RFC2292 */
1659 					if (OPTBIT(IN6P_RFC2292)) {
1660 						error = EINVAL;
1661 						break;
1662 					}
1663 					OPTSET(IN6P_HOPLIMIT);
1664 					break;
1665 
1666 				case IPV6_RECVHOPOPTS:
1667 					/* cannot mix with RFC2292 */
1668 					if (OPTBIT(IN6P_RFC2292)) {
1669 						error = EINVAL;
1670 						break;
1671 					}
1672 					OPTSET(IN6P_HOPOPTS);
1673 					break;
1674 
1675 				case IPV6_RECVDSTOPTS:
1676 					/* cannot mix with RFC2292 */
1677 					if (OPTBIT(IN6P_RFC2292)) {
1678 						error = EINVAL;
1679 						break;
1680 					}
1681 					OPTSET(IN6P_DSTOPTS);
1682 					break;
1683 
1684 				case IPV6_RECVRTHDRDSTOPTS:
1685 					/* cannot mix with RFC2292 */
1686 					if (OPTBIT(IN6P_RFC2292)) {
1687 						error = EINVAL;
1688 						break;
1689 					}
1690 					OPTSET(IN6P_RTHDRDSTOPTS);
1691 					break;
1692 
1693 				case IPV6_RECVRTHDR:
1694 					/* cannot mix with RFC2292 */
1695 					if (OPTBIT(IN6P_RFC2292)) {
1696 						error = EINVAL;
1697 						break;
1698 					}
1699 					OPTSET(IN6P_RTHDR);
1700 					break;
1701 
1702 				case IPV6_FAITH:
1703 					OPTSET(INP_FAITH);
1704 					break;
1705 
1706 				case IPV6_RECVPATHMTU:
1707 					/*
1708 					 * We ignore this option for TCP
1709 					 * sockets.
1710 					 * (RFC3542 leaves this case
1711 					 * unspecified.)
1712 					 */
1713 					if (uproto != IPPROTO_TCP)
1714 						OPTSET(IN6P_MTU);
1715 					break;
1716 
1717 				case IPV6_V6ONLY:
1718 					/*
1719 					 * make setsockopt(IPV6_V6ONLY)
1720 					 * available only prior to bind(2).
1721 					 * see ipng mailing list, Jun 22 2001.
1722 					 */
1723 					if (in6p->inp_lport ||
1724 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1725 						error = EINVAL;
1726 						break;
1727 					}
1728 					OPTSET(IN6P_IPV6_V6ONLY);
1729 					if (optval)
1730 						in6p->inp_vflag &= ~INP_IPV4;
1731 					else
1732 						in6p->inp_vflag |= INP_IPV4;
1733 					break;
1734 				case IPV6_RECVTCLASS:
1735 					/* cannot mix with RFC2292 XXX */
1736 					if (OPTBIT(IN6P_RFC2292)) {
1737 						error = EINVAL;
1738 						break;
1739 					}
1740 					OPTSET(IN6P_TCLASS);
1741 					break;
1742 				case IPV6_AUTOFLOWLABEL:
1743 					OPTSET(IN6P_AUTOFLOWLABEL);
1744 					break;
1745 
1746 				case IPV6_BINDANY:
1747 					OPTSET(INP_BINDANY);
1748 					break;
1749 				}
1750 				break;
1751 
1752 			case IPV6_TCLASS:
1753 			case IPV6_DONTFRAG:
1754 			case IPV6_USE_MIN_MTU:
1755 			case IPV6_PREFER_TEMPADDR:
1756 				if (optlen != sizeof(optval)) {
1757 					error = EINVAL;
1758 					break;
1759 				}
1760 				error = sooptcopyin(sopt, &optval,
1761 					sizeof optval, sizeof optval);
1762 				if (error)
1763 					break;
1764 				{
1765 					struct ip6_pktopts **optp;
1766 					optp = &in6p->in6p_outputopts;
1767 					error = ip6_pcbopt(optname,
1768 					    (u_char *)&optval, sizeof(optval),
1769 					    optp, (td != NULL) ? td->td_ucred :
1770 					    NULL, uproto);
1771 					break;
1772 				}
1773 
1774 			case IPV6_2292PKTINFO:
1775 			case IPV6_2292HOPLIMIT:
1776 			case IPV6_2292HOPOPTS:
1777 			case IPV6_2292DSTOPTS:
1778 			case IPV6_2292RTHDR:
1779 				/* RFC 2292 */
1780 				if (optlen != sizeof(int)) {
1781 					error = EINVAL;
1782 					break;
1783 				}
1784 				error = sooptcopyin(sopt, &optval,
1785 					sizeof optval, sizeof optval);
1786 				if (error)
1787 					break;
1788 				switch (optname) {
1789 				case IPV6_2292PKTINFO:
1790 					OPTSET2292(IN6P_PKTINFO);
1791 					break;
1792 				case IPV6_2292HOPLIMIT:
1793 					OPTSET2292(IN6P_HOPLIMIT);
1794 					break;
1795 				case IPV6_2292HOPOPTS:
1796 					/*
1797 					 * Check super-user privilege.
1798 					 * See comments for IPV6_RECVHOPOPTS.
1799 					 */
1800 					if (td != NULL) {
1801 						error = priv_check(td,
1802 						    PRIV_NETINET_SETHDROPTS);
1803 						if (error)
1804 							return (error);
1805 					}
1806 					OPTSET2292(IN6P_HOPOPTS);
1807 					break;
1808 				case IPV6_2292DSTOPTS:
1809 					if (td != NULL) {
1810 						error = priv_check(td,
1811 						    PRIV_NETINET_SETHDROPTS);
1812 						if (error)
1813 							return (error);
1814 					}
1815 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1816 					break;
1817 				case IPV6_2292RTHDR:
1818 					OPTSET2292(IN6P_RTHDR);
1819 					break;
1820 				}
1821 				break;
1822 			case IPV6_PKTINFO:
1823 			case IPV6_HOPOPTS:
1824 			case IPV6_RTHDR:
1825 			case IPV6_DSTOPTS:
1826 			case IPV6_RTHDRDSTOPTS:
1827 			case IPV6_NEXTHOP:
1828 			{
1829 				/* new advanced API (RFC3542) */
1830 				u_char *optbuf;
1831 				u_char optbuf_storage[MCLBYTES];
1832 				int optlen;
1833 				struct ip6_pktopts **optp;
1834 
1835 				/* cannot mix with RFC2292 */
1836 				if (OPTBIT(IN6P_RFC2292)) {
1837 					error = EINVAL;
1838 					break;
1839 				}
1840 
1841 				/*
1842 				 * We only ensure valsize is not too large
1843 				 * here.  Further validation will be done
1844 				 * later.
1845 				 */
1846 				error = sooptcopyin(sopt, optbuf_storage,
1847 				    sizeof(optbuf_storage), 0);
1848 				if (error)
1849 					break;
1850 				optlen = sopt->sopt_valsize;
1851 				optbuf = optbuf_storage;
1852 				optp = &in6p->in6p_outputopts;
1853 				error = ip6_pcbopt(optname, optbuf, optlen,
1854 				    optp, (td != NULL) ? td->td_ucred : NULL,
1855 				    uproto);
1856 				break;
1857 			}
1858 #undef OPTSET
1859 
1860 			case IPV6_MULTICAST_IF:
1861 			case IPV6_MULTICAST_HOPS:
1862 			case IPV6_MULTICAST_LOOP:
1863 			case IPV6_JOIN_GROUP:
1864 			case IPV6_LEAVE_GROUP:
1865 			case IPV6_MSFILTER:
1866 			case MCAST_BLOCK_SOURCE:
1867 			case MCAST_UNBLOCK_SOURCE:
1868 			case MCAST_JOIN_GROUP:
1869 			case MCAST_LEAVE_GROUP:
1870 			case MCAST_JOIN_SOURCE_GROUP:
1871 			case MCAST_LEAVE_SOURCE_GROUP:
1872 				error = ip6_setmoptions(in6p, sopt);
1873 				break;
1874 
1875 			case IPV6_PORTRANGE:
1876 				error = sooptcopyin(sopt, &optval,
1877 				    sizeof optval, sizeof optval);
1878 				if (error)
1879 					break;
1880 
1881 				INP_WLOCK(in6p);
1882 				switch (optval) {
1883 				case IPV6_PORTRANGE_DEFAULT:
1884 					in6p->inp_flags &= ~(INP_LOWPORT);
1885 					in6p->inp_flags &= ~(INP_HIGHPORT);
1886 					break;
1887 
1888 				case IPV6_PORTRANGE_HIGH:
1889 					in6p->inp_flags &= ~(INP_LOWPORT);
1890 					in6p->inp_flags |= INP_HIGHPORT;
1891 					break;
1892 
1893 				case IPV6_PORTRANGE_LOW:
1894 					in6p->inp_flags &= ~(INP_HIGHPORT);
1895 					in6p->inp_flags |= INP_LOWPORT;
1896 					break;
1897 
1898 				default:
1899 					error = EINVAL;
1900 					break;
1901 				}
1902 				INP_WUNLOCK(in6p);
1903 				break;
1904 
1905 #ifdef IPSEC
1906 			case IPV6_IPSEC_POLICY:
1907 			{
1908 				caddr_t req;
1909 				struct mbuf *m;
1910 
1911 				if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1912 					break;
1913 				if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1914 					break;
1915 				req = mtod(m, caddr_t);
1916 				error = ipsec_set_policy(in6p, optname, req,
1917 				    m->m_len, (sopt->sopt_td != NULL) ?
1918 				    sopt->sopt_td->td_ucred : NULL);
1919 				m_freem(m);
1920 				break;
1921 			}
1922 #endif /* IPSEC */
1923 
1924 			default:
1925 				error = ENOPROTOOPT;
1926 				break;
1927 			}
1928 			break;
1929 
1930 		case SOPT_GET:
1931 			switch (optname) {
1932 
1933 			case IPV6_2292PKTOPTIONS:
1934 #ifdef IPV6_PKTOPTIONS
1935 			case IPV6_PKTOPTIONS:
1936 #endif
1937 				/*
1938 				 * RFC3542 (effectively) deprecated the
1939 				 * semantics of the 2292-style pktoptions.
1940 				 * Since it was not reliable in nature (i.e.,
1941 				 * applications had to expect the lack of some
1942 				 * information after all), it would make sense
1943 				 * to simplify this part by always returning
1944 				 * empty data.
1945 				 */
1946 				sopt->sopt_valsize = 0;
1947 				break;
1948 
1949 			case IPV6_RECVHOPOPTS:
1950 			case IPV6_RECVDSTOPTS:
1951 			case IPV6_RECVRTHDRDSTOPTS:
1952 			case IPV6_UNICAST_HOPS:
1953 			case IPV6_RECVPKTINFO:
1954 			case IPV6_RECVHOPLIMIT:
1955 			case IPV6_RECVRTHDR:
1956 			case IPV6_RECVPATHMTU:
1957 
1958 			case IPV6_FAITH:
1959 			case IPV6_V6ONLY:
1960 			case IPV6_PORTRANGE:
1961 			case IPV6_RECVTCLASS:
1962 			case IPV6_AUTOFLOWLABEL:
1963 			case IPV6_BINDANY:
1964 				switch (optname) {
1965 
1966 				case IPV6_RECVHOPOPTS:
1967 					optval = OPTBIT(IN6P_HOPOPTS);
1968 					break;
1969 
1970 				case IPV6_RECVDSTOPTS:
1971 					optval = OPTBIT(IN6P_DSTOPTS);
1972 					break;
1973 
1974 				case IPV6_RECVRTHDRDSTOPTS:
1975 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1976 					break;
1977 
1978 				case IPV6_UNICAST_HOPS:
1979 					optval = in6p->in6p_hops;
1980 					break;
1981 
1982 				case IPV6_RECVPKTINFO:
1983 					optval = OPTBIT(IN6P_PKTINFO);
1984 					break;
1985 
1986 				case IPV6_RECVHOPLIMIT:
1987 					optval = OPTBIT(IN6P_HOPLIMIT);
1988 					break;
1989 
1990 				case IPV6_RECVRTHDR:
1991 					optval = OPTBIT(IN6P_RTHDR);
1992 					break;
1993 
1994 				case IPV6_RECVPATHMTU:
1995 					optval = OPTBIT(IN6P_MTU);
1996 					break;
1997 
1998 				case IPV6_FAITH:
1999 					optval = OPTBIT(INP_FAITH);
2000 					break;
2001 
2002 				case IPV6_V6ONLY:
2003 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
2004 					break;
2005 
2006 				case IPV6_PORTRANGE:
2007 				    {
2008 					int flags;
2009 					flags = in6p->inp_flags;
2010 					if (flags & INP_HIGHPORT)
2011 						optval = IPV6_PORTRANGE_HIGH;
2012 					else if (flags & INP_LOWPORT)
2013 						optval = IPV6_PORTRANGE_LOW;
2014 					else
2015 						optval = 0;
2016 					break;
2017 				    }
2018 				case IPV6_RECVTCLASS:
2019 					optval = OPTBIT(IN6P_TCLASS);
2020 					break;
2021 
2022 				case IPV6_AUTOFLOWLABEL:
2023 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2024 					break;
2025 
2026 				case IPV6_BINDANY:
2027 					optval = OPTBIT(INP_BINDANY);
2028 					break;
2029 				}
2030 				if (error)
2031 					break;
2032 				error = sooptcopyout(sopt, &optval,
2033 					sizeof optval);
2034 				break;
2035 
2036 			case IPV6_PATHMTU:
2037 			{
2038 				u_long pmtu = 0;
2039 				struct ip6_mtuinfo mtuinfo;
2040 				struct route_in6 sro;
2041 
2042 				bzero(&sro, sizeof(sro));
2043 
2044 				if (!(so->so_state & SS_ISCONNECTED))
2045 					return (ENOTCONN);
2046 				/*
2047 				 * XXX: we dot not consider the case of source
2048 				 * routing, or optional information to specify
2049 				 * the outgoing interface.
2050 				 */
2051 				error = ip6_getpmtu(&sro, NULL, NULL,
2052 				    &in6p->in6p_faddr, &pmtu, NULL,
2053 				    so->so_fibnum);
2054 				if (sro.ro_rt)
2055 					RTFREE(sro.ro_rt);
2056 				if (error)
2057 					break;
2058 				if (pmtu > IPV6_MAXPACKET)
2059 					pmtu = IPV6_MAXPACKET;
2060 
2061 				bzero(&mtuinfo, sizeof(mtuinfo));
2062 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2063 				optdata = (void *)&mtuinfo;
2064 				optdatalen = sizeof(mtuinfo);
2065 				error = sooptcopyout(sopt, optdata,
2066 				    optdatalen);
2067 				break;
2068 			}
2069 
2070 			case IPV6_2292PKTINFO:
2071 			case IPV6_2292HOPLIMIT:
2072 			case IPV6_2292HOPOPTS:
2073 			case IPV6_2292RTHDR:
2074 			case IPV6_2292DSTOPTS:
2075 				switch (optname) {
2076 				case IPV6_2292PKTINFO:
2077 					optval = OPTBIT(IN6P_PKTINFO);
2078 					break;
2079 				case IPV6_2292HOPLIMIT:
2080 					optval = OPTBIT(IN6P_HOPLIMIT);
2081 					break;
2082 				case IPV6_2292HOPOPTS:
2083 					optval = OPTBIT(IN6P_HOPOPTS);
2084 					break;
2085 				case IPV6_2292RTHDR:
2086 					optval = OPTBIT(IN6P_RTHDR);
2087 					break;
2088 				case IPV6_2292DSTOPTS:
2089 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2090 					break;
2091 				}
2092 				error = sooptcopyout(sopt, &optval,
2093 				    sizeof optval);
2094 				break;
2095 			case IPV6_PKTINFO:
2096 			case IPV6_HOPOPTS:
2097 			case IPV6_RTHDR:
2098 			case IPV6_DSTOPTS:
2099 			case IPV6_RTHDRDSTOPTS:
2100 			case IPV6_NEXTHOP:
2101 			case IPV6_TCLASS:
2102 			case IPV6_DONTFRAG:
2103 			case IPV6_USE_MIN_MTU:
2104 			case IPV6_PREFER_TEMPADDR:
2105 				error = ip6_getpcbopt(in6p->in6p_outputopts,
2106 				    optname, sopt);
2107 				break;
2108 
2109 			case IPV6_MULTICAST_IF:
2110 			case IPV6_MULTICAST_HOPS:
2111 			case IPV6_MULTICAST_LOOP:
2112 			case IPV6_MSFILTER:
2113 				error = ip6_getmoptions(in6p, sopt);
2114 				break;
2115 
2116 #ifdef IPSEC
2117 			case IPV6_IPSEC_POLICY:
2118 			  {
2119 				caddr_t req = NULL;
2120 				size_t len = 0;
2121 				struct mbuf *m = NULL;
2122 				struct mbuf **mp = &m;
2123 				size_t ovalsize = sopt->sopt_valsize;
2124 				caddr_t oval = (caddr_t)sopt->sopt_val;
2125 
2126 				error = soopt_getm(sopt, &m); /* XXX */
2127 				if (error != 0)
2128 					break;
2129 				error = soopt_mcopyin(sopt, m); /* XXX */
2130 				if (error != 0)
2131 					break;
2132 				sopt->sopt_valsize = ovalsize;
2133 				sopt->sopt_val = oval;
2134 				if (m) {
2135 					req = mtod(m, caddr_t);
2136 					len = m->m_len;
2137 				}
2138 				error = ipsec_get_policy(in6p, req, len, mp);
2139 				if (error == 0)
2140 					error = soopt_mcopyout(sopt, m); /* XXX */
2141 				if (error == 0 && m)
2142 					m_freem(m);
2143 				break;
2144 			  }
2145 #endif /* IPSEC */
2146 
2147 			default:
2148 				error = ENOPROTOOPT;
2149 				break;
2150 			}
2151 			break;
2152 		}
2153 	}
2154 	return (error);
2155 }
2156 
2157 int
2158 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2159 {
2160 	int error = 0, optval, optlen;
2161 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2162 	struct inpcb *in6p = sotoinpcb(so);
2163 	int level, op, optname;
2164 
2165 	level = sopt->sopt_level;
2166 	op = sopt->sopt_dir;
2167 	optname = sopt->sopt_name;
2168 	optlen = sopt->sopt_valsize;
2169 
2170 	if (level != IPPROTO_IPV6) {
2171 		return (EINVAL);
2172 	}
2173 
2174 	switch (optname) {
2175 	case IPV6_CHECKSUM:
2176 		/*
2177 		 * For ICMPv6 sockets, no modification allowed for checksum
2178 		 * offset, permit "no change" values to help existing apps.
2179 		 *
2180 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2181 		 * for an ICMPv6 socket will fail."
2182 		 * The current behavior does not meet RFC3542.
2183 		 */
2184 		switch (op) {
2185 		case SOPT_SET:
2186 			if (optlen != sizeof(int)) {
2187 				error = EINVAL;
2188 				break;
2189 			}
2190 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2191 					    sizeof(optval));
2192 			if (error)
2193 				break;
2194 			if ((optval % 2) != 0) {
2195 				/* the API assumes even offset values */
2196 				error = EINVAL;
2197 			} else if (so->so_proto->pr_protocol ==
2198 			    IPPROTO_ICMPV6) {
2199 				if (optval != icmp6off)
2200 					error = EINVAL;
2201 			} else
2202 				in6p->in6p_cksum = optval;
2203 			break;
2204 
2205 		case SOPT_GET:
2206 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2207 				optval = icmp6off;
2208 			else
2209 				optval = in6p->in6p_cksum;
2210 
2211 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2212 			break;
2213 
2214 		default:
2215 			error = EINVAL;
2216 			break;
2217 		}
2218 		break;
2219 
2220 	default:
2221 		error = ENOPROTOOPT;
2222 		break;
2223 	}
2224 
2225 	return (error);
2226 }
2227 
2228 /*
2229  * Set up IP6 options in pcb for insertion in output packets or
2230  * specifying behavior of outgoing packets.
2231  */
2232 static int
2233 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2234     struct socket *so, struct sockopt *sopt)
2235 {
2236 	struct ip6_pktopts *opt = *pktopt;
2237 	int error = 0;
2238 	struct thread *td = sopt->sopt_td;
2239 
2240 	/* turn off any old options. */
2241 	if (opt) {
2242 #ifdef DIAGNOSTIC
2243 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2244 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2245 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2246 			printf("ip6_pcbopts: all specified options are cleared.\n");
2247 #endif
2248 		ip6_clearpktopts(opt, -1);
2249 	} else
2250 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2251 	*pktopt = NULL;
2252 
2253 	if (!m || m->m_len == 0) {
2254 		/*
2255 		 * Only turning off any previous options, regardless of
2256 		 * whether the opt is just created or given.
2257 		 */
2258 		free(opt, M_IP6OPT);
2259 		return (0);
2260 	}
2261 
2262 	/*  set options specified by user. */
2263 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2264 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2265 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2266 		free(opt, M_IP6OPT);
2267 		return (error);
2268 	}
2269 	*pktopt = opt;
2270 	return (0);
2271 }
2272 
2273 /*
2274  * initialize ip6_pktopts.  beware that there are non-zero default values in
2275  * the struct.
2276  */
2277 void
2278 ip6_initpktopts(struct ip6_pktopts *opt)
2279 {
2280 
2281 	bzero(opt, sizeof(*opt));
2282 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2283 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2284 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2285 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2286 }
2287 
2288 static int
2289 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2290     struct ucred *cred, int uproto)
2291 {
2292 	struct ip6_pktopts *opt;
2293 
2294 	if (*pktopt == NULL) {
2295 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2296 		    M_WAITOK);
2297 		ip6_initpktopts(*pktopt);
2298 	}
2299 	opt = *pktopt;
2300 
2301 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2302 }
2303 
2304 static int
2305 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2306 {
2307 	void *optdata = NULL;
2308 	int optdatalen = 0;
2309 	struct ip6_ext *ip6e;
2310 	int error = 0;
2311 	struct in6_pktinfo null_pktinfo;
2312 	int deftclass = 0, on;
2313 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2314 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2315 
2316 	switch (optname) {
2317 	case IPV6_PKTINFO:
2318 		if (pktopt && pktopt->ip6po_pktinfo)
2319 			optdata = (void *)pktopt->ip6po_pktinfo;
2320 		else {
2321 			/* XXX: we don't have to do this every time... */
2322 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2323 			optdata = (void *)&null_pktinfo;
2324 		}
2325 		optdatalen = sizeof(struct in6_pktinfo);
2326 		break;
2327 	case IPV6_TCLASS:
2328 		if (pktopt && pktopt->ip6po_tclass >= 0)
2329 			optdata = (void *)&pktopt->ip6po_tclass;
2330 		else
2331 			optdata = (void *)&deftclass;
2332 		optdatalen = sizeof(int);
2333 		break;
2334 	case IPV6_HOPOPTS:
2335 		if (pktopt && pktopt->ip6po_hbh) {
2336 			optdata = (void *)pktopt->ip6po_hbh;
2337 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2338 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2339 		}
2340 		break;
2341 	case IPV6_RTHDR:
2342 		if (pktopt && pktopt->ip6po_rthdr) {
2343 			optdata = (void *)pktopt->ip6po_rthdr;
2344 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2345 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2346 		}
2347 		break;
2348 	case IPV6_RTHDRDSTOPTS:
2349 		if (pktopt && pktopt->ip6po_dest1) {
2350 			optdata = (void *)pktopt->ip6po_dest1;
2351 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2352 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2353 		}
2354 		break;
2355 	case IPV6_DSTOPTS:
2356 		if (pktopt && pktopt->ip6po_dest2) {
2357 			optdata = (void *)pktopt->ip6po_dest2;
2358 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2359 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2360 		}
2361 		break;
2362 	case IPV6_NEXTHOP:
2363 		if (pktopt && pktopt->ip6po_nexthop) {
2364 			optdata = (void *)pktopt->ip6po_nexthop;
2365 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2366 		}
2367 		break;
2368 	case IPV6_USE_MIN_MTU:
2369 		if (pktopt)
2370 			optdata = (void *)&pktopt->ip6po_minmtu;
2371 		else
2372 			optdata = (void *)&defminmtu;
2373 		optdatalen = sizeof(int);
2374 		break;
2375 	case IPV6_DONTFRAG:
2376 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2377 			on = 1;
2378 		else
2379 			on = 0;
2380 		optdata = (void *)&on;
2381 		optdatalen = sizeof(on);
2382 		break;
2383 	case IPV6_PREFER_TEMPADDR:
2384 		if (pktopt)
2385 			optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2386 		else
2387 			optdata = (void *)&defpreftemp;
2388 		optdatalen = sizeof(int);
2389 		break;
2390 	default:		/* should not happen */
2391 #ifdef DIAGNOSTIC
2392 		panic("ip6_getpcbopt: unexpected option\n");
2393 #endif
2394 		return (ENOPROTOOPT);
2395 	}
2396 
2397 	error = sooptcopyout(sopt, optdata, optdatalen);
2398 
2399 	return (error);
2400 }
2401 
2402 void
2403 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2404 {
2405 	if (pktopt == NULL)
2406 		return;
2407 
2408 	if (optname == -1 || optname == IPV6_PKTINFO) {
2409 		if (pktopt->ip6po_pktinfo)
2410 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2411 		pktopt->ip6po_pktinfo = NULL;
2412 	}
2413 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2414 		pktopt->ip6po_hlim = -1;
2415 	if (optname == -1 || optname == IPV6_TCLASS)
2416 		pktopt->ip6po_tclass = -1;
2417 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2418 		if (pktopt->ip6po_nextroute.ro_rt) {
2419 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2420 			pktopt->ip6po_nextroute.ro_rt = NULL;
2421 		}
2422 		if (pktopt->ip6po_nexthop)
2423 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2424 		pktopt->ip6po_nexthop = NULL;
2425 	}
2426 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2427 		if (pktopt->ip6po_hbh)
2428 			free(pktopt->ip6po_hbh, M_IP6OPT);
2429 		pktopt->ip6po_hbh = NULL;
2430 	}
2431 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2432 		if (pktopt->ip6po_dest1)
2433 			free(pktopt->ip6po_dest1, M_IP6OPT);
2434 		pktopt->ip6po_dest1 = NULL;
2435 	}
2436 	if (optname == -1 || optname == IPV6_RTHDR) {
2437 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2438 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2439 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2440 		if (pktopt->ip6po_route.ro_rt) {
2441 			RTFREE(pktopt->ip6po_route.ro_rt);
2442 			pktopt->ip6po_route.ro_rt = NULL;
2443 		}
2444 	}
2445 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2446 		if (pktopt->ip6po_dest2)
2447 			free(pktopt->ip6po_dest2, M_IP6OPT);
2448 		pktopt->ip6po_dest2 = NULL;
2449 	}
2450 }
2451 
2452 #define PKTOPT_EXTHDRCPY(type) \
2453 do {\
2454 	if (src->type) {\
2455 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2456 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2457 		if (dst->type == NULL && canwait == M_NOWAIT)\
2458 			goto bad;\
2459 		bcopy(src->type, dst->type, hlen);\
2460 	}\
2461 } while (/*CONSTCOND*/ 0)
2462 
2463 static int
2464 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2465 {
2466 	if (dst == NULL || src == NULL)  {
2467 		printf("ip6_clearpktopts: invalid argument\n");
2468 		return (EINVAL);
2469 	}
2470 
2471 	dst->ip6po_hlim = src->ip6po_hlim;
2472 	dst->ip6po_tclass = src->ip6po_tclass;
2473 	dst->ip6po_flags = src->ip6po_flags;
2474 	dst->ip6po_minmtu = src->ip6po_minmtu;
2475 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2476 	if (src->ip6po_pktinfo) {
2477 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2478 		    M_IP6OPT, canwait);
2479 		if (dst->ip6po_pktinfo == NULL)
2480 			goto bad;
2481 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2482 	}
2483 	if (src->ip6po_nexthop) {
2484 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2485 		    M_IP6OPT, canwait);
2486 		if (dst->ip6po_nexthop == NULL)
2487 			goto bad;
2488 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2489 		    src->ip6po_nexthop->sa_len);
2490 	}
2491 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2492 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2493 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2494 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2495 	return (0);
2496 
2497   bad:
2498 	ip6_clearpktopts(dst, -1);
2499 	return (ENOBUFS);
2500 }
2501 #undef PKTOPT_EXTHDRCPY
2502 
2503 struct ip6_pktopts *
2504 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2505 {
2506 	int error;
2507 	struct ip6_pktopts *dst;
2508 
2509 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2510 	if (dst == NULL)
2511 		return (NULL);
2512 	ip6_initpktopts(dst);
2513 
2514 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2515 		free(dst, M_IP6OPT);
2516 		return (NULL);
2517 	}
2518 
2519 	return (dst);
2520 }
2521 
2522 void
2523 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2524 {
2525 	if (pktopt == NULL)
2526 		return;
2527 
2528 	ip6_clearpktopts(pktopt, -1);
2529 
2530 	free(pktopt, M_IP6OPT);
2531 }
2532 
2533 /*
2534  * Set IPv6 outgoing packet options based on advanced API.
2535  */
2536 int
2537 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2538     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2539 {
2540 	struct cmsghdr *cm = 0;
2541 
2542 	if (control == NULL || opt == NULL)
2543 		return (EINVAL);
2544 
2545 	ip6_initpktopts(opt);
2546 	if (stickyopt) {
2547 		int error;
2548 
2549 		/*
2550 		 * If stickyopt is provided, make a local copy of the options
2551 		 * for this particular packet, then override them by ancillary
2552 		 * objects.
2553 		 * XXX: copypktopts() does not copy the cached route to a next
2554 		 * hop (if any).  This is not very good in terms of efficiency,
2555 		 * but we can allow this since this option should be rarely
2556 		 * used.
2557 		 */
2558 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2559 			return (error);
2560 	}
2561 
2562 	/*
2563 	 * XXX: Currently, we assume all the optional information is stored
2564 	 * in a single mbuf.
2565 	 */
2566 	if (control->m_next)
2567 		return (EINVAL);
2568 
2569 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2570 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2571 		int error;
2572 
2573 		if (control->m_len < CMSG_LEN(0))
2574 			return (EINVAL);
2575 
2576 		cm = mtod(control, struct cmsghdr *);
2577 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2578 			return (EINVAL);
2579 		if (cm->cmsg_level != IPPROTO_IPV6)
2580 			continue;
2581 
2582 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2583 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2584 		if (error)
2585 			return (error);
2586 	}
2587 
2588 	return (0);
2589 }
2590 
2591 /*
2592  * Set a particular packet option, as a sticky option or an ancillary data
2593  * item.  "len" can be 0 only when it's a sticky option.
2594  * We have 4 cases of combination of "sticky" and "cmsg":
2595  * "sticky=0, cmsg=0": impossible
2596  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2597  * "sticky=1, cmsg=0": RFC3542 socket option
2598  * "sticky=1, cmsg=1": RFC2292 socket option
2599  */
2600 static int
2601 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2602     struct ucred *cred, int sticky, int cmsg, int uproto)
2603 {
2604 	int minmtupolicy, preftemp;
2605 	int error;
2606 
2607 	if (!sticky && !cmsg) {
2608 #ifdef DIAGNOSTIC
2609 		printf("ip6_setpktopt: impossible case\n");
2610 #endif
2611 		return (EINVAL);
2612 	}
2613 
2614 	/*
2615 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2616 	 * not be specified in the context of RFC3542.  Conversely,
2617 	 * RFC3542 types should not be specified in the context of RFC2292.
2618 	 */
2619 	if (!cmsg) {
2620 		switch (optname) {
2621 		case IPV6_2292PKTINFO:
2622 		case IPV6_2292HOPLIMIT:
2623 		case IPV6_2292NEXTHOP:
2624 		case IPV6_2292HOPOPTS:
2625 		case IPV6_2292DSTOPTS:
2626 		case IPV6_2292RTHDR:
2627 		case IPV6_2292PKTOPTIONS:
2628 			return (ENOPROTOOPT);
2629 		}
2630 	}
2631 	if (sticky && cmsg) {
2632 		switch (optname) {
2633 		case IPV6_PKTINFO:
2634 		case IPV6_HOPLIMIT:
2635 		case IPV6_NEXTHOP:
2636 		case IPV6_HOPOPTS:
2637 		case IPV6_DSTOPTS:
2638 		case IPV6_RTHDRDSTOPTS:
2639 		case IPV6_RTHDR:
2640 		case IPV6_USE_MIN_MTU:
2641 		case IPV6_DONTFRAG:
2642 		case IPV6_TCLASS:
2643 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2644 			return (ENOPROTOOPT);
2645 		}
2646 	}
2647 
2648 	switch (optname) {
2649 	case IPV6_2292PKTINFO:
2650 	case IPV6_PKTINFO:
2651 	{
2652 		struct ifnet *ifp = NULL;
2653 		struct in6_pktinfo *pktinfo;
2654 
2655 		if (len != sizeof(struct in6_pktinfo))
2656 			return (EINVAL);
2657 
2658 		pktinfo = (struct in6_pktinfo *)buf;
2659 
2660 		/*
2661 		 * An application can clear any sticky IPV6_PKTINFO option by
2662 		 * doing a "regular" setsockopt with ipi6_addr being
2663 		 * in6addr_any and ipi6_ifindex being zero.
2664 		 * [RFC 3542, Section 6]
2665 		 */
2666 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2667 		    pktinfo->ipi6_ifindex == 0 &&
2668 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2669 			ip6_clearpktopts(opt, optname);
2670 			break;
2671 		}
2672 
2673 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2674 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2675 			return (EINVAL);
2676 		}
2677 
2678 		/* validate the interface index if specified. */
2679 		if (pktinfo->ipi6_ifindex > V_if_index ||
2680 		    pktinfo->ipi6_ifindex < 0) {
2681 			 return (ENXIO);
2682 		}
2683 		if (pktinfo->ipi6_ifindex) {
2684 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2685 			if (ifp == NULL)
2686 				return (ENXIO);
2687 		}
2688 
2689 		/*
2690 		 * We store the address anyway, and let in6_selectsrc()
2691 		 * validate the specified address.  This is because ipi6_addr
2692 		 * may not have enough information about its scope zone, and
2693 		 * we may need additional information (such as outgoing
2694 		 * interface or the scope zone of a destination address) to
2695 		 * disambiguate the scope.
2696 		 * XXX: the delay of the validation may confuse the
2697 		 * application when it is used as a sticky option.
2698 		 */
2699 		if (opt->ip6po_pktinfo == NULL) {
2700 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2701 			    M_IP6OPT, M_NOWAIT);
2702 			if (opt->ip6po_pktinfo == NULL)
2703 				return (ENOBUFS);
2704 		}
2705 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2706 		break;
2707 	}
2708 
2709 	case IPV6_2292HOPLIMIT:
2710 	case IPV6_HOPLIMIT:
2711 	{
2712 		int *hlimp;
2713 
2714 		/*
2715 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2716 		 * to simplify the ordering among hoplimit options.
2717 		 */
2718 		if (optname == IPV6_HOPLIMIT && sticky)
2719 			return (ENOPROTOOPT);
2720 
2721 		if (len != sizeof(int))
2722 			return (EINVAL);
2723 		hlimp = (int *)buf;
2724 		if (*hlimp < -1 || *hlimp > 255)
2725 			return (EINVAL);
2726 
2727 		opt->ip6po_hlim = *hlimp;
2728 		break;
2729 	}
2730 
2731 	case IPV6_TCLASS:
2732 	{
2733 		int tclass;
2734 
2735 		if (len != sizeof(int))
2736 			return (EINVAL);
2737 		tclass = *(int *)buf;
2738 		if (tclass < -1 || tclass > 255)
2739 			return (EINVAL);
2740 
2741 		opt->ip6po_tclass = tclass;
2742 		break;
2743 	}
2744 
2745 	case IPV6_2292NEXTHOP:
2746 	case IPV6_NEXTHOP:
2747 		if (cred != NULL) {
2748 			error = priv_check_cred(cred,
2749 			    PRIV_NETINET_SETHDROPTS, 0);
2750 			if (error)
2751 				return (error);
2752 		}
2753 
2754 		if (len == 0) {	/* just remove the option */
2755 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2756 			break;
2757 		}
2758 
2759 		/* check if cmsg_len is large enough for sa_len */
2760 		if (len < sizeof(struct sockaddr) || len < *buf)
2761 			return (EINVAL);
2762 
2763 		switch (((struct sockaddr *)buf)->sa_family) {
2764 		case AF_INET6:
2765 		{
2766 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2767 			int error;
2768 
2769 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2770 				return (EINVAL);
2771 
2772 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2773 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2774 				return (EINVAL);
2775 			}
2776 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2777 			    != 0) {
2778 				return (error);
2779 			}
2780 			break;
2781 		}
2782 		case AF_LINK:	/* should eventually be supported */
2783 		default:
2784 			return (EAFNOSUPPORT);
2785 		}
2786 
2787 		/* turn off the previous option, then set the new option. */
2788 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2789 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2790 		if (opt->ip6po_nexthop == NULL)
2791 			return (ENOBUFS);
2792 		bcopy(buf, opt->ip6po_nexthop, *buf);
2793 		break;
2794 
2795 	case IPV6_2292HOPOPTS:
2796 	case IPV6_HOPOPTS:
2797 	{
2798 		struct ip6_hbh *hbh;
2799 		int hbhlen;
2800 
2801 		/*
2802 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2803 		 * options, since per-option restriction has too much
2804 		 * overhead.
2805 		 */
2806 		if (cred != NULL) {
2807 			error = priv_check_cred(cred,
2808 			    PRIV_NETINET_SETHDROPTS, 0);
2809 			if (error)
2810 				return (error);
2811 		}
2812 
2813 		if (len == 0) {
2814 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2815 			break;	/* just remove the option */
2816 		}
2817 
2818 		/* message length validation */
2819 		if (len < sizeof(struct ip6_hbh))
2820 			return (EINVAL);
2821 		hbh = (struct ip6_hbh *)buf;
2822 		hbhlen = (hbh->ip6h_len + 1) << 3;
2823 		if (len != hbhlen)
2824 			return (EINVAL);
2825 
2826 		/* turn off the previous option, then set the new option. */
2827 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2828 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2829 		if (opt->ip6po_hbh == NULL)
2830 			return (ENOBUFS);
2831 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
2832 
2833 		break;
2834 	}
2835 
2836 	case IPV6_2292DSTOPTS:
2837 	case IPV6_DSTOPTS:
2838 	case IPV6_RTHDRDSTOPTS:
2839 	{
2840 		struct ip6_dest *dest, **newdest = NULL;
2841 		int destlen;
2842 
2843 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2844 			error = priv_check_cred(cred,
2845 			    PRIV_NETINET_SETHDROPTS, 0);
2846 			if (error)
2847 				return (error);
2848 		}
2849 
2850 		if (len == 0) {
2851 			ip6_clearpktopts(opt, optname);
2852 			break;	/* just remove the option */
2853 		}
2854 
2855 		/* message length validation */
2856 		if (len < sizeof(struct ip6_dest))
2857 			return (EINVAL);
2858 		dest = (struct ip6_dest *)buf;
2859 		destlen = (dest->ip6d_len + 1) << 3;
2860 		if (len != destlen)
2861 			return (EINVAL);
2862 
2863 		/*
2864 		 * Determine the position that the destination options header
2865 		 * should be inserted; before or after the routing header.
2866 		 */
2867 		switch (optname) {
2868 		case IPV6_2292DSTOPTS:
2869 			/*
2870 			 * The old advacned API is ambiguous on this point.
2871 			 * Our approach is to determine the position based
2872 			 * according to the existence of a routing header.
2873 			 * Note, however, that this depends on the order of the
2874 			 * extension headers in the ancillary data; the 1st
2875 			 * part of the destination options header must appear
2876 			 * before the routing header in the ancillary data,
2877 			 * too.
2878 			 * RFC3542 solved the ambiguity by introducing
2879 			 * separate ancillary data or option types.
2880 			 */
2881 			if (opt->ip6po_rthdr == NULL)
2882 				newdest = &opt->ip6po_dest1;
2883 			else
2884 				newdest = &opt->ip6po_dest2;
2885 			break;
2886 		case IPV6_RTHDRDSTOPTS:
2887 			newdest = &opt->ip6po_dest1;
2888 			break;
2889 		case IPV6_DSTOPTS:
2890 			newdest = &opt->ip6po_dest2;
2891 			break;
2892 		}
2893 
2894 		/* turn off the previous option, then set the new option. */
2895 		ip6_clearpktopts(opt, optname);
2896 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2897 		if (*newdest == NULL)
2898 			return (ENOBUFS);
2899 		bcopy(dest, *newdest, destlen);
2900 
2901 		break;
2902 	}
2903 
2904 	case IPV6_2292RTHDR:
2905 	case IPV6_RTHDR:
2906 	{
2907 		struct ip6_rthdr *rth;
2908 		int rthlen;
2909 
2910 		if (len == 0) {
2911 			ip6_clearpktopts(opt, IPV6_RTHDR);
2912 			break;	/* just remove the option */
2913 		}
2914 
2915 		/* message length validation */
2916 		if (len < sizeof(struct ip6_rthdr))
2917 			return (EINVAL);
2918 		rth = (struct ip6_rthdr *)buf;
2919 		rthlen = (rth->ip6r_len + 1) << 3;
2920 		if (len != rthlen)
2921 			return (EINVAL);
2922 
2923 		switch (rth->ip6r_type) {
2924 		case IPV6_RTHDR_TYPE_0:
2925 			if (rth->ip6r_len == 0)	/* must contain one addr */
2926 				return (EINVAL);
2927 			if (rth->ip6r_len % 2) /* length must be even */
2928 				return (EINVAL);
2929 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2930 				return (EINVAL);
2931 			break;
2932 		default:
2933 			return (EINVAL);	/* not supported */
2934 		}
2935 
2936 		/* turn off the previous option */
2937 		ip6_clearpktopts(opt, IPV6_RTHDR);
2938 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2939 		if (opt->ip6po_rthdr == NULL)
2940 			return (ENOBUFS);
2941 		bcopy(rth, opt->ip6po_rthdr, rthlen);
2942 
2943 		break;
2944 	}
2945 
2946 	case IPV6_USE_MIN_MTU:
2947 		if (len != sizeof(int))
2948 			return (EINVAL);
2949 		minmtupolicy = *(int *)buf;
2950 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2951 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
2952 		    minmtupolicy != IP6PO_MINMTU_ALL) {
2953 			return (EINVAL);
2954 		}
2955 		opt->ip6po_minmtu = minmtupolicy;
2956 		break;
2957 
2958 	case IPV6_DONTFRAG:
2959 		if (len != sizeof(int))
2960 			return (EINVAL);
2961 
2962 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2963 			/*
2964 			 * we ignore this option for TCP sockets.
2965 			 * (RFC3542 leaves this case unspecified.)
2966 			 */
2967 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2968 		} else
2969 			opt->ip6po_flags |= IP6PO_DONTFRAG;
2970 		break;
2971 
2972 	case IPV6_PREFER_TEMPADDR:
2973 		if (len != sizeof(int))
2974 			return (EINVAL);
2975 		preftemp = *(int *)buf;
2976 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2977 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2978 		    preftemp != IP6PO_TEMPADDR_PREFER) {
2979 			return (EINVAL);
2980 		}
2981 		opt->ip6po_prefer_tempaddr = preftemp;
2982 		break;
2983 
2984 	default:
2985 		return (ENOPROTOOPT);
2986 	} /* end of switch */
2987 
2988 	return (0);
2989 }
2990 
2991 /*
2992  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2993  * packet to the input queue of a specified interface.  Note that this
2994  * calls the output routine of the loopback "driver", but with an interface
2995  * pointer that might NOT be &loif -- easier than replicating that code here.
2996  */
2997 void
2998 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
2999 {
3000 	struct mbuf *copym;
3001 	struct ip6_hdr *ip6;
3002 
3003 	copym = m_copy(m, 0, M_COPYALL);
3004 	if (copym == NULL)
3005 		return;
3006 
3007 	/*
3008 	 * Make sure to deep-copy IPv6 header portion in case the data
3009 	 * is in an mbuf cluster, so that we can safely override the IPv6
3010 	 * header portion later.
3011 	 */
3012 	if ((copym->m_flags & M_EXT) != 0 ||
3013 	    copym->m_len < sizeof(struct ip6_hdr)) {
3014 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3015 		if (copym == NULL)
3016 			return;
3017 	}
3018 
3019 #ifdef DIAGNOSTIC
3020 	if (copym->m_len < sizeof(*ip6)) {
3021 		m_freem(copym);
3022 		return;
3023 	}
3024 #endif
3025 
3026 	ip6 = mtod(copym, struct ip6_hdr *);
3027 	/*
3028 	 * clear embedded scope identifiers if necessary.
3029 	 * in6_clearscope will touch the addresses only when necessary.
3030 	 */
3031 	in6_clearscope(&ip6->ip6_src);
3032 	in6_clearscope(&ip6->ip6_dst);
3033 
3034 	(void)if_simloop(ifp, copym, dst->sin6_family, 0);
3035 }
3036 
3037 /*
3038  * Chop IPv6 header off from the payload.
3039  */
3040 static int
3041 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3042 {
3043 	struct mbuf *mh;
3044 	struct ip6_hdr *ip6;
3045 
3046 	ip6 = mtod(m, struct ip6_hdr *);
3047 	if (m->m_len > sizeof(*ip6)) {
3048 		MGETHDR(mh, M_NOWAIT, MT_HEADER);
3049 		if (mh == 0) {
3050 			m_freem(m);
3051 			return ENOBUFS;
3052 		}
3053 		M_MOVE_PKTHDR(mh, m);
3054 		MH_ALIGN(mh, sizeof(*ip6));
3055 		m->m_len -= sizeof(*ip6);
3056 		m->m_data += sizeof(*ip6);
3057 		mh->m_next = m;
3058 		m = mh;
3059 		m->m_len = sizeof(*ip6);
3060 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3061 	}
3062 	exthdrs->ip6e_ip6 = m;
3063 	return 0;
3064 }
3065 
3066 /*
3067  * Compute IPv6 extension header length.
3068  */
3069 int
3070 ip6_optlen(struct inpcb *in6p)
3071 {
3072 	int len;
3073 
3074 	if (!in6p->in6p_outputopts)
3075 		return 0;
3076 
3077 	len = 0;
3078 #define elen(x) \
3079     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3080 
3081 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3082 	if (in6p->in6p_outputopts->ip6po_rthdr)
3083 		/* dest1 is valid with rthdr only */
3084 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
3085 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3086 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3087 	return len;
3088 #undef elen
3089 }
3090