xref: /freebsd/sys/netinet6/ip6_output.c (revision d6b92ffa990dc57d8a969a98e3ca4e25ba39cbb2)
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  * 3. 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_ratelimit.h"
69 #include "opt_ipsec.h"
70 #include "opt_sctp.h"
71 #include "opt_route.h"
72 #include "opt_rss.h"
73 
74 #include <sys/param.h>
75 #include <sys/kernel.h>
76 #include <sys/malloc.h>
77 #include <sys/mbuf.h>
78 #include <sys/errno.h>
79 #include <sys/priv.h>
80 #include <sys/proc.h>
81 #include <sys/protosw.h>
82 #include <sys/socket.h>
83 #include <sys/socketvar.h>
84 #include <sys/syslog.h>
85 #include <sys/ucred.h>
86 
87 #include <machine/in_cksum.h>
88 
89 #include <net/if.h>
90 #include <net/if_var.h>
91 #include <net/if_llatbl.h>
92 #include <net/netisr.h>
93 #include <net/route.h>
94 #include <net/pfil.h>
95 #include <net/rss_config.h>
96 #include <net/vnet.h>
97 
98 #include <netinet/in.h>
99 #include <netinet/in_var.h>
100 #include <netinet/ip_var.h>
101 #include <netinet6/in6_fib.h>
102 #include <netinet6/in6_var.h>
103 #include <netinet/ip6.h>
104 #include <netinet/icmp6.h>
105 #include <netinet6/ip6_var.h>
106 #include <netinet/in_pcb.h>
107 #include <netinet/tcp_var.h>
108 #include <netinet6/nd6.h>
109 #include <netinet6/in6_rss.h>
110 
111 #include <netipsec/ipsec_support.h>
112 #ifdef SCTP
113 #include <netinet/sctp.h>
114 #include <netinet/sctp_crc32.h>
115 #endif
116 
117 #include <netinet6/ip6protosw.h>
118 #include <netinet6/scope6_var.h>
119 
120 #ifdef FLOWTABLE
121 #include <net/flowtable.h>
122 #endif
123 
124 extern int in6_mcast_loop;
125 
126 struct ip6_exthdrs {
127 	struct mbuf *ip6e_ip6;
128 	struct mbuf *ip6e_hbh;
129 	struct mbuf *ip6e_dest1;
130 	struct mbuf *ip6e_rthdr;
131 	struct mbuf *ip6e_dest2;
132 };
133 
134 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
135 
136 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
137 			   struct ucred *, int);
138 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
139 	struct socket *, struct sockopt *);
140 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
141 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
142 	struct ucred *, int, int, int);
143 
144 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
145 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
146 	struct ip6_frag **);
147 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
148 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
149 static int ip6_getpmtu(struct route_in6 *, int,
150 	struct ifnet *, const struct in6_addr *, u_long *, int *, u_int,
151 	u_int);
152 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
153 	u_long *, int *, u_int);
154 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
155 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
156 
157 
158 /*
159  * Make an extension header from option data.  hp is the source, and
160  * mp is the destination.
161  */
162 #define MAKE_EXTHDR(hp, mp)						\
163     do {								\
164 	if (hp) {							\
165 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
166 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
167 		    ((eh)->ip6e_len + 1) << 3);				\
168 		if (error)						\
169 			goto freehdrs;					\
170 	}								\
171     } while (/*CONSTCOND*/ 0)
172 
173 /*
174  * Form a chain of extension headers.
175  * m is the extension header mbuf
176  * mp is the previous mbuf in the chain
177  * p is the next header
178  * i is the type of option.
179  */
180 #define MAKE_CHAIN(m, mp, p, i)\
181     do {\
182 	if (m) {\
183 		if (!hdrsplit) \
184 			panic("assumption failed: hdr not split"); \
185 		*mtod((m), u_char *) = *(p);\
186 		*(p) = (i);\
187 		p = mtod((m), u_char *);\
188 		(m)->m_next = (mp)->m_next;\
189 		(mp)->m_next = (m);\
190 		(mp) = (m);\
191 	}\
192     } while (/*CONSTCOND*/ 0)
193 
194 void
195 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
196 {
197 	u_short csum;
198 
199 	csum = in_cksum_skip(m, offset + plen, offset);
200 	if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
201 		csum = 0xffff;
202 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
203 
204 	if (offset + sizeof(u_short) > m->m_len) {
205 		printf("%s: delayed m_pullup, m->len: %d plen %u off %u "
206 		    "csum_flags=%b\n", __func__, m->m_len, plen, offset,
207 		    (int)m->m_pkthdr.csum_flags, CSUM_BITS);
208 		/*
209 		 * XXX this should not happen, but if it does, the correct
210 		 * behavior may be to insert the checksum in the appropriate
211 		 * next mbuf in the chain.
212 		 */
213 		return;
214 	}
215 	*(u_short *)(m->m_data + offset) = csum;
216 }
217 
218 int
219 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
220     int fraglen , uint32_t id)
221 {
222 	struct mbuf *m, **mnext, *m_frgpart;
223 	struct ip6_hdr *ip6, *mhip6;
224 	struct ip6_frag *ip6f;
225 	int off;
226 	int error;
227 	int tlen = m0->m_pkthdr.len;
228 
229 	KASSERT((fraglen % 8 == 0), ("Fragment length must be a multiple of 8"));
230 
231 	m = m0;
232 	ip6 = mtod(m, struct ip6_hdr *);
233 	mnext = &m->m_nextpkt;
234 
235 	for (off = hlen; off < tlen; off += fraglen) {
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 + fraglen >= tlen)
255 			fraglen = tlen - off;
256 		else
257 			ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
258 		mhip6->ip6_plen = htons((u_short)(fraglen + hlen +
259 		    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
260 		if ((m_frgpart = m_copym(m0, off, fraglen, M_NOWAIT)) == NULL) {
261 			IP6STAT_INC(ip6s_odropped);
262 			return (ENOBUFS);
263 		}
264 		m_cat(m, m_frgpart);
265 		m->m_pkthdr.len = fraglen + 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 		INP_LOCK_ASSERT(inp);
329 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
330 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
331 			/* unconditionally set flowid */
332 			m->m_pkthdr.flowid = inp->inp_flowid;
333 			M_HASHTYPE_SET(m, inp->inp_flowtype);
334 		}
335 	}
336 
337 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
338 	/*
339 	 * IPSec checking which handles several cases.
340 	 * FAST IPSEC: We re-injected the packet.
341 	 * XXX: need scope argument.
342 	 */
343 	if (IPSEC_ENABLED(ipv6)) {
344 		if ((error = IPSEC_OUTPUT(ipv6, m, inp)) != 0) {
345 			if (error == EINPROGRESS)
346 				error = 0;
347 			goto done;
348 		}
349 	}
350 #endif /* IPSEC */
351 
352 	bzero(&exthdrs, sizeof(exthdrs));
353 	if (opt) {
354 		/* Hop-by-Hop options header */
355 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
356 		/* Destination options header(1st part) */
357 		if (opt->ip6po_rthdr) {
358 			/*
359 			 * Destination options header(1st part)
360 			 * This only makes sense with a routing header.
361 			 * See Section 9.2 of RFC 3542.
362 			 * Disabling this part just for MIP6 convenience is
363 			 * a bad idea.  We need to think carefully about a
364 			 * way to make the advanced API coexist with MIP6
365 			 * options, which might automatically be inserted in
366 			 * the kernel.
367 			 */
368 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
369 		}
370 		/* Routing header */
371 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
372 		/* Destination options header(2nd part) */
373 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
374 	}
375 
376 	/*
377 	 * Calculate the total length of the extension header chain.
378 	 * Keep the length of the unfragmentable part for fragmentation.
379 	 */
380 	optlen = 0;
381 	if (exthdrs.ip6e_hbh)
382 		optlen += exthdrs.ip6e_hbh->m_len;
383 	if (exthdrs.ip6e_dest1)
384 		optlen += exthdrs.ip6e_dest1->m_len;
385 	if (exthdrs.ip6e_rthdr)
386 		optlen += exthdrs.ip6e_rthdr->m_len;
387 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
388 
389 	/* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */
390 	if (exthdrs.ip6e_dest2)
391 		optlen += exthdrs.ip6e_dest2->m_len;
392 
393 	/*
394 	 * If there is at least one extension header,
395 	 * separate IP6 header from the payload.
396 	 */
397 	if (optlen && !hdrsplit) {
398 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
399 			m = NULL;
400 			goto freehdrs;
401 		}
402 		m = exthdrs.ip6e_ip6;
403 		hdrsplit++;
404 	}
405 
406 	ip6 = mtod(m, struct ip6_hdr *);
407 
408 	/* adjust mbuf packet header length */
409 	m->m_pkthdr.len += optlen;
410 	plen = m->m_pkthdr.len - sizeof(*ip6);
411 
412 	/* If this is a jumbo payload, insert a jumbo payload option. */
413 	if (plen > IPV6_MAXPACKET) {
414 		if (!hdrsplit) {
415 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
416 				m = NULL;
417 				goto freehdrs;
418 			}
419 			m = exthdrs.ip6e_ip6;
420 			hdrsplit++;
421 		}
422 		/* adjust pointer */
423 		ip6 = mtod(m, struct ip6_hdr *);
424 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
425 			goto freehdrs;
426 		ip6->ip6_plen = 0;
427 	} else
428 		ip6->ip6_plen = htons(plen);
429 
430 	/*
431 	 * Concatenate headers and fill in next header fields.
432 	 * Here we have, on "m"
433 	 *	IPv6 payload
434 	 * and we insert headers accordingly.  Finally, we should be getting:
435 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
436 	 *
437 	 * during the header composing process, "m" points to IPv6 header.
438 	 * "mprev" points to an extension header prior to esp.
439 	 */
440 	u_char *nexthdrp = &ip6->ip6_nxt;
441 	mprev = m;
442 
443 	/*
444 	 * we treat dest2 specially.  this makes IPsec processing
445 	 * much easier.  the goal here is to make mprev point the
446 	 * mbuf prior to dest2.
447 	 *
448 	 * result: IPv6 dest2 payload
449 	 * m and mprev will point to IPv6 header.
450 	 */
451 	if (exthdrs.ip6e_dest2) {
452 		if (!hdrsplit)
453 			panic("assumption failed: hdr not split");
454 		exthdrs.ip6e_dest2->m_next = m->m_next;
455 		m->m_next = exthdrs.ip6e_dest2;
456 		*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
457 		ip6->ip6_nxt = IPPROTO_DSTOPTS;
458 	}
459 
460 	/*
461 	 * result: IPv6 hbh dest1 rthdr dest2 payload
462 	 * m will point to IPv6 header.  mprev will point to the
463 	 * extension header prior to dest2 (rthdr in the above case).
464 	 */
465 	MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
466 	MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
467 		   IPPROTO_DSTOPTS);
468 	MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
469 		   IPPROTO_ROUTING);
470 
471 	/*
472 	 * If there is a routing header, discard the packet.
473 	 */
474 	if (exthdrs.ip6e_rthdr) {
475 		 error = EINVAL;
476 		 goto bad;
477 	}
478 
479 	/* Source address validation */
480 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
481 	    (flags & IPV6_UNSPECSRC) == 0) {
482 		error = EOPNOTSUPP;
483 		IP6STAT_INC(ip6s_badscope);
484 		goto bad;
485 	}
486 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
487 		error = EOPNOTSUPP;
488 		IP6STAT_INC(ip6s_badscope);
489 		goto bad;
490 	}
491 
492 	IP6STAT_INC(ip6s_localout);
493 
494 	/*
495 	 * Route packet.
496 	 */
497 	if (ro == NULL) {
498 		ro = &ip6route;
499 		bzero((caddr_t)ro, sizeof(*ro));
500 	}
501 	ro_pmtu = ro;
502 	if (opt && opt->ip6po_rthdr)
503 		ro = &opt->ip6po_route;
504 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
505 #ifdef FLOWTABLE
506 	if (ro->ro_rt == NULL)
507 		(void )flowtable_lookup(AF_INET6, m, (struct route *)ro);
508 #endif
509 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
510 again:
511 	/*
512 	 * if specified, try to fill in the traffic class field.
513 	 * do not override if a non-zero value is already set.
514 	 * we check the diffserv field and the ecn field separately.
515 	 */
516 	if (opt && opt->ip6po_tclass >= 0) {
517 		int mask = 0;
518 
519 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
520 			mask |= 0xfc;
521 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
522 			mask |= 0x03;
523 		if (mask != 0)
524 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
525 	}
526 
527 	/* fill in or override the hop limit field, if necessary. */
528 	if (opt && opt->ip6po_hlim != -1)
529 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
530 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
531 		if (im6o != NULL)
532 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
533 		else
534 			ip6->ip6_hlim = V_ip6_defmcasthlim;
535 	}
536 	/*
537 	 * Validate route against routing table additions;
538 	 * a better/more specific route might have been added.
539 	 * Make sure address family is set in route.
540 	 */
541 	if (inp) {
542 		ro->ro_dst.sin6_family = AF_INET6;
543 		RT_VALIDATE((struct route *)ro, &inp->inp_rt_cookie, fibnum);
544 	}
545 	if (ro->ro_rt && fwd_tag == NULL && (ro->ro_rt->rt_flags & RTF_UP) &&
546 	    ro->ro_dst.sin6_family == AF_INET6 &&
547 	    IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
548 		rt = ro->ro_rt;
549 		ifp = ro->ro_rt->rt_ifp;
550 	} else {
551 		if (ro->ro_lle)
552 			LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
553 		ro->ro_lle = NULL;
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, *nexthdrp)) != 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 			if (ro->ro_lle)
824 				LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
825 			ro->ro_lle = NULL;
826 		}
827 	}
828 	/* See if fib was changed by packet filter. */
829 	if (fibnum != M_GETFIB(m)) {
830 		m->m_flags |= M_SKIP_FIREWALL;
831 		fibnum = M_GETFIB(m);
832 		RO_RTFREE(ro);
833 		needfiblookup = 1;
834 		if (ro->ro_lle)
835 			LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
836 		ro->ro_lle = NULL;
837 	}
838 	if (needfiblookup)
839 		goto again;
840 
841 	/* See if local, if yes, send it to netisr. */
842 	if (m->m_flags & M_FASTFWD_OURS) {
843 		if (m->m_pkthdr.rcvif == NULL)
844 			m->m_pkthdr.rcvif = V_loif;
845 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
846 			m->m_pkthdr.csum_flags |=
847 			    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
848 			m->m_pkthdr.csum_data = 0xffff;
849 		}
850 #ifdef SCTP
851 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
852 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
853 #endif
854 		error = netisr_queue(NETISR_IPV6, m);
855 		goto done;
856 	}
857 	/* Or forward to some other address? */
858 	if ((m->m_flags & M_IP6_NEXTHOP) &&
859 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
860 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
861 		bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
862 		m->m_flags |= M_SKIP_FIREWALL;
863 		m->m_flags &= ~M_IP6_NEXTHOP;
864 		m_tag_delete(m, fwd_tag);
865 		goto again;
866 	}
867 
868 passout:
869 	/*
870 	 * Send the packet to the outgoing interface.
871 	 * If necessary, do IPv6 fragmentation before sending.
872 	 *
873 	 * the logic here is rather complex:
874 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
875 	 * 1-a:	send as is if tlen <= path mtu
876 	 * 1-b:	fragment if tlen > path mtu
877 	 *
878 	 * 2: if user asks us not to fragment (dontfrag == 1)
879 	 * 2-a:	send as is if tlen <= interface mtu
880 	 * 2-b:	error if tlen > interface mtu
881 	 *
882 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
883 	 *	always fragment
884 	 *
885 	 * 4: if dontfrag == 1 && alwaysfrag == 1
886 	 *	error, as we cannot handle this conflicting request
887 	 */
888 	sw_csum = m->m_pkthdr.csum_flags;
889 	if (!hdrsplit) {
890 		tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
891 		sw_csum &= ~ifp->if_hwassist;
892 	} else
893 		tso = 0;
894 	/*
895 	 * If we added extension headers, we will not do TSO and calculate the
896 	 * checksums ourselves for now.
897 	 * XXX-BZ  Need a framework to know when the NIC can handle it, even
898 	 * with ext. hdrs.
899 	 */
900 	if (sw_csum & CSUM_DELAY_DATA_IPV6) {
901 		sw_csum &= ~CSUM_DELAY_DATA_IPV6;
902 		in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
903 	}
904 #ifdef SCTP
905 	if (sw_csum & CSUM_SCTP_IPV6) {
906 		sw_csum &= ~CSUM_SCTP_IPV6;
907 		sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
908 	}
909 #endif
910 	m->m_pkthdr.csum_flags &= ifp->if_hwassist;
911 	tlen = m->m_pkthdr.len;
912 
913 	if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
914 		dontfrag = 1;
915 	else
916 		dontfrag = 0;
917 	if (dontfrag && alwaysfrag) {	/* case 4 */
918 		/* conflicting request - can't transmit */
919 		error = EMSGSIZE;
920 		goto bad;
921 	}
922 	if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {	/* case 2-b */
923 		/*
924 		 * Even if the DONTFRAG option is specified, we cannot send the
925 		 * packet when the data length is larger than the MTU of the
926 		 * outgoing interface.
927 		 * Notify the error by sending IPV6_PATHMTU ancillary data if
928 		 * application wanted to know the MTU value. Also return an
929 		 * error code (this is not described in the API spec).
930 		 */
931 		if (inp != NULL)
932 			ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
933 		error = EMSGSIZE;
934 		goto bad;
935 	}
936 
937 	/*
938 	 * transmit packet without fragmentation
939 	 */
940 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
941 		struct in6_ifaddr *ia6;
942 
943 		ip6 = mtod(m, struct ip6_hdr *);
944 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
945 		if (ia6) {
946 			/* Record statistics for this interface address. */
947 			counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
948 			counter_u64_add(ia6->ia_ifa.ifa_obytes,
949 			    m->m_pkthdr.len);
950 			ifa_free(&ia6->ia_ifa);
951 		}
952 #ifdef RATELIMIT
953 		if (inp != NULL) {
954 			if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
955 				in_pcboutput_txrtlmt(inp, ifp, m);
956 			/* stamp send tag on mbuf */
957 			m->m_pkthdr.snd_tag = inp->inp_snd_tag;
958 		} else {
959 			m->m_pkthdr.snd_tag = NULL;
960 		}
961 #endif
962 		error = nd6_output_ifp(ifp, origifp, m, dst,
963 		    (struct route *)ro);
964 #ifdef RATELIMIT
965 		/* check for route change */
966 		if (error == EAGAIN)
967 			in_pcboutput_eagain(inp);
968 #endif
969 		goto done;
970 	}
971 
972 	/*
973 	 * try to fragment the packet.  case 1-b and 3
974 	 */
975 	if (mtu < IPV6_MMTU) {
976 		/* path MTU cannot be less than IPV6_MMTU */
977 		error = EMSGSIZE;
978 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
979 		goto bad;
980 	} else if (ip6->ip6_plen == 0) {
981 		/* jumbo payload cannot be fragmented */
982 		error = EMSGSIZE;
983 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
984 		goto bad;
985 	} else {
986 		u_char nextproto;
987 
988 		/*
989 		 * Too large for the destination or interface;
990 		 * fragment if possible.
991 		 * Must be able to put at least 8 bytes per fragment.
992 		 */
993 		hlen = unfragpartlen;
994 		if (mtu > IPV6_MAXPACKET)
995 			mtu = IPV6_MAXPACKET;
996 
997 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
998 		if (len < 8) {
999 			error = EMSGSIZE;
1000 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
1001 			goto bad;
1002 		}
1003 
1004 		/*
1005 		 * If the interface will not calculate checksums on
1006 		 * fragmented packets, then do it here.
1007 		 * XXX-BZ handle the hw offloading case.  Need flags.
1008 		 */
1009 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1010 			in6_delayed_cksum(m, plen, hlen);
1011 			m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
1012 		}
1013 #ifdef SCTP
1014 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1015 			sctp_delayed_cksum(m, hlen);
1016 			m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1017 		}
1018 #endif
1019 		/*
1020 		 * Change the next header field of the last header in the
1021 		 * unfragmentable part.
1022 		 */
1023 		if (exthdrs.ip6e_rthdr) {
1024 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1025 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1026 		} else if (exthdrs.ip6e_dest1) {
1027 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1028 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1029 		} else if (exthdrs.ip6e_hbh) {
1030 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1031 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1032 		} else {
1033 			nextproto = ip6->ip6_nxt;
1034 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1035 		}
1036 
1037 		/*
1038 		 * Loop through length of segment after first fragment,
1039 		 * make new header and copy data of each part and link onto
1040 		 * chain.
1041 		 */
1042 		m0 = m;
1043 		id = htonl(ip6_randomid());
1044 		if ((error = ip6_fragment(ifp, m, hlen, nextproto, len, id)))
1045 			goto sendorfree;
1046 
1047 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1048 	}
1049 
1050 	/*
1051 	 * Remove leading garbages.
1052 	 */
1053 sendorfree:
1054 	m = m0->m_nextpkt;
1055 	m0->m_nextpkt = 0;
1056 	m_freem(m0);
1057 	for (m0 = m; m; m = m0) {
1058 		m0 = m->m_nextpkt;
1059 		m->m_nextpkt = 0;
1060 		if (error == 0) {
1061 			/* Record statistics for this interface address. */
1062 			if (ia) {
1063 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1064 				counter_u64_add(ia->ia_ifa.ifa_obytes,
1065 				    m->m_pkthdr.len);
1066 			}
1067 #ifdef RATELIMIT
1068 			if (inp != NULL) {
1069 				if (inp->inp_flags2 & INP_RATE_LIMIT_CHANGED)
1070 					in_pcboutput_txrtlmt(inp, ifp, m);
1071 				/* stamp send tag on mbuf */
1072 				m->m_pkthdr.snd_tag = inp->inp_snd_tag;
1073 			} else {
1074 				m->m_pkthdr.snd_tag = NULL;
1075 			}
1076 #endif
1077 			error = nd6_output_ifp(ifp, origifp, m, dst,
1078 			    (struct route *)ro);
1079 #ifdef RATELIMIT
1080 			/* check for route change */
1081 			if (error == EAGAIN)
1082 				in_pcboutput_eagain(inp);
1083 #endif
1084 		} else
1085 			m_freem(m);
1086 	}
1087 
1088 	if (error == 0)
1089 		IP6STAT_INC(ip6s_fragmented);
1090 
1091 done:
1092 	if (ro == &ip6route)
1093 		RO_RTFREE(ro);
1094 	return (error);
1095 
1096 freehdrs:
1097 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1098 	m_freem(exthdrs.ip6e_dest1);
1099 	m_freem(exthdrs.ip6e_rthdr);
1100 	m_freem(exthdrs.ip6e_dest2);
1101 	/* FALLTHROUGH */
1102 bad:
1103 	if (m)
1104 		m_freem(m);
1105 	goto done;
1106 }
1107 
1108 static int
1109 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1110 {
1111 	struct mbuf *m;
1112 
1113 	if (hlen > MCLBYTES)
1114 		return (ENOBUFS); /* XXX */
1115 
1116 	if (hlen > MLEN)
1117 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1118 	else
1119 		m = m_get(M_NOWAIT, MT_DATA);
1120 	if (m == NULL)
1121 		return (ENOBUFS);
1122 	m->m_len = hlen;
1123 	if (hdr)
1124 		bcopy(hdr, mtod(m, caddr_t), hlen);
1125 
1126 	*mp = m;
1127 	return (0);
1128 }
1129 
1130 /*
1131  * Insert jumbo payload option.
1132  */
1133 static int
1134 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1135 {
1136 	struct mbuf *mopt;
1137 	u_char *optbuf;
1138 	u_int32_t v;
1139 
1140 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1141 
1142 	/*
1143 	 * If there is no hop-by-hop options header, allocate new one.
1144 	 * If there is one but it doesn't have enough space to store the
1145 	 * jumbo payload option, allocate a cluster to store the whole options.
1146 	 * Otherwise, use it to store the options.
1147 	 */
1148 	if (exthdrs->ip6e_hbh == NULL) {
1149 		mopt = m_get(M_NOWAIT, MT_DATA);
1150 		if (mopt == NULL)
1151 			return (ENOBUFS);
1152 		mopt->m_len = JUMBOOPTLEN;
1153 		optbuf = mtod(mopt, u_char *);
1154 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1155 		exthdrs->ip6e_hbh = mopt;
1156 	} else {
1157 		struct ip6_hbh *hbh;
1158 
1159 		mopt = exthdrs->ip6e_hbh;
1160 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1161 			/*
1162 			 * XXX assumption:
1163 			 * - exthdrs->ip6e_hbh is not referenced from places
1164 			 *   other than exthdrs.
1165 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1166 			 */
1167 			int oldoptlen = mopt->m_len;
1168 			struct mbuf *n;
1169 
1170 			/*
1171 			 * XXX: give up if the whole (new) hbh header does
1172 			 * not fit even in an mbuf cluster.
1173 			 */
1174 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1175 				return (ENOBUFS);
1176 
1177 			/*
1178 			 * As a consequence, we must always prepare a cluster
1179 			 * at this point.
1180 			 */
1181 			n = m_getcl(M_NOWAIT, MT_DATA, 0);
1182 			if (n == NULL)
1183 				return (ENOBUFS);
1184 			n->m_len = oldoptlen + JUMBOOPTLEN;
1185 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1186 			    oldoptlen);
1187 			optbuf = mtod(n, caddr_t) + oldoptlen;
1188 			m_freem(mopt);
1189 			mopt = exthdrs->ip6e_hbh = n;
1190 		} else {
1191 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1192 			mopt->m_len += JUMBOOPTLEN;
1193 		}
1194 		optbuf[0] = IP6OPT_PADN;
1195 		optbuf[1] = 1;
1196 
1197 		/*
1198 		 * Adjust the header length according to the pad and
1199 		 * the jumbo payload option.
1200 		 */
1201 		hbh = mtod(mopt, struct ip6_hbh *);
1202 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1203 	}
1204 
1205 	/* fill in the option. */
1206 	optbuf[2] = IP6OPT_JUMBO;
1207 	optbuf[3] = 4;
1208 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1209 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1210 
1211 	/* finally, adjust the packet header length */
1212 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1213 
1214 	return (0);
1215 #undef JUMBOOPTLEN
1216 }
1217 
1218 /*
1219  * Insert fragment header and copy unfragmentable header portions.
1220  */
1221 static int
1222 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1223     struct ip6_frag **frghdrp)
1224 {
1225 	struct mbuf *n, *mlast;
1226 
1227 	if (hlen > sizeof(struct ip6_hdr)) {
1228 		n = m_copym(m0, sizeof(struct ip6_hdr),
1229 		    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1230 		if (n == NULL)
1231 			return (ENOBUFS);
1232 		m->m_next = n;
1233 	} else
1234 		n = m;
1235 
1236 	/* Search for the last mbuf of unfragmentable part. */
1237 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1238 		;
1239 
1240 	if (M_WRITABLE(mlast) &&
1241 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1242 		/* use the trailing space of the last mbuf for the fragment hdr */
1243 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1244 		    mlast->m_len);
1245 		mlast->m_len += sizeof(struct ip6_frag);
1246 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1247 	} else {
1248 		/* allocate a new mbuf for the fragment header */
1249 		struct mbuf *mfrg;
1250 
1251 		mfrg = m_get(M_NOWAIT, MT_DATA);
1252 		if (mfrg == NULL)
1253 			return (ENOBUFS);
1254 		mfrg->m_len = sizeof(struct ip6_frag);
1255 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1256 		mlast->m_next = mfrg;
1257 	}
1258 
1259 	return (0);
1260 }
1261 
1262 /*
1263  * Calculates IPv6 path mtu for destination @dst.
1264  * Resulting MTU is stored in @mtup.
1265  *
1266  * Returns 0 on success.
1267  */
1268 static int
1269 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1270 {
1271 	struct nhop6_extended nh6;
1272 	struct in6_addr kdst;
1273 	uint32_t scopeid;
1274 	struct ifnet *ifp;
1275 	u_long mtu;
1276 	int error;
1277 
1278 	in6_splitscope(dst, &kdst, &scopeid);
1279 	if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0)
1280 		return (EHOSTUNREACH);
1281 
1282 	ifp = nh6.nh_ifp;
1283 	mtu = nh6.nh_mtu;
1284 
1285 	error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL, 0);
1286 	fib6_free_nh_ext(fibnum, &nh6);
1287 
1288 	return (error);
1289 }
1290 
1291 /*
1292  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1293  * and cached data in @ro_pmtu.
1294  * MTU from (successful) route lookup is saved (along with dst)
1295  * inside @ro_pmtu to avoid subsequent route lookups after packet
1296  * filter processing.
1297  *
1298  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1299  * Returns 0 on success.
1300  */
1301 static int
1302 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1303     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1304     int *alwaysfragp, u_int fibnum, u_int proto)
1305 {
1306 	struct nhop6_basic nh6;
1307 	struct in6_addr kdst;
1308 	uint32_t scopeid;
1309 	struct sockaddr_in6 *sa6_dst;
1310 	u_long mtu;
1311 
1312 	mtu = 0;
1313 	if (do_lookup) {
1314 
1315 		/*
1316 		 * Here ro_pmtu has final destination address, while
1317 		 * ro might represent immediate destination.
1318 		 * Use ro_pmtu destination since mtu might differ.
1319 		 */
1320 		sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1321 		if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1322 			ro_pmtu->ro_mtu = 0;
1323 
1324 		if (ro_pmtu->ro_mtu == 0) {
1325 			bzero(sa6_dst, sizeof(*sa6_dst));
1326 			sa6_dst->sin6_family = AF_INET6;
1327 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1328 			sa6_dst->sin6_addr = *dst;
1329 
1330 			in6_splitscope(dst, &kdst, &scopeid);
1331 			if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0,
1332 			    &nh6) == 0)
1333 				ro_pmtu->ro_mtu = nh6.nh_mtu;
1334 		}
1335 
1336 		mtu = ro_pmtu->ro_mtu;
1337 	}
1338 
1339 	if (ro_pmtu->ro_rt)
1340 		mtu = ro_pmtu->ro_rt->rt_mtu;
1341 
1342 	return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1343 }
1344 
1345 /*
1346  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1347  * hostcache data for @dst.
1348  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1349  *
1350  * Returns 0 on success.
1351  */
1352 static int
1353 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1354     u_long *mtup, int *alwaysfragp, u_int proto)
1355 {
1356 	u_long mtu = 0;
1357 	int alwaysfrag = 0;
1358 	int error = 0;
1359 
1360 	if (rt_mtu > 0) {
1361 		u_int32_t ifmtu;
1362 		struct in_conninfo inc;
1363 
1364 		bzero(&inc, sizeof(inc));
1365 		inc.inc_flags |= INC_ISIPV6;
1366 		inc.inc6_faddr = *dst;
1367 
1368 		ifmtu = IN6_LINKMTU(ifp);
1369 
1370 		/* TCP is known to react to pmtu changes so skip hc */
1371 		if (proto != IPPROTO_TCP)
1372 			mtu = tcp_hc_getmtu(&inc);
1373 
1374 		if (mtu)
1375 			mtu = min(mtu, rt_mtu);
1376 		else
1377 			mtu = rt_mtu;
1378 		if (mtu == 0)
1379 			mtu = ifmtu;
1380 		else if (mtu < IPV6_MMTU) {
1381 			/*
1382 			 * RFC2460 section 5, last paragraph:
1383 			 * if we record ICMPv6 too big message with
1384 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1385 			 * or smaller, with framgent header attached.
1386 			 * (fragment header is needed regardless from the
1387 			 * packet size, for translators to identify packets)
1388 			 */
1389 			alwaysfrag = 1;
1390 			mtu = IPV6_MMTU;
1391 		}
1392 	} else if (ifp) {
1393 		mtu = IN6_LINKMTU(ifp);
1394 	} else
1395 		error = EHOSTUNREACH; /* XXX */
1396 
1397 	*mtup = mtu;
1398 	if (alwaysfragp)
1399 		*alwaysfragp = alwaysfrag;
1400 	return (error);
1401 }
1402 
1403 /*
1404  * IP6 socket option processing.
1405  */
1406 int
1407 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1408 {
1409 	int optdatalen, uproto;
1410 	void *optdata;
1411 	struct inpcb *in6p = sotoinpcb(so);
1412 	int error, optval;
1413 	int level, op, optname;
1414 	int optlen;
1415 	struct thread *td;
1416 #ifdef	RSS
1417 	uint32_t rss_bucket;
1418 	int retval;
1419 #endif
1420 
1421 /*
1422  * Don't use more than a quarter of mbuf clusters.  N.B.:
1423  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1424  * on LP64 architectures, so cast to u_long to avoid undefined
1425  * behavior.  ILP32 architectures cannot have nmbclusters
1426  * large enough to overflow for other reasons.
1427  */
1428 #define IPV6_PKTOPTIONS_MBUF_LIMIT	((u_long)nmbclusters * MCLBYTES / 4)
1429 
1430 	level = sopt->sopt_level;
1431 	op = sopt->sopt_dir;
1432 	optname = sopt->sopt_name;
1433 	optlen = sopt->sopt_valsize;
1434 	td = sopt->sopt_td;
1435 	error = 0;
1436 	optval = 0;
1437 	uproto = (int)so->so_proto->pr_protocol;
1438 
1439 	if (level != IPPROTO_IPV6) {
1440 		error = EINVAL;
1441 
1442 		if (sopt->sopt_level == SOL_SOCKET &&
1443 		    sopt->sopt_dir == SOPT_SET) {
1444 			switch (sopt->sopt_name) {
1445 			case SO_REUSEADDR:
1446 				INP_WLOCK(in6p);
1447 				if ((so->so_options & SO_REUSEADDR) != 0)
1448 					in6p->inp_flags2 |= INP_REUSEADDR;
1449 				else
1450 					in6p->inp_flags2 &= ~INP_REUSEADDR;
1451 				INP_WUNLOCK(in6p);
1452 				error = 0;
1453 				break;
1454 			case SO_REUSEPORT:
1455 				INP_WLOCK(in6p);
1456 				if ((so->so_options & SO_REUSEPORT) != 0)
1457 					in6p->inp_flags2 |= INP_REUSEPORT;
1458 				else
1459 					in6p->inp_flags2 &= ~INP_REUSEPORT;
1460 				INP_WUNLOCK(in6p);
1461 				error = 0;
1462 				break;
1463 			case SO_SETFIB:
1464 				INP_WLOCK(in6p);
1465 				in6p->inp_inc.inc_fibnum = so->so_fibnum;
1466 				INP_WUNLOCK(in6p);
1467 				error = 0;
1468 				break;
1469 			case SO_MAX_PACING_RATE:
1470 #ifdef RATELIMIT
1471 				INP_WLOCK(in6p);
1472 				in6p->inp_flags2 |= INP_RATE_LIMIT_CHANGED;
1473 				INP_WUNLOCK(in6p);
1474 				error = 0;
1475 #else
1476 				error = EOPNOTSUPP;
1477 #endif
1478 				break;
1479 			default:
1480 				break;
1481 			}
1482 		}
1483 	} else {		/* level == IPPROTO_IPV6 */
1484 		switch (op) {
1485 
1486 		case SOPT_SET:
1487 			switch (optname) {
1488 			case IPV6_2292PKTOPTIONS:
1489 #ifdef IPV6_PKTOPTIONS
1490 			case IPV6_PKTOPTIONS:
1491 #endif
1492 			{
1493 				struct mbuf *m;
1494 
1495 				if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1496 					printf("ip6_ctloutput: mbuf limit hit\n");
1497 					error = ENOBUFS;
1498 					break;
1499 				}
1500 
1501 				error = soopt_getm(sopt, &m); /* XXX */
1502 				if (error != 0)
1503 					break;
1504 				error = soopt_mcopyin(sopt, m); /* XXX */
1505 				if (error != 0)
1506 					break;
1507 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1508 						    m, so, sopt);
1509 				m_freem(m); /* XXX */
1510 				break;
1511 			}
1512 
1513 			/*
1514 			 * Use of some Hop-by-Hop options or some
1515 			 * Destination options, might require special
1516 			 * privilege.  That is, normal applications
1517 			 * (without special privilege) might be forbidden
1518 			 * from setting certain options in outgoing packets,
1519 			 * and might never see certain options in received
1520 			 * packets. [RFC 2292 Section 6]
1521 			 * KAME specific note:
1522 			 *  KAME prevents non-privileged users from sending or
1523 			 *  receiving ANY hbh/dst options in order to avoid
1524 			 *  overhead of parsing options in the kernel.
1525 			 */
1526 			case IPV6_RECVHOPOPTS:
1527 			case IPV6_RECVDSTOPTS:
1528 			case IPV6_RECVRTHDRDSTOPTS:
1529 				if (td != NULL) {
1530 					error = priv_check(td,
1531 					    PRIV_NETINET_SETHDROPTS);
1532 					if (error)
1533 						break;
1534 				}
1535 				/* FALLTHROUGH */
1536 			case IPV6_UNICAST_HOPS:
1537 			case IPV6_HOPLIMIT:
1538 
1539 			case IPV6_RECVPKTINFO:
1540 			case IPV6_RECVHOPLIMIT:
1541 			case IPV6_RECVRTHDR:
1542 			case IPV6_RECVPATHMTU:
1543 			case IPV6_RECVTCLASS:
1544 			case IPV6_RECVFLOWID:
1545 #ifdef	RSS
1546 			case IPV6_RECVRSSBUCKETID:
1547 #endif
1548 			case IPV6_V6ONLY:
1549 			case IPV6_AUTOFLOWLABEL:
1550 			case IPV6_ORIGDSTADDR:
1551 			case IPV6_BINDANY:
1552 			case IPV6_BINDMULTI:
1553 #ifdef	RSS
1554 			case IPV6_RSS_LISTEN_BUCKET:
1555 #endif
1556 				if (optname == IPV6_BINDANY && td != NULL) {
1557 					error = priv_check(td,
1558 					    PRIV_NETINET_BINDANY);
1559 					if (error)
1560 						break;
1561 				}
1562 
1563 				if (optlen != sizeof(int)) {
1564 					error = EINVAL;
1565 					break;
1566 				}
1567 				error = sooptcopyin(sopt, &optval,
1568 					sizeof optval, sizeof optval);
1569 				if (error)
1570 					break;
1571 				switch (optname) {
1572 
1573 				case IPV6_UNICAST_HOPS:
1574 					if (optval < -1 || optval >= 256)
1575 						error = EINVAL;
1576 					else {
1577 						/* -1 = kernel default */
1578 						in6p->in6p_hops = optval;
1579 						if ((in6p->inp_vflag &
1580 						     INP_IPV4) != 0)
1581 							in6p->inp_ip_ttl = optval;
1582 					}
1583 					break;
1584 #define OPTSET(bit) \
1585 do { \
1586 	INP_WLOCK(in6p); \
1587 	if (optval) \
1588 		in6p->inp_flags |= (bit); \
1589 	else \
1590 		in6p->inp_flags &= ~(bit); \
1591 	INP_WUNLOCK(in6p); \
1592 } while (/*CONSTCOND*/ 0)
1593 #define OPTSET2292(bit) \
1594 do { \
1595 	INP_WLOCK(in6p); \
1596 	in6p->inp_flags |= IN6P_RFC2292; \
1597 	if (optval) \
1598 		in6p->inp_flags |= (bit); \
1599 	else \
1600 		in6p->inp_flags &= ~(bit); \
1601 	INP_WUNLOCK(in6p); \
1602 } while (/*CONSTCOND*/ 0)
1603 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1604 
1605 #define OPTSET2(bit, val) do {						\
1606 	INP_WLOCK(in6p);						\
1607 	if (val)							\
1608 		in6p->inp_flags2 |= bit;				\
1609 	else								\
1610 		in6p->inp_flags2 &= ~bit;				\
1611 	INP_WUNLOCK(in6p);						\
1612 } while (0)
1613 #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0)
1614 
1615 				case IPV6_RECVPKTINFO:
1616 					/* cannot mix with RFC2292 */
1617 					if (OPTBIT(IN6P_RFC2292)) {
1618 						error = EINVAL;
1619 						break;
1620 					}
1621 					OPTSET(IN6P_PKTINFO);
1622 					break;
1623 
1624 				case IPV6_HOPLIMIT:
1625 				{
1626 					struct ip6_pktopts **optp;
1627 
1628 					/* cannot mix with RFC2292 */
1629 					if (OPTBIT(IN6P_RFC2292)) {
1630 						error = EINVAL;
1631 						break;
1632 					}
1633 					optp = &in6p->in6p_outputopts;
1634 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1635 					    (u_char *)&optval, sizeof(optval),
1636 					    optp, (td != NULL) ? td->td_ucred :
1637 					    NULL, uproto);
1638 					break;
1639 				}
1640 
1641 				case IPV6_RECVHOPLIMIT:
1642 					/* cannot mix with RFC2292 */
1643 					if (OPTBIT(IN6P_RFC2292)) {
1644 						error = EINVAL;
1645 						break;
1646 					}
1647 					OPTSET(IN6P_HOPLIMIT);
1648 					break;
1649 
1650 				case IPV6_RECVHOPOPTS:
1651 					/* cannot mix with RFC2292 */
1652 					if (OPTBIT(IN6P_RFC2292)) {
1653 						error = EINVAL;
1654 						break;
1655 					}
1656 					OPTSET(IN6P_HOPOPTS);
1657 					break;
1658 
1659 				case IPV6_RECVDSTOPTS:
1660 					/* cannot mix with RFC2292 */
1661 					if (OPTBIT(IN6P_RFC2292)) {
1662 						error = EINVAL;
1663 						break;
1664 					}
1665 					OPTSET(IN6P_DSTOPTS);
1666 					break;
1667 
1668 				case IPV6_RECVRTHDRDSTOPTS:
1669 					/* cannot mix with RFC2292 */
1670 					if (OPTBIT(IN6P_RFC2292)) {
1671 						error = EINVAL;
1672 						break;
1673 					}
1674 					OPTSET(IN6P_RTHDRDSTOPTS);
1675 					break;
1676 
1677 				case IPV6_RECVRTHDR:
1678 					/* cannot mix with RFC2292 */
1679 					if (OPTBIT(IN6P_RFC2292)) {
1680 						error = EINVAL;
1681 						break;
1682 					}
1683 					OPTSET(IN6P_RTHDR);
1684 					break;
1685 
1686 				case IPV6_RECVPATHMTU:
1687 					/*
1688 					 * We ignore this option for TCP
1689 					 * sockets.
1690 					 * (RFC3542 leaves this case
1691 					 * unspecified.)
1692 					 */
1693 					if (uproto != IPPROTO_TCP)
1694 						OPTSET(IN6P_MTU);
1695 					break;
1696 
1697 				case IPV6_RECVFLOWID:
1698 					OPTSET2(INP_RECVFLOWID, optval);
1699 					break;
1700 
1701 #ifdef	RSS
1702 				case IPV6_RECVRSSBUCKETID:
1703 					OPTSET2(INP_RECVRSSBUCKETID, optval);
1704 					break;
1705 #endif
1706 
1707 				case IPV6_V6ONLY:
1708 					/*
1709 					 * make setsockopt(IPV6_V6ONLY)
1710 					 * available only prior to bind(2).
1711 					 * see ipng mailing list, Jun 22 2001.
1712 					 */
1713 					if (in6p->inp_lport ||
1714 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1715 						error = EINVAL;
1716 						break;
1717 					}
1718 					OPTSET(IN6P_IPV6_V6ONLY);
1719 					if (optval)
1720 						in6p->inp_vflag &= ~INP_IPV4;
1721 					else
1722 						in6p->inp_vflag |= INP_IPV4;
1723 					break;
1724 				case IPV6_RECVTCLASS:
1725 					/* cannot mix with RFC2292 XXX */
1726 					if (OPTBIT(IN6P_RFC2292)) {
1727 						error = EINVAL;
1728 						break;
1729 					}
1730 					OPTSET(IN6P_TCLASS);
1731 					break;
1732 				case IPV6_AUTOFLOWLABEL:
1733 					OPTSET(IN6P_AUTOFLOWLABEL);
1734 					break;
1735 
1736 				case IPV6_ORIGDSTADDR:
1737 					OPTSET2(INP_ORIGDSTADDR, optval);
1738 					break;
1739 				case IPV6_BINDANY:
1740 					OPTSET(INP_BINDANY);
1741 					break;
1742 
1743 				case IPV6_BINDMULTI:
1744 					OPTSET2(INP_BINDMULTI, optval);
1745 					break;
1746 #ifdef	RSS
1747 				case IPV6_RSS_LISTEN_BUCKET:
1748 					if ((optval >= 0) &&
1749 					    (optval < rss_getnumbuckets())) {
1750 						in6p->inp_rss_listen_bucket = optval;
1751 						OPTSET2(INP_RSS_BUCKET_SET, 1);
1752 					} else {
1753 						error = EINVAL;
1754 					}
1755 					break;
1756 #endif
1757 				}
1758 				break;
1759 
1760 			case IPV6_TCLASS:
1761 			case IPV6_DONTFRAG:
1762 			case IPV6_USE_MIN_MTU:
1763 			case IPV6_PREFER_TEMPADDR:
1764 				if (optlen != sizeof(optval)) {
1765 					error = EINVAL;
1766 					break;
1767 				}
1768 				error = sooptcopyin(sopt, &optval,
1769 					sizeof optval, sizeof optval);
1770 				if (error)
1771 					break;
1772 				{
1773 					struct ip6_pktopts **optp;
1774 					optp = &in6p->in6p_outputopts;
1775 					error = ip6_pcbopt(optname,
1776 					    (u_char *)&optval, sizeof(optval),
1777 					    optp, (td != NULL) ? td->td_ucred :
1778 					    NULL, uproto);
1779 					break;
1780 				}
1781 
1782 			case IPV6_2292PKTINFO:
1783 			case IPV6_2292HOPLIMIT:
1784 			case IPV6_2292HOPOPTS:
1785 			case IPV6_2292DSTOPTS:
1786 			case IPV6_2292RTHDR:
1787 				/* RFC 2292 */
1788 				if (optlen != sizeof(int)) {
1789 					error = EINVAL;
1790 					break;
1791 				}
1792 				error = sooptcopyin(sopt, &optval,
1793 					sizeof optval, sizeof optval);
1794 				if (error)
1795 					break;
1796 				switch (optname) {
1797 				case IPV6_2292PKTINFO:
1798 					OPTSET2292(IN6P_PKTINFO);
1799 					break;
1800 				case IPV6_2292HOPLIMIT:
1801 					OPTSET2292(IN6P_HOPLIMIT);
1802 					break;
1803 				case IPV6_2292HOPOPTS:
1804 					/*
1805 					 * Check super-user privilege.
1806 					 * See comments for IPV6_RECVHOPOPTS.
1807 					 */
1808 					if (td != NULL) {
1809 						error = priv_check(td,
1810 						    PRIV_NETINET_SETHDROPTS);
1811 						if (error)
1812 							return (error);
1813 					}
1814 					OPTSET2292(IN6P_HOPOPTS);
1815 					break;
1816 				case IPV6_2292DSTOPTS:
1817 					if (td != NULL) {
1818 						error = priv_check(td,
1819 						    PRIV_NETINET_SETHDROPTS);
1820 						if (error)
1821 							return (error);
1822 					}
1823 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1824 					break;
1825 				case IPV6_2292RTHDR:
1826 					OPTSET2292(IN6P_RTHDR);
1827 					break;
1828 				}
1829 				break;
1830 			case IPV6_PKTINFO:
1831 			case IPV6_HOPOPTS:
1832 			case IPV6_RTHDR:
1833 			case IPV6_DSTOPTS:
1834 			case IPV6_RTHDRDSTOPTS:
1835 			case IPV6_NEXTHOP:
1836 			{
1837 				/* new advanced API (RFC3542) */
1838 				u_char *optbuf;
1839 				u_char optbuf_storage[MCLBYTES];
1840 				int optlen;
1841 				struct ip6_pktopts **optp;
1842 
1843 				/* cannot mix with RFC2292 */
1844 				if (OPTBIT(IN6P_RFC2292)) {
1845 					error = EINVAL;
1846 					break;
1847 				}
1848 
1849 				/*
1850 				 * We only ensure valsize is not too large
1851 				 * here.  Further validation will be done
1852 				 * later.
1853 				 */
1854 				error = sooptcopyin(sopt, optbuf_storage,
1855 				    sizeof(optbuf_storage), 0);
1856 				if (error)
1857 					break;
1858 				optlen = sopt->sopt_valsize;
1859 				optbuf = optbuf_storage;
1860 				optp = &in6p->in6p_outputopts;
1861 				error = ip6_pcbopt(optname, optbuf, optlen,
1862 				    optp, (td != NULL) ? td->td_ucred : NULL,
1863 				    uproto);
1864 				break;
1865 			}
1866 #undef OPTSET
1867 
1868 			case IPV6_MULTICAST_IF:
1869 			case IPV6_MULTICAST_HOPS:
1870 			case IPV6_MULTICAST_LOOP:
1871 			case IPV6_JOIN_GROUP:
1872 			case IPV6_LEAVE_GROUP:
1873 			case IPV6_MSFILTER:
1874 			case MCAST_BLOCK_SOURCE:
1875 			case MCAST_UNBLOCK_SOURCE:
1876 			case MCAST_JOIN_GROUP:
1877 			case MCAST_LEAVE_GROUP:
1878 			case MCAST_JOIN_SOURCE_GROUP:
1879 			case MCAST_LEAVE_SOURCE_GROUP:
1880 				error = ip6_setmoptions(in6p, sopt);
1881 				break;
1882 
1883 			case IPV6_PORTRANGE:
1884 				error = sooptcopyin(sopt, &optval,
1885 				    sizeof optval, sizeof optval);
1886 				if (error)
1887 					break;
1888 
1889 				INP_WLOCK(in6p);
1890 				switch (optval) {
1891 				case IPV6_PORTRANGE_DEFAULT:
1892 					in6p->inp_flags &= ~(INP_LOWPORT);
1893 					in6p->inp_flags &= ~(INP_HIGHPORT);
1894 					break;
1895 
1896 				case IPV6_PORTRANGE_HIGH:
1897 					in6p->inp_flags &= ~(INP_LOWPORT);
1898 					in6p->inp_flags |= INP_HIGHPORT;
1899 					break;
1900 
1901 				case IPV6_PORTRANGE_LOW:
1902 					in6p->inp_flags &= ~(INP_HIGHPORT);
1903 					in6p->inp_flags |= INP_LOWPORT;
1904 					break;
1905 
1906 				default:
1907 					error = EINVAL;
1908 					break;
1909 				}
1910 				INP_WUNLOCK(in6p);
1911 				break;
1912 
1913 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
1914 			case IPV6_IPSEC_POLICY:
1915 				if (IPSEC_ENABLED(ipv6)) {
1916 					error = IPSEC_PCBCTL(ipv6, in6p, sopt);
1917 					break;
1918 				}
1919 				/* FALLTHROUGH */
1920 #endif /* IPSEC */
1921 
1922 			default:
1923 				error = ENOPROTOOPT;
1924 				break;
1925 			}
1926 			break;
1927 
1928 		case SOPT_GET:
1929 			switch (optname) {
1930 
1931 			case IPV6_2292PKTOPTIONS:
1932 #ifdef IPV6_PKTOPTIONS
1933 			case IPV6_PKTOPTIONS:
1934 #endif
1935 				/*
1936 				 * RFC3542 (effectively) deprecated the
1937 				 * semantics of the 2292-style pktoptions.
1938 				 * Since it was not reliable in nature (i.e.,
1939 				 * applications had to expect the lack of some
1940 				 * information after all), it would make sense
1941 				 * to simplify this part by always returning
1942 				 * empty data.
1943 				 */
1944 				sopt->sopt_valsize = 0;
1945 				break;
1946 
1947 			case IPV6_RECVHOPOPTS:
1948 			case IPV6_RECVDSTOPTS:
1949 			case IPV6_RECVRTHDRDSTOPTS:
1950 			case IPV6_UNICAST_HOPS:
1951 			case IPV6_RECVPKTINFO:
1952 			case IPV6_RECVHOPLIMIT:
1953 			case IPV6_RECVRTHDR:
1954 			case IPV6_RECVPATHMTU:
1955 
1956 			case IPV6_V6ONLY:
1957 			case IPV6_PORTRANGE:
1958 			case IPV6_RECVTCLASS:
1959 			case IPV6_AUTOFLOWLABEL:
1960 			case IPV6_BINDANY:
1961 			case IPV6_FLOWID:
1962 			case IPV6_FLOWTYPE:
1963 			case IPV6_RECVFLOWID:
1964 #ifdef	RSS
1965 			case IPV6_RSSBUCKETID:
1966 			case IPV6_RECVRSSBUCKETID:
1967 #endif
1968 			case IPV6_BINDMULTI:
1969 				switch (optname) {
1970 
1971 				case IPV6_RECVHOPOPTS:
1972 					optval = OPTBIT(IN6P_HOPOPTS);
1973 					break;
1974 
1975 				case IPV6_RECVDSTOPTS:
1976 					optval = OPTBIT(IN6P_DSTOPTS);
1977 					break;
1978 
1979 				case IPV6_RECVRTHDRDSTOPTS:
1980 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1981 					break;
1982 
1983 				case IPV6_UNICAST_HOPS:
1984 					optval = in6p->in6p_hops;
1985 					break;
1986 
1987 				case IPV6_RECVPKTINFO:
1988 					optval = OPTBIT(IN6P_PKTINFO);
1989 					break;
1990 
1991 				case IPV6_RECVHOPLIMIT:
1992 					optval = OPTBIT(IN6P_HOPLIMIT);
1993 					break;
1994 
1995 				case IPV6_RECVRTHDR:
1996 					optval = OPTBIT(IN6P_RTHDR);
1997 					break;
1998 
1999 				case IPV6_RECVPATHMTU:
2000 					optval = OPTBIT(IN6P_MTU);
2001 					break;
2002 
2003 				case IPV6_V6ONLY:
2004 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
2005 					break;
2006 
2007 				case IPV6_PORTRANGE:
2008 				    {
2009 					int flags;
2010 					flags = in6p->inp_flags;
2011 					if (flags & INP_HIGHPORT)
2012 						optval = IPV6_PORTRANGE_HIGH;
2013 					else if (flags & INP_LOWPORT)
2014 						optval = IPV6_PORTRANGE_LOW;
2015 					else
2016 						optval = 0;
2017 					break;
2018 				    }
2019 				case IPV6_RECVTCLASS:
2020 					optval = OPTBIT(IN6P_TCLASS);
2021 					break;
2022 
2023 				case IPV6_AUTOFLOWLABEL:
2024 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
2025 					break;
2026 
2027 				case IPV6_ORIGDSTADDR:
2028 					optval = OPTBIT2(INP_ORIGDSTADDR);
2029 					break;
2030 
2031 				case IPV6_BINDANY:
2032 					optval = OPTBIT(INP_BINDANY);
2033 					break;
2034 
2035 				case IPV6_FLOWID:
2036 					optval = in6p->inp_flowid;
2037 					break;
2038 
2039 				case IPV6_FLOWTYPE:
2040 					optval = in6p->inp_flowtype;
2041 					break;
2042 
2043 				case IPV6_RECVFLOWID:
2044 					optval = OPTBIT2(INP_RECVFLOWID);
2045 					break;
2046 #ifdef	RSS
2047 				case IPV6_RSSBUCKETID:
2048 					retval =
2049 					    rss_hash2bucket(in6p->inp_flowid,
2050 					    in6p->inp_flowtype,
2051 					    &rss_bucket);
2052 					if (retval == 0)
2053 						optval = rss_bucket;
2054 					else
2055 						error = EINVAL;
2056 					break;
2057 
2058 				case IPV6_RECVRSSBUCKETID:
2059 					optval = OPTBIT2(INP_RECVRSSBUCKETID);
2060 					break;
2061 #endif
2062 
2063 				case IPV6_BINDMULTI:
2064 					optval = OPTBIT2(INP_BINDMULTI);
2065 					break;
2066 
2067 				}
2068 				if (error)
2069 					break;
2070 				error = sooptcopyout(sopt, &optval,
2071 					sizeof optval);
2072 				break;
2073 
2074 			case IPV6_PATHMTU:
2075 			{
2076 				u_long pmtu = 0;
2077 				struct ip6_mtuinfo mtuinfo;
2078 
2079 				if (!(so->so_state & SS_ISCONNECTED))
2080 					return (ENOTCONN);
2081 				/*
2082 				 * XXX: we dot not consider the case of source
2083 				 * routing, or optional information to specify
2084 				 * the outgoing interface.
2085 				 */
2086 				error = ip6_getpmtu_ctl(so->so_fibnum,
2087 				    &in6p->in6p_faddr, &pmtu);
2088 				if (error)
2089 					break;
2090 				if (pmtu > IPV6_MAXPACKET)
2091 					pmtu = IPV6_MAXPACKET;
2092 
2093 				bzero(&mtuinfo, sizeof(mtuinfo));
2094 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2095 				optdata = (void *)&mtuinfo;
2096 				optdatalen = sizeof(mtuinfo);
2097 				error = sooptcopyout(sopt, optdata,
2098 				    optdatalen);
2099 				break;
2100 			}
2101 
2102 			case IPV6_2292PKTINFO:
2103 			case IPV6_2292HOPLIMIT:
2104 			case IPV6_2292HOPOPTS:
2105 			case IPV6_2292RTHDR:
2106 			case IPV6_2292DSTOPTS:
2107 				switch (optname) {
2108 				case IPV6_2292PKTINFO:
2109 					optval = OPTBIT(IN6P_PKTINFO);
2110 					break;
2111 				case IPV6_2292HOPLIMIT:
2112 					optval = OPTBIT(IN6P_HOPLIMIT);
2113 					break;
2114 				case IPV6_2292HOPOPTS:
2115 					optval = OPTBIT(IN6P_HOPOPTS);
2116 					break;
2117 				case IPV6_2292RTHDR:
2118 					optval = OPTBIT(IN6P_RTHDR);
2119 					break;
2120 				case IPV6_2292DSTOPTS:
2121 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2122 					break;
2123 				}
2124 				error = sooptcopyout(sopt, &optval,
2125 				    sizeof optval);
2126 				break;
2127 			case IPV6_PKTINFO:
2128 			case IPV6_HOPOPTS:
2129 			case IPV6_RTHDR:
2130 			case IPV6_DSTOPTS:
2131 			case IPV6_RTHDRDSTOPTS:
2132 			case IPV6_NEXTHOP:
2133 			case IPV6_TCLASS:
2134 			case IPV6_DONTFRAG:
2135 			case IPV6_USE_MIN_MTU:
2136 			case IPV6_PREFER_TEMPADDR:
2137 				error = ip6_getpcbopt(in6p->in6p_outputopts,
2138 				    optname, sopt);
2139 				break;
2140 
2141 			case IPV6_MULTICAST_IF:
2142 			case IPV6_MULTICAST_HOPS:
2143 			case IPV6_MULTICAST_LOOP:
2144 			case IPV6_MSFILTER:
2145 				error = ip6_getmoptions(in6p, sopt);
2146 				break;
2147 
2148 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
2149 			case IPV6_IPSEC_POLICY:
2150 				if (IPSEC_ENABLED(ipv6)) {
2151 					error = IPSEC_PCBCTL(ipv6, in6p, sopt);
2152 					break;
2153 				}
2154 				/* FALLTHROUGH */
2155 #endif /* IPSEC */
2156 			default:
2157 				error = ENOPROTOOPT;
2158 				break;
2159 			}
2160 			break;
2161 		}
2162 	}
2163 	return (error);
2164 }
2165 
2166 int
2167 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2168 {
2169 	int error = 0, optval, optlen;
2170 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2171 	struct inpcb *in6p = sotoinpcb(so);
2172 	int level, op, optname;
2173 
2174 	level = sopt->sopt_level;
2175 	op = sopt->sopt_dir;
2176 	optname = sopt->sopt_name;
2177 	optlen = sopt->sopt_valsize;
2178 
2179 	if (level != IPPROTO_IPV6) {
2180 		return (EINVAL);
2181 	}
2182 
2183 	switch (optname) {
2184 	case IPV6_CHECKSUM:
2185 		/*
2186 		 * For ICMPv6 sockets, no modification allowed for checksum
2187 		 * offset, permit "no change" values to help existing apps.
2188 		 *
2189 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2190 		 * for an ICMPv6 socket will fail."
2191 		 * The current behavior does not meet RFC3542.
2192 		 */
2193 		switch (op) {
2194 		case SOPT_SET:
2195 			if (optlen != sizeof(int)) {
2196 				error = EINVAL;
2197 				break;
2198 			}
2199 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2200 					    sizeof(optval));
2201 			if (error)
2202 				break;
2203 			if ((optval % 2) != 0) {
2204 				/* the API assumes even offset values */
2205 				error = EINVAL;
2206 			} else if (so->so_proto->pr_protocol ==
2207 			    IPPROTO_ICMPV6) {
2208 				if (optval != icmp6off)
2209 					error = EINVAL;
2210 			} else
2211 				in6p->in6p_cksum = optval;
2212 			break;
2213 
2214 		case SOPT_GET:
2215 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2216 				optval = icmp6off;
2217 			else
2218 				optval = in6p->in6p_cksum;
2219 
2220 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2221 			break;
2222 
2223 		default:
2224 			error = EINVAL;
2225 			break;
2226 		}
2227 		break;
2228 
2229 	default:
2230 		error = ENOPROTOOPT;
2231 		break;
2232 	}
2233 
2234 	return (error);
2235 }
2236 
2237 /*
2238  * Set up IP6 options in pcb for insertion in output packets or
2239  * specifying behavior of outgoing packets.
2240  */
2241 static int
2242 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2243     struct socket *so, struct sockopt *sopt)
2244 {
2245 	struct ip6_pktopts *opt = *pktopt;
2246 	int error = 0;
2247 	struct thread *td = sopt->sopt_td;
2248 
2249 	/* turn off any old options. */
2250 	if (opt) {
2251 #ifdef DIAGNOSTIC
2252 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2253 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2254 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2255 			printf("ip6_pcbopts: all specified options are cleared.\n");
2256 #endif
2257 		ip6_clearpktopts(opt, -1);
2258 	} else
2259 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2260 	*pktopt = NULL;
2261 
2262 	if (!m || m->m_len == 0) {
2263 		/*
2264 		 * Only turning off any previous options, regardless of
2265 		 * whether the opt is just created or given.
2266 		 */
2267 		free(opt, M_IP6OPT);
2268 		return (0);
2269 	}
2270 
2271 	/*  set options specified by user. */
2272 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2273 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2274 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2275 		free(opt, M_IP6OPT);
2276 		return (error);
2277 	}
2278 	*pktopt = opt;
2279 	return (0);
2280 }
2281 
2282 /*
2283  * initialize ip6_pktopts.  beware that there are non-zero default values in
2284  * the struct.
2285  */
2286 void
2287 ip6_initpktopts(struct ip6_pktopts *opt)
2288 {
2289 
2290 	bzero(opt, sizeof(*opt));
2291 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2292 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2293 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2294 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2295 }
2296 
2297 static int
2298 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2299     struct ucred *cred, int uproto)
2300 {
2301 	struct ip6_pktopts *opt;
2302 
2303 	if (*pktopt == NULL) {
2304 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2305 		    M_WAITOK);
2306 		ip6_initpktopts(*pktopt);
2307 	}
2308 	opt = *pktopt;
2309 
2310 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2311 }
2312 
2313 static int
2314 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2315 {
2316 	void *optdata = NULL;
2317 	int optdatalen = 0;
2318 	struct ip6_ext *ip6e;
2319 	int error = 0;
2320 	struct in6_pktinfo null_pktinfo;
2321 	int deftclass = 0, on;
2322 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2323 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2324 
2325 	switch (optname) {
2326 	case IPV6_PKTINFO:
2327 		optdata = (void *)&null_pktinfo;
2328 		if (pktopt && pktopt->ip6po_pktinfo) {
2329 			bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2330 			    sizeof(null_pktinfo));
2331 			in6_clearscope(&null_pktinfo.ipi6_addr);
2332 		} else {
2333 			/* XXX: we don't have to do this every time... */
2334 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2335 		}
2336 		optdatalen = sizeof(struct in6_pktinfo);
2337 		break;
2338 	case IPV6_TCLASS:
2339 		if (pktopt && pktopt->ip6po_tclass >= 0)
2340 			optdata = (void *)&pktopt->ip6po_tclass;
2341 		else
2342 			optdata = (void *)&deftclass;
2343 		optdatalen = sizeof(int);
2344 		break;
2345 	case IPV6_HOPOPTS:
2346 		if (pktopt && pktopt->ip6po_hbh) {
2347 			optdata = (void *)pktopt->ip6po_hbh;
2348 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2349 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2350 		}
2351 		break;
2352 	case IPV6_RTHDR:
2353 		if (pktopt && pktopt->ip6po_rthdr) {
2354 			optdata = (void *)pktopt->ip6po_rthdr;
2355 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2356 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2357 		}
2358 		break;
2359 	case IPV6_RTHDRDSTOPTS:
2360 		if (pktopt && pktopt->ip6po_dest1) {
2361 			optdata = (void *)pktopt->ip6po_dest1;
2362 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2363 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2364 		}
2365 		break;
2366 	case IPV6_DSTOPTS:
2367 		if (pktopt && pktopt->ip6po_dest2) {
2368 			optdata = (void *)pktopt->ip6po_dest2;
2369 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2370 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2371 		}
2372 		break;
2373 	case IPV6_NEXTHOP:
2374 		if (pktopt && pktopt->ip6po_nexthop) {
2375 			optdata = (void *)pktopt->ip6po_nexthop;
2376 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2377 		}
2378 		break;
2379 	case IPV6_USE_MIN_MTU:
2380 		if (pktopt)
2381 			optdata = (void *)&pktopt->ip6po_minmtu;
2382 		else
2383 			optdata = (void *)&defminmtu;
2384 		optdatalen = sizeof(int);
2385 		break;
2386 	case IPV6_DONTFRAG:
2387 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2388 			on = 1;
2389 		else
2390 			on = 0;
2391 		optdata = (void *)&on;
2392 		optdatalen = sizeof(on);
2393 		break;
2394 	case IPV6_PREFER_TEMPADDR:
2395 		if (pktopt)
2396 			optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2397 		else
2398 			optdata = (void *)&defpreftemp;
2399 		optdatalen = sizeof(int);
2400 		break;
2401 	default:		/* should not happen */
2402 #ifdef DIAGNOSTIC
2403 		panic("ip6_getpcbopt: unexpected option\n");
2404 #endif
2405 		return (ENOPROTOOPT);
2406 	}
2407 
2408 	error = sooptcopyout(sopt, optdata, optdatalen);
2409 
2410 	return (error);
2411 }
2412 
2413 void
2414 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2415 {
2416 	if (pktopt == NULL)
2417 		return;
2418 
2419 	if (optname == -1 || optname == IPV6_PKTINFO) {
2420 		if (pktopt->ip6po_pktinfo)
2421 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2422 		pktopt->ip6po_pktinfo = NULL;
2423 	}
2424 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2425 		pktopt->ip6po_hlim = -1;
2426 	if (optname == -1 || optname == IPV6_TCLASS)
2427 		pktopt->ip6po_tclass = -1;
2428 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2429 		if (pktopt->ip6po_nextroute.ro_rt) {
2430 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2431 			pktopt->ip6po_nextroute.ro_rt = NULL;
2432 		}
2433 		if (pktopt->ip6po_nexthop)
2434 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2435 		pktopt->ip6po_nexthop = NULL;
2436 	}
2437 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2438 		if (pktopt->ip6po_hbh)
2439 			free(pktopt->ip6po_hbh, M_IP6OPT);
2440 		pktopt->ip6po_hbh = NULL;
2441 	}
2442 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2443 		if (pktopt->ip6po_dest1)
2444 			free(pktopt->ip6po_dest1, M_IP6OPT);
2445 		pktopt->ip6po_dest1 = NULL;
2446 	}
2447 	if (optname == -1 || optname == IPV6_RTHDR) {
2448 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2449 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2450 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2451 		if (pktopt->ip6po_route.ro_rt) {
2452 			RTFREE(pktopt->ip6po_route.ro_rt);
2453 			pktopt->ip6po_route.ro_rt = NULL;
2454 		}
2455 	}
2456 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2457 		if (pktopt->ip6po_dest2)
2458 			free(pktopt->ip6po_dest2, M_IP6OPT);
2459 		pktopt->ip6po_dest2 = NULL;
2460 	}
2461 }
2462 
2463 #define PKTOPT_EXTHDRCPY(type) \
2464 do {\
2465 	if (src->type) {\
2466 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2467 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2468 		if (dst->type == NULL)\
2469 			goto bad;\
2470 		bcopy(src->type, dst->type, hlen);\
2471 	}\
2472 } while (/*CONSTCOND*/ 0)
2473 
2474 static int
2475 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2476 {
2477 	if (dst == NULL || src == NULL)  {
2478 		printf("ip6_clearpktopts: invalid argument\n");
2479 		return (EINVAL);
2480 	}
2481 
2482 	dst->ip6po_hlim = src->ip6po_hlim;
2483 	dst->ip6po_tclass = src->ip6po_tclass;
2484 	dst->ip6po_flags = src->ip6po_flags;
2485 	dst->ip6po_minmtu = src->ip6po_minmtu;
2486 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2487 	if (src->ip6po_pktinfo) {
2488 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2489 		    M_IP6OPT, canwait);
2490 		if (dst->ip6po_pktinfo == NULL)
2491 			goto bad;
2492 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2493 	}
2494 	if (src->ip6po_nexthop) {
2495 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2496 		    M_IP6OPT, canwait);
2497 		if (dst->ip6po_nexthop == NULL)
2498 			goto bad;
2499 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2500 		    src->ip6po_nexthop->sa_len);
2501 	}
2502 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2503 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2504 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2505 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2506 	return (0);
2507 
2508   bad:
2509 	ip6_clearpktopts(dst, -1);
2510 	return (ENOBUFS);
2511 }
2512 #undef PKTOPT_EXTHDRCPY
2513 
2514 struct ip6_pktopts *
2515 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2516 {
2517 	int error;
2518 	struct ip6_pktopts *dst;
2519 
2520 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2521 	if (dst == NULL)
2522 		return (NULL);
2523 	ip6_initpktopts(dst);
2524 
2525 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2526 		free(dst, M_IP6OPT);
2527 		return (NULL);
2528 	}
2529 
2530 	return (dst);
2531 }
2532 
2533 void
2534 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2535 {
2536 	if (pktopt == NULL)
2537 		return;
2538 
2539 	ip6_clearpktopts(pktopt, -1);
2540 
2541 	free(pktopt, M_IP6OPT);
2542 }
2543 
2544 /*
2545  * Set IPv6 outgoing packet options based on advanced API.
2546  */
2547 int
2548 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2549     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2550 {
2551 	struct cmsghdr *cm = NULL;
2552 
2553 	if (control == NULL || opt == NULL)
2554 		return (EINVAL);
2555 
2556 	ip6_initpktopts(opt);
2557 	if (stickyopt) {
2558 		int error;
2559 
2560 		/*
2561 		 * If stickyopt is provided, make a local copy of the options
2562 		 * for this particular packet, then override them by ancillary
2563 		 * objects.
2564 		 * XXX: copypktopts() does not copy the cached route to a next
2565 		 * hop (if any).  This is not very good in terms of efficiency,
2566 		 * but we can allow this since this option should be rarely
2567 		 * used.
2568 		 */
2569 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2570 			return (error);
2571 	}
2572 
2573 	/*
2574 	 * XXX: Currently, we assume all the optional information is stored
2575 	 * in a single mbuf.
2576 	 */
2577 	if (control->m_next)
2578 		return (EINVAL);
2579 
2580 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2581 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2582 		int error;
2583 
2584 		if (control->m_len < CMSG_LEN(0))
2585 			return (EINVAL);
2586 
2587 		cm = mtod(control, struct cmsghdr *);
2588 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2589 			return (EINVAL);
2590 		if (cm->cmsg_level != IPPROTO_IPV6)
2591 			continue;
2592 
2593 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2594 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2595 		if (error)
2596 			return (error);
2597 	}
2598 
2599 	return (0);
2600 }
2601 
2602 /*
2603  * Set a particular packet option, as a sticky option or an ancillary data
2604  * item.  "len" can be 0 only when it's a sticky option.
2605  * We have 4 cases of combination of "sticky" and "cmsg":
2606  * "sticky=0, cmsg=0": impossible
2607  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2608  * "sticky=1, cmsg=0": RFC3542 socket option
2609  * "sticky=1, cmsg=1": RFC2292 socket option
2610  */
2611 static int
2612 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2613     struct ucred *cred, int sticky, int cmsg, int uproto)
2614 {
2615 	int minmtupolicy, preftemp;
2616 	int error;
2617 
2618 	if (!sticky && !cmsg) {
2619 #ifdef DIAGNOSTIC
2620 		printf("ip6_setpktopt: impossible case\n");
2621 #endif
2622 		return (EINVAL);
2623 	}
2624 
2625 	/*
2626 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2627 	 * not be specified in the context of RFC3542.  Conversely,
2628 	 * RFC3542 types should not be specified in the context of RFC2292.
2629 	 */
2630 	if (!cmsg) {
2631 		switch (optname) {
2632 		case IPV6_2292PKTINFO:
2633 		case IPV6_2292HOPLIMIT:
2634 		case IPV6_2292NEXTHOP:
2635 		case IPV6_2292HOPOPTS:
2636 		case IPV6_2292DSTOPTS:
2637 		case IPV6_2292RTHDR:
2638 		case IPV6_2292PKTOPTIONS:
2639 			return (ENOPROTOOPT);
2640 		}
2641 	}
2642 	if (sticky && cmsg) {
2643 		switch (optname) {
2644 		case IPV6_PKTINFO:
2645 		case IPV6_HOPLIMIT:
2646 		case IPV6_NEXTHOP:
2647 		case IPV6_HOPOPTS:
2648 		case IPV6_DSTOPTS:
2649 		case IPV6_RTHDRDSTOPTS:
2650 		case IPV6_RTHDR:
2651 		case IPV6_USE_MIN_MTU:
2652 		case IPV6_DONTFRAG:
2653 		case IPV6_TCLASS:
2654 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2655 			return (ENOPROTOOPT);
2656 		}
2657 	}
2658 
2659 	switch (optname) {
2660 	case IPV6_2292PKTINFO:
2661 	case IPV6_PKTINFO:
2662 	{
2663 		struct ifnet *ifp = NULL;
2664 		struct in6_pktinfo *pktinfo;
2665 
2666 		if (len != sizeof(struct in6_pktinfo))
2667 			return (EINVAL);
2668 
2669 		pktinfo = (struct in6_pktinfo *)buf;
2670 
2671 		/*
2672 		 * An application can clear any sticky IPV6_PKTINFO option by
2673 		 * doing a "regular" setsockopt with ipi6_addr being
2674 		 * in6addr_any and ipi6_ifindex being zero.
2675 		 * [RFC 3542, Section 6]
2676 		 */
2677 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2678 		    pktinfo->ipi6_ifindex == 0 &&
2679 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2680 			ip6_clearpktopts(opt, optname);
2681 			break;
2682 		}
2683 
2684 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2685 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2686 			return (EINVAL);
2687 		}
2688 		if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2689 			return (EINVAL);
2690 		/* validate the interface index if specified. */
2691 		if (pktinfo->ipi6_ifindex > V_if_index)
2692 			 return (ENXIO);
2693 		if (pktinfo->ipi6_ifindex) {
2694 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2695 			if (ifp == NULL)
2696 				return (ENXIO);
2697 		}
2698 		if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2699 		    (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2700 			return (ENETDOWN);
2701 
2702 		if (ifp != NULL &&
2703 		    !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2704 			struct in6_ifaddr *ia;
2705 
2706 			in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2707 			ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2708 			if (ia == NULL)
2709 				return (EADDRNOTAVAIL);
2710 			ifa_free(&ia->ia_ifa);
2711 		}
2712 		/*
2713 		 * We store the address anyway, and let in6_selectsrc()
2714 		 * validate the specified address.  This is because ipi6_addr
2715 		 * may not have enough information about its scope zone, and
2716 		 * we may need additional information (such as outgoing
2717 		 * interface or the scope zone of a destination address) to
2718 		 * disambiguate the scope.
2719 		 * XXX: the delay of the validation may confuse the
2720 		 * application when it is used as a sticky option.
2721 		 */
2722 		if (opt->ip6po_pktinfo == NULL) {
2723 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2724 			    M_IP6OPT, M_NOWAIT);
2725 			if (opt->ip6po_pktinfo == NULL)
2726 				return (ENOBUFS);
2727 		}
2728 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2729 		break;
2730 	}
2731 
2732 	case IPV6_2292HOPLIMIT:
2733 	case IPV6_HOPLIMIT:
2734 	{
2735 		int *hlimp;
2736 
2737 		/*
2738 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2739 		 * to simplify the ordering among hoplimit options.
2740 		 */
2741 		if (optname == IPV6_HOPLIMIT && sticky)
2742 			return (ENOPROTOOPT);
2743 
2744 		if (len != sizeof(int))
2745 			return (EINVAL);
2746 		hlimp = (int *)buf;
2747 		if (*hlimp < -1 || *hlimp > 255)
2748 			return (EINVAL);
2749 
2750 		opt->ip6po_hlim = *hlimp;
2751 		break;
2752 	}
2753 
2754 	case IPV6_TCLASS:
2755 	{
2756 		int tclass;
2757 
2758 		if (len != sizeof(int))
2759 			return (EINVAL);
2760 		tclass = *(int *)buf;
2761 		if (tclass < -1 || tclass > 255)
2762 			return (EINVAL);
2763 
2764 		opt->ip6po_tclass = tclass;
2765 		break;
2766 	}
2767 
2768 	case IPV6_2292NEXTHOP:
2769 	case IPV6_NEXTHOP:
2770 		if (cred != NULL) {
2771 			error = priv_check_cred(cred,
2772 			    PRIV_NETINET_SETHDROPTS, 0);
2773 			if (error)
2774 				return (error);
2775 		}
2776 
2777 		if (len == 0) {	/* just remove the option */
2778 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2779 			break;
2780 		}
2781 
2782 		/* check if cmsg_len is large enough for sa_len */
2783 		if (len < sizeof(struct sockaddr) || len < *buf)
2784 			return (EINVAL);
2785 
2786 		switch (((struct sockaddr *)buf)->sa_family) {
2787 		case AF_INET6:
2788 		{
2789 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2790 			int error;
2791 
2792 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2793 				return (EINVAL);
2794 
2795 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2796 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2797 				return (EINVAL);
2798 			}
2799 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2800 			    != 0) {
2801 				return (error);
2802 			}
2803 			break;
2804 		}
2805 		case AF_LINK:	/* should eventually be supported */
2806 		default:
2807 			return (EAFNOSUPPORT);
2808 		}
2809 
2810 		/* turn off the previous option, then set the new option. */
2811 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2812 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2813 		if (opt->ip6po_nexthop == NULL)
2814 			return (ENOBUFS);
2815 		bcopy(buf, opt->ip6po_nexthop, *buf);
2816 		break;
2817 
2818 	case IPV6_2292HOPOPTS:
2819 	case IPV6_HOPOPTS:
2820 	{
2821 		struct ip6_hbh *hbh;
2822 		int hbhlen;
2823 
2824 		/*
2825 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2826 		 * options, since per-option restriction has too much
2827 		 * overhead.
2828 		 */
2829 		if (cred != NULL) {
2830 			error = priv_check_cred(cred,
2831 			    PRIV_NETINET_SETHDROPTS, 0);
2832 			if (error)
2833 				return (error);
2834 		}
2835 
2836 		if (len == 0) {
2837 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2838 			break;	/* just remove the option */
2839 		}
2840 
2841 		/* message length validation */
2842 		if (len < sizeof(struct ip6_hbh))
2843 			return (EINVAL);
2844 		hbh = (struct ip6_hbh *)buf;
2845 		hbhlen = (hbh->ip6h_len + 1) << 3;
2846 		if (len != hbhlen)
2847 			return (EINVAL);
2848 
2849 		/* turn off the previous option, then set the new option. */
2850 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2851 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2852 		if (opt->ip6po_hbh == NULL)
2853 			return (ENOBUFS);
2854 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
2855 
2856 		break;
2857 	}
2858 
2859 	case IPV6_2292DSTOPTS:
2860 	case IPV6_DSTOPTS:
2861 	case IPV6_RTHDRDSTOPTS:
2862 	{
2863 		struct ip6_dest *dest, **newdest = NULL;
2864 		int destlen;
2865 
2866 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2867 			error = priv_check_cred(cred,
2868 			    PRIV_NETINET_SETHDROPTS, 0);
2869 			if (error)
2870 				return (error);
2871 		}
2872 
2873 		if (len == 0) {
2874 			ip6_clearpktopts(opt, optname);
2875 			break;	/* just remove the option */
2876 		}
2877 
2878 		/* message length validation */
2879 		if (len < sizeof(struct ip6_dest))
2880 			return (EINVAL);
2881 		dest = (struct ip6_dest *)buf;
2882 		destlen = (dest->ip6d_len + 1) << 3;
2883 		if (len != destlen)
2884 			return (EINVAL);
2885 
2886 		/*
2887 		 * Determine the position that the destination options header
2888 		 * should be inserted; before or after the routing header.
2889 		 */
2890 		switch (optname) {
2891 		case IPV6_2292DSTOPTS:
2892 			/*
2893 			 * The old advacned API is ambiguous on this point.
2894 			 * Our approach is to determine the position based
2895 			 * according to the existence of a routing header.
2896 			 * Note, however, that this depends on the order of the
2897 			 * extension headers in the ancillary data; the 1st
2898 			 * part of the destination options header must appear
2899 			 * before the routing header in the ancillary data,
2900 			 * too.
2901 			 * RFC3542 solved the ambiguity by introducing
2902 			 * separate ancillary data or option types.
2903 			 */
2904 			if (opt->ip6po_rthdr == NULL)
2905 				newdest = &opt->ip6po_dest1;
2906 			else
2907 				newdest = &opt->ip6po_dest2;
2908 			break;
2909 		case IPV6_RTHDRDSTOPTS:
2910 			newdest = &opt->ip6po_dest1;
2911 			break;
2912 		case IPV6_DSTOPTS:
2913 			newdest = &opt->ip6po_dest2;
2914 			break;
2915 		}
2916 
2917 		/* turn off the previous option, then set the new option. */
2918 		ip6_clearpktopts(opt, optname);
2919 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2920 		if (*newdest == NULL)
2921 			return (ENOBUFS);
2922 		bcopy(dest, *newdest, destlen);
2923 
2924 		break;
2925 	}
2926 
2927 	case IPV6_2292RTHDR:
2928 	case IPV6_RTHDR:
2929 	{
2930 		struct ip6_rthdr *rth;
2931 		int rthlen;
2932 
2933 		if (len == 0) {
2934 			ip6_clearpktopts(opt, IPV6_RTHDR);
2935 			break;	/* just remove the option */
2936 		}
2937 
2938 		/* message length validation */
2939 		if (len < sizeof(struct ip6_rthdr))
2940 			return (EINVAL);
2941 		rth = (struct ip6_rthdr *)buf;
2942 		rthlen = (rth->ip6r_len + 1) << 3;
2943 		if (len != rthlen)
2944 			return (EINVAL);
2945 
2946 		switch (rth->ip6r_type) {
2947 		case IPV6_RTHDR_TYPE_0:
2948 			if (rth->ip6r_len == 0)	/* must contain one addr */
2949 				return (EINVAL);
2950 			if (rth->ip6r_len % 2) /* length must be even */
2951 				return (EINVAL);
2952 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2953 				return (EINVAL);
2954 			break;
2955 		default:
2956 			return (EINVAL);	/* not supported */
2957 		}
2958 
2959 		/* turn off the previous option */
2960 		ip6_clearpktopts(opt, IPV6_RTHDR);
2961 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2962 		if (opt->ip6po_rthdr == NULL)
2963 			return (ENOBUFS);
2964 		bcopy(rth, opt->ip6po_rthdr, rthlen);
2965 
2966 		break;
2967 	}
2968 
2969 	case IPV6_USE_MIN_MTU:
2970 		if (len != sizeof(int))
2971 			return (EINVAL);
2972 		minmtupolicy = *(int *)buf;
2973 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2974 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
2975 		    minmtupolicy != IP6PO_MINMTU_ALL) {
2976 			return (EINVAL);
2977 		}
2978 		opt->ip6po_minmtu = minmtupolicy;
2979 		break;
2980 
2981 	case IPV6_DONTFRAG:
2982 		if (len != sizeof(int))
2983 			return (EINVAL);
2984 
2985 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2986 			/*
2987 			 * we ignore this option for TCP sockets.
2988 			 * (RFC3542 leaves this case unspecified.)
2989 			 */
2990 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2991 		} else
2992 			opt->ip6po_flags |= IP6PO_DONTFRAG;
2993 		break;
2994 
2995 	case IPV6_PREFER_TEMPADDR:
2996 		if (len != sizeof(int))
2997 			return (EINVAL);
2998 		preftemp = *(int *)buf;
2999 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
3000 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
3001 		    preftemp != IP6PO_TEMPADDR_PREFER) {
3002 			return (EINVAL);
3003 		}
3004 		opt->ip6po_prefer_tempaddr = preftemp;
3005 		break;
3006 
3007 	default:
3008 		return (ENOPROTOOPT);
3009 	} /* end of switch */
3010 
3011 	return (0);
3012 }
3013 
3014 /*
3015  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3016  * packet to the input queue of a specified interface.  Note that this
3017  * calls the output routine of the loopback "driver", but with an interface
3018  * pointer that might NOT be &loif -- easier than replicating that code here.
3019  */
3020 void
3021 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3022 {
3023 	struct mbuf *copym;
3024 	struct ip6_hdr *ip6;
3025 
3026 	copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3027 	if (copym == NULL)
3028 		return;
3029 
3030 	/*
3031 	 * Make sure to deep-copy IPv6 header portion in case the data
3032 	 * is in an mbuf cluster, so that we can safely override the IPv6
3033 	 * header portion later.
3034 	 */
3035 	if (!M_WRITABLE(copym) ||
3036 	    copym->m_len < sizeof(struct ip6_hdr)) {
3037 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3038 		if (copym == NULL)
3039 			return;
3040 	}
3041 	ip6 = mtod(copym, struct ip6_hdr *);
3042 	/*
3043 	 * clear embedded scope identifiers if necessary.
3044 	 * in6_clearscope will touch the addresses only when necessary.
3045 	 */
3046 	in6_clearscope(&ip6->ip6_src);
3047 	in6_clearscope(&ip6->ip6_dst);
3048 	if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3049 		copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3050 		    CSUM_PSEUDO_HDR;
3051 		copym->m_pkthdr.csum_data = 0xffff;
3052 	}
3053 	if_simloop(ifp, copym, AF_INET6, 0);
3054 }
3055 
3056 /*
3057  * Chop IPv6 header off from the payload.
3058  */
3059 static int
3060 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3061 {
3062 	struct mbuf *mh;
3063 	struct ip6_hdr *ip6;
3064 
3065 	ip6 = mtod(m, struct ip6_hdr *);
3066 	if (m->m_len > sizeof(*ip6)) {
3067 		mh = m_gethdr(M_NOWAIT, MT_DATA);
3068 		if (mh == NULL) {
3069 			m_freem(m);
3070 			return ENOBUFS;
3071 		}
3072 		m_move_pkthdr(mh, m);
3073 		M_ALIGN(mh, sizeof(*ip6));
3074 		m->m_len -= sizeof(*ip6);
3075 		m->m_data += sizeof(*ip6);
3076 		mh->m_next = m;
3077 		m = mh;
3078 		m->m_len = sizeof(*ip6);
3079 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3080 	}
3081 	exthdrs->ip6e_ip6 = m;
3082 	return 0;
3083 }
3084 
3085 /*
3086  * Compute IPv6 extension header length.
3087  */
3088 int
3089 ip6_optlen(struct inpcb *in6p)
3090 {
3091 	int len;
3092 
3093 	if (!in6p->in6p_outputopts)
3094 		return 0;
3095 
3096 	len = 0;
3097 #define elen(x) \
3098     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3099 
3100 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3101 	if (in6p->in6p_outputopts->ip6po_rthdr)
3102 		/* dest1 is valid with rthdr only */
3103 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
3104 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3105 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3106 	return len;
3107 #undef elen
3108 }
3109