xref: /freebsd/sys/netinet6/ip6_output.c (revision 7cd2dcf07629713e5a3d60472cfe4701b705a167)
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;
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 	bzero(&dst_sa, sizeof(dst_sa));
640 	dst_sa.sin6_family = AF_INET6;
641 	dst_sa.sin6_len = sizeof(dst_sa);
642 	dst_sa.sin6_addr = ip6->ip6_dst;
643 	if (ro->ro_rt) {
644 		rt = ro->ro_rt;
645 		ifp = ro->ro_rt->rt_ifp;
646 	} else if ((error = in6_selectroute_fib(&dst_sa, opt, im6o, ro,
647 	    &ifp, &rt, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m))) != 0) {
648 		switch (error) {
649 		case EHOSTUNREACH:
650 			V_ip6stat.ip6s_noroute++;
651 			break;
652 		case EADDRNOTAVAIL:
653 		default:
654 			break; /* XXX statistics? */
655 		}
656 		if (ifp != NULL)
657 			in6_ifstat_inc(ifp, ifs6_out_discard);
658 		goto bad;
659 	}
660 	if (rt == NULL) {
661 		/*
662 		 * If in6_selectroute() does not return a route entry,
663 		 * dst may not have been updated.
664 		 */
665 		*dst = dst_sa;	/* XXX */
666 	}
667 
668 	/*
669 	 * then rt (for unicast) and ifp must be non-NULL valid values.
670 	 */
671 	if ((flags & IPV6_FORWARDING) == 0) {
672 		/* XXX: the FORWARDING flag can be set for mrouting. */
673 		in6_ifstat_inc(ifp, ifs6_out_request);
674 	}
675 	if (rt != NULL) {
676 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
677 		rt->rt_use++;
678 	}
679 
680 
681 	/*
682 	 * The outgoing interface must be in the zone of source and
683 	 * destination addresses.
684 	 */
685 	origifp = ifp;
686 
687 	src0 = ip6->ip6_src;
688 	if (in6_setscope(&src0, origifp, &zone))
689 		goto badscope;
690 	bzero(&src_sa, sizeof(src_sa));
691 	src_sa.sin6_family = AF_INET6;
692 	src_sa.sin6_len = sizeof(src_sa);
693 	src_sa.sin6_addr = ip6->ip6_src;
694 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
695 		goto badscope;
696 
697 	dst0 = ip6->ip6_dst;
698 	if (in6_setscope(&dst0, origifp, &zone))
699 		goto badscope;
700 	/* re-initialize to be sure */
701 	bzero(&dst_sa, sizeof(dst_sa));
702 	dst_sa.sin6_family = AF_INET6;
703 	dst_sa.sin6_len = sizeof(dst_sa);
704 	dst_sa.sin6_addr = ip6->ip6_dst;
705 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
706 		goto badscope;
707 	}
708 
709 	/* We should use ia_ifp to support the case of
710 	 * sending packets to an address of our own.
711 	 */
712 	if (ia != NULL && ia->ia_ifp)
713 		ifp = ia->ia_ifp;
714 
715 	/* scope check is done. */
716 	goto routefound;
717 
718   badscope:
719 	V_ip6stat.ip6s_badscope++;
720 	in6_ifstat_inc(origifp, ifs6_out_discard);
721 	if (error == 0)
722 		error = EHOSTUNREACH; /* XXX */
723 	goto bad;
724 
725   routefound:
726 	if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
727 		if (opt && opt->ip6po_nextroute.ro_rt) {
728 			/*
729 			 * The nexthop is explicitly specified by the
730 			 * application.  We assume the next hop is an IPv6
731 			 * address.
732 			 */
733 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
734 		}
735 		else if ((rt->rt_flags & RTF_GATEWAY))
736 			dst = (struct sockaddr_in6 *)rt->rt_gateway;
737 	}
738 
739 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
740 		m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
741 	} else {
742 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
743 		in6_ifstat_inc(ifp, ifs6_out_mcast);
744 		/*
745 		 * Confirm that the outgoing interface supports multicast.
746 		 */
747 		if (!(ifp->if_flags & IFF_MULTICAST)) {
748 			V_ip6stat.ip6s_noroute++;
749 			in6_ifstat_inc(ifp, ifs6_out_discard);
750 			error = ENETUNREACH;
751 			goto bad;
752 		}
753 		if ((im6o == NULL && in6_mcast_loop) ||
754 		    (im6o && im6o->im6o_multicast_loop)) {
755 			/*
756 			 * Loop back multicast datagram if not expressly
757 			 * forbidden to do so, even if we have not joined
758 			 * the address; protocols will filter it later,
759 			 * thus deferring a hash lookup and lock acquisition
760 			 * at the expense of an m_copym().
761 			 */
762 			ip6_mloopback(ifp, m, dst);
763 		} else {
764 			/*
765 			 * If we are acting as a multicast router, perform
766 			 * multicast forwarding as if the packet had just
767 			 * arrived on the interface to which we are about
768 			 * to send.  The multicast forwarding function
769 			 * recursively calls this function, using the
770 			 * IPV6_FORWARDING flag to prevent infinite recursion.
771 			 *
772 			 * Multicasts that are looped back by ip6_mloopback(),
773 			 * above, will be forwarded by the ip6_input() routine,
774 			 * if necessary.
775 			 */
776 			if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
777 				/*
778 				 * XXX: ip6_mforward expects that rcvif is NULL
779 				 * when it is called from the originating path.
780 				 * However, it is not always the case, since
781 				 * some versions of MGETHDR() does not
782 				 * initialize the field.
783 				 */
784 				m->m_pkthdr.rcvif = NULL;
785 				if (ip6_mforward(ip6, ifp, m) != 0) {
786 					m_freem(m);
787 					goto done;
788 				}
789 			}
790 		}
791 		/*
792 		 * Multicasts with a hoplimit of zero may be looped back,
793 		 * above, but must not be transmitted on a network.
794 		 * Also, multicasts addressed to the loopback interface
795 		 * are not sent -- the above call to ip6_mloopback() will
796 		 * loop back a copy if this host actually belongs to the
797 		 * destination group on the loopback interface.
798 		 */
799 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
800 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
801 			m_freem(m);
802 			goto done;
803 		}
804 	}
805 
806 	/*
807 	 * Fill the outgoing inteface to tell the upper layer
808 	 * to increment per-interface statistics.
809 	 */
810 	if (ifpp)
811 		*ifpp = ifp;
812 
813 	/* Determine path MTU. */
814 	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
815 	    &alwaysfrag, inp ? inp->inp_inc.inc_fibnum : M_GETFIB(m))) != 0)
816 		goto bad;
817 
818 	/*
819 	 * The caller of this function may specify to use the minimum MTU
820 	 * in some cases.
821 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
822 	 * setting.  The logic is a bit complicated; by default, unicast
823 	 * packets will follow path MTU while multicast packets will be sent at
824 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
825 	 * including unicast ones will be sent at the minimum MTU.  Multicast
826 	 * packets will always be sent at the minimum MTU unless
827 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
828 	 * See RFC 3542 for more details.
829 	 */
830 	if (mtu > IPV6_MMTU) {
831 		if ((flags & IPV6_MINMTU))
832 			mtu = IPV6_MMTU;
833 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
834 			mtu = IPV6_MMTU;
835 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
836 			 (opt == NULL ||
837 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
838 			mtu = IPV6_MMTU;
839 		}
840 	}
841 
842 	/*
843 	 * clear embedded scope identifiers if necessary.
844 	 * in6_clearscope will touch the addresses only when necessary.
845 	 */
846 	in6_clearscope(&ip6->ip6_src);
847 	in6_clearscope(&ip6->ip6_dst);
848 
849 	/*
850 	 * If the outgoing packet contains a hop-by-hop options header,
851 	 * it must be examined and processed even by the source node.
852 	 * (RFC 2460, section 4.)
853 	 */
854 	if (exthdrs.ip6e_hbh) {
855 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
856 		u_int32_t dummy; /* XXX unused */
857 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
858 
859 #ifdef DIAGNOSTIC
860 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
861 			panic("ip6e_hbh is not contiguous");
862 #endif
863 		/*
864 		 *  XXX: if we have to send an ICMPv6 error to the sender,
865 		 *       we need the M_LOOP flag since icmp6_error() expects
866 		 *       the IPv6 and the hop-by-hop options header are
867 		 *       contiguous unless the flag is set.
868 		 */
869 		m->m_flags |= M_LOOP;
870 		m->m_pkthdr.rcvif = ifp;
871 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
872 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
873 		    &dummy, &plen) < 0) {
874 			/* m was already freed at this point */
875 			error = EINVAL;/* better error? */
876 			goto done;
877 		}
878 		m->m_flags &= ~M_LOOP; /* XXX */
879 		m->m_pkthdr.rcvif = NULL;
880 	}
881 
882 	/* Jump over all PFIL processing if hooks are not active. */
883 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
884 		goto passout;
885 
886 	odst = ip6->ip6_dst;
887 	/* Run through list of hooks for output packets. */
888 	error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
889 	if (error != 0 || m == NULL)
890 		goto done;
891 	ip6 = mtod(m, struct ip6_hdr *);
892 
893 	/* See if destination IP address was changed by packet filter. */
894 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
895 		m->m_flags |= M_SKIP_FIREWALL;
896 		/* If destination is now ourself drop to ip6_input(). */
897 		if (in6_localip(&ip6->ip6_dst)) {
898 			m->m_flags |= M_FASTFWD_OURS;
899 			if (m->m_pkthdr.rcvif == NULL)
900 				m->m_pkthdr.rcvif = V_loif;
901 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
902 				m->m_pkthdr.csum_flags |=
903 				    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
904 				m->m_pkthdr.csum_data = 0xffff;
905 			}
906 #ifdef SCTP
907 			if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
908 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
909 #endif
910 			error = netisr_queue(NETISR_IPV6, m);
911 			goto done;
912 		} else
913 			goto again;	/* Redo the routing table lookup. */
914 	}
915 
916 	/* See if local, if yes, send it to netisr. */
917 	if (m->m_flags & M_FASTFWD_OURS) {
918 		if (m->m_pkthdr.rcvif == NULL)
919 			m->m_pkthdr.rcvif = V_loif;
920 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
921 			m->m_pkthdr.csum_flags |=
922 			    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
923 			m->m_pkthdr.csum_data = 0xffff;
924 		}
925 #ifdef SCTP
926 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
927 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
928 #endif
929 		error = netisr_queue(NETISR_IPV6, m);
930 		goto done;
931 	}
932 	/* Or forward to some other address? */
933 	if ((m->m_flags & M_IP6_NEXTHOP) &&
934 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
935 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
936 		bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in6));
937 		m->m_flags |= M_SKIP_FIREWALL;
938 		m->m_flags &= ~M_IP6_NEXTHOP;
939 		m_tag_delete(m, fwd_tag);
940 		goto again;
941 	}
942 
943 passout:
944 	/*
945 	 * Send the packet to the outgoing interface.
946 	 * If necessary, do IPv6 fragmentation before sending.
947 	 *
948 	 * the logic here is rather complex:
949 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
950 	 * 1-a:	send as is if tlen <= path mtu
951 	 * 1-b:	fragment if tlen > path mtu
952 	 *
953 	 * 2: if user asks us not to fragment (dontfrag == 1)
954 	 * 2-a:	send as is if tlen <= interface mtu
955 	 * 2-b:	error if tlen > interface mtu
956 	 *
957 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
958 	 *	always fragment
959 	 *
960 	 * 4: if dontfrag == 1 && alwaysfrag == 1
961 	 *	error, as we cannot handle this conflicting request
962 	 */
963 	sw_csum = m->m_pkthdr.csum_flags;
964 	if (!hdrsplit) {
965 		tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
966 		sw_csum &= ~ifp->if_hwassist;
967 	} else
968 		tso = 0;
969 	/*
970 	 * If we added extension headers, we will not do TSO and calculate the
971 	 * checksums ourselves for now.
972 	 * XXX-BZ  Need a framework to know when the NIC can handle it, even
973 	 * with ext. hdrs.
974 	 */
975 	if (sw_csum & CSUM_DELAY_DATA_IPV6) {
976 		sw_csum &= ~CSUM_DELAY_DATA_IPV6;
977 		in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
978 	}
979 #ifdef SCTP
980 	if (sw_csum & CSUM_SCTP_IPV6) {
981 		sw_csum &= ~CSUM_SCTP_IPV6;
982 		sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
983 	}
984 #endif
985 	m->m_pkthdr.csum_flags &= ifp->if_hwassist;
986 	tlen = m->m_pkthdr.len;
987 
988 	if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
989 		dontfrag = 1;
990 	else
991 		dontfrag = 0;
992 	if (dontfrag && alwaysfrag) {	/* case 4 */
993 		/* conflicting request - can't transmit */
994 		error = EMSGSIZE;
995 		goto bad;
996 	}
997 	if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {	/* case 2-b */
998 		/*
999 		 * Even if the DONTFRAG option is specified, we cannot send the
1000 		 * packet when the data length is larger than the MTU of the
1001 		 * outgoing interface.
1002 		 * Notify the error by sending IPV6_PATHMTU ancillary data as
1003 		 * well as returning an error code (the latter is not described
1004 		 * in the API spec.)
1005 		 */
1006 		u_int32_t mtu32;
1007 		struct ip6ctlparam ip6cp;
1008 
1009 		mtu32 = (u_int32_t)mtu;
1010 		bzero(&ip6cp, sizeof(ip6cp));
1011 		ip6cp.ip6c_cmdarg = (void *)&mtu32;
1012 		pfctlinput2(PRC_MSGSIZE, (struct sockaddr *)&ro_pmtu->ro_dst,
1013 		    (void *)&ip6cp);
1014 
1015 		error = EMSGSIZE;
1016 		goto bad;
1017 	}
1018 
1019 	/*
1020 	 * transmit packet without fragmentation
1021 	 */
1022 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
1023 		struct in6_ifaddr *ia6;
1024 
1025 		ip6 = mtod(m, struct ip6_hdr *);
1026 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
1027 		if (ia6) {
1028 			/* Record statistics for this interface address. */
1029 			ia6->ia_ifa.if_opackets++;
1030 			ia6->ia_ifa.if_obytes += m->m_pkthdr.len;
1031 			ifa_free(&ia6->ia_ifa);
1032 		}
1033 		error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1034 		goto done;
1035 	}
1036 
1037 	/*
1038 	 * try to fragment the packet.  case 1-b and 3
1039 	 */
1040 	if (mtu < IPV6_MMTU) {
1041 		/* path MTU cannot be less than IPV6_MMTU */
1042 		error = EMSGSIZE;
1043 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1044 		goto bad;
1045 	} else if (ip6->ip6_plen == 0) {
1046 		/* jumbo payload cannot be fragmented */
1047 		error = EMSGSIZE;
1048 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
1049 		goto bad;
1050 	} else {
1051 		struct mbuf **mnext, *m_frgpart;
1052 		struct ip6_frag *ip6f;
1053 		u_int32_t id = htonl(ip6_randomid());
1054 		u_char nextproto;
1055 
1056 		int qslots = ifp->if_snd.ifq_maxlen - ifp->if_snd.ifq_len;
1057 
1058 		/*
1059 		 * Too large for the destination or interface;
1060 		 * fragment if possible.
1061 		 * Must be able to put at least 8 bytes per fragment.
1062 		 */
1063 		hlen = unfragpartlen;
1064 		if (mtu > IPV6_MAXPACKET)
1065 			mtu = IPV6_MAXPACKET;
1066 
1067 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
1068 		if (len < 8) {
1069 			error = EMSGSIZE;
1070 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
1071 			goto bad;
1072 		}
1073 
1074 		/*
1075 		 * Verify that we have any chance at all of being able to queue
1076 		 *      the packet or packet fragments
1077 		 */
1078 		if (qslots <= 0 || ((u_int)qslots * (mtu - hlen)
1079 		    < tlen  /* - hlen */)) {
1080 			error = ENOBUFS;
1081 			V_ip6stat.ip6s_odropped++;
1082 			goto bad;
1083 		}
1084 
1085 
1086 		/*
1087 		 * If the interface will not calculate checksums on
1088 		 * fragmented packets, then do it here.
1089 		 * XXX-BZ handle the hw offloading case.  Need flags.
1090 		 */
1091 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1092 			in6_delayed_cksum(m, plen, hlen);
1093 			m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
1094 		}
1095 #ifdef SCTP
1096 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1097 			sctp_delayed_cksum(m, hlen);
1098 			m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1099 		}
1100 #endif
1101 		mnext = &m->m_nextpkt;
1102 
1103 		/*
1104 		 * Change the next header field of the last header in the
1105 		 * unfragmentable part.
1106 		 */
1107 		if (exthdrs.ip6e_rthdr) {
1108 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1109 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1110 		} else if (exthdrs.ip6e_dest1) {
1111 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1112 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1113 		} else if (exthdrs.ip6e_hbh) {
1114 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1115 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1116 		} else {
1117 			nextproto = ip6->ip6_nxt;
1118 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1119 		}
1120 
1121 		/*
1122 		 * Loop through length of segment after first fragment,
1123 		 * make new header and copy data of each part and link onto
1124 		 * chain.
1125 		 */
1126 		m0 = m;
1127 		for (off = hlen; off < tlen; off += len) {
1128 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
1129 			if (!m) {
1130 				error = ENOBUFS;
1131 				V_ip6stat.ip6s_odropped++;
1132 				goto sendorfree;
1133 			}
1134 			m->m_pkthdr.rcvif = NULL;
1135 			m->m_flags = m0->m_flags & M_COPYFLAGS;	/* incl. FIB */
1136 			*mnext = m;
1137 			mnext = &m->m_nextpkt;
1138 			m->m_data += max_linkhdr;
1139 			mhip6 = mtod(m, struct ip6_hdr *);
1140 			*mhip6 = *ip6;
1141 			m->m_len = sizeof(*mhip6);
1142 			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
1143 			if (error) {
1144 				V_ip6stat.ip6s_odropped++;
1145 				goto sendorfree;
1146 			}
1147 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
1148 			if (off + len >= tlen)
1149 				len = tlen - off;
1150 			else
1151 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
1152 			mhip6->ip6_plen = htons((u_short)(len + hlen +
1153 			    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
1154 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
1155 				error = ENOBUFS;
1156 				V_ip6stat.ip6s_odropped++;
1157 				goto sendorfree;
1158 			}
1159 			m_cat(m, m_frgpart);
1160 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
1161 			m->m_pkthdr.rcvif = NULL;
1162 			ip6f->ip6f_reserved = 0;
1163 			ip6f->ip6f_ident = id;
1164 			ip6f->ip6f_nxt = nextproto;
1165 			V_ip6stat.ip6s_ofragments++;
1166 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
1167 		}
1168 
1169 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1170 	}
1171 
1172 	/*
1173 	 * Remove leading garbages.
1174 	 */
1175 sendorfree:
1176 	m = m0->m_nextpkt;
1177 	m0->m_nextpkt = 0;
1178 	m_freem(m0);
1179 	for (m0 = m; m; m = m0) {
1180 		m0 = m->m_nextpkt;
1181 		m->m_nextpkt = 0;
1182 		if (error == 0) {
1183 			/* Record statistics for this interface address. */
1184 			if (ia) {
1185 				ia->ia_ifa.if_opackets++;
1186 				ia->ia_ifa.if_obytes += m->m_pkthdr.len;
1187 			}
1188 			error = nd6_output(ifp, origifp, m, dst, ro->ro_rt);
1189 		} else
1190 			m_freem(m);
1191 	}
1192 
1193 	if (error == 0)
1194 		V_ip6stat.ip6s_fragmented++;
1195 
1196 done:
1197 	if (ro == &ip6route)
1198 		RO_RTFREE(ro);
1199 	if (ro_pmtu == &ip6route)
1200 		RO_RTFREE(ro_pmtu);
1201 #ifdef IPSEC
1202 	if (sp != NULL)
1203 		KEY_FREESP(&sp);
1204 #endif
1205 
1206 	return (error);
1207 
1208 freehdrs:
1209 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1210 	m_freem(exthdrs.ip6e_dest1);
1211 	m_freem(exthdrs.ip6e_rthdr);
1212 	m_freem(exthdrs.ip6e_dest2);
1213 	/* FALLTHROUGH */
1214 bad:
1215 	if (m)
1216 		m_freem(m);
1217 	goto done;
1218 }
1219 
1220 static int
1221 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1222 {
1223 	struct mbuf *m;
1224 
1225 	if (hlen > MCLBYTES)
1226 		return (ENOBUFS); /* XXX */
1227 
1228 	MGET(m, M_DONTWAIT, MT_DATA);
1229 	if (!m)
1230 		return (ENOBUFS);
1231 
1232 	if (hlen > MLEN) {
1233 		MCLGET(m, M_DONTWAIT);
1234 		if ((m->m_flags & M_EXT) == 0) {
1235 			m_free(m);
1236 			return (ENOBUFS);
1237 		}
1238 	}
1239 	m->m_len = hlen;
1240 	if (hdr)
1241 		bcopy(hdr, mtod(m, caddr_t), hlen);
1242 
1243 	*mp = m;
1244 	return (0);
1245 }
1246 
1247 /*
1248  * Insert jumbo payload option.
1249  */
1250 static int
1251 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1252 {
1253 	struct mbuf *mopt;
1254 	u_char *optbuf;
1255 	u_int32_t v;
1256 
1257 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1258 
1259 	/*
1260 	 * If there is no hop-by-hop options header, allocate new one.
1261 	 * If there is one but it doesn't have enough space to store the
1262 	 * jumbo payload option, allocate a cluster to store the whole options.
1263 	 * Otherwise, use it to store the options.
1264 	 */
1265 	if (exthdrs->ip6e_hbh == 0) {
1266 		MGET(mopt, M_DONTWAIT, MT_DATA);
1267 		if (mopt == 0)
1268 			return (ENOBUFS);
1269 		mopt->m_len = JUMBOOPTLEN;
1270 		optbuf = mtod(mopt, u_char *);
1271 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1272 		exthdrs->ip6e_hbh = mopt;
1273 	} else {
1274 		struct ip6_hbh *hbh;
1275 
1276 		mopt = exthdrs->ip6e_hbh;
1277 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1278 			/*
1279 			 * XXX assumption:
1280 			 * - exthdrs->ip6e_hbh is not referenced from places
1281 			 *   other than exthdrs.
1282 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1283 			 */
1284 			int oldoptlen = mopt->m_len;
1285 			struct mbuf *n;
1286 
1287 			/*
1288 			 * XXX: give up if the whole (new) hbh header does
1289 			 * not fit even in an mbuf cluster.
1290 			 */
1291 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1292 				return (ENOBUFS);
1293 
1294 			/*
1295 			 * As a consequence, we must always prepare a cluster
1296 			 * at this point.
1297 			 */
1298 			MGET(n, M_DONTWAIT, MT_DATA);
1299 			if (n) {
1300 				MCLGET(n, M_DONTWAIT);
1301 				if ((n->m_flags & M_EXT) == 0) {
1302 					m_freem(n);
1303 					n = NULL;
1304 				}
1305 			}
1306 			if (!n)
1307 				return (ENOBUFS);
1308 			n->m_len = oldoptlen + JUMBOOPTLEN;
1309 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1310 			    oldoptlen);
1311 			optbuf = mtod(n, caddr_t) + oldoptlen;
1312 			m_freem(mopt);
1313 			mopt = exthdrs->ip6e_hbh = n;
1314 		} else {
1315 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1316 			mopt->m_len += JUMBOOPTLEN;
1317 		}
1318 		optbuf[0] = IP6OPT_PADN;
1319 		optbuf[1] = 1;
1320 
1321 		/*
1322 		 * Adjust the header length according to the pad and
1323 		 * the jumbo payload option.
1324 		 */
1325 		hbh = mtod(mopt, struct ip6_hbh *);
1326 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1327 	}
1328 
1329 	/* fill in the option. */
1330 	optbuf[2] = IP6OPT_JUMBO;
1331 	optbuf[3] = 4;
1332 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1333 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1334 
1335 	/* finally, adjust the packet header length */
1336 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1337 
1338 	return (0);
1339 #undef JUMBOOPTLEN
1340 }
1341 
1342 /*
1343  * Insert fragment header and copy unfragmentable header portions.
1344  */
1345 static int
1346 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1347     struct ip6_frag **frghdrp)
1348 {
1349 	struct mbuf *n, *mlast;
1350 
1351 	if (hlen > sizeof(struct ip6_hdr)) {
1352 		n = m_copym(m0, sizeof(struct ip6_hdr),
1353 		    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1354 		if (n == 0)
1355 			return (ENOBUFS);
1356 		m->m_next = n;
1357 	} else
1358 		n = m;
1359 
1360 	/* Search for the last mbuf of unfragmentable part. */
1361 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1362 		;
1363 
1364 	if ((mlast->m_flags & M_EXT) == 0 &&
1365 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1366 		/* use the trailing space of the last mbuf for the fragment hdr */
1367 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1368 		    mlast->m_len);
1369 		mlast->m_len += sizeof(struct ip6_frag);
1370 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1371 	} else {
1372 		/* allocate a new mbuf for the fragment header */
1373 		struct mbuf *mfrg;
1374 
1375 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1376 		if (mfrg == 0)
1377 			return (ENOBUFS);
1378 		mfrg->m_len = sizeof(struct ip6_frag);
1379 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1380 		mlast->m_next = mfrg;
1381 	}
1382 
1383 	return (0);
1384 }
1385 
1386 static int
1387 ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
1388     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
1389     int *alwaysfragp, u_int fibnum)
1390 {
1391 	u_int32_t mtu = 0;
1392 	int alwaysfrag = 0;
1393 	int error = 0;
1394 
1395 	if (ro_pmtu != ro) {
1396 		/* The first hop and the final destination may differ. */
1397 		struct sockaddr_in6 *sa6_dst =
1398 		    (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1399 		if (ro_pmtu->ro_rt &&
1400 		    ((ro_pmtu->ro_rt->rt_flags & RTF_UP) == 0 ||
1401 		     !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
1402 			RTFREE(ro_pmtu->ro_rt);
1403 			ro_pmtu->ro_rt = (struct rtentry *)NULL;
1404 		}
1405 		if (ro_pmtu->ro_rt == NULL) {
1406 			bzero(sa6_dst, sizeof(*sa6_dst));
1407 			sa6_dst->sin6_family = AF_INET6;
1408 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1409 			sa6_dst->sin6_addr = *dst;
1410 
1411 			in6_rtalloc(ro_pmtu, fibnum);
1412 		}
1413 	}
1414 	if (ro_pmtu->ro_rt) {
1415 		u_int32_t ifmtu;
1416 		struct in_conninfo inc;
1417 
1418 		bzero(&inc, sizeof(inc));
1419 		inc.inc_flags |= INC_ISIPV6;
1420 		inc.inc6_faddr = *dst;
1421 
1422 		if (ifp == NULL)
1423 			ifp = ro_pmtu->ro_rt->rt_ifp;
1424 		ifmtu = IN6_LINKMTU(ifp);
1425 		mtu = tcp_hc_getmtu(&inc);
1426 		if (mtu)
1427 			mtu = min(mtu, ro_pmtu->ro_rt->rt_rmx.rmx_mtu);
1428 		else
1429 			mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
1430 		if (mtu == 0)
1431 			mtu = ifmtu;
1432 		else if (mtu < IPV6_MMTU) {
1433 			/*
1434 			 * RFC2460 section 5, last paragraph:
1435 			 * if we record ICMPv6 too big message with
1436 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1437 			 * or smaller, with framgent header attached.
1438 			 * (fragment header is needed regardless from the
1439 			 * packet size, for translators to identify packets)
1440 			 */
1441 			alwaysfrag = 1;
1442 			mtu = IPV6_MMTU;
1443 		} else if (mtu > ifmtu) {
1444 			/*
1445 			 * The MTU on the route is larger than the MTU on
1446 			 * the interface!  This shouldn't happen, unless the
1447 			 * MTU of the interface has been changed after the
1448 			 * interface was brought up.  Change the MTU in the
1449 			 * route to match the interface MTU (as long as the
1450 			 * field isn't locked).
1451 			 */
1452 			mtu = ifmtu;
1453 			ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
1454 		}
1455 	} else if (ifp) {
1456 		mtu = IN6_LINKMTU(ifp);
1457 	} else
1458 		error = EHOSTUNREACH; /* XXX */
1459 
1460 	*mtup = mtu;
1461 	if (alwaysfragp)
1462 		*alwaysfragp = alwaysfrag;
1463 	return (error);
1464 }
1465 
1466 /*
1467  * IP6 socket option processing.
1468  */
1469 int
1470 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1471 {
1472 	int optdatalen, uproto;
1473 	void *optdata;
1474 	struct inpcb *in6p = sotoinpcb(so);
1475 	int error, optval;
1476 	int level, op, optname;
1477 	int optlen;
1478 	struct thread *td;
1479 
1480 	level = sopt->sopt_level;
1481 	op = sopt->sopt_dir;
1482 	optname = sopt->sopt_name;
1483 	optlen = sopt->sopt_valsize;
1484 	td = sopt->sopt_td;
1485 	error = 0;
1486 	optval = 0;
1487 	uproto = (int)so->so_proto->pr_protocol;
1488 
1489 	if (level != IPPROTO_IPV6) {
1490 		error = EINVAL;
1491 
1492 		if (sopt->sopt_level == SOL_SOCKET &&
1493 		    sopt->sopt_dir == SOPT_SET) {
1494 			switch (sopt->sopt_name) {
1495 			case SO_REUSEADDR:
1496 				INP_WLOCK(in6p);
1497 				if (IN_MULTICAST(ntohl(in6p->inp_laddr.s_addr))) {
1498 					if ((so->so_options &
1499 					    (SO_REUSEADDR | SO_REUSEPORT)) != 0)
1500 						in6p->inp_flags2 |= INP_REUSEPORT;
1501 					else
1502 						in6p->inp_flags2 &= ~INP_REUSEPORT;
1503 				}
1504 				INP_WUNLOCK(in6p);
1505 				error = 0;
1506 				break;
1507 			case SO_REUSEPORT:
1508 				INP_WLOCK(in6p);
1509 				if ((so->so_options & SO_REUSEPORT) != 0)
1510 					in6p->inp_flags2 |= INP_REUSEPORT;
1511 				else
1512 					in6p->inp_flags2 &= ~INP_REUSEPORT;
1513 				INP_WUNLOCK(in6p);
1514 				error = 0;
1515 				break;
1516 			case SO_SETFIB:
1517 				INP_WLOCK(in6p);
1518 				in6p->inp_inc.inc_fibnum = so->so_fibnum;
1519 				INP_WUNLOCK(in6p);
1520 				error = 0;
1521 				break;
1522 			default:
1523 				break;
1524 			}
1525 		}
1526 	} else {		/* level == IPPROTO_IPV6 */
1527 		switch (op) {
1528 
1529 		case SOPT_SET:
1530 			switch (optname) {
1531 			case IPV6_2292PKTOPTIONS:
1532 #ifdef IPV6_PKTOPTIONS
1533 			case IPV6_PKTOPTIONS:
1534 #endif
1535 			{
1536 				struct mbuf *m;
1537 
1538 				error = soopt_getm(sopt, &m); /* XXX */
1539 				if (error != 0)
1540 					break;
1541 				error = soopt_mcopyin(sopt, m); /* XXX */
1542 				if (error != 0)
1543 					break;
1544 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1545 						    m, so, sopt);
1546 				m_freem(m); /* XXX */
1547 				break;
1548 			}
1549 
1550 			/*
1551 			 * Use of some Hop-by-Hop options or some
1552 			 * Destination options, might require special
1553 			 * privilege.  That is, normal applications
1554 			 * (without special privilege) might be forbidden
1555 			 * from setting certain options in outgoing packets,
1556 			 * and might never see certain options in received
1557 			 * packets. [RFC 2292 Section 6]
1558 			 * KAME specific note:
1559 			 *  KAME prevents non-privileged users from sending or
1560 			 *  receiving ANY hbh/dst options in order to avoid
1561 			 *  overhead of parsing options in the kernel.
1562 			 */
1563 			case IPV6_RECVHOPOPTS:
1564 			case IPV6_RECVDSTOPTS:
1565 			case IPV6_RECVRTHDRDSTOPTS:
1566 				if (td != NULL) {
1567 					error = priv_check(td,
1568 					    PRIV_NETINET_SETHDROPTS);
1569 					if (error)
1570 						break;
1571 				}
1572 				/* FALLTHROUGH */
1573 			case IPV6_UNICAST_HOPS:
1574 			case IPV6_HOPLIMIT:
1575 			case IPV6_FAITH:
1576 
1577 			case IPV6_RECVPKTINFO:
1578 			case IPV6_RECVHOPLIMIT:
1579 			case IPV6_RECVRTHDR:
1580 			case IPV6_RECVPATHMTU:
1581 			case IPV6_RECVTCLASS:
1582 			case IPV6_V6ONLY:
1583 			case IPV6_AUTOFLOWLABEL:
1584 			case IPV6_BINDANY:
1585 				if (optname == IPV6_BINDANY && td != NULL) {
1586 					error = priv_check(td,
1587 					    PRIV_NETINET_BINDANY);
1588 					if (error)
1589 						break;
1590 				}
1591 
1592 				if (optlen != sizeof(int)) {
1593 					error = EINVAL;
1594 					break;
1595 				}
1596 				error = sooptcopyin(sopt, &optval,
1597 					sizeof optval, sizeof optval);
1598 				if (error)
1599 					break;
1600 				switch (optname) {
1601 
1602 				case IPV6_UNICAST_HOPS:
1603 					if (optval < -1 || optval >= 256)
1604 						error = EINVAL;
1605 					else {
1606 						/* -1 = kernel default */
1607 						in6p->in6p_hops = optval;
1608 						if ((in6p->inp_vflag &
1609 						     INP_IPV4) != 0)
1610 							in6p->inp_ip_ttl = optval;
1611 					}
1612 					break;
1613 #define OPTSET(bit) \
1614 do { \
1615 	INP_WLOCK(in6p); \
1616 	if (optval) \
1617 		in6p->inp_flags |= (bit); \
1618 	else \
1619 		in6p->inp_flags &= ~(bit); \
1620 	INP_WUNLOCK(in6p); \
1621 } while (/*CONSTCOND*/ 0)
1622 #define OPTSET2292(bit) \
1623 do { \
1624 	INP_WLOCK(in6p); \
1625 	in6p->inp_flags |= IN6P_RFC2292; \
1626 	if (optval) \
1627 		in6p->inp_flags |= (bit); \
1628 	else \
1629 		in6p->inp_flags &= ~(bit); \
1630 	INP_WUNLOCK(in6p); \
1631 } while (/*CONSTCOND*/ 0)
1632 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1633 
1634 				case IPV6_RECVPKTINFO:
1635 					/* cannot mix with RFC2292 */
1636 					if (OPTBIT(IN6P_RFC2292)) {
1637 						error = EINVAL;
1638 						break;
1639 					}
1640 					OPTSET(IN6P_PKTINFO);
1641 					break;
1642 
1643 				case IPV6_HOPLIMIT:
1644 				{
1645 					struct ip6_pktopts **optp;
1646 
1647 					/* cannot mix with RFC2292 */
1648 					if (OPTBIT(IN6P_RFC2292)) {
1649 						error = EINVAL;
1650 						break;
1651 					}
1652 					optp = &in6p->in6p_outputopts;
1653 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1654 					    (u_char *)&optval, sizeof(optval),
1655 					    optp, (td != NULL) ? td->td_ucred :
1656 					    NULL, uproto);
1657 					break;
1658 				}
1659 
1660 				case IPV6_RECVHOPLIMIT:
1661 					/* cannot mix with RFC2292 */
1662 					if (OPTBIT(IN6P_RFC2292)) {
1663 						error = EINVAL;
1664 						break;
1665 					}
1666 					OPTSET(IN6P_HOPLIMIT);
1667 					break;
1668 
1669 				case IPV6_RECVHOPOPTS:
1670 					/* cannot mix with RFC2292 */
1671 					if (OPTBIT(IN6P_RFC2292)) {
1672 						error = EINVAL;
1673 						break;
1674 					}
1675 					OPTSET(IN6P_HOPOPTS);
1676 					break;
1677 
1678 				case IPV6_RECVDSTOPTS:
1679 					/* cannot mix with RFC2292 */
1680 					if (OPTBIT(IN6P_RFC2292)) {
1681 						error = EINVAL;
1682 						break;
1683 					}
1684 					OPTSET(IN6P_DSTOPTS);
1685 					break;
1686 
1687 				case IPV6_RECVRTHDRDSTOPTS:
1688 					/* cannot mix with RFC2292 */
1689 					if (OPTBIT(IN6P_RFC2292)) {
1690 						error = EINVAL;
1691 						break;
1692 					}
1693 					OPTSET(IN6P_RTHDRDSTOPTS);
1694 					break;
1695 
1696 				case IPV6_RECVRTHDR:
1697 					/* cannot mix with RFC2292 */
1698 					if (OPTBIT(IN6P_RFC2292)) {
1699 						error = EINVAL;
1700 						break;
1701 					}
1702 					OPTSET(IN6P_RTHDR);
1703 					break;
1704 
1705 				case IPV6_FAITH:
1706 					OPTSET(INP_FAITH);
1707 					break;
1708 
1709 				case IPV6_RECVPATHMTU:
1710 					/*
1711 					 * We ignore this option for TCP
1712 					 * sockets.
1713 					 * (RFC3542 leaves this case
1714 					 * unspecified.)
1715 					 */
1716 					if (uproto != IPPROTO_TCP)
1717 						OPTSET(IN6P_MTU);
1718 					break;
1719 
1720 				case IPV6_V6ONLY:
1721 					/*
1722 					 * make setsockopt(IPV6_V6ONLY)
1723 					 * available only prior to bind(2).
1724 					 * see ipng mailing list, Jun 22 2001.
1725 					 */
1726 					if (in6p->inp_lport ||
1727 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1728 						error = EINVAL;
1729 						break;
1730 					}
1731 					OPTSET(IN6P_IPV6_V6ONLY);
1732 					if (optval)
1733 						in6p->inp_vflag &= ~INP_IPV4;
1734 					else
1735 						in6p->inp_vflag |= INP_IPV4;
1736 					break;
1737 				case IPV6_RECVTCLASS:
1738 					/* cannot mix with RFC2292 XXX */
1739 					if (OPTBIT(IN6P_RFC2292)) {
1740 						error = EINVAL;
1741 						break;
1742 					}
1743 					OPTSET(IN6P_TCLASS);
1744 					break;
1745 				case IPV6_AUTOFLOWLABEL:
1746 					OPTSET(IN6P_AUTOFLOWLABEL);
1747 					break;
1748 
1749 				case IPV6_BINDANY:
1750 					OPTSET(INP_BINDANY);
1751 					break;
1752 				}
1753 				break;
1754 
1755 			case IPV6_TCLASS:
1756 			case IPV6_DONTFRAG:
1757 			case IPV6_USE_MIN_MTU:
1758 			case IPV6_PREFER_TEMPADDR:
1759 				if (optlen != sizeof(optval)) {
1760 					error = EINVAL;
1761 					break;
1762 				}
1763 				error = sooptcopyin(sopt, &optval,
1764 					sizeof optval, sizeof optval);
1765 				if (error)
1766 					break;
1767 				{
1768 					struct ip6_pktopts **optp;
1769 					optp = &in6p->in6p_outputopts;
1770 					error = ip6_pcbopt(optname,
1771 					    (u_char *)&optval, sizeof(optval),
1772 					    optp, (td != NULL) ? td->td_ucred :
1773 					    NULL, uproto);
1774 					break;
1775 				}
1776 
1777 			case IPV6_2292PKTINFO:
1778 			case IPV6_2292HOPLIMIT:
1779 			case IPV6_2292HOPOPTS:
1780 			case IPV6_2292DSTOPTS:
1781 			case IPV6_2292RTHDR:
1782 				/* RFC 2292 */
1783 				if (optlen != sizeof(int)) {
1784 					error = EINVAL;
1785 					break;
1786 				}
1787 				error = sooptcopyin(sopt, &optval,
1788 					sizeof optval, sizeof optval);
1789 				if (error)
1790 					break;
1791 				switch (optname) {
1792 				case IPV6_2292PKTINFO:
1793 					OPTSET2292(IN6P_PKTINFO);
1794 					break;
1795 				case IPV6_2292HOPLIMIT:
1796 					OPTSET2292(IN6P_HOPLIMIT);
1797 					break;
1798 				case IPV6_2292HOPOPTS:
1799 					/*
1800 					 * Check super-user privilege.
1801 					 * See comments for IPV6_RECVHOPOPTS.
1802 					 */
1803 					if (td != NULL) {
1804 						error = priv_check(td,
1805 						    PRIV_NETINET_SETHDROPTS);
1806 						if (error)
1807 							return (error);
1808 					}
1809 					OPTSET2292(IN6P_HOPOPTS);
1810 					break;
1811 				case IPV6_2292DSTOPTS:
1812 					if (td != NULL) {
1813 						error = priv_check(td,
1814 						    PRIV_NETINET_SETHDROPTS);
1815 						if (error)
1816 							return (error);
1817 					}
1818 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1819 					break;
1820 				case IPV6_2292RTHDR:
1821 					OPTSET2292(IN6P_RTHDR);
1822 					break;
1823 				}
1824 				break;
1825 			case IPV6_PKTINFO:
1826 			case IPV6_HOPOPTS:
1827 			case IPV6_RTHDR:
1828 			case IPV6_DSTOPTS:
1829 			case IPV6_RTHDRDSTOPTS:
1830 			case IPV6_NEXTHOP:
1831 			{
1832 				/* new advanced API (RFC3542) */
1833 				u_char *optbuf;
1834 				u_char optbuf_storage[MCLBYTES];
1835 				int optlen;
1836 				struct ip6_pktopts **optp;
1837 
1838 				/* cannot mix with RFC2292 */
1839 				if (OPTBIT(IN6P_RFC2292)) {
1840 					error = EINVAL;
1841 					break;
1842 				}
1843 
1844 				/*
1845 				 * We only ensure valsize is not too large
1846 				 * here.  Further validation will be done
1847 				 * later.
1848 				 */
1849 				error = sooptcopyin(sopt, optbuf_storage,
1850 				    sizeof(optbuf_storage), 0);
1851 				if (error)
1852 					break;
1853 				optlen = sopt->sopt_valsize;
1854 				optbuf = optbuf_storage;
1855 				optp = &in6p->in6p_outputopts;
1856 				error = ip6_pcbopt(optname, optbuf, optlen,
1857 				    optp, (td != NULL) ? td->td_ucred : NULL,
1858 				    uproto);
1859 				break;
1860 			}
1861 #undef OPTSET
1862 
1863 			case IPV6_MULTICAST_IF:
1864 			case IPV6_MULTICAST_HOPS:
1865 			case IPV6_MULTICAST_LOOP:
1866 			case IPV6_JOIN_GROUP:
1867 			case IPV6_LEAVE_GROUP:
1868 			case IPV6_MSFILTER:
1869 			case MCAST_BLOCK_SOURCE:
1870 			case MCAST_UNBLOCK_SOURCE:
1871 			case MCAST_JOIN_GROUP:
1872 			case MCAST_LEAVE_GROUP:
1873 			case MCAST_JOIN_SOURCE_GROUP:
1874 			case MCAST_LEAVE_SOURCE_GROUP:
1875 				error = ip6_setmoptions(in6p, sopt);
1876 				break;
1877 
1878 			case IPV6_PORTRANGE:
1879 				error = sooptcopyin(sopt, &optval,
1880 				    sizeof optval, sizeof optval);
1881 				if (error)
1882 					break;
1883 
1884 				INP_WLOCK(in6p);
1885 				switch (optval) {
1886 				case IPV6_PORTRANGE_DEFAULT:
1887 					in6p->inp_flags &= ~(INP_LOWPORT);
1888 					in6p->inp_flags &= ~(INP_HIGHPORT);
1889 					break;
1890 
1891 				case IPV6_PORTRANGE_HIGH:
1892 					in6p->inp_flags &= ~(INP_LOWPORT);
1893 					in6p->inp_flags |= INP_HIGHPORT;
1894 					break;
1895 
1896 				case IPV6_PORTRANGE_LOW:
1897 					in6p->inp_flags &= ~(INP_HIGHPORT);
1898 					in6p->inp_flags |= INP_LOWPORT;
1899 					break;
1900 
1901 				default:
1902 					error = EINVAL;
1903 					break;
1904 				}
1905 				INP_WUNLOCK(in6p);
1906 				break;
1907 
1908 #ifdef IPSEC
1909 			case IPV6_IPSEC_POLICY:
1910 			{
1911 				caddr_t req;
1912 				struct mbuf *m;
1913 
1914 				if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1915 					break;
1916 				if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1917 					break;
1918 				req = mtod(m, caddr_t);
1919 				error = ipsec_set_policy(in6p, optname, req,
1920 				    m->m_len, (sopt->sopt_td != NULL) ?
1921 				    sopt->sopt_td->td_ucred : NULL);
1922 				m_freem(m);
1923 				break;
1924 			}
1925 #endif /* IPSEC */
1926 
1927 			default:
1928 				error = ENOPROTOOPT;
1929 				break;
1930 			}
1931 			break;
1932 
1933 		case SOPT_GET:
1934 			switch (optname) {
1935 
1936 			case IPV6_2292PKTOPTIONS:
1937 #ifdef IPV6_PKTOPTIONS
1938 			case IPV6_PKTOPTIONS:
1939 #endif
1940 				/*
1941 				 * RFC3542 (effectively) deprecated the
1942 				 * semantics of the 2292-style pktoptions.
1943 				 * Since it was not reliable in nature (i.e.,
1944 				 * applications had to expect the lack of some
1945 				 * information after all), it would make sense
1946 				 * to simplify this part by always returning
1947 				 * empty data.
1948 				 */
1949 				sopt->sopt_valsize = 0;
1950 				break;
1951 
1952 			case IPV6_RECVHOPOPTS:
1953 			case IPV6_RECVDSTOPTS:
1954 			case IPV6_RECVRTHDRDSTOPTS:
1955 			case IPV6_UNICAST_HOPS:
1956 			case IPV6_RECVPKTINFO:
1957 			case IPV6_RECVHOPLIMIT:
1958 			case IPV6_RECVRTHDR:
1959 			case IPV6_RECVPATHMTU:
1960 
1961 			case IPV6_FAITH:
1962 			case IPV6_V6ONLY:
1963 			case IPV6_PORTRANGE:
1964 			case IPV6_RECVTCLASS:
1965 			case IPV6_AUTOFLOWLABEL:
1966 			case IPV6_BINDANY:
1967 				switch (optname) {
1968 
1969 				case IPV6_RECVHOPOPTS:
1970 					optval = OPTBIT(IN6P_HOPOPTS);
1971 					break;
1972 
1973 				case IPV6_RECVDSTOPTS:
1974 					optval = OPTBIT(IN6P_DSTOPTS);
1975 					break;
1976 
1977 				case IPV6_RECVRTHDRDSTOPTS:
1978 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1979 					break;
1980 
1981 				case IPV6_UNICAST_HOPS:
1982 					optval = in6p->in6p_hops;
1983 					break;
1984 
1985 				case IPV6_RECVPKTINFO:
1986 					optval = OPTBIT(IN6P_PKTINFO);
1987 					break;
1988 
1989 				case IPV6_RECVHOPLIMIT:
1990 					optval = OPTBIT(IN6P_HOPLIMIT);
1991 					break;
1992 
1993 				case IPV6_RECVRTHDR:
1994 					optval = OPTBIT(IN6P_RTHDR);
1995 					break;
1996 
1997 				case IPV6_RECVPATHMTU:
1998 					optval = OPTBIT(IN6P_MTU);
1999 					break;
2000 
2001 				case IPV6_FAITH:
2002 					optval = OPTBIT(INP_FAITH);
2003 					break;
2004 
2005 				case IPV6_V6ONLY:
2006 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
2007 					break;
2008 
2009 				case IPV6_PORTRANGE:
2010 				    {
2011 					int flags;
2012 					flags = in6p->inp_flags;
2013 					if (flags & INP_HIGHPORT)
2014 						optval = IPV6_PORTRANGE_HIGH;
2015 					else if (flags & INP_LOWPORT)
2016 						optval = IPV6_PORTRANGE_LOW;
2017 					else
2018 						optval = 0;
2019 					break;
2020 				    }
2021 				case IPV6_RECVTCLASS:
2022 					optval = OPTBIT(IN6P_TCLASS);
2023 					break;
2024 
2025 				case IPV6_AUTOFLOWLABEL:
2026 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2027 					break;
2028 
2029 				case IPV6_BINDANY:
2030 					optval = OPTBIT(INP_BINDANY);
2031 					break;
2032 				}
2033 				if (error)
2034 					break;
2035 				error = sooptcopyout(sopt, &optval,
2036 					sizeof optval);
2037 				break;
2038 
2039 			case IPV6_PATHMTU:
2040 			{
2041 				u_long pmtu = 0;
2042 				struct ip6_mtuinfo mtuinfo;
2043 				struct route_in6 sro;
2044 
2045 				bzero(&sro, sizeof(sro));
2046 
2047 				if (!(so->so_state & SS_ISCONNECTED))
2048 					return (ENOTCONN);
2049 				/*
2050 				 * XXX: we dot not consider the case of source
2051 				 * routing, or optional information to specify
2052 				 * the outgoing interface.
2053 				 */
2054 				error = ip6_getpmtu(&sro, NULL, NULL,
2055 				    &in6p->in6p_faddr, &pmtu, NULL,
2056 				    so->so_fibnum);
2057 				if (sro.ro_rt)
2058 					RTFREE(sro.ro_rt);
2059 				if (error)
2060 					break;
2061 				if (pmtu > IPV6_MAXPACKET)
2062 					pmtu = IPV6_MAXPACKET;
2063 
2064 				bzero(&mtuinfo, sizeof(mtuinfo));
2065 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2066 				optdata = (void *)&mtuinfo;
2067 				optdatalen = sizeof(mtuinfo);
2068 				error = sooptcopyout(sopt, optdata,
2069 				    optdatalen);
2070 				break;
2071 			}
2072 
2073 			case IPV6_2292PKTINFO:
2074 			case IPV6_2292HOPLIMIT:
2075 			case IPV6_2292HOPOPTS:
2076 			case IPV6_2292RTHDR:
2077 			case IPV6_2292DSTOPTS:
2078 				switch (optname) {
2079 				case IPV6_2292PKTINFO:
2080 					optval = OPTBIT(IN6P_PKTINFO);
2081 					break;
2082 				case IPV6_2292HOPLIMIT:
2083 					optval = OPTBIT(IN6P_HOPLIMIT);
2084 					break;
2085 				case IPV6_2292HOPOPTS:
2086 					optval = OPTBIT(IN6P_HOPOPTS);
2087 					break;
2088 				case IPV6_2292RTHDR:
2089 					optval = OPTBIT(IN6P_RTHDR);
2090 					break;
2091 				case IPV6_2292DSTOPTS:
2092 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2093 					break;
2094 				}
2095 				error = sooptcopyout(sopt, &optval,
2096 				    sizeof optval);
2097 				break;
2098 			case IPV6_PKTINFO:
2099 			case IPV6_HOPOPTS:
2100 			case IPV6_RTHDR:
2101 			case IPV6_DSTOPTS:
2102 			case IPV6_RTHDRDSTOPTS:
2103 			case IPV6_NEXTHOP:
2104 			case IPV6_TCLASS:
2105 			case IPV6_DONTFRAG:
2106 			case IPV6_USE_MIN_MTU:
2107 			case IPV6_PREFER_TEMPADDR:
2108 				error = ip6_getpcbopt(in6p->in6p_outputopts,
2109 				    optname, sopt);
2110 				break;
2111 
2112 			case IPV6_MULTICAST_IF:
2113 			case IPV6_MULTICAST_HOPS:
2114 			case IPV6_MULTICAST_LOOP:
2115 			case IPV6_MSFILTER:
2116 				error = ip6_getmoptions(in6p, sopt);
2117 				break;
2118 
2119 #ifdef IPSEC
2120 			case IPV6_IPSEC_POLICY:
2121 			  {
2122 				caddr_t req = NULL;
2123 				size_t len = 0;
2124 				struct mbuf *m = NULL;
2125 				struct mbuf **mp = &m;
2126 				size_t ovalsize = sopt->sopt_valsize;
2127 				caddr_t oval = (caddr_t)sopt->sopt_val;
2128 
2129 				error = soopt_getm(sopt, &m); /* XXX */
2130 				if (error != 0)
2131 					break;
2132 				error = soopt_mcopyin(sopt, m); /* XXX */
2133 				if (error != 0)
2134 					break;
2135 				sopt->sopt_valsize = ovalsize;
2136 				sopt->sopt_val = oval;
2137 				if (m) {
2138 					req = mtod(m, caddr_t);
2139 					len = m->m_len;
2140 				}
2141 				error = ipsec_get_policy(in6p, req, len, mp);
2142 				if (error == 0)
2143 					error = soopt_mcopyout(sopt, m); /* XXX */
2144 				if (error == 0 && m)
2145 					m_freem(m);
2146 				break;
2147 			  }
2148 #endif /* IPSEC */
2149 
2150 			default:
2151 				error = ENOPROTOOPT;
2152 				break;
2153 			}
2154 			break;
2155 		}
2156 	}
2157 	return (error);
2158 }
2159 
2160 int
2161 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2162 {
2163 	int error = 0, optval, optlen;
2164 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2165 	struct inpcb *in6p = sotoinpcb(so);
2166 	int level, op, optname;
2167 
2168 	level = sopt->sopt_level;
2169 	op = sopt->sopt_dir;
2170 	optname = sopt->sopt_name;
2171 	optlen = sopt->sopt_valsize;
2172 
2173 	if (level != IPPROTO_IPV6) {
2174 		return (EINVAL);
2175 	}
2176 
2177 	switch (optname) {
2178 	case IPV6_CHECKSUM:
2179 		/*
2180 		 * For ICMPv6 sockets, no modification allowed for checksum
2181 		 * offset, permit "no change" values to help existing apps.
2182 		 *
2183 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2184 		 * for an ICMPv6 socket will fail."
2185 		 * The current behavior does not meet RFC3542.
2186 		 */
2187 		switch (op) {
2188 		case SOPT_SET:
2189 			if (optlen != sizeof(int)) {
2190 				error = EINVAL;
2191 				break;
2192 			}
2193 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2194 					    sizeof(optval));
2195 			if (error)
2196 				break;
2197 			if ((optval % 2) != 0) {
2198 				/* the API assumes even offset values */
2199 				error = EINVAL;
2200 			} else if (so->so_proto->pr_protocol ==
2201 			    IPPROTO_ICMPV6) {
2202 				if (optval != icmp6off)
2203 					error = EINVAL;
2204 			} else
2205 				in6p->in6p_cksum = optval;
2206 			break;
2207 
2208 		case SOPT_GET:
2209 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2210 				optval = icmp6off;
2211 			else
2212 				optval = in6p->in6p_cksum;
2213 
2214 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2215 			break;
2216 
2217 		default:
2218 			error = EINVAL;
2219 			break;
2220 		}
2221 		break;
2222 
2223 	default:
2224 		error = ENOPROTOOPT;
2225 		break;
2226 	}
2227 
2228 	return (error);
2229 }
2230 
2231 /*
2232  * Set up IP6 options in pcb for insertion in output packets or
2233  * specifying behavior of outgoing packets.
2234  */
2235 static int
2236 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2237     struct socket *so, struct sockopt *sopt)
2238 {
2239 	struct ip6_pktopts *opt = *pktopt;
2240 	int error = 0;
2241 	struct thread *td = sopt->sopt_td;
2242 
2243 	/* turn off any old options. */
2244 	if (opt) {
2245 #ifdef DIAGNOSTIC
2246 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2247 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2248 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2249 			printf("ip6_pcbopts: all specified options are cleared.\n");
2250 #endif
2251 		ip6_clearpktopts(opt, -1);
2252 	} else
2253 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2254 	*pktopt = NULL;
2255 
2256 	if (!m || m->m_len == 0) {
2257 		/*
2258 		 * Only turning off any previous options, regardless of
2259 		 * whether the opt is just created or given.
2260 		 */
2261 		free(opt, M_IP6OPT);
2262 		return (0);
2263 	}
2264 
2265 	/*  set options specified by user. */
2266 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2267 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2268 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2269 		free(opt, M_IP6OPT);
2270 		return (error);
2271 	}
2272 	*pktopt = opt;
2273 	return (0);
2274 }
2275 
2276 /*
2277  * initialize ip6_pktopts.  beware that there are non-zero default values in
2278  * the struct.
2279  */
2280 void
2281 ip6_initpktopts(struct ip6_pktopts *opt)
2282 {
2283 
2284 	bzero(opt, sizeof(*opt));
2285 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2286 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2287 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2288 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2289 }
2290 
2291 static int
2292 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2293     struct ucred *cred, int uproto)
2294 {
2295 	struct ip6_pktopts *opt;
2296 
2297 	if (*pktopt == NULL) {
2298 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2299 		    M_WAITOK);
2300 		ip6_initpktopts(*pktopt);
2301 	}
2302 	opt = *pktopt;
2303 
2304 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2305 }
2306 
2307 static int
2308 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2309 {
2310 	void *optdata = NULL;
2311 	int optdatalen = 0;
2312 	struct ip6_ext *ip6e;
2313 	int error = 0;
2314 	struct in6_pktinfo null_pktinfo;
2315 	int deftclass = 0, on;
2316 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2317 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2318 
2319 	switch (optname) {
2320 	case IPV6_PKTINFO:
2321 		if (pktopt && pktopt->ip6po_pktinfo)
2322 			optdata = (void *)pktopt->ip6po_pktinfo;
2323 		else {
2324 			/* XXX: we don't have to do this every time... */
2325 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2326 			optdata = (void *)&null_pktinfo;
2327 		}
2328 		optdatalen = sizeof(struct in6_pktinfo);
2329 		break;
2330 	case IPV6_TCLASS:
2331 		if (pktopt && pktopt->ip6po_tclass >= 0)
2332 			optdata = (void *)&pktopt->ip6po_tclass;
2333 		else
2334 			optdata = (void *)&deftclass;
2335 		optdatalen = sizeof(int);
2336 		break;
2337 	case IPV6_HOPOPTS:
2338 		if (pktopt && pktopt->ip6po_hbh) {
2339 			optdata = (void *)pktopt->ip6po_hbh;
2340 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2341 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2342 		}
2343 		break;
2344 	case IPV6_RTHDR:
2345 		if (pktopt && pktopt->ip6po_rthdr) {
2346 			optdata = (void *)pktopt->ip6po_rthdr;
2347 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2348 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2349 		}
2350 		break;
2351 	case IPV6_RTHDRDSTOPTS:
2352 		if (pktopt && pktopt->ip6po_dest1) {
2353 			optdata = (void *)pktopt->ip6po_dest1;
2354 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2355 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2356 		}
2357 		break;
2358 	case IPV6_DSTOPTS:
2359 		if (pktopt && pktopt->ip6po_dest2) {
2360 			optdata = (void *)pktopt->ip6po_dest2;
2361 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2362 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2363 		}
2364 		break;
2365 	case IPV6_NEXTHOP:
2366 		if (pktopt && pktopt->ip6po_nexthop) {
2367 			optdata = (void *)pktopt->ip6po_nexthop;
2368 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2369 		}
2370 		break;
2371 	case IPV6_USE_MIN_MTU:
2372 		if (pktopt)
2373 			optdata = (void *)&pktopt->ip6po_minmtu;
2374 		else
2375 			optdata = (void *)&defminmtu;
2376 		optdatalen = sizeof(int);
2377 		break;
2378 	case IPV6_DONTFRAG:
2379 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2380 			on = 1;
2381 		else
2382 			on = 0;
2383 		optdata = (void *)&on;
2384 		optdatalen = sizeof(on);
2385 		break;
2386 	case IPV6_PREFER_TEMPADDR:
2387 		if (pktopt)
2388 			optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2389 		else
2390 			optdata = (void *)&defpreftemp;
2391 		optdatalen = sizeof(int);
2392 		break;
2393 	default:		/* should not happen */
2394 #ifdef DIAGNOSTIC
2395 		panic("ip6_getpcbopt: unexpected option\n");
2396 #endif
2397 		return (ENOPROTOOPT);
2398 	}
2399 
2400 	error = sooptcopyout(sopt, optdata, optdatalen);
2401 
2402 	return (error);
2403 }
2404 
2405 void
2406 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2407 {
2408 	if (pktopt == NULL)
2409 		return;
2410 
2411 	if (optname == -1 || optname == IPV6_PKTINFO) {
2412 		if (pktopt->ip6po_pktinfo)
2413 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2414 		pktopt->ip6po_pktinfo = NULL;
2415 	}
2416 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2417 		pktopt->ip6po_hlim = -1;
2418 	if (optname == -1 || optname == IPV6_TCLASS)
2419 		pktopt->ip6po_tclass = -1;
2420 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2421 		if (pktopt->ip6po_nextroute.ro_rt) {
2422 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2423 			pktopt->ip6po_nextroute.ro_rt = NULL;
2424 		}
2425 		if (pktopt->ip6po_nexthop)
2426 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2427 		pktopt->ip6po_nexthop = NULL;
2428 	}
2429 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2430 		if (pktopt->ip6po_hbh)
2431 			free(pktopt->ip6po_hbh, M_IP6OPT);
2432 		pktopt->ip6po_hbh = NULL;
2433 	}
2434 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2435 		if (pktopt->ip6po_dest1)
2436 			free(pktopt->ip6po_dest1, M_IP6OPT);
2437 		pktopt->ip6po_dest1 = NULL;
2438 	}
2439 	if (optname == -1 || optname == IPV6_RTHDR) {
2440 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2441 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2442 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2443 		if (pktopt->ip6po_route.ro_rt) {
2444 			RTFREE(pktopt->ip6po_route.ro_rt);
2445 			pktopt->ip6po_route.ro_rt = NULL;
2446 		}
2447 	}
2448 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2449 		if (pktopt->ip6po_dest2)
2450 			free(pktopt->ip6po_dest2, M_IP6OPT);
2451 		pktopt->ip6po_dest2 = NULL;
2452 	}
2453 }
2454 
2455 #define PKTOPT_EXTHDRCPY(type) \
2456 do {\
2457 	if (src->type) {\
2458 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2459 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2460 		if (dst->type == NULL && canwait == M_NOWAIT)\
2461 			goto bad;\
2462 		bcopy(src->type, dst->type, hlen);\
2463 	}\
2464 } while (/*CONSTCOND*/ 0)
2465 
2466 static int
2467 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2468 {
2469 	if (dst == NULL || src == NULL)  {
2470 		printf("ip6_clearpktopts: invalid argument\n");
2471 		return (EINVAL);
2472 	}
2473 
2474 	dst->ip6po_hlim = src->ip6po_hlim;
2475 	dst->ip6po_tclass = src->ip6po_tclass;
2476 	dst->ip6po_flags = src->ip6po_flags;
2477 	dst->ip6po_minmtu = src->ip6po_minmtu;
2478 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2479 	if (src->ip6po_pktinfo) {
2480 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2481 		    M_IP6OPT, canwait);
2482 		if (dst->ip6po_pktinfo == NULL)
2483 			goto bad;
2484 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2485 	}
2486 	if (src->ip6po_nexthop) {
2487 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2488 		    M_IP6OPT, canwait);
2489 		if (dst->ip6po_nexthop == NULL)
2490 			goto bad;
2491 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2492 		    src->ip6po_nexthop->sa_len);
2493 	}
2494 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2495 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2496 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2497 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2498 	return (0);
2499 
2500   bad:
2501 	ip6_clearpktopts(dst, -1);
2502 	return (ENOBUFS);
2503 }
2504 #undef PKTOPT_EXTHDRCPY
2505 
2506 struct ip6_pktopts *
2507 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2508 {
2509 	int error;
2510 	struct ip6_pktopts *dst;
2511 
2512 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2513 	if (dst == NULL)
2514 		return (NULL);
2515 	ip6_initpktopts(dst);
2516 
2517 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2518 		free(dst, M_IP6OPT);
2519 		return (NULL);
2520 	}
2521 
2522 	return (dst);
2523 }
2524 
2525 void
2526 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2527 {
2528 	if (pktopt == NULL)
2529 		return;
2530 
2531 	ip6_clearpktopts(pktopt, -1);
2532 
2533 	free(pktopt, M_IP6OPT);
2534 }
2535 
2536 /*
2537  * Set IPv6 outgoing packet options based on advanced API.
2538  */
2539 int
2540 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2541     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2542 {
2543 	struct cmsghdr *cm = 0;
2544 
2545 	if (control == NULL || opt == NULL)
2546 		return (EINVAL);
2547 
2548 	ip6_initpktopts(opt);
2549 	if (stickyopt) {
2550 		int error;
2551 
2552 		/*
2553 		 * If stickyopt is provided, make a local copy of the options
2554 		 * for this particular packet, then override them by ancillary
2555 		 * objects.
2556 		 * XXX: copypktopts() does not copy the cached route to a next
2557 		 * hop (if any).  This is not very good in terms of efficiency,
2558 		 * but we can allow this since this option should be rarely
2559 		 * used.
2560 		 */
2561 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2562 			return (error);
2563 	}
2564 
2565 	/*
2566 	 * XXX: Currently, we assume all the optional information is stored
2567 	 * in a single mbuf.
2568 	 */
2569 	if (control->m_next)
2570 		return (EINVAL);
2571 
2572 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2573 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2574 		int error;
2575 
2576 		if (control->m_len < CMSG_LEN(0))
2577 			return (EINVAL);
2578 
2579 		cm = mtod(control, struct cmsghdr *);
2580 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2581 			return (EINVAL);
2582 		if (cm->cmsg_level != IPPROTO_IPV6)
2583 			continue;
2584 
2585 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2586 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2587 		if (error)
2588 			return (error);
2589 	}
2590 
2591 	return (0);
2592 }
2593 
2594 /*
2595  * Set a particular packet option, as a sticky option or an ancillary data
2596  * item.  "len" can be 0 only when it's a sticky option.
2597  * We have 4 cases of combination of "sticky" and "cmsg":
2598  * "sticky=0, cmsg=0": impossible
2599  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2600  * "sticky=1, cmsg=0": RFC3542 socket option
2601  * "sticky=1, cmsg=1": RFC2292 socket option
2602  */
2603 static int
2604 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2605     struct ucred *cred, int sticky, int cmsg, int uproto)
2606 {
2607 	int minmtupolicy, preftemp;
2608 	int error;
2609 
2610 	if (!sticky && !cmsg) {
2611 #ifdef DIAGNOSTIC
2612 		printf("ip6_setpktopt: impossible case\n");
2613 #endif
2614 		return (EINVAL);
2615 	}
2616 
2617 	/*
2618 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2619 	 * not be specified in the context of RFC3542.  Conversely,
2620 	 * RFC3542 types should not be specified in the context of RFC2292.
2621 	 */
2622 	if (!cmsg) {
2623 		switch (optname) {
2624 		case IPV6_2292PKTINFO:
2625 		case IPV6_2292HOPLIMIT:
2626 		case IPV6_2292NEXTHOP:
2627 		case IPV6_2292HOPOPTS:
2628 		case IPV6_2292DSTOPTS:
2629 		case IPV6_2292RTHDR:
2630 		case IPV6_2292PKTOPTIONS:
2631 			return (ENOPROTOOPT);
2632 		}
2633 	}
2634 	if (sticky && cmsg) {
2635 		switch (optname) {
2636 		case IPV6_PKTINFO:
2637 		case IPV6_HOPLIMIT:
2638 		case IPV6_NEXTHOP:
2639 		case IPV6_HOPOPTS:
2640 		case IPV6_DSTOPTS:
2641 		case IPV6_RTHDRDSTOPTS:
2642 		case IPV6_RTHDR:
2643 		case IPV6_USE_MIN_MTU:
2644 		case IPV6_DONTFRAG:
2645 		case IPV6_TCLASS:
2646 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2647 			return (ENOPROTOOPT);
2648 		}
2649 	}
2650 
2651 	switch (optname) {
2652 	case IPV6_2292PKTINFO:
2653 	case IPV6_PKTINFO:
2654 	{
2655 		struct ifnet *ifp = NULL;
2656 		struct in6_pktinfo *pktinfo;
2657 
2658 		if (len != sizeof(struct in6_pktinfo))
2659 			return (EINVAL);
2660 
2661 		pktinfo = (struct in6_pktinfo *)buf;
2662 
2663 		/*
2664 		 * An application can clear any sticky IPV6_PKTINFO option by
2665 		 * doing a "regular" setsockopt with ipi6_addr being
2666 		 * in6addr_any and ipi6_ifindex being zero.
2667 		 * [RFC 3542, Section 6]
2668 		 */
2669 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2670 		    pktinfo->ipi6_ifindex == 0 &&
2671 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2672 			ip6_clearpktopts(opt, optname);
2673 			break;
2674 		}
2675 
2676 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2677 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2678 			return (EINVAL);
2679 		}
2680 
2681 		/* validate the interface index if specified. */
2682 		if (pktinfo->ipi6_ifindex > V_if_index ||
2683 		    pktinfo->ipi6_ifindex < 0) {
2684 			 return (ENXIO);
2685 		}
2686 		if (pktinfo->ipi6_ifindex) {
2687 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2688 			if (ifp == NULL)
2689 				return (ENXIO);
2690 		}
2691 
2692 		/*
2693 		 * We store the address anyway, and let in6_selectsrc()
2694 		 * validate the specified address.  This is because ipi6_addr
2695 		 * may not have enough information about its scope zone, and
2696 		 * we may need additional information (such as outgoing
2697 		 * interface or the scope zone of a destination address) to
2698 		 * disambiguate the scope.
2699 		 * XXX: the delay of the validation may confuse the
2700 		 * application when it is used as a sticky option.
2701 		 */
2702 		if (opt->ip6po_pktinfo == NULL) {
2703 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2704 			    M_IP6OPT, M_NOWAIT);
2705 			if (opt->ip6po_pktinfo == NULL)
2706 				return (ENOBUFS);
2707 		}
2708 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2709 		break;
2710 	}
2711 
2712 	case IPV6_2292HOPLIMIT:
2713 	case IPV6_HOPLIMIT:
2714 	{
2715 		int *hlimp;
2716 
2717 		/*
2718 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2719 		 * to simplify the ordering among hoplimit options.
2720 		 */
2721 		if (optname == IPV6_HOPLIMIT && sticky)
2722 			return (ENOPROTOOPT);
2723 
2724 		if (len != sizeof(int))
2725 			return (EINVAL);
2726 		hlimp = (int *)buf;
2727 		if (*hlimp < -1 || *hlimp > 255)
2728 			return (EINVAL);
2729 
2730 		opt->ip6po_hlim = *hlimp;
2731 		break;
2732 	}
2733 
2734 	case IPV6_TCLASS:
2735 	{
2736 		int tclass;
2737 
2738 		if (len != sizeof(int))
2739 			return (EINVAL);
2740 		tclass = *(int *)buf;
2741 		if (tclass < -1 || tclass > 255)
2742 			return (EINVAL);
2743 
2744 		opt->ip6po_tclass = tclass;
2745 		break;
2746 	}
2747 
2748 	case IPV6_2292NEXTHOP:
2749 	case IPV6_NEXTHOP:
2750 		if (cred != NULL) {
2751 			error = priv_check_cred(cred,
2752 			    PRIV_NETINET_SETHDROPTS, 0);
2753 			if (error)
2754 				return (error);
2755 		}
2756 
2757 		if (len == 0) {	/* just remove the option */
2758 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2759 			break;
2760 		}
2761 
2762 		/* check if cmsg_len is large enough for sa_len */
2763 		if (len < sizeof(struct sockaddr) || len < *buf)
2764 			return (EINVAL);
2765 
2766 		switch (((struct sockaddr *)buf)->sa_family) {
2767 		case AF_INET6:
2768 		{
2769 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2770 			int error;
2771 
2772 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2773 				return (EINVAL);
2774 
2775 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2776 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2777 				return (EINVAL);
2778 			}
2779 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2780 			    != 0) {
2781 				return (error);
2782 			}
2783 			break;
2784 		}
2785 		case AF_LINK:	/* should eventually be supported */
2786 		default:
2787 			return (EAFNOSUPPORT);
2788 		}
2789 
2790 		/* turn off the previous option, then set the new option. */
2791 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2792 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2793 		if (opt->ip6po_nexthop == NULL)
2794 			return (ENOBUFS);
2795 		bcopy(buf, opt->ip6po_nexthop, *buf);
2796 		break;
2797 
2798 	case IPV6_2292HOPOPTS:
2799 	case IPV6_HOPOPTS:
2800 	{
2801 		struct ip6_hbh *hbh;
2802 		int hbhlen;
2803 
2804 		/*
2805 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2806 		 * options, since per-option restriction has too much
2807 		 * overhead.
2808 		 */
2809 		if (cred != NULL) {
2810 			error = priv_check_cred(cred,
2811 			    PRIV_NETINET_SETHDROPTS, 0);
2812 			if (error)
2813 				return (error);
2814 		}
2815 
2816 		if (len == 0) {
2817 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2818 			break;	/* just remove the option */
2819 		}
2820 
2821 		/* message length validation */
2822 		if (len < sizeof(struct ip6_hbh))
2823 			return (EINVAL);
2824 		hbh = (struct ip6_hbh *)buf;
2825 		hbhlen = (hbh->ip6h_len + 1) << 3;
2826 		if (len != hbhlen)
2827 			return (EINVAL);
2828 
2829 		/* turn off the previous option, then set the new option. */
2830 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2831 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2832 		if (opt->ip6po_hbh == NULL)
2833 			return (ENOBUFS);
2834 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
2835 
2836 		break;
2837 	}
2838 
2839 	case IPV6_2292DSTOPTS:
2840 	case IPV6_DSTOPTS:
2841 	case IPV6_RTHDRDSTOPTS:
2842 	{
2843 		struct ip6_dest *dest, **newdest = NULL;
2844 		int destlen;
2845 
2846 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2847 			error = priv_check_cred(cred,
2848 			    PRIV_NETINET_SETHDROPTS, 0);
2849 			if (error)
2850 				return (error);
2851 		}
2852 
2853 		if (len == 0) {
2854 			ip6_clearpktopts(opt, optname);
2855 			break;	/* just remove the option */
2856 		}
2857 
2858 		/* message length validation */
2859 		if (len < sizeof(struct ip6_dest))
2860 			return (EINVAL);
2861 		dest = (struct ip6_dest *)buf;
2862 		destlen = (dest->ip6d_len + 1) << 3;
2863 		if (len != destlen)
2864 			return (EINVAL);
2865 
2866 		/*
2867 		 * Determine the position that the destination options header
2868 		 * should be inserted; before or after the routing header.
2869 		 */
2870 		switch (optname) {
2871 		case IPV6_2292DSTOPTS:
2872 			/*
2873 			 * The old advacned API is ambiguous on this point.
2874 			 * Our approach is to determine the position based
2875 			 * according to the existence of a routing header.
2876 			 * Note, however, that this depends on the order of the
2877 			 * extension headers in the ancillary data; the 1st
2878 			 * part of the destination options header must appear
2879 			 * before the routing header in the ancillary data,
2880 			 * too.
2881 			 * RFC3542 solved the ambiguity by introducing
2882 			 * separate ancillary data or option types.
2883 			 */
2884 			if (opt->ip6po_rthdr == NULL)
2885 				newdest = &opt->ip6po_dest1;
2886 			else
2887 				newdest = &opt->ip6po_dest2;
2888 			break;
2889 		case IPV6_RTHDRDSTOPTS:
2890 			newdest = &opt->ip6po_dest1;
2891 			break;
2892 		case IPV6_DSTOPTS:
2893 			newdest = &opt->ip6po_dest2;
2894 			break;
2895 		}
2896 
2897 		/* turn off the previous option, then set the new option. */
2898 		ip6_clearpktopts(opt, optname);
2899 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2900 		if (*newdest == NULL)
2901 			return (ENOBUFS);
2902 		bcopy(dest, *newdest, destlen);
2903 
2904 		break;
2905 	}
2906 
2907 	case IPV6_2292RTHDR:
2908 	case IPV6_RTHDR:
2909 	{
2910 		struct ip6_rthdr *rth;
2911 		int rthlen;
2912 
2913 		if (len == 0) {
2914 			ip6_clearpktopts(opt, IPV6_RTHDR);
2915 			break;	/* just remove the option */
2916 		}
2917 
2918 		/* message length validation */
2919 		if (len < sizeof(struct ip6_rthdr))
2920 			return (EINVAL);
2921 		rth = (struct ip6_rthdr *)buf;
2922 		rthlen = (rth->ip6r_len + 1) << 3;
2923 		if (len != rthlen)
2924 			return (EINVAL);
2925 
2926 		switch (rth->ip6r_type) {
2927 		case IPV6_RTHDR_TYPE_0:
2928 			if (rth->ip6r_len == 0)	/* must contain one addr */
2929 				return (EINVAL);
2930 			if (rth->ip6r_len % 2) /* length must be even */
2931 				return (EINVAL);
2932 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2933 				return (EINVAL);
2934 			break;
2935 		default:
2936 			return (EINVAL);	/* not supported */
2937 		}
2938 
2939 		/* turn off the previous option */
2940 		ip6_clearpktopts(opt, IPV6_RTHDR);
2941 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2942 		if (opt->ip6po_rthdr == NULL)
2943 			return (ENOBUFS);
2944 		bcopy(rth, opt->ip6po_rthdr, rthlen);
2945 
2946 		break;
2947 	}
2948 
2949 	case IPV6_USE_MIN_MTU:
2950 		if (len != sizeof(int))
2951 			return (EINVAL);
2952 		minmtupolicy = *(int *)buf;
2953 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2954 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
2955 		    minmtupolicy != IP6PO_MINMTU_ALL) {
2956 			return (EINVAL);
2957 		}
2958 		opt->ip6po_minmtu = minmtupolicy;
2959 		break;
2960 
2961 	case IPV6_DONTFRAG:
2962 		if (len != sizeof(int))
2963 			return (EINVAL);
2964 
2965 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2966 			/*
2967 			 * we ignore this option for TCP sockets.
2968 			 * (RFC3542 leaves this case unspecified.)
2969 			 */
2970 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2971 		} else
2972 			opt->ip6po_flags |= IP6PO_DONTFRAG;
2973 		break;
2974 
2975 	case IPV6_PREFER_TEMPADDR:
2976 		if (len != sizeof(int))
2977 			return (EINVAL);
2978 		preftemp = *(int *)buf;
2979 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2980 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2981 		    preftemp != IP6PO_TEMPADDR_PREFER) {
2982 			return (EINVAL);
2983 		}
2984 		opt->ip6po_prefer_tempaddr = preftemp;
2985 		break;
2986 
2987 	default:
2988 		return (ENOPROTOOPT);
2989 	} /* end of switch */
2990 
2991 	return (0);
2992 }
2993 
2994 /*
2995  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2996  * packet to the input queue of a specified interface.  Note that this
2997  * calls the output routine of the loopback "driver", but with an interface
2998  * pointer that might NOT be &loif -- easier than replicating that code here.
2999  */
3000 void
3001 ip6_mloopback(struct ifnet *ifp, struct mbuf *m, struct sockaddr_in6 *dst)
3002 {
3003 	struct mbuf *copym;
3004 	struct ip6_hdr *ip6;
3005 
3006 	copym = m_copy(m, 0, M_COPYALL);
3007 	if (copym == NULL)
3008 		return;
3009 
3010 	/*
3011 	 * Make sure to deep-copy IPv6 header portion in case the data
3012 	 * is in an mbuf cluster, so that we can safely override the IPv6
3013 	 * header portion later.
3014 	 */
3015 	if ((copym->m_flags & M_EXT) != 0 ||
3016 	    copym->m_len < sizeof(struct ip6_hdr)) {
3017 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3018 		if (copym == NULL)
3019 			return;
3020 	}
3021 
3022 #ifdef DIAGNOSTIC
3023 	if (copym->m_len < sizeof(*ip6)) {
3024 		m_freem(copym);
3025 		return;
3026 	}
3027 #endif
3028 
3029 	ip6 = mtod(copym, struct ip6_hdr *);
3030 	/*
3031 	 * clear embedded scope identifiers if necessary.
3032 	 * in6_clearscope will touch the addresses only when necessary.
3033 	 */
3034 	in6_clearscope(&ip6->ip6_src);
3035 	in6_clearscope(&ip6->ip6_dst);
3036 
3037 	(void)if_simloop(ifp, copym, dst->sin6_family, 0);
3038 }
3039 
3040 /*
3041  * Chop IPv6 header off from the payload.
3042  */
3043 static int
3044 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3045 {
3046 	struct mbuf *mh;
3047 	struct ip6_hdr *ip6;
3048 
3049 	ip6 = mtod(m, struct ip6_hdr *);
3050 	if (m->m_len > sizeof(*ip6)) {
3051 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
3052 		if (mh == 0) {
3053 			m_freem(m);
3054 			return ENOBUFS;
3055 		}
3056 		M_MOVE_PKTHDR(mh, m);
3057 		MH_ALIGN(mh, sizeof(*ip6));
3058 		m->m_len -= sizeof(*ip6);
3059 		m->m_data += sizeof(*ip6);
3060 		mh->m_next = m;
3061 		m = mh;
3062 		m->m_len = sizeof(*ip6);
3063 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3064 	}
3065 	exthdrs->ip6e_ip6 = m;
3066 	return 0;
3067 }
3068 
3069 /*
3070  * Compute IPv6 extension header length.
3071  */
3072 int
3073 ip6_optlen(struct inpcb *in6p)
3074 {
3075 	int len;
3076 
3077 	if (!in6p->in6p_outputopts)
3078 		return 0;
3079 
3080 	len = 0;
3081 #define elen(x) \
3082     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3083 
3084 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3085 	if (in6p->in6p_outputopts->ip6po_rthdr)
3086 		/* dest1 is valid with rthdr only */
3087 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
3088 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3089 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3090 	return len;
3091 #undef elen
3092 }
3093