xref: /freebsd/sys/netinet6/ip6_output.c (revision 18849b5da0c5eaa88500b457be05b038813b51b1)
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 #include "opt_rss.h"
73 
74 #include <sys/param.h>
75 #include <sys/kernel.h>
76 #include <sys/malloc.h>
77 #include <sys/mbuf.h>
78 #include <sys/errno.h>
79 #include <sys/priv.h>
80 #include <sys/proc.h>
81 #include <sys/protosw.h>
82 #include <sys/socket.h>
83 #include <sys/socketvar.h>
84 #include <sys/syslog.h>
85 #include <sys/ucred.h>
86 
87 #include <machine/in_cksum.h>
88 
89 #include <net/if.h>
90 #include <net/if_var.h>
91 #include <net/netisr.h>
92 #include <net/route.h>
93 #include <net/pfil.h>
94 #include <net/rss_config.h>
95 #include <net/vnet.h>
96 
97 #include <netinet/in.h>
98 #include <netinet/in_var.h>
99 #include <netinet/ip_var.h>
100 #include <netinet6/in6_fib.h>
101 #include <netinet6/in6_var.h>
102 #include <netinet/ip6.h>
103 #include <netinet/icmp6.h>
104 #include <netinet6/ip6_var.h>
105 #include <netinet/in_pcb.h>
106 #include <netinet/tcp_var.h>
107 #include <netinet6/nd6.h>
108 #include <netinet6/in6_rss.h>
109 
110 #ifdef IPSEC
111 #include <netipsec/ipsec.h>
112 #include <netipsec/ipsec6.h>
113 #include <netipsec/key.h>
114 #include <netinet6/ip6_ipsec.h>
115 #endif /* IPSEC */
116 #ifdef SCTP
117 #include <netinet/sctp.h>
118 #include <netinet/sctp_crc32.h>
119 #endif
120 
121 #include <netinet6/ip6protosw.h>
122 #include <netinet6/scope6_var.h>
123 
124 #ifdef FLOWTABLE
125 #include <net/flowtable.h>
126 #endif
127 
128 extern int in6_mcast_loop;
129 
130 struct ip6_exthdrs {
131 	struct mbuf *ip6e_ip6;
132 	struct mbuf *ip6e_hbh;
133 	struct mbuf *ip6e_dest1;
134 	struct mbuf *ip6e_rthdr;
135 	struct mbuf *ip6e_dest2;
136 };
137 
138 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
139 			   struct ucred *, int);
140 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
141 	struct socket *, struct sockopt *);
142 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
143 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
144 	struct ucred *, int, int, int);
145 
146 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
147 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
148 	struct ip6_frag **);
149 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
150 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
151 static int ip6_getpmtu(struct route_in6 *, int,
152 	struct ifnet *, const struct in6_addr *, u_long *, int *, u_int);
153 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
154 	u_long *, int *);
155 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
156 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
157 
158 
159 /*
160  * Make an extension header from option data.  hp is the source, and
161  * mp is the destination.
162  */
163 #define MAKE_EXTHDR(hp, mp)						\
164     do {								\
165 	if (hp) {							\
166 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
167 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
168 		    ((eh)->ip6e_len + 1) << 3);				\
169 		if (error)						\
170 			goto freehdrs;					\
171 	}								\
172     } while (/*CONSTCOND*/ 0)
173 
174 /*
175  * Form a chain of extension headers.
176  * m is the extension header mbuf
177  * mp is the previous mbuf in the chain
178  * p is the next header
179  * i is the type of option.
180  */
181 #define MAKE_CHAIN(m, mp, p, i)\
182     do {\
183 	if (m) {\
184 		if (!hdrsplit) \
185 			panic("assumption failed: hdr not split"); \
186 		*mtod((m), u_char *) = *(p);\
187 		*(p) = (i);\
188 		p = mtod((m), u_char *);\
189 		(m)->m_next = (mp)->m_next;\
190 		(mp)->m_next = (m);\
191 		(mp) = (m);\
192 	}\
193     } while (/*CONSTCOND*/ 0)
194 
195 void
196 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
197 {
198 	u_short csum;
199 
200 	csum = in_cksum_skip(m, offset + plen, offset);
201 	if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
202 		csum = 0xffff;
203 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
204 
205 	if (offset + sizeof(u_short) > m->m_len) {
206 		printf("%s: delayed m_pullup, m->len: %d plen %u off %u "
207 		    "csum_flags=%b\n", __func__, m->m_len, plen, offset,
208 		    (int)m->m_pkthdr.csum_flags, CSUM_BITS);
209 		/*
210 		 * XXX this should not happen, but if it does, the correct
211 		 * behavior may be to insert the checksum in the appropriate
212 		 * next mbuf in the chain.
213 		 */
214 		return;
215 	}
216 	*(u_short *)(m->m_data + offset) = csum;
217 }
218 
219 int
220 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
221     int mtu, uint32_t id)
222 {
223 	struct mbuf *m, **mnext, *m_frgpart;
224 	struct ip6_hdr *ip6, *mhip6;
225 	struct ip6_frag *ip6f;
226 	int off;
227 	int error;
228 	int tlen = m0->m_pkthdr.len;
229 
230 	m = m0;
231 	ip6 = mtod(m, struct ip6_hdr *);
232 	mnext = &m->m_nextpkt;
233 
234 	for (off = hlen; off < tlen; off += mtu) {
235 		m = m_gethdr(M_NOWAIT, MT_DATA);
236 		if (!m) {
237 			IP6STAT_INC(ip6s_odropped);
238 			return (ENOBUFS);
239 		}
240 		m->m_flags = m0->m_flags & M_COPYFLAGS;
241 		*mnext = m;
242 		mnext = &m->m_nextpkt;
243 		m->m_data += max_linkhdr;
244 		mhip6 = mtod(m, struct ip6_hdr *);
245 		*mhip6 = *ip6;
246 		m->m_len = sizeof(*mhip6);
247 		error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
248 		if (error) {
249 			IP6STAT_INC(ip6s_odropped);
250 			return (error);
251 		}
252 		ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
253 		if (off + mtu >= tlen)
254 			mtu = tlen - off;
255 		else
256 			ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
257 		mhip6->ip6_plen = htons((u_short)(mtu + hlen +
258 		    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
259 		if ((m_frgpart = m_copy(m0, off, mtu)) == NULL) {
260 			IP6STAT_INC(ip6s_odropped);
261 			return (ENOBUFS);
262 		}
263 		m_cat(m, m_frgpart);
264 		m->m_pkthdr.len = mtu + hlen + sizeof(*ip6f);
265 		m->m_pkthdr.fibnum = m0->m_pkthdr.fibnum;
266 		m->m_pkthdr.rcvif = NULL;
267 		ip6f->ip6f_reserved = 0;
268 		ip6f->ip6f_ident = id;
269 		ip6f->ip6f_nxt = nextproto;
270 		IP6STAT_INC(ip6s_ofragments);
271 		in6_ifstat_inc(ifp, ifs6_out_fragcreat);
272 	}
273 
274 	return (0);
275 }
276 
277 /*
278  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
279  * header (with pri, len, nxt, hlim, src, dst).
280  * This function may modify ver and hlim only.
281  * The mbuf chain containing the packet will be freed.
282  * The mbuf opt, if present, will not be freed.
283  * If route_in6 ro is present and has ro_rt initialized, route lookup would be
284  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
285  * then result of route lookup is stored in ro->ro_rt.
286  *
287  * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and
288  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
289  * which is rt_mtu.
290  *
291  * ifpp - XXX: just for statistics
292  */
293 /*
294  * XXX TODO: no flowid is assigned for outbound flows?
295  */
296 int
297 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
298     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
299     struct ifnet **ifpp, struct inpcb *inp)
300 {
301 	struct ip6_hdr *ip6;
302 	struct ifnet *ifp, *origifp;
303 	struct mbuf *m = m0;
304 	struct mbuf *mprev = NULL;
305 	int hlen, tlen, len;
306 	struct route_in6 ip6route;
307 	struct rtentry *rt = NULL;
308 	struct sockaddr_in6 *dst, src_sa, dst_sa;
309 	struct in6_addr odst;
310 	int error = 0;
311 	struct in6_ifaddr *ia = NULL;
312 	u_long mtu;
313 	int alwaysfrag, dontfrag;
314 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
315 	struct ip6_exthdrs exthdrs;
316 	struct in6_addr src0, dst0;
317 	u_int32_t zone;
318 	struct route_in6 *ro_pmtu = NULL;
319 	int hdrsplit = 0;
320 	int sw_csum, tso;
321 	int needfiblookup;
322 	uint32_t fibnum;
323 	struct m_tag *fwd_tag = NULL;
324 	uint32_t id;
325 
326 	ip6 = mtod(m, struct ip6_hdr *);
327 	if (ip6 == NULL) {
328 		printf ("ip6 is NULL");
329 		goto bad;
330 	}
331 
332 	if (inp != NULL) {
333 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
334 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
335 			/* unconditionally set flowid */
336 			m->m_pkthdr.flowid = inp->inp_flowid;
337 			M_HASHTYPE_SET(m, inp->inp_flowtype);
338 		}
339 	}
340 
341 	bzero(&exthdrs, sizeof(exthdrs));
342 	if (opt) {
343 		/* Hop-by-Hop options header */
344 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
345 		/* Destination options header(1st part) */
346 		if (opt->ip6po_rthdr) {
347 			/*
348 			 * Destination options header(1st part)
349 			 * This only makes sense with a routing header.
350 			 * See Section 9.2 of RFC 3542.
351 			 * Disabling this part just for MIP6 convenience is
352 			 * a bad idea.  We need to think carefully about a
353 			 * way to make the advanced API coexist with MIP6
354 			 * options, which might automatically be inserted in
355 			 * the kernel.
356 			 */
357 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
358 		}
359 		/* Routing header */
360 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
361 		/* Destination options header(2nd part) */
362 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
363 	}
364 
365 #ifdef IPSEC
366 	/*
367 	 * IPSec checking which handles several cases.
368 	 * FAST IPSEC: We re-injected the packet.
369 	 * XXX: need scope argument.
370 	 */
371 	switch(ip6_ipsec_output(&m, inp, &error))
372 	{
373 	case 1:                 /* Bad packet */
374 		goto freehdrs;
375 	case -1:                /* IPSec done */
376 		goto done;
377 	case 0:                 /* No IPSec */
378 	default:
379 		break;
380 	}
381 #endif /* IPSEC */
382 
383 	/*
384 	 * Calculate the total length of the extension header chain.
385 	 * Keep the length of the unfragmentable part for fragmentation.
386 	 */
387 	optlen = 0;
388 	if (exthdrs.ip6e_hbh)
389 		optlen += exthdrs.ip6e_hbh->m_len;
390 	if (exthdrs.ip6e_dest1)
391 		optlen += exthdrs.ip6e_dest1->m_len;
392 	if (exthdrs.ip6e_rthdr)
393 		optlen += exthdrs.ip6e_rthdr->m_len;
394 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
395 
396 	/* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */
397 	if (exthdrs.ip6e_dest2)
398 		optlen += exthdrs.ip6e_dest2->m_len;
399 
400 	/*
401 	 * If there is at least one extension header,
402 	 * separate IP6 header from the payload.
403 	 */
404 	if (optlen && !hdrsplit) {
405 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
406 			m = NULL;
407 			goto freehdrs;
408 		}
409 		m = exthdrs.ip6e_ip6;
410 		hdrsplit++;
411 	}
412 
413 	/* adjust pointer */
414 	ip6 = mtod(m, struct ip6_hdr *);
415 
416 	/* adjust mbuf packet header length */
417 	m->m_pkthdr.len += optlen;
418 	plen = m->m_pkthdr.len - sizeof(*ip6);
419 
420 	/* If this is a jumbo payload, insert a jumbo payload option. */
421 	if (plen > IPV6_MAXPACKET) {
422 		if (!hdrsplit) {
423 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
424 				m = NULL;
425 				goto freehdrs;
426 			}
427 			m = exthdrs.ip6e_ip6;
428 			hdrsplit++;
429 		}
430 		/* adjust pointer */
431 		ip6 = mtod(m, struct ip6_hdr *);
432 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
433 			goto freehdrs;
434 		ip6->ip6_plen = 0;
435 	} else
436 		ip6->ip6_plen = htons(plen);
437 
438 	/*
439 	 * Concatenate headers and fill in next header fields.
440 	 * Here we have, on "m"
441 	 *	IPv6 payload
442 	 * and we insert headers accordingly.  Finally, we should be getting:
443 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
444 	 *
445 	 * during the header composing process, "m" points to IPv6 header.
446 	 * "mprev" points to an extension header prior to esp.
447 	 */
448 	u_char *nexthdrp = &ip6->ip6_nxt;
449 	mprev = m;
450 
451 	/*
452 	 * we treat dest2 specially.  this makes IPsec processing
453 	 * much easier.  the goal here is to make mprev point the
454 	 * mbuf prior to dest2.
455 	 *
456 	 * result: IPv6 dest2 payload
457 	 * m and mprev will point to IPv6 header.
458 	 */
459 	if (exthdrs.ip6e_dest2) {
460 		if (!hdrsplit)
461 			panic("assumption failed: hdr not split");
462 		exthdrs.ip6e_dest2->m_next = m->m_next;
463 		m->m_next = exthdrs.ip6e_dest2;
464 		*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
465 		ip6->ip6_nxt = IPPROTO_DSTOPTS;
466 	}
467 
468 	/*
469 	 * result: IPv6 hbh dest1 rthdr dest2 payload
470 	 * m will point to IPv6 header.  mprev will point to the
471 	 * extension header prior to dest2 (rthdr in the above case).
472 	 */
473 	MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
474 	MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
475 		   IPPROTO_DSTOPTS);
476 	MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
477 		   IPPROTO_ROUTING);
478 
479 	/*
480 	 * If there is a routing header, discard the packet.
481 	 */
482 	if (exthdrs.ip6e_rthdr) {
483 		 error = EINVAL;
484 		 goto bad;
485 	}
486 
487 	/* Source address validation */
488 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
489 	    (flags & IPV6_UNSPECSRC) == 0) {
490 		error = EOPNOTSUPP;
491 		IP6STAT_INC(ip6s_badscope);
492 		goto bad;
493 	}
494 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
495 		error = EOPNOTSUPP;
496 		IP6STAT_INC(ip6s_badscope);
497 		goto bad;
498 	}
499 
500 	IP6STAT_INC(ip6s_localout);
501 
502 	/*
503 	 * Route packet.
504 	 */
505 	if (ro == NULL) {
506 		ro = &ip6route;
507 		bzero((caddr_t)ro, sizeof(*ro));
508 	}
509 	ro_pmtu = ro;
510 	if (opt && opt->ip6po_rthdr)
511 		ro = &opt->ip6po_route;
512 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
513 #ifdef FLOWTABLE
514 	if (ro->ro_rt == NULL)
515 		(void )flowtable_lookup(AF_INET6, m, (struct route *)ro);
516 #endif
517 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
518 again:
519 	/*
520 	 * if specified, try to fill in the traffic class field.
521 	 * do not override if a non-zero value is already set.
522 	 * we check the diffserv field and the ecn field separately.
523 	 */
524 	if (opt && opt->ip6po_tclass >= 0) {
525 		int mask = 0;
526 
527 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
528 			mask |= 0xfc;
529 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
530 			mask |= 0x03;
531 		if (mask != 0)
532 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
533 	}
534 
535 	/* fill in or override the hop limit field, if necessary. */
536 	if (opt && opt->ip6po_hlim != -1)
537 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
538 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
539 		if (im6o != NULL)
540 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
541 		else
542 			ip6->ip6_hlim = V_ip6_defmcasthlim;
543 	}
544 
545 	/* adjust pointer */
546 	ip6 = mtod(m, struct ip6_hdr *);
547 
548 	/*
549 	 * Validate route against routing table additions;
550 	 * a better/more specific route might have been added.
551 	 * Make sure address family is set in route.
552 	 */
553 	if (inp) {
554 		ro->ro_dst.sin6_family = AF_INET6;
555 		RT_VALIDATE((struct route *)ro, &inp->inp_rt_cookie, fibnum);
556 	}
557 	if (ro->ro_rt && fwd_tag == NULL && (ro->ro_rt->rt_flags & RTF_UP) &&
558 	    ro->ro_dst.sin6_family == AF_INET6 &&
559 	    IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
560 		rt = ro->ro_rt;
561 		ifp = ro->ro_rt->rt_ifp;
562 	} else {
563 		if (fwd_tag == NULL) {
564 			bzero(&dst_sa, sizeof(dst_sa));
565 			dst_sa.sin6_family = AF_INET6;
566 			dst_sa.sin6_len = sizeof(dst_sa);
567 			dst_sa.sin6_addr = ip6->ip6_dst;
568 		}
569 		error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
570 		    &rt, fibnum);
571 		if (error != 0) {
572 			if (ifp != NULL)
573 				in6_ifstat_inc(ifp, ifs6_out_discard);
574 			goto bad;
575 		}
576 	}
577 	if (rt == NULL) {
578 		/*
579 		 * If in6_selectroute() does not return a route entry,
580 		 * dst may not have been updated.
581 		 */
582 		*dst = dst_sa;	/* XXX */
583 	}
584 
585 	/*
586 	 * then rt (for unicast) and ifp must be non-NULL valid values.
587 	 */
588 	if ((flags & IPV6_FORWARDING) == 0) {
589 		/* XXX: the FORWARDING flag can be set for mrouting. */
590 		in6_ifstat_inc(ifp, ifs6_out_request);
591 	}
592 	if (rt != NULL) {
593 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
594 		counter_u64_add(rt->rt_pksent, 1);
595 	}
596 
597 
598 	/*
599 	 * The outgoing interface must be in the zone of source and
600 	 * destination addresses.
601 	 */
602 	origifp = ifp;
603 
604 	src0 = ip6->ip6_src;
605 	if (in6_setscope(&src0, origifp, &zone))
606 		goto badscope;
607 	bzero(&src_sa, sizeof(src_sa));
608 	src_sa.sin6_family = AF_INET6;
609 	src_sa.sin6_len = sizeof(src_sa);
610 	src_sa.sin6_addr = ip6->ip6_src;
611 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
612 		goto badscope;
613 
614 	dst0 = ip6->ip6_dst;
615 	if (in6_setscope(&dst0, origifp, &zone))
616 		goto badscope;
617 	/* re-initialize to be sure */
618 	bzero(&dst_sa, sizeof(dst_sa));
619 	dst_sa.sin6_family = AF_INET6;
620 	dst_sa.sin6_len = sizeof(dst_sa);
621 	dst_sa.sin6_addr = ip6->ip6_dst;
622 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
623 		goto badscope;
624 	}
625 
626 	/* We should use ia_ifp to support the case of
627 	 * sending packets to an address of our own.
628 	 */
629 	if (ia != NULL && ia->ia_ifp)
630 		ifp = ia->ia_ifp;
631 
632 	/* scope check is done. */
633 	goto routefound;
634 
635   badscope:
636 	IP6STAT_INC(ip6s_badscope);
637 	in6_ifstat_inc(origifp, ifs6_out_discard);
638 	if (error == 0)
639 		error = EHOSTUNREACH; /* XXX */
640 	goto bad;
641 
642   routefound:
643 	if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
644 		if (opt && opt->ip6po_nextroute.ro_rt) {
645 			/*
646 			 * The nexthop is explicitly specified by the
647 			 * application.  We assume the next hop is an IPv6
648 			 * address.
649 			 */
650 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
651 		}
652 		else if ((rt->rt_flags & RTF_GATEWAY))
653 			dst = (struct sockaddr_in6 *)rt->rt_gateway;
654 	}
655 
656 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
657 		m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
658 	} else {
659 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
660 		in6_ifstat_inc(ifp, ifs6_out_mcast);
661 		/*
662 		 * Confirm that the outgoing interface supports multicast.
663 		 */
664 		if (!(ifp->if_flags & IFF_MULTICAST)) {
665 			IP6STAT_INC(ip6s_noroute);
666 			in6_ifstat_inc(ifp, ifs6_out_discard);
667 			error = ENETUNREACH;
668 			goto bad;
669 		}
670 		if ((im6o == NULL && in6_mcast_loop) ||
671 		    (im6o && im6o->im6o_multicast_loop)) {
672 			/*
673 			 * Loop back multicast datagram if not expressly
674 			 * forbidden to do so, even if we have not joined
675 			 * the address; protocols will filter it later,
676 			 * thus deferring a hash lookup and lock acquisition
677 			 * at the expense of an m_copym().
678 			 */
679 			ip6_mloopback(ifp, m);
680 		} else {
681 			/*
682 			 * If we are acting as a multicast router, perform
683 			 * multicast forwarding as if the packet had just
684 			 * arrived on the interface to which we are about
685 			 * to send.  The multicast forwarding function
686 			 * recursively calls this function, using the
687 			 * IPV6_FORWARDING flag to prevent infinite recursion.
688 			 *
689 			 * Multicasts that are looped back by ip6_mloopback(),
690 			 * above, will be forwarded by the ip6_input() routine,
691 			 * if necessary.
692 			 */
693 			if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
694 				/*
695 				 * XXX: ip6_mforward expects that rcvif is NULL
696 				 * when it is called from the originating path.
697 				 * However, it may not always be the case.
698 				 */
699 				m->m_pkthdr.rcvif = NULL;
700 				if (ip6_mforward(ip6, ifp, m) != 0) {
701 					m_freem(m);
702 					goto done;
703 				}
704 			}
705 		}
706 		/*
707 		 * Multicasts with a hoplimit of zero may be looped back,
708 		 * above, but must not be transmitted on a network.
709 		 * Also, multicasts addressed to the loopback interface
710 		 * are not sent -- the above call to ip6_mloopback() will
711 		 * loop back a copy if this host actually belongs to the
712 		 * destination group on the loopback interface.
713 		 */
714 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
715 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
716 			m_freem(m);
717 			goto done;
718 		}
719 	}
720 
721 	/*
722 	 * Fill the outgoing inteface to tell the upper layer
723 	 * to increment per-interface statistics.
724 	 */
725 	if (ifpp)
726 		*ifpp = ifp;
727 
728 	/* Determine path MTU. */
729 	if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
730 	    &mtu, &alwaysfrag, fibnum)) != 0)
731 		goto bad;
732 
733 	/*
734 	 * The caller of this function may specify to use the minimum MTU
735 	 * in some cases.
736 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
737 	 * setting.  The logic is a bit complicated; by default, unicast
738 	 * packets will follow path MTU while multicast packets will be sent at
739 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
740 	 * including unicast ones will be sent at the minimum MTU.  Multicast
741 	 * packets will always be sent at the minimum MTU unless
742 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
743 	 * See RFC 3542 for more details.
744 	 */
745 	if (mtu > IPV6_MMTU) {
746 		if ((flags & IPV6_MINMTU))
747 			mtu = IPV6_MMTU;
748 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
749 			mtu = IPV6_MMTU;
750 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
751 			 (opt == NULL ||
752 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
753 			mtu = IPV6_MMTU;
754 		}
755 	}
756 
757 	/*
758 	 * clear embedded scope identifiers if necessary.
759 	 * in6_clearscope will touch the addresses only when necessary.
760 	 */
761 	in6_clearscope(&ip6->ip6_src);
762 	in6_clearscope(&ip6->ip6_dst);
763 
764 	/*
765 	 * If the outgoing packet contains a hop-by-hop options header,
766 	 * it must be examined and processed even by the source node.
767 	 * (RFC 2460, section 4.)
768 	 */
769 	if (exthdrs.ip6e_hbh) {
770 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
771 		u_int32_t dummy; /* XXX unused */
772 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
773 
774 #ifdef DIAGNOSTIC
775 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
776 			panic("ip6e_hbh is not contiguous");
777 #endif
778 		/*
779 		 *  XXX: if we have to send an ICMPv6 error to the sender,
780 		 *       we need the M_LOOP flag since icmp6_error() expects
781 		 *       the IPv6 and the hop-by-hop options header are
782 		 *       contiguous unless the flag is set.
783 		 */
784 		m->m_flags |= M_LOOP;
785 		m->m_pkthdr.rcvif = ifp;
786 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
787 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
788 		    &dummy, &plen) < 0) {
789 			/* m was already freed at this point */
790 			error = EINVAL;/* better error? */
791 			goto done;
792 		}
793 		m->m_flags &= ~M_LOOP; /* XXX */
794 		m->m_pkthdr.rcvif = NULL;
795 	}
796 
797 	/* Jump over all PFIL processing if hooks are not active. */
798 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
799 		goto passout;
800 
801 	odst = ip6->ip6_dst;
802 	/* Run through list of hooks for output packets. */
803 	error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
804 	if (error != 0 || m == NULL)
805 		goto done;
806 	ip6 = mtod(m, struct ip6_hdr *);
807 
808 	needfiblookup = 0;
809 	/* See if destination IP address was changed by packet filter. */
810 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
811 		m->m_flags |= M_SKIP_FIREWALL;
812 		/* If destination is now ourself drop to ip6_input(). */
813 		if (in6_localip(&ip6->ip6_dst)) {
814 			m->m_flags |= M_FASTFWD_OURS;
815 			if (m->m_pkthdr.rcvif == NULL)
816 				m->m_pkthdr.rcvif = V_loif;
817 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
818 				m->m_pkthdr.csum_flags |=
819 				    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
820 				m->m_pkthdr.csum_data = 0xffff;
821 			}
822 #ifdef SCTP
823 			if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
824 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
825 #endif
826 			error = netisr_queue(NETISR_IPV6, m);
827 			goto done;
828 		} else {
829 			RO_RTFREE(ro);
830 			needfiblookup = 1; /* Redo the routing table lookup. */
831 		}
832 	}
833 	/* See if fib was changed by packet filter. */
834 	if (fibnum != M_GETFIB(m)) {
835 		m->m_flags |= M_SKIP_FIREWALL;
836 		fibnum = M_GETFIB(m);
837 		RO_RTFREE(ro);
838 		needfiblookup = 1;
839 	}
840 	if (needfiblookup)
841 		goto again;
842 
843 	/* See if local, if yes, send it to netisr. */
844 	if (m->m_flags & M_FASTFWD_OURS) {
845 		if (m->m_pkthdr.rcvif == NULL)
846 			m->m_pkthdr.rcvif = V_loif;
847 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
848 			m->m_pkthdr.csum_flags |=
849 			    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
850 			m->m_pkthdr.csum_data = 0xffff;
851 		}
852 #ifdef SCTP
853 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
854 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
855 #endif
856 		error = netisr_queue(NETISR_IPV6, m);
857 		goto done;
858 	}
859 	/* Or forward to some other address? */
860 	if ((m->m_flags & M_IP6_NEXTHOP) &&
861 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
862 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
863 		bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
864 		m->m_flags |= M_SKIP_FIREWALL;
865 		m->m_flags &= ~M_IP6_NEXTHOP;
866 		m_tag_delete(m, fwd_tag);
867 		goto again;
868 	}
869 
870 passout:
871 	/*
872 	 * Send the packet to the outgoing interface.
873 	 * If necessary, do IPv6 fragmentation before sending.
874 	 *
875 	 * the logic here is rather complex:
876 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
877 	 * 1-a:	send as is if tlen <= path mtu
878 	 * 1-b:	fragment if tlen > path mtu
879 	 *
880 	 * 2: if user asks us not to fragment (dontfrag == 1)
881 	 * 2-a:	send as is if tlen <= interface mtu
882 	 * 2-b:	error if tlen > interface mtu
883 	 *
884 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
885 	 *	always fragment
886 	 *
887 	 * 4: if dontfrag == 1 && alwaysfrag == 1
888 	 *	error, as we cannot handle this conflicting request
889 	 */
890 	sw_csum = m->m_pkthdr.csum_flags;
891 	if (!hdrsplit) {
892 		tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
893 		sw_csum &= ~ifp->if_hwassist;
894 	} else
895 		tso = 0;
896 	/*
897 	 * If we added extension headers, we will not do TSO and calculate the
898 	 * checksums ourselves for now.
899 	 * XXX-BZ  Need a framework to know when the NIC can handle it, even
900 	 * with ext. hdrs.
901 	 */
902 	if (sw_csum & CSUM_DELAY_DATA_IPV6) {
903 		sw_csum &= ~CSUM_DELAY_DATA_IPV6;
904 		in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
905 	}
906 #ifdef SCTP
907 	if (sw_csum & CSUM_SCTP_IPV6) {
908 		sw_csum &= ~CSUM_SCTP_IPV6;
909 		sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
910 	}
911 #endif
912 	m->m_pkthdr.csum_flags &= ifp->if_hwassist;
913 	tlen = m->m_pkthdr.len;
914 
915 	if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
916 		dontfrag = 1;
917 	else
918 		dontfrag = 0;
919 	if (dontfrag && alwaysfrag) {	/* case 4 */
920 		/* conflicting request - can't transmit */
921 		error = EMSGSIZE;
922 		goto bad;
923 	}
924 	if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {	/* case 2-b */
925 		/*
926 		 * Even if the DONTFRAG option is specified, we cannot send the
927 		 * packet when the data length is larger than the MTU of the
928 		 * outgoing interface.
929 		 * Notify the error by sending IPV6_PATHMTU ancillary data if
930 		 * application wanted to know the MTU value. Also return an
931 		 * error code (this is not described in the API spec).
932 		 */
933 		if (inp != NULL)
934 			ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
935 		error = EMSGSIZE;
936 		goto bad;
937 	}
938 
939 	/*
940 	 * transmit packet without fragmentation
941 	 */
942 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
943 		struct in6_ifaddr *ia6;
944 
945 		ip6 = mtod(m, struct ip6_hdr *);
946 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
947 		if (ia6) {
948 			/* Record statistics for this interface address. */
949 			counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
950 			counter_u64_add(ia6->ia_ifa.ifa_obytes,
951 			    m->m_pkthdr.len);
952 			ifa_free(&ia6->ia_ifa);
953 		}
954 		error = nd6_output_ifp(ifp, origifp, m, dst,
955 		    (struct route *)ro);
956 		goto done;
957 	}
958 
959 	/*
960 	 * try to fragment the packet.  case 1-b and 3
961 	 */
962 	if (mtu < IPV6_MMTU) {
963 		/* path MTU cannot be less than IPV6_MMTU */
964 		error = EMSGSIZE;
965 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
966 		goto bad;
967 	} else if (ip6->ip6_plen == 0) {
968 		/* jumbo payload cannot be fragmented */
969 		error = EMSGSIZE;
970 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
971 		goto bad;
972 	} else {
973 		u_char nextproto;
974 
975 		/*
976 		 * Too large for the destination or interface;
977 		 * fragment if possible.
978 		 * Must be able to put at least 8 bytes per fragment.
979 		 */
980 		hlen = unfragpartlen;
981 		if (mtu > IPV6_MAXPACKET)
982 			mtu = IPV6_MAXPACKET;
983 
984 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
985 		if (len < 8) {
986 			error = EMSGSIZE;
987 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
988 			goto bad;
989 		}
990 
991 		/*
992 		 * If the interface will not calculate checksums on
993 		 * fragmented packets, then do it here.
994 		 * XXX-BZ handle the hw offloading case.  Need flags.
995 		 */
996 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
997 			in6_delayed_cksum(m, plen, hlen);
998 			m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
999 		}
1000 #ifdef SCTP
1001 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1002 			sctp_delayed_cksum(m, hlen);
1003 			m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1004 		}
1005 #endif
1006 		/*
1007 		 * Change the next header field of the last header in the
1008 		 * unfragmentable part.
1009 		 */
1010 		if (exthdrs.ip6e_rthdr) {
1011 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1012 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1013 		} else if (exthdrs.ip6e_dest1) {
1014 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1015 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1016 		} else if (exthdrs.ip6e_hbh) {
1017 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1018 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1019 		} else {
1020 			nextproto = ip6->ip6_nxt;
1021 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1022 		}
1023 
1024 		/*
1025 		 * Loop through length of segment after first fragment,
1026 		 * make new header and copy data of each part and link onto
1027 		 * chain.
1028 		 */
1029 		m0 = m;
1030 		id = htonl(ip6_randomid());
1031 		if ((error = ip6_fragment(ifp, m, hlen, nextproto, len, id)))
1032 			goto sendorfree;
1033 
1034 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1035 	}
1036 
1037 	/*
1038 	 * Remove leading garbages.
1039 	 */
1040 sendorfree:
1041 	m = m0->m_nextpkt;
1042 	m0->m_nextpkt = 0;
1043 	m_freem(m0);
1044 	for (m0 = m; m; m = m0) {
1045 		m0 = m->m_nextpkt;
1046 		m->m_nextpkt = 0;
1047 		if (error == 0) {
1048 			/* Record statistics for this interface address. */
1049 			if (ia) {
1050 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1051 				counter_u64_add(ia->ia_ifa.ifa_obytes,
1052 				    m->m_pkthdr.len);
1053 			}
1054 			error = nd6_output_ifp(ifp, origifp, m, dst,
1055 			    (struct route *)ro);
1056 		} else
1057 			m_freem(m);
1058 	}
1059 
1060 	if (error == 0)
1061 		IP6STAT_INC(ip6s_fragmented);
1062 
1063 done:
1064 	if (ro == &ip6route)
1065 		RO_RTFREE(ro);
1066 	return (error);
1067 
1068 freehdrs:
1069 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1070 	m_freem(exthdrs.ip6e_dest1);
1071 	m_freem(exthdrs.ip6e_rthdr);
1072 	m_freem(exthdrs.ip6e_dest2);
1073 	/* FALLTHROUGH */
1074 bad:
1075 	if (m)
1076 		m_freem(m);
1077 	goto done;
1078 }
1079 
1080 static int
1081 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1082 {
1083 	struct mbuf *m;
1084 
1085 	if (hlen > MCLBYTES)
1086 		return (ENOBUFS); /* XXX */
1087 
1088 	if (hlen > MLEN)
1089 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1090 	else
1091 		m = m_get(M_NOWAIT, MT_DATA);
1092 	if (m == NULL)
1093 		return (ENOBUFS);
1094 	m->m_len = hlen;
1095 	if (hdr)
1096 		bcopy(hdr, mtod(m, caddr_t), hlen);
1097 
1098 	*mp = m;
1099 	return (0);
1100 }
1101 
1102 /*
1103  * Insert jumbo payload option.
1104  */
1105 static int
1106 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1107 {
1108 	struct mbuf *mopt;
1109 	u_char *optbuf;
1110 	u_int32_t v;
1111 
1112 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1113 
1114 	/*
1115 	 * If there is no hop-by-hop options header, allocate new one.
1116 	 * If there is one but it doesn't have enough space to store the
1117 	 * jumbo payload option, allocate a cluster to store the whole options.
1118 	 * Otherwise, use it to store the options.
1119 	 */
1120 	if (exthdrs->ip6e_hbh == NULL) {
1121 		mopt = m_get(M_NOWAIT, MT_DATA);
1122 		if (mopt == NULL)
1123 			return (ENOBUFS);
1124 		mopt->m_len = JUMBOOPTLEN;
1125 		optbuf = mtod(mopt, u_char *);
1126 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1127 		exthdrs->ip6e_hbh = mopt;
1128 	} else {
1129 		struct ip6_hbh *hbh;
1130 
1131 		mopt = exthdrs->ip6e_hbh;
1132 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1133 			/*
1134 			 * XXX assumption:
1135 			 * - exthdrs->ip6e_hbh is not referenced from places
1136 			 *   other than exthdrs.
1137 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1138 			 */
1139 			int oldoptlen = mopt->m_len;
1140 			struct mbuf *n;
1141 
1142 			/*
1143 			 * XXX: give up if the whole (new) hbh header does
1144 			 * not fit even in an mbuf cluster.
1145 			 */
1146 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1147 				return (ENOBUFS);
1148 
1149 			/*
1150 			 * As a consequence, we must always prepare a cluster
1151 			 * at this point.
1152 			 */
1153 			n = m_getcl(M_NOWAIT, MT_DATA, 0);
1154 			if (n == NULL)
1155 				return (ENOBUFS);
1156 			n->m_len = oldoptlen + JUMBOOPTLEN;
1157 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1158 			    oldoptlen);
1159 			optbuf = mtod(n, caddr_t) + oldoptlen;
1160 			m_freem(mopt);
1161 			mopt = exthdrs->ip6e_hbh = n;
1162 		} else {
1163 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1164 			mopt->m_len += JUMBOOPTLEN;
1165 		}
1166 		optbuf[0] = IP6OPT_PADN;
1167 		optbuf[1] = 1;
1168 
1169 		/*
1170 		 * Adjust the header length according to the pad and
1171 		 * the jumbo payload option.
1172 		 */
1173 		hbh = mtod(mopt, struct ip6_hbh *);
1174 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1175 	}
1176 
1177 	/* fill in the option. */
1178 	optbuf[2] = IP6OPT_JUMBO;
1179 	optbuf[3] = 4;
1180 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1181 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1182 
1183 	/* finally, adjust the packet header length */
1184 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1185 
1186 	return (0);
1187 #undef JUMBOOPTLEN
1188 }
1189 
1190 /*
1191  * Insert fragment header and copy unfragmentable header portions.
1192  */
1193 static int
1194 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1195     struct ip6_frag **frghdrp)
1196 {
1197 	struct mbuf *n, *mlast;
1198 
1199 	if (hlen > sizeof(struct ip6_hdr)) {
1200 		n = m_copym(m0, sizeof(struct ip6_hdr),
1201 		    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1202 		if (n == NULL)
1203 			return (ENOBUFS);
1204 		m->m_next = n;
1205 	} else
1206 		n = m;
1207 
1208 	/* Search for the last mbuf of unfragmentable part. */
1209 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1210 		;
1211 
1212 	if (M_WRITABLE(mlast) &&
1213 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1214 		/* use the trailing space of the last mbuf for the fragment hdr */
1215 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1216 		    mlast->m_len);
1217 		mlast->m_len += sizeof(struct ip6_frag);
1218 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1219 	} else {
1220 		/* allocate a new mbuf for the fragment header */
1221 		struct mbuf *mfrg;
1222 
1223 		mfrg = m_get(M_NOWAIT, MT_DATA);
1224 		if (mfrg == NULL)
1225 			return (ENOBUFS);
1226 		mfrg->m_len = sizeof(struct ip6_frag);
1227 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1228 		mlast->m_next = mfrg;
1229 	}
1230 
1231 	return (0);
1232 }
1233 
1234 /*
1235  * Calculates IPv6 path mtu for destination @dst.
1236  * Resulting MTU is stored in @mtup.
1237  *
1238  * Returns 0 on success.
1239  */
1240 static int
1241 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1242 {
1243 	struct nhop6_extended nh6;
1244 	struct in6_addr kdst;
1245 	uint32_t scopeid;
1246 	struct ifnet *ifp;
1247 	u_long mtu;
1248 	int error;
1249 
1250 	in6_splitscope(dst, &kdst, &scopeid);
1251 	if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0)
1252 		return (EHOSTUNREACH);
1253 
1254 	ifp = nh6.nh_ifp;
1255 	mtu = nh6.nh_mtu;
1256 
1257 	error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL);
1258 	fib6_free_nh_ext(fibnum, &nh6);
1259 
1260 	return (error);
1261 }
1262 
1263 /*
1264  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1265  * and cached data in @ro_pmtu.
1266  * MTU from (successful) route lookup is saved (along with dst)
1267  * inside @ro_pmtu to avoid subsequent route lookups after packet
1268  * filter processing.
1269  *
1270  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1271  * Returns 0 on success.
1272  */
1273 static int
1274 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1275     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1276     int *alwaysfragp, u_int fibnum)
1277 {
1278 	struct nhop6_basic nh6;
1279 	struct in6_addr kdst;
1280 	uint32_t scopeid;
1281 	struct sockaddr_in6 *sa6_dst;
1282 	u_long mtu;
1283 
1284 	mtu = 0;
1285 	if (do_lookup) {
1286 
1287 		/*
1288 		 * Here ro_pmtu has final destination address, while
1289 		 * ro might represent immediate destination.
1290 		 * Use ro_pmtu destination since mtu might differ.
1291 		 */
1292 		sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1293 		if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1294 			ro_pmtu->ro_mtu = 0;
1295 
1296 		if (ro_pmtu->ro_mtu == 0) {
1297 			bzero(sa6_dst, sizeof(*sa6_dst));
1298 			sa6_dst->sin6_family = AF_INET6;
1299 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1300 			sa6_dst->sin6_addr = *dst;
1301 
1302 			in6_splitscope(dst, &kdst, &scopeid);
1303 			if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0,
1304 			    &nh6) == 0)
1305 				ro_pmtu->ro_mtu = nh6.nh_mtu;
1306 		}
1307 
1308 		mtu = ro_pmtu->ro_mtu;
1309 	}
1310 
1311 	if (ro_pmtu->ro_rt)
1312 		mtu = ro_pmtu->ro_rt->rt_mtu;
1313 
1314 	return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp));
1315 }
1316 
1317 /*
1318  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1319  * hostcache data for @dst.
1320  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1321  *
1322  * Returns 0 on success.
1323  */
1324 static int
1325 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1326     u_long *mtup, int *alwaysfragp)
1327 {
1328 	u_long mtu = 0;
1329 	int alwaysfrag = 0;
1330 	int error = 0;
1331 
1332 	if (rt_mtu > 0) {
1333 		u_int32_t ifmtu;
1334 		struct in_conninfo inc;
1335 
1336 		bzero(&inc, sizeof(inc));
1337 		inc.inc_flags |= INC_ISIPV6;
1338 		inc.inc6_faddr = *dst;
1339 
1340 		ifmtu = IN6_LINKMTU(ifp);
1341 		mtu = tcp_hc_getmtu(&inc);
1342 		if (mtu)
1343 			mtu = min(mtu, rt_mtu);
1344 		else
1345 			mtu = rt_mtu;
1346 		if (mtu == 0)
1347 			mtu = ifmtu;
1348 		else if (mtu < IPV6_MMTU) {
1349 			/*
1350 			 * RFC2460 section 5, last paragraph:
1351 			 * if we record ICMPv6 too big message with
1352 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1353 			 * or smaller, with framgent header attached.
1354 			 * (fragment header is needed regardless from the
1355 			 * packet size, for translators to identify packets)
1356 			 */
1357 			alwaysfrag = 1;
1358 			mtu = IPV6_MMTU;
1359 		}
1360 	} else if (ifp) {
1361 		mtu = IN6_LINKMTU(ifp);
1362 	} else
1363 		error = EHOSTUNREACH; /* XXX */
1364 
1365 	*mtup = mtu;
1366 	if (alwaysfragp)
1367 		*alwaysfragp = alwaysfrag;
1368 	return (error);
1369 }
1370 
1371 /*
1372  * IP6 socket option processing.
1373  */
1374 int
1375 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1376 {
1377 	int optdatalen, uproto;
1378 	void *optdata;
1379 	struct inpcb *in6p = sotoinpcb(so);
1380 	int error, optval;
1381 	int level, op, optname;
1382 	int optlen;
1383 	struct thread *td;
1384 #ifdef	RSS
1385 	uint32_t rss_bucket;
1386 	int retval;
1387 #endif
1388 
1389 	level = sopt->sopt_level;
1390 	op = sopt->sopt_dir;
1391 	optname = sopt->sopt_name;
1392 	optlen = sopt->sopt_valsize;
1393 	td = sopt->sopt_td;
1394 	error = 0;
1395 	optval = 0;
1396 	uproto = (int)so->so_proto->pr_protocol;
1397 
1398 	if (level != IPPROTO_IPV6) {
1399 		error = EINVAL;
1400 
1401 		if (sopt->sopt_level == SOL_SOCKET &&
1402 		    sopt->sopt_dir == SOPT_SET) {
1403 			switch (sopt->sopt_name) {
1404 			case SO_REUSEADDR:
1405 				INP_WLOCK(in6p);
1406 				if ((so->so_options & SO_REUSEADDR) != 0)
1407 					in6p->inp_flags2 |= INP_REUSEADDR;
1408 				else
1409 					in6p->inp_flags2 &= ~INP_REUSEADDR;
1410 				INP_WUNLOCK(in6p);
1411 				error = 0;
1412 				break;
1413 			case SO_REUSEPORT:
1414 				INP_WLOCK(in6p);
1415 				if ((so->so_options & SO_REUSEPORT) != 0)
1416 					in6p->inp_flags2 |= INP_REUSEPORT;
1417 				else
1418 					in6p->inp_flags2 &= ~INP_REUSEPORT;
1419 				INP_WUNLOCK(in6p);
1420 				error = 0;
1421 				break;
1422 			case SO_SETFIB:
1423 				INP_WLOCK(in6p);
1424 				in6p->inp_inc.inc_fibnum = so->so_fibnum;
1425 				INP_WUNLOCK(in6p);
1426 				error = 0;
1427 				break;
1428 			default:
1429 				break;
1430 			}
1431 		}
1432 	} else {		/* level == IPPROTO_IPV6 */
1433 		switch (op) {
1434 
1435 		case SOPT_SET:
1436 			switch (optname) {
1437 			case IPV6_2292PKTOPTIONS:
1438 #ifdef IPV6_PKTOPTIONS
1439 			case IPV6_PKTOPTIONS:
1440 #endif
1441 			{
1442 				struct mbuf *m;
1443 
1444 				error = soopt_getm(sopt, &m); /* XXX */
1445 				if (error != 0)
1446 					break;
1447 				error = soopt_mcopyin(sopt, m); /* XXX */
1448 				if (error != 0)
1449 					break;
1450 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1451 						    m, so, sopt);
1452 				m_freem(m); /* XXX */
1453 				break;
1454 			}
1455 
1456 			/*
1457 			 * Use of some Hop-by-Hop options or some
1458 			 * Destination options, might require special
1459 			 * privilege.  That is, normal applications
1460 			 * (without special privilege) might be forbidden
1461 			 * from setting certain options in outgoing packets,
1462 			 * and might never see certain options in received
1463 			 * packets. [RFC 2292 Section 6]
1464 			 * KAME specific note:
1465 			 *  KAME prevents non-privileged users from sending or
1466 			 *  receiving ANY hbh/dst options in order to avoid
1467 			 *  overhead of parsing options in the kernel.
1468 			 */
1469 			case IPV6_RECVHOPOPTS:
1470 			case IPV6_RECVDSTOPTS:
1471 			case IPV6_RECVRTHDRDSTOPTS:
1472 				if (td != NULL) {
1473 					error = priv_check(td,
1474 					    PRIV_NETINET_SETHDROPTS);
1475 					if (error)
1476 						break;
1477 				}
1478 				/* FALLTHROUGH */
1479 			case IPV6_UNICAST_HOPS:
1480 			case IPV6_HOPLIMIT:
1481 
1482 			case IPV6_RECVPKTINFO:
1483 			case IPV6_RECVHOPLIMIT:
1484 			case IPV6_RECVRTHDR:
1485 			case IPV6_RECVPATHMTU:
1486 			case IPV6_RECVTCLASS:
1487 			case IPV6_RECVFLOWID:
1488 #ifdef	RSS
1489 			case IPV6_RECVRSSBUCKETID:
1490 #endif
1491 			case IPV6_V6ONLY:
1492 			case IPV6_AUTOFLOWLABEL:
1493 			case IPV6_BINDANY:
1494 			case IPV6_BINDMULTI:
1495 #ifdef	RSS
1496 			case IPV6_RSS_LISTEN_BUCKET:
1497 #endif
1498 				if (optname == IPV6_BINDANY && td != NULL) {
1499 					error = priv_check(td,
1500 					    PRIV_NETINET_BINDANY);
1501 					if (error)
1502 						break;
1503 				}
1504 
1505 				if (optlen != sizeof(int)) {
1506 					error = EINVAL;
1507 					break;
1508 				}
1509 				error = sooptcopyin(sopt, &optval,
1510 					sizeof optval, sizeof optval);
1511 				if (error)
1512 					break;
1513 				switch (optname) {
1514 
1515 				case IPV6_UNICAST_HOPS:
1516 					if (optval < -1 || optval >= 256)
1517 						error = EINVAL;
1518 					else {
1519 						/* -1 = kernel default */
1520 						in6p->in6p_hops = optval;
1521 						if ((in6p->inp_vflag &
1522 						     INP_IPV4) != 0)
1523 							in6p->inp_ip_ttl = optval;
1524 					}
1525 					break;
1526 #define OPTSET(bit) \
1527 do { \
1528 	INP_WLOCK(in6p); \
1529 	if (optval) \
1530 		in6p->inp_flags |= (bit); \
1531 	else \
1532 		in6p->inp_flags &= ~(bit); \
1533 	INP_WUNLOCK(in6p); \
1534 } while (/*CONSTCOND*/ 0)
1535 #define OPTSET2292(bit) \
1536 do { \
1537 	INP_WLOCK(in6p); \
1538 	in6p->inp_flags |= IN6P_RFC2292; \
1539 	if (optval) \
1540 		in6p->inp_flags |= (bit); \
1541 	else \
1542 		in6p->inp_flags &= ~(bit); \
1543 	INP_WUNLOCK(in6p); \
1544 } while (/*CONSTCOND*/ 0)
1545 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1546 
1547 #define OPTSET2(bit, val) do {						\
1548 	INP_WLOCK(in6p);						\
1549 	if (val)							\
1550 		in6p->inp_flags2 |= bit;				\
1551 	else								\
1552 		in6p->inp_flags2 &= ~bit;				\
1553 	INP_WUNLOCK(in6p);						\
1554 } while (0)
1555 #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0)
1556 
1557 				case IPV6_RECVPKTINFO:
1558 					/* cannot mix with RFC2292 */
1559 					if (OPTBIT(IN6P_RFC2292)) {
1560 						error = EINVAL;
1561 						break;
1562 					}
1563 					OPTSET(IN6P_PKTINFO);
1564 					break;
1565 
1566 				case IPV6_HOPLIMIT:
1567 				{
1568 					struct ip6_pktopts **optp;
1569 
1570 					/* cannot mix with RFC2292 */
1571 					if (OPTBIT(IN6P_RFC2292)) {
1572 						error = EINVAL;
1573 						break;
1574 					}
1575 					optp = &in6p->in6p_outputopts;
1576 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1577 					    (u_char *)&optval, sizeof(optval),
1578 					    optp, (td != NULL) ? td->td_ucred :
1579 					    NULL, uproto);
1580 					break;
1581 				}
1582 
1583 				case IPV6_RECVHOPLIMIT:
1584 					/* cannot mix with RFC2292 */
1585 					if (OPTBIT(IN6P_RFC2292)) {
1586 						error = EINVAL;
1587 						break;
1588 					}
1589 					OPTSET(IN6P_HOPLIMIT);
1590 					break;
1591 
1592 				case IPV6_RECVHOPOPTS:
1593 					/* cannot mix with RFC2292 */
1594 					if (OPTBIT(IN6P_RFC2292)) {
1595 						error = EINVAL;
1596 						break;
1597 					}
1598 					OPTSET(IN6P_HOPOPTS);
1599 					break;
1600 
1601 				case IPV6_RECVDSTOPTS:
1602 					/* cannot mix with RFC2292 */
1603 					if (OPTBIT(IN6P_RFC2292)) {
1604 						error = EINVAL;
1605 						break;
1606 					}
1607 					OPTSET(IN6P_DSTOPTS);
1608 					break;
1609 
1610 				case IPV6_RECVRTHDRDSTOPTS:
1611 					/* cannot mix with RFC2292 */
1612 					if (OPTBIT(IN6P_RFC2292)) {
1613 						error = EINVAL;
1614 						break;
1615 					}
1616 					OPTSET(IN6P_RTHDRDSTOPTS);
1617 					break;
1618 
1619 				case IPV6_RECVRTHDR:
1620 					/* cannot mix with RFC2292 */
1621 					if (OPTBIT(IN6P_RFC2292)) {
1622 						error = EINVAL;
1623 						break;
1624 					}
1625 					OPTSET(IN6P_RTHDR);
1626 					break;
1627 
1628 				case IPV6_RECVPATHMTU:
1629 					/*
1630 					 * We ignore this option for TCP
1631 					 * sockets.
1632 					 * (RFC3542 leaves this case
1633 					 * unspecified.)
1634 					 */
1635 					if (uproto != IPPROTO_TCP)
1636 						OPTSET(IN6P_MTU);
1637 					break;
1638 
1639 				case IPV6_RECVFLOWID:
1640 					OPTSET2(INP_RECVFLOWID, optval);
1641 					break;
1642 
1643 #ifdef	RSS
1644 				case IPV6_RECVRSSBUCKETID:
1645 					OPTSET2(INP_RECVRSSBUCKETID, optval);
1646 					break;
1647 #endif
1648 
1649 				case IPV6_V6ONLY:
1650 					/*
1651 					 * make setsockopt(IPV6_V6ONLY)
1652 					 * available only prior to bind(2).
1653 					 * see ipng mailing list, Jun 22 2001.
1654 					 */
1655 					if (in6p->inp_lport ||
1656 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1657 						error = EINVAL;
1658 						break;
1659 					}
1660 					OPTSET(IN6P_IPV6_V6ONLY);
1661 					if (optval)
1662 						in6p->inp_vflag &= ~INP_IPV4;
1663 					else
1664 						in6p->inp_vflag |= INP_IPV4;
1665 					break;
1666 				case IPV6_RECVTCLASS:
1667 					/* cannot mix with RFC2292 XXX */
1668 					if (OPTBIT(IN6P_RFC2292)) {
1669 						error = EINVAL;
1670 						break;
1671 					}
1672 					OPTSET(IN6P_TCLASS);
1673 					break;
1674 				case IPV6_AUTOFLOWLABEL:
1675 					OPTSET(IN6P_AUTOFLOWLABEL);
1676 					break;
1677 
1678 				case IPV6_BINDANY:
1679 					OPTSET(INP_BINDANY);
1680 					break;
1681 
1682 				case IPV6_BINDMULTI:
1683 					OPTSET2(INP_BINDMULTI, optval);
1684 					break;
1685 #ifdef	RSS
1686 				case IPV6_RSS_LISTEN_BUCKET:
1687 					if ((optval >= 0) &&
1688 					    (optval < rss_getnumbuckets())) {
1689 						in6p->inp_rss_listen_bucket = optval;
1690 						OPTSET2(INP_RSS_BUCKET_SET, 1);
1691 					} else {
1692 						error = EINVAL;
1693 					}
1694 					break;
1695 #endif
1696 				}
1697 				break;
1698 
1699 			case IPV6_TCLASS:
1700 			case IPV6_DONTFRAG:
1701 			case IPV6_USE_MIN_MTU:
1702 			case IPV6_PREFER_TEMPADDR:
1703 				if (optlen != sizeof(optval)) {
1704 					error = EINVAL;
1705 					break;
1706 				}
1707 				error = sooptcopyin(sopt, &optval,
1708 					sizeof optval, sizeof optval);
1709 				if (error)
1710 					break;
1711 				{
1712 					struct ip6_pktopts **optp;
1713 					optp = &in6p->in6p_outputopts;
1714 					error = ip6_pcbopt(optname,
1715 					    (u_char *)&optval, sizeof(optval),
1716 					    optp, (td != NULL) ? td->td_ucred :
1717 					    NULL, uproto);
1718 					break;
1719 				}
1720 
1721 			case IPV6_2292PKTINFO:
1722 			case IPV6_2292HOPLIMIT:
1723 			case IPV6_2292HOPOPTS:
1724 			case IPV6_2292DSTOPTS:
1725 			case IPV6_2292RTHDR:
1726 				/* RFC 2292 */
1727 				if (optlen != sizeof(int)) {
1728 					error = EINVAL;
1729 					break;
1730 				}
1731 				error = sooptcopyin(sopt, &optval,
1732 					sizeof optval, sizeof optval);
1733 				if (error)
1734 					break;
1735 				switch (optname) {
1736 				case IPV6_2292PKTINFO:
1737 					OPTSET2292(IN6P_PKTINFO);
1738 					break;
1739 				case IPV6_2292HOPLIMIT:
1740 					OPTSET2292(IN6P_HOPLIMIT);
1741 					break;
1742 				case IPV6_2292HOPOPTS:
1743 					/*
1744 					 * Check super-user privilege.
1745 					 * See comments for IPV6_RECVHOPOPTS.
1746 					 */
1747 					if (td != NULL) {
1748 						error = priv_check(td,
1749 						    PRIV_NETINET_SETHDROPTS);
1750 						if (error)
1751 							return (error);
1752 					}
1753 					OPTSET2292(IN6P_HOPOPTS);
1754 					break;
1755 				case IPV6_2292DSTOPTS:
1756 					if (td != NULL) {
1757 						error = priv_check(td,
1758 						    PRIV_NETINET_SETHDROPTS);
1759 						if (error)
1760 							return (error);
1761 					}
1762 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1763 					break;
1764 				case IPV6_2292RTHDR:
1765 					OPTSET2292(IN6P_RTHDR);
1766 					break;
1767 				}
1768 				break;
1769 			case IPV6_PKTINFO:
1770 			case IPV6_HOPOPTS:
1771 			case IPV6_RTHDR:
1772 			case IPV6_DSTOPTS:
1773 			case IPV6_RTHDRDSTOPTS:
1774 			case IPV6_NEXTHOP:
1775 			{
1776 				/* new advanced API (RFC3542) */
1777 				u_char *optbuf;
1778 				u_char optbuf_storage[MCLBYTES];
1779 				int optlen;
1780 				struct ip6_pktopts **optp;
1781 
1782 				/* cannot mix with RFC2292 */
1783 				if (OPTBIT(IN6P_RFC2292)) {
1784 					error = EINVAL;
1785 					break;
1786 				}
1787 
1788 				/*
1789 				 * We only ensure valsize is not too large
1790 				 * here.  Further validation will be done
1791 				 * later.
1792 				 */
1793 				error = sooptcopyin(sopt, optbuf_storage,
1794 				    sizeof(optbuf_storage), 0);
1795 				if (error)
1796 					break;
1797 				optlen = sopt->sopt_valsize;
1798 				optbuf = optbuf_storage;
1799 				optp = &in6p->in6p_outputopts;
1800 				error = ip6_pcbopt(optname, optbuf, optlen,
1801 				    optp, (td != NULL) ? td->td_ucred : NULL,
1802 				    uproto);
1803 				break;
1804 			}
1805 #undef OPTSET
1806 
1807 			case IPV6_MULTICAST_IF:
1808 			case IPV6_MULTICAST_HOPS:
1809 			case IPV6_MULTICAST_LOOP:
1810 			case IPV6_JOIN_GROUP:
1811 			case IPV6_LEAVE_GROUP:
1812 			case IPV6_MSFILTER:
1813 			case MCAST_BLOCK_SOURCE:
1814 			case MCAST_UNBLOCK_SOURCE:
1815 			case MCAST_JOIN_GROUP:
1816 			case MCAST_LEAVE_GROUP:
1817 			case MCAST_JOIN_SOURCE_GROUP:
1818 			case MCAST_LEAVE_SOURCE_GROUP:
1819 				error = ip6_setmoptions(in6p, sopt);
1820 				break;
1821 
1822 			case IPV6_PORTRANGE:
1823 				error = sooptcopyin(sopt, &optval,
1824 				    sizeof optval, sizeof optval);
1825 				if (error)
1826 					break;
1827 
1828 				INP_WLOCK(in6p);
1829 				switch (optval) {
1830 				case IPV6_PORTRANGE_DEFAULT:
1831 					in6p->inp_flags &= ~(INP_LOWPORT);
1832 					in6p->inp_flags &= ~(INP_HIGHPORT);
1833 					break;
1834 
1835 				case IPV6_PORTRANGE_HIGH:
1836 					in6p->inp_flags &= ~(INP_LOWPORT);
1837 					in6p->inp_flags |= INP_HIGHPORT;
1838 					break;
1839 
1840 				case IPV6_PORTRANGE_LOW:
1841 					in6p->inp_flags &= ~(INP_HIGHPORT);
1842 					in6p->inp_flags |= INP_LOWPORT;
1843 					break;
1844 
1845 				default:
1846 					error = EINVAL;
1847 					break;
1848 				}
1849 				INP_WUNLOCK(in6p);
1850 				break;
1851 
1852 #ifdef IPSEC
1853 			case IPV6_IPSEC_POLICY:
1854 			{
1855 				caddr_t req;
1856 				struct mbuf *m;
1857 
1858 				if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1859 					break;
1860 				if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1861 					break;
1862 				req = mtod(m, caddr_t);
1863 				error = ipsec_set_policy(in6p, optname, req,
1864 				    m->m_len, (sopt->sopt_td != NULL) ?
1865 				    sopt->sopt_td->td_ucred : NULL);
1866 				m_freem(m);
1867 				break;
1868 			}
1869 #endif /* IPSEC */
1870 
1871 			default:
1872 				error = ENOPROTOOPT;
1873 				break;
1874 			}
1875 			break;
1876 
1877 		case SOPT_GET:
1878 			switch (optname) {
1879 
1880 			case IPV6_2292PKTOPTIONS:
1881 #ifdef IPV6_PKTOPTIONS
1882 			case IPV6_PKTOPTIONS:
1883 #endif
1884 				/*
1885 				 * RFC3542 (effectively) deprecated the
1886 				 * semantics of the 2292-style pktoptions.
1887 				 * Since it was not reliable in nature (i.e.,
1888 				 * applications had to expect the lack of some
1889 				 * information after all), it would make sense
1890 				 * to simplify this part by always returning
1891 				 * empty data.
1892 				 */
1893 				sopt->sopt_valsize = 0;
1894 				break;
1895 
1896 			case IPV6_RECVHOPOPTS:
1897 			case IPV6_RECVDSTOPTS:
1898 			case IPV6_RECVRTHDRDSTOPTS:
1899 			case IPV6_UNICAST_HOPS:
1900 			case IPV6_RECVPKTINFO:
1901 			case IPV6_RECVHOPLIMIT:
1902 			case IPV6_RECVRTHDR:
1903 			case IPV6_RECVPATHMTU:
1904 
1905 			case IPV6_V6ONLY:
1906 			case IPV6_PORTRANGE:
1907 			case IPV6_RECVTCLASS:
1908 			case IPV6_AUTOFLOWLABEL:
1909 			case IPV6_BINDANY:
1910 			case IPV6_FLOWID:
1911 			case IPV6_FLOWTYPE:
1912 			case IPV6_RECVFLOWID:
1913 #ifdef	RSS
1914 			case IPV6_RSSBUCKETID:
1915 			case IPV6_RECVRSSBUCKETID:
1916 #endif
1917 			case IPV6_BINDMULTI:
1918 				switch (optname) {
1919 
1920 				case IPV6_RECVHOPOPTS:
1921 					optval = OPTBIT(IN6P_HOPOPTS);
1922 					break;
1923 
1924 				case IPV6_RECVDSTOPTS:
1925 					optval = OPTBIT(IN6P_DSTOPTS);
1926 					break;
1927 
1928 				case IPV6_RECVRTHDRDSTOPTS:
1929 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1930 					break;
1931 
1932 				case IPV6_UNICAST_HOPS:
1933 					optval = in6p->in6p_hops;
1934 					break;
1935 
1936 				case IPV6_RECVPKTINFO:
1937 					optval = OPTBIT(IN6P_PKTINFO);
1938 					break;
1939 
1940 				case IPV6_RECVHOPLIMIT:
1941 					optval = OPTBIT(IN6P_HOPLIMIT);
1942 					break;
1943 
1944 				case IPV6_RECVRTHDR:
1945 					optval = OPTBIT(IN6P_RTHDR);
1946 					break;
1947 
1948 				case IPV6_RECVPATHMTU:
1949 					optval = OPTBIT(IN6P_MTU);
1950 					break;
1951 
1952 				case IPV6_V6ONLY:
1953 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
1954 					break;
1955 
1956 				case IPV6_PORTRANGE:
1957 				    {
1958 					int flags;
1959 					flags = in6p->inp_flags;
1960 					if (flags & INP_HIGHPORT)
1961 						optval = IPV6_PORTRANGE_HIGH;
1962 					else if (flags & INP_LOWPORT)
1963 						optval = IPV6_PORTRANGE_LOW;
1964 					else
1965 						optval = 0;
1966 					break;
1967 				    }
1968 				case IPV6_RECVTCLASS:
1969 					optval = OPTBIT(IN6P_TCLASS);
1970 					break;
1971 
1972 				case IPV6_AUTOFLOWLABEL:
1973 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1974 					break;
1975 
1976 				case IPV6_BINDANY:
1977 					optval = OPTBIT(INP_BINDANY);
1978 					break;
1979 
1980 				case IPV6_FLOWID:
1981 					optval = in6p->inp_flowid;
1982 					break;
1983 
1984 				case IPV6_FLOWTYPE:
1985 					optval = in6p->inp_flowtype;
1986 					break;
1987 
1988 				case IPV6_RECVFLOWID:
1989 					optval = OPTBIT2(INP_RECVFLOWID);
1990 					break;
1991 #ifdef	RSS
1992 				case IPV6_RSSBUCKETID:
1993 					retval =
1994 					    rss_hash2bucket(in6p->inp_flowid,
1995 					    in6p->inp_flowtype,
1996 					    &rss_bucket);
1997 					if (retval == 0)
1998 						optval = rss_bucket;
1999 					else
2000 						error = EINVAL;
2001 					break;
2002 
2003 				case IPV6_RECVRSSBUCKETID:
2004 					optval = OPTBIT2(INP_RECVRSSBUCKETID);
2005 					break;
2006 #endif
2007 
2008 				case IPV6_BINDMULTI:
2009 					optval = OPTBIT2(INP_BINDMULTI);
2010 					break;
2011 
2012 				}
2013 				if (error)
2014 					break;
2015 				error = sooptcopyout(sopt, &optval,
2016 					sizeof optval);
2017 				break;
2018 
2019 			case IPV6_PATHMTU:
2020 			{
2021 				u_long pmtu = 0;
2022 				struct ip6_mtuinfo mtuinfo;
2023 
2024 				if (!(so->so_state & SS_ISCONNECTED))
2025 					return (ENOTCONN);
2026 				/*
2027 				 * XXX: we dot not consider the case of source
2028 				 * routing, or optional information to specify
2029 				 * the outgoing interface.
2030 				 */
2031 				error = ip6_getpmtu_ctl(so->so_fibnum,
2032 				    &in6p->in6p_faddr, &pmtu);
2033 				if (error)
2034 					break;
2035 				if (pmtu > IPV6_MAXPACKET)
2036 					pmtu = IPV6_MAXPACKET;
2037 
2038 				bzero(&mtuinfo, sizeof(mtuinfo));
2039 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2040 				optdata = (void *)&mtuinfo;
2041 				optdatalen = sizeof(mtuinfo);
2042 				error = sooptcopyout(sopt, optdata,
2043 				    optdatalen);
2044 				break;
2045 			}
2046 
2047 			case IPV6_2292PKTINFO:
2048 			case IPV6_2292HOPLIMIT:
2049 			case IPV6_2292HOPOPTS:
2050 			case IPV6_2292RTHDR:
2051 			case IPV6_2292DSTOPTS:
2052 				switch (optname) {
2053 				case IPV6_2292PKTINFO:
2054 					optval = OPTBIT(IN6P_PKTINFO);
2055 					break;
2056 				case IPV6_2292HOPLIMIT:
2057 					optval = OPTBIT(IN6P_HOPLIMIT);
2058 					break;
2059 				case IPV6_2292HOPOPTS:
2060 					optval = OPTBIT(IN6P_HOPOPTS);
2061 					break;
2062 				case IPV6_2292RTHDR:
2063 					optval = OPTBIT(IN6P_RTHDR);
2064 					break;
2065 				case IPV6_2292DSTOPTS:
2066 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2067 					break;
2068 				}
2069 				error = sooptcopyout(sopt, &optval,
2070 				    sizeof optval);
2071 				break;
2072 			case IPV6_PKTINFO:
2073 			case IPV6_HOPOPTS:
2074 			case IPV6_RTHDR:
2075 			case IPV6_DSTOPTS:
2076 			case IPV6_RTHDRDSTOPTS:
2077 			case IPV6_NEXTHOP:
2078 			case IPV6_TCLASS:
2079 			case IPV6_DONTFRAG:
2080 			case IPV6_USE_MIN_MTU:
2081 			case IPV6_PREFER_TEMPADDR:
2082 				error = ip6_getpcbopt(in6p->in6p_outputopts,
2083 				    optname, sopt);
2084 				break;
2085 
2086 			case IPV6_MULTICAST_IF:
2087 			case IPV6_MULTICAST_HOPS:
2088 			case IPV6_MULTICAST_LOOP:
2089 			case IPV6_MSFILTER:
2090 				error = ip6_getmoptions(in6p, sopt);
2091 				break;
2092 
2093 #ifdef IPSEC
2094 			case IPV6_IPSEC_POLICY:
2095 			  {
2096 				caddr_t req = NULL;
2097 				size_t len = 0;
2098 				struct mbuf *m = NULL;
2099 				struct mbuf **mp = &m;
2100 				size_t ovalsize = sopt->sopt_valsize;
2101 				caddr_t oval = (caddr_t)sopt->sopt_val;
2102 
2103 				error = soopt_getm(sopt, &m); /* XXX */
2104 				if (error != 0)
2105 					break;
2106 				error = soopt_mcopyin(sopt, m); /* XXX */
2107 				if (error != 0)
2108 					break;
2109 				sopt->sopt_valsize = ovalsize;
2110 				sopt->sopt_val = oval;
2111 				if (m) {
2112 					req = mtod(m, caddr_t);
2113 					len = m->m_len;
2114 				}
2115 				error = ipsec_get_policy(in6p, req, len, mp);
2116 				if (error == 0)
2117 					error = soopt_mcopyout(sopt, m); /* XXX */
2118 				if (error == 0 && m)
2119 					m_freem(m);
2120 				break;
2121 			  }
2122 #endif /* IPSEC */
2123 
2124 			default:
2125 				error = ENOPROTOOPT;
2126 				break;
2127 			}
2128 			break;
2129 		}
2130 	}
2131 	return (error);
2132 }
2133 
2134 int
2135 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2136 {
2137 	int error = 0, optval, optlen;
2138 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2139 	struct inpcb *in6p = sotoinpcb(so);
2140 	int level, op, optname;
2141 
2142 	level = sopt->sopt_level;
2143 	op = sopt->sopt_dir;
2144 	optname = sopt->sopt_name;
2145 	optlen = sopt->sopt_valsize;
2146 
2147 	if (level != IPPROTO_IPV6) {
2148 		return (EINVAL);
2149 	}
2150 
2151 	switch (optname) {
2152 	case IPV6_CHECKSUM:
2153 		/*
2154 		 * For ICMPv6 sockets, no modification allowed for checksum
2155 		 * offset, permit "no change" values to help existing apps.
2156 		 *
2157 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2158 		 * for an ICMPv6 socket will fail."
2159 		 * The current behavior does not meet RFC3542.
2160 		 */
2161 		switch (op) {
2162 		case SOPT_SET:
2163 			if (optlen != sizeof(int)) {
2164 				error = EINVAL;
2165 				break;
2166 			}
2167 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2168 					    sizeof(optval));
2169 			if (error)
2170 				break;
2171 			if ((optval % 2) != 0) {
2172 				/* the API assumes even offset values */
2173 				error = EINVAL;
2174 			} else if (so->so_proto->pr_protocol ==
2175 			    IPPROTO_ICMPV6) {
2176 				if (optval != icmp6off)
2177 					error = EINVAL;
2178 			} else
2179 				in6p->in6p_cksum = optval;
2180 			break;
2181 
2182 		case SOPT_GET:
2183 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2184 				optval = icmp6off;
2185 			else
2186 				optval = in6p->in6p_cksum;
2187 
2188 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2189 			break;
2190 
2191 		default:
2192 			error = EINVAL;
2193 			break;
2194 		}
2195 		break;
2196 
2197 	default:
2198 		error = ENOPROTOOPT;
2199 		break;
2200 	}
2201 
2202 	return (error);
2203 }
2204 
2205 /*
2206  * Set up IP6 options in pcb for insertion in output packets or
2207  * specifying behavior of outgoing packets.
2208  */
2209 static int
2210 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2211     struct socket *so, struct sockopt *sopt)
2212 {
2213 	struct ip6_pktopts *opt = *pktopt;
2214 	int error = 0;
2215 	struct thread *td = sopt->sopt_td;
2216 
2217 	/* turn off any old options. */
2218 	if (opt) {
2219 #ifdef DIAGNOSTIC
2220 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2221 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2222 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2223 			printf("ip6_pcbopts: all specified options are cleared.\n");
2224 #endif
2225 		ip6_clearpktopts(opt, -1);
2226 	} else
2227 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2228 	*pktopt = NULL;
2229 
2230 	if (!m || m->m_len == 0) {
2231 		/*
2232 		 * Only turning off any previous options, regardless of
2233 		 * whether the opt is just created or given.
2234 		 */
2235 		free(opt, M_IP6OPT);
2236 		return (0);
2237 	}
2238 
2239 	/*  set options specified by user. */
2240 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2241 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2242 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2243 		free(opt, M_IP6OPT);
2244 		return (error);
2245 	}
2246 	*pktopt = opt;
2247 	return (0);
2248 }
2249 
2250 /*
2251  * initialize ip6_pktopts.  beware that there are non-zero default values in
2252  * the struct.
2253  */
2254 void
2255 ip6_initpktopts(struct ip6_pktopts *opt)
2256 {
2257 
2258 	bzero(opt, sizeof(*opt));
2259 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2260 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2261 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2262 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2263 }
2264 
2265 static int
2266 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2267     struct ucred *cred, int uproto)
2268 {
2269 	struct ip6_pktopts *opt;
2270 
2271 	if (*pktopt == NULL) {
2272 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2273 		    M_WAITOK);
2274 		ip6_initpktopts(*pktopt);
2275 	}
2276 	opt = *pktopt;
2277 
2278 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2279 }
2280 
2281 static int
2282 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2283 {
2284 	void *optdata = NULL;
2285 	int optdatalen = 0;
2286 	struct ip6_ext *ip6e;
2287 	int error = 0;
2288 	struct in6_pktinfo null_pktinfo;
2289 	int deftclass = 0, on;
2290 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2291 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2292 
2293 	switch (optname) {
2294 	case IPV6_PKTINFO:
2295 		optdata = (void *)&null_pktinfo;
2296 		if (pktopt && pktopt->ip6po_pktinfo) {
2297 			bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2298 			    sizeof(null_pktinfo));
2299 			in6_clearscope(&null_pktinfo.ipi6_addr);
2300 		} else {
2301 			/* XXX: we don't have to do this every time... */
2302 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2303 		}
2304 		optdatalen = sizeof(struct in6_pktinfo);
2305 		break;
2306 	case IPV6_TCLASS:
2307 		if (pktopt && pktopt->ip6po_tclass >= 0)
2308 			optdata = (void *)&pktopt->ip6po_tclass;
2309 		else
2310 			optdata = (void *)&deftclass;
2311 		optdatalen = sizeof(int);
2312 		break;
2313 	case IPV6_HOPOPTS:
2314 		if (pktopt && pktopt->ip6po_hbh) {
2315 			optdata = (void *)pktopt->ip6po_hbh;
2316 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2317 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2318 		}
2319 		break;
2320 	case IPV6_RTHDR:
2321 		if (pktopt && pktopt->ip6po_rthdr) {
2322 			optdata = (void *)pktopt->ip6po_rthdr;
2323 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2324 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2325 		}
2326 		break;
2327 	case IPV6_RTHDRDSTOPTS:
2328 		if (pktopt && pktopt->ip6po_dest1) {
2329 			optdata = (void *)pktopt->ip6po_dest1;
2330 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2331 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2332 		}
2333 		break;
2334 	case IPV6_DSTOPTS:
2335 		if (pktopt && pktopt->ip6po_dest2) {
2336 			optdata = (void *)pktopt->ip6po_dest2;
2337 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2338 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2339 		}
2340 		break;
2341 	case IPV6_NEXTHOP:
2342 		if (pktopt && pktopt->ip6po_nexthop) {
2343 			optdata = (void *)pktopt->ip6po_nexthop;
2344 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2345 		}
2346 		break;
2347 	case IPV6_USE_MIN_MTU:
2348 		if (pktopt)
2349 			optdata = (void *)&pktopt->ip6po_minmtu;
2350 		else
2351 			optdata = (void *)&defminmtu;
2352 		optdatalen = sizeof(int);
2353 		break;
2354 	case IPV6_DONTFRAG:
2355 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2356 			on = 1;
2357 		else
2358 			on = 0;
2359 		optdata = (void *)&on;
2360 		optdatalen = sizeof(on);
2361 		break;
2362 	case IPV6_PREFER_TEMPADDR:
2363 		if (pktopt)
2364 			optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2365 		else
2366 			optdata = (void *)&defpreftemp;
2367 		optdatalen = sizeof(int);
2368 		break;
2369 	default:		/* should not happen */
2370 #ifdef DIAGNOSTIC
2371 		panic("ip6_getpcbopt: unexpected option\n");
2372 #endif
2373 		return (ENOPROTOOPT);
2374 	}
2375 
2376 	error = sooptcopyout(sopt, optdata, optdatalen);
2377 
2378 	return (error);
2379 }
2380 
2381 void
2382 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2383 {
2384 	if (pktopt == NULL)
2385 		return;
2386 
2387 	if (optname == -1 || optname == IPV6_PKTINFO) {
2388 		if (pktopt->ip6po_pktinfo)
2389 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2390 		pktopt->ip6po_pktinfo = NULL;
2391 	}
2392 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2393 		pktopt->ip6po_hlim = -1;
2394 	if (optname == -1 || optname == IPV6_TCLASS)
2395 		pktopt->ip6po_tclass = -1;
2396 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2397 		if (pktopt->ip6po_nextroute.ro_rt) {
2398 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2399 			pktopt->ip6po_nextroute.ro_rt = NULL;
2400 		}
2401 		if (pktopt->ip6po_nexthop)
2402 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2403 		pktopt->ip6po_nexthop = NULL;
2404 	}
2405 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2406 		if (pktopt->ip6po_hbh)
2407 			free(pktopt->ip6po_hbh, M_IP6OPT);
2408 		pktopt->ip6po_hbh = NULL;
2409 	}
2410 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2411 		if (pktopt->ip6po_dest1)
2412 			free(pktopt->ip6po_dest1, M_IP6OPT);
2413 		pktopt->ip6po_dest1 = NULL;
2414 	}
2415 	if (optname == -1 || optname == IPV6_RTHDR) {
2416 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2417 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2418 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2419 		if (pktopt->ip6po_route.ro_rt) {
2420 			RTFREE(pktopt->ip6po_route.ro_rt);
2421 			pktopt->ip6po_route.ro_rt = NULL;
2422 		}
2423 	}
2424 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2425 		if (pktopt->ip6po_dest2)
2426 			free(pktopt->ip6po_dest2, M_IP6OPT);
2427 		pktopt->ip6po_dest2 = NULL;
2428 	}
2429 }
2430 
2431 #define PKTOPT_EXTHDRCPY(type) \
2432 do {\
2433 	if (src->type) {\
2434 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2435 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2436 		if (dst->type == NULL && canwait == M_NOWAIT)\
2437 			goto bad;\
2438 		bcopy(src->type, dst->type, hlen);\
2439 	}\
2440 } while (/*CONSTCOND*/ 0)
2441 
2442 static int
2443 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2444 {
2445 	if (dst == NULL || src == NULL)  {
2446 		printf("ip6_clearpktopts: invalid argument\n");
2447 		return (EINVAL);
2448 	}
2449 
2450 	dst->ip6po_hlim = src->ip6po_hlim;
2451 	dst->ip6po_tclass = src->ip6po_tclass;
2452 	dst->ip6po_flags = src->ip6po_flags;
2453 	dst->ip6po_minmtu = src->ip6po_minmtu;
2454 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2455 	if (src->ip6po_pktinfo) {
2456 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2457 		    M_IP6OPT, canwait);
2458 		if (dst->ip6po_pktinfo == NULL)
2459 			goto bad;
2460 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2461 	}
2462 	if (src->ip6po_nexthop) {
2463 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2464 		    M_IP6OPT, canwait);
2465 		if (dst->ip6po_nexthop == NULL)
2466 			goto bad;
2467 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2468 		    src->ip6po_nexthop->sa_len);
2469 	}
2470 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2471 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2472 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2473 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2474 	return (0);
2475 
2476   bad:
2477 	ip6_clearpktopts(dst, -1);
2478 	return (ENOBUFS);
2479 }
2480 #undef PKTOPT_EXTHDRCPY
2481 
2482 struct ip6_pktopts *
2483 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2484 {
2485 	int error;
2486 	struct ip6_pktopts *dst;
2487 
2488 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2489 	if (dst == NULL)
2490 		return (NULL);
2491 	ip6_initpktopts(dst);
2492 
2493 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2494 		free(dst, M_IP6OPT);
2495 		return (NULL);
2496 	}
2497 
2498 	return (dst);
2499 }
2500 
2501 void
2502 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2503 {
2504 	if (pktopt == NULL)
2505 		return;
2506 
2507 	ip6_clearpktopts(pktopt, -1);
2508 
2509 	free(pktopt, M_IP6OPT);
2510 }
2511 
2512 /*
2513  * Set IPv6 outgoing packet options based on advanced API.
2514  */
2515 int
2516 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2517     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2518 {
2519 	struct cmsghdr *cm = NULL;
2520 
2521 	if (control == NULL || opt == NULL)
2522 		return (EINVAL);
2523 
2524 	ip6_initpktopts(opt);
2525 	if (stickyopt) {
2526 		int error;
2527 
2528 		/*
2529 		 * If stickyopt is provided, make a local copy of the options
2530 		 * for this particular packet, then override them by ancillary
2531 		 * objects.
2532 		 * XXX: copypktopts() does not copy the cached route to a next
2533 		 * hop (if any).  This is not very good in terms of efficiency,
2534 		 * but we can allow this since this option should be rarely
2535 		 * used.
2536 		 */
2537 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2538 			return (error);
2539 	}
2540 
2541 	/*
2542 	 * XXX: Currently, we assume all the optional information is stored
2543 	 * in a single mbuf.
2544 	 */
2545 	if (control->m_next)
2546 		return (EINVAL);
2547 
2548 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2549 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2550 		int error;
2551 
2552 		if (control->m_len < CMSG_LEN(0))
2553 			return (EINVAL);
2554 
2555 		cm = mtod(control, struct cmsghdr *);
2556 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2557 			return (EINVAL);
2558 		if (cm->cmsg_level != IPPROTO_IPV6)
2559 			continue;
2560 
2561 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2562 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2563 		if (error)
2564 			return (error);
2565 	}
2566 
2567 	return (0);
2568 }
2569 
2570 /*
2571  * Set a particular packet option, as a sticky option or an ancillary data
2572  * item.  "len" can be 0 only when it's a sticky option.
2573  * We have 4 cases of combination of "sticky" and "cmsg":
2574  * "sticky=0, cmsg=0": impossible
2575  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2576  * "sticky=1, cmsg=0": RFC3542 socket option
2577  * "sticky=1, cmsg=1": RFC2292 socket option
2578  */
2579 static int
2580 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2581     struct ucred *cred, int sticky, int cmsg, int uproto)
2582 {
2583 	int minmtupolicy, preftemp;
2584 	int error;
2585 
2586 	if (!sticky && !cmsg) {
2587 #ifdef DIAGNOSTIC
2588 		printf("ip6_setpktopt: impossible case\n");
2589 #endif
2590 		return (EINVAL);
2591 	}
2592 
2593 	/*
2594 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2595 	 * not be specified in the context of RFC3542.  Conversely,
2596 	 * RFC3542 types should not be specified in the context of RFC2292.
2597 	 */
2598 	if (!cmsg) {
2599 		switch (optname) {
2600 		case IPV6_2292PKTINFO:
2601 		case IPV6_2292HOPLIMIT:
2602 		case IPV6_2292NEXTHOP:
2603 		case IPV6_2292HOPOPTS:
2604 		case IPV6_2292DSTOPTS:
2605 		case IPV6_2292RTHDR:
2606 		case IPV6_2292PKTOPTIONS:
2607 			return (ENOPROTOOPT);
2608 		}
2609 	}
2610 	if (sticky && cmsg) {
2611 		switch (optname) {
2612 		case IPV6_PKTINFO:
2613 		case IPV6_HOPLIMIT:
2614 		case IPV6_NEXTHOP:
2615 		case IPV6_HOPOPTS:
2616 		case IPV6_DSTOPTS:
2617 		case IPV6_RTHDRDSTOPTS:
2618 		case IPV6_RTHDR:
2619 		case IPV6_USE_MIN_MTU:
2620 		case IPV6_DONTFRAG:
2621 		case IPV6_TCLASS:
2622 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2623 			return (ENOPROTOOPT);
2624 		}
2625 	}
2626 
2627 	switch (optname) {
2628 	case IPV6_2292PKTINFO:
2629 	case IPV6_PKTINFO:
2630 	{
2631 		struct ifnet *ifp = NULL;
2632 		struct in6_pktinfo *pktinfo;
2633 
2634 		if (len != sizeof(struct in6_pktinfo))
2635 			return (EINVAL);
2636 
2637 		pktinfo = (struct in6_pktinfo *)buf;
2638 
2639 		/*
2640 		 * An application can clear any sticky IPV6_PKTINFO option by
2641 		 * doing a "regular" setsockopt with ipi6_addr being
2642 		 * in6addr_any and ipi6_ifindex being zero.
2643 		 * [RFC 3542, Section 6]
2644 		 */
2645 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2646 		    pktinfo->ipi6_ifindex == 0 &&
2647 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2648 			ip6_clearpktopts(opt, optname);
2649 			break;
2650 		}
2651 
2652 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2653 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2654 			return (EINVAL);
2655 		}
2656 		if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2657 			return (EINVAL);
2658 		/* validate the interface index if specified. */
2659 		if (pktinfo->ipi6_ifindex > V_if_index)
2660 			 return (ENXIO);
2661 		if (pktinfo->ipi6_ifindex) {
2662 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2663 			if (ifp == NULL)
2664 				return (ENXIO);
2665 		}
2666 		if (ifp != NULL && (
2667 		    ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED))
2668 			return (ENETDOWN);
2669 
2670 		if (ifp != NULL &&
2671 		    !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2672 			struct in6_ifaddr *ia;
2673 
2674 			in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2675 			ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2676 			if (ia == NULL)
2677 				return (EADDRNOTAVAIL);
2678 			ifa_free(&ia->ia_ifa);
2679 		}
2680 		/*
2681 		 * We store the address anyway, and let in6_selectsrc()
2682 		 * validate the specified address.  This is because ipi6_addr
2683 		 * may not have enough information about its scope zone, and
2684 		 * we may need additional information (such as outgoing
2685 		 * interface or the scope zone of a destination address) to
2686 		 * disambiguate the scope.
2687 		 * XXX: the delay of the validation may confuse the
2688 		 * application when it is used as a sticky option.
2689 		 */
2690 		if (opt->ip6po_pktinfo == NULL) {
2691 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2692 			    M_IP6OPT, M_NOWAIT);
2693 			if (opt->ip6po_pktinfo == NULL)
2694 				return (ENOBUFS);
2695 		}
2696 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2697 		break;
2698 	}
2699 
2700 	case IPV6_2292HOPLIMIT:
2701 	case IPV6_HOPLIMIT:
2702 	{
2703 		int *hlimp;
2704 
2705 		/*
2706 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2707 		 * to simplify the ordering among hoplimit options.
2708 		 */
2709 		if (optname == IPV6_HOPLIMIT && sticky)
2710 			return (ENOPROTOOPT);
2711 
2712 		if (len != sizeof(int))
2713 			return (EINVAL);
2714 		hlimp = (int *)buf;
2715 		if (*hlimp < -1 || *hlimp > 255)
2716 			return (EINVAL);
2717 
2718 		opt->ip6po_hlim = *hlimp;
2719 		break;
2720 	}
2721 
2722 	case IPV6_TCLASS:
2723 	{
2724 		int tclass;
2725 
2726 		if (len != sizeof(int))
2727 			return (EINVAL);
2728 		tclass = *(int *)buf;
2729 		if (tclass < -1 || tclass > 255)
2730 			return (EINVAL);
2731 
2732 		opt->ip6po_tclass = tclass;
2733 		break;
2734 	}
2735 
2736 	case IPV6_2292NEXTHOP:
2737 	case IPV6_NEXTHOP:
2738 		if (cred != NULL) {
2739 			error = priv_check_cred(cred,
2740 			    PRIV_NETINET_SETHDROPTS, 0);
2741 			if (error)
2742 				return (error);
2743 		}
2744 
2745 		if (len == 0) {	/* just remove the option */
2746 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2747 			break;
2748 		}
2749 
2750 		/* check if cmsg_len is large enough for sa_len */
2751 		if (len < sizeof(struct sockaddr) || len < *buf)
2752 			return (EINVAL);
2753 
2754 		switch (((struct sockaddr *)buf)->sa_family) {
2755 		case AF_INET6:
2756 		{
2757 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2758 			int error;
2759 
2760 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2761 				return (EINVAL);
2762 
2763 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2764 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2765 				return (EINVAL);
2766 			}
2767 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2768 			    != 0) {
2769 				return (error);
2770 			}
2771 			break;
2772 		}
2773 		case AF_LINK:	/* should eventually be supported */
2774 		default:
2775 			return (EAFNOSUPPORT);
2776 		}
2777 
2778 		/* turn off the previous option, then set the new option. */
2779 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2780 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2781 		if (opt->ip6po_nexthop == NULL)
2782 			return (ENOBUFS);
2783 		bcopy(buf, opt->ip6po_nexthop, *buf);
2784 		break;
2785 
2786 	case IPV6_2292HOPOPTS:
2787 	case IPV6_HOPOPTS:
2788 	{
2789 		struct ip6_hbh *hbh;
2790 		int hbhlen;
2791 
2792 		/*
2793 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2794 		 * options, since per-option restriction has too much
2795 		 * overhead.
2796 		 */
2797 		if (cred != NULL) {
2798 			error = priv_check_cred(cred,
2799 			    PRIV_NETINET_SETHDROPTS, 0);
2800 			if (error)
2801 				return (error);
2802 		}
2803 
2804 		if (len == 0) {
2805 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2806 			break;	/* just remove the option */
2807 		}
2808 
2809 		/* message length validation */
2810 		if (len < sizeof(struct ip6_hbh))
2811 			return (EINVAL);
2812 		hbh = (struct ip6_hbh *)buf;
2813 		hbhlen = (hbh->ip6h_len + 1) << 3;
2814 		if (len != hbhlen)
2815 			return (EINVAL);
2816 
2817 		/* turn off the previous option, then set the new option. */
2818 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2819 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2820 		if (opt->ip6po_hbh == NULL)
2821 			return (ENOBUFS);
2822 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
2823 
2824 		break;
2825 	}
2826 
2827 	case IPV6_2292DSTOPTS:
2828 	case IPV6_DSTOPTS:
2829 	case IPV6_RTHDRDSTOPTS:
2830 	{
2831 		struct ip6_dest *dest, **newdest = NULL;
2832 		int destlen;
2833 
2834 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2835 			error = priv_check_cred(cred,
2836 			    PRIV_NETINET_SETHDROPTS, 0);
2837 			if (error)
2838 				return (error);
2839 		}
2840 
2841 		if (len == 0) {
2842 			ip6_clearpktopts(opt, optname);
2843 			break;	/* just remove the option */
2844 		}
2845 
2846 		/* message length validation */
2847 		if (len < sizeof(struct ip6_dest))
2848 			return (EINVAL);
2849 		dest = (struct ip6_dest *)buf;
2850 		destlen = (dest->ip6d_len + 1) << 3;
2851 		if (len != destlen)
2852 			return (EINVAL);
2853 
2854 		/*
2855 		 * Determine the position that the destination options header
2856 		 * should be inserted; before or after the routing header.
2857 		 */
2858 		switch (optname) {
2859 		case IPV6_2292DSTOPTS:
2860 			/*
2861 			 * The old advacned API is ambiguous on this point.
2862 			 * Our approach is to determine the position based
2863 			 * according to the existence of a routing header.
2864 			 * Note, however, that this depends on the order of the
2865 			 * extension headers in the ancillary data; the 1st
2866 			 * part of the destination options header must appear
2867 			 * before the routing header in the ancillary data,
2868 			 * too.
2869 			 * RFC3542 solved the ambiguity by introducing
2870 			 * separate ancillary data or option types.
2871 			 */
2872 			if (opt->ip6po_rthdr == NULL)
2873 				newdest = &opt->ip6po_dest1;
2874 			else
2875 				newdest = &opt->ip6po_dest2;
2876 			break;
2877 		case IPV6_RTHDRDSTOPTS:
2878 			newdest = &opt->ip6po_dest1;
2879 			break;
2880 		case IPV6_DSTOPTS:
2881 			newdest = &opt->ip6po_dest2;
2882 			break;
2883 		}
2884 
2885 		/* turn off the previous option, then set the new option. */
2886 		ip6_clearpktopts(opt, optname);
2887 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2888 		if (*newdest == NULL)
2889 			return (ENOBUFS);
2890 		bcopy(dest, *newdest, destlen);
2891 
2892 		break;
2893 	}
2894 
2895 	case IPV6_2292RTHDR:
2896 	case IPV6_RTHDR:
2897 	{
2898 		struct ip6_rthdr *rth;
2899 		int rthlen;
2900 
2901 		if (len == 0) {
2902 			ip6_clearpktopts(opt, IPV6_RTHDR);
2903 			break;	/* just remove the option */
2904 		}
2905 
2906 		/* message length validation */
2907 		if (len < sizeof(struct ip6_rthdr))
2908 			return (EINVAL);
2909 		rth = (struct ip6_rthdr *)buf;
2910 		rthlen = (rth->ip6r_len + 1) << 3;
2911 		if (len != rthlen)
2912 			return (EINVAL);
2913 
2914 		switch (rth->ip6r_type) {
2915 		case IPV6_RTHDR_TYPE_0:
2916 			if (rth->ip6r_len == 0)	/* must contain one addr */
2917 				return (EINVAL);
2918 			if (rth->ip6r_len % 2) /* length must be even */
2919 				return (EINVAL);
2920 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2921 				return (EINVAL);
2922 			break;
2923 		default:
2924 			return (EINVAL);	/* not supported */
2925 		}
2926 
2927 		/* turn off the previous option */
2928 		ip6_clearpktopts(opt, IPV6_RTHDR);
2929 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2930 		if (opt->ip6po_rthdr == NULL)
2931 			return (ENOBUFS);
2932 		bcopy(rth, opt->ip6po_rthdr, rthlen);
2933 
2934 		break;
2935 	}
2936 
2937 	case IPV6_USE_MIN_MTU:
2938 		if (len != sizeof(int))
2939 			return (EINVAL);
2940 		minmtupolicy = *(int *)buf;
2941 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2942 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
2943 		    minmtupolicy != IP6PO_MINMTU_ALL) {
2944 			return (EINVAL);
2945 		}
2946 		opt->ip6po_minmtu = minmtupolicy;
2947 		break;
2948 
2949 	case IPV6_DONTFRAG:
2950 		if (len != sizeof(int))
2951 			return (EINVAL);
2952 
2953 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2954 			/*
2955 			 * we ignore this option for TCP sockets.
2956 			 * (RFC3542 leaves this case unspecified.)
2957 			 */
2958 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2959 		} else
2960 			opt->ip6po_flags |= IP6PO_DONTFRAG;
2961 		break;
2962 
2963 	case IPV6_PREFER_TEMPADDR:
2964 		if (len != sizeof(int))
2965 			return (EINVAL);
2966 		preftemp = *(int *)buf;
2967 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2968 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2969 		    preftemp != IP6PO_TEMPADDR_PREFER) {
2970 			return (EINVAL);
2971 		}
2972 		opt->ip6po_prefer_tempaddr = preftemp;
2973 		break;
2974 
2975 	default:
2976 		return (ENOPROTOOPT);
2977 	} /* end of switch */
2978 
2979 	return (0);
2980 }
2981 
2982 /*
2983  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2984  * packet to the input queue of a specified interface.  Note that this
2985  * calls the output routine of the loopback "driver", but with an interface
2986  * pointer that might NOT be &loif -- easier than replicating that code here.
2987  */
2988 void
2989 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
2990 {
2991 	struct mbuf *copym;
2992 	struct ip6_hdr *ip6;
2993 
2994 	copym = m_copy(m, 0, M_COPYALL);
2995 	if (copym == NULL)
2996 		return;
2997 
2998 	/*
2999 	 * Make sure to deep-copy IPv6 header portion in case the data
3000 	 * is in an mbuf cluster, so that we can safely override the IPv6
3001 	 * header portion later.
3002 	 */
3003 	if (!M_WRITABLE(copym) ||
3004 	    copym->m_len < sizeof(struct ip6_hdr)) {
3005 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3006 		if (copym == NULL)
3007 			return;
3008 	}
3009 	ip6 = mtod(copym, struct ip6_hdr *);
3010 	/*
3011 	 * clear embedded scope identifiers if necessary.
3012 	 * in6_clearscope will touch the addresses only when necessary.
3013 	 */
3014 	in6_clearscope(&ip6->ip6_src);
3015 	in6_clearscope(&ip6->ip6_dst);
3016 	if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3017 		copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3018 		    CSUM_PSEUDO_HDR;
3019 		copym->m_pkthdr.csum_data = 0xffff;
3020 	}
3021 	if_simloop(ifp, copym, AF_INET6, 0);
3022 }
3023 
3024 /*
3025  * Chop IPv6 header off from the payload.
3026  */
3027 static int
3028 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3029 {
3030 	struct mbuf *mh;
3031 	struct ip6_hdr *ip6;
3032 
3033 	ip6 = mtod(m, struct ip6_hdr *);
3034 	if (m->m_len > sizeof(*ip6)) {
3035 		mh = m_gethdr(M_NOWAIT, MT_DATA);
3036 		if (mh == NULL) {
3037 			m_freem(m);
3038 			return ENOBUFS;
3039 		}
3040 		m_move_pkthdr(mh, m);
3041 		M_ALIGN(mh, sizeof(*ip6));
3042 		m->m_len -= sizeof(*ip6);
3043 		m->m_data += sizeof(*ip6);
3044 		mh->m_next = m;
3045 		m = mh;
3046 		m->m_len = sizeof(*ip6);
3047 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3048 	}
3049 	exthdrs->ip6e_ip6 = m;
3050 	return 0;
3051 }
3052 
3053 /*
3054  * Compute IPv6 extension header length.
3055  */
3056 int
3057 ip6_optlen(struct inpcb *in6p)
3058 {
3059 	int len;
3060 
3061 	if (!in6p->in6p_outputopts)
3062 		return 0;
3063 
3064 	len = 0;
3065 #define elen(x) \
3066     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3067 
3068 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3069 	if (in6p->in6p_outputopts->ip6po_rthdr)
3070 		/* dest1 is valid with rthdr only */
3071 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
3072 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3073 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3074 	return len;
3075 #undef elen
3076 }
3077