xref: /freebsd/sys/netinet6/ip6_output.c (revision 5dae51da3da0cc94d17bd67b308fad304ebec7e0)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: ip6_output.c,v 1.279 2002/01/26 06:12:30 jinmei Exp $
30  */
31 
32 /*-
33  * Copyright (c) 1982, 1986, 1988, 1990, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
61  */
62 
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
65 
66 #include "opt_inet.h"
67 #include "opt_inet6.h"
68 #include "opt_ipsec.h"
69 #include "opt_sctp.h"
70 #include "opt_route.h"
71 #include "opt_rss.h"
72 
73 #include <sys/param.h>
74 #include <sys/kernel.h>
75 #include <sys/malloc.h>
76 #include <sys/mbuf.h>
77 #include <sys/errno.h>
78 #include <sys/priv.h>
79 #include <sys/proc.h>
80 #include <sys/protosw.h>
81 #include <sys/socket.h>
82 #include <sys/socketvar.h>
83 #include <sys/syslog.h>
84 #include <sys/ucred.h>
85 
86 #include <machine/in_cksum.h>
87 
88 #include <net/if.h>
89 #include <net/if_var.h>
90 #include <net/if_llatbl.h>
91 #include <net/netisr.h>
92 #include <net/route.h>
93 #include <net/pfil.h>
94 #include <net/rss_config.h>
95 #include <net/vnet.h>
96 
97 #include <netinet/in.h>
98 #include <netinet/in_var.h>
99 #include <netinet/ip_var.h>
100 #include <netinet6/in6_fib.h>
101 #include <netinet6/in6_var.h>
102 #include <netinet/ip6.h>
103 #include <netinet/icmp6.h>
104 #include <netinet6/ip6_var.h>
105 #include <netinet/in_pcb.h>
106 #include <netinet/tcp_var.h>
107 #include <netinet6/nd6.h>
108 #include <netinet6/in6_rss.h>
109 
110 #ifdef IPSEC
111 #include <netipsec/ipsec.h>
112 #include <netipsec/ipsec6.h>
113 #include <netipsec/key.h>
114 #include <netinet6/ip6_ipsec.h>
115 #endif /* IPSEC */
116 #ifdef SCTP
117 #include <netinet/sctp.h>
118 #include <netinet/sctp_crc32.h>
119 #endif
120 
121 #include <netinet6/ip6protosw.h>
122 #include <netinet6/scope6_var.h>
123 
124 #ifdef FLOWTABLE
125 #include <net/flowtable.h>
126 #endif
127 
128 extern int in6_mcast_loop;
129 
130 struct ip6_exthdrs {
131 	struct mbuf *ip6e_ip6;
132 	struct mbuf *ip6e_hbh;
133 	struct mbuf *ip6e_dest1;
134 	struct mbuf *ip6e_rthdr;
135 	struct mbuf *ip6e_dest2;
136 };
137 
138 static MALLOC_DEFINE(M_IP6OPT, "ip6opt", "IPv6 options");
139 
140 static int ip6_pcbopt(int, u_char *, int, struct ip6_pktopts **,
141 			   struct ucred *, int);
142 static int ip6_pcbopts(struct ip6_pktopts **, struct mbuf *,
143 	struct socket *, struct sockopt *);
144 static int ip6_getpcbopt(struct ip6_pktopts *, int, struct sockopt *);
145 static int ip6_setpktopt(int, u_char *, int, struct ip6_pktopts *,
146 	struct ucred *, int, int, int);
147 
148 static int ip6_copyexthdr(struct mbuf **, caddr_t, int);
149 static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
150 	struct ip6_frag **);
151 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
152 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
153 static int ip6_getpmtu(struct route_in6 *, int,
154 	struct ifnet *, const struct in6_addr *, u_long *, int *, u_int,
155 	u_int);
156 static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
157 	u_long *, int *, u_int);
158 static int ip6_getpmtu_ctl(u_int, const struct in6_addr *, u_long *);
159 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
160 
161 
162 /*
163  * Make an extension header from option data.  hp is the source, and
164  * mp is the destination.
165  */
166 #define MAKE_EXTHDR(hp, mp)						\
167     do {								\
168 	if (hp) {							\
169 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
170 		error = ip6_copyexthdr((mp), (caddr_t)(hp),		\
171 		    ((eh)->ip6e_len + 1) << 3);				\
172 		if (error)						\
173 			goto freehdrs;					\
174 	}								\
175     } while (/*CONSTCOND*/ 0)
176 
177 /*
178  * Form a chain of extension headers.
179  * m is the extension header mbuf
180  * mp is the previous mbuf in the chain
181  * p is the next header
182  * i is the type of option.
183  */
184 #define MAKE_CHAIN(m, mp, p, i)\
185     do {\
186 	if (m) {\
187 		if (!hdrsplit) \
188 			panic("assumption failed: hdr not split"); \
189 		*mtod((m), u_char *) = *(p);\
190 		*(p) = (i);\
191 		p = mtod((m), u_char *);\
192 		(m)->m_next = (mp)->m_next;\
193 		(mp)->m_next = (m);\
194 		(mp) = (m);\
195 	}\
196     } while (/*CONSTCOND*/ 0)
197 
198 void
199 in6_delayed_cksum(struct mbuf *m, uint32_t plen, u_short offset)
200 {
201 	u_short csum;
202 
203 	csum = in_cksum_skip(m, offset + plen, offset);
204 	if (m->m_pkthdr.csum_flags & CSUM_UDP_IPV6 && csum == 0)
205 		csum = 0xffff;
206 	offset += m->m_pkthdr.csum_data;	/* checksum offset */
207 
208 	if (offset + sizeof(u_short) > m->m_len) {
209 		printf("%s: delayed m_pullup, m->len: %d plen %u off %u "
210 		    "csum_flags=%b\n", __func__, m->m_len, plen, offset,
211 		    (int)m->m_pkthdr.csum_flags, CSUM_BITS);
212 		/*
213 		 * XXX this should not happen, but if it does, the correct
214 		 * behavior may be to insert the checksum in the appropriate
215 		 * next mbuf in the chain.
216 		 */
217 		return;
218 	}
219 	*(u_short *)(m->m_data + offset) = csum;
220 }
221 
222 int
223 ip6_fragment(struct ifnet *ifp, struct mbuf *m0, int hlen, u_char nextproto,
224     int mtu, uint32_t id)
225 {
226 	struct mbuf *m, **mnext, *m_frgpart;
227 	struct ip6_hdr *ip6, *mhip6;
228 	struct ip6_frag *ip6f;
229 	int off;
230 	int error;
231 	int tlen = m0->m_pkthdr.len;
232 
233 	m = m0;
234 	ip6 = mtod(m, struct ip6_hdr *);
235 	mnext = &m->m_nextpkt;
236 
237 	for (off = hlen; off < tlen; off += mtu) {
238 		m = m_gethdr(M_NOWAIT, MT_DATA);
239 		if (!m) {
240 			IP6STAT_INC(ip6s_odropped);
241 			return (ENOBUFS);
242 		}
243 		m->m_flags = m0->m_flags & M_COPYFLAGS;
244 		*mnext = m;
245 		mnext = &m->m_nextpkt;
246 		m->m_data += max_linkhdr;
247 		mhip6 = mtod(m, struct ip6_hdr *);
248 		*mhip6 = *ip6;
249 		m->m_len = sizeof(*mhip6);
250 		error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
251 		if (error) {
252 			IP6STAT_INC(ip6s_odropped);
253 			return (error);
254 		}
255 		ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
256 		if (off + mtu >= tlen)
257 			mtu = tlen - off;
258 		else
259 			ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
260 		mhip6->ip6_plen = htons((u_short)(mtu + hlen +
261 		    sizeof(*ip6f) - sizeof(struct ip6_hdr)));
262 		if ((m_frgpart = m_copym(m0, off, mtu, M_NOWAIT)) == NULL) {
263 			IP6STAT_INC(ip6s_odropped);
264 			return (ENOBUFS);
265 		}
266 		m_cat(m, m_frgpart);
267 		m->m_pkthdr.len = mtu + hlen + sizeof(*ip6f);
268 		m->m_pkthdr.fibnum = m0->m_pkthdr.fibnum;
269 		m->m_pkthdr.rcvif = NULL;
270 		ip6f->ip6f_reserved = 0;
271 		ip6f->ip6f_ident = id;
272 		ip6f->ip6f_nxt = nextproto;
273 		IP6STAT_INC(ip6s_ofragments);
274 		in6_ifstat_inc(ifp, ifs6_out_fragcreat);
275 	}
276 
277 	return (0);
278 }
279 
280 /*
281  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
282  * header (with pri, len, nxt, hlim, src, dst).
283  * This function may modify ver and hlim only.
284  * The mbuf chain containing the packet will be freed.
285  * The mbuf opt, if present, will not be freed.
286  * If route_in6 ro is present and has ro_rt initialized, route lookup would be
287  * skipped and ro->ro_rt would be used. If ro is present but ro->ro_rt is NULL,
288  * then result of route lookup is stored in ro->ro_rt.
289  *
290  * type of "mtu": rt_mtu is u_long, ifnet.ifr_mtu is int, and
291  * nd_ifinfo.linkmtu is u_int32_t.  so we use u_long to hold largest one,
292  * which is rt_mtu.
293  *
294  * ifpp - XXX: just for statistics
295  */
296 /*
297  * XXX TODO: no flowid is assigned for outbound flows?
298  */
299 int
300 ip6_output(struct mbuf *m0, struct ip6_pktopts *opt,
301     struct route_in6 *ro, int flags, struct ip6_moptions *im6o,
302     struct ifnet **ifpp, struct inpcb *inp)
303 {
304 	struct ip6_hdr *ip6;
305 	struct ifnet *ifp, *origifp;
306 	struct mbuf *m = m0;
307 	struct mbuf *mprev = NULL;
308 	int hlen, tlen, len;
309 	struct route_in6 ip6route;
310 	struct rtentry *rt = NULL;
311 	struct sockaddr_in6 *dst, src_sa, dst_sa;
312 	struct in6_addr odst;
313 	int error = 0;
314 	struct in6_ifaddr *ia = NULL;
315 	u_long mtu;
316 	int alwaysfrag, dontfrag;
317 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
318 	struct ip6_exthdrs exthdrs;
319 	struct in6_addr src0, dst0;
320 	u_int32_t zone;
321 	struct route_in6 *ro_pmtu = NULL;
322 	int hdrsplit = 0;
323 	int sw_csum, tso;
324 	int needfiblookup;
325 	uint32_t fibnum;
326 	struct m_tag *fwd_tag = NULL;
327 	uint32_t id;
328 
329 	if (inp != NULL) {
330 		M_SETFIB(m, inp->inp_inc.inc_fibnum);
331 		if ((flags & IP_NODEFAULTFLOWID) == 0) {
332 			/* unconditionally set flowid */
333 			m->m_pkthdr.flowid = inp->inp_flowid;
334 			M_HASHTYPE_SET(m, inp->inp_flowtype);
335 		}
336 	}
337 
338 	bzero(&exthdrs, sizeof(exthdrs));
339 	if (opt) {
340 		/* Hop-by-Hop options header */
341 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
342 		/* Destination options header(1st part) */
343 		if (opt->ip6po_rthdr) {
344 			/*
345 			 * Destination options header(1st part)
346 			 * This only makes sense with a routing header.
347 			 * See Section 9.2 of RFC 3542.
348 			 * Disabling this part just for MIP6 convenience is
349 			 * a bad idea.  We need to think carefully about a
350 			 * way to make the advanced API coexist with MIP6
351 			 * options, which might automatically be inserted in
352 			 * the kernel.
353 			 */
354 			MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
355 		}
356 		/* Routing header */
357 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
358 		/* Destination options header(2nd part) */
359 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
360 	}
361 
362 #ifdef IPSEC
363 	/*
364 	 * IPSec checking which handles several cases.
365 	 * FAST IPSEC: We re-injected the packet.
366 	 * XXX: need scope argument.
367 	 */
368 	switch(ip6_ipsec_output(&m, inp, &error))
369 	{
370 	case 1:                 /* Bad packet */
371 		goto freehdrs;
372 	case -1:                /* IPSec done */
373 		goto done;
374 	case 0:                 /* No IPSec */
375 	default:
376 		break;
377 	}
378 #endif /* IPSEC */
379 
380 	/*
381 	 * Calculate the total length of the extension header chain.
382 	 * Keep the length of the unfragmentable part for fragmentation.
383 	 */
384 	optlen = 0;
385 	if (exthdrs.ip6e_hbh)
386 		optlen += exthdrs.ip6e_hbh->m_len;
387 	if (exthdrs.ip6e_dest1)
388 		optlen += exthdrs.ip6e_dest1->m_len;
389 	if (exthdrs.ip6e_rthdr)
390 		optlen += exthdrs.ip6e_rthdr->m_len;
391 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
392 
393 	/* NOTE: we don't add AH/ESP length here (done in ip6_ipsec_output) */
394 	if (exthdrs.ip6e_dest2)
395 		optlen += exthdrs.ip6e_dest2->m_len;
396 
397 	/*
398 	 * If there is at least one extension header,
399 	 * separate IP6 header from the payload.
400 	 */
401 	if (optlen && !hdrsplit) {
402 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
403 			m = NULL;
404 			goto freehdrs;
405 		}
406 		m = exthdrs.ip6e_ip6;
407 		hdrsplit++;
408 	}
409 
410 	ip6 = mtod(m, struct ip6_hdr *);
411 
412 	/* adjust mbuf packet header length */
413 	m->m_pkthdr.len += optlen;
414 	plen = m->m_pkthdr.len - sizeof(*ip6);
415 
416 	/* If this is a jumbo payload, insert a jumbo payload option. */
417 	if (plen > IPV6_MAXPACKET) {
418 		if (!hdrsplit) {
419 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
420 				m = NULL;
421 				goto freehdrs;
422 			}
423 			m = exthdrs.ip6e_ip6;
424 			hdrsplit++;
425 		}
426 		/* adjust pointer */
427 		ip6 = mtod(m, struct ip6_hdr *);
428 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
429 			goto freehdrs;
430 		ip6->ip6_plen = 0;
431 	} else
432 		ip6->ip6_plen = htons(plen);
433 
434 	/*
435 	 * Concatenate headers and fill in next header fields.
436 	 * Here we have, on "m"
437 	 *	IPv6 payload
438 	 * and we insert headers accordingly.  Finally, we should be getting:
439 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
440 	 *
441 	 * during the header composing process, "m" points to IPv6 header.
442 	 * "mprev" points to an extension header prior to esp.
443 	 */
444 	u_char *nexthdrp = &ip6->ip6_nxt;
445 	mprev = m;
446 
447 	/*
448 	 * we treat dest2 specially.  this makes IPsec processing
449 	 * much easier.  the goal here is to make mprev point the
450 	 * mbuf prior to dest2.
451 	 *
452 	 * result: IPv6 dest2 payload
453 	 * m and mprev will point to IPv6 header.
454 	 */
455 	if (exthdrs.ip6e_dest2) {
456 		if (!hdrsplit)
457 			panic("assumption failed: hdr not split");
458 		exthdrs.ip6e_dest2->m_next = m->m_next;
459 		m->m_next = exthdrs.ip6e_dest2;
460 		*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
461 		ip6->ip6_nxt = IPPROTO_DSTOPTS;
462 	}
463 
464 	/*
465 	 * result: IPv6 hbh dest1 rthdr dest2 payload
466 	 * m will point to IPv6 header.  mprev will point to the
467 	 * extension header prior to dest2 (rthdr in the above case).
468 	 */
469 	MAKE_CHAIN(exthdrs.ip6e_hbh, mprev, nexthdrp, IPPROTO_HOPOPTS);
470 	MAKE_CHAIN(exthdrs.ip6e_dest1, mprev, nexthdrp,
471 		   IPPROTO_DSTOPTS);
472 	MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev, nexthdrp,
473 		   IPPROTO_ROUTING);
474 
475 	/*
476 	 * If there is a routing header, discard the packet.
477 	 */
478 	if (exthdrs.ip6e_rthdr) {
479 		 error = EINVAL;
480 		 goto bad;
481 	}
482 
483 	/* Source address validation */
484 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
485 	    (flags & IPV6_UNSPECSRC) == 0) {
486 		error = EOPNOTSUPP;
487 		IP6STAT_INC(ip6s_badscope);
488 		goto bad;
489 	}
490 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
491 		error = EOPNOTSUPP;
492 		IP6STAT_INC(ip6s_badscope);
493 		goto bad;
494 	}
495 
496 	IP6STAT_INC(ip6s_localout);
497 
498 	/*
499 	 * Route packet.
500 	 */
501 	if (ro == NULL) {
502 		ro = &ip6route;
503 		bzero((caddr_t)ro, sizeof(*ro));
504 	} else
505 		ro->ro_flags |= RT_LLE_CACHE;
506 	ro_pmtu = ro;
507 	if (opt && opt->ip6po_rthdr)
508 		ro = &opt->ip6po_route;
509 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
510 #ifdef FLOWTABLE
511 	if (ro->ro_rt == NULL)
512 		(void )flowtable_lookup(AF_INET6, m, (struct route *)ro);
513 #endif
514 	fibnum = (inp != NULL) ? inp->inp_inc.inc_fibnum : M_GETFIB(m);
515 again:
516 	/*
517 	 * if specified, try to fill in the traffic class field.
518 	 * do not override if a non-zero value is already set.
519 	 * we check the diffserv field and the ecn field separately.
520 	 */
521 	if (opt && opt->ip6po_tclass >= 0) {
522 		int mask = 0;
523 
524 		if ((ip6->ip6_flow & htonl(0xfc << 20)) == 0)
525 			mask |= 0xfc;
526 		if ((ip6->ip6_flow & htonl(0x03 << 20)) == 0)
527 			mask |= 0x03;
528 		if (mask != 0)
529 			ip6->ip6_flow |= htonl((opt->ip6po_tclass & mask) << 20);
530 	}
531 
532 	/* fill in or override the hop limit field, if necessary. */
533 	if (opt && opt->ip6po_hlim != -1)
534 		ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
535 	else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
536 		if (im6o != NULL)
537 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
538 		else
539 			ip6->ip6_hlim = V_ip6_defmcasthlim;
540 	}
541 	/*
542 	 * Validate route against routing table additions;
543 	 * a better/more specific route might have been added.
544 	 * Make sure address family is set in route.
545 	 */
546 	if (inp) {
547 		ro->ro_dst.sin6_family = AF_INET6;
548 		RT_VALIDATE((struct route *)ro, &inp->inp_rt_cookie, fibnum);
549 	}
550 	if (ro->ro_rt && fwd_tag == NULL && (ro->ro_rt->rt_flags & RTF_UP) &&
551 	    ro->ro_dst.sin6_family == AF_INET6 &&
552 	    IN6_ARE_ADDR_EQUAL(&ro->ro_dst.sin6_addr, &ip6->ip6_dst)) {
553 		rt = ro->ro_rt;
554 		ifp = ro->ro_rt->rt_ifp;
555 	} else {
556 		if (ro->ro_lle)
557 			LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
558 		ro->ro_lle = NULL;
559 		if (fwd_tag == NULL) {
560 			bzero(&dst_sa, sizeof(dst_sa));
561 			dst_sa.sin6_family = AF_INET6;
562 			dst_sa.sin6_len = sizeof(dst_sa);
563 			dst_sa.sin6_addr = ip6->ip6_dst;
564 		}
565 		error = in6_selectroute_fib(&dst_sa, opt, im6o, ro, &ifp,
566 		    &rt, fibnum);
567 		if (error != 0) {
568 			if (ifp != NULL)
569 				in6_ifstat_inc(ifp, ifs6_out_discard);
570 			goto bad;
571 		}
572 	}
573 	if (rt == NULL) {
574 		/*
575 		 * If in6_selectroute() does not return a route entry,
576 		 * dst may not have been updated.
577 		 */
578 		*dst = dst_sa;	/* XXX */
579 	}
580 
581 	/*
582 	 * then rt (for unicast) and ifp must be non-NULL valid values.
583 	 */
584 	if ((flags & IPV6_FORWARDING) == 0) {
585 		/* XXX: the FORWARDING flag can be set for mrouting. */
586 		in6_ifstat_inc(ifp, ifs6_out_request);
587 	}
588 	if (rt != NULL) {
589 		ia = (struct in6_ifaddr *)(rt->rt_ifa);
590 		counter_u64_add(rt->rt_pksent, 1);
591 	}
592 
593 
594 	/*
595 	 * The outgoing interface must be in the zone of source and
596 	 * destination addresses.
597 	 */
598 	origifp = ifp;
599 
600 	src0 = ip6->ip6_src;
601 	if (in6_setscope(&src0, origifp, &zone))
602 		goto badscope;
603 	bzero(&src_sa, sizeof(src_sa));
604 	src_sa.sin6_family = AF_INET6;
605 	src_sa.sin6_len = sizeof(src_sa);
606 	src_sa.sin6_addr = ip6->ip6_src;
607 	if (sa6_recoverscope(&src_sa) || zone != src_sa.sin6_scope_id)
608 		goto badscope;
609 
610 	dst0 = ip6->ip6_dst;
611 	if (in6_setscope(&dst0, origifp, &zone))
612 		goto badscope;
613 	/* re-initialize to be sure */
614 	bzero(&dst_sa, sizeof(dst_sa));
615 	dst_sa.sin6_family = AF_INET6;
616 	dst_sa.sin6_len = sizeof(dst_sa);
617 	dst_sa.sin6_addr = ip6->ip6_dst;
618 	if (sa6_recoverscope(&dst_sa) || zone != dst_sa.sin6_scope_id) {
619 		goto badscope;
620 	}
621 
622 	/* We should use ia_ifp to support the case of
623 	 * sending packets to an address of our own.
624 	 */
625 	if (ia != NULL && ia->ia_ifp)
626 		ifp = ia->ia_ifp;
627 
628 	/* scope check is done. */
629 	goto routefound;
630 
631   badscope:
632 	IP6STAT_INC(ip6s_badscope);
633 	in6_ifstat_inc(origifp, ifs6_out_discard);
634 	if (error == 0)
635 		error = EHOSTUNREACH; /* XXX */
636 	goto bad;
637 
638   routefound:
639 	if (rt && !IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
640 		if (opt && opt->ip6po_nextroute.ro_rt) {
641 			/*
642 			 * The nexthop is explicitly specified by the
643 			 * application.  We assume the next hop is an IPv6
644 			 * address.
645 			 */
646 			dst = (struct sockaddr_in6 *)opt->ip6po_nexthop;
647 		}
648 		else if ((rt->rt_flags & RTF_GATEWAY))
649 			dst = (struct sockaddr_in6 *)rt->rt_gateway;
650 	}
651 
652 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
653 		m->m_flags &= ~(M_BCAST | M_MCAST); /* just in case */
654 	} else {
655 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
656 		in6_ifstat_inc(ifp, ifs6_out_mcast);
657 		/*
658 		 * Confirm that the outgoing interface supports multicast.
659 		 */
660 		if (!(ifp->if_flags & IFF_MULTICAST)) {
661 			IP6STAT_INC(ip6s_noroute);
662 			in6_ifstat_inc(ifp, ifs6_out_discard);
663 			error = ENETUNREACH;
664 			goto bad;
665 		}
666 		if ((im6o == NULL && in6_mcast_loop) ||
667 		    (im6o && im6o->im6o_multicast_loop)) {
668 			/*
669 			 * Loop back multicast datagram if not expressly
670 			 * forbidden to do so, even if we have not joined
671 			 * the address; protocols will filter it later,
672 			 * thus deferring a hash lookup and lock acquisition
673 			 * at the expense of an m_copym().
674 			 */
675 			ip6_mloopback(ifp, m);
676 		} else {
677 			/*
678 			 * If we are acting as a multicast router, perform
679 			 * multicast forwarding as if the packet had just
680 			 * arrived on the interface to which we are about
681 			 * to send.  The multicast forwarding function
682 			 * recursively calls this function, using the
683 			 * IPV6_FORWARDING flag to prevent infinite recursion.
684 			 *
685 			 * Multicasts that are looped back by ip6_mloopback(),
686 			 * above, will be forwarded by the ip6_input() routine,
687 			 * if necessary.
688 			 */
689 			if (V_ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
690 				/*
691 				 * XXX: ip6_mforward expects that rcvif is NULL
692 				 * when it is called from the originating path.
693 				 * However, it may not always be the case.
694 				 */
695 				m->m_pkthdr.rcvif = NULL;
696 				if (ip6_mforward(ip6, ifp, m) != 0) {
697 					m_freem(m);
698 					goto done;
699 				}
700 			}
701 		}
702 		/*
703 		 * Multicasts with a hoplimit of zero may be looped back,
704 		 * above, but must not be transmitted on a network.
705 		 * Also, multicasts addressed to the loopback interface
706 		 * are not sent -- the above call to ip6_mloopback() will
707 		 * loop back a copy if this host actually belongs to the
708 		 * destination group on the loopback interface.
709 		 */
710 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK) ||
711 		    IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst)) {
712 			m_freem(m);
713 			goto done;
714 		}
715 	}
716 
717 	/*
718 	 * Fill the outgoing inteface to tell the upper layer
719 	 * to increment per-interface statistics.
720 	 */
721 	if (ifpp)
722 		*ifpp = ifp;
723 
724 	/* Determine path MTU. */
725 	if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &ip6->ip6_dst,
726 		    &mtu, &alwaysfrag, fibnum, *nexthdrp)) != 0)
727 		goto bad;
728 
729 	/*
730 	 * The caller of this function may specify to use the minimum MTU
731 	 * in some cases.
732 	 * An advanced API option (IPV6_USE_MIN_MTU) can also override MTU
733 	 * setting.  The logic is a bit complicated; by default, unicast
734 	 * packets will follow path MTU while multicast packets will be sent at
735 	 * the minimum MTU.  If IP6PO_MINMTU_ALL is specified, all packets
736 	 * including unicast ones will be sent at the minimum MTU.  Multicast
737 	 * packets will always be sent at the minimum MTU unless
738 	 * IP6PO_MINMTU_DISABLE is explicitly specified.
739 	 * See RFC 3542 for more details.
740 	 */
741 	if (mtu > IPV6_MMTU) {
742 		if ((flags & IPV6_MINMTU))
743 			mtu = IPV6_MMTU;
744 		else if (opt && opt->ip6po_minmtu == IP6PO_MINMTU_ALL)
745 			mtu = IPV6_MMTU;
746 		else if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
747 			 (opt == NULL ||
748 			  opt->ip6po_minmtu != IP6PO_MINMTU_DISABLE)) {
749 			mtu = IPV6_MMTU;
750 		}
751 	}
752 
753 	/*
754 	 * clear embedded scope identifiers if necessary.
755 	 * in6_clearscope will touch the addresses only when necessary.
756 	 */
757 	in6_clearscope(&ip6->ip6_src);
758 	in6_clearscope(&ip6->ip6_dst);
759 
760 	/*
761 	 * If the outgoing packet contains a hop-by-hop options header,
762 	 * it must be examined and processed even by the source node.
763 	 * (RFC 2460, section 4.)
764 	 */
765 	if (exthdrs.ip6e_hbh) {
766 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh, struct ip6_hbh *);
767 		u_int32_t dummy; /* XXX unused */
768 		u_int32_t plen = 0; /* XXX: ip6_process will check the value */
769 
770 #ifdef DIAGNOSTIC
771 		if ((hbh->ip6h_len + 1) << 3 > exthdrs.ip6e_hbh->m_len)
772 			panic("ip6e_hbh is not contiguous");
773 #endif
774 		/*
775 		 *  XXX: if we have to send an ICMPv6 error to the sender,
776 		 *       we need the M_LOOP flag since icmp6_error() expects
777 		 *       the IPv6 and the hop-by-hop options header are
778 		 *       contiguous unless the flag is set.
779 		 */
780 		m->m_flags |= M_LOOP;
781 		m->m_pkthdr.rcvif = ifp;
782 		if (ip6_process_hopopts(m, (u_int8_t *)(hbh + 1),
783 		    ((hbh->ip6h_len + 1) << 3) - sizeof(struct ip6_hbh),
784 		    &dummy, &plen) < 0) {
785 			/* m was already freed at this point */
786 			error = EINVAL;/* better error? */
787 			goto done;
788 		}
789 		m->m_flags &= ~M_LOOP; /* XXX */
790 		m->m_pkthdr.rcvif = NULL;
791 	}
792 
793 	/* Jump over all PFIL processing if hooks are not active. */
794 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
795 		goto passout;
796 
797 	odst = ip6->ip6_dst;
798 	/* Run through list of hooks for output packets. */
799 	error = pfil_run_hooks(&V_inet6_pfil_hook, &m, ifp, PFIL_OUT, inp);
800 	if (error != 0 || m == NULL)
801 		goto done;
802 	/* adjust pointer */
803 	ip6 = mtod(m, struct ip6_hdr *);
804 
805 	needfiblookup = 0;
806 	/* See if destination IP address was changed by packet filter. */
807 	if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) {
808 		m->m_flags |= M_SKIP_FIREWALL;
809 		/* If destination is now ourself drop to ip6_input(). */
810 		if (in6_localip(&ip6->ip6_dst)) {
811 			m->m_flags |= M_FASTFWD_OURS;
812 			if (m->m_pkthdr.rcvif == NULL)
813 				m->m_pkthdr.rcvif = V_loif;
814 			if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
815 				m->m_pkthdr.csum_flags |=
816 				    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
817 				m->m_pkthdr.csum_data = 0xffff;
818 			}
819 #ifdef SCTP
820 			if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
821 				m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
822 #endif
823 			error = netisr_queue(NETISR_IPV6, m);
824 			goto done;
825 		} else {
826 			RO_RTFREE(ro);
827 			needfiblookup = 1; /* Redo the routing table lookup. */
828 			if (ro->ro_lle)
829 				LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
830 			ro->ro_lle = NULL;
831 		}
832 	}
833 	/* See if fib was changed by packet filter. */
834 	if (fibnum != M_GETFIB(m)) {
835 		m->m_flags |= M_SKIP_FIREWALL;
836 		fibnum = M_GETFIB(m);
837 		RO_RTFREE(ro);
838 		needfiblookup = 1;
839 		if (ro->ro_lle)
840 			LLE_FREE(ro->ro_lle);	/* zeros ro_lle */
841 		ro->ro_lle = NULL;
842 	}
843 	if (needfiblookup)
844 		goto again;
845 
846 	/* See if local, if yes, send it to netisr. */
847 	if (m->m_flags & M_FASTFWD_OURS) {
848 		if (m->m_pkthdr.rcvif == NULL)
849 			m->m_pkthdr.rcvif = V_loif;
850 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
851 			m->m_pkthdr.csum_flags |=
852 			    CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR;
853 			m->m_pkthdr.csum_data = 0xffff;
854 		}
855 #ifdef SCTP
856 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6)
857 			m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID;
858 #endif
859 		error = netisr_queue(NETISR_IPV6, m);
860 		goto done;
861 	}
862 	/* Or forward to some other address? */
863 	if ((m->m_flags & M_IP6_NEXTHOP) &&
864 	    (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) {
865 		dst = (struct sockaddr_in6 *)&ro->ro_dst;
866 		bcopy((fwd_tag+1), &dst_sa, sizeof(struct sockaddr_in6));
867 		m->m_flags |= M_SKIP_FIREWALL;
868 		m->m_flags &= ~M_IP6_NEXTHOP;
869 		m_tag_delete(m, fwd_tag);
870 		goto again;
871 	}
872 
873 passout:
874 	/*
875 	 * Send the packet to the outgoing interface.
876 	 * If necessary, do IPv6 fragmentation before sending.
877 	 *
878 	 * the logic here is rather complex:
879 	 * 1: normal case (dontfrag == 0, alwaysfrag == 0)
880 	 * 1-a:	send as is if tlen <= path mtu
881 	 * 1-b:	fragment if tlen > path mtu
882 	 *
883 	 * 2: if user asks us not to fragment (dontfrag == 1)
884 	 * 2-a:	send as is if tlen <= interface mtu
885 	 * 2-b:	error if tlen > interface mtu
886 	 *
887 	 * 3: if we always need to attach fragment header (alwaysfrag == 1)
888 	 *	always fragment
889 	 *
890 	 * 4: if dontfrag == 1 && alwaysfrag == 1
891 	 *	error, as we cannot handle this conflicting request
892 	 */
893 	sw_csum = m->m_pkthdr.csum_flags;
894 	if (!hdrsplit) {
895 		tso = ((sw_csum & ifp->if_hwassist & CSUM_TSO) != 0) ? 1 : 0;
896 		sw_csum &= ~ifp->if_hwassist;
897 	} else
898 		tso = 0;
899 	/*
900 	 * If we added extension headers, we will not do TSO and calculate the
901 	 * checksums ourselves for now.
902 	 * XXX-BZ  Need a framework to know when the NIC can handle it, even
903 	 * with ext. hdrs.
904 	 */
905 	if (sw_csum & CSUM_DELAY_DATA_IPV6) {
906 		sw_csum &= ~CSUM_DELAY_DATA_IPV6;
907 		in6_delayed_cksum(m, plen, sizeof(struct ip6_hdr));
908 	}
909 #ifdef SCTP
910 	if (sw_csum & CSUM_SCTP_IPV6) {
911 		sw_csum &= ~CSUM_SCTP_IPV6;
912 		sctp_delayed_cksum(m, sizeof(struct ip6_hdr));
913 	}
914 #endif
915 	m->m_pkthdr.csum_flags &= ifp->if_hwassist;
916 	tlen = m->m_pkthdr.len;
917 
918 	if ((opt && (opt->ip6po_flags & IP6PO_DONTFRAG)) || tso)
919 		dontfrag = 1;
920 	else
921 		dontfrag = 0;
922 	if (dontfrag && alwaysfrag) {	/* case 4 */
923 		/* conflicting request - can't transmit */
924 		error = EMSGSIZE;
925 		goto bad;
926 	}
927 	if (dontfrag && tlen > IN6_LINKMTU(ifp) && !tso) {	/* case 2-b */
928 		/*
929 		 * Even if the DONTFRAG option is specified, we cannot send the
930 		 * packet when the data length is larger than the MTU of the
931 		 * outgoing interface.
932 		 * Notify the error by sending IPV6_PATHMTU ancillary data if
933 		 * application wanted to know the MTU value. Also return an
934 		 * error code (this is not described in the API spec).
935 		 */
936 		if (inp != NULL)
937 			ip6_notify_pmtu(inp, &dst_sa, (u_int32_t)mtu);
938 		error = EMSGSIZE;
939 		goto bad;
940 	}
941 
942 	/*
943 	 * transmit packet without fragmentation
944 	 */
945 	if (dontfrag || (!alwaysfrag && tlen <= mtu)) {	/* case 1-a and 2-a */
946 		struct in6_ifaddr *ia6;
947 
948 		ip6 = mtod(m, struct ip6_hdr *);
949 		ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
950 		if (ia6) {
951 			/* Record statistics for this interface address. */
952 			counter_u64_add(ia6->ia_ifa.ifa_opackets, 1);
953 			counter_u64_add(ia6->ia_ifa.ifa_obytes,
954 			    m->m_pkthdr.len);
955 			ifa_free(&ia6->ia_ifa);
956 		}
957 		error = nd6_output_ifp(ifp, origifp, m, dst,
958 		    (struct route *)ro);
959 		goto done;
960 	}
961 
962 	/*
963 	 * try to fragment the packet.  case 1-b and 3
964 	 */
965 	if (mtu < IPV6_MMTU) {
966 		/* path MTU cannot be less than IPV6_MMTU */
967 		error = EMSGSIZE;
968 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
969 		goto bad;
970 	} else if (ip6->ip6_plen == 0) {
971 		/* jumbo payload cannot be fragmented */
972 		error = EMSGSIZE;
973 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
974 		goto bad;
975 	} else {
976 		u_char nextproto;
977 
978 		/*
979 		 * Too large for the destination or interface;
980 		 * fragment if possible.
981 		 * Must be able to put at least 8 bytes per fragment.
982 		 */
983 		hlen = unfragpartlen;
984 		if (mtu > IPV6_MAXPACKET)
985 			mtu = IPV6_MAXPACKET;
986 
987 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
988 		if (len < 8) {
989 			error = EMSGSIZE;
990 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
991 			goto bad;
992 		}
993 
994 		/*
995 		 * If the interface will not calculate checksums on
996 		 * fragmented packets, then do it here.
997 		 * XXX-BZ handle the hw offloading case.  Need flags.
998 		 */
999 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
1000 			in6_delayed_cksum(m, plen, hlen);
1001 			m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6;
1002 		}
1003 #ifdef SCTP
1004 		if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) {
1005 			sctp_delayed_cksum(m, hlen);
1006 			m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6;
1007 		}
1008 #endif
1009 		/*
1010 		 * Change the next header field of the last header in the
1011 		 * unfragmentable part.
1012 		 */
1013 		if (exthdrs.ip6e_rthdr) {
1014 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
1015 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
1016 		} else if (exthdrs.ip6e_dest1) {
1017 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
1018 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
1019 		} else if (exthdrs.ip6e_hbh) {
1020 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
1021 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
1022 		} else {
1023 			nextproto = ip6->ip6_nxt;
1024 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
1025 		}
1026 
1027 		/*
1028 		 * Loop through length of segment after first fragment,
1029 		 * make new header and copy data of each part and link onto
1030 		 * chain.
1031 		 */
1032 		m0 = m;
1033 		id = htonl(ip6_randomid());
1034 		if ((error = ip6_fragment(ifp, m, hlen, nextproto, len, id)))
1035 			goto sendorfree;
1036 
1037 		in6_ifstat_inc(ifp, ifs6_out_fragok);
1038 	}
1039 
1040 	/*
1041 	 * Remove leading garbages.
1042 	 */
1043 sendorfree:
1044 	m = m0->m_nextpkt;
1045 	m0->m_nextpkt = 0;
1046 	m_freem(m0);
1047 	for (m0 = m; m; m = m0) {
1048 		m0 = m->m_nextpkt;
1049 		m->m_nextpkt = 0;
1050 		if (error == 0) {
1051 			/* Record statistics for this interface address. */
1052 			if (ia) {
1053 				counter_u64_add(ia->ia_ifa.ifa_opackets, 1);
1054 				counter_u64_add(ia->ia_ifa.ifa_obytes,
1055 				    m->m_pkthdr.len);
1056 			}
1057 			error = nd6_output_ifp(ifp, origifp, m, dst,
1058 			    (struct route *)ro);
1059 		} else
1060 			m_freem(m);
1061 	}
1062 
1063 	if (error == 0)
1064 		IP6STAT_INC(ip6s_fragmented);
1065 
1066 done:
1067 	if (ro == &ip6route)
1068 		RO_RTFREE(ro);
1069 	return (error);
1070 
1071 freehdrs:
1072 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1073 	m_freem(exthdrs.ip6e_dest1);
1074 	m_freem(exthdrs.ip6e_rthdr);
1075 	m_freem(exthdrs.ip6e_dest2);
1076 	/* FALLTHROUGH */
1077 bad:
1078 	if (m)
1079 		m_freem(m);
1080 	goto done;
1081 }
1082 
1083 static int
1084 ip6_copyexthdr(struct mbuf **mp, caddr_t hdr, int hlen)
1085 {
1086 	struct mbuf *m;
1087 
1088 	if (hlen > MCLBYTES)
1089 		return (ENOBUFS); /* XXX */
1090 
1091 	if (hlen > MLEN)
1092 		m = m_getcl(M_NOWAIT, MT_DATA, 0);
1093 	else
1094 		m = m_get(M_NOWAIT, MT_DATA);
1095 	if (m == NULL)
1096 		return (ENOBUFS);
1097 	m->m_len = hlen;
1098 	if (hdr)
1099 		bcopy(hdr, mtod(m, caddr_t), hlen);
1100 
1101 	*mp = m;
1102 	return (0);
1103 }
1104 
1105 /*
1106  * Insert jumbo payload option.
1107  */
1108 static int
1109 ip6_insert_jumboopt(struct ip6_exthdrs *exthdrs, u_int32_t plen)
1110 {
1111 	struct mbuf *mopt;
1112 	u_char *optbuf;
1113 	u_int32_t v;
1114 
1115 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1116 
1117 	/*
1118 	 * If there is no hop-by-hop options header, allocate new one.
1119 	 * If there is one but it doesn't have enough space to store the
1120 	 * jumbo payload option, allocate a cluster to store the whole options.
1121 	 * Otherwise, use it to store the options.
1122 	 */
1123 	if (exthdrs->ip6e_hbh == NULL) {
1124 		mopt = m_get(M_NOWAIT, MT_DATA);
1125 		if (mopt == NULL)
1126 			return (ENOBUFS);
1127 		mopt->m_len = JUMBOOPTLEN;
1128 		optbuf = mtod(mopt, u_char *);
1129 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1130 		exthdrs->ip6e_hbh = mopt;
1131 	} else {
1132 		struct ip6_hbh *hbh;
1133 
1134 		mopt = exthdrs->ip6e_hbh;
1135 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1136 			/*
1137 			 * XXX assumption:
1138 			 * - exthdrs->ip6e_hbh is not referenced from places
1139 			 *   other than exthdrs.
1140 			 * - exthdrs->ip6e_hbh is not an mbuf chain.
1141 			 */
1142 			int oldoptlen = mopt->m_len;
1143 			struct mbuf *n;
1144 
1145 			/*
1146 			 * XXX: give up if the whole (new) hbh header does
1147 			 * not fit even in an mbuf cluster.
1148 			 */
1149 			if (oldoptlen + JUMBOOPTLEN > MCLBYTES)
1150 				return (ENOBUFS);
1151 
1152 			/*
1153 			 * As a consequence, we must always prepare a cluster
1154 			 * at this point.
1155 			 */
1156 			n = m_getcl(M_NOWAIT, MT_DATA, 0);
1157 			if (n == NULL)
1158 				return (ENOBUFS);
1159 			n->m_len = oldoptlen + JUMBOOPTLEN;
1160 			bcopy(mtod(mopt, caddr_t), mtod(n, caddr_t),
1161 			    oldoptlen);
1162 			optbuf = mtod(n, caddr_t) + oldoptlen;
1163 			m_freem(mopt);
1164 			mopt = exthdrs->ip6e_hbh = n;
1165 		} else {
1166 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1167 			mopt->m_len += JUMBOOPTLEN;
1168 		}
1169 		optbuf[0] = IP6OPT_PADN;
1170 		optbuf[1] = 1;
1171 
1172 		/*
1173 		 * Adjust the header length according to the pad and
1174 		 * the jumbo payload option.
1175 		 */
1176 		hbh = mtod(mopt, struct ip6_hbh *);
1177 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1178 	}
1179 
1180 	/* fill in the option. */
1181 	optbuf[2] = IP6OPT_JUMBO;
1182 	optbuf[3] = 4;
1183 	v = (u_int32_t)htonl(plen + JUMBOOPTLEN);
1184 	bcopy(&v, &optbuf[4], sizeof(u_int32_t));
1185 
1186 	/* finally, adjust the packet header length */
1187 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1188 
1189 	return (0);
1190 #undef JUMBOOPTLEN
1191 }
1192 
1193 /*
1194  * Insert fragment header and copy unfragmentable header portions.
1195  */
1196 static int
1197 ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
1198     struct ip6_frag **frghdrp)
1199 {
1200 	struct mbuf *n, *mlast;
1201 
1202 	if (hlen > sizeof(struct ip6_hdr)) {
1203 		n = m_copym(m0, sizeof(struct ip6_hdr),
1204 		    hlen - sizeof(struct ip6_hdr), M_NOWAIT);
1205 		if (n == NULL)
1206 			return (ENOBUFS);
1207 		m->m_next = n;
1208 	} else
1209 		n = m;
1210 
1211 	/* Search for the last mbuf of unfragmentable part. */
1212 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1213 		;
1214 
1215 	if (M_WRITABLE(mlast) &&
1216 	    M_TRAILINGSPACE(mlast) >= sizeof(struct ip6_frag)) {
1217 		/* use the trailing space of the last mbuf for the fragment hdr */
1218 		*frghdrp = (struct ip6_frag *)(mtod(mlast, caddr_t) +
1219 		    mlast->m_len);
1220 		mlast->m_len += sizeof(struct ip6_frag);
1221 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1222 	} else {
1223 		/* allocate a new mbuf for the fragment header */
1224 		struct mbuf *mfrg;
1225 
1226 		mfrg = m_get(M_NOWAIT, MT_DATA);
1227 		if (mfrg == NULL)
1228 			return (ENOBUFS);
1229 		mfrg->m_len = sizeof(struct ip6_frag);
1230 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1231 		mlast->m_next = mfrg;
1232 	}
1233 
1234 	return (0);
1235 }
1236 
1237 /*
1238  * Calculates IPv6 path mtu for destination @dst.
1239  * Resulting MTU is stored in @mtup.
1240  *
1241  * Returns 0 on success.
1242  */
1243 static int
1244 ip6_getpmtu_ctl(u_int fibnum, const struct in6_addr *dst, u_long *mtup)
1245 {
1246 	struct nhop6_extended nh6;
1247 	struct in6_addr kdst;
1248 	uint32_t scopeid;
1249 	struct ifnet *ifp;
1250 	u_long mtu;
1251 	int error;
1252 
1253 	in6_splitscope(dst, &kdst, &scopeid);
1254 	if (fib6_lookup_nh_ext(fibnum, &kdst, scopeid, NHR_REF, 0, &nh6) != 0)
1255 		return (EHOSTUNREACH);
1256 
1257 	ifp = nh6.nh_ifp;
1258 	mtu = nh6.nh_mtu;
1259 
1260 	error = ip6_calcmtu(ifp, dst, mtu, mtup, NULL, 0);
1261 	fib6_free_nh_ext(fibnum, &nh6);
1262 
1263 	return (error);
1264 }
1265 
1266 /*
1267  * Calculates IPv6 path MTU for @dst based on transmit @ifp,
1268  * and cached data in @ro_pmtu.
1269  * MTU from (successful) route lookup is saved (along with dst)
1270  * inside @ro_pmtu to avoid subsequent route lookups after packet
1271  * filter processing.
1272  *
1273  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1274  * Returns 0 on success.
1275  */
1276 static int
1277 ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
1278     struct ifnet *ifp, const struct in6_addr *dst, u_long *mtup,
1279     int *alwaysfragp, u_int fibnum, u_int proto)
1280 {
1281 	struct nhop6_basic nh6;
1282 	struct in6_addr kdst;
1283 	uint32_t scopeid;
1284 	struct sockaddr_in6 *sa6_dst;
1285 	u_long mtu;
1286 
1287 	mtu = 0;
1288 	if (do_lookup) {
1289 
1290 		/*
1291 		 * Here ro_pmtu has final destination address, while
1292 		 * ro might represent immediate destination.
1293 		 * Use ro_pmtu destination since mtu might differ.
1294 		 */
1295 		sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
1296 		if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
1297 			ro_pmtu->ro_mtu = 0;
1298 
1299 		if (ro_pmtu->ro_mtu == 0) {
1300 			bzero(sa6_dst, sizeof(*sa6_dst));
1301 			sa6_dst->sin6_family = AF_INET6;
1302 			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
1303 			sa6_dst->sin6_addr = *dst;
1304 
1305 			in6_splitscope(dst, &kdst, &scopeid);
1306 			if (fib6_lookup_nh_basic(fibnum, &kdst, scopeid, 0, 0,
1307 			    &nh6) == 0)
1308 				ro_pmtu->ro_mtu = nh6.nh_mtu;
1309 		}
1310 
1311 		mtu = ro_pmtu->ro_mtu;
1312 	}
1313 
1314 	if (ro_pmtu->ro_rt)
1315 		mtu = ro_pmtu->ro_rt->rt_mtu;
1316 
1317 	return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp, proto));
1318 }
1319 
1320 /*
1321  * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
1322  * hostcache data for @dst.
1323  * Stores mtu and always-frag value into @mtup and @alwaysfragp.
1324  *
1325  * Returns 0 on success.
1326  */
1327 static int
1328 ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
1329     u_long *mtup, int *alwaysfragp, u_int proto)
1330 {
1331 	u_long mtu = 0;
1332 	int alwaysfrag = 0;
1333 	int error = 0;
1334 
1335 	if (rt_mtu > 0) {
1336 		u_int32_t ifmtu;
1337 		struct in_conninfo inc;
1338 
1339 		bzero(&inc, sizeof(inc));
1340 		inc.inc_flags |= INC_ISIPV6;
1341 		inc.inc6_faddr = *dst;
1342 
1343 		ifmtu = IN6_LINKMTU(ifp);
1344 
1345 		/* TCP is known to react to pmtu changes so skip hc */
1346 		if (proto != IPPROTO_TCP)
1347 			mtu = tcp_hc_getmtu(&inc);
1348 
1349 		if (mtu)
1350 			mtu = min(mtu, rt_mtu);
1351 		else
1352 			mtu = rt_mtu;
1353 		if (mtu == 0)
1354 			mtu = ifmtu;
1355 		else if (mtu < IPV6_MMTU) {
1356 			/*
1357 			 * RFC2460 section 5, last paragraph:
1358 			 * if we record ICMPv6 too big message with
1359 			 * mtu < IPV6_MMTU, transmit packets sized IPV6_MMTU
1360 			 * or smaller, with framgent header attached.
1361 			 * (fragment header is needed regardless from the
1362 			 * packet size, for translators to identify packets)
1363 			 */
1364 			alwaysfrag = 1;
1365 			mtu = IPV6_MMTU;
1366 		}
1367 	} else if (ifp) {
1368 		mtu = IN6_LINKMTU(ifp);
1369 	} else
1370 		error = EHOSTUNREACH; /* XXX */
1371 
1372 	*mtup = mtu;
1373 	if (alwaysfragp)
1374 		*alwaysfragp = alwaysfrag;
1375 	return (error);
1376 }
1377 
1378 /*
1379  * IP6 socket option processing.
1380  */
1381 int
1382 ip6_ctloutput(struct socket *so, struct sockopt *sopt)
1383 {
1384 	int optdatalen, uproto;
1385 	void *optdata;
1386 	struct inpcb *in6p = sotoinpcb(so);
1387 	int error, optval;
1388 	int level, op, optname;
1389 	int optlen;
1390 	struct thread *td;
1391 #ifdef	RSS
1392 	uint32_t rss_bucket;
1393 	int retval;
1394 #endif
1395 
1396 /*
1397  * Don't use more than a quarter of mbuf clusters.  N.B.:
1398  * nmbclusters is an int, but nmbclusters * MCLBYTES may overflow
1399  * on LP64 architectures, so cast to u_long to avoid undefined
1400  * behavior.  ILP32 architectures cannot have nmbclusters
1401  * large enough to overflow for other reasons.
1402  */
1403 #define IPV6_PKTOPTIONS_MBUF_LIMIT	((u_long)nmbclusters * MCLBYTES / 4)
1404 
1405 	level = sopt->sopt_level;
1406 	op = sopt->sopt_dir;
1407 	optname = sopt->sopt_name;
1408 	optlen = sopt->sopt_valsize;
1409 	td = sopt->sopt_td;
1410 	error = 0;
1411 	optval = 0;
1412 	uproto = (int)so->so_proto->pr_protocol;
1413 
1414 	if (level != IPPROTO_IPV6) {
1415 		error = EINVAL;
1416 
1417 		if (sopt->sopt_level == SOL_SOCKET &&
1418 		    sopt->sopt_dir == SOPT_SET) {
1419 			switch (sopt->sopt_name) {
1420 			case SO_REUSEADDR:
1421 				INP_WLOCK(in6p);
1422 				if ((so->so_options & SO_REUSEADDR) != 0)
1423 					in6p->inp_flags2 |= INP_REUSEADDR;
1424 				else
1425 					in6p->inp_flags2 &= ~INP_REUSEADDR;
1426 				INP_WUNLOCK(in6p);
1427 				error = 0;
1428 				break;
1429 			case SO_REUSEPORT:
1430 				INP_WLOCK(in6p);
1431 				if ((so->so_options & SO_REUSEPORT) != 0)
1432 					in6p->inp_flags2 |= INP_REUSEPORT;
1433 				else
1434 					in6p->inp_flags2 &= ~INP_REUSEPORT;
1435 				INP_WUNLOCK(in6p);
1436 				error = 0;
1437 				break;
1438 			case SO_SETFIB:
1439 				INP_WLOCK(in6p);
1440 				in6p->inp_inc.inc_fibnum = so->so_fibnum;
1441 				INP_WUNLOCK(in6p);
1442 				error = 0;
1443 				break;
1444 			default:
1445 				break;
1446 			}
1447 		}
1448 	} else {		/* level == IPPROTO_IPV6 */
1449 		switch (op) {
1450 
1451 		case SOPT_SET:
1452 			switch (optname) {
1453 			case IPV6_2292PKTOPTIONS:
1454 #ifdef IPV6_PKTOPTIONS
1455 			case IPV6_PKTOPTIONS:
1456 #endif
1457 			{
1458 				struct mbuf *m;
1459 
1460 				if (optlen > IPV6_PKTOPTIONS_MBUF_LIMIT) {
1461 					printf("ip6_ctloutput: mbuf limit hit\n");
1462 					error = ENOBUFS;
1463 					break;
1464 				}
1465 
1466 				error = soopt_getm(sopt, &m); /* XXX */
1467 				if (error != 0)
1468 					break;
1469 				error = soopt_mcopyin(sopt, m); /* XXX */
1470 				if (error != 0)
1471 					break;
1472 				error = ip6_pcbopts(&in6p->in6p_outputopts,
1473 						    m, so, sopt);
1474 				m_freem(m); /* XXX */
1475 				break;
1476 			}
1477 
1478 			/*
1479 			 * Use of some Hop-by-Hop options or some
1480 			 * Destination options, might require special
1481 			 * privilege.  That is, normal applications
1482 			 * (without special privilege) might be forbidden
1483 			 * from setting certain options in outgoing packets,
1484 			 * and might never see certain options in received
1485 			 * packets. [RFC 2292 Section 6]
1486 			 * KAME specific note:
1487 			 *  KAME prevents non-privileged users from sending or
1488 			 *  receiving ANY hbh/dst options in order to avoid
1489 			 *  overhead of parsing options in the kernel.
1490 			 */
1491 			case IPV6_RECVHOPOPTS:
1492 			case IPV6_RECVDSTOPTS:
1493 			case IPV6_RECVRTHDRDSTOPTS:
1494 				if (td != NULL) {
1495 					error = priv_check(td,
1496 					    PRIV_NETINET_SETHDROPTS);
1497 					if (error)
1498 						break;
1499 				}
1500 				/* FALLTHROUGH */
1501 			case IPV6_UNICAST_HOPS:
1502 			case IPV6_HOPLIMIT:
1503 
1504 			case IPV6_RECVPKTINFO:
1505 			case IPV6_RECVHOPLIMIT:
1506 			case IPV6_RECVRTHDR:
1507 			case IPV6_RECVPATHMTU:
1508 			case IPV6_RECVTCLASS:
1509 			case IPV6_RECVFLOWID:
1510 #ifdef	RSS
1511 			case IPV6_RECVRSSBUCKETID:
1512 #endif
1513 			case IPV6_V6ONLY:
1514 			case IPV6_AUTOFLOWLABEL:
1515 			case IPV6_BINDANY:
1516 			case IPV6_BINDMULTI:
1517 #ifdef	RSS
1518 			case IPV6_RSS_LISTEN_BUCKET:
1519 #endif
1520 				if (optname == IPV6_BINDANY && td != NULL) {
1521 					error = priv_check(td,
1522 					    PRIV_NETINET_BINDANY);
1523 					if (error)
1524 						break;
1525 				}
1526 
1527 				if (optlen != sizeof(int)) {
1528 					error = EINVAL;
1529 					break;
1530 				}
1531 				error = sooptcopyin(sopt, &optval,
1532 					sizeof optval, sizeof optval);
1533 				if (error)
1534 					break;
1535 				switch (optname) {
1536 
1537 				case IPV6_UNICAST_HOPS:
1538 					if (optval < -1 || optval >= 256)
1539 						error = EINVAL;
1540 					else {
1541 						/* -1 = kernel default */
1542 						in6p->in6p_hops = optval;
1543 						if ((in6p->inp_vflag &
1544 						     INP_IPV4) != 0)
1545 							in6p->inp_ip_ttl = optval;
1546 					}
1547 					break;
1548 #define OPTSET(bit) \
1549 do { \
1550 	INP_WLOCK(in6p); \
1551 	if (optval) \
1552 		in6p->inp_flags |= (bit); \
1553 	else \
1554 		in6p->inp_flags &= ~(bit); \
1555 	INP_WUNLOCK(in6p); \
1556 } while (/*CONSTCOND*/ 0)
1557 #define OPTSET2292(bit) \
1558 do { \
1559 	INP_WLOCK(in6p); \
1560 	in6p->inp_flags |= IN6P_RFC2292; \
1561 	if (optval) \
1562 		in6p->inp_flags |= (bit); \
1563 	else \
1564 		in6p->inp_flags &= ~(bit); \
1565 	INP_WUNLOCK(in6p); \
1566 } while (/*CONSTCOND*/ 0)
1567 #define OPTBIT(bit) (in6p->inp_flags & (bit) ? 1 : 0)
1568 
1569 #define OPTSET2(bit, val) do {						\
1570 	INP_WLOCK(in6p);						\
1571 	if (val)							\
1572 		in6p->inp_flags2 |= bit;				\
1573 	else								\
1574 		in6p->inp_flags2 &= ~bit;				\
1575 	INP_WUNLOCK(in6p);						\
1576 } while (0)
1577 #define OPTBIT2(bit) (in6p->inp_flags2 & (bit) ? 1 : 0)
1578 
1579 				case IPV6_RECVPKTINFO:
1580 					/* cannot mix with RFC2292 */
1581 					if (OPTBIT(IN6P_RFC2292)) {
1582 						error = EINVAL;
1583 						break;
1584 					}
1585 					OPTSET(IN6P_PKTINFO);
1586 					break;
1587 
1588 				case IPV6_HOPLIMIT:
1589 				{
1590 					struct ip6_pktopts **optp;
1591 
1592 					/* cannot mix with RFC2292 */
1593 					if (OPTBIT(IN6P_RFC2292)) {
1594 						error = EINVAL;
1595 						break;
1596 					}
1597 					optp = &in6p->in6p_outputopts;
1598 					error = ip6_pcbopt(IPV6_HOPLIMIT,
1599 					    (u_char *)&optval, sizeof(optval),
1600 					    optp, (td != NULL) ? td->td_ucred :
1601 					    NULL, uproto);
1602 					break;
1603 				}
1604 
1605 				case IPV6_RECVHOPLIMIT:
1606 					/* cannot mix with RFC2292 */
1607 					if (OPTBIT(IN6P_RFC2292)) {
1608 						error = EINVAL;
1609 						break;
1610 					}
1611 					OPTSET(IN6P_HOPLIMIT);
1612 					break;
1613 
1614 				case IPV6_RECVHOPOPTS:
1615 					/* cannot mix with RFC2292 */
1616 					if (OPTBIT(IN6P_RFC2292)) {
1617 						error = EINVAL;
1618 						break;
1619 					}
1620 					OPTSET(IN6P_HOPOPTS);
1621 					break;
1622 
1623 				case IPV6_RECVDSTOPTS:
1624 					/* cannot mix with RFC2292 */
1625 					if (OPTBIT(IN6P_RFC2292)) {
1626 						error = EINVAL;
1627 						break;
1628 					}
1629 					OPTSET(IN6P_DSTOPTS);
1630 					break;
1631 
1632 				case IPV6_RECVRTHDRDSTOPTS:
1633 					/* cannot mix with RFC2292 */
1634 					if (OPTBIT(IN6P_RFC2292)) {
1635 						error = EINVAL;
1636 						break;
1637 					}
1638 					OPTSET(IN6P_RTHDRDSTOPTS);
1639 					break;
1640 
1641 				case IPV6_RECVRTHDR:
1642 					/* cannot mix with RFC2292 */
1643 					if (OPTBIT(IN6P_RFC2292)) {
1644 						error = EINVAL;
1645 						break;
1646 					}
1647 					OPTSET(IN6P_RTHDR);
1648 					break;
1649 
1650 				case IPV6_RECVPATHMTU:
1651 					/*
1652 					 * We ignore this option for TCP
1653 					 * sockets.
1654 					 * (RFC3542 leaves this case
1655 					 * unspecified.)
1656 					 */
1657 					if (uproto != IPPROTO_TCP)
1658 						OPTSET(IN6P_MTU);
1659 					break;
1660 
1661 				case IPV6_RECVFLOWID:
1662 					OPTSET2(INP_RECVFLOWID, optval);
1663 					break;
1664 
1665 #ifdef	RSS
1666 				case IPV6_RECVRSSBUCKETID:
1667 					OPTSET2(INP_RECVRSSBUCKETID, optval);
1668 					break;
1669 #endif
1670 
1671 				case IPV6_V6ONLY:
1672 					/*
1673 					 * make setsockopt(IPV6_V6ONLY)
1674 					 * available only prior to bind(2).
1675 					 * see ipng mailing list, Jun 22 2001.
1676 					 */
1677 					if (in6p->inp_lport ||
1678 					    !IN6_IS_ADDR_UNSPECIFIED(&in6p->in6p_laddr)) {
1679 						error = EINVAL;
1680 						break;
1681 					}
1682 					OPTSET(IN6P_IPV6_V6ONLY);
1683 					if (optval)
1684 						in6p->inp_vflag &= ~INP_IPV4;
1685 					else
1686 						in6p->inp_vflag |= INP_IPV4;
1687 					break;
1688 				case IPV6_RECVTCLASS:
1689 					/* cannot mix with RFC2292 XXX */
1690 					if (OPTBIT(IN6P_RFC2292)) {
1691 						error = EINVAL;
1692 						break;
1693 					}
1694 					OPTSET(IN6P_TCLASS);
1695 					break;
1696 				case IPV6_AUTOFLOWLABEL:
1697 					OPTSET(IN6P_AUTOFLOWLABEL);
1698 					break;
1699 
1700 				case IPV6_BINDANY:
1701 					OPTSET(INP_BINDANY);
1702 					break;
1703 
1704 				case IPV6_BINDMULTI:
1705 					OPTSET2(INP_BINDMULTI, optval);
1706 					break;
1707 #ifdef	RSS
1708 				case IPV6_RSS_LISTEN_BUCKET:
1709 					if ((optval >= 0) &&
1710 					    (optval < rss_getnumbuckets())) {
1711 						in6p->inp_rss_listen_bucket = optval;
1712 						OPTSET2(INP_RSS_BUCKET_SET, 1);
1713 					} else {
1714 						error = EINVAL;
1715 					}
1716 					break;
1717 #endif
1718 				}
1719 				break;
1720 
1721 			case IPV6_TCLASS:
1722 			case IPV6_DONTFRAG:
1723 			case IPV6_USE_MIN_MTU:
1724 			case IPV6_PREFER_TEMPADDR:
1725 				if (optlen != sizeof(optval)) {
1726 					error = EINVAL;
1727 					break;
1728 				}
1729 				error = sooptcopyin(sopt, &optval,
1730 					sizeof optval, sizeof optval);
1731 				if (error)
1732 					break;
1733 				{
1734 					struct ip6_pktopts **optp;
1735 					optp = &in6p->in6p_outputopts;
1736 					error = ip6_pcbopt(optname,
1737 					    (u_char *)&optval, sizeof(optval),
1738 					    optp, (td != NULL) ? td->td_ucred :
1739 					    NULL, uproto);
1740 					break;
1741 				}
1742 
1743 			case IPV6_2292PKTINFO:
1744 			case IPV6_2292HOPLIMIT:
1745 			case IPV6_2292HOPOPTS:
1746 			case IPV6_2292DSTOPTS:
1747 			case IPV6_2292RTHDR:
1748 				/* RFC 2292 */
1749 				if (optlen != sizeof(int)) {
1750 					error = EINVAL;
1751 					break;
1752 				}
1753 				error = sooptcopyin(sopt, &optval,
1754 					sizeof optval, sizeof optval);
1755 				if (error)
1756 					break;
1757 				switch (optname) {
1758 				case IPV6_2292PKTINFO:
1759 					OPTSET2292(IN6P_PKTINFO);
1760 					break;
1761 				case IPV6_2292HOPLIMIT:
1762 					OPTSET2292(IN6P_HOPLIMIT);
1763 					break;
1764 				case IPV6_2292HOPOPTS:
1765 					/*
1766 					 * Check super-user privilege.
1767 					 * See comments for IPV6_RECVHOPOPTS.
1768 					 */
1769 					if (td != NULL) {
1770 						error = priv_check(td,
1771 						    PRIV_NETINET_SETHDROPTS);
1772 						if (error)
1773 							return (error);
1774 					}
1775 					OPTSET2292(IN6P_HOPOPTS);
1776 					break;
1777 				case IPV6_2292DSTOPTS:
1778 					if (td != NULL) {
1779 						error = priv_check(td,
1780 						    PRIV_NETINET_SETHDROPTS);
1781 						if (error)
1782 							return (error);
1783 					}
1784 					OPTSET2292(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS); /* XXX */
1785 					break;
1786 				case IPV6_2292RTHDR:
1787 					OPTSET2292(IN6P_RTHDR);
1788 					break;
1789 				}
1790 				break;
1791 			case IPV6_PKTINFO:
1792 			case IPV6_HOPOPTS:
1793 			case IPV6_RTHDR:
1794 			case IPV6_DSTOPTS:
1795 			case IPV6_RTHDRDSTOPTS:
1796 			case IPV6_NEXTHOP:
1797 			{
1798 				/* new advanced API (RFC3542) */
1799 				u_char *optbuf;
1800 				u_char optbuf_storage[MCLBYTES];
1801 				int optlen;
1802 				struct ip6_pktopts **optp;
1803 
1804 				/* cannot mix with RFC2292 */
1805 				if (OPTBIT(IN6P_RFC2292)) {
1806 					error = EINVAL;
1807 					break;
1808 				}
1809 
1810 				/*
1811 				 * We only ensure valsize is not too large
1812 				 * here.  Further validation will be done
1813 				 * later.
1814 				 */
1815 				error = sooptcopyin(sopt, optbuf_storage,
1816 				    sizeof(optbuf_storage), 0);
1817 				if (error)
1818 					break;
1819 				optlen = sopt->sopt_valsize;
1820 				optbuf = optbuf_storage;
1821 				optp = &in6p->in6p_outputopts;
1822 				error = ip6_pcbopt(optname, optbuf, optlen,
1823 				    optp, (td != NULL) ? td->td_ucred : NULL,
1824 				    uproto);
1825 				break;
1826 			}
1827 #undef OPTSET
1828 
1829 			case IPV6_MULTICAST_IF:
1830 			case IPV6_MULTICAST_HOPS:
1831 			case IPV6_MULTICAST_LOOP:
1832 			case IPV6_JOIN_GROUP:
1833 			case IPV6_LEAVE_GROUP:
1834 			case IPV6_MSFILTER:
1835 			case MCAST_BLOCK_SOURCE:
1836 			case MCAST_UNBLOCK_SOURCE:
1837 			case MCAST_JOIN_GROUP:
1838 			case MCAST_LEAVE_GROUP:
1839 			case MCAST_JOIN_SOURCE_GROUP:
1840 			case MCAST_LEAVE_SOURCE_GROUP:
1841 				error = ip6_setmoptions(in6p, sopt);
1842 				break;
1843 
1844 			case IPV6_PORTRANGE:
1845 				error = sooptcopyin(sopt, &optval,
1846 				    sizeof optval, sizeof optval);
1847 				if (error)
1848 					break;
1849 
1850 				INP_WLOCK(in6p);
1851 				switch (optval) {
1852 				case IPV6_PORTRANGE_DEFAULT:
1853 					in6p->inp_flags &= ~(INP_LOWPORT);
1854 					in6p->inp_flags &= ~(INP_HIGHPORT);
1855 					break;
1856 
1857 				case IPV6_PORTRANGE_HIGH:
1858 					in6p->inp_flags &= ~(INP_LOWPORT);
1859 					in6p->inp_flags |= INP_HIGHPORT;
1860 					break;
1861 
1862 				case IPV6_PORTRANGE_LOW:
1863 					in6p->inp_flags &= ~(INP_HIGHPORT);
1864 					in6p->inp_flags |= INP_LOWPORT;
1865 					break;
1866 
1867 				default:
1868 					error = EINVAL;
1869 					break;
1870 				}
1871 				INP_WUNLOCK(in6p);
1872 				break;
1873 
1874 #ifdef IPSEC
1875 			case IPV6_IPSEC_POLICY:
1876 			{
1877 				caddr_t req;
1878 				struct mbuf *m;
1879 
1880 				if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
1881 					break;
1882 				if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
1883 					break;
1884 				req = mtod(m, caddr_t);
1885 				error = ipsec_set_policy(in6p, optname, req,
1886 				    m->m_len, (sopt->sopt_td != NULL) ?
1887 				    sopt->sopt_td->td_ucred : NULL);
1888 				m_freem(m);
1889 				break;
1890 			}
1891 #endif /* IPSEC */
1892 
1893 			default:
1894 				error = ENOPROTOOPT;
1895 				break;
1896 			}
1897 			break;
1898 
1899 		case SOPT_GET:
1900 			switch (optname) {
1901 
1902 			case IPV6_2292PKTOPTIONS:
1903 #ifdef IPV6_PKTOPTIONS
1904 			case IPV6_PKTOPTIONS:
1905 #endif
1906 				/*
1907 				 * RFC3542 (effectively) deprecated the
1908 				 * semantics of the 2292-style pktoptions.
1909 				 * Since it was not reliable in nature (i.e.,
1910 				 * applications had to expect the lack of some
1911 				 * information after all), it would make sense
1912 				 * to simplify this part by always returning
1913 				 * empty data.
1914 				 */
1915 				sopt->sopt_valsize = 0;
1916 				break;
1917 
1918 			case IPV6_RECVHOPOPTS:
1919 			case IPV6_RECVDSTOPTS:
1920 			case IPV6_RECVRTHDRDSTOPTS:
1921 			case IPV6_UNICAST_HOPS:
1922 			case IPV6_RECVPKTINFO:
1923 			case IPV6_RECVHOPLIMIT:
1924 			case IPV6_RECVRTHDR:
1925 			case IPV6_RECVPATHMTU:
1926 
1927 			case IPV6_V6ONLY:
1928 			case IPV6_PORTRANGE:
1929 			case IPV6_RECVTCLASS:
1930 			case IPV6_AUTOFLOWLABEL:
1931 			case IPV6_BINDANY:
1932 			case IPV6_FLOWID:
1933 			case IPV6_FLOWTYPE:
1934 			case IPV6_RECVFLOWID:
1935 #ifdef	RSS
1936 			case IPV6_RSSBUCKETID:
1937 			case IPV6_RECVRSSBUCKETID:
1938 #endif
1939 			case IPV6_BINDMULTI:
1940 				switch (optname) {
1941 
1942 				case IPV6_RECVHOPOPTS:
1943 					optval = OPTBIT(IN6P_HOPOPTS);
1944 					break;
1945 
1946 				case IPV6_RECVDSTOPTS:
1947 					optval = OPTBIT(IN6P_DSTOPTS);
1948 					break;
1949 
1950 				case IPV6_RECVRTHDRDSTOPTS:
1951 					optval = OPTBIT(IN6P_RTHDRDSTOPTS);
1952 					break;
1953 
1954 				case IPV6_UNICAST_HOPS:
1955 					optval = in6p->in6p_hops;
1956 					break;
1957 
1958 				case IPV6_RECVPKTINFO:
1959 					optval = OPTBIT(IN6P_PKTINFO);
1960 					break;
1961 
1962 				case IPV6_RECVHOPLIMIT:
1963 					optval = OPTBIT(IN6P_HOPLIMIT);
1964 					break;
1965 
1966 				case IPV6_RECVRTHDR:
1967 					optval = OPTBIT(IN6P_RTHDR);
1968 					break;
1969 
1970 				case IPV6_RECVPATHMTU:
1971 					optval = OPTBIT(IN6P_MTU);
1972 					break;
1973 
1974 				case IPV6_V6ONLY:
1975 					optval = OPTBIT(IN6P_IPV6_V6ONLY);
1976 					break;
1977 
1978 				case IPV6_PORTRANGE:
1979 				    {
1980 					int flags;
1981 					flags = in6p->inp_flags;
1982 					if (flags & INP_HIGHPORT)
1983 						optval = IPV6_PORTRANGE_HIGH;
1984 					else if (flags & INP_LOWPORT)
1985 						optval = IPV6_PORTRANGE_LOW;
1986 					else
1987 						optval = 0;
1988 					break;
1989 				    }
1990 				case IPV6_RECVTCLASS:
1991 					optval = OPTBIT(IN6P_TCLASS);
1992 					break;
1993 
1994 				case IPV6_AUTOFLOWLABEL:
1995 					optval = OPTBIT(IN6P_AUTOFLOWLABEL);
1996 					break;
1997 
1998 				case IPV6_BINDANY:
1999 					optval = OPTBIT(INP_BINDANY);
2000 					break;
2001 
2002 				case IPV6_FLOWID:
2003 					optval = in6p->inp_flowid;
2004 					break;
2005 
2006 				case IPV6_FLOWTYPE:
2007 					optval = in6p->inp_flowtype;
2008 					break;
2009 
2010 				case IPV6_RECVFLOWID:
2011 					optval = OPTBIT2(INP_RECVFLOWID);
2012 					break;
2013 #ifdef	RSS
2014 				case IPV6_RSSBUCKETID:
2015 					retval =
2016 					    rss_hash2bucket(in6p->inp_flowid,
2017 					    in6p->inp_flowtype,
2018 					    &rss_bucket);
2019 					if (retval == 0)
2020 						optval = rss_bucket;
2021 					else
2022 						error = EINVAL;
2023 					break;
2024 
2025 				case IPV6_RECVRSSBUCKETID:
2026 					optval = OPTBIT2(INP_RECVRSSBUCKETID);
2027 					break;
2028 #endif
2029 
2030 				case IPV6_BINDMULTI:
2031 					optval = OPTBIT2(INP_BINDMULTI);
2032 					break;
2033 
2034 				}
2035 				if (error)
2036 					break;
2037 				error = sooptcopyout(sopt, &optval,
2038 					sizeof optval);
2039 				break;
2040 
2041 			case IPV6_PATHMTU:
2042 			{
2043 				u_long pmtu = 0;
2044 				struct ip6_mtuinfo mtuinfo;
2045 
2046 				if (!(so->so_state & SS_ISCONNECTED))
2047 					return (ENOTCONN);
2048 				/*
2049 				 * XXX: we dot not consider the case of source
2050 				 * routing, or optional information to specify
2051 				 * the outgoing interface.
2052 				 */
2053 				error = ip6_getpmtu_ctl(so->so_fibnum,
2054 				    &in6p->in6p_faddr, &pmtu);
2055 				if (error)
2056 					break;
2057 				if (pmtu > IPV6_MAXPACKET)
2058 					pmtu = IPV6_MAXPACKET;
2059 
2060 				bzero(&mtuinfo, sizeof(mtuinfo));
2061 				mtuinfo.ip6m_mtu = (u_int32_t)pmtu;
2062 				optdata = (void *)&mtuinfo;
2063 				optdatalen = sizeof(mtuinfo);
2064 				error = sooptcopyout(sopt, optdata,
2065 				    optdatalen);
2066 				break;
2067 			}
2068 
2069 			case IPV6_2292PKTINFO:
2070 			case IPV6_2292HOPLIMIT:
2071 			case IPV6_2292HOPOPTS:
2072 			case IPV6_2292RTHDR:
2073 			case IPV6_2292DSTOPTS:
2074 				switch (optname) {
2075 				case IPV6_2292PKTINFO:
2076 					optval = OPTBIT(IN6P_PKTINFO);
2077 					break;
2078 				case IPV6_2292HOPLIMIT:
2079 					optval = OPTBIT(IN6P_HOPLIMIT);
2080 					break;
2081 				case IPV6_2292HOPOPTS:
2082 					optval = OPTBIT(IN6P_HOPOPTS);
2083 					break;
2084 				case IPV6_2292RTHDR:
2085 					optval = OPTBIT(IN6P_RTHDR);
2086 					break;
2087 				case IPV6_2292DSTOPTS:
2088 					optval = OPTBIT(IN6P_DSTOPTS|IN6P_RTHDRDSTOPTS);
2089 					break;
2090 				}
2091 				error = sooptcopyout(sopt, &optval,
2092 				    sizeof optval);
2093 				break;
2094 			case IPV6_PKTINFO:
2095 			case IPV6_HOPOPTS:
2096 			case IPV6_RTHDR:
2097 			case IPV6_DSTOPTS:
2098 			case IPV6_RTHDRDSTOPTS:
2099 			case IPV6_NEXTHOP:
2100 			case IPV6_TCLASS:
2101 			case IPV6_DONTFRAG:
2102 			case IPV6_USE_MIN_MTU:
2103 			case IPV6_PREFER_TEMPADDR:
2104 				error = ip6_getpcbopt(in6p->in6p_outputopts,
2105 				    optname, sopt);
2106 				break;
2107 
2108 			case IPV6_MULTICAST_IF:
2109 			case IPV6_MULTICAST_HOPS:
2110 			case IPV6_MULTICAST_LOOP:
2111 			case IPV6_MSFILTER:
2112 				error = ip6_getmoptions(in6p, sopt);
2113 				break;
2114 
2115 #ifdef IPSEC
2116 			case IPV6_IPSEC_POLICY:
2117 			  {
2118 				caddr_t req = NULL;
2119 				size_t len = 0;
2120 				struct mbuf *m = NULL;
2121 				struct mbuf **mp = &m;
2122 				size_t ovalsize = sopt->sopt_valsize;
2123 				caddr_t oval = (caddr_t)sopt->sopt_val;
2124 
2125 				error = soopt_getm(sopt, &m); /* XXX */
2126 				if (error != 0)
2127 					break;
2128 				error = soopt_mcopyin(sopt, m); /* XXX */
2129 				if (error != 0)
2130 					break;
2131 				sopt->sopt_valsize = ovalsize;
2132 				sopt->sopt_val = oval;
2133 				if (m) {
2134 					req = mtod(m, caddr_t);
2135 					len = m->m_len;
2136 				}
2137 				error = ipsec_get_policy(in6p, req, len, mp);
2138 				if (error == 0)
2139 					error = soopt_mcopyout(sopt, m); /* XXX */
2140 				if (error == 0 && m)
2141 					m_freem(m);
2142 				break;
2143 			  }
2144 #endif /* IPSEC */
2145 
2146 			default:
2147 				error = ENOPROTOOPT;
2148 				break;
2149 			}
2150 			break;
2151 		}
2152 	}
2153 	return (error);
2154 }
2155 
2156 int
2157 ip6_raw_ctloutput(struct socket *so, struct sockopt *sopt)
2158 {
2159 	int error = 0, optval, optlen;
2160 	const int icmp6off = offsetof(struct icmp6_hdr, icmp6_cksum);
2161 	struct inpcb *in6p = sotoinpcb(so);
2162 	int level, op, optname;
2163 
2164 	level = sopt->sopt_level;
2165 	op = sopt->sopt_dir;
2166 	optname = sopt->sopt_name;
2167 	optlen = sopt->sopt_valsize;
2168 
2169 	if (level != IPPROTO_IPV6) {
2170 		return (EINVAL);
2171 	}
2172 
2173 	switch (optname) {
2174 	case IPV6_CHECKSUM:
2175 		/*
2176 		 * For ICMPv6 sockets, no modification allowed for checksum
2177 		 * offset, permit "no change" values to help existing apps.
2178 		 *
2179 		 * RFC3542 says: "An attempt to set IPV6_CHECKSUM
2180 		 * for an ICMPv6 socket will fail."
2181 		 * The current behavior does not meet RFC3542.
2182 		 */
2183 		switch (op) {
2184 		case SOPT_SET:
2185 			if (optlen != sizeof(int)) {
2186 				error = EINVAL;
2187 				break;
2188 			}
2189 			error = sooptcopyin(sopt, &optval, sizeof(optval),
2190 					    sizeof(optval));
2191 			if (error)
2192 				break;
2193 			if ((optval % 2) != 0) {
2194 				/* the API assumes even offset values */
2195 				error = EINVAL;
2196 			} else if (so->so_proto->pr_protocol ==
2197 			    IPPROTO_ICMPV6) {
2198 				if (optval != icmp6off)
2199 					error = EINVAL;
2200 			} else
2201 				in6p->in6p_cksum = optval;
2202 			break;
2203 
2204 		case SOPT_GET:
2205 			if (so->so_proto->pr_protocol == IPPROTO_ICMPV6)
2206 				optval = icmp6off;
2207 			else
2208 				optval = in6p->in6p_cksum;
2209 
2210 			error = sooptcopyout(sopt, &optval, sizeof(optval));
2211 			break;
2212 
2213 		default:
2214 			error = EINVAL;
2215 			break;
2216 		}
2217 		break;
2218 
2219 	default:
2220 		error = ENOPROTOOPT;
2221 		break;
2222 	}
2223 
2224 	return (error);
2225 }
2226 
2227 /*
2228  * Set up IP6 options in pcb for insertion in output packets or
2229  * specifying behavior of outgoing packets.
2230  */
2231 static int
2232 ip6_pcbopts(struct ip6_pktopts **pktopt, struct mbuf *m,
2233     struct socket *so, struct sockopt *sopt)
2234 {
2235 	struct ip6_pktopts *opt = *pktopt;
2236 	int error = 0;
2237 	struct thread *td = sopt->sopt_td;
2238 
2239 	/* turn off any old options. */
2240 	if (opt) {
2241 #ifdef DIAGNOSTIC
2242 		if (opt->ip6po_pktinfo || opt->ip6po_nexthop ||
2243 		    opt->ip6po_hbh || opt->ip6po_dest1 || opt->ip6po_dest2 ||
2244 		    opt->ip6po_rhinfo.ip6po_rhi_rthdr)
2245 			printf("ip6_pcbopts: all specified options are cleared.\n");
2246 #endif
2247 		ip6_clearpktopts(opt, -1);
2248 	} else
2249 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
2250 	*pktopt = NULL;
2251 
2252 	if (!m || m->m_len == 0) {
2253 		/*
2254 		 * Only turning off any previous options, regardless of
2255 		 * whether the opt is just created or given.
2256 		 */
2257 		free(opt, M_IP6OPT);
2258 		return (0);
2259 	}
2260 
2261 	/*  set options specified by user. */
2262 	if ((error = ip6_setpktopts(m, opt, NULL, (td != NULL) ?
2263 	    td->td_ucred : NULL, so->so_proto->pr_protocol)) != 0) {
2264 		ip6_clearpktopts(opt, -1); /* XXX: discard all options */
2265 		free(opt, M_IP6OPT);
2266 		return (error);
2267 	}
2268 	*pktopt = opt;
2269 	return (0);
2270 }
2271 
2272 /*
2273  * initialize ip6_pktopts.  beware that there are non-zero default values in
2274  * the struct.
2275  */
2276 void
2277 ip6_initpktopts(struct ip6_pktopts *opt)
2278 {
2279 
2280 	bzero(opt, sizeof(*opt));
2281 	opt->ip6po_hlim = -1;	/* -1 means default hop limit */
2282 	opt->ip6po_tclass = -1;	/* -1 means default traffic class */
2283 	opt->ip6po_minmtu = IP6PO_MINMTU_MCASTONLY;
2284 	opt->ip6po_prefer_tempaddr = IP6PO_TEMPADDR_SYSTEM;
2285 }
2286 
2287 static int
2288 ip6_pcbopt(int optname, u_char *buf, int len, struct ip6_pktopts **pktopt,
2289     struct ucred *cred, int uproto)
2290 {
2291 	struct ip6_pktopts *opt;
2292 
2293 	if (*pktopt == NULL) {
2294 		*pktopt = malloc(sizeof(struct ip6_pktopts), M_IP6OPT,
2295 		    M_WAITOK);
2296 		ip6_initpktopts(*pktopt);
2297 	}
2298 	opt = *pktopt;
2299 
2300 	return (ip6_setpktopt(optname, buf, len, opt, cred, 1, 0, uproto));
2301 }
2302 
2303 static int
2304 ip6_getpcbopt(struct ip6_pktopts *pktopt, int optname, struct sockopt *sopt)
2305 {
2306 	void *optdata = NULL;
2307 	int optdatalen = 0;
2308 	struct ip6_ext *ip6e;
2309 	int error = 0;
2310 	struct in6_pktinfo null_pktinfo;
2311 	int deftclass = 0, on;
2312 	int defminmtu = IP6PO_MINMTU_MCASTONLY;
2313 	int defpreftemp = IP6PO_TEMPADDR_SYSTEM;
2314 
2315 	switch (optname) {
2316 	case IPV6_PKTINFO:
2317 		optdata = (void *)&null_pktinfo;
2318 		if (pktopt && pktopt->ip6po_pktinfo) {
2319 			bcopy(pktopt->ip6po_pktinfo, &null_pktinfo,
2320 			    sizeof(null_pktinfo));
2321 			in6_clearscope(&null_pktinfo.ipi6_addr);
2322 		} else {
2323 			/* XXX: we don't have to do this every time... */
2324 			bzero(&null_pktinfo, sizeof(null_pktinfo));
2325 		}
2326 		optdatalen = sizeof(struct in6_pktinfo);
2327 		break;
2328 	case IPV6_TCLASS:
2329 		if (pktopt && pktopt->ip6po_tclass >= 0)
2330 			optdata = (void *)&pktopt->ip6po_tclass;
2331 		else
2332 			optdata = (void *)&deftclass;
2333 		optdatalen = sizeof(int);
2334 		break;
2335 	case IPV6_HOPOPTS:
2336 		if (pktopt && pktopt->ip6po_hbh) {
2337 			optdata = (void *)pktopt->ip6po_hbh;
2338 			ip6e = (struct ip6_ext *)pktopt->ip6po_hbh;
2339 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2340 		}
2341 		break;
2342 	case IPV6_RTHDR:
2343 		if (pktopt && pktopt->ip6po_rthdr) {
2344 			optdata = (void *)pktopt->ip6po_rthdr;
2345 			ip6e = (struct ip6_ext *)pktopt->ip6po_rthdr;
2346 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2347 		}
2348 		break;
2349 	case IPV6_RTHDRDSTOPTS:
2350 		if (pktopt && pktopt->ip6po_dest1) {
2351 			optdata = (void *)pktopt->ip6po_dest1;
2352 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest1;
2353 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2354 		}
2355 		break;
2356 	case IPV6_DSTOPTS:
2357 		if (pktopt && pktopt->ip6po_dest2) {
2358 			optdata = (void *)pktopt->ip6po_dest2;
2359 			ip6e = (struct ip6_ext *)pktopt->ip6po_dest2;
2360 			optdatalen = (ip6e->ip6e_len + 1) << 3;
2361 		}
2362 		break;
2363 	case IPV6_NEXTHOP:
2364 		if (pktopt && pktopt->ip6po_nexthop) {
2365 			optdata = (void *)pktopt->ip6po_nexthop;
2366 			optdatalen = pktopt->ip6po_nexthop->sa_len;
2367 		}
2368 		break;
2369 	case IPV6_USE_MIN_MTU:
2370 		if (pktopt)
2371 			optdata = (void *)&pktopt->ip6po_minmtu;
2372 		else
2373 			optdata = (void *)&defminmtu;
2374 		optdatalen = sizeof(int);
2375 		break;
2376 	case IPV6_DONTFRAG:
2377 		if (pktopt && ((pktopt->ip6po_flags) & IP6PO_DONTFRAG))
2378 			on = 1;
2379 		else
2380 			on = 0;
2381 		optdata = (void *)&on;
2382 		optdatalen = sizeof(on);
2383 		break;
2384 	case IPV6_PREFER_TEMPADDR:
2385 		if (pktopt)
2386 			optdata = (void *)&pktopt->ip6po_prefer_tempaddr;
2387 		else
2388 			optdata = (void *)&defpreftemp;
2389 		optdatalen = sizeof(int);
2390 		break;
2391 	default:		/* should not happen */
2392 #ifdef DIAGNOSTIC
2393 		panic("ip6_getpcbopt: unexpected option\n");
2394 #endif
2395 		return (ENOPROTOOPT);
2396 	}
2397 
2398 	error = sooptcopyout(sopt, optdata, optdatalen);
2399 
2400 	return (error);
2401 }
2402 
2403 void
2404 ip6_clearpktopts(struct ip6_pktopts *pktopt, int optname)
2405 {
2406 	if (pktopt == NULL)
2407 		return;
2408 
2409 	if (optname == -1 || optname == IPV6_PKTINFO) {
2410 		if (pktopt->ip6po_pktinfo)
2411 			free(pktopt->ip6po_pktinfo, M_IP6OPT);
2412 		pktopt->ip6po_pktinfo = NULL;
2413 	}
2414 	if (optname == -1 || optname == IPV6_HOPLIMIT)
2415 		pktopt->ip6po_hlim = -1;
2416 	if (optname == -1 || optname == IPV6_TCLASS)
2417 		pktopt->ip6po_tclass = -1;
2418 	if (optname == -1 || optname == IPV6_NEXTHOP) {
2419 		if (pktopt->ip6po_nextroute.ro_rt) {
2420 			RTFREE(pktopt->ip6po_nextroute.ro_rt);
2421 			pktopt->ip6po_nextroute.ro_rt = NULL;
2422 		}
2423 		if (pktopt->ip6po_nexthop)
2424 			free(pktopt->ip6po_nexthop, M_IP6OPT);
2425 		pktopt->ip6po_nexthop = NULL;
2426 	}
2427 	if (optname == -1 || optname == IPV6_HOPOPTS) {
2428 		if (pktopt->ip6po_hbh)
2429 			free(pktopt->ip6po_hbh, M_IP6OPT);
2430 		pktopt->ip6po_hbh = NULL;
2431 	}
2432 	if (optname == -1 || optname == IPV6_RTHDRDSTOPTS) {
2433 		if (pktopt->ip6po_dest1)
2434 			free(pktopt->ip6po_dest1, M_IP6OPT);
2435 		pktopt->ip6po_dest1 = NULL;
2436 	}
2437 	if (optname == -1 || optname == IPV6_RTHDR) {
2438 		if (pktopt->ip6po_rhinfo.ip6po_rhi_rthdr)
2439 			free(pktopt->ip6po_rhinfo.ip6po_rhi_rthdr, M_IP6OPT);
2440 		pktopt->ip6po_rhinfo.ip6po_rhi_rthdr = NULL;
2441 		if (pktopt->ip6po_route.ro_rt) {
2442 			RTFREE(pktopt->ip6po_route.ro_rt);
2443 			pktopt->ip6po_route.ro_rt = NULL;
2444 		}
2445 	}
2446 	if (optname == -1 || optname == IPV6_DSTOPTS) {
2447 		if (pktopt->ip6po_dest2)
2448 			free(pktopt->ip6po_dest2, M_IP6OPT);
2449 		pktopt->ip6po_dest2 = NULL;
2450 	}
2451 }
2452 
2453 #define PKTOPT_EXTHDRCPY(type) \
2454 do {\
2455 	if (src->type) {\
2456 		int hlen = (((struct ip6_ext *)src->type)->ip6e_len + 1) << 3;\
2457 		dst->type = malloc(hlen, M_IP6OPT, canwait);\
2458 		if (dst->type == NULL && canwait == M_NOWAIT)\
2459 			goto bad;\
2460 		bcopy(src->type, dst->type, hlen);\
2461 	}\
2462 } while (/*CONSTCOND*/ 0)
2463 
2464 static int
2465 copypktopts(struct ip6_pktopts *dst, struct ip6_pktopts *src, int canwait)
2466 {
2467 	if (dst == NULL || src == NULL)  {
2468 		printf("ip6_clearpktopts: invalid argument\n");
2469 		return (EINVAL);
2470 	}
2471 
2472 	dst->ip6po_hlim = src->ip6po_hlim;
2473 	dst->ip6po_tclass = src->ip6po_tclass;
2474 	dst->ip6po_flags = src->ip6po_flags;
2475 	dst->ip6po_minmtu = src->ip6po_minmtu;
2476 	dst->ip6po_prefer_tempaddr = src->ip6po_prefer_tempaddr;
2477 	if (src->ip6po_pktinfo) {
2478 		dst->ip6po_pktinfo = malloc(sizeof(*dst->ip6po_pktinfo),
2479 		    M_IP6OPT, canwait);
2480 		if (dst->ip6po_pktinfo == NULL)
2481 			goto bad;
2482 		*dst->ip6po_pktinfo = *src->ip6po_pktinfo;
2483 	}
2484 	if (src->ip6po_nexthop) {
2485 		dst->ip6po_nexthop = malloc(src->ip6po_nexthop->sa_len,
2486 		    M_IP6OPT, canwait);
2487 		if (dst->ip6po_nexthop == NULL)
2488 			goto bad;
2489 		bcopy(src->ip6po_nexthop, dst->ip6po_nexthop,
2490 		    src->ip6po_nexthop->sa_len);
2491 	}
2492 	PKTOPT_EXTHDRCPY(ip6po_hbh);
2493 	PKTOPT_EXTHDRCPY(ip6po_dest1);
2494 	PKTOPT_EXTHDRCPY(ip6po_dest2);
2495 	PKTOPT_EXTHDRCPY(ip6po_rthdr); /* not copy the cached route */
2496 	return (0);
2497 
2498   bad:
2499 	ip6_clearpktopts(dst, -1);
2500 	return (ENOBUFS);
2501 }
2502 #undef PKTOPT_EXTHDRCPY
2503 
2504 struct ip6_pktopts *
2505 ip6_copypktopts(struct ip6_pktopts *src, int canwait)
2506 {
2507 	int error;
2508 	struct ip6_pktopts *dst;
2509 
2510 	dst = malloc(sizeof(*dst), M_IP6OPT, canwait);
2511 	if (dst == NULL)
2512 		return (NULL);
2513 	ip6_initpktopts(dst);
2514 
2515 	if ((error = copypktopts(dst, src, canwait)) != 0) {
2516 		free(dst, M_IP6OPT);
2517 		return (NULL);
2518 	}
2519 
2520 	return (dst);
2521 }
2522 
2523 void
2524 ip6_freepcbopts(struct ip6_pktopts *pktopt)
2525 {
2526 	if (pktopt == NULL)
2527 		return;
2528 
2529 	ip6_clearpktopts(pktopt, -1);
2530 
2531 	free(pktopt, M_IP6OPT);
2532 }
2533 
2534 /*
2535  * Set IPv6 outgoing packet options based on advanced API.
2536  */
2537 int
2538 ip6_setpktopts(struct mbuf *control, struct ip6_pktopts *opt,
2539     struct ip6_pktopts *stickyopt, struct ucred *cred, int uproto)
2540 {
2541 	struct cmsghdr *cm = NULL;
2542 
2543 	if (control == NULL || opt == NULL)
2544 		return (EINVAL);
2545 
2546 	ip6_initpktopts(opt);
2547 	if (stickyopt) {
2548 		int error;
2549 
2550 		/*
2551 		 * If stickyopt is provided, make a local copy of the options
2552 		 * for this particular packet, then override them by ancillary
2553 		 * objects.
2554 		 * XXX: copypktopts() does not copy the cached route to a next
2555 		 * hop (if any).  This is not very good in terms of efficiency,
2556 		 * but we can allow this since this option should be rarely
2557 		 * used.
2558 		 */
2559 		if ((error = copypktopts(opt, stickyopt, M_NOWAIT)) != 0)
2560 			return (error);
2561 	}
2562 
2563 	/*
2564 	 * XXX: Currently, we assume all the optional information is stored
2565 	 * in a single mbuf.
2566 	 */
2567 	if (control->m_next)
2568 		return (EINVAL);
2569 
2570 	for (; control->m_len > 0; control->m_data += CMSG_ALIGN(cm->cmsg_len),
2571 	    control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
2572 		int error;
2573 
2574 		if (control->m_len < CMSG_LEN(0))
2575 			return (EINVAL);
2576 
2577 		cm = mtod(control, struct cmsghdr *);
2578 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
2579 			return (EINVAL);
2580 		if (cm->cmsg_level != IPPROTO_IPV6)
2581 			continue;
2582 
2583 		error = ip6_setpktopt(cm->cmsg_type, CMSG_DATA(cm),
2584 		    cm->cmsg_len - CMSG_LEN(0), opt, cred, 0, 1, uproto);
2585 		if (error)
2586 			return (error);
2587 	}
2588 
2589 	return (0);
2590 }
2591 
2592 /*
2593  * Set a particular packet option, as a sticky option or an ancillary data
2594  * item.  "len" can be 0 only when it's a sticky option.
2595  * We have 4 cases of combination of "sticky" and "cmsg":
2596  * "sticky=0, cmsg=0": impossible
2597  * "sticky=0, cmsg=1": RFC2292 or RFC3542 ancillary data
2598  * "sticky=1, cmsg=0": RFC3542 socket option
2599  * "sticky=1, cmsg=1": RFC2292 socket option
2600  */
2601 static int
2602 ip6_setpktopt(int optname, u_char *buf, int len, struct ip6_pktopts *opt,
2603     struct ucred *cred, int sticky, int cmsg, int uproto)
2604 {
2605 	int minmtupolicy, preftemp;
2606 	int error;
2607 
2608 	if (!sticky && !cmsg) {
2609 #ifdef DIAGNOSTIC
2610 		printf("ip6_setpktopt: impossible case\n");
2611 #endif
2612 		return (EINVAL);
2613 	}
2614 
2615 	/*
2616 	 * IPV6_2292xxx is for backward compatibility to RFC2292, and should
2617 	 * not be specified in the context of RFC3542.  Conversely,
2618 	 * RFC3542 types should not be specified in the context of RFC2292.
2619 	 */
2620 	if (!cmsg) {
2621 		switch (optname) {
2622 		case IPV6_2292PKTINFO:
2623 		case IPV6_2292HOPLIMIT:
2624 		case IPV6_2292NEXTHOP:
2625 		case IPV6_2292HOPOPTS:
2626 		case IPV6_2292DSTOPTS:
2627 		case IPV6_2292RTHDR:
2628 		case IPV6_2292PKTOPTIONS:
2629 			return (ENOPROTOOPT);
2630 		}
2631 	}
2632 	if (sticky && cmsg) {
2633 		switch (optname) {
2634 		case IPV6_PKTINFO:
2635 		case IPV6_HOPLIMIT:
2636 		case IPV6_NEXTHOP:
2637 		case IPV6_HOPOPTS:
2638 		case IPV6_DSTOPTS:
2639 		case IPV6_RTHDRDSTOPTS:
2640 		case IPV6_RTHDR:
2641 		case IPV6_USE_MIN_MTU:
2642 		case IPV6_DONTFRAG:
2643 		case IPV6_TCLASS:
2644 		case IPV6_PREFER_TEMPADDR: /* XXX: not an RFC3542 option */
2645 			return (ENOPROTOOPT);
2646 		}
2647 	}
2648 
2649 	switch (optname) {
2650 	case IPV6_2292PKTINFO:
2651 	case IPV6_PKTINFO:
2652 	{
2653 		struct ifnet *ifp = NULL;
2654 		struct in6_pktinfo *pktinfo;
2655 
2656 		if (len != sizeof(struct in6_pktinfo))
2657 			return (EINVAL);
2658 
2659 		pktinfo = (struct in6_pktinfo *)buf;
2660 
2661 		/*
2662 		 * An application can clear any sticky IPV6_PKTINFO option by
2663 		 * doing a "regular" setsockopt with ipi6_addr being
2664 		 * in6addr_any and ipi6_ifindex being zero.
2665 		 * [RFC 3542, Section 6]
2666 		 */
2667 		if (optname == IPV6_PKTINFO && opt->ip6po_pktinfo &&
2668 		    pktinfo->ipi6_ifindex == 0 &&
2669 		    IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2670 			ip6_clearpktopts(opt, optname);
2671 			break;
2672 		}
2673 
2674 		if (uproto == IPPROTO_TCP && optname == IPV6_PKTINFO &&
2675 		    sticky && !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2676 			return (EINVAL);
2677 		}
2678 		if (IN6_IS_ADDR_MULTICAST(&pktinfo->ipi6_addr))
2679 			return (EINVAL);
2680 		/* validate the interface index if specified. */
2681 		if (pktinfo->ipi6_ifindex > V_if_index)
2682 			 return (ENXIO);
2683 		if (pktinfo->ipi6_ifindex) {
2684 			ifp = ifnet_byindex(pktinfo->ipi6_ifindex);
2685 			if (ifp == NULL)
2686 				return (ENXIO);
2687 		}
2688 		if (ifp != NULL && (ifp->if_afdata[AF_INET6] == NULL ||
2689 		    (ND_IFINFO(ifp)->flags & ND6_IFF_IFDISABLED) != 0))
2690 			return (ENETDOWN);
2691 
2692 		if (ifp != NULL &&
2693 		    !IN6_IS_ADDR_UNSPECIFIED(&pktinfo->ipi6_addr)) {
2694 			struct in6_ifaddr *ia;
2695 
2696 			in6_setscope(&pktinfo->ipi6_addr, ifp, NULL);
2697 			ia = in6ifa_ifpwithaddr(ifp, &pktinfo->ipi6_addr);
2698 			if (ia == NULL)
2699 				return (EADDRNOTAVAIL);
2700 			ifa_free(&ia->ia_ifa);
2701 		}
2702 		/*
2703 		 * We store the address anyway, and let in6_selectsrc()
2704 		 * validate the specified address.  This is because ipi6_addr
2705 		 * may not have enough information about its scope zone, and
2706 		 * we may need additional information (such as outgoing
2707 		 * interface or the scope zone of a destination address) to
2708 		 * disambiguate the scope.
2709 		 * XXX: the delay of the validation may confuse the
2710 		 * application when it is used as a sticky option.
2711 		 */
2712 		if (opt->ip6po_pktinfo == NULL) {
2713 			opt->ip6po_pktinfo = malloc(sizeof(*pktinfo),
2714 			    M_IP6OPT, M_NOWAIT);
2715 			if (opt->ip6po_pktinfo == NULL)
2716 				return (ENOBUFS);
2717 		}
2718 		bcopy(pktinfo, opt->ip6po_pktinfo, sizeof(*pktinfo));
2719 		break;
2720 	}
2721 
2722 	case IPV6_2292HOPLIMIT:
2723 	case IPV6_HOPLIMIT:
2724 	{
2725 		int *hlimp;
2726 
2727 		/*
2728 		 * RFC 3542 deprecated the usage of sticky IPV6_HOPLIMIT
2729 		 * to simplify the ordering among hoplimit options.
2730 		 */
2731 		if (optname == IPV6_HOPLIMIT && sticky)
2732 			return (ENOPROTOOPT);
2733 
2734 		if (len != sizeof(int))
2735 			return (EINVAL);
2736 		hlimp = (int *)buf;
2737 		if (*hlimp < -1 || *hlimp > 255)
2738 			return (EINVAL);
2739 
2740 		opt->ip6po_hlim = *hlimp;
2741 		break;
2742 	}
2743 
2744 	case IPV6_TCLASS:
2745 	{
2746 		int tclass;
2747 
2748 		if (len != sizeof(int))
2749 			return (EINVAL);
2750 		tclass = *(int *)buf;
2751 		if (tclass < -1 || tclass > 255)
2752 			return (EINVAL);
2753 
2754 		opt->ip6po_tclass = tclass;
2755 		break;
2756 	}
2757 
2758 	case IPV6_2292NEXTHOP:
2759 	case IPV6_NEXTHOP:
2760 		if (cred != NULL) {
2761 			error = priv_check_cred(cred,
2762 			    PRIV_NETINET_SETHDROPTS, 0);
2763 			if (error)
2764 				return (error);
2765 		}
2766 
2767 		if (len == 0) {	/* just remove the option */
2768 			ip6_clearpktopts(opt, IPV6_NEXTHOP);
2769 			break;
2770 		}
2771 
2772 		/* check if cmsg_len is large enough for sa_len */
2773 		if (len < sizeof(struct sockaddr) || len < *buf)
2774 			return (EINVAL);
2775 
2776 		switch (((struct sockaddr *)buf)->sa_family) {
2777 		case AF_INET6:
2778 		{
2779 			struct sockaddr_in6 *sa6 = (struct sockaddr_in6 *)buf;
2780 			int error;
2781 
2782 			if (sa6->sin6_len != sizeof(struct sockaddr_in6))
2783 				return (EINVAL);
2784 
2785 			if (IN6_IS_ADDR_UNSPECIFIED(&sa6->sin6_addr) ||
2786 			    IN6_IS_ADDR_MULTICAST(&sa6->sin6_addr)) {
2787 				return (EINVAL);
2788 			}
2789 			if ((error = sa6_embedscope(sa6, V_ip6_use_defzone))
2790 			    != 0) {
2791 				return (error);
2792 			}
2793 			break;
2794 		}
2795 		case AF_LINK:	/* should eventually be supported */
2796 		default:
2797 			return (EAFNOSUPPORT);
2798 		}
2799 
2800 		/* turn off the previous option, then set the new option. */
2801 		ip6_clearpktopts(opt, IPV6_NEXTHOP);
2802 		opt->ip6po_nexthop = malloc(*buf, M_IP6OPT, M_NOWAIT);
2803 		if (opt->ip6po_nexthop == NULL)
2804 			return (ENOBUFS);
2805 		bcopy(buf, opt->ip6po_nexthop, *buf);
2806 		break;
2807 
2808 	case IPV6_2292HOPOPTS:
2809 	case IPV6_HOPOPTS:
2810 	{
2811 		struct ip6_hbh *hbh;
2812 		int hbhlen;
2813 
2814 		/*
2815 		 * XXX: We don't allow a non-privileged user to set ANY HbH
2816 		 * options, since per-option restriction has too much
2817 		 * overhead.
2818 		 */
2819 		if (cred != NULL) {
2820 			error = priv_check_cred(cred,
2821 			    PRIV_NETINET_SETHDROPTS, 0);
2822 			if (error)
2823 				return (error);
2824 		}
2825 
2826 		if (len == 0) {
2827 			ip6_clearpktopts(opt, IPV6_HOPOPTS);
2828 			break;	/* just remove the option */
2829 		}
2830 
2831 		/* message length validation */
2832 		if (len < sizeof(struct ip6_hbh))
2833 			return (EINVAL);
2834 		hbh = (struct ip6_hbh *)buf;
2835 		hbhlen = (hbh->ip6h_len + 1) << 3;
2836 		if (len != hbhlen)
2837 			return (EINVAL);
2838 
2839 		/* turn off the previous option, then set the new option. */
2840 		ip6_clearpktopts(opt, IPV6_HOPOPTS);
2841 		opt->ip6po_hbh = malloc(hbhlen, M_IP6OPT, M_NOWAIT);
2842 		if (opt->ip6po_hbh == NULL)
2843 			return (ENOBUFS);
2844 		bcopy(hbh, opt->ip6po_hbh, hbhlen);
2845 
2846 		break;
2847 	}
2848 
2849 	case IPV6_2292DSTOPTS:
2850 	case IPV6_DSTOPTS:
2851 	case IPV6_RTHDRDSTOPTS:
2852 	{
2853 		struct ip6_dest *dest, **newdest = NULL;
2854 		int destlen;
2855 
2856 		if (cred != NULL) { /* XXX: see the comment for IPV6_HOPOPTS */
2857 			error = priv_check_cred(cred,
2858 			    PRIV_NETINET_SETHDROPTS, 0);
2859 			if (error)
2860 				return (error);
2861 		}
2862 
2863 		if (len == 0) {
2864 			ip6_clearpktopts(opt, optname);
2865 			break;	/* just remove the option */
2866 		}
2867 
2868 		/* message length validation */
2869 		if (len < sizeof(struct ip6_dest))
2870 			return (EINVAL);
2871 		dest = (struct ip6_dest *)buf;
2872 		destlen = (dest->ip6d_len + 1) << 3;
2873 		if (len != destlen)
2874 			return (EINVAL);
2875 
2876 		/*
2877 		 * Determine the position that the destination options header
2878 		 * should be inserted; before or after the routing header.
2879 		 */
2880 		switch (optname) {
2881 		case IPV6_2292DSTOPTS:
2882 			/*
2883 			 * The old advacned API is ambiguous on this point.
2884 			 * Our approach is to determine the position based
2885 			 * according to the existence of a routing header.
2886 			 * Note, however, that this depends on the order of the
2887 			 * extension headers in the ancillary data; the 1st
2888 			 * part of the destination options header must appear
2889 			 * before the routing header in the ancillary data,
2890 			 * too.
2891 			 * RFC3542 solved the ambiguity by introducing
2892 			 * separate ancillary data or option types.
2893 			 */
2894 			if (opt->ip6po_rthdr == NULL)
2895 				newdest = &opt->ip6po_dest1;
2896 			else
2897 				newdest = &opt->ip6po_dest2;
2898 			break;
2899 		case IPV6_RTHDRDSTOPTS:
2900 			newdest = &opt->ip6po_dest1;
2901 			break;
2902 		case IPV6_DSTOPTS:
2903 			newdest = &opt->ip6po_dest2;
2904 			break;
2905 		}
2906 
2907 		/* turn off the previous option, then set the new option. */
2908 		ip6_clearpktopts(opt, optname);
2909 		*newdest = malloc(destlen, M_IP6OPT, M_NOWAIT);
2910 		if (*newdest == NULL)
2911 			return (ENOBUFS);
2912 		bcopy(dest, *newdest, destlen);
2913 
2914 		break;
2915 	}
2916 
2917 	case IPV6_2292RTHDR:
2918 	case IPV6_RTHDR:
2919 	{
2920 		struct ip6_rthdr *rth;
2921 		int rthlen;
2922 
2923 		if (len == 0) {
2924 			ip6_clearpktopts(opt, IPV6_RTHDR);
2925 			break;	/* just remove the option */
2926 		}
2927 
2928 		/* message length validation */
2929 		if (len < sizeof(struct ip6_rthdr))
2930 			return (EINVAL);
2931 		rth = (struct ip6_rthdr *)buf;
2932 		rthlen = (rth->ip6r_len + 1) << 3;
2933 		if (len != rthlen)
2934 			return (EINVAL);
2935 
2936 		switch (rth->ip6r_type) {
2937 		case IPV6_RTHDR_TYPE_0:
2938 			if (rth->ip6r_len == 0)	/* must contain one addr */
2939 				return (EINVAL);
2940 			if (rth->ip6r_len % 2) /* length must be even */
2941 				return (EINVAL);
2942 			if (rth->ip6r_len / 2 != rth->ip6r_segleft)
2943 				return (EINVAL);
2944 			break;
2945 		default:
2946 			return (EINVAL);	/* not supported */
2947 		}
2948 
2949 		/* turn off the previous option */
2950 		ip6_clearpktopts(opt, IPV6_RTHDR);
2951 		opt->ip6po_rthdr = malloc(rthlen, M_IP6OPT, M_NOWAIT);
2952 		if (opt->ip6po_rthdr == NULL)
2953 			return (ENOBUFS);
2954 		bcopy(rth, opt->ip6po_rthdr, rthlen);
2955 
2956 		break;
2957 	}
2958 
2959 	case IPV6_USE_MIN_MTU:
2960 		if (len != sizeof(int))
2961 			return (EINVAL);
2962 		minmtupolicy = *(int *)buf;
2963 		if (minmtupolicy != IP6PO_MINMTU_MCASTONLY &&
2964 		    minmtupolicy != IP6PO_MINMTU_DISABLE &&
2965 		    minmtupolicy != IP6PO_MINMTU_ALL) {
2966 			return (EINVAL);
2967 		}
2968 		opt->ip6po_minmtu = minmtupolicy;
2969 		break;
2970 
2971 	case IPV6_DONTFRAG:
2972 		if (len != sizeof(int))
2973 			return (EINVAL);
2974 
2975 		if (uproto == IPPROTO_TCP || *(int *)buf == 0) {
2976 			/*
2977 			 * we ignore this option for TCP sockets.
2978 			 * (RFC3542 leaves this case unspecified.)
2979 			 */
2980 			opt->ip6po_flags &= ~IP6PO_DONTFRAG;
2981 		} else
2982 			opt->ip6po_flags |= IP6PO_DONTFRAG;
2983 		break;
2984 
2985 	case IPV6_PREFER_TEMPADDR:
2986 		if (len != sizeof(int))
2987 			return (EINVAL);
2988 		preftemp = *(int *)buf;
2989 		if (preftemp != IP6PO_TEMPADDR_SYSTEM &&
2990 		    preftemp != IP6PO_TEMPADDR_NOTPREFER &&
2991 		    preftemp != IP6PO_TEMPADDR_PREFER) {
2992 			return (EINVAL);
2993 		}
2994 		opt->ip6po_prefer_tempaddr = preftemp;
2995 		break;
2996 
2997 	default:
2998 		return (ENOPROTOOPT);
2999 	} /* end of switch */
3000 
3001 	return (0);
3002 }
3003 
3004 /*
3005  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
3006  * packet to the input queue of a specified interface.  Note that this
3007  * calls the output routine of the loopback "driver", but with an interface
3008  * pointer that might NOT be &loif -- easier than replicating that code here.
3009  */
3010 void
3011 ip6_mloopback(struct ifnet *ifp, struct mbuf *m)
3012 {
3013 	struct mbuf *copym;
3014 	struct ip6_hdr *ip6;
3015 
3016 	copym = m_copym(m, 0, M_COPYALL, M_NOWAIT);
3017 	if (copym == NULL)
3018 		return;
3019 
3020 	/*
3021 	 * Make sure to deep-copy IPv6 header portion in case the data
3022 	 * is in an mbuf cluster, so that we can safely override the IPv6
3023 	 * header portion later.
3024 	 */
3025 	if (!M_WRITABLE(copym) ||
3026 	    copym->m_len < sizeof(struct ip6_hdr)) {
3027 		copym = m_pullup(copym, sizeof(struct ip6_hdr));
3028 		if (copym == NULL)
3029 			return;
3030 	}
3031 	ip6 = mtod(copym, struct ip6_hdr *);
3032 	/*
3033 	 * clear embedded scope identifiers if necessary.
3034 	 * in6_clearscope will touch the addresses only when necessary.
3035 	 */
3036 	in6_clearscope(&ip6->ip6_src);
3037 	in6_clearscope(&ip6->ip6_dst);
3038 	if (copym->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) {
3039 		copym->m_pkthdr.csum_flags |= CSUM_DATA_VALID_IPV6 |
3040 		    CSUM_PSEUDO_HDR;
3041 		copym->m_pkthdr.csum_data = 0xffff;
3042 	}
3043 	if_simloop(ifp, copym, AF_INET6, 0);
3044 }
3045 
3046 /*
3047  * Chop IPv6 header off from the payload.
3048  */
3049 static int
3050 ip6_splithdr(struct mbuf *m, struct ip6_exthdrs *exthdrs)
3051 {
3052 	struct mbuf *mh;
3053 	struct ip6_hdr *ip6;
3054 
3055 	ip6 = mtod(m, struct ip6_hdr *);
3056 	if (m->m_len > sizeof(*ip6)) {
3057 		mh = m_gethdr(M_NOWAIT, MT_DATA);
3058 		if (mh == NULL) {
3059 			m_freem(m);
3060 			return ENOBUFS;
3061 		}
3062 		m_move_pkthdr(mh, m);
3063 		M_ALIGN(mh, sizeof(*ip6));
3064 		m->m_len -= sizeof(*ip6);
3065 		m->m_data += sizeof(*ip6);
3066 		mh->m_next = m;
3067 		m = mh;
3068 		m->m_len = sizeof(*ip6);
3069 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
3070 	}
3071 	exthdrs->ip6e_ip6 = m;
3072 	return 0;
3073 }
3074 
3075 /*
3076  * Compute IPv6 extension header length.
3077  */
3078 int
3079 ip6_optlen(struct inpcb *in6p)
3080 {
3081 	int len;
3082 
3083 	if (!in6p->in6p_outputopts)
3084 		return 0;
3085 
3086 	len = 0;
3087 #define elen(x) \
3088     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
3089 
3090 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
3091 	if (in6p->in6p_outputopts->ip6po_rthdr)
3092 		/* dest1 is valid with rthdr only */
3093 		len += elen(in6p->in6p_outputopts->ip6po_dest1);
3094 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
3095 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
3096 	return len;
3097 #undef elen
3098 }
3099