xref: /freebsd/sys/netinet6/ip6_output.c (revision 884d26c84cba3ffc3d4e626306098fcdfe6a0c2b)
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_ipsec.h"
69 #include "opt_sctp.h"
70 #include "opt_route.h"
71 #include "opt_rss.h"
72 
73 #include <sys/param.h>
74 #include <sys/kernel.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/errno.h>
78 #include <sys/priv.h>
79 #include <sys/proc.h>
80 #include <sys/protosw.h>
81 #include <sys/socket.h>
82 #include <sys/socketvar.h>
83 #include <sys/syslog.h>
84 #include <sys/ucred.h>
85 
86 #include <machine/in_cksum.h>
87 
88 #include <net/if.h>
89 #include <net/if_var.h>
90 #include <net/netisr.h>
91 #include <net/route.h>
92 #include <net/pfil.h>
93 #include <net/rss_config.h>
94 #include <net/vnet.h>
95 
96 #include <netinet/in.h>
97 #include <netinet/in_var.h>
98 #include <netinet/ip_var.h>
99 #include <netinet6/in6_fib.h>
100 #include <netinet6/in6_var.h>
101 #include <netinet/ip6.h>
102 #include <netinet/icmp6.h>
103 #include <netinet6/ip6_var.h>
104 #include <netinet/in_pcb.h>
105 #include <netinet/tcp_var.h>
106 #include <netinet6/nd6.h>
107 #include <netinet6/in6_rss.h>
108 
109 #ifdef IPSEC
110 #include <netipsec/ipsec.h>
111 #include <netipsec/ipsec6.h>
112 #include <netipsec/key.h>
113 #include <netinet6/ip6_ipsec.h>
114 #endif /* IPSEC */
115 #ifdef SCTP
116 #include <netinet/sctp.h>
117 #include <netinet/sctp_crc32.h>
118 #endif
119 
120 #include <netinet6/ip6protosw.h>
121 #include <netinet6/scope6_var.h>
122 
123 #ifdef FLOWTABLE
124 #include <net/flowtable.h>
125 #endif
126 
127 extern int in6_mcast_loop;
128 
129 struct ip6_exthdrs {
130 	struct mbuf *ip6e_ip6;
131 	struct mbuf *ip6e_hbh;
132 	struct mbuf *ip6e_dest1;
133 	struct mbuf *ip6e_rthdr;
134 	struct mbuf *ip6e_dest2;
135 };
136 
137 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
138 
139 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
140 			   struct ucred *, int);
141 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
142 	struct socket *, struct sockopt *);
143 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
144 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
145 	struct ucred *, int, int, int);
146 
147 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
148 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
149 	struct ip6_frag **);
150 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
151 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
152 static int ip6_getpmtu(struct route_in6 *, int,
153 	struct ifnet *, const struct in6_addr *, u_long *, int *, u_int);
154 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
155 	u_long *, int *);
156 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
157 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
158 
159 
160 /*
161  * Make an extension header from option data.  hp is the source, and
162  * mp is the destination.
163  */
164 #define MAKE_EXTHDR(hp, mp)						\
165     do {								\
166 	if (hp) {							\
167 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
168 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
169 		    ((eh)->ip6e_len + 1) << 3);				\
170 		if (error)						\
171 			goto freehdrs;					\
172 	}								\
173     } while (/*CONSTCOND*/ 0)
174 
175 /*
176  * Form a chain of extension headers.
177  * m is the extension header mbuf
178  * mp is the previous mbuf in the chain
179  * p is the next header
180  * i is the type of option.
181  */
182 #define MAKE_CHAIN(m, mp, p, i)\
183     do {\
184 	if (m) {\
185 		if (!hdrsplit) \
186 			panic("assumption failed: hdr not split"); \
187 		*mtod((m), u_char *) = *(p);\
188 		*(p) = (i);\
189 		p = mtod((m), u_char *);\
190 		(m)->m_next = (mp)->m_next;\
191 		(mp)->m_next = (m);\
192 		(mp) = (m);\
193 	}\
194     } while (/*CONSTCOND*/ 0)
195 
196 void
197 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
198 {
199 	u_short csum;
200 
201 	csum = in_cksum_skip(m, offset + plen, offset);
202 	if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
203 		csum = 0xffff;
204 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
205 
206 	if (offset + sizeof(u_short) > m->m_len) {
207 		printf("%s: delayed m_pullup, m->len: %d plen %u off %u "
208 		    "csum_flags=%b\n", __func__, m->m_len, plen, offset,
209 		    (int)m->m_pkthdr.csum_flags, CSUM_BITS);
210 		/*
211 		 * XXX this should not happen, but if it does, the correct
212 		 * behavior may be to insert the checksum in the appropriate
213 		 * next mbuf in the chain.
214 		 */
215 		return;
216 	}
217 	*(u_short *)(m->m_data + offset) = csum;
218 }
219 
220 int
221 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
222     int mtu, uint32_t id)
223 {
224 	struct mbuf *m, **mnext, *m_frgpart;
225 	struct ip6_hdr *ip6, *mhip6;
226 	struct ip6_frag *ip6f;
227 	int off;
228 	int error;
229 	int tlen = m0->m_pkthdr.len;
230 
231 	m = m0;
232 	ip6 = mtod(m, struct ip6_hdr *);
233 	mnext = &m->m_nextpkt;
234 
235 	for (off = hlen; off < tlen; off += mtu) {
236 		m = m_gethdr(M_NOWAIT, MT_DATA);
237 		if (!m) {
238 			IP6STAT_INC(ip6s_odropped);
239 			return (ENOBUFS);
240 		}
241 		m->m_flags = m0->m_flags & M_COPYFLAGS;
242 		*mnext = m;
243 		mnext = &m->m_nextpkt;
244 		m->m_data += max_linkhdr;
245 		mhip6 = mtod(m, struct ip6_hdr *);
246 		*mhip6 = *ip6;
247 		m->m_len = sizeof(*mhip6);
248 		error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
249 		if (error) {
250 			IP6STAT_INC(ip6s_odropped);
251 			return (error);
252 		}
253 		ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
254 		if (off + mtu >= tlen)
255 			mtu = tlen - off;
256 		else
257 			ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
258 		mhip6->ip6_plen = htons((u_short)(mtu + hlen +
259 		    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
260 		if ((m_frgpart = m_copy(m0, off, mtu)) == NULL) {
261 			IP6STAT_INC(ip6s_odropped);
262 			return (ENOBUFS);
263 		}
264 		m_cat(m, m_frgpart);
265 		m->m_pkthdr.len = mtu + hlen + sizeof(*ip6f);
266 		m->m_pkthdr.fibnum = m0->m_pkthdr.fibnum;
267 		m->m_pkthdr.rcvif = NULL;
268 		ip6f->ip6f_reserved = 0;
269 		ip6f->ip6f_ident = id;
270 		ip6f->ip6f_nxt = nextproto;
271 		IP6STAT_INC(ip6s_ofragments);
272 		in6_ifstat_inc(ifp, ifs6_out_fragcreat);
273 	}
274 
275 	return (0);
276 }
277 
278 /*
279  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
280  * header (with pri, len, nxt, hlim, src, dst).
281  * This function may modify ver and hlim only.
282  * The mbuf chain containing the packet will be freed.
283  * The mbuf opt, if present, will not be freed.
284  * If route_in6 ro is present and has ro_rt initialized, route lookup would be
285  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
286  * then result of route lookup is stored in ro->ro_rt.
287  *
288  * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and
289  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
290  * which is rt_mtu.
291  *
292  * ifpp - XXX: just for statistics
293  */
294 /*
295  * XXX TODO: no flowid is assigned for outbound flows?
296  */
297 int
298 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
299     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
300     struct ifnet **ifpp, struct inpcb *inp)
301 {
302 	struct ip6_hdr *ip6;
303 	struct ifnet *ifp, *origifp;
304 	struct mbuf *m = m0;
305 	struct mbuf *mprev = NULL;
306 	int hlen, tlen, len;
307 	struct route_in6 ip6route;
308 	struct rtentry *rt = NULL;
309 	struct sockaddr_in6 *dst, src_sa, dst_sa;
310 	struct in6_addr odst;
311 	int error = 0;
312 	struct in6_ifaddr *ia = NULL;
313 	u_long mtu;
314 	int alwaysfrag, dontfrag;
315 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
316 	struct ip6_exthdrs exthdrs;
317 	struct in6_addr src0, dst0;
318 	u_int32_t zone;
319 	struct route_in6 *ro_pmtu = NULL;
320 	int hdrsplit = 0;
321 	int sw_csum, tso;
322 	int needfiblookup;
323 	uint32_t fibnum;
324 	struct m_tag *fwd_tag = NULL;
325 	uint32_t id;
326 
327 	if (inp != NULL) {
328 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
329 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
330 			/* unconditionally set flowid */
331 			m->m_pkthdr.flowid = inp->inp_flowid;
332 			M_HASHTYPE_SET(m, inp->inp_flowtype);
333 		}
334 	}
335 
336 	bzero(&exthdrs, sizeof(exthdrs));
337 	if (opt) {
338 		/* Hop-by-Hop options header */
339 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
340 		/* Destination options header(1st part) */
341 		if (opt->ip6po_rthdr) {
342 			/*
343 			 * Destination options header(1st part)
344 			 * This only makes sense with a routing header.
345 			 * See Section 9.2 of RFC 3542.
346 			 * Disabling this part just for MIP6 convenience is
347 			 * a bad idea.  We need to think carefully about a
348 			 * way to make the advanced API coexist with MIP6
349 			 * options, which might automatically be inserted in
350 			 * the kernel.
351 			 */
352 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
353 		}
354 		/* Routing header */
355 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
356 		/* Destination options header(2nd part) */
357 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
358 	}
359 
360 #ifdef IPSEC
361 	/*
362 	 * IPSec checking which handles several cases.
363 	 * FAST IPSEC: We re-injected the packet.
364 	 * XXX: need scope argument.
365 	 */
366 	switch(ip6_ipsec_output(&m, inp, &error))
367 	{
368 	case 1:                 /* Bad packet */
369 		goto freehdrs;
370 	case -1:                /* IPSec done */
371 		goto done;
372 	case 0:                 /* No IPSec */
373 	default:
374 		break;
375 	}
376 #endif /* IPSEC */
377 
378 	/*
379 	 * Calculate the total length of the extension header chain.
380 	 * Keep the length of the unfragmentable part for fragmentation.
381 	 */
382 	optlen = 0;
383 	if (exthdrs.ip6e_hbh)
384 		optlen += exthdrs.ip6e_hbh->m_len;
385 	if (exthdrs.ip6e_dest1)
386 		optlen += exthdrs.ip6e_dest1->m_len;
387 	if (exthdrs.ip6e_rthdr)
388 		optlen += exthdrs.ip6e_rthdr->m_len;
389 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
390 
391 	/* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */
392 	if (exthdrs.ip6e_dest2)
393 		optlen += exthdrs.ip6e_dest2->m_len;
394 
395 	/*
396 	 * If there is at least one extension header,
397 	 * separate IP6 header from the payload.
398 	 */
399 	if (optlen && !hdrsplit) {
400 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
401 			m = NULL;
402 			goto freehdrs;
403 		}
404 		m = exthdrs.ip6e_ip6;
405 		hdrsplit++;
406 	}
407 
408 	ip6 = mtod(m, struct ip6_hdr *);
409 
410 	/* adjust mbuf packet header length */
411 	m->m_pkthdr.len += optlen;
412 	plen = m->m_pkthdr.len - sizeof(*ip6);
413 
414 	/* If this is a jumbo payload, insert a jumbo payload option. */
415 	if (plen > IPV6_MAXPACKET) {
416 		if (!hdrsplit) {
417 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
418 				m = NULL;
419 				goto freehdrs;
420 			}
421 			m = exthdrs.ip6e_ip6;
422 			hdrsplit++;
423 		}
424 		/* adjust pointer */
425 		ip6 = mtod(m, struct ip6_hdr *);
426 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
427 			goto freehdrs;
428 		ip6->ip6_plen = 0;
429 	} else
430 		ip6->ip6_plen = htons(plen);
431 
432 	/*
433 	 * Concatenate headers and fill in next header fields.
434 	 * Here we have, on "m"
435 	 *	IPv6 payload
436 	 * and we insert headers accordingly.  Finally, we should be getting:
437 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
438 	 *
439 	 * during the header composing process, "m" points to IPv6 header.
440 	 * "mprev" points to an extension header prior to esp.
441 	 */
442 	u_char *nexthdrp = &ip6->ip6_nxt;
443 	mprev = m;
444 
445 	/*
446 	 * we treat dest2 specially.  this makes IPsec processing
447 	 * much easier.  the goal here is to make mprev point the
448 	 * mbuf prior to dest2.
449 	 *
450 	 * result: IPv6 dest2 payload
451 	 * m and mprev will point to IPv6 header.
452 	 */
453 	if (exthdrs.ip6e_dest2) {
454 		if (!hdrsplit)
455 			panic("assumption failed: hdr not split");
456 		exthdrs.ip6e_dest2->m_next = m->m_next;
457 		m->m_next = exthdrs.ip6e_dest2;
458 		*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
459 		ip6->ip6_nxt = IPPROTO_DSTOPTS;
460 	}
461 
462 	/*
463 	 * result: IPv6 hbh dest1 rthdr dest2 payload
464 	 * m will point to IPv6 header.  mprev will point to the
465 	 * extension header prior to dest2 (rthdr in the above case).
466 	 */
467 	MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
468 	MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
469 		   IPPROTO_DSTOPTS);
470 	MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
471 		   IPPROTO_ROUTING);
472 
473 	/*
474 	 * If there is a routing header, discard the packet.
475 	 */
476 	if (exthdrs.ip6e_rthdr) {
477 		 error = EINVAL;
478 		 goto bad;
479 	}
480 
481 	/* Source address validation */
482 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
483 	    (flags & IPV6_UNSPECSRC) == 0) {
484 		error = EOPNOTSUPP;
485 		IP6STAT_INC(ip6s_badscope);
486 		goto bad;
487 	}
488 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
489 		error = EOPNOTSUPP;
490 		IP6STAT_INC(ip6s_badscope);
491 		goto bad;
492 	}
493 
494 	IP6STAT_INC(ip6s_localout);
495 
496 	/*
497 	 * Route packet.
498 	 */
499 	if (ro == NULL) {
500 		ro = &ip6route;
501 		bzero((caddr_t)ro, sizeof(*ro));
502 	} else
503 		ro->ro_flags |= RT_LLE_CACHE;
504 	ro_pmtu = ro;
505 	if (opt && opt->ip6po_rthdr)
506 		ro = &opt->ip6po_route;
507 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
508 #ifdef FLOWTABLE
509 	if (ro->ro_rt == NULL)
510 		(void )flowtable_lookup(AF_INET6, m, (struct route *)ro);
511 #endif
512 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
513 again:
514 	/*
515 	 * if specified, try to fill in the traffic class field.
516 	 * do not override if a non-zero value is already set.
517 	 * we check the diffserv field and the ecn field separately.
518 	 */
519 	if (opt && opt->ip6po_tclass >= 0) {
520 		int mask = 0;
521 
522 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
523 			mask |= 0xfc;
524 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
525 			mask |= 0x03;
526 		if (mask != 0)
527 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
528 	}
529 
530 	/* fill in or override the hop limit field, if necessary. */
531 	if (opt && opt->ip6po_hlim != -1)
532 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
533 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
534 		if (im6o != NULL)
535 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
536 		else
537 			ip6->ip6_hlim = V_ip6_defmcasthlim;
538 	}
539 	/*
540 	 * Validate route against routing table additions;
541 	 * a better/more specific route might have been added.
542 	 * Make sure address family is set in route.
543 	 */
544 	if (inp) {
545 		ro->ro_dst.sin6_family = AF_INET6;
546 		RT_VALIDATE((struct route *)ro, &inp->inp_rt_cookie, fibnum);
547 	}
548 	if (ro->ro_rt && fwd_tag == NULL && (ro->ro_rt->rt_flags & RTF_UP) &&
549 	    ro->ro_dst.sin6_family == AF_INET6 &&
550 	    IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
551 		rt = ro->ro_rt;
552 		ifp = ro->ro_rt->rt_ifp;
553 	} else {
554 		if (fwd_tag == NULL) {
555 			bzero(&dst_sa, sizeof(dst_sa));
556 			dst_sa.sin6_family = AF_INET6;
557 			dst_sa.sin6_len = sizeof(dst_sa);
558 			dst_sa.sin6_addr = ip6->ip6_dst;
559 		}
560 		error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
561 		    &rt, fibnum);
562 		if (error != 0) {
563 			if (ifp != NULL)
564 				in6_ifstat_inc(ifp, ifs6_out_discard);
565 			goto bad;
566 		}
567 	}
568 	if (rt == NULL) {
569 		/*
570 		 * If in6_selectroute() does not return a route entry,
571 		 * dst may not have been updated.
572 		 */
573 		*dst = dst_sa;	/* XXX */
574 	}
575 
576 	/*
577 	 * then rt (for unicast) and ifp must be non-NULL valid values.
578 	 */
579 	if ((flags & IPV6_FORWARDING) == 0) {
580 		/* XXX: the FORWARDING flag can be set for mrouting. */
581 		in6_ifstat_inc(ifp, ifs6_out_request);
582 	}
583 	if (rt != NULL) {
584 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
585 		counter_u64_add(rt->rt_pksent, 1);
586 	}
587 
588 
589 	/*
590 	 * The outgoing interface must be in the zone of source and
591 	 * destination addresses.
592 	 */
593 	origifp = ifp;
594 
595 	src0 = ip6->ip6_src;
596 	if (in6_setscope(&src0, origifp, &zone))
597 		goto badscope;
598 	bzero(&src_sa, sizeof(src_sa));
599 	src_sa.sin6_family = AF_INET6;
600 	src_sa.sin6_len = sizeof(src_sa);
601 	src_sa.sin6_addr = ip6->ip6_src;
602 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
603 		goto badscope;
604 
605 	dst0 = ip6->ip6_dst;
606 	if (in6_setscope(&dst0, origifp, &zone))
607 		goto badscope;
608 	/* re-initialize to be sure */
609 	bzero(&dst_sa, sizeof(dst_sa));
610 	dst_sa.sin6_family = AF_INET6;
611 	dst_sa.sin6_len = sizeof(dst_sa);
612 	dst_sa.sin6_addr = ip6->ip6_dst;
613 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
614 		goto badscope;
615 	}
616 
617 	/* We should use ia_ifp to support the case of
618 	 * sending packets to an address of our own.
619 	 */
620 	if (ia != NULL && ia->ia_ifp)
621 		ifp = ia->ia_ifp;
622 
623 	/* scope check is done. */
624 	goto routefound;
625 
626   badscope:
627 	IP6STAT_INC(ip6s_badscope);
628 	in6_ifstat_inc(origifp, ifs6_out_discard);
629 	if (error == 0)
630 		error = EHOSTUNREACH; /* XXX */
631 	goto bad;
632 
633   routefound:
634 	if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
635 		if (opt && opt->ip6po_nextroute.ro_rt) {
636 			/*
637 			 * The nexthop is explicitly specified by the
638 			 * application.  We assume the next hop is an IPv6
639 			 * address.
640 			 */
641 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
642 		}
643 		else if ((rt->rt_flags & RTF_GATEWAY))
644 			dst = (struct sockaddr_in6 *)rt->rt_gateway;
645 	}
646 
647 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
648 		m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
649 	} else {
650 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
651 		in6_ifstat_inc(ifp, ifs6_out_mcast);
652 		/*
653 		 * Confirm that the outgoing interface supports multicast.
654 		 */
655 		if (!(ifp->if_flags & IFF_MULTICAST)) {
656 			IP6STAT_INC(ip6s_noroute);
657 			in6_ifstat_inc(ifp, ifs6_out_discard);
658 			error = ENETUNREACH;
659 			goto bad;
660 		}
661 		if ((im6o == NULL && in6_mcast_loop) ||
662 		    (im6o && im6o->im6o_multicast_loop)) {
663 			/*
664 			 * Loop back multicast datagram if not expressly
665 			 * forbidden to do so, even if we have not joined
666 			 * the address; protocols will filter it later,
667 			 * thus deferring a hash lookup and lock acquisition
668 			 * at the expense of an m_copym().
669 			 */
670 			ip6_mloopback(ifp, m);
671 		} else {
672 			/*
673 			 * If we are acting as a multicast router, perform
674 			 * multicast forwarding as if the packet had just
675 			 * arrived on the interface to which we are about
676 			 * to send.  The multicast forwarding function
677 			 * recursively calls this function, using the
678 			 * IPV6_FORWARDING flag to prevent infinite recursion.
679 			 *
680 			 * Multicasts that are looped back by ip6_mloopback(),
681 			 * above, will be forwarded by the ip6_input() routine,
682 			 * if necessary.
683 			 */
684 			if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
685 				/*
686 				 * XXX: ip6_mforward expects that rcvif is NULL
687 				 * when it is called from the originating path.
688 				 * However, it may not always be the case.
689 				 */
690 				m->m_pkthdr.rcvif = NULL;
691 				if (ip6_mforward(ip6, ifp, m) != 0) {
692 					m_freem(m);
693 					goto done;
694 				}
695 			}
696 		}
697 		/*
698 		 * Multicasts with a hoplimit of zero may be looped back,
699 		 * above, but must not be transmitted on a network.
700 		 * Also, multicasts addressed to the loopback interface
701 		 * are not sent -- the above call to ip6_mloopback() will
702 		 * loop back a copy if this host actually belongs to the
703 		 * destination group on the loopback interface.
704 		 */
705 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
706 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
707 			m_freem(m);
708 			goto done;
709 		}
710 	}
711 
712 	/*
713 	 * Fill the outgoing inteface to tell the upper layer
714 	 * to increment per-interface statistics.
715 	 */
716 	if (ifpp)
717 		*ifpp = ifp;
718 
719 	/* Determine path MTU. */
720 	if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
721 	    &mtu, &alwaysfrag, fibnum)) != 0)
722 		goto bad;
723 
724 	/*
725 	 * The caller of this function may specify to use the minimum MTU
726 	 * in some cases.
727 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
728 	 * setting.  The logic is a bit complicated; by default, unicast
729 	 * packets will follow path MTU while multicast packets will be sent at
730 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
731 	 * including unicast ones will be sent at the minimum MTU.  Multicast
732 	 * packets will always be sent at the minimum MTU unless
733 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
734 	 * See RFC 3542 for more details.
735 	 */
736 	if (mtu > IPV6_MMTU) {
737 		if ((flags & IPV6_MINMTU))
738 			mtu = IPV6_MMTU;
739 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
740 			mtu = IPV6_MMTU;
741 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
742 			 (opt == NULL ||
743 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
744 			mtu = IPV6_MMTU;
745 		}
746 	}
747 
748 	/*
749 	 * clear embedded scope identifiers if necessary.
750 	 * in6_clearscope will touch the addresses only when necessary.
751 	 */
752 	in6_clearscope(&ip6->ip6_src);
753 	in6_clearscope(&ip6->ip6_dst);
754 
755 	/*
756 	 * If the outgoing packet contains a hop-by-hop options header,
757 	 * it must be examined and processed even by the source node.
758 	 * (RFC 2460, section 4.)
759 	 */
760 	if (exthdrs.ip6e_hbh) {
761 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
762 		u_int32_t dummy; /* XXX unused */
763 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
764 
765 #ifdef DIAGNOSTIC
766 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
767 			panic("ip6e_hbh is not contiguous");
768 #endif
769 		/*
770 		 *  XXX: if we have to send an ICMPv6 error to the sender,
771 		 *       we need the M_LOOP flag since icmp6_error() expects
772 		 *       the IPv6 and the hop-by-hop options header are
773 		 *       contiguous unless the flag is set.
774 		 */
775 		m->m_flags |= M_LOOP;
776 		m->m_pkthdr.rcvif = ifp;
777 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
778 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
779 		    &dummy, &plen) < 0) {
780 			/* m was already freed at this point */
781 			error = EINVAL;/* better error? */
782 			goto done;
783 		}
784 		m->m_flags &= ~M_LOOP; /* XXX */
785 		m->m_pkthdr.rcvif = NULL;
786 	}
787 
788 	/* Jump over all PFIL processing if hooks are not active. */
789 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
790 		goto passout;
791 
792 	odst = ip6->ip6_dst;
793 	/* Run through list of hooks for output packets. */
794 	error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
795 	if (error != 0 || m == NULL)
796 		goto done;
797 	/* adjust pointer */
798 	ip6 = mtod(m, struct ip6_hdr *);
799 
800 	needfiblookup = 0;
801 	/* See if destination IP address was changed by packet filter. */
802 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
803 		m->m_flags |= M_SKIP_FIREWALL;
804 		/* If destination is now ourself drop to ip6_input(). */
805 		if (in6_localip(&ip6->ip6_dst)) {
806 			m->m_flags |= M_FASTFWD_OURS;
807 			if (m->m_pkthdr.rcvif == NULL)
808 				m->m_pkthdr.rcvif = V_loif;
809 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
810 				m->m_pkthdr.csum_flags |=
811 				    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
812 				m->m_pkthdr.csum_data = 0xffff;
813 			}
814 #ifdef SCTP
815 			if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
816 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
817 #endif
818 			error = netisr_queue(NETISR_IPV6, m);
819 			goto done;
820 		} else {
821 			RO_RTFREE(ro);
822 			needfiblookup = 1; /* Redo the routing table lookup. */
823 		}
824 	}
825 	/* See if fib was changed by packet filter. */
826 	if (fibnum != M_GETFIB(m)) {
827 		m->m_flags |= M_SKIP_FIREWALL;
828 		fibnum = M_GETFIB(m);
829 		RO_RTFREE(ro);
830 		needfiblookup = 1;
831 	}
832 	if (needfiblookup)
833 		goto again;
834 
835 	/* See if local, if yes, send it to netisr. */
836 	if (m->m_flags & M_FASTFWD_OURS) {
837 		if (m->m_pkthdr.rcvif == NULL)
838 			m->m_pkthdr.rcvif = V_loif;
839 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
840 			m->m_pkthdr.csum_flags |=
841 			    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
842 			m->m_pkthdr.csum_data = 0xffff;
843 		}
844 #ifdef SCTP
845 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
846 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
847 #endif
848 		error = netisr_queue(NETISR_IPV6, m);
849 		goto done;
850 	}
851 	/* Or forward to some other address? */
852 	if ((m->m_flags & M_IP6_NEXTHOP) &&
853 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
854 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
855 		bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
856 		m->m_flags |= M_SKIP_FIREWALL;
857 		m->m_flags &= ~M_IP6_NEXTHOP;
858 		m_tag_delete(m, fwd_tag);
859 		goto again;
860 	}
861 
862 passout:
863 	/*
864 	 * Send the packet to the outgoing interface.
865 	 * If necessary, do IPv6 fragmentation before sending.
866 	 *
867 	 * the logic here is rather complex:
868 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
869 	 * 1-a:	send as is if tlen <= path mtu
870 	 * 1-b:	fragment if tlen > path mtu
871 	 *
872 	 * 2: if user asks us not to fragment (dontfrag == 1)
873 	 * 2-a:	send as is if tlen <= interface mtu
874 	 * 2-b:	error if tlen > interface mtu
875 	 *
876 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
877 	 *	always fragment
878 	 *
879 	 * 4: if dontfrag == 1 && alwaysfrag == 1
880 	 *	error, as we cannot handle this conflicting request
881 	 */
882 	sw_csum = m->m_pkthdr.csum_flags;
883 	if (!hdrsplit) {
884 		tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
885 		sw_csum &= ~ifp->if_hwassist;
886 	} else
887 		tso = 0;
888 	/*
889 	 * If we added extension headers, we will not do TSO and calculate the
890 	 * checksums ourselves for now.
891 	 * XXX-BZ  Need a framework to know when the NIC can handle it, even
892 	 * with ext. hdrs.
893 	 */
894 	if (sw_csum & CSUM_DELAY_DATA_IPV6) {
895 		sw_csum &= ~CSUM_DELAY_DATA_IPV6;
896 		in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
897 	}
898 #ifdef SCTP
899 	if (sw_csum & CSUM_SCTP_IPV6) {
900 		sw_csum &= ~CSUM_SCTP_IPV6;
901 		sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
902 	}
903 #endif
904 	m->m_pkthdr.csum_flags &= ifp->if_hwassist;
905 	tlen = m->m_pkthdr.len;
906 
907 	if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
908 		dontfrag = 1;
909 	else
910 		dontfrag = 0;
911 	if (dontfrag && alwaysfrag) {	/* case 4 */
912 		/* conflicting request - can't transmit */
913 		error = EMSGSIZE;
914 		goto bad;
915 	}
916 	if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {	/* case 2-b */
917 		/*
918 		 * Even if the DONTFRAG option is specified, we cannot send the
919 		 * packet when the data length is larger than the MTU of the
920 		 * outgoing interface.
921 		 * Notify the error by sending IPV6_PATHMTU ancillary data if
922 		 * application wanted to know the MTU value. Also return an
923 		 * error code (this is not described in the API spec).
924 		 */
925 		if (inp != NULL)
926 			ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
927 		error = EMSGSIZE;
928 		goto bad;
929 	}
930 
931 	/*
932 	 * transmit packet without fragmentation
933 	 */
934 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
935 		struct in6_ifaddr *ia6;
936 
937 		ip6 = mtod(m, struct ip6_hdr *);
938 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
939 		if (ia6) {
940 			/* Record statistics for this interface address. */
941 			counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
942 			counter_u64_add(ia6->ia_ifa.ifa_obytes,
943 			    m->m_pkthdr.len);
944 			ifa_free(&ia6->ia_ifa);
945 		}
946 		error = nd6_output_ifp(ifp, origifp, m, dst,
947 		    (struct route *)ro);
948 		goto done;
949 	}
950 
951 	/*
952 	 * try to fragment the packet.  case 1-b and 3
953 	 */
954 	if (mtu < IPV6_MMTU) {
955 		/* path MTU cannot be less than IPV6_MMTU */
956 		error = EMSGSIZE;
957 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
958 		goto bad;
959 	} else if (ip6->ip6_plen == 0) {
960 		/* jumbo payload cannot be fragmented */
961 		error = EMSGSIZE;
962 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
963 		goto bad;
964 	} else {
965 		u_char nextproto;
966 
967 		/*
968 		 * Too large for the destination or interface;
969 		 * fragment if possible.
970 		 * Must be able to put at least 8 bytes per fragment.
971 		 */
972 		hlen = unfragpartlen;
973 		if (mtu > IPV6_MAXPACKET)
974 			mtu = IPV6_MAXPACKET;
975 
976 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
977 		if (len < 8) {
978 			error = EMSGSIZE;
979 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
980 			goto bad;
981 		}
982 
983 		/*
984 		 * If the interface will not calculate checksums on
985 		 * fragmented packets, then do it here.
986 		 * XXX-BZ handle the hw offloading case.  Need flags.
987 		 */
988 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
989 			in6_delayed_cksum(m, plen, hlen);
990 			m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
991 		}
992 #ifdef SCTP
993 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
994 			sctp_delayed_cksum(m, hlen);
995 			m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
996 		}
997 #endif
998 		/*
999 		 * Change the next header field of the last header in the
1000 		 * unfragmentable part.
1001 		 */
1002 		if (exthdrs.ip6e_rthdr) {
1003 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1004 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1005 		} else if (exthdrs.ip6e_dest1) {
1006 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1007 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1008 		} else if (exthdrs.ip6e_hbh) {
1009 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1010 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1011 		} else {
1012 			nextproto = ip6->ip6_nxt;
1013 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1014 		}
1015 
1016 		/*
1017 		 * Loop through length of segment after first fragment,
1018 		 * make new header and copy data of each part and link onto
1019 		 * chain.
1020 		 */
1021 		m0 = m;
1022 		id = htonl(ip6_randomid());
1023 		if ((error = ip6_fragment(ifp, m, hlen, nextproto, len, id)))
1024 			goto sendorfree;
1025 
1026 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1027 	}
1028 
1029 	/*
1030 	 * Remove leading garbages.
1031 	 */
1032 sendorfree:
1033 	m = m0->m_nextpkt;
1034 	m0->m_nextpkt = 0;
1035 	m_freem(m0);
1036 	for (m0 = m; m; m = m0) {
1037 		m0 = m->m_nextpkt;
1038 		m->m_nextpkt = 0;
1039 		if (error == 0) {
1040 			/* Record statistics for this interface address. */
1041 			if (ia) {
1042 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1043 				counter_u64_add(ia->ia_ifa.ifa_obytes,
1044 				    m->m_pkthdr.len);
1045 			}
1046 			error = nd6_output_ifp(ifp, origifp, m, dst,
1047 			    (struct route *)ro);
1048 		} else
1049 			m_freem(m);
1050 	}
1051 
1052 	if (error == 0)
1053 		IP6STAT_INC(ip6s_fragmented);
1054 
1055 done:
1056 	/*
1057 	 * Release the route if using our private route, or if
1058 	 * (with flowtable) we don't have our own reference.
1059 	 */
1060 	if (ro == &ip6route || ro->ro_flags & RT_NORTREF)
1061 		RO_RTFREE(ro);
1062 	return (error);
1063 
1064 freehdrs:
1065 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1066 	m_freem(exthdrs.ip6e_dest1);
1067 	m_freem(exthdrs.ip6e_rthdr);
1068 	m_freem(exthdrs.ip6e_dest2);
1069 	/* FALLTHROUGH */
1070 bad:
1071 	if (m)
1072 		m_freem(m);
1073 	goto done;
1074 }
1075 
1076 static int
1077 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1078 {
1079 	struct mbuf *m;
1080 
1081 	if (hlen > MCLBYTES)
1082 		return (ENOBUFS); /* XXX */
1083 
1084 	if (hlen > MLEN)
1085 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1086 	else
1087 		m = m_get(M_NOWAIT, MT_DATA);
1088 	if (m == NULL)
1089 		return (ENOBUFS);
1090 	m->m_len = hlen;
1091 	if (hdr)
1092 		bcopy(hdr, mtod(m, caddr_t), hlen);
1093 
1094 	*mp = m;
1095 	return (0);
1096 }
1097 
1098 /*
1099  * Insert jumbo payload option.
1100  */
1101 static int
1102 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1103 {
1104 	struct mbuf *mopt;
1105 	u_char *optbuf;
1106 	u_int32_t v;
1107 
1108 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1109 
1110 	/*
1111 	 * If there is no hop-by-hop options header, allocate new one.
1112 	 * If there is one but it doesn't have enough space to store the
1113 	 * jumbo payload option, allocate a cluster to store the whole options.
1114 	 * Otherwise, use it to store the options.
1115 	 */
1116 	if (exthdrs->ip6e_hbh == NULL) {
1117 		mopt = m_get(M_NOWAIT, MT_DATA);
1118 		if (mopt == NULL)
1119 			return (ENOBUFS);
1120 		mopt->m_len = JUMBOOPTLEN;
1121 		optbuf = mtod(mopt, u_char *);
1122 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1123 		exthdrs->ip6e_hbh = mopt;
1124 	} else {
1125 		struct ip6_hbh *hbh;
1126 
1127 		mopt = exthdrs->ip6e_hbh;
1128 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1129 			/*
1130 			 * XXX assumption:
1131 			 * - exthdrs->ip6e_hbh is not referenced from places
1132 			 *   other than exthdrs.
1133 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1134 			 */
1135 			int oldoptlen = mopt->m_len;
1136 			struct mbuf *n;
1137 
1138 			/*
1139 			 * XXX: give up if the whole (new) hbh header does
1140 			 * not fit even in an mbuf cluster.
1141 			 */
1142 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1143 				return (ENOBUFS);
1144 
1145 			/*
1146 			 * As a consequence, we must always prepare a cluster
1147 			 * at this point.
1148 			 */
1149 			n = m_getcl(M_NOWAIT, MT_DATA, 0);
1150 			if (n == NULL)
1151 				return (ENOBUFS);
1152 			n->m_len = oldoptlen + JUMBOOPTLEN;
1153 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1154 			    oldoptlen);
1155 			optbuf = mtod(n, caddr_t) + oldoptlen;
1156 			m_freem(mopt);
1157 			mopt = exthdrs->ip6e_hbh = n;
1158 		} else {
1159 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1160 			mopt->m_len += JUMBOOPTLEN;
1161 		}
1162 		optbuf[0] = IP6OPT_PADN;
1163 		optbuf[1] = 1;
1164 
1165 		/*
1166 		 * Adjust the header length according to the pad and
1167 		 * the jumbo payload option.
1168 		 */
1169 		hbh = mtod(mopt, struct ip6_hbh *);
1170 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1171 	}
1172 
1173 	/* fill in the option. */
1174 	optbuf[2] = IP6OPT_JUMBO;
1175 	optbuf[3] = 4;
1176 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1177 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1178 
1179 	/* finally, adjust the packet header length */
1180 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1181 
1182 	return (0);
1183 #undef JUMBOOPTLEN
1184 }
1185 
1186 /*
1187  * Insert fragment header and copy unfragmentable header portions.
1188  */
1189 static int
1190 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1191     struct ip6_frag **frghdrp)
1192 {
1193 	struct mbuf *n, *mlast;
1194 
1195 	if (hlen > sizeof(struct ip6_hdr)) {
1196 		n = m_copym(m0, sizeof(struct ip6_hdr),
1197 		    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1198 		if (n == NULL)
1199 			return (ENOBUFS);
1200 		m->m_next = n;
1201 	} else
1202 		n = m;
1203 
1204 	/* Search for the last mbuf of unfragmentable part. */
1205 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1206 		;
1207 
1208 	if (M_WRITABLE(mlast) &&
1209 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1210 		/* use the trailing space of the last mbuf for the fragment hdr */
1211 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1212 		    mlast->m_len);
1213 		mlast->m_len += sizeof(struct ip6_frag);
1214 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1215 	} else {
1216 		/* allocate a new mbuf for the fragment header */
1217 		struct mbuf *mfrg;
1218 
1219 		mfrg = m_get(M_NOWAIT, MT_DATA);
1220 		if (mfrg == NULL)
1221 			return (ENOBUFS);
1222 		mfrg->m_len = sizeof(struct ip6_frag);
1223 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1224 		mlast->m_next = mfrg;
1225 	}
1226 
1227 	return (0);
1228 }
1229 
1230 /*
1231  * Calculates IPv6 path mtu for destination @dst.
1232  * Resulting MTU is stored in @mtup.
1233  *
1234  * Returns 0 on success.
1235  */
1236 static int
1237 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1238 {
1239 	struct nhop6_extended nh6;
1240 	struct in6_addr kdst;
1241 	uint32_t scopeid;
1242 	struct ifnet *ifp;
1243 	u_long mtu;
1244 	int error;
1245 
1246 	in6_splitscope(dst, &kdst, &scopeid);
1247 	if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0)
1248 		return (EHOSTUNREACH);
1249 
1250 	ifp = nh6.nh_ifp;
1251 	mtu = nh6.nh_mtu;
1252 
1253 	error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL);
1254 	fib6_free_nh_ext(fibnum, &nh6);
1255 
1256 	return (error);
1257 }
1258 
1259 /*
1260  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1261  * and cached data in @ro_pmtu.
1262  * MTU from (successful) route lookup is saved (along with dst)
1263  * inside @ro_pmtu to avoid subsequent route lookups after packet
1264  * filter processing.
1265  *
1266  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1267  * Returns 0 on success.
1268  */
1269 static int
1270 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1271     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1272     int *alwaysfragp, u_int fibnum)
1273 {
1274 	struct nhop6_basic nh6;
1275 	struct in6_addr kdst;
1276 	uint32_t scopeid;
1277 	struct sockaddr_in6 *sa6_dst;
1278 	u_long mtu;
1279 
1280 	mtu = 0;
1281 	if (do_lookup) {
1282 
1283 		/*
1284 		 * Here ro_pmtu has final destination address, while
1285 		 * ro might represent immediate destination.
1286 		 * Use ro_pmtu destination since mtu might differ.
1287 		 */
1288 		sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1289 		if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1290 			ro_pmtu->ro_mtu = 0;
1291 
1292 		if (ro_pmtu->ro_mtu == 0) {
1293 			bzero(sa6_dst, sizeof(*sa6_dst));
1294 			sa6_dst->sin6_family = AF_INET6;
1295 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1296 			sa6_dst->sin6_addr = *dst;
1297 
1298 			in6_splitscope(dst, &kdst, &scopeid);
1299 			if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0,
1300 			    &nh6) == 0)
1301 				ro_pmtu->ro_mtu = nh6.nh_mtu;
1302 		}
1303 
1304 		mtu = ro_pmtu->ro_mtu;
1305 	}
1306 
1307 	if (ro_pmtu->ro_rt)
1308 		mtu = ro_pmtu->ro_rt->rt_mtu;
1309 
1310 	return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp));
1311 }
1312 
1313 /*
1314  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1315  * hostcache data for @dst.
1316  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1317  *
1318  * Returns 0 on success.
1319  */
1320 static int
1321 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1322     u_long *mtup, int *alwaysfragp)
1323 {
1324 	u_long mtu = 0;
1325 	int alwaysfrag = 0;
1326 	int error = 0;
1327 
1328 	if (rt_mtu > 0) {
1329 		u_int32_t ifmtu;
1330 		struct in_conninfo inc;
1331 
1332 		bzero(&inc, sizeof(inc));
1333 		inc.inc_flags |= INC_ISIPV6;
1334 		inc.inc6_faddr = *dst;
1335 
1336 		ifmtu = IN6_LINKMTU(ifp);
1337 		mtu = tcp_hc_getmtu(&inc);
1338 		if (mtu)
1339 			mtu = min(mtu, rt_mtu);
1340 		else
1341 			mtu = rt_mtu;
1342 		if (mtu == 0)
1343 			mtu = ifmtu;
1344 		else if (mtu < IPV6_MMTU) {
1345 			/*
1346 			 * RFC2460 section 5, last paragraph:
1347 			 * if we record ICMPv6 too big message with
1348 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1349 			 * or smaller, with framgent header attached.
1350 			 * (fragment header is needed regardless from the
1351 			 * packet size, for translators to identify packets)
1352 			 */
1353 			alwaysfrag = 1;
1354 			mtu = IPV6_MMTU;
1355 		}
1356 	} else if (ifp) {
1357 		mtu = IN6_LINKMTU(ifp);
1358 	} else
1359 		error = EHOSTUNREACH; /* XXX */
1360 
1361 	*mtup = mtu;
1362 	if (alwaysfragp)
1363 		*alwaysfragp = alwaysfrag;
1364 	return (error);
1365 }
1366 
1367 /*
1368  * IP6 socket option processing.
1369  */
1370 int
1371 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1372 {
1373 	int optdatalen, uproto;
1374 	void *optdata;
1375 	struct inpcb *in6p = sotoinpcb(so);
1376 	int error, optval;
1377 	int level, op, optname;
1378 	int optlen;
1379 	struct thread *td;
1380 #ifdef	RSS
1381 	uint32_t rss_bucket;
1382 	int retval;
1383 #endif
1384 
1385 	level = sopt->sopt_level;
1386 	op = sopt->sopt_dir;
1387 	optname = sopt->sopt_name;
1388 	optlen = sopt->sopt_valsize;
1389 	td = sopt->sopt_td;
1390 	error = 0;
1391 	optval = 0;
1392 	uproto = (int)so->so_proto->pr_protocol;
1393 
1394 	if (level != IPPROTO_IPV6) {
1395 		error = EINVAL;
1396 
1397 		if (sopt->sopt_level == SOL_SOCKET &&
1398 		    sopt->sopt_dir == SOPT_SET) {
1399 			switch (sopt->sopt_name) {
1400 			case SO_REUSEADDR:
1401 				INP_WLOCK(in6p);
1402 				if ((so->so_options & SO_REUSEADDR) != 0)
1403 					in6p->inp_flags2 |= INP_REUSEADDR;
1404 				else
1405 					in6p->inp_flags2 &= ~INP_REUSEADDR;
1406 				INP_WUNLOCK(in6p);
1407 				error = 0;
1408 				break;
1409 			case SO_REUSEPORT:
1410 				INP_WLOCK(in6p);
1411 				if ((so->so_options & SO_REUSEPORT) != 0)
1412 					in6p->inp_flags2 |= INP_REUSEPORT;
1413 				else
1414 					in6p->inp_flags2 &= ~INP_REUSEPORT;
1415 				INP_WUNLOCK(in6p);
1416 				error = 0;
1417 				break;
1418 			case SO_SETFIB:
1419 				INP_WLOCK(in6p);
1420 				in6p->inp_inc.inc_fibnum = so->so_fibnum;
1421 				INP_WUNLOCK(in6p);
1422 				error = 0;
1423 				break;
1424 			default:
1425 				break;
1426 			}
1427 		}
1428 	} else {		/* level == IPPROTO_IPV6 */
1429 		switch (op) {
1430 
1431 		case SOPT_SET:
1432 			switch (optname) {
1433 			case IPV6_2292PKTOPTIONS:
1434 #ifdef IPV6_PKTOPTIONS
1435 			case IPV6_PKTOPTIONS:
1436 #endif
1437 			{
1438 				struct mbuf *m;
1439 
1440 				error = soopt_getm(sopt, &m); /* XXX */
1441 				if (error != 0)
1442 					break;
1443 				error = soopt_mcopyin(sopt, m); /* XXX */
1444 				if (error != 0)
1445 					break;
1446 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1447 						    m, so, sopt);
1448 				m_freem(m); /* XXX */
1449 				break;
1450 			}
1451 
1452 			/*
1453 			 * Use of some Hop-by-Hop options or some
1454 			 * Destination options, might require special
1455 			 * privilege.  That is, normal applications
1456 			 * (without special privilege) might be forbidden
1457 			 * from setting certain options in outgoing packets,
1458 			 * and might never see certain options in received
1459 			 * packets. [RFC 2292 Section 6]
1460 			 * KAME specific note:
1461 			 *  KAME prevents non-privileged users from sending or
1462 			 *  receiving ANY hbh/dst options in order to avoid
1463 			 *  overhead of parsing options in the kernel.
1464 			 */
1465 			case IPV6_RECVHOPOPTS:
1466 			case IPV6_RECVDSTOPTS:
1467 			case IPV6_RECVRTHDRDSTOPTS:
1468 				if (td != NULL) {
1469 					error = priv_check(td,
1470 					    PRIV_NETINET_SETHDROPTS);
1471 					if (error)
1472 						break;
1473 				}
1474 				/* FALLTHROUGH */
1475 			case IPV6_UNICAST_HOPS:
1476 			case IPV6_HOPLIMIT:
1477 
1478 			case IPV6_RECVPKTINFO:
1479 			case IPV6_RECVHOPLIMIT:
1480 			case IPV6_RECVRTHDR:
1481 			case IPV6_RECVPATHMTU:
1482 			case IPV6_RECVTCLASS:
1483 			case IPV6_RECVFLOWID:
1484 #ifdef	RSS
1485 			case IPV6_RECVRSSBUCKETID:
1486 #endif
1487 			case IPV6_V6ONLY:
1488 			case IPV6_AUTOFLOWLABEL:
1489 			case IPV6_BINDANY:
1490 			case IPV6_BINDMULTI:
1491 #ifdef	RSS
1492 			case IPV6_RSS_LISTEN_BUCKET:
1493 #endif
1494 				if (optname == IPV6_BINDANY && td != NULL) {
1495 					error = priv_check(td,
1496 					    PRIV_NETINET_BINDANY);
1497 					if (error)
1498 						break;
1499 				}
1500 
1501 				if (optlen != sizeof(int)) {
1502 					error = EINVAL;
1503 					break;
1504 				}
1505 				error = sooptcopyin(sopt, &optval,
1506 					sizeof optval, sizeof optval);
1507 				if (error)
1508 					break;
1509 				switch (optname) {
1510 
1511 				case IPV6_UNICAST_HOPS:
1512 					if (optval < -1 || optval >= 256)
1513 						error = EINVAL;
1514 					else {
1515 						/* -1 = kernel default */
1516 						in6p->in6p_hops = optval;
1517 						if ((in6p->inp_vflag &
1518 						     INP_IPV4) != 0)
1519 							in6p->inp_ip_ttl = optval;
1520 					}
1521 					break;
1522 #define OPTSET(bit) \
1523 do { \
1524 	INP_WLOCK(in6p); \
1525 	if (optval) \
1526 		in6p->inp_flags |= (bit); \
1527 	else \
1528 		in6p->inp_flags &= ~(bit); \
1529 	INP_WUNLOCK(in6p); \
1530 } while (/*CONSTCOND*/ 0)
1531 #define OPTSET2292(bit) \
1532 do { \
1533 	INP_WLOCK(in6p); \
1534 	in6p->inp_flags |= IN6P_RFC2292; \
1535 	if (optval) \
1536 		in6p->inp_flags |= (bit); \
1537 	else \
1538 		in6p->inp_flags &= ~(bit); \
1539 	INP_WUNLOCK(in6p); \
1540 } while (/*CONSTCOND*/ 0)
1541 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1542 
1543 #define OPTSET2(bit, val) do {						\
1544 	INP_WLOCK(in6p);						\
1545 	if (val)							\
1546 		in6p->inp_flags2 |= bit;				\
1547 	else								\
1548 		in6p->inp_flags2 &= ~bit;				\
1549 	INP_WUNLOCK(in6p);						\
1550 } while (0)
1551 #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0)
1552 
1553 				case IPV6_RECVPKTINFO:
1554 					/* cannot mix with RFC2292 */
1555 					if (OPTBIT(IN6P_RFC2292)) {
1556 						error = EINVAL;
1557 						break;
1558 					}
1559 					OPTSET(IN6P_PKTINFO);
1560 					break;
1561 
1562 				case IPV6_HOPLIMIT:
1563 				{
1564 					struct ip6_pktopts **optp;
1565 
1566 					/* cannot mix with RFC2292 */
1567 					if (OPTBIT(IN6P_RFC2292)) {
1568 						error = EINVAL;
1569 						break;
1570 					}
1571 					optp = &in6p->in6p_outputopts;
1572 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1573 					    (u_char *)&optval, sizeof(optval),
1574 					    optp, (td != NULL) ? td->td_ucred :
1575 					    NULL, uproto);
1576 					break;
1577 				}
1578 
1579 				case IPV6_RECVHOPLIMIT:
1580 					/* cannot mix with RFC2292 */
1581 					if (OPTBIT(IN6P_RFC2292)) {
1582 						error = EINVAL;
1583 						break;
1584 					}
1585 					OPTSET(IN6P_HOPLIMIT);
1586 					break;
1587 
1588 				case IPV6_RECVHOPOPTS:
1589 					/* cannot mix with RFC2292 */
1590 					if (OPTBIT(IN6P_RFC2292)) {
1591 						error = EINVAL;
1592 						break;
1593 					}
1594 					OPTSET(IN6P_HOPOPTS);
1595 					break;
1596 
1597 				case IPV6_RECVDSTOPTS:
1598 					/* cannot mix with RFC2292 */
1599 					if (OPTBIT(IN6P_RFC2292)) {
1600 						error = EINVAL;
1601 						break;
1602 					}
1603 					OPTSET(IN6P_DSTOPTS);
1604 					break;
1605 
1606 				case IPV6_RECVRTHDRDSTOPTS:
1607 					/* cannot mix with RFC2292 */
1608 					if (OPTBIT(IN6P_RFC2292)) {
1609 						error = EINVAL;
1610 						break;
1611 					}
1612 					OPTSET(IN6P_RTHDRDSTOPTS);
1613 					break;
1614 
1615 				case IPV6_RECVRTHDR:
1616 					/* cannot mix with RFC2292 */
1617 					if (OPTBIT(IN6P_RFC2292)) {
1618 						error = EINVAL;
1619 						break;
1620 					}
1621 					OPTSET(IN6P_RTHDR);
1622 					break;
1623 
1624 				case IPV6_RECVPATHMTU:
1625 					/*
1626 					 * We ignore this option for TCP
1627 					 * sockets.
1628 					 * (RFC3542 leaves this case
1629 					 * unspecified.)
1630 					 */
1631 					if (uproto != IPPROTO_TCP)
1632 						OPTSET(IN6P_MTU);
1633 					break;
1634 
1635 				case IPV6_RECVFLOWID:
1636 					OPTSET2(INP_RECVFLOWID, optval);
1637 					break;
1638 
1639 #ifdef	RSS
1640 				case IPV6_RECVRSSBUCKETID:
1641 					OPTSET2(INP_RECVRSSBUCKETID, optval);
1642 					break;
1643 #endif
1644 
1645 				case IPV6_V6ONLY:
1646 					/*
1647 					 * make setsockopt(IPV6_V6ONLY)
1648 					 * available only prior to bind(2).
1649 					 * see ipng mailing list, Jun 22 2001.
1650 					 */
1651 					if (in6p->inp_lport ||
1652 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1653 						error = EINVAL;
1654 						break;
1655 					}
1656 					OPTSET(IN6P_IPV6_V6ONLY);
1657 					if (optval)
1658 						in6p->inp_vflag &= ~INP_IPV4;
1659 					else
1660 						in6p->inp_vflag |= INP_IPV4;
1661 					break;
1662 				case IPV6_RECVTCLASS:
1663 					/* cannot mix with RFC2292 XXX */
1664 					if (OPTBIT(IN6P_RFC2292)) {
1665 						error = EINVAL;
1666 						break;
1667 					}
1668 					OPTSET(IN6P_TCLASS);
1669 					break;
1670 				case IPV6_AUTOFLOWLABEL:
1671 					OPTSET(IN6P_AUTOFLOWLABEL);
1672 					break;
1673 
1674 				case IPV6_BINDANY:
1675 					OPTSET(INP_BINDANY);
1676 					break;
1677 
1678 				case IPV6_BINDMULTI:
1679 					OPTSET2(INP_BINDMULTI, optval);
1680 					break;
1681 #ifdef	RSS
1682 				case IPV6_RSS_LISTEN_BUCKET:
1683 					if ((optval >= 0) &&
1684 					    (optval < rss_getnumbuckets())) {
1685 						in6p->inp_rss_listen_bucket = optval;
1686 						OPTSET2(INP_RSS_BUCKET_SET, 1);
1687 					} else {
1688 						error = EINVAL;
1689 					}
1690 					break;
1691 #endif
1692 				}
1693 				break;
1694 
1695 			case IPV6_TCLASS:
1696 			case IPV6_DONTFRAG:
1697 			case IPV6_USE_MIN_MTU:
1698 			case IPV6_PREFER_TEMPADDR:
1699 				if (optlen != sizeof(optval)) {
1700 					error = EINVAL;
1701 					break;
1702 				}
1703 				error = sooptcopyin(sopt, &optval,
1704 					sizeof optval, sizeof optval);
1705 				if (error)
1706 					break;
1707 				{
1708 					struct ip6_pktopts **optp;
1709 					optp = &in6p->in6p_outputopts;
1710 					error = ip6_pcbopt(optname,
1711 					    (u_char *)&optval, sizeof(optval),
1712 					    optp, (td != NULL) ? td->td_ucred :
1713 					    NULL, uproto);
1714 					break;
1715 				}
1716 
1717 			case IPV6_2292PKTINFO:
1718 			case IPV6_2292HOPLIMIT:
1719 			case IPV6_2292HOPOPTS:
1720 			case IPV6_2292DSTOPTS:
1721 			case IPV6_2292RTHDR:
1722 				/* RFC 2292 */
1723 				if (optlen != sizeof(int)) {
1724 					error = EINVAL;
1725 					break;
1726 				}
1727 				error = sooptcopyin(sopt, &optval,
1728 					sizeof optval, sizeof optval);
1729 				if (error)
1730 					break;
1731 				switch (optname) {
1732 				case IPV6_2292PKTINFO:
1733 					OPTSET2292(IN6P_PKTINFO);
1734 					break;
1735 				case IPV6_2292HOPLIMIT:
1736 					OPTSET2292(IN6P_HOPLIMIT);
1737 					break;
1738 				case IPV6_2292HOPOPTS:
1739 					/*
1740 					 * Check super-user privilege.
1741 					 * See comments for IPV6_RECVHOPOPTS.
1742 					 */
1743 					if (td != NULL) {
1744 						error = priv_check(td,
1745 						    PRIV_NETINET_SETHDROPTS);
1746 						if (error)
1747 							return (error);
1748 					}
1749 					OPTSET2292(IN6P_HOPOPTS);
1750 					break;
1751 				case IPV6_2292DSTOPTS:
1752 					if (td != NULL) {
1753 						error = priv_check(td,
1754 						    PRIV_NETINET_SETHDROPTS);
1755 						if (error)
1756 							return (error);
1757 					}
1758 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1759 					break;
1760 				case IPV6_2292RTHDR:
1761 					OPTSET2292(IN6P_RTHDR);
1762 					break;
1763 				}
1764 				break;
1765 			case IPV6_PKTINFO:
1766 			case IPV6_HOPOPTS:
1767 			case IPV6_RTHDR:
1768 			case IPV6_DSTOPTS:
1769 			case IPV6_RTHDRDSTOPTS:
1770 			case IPV6_NEXTHOP:
1771 			{
1772 				/* new advanced API (RFC3542) */
1773 				u_char *optbuf;
1774 				u_char optbuf_storage[MCLBYTES];
1775 				int optlen;
1776 				struct ip6_pktopts **optp;
1777 
1778 				/* cannot mix with RFC2292 */
1779 				if (OPTBIT(IN6P_RFC2292)) {
1780 					error = EINVAL;
1781 					break;
1782 				}
1783 
1784 				/*
1785 				 * We only ensure valsize is not too large
1786 				 * here.  Further validation will be done
1787 				 * later.
1788 				 */
1789 				error = sooptcopyin(sopt, optbuf_storage,
1790 				    sizeof(optbuf_storage), 0);
1791 				if (error)
1792 					break;
1793 				optlen = sopt->sopt_valsize;
1794 				optbuf = optbuf_storage;
1795 				optp = &in6p->in6p_outputopts;
1796 				error = ip6_pcbopt(optname, optbuf, optlen,
1797 				    optp, (td != NULL) ? td->td_ucred : NULL,
1798 				    uproto);
1799 				break;
1800 			}
1801 #undef OPTSET
1802 
1803 			case IPV6_MULTICAST_IF:
1804 			case IPV6_MULTICAST_HOPS:
1805 			case IPV6_MULTICAST_LOOP:
1806 			case IPV6_JOIN_GROUP:
1807 			case IPV6_LEAVE_GROUP:
1808 			case IPV6_MSFILTER:
1809 			case MCAST_BLOCK_SOURCE:
1810 			case MCAST_UNBLOCK_SOURCE:
1811 			case MCAST_JOIN_GROUP:
1812 			case MCAST_LEAVE_GROUP:
1813 			case MCAST_JOIN_SOURCE_GROUP:
1814 			case MCAST_LEAVE_SOURCE_GROUP:
1815 				error = ip6_setmoptions(in6p, sopt);
1816 				break;
1817 
1818 			case IPV6_PORTRANGE:
1819 				error = sooptcopyin(sopt, &optval,
1820 				    sizeof optval, sizeof optval);
1821 				if (error)
1822 					break;
1823 
1824 				INP_WLOCK(in6p);
1825 				switch (optval) {
1826 				case IPV6_PORTRANGE_DEFAULT:
1827 					in6p->inp_flags &= ~(INP_LOWPORT);
1828 					in6p->inp_flags &= ~(INP_HIGHPORT);
1829 					break;
1830 
1831 				case IPV6_PORTRANGE_HIGH:
1832 					in6p->inp_flags &= ~(INP_LOWPORT);
1833 					in6p->inp_flags |= INP_HIGHPORT;
1834 					break;
1835 
1836 				case IPV6_PORTRANGE_LOW:
1837 					in6p->inp_flags &= ~(INP_HIGHPORT);
1838 					in6p->inp_flags |= INP_LOWPORT;
1839 					break;
1840 
1841 				default:
1842 					error = EINVAL;
1843 					break;
1844 				}
1845 				INP_WUNLOCK(in6p);
1846 				break;
1847 
1848 #ifdef IPSEC
1849 			case IPV6_IPSEC_POLICY:
1850 			{
1851 				caddr_t req;
1852 				struct mbuf *m;
1853 
1854 				if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1855 					break;
1856 				if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1857 					break;
1858 				req = mtod(m, caddr_t);
1859 				error = ipsec_set_policy(in6p, optname, req,
1860 				    m->m_len, (sopt->sopt_td != NULL) ?
1861 				    sopt->sopt_td->td_ucred : NULL);
1862 				m_freem(m);
1863 				break;
1864 			}
1865 #endif /* IPSEC */
1866 
1867 			default:
1868 				error = ENOPROTOOPT;
1869 				break;
1870 			}
1871 			break;
1872 
1873 		case SOPT_GET:
1874 			switch (optname) {
1875 
1876 			case IPV6_2292PKTOPTIONS:
1877 #ifdef IPV6_PKTOPTIONS
1878 			case IPV6_PKTOPTIONS:
1879 #endif
1880 				/*
1881 				 * RFC3542 (effectively) deprecated the
1882 				 * semantics of the 2292-style pktoptions.
1883 				 * Since it was not reliable in nature (i.e.,
1884 				 * applications had to expect the lack of some
1885 				 * information after all), it would make sense
1886 				 * to simplify this part by always returning
1887 				 * empty data.
1888 				 */
1889 				sopt->sopt_valsize = 0;
1890 				break;
1891 
1892 			case IPV6_RECVHOPOPTS:
1893 			case IPV6_RECVDSTOPTS:
1894 			case IPV6_RECVRTHDRDSTOPTS:
1895 			case IPV6_UNICAST_HOPS:
1896 			case IPV6_RECVPKTINFO:
1897 			case IPV6_RECVHOPLIMIT:
1898 			case IPV6_RECVRTHDR:
1899 			case IPV6_RECVPATHMTU:
1900 
1901 			case IPV6_V6ONLY:
1902 			case IPV6_PORTRANGE:
1903 			case IPV6_RECVTCLASS:
1904 			case IPV6_AUTOFLOWLABEL:
1905 			case IPV6_BINDANY:
1906 			case IPV6_FLOWID:
1907 			case IPV6_FLOWTYPE:
1908 			case IPV6_RECVFLOWID:
1909 #ifdef	RSS
1910 			case IPV6_RSSBUCKETID:
1911 			case IPV6_RECVRSSBUCKETID:
1912 #endif
1913 			case IPV6_BINDMULTI:
1914 				switch (optname) {
1915 
1916 				case IPV6_RECVHOPOPTS:
1917 					optval = OPTBIT(IN6P_HOPOPTS);
1918 					break;
1919 
1920 				case IPV6_RECVDSTOPTS:
1921 					optval = OPTBIT(IN6P_DSTOPTS);
1922 					break;
1923 
1924 				case IPV6_RECVRTHDRDSTOPTS:
1925 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1926 					break;
1927 
1928 				case IPV6_UNICAST_HOPS:
1929 					optval = in6p->in6p_hops;
1930 					break;
1931 
1932 				case IPV6_RECVPKTINFO:
1933 					optval = OPTBIT(IN6P_PKTINFO);
1934 					break;
1935 
1936 				case IPV6_RECVHOPLIMIT:
1937 					optval = OPTBIT(IN6P_HOPLIMIT);
1938 					break;
1939 
1940 				case IPV6_RECVRTHDR:
1941 					optval = OPTBIT(IN6P_RTHDR);
1942 					break;
1943 
1944 				case IPV6_RECVPATHMTU:
1945 					optval = OPTBIT(IN6P_MTU);
1946 					break;
1947 
1948 				case IPV6_V6ONLY:
1949 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
1950 					break;
1951 
1952 				case IPV6_PORTRANGE:
1953 				    {
1954 					int flags;
1955 					flags = in6p->inp_flags;
1956 					if (flags & INP_HIGHPORT)
1957 						optval = IPV6_PORTRANGE_HIGH;
1958 					else if (flags & INP_LOWPORT)
1959 						optval = IPV6_PORTRANGE_LOW;
1960 					else
1961 						optval = 0;
1962 					break;
1963 				    }
1964 				case IPV6_RECVTCLASS:
1965 					optval = OPTBIT(IN6P_TCLASS);
1966 					break;
1967 
1968 				case IPV6_AUTOFLOWLABEL:
1969 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1970 					break;
1971 
1972 				case IPV6_BINDANY:
1973 					optval = OPTBIT(INP_BINDANY);
1974 					break;
1975 
1976 				case IPV6_FLOWID:
1977 					optval = in6p->inp_flowid;
1978 					break;
1979 
1980 				case IPV6_FLOWTYPE:
1981 					optval = in6p->inp_flowtype;
1982 					break;
1983 
1984 				case IPV6_RECVFLOWID:
1985 					optval = OPTBIT2(INP_RECVFLOWID);
1986 					break;
1987 #ifdef	RSS
1988 				case IPV6_RSSBUCKETID:
1989 					retval =
1990 					    rss_hash2bucket(in6p->inp_flowid,
1991 					    in6p->inp_flowtype,
1992 					    &rss_bucket);
1993 					if (retval == 0)
1994 						optval = rss_bucket;
1995 					else
1996 						error = EINVAL;
1997 					break;
1998 
1999 				case IPV6_RECVRSSBUCKETID:
2000 					optval = OPTBIT2(INP_RECVRSSBUCKETID);
2001 					break;
2002 #endif
2003 
2004 				case IPV6_BINDMULTI:
2005 					optval = OPTBIT2(INP_BINDMULTI);
2006 					break;
2007 
2008 				}
2009 				if (error)
2010 					break;
2011 				error = sooptcopyout(sopt, &optval,
2012 					sizeof optval);
2013 				break;
2014 
2015 			case IPV6_PATHMTU:
2016 			{
2017 				u_long pmtu = 0;
2018 				struct ip6_mtuinfo mtuinfo;
2019 
2020 				if (!(so->so_state & SS_ISCONNECTED))
2021 					return (ENOTCONN);
2022 				/*
2023 				 * XXX: we dot not consider the case of source
2024 				 * routing, or optional information to specify
2025 				 * the outgoing interface.
2026 				 */
2027 				error = ip6_getpmtu_ctl(so->so_fibnum,
2028 				    &in6p->in6p_faddr, &pmtu);
2029 				if (error)
2030 					break;
2031 				if (pmtu > IPV6_MAXPACKET)
2032 					pmtu = IPV6_MAXPACKET;
2033 
2034 				bzero(&mtuinfo, sizeof(mtuinfo));
2035 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2036 				optdata = (void *)&mtuinfo;
2037 				optdatalen = sizeof(mtuinfo);
2038 				error = sooptcopyout(sopt, optdata,
2039 				    optdatalen);
2040 				break;
2041 			}
2042 
2043 			case IPV6_2292PKTINFO:
2044 			case IPV6_2292HOPLIMIT:
2045 			case IPV6_2292HOPOPTS:
2046 			case IPV6_2292RTHDR:
2047 			case IPV6_2292DSTOPTS:
2048 				switch (optname) {
2049 				case IPV6_2292PKTINFO:
2050 					optval = OPTBIT(IN6P_PKTINFO);
2051 					break;
2052 				case IPV6_2292HOPLIMIT:
2053 					optval = OPTBIT(IN6P_HOPLIMIT);
2054 					break;
2055 				case IPV6_2292HOPOPTS:
2056 					optval = OPTBIT(IN6P_HOPOPTS);
2057 					break;
2058 				case IPV6_2292RTHDR:
2059 					optval = OPTBIT(IN6P_RTHDR);
2060 					break;
2061 				case IPV6_2292DSTOPTS:
2062 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2063 					break;
2064 				}
2065 				error = sooptcopyout(sopt, &optval,
2066 				    sizeof optval);
2067 				break;
2068 			case IPV6_PKTINFO:
2069 			case IPV6_HOPOPTS:
2070 			case IPV6_RTHDR:
2071 			case IPV6_DSTOPTS:
2072 			case IPV6_RTHDRDSTOPTS:
2073 			case IPV6_NEXTHOP:
2074 			case IPV6_TCLASS:
2075 			case IPV6_DONTFRAG:
2076 			case IPV6_USE_MIN_MTU:
2077 			case IPV6_PREFER_TEMPADDR:
2078 				error = ip6_getpcbopt(in6p->in6p_outputopts,
2079 				    optname, sopt);
2080 				break;
2081 
2082 			case IPV6_MULTICAST_IF:
2083 			case IPV6_MULTICAST_HOPS:
2084 			case IPV6_MULTICAST_LOOP:
2085 			case IPV6_MSFILTER:
2086 				error = ip6_getmoptions(in6p, sopt);
2087 				break;
2088 
2089 #ifdef IPSEC
2090 			case IPV6_IPSEC_POLICY:
2091 			  {
2092 				caddr_t req = NULL;
2093 				size_t len = 0;
2094 				struct mbuf *m = NULL;
2095 				struct mbuf **mp = &m;
2096 				size_t ovalsize = sopt->sopt_valsize;
2097 				caddr_t oval = (caddr_t)sopt->sopt_val;
2098 
2099 				error = soopt_getm(sopt, &m); /* XXX */
2100 				if (error != 0)
2101 					break;
2102 				error = soopt_mcopyin(sopt, m); /* XXX */
2103 				if (error != 0)
2104 					break;
2105 				sopt->sopt_valsize = ovalsize;
2106 				sopt->sopt_val = oval;
2107 				if (m) {
2108 					req = mtod(m, caddr_t);
2109 					len = m->m_len;
2110 				}
2111 				error = ipsec_get_policy(in6p, req, len, mp);
2112 				if (error == 0)
2113 					error = soopt_mcopyout(sopt, m); /* XXX */
2114 				if (error == 0 && m)
2115 					m_freem(m);
2116 				break;
2117 			  }
2118 #endif /* IPSEC */
2119 
2120 			default:
2121 				error = ENOPROTOOPT;
2122 				break;
2123 			}
2124 			break;
2125 		}
2126 	}
2127 	return (error);
2128 }
2129 
2130 int
2131 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2132 {
2133 	int error = 0, optval, optlen;
2134 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2135 	struct inpcb *in6p = sotoinpcb(so);
2136 	int level, op, optname;
2137 
2138 	level = sopt->sopt_level;
2139 	op = sopt->sopt_dir;
2140 	optname = sopt->sopt_name;
2141 	optlen = sopt->sopt_valsize;
2142 
2143 	if (level != IPPROTO_IPV6) {
2144 		return (EINVAL);
2145 	}
2146 
2147 	switch (optname) {
2148 	case IPV6_CHECKSUM:
2149 		/*
2150 		 * For ICMPv6 sockets, no modification allowed for checksum
2151 		 * offset, permit "no change" values to help existing apps.
2152 		 *
2153 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2154 		 * for an ICMPv6 socket will fail."
2155 		 * The current behavior does not meet RFC3542.
2156 		 */
2157 		switch (op) {
2158 		case SOPT_SET:
2159 			if (optlen != sizeof(int)) {
2160 				error = EINVAL;
2161 				break;
2162 			}
2163 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2164 					    sizeof(optval));
2165 			if (error)
2166 				break;
2167 			if ((optval % 2) != 0) {
2168 				/* the API assumes even offset values */
2169 				error = EINVAL;
2170 			} else if (so->so_proto->pr_protocol ==
2171 			    IPPROTO_ICMPV6) {
2172 				if (optval != icmp6off)
2173 					error = EINVAL;
2174 			} else
2175 				in6p->in6p_cksum = optval;
2176 			break;
2177 
2178 		case SOPT_GET:
2179 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2180 				optval = icmp6off;
2181 			else
2182 				optval = in6p->in6p_cksum;
2183 
2184 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2185 			break;
2186 
2187 		default:
2188 			error = EINVAL;
2189 			break;
2190 		}
2191 		break;
2192 
2193 	default:
2194 		error = ENOPROTOOPT;
2195 		break;
2196 	}
2197 
2198 	return (error);
2199 }
2200 
2201 /*
2202  * Set up IP6 options in pcb for insertion in output packets or
2203  * specifying behavior of outgoing packets.
2204  */
2205 static int
2206 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2207     struct socket *so, struct sockopt *sopt)
2208 {
2209 	struct ip6_pktopts *opt = *pktopt;
2210 	int error = 0;
2211 	struct thread *td = sopt->sopt_td;
2212 
2213 	/* turn off any old options. */
2214 	if (opt) {
2215 #ifdef DIAGNOSTIC
2216 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2217 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2218 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2219 			printf("ip6_pcbopts: all specified options are cleared.\n");
2220 #endif
2221 		ip6_clearpktopts(opt, -1);
2222 	} else
2223 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2224 	*pktopt = NULL;
2225 
2226 	if (!m || m->m_len == 0) {
2227 		/*
2228 		 * Only turning off any previous options, regardless of
2229 		 * whether the opt is just created or given.
2230 		 */
2231 		free(opt, M_IP6OPT);
2232 		return (0);
2233 	}
2234 
2235 	/*  set options specified by user. */
2236 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2237 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2238 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2239 		free(opt, M_IP6OPT);
2240 		return (error);
2241 	}
2242 	*pktopt = opt;
2243 	return (0);
2244 }
2245 
2246 /*
2247  * initialize ip6_pktopts.  beware that there are non-zero default values in
2248  * the struct.
2249  */
2250 void
2251 ip6_initpktopts(struct ip6_pktopts *opt)
2252 {
2253 
2254 	bzero(opt, sizeof(*opt));
2255 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2256 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2257 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2258 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2259 }
2260 
2261 static int
2262 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2263     struct ucred *cred, int uproto)
2264 {
2265 	struct ip6_pktopts *opt;
2266 
2267 	if (*pktopt == NULL) {
2268 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2269 		    M_WAITOK);
2270 		ip6_initpktopts(*pktopt);
2271 	}
2272 	opt = *pktopt;
2273 
2274 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2275 }
2276 
2277 static int
2278 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2279 {
2280 	void *optdata = NULL;
2281 	int optdatalen = 0;
2282 	struct ip6_ext *ip6e;
2283 	int error = 0;
2284 	struct in6_pktinfo null_pktinfo;
2285 	int deftclass = 0, on;
2286 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2287 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2288 
2289 	switch (optname) {
2290 	case IPV6_PKTINFO:
2291 		optdata = (void *)&null_pktinfo;
2292 		if (pktopt && pktopt->ip6po_pktinfo) {
2293 			bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2294 			    sizeof(null_pktinfo));
2295 			in6_clearscope(&null_pktinfo.ipi6_addr);
2296 		} else {
2297 			/* XXX: we don't have to do this every time... */
2298 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2299 		}
2300 		optdatalen = sizeof(struct in6_pktinfo);
2301 		break;
2302 	case IPV6_TCLASS:
2303 		if (pktopt && pktopt->ip6po_tclass >= 0)
2304 			optdata = (void *)&pktopt->ip6po_tclass;
2305 		else
2306 			optdata = (void *)&deftclass;
2307 		optdatalen = sizeof(int);
2308 		break;
2309 	case IPV6_HOPOPTS:
2310 		if (pktopt && pktopt->ip6po_hbh) {
2311 			optdata = (void *)pktopt->ip6po_hbh;
2312 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2313 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2314 		}
2315 		break;
2316 	case IPV6_RTHDR:
2317 		if (pktopt && pktopt->ip6po_rthdr) {
2318 			optdata = (void *)pktopt->ip6po_rthdr;
2319 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2320 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2321 		}
2322 		break;
2323 	case IPV6_RTHDRDSTOPTS:
2324 		if (pktopt && pktopt->ip6po_dest1) {
2325 			optdata = (void *)pktopt->ip6po_dest1;
2326 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2327 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2328 		}
2329 		break;
2330 	case IPV6_DSTOPTS:
2331 		if (pktopt && pktopt->ip6po_dest2) {
2332 			optdata = (void *)pktopt->ip6po_dest2;
2333 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2334 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2335 		}
2336 		break;
2337 	case IPV6_NEXTHOP:
2338 		if (pktopt && pktopt->ip6po_nexthop) {
2339 			optdata = (void *)pktopt->ip6po_nexthop;
2340 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2341 		}
2342 		break;
2343 	case IPV6_USE_MIN_MTU:
2344 		if (pktopt)
2345 			optdata = (void *)&pktopt->ip6po_minmtu;
2346 		else
2347 			optdata = (void *)&defminmtu;
2348 		optdatalen = sizeof(int);
2349 		break;
2350 	case IPV6_DONTFRAG:
2351 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2352 			on = 1;
2353 		else
2354 			on = 0;
2355 		optdata = (void *)&on;
2356 		optdatalen = sizeof(on);
2357 		break;
2358 	case IPV6_PREFER_TEMPADDR:
2359 		if (pktopt)
2360 			optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2361 		else
2362 			optdata = (void *)&defpreftemp;
2363 		optdatalen = sizeof(int);
2364 		break;
2365 	default:		/* should not happen */
2366 #ifdef DIAGNOSTIC
2367 		panic("ip6_getpcbopt: unexpected option\n");
2368 #endif
2369 		return (ENOPROTOOPT);
2370 	}
2371 
2372 	error = sooptcopyout(sopt, optdata, optdatalen);
2373 
2374 	return (error);
2375 }
2376 
2377 void
2378 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2379 {
2380 	if (pktopt == NULL)
2381 		return;
2382 
2383 	if (optname == -1 || optname == IPV6_PKTINFO) {
2384 		if (pktopt->ip6po_pktinfo)
2385 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2386 		pktopt->ip6po_pktinfo = NULL;
2387 	}
2388 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2389 		pktopt->ip6po_hlim = -1;
2390 	if (optname == -1 || optname == IPV6_TCLASS)
2391 		pktopt->ip6po_tclass = -1;
2392 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2393 		if (pktopt->ip6po_nextroute.ro_rt) {
2394 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2395 			pktopt->ip6po_nextroute.ro_rt = NULL;
2396 		}
2397 		if (pktopt->ip6po_nexthop)
2398 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2399 		pktopt->ip6po_nexthop = NULL;
2400 	}
2401 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2402 		if (pktopt->ip6po_hbh)
2403 			free(pktopt->ip6po_hbh, M_IP6OPT);
2404 		pktopt->ip6po_hbh = NULL;
2405 	}
2406 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2407 		if (pktopt->ip6po_dest1)
2408 			free(pktopt->ip6po_dest1, M_IP6OPT);
2409 		pktopt->ip6po_dest1 = NULL;
2410 	}
2411 	if (optname == -1 || optname == IPV6_RTHDR) {
2412 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2413 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2414 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2415 		if (pktopt->ip6po_route.ro_rt) {
2416 			RTFREE(pktopt->ip6po_route.ro_rt);
2417 			pktopt->ip6po_route.ro_rt = NULL;
2418 		}
2419 	}
2420 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2421 		if (pktopt->ip6po_dest2)
2422 			free(pktopt->ip6po_dest2, M_IP6OPT);
2423 		pktopt->ip6po_dest2 = NULL;
2424 	}
2425 }
2426 
2427 #define PKTOPT_EXTHDRCPY(type) \
2428 do {\
2429 	if (src->type) {\
2430 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2431 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2432 		if (dst->type == NULL && canwait == M_NOWAIT)\
2433 			goto bad;\
2434 		bcopy(src->type, dst->type, hlen);\
2435 	}\
2436 } while (/*CONSTCOND*/ 0)
2437 
2438 static int
2439 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2440 {
2441 	if (dst == NULL || src == NULL)  {
2442 		printf("ip6_clearpktopts: invalid argument\n");
2443 		return (EINVAL);
2444 	}
2445 
2446 	dst->ip6po_hlim = src->ip6po_hlim;
2447 	dst->ip6po_tclass = src->ip6po_tclass;
2448 	dst->ip6po_flags = src->ip6po_flags;
2449 	dst->ip6po_minmtu = src->ip6po_minmtu;
2450 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2451 	if (src->ip6po_pktinfo) {
2452 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2453 		    M_IP6OPT, canwait);
2454 		if (dst->ip6po_pktinfo == NULL)
2455 			goto bad;
2456 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2457 	}
2458 	if (src->ip6po_nexthop) {
2459 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2460 		    M_IP6OPT, canwait);
2461 		if (dst->ip6po_nexthop == NULL)
2462 			goto bad;
2463 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2464 		    src->ip6po_nexthop->sa_len);
2465 	}
2466 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2467 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2468 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2469 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2470 	return (0);
2471 
2472   bad:
2473 	ip6_clearpktopts(dst, -1);
2474 	return (ENOBUFS);
2475 }
2476 #undef PKTOPT_EXTHDRCPY
2477 
2478 struct ip6_pktopts *
2479 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2480 {
2481 	int error;
2482 	struct ip6_pktopts *dst;
2483 
2484 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2485 	if (dst == NULL)
2486 		return (NULL);
2487 	ip6_initpktopts(dst);
2488 
2489 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2490 		free(dst, M_IP6OPT);
2491 		return (NULL);
2492 	}
2493 
2494 	return (dst);
2495 }
2496 
2497 void
2498 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2499 {
2500 	if (pktopt == NULL)
2501 		return;
2502 
2503 	ip6_clearpktopts(pktopt, -1);
2504 
2505 	free(pktopt, M_IP6OPT);
2506 }
2507 
2508 /*
2509  * Set IPv6 outgoing packet options based on advanced API.
2510  */
2511 int
2512 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2513     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2514 {
2515 	struct cmsghdr *cm = NULL;
2516 
2517 	if (control == NULL || opt == NULL)
2518 		return (EINVAL);
2519 
2520 	ip6_initpktopts(opt);
2521 	if (stickyopt) {
2522 		int error;
2523 
2524 		/*
2525 		 * If stickyopt is provided, make a local copy of the options
2526 		 * for this particular packet, then override them by ancillary
2527 		 * objects.
2528 		 * XXX: copypktopts() does not copy the cached route to a next
2529 		 * hop (if any).  This is not very good in terms of efficiency,
2530 		 * but we can allow this since this option should be rarely
2531 		 * used.
2532 		 */
2533 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2534 			return (error);
2535 	}
2536 
2537 	/*
2538 	 * XXX: Currently, we assume all the optional information is stored
2539 	 * in a single mbuf.
2540 	 */
2541 	if (control->m_next)
2542 		return (EINVAL);
2543 
2544 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2545 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2546 		int error;
2547 
2548 		if (control->m_len < CMSG_LEN(0))
2549 			return (EINVAL);
2550 
2551 		cm = mtod(control, struct cmsghdr *);
2552 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2553 			return (EINVAL);
2554 		if (cm->cmsg_level != IPPROTO_IPV6)
2555 			continue;
2556 
2557 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2558 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2559 		if (error)
2560 			return (error);
2561 	}
2562 
2563 	return (0);
2564 }
2565 
2566 /*
2567  * Set a particular packet option, as a sticky option or an ancillary data
2568  * item.  "len" can be 0 only when it's a sticky option.
2569  * We have 4 cases of combination of "sticky" and "cmsg":
2570  * "sticky=0, cmsg=0": impossible
2571  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2572  * "sticky=1, cmsg=0": RFC3542 socket option
2573  * "sticky=1, cmsg=1": RFC2292 socket option
2574  */
2575 static int
2576 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2577     struct ucred *cred, int sticky, int cmsg, int uproto)
2578 {
2579 	int minmtupolicy, preftemp;
2580 	int error;
2581 
2582 	if (!sticky && !cmsg) {
2583 #ifdef DIAGNOSTIC
2584 		printf("ip6_setpktopt: impossible case\n");
2585 #endif
2586 		return (EINVAL);
2587 	}
2588 
2589 	/*
2590 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2591 	 * not be specified in the context of RFC3542.  Conversely,
2592 	 * RFC3542 types should not be specified in the context of RFC2292.
2593 	 */
2594 	if (!cmsg) {
2595 		switch (optname) {
2596 		case IPV6_2292PKTINFO:
2597 		case IPV6_2292HOPLIMIT:
2598 		case IPV6_2292NEXTHOP:
2599 		case IPV6_2292HOPOPTS:
2600 		case IPV6_2292DSTOPTS:
2601 		case IPV6_2292RTHDR:
2602 		case IPV6_2292PKTOPTIONS:
2603 			return (ENOPROTOOPT);
2604 		}
2605 	}
2606 	if (sticky && cmsg) {
2607 		switch (optname) {
2608 		case IPV6_PKTINFO:
2609 		case IPV6_HOPLIMIT:
2610 		case IPV6_NEXTHOP:
2611 		case IPV6_HOPOPTS:
2612 		case IPV6_DSTOPTS:
2613 		case IPV6_RTHDRDSTOPTS:
2614 		case IPV6_RTHDR:
2615 		case IPV6_USE_MIN_MTU:
2616 		case IPV6_DONTFRAG:
2617 		case IPV6_TCLASS:
2618 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2619 			return (ENOPROTOOPT);
2620 		}
2621 	}
2622 
2623 	switch (optname) {
2624 	case IPV6_2292PKTINFO:
2625 	case IPV6_PKTINFO:
2626 	{
2627 		struct ifnet *ifp = NULL;
2628 		struct in6_pktinfo *pktinfo;
2629 
2630 		if (len != sizeof(struct in6_pktinfo))
2631 			return (EINVAL);
2632 
2633 		pktinfo = (struct in6_pktinfo *)buf;
2634 
2635 		/*
2636 		 * An application can clear any sticky IPV6_PKTINFO option by
2637 		 * doing a "regular" setsockopt with ipi6_addr being
2638 		 * in6addr_any and ipi6_ifindex being zero.
2639 		 * [RFC 3542, Section 6]
2640 		 */
2641 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2642 		    pktinfo->ipi6_ifindex == 0 &&
2643 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2644 			ip6_clearpktopts(opt, optname);
2645 			break;
2646 		}
2647 
2648 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2649 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2650 			return (EINVAL);
2651 		}
2652 		if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2653 			return (EINVAL);
2654 		/* validate the interface index if specified. */
2655 		if (pktinfo->ipi6_ifindex > V_if_index)
2656 			 return (ENXIO);
2657 		if (pktinfo->ipi6_ifindex) {
2658 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2659 			if (ifp == NULL)
2660 				return (ENXIO);
2661 		}
2662 		if (ifp != NULL && (
2663 		    ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED))
2664 			return (ENETDOWN);
2665 
2666 		if (ifp != NULL &&
2667 		    !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2668 			struct in6_ifaddr *ia;
2669 
2670 			in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2671 			ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2672 			if (ia == NULL)
2673 				return (EADDRNOTAVAIL);
2674 			ifa_free(&ia->ia_ifa);
2675 		}
2676 		/*
2677 		 * We store the address anyway, and let in6_selectsrc()
2678 		 * validate the specified address.  This is because ipi6_addr
2679 		 * may not have enough information about its scope zone, and
2680 		 * we may need additional information (such as outgoing
2681 		 * interface or the scope zone of a destination address) to
2682 		 * disambiguate the scope.
2683 		 * XXX: the delay of the validation may confuse the
2684 		 * application when it is used as a sticky option.
2685 		 */
2686 		if (opt->ip6po_pktinfo == NULL) {
2687 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2688 			    M_IP6OPT, M_NOWAIT);
2689 			if (opt->ip6po_pktinfo == NULL)
2690 				return (ENOBUFS);
2691 		}
2692 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2693 		break;
2694 	}
2695 
2696 	case IPV6_2292HOPLIMIT:
2697 	case IPV6_HOPLIMIT:
2698 	{
2699 		int *hlimp;
2700 
2701 		/*
2702 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2703 		 * to simplify the ordering among hoplimit options.
2704 		 */
2705 		if (optname == IPV6_HOPLIMIT && sticky)
2706 			return (ENOPROTOOPT);
2707 
2708 		if (len != sizeof(int))
2709 			return (EINVAL);
2710 		hlimp = (int *)buf;
2711 		if (*hlimp < -1 || *hlimp > 255)
2712 			return (EINVAL);
2713 
2714 		opt->ip6po_hlim = *hlimp;
2715 		break;
2716 	}
2717 
2718 	case IPV6_TCLASS:
2719 	{
2720 		int tclass;
2721 
2722 		if (len != sizeof(int))
2723 			return (EINVAL);
2724 		tclass = *(int *)buf;
2725 		if (tclass < -1 || tclass > 255)
2726 			return (EINVAL);
2727 
2728 		opt->ip6po_tclass = tclass;
2729 		break;
2730 	}
2731 
2732 	case IPV6_2292NEXTHOP:
2733 	case IPV6_NEXTHOP:
2734 		if (cred != NULL) {
2735 			error = priv_check_cred(cred,
2736 			    PRIV_NETINET_SETHDROPTS, 0);
2737 			if (error)
2738 				return (error);
2739 		}
2740 
2741 		if (len == 0) {	/* just remove the option */
2742 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2743 			break;
2744 		}
2745 
2746 		/* check if cmsg_len is large enough for sa_len */
2747 		if (len < sizeof(struct sockaddr) || len < *buf)
2748 			return (EINVAL);
2749 
2750 		switch (((struct sockaddr *)buf)->sa_family) {
2751 		case AF_INET6:
2752 		{
2753 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2754 			int error;
2755 
2756 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2757 				return (EINVAL);
2758 
2759 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2760 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2761 				return (EINVAL);
2762 			}
2763 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2764 			    != 0) {
2765 				return (error);
2766 			}
2767 			break;
2768 		}
2769 		case AF_LINK:	/* should eventually be supported */
2770 		default:
2771 			return (EAFNOSUPPORT);
2772 		}
2773 
2774 		/* turn off the previous option, then set the new option. */
2775 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2776 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2777 		if (opt->ip6po_nexthop == NULL)
2778 			return (ENOBUFS);
2779 		bcopy(buf, opt->ip6po_nexthop, *buf);
2780 		break;
2781 
2782 	case IPV6_2292HOPOPTS:
2783 	case IPV6_HOPOPTS:
2784 	{
2785 		struct ip6_hbh *hbh;
2786 		int hbhlen;
2787 
2788 		/*
2789 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2790 		 * options, since per-option restriction has too much
2791 		 * overhead.
2792 		 */
2793 		if (cred != NULL) {
2794 			error = priv_check_cred(cred,
2795 			    PRIV_NETINET_SETHDROPTS, 0);
2796 			if (error)
2797 				return (error);
2798 		}
2799 
2800 		if (len == 0) {
2801 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2802 			break;	/* just remove the option */
2803 		}
2804 
2805 		/* message length validation */
2806 		if (len < sizeof(struct ip6_hbh))
2807 			return (EINVAL);
2808 		hbh = (struct ip6_hbh *)buf;
2809 		hbhlen = (hbh->ip6h_len + 1) << 3;
2810 		if (len != hbhlen)
2811 			return (EINVAL);
2812 
2813 		/* turn off the previous option, then set the new option. */
2814 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2815 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2816 		if (opt->ip6po_hbh == NULL)
2817 			return (ENOBUFS);
2818 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
2819 
2820 		break;
2821 	}
2822 
2823 	case IPV6_2292DSTOPTS:
2824 	case IPV6_DSTOPTS:
2825 	case IPV6_RTHDRDSTOPTS:
2826 	{
2827 		struct ip6_dest *dest, **newdest = NULL;
2828 		int destlen;
2829 
2830 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2831 			error = priv_check_cred(cred,
2832 			    PRIV_NETINET_SETHDROPTS, 0);
2833 			if (error)
2834 				return (error);
2835 		}
2836 
2837 		if (len == 0) {
2838 			ip6_clearpktopts(opt, optname);
2839 			break;	/* just remove the option */
2840 		}
2841 
2842 		/* message length validation */
2843 		if (len < sizeof(struct ip6_dest))
2844 			return (EINVAL);
2845 		dest = (struct ip6_dest *)buf;
2846 		destlen = (dest->ip6d_len + 1) << 3;
2847 		if (len != destlen)
2848 			return (EINVAL);
2849 
2850 		/*
2851 		 * Determine the position that the destination options header
2852 		 * should be inserted; before or after the routing header.
2853 		 */
2854 		switch (optname) {
2855 		case IPV6_2292DSTOPTS:
2856 			/*
2857 			 * The old advacned API is ambiguous on this point.
2858 			 * Our approach is to determine the position based
2859 			 * according to the existence of a routing header.
2860 			 * Note, however, that this depends on the order of the
2861 			 * extension headers in the ancillary data; the 1st
2862 			 * part of the destination options header must appear
2863 			 * before the routing header in the ancillary data,
2864 			 * too.
2865 			 * RFC3542 solved the ambiguity by introducing
2866 			 * separate ancillary data or option types.
2867 			 */
2868 			if (opt->ip6po_rthdr == NULL)
2869 				newdest = &opt->ip6po_dest1;
2870 			else
2871 				newdest = &opt->ip6po_dest2;
2872 			break;
2873 		case IPV6_RTHDRDSTOPTS:
2874 			newdest = &opt->ip6po_dest1;
2875 			break;
2876 		case IPV6_DSTOPTS:
2877 			newdest = &opt->ip6po_dest2;
2878 			break;
2879 		}
2880 
2881 		/* turn off the previous option, then set the new option. */
2882 		ip6_clearpktopts(opt, optname);
2883 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2884 		if (*newdest == NULL)
2885 			return (ENOBUFS);
2886 		bcopy(dest, *newdest, destlen);
2887 
2888 		break;
2889 	}
2890 
2891 	case IPV6_2292RTHDR:
2892 	case IPV6_RTHDR:
2893 	{
2894 		struct ip6_rthdr *rth;
2895 		int rthlen;
2896 
2897 		if (len == 0) {
2898 			ip6_clearpktopts(opt, IPV6_RTHDR);
2899 			break;	/* just remove the option */
2900 		}
2901 
2902 		/* message length validation */
2903 		if (len < sizeof(struct ip6_rthdr))
2904 			return (EINVAL);
2905 		rth = (struct ip6_rthdr *)buf;
2906 		rthlen = (rth->ip6r_len + 1) << 3;
2907 		if (len != rthlen)
2908 			return (EINVAL);
2909 
2910 		switch (rth->ip6r_type) {
2911 		case IPV6_RTHDR_TYPE_0:
2912 			if (rth->ip6r_len == 0)	/* must contain one addr */
2913 				return (EINVAL);
2914 			if (rth->ip6r_len % 2) /* length must be even */
2915 				return (EINVAL);
2916 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2917 				return (EINVAL);
2918 			break;
2919 		default:
2920 			return (EINVAL);	/* not supported */
2921 		}
2922 
2923 		/* turn off the previous option */
2924 		ip6_clearpktopts(opt, IPV6_RTHDR);
2925 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2926 		if (opt->ip6po_rthdr == NULL)
2927 			return (ENOBUFS);
2928 		bcopy(rth, opt->ip6po_rthdr, rthlen);
2929 
2930 		break;
2931 	}
2932 
2933 	case IPV6_USE_MIN_MTU:
2934 		if (len != sizeof(int))
2935 			return (EINVAL);
2936 		minmtupolicy = *(int *)buf;
2937 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2938 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
2939 		    minmtupolicy != IP6PO_MINMTU_ALL) {
2940 			return (EINVAL);
2941 		}
2942 		opt->ip6po_minmtu = minmtupolicy;
2943 		break;
2944 
2945 	case IPV6_DONTFRAG:
2946 		if (len != sizeof(int))
2947 			return (EINVAL);
2948 
2949 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2950 			/*
2951 			 * we ignore this option for TCP sockets.
2952 			 * (RFC3542 leaves this case unspecified.)
2953 			 */
2954 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2955 		} else
2956 			opt->ip6po_flags |= IP6PO_DONTFRAG;
2957 		break;
2958 
2959 	case IPV6_PREFER_TEMPADDR:
2960 		if (len != sizeof(int))
2961 			return (EINVAL);
2962 		preftemp = *(int *)buf;
2963 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2964 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2965 		    preftemp != IP6PO_TEMPADDR_PREFER) {
2966 			return (EINVAL);
2967 		}
2968 		opt->ip6po_prefer_tempaddr = preftemp;
2969 		break;
2970 
2971 	default:
2972 		return (ENOPROTOOPT);
2973 	} /* end of switch */
2974 
2975 	return (0);
2976 }
2977 
2978 /*
2979  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2980  * packet to the input queue of a specified interface.  Note that this
2981  * calls the output routine of the loopback "driver", but with an interface
2982  * pointer that might NOT be &loif -- easier than replicating that code here.
2983  */
2984 void
2985 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
2986 {
2987 	struct mbuf *copym;
2988 	struct ip6_hdr *ip6;
2989 
2990 	copym = m_copy(m, 0, M_COPYALL);
2991 	if (copym == NULL)
2992 		return;
2993 
2994 	/*
2995 	 * Make sure to deep-copy IPv6 header portion in case the data
2996 	 * is in an mbuf cluster, so that we can safely override the IPv6
2997 	 * header portion later.
2998 	 */
2999 	if (!M_WRITABLE(copym) ||
3000 	    copym->m_len < sizeof(struct ip6_hdr)) {
3001 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3002 		if (copym == NULL)
3003 			return;
3004 	}
3005 	ip6 = mtod(copym, struct ip6_hdr *);
3006 	/*
3007 	 * clear embedded scope identifiers if necessary.
3008 	 * in6_clearscope will touch the addresses only when necessary.
3009 	 */
3010 	in6_clearscope(&ip6->ip6_src);
3011 	in6_clearscope(&ip6->ip6_dst);
3012 	if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3013 		copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3014 		    CSUM_PSEUDO_HDR;
3015 		copym->m_pkthdr.csum_data = 0xffff;
3016 	}
3017 	if_simloop(ifp, copym, AF_INET6, 0);
3018 }
3019 
3020 /*
3021  * Chop IPv6 header off from the payload.
3022  */
3023 static int
3024 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3025 {
3026 	struct mbuf *mh;
3027 	struct ip6_hdr *ip6;
3028 
3029 	ip6 = mtod(m, struct ip6_hdr *);
3030 	if (m->m_len > sizeof(*ip6)) {
3031 		mh = m_gethdr(M_NOWAIT, MT_DATA);
3032 		if (mh == NULL) {
3033 			m_freem(m);
3034 			return ENOBUFS;
3035 		}
3036 		m_move_pkthdr(mh, m);
3037 		M_ALIGN(mh, sizeof(*ip6));
3038 		m->m_len -= sizeof(*ip6);
3039 		m->m_data += sizeof(*ip6);
3040 		mh->m_next = m;
3041 		m = mh;
3042 		m->m_len = sizeof(*ip6);
3043 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3044 	}
3045 	exthdrs->ip6e_ip6 = m;
3046 	return 0;
3047 }
3048 
3049 /*
3050  * Compute IPv6 extension header length.
3051  */
3052 int
3053 ip6_optlen(struct inpcb *in6p)
3054 {
3055 	int len;
3056 
3057 	if (!in6p->in6p_outputopts)
3058 		return 0;
3059 
3060 	len = 0;
3061 #define elen(x) \
3062     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3063 
3064 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3065 	if (in6p->in6p_outputopts->ip6po_rthdr)
3066 		/* dest1 is valid with rthdr only */
3067 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
3068 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3069 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3070 	return len;
3071 #undef elen
3072 }
3073